paperID
stringlengths 36
36
| pwc_id
stringlengths 8
47
| arxiv_id
stringlengths 6
16
⌀ | nips_id
float64 | url_abs
stringlengths 18
329
| url_pdf
stringlengths 18
742
| title
stringlengths 8
325
| abstract
stringlengths 1
7.27k
⌀ | authors
stringlengths 2
7.06k
| published
stringlengths 10
10
⌀ | conference
stringlengths 12
47
⌀ | conference_url_abs
stringlengths 16
198
⌀ | conference_url_pdf
stringlengths 27
199
⌀ | proceeding
stringlengths 6
47
⌀ | taskID
stringlengths 7
1.44k
| areaID
stringclasses 688
values | embedding
stringlengths 9.26k
12.5k
| umap_embedding
stringlengths 29
44
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
47f95bb5-3d3b-4439-a043-992b2ef39eab
|
towards-high-fidelity-single-view-holistic
|
2207.08656
| null |
https://arxiv.org/abs/2207.08656v2
|
https://arxiv.org/pdf/2207.08656v2.pdf
|
Towards High-Fidelity Single-view Holistic Reconstruction of Indoor Scenes
|
We present a new framework to reconstruct holistic 3D indoor scenes including both room background and indoor objects from single-view images. Existing methods can only produce 3D shapes of indoor objects with limited geometry quality because of the heavy occlusion of indoor scenes. To solve this, we propose an instance-aligned implicit function (InstPIFu) for detailed object reconstruction. Combining with instance-aligned attention module, our method is empowered to decouple mixed local features toward the occluded instances. Additionally, unlike previous methods that simply represents the room background as a 3D bounding box, depth map or a set of planes, we recover the fine geometry of the background via implicit representation. Extensive experiments on the SUN RGB-D, Pix3D, 3D-FUTURE, and 3D-FRONT datasets demonstrate that our method outperforms existing approaches in both background and foreground object reconstruction. Our code and model will be made publicly available.
|
['Xiaoguang Han', 'Shuguang Cui', 'GuanYing Chen', 'Yujian Zheng', 'Haolin Liu']
|
2022-07-18
| null | null | null | null |
['object-reconstruction']
|
['computer-vision']
|
[ 2.47269586e-01 -3.13175544e-02 3.73541385e-01 -5.17273307e-01
-6.53047144e-01 -6.85357988e-01 4.35614139e-01 -3.42447519e-01
3.03867757e-01 6.57948196e-01 2.26050213e-01 -2.34027877e-01
1.75002292e-01 -9.61597502e-01 -1.00553524e+00 -6.17701948e-01
4.89668638e-01 4.75968063e-01 2.74407834e-01 5.01487255e-02
4.75145923e-03 7.18366563e-01 -1.72338831e+00 3.30918849e-01
9.35734808e-01 1.05650973e+00 3.43361348e-01 7.55511105e-01
-4.29457366e-01 7.12460876e-01 -3.49791199e-01 -7.74113312e-02
5.57044089e-01 -3.59859578e-02 -4.79613394e-01 5.19552410e-01
1.08697629e+00 -9.18089986e-01 -4.27509636e-01 8.51335287e-01
4.10125613e-01 1.36181638e-01 3.98881614e-01 -8.49604547e-01
-3.76236469e-01 -3.55529547e-01 -4.64434743e-01 -1.50580287e-01
8.56963098e-01 -5.15024103e-02 5.20861268e-01 -1.18908298e+00
5.34757614e-01 1.42491984e+00 6.52272940e-01 4.78531629e-01
-1.19250190e+00 -4.92739797e-01 9.10884678e-01 -2.81787217e-01
-1.39393401e+00 -4.35226440e-01 1.10095453e+00 -3.51995438e-01
7.98417568e-01 5.52795053e-01 8.37870061e-01 1.17806399e+00
-1.48435324e-01 7.16638565e-01 1.13854587e+00 -1.33985773e-01
1.68886676e-01 -4.97511625e-02 1.22471787e-01 8.07534099e-01
2.59987175e-01 -8.34027454e-02 -4.06030446e-01 -2.39042193e-02
1.23846912e+00 4.81735736e-01 -4.69924182e-01 -9.16316926e-01
-1.10955608e+00 2.06335083e-01 5.87411940e-01 -3.50835025e-01
-1.41234443e-01 2.24888310e-01 -3.36041391e-01 -4.01382625e-01
6.74199522e-01 -1.44954726e-01 -6.19231164e-01 1.62692219e-01
-5.40388465e-01 5.27376533e-01 6.21594548e-01 1.42162359e+00
9.54428971e-01 -1.47474855e-01 3.29856798e-02 3.75549227e-01
5.37638426e-01 8.30179036e-01 -3.49581182e-01 -1.34225881e+00
5.83746433e-01 7.50638545e-01 5.34775615e-01 -9.17325556e-01
-2.19571277e-01 -2.24075735e-01 -6.79467320e-01 2.76658893e-01
4.12061036e-01 3.52432132e-01 -1.07800579e+00 1.12485945e+00
1.01618016e+00 3.05330217e-01 -2.68727750e-01 9.59990978e-01
1.28362906e+00 6.95569158e-01 -5.30943692e-01 1.33379087e-01
1.03217185e+00 -1.15145302e+00 -6.36406124e-01 -4.99970138e-01
-1.63073555e-01 -5.21639287e-01 1.07253015e+00 2.98419446e-01
-1.16772664e+00 -6.81218326e-01 -8.98775518e-01 -6.30335212e-01
-1.71311066e-01 -1.00388430e-01 7.09475338e-01 7.07672477e-01
-7.53355086e-01 2.98707813e-01 -1.04492438e+00 -1.16744898e-01
6.73829138e-01 2.75148422e-01 -3.74556750e-01 -5.22061408e-01
-2.44469896e-01 4.27448690e-01 -1.39667526e-01 2.17770904e-01
-1.08689189e+00 -8.39705825e-01 -1.05570352e+00 -9.46600139e-02
4.75585938e-01 -1.06199217e+00 9.94360209e-01 -5.84641933e-01
-1.43153691e+00 9.15840209e-01 -4.75634694e-01 1.54449418e-01
6.47463202e-01 -6.51054978e-01 2.87409216e-01 -9.78010222e-02
-3.63490134e-02 1.32654339e-01 6.70195758e-01 -2.01409841e+00
-3.39924306e-01 -7.14758754e-01 4.97380257e-01 4.45331573e-01
3.41175795e-01 -4.60041225e-01 -6.38371587e-01 -2.27859750e-01
7.75145471e-01 -3.87554199e-01 -4.41358596e-01 4.06136096e-01
-6.36282146e-01 4.54663396e-01 7.53504694e-01 -8.00872326e-01
6.44510269e-01 -2.14109135e+00 3.24583083e-01 -1.92200299e-02
2.32659385e-01 -2.59393126e-01 2.18835682e-01 -1.99400857e-01
2.33740270e-01 -9.36504751e-02 -2.79686391e-01 -9.95945632e-01
1.05078943e-01 4.52403992e-01 -4.72610176e-01 6.09031796e-01
8.77650306e-02 5.98788321e-01 -9.75015879e-01 -3.73336226e-01
7.52705455e-01 1.00509977e+00 -8.15690994e-01 3.78942728e-01
-3.38467032e-01 8.16173613e-01 -6.93373382e-01 1.08789539e+00
1.33448398e+00 -2.98271596e-01 1.35417908e-01 -1.69544443e-01
-1.81876510e-01 4.87062633e-01 -1.39525700e+00 2.16932702e+00
-5.03435850e-01 2.15831120e-02 5.18440962e-01 -3.39675367e-01
7.32546926e-01 1.33959781e-02 3.85845929e-01 -4.44031507e-01
-9.82469097e-02 -5.92199415e-02 -8.84605467e-01 -1.89709127e-01
4.30014491e-01 1.68371424e-01 2.22377419e-01 -3.09383664e-02
-2.37030014e-01 -5.47262609e-01 -5.71288705e-01 2.76454166e-02
9.30522501e-01 9.45650339e-01 1.21872075e-01 -2.77162008e-02
5.41419804e-01 -1.49178833e-01 6.55806422e-01 7.34980762e-01
5.64737571e-03 1.14340281e+00 -2.05591861e-02 -8.01052094e-01
-8.75230849e-01 -1.61855328e+00 -3.29993814e-01 6.00071847e-01
4.90097374e-01 -4.37617511e-01 -6.97203934e-01 -6.55413985e-01
9.48670879e-02 5.65964937e-01 -5.52690983e-01 5.76643646e-01
-7.60363042e-01 -5.26370943e-01 -2.42918268e-01 5.44006824e-01
5.07089436e-01 -4.98559862e-01 -6.73324347e-01 -3.10564917e-02
-4.47996706e-01 -1.28953505e+00 -3.73055428e-01 2.12484851e-01
-1.01776898e+00 -1.18789220e+00 -3.50917995e-01 -4.36288267e-01
9.99285221e-01 7.12602854e-01 1.42833877e+00 1.20029241e-01
-4.35394198e-01 8.55774283e-01 -1.39051795e-01 -2.75711536e-01
1.79614335e-01 -4.41030413e-01 -7.28200823e-02 -9.07612294e-02
-5.03291525e-02 -7.73835123e-01 -7.11150289e-01 2.39767656e-01
-4.47605222e-01 6.22374058e-01 -8.49047303e-02 4.32782471e-01
1.15576041e+00 -8.10773447e-02 -3.39973003e-01 -6.42182052e-01
-5.65725267e-01 -2.89591074e-01 -9.78791833e-01 -4.23083454e-02
5.53039685e-02 -4.03296500e-01 2.66146839e-01 -1.60308719e-01
-1.38227999e+00 5.25581121e-01 -7.37325028e-02 -6.81495070e-01
-4.80053455e-01 -4.79791641e-01 -9.16808903e-01 -2.20865328e-02
2.10560232e-01 3.03035080e-01 -6.75400376e-01 -7.77909815e-01
2.52212286e-01 1.00230806e-01 5.31965256e-01 -9.82102752e-01
1.02379107e+00 1.08676279e+00 5.95433488e-02 -7.86578655e-01
-1.16623867e+00 -3.26456934e-01 -1.03758931e+00 -2.24855915e-01
1.08475590e+00 -1.28930998e+00 -5.90180337e-01 3.57841164e-01
-1.38953221e+00 -4.89405721e-01 -2.76665330e-01 1.04858458e-01
-6.07496202e-01 9.94172543e-02 -3.13109487e-01 -1.21953022e+00
1.77373931e-01 -1.08099830e+00 1.80174685e+00 1.10474743e-01
2.60524690e-01 -7.03861713e-01 -1.63666993e-01 6.95834517e-01
7.10419640e-02 6.89948976e-01 6.44273937e-01 3.21140200e-01
-1.68181396e+00 1.09263211e-01 -2.96175122e-01 1.17264882e-01
4.07322645e-01 -2.38420889e-02 -1.55702972e+00 -5.72125940e-03
1.84393361e-01 -1.93167664e-02 6.59388065e-01 5.76303005e-01
1.52864122e+00 -2.00032353e-01 -4.17879552e-01 1.36070848e+00
1.54184794e+00 1.11181904e-02 6.54401660e-01 2.54237980e-01
1.25814867e+00 5.42001903e-01 5.13820052e-01 4.84512776e-01
7.99274683e-01 7.63040781e-01 1.00107825e+00 -1.65466875e-01
-4.35309857e-01 -4.21967953e-01 1.35903269e-01 4.98313248e-01
-4.85052764e-01 -1.05016701e-01 -7.33348191e-01 1.17736675e-01
-1.63344038e+00 -8.94580066e-01 -3.11628312e-01 2.27706289e+00
4.53492671e-01 -4.57164422e-02 -2.03484073e-01 5.84562756e-02
2.21407637e-01 2.28591502e-01 -6.42007589e-01 1.75639406e-01
-1.58921435e-01 5.86531013e-02 3.10106456e-01 9.38212454e-01
-1.15903699e+00 7.38214850e-01 6.07143784e+00 1.83717027e-01
-4.42213714e-01 6.27531409e-02 7.13518202e-01 -3.82813901e-01
-5.52497983e-01 -1.16118453e-01 -1.08545387e+00 2.15338349e-01
2.47955367e-01 4.83144671e-01 6.61350191e-01 9.90911961e-01
5.52047603e-03 -1.00081384e-01 -1.11935210e+00 1.17779315e+00
4.61768284e-02 -1.38211644e+00 -1.42572924e-01 2.74754703e-01
8.95307124e-01 -1.52796865e-01 -1.21294394e-01 1.57399401e-01
3.56266052e-01 -9.38044488e-01 1.08104825e+00 7.11488783e-01
7.10063279e-01 -5.53874195e-01 2.83254385e-01 3.91565233e-01
-1.44585919e+00 8.60452056e-02 -3.48143667e-01 -2.34370068e-01
1.17752634e-01 6.13525510e-01 -4.85637575e-01 8.51255238e-01
1.19748080e+00 7.43667245e-01 -4.38241631e-01 7.39060283e-01
-1.25998765e-01 5.57232462e-02 -4.03391212e-01 5.87924063e-01
-2.08511710e-01 -3.99757773e-01 4.69324142e-01 8.55332077e-01
3.20796072e-01 3.75225633e-01 4.87630129e-01 1.29730070e+00
1.04303978e-01 -3.74412626e-01 -6.87927306e-01 4.89236802e-01
1.55208021e-01 1.13407826e+00 -7.59124935e-01 -2.00659081e-01
-5.27588904e-01 1.20442748e+00 3.13837975e-01 6.43004298e-01
-1.00328290e+00 1.42634481e-01 1.00207996e+00 5.41227758e-01
4.86154646e-01 -4.59219068e-01 -5.28319299e-01 -1.49759960e+00
2.39740998e-01 -4.77307111e-01 -2.81538069e-01 -1.14062726e+00
-1.14266181e+00 3.78225833e-01 3.87426428e-02 -9.50290680e-01
3.76356691e-01 -7.87133098e-01 -2.23046765e-01 8.70873749e-01
-1.56396961e+00 -1.39717937e+00 -9.94724751e-01 7.17717707e-01
7.02744365e-01 6.54358506e-01 9.41758931e-01 3.12492937e-01
-4.54987913e-01 -1.21129118e-01 -1.61804333e-01 9.74446163e-03
2.90806055e-01 -1.38512349e+00 4.37988102e-01 7.38516748e-01
-7.40602463e-02 5.81341565e-01 5.20146847e-01 -7.29504943e-01
-1.81776977e+00 -1.16052318e+00 3.14733922e-01 -1.13616133e+00
-9.07657668e-02 -8.99248540e-01 -6.58336580e-01 8.84790897e-01
-2.92715698e-01 5.17438173e-01 4.69082057e-01 -3.84109281e-02
-4.23271090e-01 -2.48780414e-01 -1.40584731e+00 4.89478737e-01
1.75444293e+00 -3.47351938e-01 -2.68583953e-01 4.30863082e-01
1.13221264e+00 -1.19524097e+00 -6.15317762e-01 4.74116147e-01
5.14096558e-01 -1.41921520e+00 1.77296877e+00 -1.66624293e-01
2.77902454e-01 -7.85796523e-01 -9.49476480e-01 -5.85319638e-01
-3.05104554e-01 -4.42703426e-01 -6.78575099e-01 1.13584018e+00
-3.31723332e-01 -3.41590583e-01 8.65416646e-01 7.69670427e-01
-5.09616613e-01 -6.11782253e-01 -9.86032546e-01 -3.57960880e-01
-3.43670726e-01 -5.50156474e-01 1.06433439e+00 6.09188259e-01
-1.05849111e+00 -2.38867085e-02 -2.63004005e-01 7.39593506e-01
1.08555675e+00 5.64094901e-01 1.34520042e+00 -1.28623056e+00
-2.52853394e-01 -6.97822645e-02 -8.90290812e-02 -1.46786451e+00
7.37407282e-02 -5.86996436e-01 8.69936645e-02 -1.88518488e+00
1.61026582e-01 -5.74538529e-01 3.56915481e-02 2.51306444e-01
-1.46960318e-01 1.73983097e-01 -7.09186420e-02 -2.40023419e-01
-6.29598558e-01 7.56785691e-01 1.57752836e+00 -1.20695727e-02
-2.69351572e-01 9.58201438e-02 -5.24780869e-01 9.84657109e-01
2.52416700e-01 -1.36169270e-01 -2.76039034e-01 -7.09580302e-01
4.80786934e-02 -1.09371983e-01 7.93966055e-01 -1.00162053e+00
-3.09125930e-01 -3.99613589e-01 1.26233637e+00 -1.29981184e+00
1.09492397e+00 -1.27425897e+00 2.24012956e-01 -2.39227153e-02
3.83615613e-01 -1.81828395e-01 2.90510595e-01 7.54620075e-01
3.83215129e-01 3.79378945e-01 5.21080256e-01 -5.70833683e-01
-2.46585518e-01 5.95004618e-01 1.77963585e-01 -2.24789158e-01
7.59394765e-01 -5.46383977e-01 -2.80300587e-01 -1.17683269e-01
-5.07720470e-01 5.12983948e-02 1.13166296e+00 1.69362649e-01
9.28236604e-01 -1.37732148e+00 -2.70967454e-01 6.24273360e-01
-1.97606355e-01 1.18823254e+00 5.51235080e-01 4.62104976e-01
-7.71464229e-01 2.57082552e-01 1.44842297e-01 -8.88053536e-01
-1.13438225e+00 6.05273187e-01 6.01865709e-01 -1.53223276e-02
-1.02523685e+00 9.72880304e-01 9.48579848e-01 -8.91171813e-01
5.19993186e-01 -8.47515285e-01 3.89694721e-01 -5.16022086e-01
6.83871090e-01 3.69327098e-01 -9.01574045e-02 -6.48612916e-01
-5.20545900e-01 8.83950949e-01 4.20907795e-01 2.61169523e-01
1.49186480e+00 -4.00505453e-01 -2.33015895e-01 6.51250660e-01
5.93195438e-01 5.29588521e-01 -1.96749651e+00 -2.31686095e-03
-7.15527058e-01 -1.07823920e+00 -5.86530454e-02 -5.99559247e-01
-8.36807609e-01 8.59537184e-01 5.61945736e-01 -2.96035200e-01
1.16214657e+00 -3.32816616e-02 3.82923782e-01 2.67740488e-01
8.05427372e-01 -5.71773469e-01 4.92873304e-02 5.97201824e-01
9.10258770e-01 -1.18991649e+00 3.62392217e-01 -8.16988051e-01
1.78015772e-02 1.00522184e+00 9.48652208e-01 -1.00713864e-01
6.81639493e-01 4.00382906e-01 -1.80271357e-01 -1.09623477e-01
-2.23213091e-01 2.93054269e-03 4.24253643e-01 8.95284891e-01
1.98354051e-01 -2.46621948e-02 8.27729523e-01 5.93907177e-01
-1.76875263e-01 -3.99390459e-01 2.82610863e-01 1.08672178e+00
-2.92315453e-01 -8.74643683e-01 -9.59521115e-01 -4.84194346e-02
-6.34171888e-02 -2.91951615e-02 -1.45985633e-01 6.72276735e-01
3.94851267e-01 7.61471212e-01 1.71359167e-01 -1.28447503e-01
4.18563664e-01 -9.16399434e-02 1.06532371e+00 -8.34405303e-01
-1.69709265e-01 2.36177325e-01 -1.63241431e-01 -1.08309662e+00
-4.41120982e-01 -6.54711783e-01 -1.15788579e+00 -1.78218305e-01
-2.09049076e-01 -5.40770829e-01 7.45690882e-01 6.06092095e-01
2.51689523e-01 8.85508060e-01 3.49162608e-01 -1.75153995e+00
2.05736578e-01 -5.47141016e-01 -4.90907133e-01 1.84945419e-01
8.70714188e-01 -8.90909195e-01 -3.63872290e-01 1.83389317e-02]
|
[8.810627937316895, -3.018479585647583]
|
1897c957-9f6c-4aad-9fab-f08ec3ca8296
|
temporal-perceiver-a-general-architecture-for
|
2203.00307
| null |
https://arxiv.org/abs/2203.00307v2
|
https://arxiv.org/pdf/2203.00307v2.pdf
|
Temporal Perceiver: A General Architecture for Arbitrary Boundary Detection
|
Generic Boundary Detection (GBD) aims at locating the general boundaries that divide videos into semantically coherent and taxonomy-free units, and could serve as an important pre-processing step for long-form video understanding. Previous works often separately handle these different types of generic boundaries with specific designs of deep networks from simple CNN to LSTM. Instead, in this paper, we present Temporal Perceiver, a general architecture with Transformer, offering a unified solution to the detection of arbitrary generic boundaries, ranging from shot-level, event-level, to scene-level GBDs. The core design is to introduce a small set of latent feature queries as anchors to compress the redundant video input into a fixed dimension via cross-attention blocks. Thanks to this fixed number of latent units, it greatly reduces the quadratic complexity of attention operation to a linear form of input frames. Specifically, to explicitly leverage the temporal structure of videos, we construct two types of latent feature queries: boundary queries and context queries, which handle the semantic incoherence and coherence accordingly. Moreover, to guide the learning of latent feature queries, we propose an alignment loss on the cross-attention maps to explicitly encourage the boundary queries to attend on the top boundary candidates. Finally, we present a sparse detection head on the compressed representation, and directly output the final boundary detection results without any post-processing module. We test our Temporal Perceiver on a variety of GBD benchmarks. Our method obtains the state-of-the-art results on all benchmarks with RGB single-stream features: SoccerNet-v2 (81.9% avg-mAP), Kinetics-GEBD (86.0% avg-f1), TAPOS (73.2% avg-f1), MovieScenes (51.9% AP and 53.1% Miou) and MovieNet (53.3% AP and 53.2% Miou), demonstrating the generalization ability of our Temporal Perceiver.
|
['LiMin Wang', 'Gangshan Wu', 'Yuhong Wang', 'Jing Tan']
|
2022-03-01
| null | null | null | null |
['boundary-detection']
|
['computer-vision']
|
[ 7.63687771e-03 -2.62917936e-01 -2.13559479e-01 -2.55264193e-01
-7.86551058e-01 -5.92206240e-01 5.50183058e-01 -1.13762267e-01
-3.32747966e-01 1.41965359e-01 2.47738346e-01 -3.28214578e-02
4.43119258e-02 -6.13248348e-01 -9.68186855e-01 -4.72101480e-01
-4.40525055e-01 -5.10734655e-02 3.94438237e-01 -1.56321704e-01
1.00741588e-01 1.62767798e-01 -1.66570497e+00 5.57892680e-01
6.57173991e-01 1.59323514e+00 3.49989235e-01 5.59117973e-01
2.09253877e-01 7.41193593e-01 -5.34885108e-01 -3.65924500e-02
4.17134643e-01 -2.82152414e-01 -6.14233792e-01 9.08828005e-02
7.92464316e-01 -6.76516414e-01 -7.37574756e-01 1.00229406e+00
4.05083686e-01 4.07709032e-01 3.34232360e-01 -1.19096303e+00
-7.07966149e-01 3.99134815e-01 -5.71962833e-01 4.73337263e-01
6.03695631e-01 5.26441336e-01 1.17403448e+00 -1.15940905e+00
6.71887815e-01 1.35308039e+00 4.20144528e-01 3.14047903e-01
-1.01192963e+00 -5.45118153e-01 6.05604112e-01 4.14739311e-01
-1.52256465e+00 -4.72621500e-01 5.14569640e-01 -5.43134153e-01
1.16356754e+00 2.01274648e-01 7.38509834e-01 1.40196323e+00
-6.40175864e-03 8.95949483e-01 6.04468167e-01 1.09180339e-01
1.89984411e-01 -5.08131504e-01 8.28525424e-02 6.53419316e-01
-8.75115097e-02 1.13651648e-01 -9.19471920e-01 4.73028868e-01
1.04004383e+00 1.82909027e-01 -5.74885428e-01 -1.14418142e-01
-1.44785488e+00 5.10517240e-01 5.19536912e-01 6.05325699e-02
-3.71094167e-01 3.13308626e-01 6.00236952e-01 2.86937445e-01
3.75832736e-01 2.39860520e-01 -4.23352331e-01 -3.57828587e-01
-9.01893377e-01 8.09381083e-02 5.69465160e-01 1.24624431e+00
6.52689219e-01 5.69370128e-02 -4.99499500e-01 6.00434899e-01
5.09104691e-02 3.71234298e-01 3.76207441e-01 -1.13897717e+00
6.52438641e-01 5.02503335e-01 1.05349563e-01 -1.22972536e+00
-2.14238688e-01 -7.17890143e-01 -8.19696665e-01 -1.42340481e-01
2.01860994e-01 3.07489008e-01 -1.08699381e+00 1.90268278e+00
1.13245733e-01 3.50098580e-01 -2.19802454e-01 1.35221970e+00
8.88220012e-01 8.05155516e-01 -4.12664451e-02 -4.61749174e-02
1.37489748e+00 -1.36011612e+00 -5.28793097e-01 -2.29062006e-01
3.87018651e-01 -5.58545887e-01 1.45670593e+00 4.76915807e-01
-1.12300837e+00 -1.08052897e+00 -1.12010705e+00 -4.16759759e-01
-2.55845755e-01 2.65539289e-01 5.37871420e-01 -1.02539703e-01
-1.14327168e+00 5.92435241e-01 -8.91206980e-01 -2.80307263e-01
2.69095719e-01 2.75137663e-01 -3.83647412e-01 -2.39100724e-01
-1.17766583e+00 2.81847686e-01 3.77024293e-01 2.66302496e-01
-1.55799961e+00 -7.31645107e-01 -1.06900394e+00 2.20976770e-01
8.14791799e-01 -7.03422606e-01 1.01042354e+00 -8.14799666e-01
-1.25895917e+00 8.73975039e-01 -1.63370863e-01 -4.09524649e-01
5.35205245e-01 -7.09006071e-01 -3.23949099e-01 4.97345746e-01
2.14233816e-01 1.04007089e+00 8.04828048e-01 -8.85998785e-01
-6.46929502e-01 6.25370517e-02 5.01200855e-01 3.08204919e-01
-2.27860034e-01 -1.10254295e-01 -1.03911543e+00 -8.53114605e-01
3.18473518e-01 -5.44959784e-01 1.55308619e-01 1.03147119e-01
-3.29052746e-01 -3.16128224e-01 7.24249065e-01 -6.77023768e-01
1.35678792e+00 -2.43541288e+00 4.88559872e-01 -3.16317916e-01
4.63067144e-01 -1.56268254e-02 -2.74089575e-01 1.87978551e-01
-3.55845094e-02 1.72722280e-01 -6.33976087e-02 -5.97353995e-01
2.06033900e-01 1.41227677e-01 -4.84385997e-01 3.80149245e-01
2.53067374e-01 9.11164165e-01 -1.03540599e+00 -1.98360786e-01
1.98234707e-01 3.54237139e-01 -8.98278356e-01 2.96559930e-01
-2.83209443e-01 2.12346733e-01 -2.09491685e-01 8.44788194e-01
4.43830878e-01 -3.23340237e-01 -2.79171556e-01 -6.18564606e-01
-1.69448897e-01 4.76350188e-01 -1.01892674e+00 2.37296200e+00
-2.37494737e-01 5.98753810e-01 1.59215212e-01 -8.40988457e-01
7.73918390e-01 1.24246322e-01 4.74547982e-01 -8.65243614e-01
1.23169366e-02 1.48218289e-01 -1.96741328e-01 -7.06464529e-01
6.00959182e-01 3.36274922e-01 -2.60171384e-01 1.34753153e-01
3.99400860e-01 3.63978893e-01 2.85005659e-01 2.16976434e-01
1.18084347e+00 3.52398902e-01 -1.16114594e-01 -1.19288035e-01
3.29110801e-01 -3.46621871e-01 8.27518642e-01 7.26077974e-01
-3.39924395e-01 9.41205740e-01 6.09877110e-01 -6.05481505e-01
-8.04910898e-01 -1.22039187e+00 2.93516338e-01 1.21707904e+00
6.58852816e-01 -7.06487119e-01 -5.41585982e-01 -6.33108675e-01
-9.03343037e-02 3.46173316e-01 -6.99692547e-01 -4.29124743e-01
-6.62429273e-01 -1.70664832e-01 3.75875294e-01 6.37581646e-01
7.55195916e-01 -1.05759108e+00 -7.86170483e-01 1.05465010e-01
-4.56169367e-01 -1.35812330e+00 -8.94457579e-01 1.23984605e-01
-5.90781510e-01 -9.78758991e-01 -5.21030188e-01 -6.00810289e-01
3.50388706e-01 5.21104276e-01 1.20023549e+00 -4.74980325e-02
-1.57978490e-01 3.37660134e-01 -5.54076612e-01 4.00999218e-01
2.11403385e-01 4.39169630e-03 1.14206642e-01 1.77157864e-01
2.85302043e-01 -5.77250004e-01 -9.40658092e-01 5.64563930e-01
-8.86664510e-01 2.93692768e-01 3.75722289e-01 8.53667617e-01
7.50455022e-01 -2.80883223e-01 2.50737578e-01 -1.37122542e-01
2.21118808e-01 -5.52042305e-01 -4.66550827e-01 1.33576393e-01
-5.55941602e-03 -2.69411922e-01 5.64426661e-01 -5.85953772e-01
-6.54406071e-01 -2.68177748e-01 -2.30423827e-02 -1.13043737e+00
-1.79611087e-01 5.09889364e-01 -4.27145928e-01 3.04978549e-01
4.25230533e-01 3.25646222e-01 -4.02074844e-01 -5.86690426e-01
3.45262796e-01 2.38983244e-01 8.46278369e-01 -6.12210155e-01
4.50743526e-01 4.84091669e-01 -4.24965739e-01 -5.87386608e-01
-1.03673792e+00 -5.44311523e-01 -4.56769526e-01 -3.14777136e-01
1.02019036e+00 -1.16989124e+00 -8.26464057e-01 5.13696492e-01
-1.09233510e+00 -6.27155602e-01 -3.82913142e-01 5.23755074e-01
-6.85692549e-01 3.21721941e-01 -9.34719861e-01 -3.44423860e-01
-1.66411996e-01 -1.32749820e+00 1.37863088e+00 2.28272915e-01
-2.30125040e-01 -5.71774125e-01 -3.21484149e-01 2.43468940e-01
3.11655283e-01 3.14364135e-01 6.13955975e-01 -4.02526885e-01
-1.01423812e+00 2.59101897e-01 -4.05850112e-01 2.89166510e-01
6.26980513e-02 -1.86978802e-01 -1.02951145e+00 -5.35320401e-01
2.31078342e-02 -5.28096616e-01 1.25827253e+00 3.64887953e-01
1.22394001e+00 -2.28586540e-01 -2.33687103e-01 1.20566058e+00
1.18916690e+00 3.06275159e-01 4.94704664e-01 1.89061388e-01
7.34679818e-01 3.15194935e-01 7.29006946e-01 5.11026204e-01
3.80658805e-01 7.04803407e-01 6.95616484e-01 -3.65545712e-02
-3.45508188e-01 -4.60161895e-01 7.40964592e-01 9.55006003e-01
2.28691250e-02 -3.60982209e-01 -6.84646308e-01 6.13692522e-01
-1.85907102e+00 -9.02881444e-01 3.83372664e-01 2.07213449e+00
5.99133253e-01 4.55520511e-01 1.22470468e-01 -1.36965305e-01
7.55466640e-01 4.37588274e-01 -7.62326181e-01 5.76069802e-02
-2.10629418e-01 -1.12348363e-01 1.13694608e-01 4.22871768e-01
-1.41118419e+00 1.13316309e+00 5.32558107e+00 9.53694940e-01
-1.20700788e+00 1.53727382e-01 6.93137527e-01 -4.80127722e-01
-1.57374695e-01 -7.21068308e-02 -7.57216036e-01 6.03451014e-01
5.41426361e-01 1.37478650e-01 5.57054877e-01 6.36070073e-01
3.04261237e-01 -1.47659183e-01 -1.51373816e+00 1.36158514e+00
5.98041154e-02 -1.27513504e+00 1.25748083e-01 -8.83557051e-02
5.74411988e-01 1.29415110e-01 1.16720051e-01 5.69832206e-01
-1.58884659e-01 -9.92160797e-01 1.26135695e+00 5.84435403e-01
1.13511014e+00 -4.51736420e-01 4.50012326e-01 1.15919665e-01
-1.62121046e+00 -1.47897765e-01 -4.90676641e-01 -2.28007317e-01
2.80304193e-01 4.04721022e-01 -7.14649260e-02 5.67077398e-01
1.04434764e+00 1.11455572e+00 -3.60515356e-01 1.03649914e+00
-1.23262204e-01 4.32338536e-01 -5.67890108e-01 3.83007616e-01
4.34700787e-01 -1.12565421e-02 6.01963103e-01 1.17277300e+00
4.47025388e-01 2.20424347e-02 4.90296394e-01 1.08503103e+00
-1.79355815e-01 -2.78283268e-01 -3.46406490e-01 4.79159057e-02
5.29236019e-01 1.03316164e+00 -7.58285522e-01 -3.91200632e-01
-3.86381060e-01 1.22431028e+00 2.69984961e-01 6.54876232e-01
-1.19796193e+00 -3.90901566e-01 9.36770022e-01 3.75815965e-02
5.99489450e-01 -4.63902175e-01 1.42786250e-01 -1.48077273e+00
3.48504037e-01 -9.39392567e-01 5.31733632e-01 -8.03267062e-01
-1.07202518e+00 5.89241207e-01 -4.44142930e-02 -1.33439624e+00
1.45190060e-01 -4.49656844e-01 -5.44844747e-01 5.97929478e-01
-1.35176480e+00 -9.67794359e-01 -7.58455575e-01 5.95638216e-01
9.35380995e-01 1.02917612e-01 4.30062473e-01 5.37514925e-01
-7.44876087e-01 7.34926045e-01 -3.15825045e-01 2.35657305e-01
7.96672821e-01 -1.01979017e+00 5.71717918e-01 1.01675534e+00
4.10031639e-02 7.09520638e-01 4.07968670e-01 -5.38448155e-01
-1.40789723e+00 -1.07196617e+00 4.36286837e-01 -2.05421656e-01
7.12727070e-01 -6.64892733e-01 -7.93004453e-01 7.09505439e-01
1.05711669e-01 3.73008221e-01 2.68672168e-01 -4.26586531e-02
-6.14322066e-01 -2.71567106e-01 -6.46978438e-01 6.62003398e-01
1.61764526e+00 -7.87990749e-01 -5.46750247e-01 6.51377290e-02
1.32288742e+00 -6.30702794e-01 -8.91848922e-01 4.66612846e-01
4.69900995e-01 -1.28353059e+00 1.03390658e+00 -4.88872826e-01
5.82724929e-01 -4.95712817e-01 -3.65736604e-01 -8.59792411e-01
-3.65183413e-01 -7.91299284e-01 -4.63701785e-01 1.20022464e+00
-1.83279924e-02 -2.46865064e-01 6.28531575e-01 3.06388259e-01
-6.60251558e-01 -1.02140951e+00 -9.80455637e-01 -7.53057122e-01
-4.02560264e-01 -6.72296762e-01 3.44413221e-01 8.86293054e-01
-1.82805270e-01 2.92737246e-01 -4.43093657e-01 1.09879687e-01
4.01890039e-01 2.74782896e-01 6.03168786e-01 -6.50288522e-01
-5.21587968e-01 -5.38989544e-01 -4.55431759e-01 -1.99955904e+00
-1.93687499e-01 -7.75901496e-01 1.06873354e-02 -1.33359599e+00
7.64767826e-02 -3.41343999e-01 -6.05011582e-01 5.65791488e-01
-8.24727118e-03 1.95508271e-01 3.24548036e-01 2.62648016e-01
-1.15743673e+00 8.22164178e-01 1.35074496e+00 -2.49604613e-01
-2.23961055e-01 -4.67648089e-01 -4.28128213e-01 7.21700132e-01
4.49651957e-01 -2.57379889e-01 -5.85829854e-01 -7.47557819e-01
2.78699458e-01 1.12251855e-01 7.05196917e-01 -1.19105923e+00
3.77051950e-01 -2.01307032e-02 2.95330286e-01 -7.68504381e-01
5.49990773e-01 -4.93861377e-01 -3.80812511e-02 3.18353534e-01
-3.08573455e-01 1.53614506e-01 2.10437387e-01 5.60896993e-01
-5.95017612e-01 1.83089137e-01 3.90279740e-01 -1.18976086e-01
-1.12405598e+00 6.19350433e-01 -1.57392919e-01 3.85551453e-01
9.80003655e-01 -3.60825121e-01 -3.94323051e-01 -4.21059251e-01
-8.53493214e-01 4.63167280e-01 4.81957734e-01 7.11914122e-01
8.23198259e-01 -1.35117722e+00 -3.77205342e-01 2.92263836e-01
6.23719394e-02 3.39882702e-01 6.35352314e-01 1.01186419e+00
-4.83754724e-01 3.59732211e-01 -1.98218912e-01 -9.56535578e-01
-6.67195201e-01 7.76911616e-01 3.76204312e-01 -1.84300337e-02
-8.13765407e-01 1.25191677e+00 7.91544020e-01 1.60367474e-01
6.32443607e-01 -7.80111790e-01 9.73887071e-02 2.59531289e-01
4.72841710e-01 1.94944322e-01 -1.98533282e-01 -5.46235740e-01
-3.35182101e-01 6.73099816e-01 -6.92251185e-03 2.06530571e-01
1.20895827e+00 -2.03009188e-01 1.49632156e-01 5.04832327e-01
1.41807628e+00 -2.59522319e-01 -1.97288394e+00 -1.40350848e-01
-2.00033799e-01 -4.11782682e-01 -1.66108668e-01 -5.41799545e-01
-1.21718562e+00 1.13650274e+00 3.79687041e-01 8.55049044e-02
1.34112203e+00 8.07785690e-02 8.85426402e-01 2.34901085e-01
4.77690518e-01 -8.98840189e-01 4.41448241e-01 5.99951804e-01
1.09866428e+00 -1.11366260e+00 -2.96829343e-01 -3.35347295e-01
-3.39860827e-01 1.01058042e+00 8.73026609e-01 -1.85986444e-01
4.00427014e-01 1.35283485e-01 -2.34645620e-01 -2.72538722e-01
-1.05970943e+00 -2.53854066e-01 3.72974366e-01 2.02275544e-01
1.15290843e-01 -1.42367199e-01 3.69799286e-02 8.54333639e-01
2.07314473e-02 -1.63917288e-01 2.64459521e-01 5.82525313e-01
-3.63325477e-01 -5.34664452e-01 -1.91473085e-02 1.89223498e-01
-2.86532044e-01 -1.98899761e-01 -8.65681618e-02 9.24005449e-01
3.43328387e-01 7.96622932e-01 2.25378916e-01 -7.99887836e-01
3.58329475e-01 -1.31875798e-01 2.79504180e-01 -5.53798079e-01
-4.90693957e-01 3.69387507e-01 -4.94390167e-03 -1.20915234e+00
-2.94627517e-01 -3.93753856e-01 -1.14614010e+00 -1.65573478e-01
-1.83138430e-01 -1.37048766e-01 5.29465675e-02 7.65635848e-01
5.96826315e-01 7.66405642e-01 3.01653743e-01 -1.03208482e+00
-3.31880957e-01 -8.92285764e-01 -2.92378664e-01 6.47073030e-01
3.90561372e-01 -9.65366721e-01 -4.18747723e-01 1.48363501e-01]
|
[9.248205184936523, 0.5174456834793091]
|
2dedddb6-3d02-42dd-962d-82fcb75e4437
|
deepsetnet-predicting-sets-with-deep-neural
|
1611.08998
| null |
http://arxiv.org/abs/1611.08998v5
|
http://arxiv.org/pdf/1611.08998v5.pdf
|
DeepSetNet: Predicting Sets with Deep Neural Networks
|
This paper addresses the task of set prediction using deep learning. This is
important because the output of many computer vision tasks, including image
tagging and object detection, are naturally expressed as sets of entities
rather than vectors. As opposed to a vector, the size of a set is not fixed in
advance, and it is invariant to the ordering of entities within it. We define a
likelihood for a set distribution and learn its parameters using a deep neural
network. We also derive a loss for predicting a discrete distribution
corresponding to set cardinality. Set prediction is demonstrated on the problem
of multi-class image classification. Moreover, we show that the proposed
cardinality loss can also trivially be applied to the tasks of object counting
and pedestrian detection. Our approach outperforms existing methods in all
three cases on standard datasets.
|
['Vijay Kumar B G', 'S. Hamid Rezatofighi', 'Anton Milan', 'Anthony Dick', 'Ehsan Abbasnejad', 'Ian Reid']
|
2016-11-28
|
deepsetnet-predicting-sets-with-deep-neural-1
|
http://openaccess.thecvf.com/content_iccv_2017/html/Rezatofighi_DeepSetNet_Predicting_Sets_ICCV_2017_paper.html
|
http://openaccess.thecvf.com/content_ICCV_2017/papers/Rezatofighi_DeepSetNet_Predicting_Sets_ICCV_2017_paper.pdf
|
iccv-2017-10
|
['object-counting']
|
['computer-vision']
|
[ 3.77976477e-01 3.24754417e-02 -1.80276394e-01 -6.97107196e-01
-3.40750843e-01 -7.15815604e-01 6.21622264e-01 5.25574863e-01
-7.65614331e-01 6.40514135e-01 -1.32778287e-01 -1.55158758e-01
-1.32551551e-01 -1.00248885e+00 -1.00034022e+00 -4.79710430e-01
-1.56922907e-01 7.70694971e-01 2.60674536e-01 1.44445568e-01
-7.60729378e-03 4.92907614e-01 -1.70375407e+00 2.66984463e-01
1.74738929e-01 1.14192367e+00 1.26729742e-01 6.08173072e-01
-3.19727138e-02 6.42708957e-01 -6.31309748e-01 -6.42141342e-01
3.36223692e-01 9.11317319e-02 -1.03253090e+00 3.44146788e-01
8.94150257e-01 -4.77756143e-01 -3.42809081e-01 1.18119550e+00
1.95906520e-01 2.18765572e-01 9.97607589e-01 -1.50959992e+00
-6.95229471e-01 6.40743792e-01 -3.84771675e-01 1.11012988e-01
-1.40633345e-01 -3.54581326e-01 1.39558053e+00 -8.02278757e-01
3.58539522e-01 1.28704906e+00 6.26860797e-01 3.70690078e-01
-1.19040251e+00 -5.14348924e-01 1.75667569e-01 -2.87957327e-03
-1.24176955e+00 -3.51460934e-01 3.93112421e-01 -7.99074054e-01
5.21501899e-01 2.85985887e-01 5.33353686e-01 6.03870511e-01
-1.94109708e-01 8.32125127e-01 7.28924870e-01 -5.47908366e-01
1.91247657e-01 3.47214729e-01 5.88389516e-01 9.40049112e-01
5.73113203e-01 -2.68631369e-01 -2.30065361e-01 -2.60194838e-01
6.41802251e-01 3.13677967e-01 7.08347633e-02 -5.75192690e-01
-1.08902323e+00 1.07304394e+00 3.65670234e-01 1.93353668e-02
6.81197047e-02 6.71188772e-01 4.96686906e-01 8.89144614e-02
5.07798672e-01 3.02160859e-01 -4.68885809e-01 4.00656134e-01
-6.13819301e-01 3.02792102e-01 8.80836427e-01 1.16052127e+00
7.42268145e-01 -4.22142506e-01 -2.32648373e-01 7.91961849e-01
3.67272437e-01 4.69444901e-01 1.95726170e-03 -9.20708716e-01
3.53407651e-01 3.53166670e-01 3.74608934e-01 -8.93763363e-01
-4.49324429e-01 -3.42998296e-01 -7.08771825e-01 1.24976166e-01
5.04757762e-01 -1.11111756e-02 -7.18185186e-01 2.11700654e+00
2.76968688e-01 2.50027299e-01 -4.18436319e-01 6.17745936e-01
7.44498074e-01 3.34902227e-01 1.05593629e-01 -3.42183039e-02
1.44907618e+00 -5.35940170e-01 -4.84957695e-01 -7.71896094e-02
6.09368861e-01 -2.65996784e-01 5.67932546e-01 4.81434837e-02
-8.02443266e-01 -3.62355858e-01 -8.07933807e-01 -2.23593980e-01
-6.56214237e-01 2.14793012e-01 9.66682851e-01 6.77378058e-01
-1.07602632e+00 5.02066731e-01 -5.63608944e-01 -3.76993418e-01
8.21459711e-01 7.54474878e-01 -3.25242490e-01 2.02388838e-01
-9.96038139e-01 8.19482386e-01 6.63429320e-01 -9.12397355e-02
-4.75537300e-01 -4.88336116e-01 -1.07999170e+00 3.51265699e-01
2.48768732e-01 -8.80056322e-01 1.21659493e+00 -9.29129064e-01
-9.58435833e-01 1.32649052e+00 -1.00831412e-01 -9.30101514e-01
2.77744114e-01 -2.31215894e-01 -1.07504800e-01 -4.06689420e-02
3.05934310e-01 7.40168810e-01 7.27971017e-01 -1.19528115e+00
-9.40258801e-01 -2.45889261e-01 6.44139051e-01 6.29447699e-02
-6.43810630e-01 6.44195676e-02 -3.66882920e-01 -3.74148607e-01
-2.39235163e-01 -9.44950342e-01 -3.66628408e-01 3.65512937e-01
-5.19549787e-01 -5.35700858e-01 6.05596662e-01 -1.18678354e-01
1.09914958e+00 -2.02038956e+00 -1.38112620e-01 2.29098469e-01
6.05356097e-01 7.05567300e-02 3.43722757e-03 1.72158673e-01
1.35021359e-02 2.88196772e-01 -2.70277262e-01 -5.35172403e-01
3.52260709e-01 2.39859760e-01 -4.43676651e-01 7.23919332e-01
3.62941116e-01 7.07583189e-01 -1.02204299e+00 -6.31121576e-01
2.68776417e-01 2.27038294e-01 -6.25590563e-01 -2.42360860e-01
-3.09472740e-01 -5.68547025e-02 -3.88175189e-01 2.11516947e-01
6.63316369e-01 -7.69104123e-01 1.95295200e-01 -1.43338770e-01
1.23528033e-01 9.34820399e-02 -1.20822263e+00 1.11279368e+00
-3.75511765e-01 7.57983088e-01 -4.86305326e-01 -1.16747606e+00
4.85479474e-01 1.65359620e-02 8.23032618e-01 -1.87614217e-01
8.18582624e-02 -2.10308298e-01 -5.67508489e-02 -2.74931677e-02
7.43825853e-01 1.55613068e-02 -4.55104411e-01 4.51686442e-01
2.44975805e-01 1.85280606e-01 3.68685335e-01 2.62344122e-01
1.00030911e+00 -2.79088646e-01 4.38506305e-01 -2.05789015e-01
3.35953742e-01 -8.39723870e-02 4.90830064e-01 1.15344107e+00
-4.84188506e-03 4.12090719e-01 4.69413310e-01 -4.96446073e-01
-1.19752026e+00 -1.07782960e+00 -5.02886951e-01 1.17802835e+00
3.01907390e-01 -3.28861713e-01 -5.91635048e-01 -7.45745540e-01
4.01973158e-01 3.81316990e-01 -6.03882313e-01 1.41587639e-02
-5.01461625e-01 -7.16184974e-01 2.77912825e-01 6.00661635e-01
2.36605898e-01 -8.40321243e-01 -5.10660946e-01 1.40987024e-01
4.50448990e-02 -1.38672793e+00 -7.71746218e-01 1.46951184e-01
-4.61129934e-01 -1.34317636e+00 -2.66767830e-01 -1.06764019e+00
5.78463972e-01 1.55573025e-01 1.48123324e+00 2.34583467e-01
-2.48373926e-01 6.24163806e-01 -3.56003232e-02 -8.55910838e-01
-2.49709323e-01 -2.69599818e-02 1.85314849e-01 2.00469568e-01
5.40235579e-01 -2.36450583e-01 -5.36329210e-01 2.19750497e-02
-9.16264057e-01 -7.66897202e-02 3.13318372e-01 9.91907179e-01
7.20336676e-01 2.20137671e-01 5.14659047e-01 -1.16303205e+00
2.91736484e-01 -4.94080067e-01 -9.66942549e-01 3.14351857e-01
-1.39308050e-01 -1.15862321e-02 5.07378042e-01 -4.55127060e-01
-6.01294756e-01 4.32155639e-01 1.90583676e-01 -2.53909647e-01
-1.53148949e-01 1.96453631e-01 -2.31411591e-01 1.61176383e-01
1.32319316e-01 -1.47364140e-01 -2.59220123e-01 -1.97954401e-01
4.48436677e-01 5.78583837e-01 4.04391795e-01 -3.34996134e-01
6.55004799e-01 7.35875309e-01 3.43112916e-01 -8.11495602e-01
-1.40127647e+00 -7.56269097e-01 -1.09027791e+00 -7.21155778e-02
9.09493387e-01 -9.47608709e-01 -1.21906316e+00 2.82258868e-01
-1.36324668e+00 -1.05523370e-01 -4.64206725e-01 3.29002410e-01
-5.71507752e-01 2.91509897e-01 -5.65439224e-01 -1.01700282e+00
-8.91822726e-02 -7.78726876e-01 1.39461446e+00 -5.97800175e-03
2.00928431e-02 -1.20605063e+00 -2.36096084e-01 -1.80806980e-01
-1.42946675e-01 2.50724256e-01 1.03495133e+00 -9.53754961e-01
-7.03383029e-01 -2.02974096e-01 -2.79525280e-01 3.76920462e-01
-5.41344434e-02 -1.84918612e-01 -8.73501420e-01 -1.95917398e-01
-2.54256994e-01 -3.97439063e-01 1.22571886e+00 5.24004996e-01
1.64727902e+00 -5.56175649e-01 -6.52131498e-01 4.15929019e-01
1.55655587e+00 -2.15945393e-01 1.65657803e-01 2.23937765e-01
7.93779790e-01 5.04667938e-01 5.51608145e-01 7.76323617e-01
4.75010723e-01 6.56589270e-01 5.66844165e-01 -1.09140217e-01
3.48327979e-02 -9.41982493e-02 -5.06525580e-03 2.05870539e-01
3.67902756e-01 -6.34303272e-01 -7.55121529e-01 5.75490892e-01
-2.00118494e+00 -1.34510231e+00 -1.90669969e-01 2.41942668e+00
7.15610862e-01 1.37922689e-01 5.40929794e-01 5.04365750e-02
9.28134382e-01 -6.80523515e-02 -5.12485683e-01 -4.36241068e-02
-2.07107320e-01 2.74980277e-01 9.29513752e-01 3.53110641e-01
-1.71874571e+00 8.46320093e-01 7.20990658e+00 5.84978819e-01
-7.44728029e-01 1.09741986e-01 8.10837448e-01 -5.49621275e-03
-1.09607711e-01 -1.61148638e-01 -1.11373138e+00 6.88608468e-01
5.70326805e-01 -9.73844603e-02 -2.48452723e-02 8.60373437e-01
-3.54648113e-01 7.59256780e-02 -1.62960422e+00 9.42360580e-01
-7.99383000e-02 -1.35334384e+00 1.01562418e-01 2.77789891e-01
4.70272928e-01 -2.62377337e-02 3.30391914e-01 2.02235982e-01
7.07398534e-01 -1.08179784e+00 8.17463875e-01 5.61958075e-01
7.89996803e-01 -7.57531703e-01 6.16293192e-01 2.87397116e-01
-1.02116716e+00 -2.29271173e-01 -6.94765985e-01 -3.83842811e-02
8.07710516e-04 6.89639509e-01 -1.01443756e+00 -4.51557785e-02
5.89488566e-01 6.89142525e-01 -3.51404995e-01 1.24150324e+00
3.09456080e-01 6.40637040e-01 -6.58141971e-01 -1.94363460e-01
1.87062770e-02 -1.08314820e-01 3.72799903e-01 1.40218222e+00
8.44497755e-02 1.20707985e-03 5.28570652e-01 7.34984934e-01
-5.11837900e-01 -1.66895196e-01 -7.82270610e-01 1.68080717e-01
7.01402605e-01 1.22385228e+00 -6.82114542e-01 -4.26124573e-01
-4.81497407e-01 8.22396219e-01 6.10450685e-01 1.63341418e-01
-8.83232355e-01 -2.15239525e-01 8.38254213e-01 6.01006411e-02
6.72739327e-01 -2.37235129e-01 -2.17503682e-01 -1.00440907e+00
-4.96149622e-02 -2.75577992e-01 4.86091167e-01 -2.36165509e-01
-1.66668463e+00 2.88399071e-01 1.62925228e-01 -1.21012831e+00
-1.43257618e-01 -1.09534419e+00 -2.27488160e-01 5.36245465e-01
-1.38723040e+00 -1.15573573e+00 1.90797701e-01 4.11389202e-01
1.80823848e-01 -7.95887783e-02 8.72421741e-01 3.21329623e-01
-3.99087936e-01 6.45215154e-01 3.24958205e-01 6.18464649e-01
4.07170445e-01 -1.71002841e+00 3.51009101e-01 6.67071819e-01
4.20616210e-01 5.31306148e-01 6.07826948e-01 -2.21537843e-01
-1.05334854e+00 -1.52620816e+00 8.23285222e-01 -8.72666955e-01
6.57468021e-01 -8.32316518e-01 -5.05845487e-01 9.59625840e-01
-8.91490281e-02 5.19424558e-01 8.62169623e-01 3.09227258e-01
-3.93224329e-01 -8.32580179e-02 -1.24658453e+00 2.88138807e-01
1.25099218e+00 -5.07804513e-01 -1.70209810e-01 9.30725276e-01
7.93471515e-01 -3.02989960e-01 -9.33868408e-01 1.91439971e-01
4.63834673e-01 -6.05644047e-01 1.19878864e+00 -1.04780412e+00
2.98028022e-01 -1.12352088e-01 -4.00116265e-01 -9.78262544e-01
-5.75983405e-01 -1.82470456e-01 -3.51746202e-01 1.18085957e+00
3.04064184e-01 -4.98704761e-01 7.89330184e-01 6.24556065e-01
2.33275935e-01 -6.21366143e-01 -7.38052785e-01 -7.99360752e-01
1.33424237e-01 -3.18157583e-01 7.19897449e-01 8.29808354e-01
-4.43072468e-01 5.59687138e-01 -5.08317590e-01 4.75414723e-01
9.07157242e-01 2.69845575e-01 5.91949701e-01 -1.52323842e+00
-5.06271660e-01 -5.70684850e-01 -7.92450428e-01 -1.24649537e+00
4.75573152e-01 -1.06420135e+00 5.34039885e-02 -1.43927777e+00
5.83443940e-01 -9.39578235e-01 -4.48367268e-01 4.99931276e-01
-9.26468298e-02 3.21939796e-01 3.38012755e-01 1.04815610e-01
-1.15712273e+00 2.97920853e-01 7.47286975e-01 -2.76914030e-01
3.15498412e-01 4.50215250e-01 -5.97278357e-01 9.65701938e-01
6.60866439e-01 -6.84325635e-01 -1.22467428e-01 -5.26099443e-01
2.37671211e-01 -4.81092148e-02 7.54526973e-01 -7.55838275e-01
1.72109291e-01 -8.91216695e-02 4.84800100e-01 -8.08447778e-01
5.70001423e-01 -9.54402685e-01 -3.04263979e-01 2.23502696e-01
-5.05501807e-01 -1.47464156e-01 -1.74749345e-02 7.29023337e-01
-1.23947505e-02 -3.31564277e-01 6.78814948e-01 -2.51087938e-02
-8.82488668e-01 7.51554668e-01 -3.27246711e-02 2.55677812e-02
1.02574718e+00 -1.62901372e-01 -1.55827925e-01 -2.79964089e-01
-7.20690787e-01 3.41770113e-01 2.65986532e-01 1.60204202e-01
4.72709715e-01 -1.51179588e+00 -6.09147847e-01 -1.04319580e-01
2.44299382e-01 1.89008474e-01 -1.80191264e-01 3.22397292e-01
-1.48331314e-01 3.50927472e-01 1.59314200e-01 -8.21898460e-01
-1.47936273e+00 6.56041384e-01 4.64424908e-01 -3.30133289e-01
-3.63178611e-01 1.02018332e+00 6.31652355e-01 -4.89644140e-01
4.12679553e-01 -2.10250780e-01 -3.40213448e-01 -1.85373008e-01
4.96417969e-01 1.18393991e-02 -2.30519027e-01 -7.32209504e-01
-4.39073682e-01 2.11152583e-01 -1.83101699e-01 1.59787014e-01
1.41347873e+00 -1.23553865e-01 -1.76329464e-01 5.52522779e-01
1.50681007e+00 -2.61657119e-01 -1.01674736e+00 -5.40774465e-01
2.22040161e-01 -3.82731348e-01 -3.26962292e-01 -4.15802479e-01
-7.69878805e-01 6.67981803e-01 7.24141419e-01 3.12962174e-01
7.79875159e-01 2.87386745e-01 6.39309525e-01 8.27371895e-01
5.27646780e-01 -9.09711123e-01 2.40523983e-02 7.32247293e-01
5.24450302e-01 -1.63526380e+00 -4.20582220e-02 -5.01887858e-01
-2.68759578e-01 9.90832984e-01 5.01710653e-01 -2.09134385e-01
1.07511854e+00 4.45307344e-01 -6.32309675e-01 -1.31301358e-01
-7.17782557e-01 -5.62276363e-01 4.95668143e-01 6.40617132e-01
4.97940749e-01 2.09147051e-01 2.68716421e-02 4.58324641e-01
-2.23375887e-01 -1.02854125e-01 4.97568518e-01 6.53498292e-01
-6.47650778e-01 -9.37351882e-01 -1.65582508e-01 9.21119928e-01
-5.97144902e-01 -1.76780939e-01 -2.51305342e-01 7.36841559e-01
5.18863916e-01 7.18396366e-01 6.75370395e-01 -1.96551427e-01
4.41944785e-02 -3.45839202e-01 6.50459945e-01 -1.10138965e+00
-2.10717052e-01 -5.36935568e-01 3.04467860e-03 -5.69567233e-02
-5.98337531e-01 -8.53942096e-01 -1.11785913e+00 -3.04772735e-01
-6.10725522e-01 -8.68630111e-02 5.01639664e-01 8.91042113e-01
1.07439421e-01 2.80214071e-01 6.18144989e-01 -3.48523557e-01
-7.57720530e-01 -7.19525397e-01 -7.76932240e-01 7.01400101e-01
5.23198247e-01 -7.11428583e-01 9.33697969e-02 1.72025502e-01]
|
[9.380012512207031, 2.3904314041137695]
|
a72ec42a-41a8-4b6d-a925-ddbd3c2e45d8
|
time-series-classification-for-detecting
|
2304.11265
| null |
https://arxiv.org/abs/2304.11265v1
|
https://arxiv.org/pdf/2304.11265v1.pdf
|
Time Series Classification for Detecting Parkinson's Disease from Wrist Motions
|
Parkinson's disease (PD) is a neurodegenerative disease with frequently changing motor symptoms where continuous symptom monitoring enables more targeted treatment. Classical time series classification (TSC) and deep learning techniques have limited performance for PD symptom monitoring using wearable accelerometer data because PD movement patterns are complex, but datasets are small. We investigate InceptionTime and RandOm Convolutional KErnel Transform (ROCKET) because they are state-of-the-art for TSC and promising for PD symptom monitoring: InceptionTime's high learning capacity is suited to modeling complex movement patterns while ROCKET is suited to small datasets. We used a random search to find the highest-scoring InceptionTime architecture and compared it to ROCKET with a ridge classifier and a multi-layer perceptron (MLP) on wrist motions of PD patients. We find that all approaches are suitable for estimating tremor severity and bradykinesia presence but struggle with detecting dyskinesia. ROCKET performs better for dyskinesia, whereas InceptionTime is slightly better for tremor and bradykinesia but has much higher variability in performance. Both outperform the MLP. In conclusion, both InceptionTime and ROCKET are suitable for continuous symptom monitoring, with the choice depending on the symptom of interest and desired robustness.
|
['Sandra Hirche', 'Satoshi Endo', 'Neha Das', 'Cedric Donié']
|
2023-04-21
| null | null | null | null |
['time-series-classification']
|
['time-series']
|
[-2.61897117e-01 -2.95464218e-01 -5.08103251e-01 -1.19249215e-02
-8.08050215e-01 -6.97055161e-02 4.03886378e-01 -2.77612567e-01
-6.63031340e-01 7.78432906e-01 4.60963219e-01 -9.00447890e-02
-6.66085362e-01 -4.59892780e-01 -1.90585345e-01 -7.59742022e-01
-4.66828644e-01 9.02024567e-01 4.58718389e-01 -3.49673569e-01
-4.25871074e-01 2.13565931e-01 -9.90379632e-01 4.06736493e-01
7.73363173e-01 7.09150970e-01 3.98932993e-01 6.96173668e-01
5.04036546e-01 7.17698872e-01 -6.54716909e-01 3.39122087e-01
-1.71274155e-01 -2.49739558e-01 -6.58417583e-01 -3.31659801e-02
8.44256207e-02 -3.83440822e-01 -1.35328636e-01 4.55504656e-01
9.08960998e-01 -5.46970248e-01 6.75003648e-01 -1.11913228e+00
-3.17990661e-01 2.87838131e-01 -3.57061416e-01 6.07863367e-01
3.10598940e-01 5.29852509e-01 7.40859568e-01 -3.35801661e-01
3.81853372e-01 9.78622854e-01 1.45907676e+00 5.69606185e-01
-1.23438752e+00 -1.11967564e-01 -3.20432752e-01 5.51258266e-01
-5.23303568e-01 -3.55801322e-02 4.31361020e-01 -7.10835993e-01
1.49448407e+00 2.65703529e-01 1.32071078e+00 1.62513721e+00
6.67971611e-01 8.08061421e-01 9.93352532e-01 1.02139510e-01
4.20759201e-01 -5.78140914e-01 3.58645946e-01 3.35097671e-01
1.66570589e-01 1.98573694e-01 -3.57930243e-01 -5.78682303e-01
7.19725370e-01 2.51317173e-01 -3.20143819e-01 -4.79641035e-02
-1.68007743e+00 8.10293496e-01 4.10090774e-01 6.39576256e-01
-8.35699081e-01 4.73186105e-01 7.49623835e-01 5.69639087e-01
1.79669097e-01 1.31597102e-01 -8.18616807e-01 -7.40279794e-01
-1.20769620e+00 6.12649977e-01 4.01329011e-01 1.39482155e-01
-2.84053087e-01 1.00207344e-01 -4.44656461e-01 9.77079928e-01
3.12125653e-01 7.34564364e-01 1.32843876e+00 -8.56958389e-01
5.83144307e-01 8.05388451e-01 1.95193049e-02 -2.02166334e-01
-1.50963604e+00 -3.75477433e-01 -8.27610195e-01 5.71209073e-01
7.27801383e-01 -2.75527835e-01 -8.14101756e-01 1.36268377e+00
-1.27989069e-01 -2.93702036e-01 -1.93698317e-01 7.81107843e-01
3.71417761e-01 4.74169552e-02 1.50360122e-01 -8.21857825e-02
1.72500455e+00 -4.53425765e-01 -5.44964194e-01 -6.64764285e-01
1.17393339e+00 -2.55398937e-02 1.29987419e+00 6.13107681e-01
-7.51440883e-01 -9.94350091e-02 -8.51243317e-01 6.21452034e-02
-2.45839089e-01 8.62388432e-01 5.43216467e-01 7.10540414e-01
-1.06743622e+00 1.02172470e+00 -1.86537385e+00 -6.95103228e-01
6.92308724e-01 7.64053464e-01 -3.17521840e-01 3.22782576e-01
-7.70688891e-01 1.49416268e+00 1.67967901e-01 -5.52503280e-02
-2.52709627e-01 -5.55925071e-01 -3.81818473e-01 -4.64648962e-01
-4.83894736e-01 -9.74937141e-01 1.05714786e+00 -1.65865466e-01
-1.53907788e+00 5.63156843e-01 2.79673487e-01 -9.73280609e-01
8.41156840e-01 -2.08180979e-01 -6.54473484e-01 -6.47842437e-02
1.82779878e-01 3.90489578e-01 7.45663047e-01 -4.87722345e-02
-6.62919939e-01 -9.98385608e-01 -6.91606939e-01 -1.27846926e-01
-1.00181401e-01 1.02764174e-01 4.10690188e-01 -7.70820081e-01
2.48165444e-01 -1.19951785e+00 3.35563347e-02 1.47641212e-01
-3.38395596e-01 -4.45226192e-01 1.26797402e+00 -8.62530053e-01
9.89409864e-01 -1.60528827e+00 1.70672193e-01 -3.07474017e-01
5.40322065e-01 3.39924157e-01 3.88641298e-01 2.41412163e-01
-2.92631179e-01 -5.47183573e-01 -4.16958928e-01 -1.19649552e-01
-2.59518158e-02 4.92795408e-01 2.12550059e-01 8.62214804e-01
2.00932160e-01 1.23428810e+00 -7.60217309e-01 8.83997083e-02
4.90687042e-01 5.89193702e-01 -3.23713332e-01 -3.23692113e-01
-1.94325820e-01 5.59484124e-01 -4.44876939e-01 7.16646552e-01
-9.03036669e-02 -4.34455723e-01 -2.51934350e-01 -3.68650794e-01
-1.82924680e-02 3.68542135e-01 -5.08267701e-01 1.26391292e+00
-2.27483526e-01 6.46011114e-01 -2.89735615e-01 -9.29131150e-01
7.24807322e-01 5.03560960e-01 1.09102178e+00 -6.13901556e-01
1.51826933e-01 6.94724321e-01 2.97528327e-01 -9.85585213e-01
-4.12045598e-01 -1.79690659e-01 1.10171549e-01 3.31761569e-01
-2.93074608e-01 3.70815992e-01 1.46346360e-01 -6.05689466e-01
2.01178622e+00 2.01391637e-01 2.94077843e-01 -8.87159184e-02
8.15787092e-02 2.52288103e-01 4.05584663e-01 2.29516640e-01
-6.46676123e-01 6.14562571e-01 3.67170870e-01 -7.39587009e-01
-8.24093401e-01 -1.09013677e+00 -8.80865678e-02 7.48786688e-01
-6.44960999e-01 -2.56856263e-01 -3.85075271e-01 -4.71659869e-01
2.59398669e-01 3.67516965e-01 -5.33427894e-01 -1.62368536e-01
-7.19452500e-01 -1.49061012e+00 7.35440493e-01 1.01019299e+00
5.70948720e-01 -1.34842694e+00 -1.29273510e+00 6.06075525e-01
-9.95227322e-02 -6.41825378e-01 -6.74273819e-02 5.61830521e-01
-1.47601271e+00 -1.18029642e+00 -1.27820611e+00 -6.76429629e-01
-2.98669115e-02 -3.96211743e-01 6.86167121e-01 -7.53900051e-01
-2.31005028e-01 4.22607481e-01 -2.83473581e-01 -1.64968252e-01
-1.82574168e-01 2.40691435e-02 4.18234557e-01 -5.77536166e-01
7.41972089e-01 -1.18058550e+00 -5.98249912e-01 1.14052624e-01
-3.89973253e-01 -4.84444350e-01 8.30948830e-01 7.23566055e-01
5.72507858e-01 -8.45952034e-02 5.60671210e-01 -1.65970504e-01
1.15560436e+00 -3.77200454e-01 1.57073569e-02 -2.02096373e-01
-7.64286518e-01 8.51471722e-02 4.60290074e-01 -7.08855927e-01
-1.75250337e-01 2.06322283e-01 -4.32690859e-01 -3.46311629e-01
-1.73229247e-01 4.27647978e-01 2.69500375e-01 1.21444963e-01
1.19621718e+00 2.84789890e-01 2.05727339e-01 -8.37145209e-01
-1.19597599e-01 6.33632660e-01 7.10033476e-01 2.13916716e-03
1.09675527e-01 6.27307475e-01 -1.47033066e-01 -9.06918705e-01
-9.23261940e-02 -4.57538545e-01 -5.92860460e-01 -5.97368367e-02
1.12648225e+00 -7.61094272e-01 -7.79462278e-01 7.89858103e-01
-8.80919397e-01 -7.11030602e-01 -5.61258495e-01 9.70021546e-01
-1.09030330e+00 1.44415215e-01 -6.29213929e-01 -5.38802564e-01
-6.71471715e-01 -1.20167387e+00 1.47246027e+00 -4.66683447e-01
-9.34819221e-01 -9.53715682e-01 5.93727648e-01 9.36746374e-02
5.43500960e-01 5.72485864e-01 8.80844533e-01 -6.33696437e-01
3.60792875e-01 -4.13887829e-01 9.13555026e-02 1.37716839e-02
4.73828554e-01 -6.19822443e-01 -5.32281518e-01 -3.78214151e-01
4.60040390e-01 -3.16511154e-01 7.60231853e-01 1.02185464e+00
3.43184948e-01 -2.47386977e-01 -3.54269236e-01 2.56248206e-01
1.10401666e+00 2.77948499e-01 8.06871235e-01 1.15790737e+00
6.62861049e-01 -2.30687242e-02 -5.85838519e-02 4.21030670e-01
5.02670646e-01 1.10451913e+00 3.02362353e-01 2.01232702e-01
-3.82333905e-01 4.40585971e-01 1.00514615e+00 5.88963985e-01
-5.24437368e-01 3.51127654e-01 -9.33567047e-01 5.39842308e-01
-2.03245115e+00 -1.06619012e+00 -5.76457202e-01 1.95541453e+00
6.46164000e-01 3.50960582e-01 1.17709458e+00 6.35594130e-01
3.55208457e-01 -4.48069312e-02 -8.27669501e-01 1.25705944e-02
-2.15801135e-01 -4.82924208e-02 7.29232430e-01 -2.27233589e-01
-8.58411729e-01 3.38847309e-01 6.38396883e+00 4.78817463e-01
-1.22473490e+00 5.57563126e-01 -1.04519874e-01 -5.97874582e-01
5.75148880e-01 -5.40480793e-01 -4.82166559e-01 9.44397449e-01
1.23152554e+00 4.89434749e-01 3.47640246e-01 7.29044974e-01
8.85485232e-01 5.19152172e-03 -1.04933369e+00 1.16463065e+00
-5.99772632e-01 -1.00817919e+00 -5.51740527e-01 2.70683259e-01
7.62865096e-02 1.16879880e+00 -8.51559192e-02 1.57834411e-01
1.40159562e-01 -8.27889979e-01 4.23390180e-01 6.34964406e-01
6.05440080e-01 -1.58742145e-01 6.54371202e-01 2.15445012e-01
-1.09055007e+00 -3.88188750e-01 2.51519028e-02 1.34444848e-01
2.75699943e-01 6.33119464e-01 -6.48565054e-01 -1.23429909e-01
9.87304330e-01 1.14309227e+00 -5.95744967e-01 1.19841278e+00
-1.29408017e-01 8.16397607e-01 -8.13694715e-01 8.43879115e-03
2.40643412e-01 7.80482590e-02 7.94667423e-01 9.62539375e-01
6.74805403e-01 -5.05228937e-01 -2.80591492e-02 7.22766280e-01
6.75999165e-01 -2.99064130e-01 -4.02087569e-01 -8.56567696e-02
-6.86350614e-02 6.77928030e-01 -5.05742371e-01 2.46005822e-02
-3.21852684e-01 1.05165052e+00 -5.77267148e-02 -8.76185223e-02
-6.39280081e-01 5.08159362e-02 5.03112555e-01 1.64369628e-01
2.31857523e-01 -1.95819750e-01 -4.40490872e-01 -1.08477950e+00
3.73820692e-01 -9.40922916e-01 6.77360415e-01 -7.63070524e-01
-1.39078236e+00 4.85612392e-01 -3.22673649e-01 -1.82486737e+00
-5.11061251e-01 -1.02131236e+00 -8.89846206e-01 4.57371682e-01
-9.16125953e-01 -1.28520525e+00 1.56463552e-02 9.56072867e-01
5.05354464e-01 -1.88727826e-01 8.44764888e-01 1.45635173e-01
-5.32833755e-01 4.75004166e-02 3.88829589e-01 -1.26749724e-01
5.15746951e-01 -1.55599821e+00 3.33461195e-01 1.21840030e-01
-1.87927961e-01 2.35794067e-01 7.55728126e-01 -1.08852541e+00
-1.40559947e+00 -1.19108725e+00 8.70864868e-01 -6.95508242e-01
9.17163789e-01 2.90127963e-01 -6.87151253e-01 7.28926241e-01
-3.87475938e-01 -1.64363522e-03 4.05373901e-01 -9.43794250e-02
2.82736719e-02 8.85306746e-02 -1.38821328e+00 4.87193376e-01
8.27764809e-01 -3.31470251e-01 -1.05323994e+00 8.61486673e-01
-2.70132963e-02 -1.61371857e-01 -1.36435282e+00 3.40499789e-01
8.30782831e-01 -9.23419356e-01 9.83616352e-01 -3.75314206e-01
9.08494666e-02 -1.38293719e-02 9.55745056e-02 -1.49243188e+00
-4.96951520e-01 -4.84159052e-01 -6.04785919e-01 2.79956669e-01
3.43106449e-01 -8.75192225e-01 1.16407323e+00 3.30159754e-01
-1.20041236e-01 -8.71323407e-01 -1.13711822e+00 -1.37914491e+00
1.68673724e-01 -7.90089309e-01 1.25648826e-01 5.50211787e-01
5.70766330e-01 4.39618409e-01 -4.50939894e-01 -1.29583806e-01
3.66596997e-01 -3.04425806e-01 9.77579877e-02 -1.65644646e+00
-3.27047080e-01 -6.22493625e-01 -8.86291504e-01 -2.80842394e-01
-5.26142120e-01 -9.93079960e-01 -3.69068384e-01 -1.89388371e+00
-3.00409585e-01 -8.65665823e-02 -1.17352471e-01 9.18354452e-01
3.70703489e-01 2.68056214e-01 -2.39756271e-01 6.91759706e-01
-1.20874375e-01 4.93677169e-01 9.36738491e-01 -4.16465253e-01
-9.77031827e-01 5.88673532e-01 -2.69488782e-01 8.92375886e-01
1.14507103e+00 -4.91097301e-01 -3.52741063e-01 -3.46850961e-01
1.18547305e-01 -6.02344275e-02 7.06099689e-01 -1.48455548e+00
1.55555084e-01 5.00939250e-01 6.52377129e-01 -7.49223590e-01
6.48283720e-01 -5.78534663e-01 7.59342983e-02 1.15191340e+00
7.43614808e-02 2.28606820e-01 1.32902801e-01 5.82443416e-01
3.61088991e-01 1.96260706e-01 7.12351441e-01 -2.45412692e-01
-3.43138099e-01 3.26488554e-01 -9.78852808e-01 3.20949359e-03
4.82810348e-01 -7.33141124e-01 -4.15525317e-01 -5.15075177e-02
-1.41568422e+00 -5.89140430e-02 -3.39880935e-03 3.86594713e-01
4.44109857e-01 -1.51472867e+00 -6.78101063e-01 -7.62311220e-02
2.03977093e-01 -5.55594027e-01 -2.23459587e-01 1.64220822e+00
-3.73327076e-01 6.76788628e-01 -5.57805419e-01 -9.77228165e-01
-1.02974606e+00 6.71472698e-02 5.32879353e-01 -4.22056258e-01
-1.55699015e+00 3.69386405e-01 -7.56959558e-01 -3.37561339e-01
2.93105274e-01 -1.18463135e+00 -4.40200329e-01 1.42379045e-01
5.32687068e-01 8.58021975e-01 6.09658897e-01 -5.53843975e-01
-5.47836483e-01 6.19921625e-01 1.57364741e-01 4.14079763e-02
1.83636475e+00 2.43025959e-01 8.32178369e-02 5.50381303e-01
9.19083476e-01 -6.50484204e-01 -1.17954791e+00 1.93350464e-01
5.01131177e-01 4.35216159e-01 3.73187661e-01 -1.08018613e+00
-1.17332983e+00 4.67673063e-01 1.54798186e+00 2.76881725e-01
9.55873847e-01 1.20892331e-01 1.22506237e+00 7.44864702e-01
5.77189088e-01 -1.08213603e+00 -1.70333341e-01 3.25967878e-01
9.74136114e-01 -9.24365819e-01 1.77183393e-02 2.22659901e-01
-5.44992685e-01 1.31307781e+00 6.98171183e-02 -3.97446156e-01
5.98108292e-01 2.61490136e-01 -9.67354402e-02 -6.80133224e-01
-3.81437719e-01 -2.81557530e-01 6.43527731e-02 1.13115096e+00
1.76648974e-01 2.08197847e-01 -5.49850941e-01 8.00069511e-01
-5.09721279e-01 5.40498137e-01 2.25379080e-01 9.97692108e-01
-5.73467374e-01 -1.06803668e+00 -5.60182810e-01 1.27490652e+00
-3.07249069e-01 3.33573222e-01 -4.68398124e-01 8.87230814e-01
3.05700481e-01 7.47467399e-01 -1.41925320e-01 -4.53031719e-01
5.50038099e-01 1.26438573e-01 6.94897354e-01 -4.18035239e-01
-8.37598920e-01 1.46046057e-01 3.25244993e-01 -8.33809018e-01
-4.02823478e-01 -1.03162289e+00 -1.38445449e+00 2.16481298e-01
8.22099447e-02 -5.05279303e-01 5.35731912e-01 1.44098091e+00
1.68975130e-01 6.27124488e-01 -9.60226804e-02 -9.43417072e-01
-6.51400685e-01 -1.38495338e+00 -9.15377855e-01 4.38856035e-02
6.44275486e-01 -9.00197625e-01 -2.17527464e-01 6.51559653e-03]
|
[7.125436305999756, 0.34610024094581604]
|
349aafd4-053d-4093-bcee-f02341286d55
|
minimalist-and-high-quality-panoramic-imaging
|
2306.12992
| null |
https://arxiv.org/abs/2306.12992v1
|
https://arxiv.org/pdf/2306.12992v1.pdf
|
Minimalist and High-Quality Panoramic Imaging with PSF-aware Transformers
|
High-quality panoramic images with a Field of View (FoV) of 360-degree are essential for contemporary panoramic computer vision tasks. However, conventional imaging systems come with sophisticated lens designs and heavy optical components. This disqualifies their usage in many mobile and wearable applications where thin and portable, minimalist imaging systems are desired. In this paper, we propose a Panoramic Computational Imaging Engine (PCIE) to address minimalist and high-quality panoramic imaging. With less than three spherical lenses, a Minimalist Panoramic Imaging Prototype (MPIP) is constructed based on the design of the Panoramic Annular Lens (PAL), but with low-quality imaging results due to aberrations and small image plane size. We propose two pipelines, i.e. Aberration Correction (AC) and Super-Resolution and Aberration Correction (SR&AC), to solve the image quality problems of MPIP, with imaging sensors of small and large pixel size, respectively. To provide a universal network for the two pipelines, we leverage the information from the Point Spread Function (PSF) of the optical system and design a PSF-aware Aberration-image Recovery Transformer (PART), in which the self-attention calculation and feature extraction are guided via PSF-aware mechanisms. We train PART on synthetic image pairs from simulation and put forward the PALHQ dataset to fill the gap of real-world high-quality PAL images for low-level vision. A comprehensive variety of experiments on synthetic and real-world benchmarks demonstrates the impressive imaging results of PCIE and the effectiveness of plug-and-play PSF-aware mechanisms. We further deliver heuristic experimental findings for minimalist and high-quality panoramic imaging. Our dataset and code will be available at https://github.com/zju-jiangqi/PCIE-PART.
|
['Kaiwei Wang', 'Lei Sun', 'Hao Shi', 'Zhonghua Yi', 'Kailun Yang', 'Yao Gao', 'Shaohua Gao', 'Qi Jiang']
|
2023-06-22
| null | null | null | null |
['super-resolution']
|
['computer-vision']
|
[ 4.28445071e-01 -2.96893299e-01 2.61243761e-01 -2.52042830e-01
-4.51628089e-01 -3.23501468e-01 3.84922206e-01 -7.84088075e-01
-2.37550586e-01 2.94932455e-01 1.26776829e-01 -3.25375229e-01
-2.34369308e-01 -6.28712356e-01 -8.76924634e-01 -7.56890476e-01
2.02522054e-01 -1.64334103e-01 5.44222891e-01 -2.39063725e-02
2.01793492e-01 3.38973373e-01 -1.70163202e+00 -1.13008535e-02
1.11256278e+00 1.01530278e+00 7.45628774e-01 8.75759602e-01
5.52401364e-01 6.36134088e-01 -2.63495892e-01 -2.77903944e-01
6.23823404e-01 -2.44200736e-01 -3.58913869e-01 3.28310221e-01
1.03919268e+00 -8.04148793e-01 -4.87822115e-01 1.23485482e+00
6.42340720e-01 -2.18062580e-01 2.47876018e-01 -1.06496036e+00
-5.92521012e-01 -5.89177981e-02 -7.51067817e-01 3.27591449e-01
1.70878127e-01 7.04274774e-01 4.90809828e-01 -8.92537057e-01
4.73002702e-01 8.73514891e-01 6.32578850e-01 2.38640055e-01
-7.23196626e-01 -6.60585344e-01 -3.06061953e-01 2.84861088e-01
-9.83051479e-01 -5.03183722e-01 5.44453681e-01 -3.11152786e-01
6.58816099e-01 2.50157982e-01 8.77980113e-01 6.29319191e-01
6.64928019e-01 3.99774194e-01 1.45227659e+00 -1.22651301e-01
2.62403022e-02 -1.41266603e-02 4.22653146e-02 8.83440673e-01
3.45930576e-01 4.20124829e-01 -4.01156694e-01 1.77058309e-01
1.34573746e+00 2.82951832e-01 -1.06790996e+00 -2.25663364e-01
-1.32414448e+00 2.78247535e-01 5.92256248e-01 -8.37065354e-02
-4.03553665e-01 9.98392552e-02 -2.04199150e-01 7.30526894e-02
1.11658506e-01 6.34074926e-01 -3.36405933e-01 -5.62847704e-02
-6.86268926e-01 -4.89333421e-02 3.17702085e-01 9.48687971e-01
6.02638185e-01 -1.98152438e-01 -2.45240510e-01 9.11098003e-01
1.73688427e-01 8.34523797e-01 4.77346569e-01 -1.31501746e+00
-9.73107293e-04 4.67620462e-01 1.46167710e-01 -7.13006437e-01
-3.82276535e-01 -4.80101168e-01 -9.01774645e-01 4.85796630e-01
3.68106782e-01 6.69636130e-02 -9.61879313e-01 1.26035690e+00
3.83917749e-01 4.59668308e-01 -1.45547122e-01 1.41086316e+00
7.85486042e-01 6.82036400e-01 -7.53406286e-01 -3.59514683e-01
1.60928202e+00 -1.27935219e+00 -5.37470460e-01 -1.23508811e-01
4.96981144e-02 -1.05738151e+00 1.48696530e+00 5.81934929e-01
-1.30070746e+00 -4.70657736e-01 -1.03587866e+00 -3.06322575e-01
3.90210062e-01 2.30817005e-01 6.48067355e-01 5.46155035e-01
-1.07546031e+00 2.89893329e-01 -8.17492366e-01 -2.37628594e-01
4.16883856e-01 8.86102468e-02 -1.88042089e-01 -6.04482651e-01
-7.03368843e-01 5.44411242e-01 -1.77170858e-01 -1.82888150e-01
-8.81422937e-01 -1.22125399e+00 -5.68633556e-01 -1.29839271e-01
3.67876709e-01 -1.06454897e+00 1.07700467e+00 -6.02887332e-01
-1.78441930e+00 6.74936354e-01 -1.58088040e-02 -2.79285461e-01
2.72311300e-01 -4.25017983e-01 -2.31658340e-01 5.37049472e-01
-6.57830089e-02 4.64835286e-01 1.00115502e+00 -9.97583508e-01
-6.46365821e-01 -3.33388537e-01 1.05511472e-01 4.36193734e-01
-1.75769165e-01 1.75008938e-01 -1.07673895e+00 -3.28135699e-01
1.66629508e-01 -8.70957613e-01 -1.23300172e-01 3.50061238e-01
-5.58622897e-01 2.68394083e-01 7.29579449e-01 -6.14210010e-01
8.80896628e-01 -2.13996124e+00 -2.26409614e-01 -3.11010897e-01
4.41846609e-01 3.92656624e-01 -7.84756094e-02 -2.75183707e-01
9.52317566e-02 -6.03363991e-01 -2.87655622e-01 -2.15431720e-01
-5.86305141e-01 -1.14374772e-01 -2.31686935e-01 5.46077371e-01
-1.58336461e-01 8.87089908e-01 -8.09728324e-01 -2.97144741e-01
5.76931357e-01 5.52241862e-01 -6.61677241e-01 3.73592526e-01
-8.01050812e-02 5.15662551e-01 -2.15735167e-01 9.91905034e-01
9.94675696e-01 -6.19723678e-01 -3.61136794e-01 -8.06105673e-01
-3.60900015e-01 -4.61896621e-02 -9.32579875e-01 1.66197991e+00
-4.57131982e-01 6.69780076e-01 2.60218829e-01 -7.42577538e-02
4.55010206e-01 -1.19352110e-01 4.11098450e-01 -8.37476790e-01
1.44197747e-01 2.92161226e-01 3.41489725e-02 -6.51520431e-01
4.79525030e-01 2.64726341e-01 5.41361153e-01 2.82988995e-01
-9.87401530e-02 -6.18935168e-01 -1.01021007e-02 1.34826541e-01
1.20180821e+00 -1.90912575e-01 7.85648897e-02 -1.68517992e-01
4.18714911e-01 -2.14119673e-01 4.62419480e-01 5.77326179e-01
-2.17526168e-01 1.18067873e+00 1.61145329e-01 -1.90525591e-01
-1.32342017e+00 -1.20151687e+00 -5.64772785e-01 3.69702816e-01
5.76476872e-01 -4.38551366e-01 -7.53470778e-01 -1.18095711e-01
-4.73286659e-01 2.78378218e-01 -6.42384216e-02 1.11321226e-01
-4.55047041e-01 -9.66039121e-01 1.15533181e-01 7.55314380e-02
1.06824768e+00 -6.09912753e-01 -1.05642748e+00 -1.76870659e-01
-1.05018213e-01 -1.42567658e+00 -6.32823825e-01 -4.08162624e-01
-6.25501752e-01 -1.45298278e+00 -8.26800287e-01 -3.94817054e-01
5.73193192e-01 8.07332456e-01 9.18452740e-01 -2.77159214e-01
-4.77283061e-01 5.16931117e-01 4.73413914e-02 -1.80619910e-01
1.19970560e-01 -5.62698066e-01 1.64909914e-01 6.49839938e-02
-7.95843378e-02 -8.47761035e-01 -1.29394865e+00 6.50098622e-01
-9.62846220e-01 7.76360333e-01 1.03387451e+00 8.94738019e-01
7.35010266e-01 2.31495485e-01 -1.24121420e-01 -5.53092062e-01
1.65300101e-01 -1.77881569e-01 -1.15628171e+00 -2.07161698e-02
-7.75455594e-01 -4.54880625e-01 5.29031873e-01 -4.83995408e-01
-1.10448277e+00 -1.40415296e-01 4.87092547e-02 -7.29712784e-01
5.74916378e-02 4.99104597e-02 -2.93984294e-01 -3.46383870e-01
6.70400083e-01 4.82544273e-01 2.21871957e-01 -2.23717690e-01
1.63735002e-01 6.89967752e-01 9.37277436e-01 -1.15707137e-01
6.61488891e-01 9.10265088e-01 6.72747567e-02 -1.02748239e+00
-9.00314271e-01 -3.95592898e-01 5.53504676e-02 -2.98083127e-01
8.02856922e-01 -1.21476626e+00 -8.28782022e-01 1.08543313e+00
-8.83185267e-01 -5.08338571e-01 -1.86924487e-01 7.91320324e-01
-4.70424145e-01 4.22854841e-01 -7.12742329e-01 -1.93642348e-01
-4.80292112e-01 -1.51979196e+00 1.12512457e+00 7.21956909e-01
5.89314938e-01 -4.08622652e-01 -1.62074268e-01 7.54716337e-01
6.50131881e-01 -2.71069139e-01 3.66685718e-01 5.04798591e-01
-1.30316639e+00 2.16656074e-01 -7.18908370e-01 6.21728778e-01
3.06803044e-02 -2.38179952e-01 -1.22363853e+00 -3.40899080e-01
4.99064505e-01 -1.24672912e-01 7.11420119e-01 8.78542364e-01
1.31843483e+00 -1.73056751e-01 -1.41779721e-01 1.57888138e+00
1.58982456e+00 2.20684577e-02 9.86405551e-01 3.21200520e-01
1.05064631e+00 3.16786200e-01 7.03338563e-01 2.12414131e-01
4.99241889e-01 9.26247895e-01 6.08527958e-01 -1.57197043e-01
-5.88105381e-01 4.07132655e-02 5.10378480e-01 6.82488322e-01
-8.08494911e-02 -1.53146058e-01 -8.18823695e-01 2.94883519e-01
-1.41203654e+00 -6.20891929e-01 -3.08722526e-01 2.21902061e+00
1.02019429e+00 -2.38187104e-01 -5.21764815e-01 -1.72500730e-01
3.47590506e-01 4.29712348e-02 -7.96751738e-01 3.60903710e-01
-3.19594771e-01 1.01626953e-02 8.07582021e-01 5.21136343e-01
-1.00141561e+00 7.18836904e-01 4.95133162e+00 8.99732649e-01
-1.48395681e+00 1.99348643e-01 6.51963472e-01 -4.33960289e-01
-7.06489310e-02 -1.22110829e-01 -8.58256459e-01 7.52076209e-01
6.85358107e-01 6.99498430e-02 5.34824491e-01 6.72565818e-01
2.60379434e-01 -5.05256534e-01 -8.34193051e-01 1.57575023e+00
2.31701523e-01 -1.40217602e+00 -2.53967136e-01 1.14230208e-01
9.00009215e-01 6.45696819e-01 1.73955470e-01 -5.98167360e-01
-7.60861561e-02 -9.20716822e-01 2.62047440e-01 6.25645041e-01
1.17676783e+00 -3.51864249e-01 5.03039956e-01 1.21837527e-01
-7.53553450e-01 -6.69071451e-02 -4.67693001e-01 1.56449839e-01
2.14767709e-01 1.04741538e+00 -3.19999456e-01 4.43760902e-01
1.25011909e+00 8.93347561e-01 -6.48444593e-01 1.33177769e+00
-2.54145026e-01 5.08637249e-01 -5.70217788e-01 4.90101069e-01
-2.13744402e-01 -4.62172985e-01 8.91679227e-01 5.90859294e-01
4.86781299e-01 2.50959724e-01 -2.38805488e-01 9.88186240e-01
1.56891122e-01 -3.44720900e-01 -4.47297901e-01 5.20343482e-01
4.67443258e-01 1.61119139e+00 -3.79498899e-01 -1.47307470e-01
-5.75998843e-01 1.03019023e+00 -2.12640166e-01 3.31307888e-01
-7.20867217e-01 -2.52444595e-01 9.37900424e-01 4.52832609e-01
1.42159583e-02 -1.90564677e-01 -2.82058597e-01 -1.39327025e+00
2.82052845e-01 -9.38288331e-01 -6.54397905e-02 -1.33773470e+00
-1.09028375e+00 5.67559361e-01 -1.74851388e-01 -1.30051184e+00
2.51758963e-01 -7.91493654e-01 -6.95841849e-01 8.37605000e-01
-1.88442659e+00 -1.18332434e+00 -9.56356287e-01 9.03343618e-01
5.12714982e-01 3.20239775e-02 3.34777594e-01 5.12524009e-01
-5.42171299e-01 2.74174005e-01 1.28738180e-01 -8.84363949e-02
8.68890226e-01 -9.13749993e-01 1.47569373e-01 1.22716713e+00
-2.31465161e-01 4.73503262e-01 5.66703618e-01 -3.91421348e-01
-1.86717725e+00 -1.14254260e+00 2.31254578e-01 -5.56223154e-01
4.53594357e-01 -1.92880541e-01 -7.06982493e-01 4.38019037e-01
1.22209162e-01 3.64635289e-01 6.77044839e-02 -5.69734275e-01
-3.35361324e-02 -2.54000813e-01 -8.53543520e-01 6.73414290e-01
1.15146613e+00 -3.22165817e-01 -1.98055744e-01 5.18855095e-01
9.20967638e-01 -6.99262202e-01 -6.68817580e-01 6.00331008e-01
4.60978329e-01 -1.56249869e+00 1.28251112e+00 4.76634920e-01
8.01054537e-01 -7.62451112e-01 3.71856727e-02 -1.25126851e+00
-1.31366760e-01 -8.33873391e-01 -1.91208899e-01 9.05285418e-01
-2.18325749e-01 -9.19257224e-01 6.26614928e-01 2.73941308e-01
-5.22796214e-01 -8.35744500e-01 -6.22124016e-01 -5.44480205e-01
-5.94340920e-01 -4.81687188e-01 4.15954113e-01 7.15908408e-01
-6.14654183e-01 2.72091299e-01 -3.16244960e-01 5.09301960e-01
1.05823529e+00 2.94841230e-01 8.97536218e-01 -7.40463495e-01
-6.54283941e-01 -2.12366849e-01 -4.26837385e-01 -1.40395463e+00
-4.84812111e-01 -3.41248065e-01 -1.27266616e-01 -1.24355996e+00
4.07972574e-01 -5.71942329e-01 1.32727697e-01 -1.28056845e-02
-1.75696924e-01 5.50799310e-01 1.02846466e-01 5.46015203e-01
-3.10782909e-01 5.01817226e-01 1.63085008e+00 1.13270767e-01
-1.06236205e-01 7.31553733e-02 -5.58973551e-01 9.59101498e-01
4.84861284e-01 -3.39313713e-03 -6.67458892e-01 -6.61652029e-01
2.63526350e-01 6.81992546e-02 7.85081327e-01 -1.24961936e+00
6.42120183e-01 -8.54679383e-03 3.67222816e-01 -4.54880476e-01
5.41462779e-01 -7.77536094e-01 2.10740656e-01 3.32232565e-01
2.22578585e-01 -3.04703772e-01 -9.15019885e-02 4.04738814e-01
-1.90623701e-01 2.26668268e-01 1.15926278e+00 8.07260945e-02
-6.77276134e-01 6.18758202e-01 1.79214254e-01 -4.45054136e-02
8.75764847e-01 -2.43123263e-01 -9.93365407e-01 -2.25332633e-01
-3.92202400e-02 2.04299122e-01 1.03727949e+00 2.13389605e-01
1.03712535e+00 -9.34008956e-01 -6.51303589e-01 6.16700888e-01
2.11047962e-01 3.73707026e-01 6.07437789e-01 1.30368316e+00
-9.44257319e-01 3.50939125e-01 -4.31376427e-01 -9.25274134e-01
-1.24151111e+00 2.87920266e-01 6.72025025e-01 1.35792360e-01
-1.00763202e+00 9.81745303e-01 7.17818201e-01 -1.64029226e-01
3.60157453e-02 -3.95601124e-01 -5.89967892e-02 -6.24696970e-01
9.15429413e-01 3.49138737e-01 -8.29343200e-02 -3.74626279e-01
3.57953049e-02 9.22888696e-01 6.67334720e-03 1.18577011e-01
1.22486210e+00 -4.01281387e-01 -2.21537143e-01 6.54437905e-03
8.58599842e-01 1.47624612e-01 -1.78467727e+00 -3.14290553e-01
-8.24687719e-01 -1.03148937e+00 4.99523789e-01 -7.15247333e-01
-1.29116476e+00 7.15875924e-01 9.20382857e-01 -4.16595489e-01
1.57350194e+00 6.08347431e-02 1.00529003e+00 5.18792234e-02
3.65575373e-01 -6.65262818e-01 5.42856902e-02 3.42470646e-01
7.86211133e-01 -1.39648819e+00 8.17612782e-02 -7.13392675e-01
-4.11783904e-01 9.17087078e-01 8.53773594e-01 -3.43426764e-02
6.98345006e-01 3.99372756e-01 1.18226379e-01 -4.14273530e-01
-5.98148584e-01 7.66002983e-02 4.18703496e-01 5.70876777e-01
1.38428714e-02 -1.29728720e-01 4.27084416e-02 4.73152071e-01
-3.80833209e-01 1.30904928e-01 8.35313380e-01 5.11822104e-01
-2.92497456e-01 -4.69958067e-01 -4.63101923e-01 5.71388125e-01
-3.22733790e-01 -3.52997839e-01 3.04830611e-01 3.54991704e-01
2.04059377e-01 8.30632329e-01 2.41806835e-01 -2.14548185e-01
1.36104643e-01 -8.90048206e-01 6.13255680e-01 -3.94505739e-01
-2.29616940e-01 1.60789326e-01 -2.78286487e-01 -1.15519297e+00
-3.57980371e-01 -3.50191921e-01 -9.68752086e-01 -3.07300001e-01
-2.53042489e-01 -3.97133470e-01 7.50393510e-01 3.44211757e-01
5.97651839e-01 5.14831781e-01 6.00209832e-01 -9.57485974e-01
-2.30477184e-01 -8.31389606e-01 -7.03039050e-01 1.46160409e-01
3.44076753e-01 -4.43768322e-01 -4.19201195e-01 -3.22561897e-02]
|
[10.727479934692383, -2.542196035385132]
|
f4ea8d21-d928-4ec2-8e24-c495d445ce1c
|
dwelling-type-classification-for-disaster
|
2211.11636
| null |
https://arxiv.org/abs/2211.11636v1
|
https://arxiv.org/pdf/2211.11636v1.pdf
|
Dwelling Type Classification for Disaster Risk Assessment Using Satellite Imagery
|
Vulnerability and risk assessment of neighborhoods is essential for effective disaster preparedness. Existing traditional systems, due to dependency on time-consuming and cost-intensive field surveying, do not provide a scalable way to decipher warnings and assess the precise extent of the risk at a hyper-local level. In this work, machine learning was used to automate the process of identifying dwellings and their type to build a potentially more effective disaster vulnerability assessment system. First, satellite imageries of low-income settlements and vulnerable areas in India were used to identify 7 different dwelling types. Specifically, we formulated the dwelling type classification as a semantic segmentation task and trained a U-net based neural network model, namely TernausNet, with the data we collected. Then a risk score assessment model was employed, using the determined dwelling type along with an inundation model of the regions. The entire pipeline was deployed to multiple locations prior to natural hazards in India in 2020. Post hoc ground-truth data from those regions was collected to validate the efficacy of this model which showed promising performance. This work can aid disaster response organizations and communities at risk by providing household-level risk information that can inform preemptive actions.
|
['Juan Lavista Ferres', 'Rahul Dodhia', 'Sumedh Ranjan Ghatage', 'Sundeep Reddy Mallu', 'Anshu Sharma', 'Tina Sederholm', 'Md Nasir']
|
2022-11-16
| null | null | null | null |
['type']
|
['speech']
|
[ 1.97415665e-01 -1.54961543e-02 7.97189996e-02 -2.95986146e-01
-7.31927633e-01 -4.52763647e-01 2.83984184e-01 9.31590557e-01
-5.65576077e-01 5.96966505e-01 7.36830413e-01 -1.05523539e+00
-2.10974798e-01 -1.56027949e+00 -2.14691952e-01 -5.66904187e-01
-2.10698307e-01 1.65732741e-01 1.59643382e-01 -5.16570449e-01
4.69783247e-01 6.76911592e-01 -1.16777205e+00 1.69111326e-01
1.00867510e+00 6.68357193e-01 5.13296306e-01 5.80513597e-01
1.25486791e-01 4.76841986e-01 -2.03623101e-01 3.63392472e-01
3.54492992e-01 -4.88078035e-02 -9.91318464e-01 -4.00305718e-01
-8.15943256e-02 -9.59042668e-01 2.71767820e-03 9.58342314e-01
5.90887964e-01 2.11201131e-01 7.94516444e-01 -6.74498260e-01
-3.97544533e-01 5.17207742e-01 -4.34281439e-01 5.72889626e-01
4.89679515e-01 -7.84066841e-02 5.97212732e-01 -7.67220438e-01
-7.29925409e-02 1.28007221e+00 9.82876480e-01 -2.19933584e-01
-1.07919729e+00 -7.01526046e-01 7.98504502e-02 1.87535539e-01
-1.42801940e+00 -2.59330928e-01 4.10721898e-01 -7.48330414e-01
1.31724119e+00 1.58254936e-01 3.20373654e-01 3.63089770e-01
-6.24913909e-03 9.95838791e-02 1.07471526e+00 -4.37290132e-01
4.10908848e-01 -1.15474559e-01 9.51567781e-04 4.07446951e-01
3.70251358e-01 9.01794657e-02 1.27937493e-03 -2.04052359e-01
5.19783437e-01 3.71525466e-01 5.90104535e-02 6.51074767e-01
-8.14623713e-01 9.33439076e-01 1.11965752e+00 3.82511377e-01
-8.32456291e-01 -2.54783034e-01 2.08435327e-01 -1.20555826e-01
6.70378983e-01 -2.18013842e-02 -1.85522005e-01 3.70382458e-01
-1.10694110e+00 -3.24306674e-02 3.41327786e-01 -1.54448032e-01
9.04272795e-01 -8.82579312e-02 1.08748026e-01 5.96349299e-01
5.54254770e-01 6.95770502e-01 -1.30795985e-01 -6.81836367e-01
8.07047665e-01 9.28314805e-01 3.13222736e-01 -1.64623082e+00
-7.71758676e-01 2.20835716e-01 -8.91138911e-01 3.23192775e-01
3.96834284e-01 -6.65024221e-01 -9.95233595e-01 1.33503520e+00
3.95652175e-01 1.05869636e-01 6.54101670e-02 6.67778075e-01
9.87871364e-02 8.50771427e-01 6.73707843e-01 2.34819651e-01
1.19816756e+00 -3.90764102e-02 -3.10065717e-01 -2.14900270e-01
6.25393152e-01 -3.67586643e-01 7.16170669e-01 -8.53066519e-02
-3.51346225e-01 -2.87850797e-01 -9.46792483e-01 4.52190042e-01
-8.02418053e-01 -2.75016844e-01 3.84851068e-01 6.55867934e-01
-1.10415399e+00 5.24985611e-01 -8.60412657e-01 -9.61646557e-01
6.43444002e-01 3.31105515e-02 -3.54883611e-01 4.17931415e-02
-1.49937010e+00 1.26342356e+00 5.30853510e-01 5.42574346e-01
-8.55414748e-01 -7.08461285e-01 -9.68992174e-01 8.27643797e-02
-1.29335821e-01 -1.15559526e-01 6.25424564e-01 -4.53840554e-01
-4.31314379e-01 3.97943944e-01 2.22696975e-01 -3.33980560e-01
1.97952718e-01 -7.38322884e-02 -4.03144538e-01 2.94273078e-01
9.28396344e-01 3.81758422e-01 4.90290858e-02 -1.11013305e+00
-1.20358503e+00 -6.53007209e-01 -3.46120670e-02 3.59744072e-01
-3.43412071e-01 5.50907731e-01 6.73685074e-01 -6.74014568e-01
3.00465852e-01 -4.76646066e-01 -5.85307300e-01 -4.62055743e-01
-3.71242344e-01 1.41560867e-01 4.76406574e-01 -1.36225367e+00
1.64155149e+00 -1.69853270e+00 -7.03873396e-01 5.05237162e-01
-1.61582023e-01 4.68019426e-01 2.12074935e-01 9.04392421e-01
6.78061768e-02 3.74316633e-01 -6.62968814e-01 2.89348871e-01
-2.17019171e-01 1.12745270e-01 -2.68910974e-01 4.09704387e-01
3.71414423e-01 3.19916695e-01 -1.11991715e+00 -3.04412216e-01
5.48026085e-01 4.37307745e-01 -2.71212578e-01 3.78555000e-01
2.31432930e-01 3.82981151e-01 -6.74519539e-01 8.59280348e-01
8.39927137e-01 4.51683730e-01 1.64312139e-01 1.53804958e-01
-6.84449911e-01 2.35639274e-01 -1.02551270e+00 7.13975608e-01
-3.81351501e-01 3.17827523e-01 7.86512792e-02 -1.13668191e+00
1.05353498e+00 3.65771562e-01 4.95787412e-01 -6.79050982e-01
-2.30770841e-01 1.56023204e-01 -4.78206486e-01 -8.74092638e-01
4.24076408e-01 -2.95462936e-01 -3.13701957e-01 6.22955203e-01
-6.44570112e-01 3.19820940e-01 -2.92198062e-01 -1.19974673e-01
1.14490151e+00 -2.01086968e-01 2.70528734e-01 -6.05250776e-01
3.05934548e-01 3.74315411e-01 4.89635676e-01 5.44185281e-01
-4.90261853e-01 5.38765311e-01 1.54979244e-01 -8.44354391e-01
-7.49548316e-01 -1.02359927e+00 -3.12215596e-01 1.02665222e+00
-1.48071051e-01 2.36024722e-01 -8.42794895e-01 -4.54867750e-01
-1.51868105e-01 1.00770569e+00 -6.51228309e-01 1.16957806e-01
-3.99225265e-01 -1.36463141e+00 8.51457775e-01 6.47191048e-01
1.04518449e+00 -1.24877858e+00 -1.03820264e+00 5.48654079e-01
-3.47839057e-01 -4.47693974e-01 8.34446698e-02 2.98318267e-01
-6.52152717e-01 -1.29300618e+00 -5.91750622e-01 -7.15003431e-01
9.17401969e-01 5.76780260e-01 4.99954402e-01 1.62984520e-01
-6.80641830e-02 3.62725668e-02 -2.54135221e-01 -3.30258638e-01
-1.34249434e-01 -7.76316747e-02 6.64391369e-02 -1.90569445e-01
4.55001891e-01 -7.43390858e-01 -9.79303300e-01 2.69078046e-01
-8.31376255e-01 -2.84616679e-01 4.31196034e-01 2.83997744e-01
1.96861491e-01 6.74105883e-01 9.11601007e-01 -4.59127158e-01
4.54929769e-01 -1.19037831e+00 -2.81259924e-01 3.68127018e-01
-4.25368398e-01 -3.88718694e-01 2.21759021e-01 5.01366891e-02
-1.44582260e+00 2.84553766e-01 -3.82544756e-01 8.69655192e-01
-6.92105949e-01 1.02665234e+00 -3.29357237e-01 3.46956015e-01
7.25474715e-01 -2.37586826e-01 -4.39731270e-01 -5.31441808e-01
-6.12920336e-02 1.02108991e+00 6.29610598e-01 -4.89153594e-01
1.00077486e+00 4.57175165e-01 -2.67115176e-01 -9.59935308e-01
-8.16758752e-01 -6.85418367e-01 -1.03451419e+00 -3.99451792e-01
1.09343553e+00 -9.78943527e-01 -5.57611994e-02 7.44265079e-01
-9.83614683e-01 -5.86213946e-01 4.22885746e-01 2.75348812e-01
1.06642276e-01 9.87461805e-02 -1.46218255e-01 -1.14269650e+00
-5.26935637e-01 -5.35160124e-01 6.04979873e-01 4.24387991e-01
-1.08780526e-01 -1.14541221e+00 3.99733692e-01 3.95260066e-01
7.08604991e-01 8.85459363e-01 9.30346191e-01 -2.93755203e-01
-3.89097966e-02 -2.48011306e-01 -6.63051009e-01 1.12256140e-01
5.98461032e-01 -2.00601235e-01 -9.08298790e-01 2.96976995e-02
-3.26386839e-01 -4.92751673e-02 8.86598349e-01 4.58847493e-01
4.14411306e-01 -7.61624515e-01 -2.97560662e-01 1.93756729e-01
1.72738874e+00 3.19046825e-01 7.07843184e-01 8.46087337e-01
6.10235333e-01 1.14058399e+00 6.16112709e-01 5.78660548e-01
9.10062075e-01 2.74299115e-01 7.20081866e-01 -4.81707662e-01
3.20198208e-01 -3.02088678e-01 4.56067830e-01 -5.68727627e-02
-1.08773477e-01 8.32530782e-02 -1.63869345e+00 1.06212783e+00
-1.78035676e+00 -1.23315513e+00 -2.83159286e-01 2.18744183e+00
5.26607335e-01 -1.18867144e-01 2.49591932e-01 3.51314336e-01
7.50433862e-01 -2.61122845e-02 -1.14372186e-01 -2.85635978e-01
6.67100623e-02 -3.10255066e-02 8.60189438e-01 8.08588922e-01
-1.39903474e+00 1.00977004e+00 6.31610394e+00 1.92552552e-01
-9.34608877e-01 -7.52813295e-02 8.42833877e-01 4.48370457e-01
-2.12613344e-01 1.55504301e-01 -5.22395074e-01 2.78780311e-01
1.14142013e+00 2.97582000e-01 3.07693064e-01 3.83269697e-01
1.08853531e+00 -6.82006598e-01 1.30679328e-02 -7.91314542e-02
-4.80957776e-01 -1.00742567e+00 -3.04115415e-01 -2.73528006e-02
6.28262341e-01 1.47011280e-01 -4.30936158e-01 -7.73165748e-02
7.96848178e-01 -9.18143332e-01 5.16420126e-01 6.24674976e-01
5.11469960e-01 -8.35074723e-01 7.92287171e-01 4.60159004e-01
-1.47382629e+00 -4.27348137e-01 -2.73763776e-01 -6.42212570e-01
4.01427120e-01 3.54752541e-01 -9.49814439e-01 3.51780057e-01
1.04891455e+00 3.75541389e-01 -6.50845528e-01 8.03978682e-01
-3.88796061e-01 8.38307917e-01 -6.41488791e-01 4.53709215e-01
4.81101543e-01 -5.06863035e-02 1.05377898e-01 1.25577259e+00
5.77031434e-01 6.07556462e-01 2.91843265e-01 5.39535820e-01
6.51556849e-01 -1.42347306e-01 -8.27498376e-01 4.53790367e-01
7.95143068e-01 1.04246557e+00 -9.62826252e-01 -1.61979664e-02
-1.39755517e-01 5.90997458e-01 2.34110981e-01 4.60259348e-01
-4.82855886e-01 -5.20495415e-01 6.64005458e-01 4.24666047e-01
-5.35265505e-02 -3.94720882e-01 -4.70459223e-01 -5.92620313e-01
-4.16778117e-01 -3.21306735e-01 6.04796827e-01 -4.71284062e-01
-9.76131320e-01 3.97704899e-01 4.17195201e-01 -5.77575326e-01
1.54516757e-01 -1.35913998e-01 -1.48989844e+00 1.21537077e+00
-1.62511218e+00 -1.40478992e+00 -3.75596195e-01 4.60975796e-01
1.57804623e-01 5.48491515e-02 9.43179131e-01 3.44397932e-01
-9.84729648e-01 6.26241118e-02 1.83581486e-02 2.94504911e-01
4.76412661e-02 -1.24418926e+00 3.82432014e-01 1.25451291e+00
-7.58181393e-01 4.41341579e-01 4.94434178e-01 -1.27077270e+00
-5.79741299e-01 -1.57781029e+00 1.38655472e+00 -1.30747229e-01
8.24421823e-01 -4.40261420e-03 -9.10425305e-01 4.16805476e-01
-2.22001582e-01 -4.17533636e-01 7.37205923e-01 -1.56795099e-01
1.37754336e-01 -7.41785467e-02 -1.69891298e+00 4.14782137e-01
6.31645620e-01 -5.98701835e-01 -5.96998870e-01 1.82402790e-01
3.61688018e-01 5.17423689e-01 -8.15632999e-01 1.72136456e-01
3.20278674e-01 -9.58864033e-01 1.05650401e+00 -2.36485794e-01
2.42600396e-01 -2.93411702e-01 -6.71111763e-01 -1.08521283e+00
-5.33957005e-01 1.96350887e-02 6.31950915e-01 1.61010861e+00
4.51992482e-01 -7.04711974e-01 5.61352849e-01 1.02012432e+00
-7.10229799e-02 -2.64627397e-01 -9.44378078e-01 -3.72121245e-01
2.79226959e-01 -6.29731953e-01 7.89429545e-01 7.60006905e-01
-9.72717106e-02 -1.36956796e-01 4.93260585e-02 1.04593277e+00
6.93763435e-01 -5.23187757e-01 1.20946042e-01 -9.86426294e-01
6.35069251e-01 -3.17216814e-01 -1.80443808e-01 1.37214348e-01
-1.25039846e-01 -4.72663939e-01 1.18787318e-01 -1.96988761e+00
2.86334679e-02 -9.51601744e-01 -4.91494119e-01 1.17198658e+00
-5.01549721e-01 2.07563430e-01 -2.12168470e-01 2.24457517e-01
1.66399732e-01 3.05535346e-01 1.29672453e-01 -1.93910226e-01
-4.64524746e-01 2.03466401e-01 -9.55061078e-01 8.07410657e-01
1.17632806e+00 -5.23196220e-01 -2.29199335e-01 -5.85889995e-01
2.05621734e-01 -7.52453357e-02 7.93955326e-01 -1.28163218e+00
1.06387503e-01 -7.65201986e-01 2.96018660e-01 -8.91731918e-01
-3.99346948e-01 -9.40486133e-01 3.46904039e-01 7.51506031e-01
-9.55336019e-02 1.32340472e-02 2.54782081e-01 2.30414167e-01
8.42485726e-02 -2.27156445e-01 5.64497173e-01 1.67654771e-02
-7.26127207e-01 2.82044828e-01 -9.31145370e-01 -4.51311707e-01
1.05081868e+00 -3.32029015e-01 -1.98194146e-01 -4.81141657e-02
-4.29874510e-01 4.41599280e-01 4.46377784e-01 -1.85628445e-03
5.77443302e-01 -1.22640991e+00 -9.78015125e-01 1.06766008e-01
-1.44127890e-01 3.70264947e-02 4.42750275e-01 3.77832234e-01
-9.85865533e-01 2.03814879e-01 -4.95459616e-01 1.05344966e-01
-6.59849286e-01 3.46480638e-01 4.06914562e-01 -2.42011994e-01
-4.75564212e-01 3.87771457e-01 -1.29633039e-01 -5.28471410e-01
-2.15396415e-02 -2.81054169e-01 -7.92505443e-01 4.95658755e-01
9.38373506e-01 8.06104958e-01 -1.82732791e-01 -1.00915158e+00
-6.00659788e-01 3.46088082e-01 5.83358288e-01 -3.48504663e-01
1.69004989e+00 -6.14679575e-01 -1.14390671e-01 -6.13069572e-02
8.86078417e-01 -4.98250932e-01 -1.46084476e+00 -4.70347852e-02
5.36826432e-01 -2.02632889e-01 6.03845358e-01 -9.74740863e-01
-9.02449548e-01 8.49549413e-01 8.24754298e-01 2.17340082e-01
1.36995935e+00 -4.01662111e-01 8.88200879e-01 3.04315448e-01
2.69766808e-01 -1.24603522e+00 -4.88656700e-01 2.35597953e-01
9.90639687e-01 -1.51068687e+00 -5.29968031e-02 1.21292308e-01
-2.95964986e-01 9.05233502e-01 3.36537600e-01 8.40126500e-02
1.00814748e+00 3.57913896e-02 1.75718009e-01 -9.11565498e-02
1.34090394e-01 -4.04573500e-01 -1.55348837e-01 1.17659950e+00
6.60794824e-02 5.84669113e-01 2.13404745e-01 5.86341023e-01
1.96299478e-01 -1.60876572e-01 2.47709289e-01 9.28050280e-01
-1.14649975e+00 -8.01051974e-01 -8.86641324e-01 4.24448907e-01
-3.20396364e-01 -3.88208658e-01 -1.15016453e-01 3.40260774e-01
3.44586641e-01 1.54713845e+00 -1.15706004e-01 -4.08966690e-01
2.31886983e-01 -4.70465690e-01 -4.19786304e-01 -6.28047109e-01
-8.32396865e-01 -3.49717915e-01 1.72780588e-01 -1.93664372e-01
-2.56893367e-01 -9.33331132e-01 -1.34421122e+00 -5.60009062e-01
7.93243200e-02 7.06848055e-02 7.04155564e-01 1.02147663e+00
-3.25000472e-02 -3.10354177e-02 1.05756831e+00 -1.35839248e+00
-2.38878235e-01 -1.19842505e+00 -5.07072330e-01 2.97897086e-02
3.22333127e-01 -4.89427656e-01 -1.45199597e-01 -1.35917619e-01]
|
[9.444388389587402, -1.3393168449401855]
|
064daa1b-1866-4fc3-8e2e-47ce59809497
|
colonoscopy-coverage-revisited-identifying
|
2305.10026
| null |
https://arxiv.org/abs/2305.10026v1
|
https://arxiv.org/pdf/2305.10026v1.pdf
|
Colonoscopy Coverage Revisited: Identifying Scanning Gaps in Real-Time
|
Colonoscopy is the most widely used medical technique for preventing Colorectal Cancer, by detecting and removing polyps before they become malignant. Recent studies show that around one quarter of the existing polyps are routinely missed. While some of these do appear in the endoscopist's field of view, others are missed due to a partial coverage of the colon. The task of detecting and marking unseen regions of the colon has been addressed in recent work, where the common approach is based on dense 3D reconstruction, which proves to be challenging due to lack of 3D ground truth and periods with poor visual content. In this paper we propose a novel and complementary method to detect deficient local coverage in real-time for video segments where a reliable 3D reconstruction is impossible. Our method aims to identify skips along the colon caused by a drifted position of the endoscope during poor visibility time intervals. The proposed solution consists of two phases. During the first, time segments with good visibility of the colon and gaps between them are identified. During the second phase, a trained model operates on each gap, answering the question: Do you observe the same scene before and after the gap? If the answer is negative, the endoscopist is alerted and can be directed to the appropriate area in real-time. The second phase model is trained using a contrastive loss based on auto-generated examples. Our method evaluation on a dataset of 250 procedures annotated by trained physicians provides sensitivity of 0.75 with specificity of 0.9.
|
['E. Rivlin', 'M. Elad', 'R. Goldenberg', 'I. Kligvasser', 'G. Leifman']
|
2023-05-17
| null | null | null | null |
['3d-reconstruction', 'specificity']
|
['computer-vision', 'natural-language-processing']
|
[ 3.14490885e-01 3.85492712e-01 -1.42206065e-03 3.14073563e-01
-5.89893818e-01 -9.01601315e-01 1.39694393e-01 8.51968884e-01
-4.10666704e-01 4.52487767e-01 -1.10366426e-01 -6.91648364e-01
6.40697777e-02 -6.84241295e-01 -8.22766304e-01 -6.78493142e-01
-3.04565877e-01 1.86129153e-01 7.98343301e-01 9.09367502e-02
1.49997950e-01 3.64208639e-01 -1.09508562e+00 3.20221037e-01
8.76509845e-01 3.50920707e-01 6.45084620e-01 8.79599214e-01
2.05853269e-01 5.26928306e-01 -5.62569082e-01 -3.44247334e-02
4.34178352e-01 -5.73056877e-01 -4.87343937e-01 1.16250709e-01
1.90541714e-01 -2.25125566e-01 3.90259689e-03 1.19913781e+00
3.50179553e-01 -2.54801176e-02 3.63901943e-01 -3.42761427e-01
3.55114162e-01 2.18291759e-01 -5.18489420e-01 4.51433241e-01
6.65421605e-01 2.14555651e-01 3.03209722e-01 -4.13787842e-01
7.37899244e-01 5.44853330e-01 8.59708369e-01 2.99809515e-01
-1.00294852e+00 -9.02915448e-02 2.96619684e-02 -2.61327833e-01
-1.06243885e+00 1.09055817e-01 4.61507410e-01 -6.92303240e-01
5.63625932e-01 3.20865005e-01 9.59027171e-01 5.00609756e-01
4.63094354e-01 3.98065269e-01 8.03090453e-01 -7.41923928e-01
1.27409268e-02 3.87284845e-01 -2.41436675e-01 9.37131763e-01
6.79503322e-01 5.03792584e-01 1.09387875e-01 -1.63458034e-01
7.98735023e-01 3.23391795e-01 -6.34957790e-01 -6.54688537e-01
-1.37970340e+00 6.18566930e-01 5.91358721e-01 5.01488209e-01
-6.43448114e-01 -4.12632823e-01 4.48870182e-01 1.45915642e-01
5.40203452e-02 6.68624401e-01 5.88308014e-02 6.72904998e-02
-9.08159912e-01 -2.00723365e-01 8.05292785e-01 6.12365842e-01
2.26254284e-01 -4.97890234e-01 6.81805983e-02 2.15647012e-01
1.37777343e-01 2.19405517e-01 6.08259320e-01 -1.65076628e-01
2.84386009e-01 9.64821875e-01 5.21283567e-01 -1.04994118e+00
-4.67813581e-01 -6.29939616e-01 -6.37381017e-01 3.67166907e-01
7.89095640e-01 -1.88138202e-01 -8.78432989e-01 9.44805861e-01
6.49653614e-01 6.25165552e-02 6.71198666e-02 1.08879256e+00
6.64543808e-01 4.15564477e-01 -1.41771480e-01 -2.45830238e-01
1.53568351e+00 -8.54593992e-01 -4.26153749e-01 -4.32796896e-01
7.29507685e-01 -1.17059040e+00 5.38533330e-01 5.58571160e-01
-1.15102339e+00 -4.71823215e-01 -1.15951347e+00 3.76394749e-01
-9.12539363e-02 4.89147872e-01 4.06317785e-02 6.76440775e-01
-7.74253845e-01 4.31676775e-01 -1.20172346e+00 -4.50075418e-01
-3.29392441e-02 1.78920165e-01 -5.13259709e-01 -2.79205680e-01
-7.00372338e-01 9.55227792e-01 4.52957094e-01 2.05422729e-01
-8.56090665e-01 -4.48366076e-01 -1.03548217e+00 -9.59287733e-02
5.09700477e-01 -6.05033755e-01 1.14865994e+00 -8.44481289e-01
-1.09945357e+00 1.11486328e+00 -6.71565309e-02 -6.49145126e-01
8.91449392e-01 -1.18730985e-01 -2.29179516e-01 5.22237659e-01
1.86793916e-02 1.16015740e-01 8.42432201e-01 -1.19156921e+00
-1.04139078e+00 -3.98392230e-01 3.41565192e-01 2.20488444e-01
2.04267636e-01 -4.64341521e-01 -5.68056047e-01 -4.46361870e-01
3.37549835e-01 -1.14428484e+00 -4.80554879e-01 1.49727285e-01
-5.26730835e-01 6.65281355e-01 4.84342307e-01 -1.02519059e+00
1.30886436e+00 -2.14452195e+00 -2.81747103e-01 2.36641988e-01
1.28137410e-01 7.27490962e-01 3.24647009e-01 2.67034352e-01
5.28550968e-02 -1.69910297e-01 -2.05440208e-01 1.03837214e-01
-7.13632703e-01 -1.82149813e-01 8.78991261e-02 9.00721550e-01
-1.50063515e-01 3.15684170e-01 -1.30440164e+00 -4.80554432e-01
7.77002871e-01 4.52615619e-01 -3.25579852e-01 4.14507896e-01
-2.54102368e-02 7.89053500e-01 -3.49480629e-01 5.59961259e-01
7.44200230e-01 -2.22408757e-01 4.52587634e-01 -2.10750297e-01
-5.25010526e-01 1.93757802e-01 -1.30447781e+00 1.49867880e+00
-5.94907045e-01 2.83346087e-01 9.72120687e-02 -8.41229200e-01
6.40878737e-01 6.35820627e-01 2.74334818e-01 -4.10150498e-01
-3.71684022e-02 6.29087150e-01 8.63813460e-02 -9.27443981e-01
2.43480444e-01 1.47117913e-01 2.69734383e-01 7.50843063e-02
-5.22310138e-01 -9.52849314e-02 3.51256073e-01 -1.49570748e-01
1.21831429e+00 1.58633024e-01 9.81386721e-01 -1.12456664e-01
7.32580364e-01 4.62005615e-01 1.30497605e-01 7.96243966e-01
-3.86946172e-01 6.54486656e-01 4.26993042e-01 -6.18776619e-01
-8.19948614e-01 -7.62963891e-01 -2.30176952e-02 -2.50086677e-03
6.32180452e-01 1.07238263e-01 -5.49603105e-01 -9.12182152e-01
-2.59891361e-01 3.47905755e-01 -6.78369343e-01 -1.91665143e-02
-7.34230518e-01 -4.35301274e-01 -2.14332610e-01 -6.38585836e-02
3.35315794e-01 -7.64710963e-01 -1.59661233e+00 3.68189871e-01
-2.54619539e-01 -8.19575310e-01 -2.36837313e-01 6.01037033e-02
-1.14219916e+00 -1.71770895e+00 -1.11262584e+00 -1.02874887e+00
1.23317230e+00 7.38072872e-01 1.13628685e+00 3.02417040e-01
-6.48225784e-01 5.70523329e-02 -4.44347382e-01 -2.00798839e-01
-7.78538167e-01 -1.55474111e-01 -7.05325842e-01 -3.78694504e-01
9.12372544e-02 9.66279954e-02 -1.18614495e+00 5.42942703e-01
-6.82816267e-01 -4.65964898e-02 6.54532075e-01 9.06666577e-01
6.81935370e-01 -3.05667073e-02 -2.09450833e-02 -1.05373645e+00
2.74951547e-01 -5.27550757e-01 -9.41870809e-01 9.29438174e-02
-1.11799248e-01 -1.91983327e-01 6.22118175e-01 -4.56473708e-01
-7.55808711e-01 5.25248468e-01 -8.19633454e-02 -1.97737917e-01
-3.22628170e-01 4.03245360e-01 7.14866817e-01 -1.19412646e-01
1.00414252e+00 5.65059632e-02 1.06863841e-01 -2.38282725e-01
-3.59043717e-01 2.10413173e-01 4.10322368e-01 2.84822166e-01
5.17649055e-01 7.42812276e-01 -6.08374625e-02 -6.71539187e-01
-6.67933047e-01 -1.11497688e+00 -2.85026789e-01 -3.18651319e-01
7.60170043e-01 -7.09380090e-01 -3.26398969e-01 -1.73583686e-01
-9.80659664e-01 -1.96638778e-02 -4.18886393e-01 9.54904675e-01
-4.18642163e-02 6.22813404e-01 -4.10941631e-01 -8.21253538e-01
-2.47206628e-01 -1.35179162e+00 9.25170898e-01 3.69609505e-01
1.02759330e-02 -1.10320771e+00 2.14140683e-01 -8.17219391e-02
1.92384258e-01 6.74625635e-01 6.31081939e-01 -5.14056921e-01
-5.72935581e-01 -7.97427952e-01 1.15339451e-01 4.09580208e-02
3.89075965e-01 -1.61662236e-01 -7.37963676e-01 -4.90688533e-01
2.06220999e-01 4.96211171e-01 6.69627428e-01 8.47276986e-01
6.24378920e-01 5.65357469e-02 -8.22747648e-01 4.05427039e-01
1.63792181e+00 5.36445320e-01 3.07779938e-01 4.17061329e-01
1.10236257e-01 5.28024137e-01 1.00929821e+00 9.10958797e-02
-1.59246981e-01 4.52193707e-01 1.00954628e+00 -5.01873016e-01
-1.14999108e-01 -1.64839804e-01 2.14150876e-01 3.69708329e-01
-6.67344108e-02 -2.33701229e-01 -8.15891027e-01 9.38497424e-01
-1.55044663e+00 -6.62398398e-01 -3.76900136e-01 2.92634702e+00
2.63816714e-01 3.99750978e-01 1.04109526e-01 1.79854199e-01
8.08895051e-01 -2.09788904e-01 -1.92477524e-01 -7.45049939e-02
2.83054024e-01 -3.60642582e-01 6.48801029e-01 7.48332441e-01
-1.23754668e+00 1.24621131e-01 5.09388208e+00 2.00604916e-01
-1.28908241e+00 -2.63531078e-02 4.26442295e-01 9.98411477e-02
1.15342349e-01 6.24893792e-02 -6.26740932e-01 4.55498576e-01
2.36063942e-01 3.04589152e-01 -1.53126061e-01 5.81806004e-01
3.10018450e-01 -7.54012585e-01 -9.39980924e-01 7.09048569e-01
1.59505472e-01 -1.07644725e+00 -5.89727461e-01 2.43822765e-02
8.06482673e-01 -2.32035443e-01 -2.96103448e-01 -1.34009883e-01
-3.26917142e-01 -6.40357852e-01 3.61967981e-01 4.45089340e-01
7.76094854e-01 -4.43827063e-01 9.95137274e-01 6.26144946e-01
-1.16098583e+00 3.32531929e-02 -1.43300638e-01 2.07079336e-01
2.34232321e-01 7.70708382e-01 -1.51272833e+00 5.12202919e-01
3.84816170e-01 4.15811062e-01 -3.20105880e-01 1.97690034e+00
-4.33045447e-01 3.26745719e-01 -3.46808255e-01 5.06605990e-02
2.78369993e-01 2.22801752e-02 1.14207864e+00 1.29692650e+00
8.33473980e-01 -3.38644207e-01 1.84739724e-01 4.01545525e-01
5.17730117e-01 2.22956181e-01 -8.67353976e-01 5.02612948e-01
2.02039540e-01 1.12603986e+00 -9.73071814e-01 -2.49808073e-01
-4.54171419e-01 9.40034509e-01 -3.62803757e-01 7.16577917e-02
-6.00020528e-01 -2.83907413e-01 -4.70504463e-01 4.69799370e-01
2.85064548e-01 2.68475503e-01 2.93366700e-01 -9.81546521e-01
3.56113225e-01 -6.53722942e-01 5.30859947e-01 -3.83084357e-01
-5.64101636e-01 7.05465198e-01 -2.83556968e-01 -1.89404011e+00
-5.17676592e-01 -5.79165220e-01 -6.69769049e-01 7.17157304e-01
-1.68893003e+00 -7.55619109e-01 -6.33382380e-01 3.48629773e-01
6.55679166e-01 3.44854057e-01 1.04103696e+00 4.43795234e-01
-7.30612651e-02 1.64808959e-01 -3.22635937e-03 -1.96452543e-01
6.01409376e-01 -1.51135457e+00 3.71260792e-02 1.08591342e+00
-2.03647152e-01 4.02534813e-01 8.54819238e-01 -8.25614572e-01
-1.02235019e+00 -8.10614288e-01 9.69158113e-01 -6.97755292e-02
6.36787266e-02 4.38050404e-02 -7.61210203e-01 4.30059731e-01
-9.04278830e-02 1.67388618e-01 3.40215683e-01 -4.50110853e-01
2.46713176e-01 2.47341111e-01 -1.38949716e+00 3.96441549e-01
2.68894374e-01 -2.21505365e-03 -5.59243441e-01 5.27628422e-01
2.11163297e-01 -9.96671855e-01 -6.07048273e-01 2.50442892e-01
4.54481572e-01 -1.34844661e+00 8.87508333e-01 1.40875280e-01
2.09629446e-01 -4.30948973e-01 5.75659275e-01 -1.33335268e+00
2.43198663e-01 -7.09262073e-01 9.55699310e-02 2.10841075e-01
4.48794037e-01 -4.48623210e-01 9.16058540e-01 -1.49144724e-01
-2.47676611e-01 -5.11700094e-01 -8.08203280e-01 -3.05140406e-01
-5.46667218e-01 1.96528777e-01 -3.21081728e-02 7.30100989e-01
1.17583863e-01 -2.97263205e-01 5.55774793e-02 5.98180890e-01
3.97624373e-01 2.55008131e-01 6.00401819e-01 -9.83926535e-01
-4.07514483e-01 -8.22001770e-02 -4.24761832e-01 -9.96956050e-01
-9.74682152e-01 -5.00404418e-01 1.26857996e-01 -1.72699821e+00
2.40460020e-02 -2.35516369e-01 1.04661040e-01 -7.58826286e-02
-2.12329313e-01 2.40276810e-02 -1.01504810e-01 1.98813379e-01
-3.05289745e-01 -5.00782490e-01 1.45561898e+00 1.95341304e-01
-5.45471489e-01 8.30323458e-01 -8.83971974e-02 1.11180806e+00
6.61609828e-01 -4.58343297e-01 -3.09857100e-01 -1.45358577e-01
3.42856258e-01 6.03371620e-01 6.20966077e-01 -1.14867127e+00
2.88462162e-01 4.06820863e-01 2.13866577e-01 -8.12242985e-01
2.08314806e-01 -1.28139901e+00 2.52133340e-01 1.46728253e+00
5.89883737e-02 1.16327256e-01 3.01788181e-01 6.38620138e-01
-4.51157033e-01 -8.49437058e-01 9.03434455e-01 -7.26009488e-01
-3.76971304e-01 -2.08641887e-01 -4.35075641e-01 -1.70197904e-01
1.29386497e+00 -4.26992536e-01 1.19191837e-02 -4.16221991e-02
-1.12055361e+00 -5.66445738e-02 5.18136799e-01 -1.38416246e-01
6.40846550e-01 -5.73891461e-01 -7.01813042e-01 3.59000623e-01
8.08792710e-02 2.98866004e-01 6.24193072e-01 1.41540182e+00
-1.29400551e+00 4.81065422e-01 2.02011943e-01 -7.92809546e-01
-1.40247273e+00 7.77452528e-01 8.60965967e-01 -6.78088129e-01
-1.01119232e+00 5.73400319e-01 2.65009522e-01 5.02079800e-02
3.60024452e-01 -8.06797683e-01 -3.78767967e-01 -6.55393153e-02
6.28360152e-01 9.21205804e-02 3.14341694e-01 -3.20794761e-01
-1.21723369e-01 6.63416445e-01 -1.32832140e-01 2.57651180e-01
7.98603952e-01 -1.89203307e-01 3.69017988e-01 1.98679194e-01
7.52584517e-01 6.39224470e-01 -1.22490025e+00 -8.70433152e-02
-1.83473602e-01 -7.55082726e-01 -7.62319267e-02 -9.37438786e-01
-7.34604776e-01 8.24605227e-01 1.01409948e+00 4.92162317e-01
1.13040900e+00 -3.41983885e-01 5.67433834e-01 -2.20429584e-01
2.41743773e-01 -6.22125745e-01 -1.49199128e-01 -9.81118828e-02
5.45388281e-01 -1.47284055e+00 2.82560233e-02 -5.93308866e-01
-5.06325006e-01 1.10723341e+00 1.93962201e-01 -2.89986938e-01
4.38404381e-01 4.14489247e-02 1.28194451e-01 -2.85501361e-01
-1.71634808e-01 7.23962486e-02 1.30314112e-01 5.28572142e-01
4.11265105e-01 6.53023571e-02 -5.14344156e-01 -1.29005730e-01
2.46506855e-01 1.06951803e-01 6.33504808e-01 1.21762133e+00
-6.96696043e-01 -6.49804294e-01 -6.95175588e-01 2.75849462e-01
-8.84849906e-01 -1.01242051e-03 3.00339013e-01 1.05032229e+00
2.99234360e-01 9.38925385e-01 -8.07121098e-02 3.21178317e-01
7.13518202e-01 -4.82740223e-01 4.33056206e-01 -7.06935108e-01
-1.00513041e+00 5.80183148e-01 1.70569599e-01 -3.61519396e-01
-3.90748918e-01 -6.41499996e-01 -1.17069995e+00 3.54053676e-01
-3.85271102e-01 3.29906970e-01 8.39336038e-01 4.59991038e-01
-8.85599777e-02 6.00173771e-01 4.20461416e-01 -6.62890196e-01
-5.83991766e-01 -5.77662051e-01 -2.42946997e-01 4.76620585e-01
1.01433980e+00 -3.61667991e-01 -6.34148300e-01 1.76430553e-01]
|
[14.024516105651855, -3.092751979827881]
|
ef3c98d7-bf9c-4769-8903-39dc06e164a5
|
harnessing-mixed-offline-reinforcement
|
2306.13085
| null |
https://arxiv.org/abs/2306.13085v1
|
https://arxiv.org/pdf/2306.13085v1.pdf
|
Harnessing Mixed Offline Reinforcement Learning Datasets via Trajectory Weighting
|
Most offline reinforcement learning (RL) algorithms return a target policy maximizing a trade-off between (1) the expected performance gain over the behavior policy that collected the dataset, and (2) the risk stemming from the out-of-distribution-ness of the induced state-action occupancy. It follows that the performance of the target policy is strongly related to the performance of the behavior policy and, thus, the trajectory return distribution of the dataset. We show that in mixed datasets consisting of mostly low-return trajectories and minor high-return trajectories, state-of-the-art offline RL algorithms are overly restrained by low-return trajectories and fail to exploit high-performing trajectories to the fullest. To overcome this issue, we show that, in deterministic MDPs with stochastic initial states, the dataset sampling can be re-weighted to induce an artificial dataset whose behavior policy has a higher return. This re-weighted sampling strategy may be combined with any offline RL algorithm. We further analyze that the opportunity for performance improvement over the behavior policy correlates with the positive-sided variance of the returns of the trajectories in the dataset. We empirically show that while CQL, IQL, and TD3+BC achieve only a part of this potential policy improvement, these same algorithms combined with our reweighted sampling strategy fully exploit the dataset. Furthermore, we empirically demonstrate that, despite its theoretical limitation, the approach may still be efficient in stochastic environments. The code is available at https://github.com/Improbable-AI/harness-offline-rl.
|
['Romain Laroche', 'Rémi Tachet des Combes', 'Pulkit Agrawal', 'Zhang-Wei Hong']
|
2023-06-22
| null | null | null | null |
['offline-rl']
|
['playing-games']
|
[-1.04792669e-01 1.94283575e-01 -4.68938053e-01 4.97831404e-02
-9.15366948e-01 -8.84654343e-01 6.79623246e-01 2.44284913e-01
-7.71606147e-01 1.05265284e+00 1.64132893e-01 -5.39405942e-01
-3.73176634e-01 -9.19783413e-01 -9.18621063e-01 -9.18551624e-01
-4.02262628e-01 5.31342924e-01 2.90418088e-01 -1.08571835e-02
2.24392340e-01 5.75222671e-01 -1.67133331e+00 -6.20021932e-02
7.83342957e-01 6.53748989e-01 1.40295222e-01 6.89180017e-01
2.06999704e-01 8.05873275e-01 -6.24400258e-01 4.68258187e-02
4.92059857e-01 -5.58681428e-01 -5.00830412e-01 2.72923638e-03
-9.49982777e-02 -5.31154513e-01 -4.57705408e-01 8.35432053e-01
3.43855321e-01 4.53878790e-01 4.12535429e-01 -1.36522102e+00
1.65955395e-01 6.76883042e-01 -4.81110394e-01 1.95126683e-01
2.50061333e-01 7.19789207e-01 8.97369266e-01 -1.62653610e-01
5.42240977e-01 9.18406665e-01 2.62878031e-01 4.41435814e-01
-1.46209359e+00 -6.06156886e-01 3.72596771e-01 -1.47180825e-01
-1.08250022e+00 -2.74930179e-01 4.21559721e-01 -1.81285173e-01
8.11926246e-01 1.92075223e-01 8.22945774e-01 9.55502152e-01
1.52243376e-01 9.13344860e-01 1.17407072e+00 -1.34015560e-01
6.38007760e-01 1.64307296e-01 -2.07285732e-01 4.31051821e-01
4.35214192e-01 6.74289167e-01 -1.52116492e-01 -4.24500376e-01
4.27371413e-01 1.96554568e-02 -1.40802190e-01 -6.20877922e-01
-8.76367748e-01 6.88650727e-01 6.66374192e-02 1.72915831e-01
-6.93677783e-01 5.39043725e-01 2.45161921e-01 2.81284720e-01
1.37232333e-01 7.74388134e-01 -4.94823277e-01 -6.55436218e-01
-9.84174907e-01 7.76066422e-01 7.73953259e-01 7.36923516e-01
8.63733470e-01 1.15352415e-01 -4.51401412e-01 2.54153818e-01
-2.63720900e-01 6.41603172e-01 3.11156094e-01 -1.15335572e+00
4.90262508e-01 3.84633422e-01 6.10965490e-01 -4.98856634e-01
-3.41053724e-01 -4.19843823e-01 -6.25059232e-02 3.64879131e-01
8.27261090e-01 -5.21253467e-01 -5.92947006e-01 2.12098551e+00
2.94543624e-01 7.39490241e-02 3.29502039e-02 6.71323240e-01
-4.90404218e-01 6.37733638e-01 7.90165216e-02 -4.64406192e-01
6.80512607e-01 -4.67474163e-01 -4.46391761e-01 -4.94337678e-02
7.43883014e-01 -3.67013216e-01 1.20249426e+00 2.43427187e-01
-1.11947596e+00 -1.13901041e-01 -8.12150002e-01 8.15023243e-01
1.63795948e-02 -1.69141293e-01 3.63133281e-01 6.14423990e-01
-6.76192760e-01 9.46957946e-01 -9.90076184e-01 -1.99937791e-01
4.76325095e-01 3.22361231e-01 1.60813764e-01 1.11940347e-01
-8.88336957e-01 8.74217629e-01 4.55928564e-01 -5.65065205e-01
-1.34608626e+00 -9.59991455e-01 -2.19679266e-01 1.54872507e-01
9.95481491e-01 -2.02836305e-01 1.41648185e+00 -9.57002282e-01
-1.43437481e+00 1.56552717e-01 6.80849329e-02 -7.64056087e-01
9.17906404e-01 -1.10612080e-01 -6.57122284e-02 5.68229780e-02
-3.38328592e-02 2.20293134e-01 7.04231322e-01 -1.28729463e+00
-9.13001835e-01 -2.87223637e-01 2.73288548e-01 2.53362626e-01
-1.89387292e-01 -5.14684021e-01 -6.91735605e-03 -3.20843101e-01
-6.71123862e-01 -1.26415193e+00 -4.78608668e-01 -6.16310954e-01
-2.41865396e-01 -2.67768446e-02 5.94224811e-01 -2.15454891e-01
1.44179046e+00 -2.06138349e+00 -6.78621233e-02 4.75066721e-01
-2.36603111e-01 2.15496331e-01 -2.50904053e-01 8.42835784e-01
2.83358246e-01 6.25113547e-02 -1.82039529e-01 7.80920014e-02
-7.54581299e-03 2.45122239e-01 -5.39472461e-01 7.07053900e-01
1.55228572e-02 6.16397262e-01 -1.26966155e+00 7.66226277e-02
1.83005258e-01 -1.51364759e-01 -7.11146176e-01 3.03225577e-01
-5.52975893e-01 5.00988126e-01 -5.40909290e-01 2.52433240e-01
4.52638447e-01 2.94035077e-02 4.88057524e-01 2.80413836e-01
-2.81853348e-01 2.42956728e-01 -1.21061516e+00 1.16220248e+00
-3.99246603e-01 2.60674566e-01 -8.59244317e-02 -8.51968586e-01
5.13890743e-01 2.25527771e-02 8.75855088e-01 -8.01299691e-01
-7.51281753e-02 3.29552174e-01 2.28122473e-01 -4.03376877e-01
6.33088946e-01 -2.50918679e-02 -6.42140284e-02 8.83330941e-01
-2.99161375e-01 -3.44724357e-02 2.76940048e-01 1.09639861e-01
1.33801436e+00 3.99124384e-01 1.49182066e-01 -4.37988281e-01
-8.03963318e-02 2.74984866e-01 4.50658500e-01 1.05763078e+00
-2.72147715e-01 -4.31602728e-03 9.78553772e-01 -5.92541695e-02
-1.23040557e+00 -1.18184650e+00 1.22364841e-01 9.72237587e-01
9.72785354e-02 -1.47387981e-01 -6.42098248e-01 -9.37926173e-01
4.46486294e-01 1.25344872e+00 -6.00536227e-01 -4.24852550e-01
-5.74884534e-01 -5.73075414e-01 5.78624904e-01 2.66191959e-01
1.52966499e-01 -1.02293098e+00 -1.22988808e+00 3.78257781e-01
1.57226697e-01 -7.60203183e-01 -4.93236780e-01 2.39046380e-01
-8.88881505e-01 -1.19986761e+00 -4.11441326e-01 1.62471265e-01
7.30257630e-01 1.52133793e-01 8.75920892e-01 -2.09981352e-01
6.68725967e-02 7.12545753e-01 -3.05110961e-01 -4.58737284e-01
-5.52843213e-01 4.64295316e-03 2.11301714e-01 -1.41481996e-01
1.34020507e-01 -5.67090333e-01 -7.52053201e-01 1.07997417e-01
-9.72554386e-01 -4.57288563e-01 2.34491974e-01 7.42862403e-01
4.31005001e-01 2.14655966e-01 8.13166320e-01 -6.74751222e-01
8.47074389e-01 -5.81555605e-01 -1.00845015e+00 2.18732402e-01
-7.70526528e-01 5.01236022e-01 7.85330892e-01 -7.49242544e-01
-8.10305297e-01 -3.63192568e-03 1.65182739e-01 -5.12115836e-01
-2.64244378e-02 2.56466478e-01 4.97956835e-02 4.39444929e-01
5.96976101e-01 4.02428001e-01 3.15144628e-01 -2.04691082e-01
4.37378019e-01 3.83331269e-01 6.51805550e-02 -1.07466912e+00
6.64994657e-01 4.50443655e-01 4.56535630e-02 -6.03400588e-01
-6.02140546e-01 -1.03158608e-01 -1.48018356e-02 -4.29314226e-01
4.96604264e-01 -6.32111907e-01 -9.93342042e-01 1.73701808e-01
-4.56068903e-01 -9.67265725e-01 -9.11458015e-01 4.99035895e-01
-1.15702367e+00 7.58306831e-02 -6.52995929e-02 -1.38277316e+00
1.48639753e-01 -1.27416539e+00 4.91005421e-01 1.57552391e-01
-1.10170439e-01 -6.31171227e-01 1.75083399e-01 -2.03005597e-01
4.06149149e-01 2.06251144e-01 8.94859970e-01 -9.24854279e-01
-6.78664327e-01 -9.37712044e-02 2.11952925e-01 2.31741473e-01
1.19326130e-01 -4.31707129e-02 -7.01273859e-01 -6.78983510e-01
-2.33518988e-01 -3.87167305e-01 6.32356822e-01 4.77128029e-01
1.13725257e+00 -6.31881893e-01 -1.54020756e-01 1.42649800e-01
1.55179155e+00 5.05745947e-01 5.13370156e-01 5.24267197e-01
2.42952541e-01 6.53364837e-01 9.56847966e-01 9.64592576e-01
1.78892180e-01 6.82651401e-01 5.40345848e-01 2.74522156e-01
3.42616796e-01 -7.04375863e-01 7.21063495e-01 -3.73554006e-02
6.96865171e-02 -9.60114673e-02 -8.05522561e-01 6.96239293e-01
-2.02369094e+00 -1.30998302e+00 3.76669973e-01 2.93677473e+00
8.93644631e-01 3.11441213e-01 8.90164375e-01 -7.56642222e-02
3.71670187e-01 9.78670344e-02 -1.07460368e+00 -5.38455784e-01
1.97847843e-01 7.75322095e-02 1.00740874e+00 4.10691738e-01
-6.38015568e-01 7.98410535e-01 6.43583822e+00 1.05458391e+00
-9.90414619e-01 -3.09514049e-02 6.00894094e-01 -6.54861569e-01
-4.50810760e-01 8.47530365e-02 -7.96476007e-01 8.48703384e-01
1.37911952e+00 -4.98520195e-01 9.50199902e-01 8.35862458e-01
4.55741584e-01 -3.53243202e-01 -1.09553933e+00 2.83366740e-01
-5.77334821e-01 -1.21592402e+00 -2.58784235e-01 5.42737007e-01
7.58601487e-01 1.08657502e-01 6.65810034e-02 6.51006699e-01
6.99092865e-01 -7.87178934e-01 8.26014996e-01 4.41637427e-01
5.76415062e-01 -1.23479724e+00 3.30245823e-01 6.72874331e-01
-8.77141356e-01 -5.99962652e-01 -3.82482968e-02 4.77386117e-02
-5.88010550e-02 4.87617910e-01 -6.49735034e-01 2.25174710e-01
3.90968978e-01 1.06857575e-01 -1.29222944e-01 8.54917407e-01
-4.10485677e-02 7.95824230e-01 -5.05360901e-01 -3.79681647e-01
4.87284839e-01 -3.60665351e-01 7.63714910e-01 9.27973151e-01
2.91685104e-01 -1.45796090e-01 4.03361738e-01 8.22110951e-01
2.24007651e-01 -1.66561991e-01 -7.43569732e-01 -4.07306671e-01
7.19787240e-01 7.57283747e-01 -5.00789225e-01 -3.64727676e-01
-8.37324485e-02 3.93891752e-01 3.90813917e-01 4.92542565e-01
-1.01136935e+00 -1.48775548e-01 9.43691373e-01 2.04374045e-01
4.10804331e-01 -1.15823291e-01 -8.49017873e-02 -8.27646852e-01
-3.54648083e-02 -1.07572818e+00 3.05808604e-01 -1.29307598e-01
-1.00863898e+00 1.24220148e-01 3.04424703e-01 -1.27911282e+00
-5.33633828e-01 -3.74553874e-02 -3.78164202e-01 6.49999738e-01
-1.25259614e+00 -3.64591271e-01 4.10243511e-01 3.97115231e-01
4.54036146e-01 -4.60914597e-02 3.56469095e-01 9.11521900e-04
-5.51865935e-01 6.14725947e-01 5.30341446e-01 -4.06402528e-01
3.83152753e-01 -1.25388587e+00 -6.49051964e-02 7.27616608e-01
-1.37485161e-01 3.18347394e-01 1.04765844e+00 -7.11029351e-01
-1.62216830e+00 -1.11160278e+00 1.19345471e-01 -2.66098350e-01
7.83440411e-01 -3.77708599e-02 -6.48270726e-01 5.29935479e-01
-5.36054485e-02 -2.97450244e-01 2.47850031e-01 4.69917022e-02
-2.70404574e-02 -1.98568553e-01 -1.33430314e+00 8.97553444e-01
8.62537742e-01 -2.46249348e-01 -1.92132309e-01 1.83656469e-01
6.42737567e-01 -1.60830364e-01 -6.98584676e-01 3.75984102e-01
5.48181236e-01 -9.52459514e-01 6.28043294e-01 -8.89693260e-01
3.43460709e-01 -1.14900991e-02 -3.24054420e-01 -1.51182032e+00
-2.79800389e-02 -6.39236152e-01 -4.73050267e-01 9.39397454e-01
4.08667952e-01 -7.00463772e-01 8.30940604e-01 5.77598691e-01
2.19645783e-01 -1.06450105e+00 -9.89868164e-01 -1.30444074e+00
3.37593317e-01 -4.00536835e-01 7.72225261e-01 4.58383262e-01
4.96563837e-02 -2.01589733e-01 -3.65498364e-01 6.97879791e-02
7.00245023e-01 2.79659599e-01 8.86651814e-01 -5.36450326e-01
-5.66137731e-01 -5.55816352e-01 1.63595036e-01 -9.52714264e-01
4.50088754e-02 -5.83455384e-01 1.52125269e-01 -1.19953763e+00
1.62035003e-01 -6.96724057e-01 -4.29463297e-01 4.05726194e-01
-1.71353549e-01 -5.41035712e-01 5.30173898e-01 1.64105445e-01
-6.48441911e-01 6.52534604e-01 1.15316033e+00 2.87254065e-01
-8.01721573e-01 2.35541552e-01 -4.32954699e-01 3.71915281e-01
1.05624318e+00 -7.21107602e-01 -6.60470188e-01 1.44430012e-01
6.84251040e-02 4.73153174e-01 1.54376239e-01 -9.28376019e-01
-1.22857071e-01 -7.91290104e-01 -8.26394111e-02 -5.89295983e-01
1.20858885e-01 -8.36012840e-01 1.48352653e-01 9.32048619e-01
-6.34836316e-01 5.20421043e-02 2.58665860e-01 8.67697358e-01
3.96379918e-01 -2.08209962e-01 7.90199637e-01 -1.67848080e-01
-1.75140426e-01 3.05101544e-01 -7.84463048e-01 3.48410308e-01
1.31113315e+00 -7.62148127e-02 -2.92417347e-01 -5.00004113e-01
-2.47768044e-01 4.44399923e-01 5.51971197e-01 1.91864088e-01
2.16372654e-01 -1.05736232e+00 -4.39075947e-01 9.06425621e-03
-9.49731022e-02 -3.69697332e-01 1.48266748e-01 7.97274470e-01
5.54295629e-03 2.34917402e-01 -2.28346825e-01 -3.27099800e-01
-7.72522449e-01 6.64986670e-01 4.73515362e-01 -5.94443858e-01
-5.62857151e-01 2.08853215e-01 -9.56066772e-02 -3.35416377e-01
3.16392004e-01 -7.54204392e-02 4.20718819e-01 1.00389376e-01
4.08225477e-01 6.91347063e-01 -2.37077266e-01 -9.35603082e-02
-3.36184323e-01 -4.84918877e-02 8.79621059e-02 -4.81359661e-01
1.23816741e+00 1.31022915e-01 3.84643227e-01 3.74431849e-01
8.57229233e-01 2.15686440e-01 -1.87266982e+00 -3.13741937e-02
1.08544841e-01 -5.68292916e-01 -4.23229858e-02 -8.33096862e-01
-9.21164691e-01 3.87502372e-01 5.21854043e-01 4.15058464e-01
1.03385484e+00 -3.55908632e-01 6.85360074e-01 3.05459023e-01
7.65480101e-01 -1.32919586e+00 6.98967054e-02 3.36756855e-01
5.42440295e-01 -8.48313928e-01 7.36936182e-02 4.48116869e-01
-9.65264499e-01 7.33645260e-01 5.39649129e-01 -4.07244861e-01
4.11380589e-01 3.56893808e-01 -3.05388123e-01 8.62528682e-02
-1.09163761e+00 -5.27114272e-01 -3.68002951e-01 5.20129383e-01
-1.72312662e-01 3.83531511e-01 -3.49349201e-01 2.47778982e-01
4.05805707e-02 8.20590705e-02 7.24359393e-01 9.88652229e-01
-6.35048568e-01 -1.08863223e+00 -2.44528726e-01 6.38714612e-01
-3.99285406e-01 2.45797157e-01 -7.64895454e-02 9.18278337e-01
-2.33131930e-01 8.87363732e-01 2.28044659e-01 -3.24759036e-01
4.42329347e-01 1.62690617e-02 5.06587207e-01 -2.60935605e-01
-6.94743752e-01 9.61032659e-02 2.68098056e-01 -9.21175778e-01
5.12608178e-02 -8.05150628e-01 -1.41221559e+00 -5.54341912e-01
-2.57072840e-02 1.62247255e-01 5.25806785e-01 8.81883502e-01
5.40931344e-01 4.92388964e-01 9.36307609e-01 -6.30882561e-01
-1.36894929e+00 -5.82353175e-01 -7.17832625e-01 3.74836206e-01
3.62448961e-01 -7.57962883e-01 -5.15278339e-01 -6.62382722e-01]
|
[4.132997035980225, 2.3799033164978027]
|
aaa6dafa-a06d-44f0-904a-322eb503eb4e
|
compiling-a-highly-accurate-bilingual-lexicon
| null | null |
https://aclanthology.org/2022.gwll-1.6
|
https://aclanthology.org/2022.gwll-1.6.pdf
|
Compiling a Highly Accurate Bilingual Lexicon by Combining Different Approaches
|
Bilingual lexicons can be generated automatically using a wide variety of approaches. We perform a rigorous manual evaluation of four different methods: word alignments on different types of bilingual data, pivoting, machine translation and cross-lingual word embeddings. We investigate how the different setups perform using publicly available data for the English-Icelandic language pair, doing separate evaluations for each method, dataset and confidence class where it can be calculated. The results are validated by human experts, working with a random sample from all our experiments. By combining the most promising approaches and data sets, using confidence scores calculated from the data and the results of manually evaluating samples from our manual evaluation as indicators, we are able to induce lists of translations with a very high acceptance rate. We show how multiple different combinations generate lists with well over 90% acceptance rate, substantially exceeding the results for each individual approach, while still generating reasonably large candidate lists. All manually evaluated equivalence pairs are published in a new lexicon of over 232,000 pairs under an open license.
|
['Andy Way', 'Hrafn Loftsson', 'Finnur Ingimundarson', 'Luke O’Brien', 'Steinþór Steingrímsson']
| null | null | null | null |
gwll-lrec-2022-6
|
['cross-lingual-word-embeddings']
|
['natural-language-processing']
|
[-4.63998318e-02 4.50885780e-02 -3.08094472e-01 -4.09004152e-01
-1.38694870e+00 -1.18666875e+00 1.07225215e+00 3.50167155e-01
-8.25388610e-01 1.06524205e+00 5.79764187e-01 -4.86052126e-01
9.14643332e-02 -4.89304394e-01 -6.67376161e-01 -3.78783733e-01
3.44004095e-01 1.26980603e+00 1.61169425e-01 -5.69300830e-01
2.55813897e-01 2.52829462e-01 -1.37719953e+00 1.21757559e-01
1.03458738e+00 3.74603629e-01 -2.75376290e-02 3.55982006e-01
5.24626002e-02 -1.89312965e-01 -5.03506958e-01 -1.06216705e+00
5.69918990e-01 -3.69258165e-01 -7.15203941e-01 -3.19952458e-01
5.15652180e-01 1.83030292e-01 2.80786932e-01 1.08998966e+00
6.94134831e-01 -4.27812427e-01 7.36712933e-01 -6.90063715e-01
-6.65504217e-01 8.09423029e-01 -8.67684409e-02 1.19325712e-01
8.31124544e-01 2.84359425e-01 1.39986074e+00 -1.20812690e+00
1.05343020e+00 9.01383996e-01 7.44074404e-01 6.58672526e-02
-1.45995176e+00 -6.85117722e-01 -4.00515765e-01 -1.23477422e-01
-1.51855743e+00 -5.51053703e-01 2.00984165e-01 -4.39100921e-01
1.25063014e+00 1.47981629e-01 7.12532163e-01 1.07706630e+00
1.95125282e-01 2.08552301e-01 1.59552062e+00 -8.71098459e-01
-1.62239932e-02 7.71019399e-01 -7.14871883e-02 3.66792947e-01
6.15329504e-01 3.04576933e-01 -4.72873628e-01 -3.67177188e-01
2.48751089e-01 -7.67949522e-01 -2.18775049e-01 -3.74329895e-01
-1.85120881e+00 8.11048746e-01 -4.60017771e-02 6.15628958e-01
-1.43706903e-01 -5.36568105e-01 4.56597894e-01 5.72434723e-01
4.74262744e-01 6.90962851e-01 -8.73027146e-01 -6.00896813e-02
-1.00119150e+00 4.44238842e-01 1.03933227e+00 1.10655439e+00
8.31237137e-01 -3.94676059e-01 2.12795064e-02 9.62966919e-01
4.92600426e-02 7.08305180e-01 7.83877075e-01 -2.76479572e-01
7.70217359e-01 5.68866730e-01 3.97443235e-01 -8.66830707e-01
-2.96335846e-01 -1.27730370e-01 -2.36554757e-01 -1.61520541e-02
6.20428085e-01 -3.25601280e-01 -6.30220175e-01 1.78619003e+00
2.34814912e-01 -5.70514619e-01 1.72497645e-01 6.35175347e-01
5.50603032e-01 5.29275894e-01 -1.76570117e-01 -3.20075870e-01
1.47137153e+00 -8.29519689e-01 -5.77114940e-01 -1.76069155e-01
8.42454553e-01 -1.44658780e+00 1.30006206e+00 3.94365281e-01
-8.09272945e-01 -5.91584206e-01 -1.30693996e+00 3.59658971e-02
-5.85940719e-01 3.67658556e-01 3.76174271e-01 7.93967724e-01
-1.37424457e+00 4.63043213e-01 -5.50958157e-01 -6.26601040e-01
-2.60608941e-01 6.31032825e-01 -7.70173252e-01 9.51520056e-02
-1.36571026e+00 1.56427121e+00 5.17290592e-01 -2.02648029e-01
-5.15455127e-01 -2.93037742e-01 -8.41807127e-01 -4.79547411e-01
-9.00681540e-02 -3.77390027e-01 9.39582944e-01 -8.91449928e-01
-1.34986448e+00 1.50350964e+00 -1.44149177e-02 -2.78583467e-01
6.17017984e-01 -8.22158679e-02 -7.62900829e-01 -3.68338853e-01
4.49929267e-01 6.95239902e-01 1.84532791e-01 -1.03790855e+00
-5.46872556e-01 -1.84311762e-01 -2.45439157e-01 2.12420702e-01
-2.17247784e-01 4.69043761e-01 -3.70972186e-01 -7.21331477e-01
-1.34070799e-01 -1.23002470e+00 -2.38436714e-01 -6.83617055e-01
-2.32020870e-01 -1.60443529e-01 -2.22696871e-01 -9.20118093e-01
1.11773837e+00 -1.72951174e+00 1.95174694e-01 2.29479507e-01
-2.48557404e-01 2.14874685e-01 -3.82176816e-01 7.32255399e-01
-1.66259438e-01 1.02717414e-01 -2.99014866e-01 -9.39913616e-02
1.72721788e-01 1.10912479e-01 -9.02277082e-02 5.22598922e-01
2.88868010e-01 8.63767445e-01 -1.04549432e+00 -4.52647597e-01
9.68014374e-02 2.51468867e-01 -5.18854976e-01 -3.48293707e-02
1.14579327e-01 2.17234224e-01 9.73586291e-02 4.93601769e-01
4.96878505e-01 3.74612033e-01 7.72024274e-01 -1.23582833e-01
-1.66684806e-01 6.85808420e-01 -1.14782488e+00 1.66856229e+00
-6.35671556e-01 6.86540961e-01 -3.41016769e-01 -6.09078228e-01
1.19469702e+00 4.87044036e-01 -7.06482828e-02 -6.45982385e-01
1.97800279e-01 1.02036357e+00 4.10111815e-01 -2.02980354e-01
6.81167960e-01 -3.37936103e-01 -3.76473308e-01 6.51005924e-01
5.97887814e-01 -4.22608256e-01 5.32845914e-01 -5.47501184e-02
7.53054142e-01 2.51761049e-01 5.63064992e-01 -9.43701565e-01
5.19478858e-01 5.02832592e-01 4.13319170e-01 3.85677010e-01
-9.25984830e-02 6.30610943e-01 3.61092061e-01 -5.83391964e-01
-1.54820681e+00 -1.11680830e+00 -4.92887050e-01 8.50387573e-01
1.01541065e-01 -6.51229918e-01 -7.42501676e-01 -6.36638999e-01
-2.46475562e-01 6.91097081e-01 -5.19673645e-01 5.73983118e-02
-6.67866409e-01 -1.11511707e+00 5.96127868e-01 1.67939723e-01
-2.86408812e-01 -1.19223499e+00 -1.54842466e-01 2.00316116e-01
-3.82559896e-01 -1.14951241e+00 -3.90065134e-01 1.61863461e-01
-5.99751770e-01 -9.96138275e-01 -5.49811900e-01 -1.13124597e+00
5.52990079e-01 -2.20994860e-01 1.68292117e+00 -3.10288876e-01
-7.70500824e-02 -1.40435994e-01 -3.55143309e-01 -1.63912460e-01
-9.37268853e-01 3.17993313e-01 5.89773118e-01 -4.19655144e-01
8.96621287e-01 -4.83023942e-01 -1.71347961e-01 3.54165941e-01
-6.16965473e-01 -2.14025602e-01 6.80050790e-01 9.08507466e-01
4.23828274e-01 -4.76882130e-01 3.95118445e-01 -8.19949090e-01
9.24835324e-01 -3.05323035e-01 -7.05895126e-01 4.19941932e-01
-7.62756884e-01 3.39973629e-01 5.75741947e-01 -3.77494782e-01
-5.42137504e-01 -4.56118695e-02 -3.57329905e-01 3.08721870e-01
-1.51414335e-01 2.88264483e-01 -1.19568996e-01 7.06724674e-02
9.49243069e-01 2.20656954e-02 -1.16286166e-01 -3.85858536e-01
6.71587348e-01 8.30038846e-01 3.75201583e-01 -8.00785065e-01
9.08072412e-01 -3.05880345e-02 -6.89298630e-01 -2.83426017e-01
-4.31472510e-01 -2.83260733e-01 -9.96199429e-01 1.30578682e-01
6.29083872e-01 -1.01259947e+00 3.40913460e-02 1.14321791e-01
-1.17367327e+00 -6.89731091e-02 -1.42045990e-01 8.53535295e-01
-5.14239609e-01 1.10101186e-01 -3.37884873e-01 -2.88079292e-01
-3.10012519e-01 -1.55870390e+00 1.17610717e+00 -2.03573912e-01
-8.67808342e-01 -1.14983320e+00 7.34200895e-01 2.30468899e-01
1.67764321e-01 1.26540974e-01 8.80591333e-01 -1.15284181e+00
-4.04951312e-02 -3.66383433e-01 4.84434962e-02 2.76620001e-01
3.47961754e-01 1.27765685e-01 -7.74006784e-01 -3.85428876e-01
-2.65411109e-01 -5.04898787e-01 5.18796921e-01 -1.41274661e-01
1.57327075e-02 -3.80010344e-02 -2.43074521e-01 2.69629061e-01
1.64344013e+00 -1.24732750e-02 5.33732116e-01 5.40162802e-01
3.35925400e-01 6.60441995e-01 6.64031267e-01 -1.35646135e-01
2.86538243e-01 9.82013226e-01 -2.50063986e-01 9.82074812e-03
-9.59525350e-05 -1.81721225e-01 6.74652219e-01 1.51733601e+00
-1.59116775e-01 -1.34264417e-02 -1.19975173e+00 9.59706306e-01
-1.55014110e+00 -6.62717164e-01 -8.25681835e-02 2.57467103e+00
1.36692238e+00 3.08856159e-01 2.01112151e-01 1.73054654e-02
7.14153349e-01 -6.62178993e-02 1.82000577e-01 -7.63744473e-01
-3.35335046e-01 7.17740536e-01 6.54170692e-01 9.00488496e-01
-9.57619488e-01 1.23702860e+00 7.70076561e+00 8.37784767e-01
-8.03974926e-01 1.05987191e-01 3.56739104e-01 9.42917168e-02
-6.97618544e-01 2.54414171e-01 -9.51330125e-01 3.26107472e-01
1.14524937e+00 -4.26347524e-01 4.49775130e-01 3.93548250e-01
-9.65042636e-02 -4.70607392e-02 -1.18813419e+00 8.18328321e-01
2.61130273e-01 -1.20392394e+00 6.26803860e-02 -6.60857260e-02
1.14956975e+00 4.78624284e-01 -3.43102336e-01 2.34959364e-01
7.25222945e-01 -9.06641126e-01 7.74557054e-01 1.34383231e-01
1.07962644e+00 -6.57263935e-01 9.40811813e-01 1.40135765e-01
-8.63491058e-01 3.35309088e-01 -2.75996506e-01 -8.71111080e-02
2.30738476e-01 7.49821484e-01 -8.27927768e-01 7.41350472e-01
4.34775740e-01 4.81376529e-01 -8.25396240e-01 6.91617191e-01
-4.18731034e-01 4.56023574e-01 -4.77944851e-01 -2.14589179e-01
2.21169308e-01 -5.42832136e-01 4.71965909e-01 1.49911618e+00
3.91971380e-01 -5.94432890e-01 1.80309359e-02 5.22409499e-01
2.00000200e-02 6.87175214e-01 -8.97844672e-01 8.81040748e-03
4.97587919e-01 1.24954391e+00 -7.23464429e-01 -4.04868007e-01
-3.59622419e-01 7.75255561e-01 5.32402217e-01 4.18452397e-02
-8.52807403e-01 -5.22429883e-01 5.94528913e-01 -4.53211553e-02
4.57676239e-02 -2.47160390e-01 -2.21344471e-01 -1.33745396e+00
3.42186034e-01 -1.34047472e+00 9.35349390e-02 -3.81269217e-01
-1.25485229e+00 1.09040630e+00 1.26626089e-01 -1.40419114e+00
-4.79632109e-01 -9.29390848e-01 -2.26069003e-01 1.20164633e+00
-9.98339832e-01 -7.65055776e-01 2.54690409e-01 1.79142319e-02
3.02457064e-01 -4.55525577e-01 1.15517926e+00 5.96273065e-01
-2.24534854e-01 7.24452376e-01 2.43937582e-01 1.28340915e-01
1.16863406e+00 -1.20768714e+00 7.61061668e-01 7.01624572e-01
7.13079691e-01 8.48247886e-01 7.64169574e-01 -4.93384391e-01
-8.20122480e-01 -7.00900912e-01 1.65868139e+00 -8.97985220e-01
1.05885005e+00 -5.91587245e-01 -5.21353960e-01 6.51088238e-01
6.81626022e-01 -4.39236552e-01 8.96584988e-01 3.39229375e-01
-3.25020254e-01 1.00583121e-01 -9.40826833e-01 6.92993701e-01
1.07786989e+00 -5.33310533e-01 -1.00621724e+00 6.70454443e-01
4.88780916e-01 -2.89408207e-01 -8.51750612e-01 5.06048143e-01
6.49115384e-01 -9.23198879e-01 6.61469996e-01 -6.64408505e-01
5.05591691e-01 -4.32904869e-01 -3.14215809e-01 -1.68382084e+00
-7.71633163e-02 -5.06032169e-01 7.60223091e-01 1.43006229e+00
1.13745296e+00 -8.10271204e-01 1.95850357e-01 3.33063714e-02
8.31094757e-02 -7.02748775e-01 -9.76694524e-01 -7.36976564e-01
4.02421743e-01 -3.03334683e-01 6.09221220e-01 1.08826447e+00
2.06595078e-01 6.57572448e-01 -2.31160700e-01 -2.01869965e-01
3.07052016e-01 7.00632408e-02 9.03595805e-01 -9.93452072e-01
-2.86158860e-01 -5.20655155e-01 -5.27342319e-01 -6.12270117e-01
1.60004959e-01 -1.29238319e+00 1.57080725e-01 -1.29788005e+00
1.89596280e-01 -5.33233404e-01 -5.70678003e-02 3.04100126e-01
-2.81631172e-01 6.94187105e-01 -1.18035838e-01 1.73697025e-01
-1.59520850e-01 2.57131338e-01 8.63214612e-01 -9.71874446e-02
-5.11244945e-02 -4.40110952e-01 -6.75363958e-01 6.48500621e-01
7.45510995e-01 -6.95979774e-01 -1.66356325e-01 -5.62270105e-01
5.48549831e-01 -4.17246789e-01 -2.06238002e-01 -9.25477207e-01
-1.98971599e-01 4.78453860e-02 1.53159872e-01 -2.67262846e-01
-4.52780444e-03 -5.06561399e-01 3.95631075e-01 3.10272485e-01
-2.20749706e-01 8.08705807e-01 3.10293883e-01 -3.11948527e-02
-3.37271512e-01 -1.59266874e-01 6.71338379e-01 -9.26870853e-02
-4.56135631e-01 -1.76498860e-01 -1.92538142e-01 3.31412703e-01
8.81874442e-01 -5.52360304e-02 2.41722371e-02 -1.02615558e-01
-6.64822757e-01 6.37677917e-03 7.89285183e-01 7.07307220e-01
-1.31995734e-02 -1.63835788e+00 -1.03313780e+00 3.56813133e-01
5.42841256e-01 -7.60758102e-01 -7.06783593e-01 8.44331145e-01
-8.34416628e-01 3.99672657e-01 -3.82357687e-01 -4.42351490e-01
-8.17934334e-01 4.38155353e-01 4.13207151e-02 -4.53210294e-01
-2.02870607e-01 5.18735707e-01 -3.65093052e-01 -1.30004621e+00
-3.64872873e-01 -3.42513889e-01 -1.35999009e-01 2.17484042e-01
2.30023995e-01 -4.66246754e-02 5.03737152e-01 -1.04716384e+00
-5.60910702e-01 8.64521682e-01 4.11470495e-02 -6.43876970e-01
1.08035100e+00 1.69646338e-01 -2.55875766e-01 4.76406932e-01
1.04982078e+00 7.20433414e-01 -3.53980035e-01 2.67151184e-03
2.33637154e-01 -3.09973240e-01 -6.42405570e-01 -8.60212922e-01
-5.60997903e-01 5.50433397e-01 4.45905119e-01 2.91315224e-02
6.32421494e-01 -8.41842070e-02 4.85513240e-01 1.86764330e-01
7.34340072e-01 -1.16488099e+00 -6.12350881e-01 5.39649129e-01
7.36385763e-01 -1.36781287e+00 4.65444587e-02 -3.01762968e-01
-6.55833960e-01 9.72417355e-01 1.61478564e-01 -2.91297317e-01
4.33102876e-01 3.43468457e-01 5.95197022e-01 1.76272735e-01
-5.76113462e-01 -1.88832417e-01 2.87926137e-01 5.91210306e-01
8.28903437e-01 2.49688298e-01 -1.20934784e+00 4.61560577e-01
-8.47512305e-01 -3.40809256e-01 3.54903072e-01 4.25490141e-01
-6.19082525e-02 -1.81520283e+00 -3.21675837e-01 2.43272021e-01
-3.91620576e-01 -5.02603710e-01 -6.80663764e-01 1.03572607e+00
2.19542205e-01 8.51377845e-01 -8.23098347e-02 -5.55822551e-01
3.70143235e-01 1.92663208e-01 5.87708235e-01 -6.45231783e-01
-7.18455374e-01 -1.41703919e-01 5.89362264e-01 -1.65209189e-01
-4.41127181e-01 -7.36797333e-01 -7.58082986e-01 -1.58061370e-01
-4.61163640e-01 4.99708682e-01 5.29847383e-01 1.03839135e+00
1.41511306e-01 -1.88251600e-01 4.58886176e-01 -8.43925297e-01
-6.08999789e-01 -1.15590072e+00 -1.85393840e-01 6.58110559e-01
-2.04206944e-01 -5.23727059e-01 -4.05987591e-01 -1.42989410e-02]
|
[11.112781524658203, 10.133016586303711]
|
3a3a0495-d0ad-4cf0-9c5a-d7e6ebca0fdb
|
pagp-a-physics-assisted-gaussian-process
|
2204.02583
| null |
https://arxiv.org/abs/2204.02583v1
|
https://arxiv.org/pdf/2204.02583v1.pdf
|
PAGP: A physics-assisted Gaussian process framework with active learning for forward and inverse problems of partial differential equations
|
In this work, a Gaussian process regression(GPR) model incorporated with given physical information in partial differential equations(PDEs) is developed: physics-assisted Gaussian processes(PAGP). The targets of this model can be divided into two types of problem: finding solutions or discovering unknown coefficients of given PDEs with initial and boundary conditions. We introduce three different models: continuous time, discrete time and hybrid models. The given physical information is integrated into Gaussian process model through our designed GP loss functions. Three types of loss function are provided in this paper based on two different approaches to train the standard GP model. The first part of the paper introduces the continuous time model which treats temporal domain the same as spatial domain. The unknown coefficients in given PDEs can be jointly learned with GP hyper-parameters by minimizing the designed loss function. In the discrete time models, we first choose a time discretization scheme to discretize the temporal domain. Then the PAGP model is applied at each time step together with the scheme to approximate PDE solutions at given test points of final time. To discover unknown coefficients in this setting, observations at two specific time are needed and a mixed mean square error function is constructed to obtain the optimal coefficients. In the last part, a novel hybrid model combining the continuous and discrete time models is presented. It merges the flexibility of continuous time model and the accuracy of the discrete time model. The performance of choosing different models with different GP loss functions is also discussed. The effectiveness of the proposed PAGP methods is illustrated in our numerical section.
|
['Guang Lin', 'Shiqi Zhang', 'Jiahao Zhang']
|
2022-04-06
| null | null | null | null |
['gpr', 'gpr']
|
['computer-vision', 'miscellaneous']
|
[-1.98840424e-01 -1.65953469e-02 4.25126493e-01 9.57410634e-02
-8.72662067e-01 -2.17296958e-01 4.66853440e-01 1.28685802e-01
-2.99603552e-01 1.09093714e+00 -5.06452858e-01 8.54051858e-02
-6.23666704e-01 -8.08295488e-01 -6.02304816e-01 -1.29216921e+00
-1.48133785e-01 7.32449293e-01 3.02628309e-01 1.32465169e-01
3.20654541e-01 7.28015840e-01 -1.13712442e+00 -5.36702216e-01
1.14219856e+00 9.93495882e-01 2.95237213e-01 7.65403986e-01
-2.71462291e-01 4.34325933e-01 -2.46743858e-01 1.74976271e-02
3.90869468e-01 -2.46166557e-01 -2.38130808e-01 2.04885200e-01
-3.92175853e-01 1.26376554e-01 -2.64153004e-01 9.68800783e-01
7.33292341e-01 8.39350700e-01 1.13897133e+00 -1.40162826e+00
-3.82945359e-01 -1.33596018e-01 -6.36506915e-01 2.07837224e-02
-1.20248444e-01 4.07197833e-01 4.37555581e-01 -7.79258907e-01
3.19096059e-01 1.41452932e+00 8.91938329e-01 2.80152351e-01
-1.27144086e+00 -3.34644198e-01 2.05984384e-01 -1.60314336e-01
-1.57290089e+00 1.53850645e-01 8.48326564e-01 -8.00311565e-01
5.42096436e-01 -3.70200947e-02 3.74541402e-01 8.66008341e-01
8.17437649e-01 4.10350084e-01 1.27332532e+00 -8.43822956e-02
6.53144717e-01 2.73380786e-01 3.36860329e-01 5.80695987e-01
9.47045460e-02 1.34980962e-01 9.09725204e-02 -5.76055646e-01
1.03896999e+00 1.01553105e-01 -4.00716335e-01 -1.84303313e-01
-6.62729979e-01 9.75423515e-01 -7.22345989e-03 1.55828714e-01
-7.36790419e-01 -5.24307741e-03 7.84377530e-02 7.08122253e-02
7.30580509e-01 4.70722318e-01 -5.27041793e-01 -5.86046185e-03
-8.74775171e-01 6.33950233e-01 1.04750991e+00 1.10336244e+00
6.71364605e-01 6.13815002e-02 -5.70872903e-01 8.27401519e-01
5.87813258e-01 7.04658628e-01 8.00267160e-02 -9.48610127e-01
4.36748385e-01 -2.79646125e-02 5.57276070e-01 -9.50878978e-01
-2.71281123e-01 -4.43643987e-01 -9.07968163e-01 2.08849967e-01
4.36973512e-01 -6.66931272e-01 -9.30820942e-01 1.53612638e+00
6.33369327e-01 8.22519422e-01 1.72855154e-01 6.02091432e-01
3.32316756e-01 1.29325426e+00 1.01100065e-01 -5.68043470e-01
1.20428240e+00 -9.27886963e-01 -6.99427903e-01 3.60156029e-01
2.64643162e-01 -5.43801785e-01 6.96256876e-01 4.64528918e-01
-1.19606078e+00 -6.53052747e-01 -7.49896884e-01 3.54991615e-01
-2.22029611e-01 1.12929203e-01 1.36995852e-01 3.01858366e-01
-1.07220209e+00 1.11939240e+00 -1.16179740e+00 -2.69212633e-01
2.50165388e-02 2.73597389e-01 1.53483689e-01 1.91253155e-01
-1.15217149e+00 5.89057684e-01 -8.99988133e-03 3.07834357e-01
-8.57646406e-01 -1.06682193e+00 -5.79900503e-01 8.05431698e-03
1.12621494e-01 -8.74728441e-01 1.16592920e+00 -3.67184371e-01
-1.76142001e+00 4.07571286e-01 -2.48392224e-01 -2.70523757e-01
1.05548072e+00 -1.29148394e-01 -2.18995169e-01 1.57773584e-01
7.43053332e-02 1.42132372e-01 1.05939949e+00 -1.32787848e+00
-4.50385690e-01 -1.92124337e-01 -4.23525959e-01 1.58158496e-01
4.16945398e-01 -1.84959173e-01 -6.27731442e-01 -5.90413570e-01
1.80032670e-01 -9.33214843e-01 -6.29031122e-01 1.29562065e-01
-3.67710114e-01 -8.57840553e-02 9.85667467e-01 -9.03115630e-01
8.63916934e-01 -1.99786067e+00 3.12494725e-01 1.60567537e-01
9.01393443e-02 -5.97004555e-02 2.32675716e-01 6.17301285e-01
2.25560784e-01 5.78151569e-02 -8.39035988e-01 -7.51745403e-01
2.61943992e-02 3.97632658e-01 -3.91881526e-01 5.42329609e-01
4.87974018e-01 4.61422324e-01 -7.80591786e-01 -5.35716653e-01
2.62034565e-01 7.75490344e-01 -2.15087503e-01 3.10341775e-01
-2.15004086e-01 1.10721087e+00 -9.58982885e-01 3.87004316e-01
1.16975284e+00 -9.50241163e-02 -4.88477319e-01 1.19036265e-01
-5.15711188e-01 -4.56745148e-01 -1.57980275e+00 9.43072736e-01
-4.87890959e-01 9.01181251e-03 5.11782825e-01 -1.09590685e+00
1.22194707e+00 5.67572653e-01 6.98698461e-01 -1.12739392e-01
1.47472680e-01 2.85315275e-01 -3.96843433e-01 -6.19078517e-01
1.95737645e-01 -6.02286339e-01 3.74009162e-01 -2.01832458e-01
6.34483108e-03 -6.32276833e-01 -4.63449629e-03 -3.97192240e-01
9.27480876e-01 3.49989444e-01 5.20483544e-03 -5.01528382e-01
9.14880157e-01 1.28019806e-02 7.98957109e-01 7.80996978e-01
-2.13590771e-01 7.16487288e-01 6.15608811e-01 -1.40907630e-01
-1.03000569e+00 -8.84641886e-01 -4.58371490e-01 1.72370538e-01
2.88931340e-01 2.11287424e-01 -2.71087587e-01 -2.24264279e-01
1.78672135e-01 7.71535277e-01 -5.95520198e-01 -1.70358866e-02
-5.88357270e-01 -1.07999837e+00 4.93870154e-02 3.30051750e-01
8.08590651e-01 -8.52267146e-01 -9.29581821e-02 4.05060977e-01
2.91895509e-01 -9.88703430e-01 -1.09932773e-01 3.00703824e-01
-1.08052778e+00 -8.33793283e-01 -1.11594248e+00 -6.58005416e-01
5.59476495e-01 -3.47524732e-01 5.00415266e-01 -6.67397618e-01
7.49494433e-02 6.13575161e-01 -1.34756610e-01 -3.62174273e-01
-3.04958999e-01 -2.57496208e-01 -5.65295219e-02 3.44073176e-01
-1.72625691e-01 -8.06509554e-01 -4.96917933e-01 2.59418607e-01
-6.27042115e-01 -3.88532579e-01 4.04378682e-01 7.11384177e-01
9.58079159e-01 6.51311517e-01 1.93782911e-01 -5.45160115e-01
7.35324740e-01 -8.14166129e-01 -9.37259257e-01 -3.97580229e-02
-3.51426929e-01 1.10563025e-01 8.16671133e-01 -7.33061016e-01
-1.30775309e+00 -1.49800584e-01 -2.10112892e-02 -9.71225321e-01
-2.31200065e-02 4.69602823e-01 -1.74565464e-01 -1.78719699e-01
1.41955554e-01 2.85402626e-01 -3.21966559e-02 -8.02026868e-01
-1.91437170e-01 2.16431499e-01 3.10559779e-01 -1.16687715e+00
7.03405440e-01 4.18843955e-01 6.18056834e-01 -1.02652609e+00
-4.14389968e-01 -6.34588838e-01 -4.69712466e-01 -4.04175408e-02
9.78667319e-01 -5.06091595e-01 -6.27467394e-01 8.36057425e-01
-1.22736001e+00 -5.73256195e-01 -5.10700285e-01 7.41750181e-01
-8.64022851e-01 4.58330452e-01 -8.37461829e-01 -1.52278757e+00
-2.19013691e-01 -1.10053504e+00 1.19493926e+00 3.11459571e-01
2.74908304e-01 -1.60187352e+00 3.26355994e-01 -3.36482674e-01
1.85337171e-01 4.85408217e-01 6.71728194e-01 -5.52698076e-01
-3.89985353e-01 -2.38683388e-01 -3.21275182e-02 4.78374302e-01
-1.75379850e-02 2.59181350e-01 -8.33732724e-01 -2.00899482e-01
9.53604519e-01 2.71110058e-01 6.46468401e-01 1.02170777e+00
1.14370489e+00 -1.31844342e-01 -7.16714859e-01 7.64388263e-01
1.73850083e+00 6.64269567e-01 5.23702502e-01 1.15196168e-01
5.89611709e-01 5.73077321e-01 7.08486080e-01 5.53972602e-01
-9.26089138e-02 4.06542122e-01 4.06571664e-02 8.45981240e-02
4.92070317e-01 -1.51326731e-01 1.64305523e-01 5.84367454e-01
-1.79880738e-01 -4.12998438e-01 -9.99596357e-01 3.80288124e-01
-2.08560419e+00 -7.11745679e-01 -5.25695086e-01 2.42989373e+00
5.34834206e-01 -6.03031367e-02 -2.42161393e-01 -8.06568414e-02
1.01316905e+00 -2.00976431e-01 -5.56461811e-01 -2.78269708e-01
-7.72304982e-02 1.98017552e-01 6.26435697e-01 8.27811062e-01
-1.31637919e+00 3.89775723e-01 5.45640039e+00 9.89547849e-01
-1.18797088e+00 1.56286553e-01 5.68563998e-01 5.26074469e-01
9.75926071e-02 6.87201023e-02 -1.09696651e+00 8.45452368e-01
8.12966287e-01 -1.30629912e-01 7.05109760e-02 6.69683635e-01
8.69216859e-01 -8.03897977e-02 -6.98367059e-01 9.08093929e-01
-5.35836577e-01 -6.97712362e-01 -2.67466575e-01 1.72663778e-01
9.27281201e-01 -3.22927982e-01 7.27712885e-02 5.20569503e-01
3.81541938e-01 -8.74884784e-01 3.76623482e-01 1.18956208e+00
3.20944726e-01 -5.58245361e-01 7.89276302e-01 5.28561532e-01
-1.28801453e+00 -5.91301918e-02 -4.59584147e-01 1.70253702e-02
6.54246867e-01 8.59565198e-01 -2.84358799e-01 9.41585183e-01
4.06915605e-01 7.24693418e-01 7.87796639e-03 1.65433240e+00
4.30528447e-03 6.91021621e-01 -6.98528528e-01 1.38269424e-01
5.28293133e-01 -9.60367799e-01 1.14512658e+00 8.83543730e-01
8.75246108e-01 1.85000733e-01 3.49321902e-01 1.32096851e+00
6.22102797e-01 4.20309603e-02 -4.12628561e-01 1.04092278e-01
3.06708723e-01 1.14189601e+00 -6.92450702e-01 -9.56628770e-02
-2.66084701e-01 7.46676743e-01 -7.94452503e-02 8.75091851e-01
-9.97541308e-01 -1.61510646e-01 3.96484435e-01 2.28831843e-01
3.54871809e-01 -3.61691564e-01 -1.26204267e-01 -8.73895586e-01
2.29147561e-02 -1.92162469e-02 3.41434449e-01 -7.24278331e-01
-1.83984959e+00 3.19742799e-01 4.13403749e-01 -1.17070723e+00
9.02271792e-02 -7.01781452e-01 -1.06238472e+00 1.47305763e+00
-1.42587483e+00 -9.09013152e-01 -1.44608781e-01 3.84581923e-01
4.17100400e-01 1.53577611e-01 4.06950742e-01 6.56724796e-02
-6.38585269e-01 -3.22154127e-02 6.46122456e-01 -3.66156876e-01
3.73000294e-01 -1.29298353e+00 1.83479428e-01 6.19626939e-01
-8.34891140e-01 4.88929152e-01 1.13295090e+00 -1.01007974e+00
-1.00167143e+00 -1.30720758e+00 5.67259550e-01 -2.12550074e-01
8.49538207e-01 -1.01562344e-01 -1.36873090e+00 4.93305981e-01
-3.46979976e-01 9.37487334e-02 -3.04471347e-02 -4.43209171e-01
4.85782325e-01 1.14572845e-01 -1.55142915e+00 4.47555393e-01
3.66285861e-01 4.30145562e-02 -4.07459795e-01 4.31139976e-01
8.00684929e-01 -3.47225279e-01 -9.34834361e-01 6.22746170e-01
-6.23606220e-02 -1.83307156e-01 8.25045347e-01 -5.70948660e-01
1.22025825e-01 -4.59519565e-01 -2.68413592e-02 -1.31487393e+00
-3.25335562e-01 -1.04215336e+00 -1.38517335e-01 1.33579087e+00
2.82801032e-01 -1.15803874e+00 6.01155460e-01 8.28750849e-01
-3.00767273e-01 -1.11769843e+00 -9.80183244e-01 -1.14052749e+00
5.29075146e-01 -4.12277669e-01 2.13748306e-01 8.66893649e-01
-5.23240924e-01 5.56636006e-02 -1.85230941e-01 7.12607026e-01
9.04015899e-01 -2.97886431e-01 5.24144173e-01 -1.52405906e+00
-5.71789563e-01 -2.47814819e-01 -2.70658702e-01 -9.01274562e-01
2.08886396e-02 -3.02186072e-01 1.78633496e-01 -1.65565562e+00
-1.19305521e-01 -7.53801227e-01 -5.03405705e-02 -3.15560818e-01
-2.20793441e-01 -6.73973382e-01 -1.42158881e-01 2.73792505e-01
9.32467058e-02 9.80231047e-01 1.45425820e+00 -1.71645381e-03
-5.04783750e-01 6.97346628e-01 3.42609249e-02 6.05490565e-01
8.69135082e-01 -4.09274846e-01 -7.16789544e-01 1.59309074e-01
-3.40778351e-01 5.40571630e-01 5.65458715e-01 -1.22779357e+00
3.73404711e-01 -3.45838189e-01 1.26059011e-01 -6.99797392e-01
6.95388496e-01 -7.28684247e-01 3.75509858e-01 1.84732303e-01
6.76527712e-03 -1.19064309e-01 4.44223076e-01 1.02092135e+00
-2.00681105e-01 -5.24730146e-01 1.13196504e+00 -1.91731483e-01
-3.03240001e-01 6.16155326e-01 -3.03536952e-01 -8.12899992e-02
1.17916155e+00 -3.49469304e-01 3.50984246e-01 -2.92425662e-01
-1.29140997e+00 4.88588959e-01 2.01652959e-01 -1.62209004e-01
2.58475006e-01 -1.02389359e+00 -5.29044032e-01 -8.97932574e-02
-4.21822608e-01 2.97116011e-01 7.53843606e-01 1.27078021e+00
-4.82128114e-01 2.44376570e-01 1.42195314e-01 -6.13376796e-01
-6.01759911e-01 7.46421337e-01 8.44740391e-01 -6.33138001e-01
-6.24317944e-01 7.85176337e-01 4.27796572e-01 -5.81458271e-01
1.25061378e-01 -4.88419205e-01 -1.67369872e-01 -2.94132978e-01
1.52956247e-01 7.44107842e-01 -1.76946357e-01 -4.75405425e-01
1.69020921e-01 9.26651359e-01 3.47710520e-01 -4.54857975e-01
1.16046786e+00 -2.14146823e-01 -4.36779149e-02 8.05618703e-01
1.19164550e+00 -1.80375680e-01 -1.71852541e+00 4.93139476e-02
-2.24498197e-01 -6.55795112e-02 -1.26655144e-03 -3.94729942e-01
-8.98895562e-01 9.17153060e-01 4.68594879e-01 1.27027318e-01
8.65654290e-01 -4.08396751e-01 6.32590413e-01 -1.26281172e-01
1.46796525e-01 -9.70504344e-01 -2.92309403e-01 7.21485734e-01
8.19127262e-01 -8.82509291e-01 -3.58245522e-01 -8.25541258e-01
-5.86322486e-01 9.33114052e-01 4.93373781e-01 -5.05490184e-01
1.16444123e+00 2.48718649e-01 -4.54674095e-01 -3.17121685e-01
-5.49574852e-01 2.06560101e-02 1.37402698e-01 6.53647304e-01
4.93168496e-02 -2.06561804e-01 -5.67220449e-01 8.08299959e-01
1.45568416e-01 -3.24171036e-02 2.16980860e-01 9.18859899e-01
-3.25979412e-01 -8.12279463e-01 -6.89015329e-01 1.78627253e-01
-2.90453672e-01 1.70224205e-01 1.65343329e-01 8.13898742e-01
1.71582311e-01 8.73792529e-01 4.34352085e-02 1.65727094e-01
3.58080804e-01 3.78236115e-01 2.35515192e-01 -4.76849139e-01
-1.53767645e-01 2.51250267e-01 -2.68778741e-01 -1.61637127e-01
-1.08943805e-02 -8.43513668e-01 -1.31162655e+00 -1.87272921e-01
-1.55376151e-01 6.60388410e-01 4.49360520e-01 9.03422713e-01
2.32167870e-01 4.74926829e-01 5.09655535e-01 -1.09805858e+00
-6.25726640e-01 -9.56871331e-01 -1.19314051e+00 -1.49153292e-01
2.28144392e-01 -1.06542087e+00 -8.51338863e-01 -1.13348529e-01]
|
[6.609189033508301, 3.545553207397461]
|
e27d25cb-f12d-4b23-97ae-990eb664da85
|
dramatic-conversation-disentanglement
|
2305.16648
| null |
https://arxiv.org/abs/2305.16648v1
|
https://arxiv.org/pdf/2305.16648v1.pdf
|
Dramatic Conversation Disentanglement
|
We present a new dataset for studying conversation disentanglement in movies and TV series. While previous work has focused on conversation disentanglement in IRC chatroom dialogues, movies and TV shows provide a space for studying complex pragmatic patterns of floor and topic change in face-to-face multi-party interactions. In this work, we draw on theoretical research in sociolinguistics, sociology, and film studies to operationalize a conversational thread (including the notion of a floor change) in dramatic texts, and use that definition to annotate a dataset of 10,033 dialogue turns (comprising 2,209 threads) from 831 movies. We compare the performance of several disentanglement models on this dramatic dataset, and apply the best-performing model to disentangle 808 movies. We see that, contrary to expectation, average thread lengths do not decrease significantly over the past 40 years, and characters portrayed by actors who are women, while underrepresented, initiate more new conversational threads relative to their speaking time.
|
['David Bamman', 'Danica Chen', 'Kent K. Chang']
|
2023-05-26
| null | null | null | null |
['disentanglement', 'conversation-disentanglement']
|
['methodology', 'natural-language-processing']
|
[ 5.08551002e-02 6.34391844e-01 -3.01981926e-01 -4.40305203e-01
-6.02729917e-01 -1.09080815e+00 1.41783452e+00 4.89373095e-02
-1.95383444e-01 6.54956102e-01 1.37345481e+00 -3.31497401e-01
7.14886412e-02 -3.50809962e-01 -4.60649282e-02 -4.40544277e-01
2.92467419e-02 5.02836049e-01 -9.93605033e-02 -6.54911816e-01
1.69629097e-01 -5.58746532e-02 -1.12499809e+00 7.74939179e-01
2.30232388e-01 2.46407017e-01 -3.20637077e-01 1.08203411e+00
-1.35535419e-01 1.16409087e+00 -1.02448690e+00 -9.07670140e-01
-2.06082210e-01 -8.89595449e-01 -1.26071358e+00 2.82635152e-01
5.87150335e-01 -1.27141669e-01 -7.16803372e-01 3.23670357e-01
3.96857262e-01 1.27192810e-01 8.19030762e-01 -1.04631567e+00
-3.08323592e-01 1.23882568e+00 -5.12537539e-01 8.41074347e-01
9.26679850e-01 -5.51882982e-02 1.37811208e+00 -2.63969630e-01
1.25004816e+00 1.71736324e+00 5.94616175e-01 5.80096245e-01
-1.49544847e+00 -4.66000021e-01 -3.65717001e-02 -1.30768523e-01
-7.51999557e-01 -7.83755422e-01 9.61943090e-01 -7.31442332e-01
4.30849016e-01 8.90164077e-01 9.27835405e-01 1.76166737e+00
-1.48491651e-01 6.71654701e-01 1.11159623e+00 -3.80651116e-01
-3.38921487e-01 2.39978343e-01 3.20728868e-01 2.40959138e-01
-3.86312097e-01 -5.76906443e-01 -8.44340980e-01 -6.05862558e-01
4.61160302e-01 -5.15086770e-01 -6.40225410e-02 3.31237286e-01
-1.44679940e+00 1.08470583e+00 -1.11956544e-01 7.58310735e-01
-5.57396794e-03 -2.77733445e-01 7.28462696e-01 8.14182043e-01
9.11216915e-01 7.25136280e-01 -1.91533551e-01 -1.12582815e+00
-4.44219321e-01 7.07282722e-01 1.49337912e+00 5.92532516e-01
2.99326777e-01 -5.17105281e-01 -2.80206025e-01 1.13688028e+00
-1.20617092e-01 1.99408069e-01 1.50499418e-01 -1.44731033e+00
7.18314469e-01 4.22429800e-01 7.48162940e-02 -1.51142418e+00
-6.25778139e-01 3.44640344e-01 -4.00805116e-01 -6.90677643e-01
9.75395679e-01 -4.91432726e-01 2.98468947e-01 1.70684624e+00
3.04038495e-01 -1.76963776e-01 -1.21679708e-01 4.90637898e-01
1.13226271e+00 5.83374262e-01 -1.58185497e-01 -7.89705157e-01
1.73370051e+00 -7.44468570e-01 -1.00960028e+00 -4.08929735e-02
6.54701650e-01 -9.75297391e-01 1.10264313e+00 2.35575899e-01
-1.16135287e+00 -1.86695173e-01 -7.86371231e-01 -2.24008411e-01
2.30513379e-01 -3.69621456e-01 7.22296894e-01 7.33732641e-01
-5.41822791e-01 4.49144036e-01 -5.07914603e-01 -5.49529195e-01
8.23831484e-02 -2.48313978e-01 -3.35877687e-01 5.03959596e-01
-1.11850917e+00 9.15655136e-01 -3.43329489e-01 -2.65038669e-01
-3.06830078e-01 -5.55749238e-01 -6.87712431e-01 -3.13345551e-01
4.55630630e-01 -2.31678933e-01 1.63710678e+00 -9.68543410e-01
-1.86512315e+00 1.26953578e+00 -1.82713062e-01 -7.21662864e-02
5.01713097e-01 -8.94161612e-02 -4.60139811e-01 6.19436502e-02
-6.12142868e-02 6.28446415e-02 4.39161450e-01 -1.07274234e+00
-5.10384142e-01 -2.18992189e-01 6.11611903e-01 4.59521115e-01
-3.78093213e-01 7.05395758e-01 -5.53961731e-02 -4.58710790e-01
-6.67479485e-02 -1.21653306e+00 2.95588970e-01 -5.27005136e-01
-4.36619967e-01 -4.64787215e-01 6.70839131e-01 -6.95489168e-01
1.55259585e+00 -2.18378401e+00 5.27383745e-01 -3.62397432e-01
8.26931179e-01 -4.22769934e-01 2.47683093e-01 9.17565584e-01
1.53243378e-01 2.62582123e-01 1.65917128e-01 -7.15955496e-01
1.10090375e-01 2.22719535e-01 -1.22747339e-01 7.31255770e-01
-4.93995637e-01 6.53651834e-01 -9.27983403e-01 -5.95911920e-01
-5.91855608e-02 -4.95244227e-02 -6.55410171e-01 -3.28418650e-02
-1.77162558e-01 7.24113524e-01 -2.99116373e-01 6.27667084e-02
5.17288921e-03 -2.30752975e-01 9.33668673e-01 1.62810668e-01
-4.06245500e-01 9.13580179e-01 -3.64479899e-01 1.53874648e+00
-5.67339540e-01 1.77943242e+00 2.86765099e-01 -5.30638874e-01
6.97113872e-01 6.14161193e-01 2.99437821e-01 -3.34398299e-01
4.38638449e-01 -3.31381291e-01 6.46037579e-01 -7.43965805e-01
8.23097408e-01 -8.36053789e-02 -7.20593154e-01 7.37581789e-01
-1.43958718e-01 -4.07946050e-01 3.14248234e-01 4.04885799e-01
1.10964799e+00 -6.15287542e-01 3.73225033e-01 -3.92754704e-01
1.52328342e-01 3.45984916e-03 4.07811970e-01 5.76179266e-01
-3.81915957e-01 3.25961679e-01 1.39483559e+00 -6.41909182e-01
-1.07349467e+00 -8.24488699e-01 -1.40200615e-01 1.39650023e+00
3.53951156e-02 -7.69343495e-01 -7.82913268e-01 -3.56085181e-01
-3.16283643e-01 6.11049533e-01 -7.20727921e-01 4.20682222e-01
-1.03106129e+00 -6.93011582e-01 7.36946046e-01 -2.79063076e-01
1.96131602e-01 -9.50727463e-01 -1.00923635e-01 7.72596449e-02
-9.14703071e-01 -1.35244703e+00 -5.06343365e-01 -7.38161087e-01
-2.66902030e-01 -1.14627302e+00 -2.26064011e-01 -3.90927017e-01
-9.29598808e-02 1.75145969e-01 1.41688645e+00 -7.48988688e-02
-1.42652974e-01 2.17731774e-01 -4.70995486e-01 -1.37700051e-01
-1.09139323e+00 3.50184590e-01 1.23319775e-01 -2.20216382e-02
2.78149873e-01 -7.71959305e-01 -3.74237269e-01 4.59712863e-01
-3.26031387e-01 1.96606770e-01 -2.50690073e-01 5.63283801e-01
-6.88249826e-01 -6.85315907e-01 3.93769532e-01 -1.30611908e+00
1.08063304e+00 -8.25181544e-01 9.04680863e-02 -1.68381482e-01
-7.58395344e-02 -4.92817044e-01 2.35280707e-01 -6.73860073e-01
-1.22636819e+00 -7.34685123e-01 1.41999334e-01 4.05969620e-01
5.67838140e-02 3.52164030e-01 2.21827865e-01 5.98897815e-01
9.03236866e-01 -3.16998392e-01 2.21961200e-01 -4.64189351e-01
4.98264521e-01 1.01593602e+00 3.26051891e-01 -5.76592326e-01
4.04081017e-01 6.02285445e-01 -6.40465915e-01 -1.46071184e+00
-7.75229871e-01 -3.66290480e-01 -3.72810125e-01 -6.86908066e-01
1.06857753e+00 -8.58494222e-01 -1.51053357e+00 2.94265270e-01
-1.31164348e+00 -4.28841949e-01 6.00465499e-02 4.02787298e-01
-5.51040232e-01 4.02642041e-01 -1.05290663e+00 -7.30077922e-01
1.30075261e-01 -7.91431487e-01 6.88246250e-01 1.88826263e-01
-1.32414711e+00 -1.15470326e+00 4.88304526e-01 1.08902943e+00
1.08101822e-01 5.14052212e-01 7.85025597e-01 -7.95253992e-01
9.83531326e-02 7.86579326e-02 2.76402771e-01 -2.99954385e-01
3.85864407e-01 2.70933062e-01 -8.10484707e-01 1.53395563e-01
1.84516072e-01 -3.12636465e-01 1.91373155e-01 1.73838034e-01
3.77254635e-01 -8.61319125e-01 -3.46163571e-01 -1.99605092e-01
3.53702813e-01 1.53325871e-01 5.63965142e-01 1.00573614e-01
5.11188865e-01 1.02190232e+00 2.71411628e-01 6.12311125e-01
8.20238590e-01 9.74813759e-01 -2.54649907e-01 3.24369192e-01
1.67663932e-01 -2.44528189e-01 3.88845295e-01 1.32901704e+00
-3.12958330e-01 -5.07924318e-01 -9.01938617e-01 5.10761619e-01
-1.89827359e+00 -1.36387360e+00 -5.62465489e-01 1.51915038e+00
1.13815653e+00 2.44492322e-01 6.38347864e-01 -1.22092769e-01
6.53793812e-01 1.00667918e+00 1.44724667e-01 -7.69824088e-01
-9.08956230e-02 -4.27408934e-01 1.13897165e-02 8.75419438e-01
-7.84366012e-01 8.26122761e-01 6.92629719e+00 7.40117311e-01
-6.18499696e-01 2.65714854e-01 7.91624486e-01 -4.84938920e-01
-4.74854797e-01 7.64155313e-02 -3.56002063e-01 5.05498111e-01
1.01311493e+00 -2.58665055e-01 5.52340031e-01 4.72773731e-01
5.14457583e-01 -1.80356279e-01 -1.36238611e+00 9.88480270e-01
1.76880792e-01 -1.61200607e+00 -6.12630188e-01 4.95274514e-02
8.04577172e-01 -3.49121332e-01 -2.28552952e-01 2.53611058e-01
4.96950924e-01 -7.97130883e-01 5.70035756e-01 2.41831928e-01
4.24072236e-01 -2.03752667e-01 1.93948239e-01 4.91336673e-01
-5.31938016e-01 1.59032762e-01 2.49196261e-01 -8.49083245e-01
4.91686106e-01 1.14137821e-01 -7.11083651e-01 7.99534023e-02
5.27299106e-01 7.26468205e-01 -4.99155000e-02 -1.49975091e-01
-6.02022046e-03 7.80816615e-01 -1.29383355e-01 -4.06620860e-01
-7.95253515e-02 -3.81577194e-01 9.56681609e-01 1.14085174e+00
-5.11011183e-01 5.41482329e-01 -2.12903634e-01 3.90353918e-01
-3.25022727e-01 -9.52170268e-02 -5.73690951e-01 -4.65259045e-01
8.10587287e-01 1.33970594e+00 -7.40596652e-01 -3.07151556e-01
-5.66503584e-01 6.98253214e-01 3.08485866e-01 2.65111290e-02
-7.16859698e-01 9.18370113e-02 9.82188523e-01 1.67215213e-01
-2.80719012e-01 -4.31653529e-01 -2.74306297e-01 -1.10292399e+00
-1.81872234e-01 -1.30032694e+00 1.33182034e-01 -2.75816023e-01
-1.35292780e+00 6.19169831e-01 2.16137305e-01 -6.32468224e-01
-6.24946535e-01 -1.12197995e-02 -6.97944164e-01 4.23494846e-01
-2.87674725e-01 -7.28767633e-01 -1.14086866e-01 5.94963767e-02
9.40994501e-01 8.45258832e-02 7.03676343e-01 2.05797240e-01
-4.41778123e-01 4.98783678e-01 -4.64807861e-02 3.13470125e-01
7.34289467e-01 -1.07487297e+00 6.56463742e-01 6.62443936e-02
2.33470798e-01 5.54098666e-01 1.16655648e+00 -3.01226407e-01
-1.16007555e+00 1.68190077e-01 1.25385010e+00 -1.12549174e+00
1.23083293e+00 -9.80490327e-01 -3.58789951e-01 8.86105239e-01
7.63435125e-01 -6.61652207e-01 9.88945603e-01 8.97965968e-01
-3.46004248e-01 3.27736795e-01 -7.44112194e-01 1.06911123e+00
1.36218667e+00 -7.73234308e-01 -9.01090086e-01 4.65948999e-01
9.41809654e-01 -6.44602180e-01 -1.21530521e+00 -2.18747094e-01
1.01147008e+00 -1.09933972e+00 5.88254869e-01 -7.60225594e-01
1.02278888e+00 5.88527858e-01 5.33686280e-02 -1.15329683e+00
-3.69671732e-02 -1.44929278e+00 1.52679384e-01 1.47893286e+00
5.27896583e-01 -3.55029672e-01 7.74097502e-01 7.04665065e-01
-2.44506150e-02 -4.21343267e-01 -9.75023210e-01 -2.27785930e-01
1.37900516e-01 -1.45631060e-01 2.33961269e-01 1.48863804e+00
8.58525038e-01 1.02458107e+00 -6.79734051e-01 -5.25455177e-01
-3.61727700e-02 1.52330115e-01 1.24760747e+00 -1.23555934e+00
-6.01521790e-01 -5.51735342e-01 -1.42314881e-01 -1.30378842e+00
1.69385880e-01 -2.83003092e-01 -9.35407132e-02 -1.14399076e+00
4.72598881e-01 -2.51391500e-01 9.55428004e-01 -2.38477603e-01
3.54687981e-02 3.94550562e-02 2.86144257e-01 3.57784510e-01
-4.40159619e-01 4.37070817e-01 1.41647995e+00 4.34727743e-02
-4.26801950e-01 1.82694986e-01 -8.51440310e-01 9.25769627e-01
4.95637089e-01 -3.79641116e-01 -1.63780063e-01 -7.84022883e-02
6.04548693e-01 8.21595371e-01 -9.06002969e-02 -5.48902974e-02
-1.51010796e-01 -3.62447530e-01 -4.67975587e-01 9.42281168e-03
7.09238052e-01 -3.00035566e-01 2.02721953e-01 1.29237622e-01
-8.33733261e-01 1.14935003e-01 -3.94899696e-02 5.90346694e-01
-1.96989179e-01 -1.67245883e-02 1.78350255e-01 -5.82048483e-02
1.96467549e-01 -2.42510453e-01 -1.12214112e+00 6.28603101e-01
8.74190986e-01 -1.52268887e-01 -7.60331571e-01 -9.28590000e-01
-1.02905726e+00 -5.21587320e-02 3.35397482e-01 8.69647026e-01
-2.59789526e-01 -1.01529443e+00 -1.01654053e+00 -6.37004435e-01
-2.92843226e-02 -4.91692126e-01 3.70282948e-01 8.26727211e-01
-5.22142410e-01 9.94447544e-02 1.53275162e-01 -2.58582294e-01
-1.63021326e+00 6.35593459e-02 8.88060331e-02 -2.14748412e-01
-5.13280511e-01 7.08651245e-01 1.59529939e-01 -4.20536309e-01
-9.40049961e-02 -1.53625965e-01 -3.93541962e-01 7.13166475e-01
3.80708516e-01 6.32263482e-01 -5.51460922e-01 -7.63900042e-01
-1.38138682e-01 1.94995720e-02 -1.26617372e-01 -5.34256876e-01
1.23934674e+00 -6.12356246e-01 -1.74631268e-01 1.25540233e+00
1.43235528e+00 6.89309299e-01 -8.85285258e-01 -2.33327970e-01
-2.27630455e-02 -4.55410570e-01 -5.14797032e-01 -3.08715105e-01
-3.72039795e-01 1.91821530e-01 -3.41987848e-01 1.21614254e+00
2.16192991e-01 6.17480338e-01 6.10563636e-01 2.76267882e-02
-9.55445766e-02 -1.00814199e+00 3.14911276e-01 5.31300664e-01
1.01361537e+00 -1.12960148e+00 -3.79754528e-02 -8.27293992e-01
-8.92966986e-01 8.37153614e-01 3.87169421e-01 -2.12430414e-02
6.37486398e-01 1.49553776e-01 7.83857256e-02 -6.57189786e-01
-1.09806871e+00 4.21234488e-01 -4.41369377e-02 4.56350818e-02
7.64671564e-01 3.69656622e-01 -8.90272677e-01 5.12382865e-01
-9.67129052e-01 -7.26208329e-01 1.02884746e+00 5.04376173e-01
2.98308432e-02 -8.47681761e-01 -9.75635275e-03 2.80509740e-01
-7.02977419e-01 4.48827483e-02 -9.63249266e-01 1.11513638e+00
-2.83214182e-01 1.38327312e+00 6.01363897e-01 -5.57733893e-01
9.64715481e-02 -1.22259043e-01 4.75277603e-01 -7.12313890e-01
-1.01307261e+00 -2.93962508e-01 1.38545287e+00 -1.53079793e-01
-8.30136240e-01 -1.23820984e+00 -5.12717545e-01 -1.09313142e+00
-3.30126017e-01 2.43130073e-01 3.95954281e-01 1.27770102e+00
1.16538659e-01 2.94702321e-01 9.49542522e-01 -5.43038309e-01
-8.03520307e-02 -1.52843320e+00 -4.33781058e-01 5.51192522e-01
3.49062949e-01 -5.03559947e-01 -6.46856606e-01 8.25196654e-02]
|
[12.519224166870117, 8.052574157714844]
|
f27c3849-2565-4673-bba6-c628af4e89b2
|
does-william-shakespeare-really-write-hamlet
|
1705.03202
| null |
http://arxiv.org/abs/1705.03202v2
|
http://arxiv.org/pdf/1705.03202v2.pdf
|
Does William Shakespeare REALLY Write Hamlet? Knowledge Representation Learning with Confidence
|
Knowledge graphs (KGs), which could provide essential relational information
between entities, have been widely utilized in various knowledge-driven
applications. Since the overall human knowledge is innumerable that still grows
explosively and changes frequently, knowledge construction and update
inevitably involve automatic mechanisms with less human supervision, which
usually bring in plenty of noises and conflicts to KGs. However, most
conventional knowledge representation learning methods assume that all triple
facts in existing KGs share the same significance without any noises. To
address this problem, we propose a novel confidence-aware knowledge
representation learning framework (CKRL), which detects possible noises in KGs
while learning knowledge representations with confidence simultaneously.
Specifically, we introduce the triple confidence to conventional
translation-based methods for knowledge representation learning. To make triple
confidence more flexible and universal, we only utilize the internal structural
information in KGs, and propose three kinds of triple confidences considering
both local and global structural information. In experiments, We evaluate our
models on knowledge graph noise detection, knowledge graph completion and
triple classification. Experimental results demonstrate that our
confidence-aware models achieve significant and consistent improvements on all
tasks, which confirms the capability of CKRL modeling confidence with
structural information in both KG noise detection and knowledge representation
learning.
|
['Ruobing Xie', 'Fen Lin', 'Leyu Lin', 'Zhiyuan Liu']
|
2017-05-09
| null | null | null | null |
['triple-classification']
|
['graphs']
|
[-2.33861819e-01 1.07605480e-01 -4.65334535e-01 -2.10519537e-01
-3.93602908e-01 -4.20948178e-01 1.87005490e-01 5.57713509e-01
-7.81431794e-02 7.67831862e-01 1.66798593e-03 -2.78663725e-01
-4.97656584e-01 -1.30296707e+00 -7.86638618e-01 -3.84323359e-01
5.68677671e-02 3.92287105e-01 4.67257887e-01 -2.57706434e-01
-1.11654624e-01 1.23714775e-01 -1.39818537e+00 1.44475803e-01
1.28918803e+00 1.07053185e+00 1.22961858e-02 2.88156867e-02
-4.37607288e-01 1.29996061e+00 -5.79860151e-01 -7.98843920e-01
-6.37807697e-02 1.34183750e-01 -9.48948860e-01 -1.27979264e-01
1.11350223e-01 1.96505755e-01 -5.01542568e-01 1.38226521e+00
1.92083031e-01 7.39962682e-02 4.41229373e-01 -1.31654525e+00
-1.15480053e+00 1.06767869e+00 -8.01664591e-01 1.36538476e-01
5.25192440e-01 -2.28389934e-01 1.11295390e+00 -9.47476029e-01
5.36934316e-01 1.33103979e+00 7.35756278e-01 -1.04022965e-01
-7.83461571e-01 -8.01633716e-01 6.87218547e-01 7.58363426e-01
-1.89067996e+00 -8.29428658e-02 8.25776756e-01 -2.30322540e-01
7.22654521e-01 1.94310039e-01 7.24285066e-01 7.45982409e-01
-2.87599117e-01 9.59858239e-01 7.65084326e-01 -4.33723718e-01
-2.57592909e-02 1.24041781e-01 3.80174518e-01 8.98467481e-01
1.16499388e+00 -4.53023642e-01 -5.66931009e-01 -9.55990031e-02
6.35469556e-01 -1.38651170e-02 -4.18283224e-01 -5.38906157e-01
-1.18865335e+00 5.32979429e-01 3.14787537e-01 3.20547760e-01
-1.33871660e-01 3.65958773e-02 4.45983082e-01 4.45628852e-01
1.56586245e-01 1.85776930e-02 -8.22093844e-01 9.84196663e-02
-2.69698203e-01 -8.66070464e-02 7.94427872e-01 1.32947600e+00
1.05948031e+00 4.35671099e-02 6.16386160e-03 7.87850618e-01
3.52927208e-01 5.23009956e-01 3.91118944e-01 -2.68429965e-01
7.79909253e-01 1.26972461e+00 -5.13594784e-02 -1.71050024e+00
-4.44478989e-01 -7.35345960e-01 -1.08572018e+00 -6.19448900e-01
-7.95269832e-02 5.52183501e-02 -8.69843185e-01 1.63092160e+00
6.23809338e-01 5.65249622e-01 2.33028188e-01 5.24285674e-01
1.05815601e+00 2.87608415e-01 1.34883538e-01 -5.46206295e-01
1.33979225e+00 -5.88509142e-01 -9.61322904e-01 1.55006652e-03
9.30955768e-01 -5.11871517e-01 9.46863055e-01 4.25521165e-01
-4.47117239e-01 -4.08260971e-01 -1.18396378e+00 9.17626768e-02
-5.73076129e-01 7.44809285e-02 1.10881567e+00 6.82067513e-01
-4.80022371e-01 4.03501779e-01 -5.87491095e-01 -3.95996645e-02
3.44132781e-01 1.35011315e-01 -5.44795811e-01 -4.50214893e-01
-1.69711483e+00 9.01295066e-01 1.07377768e+00 3.27702701e-01
-2.35136330e-01 -6.22989833e-01 -1.10602653e+00 3.06872912e-02
1.24679267e+00 -7.44755864e-01 8.60952318e-01 -3.38770330e-01
-9.12513614e-01 3.76148045e-01 3.03697251e-02 -1.73422769e-02
2.57910609e-01 -2.09531233e-01 -1.23214471e+00 -2.22952247e-01
1.81950688e-01 -3.28880757e-01 6.34833455e-01 -1.27620089e+00
-7.22708642e-01 -3.45602661e-01 1.48236990e-01 2.48158693e-01
-4.41920489e-01 -3.30240190e-01 -7.96959043e-01 -6.18580222e-01
6.67593479e-01 -4.09648716e-01 -6.55240715e-02 -6.25159860e-01
-5.92815578e-01 -4.65689898e-01 6.18263304e-01 -5.27710378e-01
1.66079330e+00 -1.80069625e+00 1.08619891e-01 8.11113477e-01
3.92476618e-01 3.71745855e-01 6.78032935e-02 1.61596566e-01
-4.38384488e-02 2.70760417e-01 2.52309795e-02 2.14608818e-01
-7.01972768e-02 5.49708903e-01 -2.07257882e-01 -2.41597965e-02
1.26754537e-01 1.02445078e+00 -1.16851747e+00 -6.33152485e-01
3.78281474e-02 1.71876207e-01 -2.06189498e-01 -1.63045481e-01
-1.10808812e-01 -6.28750250e-02 -7.92436957e-01 1.12607467e+00
8.30079734e-01 -5.23752153e-01 6.57028019e-01 -8.06941152e-01
4.16173190e-01 -5.12062525e-03 -1.93524158e+00 1.34539270e+00
-1.21677577e-01 -3.09227645e-01 -2.71937162e-01 -1.23454320e+00
9.66485977e-01 4.85329218e-02 1.56950042e-01 -5.71642041e-01
-1.29406899e-01 2.16618970e-01 -1.10019997e-01 -5.44245064e-01
4.75695968e-01 1.62522972e-01 -1.46996658e-02 -2.34957095e-02
2.00818241e-01 2.12089136e-01 5.12268901e-01 6.95249856e-01
1.03982890e+00 -1.98302910e-01 5.60084343e-01 2.01265868e-02
7.56350219e-01 -2.28538483e-01 1.08412933e+00 5.75173795e-01
2.35144310e-02 -1.08156353e-03 7.54593551e-01 -2.60278761e-01
-3.81512046e-01 -9.22070920e-01 2.66688555e-01 8.26526046e-01
4.33753818e-01 -9.94760573e-01 -1.68704361e-01 -9.90007758e-01
3.75959963e-01 4.25046027e-01 -5.31067908e-01 -5.46213210e-01
-2.46401608e-01 -6.73630416e-01 5.46571434e-01 7.76613414e-01
5.02980649e-01 -7.05470085e-01 5.45030951e-01 2.37055555e-01
-3.38223040e-01 -1.20382345e+00 -2.24769622e-01 -1.10918446e-03
-6.80223942e-01 -1.63652980e+00 1.09729417e-01 -7.38989949e-01
7.27135539e-01 6.59387946e-01 1.08938217e+00 3.44245553e-01
-1.74191520e-01 5.41690230e-01 -6.11126840e-01 -3.03935260e-01
3.79811078e-02 -4.21313234e-02 3.97117168e-01 -9.11600292e-02
4.88496810e-01 -8.16311419e-01 -2.78859049e-01 2.89404094e-01
-9.31568742e-01 -1.44586176e-01 8.88031244e-01 6.82012916e-01
8.71090293e-01 7.12327838e-01 8.57374191e-01 -1.19545233e+00
7.13129818e-01 -5.81808448e-01 -4.73093867e-01 9.21616256e-01
-1.03704667e+00 1.82947293e-01 3.54816258e-01 -2.52324760e-01
-1.06808984e+00 -2.15113387e-01 2.22855702e-01 -5.77086568e-01
2.45390341e-01 1.39106381e+00 -6.35922849e-01 -1.66299522e-01
4.66316313e-01 2.68870294e-01 -5.41300774e-01 -3.96456271e-01
6.74893439e-01 1.82147563e-01 5.72826564e-01 -9.29626584e-01
1.00880623e+00 1.14971705e-01 6.54540658e-02 -4.34751064e-01
-1.05857503e+00 -5.68495214e-01 -5.14761209e-01 -6.90127686e-02
6.66780844e-02 -1.11278558e+00 -9.72395718e-01 2.76432425e-01
-9.07225490e-01 6.28009617e-01 -2.33968526e-01 4.32659060e-01
5.01587018e-02 8.37300837e-01 -3.58005047e-01 -7.26334751e-01
-2.48511136e-01 -6.58835292e-01 4.84893203e-01 2.83828080e-01
2.59744376e-01 -8.53011906e-01 -1.06146634e-01 3.45641911e-01
4.29198630e-02 1.36775374e-01 1.18311048e+00 -6.37510478e-01
-6.92212462e-01 -1.50246412e-01 -3.76992643e-01 2.79498994e-01
4.02248114e-01 -1.22027896e-01 -4.21495229e-01 -7.05971494e-02
-4.70755905e-01 -5.17518401e-01 1.06327128e+00 -2.47328505e-01
1.35405087e+00 -2.61051506e-01 -4.41202283e-01 2.45191321e-01
1.25712621e+00 -7.74728209e-02 5.32653511e-01 1.46623746e-01
1.22193158e+00 3.04041058e-01 7.81605065e-01 4.75022435e-01
8.80724609e-01 2.50653207e-01 2.70760149e-01 2.23706245e-01
1.26573652e-01 -6.72088861e-01 -9.10040364e-02 1.33878720e+00
-4.15401012e-01 -1.78954862e-02 -7.10836828e-01 5.41899860e-01
-2.25429726e+00 -8.96044135e-01 -2.08348438e-01 2.00853252e+00
1.22003317e+00 3.00835222e-01 -3.55368435e-01 4.62265909e-01
8.55607748e-01 -8.94303769e-02 -5.65016747e-01 4.34307277e-01
-4.00526226e-01 -8.36264491e-02 3.79211247e-01 2.13641524e-01
-9.74932730e-01 9.47772801e-01 4.98714781e+00 1.11856508e+00
-5.17327785e-01 -1.06480137e-01 2.06846103e-01 4.40396428e-01
-6.70380950e-01 8.42428952e-02 -6.93150520e-01 2.46334538e-01
2.74728417e-01 -5.57937920e-01 1.59781143e-01 1.11348414e+00
-3.18971634e-01 -5.37593029e-02 -8.63982320e-01 1.22040641e+00
-9.97116044e-02 -1.30901146e+00 5.09819388e-01 -3.59469116e-01
8.85372639e-01 -5.01793623e-01 -1.92819253e-01 7.35240102e-01
6.02383137e-01 -7.52014220e-01 3.41482222e-01 7.01902092e-01
5.16726434e-01 -1.05755103e+00 9.37804341e-01 1.99403942e-01
-1.82904530e+00 9.18560550e-02 -6.22917235e-01 1.95541531e-01
-7.78188482e-02 1.14599419e+00 -6.50744379e-01 1.53808486e+00
6.34475052e-01 8.95161510e-01 -7.74336338e-01 9.43363726e-01
-6.54879630e-01 5.16079664e-01 -3.24856222e-01 1.77005932e-01
-2.61647552e-01 -9.44903567e-02 1.47815228e-01 1.06877184e+00
2.54273623e-01 3.07873189e-01 5.46114385e-01 5.35761356e-01
-4.52841282e-01 4.30373162e-01 -4.29270685e-01 -2.45274916e-01
9.45384443e-01 1.27590895e+00 -6.42944098e-01 -3.65508050e-01
-6.02376044e-01 5.49925685e-01 7.96848953e-01 3.15883607e-01
-5.69855988e-01 -5.82769096e-01 4.75009650e-01 -1.78279147e-01
3.59813750e-01 -1.30497873e-01 1.76590402e-02 -1.56828225e+00
4.34979618e-01 -6.90588534e-01 9.10758793e-01 -5.66056669e-01
-1.54155397e+00 9.24485922e-02 1.01617286e-02 -1.02333319e+00
1.98344052e-01 -5.59433401e-01 -3.28478485e-01 4.28669989e-01
-1.70241272e+00 -1.31588221e+00 -3.82386655e-01 9.56669509e-01
-1.49606138e-01 -1.22461006e-01 6.66776597e-01 5.50700128e-01
-7.11977780e-01 8.12263966e-01 -1.06751084e-01 2.91300595e-01
6.94749892e-01 -1.22700465e+00 4.55581620e-02 8.31121504e-01
2.57384449e-01 1.15829599e+00 3.40457022e-01 -1.14391923e+00
-1.73137450e+00 -1.19319403e+00 6.28192484e-01 -2.16271788e-01
8.90806198e-01 9.04498994e-02 -1.28020871e+00 8.86878252e-01
-5.15981972e-01 4.32364583e-01 6.57789767e-01 8.14443350e-01
-9.29049909e-01 -3.34136158e-01 -8.27879012e-01 4.47153836e-01
1.39054406e+00 -6.07215941e-01 -7.40090370e-01 3.75807375e-01
1.02734375e+00 -3.57627869e-01 -1.04312992e+00 9.90986764e-01
2.65687853e-01 -4.92047131e-01 9.27240372e-01 -7.17120349e-01
-2.71990657e-01 -9.59467709e-01 -9.43983942e-02 -1.24543214e+00
-5.03929019e-01 -2.10432112e-01 -8.68486345e-01 1.43396270e+00
4.50395525e-01 -6.12168252e-01 7.60632753e-01 4.61059541e-01
-1.55670494e-01 -5.89697361e-01 -9.11517859e-01 -9.57900703e-01
-5.63363433e-01 -6.54985845e-01 7.29678810e-01 1.57281661e+00
3.67989928e-01 3.83496463e-01 -4.74124938e-01 6.39286995e-01
4.47329104e-01 2.10940808e-01 5.38773000e-01 -1.56128824e+00
-1.80742517e-01 -1.57173797e-01 -8.26208234e-01 -6.72814786e-01
1.58199817e-02 -9.78805661e-01 -4.03595626e-01 -1.81839776e+00
3.44724447e-01 -5.31835973e-01 -7.65954733e-01 8.91640961e-01
-6.12269342e-01 -2.31072873e-01 -3.40767473e-01 1.42082796e-01
-1.17203772e+00 6.04720235e-01 1.17739570e+00 -3.60653490e-01
6.55182004e-02 -1.58198819e-01 -1.16987956e+00 9.12603498e-01
3.76788318e-01 -2.10849717e-01 -8.83910298e-01 -2.46140525e-01
9.16632056e-01 -3.42990488e-01 3.57604623e-02 -8.35672975e-01
6.54558539e-01 -2.87990361e-01 3.71457994e-01 -6.54605567e-01
-9.78079066e-02 -8.71637583e-01 3.09227824e-01 2.91642696e-01
2.14326486e-01 -2.46114030e-01 7.76994452e-02 1.28753316e+00
-3.95409822e-01 5.80994710e-02 1.69755489e-01 -5.38635952e-03
-1.23059857e+00 4.88045752e-01 2.05190867e-01 1.60516471e-01
9.03802335e-01 -2.61261985e-02 -4.69388068e-01 -8.10113400e-02
-9.46348369e-01 6.08074665e-01 -6.84982687e-02 5.23333430e-01
8.42632771e-01 -1.59722126e+00 -3.93392593e-01 5.62551878e-02
5.42895913e-01 3.55122626e-01 4.99521703e-01 5.73917389e-01
2.00544931e-02 1.91444248e-01 2.65662700e-01 -2.66193092e-01
-1.00540054e+00 8.46859753e-01 1.57839004e-02 -4.65751976e-01
-4.14621651e-01 9.19592559e-01 -1.93659410e-01 -5.65200686e-01
2.99702495e-01 -3.91393781e-01 -4.67955470e-01 1.21319622e-01
5.00298023e-01 4.01782632e-01 3.97689581e-01 -3.06888223e-01
-4.34845090e-01 4.31996316e-01 -5.73016882e-01 6.91809654e-01
9.24543619e-01 -1.62642226e-01 -3.23261589e-01 3.30950081e-01
5.01351833e-01 2.77590126e-01 -4.17122900e-01 -7.61150360e-01
4.11891133e-01 -5.56470871e-01 -1.47538409e-01 -9.13249493e-01
-1.18447423e+00 2.58014977e-01 9.31973681e-02 1.32992864e-01
8.49210322e-01 1.00571170e-01 4.48998660e-01 1.09498549e+00
8.12338650e-01 -1.22254121e+00 3.62408385e-02 5.69100797e-01
7.02518821e-01 -1.27161157e+00 3.75353634e-01 -1.13852894e+00
-5.79051852e-01 8.26868594e-01 1.01602697e+00 4.21761304e-01
7.34655976e-01 -4.26758407e-03 -4.49165314e-01 -2.66836345e-01
-7.38052309e-01 -3.87458235e-01 3.69171202e-01 8.46215010e-01
1.48218200e-01 1.95182949e-01 -3.04092228e-01 1.31277263e+00
1.44491628e-01 -3.60077955e-02 3.31809968e-01 9.69065487e-01
-6.79995716e-01 -1.20996380e+00 -1.52526110e-01 6.51969612e-01
-2.14364361e-02 -1.23663276e-01 -2.45262817e-01 7.57770061e-01
3.55507165e-01 9.25930738e-01 -6.78886354e-01 -6.88896596e-01
7.79869914e-01 2.60774121e-02 4.63907361e-01 -6.27997220e-01
1.14700034e-01 -4.74877894e-01 3.12586039e-01 -4.16312665e-01
-3.81480664e-01 -3.00106406e-01 -1.47312677e+00 -5.21401167e-01
-8.07139337e-01 4.48072791e-01 8.89888555e-02 1.15435994e+00
4.13862407e-01 7.17856228e-01 2.84573674e-01 2.17556641e-01
-4.54298705e-01 -8.72500539e-01 -9.63139415e-01 3.17625016e-01
-3.33634079e-01 -1.08606601e+00 -3.62825431e-02 -1.56131029e-01]
|
[8.784164428710938, 7.895570278167725]
|
e0d5ac48-4787-467b-ba01-ee38f0c19230
|
counterfactual-explanation-based-on-gradual
|
2008.01897
| null |
https://arxiv.org/abs/2008.01897v2
|
https://arxiv.org/pdf/2008.01897v2.pdf
|
Counterfactual Explanation Based on Gradual Construction for Deep Networks
|
To understand the black-box characteristics of deep networks, counterfactual explanation that deduces not only the important features of an input space but also how those features should be modified to classify input as a target class has gained an increasing interest. The patterns that deep networks have learned from a training dataset can be grasped by observing the feature variation among various classes. However, current approaches perform the feature modification to increase the classification probability for the target class irrespective of the internal characteristics of deep networks. This often leads to unclear explanations that deviate from real-world data distributions. To address this problem, we propose a counterfactual explanation method that exploits the statistics learned from a training dataset. Especially, we gradually construct an explanation by iterating over masking and composition steps. The masking step aims to select an important feature from the input data to be classified as a target class. Meanwhile, the composition step aims to optimize the previously selected feature by ensuring that its output score is close to the logit space of the training data that are classified as the target class. Experimental results show that our method produces human-friendly interpretations on various classification datasets and verify that such interpretations can be achieved with fewer feature modification.
|
['Hee-Dong Kim', 'Hong-Gyu Jung', 'Seong-Whan Lee', 'Dong-Ok Won', 'Sin-Han Kang']
|
2020-08-05
| null | null | null | null |
['counterfactual-explanation']
|
['miscellaneous']
|
[ 4.52143610e-01 3.96112740e-01 -4.06702816e-01 -8.95577908e-01
-7.04878345e-02 -5.41016042e-01 6.28595114e-01 4.83261501e-05
-3.65716487e-01 9.04012084e-01 1.49024889e-01 -4.40562218e-01
-5.06804764e-01 -9.67757106e-01 -9.43858624e-01 -9.13105071e-01
2.13769123e-01 4.07292783e-01 -1.61865935e-01 1.63810596e-01
3.55288714e-01 5.18420398e-01 -1.73900878e+00 5.19053936e-01
1.03549147e+00 1.03470325e+00 1.73188627e-01 1.26895621e-01
-2.21913412e-01 3.28194976e-01 -6.70548499e-01 -4.58643049e-01
2.85854250e-01 -6.38598800e-01 -6.49785995e-01 3.12465653e-02
1.16012447e-01 -2.46057272e-01 -6.32032156e-02 1.33546078e+00
2.83559412e-02 1.03516936e-01 8.55691969e-01 -1.50704801e+00
-8.24124873e-01 8.02799523e-01 -2.97738433e-01 3.65392268e-02
-2.80470431e-01 3.65463525e-01 1.20226502e+00 -5.32124758e-01
2.84338981e-01 1.25730658e+00 1.87782228e-01 5.44957399e-01
-1.36043048e+00 -8.99268746e-01 5.39845586e-01 3.02180320e-01
-1.01212740e+00 -1.61613211e-01 1.01827884e+00 -1.43203154e-01
3.55982989e-01 4.32309836e-01 6.72545135e-01 1.10133421e+00
3.84030700e-01 6.79073811e-01 1.02982759e+00 -2.30478615e-01
4.54157114e-01 4.77686673e-01 1.96900338e-01 5.05743861e-01
6.03065133e-01 3.18030328e-01 -3.96348864e-01 -7.42725134e-02
5.81857443e-01 3.51012141e-01 -4.75395948e-01 -5.85540533e-01
-1.14978540e+00 9.35596108e-01 7.65857637e-01 2.69180119e-01
-6.22760952e-01 1.62389070e-01 5.04881702e-02 2.09397510e-01
2.18262419e-01 7.91479111e-01 -9.55657661e-01 2.84777671e-01
-4.69037801e-01 3.07387590e-01 5.81384659e-01 4.63324517e-01
1.03843760e+00 -5.15310057e-02 -1.73951536e-01 5.07276356e-01
2.98282534e-01 3.73796791e-01 6.69401467e-01 -7.85885215e-01
4.76642370e-01 9.40111816e-01 1.14660986e-01 -9.65440571e-01
-2.39090666e-01 -7.95849681e-01 -9.38429832e-01 2.92212129e-01
7.18825161e-01 -2.85931945e-01 -9.11747515e-01 2.12624669e+00
2.77429461e-01 -1.38613179e-01 2.26565808e-01 1.05573285e+00
3.03527206e-01 4.90589857e-01 6.24782685e-03 -9.81816724e-02
1.12039351e+00 -3.86001855e-01 -5.05276620e-01 -3.75343591e-01
5.40774524e-01 -1.42800763e-01 1.25895000e+00 1.19789720e-01
-4.45065588e-01 -5.90349913e-01 -1.15316474e+00 5.12482226e-01
-3.04769158e-01 1.67563543e-01 9.17927623e-01 5.43786705e-01
-3.87574017e-01 8.90158057e-01 -6.04820967e-01 1.07468165e-01
7.94315457e-01 5.25935531e-01 -3.03547949e-01 2.16210961e-01
-1.24910057e+00 4.98501658e-01 8.36589634e-01 1.32402062e-01
-7.61507511e-01 -7.98096657e-01 -7.81738043e-01 7.30939150e-01
5.18981099e-01 -6.89212084e-01 1.08201766e+00 -1.58709538e+00
-1.23671424e+00 3.23677331e-01 -1.72508478e-01 -4.50126588e-01
4.21934247e-01 1.76532213e-02 -2.33946487e-01 -3.18545073e-01
1.14625111e-01 5.90104938e-01 8.43722045e-01 -1.36172295e+00
-7.94187725e-01 -5.40249765e-01 1.15020573e-01 1.04673736e-01
-2.16949254e-01 -6.12552404e-01 3.47391009e-01 -4.74333376e-01
4.44786996e-01 -7.83185005e-01 -2.06575930e-01 5.81869148e-02
-7.69923747e-01 -1.84637725e-01 6.33685708e-01 -7.74372071e-02
7.49496639e-01 -2.04162383e+00 -1.83603927e-01 3.79895717e-01
2.84360081e-01 -1.53725808e-02 1.21139482e-01 -1.56111181e-01
-4.38177258e-01 3.99461389e-01 -2.83160090e-01 1.15078688e-01
2.97296531e-02 2.82466650e-01 -6.90976322e-01 3.61730695e-01
4.01309758e-01 7.26458430e-01 -8.68099093e-01 -5.84743023e-02
4.12458666e-02 2.71050148e-02 -6.79955006e-01 2.65380502e-01
-2.70738244e-01 2.92438567e-01 -4.63280559e-01 1.63105741e-01
6.01678550e-01 -3.48290980e-01 3.23614091e-01 -1.50525436e-01
1.78249538e-01 4.56299096e-01 -1.25688004e+00 9.16238546e-01
-3.77496868e-01 6.60786867e-01 -5.29405653e-01 -1.25503373e+00
1.02760339e+00 1.10497437e-01 -7.35262930e-02 -4.34811324e-01
1.63211733e-01 2.26104483e-01 7.15899706e-01 -4.07844603e-01
7.74066970e-02 -5.10884404e-01 9.58355069e-02 6.24116242e-01
-1.20692246e-01 2.43007287e-01 -3.74682665e-01 -1.24994747e-01
5.73757946e-01 -1.91788375e-01 4.53851491e-01 -3.29134762e-01
3.41610640e-01 -1.39512032e-01 9.01733816e-01 9.98412132e-01
-1.33634005e-02 4.37424898e-01 9.74637568e-01 -7.98199594e-01
-7.28804588e-01 -1.11654460e+00 -5.24532087e-02 6.47619605e-01
4.76100631e-02 3.82005304e-01 -4.88778800e-01 -1.27525938e+00
9.05351564e-02 1.11736751e+00 -1.08512759e+00 -7.02684462e-01
-2.04332963e-01 -8.65591526e-01 6.55707344e-02 4.82032150e-01
7.04017520e-01 -1.30181754e+00 -8.33276093e-01 -1.31355643e-01
9.85859241e-03 -3.50770503e-01 -1.39139980e-01 4.65676010e-01
-9.36699510e-01 -1.31250107e+00 -2.18588218e-01 -3.58527511e-01
1.05586791e+00 1.39762327e-01 7.60740519e-01 2.53563970e-01
2.79023588e-01 -4.77487028e-01 -2.38448437e-02 -6.27028763e-01
-2.49850795e-01 4.70893383e-02 1.46700323e-01 5.29298961e-01
5.10393739e-01 -6.11253798e-01 -6.92247689e-01 2.61759222e-01
-9.61630762e-01 2.65879869e-01 8.66257012e-01 9.92429256e-01
4.48375493e-01 4.18720186e-01 8.23029280e-01 -1.08500648e+00
5.41027486e-01 -6.14154816e-01 -3.89112949e-01 3.36032242e-01
-7.66877234e-01 6.16516352e-01 1.06365550e+00 -7.64475167e-01
-1.14574969e+00 4.83009554e-02 1.82069153e-01 -1.46990016e-01
-4.42070872e-01 5.76217473e-01 -6.58309996e-01 7.13010073e-01
6.02697790e-01 3.63111883e-01 -4.57423218e-02 -3.40712786e-01
2.08317190e-01 4.79743838e-01 4.17862415e-01 -5.22639453e-01
7.78266430e-01 5.18017113e-01 -7.18012154e-02 -1.29936069e-01
-1.10892522e+00 3.30578834e-01 -6.49626195e-01 4.28361781e-02
6.00114405e-01 -3.19376916e-01 -7.40237951e-01 1.60322890e-01
-1.11698782e+00 -7.49760941e-02 -4.44003999e-01 6.28738105e-01
-2.91220695e-01 -3.09468597e-01 2.40517497e-01 -7.55041122e-01
8.04484263e-02 -1.26487339e+00 6.38829827e-01 4.17641759e-01
-3.72857600e-01 -8.62802505e-01 -3.11289370e-01 -6.03101999e-02
1.57533035e-01 1.52925864e-01 1.45685172e+00 -1.20022368e+00
-7.65322268e-01 -1.82941392e-01 -3.75836462e-01 1.56145230e-01
5.13804495e-01 -5.21393940e-02 -9.83759046e-01 3.74425501e-02
1.92364544e-01 1.50982529e-01 9.63350534e-01 5.41466236e-01
1.56935227e+00 -7.57800758e-01 -3.42874795e-01 5.69114268e-01
1.08680427e+00 5.43439627e-01 3.44406188e-01 3.96276474e-01
4.14337099e-01 7.78228819e-01 4.46831852e-01 1.75488323e-01
4.97722290e-02 5.59675038e-01 8.07325363e-01 -1.12593686e-02
4.09523427e-01 -5.14221907e-01 8.35088193e-02 -1.72561064e-01
2.45852605e-01 -2.94491439e-03 -5.97522140e-01 6.45298421e-01
-1.88994861e+00 -9.66283321e-01 3.32848132e-02 2.23400617e+00
7.26987422e-01 4.26192462e-01 -2.88040906e-01 2.81184584e-01
7.23151803e-01 1.97334168e-03 -1.12316036e+00 -3.46733570e-01
1.78495143e-02 -2.32642740e-01 2.84427702e-01 3.21626276e-01
-6.87054098e-01 5.40206850e-01 5.07211256e+00 5.98862350e-01
-1.28580081e+00 -3.92478973e-01 1.07514179e+00 2.83215661e-02
-7.87125349e-01 1.97744668e-01 -6.58097565e-01 5.10585666e-01
5.85305512e-01 -4.78139043e-01 3.11808497e-01 8.41321528e-01
3.49079102e-01 -2.55301483e-02 -1.48488712e+00 5.45667946e-01
-3.83953214e-01 -1.40761745e+00 6.14010990e-01 2.06333339e-01
5.76465487e-01 -4.83216375e-01 3.25958312e-01 2.60315806e-01
2.58418828e-01 -1.03506124e+00 8.11212599e-01 4.30047899e-01
3.93279314e-01 -9.39965129e-01 9.31230664e-01 7.03766525e-01
-5.34601569e-01 -3.00372392e-01 -5.19603014e-01 -4.11826491e-01
-4.11974758e-01 7.65933812e-01 -1.19229829e+00 2.53517777e-01
5.43655336e-01 4.50352192e-01 -5.48345864e-01 7.58010745e-01
-6.90756440e-01 5.39175749e-01 5.77870198e-02 -2.30864137e-01
1.03002429e-01 -1.35968416e-03 3.80056649e-01 5.39051652e-01
3.18734407e-01 -9.83941630e-02 -3.30524832e-01 1.46156430e+00
-1.60222620e-01 -1.41579404e-01 -8.01554263e-01 1.81343809e-01
4.31804717e-01 9.29947495e-01 -7.93684602e-01 -4.90236372e-01
8.16263631e-02 6.16414607e-01 4.17850792e-01 4.99206096e-01
-8.59360874e-01 -3.59569460e-01 8.92275572e-01 -9.60048512e-02
1.91399097e-01 3.69536161e-01 -8.09054196e-01 -1.09868789e+00
4.58875783e-02 -8.35424840e-01 3.43032479e-01 -6.01779819e-01
-1.15509796e+00 5.95355213e-01 9.46115181e-02 -1.15471435e+00
-2.57590830e-01 -5.44273496e-01 -9.38597322e-01 9.10792470e-01
-1.23143709e+00 -5.65723121e-01 2.94855330e-02 2.78030813e-01
4.16964889e-01 -2.13041142e-01 5.98479569e-01 -3.08376521e-01
-5.44575691e-01 6.05196118e-01 -1.67701468e-02 1.92849234e-01
1.55369237e-01 -1.22723234e+00 1.29693240e-01 6.91517174e-01
3.01162720e-01 8.07277679e-01 7.99891412e-01 -4.72917885e-01
-6.90110862e-01 -1.03679454e+00 9.68875885e-01 -2.15940848e-01
3.92992765e-01 -2.14462787e-01 -1.14087260e+00 8.36599112e-01
-9.45911631e-02 -8.82346481e-02 5.85870445e-01 2.23562717e-01
-3.15524638e-01 -3.46882641e-01 -1.24988258e+00 8.89466882e-01
8.47153544e-01 -1.90698639e-01 -9.07019377e-01 -2.62348577e-02
6.54284954e-01 -3.63029353e-02 -2.38444462e-01 4.69220400e-01
7.26240218e-01 -1.00275421e+00 6.21982276e-01 -1.14178574e+00
7.42190599e-01 -3.43704224e-01 -1.00201480e-01 -1.84155452e+00
-1.17890283e-01 -1.61515489e-01 2.02290267e-01 1.03505898e+00
7.86353588e-01 -1.04078543e+00 8.61233115e-01 9.90514517e-01
8.90846476e-02 -9.61170793e-01 -1.02336192e+00 -4.63459849e-01
-6.79292455e-02 -3.17580551e-01 1.24237800e+00 1.03333747e+00
-1.27959996e-01 2.50584453e-01 -1.55369282e-01 4.28279042e-01
6.28960788e-01 7.36580431e-01 7.10468471e-01 -1.32121682e+00
-3.27329814e-01 -4.93277192e-01 -1.92829102e-01 -8.87980700e-01
3.67596745e-01 -8.91729176e-01 -8.72094743e-03 -1.20565200e+00
3.76221269e-01 -4.19652849e-01 -5.37395954e-01 6.52358472e-01
-4.98194218e-01 -2.91533440e-01 1.38248429e-01 4.87278216e-02
-9.00793672e-02 7.60451019e-01 1.28719330e+00 -1.71773896e-01
-3.44086409e-01 4.49632436e-01 -1.28152072e+00 9.46117222e-01
1.00396299e+00 -7.94050634e-01 -5.42208612e-01 -3.06197017e-01
1.09411888e-01 -2.30427925e-02 5.64776957e-01 -6.66101038e-01
-6.44862503e-02 -6.77853405e-01 7.89585650e-01 -2.46652082e-01
-2.93206163e-02 -1.12907684e+00 2.49081343e-01 8.45931947e-01
-9.06628966e-01 -1.10219633e-02 3.29819694e-02 6.36548281e-01
-4.17077281e-02 -3.19899619e-01 6.94564641e-01 -6.19770288e-02
-2.90160090e-01 6.63976446e-02 -2.00543523e-01 -2.46528074e-01
8.14764917e-01 -1.82316124e-01 -3.92759562e-01 -4.82741207e-01
-6.49748862e-01 1.89729735e-01 2.54369467e-01 5.28830051e-01
6.77733719e-01 -1.37612057e+00 -4.77034181e-01 5.49252868e-01
9.58087146e-02 3.17897461e-02 9.01186466e-02 3.29526186e-01
1.43035188e-01 4.70107138e-01 -1.97227806e-01 -4.85021800e-01
-9.31967080e-01 4.15959448e-01 6.25043690e-01 -2.44188026e-01
-3.57655555e-01 6.09671175e-01 7.48866737e-01 -5.03488660e-01
-2.91767363e-02 -5.93792379e-01 -2.72894114e-01 -1.12552896e-01
3.59113485e-01 -8.46666843e-03 -2.11596802e-01 -1.54020935e-01
-1.70374721e-01 -4.42454815e-02 -1.29306465e-01 1.16689354e-01
1.56542146e+00 6.72926307e-02 8.38384107e-02 5.52280366e-01
1.13316917e+00 -2.68165678e-01 -1.45647812e+00 -1.73262045e-01
7.08080605e-02 -7.62779593e-01 -2.06762210e-01 -1.03676617e+00
-1.28807807e+00 7.67355382e-01 4.76229429e-01 3.76123160e-01
1.03536141e+00 -3.50189917e-02 2.54485279e-01 4.68469471e-01
1.05749756e-01 -8.18588078e-01 -1.43132836e-01 1.73348218e-01
9.71707642e-01 -1.29247200e+00 -3.22208732e-01 -8.33248943e-02
-7.05037653e-01 1.09159541e+00 8.44469547e-01 -1.17694937e-01
6.16052806e-01 -3.05356264e-01 8.32039937e-02 -3.12713683e-01
-8.91779184e-01 9.86146033e-02 4.93205488e-01 3.66995394e-01
1.56026155e-01 2.81878352e-01 -1.75363109e-01 9.58743930e-01
-4.90998328e-01 -2.50949800e-01 4.02732760e-01 2.52362102e-01
-3.12652230e-01 -7.49278724e-01 -2.72494197e-01 8.90888214e-01
-3.52266312e-01 -1.06176427e-02 -4.58972692e-01 9.46006894e-01
1.37059420e-01 6.76184773e-01 1.77796751e-01 -3.89742494e-01
4.07877445e-01 2.51045674e-01 -3.42843793e-02 -7.69663155e-01
-2.71151930e-01 -4.72057253e-01 -1.96807921e-01 -4.23536956e-01
-1.71407878e-01 -6.58038020e-01 -1.39672494e+00 -1.44618321e-02
-3.38273436e-01 3.21814537e-01 6.06492162e-01 1.41319978e+00
2.94528902e-01 5.64292967e-01 8.63476992e-01 -6.32266760e-01
-8.99970055e-01 -7.63545871e-01 -4.79762256e-01 5.08008003e-01
6.74757838e-01 -8.71070623e-01 -7.51013100e-01 -2.09167480e-01]
|
[8.727127075195312, 5.611660480499268]
|
e4e45793-2591-4292-a5df-1a660c2e74b9
|
facedirector-continuous-control-of-facial
| null | null |
http://openaccess.thecvf.com/content_iccv_2015/html/Malleson_FaceDirector_Continuous_Control_ICCV_2015_paper.html
|
http://openaccess.thecvf.com/content_iccv_2015/papers/Malleson_FaceDirector_Continuous_Control_ICCV_2015_paper.pdf
|
FaceDirector: Continuous Control of Facial Performance in Video
|
We present a method to continuously blend between multiple facial performances of an actor, which can contain different facial expressions or emotional states. As an example, given sad and angry video takes of a scene, our method empowers the movie director to specify arbitrary weighted combinations and smooth transitions between the two takes in post-production. Our contributions include (1) a robust nonlinear audio-visual synchronization technique that exploits complementary properties of audio and visual cues to automatically determine robust, dense spatiotemporal correspondences between takes, and (2) a seamless facial blending approach that provides the director full control to interpolate timing, facial expression, and local appearance, in order to generate novel performances after filming. In contrast to most previous works, our approach operates entirely in image space, avoiding the need of 3D facial reconstruction. We demonstrate that our method can synthesize visually believable performances with applications in emotion transition, performance correction, and timing control.
|
['Alexander Sorkine-Hornung', 'Jean-Charles Bazin', 'Charles Malleson', 'Adrian Hilton', 'Thabo Beeler', 'Oliver Wang', 'Derek Bradley']
|
2015-12-01
| null | null | null |
iccv-2015-12
|
['audio-visual-synchronization', 'audio-visual-synchronization']
|
['audio', 'computer-vision']
|
[ 2.62287498e-01 -8.19579735e-02 7.15274587e-02 -3.25817436e-01
-5.42919219e-01 -8.12979758e-01 5.33073783e-01 -5.95303774e-02
2.66063541e-01 1.25549793e-01 1.69370905e-01 3.65213573e-01
2.17294574e-01 -3.05537760e-01 -6.28727734e-01 -3.10687810e-01
-1.67703498e-02 -2.03698725e-01 -1.38493665e-02 -3.75029027e-01
1.90278351e-01 8.14213276e-01 -1.84564912e+00 5.07867694e-01
5.10429204e-01 1.10469258e+00 -3.58639210e-01 1.02772593e+00
1.90189183e-01 1.10914338e+00 -6.31280422e-01 -4.93797481e-01
3.19345504e-01 -6.89817965e-01 -3.22261691e-01 5.95290422e-01
7.91899264e-01 -2.97381818e-01 -1.34286508e-01 8.18503916e-01
4.27890152e-01 2.79728115e-01 4.26565260e-01 -1.34608960e+00
-6.11565351e-01 1.67955399e-01 -7.03845441e-01 -4.71026301e-01
1.04069924e+00 2.72547722e-01 6.10439360e-01 -8.23599875e-01
8.59014869e-01 1.18650913e+00 6.73803866e-01 6.20283544e-01
-1.51557672e+00 -7.11046934e-01 9.53790992e-02 2.81260032e-02
-1.56215966e+00 -9.36975479e-01 1.21596348e+00 -5.33656597e-01
2.42530495e-01 5.82260370e-01 1.43003178e+00 1.01940167e+00
-7.83730820e-02 5.07277310e-01 9.23677921e-01 -5.87048590e-01
9.36286896e-03 7.65417367e-02 -7.05592632e-01 7.33313441e-01
-8.76820445e-01 9.42193717e-02 -9.51099873e-01 -2.29959369e-01
1.10689485e+00 -3.15158099e-01 -3.26014698e-01 -2.55514711e-01
-1.27393484e+00 2.87432343e-01 -8.15570578e-02 2.41764396e-01
-4.17044371e-01 3.87588739e-01 2.10563779e-01 3.76617402e-01
3.62078339e-01 6.31318271e-01 7.07673952e-02 -3.74025822e-01
-1.27317536e+00 5.80395043e-01 5.12896776e-01 7.67433763e-01
6.65209055e-01 4.63895738e-01 -3.31420690e-01 8.23509276e-01
-8.91543180e-02 5.24408340e-01 -9.81432758e-03 -1.77607238e+00
-2.87850022e-01 1.89012140e-01 3.83437157e-01 -1.49012613e+00
-6.36182204e-02 2.02812403e-01 -3.14823389e-01 5.34656167e-01
2.54450470e-01 2.88387090e-02 -3.63340408e-01 1.95143008e+00
6.34916544e-01 4.70672280e-01 -3.96605283e-01 9.99857903e-01
7.68413603e-01 5.87549150e-01 -2.32413217e-01 -5.44816196e-01
1.07761228e+00 -6.34096384e-01 -1.14382923e+00 2.67079979e-01
9.97922570e-02 -1.05457592e+00 1.38569784e+00 6.11023307e-01
-1.59760582e+00 -6.56792283e-01 -9.39076424e-01 -1.27445862e-01
2.54215330e-01 3.31345409e-01 6.61031425e-01 4.13100988e-01
-1.25547731e+00 7.20443070e-01 -6.32496595e-01 -4.30064760e-02
-8.31581727e-02 2.51826197e-01 -4.68840241e-01 3.58783841e-01
-7.77590513e-01 4.95212376e-01 -2.51258194e-01 1.21142520e-02
-7.99673915e-01 -8.94467175e-01 -9.77643073e-01 -1.52661800e-01
2.10458666e-01 -5.20267546e-01 1.47465050e+00 -1.90623856e+00
-2.41658258e+00 1.10695899e+00 -1.72395453e-01 2.15529576e-01
4.44153011e-01 -8.95027891e-02 -4.16794151e-01 5.26181042e-01
-1.35730147e-01 1.04858828e+00 1.15702367e+00 -1.39209580e+00
-1.44989029e-01 -1.09326109e-01 3.04624755e-02 3.69725138e-01
-3.43587637e-01 4.53943878e-01 -7.12109447e-01 -8.92513633e-01
-4.72607603e-03 -8.57395053e-01 9.02902335e-03 6.87559485e-01
-2.47832716e-01 2.56989866e-01 8.58135998e-01 -6.75848424e-01
1.19751680e+00 -2.48077464e+00 5.04893959e-01 3.34868133e-01
1.27725765e-01 -7.09273517e-02 -1.54717073e-01 3.25035632e-01
-1.78483292e-01 -3.66622597e-01 1.65397093e-01 -5.76677203e-01
-1.28526807e-01 -8.16408247e-02 -2.25286424e-01 5.29530168e-01
1.37456000e-01 6.19787991e-01 -8.16018522e-01 -6.72228575e-01
2.37417892e-01 8.50396037e-01 -6.59096122e-01 4.49680865e-01
-1.30103812e-01 8.49815965e-01 -4.21454757e-02 8.53512645e-01
3.91640037e-01 3.06587547e-01 2.15131089e-01 -3.09532672e-01
-2.52754956e-01 -7.99964666e-02 -1.17981422e+00 1.92152131e+00
-3.64453644e-01 7.31431961e-01 5.95115662e-01 -3.22128296e-01
1.15364587e+00 6.48049176e-01 7.55289197e-01 -4.96751547e-01
1.31811842e-01 -1.25637785e-01 -5.82712293e-01 -7.27184236e-01
6.81642294e-01 -2.37035140e-01 -2.39533782e-01 4.01960880e-01
-1.75244287e-01 -8.64821732e-01 -5.19408733e-02 1.03572190e-01
7.77330041e-01 5.04661858e-01 1.61200479e-01 1.12976216e-01
4.17853355e-01 -3.96271586e-01 5.91172814e-01 -1.05394600e-02
-1.87413782e-01 8.06468785e-01 7.25898981e-01 -3.61631334e-01
-1.00954318e+00 -1.02432036e+00 4.85421658e-01 1.12448478e+00
1.42677605e-01 -6.37216568e-01 -8.79685760e-01 -8.13194737e-03
-2.85967171e-01 4.15767550e-01 -7.19302773e-01 -1.64908497e-03
-7.08793819e-01 2.89144427e-01 5.46852469e-01 1.99653342e-01
-1.53799094e-02 -8.65420163e-01 -6.79392815e-01 2.48486712e-03
-4.10190791e-01 -1.09120846e+00 -1.09923804e+00 -4.46331233e-01
-5.71222842e-01 -8.50894749e-01 -6.03192508e-01 -6.85137987e-01
7.78279603e-01 1.46229491e-01 9.50022876e-01 -1.18629495e-02
-4.69721735e-01 7.62754381e-01 -1.62385330e-01 6.32881839e-03
-5.94675779e-01 -5.82811773e-01 9.81953517e-02 5.88394821e-01
-4.40600604e-01 -1.02467656e+00 -4.70507801e-01 2.80488729e-01
-8.58593822e-01 5.33736587e-01 -2.89270788e-01 4.17963386e-01
6.34173155e-01 -4.69052464e-01 2.88717180e-01 -3.72560412e-01
6.76705539e-01 6.72018752e-02 -4.90071118e-01 2.21103251e-01
7.02805668e-02 -4.10412252e-01 5.62177062e-01 -8.90046954e-01
-1.02800953e+00 5.01837015e-01 1.52819887e-01 -9.82654214e-01
-4.46974346e-03 -7.09121227e-02 -1.00707747e-02 -3.34180504e-01
7.43651211e-01 -1.45255491e-01 2.33542740e-01 2.47960072e-02
6.33258641e-01 3.25695932e-01 1.08032560e+00 -8.78886640e-01
7.36071348e-01 5.52368462e-01 -2.85793487e-02 -8.34014535e-01
-4.02432770e-01 2.29001790e-03 -7.84937680e-01 -9.96901453e-01
6.91474974e-01 -8.27243686e-01 -1.12921643e+00 2.89131582e-01
-1.28042829e+00 -5.16656697e-01 -5.14384091e-01 3.64082336e-01
-9.11712408e-01 2.92601675e-01 -7.51739681e-01 -7.68601358e-01
-8.62256810e-03 -1.03478575e+00 1.18353832e+00 2.29557201e-01
-8.00136209e-01 -6.12036705e-01 1.40321746e-01 3.02348375e-01
1.76185623e-01 9.53611612e-01 4.84062850e-01 5.83362937e-01
-3.38436574e-01 -1.68048069e-01 2.40254968e-01 1.55482635e-01
2.25723118e-01 1.01227248e+00 -8.42879236e-01 -5.18519804e-02
-2.68019646e-01 -4.81263429e-01 -8.17079470e-03 3.33341509e-01
9.76251483e-01 -5.63683808e-01 1.88320100e-01 7.71194935e-01
8.56689930e-01 3.01789463e-01 5.93375444e-01 -1.51121482e-01
4.86506134e-01 9.09882665e-01 6.75759077e-01 8.06985855e-01
2.26698726e-01 1.19148982e+00 2.55396962e-01 -2.69419521e-01
-2.54208952e-01 -4.05779600e-01 7.32595444e-01 8.11827302e-01
-4.45622683e-01 2.88165957e-01 -2.76517510e-01 3.70293766e-01
-1.74775100e+00 -1.26521683e+00 8.62353668e-02 2.05591893e+00
1.00896966e+00 -5.82999587e-01 3.07190686e-01 2.14005098e-01
7.48153329e-01 2.27054551e-01 -2.57260710e-01 -8.78614008e-01
-1.03490233e-01 3.83488238e-01 -2.27931395e-01 5.62943637e-01
-8.98855090e-01 9.20729578e-01 7.16900444e+00 6.73932910e-01
-1.38720393e+00 -2.70797163e-02 5.00259042e-01 -6.05002344e-01
-5.92787325e-01 -2.22220104e-02 2.21154820e-02 1.02955893e-01
3.96633029e-01 -1.42723486e-01 6.71571791e-01 5.86782217e-01
7.08302259e-01 -4.27354574e-02 -1.10253239e+00 1.11009395e+00
4.33799714e-01 -1.33756447e+00 -1.66441575e-01 -3.07830155e-01
8.65561306e-01 -1.10427368e+00 4.29466456e-01 -2.26729482e-01
-4.78399582e-02 -9.86762941e-01 1.31906831e+00 7.08818436e-01
1.16453242e+00 -9.00312781e-01 -2.98190564e-01 -6.05434291e-02
-1.21352005e+00 1.49181321e-01 3.86025459e-01 -5.22907525e-02
3.36223066e-01 1.55038849e-01 -4.64463592e-01 7.92976320e-02
6.57709360e-01 5.67526340e-01 -2.98983604e-01 5.00946760e-01
-4.92287993e-01 9.81240645e-02 -1.44338816e-01 2.87422359e-01
-2.86505431e-01 -1.85317874e-01 6.92057192e-01 1.10788488e+00
4.61820304e-01 2.59131372e-01 1.46338716e-01 9.95763719e-01
1.78948212e-02 3.86895567e-01 -7.16456413e-01 -1.76074021e-02
3.91472042e-01 1.35899448e+00 -4.70759004e-01 -1.87532738e-01
2.48378571e-02 1.36445785e+00 9.53727812e-02 3.46830875e-01
-1.15829694e+00 -2.45914891e-01 8.20084453e-01 4.89713371e-01
-3.08927484e-02 -4.86287922e-01 -1.79855034e-01 -1.09329295e+00
1.07585359e-03 -1.05961561e+00 -1.37993842e-01 -1.27024138e+00
-5.41945100e-01 6.23289108e-01 -1.36226311e-01 -1.33809936e+00
-4.14535731e-01 6.42930269e-02 -6.45636618e-01 3.55200827e-01
-9.08529043e-01 -1.36323214e+00 -4.85744268e-01 8.54309082e-01
4.26990807e-01 1.10151373e-01 8.46962392e-01 3.76500517e-01
-3.28208447e-01 6.32803619e-01 -6.09410882e-01 -2.67296195e-01
1.09411299e+00 -6.75197124e-01 -8.05009753e-02 6.88391089e-01
1.98833838e-01 3.08225095e-01 9.21530604e-01 -2.70705819e-01
-1.33129275e+00 -7.13954628e-01 6.24222636e-01 -2.62642860e-01
3.03405255e-01 -3.24255705e-01 -5.75377584e-01 5.44055462e-01
3.31782073e-01 3.26704867e-02 8.20577621e-01 -2.98236817e-01
-5.01799047e-01 -2.85698622e-01 -1.10231149e+00 9.50656712e-01
7.51652718e-01 -6.79816425e-01 -2.39550117e-02 1.12224966e-01
5.66724360e-01 -6.98666394e-01 -9.12958026e-01 2.85779923e-01
1.02442181e+00 -1.33702552e+00 9.09318268e-01 -1.59880489e-01
6.30548656e-01 -5.99779308e-01 -9.97180641e-02 -1.10463119e+00
-1.34593517e-01 -1.45715344e+00 -8.44890065e-03 1.27525055e+00
-4.39922437e-02 1.30598158e-01 5.24941325e-01 7.70934880e-01
-1.41320214e-01 -5.61209500e-01 -7.02416956e-01 -2.84299910e-01
-6.55659258e-01 -5.47530711e-01 5.04103720e-01 1.10695648e+00
3.16018969e-01 -1.28288865e-01 -8.54026973e-01 -3.22652906e-02
2.36733034e-01 3.03080171e-01 1.13518882e+00 -8.51360440e-01
-4.58741724e-01 -4.77452099e-01 -3.54227901e-01 -8.30598891e-01
3.19193661e-01 -4.37802136e-01 3.71422209e-02 -8.18489432e-01
-1.88348070e-01 -2.65646458e-01 2.67280012e-01 5.02504647e-01
2.09266573e-01 7.28545785e-01 4.78255451e-01 1.34437710e-01
-5.21513104e-01 7.42534757e-01 1.52370775e+00 1.04823656e-01
-5.51746130e-01 -2.31031999e-01 -4.41169739e-01 8.15952182e-01
3.36137682e-01 -2.83625036e-01 -3.80343318e-01 -2.22493052e-01
3.27791303e-01 7.51160324e-01 5.00119209e-01 -9.37798142e-01
2.56686192e-02 -5.31742096e-01 3.05679083e-01 -1.69868842e-01
8.94289792e-01 -6.04821265e-01 7.95294821e-01 3.56770903e-02
-5.72757840e-01 1.34242728e-01 2.09415779e-01 9.41576809e-02
-3.82689804e-01 1.50283933e-01 1.00339127e+00 5.56590082e-03
-3.20534647e-01 6.09551556e-02 -6.02742374e-01 -3.81949872e-01
1.18841934e+00 -4.95832413e-01 3.53997558e-01 -1.03815818e+00
-9.45328593e-01 -2.01049685e-01 9.40474391e-01 3.42093915e-01
7.25519061e-01 -1.79924917e+00 -5.19180417e-01 3.60844195e-01
-2.80414727e-02 -3.84730279e-01 5.26660860e-01 8.31530929e-01
-8.09255302e-01 -4.22370493e-01 -3.83340418e-01 -6.19331717e-01
-1.73639143e+00 2.89138079e-01 2.84656495e-01 3.61256331e-01
-3.57846588e-01 8.00168753e-01 1.44443110e-01 -1.26361698e-01
1.79426476e-01 -4.29502763e-02 4.25258055e-02 1.08242109e-01
6.07352376e-01 2.29324192e-01 -2.12330773e-01 -8.86599958e-01
-4.30032834e-02 9.25471485e-01 5.05215943e-01 -6.21389508e-01
1.06092274e+00 -2.25881264e-01 -2.09669262e-01 7.05680430e-01
9.75087821e-01 5.61650038e-01 -1.55437613e+00 1.34001419e-01
-5.71669877e-01 -8.69690359e-01 -2.60693759e-01 -3.50822657e-01
-1.18396008e+00 7.49767482e-01 2.07160413e-01 -6.62722513e-02
1.51837063e+00 -2.20408812e-01 8.33827138e-01 -3.74998115e-02
2.00691164e-01 -1.22548997e+00 6.71063781e-01 1.42653883e-01
1.18491352e+00 -5.63460410e-01 1.86751485e-02 -5.14704823e-01
-9.08644795e-01 1.31667793e+00 6.07799351e-01 -5.67963868e-02
2.64023006e-01 5.93477249e-01 3.01321000e-01 -1.25513583e-01
-8.59548509e-01 3.21011752e-01 2.85785705e-01 6.94925129e-01
6.29787326e-01 1.02781001e-02 1.01147424e-02 3.24730754e-01
-2.89090008e-01 6.37859181e-02 6.29911840e-01 7.19688892e-01
-2.95137819e-02 -1.01868808e+00 -6.36322200e-01 -4.12114143e-01
-4.74087268e-01 2.45656192e-01 -5.25164306e-01 4.55372483e-01
3.15681845e-01 8.33572745e-01 1.26009420e-01 -5.32963812e-01
5.12067676e-01 -5.68065094e-03 8.88227582e-01 -4.44985121e-01
-9.58617568e-01 3.98438603e-01 4.36929837e-02 -1.04460514e+00
-6.63921833e-01 -6.88064575e-01 -1.07983470e+00 -4.30238873e-01
-8.19022432e-02 -1.70229420e-01 5.73878884e-01 3.89642626e-01
4.77577865e-01 3.91211301e-01 1.11703742e+00 -1.43256795e+00
1.64188638e-01 -4.43892866e-01 -6.67636633e-01 6.27060413e-01
2.78771400e-01 -5.00394106e-01 -1.32455662e-01 7.41117477e-01]
|
[12.994586944580078, -0.44305336475372314]
|
11d47bd4-cbe5-4548-8089-8ea5171be8fe
|
lightx3ecg-a-lightweight-and-explainable-deep
|
2207.12381
| null |
https://arxiv.org/abs/2207.12381v1
|
https://arxiv.org/pdf/2207.12381v1.pdf
|
LightX3ECG: A Lightweight and eXplainable Deep Learning System for 3-lead Electrocardiogram Classification
|
Cardiovascular diseases (CVDs) are a group of heart and blood vessel disorders that is one of the most serious dangers to human health, and the number of such patients is still growing. Early and accurate detection plays a key role in successful treatment and intervention. Electrocardiogram (ECG) is the gold standard for identifying a variety of cardiovascular abnormalities. In clinical practices and most of the current research, standard 12-lead ECG is mainly used. However, using a lower number of leads can make ECG more prevalent as it can be conveniently recorded by portable or wearable devices. In this research, we develop a novel deep learning system to accurately identify multiple cardiovascular abnormalities by using only three ECG leads.
|
['Cuong D. Do', 'Tien N. Thanh', 'Tu A. Nguyen', 'Thao BT. Nguyen', 'Hieu H. Pham', 'Khiem H. Le']
|
2022-07-25
| null | null | null | null |
['ecg-classification']
|
['medical']
|
[-3.05045489e-02 -6.62842989e-01 -2.78265029e-01 -2.82719672e-01
-2.33075753e-01 -3.22126567e-01 -3.13271016e-01 5.23231506e-01
-1.92066237e-01 7.02810705e-01 -2.87104696e-01 -5.75635374e-01
-7.85053000e-02 -8.90692770e-01 2.04692334e-01 -5.38326144e-01
-2.13394180e-01 3.23235989e-01 1.46811903e-01 1.08933359e-01
-8.71862099e-02 7.12789714e-01 -8.02950859e-01 -1.05181538e-01
7.55199313e-01 1.29695868e+00 -2.78479040e-01 6.40569150e-01
1.43842757e-01 2.34275162e-01 -7.48918295e-01 -1.33038357e-01
2.46946737e-01 -8.81405830e-01 -3.61553609e-01 -1.81242615e-01
-1.40789822e-01 -6.24360204e-01 -4.09587473e-02 8.38513255e-01
1.13541329e+00 -3.01301271e-01 3.41349483e-01 -5.79133153e-01
-2.71465778e-01 1.66843459e-01 -6.07222319e-01 5.86636007e-01
-1.21697240e-01 4.39539133e-03 2.50228375e-01 -4.46055382e-01
-2.12812617e-01 7.23754048e-01 8.31836522e-01 4.05830234e-01
-9.85149562e-01 -6.61695421e-01 -4.28027213e-01 2.88432032e-01
-1.39659238e+00 -2.61817068e-01 8.12571108e-01 -3.55597258e-01
2.94245481e-01 3.04101110e-01 1.16320169e+00 6.13965333e-01
5.88314354e-01 1.59795612e-01 9.62450624e-01 -3.73411834e-01
6.66886196e-03 7.67777041e-02 1.35770231e-01 3.24223965e-01
6.70022428e-01 2.63724532e-02 1.03460640e-01 -2.06957296e-01
1.28448558e+00 5.96672773e-01 -2.05491111e-01 9.96608213e-02
-1.00087154e+00 6.31860495e-01 1.79279968e-01 4.37080711e-01
-4.02627707e-01 -1.86330661e-01 6.64017797e-01 2.12484732e-01
1.73903838e-01 3.46112520e-01 -4.17433947e-01 -3.23598295e-01
-7.63020933e-01 -3.63834351e-02 4.41816062e-01 1.25345767e-01
1.17237307e-01 7.35903159e-02 -1.28221944e-01 7.42023349e-01
3.66779566e-01 4.27254230e-01 5.89170218e-01 -9.54324722e-01
-2.28027608e-02 9.15468693e-01 3.96445468e-02 -1.24459684e+00
-5.85625887e-01 -9.10930336e-01 -1.54006290e+00 1.34943157e-01
5.22157311e-01 -3.28720212e-01 -5.18720865e-01 1.18370366e+00
3.44500601e-01 2.76183873e-01 -4.31762189e-01 1.03665423e+00
9.27036643e-01 2.95673072e-01 2.67174065e-01 -4.38241154e-01
1.44762719e+00 -2.27943897e-01 -6.88874841e-01 -4.67791148e-02
2.85177588e-01 -6.04758382e-01 7.26827085e-01 4.95577097e-01
-6.81784868e-01 -6.62301719e-01 -1.16655195e+00 4.87927757e-02
-1.45603800e-02 3.45562100e-01 6.62944436e-01 1.04079342e+00
-5.70541024e-01 5.42142034e-01 -9.53397036e-01 -3.17927450e-01
5.70745647e-01 1.01247400e-01 2.58167759e-02 1.28551111e-01
-1.32699561e+00 8.11079621e-01 1.92284212e-02 3.18720669e-01
-3.21780413e-01 -5.44839323e-01 -4.63995188e-01 -3.67497504e-02
-1.23378579e-02 -6.83200836e-01 8.36397290e-01 -4.05690372e-01
-1.33903372e+00 9.79925036e-01 -1.80376306e-01 -3.28912199e-01
5.02249837e-01 -5.10268033e-01 -5.93509018e-01 3.47695231e-01
-4.01567705e-02 -6.83411062e-02 6.53623819e-01 -5.06167829e-01
-3.14468056e-01 -6.08826697e-01 -3.76913220e-01 -7.58303329e-02
-2.50995785e-01 4.07006741e-01 4.47229343e-03 -4.45203215e-01
3.37555856e-01 -6.63453162e-01 -4.94415492e-01 2.68675327e-01
-2.18894616e-01 -2.38396376e-01 6.50961220e-01 -1.01208007e+00
1.40725660e+00 -2.13978004e+00 -2.22039267e-01 2.92028904e-01
7.35396504e-01 8.23173344e-01 6.19474590e-01 4.83020879e-02
-5.12586161e-02 4.66941714e-01 7.59783164e-02 2.68030405e-01
-6.52337849e-01 1.37508795e-01 8.49294215e-02 4.34306502e-01
-7.06871673e-02 8.49708915e-01 -6.34390175e-01 -5.26630104e-01
5.47469735e-01 5.78636765e-01 9.64235142e-02 9.51287150e-02
5.24703860e-01 1.09949350e+00 -7.32094944e-01 5.44123471e-01
4.03089792e-01 -4.57717389e-01 2.23199219e-01 -1.14386685e-01
4.31391485e-02 9.80081707e-02 -1.19909203e+00 1.21059299e+00
5.68237640e-02 4.43988174e-01 -3.20596755e-01 -1.27402902e+00
1.12248826e+00 7.65374422e-01 6.44205093e-01 -6.20315492e-01
3.57386053e-01 9.73801389e-02 3.46113771e-01 -7.10446894e-01
-7.27487385e-01 -1.41008273e-01 3.15449566e-01 1.84944674e-01
-6.83790684e-01 4.64822769e-01 1.56485721e-01 -1.65043652e-01
7.07865179e-01 -3.41899961e-01 9.85805809e-01 -2.48429179e-02
4.99394923e-01 -3.93502653e-01 1.17443287e+00 6.44098699e-01
-6.30257249e-01 6.11567914e-01 4.02553409e-01 -1.18763816e+00
-6.14458919e-01 -9.56853867e-01 -4.89620060e-01 1.37316585e-01
8.28194395e-02 -2.48991773e-01 -2.79231310e-01 -3.78246367e-01
-2.19317265e-02 -1.12709112e-01 -2.54968554e-01 -2.93760747e-02
-7.12985337e-01 -9.83438313e-01 7.43762314e-01 7.62038827e-01
8.83076310e-01 -8.04016531e-01 -1.21154821e+00 4.64299053e-01
-1.57228604e-01 -6.48444951e-01 -4.54220176e-02 -1.91983446e-01
-1.24063158e+00 -1.27558410e+00 -1.04013407e+00 -6.13464832e-01
3.17050576e-01 2.00428441e-02 9.70202327e-01 5.24109662e-01
-7.88120985e-01 -3.09556037e-01 -2.32414380e-01 -7.84174323e-01
-2.74114609e-01 -2.05923356e-02 1.04575329e-01 1.27048865e-01
4.09848243e-01 -7.14695573e-01 -1.13865650e+00 5.82524426e-02
-1.79228842e-01 -7.52191320e-02 6.78696990e-01 5.34714937e-01
6.82491064e-01 2.65127987e-01 8.64053547e-01 -8.16471100e-01
6.51648283e-01 -1.28521234e-01 -3.79846603e-01 1.00008562e-01
-8.29933882e-01 -6.79574490e-01 6.05789483e-01 -1.16700426e-01
-5.76031327e-01 -1.83026925e-01 -3.15771788e-01 -7.93346316e-02
-7.56438792e-01 5.88384748e-01 -1.48124605e-01 -1.08713910e-01
6.62855685e-01 1.20632388e-01 2.04444909e-03 -6.58092260e-01
-5.01622796e-01 6.94620371e-01 3.35566968e-01 -2.73383319e-01
2.47298867e-01 4.35083658e-02 5.22953749e-01 -1.09734452e+00
-7.07737565e-01 -5.63883722e-01 -5.84270537e-01 -3.94316643e-01
8.60512733e-01 -6.30362093e-01 -8.02902102e-01 5.06301463e-01
-1.05349779e+00 2.12050483e-01 6.89337403e-02 6.85136378e-01
3.66864681e-01 6.55712426e-01 -7.63116419e-01 -7.91378140e-01
-8.88158679e-01 -9.75260437e-01 7.16116190e-01 4.96503443e-01
-4.45982009e-01 -9.46654379e-01 1.60819903e-01 -2.76733302e-02
5.59142768e-01 8.65879655e-01 1.05110884e+00 -4.36494827e-01
-5.15406691e-02 -6.15926087e-01 -2.58250117e-01 5.27658820e-01
5.03566265e-01 -1.29303470e-01 -6.06578410e-01 -1.04480699e-01
3.17529678e-01 1.02736168e-01 4.52262938e-01 6.55555725e-01
1.22596204e+00 1.85792014e-01 -3.72998029e-01 7.02863157e-01
1.23854327e+00 9.33890700e-01 7.37946630e-01 1.24425150e-01
6.81647122e-01 1.28979357e-02 4.07267362e-02 3.64170879e-01
1.99141175e-01 2.51082242e-01 7.17054382e-02 -5.74904859e-01
6.93109855e-02 4.34631497e-01 -2.45827660e-01 7.55947292e-01
-6.09520853e-01 2.60031998e-01 -1.07206702e+00 1.79738432e-01
-1.48313749e+00 -9.78336930e-01 -6.59916997e-01 2.37585616e+00
8.91117215e-01 -2.96079107e-02 3.45673144e-01 6.03643417e-01
7.89051950e-01 -3.95183593e-01 -7.06772089e-01 -1.04778230e-01
3.65848765e-02 5.99287927e-01 5.22768721e-02 -3.18478107e-01
-1.17642927e+00 -1.11542486e-01 7.09283590e+00 -1.45473391e-01
-1.67801940e+00 -3.15830670e-02 9.86552477e-01 2.97944576e-01
5.34656644e-01 -3.53094965e-01 -5.33137798e-01 6.37190342e-01
5.90167820e-01 -4.60239351e-02 -6.57505095e-02 7.02338159e-01
6.20375931e-01 -6.17116615e-02 -8.94957483e-01 1.24876416e+00
-3.11788648e-01 -1.03006363e+00 -3.20790857e-01 -2.75189523e-02
2.02387229e-01 -4.83190686e-01 -7.70867690e-02 -2.09319480e-02
-5.57702839e-01 -9.39924419e-01 -2.13617682e-01 4.44543481e-01
1.04371071e+00 -5.45103312e-01 1.07188272e+00 2.89172649e-01
-1.08925903e+00 5.18824905e-02 -2.32843027e-01 -4.21873659e-01
2.70935893e-01 1.13748658e+00 -3.74870032e-01 4.03276533e-01
9.84894216e-01 8.81099999e-01 -5.67318141e-01 1.58399248e+00
-2.90087789e-01 1.08981013e+00 -3.39193374e-01 1.94653079e-01
-5.20073116e-01 -3.08327705e-01 4.69237030e-01 7.15664268e-01
3.92427087e-01 3.40403169e-01 6.63663268e-01 5.34021378e-01
1.83347195e-01 2.78722048e-01 -3.99058372e-01 2.36597016e-01
3.23022693e-01 1.20511520e+00 -9.58549678e-01 -4.50783759e-01
-5.42282760e-01 6.00408018e-01 -3.07156384e-01 2.30467483e-01
-7.62152374e-01 -6.43246412e-01 3.52103472e-01 3.78733128e-01
-4.27023113e-01 -1.11909859e-01 -8.39736700e-01 -1.16360319e+00
-6.92689568e-02 -8.95079553e-01 6.19242728e-01 -6.10695034e-02
-1.13256156e+00 2.05016106e-01 -3.00579160e-01 -1.33308947e+00
3.66423093e-02 -3.18951935e-01 -9.65110719e-01 1.26166558e+00
-1.30160356e+00 -7.18185365e-01 -5.43007731e-01 3.86307299e-01
3.16931903e-01 3.07163261e-02 1.37457216e+00 6.68302715e-01
-1.03721917e+00 5.63313544e-01 -1.87594905e-01 5.29823005e-01
8.37356150e-01 -1.15722191e+00 1.37499303e-01 7.96361148e-01
-3.69128436e-01 9.48105633e-01 2.78103620e-01 -5.59236050e-01
-8.49235117e-01 -9.97467160e-01 7.76346743e-01 -7.79315904e-02
-9.72481444e-02 3.71236295e-01 -8.18336070e-01 3.68111759e-01
-5.62433787e-02 2.68131822e-01 1.06410897e+00 7.31872097e-02
2.75679260e-01 -6.25316262e-01 -1.06342208e+00 3.55870008e-01
4.44422066e-01 -1.20955080e-01 -3.30562890e-01 8.58509764e-02
-9.72863510e-02 -4.99414682e-01 -1.16655040e+00 5.35852849e-01
9.69494045e-01 -9.99362946e-01 9.37155187e-01 -3.71944577e-01
1.81142718e-01 -4.66072351e-01 5.23184836e-01 -1.15390933e+00
-4.98314679e-01 -5.19464552e-01 -2.50893950e-01 1.12118876e+00
-9.07548890e-03 -8.76166821e-01 7.73746133e-01 5.52556515e-01
1.59205824e-01 -8.97251606e-01 -6.40406191e-01 -5.02904654e-01
-1.95393294e-01 -5.77543788e-02 4.72001433e-01 8.86942923e-01
-9.81917605e-02 4.05035943e-01 -4.47394580e-01 -8.50802734e-02
9.00270283e-01 6.20029978e-02 4.84363914e-01 -1.81352234e+00
-2.50645161e-01 -3.14309716e-01 -5.86203218e-01 -5.09910762e-01
-7.78827906e-01 -5.57681978e-01 -5.44311881e-01 -1.72612011e+00
6.82688653e-02 -3.66376191e-01 -6.42467260e-01 4.64974761e-01
-3.84886324e-01 4.38612193e-01 -3.73467416e-01 6.99838698e-02
-6.62689060e-02 -9.75972861e-02 1.24959981e+00 6.75216094e-02
-5.08106589e-01 4.13768947e-01 -8.12141240e-01 8.66161644e-01
1.22491980e+00 -3.80476147e-01 -2.95050085e-01 -2.83342391e-01
1.24477386e-01 2.80102670e-01 2.44628638e-01 -1.13850641e+00
-7.06054328e-04 -5.46417683e-02 1.20663834e+00 -3.56283098e-01
-8.75358731e-02 -6.50783181e-01 3.17765862e-01 9.15680051e-01
-8.61814320e-02 2.11356983e-01 6.01479374e-02 1.32021457e-01
-1.60300225e-01 1.97725803e-01 9.69089866e-01 -4.11330193e-01
-4.15698498e-01 3.37769270e-01 -5.83711803e-01 -3.74650396e-02
1.12196422e+00 -2.53424615e-01 8.03857818e-02 -2.36750931e-01
-8.02224457e-01 7.91006163e-02 -1.51082620e-01 7.04917610e-02
8.82167101e-01 -1.20707190e+00 -8.48827839e-01 2.96030343e-01
-1.03115126e-01 1.09547228e-01 3.71874571e-01 1.30720115e+00
-9.76729333e-01 3.03568900e-01 -3.42371881e-01 -6.52036011e-01
-1.45511103e+00 9.40934271e-02 6.03389740e-01 -1.22372605e-01
-1.07490182e+00 4.76615846e-01 -2.41122380e-01 3.87347877e-01
3.33851278e-01 -1.08412564e-01 -5.86581707e-01 -2.70875275e-01
8.84538770e-01 8.39885890e-01 4.95439246e-02 -2.15602741e-01
-4.70035791e-01 7.08247244e-01 3.86368454e-04 4.30780798e-01
1.04964066e+00 -1.76044792e-01 -3.14305246e-01 4.21762228e-01
6.85156822e-01 -1.35965869e-01 -4.34603572e-01 1.57184348e-01
-3.38695794e-01 -3.47561777e-01 2.66667306e-01 -7.76273370e-01
-1.21195042e+00 1.37802041e+00 1.16422498e+00 3.92996520e-01
1.15614700e+00 -3.88073921e-01 1.14806378e+00 1.78319052e-01
3.80090684e-01 -4.93378073e-01 -1.89377159e-01 -6.44779354e-02
4.31235403e-01 -1.10476160e+00 7.49074817e-02 -3.77711833e-01
-2.69022614e-01 1.35572565e+00 4.92651075e-01 -5.66512644e-02
1.03163218e+00 4.82594594e-02 6.24582827e-01 -7.78560489e-02
2.05477122e-02 1.27649590e-01 7.17870742e-02 5.96272647e-01
8.72936964e-01 9.19259787e-02 -7.51335442e-01 8.36636603e-01
3.51771295e-01 2.91853994e-01 6.99541792e-02 8.12633157e-01
-4.20682669e-01 -1.32576942e+00 -4.16561306e-01 9.76994336e-01
-1.19958210e+00 1.43291742e-01 -2.52614111e-01 3.43859106e-01
3.49549979e-01 1.13790810e+00 -4.41601485e-01 -1.12135582e-01
3.09697151e-01 1.71283334e-01 4.03474659e-01 -5.97837269e-01
-4.20755774e-01 2.21664011e-01 -2.17021167e-01 -3.57820630e-01
-3.20117623e-01 -5.26885748e-01 -1.06423450e+00 -1.97616786e-01
-7.45353326e-02 -3.46689634e-02 3.80624771e-01 1.09087729e+00
4.74686772e-01 7.48125315e-01 5.03127098e-01 -2.72423118e-01
-5.34259737e-01 -9.05950010e-01 -9.43852365e-01 3.12392324e-01
2.22099215e-01 -5.16421080e-01 3.96868140e-02 3.59342813e-01]
|
[14.258171081542969, 3.207213878631592]
|
8eb3796a-4960-4097-b10b-9813acdd11b6
|
you-might-think-about-slightly-revising-the-2
|
2306.14911
| null |
https://arxiv.org/abs/2306.14911v1
|
https://arxiv.org/pdf/2306.14911v1.pdf
|
"You might think about slightly revising the title": identifying hedges in peer-tutoring interactions
|
Hedges play an important role in the management of conversational interaction. In peer tutoring, they are notably used by tutors in dyads (pairs of interlocutors) experiencing low rapport to tone down the impact of instructions and negative feedback. Pursuing the objective of building a tutoring agent that manages rapport with students in order to improve learning, we used a multimodal peer-tutoring dataset to construct a computational framework for identifying hedges. We compared approaches relying on pre-trained resources with others that integrate insights from the social science literature. Our best performance involved a hybrid approach that outperforms the existing baseline while being easier to interpret. We employ a model explainability tool to explore the features that characterize hedges in peer-tutoring conversations, and we identify some novel features, and the benefits of such a hybrid model approach.
|
['Justine Cassell', 'Chloé Clavel', 'Yann Raphalen']
|
2023-06-18
| null | null | null | null |
['management']
|
['miscellaneous']
|
[ 2.36151926e-02 9.73810077e-01 -1.72002703e-01 -3.27668011e-01
-4.86109018e-01 -5.55812418e-01 7.07530856e-01 5.12265980e-01
1.35094374e-01 6.65723205e-01 8.48961353e-01 -4.27142471e-01
-3.86599869e-01 -6.22218490e-01 -4.67974275e-01 -2.94754386e-01
5.25788814e-02 4.41648334e-01 1.60254434e-01 -6.87888026e-01
6.65506423e-01 7.04498291e-02 -1.85930645e+00 5.70743978e-01
1.29050028e+00 8.03627148e-02 1.42694041e-01 7.62729228e-01
-2.76260465e-01 1.56514609e+00 -7.50486374e-01 -4.22212988e-01
-4.19991851e-01 -6.22145176e-01 -1.07038379e+00 -1.27926216e-01
5.35803735e-01 -1.10229380e-01 -8.65583122e-02 5.37256598e-01
5.06617129e-01 2.82056689e-01 6.83621585e-01 -1.56639731e+00
-2.85326481e-01 1.07585680e+00 -1.34646565e-01 8.27449486e-02
9.96435463e-01 7.67556727e-02 1.06606352e+00 -1.05627440e-01
7.41461039e-01 1.55642235e+00 8.76415491e-01 7.60539711e-01
-1.33659184e+00 -5.31033039e-01 -6.47193789e-02 4.04822677e-01
-2.09063098e-01 -4.55269128e-01 7.18463838e-01 -6.21886730e-01
6.79030657e-01 3.73491347e-01 8.67504239e-01 1.52343285e+00
-1.23103805e-01 1.00733423e+00 1.18821156e+00 -5.78513980e-01
-1.78423449e-01 5.86692810e-01 6.44003630e-01 6.38304234e-01
-3.91001791e-01 -2.72389978e-01 -1.26951635e+00 -4.40230250e-01
1.75666884e-01 -3.64009708e-01 -5.64675748e-01 -4.02810097e-01
-1.13968205e+00 1.13330650e+00 2.40694046e-01 4.32700753e-01
3.70373726e-02 -3.90681237e-01 3.07752579e-01 9.41620231e-01
1.31751984e-01 1.22954345e+00 -3.56438398e-01 -1.02197838e+00
-2.31033951e-01 2.81341970e-01 1.58897889e+00 6.87021732e-01
6.04640007e-01 -6.77609980e-01 -4.01518166e-01 1.12793136e+00
3.68027806e-01 -2.91373074e-01 6.68030083e-01 -1.40373611e+00
5.19571304e-01 1.20506608e+00 -1.37589961e-01 -6.28616452e-01
-3.59204382e-01 3.63772064e-02 -3.51838656e-02 -1.63345143e-01
5.85137248e-01 -4.48004246e-01 -7.23778531e-02 1.61705947e+00
5.84341645e-01 4.28989053e-01 1.92217410e-01 2.72547811e-01
1.26208174e+00 1.05757408e-01 -3.82563770e-02 -1.24331586e-01
1.10152233e+00 -1.37610388e+00 -8.92174959e-01 2.45946363e-01
1.27615952e+00 -9.66220081e-01 1.09695840e+00 3.43640268e-01
-9.73101735e-01 -1.15365624e-01 -8.64033699e-01 -3.18386108e-02
-4.28565890e-01 -6.23405576e-01 3.51454347e-01 6.62263930e-01
-1.05387640e+00 1.03879857e+00 -2.94268638e-01 -4.57892299e-01
5.85837290e-02 2.88937479e-01 -3.20456117e-01 3.95596236e-01
-1.19566834e+00 1.20161295e+00 -2.39931852e-01 -4.42669332e-01
-5.15631199e-01 -1.26946306e+00 -8.92152786e-01 1.56636760e-01
3.15542698e-01 -5.91649830e-01 1.72686493e+00 -7.32263923e-01
-1.97447944e+00 8.26107979e-01 3.46030779e-02 -3.30882251e-01
7.23207355e-01 -2.76226103e-01 4.26346004e-01 3.60290036e-02
-9.20403283e-03 3.93147945e-01 3.21012497e-01 -1.22753549e+00
-3.49120378e-01 -2.44894773e-01 2.14249730e-01 4.95789468e-01
-5.40684581e-01 -1.17673501e-01 2.64555991e-01 -1.42452061e-01
-2.14481488e-01 -9.70793128e-01 7.65247494e-02 -7.86748171e-01
-3.60443294e-01 -7.75967240e-01 1.03091085e+00 -2.97944039e-01
1.25990438e+00 -1.70541453e+00 3.20370793e-01 2.06533641e-01
6.55340672e-01 3.80597144e-01 -1.42624512e-01 9.76902306e-01
-1.90469399e-02 1.99889839e-01 2.15092465e-01 -4.45460737e-01
2.77092129e-01 -1.17881997e-02 -2.98827261e-01 -2.37306431e-01
6.50576279e-02 6.64699793e-01 -1.17398965e+00 -4.98735815e-01
2.71247149e-01 3.68716806e-01 -5.07578492e-01 1.20655894e+00
-1.57294080e-01 5.92227399e-01 -3.61572802e-01 3.03270668e-01
-6.99089393e-02 2.08959579e-02 3.18530500e-01 5.44016063e-01
-1.03627935e-01 9.04570699e-01 -5.38358510e-01 9.97255921e-01
-4.64254081e-01 9.90784943e-01 2.05846444e-01 -7.18411803e-01
9.83044624e-01 4.82316673e-01 4.09387290e-01 -2.85824060e-01
1.45853646e-02 -2.28860825e-01 3.91185820e-01 -8.84832025e-01
4.57897514e-01 3.96042913e-01 2.55647749e-01 1.06277323e+00
1.35396659e-01 -3.73857081e-01 -7.32002929e-02 4.54961389e-01
1.40662670e+00 -6.54896647e-02 3.28146130e-01 -1.14044860e-01
4.16884333e-01 -1.91421464e-01 1.17711313e-01 8.48797798e-01
-5.73107362e-01 3.09577137e-01 1.32471740e+00 -1.19735137e-01
-3.80200833e-01 -3.78481418e-01 -1.59614757e-02 1.63708103e+00
-3.60062927e-01 -6.49251997e-01 -1.04450560e+00 -8.84287596e-01
2.11046576e-01 7.59506762e-01 -7.41364241e-01 -2.26550922e-01
-4.51302379e-01 -7.64385983e-02 6.44360960e-01 2.38367300e-02
1.62098587e-01 -1.18551636e+00 -5.13194203e-01 2.85331041e-01
-2.07462132e-01 -7.20621645e-01 -2.00097591e-01 4.73860890e-01
-5.78521609e-01 -1.53428590e+00 -1.46899059e-01 -6.04914904e-01
3.05450886e-01 4.69506204e-01 1.36121929e+00 6.39492452e-01
7.33049139e-02 9.83290195e-01 -3.17533106e-01 -6.63161933e-01
-7.55173981e-01 4.13468570e-01 -9.35570747e-02 -6.37505412e-01
6.41124249e-01 -8.45635116e-01 -1.40589640e-01 6.13252342e-01
-4.77125555e-01 2.04139978e-01 1.95670143e-01 9.81000483e-01
-6.89645588e-01 -6.48933947e-01 9.01002586e-01 -1.41584539e+00
1.24991524e+00 -7.84781039e-01 -6.91705346e-02 3.29801619e-01
-6.39095664e-01 -4.09115897e-03 2.59570122e-01 -5.12641788e-01
-1.05839777e+00 -4.54205334e-01 1.03155889e-01 -1.53662965e-01
-2.92240828e-01 5.07071972e-01 1.45146959e-02 -5.46436787e-01
6.41849935e-01 -5.37606418e-01 4.72533047e-01 -4.63156343e-01
1.34876028e-01 8.44381690e-01 -8.92067403e-02 -8.41325104e-01
4.31795448e-01 -4.91794258e-01 -3.47054809e-01 -9.68399584e-01
-1.09716332e+00 -5.78355968e-01 -6.13762736e-01 -6.72159851e-01
4.89237338e-01 -7.47388899e-01 -1.47241974e+00 1.56388536e-01
-9.72162485e-01 -7.87073135e-01 1.48047090e-01 3.42256039e-01
-7.25590408e-01 1.77514270e-01 -6.61525846e-01 -9.17015612e-01
7.10402653e-02 -1.06036353e+00 5.49681365e-01 5.78525066e-01
-1.01078868e+00 -1.33863544e+00 6.30079091e-01 1.31598759e+00
4.67469692e-01 6.50852546e-02 1.12658167e+00 -1.33182395e+00
-2.57527024e-01 4.46900845e-01 3.18579853e-01 -5.41695990e-02
-2.15575501e-01 2.87901521e-01 -1.29335654e+00 2.52066344e-01
1.20030463e-01 -8.53624523e-01 5.62116325e-01 -9.78468731e-02
7.32737660e-01 -5.19115746e-01 -9.35359523e-02 3.01923417e-02
4.27102089e-01 -9.49790031e-02 3.68403256e-01 7.56497204e-01
7.55958855e-01 1.58306813e+00 7.48273253e-01 1.05191424e-01
9.12412763e-01 5.19029498e-01 3.60404730e-01 4.99703616e-01
2.04334840e-01 -5.43286204e-01 7.63276398e-01 1.12434840e+00
1.13287419e-01 9.48754922e-02 -1.22018337e+00 6.01164281e-01
-2.21742320e+00 -7.05530405e-01 -3.27549487e-01 1.84361207e+00
1.16168320e+00 -3.69874537e-02 2.63532609e-01 -7.85215273e-02
5.60707867e-01 1.98883146e-01 -1.61218971e-01 -8.73261392e-01
3.10583800e-01 -9.31731239e-02 -1.39161080e-01 9.27865446e-01
-6.73918486e-01 7.75777400e-01 6.27228308e+00 3.23002219e-01
-5.40925145e-01 -3.70785534e-01 4.28481340e-01 1.03249632e-01
-4.17518169e-01 -1.14794284e-01 -6.40581667e-01 1.04244269e-01
1.45600259e+00 -2.07009524e-01 3.38590205e-01 7.39546478e-01
2.88631201e-01 -2.06985906e-01 -1.57152975e+00 3.64822179e-01
9.78677869e-02 -1.05541933e+00 -3.81535262e-01 6.00209050e-02
7.65671074e-01 -2.39003256e-01 -1.08919322e-01 5.92653334e-01
8.12337458e-01 -1.18750143e+00 -1.45930484e-01 6.28404319e-01
-6.07479990e-01 -5.82152128e-01 8.11303735e-01 4.43827003e-01
-4.93281364e-01 -4.71161455e-02 1.46666139e-01 -4.54680622e-01
-5.48142731e-01 -1.93900451e-01 -1.56206131e+00 3.14504765e-02
5.80760062e-01 7.83623576e-01 -6.93634391e-01 8.85213315e-01
-3.97846371e-01 8.84605885e-01 6.03666250e-03 -4.09040928e-01
9.81887951e-02 -4.43359971e-01 7.34192014e-01 9.34800446e-01
3.22950296e-02 1.26787156e-01 -6.74338406e-03 6.81933880e-01
-2.09494919e-01 1.96447819e-01 -1.09619844e+00 -1.26631036e-01
7.34317124e-01 1.28297925e+00 -5.54334698e-03 -2.09432527e-01
-2.05267787e-01 2.33658403e-01 7.09512889e-01 1.58010036e-01
-2.32722923e-01 -1.12239242e-01 7.94645607e-01 -5.31627983e-03
-6.89645335e-02 2.11145490e-01 -3.79526615e-01 -8.82440388e-01
-2.92118430e-01 -1.45334172e+00 3.03610295e-01 -7.54481256e-01
-1.26319051e+00 3.94605547e-02 -1.67837620e-01 -9.17702913e-01
-7.44360864e-01 -1.92414939e-01 -1.18549085e+00 4.93458420e-01
-1.38691819e+00 -8.34987760e-01 -3.93697053e-01 3.27158213e-01
2.83561766e-01 -3.58466595e-01 9.97762263e-01 -3.28904420e-01
-7.08630800e-01 5.55062413e-01 -1.84842676e-01 -2.91722119e-01
1.23571885e+00 -1.93268788e+00 -2.16390088e-01 -1.63028285e-01
-1.89802051e-01 8.17510784e-01 1.17789948e+00 -4.40520078e-01
-1.48750019e+00 -4.92273837e-01 1.20852304e+00 -9.06633258e-01
1.22296846e+00 -1.77057341e-01 -1.39116395e+00 6.63176179e-01
9.61933255e-01 -1.03094947e+00 1.40301609e+00 7.20046163e-01
-2.58014530e-01 4.44515139e-01 -1.00708270e+00 6.23991668e-01
8.31793845e-01 -5.96415639e-01 -1.33065939e+00 3.62521470e-01
7.43750811e-01 -4.96199667e-01 -1.34883225e+00 -5.15454449e-02
4.79246706e-01 -1.54994440e+00 4.89001095e-01 -6.56619906e-01
8.49888682e-01 4.66840059e-01 6.44069552e-01 -1.67232788e+00
4.54361141e-01 -1.22959161e+00 -2.10257009e-01 1.61996460e+00
3.35773408e-01 -6.69555843e-01 8.66473496e-01 1.12208569e+00
-1.32779092e-01 -8.10530543e-01 -5.02042234e-01 -9.00503770e-02
1.07211031e-01 1.06175162e-01 4.82366234e-01 1.49814665e+00
8.87715578e-01 8.23863328e-01 -4.81712669e-02 -3.36681038e-01
4.02156591e-01 -1.77789420e-01 1.31464767e+00 -1.70390046e+00
-2.49935295e-02 -6.21811450e-01 -1.85145006e-01 -5.80165267e-01
6.92310572e-01 -6.00335598e-01 1.82067137e-02 -1.11032689e+00
1.24375917e-01 -2.91783094e-01 7.76134804e-02 5.14302969e-01
-3.88512015e-01 -3.82787138e-01 2.15520903e-01 4.42115357e-03
-6.92701817e-01 7.33855486e-01 1.18675852e+00 2.73643821e-01
-6.14723206e-01 1.96222126e-01 -8.55015934e-01 9.47890222e-01
6.83684111e-01 -3.31235647e-01 -3.26222897e-01 1.51007891e-01
6.30211085e-02 2.56286711e-01 -1.88520439e-02 -4.89171594e-01
4.75012124e-01 -1.90635249e-01 -2.36008897e-01 -2.19809815e-01
8.55572298e-02 -6.90189123e-01 -4.22286272e-01 1.16830580e-01
-9.26950097e-01 -2.14667246e-01 -3.15247551e-02 1.47805274e-01
-5.24084568e-01 -5.24242699e-01 3.99192542e-01 -1.54698091e-02
1.42317742e-01 -4.95598197e-01 -7.04813659e-01 -1.64909326e-02
1.20068288e+00 2.88758781e-02 -8.68816972e-01 -1.06758082e+00
-4.79335725e-01 8.48739088e-01 4.27755564e-01 7.96569169e-01
3.58272403e-01 -1.02273989e+00 -5.24230480e-01 -2.62958743e-02
6.89108819e-02 -1.95076704e-01 6.12916201e-02 9.30770397e-01
-1.96316868e-01 4.51212436e-01 -2.19683975e-01 -4.94511276e-01
-1.93277407e+00 -3.13617647e-01 4.01802421e-01 -4.88241315e-01
-3.67550939e-01 6.72911942e-01 -1.61095768e-01 -1.14227438e+00
7.90723205e-01 -1.46923080e-01 -1.08008766e+00 7.93795764e-01
5.75733125e-01 6.73385322e-01 -1.16796292e-01 -2.61149108e-01
4.08869654e-01 1.15025826e-02 -1.30643427e-01 4.40209955e-02
1.63949800e+00 -1.01830333e-01 -3.08888167e-01 8.91601205e-01
1.07559896e+00 1.29203573e-01 -8.49037051e-01 -1.94936320e-01
6.12669110e-01 -3.49071562e-01 -3.39654595e-01 -6.97767258e-01
-2.08745107e-01 7.59049892e-01 1.14367679e-01 9.03909385e-01
2.31169537e-01 -2.03260668e-02 2.82610089e-01 8.43305588e-01
-3.13103795e-01 -1.06878865e+00 3.75557214e-01 9.33729827e-01
8.37344646e-01 -1.71112084e+00 -1.73005894e-01 -4.89011109e-01
-8.06244612e-01 1.27968657e+00 1.13380384e+00 2.87494004e-01
4.23117787e-01 -3.27730849e-02 3.85453790e-01 -3.66026431e-01
-1.65683234e+00 -2.36713942e-02 3.94251466e-01 5.27545810e-01
1.11139894e+00 1.29756154e-02 -2.52694309e-01 4.83495653e-01
-7.53171682e-01 -2.97792196e-01 9.45294976e-01 9.73611355e-01
-6.39480233e-01 -1.23227453e+00 -3.56435746e-01 4.81923819e-01
-2.19390959e-01 5.42504825e-02 -1.49260664e+00 6.99481487e-01
-4.51195657e-01 1.37988734e+00 -3.20478588e-01 -4.65637535e-01
4.88613397e-01 5.00854611e-01 1.13815643e-01 -8.40825260e-01
-1.66504121e+00 -4.24053818e-01 6.50068402e-01 -6.66223228e-01
-6.63573921e-01 -8.56308699e-01 -7.87516534e-01 -4.50329959e-01
-3.52656871e-01 6.88086271e-01 4.54897940e-01 1.08004725e+00
1.79247901e-01 4.61857438e-01 8.19322288e-01 -6.03748977e-01
-5.05546510e-01 -1.46222520e+00 -2.14653194e-01 3.22035044e-01
4.31719691e-01 -6.76173568e-01 -8.10716510e-01 -3.59092206e-01]
|
[12.220253944396973, 8.051685333251953]
|
c4c61969-2a3a-40e7-a793-6939353d1f3e
|
egocentric-affordance-detection-with-the-one
|
1906.05794
| null |
https://arxiv.org/abs/1906.05794v1
|
https://arxiv.org/pdf/1906.05794v1.pdf
|
Egocentric affordance detection with the one-shot geometry-driven Interaction Tensor
|
In this abstract we describe recent [4,7] and latest work on the determination of affordances in visually perceived 3D scenes. Our method builds on the hypothesis that geometry on its own provides enough information to enable the detection of significant interaction possibilities in the environment. The motivation behind this is that geometric information is intimately related to the physical interactions afforded by objects in the world. The approach uses a generic representation for the interaction between everyday objects such as a mug or an umbrella with the environment, and also for more complex affordances such as humans Sitting or Riding a motorcycle. Experiments with synthetic and real RGB-D scenes show that the representation enables the prediction of affordance candidate locations in novel environments at fast rates and from a single (one-shot) training example. The determination of affordances is a crucial step towards systems that need to perceive and interact with their surroundings. We here illustrate output on two cases for a simulated robot and for an Augmented Reality setting, both perceiving in an egocentric manner.
|
['Walterio Mayol-Cuevas', 'Eduardo Ruiz']
|
2019-06-13
| null | null | null | null |
['affordance-detection']
|
['computer-vision']
|
[ 2.76973993e-01 2.35229999e-01 4.55605000e-01 -4.21359062e-01
5.94170019e-03 -6.51723981e-01 7.27510750e-01 2.64129937e-01
-3.85100812e-01 4.63613570e-01 1.85520351e-01 -3.72023553e-01
-2.98922598e-01 -5.01063764e-01 -5.38913488e-01 -2.99366683e-01
-6.06360316e-01 4.47366416e-01 4.66346741e-01 -5.84717512e-01
4.83916014e-01 1.14944959e+00 -2.23828292e+00 4.84893434e-02
3.01361799e-01 6.58545792e-01 7.86273539e-01 9.68926728e-01
3.21570516e-01 3.18682492e-01 -4.00047660e-01 2.23148257e-01
4.41064984e-01 -1.63816139e-01 -6.50061905e-01 4.29747581e-01
2.94933915e-01 -2.27181897e-01 -2.29960203e-01 7.93797374e-01
4.19443905e-01 6.80480063e-01 7.04616964e-01 -1.14898384e+00
-1.98015779e-01 5.53041101e-02 -1.15697771e-01 1.62835002e-01
1.39661920e+00 2.74690509e-01 7.46700048e-01 -9.89555895e-01
8.54452074e-01 1.35989869e+00 2.18409419e-01 3.54171157e-01
-1.15409601e+00 2.48578757e-01 2.22387329e-01 1.14687018e-01
-1.39986134e+00 -6.07492328e-01 4.71539557e-01 -3.78073543e-01
1.44489110e+00 7.97338009e-01 9.11761522e-01 6.72958493e-01
5.30033074e-02 6.90088749e-01 8.09647799e-01 -1.00170279e+00
3.67155105e-01 6.45722225e-02 -7.17102140e-02 6.11058116e-01
8.24440941e-02 3.17640692e-01 -2.86376655e-01 -7.35117719e-02
1.26543748e+00 -4.03735088e-03 -4.17264223e-01 -1.17126572e+00
-1.64395273e+00 1.39665231e-01 7.57650614e-01 3.85158002e-01
-2.48534322e-01 -9.16640684e-02 -9.82836559e-02 2.72027165e-01
3.87694500e-03 7.51566231e-01 -3.18313003e-01 -2.41632819e-01
8.40253904e-02 4.76697773e-01 6.64131045e-01 1.27032530e+00
5.12474835e-01 -4.90324169e-01 4.70948637e-01 2.27575272e-01
3.14657092e-01 1.02527127e-01 1.14627862e-02 -9.70896602e-01
4.02245432e-01 5.94099224e-01 6.94138348e-01 -9.11935151e-01
-7.79951215e-01 2.07684577e-01 -4.61912341e-02 8.70362043e-01
6.76569462e-01 2.29001492e-01 -7.83125043e-01 1.26053536e+00
8.20662260e-01 -1.93197086e-01 1.62942894e-02 1.21273077e+00
5.39992034e-01 1.73991278e-01 -1.11814752e-01 9.26159099e-02
1.47619307e+00 -6.19236112e-01 -4.91680682e-01 -2.46669158e-01
7.31067479e-01 -7.56186426e-01 1.25640023e+00 4.44857717e-01
-7.88925171e-01 -7.70098150e-01 -1.16188836e+00 -3.81589919e-01
-7.18421996e-01 -3.57288346e-02 9.47244704e-01 4.51374382e-01
-8.79704714e-01 7.47370541e-01 -7.84818947e-01 -9.81506228e-01
-2.68428028e-01 3.45166445e-01 -7.49654233e-01 8.95847678e-02
-5.03176928e-01 1.27347994e+00 5.66623926e-01 2.85380363e-01
-3.56116146e-01 1.02417156e-01 -1.17408836e+00 -1.78555340e-01
4.50287759e-01 -7.90859461e-01 1.12035346e+00 -6.37713969e-01
-1.33247995e+00 9.43442643e-01 -3.25380452e-02 3.25217992e-02
6.12297177e-01 -3.82957816e-01 -3.70446920e-01 -9.30121541e-03
3.42789106e-02 5.22736907e-01 4.75493044e-01 -1.47355163e+00
-5.02863765e-01 -7.31696367e-01 5.73835671e-01 8.19050550e-01
7.37079024e-01 1.62455125e-03 -4.20407504e-02 -3.69071096e-01
9.63801920e-01 -1.01350427e+00 -5.10822892e-01 2.75577515e-01
-5.22403121e-01 -2.28953004e-01 6.17156982e-01 -2.23716661e-01
4.31970507e-01 -2.23270369e+00 2.41620019e-01 3.28225166e-01
5.99343665e-02 -2.05894396e-01 1.82044730e-01 7.49704063e-01
-2.41803020e-01 -2.95880914e-01 1.39744028e-01 -2.98548937e-01
2.68085986e-01 2.27475554e-01 -8.65499228e-02 5.96015155e-01
-1.00811586e-01 6.89322889e-01 -1.22903383e+00 -2.35496089e-01
7.11775362e-01 5.37093699e-01 -3.03028017e-01 3.44636381e-01
3.12242080e-02 8.56574714e-01 -5.08317232e-01 4.69240129e-01
3.66928756e-01 2.23410487e-01 4.17215899e-02 -2.01184098e-02
-3.94251227e-01 5.20253420e-01 -1.63468075e+00 2.11870909e+00
-1.73155084e-01 4.85253334e-01 -1.41430333e-01 -3.61080855e-01
8.78708303e-01 2.72763610e-01 -8.01328644e-02 -5.05328059e-01
1.90480664e-01 1.72289982e-01 6.45253286e-02 -9.76594687e-01
8.10475111e-01 1.44908940e-02 -1.24117240e-01 4.58121955e-01
-1.46164611e-01 -6.17208421e-01 6.10441044e-02 -1.38639688e-01
7.86506295e-01 8.68400812e-01 9.21558082e-01 -1.59846678e-01
1.26292229e-01 -8.24866965e-02 -1.40553012e-01 6.31943464e-01
-2.84201533e-01 7.46074975e-01 -1.03539400e-01 -7.83870161e-01
-1.00981724e+00 -1.33913851e+00 -7.85024166e-02 8.21430683e-01
6.80073857e-01 -2.32412979e-01 -3.65787625e-01 -2.64300346e-01
-1.44898981e-01 9.32637095e-01 -6.09303892e-01 1.10823147e-01
-4.21151668e-01 -1.15901075e-01 -3.95626247e-01 3.76017928e-01
6.44412031e-03 -1.14832294e+00 -1.83079946e+00 -1.12812861e-03
8.42939839e-02 -1.09661806e+00 -4.14280258e-02 2.03592032e-01
-9.56506908e-01 -9.48491633e-01 -4.28400785e-01 -6.92141116e-01
1.03993905e+00 4.82263207e-01 9.42188084e-01 -1.24883108e-01
-5.36635041e-01 6.81221664e-01 -4.83015150e-01 -3.98820430e-01
-1.45790234e-01 -5.34683347e-01 4.16595519e-01 -1.80005968e-01
3.20777237e-01 -7.18975902e-01 -6.04070842e-01 4.21807438e-01
-4.59490061e-01 3.76969904e-01 2.44662344e-01 2.66917199e-01
3.73370796e-01 -1.20046243e-01 -2.04867736e-01 -1.88323036e-01
4.35390294e-01 -1.76564574e-01 -2.11113855e-01 1.31396756e-01
1.44979611e-01 -8.40651393e-02 -7.27818683e-02 -4.93910521e-01
-9.92814124e-01 5.93961120e-01 2.35505924e-01 -1.09682279e-02
-8.66011202e-01 8.70575756e-02 -4.10382211e-01 -8.66843462e-02
1.00962746e+00 -2.80862540e-01 -4.52371836e-01 -6.29659593e-01
6.10238910e-01 4.48877811e-01 5.86447716e-01 -6.57896519e-01
6.23996675e-01 6.78018928e-01 1.68694362e-01 -9.85570431e-01
-1.24557801e-01 -5.75709105e-01 -1.59406471e+00 -3.75096977e-01
6.99712276e-01 -5.84971428e-01 -9.95832443e-01 -9.52203870e-02
-1.31356955e+00 -2.46673346e-01 -5.49762189e-01 8.80217910e-01
-9.86691415e-01 1.75225124e-01 1.20256148e-01 -1.25257659e+00
5.94569743e-01 -1.18384969e+00 1.33273458e+00 1.78465888e-01
-6.85490370e-01 -7.07640827e-01 -3.24626751e-02 -9.77711305e-02
-2.07519054e-01 8.81707251e-01 5.98409235e-01 -2.62962073e-01
-7.18042135e-01 -3.17690760e-01 -6.10980392e-02 -4.71673608e-01
3.99810940e-01 -2.28279442e-01 -1.03009951e+00 -1.10352755e-01
-1.03545105e-02 1.33384010e-02 1.45731762e-01 -1.54712265e-02
4.91467148e-01 7.43277669e-02 -5.23089230e-01 1.64553687e-01
1.36591494e+00 5.10184407e-01 6.91974044e-01 4.54415381e-01
7.44067430e-01 1.04170239e+00 8.55691969e-01 3.75272989e-01
4.43016678e-01 8.99885356e-01 6.85823977e-01 -1.25278970e-02
2.60842174e-01 -2.47077182e-01 -1.96898058e-01 9.12495777e-02
-6.22472405e-01 8.90094563e-02 -9.77078676e-01 3.38974535e-01
-1.82403314e+00 -6.40237629e-01 -2.64876544e-01 2.40791512e+00
1.21995270e-01 2.38940746e-01 9.03025493e-02 3.96299601e-01
3.27029049e-01 -2.75938421e-01 -6.43964708e-01 -5.21070004e-01
1.23633727e-01 -2.06496686e-01 1.62583411e-01 5.55231571e-01
-9.34269190e-01 7.41887689e-01 6.53031349e+00 -3.25987907e-03
-5.91821194e-01 -3.57441276e-01 5.29068224e-02 1.19590014e-02
1.31178007e-01 1.25506952e-01 -2.06797212e-01 -2.51735061e-01
3.41784120e-01 1.47836491e-01 5.49269438e-01 8.60889256e-01
2.12038562e-01 -8.75642955e-01 -1.65727019e+00 1.06927204e+00
4.28937450e-02 -7.12075055e-01 -2.88028270e-01 2.46134084e-02
1.93189278e-01 -1.79732963e-01 -1.48182064e-01 -1.85900807e-01
1.12542041e-01 -9.48981047e-01 1.00768960e+00 7.91157722e-01
5.14902771e-01 -4.80727941e-01 2.28791609e-01 6.77826226e-01
-1.13235402e+00 -1.06252797e-01 -3.09421003e-01 -7.75404096e-01
3.35785210e-01 -1.30422488e-01 -1.24178493e+00 4.14629012e-01
6.25060081e-01 2.51956761e-01 -5.31287193e-01 1.41080356e+00
-3.03193927e-01 -6.52292609e-01 -6.11591220e-01 -2.02921510e-01
8.82297754e-02 -2.04040512e-01 8.60464990e-01 9.37268436e-01
3.72672439e-01 5.18384814e-01 2.20178459e-02 6.46101713e-01
7.75076270e-01 2.59092748e-02 -1.18323767e+00 5.24870753e-01
1.23569049e-01 9.53441858e-01 -1.20141554e+00 -1.84148803e-01
-9.03935954e-02 1.15395522e+00 1.80490732e-01 5.59524655e-01
-3.01603496e-01 -5.44936717e-01 7.19475448e-01 1.17149733e-01
1.85987368e-01 -8.64659429e-01 -9.12773013e-02 -1.12300491e+00
2.85119444e-01 -5.20366907e-01 -5.16020581e-02 -1.55847096e+00
-5.49333453e-01 7.14830220e-01 3.51541489e-01 -1.64445448e+00
-3.68865758e-01 -8.59010518e-01 -3.21221560e-01 9.83371496e-01
-9.40882683e-01 -1.02048361e+00 -5.07502317e-01 4.58533138e-01
5.83386898e-01 3.08403999e-01 1.20763397e+00 -4.35910225e-01
1.96626440e-01 -3.10604066e-01 -2.51933903e-01 -3.85021091e-01
2.65541166e-01 -1.55577743e+00 8.63784730e-01 6.79651916e-01
5.53620994e-01 8.72068644e-01 1.14916730e+00 -3.52643669e-01
-1.46227443e+00 -9.43892077e-02 8.23935926e-01 -1.19372833e+00
2.59104699e-01 -8.08695793e-01 -6.12928331e-01 8.29954684e-01
-1.45254478e-01 1.68673769e-02 3.83423686e-01 4.75595668e-02
-2.07904547e-01 4.55518425e-01 -1.38686609e+00 1.10947943e+00
1.68750489e+00 -5.95668852e-01 -1.07923293e+00 2.91634619e-01
6.52292788e-01 -9.50178385e-01 -4.72587287e-01 3.01175803e-01
9.10644591e-01 -1.41732550e+00 1.24282491e+00 -5.86448491e-01
-3.59492272e-01 -6.16557360e-01 -4.14852023e-01 -1.32553828e+00
-1.86943069e-01 -7.58081496e-01 -1.40613420e-02 2.31635407e-01
-1.04284761e-02 -5.07627785e-01 4.16256130e-01 8.79865229e-01
-3.90824452e-02 -2.64379710e-01 -1.05398440e+00 -6.58106029e-01
-7.79093981e-01 -7.18963742e-01 9.18646395e-01 6.87681079e-01
5.24659574e-01 2.73493886e-01 1.97562099e-01 5.16785443e-01
4.71224695e-01 -4.79833735e-03 1.10241711e+00 -1.38240266e+00
-2.59767860e-01 -2.88878500e-01 -1.01203680e+00 -1.26110137e+00
-2.18164861e-01 -4.00439769e-01 1.76348954e-01 -1.55836093e+00
-5.23320377e-01 -5.76598346e-01 -2.17838474e-02 6.09742217e-02
1.36230990e-01 1.43963834e-02 2.81744689e-01 1.85373813e-01
-4.40741718e-01 2.69627869e-01 1.56270897e+00 3.67931038e-01
-4.78493750e-01 2.86906064e-02 -1.79314911e-01 8.95604670e-01
4.92812753e-01 1.06290998e-02 -5.77000260e-01 -2.70753592e-01
3.72526407e-01 1.24828801e-01 6.49349749e-01 -1.17739224e+00
1.47312135e-01 -1.93610922e-01 4.37340409e-01 -6.26232207e-01
8.72005284e-01 -1.11990726e+00 3.30484092e-01 2.81808585e-01
-6.93960339e-02 2.19752103e-01 2.21681982e-01 6.46118224e-01
3.13156813e-01 -1.47629708e-01 1.16153084e-01 -5.68871260e-01
-1.25292039e+00 -2.23860189e-01 -2.44327322e-01 -6.28101587e-01
1.05753779e+00 -8.94311786e-01 2.04335913e-01 -3.25045764e-01
-1.34764588e+00 -7.12083206e-02 7.80443609e-01 6.01674557e-01
9.40236032e-01 -1.21042216e+00 -2.15037823e-01 5.99341333e-01
5.02499938e-01 1.68223187e-01 -7.41616488e-02 4.37767833e-01
-7.25929558e-01 3.28098834e-01 -5.21577954e-01 -7.49234736e-01
-1.25581169e+00 9.34873760e-01 3.49991292e-01 5.32157838e-01
-7.67348230e-01 7.14013040e-01 4.54498082e-01 -6.40241563e-01
2.49666408e-01 -8.64640176e-01 -3.59140605e-01 -4.30519730e-01
6.33930326e-01 3.47126275e-01 -1.32772803e-01 -1.04727411e+00
-1.89852908e-01 6.93512857e-01 4.53544259e-01 -3.88751209e-01
1.12426412e+00 -6.10987782e-01 1.23448469e-01 1.08562434e+00
8.62720191e-01 -2.22700715e-01 -1.47599602e+00 -7.48287737e-02
2.32072920e-01 -1.05247414e+00 -3.65966111e-01 -6.17092013e-01
1.37660891e-01 9.15913582e-01 6.83481991e-01 6.08224511e-01
8.91021013e-01 2.10918576e-01 3.91700193e-02 8.58296037e-01
1.09717155e+00 -7.67903626e-01 3.73261720e-02 2.75430650e-01
1.33052540e+00 -1.14515388e+00 1.09405763e-01 -5.81027210e-01
-4.58185554e-01 1.35772729e+00 3.13195705e-01 -8.67292285e-02
5.63577890e-01 -3.48205231e-02 -1.79239009e-02 -4.07076806e-01
-2.45190233e-01 -4.30603564e-01 4.93976772e-01 1.12390864e+00
2.01155588e-01 3.74494672e-01 3.22719455e-01 -2.00997412e-01
-4.47314948e-01 -4.03422773e-01 5.81314683e-01 1.24821520e+00
-5.46332538e-01 -7.08280027e-01 -5.74110329e-01 -3.99191752e-02
3.22201163e-01 2.94864535e-01 -3.33635956e-01 9.97138619e-01
3.50142062e-01 8.82589698e-01 2.69176275e-01 -1.12054981e-01
8.66635740e-01 1.73221286e-02 1.14753962e+00 -9.42201138e-01
-2.42014319e-01 -5.04264375e-04 2.51587272e-01 -9.87545490e-01
-6.47766709e-01 -1.00211167e+00 -1.26451838e+00 2.42897883e-01
-2.32071191e-01 -2.08764315e-01 1.01330233e+00 9.99174118e-01
1.53898329e-01 2.60427326e-01 3.24000031e-01 -1.76649010e+00
-2.06919722e-02 -7.82059133e-01 -8.14578474e-01 3.56310576e-01
7.44768441e-01 -1.02478814e+00 -2.43905932e-01 6.02983013e-02]
|
[4.937552452087402, 0.3827999234199524]
|
a3f16881-f82c-4a4d-869b-d6520ec9460b
|
unlearnable-clusters-towards-label-agnostic
|
2301.01217
| null |
https://arxiv.org/abs/2301.01217v4
|
https://arxiv.org/pdf/2301.01217v4.pdf
|
Unlearnable Clusters: Towards Label-agnostic Unlearnable Examples
|
There is a growing interest in developing unlearnable examples (UEs) against visual privacy leaks on the Internet. UEs are training samples added with invisible but unlearnable noise, which have been found can prevent unauthorized training of machine learning models. UEs typically are generated via a bilevel optimization framework with a surrogate model to remove (minimize) errors from the original samples, and then applied to protect the data against unknown target models. However, existing UE generation methods all rely on an ideal assumption called label-consistency, where the hackers and protectors are assumed to hold the same label for a given sample. In this work, we propose and promote a more practical label-agnostic setting, where the hackers may exploit the protected data quite differently from the protectors. E.g., a m-class unlearnable dataset held by the protector may be exploited by the hacker as a n-class dataset. Existing UE generation methods are rendered ineffective in this challenging setting. To tackle this challenge, we present a novel technique called Unlearnable Clusters (UCs) to generate label-agnostic unlearnable examples with cluster-wise perturbations. Furthermore, we propose to leverage VisionandLanguage Pre-trained Models (VLPMs) like CLIP as the surrogate model to improve the transferability of the crafted UCs to diverse domains. We empirically verify the effectiveness of our proposed approach under a variety of settings with different datasets, target models, and even commercial platforms Microsoft Azure and Baidu PaddlePaddle. Code is available at \url{https://github.com/jiamingzhang94/Unlearnable-Clusters}.
|
['Yu-Gang Jiang', 'Changsheng Xu', 'YaoWei Wang', 'Jitao Sang', 'Qi Yi', 'Xingjun Ma', 'Jiaming Zhang']
|
2022-12-31
| null |
http://openaccess.thecvf.com//content/CVPR2023/html/Zhang_Unlearnable_Clusters_Towards_Label-Agnostic_Unlearnable_Examples_CVPR_2023_paper.html
|
http://openaccess.thecvf.com//content/CVPR2023/papers/Zhang_Unlearnable_Clusters_Towards_Label-Agnostic_Unlearnable_Examples_CVPR_2023_paper.pdf
|
cvpr-2023-1
|
['data-poisoning']
|
['adversarial']
|
[ 2.94497669e-01 1.03745861e-02 -1.05080284e-01 5.04246578e-02
-8.53276134e-01 -1.33974683e+00 5.19499660e-01 -2.44464546e-01
-2.73352146e-01 7.85823643e-01 -1.37270287e-01 -5.26950657e-01
1.16251811e-01 -5.83918214e-01 -1.11930919e+00 -7.69528270e-01
2.46294037e-01 -5.41659258e-02 -4.12732549e-02 2.26957217e-01
5.46956733e-02 4.66938525e-01 -1.33221853e+00 3.97969097e-01
1.03041649e+00 8.17661285e-01 -1.30169794e-01 6.89183295e-01
3.53718102e-01 6.41718984e-01 -8.24963570e-01 -8.72529984e-01
8.74250948e-01 -1.75483778e-01 -3.94430757e-01 -1.54679015e-01
8.82147372e-01 -3.06146175e-01 -2.85156071e-01 1.59437704e+00
5.19660890e-01 -1.09113246e-01 5.32792032e-01 -1.77889264e+00
-1.07296336e+00 3.85149986e-01 -5.77852368e-01 -2.00694948e-01
2.85397824e-02 5.22034287e-01 5.44216037e-01 -6.20362103e-01
6.00377262e-01 1.08517194e+00 5.69465697e-01 1.18775821e+00
-1.21107197e+00 -1.39890635e+00 6.23167418e-02 2.45760068e-01
-1.39827168e+00 -3.67978483e-01 7.46907949e-01 -3.98481965e-01
2.05474403e-02 6.39116704e-01 1.08228408e-01 1.96333253e+00
-2.42486164e-01 6.86489522e-01 1.54858184e+00 -3.63861442e-01
2.72805661e-01 8.60999823e-01 1.30930305e-01 4.93636489e-01
6.23582006e-01 3.90349656e-01 -2.29563355e-01 -6.12643540e-01
3.64985168e-01 8.08961689e-02 -7.29949534e-01 -3.51484954e-01
-7.63735712e-01 7.65699625e-01 4.28122520e-01 -2.21789286e-01
1.18713696e-02 -7.18302131e-02 3.42097998e-01 1.88789383e-01
2.11413577e-01 4.70772952e-01 -5.23114562e-01 3.95622343e-01
-5.51182449e-01 1.55386716e-01 8.37325811e-01 9.45795000e-01
6.60692334e-01 -1.44852642e-02 -1.61807716e-01 4.36405867e-01
2.05080479e-01 6.49055243e-01 4.38775986e-01 -7.56460309e-01
5.75901747e-01 3.37971836e-01 3.60739172e-01 -1.05353570e+00
2.83948004e-01 -3.79040092e-01 -8.00071061e-01 6.43337905e-01
5.38469851e-01 -4.40072447e-01 -1.07040989e+00 1.95673811e+00
6.82858288e-01 6.34656668e-01 1.08001776e-01 8.07704210e-01
3.87440801e-01 5.12699068e-01 2.04672977e-01 -1.16572939e-01
1.23177791e+00 -9.98231173e-01 -4.55937326e-01 -3.89340543e-03
4.03674304e-01 -5.73109150e-01 1.46795869e+00 5.63792825e-01
-4.78816271e-01 -3.78147155e-01 -1.11663759e+00 2.88398862e-01
-5.80295980e-01 -1.42641813e-01 2.02072531e-01 1.39010251e+00
-6.91652060e-01 4.17556167e-01 -4.55534399e-01 -8.58405530e-02
8.37912261e-01 1.48627251e-01 -5.15161216e-01 -1.08255319e-01
-1.18327773e+00 4.71103698e-01 4.11186099e-01 -9.03306603e-02
-1.44804692e+00 -7.82559991e-01 -5.73482990e-01 -1.43987030e-01
6.57173812e-01 -4.19127166e-01 8.70044410e-01 -1.20544064e+00
-9.42215681e-01 8.80955458e-01 2.74960577e-01 -5.45488060e-01
8.46469641e-01 -1.97134838e-01 -5.54883659e-01 7.19327480e-02
-1.31530672e-01 2.18706354e-01 1.39735103e+00 -2.01045418e+00
-3.89275998e-01 -3.93670142e-01 4.74568941e-02 -1.61141977e-01
-8.78145218e-01 1.06497630e-01 -1.84369728e-01 -9.29775238e-01
-5.76528430e-01 -1.02703190e+00 -6.20743558e-02 8.20574090e-02
-8.33039820e-01 3.30695450e-01 1.23333693e+00 -8.72035265e-01
1.11996090e+00 -2.30684042e+00 -2.46589199e-01 3.22423041e-01
3.09020549e-01 9.50639307e-01 -3.41926366e-01 1.50525987e-01
-2.11717218e-01 6.80534422e-01 -2.38791510e-01 -3.68900150e-01
1.31045505e-01 1.37669027e-01 -8.95237327e-01 4.37471300e-01
-1.91260442e-01 7.82476544e-01 -8.12876523e-01 -1.97220072e-01
-9.12743434e-02 4.77358282e-01 -3.56764615e-01 4.16490793e-01
-4.22546208e-01 5.72047889e-01 -3.49515587e-01 7.85203516e-01
1.05497718e+00 -1.00689657e-01 -6.63666194e-03 -3.85610797e-02
4.62697804e-01 -3.70039284e-01 -1.25366056e+00 9.37740624e-01
-1.69490471e-01 3.19210291e-01 2.15342358e-01 -4.22261059e-01
6.83957994e-01 3.84107143e-01 -1.61348596e-01 -2.03648373e-01
1.51278049e-01 1.15711689e-01 -2.94362336e-01 -5.96559167e-01
1.92879021e-01 -6.65082783e-03 6.76076114e-02 4.78112370e-01
-1.49434820e-01 5.92156827e-01 -3.71186525e-01 1.83600008e-01
1.08780146e+00 -1.78555679e-02 1.76894158e-01 3.05781424e-01
4.61373806e-01 -2.61782229e-01 7.10716486e-01 1.03717017e+00
-4.83916968e-01 5.97239912e-01 2.99178809e-01 -2.50299126e-01
-1.08238816e+00 -1.22020304e+00 -8.92961100e-02 8.38558853e-01
2.43281811e-01 -2.74739176e-01 -1.00962341e+00 -1.49398363e+00
1.06310718e-01 8.11938465e-01 -6.21761620e-01 -3.67863804e-01
-2.29453325e-01 -6.27885342e-01 1.05601931e+00 9.76635888e-02
5.71876466e-01 -1.00745153e+00 -8.72019976e-02 -2.87536442e-01
-1.12489536e-01 -1.03032875e+00 -5.94353080e-01 -2.00907096e-01
-3.96095008e-01 -1.36399567e+00 -5.35159409e-01 -4.59890008e-01
9.03714240e-01 1.74909249e-01 5.02579808e-01 5.40071763e-02
-2.66511261e-01 5.89530647e-01 -4.14098084e-01 -5.58170855e-01
-6.75965726e-01 -1.68707043e-01 3.47085625e-01 5.53309739e-01
3.61727238e-01 -4.77113247e-01 -5.54719746e-01 4.18365479e-01
-1.24012518e+00 -3.37721944e-01 3.63743991e-01 7.65546620e-01
4.45890069e-01 2.61556476e-01 4.01066422e-01 -1.17658424e+00
7.27597058e-01 -6.35645747e-01 -5.94065607e-01 5.89894772e-01
-5.86051047e-01 -1.34357825e-01 1.05155492e+00 -1.23339331e+00
-8.96697640e-01 -6.58488348e-02 1.69921398e-01 -1.00742209e+00
-5.72760999e-01 -1.81629017e-01 -8.11139286e-01 -5.36875188e-01
9.94809330e-01 2.78073132e-01 -3.34461659e-01 -6.13570154e-01
5.77699602e-01 9.02695119e-01 6.01031661e-01 -8.16094100e-01
1.46284842e+00 5.52925587e-01 -2.61971444e-01 -4.34237659e-01
-6.84573352e-01 -1.16347492e-01 -6.38519824e-02 -1.91964194e-01
6.22585356e-01 -7.37324238e-01 -7.13001072e-01 7.66862452e-01
-1.13542390e+00 -2.24248827e-01 -1.39620259e-01 1.83505461e-01
-1.60616174e-01 7.34652281e-01 -3.54650080e-01 -9.89314139e-01
-4.95779157e-01 -9.77051318e-01 7.12583899e-01 4.51972038e-01
1.40843198e-01 -7.91219413e-01 -3.16663738e-03 7.18017161e-01
1.29864544e-01 6.48347139e-01 8.56192112e-01 -1.14381611e+00
-7.58654952e-01 -4.65079218e-01 -1.40567482e-01 8.24037731e-01
5.88137917e-02 -1.57223657e-01 -1.35775900e+00 -6.28296733e-01
2.75073797e-01 -4.43664163e-01 6.37989044e-01 -1.92713529e-01
1.43019891e+00 -1.03109241e+00 -3.49725276e-01 9.58922982e-01
1.44407082e+00 1.47316962e-01 6.67246640e-01 5.26179910e-01
9.84281182e-01 4.39083964e-01 2.45490193e-01 3.84044349e-01
6.24511167e-02 4.51428384e-01 8.54900777e-01 5.65862805e-02
3.36805955e-02 -6.04022324e-01 4.78708982e-01 2.18182683e-01
2.09831238e-01 -5.49011230e-01 -6.78706586e-01 3.95083368e-01
-1.74698293e+00 -9.90328610e-01 -1.13039259e-02 2.41462302e+00
9.76275980e-01 -1.93754267e-02 -6.18294291e-02 -6.14066534e-02
1.02096629e+00 1.46217607e-02 -7.97825396e-01 -6.10086508e-02
-7.46570155e-02 8.08263421e-02 7.23031104e-01 3.98825675e-01
-1.27114570e+00 9.80577886e-01 4.27902365e+00 1.18755531e+00
-1.12356937e+00 3.61492336e-01 5.38343787e-01 -1.55691534e-01
-2.76940733e-01 1.85497627e-01 -9.20219362e-01 7.69817054e-01
7.85445094e-01 -8.12546760e-02 6.92964435e-01 1.07221103e+00
-1.04949437e-01 5.39413095e-01 -8.41591120e-01 9.97670054e-01
2.58988023e-01 -1.05904162e+00 1.41433537e-01 2.59792686e-01
6.87371314e-01 -2.13455021e-01 5.81921697e-01 1.83368400e-01
5.40013433e-01 -9.34161365e-01 6.51827037e-01 5.15384495e-01
8.83497953e-01 -6.95218742e-01 3.35648745e-01 5.70786417e-01
-5.89587748e-01 -2.13431001e-01 -4.80027914e-01 4.66969520e-01
-1.04787141e-01 3.48076403e-01 -9.70603883e-01 4.66904998e-01
8.35717618e-01 2.00806931e-01 -8.21604788e-01 1.02325618e+00
-5.60661912e-01 9.67781365e-01 -1.53012261e-01 2.55776972e-01
-2.25187298e-02 -2.12305021e-02 8.51660609e-01 1.03197348e+00
2.42217064e-01 -6.35954365e-02 3.44617143e-02 8.72912765e-01
-4.65788126e-01 -2.25478653e-02 -8.45658302e-01 -1.14635579e-01
6.77932918e-01 1.38657176e+00 -2.80984312e-01 -1.04321659e-01
-1.21130347e-01 1.11147273e+00 2.84883887e-01 7.62846172e-01
-1.00865424e+00 -4.43688273e-01 9.25319195e-01 9.61037576e-02
1.73485547e-01 2.56052446e-02 -2.14004785e-01 -1.41073096e+00
1.95881963e-01 -1.48280311e+00 5.97239137e-01 -8.12249839e-01
-1.64770234e+00 6.67526007e-01 -2.93269128e-01 -1.37835932e+00
3.01246822e-01 -6.67812407e-01 -4.96025205e-01 9.01729465e-01
-1.44489467e+00 -1.40818763e+00 -3.02417725e-01 8.26442778e-01
8.55026543e-02 -4.22857940e-01 8.03874433e-01 1.03896476e-01
-5.98137319e-01 1.08841586e+00 2.51928002e-01 3.95496219e-01
9.94924486e-01 -1.06611621e+00 2.71538556e-01 1.21557319e+00
3.37378979e-01 9.49240088e-01 6.17204785e-01 -9.08673525e-01
-1.03207123e+00 -1.50540829e+00 2.83360958e-01 -8.48818898e-01
6.11743093e-01 -8.22262943e-01 -1.20442331e+00 8.89573634e-01
1.08747885e-01 1.48570538e-01 9.27026868e-01 -4.53553468e-01
-1.04092014e+00 -1.55018754e-02 -1.62682271e+00 8.86389315e-01
8.51074755e-01 -6.65486395e-01 -2.48730466e-01 3.55316699e-01
9.46987629e-01 -2.22895205e-01 -4.80185807e-01 2.98807234e-01
4.47677135e-01 -7.73642778e-01 1.04933953e+00 -9.76194561e-01
7.66300410e-02 -5.52953839e-01 -2.06696823e-01 -1.05668795e+00
8.57989639e-02 -9.22480404e-01 -4.74763215e-01 1.58202147e+00
1.99785054e-01 -8.27238142e-01 9.32769418e-01 7.69378126e-01
3.82526696e-01 -2.67006934e-01 -7.95963466e-01 -1.05504501e+00
9.20287743e-02 -4.50711757e-01 7.79220641e-01 1.38722622e+00
-4.30931062e-01 -1.85324356e-01 -9.18166578e-01 8.58168066e-01
1.18606484e+00 -3.56798261e-01 1.03763843e+00 -8.65751445e-01
-4.59096909e-01 1.38245761e-01 -1.95182949e-01 -6.21549070e-01
3.66779685e-01 -9.68639076e-01 -3.09263468e-01 -6.35263920e-01
1.83262900e-01 -4.93538558e-01 -3.24151188e-01 7.56492019e-01
-4.63921070e-01 3.24041605e-01 5.14230251e-01 3.15475881e-01
-3.71072650e-01 3.91086549e-01 1.05726945e+00 -2.42576227e-01
-2.42330544e-02 1.97440147e-01 -9.95804548e-01 6.72437966e-01
8.97839904e-01 -9.48664188e-01 -5.43644369e-01 -1.35742947e-01
1.59513019e-02 -4.97690648e-01 8.11949611e-01 -9.04312074e-01
1.57212421e-01 -2.43163019e-01 3.46226215e-01 -3.64022031e-02
8.19470547e-03 -1.25519097e+00 3.59082609e-01 2.79588461e-01
-3.18405986e-01 -2.84325391e-01 1.67627409e-01 9.69072163e-01
2.35556602e-01 -4.38103169e-01 8.81031692e-01 -2.07468092e-01
-5.38978100e-01 4.94441628e-01 2.01548204e-01 1.11287676e-01
1.41049492e+00 -2.32558072e-01 -9.03266788e-01 -2.45700598e-01
-6.63614690e-01 1.35739252e-01 7.76280284e-01 4.03096586e-01
6.49875760e-01 -1.21854329e+00 -5.50845861e-01 3.22995603e-01
2.37280667e-01 -3.43306810e-01 1.88383505e-01 2.19773933e-01
-1.89099118e-01 -1.80127636e-01 -1.53646410e-01 -1.31451771e-01
-1.61805737e+00 1.16723549e+00 3.44798326e-01 -7.78810382e-02
-3.48114669e-01 7.35343695e-01 3.97469997e-01 -7.75772512e-01
4.75355715e-01 3.71783048e-01 7.80404955e-02 -2.22190440e-01
6.47471786e-01 4.61796314e-01 -2.63313174e-01 -5.29698849e-01
-1.73304439e-01 2.88982570e-01 -2.99472302e-01 1.48776770e-01
8.91358316e-01 4.18794081e-02 3.24534290e-02 -8.52292553e-02
1.21087527e+00 2.81610012e-01 -1.52740657e+00 -2.19189450e-01
-1.00095794e-01 -8.10828507e-01 -4.91558224e-01 -9.62634623e-01
-9.69311595e-01 6.76829755e-01 8.20154488e-01 2.24688411e-01
1.06360900e+00 -2.50077337e-01 8.93106699e-01 1.12602353e-01
4.16321069e-01 -8.05835843e-01 2.02401116e-01 -9.37263519e-02
6.82614684e-01 -1.25190294e+00 -2.87302762e-01 -3.65285993e-01
-8.10351014e-01 6.93754733e-01 8.09700251e-01 4.64917868e-02
4.69497383e-01 1.31033435e-01 3.56965214e-01 2.62177467e-01
-4.65239793e-01 2.01473743e-01 1.58630580e-01 1.17765105e+00
-4.99046355e-01 5.55333048e-02 4.62068245e-02 9.36976671e-01
3.00105289e-02 -1.44610658e-01 6.66524529e-01 7.63883770e-01
-8.19776952e-02 -1.30058467e+00 -8.65462840e-01 1.51588395e-01
-6.08975649e-01 -4.48635630e-02 -6.13982141e-01 5.68460166e-01
4.96506691e-01 9.66320276e-01 -6.57979071e-01 -7.30640471e-01
2.71237046e-01 1.88813135e-01 8.85646045e-02 -5.56549013e-01
-7.38171220e-01 -3.34892750e-01 -2.50491887e-01 -4.16017026e-01
-6.69174194e-02 -3.53500336e-01 -6.74538255e-01 -3.22333604e-01
-4.11902815e-01 4.25398536e-02 5.74725688e-01 5.64318419e-01
4.51343387e-01 -8.07052553e-02 8.33746016e-01 -4.46072698e-01
-1.04420745e+00 -5.13006210e-01 -6.14489675e-01 8.01824450e-01
5.43746233e-01 -3.12704504e-01 -9.08109128e-01 2.99009651e-01]
|
[5.750309467315674, 7.46855354309082]
|
f28ac9ed-4039-458d-947b-605ccf8ad9c2
|
improving-text-independent-speaker
|
2109.09674
| null |
https://arxiv.org/abs/2109.09674v1
|
https://arxiv.org/pdf/2109.09674v1.pdf
|
Improving Text-Independent Speaker Verification with Auxiliary Speakers Using Graph
|
The paper presents a novel approach to refining similarity scores between input utterances for robust speaker verification. Given the embeddings from a pair of input utterances, a graph model is designed to incorporate additional information from a group of embeddings representing the so-called auxiliary speakers. The relations between the input utterances and the auxiliary speakers are represented by the edges and vertices in the graph. The similarity scores are refined by iteratively updating the values of the graph's vertices using an algorithm similar to the random walk algorithm on graphs. Through this updating process, the information of auxiliary speakers is involved in determining the relation between input utterances and hence contributing to the verification process. We propose to create a set of artificial embeddings through the model training process. Utilizing the generated embeddings as auxiliary speakers, no extra data are required for the graph model in the verification stage. The proposed model is trained in an end-to-end manner within the whole system. Experiments are carried out with the Voxceleb datasets. The results indicate that involving auxiliary speakers with graph is effective to improve speaker verification performance.
|
['Tan Lee', 'Si-Ioi Ng', 'Jingyu Li']
|
2021-09-20
| null | null | null | null |
['text-independent-speaker-verification']
|
['speech']
|
[ 1.72018945e-01 6.09403789e-01 2.04082757e-01 -7.48205781e-01
-4.24696922e-01 -2.53563136e-01 6.53990388e-01 4.85047460e-01
-1.82430744e-01 7.66585469e-02 4.17860389e-01 -1.99467272e-01
-5.39650172e-02 -6.12668633e-01 -2.59509385e-01 -6.74818277e-01
-1.26978800e-01 5.95714033e-01 2.00755857e-02 -2.43728980e-01
2.10662007e-01 3.10020268e-01 -1.44379342e+00 -6.86896369e-02
6.59945548e-01 6.81161106e-01 1.58142850e-01 8.40600014e-01
-2.61137813e-01 3.66537869e-01 -5.34462452e-01 -5.92431843e-01
2.33239248e-01 -6.67899668e-01 -5.83509624e-01 5.59915245e-01
1.51778907e-01 -4.13282327e-02 -3.13435823e-01 1.11763787e+00
5.90670228e-01 5.06317139e-01 4.67806041e-01 -1.31652975e+00
-6.94509327e-01 1.08999205e+00 -2.25802451e-01 -4.78591993e-02
5.83230019e-01 -2.82398283e-01 1.23808873e+00 -1.18758261e+00
4.85529721e-01 1.41123343e+00 4.69421178e-01 6.04438543e-01
-8.79284203e-01 -4.53240454e-01 3.42065603e-01 4.12836164e-01
-1.59193480e+00 -8.16472948e-01 1.16328883e+00 -1.28861174e-01
6.69582248e-01 3.36447358e-01 5.93893945e-01 4.19757068e-01
-2.35459998e-01 5.05149782e-01 3.99764538e-01 -7.56982625e-01
2.69202977e-01 5.59538364e-01 5.42789042e-01 9.43338335e-01
-2.19752401e-01 -7.96781629e-02 -6.42290711e-01 -4.82738972e-01
5.42260520e-02 -2.30598539e-01 -3.24904889e-01 -5.70194185e-01
-8.87756407e-01 1.01219440e+00 4.77559894e-01 4.13878083e-01
-3.56794626e-01 -2.95681536e-01 3.41872305e-01 3.24744552e-01
3.42709452e-01 -3.37356329e-03 -5.05212918e-02 3.18088055e-01
-6.93590760e-01 -1.35386765e-01 7.38150477e-01 8.98325861e-01
8.22765052e-01 -3.60453762e-02 9.75707695e-02 9.01903212e-01
1.08708215e+00 2.36980319e-01 3.95638049e-01 -2.18898714e-01
7.83676803e-01 9.49349761e-01 -1.53736770e-01 -1.19891953e+00
-7.51949400e-02 -3.73487890e-01 -4.76646781e-01 -1.79532632e-01
1.59636781e-01 -3.19304802e-02 -1.05055344e+00 1.68058538e+00
8.46526980e-01 3.86424601e-01 2.93397635e-01 7.32826531e-01
9.67537761e-01 6.17285132e-01 3.10651902e-02 -1.43784821e-01
1.40116465e+00 -9.45859790e-01 -9.07828689e-01 5.71969859e-02
6.98776007e-01 -8.86813760e-01 6.28772259e-01 -1.68832541e-01
-9.22671616e-01 -7.30680764e-01 -1.02087224e+00 1.73273817e-01
-3.80792260e-01 -4.70202125e-04 4.40392047e-02 7.16603041e-01
-1.32945645e+00 1.81609347e-01 -6.17013991e-01 -4.84975815e-01
7.21417367e-02 5.36452591e-01 -3.43954504e-01 -9.20112059e-02
-1.30901635e+00 9.43421721e-01 1.87257916e-01 5.41368842e-01
-7.46146858e-01 -3.03036928e-01 -1.36265028e+00 5.88680059e-02
-1.45327300e-01 -3.35409433e-01 9.40265119e-01 -6.82146966e-01
-1.39836812e+00 5.73755503e-01 -5.16828477e-01 -1.22363448e-01
2.89149284e-01 4.23884243e-01 -6.44169688e-01 9.06247552e-03
-8.20292830e-02 4.33541983e-01 9.82301235e-01 -1.43508112e+00
-5.47830403e-01 -5.11826575e-01 -1.37830913e-01 4.17549342e-01
-3.71127307e-01 1.48110986e-01 -7.34749138e-01 -2.06465229e-01
5.63377261e-01 -9.89799619e-01 -7.05197603e-02 -2.70176142e-01
-4.74364221e-01 -6.41240835e-01 1.14633703e+00 -8.81741762e-01
1.31415892e+00 -2.37838292e+00 3.01984459e-01 8.40340972e-01
2.11526319e-01 3.05430651e-01 -3.99980217e-01 5.19344330e-01
-1.59432933e-01 -1.41911164e-01 -2.62179613e-01 -6.82166278e-01
9.71821398e-02 1.31418794e-01 1.75790772e-01 5.41859329e-01
1.24954641e-01 5.16193032e-01 -8.88717592e-01 -6.07168198e-01
2.44005084e-01 8.12724769e-01 -3.80883962e-01 5.21750271e-01
1.68720812e-01 2.67289072e-01 -4.56670403e-01 2.65433401e-01
7.40269363e-01 1.52542979e-01 4.00854975e-01 -6.47978634e-02
2.77216703e-01 4.69086051e-01 -1.49584985e+00 1.45240045e+00
-3.74756724e-01 4.59980160e-01 2.66515464e-01 -8.72530401e-01
1.35354602e+00 6.18943334e-01 -3.16923596e-02 -1.58454418e-01
2.48071685e-01 -8.60522687e-02 3.22764456e-01 -4.84135836e-01
4.91939515e-01 8.14413559e-03 3.62176858e-02 6.14305258e-01
1.61619142e-01 2.08806489e-02 1.79966718e-01 5.45171559e-01
7.39511669e-01 -4.82853562e-01 1.24117941e-01 -1.19027426e-03
1.14695251e+00 -3.76372278e-01 2.80420482e-01 1.08502284e-01
-2.99912840e-01 1.47682101e-01 1.59194514e-01 -1.73333153e-01
-8.63990068e-01 -9.18153524e-01 6.04997799e-02 9.86728311e-01
3.22295167e-02 -4.41723347e-01 -8.53650451e-01 -7.88071036e-01
-7.81124011e-02 7.01085687e-01 -8.57110023e-01 -2.05856100e-01
-3.18138242e-01 -1.08900949e-01 3.01667362e-01 3.21557969e-01
2.73544759e-01 -9.30850089e-01 1.23489536e-02 2.15588287e-01
-1.67933166e-01 -9.43712592e-01 -8.65896940e-01 1.60514694e-02
-7.67460346e-01 -1.01424861e+00 -3.84018272e-01 -1.24058092e+00
1.33096337e+00 1.88634142e-01 5.27229846e-01 4.21974003e-01
1.00044198e-01 2.85632014e-01 -4.81020153e-01 -2.30742380e-01
-8.85198593e-01 -1.33698612e-01 2.23861381e-01 5.55424869e-01
5.76070964e-01 -2.33220741e-01 -2.20050901e-01 4.70383823e-01
-6.70980155e-01 -2.78548956e-01 2.73348451e-01 9.99594748e-01
2.37827778e-01 9.26661044e-02 6.83611810e-01 -7.42312968e-01
7.97520339e-01 -3.79716128e-01 -5.42033672e-01 4.43229288e-01
-4.42819387e-01 4.36666846e-01 4.89747405e-01 -4.31659222e-01
-1.04797184e+00 1.93166763e-01 -1.75146937e-01 -2.94487923e-01
5.83867431e-02 6.37703836e-01 -4.49470699e-01 -7.33450055e-02
3.04765761e-01 1.34150520e-01 2.12667659e-01 -2.47121438e-01
6.50568604e-01 1.04123986e+00 8.89311880e-02 -2.75230948e-02
1.05244076e+00 -1.00141093e-02 -4.46646661e-01 -8.70958924e-01
-2.32603461e-01 -7.23553836e-01 -7.86276042e-01 -4.61921304e-01
7.10337400e-01 -6.08928561e-01 -5.58217168e-01 2.13574573e-01
-1.11054015e+00 2.98758507e-01 -1.43635899e-01 6.23725176e-01
5.79435416e-02 4.85884249e-01 -1.83128953e-01 -1.17850852e+00
-3.14700037e-01 -1.19239390e+00 8.80879581e-01 2.92914033e-01
-3.77587050e-01 -1.32718146e+00 2.29636803e-01 5.25025368e-01
2.11961508e-01 -2.50331163e-01 9.78276551e-01 -1.12420785e+00
-4.01279420e-01 -4.87778008e-01 9.43398774e-02 3.90429676e-01
4.22866762e-01 1.86199769e-01 -1.08052564e+00 -2.89043039e-01
1.04629556e-02 9.34847072e-02 4.42184836e-01 2.50258911e-02
3.03079873e-01 -1.80427700e-01 -3.86103123e-01 1.30468547e-01
9.04740870e-01 4.21523929e-01 2.78726339e-01 -2.79449791e-01
6.76729321e-01 8.19634736e-01 5.03917456e-01 2.51858622e-01
5.26631176e-01 5.58441937e-01 1.91516876e-01 5.35561182e-02
-1.53005749e-01 -5.47092319e-01 3.81502122e-01 1.49453688e+00
2.05410540e-01 -1.96692824e-01 -7.53583372e-01 7.42547631e-01
-1.57857716e+00 -8.71986806e-01 -1.08732484e-01 2.22000623e+00
5.32493472e-01 -8.13354179e-02 -1.30273536e-01 5.42869508e-01
1.19213986e+00 2.79599577e-01 -2.20571950e-01 -7.60882318e-01
3.56534690e-01 1.04644194e-01 -1.69143781e-01 1.06463015e+00
-6.89634621e-01 7.83402741e-01 5.63416719e+00 9.52729657e-02
-7.80863822e-01 -5.85754216e-03 1.95839331e-01 4.37802136e-01
-4.69816536e-01 5.56377620e-02 -9.02270854e-01 1.08776435e-01
9.31744635e-01 -3.36523503e-01 3.80021602e-01 6.47508323e-01
2.40639746e-01 1.76734015e-01 -1.16456652e+00 6.49764538e-01
4.58075136e-01 -8.21362257e-01 6.39562160e-02 -1.46908611e-01
4.47470754e-01 -1.73134938e-01 -2.23479271e-02 1.74464256e-01
3.14816952e-01 -5.48096359e-01 5.31174719e-01 1.68399572e-01
4.36642736e-01 -7.78044641e-01 9.92729962e-01 1.71949446e-01
-1.65718591e+00 2.27574799e-02 -4.07771766e-01 2.89825767e-01
2.87423491e-01 2.24159017e-01 -1.62282217e+00 6.14448309e-01
1.38996914e-01 3.31136912e-01 -4.65790957e-01 7.98229754e-01
-4.92847741e-01 4.02897626e-01 -1.27189100e-01 -3.43946874e-01
9.79168937e-02 -3.05683017e-01 5.07291913e-01 9.61951315e-01
2.59577304e-01 -1.09545719e-02 -7.64532611e-02 5.81237316e-01
-2.56001264e-01 2.96859205e-01 -7.64162064e-01 -5.34384772e-02
6.27469778e-01 1.28690803e+00 -5.22320449e-01 -4.56104815e-01
-3.60869229e-01 8.00765038e-01 4.32908416e-01 3.68017852e-01
-7.59716153e-01 -5.20492136e-01 5.32812059e-01 -3.38513963e-02
3.87078315e-01 -1.29208684e-01 1.23163179e-01 -6.04569614e-01
6.87001199e-02 -7.39877999e-01 3.76148999e-01 -4.70256835e-01
-1.00668633e+00 9.40232575e-01 -2.20648050e-01 -1.01977420e+00
-3.05196971e-01 -1.72013670e-01 -8.60651016e-01 1.28030789e+00
-1.46161067e+00 -8.96409333e-01 -3.16386014e-01 8.51328492e-01
3.90007675e-01 -2.98127085e-01 1.05000532e+00 1.79148450e-01
-5.44426501e-01 8.96515608e-01 -2.52462447e-01 4.24676895e-01
3.27401757e-01 -9.40102518e-01 5.13671577e-01 1.07543659e+00
5.03086030e-01 7.62183547e-01 6.29960537e-01 -6.18488610e-01
-1.30843973e+00 -1.04979241e+00 1.66894805e+00 -2.47559577e-01
5.74962795e-01 -6.23835623e-01 -9.65297461e-01 5.86401343e-01
4.52684879e-01 -7.03062341e-02 9.82595563e-01 2.57211268e-01
-4.73391414e-01 -8.88231471e-02 -1.23412359e+00 3.83192450e-01
8.50551963e-01 -8.69465530e-01 -9.43810344e-01 1.44927412e-01
6.56343877e-01 -2.27414280e-01 -8.03914011e-01 -1.75539795e-02
2.96589643e-01 -4.11775649e-01 5.86082339e-01 -6.07482910e-01
-1.67681992e-01 -5.19544423e-01 -5.07500544e-02 -1.67270327e+00
-2.70295650e-01 -5.54322720e-01 9.80415717e-02 1.54801357e+00
8.71308625e-01 -6.94448531e-01 6.29536867e-01 6.41580582e-01
-1.53404817e-01 -3.01823467e-01 -1.04103613e+00 -3.99439335e-01
-7.49141037e-01 -2.66882181e-01 9.33912456e-01 9.45469916e-01
4.96511728e-01 6.22098327e-01 -3.94234583e-02 7.31090367e-01
7.54681468e-01 -1.38945937e-01 7.71830320e-01 -1.09759712e+00
-9.47841704e-02 5.50401746e-04 -7.69725561e-01 -8.62981200e-01
3.78046215e-01 -1.13288307e+00 2.49100164e-01 -1.62129712e+00
-2.42774300e-02 -6.06176317e-01 -3.62364352e-01 3.61746252e-01
-4.70001519e-01 -1.61904901e-01 1.77852824e-01 -1.44228518e-01
-3.00306171e-01 6.78995430e-01 9.61439371e-01 -2.44636267e-01
-3.73922259e-01 2.24760279e-01 -5.06189346e-01 3.63412619e-01
8.00235987e-01 -4.68948334e-01 -7.66262710e-01 -3.00278902e-01
-4.97918665e-01 2.12450430e-01 -1.01982549e-01 -6.17559612e-01
5.06948411e-01 3.71704429e-01 2.04452928e-02 -5.38879514e-01
3.76511633e-01 -1.05601823e+00 1.98774293e-01 5.00520825e-01
-5.28650582e-01 2.61652857e-01 -2.41585616e-02 5.79085350e-01
-4.76237714e-01 -3.10515910e-01 7.39561498e-01 4.03381824e-01
-3.93395185e-01 3.27709764e-01 -6.39819726e-02 -3.53135228e-01
1.05107641e+00 -2.64843374e-01 3.41451094e-02 -5.59121311e-01
-1.15787005e+00 3.95919234e-01 5.87710701e-02 5.47071457e-01
9.76593316e-01 -1.49424517e+00 -7.70326316e-01 6.47416413e-01
4.02253330e-01 -2.40757301e-01 1.29634336e-01 5.51135480e-01
-7.15421736e-02 2.84965962e-01 9.45023298e-02 -6.00789249e-01
-2.04235744e+00 4.48456466e-01 2.21252292e-01 -1.74597278e-01
-2.07443208e-01 1.19379199e+00 -2.03096688e-01 -6.69587791e-01
5.48175037e-01 -3.69275630e-01 -5.06357253e-01 3.18004012e-01
4.36767876e-01 1.11024611e-01 2.19320282e-01 -1.22339010e+00
-5.44758976e-01 4.12279785e-01 -2.93012470e-01 -3.18151355e-01
1.28941369e+00 -2.82322496e-01 -8.23653713e-02 2.96676248e-01
1.36449671e+00 1.69434458e-01 -8.20742130e-01 -4.68074888e-01
3.56095880e-02 -5.57496190e-01 -5.94889298e-02 -3.40134680e-01
-1.25032032e+00 8.84484529e-01 5.95126092e-01 2.99149960e-01
9.43352818e-01 1.32890403e-01 4.45012271e-01 1.04041360e-01
2.15784013e-01 -8.98305237e-01 -2.10378170e-01 2.86691308e-01
7.85131037e-01 -1.08856392e+00 -2.85390824e-01 -4.73784745e-01
-7.85131454e-01 9.33248520e-01 2.45584384e-01 2.25115746e-01
8.24977040e-01 5.05741744e-04 4.88238305e-01 -2.05813617e-01
-6.09707236e-01 -8.03392604e-02 3.00818324e-01 7.34068155e-01
4.60102051e-01 2.24600479e-01 -2.27097720e-01 1.12758227e-01
-2.63694227e-01 -5.50187051e-01 2.54861146e-01 6.73735917e-01
-4.78212923e-01 -1.42947459e+00 -4.18476492e-01 1.42080739e-01
2.56411731e-01 -7.67263994e-02 -5.73731124e-01 4.99096155e-01
-6.50673583e-02 1.47098362e+00 2.91581284e-02 -6.55083239e-01
5.80998719e-01 4.54196513e-01 1.65173963e-01 -6.95475817e-01
-7.11249888e-01 -2.15738818e-01 2.32445806e-01 -1.59300312e-01
-3.73283684e-01 -7.11485744e-01 -1.42931294e+00 -2.65368044e-01
-8.56573164e-01 7.90073216e-01 9.86233413e-01 7.71457791e-01
2.99436659e-01 5.55809796e-01 1.25103414e+00 -4.94174480e-01
-5.89736879e-01 -1.12007678e+00 -5.50741196e-01 4.78132933e-01
3.45584333e-01 -5.14167607e-01 -6.22048318e-01 8.88378266e-03]
|
[14.312861442565918, 6.19175386428833]
|
92f30401-5319-497c-a5d7-c08c9c5bcd5a
|
subject-independent-brain-computer-interfaces
|
2301.07894
| null |
https://arxiv.org/abs/2301.07894v1
|
https://arxiv.org/pdf/2301.07894v1.pdf
|
Subject-Independent Brain-Computer Interfaces with Open-Set Subject Recognition
|
A brain-computer interface (BCI) can't be effectively used since electroencephalography (EEG) varies between and within subjects. BCI systems require calibration steps to adjust the model to subject-specific data. It is widely acknowledged that this is a major obstacle to the development of BCIs. To address this issue, previous studies have trained a generalized model by removing the subjects' information. In contrast, in this work, we introduce a style information encoder as an auxiliary task that classifies various source domains and recognizes open-set domains. Open-set recognition method was used as an auxiliary task to learn subject-related style information from the source subjects, while at the same time helping the shared feature extractor map features in an unseen target. This paper compares various OSR methods within an open-set subject recognition (OSSR) framework. As a result of our experiments, we found that the OSSR auxiliary network that encodes domain information improves generalization performance.
|
['Geun-Deok Jang', 'Dong-Young Kim', 'Dong-Kyun Han']
|
2023-01-19
| null | null | null | null |
['open-set-learning']
|
['miscellaneous']
|
[ 3.88340443e-01 2.43577808e-02 2.76717991e-01 -8.29718828e-01
-5.30866265e-01 -5.31621218e-01 2.90781111e-01 -4.64310616e-01
-3.50245446e-01 1.16340756e+00 7.00219721e-02 1.28044456e-01
-2.44898573e-01 -4.05143917e-01 -4.95724291e-01 -4.95932192e-01
-1.44210393e-02 4.43515062e-01 5.71127571e-02 -3.40194821e-01
5.17375529e-01 4.89326984e-01 -1.49765539e+00 6.00982726e-01
1.09427285e+00 1.05821550e+00 4.47724551e-01 2.28867665e-01
-1.69410497e-01 2.65878350e-01 -1.26695514e+00 -4.06291336e-03
2.99143404e-01 -4.95977640e-01 -7.53609776e-01 -3.39933008e-01
2.17926756e-01 8.27659145e-02 -4.24495012e-01 1.13626230e+00
7.28658080e-01 -7.80968666e-02 1.04672039e+00 -1.48936033e+00
-1.03353977e+00 3.67507696e-01 -1.47616327e-01 5.40833533e-01
4.97891724e-01 -1.39212415e-01 3.09914023e-01 -5.49596667e-01
3.84346664e-01 8.35841715e-01 4.05202568e-01 1.08094466e+00
-1.14038074e+00 -1.29358757e+00 5.34236357e-02 5.38114607e-01
-1.67061436e+00 -3.76935959e-01 9.44937825e-01 -3.41167539e-01
8.96230757e-01 3.13065350e-01 6.33574903e-01 1.61587012e+00
5.40391505e-01 6.81148648e-01 1.53186965e+00 -4.61535394e-01
3.94507289e-01 7.21532047e-01 7.02744365e-01 -1.67111591e-01
3.02531779e-01 2.60347307e-01 -8.04321289e-01 8.47369358e-02
8.68500054e-01 -1.00491896e-01 -7.17101455e-01 -2.03238517e-01
-9.86145735e-01 3.68949920e-01 3.75020057e-01 8.02716553e-01
-1.89431086e-01 -5.30960262e-01 3.76291573e-01 9.11888480e-01
3.71161908e-01 7.47376502e-01 -5.54273486e-01 -3.12791288e-01
-8.77674520e-01 -1.53180540e-01 1.09647298e+00 1.26329839e+00
4.99363005e-01 1.27633050e-01 -1.87312648e-01 1.09256673e+00
-1.32236004e-01 4.44848657e-01 1.22830558e+00 -1.53178766e-01
3.10975313e-01 6.17960215e-01 -2.79716998e-01 -6.45171285e-01
-2.27418542e-01 -6.63848877e-01 -7.35617518e-01 2.46661425e-01
1.45278871e-01 -9.38312933e-02 -1.08320272e+00 1.63972461e+00
-4.10970300e-01 2.85187423e-01 2.08906606e-01 8.20785761e-01
9.14397895e-01 4.24242020e-01 -1.23930275e-01 -8.14594626e-02
1.20487356e+00 -4.92276549e-01 -1.01676857e+00 -2.97602594e-01
3.91048998e-01 -2.81341523e-01 9.87665176e-01 7.81238198e-01
-6.05657756e-01 -6.28587425e-01 -1.33059955e+00 2.64954269e-01
-9.87564862e-01 7.90043622e-02 3.24458897e-01 8.73295486e-01
-1.01607537e+00 4.05131161e-01 -3.04321438e-01 -5.12583673e-01
5.60558856e-01 7.13997841e-01 -6.38818502e-01 1.78507775e-01
-1.35459614e+00 1.32539117e+00 4.60958481e-01 -1.23519823e-01
-8.47967148e-01 -6.64541006e-01 -4.85418648e-01 1.26479454e-02
-3.27845700e-02 -1.40439287e-01 7.90299177e-01 -1.48466516e+00
-1.65808761e+00 8.02020371e-01 -8.05547908e-02 -3.99110347e-01
5.17702624e-02 -8.52903575e-02 -9.33301032e-01 -1.93375394e-01
4.18803543e-02 3.99790764e-01 9.44516242e-01 -1.10484874e+00
-4.26940382e-01 -8.14450502e-01 -3.12591165e-01 1.31975651e-01
-6.23892307e-01 1.47330225e-01 1.65300146e-01 -8.03709090e-01
6.38266131e-02 -6.85695767e-01 4.52128917e-01 -4.65553790e-01
-1.00838654e-01 -3.51920456e-01 8.15242708e-01 -7.70616710e-01
1.31001985e+00 -2.34497070e+00 2.43900955e-01 4.65649784e-01
1.27011582e-01 3.46141845e-01 -5.20891882e-02 -8.66962597e-02
-6.52045369e-01 -2.66876251e-01 -3.20550472e-01 3.47429156e-01
-1.84637308e-01 2.26982743e-01 -2.29595318e-01 2.07419127e-01
3.12234424e-02 6.20203614e-01 -4.69551742e-01 -1.30010378e-02
-2.15059891e-02 3.45349461e-01 -3.08186710e-01 4.70753551e-01
3.99086356e-01 5.94100237e-01 -8.28703269e-02 3.64697337e-01
5.71453512e-01 1.50199190e-01 -2.05973044e-01 -3.72363955e-01
1.77121967e-01 2.07593188e-01 -1.38142788e+00 1.95297623e+00
-2.86758631e-01 8.51962864e-01 -2.08203420e-01 -1.36272013e+00
1.33089602e+00 6.31397247e-01 4.50165033e-01 -7.93885887e-01
3.83101106e-01 4.81514901e-01 4.10064399e-01 -4.33839321e-01
-6.41414896e-02 1.19482197e-01 1.16088085e-01 3.66717577e-01
5.22974312e-01 8.78510326e-02 -3.53511751e-01 -1.47389080e-02
1.07020581e+00 1.03646025e-01 4.31998640e-01 -6.96067154e-01
6.70429647e-01 -2.03425780e-01 5.60800672e-01 5.77684402e-01
-3.19276631e-01 6.89527810e-01 9.38875899e-02 -4.41971362e-01
-5.87511122e-01 -1.02441573e+00 -6.36478424e-01 6.52085304e-01
-5.47327846e-03 -2.07697377e-01 -1.05596399e+00 -5.53010762e-01
-1.62531585e-01 8.63434076e-01 -7.47399986e-01 -6.04043901e-01
-2.10277244e-01 -5.15465617e-01 5.96715331e-01 6.53429449e-01
5.69452047e-01 -1.02674341e+00 -6.06743217e-01 1.88932523e-01
-1.32012859e-01 -6.24024272e-01 -4.04784381e-01 7.54599154e-01
-7.39904284e-01 -8.91962111e-01 -9.70885873e-01 -9.97428000e-01
5.64659774e-01 4.10831757e-02 7.63604164e-01 -5.58546245e-01
-3.43484402e-01 2.46152118e-01 -4.85987633e-01 -9.04004931e-01
-1.82806745e-01 1.24662876e-01 3.78351808e-01 1.77832827e-01
1.20806158e+00 -8.28429580e-01 -3.00420076e-01 5.33933938e-01
-6.14188612e-01 -1.65259331e-01 4.85691041e-01 9.12332058e-01
1.11371674e-01 -9.81613025e-02 1.14903843e+00 -6.36040449e-01
1.05170059e+00 -4.48226273e-01 -1.36538759e-01 5.82189620e-01
-5.39859533e-01 1.23262249e-01 4.14536446e-01 -7.89201915e-01
-1.12724185e+00 -1.32574648e-01 1.10947564e-01 -2.39124358e-01
-5.56566894e-01 1.36746898e-01 -5.61024904e-01 -5.24685860e-01
1.11979985e+00 5.53389549e-01 1.06891185e-01 -4.69880015e-01
-2.26313442e-01 1.72651613e+00 4.53230947e-01 -4.12235469e-01
4.61869627e-01 1.35906576e-03 -5.65811455e-01 -1.02369833e+00
-6.39142692e-01 -3.54392618e-01 -7.96784163e-01 -6.68420345e-02
6.09032452e-01 -9.43335235e-01 -2.11132094e-01 5.30110478e-01
-1.23616052e+00 -2.71193415e-01 -1.56179816e-01 5.96922338e-01
-4.91690695e-01 -8.78669918e-02 -5.69648445e-02 -5.07049918e-01
-3.47326547e-01 -1.20701766e+00 6.76809371e-01 1.38464525e-01
-5.67590654e-01 -6.65434420e-01 -5.49992546e-02 4.75091487e-02
6.03969574e-01 -2.66444474e-01 5.46410263e-01 -1.25212443e+00
-9.87245068e-02 -1.82947338e-01 -1.93284586e-01 7.79813528e-01
5.16358674e-01 -8.75024796e-01 -1.35835171e+00 -2.34530687e-01
5.24725914e-01 -1.40762612e-01 3.30410540e-01 2.41447136e-01
1.37911785e+00 1.99949350e-02 -3.32880080e-01 7.94369638e-01
1.21825564e+00 9.79897439e-01 1.10540986e+00 3.01892966e-01
4.21119958e-01 3.80779296e-01 1.72428608e-01 2.22806856e-01
-5.07872030e-02 5.39393723e-01 -4.96866792e-01 2.09778264e-01
-1.77980900e-01 1.11812130e-01 3.95927995e-01 5.77852190e-01
-7.95026943e-02 2.17266783e-01 -6.85031474e-01 2.81600773e-01
-1.24807572e+00 -7.95051098e-01 3.05700749e-01 2.18339729e+00
1.15579545e+00 5.71819255e-03 -1.16820455e-01 3.95020217e-01
5.21250129e-01 -5.63481987e-01 -6.77280188e-01 -4.01753724e-01
-1.51393309e-01 6.94790900e-01 3.13238978e-01 1.60281584e-01
-5.76418161e-01 7.31956184e-01 6.39934635e+00 8.13289881e-01
-1.26514745e+00 2.14517385e-01 1.74835026e-01 -4.77606505e-02
2.27777269e-02 -2.75769234e-01 -8.53288829e-01 7.92466700e-01
1.10108411e+00 -3.14834476e-01 6.61903203e-01 7.56566286e-01
-2.13243052e-01 7.17881769e-02 -1.66169298e+00 1.64662826e+00
6.76364720e-01 -8.81446123e-01 8.69353414e-02 2.34254077e-02
2.62475103e-01 -1.97374374e-01 -3.30485940e-01 5.23954451e-01
-3.29451501e-01 -1.09327483e+00 5.18365920e-01 6.25244021e-01
1.01605761e+00 -5.23583710e-01 6.05353236e-01 2.00660855e-01
-7.75496602e-01 -2.59812832e-01 -4.26873058e-01 -2.34506890e-01
-4.75695252e-01 -3.28791551e-02 -6.01694167e-01 2.60748416e-01
8.31960618e-01 8.11102867e-01 -8.04610908e-01 1.13261306e+00
-4.17540148e-02 5.15666783e-01 -1.90793529e-01 -1.11540802e-01
-2.96405941e-01 -2.44596556e-01 5.09073079e-01 9.35365021e-01
3.23791474e-01 2.94474542e-01 -3.26042682e-01 9.69878078e-01
3.65648389e-01 8.41621833e-04 -1.11014438e+00 1.86016798e-01
5.15572190e-01 6.57519042e-01 -3.41108531e-01 -2.41085112e-01
-4.08790439e-01 1.43780410e+00 2.41658419e-01 4.55875695e-01
-6.54122293e-01 -9.09339428e-01 5.01504362e-01 -3.45770363e-03
-3.32183152e-01 9.06428099e-02 -6.58489883e-01 -1.46034455e+00
1.55201298e-03 -1.09948230e+00 3.88256818e-01 -9.68172193e-01
-1.42380273e+00 9.73047078e-01 2.34976515e-01 -1.33926141e+00
1.46816494e-02 -9.67583954e-01 -3.22956830e-01 1.22609496e+00
-1.29499316e+00 -6.82664454e-01 -3.74760866e-01 1.09212494e+00
6.21869922e-01 -7.87836790e-01 1.17272604e+00 3.15602362e-01
-5.58990777e-01 9.01688576e-01 1.96729049e-01 2.21208841e-01
1.00023401e+00 -1.01597559e+00 -1.39633551e-01 5.66641808e-01
1.42483801e-01 1.01873195e+00 4.61850852e-01 -5.67612469e-01
-1.09701765e+00 -6.56319320e-01 5.90438366e-01 -4.99032587e-01
3.29360932e-01 -6.26491964e-01 -1.16092575e+00 7.23333299e-01
3.65926445e-01 -2.30831355e-01 1.05074310e+00 9.35333297e-02
-2.37144202e-01 -6.55247748e-01 -1.20242715e+00 4.86686081e-01
1.25707531e+00 -5.89848578e-01 -1.37850189e+00 7.83200413e-02
2.92367667e-01 -9.73036513e-02 -9.79497671e-01 1.23067766e-01
5.03658772e-01 -5.19192100e-01 7.67602444e-01 -5.81337214e-01
-9.70755816e-02 -2.64854450e-02 -2.81610131e-01 -1.86137927e+00
-3.42936099e-01 -4.66434270e-01 1.85987905e-01 1.02396393e+00
3.52364093e-01 -1.11400020e+00 3.87922257e-01 1.03567827e+00
-2.18127668e-01 -3.40128601e-01 -9.55895603e-01 -1.07580030e+00
1.34753942e-01 -3.48272562e-01 8.11098039e-01 7.73912013e-01
6.56406999e-01 4.70317632e-01 -1.81602105e-01 -4.60380241e-02
5.10621309e-01 -2.54358351e-01 5.06326675e-01 -1.59583724e+00
-8.36224947e-03 -1.37249246e-01 -8.53814781e-01 -6.93060458e-01
3.64579916e-01 -1.04368603e+00 -7.49052241e-02 -1.18646967e+00
1.01071432e-01 -5.13250470e-01 -8.16529572e-01 4.27508384e-01
2.32408866e-01 2.59798355e-02 -9.73332599e-02 1.18963249e-01
1.13948397e-02 3.99901420e-01 1.00651968e+00 -1.54582918e-01
-3.00841242e-01 -1.65684093e-02 -1.04757440e+00 5.33641160e-01
1.17008758e+00 -4.80708331e-01 -7.95369864e-01 -4.29236233e-01
-3.66749406e-01 -2.03677222e-01 2.85388798e-01 -1.53170061e+00
2.19782755e-01 1.98235922e-02 7.50220656e-01 -2.51242250e-01
2.88042217e-01 -1.07301247e+00 6.03470281e-02 2.68316865e-01
-4.53980267e-01 -3.26741993e-01 4.12119836e-01 4.00818169e-01
-2.04099894e-01 -3.98207605e-01 7.50975013e-01 -1.83000863e-01
-8.50301921e-01 1.13594793e-01 -3.77325624e-01 1.17130257e-01
1.13657558e+00 -9.67830122e-01 2.46035419e-02 -9.19962674e-02
-6.81071639e-01 -1.29557729e-01 1.53848544e-01 6.44875646e-01
8.61935079e-01 -1.21453714e+00 -4.71884817e-01 9.23208952e-01
3.43906254e-01 -3.92304629e-01 1.13083236e-02 4.14403528e-01
-3.35295461e-02 5.46164334e-01 -9.41962719e-01 -3.99344116e-01
-1.11268520e+00 3.31810772e-01 4.70090061e-01 3.87674779e-01
-7.08132744e-01 8.99749815e-01 2.33040959e-01 -2.74869978e-01
7.23082244e-01 -2.51027286e-01 -6.33814573e-01 -1.07121088e-01
8.74367177e-01 1.62003085e-01 2.68427253e-01 -4.43882912e-01
-4.32119459e-01 3.35370451e-01 -2.61245370e-01 -1.75277710e-01
1.44083667e+00 4.67046313e-02 -2.64725578e-03 6.48114264e-01
1.28847337e+00 -6.51682436e-01 -6.98208869e-01 -2.17302665e-01
-2.68114638e-02 -4.90776062e-01 -1.14626810e-03 -1.27239895e+00
-8.27551305e-01 9.03204262e-01 1.18592858e+00 -1.96705461e-01
1.36799550e+00 -3.99098575e-01 4.15194005e-01 5.05616724e-01
9.64366794e-01 -1.46622550e+00 -9.82810780e-02 2.59528250e-01
1.30670691e+00 -1.20615566e+00 -3.39603096e-01 -1.97127372e-01
-8.26104045e-01 1.16478348e+00 9.58297670e-01 -2.38945872e-01
9.93525147e-01 4.97520566e-01 -1.25169590e-01 -2.19401289e-02
-3.85066390e-01 2.78042108e-01 4.37182516e-01 1.07937062e+00
4.23764586e-01 1.28812909e-01 -4.46474254e-01 1.27734053e+00
-2.63405830e-01 5.54927468e-01 4.58429545e-01 1.01237893e+00
-3.56134623e-01 -1.14052546e+00 -4.87736672e-01 8.47840428e-01
-1.41343310e-01 -1.39608383e-01 -3.10778975e-01 4.68565822e-01
1.67618573e-01 1.00114632e+00 -1.25158634e-02 -6.26745462e-01
3.98640871e-01 4.60553944e-01 7.10690498e-01 -8.23467910e-01
-5.46946585e-01 -4.59244519e-01 -1.82939678e-01 -5.95983028e-01
-7.66023174e-02 -5.31611383e-01 -8.02110076e-01 4.37728882e-01
-4.18654621e-01 2.70571291e-01 6.79820955e-01 8.78466249e-01
4.52440262e-01 6.95294976e-01 4.15273905e-01 -5.31142652e-01
-7.16730475e-01 -1.36112499e+00 -9.66328442e-01 5.72009802e-01
7.97699317e-02 -1.08899605e+00 -3.35541695e-01 2.13962421e-01]
|
[13.128351211547852, 3.4553775787353516]
|
bcf3e15e-a4f6-4dfb-9809-8e464c1c97c5
|
entanglement-as-a-method-to-reduce
|
2302.05898
| null |
https://arxiv.org/abs/2302.05898v1
|
https://arxiv.org/pdf/2302.05898v1.pdf
|
Entanglement as a Method to Reduce Uncertainty
|
In physics, entanglement 'reduces' the entropy of an entity, because the (von Neumann) entropy of, e.g., a composite bipartite entity in a pure entangled state is systematically lower than the entropy of the component sub-entities. We show here that this 'genuinely non-classical reduction of entropy as a result of composition' also holds whenever two concepts combine in human cognition and, more generally, it is valid in human culture. We exploit these results and make a 'new hypothesis' on the nature of entanglement, namely, the production of entanglement in the preparation of a composite entity can be seen as a 'dynamical process of collaboration between its sub-entities to reduce uncertainty', because the composite entity is in a pure state while its sub-entities are in a non-pure, or density, state, as a result of the preparation. We identify within the nature of this entanglement a mechanism of contextual updating and illustrate the mechanism in the example we analyze. Our hypothesis naturally explains the 'non-classical nature' of some quantum logical connectives, as due to Bell-type correlations.
|
['Sandro Sozzo', 'Suzette Geriente', 'Lester Beltran', 'Jonito Aerts Argëlles', 'Diederik Aerts']
|
2023-02-12
| null | null | null | null |
['culture']
|
['speech']
|
[ 1.45739734e-01 7.42112398e-01 4.38712120e-01 4.71890345e-03
2.09228784e-01 -8.33496034e-01 8.95489573e-01 2.67329097e-01
-5.79384208e-01 1.04195452e+00 3.81195277e-01 -2.94835746e-01
-2.42817774e-01 -1.18689084e+00 -5.09482563e-01 -1.07836437e+00
-2.70186216e-01 4.32501435e-01 1.83281735e-01 -4.62003231e-01
2.24071085e-01 3.19097698e-01 -1.36474013e+00 6.15186058e-03
7.03293502e-01 5.43978274e-01 -1.02811009e-01 8.06141198e-01
-1.09034218e-02 1.07780528e+00 -5.47800243e-01 -6.41161442e-01
1.60790607e-01 -1.02558327e+00 -1.34314919e+00 -3.77970278e-01
-2.49541879e-01 2.49745309e-01 -5.40666103e-01 1.61976910e+00
-7.55812600e-02 1.71331823e-01 4.71679479e-01 -1.04494417e+00
-1.01415145e+00 9.05944824e-01 2.54485190e-01 4.16352749e-02
6.44281387e-01 -1.38121815e-02 1.24062884e+00 -9.69100595e-02
1.00951755e+00 1.00862432e+00 4.93258268e-01 5.37675977e-01
-1.72659516e+00 -5.61052747e-02 -5.30520082e-01 1.29033968e-01
-1.51011431e+00 -4.17504162e-01 3.23252618e-01 -4.47390884e-01
8.47018361e-01 5.64350545e-01 1.12526202e+00 6.36655807e-01
9.57026362e-01 1.16398007e-01 1.54555488e+00 -7.76955962e-01
6.72030032e-01 3.33016425e-01 3.60890836e-01 6.13778770e-01
8.06653082e-01 4.42642450e-01 -6.83209836e-01 7.43143260e-02
4.40959066e-01 -4.88422841e-01 -2.15888351e-01 -2.22651720e-01
-1.19704640e+00 3.31662536e-01 4.42577392e-01 1.00913191e+00
-4.67860699e-01 3.65854084e-01 -2.54020058e-02 4.82289553e-01
-6.94928318e-02 6.40209854e-01 -2.40592822e-01 -2.25555092e-01
-2.80320227e-01 1.53572902e-01 1.49069738e+00 7.51278162e-01
9.18489635e-01 -6.32703662e-01 2.74418473e-01 -3.11041117e-01
5.19359767e-01 5.72176218e-01 1.62258789e-01 -1.00027752e+00
-9.11199450e-02 4.29913104e-01 1.55646786e-01 -7.88778424e-01
-3.21492285e-01 -2.22350284e-01 -1.07524168e+00 1.72664553e-01
3.20465565e-01 -8.43208097e-03 -3.99742126e-01 2.33515716e+00
4.06406820e-02 -9.49023105e-03 5.51396728e-01 6.50632620e-01
6.41887367e-01 5.39697707e-01 5.72876520e-02 -6.89399302e-01
1.33446372e+00 -1.79582149e-01 -1.14982343e+00 3.60049427e-01
8.00766647e-01 -5.52336037e-01 1.40834048e-01 2.33753398e-01
-1.06886923e+00 -1.52201429e-01 -9.07197416e-01 6.48363903e-02
-6.72752023e-01 -7.20730126e-01 9.83440161e-01 8.54459643e-01
-1.19747066e+00 1.00374246e+00 -5.49762309e-01 -4.52176899e-01
-1.05559327e-01 3.08826894e-01 -7.83972800e-01 3.82592052e-01
-1.52366781e+00 1.52814770e+00 7.41445065e-01 4.91152436e-01
-3.76723766e-01 -1.58446521e-01 -3.79639953e-01 8.80660340e-02
1.17747322e-01 -8.56004119e-01 9.47665751e-01 -8.43107939e-01
-1.48802710e+00 1.01583982e+00 1.45429090e-01 -1.51745379e-01
7.71642476e-02 3.33752990e-01 -6.51134789e-01 -4.24676947e-02
-1.07758403e-01 3.45294178e-01 1.76820472e-01 -1.36510968e+00
-9.73458961e-02 -2.52649069e-01 4.37784642e-01 5.32176457e-02
4.88505185e-01 -3.73296171e-01 3.53998750e-01 3.31174999e-01
6.26494408e-01 -1.20649135e+00 -1.05347931e-01 -7.53507674e-01
-5.20244181e-01 -2.89305717e-01 8.39744881e-02 -7.63538629e-02
1.37047362e+00 -2.20824766e+00 5.76372087e-01 6.51067317e-01
5.62829673e-01 -6.65448830e-02 1.31426290e-01 9.22494411e-01
-4.10443217e-01 4.11413342e-01 -3.52256268e-01 1.06490128e-01
5.41523635e-01 4.30391848e-01 4.54129167e-02 6.14243686e-01
6.28316626e-02 7.86775470e-01 -1.18416011e+00 -5.77214360e-01
1.05403863e-01 1.97917208e-01 -5.14206111e-01 1.13291722e-02
1.66148856e-01 6.00312829e-01 -2.61199325e-01 1.92954779e-01
6.51137829e-01 -2.07007885e-01 7.06354439e-01 1.57721899e-02
-4.69022959e-01 4.45017189e-01 -1.24990463e+00 1.45150256e+00
-1.93765219e-02 6.23864472e-01 -6.91184998e-02 -6.36720359e-01
4.08117056e-01 4.15609539e-01 1.57255128e-01 -7.71639764e-01
4.30196196e-01 3.82609874e-01 9.35101032e-01 -5.68819284e-01
6.98408127e-01 -7.12713599e-01 -4.27788079e-01 7.04270601e-01
1.78282708e-01 -5.38552642e-01 3.86647314e-01 5.54652274e-01
1.45382881e+00 -1.12421021e-01 5.44387698e-01 -6.18609905e-01
7.81448364e-01 -5.43773890e-01 4.86694813e-01 9.40074801e-01
-4.98706549e-01 -2.06116170e-01 8.80252838e-01 -3.28895934e-02
-1.21546388e+00 -1.35907829e+00 -4.79453325e-01 3.49489748e-01
6.58687472e-01 -6.22572422e-01 -7.91480720e-01 1.05862997e-01
-1.85092404e-01 9.71662760e-01 -7.71891117e-01 -3.82551223e-01
-2.82502025e-01 -5.90993881e-01 3.47332329e-01 -1.97931230e-01
5.15827358e-01 -1.15704215e+00 -7.54009604e-01 1.93419963e-01
-1.53689802e-01 -7.90966094e-01 2.94014633e-01 4.96931493e-01
-7.07882106e-01 -8.75666499e-01 4.24982965e-01 -1.46257669e-01
6.25438094e-01 -3.74383420e-01 9.77518976e-01 4.34325159e-01
5.06956913e-02 2.91559100e-01 -4.16277677e-01 -1.15458526e-01
-9.84473944e-01 -3.52642834e-01 3.30841154e-01 -1.14799805e-01
2.31896520e-01 -7.31622875e-01 -2.88966477e-01 -5.51256061e-01
-1.33389664e+00 -1.67672690e-02 3.65522265e-01 6.78366899e-01
-1.20866589e-01 4.42018092e-01 -1.37886703e-02 -6.68626189e-01
8.13993335e-01 -4.60916072e-01 -3.43156129e-01 2.75025666e-01
-3.87925684e-01 6.52782917e-01 4.83918548e-01 3.89625393e-02
-1.08545470e+00 -4.22975659e-01 2.05123052e-01 5.68247139e-01
-2.42115539e-02 6.75939143e-01 -1.23649828e-01 -3.54327150e-02
5.20542979e-01 1.59149095e-01 2.08645500e-02 1.13664612e-01
5.89809716e-01 4.90553737e-01 5.82451165e-01 -7.43847251e-01
8.99531424e-01 4.87283856e-01 9.40851212e-01 -5.62319636e-01
-6.26457751e-01 -1.23367786e-01 -9.16420579e-01 -3.41978908e-01
9.98121381e-01 -3.99677604e-01 -1.41477299e+00 3.77939731e-01
-1.39186680e+00 4.79269288e-02 -8.35239589e-01 7.82667756e-01
-4.46184933e-01 4.65546846e-01 -6.81336284e-01 -1.27313781e+00
1.27951607e-01 -8.40731502e-01 4.04697865e-01 5.42100787e-01
-3.52168322e-01 -1.25382710e+00 5.49870014e-01 -1.93988025e-01
2.42386952e-01 1.22086681e-01 8.99791539e-01 -6.33889019e-01
-9.55276966e-01 -2.85457492e-01 -4.67147119e-02 3.43143642e-01
-1.74521208e-01 1.62738949e-01 -5.94146430e-01 5.08943573e-02
2.32979402e-01 9.88968164e-02 4.37610269e-01 -1.37743458e-01
5.93723841e-02 -1.82848647e-01 -6.50084838e-02 -8.24261829e-02
1.77198970e+00 1.83558241e-01 1.09414220e+00 1.48608640e-01
3.49654883e-01 6.77809238e-01 1.02834776e-01 1.69129938e-01
2.07364678e-01 3.57841641e-01 1.63571417e-01 6.45940304e-01
2.50710994e-01 8.00002888e-02 3.84346753e-01 1.54192758e+00
-7.71888316e-01 -6.02748757e-03 -5.97408295e-01 1.20289661e-01
-1.72486055e+00 -1.63562095e+00 -7.86254227e-01 2.37407875e+00
1.12735152e+00 1.38328865e-01 -5.67783356e-01 -1.61866680e-01
6.67525828e-01 -1.27003595e-01 3.63493741e-01 -8.70019674e-01
-3.57189208e-01 6.64481163e-01 4.52498794e-01 9.25266087e-01
-5.34872651e-01 9.26622450e-01 6.85737944e+00 3.17755878e-01
-4.69187677e-01 4.14104044e-01 -2.59251773e-01 2.31104702e-01
-6.55639350e-01 8.23177338e-01 -1.43964931e-01 8.01040292e-01
1.31690001e+00 -5.13395607e-01 7.44385839e-01 5.99431135e-02
-8.97607654e-02 -9.97418225e-01 -1.39415205e+00 7.09636211e-01
-2.85243183e-01 -1.12520206e+00 -4.37702209e-01 3.36689442e-01
7.27257371e-01 -3.56123865e-01 -4.77489501e-01 1.39444336e-01
4.70567346e-01 -8.77882838e-01 1.00073004e+00 1.13577700e+00
3.51713270e-01 -5.28041005e-01 1.10809553e+00 4.20066327e-01
-9.10231471e-01 2.97219306e-02 -3.15500237e-02 -6.28670394e-01
4.32356954e-01 8.18349838e-01 -1.23173445e-01 7.13287354e-01
1.86179206e-01 -1.94165111e-01 -3.50183368e-01 7.59280145e-01
-4.01593089e-01 5.65005355e-02 -4.63330179e-01 -3.22502583e-01
2.65012868e-02 -1.07927942e+00 7.57197261e-01 9.60137367e-01
1.39351860e-01 7.26842880e-01 -7.82058299e-01 1.24693549e+00
2.72013340e-02 -3.11630905e-01 -6.34731591e-01 -3.14682424e-01
4.74907905e-01 1.13527954e+00 -8.29208314e-01 -6.11303449e-01
8.16608146e-02 7.84949839e-01 2.35801995e-01 -6.55196011e-02
-8.25273573e-01 -2.81522959e-01 2.89231479e-01 -6.49680912e-01
-1.18108019e-01 -3.22515339e-01 -2.11889684e-01 -1.34752357e+00
-6.58630505e-02 -3.22187871e-01 -4.45907414e-01 -6.99849188e-01
-1.07908905e+00 3.62486780e-01 -1.79978788e-01 -7.88450301e-01
3.69328149e-02 -4.17322248e-01 -2.96154231e-01 9.88140047e-01
-7.44983673e-01 -5.04203200e-01 -5.37287723e-03 3.12576979e-01
-1.01147771e+00 2.75562316e-01 1.27627540e+00 4.65069152e-02
-4.68917131e-01 -1.01302370e-01 3.49108607e-01 -1.30704820e-01
2.75183171e-01 -1.54759789e+00 3.62404878e-03 1.09604609e+00
4.83544692e-02 1.10250914e+00 1.10439050e+00 -7.47459769e-01
-1.75475860e+00 -1.43808171e-01 1.58208990e+00 -6.92621112e-01
1.03122401e+00 -4.58588421e-01 -7.05081403e-01 6.54997826e-01
5.92305243e-01 -2.48316273e-01 8.12155604e-01 3.68941963e-01
-2.95724839e-01 -2.42610890e-02 -1.08376086e+00 6.25988960e-01
1.24233615e+00 -1.06015325e+00 -1.21339560e+00 3.59527022e-01
5.61422348e-01 -2.31690630e-01 -1.20453167e+00 2.73894519e-01
7.95034170e-01 -1.50244308e+00 4.02775496e-01 -5.95306933e-01
3.73515129e-01 -4.97302979e-01 -8.45851451e-02 -1.12257266e+00
-6.11695170e-01 -7.25843668e-01 2.17794087e-02 8.54589701e-01
2.00831220e-01 -1.07174492e+00 7.99902007e-02 1.03716385e+00
1.30564803e-02 -1.04244046e-01 -1.41082680e+00 -7.03054726e-01
9.83531326e-02 -2.70090699e-01 4.94311512e-01 1.11524308e+00
8.97384167e-01 5.12283802e-01 6.52651936e-02 5.21453321e-02
5.10604680e-01 -8.74788761e-02 2.52518743e-01 -1.23575318e+00
-3.54383975e-01 -5.69240391e-01 -7.68837094e-01 -5.59700966e-01
1.53662771e-01 -1.06620038e+00 3.42843145e-01 -1.13514972e+00
3.88486803e-01 -5.43398857e-01 -1.94256783e-01 -1.29991829e-01
3.13894928e-01 1.25514856e-02 5.28150618e-01 3.61053199e-01
-7.09289432e-01 2.39693746e-01 1.25747716e+00 3.21790904e-01
-9.81018320e-02 -5.96217453e-01 -7.21362770e-01 3.06108713e-01
4.17975694e-01 -5.57031512e-01 1.25477850e-01 2.96511024e-01
1.06662977e+00 3.00131172e-01 5.46316206e-01 -8.92499506e-01
3.31946343e-01 -2.19406579e-02 -2.31239676e-01 -2.02972978e-01
1.37709767e-01 -9.27366316e-01 1.08313751e+00 7.86946893e-01
-1.34262234e-01 -2.73289293e-01 -2.55254030e-01 4.05357927e-01
-2.73327883e-02 -7.11503625e-01 6.49528801e-01 -1.88847497e-01
-4.72532630e-01 -4.57330585e-01 -6.13609016e-01 -3.73268038e-01
1.00531220e+00 -1.11242458e-01 -7.12588310e-01 -1.42039536e-02
-9.62657392e-01 -3.16325814e-01 6.48638129e-01 -4.63916153e-01
2.34199613e-02 -1.31100452e+00 -2.40863070e-01 -6.11160211e-02
-3.55866432e-01 -2.77238548e-01 4.54052597e-01 1.49734986e+00
-8.63798559e-01 6.80861413e-01 -3.91349465e-01 -2.19602138e-01
-8.14503968e-01 5.31976938e-01 3.12353760e-01 -1.59348369e-01
-2.59280354e-01 6.89933717e-01 -3.67123298e-02 -1.96320713e-01
-5.79849660e-01 -4.38645363e-01 1.71938166e-01 2.02235319e-02
7.83304870e-02 1.41099751e-01 -2.98470259e-01 -1.03057647e+00
-5.04561841e-01 2.53821790e-01 2.83859223e-01 -4.64039564e-01
9.66212809e-01 -2.84663767e-01 -1.26379478e+00 1.04418623e+00
1.10046029e+00 3.53673071e-01 -1.62892833e-01 1.32645994e-01
-1.51114851e-01 -3.52252275e-01 -3.10708731e-01 -5.72872460e-01
-2.01819196e-01 3.61776620e-01 2.64671385e-01 1.35278237e+00
8.47247720e-01 3.57091457e-01 2.34557763e-01 4.64816749e-01
6.34700239e-01 -1.19932425e+00 -4.26320136e-01 7.14312494e-01
6.55944943e-01 -7.31640399e-01 8.37877393e-02 -4.00175422e-01
-1.82314247e-01 1.16704571e+00 -1.31965771e-01 -1.88659668e-01
8.32850218e-01 9.31420028e-02 -5.62916338e-01 -3.70942980e-01
-8.08172703e-01 -4.42636460e-01 -9.56061482e-02 2.16462165e-01
5.82723737e-01 7.75970757e-01 -9.03966308e-01 1.37182236e-01
-6.56616628e-01 -2.80886423e-02 1.12716293e+00 9.86435056e-01
-3.19669247e-01 -1.30395174e+00 -5.10782182e-01 1.83465332e-01
-4.17521596e-02 -2.39852265e-01 -3.81185442e-01 7.85605550e-01
8.89875472e-01 9.18706715e-01 2.54181415e-01 -4.19159144e-01
1.43767623e-02 2.11408630e-01 1.18206155e+00 -6.48129463e-01
-3.51432055e-01 -4.95085716e-01 2.33565569e-01 -4.82544214e-01
-1.14017415e+00 -9.47387159e-01 -1.46029484e+00 -1.11746156e+00
-6.53078198e-01 5.30321956e-01 5.58356822e-01 1.36348510e+00
2.93928832e-02 5.90466082e-01 3.22640657e-01 -3.46232176e-01
-4.76224124e-01 -9.63719428e-01 -1.25710368e+00 5.74377537e-01
1.87121391e-01 -5.84906995e-01 -8.65437448e-01 -7.08498582e-02]
|
[5.646490573883057, 4.915651321411133]
|
067e6e15-29d9-4cb4-8712-5b621e9feb3a
|
rawgment-noise-accounted-raw-augmentation
|
2210.16046
| null |
https://arxiv.org/abs/2210.16046v2
|
https://arxiv.org/pdf/2210.16046v2.pdf
|
Rawgment: Noise-Accounted RAW Augmentation Enables Recognition in a Wide Variety of Environments
|
Image recognition models that work in challenging environments (e.g., extremely dark, blurry, or high dynamic range conditions) must be useful. However, creating training datasets for such environments is expensive and hard due to the difficulties of data collection and annotation. It is desirable if we could get a robust model without the need for hard-to-obtain datasets. One simple approach is to apply data augmentation such as color jitter and blur to standard RGB (sRGB) images in simple scenes. Unfortunately, this approach struggles to yield realistic images in terms of pixel intensity and noise distribution due to not considering the non-linearity of Image Signal Processors (ISPs) and noise characteristics of image sensors. Instead, we propose a noise-accounted RAW image augmentation method. In essence, color jitter and blur augmentation are applied to a RAW image before applying non-linear ISP, resulting in realistic intensity. Furthermore, we introduce a noise amount alignment method that calibrates the domain gap in the noise property caused by the augmentation. We show that our proposed noise-accounted RAW augmentation method doubles the image recognition accuracy in challenging environments only with simple training data.
|
['Takeshi Ohashi', 'Atsushi Irie', 'Junji Otsuka', 'Masakazu Yoshimura']
|
2022-10-28
| null |
http://openaccess.thecvf.com//content/CVPR2023/html/Yoshimura_Rawgment_Noise-Accounted_RAW_Augmentation_Enables_Recognition_in_a_Wide_Variety_CVPR_2023_paper.html
|
http://openaccess.thecvf.com//content/CVPR2023/papers/Yoshimura_Rawgment_Noise-Accounted_RAW_Augmentation_Enables_Recognition_in_a_Wide_Variety_CVPR_2023_paper.pdf
|
cvpr-2023-1
|
['image-augmentation']
|
['computer-vision']
|
[ 7.74935126e-01 -3.40574443e-01 4.94967520e-01 -4.40835238e-01
-4.04619098e-01 -6.20088518e-01 3.38856220e-01 -3.02128702e-01
-5.65730512e-01 6.20458484e-01 -2.89277732e-01 -2.49304205e-01
-8.72458369e-02 -5.24212658e-01 -8.57208490e-01 -8.17960560e-01
3.42518985e-01 -1.44311875e-01 1.03218071e-01 7.09875673e-02
9.56463888e-02 5.23966014e-01 -1.71637690e+00 -6.36892021e-02
1.14917338e+00 1.10913110e+00 4.84606415e-01 7.66008794e-01
-2.12087244e-01 6.81720138e-01 -7.88318217e-01 -1.11857690e-01
5.98859012e-01 -5.05436480e-01 -2.85907686e-01 3.91726136e-01
4.71223116e-01 -5.71215749e-01 -1.86993718e-01 1.31411910e+00
3.58104676e-01 9.23040658e-02 2.54693121e-01 -1.27736890e+00
-7.07595468e-01 6.20244183e-02 -7.97735512e-01 -1.23162888e-01
-2.59639714e-02 5.00537992e-01 6.77553713e-02 -5.26195645e-01
3.00423563e-01 8.81043673e-01 5.68634450e-01 5.51859140e-01
-1.44774389e+00 -6.99759722e-01 -6.69377819e-02 2.22204074e-01
-1.21496356e+00 -4.81685281e-01 9.07846808e-01 -1.98190898e-01
5.20613968e-01 4.21696335e-01 4.81037557e-01 1.11386204e+00
-3.08499634e-01 2.66280323e-01 1.70707738e+00 -5.00295877e-01
3.69675189e-01 1.79966182e-01 3.51791154e-03 8.93102065e-02
5.53906202e-01 -3.59356813e-02 -2.72736043e-01 2.04335943e-01
8.72518659e-01 -1.79418504e-01 -5.07748604e-01 -1.63429052e-01
-1.18809855e+00 5.16566187e-02 4.29992139e-01 1.39852315e-01
-3.98594320e-01 2.03457639e-01 1.32897675e-01 1.73224375e-01
8.71732160e-02 5.73835790e-01 -5.45264602e-01 -3.62685233e-01
-9.31209147e-01 -2.27113396e-01 4.38318878e-01 9.06557977e-01
7.78754950e-01 2.27569878e-01 8.46324936e-02 8.55389655e-01
9.67076197e-02 7.66379058e-01 3.56536031e-01 -1.06281912e+00
3.12790185e-01 4.23100382e-01 4.65500802e-01 -9.72780943e-01
-3.07999849e-01 -3.26835126e-01 -9.89577770e-01 4.88429397e-01
8.17768693e-01 -1.52330667e-01 -1.17819488e+00 1.66315258e+00
1.50591075e-01 2.26820439e-01 1.49283126e-01 1.26276028e+00
3.68930101e-01 5.30703843e-01 -8.79294425e-02 -4.76429254e-01
1.34500194e+00 -6.58566236e-01 -9.96113539e-01 -2.93370366e-01
1.08997829e-01 -1.01571190e+00 1.58625090e+00 5.86037397e-01
-8.02731395e-01 -6.41354322e-01 -1.09044647e+00 1.41400909e-02
-2.06353769e-01 2.03790933e-01 4.79492009e-01 9.88959491e-01
-8.80375981e-01 3.16547215e-01 -6.88657165e-01 -3.96204263e-01
3.27353515e-02 2.14200258e-01 -2.83429682e-01 -2.49439687e-01
-8.78377438e-01 7.76158333e-01 1.96434945e-01 5.57950199e-01
-3.66615623e-01 -5.34595788e-01 -5.13725162e-01 -1.30537346e-01
4.21252549e-01 -4.31342661e-01 9.82639074e-01 -1.25334692e+00
-1.69751549e+00 3.99263829e-01 -2.68244855e-02 -2.28436336e-01
5.60904384e-01 -5.05575299e-01 -3.18355024e-01 3.13910320e-02
-4.50223267e-01 3.93950313e-01 1.01918101e+00 -1.61177635e+00
-8.14621821e-02 -2.19011977e-01 -6.47330508e-02 1.19450606e-01
-4.19001728e-01 -1.74025018e-02 -5.61194599e-01 -3.77709240e-01
3.28129172e-01 -8.80635440e-01 -3.08596283e-01 3.24710384e-02
-3.78425032e-01 7.48728216e-01 1.07423866e+00 -7.41453469e-01
8.23337436e-01 -2.38528943e+00 -5.27933180e-01 1.28714502e-01
-1.18644200e-01 5.31790912e-01 -2.20243052e-01 2.00686008e-02
-5.23729101e-02 -2.77649313e-02 -2.23775297e-01 -1.12878725e-01
-1.95805669e-01 2.95488507e-01 -8.73704329e-02 3.13221067e-01
3.56012851e-01 4.69925463e-01 -7.04158127e-01 -3.14055830e-01
5.05576253e-01 5.99849939e-01 -3.53802174e-01 3.75875056e-01
-2.65294492e-01 7.50404418e-01 -4.78661470e-02 5.57608843e-01
1.17365670e+00 -7.54430294e-02 1.80408228e-02 -7.36550093e-01
-1.87875256e-01 -5.88277355e-02 -1.50018418e+00 1.36380446e+00
-5.66364586e-01 6.59171999e-01 2.65771061e-01 -6.90044999e-01
1.06438756e+00 -1.05476201e-01 3.05588096e-01 -8.37070942e-01
2.06654966e-01 2.85861641e-01 2.19162717e-01 -6.98560059e-01
5.03322661e-01 9.22074914e-02 3.07829559e-01 2.92501479e-01
-4.36817378e-01 -3.82138610e-01 -2.53622420e-02 -2.22806916e-01
1.02159417e+00 2.34520659e-01 -3.67302261e-02 1.95428714e-01
3.55673850e-01 -1.50205329e-01 6.21438742e-01 7.91819513e-01
-2.45375991e-01 9.52592731e-01 2.41844043e-01 -2.15169400e-01
-1.41939867e+00 -9.16614234e-01 -3.87384254e-03 4.12734360e-01
3.70867670e-01 1.48959039e-02 -8.35214913e-01 -1.31597072e-01
-3.97982806e-01 5.42749107e-01 -1.50454015e-01 -1.38579652e-01
-4.48634893e-01 -1.06911981e+00 5.90223551e-01 4.08080012e-01
1.17270100e+00 -5.69352269e-01 -8.07742655e-01 2.99210865e-02
-1.24124400e-01 -1.57169461e+00 -3.78715843e-01 3.48323941e-01
-7.92370677e-01 -9.80064154e-01 -4.27449942e-01 -4.58227664e-01
9.78141546e-01 5.13271272e-01 7.02491581e-01 -4.27278318e-03
-2.95866728e-01 1.76472887e-01 -3.60193700e-01 -1.70523971e-01
-2.19405711e-01 -5.18469810e-01 1.22993290e-01 2.38622293e-01
2.24669293e-01 -5.12151122e-01 -8.51733983e-01 4.25116271e-01
-1.13699222e+00 3.50043505e-01 8.18603754e-01 9.26668584e-01
3.97771418e-01 2.80297667e-01 1.60056338e-01 -5.72130978e-01
4.18103367e-01 2.06002295e-01 -9.28729713e-01 2.64264435e-01
-6.49110317e-01 -2.22538766e-02 7.85558224e-01 -8.35606456e-01
-1.34635413e+00 3.88730705e-01 1.35081917e-01 -3.78867209e-01
-4.88027990e-01 7.01783821e-02 -3.25502604e-01 -3.05252373e-01
7.83949614e-01 1.89144388e-01 6.43192902e-02 -3.89353663e-01
3.99299055e-01 1.05107594e+00 7.42094278e-01 -4.80565995e-01
1.00379980e+00 3.60553354e-01 6.94161505e-02 -1.08337665e+00
-3.26757312e-01 -1.26026928e-01 -3.97030205e-01 -2.37261489e-01
5.63310623e-01 -7.56512284e-01 -7.16205955e-01 9.63107407e-01
-1.05175447e+00 -3.50602955e-01 -1.68982342e-01 6.78827167e-01
-2.48626351e-01 4.70400512e-01 -4.73396689e-01 -1.10500240e+00
-1.49376243e-01 -1.13124037e+00 7.79903769e-01 5.37309587e-01
2.76417524e-01 -2.89499909e-01 -4.58714485e-01 4.37388331e-01
8.90940309e-01 2.71688849e-01 6.51041150e-01 4.94333915e-02
-7.77674437e-01 -5.23308180e-02 -7.48540103e-01 7.81390727e-01
5.11247456e-01 1.83534890e-01 -1.31624222e+00 1.48750618e-02
3.82108212e-01 -2.50605673e-01 4.48134631e-01 2.68714577e-01
1.33405650e+00 -2.09657416e-01 2.53670871e-01 6.38262153e-01
1.65988004e+00 3.60304326e-01 9.98038948e-01 5.47077179e-01
7.98549592e-01 4.40468311e-01 5.51670492e-01 2.75594085e-01
1.12468963e-02 8.10869634e-01 3.79496187e-01 -5.15124261e-01
-3.45213205e-01 1.79021671e-01 2.62386769e-01 4.56371993e-01
-3.10002379e-02 -2.06898199e-03 -8.17362547e-01 3.33177865e-01
-1.47616136e+00 -6.24455035e-01 -5.08016586e-01 2.44420862e+00
1.08965421e+00 4.95294891e-02 -2.26530403e-01 2.93422252e-01
7.10881412e-01 -2.99974173e-01 -5.62166154e-01 -3.02895874e-01
-4.95087683e-01 6.62227795e-02 7.75268495e-01 3.01125675e-01
-7.48924494e-01 5.40874541e-01 5.72556925e+00 4.59120065e-01
-1.45521080e+00 -4.16279621e-02 6.98939145e-01 -9.35302973e-02
-1.01614483e-01 5.58498316e-02 -2.95872480e-01 6.10599756e-01
8.19168687e-01 1.57816768e-01 7.65529096e-01 6.98837996e-01
5.73121250e-01 -5.44195592e-01 -8.14755201e-01 1.22081542e+00
-5.26803173e-02 -7.31145442e-01 -3.84457201e-01 -1.44860178e-01
6.34536684e-01 -2.93017566e-01 9.74529162e-02 -1.94647521e-01
4.44155894e-02 -7.77056932e-01 4.90765691e-01 6.67791843e-01
8.27465951e-01 -2.38297641e-01 8.26456964e-01 1.79771677e-01
-7.03014553e-01 -1.78716704e-02 -3.17960203e-01 -1.22394539e-01
-2.84329057e-02 9.05172944e-01 -7.97826350e-01 2.84229785e-01
8.23161423e-01 -1.85252167e-03 -7.28647113e-01 1.26322377e+00
-1.71775937e-01 5.95443130e-01 -6.54744923e-01 1.24138623e-01
-2.13363856e-01 -4.73551154e-01 1.68900922e-01 9.93451774e-01
4.71127748e-01 3.76176946e-02 -1.88033491e-01 7.98907399e-01
4.23948132e-02 -2.19346583e-01 -5.14805734e-01 1.86367368e-03
4.10885811e-01 1.34287858e+00 -6.68031394e-01 -7.46906102e-02
-4.80144769e-01 1.13526022e+00 -1.74079880e-01 6.42634034e-01
-9.46539521e-01 -4.09794718e-01 6.40509665e-01 1.10330889e-02
-3.15186866e-02 -3.92261267e-01 -6.63595319e-01 -9.91895378e-01
3.22458446e-01 -1.06315053e+00 -2.04972208e-01 -1.24216437e+00
-1.13595247e+00 6.33257091e-01 -2.38816500e-01 -1.35606039e+00
7.55829290e-02 -6.11087561e-01 -3.02197039e-01 9.04165626e-01
-1.58268940e+00 -1.09921861e+00 -1.07831395e+00 5.04420519e-01
2.52543420e-01 3.84282619e-01 5.80887318e-01 5.68458080e-01
-5.90828121e-01 4.87395376e-01 1.87579691e-01 -4.68717664e-02
8.98635209e-01 -1.16879761e+00 2.30143279e-01 1.21803999e+00
-2.20432624e-01 6.79146230e-01 9.73779678e-01 -4.11056638e-01
-1.68267894e+00 -1.01034951e+00 1.85265183e-01 -9.68263000e-02
5.44663370e-01 -5.24278939e-01 -1.08619142e+00 2.48887926e-01
2.03439355e-01 2.30800495e-01 2.39540204e-01 -3.83006632e-01
-2.96993583e-01 -6.93145275e-01 -1.19368672e+00 6.04128003e-01
7.07891345e-01 -4.34300423e-01 -1.13223299e-01 2.06434906e-01
6.11204207e-01 -4.94214803e-01 -5.39487779e-01 3.98149103e-01
5.53792000e-01 -8.47634315e-01 7.99079657e-01 -1.15146801e-01
2.13827491e-01 -1.00943005e+00 -1.85822666e-01 -1.16919076e+00
3.75780128e-02 -6.67875767e-01 3.09193462e-01 1.56426048e+00
3.33258092e-01 -5.87248683e-01 5.98854244e-01 1.07088184e+00
5.27578071e-02 -1.21143311e-01 -5.83598971e-01 -9.44403708e-01
-5.55624902e-01 -5.22812307e-01 4.72642541e-01 8.04596603e-01
-3.81130785e-01 -1.10471666e-01 -6.60657823e-01 4.36218858e-01
7.01651454e-01 -5.15028611e-02 1.10688138e+00 -7.35470176e-01
-3.56254280e-01 7.54897743e-02 -3.50031674e-01 -8.61987054e-01
-5.29835343e-01 1.05216056e-01 3.96684319e-01 -1.33919060e+00
3.11357956e-02 -8.13120902e-01 -5.36552854e-02 2.55665183e-01
-3.37797880e-01 5.62018871e-01 2.53562540e-01 2.47600794e-01
-3.11846852e-01 3.07932824e-01 1.14139521e+00 -2.22069789e-02
-2.88009316e-01 -2.27883145e-01 -5.27123749e-01 6.62603915e-01
8.76943111e-01 -1.25544220e-01 -4.91096258e-01 -7.41591275e-01
1.38222560e-01 -2.72126555e-01 4.62002069e-01 -1.24240744e+00
3.38720411e-01 -2.23552987e-01 6.00924909e-01 -3.67568642e-01
3.69740337e-01 -1.22115231e+00 5.60294390e-01 2.03536138e-01
-3.99294123e-02 -8.07297379e-02 2.71783352e-01 3.22403133e-01
-1.82004482e-01 -2.08119914e-01 8.49771500e-01 -2.16868669e-02
-7.21192062e-01 -2.80579805e-01 -7.58332238e-02 -3.47947896e-01
9.78423476e-01 -5.01914978e-01 -7.70985544e-01 -3.81960928e-01
-2.46414378e-01 -1.68545514e-01 8.88754189e-01 1.52543530e-01
5.03834665e-01 -1.06668186e+00 -2.35574633e-01 4.10676569e-01
7.92258307e-02 7.90230259e-02 3.31977218e-01 9.46647108e-01
-7.42564619e-01 -1.34066418e-01 -3.50976795e-01 -6.67081237e-01
-1.15156281e+00 5.80686748e-01 2.90301174e-01 1.81561604e-01
-3.98806095e-01 5.39550841e-01 -1.00427568e-01 2.98829135e-02
3.33729148e-01 -6.03670895e-01 2.89516658e-01 -3.54107678e-01
6.04417682e-01 1.18676372e-01 3.14384967e-01 -3.32398206e-01
-2.61721224e-01 6.39770389e-01 1.43211976e-01 1.23669468e-02
1.12830305e+00 -4.11169469e-01 -1.06327765e-01 2.62799740e-01
8.93805444e-01 -6.53208494e-02 -1.45827198e+00 -2.76686456e-02
-9.44257975e-02 -8.62998903e-01 1.72154441e-01 -8.84467125e-01
-1.05015230e+00 7.14291930e-01 1.06056821e+00 2.41748720e-01
1.89020252e+00 -6.03660405e-01 6.57312155e-01 3.13825935e-01
2.74213821e-01 -1.20020270e+00 3.33525315e-02 2.29907542e-01
5.52685320e-01 -1.27399242e+00 -1.16346270e-01 -5.32121360e-01
-5.43809175e-01 1.15164316e+00 8.82285893e-01 2.71623939e-01
2.21995413e-01 6.79021358e-01 4.24290895e-01 3.50749999e-01
-3.57101917e-01 -3.46488416e-01 -5.27055450e-02 8.07006955e-01
1.93448767e-01 -2.58942157e-01 -1.99508712e-01 2.51613379e-01
-5.25624827e-02 2.93336451e-01 8.54370356e-01 8.64793122e-01
-2.91164070e-01 -9.04745340e-01 -9.13528085e-01 2.72090226e-01
-2.99162090e-01 -7.40208626e-02 -1.93670616e-01 6.31359935e-01
3.71834040e-01 1.20739937e+00 5.51993307e-03 -5.52283883e-01
5.25431514e-01 -6.21610023e-02 4.57491785e-01 -1.59884468e-01
-2.66575992e-01 3.15017611e-01 -5.92566375e-03 -5.31679094e-01
-5.67578077e-01 -4.45332229e-01 -9.70613539e-01 -2.87723035e-01
-4.89214122e-01 -2.38528192e-01 1.26134038e+00 7.48312533e-01
2.59993672e-01 5.23335278e-01 6.09542251e-01 -5.76806843e-01
-4.27533537e-01 -1.07041132e+00 -5.41495562e-01 7.64292598e-01
3.72842580e-01 -4.18228984e-01 -4.99540359e-01 3.97849351e-01]
|
[10.520033836364746, -2.497134208679199]
|
a7bf988e-5e2f-48c8-b8c7-56153bb46d40
|
detectorguard-provably-securing-object
|
2102.02956
| null |
https://arxiv.org/abs/2102.02956v3
|
https://arxiv.org/pdf/2102.02956v3.pdf
|
DetectorGuard: Provably Securing Object Detectors against Localized Patch Hiding Attacks
|
State-of-the-art object detectors are vulnerable to localized patch hiding attacks, where an adversary introduces a small adversarial patch to make detectors miss the detection of salient objects. The patch attacker can carry out a physical-world attack by printing and attaching an adversarial patch to the victim object. In this paper, we propose DetectorGuard as the first general framework for building provably robust object detectors against localized patch hiding attacks. DetectorGuard is inspired by recent advancements in robust image classification research; we ask: can we adapt robust image classifiers for robust object detection? Unfortunately, due to their task difference, an object detector naively adapted from a robust image classifier 1) may not necessarily be robust in the adversarial setting or 2) even maintain decent performance in the clean setting. To build a high-performance robust object detector, we propose an objectness explaining strategy: we adapt a robust image classifier to predict objectness for every image location and then explain each objectness using the bounding boxes predicted by a conventional object detector. If all objectness is well explained, we output the predictions made by the conventional object detector; otherwise, we issue an attack alert. Notably, 1) in the adversarial setting, we formally prove the end-to-end robustness of DetectorGuard on certified objects, i.e., it either detects the object or triggers an alert, against any patch hiding attacker within our threat model; 2) in the clean setting, we have almost the same performance as state-of-the-art object detectors. Our evaluation on the PASCAL VOC, MS COCO, and KITTI datasets further demonstrates that DetectorGuard achieves the first provable robustness against localized patch hiding attacks at a negligible cost (<1%) of clean performance.
|
['Prateek Mittal', 'Chong Xiang']
|
2021-02-05
| null | null | null | null |
['robust-object-detection']
|
['computer-vision']
|
[ 5.15494645e-01 1.59871295e-01 -5.83240413e-04 6.14411496e-02
-1.18522429e+00 -1.32912576e+00 3.81490171e-01 1.85480177e-01
-1.76102355e-01 8.54072496e-02 -4.34970081e-01 -3.26206923e-01
2.49823630e-01 -6.47698283e-01 -1.42557716e+00 -9.04059887e-01
-2.18786523e-01 -1.39268890e-01 7.31929302e-01 8.34948849e-04
2.01731458e-01 8.35146606e-01 -1.40761745e+00 4.12730634e-01
3.63913029e-01 1.12626481e+00 -3.41050833e-01 1.09541619e+00
7.12596238e-01 7.70244241e-01 -8.67646217e-01 -4.52681869e-01
8.50684881e-01 -1.15054935e-01 -6.86913490e-01 1.04388744e-01
9.70666289e-01 -5.15478730e-01 -4.77386564e-01 1.45853734e+00
1.76105067e-01 -3.70267332e-01 2.92580068e-01 -1.65174294e+00
-8.15359592e-01 6.87745869e-01 -6.07527792e-01 1.62113398e-01
3.19427580e-01 6.30056620e-01 8.09127569e-01 -6.02778316e-01
3.14805537e-01 1.12561619e+00 4.26748306e-01 7.62078404e-01
-1.40025008e+00 -8.33979785e-01 3.67947876e-01 -8.31797794e-02
-1.54305708e+00 -3.39349717e-01 6.48091972e-01 -2.07300097e-01
5.54082632e-01 7.23049641e-01 -8.80551189e-02 1.24524939e+00
2.42390856e-01 7.50978112e-01 1.19550741e+00 -1.19727934e-02
3.78129870e-01 2.56251663e-01 2.84537882e-01 6.10662103e-01
5.47428250e-01 6.79795504e-01 -2.08222687e-01 -6.57501042e-01
2.02403292e-01 1.43435970e-01 -4.06398535e-01 -2.74685472e-01
-1.15970397e+00 7.20967710e-01 8.60439539e-01 -1.40414238e-01
8.06237683e-02 3.49115133e-01 3.58355403e-01 3.25757444e-01
-8.46109763e-02 4.60472912e-01 -2.96605974e-01 7.02705920e-01
-4.22843903e-01 4.45432186e-01 7.45576680e-01 9.73733306e-01
5.63744247e-01 1.02286965e-01 -2.08448619e-01 -1.68013990e-01
1.24661662e-01 1.09438717e+00 -1.38562679e-01 -5.58074892e-01
2.16783836e-01 4.24763441e-01 1.30350009e-01 -1.02964735e+00
-1.18638976e-02 -3.63772631e-01 -6.08149946e-01 7.63344646e-01
2.27224812e-01 1.01401366e-01 -9.99211848e-01 1.77641070e+00
6.31893158e-01 2.83257604e-01 3.34301859e-01 1.06865358e+00
5.35634696e-01 5.21831989e-01 5.60233146e-02 6.18080869e-02
1.54332078e+00 -6.35792196e-01 -1.79993585e-01 -3.20878863e-01
3.22958738e-01 -6.99910283e-01 8.14326108e-01 4.35476452e-01
-6.42732084e-01 -6.75883889e-01 -1.49820507e+00 3.29246998e-01
-3.84901941e-01 -2.55628854e-01 1.97835445e-01 9.40573156e-01
-5.17715454e-01 2.93466181e-01 -8.06386113e-01 7.88010359e-02
5.23327351e-01 4.08440709e-01 -6.61034703e-01 1.77981351e-02
-7.76817977e-01 6.86174214e-01 3.89756739e-01 -1.96334153e-01
-1.66266930e+00 -6.71767294e-01 -8.44865620e-01 -9.76699963e-02
6.70079172e-01 -4.23400879e-01 1.10698307e+00 -1.09455454e+00
-9.28295910e-01 8.75926793e-01 2.98797667e-01 -7.54791439e-01
5.63257039e-01 -2.59778529e-01 -4.67833459e-01 4.26148862e-01
7.44229779e-02 5.29967904e-01 1.63244188e+00 -1.66935992e+00
-7.25387514e-01 -3.58308464e-01 4.92045224e-01 -4.84126508e-01
-1.25040919e-01 2.62330562e-01 -7.82815963e-02 -6.87827110e-01
-1.23817809e-01 -1.27510834e+00 -1.65273726e-01 4.81620997e-01
-8.74095917e-01 3.04522514e-01 1.38918591e+00 -3.36429656e-01
6.28955126e-01 -2.76512980e+00 -4.33100581e-01 2.63626248e-01
3.81353378e-01 4.61982012e-01 -3.38147968e-01 -5.78455478e-02
-1.94203332e-01 2.83183932e-01 -4.56864595e-01 -1.58946902e-01
2.37841718e-02 1.02748148e-01 -1.22276282e+00 1.17168236e+00
4.70549017e-01 8.33507359e-01 -7.59683788e-01 -1.54546708e-01
1.10150632e-02 3.31386685e-01 -5.49447060e-01 2.81595290e-01
-2.04428881e-01 1.02474824e-01 -5.20729542e-01 8.90908599e-01
1.06405616e+00 3.49673629e-02 -1.91352293e-01 -1.39586762e-01
1.92966446e-01 -2.23009616e-01 -1.41811609e+00 9.05645847e-01
2.43570022e-02 4.22111601e-01 3.74123693e-01 -5.39094388e-01
6.52884781e-01 2.05595553e-01 -1.02938965e-01 -4.29663509e-01
-5.14613055e-02 1.69723272e-01 -1.86631344e-02 -7.24343657e-02
3.55938286e-01 1.87777817e-01 -4.99610931e-01 5.32077849e-01
-2.38471091e-01 -2.76557496e-03 -5.08126378e-01 5.02085209e-01
1.47316599e+00 -1.45057216e-01 6.19053245e-02 -1.95656031e-01
4.74021494e-01 -5.70967905e-02 5.59271872e-01 1.36463594e+00
-3.90195072e-01 7.28980720e-01 3.93174827e-01 -4.99392092e-01
-7.68092632e-01 -1.51975119e+00 -1.13913208e-01 1.02223933e+00
5.75187624e-01 -1.67744011e-01 -8.66163552e-01 -1.22702098e+00
3.54272008e-01 3.51281345e-01 -7.68686116e-01 -5.48286974e-01
-5.36514997e-01 -3.33080173e-01 1.09557593e+00 5.73065400e-01
5.31992614e-01 -8.62482429e-01 -7.75856674e-01 -1.28454238e-01
2.19386145e-01 -1.29902792e+00 -7.06971049e-01 1.35994971e-01
-3.37820262e-01 -1.39013839e+00 -3.42325941e-02 -7.16807127e-01
9.22430634e-01 6.51517153e-01 7.53983557e-01 6.66324437e-01
-4.61738616e-01 4.04508293e-01 -3.25249434e-01 -6.71924055e-01
-7.40105629e-01 -2.76072383e-01 3.33821952e-01 2.87582666e-01
-1.18681766e-01 -1.88011855e-01 -5.69429636e-01 4.93825823e-01
-1.40278840e+00 -5.46855986e-01 4.53980803e-01 4.85149652e-01
6.33311033e-01 1.48293570e-01 3.53940457e-01 -6.74142599e-01
-1.57710053e-02 -2.38936052e-01 -1.20345068e+00 1.29417852e-01
-1.80487409e-01 -8.86952057e-02 7.95422614e-01 -9.42333281e-01
-4.77885634e-01 4.48286176e-01 3.16829830e-02 -5.68623185e-01
-3.48961204e-01 -2.94237107e-01 -5.68841457e-01 -6.08099520e-01
1.18431902e+00 2.16166914e-01 -2.89659858e-01 -1.45409688e-01
5.19288063e-01 4.98899877e-01 9.93417919e-01 -7.15043604e-01
1.83009839e+00 9.35028493e-01 9.02983472e-02 -4.11698580e-01
-8.10343504e-01 -3.12347174e-01 -2.53761321e-01 8.09522122e-02
6.70111895e-01 -8.64753664e-01 -1.09282017e+00 5.95548093e-01
-1.20292723e+00 -1.22201800e-01 -2.67253697e-01 -1.51772171e-01
-3.42092335e-01 4.28702503e-01 -4.52275634e-01 -7.87383199e-01
-4.25001174e-01 -1.35994089e+00 1.27149081e+00 6.28396645e-02
2.23296821e-01 -2.81512588e-01 -2.22469479e-01 1.71387792e-01
2.70266563e-01 6.74199224e-01 6.12040520e-01 -8.83646011e-01
-1.03011453e+00 -7.46690392e-01 -1.16646305e-01 4.98060226e-01
-1.27193391e-01 5.99470222e-03 -1.35699761e+00 -6.17925584e-01
2.02479288e-01 -3.94488633e-01 7.90144265e-01 -3.01042587e-01
1.16970241e+00 -8.77720475e-01 -3.56634468e-01 6.03302836e-01
1.54802370e+00 -1.23959348e-01 6.39483511e-01 3.21764529e-01
6.48020387e-01 2.23031461e-01 4.96404380e-01 7.35077858e-02
-2.51730651e-01 5.88773310e-01 8.53113949e-01 -1.82684362e-01
3.76701578e-02 -2.77970523e-01 7.61419535e-01 -3.43475461e-01
5.18122256e-01 -1.93275854e-01 -5.09812832e-01 4.99693960e-01
-1.71839774e+00 -9.61939454e-01 5.51984832e-03 2.33724833e+00
7.06784487e-01 4.59894657e-01 1.28878504e-01 2.39420071e-01
5.34125984e-01 2.13532239e-01 -6.55303717e-01 -2.90237069e-01
-3.64806950e-01 9.77834761e-02 9.53176498e-01 3.75832826e-01
-1.50712168e+00 8.51219237e-01 5.67196703e+00 6.77468359e-01
-1.15156066e+00 2.10812375e-01 4.16421056e-01 1.58210788e-02
1.02823116e-01 2.50420094e-01 -8.48653078e-01 2.60898024e-01
6.42550230e-01 -4.18791138e-02 3.62148255e-01 1.24695194e+00
-3.87311488e-01 1.80154860e-01 -1.32743227e+00 3.99875492e-01
2.33881310e-01 -1.25209856e+00 -1.83200371e-02 1.72563136e-01
5.69093823e-01 -9.65807810e-02 5.63313484e-01 1.83465496e-01
3.66330028e-01 -8.55952084e-01 1.20101404e+00 -1.16487943e-01
5.72917759e-01 -8.06908488e-01 5.33721626e-01 4.68134463e-01
-1.24300706e+00 -2.77835965e-01 -3.92563373e-01 2.47058928e-01
-2.36187518e-01 1.74485296e-01 -6.15587473e-01 3.69027108e-01
7.87172556e-01 -8.48513395e-02 -6.95246875e-01 9.40200150e-01
-5.73983431e-01 7.48153150e-01 -4.58536834e-01 5.00183702e-01
1.28443420e-01 7.29151964e-01 1.03324723e+00 1.17113578e+00
-2.63889730e-01 1.49424672e-01 5.73516130e-01 1.02679563e+00
-2.94581920e-01 -3.61038327e-01 -7.18117476e-01 4.12628591e-01
4.20841992e-01 1.25590050e+00 -6.94847763e-01 -1.78335354e-01
-8.88669342e-02 1.04403818e+00 4.33772467e-02 1.99562952e-01
-1.19855940e+00 -3.50254864e-01 9.22895908e-01 -7.95223266e-02
7.01256335e-01 -7.32224062e-03 -2.89283395e-01 -9.82608736e-01
3.56232911e-01 -1.32253182e+00 5.40957570e-01 -4.57374185e-01
-1.34726536e+00 5.34570873e-01 -1.69827402e-01 -1.25825214e+00
2.78769523e-01 -7.39311457e-01 -7.76884854e-01 4.81910080e-01
-1.29380631e+00 -1.40912783e+00 -2.52507068e-02 7.88675845e-01
1.46729022e-01 1.37321934e-01 8.32317531e-01 -1.23977959e-01
-4.50348914e-01 1.04175878e+00 -4.09094155e-01 5.27956724e-01
5.78428566e-01 -1.12890685e+00 6.79155648e-01 1.62847733e+00
4.01590437e-01 7.31871605e-01 7.84577072e-01 -6.17392242e-01
-1.88419700e+00 -1.58877003e+00 1.63080513e-01 -9.35829222e-01
7.76471376e-01 -7.03231156e-01 -1.14979959e+00 7.85713494e-01
-1.37026057e-01 7.25494385e-01 3.01759422e-01 -4.09538180e-01
-1.35866511e+00 -2.18583137e-01 -1.59537411e+00 6.22742295e-01
7.34056175e-01 -8.13733637e-01 -6.19446337e-01 4.80536312e-01
1.17875350e+00 -4.81057197e-01 -5.06069183e-01 4.92383868e-01
4.14567620e-01 -6.14021361e-01 1.13420212e+00 -5.03066599e-01
-1.96810678e-01 -1.00220156e+00 -4.16876465e-01 -6.58636272e-01
-3.29820484e-01 -9.68365908e-01 -4.12207156e-01 1.09383678e+00
3.48070264e-01 -7.75474310e-01 4.85005975e-01 5.02015114e-01
3.78768183e-02 -1.94417894e-01 -1.07652593e+00 -1.23849118e+00
-1.97228370e-03 -6.26974523e-01 7.02352285e-01 5.96311510e-01
-3.37982416e-01 -2.14467689e-01 -3.46784323e-01 1.44090438e+00
1.07419455e+00 2.41829991e-01 1.07921517e+00 -6.75493300e-01
-5.62768877e-01 -2.42538199e-01 -7.87389815e-01 -7.24512279e-01
2.66015887e-01 -6.38566375e-01 4.20228660e-01 -5.08170962e-01
4.23489958e-01 -2.43751869e-01 -4.16976601e-01 8.63979220e-01
-4.30718958e-01 7.13279784e-01 6.10701084e-01 1.83422923e-01
-6.62462294e-01 -4.39377688e-02 5.48706055e-01 -6.45975828e-01
2.66736317e-02 3.83734256e-02 -9.63719308e-01 6.35962725e-01
5.14175773e-01 -9.62542832e-01 4.47293408e-02 -1.91606745e-01
3.07760593e-02 -4.19813305e-01 1.15429556e+00 -1.04165006e+00
2.12617934e-01 -1.39312461e-01 2.59809256e-01 -2.49267414e-01
8.03633034e-02 -1.11634636e+00 -4.18481492e-02 7.83638239e-01
-1.05635233e-01 -1.76125497e-01 4.68665898e-01 8.06446791e-01
1.35358632e-01 1.54442936e-01 1.13330185e+00 2.55620092e-01
-5.28330386e-01 4.07336235e-01 -1.77256718e-01 -1.55711651e-01
1.35645866e+00 -2.06735492e-01 -8.95303249e-01 1.16964646e-01
-3.61279994e-01 -3.89948860e-02 8.20288479e-01 5.26596308e-01
7.62730122e-01 -1.04409325e+00 -7.93085217e-01 3.91388506e-01
5.20453334e-01 -1.01398475e-01 1.43976077e-01 4.87512678e-01
-2.17797503e-01 -1.70119375e-01 1.64983317e-01 -7.72936165e-01
-1.48636866e+00 1.38301790e+00 4.84501123e-01 1.99892506e-01
-8.28664064e-01 7.95221746e-01 6.66731238e-01 -7.20064342e-02
3.12585592e-01 -3.25316519e-01 5.29081881e-01 -7.00827122e-01
1.03562260e+00 -1.90712791e-02 -3.17070000e-02 -7.51673698e-01
-7.14587748e-01 4.11276788e-01 -2.63996124e-01 7.52582476e-02
8.40888619e-01 3.99375439e-01 3.47416028e-02 -2.49518916e-01
1.13991177e+00 4.62691993e-01 -1.39192820e+00 -1.62447631e-01
-2.17882290e-01 -5.74925423e-01 -1.73494756e-01 -6.69654131e-01
-1.03737891e+00 4.27377880e-01 7.90282786e-01 3.40751886e-01
1.09752500e+00 3.01323861e-01 7.08143890e-01 4.34406906e-01
5.64001739e-01 -4.74335492e-01 3.81500423e-02 1.19092718e-01
9.69547987e-01 -1.13436878e+00 7.00986339e-03 -4.69962448e-01
-4.67716247e-01 7.02713370e-01 5.69085956e-01 -4.73697692e-01
5.43376923e-01 6.35608852e-01 1.30441144e-01 -8.67075324e-02
-5.78897059e-01 -7.76536465e-02 3.21275383e-01 7.58296072e-01
-6.80955529e-01 1.55388758e-01 5.43000281e-01 5.97189963e-01
4.99152625e-03 -6.38722181e-01 4.93203133e-01 1.09397185e+00
-5.61572254e-01 -6.82403684e-01 -1.01403832e+00 -2.50808328e-01
-7.61675119e-01 5.08637242e-02 -6.84772134e-01 7.09302187e-01
2.26650819e-01 1.23344994e+00 -2.37628743e-01 -5.39284348e-01
4.81764972e-01 -4.63952601e-01 1.93945199e-01 -4.61378396e-01
-8.63702118e-01 -2.21322179e-01 -3.61201674e-01 -7.88285315e-01
4.29477106e-04 -3.09209228e-01 -1.35101986e+00 -1.79114491e-01
-4.83693182e-01 -1.93745911e-01 6.13157928e-01 8.86132419e-01
3.69784653e-01 4.14529406e-02 1.05665219e+00 -6.26078904e-01
-9.35775161e-01 -3.57028306e-01 -3.40446681e-01 5.78636587e-01
9.02407825e-01 -2.53299385e-01 -7.47447610e-01 1.05117492e-01]
|
[5.552002906799316, 7.920490741729736]
|
67707be0-31c1-4644-932d-d0017ce1c27b
|
3d-point-cloud-classification-and
|
1711.08241
| null |
http://arxiv.org/abs/1711.08241v1
|
http://arxiv.org/pdf/1711.08241v1.pdf
|
3D Point Cloud Classification and Segmentation using 3D Modified Fisher Vector Representation for Convolutional Neural Networks
|
The point cloud is gaining prominence as a method for representing 3D shapes,
but its irregular format poses a challenge for deep learning methods. The
common solution of transforming the data into a 3D voxel grid introduces its
own challenges, mainly large memory size. In this paper we propose a novel 3D
point cloud representation called 3D Modified Fisher Vectors (3DmFV). Our
representation is hybrid as it combines the discrete structure of a grid with
continuous generalization of Fisher vectors, in a compact and computationally
efficient way. Using the grid enables us to design a new CNN architecture for
point cloud classification and part segmentation. In a series of experiments we
demonstrate competitive performance or even better than state-of-the-art on
challenging benchmark datasets.
|
['Michael Lindenbaum', 'Yizhak Ben-Shabat', 'Anath Fischer']
|
2017-11-22
| null | null | null | null |
['3d-part-segmentation']
|
['computer-vision']
|
[-4.46591794e-01 -4.20589477e-01 9.68359262e-02 -2.52868503e-01
-6.76373959e-01 -4.71890539e-01 5.57354212e-01 1.68242186e-01
-2.66221613e-01 2.47784048e-01 -3.80466461e-01 -3.97041887e-01
1.29578769e-01 -1.04869008e+00 -9.92845774e-01 -4.22655761e-01
-1.74725503e-01 6.95870101e-01 3.50090832e-01 -1.17385417e-01
1.79870293e-01 1.27110672e+00 -1.58230257e+00 7.65694603e-02
6.76178515e-01 1.36081624e+00 1.08815446e-01 7.42959380e-02
-7.13999271e-01 -9.81409848e-02 -5.10292172e-01 -3.37571383e-01
5.78393817e-01 3.77980798e-01 -6.62915170e-01 1.36337772e-01
6.42246008e-01 -7.36827999e-02 -1.51142433e-01 9.66280520e-01
3.07428718e-01 1.28224939e-01 8.51279855e-01 -1.23915863e+00
-6.17180049e-01 1.37798741e-01 -7.01896727e-01 -9.44749340e-02
6.92062601e-02 -1.70529917e-01 7.98003018e-01 -1.14190102e+00
4.06821221e-01 1.36210108e+00 1.10922134e+00 2.64288217e-01
-1.15493155e+00 -6.84754729e-01 9.73758176e-02 2.22388823e-02
-1.61014271e+00 2.66336828e-01 9.87123311e-01 -5.59393883e-01
1.29060400e+00 3.68502051e-01 1.04095089e+00 7.44545162e-01
1.75129965e-01 8.71234894e-01 8.87753963e-01 -1.42947450e-01
2.89772123e-01 -4.50351626e-01 1.84430867e-01 6.81208074e-01
2.28683576e-01 -9.98963416e-02 -1.27655968e-01 -3.58437091e-01
1.15425861e+00 3.25000316e-01 -6.11479953e-02 -5.89940667e-01
-9.66488421e-01 9.49400425e-01 8.43631506e-01 2.94138640e-01
-2.89269626e-01 3.53175312e-01 1.64208516e-01 -5.38793355e-02
8.46814513e-01 6.04694039e-02 -4.99364763e-01 3.01991589e-02
-9.95464742e-01 5.68515420e-01 5.14382899e-01 9.59356964e-01
8.83996964e-01 9.02296752e-02 -2.79703625e-02 8.21680486e-01
4.71816510e-01 5.04017234e-01 2.73070276e-01 -6.61751032e-01
2.36851454e-01 9.84612167e-01 -3.21845673e-02 -1.00091052e+00
-4.90215123e-01 -6.26034200e-01 -1.16645885e+00 5.41438878e-01
9.44378525e-02 4.31139588e-01 -1.17649484e+00 1.19531095e+00
5.63502848e-01 4.85800713e-01 -3.65911245e-01 8.34307671e-01
1.01276755e+00 6.72701001e-01 -1.76223427e-01 3.47019434e-01
1.24659228e+00 -5.41803896e-01 -2.32444048e-01 8.60754848e-02
3.32053363e-01 -5.90641499e-01 8.76084685e-01 3.71032566e-01
-1.14592969e+00 -7.17325628e-01 -1.13116145e+00 -3.53963703e-01
-4.16986465e-01 4.75772172e-02 8.40781212e-01 5.27849674e-01
-1.09618235e+00 8.44154358e-01 -1.09515953e+00 6.68530911e-02
9.22033310e-01 5.88647127e-01 -5.26984274e-01 1.59278572e-01
-5.68957865e-01 6.38233066e-01 1.52725279e-01 7.60327429e-02
-1.77335039e-01 -7.95583248e-01 -1.08290589e+00 2.90914446e-01
-1.79379657e-01 -7.93487251e-01 1.16891456e+00 -1.66185156e-01
-1.39600575e+00 1.06820762e+00 -3.51028889e-02 -3.66968662e-01
5.60918927e-01 -2.32798532e-01 1.39363468e-01 -1.04356125e-01
8.80239084e-02 7.72886872e-01 8.29192340e-01 -1.32392156e+00
-5.29208064e-01 -6.81879759e-01 -1.86661705e-01 -6.78973198e-02
-9.59259085e-03 -3.03482920e-01 -6.00601792e-01 -6.31172061e-01
7.47959197e-01 -8.08166146e-01 -3.48734826e-01 2.56623626e-01
-3.29049498e-01 -7.78060555e-01 9.11738336e-01 -1.43763959e-01
6.09861195e-01 -2.15857005e+00 -5.88157680e-03 3.06868136e-01
5.41365683e-01 3.49215031e-01 1.15144126e-01 1.14338353e-01
-2.30443478e-01 3.28124404e-01 -4.23558056e-01 -7.13182867e-01
1.50113687e-01 2.23379195e-01 -4.00715828e-01 6.06181026e-01
5.86631060e-01 1.10083234e+00 -5.96369624e-01 -3.30407530e-01
4.08384860e-01 8.78858924e-01 -6.49009228e-01 -8.52980018e-02
-1.03320248e-01 1.41139731e-01 -4.96068150e-01 6.69501543e-01
1.37469888e+00 -4.35283571e-01 -4.73823547e-01 -2.83265889e-01
-2.08340719e-01 1.23024024e-01 -1.15399575e+00 1.81542766e+00
-2.55407333e-01 3.70371938e-01 -9.25234184e-02 -1.06647897e+00
1.28760183e+00 -4.65806611e-02 7.42273450e-01 -3.45259756e-01
2.96558052e-01 2.68525273e-01 -4.07632053e-01 1.97854772e-01
3.89403045e-01 -1.48019776e-01 3.18611553e-03 9.41330567e-02
7.42602274e-02 -5.64986289e-01 -2.90960610e-01 -2.36971140e-01
8.09979916e-01 8.16867203e-02 3.62669677e-02 -3.88826758e-01
2.77171671e-01 4.11103331e-02 4.17118996e-01 4.99081552e-01
2.90111303e-02 8.85295808e-01 2.61084855e-01 -7.83492386e-01
-8.35501134e-01 -1.27168453e+00 -5.13787329e-01 1.42874360e-01
1.38251990e-01 -3.40914160e-01 -4.91647094e-01 -5.35970986e-01
6.09781504e-01 3.34522724e-01 -5.60510635e-01 1.41517110e-02
-8.43080580e-01 -4.82192427e-01 3.31866175e-01 7.00849593e-01
5.56465626e-01 -6.59503639e-01 -5.91632009e-01 2.14793116e-01
2.09333301e-01 -1.13022518e+00 -2.15672210e-01 2.83063710e-01
-1.34663570e+00 -9.22976732e-01 -7.77959347e-01 -9.25326586e-01
4.94127989e-01 5.79363406e-01 1.28844738e+00 1.95563182e-01
-1.95216179e-01 2.72027347e-02 -2.58698344e-01 -5.79062879e-01
2.92492397e-02 1.31854355e-01 -1.20703638e-01 -9.45676193e-02
4.75284666e-01 -7.46541321e-01 -4.98641759e-01 9.59616378e-02
-8.52579236e-01 -6.24135993e-02 4.26932275e-01 6.80763245e-01
1.09261286e+00 -7.07092807e-02 2.57762205e-02 -4.78433400e-01
3.56522262e-01 -2.76972890e-01 -8.06452334e-01 -3.50582242e-01
-1.15888506e-01 5.61096985e-03 3.94413561e-01 -1.75735667e-01
-4.35538530e-01 3.08112174e-01 -7.05628812e-01 -1.10984325e+00
-2.90431887e-01 2.78978646e-01 4.33381385e-04 -4.97043580e-01
3.61349344e-01 1.38580829e-01 1.42951787e-01 -9.83049631e-01
2.96951026e-01 4.98275250e-01 4.13309813e-01 -5.95565557e-01
1.00362134e+00 8.76570344e-01 3.60175580e-01 -8.17354739e-01
-5.00919223e-01 -5.07021904e-01 -8.87460887e-01 -8.84717405e-02
9.56140637e-01 -8.32308650e-01 -9.23254013e-01 5.50649941e-01
-1.51317883e+00 -3.00075933e-02 -3.25723052e-01 3.00195783e-01
-5.87317348e-01 3.93467933e-01 -5.21209896e-01 -4.81339425e-01
-4.24352825e-01 -1.30376565e+00 1.58212769e+00 -1.55674131e-03
1.43650502e-01 -6.32645369e-01 2.60068271e-02 1.44172803e-01
1.68478280e-01 5.88190079e-01 1.04176056e+00 -4.33004439e-01
-8.13241839e-01 -4.16267574e-01 -4.58002985e-01 3.12821448e-01
-8.70206729e-02 5.51881976e-02 -9.14857924e-01 -1.93242714e-01
2.24229977e-01 9.12098289e-02 1.00193751e+00 3.78571212e-01
1.72762573e+00 1.46956757e-01 -4.73048210e-01 1.07736838e+00
1.56402469e+00 1.57096852e-02 4.84142542e-01 1.49182618e-01
8.05327356e-01 -1.30859725e-02 2.29229465e-01 3.74456018e-01
3.80999893e-01 7.15106606e-01 6.68542862e-01 -1.58554718e-01
-7.11481348e-02 -1.75436527e-01 -2.79400408e-01 1.00693893e+00
-2.50167996e-01 2.56410218e-03 -1.10070539e+00 3.71165395e-01
-1.69888675e+00 -6.68995559e-01 -4.07733232e-01 1.99868405e+00
4.16082621e-01 1.75525352e-01 -1.27631038e-01 3.30030531e-01
5.13616323e-01 5.59013039e-02 -3.40128303e-01 -2.20910624e-01
-1.04204692e-01 7.56652832e-01 4.22614247e-01 2.12981001e-01
-1.36388254e+00 8.38624835e-01 6.59991264e+00 1.05320776e+00
-1.31586742e+00 8.32317397e-02 5.16433120e-01 2.57973522e-01
-2.25793675e-01 -4.33408231e-01 -8.12639773e-01 3.20114762e-01
3.55560392e-01 2.01869935e-01 -7.30308332e-03 1.01434433e+00
-3.39203596e-01 3.79213423e-01 -9.99959469e-01 1.55112362e+00
-1.17628790e-01 -1.80605817e+00 2.93075383e-01 2.70005047e-01
5.78410625e-01 5.00891447e-01 1.13577671e-01 1.98883876e-01
1.23748012e-01 -1.17230964e+00 9.98218596e-01 3.29934120e-01
8.06900620e-01 -9.29165423e-01 6.08293295e-01 3.98853928e-01
-1.33088112e+00 2.55422324e-01 -7.79252291e-01 -1.17538437e-01
-3.85144874e-02 6.78284228e-01 -6.09130979e-01 5.81680238e-01
1.05249822e+00 8.07982326e-01 -4.07182544e-01 1.41450524e+00
3.09998333e-01 2.52903610e-01 -6.70732260e-01 1.00009469e-03
4.61107612e-01 -3.91819090e-01 5.59058964e-01 1.10398507e+00
6.10661268e-01 8.18499401e-02 2.68603712e-01 1.25312614e+00
-1.90427557e-01 5.46318144e-02 -7.67143905e-01 2.01428950e-01
2.93392420e-01 1.14387929e+00 -1.04837322e+00 -1.88829482e-01
-4.64531362e-01 7.04447865e-01 4.54952478e-01 -2.06183493e-02
-5.11831045e-01 -2.70817101e-01 1.05057132e+00 7.09981993e-02
6.82549059e-01 -6.93276823e-01 -6.29846454e-01 -1.10285079e+00
2.17563957e-01 -3.22336435e-01 -3.34445499e-02 -7.09574223e-01
-1.59717834e+00 7.10458636e-01 7.13506714e-02 -1.50244534e+00
2.40206182e-01 -9.98415351e-01 -6.17618561e-01 8.96158338e-01
-1.67684174e+00 -1.15441287e+00 -5.09364307e-01 6.03152514e-01
3.85256678e-01 -5.38484193e-03 7.71916866e-01 4.16518301e-01
-9.30691659e-02 3.60150397e-01 1.76864192e-01 2.89490491e-01
-1.90914162e-02 -1.42654884e+00 1.10575378e+00 3.53210479e-01
3.71725768e-01 3.85336667e-01 1.97157070e-01 -6.11108780e-01
-1.63012302e+00 -1.12692857e+00 5.81141353e-01 -4.14237469e-01
1.98173344e-01 -6.41108513e-01 -1.14926326e+00 4.38479006e-01
-2.95432180e-01 1.71466008e-01 5.54165959e-01 -1.12047262e-01
-5.01174331e-01 -6.80894852e-02 -1.30052090e+00 3.92141014e-01
1.16044390e+00 -2.93405533e-01 -5.76406360e-01 5.12012899e-01
9.39477086e-01 -7.04131484e-01 -8.86720002e-01 6.56890035e-01
1.89864099e-01 -9.10154998e-01 1.30150974e+00 -3.98988754e-01
9.63987857e-02 -3.56128216e-01 -2.79609591e-01 -1.21946025e+00
-6.33634925e-01 -3.77824724e-01 -3.46364863e-02 6.56849742e-01
-7.90995508e-02 -6.44127011e-01 1.15221977e+00 2.41729096e-01
-6.52889967e-01 -1.17446566e+00 -1.40013301e+00 -8.08252752e-01
4.74637538e-01 -8.42500567e-01 1.04875219e+00 7.88867414e-01
-5.45643926e-01 -7.65744671e-02 4.06553268e-01 1.30802304e-01
7.48893023e-01 3.88813883e-01 6.50870562e-01 -1.76245368e+00
1.30528465e-01 -9.19056714e-01 -1.02986073e+00 -1.24141216e+00
2.23042279e-01 -1.25514007e+00 -2.24751130e-01 -1.68223190e+00
-1.37119427e-01 -6.46294236e-01 -9.40536782e-02 4.41137314e-01
1.31263569e-01 5.66676080e-01 3.40619326e-01 3.02355975e-01
-1.57341003e-01 8.47107232e-01 1.36976075e+00 -3.91173452e-01
-5.44340648e-02 2.06765190e-01 -2.75783718e-01 7.77727723e-01
6.25839114e-01 -3.89245421e-01 -3.02504022e-02 -7.52165616e-01
5.32605350e-02 -1.51283056e-01 5.52263081e-01 -1.17439997e+00
-2.25192532e-02 2.67543912e-01 7.00151682e-01 -1.40928829e+00
7.27689624e-01 -9.64862406e-01 1.50017858e-01 3.05218399e-01
4.23224688e-01 3.01468343e-01 5.44026494e-01 4.51407731e-01
-3.62080961e-01 -7.82342777e-02 7.81721830e-01 -2.16228530e-01
-3.84631395e-01 7.38013983e-01 1.09882444e-01 -3.07612419e-01
8.37055206e-01 -4.84260499e-01 1.18435599e-01 1.46186367e-01
-5.34943044e-01 6.61886111e-02 5.53943574e-01 4.37987775e-01
9.81815219e-01 -1.72716486e+00 -5.98036826e-01 5.73897183e-01
-1.40871048e-01 6.58537924e-01 1.98970646e-01 3.88262182e-01
-9.47580814e-01 4.85602766e-01 -3.34684849e-01 -1.19597363e+00
-8.59603345e-01 3.39890510e-01 3.78973216e-01 5.33060022e-02
-1.03149962e+00 1.02920520e+00 1.98503494e-01 -5.72996438e-01
2.24873081e-01 -8.70087802e-01 -7.05084950e-02 -1.09664530e-01
1.69087797e-01 1.21380568e-01 6.00911200e-01 -6.94451332e-01
-3.44924867e-01 9.27315295e-01 1.08447619e-01 2.79986650e-01
1.66605031e+00 6.22230291e-01 -2.59674489e-01 3.84438574e-01
1.46806753e+00 -3.61205488e-01 -1.20082819e+00 -1.55197546e-01
-1.59121349e-01 -7.66232073e-01 1.88493177e-01 -2.05197453e-01
-1.26782906e+00 1.05715299e+00 6.84532940e-01 3.34741682e-01
7.54243553e-01 1.89809754e-01 1.10501468e+00 3.87351543e-01
5.72910905e-01 -4.46344614e-01 -2.63726681e-01 8.49385500e-01
1.02638710e+00 -1.13949227e+00 -1.39156282e-01 -6.18310988e-01
1.65850818e-02 1.20593131e+00 3.86889696e-01 -7.89295971e-01
1.21296406e+00 1.78772286e-01 -7.58219212e-02 -4.62052196e-01
-2.96571225e-01 -1.35060981e-01 5.93148947e-01 6.54987574e-01
1.88567787e-01 1.81718215e-01 3.04735564e-02 6.97919130e-01
-4.22421187e-01 -2.35500872e-01 6.55110227e-04 8.85630548e-01
-3.64140630e-01 -1.13236356e+00 -4.27931905e-01 6.43016875e-01
-2.72660792e-01 2.48547748e-01 -2.34490767e-01 8.46229374e-01
2.42854923e-01 2.83527613e-01 6.26734078e-01 -3.96897137e-01
4.18319851e-01 4.03242931e-02 5.46119690e-01 -5.75044036e-01
-6.07627630e-01 1.46033615e-01 -6.95809245e-01 -6.52667284e-01
-3.25348049e-01 -6.12966061e-01 -1.43100059e+00 -2.81596363e-01
-3.37243915e-01 9.43642780e-02 1.06082833e+00 7.47013092e-01
5.72454333e-01 2.61409163e-01 5.35948038e-01 -1.53433120e+00
-5.97082555e-01 -7.41376817e-01 -5.79056680e-01 4.16486561e-01
3.22000653e-01 -1.08736885e+00 -9.06002000e-02 -4.64260131e-01]
|
[7.957791328430176, -3.612375497817993]
|
c3e03246-6ee3-4719-9d9d-700fc43f28ab
|
an-accurate-non-accelerometer-based-ppg
|
2106.11512
| null |
https://arxiv.org/abs/2106.11512v1
|
https://arxiv.org/pdf/2106.11512v1.pdf
|
An Accurate Non-accelerometer-based PPG Motion Artifact Removal Technique using CycleGAN
|
A photoplethysmography (PPG) is an uncomplicated and inexpensive optical technique widely used in the healthcare domain to extract valuable health-related information, e.g., heart rate variability, blood pressure, and respiration rate. PPG signals can easily be collected continuously and remotely using portable wearable devices. However, these measuring devices are vulnerable to motion artifacts caused by daily life activities. The most common ways to eliminate motion artifacts use extra accelerometer sensors, which suffer from two limitations: i) high power consumption and ii) the need to integrate an accelerometer sensor in a wearable device (which is not required in certain wearables). This paper proposes a low-power non-accelerometer-based PPG motion artifacts removal method outperforming the accuracy of the existing methods. We use Cycle Generative Adversarial Network to reconstruct clean PPG signals from noisy PPG signals. Our novel machine-learning-based technique achieves 9.5 times improvement in motion artifact removal compared to the state-of-the-art without using extra sensors such as an accelerometer.
|
['Fadi Kurdahi', 'Amir M. Rahmani', 'Hadi Khodabandeh', 'Seyed Amir Hossein Aqajari', 'Amir Hosein Afandizadeh Zargari']
|
2021-06-22
| null | null | null | null |
['photoplethysmography-ppg', 'heart-rate-variability']
|
['medical', 'medical']
|
[ 3.94199699e-01 -2.29192488e-02 2.68584579e-01 8.03093910e-02
-4.59005356e-01 -3.29101712e-01 -2.04067245e-01 -2.31408879e-01
-2.59575993e-01 8.84518266e-01 2.70038396e-01 -1.01343147e-01
2.12457702e-01 -4.63495463e-01 -4.30449516e-01 -8.36850584e-01
1.50442824e-01 -3.66344601e-01 -3.28959413e-02 2.89457500e-01
-6.58416003e-02 2.70885855e-01 -1.01383352e+00 -4.05206710e-01
9.28460360e-01 9.82363224e-01 -3.16024780e-01 7.04734683e-01
4.23703104e-01 3.09617490e-01 -1.01717436e+00 1.01617336e-01
3.43336284e-01 -9.95656192e-01 9.62652490e-02 -1.10716701e-01
1.28852427e-01 -6.07087672e-01 -2.60644555e-01 8.09558868e-01
9.82672334e-01 -2.20713228e-01 1.96592391e-01 -9.26037431e-01
-3.97634357e-01 1.45797521e-01 -6.59596145e-01 2.06912711e-01
4.05092627e-01 2.49297947e-01 -2.54477151e-02 -3.98038685e-01
2.46675968e-01 4.81850266e-01 1.13182771e+00 7.41203487e-01
-1.26152158e+00 -6.32234156e-01 -7.33622491e-01 -1.52059679e-03
-1.23889816e+00 -4.76681352e-01 1.16053379e+00 -3.07482749e-01
6.83555484e-01 6.59039974e-01 9.64239895e-01 1.23418880e+00
6.74250484e-01 -6.85860887e-02 1.39826870e+00 -2.69662857e-01
3.45345110e-01 1.28331304e-01 -1.04152001e-01 2.93231905e-01
8.17457318e-01 -5.38995415e-02 -6.05353415e-01 -3.94840449e-01
9.65414464e-01 3.02293777e-01 -6.46054268e-01 6.28865659e-02
-1.21039903e+00 1.68093115e-01 -5.99880219e-02 4.12475079e-01
-5.25765777e-01 3.19624484e-01 1.93279997e-01 8.69380012e-02
3.43580663e-01 4.74793702e-01 -1.94434792e-01 -6.67683363e-01
-9.49196994e-01 -2.69974858e-01 7.79650927e-01 5.52551329e-01
1.71619967e-01 3.78918082e-01 -7.09845573e-02 3.85218352e-01
4.47258770e-01 6.77451551e-01 9.07550335e-01 -8.30866873e-01
3.25204343e-01 3.50635141e-01 4.14486796e-01 -1.15758729e+00
-7.19664812e-01 -2.89493859e-01 -1.04263210e+00 -7.48133734e-02
3.75276119e-01 -4.78646517e-01 -6.06607199e-01 1.46707737e+00
4.37392473e-01 8.74428034e-01 -6.92342520e-02 1.07458329e+00
1.05437088e+00 2.42688462e-01 -1.51488688e-02 -6.45980775e-01
1.26881957e+00 -3.89791220e-01 -1.13929880e+00 -1.98147908e-01
1.39283404e-01 -7.07524836e-01 9.57508385e-01 4.47419584e-01
-8.69232059e-01 -4.93340075e-01 -1.44516397e+00 -1.87366586e-02
1.75956726e-01 1.34601653e-01 3.48941356e-01 1.35501397e+00
-7.46503353e-01 8.99473667e-01 -1.24214768e+00 -2.77144641e-01
1.96712956e-01 2.65676677e-01 -1.59605950e-01 3.85844409e-01
-8.88925493e-01 7.30071604e-01 -2.87081748e-01 2.66355813e-01
-1.03950605e-01 -7.00944662e-01 -8.21011662e-01 -7.20468983e-02
9.40028299e-03 -8.08084428e-01 7.59643912e-01 -3.40203375e-01
-2.33956289e+00 4.11759555e-01 -3.18484664e-01 -3.44795555e-01
5.90236604e-01 -5.48760653e-01 -7.09107876e-01 3.22970778e-01
-3.16783458e-01 -2.00791240e-01 1.10857916e+00 -4.31616902e-01
3.78854662e-01 -4.71745670e-01 -7.84565866e-01 -3.96333337e-02
-4.34355289e-01 -1.91024795e-01 -8.82702395e-02 -7.17942655e-01
5.37974238e-01 -1.09710753e+00 -6.16351981e-03 -1.04808487e-01
-3.69272918e-01 4.65256304e-01 6.94149375e-01 -9.97277319e-01
1.20078981e+00 -2.08503485e+00 -2.19193816e-01 7.33719766e-02
2.74161458e-01 5.24811924e-01 3.94214779e-01 2.48669997e-01
1.05201468e-01 1.65887967e-01 -1.75993070e-01 -4.57985133e-01
-3.98628086e-01 4.06435430e-02 -5.57749942e-02 8.86405110e-01
-1.77900866e-01 8.99357915e-01 -7.30726063e-01 -7.99707845e-02
6.77236259e-01 8.07554662e-01 -4.49679233e-02 1.94590911e-01
5.41569948e-01 1.10006964e+00 -1.33944035e-01 7.35584974e-01
6.48000538e-01 -1.17703311e-01 2.24372506e-01 -5.04179418e-01
1.54333875e-01 4.70796406e-01 -1.21003544e+00 1.79683435e+00
-3.61311048e-01 5.20107210e-01 -2.21622884e-01 -7.32102573e-01
1.21054351e+00 7.45907426e-01 5.92051864e-01 -4.41205233e-01
2.18264788e-01 2.45860696e-01 -3.26228850e-02 -9.29140985e-01
-2.66354363e-02 -5.32273471e-01 -2.25534849e-02 4.71539527e-01
-2.73459256e-01 1.13104105e-01 -3.90528262e-01 -4.45072979e-01
1.28605092e+00 3.53790224e-01 6.11514509e-01 -1.76853657e-01
3.43423188e-01 -6.05932057e-01 9.49206293e-01 5.81826389e-01
-5.16579092e-01 9.02173102e-01 3.74751315e-02 -5.01722634e-01
-6.60613418e-01 -8.73631179e-01 1.76448897e-02 -2.01938331e-01
2.11230025e-01 -4.73383099e-01 -5.83545923e-01 -1.24266110e-01
1.09200187e-01 1.45298824e-01 -3.99307817e-01 -2.49654844e-01
-6.62879646e-01 -9.99478281e-01 7.36378133e-01 6.78803205e-01
6.43864095e-01 -5.95843732e-01 -9.76191282e-01 6.05342209e-01
-4.08551574e-01 -1.27813983e+00 -4.83306170e-01 -4.78333712e-01
-1.39533985e+00 -7.91401625e-01 -7.65516460e-01 -2.69653834e-02
4.90863383e-01 8.02605525e-02 7.32360363e-01 -3.80096585e-01
-6.63361907e-01 4.45433497e-01 -4.31090221e-02 -7.93359101e-01
-6.94236308e-02 -3.33409697e-01 5.14800072e-01 3.19564104e-01
6.96173728e-01 -1.13478041e+00 -1.18553877e+00 1.35257483e-01
-2.76678562e-01 -1.98075309e-01 2.94366598e-01 3.04938585e-01
6.31013513e-01 -2.97567338e-01 7.10246027e-01 -5.63502610e-01
7.57501423e-01 -2.09254920e-01 -4.99976963e-01 -2.88419783e-01
-6.33624077e-01 -2.88887531e-01 6.07303083e-01 -6.38969243e-01
-5.60328782e-01 -1.67704728e-02 4.46035787e-02 -2.23837048e-01
-1.22413121e-01 8.29538181e-02 -9.20683742e-02 -2.12468594e-01
7.06070304e-01 2.39950553e-01 2.09702000e-01 -6.34200990e-01
-1.95507243e-01 7.04702258e-01 8.94992173e-01 -1.29353896e-01
6.88184857e-01 4.65311974e-01 4.57907796e-01 -1.17750645e+00
-1.87653631e-01 -4.47187483e-01 -2.08718464e-01 -2.32007101e-01
8.10989380e-01 -9.61835206e-01 -9.21643615e-01 6.73143387e-01
-9.41937387e-01 9.66616943e-02 -2.18888938e-01 9.76577103e-01
-2.47187227e-01 7.55275607e-01 -6.71346724e-01 -9.94522095e-01
-9.75526273e-01 -6.09622002e-01 7.26841152e-01 6.23025835e-01
-6.55081749e-01 -6.09694242e-01 1.85749024e-01 4.52148557e-01
6.93976581e-01 1.06825423e+00 -4.30943854e-02 2.30850309e-01
-3.05903912e-01 -4.65455741e-01 3.45185280e-01 4.78126854e-01
6.38727188e-01 -4.12974715e-01 -1.11080468e+00 -3.88247281e-01
7.90973186e-01 2.67881453e-01 2.21113890e-01 6.30789459e-01
1.05635619e+00 -4.44473565e-01 -1.80129975e-01 8.87354374e-01
1.54514539e+00 3.82197231e-01 1.23628199e+00 -6.42229170e-02
7.28071332e-01 -5.27804065e-03 5.13782427e-02 4.76815104e-01
-7.52820633e-05 4.77209389e-01 -4.21163253e-02 -1.90571994e-01
-7.94445258e-03 -6.90624267e-02 2.72350252e-01 9.38316822e-01
-5.83760381e-01 6.99435025e-02 -2.99216866e-01 3.24752867e-01
-1.57071793e+00 -7.06853390e-01 -3.83874804e-01 2.50538826e+00
7.84144163e-01 -1.15338571e-01 1.51986852e-01 5.07041752e-01
4.93577331e-01 -1.06555380e-01 -7.56647468e-01 -2.57202834e-01
2.04752818e-01 5.42287409e-01 7.10595489e-01 8.97342786e-02
-8.64215195e-01 -1.28423959e-01 6.15337610e+00 -3.77294451e-01
-1.45907259e+00 3.77201408e-01 1.38129070e-01 -4.49613214e-01
1.48201987e-01 -3.93569529e-01 -5.04557908e-01 1.07597125e+00
1.15276325e+00 1.58657134e-02 8.76899436e-02 5.69916368e-01
6.05712414e-01 -2.69555241e-01 -8.55715811e-01 1.58863378e+00
2.90344715e-01 -9.66223717e-01 -7.13743687e-01 1.07244842e-01
3.26960951e-01 -3.19156259e-01 -3.23506743e-01 -3.66142929e-01
-8.55196297e-01 -7.45730996e-01 -4.54003587e-02 7.91123927e-01
9.03994858e-01 -4.20115501e-01 7.68973351e-01 1.42760202e-01
-8.86058807e-01 4.08679694e-01 -3.31533879e-01 -3.25534612e-01
2.27549061e-01 1.26177239e+00 -4.77365911e-01 3.26717317e-01
7.25148141e-01 5.40380895e-01 -1.37756988e-01 1.17313182e+00
-5.03298461e-01 9.80842412e-01 -6.67423904e-01 -2.65429374e-02
-5.74121952e-01 -4.34056520e-01 7.78880060e-01 8.29877317e-01
8.08974624e-01 2.98633814e-01 -3.68039221e-01 9.34776127e-01
2.61513479e-02 -1.30916223e-01 -7.04861522e-01 1.81612223e-01
4.04100090e-01 1.16613722e+00 -4.27514017e-01 -2.41520375e-01
-6.34187460e-01 1.07920754e+00 -7.20455050e-01 1.79104134e-01
-8.46098125e-01 -9.00836229e-01 6.77001357e-01 5.90753555e-01
-1.29463315e-01 -3.64515573e-01 -6.93113446e-01 -1.35706651e+00
6.78826630e-01 -6.30420029e-01 3.38460281e-02 -6.97633862e-01
-8.72698069e-01 2.66427547e-01 -4.81636405e-01 -1.76522493e+00
-3.65594387e-01 -1.10090353e-01 -6.89944565e-01 1.28467941e+00
-1.25623274e+00 -5.21353066e-01 -8.84024382e-01 5.87760508e-01
-1.35146201e-01 2.34785765e-01 1.12134087e+00 4.79529262e-01
-5.94360828e-01 4.54419494e-01 -1.61429554e-01 -4.79200669e-02
9.22908902e-01 -1.13920557e+00 3.60681206e-01 1.22859561e+00
-2.25151628e-01 9.22761440e-01 7.12718844e-01 -7.32680082e-01
-1.67017746e+00 -8.64948273e-01 1.05966175e+00 -2.88749546e-01
1.47500828e-01 -2.48388320e-01 -7.84223974e-01 4.01540309e-01
9.95261669e-02 4.11206812e-01 1.12207615e+00 -4.82439667e-01
2.36751169e-01 -6.76526785e-01 -1.53279734e+00 4.63669330e-01
6.38019621e-01 -3.90076876e-01 -7.61848152e-01 -3.80884632e-02
2.11443231e-01 -6.69765115e-01 -1.20135295e+00 3.42926234e-01
9.68859315e-01 -7.28808045e-01 7.86145985e-01 8.47938508e-02
4.63374779e-02 -6.23415649e-01 4.87248987e-01 -1.15524077e+00
-1.44875452e-01 -1.46047413e+00 -5.73357105e-01 1.17304754e+00
-1.06885813e-01 -1.22687626e+00 7.91986942e-01 1.06873429e+00
2.01030403e-01 -3.61264735e-01 -9.15816426e-01 -8.55787754e-01
-7.11880803e-01 -9.41207036e-02 4.48737502e-01 8.16200316e-01
3.82733345e-01 2.61816144e-01 -9.58809137e-01 2.12069333e-01
8.95869851e-01 -4.60170470e-02 7.63719678e-01 -1.22972345e+00
-5.78269124e-01 3.80556136e-01 -7.27965593e-01 -7.09822655e-01
-7.81866848e-01 -1.26812279e-01 -2.48824075e-01 -1.43879223e+00
-2.72349596e-01 1.69747323e-01 -4.50700223e-01 2.63804287e-01
-3.07823330e-01 9.13038254e-01 1.34586515e-02 8.46615806e-02
1.28402859e-01 1.53190374e-01 9.77223158e-01 2.10604295e-01
-8.11578512e-01 8.37054104e-02 -7.91784942e-01 6.69899940e-01
9.54819381e-01 -5.86010277e-01 -4.98683780e-01 1.45697463e-02
8.13752562e-02 3.17152321e-01 3.54383647e-01 -1.51829410e+00
1.04995124e-01 2.51386762e-01 7.76203871e-01 -2.09775507e-01
3.70057166e-01 -6.92955196e-01 8.17886412e-01 7.56699264e-01
4.80572730e-01 2.93821879e-02 3.32871199e-01 3.91034633e-01
-3.42759676e-02 1.27951026e-01 6.90637887e-01 -2.20923260e-01
1.32152542e-01 -1.32210910e-01 -5.11817813e-01 -1.35409191e-01
7.36434281e-01 -5.53584576e-01 -4.66902614e-01 -4.76754904e-01
-7.78206408e-01 -4.80454803e-01 3.10348183e-01 2.40532130e-01
6.62983358e-01 -1.32503712e+00 -4.24009502e-01 4.39811677e-01
-1.79358616e-01 -2.31178075e-01 3.15640599e-01 1.28369212e+00
-5.84732234e-01 2.23412618e-01 -3.58427584e-01 -5.82172036e-01
-1.22351837e+00 1.82363659e-01 4.94477034e-01 5.41072451e-02
-1.16678190e+00 2.84202933e-01 -6.14201844e-01 4.57475036e-01
-8.94884951e-03 -5.93608439e-01 4.23314534e-02 -3.81116182e-01
7.86614656e-01 7.53977716e-01 3.53563249e-01 1.95351755e-03
-5.25552154e-01 9.61195230e-01 6.08699441e-01 4.62365560e-02
1.00433695e+00 -3.87907028e-01 6.23695403e-02 3.66466135e-01
8.90030980e-01 2.02948600e-01 -8.65009785e-01 1.19375058e-01
-3.58395994e-01 -5.27817070e-01 -9.24537033e-02 -7.18242645e-01
-1.03538072e+00 7.46286750e-01 1.04393518e+00 8.99191499e-02
1.42754328e+00 -7.44623482e-01 1.19193983e+00 -7.54138082e-02
5.05576193e-01 -9.22301531e-01 -6.76043704e-02 -4.66415465e-01
6.45265102e-01 -8.51054966e-01 2.80858546e-01 -4.26781476e-01
-2.37078100e-01 9.66072619e-01 1.88276514e-01 -2.21058011e-01
7.30692267e-01 3.56875151e-01 4.13513720e-01 3.16127479e-01
1.40968040e-02 2.90062785e-01 2.68986374e-01 8.43572855e-01
5.42060792e-01 1.12071402e-01 -1.04552841e+00 6.84418082e-01
-9.33868140e-02 6.15426362e-01 8.63754392e-01 1.05733407e+00
1.16368525e-01 -9.45983410e-01 -4.47634280e-01 5.46579480e-01
-1.07739890e+00 1.69990212e-01 -1.09572805e-01 3.86259913e-01
5.13412543e-02 1.25911629e+00 -2.16373175e-01 -2.92940378e-01
4.32162791e-01 2.66443163e-01 5.86076200e-01 -3.71140569e-01
-5.56464016e-01 3.34896594e-01 -4.12926115e-02 -8.69080067e-01
-7.43022084e-01 -5.63854456e-01 -8.28180790e-01 -4.64597009e-02
5.63114230e-03 -2.88694978e-01 8.32804620e-01 6.48440838e-01
8.03719938e-01 5.50616145e-01 3.32337290e-01 -5.81146359e-01
-3.80385965e-01 -1.20053339e+00 -8.70621204e-01 4.67618138e-01
5.90519965e-01 -3.97859037e-01 -5.08799970e-01 2.08692849e-01]
|
[13.942517280578613, 2.9940614700317383]
|
1a4620a3-956d-4ab6-8342-6aa35b05ff14
|
direction-of-arrival-estimation-of-noisy
|
2102.09853
| null |
https://arxiv.org/abs/2102.09853v3
|
https://arxiv.org/pdf/2102.09853v3.pdf
|
Direction of Arrival Estimation of Noisy Speech Using Convolutional Recurrent Neural Networks with Higher-Order Ambisonics Signals
|
Training convolutional recurrent neural networks on first-order Ambisonics signals is a well-known approach when estimating the direction of arrival for speech/sound signals. In this work, we investigate whether increasing the order of Ambisonics up to the fourth order further improves the estimation performance of convolutional recurrent neural networks. While our results on data based on simulated spatial room impulse responses show that the use of higher Ambisonics orders does have the potential to provide better localization results, no further improvement was shown on data based on real spatial room impulse responses from order two onwards. Rather, it seems to be crucial to extract meaningful features from the raw data. First order features derived from the acoustic intensity vector were superior to pure higher-order magnitude and phase features in almost all scenarios.
|
['Jürgen Peissig', 'Stephan Preihs', 'Robert Hupke', 'Nils Poschadel']
|
2021-02-19
| null | null | null | null |
['direction-of-arrival-estimation']
|
['audio']
|
[-1.09238297e-01 -4.05234784e-01 7.31500804e-01 -2.14329302e-01
-1.00308764e+00 -3.46111387e-01 6.01366460e-01 7.96011388e-02
-5.12958646e-01 5.33058941e-01 5.37971199e-01 -5.20411789e-01
-3.52813572e-01 -5.30515075e-01 -4.21831429e-01 -9.40088928e-01
-5.34808874e-01 -4.38851677e-02 -1.06363349e-01 -3.29674512e-01
1.13645516e-01 6.39953196e-01 -1.57195985e+00 2.63969809e-01
2.40836456e-01 1.05405056e+00 1.33809209e-01 1.03589165e+00
-1.39822625e-02 6.19026303e-01 -1.00079322e+00 3.39427859e-01
3.17274421e-01 -4.17843491e-01 -4.07764465e-01 -6.04859769e-01
2.98473448e-01 -2.71833628e-01 -3.13828796e-01 5.18002450e-01
9.62967098e-01 3.89293432e-01 4.91156071e-01 -6.57920301e-01
-4.46041413e-05 6.53052032e-01 8.18581060e-02 4.34790671e-01
2.93198794e-01 8.96594822e-02 9.29850280e-01 -8.72443557e-01
3.67558524e-02 9.28332806e-01 1.02982426e+00 -9.77088884e-02
-8.92409801e-01 -5.98320305e-01 -3.17568839e-01 -8.06263182e-03
-1.16254890e+00 -6.49683535e-01 1.05877924e+00 -2.09320024e-01
1.13274193e+00 4.94698256e-01 3.27467442e-01 1.08981299e+00
-2.00612191e-02 4.71126467e-01 1.22533727e+00 -6.64740622e-01
2.01961800e-01 3.75612266e-02 1.32596314e-01 4.35452051e-02
-1.39015079e-01 3.37488502e-01 -3.14724505e-01 -5.47037050e-02
5.48529923e-01 -3.79956305e-01 -4.17447895e-01 3.44723612e-02
-1.32191217e+00 6.93534553e-01 8.06218445e-01 1.06425846e+00
-7.20655024e-01 5.42723238e-01 5.14538765e-01 3.31856310e-01
4.60947573e-01 9.80008841e-01 -4.19318527e-01 -5.07560849e-01
-9.86940444e-01 -1.08094461e-01 5.37550390e-01 2.31736690e-01
4.20148224e-01 6.50042534e-01 1.29182309e-01 1.00307906e+00
1.83251157e-01 7.46593535e-01 1.95009690e-02 -7.53963888e-01
3.99060905e-01 -2.59521902e-01 1.48865148e-01 -1.13641834e+00
-9.85647142e-01 -1.04061866e+00 -7.83334732e-01 3.81547846e-02
6.64481044e-01 -4.94174272e-01 -8.09666991e-01 1.51295722e+00
-1.72745556e-01 2.15618238e-01 2.28627667e-01 1.24394262e+00
7.75943100e-01 9.01381671e-01 -1.79647237e-01 2.73648594e-02
1.15281212e+00 -4.44767386e-01 -7.30863333e-01 -1.56184226e-01
5.06493449e-01 -1.02595270e+00 7.79181063e-01 4.47319031e-01
-7.25861609e-01 -6.88033223e-01 -1.02543616e+00 5.48789322e-01
-3.82886261e-01 3.52780342e-01 6.57207191e-01 9.05136704e-01
-1.02292538e+00 5.00658751e-01 -5.20052195e-01 7.46619627e-02
-3.35118681e-01 3.33906822e-02 -2.22408876e-01 1.20307803e-01
-1.54736876e+00 7.29803622e-01 -2.66754121e-01 5.26515603e-01
-5.18767655e-01 -8.97232473e-01 -7.71790028e-01 1.87777564e-01
-1.69666931e-01 -3.34594771e-02 1.04927254e+00 -5.88414967e-01
-1.52824235e+00 1.49368256e-01 -5.90224266e-02 -5.80710411e-01
3.20262372e-01 -2.54854530e-01 -6.78360879e-01 1.07176118e-02
-1.61190748e-01 3.81993026e-01 7.52334118e-01 -1.33105588e+00
-3.22780162e-01 6.82139248e-02 5.55118471e-02 -1.38285607e-01
-2.67503411e-01 -1.63155645e-01 3.45746100e-01 -4.95167762e-01
3.47124130e-01 -1.05208552e+00 -2.60765702e-01 -8.24176133e-01
-4.17962283e-01 -6.17873482e-02 3.27234417e-01 -6.76756322e-01
9.88107145e-01 -2.34223723e+00 -2.17386633e-01 3.53857040e-01
-1.83247313e-01 2.38922507e-01 -1.29904166e-01 5.32202065e-01
-4.63901192e-01 -1.12251088e-01 7.58043751e-02 -1.96503475e-01
2.42334371e-03 -3.51630688e-01 -5.22683382e-01 4.72813278e-01
1.62393525e-01 3.23562980e-01 -7.00641453e-01 1.14996076e-01
4.17948395e-01 9.35032785e-01 -5.16661704e-01 8.69420881e-04
3.44459206e-01 5.54769039e-01 -2.03241825e-01 4.68245111e-02
7.90133774e-01 1.12675376e-01 -3.41718137e-01 -3.08038801e-01
-4.35954660e-01 8.27846646e-01 -1.00312006e+00 1.10521054e+00
-1.36314857e+00 1.29977214e+00 1.41874999e-02 -8.51575196e-01
9.74282861e-01 6.05123937e-01 4.64384407e-01 -9.88528728e-01
3.56677361e-02 4.76159602e-01 3.94814253e-01 -3.64023685e-01
3.91795963e-01 -1.89997226e-01 -1.64871395e-01 2.43373826e-01
-1.71343490e-01 -2.42641717e-01 -2.88866192e-01 -3.74731809e-01
1.09830141e+00 -3.20198953e-01 -2.31052935e-01 -2.20215261e-01
5.96726716e-01 -3.60211045e-01 -1.46644181e-02 7.30497003e-01
1.97605476e-01 7.72767186e-01 1.88850924e-01 -3.78537506e-01
-8.85751486e-01 -8.70004535e-01 -2.52436042e-01 1.17493272e+00
-4.40417349e-01 -3.27860624e-01 -6.37220383e-01 -1.27458572e-01
-3.74092937e-01 1.01310050e+00 -5.37496150e-01 6.38530869e-03
-7.62622833e-01 -4.77852255e-01 8.17061961e-01 4.79320914e-01
6.88573837e-01 -9.78879809e-01 -9.15427685e-01 3.98148268e-01
-2.94745803e-01 -1.17336762e+00 -5.19621447e-02 7.72369504e-01
-4.85583097e-01 -3.19141746e-01 -1.01342225e+00 -3.18347424e-01
1.63355693e-01 2.34597653e-01 7.97845721e-01 -8.67831036e-02
-1.69007465e-01 5.67980468e-01 -3.88274938e-01 -5.23441017e-01
-2.97893792e-01 1.18220724e-01 1.72546148e-01 -1.45062149e-01
-5.58937863e-02 -6.51973605e-01 -6.73718512e-01 2.24737853e-01
-5.79059899e-01 -4.68276590e-01 5.97453654e-01 6.46685481e-01
-1.92676231e-01 8.35337788e-02 5.41666269e-01 -1.09468311e-01
6.61678791e-01 -1.64762568e-02 -3.84753883e-01 -2.68230110e-01
9.82001802e-05 1.97609589e-01 8.74653518e-01 -1.93115622e-01
-9.21829820e-01 -1.81075528e-01 -7.65315354e-01 -3.49399090e-01
-4.80537802e-01 5.67266583e-01 3.26880395e-01 3.06032337e-02
7.60445893e-01 1.93136752e-01 -5.07012069e-01 -4.27013189e-01
1.41578084e-02 6.78898215e-01 1.50451392e-01 -2.13226289e-01
6.41893625e-01 3.27454537e-01 2.36610100e-01 -1.49002719e+00
-3.82350177e-01 -6.86492443e-01 -2.84660518e-01 -1.49478272e-01
9.61051941e-01 -7.34773517e-01 -8.33725929e-01 2.52234280e-01
-1.23160577e+00 -3.74001831e-01 1.49917905e-04 1.07281697e+00
-3.62401068e-01 -1.94789730e-02 -4.79813844e-01 -1.22560394e+00
-7.40294978e-02 -1.26261926e+00 9.03072834e-01 -2.59102255e-01
-3.70461524e-01 -9.75342214e-01 1.78177893e-01 1.04519896e-01
1.06503880e+00 -1.74262356e-02 6.83174789e-01 -5.71235955e-01
-4.08629358e-01 -3.69261414e-01 1.77058443e-01 3.17605406e-01
2.38300487e-01 -2.84720212e-01 -1.44043374e+00 -7.45318830e-02
2.90583104e-01 5.43624023e-03 9.39137161e-01 6.79859519e-01
6.98139846e-01 -1.87189188e-02 1.90733727e-02 4.33710039e-01
1.16110802e+00 2.71895885e-01 8.09603095e-01 3.71485382e-01
5.34545660e-01 5.72864890e-01 4.75168020e-01 4.25642341e-01
-2.25691766e-01 7.11068988e-01 4.37741846e-01 -3.19137961e-01
-3.17939848e-01 1.20838925e-01 2.16233402e-01 9.52085853e-01
-2.65919060e-01 -2.63599724e-01 -1.08976388e+00 5.87795258e-01
-9.41264629e-01 -9.84470487e-01 -3.16161066e-01 2.26358843e+00
1.79782584e-01 1.31597489e-01 -6.80327192e-02 5.15248001e-01
4.27438498e-01 3.41033936e-01 5.02897687e-02 -7.13640153e-01
-2.68340595e-02 2.09833071e-01 6.33138120e-01 5.38713396e-01
-8.96907687e-01 4.06528950e-01 6.48080969e+00 4.83184665e-01
-1.75815606e+00 -7.89168552e-02 6.44288480e-01 4.90954183e-02
-4.98001695e-01 -2.92069286e-01 -5.14450908e-01 1.89704880e-01
1.34982789e+00 6.84415579e-01 3.76690149e-01 6.23056829e-01
2.90428489e-01 -1.29897460e-01 -8.44092906e-01 9.39097881e-01
-1.80815414e-01 -9.96633470e-01 -5.63100696e-01 1.18861638e-01
3.21939468e-01 2.46599704e-01 4.97985303e-01 3.42200160e-01
-1.52037814e-01 -1.24100089e+00 5.88490844e-01 4.25367206e-01
4.67706233e-01 -8.13265979e-01 7.95413911e-01 3.59482825e-01
-1.08020973e+00 -9.75598544e-02 -2.89259464e-01 -1.24918379e-01
2.25517109e-01 6.90128863e-01 -1.34002316e+00 3.75456542e-01
8.25619638e-01 1.74732104e-01 -3.83373946e-01 1.19922876e+00
1.14737945e-02 1.06847048e+00 -8.20758998e-01 -3.57612133e-01
7.52058506e-01 7.63583556e-02 6.12203062e-01 1.48735809e+00
5.25181115e-01 -2.59518195e-02 -4.80735481e-01 3.64186198e-01
3.00517648e-01 -1.18473738e-01 -7.46849716e-01 6.09478578e-02
1.15778387e-01 1.17502534e+00 -8.36940110e-01 1.42222866e-01
-7.77943879e-02 4.60218042e-01 -3.04698616e-01 6.27213538e-01
-9.59382296e-01 -7.00119555e-01 5.30875683e-01 2.25009605e-01
6.44037306e-01 -6.94824159e-01 -1.77968949e-01 -4.36535329e-01
-1.55788869e-01 -6.54014587e-01 -1.48235977e-01 -8.62347424e-01
-8.04131866e-01 9.14334893e-01 -7.29635581e-02 -1.36702800e+00
-7.10631549e-01 -5.36968887e-01 -7.81485558e-01 9.79695141e-01
-1.34268379e+00 -7.24236429e-01 -5.01454994e-02 2.39187285e-01
3.97846013e-01 1.51429132e-01 9.69343543e-01 3.76640350e-01
9.97914001e-02 4.80510026e-01 2.14063451e-01 8.20372328e-02
5.51385581e-01 -1.00416660e+00 3.71371150e-01 6.62522852e-01
3.76915991e-01 8.70188296e-01 1.41565907e+00 3.10361851e-03
-1.05716705e+00 -5.90021014e-01 6.24962032e-01 -2.21896783e-01
3.93211782e-01 -5.22685111e-01 -7.37362683e-01 2.87829995e-01
5.01267672e-01 4.24667820e-03 5.71662605e-01 3.41843635e-01
-1.35347933e-01 -3.07612538e-01 -6.89058244e-01 4.82292473e-01
6.69264555e-01 -6.05297744e-01 -4.09312308e-01 1.80687174e-01
5.19927204e-01 -2.02130377e-01 -8.07202935e-01 4.74655479e-01
2.56774455e-01 -1.07317877e+00 1.34147227e+00 5.03769740e-02
1.56767726e-01 -2.24053755e-01 -5.04109979e-01 -1.62763572e+00
-1.60321489e-01 -6.22414529e-01 6.06552780e-01 1.06571627e+00
6.76524162e-01 -9.25466418e-01 6.35163248e-01 -1.56468358e-02
-2.46500120e-01 -4.20668095e-01 -1.17074800e+00 -7.41897881e-01
1.25647575e-01 -1.08285367e+00 5.73501468e-01 6.90478146e-01
-2.17782184e-01 4.34808820e-01 -4.51345682e-01 5.59513509e-01
3.24500799e-01 1.35269597e-01 6.76188648e-01 -1.02101099e+00
-2.61851102e-01 -4.24419791e-01 -3.24248105e-01 -1.10205925e+00
2.13953167e-01 -5.11673748e-01 4.60848868e-01 -1.71955717e+00
-9.58857417e-01 -9.28290367e-01 -5.23847580e-01 2.10679378e-02
1.75519943e-01 3.39854598e-01 1.67537168e-01 -3.33979040e-01
-1.89139068e-01 4.47274536e-01 8.25821519e-01 -4.98483796e-03
-1.25393867e-01 6.10296547e-01 -1.18409939e-01 6.06536984e-01
8.10373187e-01 -3.74013960e-01 -4.14419800e-01 -4.47990090e-01
3.88452202e-01 2.94002861e-01 4.13569480e-01 -1.63529897e+00
3.41522276e-01 4.16917175e-01 5.65393507e-01 -6.69475734e-01
8.68036032e-01 -8.29919577e-01 5.82557283e-02 3.47356558e-01
-4.71676767e-01 -1.56424120e-02 6.46003008e-01 2.82659769e-01
-3.27270567e-01 -1.17603935e-01 5.23643970e-01 1.56222703e-02
-5.02609670e-01 -4.29540873e-01 -7.29932845e-01 -2.91203856e-01
3.15512598e-01 -1.91942416e-02 1.96894839e-01 -9.80690479e-01
-4.32115227e-01 -5.53176224e-01 -1.64044037e-01 3.91852707e-01
5.62574267e-01 -1.18744051e+00 -5.55030942e-01 1.23280816e-01
-1.56179160e-01 -5.21677017e-01 2.95988232e-01 9.30459738e-01
-4.13669109e-01 1.11478949e+00 -1.41728371e-02 -9.36999321e-01
-1.17548621e+00 2.98483670e-01 5.41022360e-01 -4.30752598e-02
-3.54209155e-01 9.45773661e-01 2.37082556e-01 -5.71531653e-01
1.78141057e-01 -7.36124694e-01 -4.24613714e-01 1.27405033e-01
3.22179407e-01 3.46823990e-01 5.17158508e-01 -8.83393288e-01
-6.32733345e-01 7.66808331e-01 4.44140226e-01 -5.30508697e-01
1.51558554e+00 -9.56127569e-02 1.99283332e-01 8.13206911e-01
1.69806087e+00 5.40687442e-01 -1.04623604e+00 1.64594755e-01
-1.05855562e-01 -3.12475532e-01 4.65426236e-01 -8.07845473e-01
-8.43013644e-01 1.34261131e+00 8.16421032e-01 5.58148563e-01
1.09950829e+00 -1.02116950e-01 6.77830815e-01 6.18204951e-01
4.37476039e-01 -7.24322379e-01 9.17242169e-02 8.30028355e-01
8.90088201e-01 -1.04395425e+00 -2.84784406e-01 8.02060962e-02
-4.25266206e-01 1.23256099e+00 -1.41736241e-02 -3.20587754e-02
7.27660358e-01 4.33515698e-01 4.55805361e-01 -9.94403884e-02
-3.59704971e-01 -1.35388032e-01 2.17055202e-01 5.18206656e-01
7.47771204e-01 6.11196598e-03 3.25463265e-01 7.99374804e-02
-6.18206501e-01 -5.90307176e-01 3.69955897e-01 4.41034406e-01
-4.14707929e-01 -7.48731136e-01 -9.76995528e-01 2.01430142e-01
-5.89507699e-01 -3.45825404e-01 -2.78445948e-02 7.78030574e-01
-2.22768635e-01 1.11728728e+00 3.45358372e-01 -3.60053480e-01
5.76499045e-01 -1.72975864e-02 2.69688964e-01 -9.66828167e-02
-7.46964276e-01 1.98037103e-01 3.44034821e-01 -1.84055090e-01
-2.10609391e-01 -6.14623547e-01 -1.27570641e+00 -1.46330446e-01
-4.17825013e-01 5.80090940e-01 1.33111978e+00 7.02676892e-01
1.62998095e-01 1.16631746e+00 8.85278463e-01 -1.16816139e+00
-3.28369975e-01 -1.03867555e+00 -2.56881624e-01 -5.45448670e-03
1.04729152e+00 -4.29129958e-01 -7.72664666e-01 -5.47833979e-01]
|
[15.10775089263916, 5.769680976867676]
|
6ce40275-a633-461d-a76a-64b4ec1ccee7
|
classification-with-costly-features-using
|
1711.07364
| null |
http://arxiv.org/abs/1711.07364v2
|
http://arxiv.org/pdf/1711.07364v2.pdf
|
Classification with Costly Features using Deep Reinforcement Learning
|
We study a classification problem where each feature can be acquired for a
cost and the goal is to optimize a trade-off between the expected
classification error and the feature cost. We revisit a former approach that
has framed the problem as a sequential decision-making problem and solved it by
Q-learning with a linear approximation, where individual actions are either
requests for feature values or terminate the episode by providing a
classification decision. On a set of eight problems, we demonstrate that by
replacing the linear approximation with neural networks the approach becomes
comparable to the state-of-the-art algorithms developed specifically for this
problem. The approach is flexible, as it can be improved with any new
reinforcement learning enhancement, it allows inclusion of pre-trained
high-performance classifier, and unlike prior art, its performance is robust
across all evaluated datasets.
|
['Viliam Lisý', 'Tomáš Pevný', 'Jaromír Janisch']
|
2017-11-20
| null | null | null | null |
['classification-with-costly-features']
|
['miscellaneous']
|
[ 4.69239026e-01 4.15825397e-01 -3.88918400e-01 -6.07986271e-01
-8.68574500e-01 -4.53897119e-01 6.54573560e-01 2.72193581e-01
-7.67412901e-01 1.11769092e+00 -2.87408173e-01 -2.02953845e-01
-6.73634410e-01 -7.65537441e-01 -5.87111354e-01 -7.57825315e-01
-3.20752829e-01 7.56422520e-01 1.11671582e-01 -9.03788060e-02
6.57171309e-01 4.38555777e-01 -1.84878051e+00 1.06070094e-01
7.71375179e-01 1.53703594e+00 -1.74910098e-01 6.26117289e-01
1.53203517e-01 8.89413953e-01 -7.00724483e-01 -4.36284184e-01
5.04747927e-01 -3.39290172e-01 -1.17315102e+00 2.79500365e-01
8.00247192e-02 -2.70699054e-01 2.10662201e-01 7.51734912e-01
5.01315176e-01 3.32456380e-01 7.77684927e-01 -1.61479640e+00
-4.96576041e-01 2.48403862e-01 -1.06427871e-01 1.60922736e-01
5.82328141e-01 3.41442674e-01 1.31286788e+00 -5.15904367e-01
4.06357586e-01 1.10311437e+00 7.07596183e-01 6.64814472e-01
-1.49319792e+00 -2.37125173e-01 2.07572818e-01 3.96384239e-01
-8.30223203e-01 -2.45714173e-01 3.77794951e-01 -3.20970714e-01
1.67689013e+00 1.61953971e-01 9.63251352e-01 8.64542544e-01
2.74702281e-01 1.03579581e+00 1.19186521e+00 -5.80579817e-01
7.52545595e-01 1.83685049e-01 -9.53314975e-02 4.98202026e-01
-9.01275203e-02 6.85764551e-01 -1.60613656e-01 -3.65029901e-01
3.22363615e-01 3.73288877e-02 1.73574224e-01 -6.96236074e-01
-8.32472205e-01 1.20783353e+00 3.29705536e-01 1.13559254e-01
-8.10619533e-01 7.79635534e-02 3.70956689e-01 7.60056615e-01
3.73894572e-01 9.28120494e-01 -8.71844292e-01 -2.38974422e-01
-9.06062782e-01 5.84415615e-01 1.03397429e+00 4.45199281e-01
7.80972779e-01 2.50515249e-02 -2.80815154e-01 6.50064945e-01
1.09466620e-01 -3.05457320e-02 7.47870803e-01 -1.25006902e+00
1.07159607e-01 5.37040055e-01 4.13719118e-01 -2.19171688e-01
-4.70548630e-01 -3.14427048e-01 -2.42864504e-01 6.74663424e-01
3.11734736e-01 -5.18983603e-01 -7.21826792e-01 1.44815004e+00
2.48677149e-01 -6.85305800e-03 3.24480027e-01 5.55765927e-01
1.75540015e-01 5.45681477e-01 2.09376942e-02 -6.08549953e-01
9.51250434e-01 -1.06136560e+00 -6.42946959e-01 -1.84931666e-01
3.49109888e-01 -4.31040555e-01 6.66464090e-01 9.32470739e-01
-1.24136174e+00 -5.99967897e-01 -1.13764000e+00 4.42399621e-01
-4.81875926e-01 -2.31042057e-01 6.95795000e-01 7.41954386e-01
-1.20117545e+00 1.23811996e+00 -6.45528495e-01 -1.73473522e-01
3.30847949e-01 7.89390326e-01 -6.55634329e-02 4.86285612e-02
-1.17894435e+00 1.18557060e+00 6.10734642e-01 -5.59784211e-02
-9.40748870e-01 -4.10125107e-01 -6.68528914e-01 2.13352069e-02
6.10511899e-01 -7.00300097e-01 1.85107911e+00 -1.61433315e+00
-1.96816909e+00 5.51680207e-01 2.86797523e-01 -6.58721566e-01
6.98428929e-01 -9.07889232e-02 -2.83767968e-01 -9.22706947e-02
4.31110859e-02 7.77393997e-01 1.04818058e+00 -9.36543703e-01
-1.04429650e+00 -1.65411994e-01 1.04225092e-01 3.69131058e-01
-1.75144583e-01 -4.10674550e-02 2.76509732e-01 -3.72847557e-01
-5.05073249e-01 -8.65201294e-01 -5.62796235e-01 -1.69902235e-01
7.42668062e-02 -5.15242457e-01 5.38923502e-01 -2.19830409e-01
1.00221658e+00 -1.87003005e+00 1.03290088e-01 1.85431138e-01
-4.13216949e-01 1.59427807e-01 -2.66308814e-01 5.78193605e-01
-2.21936852e-01 1.58073738e-01 -3.42952162e-01 -2.05988899e-01
1.19988278e-01 3.51229399e-01 -1.92343630e-02 4.07706887e-01
6.58187389e-01 7.42321610e-01 -1.02898598e+00 -7.93501213e-02
-1.70380976e-02 3.53968740e-02 -6.15512967e-01 5.42733133e-01
-4.26412702e-01 6.61736280e-02 -5.59085667e-01 4.96721566e-01
3.47318262e-01 -3.14336121e-01 1.33301079e-01 4.46948469e-01
5.11783846e-02 1.05483212e-01 -1.52728462e+00 1.45052099e+00
-4.66848224e-01 2.62920946e-01 -1.78745314e-01 -1.48002052e+00
7.76485980e-01 2.74317026e-01 6.09961808e-01 -7.34870851e-01
-1.25111323e-02 2.28466064e-01 7.92008936e-02 -6.45006180e-01
2.15106457e-01 -4.02917683e-01 8.06068033e-02 5.47211647e-01
3.54045480e-01 -3.40723693e-01 1.94071218e-01 -4.69869733e-01
1.22403479e+00 3.09219122e-01 5.45655966e-01 -1.42388880e-01
4.55363482e-01 1.85696959e-01 5.44064939e-01 9.20791924e-01
-1.43941596e-01 1.93666905e-01 5.30064702e-01 -5.68848312e-01
-9.14112926e-01 -7.81507194e-01 -2.15852246e-01 1.36891162e+00
-2.20895484e-01 1.07795455e-01 -6.58672154e-01 -1.01044929e+00
3.53117079e-01 9.01623189e-01 -8.63852024e-01 -2.65235037e-01
-2.77303457e-01 -9.96795833e-01 -4.07649800e-02 6.08258843e-01
3.81984800e-01 -1.47132528e+00 -1.00192750e+00 4.86433387e-01
3.79010737e-01 -5.59874654e-01 5.02813868e-02 7.68367767e-01
-1.09984529e+00 -1.03399169e+00 -3.88961792e-01 -6.10666990e-01
4.31498766e-01 -4.50505972e-01 1.43402624e+00 1.48414105e-01
-3.81024688e-01 4.72291410e-01 -4.37978238e-01 -6.38175666e-01
-2.11461022e-01 1.29831821e-01 -4.93398085e-02 2.64525376e-02
3.13636333e-01 -2.85131991e-01 -6.79303944e-01 1.47095367e-01
-7.87789881e-01 -3.60107034e-01 7.14145243e-01 1.30089188e+00
4.22232628e-01 1.21422477e-01 8.65951538e-01 -1.01571679e+00
9.35327947e-01 -5.37420571e-01 -6.67962790e-01 1.51448324e-01
-1.16141427e+00 2.46836558e-01 8.42836738e-01 -3.85622323e-01
-8.71853590e-01 2.59871602e-01 -1.64406329e-01 -7.39860162e-02
-2.20566168e-01 3.16219300e-01 2.09530264e-01 -1.58529609e-01
6.75046086e-01 2.00531229e-01 2.42693365e-01 -2.14171037e-01
2.43710324e-01 5.87659001e-01 -6.04339354e-02 -4.34427232e-01
3.45324576e-01 -4.57141027e-02 2.30232887e-02 -3.20952326e-01
-6.59225762e-01 -2.36334220e-01 -5.27077615e-01 -3.47337425e-02
4.63172972e-01 -5.45120835e-01 -1.03633797e+00 2.12774724e-01
-6.25255108e-01 -5.39318800e-01 -8.13215256e-01 3.97425115e-01
-1.11831343e+00 -6.58088848e-02 -3.26291114e-01 -9.86307263e-01
-2.72888035e-01 -1.06747127e+00 7.79277325e-01 2.46832684e-01
-1.04461089e-01 -9.73638117e-01 1.46588176e-01 -7.96584412e-03
4.74521786e-01 2.17037231e-01 7.98154533e-01 -1.02184141e+00
-2.72530764e-01 -3.55926454e-01 1.37014613e-01 5.63852906e-01
-2.27480028e-02 -5.51444627e-02 -8.88174176e-01 -6.42807484e-01
-1.16880842e-01 -9.89544868e-01 5.63602746e-01 3.52902353e-01
1.30832946e+00 -4.41016912e-01 -9.21882689e-03 2.12248340e-01
1.58521426e+00 5.83550334e-01 3.03604722e-01 7.73973644e-01
-2.07941636e-01 6.23640656e-01 9.92907286e-01 6.33128166e-01
8.08314160e-02 4.99709547e-01 6.83158576e-01 6.38466189e-03
4.47446078e-01 1.35457680e-01 2.09505409e-01 -5.23563884e-02
-1.61100581e-01 -1.85402513e-01 -7.87633836e-01 4.69450355e-01
-2.24941516e+00 -9.40620124e-01 5.40592313e-01 2.17591619e+00
5.72303116e-01 2.50120640e-01 6.11658216e-01 4.12770569e-01
3.08564603e-01 -1.93253502e-01 -9.96065319e-01 -8.59306037e-01
2.96390831e-01 3.92863750e-01 2.59061754e-01 4.65148330e-01
-1.01685810e+00 7.12721944e-01 8.32370281e+00 7.36605346e-01
-8.92547488e-01 -1.16124012e-01 8.73032272e-01 -3.17211866e-01
-1.80905256e-02 -2.23214597e-01 -5.13745308e-01 2.58948654e-01
1.17313623e+00 -6.80732876e-02 7.80833244e-01 8.98967326e-01
-3.37963179e-03 -2.71549046e-01 -1.53614306e+00 4.76835430e-01
-1.49239197e-01 -1.02109647e+00 -2.32780486e-01 -8.14196989e-02
5.90668082e-01 -1.52666094e-02 5.86581416e-02 7.27321446e-01
7.58003294e-01 -1.14703119e+00 5.94780564e-01 4.35857177e-01
5.03000796e-01 -1.02770054e+00 8.86753380e-01 5.14728785e-01
-5.45384645e-01 -8.48270595e-01 -2.57212609e-01 -4.67809618e-01
-3.39506775e-01 2.51735240e-01 -8.30731809e-01 6.02013707e-01
5.81739068e-01 5.54521143e-01 -5.86920738e-01 1.23310208e+00
-1.98410377e-01 4.60121006e-01 -1.26392573e-01 -5.99307060e-01
5.55435479e-01 -1.65569261e-01 1.03637777e-01 1.05535066e+00
1.87300533e-01 -3.94636719e-03 5.49629867e-01 3.62063169e-01
2.40649909e-01 2.99843848e-01 -3.74663442e-01 2.76572585e-01
2.10985765e-01 1.12897384e+00 -6.01429403e-01 -3.36851090e-01
-4.17487741e-01 8.91081691e-01 6.93975925e-01 2.02685207e-01
-3.79594237e-01 -4.17129278e-01 3.47339243e-01 -2.61945605e-01
6.67241812e-01 5.27133346e-01 -7.40430132e-02 -8.62115800e-01
-6.58588624e-03 -9.64173377e-01 6.16472960e-01 -5.45847893e-01
-1.60541725e+00 5.92820048e-01 -8.31743479e-02 -1.29917514e+00
-9.79401112e-01 -7.39870965e-01 -5.25154412e-01 7.44181335e-01
-1.76195633e+00 -6.13547206e-01 6.47094771e-02 4.59306955e-01
4.90078866e-01 -3.51840109e-01 1.03889191e+00 -9.61345732e-02
-4.01636690e-01 4.43015009e-01 4.62959439e-01 -2.30853006e-01
3.86835605e-01 -1.70616829e+00 7.45735988e-02 2.47430041e-01
-1.36735201e-01 -4.63267416e-02 7.91806698e-01 -2.06767648e-01
-1.24121821e+00 -7.65947878e-01 5.34744561e-01 -2.79986352e-01
6.06936216e-01 -1.64828613e-01 -5.84345877e-01 4.92747247e-01
4.16834265e-01 3.42631161e-01 8.10091734e-01 1.79149032e-01
1.91107363e-01 -2.76392490e-01 -1.54680085e+00 3.13196555e-02
7.97536910e-01 -1.18186496e-01 -7.72823691e-01 4.32408988e-01
2.87473083e-01 -2.27964088e-01 -9.31888998e-01 3.75097513e-01
5.90537071e-01 -9.79638398e-01 7.11753011e-01 -1.06871521e+00
5.17178237e-01 8.16927552e-02 5.86341415e-03 -1.78489280e+00
-4.16173339e-01 -5.64067900e-01 -2.29392007e-01 8.21659088e-01
5.98677516e-01 -7.04416990e-01 8.78788948e-01 7.05423117e-01
3.30043972e-01 -1.29887092e+00 -1.12154424e+00 -8.43827546e-01
2.49281198e-01 -1.65742368e-01 5.75645983e-01 6.78373158e-01
5.91000654e-02 2.45757714e-01 -3.39580655e-01 -2.00677559e-01
4.44208980e-01 2.97857463e-01 3.06399375e-01 -1.27684438e+00
-6.62137210e-01 -5.43503463e-01 -3.69396836e-01 -6.81068063e-01
2.89350539e-01 -6.54892147e-01 1.46456718e-01 -1.28048396e+00
9.97100212e-03 -4.38805908e-01 -6.52690470e-01 8.10132265e-01
-2.35388502e-01 -1.10953294e-01 1.32723004e-01 -1.19311690e-01
-7.42132664e-01 5.12339532e-01 9.67796087e-01 -9.78007540e-03
-2.70399451e-01 5.36718607e-01 -6.93027675e-01 4.14296180e-01
6.75160050e-01 -6.92604125e-01 -3.88887912e-01 1.28039792e-01
3.40322196e-01 4.91046757e-01 5.24235051e-03 -8.93573701e-01
1.78465500e-01 -3.73027980e-01 7.42957473e-01 -1.82612389e-01
1.15226105e-01 -1.05646324e+00 -1.30937070e-01 7.66720176e-01
-7.11485803e-01 2.92392552e-01 1.15293421e-01 6.02833331e-01
-2.98828214e-01 -7.16883123e-01 8.34106624e-01 -3.70864213e-01
-6.92729354e-01 1.51477158e-01 -5.60769618e-01 -5.34338653e-02
1.53854573e+00 -2.66814947e-01 1.42616570e-01 -2.47933090e-01
-9.76755261e-01 5.99805772e-01 1.98258042e-01 3.04568410e-01
4.48606849e-01 -1.17009890e+00 -7.56311238e-01 2.86084205e-01
3.55687551e-02 -1.69607535e-01 -1.49251625e-01 2.92919010e-01
-2.97750607e-02 1.95235044e-01 -5.47735631e-01 -4.01599765e-01
-1.03241861e+00 8.24450731e-01 6.85471654e-01 -7.20008969e-01
-8.14995915e-02 6.05115950e-01 -5.47788680e-01 -5.52509487e-01
3.90686214e-01 -3.16494219e-02 -4.08157080e-01 3.51871252e-01
4.97081131e-01 3.68785173e-01 3.97375405e-01 1.48427472e-01
-1.26336962e-01 2.92840838e-01 -1.55507922e-01 -2.55288571e-01
1.73022127e+00 9.70795676e-02 1.93734273e-01 5.03864408e-01
1.05020213e+00 -8.13564658e-01 -1.53939509e+00 -9.41487700e-02
1.46423206e-01 -4.33313072e-01 1.28713533e-01 -1.29684711e+00
-8.75765502e-01 5.56754947e-01 9.99716222e-01 5.82056761e-01
1.37688148e+00 -4.08691406e-01 6.36581555e-02 8.87346625e-01
3.26936930e-01 -1.43094563e+00 3.12884957e-01 4.22983408e-01
7.31779993e-01 -1.48414028e+00 1.81972727e-01 9.21257287e-02
-8.19335938e-01 1.31763005e+00 5.80959916e-01 -5.83457530e-01
6.29131079e-01 1.17096908e-01 -1.54835314e-01 2.35339832e-02
-1.30349636e+00 -1.75208539e-01 3.13746929e-01 6.12299204e-01
2.81417400e-01 -5.37209585e-03 -2.53793299e-01 2.94637144e-01
-5.17706387e-02 3.24220240e-01 3.10531288e-01 1.25345469e+00
-3.58348697e-01 -1.30108356e+00 -1.41173765e-01 9.10303354e-01
-7.73382783e-01 7.28606805e-02 -2.57030874e-01 7.68701792e-01
2.14804530e-01 8.65516603e-01 2.33203024e-01 -1.52091935e-01
5.52057683e-01 2.18969196e-01 5.59983790e-01 -6.39834106e-01
-1.07323968e+00 -1.29918590e-01 1.72560096e-01 -6.27670825e-01
-5.19030273e-01 -1.04182756e+00 -8.88702512e-01 1.22617528e-01
-4.13337618e-01 3.88279974e-01 6.27870023e-01 1.07425070e+00
2.53262728e-01 5.51699698e-01 1.07465148e+00 -9.80200112e-01
-1.14578068e+00 -7.88637698e-01 -6.66421056e-01 4.01090235e-01
6.40773177e-01 -7.14381754e-01 -3.01044405e-01 -2.86081731e-01]
|
[4.292187213897705, 2.164916515350342]
|
947832eb-560b-44a0-aa66-bb125d26a0a7
|
the-lambada-dataset-word-prediction-requiring
|
1606.06031
| null |
http://arxiv.org/abs/1606.06031v1
|
http://arxiv.org/pdf/1606.06031v1.pdf
|
The LAMBADA dataset: Word prediction requiring a broad discourse context
|
We introduce LAMBADA, a dataset to evaluate the capabilities of computational
models for text understanding by means of a word prediction task. LAMBADA is a
collection of narrative passages sharing the characteristic that human subjects
are able to guess their last word if they are exposed to the whole passage, but
not if they only see the last sentence preceding the target word. To succeed on
LAMBADA, computational models cannot simply rely on local context, but must be
able to keep track of information in the broader discourse. We show that
LAMBADA exemplifies a wide range of linguistic phenomena, and that none of
several state-of-the-art language models reaches accuracy above 1% on this
novel benchmark. We thus propose LAMBADA as a challenging test set, meant to
encourage the development of new models capable of genuine understanding of
broad context in natural language text.
|
['Raquel Fernández', 'Quan Ngoc Pham', 'Germán Kruszewski', 'Sandro Pezzelle', 'Raffaella Bernardi', 'Angeliki Lazaridou', 'Gemma Boleda', 'Denis Paperno', 'Marco Baroni']
|
2016-06-20
|
the-lambada-dataset-word-prediction-requiring-1
|
https://aclanthology.org/P16-1144
|
https://aclanthology.org/P16-1144.pdf
|
acl-2016-8
|
['lambada']
|
['natural-language-processing']
|
[ 1.09598242e-01 1.92139164e-01 -4.31144863e-01 -1.31957754e-01
-7.51060367e-01 -9.43857789e-01 1.32516742e+00 8.28574002e-01
-5.95382214e-01 6.60748243e-01 7.45606244e-01 -5.38340747e-01
1.52659684e-01 -9.52543557e-01 -4.31210220e-01 -5.76790646e-02
2.23279186e-02 5.62163472e-01 5.31098902e-01 -7.47011185e-01
3.96782428e-01 1.22559905e-01 -1.03873968e+00 8.78501058e-01
4.11251634e-01 3.32082897e-01 2.95446813e-01 7.65453458e-01
-3.03688824e-01 1.22137785e+00 -6.72974467e-01 -4.58392620e-01
-5.15237272e-01 -6.75495565e-01 -1.37607014e+00 -2.30375260e-01
2.28702068e-01 -2.59741414e-02 -3.41482729e-01 5.25610864e-01
1.35410979e-01 3.54054302e-01 8.96055043e-01 -6.67789996e-01
-6.48633540e-01 1.04519546e+00 -6.63432702e-02 6.64616585e-01
9.84269023e-01 8.57261568e-02 1.48520410e+00 -8.41577470e-01
1.12584686e+00 1.32917416e+00 6.22184992e-01 5.04083931e-01
-1.26190615e+00 4.17453386e-02 3.66117150e-01 5.19916713e-01
-8.04448962e-01 -4.25653934e-01 7.08767772e-01 -5.81182778e-01
1.52820945e+00 3.74019861e-01 3.18588585e-01 1.62414193e+00
3.09426516e-01 7.32589245e-01 9.76368725e-01 -8.24033856e-01
2.06923902e-01 5.39917760e-02 7.32095659e-01 5.74152946e-01
-1.25922441e-01 -1.81027368e-01 -8.43821228e-01 -3.24299522e-02
2.50571151e-03 -5.11425197e-01 -3.28937769e-01 2.12321401e-01
-1.48072314e+00 8.77397358e-01 1.94519222e-01 8.09109688e-01
-1.53000325e-01 -1.69541866e-01 5.98680913e-01 2.88850904e-01
5.28226674e-01 8.09978783e-01 -3.76946628e-01 -4.88223523e-01
-7.12085783e-01 6.65564656e-01 1.22677612e+00 4.86477822e-01
3.98279756e-01 -5.65959036e-01 -2.02764884e-01 6.20865047e-01
2.57589133e-03 2.16773987e-01 5.25574327e-01 -6.56288564e-01
6.70066953e-01 6.30824506e-01 1.26909688e-01 -1.23993897e+00
-5.94566286e-01 -2.41039246e-01 -2.92198330e-01 -1.61820844e-01
7.50084221e-01 6.44017896e-03 -3.66974264e-01 1.63899982e+00
-1.09766088e-01 -7.17155961e-03 3.09016675e-01 5.76050758e-01
9.99139547e-01 1.05579042e+00 2.00782910e-01 -4.07008618e-01
1.44584596e+00 -9.04649734e-01 -6.52506709e-01 -8.24426055e-01
1.07197213e+00 -6.53371215e-01 1.53060150e+00 3.30818236e-01
-1.06080985e+00 -3.71295869e-01 -1.26589596e+00 -5.09974420e-01
-5.20794511e-01 -1.81840941e-01 3.88475716e-01 1.94558859e-01
-6.79626465e-01 3.70056063e-01 -6.98323607e-01 -5.23503125e-01
2.21917897e-01 -4.04026002e-01 -2.83776879e-01 -5.80232739e-02
-1.39752138e+00 1.38793635e+00 4.71908659e-01 -1.72259510e-01
-7.28219509e-01 -7.17743993e-01 -8.99451077e-01 1.78244308e-01
5.41131914e-01 -5.12312829e-01 1.29554021e+00 -7.76400030e-01
-9.80986714e-01 1.43523550e+00 -5.28447747e-01 -6.81822360e-01
5.64127445e-01 -3.99175733e-01 -4.95921701e-01 2.05779016e-01
1.31464675e-01 6.69430122e-02 4.36513603e-01 -1.29864752e+00
-3.63656461e-01 -1.42297387e-01 4.52288330e-01 -6.49335086e-02
-2.01788291e-01 1.93457857e-01 7.50575662e-02 -7.00376451e-01
-2.85612673e-01 -6.75787866e-01 4.08740342e-02 -3.24426830e-01
-5.94902635e-01 -4.82079417e-01 6.46339417e-01 -6.00286305e-01
1.55647433e+00 -1.78739321e+00 1.44682705e-01 -2.11335644e-01
1.67474538e-01 2.01813459e-01 -2.09370762e-01 1.10467935e+00
4.89350259e-02 5.13563037e-01 -1.72846287e-01 -3.60500365e-01
8.91945735e-02 9.31326300e-02 -8.37358117e-01 2.29409620e-01
2.64602602e-01 9.69103277e-01 -1.02521336e+00 -4.42649603e-01
-3.09974067e-02 5.40548526e-02 -4.08535898e-01 -1.54171847e-02
-7.54602730e-01 3.40527147e-02 -5.49621999e-01 -5.28016165e-02
-1.86258834e-02 -4.43660349e-01 1.90054625e-01 2.09006160e-01
-7.36412555e-02 9.58972991e-01 -7.07505345e-01 1.56535542e+00
-5.36598086e-01 1.19335139e+00 -6.07129693e-01 -1.01954639e+00
6.32589757e-01 3.70207936e-01 -2.27251068e-01 -6.45933986e-01
2.43780151e-01 -3.07491608e-03 8.51614028e-02 -7.51945734e-01
5.79318762e-01 -3.34304303e-01 -4.08709556e-01 7.43241251e-01
-1.42422110e-01 -5.14666140e-02 5.02772748e-01 5.11211455e-01
1.05885112e+00 -2.18442187e-01 6.08055055e-01 -3.69040668e-01
8.34844112e-01 4.54945952e-01 3.72201353e-02 1.06498575e+00
-1.97166707e-02 4.69180226e-01 7.61711240e-01 -6.43850267e-01
-9.62017357e-01 -9.69217360e-01 -5.25027178e-02 1.44641387e+00
2.88343608e-01 -9.03250277e-01 -6.00157738e-01 -6.36470079e-01
-2.87494451e-01 1.51313210e+00 -9.72623765e-01 4.12712954e-02
-7.95061827e-01 -5.25770485e-01 6.18644416e-01 4.00775343e-01
2.08597243e-01 -1.35821664e+00 -7.07852900e-01 3.71179730e-01
-5.34789681e-01 -1.23859847e+00 -1.80687621e-01 -9.60110575e-02
-3.61035913e-01 -1.12804782e+00 6.74051931e-03 -9.16859925e-01
-3.42919379e-02 5.71357757e-02 1.57675076e+00 5.06187797e-01
-6.63203821e-02 4.13064450e-01 -5.81664801e-01 -4.92676586e-01
-9.88819540e-01 3.72059315e-01 -3.22061807e-01 -3.35917413e-01
5.72511077e-01 -3.82498205e-01 -1.07378364e-01 -4.50905673e-02
-7.56361127e-01 2.09175646e-01 -1.88345507e-01 9.08003390e-01
1.07031547e-01 -3.94192308e-01 7.21626043e-01 -1.08572137e+00
9.52567756e-01 -7.28804171e-01 5.58991507e-02 4.40995991e-01
-6.76998049e-02 2.23506126e-03 6.42380118e-01 -5.70260584e-01
-9.97884274e-01 -6.48550272e-01 -2.83176422e-01 5.01159847e-01
-1.21119164e-01 1.01088750e+00 1.22607417e-01 7.50456750e-01
9.85382974e-01 2.96483099e-01 -2.27326214e-01 -4.62122262e-01
4.92808789e-01 3.65027338e-01 6.93195522e-01 -6.95663154e-01
5.27538836e-01 3.25491607e-01 -6.85015097e-02 -1.13174284e+00
-1.33480775e+00 -4.28548694e-01 -7.25351691e-01 -1.26072109e-01
8.95117342e-01 -7.27171123e-01 -7.12325156e-01 1.02893613e-01
-1.57616174e+00 -4.68170166e-01 -1.03015065e-01 2.24218786e-01
-6.48896873e-01 2.38011152e-01 -6.00431383e-01 -6.03663862e-01
-6.39848858e-02 -6.39415324e-01 6.61521256e-01 -1.61941368e-02
-1.00954390e+00 -1.49579346e+00 1.68074325e-01 3.40647459e-01
1.57310925e-02 4.39704657e-01 1.39424336e+00 -1.35004866e+00
-2.05268294e-01 -3.12916547e-01 2.48102784e-01 8.82211514e-03
-4.53291554e-03 -2.31913431e-03 -9.61151123e-01 -6.88397363e-02
1.76778585e-01 -7.01960981e-01 9.16074693e-01 -5.46074882e-02
5.93366742e-01 -3.94493788e-01 -5.07789493e-01 -7.96437487e-02
1.26100862e+00 -3.34735885e-02 5.38858533e-01 6.68021798e-01
3.81104380e-01 6.11721337e-01 3.23024958e-01 1.27632022e-01
6.08767152e-01 4.67813075e-01 1.26954779e-01 2.57017493e-01
-2.04415992e-01 -5.24000704e-01 1.92094296e-01 7.88629651e-01
1.53085604e-01 -8.79592061e-01 -1.25656557e+00 1.04347801e+00
-1.79938209e+00 -1.27209187e+00 -3.75180364e-01 1.68329072e+00
1.00578141e+00 6.18740022e-01 -9.94612128e-02 2.27085650e-01
2.62160480e-01 7.59249330e-01 -1.05121225e-01 -5.96518338e-01
-3.73331189e-01 1.72561347e-01 -2.39136145e-01 1.04946303e+00
-1.06792223e+00 1.12128043e+00 7.19853926e+00 7.65797079e-01
-7.77309537e-01 7.47604622e-03 5.67923248e-01 1.54950157e-01
-4.28933501e-01 2.87349243e-02 -7.24431753e-01 2.68555105e-01
9.55687404e-01 -4.86197025e-01 2.03099340e-01 5.94529688e-01
2.37606019e-01 -4.26994175e-01 -1.47257483e+00 4.24096882e-01
3.82701993e-01 -1.86798561e+00 1.74318388e-01 -3.89253616e-01
5.72127938e-01 -1.01592861e-01 -3.92153323e-01 2.49696806e-01
1.22863136e-01 -1.42291462e+00 7.62143850e-01 4.45329010e-01
2.25023955e-01 -4.83834535e-01 5.82314670e-01 9.71902311e-01
-6.94333136e-01 1.22323439e-01 -9.04141515e-02 -6.21487081e-01
4.78677154e-01 -1.23254880e-01 -8.43335032e-01 2.62402087e-01
2.20466390e-01 7.01538920e-01 -7.96788037e-01 4.37027216e-01
-4.91258353e-01 9.54800487e-01 -8.86461735e-02 -5.28269768e-01
4.79944974e-01 2.69905239e-01 8.58077347e-01 1.58893216e+00
3.64955999e-02 4.52150702e-01 2.24118590e-01 8.43967676e-01
-2.58403812e-02 2.85667688e-01 -5.37116587e-01 -2.30761856e-01
4.75973010e-01 7.54589081e-01 -6.15508497e-01 -4.09327954e-01
-5.46243131e-01 6.67114139e-01 5.17719090e-01 1.74817935e-01
-4.08130616e-01 -1.18415497e-01 4.60684925e-01 2.53011584e-01
1.63551331e-01 -4.29346383e-01 -5.81891060e-01 -1.12662005e+00
4.02099155e-02 -9.31549430e-01 4.71146047e-01 -9.54236448e-01
-1.27515984e+00 7.46073604e-01 -1.64439697e-02 -6.28003061e-01
-6.37415707e-01 -7.29885519e-01 -9.93335366e-01 8.07264626e-01
-1.23753369e+00 -1.20277083e+00 1.73403658e-02 3.68697912e-01
9.19522464e-01 -6.89414889e-02 1.05667281e+00 -4.21300262e-01
-4.69099134e-02 1.78294241e-01 -1.37559354e-01 3.74710858e-01
6.57914579e-01 -1.08080149e+00 5.65689981e-01 7.33387947e-01
6.53069854e-01 6.16984367e-01 1.34563720e+00 -4.60721880e-01
-1.03051507e+00 -4.67168361e-01 1.64908898e+00 -1.18028843e+00
1.06980836e+00 -4.43939656e-01 -1.37283385e+00 1.02492726e+00
5.79279900e-01 -3.64445567e-01 6.25572622e-01 4.88516986e-01
-5.72601795e-01 4.57364142e-01 -5.87288320e-01 7.46934056e-01
9.77001667e-01 -9.55666184e-01 -1.66677654e+00 4.21755403e-01
8.70560884e-01 -3.58723491e-01 -5.57104349e-01 4.19328883e-02
3.30681801e-01 -9.14646506e-01 9.63541269e-01 -1.21217024e+00
1.07609463e+00 -1.69897135e-02 -1.57351270e-01 -1.21333075e+00
-2.26890538e-02 -4.35836613e-01 -6.95056170e-02 1.36023796e+00
7.42000043e-01 -3.30902696e-01 3.79649818e-01 3.06541771e-01
2.47580577e-02 -5.08565247e-01 -8.88711631e-01 -6.64377570e-01
6.78977132e-01 -8.00550461e-01 2.41463795e-01 9.91994381e-01
6.99076176e-01 8.90093207e-01 -8.26628208e-02 -1.08003788e-01
-2.35821363e-02 4.69976813e-02 6.06829703e-01 -1.01094639e+00
-2.02398315e-01 -5.90353608e-01 -1.57171726e-01 -1.13571525e+00
4.46224332e-01 -9.51566517e-01 1.47513732e-01 -1.69981658e+00
2.78082550e-01 4.14804779e-02 5.33957966e-02 2.58743435e-01
-3.48719478e-01 6.76400363e-02 2.85611987e-01 9.18253139e-02
-5.94868481e-01 3.91351193e-01 1.21399283e+00 -1.69192269e-01
-1.38724878e-01 -2.62260377e-01 -7.13085353e-01 9.97771323e-01
4.49521899e-01 -3.86420399e-01 -4.68979239e-01 -4.35840189e-01
5.98357499e-01 1.51561916e-01 5.37572801e-01 -5.77282965e-01
3.00782919e-01 -5.37483633e-01 1.71819478e-01 -5.27505159e-01
3.02977622e-01 -3.64353031e-01 -3.07752132e-01 2.90391922e-01
-9.28995073e-01 2.05505684e-01 3.31893414e-01 3.83590877e-01
-3.04247528e-01 -4.17921066e-01 3.82811785e-01 -1.94932356e-01
-9.62556660e-01 -3.78665775e-01 -9.23406661e-01 7.04094410e-01
8.14838111e-01 3.29221413e-02 -7.68190444e-01 -5.29629946e-01
-8.16892087e-01 9.34575573e-02 4.67711449e-01 7.41437852e-01
3.97141576e-01 -9.71650362e-01 -9.53940392e-01 -2.75491357e-01
3.00722212e-01 -2.53831506e-01 1.24440536e-01 2.88924068e-01
-5.78816056e-01 5.86673021e-01 3.02248895e-01 -3.97960693e-01
-1.19374704e+00 7.23353207e-01 3.04972917e-01 -5.90929747e-01
-8.00428152e-01 7.56987751e-01 1.91381291e-01 -1.95289120e-01
-9.41873640e-02 -2.89398342e-01 -4.08721119e-01 2.48894557e-01
9.74152982e-01 -4.99700233e-02 -1.56082362e-01 -7.80495882e-01
-4.95376408e-01 4.09649193e-01 -1.01648569e-01 -2.96769083e-01
1.24121964e+00 -4.31434005e-01 -1.11133620e-01 1.08773601e+00
1.08329690e+00 2.40434632e-01 -7.88255334e-01 -4.98432666e-01
5.92371523e-01 -2.97724515e-01 -1.39383465e-01 -1.12077057e+00
-1.37042720e-03 8.79004955e-01 -2.47671232e-01 6.07291758e-01
7.38630712e-01 3.71986330e-01 6.76082730e-01 7.32526600e-01
7.93715641e-02 -8.48696172e-01 2.55356640e-01 1.04974270e+00
1.28809965e+00 -1.11369622e+00 7.59872720e-02 -3.94217759e-01
-7.68407881e-01 1.34495878e+00 3.50565761e-01 -2.57286340e-01
5.11311948e-01 1.58353239e-01 1.88727409e-01 -3.58723611e-01
-1.42006612e+00 2.79888883e-02 4.36915666e-01 4.08078700e-01
5.55333972e-01 -2.03070581e-01 -4.73882645e-01 9.03221548e-01
-5.78965127e-01 -2.69662201e-01 7.76917696e-01 7.13214874e-01
-5.30330300e-01 -9.97645915e-01 -1.36532143e-01 -5.23826368e-02
-5.82381189e-01 -1.77924961e-01 -6.72918379e-01 1.11999929e+00
-8.87226611e-02 1.14814937e+00 1.70330971e-01 -2.35237963e-02
2.69564480e-01 3.92699152e-01 3.97901505e-01 -6.93207681e-01
-6.65850401e-01 -3.34709167e-01 7.89100409e-01 -2.16456130e-01
-4.68604594e-01 -6.94055557e-01 -1.29213083e+00 -4.42759037e-01
8.30991417e-02 1.51618958e-01 3.24971855e-01 1.64840901e+00
-2.29939923e-01 3.37525576e-01 2.85564661e-01 -4.83963251e-01
-2.93399572e-01 -1.00957477e+00 -1.59752831e-01 6.71089292e-01
6.31107390e-01 -3.98741037e-01 -2.98400819e-01 4.43009228e-01]
|
[11.106837272644043, 8.879469871520996]
|
c08c3ae6-01a7-4d75-85f6-a095b0d85f2c
|
ufal-corpipe-at-crac-2022-effectivity-of
|
2209.07278
| null |
https://arxiv.org/abs/2209.07278v1
|
https://arxiv.org/pdf/2209.07278v1.pdf
|
ÚFAL CorPipe at CRAC 2022: Effectivity of Multilingual Models for Coreference Resolution
|
We describe the winning submission to the CRAC 2022 Shared Task on Multilingual Coreference Resolution. Our system first solves mention detection and then coreference linking on the retrieved spans with an antecedent-maximization approach, and both tasks are fine-tuned jointly with shared Transformer weights. We report results of fine-tuning a wide range of pretrained models. The center of this contribution are fine-tuned multilingual models. We found one large multilingual model with sufficiently large encoder to increase performance on all datasets across the board, with the benefit not limited only to the underrepresented languages or groups of typologically relative languages. The source code is available at https://github.com/ufal/crac2022-corpipe.
|
['Jana Straková', 'Milan Straka']
|
2022-09-15
| null |
https://aclanthology.org/2022.crac-mcr.4
|
https://aclanthology.org/2022.crac-mcr.4.pdf
|
crac-acl-2022-10
|
['coreference-resolution']
|
['natural-language-processing']
|
[-4.00325507e-01 4.06187057e-01 -6.17242157e-01 -3.25468004e-01
-1.59578645e+00 -8.88849795e-01 7.22129643e-01 -3.00979242e-02
-6.34608746e-01 1.18626773e+00 8.98179293e-01 -3.11677575e-01
-2.14202836e-01 -3.02859664e-01 -7.61291504e-01 -1.10811420e-01
-9.97675210e-02 1.24995577e+00 -5.07569611e-02 -6.62014842e-01
2.09922567e-02 2.10562736e-01 -1.18318367e+00 6.52947247e-01
9.81906891e-01 1.88858986e-01 2.58857638e-01 3.44103068e-01
-4.85350043e-02 6.75752759e-01 -1.77946255e-01 -8.00906062e-01
2.83078905e-02 1.98149178e-02 -1.34435451e+00 -1.05943024e+00
9.07801628e-01 2.30155900e-01 -5.40002465e-01 1.08676910e+00
6.28621757e-01 1.23333953e-01 3.99929583e-01 -9.11708236e-01
-3.49102259e-01 1.68008041e+00 -3.53643358e-01 4.06731635e-01
6.87729180e-01 -3.06245536e-01 1.59426486e+00 -9.27613318e-01
1.06009686e+00 1.66640055e+00 8.48305881e-01 5.40111303e-01
-1.15821242e+00 -1.29037499e+00 2.15430811e-01 3.60171139e-01
-1.46771574e+00 -9.98332322e-01 3.83652329e-01 -1.65276974e-01
1.50020969e+00 3.14001679e-01 1.13936938e-01 1.14845562e+00
-2.29597807e-01 6.71790600e-01 7.98180401e-01 -3.93345654e-01
-5.72020471e-01 -1.40672950e-02 2.42034957e-01 3.52539867e-01
4.20358896e-01 2.37051472e-01 -6.20635569e-01 -2.78366387e-01
2.28189826e-01 -6.80261672e-01 -5.46974659e-01 -2.78635845e-02
-1.30057395e+00 8.35328221e-01 4.37369168e-01 7.38584220e-01
-1.00443631e-01 -5.85670285e-02 4.53493714e-01 4.61703628e-01
2.91575909e-01 8.55822384e-01 -8.68603349e-01 -1.90214708e-03
-1.00261259e+00 3.29990923e-01 1.01082695e+00 1.24713171e+00
5.89875877e-01 -4.17716384e-01 9.79177281e-02 1.07411492e+00
2.27951154e-01 7.34312773e-01 1.19673647e-01 -1.29813147e+00
8.70739698e-01 2.47911498e-01 1.27696455e-01 -7.89976537e-01
-7.04347432e-01 -4.53541219e-01 -3.30349535e-01 -4.83885258e-01
5.11265993e-01 -3.81561279e-01 -2.80450702e-01 2.19637918e+00
2.49177013e-02 -1.14592835e-01 8.72124508e-02 8.27815056e-01
1.18141365e+00 3.02519709e-01 4.52221036e-01 -1.45423785e-01
1.71036208e+00 -8.59324038e-01 -8.21960747e-01 -3.23410004e-01
7.48802066e-01 -1.15781581e+00 5.07157087e-01 1.36464313e-02
-1.25320470e+00 -2.31210440e-01 -8.14054906e-01 -4.14130956e-01
-3.25393111e-01 -1.23627834e-01 7.66849995e-01 1.78611562e-01
-1.16689718e+00 3.37416619e-01 -4.40949976e-01 -7.38032818e-01
-7.00777844e-02 2.94188887e-01 -5.56716681e-01 -9.10654515e-02
-2.05721498e+00 1.42671072e+00 7.23820984e-01 -1.35269329e-01
-5.85749209e-01 -9.91808295e-01 -9.39462006e-01 8.35821778e-02
3.43213499e-01 -5.80105662e-01 1.35649180e+00 -5.55866897e-01
-7.69994140e-01 1.35978127e+00 -1.90690383e-01 -5.54103136e-01
3.77203226e-01 -4.63528156e-01 -8.96905422e-01 -1.42639801e-01
5.53568602e-01 9.97145712e-01 -1.86176822e-02 -1.12301779e+00
-9.50879455e-01 -4.37028557e-02 2.13119518e-02 3.84909481e-01
1.46307334e-01 5.44101596e-01 -5.20203710e-01 -4.59940255e-01
-2.54386514e-01 -9.05468404e-01 1.69647947e-01 -1.03426385e+00
-2.86057383e-01 -5.19574761e-01 2.31828481e-01 -9.02003586e-01
1.35820055e+00 -1.76628339e+00 1.50093496e-01 -1.49902612e-01
-1.61135420e-01 1.58956096e-01 -4.21716124e-01 6.50357127e-01
-4.93148953e-01 1.06833667e-01 -4.37150593e-04 -2.30912745e-01
2.42399201e-01 -1.01188578e-01 -4.87971693e-01 3.52000952e-01
-1.43644273e-01 8.05511355e-01 -1.12279713e+00 -6.93780959e-01
-9.15842652e-02 2.77508557e-01 -6.05817795e-01 -5.87239601e-02
-1.34830788e-01 3.21191519e-01 -1.59290195e-01 5.04679680e-01
6.54982209e-01 6.82465434e-02 8.25753212e-01 -5.75220942e-01
-4.28234428e-01 9.74760175e-01 -1.17413640e+00 2.18626142e+00
-3.71329874e-01 4.52629775e-01 5.50836504e-01 -6.83393180e-01
7.25639284e-01 5.79885364e-01 3.46108258e-01 -7.50908077e-01
-6.25573918e-02 6.93629146e-01 3.39732431e-02 -1.06156781e-01
8.44690382e-01 -7.92898461e-02 -5.96154153e-01 4.41752523e-01
5.09672999e-01 8.31673071e-02 3.69167566e-01 5.97438157e-01
6.79787159e-01 1.60554931e-01 4.15151596e-01 -9.37047422e-01
5.77157795e-01 3.92765731e-01 1.00517225e+00 4.99251783e-01
-1.54466584e-01 3.20597023e-01 1.96369991e-01 -2.59515077e-01
-8.53154004e-01 -8.66227269e-01 -5.54039419e-01 1.43752170e+00
-1.20848849e-01 -5.48867404e-01 -4.54841852e-01 -6.05334818e-01
1.86712205e-01 1.00203013e+00 -3.57823402e-01 2.08298758e-01
-1.00037396e+00 -5.62261641e-01 1.11426830e+00 2.51470715e-01
8.36760327e-02 -1.22088718e+00 -7.51391286e-03 1.91874132e-01
-9.75579977e-01 -1.08518016e+00 -8.59288335e-01 2.46210739e-01
-5.61765850e-01 -1.46636224e+00 -4.13615435e-01 -1.05862010e+00
-3.31582576e-02 -8.71968716e-02 1.72732210e+00 -8.92064124e-02
-2.93100458e-02 3.60764563e-01 1.75584145e-02 -9.75030810e-02
-1.59239218e-01 7.10830986e-01 3.09409589e-01 -7.31503546e-01
8.11494172e-01 -3.50051045e-01 -1.32105842e-01 -2.42059212e-02
-1.13454007e-01 -1.32753879e-01 8.37292150e-02 7.93005347e-01
3.13988000e-01 -4.76263613e-01 6.74439073e-01 -1.18701446e+00
4.86182123e-01 -5.19580960e-01 -5.76924324e-01 5.32267272e-01
-5.39759755e-01 6.54280633e-02 -2.88331266e-02 -3.82332280e-02
-1.33971989e+00 -2.96391398e-01 -1.41254053e-01 -7.11387247e-02
-1.40235424e-01 5.65796494e-01 -3.89652163e-01 3.51435810e-01
4.54249948e-01 -4.64959741e-01 -4.48123544e-01 -8.13518405e-01
6.80867791e-01 5.25823116e-01 1.00297356e+00 -1.26001680e+00
3.60352665e-01 -9.69822183e-02 -6.17907941e-01 -4.17520851e-01
-8.81461859e-01 -4.69293237e-01 -6.89112902e-01 1.47441551e-01
6.75274372e-01 -1.46395481e+00 -7.84360111e-01 3.87708806e-02
-1.34182858e+00 -4.25885528e-01 1.65210292e-02 7.37816811e-01
-3.17778826e-01 6.79478645e-02 -9.97140825e-01 -2.55714923e-01
-7.47439861e-01 -8.72200787e-01 7.71039784e-01 1.46463245e-01
-7.11284459e-01 -1.22964001e+00 6.76089346e-01 5.35676420e-01
1.87351495e-01 -1.17043413e-01 1.00127423e+00 -9.94469464e-01
-1.87401190e-01 2.55216599e-01 -2.83445925e-01 -4.41661805e-01
-2.70462126e-01 -8.78201872e-02 -8.12128484e-01 -5.95211506e-01
-5.90870559e-01 -4.19439286e-01 1.08494782e+00 4.91713583e-01
2.03906715e-01 -1.90998688e-02 -7.72451341e-01 8.39555919e-01
1.35035229e+00 2.83571309e-03 4.01611030e-01 7.14958727e-01
4.85822499e-01 6.06778979e-01 5.55279016e-01 1.15011411e-03
9.03509021e-01 8.14733863e-01 8.23680591e-03 2.09536701e-01
-4.55356568e-01 -4.05548066e-01 2.17166424e-01 1.09478819e+00
-1.76030129e-01 -3.64184119e-02 -1.28059697e+00 1.08310759e+00
-1.97757459e+00 -1.14777088e+00 -2.39743695e-01 1.99851918e+00
1.32977533e+00 -1.87418699e-01 -8.47140029e-02 -6.77234650e-01
1.01416647e+00 6.60951883e-02 -9.36602056e-02 -4.44768876e-01
-6.41757369e-01 4.42771018e-01 4.78189260e-01 1.19705594e+00
-1.21071827e+00 1.43880761e+00 6.82080698e+00 6.35194361e-01
-6.66580677e-01 3.42972249e-01 -7.38498494e-02 -3.82675588e-01
-6.38593614e-01 3.54628056e-01 -1.29853404e+00 7.31196627e-02
1.13578463e+00 -3.71173590e-01 5.39255559e-01 1.94848821e-01
-3.19311976e-01 2.32004106e-01 -1.00104761e+00 5.70404232e-01
-2.63606180e-02 -1.20901370e+00 -1.06684640e-01 -1.09728843e-01
6.74707472e-01 6.89832628e-01 -2.36457944e-01 7.52285957e-01
1.06779790e+00 -9.48258221e-01 6.55870795e-01 4.96136129e-01
1.09050155e+00 -8.14975739e-01 7.86140323e-01 -1.63043320e-01
-1.18767893e+00 2.21745968e-02 -3.17025036e-01 2.94931412e-01
3.87492299e-01 3.38138491e-01 -3.42443883e-01 8.89944851e-01
8.40485215e-01 7.79886365e-01 -3.26361150e-01 7.96697676e-01
-4.27502096e-01 4.20157254e-01 -2.60420293e-01 6.10351443e-01
1.42175481e-01 1.31127894e-01 8.38009775e-01 1.78692353e+00
2.54011244e-01 1.26688689e-01 9.41456929e-02 5.79225361e-01
-4.18055475e-01 2.24868968e-01 -3.72899711e-01 2.51891881e-01
1.29741037e+00 1.27157509e+00 1.78760841e-01 -1.92244932e-01
-4.40588355e-01 3.47079664e-01 9.53395665e-01 3.62576962e-01
-7.22442925e-01 -4.03948307e-01 6.96245790e-01 -2.69418418e-01
1.24680385e-01 1.88752785e-01 -1.63427014e-02 -1.30266893e+00
-6.81254983e-01 -1.28055084e+00 1.29653752e+00 -4.39547122e-01
-1.24418151e+00 6.07132316e-01 1.13278158e-01 -6.50590479e-01
-4.56942081e-01 -3.10056508e-01 -4.00182158e-01 1.15667367e+00
-1.51628566e+00 -1.33788908e+00 2.61587858e-01 7.89136291e-01
8.36296976e-02 -4.51041669e-01 1.17682910e+00 7.99962342e-01
-5.25027931e-01 1.05909002e+00 -1.18537553e-01 3.45642924e-01
1.59040940e+00 -1.20209587e+00 2.92040169e-01 6.98815525e-01
5.15413471e-02 1.05533481e+00 7.63569355e-01 -7.63326466e-01
-1.00587153e+00 -8.45249057e-01 1.78196990e+00 -5.47512591e-01
9.77258146e-01 -1.98045492e-01 -7.44555354e-01 1.27744901e+00
8.60169053e-01 -4.90850121e-01 5.67917466e-01 1.13448513e+00
-7.38825083e-01 -6.35143071e-02 -9.35561657e-01 2.81689227e-01
1.11236560e+00 -7.78506517e-01 -1.14802098e+00 8.12716335e-02
6.42935693e-01 -5.97518921e-01 -1.44754469e+00 5.57155550e-01
4.80670959e-01 -3.59972209e-01 1.02925932e+00 -8.84889305e-01
9.17958394e-02 -5.91334179e-02 -4.80054259e-01 -1.42992234e+00
-8.23804736e-01 -5.83068073e-01 3.88709493e-02 1.56615245e+00
9.44183528e-01 -6.59091592e-01 7.35688210e-02 3.18413764e-01
-2.67482728e-01 9.43580568e-02 -9.50307012e-01 -4.51451421e-01
6.13862455e-01 -1.99309856e-01 6.98947251e-01 1.55364227e+00
5.74101090e-01 7.06768513e-01 -2.25161627e-01 1.38876513e-01
8.27530026e-01 4.22712147e-01 3.35665822e-01 -1.37852585e+00
-3.70360613e-02 -5.72635710e-01 3.97494733e-01 -4.66597199e-01
6.74828589e-01 -1.52385557e+00 -7.00066909e-02 -1.30031669e+00
4.46945399e-01 -5.84487438e-01 -4.18756366e-01 6.71597660e-01
-2.97896534e-01 6.90936903e-03 3.39919657e-01 4.25635248e-01
-7.75598884e-01 -9.41124000e-03 8.34916890e-01 -2.58738786e-01
3.19442563e-02 -4.16577309e-01 -1.04872954e+00 4.09140408e-01
7.17257082e-01 -5.86514771e-01 1.52845904e-01 -8.40052783e-01
3.05490047e-01 1.55645788e-01 -1.37656212e-01 -5.57294786e-01
5.05101442e-01 1.92590542e-02 5.33022173e-02 -6.32546067e-01
4.30638604e-02 -3.22332501e-01 4.27658170e-01 3.99511486e-01
-5.28766036e-01 2.38001212e-01 5.85632861e-01 -9.53165069e-02
-3.03799450e-01 -1.22534461e-01 5.84857345e-01 -2.90243417e-01
-9.38881397e-01 -2.46998463e-02 -4.74582314e-02 7.81807423e-01
3.73886764e-01 7.18563557e-01 -7.83275604e-01 -1.95766419e-01
-7.84874797e-01 7.91414440e-01 2.68708378e-01 7.65534818e-01
-1.75919011e-01 -1.51050019e+00 -1.31881762e+00 -1.33396968e-01
1.65875077e-01 -4.83078986e-01 2.65490085e-01 9.00413692e-01
-1.31672338e-01 1.32701778e+00 -2.87679404e-01 -7.86136016e-02
-1.23751032e+00 5.65393209e-01 5.24223745e-01 -6.24149323e-01
-4.12427396e-01 8.01778257e-01 -5.62586486e-02 -1.16111660e+00
2.94423670e-01 3.22921306e-01 -4.41926360e-01 4.28750724e-01
5.45327008e-01 3.52710158e-01 4.82753851e-02 -8.72433007e-01
-9.28862333e-01 4.65528846e-01 -3.96421373e-01 -1.63348436e-01
1.26749194e+00 -1.53185099e-01 -4.20622796e-01 1.61308840e-01
9.48122740e-01 2.90405035e-01 -4.86390471e-01 -4.94945765e-01
4.50244784e-01 2.50250787e-01 4.13897298e-02 -1.25550270e+00
-1.09086561e+00 4.83239979e-01 2.56878883e-01 -4.65517730e-01
5.93546927e-01 2.57584929e-01 6.36395574e-01 2.83477575e-01
5.69675684e-01 -1.19729292e+00 -9.67463672e-01 1.02291274e+00
9.58509266e-01 -1.07516074e+00 1.40105663e-02 -3.99729647e-02
-6.10514700e-01 7.32760549e-01 6.41503930e-01 -1.35100055e-02
5.21250427e-01 6.70065343e-01 2.53826737e-01 -3.53265643e-01
-1.02157378e+00 -2.65386224e-01 4.03453141e-01 5.09649754e-01
1.14648604e+00 5.21810412e-01 -7.30528414e-01 1.00340092e+00
-4.49402958e-01 -2.53512353e-01 4.06764448e-02 3.04411590e-01
-2.35076308e-01 -1.22625554e+00 -4.00757283e-01 5.16164489e-02
-6.83175266e-01 -7.00571895e-01 -4.61153060e-01 1.15783322e+00
8.51664096e-02 9.50307608e-01 2.25327864e-01 -2.51888901e-01
5.12081265e-01 2.68434018e-01 6.28503740e-01 -4.42883909e-01
-8.44767869e-01 5.53764030e-02 9.35633123e-01 -5.85209787e-01
-4.49766308e-01 -1.14312959e+00 -1.20475495e+00 -7.12784410e-01
-5.54290675e-02 6.30333424e-01 5.60194068e-03 6.18259728e-01
3.65942955e-01 2.67052203e-01 2.19814759e-02 -5.26830137e-01
-3.16294640e-01 -1.40457451e+00 -3.57317895e-01 3.65986854e-01
7.35441297e-02 -6.21071398e-01 -2.53959507e-01 -3.77063990e-01]
|
[9.31447696685791, 9.582733154296875]
|
f9bb6dfc-38b5-478e-b1ca-9b7e886a4dd5
|
dote-rethinking-predictive-wan-traffic
|
2303.00735
| null |
https://arxiv.org/abs/2303.00735v2
|
https://arxiv.org/pdf/2303.00735v2.pdf
|
A Deep Learning Perspective on Network Routing
|
Routing is, arguably, the most fundamental task in computer networking, and the most extensively studied one. A key challenge for routing in real-world environments is the need to contend with uncertainty about future traffic demands. We present a new approach to routing under demand uncertainty: tackling this challenge as stochastic optimization, and employing deep learning to learn complex patterns in traffic demands. We show that our method provably converges to the global optimum in well-studied theoretical models of multicommodity flow. We exemplify the practical usefulness of our approach by zooming in on the real-world challenge of traffic engineering (TE) on wide-area networks (WANs). Our extensive empirical evaluation on real-world traffic and network topologies establishes that our approach's TE quality almost matches that of an (infeasible) omniscient oracle, outperforming previously proposed approaches, and also substantially lowers runtimes.
|
['Aviv Tamar', 'Michael Schapira', 'Ishai Menache', 'Srikanth Kandula', 'Chaim Hoch', 'Felipe Vieira Frujeri', 'Yarin Perry']
|
2023-03-01
| null | null | null | null |
['stochastic-optimization']
|
['methodology']
|
[-1.09814825e-02 -6.31595179e-02 -6.97965801e-01 -4.38733339e-01
-7.09606290e-01 -6.74652338e-01 2.35569272e-02 -3.76754463e-01
-1.55817091e-01 1.43886566e+00 -1.52255818e-01 -1.15241015e+00
-6.89331532e-01 -6.85785353e-01 -6.12978041e-01 -4.90514308e-01
-8.20370436e-01 1.22155666e+00 1.81289792e-01 -1.69774845e-01
4.18147802e-01 9.44446802e-01 -8.54520261e-01 -1.70299202e-01
4.04584348e-01 1.49066496e+00 -2.15706825e-01 6.73573256e-01
-3.34084302e-01 5.17719090e-01 -5.77003062e-01 -5.61157942e-01
6.15117192e-01 2.65915304e-01 -8.54931593e-01 1.77083045e-01
1.94762126e-01 -2.73147881e-01 -4.72301096e-01 7.25308836e-01
4.09026414e-01 -2.89094836e-01 2.85448641e-01 -1.95269501e+00
-1.23758316e-02 9.89903569e-01 -5.89555144e-01 6.11092627e-01
-2.32283950e-01 3.33058059e-01 1.30301595e+00 -5.18247485e-02
4.53739494e-01 1.40933084e+00 3.86951327e-01 3.32832813e-01
-1.61953950e+00 -8.08699608e-01 6.01150930e-01 1.68665394e-01
-1.01413977e+00 -7.69296050e-01 6.41729116e-01 -1.74640473e-02
7.13845968e-01 1.99477807e-01 1.11631952e-01 7.95572340e-01
2.01803938e-01 7.84725487e-01 5.92857480e-01 1.03148647e-01
4.79790241e-01 1.87398493e-01 -5.90339720e-01 3.19764197e-01
3.08616042e-01 2.18371436e-01 1.10635765e-01 -1.57677144e-01
7.51135826e-01 -4.32345450e-01 -4.31613959e-02 -8.13485086e-01
-1.18134081e+00 6.35066688e-01 1.69094786e-01 -2.37837732e-01
-4.74321514e-01 8.26819658e-01 4.24608022e-01 8.21757138e-01
3.00390095e-01 2.87931353e-01 -9.11511421e-01 -4.11193639e-01
-8.81465614e-01 3.12465932e-02 1.42187011e+00 1.32361233e+00
5.62891543e-01 4.12791461e-01 2.10590467e-01 3.55790615e-01
3.16299111e-01 5.83754659e-01 -5.37702680e-01 -1.54343510e+00
8.72674584e-01 -3.67574185e-01 3.76214892e-01 -1.00117874e+00
-5.87372780e-01 -8.22281122e-01 -7.28152692e-01 1.92970648e-01
6.17573023e-01 -9.26369369e-01 -3.31474543e-01 1.96778154e+00
2.07326226e-02 4.78780806e-01 -1.43924743e-01 5.37732840e-01
-2.92186409e-01 6.88277185e-01 -1.21570885e-01 -5.51438451e-01
5.02806127e-01 -8.76607299e-01 -5.30997515e-01 -1.52998134e-01
2.99425811e-01 -6.13477767e-01 1.68799192e-01 3.97987396e-01
-1.35162282e+00 -3.45525518e-02 -6.53943241e-01 6.63097799e-01
-7.74989650e-02 -6.38459325e-01 8.17914784e-01 1.16008186e+00
-1.23902333e+00 4.78236675e-01 -4.88960654e-01 -1.19575039e-01
7.16202617e-01 7.03072906e-01 6.55284971e-02 3.28978384e-03
-1.08625579e+00 5.57240963e-01 2.05466047e-01 1.62732735e-01
-1.14927864e+00 -1.10969532e+00 -5.17821312e-01 3.49952698e-01
1.17177892e+00 -2.65311062e-01 1.40551233e+00 -5.61989367e-01
-1.66039228e+00 1.63707942e-01 1.48538619e-01 -6.50134921e-01
7.99292028e-01 1.14339627e-01 -6.99376643e-01 -3.17198001e-02
3.33740301e-02 3.85780334e-01 7.55724013e-01 -1.31909370e+00
-8.38371038e-01 1.64776683e-01 4.58212078e-01 -4.69194055e-01
-4.30171043e-02 -5.98118939e-02 -1.00558467e-01 -2.00618848e-01
-1.67402104e-01 -8.53219390e-01 -7.70570338e-01 2.79191345e-01
-5.64105809e-01 -2.71107852e-01 7.85040736e-01 3.10327291e-01
1.31794381e+00 -1.61036527e+00 -2.90702283e-01 7.45979071e-01
2.95682669e-01 1.70735016e-01 -4.10604239e-01 6.58337831e-01
-6.69987574e-02 5.51708162e-01 9.85386446e-02 -1.01707399e-01
4.18191731e-01 4.45183337e-01 -4.32137221e-01 4.39993262e-01
2.12278500e-01 8.03834438e-01 -1.11347425e+00 -3.33498806e-01
2.26557776e-01 -1.27081424e-01 -8.30181658e-01 -7.45947212e-02
-7.20250309e-01 3.61500531e-01 -4.91573691e-01 5.05395055e-01
8.10882568e-01 -4.60947543e-01 6.22255385e-01 1.16885370e-02
5.71906380e-02 1.53994277e-01 -1.58255625e+00 1.14893138e+00
-1.00931454e+00 7.94769764e-01 4.81509626e-01 -1.38722408e+00
5.78064024e-01 3.29911828e-01 9.10052061e-01 -6.56131208e-01
1.27463013e-01 4.31400359e-01 1.47169605e-01 -4.18827683e-01
-6.49562776e-02 -9.35140252e-02 2.39739530e-02 8.46245885e-01
-2.02519819e-01 5.13416938e-02 3.59572560e-01 1.36457637e-01
1.21088243e+00 -4.84119266e-01 -2.36214906e-01 -4.29891467e-01
5.89837670e-01 -4.34319168e-01 8.16439390e-01 9.48924124e-01
-6.38733923e-01 -1.12715662e-02 1.41564667e+00 -6.69506609e-01
-9.79659379e-01 -1.43478060e+00 -8.08153581e-03 8.67125988e-01
6.37778267e-02 2.02702925e-01 -2.74721891e-01 -6.84013128e-01
2.52142966e-01 4.72907603e-01 -1.43415287e-01 3.67546082e-01
-7.30574608e-01 -7.08175600e-01 3.08414340e-01 2.19328865e-01
2.08259180e-01 -9.97926295e-01 -1.71869725e-01 8.65949273e-01
1.54727940e-02 -1.69774342e+00 -4.71056014e-01 3.75490785e-02
-5.44139445e-01 -9.60067391e-01 -2.89631635e-01 -5.00560880e-01
2.52582282e-01 1.38198659e-01 1.54216361e+00 -2.04918422e-02
-5.87346733e-01 4.45295095e-01 3.75224084e-01 5.31225465e-02
-4.85184222e-01 5.83830953e-01 2.02699572e-01 3.00068021e-01
4.30921046e-03 -9.74591672e-01 -5.23522317e-01 4.53274131e-01
-7.34524310e-01 -7.43796587e-01 6.04276180e-01 4.65240300e-01
2.16034785e-01 6.93412304e-01 1.24917507e+00 -1.08031380e+00
6.23987973e-01 -1.07070220e+00 -1.22353506e+00 8.54617879e-02
-6.29286408e-01 1.53683379e-01 7.83803165e-01 -1.31739900e-01
-7.19915986e-01 -4.09540534e-01 -1.59139976e-01 -9.15152803e-02
-8.22229981e-02 1.62137747e-01 -3.09568495e-01 -2.73635268e-01
8.30163956e-02 -2.56882221e-01 -6.70657083e-02 -1.06768921e-01
3.03980321e-01 4.71434355e-01 1.88284030e-03 -1.19589758e+00
1.05302334e+00 5.58840752e-01 7.12163627e-01 -6.64836764e-01
-5.88492990e-01 1.65420175e-02 -2.26119697e-01 -1.78081349e-01
1.10995106e-01 -3.83662730e-01 -1.48111296e+00 -1.34850517e-01
-9.59008932e-01 -4.05213416e-01 -1.90512687e-01 3.41609299e-01
-1.05496204e+00 2.03368932e-01 -5.39617956e-01 -9.76332188e-01
3.26122075e-01 -1.29646945e+00 4.31480110e-01 3.48675400e-02
2.67626107e-01 -1.28196216e+00 -1.27462566e-01 2.21086279e-01
1.12779260e+00 1.78392142e-01 1.31877887e+00 -6.07955873e-01
-1.30354238e+00 6.25705644e-02 -8.15106988e-01 2.40924135e-01
-5.10961749e-02 2.45779932e-01 -5.09976566e-01 -3.43434304e-01
-4.21435446e-01 -2.90546834e-01 4.44849044e-01 5.21130204e-01
1.57452226e+00 -2.50215560e-01 -1.76159829e-01 5.64739525e-01
1.81891370e+00 1.74070984e-01 4.51908439e-01 2.14920044e-01
2.23071605e-01 7.54040062e-01 2.57413089e-01 8.02842438e-01
5.76929748e-01 3.53742450e-01 1.01769948e+00 1.81747988e-01
2.86018580e-01 8.28481168e-02 9.23614278e-02 3.59723896e-01
4.95301306e-01 -9.83112633e-01 -7.30996966e-01 6.04996204e-01
-1.67447662e+00 -1.00433135e+00 2.26684406e-01 1.98906767e+00
4.10528839e-01 9.46352363e-01 2.37916768e-01 7.92343840e-02
6.60877943e-01 3.69033903e-01 -8.60147059e-01 -8.77467692e-01
1.60301700e-01 1.03892289e-01 9.37068701e-01 7.52682447e-01
-8.28873932e-01 7.42924988e-01 7.52058649e+00 7.27998495e-01
-8.55527520e-01 -4.22519505e-01 9.47798729e-01 -1.17147438e-01
-3.22981983e-01 -7.63616413e-02 -6.08938456e-01 6.38767660e-01
1.41215694e+00 -5.87589085e-01 9.58143950e-01 7.38336742e-01
4.57720160e-01 2.90777266e-01 -1.36129856e+00 6.81007922e-01
-6.00827396e-01 -1.49853277e+00 -7.04455599e-02 2.87760347e-01
8.78403783e-01 1.75963134e-01 4.10385460e-01 3.91716629e-01
7.93826699e-01 -1.05722690e+00 5.17992710e-04 2.29949206e-01
6.67916834e-01 -9.72728491e-01 6.16479039e-01 -2.67399084e-02
-1.04576206e+00 -5.85264444e-01 -1.01235742e-02 2.00261489e-01
7.74439871e-01 6.83958471e-01 -6.61602616e-01 1.21199384e-01
2.13259622e-01 4.80539322e-01 1.25753745e-01 1.39418399e+00
2.89517492e-01 5.26535988e-01 -6.38561189e-01 1.02676474e-01
6.17496789e-01 1.29524663e-01 6.79306269e-01 1.13136554e+00
-1.08511634e-01 -3.59463781e-01 5.59002399e-01 7.69635558e-01
-7.15331018e-01 -2.52122939e-01 -3.93075019e-01 -7.23802969e-02
7.34929740e-01 1.14941108e+00 -6.95544243e-01 -1.72611758e-01
-4.23348367e-01 2.24460244e-01 -1.35174006e-01 7.84027934e-01
-9.53476012e-01 -6.36516452e-01 1.33016884e+00 8.18095803e-02
5.97278178e-01 -3.27190638e-01 -2.67323375e-01 -7.62757003e-01
1.33040369e-01 -7.54192531e-01 2.23246247e-01 -2.08479047e-01
-1.46418822e+00 5.54025412e-01 -1.82433724e-01 -1.00836134e+00
-1.28311932e-01 -7.23923683e-01 -5.45265317e-01 5.29269457e-01
-2.26795959e+00 -3.07066917e-01 2.87852138e-01 3.30499321e-01
7.18154967e-01 -1.17628507e-01 2.82358766e-01 7.83242166e-01
-6.96108580e-01 6.41864240e-01 3.40064496e-01 -9.56523195e-02
2.96781391e-01 -1.27013421e+00 5.39268792e-01 6.29836261e-01
-8.21264088e-02 2.44245995e-02 8.09817851e-01 -1.42119691e-01
-1.53505290e+00 -8.47409606e-01 5.30600488e-01 -2.41734967e-01
1.35969901e+00 -1.45726696e-01 -3.25613588e-01 7.73164034e-01
-7.23337196e-03 2.71508753e-01 5.55524647e-01 -1.29751787e-01
-1.65314794e-01 -6.38621569e-01 -1.38374829e+00 5.94645858e-01
1.11595833e+00 -1.52144864e-01 1.65513635e-01 5.79085946e-01
9.21594024e-01 -7.54247010e-02 -8.97881269e-01 2.25300357e-01
5.90703130e-01 -7.89519012e-01 9.08469200e-01 -1.08438349e+00
-1.47705063e-01 5.59759811e-02 -4.26888704e-01 -1.26878572e+00
-8.61080810e-02 -1.45434904e+00 -2.15814754e-01 9.99196768e-01
6.53946340e-01 -8.19687247e-01 1.28943014e+00 5.37584424e-01
3.07044744e-01 -8.73677015e-01 -1.23227096e+00 -1.13905811e+00
2.11222857e-01 -6.46045327e-01 1.02486527e+00 5.57756603e-01
-5.07756591e-01 -1.11793853e-01 -4.93026495e-01 4.63044405e-01
1.13426161e+00 1.06693164e-01 6.29341483e-01 -1.40080595e+00
-7.64956651e-03 -8.24630737e-01 -4.00182664e-01 -1.46738553e+00
5.62663794e-01 -4.98966664e-01 -1.22842588e-01 -1.19760299e+00
-5.60252249e-01 -1.04271770e+00 -7.13491678e-01 -2.68903404e-01
6.92651093e-01 -6.27902523e-02 1.03497945e-01 -3.85747850e-01
-1.11222029e+00 3.72122705e-01 1.09557629e+00 -1.58554502e-02
4.48818058e-02 6.79854393e-01 -8.21443617e-01 3.18387985e-01
1.20741618e+00 -5.29432416e-01 -6.24142528e-01 -7.16815591e-01
4.33297187e-01 5.17173529e-01 -5.11094294e-02 -7.96884716e-01
3.82439464e-01 -7.54727542e-01 -2.17405200e-01 -5.61806977e-01
1.04278646e-01 -1.37974930e+00 -1.71149492e-01 4.57151771e-01
-4.37705815e-01 2.89404213e-01 9.88398194e-02 8.77520800e-01
3.30744416e-01 1.62406027e-01 6.95454955e-01 5.38615547e-02
-6.57145023e-01 1.07237828e+00 -6.44078434e-01 7.44836569e-01
1.03887081e+00 1.39670298e-01 -2.97064245e-01 -7.57518113e-01
-9.41610754e-01 1.02658188e+00 -1.26291603e-01 4.89315182e-01
3.09746474e-01 -1.09068811e+00 -7.03460693e-01 2.27394357e-01
-4.00725305e-01 -5.57218909e-01 5.55875152e-02 6.82279706e-01
-6.26133263e-01 8.02424729e-01 -1.59881070e-01 -4.18723375e-01
-4.04926181e-01 8.31898570e-01 4.33992505e-01 -4.93433148e-01
-7.37138912e-02 5.87023854e-01 -4.13341880e-01 -2.85189509e-01
7.68093646e-01 -7.97378272e-02 3.55004340e-01 -2.25869030e-01
2.65555948e-01 9.26535249e-01 -2.71522850e-01 -8.29428881e-02
-4.20691550e-01 3.23813766e-01 -1.97856918e-01 -1.35995597e-01
1.40574408e+00 -7.15222418e-01 -2.31610209e-01 1.73355520e-01
1.34394395e+00 -1.39635012e-01 -1.08302033e+00 -6.28800988e-01
4.27206337e-01 -6.99115455e-01 3.13996784e-02 -7.51744568e-01
-1.60443735e+00 9.80289340e-01 4.00045604e-01 8.45320702e-01
9.38387811e-01 -3.63474965e-01 1.10956788e+00 9.02725160e-01
8.11934412e-01 -1.04396677e+00 -6.95203468e-02 3.84305656e-01
1.31848589e-01 -1.31366372e+00 -2.96652555e-01 -4.03040379e-01
-2.02641740e-01 1.33559930e+00 4.96482462e-01 -1.80040032e-01
1.16025424e+00 4.24512446e-01 6.56823888e-02 1.03042111e-01
-1.43936634e+00 -6.98089525e-02 -5.78444958e-01 4.76341933e-01
-5.59237562e-02 1.18538566e-01 2.09287778e-01 -2.61376977e-01
1.84929952e-01 -7.51906782e-02 9.27994192e-01 6.70302987e-01
-4.84046131e-01 -1.36686599e+00 1.37807906e-01 5.72580278e-01
-7.02588558e-01 -1.27482682e-03 3.75034720e-01 6.24083400e-01
-4.00886327e-01 1.15208054e+00 1.16919778e-01 1.03763165e-02
7.93391988e-02 -3.99810314e-01 3.53481360e-02 -3.72763366e-01
7.32386783e-02 -3.28514963e-01 2.26849735e-01 -1.03418410e+00
-2.93509401e-02 -3.26465160e-01 -6.15305007e-01 -9.38328505e-01
-9.63211805e-02 3.80125523e-01 7.56720722e-01 8.31372857e-01
4.47955221e-01 6.24111056e-01 1.52945018e+00 -3.37112039e-01
-8.65990937e-01 -1.06733805e-02 -5.03363550e-01 -2.90830672e-01
9.88094926e-01 -6.18927360e-01 -7.02248991e-01 -6.13981187e-01]
|
[5.6350321769714355, 1.7051268815994263]
|
4115c9b7-941e-47ce-acd8-54a3f1ceac83
|
pixcue-joint-uncertainty-estimation-and-image
|
2303.00111
| null |
https://arxiv.org/abs/2303.00111v2
|
https://arxiv.org/pdf/2303.00111v2.pdf
|
PixCUE: Joint Uncertainty Estimation and Image Reconstruction in MRI using Deep Pixel Classification
|
Deep learning (DL) models are capable of successfully exploiting latent representations in MR data and have become state-of-the-art for accelerated MRI reconstruction. However, undersampling the measurements in k-space as well as the over- or under-parameterized and non-transparent nature of DL make these models exposed to uncertainty. Consequently, uncertainty estimation has become a major issue in DL MRI reconstruction. To estimate uncertainty, Monte Carlo (MC) inference techniques have become a common practice where multiple reconstructions are utilized to compute the variance in reconstruction as a measurement of uncertainty. However, these methods demand high computational costs as they require multiple inferences through the DL model. To this end, we introduce a method to estimate uncertainty during MRI reconstruction using a pixel classification framework. The proposed method, PixCUE (stands for Pixel Classification Uncertainty Estimation) produces the reconstructed image along with an uncertainty map during a single forward pass through the DL model. We demonstrate that this approach generates uncertainty maps that highly correlate with the reconstruction errors with respect to various MR imaging sequences and under numerous adversarial conditions. We also show that the estimated uncertainties are correlated to that of the conventional MC method. We further provide an empirical relationship between the uncertainty estimations using PixCUE and well-established reconstruction metrics such as NMSE, PSNR, and SSIM. We conclude that PixCUE is capable of reliably estimating the uncertainty in MRI reconstruction with a minimum additional computational cost.
|
['Zhaolin Chen', 'Gary Egan', 'Kamlesh Pawar', 'Mevan Ekanayake']
|
2023-02-28
| null | null | null | null |
['mri-reconstruction']
|
['computer-vision']
|
[ 2.93523848e-01 -4.98206429e-02 2.36668438e-01 -3.89556587e-01
-1.43529439e+00 -3.11972171e-01 5.56981027e-01 1.09489024e-01
-6.26290143e-01 1.08064950e+00 1.06080964e-01 -1.01947144e-01
-1.79872841e-01 -7.04638481e-01 -1.04121220e+00 -1.05368316e+00
-1.50197089e-01 2.14916468e-01 -2.09637657e-02 5.97651124e-01
5.70780262e-02 4.00748342e-01 -9.54399467e-01 -6.75069122e-03
1.05912936e+00 1.26794541e+00 1.32427141e-01 4.23988819e-01
1.57449678e-01 7.05607057e-01 -5.79822540e-01 -3.84644806e-01
9.13130194e-02 -3.41731399e-01 -5.85388422e-01 9.22208186e-03
3.73185202e-02 -7.59077489e-01 -3.66286010e-01 1.43507504e+00
6.91651821e-01 3.39747183e-02 9.35780108e-01 -7.57842362e-01
-1.03549927e-01 9.47192013e-01 -5.33857286e-01 2.50622571e-01
2.03293245e-02 1.05807940e-02 3.26946288e-01 -9.62902486e-01
4.30031210e-01 7.36447930e-01 7.33486235e-01 2.29739845e-01
-1.32232153e+00 -4.23074424e-01 -4.05587584e-01 2.81339198e-01
-1.49811029e+00 -3.70389074e-01 6.24263406e-01 -5.43026805e-01
3.14230591e-01 2.37712964e-01 2.55079627e-01 1.06844223e+00
9.16405082e-01 7.45503902e-01 1.67377222e+00 -3.11359465e-01
6.46925211e-01 8.41804594e-02 -1.29167065e-01 4.93158311e-01
2.93753713e-01 4.29363072e-01 -3.12257141e-01 -2.42647931e-01
9.90758538e-01 -1.81340709e-01 -5.37974477e-01 -3.75987262e-01
-1.22784221e+00 7.48750925e-01 4.78896230e-01 4.50890362e-02
-6.20117605e-01 6.93412066e-01 4.32309031e-01 -1.22698031e-01
5.09870529e-01 2.33368799e-01 9.63029172e-03 -1.57087505e-01
-1.34108555e+00 1.04801752e-01 3.81919622e-01 5.36255717e-01
2.69146442e-01 2.57291615e-01 -2.25595385e-01 5.84267557e-01
3.06393802e-01 5.82830906e-01 4.81974453e-01 -1.04910493e+00
6.65058717e-02 -4.00790244e-01 8.16266686e-02 -8.44002545e-01
-2.49263868e-01 -8.88800025e-01 -9.93179440e-01 4.66062009e-01
4.30884987e-01 -1.34866267e-01 -8.60429108e-01 1.86010587e+00
3.22258621e-01 4.27479506e-01 3.16496268e-02 8.10167015e-01
4.39048648e-01 4.42383200e-01 3.44530903e-02 -5.93467891e-01
1.13461578e+00 -6.30868793e-01 -9.68795240e-01 2.04907469e-02
1.29389659e-01 -6.73572481e-01 7.20202744e-01 5.27728200e-01
-1.28665996e+00 -2.94455081e-01 -1.28693163e+00 4.56372976e-01
3.14759314e-01 -4.40686382e-02 3.19417626e-01 9.18827057e-01
-7.28154778e-01 8.95893633e-01 -1.28270721e+00 4.52040493e-01
5.41397393e-01 4.94141281e-02 -1.45603016e-01 -2.32489318e-01
-1.28695977e+00 1.14807332e+00 2.12979585e-01 3.21850121e-01
-1.41197550e+00 -8.52069795e-01 -8.44783783e-01 -2.30590627e-01
3.79569650e-01 -4.62735802e-01 1.13515365e+00 -3.19638968e-01
-1.58033705e+00 3.68301541e-01 1.91068396e-01 -6.24818087e-01
1.04870582e+00 -2.92310387e-01 -3.40300947e-01 4.69382226e-01
-6.93409517e-02 2.96531469e-01 1.21819425e+00 -1.28242052e+00
2.20253929e-01 -2.08854541e-01 -1.60202101e-01 7.69781843e-02
3.99394602e-01 -2.37247735e-01 -1.34717688e-01 -7.37793267e-01
6.18024230e-01 -8.25094223e-01 -3.96540135e-01 1.27793059e-01
-4.00883377e-01 7.51866400e-01 5.54850027e-02 -9.83801782e-01
1.06774306e+00 -1.85168588e+00 6.41862527e-02 3.46433580e-01
3.48161012e-01 -2.65671164e-01 3.80540103e-01 -1.19054196e-02
9.60227251e-02 -8.98475796e-02 -7.88713753e-01 -2.80804306e-01
-8.96941870e-02 1.79425627e-01 -1.92219794e-01 9.19512033e-01
-1.09307550e-01 8.37752044e-01 -9.78687227e-01 -6.46185040e-01
4.81810987e-01 7.43992865e-01 -3.27753454e-01 8.94888118e-02
1.29834130e-01 9.75948632e-01 -2.78882176e-01 3.84219199e-01
8.90934646e-01 -2.05151409e-01 1.05414212e-01 -6.30316913e-01
1.25506550e-01 -1.10234618e-01 -1.13428771e+00 1.81806743e+00
-6.89806938e-01 3.38511854e-01 -5.49407452e-02 -7.90920138e-01
3.83985192e-01 4.39243197e-01 4.45182979e-01 -2.37962782e-01
3.03817391e-01 5.07694721e-01 -1.15454882e-01 -2.78728247e-01
2.63887286e-01 -4.86463010e-01 -7.51757473e-02 5.29370248e-01
1.61337648e-02 -5.59741795e-01 -2.29315802e-01 1.25732481e-01
9.76978660e-01 5.99178486e-02 3.80133122e-01 -3.83445859e-01
4.27108735e-01 -4.54567552e-01 3.78570169e-01 9.56841290e-01
-4.56590772e-01 7.04081714e-01 2.61275619e-01 -1.79273412e-02
-1.08599198e+00 -1.52553964e+00 -8.33818495e-01 -1.42144099e-01
-4.55706939e-02 2.47350886e-01 -9.62778032e-01 -4.45736706e-01
-2.19399586e-01 8.39697480e-01 -4.53516334e-01 -1.80284038e-01
-3.60906243e-01 -1.06145453e+00 5.13454080e-01 5.00710011e-01
5.16995490e-01 -5.20647585e-01 -7.66276300e-01 3.16653073e-01
-3.72472167e-01 -1.32796478e+00 -3.16458255e-01 1.32654890e-01
-1.09299684e+00 -7.92304814e-01 -9.41749394e-01 1.29761189e-01
9.05600488e-01 -4.48633045e-01 9.86641169e-01 -3.83694798e-01
-4.11027104e-01 4.80529666e-01 -1.64138556e-01 7.47017488e-02
-7.28267074e-01 -5.98043382e-01 3.25234145e-01 -1.08683400e-01
-3.23614955e-01 -6.60842717e-01 -8.40383053e-01 2.05206111e-01
-1.21114278e+00 2.71878485e-02 6.61047161e-01 1.02440751e+00
1.07296562e+00 3.78179491e-01 5.21217406e-01 -7.96910465e-01
6.00744009e-01 -4.27196145e-01 -7.60697663e-01 2.03741819e-01
-7.58709252e-01 4.73231703e-01 1.46091729e-01 -4.73405153e-01
-1.09943116e+00 -1.70468524e-01 -2.91198701e-01 -5.07168293e-01
2.47451991e-01 7.74401605e-01 5.85779659e-02 -3.67656678e-01
6.55613542e-01 2.91101694e-01 6.72737882e-02 -1.79682791e-01
4.02634293e-01 5.03547311e-01 7.27502644e-01 -7.12819219e-01
4.94448304e-01 5.93661129e-01 2.54296988e-01 -4.83758986e-01
-6.01470828e-01 1.15022525e-01 -3.26046228e-01 -5.61503291e-01
8.65723848e-01 -8.76819611e-01 -4.86923844e-01 6.60830379e-01
-8.80943358e-01 -2.55850311e-02 -3.69976729e-01 1.05422711e+00
-8.09639335e-01 8.06004047e-01 -8.51726830e-01 -9.23600495e-01
-4.28388923e-01 -1.88976085e+00 7.32476294e-01 -1.66741371e-01
3.98757122e-02 -8.58101666e-01 -2.58533448e-01 2.71991402e-01
5.45768559e-01 6.58715069e-01 8.54707539e-01 -7.26162717e-02
-7.21455038e-01 -2.50983953e-01 -1.75255865e-01 6.82265282e-01
-8.71923938e-02 -5.23183584e-01 -1.03794193e+00 -2.97796875e-01
7.23077297e-01 -1.57620400e-01 5.96146762e-01 1.00108349e+00
1.41940629e+00 1.03020342e-02 5.66229783e-02 5.64574063e-01
1.70230114e+00 -1.60202049e-02 7.64072537e-01 -3.32352147e-02
2.19015568e-01 2.03853503e-01 4.15040433e-01 5.31267524e-01
-1.90008521e-01 6.34711683e-01 4.33593690e-01 4.18363541e-01
-1.08901948e-01 -7.33708516e-02 2.06125319e-01 1.11250436e+00
1.27078161e-01 3.05545924e-04 -8.08971584e-01 3.26800257e-01
-1.47851515e+00 -6.67842269e-01 1.36229638e-02 2.46873879e+00
1.06846035e+00 2.18975231e-01 -6.06862128e-01 3.04441661e-01
6.09936833e-01 1.14210986e-01 -7.20384598e-01 -4.57845181e-02
8.65443125e-02 2.37795666e-01 8.81845415e-01 7.17133045e-01
-8.69909346e-01 2.84675628e-01 6.85050917e+00 1.13598251e+00
-9.02412474e-01 5.82685709e-01 8.48443747e-01 -5.24409348e-03
-5.14122784e-01 -2.64722049e-01 -2.33367756e-01 7.11695433e-01
9.86278832e-01 -2.93623526e-02 4.15235341e-01 6.52154028e-01
9.45694447e-02 -6.46085918e-01 -9.11410451e-01 1.14577675e+00
-2.50177924e-02 -1.29773307e+00 -3.09631974e-01 -3.16636600e-02
9.16603327e-01 -1.35717198e-01 2.50098377e-01 -8.85832906e-02
1.59815904e-02 -1.02716446e+00 7.81409740e-01 8.50889564e-01
1.13958514e+00 -7.62627661e-01 8.90424550e-01 2.79449940e-01
-5.70364892e-01 3.70493591e-01 -3.93761277e-01 4.65532809e-01
7.45786786e-01 1.41767967e+00 -6.94583595e-01 5.60632765e-01
5.06577075e-01 -3.99255119e-02 3.65021303e-02 1.05349791e+00
-5.45954764e-01 4.56831813e-01 -3.49193066e-01 3.94767940e-01
-2.49536112e-02 -1.60628691e-01 7.13301659e-01 8.11578929e-01
5.18225491e-01 -1.04438901e-01 -2.73255408e-01 1.13721168e+00
2.86349989e-02 -2.14476079e-01 -1.12504609e-01 2.02248599e-02
3.91355515e-01 8.76102984e-01 -7.22411811e-01 -2.70865381e-01
-2.94772126e-02 1.09298944e+00 -1.12139441e-01 2.55406320e-01
-1.03640449e+00 3.84983793e-02 3.65727037e-01 -5.82691282e-02
-6.81758597e-02 -3.47533286e-01 -3.94058168e-01 -1.10693812e+00
-1.57188345e-02 -7.26491213e-01 -2.07955763e-02 -8.48546326e-01
-1.26723158e+00 8.47301483e-01 3.99651736e-01 -1.21882176e+00
-5.08608401e-01 -4.68725950e-01 -7.70453066e-02 9.70879614e-01
-1.53960884e+00 -6.18984759e-01 -1.43313512e-01 3.90131265e-01
2.98084557e-01 1.00374766e-01 7.83899248e-01 4.25622463e-01
-2.87869364e-01 4.81385022e-01 4.81604308e-01 -1.53584674e-01
3.41192186e-01 -1.11137605e+00 1.40840188e-02 1.04475224e+00
-3.68782841e-02 5.25900960e-01 1.04334247e+00 -8.00368547e-01
-1.10229957e+00 -6.93393528e-01 9.25532281e-02 -1.82448566e-01
6.65522814e-01 -2.09553894e-02 -6.92030191e-01 4.46586370e-01
-2.41516426e-01 4.68080372e-01 5.71754694e-01 -4.82820749e-01
2.48727445e-02 8.59975144e-02 -1.63136518e+00 3.73670578e-01
5.11395454e-01 -8.31205547e-01 -5.28386712e-01 1.38616309e-01
6.93059981e-01 -9.17694330e-01 -1.19717860e+00 5.90954006e-01
6.34444535e-01 -9.03682709e-01 1.02778268e+00 1.06289066e-01
5.24763823e-01 -2.87364215e-01 -4.46101308e-01 -1.38495767e+00
2.64399558e-01 -2.45753437e-01 -3.09209704e-01 7.30601668e-01
3.30537021e-01 -5.48925817e-01 5.95592320e-01 7.73185968e-01
-9.10401791e-02 -7.44556248e-01 -1.40046144e+00 -8.02928448e-01
1.34934366e-01 -1.02697432e+00 3.75573963e-01 6.22081697e-01
-2.70654052e-01 -5.97305298e-01 -5.70131183e-01 4.75363463e-01
1.37953150e+00 -2.89016157e-01 1.72940120e-02 -7.06365764e-01
-6.72862351e-01 -3.45459655e-02 -5.51868796e-01 -8.08798313e-01
-6.12359829e-02 -7.38236368e-01 2.30307043e-01 -1.24239004e+00
1.87300056e-01 -5.99715531e-01 -4.78685170e-01 -3.46697152e-01
-8.10844302e-02 2.84470499e-01 -1.24601617e-01 3.15944761e-01
3.11839599e-02 5.71788490e-01 1.29782152e+00 -4.58488204e-02
2.94943005e-01 9.90091637e-02 -8.08001533e-02 6.55640304e-01
4.91878897e-01 -6.85156882e-01 -4.46107805e-01 -3.37518156e-01
3.42956275e-01 4.99101281e-01 4.59013611e-01 -1.30639410e+00
1.24275289e-01 2.70324379e-01 4.87826586e-01 -6.10455155e-01
4.03881431e-01 -9.84029710e-01 6.45060718e-01 5.97027600e-01
-3.89960170e-01 -2.64218032e-01 1.92778200e-01 6.77341163e-01
-2.09831044e-01 -6.95271969e-01 1.02851617e+00 -2.07190901e-01
-3.72872472e-01 2.04813495e-01 -2.19587490e-01 -1.83694437e-01
8.09720337e-01 1.22923523e-01 6.46135658e-02 -4.72437412e-01
-1.02707553e+00 -2.98663586e-01 1.25169843e-01 -2.48074561e-01
8.71387601e-01 -1.33009446e+00 -6.26426339e-01 8.40017498e-02
-1.85287908e-01 -1.48037568e-01 7.48661637e-01 1.15580678e+00
-8.24757159e-01 1.64610088e-01 -8.71419311e-02 -8.42301369e-01
-6.68333888e-01 4.54024732e-01 5.08327901e-01 -5.20480514e-01
-6.79232955e-01 8.51184785e-01 8.86885896e-02 -1.57817066e-01
6.58366829e-02 -4.02002901e-01 3.00278038e-01 -3.03909153e-01
6.31544948e-01 4.19645786e-01 1.72183633e-01 -3.42851788e-01
-1.92561060e-01 3.44077796e-01 1.60944134e-01 -6.24102056e-01
1.01848471e+00 -3.96597683e-01 7.24401996e-02 5.47625542e-01
1.09146333e+00 2.82913987e-02 -1.46285748e+00 -3.22764844e-01
-2.04820096e-01 -5.17806768e-01 6.71423674e-01 -9.43689942e-01
-1.11385918e+00 9.90681767e-01 1.05918825e+00 -2.90515542e-01
9.64716792e-01 -2.45578632e-01 7.08379149e-01 -2.27888837e-01
7.83785939e-01 -1.05175781e+00 -1.54670700e-01 -2.23594606e-01
9.85353708e-01 -1.29235971e+00 4.14466321e-01 -3.87935430e-01
-6.58602178e-01 8.12477887e-01 -3.43917236e-02 5.16126342e-02
8.23727190e-01 5.71960866e-01 -2.80983672e-02 4.58253771e-02
-1.46772042e-01 2.83012301e-01 2.38814965e-01 4.32891935e-01
2.01399565e-01 3.90942663e-01 -5.62006891e-01 3.99816513e-01
6.74112216e-02 1.93819836e-01 6.83079779e-01 7.95793772e-01
-1.08223945e-01 -8.82593751e-01 -5.83181202e-01 4.00272280e-01
-8.09579372e-01 -2.39043877e-01 8.31400275e-01 4.07301366e-01
-2.62246877e-01 7.26270974e-01 -2.72718638e-01 -1.00387976e-01
-1.28831431e-01 -1.43116310e-01 8.18542242e-01 -1.47996962e-01
-1.39302745e-01 8.57479423e-02 -4.80662938e-03 -6.24924958e-01
-3.34964365e-01 -5.68338275e-01 -1.27614534e+00 -1.87487096e-01
-4.78696823e-01 6.95876256e-02 1.19866300e+00 1.00884414e+00
-1.03405781e-01 7.35100567e-01 4.96229261e-01 -7.25622833e-01
-1.11440396e+00 -8.22543681e-01 -8.51109445e-01 1.34370610e-01
1.31436154e-01 -8.38798344e-01 -6.14867508e-01 -3.40644330e-01]
|
[13.531387329101562, -2.341578483581543]
|
3cc38e25-0d0a-46d6-b49f-f79d689bc74c
|
an-image-processing-based-object-counting
|
1802.05911
| null |
http://arxiv.org/abs/1802.05911v1
|
http://arxiv.org/pdf/1802.05911v1.pdf
|
An Image Processing based Object Counting Approach for Machine Vision Application
|
Machine vision applications are low cost and high precision measurement
systems which are frequently used in production lines. With these systems that
provide contactless control and measurement, production facilities are able to
reach high production numbers without errors. Machine vision operations such as
product counting, error control, dimension measurement can be performed through
a camera. In this paper, a machine vision application is proposed, which can
perform object-independent product counting. The proposed approach is based on
Otsu thresholding and Hough transformation and performs automatic counting
independently of product type and color. Basically one camera is used in the
system. Through this camera, an image of the products passing through a
conveyor is taken and various image processing algorithms are applied to these
images. In this approach using images obtained from a real experimental setup,
a real-time machine vision application was installed. As a result of the
experimental studies performed, it has been determined that the proposed
approach gives fast, accurate and reliable results.
|
['Erhan Akin', 'Alisan Sarimaden', 'Mehmet Karakose', 'Mehmet Baygin']
|
2018-02-16
| null | null | null | null |
['object-counting']
|
['computer-vision']
|
[ 3.48815709e-01 -5.04737556e-01 4.92920205e-02 -5.93783110e-02
2.18787834e-01 -5.58964610e-01 4.19386417e-01 5.06329656e-01
-5.80290556e-01 3.07161450e-01 -8.97869408e-01 -2.49881044e-01
1.62906677e-01 -9.39843237e-01 -2.31656492e-01 -4.73042995e-01
5.13265491e-01 6.79531097e-01 1.97904497e-01 7.73830665e-03
6.08570755e-01 8.85572672e-01 -1.61788201e+00 -2.42225528e-01
4.78331327e-01 1.18237662e+00 5.09004176e-01 1.05644941e+00
-8.84109512e-02 4.74460393e-01 -6.13994718e-01 -9.88293886e-02
4.66146290e-01 -4.67749238e-01 -3.40859085e-01 8.66055310e-01
-5.34590185e-02 -5.20675600e-01 4.83948231e-01 1.09690404e+00
-7.02631101e-02 5.90020828e-02 9.32200789e-01 -9.19957995e-01
-3.31990987e-01 -3.23121101e-02 -8.45693886e-01 -1.25051692e-01
4.01618749e-01 8.57535452e-02 5.89612484e-01 -8.79855037e-01
2.55318284e-01 9.99316871e-01 3.17731857e-01 -1.80217803e-01
-1.32498908e+00 -3.34279627e-01 -5.05888045e-01 1.40018001e-01
-1.29082584e+00 -3.64620835e-02 6.85040534e-01 -5.09395123e-01
6.95933640e-01 1.21024333e-01 8.05073857e-01 -5.76843582e-02
5.93479395e-01 3.10776383e-01 1.21536839e+00 -1.11704004e+00
3.45413536e-01 6.55818820e-01 3.51686805e-01 7.19279170e-01
7.21804142e-01 -4.72570397e-02 2.89450258e-01 3.69520009e-01
1.03648531e+00 3.54961604e-01 -8.74376372e-02 -4.92257237e-01
-8.60588610e-01 7.94983208e-01 2.05052823e-01 7.70967007e-01
-5.26535034e-01 -1.20354339e-01 3.05436224e-01 2.76907571e-02
3.90980247e-04 2.62740970e-01 -2.35666528e-01 -1.43589243e-01
-8.86748374e-01 -2.17051864e-01 1.05320144e+00 9.11538661e-01
4.79991257e-01 -1.78880230e-01 3.63879025e-01 5.35216510e-01
3.49209577e-01 8.26859653e-01 2.82386482e-01 -5.92947602e-01
1.71684965e-01 7.41266727e-01 2.87098348e-01 -1.24132001e+00
-3.75958979e-01 5.95503263e-02 -6.69652700e-01 9.16106939e-01
4.73294884e-01 -8.24367802e-04 -8.90483975e-01 4.59665507e-01
3.74934137e-01 -2.73379207e-01 -9.04810615e-03 8.65474999e-01
2.84319758e-01 8.24321508e-01 -1.06235132e-01 -3.73369932e-01
1.57569611e+00 -7.16836631e-01 -9.66092825e-01 2.31725484e-01
4.76667643e-01 -1.44586849e+00 8.92101645e-01 1.07728148e+00
-7.90096045e-01 -8.31898332e-01 -1.57825732e+00 9.82420221e-02
-6.24919534e-01 8.20048690e-01 4.48809475e-01 6.88901246e-01
-2.35984638e-01 5.07262170e-01 -7.17006981e-01 -5.48790574e-01
-2.40977600e-01 2.28300840e-01 -4.51564819e-01 -6.40472621e-02
-3.71315747e-01 1.18193352e+00 6.73558116e-01 5.97962558e-01
-3.73388886e-01 1.17961150e-02 -7.01978266e-01 3.88532467e-02
2.38488004e-01 -2.62624174e-01 1.35866475e+00 -8.07083786e-01
-1.84560394e+00 8.05913448e-01 2.27030054e-01 -2.33977333e-01
4.06781733e-01 -2.17430755e-01 -2.26126462e-01 3.30703884e-01
-6.85913712e-02 -1.57207921e-01 8.94957602e-01 -1.05574131e+00
-1.01823676e+00 -5.33590734e-01 -3.16605926e-01 -7.54019171e-02
1.04436807e-01 1.83635876e-02 -1.94483906e-01 7.41040483e-02
2.38981232e-01 -8.68522763e-01 4.25162092e-02 -8.91192257e-02
-3.07018459e-01 1.71420410e-01 9.28695738e-01 -4.89018142e-01
6.57135546e-01 -2.00956440e+00 -3.24081212e-01 5.47790766e-01
-1.83491558e-01 5.01780093e-01 5.23833454e-01 2.53815711e-01
2.79464990e-01 -4.82761770e-01 1.37010366e-02 1.09775383e-02
-2.70923644e-01 6.76300079e-02 4.73501086e-01 4.50242996e-01
-1.23913787e-01 2.64591634e-01 -4.12803262e-01 -6.94194853e-01
1.13255668e+00 3.17689896e-01 7.57708251e-02 1.41808301e-01
2.15203583e-01 1.36680409e-01 -3.27844411e-01 6.94692552e-01
8.25355887e-01 1.26635462e-01 1.39296830e-01 -3.79429668e-01
-3.26937079e-01 -6.79111123e-01 -1.46569788e+00 1.01287901e+00
-7.97670782e-01 6.37064338e-01 3.22146356e-01 -1.01099253e+00
1.38515437e+00 3.27661604e-01 2.65182972e-01 -6.07109964e-01
7.94706106e-01 5.25682271e-01 -1.03390858e-01 -5.59349179e-01
6.45619452e-01 -5.56150265e-02 2.06320956e-01 -1.04863442e-01
-3.64456326e-02 -7.75937557e-01 7.12383986e-01 -3.70560855e-01
4.14940089e-01 -6.21326901e-02 7.15530097e-01 -6.53039245e-03
9.93820190e-01 3.02381396e-01 5.71725294e-02 2.89546072e-01
2.28216108e-02 2.75204211e-01 5.21120466e-02 -4.61064309e-01
-1.28373539e+00 -7.88972080e-01 -2.70880401e-01 3.81684780e-01
3.95852506e-01 1.92990765e-01 -7.02631712e-01 5.96130863e-02
-1.66406552e-03 4.72643375e-01 -1.75137550e-01 1.94905147e-01
-2.94501543e-01 -1.74580142e-01 -3.37532550e-01 2.46655226e-01
6.95989668e-01 -9.26523745e-01 -1.11298060e+00 5.84031880e-01
4.79313463e-01 -1.20041299e+00 1.70469746e-01 -3.95213179e-02
-1.15872800e+00 -1.36029601e+00 -7.72980034e-01 -8.68380487e-01
7.70473778e-01 5.42726457e-01 7.38563478e-01 1.19042881e-01
-9.54714656e-01 2.44655669e-01 -4.51956689e-01 -8.30269456e-01
-5.91405630e-01 -1.10566340e-01 -2.00515330e-01 6.85399845e-02
6.84884131e-01 -9.52500850e-02 -4.84469444e-01 3.07091653e-01
-5.64097762e-01 -2.99721718e-01 8.71714175e-01 5.58866918e-01
4.23718393e-01 4.73854542e-01 3.31086330e-02 -6.95978284e-01
8.15018594e-01 2.81888604e-01 -1.32778180e+00 2.20478892e-01
-8.76552105e-01 -2.77075589e-01 8.57629001e-01 -5.02558313e-02
-1.06505394e+00 3.55199933e-01 1.53463453e-01 -2.69369423e-01
-4.95398313e-01 2.80737042e-01 -1.26858931e-02 -9.32937041e-02
4.69739020e-01 1.18456148e-01 4.10141945e-01 -5.27068138e-01
5.76573834e-02 1.03956330e+00 6.33383036e-01 1.90167911e-02
5.08863986e-01 1.22596338e-01 5.14419496e-01 -1.24796236e+00
-2.80575991e-01 -7.34700620e-01 -9.17260587e-01 -4.89226043e-01
1.00119638e+00 -4.18326169e-01 -1.03485656e+00 6.94334686e-01
-1.21669686e+00 3.79336208e-01 -4.31725383e-03 1.03848839e+00
-4.32982922e-01 2.42165774e-01 -7.16244042e-01 -1.22566104e+00
-4.62863117e-01 -1.10015440e+00 7.68766582e-01 5.75974941e-01
-9.85495746e-02 -8.47221673e-01 -3.09215486e-01 2.89469898e-01
8.78806189e-02 2.17484415e-01 5.51689744e-01 -1.95565775e-01
-4.72866684e-01 -9.64618921e-01 -1.40632033e-01 6.69938624e-01
4.86045003e-01 5.94746947e-01 -5.40942192e-01 2.38671135e-02
1.85795814e-01 1.21865496e-01 4.52486128e-01 4.74524200e-01
5.75589895e-01 1.62180841e-01 -2.93640941e-01 7.11235851e-02
1.86364102e+00 7.83177257e-01 6.32536948e-01 3.10804904e-01
5.10451078e-01 4.50666428e-01 1.46328366e+00 4.50945139e-01
-3.22629809e-01 6.46514952e-01 3.79575282e-01 -2.66616583e-01
3.07560712e-01 1.06779046e-01 3.80269135e-03 5.72149158e-01
-4.30808872e-01 -2.62721255e-02 -4.40256536e-01 2.49777108e-01
-1.53875113e+00 -8.47889662e-01 -6.80234194e-01 2.40097213e+00
1.79166660e-01 4.14037287e-01 -2.40055900e-02 9.09666717e-01
8.66197705e-01 -5.58122396e-01 6.48119375e-02 -1.07719314e+00
5.76697707e-01 2.71655828e-01 7.95782745e-01 6.03735805e-01
-1.03195620e+00 4.04216915e-01 5.54434681e+00 5.02848208e-01
-1.35026026e+00 -1.92662865e-01 1.98478416e-01 2.58582503e-01
5.71391582e-01 -2.87162036e-01 -7.01131463e-01 4.76756305e-01
4.32104617e-01 1.06868863e-01 3.32833268e-02 1.04046071e+00
2.94463784e-01 -1.01804221e+00 -9.52234149e-01 1.16995609e+00
-1.06063969e-01 -9.76119101e-01 -1.72757328e-01 1.74763069e-01
4.19727653e-01 -9.71038997e-01 -1.38653323e-01 -2.26096675e-01
-1.57633305e-01 -5.74058354e-01 5.49052298e-01 4.48257357e-01
3.58565718e-01 -9.21489120e-01 1.31046736e+00 2.68131882e-01
-1.15152061e+00 -1.22977339e-01 -4.32843715e-01 -3.45731854e-01
3.46597314e-01 8.20573270e-01 -1.21165907e+00 2.87722647e-01
1.80070564e-01 1.44526273e-01 -1.98131785e-01 1.05418766e+00
-2.48482674e-01 3.03614527e-01 -5.24270713e-01 -4.68405724e-01
1.27100483e-01 -9.23064709e-01 1.26886219e-01 1.15563536e+00
6.27671242e-01 1.08844079e-01 4.41928096e-02 8.81699741e-01
4.65656787e-01 4.63600636e-01 -7.46885836e-01 -7.69770369e-02
2.42812917e-01 1.52953196e+00 -1.42558229e+00 -3.59568685e-01
-3.52288693e-01 1.00729930e+00 -5.67836285e-01 -8.29417184e-02
-7.12688804e-01 -8.62854004e-01 1.36038521e-02 1.66078284e-01
5.36659479e-01 -5.60876429e-01 -3.77748668e-01 -6.07873082e-01
-3.00260745e-02 -2.21088424e-01 -8.92897323e-03 -7.93563426e-01
-5.77394009e-01 1.02058329e-01 -1.96796078e-02 -1.29821229e+00
-4.51317847e-01 -1.26529479e+00 -4.42377448e-01 8.09047341e-01
-7.83522844e-01 -9.91011083e-01 -2.88143069e-01 3.40708315e-01
7.86191583e-01 -2.89771825e-01 6.76487327e-01 1.47218689e-01
-5.84581375e-01 1.23677090e-01 3.36769789e-01 -1.17182977e-01
9.34599712e-02 -1.20976889e+00 -3.52045745e-01 9.80878711e-01
1.38781846e-01 3.99794817e-01 8.97439897e-01 -4.43917155e-01
-1.16958165e+00 -2.93034196e-01 6.28851116e-01 1.71724483e-01
4.22561765e-01 -1.63013339e-01 -4.86615658e-01 3.94807518e-01
1.30584121e-01 5.23991622e-02 4.00892884e-01 -3.41137439e-01
3.96729469e-01 -3.94212902e-01 -1.42793095e+00 -4.11200710e-02
-2.84790322e-02 -1.02269538e-01 -5.41986048e-01 2.48591155e-01
-3.76328453e-02 -2.05590740e-01 -9.26060379e-01 -7.77394250e-02
6.50766969e-01 -1.05934942e+00 6.16514742e-01 1.94093794e-01
2.52568305e-01 -4.62964356e-01 3.54639590e-01 -1.05834115e+00
-1.04638554e-01 -6.84226453e-02 3.33653420e-01 1.29099202e+00
3.77890378e-01 -6.82128131e-01 6.24196231e-01 5.10597885e-01
2.08263874e-01 -2.43546993e-01 -5.04045367e-01 -6.71422482e-01
-6.22774482e-01 1.80918667e-02 1.73877418e-01 3.47245514e-01
2.16940939e-02 4.05834377e-01 2.23335437e-02 1.94901705e-01
6.81955040e-01 8.31755698e-02 7.90515363e-01 -1.55745864e+00
-2.34148189e-01 -1.70153618e-01 -9.33581829e-01 -5.53630888e-01
-4.64401811e-01 -1.20861225e-01 8.13613907e-02 -1.55781758e+00
-1.67006880e-01 -1.22471452e-01 1.15156390e-01 -6.40804246e-02
3.09060425e-01 3.70135874e-01 4.10925657e-01 1.93358302e-01
-8.49254653e-02 -1.73249513e-01 1.21433580e+00 -2.71165464e-02
-1.74730733e-01 4.98121977e-01 1.72774911e-01 9.19571221e-01
9.97998059e-01 -1.28841266e-01 -3.12303334e-01 5.44929914e-02
-1.98276043e-01 -4.04695384e-02 1.02679618e-01 -1.14710987e+00
1.41530737e-01 3.52148190e-02 8.28468502e-01 -7.06566155e-01
3.80670279e-01 -1.50426221e+00 1.39457062e-01 9.27474082e-01
2.95863658e-01 4.63758036e-03 -1.05130620e-01 1.53253764e-01
-2.98428565e-01 -9.89852011e-01 8.25569570e-01 -3.28935087e-01
-8.37496519e-01 -4.60639000e-01 -5.76207578e-01 -9.60261881e-01
1.53542387e+00 -8.36434245e-01 1.53546231e-02 -1.43201992e-01
-6.86997890e-01 -2.11136505e-01 4.33554918e-01 -7.96947256e-02
5.02762556e-01 -1.08678389e+00 -3.40987295e-01 2.69371450e-01
1.25769988e-01 -7.91885182e-02 -1.77418850e-02 6.91967487e-01
-1.46541464e+00 6.45731688e-01 -5.14912963e-01 -6.40561640e-01
-1.62745929e+00 8.97546828e-01 1.09841824e-01 1.09302783e-02
-3.38946849e-01 2.73615211e-01 -5.20630479e-01 1.60963908e-01
-1.62508994e-01 -8.05709898e-01 -3.52751881e-01 6.88984767e-02
5.78713953e-01 7.85729229e-01 3.48490596e-01 -5.67053854e-01
-6.97599053e-02 1.11533332e+00 1.66859608e-02 -3.54125738e-01
8.88326764e-01 -1.55112460e-01 -4.40803207e-02 6.26136005e-01
1.06320500e+00 3.76200750e-02 -9.28891480e-01 2.77351171e-01
-9.16461721e-02 -7.71349192e-01 1.10353902e-01 -5.21066606e-01
-8.49276066e-01 9.93845463e-01 9.51522171e-01 7.46513307e-01
1.20846307e+00 -4.64752883e-01 4.36506331e-01 3.57611686e-01
6.51812553e-01 -1.64208376e+00 -3.46046805e-01 9.65032429e-02
6.34271324e-01 -1.41311038e+00 3.17775577e-01 -7.04394579e-01
-3.69035482e-01 1.66372156e+00 4.89112943e-01 -1.94505617e-01
3.28035414e-01 4.40250486e-01 1.93222985e-02 -1.71832845e-01
-2.85464954e-02 -3.30319047e-01 -2.16844350e-01 6.33118093e-01
4.99971628e-01 2.87188053e-01 -9.18764949e-01 -1.11653596e-01
-5.19325882e-02 5.39598286e-01 8.15627813e-01 1.24784982e+00
-8.73265445e-01 -8.66512895e-01 -1.06093574e+00 3.46873730e-01
-5.36584079e-01 5.87778032e-01 -7.45003521e-02 1.08668399e+00
1.30000770e-01 1.17767203e+00 3.49448830e-01 -3.76772061e-02
5.94970226e-01 -3.38638633e-01 8.40645015e-01 -2.87464321e-01
-3.58977348e-01 -6.78659836e-03 -5.99740781e-02 -2.39049777e-01
-1.90155268e-01 -5.02070308e-01 -1.31455410e+00 -2.15347573e-01
-6.46865249e-01 2.57139802e-01 1.55200577e+00 6.14689350e-01
-1.90859795e-01 3.25817853e-01 7.15712786e-01 -9.09804106e-01
-5.32005847e-01 -1.08544016e+00 -1.21762133e+00 2.75514483e-01
-5.62421940e-02 -7.17815757e-01 -3.01410049e-01 2.79290140e-01]
|
[9.317381858825684, -1.5608540773391724]
|
18ffe972-723e-4339-bc27-8c1b246a86e0
|
what-evidence-does-deep-learning-model-use-to
|
1811.01051
| null |
http://arxiv.org/abs/1811.01051v3
|
http://arxiv.org/pdf/1811.01051v3.pdf
|
What evidence does deep learning model use to classify Skin Lesions?
|
Melanoma is a type of skin cancer with the most rapidly increasing incidence.
Early detection of melanoma using dermoscopy images significantly increases
patients' survival rate. However, accurately classifying skin lesions by eye,
especially in the early stage of melanoma, is extremely challenging for the
dermatologists. Hence, the discovery of reliable biomarkers will be meaningful
for melanoma diagnosis. Recent years, the value of deep learning empowered
computer-assisted diagnose has been shown in biomedical imaging based decision
making. However, much research focuses on improving disease detection accuracy
but not exploring the evidence of pathology. In this paper, we propose a method
to interpret the deep learning classification findings. Firstly, we propose an
accurate neural network architecture to classify skin lesions. Secondly, we
utilize a prediction difference analysis method that examines each patch on the
image through patch-wised corrupting to detect the biomarkers. Lastly, we
validate that our biomarker findings are corresponding to the patterns in the
literature. The findings can be significant and useful to guide clinical
diagnosis.
|
['Hongda Jiang', 'Xiaoxiao Li', 'Eric Z. Chen', 'Junyan Wu']
|
2018-11-02
| null | null | null | null |
['melanoma-diagnosis']
|
['computer-vision']
|
[ 4.95832324e-01 -1.26640886e-01 -4.89855111e-01 -1.09638326e-01
-5.25993407e-01 -1.93289340e-01 1.88164234e-01 3.44307214e-01
-3.53298426e-01 8.00805509e-01 -1.55102581e-01 -3.34450066e-01
-1.73961341e-01 -7.16318607e-01 -1.86813623e-01 -1.17153347e+00
9.26649868e-02 -1.62379727e-01 -1.24580618e-02 -3.50365788e-02
2.34976470e-01 5.23565114e-01 -1.15468097e+00 5.12271345e-01
1.13921940e+00 1.21393919e+00 -2.53714472e-02 6.53212368e-01
-1.24601901e-01 6.23191237e-01 -3.25198054e-01 -4.07002032e-01
1.44138217e-01 -5.78697741e-01 -5.93187094e-01 1.38273388e-01
1.96684510e-01 -4.62488413e-01 3.26779298e-03 1.09889722e+00
4.50201124e-01 -6.24031961e-01 7.40507841e-01 -8.18332314e-01
-5.23423851e-01 3.87855917e-02 -8.70347917e-01 1.15054406e-01
1.22187249e-01 2.50190467e-01 5.69361567e-01 -5.31237006e-01
5.53867102e-01 5.76746464e-01 7.06881464e-01 9.09442425e-01
-7.18487263e-01 -4.11736190e-01 -2.03459665e-01 4.39261019e-01
-1.14932668e+00 -1.14329875e-01 5.49410641e-01 -5.39184868e-01
2.96869963e-01 5.48564553e-01 1.04134512e+00 1.04867566e+00
5.83463490e-01 8.04357827e-01 1.47792232e+00 -5.02195299e-01
1.20414391e-01 1.68365076e-01 1.22408150e-02 9.52868223e-01
4.25716996e-01 1.21831492e-01 -4.70498443e-01 1.18829049e-01
6.30469918e-01 2.33836100e-01 -3.57768595e-01 2.83182114e-01
-8.30918193e-01 5.85632563e-01 6.09514296e-01 2.65414685e-01
-5.45561194e-01 7.39897937e-02 4.23063308e-01 -7.31722359e-03
6.84632361e-01 2.72099882e-01 -1.76184461e-01 3.72865014e-02
-7.17624664e-01 -4.03210491e-01 3.61453056e-01 -1.83525473e-01
3.10958683e-01 -2.29821011e-01 -6.25787675e-02 9.05576348e-01
4.83175814e-01 3.48593056e-01 5.16774595e-01 -4.15961981e-01
-3.54245007e-01 9.42777991e-01 -2.93495119e-01 -8.56370211e-01
-5.48885286e-01 -4.53013659e-01 -1.10217261e+00 3.02185118e-01
5.59287012e-01 -2.53250748e-01 -9.99839962e-01 1.05051351e+00
3.67650568e-01 1.59595072e-01 -9.05326307e-02 1.16985738e+00
7.18860805e-01 6.67354390e-02 3.08884025e-01 -1.07875273e-01
1.39959872e+00 -5.70217907e-01 -6.88233972e-01 8.67903158e-02
5.89660287e-01 -4.55611944e-01 5.81658244e-01 5.50534725e-01
-4.16867226e-01 -2.23555550e-01 -9.53870714e-01 1.10825770e-01
-3.11567128e-01 4.56522822e-01 1.08696461e+00 7.87863791e-01
-7.80240774e-01 4.58673239e-01 -9.87731934e-01 -8.29634428e-01
7.98568726e-01 3.29773784e-01 -4.30667639e-01 -6.27901331e-02
-1.01757038e+00 8.09015334e-01 8.97342525e-03 3.58681262e-01
-5.87953389e-01 -6.69738770e-01 -4.17747051e-01 -5.01114428e-01
-9.04212743e-02 -7.55349755e-01 9.45500791e-01 -1.05456233e+00
-1.36518431e+00 1.12306988e+00 -3.78063202e-01 -3.98971796e-01
4.29104924e-01 1.30488932e-01 -5.01996756e-01 4.00564611e-01
-2.04678804e-01 6.19257867e-01 7.94264138e-01 -1.06911683e+00
-9.13343370e-01 -6.35868967e-01 -2.11527169e-01 -9.51942131e-02
-6.58671856e-01 -1.24149531e-01 -1.19780019e-01 -2.15231672e-01
7.54284847e-05 -7.34515369e-01 -4.78867263e-01 6.47919238e-01
-6.73911273e-01 -9.79474485e-02 5.03519356e-01 -1.06163788e+00
1.04977703e+00 -2.08062029e+00 -2.05326736e-01 2.73541808e-01
5.80653787e-01 3.93346906e-01 1.18063958e-02 2.86072880e-01
6.02697022e-02 5.30047774e-01 1.08584873e-02 1.05936535e-01
-5.17372370e-01 -2.64897197e-01 1.10966779e-01 5.52196741e-01
7.11248159e-01 1.01528203e+00 -7.40306139e-01 -5.24259746e-01
2.05503479e-01 5.38803279e-01 1.50539130e-01 -1.49442345e-01
-8.59864615e-03 3.70826185e-01 -4.36803669e-01 1.21472788e+00
5.30568004e-01 -5.02288520e-01 2.70337790e-01 -4.01176989e-01
4.31013778e-02 -1.99393466e-01 -2.64779180e-01 1.29981506e+00
-1.81523815e-01 9.62879896e-01 -7.42464885e-02 -6.42496407e-01
7.47847617e-01 1.53980568e-01 4.57383662e-01 -6.55855894e-01
2.01626346e-01 2.45343164e-01 4.37260479e-01 -9.83091295e-01
-1.20727509e-01 -3.90431225e-01 5.65677166e-01 8.34216699e-02
-5.34218967e-01 3.09340596e-01 9.55390185e-02 -2.36096993e-01
1.04796040e+00 -2.13212892e-01 4.37438548e-01 2.14978352e-01
3.57649684e-01 2.99192339e-01 2.96749204e-01 2.69920796e-01
-4.28861499e-01 3.51853520e-01 5.64247191e-01 -4.26773757e-01
-6.09134436e-01 -8.29036891e-01 -5.49102843e-01 3.93561125e-01
1.23556979e-01 2.06666380e-01 -6.90582275e-01 -6.25778079e-01
1.91822201e-01 5.03589995e-02 -9.72116292e-01 -6.96109235e-02
8.80565196e-02 -1.23546231e+00 6.21503115e-01 4.92657661e-01
6.61799550e-01 -5.57839632e-01 -4.95047867e-01 -8.72339457e-02
1.67274296e-01 -6.30590081e-01 9.91116911e-02 2.46452522e-02
-8.75452101e-01 -1.56823969e+00 -1.02221096e+00 -7.05944300e-01
1.07377362e+00 3.26464057e-01 4.94667649e-01 4.48014140e-01
-1.09273589e+00 5.93755357e-02 -3.61005247e-01 -8.52440655e-01
-4.33423042e-01 -3.39077637e-02 -1.38007388e-01 2.67901242e-01
7.67111063e-01 7.98999369e-02 -8.44622493e-01 3.03947460e-03
-7.62267888e-01 1.94365248e-01 1.22636259e+00 9.73057270e-01
6.87227309e-01 1.30892426e-01 6.26919091e-01 -9.30575550e-01
8.12352002e-01 -5.90241611e-01 -1.08109415e-01 3.07910413e-01
-6.08150661e-01 -2.91745156e-01 3.77311021e-01 -1.21820755e-01
-8.48713636e-01 -3.08486577e-02 -2.40764081e-01 -1.58917194e-03
-4.10961837e-01 8.53020310e-01 2.76539236e-01 -3.47964972e-01
7.43954480e-01 7.16856793e-02 5.52167714e-01 -2.04632640e-01
-3.94452125e-01 9.88549471e-01 2.12675016e-02 1.35758698e-01
2.81054318e-01 6.89388454e-01 4.66978461e-01 -1.10527575e+00
-8.18033934e-01 -5.66323638e-01 -4.27063972e-01 -5.70710778e-01
9.06121194e-01 -7.62617171e-01 -6.86622083e-01 8.98695827e-01
-8.28841627e-01 -2.88608491e-01 5.01408763e-02 4.33127701e-01
2.07702786e-01 5.66332996e-01 -6.41633451e-01 -9.25565660e-01
-4.12697613e-01 -8.98876369e-01 1.15790749e+00 6.00574255e-01
-4.05184060e-01 -1.43223882e+00 8.38309005e-02 5.27958453e-01
2.26822242e-01 4.89323646e-01 9.27969277e-01 -3.85311306e-01
-3.83435369e-01 -4.93663192e-01 -4.50872868e-01 3.64885241e-01
5.03871739e-01 6.20780706e-01 -1.12570393e+00 7.86775425e-02
-3.53595585e-01 -2.57296324e-01 1.07686949e+00 5.90192616e-01
1.32363582e+00 1.80720508e-01 -6.71934187e-01 5.56596577e-01
1.50858307e+00 3.67251396e-01 6.38131499e-01 2.17007414e-01
4.60965306e-01 7.63178706e-01 6.17953479e-01 1.89419821e-01
1.31724417e-01 1.36932641e-01 5.35436273e-01 -6.97264016e-01
-3.61531168e-01 -1.75995566e-02 2.46977899e-02 4.63578403e-01
-4.69592512e-01 -1.72908023e-01 -1.01220036e+00 5.49667954e-01
-1.39296293e+00 -7.93480158e-01 -5.94041169e-01 1.93439007e+00
7.53827155e-01 -9.08417106e-02 4.05474193e-03 2.75962412e-01
7.39679039e-01 -3.88857812e-01 -8.03673089e-01 -2.80182183e-01
-1.73303559e-01 3.26165289e-01 3.54054093e-01 7.77519718e-02
-9.78379071e-01 4.67466742e-01 6.69489384e+00 6.02722526e-01
-1.84769559e+00 -1.80362701e-01 1.05867708e+00 5.99459745e-02
-1.43323481e-01 -2.91260868e-01 -4.98484582e-01 5.10689795e-01
5.70731282e-01 1.12344116e-01 -9.78512806e-04 3.68900806e-01
4.60472405e-01 -6.13609254e-01 -1.00651443e+00 7.75040567e-01
2.49486025e-02 -1.32886410e+00 1.41631039e-02 4.22549397e-01
5.36499023e-01 -3.67383659e-01 4.23369825e-01 -4.49275821e-01
-3.48126233e-01 -1.23619580e+00 -2.01161191e-01 1.03274000e+00
1.12231886e+00 -4.99199688e-01 1.13181353e+00 2.55263984e-01
-5.95252156e-01 -3.78330983e-03 -1.63829729e-01 -1.44286543e-01
-3.47696871e-01 9.74914551e-01 -1.48491609e+00 3.65235716e-01
4.22835171e-01 8.84773791e-01 -7.19091356e-01 1.36873257e+00
-3.29028368e-01 8.55814815e-01 3.45668383e-02 -5.31383753e-01
-1.05047904e-01 -2.20939238e-02 5.67448474e-02 8.34968567e-01
3.10589880e-01 -2.02910036e-01 -1.38823569e-01 6.49821639e-01
2.51772583e-01 2.32487261e-01 -3.71439666e-01 -4.11444366e-01
6.01101220e-02 1.43300784e+00 -7.73032010e-01 1.47654071e-01
-3.89575720e-01 9.55482543e-01 -3.63228992e-02 1.86472431e-01
-5.05208731e-01 -2.94123501e-01 6.66222632e-01 2.78813511e-01
-4.14518952e-01 1.74534336e-01 -5.47531843e-01 -7.71257937e-01
-1.39425546e-01 -5.74959278e-01 2.19646111e-01 -4.88447636e-01
-1.57713556e+00 2.10125268e-01 -6.53445721e-01 -1.08236074e+00
4.39499766e-02 -1.15643167e+00 -8.55194330e-01 9.22523737e-01
-1.71546221e+00 -1.22509170e+00 -7.54353940e-01 5.79339303e-02
2.62686193e-01 -8.75013322e-02 1.01296902e+00 -7.06863925e-02
-1.07145751e+00 5.77138722e-01 1.38591573e-01 2.93865860e-01
8.80393505e-01 -1.25925469e+00 -1.87559038e-01 5.67147017e-01
-3.67043465e-01 4.88029122e-01 2.84117818e-01 -6.64722145e-01
-1.33445072e+00 -1.00049484e+00 5.10362983e-01 -1.89840540e-01
7.30519712e-01 1.62998587e-01 -6.47330940e-01 2.49377126e-03
1.70156702e-01 -8.40932280e-02 1.42586255e+00 1.50617376e-01
-1.66590400e-02 -3.31377208e-01 -1.34332180e+00 5.82965851e-01
4.59826440e-01 -4.83913809e-01 9.19074565e-02 4.80111927e-01
-6.73981011e-02 -2.37179875e-01 -1.03395784e+00 3.84056538e-01
9.00345325e-01 -8.45937908e-01 5.72176456e-01 -7.03938544e-01
8.39845300e-01 -1.70731738e-01 1.98851645e-01 -1.34917676e+00
-2.37103760e-01 1.59248710e-01 4.04690169e-02 7.81627953e-01
4.85765040e-01 -5.87340832e-01 1.25353789e+00 5.74833810e-01
2.26779282e-01 -1.40930665e+00 -7.88845658e-01 -3.04502308e-01
-1.59235075e-02 -8.97236615e-02 1.62472114e-01 8.68140996e-01
1.75041556e-01 -9.18992609e-02 5.46998791e-02 4.57171164e-02
5.74843228e-01 -1.66442677e-01 4.18469071e-01 -1.33871150e+00
-5.80270439e-02 -4.36883837e-01 -6.47298276e-01 -3.10948014e-01
-1.53794467e-01 -8.57592344e-01 -3.45519632e-01 -1.77482021e+00
5.27863264e-01 -2.92066157e-01 -4.21654820e-01 5.91344893e-01
-4.69079524e-01 5.94762862e-01 -2.57107913e-01 5.30480519e-02
-1.00636937e-01 -7.29543865e-02 1.49422467e+00 -4.81791228e-01
1.84955329e-01 -2.51479428e-02 -9.99823749e-01 6.54448390e-01
1.07894909e+00 4.92726564e-02 -1.40650392e-01 -1.73558652e-01
3.03893685e-01 -1.90488890e-01 7.35292971e-01 -9.70544577e-01
3.47974628e-01 -3.15878749e-01 8.73669267e-01 -3.32034081e-01
2.46096462e-01 -6.06419086e-01 -7.23470747e-02 9.50163722e-01
-2.15475649e-01 -7.03097820e-01 1.99884757e-01 5.59428453e-01
-2.84649938e-01 -2.93841392e-01 7.55311430e-01 -4.46082801e-02
-1.04832458e+00 2.26802409e-01 -6.73452318e-01 -5.46268940e-01
1.44004428e+00 -5.52925467e-01 -5.00105083e-01 -1.02335684e-01
-6.01219714e-01 3.82956043e-02 3.88818353e-01 -8.60824361e-02
8.93010557e-01 -1.03123558e+00 -7.31440961e-01 6.11491054e-02
2.82064706e-01 -1.95262715e-01 6.68480694e-01 1.23231161e+00
-7.75808573e-01 3.55878025e-01 -2.97695726e-01 -7.78997719e-01
-1.44929779e+00 1.09139822e-01 5.11893392e-01 -1.14410289e-01
-1.09899780e-02 1.13667059e+00 -1.40450135e-01 1.74169332e-01
3.87931503e-02 -2.42807776e-01 -3.89233947e-01 1.72645822e-01
6.90972924e-01 3.56075525e-01 5.77879064e-02 -4.28971685e-02
-2.63548940e-01 5.58544755e-01 -4.63284492e-01 4.63285834e-01
1.04456341e+00 1.74951479e-01 -4.24605191e-01 4.37037826e-01
8.29376757e-01 -4.74203676e-02 -1.01717329e+00 3.28231961e-01
-1.42705053e-01 -4.78615671e-01 2.81065255e-01 -1.28998494e+00
-1.10970688e+00 1.14500690e+00 1.06790447e+00 2.40279183e-01
1.32374418e+00 -3.47240180e-01 7.06077099e-01 2.53182381e-01
2.74671316e-01 -9.62201357e-01 -5.60024418e-02 -2.44601473e-01
4.92985457e-01 -1.58753753e+00 1.46905407e-01 -7.25086033e-01
-5.18486857e-01 1.38685107e+00 6.74210191e-01 9.98699442e-02
5.37212193e-01 2.82376468e-01 5.40408731e-01 -7.73093775e-02
-7.21705317e-01 -3.31690431e-01 3.28058153e-01 7.77794421e-01
7.49973655e-01 2.49827191e-01 -5.28892338e-01 4.90949631e-01
3.63585114e-01 2.80644029e-01 3.58631402e-01 6.64074004e-01
-3.95731151e-01 -1.19263828e+00 -3.19500089e-01 1.17963827e+00
-6.62839413e-01 -1.40933812e-01 -1.04259539e+00 8.00556481e-01
2.13904798e-01 8.74618292e-01 2.65296120e-02 -5.49668074e-01
-2.10595787e-01 1.59094762e-02 4.35151696e-01 -4.79371876e-01
-3.16352010e-01 -1.17121041e-01 1.28987536e-01 -1.49378434e-01
-6.18143976e-01 -5.64972281e-01 -1.16965413e+00 -3.69453467e-02
-3.61141115e-01 -2.83615351e-01 9.50931489e-01 9.52817678e-01
3.07872146e-01 6.30020976e-01 6.47797108e-01 -1.64174885e-01
-2.04817921e-01 -8.90929163e-01 -8.06220710e-01 7.22432807e-02
2.45002061e-01 -4.60430413e-01 -2.12282106e-01 7.59659633e-02]
|
[15.600715637207031, -3.0209829807281494]
|
42ddec2b-1a30-492e-a686-edff3f53bee4
|
discovering-multiple-algorithm-configurations
|
2303.07434
| null |
https://arxiv.org/abs/2303.07434v1
|
https://arxiv.org/pdf/2303.07434v1.pdf
|
Discovering Multiple Algorithm Configurations
|
Many practitioners in robotics regularly depend on classic, hand-designed algorithms. Often the performance of these algorithms is tuned across a dataset of annotated examples which represent typical deployment conditions. Automatic tuning of these settings is traditionally known as algorithm configuration. In this work, we extend algorithm configuration to automatically discover multiple modes in the tuning dataset. Unlike prior work, these configuration modes represent multiple dataset instances and are detected automatically during the course of optimization. We propose three methods for mode discovery: a post hoc method, a multi-stage method, and an online algorithm using a multi-armed bandit. Our results characterize these methods on synthetic test functions and in multiple robotics application domains: stereoscopic depth estimation, differentiable rendering, motion planning, and visual odometry. We show the clear benefits of detecting multiple modes in algorithm configuration space.
|
['Martial Hebert', 'Leonid Keselman']
|
2023-03-13
| null | null | null | null |
['motion-planning']
|
['robots']
|
[ 3.08536679e-01 -6.88131303e-02 -5.36458731e-01 -1.17340133e-01
-9.79073167e-01 -1.12706041e+00 6.25157475e-01 -4.55210768e-02
-3.21335077e-01 7.23722398e-01 -4.60288301e-03 -4.48239982e-01
-8.97135973e-01 -2.97190815e-01 -8.28957498e-01 -6.08177781e-01
-3.91212553e-01 1.11941850e+00 1.80558115e-01 5.17055281e-02
8.25777471e-01 4.35215056e-01 -1.63506711e+00 1.19014360e-01
7.27348089e-01 5.90061307e-01 4.36229736e-01 9.11167979e-01
1.25562713e-01 2.22179070e-01 -5.96802413e-01 7.75393844e-02
8.27138007e-01 -1.86310828e-01 -7.99191356e-01 3.84127080e-01
1.44249141e-01 -1.95634648e-01 2.81416953e-01 9.83126581e-01
5.22088408e-01 2.00059742e-01 4.99893963e-01 -1.39036071e+00
1.22941725e-01 8.29379201e-01 -7.22924948e-01 1.36435404e-01
5.16506612e-01 4.92255569e-01 1.04300499e+00 -5.70743620e-01
5.52227437e-01 1.17493105e+00 5.59355319e-01 2.09135383e-01
-1.51143444e+00 -3.85059893e-01 3.39778334e-01 -4.51500062e-03
-1.28080261e+00 -4.20704275e-01 7.80473650e-01 -1.00023568e+00
7.51076639e-01 6.96064830e-02 7.15413392e-01 8.02358389e-01
-2.22276151e-01 8.60969484e-01 9.66823220e-01 -2.90497154e-01
6.95764542e-01 -1.35083705e-01 -1.56810939e-01 8.25876892e-01
4.08501148e-01 4.00925159e-01 -6.51180267e-01 -4.90615338e-01
9.26367581e-01 -3.69033635e-01 -1.78376228e-01 -8.62276673e-01
-1.51740682e+00 6.33666098e-01 -9.85240750e-03 -3.66114497e-01
-3.83328617e-01 3.71851593e-01 5.26515394e-02 1.09274186e-01
1.28200769e-01 1.15154195e+00 -9.17554796e-01 -4.05093014e-01
-6.54038191e-01 4.11088377e-01 8.71166706e-01 1.05425692e+00
1.01062942e+00 -4.56453234e-01 1.32166445e-02 8.37276399e-01
2.26264119e-01 2.25356892e-01 7.05026031e-01 -1.23849988e+00
5.13238668e-01 5.54480791e-01 6.92144692e-01 -6.24607682e-01
-5.95238864e-01 -2.51408696e-01 -4.76253554e-02 1.41138822e-01
5.00535071e-01 -2.82713294e-01 -8.93110991e-01 1.43009150e+00
6.38424635e-01 1.08473644e-01 -6.12712055e-02 9.54076767e-01
1.41350225e-01 2.08660617e-01 -4.25183654e-01 -1.91412970e-01
8.66082549e-01 -1.13873589e+00 -3.69517535e-01 -5.67180037e-01
4.72463578e-01 -6.48572505e-01 1.27256143e+00 8.35772157e-01
-8.94547224e-01 -1.15634799e-01 -9.85828936e-01 7.53034174e-01
-3.44878510e-02 2.56653488e-01 8.38784158e-01 6.48743033e-01
-7.11336374e-01 7.02407300e-01 -8.81717265e-01 -1.51753530e-01
2.38660783e-01 5.48627377e-01 1.58662111e-01 2.30288029e-01
-3.46094221e-01 6.69623435e-01 4.19681579e-01 -5.02622314e-02
-1.18270695e+00 -5.63475609e-01 -3.88647377e-01 -4.22684908e-01
8.38488936e-01 -7.09070265e-01 1.72609675e+00 -9.37178373e-01
-2.03124881e+00 8.10709476e-01 8.00209194e-02 -2.28597790e-01
5.60112536e-01 -4.25429285e-01 2.26730973e-01 -1.88276738e-01
1.91946685e-01 5.37192702e-01 1.00428259e+00 -1.36396527e+00
-8.64567161e-01 -2.33201504e-01 3.53231132e-01 3.22039932e-01
1.24684744e-01 -2.71190643e-01 -5.97577810e-01 -3.05713087e-01
3.33360851e-01 -1.34159410e+00 -5.87000549e-01 -2.99295396e-01
-8.02476346e-01 9.07477736e-02 2.53142089e-01 1.53497988e-02
9.52685714e-01 -1.67560291e+00 5.51385641e-01 4.76081610e-01
-2.11704135e-01 -5.25222123e-01 -9.52367410e-02 2.94438243e-01
1.50209114e-01 8.14052224e-02 -1.25300378e-01 -2.75154859e-02
1.48106560e-01 2.54435092e-01 -2.34330848e-01 7.02610552e-01
5.36889844e-02 4.51851666e-01 -1.08390868e+00 -3.24885190e-01
1.16718464e-01 -5.38734972e-01 -8.81980956e-01 2.58124024e-01
-7.89149642e-01 7.31672227e-01 -4.69115943e-01 8.59746397e-01
1.79293409e-01 -4.34549868e-01 2.48206079e-01 5.99085130e-02
-2.19693735e-01 5.49648590e-02 -1.60070062e+00 1.95772099e+00
-2.40424976e-01 2.44425148e-01 -7.75805488e-02 -8.74372423e-01
6.54488206e-01 -2.03182057e-01 7.86063194e-01 -7.21725821e-02
2.56276399e-01 1.05416268e-01 1.19852737e-01 -8.34844351e-01
6.51534736e-01 4.13864493e-01 -2.15839222e-01 6.44977391e-01
-2.03441501e-01 -5.64025760e-01 3.72913092e-01 -3.14836830e-01
1.38265324e+00 3.91275316e-01 4.99548584e-01 -3.20659399e-01
-9.23084468e-02 4.44169581e-01 4.59201962e-01 1.13690817e+00
-1.74802542e-01 5.11533082e-01 4.41479504e-01 -4.73677844e-01
-8.68204653e-01 -8.71689320e-01 -2.38871440e-01 1.36158323e+00
5.70289552e-01 -2.69147485e-01 -5.87113082e-01 -4.24221247e-01
2.73274422e-01 3.73807192e-01 -6.76657736e-01 -7.07010925e-02
-5.06840587e-01 -8.57914209e-01 2.83538967e-01 2.38815382e-01
1.79067880e-01 -9.78818953e-01 -1.31759787e+00 1.61100909e-01
-7.83881769e-02 -1.07561946e+00 -3.16693664e-01 5.13020992e-01
-8.39231849e-01 -1.39053857e+00 -1.40344545e-01 -3.88615310e-01
5.98030031e-01 1.55236453e-01 1.15151715e+00 -8.94649997e-02
-3.56499285e-01 5.36601722e-01 -3.63938540e-01 -5.72430432e-01
-2.38593757e-01 1.94444835e-01 2.89014131e-01 -1.77572772e-01
-2.42643178e-01 -4.34649765e-01 -5.07619023e-01 6.87344074e-01
-5.30366004e-01 2.15662479e-01 8.64413738e-01 7.91860104e-01
8.94813836e-01 -7.64750466e-02 -1.84232101e-03 -8.65164995e-01
5.58628619e-01 -7.26733327e-01 -1.20664465e+00 2.96874732e-01
-5.89809537e-01 5.14555752e-01 1.00252256e-01 -9.98060584e-01
-5.17195523e-01 7.88596392e-01 7.68876791e-01 -6.83269143e-01
-1.31032795e-01 7.03712642e-01 -1.51772785e-03 -1.60630777e-01
1.07425320e+00 -2.31629893e-01 -1.25067696e-01 -3.01863164e-01
3.92683417e-01 1.64028451e-01 4.70399618e-01 -1.22356856e+00
8.58906627e-01 3.73831898e-01 -8.17838609e-02 -4.96287346e-01
-7.85166264e-01 -5.74516475e-01 -5.30530691e-01 -4.70165193e-01
3.23144764e-01 -5.82453012e-01 -7.29959130e-01 2.48387486e-01
-8.91598761e-01 -9.59591031e-01 -2.36218497e-01 3.19852948e-01
-1.04704118e+00 -6.85687885e-02 1.27462566e-01 -9.46799994e-01
2.36364797e-01 -1.57151318e+00 1.34439898e+00 1.73438802e-01
-2.26502046e-01 -5.42965412e-01 4.08383071e-01 3.44317585e-01
7.39952084e-03 3.53371203e-01 5.83291233e-01 -3.59426141e-01
-9.73683834e-01 4.07627970e-02 1.90818623e-01 -5.39472759e-01
1.57134935e-01 3.34897876e-01 -7.61720359e-01 -4.02387083e-01
-5.89456201e-01 -2.68542379e-01 4.97253239e-01 6.16952538e-01
1.54149914e+00 -3.74718815e-01 -4.41872388e-01 8.34189296e-01
1.21044445e+00 3.41087848e-01 8.65126625e-02 8.27867329e-01
3.45743448e-01 5.26124477e-01 9.68649209e-01 7.90514231e-01
8.35827067e-02 8.94440055e-01 8.34699035e-01 3.33467007e-01
5.71555018e-01 -4.53839190e-02 2.16535047e-01 -3.60449776e-02
1.08773850e-01 -3.89061384e-02 -1.40090215e+00 6.97213650e-01
-2.27077818e+00 -5.87656260e-01 3.82171214e-01 2.25494194e+00
6.88759089e-01 3.01749349e-01 5.32663465e-01 5.54624833e-02
8.49652350e-01 -3.64281923e-01 -1.08754599e+00 -3.41426916e-02
2.15247139e-01 -2.25163758e-01 9.72078443e-01 3.02489817e-01
-1.27625358e+00 8.41013014e-01 6.58688736e+00 3.80866647e-01
-7.70158589e-01 -2.34875992e-01 2.77740180e-01 -4.58089292e-01
-3.21942242e-03 1.58502564e-01 -7.19374597e-01 3.64011675e-01
5.61924934e-01 -2.45661773e-02 9.39552605e-01 1.05828762e+00
1.83030277e-01 -5.33362806e-01 -1.25821149e+00 1.17212927e+00
-2.85282701e-01 -1.50423431e+00 -4.36843127e-01 2.16876730e-01
9.68556702e-01 1.20941117e-01 1.66284591e-01 -4.65157293e-02
9.92274821e-01 -8.48082960e-01 1.08614147e+00 2.68366069e-01
6.90893352e-01 -6.23822212e-01 1.74765706e-01 4.57975030e-01
-7.86190331e-01 -7.24801421e-01 -1.87602397e-02 -8.28257471e-04
4.15651873e-03 3.89297903e-01 -1.26878905e+00 1.83687076e-01
9.08037603e-01 4.66374427e-01 -3.37455094e-01 1.59410286e+00
-1.50529832e-01 4.45497811e-01 -6.35205269e-01 -3.17374527e-01
1.18127808e-01 -9.16415453e-02 9.67831433e-01 6.97712541e-01
7.77483732e-02 -1.21914662e-01 5.57086349e-01 7.67686129e-01
4.39013392e-01 -2.22418979e-01 -6.40951633e-01 1.02385178e-01
7.04307437e-01 1.03310621e+00 -9.98359144e-01 2.25870863e-01
1.02486178e-01 5.29508650e-01 2.09549084e-01 3.07295650e-01
-8.63672256e-01 9.63988341e-03 8.49162877e-01 -1.54266104e-01
5.17055154e-01 -4.38898712e-01 -4.98583138e-01 -8.59877169e-01
-2.71865390e-02 -9.22093451e-01 3.99843723e-01 -6.39840424e-01
-1.04579377e+00 1.36658818e-01 4.44891781e-01 -1.37170851e+00
-3.77416521e-01 -5.80703855e-01 -3.14619720e-01 9.06619579e-02
-1.10913110e+00 -7.30154812e-01 -4.36532408e-01 4.09656912e-01
7.57180333e-01 -2.98764467e-01 5.49870849e-01 5.82590234e-03
-8.74165177e-01 3.82989883e-01 2.18818024e-01 -2.51050502e-01
5.87053776e-01 -1.39457119e+00 4.75658067e-02 7.33188272e-01
1.99475214e-01 4.99703735e-01 1.06794691e+00 -6.46106303e-01
-1.68555987e+00 -1.05510485e+00 -4.78944272e-01 -5.33355892e-01
8.06816876e-01 -5.69148846e-02 -2.49358743e-01 6.83572292e-01
-2.62128294e-01 -2.35260814e-01 3.60353768e-01 3.45106602e-01
-2.86484491e-02 2.15972662e-02 -1.01119494e+00 5.60343266e-01
1.19845080e+00 9.27724168e-02 -2.84755141e-01 6.97217882e-01
5.71607113e-01 -9.69473720e-01 -5.50334930e-01 4.41393733e-01
6.26272380e-01 -7.14906931e-01 9.77325261e-01 -6.93609297e-01
2.91278601e-01 -5.71010351e-01 -2.20330030e-01 -1.47178876e+00
-9.79716331e-02 -1.26576018e+00 -2.46783122e-01 6.50020421e-01
6.09920681e-01 -3.43374312e-01 9.32446301e-01 4.26243126e-01
-2.43696034e-01 -7.08174884e-01 -8.80070984e-01 -7.96676517e-01
-3.74764889e-01 -4.15151924e-01 7.62683213e-01 8.71139824e-01
-1.55808032e-01 3.61301988e-01 -9.60027128e-02 6.10704660e-01
6.95359588e-01 3.99582505e-01 1.34428477e+00 -1.21019626e+00
-8.03188980e-01 -7.88222134e-01 -1.20149061e-01 -1.27808833e+00
6.24844944e-03 -3.07212174e-01 7.34595954e-01 -9.88715589e-01
8.27310234e-02 -7.32102513e-01 3.82367410e-02 3.84300649e-01
-4.07291539e-02 -4.11562383e-01 -3.78661335e-01 4.53612089e-01
-9.13246095e-01 1.94431558e-01 1.03981888e+00 -7.25919530e-02
-6.02207601e-01 3.54140669e-01 -5.49336851e-01 7.69716620e-01
8.71400654e-01 -4.76756603e-01 -5.35623014e-01 -5.89040935e-01
6.99150622e-01 1.55246615e-01 2.92422384e-01 -9.27264333e-01
3.30925256e-01 -8.78718317e-01 4.05379292e-03 -6.89271271e-01
2.93119729e-01 -7.91320801e-01 1.77693054e-01 3.45303446e-01
-3.30934048e-01 2.52915114e-01 1.53866053e-01 8.88364375e-01
3.86036277e-01 -2.01117456e-01 8.51426721e-01 -2.55260348e-01
-9.56254601e-01 1.98108837e-01 -3.88171107e-01 7.65656233e-02
1.21388757e+00 -3.02621961e-01 -2.88479328e-01 -2.08714619e-01
-5.22863269e-01 5.63508034e-01 5.51409245e-01 3.34727228e-01
3.73826832e-01 -8.00224721e-01 -2.99139261e-01 4.42227013e-02
3.26327413e-01 4.99040782e-01 -3.70440304e-01 7.81874597e-01
-4.66437608e-01 -6.68301359e-02 -2.52858363e-02 -1.17831278e+00
-9.42603111e-01 3.33360404e-01 5.22913337e-01 -4.67035063e-02
-2.54378021e-01 9.60192502e-01 -2.24931240e-01 -7.36530304e-01
4.29837763e-01 -4.72829610e-01 -1.03579894e-01 -1.26945481e-01
1.27150446e-01 3.56358588e-01 -3.01810820e-02 -9.22221988e-02
-3.10518891e-01 6.05092287e-01 4.59167570e-01 -3.55875492e-01
1.43504548e+00 -1.12177499e-01 3.59486192e-02 6.93361759e-01
4.55905139e-01 -2.53085464e-01 -1.50851965e+00 -7.70024583e-02
3.21670592e-01 -6.12313747e-01 -1.04953282e-01 -8.14610243e-01
-7.57415712e-01 1.06161095e-01 4.81949270e-01 3.64525318e-01
8.55926991e-01 7.66592324e-02 2.18455382e-02 9.67745721e-01
7.38944411e-01 -1.33763278e+00 3.30957294e-01 5.85094333e-01
8.38914156e-01 -1.39657223e+00 1.41976789e-01 -1.69583082e-01
-4.49244261e-01 1.00081658e+00 8.25279832e-01 -7.38037080e-02
5.73985457e-01 4.22073007e-01 -1.18663825e-01 -3.02863687e-01
-7.92724431e-01 -2.84053475e-01 8.10942352e-02 3.27190995e-01
-2.82479405e-01 -2.71068923e-02 3.45087886e-01 3.06552917e-01
-5.80615997e-01 -4.02241558e-01 5.62956035e-01 1.14671850e+00
-5.92412949e-01 -7.62232721e-01 -5.84608972e-01 6.03336215e-01
-4.48101088e-02 2.82168657e-01 -5.94490647e-01 6.38735831e-01
-1.81570664e-01 9.47846949e-01 -1.22168064e-01 -4.81260687e-01
4.97922122e-01 -1.96464583e-01 7.89003432e-01 -6.59998000e-01
-4.63956058e-01 9.21385922e-03 3.18143249e-01 -8.41856897e-01
-4.83184963e-01 -9.46464956e-01 -9.45750773e-01 1.26105666e-01
-6.30572438e-01 3.81250381e-02 9.02523696e-01 9.09993410e-01
4.29651618e-01 4.04041141e-01 6.21972263e-01 -1.34912038e+00
-5.99956751e-01 -6.59938931e-01 -9.93801728e-02 1.61074534e-01
5.64794123e-01 -1.29465818e+00 -2.90945292e-01 1.55082777e-01]
|
[4.348991394042969, 1.9167594909667969]
|
e17a1279-d45b-46b7-832b-d737374a907d
|
harmonizing-pathological-and-normal-pixels
|
2203.15347
| null |
https://arxiv.org/abs/2203.15347v1
|
https://arxiv.org/pdf/2203.15347v1.pdf
|
Harmonizing Pathological and Normal Pixels for Pseudo-healthy Synthesis
|
Synthesizing a subject-specific pathology-free image from a pathological image is valuable for algorithm development and clinical practice. In recent years, several approaches based on the Generative Adversarial Network (GAN) have achieved promising results in pseudo-healthy synthesis. However, the discriminator (i.e., a classifier) in the GAN cannot accurately identify lesions and further hampers from generating admirable pseudo-healthy images. To address this problem, we present a new type of discriminator, the segmentor, to accurately locate the lesions and improve the visual quality of pseudo-healthy images. Then, we apply the generated images into medical image enhancement and utilize the enhanced results to cope with the low contrast problem existing in medical image segmentation. Furthermore, a reliable metric is proposed by utilizing two attributes of label noise to measure the health of synthetic images. Comprehensive experiments on the T2 modality of BraTS demonstrate that the proposed method substantially outperforms the state-of-the-art methods. The method achieves better performance than the existing methods with only 30\% of the training data. The effectiveness of the proposed method is also demonstrated on the LiTS and the T1 modality of BraTS. The code and the pre-trained model of this study are publicly available at https://github.com/Au3C2/Generator-Versus-Segmentor.
|
['Yizhou Yu', 'Lin Yang', 'Guisheng Wang', 'Xinghao Ding', 'Yue Huang', 'LiyanSun', 'Yihong Zhuang', 'Xin Lin', 'Yunlong Zhang']
|
2022-03-29
| null | null | null | null |
['medical-image-enhancement']
|
['computer-vision']
|
[ 5.19011974e-01 1.01193063e-01 -1.63240179e-01 -4.63782176e-02
-1.05423212e+00 -2.47402146e-01 1.94481596e-01 -3.82576138e-01
-1.70290709e-01 7.95733988e-01 -7.09299073e-02 -4.34016064e-03
2.84340799e-01 -7.17263579e-01 -4.41872448e-01 -1.10438502e+00
3.41702044e-01 1.93974942e-01 3.22387069e-01 -1.96573753e-02
-1.25785396e-01 2.32428342e-01 -1.13181138e+00 2.93626279e-01
1.33026397e+00 1.06229603e+00 4.35560226e-01 4.79259759e-01
4.01570946e-02 5.72529197e-01 -6.95220530e-01 -3.03174168e-01
2.76738256e-01 -1.04918277e+00 -5.94764829e-01 2.02771336e-01
1.24604657e-01 -2.57083148e-01 -3.34628463e-01 1.25348318e+00
8.02307189e-01 -1.66469112e-01 7.11318910e-01 -1.20611632e+00
-7.16180444e-01 3.56577337e-01 -6.80209935e-01 2.87127852e-01
-9.08893123e-02 3.39354128e-01 3.13981593e-01 -5.23926139e-01
5.68879128e-01 8.25914741e-01 4.96545076e-01 7.24676907e-01
-1.01019359e+00 -8.83104444e-01 -3.91885757e-01 2.00593367e-01
-1.38447201e+00 -2.55932719e-01 9.11624670e-01 -3.81873280e-01
3.26597616e-02 5.24084687e-01 6.37115419e-01 1.06086218e+00
3.77630115e-01 9.20877814e-01 1.59049869e+00 -2.67966598e-01
-6.45211479e-03 1.15491956e-01 -3.52989107e-01 8.02509427e-01
6.22758493e-02 2.32598022e-01 -7.71053657e-02 4.22695540e-02
9.38493967e-01 2.74265595e-02 -6.79034889e-01 -6.49097189e-02
-1.10304737e+00 6.32843673e-01 5.96221268e-01 5.34490764e-01
-4.80069518e-01 6.48593605e-02 3.44845265e-01 -1.37584984e-01
3.87908727e-01 1.92157239e-01 2.89342850e-01 1.76814899e-01
-9.75637317e-01 -1.23453759e-01 1.22843325e-01 7.12841153e-01
1.66408300e-01 4.86504346e-01 -4.50636089e-01 9.93111014e-01
1.50311619e-01 6.53696954e-01 9.15509820e-01 -7.84841716e-01
4.76363711e-02 5.25732517e-01 -1.31737784e-01 -8.48882079e-01
-9.60791409e-02 -8.07277322e-01 -1.14080882e+00 2.40323871e-01
2.72920936e-01 -2.89335232e-02 -1.37059855e+00 1.59418213e+00
4.80122507e-01 3.13949674e-01 1.16021655e-01 1.05796266e+00
8.89955640e-01 6.30341351e-01 1.30669743e-01 -3.30089688e-01
1.19542503e+00 -1.05312705e+00 -1.03119648e+00 -1.89492598e-01
2.12014154e-01 -8.82618487e-01 1.11352694e+00 3.55447710e-01
-1.25639045e+00 -6.44982040e-01 -1.11289573e+00 4.35652524e-01
-2.81762853e-02 2.80510664e-01 3.16465229e-01 7.29407072e-01
-8.32314968e-01 2.54373223e-01 -8.80294621e-01 -1.52620912e-01
6.21758461e-01 9.86634940e-02 -1.99504763e-01 -2.88742781e-01
-1.24739921e+00 8.09740722e-01 3.30579519e-01 1.06254078e-01
-1.13999069e+00 -6.76623285e-01 -5.93674898e-01 -4.50778335e-01
2.16631293e-01 -6.61497593e-01 1.07206547e+00 -1.21639490e+00
-1.25655401e+00 1.01658154e+00 1.47383556e-01 -2.76912779e-01
7.46967435e-01 2.75839001e-01 -5.10494053e-01 5.06099820e-01
1.92731217e-01 6.00730062e-01 9.33354855e-01 -1.54418504e+00
-4.51762408e-01 -2.36313924e-01 -3.55709702e-01 1.17763773e-01
4.30826321e-02 -1.41738504e-01 -5.03144085e-01 -1.11560476e+00
8.05743933e-02 -1.03742814e+00 -1.44411862e-01 7.52812326e-02
-6.52347028e-01 3.08758974e-01 8.17136884e-01 -1.02034867e+00
9.41830933e-01 -2.15463734e+00 -5.20539768e-02 1.09766170e-01
2.05425188e-01 5.00764370e-01 -8.21150616e-02 -1.86297055e-02
-1.80083886e-01 1.98919177e-01 -6.60876811e-01 -5.68878166e-02
-3.40880841e-01 1.10175394e-01 1.68446124e-01 5.86161792e-01
2.77939700e-02 1.00264049e+00 -7.96089947e-01 -7.74912655e-01
2.64874011e-01 4.89932984e-01 -1.98708460e-01 3.88804972e-01
1.83640227e-01 9.95352209e-01 -5.83947062e-01 7.92229652e-01
7.62914062e-01 -1.43610448e-01 5.59077524e-02 -4.31959271e-01
3.22315097e-01 -4.30501312e-01 -8.18058729e-01 1.55314732e+00
-2.76292443e-01 2.41459653e-01 1.23365022e-01 -1.11127615e+00
7.54960895e-01 5.29792488e-01 5.58961868e-01 -8.27714145e-01
3.38183701e-01 4.53413874e-01 2.07955629e-01 -7.42518425e-01
-7.51773790e-02 -4.20945913e-01 1.53842807e-01 2.53613174e-01
-2.08979741e-01 -2.62829900e-01 1.18989676e-01 -4.95528243e-02
9.54111755e-01 -8.86585042e-02 5.05995043e-02 1.75892077e-02
5.86551845e-01 5.70918582e-02 7.75054216e-01 4.21339929e-01
-4.92393255e-01 8.90877128e-01 2.27820590e-01 1.13207802e-01
-1.02470326e+00 -1.10577166e+00 -2.25460276e-01 2.58457154e-01
2.76878327e-01 1.92287147e-01 -1.02478874e+00 -7.40219176e-01
-3.28877687e-01 6.40204608e-01 -6.44568086e-01 -3.77214879e-01
-4.55454975e-01 -1.00187206e+00 6.94833755e-01 5.38798571e-01
9.39685822e-01 -1.04783392e+00 -3.86389315e-01 1.49252445e-01
-5.73075533e-01 -1.04905701e+00 -5.64319968e-01 -2.68061906e-01
-7.99025536e-01 -1.06078613e+00 -1.24257898e+00 -9.43118989e-01
9.95061696e-01 1.52720436e-02 8.04377317e-01 2.58997500e-01
-4.33733940e-01 7.53322914e-02 -5.70005774e-01 -3.11077386e-01
-8.88452351e-01 -3.00862938e-01 -3.44882667e-01 1.36277691e-01
-3.03084850e-01 -4.66362089e-01 -1.02820206e+00 5.14594972e-01
-1.21823084e+00 3.85636330e-01 9.56236482e-01 1.14582348e+00
9.75377560e-01 2.93647796e-01 6.84175313e-01 -1.01760757e+00
4.56294805e-01 -3.62203598e-01 -1.59783021e-01 3.36204976e-01
-6.16209626e-01 -1.76328257e-01 5.91825843e-01 -5.00033438e-01
-1.14359665e+00 -5.28678261e-02 -3.29682201e-01 -4.79736447e-01
-9.12605077e-02 2.21103370e-01 -1.74620166e-01 -2.81020522e-01
5.27517736e-01 5.84248185e-01 2.16032952e-01 -2.10440040e-01
1.76797584e-01 7.17333436e-01 7.50176132e-01 -3.27217013e-01
8.06026638e-01 5.75277627e-01 -1.09596565e-01 -4.02830213e-01
-5.55548310e-01 -1.96710870e-01 -1.60038143e-01 -4.31595653e-01
9.20446157e-01 -6.94071949e-01 -1.16749533e-01 8.41499746e-01
-8.42625558e-01 -3.81773621e-01 -2.91567564e-01 5.40238082e-01
-6.00116074e-01 2.64851958e-01 -7.54421055e-01 -5.33960164e-01
-6.95969045e-01 -1.56405270e+00 7.75372446e-01 4.49192107e-01
1.70488447e-01 -8.54372621e-01 -1.28713578e-01 7.27272034e-01
5.35488486e-01 8.76619577e-01 9.91584480e-01 -3.65051985e-01
-5.01515388e-01 -3.30055505e-01 -9.35228765e-02 7.81508863e-01
3.64208996e-01 -2.83204854e-01 -7.74818301e-01 -2.63549894e-01
2.51820952e-01 -1.96270704e-01 6.79991007e-01 6.35463655e-01
1.20021331e+00 -1.54261068e-01 -2.12626263e-01 7.04908669e-01
1.39501619e+00 6.03592396e-01 9.18591619e-01 5.84652908e-02
6.56284869e-01 3.17670524e-01 8.29562604e-01 7.99170062e-02
8.19727592e-03 5.69121480e-01 4.14288223e-01 -6.94184124e-01
-7.48470902e-01 -1.94042444e-01 1.24755621e-01 9.03862596e-01
-6.29569590e-02 -3.46598178e-01 -8.51379156e-01 5.67094326e-01
-1.47997522e+00 -6.61288619e-01 -9.57971513e-02 1.89517450e+00
1.09062696e+00 -1.70992576e-02 -1.75237522e-01 2.97202855e-01
9.74002779e-01 -2.98132841e-02 -6.59083903e-01 4.77728061e-02
2.06664699e-04 4.10322726e-01 5.61513901e-01 2.04369769e-01
-8.67723167e-01 6.79522932e-01 6.04044247e+00 1.29334855e+00
-1.37863684e+00 5.18046081e-01 9.32012618e-01 1.49947003e-01
-3.06614280e-01 -3.41137260e-01 -2.40209013e-01 8.04753184e-01
6.20634854e-01 -2.36640759e-02 2.71110505e-01 5.20795107e-01
3.68807763e-01 -1.56824335e-01 -5.23333311e-01 8.68056655e-01
2.98744202e-01 -1.03412485e+00 -1.49161322e-02 2.74548976e-04
7.97147274e-01 -3.89224678e-01 4.02347326e-01 -1.06758652e-02
-5.95338643e-02 -1.06660616e+00 5.28277993e-01 6.18635058e-01
1.36196470e+00 -7.08790421e-01 9.57101166e-01 2.25775242e-01
-9.47336733e-01 2.31719479e-01 -5.66526502e-02 7.29331434e-01
1.40237451e-01 5.91177642e-01 -8.94840717e-01 6.92230523e-01
4.96290505e-01 4.60697889e-01 -6.54603481e-01 1.09229338e+00
-4.95922297e-01 7.93113232e-01 1.18784733e-01 4.93850052e-01
4.91655879e-02 -1.88746721e-01 6.05119169e-01 8.72307777e-01
4.71667141e-01 2.42306262e-01 -7.96204582e-02 9.90689576e-01
-5.92748448e-03 2.50687927e-01 -3.28667223e-01 -9.74734053e-02
2.13386506e-01 1.32257283e+00 -8.93553913e-01 -4.33503598e-01
-9.46858749e-02 1.00114226e+00 -3.15877289e-01 4.23463494e-01
-1.12418389e+00 -3.21428716e-01 1.42671531e-02 3.33262265e-01
-1.06790503e-02 2.66299337e-01 -2.40562215e-01 -9.09919441e-01
-1.17483869e-01 -1.29060030e+00 2.69195348e-01 -1.05739307e+00
-1.07537150e+00 8.95236731e-01 -1.23223610e-01 -1.42556643e+00
-1.20778613e-01 -2.26859912e-01 -7.32118249e-01 9.09841180e-01
-1.42081571e+00 -1.30268848e+00 -7.28790283e-01 6.43748224e-01
4.45708662e-01 -1.53563976e-01 6.10122681e-01 5.16897678e-01
-6.87557399e-01 8.15281987e-01 2.03173608e-01 1.52924821e-01
8.09655428e-01 -1.05202091e+00 -2.47627068e-02 1.08557487e+00
-4.12678868e-01 2.83506244e-01 6.21745884e-01 -8.35945487e-01
-8.68866742e-01 -1.16989350e+00 1.14194013e-01 1.22410417e-01
2.65706718e-01 4.31099012e-02 -8.64256620e-01 3.24870825e-01
3.60125184e-01 1.86512560e-01 5.79791903e-01 -8.63411665e-01
3.15770686e-01 -2.20924586e-01 -1.67444789e+00 4.48288411e-01
6.91681325e-01 -1.76603973e-01 -2.81353176e-01 2.30421886e-01
5.61142385e-01 -7.11891472e-01 -1.03518379e+00 7.05595732e-01
3.37109536e-01 -7.89552271e-01 8.90871167e-01 -3.55590172e-02
4.82777447e-01 -4.40196335e-01 9.45954919e-02 -1.42629766e+00
-5.46807982e-02 -3.28644216e-01 1.81236058e-01 1.16815257e+00
1.48988724e-01 -6.65221334e-01 8.64401579e-01 1.75823525e-01
-3.64526421e-01 -9.47456300e-01 -7.70426810e-01 -6.30471051e-01
4.67690714e-02 -3.73413190e-02 3.82390082e-01 9.04915214e-01
-5.68541765e-01 -1.60075843e-01 -3.24424148e-01 -9.18625221e-02
7.82211483e-01 1.70975327e-02 3.93146515e-01 -6.26059294e-01
-2.17416167e-01 -2.11112589e-01 -3.72755736e-01 -5.42489409e-01
-1.95281003e-02 -1.04936969e+00 1.34327397e-01 -1.52050376e+00
3.44227344e-01 -7.33729303e-01 -3.97227079e-01 5.17504990e-01
-4.54942673e-01 7.13428319e-01 1.13242187e-01 4.27828103e-01
-6.66364357e-02 5.62091887e-01 2.07083607e+00 -4.20299917e-01
2.00488776e-01 -4.06571291e-02 -6.25865936e-01 5.68397462e-01
1.06872869e+00 -6.13065064e-01 -5.39635539e-01 -4.22711810e-03
-5.12608171e-01 1.93333969e-01 5.33474684e-01 -1.15385520e+00
-1.14760734e-01 -6.85016587e-02 5.07924736e-01 -4.67248529e-01
1.70792401e-01 -7.31996953e-01 5.04369676e-01 8.46056342e-01
-1.78516686e-01 -1.12592392e-01 6.72305450e-02 3.75802159e-01
-4.12525892e-01 -2.62554079e-01 1.31150258e+00 -2.69951224e-01
-4.46477979e-01 3.15923631e-01 -9.43632647e-02 2.12966993e-01
1.21032023e+00 -2.08349898e-01 -3.60398352e-01 -3.26931268e-01
-6.71793222e-01 6.19793907e-02 5.17423928e-01 1.22095682e-01
7.87684202e-01 -1.43117523e+00 -7.97561049e-01 1.77106202e-01
-1.18872568e-01 -6.43850192e-02 6.25295997e-01 1.33897913e+00
-8.05586815e-01 4.68361787e-02 -4.59245533e-01 -5.34116626e-01
-1.12156320e+00 4.15666223e-01 6.93581760e-01 -4.37965691e-01
-5.39656341e-01 5.97486794e-01 4.15116966e-01 -1.24855423e-02
9.05175228e-03 -1.41460121e-01 -3.39594781e-02 -3.67030591e-01
3.05285513e-01 2.06595317e-01 2.36639492e-02 -7.51214623e-01
-2.37589225e-01 4.39294100e-01 -4.12081443e-02 -6.12922795e-02
9.62969661e-01 -3.96997482e-02 9.06312987e-02 6.39023557e-02
9.85034645e-01 -4.70987186e-02 -1.09150338e+00 -3.12896110e-02
-6.35716856e-01 -5.32108843e-01 2.70047486e-01 -1.09338450e+00
-1.73220706e+00 7.14394033e-01 1.18019688e+00 -4.24123332e-02
1.50573671e+00 -2.05531240e-01 1.31871307e+00 -6.46772325e-01
1.90912589e-01 -8.02635968e-01 1.69745520e-01 -3.39884311e-01
9.55184221e-01 -1.20975792e+00 -1.19726785e-01 -6.21008039e-01
-9.07502651e-01 6.72672868e-01 6.37151778e-01 -2.40115114e-02
2.56746501e-01 2.97600806e-01 4.48214799e-01 -1.30326739e-02
-1.71621054e-01 -2.09808022e-01 3.22209716e-01 8.03910375e-01
2.69428045e-01 1.07360177e-01 -6.74061596e-01 5.66927314e-01
1.71129778e-02 -3.74450013e-02 4.35761631e-01 8.51196289e-01
-8.70857835e-02 -1.20797658e+00 -4.75624233e-01 4.61650461e-01
-8.33428025e-01 2.79704866e-04 -1.08145416e-01 8.52134466e-01
3.74584138e-01 8.68536294e-01 -3.73852313e-01 -4.48176980e-01
2.68169284e-01 -1.91615492e-01 4.24210519e-01 -3.86387229e-01
-4.99599218e-01 3.44516277e-01 -2.35709757e-01 -4.75864381e-01
-5.26788175e-01 -3.57823491e-01 -1.28959739e+00 -1.25278622e-01
-3.08050305e-01 1.41021192e-01 5.59873402e-01 5.44502497e-01
5.41524589e-02 9.87788081e-01 7.74490178e-01 -4.96864349e-01
-2.97864676e-01 -9.41546857e-01 -7.53263295e-01 6.18808687e-01
6.33426905e-02 -6.67357564e-01 -3.19478273e-01 2.94129908e-01]
|
[13.880086898803711, -2.1758463382720947]
|
24e343be-ae02-4eec-9f68-7a3b3d567c01
|
using-deep-convolutional-neural-networks-to-2
| null | null |
https://www.medrxiv.org/content/10.1101/2022.10.03.22280640v3
|
https://www.medrxiv.org/content/10.1101/2022.10.03.22280640v3.full.pdf
|
Using deep convolutional neural networks to predict patients age based on ECGs from an independent test cohort
|
Electrocardiography is one of the most frequently used methods to evaluate cardiovascular diseases. However, the last decade has shown that deep convolutional neural networks (CNN) can extract information from the electrocardiogram (ECG) that goes beyond traditional diagnostics, such as predicting a persons age. In this study, we trained two different 1-dimensional CNNs on open datasets to predict age from a persons ECG.
The models were trained and validated using 10 seconds long 12-lead ECG records, resampled to 100Hz. 59355 ECGs were used for training and cross-validation, while 21748 ECGs from a separate cohort were used as the test set. We compared the performance achieved on the cross-validation with the performance on the test set. Furthermore, we used cardiologist-annotated cardiovascular conditions to categorize the patients in the test set in order to assess whether some cardiac condition leads to greater discrepancies between CNN-predicted age and chronological age.
The best CNN model, using an Inception Time architecture, showed a significant drop in performance, in terms of mean absolute error (MAE), from cross-validation on the training set (7.90 ± 0.04 years) to the performance on the test set (8.3 years). On the other hand, the mean squared error (MSE) improved from the training set (117.5 ± 2.7 years^2) to the test set (111 years^2). We also observed that the cardiovascular condition that showed the highest deviation between predicted and biological age, in terms of MAE, was the patients with pacing rhythm (10.5 years), while the patients with prolonged QT-interval had the smallest deviation (7.4 years) in terms of MAE.
This work contributes to existing knowledge of age prediction using deep CNNs on ECGs by showing how a trained model performs on a test set from a separate cohort to that used in the training set.
|
['Belal Tavashi', 'Bjørn-Jostein Singstad']
|
2022-10-06
| null | null | null |
medrxiv-2022-10
|
['age-estimation', 'age-estimation']
|
['computer-vision', 'miscellaneous']
|
[-5.30012026e-02 7.49881715e-02 4.01332736e-01 -4.09987122e-01
-4.40486461e-01 -3.67354572e-01 -3.65317389e-02 5.44975996e-01
-7.55567312e-01 8.89309585e-01 -2.09311798e-01 -4.92533088e-01
-2.64372498e-01 -9.41895843e-01 -3.60358834e-01 -7.35779285e-01
-6.07926369e-01 4.92917448e-01 -3.26155275e-01 1.07531538e-02
-8.88779163e-02 4.10118997e-01 -1.09763658e+00 -1.66554395e-02
8.39805722e-01 1.29439390e+00 -5.12588918e-01 8.09503078e-01
5.39825022e-01 1.95739210e-01 -9.34247375e-01 -2.57796973e-01
2.74641871e-01 -6.95069909e-01 -4.27092969e-01 -4.59877819e-01
5.70533752e-01 -2.22393870e-01 -9.40300673e-02 3.27004313e-01
1.14443409e+00 -3.99139404e-01 4.29174483e-01 -7.80167341e-01
-3.11193198e-01 5.71781158e-01 -1.06886610e-01 4.15224791e-01
1.01893814e-02 1.63912863e-01 4.07539725e-01 -2.67502785e-01
3.83605391e-01 5.67062259e-01 1.28487074e+00 6.10447109e-01
-1.10372627e+00 -6.43449664e-01 -5.30520678e-01 -7.34601840e-02
-1.46037972e+00 -8.34493116e-02 4.14478779e-01 -6.09492362e-01
6.45250857e-01 2.27057159e-01 1.19661713e+00 9.30091739e-01
5.67768514e-01 -1.05575994e-01 1.09968603e+00 -2.90225297e-01
2.31973737e-01 -1.12484463e-01 1.72701716e-01 3.44228387e-01
4.43130225e-01 4.29470062e-01 -5.37687950e-02 -2.83474833e-01
8.26051176e-01 -1.75692245e-01 -2.48675928e-01 3.46628547e-01
-1.12956774e+00 5.31001389e-01 1.85143501e-01 4.94263470e-01
-4.95327801e-01 4.69606966e-02 5.53071439e-01 5.87146521e-01
4.63321030e-01 7.63032138e-01 -1.06465197e+00 -3.60595554e-01
-9.84735489e-01 3.01281899e-01 8.48850250e-01 -3.67037728e-02
1.76949918e-01 2.14610115e-01 -1.25250295e-01 6.06711626e-01
-7.49999434e-02 5.17917395e-01 6.17765129e-01 -5.69584072e-01
1.22798160e-01 6.16331816e-01 -8.62054080e-02 -7.71965146e-01
-8.88167083e-01 -1.20525765e+00 -1.19949067e+00 1.10450804e-01
1.02352524e+00 -5.96777678e-01 -8.90653789e-01 1.75753403e+00
-1.57109827e-01 -1.48069516e-01 9.14533809e-02 8.63860309e-01
8.62455785e-01 9.87368897e-02 1.58626512e-01 -1.75917968e-01
1.47598982e+00 -7.52226962e-03 -2.00032249e-01 7.70633295e-02
8.39663863e-01 -3.17613035e-01 5.38897872e-01 5.98168075e-01
-1.05487537e+00 -7.67360151e-01 -1.13567126e+00 4.57114607e-01
-2.09251285e-01 1.86415017e-01 2.59815186e-01 1.03751016e+00
-1.08655214e+00 1.12795842e+00 -6.70891464e-01 -5.83400667e-01
5.04417062e-01 3.74612302e-01 -2.80626327e-01 3.19669187e-01
-1.73288310e+00 9.68238592e-01 2.63877541e-01 1.35629196e-02
-7.16219127e-01 -1.13325644e+00 -4.89976019e-01 3.79547887e-02
-2.64823765e-01 -9.45978463e-01 6.15521967e-01 -8.75515640e-01
-9.30634499e-01 1.04542482e+00 4.06947494e-01 -9.27077234e-01
7.93349028e-01 -3.51649970e-01 -5.64436078e-01 -8.92614760e-03
-1.60772502e-01 3.51719350e-01 4.39790249e-01 -5.83172083e-01
-2.61110842e-01 -6.83320642e-01 -2.98726648e-01 -3.62255245e-01
-1.40653402e-01 -9.64371860e-02 1.16316088e-01 -6.20038509e-01
1.01481043e-01 -9.78567779e-01 -2.20460624e-01 -2.05045566e-01
-1.71242595e-01 -8.23587179e-03 3.10110092e-01 -9.87099707e-01
1.41454983e+00 -1.94888186e+00 -9.41885188e-02 4.08016384e-01
7.48149693e-01 3.21748167e-01 3.89774710e-01 3.79993051e-01
-5.86954653e-01 3.09933573e-01 -2.31892109e-01 3.35096419e-02
-6.29948676e-01 -5.66372201e-02 3.05987984e-01 4.87634987e-01
2.53878325e-01 7.51789033e-01 -6.20748997e-01 -2.81101018e-01
1.33054587e-03 5.43409526e-01 -1.41423985e-01 -1.06592597e-02
4.32467997e-01 8.40595305e-01 -4.90280166e-02 3.40890229e-01
4.72848386e-01 -8.30113366e-02 9.70417261e-02 -7.06148744e-02
-9.42353830e-02 -1.49066329e-01 -7.06973314e-01 1.50128210e+00
-2.67488480e-01 6.82348251e-01 -5.52262068e-01 -8.76725256e-01
1.46241260e+00 6.89746141e-01 7.88309455e-01 -7.03001678e-01
3.54290456e-01 3.65031868e-01 8.94912660e-01 -3.81772131e-01
-2.27351367e-01 -3.48175526e-01 -3.43931615e-02 3.91996682e-01
-1.19592398e-01 1.91576630e-01 1.41060710e-01 -8.66231248e-02
1.18588436e+00 -2.57853478e-01 1.92300439e-01 -4.02508706e-01
4.34968978e-01 -3.51053268e-01 8.09826970e-01 7.00242281e-01
-3.25806320e-01 9.19991732e-01 8.22742403e-01 -1.24816382e+00
-1.15359962e+00 -1.01491833e+00 -6.03943706e-01 4.01235707e-02
-6.30763769e-01 -4.91590500e-01 -8.76215279e-01 -5.11467040e-01
2.42588669e-02 2.35859856e-01 -8.28525186e-01 -3.13213348e-01
-5.02646327e-01 -1.03466451e+00 1.15332198e+00 7.35576987e-01
4.40566242e-01 -1.07244575e+00 -1.16617107e+00 3.05119574e-01
-8.73242617e-02 -6.08158171e-01 1.45262912e-01 2.38851547e-01
-1.15377617e+00 -1.18035698e+00 -9.34997261e-01 -3.55277330e-01
3.29079479e-01 -1.02088940e+00 1.42394590e+00 4.08634514e-01
-5.43328404e-01 1.11772222e-02 -1.52675807e-01 -1.01610994e+00
-4.10477489e-01 2.75703043e-01 2.44372785e-01 -1.01964794e-01
3.28909874e-01 -7.68514693e-01 -1.02018356e+00 9.83288735e-02
-3.98546278e-01 -2.82653630e-01 5.23816884e-01 8.95754278e-01
4.38733757e-01 -3.18449467e-01 1.14476717e+00 -5.18574297e-01
4.45862889e-01 -3.68997395e-01 -3.58446479e-01 -7.06145763e-02
-1.19285548e+00 -2.66597092e-01 5.31639218e-01 -3.39560658e-01
-1.52763769e-01 -9.56607684e-02 -2.02780753e-01 -2.67224729e-01
-3.87511432e-01 6.70350194e-01 2.57690400e-01 3.30290824e-01
9.89235997e-01 -7.78247342e-02 3.13018113e-01 -4.06870693e-01
-4.79649901e-01 4.32819456e-01 5.19356787e-01 -4.83697593e-01
3.45610023e-01 7.67761469e-02 4.32080150e-01 -6.58273637e-01
-5.22391796e-01 3.07321791e-02 -8.47431719e-01 -3.06637406e-01
9.70171988e-01 -7.58803904e-01 -7.53636360e-01 8.20998788e-01
-8.12101185e-01 -4.03325707e-01 -3.94760698e-01 6.52805746e-01
-3.38195950e-01 9.49206576e-02 -5.75224757e-01 -7.46479452e-01
-1.03063595e+00 -6.66635692e-01 6.25890911e-01 3.48316878e-02
-8.10394347e-01 -1.11418366e+00 2.45436862e-01 -4.89601269e-02
4.76075828e-01 1.10765004e+00 1.07682025e+00 -8.49934340e-01
3.51686418e-01 -6.85409188e-01 1.39992433e-02 5.74035048e-01
1.04185633e-01 3.61533626e-03 -9.68404591e-01 -2.35891536e-01
-6.39828667e-02 -3.06390300e-02 6.63018286e-01 6.83039188e-01
1.13222325e+00 3.71192634e-01 -4.50555161e-02 5.08701205e-01
1.29700887e+00 5.38773239e-01 1.12434030e+00 3.87289107e-01
4.53364164e-01 2.89840788e-01 1.46806762e-01 4.35044080e-01
5.83881587e-02 3.77010465e-01 2.39968330e-01 -5.49249113e-01
1.11976497e-01 3.03644329e-01 -1.33184925e-01 3.77666235e-01
-6.50567234e-01 2.08960056e-01 -1.34965467e+00 4.82399970e-01
-1.44605625e+00 -6.83317304e-01 -5.22968948e-01 2.59371138e+00
6.85021818e-01 3.80150944e-01 4.31133538e-01 7.13539481e-01
4.98275101e-01 -3.39804381e-01 -5.12475371e-01 -6.54260933e-01
-1.50669336e-01 6.52082622e-01 3.36311281e-01 -1.43088013e-01
-9.76731718e-01 8.77367780e-02 6.20015764e+00 -3.16194519e-02
-1.56257784e+00 -2.29488209e-01 1.18956161e+00 -7.65901431e-02
5.23934960e-01 -2.43442431e-01 -1.14810921e-01 6.54649019e-01
1.69738245e+00 -9.03271511e-02 -3.67238522e-02 4.84932452e-01
3.34499776e-01 -3.36947776e-02 -1.26679182e+00 1.01301193e+00
-9.64590237e-02 -9.25064147e-01 -5.90433359e-01 1.41651705e-01
4.42705065e-01 -8.13455805e-02 -6.88030496e-02 2.08031535e-01
-5.38429379e-01 -1.36348748e+00 5.11944592e-01 8.08627129e-01
1.29935753e+00 -7.32120275e-01 1.38675094e+00 2.10716262e-01
-8.11622322e-01 -1.55060012e-02 2.00415447e-01 -5.03984690e-01
3.12346648e-02 9.56278920e-01 -8.41450930e-01 6.02552891e-01
1.05564892e+00 7.68582821e-01 -7.31224656e-01 1.05462790e+00
1.69555858e-01 8.12438190e-01 -1.24218337e-01 2.83776402e-01
-1.78980619e-01 -4.89019305e-02 2.85535455e-01 8.26406419e-01
4.77648735e-01 -5.40499873e-02 -1.64467111e-01 7.32179880e-01
1.40018910e-01 9.32637379e-02 -3.02494824e-01 5.41995168e-02
1.53177917e-01 1.03516686e+00 -6.71383262e-01 -4.66202408e-01
-1.47454277e-01 4.52954710e-01 -1.41785949e-01 6.15051351e-02
-8.19620550e-01 -8.35304141e-01 3.65896732e-01 4.49167848e-01
-1.60878748e-01 2.35662997e-01 -7.46084154e-01 -7.45999575e-01
-1.58714727e-02 -7.67930269e-01 5.05220234e-01 -5.31250000e-01
-1.29525638e+00 6.94941521e-01 -1.30555853e-01 -1.23209059e+00
-2.91747928e-01 -3.88760448e-01 -7.74273634e-01 1.53451443e+00
-8.47465277e-01 -5.16461253e-01 -4.07203436e-01 1.89471141e-01
2.66378354e-02 -5.34183271e-02 1.30456758e+00 6.18288279e-01
-4.94230956e-01 7.62265921e-01 -2.38519788e-01 5.99652350e-01
8.28208685e-01 -1.43576598e+00 3.22009146e-01 4.07308161e-01
-3.68140578e-01 7.07582295e-01 4.78496552e-01 -5.77476382e-01
-6.55789077e-01 -1.06406498e+00 1.13848615e+00 -5.30450642e-01
2.29432315e-01 1.34629577e-01 -8.76669943e-01 2.69098103e-01
-1.34542808e-01 5.38642406e-02 9.56953943e-01 3.77323151e-01
-1.67684108e-01 -3.04551065e-01 -1.27541232e+00 1.77874655e-01
5.22560120e-01 -2.20222294e-01 -4.07864600e-01 -2.13362992e-01
1.15169071e-01 -5.46309233e-01 -1.68216789e+00 6.99453592e-01
1.31579244e+00 -1.08799744e+00 8.44172478e-01 -6.80905163e-01
6.12335920e-01 -2.84345336e-02 4.49988961e-01 -1.30122721e+00
-2.19704375e-01 -2.74583578e-01 6.24796860e-02 8.71273696e-01
6.34855032e-01 -9.07906294e-01 7.59158075e-01 4.55699980e-01
-1.28146812e-01 -1.50663757e+00 -7.75862038e-01 -7.31657624e-01
4.88695443e-01 -3.66467178e-01 5.61832428e-01 7.91369021e-01
-3.59150827e-01 2.11683929e-01 -2.22852185e-01 -1.52346000e-01
3.76261890e-01 -1.60220340e-01 5.20587265e-01 -1.87478399e+00
-2.51367897e-01 -1.96555972e-01 -8.18690181e-01 1.59780219e-01
-3.95188302e-01 -7.75955558e-01 -3.27504992e-01 -1.57640004e+00
1.13849435e-02 -5.92174828e-01 -8.77478182e-01 4.16609764e-01
-1.40454814e-01 4.30613756e-01 1.53220326e-01 -6.72485903e-02
1.84694976e-01 -1.54041380e-01 1.07582331e+00 9.13244486e-02
-4.66244787e-01 2.51468807e-01 -5.58274388e-01 5.72895408e-01
1.29357982e+00 -5.41114748e-01 -2.06463590e-01 -6.12773672e-02
3.06091547e-01 3.44985068e-01 3.90315175e-01 -1.59508598e+00
-4.88340586e-01 4.81651455e-01 1.15140808e+00 -4.38853294e-01
5.63645512e-02 -4.17924941e-01 4.33409423e-01 1.15719354e+00
-3.05841714e-01 3.83930147e-01 2.54718125e-01 1.37720659e-01
9.06032324e-02 1.59142926e-01 5.68482935e-01 -1.71608970e-01
-1.49948988e-02 3.86004478e-01 -4.43409145e-01 1.47016808e-01
8.30982089e-01 -4.58731234e-01 -2.56376178e-03 -2.91685253e-01
-1.31131268e+00 -8.93551037e-02 2.32592463e-01 1.50475621e-01
5.06202579e-01 -1.04075575e+00 -1.05831003e+00 1.60745457e-01
1.92216665e-01 -2.25826815e-01 3.60800952e-01 1.30514348e+00
-6.74933255e-01 2.70080328e-01 -5.82759023e-01 -6.67703807e-01
-1.35836196e+00 1.55242160e-01 7.20557034e-01 -3.48213673e-01
-6.05687261e-01 5.56241214e-01 -3.54830325e-01 -5.09584174e-02
1.61266178e-01 -5.35215616e-01 -2.91004390e-01 1.72554851e-01
4.37286198e-01 5.77246428e-01 4.03463513e-01 -2.93911964e-01
-4.46133912e-01 7.09145546e-01 1.78280756e-01 1.95161149e-01
1.28200603e+00 2.41348118e-01 -1.93558082e-01 7.47099817e-01
1.04404891e+00 -2.94848412e-01 -7.52832413e-01 3.28593224e-01
-1.75760508e-01 -9.07766167e-03 -2.04254687e-01 -1.18515110e+00
-1.29502726e+00 1.02769578e+00 1.42328727e+00 3.82852376e-01
1.09207320e+00 -2.72132128e-01 6.89215481e-01 2.14161538e-03
2.91144609e-01 -8.14978004e-01 -2.67222911e-01 3.43145519e-01
6.62907481e-01 -8.36962104e-01 -1.56612061e-02 6.17514849e-02
-4.79819238e-01 1.32751811e+00 5.35202682e-01 -2.14044571e-01
8.17638993e-01 -1.20230682e-01 4.14613008e-01 -3.60264480e-01
-5.15786588e-01 2.59377122e-01 1.21601947e-01 7.29656458e-01
8.44774902e-01 2.04214722e-01 -7.01414645e-01 8.75553489e-01
-5.09517074e-01 3.23296845e-01 5.36760211e-01 6.48206472e-01
7.60059431e-02 -1.05272174e+00 -2.66237527e-01 9.22155917e-01
-9.05075014e-01 -9.75225642e-02 -2.96677709e-01 7.14125335e-01
6.09395683e-01 8.20054531e-01 2.96753943e-01 -4.46616709e-01
3.08255404e-01 5.55346549e-01 4.14837956e-01 -4.35463905e-01
-1.15721393e+00 -2.24160150e-01 1.59648612e-01 -1.87115029e-01
-1.99020430e-01 -6.21398747e-01 -1.17779005e+00 -1.31379589e-01
-1.40158772e-01 2.14460179e-01 6.68089747e-01 7.04138637e-01
3.40404004e-01 1.00439572e+00 3.42441440e-01 -2.74311781e-01
-3.53991091e-01 -1.21147573e+00 -7.30500877e-01 3.47151577e-01
3.03640515e-01 -2.48073786e-01 -2.51188487e-01 9.95983258e-02]
|
[14.321404457092285, 3.2962570190429688]
|
d7bfd61e-b106-4290-98d4-c27166df2f4a
|
confound-leakage-confound-removal-in-machine
|
2210.09232
| null |
https://arxiv.org/abs/2210.09232v2
|
https://arxiv.org/pdf/2210.09232v2.pdf
|
Confound-leakage: Confound Removal in Machine Learning Leads to Leakage
|
Machine learning (ML) approaches to data analysis are now widely adopted in many fields including epidemiology and medicine. To apply these approaches, confounds must first be removed as is commonly done by featurewise removal of their variance by linear regression before applying ML. Here, we show this common approach to confound removal biases ML models, leading to misleading results. Specifically, this common deconfounding approach can leak information such that what are null or moderate effects become amplified to near-perfect prediction when nonlinear ML approaches are subsequently applied. We identify and evaluate possible mechanisms for such confound-leakage and provide practical guidance to mitigate its negative impact. We demonstrate the real-world importance of confound-leakage by analyzing a clinical dataset where accuracy is overestimated for predicting attention deficit hyperactivity disorder (ADHD) with depression as a confound. Our results have wide-reaching implications for implementation and deployment of ML workflows and beg caution against na\"ive use of standard confound removal approaches.
|
['Kaustubh R. Patil', 'Simon B. Eickhoff', 'Holger Schwender', 'Susanne Weis', 'Georg G. von Polier', 'Bradley C. Love', 'Sami Hamdan']
|
2022-10-17
| null | null | null | null |
['epidemiology']
|
['medical']
|
[ 5.85585535e-01 1.19393930e-01 -4.83073413e-01 -5.67073405e-01
-8.36759984e-01 -5.50041974e-01 4.02177989e-01 6.04179084e-01
-4.72841144e-01 8.68515968e-01 5.12790084e-01 -9.57512736e-01
-2.92165726e-01 -2.74143875e-01 -8.49809587e-01 -3.55783761e-01
-1.56839296e-01 8.23225901e-02 -6.02824509e-01 4.61153507e-01
1.59744918e-01 4.92854983e-01 -1.14042950e+00 2.98350781e-01
1.04155993e+00 -1.91384498e-02 -2.00676799e-01 5.43014705e-01
-1.89789116e-01 8.37371349e-01 -8.38751197e-01 -4.24272358e-01
8.96620154e-02 -4.17639256e-01 -2.82207042e-01 -3.58366072e-01
5.58575869e-01 -4.95049804e-01 2.48249620e-02 6.27484560e-01
6.55427635e-01 -2.87093997e-01 8.09165835e-01 -1.31866050e+00
-4.58515108e-01 7.51027048e-01 -8.23210001e-01 3.05322826e-01
1.51443720e-01 6.55863225e-01 5.06576478e-01 -6.98350728e-01
5.19925117e-01 1.28822434e+00 1.02180505e+00 2.89331287e-01
-1.65628064e+00 -1.12612939e+00 3.57344866e-01 -2.30871633e-01
-1.16690612e+00 -6.65638804e-01 -2.66152378e-02 -9.62613404e-01
1.05985022e+00 3.61214370e-01 5.06497860e-01 8.71928215e-01
5.02885401e-01 2.73721427e-01 1.28327858e+00 -3.03945243e-01
1.59236386e-01 3.53606790e-02 3.06540608e-01 3.23220491e-01
8.50460589e-01 4.38628122e-02 -4.37551916e-01 -7.77004004e-01
5.45278609e-01 1.28721133e-01 2.38344278e-02 1.65610522e-01
-9.21632349e-01 7.52957344e-01 1.06946059e-01 -2.64325384e-02
-3.71379107e-01 8.94218981e-02 3.65619928e-01 -8.19501430e-02
5.61208546e-01 4.90968645e-01 -7.31560588e-01 -5.86883277e-02
-1.19727516e+00 4.45930988e-01 4.70197558e-01 5.59762895e-01
1.48065165e-01 -1.47318289e-01 -3.05005461e-01 8.61220598e-01
3.03597569e-01 6.42993450e-01 1.88849598e-01 -7.87187934e-01
4.31303889e-01 6.79733932e-01 1.31728724e-01 -7.50602782e-01
-8.83916080e-01 -5.88451564e-01 -7.03497231e-01 1.51188090e-01
5.05894125e-01 -4.94115591e-01 -7.51123250e-01 1.78806210e+00
3.32433373e-01 3.55630368e-01 -2.86897600e-01 4.81853873e-01
7.45009661e-01 -2.50937313e-01 7.85463691e-01 -3.69949728e-01
1.39645004e+00 -2.46999770e-01 -9.10923898e-01 -2.82761186e-01
1.06487155e+00 -6.73923850e-01 8.10693383e-01 4.86149490e-01
-1.12358415e+00 -1.86580159e-02 -7.77471304e-01 -8.83786101e-03
-2.50123233e-01 -1.33518904e-01 6.76017880e-01 9.09850836e-01
-5.18132627e-01 6.41707659e-01 -8.16431344e-01 -1.29175529e-01
9.40762162e-01 5.98508656e-01 -2.15413824e-01 -1.22804075e-01
-7.94543386e-01 9.60417688e-01 -9.66244265e-02 1.40545994e-01
-4.61499155e-01 -1.72542584e+00 -6.07483745e-01 -3.33855972e-02
2.75246352e-01 -7.90163517e-01 9.95342970e-01 -6.48810685e-01
-5.64719737e-01 7.93401837e-01 -5.35773456e-01 -3.07616442e-01
4.70262140e-01 -4.04044241e-01 -5.11820912e-01 -3.44403118e-01
2.27304369e-01 2.60323733e-01 5.83897948e-01 -8.41573596e-01
-6.14876807e-01 -5.78685760e-01 -4.62016970e-01 -8.10774043e-03
-1.75011493e-02 4.98138040e-01 4.13696676e-01 -5.91075420e-01
-1.55821919e-01 -5.52298009e-01 -5.05931139e-01 -9.49211717e-02
-3.10746372e-01 1.35661334e-01 2.84313172e-01 -7.30796218e-01
1.63130832e+00 -1.77996886e+00 -4.90584701e-01 -2.16989759e-02
7.06818044e-01 5.09326756e-01 -2.72243563e-02 4.11532849e-01
-3.26792628e-01 5.32178998e-01 -2.37209857e-01 2.54837386e-02
-1.82282776e-01 -5.00750029e-03 -1.73693866e-01 7.70082474e-01
7.40818143e-01 1.07678628e+00 -9.38363433e-01 -2.25582141e-02
3.69876683e-01 6.03428066e-01 -7.55837083e-01 8.53158236e-02
1.32816628e-01 4.17385131e-01 -3.62091623e-02 5.27132869e-01
9.65052605e-01 -2.18887374e-01 4.63367641e-01 1.18562929e-01
-3.49562198e-01 5.82440913e-01 -1.00380182e+00 7.19670057e-01
-1.39843658e-01 3.60660166e-01 1.47741213e-01 -6.53277636e-01
6.38608754e-01 3.95421721e-02 4.15674061e-01 -2.97860503e-01
-1.03795424e-01 1.79137260e-01 3.78848255e-01 -5.97113192e-01
-2.51900971e-01 -4.44022089e-01 2.16752619e-01 2.41512120e-01
-3.59093100e-01 9.72438976e-03 -4.16540980e-01 -9.65421200e-02
1.41072345e+00 -9.71946493e-02 7.36907184e-01 -3.53161395e-01
1.52945876e-01 1.35710970e-01 6.14665508e-01 1.07344413e+00
-5.74847162e-02 5.28637409e-01 8.60317528e-01 -1.16840214e-01
-6.29984200e-01 -9.34865475e-01 -6.33927882e-01 6.42024517e-01
-9.32529211e-01 -4.40675348e-01 -3.73178244e-01 -6.53227329e-01
5.62715471e-01 9.75084960e-01 -7.45636642e-01 -4.50097740e-01
-3.83796096e-01 -1.40853333e+00 9.79786396e-01 4.26137745e-01
-2.77549952e-01 -5.82090437e-01 -7.20622778e-01 3.94979119e-01
3.24651212e-01 -7.74203181e-01 -5.32906912e-02 4.40196007e-01
-9.34469163e-01 -1.54842913e+00 -4.89974171e-01 6.10289052e-02
6.44290328e-01 1.19479932e-01 9.72345293e-01 1.25990629e-01
-6.87601566e-01 3.20226222e-01 1.12607665e-01 -9.44622755e-01
-4.40885931e-01 -2.98830122e-01 3.78013477e-02 -4.67597634e-01
8.82623792e-01 -3.61726046e-01 -8.54219437e-01 -2.25685686e-01
-7.69484699e-01 -2.99073905e-01 6.31311953e-01 6.40825152e-01
2.64627695e-01 -3.72933269e-01 1.06590617e+00 -1.26242530e+00
8.06298554e-01 -9.62211668e-01 -5.75419784e-01 -1.86461046e-01
-1.05262852e+00 -3.09400469e-01 3.84453058e-01 -5.87669969e-01
-8.14773262e-01 -2.41469920e-01 9.96056199e-02 -2.29750529e-01
-4.90406364e-01 7.12968767e-01 -1.46933750e-03 2.62409627e-01
7.19099581e-01 -4.51941252e-01 1.29937902e-01 -6.20125234e-01
1.30323723e-01 6.14429533e-01 1.69688642e-01 -2.15919495e-01
3.64555359e-01 4.81939822e-01 2.78789371e-01 -7.79939473e-01
-7.61465251e-01 -3.54438037e-01 -5.32644212e-01 1.77563518e-01
6.72815561e-01 -1.00248098e+00 -8.47354591e-01 1.22618154e-01
-8.56445968e-01 -4.98443514e-01 -7.03209415e-02 7.53348053e-01
4.33787368e-02 -8.66525546e-02 -1.94054227e-02 -1.00601768e+00
-2.40046471e-01 -9.61634159e-01 7.77614236e-01 1.12370074e-01
-7.86941051e-01 -9.40181792e-01 3.67919773e-01 3.72959524e-01
2.22716600e-01 3.86066258e-01 1.38345230e+00 -9.20013845e-01
7.90443048e-02 9.19197872e-02 -4.00500804e-01 6.71803877e-02
5.65224349e-01 2.14070871e-01 -1.23690641e+00 -7.05349520e-02
-1.97746500e-01 8.64735916e-02 4.86462057e-01 8.43534768e-01
1.02480567e+00 -3.69558364e-01 -4.27604407e-01 5.31157494e-01
1.21606970e+00 3.19213718e-01 1.59155503e-01 -2.55829077e-02
7.75147140e-01 7.34362721e-01 3.14929754e-01 4.79411423e-01
4.76475984e-01 5.70226252e-01 -1.35797143e-01 -5.03272951e-01
-1.64037168e-01 -2.99948245e-01 1.65588602e-01 1.29690707e-01
2.33784094e-01 2.36675918e-01 -1.12332940e+00 3.89852524e-01
-1.57095325e+00 -7.43906081e-01 -9.20067608e-01 2.60574365e+00
9.38295186e-01 9.03646424e-02 3.18399191e-01 -2.73510396e-01
3.45148504e-01 -3.31030518e-01 -6.25113130e-01 -7.64574647e-01
-1.04220666e-01 2.44751453e-01 6.31567061e-01 2.83794224e-01
-8.17396879e-01 6.01469696e-01 8.12587070e+00 1.25204518e-01
-1.11276817e+00 1.77338481e-01 6.31552696e-01 -6.02678478e-01
-2.90290922e-01 -1.10464148e-01 -7.37491369e-01 4.38921154e-01
1.27468181e+00 -1.22299775e-01 1.38540566e-01 1.86880738e-01
9.62467790e-01 -3.78532618e-01 -1.19414198e+00 6.41528130e-01
-2.53418088e-01 -1.08232105e+00 -4.08327878e-01 1.75030738e-01
5.26403606e-01 -1.61712337e-03 1.28078997e-01 2.84688979e-01
5.10949314e-01 -1.49139690e+00 1.73408166e-01 5.25917590e-01
8.98781002e-01 -7.02143192e-01 6.29866779e-01 5.57340942e-02
-5.81691921e-01 6.06495654e-03 -3.46097052e-01 -4.84262526e-01
3.19136158e-02 1.19702435e+00 -1.48618877e+00 4.00541931e-01
4.24873978e-01 5.64697683e-01 -4.91270304e-01 1.13300443e+00
-5.11863455e-02 1.17801607e+00 -1.79019153e-01 3.97327125e-01
-1.79365613e-02 -4.28444631e-02 3.60147059e-01 1.60783792e+00
8.99323896e-02 1.79463729e-01 -4.15598720e-01 1.01574242e+00
1.33976340e-01 2.87232608e-01 -8.52625191e-01 -1.91560373e-01
4.27825361e-01 9.68468010e-01 -5.09265363e-01 -3.05428505e-01
-8.67817640e-01 2.65406430e-01 2.84393698e-01 3.47093284e-01
-6.59219444e-01 9.23590660e-02 9.98190224e-01 2.62499303e-01
-9.95391309e-02 2.76404828e-01 -8.13883662e-01 -8.41868758e-01
-2.38525823e-01 -1.19372594e+00 6.34669781e-01 -3.26402038e-01
-1.24215031e+00 -3.39379758e-01 2.01335743e-01 -7.56520987e-01
8.33290443e-02 -3.35647613e-01 -4.34007108e-01 1.29340804e+00
-1.27770114e+00 -8.29587281e-01 1.81756511e-01 9.47765335e-02
1.41873434e-01 2.50259429e-01 6.79856777e-01 3.40643227e-01
-8.21261108e-01 6.47259474e-01 1.66757442e-02 -3.35604191e-01
1.26989269e+00 -1.15039766e+00 1.88250050e-01 5.71053267e-01
-3.61186802e-01 1.08513522e+00 6.52818084e-01 -1.23935044e+00
-1.13801467e+00 -1.20865560e+00 1.02167356e+00 -7.74689972e-01
7.09370017e-01 -4.29022878e-01 -1.10880637e+00 8.70944023e-01
-2.37831637e-01 -5.24735510e-01 1.21874106e+00 5.08670568e-01
-4.42980528e-01 9.63595957e-02 -1.25786746e+00 7.85005689e-01
9.02283907e-01 -1.53398857e-01 -6.47182107e-01 2.89086252e-01
4.08308893e-01 -1.04595102e-01 -1.02012980e+00 5.49589396e-01
7.92328358e-01 -7.54248798e-01 8.72716427e-01 -1.24207497e+00
4.77325439e-01 -5.49053773e-02 2.18646064e-01 -1.12345922e+00
-1.76852658e-01 -5.24152100e-01 7.16071129e-02 1.11318016e+00
5.04266441e-01 -7.79349029e-01 3.64033937e-01 1.12195158e+00
1.45211682e-01 -6.45673394e-01 -6.37342930e-01 -3.62120956e-01
4.71494317e-01 -6.33342087e-01 6.87546492e-01 1.28140664e+00
8.06232393e-02 4.12200093e-02 -2.58673251e-01 3.76407474e-01
5.50460577e-01 -4.63999629e-01 7.28639781e-01 -1.20648289e+00
-1.38790607e-01 -4.42647010e-01 -7.56225586e-02 -1.52247608e-01
-8.02722052e-02 -9.66364682e-01 -3.76774937e-01 -1.38287818e+00
5.37265241e-01 -5.27596176e-01 -4.00904387e-01 6.45069957e-01
-6.86205864e-01 -1.01892799e-01 6.02308288e-02 -7.77077079e-02
-5.14926314e-02 -6.18430227e-02 9.27190185e-01 1.32982120e-01
-5.96673727e-01 1.36605958e-02 -1.23208714e+00 7.24319994e-01
7.15207398e-01 -9.36176300e-01 -4.03457880e-01 -1.54096857e-01
2.92325795e-01 -1.84728771e-01 6.06139004e-01 -3.71332735e-01
-1.85033828e-01 -4.77274507e-01 7.78569818e-01 -4.99258786e-01
-2.08011001e-01 -7.87816703e-01 3.96116883e-01 7.80618012e-01
-4.64325070e-01 1.57935753e-01 7.68914044e-01 1.07374653e-01
5.44896603e-01 -1.74582273e-01 5.10402083e-01 1.05221294e-01
2.99145401e-01 -2.41840988e-01 -7.41409481e-01 2.74968445e-01
7.60503888e-01 2.65764773e-01 -5.81111193e-01 -2.03814954e-01
-6.47261322e-01 2.56394744e-01 3.93579304e-02 3.82614821e-01
3.99198085e-01 -7.42865682e-01 -9.36641037e-01 3.14573765e-01
-7.93481097e-02 -3.24876070e-01 3.92764628e-01 1.39069724e+00
-2.56899148e-02 6.07263505e-01 1.36712894e-01 -3.44666779e-01
-1.56314683e+00 6.18221998e-01 1.88362658e-01 -5.09976074e-02
-7.49052823e-01 4.11939025e-01 4.53539938e-01 -3.71074021e-01
1.53573945e-01 -5.62079370e-01 -2.01924257e-02 1.18527934e-01
8.19162428e-01 8.46590221e-01 2.26033375e-01 -1.59462690e-01
-7.22012579e-01 1.28194988e-01 -2.10358262e-01 1.73817929e-02
1.29829776e+00 -9.95118693e-02 -2.37365842e-01 5.12337387e-01
8.71309757e-01 3.74368191e-01 -9.07718837e-01 2.35821575e-01
1.33248180e-01 -2.54801154e-01 -3.74016874e-02 -1.27569997e+00
-4.85950172e-01 7.05681682e-01 6.99610472e-01 -1.82224408e-01
8.48234594e-01 -1.37823880e-01 -2.88672149e-01 3.26608531e-02
-2.20085904e-01 -7.01239765e-01 -7.12100983e-01 -9.26790759e-02
7.66265392e-01 -1.10573673e+00 4.43906069e-01 -2.19884038e-01
-4.45456266e-01 7.59736001e-01 4.22481149e-01 5.65732596e-03
5.62666655e-01 6.18696392e-01 1.74933225e-01 -3.41606975e-01
-9.95972633e-01 8.90203640e-02 3.10624927e-01 8.37496638e-01
6.92499399e-01 3.01673889e-01 -8.38619113e-01 1.01145756e+00
-3.76824103e-02 3.94173294e-01 5.73605657e-01 8.16488028e-01
1.01520225e-01 -1.13783848e+00 -6.53277397e-01 1.22785318e+00
-9.22583282e-01 -3.59459132e-01 -4.66912955e-01 1.12979937e+00
3.30265552e-01 1.12778568e+00 1.02623943e-02 1.05552435e-01
4.64835376e-01 4.62429672e-01 3.23386997e-01 -7.66918778e-01
-1.01819372e+00 4.42673773e-01 3.52626115e-01 -6.08427525e-01
-2.50975192e-01 -1.00130534e+00 -1.15338922e+00 -1.42957762e-01
-1.91982806e-01 -2.96374619e-01 4.75531280e-01 9.93701398e-01
6.05675757e-01 8.32544148e-01 1.83813408e-01 -1.25333771e-01
-4.68872696e-01 -6.69814646e-01 -2.80571252e-01 9.65417400e-02
7.26323009e-01 -6.95720077e-01 -5.48803508e-01 -7.18689412e-02]
|
[8.028822898864746, 5.5230512619018555]
|
d67e674b-2d94-4de0-a48a-2d9acb23fe70
|
syntactic-and-semantic-driven-learning-for
|
2103.03448
| null |
https://arxiv.org/abs/2103.03448v1
|
https://arxiv.org/pdf/2103.03448v1.pdf
|
Syntactic and Semantic-driven Learning for Open Information Extraction
|
One of the biggest bottlenecks in building accurate, high coverage neural open IE systems is the need for large labelled corpora. The diversity of open domain corpora and the variety of natural language expressions further exacerbate this problem. In this paper, we propose a syntactic and semantic-driven learning approach, which can learn neural open IE models without any human-labelled data by leveraging syntactic and semantic knowledge as noisier, higher-level supervisions. Specifically, we first employ syntactic patterns as data labelling functions and pretrain a base model using the generated labels. Then we propose a syntactic and semantic-driven reinforcement learning algorithm, which can effectively generalize the base model to open situations with high accuracy. Experimental results show that our approach significantly outperforms the supervised counterparts, and can even achieve competitive performance to supervised state-of-the-art (SoA) model
|
['Hua Wu', 'Xinyan Xiao', 'Le Sun', 'Xianpei Han', 'Hongyu Lin', 'Yaojie Lu', 'Jialong Tang']
|
2021-03-05
| null |
https://aclanthology.org/2020.findings-emnlp.69
|
https://aclanthology.org/2020.findings-emnlp.69.pdf
|
findings-of-the-association-for-computational
|
['open-information-extraction']
|
['natural-language-processing']
|
[ 2.82005221e-01 6.61044300e-01 -4.88732040e-01 -6.15426481e-01
-8.42400908e-01 -6.86596990e-01 4.16080177e-01 -1.56754717e-01
-5.59913933e-01 1.00801468e+00 2.25816116e-01 -4.37153727e-01
-1.73048433e-02 -8.22894514e-01 -1.08101463e+00 -2.14957729e-01
2.20528007e-01 9.15022850e-01 2.50879526e-01 -2.23941430e-01
-1.14693850e-01 -1.30563319e-01 -1.57886779e+00 4.46292222e-01
1.26022077e+00 1.11911535e+00 3.21465224e-01 2.73831159e-01
-6.62284017e-01 8.92276704e-01 -3.08529735e-01 -5.15499651e-01
7.00061098e-02 -1.58887625e-01 -1.28948534e+00 -8.68256092e-02
2.11368188e-01 -1.17861785e-01 -1.20864697e-02 1.02246141e+00
1.56300589e-01 1.09784074e-01 6.12644017e-01 -1.06240010e+00
-9.78615105e-01 8.41912270e-01 -2.04044849e-01 2.50960514e-02
2.11330563e-01 5.79011925e-02 1.41794932e+00 -5.75594604e-01
9.99757707e-01 1.26469719e+00 3.50725830e-01 9.05036330e-01
-1.38011897e+00 -5.53623557e-01 4.71829593e-01 8.00056756e-02
-9.16338921e-01 -3.15226942e-01 6.74067020e-01 -1.95458800e-01
1.14779997e+00 -2.67039746e-01 3.35541427e-01 1.37791026e+00
-2.57387966e-01 1.04879498e+00 1.03200090e+00 -6.83455110e-01
3.30258191e-01 9.64713991e-02 4.84341592e-01 7.48778403e-01
9.88117903e-02 6.91141784e-02 -6.42850697e-02 3.10768243e-02
5.18091381e-01 -1.22151628e-01 -1.36921436e-01 -5.05311310e-01
-1.05482852e+00 9.84228313e-01 6.04274213e-01 2.32335761e-01
-2.42172256e-01 -7.35006258e-02 5.87369502e-01 4.69240606e-01
6.29563868e-01 6.49742544e-01 -1.01271415e+00 -8.57221335e-02
-5.04662275e-01 2.68021077e-01 1.11301017e+00 1.07576096e+00
9.85216022e-01 -3.44285101e-01 2.35182539e-01 1.21285427e+00
4.58206981e-01 2.33652219e-01 6.08226657e-01 -9.75314677e-01
6.42827392e-01 8.25601399e-01 -2.91165978e-01 -3.59440953e-01
-3.78777027e-01 -3.28242004e-01 -4.34068292e-01 -1.48858413e-01
3.13408643e-01 -3.68884355e-01 -1.03977156e+00 2.07114339e+00
2.53277153e-01 -2.98718847e-02 6.43093169e-01 6.48360312e-01
9.02975678e-01 6.94786668e-01 3.79200310e-01 2.98547465e-02
1.34980261e+00 -1.20175350e+00 -4.68567282e-01 -7.01583266e-01
9.62185621e-01 -1.01195080e-02 1.12598908e+00 1.38566315e-01
-8.67277980e-01 -3.11091512e-01 -8.39302480e-01 -2.19985977e-01
-5.17297328e-01 -1.76947236e-01 6.48685455e-01 1.20232619e-01
-8.53525102e-01 2.51478225e-01 -6.38755322e-01 -4.94158745e-01
6.08751178e-01 3.72769713e-01 -4.04018641e-01 -2.40987480e-01
-1.46354222e+00 8.07814240e-01 1.11139417e+00 -4.51306432e-01
-6.78737819e-01 -7.13694274e-01 -1.22100830e+00 2.82823026e-01
7.44242549e-01 -6.41817272e-01 1.45704436e+00 -1.35221362e+00
-1.47726619e+00 1.10628688e+00 4.80117835e-02 -5.64673603e-01
1.23401821e-01 -9.98715386e-02 -3.00957829e-01 6.75814599e-02
2.41415009e-01 1.24326503e+00 3.77158910e-01 -1.37230146e+00
-5.55584669e-01 -4.73640323e-01 3.06619793e-01 1.19838916e-01
-3.91190022e-01 1.46921903e-01 -2.38157302e-01 -3.06781769e-01
-8.60966966e-02 -8.35969627e-01 -4.69297171e-01 -2.92950541e-01
-2.10786983e-01 -8.58108103e-01 5.39402366e-01 -3.94591898e-01
1.05945194e+00 -1.97515333e+00 3.06159973e-01 6.55849501e-02
2.52970248e-01 3.56724113e-01 -2.60377347e-01 7.81454891e-02
1.19163729e-02 1.28920749e-01 -3.73583615e-01 -1.03709050e-01
1.03146896e-01 8.63361180e-01 -3.61397535e-01 -2.24974290e-01
4.98999387e-01 1.10191238e+00 -1.00883901e+00 -5.18440187e-01
-2.16015190e-01 8.37852731e-02 -8.65998030e-01 3.29899848e-01
-1.04457498e+00 5.51896632e-01 -6.85470223e-01 2.24944398e-01
1.53616548e-01 -6.43752813e-01 3.00216407e-01 2.41371170e-01
8.48998725e-02 7.44528413e-01 -9.61032987e-01 1.99196625e+00
-8.09936702e-01 2.20653638e-01 -1.43570095e-01 -1.34792781e+00
1.04605389e+00 3.63126576e-01 7.94477537e-02 -6.93812251e-01
3.23239923e-01 4.64000851e-01 -1.08715259e-01 -5.74246109e-01
-2.46239770e-02 -2.17274040e-01 -4.16262925e-01 5.26036918e-01
6.43785596e-01 2.27989212e-01 2.82416523e-01 1.12019069e-02
9.65745091e-01 3.20995033e-01 2.80347437e-01 -3.94281745e-01
6.69229627e-01 9.79164541e-02 9.07044351e-01 4.76324975e-01
-1.49173424e-01 2.11460665e-01 6.65122390e-01 -7.02878475e-01
-8.87074232e-01 -7.24619567e-01 -1.97861344e-01 1.53181434e+00
-5.83066009e-02 -1.59354538e-01 -1.21741486e+00 -1.08570361e+00
-1.93219006e-01 6.91052973e-01 -4.93469417e-01 -5.55298440e-02
-5.68117797e-01 -4.12549287e-01 5.26069224e-01 7.30530500e-01
4.91445184e-01 -1.43543255e+00 -3.82282466e-01 1.67606905e-01
-3.55161071e-01 -1.39962029e+00 -8.86444524e-02 6.01668417e-01
-9.34594750e-01 -8.12630475e-01 -4.12430793e-01 -1.35929716e+00
6.70612693e-01 -3.52737635e-01 1.28018045e+00 -1.14143208e-01
3.56556326e-02 -1.79670185e-01 -3.32811981e-01 -5.23107827e-01
-6.31759405e-01 6.94313288e-01 -4.14047502e-02 -1.21253021e-01
7.77404010e-01 -4.66104329e-01 -1.06162690e-01 1.61499262e-01
-8.48739028e-01 1.28399879e-01 5.07508457e-01 1.08496106e+00
5.66749156e-01 -2.02800900e-01 1.12047410e+00 -1.19108319e+00
5.19805551e-01 -6.41946256e-01 -6.67982519e-01 3.80410701e-01
-6.25530124e-01 6.36910677e-01 8.17156076e-01 -2.72952914e-01
-1.37614107e+00 7.81250298e-02 -3.58654052e-01 -2.42580041e-01
-7.68377364e-01 6.01760805e-01 -4.76606280e-01 4.24554013e-02
8.65910232e-01 -2.57470571e-02 -9.24176574e-02 -6.88576877e-01
5.64558864e-01 9.99666631e-01 5.36283553e-01 -9.17087317e-01
4.68012750e-01 1.93879619e-01 -5.53693295e-01 -5.70791900e-01
-1.57042348e+00 -3.96668762e-01 -8.69249284e-01 3.11955273e-01
9.69700515e-01 -9.01237071e-01 -2.33172610e-01 1.89394191e-01
-1.24590182e+00 -5.00843048e-01 -3.02991509e-01 1.68685585e-01
-8.53414059e-01 2.13760268e-02 -6.97630703e-01 -3.25271308e-01
-3.23685765e-01 -1.15870821e+00 8.98479462e-01 3.02631259e-01
-2.02910170e-01 -1.26519954e+00 3.42642188e-01 6.59242213e-01
1.85447127e-01 -7.50887841e-02 1.15099967e+00 -1.27208531e+00
-5.19771636e-01 8.81915316e-02 -3.69318634e-01 3.82551193e-01
-2.55379736e-01 -5.61375380e-01 -1.10829616e+00 1.64636806e-01
-3.48461121e-01 -1.11238289e+00 7.01011062e-01 9.36443452e-03
1.10167193e+00 -4.02676940e-01 -4.52565193e-01 7.17513621e-01
1.31425035e+00 -8.96283165e-02 4.66184288e-01 5.78331292e-01
7.19080091e-01 8.52316976e-01 4.18315053e-01 -3.28559093e-02
6.82669997e-01 3.17272812e-01 2.12032169e-01 -6.94893152e-02
9.78555307e-02 -6.27272546e-01 1.50534809e-01 7.53888130e-01
1.19944297e-01 -1.87783409e-02 -1.01897514e+00 6.87419355e-01
-1.95162356e+00 -7.49568462e-01 2.01067358e-01 1.69471359e+00
1.25671780e+00 1.89610794e-01 -8.71185809e-02 -1.31062284e-01
6.07078850e-01 7.49914274e-02 -6.57463193e-01 -5.60152292e-01
4.92774621e-02 3.27683270e-01 3.75057280e-01 3.18061113e-01
-1.09020388e+00 1.47506642e+00 6.18568659e+00 6.51069224e-01
-8.87547076e-01 2.13908345e-01 5.81665814e-01 2.79484659e-01
-3.94113541e-01 -3.03313825e-02 -1.10646987e+00 2.12911442e-01
1.25409663e+00 1.67497531e-01 4.01934922e-01 9.62894917e-01
-3.48666400e-01 3.12204868e-01 -1.30296409e+00 6.12010598e-01
1.32496208e-01 -1.21566868e+00 1.57432601e-01 2.37473115e-01
8.74771059e-01 6.06494665e-01 -4.68022138e-01 7.09715009e-01
8.65645826e-01 -7.98400342e-01 5.49460351e-01 1.15482695e-01
6.81388915e-01 -5.94928682e-01 6.38725221e-01 7.15094209e-01
-7.30280638e-01 -2.55499691e-01 -4.31045741e-01 -2.08602577e-01
9.38214436e-02 2.49868155e-01 -5.07338166e-01 2.87426949e-01
5.63955724e-01 7.23584354e-01 -2.30919063e-01 7.12554276e-01
-7.44991660e-01 6.88749611e-01 -4.00390536e-01 -1.47523448e-01
5.18334448e-01 6.85585737e-02 1.80035949e-01 9.54449236e-01
-2.13223815e-01 1.07457444e-01 5.52890480e-01 1.04313576e+00
-5.06101370e-01 2.05034092e-01 -8.43870223e-01 -7.86972791e-02
1.18900508e-01 1.11450374e+00 -4.28376347e-01 -3.54051679e-01
-9.21824872e-01 8.38271916e-01 1.05134594e+00 3.11523318e-01
-6.60542250e-01 -3.47377241e-01 5.35957575e-01 -2.57706255e-01
3.96881849e-01 1.79635853e-01 -2.20206633e-01 -1.45339096e+00
-7.01833190e-03 -8.69343877e-01 6.98911726e-01 -5.30270219e-01
-1.48594117e+00 6.72079802e-01 -1.36636615e-01 -6.77846491e-01
-4.31434423e-01 -9.04197574e-01 -4.46316063e-01 6.57471895e-01
-2.00172043e+00 -1.15727007e+00 1.07765630e-01 4.61537600e-01
7.81893373e-01 -1.74457356e-01 1.00030398e+00 1.79454505e-01
-5.88263214e-01 6.43139482e-01 1.12412125e-01 4.38811272e-01
4.59613204e-01 -1.23015428e+00 6.19988680e-01 4.56902742e-01
2.38021672e-01 3.26034069e-01 3.05084318e-01 -4.67619836e-01
-9.71459806e-01 -1.06942213e+00 1.14706433e+00 -5.27035654e-01
7.65062571e-01 -5.36227465e-01 -1.21520507e+00 1.06268537e+00
1.22022718e-01 2.57039964e-01 8.14797938e-01 5.88782430e-01
-6.36579812e-01 1.11269407e-01 -9.77647901e-01 3.63379061e-01
1.26409972e+00 -3.84040654e-01 -1.25257719e+00 1.80394024e-01
1.10963237e+00 -2.77641118e-01 -9.41921473e-01 4.71268773e-01
1.81863531e-01 -4.84148264e-01 7.81462908e-01 -1.09250450e+00
6.50964141e-01 1.79626405e-01 9.00017563e-03 -1.45055783e+00
-1.77019238e-01 -3.29919100e-01 9.74628795e-03 1.22558439e+00
9.26754951e-01 -7.35035062e-01 6.43041253e-01 7.51729488e-01
-1.52048901e-01 -8.50779235e-01 -7.26702034e-01 -8.04098010e-01
4.73642588e-01 -4.25537944e-01 5.26842058e-01 1.09604788e+00
5.22893429e-01 8.47619116e-01 1.27930298e-01 3.93670686e-02
4.82988715e-01 1.63256973e-01 5.24800539e-01 -1.68088496e+00
-3.27511668e-01 -2.07730249e-01 -2.87463576e-01 -1.15553725e+00
9.09739256e-01 -1.26548648e+00 1.77374110e-01 -1.50355089e+00
2.33879238e-01 -7.88928986e-01 -4.20463175e-01 8.18861783e-01
-1.67006940e-01 -1.65099613e-02 -1.77792534e-01 1.32168278e-01
-1.03069949e+00 5.59738278e-01 1.11396015e+00 4.71064150e-02
-2.20788613e-01 -3.63497794e-01 -9.13644969e-01 9.32273626e-01
7.07656145e-01 -4.18275028e-01 -4.86574173e-01 -8.82912457e-01
2.97140509e-01 -1.63096011e-01 1.75900757e-02 -8.04155827e-01
7.77858347e-02 -7.38797337e-02 1.62726529e-02 3.76925059e-02
-2.00714290e-01 -7.66366720e-01 -5.78716934e-01 7.33445063e-02
-6.84708595e-01 -3.67593557e-01 1.62290186e-01 5.58638930e-01
-2.97848046e-01 -4.17136580e-01 7.12501764e-01 -3.52002800e-01
-8.87626052e-01 3.26401442e-01 -6.55210316e-02 7.94750452e-01
1.05293572e+00 1.53125629e-01 -3.77390802e-01 -4.03759768e-04
-6.66238368e-01 7.06535280e-01 4.20018971e-01 7.82432795e-01
1.90296188e-01 -1.02676308e+00 -5.04875004e-01 4.72542912e-01
5.92869163e-01 5.08734524e-01 -1.16100520e-01 2.09844559e-01
-1.54337808e-01 7.21616268e-01 -8.11298415e-02 -5.46421528e-01
-7.72900164e-01 6.80999994e-01 2.85328537e-01 -5.07540762e-01
-7.03303933e-01 6.66705847e-01 5.43533087e-01 -1.16829443e+00
4.10484105e-01 -2.44568273e-01 -2.88373053e-01 -2.87151307e-01
5.31396687e-01 -1.72843128e-01 1.82893127e-03 -5.74358642e-01
-1.24102263e-02 4.70314592e-01 -3.94020706e-01 7.76991025e-02
1.54029822e+00 -2.06575617e-02 1.50287524e-01 3.14919800e-01
1.24410892e+00 -5.92350543e-01 -1.26423967e+00 -7.14005589e-01
4.06592965e-01 9.35407728e-02 4.44721766e-02 -9.77685750e-01
-8.69090319e-01 9.94225919e-01 6.70721158e-02 2.76682223e-03
8.96595001e-01 7.30256796e-01 1.11254442e+00 7.31726408e-01
4.83803779e-01 -1.12682426e+00 -2.96593726e-01 9.93927240e-01
3.90255392e-01 -1.46027303e+00 -7.05757380e-01 -4.90734071e-01
-6.40263498e-01 8.52914035e-01 9.62577224e-01 -1.76277965e-01
4.47688490e-01 2.48477653e-01 2.39471436e-01 -2.50488997e-01
-1.09784448e+00 -3.64522308e-01 2.32318006e-02 6.55058920e-01
3.69171977e-01 -7.00160787e-02 -1.25083774e-01 9.39175546e-01
-2.41938476e-02 2.33862087e-01 9.48083252e-02 8.44700754e-01
-7.30671704e-01 -1.34104359e+00 -2.84125954e-02 4.98972028e-01
-4.87437099e-01 -2.39288598e-01 -3.30610126e-01 4.03041899e-01
1.21036954e-01 5.52181900e-01 1.68626666e-01 9.05986875e-03
3.51935238e-01 7.41362393e-01 3.10824841e-01 -9.10562277e-01
-3.37305665e-01 -2.44860247e-01 4.04705018e-01 -5.17828703e-01
-2.91256547e-01 -5.41554511e-01 -1.70615852e+00 2.59997606e-01
-3.71611625e-01 1.94638908e-01 5.63602269e-01 1.36618805e+00
5.75773835e-01 3.54485542e-01 3.74009162e-01 -4.31189775e-01
-7.15298414e-01 -8.88277650e-01 -1.38132706e-01 6.26745105e-01
2.45785311e-01 -7.22347736e-01 -3.62720013e-01 1.56849742e-01]
|
[10.517142295837402, 8.877142906188965]
|
ace8b8eb-8864-4e9f-b6fa-54471c2ad104
|
generative-multimodal-entity-linking
|
2306.12725
| null |
https://arxiv.org/abs/2306.12725v1
|
https://arxiv.org/pdf/2306.12725v1.pdf
|
Generative Multimodal Entity Linking
|
Multimodal Entity Linking (MEL) is the task of mapping mentions with multimodal contexts to the referent entities from a knowledge base (e.g., Wikipedia). Prior MEL methods mainly focus on designing complex multimodal interaction mechanisms and require fine-tuning all model parameters, which can be prohibitively costly and difficult to scale in the era of Large Language Models (LLMs). In this work, we propose GEMEL, a simple yet effective Generative Multimodal Entity Linking method, which leverages the capabilities of LLMs from large-scale pre-training to directly generate target entity names. We keep the vision and language model frozen and only train a linear layer to enable cross-modality interactions. To adapt LLMs to the MEL task, we take advantage of the emerging in-context learning (ICL) capability of LLMs by retrieving multimodal instances as demonstrations. Extensive experiments show that with only ~0.3% of the model parameters fine-tuned, GEMEL achieves state-of-the-art results on two well-established MEL datasets (4.1% accuracy gains on WikiDiverse and 15.4% accuracy gains on WikiMEL). Our approach is compatible with any off-the-shelf language model, paving the way towards an efficient and general solution for utilizing LLMs in the MEL task.
|
['Min Zhang', 'Baotian Hu', 'Zhenran Xu', 'Senbao Shi']
|
2023-06-22
| null | null | null | null |
['entity-linking']
|
['natural-language-processing']
|
[-1.14196643e-01 4.49430674e-01 -6.44900948e-02 -5.66348247e-02
-1.23858476e+00 -8.60281229e-01 9.32426572e-01 -5.32775279e-03
-7.57205427e-01 7.15194404e-01 1.85612559e-01 -1.15521990e-01
2.55606651e-01 -5.48728287e-01 -1.10555518e+00 -3.26511681e-01
4.04295512e-02 7.60112286e-01 2.76365746e-02 -4.04937565e-01
-2.82935947e-01 8.12721774e-02 -1.41553187e+00 4.97919023e-01
8.93571734e-01 4.42518383e-01 2.63732016e-01 7.59948492e-01
-4.73297238e-01 5.68909466e-01 -4.72567350e-01 -1.06240606e+00
-2.00865254e-01 -3.08875620e-01 -9.07075882e-01 -4.34481353e-01
6.88007474e-01 -6.44667298e-02 -3.68405700e-01 6.27942204e-01
7.45426238e-01 3.85023504e-02 4.88578379e-01 -1.43199229e+00
-8.67435396e-01 9.53469872e-01 -2.51570910e-01 -3.89004171e-01
5.32384217e-01 2.52066284e-01 9.42847311e-01 -1.17473435e+00
9.54966247e-01 1.28839409e+00 7.34566092e-01 9.44079340e-01
-1.36474311e+00 -4.54650700e-01 1.65698022e-01 1.29177421e-02
-1.37684083e+00 -6.18987501e-01 2.99313635e-01 -3.44637454e-01
1.12383735e+00 1.79596663e-01 2.68794328e-01 1.47772717e+00
-4.04438913e-01 9.22600269e-01 7.70496488e-01 -5.92355192e-01
-2.80404806e-01 3.42664301e-01 -1.53800502e-01 8.14098358e-01
7.31773376e-02 -2.19692305e-01 -7.29095101e-01 -1.02513649e-01
4.78569388e-01 -5.50685048e-01 -2.62546986e-01 -4.52946275e-01
-1.60357654e+00 6.47434056e-01 5.30245543e-01 2.02087000e-01
-1.24562927e-01 3.53377402e-01 3.60962689e-01 1.80091068e-01
-2.62751076e-02 6.32741511e-01 -3.53737235e-01 -1.38730958e-01
-6.75376475e-01 9.85142440e-02 9.53397512e-01 1.23570371e+00
6.78560972e-01 -3.47341031e-01 -2.69110858e-01 8.95927131e-01
4.06872153e-01 7.35503495e-01 1.60769582e-01 -7.57972360e-01
8.66778314e-01 6.69503987e-01 2.49578536e-01 -6.11775637e-01
-4.46666330e-01 -1.40122682e-01 -4.54044878e-01 -3.14699858e-01
4.64170367e-01 -3.98382783e-01 -8.42001319e-01 2.11501145e+00
4.30363446e-01 1.92327216e-01 3.81079644e-01 6.44117594e-01
1.36919141e+00 6.52344286e-01 6.25171721e-01 3.01475108e-01
1.23728323e+00 -1.03292608e+00 -4.45936322e-01 -2.83680052e-01
8.27071548e-01 -6.58083320e-01 1.44303858e+00 -4.11691777e-02
-1.24563861e+00 -3.81489128e-01 -7.31824994e-01 -4.02919561e-01
-7.82464325e-01 4.24630851e-01 7.21157014e-01 2.93243468e-01
-1.02531075e+00 1.28324345e-01 -7.86289632e-01 -7.50499487e-01
1.12441853e-01 4.56517220e-01 -6.91018939e-01 -1.84547365e-01
-1.32840204e+00 1.06104231e+00 5.56386650e-01 2.14550480e-01
-7.95278251e-01 -7.90383637e-01 -1.09380507e+00 -1.09965764e-01
3.93510044e-01 -1.00641012e+00 1.21823955e+00 -6.74577653e-01
-1.32287610e+00 9.18030977e-01 -1.16386876e-01 -9.83163193e-02
4.60056752e-01 -3.09340060e-01 -5.37182927e-01 1.03646621e-01
-7.59685040e-02 1.25589430e+00 3.47355843e-01 -1.59416950e+00
-4.56637710e-01 1.31174564e-01 3.39993119e-01 3.11631382e-01
-4.15137410e-01 -1.59282580e-01 -1.09706402e+00 -3.08759212e-01
-5.14033675e-01 -1.24018991e+00 1.86420493e-02 -3.96575361e-01
-5.76349556e-01 -1.73336968e-01 4.68274295e-01 -8.91160667e-01
1.38351357e+00 -1.89381063e+00 5.26223779e-01 -1.11661023e-02
7.28992745e-02 4.12279636e-01 -6.37431204e-01 9.86496031e-01
2.80455977e-01 1.09585471e-01 -2.20016137e-01 -8.14716995e-01
5.22602975e-01 6.94395304e-02 -1.59217581e-01 -1.29643202e-01
2.21405938e-01 1.50466883e+00 -8.61420274e-01 -4.87538785e-01
6.49608895e-02 8.14004958e-01 -4.34010148e-01 4.44438696e-01
-4.88117248e-01 4.75826204e-01 -3.83128785e-02 5.64575195e-01
2.89407462e-01 -4.27739203e-01 5.03394783e-01 -5.83821476e-01
2.31594127e-02 2.12152183e-01 -1.04330277e+00 2.09307575e+00
-8.35311472e-01 5.61506093e-01 -1.31316856e-02 -3.33653837e-01
4.24383163e-01 4.06865090e-01 -7.27415979e-02 -6.50792360e-01
-1.17802829e-01 2.15786591e-01 -3.46330434e-01 -7.54139304e-01
6.52633071e-01 1.17457926e-01 -5.49970448e-01 2.98346311e-01
4.48056132e-01 3.20810229e-01 4.06069398e-01 5.66355467e-01
7.70088136e-01 4.47717488e-01 5.14354603e-03 4.32052433e-01
3.16138417e-01 5.91204055e-02 5.56486063e-02 7.25249112e-01
3.76133978e-01 3.43007386e-01 3.25907856e-01 1.48287117e-01
-8.78127933e-01 -1.07406473e+00 2.32971117e-01 1.39733636e+00
9.81947035e-02 -5.13915002e-01 -8.23817670e-01 -8.78708303e-01
4.93588373e-02 8.23740065e-01 -5.11802197e-01 -4.81859362e-03
-5.28045952e-01 -7.86538839e-01 9.34023380e-01 5.03330529e-01
4.07704622e-01 -1.06494975e+00 -1.44056767e-01 1.34154707e-01
-4.82525557e-01 -1.41804123e+00 -4.33802843e-01 -2.33552620e-01
-3.59969229e-01 -8.44661117e-01 -7.17845976e-01 -7.86900163e-01
6.94196403e-01 -1.31602764e-01 1.54100704e+00 -7.18884841e-02
-1.56636000e-01 9.21338916e-01 -2.51995355e-01 -5.37865460e-02
-4.66300666e-01 5.82657039e-01 -2.59299636e-01 -1.65580630e-01
2.86190748e-01 -3.90640914e-01 -4.23948258e-01 9.32282507e-02
-8.06752801e-01 3.75446200e-01 8.19531739e-01 8.47165465e-01
4.16310757e-01 -6.02185130e-01 7.65203893e-01 -1.01836061e+00
5.09873152e-01 -6.29390419e-01 -4.38739836e-01 7.97926188e-01
-3.76707733e-01 9.94288400e-02 1.79612026e-01 -6.45918846e-01
-1.17273533e+00 8.02400410e-02 -1.87563635e-02 -2.37509012e-01
-1.97491407e-01 9.45546627e-01 -4.99491870e-01 -4.63773981e-02
4.56434935e-01 -3.92693765e-02 -3.91737729e-01 -6.20498240e-01
1.07562745e+00 5.36054730e-01 9.26329017e-01 -8.35362017e-01
7.78824925e-01 2.52867136e-02 -2.02876374e-01 -4.63967085e-01
-7.21627951e-01 -2.27054447e-01 -6.21570349e-01 -1.66907474e-01
9.59141135e-01 -1.25941622e+00 -9.19179022e-01 2.67017186e-01
-1.10495555e+00 -5.94860077e-01 1.46574497e-01 2.95542449e-01
-3.89976084e-01 1.31032467e-01 -7.44100332e-01 -5.36546350e-01
-4.00533795e-01 -8.00345600e-01 1.27406573e+00 3.31844091e-01
-2.47427046e-01 -1.18418384e+00 1.76286235e-01 6.46041811e-01
3.53093237e-01 2.28789091e-01 9.43927169e-01 -5.87604225e-01
-8.01211536e-01 -1.85354427e-01 -2.66281247e-01 -1.50572434e-01
-3.22030336e-01 5.70472665e-02 -1.01049089e+00 -2.45167226e-01
-1.04233611e+00 -7.18313932e-01 7.77672470e-01 -2.45822236e-01
6.02701128e-01 -2.86282450e-01 -6.12576306e-01 5.22670984e-01
1.35696924e+00 -1.54856354e-01 5.48641503e-01 3.56922448e-01
1.22339571e+00 6.33045971e-01 5.34481823e-01 6.53522313e-02
1.15047324e+00 8.09465528e-01 3.46123159e-01 -1.87330544e-01
-4.16925639e-01 -6.89949334e-01 3.89548212e-01 8.44694316e-01
1.55412093e-01 -5.23354888e-01 -1.18078864e+00 7.82277882e-01
-2.07091331e+00 -9.03642952e-01 -3.56468633e-02 2.17819691e+00
1.30051899e+00 -3.34939778e-01 4.37148549e-02 -8.36699069e-01
5.86064458e-01 -2.05018237e-01 -4.03544873e-01 -5.69220968e-02
-1.83839425e-01 -1.03007657e-02 1.94112524e-01 7.16495574e-01
-1.26106894e+00 1.31280828e+00 5.36153221e+00 6.60010636e-01
-9.39043105e-01 1.79115266e-01 6.39577284e-02 -2.59522378e-01
-5.20844758e-01 -7.89046586e-02 -1.07487881e+00 4.83563155e-01
1.15254545e+00 1.25233248e-01 6.38800383e-01 3.68007839e-01
-2.85502046e-01 -4.81667481e-02 -1.41931534e+00 1.11357391e+00
1.18717819e-01 -1.25924647e+00 1.92046866e-01 -9.71046761e-02
6.50726438e-01 1.56831816e-01 2.92172283e-02 7.83052564e-01
4.69145268e-01 -1.14805508e+00 5.64524114e-01 8.28533769e-01
1.07681656e+00 -7.60677576e-01 7.06913352e-01 2.36482933e-01
-1.04489720e+00 1.74357101e-01 1.57336637e-01 5.80013096e-01
4.74177599e-01 1.76481232e-01 -9.32926595e-01 5.18241942e-01
5.26011407e-01 3.54294091e-01 -7.86396742e-01 9.19248462e-01
-2.98930854e-01 2.62738466e-01 -4.11957979e-01 2.18507737e-01
7.35274404e-02 1.92647606e-01 3.11101198e-01 1.64078224e+00
4.27944541e-01 -1.67166531e-01 1.04508370e-01 7.78490722e-01
-6.13371074e-01 2.57963035e-02 -5.58234930e-01 -4.62216586e-01
6.97946966e-01 1.47531748e+00 -1.53438032e-01 -3.08081359e-01
-5.60613751e-01 1.14338148e+00 6.80360496e-01 5.42310953e-01
-1.00266993e+00 -4.69617724e-01 3.43964696e-01 -2.25372702e-01
3.58894646e-01 -3.28276068e-01 8.76399726e-02 -1.34446251e+00
-5.77675626e-02 -9.19882774e-01 4.45266783e-01 -9.46994185e-01
-1.40702844e+00 6.62899613e-01 9.22898799e-02 -9.18600798e-01
-3.58639389e-01 -4.37130779e-01 -3.10214400e-01 9.39856291e-01
-1.50348127e+00 -1.96880281e+00 -3.86635631e-01 5.66339910e-01
1.30448595e-01 -5.40040433e-02 9.80123162e-01 6.19532108e-01
-6.84013009e-01 1.05156517e+00 -1.66250467e-01 2.45249614e-01
1.19706619e+00 -1.31597936e+00 5.11040092e-01 6.10036314e-01
2.68869400e-01 9.41172063e-01 5.39008379e-01 -7.16780722e-01
-1.66133356e+00 -1.06638277e+00 1.36410189e+00 -9.67718303e-01
8.06204975e-01 -6.88314736e-01 -8.54253590e-01 9.34148192e-01
5.94605505e-01 -2.13081479e-01 9.74642754e-01 3.07303458e-01
-5.55763662e-01 2.65099227e-01 -8.16128850e-01 9.42414045e-01
1.14629674e+00 -7.91745007e-01 -5.13451993e-01 2.15131506e-01
7.85820842e-01 -6.41862929e-01 -1.14268911e+00 4.20704335e-01
6.00312769e-01 -3.64478469e-01 1.07442939e+00 -8.07819426e-01
3.41556966e-01 -4.16280776e-01 -2.16539860e-01 -1.27302563e+00
5.75821735e-02 -6.89765334e-01 -4.42536533e-01 1.82057929e+00
9.05217826e-01 -3.52220386e-01 3.83618444e-01 9.74998295e-01
-4.90928069e-02 -4.66198325e-01 -6.39857173e-01 -4.80136812e-01
-4.65973243e-02 -2.89430469e-01 5.22884548e-01 1.14555120e+00
2.09054753e-01 8.11146438e-01 -3.22095454e-01 4.57866162e-01
5.07078588e-01 4.90999669e-02 1.04934275e+00 -9.52615380e-01
-3.91762108e-01 -2.09549069e-01 1.54406190e-01 -8.78943920e-01
3.65607440e-01 -1.05695236e+00 -4.29747021e-03 -1.74916065e+00
3.17564458e-01 -6.45652473e-01 -1.80683494e-01 8.78379524e-01
-3.58603448e-01 3.29741240e-01 5.48449993e-01 1.26012087e-01
-1.01366615e+00 6.07167244e-01 8.76024127e-01 -1.15401439e-01
-2.04929754e-01 -5.40986121e-01 -6.54070973e-01 4.87711221e-01
5.55987000e-01 -1.51219562e-01 -5.42066813e-01 -8.07270825e-01
5.79464018e-01 3.57536115e-02 5.85547805e-01 -6.51191473e-01
3.64935935e-01 1.01637961e-02 1.80601493e-01 -3.96849066e-01
5.67922950e-01 -7.50674486e-01 3.24036926e-01 -2.78519511e-01
-5.47831833e-01 2.75412276e-02 5.19563317e-01 4.36483026e-01
-1.48614049e-01 -1.34972557e-01 2.07609102e-01 -2.25074850e-02
-1.05778670e+00 1.45260533e-02 3.18380967e-02 1.80349380e-01
8.43900800e-01 3.11084270e-01 -8.59074712e-01 -3.46977472e-01
-1.01430333e+00 5.23754835e-01 5.82461119e-01 6.38354361e-01
2.92509198e-01 -1.53006780e+00 -4.58400130e-01 -2.67317384e-01
4.60647970e-01 -1.02950513e-01 3.54792476e-01 9.10440803e-01
-7.23196119e-02 4.53873754e-01 6.88329786e-02 -4.09152120e-01
-1.26432264e+00 5.47049701e-01 2.76422471e-01 -2.65806347e-01
-2.82550454e-01 9.72543240e-01 1.90241665e-01 -9.33190942e-01
3.30617011e-01 1.09931588e-01 -1.99089706e-01 2.46175572e-01
4.09351021e-01 1.40593529e-01 -4.39031795e-02 -7.09962666e-01
-3.11366946e-01 4.73201752e-01 1.13072491e-03 -4.21839774e-01
1.16125679e+00 -3.96431118e-01 -9.15118903e-02 4.87927526e-01
1.12953401e+00 1.63841069e-01 -1.20292425e+00 -2.08618924e-01
1.21619470e-01 1.19922258e-01 -2.11311877e-01 -1.51830328e+00
-8.01051736e-01 7.57330656e-01 4.51634347e-01 -2.49357089e-01
8.59210134e-01 3.40207875e-01 9.74584281e-01 6.56419575e-01
4.77101266e-01 -9.53978717e-01 -3.55565138e-02 4.92096186e-01
8.24059367e-01 -1.41328752e+00 -4.66327071e-01 -3.68204057e-01
-9.61358368e-01 7.51261294e-01 8.50823939e-01 4.71954912e-01
7.60331899e-02 2.15180486e-01 1.51503175e-01 -1.93767220e-01
-9.52104092e-01 -4.00656164e-01 5.90733469e-01 6.09016359e-01
6.61375642e-01 7.11161569e-02 7.86298364e-02 7.65874743e-01
8.85326415e-02 -7.58772641e-02 1.72748059e-01 7.69188523e-01
-3.93585078e-02 -1.29882753e+00 -9.37571600e-02 -1.00696810e-01
-1.81805074e-01 -4.97367769e-01 -4.35937256e-01 1.14661324e+00
1.30165741e-01 8.08594644e-01 -1.95819810e-01 -2.78559625e-01
5.76543629e-01 5.07682562e-01 6.03848696e-01 -5.69990873e-01
-6.97717369e-01 -1.03959359e-01 6.12461746e-01 -5.90454340e-01
-4.07070935e-01 -6.03758812e-01 -1.41372550e+00 -2.45951310e-01
-8.54597762e-02 -5.07502593e-02 6.82327271e-01 7.56745100e-01
8.03431690e-01 3.39917094e-01 -9.69724953e-02 -8.90737891e-01
-4.92528751e-02 -8.19214582e-01 6.62456974e-02 4.38315928e-01
1.44521892e-01 -6.37405217e-01 8.49160999e-02 2.73973227e-01]
|
[10.882445335388184, 1.6278167963027954]
|
baa52932-660c-4894-9cac-04b34baca1fb
|
smae-few-shot-learning-for-hdr-deghosting
|
2304.06914
| null |
https://arxiv.org/abs/2304.06914v1
|
https://arxiv.org/pdf/2304.06914v1.pdf
|
SMAE: Few-shot Learning for HDR Deghosting with Saturation-Aware Masked Autoencoders
|
Generating a high-quality High Dynamic Range (HDR) image from dynamic scenes has recently been extensively studied by exploiting Deep Neural Networks (DNNs). Most DNNs-based methods require a large amount of training data with ground truth, requiring tedious and time-consuming work. Few-shot HDR imaging aims to generate satisfactory images with limited data. However, it is difficult for modern DNNs to avoid overfitting when trained on only a few images. In this work, we propose a novel semi-supervised approach to realize few-shot HDR imaging via two stages of training, called SSHDR. Unlikely previous methods, directly recovering content and removing ghosts simultaneously, which is hard to achieve optimum, we first generate content of saturated regions with a self-supervised mechanism and then address ghosts via an iterative semi-supervised learning framework. Concretely, considering that saturated regions can be regarded as masking Low Dynamic Range (LDR) input regions, we design a Saturated Mask AutoEncoder (SMAE) to learn a robust feature representation and reconstruct a non-saturated HDR image. We also propose an adaptive pseudo-label selection strategy to pick high-quality HDR pseudo-labels in the second stage to avoid the effect of mislabeled samples. Experiments demonstrate that SSHDR outperforms state-of-the-art methods quantitatively and qualitatively within and across different datasets, achieving appealing HDR visualization with few labeled samples.
|
['Yanning Zhang', 'Luc van Gool', 'Jinqiu Sun', 'Yu Zhu', 'Hao Tang', 'Weiye Chen', 'Song Zhang', 'Qingsen Yan']
|
2023-04-14
| null |
http://openaccess.thecvf.com//content/CVPR2023/html/Yan_SMAE_Few-Shot_Learning_for_HDR_Deghosting_With_Saturation-Aware_Masked_Autoencoders_CVPR_2023_paper.html
|
http://openaccess.thecvf.com//content/CVPR2023/papers/Yan_SMAE_Few-Shot_Learning_for_HDR_Deghosting_With_Saturation-Aware_Masked_Autoencoders_CVPR_2023_paper.pdf
|
cvpr-2023-1
|
['pseudo-label']
|
['miscellaneous']
|
[ 4.86068308e-01 -3.23636122e-02 -1.40643362e-02 -3.65224123e-01
-5.54814875e-01 -2.10856035e-01 3.91544431e-01 -4.60111171e-01
-2.60859847e-01 8.08622718e-01 6.08818121e-02 7.78080449e-02
3.00491787e-02 -9.08817589e-01 -7.56402850e-01 -1.09207916e+00
4.11317378e-01 3.25941741e-01 3.55865359e-01 -2.89431095e-01
-1.12183936e-01 5.79172671e-01 -1.73591375e+00 7.66158942e-03
1.06320381e+00 8.90215635e-01 5.88701069e-01 3.59208167e-01
-8.39208905e-03 1.12005746e+00 -4.81430799e-01 3.49507667e-03
4.33652043e-01 -8.08615029e-01 -5.79028010e-01 4.69570756e-01
2.16134533e-01 -7.77888417e-01 -6.05062366e-01 1.06026435e+00
6.61595523e-01 1.92143559e-01 5.99675059e-01 -8.87133181e-01
-8.52659166e-01 3.35817873e-01 -6.47127271e-01 9.90984738e-02
6.24309666e-03 5.47827184e-01 5.48352182e-01 -8.37506711e-01
7.14088500e-01 8.46519947e-01 3.15357834e-01 7.80505598e-01
-1.46754062e+00 -6.47495747e-01 -2.53121555e-01 -4.24134098e-02
-1.26119602e+00 -6.34327769e-01 1.06220567e+00 -3.32064390e-01
3.80373150e-01 1.32531658e-01 7.34356463e-01 9.80722368e-01
-4.09940369e-02 4.85886991e-01 1.65713274e+00 -3.43685031e-01
3.91242743e-01 1.05354704e-01 -1.48839384e-01 6.67525709e-01
7.52686486e-02 5.78140259e-01 -2.95303494e-01 2.70038933e-01
1.19377720e+00 2.70479649e-01 -5.39839983e-01 -2.78072596e-01
-1.24770272e+00 6.78735673e-01 5.69101632e-01 4.80414450e-01
-3.69284362e-01 -2.39167944e-01 -3.73695628e-04 2.61673182e-01
4.15014058e-01 2.63146222e-01 -1.00852206e-01 4.21910405e-01
-1.09731674e+00 -1.90009400e-01 3.26761872e-01 6.73489571e-01
1.05906272e+00 4.54973787e-01 -2.16994211e-01 1.10739422e+00
6.74504042e-02 7.16164470e-01 4.56174076e-01 -1.17012334e+00
-2.03222319e-01 3.75037730e-01 2.27701485e-01 -7.16335297e-01
-2.23979682e-01 -4.32300329e-01 -1.40218878e+00 5.82272887e-01
3.29245687e-01 7.08455220e-02 -1.35988510e+00 1.46749711e+00
3.98130506e-01 1.17649578e-01 2.87731797e-01 1.36307001e+00
1.08999419e+00 9.59029019e-01 -7.52389012e-03 -7.54352629e-01
9.32675898e-01 -9.62498546e-01 -8.96209061e-01 -1.74624905e-01
1.31692663e-01 -5.26531756e-01 1.19747972e+00 4.44196135e-01
-1.13284051e+00 -6.59231961e-01 -1.16693199e+00 -2.40986496e-01
-4.90575209e-02 6.98659793e-02 4.13215965e-01 3.56347680e-01
-8.28722298e-01 6.85205996e-01 -3.59398901e-01 3.07420529e-02
4.25625980e-01 3.21123041e-02 -2.72358477e-01 -5.21211386e-01
-1.21259344e+00 6.61222517e-01 4.57461506e-01 2.07784086e-01
-1.35212159e+00 -7.66461492e-01 -7.18682230e-01 -2.80863941e-01
4.49710280e-01 -6.16270840e-01 7.31865883e-01 -9.23248708e-01
-1.91087937e+00 1.06146097e+00 1.29889563e-01 -2.02756599e-01
4.20756400e-01 1.66552663e-01 -5.66490948e-01 4.90171224e-01
-6.12629876e-02 7.37535417e-01 1.10954738e+00 -1.74961472e+00
-2.19364896e-01 -1.08583383e-01 -2.09256142e-01 1.53111607e-01
-1.32366002e-01 -5.11168167e-02 -1.67069957e-01 -5.77153325e-01
2.36058891e-01 -7.13834524e-01 -2.50481665e-01 2.14418814e-01
-4.12025750e-01 1.92187443e-01 8.66530299e-01 -7.45860457e-01
8.69988918e-01 -2.11672544e+00 4.03676406e-02 -1.68005660e-01
5.22897124e-01 5.64895749e-01 -1.65243506e-01 -9.86645073e-02
-1.67582363e-01 -2.95406401e-01 -6.73831820e-01 -6.96482062e-02
-2.91911393e-01 2.00024039e-01 -3.96159619e-01 7.15801716e-01
1.24843597e-01 9.33038414e-01 -9.43793237e-01 -6.75510943e-01
8.11721146e-01 6.25862956e-01 -1.00844607e-01 5.77175498e-01
-3.41723502e-01 7.68986642e-01 -1.31289707e-03 8.21438968e-01
8.89146447e-01 -4.15866733e-01 -2.57277712e-02 -5.63954234e-01
-3.02959681e-01 -3.72624218e-01 -1.04258764e+00 1.49392474e+00
-4.88477945e-01 4.23889041e-01 -4.86821216e-03 -1.07752323e+00
1.27594900e+00 2.80032512e-02 6.85268104e-01 -1.14083302e+00
2.18521044e-01 4.68370885e-01 -2.56635517e-01 -5.84888935e-01
2.30929092e-01 -6.99332595e-01 1.75651327e-01 5.53818107e-01
1.42359376e-01 -2.09405839e-01 2.23539807e-02 7.42617920e-02
8.23950529e-01 -9.23638567e-02 1.71960652e-01 5.99595159e-02
4.59757298e-01 -1.50859326e-01 5.90521574e-01 7.49752939e-01
-2.55306035e-01 1.19449615e+00 -5.54600433e-02 -5.22583306e-01
-1.50351751e+00 -1.17933774e+00 -1.58207431e-01 6.66385591e-01
5.12482226e-01 3.16362143e-01 -3.98911625e-01 -3.41699123e-01
-4.45905179e-01 5.25382936e-01 -4.69526410e-01 -2.23788485e-01
-6.48945391e-01 -9.09806967e-01 1.75591916e-01 1.73358992e-01
8.96947384e-01 -1.33551157e+00 -5.66869259e-01 2.04471692e-01
-3.10848117e-01 -1.16730917e+00 -9.43065509e-02 3.72252256e-01
-6.07192159e-01 -7.37449408e-01 -1.18899953e+00 -7.91706264e-01
6.29852414e-01 5.73625207e-01 1.10591447e+00 1.04370274e-01
-4.81576860e-01 -9.48025435e-02 -3.99889410e-01 2.02423826e-01
-5.82766533e-01 -3.94649059e-01 -1.32153183e-01 1.43351391e-01
-9.70253944e-02 -6.32331610e-01 -9.73944724e-01 4.45216149e-01
-1.22969651e+00 4.21846181e-01 9.88966107e-01 9.71896231e-01
9.88143623e-01 2.86343068e-01 5.85755229e-01 -7.94360161e-01
3.52573469e-02 -3.38147402e-01 -5.21112978e-01 2.41369411e-01
-5.71303368e-01 -4.45345119e-02 9.57510293e-01 -5.98030031e-01
-1.33819938e+00 1.77852139e-01 -1.88027322e-01 -8.43783855e-01
-3.62042278e-01 -1.75574943e-01 -2.56413817e-01 -2.72131294e-01
7.04814017e-01 7.32242882e-01 2.08474338e-01 -1.09170258e-01
5.44227958e-01 7.17486560e-01 7.45347083e-01 -2.84230411e-01
9.36236441e-01 9.42411840e-01 -1.09939098e-01 -9.27864254e-01
-1.15985310e+00 -3.67992163e-01 -3.66104841e-01 -5.38210750e-01
9.61679995e-01 -1.00443840e+00 -3.86256725e-01 7.69467056e-01
-7.16034770e-01 -7.99510539e-01 -6.09002054e-01 2.79569626e-01
-7.53663898e-01 2.47350127e-01 -7.20053017e-01 -5.91878831e-01
-4.49916661e-01 -9.89596844e-01 1.11115170e+00 3.63187462e-01
4.01123524e-01 -6.86285436e-01 7.68085592e-04 4.39098001e-01
5.77761292e-01 3.29646438e-01 6.67463779e-01 -6.18886501e-02
-8.35835874e-01 2.38714233e-01 -6.17158413e-01 6.24021769e-01
3.98226641e-02 -2.92792052e-01 -1.05784428e+00 -2.66738892e-01
3.55236739e-01 -6.70078158e-01 9.94175971e-01 5.15355289e-01
1.28943539e+00 -1.61344871e-01 1.84765086e-02 8.83787513e-01
1.73436069e+00 3.98682579e-02 1.07576454e+00 2.20398217e-01
8.59550595e-01 6.40255332e-01 7.25726843e-01 3.69706571e-01
-1.27798738e-02 5.20056427e-01 2.61579692e-01 -5.91691256e-01
-6.31755352e-01 -1.51672870e-01 2.29710594e-01 7.29983389e-01
1.34809390e-01 -2.94817001e-01 -4.50268239e-01 4.06309277e-01
-1.45534146e+00 -9.86826599e-01 -9.64970738e-02 2.03946662e+00
1.07800102e+00 -2.24974185e-01 6.45742640e-02 2.87075400e-01
8.77712727e-01 4.30181265e-01 -8.28175962e-01 3.29071164e-01
-5.05931556e-01 6.46686852e-02 3.62059087e-01 2.28808567e-01
-9.11229849e-01 8.84418130e-01 5.91955280e+00 1.00333846e+00
-1.43990016e+00 2.84408212e-01 1.04849064e+00 -6.90091923e-02
-4.49330926e-01 -9.40929055e-02 -5.32451212e-01 6.54870629e-01
6.85279608e-01 2.08329201e-01 6.30732536e-01 5.44309080e-01
2.91600049e-01 -1.30825698e-01 -6.15988135e-01 1.31983685e+00
2.48600841e-01 -1.29464948e+00 -8.72715190e-02 1.10368002e-02
1.11338985e+00 -1.31999016e-01 7.02902600e-02 4.45277654e-02
4.85331893e-01 -9.47561622e-01 5.74846804e-01 7.62383878e-01
1.25084722e+00 -5.67534089e-01 4.51697290e-01 2.09154025e-01
-8.57537031e-01 -2.22300485e-01 -6.26776338e-01 4.50753689e-01
4.30919230e-01 1.26625729e+00 -8.36546421e-02 3.58050942e-01
7.36885607e-01 7.57549822e-01 -3.88833761e-01 7.90565193e-01
-1.78619117e-01 3.07633072e-01 -9.21472237e-02 3.88586849e-01
-1.37069464e-01 -3.97683769e-01 3.76995504e-01 7.30526984e-01
2.81985193e-01 5.34590304e-01 1.49863422e-01 1.28673184e+00
-9.74410102e-02 -2.68009037e-01 -5.47585130e-01 1.12987466e-01
3.17576081e-01 1.44674158e+00 -8.62454236e-01 -3.93731743e-01
-2.27455452e-01 1.21425092e+00 1.51795253e-01 5.15782297e-01
-7.50753343e-01 -3.82229686e-01 -6.31071553e-02 3.75286132e-01
1.72498196e-01 -3.86012904e-02 -5.88631146e-02 -1.28224969e+00
-2.37278253e-01 -8.39582443e-01 1.30796179e-01 -1.19857383e+00
-1.36261237e+00 6.38400614e-01 -1.91345707e-01 -1.47442758e+00
-5.13764620e-02 -1.98395938e-01 -4.55974549e-01 5.40299475e-01
-1.85810697e+00 -1.05245638e+00 -8.49917591e-01 6.32543504e-01
4.06781942e-01 -2.57084686e-02 4.97176379e-01 4.97974992e-01
-5.35351217e-01 1.18981726e-01 1.69618219e-01 -1.37916937e-01
8.05135131e-01 -1.07574165e+00 -2.07903832e-01 8.74868214e-01
-1.68555439e-01 2.22279966e-01 7.77815938e-01 -5.15347719e-01
-1.23515129e+00 -1.25193691e+00 2.17618644e-01 2.15807065e-01
2.98858702e-01 -1.60885587e-01 -1.15589893e+00 1.86333090e-01
7.07543194e-02 5.30447841e-01 3.46762240e-01 -6.28119946e-01
-1.23465016e-01 -3.90466213e-01 -1.36321497e+00 4.17688757e-01
1.15137517e+00 -4.56638217e-01 -1.85123608e-01 2.97112823e-01
8.66478205e-01 -2.45205268e-01 -8.89575601e-01 5.72921097e-01
2.82532275e-01 -1.34588218e+00 1.03727448e+00 3.84481966e-01
7.35270798e-01 -7.18776524e-01 -1.08038455e-01 -1.31458795e+00
-1.84496209e-01 -4.05591875e-01 -2.55545318e-01 1.19211662e+00
-2.56930023e-01 -3.72399271e-01 6.46195114e-01 2.73471475e-01
-2.85177618e-01 -5.90588093e-01 -5.04866660e-01 -8.31487358e-01
-1.22390613e-01 1.57699302e-01 4.06625897e-01 1.05645311e+00
-7.33225882e-01 1.58344448e-01 -7.92958915e-01 -8.48122016e-02
1.12494874e+00 2.72461832e-01 6.16205335e-01 -1.09833455e+00
-3.19881260e-01 -1.43464014e-01 -4.62973490e-02 -9.29385781e-01
2.17209041e-01 -6.40820861e-01 4.94487792e-01 -1.57878947e+00
4.06969428e-01 -6.91921532e-01 -1.78367779e-01 2.16466948e-01
-4.68815619e-04 7.07425296e-01 -8.15297514e-02 5.00796318e-01
-7.52400875e-01 9.26223218e-01 1.66132498e+00 -1.23802923e-01
-1.77808091e-01 -4.90462512e-01 -5.43126583e-01 3.75915796e-01
6.51348472e-01 -3.21256071e-01 -3.77805412e-01 -1.70139417e-01
-1.20631792e-01 2.49508873e-01 5.70647180e-01 -1.16643643e+00
2.01843493e-02 -3.17391038e-01 6.84397936e-01 -5.60686827e-01
1.84599891e-01 -6.64096057e-01 4.16223705e-01 2.72509634e-01
-2.10407153e-01 -7.53181100e-01 -3.84934336e-01 5.70489407e-01
-2.71313727e-01 -2.33621106e-01 1.48849356e+00 -3.15978974e-01
-8.22485745e-01 4.40564752e-01 -1.55637369e-01 -9.35650766e-02
1.08448482e+00 -2.00830176e-01 -4.97972786e-01 -2.70626485e-01
-5.41452765e-01 -1.32235855e-01 8.05718124e-01 7.86617547e-02
9.73470867e-01 -1.33505285e+00 -4.99791175e-01 3.81923109e-01
8.75132605e-02 2.56685615e-01 8.65831196e-01 5.80005527e-01
-5.86526513e-01 -2.09567398e-02 -4.35099691e-01 -6.03329360e-01
-7.32756078e-01 8.04573298e-01 5.01090586e-01 -9.62642506e-02
-1.06614530e+00 4.22404081e-01 1.79471597e-01 -3.48393083e-01
-3.42208589e-03 2.13131443e-01 -1.45392179e-01 -2.46738404e-01
7.39115119e-01 2.50731975e-01 -1.70208588e-01 -6.69482946e-01
1.19426981e-01 6.91675067e-01 4.64749187e-02 -3.30233052e-02
1.50453007e+00 -4.36124414e-01 -9.42971185e-02 4.80857283e-01
1.24801779e+00 -4.04697180e-01 -1.70201921e+00 -3.46257299e-01
-5.92824757e-01 -6.34769440e-01 4.38868225e-01 -7.03706503e-01
-1.55228150e+00 7.34834492e-01 9.34467614e-01 4.92132977e-02
1.37601018e+00 1.94441855e-01 1.11504865e+00 -2.82970890e-02
2.58121192e-01 -1.05621433e+00 6.91727519e-01 -4.99473847e-02
7.39815950e-01 -1.56245887e+00 2.96686236e-02 -3.58402908e-01
-6.31333768e-01 1.06019199e+00 7.04962492e-01 -2.04506472e-01
4.11678463e-01 1.72229707e-01 1.14708155e-01 -2.12768078e-01
-4.90108609e-01 -3.55107576e-01 -1.00888237e-02 7.64080822e-01
-2.38839481e-02 -1.54242024e-01 -1.02607675e-01 2.42968962e-01
2.99150914e-01 2.22458124e-01 7.31666028e-01 7.04421163e-01
-6.31388366e-01 -7.08110809e-01 -1.85537562e-01 4.76748139e-01
-1.51985437e-01 1.56876259e-02 1.63944587e-01 4.37213540e-01
2.01437429e-01 8.49920094e-01 1.20373644e-01 -4.81221169e-01
1.52942725e-02 -3.97373855e-01 4.88325030e-01 -4.66454566e-01
2.86833402e-02 2.64168352e-01 -3.74169350e-01 -4.37201291e-01
-8.14675689e-01 -2.41525561e-01 -1.20382833e+00 -3.51876199e-01
-2.82794923e-01 -2.95513541e-01 4.50401455e-01 8.26638043e-01
1.58728391e-01 5.66476643e-01 9.95289266e-01 -9.46093976e-01
-2.34512806e-01 -7.41424561e-01 -9.51952279e-01 6.43003464e-01
4.45486248e-01 -6.98853433e-01 -6.94265246e-01 1.59091190e-01]
|
[10.860905647277832, -2.214545965194702]
|
2abd0637-83de-485a-95fc-daa92794ec16
|
relational-space-time-query-in-long-form
| null | null |
http://openaccess.thecvf.com//content/CVPR2023/html/Yang_Relational_Space-Time_Query_in_Long-Form_Videos_CVPR_2023_paper.html
|
http://openaccess.thecvf.com//content/CVPR2023/papers/Yang_Relational_Space-Time_Query_in_Long-Form_Videos_CVPR_2023_paper.pdf
|
Relational Space-Time Query in Long-Form Videos
|
Egocentric videos are often available in the form of uninterrupted, uncurated long videos capturing the camera wearers' daily life activities.Understanding these videos requires models to be able to reason about activities, objects, and their interactions. However, current video benchmarks study these problems independently and under short, curated clips. In contrast, real-world applications, e.g., AR assistants, require bundling these problems for both model development and evaluation. In this paper, we propose to study these problems in a joint framework for long video understanding. Our contributions are three-fold. First, we propose an integrated framework, namely Relational Space-Time Query (ReST), for evaluating video understanding models via templated spatiotemporal queries. Second, we introduce two new benchmarks, ReST-ADL and ReST-Ego4D, which augment the existing egocentric video datasets with abundant query annotations generated by the ReST framework. Finally, we present a set of baselines and in-depth analysis on the two benchmarks and provide insights about the query tasks. We view our integrated framework and benchmarks as a step towards comprehensive, multi-step reasoning in long videos, and believe it will facilitate the development of next generations of video understanding models.
|
['Du Tran', 'Lorenzo Torresani', 'Raghav Goyal', 'Matt Feiszli', 'Fu-Jen Chu', 'Xitong Yang']
|
2023-01-01
| null | null | null |
cvpr-2023-1
|
['video-understanding']
|
['computer-vision']
|
[ 3.94864790e-02 -7.68987983e-02 -4.81866807e-01 -5.27640998e-01
-6.05718672e-01 -7.48979628e-01 5.87832868e-01 -4.91787314e-01
-1.32095441e-01 1.78761795e-01 7.24576354e-01 7.14024678e-02
-1.15037397e-01 -3.89654726e-01 -9.75977004e-01 -9.93411243e-03
-1.14697061e-01 3.69337946e-01 2.34522954e-01 4.09095213e-02
4.88135852e-02 2.15539023e-01 -1.77763355e+00 6.31210089e-01
3.19893181e-01 1.06062353e+00 1.03304274e-01 1.00815487e+00
5.10075212e-01 1.57303011e+00 -1.83159605e-01 -6.19075179e-01
8.67223889e-02 -1.65572092e-01 -1.11727202e+00 3.37340236e-01
8.29291523e-01 -1.11181998e+00 -1.02774012e+00 7.60431230e-01
8.12584758e-02 4.54191655e-01 4.15436566e-01 -1.61427617e+00
-5.10795891e-01 4.07056093e-01 -1.98999286e-01 2.88595796e-01
1.02625847e+00 2.80870378e-01 1.18114424e+00 -8.51141155e-01
1.11641586e+00 1.35029185e+00 4.70389962e-01 4.50111330e-01
-7.97122002e-01 -4.35479730e-01 4.71659720e-01 9.27415371e-01
-1.28702092e+00 -9.38035786e-01 7.02294946e-01 -5.21455526e-01
1.12032700e+00 2.96612054e-01 7.04886377e-01 1.56942689e+00
-2.70510644e-01 1.38589752e+00 2.52062529e-01 7.34085515e-02
2.57282909e-02 -2.17835233e-01 1.22540720e-01 6.52964830e-01
-1.20020002e-01 -2.54824877e-01 -1.04155993e+00 7.70586208e-02
8.31962466e-01 3.54356945e-01 -3.50110710e-01 -7.94130027e-01
-1.45450437e+00 4.86557543e-01 1.88209459e-01 4.03552353e-02
-4.56915885e-01 5.76662958e-01 6.22194052e-01 1.41203582e-01
3.71651351e-01 1.48358047e-01 -2.33855933e-01 -9.41032946e-01
-7.80000865e-01 5.35861492e-01 9.49919045e-01 1.63601780e+00
4.29989189e-01 -4.66378480e-01 -2.33158901e-01 5.38666308e-01
1.66583151e-01 4.51438338e-01 -1.56039968e-02 -1.73015475e+00
5.70527554e-01 3.63409787e-01 1.89017862e-01 -1.15543902e+00
-1.25308156e-01 -8.38773027e-02 -1.76290229e-01 -6.30546153e-01
1.29199922e-01 2.20606387e-01 -4.44503695e-01 1.78610444e+00
1.12441234e-01 6.56634152e-01 -2.20400952e-02 9.87509370e-01
7.77382791e-01 4.21510130e-01 7.56940693e-02 -1.54054351e-03
1.56999600e+00 -1.51307023e+00 -9.38851297e-01 -2.14771390e-01
7.80820310e-01 -3.87611359e-01 1.13466763e+00 3.72254282e-01
-1.44661736e+00 -7.02974081e-01 -7.95394421e-01 -6.03782892e-01
-4.29395795e-01 3.19611393e-02 7.06924915e-01 6.63962588e-02
-1.05046701e+00 2.76790202e-01 -1.28038299e+00 -7.70573854e-01
6.47472143e-01 1.24379292e-01 -5.88283718e-01 -3.80709082e-01
-9.44400609e-01 4.71049517e-01 1.54959619e-01 1.06059592e-02
-1.38422430e+00 -7.08414674e-01 -1.16850615e+00 1.17669413e-02
8.41521382e-01 -8.76319885e-01 1.80748630e+00 -7.23421872e-01
-8.94403875e-01 9.15389180e-01 -4.91561919e-01 -4.84667420e-01
5.54297745e-01 -8.99824321e-01 -2.58037597e-01 6.24414802e-01
2.57030278e-01 8.31816554e-01 5.58768094e-01 -9.06211495e-01
-5.84392726e-01 -5.92391253e-01 9.25229132e-01 4.33564305e-01
-3.99603873e-01 8.49681869e-02 -1.34228122e+00 -5.91921389e-01
-2.53964573e-01 -1.01605558e+00 2.97757685e-01 1.94726855e-01
-2.56084740e-01 -1.85309201e-01 1.26357651e+00 -6.31433010e-01
1.30361915e+00 -2.23177052e+00 5.15348494e-01 -3.81002188e-01
3.73850882e-01 -1.39145732e-01 -5.51152825e-02 4.76165861e-01
-4.36450131e-02 -4.88404185e-02 2.25284249e-01 -7.14557827e-01
2.35210717e-01 3.28041077e-01 -4.23825443e-01 2.99648106e-01
-1.00765809e-01 1.15379775e+00 -1.18291175e+00 -6.36465430e-01
3.87406349e-01 4.84823555e-01 -8.72722268e-01 4.28267509e-01
-4.09226924e-01 1.84693024e-01 -6.27998352e-01 9.80499625e-01
1.36258572e-01 -4.61524159e-01 -9.82970893e-02 -6.82175696e-01
3.48171204e-01 3.38166282e-02 -7.74885118e-01 2.39704299e+00
-3.92015010e-01 9.51912701e-01 -7.55026415e-02 -7.78945446e-01
8.33596885e-02 4.95661378e-01 8.44065666e-01 -5.89437783e-01
-6.92908987e-02 -3.22224617e-01 -6.08442366e-01 -1.11407351e+00
6.59128368e-01 4.38681453e-01 -5.67509532e-02 4.77158576e-01
1.75707206e-01 3.59319933e-02 4.71202344e-01 7.60423124e-01
1.27704978e+00 6.11921012e-01 5.92538156e-02 3.24675232e-01
4.15229589e-01 4.44934778e-02 3.06549251e-01 8.82632792e-01
-4.76683319e-01 7.31489360e-01 6.42390370e-01 -5.25858700e-01
-9.02778208e-01 -9.11986828e-01 3.06824207e-01 1.36332786e+00
4.31358784e-01 -1.01727498e+00 -8.21349263e-01 -8.94119143e-01
-1.79162726e-01 6.26440823e-01 -5.39940298e-01 -1.34615019e-01
-5.80111921e-01 -4.44029532e-02 5.34148812e-01 1.01383877e+00
7.36128688e-01 -9.06845808e-01 -7.29159296e-01 -9.73729566e-02
-7.19836652e-01 -1.91136837e+00 -6.35473609e-01 -5.96518040e-01
-7.15511382e-01 -1.50794041e+00 -3.71091574e-01 -3.56602967e-01
3.40646446e-01 7.90940821e-01 1.51944649e+00 2.49414216e-03
-1.28084987e-01 1.52044725e+00 -6.22034669e-01 -3.41506183e-01
2.41480246e-01 -3.48917767e-02 1.50462285e-01 4.79807593e-02
7.53334939e-01 -5.95690310e-01 -8.07895303e-01 5.33516824e-01
-9.46572304e-01 2.43129522e-01 2.68996388e-01 3.87215316e-01
5.98296285e-01 -2.48035982e-01 6.52144775e-02 -8.85084152e-01
2.96961188e-01 -7.14416504e-01 -1.37633875e-01 4.85870600e-01
7.38466680e-02 -2.57960677e-01 1.87195212e-01 -2.77486473e-01
-1.11218858e+00 6.01475537e-02 1.20988950e-01 -1.12771034e+00
-1.51324406e-01 3.46341640e-01 -2.94047892e-01 2.33513340e-01
4.45816040e-01 1.65912807e-01 -2.32256219e-01 -3.59640867e-01
5.40241361e-01 4.57093924e-01 9.75088477e-01 -7.00923324e-01
6.05746925e-01 9.03220356e-01 -3.55823994e-01 -8.46353590e-01
-1.05201995e+00 -8.52411389e-01 -7.00674832e-01 -4.21201617e-01
1.16233826e+00 -1.39360237e+00 -9.73669052e-01 2.83626407e-01
-1.27541602e+00 -4.56736267e-01 -1.50308639e-01 3.32056820e-01
-1.25618386e+00 5.41556001e-01 -4.82407749e-01 -6.37882590e-01
8.27522129e-02 -1.27428746e+00 1.59816408e+00 3.26855271e-03
-4.37268257e-01 -1.04805231e+00 -1.66440248e-01 1.02149367e+00
1.34369254e-01 1.60515606e-01 3.72241884e-01 -5.72597146e-01
-1.04861724e+00 -1.72953948e-01 -3.75050247e-01 1.17336400e-01
1.32603357e-02 -2.05653653e-01 -1.05543709e+00 -1.54506803e-01
4.10478078e-02 -7.14381218e-01 6.61169827e-01 2.20791265e-01
1.60288382e+00 -1.82635248e-01 -4.24444288e-01 8.39038789e-01
9.38105822e-01 9.67461467e-02 6.95000589e-01 5.18689789e-02
8.71411264e-01 4.16946083e-01 9.86930788e-01 4.22399223e-01
9.25390601e-01 9.27575171e-01 6.69164360e-01 2.72070974e-01
-1.13886394e-01 -4.18009043e-01 4.53611881e-01 6.12403810e-01
-5.06305814e-01 -3.63430291e-01 -7.15001881e-01 6.73784733e-01
-2.23705268e+00 -1.34154975e+00 2.52909869e-01 1.62942874e+00
2.09156826e-01 -8.58081132e-02 1.70672789e-01 -2.95467317e-01
2.64640987e-01 5.15531600e-01 -9.07258809e-01 1.83814526e-01
8.20968747e-02 -3.29244494e-01 1.57789797e-01 3.18110734e-01
-1.27941823e+00 1.03995121e+00 6.39353228e+00 2.66665518e-01
-4.92000967e-01 1.19722508e-01 3.10630381e-01 -5.89041889e-01
-1.55104068e-03 -1.19112888e-02 -5.88581443e-01 1.05094038e-01
8.32295716e-01 -1.24923050e-01 6.95097506e-01 1.04351425e+00
3.62503201e-01 -2.20406689e-02 -1.82461083e+00 1.57915998e+00
5.30023336e-01 -1.21446300e+00 1.89952403e-01 -6.75538033e-02
5.06472528e-01 1.25068992e-01 -1.00939557e-01 4.59944338e-01
8.84348676e-02 -7.90479481e-01 8.10316682e-01 7.91294873e-01
6.52587295e-01 -3.01560163e-01 5.19644082e-01 1.44750312e-01
-1.41457260e+00 -2.90309578e-01 1.78715661e-02 -7.75926635e-02
4.75677133e-01 -1.98284641e-01 -2.59271681e-01 6.30265236e-01
1.06511855e+00 1.57005155e+00 -6.89384699e-01 5.46278119e-01
3.80554646e-02 2.52145827e-01 -1.90230161e-02 5.34867704e-01
1.09087832e-01 -4.04158123e-02 5.13801694e-01 1.03393137e+00
1.71165720e-01 4.06950414e-01 1.57551855e-01 7.34517932e-01
-2.57985055e-01 -3.95172596e-01 -9.08145487e-01 -3.80907685e-01
4.51660067e-01 9.21903551e-01 -2.30428413e-01 -6.06745183e-01
-8.97974432e-01 1.15477931e+00 2.88834840e-01 6.64246202e-01
-1.22041214e+00 -1.09014280e-01 1.04387057e+00 1.88087568e-01
1.30042568e-01 -3.54360670e-01 3.55395079e-01 -1.73509693e+00
2.63070613e-01 -1.08175051e+00 6.40730202e-01 -1.37796366e+00
-9.90946651e-01 2.49817744e-01 5.69966018e-01 -1.09424150e+00
-6.75812542e-01 -5.38719952e-01 -3.31152469e-01 6.81808079e-03
-1.32308555e+00 -1.50272548e+00 -1.00299728e+00 9.67765033e-01
1.23355198e+00 9.72292200e-02 5.11448622e-01 4.91319716e-01
-5.41626930e-01 2.73886174e-01 -3.94847900e-01 2.98197091e-01
8.20988238e-01 -1.03218031e+00 3.97128910e-01 6.49177372e-01
4.61140186e-01 6.49866223e-01 6.43784285e-01 -5.06729484e-01
-1.89921868e+00 -9.60746884e-01 5.19540370e-01 -1.09399092e+00
8.03668678e-01 -5.61103582e-01 -5.78766584e-01 1.54047632e+00
9.90424007e-02 1.86561286e-01 6.09846056e-01 1.53717950e-01
-3.98116946e-01 -1.29152343e-01 -5.72449088e-01 6.81965113e-01
1.78520906e+00 -1.03541207e+00 -5.58872461e-01 6.76348507e-01
9.27613676e-01 -4.91280764e-01 -1.05719328e+00 4.10198778e-01
9.55909193e-01 -1.25928164e+00 1.28214717e+00 -9.61621225e-01
6.75204217e-01 -6.58301413e-02 -4.50135797e-01 -6.69644773e-01
1.22579131e-02 -6.81794941e-01 -7.33868062e-01 9.71948981e-01
-1.36362493e-01 9.18909311e-02 9.53593314e-01 9.77123022e-01
-1.81065083e-01 -5.78931570e-01 -3.83271962e-01 -6.52092874e-01
-7.85387933e-01 -9.43863273e-01 4.96588737e-01 7.20207930e-01
2.93862615e-02 1.95163876e-01 -5.53794563e-01 2.43904553e-02
5.54467976e-01 -1.71868414e-01 1.26168358e+00 -8.69687736e-01
-2.65358508e-01 -4.02829945e-02 -6.97989523e-01 -1.64730108e+00
3.44799012e-01 -2.90188253e-01 -2.37415478e-01 -1.41336942e+00
5.70896626e-01 1.91922858e-01 -3.87527533e-02 2.43912399e-01
1.03409216e-01 2.76864797e-01 2.63340145e-01 3.28604281e-01
-1.51683247e+00 5.07127106e-01 1.12159002e+00 -1.46448046e-01
-2.43303981e-02 -1.42631888e-01 -5.12738109e-01 9.24369395e-01
1.40431523e-01 -1.16152596e-02 -1.11236405e+00 -9.31127250e-01
3.28731656e-01 1.76039636e-01 7.06797719e-01 -1.12717485e+00
6.02771997e-01 -1.45144209e-01 5.05859889e-02 -7.95758188e-01
9.75369632e-01 -9.91488636e-01 1.82366759e-01 -8.41639936e-02
-4.28549141e-01 2.23023519e-01 -3.54606397e-02 9.26142395e-01
-4.95718420e-01 2.88431913e-01 2.37824628e-03 -3.46380621e-01
-1.16312587e+00 6.06810033e-01 7.45664835e-02 2.12293699e-01
1.20657909e+00 -3.98153871e-01 -3.75311106e-01 -1.12310481e+00
-9.08887863e-01 6.28374875e-01 5.92769444e-01 8.92238021e-01
6.30327761e-01 -1.30708671e+00 -3.50863397e-01 -1.39376029e-01
5.62927783e-01 6.56777620e-02 6.01496220e-01 8.77673626e-01
-5.33335209e-01 7.57650316e-01 -2.73463316e-02 -7.89252937e-01
-1.36538160e+00 7.33833253e-01 1.53557494e-01 8.83607864e-02
-5.22503316e-01 6.48586690e-01 6.98119998e-01 -1.84624214e-02
6.98826253e-01 -3.31860960e-01 -1.93690032e-01 -9.85191315e-02
7.10996568e-01 4.51764643e-01 -2.84737110e-01 -8.38171780e-01
-2.23256901e-01 5.75975955e-01 -1.63799658e-01 3.87163237e-02
1.29578948e+00 -4.47278112e-01 7.65293092e-02 5.05735934e-01
1.24497235e+00 -4.54645753e-01 -1.46313739e+00 -2.06818059e-01
-3.34886052e-02 -6.83743417e-01 -1.42625540e-01 -2.98198789e-01
-9.31200922e-01 7.47520030e-01 2.01797083e-01 -2.63940706e-03
1.26982653e+00 3.94195020e-01 8.33799720e-01 8.25913548e-01
3.40892076e-01 -9.97665882e-01 4.99512702e-01 2.74134010e-01
9.70664561e-01 -1.26812553e+00 1.05862655e-01 -5.61472118e-01
-7.11340547e-01 9.35667098e-01 8.80467772e-01 1.98153287e-01
4.98468637e-01 -1.49087429e-01 -3.03470880e-01 -6.34251595e-01
-1.09445310e+00 -1.26912281e-01 3.12699229e-01 5.31244040e-01
2.39161298e-01 -2.46288449e-01 4.40468192e-01 5.62668085e-01
-5.98721169e-02 4.08646196e-01 3.96017164e-01 1.06796300e+00
7.37758577e-02 -7.32964337e-01 -5.99487647e-02 2.01075628e-01
-3.04194033e-01 3.58293891e-01 -4.41903770e-01 9.38767016e-01
-1.33175656e-01 9.73411143e-01 3.69510859e-01 -4.35470343e-01
5.96213579e-01 -2.10891236e-02 5.35951138e-01 -5.03814161e-01
3.34193558e-02 -2.71532476e-01 2.77385741e-01 -1.41754222e+00
-8.14799607e-01 -6.52813852e-01 -7.71609366e-01 -2.39393562e-01
4.83685313e-03 -1.06103577e-01 4.02853876e-01 1.00117838e+00
6.13352060e-01 3.39695215e-01 1.07527010e-01 -8.90816510e-01
-1.99485332e-01 -8.37344229e-01 -2.52496153e-01 8.26904058e-01
2.90564746e-01 -8.57682765e-01 -1.09030172e-01 6.14845753e-01]
|
[9.899053573608398, 0.8202782869338989]
|
8a3987f2-f95c-434a-8671-3ec768d1155d
|
understanding-satirical-articles-using-common
| null | null |
https://aclanthology.org/Q16-1038
|
https://aclanthology.org/Q16-1038.pdf
|
Understanding Satirical Articles Using Common-Sense
|
Automatic satire detection is a subtle text classification task, for machines and at times, even for humans. In this paper we argue that satire detection should be approached using common-sense inferences, rather than traditional text classification methods. We present a highly structured latent variable model capturing the required inferences. The model abstracts over the specific entities appearing in the articles, grouping them into generalized categories, thus allowing the model to adapt to previously unseen situations.
|
['Xiao Zhang', 'Dan Goldwasser']
|
2016-01-01
| null | null | null |
tacl-2016-1
|
['satire-detection']
|
['natural-language-processing']
|
[ 2.09537238e-01 1.81647107e-01 -6.14457548e-01 -3.82279903e-01
-3.08286071e-01 -7.91608810e-01 8.16361666e-01 7.10730612e-01
-4.36612934e-01 7.21396565e-01 6.48817718e-01 -5.61007679e-01
7.62112290e-02 -5.96129000e-01 1.56333774e-01 -4.76712942e-01
1.98153690e-01 7.53505170e-01 1.33578360e-01 -7.18257204e-02
7.70201147e-01 3.44395697e-01 -1.30150962e+00 5.32907665e-01
6.13745987e-01 6.97950006e-01 -1.23076051e-01 6.01381361e-01
-2.20818713e-01 1.17078102e+00 -7.38834918e-01 -7.42518187e-01
-1.48637116e-01 -3.11957985e-01 -8.48009944e-01 1.15870781e-01
2.18360275e-01 -3.84099841e-01 -2.34964311e-01 1.26645601e+00
4.15302515e-02 3.71308267e-01 9.07587230e-01 -1.16691256e+00
-7.77731776e-01 1.05042505e+00 -2.84010857e-01 2.74011195e-01
6.41597807e-01 -2.90579021e-01 1.37803590e+00 -7.88274765e-01
6.57556474e-01 1.57247066e+00 5.56591213e-01 4.59366441e-01
-1.35039341e+00 -6.79329813e-01 3.78817976e-01 3.33071589e-01
-1.19470620e+00 -3.74512047e-01 8.64859879e-01 -7.49842048e-01
7.79386222e-01 3.91652197e-01 2.29979590e-01 1.53115475e+00
5.64423501e-01 7.87498534e-01 1.19228089e+00 -3.16200167e-01
6.44737005e-01 5.18716335e-01 6.00251257e-01 5.77044547e-01
5.63562512e-01 -5.56287050e-01 -6.56760991e-01 -7.91313827e-01
2.52933502e-01 2.01876089e-01 -2.60179758e-01 -1.92124546e-01
-1.07269239e+00 1.30778968e+00 -7.64873549e-02 5.06073236e-01
-5.97108424e-01 -2.47253045e-01 5.72229505e-01 2.44948462e-01
8.14734042e-01 5.99674165e-01 -7.36328602e-01 -2.34704018e-02
-1.15467489e+00 3.36130321e-01 1.21579278e+00 6.41607642e-01
8.17222670e-02 -2.06801087e-01 -6.51552528e-02 5.59330106e-01
4.43700939e-01 2.05181539e-01 7.57869482e-01 -4.98324126e-01
1.17733628e-01 7.44214058e-01 3.40363592e-01 -1.46964371e+00
-3.14140856e-01 -2.69843102e-01 -6.34908557e-01 1.50305405e-01
2.86012769e-01 -8.71949643e-03 -4.85445797e-01 1.33502614e+00
-8.15301910e-02 -3.88571680e-01 1.78290904e-01 5.33680201e-01
6.83657765e-01 4.25338298e-01 5.10918319e-01 -5.72895825e-01
1.60296190e+00 -6.16995513e-01 -1.02390254e+00 -2.71911204e-01
6.14502668e-01 -8.67784083e-01 8.73711050e-01 8.15590799e-01
-8.84985387e-01 2.84370743e-02 -1.07794631e+00 -2.33164176e-01
-5.78552485e-01 -2.13436469e-01 6.61603212e-01 4.55045104e-01
-5.18263400e-01 7.76871562e-01 -5.19781113e-01 -3.27643037e-01
2.54099339e-01 5.44198118e-02 -8.72851461e-02 3.92933786e-01
-1.46354294e+00 1.11377680e+00 5.30188441e-01 -2.17639029e-01
-3.69414538e-01 -1.02683179e-01 -9.00545299e-01 1.42959401e-01
6.43571138e-01 -6.39841676e-01 1.28889453e+00 -8.94016623e-01
-1.40864110e+00 1.18721068e+00 -3.27721208e-01 -4.55793053e-01
5.28620601e-01 -2.80055732e-01 -5.27767837e-01 1.19460799e-01
1.50904611e-01 -4.90901759e-03 1.16386521e+00 -1.08231032e+00
-3.34762305e-01 -3.29394370e-01 6.59885630e-02 1.34118930e-01
-1.98702335e-01 4.06137824e-01 3.93790573e-01 -7.83734858e-01
4.00884181e-01 -7.69201279e-01 -2.40843371e-01 -1.91777915e-01
-4.53981072e-01 -6.61175191e-01 8.28028440e-01 -5.80609500e-01
1.28923726e+00 -2.22605801e+00 1.60067588e-01 6.15957007e-03
7.45610356e-01 -1.66584462e-01 6.22573853e-01 3.73059273e-01
-7.71495625e-02 3.58896434e-01 2.58026749e-01 -3.66675496e-01
4.10160571e-01 2.32362702e-01 -7.41380155e-01 6.51448071e-01
-1.91188827e-01 6.57963812e-01 -1.10589242e+00 -5.29032946e-01
-2.24587682e-04 6.08625007e-04 -3.60021919e-01 -1.62128851e-01
-3.88663083e-01 2.76650079e-02 -7.65767336e-01 5.14381945e-01
2.46094763e-01 -4.61996228e-01 2.92042494e-01 7.81584233e-02
-4.30309176e-02 7.04884291e-01 -8.76317680e-01 1.03845799e+00
-1.05388686e-01 8.68462026e-01 -3.28977290e-03 -8.32596898e-01
7.55795419e-01 4.68132228e-01 2.19055906e-01 -8.80689546e-02
5.63630044e-01 -8.75413511e-03 -2.95901984e-01 -3.31537813e-01
6.93122625e-01 -7.91865945e-01 -6.04296803e-01 5.79054296e-01
-1.68263212e-01 1.06394373e-01 -5.99205792e-02 5.60778439e-01
7.64356077e-01 -4.94977683e-01 9.74378169e-01 -6.53611839e-01
4.14235055e-01 7.59077370e-02 3.75036091e-01 1.01422298e+00
-2.61044472e-01 -1.64618287e-02 6.26757622e-01 -5.91504216e-01
-6.34709537e-01 -1.05174315e+00 -4.25958902e-01 1.37230694e+00
1.24703854e-01 -6.24496102e-01 -2.28217572e-01 -8.16481352e-01
1.54578418e-01 1.27103770e+00 -6.65352046e-01 -1.53825238e-01
4.91390862e-02 -4.57485825e-01 2.82805681e-01 3.74082923e-01
6.60096481e-02 -9.61185694e-01 -8.78043056e-01 2.09511355e-01
-2.85299122e-01 -9.82804596e-01 -1.74925834e-01 5.31024814e-01
-6.86019003e-01 -8.75015199e-01 -1.47779822e-01 -5.41875303e-01
4.08071160e-01 3.08184549e-02 1.08917570e+00 2.16391638e-01
-3.09715997e-02 2.49942780e-01 -4.39952642e-01 -5.12796700e-01
-7.36903071e-01 -2.68973887e-01 3.74381483e-01 -2.12655246e-01
8.59299362e-01 -2.64721006e-01 -1.46047741e-01 1.48437649e-01
-7.69889832e-01 4.69260514e-02 2.27560133e-01 8.40547502e-01
2.05571391e-02 3.09628963e-01 1.35792673e-01 -1.13293183e+00
8.89505744e-01 -5.14002383e-01 -2.24469572e-01 1.24551825e-01
-6.77917004e-01 1.25015303e-01 4.59201276e-01 -8.03412616e-01
-9.25106287e-01 -3.76774162e-01 1.28898054e-01 -8.45074877e-02
-3.26945812e-01 6.82056546e-01 -1.96122393e-01 4.07647938e-01
8.69533062e-01 -1.24652624e-01 -4.90389735e-01 -4.72359955e-01
5.36962986e-01 9.94372129e-01 2.52091944e-01 -3.62322718e-01
6.32189155e-01 4.19067979e-01 -1.39339253e-01 -8.53088856e-01
-1.31936288e+00 -7.33797014e-01 -8.33040595e-01 -1.81108695e-02
8.27509463e-01 -7.22801208e-01 -6.15585804e-01 -2.09923517e-02
-1.17474365e+00 -1.76794052e-01 -7.73012862e-02 3.98373187e-01
-3.80154699e-01 4.96731937e-01 -9.13920999e-01 -9.23116982e-01
-1.55141264e-01 -6.50961578e-01 7.93375850e-01 -1.51496261e-01
-1.24970555e+00 -1.41880953e+00 2.97288969e-02 3.86696965e-01
1.67454347e-01 1.21610701e-01 1.05217290e+00 -1.42320001e+00
1.65639788e-01 -3.86163831e-01 -3.99338417e-02 -9.15532280e-03
2.17224583e-01 -1.17335007e-01 -8.98702621e-01 -1.44708186e-01
6.91170394e-01 -3.23890626e-01 9.04315174e-01 1.53972268e-01
9.42461610e-01 -8.51186037e-01 -5.41875124e-01 -1.08220801e-01
1.01885867e+00 -5.42313568e-02 2.91406572e-01 2.35408381e-01
3.28865141e-01 4.94318694e-01 2.38731220e-01 6.06502175e-01
1.32340804e-01 3.93681705e-01 -1.23579443e-01 1.33815214e-01
5.25530696e-01 -2.43654862e-01 2.81977534e-01 6.24338984e-01
3.36997777e-01 -3.58383358e-01 -8.07934999e-01 2.03996405e-01
-1.88358331e+00 -1.30669558e+00 -3.79619628e-01 1.89224315e+00
9.37409282e-01 6.77102268e-01 -1.74896151e-01 2.26507768e-01
7.46559381e-01 1.42954782e-01 -2.16720462e-01 -6.87841892e-01
-1.34509146e-01 3.92257795e-02 4.25619066e-01 7.78442144e-01
-1.15703142e+00 9.67712879e-01 7.84685993e+00 4.67468768e-01
-9.68764007e-01 -1.83529295e-02 4.12175030e-01 1.76690191e-01
-3.27296436e-01 2.08520636e-01 -6.50033653e-01 2.69409329e-01
5.62799752e-01 -6.90457225e-01 -5.56202829e-02 1.23752964e+00
2.08226576e-01 -2.36459121e-01 -1.19486129e+00 8.15030217e-01
6.00670040e-01 -8.58164787e-01 3.59465867e-01 -1.35576665e-01
5.34159422e-01 -4.06722605e-01 -7.97242522e-02 3.53574574e-01
5.60779750e-01 -8.95212948e-01 7.00865507e-01 3.01229447e-01
1.61532648e-02 -2.60162681e-01 7.38587320e-01 5.68868041e-01
-4.63775873e-01 -1.50143355e-01 -2.42059022e-01 -6.98112547e-01
2.64803646e-03 5.30594051e-01 -6.71872616e-01 -9.59364101e-02
3.85662824e-01 6.47647798e-01 -6.89750969e-01 4.81379658e-01
-5.60131013e-01 5.86254776e-01 -4.31667566e-02 -4.16711837e-01
9.75004882e-02 8.66281092e-02 7.64047980e-01 1.33594322e+00
-1.46459222e-01 3.66677463e-01 5.75850964e-01 7.81359911e-01
-3.81016508e-02 2.60601580e-01 -5.72872519e-01 -2.06867561e-01
2.44831502e-01 1.22196889e+00 -1.06064022e+00 -9.73746777e-01
-2.50901282e-01 1.23084140e+00 7.40146339e-02 2.59523213e-01
-4.43614990e-01 -1.64859965e-01 2.45985016e-01 -3.98365222e-03
1.64450482e-02 -1.86590761e-01 -5.22118032e-01 -1.60835385e+00
-3.62652510e-01 -7.67062366e-01 7.53174424e-01 -8.37866545e-01
-1.72377098e+00 2.66545922e-01 1.74753949e-01 -7.64741004e-01
-2.09306240e-01 -7.67599761e-01 -5.52006841e-01 5.09080708e-01
-8.77035797e-01 -8.29741418e-01 1.33760646e-01 4.36059594e-01
6.35870099e-01 1.47150502e-01 1.04327619e+00 -2.65428185e-01
-3.61534804e-01 2.29896292e-01 1.63659099e-02 2.43302613e-01
8.37565899e-01 -1.40776420e+00 1.40903249e-01 5.97899556e-01
3.55076551e-01 1.26540017e+00 1.50515914e+00 -8.32986116e-01
-6.23613894e-01 -5.14830351e-01 1.51117516e+00 -8.16782176e-01
1.40770912e+00 -5.41219115e-01 -1.13663840e+00 7.95347631e-01
3.16129804e-01 -7.55992353e-01 1.15545106e+00 5.85741818e-01
-7.63730645e-01 6.37123585e-01 -1.04040396e+00 7.61394978e-01
5.19153714e-01 -7.67452955e-01 -1.51840830e+00 4.80995595e-01
3.62766564e-01 7.97960237e-02 -4.37114745e-01 5.33204600e-02
5.38741469e-01 -3.73033285e-01 6.33572817e-01 -1.09260154e+00
5.06417036e-01 -1.27817720e-01 -1.61459371e-01 -1.14085019e+00
-8.04492831e-01 -6.17405415e-01 -1.56832621e-01 7.98191607e-01
4.64671642e-01 -4.64092642e-01 5.42887270e-01 9.59377170e-01
4.67662007e-01 -2.67745219e-02 -7.20544755e-01 -4.48334575e-01
2.27109492e-01 -1.92900062e-01 -1.34743735e-01 1.63630629e+00
9.78923857e-01 1.01170444e+00 -4.89651889e-01 2.46560331e-02
5.87957084e-01 3.69356275e-01 2.63855219e-01 -1.79011810e+00
-4.06869560e-01 -8.09722722e-01 -6.37233078e-01 -1.13763475e+00
3.54181319e-01 -7.75921345e-01 8.25848654e-02 -1.27504194e+00
5.96068561e-01 1.68582752e-01 -2.64220536e-01 5.48591256e-01
-3.79015505e-01 1.97396755e-01 -1.21527240e-01 3.79135311e-01
-9.61655855e-01 3.16435635e-01 7.40670860e-01 -2.73911089e-01
-2.19613940e-01 -3.31525058e-02 -1.04563022e+00 1.13650489e+00
1.09523833e+00 -7.98163533e-01 -1.32848650e-01 3.16126764e-01
6.85431838e-01 -1.74519226e-01 5.83065212e-01 -5.52216411e-01
3.38198841e-01 -5.27301729e-01 3.88777524e-01 -5.80249727e-01
-8.52244627e-03 -6.15443408e-01 -1.45624146e-01 4.43582296e-01
-8.14287305e-01 -1.27308205e-01 -2.67754178e-02 6.08031034e-01
9.32918712e-02 -4.93452877e-01 7.46930122e-01 -2.48007849e-01
-4.66310561e-01 -3.58320475e-01 -1.04531312e+00 -7.40190148e-02
7.09940076e-01 1.12299353e-01 -2.14747459e-01 -5.45218945e-01
-1.17532432e+00 -1.04722701e-01 5.21056652e-01 4.28498477e-01
3.65561426e-01 -9.84543204e-01 -6.24272287e-01 -2.04697236e-01
3.89280796e-01 -5.69115579e-01 -1.58582285e-01 5.58262169e-01
-9.02170464e-02 3.33282501e-01 1.10326864e-01 -1.70564696e-01
-1.38373041e+00 1.15945673e+00 -1.74777478e-01 -7.55568370e-02
-8.48052502e-01 4.46868181e-01 2.17196941e-01 2.40027592e-01
2.04488799e-01 -2.07934052e-01 -5.39974332e-01 2.29003400e-01
7.53090382e-01 7.05486685e-02 -2.77291864e-01 -6.80620790e-01
-3.56658727e-01 -1.03706867e-01 -4.61437970e-01 -6.67264014e-02
8.63615096e-01 -2.46587768e-01 -3.19973320e-01 1.25499547e+00
8.68260741e-01 4.46261317e-01 -3.43899488e-01 -3.14380616e-01
5.12869298e-01 -3.83326858e-01 1.48623973e-01 -8.49706054e-01
-4.23314534e-02 5.72212875e-01 -8.73238891e-02 6.68875277e-01
5.42269349e-01 3.26268584e-01 3.29225868e-01 7.05005705e-01
2.50422418e-01 -1.10121977e+00 5.47121689e-02 4.19029474e-01
7.87487507e-01 -1.08459640e+00 5.40496171e-01 -4.51044530e-01
-7.33177602e-01 1.25562048e+00 -6.60490915e-02 -1.94778398e-01
6.40229106e-01 2.06609741e-01 1.85320258e-01 -5.26482522e-01
-1.05348265e+00 1.91431463e-01 3.07768047e-01 1.00301117e-01
6.72555864e-01 1.84608549e-01 -6.98858321e-01 8.41380000e-01
-3.36929053e-01 -2.34458447e-01 6.28097951e-01 9.03726995e-01
-7.40336478e-01 -9.21255231e-01 -3.00421864e-01 5.56888878e-01
-7.03474283e-01 -2.81270266e-01 -1.07919931e+00 3.93866390e-01
-6.98600411e-02 1.14324987e+00 -2.06205770e-02 -1.48873672e-01
1.13758840e-01 4.32926595e-01 4.38935272e-02 -7.57035673e-01
-4.80629683e-01 7.35308230e-02 1.68124199e-01 -8.43505040e-02
-2.66815722e-01 -7.44240284e-01 -9.89645123e-01 -2.73108989e-01
-6.76171303e-01 4.50946599e-01 3.57384324e-01 1.35766745e+00
-1.08259954e-01 8.02943558e-02 4.28332746e-01 -2.97861069e-01
-8.73574734e-01 -1.00277019e+00 -6.20604217e-01 7.85430908e-01
3.02013427e-01 -8.18692029e-01 -9.10976052e-01 2.92179435e-01]
|
[8.91259479522705, 10.025059700012207]
|
58e0c618-fed0-43d7-baec-5c2d58230697
|
analyzing-the-influence-of-dataset
|
2103.03700
| null |
https://arxiv.org/abs/2103.03700v1
|
https://arxiv.org/pdf/2103.03700v1.pdf
|
Analyzing the Influence of Dataset Composition for Emotion Recognition
|
Recognizing emotions from text in multimodal architectures has yielded promising results, surpassing video and audio modalities under certain circumstances. However, the method by which multimodal data is collected can be significant for recognizing emotional features in language. In this paper, we address the influence data collection methodology has on two multimodal emotion recognition datasets, the IEMOCAP dataset and the OMG-Emotion Behavior dataset, by analyzing textual dataset compositions and emotion recognition accuracy. Experiments with the full IEMOCAP dataset indicate that the composition negatively influences generalization performance when compared to the OMG-Emotion Behavior dataset. We conclude by discussing the impact this may have on HRI experiments.
|
['S. Wermter', 'C. Weber', 'S. Magg', 'A. Sutherland']
|
2021-03-05
| null | null | null | null |
['multimodal-emotion-recognition', 'multimodal-emotion-recognition']
|
['computer-vision', 'speech']
|
[-1.30513132e-01 -3.25302243e-01 -2.25544333e-01 -5.97487271e-01
-5.60771167e-01 -6.48922741e-01 5.90203822e-01 1.55694604e-01
-6.30103528e-01 6.41412914e-01 5.35532892e-01 2.50835270e-02
-1.59471527e-01 -2.85753876e-01 -1.67288527e-01 -4.90495682e-01
-1.38239339e-02 8.57062936e-02 -8.34953487e-01 -3.50887090e-01
2.32657969e-01 3.68959188e-01 -1.78292322e+00 7.77411342e-01
3.17117691e-01 1.36207104e+00 -5.39931655e-01 8.88973534e-01
-2.84673303e-01 1.15125167e+00 -7.18525350e-01 -5.59495509e-01
-3.43051493e-01 -3.72967929e-01 -8.24906588e-01 1.82354182e-01
1.27250403e-01 -8.12228471e-02 -3.20060432e-01 5.62097132e-01
7.12840497e-01 3.85683745e-01 9.16469395e-01 -1.51276588e+00
-3.29160064e-01 6.21940315e-01 -1.17078394e-01 6.54495433e-02
9.34294343e-01 -2.72700399e-01 8.57352972e-01 -9.61130500e-01
9.25514638e-01 1.23632538e+00 4.69693065e-01 7.06087232e-01
-9.67589974e-01 -4.75448906e-01 -1.30497143e-01 1.41953200e-01
-1.39540827e+00 -7.99582839e-01 8.25720191e-01 -2.62159973e-01
1.23673248e+00 3.61089319e-01 3.71565163e-01 1.48196268e+00
7.45714922e-03 8.90561283e-01 1.16518891e+00 -4.43180114e-01
1.63626999e-01 5.29224694e-01 6.20106995e-01 2.45656118e-01
-3.12291533e-01 -3.23045194e-01 -8.61517131e-01 -3.17131877e-01
-1.99266955e-01 -4.06016350e-01 -2.66935468e-01 8.95343274e-02
-7.83498883e-01 7.74072468e-01 -1.29063874e-01 4.91187990e-01
-4.13994431e-01 2.26749550e-03 1.00325263e+00 8.10926199e-01
3.74877244e-01 4.34637576e-01 -5.79142928e-01 -7.67203867e-01
-5.67803621e-01 -1.17048502e-01 1.04660320e+00 6.06833696e-01
5.01580536e-01 1.53072208e-01 1.39620960e-01 1.31609678e+00
4.29909118e-02 4.08284277e-01 5.87941945e-01 -6.25927806e-01
3.27437878e-01 6.28503859e-01 8.48183334e-02 -1.14332855e+00
-7.32553005e-01 2.78445512e-01 -3.91304761e-01 -3.12935054e-01
4.01264697e-01 -7.04214036e-01 -5.59680462e-01 1.70581198e+00
-1.74533755e-01 -6.71077728e-01 5.49948156e-01 9.20338511e-01
1.40252328e+00 7.71256506e-01 4.63055938e-01 -1.90531209e-01
1.27303028e+00 -3.91929567e-01 -9.74582076e-01 -2.22325787e-01
9.58018839e-01 -7.36399949e-01 1.06768572e+00 7.68689394e-01
-7.86261857e-01 -2.83045769e-01 -7.85349667e-01 5.25795780e-02
-8.54279041e-01 4.11122382e-01 6.77481771e-01 1.04311800e+00
-7.34756231e-01 1.29834330e-02 -3.35245013e-01 -6.72230303e-01
2.88183112e-02 2.57914245e-01 -7.16251135e-01 8.80189687e-02
-1.24240124e+00 9.51568007e-01 5.14143348e-01 1.14246324e-01
-5.69378614e-01 -9.65169519e-02 -9.26516294e-01 5.12394309e-02
9.74963158e-02 1.25378311e-01 1.06908166e+00 -1.48628223e+00
-1.50973666e+00 8.30329716e-01 -3.00090134e-01 -9.27921683e-02
1.45997956e-01 -1.11372843e-02 -1.11487377e+00 4.38547939e-01
-7.41891563e-01 7.43418276e-01 4.73059118e-01 -1.18857241e+00
-3.40360940e-01 -5.34225106e-01 -1.89856738e-01 3.32999885e-01
-7.87087023e-01 4.05482322e-01 -3.76486123e-01 -2.59858251e-01
-1.60355002e-01 -1.06784415e+00 3.38393152e-01 -9.34184015e-01
-1.79309428e-01 -3.65076482e-01 9.10302341e-01 -5.82681119e-01
1.39937615e+00 -2.36428881e+00 2.06616729e-01 4.90890890e-01
-1.70706734e-01 -2.67320454e-01 -3.87802124e-01 7.63967216e-01
-2.36260414e-01 4.44602698e-01 4.17515963e-01 -2.09051982e-01
1.19764082e-01 1.36114135e-01 -2.36688629e-01 1.65751994e-01
1.56364560e-01 8.01328957e-01 -3.18705142e-01 -4.67448264e-01
4.04861309e-02 6.58888042e-01 -2.67718524e-01 1.54750139e-01
1.08230166e-01 8.17532912e-02 -2.35673323e-01 9.56060767e-01
1.83281168e-01 1.20187953e-01 3.11102778e-01 -2.35996246e-01
2.27468438e-03 1.42620085e-02 -6.77217484e-01 1.46001184e+00
-5.20698845e-01 1.10649920e+00 -2.91314553e-02 -8.05699170e-01
1.25950563e+00 7.28811204e-01 5.11079192e-01 -8.23919475e-01
6.69591010e-01 -4.60600033e-02 3.31218839e-02 -7.69905865e-01
1.11622417e+00 -3.43957961e-01 -6.67659283e-01 3.78161222e-01
1.81703791e-01 7.39025921e-02 3.73650998e-01 2.01707840e-01
6.83921576e-01 -4.44373369e-01 2.79465653e-02 3.44966985e-02
3.89058322e-01 1.67757079e-01 3.90141696e-01 6.48599565e-01
-4.05596942e-01 2.82022327e-01 7.77931452e-01 -4.71027717e-02
-5.30059040e-01 -6.01530552e-01 -3.55401814e-01 1.31043112e+00
-2.28797510e-01 -6.17807209e-01 -4.61761236e-01 -6.92372620e-01
-2.37239212e-01 9.33732152e-01 -6.68426156e-01 -4.89144921e-01
1.73713416e-01 -7.48164475e-01 9.99191403e-01 4.79418516e-01
1.02389015e-01 -1.04462492e+00 -5.80357611e-01 -7.79697374e-02
-5.29241741e-01 -1.17473817e+00 3.52364033e-02 3.65734667e-01
-5.98221183e-01 -7.48950183e-01 -4.29772794e-01 -6.20317996e-01
3.09666723e-01 9.69409291e-03 1.15319407e+00 -1.93494424e-01
-1.20409474e-01 1.24288177e+00 -8.07377577e-01 -5.98836780e-01
-3.25591743e-01 2.25433350e-01 -8.51152465e-03 6.91931173e-02
8.88833463e-01 -2.67324410e-02 5.34938788e-03 4.05232191e-01
-1.03676498e+00 -2.13266626e-01 1.95673302e-01 5.35494208e-01
7.97096938e-02 2.81450897e-01 8.03647935e-01 -6.87274456e-01
1.02383971e+00 -6.88651383e-01 2.73687422e-01 3.86371642e-01
-3.16816568e-01 -3.62344325e-01 4.30630207e-01 -5.45227051e-01
-1.30343437e+00 -1.42968699e-01 1.28049508e-01 -3.61416876e-01
-4.92776871e-01 1.01661372e+00 -1.67812228e-01 -5.22394478e-02
4.14322555e-01 -8.38473141e-02 8.81694816e-03 -2.66085178e-01
3.41535769e-02 1.09637415e+00 2.14709774e-01 -6.06308579e-01
-1.06635235e-01 1.56681582e-01 -5.14780343e-01 -1.30801773e+00
-2.32072800e-01 -3.58874887e-01 -2.52306610e-01 -7.49553502e-01
8.90875757e-01 -9.20818686e-01 -8.76291871e-01 4.76277560e-01
-7.78815687e-01 -1.55705824e-01 2.07310647e-01 8.34404111e-01
-3.74478430e-01 -1.12558663e-01 -8.59277844e-01 -1.19547331e+00
-1.30766854e-01 -9.64828372e-01 6.99687779e-01 2.63943821e-01
-7.88146734e-01 -1.10448503e+00 -7.75319636e-02 5.17949700e-01
3.27287436e-01 1.78489000e-01 9.40239310e-01 -9.01166916e-01
5.37307262e-01 -4.43530113e-01 -2.19699964e-02 3.58077586e-01
-2.40416408e-01 2.77468830e-01 -1.15837348e+00 -1.14558060e-02
-1.45217568e-01 -1.21490884e+00 6.72470927e-01 1.27903551e-01
9.72234547e-01 1.13213927e-01 1.03624657e-01 -3.68575528e-02
1.26115358e+00 5.11017561e-01 6.85476840e-01 2.54167229e-01
3.95022839e-01 1.07017815e+00 6.83516026e-01 6.03220403e-01
2.48341575e-01 2.13538006e-01 -5.41391671e-02 1.66756570e-01
4.78571236e-01 1.64745390e-01 7.14196026e-01 8.90143216e-01
1.60373867e-01 -5.73915839e-01 -1.17005134e+00 2.36735359e-01
-1.66308033e+00 -9.32181537e-01 1.54462770e-01 1.67829180e+00
4.84958231e-01 -4.04739857e-01 3.26073527e-01 2.96105981e-01
3.72154593e-01 2.10450485e-01 -1.89088717e-01 -1.21986139e+00
-5.82646549e-01 -3.78484994e-01 4.88332473e-02 2.41793126e-01
-9.17398870e-01 5.86650312e-01 6.86136055e+00 4.07470465e-01
-1.31764185e+00 -2.80285627e-01 8.36733103e-01 -5.36215246e-01
-3.01807076e-01 -4.85809296e-01 -5.70439875e-01 2.81180471e-01
1.41853511e+00 -2.43554905e-01 4.06567961e-01 7.32613504e-01
2.20367879e-01 -4.15885329e-01 -1.17443073e+00 1.35949492e+00
6.99413002e-01 -4.41492230e-01 -4.88608740e-02 -7.01048672e-02
2.22051471e-01 -7.04845190e-02 5.16762547e-02 6.93062246e-01
-2.53763765e-01 -1.08681214e+00 5.63243270e-01 7.03376472e-01
5.97649336e-01 -1.07942545e+00 7.70868540e-01 4.72876802e-02
-6.96778595e-01 -1.57785580e-01 -7.94901475e-02 -4.94747534e-02
-9.46097597e-02 1.48751453e-01 -7.09966660e-01 3.88433427e-01
8.34630668e-01 5.20443857e-01 -7.64055908e-01 4.76656526e-01
4.00653452e-01 7.84781933e-01 -1.94756195e-01 -4.35470432e-01
2.71892458e-01 -2.50403315e-01 3.42579454e-01 1.75489664e+00
2.94658333e-01 1.62143528e-01 -1.37760222e-01 1.67773366e-01
-2.55683184e-01 4.90397781e-01 -8.61516535e-01 -9.32266474e-01
1.54993609e-01 1.41247773e+00 -6.53831482e-01 -2.72513896e-01
-7.15135098e-01 7.78278887e-01 1.84065655e-01 6.54934645e-01
-6.82695150e-01 -5.21951616e-01 4.89326864e-01 -5.31566083e-01
-1.09517261e-01 -6.40770793e-02 -2.08767951e-01 -1.19778585e+00
-2.90675372e-01 -1.23750365e+00 7.55463660e-01 -1.06240213e+00
-1.22712004e+00 5.12045681e-01 4.56547178e-03 -9.40299273e-01
-2.80578613e-01 -7.24351883e-01 -2.32693091e-01 4.88862514e-01
-8.55868876e-01 -8.14279795e-01 -2.89712489e-01 5.99895179e-01
2.55626887e-01 -2.99389690e-01 9.20304775e-01 4.91590053e-01
-8.70258808e-01 7.72250116e-01 -1.35280788e-01 2.97451645e-01
1.09622192e+00 -8.10957670e-01 -8.69793415e-01 2.27204397e-01
1.84739362e-02 4.56357926e-01 6.32695794e-01 -2.93091267e-01
-1.92178559e+00 -4.57722396e-01 7.80175328e-01 -5.51285148e-01
5.16692698e-01 -1.26745537e-01 -7.85280406e-01 6.72182798e-01
5.14086783e-01 -7.36387432e-01 1.32275641e+00 6.11624002e-01
-4.41002488e-01 3.01163532e-02 -1.12853825e+00 6.84629440e-01
5.94393611e-01 -1.03888929e+00 -4.70365345e-01 -3.39340210e-01
1.45037666e-01 -5.88847287e-02 -1.30594516e+00 3.96600246e-01
8.62361431e-01 -7.34267533e-01 5.73176146e-01 -1.04272699e+00
8.15655947e-01 2.47434944e-01 -5.15750527e-01 -1.42922139e+00
2.86222864e-02 1.25725549e-02 1.15832143e-01 1.67303491e+00
7.40892649e-01 -4.54221398e-01 7.64297962e-01 1.39430344e+00
2.11090565e-01 -4.16721255e-01 -6.83559418e-01 -3.65480483e-01
-4.63540331e-02 -9.59938824e-01 3.90546918e-01 1.21716189e+00
6.16894543e-01 4.46919531e-01 -5.14792264e-01 -4.04463708e-01
-7.26424679e-02 -1.59029588e-01 8.41732204e-01 -7.44551539e-01
9.23513025e-02 -5.12540519e-01 -5.18447995e-01 -1.29305601e-01
6.24242544e-01 -8.63889754e-01 -2.09853813e-01 -9.26688790e-01
2.32022732e-01 7.45064691e-02 -3.86750460e-01 5.46812296e-01
5.22361957e-02 3.06628078e-01 3.34584773e-01 -8.16357210e-02
-7.74915338e-01 5.31273782e-01 6.53781652e-01 -1.45699933e-01
-3.84995103e-01 -6.16638184e-01 -6.06776893e-01 6.80344284e-01
1.05013251e+00 -1.77919835e-01 -4.06781226e-01 -2.51315445e-01
5.54299176e-01 2.21541584e-01 5.95751405e-03 -7.89062798e-01
1.06692791e-01 1.49203360e-01 5.64987063e-01 -4.82754558e-01
6.57906353e-01 -9.48731303e-01 8.66665319e-02 -1.88028529e-01
-8.05166483e-01 2.66263872e-01 7.26568043e-01 2.68383145e-01
-5.93102276e-01 -4.09134448e-01 2.82204151e-01 2.96622097e-01
-1.08888471e+00 -2.44655773e-01 -1.12318027e+00 1.11886181e-01
6.21134102e-01 -1.79354295e-01 -2.83834815e-01 -8.68682742e-01
-8.31856191e-01 2.28968099e-01 4.50280488e-01 7.87343919e-01
5.99910796e-01 -1.34741116e+00 -4.11544621e-01 -1.61488116e-01
5.97933352e-01 -1.14205849e+00 3.49669576e-01 1.07581389e+00
1.46732226e-01 4.33763385e-01 -3.90940040e-01 -2.92462736e-01
-1.65309000e+00 1.94315866e-01 2.40272626e-01 2.84844935e-01
1.12487316e-01 4.43798453e-01 -3.14068168e-01 -4.87999320e-01
5.85972965e-01 1.65815920e-01 -5.31379521e-01 8.18042576e-01
5.26937068e-01 5.49207211e-01 2.28988584e-02 -8.95704806e-01
-3.23417366e-01 1.84660688e-01 -1.71552911e-01 -6.59335315e-01
1.04366302e+00 -3.26511949e-01 -1.12272806e-01 1.18514824e+00
1.61418283e+00 -3.81842218e-02 -2.75925756e-01 1.07093714e-01
2.19124675e-01 -1.45309702e-01 2.27312729e-01 -1.10365212e+00
-8.53885651e-01 7.62797058e-01 6.78711534e-01 3.33158821e-01
1.20599079e+00 -7.98212513e-02 4.61778402e-01 8.31636131e-01
9.70688313e-02 -1.72068059e+00 -3.56511213e-02 7.40098417e-01
9.15172577e-01 -1.47778058e+00 -4.45272267e-01 5.39513640e-02
-1.43582571e+00 1.34729409e+00 5.17135620e-01 4.87598807e-01
6.39323890e-01 1.89864635e-01 5.09602368e-01 -3.76045674e-01
-1.02105331e+00 3.63471135e-02 3.38928998e-01 2.26922855e-01
1.03482759e+00 1.51182696e-01 -4.99439269e-01 8.27705801e-01
-1.88909307e-01 7.95829520e-02 4.47282642e-01 9.60329235e-01
-7.30325580e-02 -9.65459347e-01 -6.71856999e-01 6.37719870e-01
-6.09658659e-01 1.20551452e-01 -1.36177576e+00 9.16091323e-01
-4.46407318e-01 1.43277240e+00 2.41769195e-01 -8.10867548e-01
3.15206438e-01 7.50537276e-01 2.26459011e-01 -1.56170785e-01
-7.92181075e-01 4.00363002e-04 1.07555199e+00 -4.81962383e-01
-5.22721469e-01 -8.26832235e-01 -1.20076764e+00 -3.45699519e-01
-8.91010538e-02 4.20033902e-01 8.22740555e-01 6.88740492e-01
4.13587332e-01 2.68889844e-01 5.64940929e-01 -7.36489952e-01
-4.11025882e-02 -9.93403256e-01 -6.74454510e-01 6.69061959e-01
-1.63034901e-01 -3.67070943e-01 -6.31821156e-01 -1.66622758e-01]
|
[13.201798439025879, 5.544878005981445]
|
ab08cc4b-a155-4b6e-a02d-fc194dcdd0d9
|
hybrid-fusion-based-interpretable-multimodal
|
2208.11450
| null |
https://arxiv.org/abs/2208.11450v1
|
https://arxiv.org/pdf/2208.11450v1.pdf
|
Hybrid Fusion Based Interpretable Multimodal Emotion Recognition with Insufficient Labelled Data
|
This paper proposes a multimodal emotion recognition system, VIsual Spoken Textual Additive Net (VISTA Net), to classify the emotions reflected by a multimodal input containing image, speech, and text into discrete classes. A new interpretability technique, K-Average Additive exPlanation (KAAP), has also been developed to identify the important visual, spoken, and textual features leading to predicting a particular emotion class. The VISTA Net fuses the information from image, speech & text modalities using a hybrid of early and late fusion. It automatically adjusts the weights of their intermediate outputs while computing the weighted average without human intervention. The KAAP technique computes the contribution of each modality and corresponding features toward predicting a particular emotion class. To mitigate the insufficiency of multimodal emotion datasets labeled with discrete emotion classes, we have constructed a large-scale IIT-R MMEmoRec dataset consisting of real-life images, corresponding speech & text, and emotion labels ('angry,' 'happy,' 'hate,' and 'sad.'). The VISTA Net has resulted in 95.99% emotion recognition accuracy on considering image, speech, and text modalities, which is better than the performance on considering the inputs of any one or two modalities.
|
['Balasubramanian Raman', 'Sarthak Malik', 'Puneet Kumar']
|
2022-08-24
| null | null | null | null |
['multimodal-emotion-recognition', 'multimodal-emotion-recognition']
|
['computer-vision', 'speech']
|
[ 3.15090775e-01 5.98260108e-03 4.58077453e-02 -6.14467144e-01
-6.19985223e-01 -3.09799075e-01 8.46081138e-01 1.27925158e-01
-3.53701651e-01 4.00664985e-01 4.17948216e-01 8.09665993e-02
-2.62572709e-02 -1.35630235e-01 -9.77655873e-02 -7.32905924e-01
2.44895771e-01 7.83932880e-02 -5.27267754e-01 -1.27209112e-01
3.70022058e-01 3.95269156e-01 -1.92790437e+00 1.06429315e+00
7.57562399e-01 1.58100438e+00 -8.85211602e-02 1.02878940e+00
-5.39577425e-01 1.26011169e+00 -6.37812674e-01 -5.35758913e-01
-2.63231307e-01 -6.48924530e-01 -4.89147097e-01 4.12858695e-01
1.17235161e-01 1.18942903e-02 -8.91220011e-03 8.65571678e-01
5.06451130e-01 2.19609186e-01 9.67532456e-01 -1.87497914e+00
-8.74710619e-01 3.34219843e-01 -6.43331766e-01 -1.30047202e-01
5.67318678e-01 -2.10208874e-02 6.94132149e-01 -1.30529845e+00
4.94340092e-01 1.43319249e+00 3.33452374e-01 4.81982619e-01
-7.45645106e-01 -3.68340731e-01 1.18822008e-01 3.42893839e-01
-1.19704103e+00 -5.13671875e-01 9.64095533e-01 -4.42787141e-01
1.03866732e+00 6.57672703e-01 5.26011527e-01 9.69080150e-01
2.84452528e-01 1.03362215e+00 1.28411400e+00 -5.97734034e-01
4.65442911e-02 7.66214788e-01 3.28584880e-01 6.93961561e-01
-7.70256639e-01 -1.91955164e-01 -8.65207136e-01 -1.27795637e-01
1.04463525e-01 1.28776863e-01 -1.64228976e-01 1.49534613e-01
-1.24396431e+00 4.75128233e-01 2.46820763e-01 2.46591091e-01
-6.01843953e-01 -2.02674627e-01 5.75092733e-01 4.91842419e-01
6.89302742e-01 -2.73314249e-02 -4.25807774e-01 -2.12267235e-01
-4.60910439e-01 -4.64165479e-01 4.81335461e-01 5.12305558e-01
7.81144202e-01 2.71641999e-01 -2.23856017e-01 1.16420448e+00
4.33537394e-01 6.50190771e-01 5.06051004e-01 -9.89538431e-01
2.08754733e-01 1.00776863e+00 -2.50144657e-02 -1.36444736e+00
-3.78973126e-01 1.66841835e-01 -9.54227507e-01 3.80720675e-01
-1.85988292e-01 -2.64701873e-01 -1.04726326e+00 1.45773089e+00
1.14207231e-02 -3.01639736e-01 7.45556593e-01 1.05967772e+00
1.41312170e+00 1.08495092e+00 3.24325621e-01 -5.35708129e-01
1.57671297e+00 -9.09688890e-01 -1.18202364e+00 -1.87460020e-01
2.68087268e-01 -8.97644401e-01 1.01565468e+00 6.45783663e-01
-7.93625891e-01 -7.17857897e-01 -7.28996634e-01 6.16632365e-02
-8.02078247e-01 6.87694430e-01 5.58735609e-01 4.68845010e-01
-9.17654812e-01 6.99128360e-02 -7.26695731e-02 -3.80395949e-01
-1.12814426e-01 2.51222312e-01 -7.61941969e-01 1.69446066e-01
-1.13723230e+00 9.77334261e-01 1.75157249e-01 3.85162830e-01
-5.55825770e-01 -1.53588846e-01 -9.81707156e-01 -8.14872235e-03
9.61162671e-02 -2.43595898e-01 8.59563887e-01 -2.07667017e+00
-1.54645872e+00 7.58446097e-01 -4.51399237e-01 1.68527424e-01
7.80618861e-02 -5.84703796e-02 -9.34806943e-01 4.32121038e-01
-2.98911214e-01 9.14433420e-01 1.07671928e+00 -1.53902400e+00
-5.61763883e-01 -5.28306663e-01 -3.79340976e-01 5.95697165e-01
-4.54410881e-01 3.71688396e-01 -3.86812955e-01 -4.76673752e-01
1.36824772e-01 -5.46418130e-01 4.31345180e-02 -5.32694273e-02
-2.20402271e-01 -3.93615425e-01 1.15181243e+00 -8.03407609e-01
9.63663995e-01 -2.51791549e+00 3.08882177e-01 2.72915602e-01
1.09367311e-01 1.95406880e-02 -1.64908335e-01 3.62853646e-01
-4.05859053e-01 1.62578020e-02 -1.40539467e-01 -5.07242799e-01
7.47050121e-02 2.86953896e-01 -2.11945817e-01 4.29056101e-02
3.62733245e-01 6.34525239e-01 -6.36033058e-01 -6.67120934e-01
5.46411514e-01 7.26588905e-01 5.72226718e-02 4.89388913e-01
1.50946110e-01 5.43901809e-02 -1.21276677e-01 8.88876498e-01
4.55450118e-01 9.99743864e-02 5.50523996e-02 -5.34562409e-01
-1.29543200e-01 -7.09132850e-01 -1.04404700e+00 1.12959647e+00
-2.68869340e-01 9.38107908e-01 -2.03072093e-02 -7.29283214e-01
1.24499214e+00 6.30947232e-01 3.96617740e-01 -7.56241262e-01
5.01990438e-01 3.19238529e-02 -4.30155158e-01 -1.06819308e+00
7.12028980e-01 -1.15748093e-01 -3.57336223e-01 2.58175373e-01
1.98226988e-01 1.05059005e-01 -2.69138873e-01 3.24879259e-01
5.20127833e-01 -1.51121065e-01 1.98921114e-01 2.26346835e-01
7.16376185e-01 -6.04058541e-02 3.25971007e-01 5.08848906e-01
-3.66546929e-01 4.89101529e-01 6.50673211e-01 -4.22951013e-01
-6.67909324e-01 -9.44715500e-01 4.21950400e-01 1.36550498e+00
2.56736070e-01 -1.62686452e-01 -4.29916263e-01 -6.08138859e-01
-2.72757679e-01 7.23688781e-01 -7.76959181e-01 -2.96764344e-01
2.42827028e-01 -6.79023385e-01 4.23874825e-01 3.30529988e-01
6.40042961e-01 -1.26790345e+00 -4.07272100e-01 -2.25730881e-01
-5.04338801e-01 -1.00787354e+00 -2.79582024e-01 4.66524288e-02
-2.45942146e-01 -7.99061537e-01 -5.77490568e-01 -5.50952792e-01
9.11225796e-01 3.94786000e-02 7.57196784e-01 -2.14232460e-01
-1.93260655e-01 9.75578010e-01 -6.62429273e-01 -4.08023089e-01
-2.43190765e-01 -7.39292562e-01 -5.56065440e-02 8.21605086e-01
2.07884684e-01 2.32111686e-03 -3.27347666e-01 1.59183741e-01
-8.71896625e-01 4.19228315e-01 6.02594137e-01 6.75946593e-01
3.71508718e-01 4.65681888e-02 5.12506187e-01 -3.91952813e-01
8.42452407e-01 -6.92502737e-01 3.12724560e-01 5.19157708e-01
-3.68914634e-01 -2.53611386e-01 4.17969346e-01 -7.02860534e-01
-1.47329700e+00 1.79925233e-01 -1.26543082e-03 -6.38040304e-01
-4.64348644e-01 6.90313101e-01 -1.96297154e-01 1.08408742e-01
3.25150013e-01 2.43016601e-01 1.63269728e-01 3.50601040e-02
3.83930266e-01 1.07336771e+00 7.74021387e-01 -1.43693686e-01
4.14721109e-02 1.84110701e-01 -3.73105794e-01 -8.48542988e-01
-3.65692258e-01 -3.51460516e-01 -2.63033986e-01 -1.15963519e+00
1.05829430e+00 -7.85093784e-01 -9.22336757e-01 7.57141113e-01
-1.23577380e+00 2.09347576e-01 8.52294713e-02 6.12978339e-01
-3.21031898e-01 3.63050610e-01 -5.95048487e-01 -1.42129672e+00
-5.81910014e-01 -1.11960435e+00 9.71659362e-01 5.44070542e-01
-3.81842315e-01 -6.63538814e-01 -4.74161834e-01 4.28351253e-01
2.11396396e-01 2.55758792e-01 8.48297477e-01 -5.91439903e-01
1.18756711e-01 -2.97390223e-01 -6.41575277e-01 5.65337539e-01
-2.47813109e-02 5.59756815e-01 -1.09673703e+00 2.11900786e-01
-1.69011056e-01 -4.81502444e-01 7.32255042e-01 8.07582662e-02
9.28767323e-01 -2.71998316e-01 1.23605296e-01 3.32044587e-02
1.15188682e+00 6.08266354e-01 6.02335036e-01 -2.40234926e-01
5.77314317e-01 7.38036394e-01 8.61500740e-01 6.63998544e-01
4.41735953e-01 3.34563732e-01 5.28819442e-01 -3.73758942e-01
2.04255939e-01 1.82393894e-01 8.52501214e-01 1.01609528e+00
-1.64067164e-01 -5.14503002e-01 -7.01659322e-01 3.19445163e-01
-2.00147438e+00 -1.06651890e+00 -1.93028137e-01 1.59921038e+00
3.37139547e-01 -3.24651927e-01 -2.27515742e-01 2.22854272e-01
7.81025112e-01 2.54435377e-04 -3.72798294e-01 -1.09052312e+00
-5.47544420e-01 -2.54325479e-01 -1.99254081e-01 5.12705863e-01
-8.86403620e-01 7.37805068e-01 5.87435436e+00 7.46109307e-01
-1.35321367e+00 -8.07314292e-02 8.24330986e-01 1.66524425e-01
-4.36731368e-01 -3.03351969e-01 -5.38156033e-02 3.76818240e-01
9.90555584e-01 4.83797900e-02 4.70168531e-01 8.69515657e-01
3.03571552e-01 -4.00715321e-01 -6.00985527e-01 1.31447816e+00
7.41946280e-01 -8.34833384e-01 3.31208646e-01 -4.41334695e-01
2.83666283e-01 -4.02835488e-01 2.02143729e-01 4.05893266e-01
-1.39608830e-01 -8.74925852e-01 8.19802999e-01 1.03230298e+00
7.58165777e-01 -9.17651594e-01 9.11723256e-01 2.44296536e-01
-1.02385271e+00 -1.76712558e-01 1.14444412e-01 -2.43953951e-02
-6.04775026e-02 2.22566634e-01 -5.50043941e-01 5.47990501e-01
7.87073970e-01 5.92414796e-01 -5.38323224e-01 2.86589533e-01
6.43738806e-02 2.76701123e-01 -1.37499109e-01 -1.43857673e-01
3.44980396e-02 -1.08087547e-01 3.72941703e-01 1.35571456e+00
5.08110523e-01 4.88797456e-01 -8.78949929e-03 3.64953607e-01
7.37739578e-02 4.76783007e-01 -5.48144460e-01 -3.18025291e-01
1.31996304e-01 1.64392018e+00 -6.94650173e-01 -7.06245303e-01
-2.50361264e-01 1.29792702e+00 6.36927634e-02 5.31856656e-01
-8.73297811e-01 -5.29226840e-01 3.51724833e-01 -8.43783438e-01
-2.47161686e-01 2.37447232e-01 -2.96677679e-01 -9.51973796e-01
-2.19388857e-01 -8.05099308e-01 6.02671921e-01 -1.81440496e+00
-1.22337747e+00 9.18961465e-01 -1.50323927e-01 -1.13333142e+00
-1.69067815e-01 -7.07924664e-01 -5.60041249e-01 7.23680854e-01
-9.41192150e-01 -1.11062682e+00 -6.08064055e-01 7.20812798e-01
6.98629856e-01 -3.48845333e-01 8.32315564e-01 1.99929848e-01
-6.18048429e-01 3.88277024e-01 -4.00629491e-01 -8.23074579e-02
7.62766242e-01 -1.03017461e+00 -7.98587501e-01 4.91499215e-01
-1.83159471e-01 1.34073392e-01 8.34010124e-01 -3.47330451e-01
-1.46326613e+00 -6.23850286e-01 1.06056166e+00 -1.58549413e-01
2.74490952e-01 9.52677652e-02 -7.23870873e-01 3.56570601e-01
8.20077002e-01 -2.86722332e-01 9.01139736e-01 -1.13458715e-01
-2.99960107e-01 -2.44383693e-01 -1.34625888e+00 6.24665618e-01
2.56509811e-01 -7.21922576e-01 -5.97535610e-01 5.77324489e-03
6.02380395e-01 -6.66661635e-02 -1.01423395e+00 6.44628704e-01
7.53270030e-01 -1.08354175e+00 6.33048832e-01 -4.44491535e-01
7.71141291e-01 -3.09183866e-01 -3.88021052e-01 -1.32437003e+00
5.96125722e-02 -1.08879320e-01 1.65619820e-01 1.45490456e+00
5.59875607e-01 -4.94104505e-01 2.17961982e-01 8.66953492e-01
-2.30201229e-01 -5.29424012e-01 -7.65262365e-01 1.63398311e-01
-8.39502573e-01 -7.03465581e-01 3.56509656e-01 1.29288042e+00
4.54203159e-01 3.58854413e-01 -7.08639622e-01 2.34716788e-01
2.79266596e-01 -2.93296259e-02 5.31064689e-01 -8.06168199e-01
2.72452325e-01 -4.35910970e-01 -3.87342781e-01 -2.82320350e-01
2.30373815e-01 -6.25098944e-01 6.01975173e-02 -1.40397882e+00
3.82036179e-01 2.36923575e-01 -6.49729609e-01 8.59685957e-01
-1.03890710e-01 4.73945349e-01 3.60262692e-01 2.21154168e-02
-6.84848964e-01 7.20722139e-01 1.01632297e+00 -2.12679625e-01
-9.48955640e-02 -4.51272011e-01 -5.41243136e-01 8.51539254e-01
7.88717449e-01 -1.02183139e-02 -3.91347051e-01 -2.52322316e-01
9.27431658e-02 5.75712502e-01 4.18645740e-01 -6.48374438e-01
2.86894411e-01 -2.79643923e-01 7.28835762e-01 -8.90134752e-01
7.93776333e-01 -1.04916334e+00 3.93628210e-01 8.19033757e-02
-6.31584764e-01 3.42548937e-01 2.20046028e-01 2.84345359e-01
-6.57605767e-01 9.61911008e-02 6.31917536e-01 4.44155112e-02
-1.26146853e+00 -1.45785302e-01 -7.71340549e-01 -6.76397562e-01
1.01924276e+00 -4.82226312e-01 -4.27394956e-01 -7.11525321e-01
-1.20217657e+00 1.81525350e-01 -1.34549392e-02 6.87196434e-01
1.35347795e+00 -1.50402474e+00 -4.36357051e-01 1.79061189e-01
4.23828095e-01 -8.28965187e-01 8.42170656e-01 9.14351523e-01
-1.05468079e-01 -1.62678182e-01 -5.43040752e-01 -4.53668326e-01
-1.77294791e+00 3.76357973e-01 2.83411711e-01 2.02795640e-01
-1.87428351e-02 5.29066205e-01 7.71481218e-03 -4.22254086e-01
2.49649465e-01 1.98100761e-01 -6.91185594e-01 5.34595370e-01
4.69538510e-01 3.35685462e-01 -2.72026390e-01 -1.17937160e+00
-2.66765803e-01 4.90104765e-01 1.53261095e-01 -4.17059958e-01
9.56243277e-01 -4.88547385e-01 -4.05612886e-01 9.00321186e-01
1.17074633e+00 -2.19500929e-01 -6.33601248e-01 -4.21848781e-02
-3.19551229e-01 -1.86861962e-01 6.76074550e-02 -1.28547204e+00
-1.12712669e+00 9.31747496e-01 9.17413414e-01 2.65163839e-01
1.73918223e+00 -1.12034358e-01 2.93625742e-01 6.64927587e-02
-1.34635285e-01 -1.35278523e+00 3.93063188e-01 3.22401673e-01
1.18905699e+00 -1.35222876e+00 -2.48031899e-01 -2.71966547e-01
-1.48305738e+00 1.19613421e+00 7.44012594e-01 3.50867778e-01
4.06914800e-01 7.63041601e-02 4.96693760e-01 -2.19308868e-01
-1.03689766e+00 -1.58028781e-01 5.40360510e-01 2.57213086e-01
2.78373301e-01 1.54497832e-01 -1.54667320e-02 8.11775744e-01
2.88700849e-01 -7.05217719e-02 2.70564467e-01 6.68600738e-01
-4.43297416e-01 -4.33657408e-01 -5.81864893e-01 3.79644364e-01
-2.77897030e-01 1.95108019e-02 -9.04853344e-01 4.53762978e-01
3.49386871e-01 1.19077837e+00 1.82691172e-01 -8.99086714e-01
2.10967273e-01 3.54867041e-01 6.57026917e-02 5.31634688e-03
-6.65715873e-01 8.92289579e-02 2.12446183e-01 -3.84625852e-01
-7.70967901e-01 -4.43936318e-01 -1.34298551e+00 -2.81379759e-01
3.36843915e-02 2.33324438e-01 8.23309422e-01 9.67250526e-01
3.99863482e-01 4.54126000e-01 8.64575624e-01 -9.98254895e-01
2.21608534e-01 -8.81629586e-01 -4.15842682e-01 5.83382130e-01
2.10982293e-01 -4.27026033e-01 -6.32667899e-01 2.94066668e-01]
|
[13.250375747680664, 5.147392272949219]
|
80705e98-acd9-4e2f-8dd6-f15f22de40f1
|
identifying-the-causes-of-pyrocumulonimbus
|
2211.08883
| null |
https://arxiv.org/abs/2211.08883v3
|
https://arxiv.org/pdf/2211.08883v3.pdf
|
Identifying the Causes of Pyrocumulonimbus (PyroCb)
|
A first causal discovery analysis from observational data of pyroCb (storm clouds generated from extreme wildfires) is presented. Invariant Causal Prediction was used to develop tools to understand the causal drivers of pyroCb formation. This includes a conditional independence test for testing $Y$ conditionally independent of $E$ given $X$ for binary variable $Y$ and multivariate, continuous variables $X$ and $E$, and a greedy-ICP search algorithm that relies on fewer conditional independence tests to obtain a smaller more manageable set of causal predictors. With these tools, we identified a subset of seven causal predictors which are plausible when contrasted with domain knowledge: surface sensible heat flux, relative humidity at $850$ hPa, a component of wind at $250$ hPa, $13.3$ micro-meters, thermal emissions, convective available potential energy, and altitude.
|
['Nis Meinert', 'Paula Harder', 'Duncan Watson-Parris', 'Kara D. Lamb', 'Daniel Okoh', 'Ashwin Braude', 'Kenza Tazi', 'Emiliano Díaz Salas-Porras']
|
2022-11-16
| null | null | null | null |
['causal-discovery']
|
['knowledge-base']
|
[ 7.68308109e-03 -1.08586989e-01 -1.01165533e-01 -2.14647740e-01
-2.85088599e-01 -3.15465599e-01 5.80369115e-01 1.38682172e-01
4.48519830e-04 1.48682201e+00 2.90848404e-01 -9.46757615e-01
-5.92726290e-01 -1.47345901e+00 -5.85835755e-01 -6.62103951e-01
-1.06848824e+00 1.47184297e-01 -5.93771040e-02 -8.55185688e-02
2.55786180e-01 4.57107723e-01 -1.85213661e+00 -1.05674982e-01
8.66061330e-01 7.75175273e-01 1.11434281e-01 5.72578430e-01
1.61446318e-01 7.93409228e-01 -1.87082648e-01 3.94535094e-01
2.34114602e-01 -4.45034742e-01 -6.19402111e-01 -8.32160056e-01
-1.35916084e-01 -1.98931277e-01 1.97052270e-01 5.29639304e-01
1.08709959e-02 1.89509839e-01 8.87922347e-01 -1.34606266e+00
-7.24961385e-02 6.11923754e-01 -4.75164890e-01 5.26859999e-01
1.44739941e-01 3.57313067e-01 1.19060588e+00 -6.19052052e-01
6.28245354e-01 1.29036367e+00 5.24087012e-01 -4.61526290e-02
-1.36008549e+00 -1.31480396e+00 1.11039907e-01 3.78980897e-02
-1.43238688e+00 -1.63984805e-01 1.87643558e-01 -8.15284073e-01
1.32104278e+00 7.25412607e-01 8.91033947e-01 6.00616574e-01
4.15434122e-01 -2.81539500e-01 1.40317118e+00 -3.10206831e-01
4.84220326e-01 -1.06587633e-01 -5.21630161e-02 6.14222348e-01
3.44130635e-01 1.10447025e+00 -8.09606314e-01 -6.47314429e-01
4.36916322e-01 -3.07009667e-01 -2.65228868e-01 2.18593314e-01
-6.30412459e-01 9.74560022e-01 5.75595737e-01 -1.21249989e-01
-4.69063759e-01 5.13673246e-01 -2.17353478e-01 2.79384311e-02
5.66014051e-01 5.87060511e-01 -9.32297528e-01 -1.45904765e-01
-6.97592556e-01 6.07836366e-01 6.99681640e-01 5.84116042e-01
9.46073353e-01 2.56430537e-01 3.28231275e-01 4.82240990e-02
6.25147462e-01 1.47841132e+00 -2.87567973e-01 -8.38490486e-01
2.38846987e-01 3.65137190e-01 5.89397013e-01 -9.67908323e-01
-2.83636868e-01 1.89006776e-02 -6.52298868e-01 3.46749812e-01
-3.26064713e-02 -5.71752369e-01 -9.53755379e-01 1.71594417e+00
1.31757438e-01 1.59768507e-01 -1.08425558e-01 7.09431052e-01
3.02590340e-01 7.86757052e-01 6.58299923e-01 -6.07891381e-01
8.07579696e-01 6.45086765e-02 -3.48257720e-01 -3.76186997e-01
1.05196610e-01 -4.65636849e-01 6.71617329e-01 -4.20714244e-02
-6.36026502e-01 -2.06411347e-01 -7.58755863e-01 8.36044848e-01
-6.42930150e-01 -5.78368783e-01 9.77048337e-01 5.86645186e-01
-9.43275034e-01 8.39220822e-01 -9.31351125e-01 -3.01451296e-01
-1.30806267e-02 1.47120804e-01 1.92919327e-03 1.42738253e-01
-1.49780321e+00 8.76078069e-01 1.39492631e-01 2.17623115e-02
-1.06459868e+00 -1.02964079e+00 -5.12870073e-01 2.21552253e-01
-2.21923869e-02 -7.30473220e-01 7.12741375e-01 -3.78026903e-01
-9.37289238e-01 2.68543720e-01 -5.86137354e-01 -3.59879076e-01
1.07551530e-01 -2.67094821e-01 -5.83322465e-01 8.11935868e-03
3.21916491e-01 4.19033378e-01 7.01120675e-01 -1.16047430e+00
-8.46790612e-01 -4.57370967e-01 -3.83784264e-01 -9.81065445e-03
2.29281157e-01 1.60197362e-01 7.59899020e-01 -1.71849892e-01
1.39544562e-01 -7.73204327e-01 -6.27263248e-01 -5.90758860e-01
-3.36855680e-01 6.17570914e-02 6.16466522e-01 -6.28921270e-01
1.05678797e+00 -1.87648749e+00 -2.46961433e-02 6.61706984e-01
5.41813560e-02 -3.29145372e-01 3.11667860e-01 5.08702159e-01
-6.35681570e-01 9.96064186e-01 -5.50485253e-01 6.04288518e-01
-1.53002858e-01 2.00518101e-01 -8.17194283e-01 2.19361812e-01
5.60592771e-01 3.76530141e-01 -9.90251899e-01 -1.72120899e-01
5.19330442e-01 4.54419032e-02 -4.58628118e-01 1.43950656e-01
-4.90695328e-01 3.98648322e-01 -3.95235687e-01 6.24936640e-01
7.02140749e-01 3.00899334e-02 1.62921682e-01 6.85321331e-01
-9.98450756e-01 2.88589597e-01 -9.67597842e-01 5.46813071e-01
-1.37812451e-01 4.29579020e-01 1.86250344e-01 -6.57049358e-01
1.06657887e+00 3.79990309e-01 6.30998373e-01 -5.68842053e-01
-1.27986178e-01 2.09338918e-01 8.30535218e-02 -4.93893862e-01
4.96504642e-02 -8.02160323e-01 1.31810740e-01 4.18649942e-01
-6.04607344e-01 -5.62072515e-01 4.14649432e-04 1.66787207e-01
1.49021697e+00 -1.07331350e-01 3.25157166e-01 -8.27947319e-01
-1.07242232e-02 3.10277492e-01 7.89666414e-01 7.50910878e-01
-1.88169599e-01 3.70150693e-02 4.20326054e-01 -5.05068123e-01
-7.51836538e-01 -1.56236207e+00 -4.31760371e-01 7.04371572e-01
5.49957938e-02 -6.53704047e-01 2.10485667e-01 2.56214309e-02
3.87778580e-01 1.20939803e+00 -6.81452453e-01 1.50360599e-01
-4.92568940e-01 -1.43394601e+00 5.34597561e-02 3.92532736e-01
4.64133263e-01 -1.05892909e+00 -9.49348688e-01 7.95345306e-02
-3.28055501e-01 -3.73761326e-01 7.82697260e-01 7.55999386e-01
-1.11441314e+00 -1.27788174e+00 2.36209393e-01 1.76241532e-01
2.80346543e-01 -7.71727785e-02 1.51612794e+00 -7.94345737e-02
-7.22776353e-01 5.33138588e-02 -3.55890512e-01 -8.70836198e-01
9.61981192e-02 -4.33389187e-01 6.92677945e-02 -8.12426746e-01
1.92774370e-01 -1.04266953e+00 -7.81621456e-01 4.45399322e-02
-4.12088305e-01 -1.28838986e-01 2.89999753e-01 4.01737750e-01
3.61996949e-01 4.07888353e-01 2.94261962e-01 -4.39212024e-01
2.07233146e-01 -1.00440764e+00 -6.48662210e-01 -6.82913661e-02
-7.83782065e-01 -1.20287105e-01 4.52706546e-01 4.23624277e-01
-1.13996208e+00 2.41508856e-02 2.11467728e-01 -1.13136359e-01
-3.87677878e-01 8.89907062e-01 1.94740444e-01 5.72954476e-01
8.28228951e-01 -8.35665539e-02 -5.00536203e-01 -3.12979400e-01
3.67553532e-01 1.46277040e-01 3.10154498e-01 -7.15366602e-01
9.49468970e-01 7.57459223e-01 1.88154995e-01 -9.64595497e-01
-9.10431817e-02 -4.22557205e-01 -4.36817199e-01 -3.15996438e-01
8.88132572e-01 -9.48592901e-01 -7.98547387e-01 1.14856832e-01
-7.21092403e-01 -5.15826643e-01 -1.01052389e-01 6.24654949e-01
-2.69098133e-01 -3.46768320e-01 -1.10575303e-01 -1.24479365e+00
-3.26839089e-01 -5.40834546e-01 5.60361266e-01 1.96011201e-01
-5.12285352e-01 -6.03657484e-01 5.04082084e-01 -1.05985783e-01
4.39581394e-01 6.12452984e-01 1.41234386e+00 2.64649093e-01
-7.91369975e-01 1.19591437e-01 -5.86257398e-01 -2.19766453e-01
3.10719401e-01 9.13967133e-01 -8.34643483e-01 -5.07607162e-02
-3.83593321e-01 1.27108991e-01 8.16693306e-01 6.60081446e-01
9.63451266e-01 -3.04800808e-01 -8.37012231e-01 5.92643380e-01
1.57461393e+00 5.18267989e-01 6.37949288e-01 2.15883791e-01
1.10562831e-01 4.72440451e-01 4.61382359e-01 9.30117190e-01
2.80165505e-02 -1.02745906e-01 8.84042680e-01 -2.54297227e-01
4.77169633e-01 1.03251822e-01 1.59564421e-01 1.06202275e-01
-7.97608972e-01 -7.81516954e-02 -1.39162123e+00 6.57435894e-01
-1.43183303e+00 -1.31307459e+00 -6.09440684e-01 1.99026954e+00
5.91050506e-01 -5.35333343e-02 -1.31385341e-01 -2.00356439e-01
5.64379632e-01 1.69996560e-01 -4.20203567e-01 -3.66687715e-01
1.29710346e-01 1.14744735e+00 9.86108243e-01 6.22302055e-01
-1.00860095e+00 7.80422628e-01 8.18012810e+00 -4.22103703e-02
-1.12606442e+00 -1.88401446e-01 4.38201308e-01 -3.17304015e-01
-7.15146244e-01 5.09047568e-01 -5.77979982e-01 4.47346747e-01
1.27724397e+00 -3.08729261e-01 3.87706041e-01 7.64698565e-01
7.94018388e-01 -6.09987855e-01 -6.10528111e-01 2.42514387e-01
-8.28719139e-01 -1.48929667e+00 -2.69890606e-01 1.34339377e-01
5.97915709e-01 4.96748835e-01 -2.16274023e-01 2.19310462e-01
1.39695334e+00 -1.16348898e+00 3.87743175e-01 6.52176261e-01
8.85824561e-01 -8.31651807e-01 3.15962285e-01 2.36319125e-01
-1.29034698e+00 -1.95748985e-01 -1.43688470e-01 -9.87672925e-01
5.21503128e-02 1.12054813e+00 -9.52423036e-01 5.23889005e-01
1.57299459e+00 4.94743526e-01 -1.94959626e-01 3.86563778e-01
-4.92940575e-01 9.54495430e-01 -7.52374470e-01 7.54352808e-02
-5.47214504e-03 -1.83700606e-01 5.10196567e-01 1.06479132e+00
6.04595184e-01 8.80633712e-01 -2.27807924e-01 1.13817573e+00
5.39799392e-01 -1.30619392e-01 -8.55965376e-01 1.89345013e-02
7.33045280e-01 7.78046191e-01 -6.85881495e-01 -2.68706590e-01
1.55069530e-01 1.68394521e-01 -9.38839614e-02 3.74180079e-01
-9.18038845e-01 -1.16220854e-01 1.03268111e+00 1.67329565e-01
1.30439609e-01 -3.28407556e-01 -5.14268041e-01 -8.18079650e-01
-4.08567756e-01 -2.48510048e-01 3.81959409e-01 -1.01824045e+00
-1.07314789e+00 1.25500500e-01 2.55994052e-01 -6.09139919e-01
-4.16148841e-01 -3.25769037e-01 -1.15039420e+00 1.54052031e+00
-1.54806638e+00 -3.83968592e-01 -2.65176237e-01 5.40932596e-01
-2.63350844e-01 2.45780841e-01 1.41109562e+00 -2.21814200e-01
-7.39132226e-01 -5.00115693e-01 3.32114160e-01 -3.30335528e-01
4.87318635e-01 -1.25669658e+00 2.75866449e-01 1.01947844e+00
-6.58563912e-01 8.15906644e-01 9.58456695e-01 -1.26643014e+00
-1.08522558e+00 -1.19579697e+00 1.12181401e+00 -2.25366339e-01
9.07344818e-01 -8.93074274e-02 -3.11387062e-01 6.24725103e-01
2.24066645e-01 -2.06870094e-01 8.07045877e-01 7.36884415e-01
-2.40244538e-01 -1.72070757e-01 -9.45637584e-01 4.98836398e-01
9.07658935e-01 -2.03944653e-01 -4.43089873e-01 2.70564735e-01
4.90558863e-01 2.22508505e-01 -9.58412230e-01 9.13817167e-01
5.97873628e-01 -1.03545737e+00 9.44071054e-01 -6.19789839e-01
6.16715848e-01 -2.32689336e-01 -4.32940841e-01 -1.27403235e+00
-4.08021927e-01 -2.55786359e-01 5.91797113e-01 8.75461757e-01
5.66507995e-01 -6.44051790e-01 5.38873196e-01 9.34407413e-01
3.80678922e-02 -6.60339296e-02 -1.26091278e+00 -5.22328734e-01
2.49224454e-01 -6.64468229e-01 4.88831908e-01 1.14723790e+00
2.09453911e-01 -2.40263306e-02 -4.88356836e-02 5.46675742e-01
6.92142487e-01 5.75905561e-01 4.19132292e-01 -1.44191480e+00
-4.21664305e-02 -1.65368885e-01 1.05108410e-01 -3.13729458e-02
-9.09860805e-02 -6.17393732e-01 3.03505421e-01 -1.16201246e+00
1.02644652e-01 -1.09522736e+00 -2.96074182e-01 6.58533454e-01
-1.85391322e-01 -1.70199603e-01 -4.14098687e-02 1.36294276e-01
4.28120196e-01 5.63839078e-01 5.87903857e-01 -8.02483112e-02
-3.07767421e-01 -3.56585860e-01 -2.98914611e-01 9.76940215e-01
1.03146636e+00 -6.93928838e-01 -2.64914930e-01 -1.36203095e-01
5.80356598e-01 5.21455228e-01 6.70270503e-01 -7.85833895e-01
-3.85902703e-01 -1.09104395e+00 5.94636738e-01 -7.08321631e-01
1.48185790e-01 -6.58574641e-01 5.85351586e-01 7.25663185e-01
1.59931868e-01 -8.75068307e-02 3.53053302e-01 4.51013058e-01
6.09900244e-02 2.84022003e-01 5.02193153e-01 -4.33300793e-01
-7.11078286e-01 1.14413507e-01 -9.36092556e-01 -2.36622140e-01
8.95614445e-01 2.48262987e-01 -5.94613373e-01 -2.83093840e-01
-7.84844816e-01 4.57670212e-01 1.78818807e-01 1.99852556e-01
5.34235358e-01 -7.95686722e-01 -8.00023913e-01 9.86539647e-02
-2.49125674e-01 -2.42060021e-01 6.07162379e-02 3.23608935e-01
-5.05558312e-01 4.63144541e-01 -2.67253578e-01 -3.91917080e-01
-8.87509227e-01 2.49603614e-01 3.13906670e-01 -1.52738318e-01
-2.74689406e-01 1.07848394e+00 -4.12996300e-02 -1.99785292e-01
-5.64695239e-01 -3.81280154e-01 1.14225410e-01 7.02727884e-02
2.43652388e-01 6.69762373e-01 -1.06633276e-01 -4.66672285e-03
-7.06391156e-01 1.38433844e-01 6.94162786e-01 -3.82919431e-01
1.48967004e+00 -2.09093112e-02 -5.74669540e-01 5.59080780e-01
5.97007334e-01 -1.54465269e-02 -1.06012833e+00 5.08860826e-01
-1.00799680e-01 -5.73526204e-01 9.87512320e-02 -1.08587110e+00
-8.50246251e-01 6.20413661e-01 5.89458287e-01 3.08945030e-01
1.05380070e+00 -1.85744405e-01 4.93704388e-03 1.08213753e-01
2.56665289e-01 -8.53886187e-01 -6.04226410e-01 5.53395391e-01
9.91204858e-01 -1.00622344e+00 3.36644381e-01 -5.16279936e-01
1.79701686e-01 8.65603387e-01 5.57278335e-01 2.35907044e-02
1.37813175e+00 5.11183560e-01 -4.38467525e-02 -4.97627825e-01
-1.05686891e+00 -3.25405717e-01 -5.07798076e-01 3.45266581e-01
4.06905472e-01 7.44906008e-01 -3.79840791e-01 2.70169944e-01
-5.58866739e-01 -5.44202961e-02 1.10365078e-01 8.75173688e-01
-6.51119411e-01 -7.29536116e-01 -7.32136071e-01 7.67277122e-01
2.02859417e-02 -6.81027234e-01 -3.40054274e-01 1.02325392e+00
5.26012838e-01 1.33379102e+00 4.17257726e-01 -3.45282733e-01
-3.66182514e-02 3.21784914e-01 -7.67341554e-02 -4.73629057e-01
-2.44411334e-01 -3.66977267e-02 2.54813582e-01 -7.21070766e-01
-5.48581481e-01 -1.13431096e+00 -1.37294054e+00 -7.96680272e-01
-1.78105310e-01 4.09426868e-01 5.63916028e-01 9.16868389e-01
1.38923690e-01 4.25490499e-01 7.67318130e-01 -6.87257528e-01
2.69287109e-01 -1.02550113e+00 -8.95922482e-01 -4.01935205e-02
2.85180509e-02 -7.80252159e-01 -9.39805925e-01 2.52027392e-01]
|
[6.513112545013428, 3.2179112434387207]
|
4129856a-e770-4297-b4e6-aebe212aa64e
|
study-on-sparse-representation-based
|
1502.06073
| null |
http://arxiv.org/abs/1502.06073v2
|
http://arxiv.org/pdf/1502.06073v2.pdf
|
Study on Sparse Representation based Classification for Biometric Verification
|
In this paper, we propose a multimodal verification system integrating face
and ear based on sparse representation based classification (SRC). The face and
ear query samples are first encoded separately to derive sparsity-based match
scores, and which are then combined with sum-rule fusion for verification.
Apart from validating the encouraging performance of SRC-based multimodal
verification, this paper also dedicates to provide a clear understanding about
the characteristics of SRC-based biometric verification. To this end, two
sparsity-based metrics, i.e. spare coding error (SCE) and sparse contribution
rate (SCR), are involved, together with face and ear unimodal SRC-based
verification. As for the issue that SRC-based biometric verification may suffer
from heavy computational burden and verification accuracy degradation with
increase of enrolled subjects, we argue that it could be properly resolved by
exploiting small random dictionary for sparsity-based score computation, which
consists of training samples from a limited number of randomly selected
subjects. Experimental results demonstrate the superiority of SRC-based
multimodal verification compared to the state-of-the-art multimodal methods
like likelihood ratio (LLR), support vector machine (SVM), and the sum-rule
fusion methods using cosine similarity, meanwhile the idea of using small
random dictionary is feasible in both effectiveness and efficiency.
|
['Zengxi Huang', 'Yiguang Liu', 'Xiaoming Wang', 'Jinrong Hu']
|
2015-02-21
| null | null | null | null |
['sparse-representation-based-classification']
|
['computer-vision']
|
[ 4.09897029e-01 -3.54726702e-01 -1.18333228e-01 -4.42382276e-01
-1.09956896e+00 -2.97338039e-01 2.27985144e-01 1.02256157e-01
-1.93400979e-02 7.09578574e-01 2.78306991e-01 1.17392860e-01
-3.01874965e-01 -3.85205954e-01 -9.62300971e-02 -1.05768204e+00
2.30108425e-01 -1.03911266e-01 -4.08619642e-01 -1.37424067e-01
2.38748118e-01 6.89288080e-01 -1.88086045e+00 1.57342270e-01
9.71849442e-01 1.37497044e+00 -4.38295871e-01 1.56665504e-01
1.64195463e-01 3.73251438e-01 -4.56711471e-01 -9.13890541e-01
8.31765160e-02 -5.25738537e-01 -1.32228330e-01 -9.59304944e-02
4.47433054e-01 -1.49409607e-01 -3.29808146e-01 1.16281199e+00
1.06097531e+00 -7.18834698e-02 7.74220526e-01 -1.34100926e+00
-7.56828785e-01 3.66355598e-01 -7.66531229e-01 -7.94368908e-02
9.10422921e-01 7.31105953e-02 7.54425168e-01 -1.32082593e+00
2.43083194e-01 1.09400094e+00 8.68625760e-01 4.69751120e-01
-9.31858718e-01 -9.96499658e-01 -4.06605005e-01 2.17278093e-01
-2.00499558e+00 -8.16359043e-01 1.00965691e+00 -2.41447583e-01
2.61796385e-01 6.13797367e-01 4.70553368e-01 8.46237361e-01
-8.45147893e-02 4.95540917e-01 1.28103018e+00 -5.04485965e-01
1.48479626e-01 3.46951306e-01 1.50214359e-01 7.62747586e-01
5.67615688e-01 4.15429235e-01 -8.42721820e-01 -5.84763646e-01
5.00966191e-01 7.87469223e-02 -5.14598548e-01 -1.76450491e-01
-1.12374675e+00 7.84231424e-01 7.50402687e-03 4.88068849e-01
-3.52191120e-01 -4.09413636e-01 2.84555167e-01 1.43285483e-01
-4.06367742e-02 -1.45186797e-01 -2.00438365e-01 2.87160933e-01
-1.35566473e+00 -5.52786142e-02 6.69380844e-01 5.69184542e-01
4.99868125e-01 3.08827281e-01 -4.75416034e-01 9.76779521e-01
8.19131732e-01 1.07265711e+00 5.88872492e-01 -5.68027198e-01
2.56823510e-01 6.07967317e-01 -2.53895700e-01 -1.25464714e+00
2.01762915e-02 -5.00339031e-01 -1.24232221e+00 -1.81625605e-01
2.38677159e-01 2.23022029e-02 -6.25168383e-01 1.57836854e+00
2.95893729e-01 3.44452798e-01 2.33960405e-01 1.00866330e+00
1.21144855e+00 1.33911341e-01 -2.30253972e-02 -4.88480210e-01
1.54172266e+00 -2.94151366e-01 -9.16800737e-01 6.10780418e-01
2.69411087e-01 -1.16842389e+00 4.62048739e-01 3.02666694e-01
-1.04744506e+00 -6.40782475e-01 -9.63580549e-01 4.12298381e-01
-2.32156396e-01 5.82772851e-01 5.72921634e-01 1.32331777e+00
-9.52975571e-01 6.97392374e-02 -2.76792407e-01 -3.08205843e-01
2.85037428e-01 6.11070693e-01 -7.56138802e-01 -2.63988137e-01
-1.08933032e+00 7.31258810e-01 -1.96600676e-01 4.92995173e-01
-3.83515060e-01 -6.37252688e-01 -1.25586820e+00 1.40277743e-01
-2.73164243e-01 -4.45168555e-01 5.56088686e-01 -4.25377220e-01
-1.40440929e+00 7.06394494e-01 -6.10880315e-01 -5.30610606e-02
-1.96551606e-01 3.46379727e-01 -8.47855270e-01 3.01256269e-01
-1.78432167e-01 2.44795740e-01 1.04665506e+00 -1.22355759e+00
-7.87247643e-02 -6.62666202e-01 -5.38311183e-01 -1.52026355e-01
-3.13800126e-01 4.31532443e-01 2.56528221e-02 -6.82869315e-01
4.91076082e-01 -5.92153549e-01 1.27593338e-01 -8.57502148e-02
-2.06569821e-01 -1.48102865e-01 7.54492044e-01 -1.09666228e+00
1.39620483e+00 -2.28167129e+00 -1.66674331e-01 7.32075751e-01
-1.10695407e-01 3.20533544e-01 -3.44419092e-01 3.95720303e-01
-2.12688521e-01 -1.78005219e-01 -3.09963077e-01 -3.52489829e-01
-3.27372290e-02 -3.58114302e-01 -2.77804375e-01 8.67638707e-01
2.57828772e-01 8.04711223e-01 -4.62579578e-01 -7.07481384e-01
1.85654238e-01 9.51362371e-01 -3.48802835e-01 4.00974713e-02
7.28112698e-01 3.67300540e-01 -3.51136863e-01 1.72604954e+00
1.23531473e+00 -9.24976617e-02 1.33429483e-01 -8.43641520e-01
2.36640647e-01 -3.49877447e-01 -1.56198263e+00 1.40029955e+00
-3.72852050e-02 1.95120186e-01 2.23246157e-01 -9.23546672e-01
1.29063296e+00 6.57373726e-01 4.86411333e-01 -5.23977757e-01
3.50892931e-01 4.47295845e-01 -1.77683368e-01 -3.50110203e-01
3.75879735e-01 -2.41945490e-01 4.06468920e-02 2.43017793e-01
2.51347274e-01 2.89902806e-01 -2.01708242e-01 -1.03029562e-03
4.95885640e-01 -3.49043995e-01 4.44223791e-01 1.87466796e-02
1.31111050e+00 -5.71973860e-01 4.56480473e-01 1.75981998e-01
-2.57967740e-01 6.12138808e-01 -6.94102123e-02 1.37973845e-01
-5.82772195e-01 -8.43282700e-01 -5.58227658e-01 3.70961636e-01
2.45656729e-01 -2.03501329e-01 -4.30449903e-01 -5.74103415e-01
3.21259826e-01 1.26854554e-01 -3.68071854e-01 -6.07061572e-02
-3.47292989e-01 -8.05276930e-01 8.19720924e-01 3.49439621e-01
5.37801206e-01 -4.43821073e-01 9.61319078e-03 -1.68390274e-01
-1.77056566e-01 -8.83110762e-01 -6.21654689e-01 -4.41585809e-01
-7.49145210e-01 -1.18734360e+00 -1.08145356e+00 -6.90818071e-01
8.73156011e-01 4.02555496e-01 3.60219926e-01 2.55018026e-01
-3.96694779e-01 7.15460777e-01 -3.74355167e-01 -5.69240712e-02
-3.35422978e-02 -3.76864582e-01 3.17854941e-01 7.44531691e-01
5.35426915e-01 -3.96573335e-01 -7.02671766e-01 5.45761943e-01
-5.85026443e-01 -8.35853040e-01 6.65962338e-01 1.13247848e+00
4.72957790e-01 -2.28317544e-01 6.35213256e-01 -1.61093827e-02
6.07197106e-01 -2.97193795e-01 -3.45163435e-01 6.25865102e-01
-5.97754121e-01 -1.83325931e-01 8.76267254e-02 -4.53474998e-01
-8.91505957e-01 1.66218534e-01 -1.51420891e-01 -5.74803233e-01
1.02412492e-01 6.18879855e-01 -1.13520429e-01 -9.04579759e-01
3.01406205e-01 8.44856858e-01 4.07330930e-01 -4.31020379e-01
1.06738068e-01 9.99405026e-01 4.48194265e-01 -4.41533923e-01
9.79355574e-01 3.66103590e-01 2.35752001e-01 -7.01731801e-01
-3.14940065e-01 -7.05561340e-01 -1.81808919e-01 -2.65427232e-01
4.88648742e-01 -1.11568749e+00 -1.23944068e+00 4.08410937e-01
-1.00888646e+00 8.19908023e-01 -1.11667268e-01 8.57765317e-01
-1.12105623e-01 8.82080793e-01 -4.06399220e-01 -1.43641078e+00
-6.09088004e-01 -1.09392488e+00 1.44498348e+00 5.69548070e-01
-1.46426270e-02 -6.28485322e-01 -1.34741604e-01 8.12858880e-01
6.85318828e-01 1.39760688e-01 4.05027658e-01 -7.57395208e-01
-4.32666719e-01 -6.85255289e-01 -4.47333366e-01 3.51183951e-01
1.06189355e-01 -2.76810795e-01 -1.17786050e+00 -6.49248123e-01
2.06839651e-01 -3.31240356e-01 5.45107126e-01 2.38346949e-01
9.00921345e-01 -1.20519646e-01 -2.19928741e-01 4.47584987e-01
1.46279061e+00 1.18863709e-01 8.23756099e-01 -5.38406610e-01
2.48418480e-01 5.14298320e-01 6.02011561e-01 6.41771197e-01
3.68308276e-01 7.44576097e-01 3.93298047e-04 -1.27822086e-01
-2.46501476e-01 -1.41094998e-01 3.90622824e-01 9.41705048e-01
-2.28342816e-01 1.50559828e-01 -4.76426840e-01 1.80517435e-01
-1.51112700e+00 -1.20661294e+00 6.66488260e-02 2.38572335e+00
6.14367545e-01 -8.02495897e-01 2.65452731e-02 6.75099015e-01
9.54858482e-01 -1.06109843e-01 -2.16063872e-01 -1.88377295e-02
-6.39134645e-01 4.89797592e-01 2.45191023e-01 3.63902330e-01
-7.60173500e-01 3.12540352e-01 6.14445257e+00 1.11125398e+00
-1.37306714e+00 2.55864233e-01 4.85922128e-01 2.35067219e-01
-2.83920944e-01 -6.28997460e-02 -1.17883623e+00 5.69422185e-01
6.46184921e-01 8.60600695e-02 2.85468221e-01 3.60376745e-01
4.06589806e-02 -1.35616601e-01 -7.98234940e-01 1.62175870e+00
8.15076649e-01 -1.06897473e+00 1.17942676e-01 1.39239103e-01
5.13786674e-01 -6.97499692e-01 2.31191665e-01 9.07163396e-02
-3.82402331e-01 -1.12061691e+00 2.84606427e-01 8.12770188e-01
1.09684336e+00 -3.80384356e-01 1.21432710e+00 -1.24221832e-01
-1.49021113e+00 -1.18503282e-02 -2.59149879e-01 4.87358361e-01
1.08887248e-01 7.80539989e-01 -3.24762464e-01 1.01607323e+00
3.18191051e-01 3.67007792e-01 -4.14961994e-01 1.14933193e+00
1.49524361e-01 4.84756023e-01 -3.19141626e-01 -1.45917377e-02
-4.89867657e-01 -1.70850232e-01 4.08993423e-01 1.03955626e+00
8.05783570e-01 2.97757596e-01 -1.81361735e-01 7.29205191e-01
1.62865475e-01 4.02383357e-01 -4.21935499e-01 3.13524418e-02
6.72437012e-01 1.25898111e+00 -2.20303103e-01 -1.09718643e-01
-4.71415192e-01 7.41728127e-01 -3.96986693e-01 3.57048213e-01
-7.07001925e-01 -4.98130739e-01 3.54935408e-01 -7.17031732e-02
5.65987945e-01 -1.75661929e-02 -4.12803888e-01 -1.45683503e+00
1.93752751e-01 -1.02062845e+00 4.76375997e-01 -5.38603783e-01
-1.49425590e+00 4.19820130e-01 -3.57048899e-01 -1.67786992e+00
1.58778772e-01 -2.62305558e-01 -3.34183276e-01 1.21632135e+00
-1.91083515e+00 -1.42252433e+00 -3.13775867e-01 9.42293167e-01
-1.13898538e-01 -6.44278288e-01 9.69177783e-01 6.36311531e-01
-4.70854670e-01 1.40826166e+00 -2.04358920e-01 2.00419992e-01
7.27936327e-01 -3.48118365e-01 -7.36024141e-01 5.36318898e-01
8.79191086e-02 9.21159387e-01 2.69222140e-01 -4.51645702e-01
-1.78923142e+00 -6.18171811e-01 1.32419384e+00 -1.68651059e-01
2.15515748e-01 1.69553697e-01 -5.98866940e-01 -1.99430615e-01
-1.13298055e-02 2.57718772e-01 1.36898863e+00 -4.86157499e-02
-5.97214043e-01 -4.91908014e-01 -1.83483708e+00 2.14625612e-01
7.57298946e-01 -7.85838604e-01 -5.03550410e-01 2.75269262e-02
1.94373295e-01 -7.87997767e-02 -1.39042008e+00 8.72508228e-01
9.36433733e-01 -8.62237036e-01 1.08387220e+00 -8.15194622e-02
-1.10029496e-01 -6.99582279e-01 -7.17570126e-01 -5.42935431e-01
-3.57470661e-02 -4.58994329e-01 -2.78231233e-01 1.53677154e+00
3.26479495e-01 -7.40475357e-01 6.67590201e-01 4.66481209e-01
4.37305748e-01 -6.91980124e-01 -1.22010446e+00 -7.73837745e-01
-5.43343186e-01 -1.47048873e-03 7.99958467e-01 1.00662613e+00
3.11115026e-01 -4.16535810e-02 -5.36482811e-01 5.16005814e-01
9.28722322e-01 2.07987472e-01 4.50306147e-01 -1.31392407e+00
-8.01979527e-02 -1.08936965e-01 -9.23681796e-01 -5.98775685e-01
-1.66389393e-03 -7.69514382e-01 -4.21300471e-01 -8.75285923e-01
4.30126369e-01 -3.76654953e-01 -6.50304377e-01 4.30826455e-01
-2.92617511e-02 6.89817071e-01 1.41054943e-01 3.04393053e-01
-2.57982343e-01 7.45350957e-01 1.06235409e+00 -2.46523112e-01
-7.16706216e-02 2.35891230e-02 -8.94613922e-01 7.38684982e-02
4.35773045e-01 -1.66324019e-01 -1.08653381e-01 2.74468333e-01
-1.91550568e-01 4.49785411e-01 5.29939651e-01 -9.81964707e-01
5.52225113e-01 1.39165208e-01 5.68103969e-01 -5.94805300e-01
6.10912383e-01 -8.90128136e-01 1.76590174e-01 5.62865257e-01
-1.59633845e-01 -1.74447715e-01 -7.46113621e-03 4.76953477e-01
-7.77455568e-01 -4.76443283e-02 7.42141604e-01 4.26517695e-01
-1.09087296e-01 2.39780292e-01 2.88213287e-02 -6.14795566e-01
9.97448742e-01 -7.93568790e-01 -1.27476335e-01 -3.63534719e-01
-7.43166387e-01 -2.13844836e-01 4.81297858e-02 2.44401153e-02
1.08157992e+00 -1.63099885e+00 -9.04028773e-01 7.96250701e-01
1.67967409e-01 -8.52241218e-01 5.32818794e-01 1.32834947e+00
1.00491144e-01 5.73561549e-01 -1.87542010e-02 -5.77962101e-01
-1.75818157e+00 2.84332365e-01 1.24750726e-01 3.44122015e-02
8.91028792e-02 7.37791538e-01 -4.10600752e-01 -2.81201184e-01
3.90768379e-01 6.65392801e-02 -2.38436103e-01 1.80830330e-01
5.93104482e-01 2.48601824e-01 1.90393716e-01 -9.75652933e-01
-6.81394339e-01 1.14760816e+00 1.69724897e-01 -6.70376346e-02
9.56873536e-01 -3.31386924e-02 -2.87852824e-01 -2.53086835e-01
1.24090350e+00 4.27535534e-01 -3.99273694e-01 -3.87601942e-01
-3.44138443e-01 -7.00636387e-01 -4.24998030e-02 -8.72787833e-01
-1.25992501e+00 6.33805156e-01 1.08267713e+00 -2.66982257e-01
1.41319788e+00 -5.94057180e-02 8.78252566e-01 6.29663318e-02
5.17073452e-01 -6.07216716e-01 -1.36668742e-01 -4.78507997e-03
9.10170019e-01 -1.41110516e+00 1.20169558e-01 -7.16805220e-01
-4.33749557e-01 1.09900558e+00 8.27434212e-02 1.36544093e-01
9.65633094e-01 2.06407905e-01 -2.24110171e-01 -4.96261008e-02
-3.08267325e-01 -1.35100096e-01 8.02771807e-01 7.65706897e-01
5.13572395e-01 -8.46054107e-02 -7.20556557e-01 9.82915878e-01
-6.74316436e-02 2.64830142e-01 -1.27772123e-01 8.58888209e-01
-1.82341769e-01 -1.29012120e+00 -8.74929905e-01 4.44829732e-01
-2.43277431e-01 -1.89661235e-01 -1.49137020e-01 2.07263455e-01
3.12992215e-01 1.08859777e+00 -4.71471727e-01 -6.84959412e-01
7.33603090e-02 1.82394311e-01 6.96076274e-01 -1.53075755e-01
-6.38736010e-01 -1.00578807e-01 -1.74635634e-01 -4.36387986e-01
-8.71695638e-01 -8.22811782e-01 -6.62562132e-01 -3.81205380e-01
-8.12397003e-01 1.99544460e-01 8.32605124e-01 8.00940394e-01
3.86662453e-01 -2.09727883e-01 9.81070876e-01 -4.50853199e-01
-8.43751431e-01 -7.15355694e-01 -7.63348520e-01 1.59045234e-01
3.71884495e-01 -6.33360565e-01 -5.85337222e-01 1.05079068e-02]
|
[12.821490287780762, 0.4919026494026184]
|
ad5a6678-368f-49b5-adf4-93fb4970ffec
|
climax-an-exploration-of-classifier-based
|
2307.00680
| null |
https://arxiv.org/abs/2307.00680v1
|
https://arxiv.org/pdf/2307.00680v1.pdf
|
CLIMAX: An exploration of Classifier-Based Contrastive Explanations
|
Explainable AI is an evolving area that deals with understanding the decision making of machine learning models so that these models are more transparent, accountable, and understandable for humans. In particular, post-hoc model-agnostic interpretable AI techniques explain the decisions of a black-box ML model for a single instance locally, without the knowledge of the intrinsic nature of the ML model. Despite their simplicity and capability in providing valuable insights, existing approaches fail to deliver consistent and reliable explanations. Moreover, in the context of black-box classifiers, existing approaches justify the predicted class, but these methods do not ensure that the explanation scores strongly differ as compared to those of another class. In this work we propose a novel post-hoc model agnostic XAI technique that provides contrastive explanations justifying the classification of a black box classifier along with a reasoning as to why another class was not predicted. Our method, which we refer to as CLIMAX which is short for Contrastive Label-aware Influence-based Model Agnostic XAI, is based on local classifiers . In order to ensure model fidelity of the explainer, we require the perturbations to be such that it leads to a class-balanced surrogate dataset. Towards this, we employ a label-aware surrogate data generation method based on random oversampling and Gaussian Mixture Model sampling. Further, we propose influence subsampling in order to retaining effective samples and hence ensure sample complexity. We show that we achieve better consistency as compared to baselines such as LIME, BayLIME, and SLIME. We also depict results on textual and image based datasets, where we generate contrastive explanations for any black-box classification model where one is able to only query the class probabilities for an instance of interest.
|
['Ranjitha Prasad', 'Praharsh Nanavati']
|
2023-07-02
| null | null | null | null |
['decision-making']
|
['reasoning']
|
[ 5.42777359e-01 8.15041542e-01 -4.64670986e-01 -5.53544402e-01
-6.43318713e-01 -5.45850158e-01 1.14039147e+00 2.52830714e-01
8.66914093e-02 8.54342282e-01 1.43018559e-01 -4.64686036e-01
-5.52214205e-01 -6.22960091e-01 -9.13807809e-01 -6.83372378e-01
3.10613692e-01 9.94377315e-01 -1.49722517e-01 1.82594761e-01
2.65236139e-01 3.05267870e-01 -1.74992335e+00 5.92524350e-01
9.43096280e-01 7.48828411e-01 -2.03675702e-01 6.21446252e-01
8.45593959e-02 9.01303828e-01 -6.60286725e-01 -4.30142134e-01
1.31204531e-01 -7.72695541e-01 -8.46131086e-01 1.42140493e-01
4.96217132e-01 -1.42695792e-02 2.89730459e-01 6.96745634e-01
-1.95068747e-01 -1.58195123e-01 1.18716800e+00 -1.80347562e+00
-5.90856314e-01 1.06812119e+00 -3.17099035e-01 -3.26284766e-01
2.10437011e-02 3.00591886e-01 1.07337976e+00 -4.54165995e-01
6.44526005e-01 1.48515737e+00 4.97111797e-01 8.16565037e-01
-1.58133674e+00 -5.72871089e-01 5.78662992e-01 3.43035787e-01
-1.03462636e+00 -3.46859932e-01 6.85310304e-01 -3.59673619e-01
4.83131945e-01 8.59380126e-01 4.80910480e-01 1.33294272e+00
3.50139469e-01 6.89845681e-01 1.31875277e+00 -6.14451885e-01
7.60254622e-01 4.94893819e-01 2.68820047e-01 5.13604999e-01
5.30185759e-01 2.30070323e-01 -6.46736860e-01 -4.29787099e-01
1.59811646e-01 -3.09101716e-02 -1.86300769e-01 -5.64052939e-01
-1.24270034e+00 9.72203434e-01 6.17807269e-01 1.20100982e-01
-5.45608699e-01 5.70606828e-01 -3.42040546e-02 6.93964511e-02
5.38720131e-01 8.29408467e-01 -7.29302108e-01 1.69029534e-01
-1.01132250e+00 5.66830218e-01 6.66306376e-01 7.90346861e-01
7.53753901e-01 -4.43886854e-02 -1.98530361e-01 1.82042882e-01
6.02593005e-01 2.66014129e-01 3.77532393e-01 -1.19713354e+00
3.26509327e-02 9.35475051e-01 1.99680571e-02 -8.16805482e-01
-1.02414414e-01 -6.08689725e-01 -6.46358907e-01 6.03164792e-01
3.49285126e-01 2.60645777e-01 -1.03254151e+00 1.96515119e+00
4.04287398e-01 4.77090105e-02 1.84118673e-01 8.32941055e-01
4.20742184e-01 4.56660509e-01 2.27377176e-01 -4.24112827e-02
1.28055704e+00 -1.07399249e+00 -5.23464084e-01 -5.44287503e-01
6.22207761e-01 -2.00705737e-01 1.26121962e+00 4.97299492e-01
-8.30018401e-01 -4.45280403e-01 -1.18726790e+00 1.63173497e-01
-2.45717883e-01 -7.51616135e-02 7.62321174e-01 6.10793233e-01
-6.44346058e-01 6.76724195e-01 -8.34467113e-01 -2.63713956e-01
6.04646862e-01 2.08969250e-01 -2.34176219e-01 7.31544420e-02
-7.85057306e-01 1.05164027e+00 3.30656379e-01 -1.35910138e-01
-1.16293097e+00 -9.25001204e-01 -5.94809175e-01 2.48160094e-01
4.89795566e-01 -9.77889717e-01 1.28660226e+00 -1.33973312e+00
-1.01240158e+00 5.14305353e-01 -2.93427289e-01 -7.54988551e-01
8.82777393e-01 -3.62936780e-02 6.82164449e-03 -1.36429414e-01
1.30360752e-01 9.76045549e-01 9.16749179e-01 -1.94880939e+00
-5.71388841e-01 -2.65451163e-01 2.39428356e-01 2.19750255e-02
-4.75789560e-03 -3.86648595e-01 1.74658820e-01 -3.71658534e-01
2.64115632e-01 -1.06695569e+00 -3.51493448e-01 2.99558491e-01
-7.59621203e-01 2.88667306e-02 8.41398001e-01 -1.79427564e-01
9.80521798e-01 -1.83843803e+00 -3.64110339e-03 2.72138059e-01
6.03233635e-01 -2.58359462e-01 4.09207232e-02 1.88982889e-01
-2.13025495e-01 5.98432064e-01 -6.14239573e-01 -3.66963148e-01
2.46736184e-01 5.73698401e-01 -6.55076444e-01 3.22615653e-01
3.95792544e-01 7.20171690e-01 -7.69025207e-01 -2.94671029e-01
1.70788184e-01 3.37843269e-01 -7.33493924e-01 1.66634601e-02
-6.86870694e-01 5.28778434e-01 -3.94351721e-01 3.95277411e-01
3.71325165e-01 -3.95815462e-01 8.27201605e-02 -1.99058969e-02
1.72060773e-01 2.60393620e-01 -1.03900290e+00 1.16456294e+00
-3.82130027e-01 5.45472383e-01 -2.47805044e-01 -8.52683842e-01
7.78227091e-01 2.98585892e-01 -2.04548668e-02 -3.34887989e-02
3.74872759e-02 3.44165087e-01 2.90683419e-01 -4.16795090e-02
1.27693355e-01 -4.06052202e-01 8.97303075e-02 8.11127245e-01
-4.58389640e-01 -3.45035583e-01 -1.37010291e-01 2.51704067e-01
1.04760838e+00 3.69930357e-01 5.64405203e-01 -4.19904411e-01
3.75484854e-01 4.20198113e-01 4.64176923e-01 1.09731877e+00
-4.09105457e-02 8.02569807e-01 5.62873363e-01 -6.89034760e-01
-1.00555634e+00 -7.73190498e-01 -3.00583959e-01 6.02437615e-01
1.41734868e-01 -4.35135096e-01 -9.21883345e-01 -1.00875938e+00
1.07596859e-01 1.52030432e+00 -1.08488858e+00 -5.12982965e-01
-1.32054552e-01 -6.25771523e-01 2.12278008e-01 3.36364895e-01
2.81135172e-01 -1.06679022e+00 -9.86052155e-01 7.50022894e-03
-2.37608589e-02 -5.73621392e-01 8.46794024e-02 5.13179243e-01
-8.98584008e-01 -1.25423002e+00 -1.19593158e-01 -4.41514812e-02
1.03998137e+00 -2.21501887e-01 1.12871242e+00 3.70533824e-01
3.56774814e-02 2.23940089e-01 -2.58427650e-01 -7.65736282e-01
-9.12457228e-01 8.05056021e-02 6.72962889e-02 3.16244327e-02
3.90629470e-01 -4.40606356e-01 -4.50714976e-01 2.92045265e-01
-9.34477627e-01 5.14791667e-01 5.21318555e-01 8.35234642e-01
5.31205475e-01 -4.82864976e-02 4.21322078e-01 -1.32408977e+00
4.94817495e-01 -4.86737251e-01 -1.67928889e-01 5.02109945e-01
-1.31270015e+00 4.71188188e-01 6.05784953e-01 -5.34612477e-01
-1.02648211e+00 -3.56003791e-02 5.08629978e-01 -2.78684229e-01
-4.24776703e-01 4.65632170e-01 -2.20396012e-01 3.76855791e-01
1.18602574e+00 -1.61877975e-01 -1.96653064e-02 -2.83493698e-01
6.35350049e-01 6.38842344e-01 2.91738570e-01 -7.93401480e-01
9.80403602e-01 5.29192805e-01 -1.66930854e-02 -4.32383828e-02
-8.77470791e-01 3.92508626e-01 -5.42357564e-01 -2.55346090e-01
5.11452794e-01 -4.70411867e-01 -5.81882596e-01 -1.71936080e-01
-1.11945987e+00 -4.77773219e-01 -5.87063074e-01 3.64354134e-01
-8.27675521e-01 -2.02875659e-01 -3.33306193e-03 -1.02736115e+00
-1.00350797e-01 -1.23256862e+00 1.02363920e+00 6.07204437e-03
-9.59337354e-01 -9.49918568e-01 -1.78559735e-01 4.58856493e-01
4.39609021e-01 4.07833189e-01 1.32399368e+00 -9.90304828e-01
-7.03123808e-01 -2.94906467e-01 -5.72413653e-02 2.57498100e-02
2.39940062e-02 3.10712844e-01 -1.25746059e+00 2.02216320e-02
5.55157997e-02 -2.43685767e-01 7.06733882e-01 3.44789565e-01
1.18760371e+00 -7.77759433e-01 -5.79303443e-01 1.32983342e-01
1.08047056e+00 1.59123521e-02 3.21100920e-01 6.82371199e-01
5.19909203e-01 9.01596725e-01 6.82493091e-01 2.45459124e-01
3.36510450e-01 5.98583281e-01 7.03913212e-01 -1.80246755e-01
-7.15050101e-02 -3.44383925e-01 1.60845947e-02 -6.65048435e-02
9.73090604e-02 -4.02933419e-01 -9.60575402e-01 5.67977369e-01
-2.15272665e+00 -7.99326479e-01 -2.66736776e-01 2.17402172e+00
8.52784395e-01 2.33408004e-01 -2.86386430e-01 4.17981714e-01
4.46186423e-01 -1.19887933e-01 -7.79197872e-01 -6.08386040e-01
-7.68853202e-02 -1.94726318e-01 2.32682317e-01 7.52732933e-01
-6.24668896e-01 6.99440956e-01 5.97468090e+00 5.39025962e-01
-8.00524950e-01 4.14697789e-02 9.28668261e-01 -1.29946724e-01
-9.48876977e-01 6.05629623e-01 -4.25896108e-01 2.22542703e-01
1.11084068e+00 -3.21265340e-01 3.39981049e-01 1.08570266e+00
3.33291680e-01 -1.67254537e-01 -1.67098224e+00 3.71088535e-01
-9.07984078e-02 -1.36888027e+00 3.26030403e-01 2.24013492e-01
6.78610981e-01 -5.13606369e-01 6.37603849e-02 1.07346669e-01
4.09449220e-01 -1.19810808e+00 1.33726144e+00 6.36160135e-01
3.68185550e-01 -6.17095709e-01 6.17523968e-01 6.57981575e-01
-6.02238774e-01 -1.73531666e-01 -8.89727548e-02 -2.36050755e-01
-1.87914982e-01 4.79273617e-01 -9.63856995e-01 4.20631170e-01
4.90436435e-01 5.25547922e-01 -6.57850444e-01 5.70794702e-01
-5.82112789e-01 8.89578819e-01 -1.82180107e-01 6.43420666e-02
1.65939316e-01 1.21838614e-01 5.16411960e-01 7.12709188e-01
9.44514051e-02 -4.91380617e-02 -6.16755486e-02 1.43728971e+00
2.28252396e-01 -3.09434593e-01 -5.00097871e-01 1.41782105e-01
4.76118118e-01 1.05336130e+00 -6.13173604e-01 -5.72927237e-01
1.78668842e-01 5.92592478e-01 1.43240497e-01 2.89757639e-01
-9.07664239e-01 2.87613034e-01 5.05649030e-01 2.12561771e-01
-6.98734745e-02 3.76646727e-01 -9.59064543e-01 -9.25256968e-01
-1.17530450e-01 -1.24696708e+00 2.19264060e-01 -1.11637938e+00
-1.22448063e+00 7.13940680e-01 4.21948165e-01 -1.04953671e+00
-6.44065380e-01 -5.17720282e-01 -6.75305247e-01 8.39806616e-01
-1.41388607e+00 -1.38972640e+00 -3.68416071e-01 9.14019272e-02
5.21320283e-01 1.38245448e-02 9.57328796e-01 -4.47035253e-01
-3.94430399e-01 3.47808838e-01 -2.35390469e-01 -5.43127120e-01
6.17761135e-01 -1.44411123e+00 2.41346851e-01 7.84551620e-01
3.48441333e-01 9.12342548e-01 1.27120852e+00 -7.00441301e-01
-9.44810331e-01 -1.07030785e+00 9.87942517e-01 -1.03003895e+00
3.91685903e-01 -2.94617146e-01 -1.02953613e+00 9.22801077e-01
-2.84756981e-02 -1.60039738e-01 5.88958085e-01 2.36018881e-01
-3.42968494e-01 -8.27243924e-02 -1.26543117e+00 7.80316591e-01
8.40403616e-01 -1.38230711e-01 -6.00199878e-01 4.09401745e-01
5.62989891e-01 -1.37950420e-01 -4.10410374e-01 5.02651215e-01
4.79508609e-01 -9.78246808e-01 5.58396280e-01 -9.08637524e-01
7.74242103e-01 -6.02414787e-01 -2.08522692e-01 -1.26548266e+00
-2.46471539e-01 -2.78429985e-01 -1.80587575e-01 1.24011123e+00
8.82623672e-01 -6.98090017e-01 8.96026194e-01 1.33846867e+00
-9.69258696e-03 -7.61033475e-01 -7.92373002e-01 -6.24216616e-01
-2.73745749e-02 -7.22128987e-01 9.57719147e-01 8.79892707e-01
-9.77813303e-02 1.28729269e-01 -2.26049751e-01 3.06502551e-01
7.38459051e-01 3.85141045e-01 1.01879382e+00 -1.42090118e+00
-3.45593393e-01 -4.06892180e-01 -2.44887128e-01 -3.85389328e-01
1.85471669e-01 -7.22529292e-01 8.03472102e-02 -1.57988679e+00
6.31327987e-01 -6.02120936e-01 -4.77505662e-02 8.95677447e-01
-3.53553206e-01 1.07389942e-01 2.39160135e-01 5.46426296e-01
-9.63766426e-02 4.82241452e-01 9.19183910e-01 -1.91123456e-01
-3.54647823e-02 -1.02650277e-01 -1.15555751e+00 8.45836699e-01
6.38816118e-01 -9.41727936e-01 -6.85483515e-01 -1.41700745e-01
1.84858277e-01 -2.94478297e-01 8.79852533e-01 -8.12276959e-01
8.48435238e-03 -3.46979529e-01 2.88456351e-01 -1.44460261e-01
2.42440209e-01 -9.42098916e-01 6.62142396e-01 6.35144114e-01
-9.72881377e-01 -2.21242577e-01 6.80611134e-02 7.08585739e-01
3.34217139e-02 -3.33296895e-01 4.78699654e-01 -3.33508328e-02
-2.46894121e-01 2.10512783e-02 -3.10754061e-01 -2.71556616e-01
1.03432333e+00 -2.16805965e-01 -4.91187632e-01 -5.69381177e-01
-4.66440618e-01 4.50310521e-02 7.65800357e-01 2.93034226e-01
3.30488563e-01 -1.05400670e+00 -8.30786467e-01 4.92270477e-02
4.47243452e-01 6.08384050e-02 -9.80316997e-02 6.17134213e-01
-2.58843452e-01 3.71763557e-01 2.99273487e-02 -6.46864235e-01
-1.11821818e+00 5.23311853e-01 4.47996795e-01 -1.03147283e-01
-4.61548626e-01 5.34913957e-01 4.81653452e-01 -5.89517832e-01
3.97608355e-02 -4.21458632e-01 -5.02787977e-02 -2.97604561e-01
4.58525777e-01 -7.72514381e-03 -1.47753045e-01 -2.36346796e-01
-3.88525665e-01 1.28975242e-01 1.47219617e-02 -2.95096159e-01
1.28380930e+00 -1.01039195e-02 -1.14930905e-02 7.38763332e-01
4.72966611e-01 -8.42439905e-02 -1.49005592e+00 7.31973648e-02
9.36745778e-02 -4.94679809e-01 1.57231525e-01 -1.36671674e+00
-6.93893015e-01 7.45058298e-01 2.86140651e-01 2.87457466e-01
8.89749825e-01 2.39718422e-01 -6.42584264e-02 1.39980733e-01
1.76717833e-01 -5.19416511e-01 -2.82659262e-01 -2.44636342e-01
1.16939807e+00 -1.31328082e+00 2.57276952e-01 -2.45921671e-01
-7.63491213e-01 1.04737103e+00 7.41370201e-01 1.79043055e-01
1.21644728e-01 3.17395641e-03 2.70736516e-01 -3.32070589e-01
-1.26396763e+00 2.65797615e-01 5.04274011e-01 6.06902182e-01
1.92517877e-01 2.16757700e-01 -5.77375777e-02 7.07965314e-01
-4.31190163e-01 -1.56904459e-01 5.53814054e-01 6.40921950e-01
-3.63986969e-01 -9.95440125e-01 -5.73583901e-01 3.77950817e-01
2.51472406e-02 -5.63964397e-02 -8.76537383e-01 1.07064736e+00
5.68716526e-02 1.14314318e+00 -1.69773251e-01 -2.41090998e-01
6.27334835e-03 2.66684562e-01 1.47388056e-01 -6.12652421e-01
-5.48091114e-01 -2.69811660e-01 1.05363376e-01 -4.10988122e-01
-3.38605613e-01 -6.33949399e-01 -1.34361255e+00 -4.46760923e-01
-4.19223100e-01 2.61138648e-01 8.40885878e-01 1.30544865e+00
3.54256362e-01 5.81917882e-01 4.19331312e-01 -6.50879323e-01
-6.72418773e-01 -8.27788174e-01 -2.40526244e-01 5.12742519e-01
2.71006048e-01 -6.55434370e-01 -7.31818914e-01 6.79165423e-02]
|
[8.833276748657227, 5.699516773223877]
|
32d57297-6be8-4e1c-b2fd-563e11c73e39
|
learning-networks-from-random-walk-based-node
|
1801.07386
| null |
http://arxiv.org/abs/1801.07386v1
|
http://arxiv.org/pdf/1801.07386v1.pdf
|
Learning Networks from Random Walk-Based Node Similarities
|
Digital presence in the world of online social media entails significant
privacy risks. In this work we consider a privacy threat to a social network in
which an attacker has access to a subset of random walk-based node
similarities, such as effective resistances (i.e., commute times) or
personalized PageRank scores. Using these similarities, the attacker's goal is
to infer as much information as possible about the underlying network,
including any remaining unknown pairwise node similarities and edges.
For the effective resistance metric, we show that with just a small subset of
measurements, the attacker can learn a large fraction of edges in a social
network, even when the measurements are noisy. We also show that it is possible
to learn a graph which accurately matches the underlying network on all other
effective resistances. This second observation is interesting from a data
mining perspective, since it can be expensive to accurately compute all
effective resistances. As an alternative, our graphs learned from just a subset
of approximate effective resistances can be used as surrogates in a wide range
of applications that use effective resistances to probe graph structure,
including for graph clustering, node centrality evaluation, and anomaly
detection.
We obtain our results by formalizing the graph learning objective
mathematically, using two optimization problems. One formulation is convex and
can be solved provably in polynomial time. The other is not, but we solve it
efficiently with projected gradient and coordinate descent. We demonstrate the
effectiveness of these methods on a number of social networks obtained from
Facebook. We also discuss how our methods can be generalized to other random
walk-based similarities, such as personalized PageRank. Our code is available
at https://github.com/cnmusco/graph-similarity-learning.
|
['Cameron Musco', 'Jeremy G. Hoskins', 'Charalampos E. Tsourakakis', 'Christopher Musco']
|
2018-01-23
| null | null | null | null |
['graph-similarity']
|
['graphs']
|
[ 9.26794037e-02 5.56099534e-01 -2.94534802e-01 -6.16094135e-02
-5.50719082e-01 -9.71746385e-01 2.18641371e-01 6.29550993e-01
-3.07509124e-01 5.15612364e-01 -7.10883662e-02 -4.80424315e-01
-5.43318033e-01 -1.24338889e+00 -7.30688751e-01 -5.24810731e-01
-7.28061676e-01 5.13803899e-01 7.17831329e-02 -1.23002872e-01
1.45297199e-01 6.18851781e-01 -6.33623064e-01 -3.28096658e-01
5.28966844e-01 5.91053009e-01 -4.72944468e-01 8.45752716e-01
3.14759642e-01 5.79568446e-01 -3.25369358e-01 -7.27482259e-01
6.03520393e-01 -5.25207855e-02 -8.14498723e-01 -1.22369222e-01
2.85709351e-01 -2.67601609e-01 -7.53486633e-01 1.34496021e+00
3.56869489e-01 -3.82314064e-02 2.80504882e-01 -1.71613085e+00
-1.97613478e-01 7.45324969e-01 -6.97481275e-01 8.34822357e-02
7.34605074e-01 -2.45260864e-01 1.37938845e+00 -1.37114853e-01
5.53855360e-01 9.71069336e-01 9.31981206e-01 9.05874670e-02
-1.47956920e+00 -6.59828126e-01 5.57906777e-02 2.31571309e-02
-1.50363874e+00 5.51486202e-03 8.82355511e-01 -3.08586121e-01
3.40197176e-01 7.98835695e-01 6.41802490e-01 1.01672220e+00
-1.94027618e-01 3.76040518e-01 8.34974766e-01 -2.67283190e-02
2.06586480e-01 1.70837253e-01 2.67884821e-01 9.06709552e-01
7.84768045e-01 -5.00188358e-02 -1.58895165e-01 -8.97284508e-01
4.12029266e-01 3.11576575e-01 -5.21966338e-01 -5.55718541e-01
-8.64630938e-01 9.89809990e-01 5.71024895e-01 1.22577004e-01
4.87830266e-02 1.74062699e-01 1.64149702e-01 6.64064825e-01
3.94717455e-01 5.72424948e-01 -4.21918482e-01 5.05445004e-02
-5.01460433e-01 6.71292543e-02 1.61288488e+00 6.91664696e-01
1.12683094e+00 -5.23023367e-01 3.59685987e-01 2.50196755e-01
1.76030725e-01 5.98987520e-01 -2.92364717e-01 -1.01091576e+00
5.46417654e-01 4.34496760e-01 1.80728510e-01 -1.91260612e+00
-3.67991537e-01 -2.24824637e-01 -1.01449549e+00 -2.59910971e-01
7.40961492e-01 -3.61838520e-01 -1.64170995e-01 1.89320004e+00
5.69484115e-01 4.27974164e-01 -3.86832088e-01 5.74790955e-01
1.66131452e-01 3.36625934e-01 -4.49226648e-01 -2.96928257e-01
9.05674398e-01 -4.25530732e-01 -2.64035374e-01 -1.86833099e-01
9.10034180e-01 -2.47004971e-01 7.91402042e-01 1.96235597e-01
-7.46273100e-01 3.22613865e-01 -9.26928401e-01 1.60052136e-01
-4.07231390e-01 -6.21559858e-01 7.76469231e-01 1.10585344e+00
-1.15349245e+00 1.00461698e+00 -8.90331268e-01 -5.15476286e-01
3.80029231e-01 5.25243461e-01 -4.76886362e-01 -4.07519378e-02
-1.23129654e+00 3.20466816e-01 -6.62507303e-03 -2.02262938e-01
-5.34882247e-01 -7.06178248e-01 -7.82607138e-01 1.56490669e-01
6.98677123e-01 -3.99958223e-01 7.71009207e-01 -6.69120193e-01
-9.94214594e-01 6.95999980e-01 -3.59401964e-02 -5.12803853e-01
7.33524144e-01 2.27305174e-01 -2.39489704e-01 3.71012598e-01
2.87236664e-02 -3.41023535e-01 5.03358126e-01 -9.59497333e-01
-1.36235550e-01 -5.91689646e-01 4.48149800e-01 -6.23246059e-02
-8.00623953e-01 -3.30841988e-01 -2.94402272e-01 -2.88038284e-01
1.33448064e-01 -1.20846057e+00 -5.43449461e-01 3.44851792e-01
-8.64624441e-01 2.06904128e-01 6.21913195e-01 -5.41959763e-01
1.30840492e+00 -1.76349533e+00 -2.70980783e-02 1.23584890e+00
9.14201796e-01 -7.50758350e-02 -1.56043783e-01 8.37282419e-01
6.56910762e-02 6.89984202e-01 -3.50356042e-01 -1.83978364e-01
-3.23195606e-02 1.64398476e-02 -8.24074596e-02 1.06837320e+00
-4.91975754e-01 7.73200989e-01 -1.12667346e+00 -1.47297964e-01
-7.85732269e-02 1.61623180e-01 -4.88264203e-01 1.24771288e-02
6.97699338e-02 3.77286822e-01 -7.29976118e-01 2.16931581e-01
8.24669421e-01 -6.47721827e-01 7.35384226e-01 1.62103668e-01
5.61923027e-01 1.94499835e-01 -1.31027651e+00 1.25159013e+00
-3.19773108e-01 5.18240571e-01 3.18540394e-01 -1.14030504e+00
5.80034435e-01 1.54440731e-01 7.41522431e-01 -5.36373667e-02
3.73556353e-02 8.99953321e-02 -1.41686216e-01 -2.56821781e-01
-8.91848579e-02 3.32497090e-01 -1.29375875e-01 1.00707030e+00
-4.25327569e-01 3.73892844e-01 3.42099555e-02 7.67467439e-01
1.95700324e+00 -9.18742597e-01 3.61279517e-01 -1.83941245e-01
3.59038860e-01 -4.26424742e-01 3.62441897e-01 8.21615160e-01
6.31445972e-03 2.62429804e-01 1.17530870e+00 -1.18035547e-01
-1.00131869e+00 -1.34574997e+00 1.30597174e-01 5.05455911e-01
2.18317524e-01 -9.18431401e-01 -6.51444077e-01 -9.75647867e-01
4.00657088e-01 7.92287812e-02 -5.74323356e-01 -2.13903159e-01
-2.77288169e-01 -7.35505044e-01 5.19197583e-01 -8.83013308e-02
2.12627187e-01 -1.42677709e-01 3.80570173e-01 -1.31297320e-01
-1.89593300e-01 -1.21227217e+00 -7.43769288e-01 -3.53976279e-01
-7.13311434e-01 -1.65678012e+00 -7.00361058e-02 -2.70755112e-01
9.94128942e-01 5.53034842e-01 8.73684168e-01 6.61917686e-01
-5.61800152e-02 7.82469630e-01 -1.64518222e-01 6.96783066e-02
-2.56305128e-01 3.27528924e-01 1.28261030e-01 3.38731974e-01
2.75696784e-01 -1.30246747e+00 -5.31802058e-01 2.65718311e-01
-7.11252153e-01 -4.93358701e-01 1.45225570e-01 3.91334057e-01
3.23241025e-01 2.51640975e-01 2.62865305e-01 -1.63525677e+00
8.22481751e-01 -1.00222170e+00 -8.67532432e-01 1.19199961e-01
-6.92374051e-01 -3.01981475e-02 7.57195473e-01 -4.61744487e-01
-8.23867396e-02 -4.66342345e-02 1.45171210e-01 -2.10722283e-01
3.67342144e-01 5.33754110e-01 -3.36075634e-01 -6.53262019e-01
6.90403402e-01 -8.36238042e-02 9.80994552e-02 -2.26264715e-01
4.05935645e-01 5.79995215e-01 -6.86753318e-02 -5.23938894e-01
1.37971795e+00 6.89248919e-01 5.99723220e-01 -1.00169599e+00
-6.24700785e-01 -6.18959010e-01 -2.81448394e-01 2.98251584e-02
1.32740870e-01 -5.54781199e-01 -1.40300441e+00 1.66856766e-01
-8.85704935e-01 -2.59647489e-01 -2.00913101e-01 2.81895399e-01
-2.56958097e-01 9.89623606e-01 -6.80241704e-01 -8.48365963e-01
-2.03777626e-01 -6.05184734e-01 5.32285631e-01 -2.20860705e-01
-3.75522316e-01 -1.51735497e+00 1.06394112e-01 2.73084342e-01
1.09347567e-01 7.48154581e-01 6.64974391e-01 -9.81286943e-01
-8.18202436e-01 -7.79898882e-01 -2.61409968e-01 4.03925730e-03
1.56768262e-01 -1.99028254e-01 -6.13671660e-01 -6.71670318e-01
-1.61813065e-01 1.89364366e-02 4.99287099e-01 -8.65156502e-02
1.40793490e+00 -1.03474903e+00 -5.07122159e-01 8.55156839e-01
1.28933108e+00 -5.37801266e-01 4.66026932e-01 -2.70873439e-02
9.53239858e-01 5.37706614e-01 1.64789557e-01 6.21338248e-01
4.58034217e-01 4.20849383e-01 4.70600128e-01 1.08029157e-01
5.71534574e-01 -6.86514080e-01 1.82307467e-01 5.69232702e-01
1.05184399e-01 -3.11410874e-01 -7.50880957e-01 2.86964804e-01
-1.84171188e+00 -8.73572528e-01 -2.57012963e-01 2.65137672e+00
5.70110738e-01 1.88790839e-02 4.45434660e-01 1.72830909e-01
8.73436511e-01 3.08495760e-01 -6.73497856e-01 -1.32102028e-01
2.40642130e-02 1.85596377e-01 1.14105380e+00 1.06265092e+00
-8.77930820e-01 3.83530408e-01 5.26995516e+00 5.78701913e-01
-7.10733473e-01 1.16006240e-01 7.03476369e-01 -9.05948654e-02
-6.19981349e-01 4.09057438e-01 -2.44894013e-01 5.14164567e-01
1.05596685e+00 -5.86469710e-01 1.09824586e+00 7.74675071e-01
2.73455400e-02 1.20194070e-01 -1.21097946e+00 9.31052029e-01
-2.14187890e-01 -1.12062871e+00 -3.57928932e-01 7.89955497e-01
6.46789789e-01 1.92581248e-02 -6.35306314e-02 -2.55152792e-01
6.10996246e-01 -9.37919199e-01 -1.15757912e-01 3.07231039e-01
8.17154467e-01 -8.01931679e-01 2.96764970e-01 4.05239642e-01
-1.25882542e+00 -7.48456791e-02 -3.72632951e-01 -2.99270675e-02
-1.33645192e-01 1.00864327e+00 -8.89864802e-01 4.84991908e-01
3.74791801e-01 7.70160079e-01 -4.18325126e-01 8.90783906e-01
-3.84719759e-01 6.72963679e-01 -9.26161826e-01 -1.82303667e-01
-1.98452711e-01 -5.00407577e-01 9.64742064e-01 6.87011957e-01
3.46193969e-01 1.28966525e-01 1.80474043e-01 6.77404344e-01
-6.19053483e-01 2.53502876e-01 -8.76771271e-01 -2.79526412e-01
8.19825470e-01 1.47711766e+00 -4.89588767e-01 1.68270081e-01
-3.18550140e-01 1.03560281e+00 7.12058663e-01 5.13946116e-01
-5.54196656e-01 -5.75319767e-01 9.16483879e-01 7.44612396e-01
-1.72945306e-01 -3.78763765e-01 1.84397474e-02 -1.42432868e+00
2.78428465e-01 -7.39348292e-01 5.68163514e-01 -1.08016737e-01
-1.52250266e+00 2.02791899e-01 -2.34632120e-01 -8.79285514e-01
-1.63048327e-01 -4.07550097e-01 -7.28128850e-01 6.17211699e-01
-1.02590144e+00 -6.79539979e-01 -1.95616573e-01 7.55103528e-01
-6.94334626e-01 1.83193520e-01 8.01012814e-01 2.27237120e-01
-4.87065971e-01 8.23941946e-01 2.44348437e-01 4.20535058e-01
3.48668367e-01 -1.49874198e+00 5.81301212e-01 9.26765859e-01
4.52781498e-01 5.40327132e-01 5.55864751e-01 -7.74829447e-01
-1.74755108e+00 -1.02083468e+00 8.04483831e-01 -5.10515749e-01
1.16495538e+00 -7.15548873e-01 -6.69456542e-01 1.02443218e+00
-4.64908153e-01 4.76058245e-01 8.00855935e-01 3.84539008e-01
-4.68586147e-01 -4.34688121e-01 -1.53066647e+00 7.50565529e-01
1.35042691e+00 -7.34689772e-01 3.92042637e-01 8.36881042e-01
5.58187246e-01 -2.08348826e-01 -1.03573859e+00 -1.04957931e-02
4.03714567e-01 -7.60352850e-01 1.00429237e+00 -6.91835880e-01
-2.98126459e-01 -5.76451160e-02 -7.64162242e-02 -1.18537199e+00
-1.81664482e-01 -1.39723611e+00 -6.03865802e-01 9.12055612e-01
5.52240670e-01 -1.32902837e+00 1.31257510e+00 8.27748835e-01
9.93481994e-01 -7.84000635e-01 -8.92449319e-01 -6.98569119e-01
-1.46115318e-01 -3.06541532e-01 6.49710655e-01 1.27144635e+00
2.27709919e-01 3.17202419e-01 -4.62355018e-01 6.81713760e-01
1.07844281e+00 -1.03438087e-01 1.02564788e+00 -1.50273609e+00
-5.54501534e-01 -9.00772363e-02 -7.05788195e-01 -9.18591857e-01
3.05555046e-01 -1.17704737e+00 -8.14538360e-01 -1.16189384e+00
1.72713801e-01 -6.72747672e-01 -4.99771489e-03 2.30192408e-01
1.10987708e-01 7.67288134e-02 1.30702868e-01 5.70941530e-02
-5.35158396e-01 3.43023501e-02 1.03816581e+00 -2.85539608e-02
-2.38829762e-01 6.36625350e-01 -9.96501923e-01 6.80687189e-01
1.06506705e+00 -6.94073021e-01 -5.86027861e-01 1.79316089e-01
7.51722455e-01 2.67245412e-01 3.67137194e-01 -7.46863782e-01
4.60073471e-01 -9.38966125e-02 6.44849846e-03 3.68269607e-02
3.97427619e-01 -9.70996916e-01 3.00363064e-01 4.84112769e-01
-3.09967071e-01 -9.55249667e-02 -3.71254534e-01 1.17184401e+00
4.00344580e-01 -2.76021771e-02 5.99855423e-01 -6.24286593e-04
2.31139243e-01 9.11049306e-01 4.53853123e-02 4.08714473e-01
1.09618151e+00 -9.98164564e-02 -4.69362557e-01 -1.19891310e+00
-7.18394399e-01 3.35894972e-01 6.80592120e-01 -4.98093739e-02
5.46602547e-01 -1.17955792e+00 -5.89567602e-01 -1.90723259e-02
-4.53429893e-02 -3.68326366e-01 6.19246112e-03 1.03550541e+00
-4.23437774e-01 -1.98222101e-02 3.30416352e-01 -1.96558580e-01
-1.34606707e+00 6.98758245e-01 3.89012039e-01 -4.05083328e-01
-3.65798444e-01 5.35492301e-01 -1.41924694e-01 -5.86336732e-01
1.82339564e-01 2.08710730e-01 3.66569221e-01 -2.57536054e-01
4.63024020e-01 7.05402017e-01 -2.59645641e-01 -4.64571118e-01
-2.85542399e-01 4.22569722e-01 -6.77711442e-02 4.12493385e-03
1.24100471e+00 -3.52236718e-01 -4.21736121e-01 1.62302569e-01
1.76691842e+00 6.53517187e-01 -8.31488132e-01 -3.71772528e-01
1.30285248e-01 -8.59333575e-01 -2.27952942e-01 -2.03137860e-01
-1.36081433e+00 5.42668104e-01 7.34212101e-02 8.62285376e-01
9.88985360e-01 1.47539601e-02 8.56023908e-01 7.66059339e-01
5.37745118e-01 -6.18378997e-01 -2.00685859e-01 9.45605487e-02
3.67372423e-01 -1.11532009e+00 1.78630725e-01 -9.20649767e-01
-6.91792890e-02 9.46251810e-01 2.05885425e-01 -3.20123136e-01
1.20970070e+00 1.11980386e-01 -5.23781598e-01 -1.18331358e-01
-4.15777177e-01 5.88046685e-02 -4.26917300e-02 6.35925710e-01
8.89524911e-03 2.67890126e-01 -1.79940119e-01 4.30758268e-01
-3.02845269e-01 -6.03133023e-01 8.98392141e-01 5.80963194e-01
-2.47975066e-01 -1.40313268e+00 8.92094895e-02 8.87908936e-01
-7.08084524e-01 -4.57829759e-02 -7.87491024e-01 5.17381012e-01
-5.94437063e-01 1.05741525e+00 -3.38957638e-01 -5.84698260e-01
8.53146091e-02 -6.33208275e-01 2.46134296e-01 -6.01022959e-01
-2.17947230e-01 -6.36236668e-01 3.56414080e-01 -8.74162972e-01
4.01561745e-02 -6.46881759e-01 -8.16789031e-01 -1.16533649e+00
-2.39445359e-01 4.55916703e-01 4.80908513e-01 5.55985987e-01
4.30581748e-01 -3.10070544e-01 1.29136777e+00 -2.77862430e-01
-6.73676431e-01 -5.54794669e-01 -9.95051980e-01 3.75653118e-01
4.22122955e-01 -8.35868493e-02 -1.02185285e+00 -6.70476794e-01]
|
[6.775928497314453, 5.600727081298828]
|
21d3e58c-312a-41ff-97f7-c5bf366835c8
|
adversarial-audio-synthesis
|
1802.04208
| null |
http://arxiv.org/abs/1802.04208v3
|
http://arxiv.org/pdf/1802.04208v3.pdf
|
Adversarial Audio Synthesis
|
Audio signals are sampled at high temporal resolutions, and learning to
synthesize audio requires capturing structure across a range of timescales.
Generative adversarial networks (GANs) have seen wide success at generating
images that are both locally and globally coherent, but they have seen little
application to audio generation. In this paper we introduce WaveGAN, a first
attempt at applying GANs to unsupervised synthesis of raw-waveform audio.
WaveGAN is capable of synthesizing one second slices of audio waveforms with
global coherence, suitable for sound effect generation. Our experiments
demonstrate that, without labels, WaveGAN learns to produce intelligible words
when trained on a small-vocabulary speech dataset, and can also synthesize
audio from other domains such as drums, bird vocalizations, and piano. We
compare WaveGAN to a method which applies GANs designed for image generation on
image-like audio feature representations, finding both approaches to be
promising.
|
['Chris Donahue', 'Miller Puckette', 'Julian McAuley']
|
2018-02-12
|
adversarial-audio-synthesis-1
|
https://openreview.net/forum?id=ByMVTsR5KQ
|
https://openreview.net/pdf?id=ByMVTsR5KQ
|
iclr-2019-5
|
['audio-generation']
|
['audio']
|
[ 5.75289965e-01 4.93267864e-01 3.10206592e-01 5.65681420e-02
-1.45416760e+00 -9.55822408e-01 8.73904526e-01 -7.88560152e-01
5.40755510e-01 9.37590063e-01 6.54901862e-01 3.67744304e-02
3.82607549e-01 -9.89979565e-01 -8.83057475e-01 -7.54710853e-01
-2.46717334e-01 2.65302867e-01 -2.20642820e-01 -2.69761622e-01
-2.13021591e-01 2.41490543e-01 -1.63131368e+00 5.20887673e-01
2.29008183e-01 7.50167787e-01 -1.68828681e-01 1.36374557e+00
1.09880783e-01 1.01178348e+00 -1.36126363e+00 -2.27219686e-01
2.33290151e-01 -1.24730396e+00 -6.49060965e-01 1.28302742e-02
5.34855008e-01 -2.91877478e-01 -3.35593253e-01 5.81569374e-01
8.95782232e-01 5.52486032e-02 8.96305203e-01 -1.33698773e+00
-9.10572648e-01 9.56132054e-01 3.89884785e-02 -9.07787979e-02
6.34266913e-01 5.59397459e-01 1.12030542e+00 -6.37995422e-01
7.13773608e-01 1.33509648e+00 7.11309373e-01 1.05577946e+00
-1.50585735e+00 -9.73794758e-01 -7.14514673e-01 -3.21154386e-01
-1.01763999e+00 -7.99315751e-01 8.11798275e-01 -4.11674529e-01
6.95561290e-01 3.76160830e-01 7.74309695e-01 1.60747826e+00
1.42387003e-01 5.44860780e-01 1.10457861e+00 -4.68017876e-01
2.34236792e-01 -3.17047030e-01 -1.17515802e+00 3.65014911e-01
-5.13442516e-01 6.46587610e-01 -9.87283587e-01 -1.53652102e-01
1.02908480e+00 -7.98320234e-01 -2.64168620e-01 3.20299745e-01
-1.49205363e+00 9.88611937e-01 2.87771910e-01 2.84214407e-01
-3.20870668e-01 1.05913556e+00 2.26117074e-01 6.76790714e-01
3.35025847e-01 8.79870951e-01 3.51687558e-02 -3.33547294e-01
-1.07819700e+00 4.77078199e-01 6.57907307e-01 9.02311623e-01
3.16799402e-01 1.28742218e+00 -3.14584196e-01 5.68022609e-01
2.06865254e-04 8.61941993e-01 7.65611589e-01 -1.31538963e+00
9.46293399e-02 -5.48287332e-01 -3.30324546e-02 -7.24376261e-01
1.65073544e-01 -9.82462168e-02 -8.08203518e-01 4.11795646e-01
1.41891912e-01 -6.27477765e-01 -1.08112741e+00 1.90122211e+00
3.79802473e-02 5.92967689e-01 4.15601373e-01 5.44074178e-01
1.21123564e+00 1.26238358e+00 -1.86192021e-01 6.27040341e-02
1.05414748e+00 -8.08212757e-01 -7.90524662e-01 8.10472369e-02
-2.02962115e-01 -1.17562830e+00 1.00195479e+00 4.59707707e-01
-1.53921962e+00 -8.45347345e-01 -1.02097809e+00 1.66593507e-01
-8.98868591e-02 -3.94982845e-01 5.78060627e-01 8.77610683e-01
-1.54117250e+00 4.05428529e-01 -3.41104209e-01 1.28396094e-01
1.36568993e-01 2.86468547e-02 -1.72856361e-01 4.24227655e-01
-1.38062978e+00 3.72553080e-01 1.18303616e-02 -4.12293911e-01
-1.84933627e+00 -9.56877351e-01 -8.65675271e-01 -2.35274378e-02
-3.41009468e-01 -8.59762073e-01 1.67273915e+00 -1.11358571e+00
-2.10738444e+00 4.53432769e-01 1.25309601e-01 -8.16923976e-01
2.37234965e-01 2.26227731e-01 -6.79065645e-01 4.68225032e-01
4.05667014e-02 1.37174082e+00 1.59615338e+00 -1.28643298e+00
-3.10786933e-01 5.60869455e-01 -8.22688863e-02 6.35225847e-02
-4.73664552e-02 -6.67733653e-03 3.41890186e-01 -1.20175827e+00
-4.94349837e-01 -8.59492660e-01 -8.73263329e-02 -4.09894109e-01
-3.33591908e-01 1.54890090e-01 8.49057198e-01 -4.96513665e-01
5.94217837e-01 -1.97802103e+00 2.31288552e-01 -4.16490361e-02
-3.68504599e-02 -6.99897036e-02 -5.70982277e-01 8.64231825e-01
-2.15387806e-01 4.25771236e-01 -9.71758887e-02 -1.83965549e-01
8.20998549e-02 3.01007181e-02 -1.11378074e+00 -9.32163000e-02
5.53980350e-01 1.14921701e+00 -9.45097387e-01 -1.90197513e-01
-4.56252396e-02 8.12216997e-01 -7.42445230e-01 5.36274254e-01
-5.56371152e-01 1.10263264e+00 -1.69107735e-01 4.59451973e-01
3.30158435e-02 2.57158428e-01 -2.47259632e-01 2.77707994e-01
1.12785712e-01 4.68654156e-01 -7.33972907e-01 1.68106401e+00
-9.02592719e-01 1.14239073e+00 -8.96657910e-03 -5.58975279e-01
1.05938578e+00 1.11113214e+00 4.51674342e-01 -3.36514920e-01
-1.26231670e-01 2.15757415e-01 -6.41055405e-02 -2.46575534e-01
3.06504518e-01 -7.36169934e-01 -2.82973647e-01 7.48266578e-01
5.02359331e-01 -1.40375555e+00 -9.01653320e-02 1.74719878e-02
1.16128111e+00 5.53796999e-02 -3.40834148e-02 1.02458589e-01
8.59794244e-02 -2.16829479e-01 -6.95450604e-02 6.77728593e-01
4.84280080e-01 1.26252711e+00 2.57425696e-01 -2.21899495e-01
-1.38985252e+00 -1.42298496e+00 2.49971196e-01 8.01420271e-01
-5.64334393e-01 -3.64173084e-01 -9.88708675e-01 -2.22070497e-02
-4.40286070e-01 7.05912769e-01 -4.64157730e-01 -2.27907032e-01
-5.63743114e-01 -1.23492487e-01 1.33912277e+00 3.34437668e-01
2.37325862e-01 -1.78626060e+00 -5.26019096e-01 5.27015984e-01
-2.97105998e-01 -8.57130349e-01 -6.77718341e-01 8.27346519e-02
-6.26762211e-01 -4.71294880e-01 -1.11362338e+00 -9.10766184e-01
1.43321484e-01 -2.08838791e-01 1.51043963e+00 -3.28121096e-01
-4.71069455e-01 6.75983965e-01 -4.29629415e-01 -8.62364650e-01
-1.06427419e+00 -8.69856179e-02 1.14782438e-01 -6.70916885e-02
-4.68096763e-01 -1.08400989e+00 -4.21445847e-01 1.77177973e-02
-1.26958692e+00 -7.31597468e-02 2.29666725e-01 1.00184166e+00
4.66116071e-01 1.82154298e-01 1.12919521e+00 -6.28369331e-01
1.06482136e+00 -3.63723874e-01 -2.17342108e-01 -3.07779908e-01
-2.12914571e-02 -1.27896473e-01 8.51553500e-01 -6.56457365e-01
-9.50234830e-01 -1.60067141e-01 -4.43136454e-01 -5.37730694e-01
-2.10616767e-01 2.68426269e-01 2.26611570e-01 1.82147712e-01
1.11941731e+00 3.78993630e-01 8.23680758e-02 -2.94747092e-02
7.97304571e-01 5.09537220e-01 9.33409810e-01 -6.92016602e-01
1.21006489e+00 3.51295084e-01 -1.80781916e-01 -1.00233269e+00
-4.48038667e-01 3.74831349e-01 9.46225785e-03 -2.81597227e-01
9.16289210e-01 -1.17461157e+00 -3.22044671e-01 3.58513504e-01
-1.08515120e+00 -7.74603903e-01 -1.14965558e+00 2.87960052e-01
-1.24810731e+00 -4.01511699e-01 -8.20647776e-01 -6.26200855e-01
-5.56135237e-01 -7.64827728e-01 1.32912326e+00 1.85124636e-01
-5.28897643e-01 -8.30898702e-01 4.45329607e-01 -1.30563276e-02
7.07306147e-01 7.51618266e-01 6.30490005e-01 -7.59428889e-02
-6.89299703e-01 8.23031142e-02 5.00959039e-01 4.19468284e-01
3.78387243e-01 2.34332517e-01 -1.23969305e+00 -2.69466072e-01
-3.97575870e-02 -8.41952980e-01 5.74524641e-01 5.12632549e-01
8.58894646e-01 -7.12960660e-01 3.23164314e-01 6.12736583e-01
1.02570117e+00 4.79939997e-01 9.09103334e-01 -3.46279174e-01
3.98527920e-01 4.45278734e-01 5.64401597e-02 2.82874197e-01
-1.72485322e-01 6.29416108e-01 2.46839449e-01 -2.38292113e-01
-9.74153459e-01 -7.39520490e-01 8.22662830e-01 1.15423584e+00
7.96305854e-03 -4.25092191e-01 -4.36774492e-01 8.26994658e-01
-1.04028499e+00 -1.42863655e+00 3.55872869e-01 1.54086316e+00
1.29933536e+00 -1.21093102e-01 2.57711619e-01 3.61694872e-01
5.49250841e-01 3.53753507e-01 -3.21082264e-01 -8.38133931e-01
-1.49548724e-01 1.24474430e+00 -1.13587208e-01 4.17492598e-01
-8.01948071e-01 8.79730880e-01 7.95325661e+00 8.62350285e-01
-1.27448440e+00 1.39106363e-01 5.78262269e-01 -1.42338693e-01
-8.65290582e-01 -2.23012209e-01 -2.04914451e-01 3.07743877e-01
1.40530777e+00 -3.59529495e-01 8.35075855e-01 5.37117898e-01
7.57144317e-02 6.33533001e-01 -1.01349342e+00 7.51515031e-01
-3.02812643e-02 -1.76415730e+00 3.44261140e-01 -1.83200061e-01
1.35480773e+00 -3.69181931e-01 8.11508358e-01 2.36863688e-01
7.73962200e-01 -1.80347013e+00 1.00678992e+00 3.41835111e-01
1.50277030e+00 -8.73698235e-01 2.06235245e-01 3.45072784e-02
-1.21144044e+00 2.89375216e-01 -1.03229754e-01 7.87164643e-03
4.16939199e-01 3.23984861e-01 -1.29600000e+00 1.09120689e-01
5.71039438e-01 3.83837998e-01 -9.30202380e-02 6.27159894e-01
-4.64274347e-01 1.11398494e+00 -1.13126732e-01 1.18905395e-01
3.26292694e-01 9.29128081e-02 6.42996550e-01 1.04459214e+00
9.39746141e-01 -1.02445558e-01 -2.58952856e-01 1.03828859e+00
-3.45966995e-01 -2.41363406e-01 -1.25361586e+00 -5.12613952e-01
5.10491788e-01 9.82758045e-01 -3.48166347e-01 -2.36422300e-01
6.82533383e-02 7.05776215e-01 -5.52971244e-01 3.25699508e-01
-9.51108098e-01 -5.42609990e-01 6.98325813e-01 -9.96044371e-03
4.12943274e-01 -1.70839489e-01 -1.97720937e-02 -7.65108526e-01
-5.24620414e-01 -1.41108477e+00 -1.10447094e-01 -1.34587049e+00
-1.20066428e+00 9.73386526e-01 -2.66757280e-01 -1.51682699e+00
-1.20625019e+00 4.42689769e-02 -9.62452054e-01 7.51554072e-01
-9.95794415e-01 -1.33137310e+00 -7.01486990e-02 6.20259881e-01
8.34800363e-01 -4.94086891e-01 1.28142262e+00 -2.34960336e-02
3.92127037e-01 4.96146530e-01 -2.33038545e-01 -3.32757458e-02
6.47655666e-01 -1.27357638e+00 9.26789939e-01 6.46102965e-01
8.74921679e-01 1.94080293e-01 1.03257859e+00 -2.48814732e-01
-1.12553477e+00 -1.25763059e+00 4.70176935e-01 -4.56957012e-01
4.82835412e-01 -3.03236544e-01 -3.36978376e-01 6.77528441e-01
9.53951120e-01 -3.19714755e-01 7.16830313e-01 -5.73634267e-01
-4.32037383e-01 3.96495946e-02 -1.23251963e+00 6.64588571e-01
9.24047589e-01 -7.90071189e-01 -5.03273070e-01 3.18784028e-01
1.15575886e+00 -4.97179121e-01 -1.05509460e+00 1.74907178e-01
3.75422090e-01 -7.93311298e-01 1.18739831e+00 -5.12074769e-01
1.01280117e+00 -3.45476717e-01 -2.67011812e-03 -1.79622746e+00
-2.98158117e-02 -1.51977336e+00 -1.05137834e-02 1.52097011e+00
3.67117971e-01 -3.55187684e-01 4.88255054e-01 -4.03054565e-01
-2.20455423e-01 -1.75962579e-02 -8.31637204e-01 -7.90381491e-01
2.72842199e-01 -2.44107276e-01 9.87768948e-01 6.65100276e-01
-3.24963123e-01 5.14393032e-01 -7.77080238e-01 -1.09415345e-01
4.24677312e-01 1.86306477e-01 1.12072587e+00 -7.42793024e-01
-5.27639031e-01 -3.96150857e-01 -3.47299993e-01 -7.02504158e-01
1.05851389e-01 -7.82843769e-01 4.69849795e-01 -1.18687940e+00
-5.73481262e-01 -2.75230139e-01 2.45896176e-01 2.49993786e-01
3.05910170e-01 1.01645803e+00 3.27274472e-01 -4.00962867e-03
2.49966279e-01 5.81824780e-01 1.61759508e+00 -4.11709905e-01
6.35789633e-02 -1.03289783e-02 -6.72138631e-01 3.80543947e-01
8.51327181e-01 -5.22635758e-01 -7.99093127e-01 -2.82645255e-01
1.18047819e-01 4.72210169e-01 3.45623821e-01 -1.37347376e+00
-1.45849258e-01 -4.77680974e-02 4.17680800e-01 -5.99730536e-02
6.77683830e-01 -2.49675274e-01 9.05983984e-01 3.08985323e-01
-7.31368959e-01 -4.28557768e-02 1.65576696e-01 4.64220136e-01
-7.36100435e-01 -1.03638612e-01 8.56273890e-01 -4.05019462e-01
-1.97357431e-01 1.75461635e-01 -7.29448974e-01 4.21292245e-01
6.89434052e-01 1.51144788e-01 -1.78422734e-01 -1.34894395e+00
-6.41834140e-01 -5.36825836e-01 2.86509573e-01 4.91449654e-01
7.78059304e-01 -1.81425667e+00 -1.06116521e+00 2.79243648e-01
-2.64535248e-01 -7.88238496e-02 7.18028843e-02 -1.07163703e-02
-6.26176715e-01 4.23238575e-01 -3.67274165e-01 -5.57671487e-01
-9.53509569e-01 3.23178262e-01 2.18042329e-01 -1.87314786e-02
-6.22296512e-01 1.13880610e+00 2.81412721e-01 -2.21778929e-01
-5.18420227e-02 -1.51479587e-01 -6.99469680e-03 1.91121046e-02
5.18634737e-01 -1.64548233e-01 -2.14716047e-01 -5.12882471e-01
1.72928944e-01 5.59674561e-01 7.98629701e-01 -8.77953887e-01
1.29208112e+00 3.31565619e-01 1.05475709e-01 4.88909960e-01
1.07932782e+00 4.90032434e-01 -1.32465553e+00 2.82277495e-01
-6.29510105e-01 -2.31492877e-01 -3.00947964e-01 -5.40090382e-01
-1.24964654e+00 9.83011901e-01 2.97578126e-01 7.28928685e-01
1.25099504e+00 1.64907873e-02 1.18782413e+00 -3.70880812e-02
5.11995077e-01 -5.31527936e-01 8.34579706e-01 2.80656755e-01
1.19425809e+00 -4.65055943e-01 -4.97158557e-01 1.82182509e-02
-7.60780513e-01 1.02371478e+00 1.23160064e-01 -4.02252376e-01
3.32930297e-01 6.26050651e-01 3.16813558e-01 3.00601800e-03
-1.11050916e+00 -4.25841846e-02 3.45878601e-01 1.04003060e+00
4.69928265e-01 1.15556665e-01 2.45034888e-01 7.73040652e-02
-1.14630687e+00 -2.43783206e-01 7.64955103e-01 5.00244498e-01
-1.69450790e-01 -1.14833462e+00 -5.26842177e-01 7.87111372e-02
-7.32079268e-01 -2.64546752e-01 -5.30055702e-01 5.62452197e-01
1.17357641e-01 1.15058112e+00 1.42064720e-01 -4.84538138e-01
-1.08770087e-01 2.26944327e-01 6.17613614e-01 -7.58950651e-01
-7.51095295e-01 2.51420110e-01 2.75303692e-01 -3.11299682e-01
-5.83761930e-01 -4.64369506e-01 -1.11220157e+00 -1.74058586e-01
-4.21778671e-02 3.77210885e-01 5.87090015e-01 3.38382155e-01
2.01312378e-01 1.02630937e+00 1.01921034e+00 -1.09791982e+00
-4.46049541e-01 -9.11885917e-01 -4.77387697e-01 3.70810598e-01
5.75141132e-01 9.16476082e-03 -4.53172863e-01 8.38757813e-01]
|
[15.583807945251465, 5.996086597442627]
|
4a5b1230-b322-41fc-9a6f-8b3bfffe3117
|
a-batch-noise-contrastive-estimation-approach
|
1708.05997
| null |
http://arxiv.org/abs/1708.05997v2
|
http://arxiv.org/pdf/1708.05997v2.pdf
|
A Batch Noise Contrastive Estimation Approach for Training Large Vocabulary Language Models
|
Training large vocabulary Neural Network Language Models (NNLMs) is a
difficult task due to the explicit requirement of the output layer
normalization, which typically involves the evaluation of the full softmax
function over the complete vocabulary. This paper proposes a Batch Noise
Contrastive Estimation (B-NCE) approach to alleviate this problem. This is
achieved by reducing the vocabulary, at each time step, to the target words in
the batch and then replacing the softmax by the noise contrastive estimation
approach, where these words play the role of targets and noise samples at the
same time. In doing so, the proposed approach can be fully formulated and
implemented using optimal dense matrix operations. Applying B-NCE to train
different NNLMs on the Large Text Compression Benchmark (LTCB) and the One
Billion Word Benchmark (OBWB) shows a significant reduction of the training
time with no noticeable degradation of the models performance. This paper also
presents a new baseline comparative study of different standard NNLMs on the
large OBWB on a single Titan-X GPU.
|
['Dietrich Klakow', 'Youssef Oualil']
|
2017-08-20
| null | null | null | null |
['text-compression']
|
['natural-language-processing']
|
[ 2.59999037e-01 -2.14089662e-01 3.67279857e-01 -5.02926707e-01
-7.92143822e-01 -4.66788262e-02 6.31503463e-01 1.84563190e-01
-1.22013879e+00 4.95068699e-01 1.15470499e-01 -5.57821810e-01
3.85691345e-01 -5.87183475e-01 -7.06979156e-01 -8.58628452e-01
3.25592399e-01 4.67918187e-01 1.06905118e-01 -1.28068760e-01
-2.41567772e-02 3.73544544e-01 -1.31276798e+00 2.46458203e-01
3.49036455e-01 9.63319480e-01 5.66354096e-01 9.52575922e-01
-5.09529471e-01 9.35580850e-01 -7.24611819e-01 -2.29213640e-01
2.34761342e-01 -9.54811871e-02 -3.05941671e-01 -1.53437987e-01
6.70614004e-01 -3.73771995e-01 -2.60499239e-01 1.11505342e+00
8.67099166e-01 5.43389976e-01 3.98176223e-01 -6.27743423e-01
-2.13530604e-02 7.42910683e-01 -5.41426301e-01 2.68656373e-01
-4.48544055e-01 -1.37592256e-01 8.05777192e-01 -1.06091368e+00
3.13478500e-01 1.39854980e+00 6.56824589e-01 3.69470596e-01
-1.08846295e+00 -8.08025599e-01 1.07771523e-01 1.25959381e-01
-1.49897707e+00 -6.74282551e-01 3.56739253e-01 -2.52625138e-01
1.43873084e+00 3.17515939e-01 2.35366359e-01 7.85204828e-01
1.66812935e-03 6.33585811e-01 7.55291760e-01 -7.64510512e-01
4.28759128e-01 1.38152912e-01 4.73743141e-01 5.77309072e-01
2.15633124e-01 -1.52478978e-01 -3.69999290e-01 -1.20036811e-01
2.99434781e-01 -2.26781443e-01 5.73263085e-03 2.06344604e-01
-8.51174116e-01 8.16307902e-01 6.45468161e-02 3.69907051e-01
-3.73187780e-01 4.20632899e-01 9.17359471e-01 3.96085769e-01
7.90003121e-01 8.45751911e-03 -6.28258228e-01 -2.08145350e-01
-1.41157174e+00 4.81702425e-02 9.20590699e-01 7.51888931e-01
6.63358033e-01 4.22494739e-01 -2.67312020e-01 1.13484061e+00
2.85060912e-01 5.04243553e-01 8.28382373e-01 -3.52565706e-01
6.49377704e-01 1.78346843e-01 -4.08439070e-01 -8.02004635e-01
-3.65045935e-01 -7.21180677e-01 -1.16957343e+00 1.20060816e-01
1.99982047e-01 -4.55321848e-01 -1.10618174e+00 1.54812574e+00
3.12879741e-01 3.10578108e-01 -1.41674802e-02 5.87876379e-01
8.14231753e-01 1.15667701e+00 3.11064273e-01 -2.74173409e-01
1.24287057e+00 -1.32038474e+00 -8.54963839e-01 -5.22013009e-01
1.07457960e+00 -9.26831901e-01 1.15617371e+00 5.83290160e-01
-9.07325447e-01 -6.73325896e-01 -1.15873253e+00 -3.32364917e-01
-4.70619738e-01 5.06568611e-01 2.51216412e-01 7.99063027e-01
-9.29671228e-01 4.56092894e-01 -9.74712431e-01 -6.97541833e-02
1.24373473e-01 6.46259964e-01 -1.39067248e-01 -5.55552281e-02
-1.03951764e+00 9.00030017e-01 7.08241582e-01 3.07391196e-01
-5.55365801e-01 -5.20056188e-01 -8.31700683e-01 3.56461227e-01
5.12072384e-01 -1.99311674e-01 1.17863321e+00 -9.87126112e-01
-1.68303525e+00 5.11349380e-01 -2.80154526e-01 -7.81571448e-01
3.61456752e-01 -3.81964356e-01 -2.19301328e-01 -3.28827739e-01
-4.19049114e-01 5.09090543e-01 1.03389096e+00 -7.14939773e-01
-4.91309464e-01 -1.12931885e-01 -2.68241823e-01 2.08825782e-01
-7.48863339e-01 2.73660541e-01 -7.79758930e-01 -1.02842224e+00
-2.02566579e-01 -7.82583117e-01 -3.38541925e-01 -3.38037252e-01
-2.10469708e-01 -1.97070763e-01 7.62287617e-01 -9.11767423e-01
1.68834817e+00 -2.27398419e+00 7.69831017e-02 3.47305328e-01
-3.47610116e-02 6.95778489e-01 -3.69479716e-01 1.03719078e-01
-4.90888162e-03 -2.19894513e-01 -1.71438083e-02 -9.91353691e-01
-7.50417188e-02 4.13079381e-01 -3.70413810e-01 3.72123927e-01
-1.04877397e-01 7.43705750e-01 -3.70929301e-01 -4.82023954e-01
2.45399639e-01 6.61664009e-01 -5.62361181e-01 1.49887264e-01
-2.77137578e-01 -2.21396863e-01 7.39765167e-02 -5.94923086e-02
6.80230200e-01 -6.04050383e-02 1.48707643e-01 -1.13676667e-01
-3.72354575e-02 4.37974453e-01 -1.42669010e+00 1.53315306e+00
-8.14167500e-01 8.51824939e-01 2.34128445e-01 -1.05639386e+00
8.01813245e-01 2.39889905e-01 8.16477090e-03 -6.45050466e-01
3.61629784e-01 3.15129250e-01 1.01770170e-01 -1.89156130e-01
5.96096992e-01 -1.28282115e-01 6.96521550e-02 4.26582277e-01
2.75097281e-01 -1.87306419e-01 2.53755152e-01 1.36168510e-01
8.05547059e-01 -1.71260864e-01 4.29093838e-01 -1.76103801e-01
7.71340251e-01 -2.58295685e-01 1.53155521e-01 1.00219738e+00
2.07640335e-01 3.59612137e-01 1.02886558e-01 -5.27854919e-01
-1.14547181e+00 -5.06977797e-01 1.97111279e-01 1.49292660e+00
-6.03358448e-01 -5.66140234e-01 -9.22695100e-01 -3.07317853e-01
-3.22764546e-01 8.97349000e-01 -3.38623255e-01 -1.00277863e-01
-8.61713767e-01 -8.46886516e-01 6.53924346e-01 3.48004818e-01
3.77827227e-01 -1.03651166e+00 -4.57470000e-01 2.60285854e-01
1.08077839e-01 -1.12663877e+00 -6.29916787e-01 7.15191662e-01
-7.25394487e-01 -2.64796406e-01 -5.89430690e-01 -9.18939471e-01
5.60866535e-01 3.00052967e-02 8.18652332e-01 1.47592887e-01
-9.89541113e-02 -3.26124519e-01 -8.03084448e-02 -5.59657753e-01
-6.18911743e-01 3.33538085e-01 1.31063968e-01 -1.10527791e-01
5.21520019e-01 -3.81505817e-01 -1.34127513e-01 -1.90117493e-01
-1.09985614e+00 8.23561326e-02 6.39707446e-01 9.86956298e-01
7.17710912e-01 1.17205299e-01 2.47265473e-01 -7.84253240e-01
8.15990806e-01 -1.93670347e-01 -1.00430608e+00 2.03098103e-01
-4.56913412e-01 2.30712965e-01 1.02042735e+00 -9.12819386e-01
-8.32591295e-01 9.97089967e-02 -5.32658219e-01 -4.28814650e-01
3.42667811e-02 6.18356109e-01 -1.85997132e-02 -8.06298703e-02
5.14142871e-01 3.15024465e-01 -2.77537674e-01 -6.78007245e-01
4.17152971e-01 9.68126178e-01 3.24614227e-01 -1.93214700e-01
5.96653402e-01 7.52322450e-02 -2.32594222e-01 -1.19787955e+00
-6.63117945e-01 -6.37879252e-01 -4.08894300e-01 5.19302040e-02
6.56888425e-01 -9.72240388e-01 -4.78229165e-01 6.27335966e-01
-1.25312316e+00 -6.69587195e-01 -3.02941859e-01 6.97211206e-01
-6.77899793e-02 4.13636982e-01 -7.82631040e-01 -7.11343884e-01
-8.50467026e-01 -1.21789908e+00 9.94178772e-01 -1.07147202e-01
-1.47146314e-01 -8.99389207e-01 -9.11051966e-03 1.11575253e-01
6.47547245e-01 -4.96200681e-01 9.54477966e-01 -1.03385758e+00
-1.63022041e-01 -4.59748149e-01 -3.06481957e-01 9.62012112e-01
-1.29338861e-01 -3.66780758e-01 -9.41799402e-01 -5.03486514e-01
4.42507207e-01 -3.54334235e-01 9.64684963e-01 5.19119501e-01
1.15611482e+00 -3.28588784e-01 6.40229210e-02 6.93442285e-01
1.54349566e+00 5.22911921e-03 4.73137081e-01 2.52110541e-01
7.32887268e-01 1.12757176e-01 2.64706343e-01 2.87202179e-01
-4.36718613e-02 6.67480707e-01 -1.06213344e-02 -1.96892977e-01
-2.20794931e-01 1.86556116e-01 4.75648642e-01 1.52601647e+00
3.26216489e-01 -4.53865111e-01 -8.24940562e-01 3.40273082e-01
-1.54040933e+00 -5.37298441e-01 -4.63897176e-02 2.16581845e+00
9.64477956e-01 2.70678699e-01 -4.22739685e-01 5.01821265e-02
4.66155946e-01 2.66460538e-01 -2.84197599e-01 -7.77730346e-01
-1.88445076e-01 5.73438466e-01 6.51463211e-01 9.30985332e-01
-1.07405031e+00 1.15178740e+00 5.64119864e+00 1.63874698e+00
-1.48843396e+00 4.69912291e-01 7.63298154e-01 -6.82758093e-01
2.81044900e-01 -2.78073013e-01 -1.17938697e+00 3.88854295e-01
1.31273019e+00 1.72773972e-01 6.96164429e-01 8.54860544e-01
3.49596858e-01 -2.66746223e-01 -7.22629786e-01 1.15601861e+00
1.60794929e-01 -1.06253600e+00 8.69275779e-02 -1.60006285e-01
5.65836191e-01 4.66161251e-01 -2.13463664e-01 6.45353734e-01
5.89689314e-02 -9.44209993e-01 6.28629208e-01 2.39556581e-01
6.48679078e-01 -8.33712757e-01 9.40248549e-01 6.27922773e-01
-9.83446062e-01 1.07600600e-01 -7.53561914e-01 -7.49951601e-02
1.86790720e-01 7.72775054e-01 -6.91835225e-01 -2.19870694e-02
5.62620699e-01 2.10250523e-02 -3.63273650e-01 6.22552752e-01
1.63531318e-01 1.03206766e+00 -8.15918028e-01 -2.65700370e-01
4.51258391e-01 -2.13876367e-01 4.73073095e-01 1.65512967e+00
2.46713996e-01 -1.53636009e-01 5.70945516e-02 3.17531347e-01
-2.88735688e-01 5.82581162e-01 -1.34395706e-02 2.54256222e-02
2.82448679e-02 1.24877608e+00 -6.21584356e-01 -7.56909907e-01
-5.67955792e-01 9.95397389e-01 5.77717185e-01 3.14435959e-01
-6.93378031e-01 -4.83523667e-01 4.11962003e-01 -1.99859977e-01
6.24313951e-01 -3.47112775e-01 -2.34398574e-01 -1.08767796e+00
1.29208535e-01 -1.08330929e+00 -1.03239520e-02 -5.20466506e-01
-7.96739340e-01 7.70350337e-01 -2.01464772e-01 -7.03188241e-01
-1.46689877e-01 -6.47214353e-01 -2.71116227e-01 1.20752430e+00
-1.44843221e+00 -8.48980427e-01 -8.73709768e-02 4.28097486e-01
8.53658438e-01 -2.11634696e-01 8.54795396e-01 6.94513977e-01
-6.67856157e-01 7.92418301e-01 5.89624882e-01 3.35286632e-02
6.69928789e-01 -9.19715226e-01 6.32298827e-01 9.59495604e-01
1.50719687e-01 6.35711730e-01 6.96699679e-01 -7.36849487e-01
-9.76365089e-01 -1.21736848e+00 1.19457185e+00 2.10785970e-01
6.95367873e-01 -8.82192850e-01 -1.01738691e+00 5.36915183e-01
2.14010671e-01 -4.43051904e-02 4.76171106e-01 -2.23700687e-01
-8.59900787e-02 -6.31769001e-02 -6.95546448e-01 7.64904618e-01
5.48750520e-01 -4.70684975e-01 -2.63455421e-01 4.26642656e-01
9.48302269e-01 -5.24585307e-01 -4.73074347e-01 1.49001673e-01
4.31101590e-01 -3.67531389e-01 8.41915369e-01 -5.99969923e-01
1.90191895e-01 -4.90420796e-02 -3.14296842e-01 -1.09856880e+00
-3.61596630e-03 -4.54615742e-01 -2.49528483e-01 1.06759441e+00
3.35176140e-01 -3.93467575e-01 5.48705637e-01 3.46540928e-01
1.17553659e-01 -7.98383474e-01 -1.09716427e+00 -6.20154917e-01
1.45143140e-02 -8.35578501e-01 2.80913323e-01 6.18227065e-01
-4.73569006e-01 4.69534278e-01 -7.26329207e-01 -7.83483014e-02
2.17473343e-01 -5.14816582e-01 7.29821563e-01 -8.36930752e-01
-5.60641050e-01 -2.39714533e-01 -2.33127281e-01 -1.29285002e+00
1.26976177e-01 -8.17742169e-01 2.94332743e-01 -1.06780195e+00
-5.99164553e-02 -1.67057231e-01 -1.43514693e-01 4.30347264e-01
-2.82751918e-01 2.94628710e-01 1.88652217e-01 -2.53551872e-03
-4.52325642e-01 5.19926906e-01 6.78974211e-01 -2.29407564e-01
-2.56879300e-01 -4.16240245e-02 -1.92545637e-01 7.27831960e-01
6.76513791e-01 -7.20993400e-01 -3.75033677e-01 -7.86190987e-01
2.76699483e-01 -3.10129791e-01 -2.62426212e-02 -1.06159008e+00
5.01118183e-01 3.50295693e-01 1.43773779e-01 -6.24817491e-01
3.95725250e-01 -9.08599854e-01 -8.48201737e-02 4.88543212e-01
-4.42896008e-01 1.03320070e-01 6.95580244e-01 3.31269681e-01
-1.73466593e-01 -6.76942348e-01 8.53781700e-01 4.63085882e-02
-4.20296103e-01 -5.69526479e-02 -6.25794053e-01 -1.34281263e-01
4.87995565e-01 2.86030956e-02 1.04207203e-01 -3.92294168e-01
-7.00896919e-01 7.77555513e-04 -6.98588863e-02 2.27117181e-01
2.81596243e-01 -9.50323939e-01 -7.49112129e-01 3.64070237e-01
-3.91125441e-01 1.71861574e-01 2.30095893e-01 8.09303522e-01
-6.32473469e-01 5.59410989e-01 3.46014351e-01 -4.04411882e-01
-1.63391924e+00 5.01271307e-01 4.36103702e-01 -8.77822280e-01
-3.78907144e-01 1.02084196e+00 1.25782549e-01 -4.97210801e-01
7.07825840e-01 -4.50750470e-01 -2.63123214e-01 -4.36505955e-03
6.40564561e-01 3.23683530e-01 5.93086839e-01 -6.28164947e-01
-6.12098388e-02 4.25341725e-01 -3.86170149e-01 -1.81407169e-01
1.40513730e+00 -1.53371960e-01 -2.68590808e-01 4.15598601e-01
1.45453107e+00 1.78331092e-01 -6.97978616e-01 -5.63542366e-01
4.67890091e-02 7.41137415e-02 5.37016451e-01 -5.51755369e-01
-1.06604517e+00 9.91016686e-01 7.53324628e-01 -2.82746047e-01
1.21545732e+00 -4.14295584e-01 8.77514601e-01 8.33096802e-01
-1.03705280e-01 -1.29393852e+00 -3.69561821e-01 9.24891353e-01
6.87102556e-01 -1.03178954e+00 1.04025282e-01 -4.07214612e-02
-2.42373630e-01 1.16035461e+00 3.66056293e-01 -1.87625915e-01
7.96852410e-01 6.13706529e-01 1.92413792e-01 9.84934121e-02
-7.86105692e-01 1.19469641e-02 3.18610400e-01 -1.58486977e-01
4.25479740e-01 -1.78901851e-01 -5.05191922e-01 5.19659996e-01
-1.61942050e-01 -4.05792333e-02 1.26182497e-01 6.68121696e-01
-3.82947952e-01 -1.15902269e+00 -3.26864064e-01 6.45989716e-01
-7.25887775e-01 -7.53419399e-01 -3.43012135e-03 5.45523643e-01
5.56336455e-02 7.45321691e-01 1.66370988e-01 -2.49662131e-01
1.48396492e-01 2.98957109e-01 4.91396375e-02 -6.94447398e-01
-9.52837110e-01 4.84368116e-01 3.55581380e-02 -1.95912451e-01
-2.19796579e-02 -2.33812764e-01 -1.14530611e+00 -2.13177800e-01
-6.94989383e-01 8.60887915e-02 1.15261292e+00 1.03035545e+00
6.29047155e-02 5.58169901e-01 8.67354646e-02 -8.66899192e-01
-8.97368729e-01 -1.34935367e+00 -3.58242512e-01 2.81804651e-01
3.51094127e-01 -6.12153858e-02 -4.17859346e-01 -3.98902968e-02]
|
[14.021909713745117, 6.506819248199463]
|
30c88ba2-1810-4510-a0b8-6a496bd48cdb
|
concept-based-explanations-to-test-for-false
|
2307.01900
| null |
https://arxiv.org/abs/2307.01900v1
|
https://arxiv.org/pdf/2307.01900v1.pdf
|
Concept-Based Explanations to Test for False Causal Relationships Learned by Abusive Language Classifiers
|
Classifiers tend to learn a false causal relationship between an over-represented concept and a label, which can result in over-reliance on the concept and compromised classification accuracy. It is imperative to have methods in place that can compare different models and identify over-reliances on specific concepts. We consider three well-known abusive language classifiers trained on large English datasets and focus on the concept of negative emotions, which is an important signal but should not be learned as a sufficient feature for the label of abuse. Motivated by the definition of global sufficiency, we first examine the unwanted dependencies learned by the classifiers by assessing their accuracy on a challenge set across all decision thresholds. Further, recognizing that a challenge set might not always be available, we introduce concept-based explanation metrics to assess the influence of the concept on the labels. These explanations allow us to compare classifiers regarding the degree of false global sufficiency they have learned between a concept and a label.
|
['Esma Balkir', 'Kathleen C. Fraser', 'Svetlana Kiritchenko', 'Isar Nejadgholi']
|
2023-07-04
| null | null | null | null |
['abusive-language']
|
['natural-language-processing']
|
[ 2.74235815e-01 2.30151281e-01 -3.89859825e-01 -8.70105326e-01
-3.74286324e-01 -7.13577986e-01 7.50404418e-01 8.13454866e-01
-4.73377377e-01 6.96826100e-01 1.67653501e-01 -5.53953052e-01
-1.49208769e-01 -6.15709484e-01 -5.09446859e-01 -5.70916414e-01
-4.18735482e-03 1.65178195e-01 -1.84028104e-01 4.36097346e-02
3.83217096e-01 4.62395430e-01 -1.22309053e+00 5.20436227e-01
7.49380648e-01 7.12156117e-01 -4.39834028e-01 8.34959373e-02
-5.17342500e-02 1.09876692e+00 -9.20988619e-01 -6.97556853e-01
-1.33261532e-01 -7.91631699e-01 -8.39957476e-01 1.01963416e-01
4.41983849e-01 -1.45914838e-01 3.34373862e-01 1.11482370e+00
-1.92970000e-02 -2.18447030e-01 1.03126287e+00 -1.43220055e+00
-3.64468634e-01 8.87513816e-01 -3.58220220e-01 5.77105641e-01
3.72416854e-01 9.11288634e-02 1.27061915e+00 -5.20270169e-01
6.96436763e-01 1.24495244e+00 6.66673303e-01 6.01205587e-01
-1.34240818e+00 -1.11605132e+00 2.94942319e-01 2.06956506e-01
-1.05459750e+00 -3.47250432e-01 5.89315891e-01 -6.67578220e-01
7.03567505e-01 2.49554336e-01 4.50878114e-01 1.43440747e+00
1.86564311e-01 4.54607457e-02 1.49196136e+00 -4.61762518e-01
2.93093532e-01 5.66769838e-01 5.18553376e-01 6.01914823e-01
6.80329025e-01 2.14775041e-01 -7.29478538e-01 -4.11469221e-01
1.03765368e-01 -1.85951307e-01 -2.34126240e-01 -1.44124910e-01
-5.34808815e-01 1.20490658e+00 6.49387956e-01 6.86307669e-01
-3.18735123e-01 -1.34330258e-01 4.01407331e-01 2.56373376e-01
4.35229242e-01 8.28600109e-01 -5.16451180e-01 -7.33118802e-02
-8.60263705e-01 7.05552995e-02 7.68597901e-01 1.48947001e-01
5.69305599e-01 -2.52240956e-01 5.72456457e-02 4.93326247e-01
7.51044229e-02 1.69627964e-02 5.65068662e-01 -5.01399636e-01
-2.25681011e-02 6.89288199e-01 4.39230017e-02 -1.04054368e+00
-4.52976942e-01 -5.78565657e-01 -2.69982100e-01 1.15703695e-01
4.54115361e-01 -5.94268106e-02 -5.68901539e-01 2.23481274e+00
1.09134734e-01 -1.95361562e-02 1.18648954e-01 8.36953282e-01
5.48126042e-01 1.34237418e-02 7.70463169e-01 -3.82855862e-01
1.11131084e+00 -2.62679935e-01 -6.01346731e-01 -5.78429520e-01
1.18051493e+00 -3.27528656e-01 9.14596736e-01 3.78856778e-01
-4.18411016e-01 -2.33590335e-01 -1.36055100e+00 1.09111562e-01
-5.18562257e-01 -3.58225405e-01 8.82367432e-01 8.60847890e-01
-3.68783087e-01 7.92433560e-01 -4.43771362e-01 -3.89805764e-01
3.22942227e-01 1.15225077e-01 -4.93231922e-01 -2.65567511e-01
-1.20568573e+00 1.44788468e+00 4.67235118e-01 -1.50286987e-01
-6.47858739e-01 -5.45867860e-01 -6.65827453e-01 2.47017190e-01
2.14689180e-01 -1.14820592e-01 8.29710066e-01 -1.54914725e+00
-4.10620600e-01 1.09723794e+00 1.48485392e-01 -3.50140214e-01
3.38765115e-01 -4.98951674e-02 -5.91218233e-01 -1.02018472e-02
4.36729044e-01 3.58194888e-01 7.68754423e-01 -1.38069904e+00
-5.58073640e-01 -5.93151748e-01 1.53769076e-01 -1.26034498e-01
-2.96644479e-01 7.97487348e-02 5.99494874e-01 -4.70073253e-01
1.09555654e-01 -8.41343284e-01 2.20529690e-01 -1.52835950e-01
-4.14468557e-01 -2.22971499e-01 7.17386186e-01 -4.85787481e-01
1.05948663e+00 -2.23682356e+00 -2.44809777e-01 2.99505383e-01
1.41856343e-01 1.16717346e-01 2.71694571e-01 1.87007144e-01
-5.00645995e-01 6.22594118e-01 -4.09032732e-01 -1.11108892e-01
-2.99140275e-01 4.47632611e-01 -5.37779868e-01 6.15862072e-01
6.30244792e-01 3.73942971e-01 -1.04175031e+00 -2.96421021e-01
-8.66238028e-02 3.68659407e-01 -4.14274246e-01 1.80294022e-01
5.97554967e-02 2.37225398e-01 -2.70153075e-01 3.57775718e-01
4.02611583e-01 1.01383127e-01 5.50707340e-01 -9.26923826e-02
3.34487885e-01 6.50135577e-01 -6.75644159e-01 7.93131292e-01
-2.60755688e-01 6.00342751e-01 -1.75906613e-01 -1.06260407e+00
9.92465854e-01 1.33913055e-01 1.48448348e-01 -5.65360725e-01
2.24822521e-01 4.58078325e-01 3.58451992e-01 -5.91205955e-01
-1.99963176e-03 -1.02656960e+00 3.90788540e-02 5.69206893e-01
4.34248261e-02 1.02738388e-01 -2.78411321e-02 2.70184487e-01
1.13770640e+00 -2.04621330e-01 3.93656909e-01 -4.17684376e-01
3.70177746e-01 -1.97380851e-03 8.16597879e-01 5.85490823e-01
-9.47889239e-02 2.01293305e-01 9.79421198e-01 -1.56241462e-01
-6.38835549e-01 -7.68655241e-01 -4.39779371e-01 1.01005638e+00
-1.35190234e-01 -5.06781340e-01 -3.12687188e-01 -1.12522674e+00
1.33596703e-01 1.48584461e+00 -1.10420883e+00 -6.66485488e-01
-4.66858223e-03 -9.40794706e-01 6.16380870e-01 4.85946923e-01
-2.95667513e-03 -7.42618620e-01 -1.00262415e+00 -3.06284502e-02
-1.13521479e-01 -1.12211919e+00 2.40321264e-01 7.28114069e-01
-7.08555460e-01 -1.51670790e+00 2.11103946e-01 -2.11487696e-01
7.87593782e-01 -6.08998761e-02 1.24222207e+00 7.72819936e-01
5.15326448e-02 2.22047761e-01 -5.37179828e-01 -5.27952015e-01
-7.71457255e-01 -9.51521397e-02 1.05011351e-01 3.26761752e-02
8.88985097e-01 -4.29518700e-01 -4.54898924e-02 6.89486042e-02
-9.53182340e-01 -3.55081886e-01 2.71618456e-01 6.93362415e-01
-9.81638059e-02 -1.11632595e-04 7.35295355e-01 -1.38482380e+00
7.24793553e-01 -9.64226663e-01 -3.73221748e-02 2.89822578e-01
-7.52705634e-01 1.82910815e-01 4.13754791e-01 -5.44712126e-01
-9.47860420e-01 -2.45312184e-01 4.11882885e-02 -1.03065312e-01
-1.78238645e-01 5.94514370e-01 1.30634685e-03 3.15183580e-01
9.45372164e-01 -3.79146129e-01 -7.14154541e-02 -2.87357330e-01
1.24104992e-01 6.60551429e-01 1.09584384e-01 -6.49337113e-01
4.89311218e-01 2.57216513e-01 5.71140507e-03 -5.54514766e-01
-1.46097648e+00 -3.63690436e-01 -6.29633784e-01 -3.04643977e-02
7.30158567e-01 -5.33469677e-01 -4.43763494e-01 -1.04552366e-01
-9.93286848e-01 -1.23766631e-01 -2.52106320e-03 5.25743425e-01
-2.75359809e-01 1.23873420e-01 -2.93762892e-01 -9.40243840e-01
9.30808336e-02 -9.03517067e-01 4.54172879e-01 -5.83411008e-02
-8.63842428e-01 -1.06326056e+00 -9.81944725e-02 2.73853153e-01
1.88678563e-01 4.15214688e-01 1.39370203e+00 -1.33232975e+00
3.47431511e-01 -3.32092255e-01 -1.19525820e-01 3.52648437e-01
2.51852363e-01 1.21721230e-01 -1.23678994e+00 -2.01822385e-01
4.30901021e-01 -7.98990369e-01 8.95843983e-01 -9.00790393e-02
7.29025424e-01 -3.51099670e-01 -2.98988163e-01 6.69824779e-02
1.33116281e+00 2.94684265e-02 2.52920181e-01 1.70534700e-01
3.33431035e-01 1.06971526e+00 5.36307395e-01 4.83205952e-02
-6.72609955e-02 5.50358593e-01 2.94929862e-01 1.19406201e-01
3.19969147e-01 -3.92172992e-01 4.66057092e-01 1.20406374e-01
4.38894808e-01 -3.33984345e-02 -8.73870850e-01 4.21533763e-01
-1.43386936e+00 -7.09437251e-01 -3.16666096e-01 2.15476775e+00
8.83380949e-01 4.51499850e-01 6.61615431e-02 4.04181659e-01
5.54201245e-01 -8.21443945e-02 -4.26063389e-01 -8.71157825e-01
-9.66304615e-02 9.95647609e-02 9.16295871e-02 6.40088558e-01
-7.80457675e-01 7.56675959e-01 6.68077564e+00 4.20861065e-01
-1.29970276e+00 1.27808720e-01 8.99495006e-01 -3.46069224e-02
-3.89817238e-01 3.98636721e-02 -4.22172517e-01 3.39171916e-01
1.01282597e+00 -1.11980028e-01 -1.25452191e-01 8.21439624e-01
-8.22759345e-02 -3.07767302e-01 -1.46019030e+00 4.06587005e-01
2.75447994e-01 -6.43923163e-01 -8.55697542e-02 7.51609132e-02
1.60153955e-01 -3.99573475e-01 -1.76800653e-01 3.28083098e-01
3.63379538e-01 -1.41361308e+00 7.11994350e-01 1.30457193e-01
5.62524915e-01 -6.43069208e-01 9.06948924e-01 4.46452886e-01
-3.07709008e-01 -2.91475117e-01 -2.86192328e-01 -4.95405644e-01
-2.58500397e-01 5.90961874e-01 -8.73327851e-01 -9.25809424e-03
5.86786509e-01 5.06643772e-01 -8.72696102e-01 4.63427663e-01
-7.57368624e-01 9.64925349e-01 -2.71771371e-01 1.29783630e-01
2.14639157e-01 1.93840474e-01 1.95063919e-01 1.21615863e+00
-7.91440830e-02 3.17259192e-01 -2.63486039e-02 1.03224564e+00
1.18453942e-01 7.00664446e-02 -7.93179452e-01 -2.50963420e-01
3.91795039e-01 1.27370095e+00 -8.49251747e-01 -2.98636705e-01
-3.02394211e-01 7.28978395e-01 5.51232755e-01 1.16233202e-02
-7.39143312e-01 3.14605355e-01 6.94948196e-01 1.32347330e-01
-1.16110049e-01 1.48963600e-01 -4.60348994e-01 -9.02403414e-01
-1.90094650e-01 -8.06878209e-01 7.14495361e-01 -7.64250040e-01
-1.53524768e+00 3.60429972e-01 -7.59339472e-03 -5.81130803e-01
-3.84123921e-01 -6.49075747e-01 -4.51017201e-01 7.25940526e-01
-1.27229536e+00 -9.71432805e-01 -9.91147459e-02 2.52148390e-01
1.50002884e-02 3.05910498e-01 9.98912752e-01 -6.86682835e-02
-3.96911353e-01 4.45993274e-01 -7.15294421e-01 2.53052711e-01
7.92961180e-01 -1.09267092e+00 -2.64236987e-01 7.36073196e-01
3.10485393e-01 7.34574437e-01 9.63019907e-01 -7.46996403e-01
-4.81696427e-01 -6.23483777e-01 1.11734891e+00 -8.18540215e-01
6.24987960e-01 -1.34529218e-01 -1.27595353e+00 8.60859573e-01
-2.39642531e-01 -2.43576840e-01 1.12325811e+00 5.90772688e-01
-1.05928206e+00 2.16899768e-01 -1.36764073e+00 3.09493482e-01
9.26497638e-01 -5.94557285e-01 -7.89873064e-01 1.41951442e-01
2.52058089e-01 1.63557336e-01 -4.37950701e-01 2.82835454e-01
3.56145531e-01 -1.22993970e+00 4.84481931e-01 -1.21767008e+00
5.85117757e-01 1.84346929e-01 -1.95367545e-01 -1.32221234e+00
-2.64306158e-01 2.29258373e-01 4.05545294e-01 1.34940219e+00
6.16959453e-01 -4.94298905e-01 3.59588772e-01 1.10243368e+00
3.75365734e-01 -6.08515203e-01 -9.59411085e-01 -5.55911601e-01
3.69985312e-01 -6.43675983e-01 3.25548381e-01 1.63006747e+00
3.55370700e-01 7.08788574e-01 -1.12970471e-01 1.53599352e-01
3.89096260e-01 -1.51059642e-01 1.54777691e-01 -1.36542118e+00
-2.07419619e-02 -3.17573875e-01 -4.96594310e-01 1.23103864e-01
5.06427109e-01 -9.46448505e-01 -1.83022365e-01 -9.37835932e-01
4.66940790e-01 -5.76553345e-01 -3.66759717e-01 7.39974916e-01
-3.18232328e-01 2.75685817e-01 3.44788611e-01 8.31665769e-02
-1.22249000e-01 1.96187347e-01 4.68415767e-01 4.18409556e-02
1.18390694e-01 -3.32233310e-01 -1.16367841e+00 9.71101761e-01
7.28642762e-01 -9.00003850e-01 -4.22002047e-01 -1.33424431e-01
4.80556577e-01 -1.70826063e-01 5.10708630e-01 -8.23880434e-01
-2.28220910e-01 -2.79615462e-01 4.60767955e-01 3.58644754e-01
3.50765914e-01 -9.03813899e-01 1.88774634e-02 7.28443205e-01
-8.31868649e-01 -3.12301945e-02 1.89834118e-01 3.24454457e-01
-1.94467157e-02 -7.25019813e-01 8.11901629e-01 -1.23128988e-01
-4.96784687e-01 -3.99093866e-01 -2.38826245e-01 1.07603095e-01
9.27351952e-01 -1.56212905e-02 -2.81171620e-01 -3.33259106e-01
-7.87354887e-01 -1.75910950e-01 4.92974520e-01 6.36926532e-01
4.27201957e-01 -9.73384500e-01 -7.26523161e-01 8.73786137e-02
3.05638850e-01 -8.06571007e-01 -2.09940016e-01 6.30424500e-01
-3.36332731e-02 2.27493584e-01 -2.16761619e-01 -1.55640498e-01
-1.37447858e+00 7.22403049e-01 5.93985736e-01 -1.52148664e-01
-2.01462194e-01 8.12091231e-01 2.10030109e-01 -9.70560238e-02
-1.89802751e-01 -3.26998420e-02 -2.95555085e-01 4.07399476e-01
3.67388278e-01 -5.42184338e-02 1.15466490e-01 -8.23168218e-01
-6.78489745e-01 3.55821177e-02 -1.00988336e-01 -5.42038120e-03
1.22639453e+00 2.07061648e-01 -7.94931799e-02 7.96809614e-01
9.32354391e-01 1.37698069e-01 -6.98547661e-01 -3.19927633e-02
3.55251163e-01 -6.38843358e-01 7.70260170e-02 -1.19924915e+00
-8.05985689e-01 8.95836294e-01 5.38213015e-01 2.31879830e-01
6.87320769e-01 -6.36361260e-03 -3.41491820e-03 -3.69618693e-03
1.93715185e-01 -8.87495041e-01 1.03770912e-01 1.23058960e-01
8.64455402e-01 -1.24482167e+00 1.49105698e-01 -6.67521834e-01
-7.35688686e-01 1.03950322e+00 6.75702810e-01 5.52813187e-02
4.25594330e-01 5.58485724e-02 2.32636139e-01 -4.06677395e-01
-7.92310536e-01 -1.27595529e-01 6.34272248e-02 5.28750598e-01
7.91881263e-01 2.49187991e-01 -7.57229388e-01 8.72126222e-01
-3.43771428e-01 -1.62186190e-01 6.53509736e-01 6.83039308e-01
-2.11624503e-01 -8.91245961e-01 -2.38479882e-01 7.27622747e-01
-8.14345002e-01 3.37105175e-03 -1.12984157e+00 8.05676162e-01
4.42444563e-01 1.06627572e+00 2.48497292e-01 -4.77823764e-01
-3.34630683e-02 5.58171093e-01 3.53095919e-01 -8.94696832e-01
-9.10672963e-01 -5.81571937e-01 5.37763417e-01 -4.30101693e-01
-4.70017016e-01 -6.18357420e-01 -1.08493125e+00 -2.86807001e-01
-4.83959347e-01 2.50996768e-01 4.96583879e-01 1.38126445e+00
6.11409657e-02 7.16794357e-02 2.14549452e-01 -2.75331922e-03
-6.39565706e-01 -1.05030763e+00 -5.87437749e-01 1.08704245e+00
2.53726363e-01 -9.63265657e-01 -8.21167171e-01 -3.37134063e-01]
|
[8.739704132080078, 5.422821998596191]
|
76f1ef95-87ce-4fe0-a84c-05ae111f8e17
|
rethinking-self-attention-an-interpretable
|
1911.03875
| null |
https://arxiv.org/abs/1911.03875v3
|
https://arxiv.org/pdf/1911.03875v3.pdf
|
Rethinking Self-Attention: Towards Interpretability in Neural Parsing
|
Attention mechanisms have improved the performance of NLP tasks while allowing models to remain explainable. Self-attention is currently widely used, however interpretability is difficult due to the numerous attention distributions. Recent work has shown that model representations can benefit from label-specific information, while facilitating interpretation of predictions. We introduce the Label Attention Layer: a new form of self-attention where attention heads represent labels. We test our novel layer by running constituency and dependency parsing experiments and show our new model obtains new state-of-the-art results for both tasks on both the Penn Treebank (PTB) and Chinese Treebank. Additionally, our model requires fewer self-attention layers compared to existing work. Finally, we find that the Label Attention heads learn relations between syntactic categories and show pathways to analyze errors.
|
['Quan Tran', 'Franck Dernoncourt', 'Walter Chang', 'Trung Bui', 'Khalil Mrini', 'Ndapa Nakashole']
|
2019-11-10
| null |
https://aclanthology.org/2020.findings-emnlp.65
|
https://aclanthology.org/2020.findings-emnlp.65.pdf
|
findings-of-the-association-for-computational
|
['constituency-parsing']
|
['natural-language-processing']
|
[ 6.24397323e-02 9.32606697e-01 -2.32009038e-01 -9.42456007e-01
-1.09703422e+00 -5.04788101e-01 2.57963359e-01 2.03922838e-01
-2.90781587e-01 8.01978350e-01 6.83769405e-01 -6.02976024e-01
4.04946834e-01 -5.60634077e-01 -8.04382443e-01 -1.48935392e-01
1.90151438e-01 7.38579988e-01 -2.27970108e-02 -2.00008959e-01
2.96732802e-02 1.93047628e-01 -7.32425034e-01 7.08164930e-01
8.34735036e-01 6.40447259e-01 2.29101583e-01 5.82065880e-01
-2.56552845e-01 9.57468212e-01 -6.72113359e-01 -7.54447520e-01
-2.22153902e-01 -3.47341537e-01 -1.34449399e+00 -4.41883564e-01
5.24909794e-01 -1.59373075e-01 -6.72133490e-02 9.17337060e-01
1.99315965e-01 9.26604320e-04 5.00317335e-01 -7.68535852e-01
-1.50328863e+00 1.02422810e+00 -3.87395203e-01 6.22371078e-01
5.84452450e-02 1.12651616e-01 1.84044063e+00 -8.56067657e-01
6.96138799e-01 1.65245116e+00 7.98173964e-01 1.03412819e+00
-1.57527995e+00 -7.61747718e-01 6.15262389e-01 2.43033946e-01
-5.16065896e-01 -2.99127400e-01 3.04493845e-01 -3.09424430e-01
1.72483253e+00 -1.06023028e-01 3.77528131e-01 1.26982498e+00
3.16439986e-01 8.47356498e-01 9.27281320e-01 -6.02048993e-01
-7.39398673e-02 -2.58527011e-01 9.94279265e-01 7.37870276e-01
2.38224402e-01 -3.42337042e-02 -3.87902647e-01 1.29939541e-01
3.90511811e-01 -2.55006939e-01 -2.30128348e-01 3.47999513e-01
-9.90555823e-01 9.41174626e-01 7.54738092e-01 2.78250605e-01
-2.22207189e-01 5.46217203e-01 3.51060510e-01 -2.70322077e-02
8.70057464e-01 9.03234363e-01 -1.14413583e+00 -6.17243461e-02
-3.85612071e-01 1.20081782e-01 7.00491250e-01 9.44585145e-01
6.48251534e-01 2.54865140e-02 -2.55124390e-01 9.16017711e-01
4.48631525e-01 2.42384508e-01 4.37063485e-01 -9.87763584e-01
8.91606629e-01 3.68691206e-01 1.86703820e-02 -4.94734555e-01
-8.00801277e-01 -3.77097040e-01 -2.94346213e-01 -5.33900224e-02
5.09306431e-01 -4.63450938e-01 -8.73427570e-01 2.20467401e+00
-1.14615858e-01 -8.68334249e-02 1.32945120e-01 6.56019509e-01
8.02121222e-01 6.40921354e-01 1.02364981e+00 1.07275113e-01
1.59563220e+00 -1.16182292e+00 -9.70689297e-01 -8.30431998e-01
1.21119940e+00 -3.78143311e-01 1.19284260e+00 -8.60893950e-02
-1.38701344e+00 -5.80751657e-01 -7.48938739e-01 -7.60624409e-01
-1.72812507e-01 -1.87631369e-01 8.62548709e-01 3.43697995e-01
-1.13767040e+00 7.19680488e-01 -1.02653027e+00 -3.89577091e-01
8.03987086e-01 5.04758060e-01 -4.03405309e-01 8.39225650e-02
-1.31125391e+00 1.41850734e+00 2.77432293e-01 -8.93870369e-02
-5.63164532e-01 -9.65703011e-01 -1.21795189e+00 5.58163702e-01
-2.05031320e-01 -5.83571911e-01 1.69859827e+00 -8.69615078e-01
-1.27789176e+00 9.84055400e-01 -5.48736215e-01 -6.21479034e-01
-2.99938947e-01 -5.35036862e-01 -1.87819585e-01 -1.50014907e-01
2.25620523e-01 1.06420112e+00 -4.75998931e-02 -9.77431774e-01
-6.14987910e-01 -2.28615716e-01 1.65389046e-01 -4.76937741e-03
-1.28463253e-01 3.31267267e-01 2.19246984e-01 -4.57256287e-01
1.07305266e-01 -8.35378170e-01 -2.29257360e-01 -4.66201782e-01
-4.00908530e-01 -6.91979825e-01 5.38416564e-01 -1.08368683e+00
1.13186967e+00 -1.89993477e+00 1.11753896e-01 -5.79391539e-01
1.06054999e-01 1.73331261e-01 -2.58245677e-01 2.27646418e-02
-5.30956745e-01 9.34574068e-01 -4.17585552e-01 -8.61675918e-01
1.59244046e-01 4.90158349e-01 -4.70221817e-01 1.52827188e-01
1.00669432e+00 1.21856666e+00 -9.27779317e-01 -4.06875193e-01
-1.14023551e-01 1.35887116e-01 -9.22593653e-01 2.83216923e-01
-5.37647545e-01 5.61438978e-01 -2.44440421e-01 5.82984149e-01
4.51116592e-01 -3.93322855e-01 3.19293588e-01 -2.64728293e-02
-4.47005555e-02 1.25892317e+00 -1.99061483e-01 1.66317034e+00
-7.27977514e-01 8.13816547e-01 -5.05438261e-02 -9.93959606e-01
6.30643785e-01 4.31776643e-01 -2.53108501e-01 -4.87253189e-01
9.00540650e-02 3.12101617e-02 7.09405184e-01 -3.68268907e-01
2.66420901e-01 -4.62793738e-01 -1.08163752e-01 6.41241372e-01
4.49245721e-01 1.33297682e-01 4.81232367e-02 1.36609346e-01
1.20845282e+00 3.23414505e-01 4.51124609e-01 -6.40904427e-01
6.98831230e-02 7.24230781e-02 8.83641541e-01 5.55569232e-01
-2.88178205e-01 5.01778126e-01 7.90223956e-01 -5.90915442e-01
-9.79475319e-01 -8.06655884e-01 -4.11842763e-01 1.50662172e+00
-5.42734206e-01 -2.60976851e-01 -7.71202624e-01 -1.12190270e+00
-2.26646848e-02 1.31009758e+00 -9.64018166e-01 -9.18777362e-02
-9.59918797e-01 -7.03128815e-01 4.97966647e-01 1.13832903e+00
1.26848459e-01 -1.59226286e+00 -2.18024597e-01 3.75061423e-01
-2.35176623e-01 -1.08813024e+00 -2.00578332e-01 6.07403874e-01
-9.50619638e-01 -8.20498228e-01 -1.40250310e-01 -1.01033258e+00
6.35455012e-01 -2.95831174e-01 1.65433824e+00 3.41402352e-01
9.90533084e-02 -4.03320789e-02 -3.05953205e-01 -7.20015585e-01
-5.82287192e-01 5.09359181e-01 -3.68345112e-01 -5.83277225e-01
6.85543299e-01 -3.75971109e-01 -1.35143414e-01 -1.37662753e-01
-2.87467152e-01 8.67766887e-02 4.09194946e-01 1.10149145e+00
3.92574579e-01 -9.04966652e-01 9.59980607e-01 -1.49413049e+00
3.42787862e-01 -8.41569781e-01 -3.21452379e-01 1.23501971e-01
-4.34331417e-01 3.91358465e-01 4.86844778e-01 6.67257980e-02
-1.47184503e+00 -1.11539856e-01 -5.18071830e-01 2.71595716e-01
-4.34748828e-01 3.57551068e-01 -3.24381948e-01 3.62709194e-01
4.58683640e-01 -5.75262606e-01 -3.11568081e-01 -7.00320959e-01
4.21109021e-01 3.47313195e-01 2.94823945e-01 -6.79258168e-01
2.09093034e-01 1.71455413e-01 -2.70972937e-01 -3.46358657e-01
-1.81180882e+00 -9.69796702e-02 -8.79588127e-01 4.58098292e-01
1.42047322e+00 -8.10848415e-01 -6.97658956e-01 -1.93462208e-01
-1.81877422e+00 -6.81938231e-01 -2.09125400e-01 2.98126310e-01
-3.99383545e-01 1.01115592e-01 -1.29685175e+00 -5.40197849e-01
-2.73215294e-01 -1.05085301e+00 1.04450583e+00 4.34452258e-02
-7.51523256e-01 -1.47630942e+00 8.68885890e-02 2.18729645e-01
1.27652392e-01 -6.70470223e-02 1.44524753e+00 -1.07172835e+00
-3.26270819e-01 1.57162204e-01 -4.57151860e-01 2.10137963e-01
-1.14385486e-01 -3.20200354e-01 -1.27429950e+00 1.55198857e-01
-1.61306486e-01 -3.57156038e-01 1.11236989e+00 5.15205920e-01
1.21419752e+00 -2.43978158e-01 -4.94161934e-01 5.94172835e-01
1.02205920e+00 1.66565888e-02 4.36607540e-01 2.76105076e-01
7.44090557e-01 8.23728859e-01 4.17891890e-01 -1.15261123e-01
6.21966362e-01 3.64451677e-01 4.36722279e-01 -2.08533674e-01
-3.16986799e-01 -1.21564008e-01 3.12925339e-01 6.98363125e-01
1.28218055e-01 -4.13582176e-01 -1.14631021e+00 7.86434591e-01
-1.80655372e+00 -6.58980012e-01 -3.53860319e-01 1.51648080e+00
1.08592141e+00 1.76905721e-01 -4.73281622e-01 -3.30443501e-01
6.78479314e-01 3.46421190e-02 -4.64205027e-01 -1.05277896e+00
-5.17193712e-02 7.31405735e-01 3.36851567e-01 8.82221043e-01
-1.11835337e+00 1.55234814e+00 6.93213797e+00 1.84207574e-01
-6.15897417e-01 5.65720856e-01 9.47936475e-01 1.61708176e-01
-3.89285594e-01 1.37579396e-01 -1.21564782e+00 1.81104526e-01
1.47020149e+00 1.73856199e-01 -4.75210994e-02 9.49667513e-01
-1.64383143e-01 1.91102102e-01 -1.60273194e+00 1.03581361e-01
-8.17884877e-02 -1.28921354e+00 -1.22699469e-01 1.69168413e-01
7.34207034e-01 3.43467325e-01 -1.54220343e-01 7.13482320e-01
1.05278695e+00 -1.21860063e+00 7.25331485e-01 1.30068332e-01
6.89218700e-01 -4.70006645e-01 9.52493370e-01 1.75975844e-01
-8.29208016e-01 -2.10667148e-01 -6.43345773e-01 -5.16618490e-01
4.64338005e-01 2.49078110e-01 -9.12332714e-01 -8.88696462e-02
4.91915703e-01 8.76058102e-01 -4.43656206e-01 4.16960269e-01
-1.03194273e+00 1.07626021e+00 6.22295253e-02 -7.06321327e-03
3.34107012e-01 3.57480228e-01 1.76389709e-01 1.54586494e+00
-2.42157150e-02 3.45131814e-01 1.10654812e-02 1.12970984e+00
-3.57671767e-01 1.73720606e-02 -3.29465449e-01 1.07087553e-01
4.49114293e-01 1.13522804e+00 -4.20774311e-01 -4.33469713e-01
-5.63455045e-01 7.04486489e-01 1.27680230e+00 2.62550950e-01
-8.32351208e-01 7.03402236e-02 1.01974380e+00 -1.17597088e-01
3.35301965e-01 -1.88229173e-01 -8.58894587e-01 -1.01054192e+00
-5.07611990e-01 -4.30665761e-01 4.86769080e-01 -8.30999911e-01
-1.34439075e+00 5.57858050e-01 -3.70471358e-01 -2.69959629e-01
-2.53667712e-01 -9.88118112e-01 -6.06478393e-01 1.26296365e+00
-1.73144555e+00 -1.23853004e+00 1.47909254e-01 -1.03161499e-01
9.62077796e-01 1.19134635e-01 1.38450468e+00 2.06871331e-02
-8.13613534e-01 6.87849641e-01 -2.84828454e-01 4.56200898e-01
7.65113771e-01 -1.68654215e+00 1.17096627e+00 7.06147850e-01
2.72871792e-01 8.52064848e-01 3.88110667e-01 -7.13827848e-01
-6.47847176e-01 -1.00892961e+00 1.51361644e+00 -9.75853741e-01
7.54101217e-01 -4.86945689e-01 -1.13723326e+00 1.54393721e+00
5.26834428e-01 2.18382761e-01 9.93729115e-01 9.19931233e-01
-5.96643925e-01 4.72169220e-01 -8.90055001e-01 3.02895725e-01
1.07967687e+00 -3.36778164e-01 -1.13160300e+00 3.74183774e-01
1.23921621e+00 -3.97348076e-01 -8.14625204e-01 1.56105533e-01
2.22680360e-01 -6.48808122e-01 6.65376186e-01 -1.34134769e+00
9.64073300e-01 2.60278285e-01 1.70842614e-02 -1.59581995e+00
-9.54203606e-01 -1.92120552e-01 -5.09026349e-02 1.53915608e+00
1.15116990e+00 -5.37780643e-01 6.00144088e-01 1.13666749e+00
-8.47299278e-01 -6.91458404e-01 -7.34228611e-01 -3.91555130e-01
7.26332426e-01 -4.18280184e-01 6.08566344e-01 1.08763623e+00
4.51205105e-01 7.55415916e-01 1.95025839e-02 1.96204156e-01
2.38549918e-01 -1.98042959e-01 9.03174952e-02 -1.40868735e+00
-3.99603337e-01 -3.25824469e-01 1.07334264e-01 -9.14285839e-01
9.81272399e-01 -1.31137109e+00 2.56333560e-01 -1.76034188e+00
2.18039438e-01 -5.33716083e-01 -3.68960917e-01 1.02600574e+00
-5.94086707e-01 3.85436900e-02 3.84833574e-01 4.34309281e-02
-5.26511252e-01 2.08058253e-01 1.06693494e+00 1.26762480e-01
2.86689103e-01 -4.30587381e-01 -9.78410900e-01 9.66162562e-01
9.17234480e-01 -7.64015496e-01 1.82265610e-01 -1.17102635e+00
1.71982989e-01 -1.98582232e-01 -1.98028460e-02 -5.98975599e-01
-3.33249390e-01 1.72798280e-02 4.28000331e-01 -1.88567430e-01
1.75228730e-01 -3.92210096e-01 -7.38821685e-01 5.14970183e-01
-8.29880595e-01 1.89916998e-01 4.53835517e-01 4.02998716e-01
-5.93848415e-02 -4.72823262e-01 8.54931056e-01 -4.69442815e-01
-5.91381311e-01 1.78089589e-02 -3.40304524e-01 4.71719652e-01
4.30784911e-01 3.46085966e-01 -6.25546157e-01 -7.37868100e-02
-1.02063179e+00 2.79380977e-01 7.31023178e-02 2.11967006e-01
-2.45053717e-03 -1.00982630e+00 -7.89197981e-01 -4.23573069e-02
-1.24875695e-01 1.62203953e-01 1.59743518e-01 4.70771968e-01
-4.13055182e-01 6.79309309e-01 -1.34408027e-01 -2.86721408e-01
-9.01902080e-01 5.22392333e-01 9.25262868e-02 -7.01402187e-01
-5.12148738e-01 1.33786201e+00 7.10909545e-01 -5.05314291e-01
-7.26122931e-02 -8.54305685e-01 -4.22149241e-01 -5.49527034e-02
5.38752913e-01 5.43200597e-02 -1.13506854e-01 -5.46660960e-01
-3.78076136e-01 2.51486570e-01 -2.73859888e-01 6.99384511e-02
1.53394675e+00 4.27074768e-02 -1.44507691e-01 4.55740243e-01
1.06248713e+00 -1.41671941e-01 -1.38734388e+00 2.05111485e-02
5.10486722e-01 9.31983069e-02 -1.68692499e-01 -1.05979013e+00
-8.16707373e-01 1.55013394e+00 -4.92339060e-02 1.11583687e-01
4.96482342e-01 3.50342274e-01 1.03093076e+00 2.65656561e-01
-6.96850792e-02 -7.82946885e-01 -1.28215536e-01 1.04296744e+00
8.00560713e-01 -1.42094433e+00 -4.10777450e-01 -5.97302854e-01
-6.29305482e-01 9.41671908e-01 1.13022614e+00 -2.54811138e-01
5.36923110e-01 5.03854752e-01 4.68977764e-02 -1.04980387e-01
-1.03767693e+00 -1.59998626e-01 1.34393379e-01 6.93247199e-01
1.26327240e+00 6.61469474e-02 -3.73606861e-01 1.29422414e+00
-3.09609979e-01 -6.55292630e-01 4.36789662e-01 5.50205350e-01
-4.91937280e-01 -1.26975513e+00 -7.77584240e-02 1.91401586e-01
-8.66859376e-01 -7.32185900e-01 -3.75046194e-01 8.89140844e-01
8.66338611e-02 8.18267703e-01 4.44353551e-01 4.13562618e-02
2.77390510e-01 7.33253121e-01 1.52714074e-01 -1.41986454e+00
-9.91533518e-01 -2.17240825e-01 5.23029566e-01 -4.74871218e-01
-1.97310418e-01 -7.66860664e-01 -1.69337785e+00 1.29474118e-01
-4.46257323e-01 1.69051096e-01 4.88541573e-01 1.05977023e+00
5.04937172e-01 8.96502554e-01 5.83206944e-04 -7.49552011e-01
-5.44922709e-01 -1.47481191e+00 -1.22386813e-01 5.32471061e-01
1.70912519e-01 -6.16622567e-01 -1.87590882e-01 1.65142670e-01]
|
[10.4799165725708, 9.377573013305664]
|
06ec0b5e-5bbd-4b6b-aca3-c851924d0f5d
|
does-long-term-series-forecasting-need
|
2306.05035
| null |
https://arxiv.org/abs/2306.05035v2
|
https://arxiv.org/pdf/2306.05035v2.pdf
|
Does Long-Term Series Forecasting Need Complex Attention and Extra Long Inputs?
|
As Transformer-based models have achieved impressive performance on various time series tasks, Long-Term Series Forecasting (LTSF) tasks have also received extensive attention in recent years. However, due to the inherent computational complexity and long sequences demanding of Transformer-based methods, its application on LTSF tasks still has two major issues that need to be further investigated: 1) Whether the sparse attention mechanism designed by these methods actually reduce the running time on real devices; 2) Whether these models need extra long input sequences to guarantee their performance? The answers given in this paper are negative. Therefore, to better copy with these two issues, we design a lightweight Period-Attention mechanism (Periodformer), which renovates the aggregation of long-term subseries via explicit periodicity and short-term subseries via built-in proximity. Meanwhile, a gating mechanism is embedded into Periodformer to regulate the influence of the attention module on the prediction results. Furthermore, to take full advantage of GPUs for fast hyperparameter optimization (e.g., finding the suitable input length), a Multi-GPU Asynchronous parallel algorithm based on Bayesian Optimization (MABO) is presented. MABO allocates a process to each GPU via a queue mechanism, and then creates multiple trials at a time for asynchronous parallel search, which greatly reduces the search time. Compared with the state-of-the-art methods, the prediction error of Periodformer reduced by 13% and 26% for multivariate and univariate forecasting, respectively. In addition, MABO reduces the average search time by 46% while finding better hyperparameters. As a conclusion, this paper indicates that LTSF may not need complex attention and extra long input sequences. The code has been open sourced on Github.
|
['Minggao Zhang', 'Dongyang Li', 'Xiaoyan Ma', 'Dongfeng Yuan', 'Haixia Zhang', 'Daojun Liang']
|
2023-06-08
| null | null | null | null |
['hyperparameter-optimization', 'bayesian-optimization']
|
['methodology', 'methodology']
|
[-6.22112602e-02 -4.58196431e-01 2.87359431e-02 -3.15954298e-01
-6.26142204e-01 -2.60666072e-01 3.31315935e-01 4.99660848e-03
-2.16145217e-01 5.08335233e-01 -8.05729628e-02 -4.17977870e-01
-1.89993754e-01 -8.01463068e-01 -5.83322406e-01 -1.10607827e+00
-7.96252936e-02 2.88712591e-01 3.23442489e-01 -1.24878094e-01
3.44076455e-01 3.06360513e-01 -1.78070271e+00 1.74611911e-01
9.86750960e-01 1.46842229e+00 5.99337697e-01 4.52488005e-01
-1.71027556e-01 3.68674725e-01 -5.55771232e-01 -4.20369580e-02
9.16567296e-02 -1.19989827e-01 -2.80033857e-01 -2.04264760e-01
2.31103916e-02 -2.77862132e-01 2.40760110e-02 9.00048673e-01
8.28756869e-01 2.67352521e-01 1.76776469e-01 -1.04360676e+00
-6.48505837e-02 4.90777373e-01 -5.45774579e-01 4.62512285e-01
6.11650534e-02 2.86303490e-01 8.70729446e-01 -1.07399213e+00
-2.06008889e-02 1.00173414e+00 8.25578272e-01 -2.04651859e-02
-1.04298651e+00 -8.45566094e-01 3.76933217e-01 3.19641918e-01
-1.33719230e+00 -4.17404473e-01 7.65417755e-01 -3.66103023e-01
1.37964261e+00 5.65861046e-01 7.76341200e-01 8.26144218e-01
4.31684643e-01 5.66417754e-01 8.98129404e-01 -2.09485859e-01
1.75906390e-01 -1.96341872e-01 4.16085601e-01 3.24588478e-01
-8.55500922e-02 2.39514306e-01 -6.47536039e-01 -4.53174680e-01
6.89505160e-01 6.04046136e-02 -3.88649613e-01 4.44149941e-01
-1.13414288e+00 8.07739258e-01 8.68891329e-02 2.85815388e-01
-5.87705135e-01 5.37608936e-02 6.83779538e-01 3.18436563e-04
6.76062047e-01 2.95275241e-01 -6.77051067e-01 -5.12942433e-01
-9.87278402e-01 1.89385176e-01 6.95095897e-01 7.20085740e-01
5.63310981e-01 2.11950719e-01 -2.63319194e-01 9.25517499e-01
2.41390824e-01 7.57462680e-01 6.72861993e-01 -5.81745863e-01
4.32277620e-01 3.61900777e-01 1.17978593e-02 -1.18985820e+00
-6.13380790e-01 -7.41012633e-01 -1.21023619e+00 -3.17664593e-01
1.69868723e-01 -1.11136073e-02 -4.20131207e-01 1.51488471e+00
4.99547988e-01 4.48525399e-01 -3.17779899e-01 1.00173044e+00
6.07763886e-01 1.20290446e+00 -1.28577709e-01 -6.34504378e-01
1.55905473e+00 -1.12833858e+00 -9.26633596e-01 -1.19145051e-01
4.89790529e-01 -1.08758521e+00 1.18743980e+00 4.01684225e-01
-7.84019113e-01 -7.47228861e-01 -8.94040406e-01 1.32919580e-01
-1.11227028e-01 2.76670218e-01 7.07307339e-01 5.14584422e-01
-7.62611568e-01 6.35549545e-01 -1.07255650e+00 -7.29304105e-02
-4.16182056e-02 3.83982569e-01 4.39180493e-01 3.67837965e-01
-1.36016905e+00 4.34075505e-01 1.42995432e-01 5.40887475e-01
-4.38139051e-01 -8.49866927e-01 -4.11425918e-01 4.16877180e-01
2.85813868e-01 -5.89973152e-01 1.16550791e+00 -8.36730778e-01
-1.89740658e+00 2.57400483e-01 -5.41512311e-01 -4.44359750e-01
1.77422151e-01 -3.05590093e-01 -4.37274694e-01 -2.33355716e-01
-1.41069621e-01 1.81782678e-01 1.01981044e+00 -4.55536991e-01
-5.83285332e-01 -4.22360003e-01 -3.64380300e-01 1.12643518e-01
-5.02142906e-01 1.54375285e-01 -6.28308594e-01 -1.01905942e+00
9.37319398e-02 -1.10209513e+00 -3.53953749e-01 -5.20554125e-01
-5.00765182e-02 -5.11789441e-01 7.54852712e-01 -6.35637283e-01
2.06264234e+00 -2.28706121e+00 -1.92035988e-01 2.68595338e-01
-1.05580233e-01 2.70720363e-01 9.97451916e-02 3.18777412e-01
-1.40152425e-02 -1.40133128e-01 -7.60209560e-02 -1.72022253e-01
-1.30193621e-01 1.28190294e-01 -7.88195014e-01 2.56368965e-01
-1.19441114e-01 6.95177674e-01 -6.16892815e-01 -2.29395285e-01
1.73196450e-01 4.93949771e-01 -4.16216254e-01 2.42906705e-01
-1.92968979e-01 2.87844151e-01 -5.10215998e-01 3.66928905e-01
8.06624889e-01 -5.99987984e-01 1.62240133e-01 -1.91801220e-01
-4.95685250e-01 5.40167511e-01 -1.16281867e+00 1.30403590e+00
-6.31096721e-01 3.59465390e-01 -1.21333033e-01 -9.26580369e-01
1.15137839e+00 4.20953721e-01 5.79298019e-01 -8.48399699e-01
-1.69547889e-02 4.86895233e-01 -7.51023516e-02 -4.21602547e-01
4.88303691e-01 1.92293778e-01 1.31592274e-01 4.11709726e-01
-5.18223822e-01 1.23135760e-01 1.04608215e-01 -1.84545487e-01
7.68819690e-01 1.18436262e-01 6.05047531e-02 -5.04660308e-01
6.98419750e-01 -8.69926959e-02 8.48373652e-01 6.07373118e-01
6.48733824e-02 3.72272402e-01 2.51645029e-01 -8.96476746e-01
-8.27521324e-01 -4.18553472e-01 -2.81682163e-01 1.24902833e+00
-1.12032391e-01 -7.37352908e-01 -4.86367673e-01 -1.11999288e-01
-1.05897963e-01 4.95978802e-01 -3.34057927e-01 1.82261124e-01
-8.41280460e-01 -1.20244122e+00 2.06679761e-01 4.59391683e-01
4.52385604e-01 -1.04763746e+00 -9.34726775e-01 6.87611043e-01
-3.17770332e-01 -8.85622501e-01 -7.53225327e-01 2.51298696e-01
-1.17102718e+00 -6.02208674e-01 -6.18949115e-01 -4.60097998e-01
2.91951150e-01 3.62059593e-01 1.08617115e+00 -7.72837698e-02
1.49685755e-01 -2.50922948e-01 -3.18832397e-01 -3.38542759e-01
2.33445853e-01 4.76422429e-01 -1.62297234e-01 9.43581462e-02
2.92640507e-01 -7.04385400e-01 -7.10848927e-01 5.92572093e-01
-5.14775753e-01 3.41416001e-01 3.54990453e-01 1.03667367e+00
8.16137433e-01 -6.84034526e-02 5.37266433e-01 -5.72350919e-01
4.56268817e-01 -4.38188463e-01 -1.12839651e+00 2.18547747e-01
-8.29118729e-01 4.32652608e-03 1.01626146e+00 -5.97285032e-01
-9.16203737e-01 -3.97455506e-02 -3.04256052e-01 -5.32213092e-01
3.48287672e-01 7.87721455e-01 1.43352598e-01 2.93579519e-01
3.36201876e-01 4.08093601e-01 -1.11327861e-02 -5.28884172e-01
-3.04134071e-01 6.85443103e-01 4.68319319e-02 -5.51079750e-01
6.74694777e-02 3.07895362e-01 -1.20626315e-01 -8.12449634e-01
-8.03874612e-01 -4.38390434e-01 3.48070301e-02 -2.48465821e-01
6.61803603e-01 -9.12775338e-01 -1.05510688e+00 6.61812544e-01
-1.36032891e+00 -4.39463258e-01 1.33302540e-01 5.25470316e-01
-1.97158128e-01 1.80530772e-01 -6.44668937e-01 -9.03611898e-01
-1.00244069e+00 -1.31152916e+00 1.31495249e+00 4.05518897e-02
-3.03045988e-01 -7.69904137e-01 -7.75023326e-02 2.08715126e-01
7.17809677e-01 -1.59109920e-01 8.01895261e-01 -3.18074614e-01
-6.50189459e-01 -2.01473781e-03 -1.73816621e-01 1.63029023e-02
-2.99190938e-01 -3.76957357e-02 -9.77648914e-01 -4.10554826e-01
3.82269472e-01 1.50676012e-01 5.49588501e-01 7.15551138e-01
1.53921986e+00 -3.54540944e-01 -3.19306761e-01 8.16965818e-01
1.18181241e+00 5.02547204e-01 5.78630149e-01 3.81234258e-01
5.17201006e-01 3.62124711e-01 7.66943276e-01 8.35475981e-01
3.28287303e-01 9.46756363e-01 9.53384861e-02 1.07878543e-01
2.12321594e-01 1.03238493e-01 5.18995166e-01 1.55319154e+00
-8.23114440e-02 -2.55957246e-01 -9.73922670e-01 1.54757053e-01
-2.03948998e+00 -8.31970215e-01 -4.76446927e-01 2.41396427e+00
7.73803711e-01 4.21290994e-02 -6.16765693e-02 1.35828197e-01
5.53062439e-01 4.65962589e-01 -6.01149678e-01 -4.80919302e-01
-4.04480733e-02 1.26317322e-01 3.65161508e-01 3.17114562e-01
-9.72529948e-01 5.51306784e-01 5.35487127e+00 1.28994453e+00
-1.67800915e+00 2.26076439e-01 8.43508184e-01 -1.00185789e-01
-2.02082634e-01 5.27997725e-02 -1.15605843e+00 8.99869680e-01
1.18564093e+00 -1.65615171e-01 5.94981730e-01 8.73432279e-01
6.30749583e-01 -9.29520503e-02 -6.77666485e-01 1.30057669e+00
-2.15311199e-01 -1.21412241e+00 -3.13490868e-01 -5.93803776e-03
5.85986257e-01 2.29181275e-01 -4.16420884e-02 3.06680053e-01
-4.35782194e-01 -7.17675269e-01 6.42292023e-01 5.02898216e-01
5.36563039e-01 -7.58623123e-01 9.12768900e-01 5.05669653e-01
-1.53369212e+00 -3.10415402e-02 -3.59440297e-01 -2.59517759e-01
2.89432526e-01 1.33870435e+00 -3.53248000e-01 6.50450945e-01
9.59724903e-01 6.09329581e-01 -2.83523291e-01 8.50018740e-01
3.72506380e-02 9.65997756e-01 -8.06442201e-01 -2.99967796e-01
1.33477613e-01 -4.54468578e-01 4.97165859e-01 1.03593349e+00
6.91969335e-01 9.82517526e-02 2.74356365e-01 4.15398061e-01
5.41266382e-01 2.39852354e-01 -8.17010254e-02 1.92498475e-01
4.84105170e-01 1.12383151e+00 -6.52848423e-01 -4.80906308e-01
-3.52027744e-01 5.39969087e-01 3.56525579e-03 3.46053004e-01
-1.11081350e+00 -3.39839548e-01 3.85498464e-01 1.49879411e-01
4.07237053e-01 -2.45340139e-01 -6.63183033e-01 -1.18791974e+00
2.60282069e-01 -1.00744689e+00 3.69430512e-01 -6.44490004e-01
-1.10633266e+00 8.99050057e-01 -3.70159149e-01 -1.25709379e+00
-1.88958526e-01 -7.34741315e-02 -4.98545825e-01 1.07187951e+00
-1.27958071e+00 -6.11665726e-01 -4.28459436e-01 3.62604111e-01
7.74080873e-01 5.27832359e-02 8.54275584e-01 5.93736947e-01
-8.89034986e-01 4.88539010e-01 3.03843528e-01 -5.12139797e-01
5.18767595e-01 -5.39201140e-01 4.71594155e-01 7.34409750e-01
-1.96503982e-01 6.71140850e-01 5.74899614e-01 -5.95991373e-01
-1.50544691e+00 -9.74212885e-01 1.19772947e+00 5.64587191e-02
6.34731650e-01 -4.01790172e-01 -1.27064490e+00 1.36755243e-01
1.14533149e-01 4.78175469e-02 5.00932693e-01 3.16671073e-01
-9.93833318e-03 -5.24049401e-01 -3.66545767e-01 3.77257884e-01
7.07763493e-01 -3.00294071e-01 -1.98030677e-02 3.98495436e-01
6.66433930e-01 -6.02667212e-01 -8.55825484e-01 4.95898664e-01
7.38115609e-01 -9.90692973e-01 7.44029403e-01 1.14010267e-01
2.05811530e-01 -3.64802063e-01 -6.97373003e-02 -9.34913456e-01
-3.07304561e-01 -9.90700662e-01 -3.34752351e-01 1.16081297e+00
2.79535204e-01 -1.08662868e+00 4.75294530e-01 3.11615735e-01
-3.18964720e-01 -1.36238217e+00 -1.04381752e+00 -8.65104377e-01
-3.85368913e-01 -7.04285324e-01 8.59886944e-01 7.34563410e-01
-3.08449298e-01 5.02056003e-01 -5.14390707e-01 2.06967786e-01
2.55305409e-01 6.19771123e-01 6.18438482e-01 -9.94379282e-01
-4.22612250e-01 -4.48806465e-01 2.60503262e-01 -1.40771592e+00
-1.18768819e-01 -6.28667414e-01 1.81782506e-02 -8.98153543e-01
1.12107373e-03 -7.95453846e-01 -2.43214697e-01 3.21591049e-01
-3.20911229e-01 -1.56579047e-01 1.18150502e-01 5.25438368e-01
-3.52090478e-01 8.16964447e-01 1.07238412e+00 2.29879841e-01
-2.95417070e-01 2.39582866e-01 -7.32805654e-02 5.46433806e-01
7.99991548e-01 -5.44279993e-01 -2.52947986e-01 -5.71464360e-01
3.97722006e-01 4.12930757e-01 1.14136711e-01 -7.86091149e-01
3.69488150e-01 1.43844159e-02 1.01768538e-01 -8.84928763e-01
1.90961301e-01 -6.76485062e-01 4.66159225e-01 5.73796511e-01
1.26165420e-01 5.89698136e-01 3.49756390e-01 3.96705657e-01
-4.68149692e-01 -8.05015862e-02 5.92633128e-01 2.59102136e-01
-4.00162637e-01 3.61701310e-01 -5.05115151e-01 -2.18972623e-01
7.08960533e-01 8.27489514e-03 -1.07414804e-01 -1.48327976e-01
-4.77043271e-01 3.38044614e-01 2.73120832e-02 2.34872326e-01
2.99712330e-01 -1.25452328e+00 -4.94819522e-01 4.48027015e-01
-3.33161980e-01 -2.06170101e-02 6.61385655e-01 1.27106738e+00
-4.04128462e-01 7.23030388e-01 2.29457498e-01 -8.64916980e-01
-1.20347214e+00 5.82126796e-01 -3.88881192e-02 -7.62019277e-01
-5.54418325e-01 7.59681702e-01 1.76437050e-01 -2.37917602e-01
2.30361894e-01 -3.87062252e-01 -6.99300040e-03 2.88480043e-01
4.91521686e-01 5.55326283e-01 4.15800154e-01 -1.70271724e-01
-4.69143152e-01 6.73391640e-01 1.40034333e-01 1.27950400e-01
1.37817180e+00 -1.36551648e-01 -3.06875944e-01 3.89829189e-01
1.12741029e+00 -5.52574582e-02 -1.13802505e+00 -6.55123517e-02
-3.37654762e-02 -4.41973358e-01 1.99936435e-01 -4.94127303e-01
-1.09300256e+00 9.84133005e-01 6.76505268e-01 3.58413637e-01
1.55878901e+00 -5.16712546e-01 1.25162148e+00 3.73251028e-02
3.72532547e-01 -1.06931293e+00 -2.34806925e-01 9.52889442e-01
1.00568259e+00 -1.01152563e+00 1.60711817e-02 -4.90606010e-01
-3.14462572e-01 1.27725470e+00 4.32709545e-01 5.47548160e-02
7.96554506e-01 4.26339865e-01 -4.71241213e-02 -1.74098797e-02
-1.11258733e+00 1.95419908e-01 3.16266000e-01 -1.45537108e-01
5.08956671e-01 3.33237462e-02 -6.64575219e-01 8.54650557e-01
-2.72956461e-01 5.13252467e-02 -1.44372150e-01 5.65100312e-01
-1.79534391e-01 -9.36836362e-01 -5.60307980e-01 5.38823783e-01
-4.91253495e-01 -3.78297001e-01 5.42966962e-01 3.04832280e-01
3.74245159e-02 8.17978144e-01 2.39975363e-01 -3.25157583e-01
2.67895728e-01 -4.51103970e-02 3.87924798e-02 -2.29486257e-01
-1.09475005e+00 7.12639034e-01 -3.93908359e-02 -6.60529613e-01
-1.41537294e-01 -7.10257947e-01 -9.69997704e-01 -3.89289975e-01
-5.59034944e-01 1.91032395e-01 6.06976032e-01 8.48230779e-01
8.48262906e-01 6.22867405e-01 6.91808105e-01 -9.30546820e-01
-6.08140051e-01 -9.17222738e-01 -2.20973060e-01 -6.33760765e-02
-6.44404441e-02 -6.04030192e-01 -3.76956224e-01 -2.27172658e-01]
|
[7.064192295074463, 2.9019718170166016]
|
16e17fb2-cfa2-44c9-9128-f97769d29023
|
pcb-randnet-rethinking-random-sampling-for
|
2209.13797
| null |
https://arxiv.org/abs/2209.13797v1
|
https://arxiv.org/pdf/2209.13797v1.pdf
|
PCB-RandNet: Rethinking Random Sampling for LIDAR Semantic Segmentation in Autonomous Driving Scene
|
Fast and efficient semantic segmentation of large-scale LiDAR point clouds is a fundamental problem in autonomous driving. To achieve this goal, the existing point-based methods mainly choose to adopt Random Sampling strategy to process large-scale point clouds. However, our quantative and qualitative studies have found that Random Sampling may be less suitable for the autonomous driving scenario, since the LiDAR points follow an uneven or even long-tailed distribution across the space, which prevents the model from capturing sufficient information from points in different distance ranges and reduces the model's learning capability. To alleviate this problem, we propose a new Polar Cylinder Balanced Random Sampling method that enables the downsampled point clouds to maintain a more balanced distribution and improve the segmentation performance under different spatial distributions. In addition, a sampling consistency loss is introduced to further improve the segmentation performance and reduce the model's variance under different sampling methods. Extensive experiments confirm that our approach produces excellent performance on both SemanticKITTI and SemanticPOSS benchmarks, achieving a 2.8% and 4.0% improvement, respectively.
|
['GuoQiang Xiao', 'Dehong He', 'Hang Jiang', 'XianFeng Han', 'Huixian Cheng']
|
2022-09-28
| null | null | null | null |
['lidar-semantic-segmentation']
|
['computer-vision']
|
[-1.50442541e-01 -3.21953923e-01 -3.38691324e-01 -5.39609730e-01
-5.58843613e-01 -3.03265035e-01 3.63975048e-01 8.82714912e-02
-5.23667037e-01 5.70761859e-01 -3.58371168e-01 -1.82587802e-01
-1.30536169e-01 -1.22671664e+00 -7.18756735e-01 -6.78778708e-01
4.86947566e-01 8.28593850e-01 9.11748230e-01 -1.02949515e-01
4.70607847e-01 6.02762938e-01 -1.87638175e+00 -4.54887986e-01
1.38920081e+00 9.85400796e-01 6.06731892e-01 -1.45316839e-01
-7.45428145e-01 5.55385016e-02 -4.77306962e-01 -8.77270326e-02
4.94597375e-01 1.30777612e-01 -3.71899098e-01 -7.46543407e-02
2.00717106e-01 -1.42569542e-01 -5.40190600e-02 1.30118155e+00
3.84371340e-01 2.17672631e-01 5.42599976e-01 -1.36296606e+00
-1.70417696e-01 1.54095754e-01 -8.49970698e-01 -3.26039605e-02
-2.31329456e-01 1.71358243e-01 6.54230058e-01 -6.36726201e-01
8.30776021e-02 1.31075966e+00 6.29106402e-01 1.78266943e-01
-8.45177650e-01 -1.10720229e+00 2.66666561e-01 2.31700554e-01
-1.63918650e+00 -7.20550027e-03 8.20195615e-01 -2.75056601e-01
3.41692865e-01 2.08305642e-01 7.86991179e-01 4.56420690e-01
1.91768408e-01 5.13431549e-01 1.13261390e+00 1.45273894e-01
2.94345319e-01 1.23767532e-01 1.41602270e-02 3.20486575e-01
5.91584444e-01 4.00975272e-02 -2.78582960e-01 1.32191395e-02
5.51984251e-01 2.88052976e-01 2.61073858e-02 -6.20908618e-01
-8.26458275e-01 9.38054919e-01 6.18303597e-01 1.18871813e-03
-2.38665715e-01 -1.52137741e-01 3.63553613e-01 -2.04431415e-01
4.27935183e-01 5.49283847e-02 -2.47977287e-01 -1.45391703e-01
-1.01225865e+00 4.96891648e-01 3.52136612e-01 1.21181464e+00
1.20716751e+00 -2.02691510e-01 1.25098929e-01 9.67953086e-01
3.22333902e-01 9.21463251e-01 4.39169168e-01 -1.05048132e+00
6.94356024e-01 7.94988930e-01 1.81233153e-01 -1.03848541e+00
-2.20111996e-01 -4.56396341e-01 -7.85468459e-01 2.78632998e-01
3.08843464e-01 2.15443492e-01 -9.67016637e-01 1.20015121e+00
4.24163491e-01 8.38504136e-02 -1.90668389e-01 1.04278827e+00
5.43809772e-01 6.77952170e-01 1.90516561e-01 -9.35019180e-02
1.18877649e+00 -7.04044402e-01 -4.77473080e-01 -4.75914031e-01
3.49344313e-01 -6.58727348e-01 1.23807526e+00 1.11764215e-01
-7.00566590e-01 -6.66899085e-01 -9.96354818e-01 -8.11043102e-03
-2.54148662e-01 -1.74579963e-01 5.16090274e-01 5.02066970e-01
-4.54493493e-01 4.00605679e-01 -8.80054891e-01 -1.53119013e-01
7.89920270e-01 -1.04619572e-02 8.58484432e-02 -3.71649384e-01
-1.03402078e+00 7.30545282e-01 4.41784650e-01 2.67566275e-02
-2.99688250e-01 -7.45884120e-01 -6.54887319e-01 -2.45215073e-02
3.88826221e-01 -4.14861113e-01 1.08829033e+00 -3.31941813e-01
-1.10659659e+00 5.94534874e-01 -4.89469141e-01 -4.88182247e-01
6.57483578e-01 -1.52026400e-01 -1.19206406e-01 -3.48254107e-02
5.78546941e-01 9.14236009e-01 4.88614321e-01 -1.41444647e+00
-8.63214791e-01 -6.26763582e-01 -2.20868677e-01 4.44056183e-01
-6.47777459e-03 -4.45165396e-01 -6.67450011e-01 -2.85500139e-01
6.37212217e-01 -1.10948098e+00 -5.15003204e-01 -1.15963869e-01
-2.03454539e-01 -3.38029593e-01 1.00007820e+00 -1.09731838e-01
9.48157668e-01 -2.22385311e+00 -3.90074760e-01 2.32332900e-01
1.21842548e-01 2.31500968e-01 2.61025071e-01 1.18508480e-01
5.04251420e-01 3.85019630e-02 -4.15531069e-01 -1.68217435e-01
-1.90178171e-01 5.32618821e-01 -2.06654653e-01 2.42608011e-01
2.72180000e-03 6.47849500e-01 -9.07212377e-01 -6.85129881e-01
5.33877492e-01 2.13384777e-01 -4.24411386e-01 -1.17080994e-01
-2.06672385e-01 4.09347564e-01 -8.97518814e-01 5.75829208e-01
1.27198946e+00 9.13410913e-03 -3.11247349e-01 -1.90482363e-02
-2.92204618e-01 1.64901912e-01 -1.19684279e+00 1.51488984e+00
-4.03501838e-01 4.13181335e-01 -1.04180045e-01 -8.22856903e-01
1.41907716e+00 -2.36615464e-01 6.14289045e-01 -8.93653691e-01
-2.71066967e-02 3.93127292e-01 -7.17697218e-02 -2.43338048e-01
7.84895182e-01 -3.75775605e-01 8.49029806e-04 -1.25815466e-01
-7.90275872e-01 -5.76828659e-01 -7.19867274e-02 -8.26098397e-02
5.49543917e-01 1.13104351e-01 -8.98394585e-02 -3.55910540e-01
3.96679014e-01 5.09753108e-01 8.62793684e-01 5.49085200e-01
-3.02038044e-01 6.55331731e-01 2.30648503e-01 -2.82312065e-01
-9.05853987e-01 -9.70237613e-01 -4.52686995e-01 4.15248752e-01
9.72946942e-01 -1.06649101e-01 -6.90156877e-01 -4.72337425e-01
2.67094612e-01 8.72501910e-01 -1.01673231e-01 -2.02403426e-01
-3.93719524e-01 -6.99652314e-01 1.40307054e-01 5.84695160e-01
9.04477119e-01 -8.57207239e-01 -6.70273542e-01 1.57090545e-01
-3.62848848e-01 -1.06284928e+00 -1.37019590e-01 4.74321954e-02
-1.07340479e+00 -1.04475713e+00 -4.16614950e-01 -5.59256494e-01
6.12423062e-01 8.42677414e-01 7.23698795e-01 -1.24315090e-01
1.64941132e-01 -3.95448238e-01 -3.68224144e-01 -7.12817311e-01
1.54941634e-03 4.00027215e-01 -8.68527740e-02 -3.81977230e-01
9.10667896e-01 -4.79122967e-01 -5.96365929e-01 7.29584157e-01
-7.66666174e-01 6.65166676e-02 4.86745983e-01 5.68338335e-01
8.80309343e-01 5.60099423e-01 3.72033060e-01 -5.99686861e-01
2.51702011e-01 -4.14784938e-01 -9.03168559e-01 -3.15568268e-01
-7.98711240e-01 -7.22427294e-02 5.65525770e-01 -1.87791973e-01
-8.21949720e-01 7.55831823e-02 -2.24755123e-01 -5.71434975e-01
-1.61031067e-01 1.77054897e-01 -3.36791843e-01 -1.21055812e-01
3.77965659e-01 2.70478845e-01 2.79278666e-01 -4.58421737e-01
1.34407565e-01 7.74527013e-01 2.30565906e-01 -5.64057767e-01
1.05052865e+00 7.75474131e-01 1.33997723e-01 -7.64677107e-01
-8.39789867e-01 -8.08310330e-01 -4.54362422e-01 -2.00269092e-02
8.06505382e-01 -1.03577495e+00 -4.48249578e-01 4.86483008e-01
-8.31065714e-01 -1.76707163e-01 -2.77867526e-01 5.32823920e-01
-4.26445335e-01 3.21331054e-01 -7.76410699e-02 -7.52826095e-01
-1.07228957e-01 -1.45404291e+00 1.25391281e+00 5.67354858e-01
1.04397252e-01 -5.32253146e-01 -1.84316888e-01 3.72304320e-01
2.63703287e-01 5.05902544e-02 6.23676181e-01 -2.72579432e-01
-8.72595251e-01 -2.38386452e-01 -4.92971867e-01 3.27225864e-01
1.78592712e-01 -6.97307140e-02 -7.31475413e-01 -3.97338457e-02
3.18891145e-02 3.30383703e-02 7.85281956e-01 3.40563715e-01
1.43365669e+00 2.32609183e-01 -6.29116595e-01 6.46477342e-01
1.45548427e+00 2.40384370e-01 7.37957001e-01 5.99714994e-01
7.51444638e-01 7.76777387e-01 1.37362933e+00 1.47641048e-01
6.28706813e-01 6.15851700e-01 7.39104867e-01 6.98834844e-03
1.06523961e-01 -4.89580333e-01 -1.74283668e-01 5.14574409e-01
1.44328952e-01 5.03006168e-02 -1.02150726e+00 6.50514722e-01
-1.86273563e+00 -8.86079013e-01 -4.38161135e-01 2.32314277e+00
5.24708688e-01 4.94535357e-01 3.51378582e-02 1.54971540e-01
7.65951395e-01 2.88711220e-01 -6.82190657e-01 -8.73544347e-03
1.28697887e-01 3.76393683e-02 1.07726252e+00 3.57296586e-01
-8.62074614e-01 1.12779880e+00 5.57484341e+00 1.26981843e+00
-1.21914971e+00 -3.34842205e-02 3.93235624e-01 -1.98934935e-02
-4.19060230e-01 1.23035265e-02 -1.19395924e+00 8.98339212e-01
6.78283811e-01 -1.34600773e-01 1.67094097e-02 1.09462082e+00
5.50686538e-01 -3.07720035e-01 -3.16358984e-01 1.03346109e+00
-4.71924603e-01 -1.15871310e+00 -6.55819848e-02 3.42835069e-01
6.32965803e-01 4.21693444e-01 -1.21104293e-01 3.59201849e-01
2.76723176e-01 -8.41116905e-01 7.99114108e-01 3.00096065e-01
6.83007479e-01 -9.71539259e-01 7.89786160e-01 7.92222083e-01
-1.33436179e+00 1.35437315e-02 -8.43813539e-01 -1.34278819e-01
2.65584618e-01 9.31404889e-01 -6.64710820e-01 6.33716047e-01
9.52818274e-01 6.26189530e-01 -6.04771733e-01 1.22361958e+00
-7.01273680e-02 4.76841033e-01 -5.95457554e-01 -1.15918167e-01
3.28089535e-01 -7.10428774e-01 5.52414894e-01 6.90775752e-01
4.78878945e-01 -5.16806729e-02 4.31551158e-01 6.46758735e-01
2.78730899e-01 2.03327954e-01 -6.80038214e-01 3.89111072e-01
9.24602628e-01 1.16692257e+00 -7.83493161e-01 -2.92883247e-01
-3.35301459e-01 3.16732138e-01 1.34539098e-01 2.44121924e-01
-9.38478947e-01 -4.17579412e-01 7.95654297e-01 6.50921702e-01
3.00013423e-01 -5.36908925e-01 -7.14155018e-01 -8.20647001e-01
3.06853563e-01 -4.42223966e-01 8.56501702e-03 -7.70731449e-01
-1.00770664e+00 3.91780347e-01 2.28625774e-01 -1.58766711e+00
7.08341599e-02 -3.48455220e-01 -4.87537235e-01 1.03903985e+00
-1.94362545e+00 -9.56884623e-01 -7.22233593e-01 2.35458165e-01
5.97351491e-01 1.81464791e-01 1.91498801e-01 3.08986038e-01
-3.43837529e-01 1.46810114e-01 9.67996269e-02 -2.44302019e-01
4.68248397e-01 -1.06021941e+00 2.88056970e-01 6.96132183e-01
-3.29170763e-01 4.71074849e-01 6.74509704e-01 -7.32834280e-01
-9.66671705e-01 -1.43367183e+00 4.92051899e-01 -3.50094795e-01
3.42108428e-01 -8.83371383e-03 -1.18831289e+00 2.60388196e-01
-4.11023766e-01 2.56911130e-03 1.51652575e-01 -3.07223760e-02
-1.09460160e-01 -5.21861911e-01 -1.27192891e+00 5.94036400e-01
1.10169017e+00 -1.57876089e-01 -4.56785321e-01 1.68100044e-01
7.51575768e-01 -5.50498486e-01 -6.92357481e-01 8.51765931e-01
3.67674887e-01 -1.04023397e+00 8.86149108e-01 1.24053612e-01
1.85196340e-01 -6.98013783e-01 -1.51559219e-01 -1.16918254e+00
-1.96541086e-01 -1.36408001e-01 3.16120118e-01 1.16780221e+00
1.27600923e-01 -1.01558876e+00 1.04750621e+00 3.29841733e-01
-3.22393000e-01 -7.27879345e-01 -1.00276458e+00 -9.39448893e-01
2.24273756e-01 -6.29567921e-01 1.02600551e+00 5.91166556e-01
-6.00793421e-01 1.49823129e-01 2.14775875e-01 3.17920297e-01
6.76454782e-01 3.72742623e-01 1.05647767e+00 -1.55668962e+00
4.47140127e-01 -4.98581290e-01 -4.49604332e-01 -1.34005666e+00
1.02648191e-01 -6.50502801e-01 4.26237583e-01 -1.62791073e+00
1.27919680e-02 -1.18881989e+00 4.67265360e-02 7.43810609e-02
-2.75227696e-01 2.25358397e-01 -5.90234958e-02 5.34321547e-01
-3.97173643e-01 7.69774258e-01 1.40496659e+00 -2.72287289e-03
-1.76048025e-01 3.91268641e-01 -6.57224357e-01 7.96495914e-01
1.04871583e+00 -5.68475962e-01 -6.64872706e-01 -4.40756172e-01
1.59432348e-02 -3.19056779e-01 2.84153491e-01 -1.28067231e+00
1.23423964e-01 -5.76109648e-01 2.00642332e-01 -1.20362508e+00
3.09645653e-01 -9.77287889e-01 1.49338692e-01 4.19741213e-01
2.82372624e-01 -9.39988494e-02 2.26338506e-01 6.54214501e-01
-3.64234954e-01 -1.82314157e-01 1.00950396e+00 -7.56360814e-02
-6.70673907e-01 6.30477250e-01 1.98377091e-02 -6.44983873e-02
1.12372613e+00 -6.36135578e-01 -1.22356825e-01 -9.48465019e-02
-4.21302617e-02 6.90732360e-01 9.60725725e-01 4.49021667e-01
4.63579178e-01 -1.35561216e+00 -4.04600203e-01 2.35927105e-01
2.83324122e-01 9.27245975e-01 2.64849067e-01 7.57890463e-01
-6.44014657e-01 4.06218708e-01 -2.10891236e-02 -1.05947149e+00
-9.14839566e-01 3.85493815e-01 2.34369665e-01 9.19240564e-02
-7.24434853e-01 5.69112241e-01 1.92486972e-01 -6.56473696e-01
-1.00240856e-01 -5.94686568e-01 -1.99299634e-01 -5.83065003e-02
1.16402932e-01 4.81217474e-01 1.10502392e-01 -7.24156499e-01
-3.19914043e-01 9.10489559e-01 7.38253221e-02 1.38178423e-01
9.39659894e-01 -3.24388176e-01 8.69631693e-02 5.32061160e-01
8.93469036e-01 2.33172774e-01 -1.48046243e+00 -2.39372641e-01
-1.26098201e-01 -8.76837134e-01 1.24484554e-01 -1.60862699e-01
-9.40692782e-01 9.51559424e-01 4.54880267e-01 4.18594301e-01
8.05831850e-01 -5.61532676e-02 1.20409179e+00 1.95445754e-02
8.00302863e-01 -1.17749035e+00 -4.04686868e-01 4.53082412e-01
4.59860146e-01 -1.33845055e+00 1.87999979e-01 -8.30099463e-01
-7.32601285e-01 7.64786541e-01 7.60984063e-01 -1.17298074e-01
5.19462883e-01 3.33599374e-02 1.18751511e-01 -2.09868759e-01
-1.81940734e-01 -3.55125606e-01 1.86677519e-02 6.09868467e-01
-8.10940862e-02 2.38044351e-01 -3.31095368e-01 2.37013549e-01
-5.69973469e-01 -7.21714497e-02 1.29095212e-01 7.01044917e-01
-9.55541611e-01 -1.02263141e+00 -4.03520912e-01 5.69111049e-01
1.46618849e-02 2.22947747e-01 2.27590129e-01 8.79004776e-01
3.71442974e-01 8.41476202e-01 3.27857554e-01 -2.04449609e-01
5.15375972e-01 -1.37351647e-01 9.93150175e-02 -6.62802279e-01
1.49633944e-01 3.32770087e-02 -2.75563300e-01 -6.03239417e-01
-3.44985932e-01 -8.58374059e-01 -1.66359770e+00 -3.78071308e-01
-5.23630738e-01 3.65936279e-01 8.45953166e-01 8.31095159e-01
3.07488739e-01 4.35956508e-01 6.75478697e-01 -6.98497772e-01
-6.46847665e-01 -8.08314204e-01 -6.12070382e-01 2.19317794e-01
-6.85726292e-03 -1.01396155e+00 -3.82616222e-01 -4.99507964e-01]
|
[8.060661315917969, -2.7728488445281982]
|
896a35a2-7ed3-4f91-baae-b7dddee2e574
|
gcpg-a-general-framework-for-controllable
| null | null |
https://openreview.net/forum?id=3YX-sCVoGl
|
https://openreview.net/pdf?id=3YX-sCVoGl
|
GCPG: A General Framework for Controllable Paraphrase Generation
|
Controllable paraphrase generation (CPG) incorporates various external conditions to obtain desirable paraphrases. However, existing works only highlight a special condition under two indispensable aspects of CPG (i.e., lexically and syntactically CPG) individually, lacking a unified circumstance to explore and analyze their effectiveness. In this paper, we propose a general controllable paraphrase generation framework (GCPG), which represents both lexical and syntactical conditions as text sequences and uniformly processes them in an encoder-decoder paradigm. Under GCPG, we reconstruct commonly adopted lexical condition (i.e., Keywords) and syntactical conditions (i.e., Part-Of-Speech sequence, Constituent Tree, Masked Template and Sentential Exemplar) and study the combination of the two types. In particular, for Sentential Exemplar condition, we propose a novel exemplar construction method --- Syntax-Similarity based Exemplar (SSE). SSE retrieves a syntactically similar but lexically different sentence as the exemplar for each target sentence, avoiding exemplar-side words copying problem. Extensive experiments demonstrate that GCPG with SSE achieves state-of-the-art performance on two popular benchmarks. In addition, the combination of lexical and syntactical conditions shows the significant controllable ability of paraphrase generation, and these empirical results could provide novel insight to user-oriented paraphrasing.
|
['Anonymous']
|
2021-10-16
| null | null | null |
acl-arr-october-2021-10
|
['paraphrase-generation', 'paraphrase-generation']
|
['computer-code', 'natural-language-processing']
|
[ 3.46883178e-01 -3.75582308e-01 -2.08086759e-01 -3.09107423e-01
-7.03200698e-01 -6.19414508e-01 6.75742567e-01 -8.95074382e-02
-1.25891501e-02 8.33609760e-01 6.39264822e-01 -2.96830416e-01
-1.39482722e-01 -8.09589446e-01 -8.24502051e-01 -6.00591779e-01
6.04578555e-01 1.01843707e-01 4.64397743e-02 -5.64112604e-01
6.67882562e-01 2.82879323e-01 -1.64620233e+00 5.51161468e-01
1.26362944e+00 4.44118708e-01 6.84149563e-01 2.94607133e-01
-5.10515273e-01 4.01714325e-01 -8.74491632e-01 -7.17590332e-01
1.16105922e-01 -9.11679447e-01 -5.14389098e-01 1.30380839e-01
1.93947271e-01 3.20599750e-02 -1.97139010e-01 1.16519427e+00
6.59784555e-01 -1.90582201e-02 6.42052531e-01 -1.10925353e+00
-1.18598366e+00 8.13581169e-01 -4.47544813e-01 3.01447958e-01
8.43152344e-01 3.35022837e-01 9.77110088e-01 -1.08345056e+00
5.50246775e-01 1.19277322e+00 4.09867793e-01 4.32216197e-01
-1.10140550e+00 -5.84721863e-01 -4.70515282e-04 3.14050317e-01
-1.37176883e+00 -3.60263616e-01 9.21404183e-01 -2.44257785e-02
1.06862426e+00 4.49491888e-01 6.02993250e-01 1.41768169e+00
3.84655327e-01 8.80122125e-01 1.40416479e+00 -6.16575837e-01
1.81946561e-01 2.74433434e-01 2.55958647e-01 3.78296912e-01
2.54046232e-01 1.29725588e-02 -4.97397602e-01 1.91715825e-02
6.25817478e-01 3.75336567e-05 -5.07452965e-01 -1.41914800e-01
-1.31995666e+00 6.21351182e-01 8.65167007e-02 3.21634352e-01
-2.28289217e-01 -2.16471151e-01 7.38419712e-01 7.31334150e-01
-1.07418492e-01 4.03901130e-01 -2.89821446e-01 -7.20704049e-02
-8.86124372e-01 3.56950194e-01 7.76335239e-01 1.54759026e+00
6.19982004e-01 1.03968903e-01 -6.39864028e-01 1.17191339e+00
-5.64363822e-02 7.73763478e-01 1.16485727e+00 -3.77410948e-01
6.74066961e-01 4.96711552e-01 -7.10225776e-02 -8.91036570e-01
1.47602772e-02 -6.08433068e-01 -8.09071302e-01 -5.44968665e-01
-2.57440567e-01 1.95469484e-01 -5.53519189e-01 1.87015021e+00
6.61303625e-02 1.50229663e-01 3.34633321e-01 7.44531631e-01
1.01485932e+00 8.23171735e-01 -1.42381489e-01 -7.49713361e-01
1.38430631e+00 -1.15106177e+00 -7.28330135e-01 -3.00273150e-01
3.70775998e-01 -9.73189712e-01 1.83894634e+00 1.17990315e-01
-1.16915417e+00 -7.77046084e-01 -1.10213614e+00 1.61988482e-01
-2.42860883e-01 7.97493756e-02 2.02256367e-01 5.75177014e-01
-7.90532708e-01 4.61209029e-01 -6.94498839e-03 -3.90139401e-01
-1.92455694e-01 -7.14755356e-02 -2.89778411e-01 -2.70121377e-02
-1.58369517e+00 9.51952815e-01 6.88656449e-01 -9.62387323e-02
-6.41934693e-01 -5.85511386e-01 -7.84814835e-01 1.40313521e-01
2.76018173e-01 -9.37213063e-01 1.24850821e+00 -9.82939184e-01
-1.49604905e+00 8.51064920e-01 -3.02970737e-01 -3.51245880e-01
2.83987641e-01 -6.01493642e-02 -6.00190699e-01 1.38911894e-02
2.10290432e-01 2.88741976e-01 1.03949964e+00 -1.16338563e+00
-3.52123499e-01 -1.03871860e-01 -2.26067584e-02 5.89890778e-01
-4.08703685e-01 2.43849844e-01 -3.40422571e-01 -1.00149953e+00
7.86604583e-02 -6.98382080e-01 1.53423399e-01 -4.73126531e-01
-5.27630150e-01 -1.61586612e-01 5.69015682e-01 -5.98046124e-01
1.74274826e+00 -2.16333270e+00 4.47705448e-01 -1.56444907e-02
-2.57150024e-01 3.48285526e-01 -1.87511966e-01 9.60297942e-01
-3.12751591e-01 1.78685859e-01 -3.68643463e-01 -1.93164736e-01
9.49881971e-02 3.28030944e-01 -5.29531598e-01 -1.00835674e-01
-3.22353654e-02 1.13593876e+00 -1.04779291e+00 -6.97446942e-01
1.41072065e-01 -1.73511595e-01 -4.46064264e-01 2.53115445e-01
-5.12812324e-02 8.89098942e-02 -4.10677165e-01 6.07585847e-01
5.47729075e-01 6.27101958e-02 -7.45066032e-02 -2.79141098e-01
-1.84807796e-02 4.45895761e-01 -9.77061331e-01 1.81583714e+00
-8.11110795e-01 -4.25663358e-03 -3.53499919e-01 -8.16638947e-01
1.14771092e+00 1.88595340e-01 -2.64541626e-01 -7.55262375e-01
6.09959997e-02 5.47449946e-01 -6.15420528e-02 -7.90323734e-01
5.18189967e-01 -4.71876383e-01 -2.30899066e-01 4.14090097e-01
-1.33702382e-01 -3.51093262e-01 2.70897627e-01 2.57220656e-01
9.95546281e-01 1.90421820e-01 8.17519546e-01 -3.03462356e-01
9.57805216e-01 -6.64893389e-02 4.46930766e-01 8.72874141e-01
-1.90926194e-01 7.48252094e-01 3.29993248e-01 1.54666916e-01
-1.21880484e+00 -1.23935676e+00 -2.98818853e-02 7.98197269e-01
4.23932999e-01 -5.50212622e-01 -7.61839867e-01 -6.01571083e-01
-1.39736623e-01 1.19939399e+00 -2.90196955e-01 -5.57911277e-01
-7.98213780e-01 -5.42249620e-01 6.91641986e-01 4.27290857e-01
7.34971881e-01 -1.58812094e+00 -2.30233863e-01 2.31689990e-01
-3.10910761e-01 -8.10593903e-01 -7.21260548e-01 -1.20995700e-01
-5.51516473e-01 -6.43174350e-01 -6.30017936e-01 -1.01523554e+00
3.77574384e-01 5.74711859e-01 1.17391360e+00 7.42722601e-02
4.12001871e-02 -3.24003473e-02 -9.00361717e-01 8.90566595e-03
-8.29500675e-01 -6.66861460e-02 -6.41914755e-02 -1.04198635e-01
3.33834648e-01 -9.12958264e-01 -5.90619922e-01 2.21975937e-01
-1.05704260e+00 4.18885678e-01 9.58517253e-01 1.06656587e+00
5.52802682e-01 -2.41756156e-01 1.01096129e+00 -8.22854996e-01
1.41179776e+00 -6.37938023e-01 1.09111533e-01 6.57961786e-01
-4.64425772e-01 9.89279076e-02 1.23209798e+00 -4.32322234e-01
-1.06936097e+00 -4.07832116e-01 -2.88934022e-01 -4.86575067e-01
-1.72841370e-01 6.83239758e-01 -4.13485974e-01 3.35195005e-01
6.96334958e-01 1.24347031e+00 -1.06061175e-01 -3.31937522e-01
5.45787692e-01 1.10621762e+00 5.75283289e-01 -9.25254583e-01
8.37352097e-01 -1.19302757e-01 -2.22155809e-01 -5.66260457e-01
-5.88613927e-01 -2.83836305e-01 -4.09220815e-01 1.02129765e-02
4.18926001e-01 -6.64155245e-01 -7.97227696e-02 2.97639996e-01
-1.30031121e+00 2.63395101e-01 -3.11386079e-01 1.52452871e-01
-7.39175916e-01 7.96430588e-01 -5.71339071e-01 -3.84472430e-01
-6.86500013e-01 -1.25633347e+00 1.24324536e+00 8.60320255e-02
-2.96783715e-01 -5.57323456e-01 -8.72206911e-02 1.25238299e-01
3.45768839e-01 -6.27934486e-02 1.23428416e+00 -7.58292496e-01
-3.24453503e-01 4.85810824e-02 -2.54999012e-01 5.23377240e-01
1.94351241e-01 -1.17525809e-01 -5.50716221e-01 -1.34010628e-01
3.13399404e-01 -1.71034545e-01 4.45092350e-01 -3.68160866e-02
1.05719090e+00 -5.70518613e-01 -1.62927881e-01 4.11241025e-01
1.37593627e+00 3.49755049e-01 8.92080128e-01 2.73558289e-01
3.09610665e-01 4.68145370e-01 6.95678711e-01 4.36434001e-01
1.37122953e-02 8.53825927e-01 -6.96354434e-02 3.68346751e-01
-3.68738919e-01 -6.19958103e-01 5.50382376e-01 1.45017064e+00
3.84985775e-01 -2.94647634e-01 -4.75836158e-01 3.56432170e-01
-1.71415079e+00 -1.11706293e+00 -8.52265954e-02 2.21855140e+00
1.26780069e+00 2.25856751e-01 -1.24899551e-01 1.97209328e-01
1.06549656e+00 2.21854612e-01 -3.59094918e-01 -6.22192919e-01
-4.22853380e-01 3.64253700e-01 -1.19771697e-01 1.81488261e-01
-5.91551483e-01 1.05619681e+00 5.22092533e+00 1.50109661e+00
-8.53093863e-01 1.09503917e-01 3.11309826e-02 3.77603173e-02
-7.67266989e-01 1.80945978e-01 -7.93416739e-01 9.63937938e-01
6.81365788e-01 -7.16014087e-01 4.73858714e-01 6.74820960e-01
4.25342351e-01 3.19472075e-01 -9.94067311e-01 9.74714637e-01
3.53133172e-01 -1.12763655e+00 7.24450469e-01 -5.13109088e-01
5.44555008e-01 -5.84284604e-01 -2.64104530e-02 6.20755136e-01
-1.98644504e-01 -7.82584131e-01 8.10202658e-01 2.74825841e-01
8.56358230e-01 -6.24016821e-01 5.76134801e-01 5.45905828e-01
-1.22068799e+00 -7.91055113e-02 -5.55261493e-01 7.46298432e-02
4.38870311e-01 3.80867213e-01 -3.29418153e-01 1.10060430e+00
2.08344147e-01 8.17877054e-01 -7.40895212e-01 9.55737650e-01
-4.94985849e-01 6.06027901e-01 1.02095455e-02 -4.80283171e-01
2.82910138e-01 -3.95489931e-01 8.46499324e-01 1.39174974e+00
4.62952316e-01 -1.06067117e-02 -1.01994962e-01 1.02927256e+00
2.02663153e-01 5.80266058e-01 -7.58528233e-01 1.83651716e-01
1.08902740e+00 8.30446184e-01 -4.41276461e-01 -4.62567538e-01
-4.03069317e-01 1.23046184e+00 3.64842236e-01 2.88227737e-01
-1.07627821e+00 -5.80363274e-01 2.01843902e-01 2.14558281e-02
1.79683208e-01 7.76468143e-02 -2.60054171e-01 -1.47492898e+00
3.93532157e-01 -1.34831762e+00 6.20229058e-02 -1.14355123e+00
-1.52366877e+00 5.52482903e-01 3.96480024e-01 -1.49902105e+00
-1.33382469e-01 -2.05760375e-01 -9.89158273e-01 1.00272179e+00
-1.21143126e+00 -1.11450326e+00 -3.50856215e-01 5.98540962e-01
1.13845658e+00 -2.27157697e-01 6.56599581e-01 2.61813164e-01
-7.25895882e-01 8.19599211e-01 1.56196862e-01 -2.76161104e-01
5.88119507e-01 -9.49240506e-01 5.11653304e-01 8.60751629e-01
1.72235519e-01 1.28017533e+00 8.77329051e-01 -6.81725681e-01
-1.42796707e+00 -1.00665343e+00 1.03255653e+00 -1.02172747e-01
6.93572044e-01 -2.54970819e-01 -9.57542598e-01 4.57431376e-01
4.10286993e-01 -7.26271331e-01 4.11998183e-01 -2.27469206e-01
-4.10891086e-01 -8.20670128e-02 -1.01734591e+00 1.24646008e+00
1.50755668e+00 -4.94554073e-01 -1.29326677e+00 5.29954672e-01
1.20136499e+00 -3.27967733e-01 -4.61195022e-01 5.42524993e-01
3.04395974e-01 -1.20907986e+00 1.02058077e+00 -5.94236016e-01
9.67628956e-01 -3.71751487e-01 -2.67182052e-01 -1.45028269e+00
-3.58566552e-01 -5.86113691e-01 -1.03287764e-01 1.45626652e+00
2.80497789e-01 -7.01690972e-01 3.92018437e-01 -4.24400829e-02
-7.24332035e-01 -1.04721653e+00 -7.77895570e-01 -1.20472920e+00
1.70545816e-01 -4.51387428e-02 9.13323581e-01 8.39656532e-01
2.18717992e-01 6.89158201e-01 -2.63840973e-01 -2.94893563e-01
2.40491942e-01 4.81794327e-01 7.00803936e-01 -3.83984596e-01
-5.94254494e-01 -6.56477988e-01 -2.19901562e-01 -1.32395566e+00
3.76940966e-01 -1.21547341e+00 -2.04420537e-01 -1.38793564e+00
3.53288680e-01 -2.75011957e-01 -1.95768699e-01 4.62771691e-02
-5.33981919e-01 -1.77414104e-01 1.94372654e-01 4.52088743e-01
-2.22806677e-01 8.36719692e-01 1.36625707e+00 -2.65603289e-02
-1.46128818e-01 -5.24339154e-02 -7.74454296e-01 2.58022308e-01
9.86813188e-01 -3.20682436e-01 -8.15181434e-01 -2.23150596e-01
7.60455877e-02 2.60914773e-01 3.24644744e-01 -8.10326040e-01
6.85745776e-02 -4.21118468e-01 -1.43677875e-01 -5.48113465e-01
1.39000103e-01 -5.33296406e-01 3.69288415e-01 5.57172000e-01
-5.35970330e-01 4.94470388e-01 1.86964218e-02 6.84821904e-01
-4.10675436e-01 -8.84034097e-01 5.85720778e-01 -3.18721682e-01
-8.11822414e-01 -2.10472092e-01 -9.20754224e-02 3.74059737e-01
1.08963931e+00 -5.73769391e-01 -4.74678397e-01 -2.20268890e-01
-2.50958502e-01 4.23698639e-03 5.55634379e-01 5.47098339e-01
7.98131943e-01 -1.47093689e+00 -9.79144633e-01 4.47298616e-01
4.30200487e-01 -4.87376183e-01 2.74897456e-01 5.40712714e-01
-5.35920560e-01 3.06286961e-01 -1.58424512e-01 -3.35333198e-01
-1.21507680e+00 9.16740954e-01 1.66383296e-01 -2.38229245e-01
-6.67152226e-01 7.26724565e-01 2.01366723e-01 -3.97033840e-01
-5.27781211e-02 -1.78447336e-01 -1.66195691e-01 -2.92284042e-01
2.99844116e-01 1.49865672e-02 7.63474079e-03 -4.50005919e-01
-8.84103552e-02 4.46310014e-01 -2.68237203e-01 -7.68340454e-02
6.85652912e-01 -2.21604198e-01 -1.92990690e-01 3.51125926e-01
1.18913746e+00 1.14231691e-01 -4.17385191e-01 -2.85544366e-01
-1.61802292e-01 -5.63636303e-01 -6.87582254e-01 -6.88142478e-01
-6.00428700e-01 6.88983023e-01 -1.39157727e-01 1.40532821e-01
1.21525609e+00 -1.71260908e-01 1.08995986e+00 4.15418923e-01
6.07844234e-01 -8.08499157e-01 1.23056985e-01 4.12198961e-01
1.27907562e+00 -6.12638354e-01 -7.08423927e-02 -6.02568150e-01
-8.38823080e-01 8.52217853e-01 7.10936010e-01 -1.83769420e-01
2.44886637e-01 2.47908439e-02 -3.70589882e-01 2.41908189e-02
-7.75520861e-01 8.89148489e-02 7.68134892e-02 3.83343875e-01
3.57261091e-01 -5.06408513e-02 -1.08968961e+00 7.84531891e-01
-6.89007223e-01 -1.22870035e-01 5.13543248e-01 9.96023357e-01
-6.01987779e-01 -1.14760339e+00 -1.58020273e-01 4.97682512e-01
-3.74940597e-02 -7.19037652e-01 -4.36066389e-01 7.39708006e-01
4.78182510e-02 8.65646780e-01 -3.64152551e-01 -3.84832382e-01
6.71217263e-01 2.09217906e-01 5.25896192e-01 -8.90341938e-01
-9.45352733e-01 -1.45335719e-01 4.60327081e-02 -2.44079694e-01
-1.65872157e-01 -4.57519323e-01 -9.57457840e-01 -3.94149929e-01
-4.77507025e-01 3.36222202e-01 3.20279419e-01 9.25024867e-01
3.27166945e-01 4.00988966e-01 8.12742352e-01 -3.98655117e-01
-1.11315989e+00 -1.10603082e+00 -5.61945617e-01 8.18439484e-01
-2.39008874e-01 -5.22042751e-01 -4.91173089e-01 -3.29588237e-03]
|
[11.692728042602539, 9.364376068115234]
|
b7e565e8-af33-4e48-b717-607b4b6c7a72
|
fine-tuning-large-language-models-for
| null | null |
https://link.springer.com/chapter/10.1007/978-3-031-36021-3_15
|
https://link.springer.com/chapter/10.1007/978-3-031-36021-3_15
|
Fine-Tuning Large Language Models for Answering Programming Questions with Code Snippets
|
We study the ability of pretrained large language models (LLM) to answer questions from online question answering fora such as Stack Overflow. We consider question-answer pairs where the main part of the answer consists of source code. On two benchmark datasets—CoNaLa and a newly collected dataset based on Stack Overflow—we investigate how a closed-book question answering system can be improved by fine-tuning the LLM for the downstream task, prompt engineering, and data preprocessing. We use publicly available autoregressive language models such as GPT-Neo, CodeGen, and PanGu-Coder, and after the proposed fine-tuning achieve a BLEU score of 0.4432 on the CoNaLa test set, significantly exceeding previous state of the art for this task.
|
['Artem Aliev', 'Sergey Nikolenko', 'Maxim Omelchenko', 'Sergey Kovalchuk', 'Vadim Lomshakov']
|
2023-06-26
| null | null | null |
iccs-international-conference-on
|
['code-generation', 'program-synthesis', 'text-to-code-generation', 'question-answering', 'prompt-engineering']
|
['computer-code', 'computer-code', 'computer-code', 'natural-language-processing', 'natural-language-processing']
|
[-3.35070372e-01 2.40087450e-01 2.73513943e-01 -3.38514477e-01
-1.49247754e+00 -9.23607290e-01 8.38330016e-02 2.51964748e-01
-3.22351217e-01 7.42178932e-02 2.02074081e-01 -1.00905418e+00
-6.62430227e-02 -7.58003473e-01 -9.57511187e-01 2.64656961e-01
2.47138157e-01 4.35866624e-01 4.87361729e-01 -4.23027098e-01
4.73077625e-01 -3.35034311e-01 -1.18835628e+00 6.85121834e-01
1.29855537e+00 7.96153486e-01 2.34456226e-01 1.36273098e+00
-9.01874542e-01 1.68078482e+00 -6.63829148e-01 -9.26180482e-01
-3.26859429e-02 -1.49612084e-01 -1.23997629e+00 -5.53140163e-01
7.84313977e-01 -2.87697524e-01 -2.51478314e-01 9.20843601e-01
3.68260682e-01 9.12249908e-02 2.79589742e-01 -8.40845883e-01
-1.19421685e+00 8.57876241e-01 -3.19939286e-01 3.90874743e-01
5.80002785e-01 3.18282872e-01 1.49217570e+00 -8.72491837e-01
3.71786207e-01 1.27551639e+00 7.03609705e-01 6.25104845e-01
-1.20290935e+00 -1.81694686e-01 -1.15160376e-01 3.33089203e-01
-6.46796048e-01 -2.82240957e-01 4.99061912e-01 -6.32154822e-01
1.40349591e+00 2.14230701e-01 -3.11862588e-01 7.94639468e-01
2.11203650e-01 8.04132044e-01 6.25101686e-01 -5.78818500e-01
7.42222741e-02 1.61679983e-01 1.14874971e+00 7.94380665e-01
-8.91043022e-02 -5.30672908e-01 -2.47976869e-01 -5.57298601e-01
-8.68175700e-02 -1.88701808e-01 -1.20100230e-01 1.51620299e-01
-7.40599573e-01 8.44186068e-01 1.42744377e-01 3.34064931e-01
-2.49726757e-01 2.94288963e-01 3.64874721e-01 1.05374551e+00
1.14299655e-01 9.28698301e-01 -1.07647741e+00 -5.44167519e-01
-5.24190426e-01 3.58770758e-01 1.43281400e+00 9.85403359e-01
8.72753680e-01 -2.04602018e-01 -5.23362637e-01 9.79752839e-01
3.04448426e-01 4.96629059e-01 7.46005714e-01 -1.20302689e+00
1.09159303e+00 1.11150098e+00 1.77213531e-02 -7.15699315e-01
-3.01188380e-01 -1.93359554e-01 -2.05644891e-02 -4.38722730e-01
7.41375268e-01 -5.80339253e-01 -4.51623708e-01 1.51701903e+00
-1.13888033e-01 -5.74156642e-02 1.44375607e-01 2.47851714e-01
1.32510424e+00 8.25419545e-01 1.23690367e-01 5.58517456e-01
1.46712804e+00 -1.62737584e+00 -4.67315286e-01 -6.95631981e-01
1.25464296e+00 -6.74742758e-01 1.77030027e+00 1.40066981e-01
-1.08351183e+00 -6.98663116e-01 -5.28730333e-01 -7.57247806e-01
-2.05204472e-01 3.00091267e-01 3.60446751e-01 6.93220973e-01
-1.24623036e+00 2.59648889e-01 -4.13513124e-01 -2.93106288e-01
-9.19950902e-02 -1.97093368e-01 -6.20505214e-02 -2.23878950e-01
-9.58840549e-01 4.72669661e-01 -1.24345608e-01 -2.57252783e-01
-6.33382499e-01 -1.26381981e+00 -6.28363848e-01 4.70216483e-01
5.01255810e-01 -7.45918930e-01 1.77652347e+00 -4.01046813e-01
-1.54063237e+00 8.96816552e-01 -3.58881921e-01 -4.39941019e-01
6.84277713e-02 -6.90144718e-01 -1.58938855e-01 -3.26181278e-02
-8.45292583e-02 1.21950723e-01 7.30514526e-01 -5.47291219e-01
-2.87375271e-01 -2.68517017e-01 5.74873030e-01 -6.20315969e-01
-5.09290934e-01 5.02995312e-01 -4.58764255e-01 -1.83065385e-01
-4.66305286e-01 -6.62404895e-01 -2.74757713e-01 -4.57619160e-01
-2.21519813e-01 -4.50333238e-01 2.95545310e-01 -1.52208865e+00
1.78056991e+00 -1.92821622e+00 -2.10337341e-02 -1.83368251e-01
2.64626890e-01 3.36879611e-01 -9.15518582e-01 2.86940664e-01
-7.54432753e-02 4.26304132e-01 -1.60625920e-01 -1.62635103e-01
2.71662921e-01 -1.03034586e-01 -6.96146607e-01 -3.24658513e-01
4.29035157e-01 1.45636296e+00 -7.58613944e-01 -8.75642747e-02
-4.91527677e-01 -1.33057863e-01 -1.29672170e+00 8.14132810e-01
-8.95609200e-01 -9.03491024e-03 -3.97417545e-01 5.54406106e-01
2.98783928e-01 -6.77842975e-01 -2.41436929e-01 5.64028382e-01
3.07460755e-01 7.23055899e-01 -6.05021894e-01 1.57487428e+00
-9.09019053e-01 6.30858600e-01 9.66688916e-02 -4.07043993e-01
1.18805730e+00 9.67266485e-02 -2.49921441e-01 -8.97465587e-01
-8.51493552e-02 2.71045566e-01 9.60006863e-02 -1.11607230e+00
6.23109758e-01 5.49873233e-01 -2.62420446e-01 5.58069050e-01
3.75703424e-01 -1.57427847e-01 3.34484428e-01 1.90317586e-01
1.88444877e+00 -8.74336138e-02 -1.26358002e-01 -2.91942656e-01
1.03545225e+00 -2.98839122e-01 5.14790714e-02 1.19098616e+00
-1.12077042e-01 4.46500540e-01 1.16854334e+00 -7.69117698e-02
-8.26477468e-01 -7.57335126e-01 3.45753431e-01 1.85573232e+00
-6.44442737e-01 -7.37692356e-01 -1.08288848e+00 -1.01053989e+00
6.55840486e-02 1.14082003e+00 -3.60759974e-01 -2.51722068e-01
-8.77230406e-01 -4.72322583e-01 6.26360059e-01 5.60925424e-01
1.67768270e-01 -1.04725897e+00 -3.31033647e-01 3.75314593e-01
-3.77255887e-01 -1.18532419e+00 -5.45778394e-01 -4.84700762e-02
-7.63145089e-01 -1.14043486e+00 -6.15782797e-01 -6.66325390e-01
2.28478014e-01 -1.50435716e-01 2.06761479e+00 4.82511491e-01
1.43584967e-01 6.43512845e-01 -2.94518828e-01 -1.91863254e-01
-7.38836586e-01 7.15969086e-01 -7.82028317e-01 -2.64657676e-01
6.28236949e-01 -4.92909431e-01 -3.86371672e-01 2.12547347e-01
-7.68824577e-01 -5.58068335e-01 2.55730450e-01 4.73745257e-01
1.04955830e-01 -7.55745769e-01 8.91489804e-01 -1.12269080e+00
8.67329001e-01 -9.10326898e-01 -9.02827740e-01 6.41908646e-01
-3.14033061e-01 4.08638149e-01 7.16917038e-01 -2.97989905e-01
-1.06270158e+00 -6.70761883e-01 -6.26130998e-01 8.11828673e-02
5.89837432e-02 7.91902721e-01 -1.48067415e-01 -1.64629444e-01
1.05638945e+00 -1.32184312e-01 -4.54473794e-01 -9.24448609e-01
5.70545077e-01 6.70034111e-01 5.75574696e-01 -8.43685687e-01
7.14691758e-01 -3.13052863e-01 -5.16089380e-01 -5.57095289e-01
-9.10007179e-01 -5.79226375e-01 -2.87760317e-01 2.73142636e-01
8.92825544e-01 -5.96351326e-01 -7.94872880e-01 4.05334979e-01
-1.53677607e+00 -5.54203331e-01 -9.47914571e-02 -1.99579433e-01
-4.91013557e-01 5.01513064e-01 -9.42475915e-01 -6.37518048e-01
-5.63975334e-01 -1.08322108e+00 8.07238519e-01 2.64433175e-01
-4.24117416e-01 -1.21973169e+00 3.64479214e-01 9.36370075e-01
8.55692387e-01 -3.20970863e-01 1.80511320e+00 -1.20530653e+00
-7.55626142e-01 -3.05012912e-01 -2.47622669e-01 5.67692697e-01
-4.59111899e-01 -1.29648849e-01 -9.41027999e-01 4.49556001e-02
3.20055783e-01 -4.73643988e-01 7.51783907e-01 -1.76800534e-01
1.38621390e+00 -3.14235181e-01 3.02782923e-01 3.93901974e-01
1.20018995e+00 -1.85715690e-01 6.80676281e-01 4.13683593e-01
7.19059050e-01 8.46339524e-01 1.18363775e-01 5.88873364e-02
8.87375772e-01 7.91581050e-02 2.54944623e-01 7.48863339e-01
-7.64957489e-03 -4.77923751e-01 7.46096671e-01 1.45159745e+00
7.35872507e-01 -2.64740855e-01 -1.54064965e+00 7.83512056e-01
-1.76166224e+00 -1.31361678e-01 -5.88671088e-01 1.84708452e+00
9.04955268e-01 -3.14990729e-02 -1.58630297e-01 -3.67198080e-01
2.57062078e-01 1.52938738e-01 -5.66632032e-01 -5.72393596e-01
1.40961930e-01 5.12476563e-01 1.77759584e-02 7.84959197e-01
-7.14336812e-01 7.68003941e-01 6.28573322e+00 6.01394951e-01
-4.89364415e-01 3.05049539e-01 4.64839756e-01 1.98913261e-01
-6.79060519e-01 2.05242127e-01 -9.98431146e-01 4.40976232e-01
1.97717834e+00 -3.36051852e-01 5.14068604e-01 1.22098172e+00
-2.98956573e-01 8.81323814e-02 -1.24580097e+00 7.15481579e-01
1.44771874e-01 -1.14026332e+00 1.94657929e-02 -4.73829031e-01
6.88537776e-01 2.94554532e-01 -5.56411371e-02 1.28251708e+00
5.71478426e-01 -9.02405679e-01 2.60780543e-01 9.43127275e-01
1.58374891e-01 -2.85612077e-01 7.36585557e-01 6.86193883e-01
-7.63108194e-01 -6.05152905e-01 -3.79715174e-01 -6.73783123e-02
-4.52692956e-02 4.86395091e-01 -5.00872493e-01 2.23433867e-01
7.40535021e-01 2.46452183e-01 -1.51437223e+00 1.09752548e+00
-4.91083533e-01 1.31417978e+00 -2.81618387e-02 -3.10826242e-01
7.95856491e-02 6.72478974e-02 3.93325448e-01 1.00321865e+00
2.97146410e-01 -2.52868712e-01 -2.14415535e-01 1.34219420e+00
-5.48485637e-01 2.24498481e-01 -5.84714897e-02 -2.54218191e-01
2.25956157e-01 1.14422715e+00 1.77924708e-01 -3.14068288e-01
-8.68299961e-01 6.52268529e-01 7.11787641e-01 5.52592754e-01
-6.38838470e-01 -6.54395938e-01 3.76119316e-01 1.53860748e-01
2.00002149e-01 -1.43586129e-01 -4.53514427e-01 -1.51127827e+00
4.71827328e-01 -1.46897101e+00 5.33819675e-01 -9.49736655e-01
-1.34933722e+00 5.96168160e-01 -4.33630526e-01 -3.32864881e-01
-6.40130341e-01 -7.59575903e-01 -7.10504174e-01 1.26709425e+00
-1.78597045e+00 -7.80509770e-01 -2.89006472e-01 8.73442367e-02
4.92957234e-01 -3.16810817e-01 7.50780225e-01 5.35696387e-01
-5.25730729e-01 9.67685580e-01 1.55938998e-01 2.39256114e-01
7.22670555e-01 -1.51922321e+00 1.13187325e+00 9.09495771e-01
2.36769151e-02 9.82336342e-01 4.25182760e-01 -2.93182760e-01
-1.67898738e+00 -1.24140275e+00 1.27468956e+00 -1.46181011e+00
1.32447469e+00 -5.05109310e-01 -1.67073941e+00 8.04404080e-01
1.06541090e-01 -1.21298395e-01 5.20576060e-01 2.29716942e-01
-7.42762864e-01 1.26917973e-01 -8.13729346e-01 3.34636509e-01
4.92847890e-01 -1.11960232e+00 -9.98628378e-01 2.21946955e-01
1.47016406e+00 -2.00639114e-01 -9.81789947e-01 1.24889158e-01
8.82237777e-02 -8.19773674e-01 7.37464666e-01 -1.13062215e+00
8.51593018e-01 1.11198545e-01 -2.61456102e-01 -9.13591981e-01
-1.09693676e-01 -9.58718002e-01 -5.87188423e-01 1.52079058e+00
7.32382476e-01 -5.23815334e-01 6.94501340e-01 1.01785219e+00
6.74248859e-02 -6.33138359e-01 -6.12764060e-01 -5.01459718e-01
7.22557545e-01 -5.73751628e-01 6.22225523e-01 4.83873010e-01
-3.20296794e-01 5.71897924e-01 3.36391211e-01 2.07694471e-02
2.56123215e-01 3.76412533e-02 1.06106830e+00 -1.23153341e+00
-8.60974133e-01 -4.42441851e-01 2.23568529e-01 -1.43488204e+00
7.22814560e-01 -1.07630992e+00 -1.48908839e-01 -1.31580365e+00
1.60958692e-01 9.54703912e-02 5.17040603e-02 2.81665593e-01
-8.22362244e-01 -5.81278265e-01 2.74417549e-01 -1.44036427e-01
-8.23797703e-01 4.57260907e-01 6.46500885e-01 -1.61071435e-01
-5.87809598e-03 2.37028487e-02 -8.77888799e-01 8.16568732e-01
4.16614026e-01 -5.27784288e-01 -8.39621574e-02 -8.83538187e-01
9.26166534e-01 5.19172251e-01 1.85517579e-01 -7.08171606e-01
1.80684686e-01 3.16004068e-01 -6.47429764e-01 -3.21612448e-01
-3.36541951e-01 -5.34395516e-01 -7.31096864e-01 2.22432449e-01
-7.80365884e-01 3.36652666e-01 3.81524712e-01 4.76069957e-01
-2.84531027e-01 -1.06266856e+00 3.93041730e-01 -1.82682574e-01
-7.38674462e-01 2.90028341e-02 -3.15521151e-01 8.41433764e-01
1.87765852e-01 6.09892070e-01 -8.02864432e-01 -4.79160279e-01
-3.54909986e-01 6.92496538e-01 3.78062241e-02 8.66529942e-01
1.38259843e-01 -8.91882539e-01 -8.95552993e-01 -3.74495215e-03
3.44145894e-01 -2.82092899e-01 3.33348691e-01 5.93238652e-01
-5.11439800e-01 6.71787441e-01 4.38604355e-01 -3.51883024e-01
-8.41822267e-01 5.34176946e-01 5.82521319e-01 -8.57721984e-01
-1.18397824e-01 1.00836456e+00 2.11544067e-01 -1.39460444e+00
4.38869484e-02 -9.30138648e-01 -3.47359776e-01 -1.42436981e-01
7.03788519e-01 6.19050682e-01 4.28034484e-01 2.33189985e-01
-1.04224451e-01 4.00286198e-01 -1.05945639e-01 2.75790840e-01
1.28247750e+00 -8.58676359e-02 -5.95925689e-01 5.23411036e-01
1.40045321e+00 1.45370975e-01 -6.94520950e-01 -4.65521842e-01
8.53260100e-01 -8.40341449e-02 -2.31452972e-01 -8.25696528e-01
-6.76697254e-01 1.13061869e+00 2.62008727e-01 2.38574758e-01
7.28050947e-01 1.31535158e-01 9.99227345e-01 1.22380519e+00
1.89264283e-01 -6.50619686e-01 3.83318424e-01 1.34983230e+00
1.07631397e+00 -1.22071922e+00 -9.22434509e-01 -1.59123559e-02
-1.52934954e-01 9.63563383e-01 8.55970860e-01 -3.45607698e-01
6.41892672e-01 1.95584193e-01 1.15776934e-01 -1.66076228e-01
-1.32197154e+00 1.72974586e-01 4.02789742e-01 2.67179251e-01
7.32041717e-01 -4.55506146e-01 9.31961387e-02 1.50633478e+00
-4.32474077e-01 -6.47853911e-02 7.38319099e-01 7.30475724e-01
-5.18346488e-01 -8.30931723e-01 -2.99020231e-01 5.94382465e-01
-8.95017266e-01 -4.95445102e-01 -4.06810433e-01 2.20630080e-01
-7.04504609e-01 1.30590749e+00 -2.05721036e-01 -2.02886522e-01
7.20842600e-01 6.15937948e-01 4.21051756e-02 -8.65163028e-01
-1.15233505e+00 -8.05551291e-01 2.36020088e-01 -6.86686337e-01
4.19864446e-01 -2.66764522e-01 -1.06637597e+00 2.70049535e-02
-2.68093973e-01 2.81069428e-01 3.88478518e-01 8.81980777e-01
7.94973969e-01 7.37539232e-01 1.96252152e-01 1.41875878e-01
-1.05487299e+00 -1.10267186e+00 7.06017986e-02 3.19163501e-01
5.01011312e-01 4.45665307e-02 -5.82316041e-01 3.47767621e-02]
|
[11.286859512329102, 8.011332511901855]
|
dbf98535-4f36-40f3-a08b-7f89f3b3db0e
|
expnet-landmark-free-deep-3d-facial
|
1802.00542
| null |
http://arxiv.org/abs/1802.00542v1
|
http://arxiv.org/pdf/1802.00542v1.pdf
|
ExpNet: Landmark-Free, Deep, 3D Facial Expressions
|
We describe a deep learning based method for estimating 3D facial expression
coefficients. Unlike previous work, our process does not relay on facial
landmark detection methods as a proxy step. Recent methods have shown that a
CNN can be trained to regress accurate and discriminative 3D morphable model
(3DMM) representations, directly from image intensities. By foregoing facial
landmark detection, these methods were able to estimate shapes for occluded
faces appearing in unprecedented in-the-wild viewing conditions. We build on
those methods by showing that facial expressions can also be estimated by a
robust, deep, landmark-free approach. Our ExpNet CNN is applied directly to the
intensities of a face image and regresses a 29D vector of 3D expression
coefficients. We propose a unique method for collecting data to train this
network, leveraging on the robustness of deep networks to training label noise.
We further offer a novel means of evaluating the accuracy of estimated
expression coefficients: by measuring how well they capture facial emotions on
the CK+ and EmotiW-17 emotion recognition benchmarks. We show that our ExpNet
produces expression coefficients which better discriminate between facial
emotions than those obtained using state of the art, facial landmark detection
techniques. Moreover, this advantage grows as image scales drop, demonstrating
that our ExpNet is more robust to scale changes than landmark detection
methods. Finally, at the same level of accuracy, our ExpNet is orders of
magnitude faster than its alternatives.
|
['Iacopo Masi', 'Feng-Ju Chang', 'Ram Nevatia', 'Gerard Medioni', 'Anh Tuan Tran', 'Tal Hassner']
|
2018-02-02
| null | null | null | null |
['3d-facial-expression-recognition']
|
['computer-vision']
|
[-5.06476685e-02 1.06636554e-01 9.68520865e-02 -7.83613265e-01
-7.34578550e-01 -4.78314072e-01 4.96523976e-01 -3.67520422e-01
-5.05111933e-01 3.29105020e-01 -1.83030710e-01 2.28619084e-01
2.96214163e-01 -5.72614849e-01 -5.77381849e-01 -5.55169523e-01
-3.77114624e-01 2.05569044e-01 -3.77974898e-01 -2.83532649e-01
-1.50336266e-01 1.24405432e+00 -1.63779664e+00 1.14050716e-01
-2.70906724e-02 1.65621209e+00 -8.00736547e-01 5.70142150e-01
3.82697359e-02 5.72999239e-01 -5.93962789e-01 -4.67284620e-01
5.39194882e-01 -3.37554574e-01 -4.57343429e-01 2.03945011e-01
1.04405415e+00 -5.61144352e-01 -2.87677553e-02 8.48167598e-01
5.69424391e-01 -2.98578423e-02 7.73380101e-01 -1.45710731e+00
-5.05599499e-01 -3.46935749e-01 -6.85178518e-01 -1.51574582e-01
4.93582278e-01 -1.45369619e-02 8.00079107e-01 -1.18305767e+00
8.34562898e-01 1.47975671e+00 1.05294371e+00 8.28814149e-01
-1.45919108e+00 -7.96868145e-01 -3.15895006e-02 -2.28122622e-01
-1.52364254e+00 -9.06106234e-01 8.12325597e-01 -2.91833311e-01
9.98824358e-01 1.19635247e-01 5.22104383e-01 1.10427332e+00
-7.34616145e-02 4.12017971e-01 1.32310975e+00 -3.91371518e-01
1.77509531e-01 1.09122843e-01 -3.52018297e-01 1.21403885e+00
-3.45421523e-01 1.01188011e-01 -5.18733680e-01 -1.97375059e-01
8.43579292e-01 -2.20412523e-01 -1.08530581e-01 -2.16687873e-01
-4.06117618e-01 7.24525928e-01 3.69944721e-01 2.17112899e-01
-2.75121391e-01 3.82698804e-01 4.38619256e-01 5.48936367e-01
9.16166008e-01 2.84470230e-01 -6.17573798e-01 -3.63822244e-02
-1.03038812e+00 7.34042898e-02 7.64987409e-01 6.41705871e-01
1.06304061e+00 1.79813579e-01 1.32177770e-01 9.17457104e-01
2.48151034e-01 4.87693608e-01 5.59112616e-02 -1.25096035e+00
-2.67736018e-01 5.81553042e-01 -6.78871199e-02 -1.23015428e+00
-6.31577432e-01 1.42426407e-02 -6.25617743e-01 9.26604927e-01
5.79656482e-01 -2.00054586e-01 -1.06392932e+00 2.07298279e+00
4.24298495e-01 -6.12563342e-02 -2.12373033e-01 8.24121058e-01
7.88164914e-01 2.76840746e-01 1.51116163e-01 -1.05795532e-01
1.15574753e+00 -4.71702874e-01 -4.29518491e-01 1.34956732e-01
8.08369040e-01 -6.37660325e-01 9.00814891e-01 4.99031723e-01
-1.10125351e+00 -5.01155436e-01 -8.89267206e-01 -1.58931717e-01
-4.71168637e-01 1.98471501e-01 7.77856588e-01 8.49186182e-01
-1.59550381e+00 7.82972276e-01 -6.32886529e-01 -5.26434898e-01
7.61919439e-01 6.21890426e-01 -1.00101399e+00 2.68269151e-01
-8.39838266e-01 1.04740918e+00 -3.08736086e-01 2.34064162e-01
-7.81770527e-01 -7.03006744e-01 -1.04053843e+00 -1.95577163e-02
-1.73277572e-01 -1.74077079e-01 1.03521585e+00 -1.69978333e+00
-1.70310628e+00 1.58443880e+00 -3.15178901e-01 -8.00138712e-02
4.77187514e-01 1.17544368e-01 -2.81353086e-01 4.81342584e-01
-1.72839180e-01 1.11740363e+00 1.09691548e+00 -1.24448860e+00
-1.65220171e-01 -6.11404598e-01 -3.16336588e-03 -2.02212900e-01
-3.87634903e-01 3.66631210e-01 -2.50328302e-01 -1.21542409e-01
-4.99800891e-02 -8.36286724e-01 -5.92423603e-02 9.06602263e-01
1.43412249e-02 -2.46385902e-01 8.08870733e-01 -4.34757620e-01
4.40405697e-01 -2.21508908e+00 -1.81687608e-01 3.65383297e-01
3.93351406e-01 2.12307379e-01 -5.45113623e-01 -7.62367900e-03
-2.41644904e-01 3.28710943e-01 -1.57515630e-01 -8.60586643e-01
2.94128865e-01 1.40064448e-01 -1.24441236e-01 8.02193403e-01
6.68341517e-01 1.01819873e+00 -4.62912738e-01 -4.87939775e-01
8.16730261e-02 9.25849736e-01 -4.49241817e-01 1.66040108e-01
-9.68934875e-03 2.40019739e-01 -1.12548436e-03 1.02427006e+00
1.02017057e+00 1.83776826e-01 6.49733171e-02 -2.47374058e-01
1.09495729e-01 -3.92684251e-01 -6.81686163e-01 1.49724531e+00
-7.34801769e-01 1.06361699e+00 4.38692003e-01 -9.28924024e-01
1.29240382e+00 3.12864989e-01 5.94451129e-01 -7.82660842e-01
4.41654831e-01 2.77641416e-01 -4.70668346e-01 -3.56069565e-01
1.00370087e-01 -4.09983516e-01 1.24107867e-01 3.58004898e-01
5.09689689e-01 -2.73726463e-01 -1.35550469e-01 -1.91418707e-01
1.01238513e+00 2.47743696e-01 1.82964485e-02 -2.58521050e-01
3.98422241e-01 -5.59543014e-01 3.67296726e-01 2.84564137e-01
-5.14533699e-01 5.61303675e-01 8.47267807e-01 -7.32690871e-01
-9.36224401e-01 -9.17485774e-01 -3.38954568e-01 1.26128411e+00
-3.55847657e-01 -2.65798718e-01 -8.89037728e-01 -8.10004056e-01
2.38703370e-01 -5.34859113e-02 -1.20259380e+00 3.04436628e-02
-3.31280917e-01 -5.62884152e-01 9.04066145e-01 5.06374300e-01
3.35627079e-01 -9.07515585e-01 -4.00266051e-01 -1.58549935e-01
3.09900939e-01 -1.35335338e+00 -9.15210042e-03 1.80995449e-01
-6.88065827e-01 -1.03570879e+00 -6.79491341e-01 -7.36162245e-01
9.55766618e-01 -2.19071820e-01 1.15650034e+00 3.57573479e-01
-5.00948787e-01 6.84378624e-01 -1.20425686e-01 -3.45133752e-01
-3.37657094e-01 -1.99971765e-01 1.19965211e-01 2.80051082e-01
6.24563575e-01 -8.06972444e-01 -5.54162800e-01 3.51186037e-01
-7.10278213e-01 -4.37715232e-01 4.14509207e-01 5.43035090e-01
4.31054890e-01 -4.52527136e-01 4.48409140e-01 -6.49163008e-01
4.77439642e-01 -1.38187602e-01 -5.55116951e-01 -1.19258864e-02
-4.48014587e-01 -1.60684615e-01 5.12619197e-01 -3.36924285e-01
-8.03779066e-01 3.93717736e-01 -5.88493347e-01 -6.47411764e-01
-5.07729352e-01 1.93597302e-02 -3.49392965e-02 -8.27095509e-01
7.60192752e-01 -3.09010714e-01 4.07772869e-01 -2.24362418e-01
3.74730736e-01 3.56830299e-01 3.41907263e-01 -5.59491754e-01
7.27594435e-01 9.12795842e-01 5.08673608e-01 -8.56103361e-01
-6.99478328e-01 -1.96520925e-01 -8.93755853e-01 -4.94182736e-01
8.69258642e-01 -8.31484973e-01 -1.12082505e+00 4.83695954e-01
-1.24916577e+00 -5.89552283e-01 -7.28524774e-02 3.94366868e-02
-6.55291617e-01 1.15306959e-01 -7.98880577e-01 -9.74887729e-01
-1.89805001e-01 -9.34070110e-01 1.51367843e+00 1.11647462e-02
-3.44456732e-01 -1.11094511e+00 -8.11383594e-03 -1.83527470e-02
4.71937001e-01 8.68000746e-01 5.73065162e-01 -2.88261294e-01
-1.17821991e-01 -3.91889840e-01 -4.43997681e-01 5.28583467e-01
1.41473830e-01 4.78575975e-01 -1.51618075e+00 -1.04269788e-01
-7.61428010e-03 -7.18431294e-01 8.37014556e-01 2.02723935e-01
1.03794920e+00 -1.61132157e-01 1.00456633e-01 1.02514815e+00
1.40738201e+00 -2.13036180e-01 7.14089751e-01 1.52992651e-01
5.28428674e-01 8.47192824e-01 1.86386004e-01 3.10470700e-01
-8.18311144e-03 7.39848733e-01 5.39917707e-01 -7.11142719e-01
-7.46342987e-02 2.18865764e-03 3.57622743e-01 2.46180147e-01
-3.68061155e-01 1.62297860e-01 -5.54814816e-01 2.17669189e-01
-1.41569066e+00 -7.95527518e-01 2.63877988e-01 1.80799234e+00
7.84501851e-01 -1.68739080e-01 1.61847115e-01 3.10585275e-02
2.24851519e-01 1.27795547e-01 -4.91471708e-01 -8.05253386e-01
-2.63948441e-01 7.49473035e-01 3.30086172e-01 4.29897547e-01
-1.12366760e+00 1.07620072e+00 6.94844723e+00 5.68308771e-01
-1.46788442e+00 5.12938201e-02 1.06425488e+00 -1.57670900e-01
-4.26258072e-02 -4.14982468e-01 -5.28590977e-01 -1.56125687e-02
8.65067780e-01 4.81298000e-01 2.80633479e-01 1.06545460e+00
1.43670753e-01 -3.94426696e-02 -1.24960947e+00 1.16665208e+00
2.46655241e-01 -1.04171360e+00 -1.56214118e-01 1.02569252e-01
6.13752007e-01 -2.05894381e-01 2.83531010e-01 2.30643883e-01
2.77727116e-02 -1.48963165e+00 5.93415916e-01 5.10035098e-01
1.14601922e+00 -7.89469361e-01 5.04243314e-01 -2.67248809e-01
-9.63942051e-01 2.30002582e-01 -3.98676097e-01 -1.32793590e-01
-7.63744637e-02 3.62715393e-01 -6.60852194e-01 -5.81634864e-02
6.14517391e-01 5.88558078e-01 -4.54289228e-01 3.76125336e-01
-2.86164761e-01 3.30346555e-01 -4.56389427e-01 1.06896982e-01
1.87654734e-01 -2.86828652e-02 7.42187202e-02 1.47417831e+00
2.57401854e-01 2.87071578e-02 -2.43101329e-01 1.00831652e+00
-4.51940566e-01 2.76769459e-01 -8.24771583e-01 2.04528108e-01
1.11571318e-02 1.83408606e+00 -7.13018417e-01 -5.22610024e-02
-3.51868510e-01 1.04852676e+00 5.79623282e-01 4.54493940e-01
-7.19866037e-01 -1.75784245e-01 1.04657233e+00 -3.73495300e-03
2.30390325e-01 -3.63978207e-01 -2.11263169e-03 -9.58735645e-01
-7.28221461e-02 -8.06119025e-01 -5.42887375e-02 -9.12555754e-01
-1.27303827e+00 7.95623779e-01 -3.48471880e-01 -6.68276131e-01
-2.20704421e-01 -1.20850372e+00 -5.38628459e-01 7.26829052e-01
-1.72509277e+00 -1.15740871e+00 -5.68272412e-01 6.61835670e-01
9.12384540e-02 7.33840615e-02 1.14315879e+00 2.64746964e-01
-4.50687528e-01 9.40262020e-01 -4.26009029e-01 4.45888996e-01
7.95898020e-01 -1.02950823e+00 2.42652595e-01 3.79144222e-01
3.75517160e-01 4.51833487e-01 4.24282223e-01 6.18147105e-02
-1.37403405e+00 -8.83159041e-01 5.15234053e-01 -5.83866477e-01
5.07299244e-01 -7.82131612e-01 -7.81776905e-01 6.22987509e-01
-7.64768943e-02 5.07065833e-01 6.77188814e-01 2.17358813e-01
-8.19161594e-01 -3.17305475e-01 -1.45830762e+00 3.63581002e-01
1.08834958e+00 -8.65217686e-01 2.01651156e-02 2.50447512e-01
2.37408414e-01 -3.53096485e-01 -9.89370763e-01 5.15971839e-01
9.73170578e-01 -1.26565707e+00 8.12826455e-01 -6.90562308e-01
4.73700374e-01 1.88116699e-01 -2.65989363e-01 -1.03000546e+00
-2.52591223e-02 -6.86546504e-01 1.13275535e-01 1.05980146e+00
4.22364742e-01 -7.37342834e-01 1.00746310e+00 8.06493819e-01
2.20151499e-01 -9.35802162e-01 -1.03660893e+00 -6.65150285e-01
8.26169550e-02 -5.01474798e-01 5.24853408e-01 9.31312382e-01
-2.62587041e-01 1.00812735e-02 -1.08536109e-01 -7.58485198e-02
5.09080231e-01 -2.41752699e-01 9.06964779e-01 -1.28587139e+00
1.16780832e-01 -7.12973118e-01 -1.01213324e+00 -7.23419726e-01
7.92243540e-01 -7.41340101e-01 -1.39567479e-02 -8.50760758e-01
-9.58524719e-02 -3.80206436e-01 -3.09625238e-01 8.47570240e-01
3.39816511e-01 1.05732715e+00 1.76002264e-01 -1.16206601e-01
-3.91020358e-01 4.00004685e-01 1.01998270e+00 1.26209604e-02
1.10336222e-01 -4.20146674e-01 -3.92398655e-01 9.41109061e-01
7.99884021e-01 -2.26501673e-01 -8.21711496e-02 -1.53564051e-01
1.93455294e-01 -2.46867478e-01 6.74737930e-01 -8.31698000e-01
-1.89897060e-01 1.24927200e-01 8.56949091e-01 5.03601432e-02
7.31981099e-01 -8.05222690e-01 -1.58660367e-01 -4.11635414e-02
-2.88325876e-01 -8.66490304e-02 5.99220097e-01 7.80232400e-02
-8.92518014e-02 2.11993065e-02 9.89276111e-01 -1.69707224e-01
-6.99570596e-01 5.81374943e-01 -2.88308859e-01 -1.38985321e-01
6.53234065e-01 -3.56622756e-01 -1.74562428e-02 -6.19417250e-01
-8.46414089e-01 -3.44115347e-01 6.36358380e-01 1.28370121e-01
6.57268345e-01 -1.33592570e+00 -5.82387865e-01 4.39879537e-01
-3.09923813e-02 -3.39002073e-01 -8.80404785e-02 1.05658638e+00
-6.12821460e-01 -1.74505953e-02 -5.06143391e-01 -6.56305909e-01
-1.47774255e+00 2.90821940e-01 8.69354188e-01 1.72659442e-01
-1.19536743e-01 1.08267415e+00 2.38765761e-01 -4.23813939e-01
2.39201561e-01 -2.04880401e-01 1.03871346e-01 2.27283970e-01
4.69826579e-01 -1.77388657e-02 2.35977948e-01 -9.24538136e-01
-4.52559829e-01 9.43789601e-01 2.12281078e-01 -1.83113471e-01
1.43042648e+00 1.10697702e-01 -3.60124677e-01 1.25706270e-01
1.95687413e+00 2.23642178e-02 -1.38682950e+00 1.93598315e-01
-1.85542479e-01 -4.55966115e-01 2.84122843e-02 -5.78947663e-01
-1.39462376e+00 1.03336513e+00 7.93155670e-01 1.47571594e-01
1.40644014e+00 5.74590378e-02 5.55866659e-01 2.72680998e-01
2.13376999e-01 -1.00098193e+00 2.05635116e-01 2.47024164e-01
8.97475719e-01 -1.44970465e+00 -7.75644630e-02 -2.31201649e-01
-2.27342978e-01 1.45871842e+00 4.96283025e-01 -2.26261705e-01
7.14132726e-01 5.26740611e-01 5.76263547e-01 -5.51254749e-01
-4.94333297e-01 -3.09917927e-01 3.45990211e-01 7.20609426e-01
5.26967049e-01 -1.62131384e-01 2.12733656e-01 6.79022521e-02
-1.28969923e-01 -9.99452919e-02 1.57395825e-01 5.70608556e-01
-2.42231607e-01 -9.26788211e-01 -1.31325781e-01 1.28467888e-01
-6.26013041e-01 1.05820641e-01 -7.65297174e-01 1.14812005e+00
1.21286415e-01 6.21633708e-01 2.05241039e-01 -3.53781521e-01
3.04757178e-01 1.90403953e-01 8.84236515e-01 -2.10385069e-01
-3.51954401e-01 -1.93910748e-01 4.00968343e-02 -9.84361410e-01
-7.26975203e-01 -4.34590340e-01 -1.04392195e+00 -2.39536434e-01
-1.26531810e-01 -2.62854487e-01 9.12836254e-01 7.06535637e-01
3.68647873e-01 -1.55736119e-01 9.04653847e-01 -1.35585880e+00
-2.27363884e-01 -7.06857264e-01 -6.41210973e-01 6.47939742e-01
4.22980905e-01 -8.24505568e-01 -7.01710224e-01 -4.54914011e-02]
|
[13.470171928405762, 1.331436038017273]
|
cfee0317-33c5-46a9-9d5e-50440e27d482
|
dagformer-directed-acyclic-graph-transformer
|
2210.13148
| null |
https://arxiv.org/abs/2210.13148v5
|
https://arxiv.org/pdf/2210.13148v5.pdf
|
Transformers over Directed Acyclic Graphs
|
Transformer models have recently gained popularity in graph representation learning as they have the potential to learn complex relationships beyond the ones captured by regular graph neural networks. The main research question is how to inject the structural bias of graphs into the transformer architecture, and several proposals have been made for undirected molecular graphs and, recently, also for larger network graphs. In this paper, we study transformers over directed acyclic graphs (DAGs) and propose architecture adaptations tailored to DAGs: (1) An attention mechanism that is considerably more efficient than the regular quadratic complexity of transformers and at the same time faithfully captures the DAG structure, and (2) a positional encoding of the DAG's partial order, complementing the former. We rigorously evaluate our approach over various types of tasks, ranging from classifying source code graphs to nodes in citation networks, and show that it is effective in two important aspects: in making graph transformers generally outperform graph neural networks tailored to DAGs and in improving SOTA graph transformer performance in terms of both quality and efficiency.
|
['Lei Shi', 'Veronika Thost', 'Yuankai Luo']
|
2022-10-24
| null | null | null | null |
['graph-property-prediction']
|
['graphs']
|
[ 3.20953399e-01 5.66360712e-01 -2.82433212e-01 -8.69564414e-02
-1.22322112e-01 -7.44716465e-01 6.20778382e-01 7.29642332e-01
-2.49283239e-02 5.41588545e-01 3.25275064e-01 -8.81671965e-01
-5.17883658e-01 -1.17586231e+00 -8.54822814e-01 -5.88679552e-01
-4.71841305e-01 8.54107141e-01 4.16750014e-01 -3.59954029e-01
2.10194021e-01 7.55164146e-01 -1.18870616e+00 -5.43261729e-02
7.77107537e-01 7.36582220e-01 -5.63775413e-02 5.85675478e-01
-3.05128306e-01 1.16833031e+00 -2.30517194e-01 -7.00234592e-01
-8.45644474e-02 -3.32028180e-01 -9.86249745e-01 -9.04111415e-02
5.94479620e-01 2.27017641e-01 -6.57776177e-01 1.03549004e+00
3.60971481e-01 -3.62498201e-02 6.91294611e-01 -9.83187854e-01
-8.59727144e-01 9.85855341e-01 -4.04652506e-01 4.21396732e-01
2.70196974e-01 -3.29010449e-02 1.55829525e+00 -2.99396098e-01
6.32223606e-01 1.35928142e+00 8.86108100e-01 2.53394455e-01
-1.42015493e+00 -4.60959792e-01 3.06163073e-01 1.56930983e-01
-9.53942895e-01 -2.68331796e-01 9.35855627e-01 -4.76838976e-01
1.21043241e+00 1.41942054e-01 7.14787006e-01 9.98264492e-01
4.72791791e-01 3.59997064e-01 8.97896111e-01 -3.98286164e-01
1.00678869e-01 -2.61625111e-01 4.98367310e-01 1.10627747e+00
6.78366184e-01 1.34641364e-01 -3.45583797e-01 -1.47278070e-01
7.28865743e-01 -2.28666887e-01 -2.44514182e-01 -7.11390793e-01
-9.09086108e-01 9.90116060e-01 8.18818629e-01 5.41072667e-01
-2.36757785e-01 5.63811541e-01 4.82928813e-01 5.44178307e-01
4.31430459e-01 8.32363665e-01 -2.79262900e-01 2.69375920e-01
-4.57174122e-01 1.65952846e-01 9.85411704e-01 8.42345238e-01
7.17510700e-01 1.85229957e-01 -2.87714656e-02 5.56190550e-01
1.15162492e-01 8.62791762e-02 3.54685873e-01 -4.12836909e-01
4.99692976e-01 9.38574016e-01 -5.49532473e-01 -1.21539474e+00
-6.78072214e-01 -7.11272836e-01 -9.98893023e-01 -1.75706923e-01
4.85666305e-01 3.15108895e-01 -9.50707793e-01 1.79609060e+00
-1.20330051e-01 -1.36392906e-01 -2.86434978e-01 2.89178312e-01
9.15607572e-01 3.87440503e-01 1.03522643e-01 5.82252890e-02
1.33812702e+00 -6.15274847e-01 -3.99508417e-01 -4.62526113e-01
6.79527342e-01 -1.89719632e-01 8.23625445e-01 7.15026036e-02
-1.08271790e+00 -3.34978819e-01 -1.02289391e+00 -1.73146039e-01
-5.79740226e-01 -3.07579368e-01 1.19577610e+00 6.30664885e-01
-1.59619915e+00 9.85846162e-01 -6.45403504e-01 -4.83271480e-01
4.03865248e-01 7.19614208e-01 -4.29579616e-01 -6.64930195e-02
-1.22313094e+00 8.01346481e-01 4.21026617e-01 -5.34221865e-02
-7.52673328e-01 -5.55922449e-01 -1.07666647e+00 5.98240018e-01
3.42695624e-01 -7.60261178e-01 7.84443676e-01 -1.03820038e+00
-1.14228737e+00 9.31213975e-01 7.15817884e-02 -5.94189167e-01
1.69402599e-01 4.81071413e-01 -1.52923748e-01 -3.91853005e-02
-1.43916994e-01 4.30741400e-01 5.80588579e-01 -7.67914474e-01
-5.78759797e-02 -5.75477540e-01 5.02599239e-01 -9.58957821e-02
-5.54590523e-01 -2.59066492e-01 -1.62873238e-01 -6.77178979e-01
4.30824468e-03 -1.10968375e+00 -3.24309647e-01 -3.33847255e-01
-5.70025206e-01 -5.34672618e-01 3.22509259e-01 -4.74495828e-01
1.27453542e+00 -1.74135983e+00 5.84666431e-01 4.70894665e-01
9.11130905e-01 1.78781345e-01 -2.50171781e-01 5.21436214e-01
-5.28265417e-01 2.39955500e-01 -1.73214287e-01 7.28975534e-02
-4.03599665e-02 2.65656114e-01 -1.23008721e-01 2.95293748e-01
3.43728215e-01 1.30576479e+00 -1.02277327e+00 -2.12193549e-01
-1.58299923e-01 2.79615790e-01 -7.28373766e-01 5.85919023e-02
-3.63044262e-01 8.30392167e-02 -4.94499922e-01 3.97482574e-01
3.30958128e-01 -7.38857687e-01 7.10289061e-01 -2.16544107e-01
1.68282837e-01 6.33255005e-01 -7.19397962e-01 1.28550613e+00
-1.27748042e-01 6.98769033e-01 -9.15451348e-02 -1.31698966e+00
8.96737278e-01 1.03768006e-01 3.86248112e-01 -7.54326582e-01
-4.00417522e-02 4.03016508e-02 4.06848758e-01 -1.42539576e-01
4.15536374e-01 -2.13140577e-01 3.74786258e-02 3.98630083e-01
3.95358592e-01 4.16844189e-02 4.12356764e-01 3.99256378e-01
1.44048512e+00 -1.78476468e-01 3.76017749e-01 -6.17985487e-01
3.80605072e-01 -2.87576109e-01 1.42418832e-01 7.13757038e-01
1.65096909e-01 1.82225499e-02 1.25140727e+00 -5.95407963e-01
-9.52116430e-01 -9.31164801e-01 2.39305660e-01 1.16308951e+00
-1.70699179e-01 -6.30505025e-01 -5.27529657e-01 -6.58661187e-01
1.66372895e-01 2.45087594e-01 -9.29810166e-01 -5.70427895e-01
-6.18310809e-01 -7.33682156e-01 5.56359112e-01 6.53837621e-01
1.62433162e-02 -9.25080478e-01 -1.12638548e-01 2.01725781e-01
1.45771161e-01 -9.32611704e-01 -4.48155731e-01 6.80195630e-01
-1.14444709e+00 -1.38300240e+00 -3.86751503e-01 -8.04647207e-01
6.65780663e-01 1.96148202e-01 1.44662213e+00 3.62740964e-01
-2.00868994e-02 3.12103599e-01 -1.56965688e-01 -2.27567568e-01
-5.15904784e-01 5.75509429e-01 -2.45798975e-01 -1.85378775e-01
1.32787973e-01 -9.75549638e-01 -2.83238664e-02 -1.79811213e-02
-9.04910922e-01 -1.15228757e-01 6.34721279e-01 8.10057282e-01
1.24963269e-01 6.44505620e-02 4.26726103e-01 -1.36815894e+00
8.12123477e-01 -4.78894144e-01 -7.11813152e-01 2.58588165e-01
-9.12936330e-01 5.95453620e-01 7.88129628e-01 -1.91345662e-01
-4.41451252e-01 -1.48302719e-01 -2.64614046e-01 -2.11794615e-01
3.49716127e-01 8.07888389e-01 -1.22421965e-01 -4.70925331e-01
7.01689184e-01 1.53331727e-01 7.95117021e-03 -4.31205392e-01
3.22353542e-01 -8.94676298e-02 1.87425777e-01 -8.15312862e-01
8.40715706e-01 1.11687817e-01 6.14499450e-01 -7.12347686e-01
-4.75293070e-01 -5.70998788e-02 -6.76464200e-01 7.84616768e-02
7.23354161e-01 -5.34670472e-01 -7.82936752e-01 2.36232609e-01
-1.01035929e+00 -3.05828899e-01 -1.01573907e-01 5.78251071e-02
-4.06699419e-01 4.73878801e-01 -1.01598358e+00 -3.59253198e-01
-2.34223083e-01 -1.14784431e+00 7.82260656e-01 -5.09462282e-02
-1.93890464e-02 -1.47979271e+00 1.68416664e-01 -5.80643713e-02
6.02577448e-01 3.93920779e-01 1.81783307e+00 -7.02366948e-01
-6.96357191e-01 -4.47610654e-02 -4.68325555e-01 4.16227914e-02
2.33135037e-02 -1.21613771e-01 -7.79047132e-01 -5.08848846e-01
-5.56349993e-01 -8.61468464e-02 1.14646244e+00 3.42714190e-01
1.10564840e+00 -4.35456067e-01 -4.80263561e-01 5.98693311e-01
1.32212520e+00 3.30420732e-02 5.89534760e-01 7.65978321e-02
9.94251132e-01 5.95692754e-01 -4.39455569e-01 -1.20040894e-01
4.88298953e-01 6.53368592e-01 7.47530878e-01 -1.09244451e-01
-3.34486574e-01 -5.42606711e-01 2.98745096e-01 1.00690842e+00
-4.15658742e-01 -4.41689253e-01 -9.35240388e-01 3.89895171e-01
-1.65028071e+00 -8.39959919e-01 -2.31179789e-01 2.05228233e+00
4.95838165e-01 4.11448151e-01 2.62442708e-01 1.56310052e-01
5.85794330e-01 4.02554840e-01 -4.88980711e-01 -5.68051398e-01
-1.40682325e-01 5.58532953e-01 5.80411494e-01 4.76558626e-01
-8.77216518e-01 7.96679080e-01 6.93741465e+00 4.21557039e-01
-1.11599493e+00 -7.40040615e-02 5.31526268e-01 4.70832676e-01
-6.91841722e-01 1.34217441e-01 -3.06760758e-01 1.71110213e-01
1.04645598e+00 -2.63845921e-01 7.55541623e-01 6.33986413e-01
-3.58759135e-01 6.13078356e-01 -1.37691903e+00 7.15524316e-01
-1.52106047e-01 -1.62765479e+00 4.59835261e-01 5.06412387e-01
3.70453209e-01 8.26101154e-02 -7.59566575e-02 3.81040245e-01
5.94516933e-01 -1.37205338e+00 5.40993571e-01 3.00749630e-01
6.92149997e-01 -5.85972667e-01 6.06872082e-01 4.74339463e-02
-1.45887649e+00 -1.47573113e-01 -4.11270350e-01 -1.67966411e-01
-3.32385749e-01 5.31962335e-01 -7.51443803e-01 7.48331547e-01
3.14061731e-01 9.81169939e-01 -9.53059375e-01 7.92779148e-01
-2.28432670e-01 6.34551167e-01 1.15733147e-02 -1.17691450e-01
5.12915909e-01 -3.04593146e-01 5.42680323e-01 1.28304362e+00
9.86938477e-02 -2.49519423e-01 4.23135161e-02 8.73960555e-01
-5.45449615e-01 -6.61999136e-02 -9.79731143e-01 -5.92478037e-01
3.32144946e-01 1.08593678e+00 -9.21839833e-01 -1.19076587e-01
-2.51765162e-01 5.32192647e-01 9.17703569e-01 2.65373856e-01
-5.79692781e-01 -4.48018640e-01 4.58360106e-01 3.73031139e-01
4.95144546e-01 -1.71936035e-01 1.33565292e-01 -1.02818274e+00
-1.43086314e-01 -8.68930936e-01 6.43105805e-01 -6.39112830e-01
-1.23893011e+00 8.23370874e-01 -1.90996379e-01 -5.31205356e-01
-1.32295430e-01 -1.10150731e+00 -6.40507579e-01 6.32620811e-01
-1.56366289e+00 -1.16194236e+00 -1.26637429e-01 5.86700201e-01
-9.45076793e-02 -4.56280895e-02 8.98803592e-01 4.77013141e-01
-5.06914854e-01 5.30671060e-01 6.37850258e-03 1.18815280e-01
1.61936268e-01 -1.59904647e+00 6.85875952e-01 6.57342136e-01
3.99488658e-01 8.16649139e-01 6.63625300e-01 -4.30752546e-01
-1.89010823e+00 -1.00389922e+00 9.46067810e-01 -5.38185060e-01
9.23547566e-01 -5.85703969e-01 -1.00595009e+00 9.70417261e-01
1.26647204e-02 -7.57673662e-03 3.26967984e-01 5.30874610e-01
-8.09179485e-01 -1.01065204e-01 -8.11385214e-01 4.03917223e-01
1.45857608e+00 -7.49504924e-01 -3.56313199e-01 4.64176893e-01
7.23107219e-01 -2.07095280e-01 -9.72520113e-01 2.74642169e-01
4.13554847e-01 -9.61546361e-01 1.02535534e+00 -1.13248241e+00
4.27146554e-01 -1.31857162e-02 3.80117260e-02 -1.42228651e+00
-8.80037308e-01 -6.68702900e-01 -2.37402648e-01 9.90849376e-01
3.96076828e-01 -9.40783381e-01 7.88366795e-01 -5.05958311e-03
-1.36473283e-01 -8.19987297e-01 -9.11044836e-01 -6.83664560e-01
2.04703391e-01 -8.99149776e-02 5.81223190e-01 1.07203865e+00
4.03158646e-03 9.68958795e-01 -1.63785011e-01 -1.09261803e-01
5.08007288e-01 1.27465725e-01 4.01024610e-01 -1.67316151e+00
-4.31223452e-01 -9.66895223e-01 -7.54595995e-01 -1.05996907e+00
4.17212248e-01 -1.44335687e+00 -4.32839990e-01 -1.59628129e+00
2.44237512e-01 -4.01228935e-01 -3.78870279e-01 5.51387846e-01
3.11046336e-02 -7.95713440e-02 3.28351222e-02 -5.51798381e-02
-4.28135097e-01 2.85886526e-01 1.11256158e+00 -3.67941856e-01
2.97933102e-01 -9.14732739e-02 -1.17885602e+00 5.00888467e-01
4.34971869e-01 -4.45293009e-01 -5.03731251e-01 -3.79949033e-01
6.37084067e-01 7.77850226e-02 3.19750726e-01 -7.04303265e-01
2.10872427e-01 1.45132467e-01 1.30690381e-01 -1.56302117e-02
8.70994627e-02 -6.08891249e-01 2.10403472e-01 5.63634872e-01
-4.14858550e-01 4.62973058e-01 1.46387637e-01 8.39929998e-01
6.09655306e-02 -3.92371714e-02 7.40028441e-01 -3.78630400e-01
-3.32165688e-01 5.09952188e-01 -3.27563614e-01 7.88335800e-02
4.98610646e-01 -1.34376466e-01 -5.22502065e-01 -3.87971520e-01
-5.90983212e-01 6.42149225e-02 4.32373196e-01 3.78431380e-01
2.98908561e-01 -1.18341315e+00 -4.33915377e-01 1.98335014e-02
1.34697899e-01 -3.91315430e-01 -3.14523458e-01 8.04816186e-01
-4.52236801e-01 7.10357547e-01 -1.53560013e-01 -3.35734963e-01
-1.11502075e+00 8.43952298e-01 5.43683290e-01 -9.05674934e-01
-5.27715743e-01 7.20151603e-01 4.26645309e-01 -4.80820715e-01
2.10799918e-01 -6.78743660e-01 -3.09461147e-01 5.59339002e-02
9.15470272e-02 2.08361953e-01 3.36502969e-01 -4.92159396e-01
-4.14820999e-01 6.26076043e-01 -6.24566302e-02 7.19096780e-01
1.49632692e+00 3.34610134e-01 -5.59410989e-01 2.60833025e-01
9.95392203e-01 5.04426882e-02 -7.28836179e-01 -2.41695434e-01
4.09587085e-01 -4.29888591e-02 -3.57578434e-02 -6.07487261e-01
-1.21996546e+00 9.23579335e-01 1.47502199e-01 8.09532583e-01
8.43958974e-01 1.89588651e-01 5.17383158e-01 5.31808555e-01
3.83252978e-01 -3.46185476e-01 1.14459485e-01 6.42016828e-01
6.96999550e-01 -7.45702207e-01 3.04648075e-02 -4.68951017e-01
-1.03174500e-01 1.38858950e+00 2.37888739e-01 -3.06349665e-01
5.57742834e-01 1.68930382e-01 -6.05808616e-01 -7.94922590e-01
-8.13354254e-01 -2.08489373e-01 4.46766853e-01 6.38951838e-01
6.96543157e-01 1.19925678e-01 -1.08557485e-01 2.16466695e-01
-2.63914485e-02 -2.06251279e-01 4.84988898e-01 5.00914335e-01
-3.40041012e-01 -1.17391062e+00 1.08524337e-01 8.45158994e-01
-3.57061356e-01 -2.78714925e-01 -8.05278420e-01 8.37087393e-01
-2.16178805e-01 6.90271914e-01 -1.33420855e-01 -4.11231577e-01
3.98368806e-01 -7.04942644e-02 8.78780603e-01 -5.59362471e-01
-6.63400471e-01 -5.15711129e-01 4.28710490e-01 -4.88821745e-01
-2.19415024e-01 -3.77407491e-01 -8.03747177e-01 -5.91433048e-01
-3.80744934e-01 4.95322585e-01 2.72672534e-01 8.21872592e-01
4.34491247e-01 8.89678538e-01 3.40880305e-01 -7.71823943e-01
-5.08623779e-01 -8.64683151e-01 -5.58685303e-01 2.44494379e-01
3.92926514e-01 -7.68114030e-01 -3.12285602e-01 -4.20508057e-01]
|
[6.926620960235596, 6.22547721862793]
|
3d09fcfb-df22-42a1-99e5-41fafd59aede
|
togethernet-bridging-image-restoration-and
|
2209.01373
| null |
https://arxiv.org/abs/2209.01373v1
|
https://arxiv.org/pdf/2209.01373v1.pdf
|
TogetherNet: Bridging Image Restoration and Object Detection Together via Dynamic Enhancement Learning
|
Adverse weather conditions such as haze, rain, and snow often impair the quality of captured images, causing detection networks trained on normal images to generalize poorly in these scenarios. In this paper, we raise an intriguing question - if the combination of image restoration and object detection, can boost the performance of cutting-edge detectors in adverse weather conditions. To answer it, we propose an effective yet unified detection paradigm that bridges these two subtasks together via dynamic enhancement learning to discern objects in adverse weather conditions, called TogetherNet. Different from existing efforts that intuitively apply image dehazing/deraining as a pre-processing step, TogetherNet considers a multi-task joint learning problem. Following the joint learning scheme, clean features produced by the restoration network can be shared to learn better object detection in the detection network, thus helping TogetherNet enhance the detection capacity in adverse weather conditions. Besides the joint learning architecture, we design a new Dynamic Transformer Feature Enhancement module to improve the feature extraction and representation capabilities of TogetherNet. Extensive experiments on both synthetic and real-world datasets demonstrate that our TogetherNet outperforms the state-of-the-art detection approaches by a large margin both quantitatively and qualitatively. Source code is available at https://github.com/yz-wang/TogetherNet.
|
['Mingqiang Wei', 'Fu Lee Wang', 'Haoran Xie', 'Lina Gong', 'Kaiwen Zhang', 'Xuefeng Yan', 'Yongzhen Wang']
|
2022-09-03
| null | null | null | null |
['image-dehazing']
|
['computer-vision']
|
[ 3.84922117e-01 -4.30898696e-01 1.41533598e-01 -1.88059047e-01
-5.01121938e-01 -5.21620214e-01 5.50723970e-01 -8.10798630e-02
-4.03468817e-01 4.58413482e-01 -3.63311879e-02 -2.27434799e-01
4.35019098e-02 -8.28976452e-01 -8.28477144e-01 -9.96251345e-01
2.72969659e-02 -3.49905610e-01 6.26391649e-01 -3.65087360e-01
1.74656048e-01 5.01923680e-01 -1.68295884e+00 1.97215050e-01
1.17345226e+00 8.58294964e-01 4.91570264e-01 7.87171006e-01
2.97349930e-01 7.33639479e-01 -5.99546492e-01 -1.13740414e-01
6.20949686e-01 -1.42706126e-01 -1.92188710e-01 3.29029649e-01
7.81147420e-01 -6.89514041e-01 -4.56391126e-01 1.25121522e+00
5.62878191e-01 1.57613456e-01 4.29143935e-01 -1.15400326e+00
-7.04333901e-01 1.08635813e-01 -8.93444538e-01 6.95605159e-01
1.27484128e-01 5.76170325e-01 8.55720818e-01 -1.07690132e+00
1.05773985e-01 1.00983274e+00 8.21251750e-01 3.73605937e-01
-8.81445885e-01 -8.96848738e-01 3.23137522e-01 3.58386755e-01
-1.30805683e+00 -4.61030275e-01 5.41112542e-01 -2.90969759e-01
6.64392650e-01 2.46893287e-01 4.67829347e-01 6.98400855e-01
1.06471576e-01 9.55256760e-01 1.07524443e+00 -2.62518287e-01
-1.39932916e-01 2.86525711e-02 1.31542102e-01 7.41887391e-01
6.09528959e-01 5.40617704e-01 -4.01534975e-01 1.58507884e-01
6.60676479e-01 3.52347851e-01 -6.33455575e-01 -9.59630832e-02
-8.33394885e-01 6.58862710e-01 9.42277908e-01 6.10290319e-02
-3.94162416e-01 5.29781990e-02 4.21755342e-03 3.04758191e-01
6.80334270e-01 4.00480241e-01 -3.41785699e-01 6.03776097e-01
-8.96609545e-01 2.36051917e-01 3.96544605e-01 4.92547035e-01
1.08578873e+00 1.24312915e-01 -2.91977763e-01 7.03383803e-01
2.22372338e-01 8.26456070e-01 -8.47823471e-02 -6.26522720e-01
2.55025268e-01 6.79710627e-01 2.38202944e-01 -8.18530083e-01
-2.93760359e-01 -8.91869545e-01 -7.91193664e-01 5.89802921e-01
3.30063581e-01 -3.16831499e-01 -1.18456280e+00 1.33658922e+00
4.62974638e-01 7.53863335e-01 9.31055173e-02 1.19280791e+00
8.14051270e-01 8.63490820e-01 -8.96226913e-02 -5.44594079e-02
1.43273735e+00 -1.16608191e+00 -5.40411353e-01 -6.69866383e-01
2.90526569e-01 -8.68467629e-01 7.66822696e-01 3.34885329e-01
-8.27066720e-01 -7.18058407e-01 -1.20983672e+00 9.60703567e-02
-4.18975115e-01 3.59618962e-01 5.18218577e-01 5.01430809e-01
-9.96345580e-01 2.58570105e-01 -6.16310120e-01 -3.45810235e-01
4.75730598e-01 2.41622198e-02 -2.35286340e-01 -5.78916013e-01
-1.12596738e+00 1.05777740e+00 2.40220219e-01 4.08010602e-01
-1.32612109e+00 -7.49201059e-01 -8.20575118e-01 2.22074613e-02
6.33061051e-01 -6.63569033e-01 1.03592432e+00 -7.37239778e-01
-8.86708379e-01 5.51582754e-01 -1.30964965e-01 -5.43547392e-01
3.72532099e-01 -6.01704240e-01 -4.26443100e-01 1.43538550e-01
1.29562095e-02 5.32371640e-01 1.08669090e+00 -1.49433625e+00
-1.06542778e+00 -2.26290688e-01 2.48647779e-01 3.17302078e-01
-3.44143689e-01 8.53201747e-02 -4.49932814e-01 -6.41526043e-01
-2.24128872e-01 -6.64951146e-01 -4.17620301e-01 3.25263709e-01
-1.86386824e-01 -7.99180418e-02 1.11143291e+00 -6.73983991e-01
9.49455321e-01 -2.18010139e+00 -2.24638104e-01 -1.71127409e-01
3.41534168e-01 8.02732289e-01 -3.78388673e-01 1.31420895e-01
1.05471186e-01 -1.17665626e-01 -4.81864601e-01 -2.74813384e-01
-4.25505012e-01 2.40560070e-01 -3.44838262e-01 7.94910431e-01
5.98724365e-01 7.07416654e-01 -9.73221719e-01 -2.43179426e-01
3.87460113e-01 5.76331019e-01 -1.64676979e-01 4.42101538e-01
3.94751094e-02 2.19299287e-01 -4.13002700e-01 8.89473259e-01
1.08263087e+00 1.71015374e-02 -4.52993453e-01 -1.67433262e-01
-2.63956904e-01 -4.08007652e-02 -1.17420101e+00 1.11906910e+00
-3.77790511e-01 8.14828217e-01 3.46418113e-01 -9.70246077e-01
9.54898417e-01 7.36494139e-02 -7.52526969e-02 -8.84660363e-01
-4.96603213e-02 1.66995078e-01 -3.70683260e-02 -6.99778736e-01
4.34362590e-01 -6.59158602e-02 2.74312735e-01 7.42852986e-02
-3.15768905e-02 -4.07097079e-02 2.15602875e-01 1.62553355e-01
1.03201056e+00 -8.82533118e-02 2.30421424e-01 -9.47413326e-04
5.44836223e-01 -7.55527988e-02 7.78076351e-01 1.09859097e+00
-3.41469258e-01 6.32397056e-01 -7.54247680e-02 -4.50604379e-01
-8.20452392e-01 -1.18650877e+00 -2.58905470e-01 1.24009037e+00
5.74329376e-01 1.22099491e-02 -4.86077160e-01 -7.27540672e-01
7.18410835e-02 3.62322241e-01 -7.43583500e-01 -3.48601997e-01
-4.02312011e-01 -1.26998210e+00 5.73813736e-01 5.63859522e-01
8.36666882e-01 -8.64461541e-01 -6.84049964e-01 2.74594445e-02
-1.41023338e-01 -1.19848657e+00 -3.33843231e-01 1.73720703e-01
-5.69897652e-01 -1.20080221e+00 -7.00530052e-01 -6.68461382e-01
5.32070339e-01 1.14083052e+00 8.56270790e-01 5.84135890e-01
-6.46917224e-01 1.81054190e-01 -6.19300067e-01 -7.66677797e-01
-1.17987484e-01 -1.93406373e-01 -1.42031506e-01 2.16730237e-01
9.07003060e-02 -4.95936602e-01 -8.98390949e-01 3.06356490e-01
-1.13569796e+00 -1.43398210e-01 1.02011335e+00 7.92223692e-01
1.48553401e-01 1.23845361e-01 3.55081469e-01 -5.20959973e-01
2.91468024e-01 -4.27268952e-01 -7.30427504e-01 1.99624449e-01
-5.57971239e-01 -2.15256929e-01 2.32791051e-01 -2.46520534e-01
-1.24988973e+00 1.24492131e-01 -3.58264223e-02 -2.68726498e-01
-3.61035913e-01 2.00809285e-01 -5.80342375e-02 -4.13390875e-01
8.00465524e-01 4.94157761e-01 -6.52528852e-02 -4.70978290e-01
2.45950803e-01 6.30590379e-01 8.46790731e-01 5.23422137e-02
1.45207286e+00 7.95368910e-01 -1.46526992e-01 -8.08418870e-01
-1.23402536e+00 -9.40626800e-01 -3.12116891e-01 -3.69613767e-01
7.70133555e-01 -1.44990599e+00 -2.05340534e-01 7.51632154e-01
-1.01923120e+00 -3.90066385e-01 7.47869909e-02 2.82121032e-01
1.32377133e-01 3.75404269e-01 -3.89775485e-01 -9.91679907e-01
-4.45118785e-01 -7.93881059e-01 1.12753391e+00 6.55173719e-01
6.34382188e-01 -8.34890306e-01 -1.07167780e-01 3.89642835e-01
5.37998021e-01 1.21963367e-01 1.69382066e-01 -3.27739656e-01
-6.99470639e-01 -2.66709805e-01 -6.28020346e-01 6.56961560e-01
1.08013362e-01 -3.99322733e-02 -1.12027895e+00 -4.55286026e-01
-3.88958752e-02 -1.18758574e-01 1.62245131e+00 2.64702499e-01
8.03561926e-01 -2.64839921e-02 -3.89639258e-01 8.44841838e-01
1.58992815e+00 -3.85799929e-02 8.35899711e-01 4.45573717e-01
7.01860428e-01 5.08427203e-01 8.15481365e-01 3.85148913e-01
2.81579942e-01 4.60246712e-01 8.39749336e-01 -6.67363286e-01
-5.18356025e-01 2.14951858e-01 5.54321647e-01 -1.74167787e-03
-1.00583524e-01 -3.61350983e-01 -8.02227259e-01 7.13311255e-01
-1.78978670e+00 -9.85281646e-01 -2.67897785e-01 1.93926573e+00
5.20092726e-01 8.02532211e-02 -1.42131329e-01 -1.97223406e-02
8.08060467e-01 3.71317744e-01 -4.55223441e-01 1.92376897e-01
-4.17152822e-01 -2.13103648e-02 7.56913602e-01 4.94806230e-01
-1.47060275e+00 8.63525391e-01 5.55212069e+00 7.07469821e-01
-1.20548499e+00 1.87111720e-01 4.69987780e-01 9.06946696e-03
1.23407297e-01 -3.37065198e-02 -9.62402344e-01 2.97574103e-01
4.61340070e-01 8.26666281e-02 2.40541130e-01 5.96877992e-01
4.15834486e-01 -2.45094076e-01 -4.97767895e-01 7.05938399e-01
2.31695741e-01 -1.26930285e+00 -5.24463765e-02 -7.97599480e-02
8.18965375e-01 3.72425407e-01 1.01875179e-01 2.71139979e-01
2.61839777e-01 -7.95524061e-01 5.77986181e-01 4.41837519e-01
2.77754933e-01 -4.51727837e-01 8.82391155e-01 4.31118965e-01
-1.39011538e+00 -4.01490867e-01 -3.77713591e-01 -3.43863994e-01
1.85975417e-01 9.71494675e-01 -7.71809280e-01 6.66251957e-01
9.67193902e-01 8.39744925e-01 -9.34502363e-01 1.68594027e+00
-6.55758142e-01 7.60424137e-01 -2.30371326e-01 3.88194114e-01
3.25941890e-01 -7.46498704e-02 7.76091635e-01 1.46337867e+00
1.32854730e-01 1.27238274e-01 4.27042454e-01 6.67934477e-01
1.03004329e-01 -4.00365204e-01 -4.75141704e-01 3.86397719e-01
2.59012222e-01 1.50215197e+00 -5.75621963e-01 -2.96811074e-01
-3.98598731e-01 8.75494361e-01 1.05851620e-01 5.21906078e-01
-9.98030841e-01 -3.80458951e-01 8.22999656e-01 9.61016491e-02
5.22634625e-01 -1.69094026e-01 -2.30368197e-01 -1.05282867e+00
2.57198632e-01 -6.34438097e-01 2.82413006e-01 -7.70576835e-01
-1.32113957e+00 4.77691591e-01 -1.92085266e-01 -1.35090125e+00
4.23742652e-01 -7.10646987e-01 -1.03203499e+00 7.33063579e-01
-2.35557985e+00 -1.23667932e+00 -8.49954605e-01 5.16025364e-01
6.10504866e-01 1.80770159e-01 1.93190917e-01 4.43974167e-01
-8.33794057e-01 3.62143934e-01 6.67695999e-02 3.00532043e-01
8.05845320e-01 -1.17617786e+00 3.55262250e-01 1.61879671e+00
1.40465751e-01 1.93261102e-01 8.53303969e-01 -5.10643303e-01
-1.29744673e+00 -1.51166141e+00 2.90889591e-01 -2.39131585e-01
4.65247840e-01 -3.23668063e-01 -1.29015422e+00 2.90968448e-01
2.19494909e-01 5.08473635e-01 2.44348109e-01 -1.60984933e-01
-5.76055408e-01 -3.16954762e-01 -8.31230342e-01 2.93424338e-01
9.62203085e-01 -3.57843459e-01 -6.43713951e-01 4.22433019e-01
6.31461442e-01 -2.85008818e-01 -3.14534009e-01 7.57889569e-01
2.04841554e-01 -1.06708407e+00 1.16762412e+00 -2.56696641e-01
2.22394899e-01 -7.48163819e-01 -8.39618370e-02 -1.45212603e+00
-4.21297550e-01 -3.94022197e-01 -1.47537664e-01 1.17174637e+00
4.26255077e-01 -8.05022359e-01 4.97984558e-01 -2.08338454e-01
-4.38421190e-01 -6.22941315e-01 -6.53244436e-01 -8.61412585e-01
-1.74915507e-01 -2.49574214e-01 3.63140702e-01 6.97882891e-01
-6.33898258e-01 8.20927545e-02 -5.76724231e-01 9.51664805e-01
7.36224651e-01 2.08709225e-01 7.69415736e-01 -1.22217345e+00
-1.65703923e-01 -3.49405795e-01 -2.71494716e-01 -1.03109467e+00
-1.33088067e-01 -4.38216001e-01 5.21190882e-01 -1.73272324e+00
2.87363082e-01 -1.70789421e-01 -3.83388579e-01 6.31033301e-01
-8.48972559e-01 6.82737529e-01 2.83183604e-01 2.66419739e-01
-7.91893005e-01 6.00077033e-01 1.23335683e+00 -1.39847651e-01
-8.67675394e-02 1.59443431e-02 -8.50534379e-01 7.90698588e-01
7.64116287e-01 -5.09308338e-01 -5.73738664e-02 -6.06169939e-01
-9.14557558e-03 -3.31128359e-01 8.10293496e-01 -1.27655494e+00
3.78825605e-01 -1.61695600e-01 5.73735476e-01 -3.28570664e-01
2.66227156e-01 -5.87497056e-01 -3.98761034e-01 5.84856153e-01
1.15618542e-01 -2.71134377e-01 2.83673793e-01 7.79760003e-01
-3.31950814e-01 -7.89978169e-03 9.29279685e-01 -4.16387767e-02
-1.20903838e+00 3.48161638e-01 -3.84672433e-01 -2.14964464e-01
1.05739224e+00 -3.56448889e-01 -8.04580331e-01 -2.27905825e-01
-3.56763631e-01 5.91388643e-01 3.06830376e-01 5.50110221e-01
8.11730862e-01 -7.22762525e-01 -1.14736819e+00 1.49208218e-01
2.79846549e-01 5.86105138e-02 3.99665594e-01 1.08763802e+00
-3.70147526e-01 -7.81945735e-02 -7.59483278e-02 -5.28916001e-01
-1.25081277e+00 4.39836651e-01 6.02073252e-01 -5.88980727e-02
-5.19180954e-01 9.83627737e-01 5.76241553e-01 -2.01068193e-01
5.54329306e-02 -1.23352371e-01 -2.01445997e-01 8.96509215e-02
1.00711942e+00 3.46128404e-01 1.06399566e-01 -4.12962198e-01
-2.70571887e-01 4.13956106e-01 -1.93841115e-01 3.32881361e-01
1.40092826e+00 -2.36985177e-01 -2.04666331e-02 -4.75884005e-02
7.46165514e-01 -1.06386833e-01 -1.68112147e+00 -3.81945789e-01
-3.41997147e-01 -6.60155296e-01 4.45761412e-01 -8.40232015e-01
-1.30553913e+00 8.48999858e-01 9.57016349e-01 1.23106949e-01
1.58994579e+00 4.40929420e-02 7.30046332e-01 4.63707983e-01
-5.78481369e-02 -6.05184853e-01 3.80869299e-01 4.93809670e-01
8.03734303e-01 -1.43537784e+00 6.53027967e-02 -6.11437261e-01
-4.93216783e-01 8.57988596e-01 8.26157689e-01 -3.83922130e-01
5.15861750e-01 4.67336953e-01 2.58855730e-01 -1.82468697e-01
-6.58148408e-01 -7.27242291e-01 2.92829871e-01 6.69690609e-01
4.67396490e-02 -1.03774585e-01 7.02330321e-02 2.68925339e-01
3.73462558e-01 -2.23533690e-01 6.60290599e-01 9.56260204e-01
-1.08440650e+00 -7.02795386e-01 -7.70873487e-01 4.15250808e-01
-3.46862257e-01 -2.13269234e-01 -4.01457399e-01 6.49393737e-01
5.76981008e-01 1.18215096e+00 -1.65308528e-02 -4.21609640e-01
3.50740403e-01 -3.70769411e-01 1.24477752e-01 -6.19345725e-01
-6.10072434e-01 -1.42220631e-02 -1.49065062e-01 -4.41138208e-01
-4.88000572e-01 -6.23158097e-01 -8.94442260e-01 1.35085240e-01
-7.53714979e-01 -8.43800530e-02 4.14454728e-01 1.04244041e+00
2.90826380e-01 6.48057997e-01 6.92968905e-01 -1.08371246e+00
-4.78376716e-01 -9.86850560e-01 -5.61737180e-01 2.65115172e-01
9.69523370e-01 -7.50473976e-01 -6.51767969e-01 -4.91838157e-03]
|
[10.767985343933105, -3.000232219696045]
|
8b5dc85d-1549-4f8a-b145-8a474abf5bfd
|
robust-unstructured-knowledge-access-in
|
2211.03990
| null |
https://arxiv.org/abs/2211.03990v1
|
https://arxiv.org/pdf/2211.03990v1.pdf
|
Robust Unstructured Knowledge Access in Conversational Dialogue with ASR Errors
|
Performance of spoken language understanding (SLU) can be degraded with automatic speech recognition (ASR) errors. We propose a novel approach to improve SLU robustness by randomly corrupting clean training text with an ASR error simulator, followed by self-correcting the errors and minimizing the target classification loss in a joint manner. In the proposed error simulator, we leverage confusion networks generated from an ASR decoder without human transcriptions to generate a variety of error patterns for model training. We evaluate our approach on the DSTC10 challenge targeted for knowledge-grounded task-oriented conversational dialogues with ASR errors. Experimental results show the effectiveness of our proposed approach, boosting the knowledge-seeking turn detection (KTD) F1 significantly from 0.9433 to 0.9904. Knowledge cluster classification is boosted from 0.7924 to 0.9333 in Recall@1. After knowledge document re-ranking, our approach shows significant improvement in all knowledge selection metrics, from 0.7358 to 0.7806 in Recall@1, from 0.8301 to 0.9333 in Recall@5, and from 0.7798 to 0.8460 in MRR@5 on the test set. In the recent DSTC10 evaluation, our approach demonstrates significant improvement in knowledge selection, boosting Recall@1 from 0.495 to 0.7144 compared to the official baseline. Our source code is released in GitHub https://github.com/yctam/dstc10_track2_task2.git.
|
['Shuhan Yuan', 'Tinglong Liao', 'Zecheng Wang', 'Jiakai Zou', 'Jiacheng Xu', 'Yik-Cheung Tam']
|
2022-11-08
| null | null | null | null |
['spoken-language-understanding', 'spoken-language-understanding']
|
['natural-language-processing', 'speech']
|
[ 1.26767471e-01 2.01239392e-01 2.45967880e-01 -2.32434288e-01
-1.54473305e+00 -9.00015116e-01 4.69929159e-01 -1.35405406e-01
-4.89383966e-01 7.58332610e-01 3.17749262e-01 -4.16241646e-01
3.29528451e-02 -2.70587951e-01 -7.99771011e-01 -4.52910990e-01
2.05048963e-01 5.24855494e-01 2.40248963e-01 -4.46905583e-01
1.21460438e-01 6.93746004e-03 -1.17030036e+00 5.32864571e-01
1.35435247e+00 8.39462519e-01 1.60066485e-01 1.22938955e+00
2.28288621e-01 8.99238944e-01 -1.14600432e+00 -3.35504979e-01
-1.36898652e-01 -4.76077795e-01 -1.08150613e+00 -4.55678478e-02
1.41890004e-01 -1.10634089e-01 -4.37079132e-01 9.64004278e-01
6.47113860e-01 4.52642292e-01 4.62455481e-01 -8.17018867e-01
-5.39895773e-01 7.37834334e-01 -2.38398284e-01 2.89335195e-02
5.98147452e-01 1.86094761e-01 6.97716236e-01 -8.80085766e-01
2.98846513e-01 1.30661392e+00 5.40681720e-01 9.15608525e-01
-1.11088467e+00 -7.38388538e-01 -1.23253144e-01 4.20667857e-01
-1.52935719e+00 -8.21055472e-01 1.08522080e-01 -1.38092250e-01
1.27763557e+00 4.47117805e-01 -1.05357528e-01 1.28786838e+00
-3.89573097e-01 9.59248006e-01 1.23903394e+00 -6.87526345e-01
2.09814832e-01 3.21937740e-01 5.92857778e-01 5.93933344e-01
-1.83214888e-01 -1.27078280e-01 -9.46323454e-01 -7.39113018e-02
1.40509769e-01 -6.56977832e-01 -7.23784924e-01 4.64745253e-01
-9.87466812e-01 5.73864520e-01 1.97750703e-01 2.30239540e-01
-1.58036277e-01 -1.64011717e-01 3.26977372e-01 5.39633989e-01
5.27987778e-01 4.39559877e-01 -6.43956482e-01 -6.20277464e-01
-5.87164760e-01 7.72429407e-02 1.02376580e+00 1.04937387e+00
5.21647811e-01 1.55910194e-01 -4.64323610e-01 1.48285782e+00
-8.00763816e-03 8.94697249e-01 5.92233777e-01 -9.28354621e-01
6.03687286e-01 2.31604055e-01 2.86548704e-01 -5.13078332e-01
-2.25471303e-01 -6.51484132e-01 -6.06664181e-01 -2.71326452e-01
2.99890250e-01 -2.54731119e-01 -1.11411631e+00 1.80307698e+00
7.57758692e-02 2.22210921e-02 6.72016680e-01 7.22909451e-01
8.04398656e-01 8.61730099e-01 -1.44902468e-01 -1.47350192e-01
1.17966449e+00 -1.16152060e+00 -8.18860412e-01 -1.80542514e-01
9.94713128e-01 -1.02118707e+00 1.34346962e+00 5.44659793e-01
-8.89537752e-01 -1.67368487e-01 -8.69040966e-01 3.61930072e-01
-1.37812138e-01 1.17066644e-01 -1.37066066e-01 8.06047678e-01
-1.20655918e+00 2.03075960e-01 -5.86236417e-01 -3.80670935e-01
1.49983615e-01 7.98838362e-02 -2.29663223e-01 -4.46904361e-01
-1.35232329e+00 9.76564407e-01 4.54214662e-01 -6.11189716e-02
-1.00445151e+00 -7.36755729e-01 -6.30865991e-01 -6.23194985e-02
6.15077615e-01 -2.58844227e-01 1.89174342e+00 -6.86474502e-01
-1.81772411e+00 4.92870390e-01 -4.21378314e-01 -6.68508351e-01
6.99776351e-01 -5.63954592e-01 -5.77973843e-01 1.04888856e-01
-4.39123027e-02 3.64787817e-01 3.16227645e-01 -1.37367284e+00
-7.02771366e-01 -3.79849523e-02 -6.14874884e-02 4.73329872e-01
-1.31640896e-01 7.78740942e-02 -5.17142296e-01 -4.87034976e-01
-5.01816440e-03 -1.01675057e+00 2.40003690e-01 -9.25547659e-01
-6.12768233e-01 -1.94672197e-01 5.99890471e-01 -1.15499949e+00
1.33006442e+00 -1.88209856e+00 -8.62992033e-02 1.97918192e-01
-2.12681949e-01 8.52002203e-01 -1.86010778e-01 3.81889611e-01
2.62919009e-01 2.12694660e-01 -5.72248936e-01 -4.33906138e-01
-3.60202901e-02 1.37068778e-01 -1.92878544e-01 1.86319146e-02
1.83664292e-01 8.04608524e-01 -1.07269251e+00 7.44928941e-02
1.95039392e-01 5.16834795e-01 -4.34042096e-01 4.73637223e-01
-1.00637846e-01 3.53378713e-01 -5.33716604e-02 4.32531238e-01
4.91695791e-01 4.48326766e-02 1.29931271e-01 2.46307448e-01
1.19863637e-01 7.33869493e-01 -9.30041075e-01 1.40321231e+00
-8.15939665e-01 7.08837748e-01 9.76054445e-02 -6.45516753e-01
8.40631306e-01 5.95799506e-01 -2.06931680e-01 -6.74653411e-01
2.88140960e-02 2.16938123e-01 1.55667007e-01 -1.83135077e-01
5.73970258e-01 -6.54196814e-02 -1.30115509e-01 4.30214107e-01
2.16023147e-01 -1.13188580e-01 -2.37831660e-02 5.67663193e-01
1.38855517e+00 -4.87465829e-01 3.10477726e-02 -1.94390759e-01
5.72428644e-01 1.84442028e-02 2.24518135e-01 1.04792368e+00
-2.90896326e-01 6.66428804e-01 2.51361549e-01 3.02976072e-01
-6.10395014e-01 -9.54863429e-01 -6.27545342e-02 1.03245890e+00
-3.20227653e-01 -2.55119056e-01 -1.18778002e+00 -7.64162540e-01
-2.41101801e-01 1.45513844e+00 -3.54501069e-01 -4.65046525e-01
-4.83964920e-01 -4.54640955e-01 1.02433777e+00 2.88685113e-01
6.83349967e-01 -1.10757625e+00 2.21359640e-01 5.81632257e-02
-6.31983638e-01 -1.36079085e+00 -5.90792358e-01 1.32361963e-01
-3.85057926e-01 -8.76316130e-01 -6.68619215e-01 -6.14188015e-01
3.74462277e-01 4.78875756e-01 9.63598073e-01 1.65517449e-01
1.43281445e-01 5.37491500e-01 -9.39336658e-01 -2.58163095e-01
-1.14323008e+00 1.58125639e-01 4.18386817e-01 4.45290580e-02
2.59403557e-01 -5.72902039e-02 -1.96325123e-01 5.06079912e-01
-6.86123431e-01 -1.87322602e-01 3.67563039e-01 1.16504192e+00
1.47146493e-01 -1.75633520e-01 9.43406522e-01 -7.34191418e-01
7.65116036e-01 -3.86031121e-01 -3.08280021e-01 5.13690889e-01
-6.33350849e-01 -1.13639288e-01 4.75169539e-01 -3.89479339e-01
-1.21078801e+00 -3.41423452e-01 -2.73586512e-01 -3.99813354e-01
-2.61260808e-01 4.60081011e-01 -1.97541729e-01 9.47081223e-02
8.89227927e-01 5.72196305e-01 -5.87906539e-02 -4.36725259e-01
4.78024542e-01 1.30217779e+00 5.38476825e-01 -3.94243687e-01
4.57721561e-01 -2.06035718e-01 -1.03519225e+00 -1.12351191e+00
-9.15122032e-01 -5.95245481e-01 -2.78561085e-01 -1.85069546e-01
4.12042439e-01 -9.96450067e-01 -6.20895505e-01 7.46559739e-01
-1.10728312e+00 -7.63696134e-01 1.51991069e-01 3.57502460e-01
-4.40607280e-01 4.07792985e-01 -5.85202277e-01 -1.05273581e+00
-6.55942142e-01 -1.04544866e+00 6.55029178e-01 5.92188127e-02
-4.73241359e-01 -7.66659558e-01 3.56817916e-02 9.39661622e-01
4.92801607e-01 -2.89945096e-01 4.81933355e-01 -1.18724847e+00
-3.03158909e-01 -1.32100672e-01 -4.18074504e-02 8.44695210e-01
2.52953321e-01 -3.84004593e-01 -1.26801360e+00 -4.01878655e-01
-9.20895040e-02 -5.95394790e-01 8.26955736e-01 7.96544552e-02
7.79997945e-01 -5.42820334e-01 -9.82494578e-02 5.15080318e-02
8.27252626e-01 5.26361227e-01 6.68559313e-01 1.61474049e-01
6.27435565e-01 5.96211255e-01 6.91294789e-01 1.33797497e-01
3.53110015e-01 8.39006543e-01 -1.58663169e-01 4.28209245e-01
-4.28239495e-01 -1.40415847e-01 7.27903306e-01 1.02988935e+00
3.36543322e-01 -4.67338532e-01 -1.27508533e+00 8.29388082e-01
-1.70196414e+00 -7.41157413e-01 -1.13893524e-01 2.36576915e+00
1.25448191e+00 3.35934088e-02 -3.75427566e-02 6.28365949e-02
8.90243411e-01 -1.71847880e-01 -2.30963781e-01 -4.84717339e-01
-1.87619403e-01 1.34334534e-01 2.07377896e-01 1.17275548e+00
-6.92933440e-01 1.39611745e+00 4.83566284e+00 1.05059326e+00
-8.45350385e-01 4.34169412e-01 6.56151414e-01 -1.34682328e-01
-1.70086455e-02 -2.74899602e-01 -7.66588449e-01 5.52387476e-01
1.46385729e+00 -2.77654648e-01 6.04260445e-01 5.36006749e-01
2.90457606e-01 -2.74715006e-01 -7.10595906e-01 9.74057496e-01
2.38114119e-01 -8.77648771e-01 2.32702158e-02 -3.81679922e-01
8.39490950e-01 1.40743494e-01 4.49823588e-02 5.87643266e-01
6.69094861e-01 -1.10052514e+00 6.32291734e-01 3.23858529e-01
7.05618978e-01 -9.66141820e-01 9.06806946e-01 3.50426078e-01
-5.12442708e-01 1.93822220e-01 -7.93316588e-02 1.00016944e-01
7.20604733e-02 5.11168838e-01 -1.72919858e+00 5.02967238e-01
7.70952404e-01 -9.15322918e-03 -1.14869356e-01 7.09844291e-01
-4.96046066e-01 1.23348176e+00 -2.82332838e-01 -2.27096111e-01
1.13495797e-01 2.97662884e-01 7.17226386e-01 1.67844927e+00
5.54920100e-02 3.78080398e-01 -1.74085200e-01 4.10675824e-01
-3.74120712e-01 8.66238996e-02 -4.26960111e-01 2.13283956e-01
9.63777840e-01 8.14794540e-01 -1.60510227e-01 -5.03437281e-01
7.57633075e-02 1.34244549e+00 4.55359519e-01 5.15147924e-01
-7.71062255e-01 -6.40916467e-01 6.78546786e-01 -3.47702801e-01
2.02686295e-01 -4.83255684e-02 -2.15159521e-01 -1.05629706e+00
1.09599128e-01 -1.25354397e+00 1.80060491e-01 -7.28196740e-01
-1.06939912e+00 9.78675365e-01 -3.15740198e-01 -8.65187109e-01
-2.96387613e-01 -3.35632175e-01 -3.38728756e-01 1.04368281e+00
-1.19594038e+00 -8.45615089e-01 -1.61578238e-01 2.87730336e-01
9.23903108e-01 -2.92025477e-01 9.24512565e-01 1.51923448e-01
-6.17710888e-01 1.16528058e+00 3.41906548e-01 2.07921371e-01
8.80706906e-01 -1.12870693e+00 4.30615097e-01 7.97920585e-01
6.30239323e-02 6.38027489e-01 6.56530857e-01 -6.21608555e-01
-9.24150348e-01 -1.21586478e+00 1.01585221e+00 -9.24547911e-01
6.62041068e-01 -4.61998731e-01 -1.28875756e+00 6.24327481e-01
3.37437838e-01 -5.34381807e-01 5.53876996e-01 1.51001140e-01
-5.26919901e-01 3.76198366e-02 -1.15771997e+00 5.82258403e-01
7.99505770e-01 -7.36872017e-01 -5.43911278e-01 4.34267670e-01
1.04995537e+00 -6.27328038e-01 -8.56845379e-01 3.04697245e-01
3.24760407e-01 -6.30836904e-01 5.36781073e-01 -4.57719028e-01
1.13077678e-01 -2.45381489e-01 -3.61049145e-01 -1.75903249e+00
-8.03067759e-02 -8.91983151e-01 -7.84258172e-02 1.40600836e+00
9.44818139e-01 -6.82738185e-01 3.80977780e-01 3.83969992e-01
-5.18587649e-01 -4.47660059e-01 -1.17435861e+00 -8.77506614e-01
2.02684686e-01 -6.51888192e-01 1.03399508e-01 9.22748804e-01
8.00078064e-02 4.65572745e-01 -3.24677736e-01 4.73444939e-01
4.45861906e-01 -5.91342390e-01 7.84105480e-01 -4.61459488e-01
-2.09006056e-01 -2.46458724e-01 5.85131645e-02 -1.10092139e+00
1.09309852e-01 -8.93725157e-01 4.38131928e-01 -1.48578465e+00
2.29342669e-01 -2.98844665e-01 -3.55910718e-01 6.58634603e-01
-4.78849858e-01 9.34145749e-02 1.78673834e-01 1.07418381e-01
-6.94208443e-01 7.20508218e-01 6.90904260e-01 -6.71195239e-02
-3.40271115e-01 2.33549047e-02 -7.35049963e-01 3.38953882e-01
1.31449044e+00 -5.25137663e-01 -3.15325797e-01 -3.61538470e-01
-5.06271362e-01 3.52051179e-03 1.48484766e-01 -1.01681137e+00
1.43418029e-01 2.80133694e-01 -2.13428557e-01 -3.61502498e-01
3.39571476e-01 -2.50154048e-01 -2.60804564e-01 5.36489367e-01
-4.25556511e-01 -3.35242331e-01 5.62625289e-01 4.15024132e-01
-1.32818103e-01 -1.13006517e-01 8.72646272e-01 1.00756995e-01
-5.77545345e-01 -4.35353130e-01 -6.95834994e-01 2.90069461e-01
5.52069247e-01 1.12256192e-01 -6.19021058e-01 -8.23497713e-01
-7.03095257e-01 3.50844890e-01 1.43216670e-01 6.89086676e-01
6.03693426e-01 -8.81349266e-01 -1.00858629e+00 -3.98477316e-02
3.07610691e-01 -1.83700606e-01 5.49619794e-01 7.99496293e-01
-2.90776372e-01 6.72623992e-01 4.87556279e-01 -5.94413102e-01
-1.50447488e+00 -8.83972794e-02 5.52214801e-01 -1.16531037e-01
-1.30171135e-01 1.10485649e+00 7.74388388e-02 -9.37128544e-01
4.99463886e-01 -2.81241626e-01 1.00689419e-01 -1.87371641e-01
7.18870103e-01 6.47507310e-01 4.94954288e-01 -5.66341519e-01
-4.78167236e-01 2.05411658e-01 -3.47851068e-01 -5.04020274e-01
8.87571335e-01 -2.36850008e-01 2.32144222e-01 3.58800560e-01
1.15187585e+00 7.50928745e-02 -8.88082623e-01 -3.54291648e-01
-9.17074271e-03 -1.64586842e-01 3.62962820e-02 -1.64883029e+00
-5.22032678e-01 5.43857276e-01 5.75622439e-01 6.43358827e-02
8.60222340e-01 9.03697535e-02 6.86334372e-01 6.57878160e-01
4.56035137e-01 -1.11190557e+00 2.83633500e-01 1.18915546e+00
1.22437489e+00 -1.31386697e+00 -5.80025911e-01 -3.47383171e-01
-1.13757014e+00 5.41195571e-01 5.35083115e-01 3.24910194e-01
1.89772174e-01 1.28540713e-02 6.41702175e-01 1.30466729e-01
-8.80435348e-01 -1.16443336e-01 1.66268498e-01 5.40138364e-01
3.21422815e-01 4.15397763e-01 -1.25721350e-01 6.17087960e-01
-5.51181734e-01 -3.46617222e-01 6.95748031e-01 7.57610977e-01
-6.09006405e-01 -8.95334661e-01 -4.83831674e-01 1.83895215e-01
-3.41182470e-01 -3.58294576e-01 -6.31025136e-01 4.66916293e-01
-6.42968655e-01 1.64270329e+00 -1.76499754e-01 -7.62188971e-01
7.58948326e-01 5.38056433e-01 3.22648622e-02 -7.20651627e-01
-7.16696203e-01 3.84568125e-02 6.86720431e-01 -3.68459404e-01
2.47534947e-04 -4.98003632e-01 -1.26556015e+00 -2.42936969e-01
-6.34607792e-01 4.80410039e-01 4.48122770e-01 8.10437381e-01
5.90932965e-01 6.52837694e-01 6.97679162e-01 -1.56224251e-01
-6.80317461e-01 -1.46282697e+00 -1.94216922e-01 1.79662421e-01
3.26085538e-01 -4.41206604e-01 -6.93642139e-01 -8.58557895e-02]
|
[14.373149871826172, 6.902366638183594]
|
d1b0b772-fe87-4187-baaf-95cca99ed464
|
bert-meets-ctc-new-formulation-of-end-to-end
|
2210.16663
| null |
https://arxiv.org/abs/2210.16663v2
|
https://arxiv.org/pdf/2210.16663v2.pdf
|
BERT Meets CTC: New Formulation of End-to-End Speech Recognition with Pre-trained Masked Language Model
|
This paper presents BERT-CTC, a novel formulation of end-to-end speech recognition that adapts BERT for connectionist temporal classification (CTC). Our formulation relaxes the conditional independence assumptions used in conventional CTC and incorporates linguistic knowledge through the explicit output dependency obtained by BERT contextual embedding. BERT-CTC attends to the full contexts of the input and hypothesized output sequences via the self-attention mechanism. This mechanism encourages a model to learn inner/inter-dependencies between the audio and token representations while maintaining CTC's training efficiency. During inference, BERT-CTC combines a mask-predict algorithm with CTC decoding, which iteratively refines an output sequence. The experimental results reveal that BERT-CTC improves over conventional approaches across variations in speaking styles and languages. Finally, we show that the semantic representations in BERT-CTC are beneficial towards downstream spoken language understanding tasks.
|
['Shinji Watanabe', 'Tetsunori Kobayashi', 'Tetsuji Ogawa', 'Siddhant Arora', 'Brian Yan', 'Yosuke Higuchi']
|
2022-10-29
| null | null | null | null |
['spoken-language-understanding', 'spoken-language-understanding']
|
['natural-language-processing', 'speech']
|
[ 2.31472328e-01 3.74216408e-01 -2.82805830e-01 -8.06395948e-01
-9.63283658e-01 -5.51384628e-01 7.11607814e-01 -2.49244735e-01
-4.10594583e-01 4.27852064e-01 7.76384294e-01 -5.34426987e-01
1.57160237e-01 -2.12243140e-01 -4.38797444e-01 -5.99819005e-01
-1.35600790e-01 5.82290053e-01 1.02840446e-01 -1.05555855e-01
-7.30179101e-02 1.11249782e-01 -1.28724289e+00 8.49555433e-01
5.77411532e-01 8.09823930e-01 4.27916169e-01 1.19783437e+00
-2.04120547e-01 1.09071469e+00 -6.07700109e-01 -1.06338896e-01
-3.38730067e-01 -6.00359261e-01 -9.22369063e-01 -1.13651134e-01
-2.17267439e-01 -2.24059045e-01 -5.62204778e-01 4.30319339e-01
3.03222865e-01 4.67621118e-01 6.96070969e-01 -1.06012249e+00
-6.15987062e-01 1.02203393e+00 4.91605364e-02 5.27542889e-01
2.58475691e-01 9.41810571e-03 1.34356129e+00 -1.04436588e+00
1.53513610e-01 1.55117607e+00 4.71926183e-01 8.32293272e-01
-1.25386751e+00 -3.98745924e-01 7.86531448e-01 4.96420115e-01
-1.14371979e+00 -9.25620317e-01 5.86073756e-01 -1.45013958e-01
1.59675932e+00 5.08956134e-01 5.87333024e-01 1.28660476e+00
-1.78449720e-01 1.35024738e+00 7.04724371e-01 -8.48353446e-01
3.02779496e-01 3.01281773e-02 3.10350299e-01 4.77787942e-01
-9.40796316e-01 4.83463019e-01 -1.14083707e+00 -4.72692549e-02
5.50040245e-01 -5.71340203e-01 -2.59985358e-01 1.63689390e-01
-1.12209415e+00 7.86769390e-01 5.03816716e-02 4.38556403e-01
2.20133606e-02 4.19302523e-01 4.98719871e-01 4.47822869e-01
5.66592932e-01 -1.62178744e-02 -7.24505126e-01 -5.84132254e-01
-8.42267275e-01 -4.61388499e-01 9.06070590e-01 1.06290853e+00
2.23841876e-01 3.89647394e-01 -3.41837972e-01 9.34506297e-01
5.96145809e-01 3.73324692e-01 6.98778629e-01 -8.26124847e-01
4.97633696e-01 -2.00105071e-01 -2.01730356e-01 -1.75915569e-01
1.06062964e-01 -4.42628443e-01 -4.88653034e-01 -3.11127007e-01
4.60039340e-02 -5.73391095e-02 -9.75272000e-01 1.90377879e+00
9.37473327e-02 6.41038537e-01 4.11436498e-01 6.54075682e-01
5.82464874e-01 8.97210896e-01 1.25014216e-01 -2.93442190e-01
9.69967723e-01 -1.37113500e+00 -1.00881851e+00 -2.62596041e-01
5.65062106e-01 -5.61228156e-01 1.16931415e+00 3.32622796e-01
-9.56719935e-01 -7.10565567e-01 -9.80544686e-01 -1.64039031e-01
-2.81608403e-01 1.92682505e-01 4.48902607e-01 5.77423155e-01
-1.26732385e+00 4.24611598e-01 -1.09431219e+00 -1.08423889e-01
-9.94206220e-02 4.54739720e-01 2.45283991e-02 1.47590160e-01
-1.23773181e+00 7.00030625e-01 4.38129306e-01 4.35071141e-01
-1.22952223e+00 -2.41828904e-01 -9.42623794e-01 1.49244204e-01
1.52942672e-01 -3.47535461e-01 1.93586326e+00 -1.16389310e+00
-2.33076429e+00 4.26147729e-01 -1.04390669e+00 -5.29134154e-01
1.33232340e-01 -5.32076538e-01 -5.94764471e-01 2.20642835e-01
-2.10582957e-01 6.96642101e-01 9.66039121e-01 -1.02423739e+00
-5.93943834e-01 1.34512424e-01 -1.47900015e-01 3.66843373e-01
-2.68822074e-01 6.37586340e-02 -7.68713236e-01 -7.44275689e-01
6.80908635e-02 -7.22884059e-01 -9.24058110e-02 -4.61820662e-01
-3.31301838e-01 -5.34875214e-01 8.67147148e-01 -5.16771972e-01
1.29656076e+00 -2.38062859e+00 4.81102020e-01 1.86318055e-01
-2.15708017e-01 7.87770152e-02 -3.26912105e-01 7.68116772e-01
-4.52545807e-02 3.29338908e-02 -7.49666467e-02 -1.01401472e+00
1.46861985e-01 8.00724506e-01 -8.09770942e-01 2.39421666e-01
3.04494351e-01 9.51711118e-01 -9.23996925e-01 -3.60883176e-01
4.77489442e-01 4.98510778e-01 -4.72815990e-01 5.41148245e-01
-6.16547227e-01 7.02414215e-01 -2.91512758e-01 2.43326694e-01
9.65975374e-02 -7.85725117e-02 3.40078950e-01 2.57517159e-01
-2.02641711e-01 1.07800686e+00 -6.77188396e-01 1.72310805e+00
-6.32011354e-01 7.95348406e-01 6.84575513e-02 -1.07603025e+00
7.38080263e-01 9.82240200e-01 1.56100178e-02 -4.80064571e-01
8.19528475e-03 9.73013639e-02 -1.01424217e-01 -4.47109669e-01
2.31068075e-01 -2.62866795e-01 1.30704464e-02 6.24123156e-01
2.80248851e-01 -1.01141512e-01 -3.95619482e-01 2.71994561e-01
6.13929808e-01 3.02271336e-01 4.98173498e-02 -3.98558080e-02
5.19128144e-01 -2.20370516e-01 5.31630695e-01 7.65803397e-01
-1.84695408e-01 4.46210086e-01 3.13280046e-01 2.25388203e-02
-6.88710690e-01 -1.25637090e+00 9.84227210e-02 1.52903593e+00
-1.64873391e-01 -5.39402604e-01 -4.76937532e-01 -5.85692644e-01
-2.97034681e-01 1.10051370e+00 -5.53209782e-01 -2.07881927e-01
-6.33890390e-01 -1.57668650e-01 8.54383290e-01 8.28773141e-01
9.69924480e-02 -1.18932426e+00 -4.51162793e-02 5.47361791e-01
-3.90234798e-01 -1.12975430e+00 -8.43031108e-01 7.59301126e-01
-9.56212699e-01 -4.80790406e-01 -2.12113872e-01 -1.11270595e+00
3.10193807e-01 2.70274393e-02 7.57421136e-01 -1.44632354e-01
7.73410872e-02 4.24321145e-01 -5.55397928e-01 -1.83878705e-01
-4.79716688e-01 2.87683960e-02 6.64389059e-02 2.05265135e-01
3.99516344e-01 -5.15938163e-01 -1.92228213e-01 2.91659623e-01
-4.16452199e-01 1.53279528e-01 2.43681341e-01 1.24236906e+00
3.70813757e-01 -2.26118371e-01 6.82589352e-01 -5.14152110e-01
3.81762981e-01 -2.59346634e-01 -1.93032071e-01 2.49132141e-01
-3.09227824e-01 2.88275063e-01 5.88519990e-01 -7.11522877e-01
-1.44238293e+00 -2.97808107e-02 -4.51841980e-01 -4.47560102e-01
-1.92989215e-01 6.08225167e-01 -2.30244562e-01 6.58515215e-01
9.34184417e-02 4.73667443e-01 -2.73202121e-01 -6.30924284e-01
7.66174138e-01 9.55533087e-01 7.68443942e-01 -7.37947762e-01
3.52900594e-01 4.09387648e-01 -1.02432573e+00 -1.06384504e+00
-1.07489026e+00 -7.78806388e-01 -9.10731494e-01 -1.71299204e-01
9.08437610e-01 -9.76562500e-01 -7.07039654e-01 2.42624924e-01
-1.27121079e+00 -7.43816137e-01 -3.90356779e-01 8.88098001e-01
-7.02945411e-01 3.22376698e-01 -1.09430873e+00 -1.35445273e+00
-1.20176122e-01 -8.91756058e-01 1.30268669e+00 -1.62766203e-01
-4.37788516e-01 -1.17202437e+00 -1.10915164e-02 3.00351590e-01
2.05769524e-01 -5.73523462e-01 1.01653945e+00 -5.14491439e-01
-5.32808661e-01 3.48234534e-01 2.32848108e-01 4.08994734e-01
-6.49227016e-03 -6.53351843e-02 -1.70500231e+00 -2.94380307e-01
7.38969073e-02 -2.92131364e-01 9.18981612e-01 4.58200425e-01
9.83654141e-01 -4.45169300e-01 -3.58772784e-01 5.21677673e-01
8.67855012e-01 5.70963025e-01 4.26734746e-01 -4.57334444e-02
4.84510392e-01 7.04865217e-01 4.72157091e-01 2.76724190e-01
2.69591272e-01 5.14137268e-01 1.21023662e-01 2.63947994e-03
-1.26517981e-01 -5.18290222e-01 9.23358619e-01 1.72824955e+00
3.33627909e-01 -5.14126062e-01 -6.60675466e-01 7.16140568e-01
-2.04526687e+00 -1.01612103e+00 6.09310158e-02 1.88997769e+00
1.02516818e+00 2.31665224e-01 -6.17976561e-02 2.29954004e-01
5.88815868e-01 1.75034329e-01 -4.88702893e-01 -8.32586348e-01
-1.31542787e-01 4.82170463e-01 -7.42501244e-02 1.04790545e+00
-8.05176854e-01 1.44761384e+00 7.27222109e+00 8.19388568e-01
-1.03228903e+00 3.67163420e-01 4.82433319e-01 -7.96135515e-02
-4.55717117e-01 1.98982999e-01 -8.01769376e-01 1.24224490e-02
1.43322742e+00 -4.21569869e-02 4.33484823e-01 4.51007217e-01
4.78299558e-01 1.40555084e-01 -1.56765568e+00 6.22359693e-01
1.18219499e-02 -1.20959342e+00 -5.17833866e-02 -2.45746613e-01
2.71369904e-01 2.36500859e-01 1.83316290e-01 5.41014612e-01
6.84915364e-01 -1.24520075e+00 1.13992465e+00 2.13391006e-01
9.05256748e-01 -5.96225679e-01 5.32923341e-01 5.61648071e-01
-1.59877658e+00 -8.48873779e-02 1.79110840e-02 -1.16326123e-01
4.17873830e-01 7.12146312e-02 -1.17501843e+00 4.46200848e-01
4.50688869e-01 8.30197453e-01 1.39798835e-01 5.38210094e-01
-6.60502553e-01 1.33567500e+00 -2.36794084e-01 -1.49773523e-01
5.63952327e-01 3.13540585e-02 5.60501218e-01 1.69537020e+00
-2.01791078e-02 9.78156999e-02 2.10370257e-01 6.63958907e-01
4.03010309e-01 -1.19140953e-01 -4.25549388e-01 -2.54993379e-01
6.08604610e-01 4.63742137e-01 -3.76397312e-01 -5.90445757e-01
-3.38701010e-01 1.34622335e+00 4.59387422e-01 7.90615201e-01
-8.08090746e-01 -6.24865890e-02 5.99197388e-01 -5.54455876e-01
7.43697047e-01 -6.50656402e-01 -8.39817524e-02 -1.08728433e+00
-1.25279739e-01 -6.69130266e-01 4.08465207e-01 -7.22344697e-01
-1.09273601e+00 7.79395163e-01 -8.94086957e-02 -8.69860232e-01
-5.51091254e-01 -4.68302816e-01 -7.30447292e-01 8.47394049e-01
-1.61223066e+00 -1.17328262e+00 3.80157858e-01 5.94340622e-01
1.16367948e+00 -1.32661283e-01 1.27211082e+00 -2.57190922e-03
-6.40141189e-01 6.30405843e-01 2.97725588e-01 1.75277159e-01
4.23132807e-01 -1.30604768e+00 5.40641248e-01 7.93861449e-01
6.16734147e-01 6.59670055e-01 4.86992508e-01 -3.98699433e-01
-1.10755289e+00 -9.73041832e-01 1.33694756e+00 -2.83312976e-01
7.89489925e-01 -7.40600765e-01 -8.27769995e-01 1.04977846e+00
5.04703104e-01 -3.15628171e-01 8.78294468e-01 3.82693052e-01
-6.44494236e-01 3.97343598e-02 -3.41677427e-01 6.92516387e-01
1.04568589e+00 -1.41332972e+00 -1.09223700e+00 1.18741274e-01
1.21185195e+00 -1.75978333e-01 -5.14814198e-01 -6.00636043e-02
5.97706497e-01 -6.23225749e-01 7.69728124e-01 -6.67966664e-01
-2.40939543e-01 -1.62636831e-01 -3.98511767e-01 -1.27494204e+00
-2.58109391e-01 -9.99928176e-01 -2.08576471e-01 1.16322219e+00
6.65472209e-01 -4.62350875e-01 4.82244194e-01 2.64188230e-01
-6.99911118e-01 -3.10064524e-01 -1.42495382e+00 -8.65327597e-01
-4.29485022e-04 -1.06251860e+00 4.49545205e-01 8.02434266e-01
4.56712037e-01 5.56306183e-01 -3.10694247e-01 4.59460109e-01
2.24030748e-01 -2.80618947e-02 3.23264599e-01 -7.87259519e-01
-8.31108153e-01 -2.74745941e-01 1.50791898e-01 -1.88426757e+00
5.42547703e-01 -1.01151431e+00 5.31399667e-01 -1.22048509e+00
-2.18799770e-01 -4.18164909e-01 -5.33443987e-01 6.54781759e-01
6.50774464e-02 -2.06864104e-01 1.47104729e-02 1.77069366e-01
-5.43560445e-01 1.06775844e+00 9.77165282e-01 -2.35456482e-01
-3.83624375e-01 -2.59992220e-02 -1.49693474e-01 4.65125680e-01
7.01560199e-01 -4.44049925e-01 -6.35727704e-01 -7.40370154e-01
-4.21376228e-01 3.23114038e-01 1.21203095e-01 -6.83610260e-01
4.30800289e-01 -3.17554572e-03 -7.31715932e-03 -7.75580764e-01
8.42844307e-01 -5.88974655e-01 -1.70406327e-01 4.06141520e-01
-8.45851779e-01 -1.65484950e-01 3.38834882e-01 8.39515746e-01
-6.40831769e-01 1.18517140e-02 5.38840473e-01 1.33191824e-01
-7.07159400e-01 -2.86917519e-02 -1.04026496e+00 -6.50023073e-02
5.58797598e-01 -3.08717310e-01 8.49005654e-02 -6.71452940e-01
-1.30918503e+00 3.88510942e-01 -5.08294344e-01 6.46259010e-01
8.53120923e-01 -1.14086485e+00 -5.59372187e-01 4.67907429e-01
1.45064980e-01 -1.50590315e-01 -1.17752612e-01 8.26015353e-01
2.00398967e-01 7.46053219e-01 5.71325779e-01 -8.39520693e-01
-1.39932668e+00 2.25906193e-01 4.20653582e-01 -7.27756843e-02
-8.31557274e-01 1.24879968e+00 3.53202254e-01 -4.83467877e-01
8.48836422e-01 -7.04823613e-01 -8.47197846e-02 -2.13594988e-01
3.73728037e-01 -1.21269047e-01 -1.08263902e-01 -5.18679678e-01
-3.58341724e-01 8.95686075e-02 -2.24938452e-01 -1.08476627e+00
1.23864317e+00 -5.04310429e-01 2.80773401e-01 1.19158471e+00
1.10448277e+00 -7.55344778e-02 -1.36268163e+00 -4.12774980e-01
3.71465921e-01 3.64596397e-02 9.85109583e-02 -7.74531662e-01
-7.11458623e-01 1.19362259e+00 2.31709957e-01 -1.85671583e-01
8.38570356e-01 1.79241449e-01 6.98250890e-01 4.64085251e-01
1.89887494e-01 -1.15253758e+00 2.78971642e-01 1.08505321e+00
9.60913658e-01 -7.93887198e-01 -7.95109689e-01 -3.77688229e-01
-8.00441146e-01 1.08587420e+00 5.06591737e-01 3.45415920e-01
7.77980506e-01 5.46840906e-01 2.76337236e-01 1.49986282e-01
-1.43829131e+00 -2.62508035e-01 1.74782723e-01 5.45799315e-01
6.12488449e-01 1.93541095e-01 2.49365494e-01 4.69177693e-01
-2.26652831e-01 -3.62646818e-01 3.54185887e-02 8.69369149e-01
-3.88757735e-01 -1.27848530e+00 -2.37741470e-01 -3.15528810e-02
-6.76186085e-02 -4.84684944e-01 -5.10225475e-01 5.10527790e-01
-1.49039865e-01 1.50782192e+00 2.31671959e-01 -4.39963073e-01
1.13331564e-02 5.65996289e-01 2.38469347e-01 -8.64239991e-01
-6.77111030e-01 8.20022404e-01 5.02463162e-01 -4.99368638e-01
-3.49999040e-01 -7.34599829e-01 -1.68320143e+00 2.07388431e-01
-5.68971992e-01 7.18456507e-01 4.40217286e-01 1.10005105e+00
2.59058535e-01 7.63375998e-01 6.88247085e-01 -5.37215352e-01
-6.02710664e-01 -1.16209936e+00 -4.18950647e-01 -8.98568928e-02
7.37414718e-01 -3.62379432e-01 -5.62271833e-01 4.76377994e-01]
|
[14.270590782165527, 6.815913677215576]
|
72209d75-aa3d-43c3-8073-579a1d9ce9d7
|
ascm-an-answer-space-clustered-prompting
| null | null |
https://aclanthology.org/2022.findings-acl.193
|
https://aclanthology.org/2022.findings-acl.193.pdf
|
ASCM: An Answer Space Clustered Prompting Method without Answer Engineering
|
Prompt-based learning, which exploits knowledge from pre-trained language models by providing textual prompts and designing appropriate answer-category mapping methods, has achieved impressive successes on few-shot text classification and natural language inference (NLI). Because of the diverse linguistic expression, there exist many answer tokens for the same category. However, both manual answer design and automatic answer search constrain answer space and therefore hardly achieve ideal performance. To address this issue, we propose an answer space clustered prompting model (ASCM) together with a synonym initialization method (SI) which automatically categorizes all answer tokens in a semantic-clustered embedding space. We also propose a stable semi-supervised method named stair learning (SL) that orderly distills knowledge from better models to weaker models. Extensive experiments demonstrate that our ASCM+SL significantly outperforms existing state-of-the-art techniques in few-shot settings.
|
['Azmat Anwar', 'Rui Dong', 'Lei Wang', 'Bo Ma', 'Zhou Xi', 'Yating Yang', 'Zhen Wang']
| null | null | null | null |
findings-acl-2022-5
|
['few-shot-text-classification']
|
['natural-language-processing']
|
[ 2.42119655e-01 2.03431193e-02 -6.90969646e-01 -5.77623427e-01
-9.38027978e-01 -4.71283704e-01 8.88527453e-01 5.58966994e-01
-6.16618574e-01 5.41490734e-01 4.78767604e-01 -3.94860893e-01
-2.48363808e-01 -6.96492374e-01 -4.16670144e-02 -6.66319281e-02
5.18152356e-01 6.91726089e-01 5.69968164e-01 -4.91523564e-01
6.32990897e-01 -2.30371222e-01 -1.61833131e+00 4.89711940e-01
1.18200135e+00 7.13570595e-01 1.59685493e-01 5.78976989e-01
-1.06322455e+00 1.37774360e+00 -4.87137347e-01 -5.39978147e-01
-2.19746858e-01 -6.27744913e-01 -1.11377978e+00 -5.27728856e-01
4.69848663e-01 -1.45200819e-01 -2.01061562e-01 7.63618290e-01
4.08225566e-01 5.85911274e-01 8.30539823e-01 -1.22676289e+00
-9.02868807e-01 7.58411944e-01 -8.60432833e-02 3.79974604e-01
8.35968852e-01 -4.32551913e-02 1.44093382e+00 -1.47308707e+00
6.09554172e-01 1.32663023e+00 4.97379690e-01 7.86669433e-01
-1.40176225e+00 -5.95555961e-01 1.79569900e-01 6.20727777e-01
-1.17670119e+00 -2.42383495e-01 9.98196304e-01 -3.55521262e-01
1.02489448e+00 3.23515415e-01 1.10176973e-01 1.05234516e+00
-1.85893297e-01 1.04772103e+00 9.76479352e-01 -8.33458066e-01
4.97886300e-01 3.89611721e-01 9.29675639e-01 6.69583678e-01
-1.39258042e-01 -3.00935626e-01 -7.93962717e-01 -4.05992568e-01
-3.06275277e-03 1.56323299e-01 -3.47141065e-02 -3.87753874e-01
-9.80374396e-01 1.22645795e+00 2.07639992e-01 3.09819967e-01
-1.05677314e-01 -1.86295137e-01 4.43354785e-01 6.82262242e-01
4.75785822e-01 8.70946527e-01 -4.05214667e-01 4.53515090e-02
-7.62416124e-01 4.20594752e-01 1.02310789e+00 9.78148699e-01
1.02203059e+00 -3.91729772e-01 -7.51474440e-01 1.13445175e+00
2.92305887e-01 1.74274296e-01 6.12320304e-01 -9.21892643e-01
5.59029520e-01 9.46607411e-01 2.27061599e-01 -9.56010342e-01
-3.12158495e-01 -1.42706618e-01 -2.61607111e-01 -3.92797947e-01
2.54352480e-01 4.04183902e-02 -3.53187561e-01 1.59217370e+00
4.94101912e-01 1.19721584e-01 1.02151170e-01 6.49822414e-01
1.28606522e+00 5.47116220e-01 3.74626875e-01 -1.03140570e-01
1.27766848e+00 -1.18214071e+00 -8.78841400e-01 -4.57365513e-01
1.01438975e+00 -5.94946086e-01 1.72664070e+00 -1.11867510e-01
-5.64617693e-01 -7.04943657e-01 -8.97528231e-01 -2.35441551e-01
-4.94268060e-01 -1.09241508e-01 6.01139009e-01 3.87077212e-01
-5.77268302e-01 3.72383654e-01 -7.67563581e-02 -5.47492206e-01
3.00437033e-01 -9.16460808e-03 -1.44249685e-02 -3.73902678e-01
-1.60552669e+00 1.01211417e+00 2.13060290e-01 -6.56575799e-01
-3.45052451e-01 -8.40401590e-01 -1.02877116e+00 1.66459888e-01
6.71716869e-01 -6.60072267e-01 1.37149060e+00 -4.33974892e-01
-1.38879275e+00 9.43543971e-01 -5.08355439e-01 -3.80393684e-01
-1.91533595e-01 -2.02810004e-01 -3.50202233e-01 2.00960666e-01
3.50061864e-01 7.74688065e-01 9.01931524e-01 -1.11303520e+00
-5.60607493e-01 -7.28820115e-02 1.43130824e-01 2.13051170e-01
-8.82594585e-01 4.69690531e-01 -1.51674887e-02 -4.53067482e-01
-5.14209345e-02 -3.47558349e-01 -8.56809691e-02 1.16027230e-02
-1.01079784e-01 -1.32421863e+00 5.82665741e-01 -1.80528730e-01
1.94618428e+00 -1.78768575e+00 -4.13588807e-02 -1.15089417e-01
4.37096685e-01 4.94526744e-01 -3.96932691e-01 6.10011101e-01
7.37703890e-02 -2.18166873e-01 -6.20902376e-03 -3.13281298e-01
3.42783570e-01 3.91977847e-01 -7.31542051e-01 -1.17665038e-01
1.69629008e-01 1.22130668e+00 -1.49473023e+00 -9.57312822e-01
2.57059097e-01 -2.96734363e-01 -6.94541156e-01 6.04879141e-01
-5.58555007e-01 7.00847525e-03 -3.86341155e-01 6.15790367e-01
1.60958663e-01 -5.87117195e-01 -1.48578798e-02 1.08964458e-01
2.08546817e-01 6.72368824e-01 -9.34334338e-01 1.69858265e+00
-5.94815314e-01 2.88964927e-01 -4.08862501e-01 -1.11928403e+00
1.20475316e+00 2.66416520e-01 1.52440354e-01 -7.24266589e-01
4.42162119e-02 2.27542564e-01 -9.07758549e-02 -8.64013433e-01
4.82103407e-01 -9.13891122e-02 -2.38376290e-01 6.83358371e-01
4.21136558e-01 -1.75708562e-01 2.46737733e-01 6.86975420e-01
1.26884925e+00 -8.64166184e-04 5.43886364e-01 -2.99973249e-01
7.36060143e-01 6.70117661e-02 3.18271995e-01 1.13133836e+00
-6.09615862e-01 2.88601816e-01 1.96905226e-01 -3.63931537e-01
-4.87327009e-01 -1.10379982e+00 2.32370421e-02 1.68456519e+00
2.00185433e-01 -7.34789193e-01 -4.80551153e-01 -1.04890800e+00
3.19425836e-02 1.16254139e+00 -5.79961002e-01 -4.40281719e-01
-4.46420401e-01 -4.58021415e-04 1.85656086e-01 5.46279430e-01
7.52753988e-02 -1.16423965e+00 -4.85032529e-01 2.96093673e-01
-4.57385093e-01 -7.43273437e-01 -7.06579864e-01 3.54049683e-01
-5.10081530e-01 -1.06112492e+00 -3.60306025e-01 -9.41860974e-01
5.73276520e-01 4.34478015e-01 1.42851079e+00 1.62798107e-01
-8.27547014e-02 6.45100772e-01 -6.43435657e-01 -3.02033097e-01
-2.37762168e-01 1.97763473e-01 1.10143095e-01 -1.60649288e-02
1.38438773e+00 -3.28616202e-01 -4.68174607e-01 2.22280815e-01
-6.91086829e-01 -2.41655663e-01 -2.23153140e-02 9.89536703e-01
1.40083045e-01 -3.76505435e-01 1.19659483e+00 -1.15188348e+00
9.66635287e-01 -7.66719103e-01 -1.04828194e-01 6.65044665e-01
-9.14054096e-01 1.74010113e-01 8.03363681e-01 -5.41040838e-01
-1.19508433e+00 -2.51584023e-01 -9.37137902e-02 -1.51130125e-01
-2.47048870e-01 4.80434656e-01 2.23664671e-01 1.45829037e-01
1.14303946e+00 5.01060784e-02 -1.57375321e-01 -4.32751864e-01
7.44745016e-01 9.52112794e-01 5.14693737e-01 -6.62368596e-01
8.41491044e-01 2.17785791e-01 -5.89499295e-01 -6.23181045e-01
-1.73622453e+00 -1.19610465e+00 -5.18857777e-01 -2.81588167e-01
5.92608750e-01 -5.73234499e-01 -3.74243975e-01 -1.20911211e-01
-1.37144756e+00 -2.24541530e-01 -4.09055710e-01 2.36563727e-01
-3.83794218e-01 2.52743691e-01 -4.89523590e-01 -9.39882874e-01
-3.86205733e-01 -4.35427845e-01 7.30609775e-01 2.21525237e-01
-1.02920723e+00 -1.26630068e+00 3.20302606e-01 5.76908588e-01
5.88904977e-01 -4.30518448e-01 1.28835666e+00 -1.26397026e+00
-2.06497774e-01 -1.35560095e-01 -1.45195276e-01 -1.81743298e-02
1.06008366e-01 -6.29440665e-01 -8.78187776e-01 1.60308689e-01
2.38003299e-01 -9.45631683e-01 9.94863987e-01 -1.37839407e-01
9.59452212e-01 -4.61941332e-01 -2.19559148e-01 -4.89836512e-03
1.11368096e+00 -5.45971952e-02 3.22604656e-01 3.53202105e-01
5.67975163e-01 9.39038157e-01 9.21998143e-01 3.71936530e-01
6.28020108e-01 6.09543562e-01 -7.20965862e-02 2.26381153e-01
-2.95536935e-01 -6.77953541e-01 6.50539994e-02 1.03058422e+00
8.21228027e-01 1.25781909e-04 -7.55905211e-01 6.55040443e-01
-1.97093952e+00 -1.10417509e+00 -7.77442008e-02 2.07519174e+00
1.45087326e+00 1.12230167e-01 9.64886416e-03 1.96080551e-01
4.54851717e-01 2.00514227e-01 -4.71331954e-01 -3.67153198e-01
1.24325894e-01 6.06899142e-01 -1.64052740e-01 7.71252632e-01
-9.73476112e-01 1.15756238e+00 5.95511627e+00 1.16552937e+00
-4.49232727e-01 3.74233425e-01 1.54946834e-01 1.78861812e-01
-7.12881565e-01 2.36307636e-01 -1.05027950e+00 3.98072928e-01
7.46021688e-01 -4.36434329e-01 1.50532201e-01 1.08341753e+00
-3.02215487e-01 -2.72557456e-02 -1.27943861e+00 9.02866185e-01
5.73448777e-01 -1.43126106e+00 1.20478123e-01 -7.40662336e-01
8.24707031e-01 -3.21270764e-01 -1.94599405e-01 9.65855181e-01
4.50764328e-01 -7.11252809e-01 2.83097863e-01 5.74151933e-01
6.24416888e-01 -4.78260338e-01 4.98533517e-01 6.25417411e-01
-1.20354867e+00 -4.04716998e-01 -4.88118321e-01 -3.47917825e-01
-4.08172160e-02 2.29002565e-01 -6.51023686e-01 1.41552389e-01
3.34113002e-01 6.73214853e-01 -7.90600419e-01 7.33728945e-01
-7.68704295e-01 9.09900427e-01 -3.09051364e-03 -7.13869512e-01
1.85588121e-01 1.00910947e-01 3.44791502e-01 1.07211781e+00
-2.79857934e-01 3.22625577e-01 4.74095762e-01 1.02737987e+00
-3.07969246e-02 3.37241918e-01 -7.04269230e-01 1.37955740e-01
9.86209631e-01 1.15735972e+00 -4.44832891e-01 -5.79701006e-01
-7.12082565e-01 7.87772954e-01 6.44640923e-01 3.63287389e-01
-4.63395000e-01 -8.47374260e-01 3.32539290e-01 -1.17387526e-01
1.04915919e-02 1.57570824e-01 -4.59886670e-01 -1.22681057e+00
-1.62773460e-01 -7.69902468e-01 8.24190021e-01 -5.51721454e-01
-1.72757852e+00 1.86045200e-01 -5.30871302e-02 -1.16077268e+00
-4.65030402e-01 -4.10699517e-01 -9.71679747e-01 5.87397695e-01
-1.62310576e+00 -7.75682628e-01 5.62362559e-03 4.04316097e-01
9.20735598e-01 -3.23490709e-01 1.08372772e+00 2.27339193e-01
-3.19327027e-01 8.59847784e-01 -1.07617371e-01 -4.69328202e-02
9.45437551e-01 -1.33649790e+00 9.85408127e-02 5.87744772e-01
4.18111354e-01 1.02976227e+00 6.83140039e-01 -4.89320904e-01
-1.18517709e+00 -9.67067719e-01 1.82167149e+00 -1.02855968e+00
9.55261707e-01 -3.02215815e-01 -1.23697639e+00 3.54041189e-01
3.30173641e-01 -2.77992278e-01 1.11707664e+00 4.22851384e-01
-7.80297339e-01 -9.79351550e-02 -8.13041389e-01 7.32597470e-01
8.00171614e-01 -9.00793433e-01 -1.44054019e+00 5.99519968e-01
1.16618598e+00 7.21032768e-02 -4.07904387e-01 1.69768661e-01
5.55797406e-02 -6.68008506e-01 1.04216743e+00 -8.95590603e-01
5.02706110e-01 -1.73787579e-01 -2.32973710e-01 -1.12287593e+00
-2.89285779e-01 -5.34378469e-01 -6.60744667e-01 1.30939269e+00
3.79749537e-01 -4.26814526e-01 5.67284107e-01 9.39812064e-01
2.44847164e-02 -9.59443152e-01 -7.85756230e-01 -9.11422789e-01
3.55155990e-02 -3.50669235e-01 2.91396558e-01 1.16111720e+00
6.76916242e-01 1.04295421e+00 -1.06881291e-01 -4.45857376e-01
6.24228954e-01 3.19843471e-01 6.92511499e-01 -1.54602361e+00
-1.30673423e-01 -4.47526127e-01 1.71016604e-01 -1.29941249e+00
6.78224444e-01 -1.21928918e+00 1.49409667e-01 -1.61979425e+00
4.49202657e-01 -4.06195760e-01 -4.51599449e-01 4.97901052e-01
-9.15634871e-01 -1.60817578e-01 -1.17523588e-01 5.86560518e-02
-1.38328159e+00 7.11955488e-01 7.47331142e-01 -1.77279070e-01
-1.23005554e-01 -9.00292695e-02 -7.82813072e-01 7.34931409e-01
6.74654484e-01 -7.13624537e-01 -7.19184697e-01 -2.04691082e-01
3.76656115e-01 -2.53095180e-01 1.60944253e-01 -5.77787161e-01
7.51486540e-01 -1.36931136e-01 -1.91499427e-01 -5.91111124e-01
6.17258027e-02 -5.31939626e-01 -9.42743063e-01 4.14497077e-01
-1.15149748e+00 -3.81801546e-01 -3.15869927e-01 6.83183610e-01
-2.16915354e-01 -7.55027175e-01 5.90669930e-01 -4.05657217e-02
-9.04675782e-01 1.01146653e-01 -2.74596840e-01 7.49171555e-01
5.53473175e-01 -4.49737720e-02 -4.38901991e-01 -4.24237967e-01
-2.67522275e-01 7.06290066e-01 -2.32846178e-02 7.73665130e-01
8.79559517e-01 -1.61052656e+00 -6.18224502e-01 7.14072958e-02
9.39503610e-01 -3.52062523e-01 1.49438843e-01 5.76320112e-01
2.90533930e-01 6.50387228e-01 5.52396894e-01 -4.07052517e-01
-1.17917204e+00 8.17755938e-01 -1.33153021e-01 -3.72446120e-01
-4.85609382e-01 1.14689815e+00 -9.07176659e-02 -1.04863906e+00
5.86663485e-01 4.04534414e-02 -6.28279388e-01 1.81033283e-01
8.32217813e-01 2.43476763e-01 -2.04140857e-01 -2.06756126e-02
-2.26930603e-01 3.22143584e-01 -2.82696456e-01 3.80562507e-02
9.01769459e-01 -2.05161512e-01 -8.73454735e-02 7.76165605e-01
1.25802600e+00 -3.18929285e-01 -7.69583404e-01 -1.06364834e+00
7.21323431e-01 -5.16661167e-01 -3.03525567e-01 -6.62455440e-01
2.55120918e-02 9.11581695e-01 1.55230626e-01 9.06394646e-02
6.11438215e-01 3.98144364e-01 9.56883252e-01 1.01218867e+00
2.78835982e-01 -1.50717115e+00 8.14670205e-01 9.63453948e-01
5.85630774e-01 -1.46164572e+00 -4.34004307e-01 -2.18333304e-01
-5.84859133e-01 9.64773536e-01 1.07944143e+00 9.67394263e-02
4.69815075e-01 -5.73559590e-02 2.56696850e-01 -2.48385802e-01
-1.15388691e+00 -3.50424111e-01 3.63732159e-01 5.62800825e-01
5.50766826e-01 -2.55526125e-01 -5.72124362e-01 9.12069440e-01
8.18069279e-02 -1.85523987e-01 6.67399094e-02 9.72347438e-01
-1.02700305e+00 -1.15159130e+00 -1.87065080e-01 6.74290180e-01
1.24332786e-01 -3.57203811e-01 -4.86068398e-01 2.59317517e-01
-1.94529265e-01 1.45240819e+00 -1.06769189e-01 -4.17078465e-01
1.84854686e-01 6.88266754e-01 2.49738663e-01 -1.21014273e+00
-5.90779841e-01 -6.32794917e-01 5.12160212e-02 -4.14954036e-01
-1.36657462e-01 -3.45618904e-01 -1.30533862e+00 5.86674288e-02
-5.28653443e-01 4.84143406e-01 8.75517055e-02 1.39727402e+00
1.64632410e-01 4.49545830e-01 8.02663267e-01 -8.70611444e-02
-1.16940403e+00 -1.01620770e+00 -1.99390963e-01 6.10172510e-01
1.04384273e-01 -7.34944046e-01 -6.19762301e-01 -2.01278478e-01]
|
[10.853470802307129, 7.731945991516113]
|
5a83c446-4148-44b3-9c1d-f7ec8be47997
|
aerial-spectral-super-resolution-using
|
1712.08690
| null |
http://arxiv.org/abs/1712.08690v1
|
http://arxiv.org/pdf/1712.08690v1.pdf
|
Aerial Spectral Super-Resolution using Conditional Adversarial Networks
|
Inferring spectral signatures from ground based natural images has acquired a
lot of interest in applied deep learning. In contrast to the spectra of ground
based images, aerial spectral images have low spatial resolution and suffer
from higher noise interference. In this paper, we train a conditional
adversarial network to learn an inverse mapping from a trichromatic space to 31
spectral bands within 400 to 700 nm. The network is trained on AeroCampus, a
first of its kind aerial hyperspectral dataset. AeroCampus consists of high
spatial resolution color images and low spatial resolution hyperspectral images
(HSI). Color images synthesized from 31 spectral bands are used to train our
network. With a baseline root mean square error of 2.48 on the synthesized RGB
test data, we show that it is possible to generate spectral signatures in
aerial imagery.
|
['Matthew Hoffman', 'Emmett Ientilucci', 'Nilay Mokashi', 'Christopher Kanan', 'Aneesh Rangnekar']
|
2017-12-23
| null | null | null | null |
['spectral-super-resolution']
|
['computer-vision']
|
[ 1.21089280e+00 -1.99687198e-01 3.28899652e-01 -1.81648612e-01
-7.41344988e-01 -1.06747055e+00 2.77541727e-01 -5.78153551e-01
-3.34352642e-01 1.07699907e+00 -4.18029606e-01 -4.60223168e-01
-3.73715580e-01 -1.37894797e+00 -9.08573091e-01 -8.88506174e-01
-3.17240477e-01 -6.30684718e-02 -1.52211979e-01 -2.26316810e-01
-3.72011453e-01 8.16698074e-01 -1.49930894e+00 4.60824758e-01
1.01625752e+00 1.14439821e+00 1.95680618e-01 9.77625549e-01
4.94469821e-01 3.07961076e-01 -5.65053523e-01 5.51882423e-02
9.50019300e-01 -4.85469967e-01 -8.13748777e-01 4.25759673e-01
7.06977367e-01 -4.61909294e-01 -3.55662137e-01 1.71003830e+00
2.63105005e-01 7.23524690e-02 5.89589179e-01 -1.23796499e+00
-8.27789307e-01 7.15453446e-01 -5.43999732e-01 -2.99126297e-01
-9.78767350e-02 3.31918597e-01 8.12010288e-01 -3.99051249e-01
4.13530529e-01 7.45374382e-01 5.56464672e-01 1.55017287e-01
-1.45352554e+00 -5.08596838e-01 -3.41769069e-01 -5.71297063e-03
-1.62825561e+00 2.13180453e-01 7.38404930e-01 -4.37533289e-01
4.54549521e-01 4.56963003e-01 8.72679174e-01 9.64340210e-01
-1.39512062e-01 1.57857433e-01 1.67692924e+00 -3.78774583e-01
4.24394608e-02 -6.25629202e-02 -6.40246570e-01 6.74804688e-01
3.31134439e-01 6.81549728e-01 9.30551346e-03 1.78773403e-01
7.42536306e-01 4.76185195e-02 -7.14590192e-01 4.26580682e-02
-1.09452426e+00 8.14712524e-01 1.22126353e+00 4.96799275e-02
-5.42781472e-01 -1.10550672e-01 -3.60282093e-01 3.86366040e-01
1.76242590e-01 6.87230527e-01 -2.42207140e-01 6.91904366e-01
-8.92115533e-01 -1.00417323e-01 5.16706407e-01 6.74924016e-01
1.03343368e+00 5.25118113e-01 3.50279450e-01 8.35207045e-01
1.24173546e-02 1.15389168e+00 2.01342002e-01 -9.15112555e-01
-1.74995884e-02 4.41996336e-01 1.00010775e-01 -7.88596153e-01
-2.74540037e-01 -3.15622896e-01 -1.08834863e+00 6.94487453e-01
2.65961498e-01 -4.67439115e-01 -1.12197447e+00 1.45666814e+00
-2.26359591e-01 8.06644931e-02 3.48919868e-01 1.26555204e+00
4.64613348e-01 9.26236212e-01 -2.20016390e-01 2.63101440e-02
7.37347841e-01 -3.09484541e-01 -1.71154588e-01 -3.15100014e-01
7.87524320e-03 -6.19931817e-01 1.09378660e+00 6.24405682e-01
-6.34711385e-01 -4.08064753e-01 -1.42289174e+00 4.36409265e-01
-8.52815807e-01 3.25888932e-01 7.13128030e-01 7.74690449e-01
-9.35143471e-01 7.15120375e-01 -3.94413054e-01 -1.63399741e-01
5.48625886e-01 1.61159277e-01 -5.17907202e-01 -6.04664907e-02
-1.38398314e+00 5.42720675e-01 9.20916438e-01 1.37868971e-01
-8.88947964e-01 -6.35477722e-01 -6.53247654e-01 -9.28547978e-02
2.36835599e-01 -1.40810356e-01 6.29074872e-01 -1.56993008e+00
-1.46557117e+00 1.13510156e+00 7.49416173e-01 -4.78014052e-01
3.11461419e-01 1.14966914e-01 -7.59051383e-01 3.38407040e-01
-5.87930381e-02 5.26957095e-01 9.22273636e-01 -1.41151106e+00
-5.80074012e-01 -4.62832272e-01 2.70995408e-01 -8.19871202e-02
-2.72271842e-01 -3.56441557e-01 4.27883863e-01 -4.74352479e-01
2.44692340e-01 -1.34559619e+00 -7.03729019e-02 8.66495818e-02
-8.37162912e-01 8.98471057e-01 7.55872548e-01 -5.60102284e-01
2.43755132e-01 -2.14291382e+00 1.54035985e-01 4.68612105e-01
-3.45418900e-01 4.58946854e-01 -4.49027359e-01 1.53099641e-01
-6.00825906e-01 3.54306817e-01 -8.05674255e-01 7.91421771e-01
-1.21896498e-01 -1.18643697e-02 -5.60755849e-01 5.72017729e-01
2.64375746e-01 6.28813565e-01 -7.38914251e-01 2.29977831e-01
2.00908095e-01 5.01699090e-01 -1.34047896e-01 2.90578485e-01
-4.15092707e-01 3.21231097e-01 -2.10918233e-01 8.16752017e-01
1.18613291e+00 -5.65959364e-02 9.28164795e-02 -4.88648295e-01
-1.37893558e-01 -4.04072285e-01 -8.92204106e-01 1.59764838e+00
-3.36789310e-01 7.09280252e-01 3.04230511e-01 -8.70321929e-01
7.86221802e-01 3.96990217e-02 4.32941258e-01 -4.55928326e-01
1.52008533e-02 5.14017455e-02 1.07181571e-01 -2.46915892e-01
4.27611917e-01 -3.85944098e-01 1.67708337e-01 2.38355443e-01
-1.54373363e-01 -9.10489678e-01 8.31299368e-03 -3.42145264e-01
7.20598400e-01 1.63247690e-01 1.00136101e-01 -1.28735229e-01
5.65104663e-01 4.12746280e-01 1.69376612e-01 4.41061705e-01
1.39626592e-01 5.81524491e-01 2.72408426e-01 -3.09487909e-01
-1.09948885e+00 -1.32862651e+00 -3.53360862e-01 6.49759293e-01
-1.02275237e-01 3.54959249e-01 -6.74224496e-01 -2.16493204e-01
1.63880169e-01 6.97522342e-01 -6.28525853e-01 -1.98542058e-01
4.60198484e-02 -1.22232997e+00 8.84095728e-01 1.99083298e-01
1.19511867e+00 -1.01328170e+00 -6.51911438e-01 -1.60215944e-01
-1.09282201e-02 -1.25509846e+00 1.66735426e-01 2.69215375e-01
-6.22733057e-01 -1.38101196e+00 -5.02016962e-01 -3.59981239e-01
4.33952242e-01 2.94559300e-01 1.02296698e+00 -4.72001076e-01
-8.79076838e-01 3.27528268e-01 -4.27129358e-01 -6.15473449e-01
-5.60825288e-01 -1.60693154e-01 -1.01480812e-01 2.12050408e-01
-2.74426639e-02 -6.91652000e-01 -4.19765800e-01 6.19387766e-03
-1.41136456e+00 7.34283924e-02 7.50122726e-01 9.91791546e-01
6.59569383e-01 5.92097938e-01 -8.58715698e-02 -7.58621931e-01
1.39652848e-01 -2.67472148e-01 -1.20352054e+00 1.69827431e-01
-1.59389570e-01 -2.89562404e-01 9.83757675e-01 -8.60736743e-02
-9.22220469e-01 4.42291826e-01 2.31375501e-01 -1.93763494e-01
-6.21735573e-01 7.20852137e-01 1.68967526e-02 -4.63054806e-01
1.17476809e+00 3.66856188e-01 -2.30988637e-01 5.43989092e-02
2.83511311e-01 6.17826283e-01 8.21265996e-01 -3.40800017e-01
1.33329093e+00 6.24538004e-01 3.65862966e-01 -1.39940369e+00
-9.39958930e-01 1.34260416e-01 -7.56552100e-01 -1.80460677e-01
1.01151347e+00 -9.86330032e-01 -4.83728588e-01 5.47349095e-01
-6.21056616e-01 -7.79536426e-01 -1.80732459e-01 5.05706906e-01
-5.46378911e-01 6.29958659e-02 -3.43503892e-01 -6.51195109e-01
-9.24074203e-02 -8.59078765e-01 6.46000624e-01 1.95958644e-01
6.13731027e-01 -7.62659073e-01 -3.80410522e-01 1.65305287e-01
3.22608382e-01 1.01940954e+00 9.14295554e-01 8.75106305e-02
-9.39950049e-01 -1.25762507e-01 -6.19182169e-01 8.18120778e-01
4.94395316e-01 3.63601148e-01 -1.12354624e+00 -2.78077960e-01
-1.64165914e-01 -7.72191167e-01 8.49631906e-01 4.22825843e-01
1.59193587e+00 -6.57201260e-02 2.28269577e-01 1.28999805e+00
2.13242245e+00 2.50883847e-01 7.38630474e-01 3.77388030e-01
6.32018864e-01 4.88006860e-01 5.86539984e-01 3.68826747e-01
-5.53093970e-01 6.97265938e-02 1.21180367e+00 -5.81947148e-01
1.45652846e-01 1.46730110e-01 9.58799645e-02 -3.09429646e-01
-3.51252317e-01 -2.94596493e-01 -9.51992929e-01 3.10615808e-01
-1.11078942e+00 -1.14277267e+00 -3.18681031e-01 2.22845411e+00
8.11366498e-01 -3.83117348e-01 -2.77183473e-01 3.54245454e-01
7.25585282e-01 3.72083157e-01 -6.09719396e-01 1.99290454e-01
-7.39612699e-01 5.23390114e-01 1.11923158e+00 4.60613400e-01
-1.45732617e+00 9.05462384e-01 5.91330242e+00 -1.17615266e-02
-1.55270839e+00 -4.45588857e-01 4.61647987e-01 1.33592114e-01
-8.99291486e-02 -2.77522177e-01 9.08628181e-02 1.78520516e-01
7.96955645e-01 -1.84497580e-01 1.20662773e+00 6.19719863e-01
-1.95177093e-01 -1.06807612e-01 -6.32448554e-01 8.56337965e-01
-1.44335762e-01 -1.11957061e+00 1.76186472e-01 1.42544761e-01
1.02229166e+00 2.98512340e-01 6.07160032e-01 -3.11955303e-01
7.54073739e-01 -1.41387677e+00 3.33944529e-01 5.44441998e-01
1.44140351e+00 -9.12734628e-01 4.94665444e-01 2.96208948e-01
-8.78963351e-01 -2.44723141e-01 -7.55453408e-01 2.22120970e-01
-4.76265222e-01 4.95205909e-01 -8.34093630e-01 7.22871184e-01
7.69622266e-01 5.88561416e-01 -5.02524853e-01 7.27967918e-01
-4.05762762e-01 6.17930114e-01 -2.91370511e-01 4.41616714e-01
5.56157351e-01 -8.05014908e-01 3.42157722e-01 9.14599776e-01
7.25095510e-01 5.32061279e-01 1.94150954e-01 1.24515975e+00
-2.17001781e-01 -2.97072411e-01 -1.07486880e+00 -6.81804717e-01
1.37715921e-01 1.28114378e+00 -4.09664959e-01 -1.83233693e-02
-2.43427932e-01 1.19359386e+00 -3.36730242e-01 6.91665292e-01
-9.14394736e-01 -4.55241531e-01 6.14936352e-01 -2.02341974e-01
5.80402017e-02 -1.73856065e-01 5.92690296e-02 -1.07651055e+00
-5.21087170e-01 -1.03476608e+00 8.15346837e-02 -1.31193113e+00
-1.16584933e+00 7.91070163e-01 -1.20689347e-01 -1.37700582e+00
-1.16575100e-02 -1.28983331e+00 -2.00769350e-01 1.33937991e+00
-1.78091228e+00 -1.27849376e+00 -1.08870888e+00 8.73391807e-01
-1.01718247e-01 -3.63175333e-01 1.29236269e+00 -2.18897730e-01
-3.00879240e-01 -5.09126149e-02 3.34226429e-01 3.41194987e-01
4.83054459e-01 -1.48255408e+00 7.28698522e-02 9.44301069e-01
3.91471535e-02 7.30744973e-02 5.39998770e-01 -2.90209174e-01
-1.37168002e+00 -1.70990598e+00 -1.25686988e-01 2.27712914e-01
7.52705395e-01 3.24315310e-01 -6.68524861e-01 7.32753217e-01
2.51442850e-01 3.11848432e-01 9.28225100e-01 -6.87606573e-01
-4.36134070e-01 -5.15423298e-01 -1.32838798e+00 4.63014007e-01
7.66269207e-01 -7.30467916e-01 -2.80669611e-02 8.85692060e-01
2.99935013e-01 -2.01943383e-01 -1.01901984e+00 5.38612723e-01
3.21714848e-01 -1.27760875e+00 1.12351155e+00 -3.25080246e-01
6.49278939e-01 -4.70829099e-01 -6.91353858e-01 -1.80174387e+00
-2.97341198e-01 -1.65600970e-01 9.92087960e-01 5.10857284e-01
4.52389568e-01 -6.55453086e-01 7.81983972e-01 1.06441289e-01
-6.92840144e-02 2.32027382e-01 -3.33302379e-01 -8.73408020e-01
2.16931760e-01 -9.96292010e-02 6.97248220e-01 1.15285647e+00
-7.02770650e-01 -3.71674001e-02 -1.58143967e-01 9.84240890e-01
1.02228165e+00 5.59450507e-01 5.61985254e-01 -1.55266857e+00
-2.69701421e-01 -3.02407056e-01 -2.49471977e-01 -2.20155433e-01
3.92856061e-01 -1.03307998e+00 1.23057835e-01 -1.21685791e+00
-1.47045061e-01 -1.79119080e-01 -2.38454282e-01 7.22508490e-01
2.91752547e-01 8.62245142e-01 9.50095206e-02 -2.84133971e-01
4.69280481e-01 2.75744140e-01 1.14336741e+00 -7.17755020e-01
-4.13663611e-02 -1.37559026e-01 -2.37332433e-01 7.07779586e-01
1.07706499e+00 -2.63815314e-01 -5.38204789e-01 -3.36105019e-01
2.11090937e-01 1.49098873e-01 8.51659417e-01 -1.29538929e+00
-2.57202238e-01 -6.31850839e-01 6.68757558e-01 -3.77671510e-01
4.99149382e-01 -1.20757031e+00 5.07489979e-01 5.72720289e-01
-2.33430326e-01 -5.33914745e-01 4.52065498e-01 2.41718173e-01
-2.01348215e-01 -1.57020569e-01 1.21145809e+00 -4.81717229e-01
-9.48920846e-01 4.99597818e-01 -3.38142276e-01 -3.80770445e-01
9.82786655e-01 -1.47482246e-01 -4.35100406e-01 -2.49891639e-01
-4.86338675e-01 -4.24270570e-01 7.07468033e-01 -9.47817415e-03
6.51475012e-01 -1.11151898e+00 -9.12498772e-01 5.27355015e-01
9.44684222e-02 -2.31316268e-01 1.32695481e-01 2.03470841e-01
-1.15666986e+00 2.09546790e-01 -8.94693613e-01 -5.23760974e-01
-1.12782681e+00 2.94555962e-01 8.77940118e-01 3.61054420e-01
-2.23231182e-01 9.33381140e-01 -2.19424292e-02 -5.09226680e-01
-4.19745892e-01 -2.04023689e-01 -1.77617595e-02 -1.80217907e-01
2.13678434e-01 8.39801729e-02 -4.38304655e-02 -6.61166728e-01
-9.76606160e-02 4.92194474e-01 7.98891425e-01 -1.97498217e-01
1.49175441e+00 4.25880432e-01 -4.76933330e-01 2.61216491e-01
1.13057315e+00 -1.08324476e-02 -1.32005143e+00 -4.07637358e-02
-5.14017642e-01 -7.74630725e-01 3.77652407e-01 -1.18266523e+00
-1.27810824e+00 8.55612934e-01 8.66183996e-01 6.98031306e-01
1.74931502e+00 -7.54711509e-01 1.92514673e-01 8.10737729e-01
3.74170572e-01 -8.98995876e-01 -2.41082668e-01 5.44265091e-01
9.58173692e-01 -1.40140975e+00 8.97455364e-02 -5.17220259e-01
-4.13597167e-01 1.30522001e+00 3.39371443e-01 -2.79832959e-01
4.04289305e-01 2.37566203e-01 2.16063321e-01 5.99006377e-02
-1.71373248e-01 -5.58991253e-01 1.76543742e-01 1.00324178e+00
2.96976268e-01 3.70429754e-01 3.49701554e-01 -7.57410750e-02
-4.44591850e-01 -8.94965678e-02 7.64347672e-01 3.35884750e-01
-3.85675043e-01 -6.70767725e-01 -5.94267309e-01 3.24000686e-01
-3.06777358e-01 -2.92715669e-01 -6.04580224e-01 8.47965181e-01
2.37539127e-01 8.72959673e-01 -1.75668627e-01 -5.11538923e-01
1.47844449e-01 -1.92409772e-02 5.39408207e-01 -3.87629420e-01
-1.59690544e-01 -2.40717471e-01 -1.51988134e-01 -5.26619554e-01
-6.50694668e-01 -2.95253605e-01 -9.92554486e-01 -3.15452844e-01
-3.35336314e-03 -1.92437116e-02 8.80306304e-01 5.05158007e-01
-1.74029887e-01 6.33241057e-01 8.63129616e-01 -9.40592051e-01
-4.88174915e-01 -8.27418745e-01 -1.35714829e+00 3.02119136e-01
4.41835999e-01 -1.79446563e-01 -4.92822617e-01 3.72627765e-01]
|
[10.166072845458984, -1.9993056058883667]
|
bc857548-3bf6-4b0a-a379-ae5a597c8bbc
|
determinant-free-fermionic-wave-function
|
2108.08631
| null |
https://arxiv.org/abs/2108.08631v2
|
https://arxiv.org/pdf/2108.08631v2.pdf
|
Determinant-free fermionic wave function using feed-forward neural networks
|
We propose a general framework for finding the ground state of many-body fermionic systems by using feed-forward neural networks. The anticommutation relation for fermions is usually implemented to a variational wave function by the Slater determinant (or Pfaffian), which is a computational bottleneck because of the numerical cost of $O(N^3)$ for $N$ particles. We bypass this bottleneck by explicitly calculating the sign changes associated with particle exchanges in real space and using fully connected neural networks for optimizing the rest parts of the wave function. This reduces the computational cost to $O(N^2)$ or less. We show that the accuracy of the approximation can be improved by optimizing the "variance" of the energy simultaneously with the energy itself. We also find that a reweighting method in Monte Carlo sampling can stabilize the calculation. These improvements can be applied to other approaches based on variational Monte Carlo methods. Moreover, we show that the accuracy can be further improved by using the symmetry of the system, the representative states, and an additional neural network implementing a generalized Gutzwiller-Jastrow factor. We demonstrate the efficiency of the method by applying it to a two-dimensional Hubbard model.
|
['Yukitoshi Motome', 'Yasuyuki Kato', 'Koji Inui']
|
2021-08-19
| null | null | null | null |
['variational-monte-carlo']
|
['miscellaneous']
|
[ 7.00738877e-02 -3.16584557e-01 1.10318279e-02 -2.60166675e-01
-4.50527161e-01 -2.32384712e-01 5.89363575e-01 -1.26015216e-01
-7.49720156e-01 1.13724053e+00 -1.65050134e-01 -3.77484232e-01
-1.59537226e-01 -1.13042498e+00 -6.84595227e-01 -1.19293487e+00
-2.03916863e-01 5.32164037e-01 2.13550311e-02 -5.28219283e-01
3.76078278e-01 5.74667275e-01 -1.54759884e+00 1.22289158e-01
6.60755932e-01 1.05320084e+00 -3.03706884e-01 5.44269383e-01
1.40663102e-01 6.46492839e-01 -2.31872186e-01 -1.02031969e-01
5.32727957e-01 -7.40554094e-01 -1.04787362e+00 -5.35089672e-01
3.20797294e-01 -4.03408669e-02 -5.27008533e-01 1.54827595e+00
4.02964622e-01 7.18215942e-01 5.78696489e-01 -8.07045698e-01
-5.53367794e-01 6.01951838e-01 -5.31011745e-02 -8.18459094e-02
-2.54521012e-01 2.14351073e-01 1.16865778e+00 -5.99910975e-01
6.33236468e-01 9.77888048e-01 8.81360650e-01 6.52781069e-01
-1.51394856e+00 -4.39958394e-01 -3.51101995e-01 5.00013232e-01
-1.62069356e+00 -5.40126204e-01 7.96494424e-01 -1.25472769e-01
1.49673975e+00 4.10278231e-01 8.47811282e-01 3.58629078e-01
5.56261122e-01 1.08968876e-01 1.06846786e+00 -9.35071290e-01
5.87530732e-01 -3.32126051e-01 4.84648377e-01 1.03770149e+00
4.49839503e-01 4.89811182e-01 -6.92719072e-02 -3.23532104e-01
4.88935143e-01 8.18933994e-02 -1.76384807e-01 -6.55628681e-01
-8.64403367e-01 1.11942244e+00 6.94123387e-01 3.11608702e-01
-2.81481475e-01 8.10634971e-01 5.17360449e-01 2.42611825e-01
1.35613009e-01 8.30852270e-01 -2.52941519e-01 6.59459755e-02
-1.15152526e+00 4.30580080e-01 1.00420833e+00 1.06248938e-01
8.84991050e-01 1.93954676e-01 -3.37389484e-02 3.07635337e-01
1.51628315e-01 5.84053099e-01 2.51041114e-01 -1.22497725e+00
-1.50175869e-01 2.84598470e-01 4.52991515e-01 -5.42520583e-01
-4.01478678e-01 -2.00452343e-01 -1.20030928e+00 5.78827202e-01
5.64687788e-01 -2.79192835e-01 -8.84756029e-01 1.82760966e+00
9.20758098e-02 -5.26082754e-01 -2.59307921e-01 8.07870507e-01
3.12665641e-01 7.84621358e-01 -1.96829721e-01 -4.61222112e-01
1.12213278e+00 -1.02501845e+00 -5.40752530e-01 1.58681497e-01
5.86812139e-01 -2.79448330e-01 6.77972138e-01 1.96074933e-01
-1.49966574e+00 -2.21106797e-01 -1.15420747e+00 -3.33528034e-02
-4.25363183e-01 -1.11685015e-01 9.62359190e-01 6.63258791e-01
-1.16611826e+00 1.42184329e+00 -1.17639482e+00 -2.04990786e-02
3.83404084e-02 8.82228851e-01 -5.30356914e-02 2.96345830e-01
-1.36672771e+00 1.09424591e+00 2.95908272e-01 1.61237776e-01
-4.75069880e-01 -2.65026659e-01 -6.42472684e-01 2.22129241e-01
1.68064415e-01 -6.47487223e-01 1.08100796e+00 -8.40835690e-01
-1.61644697e+00 3.82557809e-01 -4.40219641e-01 -4.14301842e-01
1.33029535e-01 4.90089953e-01 -2.99929857e-01 -1.26380801e-01
-2.08717600e-01 3.45672429e-01 4.49711233e-01 -6.64478004e-01
1.94316372e-01 -2.22046539e-01 3.33864279e-02 -1.48907319e-01
8.80722851e-02 -8.98038000e-02 7.04591125e-02 -3.23294066e-02
2.47861564e-01 -1.06862390e+00 -6.62274122e-01 -3.13594103e-01
-3.44400108e-01 -1.00457467e-01 2.85982013e-01 -3.47658783e-01
1.08766568e+00 -1.69782960e+00 1.62235603e-01 6.44815147e-01
2.50531524e-01 2.13780150e-01 1.95464697e-02 6.48197889e-01
-2.19741225e-01 -1.15300335e-01 -2.88828701e-01 1.66543603e-01
6.23986907e-02 7.00109303e-02 2.72517651e-02 6.15078390e-01
1.68236031e-03 1.01680589e+00 -7.06640303e-01 -3.65446694e-02
7.62119740e-02 5.61238766e-01 -9.99695957e-01 -4.77858394e-01
-1.44938469e-01 2.34266341e-01 -1.02493703e-01 8.03489089e-02
6.38478816e-01 -6.53930306e-01 5.39506257e-01 -3.74331713e-01
-4.06931967e-01 7.10609496e-01 -1.17052102e+00 1.35288095e+00
-3.58784080e-01 3.88536394e-01 1.77454248e-01 -1.41200089e+00
3.96525085e-01 1.56734943e-01 3.77716511e-01 -7.03319967e-01
3.84597480e-01 3.49599153e-01 6.10235870e-01 2.59697624e-02
5.16650796e-01 -6.49980485e-01 -7.78555423e-02 7.65834987e-01
6.94339126e-02 -2.00216845e-01 4.92599845e-01 -7.96587095e-02
1.05259776e+00 -2.07553983e-01 5.26606858e-01 -7.71984220e-01
4.70649034e-01 1.51522920e-01 2.13690400e-01 1.04391897e+00
-2.16042697e-01 3.12160254e-02 4.71215487e-01 -9.06525493e-01
-1.30215716e+00 -7.48033226e-01 -3.63688141e-01 7.42271543e-01
3.95434303e-03 -3.97904903e-01 -9.57291663e-01 -1.75306723e-01
-1.39714643e-01 7.96694577e-01 -7.44579673e-01 -3.12615484e-01
-7.76717246e-01 -1.34377110e+00 2.36157507e-01 3.39894533e-01
5.48371732e-01 -1.27007580e+00 -7.46043146e-01 2.37516880e-01
1.92993388e-01 -2.40443230e-01 -1.57534346e-01 9.12163377e-01
-8.73292863e-01 -7.81029046e-01 -2.48885497e-01 -3.71694505e-01
6.98646545e-01 -1.95870876e-01 9.86009777e-01 2.50374228e-01
-3.52132916e-01 -3.63918811e-01 3.87720913e-01 4.82054323e-01
-4.52992231e-01 -1.88508220e-02 3.90119672e-01 -5.77205539e-01
4.86028850e-01 -6.66706443e-01 -5.83946526e-01 -8.75499174e-02
-5.31678736e-01 1.11795124e-02 7.97465667e-02 1.32980072e+00
4.96134996e-01 2.72420615e-01 -1.22612461e-01 -6.01252675e-01
6.27228498e-01 4.42960747e-02 -9.93606925e-01 -3.68804410e-02
-6.74505234e-01 8.24962556e-01 8.90523553e-01 -1.50488213e-01
-7.04793572e-01 2.94436067e-02 -3.22345287e-01 5.61003685e-02
3.61541033e-01 3.03842992e-01 2.39664868e-01 -7.34519839e-01
7.04471111e-01 1.97527215e-01 -8.76015574e-02 -2.75897235e-01
4.43083823e-01 2.59178728e-01 2.28916675e-01 -6.75179780e-01
5.07805824e-01 4.83665645e-01 6.40983582e-01 -5.36567450e-01
-6.89313650e-01 6.22911565e-02 -4.78343606e-01 1.72045842e-01
6.97823286e-01 -3.72985035e-01 -1.61487377e+00 2.17496321e-01
-1.19706738e+00 -1.22064084e-01 -6.15157008e-01 6.24987304e-01
-4.36117560e-01 3.17597985e-01 -9.90719736e-01 -9.68102157e-01
-5.35867929e-01 -1.41603696e+00 3.99617255e-01 -9.07983184e-02
-3.31114382e-01 -8.66499305e-01 4.44870889e-01 -1.02402426e-01
8.20932448e-01 -2.62952626e-01 1.42918372e+00 -4.58040476e-01
-5.01938939e-01 -3.75711799e-01 -4.04185385e-01 2.25968421e-01
-3.28657657e-01 -2.07450867e-01 -6.07032895e-01 -3.38624775e-01
1.50027126e-01 -3.10232759e-01 1.38502359e+00 6.35908186e-01
1.20967662e+00 -5.00927866e-01 -3.10564369e-01 4.52437252e-01
1.60423911e+00 4.04407561e-01 4.90219086e-01 8.04793555e-03
6.42340541e-01 8.17520097e-02 -5.56945324e-01 2.30283275e-01
-7.69092068e-02 6.42093599e-01 2.39959702e-01 2.41814330e-01
1.50008744e-03 9.26238820e-02 2.17424482e-01 1.06529880e+00
-6.79755688e-01 5.92241734e-02 -8.67042005e-01 1.76701427e-01
-1.75343573e+00 -1.48930502e+00 -1.32224455e-01 2.28314710e+00
9.25351322e-01 -1.11214491e-02 -1.11487776e-01 1.70579210e-01
4.88636166e-01 2.13188499e-01 -5.20685434e-01 -7.26836085e-01
1.99778184e-01 9.37777758e-01 8.18636596e-01 9.05491889e-01
-9.98435378e-01 7.36611843e-01 7.92980576e+00 8.39666069e-01
-1.25253403e+00 3.48485976e-01 3.56593341e-01 -4.42572147e-01
-2.08001778e-01 1.83634028e-01 -6.71666920e-01 4.91338909e-01
1.16237056e+00 1.07744232e-01 1.19642174e+00 6.59992456e-01
1.40855312e-01 -1.15494497e-01 -9.99777257e-01 7.92142630e-01
-2.78653830e-01 -1.72196591e+00 -1.72400564e-01 1.30960613e-01
9.55858409e-01 3.34971428e-01 -2.59825349e-01 2.80311316e-01
5.78141809e-01 -1.06771481e+00 4.47745919e-01 3.42155814e-01
7.24910736e-01 -8.23155284e-01 5.88782907e-01 1.20287031e-01
-8.51789892e-01 1.40773311e-01 -5.65247416e-01 -5.80487907e-01
1.15435638e-01 7.04807997e-01 -1.69846132e-01 -1.46298520e-02
3.19590181e-01 4.73639742e-02 -4.23332304e-02 8.20835531e-01
1.15320839e-01 2.96278000e-01 -5.07639825e-01 -6.03735745e-01
4.95519280e-01 -6.31178319e-01 2.51125872e-01 6.99200511e-01
4.10931319e-01 6.83108345e-02 -1.01806253e-01 1.19632173e+00
-1.93883985e-01 -3.09319403e-02 -3.47135425e-01 -2.95886368e-01
1.11871228e-01 9.57265198e-01 -7.24876761e-01 -4.15408164e-01
-8.05981681e-02 8.43823016e-01 3.86933953e-01 3.85264993e-01
-4.41262394e-01 -7.49540269e-01 5.64117312e-01 9.94325802e-03
5.92281342e-01 -2.49756932e-01 -1.03226431e-01 -1.30497420e+00
-2.00187325e-01 -6.29744112e-01 -1.29174694e-01 -5.36468387e-01
-1.09640229e+00 5.36135912e-01 -1.41129762e-01 -4.63328630e-01
-3.98357511e-01 -1.03250933e+00 -7.26385415e-01 1.01379097e+00
-9.64290857e-01 -4.11908358e-01 4.30500776e-01 3.85007024e-01
-3.51884961e-01 1.05462130e-02 1.18079889e+00 2.30577618e-01
-6.49471223e-01 3.56516510e-01 7.73044527e-01 -3.59902866e-02
1.05871648e-01 -1.02980614e+00 4.94540483e-01 6.33071542e-01
-1.79900471e-02 1.03745282e+00 8.74732137e-01 -5.19445717e-01
-1.64952374e+00 -6.77950799e-01 1.13300371e+00 1.57461669e-02
8.16393793e-01 -3.71676326e-01 -5.84803820e-01 5.11288643e-01
2.94768959e-01 2.23495394e-01 4.17561591e-01 3.87727559e-01
-1.48484290e-01 7.18223229e-02 -1.17128778e+00 6.99458599e-01
1.02407694e+00 -6.77931905e-01 -1.40288502e-01 5.54383218e-01
3.51293862e-01 -1.08792916e-01 -4.62815255e-01 4.60346378e-02
6.84627175e-01 -1.01498628e+00 9.41941917e-01 -9.77566123e-01
3.74950469e-01 -2.27103785e-01 -3.36032778e-01 -1.31630361e+00
-6.17212713e-01 -5.73735058e-01 -1.47809178e-01 9.04672593e-02
3.85206848e-01 -6.72761500e-01 8.51460874e-01 6.56312644e-01
2.79307216e-01 -6.97853684e-01 -1.32155752e+00 -6.08567595e-01
5.03906190e-01 -4.28846925e-01 5.32516360e-01 8.32248092e-01
4.29349959e-01 3.28613132e-01 -5.59339762e-01 -2.28041187e-01
6.11334026e-01 3.36813778e-01 -9.08177048e-02 -1.25077331e+00
-6.46198392e-01 -6.06705248e-01 -2.97566861e-01 -8.14397633e-01
2.55659729e-01 -1.30462325e+00 1.10919848e-01 -1.04364181e+00
4.68488723e-01 -1.70243233e-01 -5.23456037e-01 4.53288972e-01
2.82320946e-01 4.62409586e-01 -7.22817779e-02 1.61646068e-01
-6.82111323e-01 5.58877409e-01 1.07056916e+00 -1.00669295e-01
-1.25739397e-02 -2.95874804e-01 -3.88204485e-01 9.00174201e-01
6.87834084e-01 -6.08431637e-01 2.75817245e-01 -1.64365098e-01
7.29737103e-01 1.35056093e-01 5.63787758e-01 -1.10605633e+00
2.62530863e-01 -4.04389836e-02 4.21638578e-01 -4.12711978e-01
5.74691415e-01 -3.73697728e-01 1.94984049e-01 9.84093249e-01
-3.37874055e-01 -5.80942072e-02 7.13247359e-02 1.33365601e-01
8.07903707e-02 -7.15920091e-01 9.68753517e-01 -4.88805354e-01
-1.11673571e-01 7.66603798e-02 -5.76305985e-01 -1.70794770e-01
4.15496767e-01 1.66011974e-01 -1.67356029e-01 -2.43134350e-01
-5.97862959e-01 -2.85843760e-01 5.98444521e-01 -6.15033567e-01
2.01920811e-02 -1.61110008e+00 -1.30892158e-01 5.55051386e-01
-4.47964638e-01 -8.03348422e-01 2.19465807e-01 1.03110015e+00
-7.82709956e-01 7.22821355e-01 -3.55326116e-01 -1.23705201e-01
-8.22467089e-01 3.94525737e-01 7.43122578e-01 -6.19004846e-01
-2.67773837e-01 6.63913012e-01 -3.48805636e-01 -4.44242597e-01
-3.46842498e-01 -2.14176834e-01 4.30988520e-01 -3.28720659e-01
5.07228494e-01 4.91867870e-01 2.82155305e-01 -5.53663790e-01
-6.15230799e-01 4.23609585e-01 -3.28632444e-02 -3.31925213e-01
1.41543043e+00 3.94630790e-01 -7.02874601e-01 1.67204738e-01
1.33941901e+00 1.68409169e-01 -8.40535879e-01 -6.52449429e-02
-4.96042401e-01 3.47278640e-02 4.12655443e-01 -6.29996002e-01
-1.20453382e+00 1.07032394e+00 5.87415040e-01 4.80260104e-01
5.58667064e-01 -3.10423851e-01 7.85365939e-01 1.18537116e+00
4.81749535e-01 -1.01919091e+00 -7.09312618e-01 7.53288865e-01
3.28103453e-01 -1.05932474e+00 7.39215910e-02 2.32699975e-01
-7.57530332e-03 1.14476621e+00 3.39383632e-02 -5.29404283e-01
8.49345028e-01 1.68696225e-01 -4.78427798e-01 -4.11813110e-01
-5.90356052e-01 8.78644586e-02 2.90097028e-01 -9.73346382e-02
6.39943302e-01 2.94808626e-01 -6.08385503e-01 3.05416346e-01
-3.15024704e-01 2.51199529e-02 3.20995659e-01 6.93171263e-01
-5.18430710e-01 -1.36672854e+00 -1.32645458e-01 6.56411529e-01
-3.42625856e-01 -5.50316155e-01 -3.92147526e-02 6.07318759e-01
1.09709769e-01 6.61258221e-01 2.04757363e-01 -1.40297964e-01
-2.07039714e-01 5.24065018e-01 9.32166040e-01 -2.32187584e-01
-5.15982926e-01 -3.42642933e-01 6.98531047e-02 -6.47692382e-01
-4.35351819e-01 -4.40297693e-01 -1.27082455e+00 -9.93218601e-01
-4.38819170e-01 5.51900327e-01 5.83400488e-01 1.12385237e+00
2.91371971e-01 3.85329396e-01 1.77135125e-01 -1.06911981e+00
-8.39416623e-01 -6.94961846e-01 -6.47282958e-01 9.99210998e-02
2.24914595e-01 -6.53322399e-01 -4.41630960e-01 -5.01811206e-01]
|
[5.504789352416992, 5.0378899574279785]
|
014c7024-0603-4ca4-9aa2-9fed9cad6751
|
identifying-computer-translated-paragraphs
|
1812.10896
| null |
http://arxiv.org/abs/1812.10896v1
|
http://arxiv.org/pdf/1812.10896v1.pdf
|
Identifying Computer-Translated Paragraphs using Coherence Features
|
We have developed a method for extracting the coherence features from a
paragraph by matching similar words in its sentences. We conducted an
experiment with a parallel German corpus containing 2000 human-created and 2000
machine-translated paragraphs. The result showed that our method achieved the
best performance (accuracy = 72.3%, equal error rate = 29.8%) when it is
compared with previous methods on various computer-generated text including
translation and paper generation (best accuracy = 67.9%, equal error rate =
32.0%). Experiments on Dutch, another rich resource language, and a low
resource one (Japanese) attained similar performances. It demonstrated the
efficiency of the coherence features at distinguishing computer-translated from
human-created paragraphs on diverse languages.
|
['Junichi Yamagishi', 'Ngoc-Dung T. Tieu', 'Hoang-Quoc Nguyen-Son', 'Isao Echizen', 'Huy H. Nguyen']
|
2018-12-28
| null |
https://aclanthology.org/Y18-1056
|
https://aclanthology.org/Y18-1056.pdf
|
paclic-2018-12
|
['paper-generation']
|
['natural-language-processing']
|
[ 1.72507875e-02 1.52699426e-02 -1.98390171e-01 -2.49181199e-03
-1.35997975e+00 -5.99063277e-01 1.11217833e+00 -2.20995769e-01
-4.40124810e-01 1.39511752e+00 6.05775774e-01 -2.67319471e-01
3.18194151e-01 -6.98364675e-01 -1.59372196e-01 -3.77589405e-01
1.85675412e-01 7.40401566e-01 1.28761664e-01 -3.79281998e-01
9.96651173e-01 1.95649400e-01 -1.09375703e+00 4.61330146e-01
1.36674833e+00 9.32617337e-02 5.64803720e-01 8.04507136e-01
-4.41717595e-01 4.14675087e-01 -1.19388950e+00 -5.72179914e-01
-1.58144593e-01 -6.41740918e-01 -1.21457934e+00 1.94995955e-01
3.36493015e-01 1.04752190e-01 -1.02473088e-01 1.01933968e+00
4.62336600e-01 -1.67738929e-01 1.00702167e+00 -8.24539244e-01
-1.06007004e+00 8.28076303e-01 -6.53944492e-01 4.95980322e-01
9.99216855e-01 -3.59322757e-01 8.66063178e-01 -1.07125866e+00
1.00118673e+00 1.13390183e+00 5.59487581e-01 3.79798651e-01
-9.87923503e-01 -6.15667582e-01 -6.40483201e-01 -9.74014178e-02
-1.46553957e+00 -4.38255519e-01 3.41688871e-01 -5.09327352e-01
1.42735696e+00 3.42012614e-01 4.46814328e-01 1.06592500e+00
6.92409158e-01 6.11513376e-01 1.40365076e+00 -9.33294356e-01
-3.65022480e-01 5.40892839e-01 3.32388550e-01 5.20283282e-01
3.64211470e-01 -4.06667501e-01 -3.09291691e-01 -2.72739172e-01
4.07204032e-01 -9.18080091e-01 -2.74667919e-01 7.82301188e-01
-1.58200169e+00 7.60016084e-01 -3.99995148e-01 1.04047000e+00
-2.11971909e-01 -7.14010119e-01 3.59784454e-01 5.26989758e-01
5.79723299e-01 6.87291682e-01 -4.59297419e-01 -5.32377958e-01
-1.03955400e+00 1.32180676e-01 1.27851510e+00 1.56236362e+00
4.93829280e-01 -1.47632137e-01 -3.63126993e-01 1.10264051e+00
-1.26034796e-01 9.60481346e-01 1.13064468e+00 -4.88609016e-01
1.15347314e+00 3.65458310e-01 4.24665093e-01 -1.26224720e+00
-3.26161742e-01 -4.38304782e-01 -9.03940976e-01 -5.77028990e-01
1.57656938e-01 -3.55126441e-01 -4.95418727e-01 1.33241594e+00
-2.81110287e-01 -7.56559253e-01 5.18436790e-01 4.33249176e-01
9.53893721e-01 1.22606921e+00 -4.10615176e-01 -7.25790739e-01
1.23922575e+00 -1.15483689e+00 -1.03069520e+00 1.27757892e-01
4.83875811e-01 -1.64607322e+00 1.09210062e+00 2.72825122e-01
-1.30816948e+00 -7.31170774e-01 -9.26456690e-01 1.04542755e-01
-1.71209723e-01 4.13202643e-01 1.26446247e-01 5.92684567e-01
-9.99255478e-01 4.40294385e-01 -1.04690336e-01 -8.12113464e-01
-2.63061076e-01 1.70000374e-01 -5.16791403e-01 2.90475339e-01
-1.33614457e+00 1.15809393e+00 3.93907040e-01 -2.57076532e-01
-3.32771838e-02 4.00486887e-02 -3.25406611e-01 -9.15918052e-02
-2.54783273e-01 -8.84680927e-01 1.20437849e+00 -1.10846674e+00
-1.62279487e+00 1.06829369e+00 -3.74471098e-01 -7.77230635e-02
4.59139496e-01 -2.51753241e-01 -1.03680503e+00 2.37844408e-01
6.01729214e-01 2.40817219e-01 4.23286200e-01 -7.90657043e-01
-8.96839678e-01 2.44239017e-01 -4.71624106e-01 1.40349433e-01
-3.62096280e-01 4.42017853e-01 -4.50484127e-01 -8.80836427e-01
-2.51517296e-02 -8.46802473e-01 3.34399104e-01 -1.04383373e+00
-4.61176485e-01 -4.81519490e-01 1.03835955e-01 -1.19510448e+00
1.76856720e+00 -1.62765944e+00 -3.01766000e-03 1.96121976e-01
-1.73236966e-01 2.77318954e-01 -2.26636037e-01 1.04796481e+00
2.28293061e-01 5.92096925e-01 -3.94325182e-02 2.94148147e-01
-1.49248376e-01 -3.06086570e-01 -1.12721816e-01 9.53118876e-02
4.27834578e-02 5.96898973e-01 -8.58714640e-01 -8.83058131e-01
-2.99874216e-01 -8.62234980e-02 -1.17326468e-01 1.55515656e-01
3.15148711e-01 -2.41288505e-02 -5.49808919e-01 6.03880823e-01
5.34092307e-01 -7.73819834e-02 3.24202627e-01 2.07630143e-01
-3.94884765e-01 4.95946020e-01 -8.34874868e-01 1.32318497e+00
-5.34213960e-01 1.28416419e+00 -4.96541619e-01 -3.11108381e-01
1.22726810e+00 6.34106278e-01 5.29488325e-02 -8.25370550e-01
1.42230196e-02 5.82050383e-01 1.77221507e-01 -9.50395167e-01
1.03711700e+00 1.34081736e-01 -3.30395222e-01 4.56773072e-01
-1.99862570e-01 -4.08498138e-01 6.53136313e-01 3.42072397e-01
9.68126595e-01 -2.67784923e-01 6.89877272e-01 -7.93549716e-01
9.14576232e-01 4.26815599e-01 3.09439719e-01 7.01560080e-01
-5.36045060e-02 6.90277815e-01 3.90746415e-01 -1.11421481e-01
-1.27901363e+00 -8.50674987e-01 -2.82953918e-01 4.86495405e-01
-1.66421294e-01 -6.06705606e-01 -1.01021898e+00 -3.96346182e-01
-2.98198581e-01 7.35290468e-01 -1.31782860e-01 3.56960535e-01
-7.66935945e-01 -7.04168081e-01 5.47920942e-01 1.97631687e-01
5.93408167e-01 -1.00501478e+00 2.51640882e-02 4.07696426e-01
-9.58535910e-01 -1.18336868e+00 -6.79081798e-01 -5.21158814e-01
-8.08930039e-01 -8.62582982e-01 -8.10752153e-01 -1.31455660e+00
5.34531116e-01 3.28543603e-01 1.44781840e+00 -3.99777889e-02
8.24716762e-02 -1.25978529e-01 -4.60861206e-01 -1.22415602e-01
-7.29618430e-01 4.14095789e-01 1.71314552e-01 -6.76800132e-01
5.03418565e-01 -2.82617182e-01 3.28909270e-02 2.84705549e-01
-5.70984900e-01 2.34361455e-01 8.16587627e-01 1.06876254e+00
-4.13573459e-02 -6.96123615e-02 7.70096958e-01 -7.24916458e-01
1.26240385e+00 -3.73264968e-01 -3.88143152e-01 4.81285542e-01
-6.54668748e-01 1.50516659e-01 7.46286035e-01 -4.05324280e-01
-1.07550132e+00 -6.38535202e-01 -2.74454337e-02 7.54435301e-01
1.70905605e-01 5.03389120e-01 1.81518719e-01 3.51236671e-01
6.41464055e-01 2.59324551e-01 -2.48235494e-01 -3.94397825e-01
-7.35853091e-02 1.46327579e+00 4.90282774e-01 -7.09144831e-01
6.84743702e-01 -2.76840985e-01 -4.16689366e-01 -1.01670814e+00
-3.91582370e-01 -4.75997716e-01 -6.44923389e-01 -4.50484967e-03
6.94102108e-01 -1.01421273e+00 -5.20819090e-02 4.28041637e-01
-1.48108411e+00 1.53276101e-01 4.20251250e-01 8.75434101e-01
-3.57089430e-01 5.70683837e-01 -6.78889275e-01 -4.15314943e-01
-8.41823459e-01 -6.62449896e-01 9.35945928e-01 1.56388953e-01
-8.66171837e-01 -8.97595108e-01 4.23007220e-01 5.46568632e-01
2.21638232e-01 4.92915288e-02 9.68958676e-01 -6.24625027e-01
2.13824809e-02 -1.64916784e-01 -2.04924583e-01 1.46010205e-01
3.99269909e-01 5.54969728e-01 -6.20769978e-01 -1.35806063e-02
-1.67245060e-01 -6.55978024e-02 4.47413474e-01 1.17955431e-01
2.56778806e-01 -4.89215732e-01 -2.89578974e-01 -1.79112434e-01
1.33360386e+00 2.93605417e-01 6.96276248e-01 5.95580101e-01
8.14495683e-02 4.72767800e-01 7.25577533e-01 2.17699468e-01
2.21705124e-01 6.88227952e-01 -6.01525724e-01 1.62836209e-01
-1.96553513e-01 -7.42011666e-02 6.23643756e-01 1.82342815e+00
-2.34969586e-01 -7.00577438e-01 -1.15738070e+00 5.37430704e-01
-1.68321121e+00 -1.33158624e+00 -6.95549726e-01 1.82334769e+00
1.28879941e+00 4.22867894e-01 1.47824928e-01 -9.85311270e-02
1.04980040e+00 -1.21491075e-01 4.10355538e-01 -9.90091562e-01
-4.61688727e-01 1.76712796e-01 3.24436277e-02 7.67053545e-01
-8.53472888e-01 1.03959608e+00 7.21557617e+00 9.46765006e-01
-7.80238032e-01 -1.50756106e-01 3.64885598e-01 3.14571112e-01
-3.71287972e-01 -2.51384884e-01 -8.01317930e-01 6.85745478e-01
1.21775556e+00 -8.54448736e-01 2.63399091e-02 3.59688640e-01
4.25029367e-01 -3.78168106e-01 -8.67292583e-01 8.73130977e-01
3.89163882e-01 -1.27603638e+00 2.49742925e-01 -2.64695168e-01
1.21011782e+00 -1.15048997e-02 -3.71957242e-01 2.40184531e-01
-4.70158122e-02 -7.40150094e-01 6.82364464e-01 6.93964601e-01
7.56106734e-01 -6.49228275e-01 1.13415015e+00 5.99235952e-01
-9.51004148e-01 5.03627598e-01 -5.86642027e-01 -4.06494558e-01
-6.61650160e-03 8.25493991e-01 -1.01705062e+00 8.79512131e-01
3.72505039e-01 4.74327534e-01 -6.13230646e-01 7.04887390e-01
-1.28816813e-01 4.73349065e-01 -2.09758114e-02 -7.85302460e-01
3.98141861e-01 -4.80338305e-01 5.43316960e-01 1.84346843e+00
5.80938041e-01 -1.33331761e-01 -9.57035795e-02 3.41877103e-01
-6.28090650e-02 7.63469756e-01 -6.15909278e-01 -1.47196591e-01
5.14907002e-01 9.79141355e-01 -6.43634737e-01 -7.00203061e-01
-5.25090098e-01 1.02728248e+00 3.22442114e-01 1.62624791e-01
-7.76962638e-01 -9.31415677e-01 -1.80146135e-02 -2.37790897e-01
9.22214389e-02 -2.22768545e-01 -3.56915593e-01 -1.33469272e+00
3.41655195e-01 -1.08937037e+00 -1.00670338e-01 -9.08330023e-01
-1.40860379e+00 1.09601736e+00 -4.43938076e-02 -1.44663167e+00
-3.91829491e-01 -5.56963861e-01 -5.23120224e-01 1.38397360e+00
-9.00523186e-01 -7.32482493e-01 1.20246410e-01 2.44503677e-01
9.08432722e-01 -7.35428452e-01 9.63087499e-01 1.82142049e-01
-5.45694888e-01 4.91769999e-01 6.22211158e-01 1.51095256e-01
1.03730989e+00 -1.14798939e+00 4.75693047e-01 7.61361122e-01
1.05682284e-01 8.86853933e-01 7.19729304e-01 -6.85281515e-01
-1.08115137e+00 -6.25463367e-01 2.39637661e+00 -4.36208755e-01
8.90236139e-01 -4.95589413e-02 -4.51060116e-01 1.64420471e-01
9.97730494e-01 -1.11165226e+00 8.12680066e-01 1.02095798e-01
8.91596731e-03 5.48717342e-02 -9.85195339e-01 8.22576344e-01
9.20647025e-01 -5.27604461e-01 -1.29144287e+00 6.71097159e-01
4.67500806e-01 -1.36647716e-01 -1.02576983e+00 1.16742946e-01
7.19487607e-01 -8.28584254e-01 3.29719037e-01 -3.76659423e-01
8.32276881e-01 -1.85628563e-01 -3.86091322e-02 -1.18505251e+00
-7.54383743e-01 -8.17813158e-01 2.93647200e-01 1.51433539e+00
9.32915866e-01 -6.79823279e-01 5.11408560e-02 3.34916323e-01
-1.25674292e-01 -5.96493900e-01 -5.56614041e-01 -9.31279242e-01
3.55699837e-01 2.02250913e-01 3.87364209e-01 1.02928948e+00
4.22936141e-01 8.58815789e-01 -2.77901709e-01 -3.17564338e-01
-1.11432701e-01 4.07848358e-01 7.17862904e-01 -9.48173165e-01
4.87962700e-02 -6.18734777e-01 -1.38226032e-01 -8.05116236e-01
4.43839937e-01 -7.25216985e-01 -2.08168417e-01 -1.68025756e+00
7.67216563e-01 6.90976977e-02 2.62083799e-01 4.47334759e-02
-4.06757683e-01 1.67645738e-01 -3.81987244e-02 5.54711342e-01
-1.05005614e-01 1.37069166e-01 1.41734564e+00 -1.89880669e-01
-7.56948814e-02 1.47263393e-01 -6.64474010e-01 3.84420633e-01
1.18208408e+00 -5.57303429e-01 -1.88060388e-01 -4.77725565e-01
1.18083045e-01 6.19213544e-02 -4.77504551e-01 -8.55339408e-01
4.01749834e-02 -2.97510624e-01 2.89928079e-01 -7.52608716e-01
-3.84422243e-01 -3.16475660e-01 1.84446275e-01 5.35468638e-01
-2.76361883e-01 8.20126295e-01 1.30554959e-01 -2.94609577e-04
-5.43302715e-01 -5.74198663e-01 3.53950083e-01 -1.79901525e-01
-4.55412596e-01 -4.67649847e-01 -9.02107000e-01 3.17242622e-01
7.91505933e-01 -2.83323318e-01 -5.26508272e-01 -3.02356154e-01
-1.62190482e-01 -9.75532681e-02 4.78355765e-01 5.05368292e-01
3.22328240e-01 -1.28627062e+00 -1.19326866e+00 3.57785914e-03
-4.73145321e-02 -1.04060435e+00 -4.38163519e-01 8.32031369e-01
-1.02713215e+00 8.80069673e-01 -3.90918612e-01 -3.22579831e-01
-1.47095072e+00 2.19281331e-01 -2.26100132e-01 -4.52647895e-01
-3.54553878e-01 3.25068712e-01 -4.65116084e-01 -1.26673386e-01
-1.45171463e-01 -3.60794216e-01 -3.02563995e-01 2.15094760e-02
3.04046988e-01 6.04512572e-01 1.14356630e-01 -7.33198524e-01
-2.74198622e-01 8.79446268e-01 -1.46241754e-01 -5.55674672e-01
7.49254346e-01 -3.12692523e-01 -5.36224365e-01 6.06117725e-01
1.39162242e+00 7.83580899e-01 5.03037870e-02 -5.52854091e-02
3.26284766e-01 -4.50920671e-01 -4.52607453e-01 -8.23543370e-01
-2.70457327e-01 3.51203859e-01 -4.63391691e-02 4.18135852e-01
9.42619503e-01 -2.39169315e-01 7.89949358e-01 7.52354801e-01
4.86330450e-01 -1.60861576e+00 -1.61117539e-01 9.59425986e-01
1.09037793e+00 -1.27127159e+00 3.36175472e-01 -2.80952543e-01
-6.41783357e-01 1.57250428e+00 2.33482480e-01 3.03883012e-02
3.83240044e-01 6.59804419e-02 1.20743722e-01 2.20727503e-01
-8.51767242e-01 2.94134319e-01 5.00565588e-01 2.91242242e-01
1.09702420e+00 3.07135522e-01 -1.50677288e+00 4.97101337e-01
-6.13073230e-01 -1.57786012e-01 7.68054962e-01 1.01700830e+00
-6.73906147e-01 -1.34712386e+00 -4.28909808e-01 3.77760082e-01
-6.08755529e-01 -4.56690699e-01 -7.62617886e-01 1.24533463e+00
-1.92561984e-01 1.36679506e+00 9.11412090e-02 -4.25622702e-01
1.58594996e-01 1.03779420e-01 5.66881537e-01 -6.59636736e-01
-8.96981776e-01 3.01649749e-01 9.39483464e-01 -1.00955039e-01
-8.59612644e-01 -7.30936229e-01 -9.73501980e-01 -7.56323516e-01
-4.18605834e-01 8.27631652e-01 3.93623918e-01 7.82867312e-01
1.79714635e-01 9.29649100e-02 1.02408159e+00 -1.69463351e-01
-4.58883315e-01 -1.46609867e+00 -4.74532783e-01 3.40916395e-01
-3.05188224e-02 -6.66627362e-02 -4.29216206e-01 4.14741039e-01]
|
[11.508429527282715, 10.196490287780762]
|
44633150-9d7d-4e5c-a8b4-d8723a874254
|
transformers-on-sarcasm-detection-with
| null | null |
https://aclanthology.org/2020.figlang-1.13
|
https://aclanthology.org/2020.figlang-1.13.pdf
|
Transformers on Sarcasm Detection with Context
|
Sarcasm Detection with Context, a shared task of Second Workshop on Figurative Language Processing (co-located with ACL 2020), is study of effect of context on Sarcasm detection in conversations of Social media. We present different techniques and models, mostly based on transformer for Sarcasm Detection with Context. We extended latest pre-trained transformers like BERT, RoBERTa, spanBERT on different task objectives like single sentence classification, sentence pair classification, etc. to understand role of conversation context for sarcasm detection on Twitter conversations and conversation threads from Reddit. We also present our own architecture consisting of LSTM and Transformers to achieve the objective.
|
[]
|
2020-07-01
| null | null | null |
acl-2020-7
|
['sentence-pair-classification']
|
['natural-language-processing']
|
[-2.20494613e-01 2.68594861e-01 3.19292247e-01 -5.13154566e-01
-6.89878345e-01 -3.22479904e-01 8.63950491e-01 3.84588420e-01
-3.27085167e-01 4.85893130e-01 1.12207508e+00 -2.11380899e-01
4.41143960e-01 -4.11426604e-01 8.61011520e-02 -1.65649071e-01
2.22278833e-01 5.06507576e-01 -3.49002667e-02 -8.99377167e-01
6.77374661e-01 -1.26447231e-01 -8.24850798e-01 1.44230556e+00
-1.19094804e-01 5.22607625e-01 -6.70612603e-02 1.25390255e+00
-5.51300406e-01 1.94797933e+00 -1.08814275e+00 -5.23772717e-01
-7.70565152e-01 -5.61766624e-01 -1.51091242e+00 -1.65033489e-01
3.74793530e-01 1.16703585e-01 4.54385951e-02 5.82638621e-01
9.01576459e-01 2.12681308e-01 3.61831248e-01 -9.67537582e-01
-1.73184603e-01 1.60206580e+00 -2.93095499e-01 7.16238558e-01
7.34266758e-01 -2.24153906e-01 9.48732674e-01 -9.60095942e-01
4.15349424e-01 1.65311468e+00 1.20294321e+00 7.91678011e-01
-7.75832474e-01 -2.86548078e-01 -2.59034604e-01 2.82873452e-01
-4.89717990e-01 -4.77305353e-01 9.74333584e-01 -4.02870476e-01
1.56341577e+00 6.62469268e-01 7.15118289e-01 1.68943167e+00
1.08113989e-01 1.21428788e+00 9.48759139e-01 -4.39720213e-01
-3.03560078e-01 3.13261241e-01 7.75395036e-01 4.66411084e-01
-8.78916323e-01 -5.57219088e-01 -1.04530525e+00 -3.28466207e-01
-8.96899253e-02 -5.33438265e-01 1.91958800e-01 4.11788225e-01
-1.24698830e+00 1.08346832e+00 2.07756221e-01 8.82488370e-01
6.43601932e-04 7.51601160e-02 1.66857505e+00 7.63317108e-01
8.77877414e-01 4.69848365e-01 -2.69954026e-01 -5.40132046e-01
-7.36786127e-01 2.61153489e-01 9.94496882e-01 5.02077281e-01
-7.36804232e-02 8.33961740e-02 -7.03050673e-01 1.46159923e+00
9.07349735e-02 1.93153486e-01 8.52487743e-01 -5.62925518e-01
5.95176518e-01 5.14481306e-01 7.81386420e-02 -9.93984580e-01
-1.00613678e+00 -3.36429298e-01 -6.54316545e-01 -4.89900649e-01
2.51450062e-01 -3.81309301e-01 -3.44211049e-02 1.42834580e+00
-1.43061608e-01 -4.64295819e-02 3.75465453e-01 7.51459181e-01
1.86045194e+00 6.64778888e-01 3.10718060e-01 -2.79680312e-01
1.58947039e+00 -1.30504191e+00 -7.22639978e-01 -3.21169019e-01
1.42328763e+00 -1.30565178e+00 1.38526845e+00 5.28518796e-01
-1.39511049e+00 -6.20517373e-01 -8.20265234e-01 -4.89845395e-01
-2.06355393e-01 3.39845717e-01 3.18108290e-01 3.66727471e-01
-9.42378759e-01 5.30849576e-01 -9.21920389e-02 -8.29189897e-01
-2.84089297e-01 2.13420540e-02 7.22889602e-03 7.71477103e-01
-1.39381742e+00 1.33193839e+00 -1.47765592e-01 1.56572670e-01
-6.73854411e-01 -3.59499812e-01 -7.02135742e-01 -3.05345386e-01
-3.28300625e-01 -5.88224590e-01 1.91265023e+00 -1.27873302e+00
-1.58656287e+00 1.64430749e+00 -1.09664597e-01 -9.83797371e-01
3.85292619e-01 -5.04336298e-01 -3.92357856e-01 -1.09078132e-01
-9.95765775e-02 3.22765678e-01 5.08073986e-01 -6.47095680e-01
-2.38372117e-01 -7.70878121e-02 1.43169761e-01 4.44075942e-01
-3.60547543e-01 9.70608115e-01 7.45101273e-01 -3.63920540e-01
-3.40150595e-01 -5.27564049e-01 1.02438562e-01 -1.03795171e+00
-6.56784892e-01 -9.66764390e-01 1.18920279e+00 -7.30389416e-01
1.12269497e+00 -1.80922008e+00 1.92603990e-02 -6.85152113e-01
4.27659571e-01 3.99030328e-01 -4.69787344e-02 1.03933930e+00
-7.37268627e-02 -4.03839513e-05 3.78441960e-01 -9.99898970e-01
-2.16097116e-01 2.32798867e-02 -5.78899860e-01 1.32893324e-01
-1.43968418e-01 9.22181189e-01 -1.00729120e+00 -7.01382637e-01
3.17348331e-01 4.04755920e-01 -3.66333239e-02 3.64520520e-01
-8.92683491e-02 2.95021743e-01 -1.84635282e-01 5.96925169e-02
2.55546749e-01 -8.38099718e-02 1.44648686e-01 -2.69423760e-02
-3.01644862e-01 1.10015547e+00 -1.45572513e-01 1.49658966e+00
-8.50597441e-01 1.09533024e+00 7.82593936e-02 -1.15158677e+00
1.42286932e+00 8.69946897e-01 -3.44804004e-02 -4.19779748e-01
6.20911956e-01 2.29354333e-02 -1.81336747e-03 -8.13048542e-01
9.03204739e-01 -6.05043411e-01 -4.83716339e-01 6.12971425e-01
-1.22258738e-01 -3.61542702e-01 1.69230521e-01 6.81744635e-01
1.05863953e+00 -3.61484289e-01 2.61819124e-01 -4.93220598e-01
1.05343258e+00 7.35820979e-02 -9.00778361e-03 7.67660141e-01
-6.58447623e-01 4.55330372e-01 7.93177843e-01 -7.72723913e-01
-9.35017049e-01 -7.35015869e-01 3.44135284e-01 1.67924464e+00
-4.96604562e-01 -7.03387618e-01 -6.25712752e-01 -7.78544664e-01
-7.38745272e-01 9.95007396e-01 -7.82463431e-01 2.69681290e-02
-8.90297532e-01 -5.58269978e-01 7.48511672e-01 2.82930851e-01
5.06970406e-01 -1.72340262e+00 -5.98574460e-01 2.77638763e-01
-6.11607611e-01 -1.17082489e+00 -4.08952951e-01 1.99591637e-01
-7.75126696e-01 -9.18555081e-01 -2.48599514e-01 -1.25959265e+00
-2.52558172e-01 1.58745065e-01 1.70209527e+00 1.55074015e-01
3.73316333e-02 3.15630168e-01 -5.98754644e-01 -4.89513516e-01
-1.07874656e+00 2.00509671e-02 -3.92280608e-01 -4.33200032e-01
7.74191320e-01 -5.14379442e-01 -1.87106565e-01 1.44927874e-01
-6.33481666e-02 3.60795736e-01 -1.58900902e-01 1.02317274e+00
-5.75535536e-01 -8.55452657e-01 8.06577265e-01 -1.03410387e+00
1.45529294e+00 -3.67158264e-01 7.04703331e-01 -8.64730477e-02
2.84221023e-01 -4.58085597e-01 6.78961813e-01 -3.70889753e-01
-9.72025990e-01 -4.30931330e-01 -4.46268201e-01 1.17020622e-01
1.63014978e-02 4.80247676e-01 6.72892392e-01 4.90332127e-01
1.04249489e+00 5.02522998e-02 3.54817994e-02 -4.30023879e-01
1.08043350e-01 1.07746315e+00 4.33256328e-01 -1.99036241e-01
-2.82131195e-01 -1.03982782e-03 -6.13130152e-01 -1.06274343e+00
-1.34521139e+00 -8.42273116e-01 -3.15991104e-01 -6.49852574e-01
7.47933030e-01 -8.03152621e-01 -9.84327257e-01 6.13261223e-01
-1.56734598e+00 -3.05249304e-01 -2.38719583e-01 1.27167821e-01
-6.08843923e-01 5.44201612e-01 -1.48688579e+00 -1.12126577e+00
-1.35720873e+00 -5.25114715e-01 8.37163627e-01 -1.19904853e-01
-1.05937719e+00 -1.26251686e+00 5.51494002e-01 9.72603321e-01
4.01300222e-01 -6.54223338e-02 7.79273033e-01 -9.99603033e-01
8.32433045e-01 -1.41066924e-01 -5.35820685e-02 5.43635607e-01
-3.58622283e-01 -5.08396775e-02 -1.13128734e+00 -5.01932725e-02
3.60880017e-01 -1.20023096e+00 6.97503865e-01 3.47208977e-01
5.13819158e-01 -4.04780567e-01 1.50189549e-02 -4.82196659e-01
3.92170638e-01 -3.74702424e-01 6.48947537e-01 1.96368515e-01
3.70043874e-01 9.18503702e-01 5.28046072e-01 5.05337596e-01
4.37575370e-01 4.75222796e-01 1.88622370e-01 1.07931651e-01
-3.47038388e-01 -2.64842153e-01 9.22874868e-01 1.49804986e+00
3.21728319e-01 -2.53571808e-01 -7.61187494e-01 5.54609537e-01
-1.96387434e+00 -1.40878665e+00 -9.81221974e-01 1.60290992e+00
7.62756586e-01 3.65436912e-01 7.29076982e-01 4.61248040e-01
7.19645798e-01 4.70682144e-01 2.13628322e-01 -1.61551476e+00
-1.88865349e-01 -1.49462879e-01 -3.13427716e-01 9.42648470e-01
-1.05710137e+00 1.18872547e+00 7.19332743e+00 6.86800301e-01
-1.23970282e+00 5.42149723e-01 6.36339009e-01 -1.63022414e-01
-2.19795071e-02 -2.54854679e-01 -6.56562209e-01 4.50278491e-01
1.22026277e+00 -5.92696071e-02 -3.95380288e-01 8.92967403e-01
9.02526140e-01 -9.02496427e-02 -8.28315020e-01 7.15468109e-01
4.01959211e-01 -1.25444078e+00 -3.20959091e-01 -8.54559362e-01
2.62859017e-01 3.62909824e-01 -2.72784829e-01 8.34146261e-01
3.49410594e-01 -9.82461035e-01 5.13898611e-01 1.43884569e-01
-7.50106573e-02 -4.68468040e-01 1.05174839e+00 4.25295353e-01
-7.70962358e-01 4.33391668e-02 -3.49266171e-01 -8.36918414e-01
3.83184016e-01 5.36281407e-01 -1.39402997e+00 -1.90386951e-01
5.59356451e-01 1.18374407e+00 -5.20533264e-01 4.46536779e-01
-2.09231064e-01 9.77238536e-01 -2.35140640e-02 -9.48753357e-01
4.08604473e-01 2.93651998e-01 8.47478986e-01 1.85208416e+00
-2.58618087e-01 -3.19263637e-01 2.84715384e-01 3.89481753e-01
3.15757394e-01 5.66962242e-01 -4.76789355e-01 1.83335662e-01
1.22429296e-01 1.57380497e+00 -4.97799426e-01 -6.00695968e-01
-7.51056075e-02 8.45125556e-01 3.99376959e-01 -2.45287195e-01
-5.71329951e-01 8.84289145e-02 4.51286584e-02 7.43557140e-02
-4.41256195e-01 3.71337198e-02 -5.61630845e-01 -8.15932631e-01
-4.51787382e-01 -8.28657866e-01 7.14478195e-01 -1.13070929e+00
-1.55654442e+00 7.95136213e-01 -2.04008371e-01 -8.97439182e-01
-2.77720064e-01 -4.18281913e-01 -1.60375118e+00 6.16367042e-01
-6.76700234e-01 -1.64811862e+00 -1.06390491e-01 5.64412117e-01
1.19338036e+00 -2.02014863e-01 9.14518416e-01 -2.11336240e-02
-5.18916368e-01 4.74588633e-01 -6.18030429e-01 9.53390375e-02
8.95825386e-01 -1.30505204e+00 4.94934440e-01 2.76771933e-01
-1.98279917e-02 2.62079269e-01 1.41255295e+00 -3.70283306e-01
-6.19382083e-01 -5.15618324e-01 1.84810221e+00 -7.33192444e-01
1.22535193e+00 -4.38885987e-01 -4.20694292e-01 6.13084793e-01
9.25338089e-01 -7.83401310e-01 7.94385254e-01 6.93520188e-01
-3.34693044e-01 4.01486248e-01 -9.75897908e-01 4.46537942e-01
8.47894967e-01 -5.50558984e-01 -1.09533155e+00 9.36515331e-01
6.18464112e-01 -9.68866646e-02 -4.60035801e-01 1.72623083e-01
2.14676306e-01 -1.42956388e+00 7.48090208e-01 -9.68797743e-01
1.08363688e+00 4.73752975e-01 1.80007592e-01 -1.22071397e+00
1.37794390e-01 -7.94150651e-01 2.28180185e-01 1.20391655e+00
4.48376566e-01 -2.02996716e-01 9.64409709e-01 7.22581670e-02
-5.56787431e-01 -4.65351760e-01 -8.24560225e-01 -1.86622903e-01
4.30379808e-01 -4.83585030e-01 -3.24608207e-01 1.01728606e+00
1.00891972e+00 1.68089485e+00 -9.04290378e-01 -8.63258004e-01
-1.12550497e-01 6.99026212e-02 9.51881051e-01 -8.04240823e-01
-4.04251039e-01 -5.75867772e-01 -1.99328333e-01 -1.06874514e+00
5.83050027e-02 -8.38450134e-01 -6.96584210e-02 -1.50443411e+00
4.83368337e-01 5.97244687e-02 6.42743707e-02 4.12906647e-01
1.79409608e-01 2.94054836e-01 3.13379139e-01 7.07001686e-02
-8.87902439e-01 3.27558339e-01 1.04548872e+00 -1.52008131e-01
-1.89459026e-01 1.94799289e-01 -4.38003004e-01 9.44202065e-01
1.18095684e+00 -3.02149981e-01 -3.28706115e-01 -1.33329570e-01
6.02734745e-01 2.91129053e-01 3.38497221e-01 -7.88485527e-01
-6.04226701e-02 2.35812113e-01 -3.14278454e-01 -1.02562785e+00
6.53670788e-01 2.05077171e-01 -6.10556781e-01 6.49480641e-01
-1.07076132e+00 3.70081306e-01 2.05042157e-02 -9.81031284e-02
-2.62480825e-01 -6.23823345e-01 6.86914742e-01 -4.14406657e-01
-1.82263911e-01 -7.28367209e-01 -1.22058761e+00 4.63939637e-01
4.35541272e-01 1.14803962e-01 -7.07491517e-01 -9.63430345e-01
-1.09677804e+00 1.48628473e-01 -2.35422969e-01 6.54290855e-01
4.83179897e-01 -9.21883345e-01 -1.21512926e+00 -4.92484629e-01
8.13206565e-03 -7.56900668e-01 4.10668850e-01 1.32761955e+00
-3.22613388e-01 2.82610923e-01 4.30445001e-02 -5.00592291e-01
-2.14034343e+00 3.32147032e-01 6.02219760e-01 -7.78549552e-01
-6.47341192e-01 1.41157663e+00 -1.21801794e-01 -6.80572748e-01
1.23015113e-01 -1.32174551e-01 -8.84408116e-01 2.95774639e-01
8.02601576e-01 2.77247339e-01 6.32493421e-02 -7.30103791e-01
-1.74776539e-01 7.99338669e-02 -2.19302893e-01 1.25788199e-02
8.91942024e-01 -1.79982156e-01 -3.38984966e-01 1.11219740e+00
1.17992771e+00 -5.77697642e-02 -1.41276464e-01 -1.23664439e-02
1.56791478e-01 1.95180655e-01 -5.97645231e-02 -1.16037452e+00
-5.60706973e-01 1.19394314e+00 1.29193634e-01 8.58452678e-01
4.97707427e-01 8.13093632e-02 1.11060309e+00 4.05879617e-01
-1.95062473e-01 -1.34982944e+00 6.96321785e-01 1.08859825e+00
1.40380430e+00 -1.22373056e+00 -2.91593462e-01 -1.94811627e-01
-1.00575650e+00 1.43337059e+00 6.29937172e-01 -3.55512649e-01
5.33880293e-01 2.24545240e-01 5.42573035e-01 -6.41435087e-01
-1.09803176e+00 2.58833677e-01 -6.58297241e-02 2.88155168e-01
1.24357033e+00 7.75696263e-02 -9.64445651e-01 6.40275538e-01
-7.75597453e-01 -2.29915410e-01 9.64909971e-01 5.19187391e-01
-6.60068631e-01 -9.52542424e-01 -1.66546941e-01 2.54993916e-01
-6.50848567e-01 -4.04202908e-01 -1.23329568e+00 4.08519328e-01
-3.14585418e-01 1.54984975e+00 2.09301673e-02 -7.93805301e-01
3.04754019e-01 1.22182935e-01 1.12784676e-01 -8.28764021e-01
-1.78734922e+00 -1.52346537e-01 1.35861778e+00 -1.12017691e-01
-7.64450550e-01 -6.62945271e-01 -8.99153173e-01 -6.61948919e-01
-1.56410098e-01 5.09663761e-01 3.83232355e-01 1.07770252e+00
-5.69564514e-02 2.04834044e-01 5.42672992e-01 -7.00326025e-01
-4.84913677e-01 -1.84966576e+00 -2.78365582e-01 5.77210009e-01
-1.85131561e-02 -9.10371095e-02 -6.32284164e-01 -2.21041441e-01]
|
[9.092945098876953, 10.72872257232666]
|
60772674-787a-4c1c-9f22-c691e6558dba
|
leveraging-alignment-and-phonology-for-low
| null | null |
https://aclanthology.org/2020.icon-main.51
|
https://aclanthology.org/2020.icon-main.51.pdf
|
Leveraging Alignment and Phonology for low-resource Indic to English Neural Machine Transliteration
|
In this paper we present a novel transliteration technique based on Orthographic Syllable(OS) segmentation for low-resource Indian languages (ILs). Given that alignment has produced promising results in Statistical Machine Transliteration systems and phonology plays an important role in transliteration, we introduce a new model which uses alignment representation similar to that of IBM model 3 to pre-process the tokenized input sequence and then use pre-trained source and target OS-embeddings for training. We apply our model for transliteration from ILs to English and report our accuracy based on Top-1 Exact Match. We also compare our accuracy with a previously proposed Phrase-Based model and report improvements.
|
['Arjun Atreya', 'Pushpak Bhattacharya', 'Manthan Mehta', 'Parth Patel']
| null | null | null | null |
icon-2020-12
|
['transliteration']
|
['natural-language-processing']
|
[ 3.57875288e-01 -2.16686606e-01 -4.29851651e-01 -3.84208828e-01
-7.53756166e-01 -5.71292639e-01 4.19441044e-01 7.23353252e-02
-9.08277690e-01 7.23766804e-01 4.78089362e-01 -1.15287673e+00
4.48778123e-01 -5.54162264e-01 -6.48695052e-01 -8.15768912e-02
4.53985363e-01 9.93296504e-01 7.39620905e-03 -3.09705466e-01
2.98881710e-01 1.97252870e-01 -6.75062656e-01 -1.33077294e-01
1.27867186e+00 2.65247412e-02 3.46525282e-01 1.03348827e+00
-2.67876625e-01 2.03168392e-02 -5.52760839e-01 -5.44475853e-01
2.49333903e-01 -8.24408829e-01 -1.34607077e+00 -1.93753578e-02
3.06152880e-01 -3.50261703e-02 1.30784586e-01 8.18508685e-01
6.25364482e-01 2.07423717e-01 9.55070078e-01 -4.48958218e-01
-1.32365286e+00 9.98568296e-01 -2.08811745e-01 6.06235266e-01
4.50028837e-01 -2.50463158e-01 1.09114730e+00 -1.10967493e+00
7.91588545e-01 1.10593987e+00 4.70241070e-01 7.19888687e-01
-1.04044032e+00 -4.25651163e-01 -1.16867132e-01 6.06373101e-02
-1.36496031e+00 -3.09686154e-01 2.08389655e-01 1.52580719e-02
1.58039665e+00 2.50805736e-01 4.17776316e-01 7.13805079e-01
4.44372535e-01 6.05488479e-01 1.19892669e+00 -1.28877962e+00
-1.84911370e-01 2.00879499e-01 1.87372729e-01 5.49563706e-01
4.37290668e-02 -9.70956013e-02 -2.44774580e-01 3.68581235e-01
6.82495475e-01 -4.94609773e-01 2.90781826e-01 7.21487343e-01
-1.43449211e+00 7.77080417e-01 1.27621433e-02 4.74720597e-01
-2.60963768e-01 1.08193099e-01 5.09162307e-01 3.29485774e-01
5.24637997e-01 4.60556686e-01 -6.15350902e-01 -4.34084684e-01
-1.07222772e+00 -1.08705096e-01 6.19854271e-01 1.07323587e+00
6.04904711e-01 4.41953272e-01 -1.32648885e-01 1.04699790e+00
2.21281216e-01 8.72278988e-01 1.19126630e+00 -2.89510787e-01
6.28197908e-01 1.89282522e-01 -2.02056170e-01 -1.21040888e-01
7.72846118e-02 -2.92015672e-01 -2.57967561e-01 -5.07103860e-01
1.50931448e-01 -5.25095575e-02 -1.57378197e+00 1.21522868e+00
1.89575851e-01 -2.54082531e-01 3.17632854e-01 6.37028515e-01
5.71139634e-01 1.00184691e+00 1.66499764e-01 -8.00463930e-02
1.30978525e+00 -1.23465276e+00 -8.37165177e-01 -3.15712154e-01
8.40258718e-01 -1.41129315e+00 1.58985198e+00 -2.42878608e-02
-1.13031375e+00 -7.89342940e-01 -9.54068422e-01 -3.40869308e-01
-4.85429317e-01 3.59426260e-01 3.98404956e-01 1.10014582e+00
-1.22341323e+00 5.01852572e-01 -1.01563323e+00 -8.35752428e-01
-3.17218721e-01 5.52601516e-01 -3.33870798e-01 1.85049653e-01
-1.18112767e+00 1.06930232e+00 6.23546183e-01 -1.10554725e-01
-3.11319798e-01 -1.50567904e-01 -8.59805524e-01 -3.24847490e-01
-2.02399820e-01 -3.62835348e-01 1.31352711e+00 -8.23616445e-01
-1.99695563e+00 8.94094229e-01 -8.32083106e-01 -1.81922153e-01
-4.25244635e-03 -2.31005743e-01 -6.18007779e-01 -1.91062793e-01
4.34114300e-02 5.73958457e-01 5.88542044e-01 -7.69529402e-01
-9.71010268e-01 1.36896506e-01 -5.55237949e-01 3.17566395e-01
-2.91489452e-01 9.77126181e-01 -4.91824657e-01 -8.35638940e-01
2.18801469e-01 -1.16398907e+00 -1.96864411e-01 -1.04556978e+00
-1.77538738e-01 -4.25818354e-01 3.26519012e-01 -1.34469080e+00
1.69729030e+00 -1.60491395e+00 1.80640370e-01 1.23787582e-01
-5.94452024e-01 5.01713991e-01 -3.12152475e-01 5.26961029e-01
-4.64195944e-02 6.97430074e-01 -2.92027414e-01 -7.93529510e-01
1.70810595e-01 7.17423379e-01 -2.95409173e-01 1.98545530e-01
6.01356149e-01 1.19899213e+00 -8.15711677e-01 -6.77480876e-01
9.55659151e-02 3.37071717e-01 -4.44149166e-01 1.87674925e-01
1.91082299e-01 4.51954484e-01 1.51705921e-01 9.32971299e-01
4.07831043e-01 6.09932482e-01 -4.94354255e-02 5.14984787e-01
-3.86941135e-01 1.08761489e+00 -6.48437738e-01 1.47044814e+00
-9.37513411e-01 3.27714473e-01 -7.87717879e-01 -6.14470243e-01
1.11207640e+00 3.55986953e-01 -2.00481653e-01 -5.92064798e-01
1.84702024e-01 8.16269040e-01 3.53246540e-01 -1.71384186e-01
1.24485672e+00 -1.97043613e-01 -3.10399622e-01 8.31601858e-01
3.42028469e-01 -6.07096910e-01 3.97621602e-01 -1.98313057e-01
6.77556276e-01 4.38673377e-01 6.13947272e-01 -3.07243347e-01
4.75128025e-01 3.01321417e-01 4.94755894e-01 4.70330924e-01
-1.84117973e-01 8.92950654e-01 -2.14323938e-01 -2.51267016e-01
-1.38245809e+00 -1.17176378e+00 -6.60046637e-02 1.39408255e+00
-4.01802927e-01 -3.08662474e-01 -8.17467690e-01 -8.20186317e-01
-5.76568902e-01 9.09911990e-01 -3.04796755e-01 1.09224364e-01
-1.31366205e+00 -5.20944118e-01 8.06747079e-01 5.87588072e-01
1.53966218e-01 -1.40276027e+00 1.59635425e-01 4.54370111e-01
-3.19194257e-01 -1.11175454e+00 -1.10465407e+00 2.60472983e-01
-9.03190613e-01 -2.09104612e-01 -8.14077973e-01 -1.39203501e+00
7.81517327e-01 -2.19069719e-01 9.25713241e-01 -1.14357784e-01
-5.86851947e-02 -1.12779789e-01 -9.14770484e-01 -5.41705549e-01
-8.67239237e-01 6.01863623e-01 4.53964442e-01 -5.50447106e-01
7.23176718e-01 -1.28772467e-01 3.11677661e-02 1.59245357e-02
-9.01439011e-01 -3.84874135e-01 4.77697760e-01 6.91949487e-01
6.18587911e-01 -3.91473114e-01 2.55502909e-01 -8.93119037e-01
8.07025254e-01 -1.60971701e-01 -2.71053463e-01 2.17800334e-01
-8.02193761e-01 2.49493212e-01 9.58135307e-01 -4.97613758e-01
-8.01669300e-01 -1.18895515e-03 -6.33638978e-01 1.00195572e-01
-5.28708398e-02 7.15401232e-01 6.92840293e-02 5.31931520e-02
4.40554321e-01 4.74752158e-01 -4.64852303e-01 -5.85005820e-01
3.78441840e-01 1.16281784e+00 5.85966170e-01 -5.71421385e-01
9.59983170e-01 -1.53137386e-01 -4.51772839e-01 -6.73737526e-01
-1.86587065e-01 -4.22690541e-01 -1.11607885e+00 2.76977658e-01
8.68363857e-01 -4.61711556e-01 -6.17485605e-02 2.39724353e-01
-1.46022928e+00 -2.46071741e-01 -3.61600101e-01 9.32273865e-01
-3.78617197e-01 5.11010110e-01 -8.01542640e-01 -6.32108867e-01
-7.28407860e-01 -1.13655078e+00 8.84617507e-01 1.54097274e-01
-7.80527413e-01 -1.39140236e+00 5.11180639e-01 1.48837805e-01
6.42540395e-01 -3.76549333e-01 9.53295231e-01 -8.32823932e-01
3.36813852e-02 -1.22100234e-01 5.13998792e-02 6.09890938e-01
5.32275498e-01 2.12292135e-01 -5.12322843e-01 -8.59147534e-02
-4.12181884e-01 7.26903528e-02 5.72736442e-01 1.76678166e-01
3.17874432e-01 -2.29184180e-01 2.08195716e-01 6.87048852e-01
1.19138861e+00 1.89586356e-01 5.39509237e-01 4.48018640e-01
7.76396513e-01 3.40013057e-01 7.55013168e-01 -3.05766929e-02
6.92081034e-01 4.00700867e-01 -3.08141947e-01 -2.09208041e-01
-3.94325197e-01 -4.63804632e-01 9.88288581e-01 1.93814087e+00
-2.02080324e-01 -3.85751724e-01 -1.15898716e+00 1.03977585e+00
-1.45514107e+00 -3.74858260e-01 -1.42396718e-01 2.27999187e+00
1.32420158e+00 1.98048223e-02 1.32939190e-01 6.48967996e-02
6.46754444e-01 -1.03451610e-01 2.39916146e-01 -1.93718112e+00
-1.17337145e-01 1.35881960e+00 8.99561644e-01 1.14471376e+00
-8.04284275e-01 2.03412390e+00 7.41318703e+00 7.18466818e-01
-1.27941358e+00 4.85525519e-01 2.64920086e-01 5.32120049e-01
-5.01399279e-01 2.72760510e-01 -1.28849339e+00 1.95419312e-01
1.51450276e+00 -3.39024246e-01 4.79218841e-01 5.25267899e-01
3.02855372e-01 -5.66931302e-03 -1.02124071e+00 5.39216042e-01
2.67035991e-01 -8.38083982e-01 2.77206153e-01 -1.58858925e-01
9.09947395e-01 1.00850716e-01 1.24653079e-01 4.33494359e-01
5.17670751e-01 -1.17656338e+00 6.82668746e-01 -6.97664246e-02
1.17980242e+00 -9.12373662e-01 9.17254448e-01 1.19867530e-02
-1.24532056e+00 6.74553871e-01 -6.55170679e-01 -2.35500664e-01
3.59728307e-01 -1.95471887e-02 -1.46198535e+00 5.83603144e-01
1.76798761e-01 4.81037796e-01 -5.15324950e-01 7.27710843e-01
-4.94985580e-01 1.30837750e+00 -3.56231272e-01 -3.14702868e-01
3.88260335e-01 -4.36641663e-01 3.14737648e-01 1.77417386e+00
7.01402843e-01 -3.83551180e-01 1.26291364e-01 4.69161510e-01
9.45308711e-03 7.64731228e-01 -1.85133085e-01 -5.62352836e-01
3.18031251e-01 8.46072853e-01 -8.31816912e-01 -4.67789680e-01
-2.33178571e-01 1.72473752e+00 2.23271340e-01 3.02637249e-01
-6.45359337e-01 -7.98446715e-01 6.29846275e-01 -9.52714384e-02
4.38986540e-01 -7.66744256e-01 -2.80596048e-01 -8.68800044e-01
-2.52977088e-02 -7.96664000e-01 6.28325194e-02 -2.00154111e-01
-8.47806454e-01 1.04439139e+00 -2.09334597e-01 -1.03314078e+00
-4.27537650e-01 -7.23791897e-01 -6.57090724e-01 1.59763503e+00
-1.75638258e+00 -1.39262736e+00 6.03513181e-01 1.20376535e-01
8.85725379e-01 -6.56865686e-02 1.12005830e+00 3.56499195e-01
-6.12271965e-01 1.08618486e+00 8.38281140e-02 2.45464653e-01
9.81038570e-01 -1.48569477e+00 1.23556435e+00 1.33950746e+00
5.86430490e-01 9.69220936e-01 6.01866543e-01 -8.37954998e-01
-1.11627138e+00 -1.11307085e+00 2.02657318e+00 -6.52332842e-01
7.20355630e-01 -2.76617348e-01 -6.25613391e-01 1.00884819e+00
5.42848885e-01 -2.07570508e-01 8.26482713e-01 1.13797687e-01
1.59202442e-02 2.81657517e-01 -7.95916021e-01 8.17872822e-01
9.87220526e-01 -6.10706806e-01 -9.25626934e-01 2.45116666e-01
1.09743094e+00 -5.47193706e-01 -6.60094976e-01 6.26768824e-03
3.05953592e-01 -4.86564375e-02 4.31894243e-01 -7.26474106e-01
3.32699001e-01 -2.23807976e-01 -9.30870026e-02 -1.72866952e+00
-4.85346973e-01 -9.48631525e-01 4.63118911e-01 1.31212091e+00
7.79854715e-01 -7.15789199e-01 4.43626523e-01 9.47811007e-02
-5.61762333e-01 -2.45028496e-01 -9.13079739e-01 -1.09897137e+00
4.99126583e-01 -4.65850174e-01 6.41503036e-01 6.70640469e-01
2.03359798e-01 3.50687087e-01 -3.78312528e-01 1.25772312e-01
-2.08007805e-02 -2.31376782e-01 4.63709474e-01 -4.21729267e-01
-3.52084696e-01 -1.61305070e-01 -3.94688129e-01 -9.83810842e-01
3.09031576e-01 -1.30621469e+00 2.93860197e-01 -1.49970746e+00
-4.32315432e-02 -5.02907574e-01 -3.38781148e-01 4.96726722e-01
-5.98121762e-01 7.38164544e-01 1.13385476e-01 -2.33214330e-02
-1.04062743e-01 3.82722139e-01 1.00753808e+00 2.58166790e-01
-4.94246185e-01 1.83603112e-02 -3.79528522e-01 1.91609263e-01
1.03896093e+00 -6.82064056e-01 9.50327218e-02 -8.35120261e-01
8.24280158e-02 -4.95080888e-01 -6.97349787e-01 -7.41400242e-01
-1.31285906e-01 -2.90780365e-01 2.82345861e-02 -4.10630256e-01
-1.81616351e-01 -4.29991573e-01 -3.91072243e-01 5.03694475e-01
-1.35685131e-01 8.59058201e-01 3.19837481e-01 -2.68917233e-01
-2.84725398e-01 -5.81251442e-01 4.55648959e-01 -2.45662276e-02
-6.43653512e-01 2.52024949e-01 -7.94434488e-01 -8.62040818e-02
5.92999458e-01 -3.68838161e-01 1.36705637e-01 -6.24485500e-02
-5.32727480e-01 -1.24365643e-01 5.59882879e-01 6.79508567e-01
5.69526374e-01 -1.22670281e+00 -1.01939797e+00 3.70069385e-01
1.56002063e-02 -4.75564152e-01 -7.40607679e-01 5.98418593e-01
-1.21139646e+00 6.82544172e-01 -1.70155168e-01 -2.17169836e-01
-1.48242033e+00 1.54188916e-01 -1.46434665e-01 -3.53276670e-01
-8.11472088e-02 1.08128381e+00 -6.17125511e-01 -1.07965314e+00
-9.87392813e-02 -7.56722450e-01 -2.90859048e-03 -4.80138838e-01
3.73425782e-01 2.11675182e-01 2.62611300e-01 -1.10504878e+00
-5.04527628e-01 6.47414923e-01 -3.49115819e-01 -6.73917711e-01
8.11467290e-01 -3.32326680e-01 -3.69041175e-01 6.33440495e-01
1.00738204e+00 5.92745304e-01 -2.53592730e-01 -3.69864196e-01
1.88299507e-01 -3.03988218e-01 -1.86982587e-01 -5.36244571e-01
-3.11644971e-01 1.07730234e+00 3.86985242e-01 -4.96402085e-01
8.11798513e-01 -2.16628283e-01 1.25960195e+00 2.86649644e-01
5.89883588e-02 -1.58011949e+00 -2.76405454e-01 1.38636804e+00
3.44580889e-01 -1.35741925e+00 -2.82302260e-01 -2.41110086e-01
-7.13586211e-01 1.19474363e+00 2.99819469e-01 -3.01960796e-01
3.63182187e-01 9.77443531e-02 4.55033451e-01 5.58478236e-01
-2.59003103e-01 -4.08052236e-01 3.67471099e-01 6.66165411e-01
1.09857178e+00 4.84336376e-01 -1.23882353e+00 3.21623921e-01
-7.98035443e-01 -2.63983697e-01 4.87380534e-01 1.00636292e+00
-4.56529558e-01 -2.04310608e+00 -4.91903305e-01 1.43100783e-01
-8.24181974e-01 -8.91678512e-01 -4.91004288e-01 5.50690651e-01
1.03747703e-01 8.15322459e-01 2.23711312e-01 -4.68823791e-01
4.71625850e-03 3.56263965e-01 6.57612562e-01 -1.16430295e+00
-8.96034360e-01 9.00494829e-02 1.54208377e-01 -1.46878794e-01
5.99795952e-02 -6.56554878e-01 -1.51527417e+00 -4.65750515e-01
-3.18369746e-01 2.66832888e-01 9.76689160e-01 1.13301504e+00
-3.50139700e-02 3.97184700e-01 7.00197577e-01 -6.22546911e-01
-4.04099226e-01 -1.38527095e+00 -3.65697086e-01 3.50838788e-02
-2.34633521e-03 1.91699654e-01 -2.32348680e-01 2.90147245e-01]
|
[11.397391319274902, 10.359014511108398]
|
da66f291-232b-4c6d-92fc-2b90a0f702a4
|
practical-license-plate-recognition-in
|
1910.04324
| null |
https://arxiv.org/abs/1910.04324v1
|
https://arxiv.org/pdf/1910.04324v1.pdf
|
Practical License Plate Recognition in Unconstrained Surveillance Systems with Adversarial Super-Resolution
|
Although most current license plate (LP) recognition applications have been significantly advanced, they are still limited to ideal environments where training data are carefully annotated with constrained scenes. In this paper, we propose a novel license plate recognition method to handle unconstrained real world traffic scenes. To overcome these difficulties, we use adversarial super-resolution (SR), and one-stage character segmentation and recognition. Combined with a deep convolutional network based on VGG-net, our method provides simple but reasonable training procedure. Moreover, we introduce GIST-LP, a challenging LP dataset where image samples are effectively collected from unconstrained surveillance scenes. Experimental results on AOLP and GIST-LP dataset illustrate that our method, without any scene-specific adaptation, outperforms current LP recognition approaches in accuracy and provides visual enhancement in our SR results that are easier to understand than original data.
|
['Yoojin Hong', 'Younkwan Lee', 'Moongu Jeon', 'Jiwon Jun']
|
2019-10-10
| null | null | null | null |
['license-plate-recognition']
|
['computer-vision']
|
[ 5.72716296e-01 -4.94121462e-01 3.74610685e-02 -2.07131490e-01
-9.31845844e-01 -7.12365866e-01 5.34238219e-01 -9.90757227e-01
-2.35801712e-01 7.84651458e-01 -3.46628070e-01 -2.63450414e-01
4.93415207e-01 -6.60367906e-01 -1.00463736e+00 -6.11138463e-01
5.30882716e-01 2.42362976e-01 9.04354692e-01 -2.69461758e-02
3.61493498e-01 6.22518480e-01 -1.00903690e+00 3.11314702e-01
1.06359494e+00 9.29351747e-01 3.76435034e-02 7.14907825e-01
-1.83068588e-02 9.18951631e-01 -6.64571166e-01 -8.28523159e-01
6.93862915e-01 -8.46839417e-03 -3.60218078e-01 7.14187860e-01
8.85431528e-01 -6.09827578e-01 -8.52970839e-01 1.17835379e+00
1.95371985e-01 1.09614059e-01 5.04568160e-01 -1.11079144e+00
-9.19840097e-01 7.85320066e-03 -6.85262799e-01 3.30776632e-01
1.40299171e-01 4.44997758e-01 3.51729453e-01 -9.05466020e-01
5.93047142e-01 8.57129633e-01 1.14207506e+00 6.86799884e-01
-9.19111371e-01 -6.00114644e-01 1.55517422e-02 1.74657956e-01
-1.43560708e+00 -5.16451836e-01 8.71227086e-01 -2.44635567e-01
7.15749979e-01 2.10999250e-01 2.37854362e-01 1.31247211e+00
7.80378580e-02 1.14191282e+00 1.18478382e+00 -2.25757107e-01
1.25000002e-02 1.50917068e-01 6.45885170e-02 5.96786737e-01
1.92882478e-01 -2.02895671e-01 -5.45723923e-03 3.31611186e-01
1.17035317e+00 1.20367147e-01 -4.51494247e-01 -1.53174937e-01
-9.27897871e-01 3.72609615e-01 9.59994942e-02 -8.70612860e-02
-3.15393955e-02 -4.35811989e-02 2.48736396e-01 -1.28244132e-01
3.35820764e-01 1.64904788e-01 -2.69069001e-02 1.34347761e-02
-1.03605485e+00 8.79576579e-02 6.20394945e-01 1.40566027e+00
4.28166598e-01 5.23149192e-01 -2.21288458e-01 1.15887642e+00
1.47418514e-01 8.21811736e-01 1.54977351e-01 -9.00695741e-01
8.43618870e-01 3.77494007e-01 1.25871345e-01 -1.25472212e+00
2.25318372e-02 -3.68562967e-01 -6.96872771e-01 2.76315898e-01
5.69116831e-01 5.35289617e-03 -1.21739650e+00 9.97268260e-01
-4.04687136e-01 5.34570336e-01 3.38518977e-01 9.31830108e-01
6.67301714e-01 8.90107810e-01 2.59329379e-02 6.80790395e-02
1.23920500e+00 -1.46615398e+00 -4.71462876e-01 -6.88624382e-01
3.02410871e-02 -5.80598950e-01 9.87552226e-01 4.53514725e-01
-8.55672598e-01 -5.80803514e-01 -1.09624112e+00 6.35904968e-02
-3.70784879e-01 4.54915822e-01 5.56342304e-01 8.85819256e-01
-8.08689535e-01 1.47916511e-01 -5.29452682e-01 -3.81168365e-01
9.92935121e-01 3.31245661e-01 -5.50630450e-01 -4.62423354e-01
-8.73757660e-01 7.22019792e-01 2.00514287e-01 2.77271628e-01
-9.21695948e-01 -3.64335120e-01 -7.60983586e-01 -8.21445584e-02
6.86973035e-01 -1.28763869e-01 9.58963871e-01 -1.04407549e+00
-1.63020205e+00 9.13059175e-01 1.99629202e-01 -4.93796915e-01
6.99204564e-01 -5.26057184e-02 -8.41697335e-01 4.69435632e-01
-1.91284850e-01 5.21085024e-01 1.01242566e+00 -1.45325005e+00
-6.48905814e-01 -1.46495089e-01 -7.43167698e-02 7.90207461e-02
-3.30181457e-02 3.84349078e-01 -1.16016281e+00 -6.97889745e-01
-1.95249379e-01 -8.12671065e-01 -2.07940429e-01 1.60119161e-01
-6.01996660e-01 1.88759863e-01 1.14252973e+00 -1.04070032e+00
6.46008611e-01 -2.23713970e+00 -5.39351165e-01 3.29192057e-02
1.23539239e-01 9.52547371e-01 -1.65588304e-01 -8.57221559e-02
3.69182020e-01 1.16527833e-01 -4.81307447e-01 -4.26198483e-01
8.38821381e-02 2.13930532e-01 -5.26686847e-01 5.33891261e-01
6.34896398e-01 1.03981853e+00 -3.98921609e-01 -4.85876203e-01
3.75196725e-01 3.57350260e-01 -2.07238242e-01 3.63133699e-02
-1.50592938e-01 3.09894174e-01 -4.20413673e-01 1.04986489e+00
1.33224535e+00 9.07819346e-02 -3.25420469e-01 3.98564301e-02
-5.58749288e-02 -4.05632854e-01 -8.27879965e-01 1.18651068e+00
-8.33198652e-02 1.18621528e+00 5.12770116e-02 -9.38323557e-01
1.19830370e+00 -1.76090911e-01 2.95352209e-02 -7.69754827e-01
9.63958167e-03 2.30354041e-01 -6.29289925e-01 -6.53841436e-01
6.16479278e-01 7.03404322e-02 -1.66063905e-01 -3.11829627e-01
-1.78229794e-01 -4.08930151e-04 1.32948726e-01 -5.43290079e-02
9.12586749e-01 2.03195587e-01 -1.62839621e-01 2.03359291e-01
8.63036513e-01 1.99470922e-01 9.34457839e-01 8.45541656e-01
-4.66432959e-01 1.04253602e+00 4.59467679e-01 -4.16478932e-01
-1.35028768e+00 -1.24537921e+00 -1.71497986e-01 5.27426243e-01
5.33500850e-01 4.38265324e-01 -8.59133005e-01 -8.05514872e-01
-2.33230919e-01 4.72383887e-01 -1.60410404e-01 2.32210651e-01
-9.16675329e-01 -5.97178340e-01 1.26078641e+00 7.62251914e-01
1.42603958e+00 -9.12705362e-01 3.99491414e-02 7.15329638e-03
1.29044220e-01 -1.93963122e+00 -7.87399232e-01 -6.08397126e-01
-4.68330204e-01 -1.01921809e+00 -9.63671505e-01 -1.08898532e+00
7.68883288e-01 3.74198645e-01 8.14739466e-01 -2.43175596e-01
-2.09463924e-01 2.70321250e-01 -2.22030327e-01 -2.36806363e-01
-3.63368750e-01 -3.83500069e-01 -1.06661431e-01 4.58653122e-01
5.42764962e-01 -2.06397206e-01 -3.46110374e-01 5.77226162e-01
-9.28400815e-01 1.56363603e-02 9.09901381e-01 8.37612212e-01
4.16000098e-01 1.43373758e-01 4.00311321e-01 -7.08563447e-01
4.43360031e-01 -1.45033568e-01 -9.66547310e-01 4.47030097e-01
-1.78268865e-01 -6.22624397e-01 8.61409724e-01 -4.52508539e-01
-1.36740005e+00 1.91598311e-01 -2.23627374e-01 -8.20948243e-01
-5.74874341e-01 4.12623398e-02 -4.59044695e-01 -6.34621918e-01
2.20423147e-01 9.81282413e-01 2.57737972e-02 -3.14187437e-01
-7.10307155e-03 9.60529745e-01 1.24034309e+00 -3.32135677e-01
1.00251508e+00 4.35089827e-01 -3.49003673e-01 -1.09392476e+00
-3.41742218e-01 -3.57083887e-01 -6.37582958e-01 -3.48306537e-01
8.90194356e-01 -8.85254323e-01 -6.41581178e-01 1.14124489e+00
-9.34096515e-01 -2.12691158e-01 5.37405573e-02 3.68619651e-01
-4.71149057e-01 8.94533277e-01 -7.36416042e-01 -8.15740108e-01
-8.16292241e-02 -1.11075377e+00 1.03786516e+00 4.93916392e-01
7.03051448e-01 -9.24745679e-01 -2.77833730e-01 9.93051469e-01
5.61067283e-01 3.18669945e-01 1.94502980e-01 -7.46423423e-01
-1.20738208e+00 -5.76941013e-01 -7.24011898e-01 7.78304279e-01
-2.55228072e-01 2.49727488e-01 -9.25011277e-01 -1.04024686e-01
-2.81662554e-01 -3.71887237e-01 1.06441545e+00 1.54678136e-01
1.16516936e+00 -4.29609865e-01 -2.38871425e-01 1.01183867e+00
1.60208416e+00 2.52461106e-01 1.24064314e+00 5.10439157e-01
8.55108202e-01 1.95394740e-01 4.55885202e-01 6.44035041e-02
3.05402607e-01 6.83546662e-01 1.14369877e-01 -3.06408495e-01
-3.27229857e-01 -3.36498141e-01 7.72942305e-01 3.82110506e-01
-2.81369299e-01 -5.64595342e-01 -1.04323328e+00 2.41447091e-01
-1.63057232e+00 -1.21716666e+00 -2.32870698e-01 1.89943361e+00
4.17375565e-01 2.37998560e-01 7.87973255e-02 -1.71889350e-01
9.43581045e-01 1.87984064e-01 -6.11346364e-01 -1.14932790e-01
-6.30766094e-01 -2.70868868e-01 9.26176012e-01 2.35432312e-01
-1.54190493e+00 1.26818097e+00 6.68480539e+00 1.04959297e+00
-1.31057048e+00 -1.84789583e-01 6.95351422e-01 4.82469350e-01
3.42556387e-02 -4.74197656e-01 -8.54249477e-01 5.92701674e-01
5.32636583e-01 9.99077633e-02 2.63184696e-01 8.96372318e-01
-1.03868596e-01 1.10844016e-01 -6.98253393e-01 1.08775723e+00
4.93965030e-01 -1.57179224e+00 3.65636423e-02 2.48137433e-02
7.08024442e-01 9.61013585e-02 3.90676320e-01 5.86192071e-01
9.88271311e-02 -1.04692626e+00 4.70651537e-01 6.85520411e-01
9.98524308e-01 -7.81622231e-01 8.85912359e-01 3.41933191e-01
-9.19321179e-01 -8.92821997e-02 -7.14671314e-01 3.76069993e-01
-4.42848657e-04 2.89315768e-02 -7.29384243e-01 4.29828018e-01
4.27128285e-01 9.64246869e-01 -7.71826267e-01 1.27605212e+00
-9.43247005e-02 7.69352019e-01 -1.26312122e-01 1.27689719e-01
4.15431857e-01 -3.22589040e-01 7.40825176e-01 1.56249678e+00
2.27127612e-01 -4.42345487e-03 2.46082842e-01 1.14867902e+00
-1.39778882e-01 -9.38211679e-02 -7.44730830e-01 -3.99703626e-03
2.24231407e-01 1.07203031e+00 -6.86965764e-01 -1.55195832e-01
-8.56020212e-01 1.37703550e+00 8.87983071e-04 5.95405519e-01
-1.17454445e+00 -3.46261114e-01 5.66964269e-01 1.20348781e-02
6.19528472e-01 -3.48420352e-01 -1.44564942e-01 -1.50137031e+00
2.68025339e-01 -7.86210775e-01 5.06115966e-02 -8.29736233e-01
-1.42132115e+00 5.73094666e-01 -4.22236592e-01 -1.71765423e+00
1.84679553e-01 -1.14766622e+00 -7.75639534e-01 5.04254341e-01
-1.75513172e+00 -1.49909616e+00 -2.33390942e-01 7.69024134e-01
1.06556487e+00 -6.73424304e-01 3.93107027e-01 3.28380942e-01
-1.12522268e+00 9.53546286e-01 5.80148637e-01 8.34127367e-01
5.34463227e-01 -8.17002237e-01 5.67221582e-01 1.45420253e+00
-2.80518293e-01 1.83269411e-01 2.80063629e-01 -5.89601934e-01
-1.38858676e+00 -1.47466028e+00 1.56063035e-01 -4.88679409e-01
5.27203202e-01 -4.98903573e-01 -9.60107327e-01 6.67320013e-01
7.30848014e-02 4.63630021e-01 5.03643572e-01 -7.85813093e-01
-3.83459926e-01 -2.90305793e-01 -1.47581100e+00 5.81293702e-01
7.39927948e-01 -3.58602732e-01 -5.80985069e-01 2.24351779e-01
4.86907870e-01 -4.05135810e-01 -7.13476956e-01 2.29946688e-01
3.92221570e-01 -7.08111227e-01 1.22461367e+00 -3.74426901e-01
4.37535226e-01 -6.45433724e-01 -2.28887737e-01 -8.28199327e-01
-2.16584057e-01 -3.49186063e-01 1.88151464e-01 1.33080244e+00
3.74054134e-01 -6.44904017e-01 9.45444405e-01 9.91718054e-01
-4.82186079e-01 -3.35475743e-01 -8.46742153e-01 -1.17562532e+00
5.82803832e-03 -4.75961298e-01 2.88703740e-01 8.24976265e-01
-4.46701407e-01 -2.45439127e-01 -9.13808405e-01 6.42065823e-01
1.12472069e+00 -1.73757419e-01 8.49978983e-01 -8.19163144e-01
-1.89867482e-01 -3.94663423e-01 -9.00700867e-01 -1.12825930e+00
2.32737675e-01 -5.37413716e-01 3.19793880e-01 -1.34982741e+00
2.56074935e-01 -1.77646115e-01 -1.21645585e-01 3.54084402e-01
-1.86529115e-01 7.92563319e-01 3.11223984e-01 4.22091812e-01
-1.02315509e+00 2.87828922e-01 1.19392872e+00 -5.77639163e-01
1.41657785e-01 1.34652421e-01 -4.34978485e-01 7.86189079e-01
8.10387909e-01 2.55801920e-02 -1.47924647e-01 -5.43403804e-01
-4.20759588e-01 -4.14746590e-02 6.34690166e-01 -1.27500033e+00
5.83175659e-01 -1.15035854e-01 7.65871823e-01 -5.08815348e-01
3.80976200e-01 -8.90847147e-01 -5.69265373e-02 1.14109196e-01
-3.91093008e-02 -5.59870362e-01 4.81458455e-01 7.95997202e-01
-3.29180300e-01 -1.44672677e-01 9.24944699e-01 -4.84941825e-02
-1.53009319e+00 5.35373926e-01 -6.85138941e-01 1.92464504e-03
1.35890806e+00 -8.24989140e-01 -7.35468507e-01 -1.87438115e-01
-3.94105434e-01 2.25846007e-01 6.96291208e-01 4.60593998e-01
1.00107145e+00 -1.26471901e+00 -8.97942901e-01 4.82565403e-01
1.76744640e-01 6.77960506e-03 4.38541114e-01 5.63005030e-01
-1.19383121e+00 6.04460657e-01 -5.66453815e-01 -5.69552422e-01
-1.11554384e+00 6.33656859e-01 3.64207208e-01 -1.04121655e-01
-8.58200252e-01 7.18816876e-01 4.86554615e-02 -4.73846823e-01
1.64812714e-01 -9.19209942e-02 -2.67886043e-01 -5.63618481e-01
6.79808259e-01 2.81041205e-01 -2.48041004e-01 -8.24272871e-01
-3.18613172e-01 7.63809443e-01 -2.47414231e-01 3.53701115e-01
1.33665502e+00 -5.67714497e-02 2.09249482e-01 -1.66901618e-01
8.97597730e-01 1.21206172e-01 -1.93062270e+00 -2.50852883e-01
-1.02374777e-01 -9.42831278e-01 -1.60860643e-01 -6.17172301e-01
-1.30690372e+00 6.42250478e-01 2.88380116e-01 -1.53223932e-01
1.23005569e+00 -3.03249806e-01 9.34507489e-01 6.74115837e-01
5.08303463e-01 -1.16055286e+00 -1.00561693e-01 4.04785246e-01
8.12293112e-01 -1.35687995e+00 -2.69375801e-01 -7.28429973e-01
-9.59344208e-01 1.34430742e+00 8.99513960e-01 -3.10771018e-01
-1.05995452e-02 4.51053083e-01 1.82755396e-01 3.97584081e-01
-3.34720314e-01 2.74282563e-02 1.23522751e-01 1.04104424e+00
-2.63087124e-01 -2.18968943e-01 2.70721436e-01 7.08998680e-01
3.36860210e-01 8.15064609e-02 8.43267024e-01 6.19980216e-01
-4.71330345e-01 -8.17715466e-01 -6.28153265e-01 2.61741191e-01
-5.91025770e-01 -3.67179736e-02 -3.25347364e-01 1.04660082e+00
-9.35992692e-03 6.77442729e-01 2.04554707e-01 -3.73887718e-01
3.88639927e-01 -8.75721425e-02 1.98125437e-01 -1.76516593e-01
-8.73429980e-03 -9.22216624e-02 1.63028628e-01 -3.67749602e-01
-2.47883379e-01 -7.19096541e-01 -9.04085100e-01 -4.35855746e-01
-2.37080902e-01 -3.61146808e-01 4.43456024e-01 8.46159458e-01
2.34293446e-01 3.74307007e-01 7.28681803e-01 -8.54529262e-01
-3.16841632e-01 -6.97268784e-01 -7.87670135e-01 3.15388650e-01
2.71446317e-01 -2.55187631e-01 -1.63256600e-01 3.45900297e-01]
|
[9.869264602661133, -4.878361701965332]
|
991f8802-b311-48f5-9f5c-d104f494c77b
|
weakly-supervised-convolutional-lstm-approach
|
1812.01366
| null |
http://arxiv.org/abs/1812.01366v2
|
http://arxiv.org/pdf/1812.01366v2.pdf
|
Weakly Supervised Convolutional LSTM Approach for Tool Tracking in Laparoscopic Videos
|
Purpose: Real-time surgical tool tracking is a core component of the future
intelligent operating room (OR), because it is highly instrumental to analyze
and understand the surgical activities. Current methods for surgical tool
tracking in videos need to be trained on data in which the spatial positions of
the tools are manually annotated. Generating such training data is difficult
and time-consuming. Instead, we propose to use solely binary presence
annotations to train a tool tracker for laparoscopic videos. Methods: The
proposed approach is composed of a CNN + Convolutional LSTM (ConvLSTM) neural
network trained end-to-end, but weakly supervised on tool binary presence
labels only. We use the ConvLSTM to model the temporal dependencies in the
motion of the surgical tools and leverage its spatio-temporal ability to smooth
the class peak activations in the localization heat maps (Lh-maps).
Results: We build a baseline tracker on top of the CNN model and demonstrate
that our approach based on the ConvLSTM outperforms the baseline in tool
presence detection, spatial localization, and motion tracking by over 5.0%,
13.9%, and 12.6%, respectively.
Conclusions: In this paper, we demonstrate that binary presence labels are
sufficient for training a deep learning tracking model using our proposed
method. We also show that the ConvLSTM can leverage the spatio-temporal
coherence of consecutive image frames across a surgical video to improve tool
presence detection, spatial localization, and motion tracking.
keywords: Surgical workflow analysis, tool tracking, weak supervision,
spatio-temporal coherence, ConvLSTM, endoscopic videos
|
['Jacques Marescaux', 'Didier Mutter', 'Nicolas Padoy', 'Chinedu Innocent Nwoye']
|
2018-12-04
| null | null | null | null |
['instrument-recognition', 'video-object-tracking', 'surgical-tool-detection']
|
['audio', 'computer-vision', 'computer-vision']
|
[ 2.24335104e-01 3.96833494e-02 -5.73708653e-01 1.01554126e-01
-6.66594267e-01 -7.86972404e-01 2.94517815e-01 -1.04592830e-01
-6.01758361e-01 -3.93992253e-02 3.87946493e-03 -3.99650455e-01
-4.96009625e-02 7.90310502e-02 -9.06515956e-01 -6.08267069e-01
-2.14294165e-01 -1.70732036e-01 3.32574964e-01 4.74309511e-02
8.66103992e-02 6.33604944e-01 -9.95285809e-01 4.54844952e-01
3.66364658e-01 9.32557106e-01 4.10434037e-01 8.95026207e-01
1.65283382e-01 1.11031318e+00 -3.45360130e-01 3.68008554e-01
3.30846965e-01 -3.05747151e-01 -5.28198481e-01 -1.89399049e-01
6.21971786e-01 -4.84703779e-01 -6.77402556e-01 1.06858695e+00
4.09080684e-01 5.93943298e-02 2.54161775e-01 -1.05555069e+00
-9.44557637e-02 4.22023594e-01 -5.98847806e-01 2.49268696e-01
2.11175367e-01 4.50035632e-01 1.80376515e-01 -7.63007343e-01
1.09390283e+00 7.26082206e-01 1.16574836e+00 7.30814338e-01
-6.91777766e-01 -8.15347433e-01 4.93724123e-02 -2.53395438e-01
-8.75397086e-01 -1.81147411e-01 6.80808246e-01 -8.60488296e-01
7.11698890e-01 1.67107895e-01 9.23113465e-01 1.36946797e+00
9.26235855e-01 1.02256751e+00 4.78320062e-01 -4.84875888e-01
-2.95738876e-01 -2.52240658e-01 -1.94995403e-01 1.30346549e+00
5.94254620e-02 4.91569221e-01 -5.74093223e-01 1.16063394e-01
1.34184420e+00 7.51406431e-01 -4.51496452e-01 -8.32269430e-01
-1.82090473e+00 5.87579429e-01 8.79567385e-01 7.05250919e-01
-3.77121031e-01 7.76591599e-01 6.40675306e-01 -1.42411068e-01
6.22057728e-02 6.64641678e-01 -2.08401650e-01 -4.15972888e-01
-9.90311146e-01 -3.59668165e-01 4.87753570e-01 1.23331213e+00
7.22368062e-02 -1.62985802e-01 -5.39955497e-01 1.92205518e-01
1.58519834e-01 2.24664569e-01 7.01759994e-01 -1.10770380e+00
1.81291580e-01 5.59299231e-01 3.95969123e-01 -8.23528707e-01
-1.00735509e+00 -6.95341349e-01 -5.95541835e-01 3.31299216e-01
3.64202410e-01 -2.78586328e-01 -1.36361790e+00 1.43442035e+00
4.56389040e-02 3.56567621e-01 -4.15116608e-01 8.93983841e-01
1.00688350e+00 -1.91568565e-02 2.18566582e-01 5.63070131e-03
1.10584521e+00 -1.44281530e+00 -1.04175234e+00 -8.88087004e-02
1.52697885e+00 -9.05677021e-01 9.34076071e-01 1.89797744e-01
-1.00340116e+00 -6.35268509e-01 -1.02643919e+00 -1.08428940e-01
-1.83680862e-01 1.04323280e+00 8.17058504e-01 2.82381415e-01
-9.41224873e-01 6.66168094e-01 -1.74324441e+00 -1.83757558e-01
4.33745027e-01 7.45908737e-01 -4.99844909e-01 -5.54182567e-02
-6.23152316e-01 1.01658404e+00 2.33572200e-01 6.18806005e-01
-9.95561719e-01 -7.80030370e-01 -1.19643092e+00 -3.28102827e-01
3.62387359e-01 -6.34235680e-01 1.43076146e+00 -8.67117584e-01
-1.35025311e+00 9.71555710e-01 9.64970738e-02 -3.67766261e-01
8.67761374e-01 -4.49215770e-01 -1.85845554e-01 1.66199982e-01
-4.92103323e-02 6.91351295e-01 4.61799324e-01 -1.00774682e+00
-7.91882396e-01 7.02883825e-02 3.25587392e-02 -1.76679015e-01
-1.48685247e-01 3.68839689e-02 -6.10692978e-01 -7.33329058e-01
6.32555857e-02 -1.53782594e+00 -3.61545682e-01 8.10333908e-01
-4.96561944e-01 9.58674625e-02 8.54484260e-01 -7.93447137e-01
1.14709902e+00 -2.39091325e+00 -1.32286161e-01 -1.28743529e-01
3.34403157e-01 2.99959958e-01 -5.43547142e-03 -9.13976058e-02
1.59082279e-01 -4.11825240e-01 1.53017655e-01 -4.12577331e-01
-3.91924709e-01 3.98541056e-02 1.11769505e-01 1.03587174e+00
-2.56109923e-01 1.14582086e+00 -1.35819101e+00 -4.93424922e-01
8.17332208e-01 6.68363690e-01 -5.06474137e-01 1.76315889e-01
-1.89381726e-02 9.44923699e-01 -2.05456596e-02 7.65920818e-01
2.58554131e-01 -3.04421633e-01 2.04810590e-01 -5.51697850e-01
-2.58699894e-01 -1.75804317e-01 -6.67573452e-01 2.76598191e+00
-9.32346225e-01 9.71169770e-01 1.48858249e-01 -3.25535268e-01
4.27364886e-01 4.16061103e-01 1.00740266e+00 -6.17517889e-01
5.01790047e-01 5.22193849e-01 3.86902213e-01 -7.19865561e-01
2.14931011e-01 1.04133762e-01 -1.40695915e-01 -1.89747259e-01
3.77074480e-01 3.86856854e-01 1.42363794e-02 -6.47052154e-02
1.37694514e+00 5.21409214e-01 -7.71021703e-03 -9.80407149e-02
3.26338783e-02 3.17368507e-01 2.87157655e-01 9.57744360e-01
-4.93751228e-01 5.00992119e-01 2.68079072e-01 -6.48819149e-01
-6.31537616e-01 -9.18112576e-01 4.19823006e-02 9.42174613e-01
6.82919919e-01 -5.89376759e-05 -4.41013485e-01 -1.10258079e+00
-1.11849688e-01 2.05423862e-01 -1.20904624e+00 -5.06441355e-01
-1.20946479e+00 4.94431481e-02 4.26735610e-01 9.98400331e-01
-1.05602384e-01 -1.15024328e+00 -1.31486380e+00 3.04954886e-01
1.13890588e-01 -1.35987723e+00 -6.68688118e-01 5.90234697e-01
-8.45080793e-01 -1.25612223e+00 -9.62158084e-01 -1.19625998e+00
7.82451749e-01 3.16009194e-01 7.05532193e-01 4.01503481e-02
-6.88440144e-01 4.22598243e-01 -1.59759298e-01 -2.94752568e-01
-3.22488964e-01 5.21447919e-02 -1.43764943e-01 -5.25366306e-01
2.92303599e-02 1.18594086e-02 -1.02448940e+00 1.35953188e-01
-6.41025901e-01 2.48472139e-01 8.94613326e-01 1.05821955e+00
2.36620426e-01 -7.53433764e-01 -2.38412037e-01 -8.19441080e-01
1.00288517e-03 -1.81257904e-01 -6.19178653e-01 1.66920543e-01
3.42072062e-02 -3.19454633e-02 4.64003265e-01 -8.57374370e-01
-5.99508643e-01 7.51593888e-01 1.15967110e-01 -1.20926607e+00
-7.75350025e-03 1.81546390e-01 5.26247919e-01 -6.29931748e-01
5.85696757e-01 -1.24706440e-01 8.42657909e-02 -1.36761919e-01
2.10592076e-01 1.69699803e-01 9.80134606e-01 -3.72410715e-02
4.34473366e-01 7.26397097e-01 1.32121682e-01 -2.45804369e-01
-1.09491134e+00 -1.09100556e+00 -9.12556887e-01 -4.89458412e-01
1.15718794e+00 -8.81359816e-01 -9.44711208e-01 -2.08141759e-01
-1.41161203e+00 -6.41188443e-01 -8.01145360e-02 1.04145420e+00
-5.95049500e-01 -2.06228066e-02 -8.84199321e-01 -3.56479049e-01
-4.70971227e-01 -1.45300913e+00 1.48935533e+00 9.48358178e-02
-2.11502656e-01 -1.30399072e+00 1.17742874e-01 -2.88670748e-01
3.73024434e-01 1.01604021e+00 1.71162948e-01 -5.44654548e-01
-5.71928501e-01 -6.97962224e-01 -5.13936281e-02 3.98339368e-02
4.64215666e-01 -1.11066423e-01 -8.02614093e-01 -5.12323141e-01
-2.40275666e-01 2.30725467e-01 7.58695841e-01 9.56000507e-01
1.29606819e+00 -7.10397810e-02 -8.88605654e-01 1.11388624e+00
1.16737020e+00 2.02892110e-01 2.41123363e-02 5.15114129e-01
1.12989926e+00 1.74824923e-01 9.55576658e-01 8.41358379e-02
-2.85064846e-01 7.70381749e-01 6.45103812e-01 -8.06411147e-01
-4.28606302e-01 -8.48257691e-02 3.51541579e-01 6.96937799e-01
-9.62311849e-02 2.79983759e-01 -9.64135945e-01 6.00334406e-01
-1.92719316e+00 -5.92880309e-01 -1.58009037e-01 2.02151871e+00
5.32092988e-01 4.26007286e-02 -1.35136947e-01 -2.77542293e-01
5.16702354e-01 -8.70073289e-02 -3.26212585e-01 4.86127473e-02
4.80003208e-01 8.26499313e-02 1.08554935e+00 3.43296140e-01
-1.61345768e+00 7.53927231e-01 5.24457264e+00 2.28084803e-01
-1.72420919e+00 2.42103338e-01 -1.30221397e-01 -4.94346380e-01
5.83592832e-01 -3.51521611e-01 -3.48544985e-01 5.57778835e-01
5.51021218e-01 2.52067506e-01 -2.83475548e-01 1.02805054e+00
2.20833749e-01 1.16029598e-01 -1.62898588e+00 1.25550961e+00
1.74154565e-01 -1.60189748e+00 -5.09826064e-01 1.28542529e-02
8.43310416e-01 8.15012828e-02 2.56487310e-01 2.99719453e-01
1.17576204e-01 -1.06349707e+00 6.34350538e-01 5.87495208e-01
1.00367856e+00 -3.50872606e-01 1.03241885e+00 2.85398185e-01
-1.19399786e+00 -1.67797416e-01 7.92786404e-02 5.45209348e-01
1.75431982e-01 -9.09484327e-02 -1.20479357e+00 3.07122737e-01
5.72201192e-01 1.00261283e+00 -3.73092741e-01 1.48076177e+00
-1.48109511e-01 -5.06570376e-02 -1.43922269e-01 3.06804091e-01
6.36170387e-01 5.57198822e-01 3.35741997e-01 1.63352561e+00
5.37911236e-01 -4.37958151e-01 5.87906718e-01 4.94716555e-01
-1.36289120e-01 -3.69589806e-01 -7.11616635e-01 2.45641753e-01
3.40224467e-02 1.29424822e+00 -8.50587487e-01 -1.72823727e-01
-3.64999592e-01 9.93582428e-01 -1.41830161e-01 1.73174366e-01
-1.03549099e+00 -4.82376188e-01 2.57891297e-01 7.19556957e-02
6.18536212e-02 -3.26026976e-01 -1.94604352e-01 -8.60311985e-01
2.24376097e-01 -4.63786691e-01 3.20407599e-01 -6.68637156e-01
-4.58064497e-01 5.31121194e-01 -5.64259768e-01 -2.07072425e+00
-5.20660281e-01 -1.06915212e+00 -4.91193026e-01 4.54932451e-01
-1.50293374e+00 -1.49927950e+00 -1.00047576e+00 5.29846847e-01
6.24119699e-01 3.57596397e-01 7.61231959e-01 3.15219730e-01
-2.78322786e-01 6.28770709e-01 5.25682382e-02 6.46199226e-01
9.07310009e-01 -1.22102010e+00 5.64273121e-03 7.29019165e-01
3.83525528e-02 9.36138272e-01 5.55827320e-01 -4.75154191e-01
-1.64958501e+00 -1.22840500e+00 6.94592670e-02 -8.81443262e-01
7.56232262e-01 -4.08876091e-01 -4.46002722e-01 1.10770452e+00
3.81633900e-02 7.71002114e-01 5.71209311e-01 -2.37160653e-01
5.35443760e-02 3.45907956e-01 -9.53555167e-01 4.86003816e-01
1.15634501e+00 -3.45981777e-01 -2.85804689e-01 7.61033535e-01
6.36003494e-01 -1.44796801e+00 -1.02469409e+00 6.75332487e-01
8.59752595e-01 -5.79896867e-01 9.82257724e-01 -4.51134026e-01
4.81692642e-01 -2.76113927e-01 5.54573715e-01 -9.66110051e-01
-5.57186157e-02 -6.78429425e-01 -3.40398014e-01 1.14163443e-01
9.48975384e-02 -3.86511348e-02 9.35496449e-01 1.72964409e-01
-7.15546310e-01 -8.39658916e-01 -9.82515574e-01 -8.51137936e-01
-2.60658205e-01 -2.61638790e-01 -3.26026917e-01 1.09426951e+00
7.12353438e-02 -3.58538181e-01 -3.53192389e-01 3.11063468e-01
4.00451392e-01 6.17303476e-02 6.05003715e-01 -9.78053391e-01
-1.74563825e-01 -5.33995152e-01 -5.29515922e-01 -1.02876914e+00
-5.23668434e-03 -9.06893849e-01 4.59263474e-01 -1.76914191e+00
2.47401949e-02 -3.33336890e-01 -6.46947086e-01 6.47290707e-01
8.43482018e-02 2.51833916e-01 2.46976003e-01 3.27661067e-01
-9.20206964e-01 -2.57829838e-02 1.51455891e+00 -2.01619759e-01
-4.28469658e-01 2.00545058e-01 2.95241743e-01 8.54825079e-01
2.45175108e-01 -6.74352288e-01 1.42573297e-01 -3.65353823e-01
-2.23203748e-01 3.06431264e-01 6.33478761e-01 -1.09923363e+00
7.63243794e-01 1.87660813e-01 6.61443472e-01 -5.78130960e-01
3.73237759e-01 -1.14907253e+00 -1.04748517e-01 1.23739898e+00
-3.94327402e-01 -1.15355507e-01 4.86238658e-01 3.85019243e-01
-1.96282923e-01 6.85331598e-02 7.25632787e-01 -2.12459803e-01
-6.90925121e-01 3.92331123e-01 -1.29319474e-01 -3.41765910e-01
1.10024381e+00 -3.32571775e-01 -1.35674149e-01 1.74916521e-01
-9.70325589e-01 7.22907856e-02 4.38475072e-01 7.97742426e-01
6.03654683e-01 -1.15434802e+00 -2.05899179e-01 3.67273018e-02
3.83180171e-01 1.54695779e-01 6.07085526e-01 1.79598844e+00
-8.26272130e-01 6.52704835e-01 -1.46635056e-01 -1.13230872e+00
-1.21253252e+00 8.02720666e-01 7.78211236e-01 -2.18101844e-01
-1.16993892e+00 9.50171947e-01 4.96912509e-01 -3.39216977e-01
7.86257327e-01 -1.06746936e+00 -4.86826599e-02 -4.53488648e-01
3.23272258e-01 -1.30986229e-01 2.19551831e-01 -2.63388366e-01
-5.65590620e-01 7.61788309e-01 -6.53676540e-02 4.11646485e-01
1.04981184e+00 4.47719812e-01 4.19544667e-01 4.09933537e-01
1.33626223e+00 -4.74041104e-02 -1.58177209e+00 4.55421917e-02
2.70499885e-01 -4.45393175e-01 2.44545132e-01 -8.32799315e-01
-1.18058145e+00 7.94406950e-01 1.15516484e+00 -2.22098917e-01
7.42395043e-01 -7.54271150e-02 8.09514165e-01 1.21290907e-01
3.24109703e-01 -6.86788023e-01 1.85963288e-01 3.12704146e-01
7.80741453e-01 -1.35792601e+00 -4.15269583e-01 -3.05691779e-01
-4.29081053e-01 1.37053120e+00 6.08971179e-01 -2.05247015e-01
4.51806366e-01 8.07974637e-01 4.73460615e-01 -3.28606546e-01
-2.72739261e-01 -9.90010565e-04 5.60590565e-01 3.70292634e-01
7.15948045e-01 -7.60608017e-02 9.51257274e-02 2.31393531e-01
2.30860412e-01 2.75715590e-01 4.07193780e-01 1.45737791e+00
-1.26743749e-01 -4.24681157e-01 5.41401282e-02 3.87041777e-01
-6.01233542e-01 3.93539369e-02 1.72942951e-01 1.12831855e+00
1.71203539e-01 5.36955416e-01 -7.13233128e-02 -2.58052826e-01
5.94017982e-01 -4.64323878e-01 7.02476442e-01 -6.67525470e-01
-1.07143044e+00 2.21813112e-01 -1.24531828e-01 -1.08523989e+00
-3.50608200e-01 -2.71625876e-01 -1.36093175e+00 3.50069940e-01
-4.57506776e-01 -1.40266448e-01 9.15250063e-01 7.07451999e-01
1.99193269e-01 1.46802485e+00 2.24043369e-01 -1.29804766e+00
-2.70752639e-01 -8.79242301e-01 -2.28195742e-01 2.94064939e-01
8.79188418e-01 -1.07345593e+00 -3.81419212e-01 1.10683322e-01]
|
[14.04632568359375, -3.319890022277832]
|
87b7dea2-ffe3-4302-97ab-ac3b9bc96ce4
|
improving-performance-of-federated-learning
|
2112.06194
| null |
https://arxiv.org/abs/2112.06194v2
|
https://arxiv.org/pdf/2112.06194v2.pdf
|
Improving Performance of Federated Learning based Medical Image Analysis in Non-IID Settings using Image Augmentation
|
Federated Learning (FL) is a suitable solution for making use of sensitive data belonging to patients, people, companies, or industries that are obligatory to work under rigid privacy constraints. FL mainly or partially supports data privacy and security issues and provides an alternative to model problems facilitating multiple edge devices or organizations to contribute a training of a global model using a number of local data without having them. Non-IID data of FL caused from its distributed nature presents a significant performance degradation and stabilization skews. This paper introduces a novel method dynamically balancing the data distributions of clients by augmenting images to address the non-IID data problem of FL. The introduced method remarkably stabilizes the model training and improves the model's test accuracy from 83.22% to 89.43% for multi-chest diseases detection of chest X-ray images in highly non-IID FL setting. The results of IID, non-IID and non-IID with proposed method federated trainings demonstrated that the proposed method might help to encourage organizations or researchers in developing better systems to get values from data with respect to data privacy not only for healthcare but also other fields.
|
['Seref Sagiroglu', 'Murat Akin', 'Alper Emin Cetinkaya']
|
2021-12-12
| null | null | null | null |
['image-augmentation']
|
['computer-vision']
|
[ 8.90789777e-02 1.21252723e-01 -3.96704406e-01 -6.47681653e-01
-6.28012776e-01 -5.71361899e-01 1.98649645e-01 1.37476236e-01
-3.34400088e-01 9.83084679e-01 1.28566816e-01 -5.26350260e-01
-4.65697318e-01 -5.51709592e-01 -4.07872826e-01 -9.87873375e-01
1.22233145e-01 4.54097927e-01 -7.38074556e-02 1.46718651e-01
-1.20601594e-01 7.93874681e-01 -1.33676732e+00 8.01705241e-01
3.96955967e-01 1.26420927e+00 -3.59663993e-01 4.76898700e-01
-1.95192605e-01 9.52420890e-01 -4.33953911e-01 -5.18123746e-01
9.08625841e-01 -6.53166464e-03 -6.12717628e-01 -1.16937287e-01
5.11551023e-01 -4.50401336e-01 -1.01928629e-01 1.00070727e+00
7.31611252e-01 -2.98808813e-02 3.24433208e-01 -1.63530171e+00
-3.71957660e-01 1.96017668e-01 -8.71831536e-01 1.42970234e-02
1.03059888e-01 6.57767290e-03 2.80750662e-01 -2.65230119e-01
8.09235275e-01 7.32367456e-01 9.26721573e-01 6.76929772e-01
-7.58974791e-01 -7.54670441e-01 -2.84085095e-01 -1.43619940e-01
-1.14806533e+00 -3.02798182e-01 7.40781248e-01 -1.30078495e-01
6.60964370e-01 1.03466201e+00 1.00904301e-01 8.37619722e-01
3.20039988e-01 4.39457059e-01 1.25063217e+00 -4.41061735e-01
2.40375504e-01 7.50686944e-01 3.84010166e-01 4.86133486e-01
6.50275588e-01 3.58436406e-02 -5.50248146e-01 -8.27489614e-01
3.93846661e-01 3.93124431e-01 -2.95020752e-02 -6.99492991e-01
-8.76625180e-01 6.33655906e-01 2.41392180e-01 4.20782059e-01
-4.22007084e-01 -3.98770213e-01 6.59233868e-01 3.90828103e-01
1.07222274e-01 -1.55689707e-02 -8.96488249e-01 2.01807067e-01
-6.03184044e-01 1.05958365e-01 6.87691867e-01 9.75066245e-01
4.46181625e-01 -2.62510449e-01 -8.53648260e-02 3.66807520e-01
3.58778350e-02 1.16314113e-01 6.65904045e-01 -8.57767463e-01
6.39786065e-01 9.44269180e-01 2.83819884e-01 -9.52203393e-01
-4.11198646e-01 -4.04809207e-01 -1.10410357e+00 2.84919828e-01
3.10008973e-01 -4.25277263e-01 -5.42605460e-01 1.58352828e+00
9.02330518e-01 -2.40363166e-01 1.76894620e-01 7.35513866e-01
5.50897598e-01 7.66568631e-02 -2.59038843e-02 -5.29686868e-01
1.41885602e+00 -7.24494457e-01 -1.11761785e+00 2.93218225e-01
6.16649330e-01 -5.62310040e-01 5.98451197e-01 5.73327482e-01
-9.33303535e-01 -2.86041468e-01 -6.60256505e-01 2.18263537e-01
-4.44734424e-01 -2.29232952e-01 7.49895573e-01 1.29988527e+00
-1.10785520e+00 3.00592244e-01 -7.51221716e-01 -4.77013111e-01
9.71583486e-01 9.16536868e-01 -8.45625222e-01 -1.97168782e-01
-8.81865561e-01 4.35810834e-01 1.11650348e-01 -2.90615320e-01
-3.00016761e-01 -7.47934937e-01 -4.61342514e-01 -1.04091793e-01
9.25946683e-02 -8.80749941e-01 8.05373192e-01 -7.79918671e-01
-5.69975853e-01 9.76621211e-01 1.91444755e-01 -6.51692867e-01
1.00406063e+00 8.96153152e-02 -7.64589071e-01 -7.91864768e-02
-1.70029342e-01 1.21868931e-01 6.37040317e-01 -9.89130437e-01
-8.02461088e-01 -1.15187430e+00 -4.32228744e-01 9.95801464e-02
-7.48517096e-01 1.14359476e-01 -5.20677678e-02 -3.88925374e-01
-2.99830902e-02 -7.43605673e-01 -1.13675281e-01 3.11191887e-01
-4.06949311e-01 1.31889254e-01 1.65832806e+00 -5.79266608e-01
1.18479681e+00 -2.05429387e+00 -7.37650275e-01 4.26500350e-01
2.65561521e-01 4.37562525e-01 4.02008891e-01 4.86604929e-01
-1.59494489e-01 3.59597564e-01 -1.80453002e-01 -3.42915744e-01
-2.70714879e-01 2.75916964e-01 2.51091391e-01 5.90522289e-01
-4.59061891e-01 4.34310853e-01 -4.04872298e-01 -6.47354186e-01
1.36528268e-01 4.78238106e-01 -2.14400709e-01 3.47897410e-01
1.71140909e-01 5.97728133e-01 -5.74140191e-01 1.02874744e+00
1.10151768e+00 -1.35525525e-01 4.51977551e-01 -2.15892300e-01
1.84906527e-01 -5.36855698e-01 -1.38322127e+00 1.40605855e+00
-9.30412635e-02 5.08769825e-02 5.85712969e-01 -7.20466137e-01
8.75292420e-01 7.84965575e-01 1.06762683e+00 -4.00461912e-01
4.78654623e-01 2.12239400e-01 -3.93894315e-01 -8.44396949e-01
1.78997248e-01 -2.16756523e-01 6.10299446e-02 7.13327467e-01
-2.85344303e-01 7.12363780e-01 -7.00693727e-01 -5.25234686e-03
1.29707348e+00 -4.11636949e-01 4.33513671e-01 -2.32326984e-01
7.61287034e-01 -3.59425575e-01 6.09902501e-01 7.54018426e-01
-8.34637702e-01 3.91112715e-01 9.57690999e-02 -8.17391217e-01
-8.35277200e-01 -7.52048314e-01 -2.89292157e-01 8.31877828e-01
-1.13524884e-01 5.78875616e-02 -4.69590068e-01 -1.20545661e+00
4.63127971e-01 5.32083511e-01 -6.12190425e-01 -1.47198588e-01
-3.61576885e-01 -6.94100499e-01 5.17070591e-01 1.50294006e-01
7.36092150e-01 -7.80395329e-01 -8.54851067e-01 -8.94626081e-02
3.78721356e-02 -5.97980857e-01 -5.04253209e-01 3.44639808e-01
-9.79416132e-01 -1.22890830e+00 -4.19257253e-01 -7.29168415e-01
8.16378355e-01 1.52622148e-01 7.13550985e-01 1.24615885e-01
-5.55826306e-01 4.03153479e-01 -3.70825790e-02 -7.39114165e-01
-3.06141883e-01 -5.93925938e-02 8.90521407e-02 3.92278016e-01
5.26696384e-01 -3.39556038e-01 -7.54656553e-01 5.69316804e-01
-1.07356560e+00 -5.56612074e-01 1.38944373e-01 7.86554873e-01
4.94795054e-01 4.01047617e-01 8.51682723e-01 -1.68721533e+00
6.63050771e-01 -8.24984193e-01 -2.84848124e-01 5.06125987e-01
-1.14909589e+00 -2.40649328e-01 5.58845937e-01 -1.06310740e-01
-1.41753173e+00 5.11477768e-01 2.35796601e-01 -6.40376985e-01
-3.50179791e-01 -2.62762189e-01 -4.47858959e-01 -2.38803431e-01
6.92758620e-01 -1.59480765e-01 4.52525944e-01 -6.58908129e-01
-1.71848554e-02 1.34294915e+00 5.64407766e-01 -2.37662941e-01
3.99463207e-01 7.08054245e-01 7.21576437e-02 -2.38662094e-01
-2.15395316e-01 -7.96743572e-01 -2.60017753e-01 1.02643564e-03
6.55768871e-01 -7.48266637e-01 -8.57382715e-01 4.58440781e-01
-7.25436509e-01 4.61161345e-01 -3.12478781e-01 2.34179527e-01
-1.57030150e-01 3.12817305e-01 -2.83564925e-01 -1.17816782e+00
-8.99607837e-01 -5.77501476e-01 6.61646545e-01 1.67685300e-01
-9.79129374e-02 -8.90664399e-01 -9.73400548e-02 8.37963581e-01
7.40298986e-01 7.90867329e-01 8.74236107e-01 -1.03493166e+00
-2.23984182e-01 -7.31339216e-01 1.94070563e-01 3.52528632e-01
7.07359791e-01 -4.30105269e-01 -1.19550335e+00 -7.30465531e-01
7.28638589e-01 -2.72533804e-01 6.36423603e-02 5.36658242e-02
1.30786681e+00 -8.17305267e-01 -4.35089320e-01 5.31104088e-01
1.68032789e+00 4.41383928e-01 3.45758289e-01 2.53463835e-01
3.79083157e-01 7.61084139e-01 6.43228412e-01 1.00216484e+00
1.02855437e-01 2.72197813e-01 7.43175209e-01 -1.94584355e-01
1.72320887e-01 6.98106810e-02 -3.09172571e-01 1.61041796e-01
5.63360035e-01 -2.65458226e-01 -7.43489802e-01 5.84133863e-01
-1.85951352e+00 -8.72880638e-01 -2.98290759e-01 2.41759157e+00
5.92870355e-01 -2.90360659e-01 1.21708073e-01 1.87434450e-01
9.52538133e-01 -2.66309381e-01 -9.01600540e-01 -7.34779656e-01
-1.61220450e-02 -2.45082080e-02 8.70008349e-01 -3.29467319e-02
-1.04411876e+00 3.54001969e-02 5.50211096e+00 3.56684148e-01
-1.04158187e+00 3.19154650e-01 1.04197454e+00 -2.06322014e-01
-2.53618926e-01 -3.67678255e-01 -4.66940969e-01 4.58849341e-01
8.33200276e-01 -2.10124746e-01 1.74163997e-01 1.07147849e+00
1.91375930e-02 -6.16611913e-03 -9.56883848e-01 1.30022061e+00
-3.31288755e-01 -1.46849835e+00 -1.81866363e-01 2.82243848e-01
1.04463291e+00 -2.28710659e-02 2.60192841e-01 -1.40648186e-01
3.37326862e-02 -7.39296734e-01 7.46169910e-02 5.13462543e-01
8.05460811e-01 -9.11111236e-01 1.02333832e+00 7.11784244e-01
-6.86722636e-01 -3.96843702e-01 -1.21786110e-01 2.84970075e-01
-3.67557794e-01 2.64776260e-01 -1.05120945e+00 7.71354377e-01
9.99772131e-01 -6.56362325e-02 -4.61699456e-01 7.71130621e-01
9.37040627e-01 2.54251331e-01 -4.25163925e-01 4.59349424e-01
-7.83140808e-02 1.36460781e-01 2.43356511e-01 8.40863883e-01
2.43404910e-01 1.47179931e-01 -2.99844056e-01 8.03194866e-02
-2.90023535e-01 3.41973215e-01 -1.01563346e+00 2.57522434e-01
6.69340611e-01 1.33574128e+00 -2.65966713e-01 -1.38847511e-02
-3.72748584e-01 6.30802751e-01 -5.16360514e-02 -1.75446853e-01
-5.55287004e-01 -2.44556949e-01 5.65633416e-01 4.40089762e-01
2.27753986e-02 4.84186232e-01 -5.19663036e-01 -6.44022524e-01
1.10568479e-01 -1.18637896e+00 1.19334638e+00 -3.21294010e-01
-1.46089470e+00 8.65149975e-01 -2.70239562e-01 -1.23367143e+00
-2.70729065e-01 -1.03364252e-01 -4.10627216e-01 6.69211149e-01
-1.19160438e+00 -1.46107996e+00 -4.63044256e-01 1.27977955e+00
-6.23481199e-02 -7.01693296e-01 1.03530157e+00 4.43816900e-01
-4.85551387e-01 9.53394175e-01 4.86579210e-01 -2.87150502e-01
9.20332730e-01 -9.53332245e-01 -3.08330983e-01 7.40982890e-01
-2.48295262e-01 7.96018422e-01 3.98251802e-01 -7.79991806e-01
-1.51203454e+00 -1.33420551e+00 8.15032482e-01 -5.44681370e-01
-1.87882543e-01 -2.09711015e-01 -6.08500123e-01 7.13495016e-01
3.05914640e-01 7.00439572e-01 1.18211842e+00 -2.82049626e-01
-5.87099884e-03 -7.37979829e-01 -2.50579166e+00 -7.55861476e-02
6.49360895e-01 -3.39778990e-01 -1.07520461e-01 5.89915037e-01
4.80231375e-01 -2.53355742e-01 -1.04776669e+00 2.08925411e-01
3.73249084e-01 -1.39024341e+00 6.64295971e-01 -8.88986349e-01
-4.43449229e-01 -2.04352923e-02 -4.47074950e-01 -2.85985798e-01
-1.08621508e-01 -1.14444816e+00 -2.55027294e-01 1.51493132e+00
1.78110115e-02 -8.28981578e-01 1.57787406e+00 1.50932741e+00
2.72833854e-01 -7.30130434e-01 -1.10255754e+00 -6.02729440e-01
-3.49042207e-01 -2.65870214e-01 1.03325915e+00 1.33917236e+00
-3.71537685e-01 -4.99074996e-01 -6.18540525e-01 2.53374577e-01
8.88661087e-01 -1.55653194e-01 8.24443698e-01 -1.02733612e+00
-1.13133579e-01 5.11787474e-01 -2.67270625e-01 6.06435537e-02
-3.80700111e-01 -7.30649233e-01 -6.92064583e-01 -1.07149351e+00
3.57179977e-02 -8.67809236e-01 -8.73576701e-01 8.30643952e-01
2.99281806e-01 1.71337008e-01 3.68314050e-02 3.43026549e-01
-4.05662179e-01 -1.35195717e-01 1.05493200e+00 1.33198708e-01
-6.90149963e-02 5.09260178e-01 -1.05382979e+00 2.85241961e-01
9.09845889e-01 -7.35201836e-01 -7.24402070e-01 -2.32751593e-01
-1.68230250e-01 3.21770132e-01 1.58788666e-01 -9.10520315e-01
5.54719687e-01 -1.81192651e-01 6.01875186e-01 -6.11710131e-01
-2.13780440e-02 -1.86368144e+00 9.09125388e-01 5.35877705e-01
-2.03533247e-01 1.04868300e-01 4.27444652e-02 5.33546150e-01
-8.89484361e-02 -4.91752289e-02 6.78463697e-01 -2.95760334e-01
-1.80564582e-01 3.70874345e-01 1.03372939e-01 -2.80576676e-01
1.76951945e+00 -5.07840037e-01 -4.49434698e-01 -3.44461501e-01
-7.22641110e-01 3.75276029e-01 5.51723123e-01 2.65664667e-01
3.64095151e-01 -1.17236710e+00 -2.74974436e-01 6.79766953e-01
-1.69557016e-02 2.96023749e-02 6.02703571e-01 7.36111999e-01
-4.26772237e-01 5.26358306e-01 -5.28926909e-01 -3.42129439e-01
-1.75958419e+00 8.39611888e-01 2.98931032e-01 -3.34256500e-01
-4.44572479e-01 7.23359525e-01 -1.09561801e-01 -6.59033954e-01
5.22880435e-01 8.34291503e-02 2.20633298e-01 2.96161547e-02
4.06695813e-01 7.58656681e-01 5.95993340e-01 -4.88238692e-01
-7.05634236e-01 -8.55929926e-02 -4.26148385e-01 3.99984419e-01
1.42911530e+00 -2.94019580e-01 -1.07392885e-01 -1.02224648e-01
1.54830050e+00 1.29479188e-02 -1.11782825e+00 -1.27559572e-01
-1.29779503e-01 -8.47661436e-01 -3.39728594e-02 -1.28225219e+00
-1.41108274e+00 3.59916449e-01 1.30843735e+00 8.04117471e-02
1.52630305e+00 -3.85161966e-01 9.10190403e-01 -3.99831086e-02
5.79139173e-01 -9.41864491e-01 -2.81687081e-01 -4.09305513e-01
4.75823849e-01 -1.27467370e+00 1.53312206e-01 5.86402640e-02
-8.85983586e-01 1.04092002e+00 5.29646993e-01 4.03298199e-01
8.90971720e-01 4.03494924e-01 3.51236045e-01 -1.49380460e-01
-7.65321434e-01 7.93116391e-01 -2.03067541e-01 9.41006303e-01
-5.41323423e-02 -6.59764484e-02 -2.51068562e-01 7.61129797e-01
2.16956168e-01 3.30895513e-01 2.58874476e-01 1.37214339e+00
9.47244391e-02 -1.50386620e+00 -6.48933232e-01 8.26879919e-01
-8.96429300e-01 5.67286849e-01 -4.51778859e-01 8.58515978e-01
7.23396182e-01 9.24840331e-01 -1.44333825e-01 -3.78366053e-01
2.26481140e-01 2.36336559e-01 -1.11626334e-01 -1.47139817e-01
-1.34461725e+00 -4.45289500e-02 -2.97663342e-02 -5.78131855e-01
-2.28872821e-01 -7.65720189e-01 -1.14382505e+00 -3.45745862e-01
-2.28650585e-01 1.97965711e-01 9.71290886e-01 2.63118356e-01
6.85280800e-01 1.87247545e-01 1.02421653e+00 1.36090383e-01
-1.01744509e+00 -3.14931244e-01 -8.68333817e-01 6.13824129e-01
5.48236847e-01 -8.63624737e-03 -2.30641082e-01 -1.21902339e-01]
|
[6.079319477081299, 6.512689113616943]
|
307ac3ce-e73e-4926-a8d4-52ca760cb311
|
caching-historical-embeddings-in
|
2211.14155
| null |
https://arxiv.org/abs/2211.14155v1
|
https://arxiv.org/pdf/2211.14155v1.pdf
|
Caching Historical Embeddings in Conversational Search
|
Rapid response, namely low latency, is fundamental in search applications; it is particularly so in interactive search sessions, such as those encountered in conversational settings. An observation with a potential to reduce latency asserts that conversational queries exhibit a temporal locality in the lists of documents retrieved. Motivated by this observation, we propose and evaluate a client-side document embedding cache, improving the responsiveness of conversational search systems. By leveraging state-of-the-art dense retrieval models to abstract document and query semantics, we cache the embeddings of documents retrieved for a topic introduced in the conversation, as they are likely relevant to successive queries. Our document embedding cache implements an efficient metric index, answering nearest-neighbor similarity queries by estimating the approximate result sets returned. We demonstrate the efficiency achieved using our cache via reproducible experiments based on TREC CAsT datasets, achieving a hit rate of up to 75% without degrading answer quality. Our achieved high cache hit rates significantly improve the responsiveness of conversational systems while likewise reducing the number of queries managed on the search back-end.
|
['Nicola Tonellotto', 'Raffaele Perego', 'Franco Maria Nardini', 'Cristina Ioana Muntean', 'Ida Mele', 'Ophir Frieder']
|
2022-11-25
| null | null | null | null |
['document-embedding', 'conversational-search']
|
['methodology', 'natural-language-processing']
|
[-2.88041115e-01 -2.23635495e-01 -3.93285602e-01 -3.59426796e-01
-1.09257603e+00 -5.93300641e-01 9.69078481e-01 4.89254951e-01
-7.45353460e-01 2.51045763e-01 9.75330770e-01 -2.46854439e-01
-4.00798738e-01 -9.00963604e-01 -3.89210582e-01 -1.95680216e-01
-3.16761076e-01 1.09434581e+00 5.72889507e-01 -3.80330771e-01
6.00253403e-01 3.48770440e-01 -1.85388887e+00 9.02701139e-01
2.75590539e-01 8.21902752e-01 -6.77728206e-02 9.32293594e-01
-4.29279745e-01 5.96188068e-01 -5.59280157e-01 -2.88048148e-01
-7.23000914e-02 7.47798979e-02 -1.21756709e+00 -6.66244447e-01
1.85154915e-01 -8.01543355e-01 -9.14575756e-01 3.45483333e-01
4.21310008e-01 3.87039334e-01 5.63374937e-01 -1.03549933e+00
-4.46723461e-01 4.43739533e-01 -8.61562863e-02 5.69599271e-01
7.70637631e-01 -1.15027182e-01 1.35189807e+00 -9.71535146e-01
6.61897898e-01 1.35337663e+00 2.46794328e-01 3.00433844e-01
-1.34103894e+00 -3.15355480e-01 -1.77884568e-02 2.22992629e-01
-1.44701576e+00 -6.33489311e-01 2.46349782e-01 -2.83061694e-02
1.46388626e+00 7.38303721e-01 3.79288137e-01 8.68452191e-01
9.90880430e-02 8.85762811e-01 5.25055647e-01 -3.44831079e-01
3.52391124e-01 3.34754109e-01 5.05662501e-01 3.06831390e-01
-4.10655551e-02 -2.24961098e-02 -1.06197703e+00 -9.73273456e-01
1.60312414e-01 1.50280640e-01 -2.97331035e-01 -3.68111968e-01
-9.25276637e-01 9.69426930e-01 4.71497327e-01 3.41250807e-01
-5.04747510e-01 4.33439553e-01 6.42267644e-01 6.06744230e-01
4.96945024e-01 2.87549555e-01 -2.11728811e-01 -6.05931818e-01
-8.79342437e-01 6.46037221e-01 1.54599631e+00 9.38563526e-01
6.43320858e-01 -8.71023536e-01 -4.41716641e-01 8.23136449e-01
1.65722176e-01 4.72928613e-01 4.69399929e-01 -1.05281365e+00
3.28744024e-01 5.59096873e-01 2.72972941e-01 -1.05246246e+00
-4.00611721e-02 -2.36664280e-01 -1.11481771e-01 -5.84821522e-01
-7.12767662e-03 6.01834238e-01 -1.32285163e-01 1.42263663e+00
3.52473289e-01 -2.27562562e-01 -6.44756258e-02 7.86275029e-01
3.41622949e-01 7.75363088e-01 -1.00677595e-01 -1.02960616e-01
1.56297922e+00 -8.50120425e-01 -4.85585570e-01 1.78708926e-01
1.05781138e+00 -1.05834246e+00 1.33225310e+00 -8.65744799e-02
-1.06381667e+00 -1.42977372e-01 -5.91980815e-01 -3.95930111e-01
-3.39221299e-01 -6.33483469e-01 6.88098192e-01 3.77611548e-01
-1.31799364e+00 2.49498263e-01 -9.32943642e-01 -6.23594463e-01
-3.72571319e-01 4.35498953e-01 -1.25561297e-01 -6.53885677e-02
-9.43513811e-01 6.31486773e-01 -8.44625533e-02 -3.53919864e-01
-6.29895329e-01 -8.06587696e-01 -7.05340803e-02 5.89696407e-01
2.22560003e-01 -5.90915561e-01 1.81092024e+00 7.98389968e-03
-9.95037973e-01 6.17555082e-01 -6.95111036e-01 -4.12501812e-01
1.27720580e-01 -4.67729062e-01 -3.60815525e-01 3.43549788e-01
-3.32315639e-02 5.71077466e-01 3.03395152e-01 -8.42628896e-01
-6.80284262e-01 -3.25940400e-01 1.20109342e-01 4.33091432e-01
-9.81718063e-01 9.55714807e-02 -9.31288362e-01 -8.56056530e-03
8.66539031e-02 -9.79738891e-01 1.66775972e-01 1.26122594e-01
4.90508862e-02 -7.74826407e-01 1.02455509e+00 -1.10223576e-01
1.63742113e+00 -2.22732353e+00 -1.19490333e-01 5.41070938e-01
3.48900884e-01 -4.86960523e-02 -2.81344801e-01 1.22919905e+00
7.21004963e-01 4.50173318e-02 5.18650174e-01 -1.57798141e-01
3.63395363e-01 1.51535496e-01 -7.99881756e-01 1.20637268e-01
-4.31292295e-01 9.77389932e-01 -7.78699279e-01 -5.00606239e-01
-1.68496832e-01 3.43350053e-01 -8.84258270e-01 4.67560709e-01
-3.29726666e-01 -3.52489412e-01 -8.78227115e-01 3.96394908e-01
2.17148721e-01 -5.33338189e-01 9.78876129e-02 -1.15573801e-01
2.56063454e-02 9.83131349e-01 -6.19561613e-01 1.58554029e+00
-7.89184868e-01 6.77689016e-01 2.20682651e-01 -3.69407594e-01
6.81383729e-01 3.30528140e-01 4.54887778e-01 -1.16832912e+00
-5.16749084e-01 3.32472682e-01 -5.34846365e-01 -4.85995263e-01
1.14992988e+00 6.35674894e-01 -1.96530193e-01 1.15931559e+00
-4.93970335e-01 -3.97545137e-02 1.67883739e-01 7.12382734e-01
1.43042302e+00 -5.34005702e-01 -2.92729288e-01 -4.29812938e-01
3.57254982e-01 2.01455042e-01 -3.02555531e-01 1.17413294e+00
9.33286399e-02 -1.19624073e-02 4.21981514e-01 -6.79748595e-01
-1.12456918e+00 -1.07820952e+00 -1.65611103e-01 1.68215883e+00
2.21319988e-01 -7.83269405e-01 -5.18383980e-01 -2.52896458e-01
4.04793173e-01 5.93928218e-01 -2.18955308e-01 -3.43990207e-01
-8.58101189e-01 -1.53790951e-01 3.78307819e-01 4.21173781e-01
2.89687216e-02 -8.90493393e-01 -5.20773113e-01 3.52706254e-01
-2.62703866e-01 -8.55866194e-01 -8.23830843e-01 -7.60869309e-02
-8.65053415e-01 -9.12741363e-01 -5.34932256e-01 -6.95893049e-01
1.91855773e-01 7.36275256e-01 1.42448461e+00 3.58628958e-01
-2.74460673e-01 9.10836637e-01 -2.66089380e-01 3.14891785e-01
-1.88320234e-01 3.95715982e-01 5.83915180e-03 -3.76398444e-01
1.11950266e+00 -6.45447433e-01 -1.21519160e+00 5.35347164e-01
-9.89426017e-01 -4.03211772e-01 2.38810778e-01 7.86301792e-01
3.25350314e-01 -4.76564586e-01 3.07247311e-01 -6.51269674e-01
1.30281126e+00 -6.62507236e-01 -4.35551465e-01 2.97755539e-01
-1.02874231e+00 3.67799431e-01 1.52439192e-01 -3.15356255e-01
-7.26249158e-01 -6.43912017e-01 3.80677450e-03 2.40654256e-02
1.79404974e-01 3.51355046e-01 7.60784149e-01 1.27837241e-01
8.07828248e-01 4.10380483e-01 4.27514464e-02 -5.90207160e-01
5.25962710e-01 1.20233667e+00 4.07225192e-01 -7.88764715e-01
2.39765465e-01 4.21777904e-01 -3.74618888e-01 -8.39075387e-01
-3.66437465e-01 -1.24577343e+00 1.61102396e-02 8.37988332e-02
3.58015120e-01 -8.71095181e-01 -1.35964680e+00 -2.91273057e-01
-1.15482628e+00 -2.17069872e-02 -1.27560541e-01 4.92409945e-01
-4.36096251e-01 1.63507417e-01 -9.53099370e-01 -7.73207486e-01
-6.66897655e-01 -9.96292770e-01 1.45148635e+00 -3.56426602e-03
-6.48614526e-01 -7.59661734e-01 5.84617138e-01 3.33360851e-01
9.96182084e-01 -7.24281549e-01 1.31145406e+00 -9.95346904e-01
-1.04566681e+00 -4.12613094e-01 -4.45860595e-01 -4.20881748e-01
-3.32097471e-01 -5.37821352e-01 -9.61107075e-01 -5.25109828e-01
-3.50085616e-01 -4.43186313e-01 7.04413414e-01 -1.43944427e-01
9.06059742e-01 -3.87758970e-01 -7.19160318e-01 1.95595875e-01
1.16437066e+00 -6.98003639e-03 2.24422008e-01 2.28707075e-01
-8.61973614e-02 6.65201128e-01 3.75960231e-01 5.84551930e-01
3.27235758e-01 1.21010065e+00 1.01714097e-02 4.77784127e-01
-5.45102321e-02 -4.26133424e-01 1.61902383e-01 1.02148449e+00
5.82725227e-01 -6.23451412e-01 -7.94273138e-01 7.29313672e-01
-1.86428416e+00 -8.39203358e-01 1.48398817e-01 2.41739559e+00
8.93228471e-01 -8.88504740e-03 -2.21408661e-02 -3.51428241e-01
2.47899905e-01 2.51760334e-01 -4.12698418e-01 -7.16919899e-01
4.53606367e-01 3.65800709e-01 1.51286870e-01 7.70270407e-01
-4.39239293e-01 8.31012368e-01 7.02941275e+00 5.67890882e-01
-8.37445319e-01 4.09561843e-02 2.43936762e-01 -6.30274534e-01
-8.15821886e-01 6.14152998e-02 -1.00925529e+00 3.68759185e-01
1.63641286e+00 -6.02572024e-01 6.42958760e-01 1.03281045e+00
1.87717274e-01 -6.61467537e-02 -1.45953107e+00 8.39320958e-01
-3.72495987e-02 -1.50163758e+00 2.76368141e-01 2.80006319e-01
2.47757599e-01 1.95915416e-01 1.19980469e-01 5.61719835e-01
2.11687863e-01 -7.09248006e-01 1.00955747e-01 5.61186910e-01
4.37268108e-01 -6.15074217e-01 5.74922144e-01 2.23679125e-01
-1.11332273e+00 -4.75223958e-02 -3.63898039e-01 1.47273406e-01
1.75831109e-01 4.30776179e-01 -1.10116422e+00 -2.22551122e-01
7.65473008e-01 -9.84653607e-02 -3.57630700e-01 8.82965803e-01
5.88725030e-01 4.61585999e-01 -7.10937202e-01 -6.19078159e-01
4.25535262e-01 1.87875256e-01 4.03124034e-01 1.31716859e+00
2.76125580e-01 9.22392774e-03 1.35620952e-01 5.62377036e-01
-3.53035003e-01 2.95613170e-01 -7.27696717e-01 -1.84148729e-01
1.09111798e+00 7.96805739e-01 -2.46378243e-01 -4.44978744e-01
-3.72031420e-01 1.07451248e+00 3.48248750e-01 4.12222594e-01
-3.64102989e-01 -4.76827562e-01 8.75326931e-01 3.04083824e-01
1.24601096e-01 -3.02214324e-01 3.78224462e-01 -7.92583346e-01
4.72958952e-01 -9.03958321e-01 3.40788066e-01 -5.33863366e-01
-1.04753029e+00 7.58934796e-01 1.59016386e-01 -7.16058135e-01
-9.25833881e-01 -1.45380676e-01 -3.49745631e-01 8.75923395e-01
-1.15500438e+00 -6.32921934e-01 -3.42487186e-01 5.37967324e-01
4.93886888e-01 -3.94124575e-02 1.18601286e+00 4.03261393e-01
1.99827433e-01 7.32724071e-01 5.60616553e-01 -4.45291668e-01
7.24977136e-01 -8.96050274e-01 5.51087976e-01 -5.30173108e-02
4.10991460e-01 1.30564225e+00 5.77036142e-01 -3.14993203e-01
-1.96095502e+00 -3.39149654e-01 1.52678549e+00 -4.64392781e-01
8.43540072e-01 -5.13713002e-01 -1.04479218e+00 2.76170224e-01
1.74583614e-01 -2.09472403e-01 7.75702417e-01 6.45642340e-01
-6.34626269e-01 -3.28346968e-01 -6.06469095e-01 7.93002725e-01
8.65272760e-01 -1.31325734e+00 -4.59251553e-01 8.76086712e-01
1.11206174e+00 -1.51091963e-01 -8.49201202e-01 -8.66897777e-02
8.03578138e-01 -8.26821208e-01 1.14913547e+00 -7.26357341e-01
3.57121751e-02 2.02003136e-01 -3.76577020e-01 -6.08111739e-01
-1.63663819e-01 -7.16813922e-01 -6.24401808e-01 8.58596146e-01
4.56269234e-01 -3.95750880e-01 1.20670342e+00 1.04732645e+00
1.93171397e-01 -8.38147998e-01 -9.06051040e-01 -6.25763774e-01
-2.37828955e-01 -1.22782521e-01 6.59117460e-01 1.26064271e-01
2.47777462e-01 1.98596254e-01 4.25940268e-02 -1.44391552e-01
2.73178428e-01 3.53097737e-01 9.76547360e-01 -1.03721941e+00
-4.65262443e-01 -3.43165606e-01 -1.33002281e-01 -1.77843547e+00
3.34472605e-03 -8.71772528e-01 -1.75119594e-01 -1.13669026e+00
4.32364315e-01 -6.47016287e-01 -3.25935751e-01 -6.61407411e-02
2.32079595e-01 -4.87249009e-02 -2.28618503e-01 7.49358296e-01
-9.30982828e-01 5.66432595e-01 6.29635930e-01 -1.67755812e-01
-2.95232743e-01 3.23747322e-02 -4.89916414e-01 1.87559605e-01
2.08260819e-01 -6.56861484e-01 -6.04561687e-01 -6.79400504e-01
3.82544219e-01 1.90315858e-01 1.20537378e-01 -6.66951060e-01
9.05478477e-01 9.75794941e-02 -2.37274796e-01 -7.51618445e-01
8.56071353e-01 -7.33927429e-01 -1.21743180e-01 4.21619505e-01
-1.18121290e+00 5.28655410e-01 -2.35040425e-04 9.07954395e-01
-4.53710407e-01 -2.08729818e-01 7.96177909e-02 1.49772093e-01
-4.68861759e-01 1.21185839e-01 -4.22760367e-01 2.64637113e-01
5.74000955e-01 1.18897021e-01 -3.48363787e-01 -8.21774423e-01
-2.91831493e-01 3.62128019e-01 2.77497917e-01 5.21488607e-01
6.46717787e-01 -1.16952264e+00 -4.25502390e-01 2.23605454e-01
5.66608548e-01 -5.72007060e-01 -3.25775295e-02 5.06510675e-01
-6.02795362e-01 1.19672191e+00 4.87690598e-01 -6.34421468e-01
-1.43517375e+00 4.09268290e-01 -1.39188290e-01 -3.26779783e-01
-7.03512132e-01 8.03459585e-01 -4.19301055e-02 -1.90361083e-01
7.21635103e-01 -4.04446684e-02 3.42154205e-01 -2.28100777e-01
9.38775778e-01 3.62296194e-01 3.12973589e-01 1.93618581e-01
-3.88492674e-01 2.55159050e-01 -7.81128705e-01 -5.11009097e-01
1.09700370e+00 -3.25571656e-01 -2.83574432e-01 2.22564697e-01
1.87900412e+00 8.50586146e-02 -5.35295486e-01 -6.73393130e-01
4.22858804e-01 -6.43795133e-01 1.84051573e-01 -4.82841015e-01
-3.32526267e-01 5.36644459e-01 6.25189364e-01 3.37145239e-01
7.00033545e-01 2.29663849e-01 1.35405147e+00 1.31807876e+00
5.40864348e-01 -1.02341568e+00 2.11163595e-01 4.85655129e-01
7.09265828e-01 -9.27641392e-01 -6.34737089e-02 3.84333245e-02
-8.77576023e-02 1.10671186e+00 1.17708795e-01 1.37692049e-01
6.67138815e-01 2.87581384e-01 -2.59308517e-02 -4.79402721e-01
-1.51641464e+00 2.00276151e-01 2.06355855e-01 7.91061446e-02
6.19947493e-01 -1.51949048e-01 -4.94016796e-01 -2.21996039e-01
-1.18743777e-01 -9.12315920e-02 -5.08454703e-02 1.06268644e+00
-7.73199260e-01 -1.27128017e+00 7.47142509e-02 4.78575379e-01
-3.75238329e-01 -3.47013146e-01 -3.07076067e-01 5.90280652e-01
-1.03063345e+00 1.13131845e+00 5.35851717e-01 -3.53901178e-01
1.89462334e-01 3.37051392e-01 -6.02020472e-02 -3.82867098e-01
-7.73032725e-01 -1.11073904e-01 3.41567636e-01 -1.10017872e+00
1.92990750e-01 -2.53063232e-01 -9.37121153e-01 -8.13989699e-01
-2.78549731e-01 9.86392438e-01 8.59265983e-01 4.41159159e-01
9.71403956e-01 -1.11330286e-01 7.13411450e-01 -2.12934792e-01
-1.15373075e+00 -7.40047455e-01 -2.62005985e-01 4.96451080e-01
1.37012735e-01 -2.49595270e-01 -5.34881115e-01 -3.67856115e-01]
|
[11.659317970275879, 7.652909278869629]
|
22a85dbf-ffa7-4834-86e9-151843e3ff74
|
on-semidefinite-relaxations-for-the-block
|
1406.5647
| null |
http://arxiv.org/abs/1406.5647v3
|
http://arxiv.org/pdf/1406.5647v3.pdf
|
On semidefinite relaxations for the block model
|
The stochastic block model (SBM) is a popular tool for community detection in
networks, but fitting it by maximum likelihood (MLE) involves a computationally
infeasible optimization problem. We propose a new semidefinite programming
(SDP) solution to the problem of fitting the SBM, derived as a relaxation of
the MLE. We put ours and previously proposed SDPs in a unified framework, as
relaxations of the MLE over various sub-classes of the SBM, revealing a
connection to sparse PCA. Our main relaxation, which we call SDP-1, is tighter
than other recently proposed SDP relaxations, and thus previously established
theoretical guarantees carry over. However, we show that SDP-1 exactly recovers
true communities over a wider class of SBMs than those covered by current
results. In particular, the assumption of strong assortativity of the SBM,
implicit in consistency conditions for previously proposed SDPs, can be relaxed
to weak assortativity for our approach, thus significantly broadening the class
of SBMs covered by the consistency results. We also show that strong
assortativity is indeed a necessary condition for exact recovery for previously
proposed SDP approaches and not an artifact of the proofs. Our analysis of SDPs
is based on primal-dual witness constructions, which provides some insight into
the nature of the solutions of various SDPs. We show how to combine features
from SDP-1 and already available SDPs to achieve the most flexibility in terms
of both assortativity and block-size constraints, as our relaxation has the
tendency to produce communities of similar sizes. This tendency makes it the
ideal tool for fitting network histograms, a method gaining popularity in the
graphon estimation literature, as we illustrate on an example of a social
networks of dolphins. We also provide empirical evidence that SDPs outperform
spectral methods for fitting SBMs with a large number of blocks.
|
['Arash A. Amini', 'Elizaveta Levina']
|
2014-06-21
| null | null | null | null |
['graphon-estimation']
|
['graphs']
|
[ 2.51366645e-01 2.28731036e-01 -2.96090990e-01 1.37169927e-01
-5.72835982e-01 -8.69874656e-01 2.65003145e-01 1.90921277e-01
-9.78867039e-02 7.92790055e-01 7.09125698e-02 -3.66835237e-01
-6.49460614e-01 -7.96464145e-01 -9.87746894e-01 -9.73183274e-01
-6.16981387e-01 7.91889966e-01 2.68125355e-01 -2.57480085e-01
6.17074482e-02 6.78740323e-01 -9.74717796e-01 2.97901072e-02
6.72525525e-01 5.76219976e-01 -3.09569873e-02 6.13026619e-01
4.82316017e-02 3.29125911e-01 -1.05532683e-01 -7.19573677e-01
6.08481765e-01 -1.53915167e-01 -8.46400380e-01 5.11960030e-01
4.24346358e-01 4.49674912e-02 -4.66340542e-01 1.23979390e+00
1.67959452e-01 -2.82212734e-01 5.41863263e-01 -1.86916327e+00
-3.27029824e-01 8.72528255e-01 -1.18525898e+00 1.89157296e-02
4.75969166e-01 -2.03328252e-01 1.42021537e+00 -5.78837991e-01
7.80910730e-01 1.27833498e+00 1.03355944e+00 3.06574643e-01
-1.81030095e+00 -6.29853308e-01 2.31413846e-03 -2.56394520e-02
-1.61261654e+00 -3.39647532e-01 6.50542796e-01 -4.56263244e-01
3.29601228e-01 5.22601545e-01 8.11884880e-01 8.74991059e-01
-2.69966692e-01 8.10497880e-01 1.32577765e+00 -3.84106874e-01
1.05688907e-01 1.24378949e-01 3.13024908e-01 6.35281980e-01
1.06931829e+00 -2.19282568e-01 -3.31822336e-01 -7.70479739e-01
7.92643189e-01 -4.69264276e-02 -5.75079441e-01 -1.06442297e+00
-1.06296968e+00 1.13483298e+00 1.58193499e-01 1.51883230e-01
-1.10931672e-01 1.30131319e-01 3.38135391e-01 1.07632034e-01
3.05917710e-01 1.02734521e-01 -7.59627521e-02 2.32065767e-01
-1.11667216e+00 2.85427451e-01 1.29695749e+00 1.17416143e+00
7.58149207e-01 -2.16026738e-01 2.10785702e-01 6.05803788e-01
3.13432455e-01 6.13726676e-01 -5.67781150e-01 -9.99495625e-01
6.68352187e-01 2.44310796e-01 9.80814919e-02 -1.37424481e+00
-2.30435744e-01 -6.12856984e-01 -1.15487945e+00 -1.77915409e-01
6.16141081e-01 8.82437304e-02 -2.74200827e-01 2.07060218e+00
2.43510917e-01 1.54802829e-01 -1.33747861e-01 7.21647322e-01
1.16349898e-01 5.93004286e-01 -4.39592808e-01 -6.02307439e-01
1.14915383e+00 -5.70605338e-01 -3.23967785e-01 -1.94098130e-01
3.75889301e-01 -3.67317230e-01 3.89696091e-01 4.85601693e-01
-1.10412014e+00 1.63424656e-01 -9.80715275e-01 4.71814394e-01
3.29326093e-01 -2.25539207e-01 9.77168322e-01 9.04414237e-01
-1.31825316e+00 5.33021986e-01 -5.44072866e-01 -5.67053914e-01
4.32137102e-01 5.07568717e-01 -6.91887975e-01 -2.94115186e-01
-7.98174322e-01 4.51693624e-01 1.34914517e-01 2.05767244e-01
-7.06156373e-01 -4.29652452e-01 -7.62691319e-01 2.55128622e-01
6.66028440e-01 -6.14007413e-01 5.29371798e-01 -8.00597191e-01
-6.98977172e-01 1.06630886e+00 -1.36382744e-01 -6.80788815e-01
5.85721672e-01 5.75028062e-01 -3.38285863e-02 4.42388117e-01
1.70296937e-01 1.59684509e-01 8.21836293e-01 -1.51190269e+00
1.97339281e-02 -1.34399474e-01 2.13258132e-01 -2.93240547e-01
-2.35633314e-01 -8.69736299e-02 -3.25498790e-01 -3.44658345e-01
3.54063272e-01 -1.17248392e+00 -5.39464593e-01 -4.07520980e-02
-5.60555220e-01 1.34826317e-01 2.29007572e-01 -4.71491843e-01
1.21223664e+00 -1.98324335e+00 4.47300762e-01 1.02990282e+00
7.01378107e-01 -9.99917556e-03 -3.25461775e-01 9.25424218e-01
-3.00359070e-01 2.15959579e-01 -6.37332439e-01 -5.17941475e-01
2.90182769e-01 6.25015855e-01 -3.85901362e-01 1.06962073e+00
-9.77217853e-02 3.88157070e-01 -8.39156210e-01 -5.02709389e-01
-2.87425399e-01 7.67258257e-02 -7.96363711e-01 -3.31257761e-01
2.33590752e-01 3.82217839e-02 -1.94521397e-01 3.85798693e-01
1.32894719e+00 -5.39913058e-01 8.37670922e-01 -6.03513494e-02
6.35425746e-02 3.76226865e-02 -1.61773884e+00 1.29747772e+00
7.73875415e-02 4.38372821e-01 8.66561115e-01 -1.42970788e+00
5.94269216e-01 1.83929682e-01 7.23549247e-01 8.81198328e-03
-1.00276522e-01 2.40774959e-01 6.08423688e-02 -6.21997677e-02
2.84577578e-01 -4.34273094e-01 -2.94453371e-02 6.62494183e-01
-1.38440663e-02 1.85947105e-01 6.03913963e-01 8.72990787e-01
1.21967435e+00 -4.28620428e-01 4.28466648e-01 -7.17171848e-01
4.42928135e-01 -1.59039035e-01 7.47378290e-01 1.11805999e+00
-1.89502031e-01 5.81349909e-01 1.13673687e+00 7.64399618e-02
-1.34026635e+00 -1.12886453e+00 -2.57118881e-01 4.43331122e-01
3.60472798e-02 -6.26370609e-01 -6.89434707e-01 -4.96853590e-01
2.53313273e-01 -7.41633028e-02 -5.75513601e-01 2.99531907e-01
-3.41630101e-01 -1.11225355e+00 5.93918741e-01 1.21293418e-01
4.75154221e-02 -2.38110676e-01 1.95436865e-01 8.75944495e-02
-3.46629411e-01 -1.42513764e+00 -5.12383819e-01 1.21398121e-01
-9.24283266e-01 -1.36057818e+00 -7.74025917e-01 -5.50866425e-01
8.20792258e-01 5.58532774e-01 1.16581059e+00 2.95998305e-01
-5.13679348e-02 6.73148870e-01 -1.42317548e-01 1.77343816e-01
-5.43751597e-01 -7.57686198e-02 2.81348675e-01 1.74911335e-01
4.02798094e-02 -1.03673947e+00 -3.06654602e-01 2.02105030e-01
-9.73575532e-01 5.64671010e-02 4.63107109e-01 7.82486379e-01
3.75250638e-01 1.95698351e-01 2.97601014e-01 -1.01944947e+00
4.00425851e-01 -8.17561150e-01 -8.75287473e-01 1.22678354e-01
-5.93524694e-01 1.17749885e-01 3.37459475e-01 -1.94991693e-01
-5.25105953e-01 2.58206856e-02 4.09570262e-02 -3.25287670e-01
3.77971172e-01 5.92522562e-01 -1.53117493e-01 -4.64418322e-01
3.67798448e-01 2.02278778e-01 2.71087319e-01 -3.40588361e-01
2.01473385e-01 3.36992860e-01 2.68788278e-01 -9.01247680e-01
1.09409654e+00 9.66652513e-01 5.68992078e-01 -8.39624465e-01
-6.89017117e-01 -8.30441773e-01 -3.31776619e-01 -2.32349709e-02
1.64742216e-01 -8.13964367e-01 -1.08349693e+00 1.78526163e-01
-1.06104171e+00 2.85758544e-02 -4.77809347e-02 2.72502750e-01
-6.23868644e-01 1.14853406e+00 -8.38510692e-01 -1.22235537e+00
2.59880740e-02 -8.62180054e-01 7.72853732e-01 -5.46666503e-01
1.53393233e-02 -1.07792866e+00 3.92366648e-01 3.91337276e-01
-1.62502620e-02 3.54771584e-01 7.31617451e-01 -5.56375086e-01
-7.56695390e-01 -8.01285505e-02 -4.68850285e-01 2.56567150e-01
-3.68232638e-01 1.29369602e-01 -5.81126571e-01 -7.12766707e-01
-1.14893764e-02 1.07778043e-01 1.03660440e+00 4.61591631e-01
7.40809023e-01 -7.58469701e-01 -4.90498662e-01 5.62240899e-01
1.78268325e+00 -6.34313166e-01 7.19357193e-01 1.72732219e-01
5.91552377e-01 7.19099879e-01 1.97629347e-01 6.23308182e-01
2.66944736e-01 7.38416255e-01 6.17949069e-01 4.04758118e-02
2.83511579e-01 -9.37743410e-02 4.27630633e-01 9.91539061e-01
-1.58028200e-01 -7.20104426e-02 -8.20449412e-01 6.37817442e-01
-2.10225630e+00 -1.25121891e+00 -7.42404878e-01 2.42810655e+00
9.49219048e-01 5.26506966e-03 6.48077607e-01 2.94434905e-01
8.82236660e-01 5.66772968e-02 9.13818702e-02 -2.65580982e-01
-5.46658754e-01 2.30510905e-01 1.00287890e+00 6.11175239e-01
-9.21246350e-01 1.89595193e-01 6.49691868e+00 1.01277435e+00
-3.95700812e-01 2.35792592e-01 1.07091412e-01 6.45583346e-02
-5.19334853e-01 4.35189664e-01 -6.64990962e-01 3.06415141e-01
6.22861326e-01 -2.00198188e-01 6.21380746e-01 8.06774914e-01
6.12078570e-02 -1.72123834e-01 -1.09065855e+00 1.00839531e+00
1.59115627e-01 -1.29830897e+00 -2.29103014e-01 6.91792727e-01
1.01061881e+00 -1.20011993e-01 -1.39594510e-01 -2.05281183e-01
4.81295109e-01 -7.92628884e-01 6.13633811e-01 1.21537596e-01
5.37615657e-01 -7.36558557e-01 5.72699904e-01 3.60354781e-01
-1.22572029e+00 -2.49428358e-02 -5.89767814e-01 -9.10947099e-02
1.64831042e-01 1.07146919e+00 -6.85737610e-01 6.63415074e-01
2.59537309e-01 7.08369136e-01 -2.10423648e-01 1.28519833e+00
1.30352870e-01 6.81549132e-01 -7.59029865e-01 2.69122332e-01
1.17578276e-01 -6.57720268e-01 1.03910625e+00 1.45712781e+00
2.78489947e-01 -1.87225267e-01 1.20572403e-01 9.64306653e-01
-1.51869655e-01 5.93990013e-02 -5.07205725e-01 -1.29194647e-01
4.05603796e-01 1.23649907e+00 -9.26076710e-01 -1.30787507e-01
-3.08576792e-01 6.52037978e-01 3.24283332e-01 3.36000562e-01
-7.65323639e-01 8.17054063e-02 6.74365938e-01 3.03593576e-01
6.49756372e-01 -3.42312574e-01 -2.17450410e-01 -1.46922743e+00
1.57006130e-01 -1.11584508e+00 4.34796989e-01 -2.87661105e-01
-1.67606843e+00 1.95468552e-02 1.97095424e-01 -1.18618000e+00
1.33878082e-01 -4.98030573e-01 -3.77347052e-01 7.04861403e-01
-1.23651564e+00 -1.00225961e+00 9.56469700e-02 5.37277579e-01
-2.66382247e-01 3.87521207e-01 5.36644936e-01 4.91523594e-01
-5.30591667e-01 3.40412438e-01 3.40703994e-01 -7.51337856e-02
4.22627777e-01 -1.28854048e+00 1.05610244e-01 1.18958235e+00
3.45794886e-01 7.98887074e-01 1.08420682e+00 -5.92875242e-01
-1.47507167e+00 -6.87287986e-01 6.46532953e-01 -2.32139647e-01
1.14302051e+00 -6.53147936e-01 -7.49029160e-01 7.76647449e-01
-1.32257506e-01 -9.59000736e-02 6.85245514e-01 3.20621252e-01
-5.99297166e-01 -1.73657566e-01 -1.03374994e+00 3.55601907e-01
1.26466346e+00 -4.65301275e-01 -2.04876035e-01 5.86100101e-01
2.47290581e-01 -2.77478620e-02 -7.80352771e-01 7.95681402e-02
4.23063099e-01 -1.06287670e+00 1.16968262e+00 -3.47920179e-01
2.42976010e-01 -4.00827706e-01 -3.05139959e-01 -9.01310682e-01
-4.21057314e-01 -9.08305764e-01 -1.97922811e-01 1.21249771e+00
1.72473997e-01 -7.15510964e-01 8.56398463e-01 1.76040769e-01
2.23986149e-01 -5.96941233e-01 -1.05000424e+00 -1.01050413e+00
-3.33147645e-02 -5.00397444e-01 2.90890396e-01 1.18053365e+00
2.38107324e-01 9.69350860e-02 -6.93819225e-01 3.94443035e-01
1.36933661e+00 1.96919039e-01 8.53544176e-01 -1.40029049e+00
-8.74872684e-01 -4.05681759e-01 -4.52914923e-01 -1.01679349e+00
1.54143155e-01 -1.06775987e+00 -1.76762179e-01 -1.30961192e+00
9.67721462e-01 -7.48183608e-01 1.19416222e-01 2.20907897e-01
2.48708427e-01 4.66867954e-01 3.59229505e-01 4.22879249e-01
-5.86010993e-01 1.53150812e-01 1.01755202e+00 -2.00935468e-01
1.51864633e-01 2.00842083e-01 -8.16208005e-01 6.91331565e-01
3.11638057e-01 -7.08560288e-01 -1.34790599e-01 1.33551598e-01
8.76464367e-01 2.38345176e-01 5.43286324e-01 -7.23749697e-01
2.69298911e-01 -1.54967472e-01 -2.49635264e-01 -5.75510919e-01
2.79463857e-01 -8.96906853e-01 6.49732113e-01 5.49878955e-01
-1.62136257e-02 -3.53500813e-01 -2.30690725e-02 9.42402959e-01
1.00235999e-01 -6.05992258e-01 6.20985806e-01 4.47408371e-02
-1.75968096e-01 3.64775985e-01 -3.28762114e-01 9.99512151e-02
8.18729401e-01 -4.28210080e-01 -2.65977383e-01 -6.84295356e-01
-7.66354382e-01 2.56588221e-01 6.43165410e-01 -4.07698274e-01
4.26484734e-01 -1.16897893e+00 -9.17708993e-01 -1.51858881e-01
2.15174146e-02 -3.39420646e-01 2.02311501e-01 1.55028117e+00
-5.36128402e-01 2.44526535e-01 -2.40180213e-02 -7.43616819e-01
-1.46542001e+00 7.77261853e-01 4.93316054e-02 -5.20370841e-01
-4.20044899e-01 7.23854959e-01 3.17078739e-01 -1.55093744e-01
3.11073232e-02 2.58419104e-02 2.13687629e-01 2.44166274e-02
1.94956124e-01 4.75131214e-01 -1.94970831e-01 -6.26814127e-01
-4.69296962e-01 4.20605659e-01 2.10191518e-01 -1.07271103e-02
1.60006070e+00 -4.33386117e-01 -7.54770517e-01 5.51788695e-02
1.12154710e+00 7.73426831e-01 -1.06247056e+00 -1.35789514e-01
-8.70760754e-02 -4.15917695e-01 -4.82067645e-01 -4.81023826e-02
-1.10395265e+00 4.99365300e-01 -1.60273537e-01 6.18174911e-01
9.54684317e-01 2.13497072e-01 5.65357924e-01 2.31195539e-01
6.69621944e-01 -5.78801274e-01 -2.92391717e-01 2.64394581e-01
7.42557704e-01 -8.23552132e-01 3.78114074e-01 -1.00568414e+00
-2.26707414e-01 9.59781587e-01 -1.01237983e-01 -3.51263136e-01
3.45392197e-01 3.05906534e-01 -8.84990633e-01 -2.21432313e-01
-5.05269408e-01 -2.11421207e-01 1.90075638e-03 6.31771624e-01
-1.17754057e-01 1.31049290e-01 -4.90610987e-01 5.98488450e-01
5.87710692e-03 -3.58849317e-01 9.63057876e-01 5.68526089e-01
-2.38584697e-01 -1.33343589e+00 -5.64528763e-01 3.43575567e-01
-3.90445173e-01 -2.33149916e-01 -4.24665213e-01 8.86954904e-01
-2.64921244e-02 8.03296387e-01 -2.27554381e-01 -7.44728968e-02
-1.71931237e-01 -2.85987854e-01 7.50789642e-01 -5.82618356e-01
-2.12218300e-01 1.44540876e-01 3.59426379e-01 -3.29969108e-01
-7.95470417e-01 -6.57690942e-01 -6.79956973e-01 -9.95509982e-01
-6.24230862e-01 4.86367315e-01 3.24443281e-01 8.26629698e-01
-7.45013030e-03 -9.40250754e-02 6.12152398e-01 -6.37034178e-01
-7.09664524e-01 -6.00272536e-01 -1.07680285e+00 3.20554048e-01
3.30172688e-01 -6.15763187e-01 -8.29027951e-01 -1.55625001e-01]
|
[6.897597312927246, 5.079958438873291]
|
31295121-1fc7-4a05-8158-bcbba7fcd3cd
|
context-guided-triple-matching-for-multiple
|
2109.12996
| null |
https://arxiv.org/abs/2109.12996v1
|
https://arxiv.org/pdf/2109.12996v1.pdf
|
Context-guided Triple Matching for Multiple Choice Question Answering
|
The task of multiple choice question answering (MCQA) refers to identifying a suitable answer from multiple candidates, by estimating the matching score among the triple of the passage, question and answer. Despite the general research interest in this regard, existing methods decouple the process into several pair-wise or dual matching steps, that limited the ability of assessing cases with multiple evidence sentences. To alleviate this issue, this paper introduces a novel Context-guided Triple Matching algorithm, which is achieved by integrating a Triple Matching (TM) module and a Contrastive Regularization (CR). The former is designed to enumerate one component from the triple as the background context, and estimate its semantic matching with the other two. Additionally, the contrastive term is further proposed to capture the dissimilarity between the correct answer and distractive ones. We validate the proposed algorithm on several benchmarking MCQA datasets, which exhibits competitive performances against state-of-the-arts.
|
['Wanqing Li', 'Jie Yang', 'Junping Liu', 'Xinrong Hu', 'Junlong Ma', 'Xun Yao']
|
2021-09-27
| null | null | null | null |
['multiple-choice-qa']
|
['natural-language-processing']
|
[ 1.28667757e-01 -1.63727209e-01 1.25534832e-01 -4.07481730e-01
-1.41173089e+00 -4.74799693e-01 5.94913423e-01 5.05758584e-01
-5.50906539e-01 6.20106876e-01 2.38103867e-01 -2.07587391e-01
-3.50239128e-01 -5.65789998e-01 -4.91464078e-01 -5.30627728e-01
6.16899490e-01 4.10795897e-01 6.98666513e-01 -2.14036837e-01
6.72425151e-01 -1.12268724e-01 -1.72276270e+00 6.02147639e-01
1.41051948e+00 1.21626437e+00 2.61392891e-01 3.73694897e-01
-6.67248905e-01 1.13270974e+00 -4.98616934e-01 -6.89374387e-01
-8.80404860e-02 -8.44079196e-01 -9.14617360e-01 -1.12605281e-01
7.19239771e-01 1.53824404e-01 7.92392567e-02 1.11615205e+00
5.80673516e-01 3.97934109e-01 5.83250344e-01 -1.03974080e+00
-4.21569586e-01 1.89229354e-01 -3.31251711e-01 5.53207338e-01
7.21551836e-01 6.53886274e-02 1.39040160e+00 -1.25211203e+00
4.10544366e-01 1.33060980e+00 2.48818666e-01 4.20463264e-01
-9.22563493e-01 -3.46006155e-01 2.00986415e-01 7.17227697e-01
-1.06089330e+00 -2.75082111e-01 9.38202620e-01 -4.15556997e-01
5.28941393e-01 4.66866195e-01 1.69465482e-01 7.80184507e-01
-1.42495260e-01 9.88042772e-01 1.27159023e+00 -4.95909840e-01
3.16777498e-01 2.31854588e-01 5.54252028e-01 4.97380883e-01
-3.32161754e-01 -2.96915919e-01 -4.12135184e-01 -3.91952276e-01
-1.13441251e-01 -2.88490504e-01 -3.81923020e-01 -2.54938662e-01
-8.94025207e-01 6.45751476e-01 3.21642041e-01 3.97418737e-01
-4.59882140e-01 -4.16741788e-01 1.34656459e-01 3.77788872e-01
2.15439901e-01 1.71965212e-01 -2.10510820e-01 4.15687710e-02
-1.00366306e+00 5.56753278e-01 8.86331499e-01 5.99689543e-01
7.71073043e-01 -6.33880794e-01 -8.62948179e-01 9.35930550e-01
4.44890559e-01 2.74611622e-01 5.64458430e-01 -8.11923385e-01
9.01759148e-01 1.03694630e+00 3.03784937e-01 -1.11817575e+00
-1.64333880e-01 -5.33430278e-01 -3.54612589e-01 -1.92028210e-01
5.95354199e-01 1.73848376e-01 -2.87829369e-01 1.73638797e+00
6.92250252e-01 1.32817656e-01 -2.13359430e-01 1.09261334e+00
1.10970902e+00 4.63035136e-01 2.32001200e-01 -2.02045441e-01
1.45037901e+00 -1.25523329e+00 -7.80426681e-01 -2.20093429e-01
3.50969762e-01 -9.61614430e-01 1.36275339e+00 1.79104224e-01
-1.15272200e+00 -5.08942068e-01 -1.05756950e+00 -2.30575681e-01
-1.04322329e-01 1.97189987e-01 -1.11622445e-01 2.92108685e-01
-4.87212092e-01 2.23837629e-01 -1.93368383e-02 5.21851592e-02
2.30640965e-03 -8.28899965e-02 1.31779453e-02 -3.00983310e-01
-1.50725782e+00 1.07721114e+00 9.01510045e-02 3.44426744e-02
-4.11369890e-01 -5.23431659e-01 -5.24030328e-01 2.69079119e-01
5.79148829e-01 -9.34149206e-01 1.11662412e+00 -1.04582024e+00
-1.20409477e+00 9.88800526e-01 -4.55467880e-01 -9.05884206e-02
6.29814863e-01 -1.54543489e-01 -5.89397848e-01 2.74731487e-01
2.89056867e-01 2.16794237e-01 9.43743587e-01 -1.29898512e+00
-7.76816189e-01 -4.67784971e-01 2.13540703e-01 4.24182355e-01
-1.04366496e-01 1.98710203e-01 -6.62827313e-01 -4.02195394e-01
3.11251134e-01 -4.70525384e-01 1.23142324e-01 -2.32181102e-01
-1.53120965e-01 -5.77878237e-01 4.47294533e-01 -8.70004773e-01
1.69356549e+00 -2.01706457e+00 3.32168281e-01 8.52879286e-02
1.01512507e-01 2.15286463e-01 -2.99319476e-01 2.83651948e-01
1.48718148e-01 -2.56517649e-01 -2.83885539e-01 -3.92625451e-01
7.87968561e-02 -1.20113656e-01 -1.22629523e-01 2.52542853e-01
3.70587498e-01 6.34043336e-01 -1.03412437e+00 -7.48933017e-01
-1.37477010e-01 -5.82700712e-04 -4.66279179e-01 4.78758216e-01
-4.24669147e-01 4.69632417e-01 -6.61944270e-01 6.41774416e-01
8.02153587e-01 -3.57359380e-01 1.54416963e-01 -2.07792297e-01
-9.23756231e-03 6.01161122e-01 -1.46567893e+00 1.39840126e+00
-3.26485783e-01 -1.07260589e-02 2.51624525e-01 -9.49980855e-01
9.29717720e-01 8.93600360e-02 9.94574502e-02 -1.13585532e+00
1.88210219e-01 5.46824753e-01 8.07184130e-02 -1.04013491e+00
5.13243794e-01 -8.14552754e-02 6.14300296e-02 3.23482156e-01
8.71579349e-03 9.13373083e-02 3.73321235e-01 1.92268148e-01
1.09943151e+00 5.06559014e-02 2.05459103e-01 -2.97826797e-01
1.32768714e+00 -1.08060218e-01 5.51952541e-01 5.86919725e-01
-4.11583424e-01 6.30685568e-01 5.43423116e-01 -8.58310536e-02
-5.56193054e-01 -1.08146799e+00 2.47523248e-01 1.04363823e+00
5.12819827e-01 -1.68371260e-01 -8.31979811e-01 -9.96541142e-01
-1.39888182e-01 8.10422480e-01 -4.14275378e-01 -1.22860625e-01
-8.36373210e-01 -5.88378966e-01 2.13650599e-01 1.67419270e-01
5.46805024e-01 -9.78743315e-01 -3.51313412e-01 1.75854072e-01
-8.62069368e-01 -8.67648363e-01 -6.69285893e-01 -2.40428999e-01
-6.17854238e-01 -1.36704290e+00 -3.59073162e-01 -8.43263090e-01
3.55986089e-01 3.29371959e-01 1.47813725e+00 4.12603676e-01
1.71293423e-01 3.58472526e-01 -4.88127291e-01 8.87688175e-02
-2.92758822e-01 -1.10154733e-01 -3.87702614e-01 4.90700245e-01
5.38849890e-01 -3.40123326e-01 -7.15269387e-01 3.92671704e-01
-9.39657032e-01 -2.18314439e-01 5.21530449e-01 8.57975602e-01
6.72943950e-01 -4.24459696e-01 9.48081493e-01 -6.82874382e-01
1.01132262e+00 -7.89828062e-01 -3.15559953e-01 7.24162340e-01
-4.61999744e-01 6.85471147e-02 5.73465645e-01 -2.92713761e-01
-1.21989119e+00 -3.18184495e-01 -3.60916466e-01 -3.96759398e-02
9.69074517e-02 5.13334930e-01 -4.17416275e-01 6.89786822e-02
4.66540635e-01 1.68568596e-01 -2.45637774e-01 -4.71680135e-01
4.58734542e-01 5.98400593e-01 3.47374320e-01 -5.55777311e-01
5.37237108e-01 3.92439999e-02 -1.82511598e-01 -3.04437220e-01
-1.22949815e+00 -8.32522452e-01 -2.79671937e-01 -4.95561570e-01
7.09270418e-01 -4.86038536e-01 -6.70004308e-01 9.86368135e-02
-1.26971173e+00 3.06939423e-01 -1.18480036e-02 3.99656504e-01
-2.87856013e-01 5.72874129e-01 -3.92657667e-01 -8.56936276e-01
-2.64256895e-01 -1.06164503e+00 7.21281588e-01 3.23942840e-01
-2.11909190e-01 -6.83375001e-01 2.21960291e-01 1.07588041e+00
3.40761632e-01 -6.36015385e-02 1.24790585e+00 -1.05900860e+00
-6.87231839e-01 -3.22266102e-01 -2.12253615e-01 4.55494732e-01
-2.53364861e-01 -3.33780378e-01 -8.42644215e-01 -3.23192738e-02
4.85604018e-01 -4.77394253e-01 8.96809995e-01 -4.57119644e-02
8.61734688e-01 -1.35023311e-01 -3.04213935e-03 -1.47055954e-01
1.30361545e+00 4.28454615e-02 6.17820024e-01 3.95197123e-01
3.97276610e-01 7.49219120e-01 9.87145424e-01 1.76987767e-01
6.61856890e-01 6.91971123e-01 4.82467294e-01 2.96699613e-01
-2.44750872e-01 -2.23969758e-01 1.56301379e-01 1.16867495e+00
3.99281234e-01 -4.20135826e-01 -6.91946805e-01 5.81171691e-01
-1.90484786e+00 -9.41556752e-01 -4.02045995e-01 2.31472135e+00
7.37019777e-01 1.83937237e-01 -6.45131022e-02 2.67732382e-01
9.27103460e-01 1.93733394e-01 -4.36544418e-01 -5.89555688e-02
-2.94291854e-01 1.92594528e-01 -4.26906735e-01 6.49761438e-01
-8.92487884e-01 4.13970202e-01 5.38162708e+00 1.38623405e+00
-5.71272016e-01 2.60077864e-01 5.44933021e-01 5.60594834e-02
-7.08077312e-01 1.79605529e-01 -6.39480412e-01 5.61788142e-01
6.72419727e-01 3.37158442e-02 1.72482982e-01 4.26784635e-01
2.29902878e-01 -3.63483518e-01 -9.15056944e-01 6.02337062e-01
3.47522289e-01 -9.88783240e-01 1.07691802e-01 -6.72613800e-01
4.56143558e-01 -3.95745665e-01 -3.95192914e-02 4.27023023e-01
-3.36318940e-01 -5.66951692e-01 7.91835546e-01 8.81719053e-01
4.13461849e-02 -5.55432856e-01 9.24562991e-01 7.59800196e-01
-1.16686976e+00 -2.58313268e-01 2.28278581e-02 -8.45554546e-02
4.33375575e-02 5.86122572e-01 -1.91219673e-01 8.33573580e-01
5.60481489e-01 1.95845321e-01 -8.10320258e-01 1.25586843e+00
-3.20061773e-01 6.26086593e-01 4.69208620e-02 -2.77076155e-01
1.43472463e-01 -4.24511939e-01 7.59465992e-01 9.40719366e-01
1.59247607e-01 1.05062403e-01 5.32809831e-02 9.63265300e-01
-2.00606450e-01 6.21330380e-01 1.46580145e-01 4.18104202e-01
7.02413976e-01 1.33217680e+00 -3.50007236e-01 -2.43602380e-01
-6.64733052e-01 8.73919308e-01 7.23435283e-01 9.46780592e-02
-7.71018147e-01 -2.54743874e-01 4.72559370e-02 -6.24851882e-02
2.29104221e-01 3.48537326e-01 -3.03093612e-01 -1.19206524e+00
6.55284524e-01 -1.07662880e+00 7.33728111e-01 -6.64519608e-01
-1.69911683e+00 5.66549599e-01 -3.90211195e-01 -1.43144226e+00
6.98653981e-02 -2.14571103e-01 -6.37986541e-01 1.10024905e+00
-1.75235295e+00 -8.76380682e-01 -3.53362232e-01 3.68176252e-01
5.78574717e-01 1.05894782e-01 5.20136297e-01 8.07783425e-01
-6.66463614e-01 6.11053586e-01 -2.37848554e-02 -2.94514358e-01
7.96703398e-01 -1.14666867e+00 -2.50503480e-01 9.37735021e-01
-6.24837354e-02 4.75495726e-01 7.12373018e-01 -3.74984443e-01
-9.74590719e-01 -6.71433389e-01 1.39821410e+00 -5.02724349e-01
4.71508265e-01 1.08276583e-01 -1.28280985e+00 -2.06046086e-02
2.42464721e-01 -5.58633842e-02 5.57707846e-01 -1.81877792e-01
-5.02280891e-01 -1.47076756e-01 -1.25649416e+00 4.12508547e-01
5.34778774e-01 -5.64536810e-01 -9.88706946e-01 -1.08870575e-02
7.02740908e-01 -2.99592584e-01 -5.16605556e-01 4.89661574e-01
2.64565676e-01 -1.22350347e+00 8.74728560e-01 -5.20907760e-01
6.29246116e-01 -5.49050570e-01 -2.74864346e-01 -9.47951555e-01
-2.22326964e-01 -5.16787320e-02 -3.64916086e-01 1.58683693e+00
4.95729893e-01 -2.00461015e-01 5.04808545e-01 4.45612311e-01
-2.16306120e-01 -9.23932433e-01 -1.12392569e+00 -4.19766605e-01
4.80478071e-02 -5.84683497e-04 4.69567657e-01 7.94441223e-01
-1.04713859e-02 6.76956236e-01 -1.50587454e-01 2.78303772e-02
3.64475816e-01 4.25381929e-01 2.10146159e-01 -1.02679944e+00
-3.32463413e-01 -6.07743442e-01 8.16670209e-02 -1.31001604e+00
2.09714234e-01 -8.87736380e-01 2.17515007e-01 -1.47301877e+00
3.99830192e-01 -2.81613737e-01 -5.02337575e-01 -2.68490136e-01
-9.78780150e-01 -1.41786799e-01 8.77089575e-02 1.99248821e-01
-9.98850226e-01 7.76338458e-01 1.19655955e+00 -2.34252602e-01
2.33912952e-02 2.50861704e-01 -6.23901129e-01 6.66409194e-01
6.83001280e-01 -6.60479903e-01 -4.59833264e-01 -4.07697976e-01
3.71820420e-01 3.41630489e-01 3.80631953e-01 -9.65875804e-01
5.22211671e-01 -9.78263468e-02 -1.59080505e-01 -6.88197494e-01
2.50870913e-01 -6.54554188e-01 -3.62147540e-01 2.79093653e-01
-7.51320779e-01 1.45779967e-01 -1.03957362e-01 7.66267955e-01
-5.59237659e-01 -7.15618551e-01 6.89068139e-01 -1.18140079e-01
-6.05586112e-01 -2.16713957e-02 -1.48687854e-01 6.42507315e-01
6.63969219e-01 1.90806612e-01 -4.00977761e-01 -1.89629927e-01
-4.97829914e-01 5.32067537e-01 1.61051620e-02 4.40693945e-01
7.86568284e-01 -1.37776148e+00 -9.25635993e-01 -3.44313562e-01
3.72035563e-01 -4.67374027e-01 5.78282058e-01 1.09270334e+00
-1.97459668e-01 1.77252620e-01 2.48535767e-01 -2.59271652e-01
-1.33983254e+00 5.80461085e-01 4.85293537e-01 -6.90285206e-01
-8.14572945e-02 8.07266414e-01 -2.09790051e-01 -5.46915472e-01
3.64542603e-01 8.71783346e-02 -7.92913020e-01 3.07938755e-01
5.55142939e-01 6.04990184e-01 3.44050467e-01 -6.88944578e-01
-5.17480254e-01 5.14373720e-01 7.90829137e-02 -1.61899164e-01
7.36097515e-01 -4.42315370e-01 -4.59810942e-01 4.66195703e-01
1.03750420e+00 2.40401447e-01 -7.64409959e-01 -5.39762080e-01
4.41812366e-01 -4.85596597e-01 -1.60766438e-01 -9.11216915e-01
-6.69474065e-01 7.38942087e-01 5.70972264e-01 1.99158669e-01
1.15826762e+00 4.72080670e-02 9.44695354e-01 2.89726347e-01
-3.69139090e-02 -1.12336993e+00 3.22558820e-01 5.31008899e-01
9.17076170e-01 -1.33647978e+00 -2.77061313e-01 -4.16583121e-01
-5.57499826e-01 8.50364208e-01 8.40560973e-01 -1.75220501e-02
4.76550698e-01 -3.89703304e-01 7.35536665e-02 -2.98081696e-01
-9.03974891e-01 -2.68806666e-01 6.99393034e-01 -4.13386375e-02
5.24726689e-01 -1.55568510e-01 -1.06935346e+00 9.42062676e-01
2.77293831e-01 -2.30449945e-01 6.02919906e-02 8.21460962e-01
-6.54461503e-01 -1.03561354e+00 -2.80888766e-01 5.23377955e-01
-4.95007604e-01 -1.39942378e-01 -3.11316550e-01 3.00032735e-01
3.84252667e-01 1.50351357e+00 -2.47363552e-01 -3.43013555e-01
6.72897279e-01 2.08891034e-01 3.14546436e-01 -2.89050668e-01
-1.08063745e+00 -2.24204004e-01 1.61661267e-01 -4.54379827e-01
-7.22902000e-01 -6.43173277e-01 -9.28303182e-01 -2.36094706e-02
-5.70264101e-01 4.29987878e-01 2.46628150e-01 1.28837419e+00
2.14780688e-01 5.63344479e-01 6.73177779e-01 -1.09748088e-01
-9.18474793e-01 -9.69536304e-01 -2.16752112e-01 8.59619558e-01
3.48675996e-01 -5.95245361e-01 -5.89215457e-01 -3.77405375e-01]
|
[11.3480806350708, 8.034926414489746]
|
dd0ddb2d-3510-411f-9b58-e5a5b3f15713
|
how-bayesian-should-bayesian-optimisation-be
|
2105.00894
| null |
https://arxiv.org/abs/2105.00894v1
|
https://arxiv.org/pdf/2105.00894v1.pdf
|
How Bayesian Should Bayesian Optimisation Be?
|
Bayesian optimisation (BO) uses probabilistic surrogate models - usually Gaussian processes (GPs) - for the optimisation of expensive black-box functions. At each BO iteration, the GP hyperparameters are fit to previously-evaluated data by maximising the marginal likelihood. However, this fails to account for uncertainty in the hyperparameters themselves, leading to overconfident model predictions. This uncertainty can be accounted for by taking the Bayesian approach of marginalising out the model hyperparameters. We investigate whether a fully-Bayesian treatment of the Gaussian process hyperparameters in BO (FBBO) leads to improved optimisation performance. Since an analytic approach is intractable, we compare FBBO using three approximate inference schemes to the maximum likelihood approach, using the Expected Improvement (EI) and Upper Confidence Bound (UCB) acquisition functions paired with ARD and isotropic Matern kernels, across 15 well-known benchmark problems for 4 observational noise settings. FBBO using EI with an ARD kernel leads to the best performance in the noise-free setting, with much less difference between combinations of BO components when the noise is increased. FBBO leads to over-exploration with UCB, but is not detrimental with EI. Therefore, we recommend that FBBO using EI with an ARD kernel as the default choice for BO.
|
['Jonathan Fieldsend', 'Richard Everson', 'George De Ath']
|
2021-05-03
| null | null | null | null |
['bayesian-optimisation']
|
['methodology']
|
[ 9.98665616e-02 2.05250010e-01 4.91421252e-01 -1.18053630e-01
-8.34732473e-01 -3.59596074e-01 9.28119540e-01 2.32770741e-01
-6.23114109e-01 8.66622567e-01 1.53657317e-01 -4.91261005e-01
-7.88236618e-01 -6.99570954e-01 -5.14055669e-01 -1.27752101e+00
-6.47441298e-03 7.65236795e-01 4.10469830e-01 1.43009812e-01
2.21005827e-01 4.46395516e-01 -1.38122284e+00 -3.64377797e-01
6.97013378e-01 7.88830221e-01 1.35599356e-02 7.92410076e-01
1.57886863e-01 3.13370943e-01 -6.19625926e-01 -6.83477461e-01
4.21241760e-01 -2.58999705e-01 -5.16201138e-01 -3.92762363e-01
-1.49389029e-01 4.77378480e-02 1.67622805e-01 9.27820325e-01
7.64705360e-01 5.02308488e-01 9.87197995e-01 -9.07452881e-01
2.81078555e-02 3.76190037e-01 -5.23127437e-01 4.94983867e-02
1.68533117e-01 3.45907152e-01 7.12171376e-01 -4.98399585e-01
3.23928565e-01 1.60763574e+00 9.87537265e-01 8.41771066e-02
-2.04675007e+00 -2.28771329e-01 -2.89036274e-01 -2.70819664e-01
-1.60689163e+00 -4.62022692e-01 2.41419092e-01 -5.62164366e-01
8.61416996e-01 3.15274477e-01 4.12805080e-01 9.35051084e-01
5.21588981e-01 8.49292725e-02 1.21497262e+00 -4.48804855e-01
6.27445519e-01 1.29501536e-01 -7.74442032e-02 2.64417995e-02
5.02762914e-01 5.80666721e-01 -2.70032376e-01 -7.55956054e-01
6.57796383e-01 -4.62997973e-01 -2.57359266e-01 -3.36700290e-01
-1.04920840e+00 9.72692490e-01 -8.69597420e-02 -7.86759034e-02
-6.30736828e-01 4.13985044e-01 2.09263414e-01 1.31841809e-01
5.27951956e-01 7.83043504e-01 -5.68334937e-01 -5.02967894e-01
-9.10594583e-01 6.17316246e-01 1.07271647e+00 6.06622696e-01
6.19986773e-01 -8.39696527e-02 -4.36351448e-01 8.82063627e-01
7.39378572e-01 5.82866132e-01 -3.29031274e-02 -1.25843620e+00
1.29817054e-01 -6.77954555e-02 6.01730585e-01 -8.00818563e-01
-4.48079407e-01 -3.99470568e-01 -5.98836482e-01 3.96154225e-01
8.21487665e-01 -4.06581223e-01 -9.82167125e-01 1.41846752e+00
5.83951712e-01 1.25151843e-01 1.49832457e-01 5.62479377e-01
4.64333892e-01 7.23986149e-01 1.40017971e-01 -3.30334663e-01
1.32576561e+00 -4.83609676e-01 -6.88301325e-01 -1.86666414e-01
4.51683939e-01 -9.28574324e-01 7.51215160e-01 7.30962157e-01
-9.70735431e-01 -3.24867189e-01 -8.21857154e-01 4.42852855e-01
6.17565140e-02 -3.48243266e-01 3.49318594e-01 1.04954195e+00
-1.02224135e+00 8.89743805e-01 -1.12337041e+00 -1.95175484e-01
1.69019312e-01 3.56512040e-01 -1.24514461e-01 7.90575221e-02
-9.27913249e-01 1.03924668e+00 6.04905069e-01 3.70461017e-01
-7.63571620e-01 -7.59676456e-01 -7.85586834e-01 1.14199623e-01
4.68849123e-01 -6.88226163e-01 1.16253507e+00 -3.50028664e-01
-1.83733130e+00 3.85197550e-01 -1.46330819e-01 -5.44840455e-01
6.56939507e-01 -3.48437011e-01 -5.14517054e-02 -1.52645051e-01
-2.60952711e-01 5.94455004e-01 1.00690293e+00 -1.26781893e+00
-2.20279604e-01 6.02112990e-03 -2.15871751e-01 7.93315247e-02
4.80546355e-01 2.41390511e-01 -4.00747061e-01 -4.45503414e-01
2.41556376e-01 -1.12550950e+00 -5.32478929e-01 -4.55634058e-01
-2.55667180e-01 2.80188490e-02 6.90183863e-02 -7.41683066e-01
1.17533255e+00 -2.20866632e+00 5.93967512e-02 5.35820484e-01
-1.43421814e-01 -8.08546171e-02 2.87151009e-01 3.14181745e-01
2.92635616e-02 1.65834844e-01 -4.85297740e-01 -3.99856448e-01
2.16069743e-01 5.24992287e-01 -7.09991977e-02 4.94838566e-01
2.95536071e-01 5.63029528e-01 -8.83054733e-01 -4.22476530e-01
2.43352517e-01 6.36745811e-01 -6.24956846e-01 3.87863144e-02
3.59875560e-02 5.23321152e-01 -1.41680822e-01 2.74107456e-01
9.01092649e-01 9.11363065e-02 -3.33555602e-02 1.84454575e-01
-1.07675120e-01 1.52011186e-01 -1.65781021e+00 9.66779709e-01
-1.90116718e-01 4.37463641e-01 2.18670040e-01 -6.96124673e-01
1.12703490e+00 2.91632235e-01 2.21422821e-01 -1.99883372e-01
1.26756243e-02 3.08797300e-01 3.10503900e-01 -1.71635479e-01
2.63703734e-01 -5.56552529e-01 1.29761934e-01 7.00145662e-02
6.40770793e-02 -6.65870667e-01 7.24751204e-02 -1.77201435e-01
1.08513021e+00 5.30052543e-01 3.32305521e-01 -5.98073363e-01
2.26369455e-01 -2.54246533e-01 6.06317043e-01 1.15681005e+00
3.85760628e-02 9.00107861e-01 8.49571764e-01 -7.88977221e-02
-8.60529184e-01 -1.19453192e+00 -6.80470467e-01 6.69440031e-01
-2.82407343e-01 -3.97924721e-01 -6.20108128e-01 -2.93798059e-01
6.29658438e-03 1.07940578e+00 -4.58721787e-01 -1.86604500e-01
-3.02153945e-01 -1.64294958e+00 4.67309743e-01 2.21342385e-01
2.30783876e-02 -7.96711981e-01 -7.19560862e-01 3.79992485e-01
1.36454254e-01 -6.81393385e-01 1.30717456e-01 6.81116581e-01
-9.19230700e-01 -5.36773503e-01 -8.37599039e-01 2.50541329e-01
2.95778036e-01 -5.30781567e-01 9.95716572e-01 -2.54884541e-01
1.26663372e-01 3.23082626e-01 -2.66225547e-01 -5.95610380e-01
-6.76572204e-01 -3.50616604e-01 -1.04998320e-01 -1.56084418e-01
1.99053824e-01 -3.21193963e-01 -4.62775320e-01 5.32776535e-01
-7.26674795e-01 -3.80165726e-01 2.99239516e-01 1.05656385e+00
5.43929040e-01 2.08714187e-01 1.84968591e-01 -5.72445869e-01
5.02726853e-01 -5.15888870e-01 -9.20230031e-01 -1.02713127e-02
-6.57394409e-01 4.40721214e-01 -3.53553146e-03 -4.78466630e-01
-1.39620590e+00 -2.24695042e-01 -2.40261734e-01 -3.82875025e-01
-3.28955829e-01 5.82929492e-01 -7.75438547e-02 1.47633106e-01
8.12486172e-01 -4.52517986e-01 -1.18398637e-01 -6.41126931e-01
2.05898620e-02 4.17298496e-01 4.19038326e-01 -9.70866203e-01
3.85249376e-01 3.53356034e-01 4.89041746e-01 -7.63075829e-01
-6.38820410e-01 -4.47926879e-01 -4.30459917e-01 -3.19145247e-02
9.52410638e-01 -6.25965953e-01 -5.48275054e-01 4.75166798e-01
-1.01105642e+00 -4.89357680e-01 -5.53249061e-01 8.67329240e-01
-7.41005838e-01 3.61991107e-01 -2.17114300e-01 -1.36496878e+00
6.41587526e-02 -1.53512073e+00 1.11818361e+00 9.34378058e-02
-4.51482445e-01 -1.16445971e+00 2.39537328e-01 2.64408216e-02
3.43569130e-01 3.96623939e-01 7.07130611e-01 -8.01959276e-01
-3.07602026e-02 -1.81178778e-01 -4.67599519e-02 4.58486259e-01
-1.78790376e-01 2.91085333e-01 -1.26847768e+00 -2.20084518e-01
2.76730955e-01 1.69346929e-01 7.19550431e-01 8.78634393e-01
6.54222369e-01 -9.60872248e-02 -3.05223554e-01 5.94866693e-01
1.28393459e+00 1.71808258e-01 7.41427064e-01 5.99805593e-01
3.07026505e-01 7.38753438e-01 6.54472589e-01 5.56824803e-01
-1.23447813e-01 8.18181694e-01 3.09258759e-01 2.07645550e-01
3.25998694e-01 1.69653818e-03 3.62318158e-01 4.68597785e-02
-3.97125274e-01 -2.82432467e-01 -1.28783154e+00 3.23388040e-01
-1.75893569e+00 -8.86261225e-01 -5.45716286e-01 2.77452350e+00
8.32086265e-01 5.83308935e-01 1.92964450e-02 1.38427140e-02
5.82840085e-01 -1.78281963e-01 -9.61441845e-02 -6.36927485e-01
1.55084997e-01 3.31782401e-01 8.37450027e-01 6.92288160e-01
-1.03339374e+00 3.10815573e-01 6.83471298e+00 9.55071449e-01
-6.43043637e-01 2.24459201e-01 6.39127970e-01 -1.30094826e-01
1.63466334e-02 4.46281970e-01 -9.79635060e-01 7.47308373e-01
1.36742830e+00 2.96080709e-01 3.24871540e-01 5.13926744e-01
4.83861357e-01 -7.78562129e-01 -8.42475653e-01 6.70364857e-01
-5.32819569e-01 -7.89556384e-01 -5.40158391e-01 4.02670592e-01
7.19790101e-01 5.61918616e-02 -2.83912212e-01 2.64961839e-01
7.04336405e-01 -1.12278855e+00 8.63536775e-01 9.88277078e-01
2.74655938e-01 -8.36892307e-01 1.30213213e+00 3.86965036e-01
-5.90591431e-01 6.06228001e-02 -5.62557101e-01 -2.67360061e-02
4.40706313e-01 8.52274418e-01 -8.87387395e-01 6.55023217e-01
1.14959002e+00 -3.09828669e-02 -3.99360061e-01 1.53694940e+00
-1.38915777e-01 1.01698697e+00 -1.07358205e+00 2.42622644e-01
2.64097750e-01 -6.41398668e-01 9.25468862e-01 1.20841455e+00
4.85507101e-01 -2.64307439e-01 -2.67694354e-01 9.63127375e-01
8.03272009e-01 -1.17041394e-01 -2.52939165e-01 2.99951732e-01
4.99978215e-01 9.33823109e-01 -9.93689299e-01 -1.58047378e-02
-8.90678316e-02 4.05058533e-01 -1.23567730e-01 5.01960456e-01
-6.06171906e-01 -2.34747186e-01 4.60613787e-01 1.14280134e-02
5.74946404e-01 -1.14821978e-02 -3.78167838e-01 -4.06799912e-01
-2.75168687e-01 -7.84492552e-01 4.45522428e-01 -8.34083736e-01
-1.20985973e+00 2.38643333e-01 8.01475167e-01 -7.13245332e-01
-4.83398259e-01 -6.30899549e-01 -7.64156640e-01 1.38828683e+00
-9.93844926e-01 -6.54333174e-01 1.08848587e-01 -2.07990259e-01
-1.14386594e-02 3.59113544e-01 6.66421950e-01 -1.89111605e-01
-3.40969086e-01 2.20153362e-01 6.33494973e-01 -5.80912173e-01
6.99556053e-01 -1.53916669e+00 3.69408369e-01 5.91618657e-01
-1.83762580e-01 6.77662969e-01 1.45104170e+00 -7.18057036e-01
-7.85562813e-01 -5.70812583e-01 5.48494697e-01 -7.18287349e-01
6.68681145e-01 -2.19842479e-01 -1.17826009e+00 2.97816306e-01
-1.24846950e-01 -9.12270620e-02 4.33416605e-01 1.77754983e-01
2.99609125e-01 2.76798218e-01 -1.24657893e+00 3.97257060e-01
4.04822081e-01 7.21850544e-02 -5.95854878e-01 1.75273214e-02
7.03377008e-01 -2.79551923e-01 -1.35991395e+00 6.82962954e-01
4.12794530e-01 -8.06479096e-01 1.07355738e+00 -3.05319041e-01
-1.64605692e-01 -5.03308713e-01 -1.24352314e-01 -1.46540189e+00
-1.79283977e-01 -9.24587011e-01 5.18403351e-02 1.41208768e+00
4.78714347e-01 -8.59452844e-01 3.11217546e-01 7.52859354e-01
7.73272514e-02 -7.00546145e-01 -1.27399445e+00 -8.90171289e-01
1.26870602e-01 -9.45500612e-01 6.81320190e-01 4.27523255e-01
-5.29116690e-01 -1.32794054e-02 -1.90274402e-01 2.88784713e-01
6.21438146e-01 -5.44872940e-01 8.45053554e-01 -1.47245133e+00
-7.39743710e-01 -5.32712340e-01 -3.85140955e-01 -6.58965945e-01
-2.47028202e-01 -2.95488656e-01 4.16461200e-01 -1.29853117e+00
-1.16638407e-01 -5.14603376e-01 1.88541815e-01 1.87879503e-01
-4.52280402e-01 1.54738147e-02 -6.85021058e-02 2.23700777e-02
-6.28752112e-02 5.33711433e-01 8.38341951e-01 3.74237776e-01
-3.54287624e-01 4.33027536e-01 -2.63003796e-01 9.29485023e-01
5.31105936e-01 -7.96828687e-01 -1.47284344e-01 3.50961909e-02
5.55341899e-01 1.94762219e-02 6.82965815e-01 -7.87956953e-01
-4.43483517e-02 -1.16842255e-01 2.95894384e-01 -5.58422029e-01
4.24739301e-01 -7.12975979e-01 7.40698814e-01 2.08423629e-01
7.41086304e-02 -1.27085879e-01 5.13105035e-01 6.80735350e-01
3.51193473e-02 -1.04699385e+00 8.19006085e-01 8.63019675e-02
-2.33722571e-03 -4.54355657e-01 -5.44131637e-01 -1.13039896e-01
5.73512673e-01 -3.45517844e-01 6.63350001e-02 -4.01367456e-01
-1.12770987e+00 2.47758076e-01 5.45311451e-01 -2.00783852e-02
1.02369882e-01 -9.13700700e-01 -7.84535110e-01 1.33378536e-01
-1.71717465e-01 3.66235733e-01 9.06346142e-02 1.28579307e+00
-6.37713969e-01 1.70850962e-01 4.54750270e-01 -8.65521073e-01
-9.43343759e-01 4.13830340e-01 5.89287639e-01 -4.18878078e-01
-3.54017764e-01 8.63644600e-01 1.88117832e-01 -6.81531668e-01
1.81464106e-01 -3.76471817e-01 -3.85841094e-02 3.08595747e-02
2.45536178e-01 8.53648961e-01 1.52975112e-01 -4.66401398e-01
-6.05162792e-02 2.85183072e-01 3.81087244e-01 -4.91126418e-01
1.16891098e+00 -2.67761886e-01 -4.17669900e-02 5.96404493e-01
5.70269942e-01 6.44407719e-02 -1.60792232e+00 6.30085021e-02
3.73583019e-01 -4.13743585e-01 5.03103852e-01 -6.77250564e-01
-4.18825537e-01 7.45821238e-01 5.67265093e-01 2.13123232e-01
7.26248384e-01 -9.36985165e-02 -2.26538032e-01 2.57621914e-01
2.20028967e-01 -8.60293508e-01 -6.46637440e-01 4.15915489e-01
8.80368054e-01 -1.02468824e+00 3.87715995e-01 -2.05711246e-01
-6.90224767e-01 8.81444693e-01 1.96742684e-01 1.98800206e-01
9.25589383e-01 3.16851377e-01 -2.17935205e-01 -2.55718887e-01
-6.43401682e-01 -1.63019508e-01 4.18152303e-01 5.37249267e-01
8.10402036e-02 -4.11336571e-02 -2.07985803e-01 5.40576756e-01
-3.19204807e-01 -3.64863813e-01 2.45606214e-01 7.94957161e-01
-2.19883561e-01 -9.71458077e-01 -1.14540982e+00 5.90146780e-01
-6.38557732e-01 -2.72621155e-01 1.10097960e-01 8.54080379e-01
1.64275646e-01 9.87130761e-01 2.85915494e-01 2.24855989e-01
2.85643905e-01 3.41206104e-01 3.09874296e-01 -4.42315757e-01
-4.78514552e-01 7.48792350e-01 3.52939337e-01 -3.11466098e-01
-3.01397741e-01 -1.22433972e+00 -8.20946217e-01 -2.13264585e-01
-8.41430426e-01 2.81508148e-01 6.58224702e-01 9.92607594e-01
-2.30886303e-02 3.72555792e-01 1.25851527e-01 -1.07147598e+00
-8.96903932e-01 -1.26713467e+00 -5.25310636e-01 7.00049754e-03
7.89341927e-02 -9.92260337e-01 -8.01718712e-01 -2.43844748e-01]
|
[6.394711971282959, 3.7407796382904053]
|
78b83287-6d1f-4e37-9354-448267212c3b
|
rankpose-learning-generalised-feature-with
|
2005.10984
| null |
https://arxiv.org/abs/2005.10984v1
|
https://arxiv.org/pdf/2005.10984v1.pdf
|
RankPose: Learning Generalised Feature with Rank Supervision for Head Pose Estimation
|
We address the challenging problem of RGB image-based head pose estimation. We first reformulate head pose representation learning to constrain it to a bounded space. Head pose represented as vector projection or vector angles shows helpful to improving performance. Further, a ranking loss combined with MSE regression loss is proposed. The ranking loss supervises a neural network with paired samples of the same person and penalises incorrect ordering of pose prediction. Analysis on this new loss function suggests it contributes to a better local feature extractor, where features are generalised to Abstract Landmarks which are pose-related features instead of pose-irrelevant information such as identity, age, and lighting. Extensive experiments show that our method significantly outperforms the current state-of-the-art schemes on public datasets: AFLW2000 and BIWI. Our model achieves significant improvements over previous SOTA MAE on AFLW2000 and BIWI from 4.50 to 3.66 and from 4.0 to 3.71 respectively. Source code will be made available at: https://github.com/seathiefwang/RankHeadPose.
|
['Donggen Dai', 'Zhuojun Chen', 'Wangkit Wong']
|
2020-05-22
| null | null | null | null |
['head-pose-estimation']
|
['computer-vision']
|
[-2.77806491e-01 3.37415546e-01 -1.79175824e-01 -1.00032449e+00
-1.28502309e+00 -2.38880560e-01 5.18911421e-01 -2.12655097e-01
-6.96334720e-01 7.91719973e-01 8.07538092e-01 1.99103385e-01
-1.36387050e-02 -2.94890851e-01 -7.65858769e-01 -6.90730214e-01
-3.55399370e-01 4.85194355e-01 -1.08652472e-01 -1.72512367e-01
4.55610752e-02 5.60568631e-01 -1.56388390e+00 -1.23982087e-01
3.05145085e-01 9.89470780e-01 -1.43615350e-01 3.76275122e-01
6.92570806e-01 2.34120816e-01 -3.99069995e-01 -3.55179161e-01
4.01779205e-01 4.96166646e-02 -6.50749624e-01 -1.41576543e-01
1.07208228e+00 -3.81848663e-01 -4.04952705e-01 7.23202765e-01
1.22675025e+00 1.54794723e-01 5.36893785e-01 -1.44767439e+00
-3.06022793e-01 2.69524425e-01 -7.00786471e-01 -6.21405616e-02
8.92455637e-01 -7.28807375e-02 1.00280297e+00 -1.24303222e+00
4.62943047e-01 1.45357049e+00 9.63846087e-01 6.15666509e-01
-9.52846229e-01 -9.78836477e-01 2.62358159e-01 4.84946102e-01
-1.86684668e+00 -8.11319292e-01 7.69961774e-01 -6.85729012e-02
6.51946902e-01 5.11996984e-01 6.55282199e-01 1.01330388e+00
-1.54334307e-01 1.14059269e+00 1.01415670e+00 -3.30995113e-01
-2.70970929e-02 -2.07412750e-01 1.45365313e-01 8.97303224e-01
2.66307145e-01 -2.26868577e-02 -1.09412563e+00 -2.65742868e-01
2.76092470e-01 -1.77597761e-01 -6.08084500e-01 -5.34770012e-01
-9.04038072e-01 8.45519841e-01 7.46366858e-01 -2.49488860e-01
-2.19358653e-01 4.78144661e-02 1.72090188e-01 -9.12658945e-02
4.71419990e-01 1.22321978e-01 -5.20671308e-01 3.53233702e-02
-1.05608940e+00 5.71384311e-01 6.81686580e-01 9.99885440e-01
6.13447607e-01 -3.08786929e-01 -1.46000281e-01 9.54157948e-01
8.59233737e-01 7.84489036e-01 3.63973677e-01 -8.76109302e-01
4.95152473e-01 2.86008805e-01 -6.44182712e-02 -6.50388539e-01
-9.77179408e-01 -1.95263237e-01 -4.07153636e-01 2.38853483e-03
3.39056760e-01 -1.08554877e-01 -1.08847892e+00 1.85890424e+00
5.37598133e-01 -2.55215075e-02 -4.74589169e-01 9.81129229e-01
1.21627808e+00 1.82708099e-01 -4.25291844e-02 4.29423824e-02
1.52325761e+00 -7.89500415e-01 -7.21964777e-01 -5.33882678e-01
3.77270609e-01 -6.69687390e-01 9.25622106e-01 3.85090500e-01
-1.21147895e+00 -1.66011453e-01 -8.55146229e-01 -3.29203814e-01
-3.22479934e-01 1.70807093e-01 7.33541787e-01 8.23760808e-01
-1.40870082e+00 1.51675746e-01 -8.48741829e-01 -4.38957691e-01
5.05041182e-01 8.93410385e-01 -6.52033687e-01 -8.14684704e-02
-9.49195802e-01 8.47095370e-01 -3.95920780e-03 3.90647262e-01
-3.43898058e-01 -6.79853261e-01 -1.33624816e+00 -4.45388526e-01
9.46725532e-02 -4.93494838e-01 1.34626985e+00 -2.11898252e-01
-1.54613137e+00 1.08601654e+00 -5.84683001e-01 -1.71549961e-01
7.45234787e-01 -6.33734405e-01 -8.62401351e-02 -6.46994114e-02
2.22322494e-01 8.68752778e-01 4.83283937e-01 -1.15254486e+00
-5.82489967e-01 -8.81935716e-01 -2.74659723e-01 3.80649656e-01
-2.53954887e-01 1.87207773e-01 -9.97606814e-01 -4.41395134e-01
4.98187602e-01 -1.20743358e+00 7.78082237e-02 2.08503112e-01
-7.18109965e-01 -4.84182358e-01 2.80922592e-01 -8.87220323e-01
1.09348714e+00 -1.80702007e+00 -3.12716216e-02 3.64903420e-01
5.57333753e-02 -1.43985108e-01 4.68734913e-02 3.12343333e-02
-1.61594391e-01 -3.39314938e-01 2.28189193e-02 -8.90230238e-01
1.45045266e-01 -4.56819236e-02 2.48047486e-01 1.27368295e+00
-2.13645130e-01 8.24848294e-01 -5.95536768e-01 -4.87383842e-01
6.32357150e-02 8.11379910e-01 -7.04270184e-01 1.31695136e-01
5.30113995e-01 3.50829452e-01 -1.19775072e-01 9.64259148e-01
1.03252769e+00 2.00223923e-01 -1.10452719e-01 -4.08071309e-01
4.11490910e-02 3.89197350e-01 -1.18413544e+00 1.80972838e+00
-1.36599824e-01 5.32431304e-01 1.73338071e-01 -4.07934517e-01
7.04327464e-01 1.16516419e-01 4.18523103e-01 -6.57937229e-01
2.75237143e-01 -2.91606672e-02 -2.88011342e-01 -2.52384245e-01
3.59427959e-01 1.44306257e-01 -1.78205013e-01 1.30109891e-01
4.94469143e-02 -1.68029573e-02 -1.54871702e-01 -5.76458946e-02
6.22340858e-01 3.22464317e-01 2.08650723e-01 -3.36867094e-01
5.72950065e-01 -6.77415550e-01 6.68133199e-01 5.03230453e-01
-4.78350490e-01 1.10814786e+00 -7.71347014e-03 -3.05511564e-01
-5.80030441e-01 -1.16178787e+00 -4.66403335e-01 1.52517629e+00
-9.54334363e-02 -4.83451843e-01 -6.96243227e-01 -5.40857732e-01
2.62275875e-01 5.61042249e-01 -9.18421090e-01 -4.63219658e-02
-8.50927413e-01 -8.48005176e-01 5.30746520e-01 7.79889286e-01
3.12682688e-01 -7.25132287e-01 -2.99286783e-01 -2.83598274e-01
-2.59052098e-01 -8.18372071e-01 -7.39248574e-01 1.27653003e-01
-5.88133812e-01 -8.60484421e-01 -9.66865659e-01 -7.31395423e-01
8.30830157e-01 -3.23431306e-02 9.79506493e-01 1.39994109e-02
-2.92802334e-01 4.07666832e-01 -1.51295409e-01 -5.98433077e-01
6.01766706e-01 1.92095891e-01 4.16110992e-01 -1.50355905e-01
6.27315402e-01 -2.60665566e-01 -9.99467075e-01 2.34754398e-01
-2.50907719e-01 -3.26001018e-01 2.50783056e-01 7.24327683e-01
4.36677247e-01 -9.00029719e-01 6.15458079e-02 -5.78143120e-01
4.29314882e-01 -2.95265615e-01 -3.04221392e-01 2.27771297e-01
-4.93454605e-01 1.05779067e-01 -1.45656347e-01 -1.57253653e-01
-8.40743899e-01 5.01828969e-01 -3.65827143e-01 3.74878980e-02
-1.63976982e-01 4.54021506e-02 -4.32912648e-01 -1.28402263e-01
4.33410674e-01 8.36392120e-02 -2.20382698e-02 -5.95218360e-01
3.43805730e-01 7.11993635e-01 7.77334750e-01 -4.60583270e-01
7.39760280e-01 6.18824661e-01 5.34361787e-02 -6.98464096e-01
-7.69981205e-01 -7.80242860e-01 -9.16824937e-01 -1.44768387e-01
4.94727165e-01 -1.18519211e+00 -9.00209546e-01 4.89299297e-01
-8.76729786e-01 -1.16285272e-01 1.78441599e-01 4.87805158e-01
-5.26470304e-01 2.10420340e-01 -5.03056049e-01 -8.88026357e-01
-5.56679547e-01 -1.01547158e+00 1.58699429e+00 2.05363438e-01
-4.94246453e-01 -6.60275996e-01 7.48833343e-02 4.73133832e-01
1.25971898e-01 2.99154818e-01 7.79910609e-02 -4.63122517e-01
-1.62383959e-01 -5.55099189e-01 3.61753511e-03 -2.31167927e-01
-5.22906519e-02 -3.44629019e-01 -1.35826647e+00 -7.30605662e-01
-3.53467196e-01 -3.86763334e-01 8.68262947e-01 6.01025641e-01
1.10908997e+00 -2.46421888e-01 -4.36540335e-01 1.09683120e+00
9.35987532e-01 -7.21007213e-02 5.27044833e-01 6.19483292e-01
7.13711798e-01 6.45138204e-01 4.14814740e-01 7.19843566e-01
9.63743150e-01 9.16549563e-01 3.33131462e-01 -1.46431252e-01
-2.22683221e-01 -2.51327187e-01 3.37707072e-01 4.84069884e-01
-3.13705862e-01 8.91206339e-02 -9.31377351e-01 3.82433146e-01
-1.74874115e+00 -5.65300584e-01 1.62503943e-01 2.41916323e+00
7.63783336e-01 -1.75170794e-01 5.15525818e-01 1.44581661e-01
4.84865606e-01 1.88914776e-01 -4.51667398e-01 -1.21136837e-01
-7.13591725e-02 2.59119004e-01 7.31258094e-01 7.33435690e-01
-1.28382874e+00 8.83832395e-01 6.17805195e+00 4.10352379e-01
-1.09978700e+00 2.37924173e-01 5.75665176e-01 -6.12546623e-01
1.77941173e-02 -6.20348394e-01 -1.16772115e+00 3.55263323e-01
6.20682120e-01 7.50765279e-02 3.28388959e-01 8.36918294e-01
1.38710991e-01 -9.68971625e-02 -1.19195724e+00 1.32767189e+00
5.54882646e-01 -5.33601403e-01 -3.38854074e-01 2.10190535e-01
2.90287018e-01 2.60521203e-01 3.03294480e-01 3.17174882e-01
-6.55444935e-02 -1.20877051e+00 1.15014172e+00 4.66270924e-01
7.96021342e-01 -8.74830365e-01 7.96401083e-01 -1.15032256e-01
-1.36011565e+00 8.14027116e-02 -2.25672692e-01 5.88664003e-02
1.28477827e-01 -1.50032444e-02 -9.80799019e-01 1.92441225e-01
1.04060256e+00 4.44999069e-01 -8.46017182e-01 1.31362689e+00
-5.75261772e-01 3.76919657e-01 -7.34479487e-01 9.29505099e-03
-9.75800604e-02 3.05259258e-01 3.36004704e-01 1.27612746e+00
2.10714102e-01 1.47729263e-01 6.29019812e-02 1.94112256e-01
-1.93824731e-02 2.08900079e-01 -3.88899297e-01 9.59139287e-01
5.76753139e-01 1.12604880e+00 -3.11599404e-01 1.13649555e-01
-4.09431279e-01 1.09874475e+00 4.95618582e-01 3.18950087e-01
-7.54284739e-01 -2.55085260e-01 8.56092453e-01 1.81188539e-01
2.54544765e-01 -4.39829752e-02 -2.50339210e-01 -9.32149470e-01
9.34116021e-02 -5.20549178e-01 5.36625803e-01 -5.92970788e-01
-1.06040156e+00 6.45428658e-01 7.35074803e-02 -9.59322095e-01
-3.69759858e-01 -7.47293472e-01 -4.94363517e-01 8.48724484e-01
-1.41131127e+00 -1.53488243e+00 -3.60320419e-01 7.11312950e-01
3.19205582e-01 3.84569168e-02 7.73855627e-01 4.49947506e-01
-6.52037203e-01 1.45159042e+00 -1.32162064e-01 2.70621330e-01
9.96652722e-01 -1.35029066e+00 3.41806918e-01 6.20668769e-01
-1.86587051e-01 8.11382830e-01 7.40306199e-01 -4.61644232e-01
-1.47842133e+00 -8.14370513e-01 1.12394679e+00 -7.53426850e-01
1.89390972e-01 -6.30388081e-01 -5.02223611e-01 7.99024522e-01
-2.87954547e-02 2.97566622e-01 7.31240928e-01 4.45873052e-01
-3.54756624e-01 -2.40222618e-01 -1.35041618e+00 4.53530043e-01
1.27879786e+00 -4.53662783e-01 -4.97292936e-01 3.43111575e-01
1.59213036e-01 -7.19986558e-01 -6.93349302e-01 4.60923135e-01
8.96537304e-01 -7.76781857e-01 1.32190096e+00 -4.36521441e-01
-2.22825542e-01 -1.57467216e-01 -2.98094213e-01 -1.15841985e+00
-5.07115901e-01 -6.35723650e-01 -1.68736249e-01 1.04430318e+00
5.02807677e-01 -6.02503657e-01 1.11636758e+00 1.12396753e+00
3.59330103e-02 -9.66278017e-01 -1.31853759e+00 -6.32033646e-01
1.05008587e-01 -4.28062677e-01 7.44063437e-01 5.92057884e-01
-7.07349600e-03 9.73503962e-02 -4.82987195e-01 3.87877792e-01
9.81598318e-01 -2.12079793e-01 7.97811627e-01 -1.25410938e+00
1.84752062e-01 -4.36286747e-01 -6.34527326e-01 -8.68695736e-01
2.72271633e-01 -8.48732531e-01 2.10594103e-01 -1.45590878e+00
3.46823692e-01 -1.13155857e-01 -2.98204154e-01 8.53015482e-01
-3.31827372e-01 6.68836296e-01 2.21173763e-01 -6.15694784e-02
-5.72401166e-01 5.65475464e-01 8.00363839e-01 -3.67929116e-02
-1.31027609e-01 1.68822750e-01 -6.20821834e-01 9.33842063e-01
8.87897611e-01 -2.97470838e-01 -5.75026013e-02 -3.77852499e-01
-6.01477847e-02 -2.98313886e-01 1.60257012e-01 -9.64248776e-01
4.01198804e-01 2.14761764e-01 8.09109330e-01 -7.76468217e-01
8.43114555e-01 -5.91132104e-01 -2.02900976e-01 3.77750665e-01
-1.09113336e-01 1.39850512e-01 3.85954790e-02 1.10999808e-01
1.04446374e-01 9.06860679e-02 6.57203853e-01 9.91887525e-02
-6.64528787e-01 4.64225590e-01 1.87118292e-01 1.48186702e-02
6.38780296e-01 -2.29956523e-01 -1.34601388e-02 -6.86950266e-01
-7.10503995e-01 4.03097004e-01 2.59702384e-01 4.34139311e-01
6.57015026e-01 -1.53777397e+00 -8.27529907e-01 4.45309132e-01
3.61515909e-01 -7.06906244e-02 -2.65871078e-01 7.96867907e-01
-3.61072540e-01 5.06454527e-01 -9.52325687e-02 -5.59794247e-01
-1.77754486e+00 -1.02837898e-01 2.20879525e-01 3.10599595e-01
-4.30755407e-01 1.46442759e+00 3.86720859e-02 -8.27810705e-01
8.72649848e-01 -2.49470621e-01 -2.55572408e-01 3.15459400e-01
7.61661410e-01 4.94279623e-01 1.36884823e-01 -1.26181495e+00
-9.94906962e-01 7.42792130e-01 -1.27398208e-01 -1.40994087e-01
1.68157375e+00 -4.02650625e-01 -1.00680970e-01 2.26510260e-02
1.52923644e+00 1.31506234e-01 -1.29350877e+00 -3.10775757e-01
-1.56048620e-02 -4.75224078e-01 1.66812152e-01 -8.24526250e-01
-1.10486495e+00 5.03308117e-01 1.15450931e+00 -5.62461257e-01
1.05726898e+00 3.58655661e-01 6.49617314e-01 3.12227577e-01
6.11880004e-01 -1.01866603e+00 -1.61490843e-01 5.84178030e-01
1.14412856e+00 -1.54959130e+00 4.71652776e-01 -3.06759089e-01
-4.28391546e-01 8.56566131e-01 6.02528989e-01 5.80432154e-02
7.65386164e-01 2.26288080e-01 1.82084456e-01 -1.55409411e-01
-4.70578447e-02 -3.30726296e-01 8.05609167e-01 6.16853595e-01
8.37632179e-01 3.02611977e-01 -2.54533738e-01 6.87350690e-01
-9.54646289e-01 -1.74539551e-01 -6.55054078e-02 8.81971657e-01
-2.37876579e-01 -1.01560235e+00 -7.61860728e-01 3.16308200e-01
-5.00605583e-01 -3.06626167e-02 -2.12234944e-01 8.15373838e-01
1.40583679e-01 7.14376986e-01 -3.31202671e-02 -2.61364609e-01
4.70839918e-01 5.98765835e-02 6.73180461e-01 -4.43215787e-01
-2.93148041e-01 1.51366711e-01 1.07913934e-01 -9.30362046e-01
-3.01922083e-01 -9.80667233e-01 -1.10542369e+00 -2.65287578e-01
-2.84748584e-01 -9.73882154e-02 7.17700005e-01 6.02123797e-01
1.06929950e-01 3.59000228e-02 4.98987645e-01 -1.43420243e+00
-4.15782034e-01 -1.05034459e+00 -6.01001441e-01 4.77489054e-01
6.40527070e-01 -8.40621412e-01 -3.39891344e-01 -1.95712581e-01]
|
[13.662213325500488, 0.29411616921424866]
|
def9b274-8354-4acf-9379-574b6624cc08
|
simulation-toolkit-for-digital-material
| null | null |
https://www.sciencedirect.com/science/article/abs/pii/S0927025623000150
|
https://www.sciencedirect.com/science/article/pii/S0927025623000150/pdfft?isDTMRedir=true&download=true
|
Simulation toolkit for digital material characterization of large image-based microstructures
|
In this paper, an efficient image-based simulation toolkit for material characterization is presented, which is scalable to work from personal computers to workstations. The effective thermal conductivity, elasticity, and permeability are evaluated employing a computational homogenization framework based on the Finite Element Method (FEM). Two complementary open-source packages are presented: one developed in Python, which can convert digital images into voxel meshes (pyTomoviewer); the other developed in Julia, that can run numerical simulations to compute effective material properties (chpack). Also, a CUDA C version of chpack is provided (chfem_gpu). They were designed to deal with large multi-phase models, so strategies were devised to minimize their memory footprint, while avoiding a high toll on execution time. The voxel-based approach significantly simplifies the FEM meshes and allows efficient matrix-free implementations. In that sense, to handle large linear systems of equations, the element-by-element (EBE) technique is adopted, in conjunction with a low-memory implementation of the Preconditioned Conjugate Gradient (PCG) method. The code was thoroughly tested on an artificial geometry made of a square array of cylinders, for which analytical solutions exist, as well as on a real micro-tomographic reconstruction of FiberFormTM, a carbon preform commonly used in thermal protection systems.
|
['André M.B. Pereira', 'Ricardo Leiderman', 'Federico Semeraro', 'Victor W. Sapucaia', 'Rafael S. Vianna', 'Pedro C.F. Lopes']
|
2023-01-18
| null | null | null |
computational-materials-science-2023-1
|
['physical-simulations']
|
['miscellaneous']
|
[ 1.38709811e-03 -3.56475621e-01 6.19669199e-01 3.05274665e-01
-3.18294346e-01 -1.07723989e-01 3.73267949e-01 6.20740578e-02
-5.23278236e-01 7.37502038e-01 -3.26972425e-01 -4.35071200e-01
-2.30006069e-01 -1.17761183e+00 -4.12907451e-01 -1.09067667e+00
4.25573774e-02 6.42653942e-01 3.98922235e-01 -8.17824900e-02
3.96860033e-01 7.22428739e-01 -1.64439738e+00 2.28102338e-02
9.88952875e-01 1.04576147e+00 3.23263943e-01 4.93111581e-01
1.84033543e-01 6.25056207e-01 1.03917532e-01 -1.19072236e-02
-2.78905593e-02 -5.37325554e-02 -1.04334855e+00 2.48548314e-01
-1.27075762e-01 -3.51058781e-01 2.95453221e-01 5.46350181e-01
3.89466166e-01 3.04581344e-01 9.55857813e-01 -6.41974747e-01
1.38734385e-01 6.35697842e-02 -3.34002078e-01 -2.91031480e-01
4.82793808e-01 2.81279355e-01 2.55435556e-01 -1.09617865e+00
6.36991739e-01 8.28480363e-01 9.76548433e-01 2.52968017e-02
-1.43651056e+00 -1.02980465e-01 -6.71196342e-01 1.21423453e-01
-1.42326581e+00 6.19842820e-02 6.60901308e-01 -7.93653905e-01
1.20127428e+00 7.93515563e-01 1.04177999e+00 5.79155982e-01
6.95263922e-01 -6.19894750e-02 1.51780498e+00 -5.15429497e-01
7.30000556e-01 1.23466879e-01 5.70935607e-02 4.36889976e-01
2.72887230e-01 8.91569331e-02 1.13860123e-01 -7.12412000e-01
9.60173786e-01 -4.15260106e-01 -1.86764002e-01 -4.04790848e-01
-9.21207488e-01 5.70273459e-01 5.98390959e-02 6.38768792e-01
-7.25736856e-01 2.15167087e-02 4.71388489e-01 -1.40601918e-01
6.96971893e-01 3.15019011e-01 -2.52847493e-01 -8.86877328e-02
-1.03071272e+00 5.72903693e-01 1.12867939e+00 3.01732242e-01
9.00970340e-01 9.23151895e-02 2.87726372e-01 5.96033931e-01
6.62961006e-01 6.57401502e-01 2.69804895e-01 -9.55607235e-01
-1.56296760e-01 3.75427455e-01 5.72605953e-02 -9.99140620e-01
-3.28725249e-01 -1.09269112e-01 -7.74616301e-01 5.59272230e-01
9.67456475e-02 -6.65421635e-02 -3.81723017e-01 7.69634724e-01
9.15402532e-01 1.11370990e-02 -1.33917630e-01 9.60463881e-01
5.86698890e-01 7.28062451e-01 1.28455460e-01 -4.14391488e-01
1.30375123e+00 -6.93332195e-01 -4.92988348e-01 2.50073910e-01
8.39878142e-01 -1.17969060e+00 7.42381573e-01 5.37907064e-01
-1.23400104e+00 -3.31588805e-01 -7.98180997e-01 2.19711944e-01
-3.23997736e-01 -1.07576191e-01 3.69247615e-01 7.08188057e-01
-1.09698391e+00 1.05325401e+00 -9.65982139e-01 -1.39439493e-01
-1.17997453e-01 1.61543638e-01 -2.03633457e-01 1.97218046e-01
-9.50706482e-01 8.41949999e-01 1.85301289e-01 1.57763079e-01
-4.05884147e-01 -7.89949536e-01 -6.27290368e-01 -2.28043452e-01
-6.20063730e-02 -1.09000623e+00 8.38584065e-01 -6.71908319e-01
-1.92516160e+00 8.27477157e-01 1.89278573e-01 -3.21441777e-02
8.46696079e-01 5.74318133e-02 -1.34162813e-01 3.98957789e-01
-4.28152606e-02 -1.60261482e-01 8.37036550e-01 -1.35860443e+00
3.35907221e-01 2.99735330e-02 -3.92337769e-01 1.17707394e-01
-1.52469143e-01 9.51270312e-02 -6.74349070e-02 -5.07825494e-01
1.26812682e-01 -8.67370903e-01 -6.62817061e-01 -9.13149714e-02
-3.59848410e-01 1.73018157e-01 8.62328410e-01 -9.67590749e-01
1.18120933e+00 -1.87721944e+00 1.70948282e-01 7.51692951e-01
-1.08716756e-01 1.56626120e-01 5.86553872e-01 1.05766404e+00
-1.67910904e-01 -3.08802664e-01 -8.30985665e-01 -2.55596250e-01
-2.08286688e-01 1.33423522e-01 1.21689908e-01 8.59005034e-01
-3.77732515e-01 3.04082572e-01 -5.40578306e-01 -5.91770589e-01
6.23811424e-01 7.00508058e-01 -6.14960909e-01 5.27273715e-02
-8.48848373e-02 5.30601799e-01 -4.74485576e-01 2.65698671e-01
1.33655560e+00 4.59244996e-02 1.74293965e-01 -3.74280840e-01
-9.13536847e-01 -1.53948396e-01 -1.48239899e+00 1.33696365e+00
-5.93020260e-01 -2.47246951e-01 1.00404716e+00 -8.97525609e-01
8.10197949e-01 5.97538412e-01 7.77015388e-01 -4.19503659e-01
3.25062990e-01 5.49518406e-01 -3.94631803e-01 -6.47811770e-01
5.36483169e-01 -2.29924619e-01 3.99551988e-01 4.73515183e-01
-5.34807801e-01 -7.94005930e-01 2.21472606e-01 1.52226567e-01
8.57326746e-01 4.12921190e-01 4.82060499e-02 -1.10778403e+00
9.57003832e-01 3.66903514e-01 1.28954828e-01 -2.02171691e-02
4.95305836e-01 3.49822223e-01 2.11502582e-01 -3.85002494e-01
-1.36196160e+00 -7.72907495e-01 -7.50883877e-01 3.15355182e-01
4.64088842e-02 -4.51172918e-01 -1.18601990e+00 4.84954059e-01
-1.07421026e-01 5.33116043e-01 -2.60030657e-01 3.37898761e-01
-5.49031734e-01 -9.29745317e-01 -1.11685753e-01 8.84209275e-02
4.35724407e-01 -1.02156711e+00 -9.45829868e-01 5.13008237e-01
2.19041467e-01 -9.33391571e-01 2.60189474e-01 -2.15662375e-01
-1.16854894e+00 -1.21884239e+00 -7.39060760e-01 -5.08496940e-01
6.93034887e-01 -1.39186323e-01 1.12699950e+00 4.53386217e-01
-4.20741200e-01 7.82530546e-01 -2.61889696e-01 2.21789300e-01
-7.55737841e-01 -3.67520988e-01 1.11750849e-01 -1.08961083e-01
-3.54096860e-01 -6.91205502e-01 -6.47690296e-01 4.40847754e-01
-1.13171422e+00 3.70257020e-01 2.19518185e-01 6.32772982e-01
8.13846707e-01 2.95240015e-01 -1.96609292e-02 -7.33574748e-01
5.70310831e-01 -3.93345535e-01 -7.83108950e-01 -1.37343422e-01
-6.23525262e-01 -4.82946336e-01 9.32852447e-01 -2.45837290e-02
-1.19257843e+00 -1.66097600e-02 -7.81985521e-01 -1.79675743e-01
-2.18270883e-01 6.29071534e-01 -2.63346191e-02 -5.79597950e-01
2.96602368e-01 2.16490999e-01 2.28948459e-01 -6.83787525e-01
-9.68571752e-02 5.11614442e-01 3.66634965e-01 -1.04456294e+00
6.57709181e-01 5.55926442e-01 3.94799113e-01 -1.42902815e+00
2.95880079e-01 -4.00733411e-01 -4.74174589e-01 -6.76525593e-01
8.38653982e-01 -4.73987967e-01 -6.93797350e-01 1.02556396e+00
-8.71831656e-01 -6.35898888e-01 -2.58720487e-01 6.01448178e-01
-7.36594141e-01 8.34370434e-01 -8.72285903e-01 -6.20741606e-01
-6.70617878e-01 -1.37172151e+00 7.87542045e-01 7.55429119e-02
-2.15594396e-01 -1.23021340e+00 4.60737407e-01 3.61063153e-01
8.35165679e-01 6.04300678e-01 8.65693390e-01 3.01090807e-01
-3.06814730e-01 -5.28219827e-02 4.36946712e-02 4.19121236e-01
-3.27260584e-01 5.18307686e-01 -8.41907978e-01 -6.18858993e-01
4.79254365e-01 1.83214247e-02 2.35292912e-01 4.69835758e-01
1.06504989e+00 -2.35190928e-01 -5.10800421e-01 5.18988609e-01
1.97659695e+00 -1.55023381e-01 8.72345626e-01 3.24880421e-01
6.63113654e-01 4.39979672e-01 4.96593356e-01 7.13023007e-01
4.20479178e-02 7.20839977e-01 4.81741160e-01 -3.52150619e-01
1.67114794e-01 4.35337692e-01 -3.93948369e-02 1.12837243e+00
-6.65648401e-01 2.34014943e-01 -1.02411067e+00 1.19997174e-01
-1.48464763e+00 -5.84490418e-01 -1.10809755e+00 2.24962854e+00
5.43678641e-01 -2.04747558e-01 -3.87904756e-02 4.28862482e-01
4.49935257e-01 -4.03935760e-01 7.07805902e-02 -8.24851692e-01
1.36124983e-01 4.63315189e-01 4.73181754e-01 5.83677173e-01
-9.30330396e-01 3.97260875e-01 6.14217663e+00 9.96728301e-01
-1.27874374e+00 2.55730212e-01 3.70705843e-01 5.14068663e-01
-3.40612173e-01 2.52248943e-01 -1.77653328e-01 7.35559762e-01
1.14031982e+00 5.02758622e-02 3.36533666e-01 7.67513216e-01
4.26202893e-01 -8.79648805e-01 -3.90178740e-01 5.59937835e-01
-4.53807414e-01 -1.30210030e+00 -5.32732606e-01 3.05072099e-01
5.05025864e-01 -8.50601047e-02 -3.30707252e-01 -3.24496239e-01
-3.59886825e-01 -4.78508919e-01 6.43237889e-01 5.94435215e-01
8.15947473e-01 -7.24968433e-01 7.21289933e-01 4.15760070e-01
-1.23427951e+00 3.63480687e-01 -3.18413347e-01 -2.54532129e-01
5.62942147e-01 1.24025118e+00 -4.65615690e-01 9.18587267e-01
9.26321328e-01 2.80757129e-01 -1.51925996e-01 1.12057459e+00
2.78610498e-01 5.92032254e-01 -6.01184785e-01 1.13133088e-01
1.55405521e-01 -1.08478510e+00 7.30444074e-01 1.02813339e+00
4.30410087e-01 1.10713365e-02 -1.18139658e-04 9.73690093e-01
7.80785263e-01 5.57990730e-01 -3.73737127e-01 4.56485242e-01
1.06807716e-01 1.54113424e+00 -1.16684079e+00 -2.83746958e-01
-2.43203402e-01 6.56533957e-01 -2.68745869e-01 1.88407093e-01
-7.53010452e-01 -1.42465696e-01 2.98802167e-01 7.68315375e-01
1.62649646e-01 -4.59698737e-01 -2.90723503e-01 -6.79505408e-01
-2.31996074e-01 -5.06960273e-01 -3.08619142e-02 -8.15501332e-01
-1.32391512e+00 5.30717552e-01 4.10465389e-01 -1.11030447e+00
-1.68690741e-01 -7.68026471e-01 -7.52702057e-01 1.11741006e+00
-8.72932911e-01 -1.08356571e+00 -2.35136002e-01 4.71302003e-01
-1.50056824e-01 4.17674392e-01 1.00906980e+00 5.25591612e-01
-5.85779667e-01 -2.74155706e-01 5.13218999e-01 -3.22195381e-01
2.46534199e-01 -9.02406991e-01 2.59715691e-03 6.46229208e-01
-1.03500974e+00 4.12205815e-01 1.08075416e+00 -8.47296596e-01
-1.70080972e+00 -6.35904670e-01 5.24602711e-01 3.37354243e-01
7.21458435e-01 -8.32766965e-02 -1.27168143e+00 9.88954455e-02
3.53224993e-01 6.27859822e-03 5.30823946e-01 -4.81364965e-01
4.00223732e-01 3.62204701e-01 -1.51266694e+00 2.58612305e-01
2.16491178e-01 -3.21139693e-01 -1.17983513e-01 4.34549123e-01
-5.51017327e-03 -5.68892241e-01 -1.58569157e+00 4.32874084e-01
4.04359162e-01 -1.38717318e+00 1.08927274e+00 4.43988025e-01
3.11336875e-01 -3.67790490e-01 1.37161613e-01 -1.00608885e+00
-3.64433557e-01 -5.38049996e-01 6.04203381e-02 1.13454866e+00
-1.16554007e-01 -1.02795744e+00 5.28407454e-01 8.92562151e-01
-4.92290705e-01 -1.06351101e+00 -1.19810498e+00 -5.63231647e-01
2.45405436e-01 -3.15251350e-01 3.51098806e-01 8.18935871e-01
-6.42809868e-02 -2.44200990e-01 7.62710273e-02 -8.75440836e-02
6.66674256e-01 8.03488865e-02 3.63248825e-01 -1.24562836e+00
-3.50853562e-01 -2.40415171e-01 -2.44334936e-01 -2.14806184e-01
-3.35485628e-03 -5.15132844e-01 -1.12834774e-01 -1.40580392e+00
9.10547283e-03 -7.57342875e-01 5.37202954e-01 -1.37737142e-02
2.76143402e-01 3.29609662e-01 -2.67635763e-01 2.52278447e-01
2.32021958e-01 3.39130312e-01 1.26906335e+00 1.73210248e-01
2.08490714e-02 -1.23174913e-01 2.65227228e-01 7.74242997e-01
5.59936225e-01 -3.45477670e-01 8.63621105e-03 5.81612438e-02
3.07847470e-01 8.22505429e-02 6.06927395e-01 -1.37023425e+00
3.63295525e-02 6.51776120e-02 1.28695354e-01 -6.75217628e-01
3.74277711e-01 -1.07743573e+00 8.92247081e-01 7.76225150e-01
5.31455696e-01 6.43421486e-02 2.91174263e-01 -1.50822503e-02
-5.44864498e-02 -6.93780601e-01 1.05122185e+00 -2.45916754e-01
-4.29148108e-01 3.15701216e-02 -9.89263594e-01 -5.75078547e-01
1.29639947e+00 -3.93933326e-01 -6.01455979e-02 2.66889930e-01
-5.54766953e-01 -3.54834527e-01 1.21754456e+00 -7.03942120e-01
3.78333807e-01 -1.16844428e+00 -5.50773203e-01 3.64465624e-01
-4.96323973e-01 4.66004387e-02 9.99817908e-01 1.21550465e+00
-1.65829694e+00 1.11055516e-01 -2.68899590e-01 -6.33726597e-01
-1.09512770e+00 5.71211040e-01 4.05307353e-01 -4.16237175e-01
-7.77719080e-01 5.34398973e-01 -1.62242100e-01 -3.49928707e-01
-6.53077841e-01 -1.58132091e-01 -1.53492272e-01 -1.35787427e-01
2.35266730e-01 9.82090294e-01 6.91298068e-01 -1.00317800e+00
-3.53237957e-01 9.35852289e-01 6.35965705e-01 -5.21904491e-02
1.41597271e+00 -4.68129627e-02 -9.38267946e-01 4.59496640e-02
1.13854146e+00 6.56705424e-02 -1.01755035e+00 2.92073816e-01
-3.19420815e-01 -2.49996841e-01 4.74288434e-01 -2.43812844e-01
-8.50693524e-01 4.39006001e-01 3.87219071e-01 4.61331457e-01
1.15542269e+00 -4.64539856e-01 7.92858124e-01 -1.67515069e-01
5.27104378e-01 -1.28654683e+00 -5.38176060e-01 4.85817105e-01
8.37629080e-01 -4.95878249e-01 5.80344737e-01 -9.34121072e-01
-8.36416781e-02 1.38352191e+00 3.33422691e-01 -2.47135594e-01
9.14151073e-01 7.65830755e-01 -2.69627631e-01 -2.62128830e-01
-2.66418010e-01 1.89079285e-01 -3.74660408e-03 2.02623680e-01
5.49327970e-01 1.37699232e-01 -7.61289537e-01 1.48969039e-01
3.68429534e-02 1.14703432e-01 3.99180859e-01 1.34395218e+00
-1.81792706e-01 -1.28240955e+00 -1.10152650e+00 3.73423129e-01
-3.21723342e-01 9.19668227e-02 1.41861126e-01 6.69745326e-01
3.15331034e-02 6.51619315e-01 2.23420918e-01 -2.75890827e-02
2.20865328e-02 -1.91393852e-01 5.45673072e-01 -1.53916299e-01
-8.72477472e-01 2.34258339e-01 1.84508801e-01 -5.15219510e-01
-4.96953070e-01 -6.50449812e-01 -1.22755682e+00 -5.63428462e-01
-2.60856748e-01 4.45090413e-01 9.55944359e-01 8.77925038e-01
-4.48137373e-02 4.99112427e-01 5.37473381e-01 -1.57965720e+00
-7.67353401e-02 -8.32891643e-01 -9.57875311e-01 1.24984197e-01
-5.01141608e-01 -7.20720291e-01 -3.76192451e-01 -2.23305359e-01]
|
[6.406203269958496, 3.2507855892181396]
|
fb5ea43f-82f7-4ea3-a9bf-f0df0d5f3f46
|
joint-learning-of-interpretation-and
|
2005.11638
| null |
https://arxiv.org/abs/2005.11638v1
|
https://arxiv.org/pdf/2005.11638v1.pdf
|
Joint learning of interpretation and distillation
|
The extra trust brought by the model interpretation has made it an indispensable part of machine learning systems. But to explain a distilled model's prediction, one may either work with the student model itself, or turn to its teacher model. This leads to a more fundamental question: if a distilled model should give a similar prediction for a similar reason as its teacher model on the same input? This question becomes even more crucial when the two models have dramatically different structure, taking GBDT2NN for example. This paper conducts an empirical study on the new approach to explaining each prediction of GBDT2NN, and how imitating the explanation can further improve the distillation process as an auxiliary learning task. Experiments on several benchmarks show that the proposed methods achieve better performance on both explanations and predictions.
|
['Shenghong Li', 'Fucai Luo', 'Zhicong Yan', 'Jinchao Huang', 'Guofu Li']
|
2020-05-24
| null | null | null | null |
['auxiliary-learning']
|
['methodology']
|
[ 2.46483132e-01 1.13313675e+00 -6.04540527e-01 -7.63258576e-01
-2.12175325e-01 -3.28994125e-01 6.95622921e-01 5.79674765e-02
2.49061435e-01 8.76540840e-01 1.54532820e-01 -1.25722528e+00
-1.57452817e-03 -7.77931511e-01 -5.80001175e-01 -7.11808681e-01
3.29085678e-01 9.08966243e-01 2.82343388e-01 -2.96496361e-01
1.73551843e-01 3.33256185e-01 -1.21826446e+00 5.67317367e-01
1.08058774e+00 8.09488416e-01 5.89556992e-02 5.02703071e-01
-4.50489223e-01 1.21107149e+00 -5.25092900e-01 -1.08684230e+00
1.20706342e-01 -6.15617752e-01 -1.29590106e+00 2.28755083e-02
3.71320307e-01 -1.93752795e-01 -2.49401554e-02 9.49814439e-01
-2.66094029e-01 6.36212751e-02 8.91545951e-01 -1.62579668e+00
-7.46824622e-01 1.22934878e+00 -2.26762965e-01 5.27728721e-02
1.00529894e-01 -1.26201555e-01 1.24689698e+00 -6.70549572e-01
1.14321567e-01 1.28003252e+00 4.77882475e-01 9.26402211e-01
-1.40478206e+00 -5.97425461e-01 4.20051008e-01 6.07936978e-01
-6.35140777e-01 2.29376405e-01 6.45452380e-01 -1.06761672e-01
8.40796828e-01 5.09299397e-01 6.46751642e-01 1.15611029e+00
-2.92219706e-02 9.60903049e-01 1.10367465e+00 -3.57343227e-01
-1.31516874e-01 7.54970372e-01 6.15201831e-01 6.41990900e-01
4.41824198e-01 9.22769457e-02 -2.21093565e-01 1.74775869e-01
3.27446282e-01 8.35007802e-02 -4.22591597e-01 -2.86985874e-01
-7.16594517e-01 1.05665994e+00 5.72201371e-01 1.22565307e-01
2.48318370e-02 3.11948415e-02 -1.12765087e-02 5.77962101e-01
3.81457567e-01 6.21884763e-01 -9.75796700e-01 8.47738236e-02
-5.92976272e-01 6.04692884e-02 1.03972471e+00 9.17181313e-01
8.89676869e-01 5.35719022e-02 3.60582769e-01 4.02103186e-01
4.55090523e-01 1.74210876e-01 4.30418104e-01 -6.87816620e-01
4.66663539e-01 8.04983497e-01 -1.66287735e-01 -8.33481789e-01
-1.25661686e-01 -8.35201859e-01 -1.10222626e+00 1.94303200e-01
4.96663660e-01 1.24522589e-01 -9.60988998e-01 1.49930191e+00
2.10202426e-01 3.22285593e-01 2.74201334e-01 7.61216521e-01
1.12799883e+00 7.11334467e-01 -1.21279731e-02 -1.21416926e-01
8.79433036e-01 -1.56171691e+00 -4.55466509e-01 -4.15037245e-01
8.14408123e-01 -5.06757498e-01 1.02894235e+00 6.66968703e-01
-8.07819724e-01 -7.39709795e-01 -7.40086854e-01 -1.99392945e-01
-3.53453249e-01 -5.67894876e-02 8.65205586e-01 4.38949257e-01
-7.64576375e-01 7.07397640e-01 -7.40353882e-01 -2.28820682e-01
2.34601736e-01 7.08051145e-01 -2.87654966e-01 6.26881272e-02
-1.06077921e+00 1.22280860e+00 4.18813884e-01 -9.35431495e-02
-6.58632100e-01 -6.81383014e-01 -5.79726398e-01 3.05875242e-01
2.53029108e-01 -6.57554746e-01 1.70007360e+00 -1.10306966e+00
-1.35101175e+00 7.46094882e-01 -3.89993042e-01 -8.47946525e-01
5.53375959e-01 -8.68013874e-02 -2.59702533e-01 -4.44062054e-01
-6.18175268e-02 6.57607555e-01 6.82265937e-01 -1.46591532e+00
-7.86392033e-01 -2.30594069e-01 4.52684969e-01 -1.82698756e-01
-1.49629727e-01 -4.98258203e-01 -1.10759221e-01 -4.67591703e-01
5.22803664e-01 -9.46126103e-01 -4.07673895e-01 -2.90674031e-01
-6.44097507e-01 -5.10155857e-01 9.35018539e-01 -3.69930834e-01
1.14040124e+00 -1.68976974e+00 2.28802770e-01 2.72705406e-01
6.50000095e-01 3.84608328e-01 5.51885962e-02 2.81870455e-01
-5.30487657e-01 5.14953852e-01 4.96920757e-02 -4.12547290e-01
5.40648215e-02 7.96213567e-01 -8.87362480e-01 -2.56314278e-01
1.37175247e-01 8.68979931e-01 -8.55256677e-01 -4.05134290e-01
7.16792941e-02 2.25699380e-01 -6.20777547e-01 3.82007897e-01
-2.75220335e-01 5.87108254e-01 -5.09036422e-01 1.48633465e-01
4.97742653e-01 -6.69626355e-01 3.50457013e-01 -2.79429406e-02
2.41861388e-01 6.39167130e-01 -8.40062797e-01 9.56093013e-01
-2.55559385e-01 7.91394413e-01 -5.58460593e-01 -1.42675459e+00
1.25918841e+00 3.18164557e-01 -1.37545109e-01 -4.02027994e-01
-1.11693330e-02 3.62226903e-01 4.96932268e-01 -4.68163490e-01
4.15429771e-01 -5.35060704e-01 3.30226630e-01 7.05699086e-01
-8.84601325e-02 -2.87491113e-01 -1.94874287e-01 4.51532453e-01
6.79571748e-01 -8.36882554e-03 4.01582718e-01 -3.49421129e-02
6.13826811e-01 2.12112039e-01 5.65089226e-01 6.95701301e-01
1.15166746e-01 2.61703134e-01 7.93146789e-01 -9.90319908e-01
-7.29729533e-01 -7.61430264e-01 6.67897090e-02 7.05786824e-01
1.37492090e-01 -5.48230708e-01 -4.52488720e-01 -1.55439925e+00
2.31806301e-02 1.24443114e+00 -9.69549775e-01 -4.47887838e-01
-3.16950917e-01 -3.59133959e-01 2.91524321e-01 7.01584995e-01
4.80681568e-01 -7.16488659e-01 -3.87114108e-01 6.18749261e-02
-2.64609247e-01 -7.52098739e-01 1.62920922e-01 9.35255885e-01
-1.26130009e+00 -1.07368851e+00 -1.20646581e-01 -7.23400474e-01
9.55075264e-01 3.45928490e-01 1.48762107e+00 7.47681141e-01
6.00134254e-01 -1.29112080e-01 -3.98416996e-01 -7.16476619e-01
-8.97146046e-01 4.45171207e-01 -1.39764518e-01 -4.12792444e-01
6.61923051e-01 -6.91479325e-01 -1.43401682e-01 5.17019033e-01
-7.12721467e-01 7.18274295e-01 4.80670393e-01 9.76346970e-01
2.39195779e-01 1.32140711e-01 4.35355574e-01 -1.39990067e+00
4.37170506e-01 -4.28110212e-01 -2.98479229e-01 3.98573130e-01
-1.10271513e+00 5.63781321e-01 1.02159107e+00 -4.65068966e-01
-9.63332176e-01 -5.25338948e-03 -1.99343652e-01 -3.22629869e-01
-1.56926513e-01 5.57952583e-01 -2.54762083e-01 2.11438015e-01
5.08399189e-01 2.43842959e-01 -1.26098841e-01 -5.80800116e-01
2.19097644e-01 3.86371166e-01 4.40696895e-01 -4.28241014e-01
1.36076713e+00 3.97818541e-04 1.42862290e-01 -3.00575644e-01
-1.48132288e+00 -6.95799962e-02 -8.64439011e-01 1.05539143e-01
5.05726635e-01 -4.79675293e-01 -6.84729457e-01 -1.53239042e-01
-1.39836621e+00 -5.46711087e-01 -2.52361715e-01 2.95257896e-01
-3.18596125e-01 1.26390249e-01 -1.89846873e-01 -4.33408916e-01
3.79201770e-02 -1.23050976e+00 2.66171157e-01 2.67503679e-01
-4.75227058e-01 -1.47682464e+00 -7.00264797e-02 5.39464593e-01
3.53127539e-01 -2.33463645e-01 1.50172210e+00 -1.35764980e+00
-6.55414402e-01 -2.55154818e-02 -1.81103110e-01 5.37312210e-01
1.33433297e-01 -1.30373577e-03 -9.94893253e-01 1.97791249e-01
8.21656808e-02 -2.60272413e-01 9.00469959e-01 -8.32298994e-02
1.28131258e+00 -6.36091650e-01 -3.89069140e-01 4.96486008e-01
1.16918552e+00 1.16968416e-01 4.70569521e-01 4.23198462e-01
7.25954890e-01 6.09287500e-01 7.51871467e-01 -1.44185098e-02
4.58844274e-01 5.14913678e-01 9.58971381e-01 -1.49332091e-01
-4.65816557e-02 -5.70900857e-01 1.16430940e-02 9.62837040e-01
-2.97008097e-01 -4.27000195e-01 -1.12717462e+00 2.24317700e-01
-1.98045659e+00 -9.42925155e-01 -7.29859710e-01 1.90987015e+00
6.60354972e-01 2.40634918e-01 -2.46964335e-01 4.45215046e-01
2.78277695e-01 -6.95995465e-02 -4.61889267e-01 -7.91476548e-01
-2.14869995e-02 1.60486743e-01 9.13307816e-03 6.85777307e-01
-6.35366857e-01 8.85549545e-01 6.55435038e+00 5.14481664e-01
-1.19008839e+00 -3.15463915e-02 9.42240894e-01 5.01253724e-01
-6.23116076e-01 2.68124431e-01 -9.50234592e-01 1.74200892e-01
9.45643425e-01 -3.26988786e-01 1.33793384e-01 9.05413151e-01
9.93570238e-02 1.43674105e-01 -1.59171557e+00 5.72377384e-01
-9.28929001e-02 -1.30655217e+00 5.86633325e-01 2.06275612e-01
9.42868829e-01 -3.79975498e-01 1.58203557e-01 6.01335168e-01
4.53019857e-01 -1.39280474e+00 5.24706244e-01 2.34201044e-01
3.93180586e-02 -7.29964077e-01 1.06991029e+00 8.74405980e-01
-8.31162930e-01 -6.75922632e-02 -5.67363024e-01 -6.10401273e-01
-4.32048619e-01 1.44248769e-01 -1.48565161e+00 5.08286953e-01
4.02827263e-01 8.77241910e-01 -6.19481802e-01 8.38401496e-01
-1.02330852e+00 1.04888380e+00 1.45318180e-01 -2.01967388e-01
4.51556295e-01 -2.22634703e-01 2.01217204e-01 8.02719295e-01
2.54074991e-01 3.83999497e-01 -1.27826571e-01 8.65508676e-01
-1.14031747e-01 -1.36948496e-01 -5.42548001e-01 1.19026974e-01
1.30797356e-01 1.11120939e+00 -6.07036710e-01 -7.76944339e-01
-4.71623719e-01 8.12224030e-01 4.48319525e-01 3.08740437e-01
-1.04586577e+00 1.67450577e-01 5.51146746e-01 2.77461231e-01
2.71092981e-01 3.17643940e-01 -4.17422056e-01 -1.12451434e+00
-2.50017494e-01 -1.11094975e+00 3.68334830e-01 -1.02357411e+00
-1.09562778e+00 8.31955791e-01 -1.02522098e-01 -1.32044923e+00
-4.33249772e-01 -6.42484665e-01 -9.53655362e-01 8.66923332e-01
-1.75841558e+00 -9.87918258e-01 -3.82234931e-01 5.04196942e-01
6.37800515e-01 -3.71444225e-02 9.88538384e-01 -3.55764739e-02
-3.92744005e-01 5.70193946e-01 -7.94110820e-02 -1.14178278e-01
4.89521801e-01 -1.65308034e+00 3.18673044e-01 6.35377228e-01
4.76458639e-01 7.49180615e-01 1.16530061e+00 -3.61694247e-01
-1.03399146e+00 -1.19040430e+00 1.41415095e+00 -7.25273490e-01
6.29989803e-01 2.37157509e-01 -1.27409577e+00 1.14060009e+00
3.16613823e-01 -2.43496329e-01 7.86104321e-01 3.29075992e-01
-5.08034527e-01 6.38619345e-03 -9.29820359e-01 5.47914147e-01
7.87802160e-01 -3.17900181e-01 -1.21462333e+00 2.56320059e-01
8.67269158e-01 -3.36530596e-01 -6.16174757e-01 3.90912443e-01
5.05679250e-01 -1.23501885e+00 6.62312806e-01 -1.26889265e+00
1.00649333e+00 -1.97437078e-01 -1.35777220e-02 -1.59848809e+00
-3.10737848e-01 -3.46367985e-01 -2.55794942e-01 1.08316612e+00
9.78594005e-01 -7.27822661e-01 1.20952010e+00 8.23927760e-01
-1.22152172e-01 -1.17128587e+00 -4.52861220e-01 -7.33909905e-01
1.62094310e-01 -6.49152577e-01 9.21304345e-01 1.10556185e+00
6.44823611e-02 6.78511143e-01 -3.87254179e-01 3.30079287e-01
3.96133482e-01 4.09872383e-01 1.17208862e+00 -1.75373876e+00
-3.53636026e-01 -3.16390574e-01 -2.84639806e-01 -1.52095318e+00
3.64141881e-01 -1.13011205e+00 -1.65307999e-01 -1.69519687e+00
2.04715803e-01 -4.91633832e-01 -3.31004798e-01 7.80434430e-01
-2.57883966e-01 1.54425383e-01 1.44520715e-01 9.51604992e-02
-3.61557573e-01 4.06315982e-01 1.41386247e+00 -2.91130126e-01
1.14292406e-01 6.01923764e-01 -9.97528076e-01 1.00875461e+00
7.59705424e-01 -8.68362546e-01 -7.11447954e-01 -5.27323008e-01
3.62486243e-01 3.92248780e-02 5.09117246e-01 -6.65184796e-01
3.03546011e-01 -2.97400087e-01 1.99820235e-01 -5.89243352e-01
2.08706141e-01 -1.34253192e+00 2.87659973e-01 8.18552136e-01
-6.07431412e-01 4.04720977e-02 1.00008056e-01 3.99217665e-01
-3.79883945e-01 -3.66539240e-01 5.93168616e-01 -3.30192931e-02
-4.39657360e-01 2.60056674e-01 -2.47695968e-01 -3.06362540e-01
7.15757072e-01 -5.52337289e-01 -3.92394930e-01 -6.36091352e-01
-8.61431956e-01 2.34637290e-01 2.58471787e-01 3.32117647e-01
6.76968932e-01 -1.19995272e+00 -4.63999927e-01 2.80180842e-01
-2.43269447e-02 -6.14237674e-02 -2.95739412e-01 8.64183247e-01
-3.22171636e-02 7.02393293e-01 -8.88242722e-02 -5.78180134e-01
-1.58487391e+00 5.25401354e-01 4.17691410e-01 -6.10864520e-01
-3.97307575e-01 9.62365150e-01 3.88285935e-01 -8.22023273e-01
3.29984367e-01 -7.79523253e-01 -2.83871830e-01 -2.19529063e-01
3.31238449e-01 1.58499345e-01 -1.45366535e-01 -4.22013223e-01
4.05419841e-02 3.80817533e-01 -2.33998179e-01 3.90791059e-01
1.48349547e+00 -3.28155123e-02 -1.41042201e-02 5.78047276e-01
8.40006232e-01 -1.97658718e-01 -8.82610917e-01 -2.81235397e-01
7.71350712e-02 -4.71075147e-01 -2.24878505e-01 -1.03167844e+00
-1.17886376e+00 1.33971334e+00 -1.80133730e-01 7.60871291e-01
9.67711329e-01 1.78036064e-01 4.66875464e-01 7.47227788e-01
3.24266851e-01 -2.17428640e-01 6.17956407e-02 7.12306619e-01
8.57459664e-01 -1.47013867e+00 -2.28061862e-02 -5.81787467e-01
-7.16204941e-01 1.45886803e+00 1.03099465e+00 1.43310115e-01
4.15587246e-01 -2.17902750e-01 1.55407816e-01 -1.40906185e-01
-1.34412456e+00 2.19397739e-01 5.65551937e-01 4.39239204e-01
6.03434145e-01 7.62719661e-02 8.15422088e-02 9.19217885e-01
-5.75934589e-01 -2.27972463e-01 5.97656250e-01 2.60687888e-01
-5.75027943e-01 -1.49246442e+00 -7.39341825e-02 6.07561588e-01
-2.56788969e-01 -1.27914727e-01 -8.24330151e-01 1.08197761e+00
2.93948472e-01 9.97366130e-01 -3.50247920e-01 -6.71346307e-01
1.00364015e-01 1.22650005e-01 1.17771640e-01 -9.51534927e-01
-7.96583831e-01 -3.54773372e-01 4.83550467e-02 -3.22447181e-01
-4.85778570e-01 -2.40850776e-01 -1.57259536e+00 -5.66866100e-01
-4.35506254e-01 9.11702514e-01 3.89273584e-01 1.29900646e+00
-1.53772980e-01 5.22397220e-01 5.52014232e-01 -2.99757898e-01
-6.95934594e-01 -9.11547065e-01 -3.57785523e-01 1.30433202e-01
4.13805485e-01 -3.79018873e-01 -6.70951009e-01 7.17015043e-02]
|
[9.02055835723877, 6.1263885498046875]
|
bef96a1f-a9c5-4c8b-8efa-71ab02f4d2de
|
learning-dynamics-via-graph-neural-networks
|
2106.03772
| null |
https://arxiv.org/abs/2106.03772v1
|
https://arxiv.org/pdf/2106.03772v1.pdf
|
Learning Dynamics via Graph Neural Networks for Human Pose Estimation and Tracking
|
Multi-person pose estimation and tracking serve as crucial steps for video understanding. Most state-of-the-art approaches rely on first estimating poses in each frame and only then implementing data association and refinement. Despite the promising results achieved, such a strategy is inevitably prone to missed detections especially in heavily-cluttered scenes, since this tracking-by-detection paradigm is, by nature, largely dependent on visual evidences that are absent in the case of occlusion. In this paper, we propose a novel online approach to learning the pose dynamics, which are independent of pose detections in current fame, and hence may serve as a robust estimation even in challenging scenarios including occlusion. Specifically, we derive this prediction of dynamics through a graph neural network~(GNN) that explicitly accounts for both spatial-temporal and visual information. It takes as input the historical pose tracklets and directly predicts the corresponding poses in the following frame for each tracklet. The predicted poses will then be aggregated with the detected poses, if any, at the same frame so as to produce the final pose, potentially recovering the occluded joints missed by the estimator. Experiments on PoseTrack 2017 and PoseTrack 2018 datasets demonstrate that the proposed method achieves results superior to the state of the art on both human pose estimation and tracking tasks.
|
['Gang Hua', 'Xinchao Wang', 'Chunluan Zhou', 'Haoxiang Li', 'Zhou Ren', 'Yiding Yang']
|
2021-06-07
| null |
http://openaccess.thecvf.com//content/CVPR2021/html/Yang_Learning_Dynamics_via_Graph_Neural_Networks_for_Human_Pose_Estimation_CVPR_2021_paper.html
|
http://openaccess.thecvf.com//content/CVPR2021/papers/Yang_Learning_Dynamics_via_Graph_Neural_Networks_for_Human_Pose_Estimation_CVPR_2021_paper.pdf
|
cvpr-2021-1
|
['multi-person-pose-estimation-and-tracking']
|
['computer-vision']
|
[ 7.07102716e-02 -1.74498141e-01 3.61989774e-02 -2.93722861e-02
-5.71020782e-01 -3.62232983e-01 4.66987342e-01 1.27071915e-02
-5.19808412e-01 7.45142639e-01 1.30091935e-01 3.98264438e-01
-3.70137095e-02 -5.11073053e-01 -8.96167397e-01 -5.73066652e-01
-1.00627027e-01 7.97849715e-01 4.63261127e-01 -1.45118877e-01
-2.66991049e-01 4.13672984e-01 -1.72849631e+00 -7.96962306e-02
6.35981143e-01 9.35039997e-01 1.20210789e-01 6.70643330e-01
2.84323901e-01 5.71079910e-01 -4.73275930e-01 -5.61058223e-01
2.74601579e-01 -1.22865438e-01 -1.77833259e-01 4.45092857e-01
9.60520566e-01 -4.30294037e-01 -5.96131980e-01 8.80896866e-01
4.63488191e-01 2.11942598e-01 2.89707035e-01 -1.00572085e+00
1.79343715e-01 1.77908450e-01 -8.01452041e-01 1.89466134e-01
7.62348473e-01 1.39897212e-01 7.07058132e-01 -9.78142738e-01
9.13653851e-01 1.14530492e+00 9.37027156e-01 4.06896472e-01
-1.14651668e+00 -6.34339690e-01 5.10075033e-01 2.98334867e-01
-1.44424355e+00 -3.97844791e-01 7.46511340e-01 -6.13917530e-01
4.30249542e-01 1.85957402e-02 1.17789435e+00 1.32584739e+00
1.21559322e-01 9.53043759e-01 6.44746542e-01 -1.68656051e-01
-1.24256000e-01 -2.56110817e-01 -6.00082278e-02 8.89015913e-01
5.26740372e-01 2.07773507e-01 -8.89070511e-01 1.04454989e-02
7.58292675e-01 1.19470060e-01 -3.18171501e-01 -6.88881636e-01
-1.33731091e+00 4.04647470e-01 5.15313923e-01 -1.07531011e-01
-8.17547619e-01 1.93820745e-01 2.88448960e-01 2.89122947e-03
5.78671753e-01 -1.19276293e-01 -3.61488402e-01 -6.30983524e-03
-1.10258412e+00 5.76169729e-01 5.02272427e-01 8.69712710e-01
4.61371303e-01 -3.65465023e-02 -2.69299001e-01 4.62863028e-01
3.34862411e-01 5.66397667e-01 -2.38081560e-01 -6.32829845e-01
6.54172242e-01 5.15216351e-01 4.59672600e-01 -1.08594167e+00
-7.38877356e-01 -8.86698484e-01 -6.16984606e-01 -4.09014989e-03
8.29065025e-01 -2.35634789e-01 -8.15730035e-01 1.79223323e+00
8.02024603e-01 3.79939437e-01 -5.49422085e-01 1.10167015e+00
5.36029458e-01 3.20168823e-01 1.45645529e-01 -3.54081571e-01
1.29598284e+00 -8.09589267e-01 -7.84055173e-01 -4.05253053e-01
1.45166278e-01 -7.10542798e-01 2.45215029e-01 5.43523610e-01
-8.17875326e-01 -9.03592408e-01 -8.91863704e-01 2.82728374e-01
8.15196037e-02 5.99187791e-01 4.06831324e-01 5.34829378e-01
-6.78247154e-01 5.93041658e-01 -1.18151832e+00 -5.37419677e-01
2.51527578e-01 3.93044174e-01 -3.70738477e-01 -3.23813446e-02
-8.58888388e-01 9.34356451e-01 4.53538328e-01 6.23443365e-01
-7.91854918e-01 -5.11679649e-01 -8.32658291e-01 -2.46861041e-01
9.55378771e-01 -9.30512428e-01 9.71591771e-01 -4.23672199e-01
-1.11708045e+00 3.47172290e-01 -3.00470293e-01 -6.07052863e-01
1.19650280e+00 -9.25488114e-01 -2.56267071e-01 1.34866580e-01
7.45935440e-02 5.98225832e-01 1.09712207e+00 -1.19745672e+00
-8.47206295e-01 -5.61566770e-01 1.54125076e-02 3.23068142e-01
-1.39147565e-01 -2.62763470e-01 -1.12351632e+00 -5.63254416e-01
3.43699336e-01 -1.28210759e+00 -1.82940140e-01 3.93958479e-01
-3.58921409e-01 -2.15386584e-01 6.02699518e-01 -1.10850203e+00
1.09119308e+00 -1.66362298e+00 6.23988330e-01 2.37674937e-01
2.10341111e-01 2.39002869e-01 2.21271768e-01 2.90067762e-01
2.36873344e-01 -7.37815857e-01 1.96847141e-01 -6.94052339e-01
-2.25217819e-01 1.00843748e-02 9.48027615e-03 9.44569647e-01
1.40844166e-01 8.17121327e-01 -9.02064443e-01 -5.42071283e-01
6.72238648e-01 7.63622284e-01 -5.72924793e-01 1.21049590e-01
-4.22614306e-01 1.00469589e+00 -3.91996354e-01 5.97651660e-01
4.65544373e-01 -1.57447413e-01 3.47553045e-01 -5.25607646e-01
-1.95704535e-01 -1.16039783e-01 -1.67605722e+00 1.88275659e+00
-2.47489009e-02 4.95553732e-01 -9.42887142e-02 -6.33597851e-01
5.71128845e-01 4.41630751e-01 8.10164869e-01 -2.81862229e-01
2.16736346e-01 -6.15905412e-02 -1.03175998e-01 -3.53687346e-01
4.46170151e-01 1.85959533e-01 9.38878134e-02 -8.27359408e-02
1.35871977e-01 5.72656691e-01 5.01245618e-01 1.09495260e-01
8.75785351e-01 8.54262292e-01 2.38020301e-01 1.76233590e-01
6.71019852e-01 -1.00702561e-01 8.47749591e-01 6.42764449e-01
-2.71433711e-01 5.60097337e-01 1.08245306e-01 -6.37332022e-01
-1.04759896e+00 -1.15524125e+00 1.76665857e-01 8.74342322e-01
2.52760291e-01 -5.80012918e-01 -5.79039335e-01 -5.02936006e-01
3.65332700e-02 1.90515816e-02 -6.16690636e-01 -5.00654150e-03
-8.75828087e-01 -4.74517763e-01 1.42541692e-01 5.25288105e-01
3.29874009e-01 -9.13283169e-01 -6.97100759e-01 5.32783508e-01
-5.12132466e-01 -1.40236092e+00 -4.50517684e-01 -4.11209792e-01
-6.63812935e-01 -1.41110241e+00 -7.10895002e-01 -4.08259779e-01
6.05512202e-01 1.40658781e-01 8.74596655e-01 1.23545498e-01
-3.26731682e-01 3.35484266e-01 -2.26764813e-01 -1.96647972e-01
1.00419717e-02 -6.64380938e-02 4.07501906e-01 3.76916438e-01
6.94121839e-03 -3.10421228e-01 -8.41516793e-01 2.28612527e-01
-2.32825056e-01 1.52658358e-01 5.81510663e-01 7.36953557e-01
6.32041931e-01 -3.49300615e-02 2.44498640e-01 -7.07146466e-01
-4.71092686e-02 -5.11417761e-02 -7.62757301e-01 3.27900827e-01
-1.56754166e-01 -1.71864375e-01 3.30729157e-01 -5.21533847e-01
-1.09369695e+00 6.99960947e-01 -4.91060987e-02 -6.77691698e-01
-1.48986764e-02 3.15777808e-01 -1.20948464e-01 -9.56096947e-02
4.70317096e-01 1.80206940e-01 -5.44056483e-02 -6.61691010e-01
2.87279099e-01 -1.03897110e-01 9.07418311e-01 -4.58608896e-01
1.11933386e+00 6.14241183e-01 1.89755648e-01 -8.53504837e-01
-1.00531292e+00 -7.39164233e-01 -1.02158785e+00 -9.23146605e-01
9.67139482e-01 -1.22855830e+00 -9.64149058e-01 4.55517054e-01
-1.20129585e+00 3.42594117e-01 1.12694390e-01 6.66561842e-01
-4.14526403e-01 4.44371849e-01 -4.74443048e-01 -1.26445472e+00
-1.13507226e-01 -1.00569856e+00 1.23763192e+00 2.75918365e-01
-3.25535506e-01 -6.77549422e-01 -2.14127973e-02 4.46417511e-01
-2.09814817e-01 7.13626504e-01 7.38735572e-02 -1.19824402e-01
-8.57126415e-01 -5.40684640e-01 1.01554781e-01 -1.16137944e-01
-8.89482275e-02 -9.12533626e-02 -7.98231184e-01 -5.86452186e-01
-3.22683871e-01 3.80750783e-02 7.97257125e-01 6.32319808e-01
6.94284797e-01 -2.39571352e-02 -5.37533343e-01 5.18888533e-01
1.16436255e+00 -3.60241085e-01 2.77018189e-01 2.70370424e-01
1.03835690e+00 5.89006841e-01 9.01371300e-01 6.25355184e-01
4.14342344e-01 1.19026101e+00 5.53075612e-01 4.06256504e-02
-3.20979416e-01 -4.02451098e-01 3.28425169e-01 5.18817186e-01
-6.71433628e-01 5.21238260e-02 -6.33001506e-01 4.36404407e-01
-2.25230455e+00 -1.05931842e+00 -3.34021807e-01 2.39980745e+00
4.88261282e-01 3.99729908e-01 4.99203116e-01 9.29125622e-02
8.59159231e-01 1.94289938e-01 -4.94587064e-01 6.44114971e-01
-1.39016076e-03 1.17721103e-01 5.12051582e-01 2.12478086e-01
-1.41742659e+00 8.33298147e-01 4.95572376e+00 3.37306976e-01
-7.00208008e-01 2.08747815e-02 -1.65952787e-01 -3.27160627e-01
4.91133481e-01 -1.07194811e-01 -9.36491311e-01 3.35342169e-01
5.17762065e-01 1.96449161e-01 2.14664176e-01 5.79419553e-01
3.41112524e-01 -2.58962661e-01 -1.10459137e+00 9.68925714e-01
-1.60122477e-02 -1.12215388e+00 -1.67857453e-01 -3.46037420e-03
6.35038376e-01 -2.81230748e-01 -2.21141711e-01 1.76874176e-01
-5.62582687e-02 -4.66518849e-01 1.10120976e+00 7.87551224e-01
3.88408065e-01 -8.01360130e-01 6.55502737e-01 5.54678440e-01
-1.67395806e+00 -6.11012653e-02 -1.18578732e-01 -8.79852250e-02
5.38381934e-01 4.76315022e-01 -5.91029286e-01 9.07192886e-01
7.24599361e-01 9.75798368e-01 -6.71735108e-01 1.39831829e+00
-4.51797545e-01 2.76722819e-01 -4.28061008e-01 2.41917357e-01
-2.38527521e-01 -6.89683808e-03 7.87286460e-01 9.03271198e-01
2.65476882e-01 -1.56651467e-01 7.55739927e-01 4.68864679e-01
2.03451648e-01 -1.70317695e-01 -2.30945498e-01 4.14369076e-01
3.67892027e-01 1.23509455e+00 -7.47860670e-01 -3.79754573e-01
-5.30526757e-01 9.02921796e-01 5.33325672e-01 2.02728897e-01
-9.64612126e-01 4.16700035e-01 6.95880175e-01 2.94996172e-01
5.53151786e-01 -5.52353561e-01 5.89085668e-02 -1.32842791e+00
4.00420785e-01 -6.88985705e-01 5.85339069e-01 -3.92656416e-01
-1.00744319e+00 3.05308878e-01 -1.81749149e-03 -1.49661553e+00
-4.67880726e-01 -5.14664650e-01 -2.27681547e-01 5.10849774e-01
-1.09475410e+00 -1.27413130e+00 -4.40721720e-01 5.48589885e-01
6.24793470e-01 2.51257360e-01 4.93678600e-01 5.66983044e-01
-8.00751626e-01 3.94549727e-01 -2.82670349e-01 3.46125424e-01
7.64144838e-01 -9.79193389e-01 3.49034071e-01 1.17901576e+00
4.84506369e-01 5.30821085e-01 1.18786645e+00 -1.09904432e+00
-1.59765911e+00 -1.05809247e+00 7.26186991e-01 -5.69789648e-01
4.94187772e-01 -4.97761250e-01 -7.24060416e-01 6.98046029e-01
-4.24147338e-01 2.19734326e-01 1.04882777e-01 2.10863337e-01
-7.70511776e-02 -6.21435754e-02 -6.00391150e-01 5.13287842e-01
1.43594241e+00 -3.14750940e-01 -4.52319950e-01 3.95486593e-01
2.99326777e-01 -8.89813542e-01 -6.20061338e-01 4.79141921e-01
8.67401600e-01 -8.96680713e-01 1.27736282e+00 -4.65513825e-01
-9.34326425e-02 -6.21502399e-01 2.02832058e-01 -1.01503730e+00
-3.17065209e-01 -5.19554377e-01 -6.23220444e-01 9.15506124e-01
-1.81804709e-02 -1.00940190e-01 1.07528746e+00 3.78377110e-01
1.87195122e-01 -5.12544096e-01 -1.07033515e+00 -8.21316421e-01
-8.11152875e-01 -5.28251529e-01 1.38688043e-01 3.71809274e-01
-5.41701853e-01 1.94005936e-01 -1.10903108e+00 5.39853811e-01
1.19273698e+00 1.07916206e-01 1.16598833e+00 -1.44346595e+00
-3.79816443e-01 1.16254967e-02 -6.98231697e-01 -1.17408752e+00
4.95764688e-02 -3.38720828e-01 2.60962456e-01 -1.39081001e+00
2.16523990e-01 -2.25664452e-02 -8.83582756e-02 1.67436764e-01
-4.32848096e-01 5.23587406e-01 4.60185945e-01 2.99029469e-01
-8.76329839e-01 5.40893555e-01 1.19485354e+00 -1.88745447e-02
-1.22714274e-01 3.80168945e-01 -9.10125673e-02 9.11060333e-01
1.53695747e-01 -4.66617674e-01 -8.65913481e-02 -2.41816744e-01
1.09862044e-01 3.04792792e-01 8.12543809e-01 -1.45625830e+00
5.19564271e-01 4.83346283e-02 9.72665906e-01 -1.21479106e+00
7.35010207e-01 -9.60776806e-01 6.11876726e-01 7.75085270e-01
3.55525203e-02 2.92631052e-02 5.46618737e-02 1.04883850e+00
1.15013950e-01 2.96445757e-01 4.10084099e-01 -7.19687417e-02
-8.95555079e-01 6.94893420e-01 1.04978524e-01 -1.71277806e-01
9.48231280e-01 -3.32495987e-01 9.00261775e-02 -4.75298285e-01
-1.10020792e+00 2.94418514e-01 3.16543907e-01 6.34709179e-01
4.20295715e-01 -1.33592153e+00 -8.51952255e-01 1.57536753e-02
1.18343703e-01 9.34693888e-02 4.05853808e-01 1.12211347e+00
-2.16401592e-01 2.94591606e-01 -1.15018129e-01 -9.80252326e-01
-1.33943999e+00 4.99978781e-01 1.36369377e-01 -3.49350095e-01
-9.40343499e-01 5.71325481e-01 1.38018548e-01 9.57206637e-02
4.32595968e-01 8.10095742e-02 -3.80582452e-01 2.51357645e-01
4.81030196e-01 4.19632733e-01 9.98586491e-02 -1.08857894e+00
-5.23064554e-01 8.04395735e-01 -2.01636143e-02 -6.43404126e-02
1.34527218e+00 -2.20867470e-01 2.94151187e-01 3.24839234e-01
7.75521398e-01 -3.73067781e-02 -1.83155823e+00 -5.15350938e-01
4.68788967e-02 -6.47799432e-01 -3.54044974e-01 -4.37735021e-01
-1.11843026e+00 5.43755531e-01 5.78633547e-01 -2.66897798e-01
8.00807774e-01 -9.44135245e-04 6.75233960e-01 1.99705482e-01
6.81954265e-01 -1.12287951e+00 7.41344318e-02 3.24035704e-01
7.78097212e-01 -1.05901909e+00 4.18923199e-01 -7.04925776e-01
-2.39785910e-01 1.09302843e+00 7.07023621e-01 -2.51364917e-01
3.15782040e-01 3.16333398e-03 -1.74732894e-01 -1.20811097e-01
-4.85849470e-01 -5.12682557e-01 8.06162596e-01 5.84918857e-01
2.67713279e-01 -6.77895620e-02 -2.05646157e-01 1.92903042e-01
-2.59726137e-01 1.01793825e-03 5.70349991e-02 9.47282255e-01
-3.59813213e-01 -1.03308952e+00 -6.87405288e-01 3.68364751e-01
-2.89715141e-01 4.11672294e-01 -3.06807637e-01 9.17766213e-01
3.05319786e-01 7.82098711e-01 -1.19542539e-01 -3.99976641e-01
6.08153880e-01 -1.20627023e-02 9.37332451e-01 -4.65647936e-01
-4.80102062e-01 4.28236097e-01 2.63859570e-01 -7.90207684e-01
-6.04877353e-01 -1.20079470e+00 -1.05709136e+00 -2.28850260e-01
-4.14165050e-01 -2.64358997e-01 4.40067679e-01 1.20809317e+00
6.09979481e-02 7.68328547e-01 6.45455644e-02 -1.33473849e+00
-3.97698730e-01 -8.38565052e-01 -2.73641229e-01 6.88029289e-01
3.97625893e-01 -1.27580070e+00 1.69896156e-01 1.13145366e-01]
|
[7.013377666473389, -0.9597166776657104]
|
8e4dafa7-c3ea-4723-b982-766d28d943de
|
weighted-sampling-for-masked-language
|
2302.14225
| null |
https://arxiv.org/abs/2302.14225v2
|
https://arxiv.org/pdf/2302.14225v2.pdf
|
Weighted Sampling for Masked Language Modeling
|
Masked Language Modeling (MLM) is widely used to pretrain language models. The standard random masking strategy in MLM causes the pre-trained language models (PLMs) to be biased toward high-frequency tokens. Representation learning of rare tokens is poor and PLMs have limited performance on downstream tasks. To alleviate this frequency bias issue, we propose two simple and effective Weighted Sampling strategies for masking tokens based on the token frequency and training loss. We apply these two strategies to BERT and obtain Weighted-Sampled BERT (WSBERT). Experiments on the Semantic Textual Similarity benchmark (STS) show that WSBERT significantly improves sentence embeddings over BERT. Combining WSBERT with calibration methods and prompt learning further improves sentence embeddings. We also investigate fine-tuning WSBERT on the GLUE benchmark and show that Weighted Sampling also improves the transfer learning capability of the backbone PLM. We further analyze and provide insights into how WSBERT improves token embeddings.
|
['Wei Wang', 'Yuxin Jiang', 'Kongzhang Hao', 'Xin Cao', 'Chong Deng', 'Wen Wang', 'Qian Chen', 'Linhan Zhang']
|
2023-02-28
| null | null | null | null |
['sentence-embeddings', 'sentence-embeddings', 'semantic-textual-similarity']
|
['methodology', 'natural-language-processing', 'natural-language-processing']
|
[ 1.39528349e-01 5.81818447e-02 -6.24443054e-01 -4.68127251e-01
-1.09771454e+00 -4.72140014e-01 6.86889946e-01 4.48065251e-01
-8.71515989e-01 4.93368864e-01 4.86697614e-01 -4.55742091e-01
3.75157773e-01 -6.64481640e-01 -6.87452674e-01 -4.22375470e-01
-5.73644787e-02 2.95948446e-01 4.02922720e-01 -1.91473275e-01
1.71462089e-01 1.55386880e-01 -1.07030272e+00 7.29246736e-01
1.05230188e+00 4.55167979e-01 6.66638494e-01 4.65901107e-01
-5.85146546e-01 7.20281184e-01 -6.71899617e-01 -4.34807628e-01
1.57462969e-01 -3.55545819e-01 -8.50379407e-01 -2.30244622e-01
1.96657434e-01 -1.48581341e-01 -1.20028816e-01 7.76938796e-01
4.64818627e-01 3.11298341e-01 7.36572266e-01 -1.00767803e+00
-8.32245171e-01 1.54982650e+00 -6.48913383e-01 3.92709374e-01
-7.54235014e-02 5.62037565e-02 1.28455997e+00 -1.21961319e+00
3.24112862e-01 1.49966931e+00 7.87577391e-01 6.49113059e-01
-1.48662722e+00 -6.94654644e-01 5.36277115e-01 1.55953720e-01
-1.26200914e+00 -5.00271797e-01 5.01605272e-01 -3.01039338e-01
1.28103876e+00 5.88721521e-02 6.39800876e-02 9.99652565e-01
2.31129136e-02 1.01457036e+00 9.67261016e-01 -8.61497045e-01
1.93677157e-01 5.12320876e-01 3.56502354e-01 5.24208546e-01
2.72141427e-01 5.25187999e-02 -9.37004566e-01 -1.76989511e-01
2.10041493e-01 -1.06520861e-01 1.16421273e-02 -6.54219389e-02
-1.12362194e+00 9.81917083e-01 4.34852600e-01 4.26938295e-01
-3.32647637e-02 2.49329507e-01 5.95470130e-01 3.94780844e-01
5.96526027e-01 6.96742654e-01 -6.49464250e-01 -1.20208085e-01
-1.14741302e+00 1.74430553e-02 4.33573365e-01 9.28601921e-01
8.90579462e-01 2.61731148e-01 -6.22659504e-01 1.30063426e+00
3.01113546e-01 2.33884797e-01 8.04338992e-01 -5.97405314e-01
7.06246138e-01 4.32013810e-01 -1.57497942e-01 -2.85142839e-01
-9.91162211e-02 -3.60358715e-01 -3.73775542e-01 -1.73802629e-01
2.61367798e-01 -1.02874614e-01 -8.79469454e-01 1.97838306e+00
-1.50825098e-01 2.14844838e-01 4.60987389e-02 5.88752270e-01
6.08815014e-01 6.31227612e-01 5.67084610e-01 1.97655648e-01
1.25122726e+00 -1.19072151e+00 -4.69007283e-01 -6.04345560e-01
1.28835475e+00 -7.18446314e-01 1.59613252e+00 1.11997291e-01
-1.04019845e+00 -5.80651760e-01 -1.09203422e+00 -1.57570735e-01
-4.09940511e-01 1.70877039e-01 5.07456243e-01 6.02872968e-01
-9.21816647e-01 7.91524470e-01 -8.76333773e-01 -2.55049288e-01
4.72563475e-01 1.18292160e-01 -1.40266465e-02 -1.46689832e-01
-1.53331637e+00 1.12923658e+00 3.08045477e-01 -2.54597098e-01
-9.50435638e-01 -9.85837698e-01 -9.37248826e-01 2.26312682e-01
-4.95946873e-03 -3.86666924e-01 1.38911784e+00 -7.39463985e-01
-1.33278394e+00 8.96601737e-01 -5.20510733e-01 -7.32749879e-01
3.16244245e-01 -2.53551364e-01 -4.59667519e-02 -1.00807324e-02
1.58708140e-01 9.92465675e-01 6.37996256e-01 -1.18157792e+00
-3.65212381e-01 2.62401730e-01 -2.42074654e-01 1.49502605e-02
-8.40227246e-01 1.82852983e-01 -2.30265018e-02 -8.51156473e-01
-3.43889624e-01 -4.81863230e-01 -2.14809969e-01 -3.55759978e-01
-2.40523845e-01 -4.04870927e-01 2.11247966e-01 -4.50967193e-01
1.42054105e+00 -2.03511477e+00 -2.23476723e-01 -4.60490659e-02
-7.77806640e-02 4.49976057e-01 -7.24076569e-01 6.15804553e-01
-7.34330118e-02 2.96811104e-01 -2.00082958e-01 -8.80004227e-01
1.75670728e-01 3.74366522e-01 -6.24993324e-01 -4.27134782e-02
4.57031399e-01 1.09363687e+00 -1.01595509e+00 -4.16352302e-01
1.01581022e-01 2.11098030e-01 -8.85840178e-01 3.27057600e-01
-2.68833548e-01 1.72021985e-02 6.28537461e-02 2.81552792e-01
6.58033431e-01 -1.43173367e-01 2.10498258e-01 3.31146978e-02
-1.31352646e-02 1.04641545e+00 -6.09016120e-01 1.65129673e+00
-9.48037028e-01 4.89496499e-01 -2.29935586e-01 -1.02519822e+00
9.37270403e-01 1.47733897e-01 8.15310627e-02 -5.47511578e-01
-2.42788475e-02 8.37966278e-02 6.61089346e-02 -4.25323814e-01
7.14101672e-01 -4.34584022e-01 -5.11847809e-02 4.90820974e-01
2.19381422e-01 -3.24093252e-02 4.72326428e-02 4.16471034e-01
1.01455104e+00 5.09621017e-02 1.46021560e-01 -3.51316929e-01
2.66232908e-01 -1.37564808e-01 6.25524521e-01 9.56091881e-01
-1.08095832e-01 3.86074364e-01 4.26803350e-01 2.45153770e-01
-8.82821739e-01 -1.36767697e+00 -3.35060502e-03 1.75734735e+00
-2.40095124e-01 -7.49676526e-01 -6.36518002e-01 -8.61098647e-01
3.40095669e-01 1.24451959e+00 -3.86954486e-01 -7.19426751e-01
-5.25117576e-01 -8.77336740e-01 6.57093823e-01 8.42811763e-01
6.03150614e-02 -1.07914257e+00 -5.74966557e-02 2.38512352e-01
-1.29930153e-01 -1.00390899e+00 -6.96204066e-01 4.78487968e-01
-8.15045297e-01 -4.67400283e-01 -5.19834578e-01 -9.21825528e-01
7.36351132e-01 4.96385783e-01 1.24459028e+00 1.78891718e-01
4.59945686e-02 -3.60016525e-02 -5.64263940e-01 -3.83388102e-01
-6.48990452e-01 4.24933136e-01 1.51253179e-01 -1.62378684e-01
7.32650936e-01 -3.63016397e-01 -2.53961653e-01 1.32515028e-01
-8.19754720e-01 -8.75492021e-02 6.36892974e-01 1.10503471e+00
1.03256799e-01 -3.42515796e-01 9.26056564e-01 -1.11365199e+00
7.00492918e-01 -5.01506984e-01 -3.42848212e-01 2.55900085e-01
-7.12733865e-01 3.92423958e-01 6.40855253e-01 -8.20994794e-01
-9.52613235e-01 -3.01193476e-01 -2.42258325e-01 -2.87564307e-01
1.40735030e-01 6.02237582e-01 -1.25611462e-02 7.57474974e-02
6.42140925e-01 3.82034741e-02 -1.61639720e-01 -7.31545329e-01
4.83121663e-01 7.19253302e-01 1.09272704e-01 -9.51066375e-01
7.05012262e-01 2.00251922e-01 -6.21394873e-01 -6.88848078e-01
-1.07035196e+00 -4.86243337e-01 -5.20533621e-01 2.37159193e-01
4.74966437e-01 -1.07398641e+00 -3.24263483e-01 1.22681186e-01
-1.13208020e+00 -7.76079297e-01 -3.24196488e-01 4.88512158e-01
-2.67294735e-01 2.66566724e-01 -7.52947450e-01 -8.50943089e-01
-1.70394465e-01 -1.06736875e+00 9.92730796e-01 -2.03105316e-01
-6.62086964e-01 -1.20489562e+00 -1.62554801e-01 1.33842662e-01
6.20472252e-01 -5.97481668e-01 1.31177092e+00 -8.88611794e-01
-3.79974663e-01 6.11090995e-02 -1.27198964e-01 6.34861827e-01
3.79601344e-02 -2.66521603e-01 -1.17320645e+00 -3.46307427e-01
-2.09203467e-01 -4.30185825e-01 1.48712659e+00 3.08860242e-01
1.20197248e+00 -1.68429196e-01 -3.01673412e-01 4.46304083e-01
1.23664641e+00 -2.51460433e-01 2.56315023e-01 3.99763316e-01
5.72120309e-01 6.16136253e-01 5.55532873e-01 1.94219291e-01
4.41575348e-01 5.28186500e-01 -9.37049985e-02 4.96080741e-02
-3.41502339e-01 -6.86473310e-01 9.65190768e-01 1.23613226e+00
5.94879150e-01 1.00981118e-02 -8.67141485e-01 7.48631954e-01
-1.59229493e+00 -7.56270111e-01 2.27125525e-01 2.13912368e+00
1.37469029e+00 5.02800941e-01 -3.44978198e-02 1.32773206e-01
6.08753562e-01 1.45583257e-01 -1.84117287e-01 -6.66230083e-01
-1.36446252e-01 4.89749163e-01 7.24740982e-01 8.30692708e-01
-7.27844179e-01 1.34636819e+00 6.70817614e+00 1.17198241e+00
-1.05311704e+00 5.26277900e-01 5.01457155e-01 -3.41414720e-01
-7.29999721e-01 1.53959930e-01 -1.29735720e+00 6.00743651e-01
1.22206938e+00 -1.97832763e-01 2.95931250e-01 7.74036825e-01
2.91607022e-01 -1.38552427e-01 -1.27585649e+00 5.66300690e-01
-7.67541677e-02 -1.41684175e+00 3.18097055e-01 -2.94338137e-01
6.88594401e-01 3.01050305e-01 7.68330917e-02 8.86772096e-01
6.97560370e-01 -9.97126877e-01 7.95586467e-01 2.09962249e-01
6.77217603e-01 -7.25952983e-01 6.08177245e-01 5.11537015e-01
-1.08948851e+00 -1.15117066e-01 -6.87354267e-01 -4.22357023e-02
1.10984802e-01 6.74454927e-01 -1.23763192e+00 1.90301374e-01
2.32579380e-01 5.33785164e-01 -6.47807121e-01 8.69758964e-01
-4.53692675e-01 1.13772178e+00 -7.72097614e-03 -2.14487821e-01
2.08566681e-01 2.27153637e-02 2.89856434e-01 1.66231251e+00
-3.60864424e-03 -5.43625891e-01 2.67522216e-01 1.09254050e+00
-1.38150573e-01 -8.08242634e-02 -2.63077825e-01 -1.65880993e-01
1.00498378e+00 1.10305750e+00 -4.08343554e-01 -5.28668642e-01
-5.33792198e-01 8.15921783e-01 7.69513488e-01 4.13620561e-01
-6.54902160e-01 -3.70001286e-01 8.64379585e-01 1.37582153e-01
3.17414224e-01 -4.35995489e-01 -6.10978961e-01 -1.17300725e+00
-2.28810206e-01 -6.01990163e-01 4.04931873e-01 -4.99628276e-01
-1.61706972e+00 3.93911839e-01 7.76808411e-02 -9.80641246e-01
-6.00801557e-02 -6.24700069e-01 -7.59193480e-01 1.18369949e+00
-2.00098777e+00 -9.48595643e-01 1.72311485e-01 1.01490274e-01
7.95566797e-01 -1.89543754e-01 8.18432629e-01 2.35485062e-01
-6.17457569e-01 1.09866166e+00 3.54481936e-02 2.27681994e-01
1.07411599e+00 -1.24143910e+00 7.69262850e-01 8.23987186e-01
2.12374002e-01 9.73552167e-01 4.36460704e-01 -5.66576600e-01
-9.71544862e-01 -1.48565054e+00 1.36702919e+00 -4.66599375e-01
7.75742650e-01 -8.62187684e-01 -1.21507585e+00 6.88422680e-01
2.95044810e-01 -3.23318243e-01 8.35707545e-01 2.46392637e-01
-7.36765325e-01 -1.73890501e-01 -8.80140185e-01 7.04950333e-01
7.96324432e-01 -7.18466520e-01 -9.74850118e-01 2.88673878e-01
1.07865036e+00 2.82825064e-02 -5.63038886e-01 2.51951128e-01
3.05462442e-02 -5.87823808e-01 9.58893478e-01 -8.11445534e-01
3.23753297e-01 8.66455361e-02 -1.98881581e-01 -1.77333283e+00
-3.05064380e-01 -4.86728132e-01 4.28017750e-02 1.55860865e+00
5.26880860e-01 -7.34965444e-01 6.54502392e-01 3.11211765e-01
-1.57273874e-01 -7.97618985e-01 -6.86807632e-01 -1.21324015e+00
5.53713083e-01 -5.54489315e-01 5.63748121e-01 1.05018091e+00
3.74979973e-01 2.13806078e-01 -2.58706696e-02 -1.29779994e-01
3.60849708e-01 -1.80194750e-01 2.95696735e-01 -8.19608152e-01
-3.49005789e-01 -6.17820919e-01 8.41472000e-02 -1.28358209e+00
5.94824791e-01 -1.59668148e+00 2.12202832e-01 -1.26684320e+00
2.02688813e-01 -7.00289667e-01 -5.80463171e-01 5.88548243e-01
-5.62995791e-01 6.12403452e-02 1.61506996e-01 -6.76597953e-02
-3.16070229e-01 6.57549918e-01 7.46686876e-01 -8.23708065e-03
6.04119226e-02 -2.60009438e-01 -7.77816057e-01 5.17250836e-01
8.72620225e-01 -6.38622642e-01 -3.65643024e-01 -7.72339046e-01
4.11788933e-02 -5.42852044e-01 1.43802688e-01 -5.46190619e-01
-5.17058708e-02 -1.31116658e-01 1.82773143e-01 -6.14864826e-01
1.98284164e-01 -3.03957194e-01 -8.25770080e-01 4.77827579e-01
-9.97320652e-01 4.93297577e-02 3.28810573e-01 1.56159699e-01
-1.67903364e-01 -6.60787165e-01 9.23307061e-01 -1.17490612e-01
-5.93639791e-01 -1.56034222e-02 -6.08035803e-01 3.88885617e-01
4.44370031e-01 2.78103370e-02 -2.34060913e-01 -6.95583522e-02
-4.12211210e-01 5.01777470e-01 2.51833558e-01 6.58527136e-01
5.51853538e-01 -1.48560750e+00 -7.09830642e-01 3.47592592e-01
2.32407764e-01 -1.57252371e-01 -8.75595808e-02 8.05958152e-01
-4.41642739e-02 4.97358114e-01 2.47072920e-01 -3.79986614e-01
-1.00846076e+00 5.06077349e-01 2.48819321e-01 -2.56491631e-01
-2.25883558e-01 1.42029786e+00 8.30883309e-02 -7.31932104e-01
3.23420167e-01 -5.60404301e-01 1.04036465e-01 1.56690523e-01
4.65525270e-01 2.71960765e-01 2.66292751e-01 -1.70662493e-01
-4.00748640e-01 1.33476272e-01 -3.74513805e-01 -2.78963894e-01
1.29480660e+00 -2.53875852e-02 -1.10393301e-01 6.90848470e-01
1.33298349e+00 2.86254764e-01 -1.00325465e+00 -6.63841486e-01
5.86380720e-01 -3.41489583e-01 -2.31747422e-02 -5.63190341e-01
-6.17306471e-01 1.18874609e+00 -6.05400205e-02 -1.44501194e-01
7.13035464e-01 -7.85979852e-02 9.49390411e-01 2.30095118e-01
1.99779004e-01 -1.21446168e+00 2.46783420e-01 7.79996991e-01
6.49389029e-01 -1.07636380e+00 -3.23353797e-01 -3.67775291e-01
-8.27864528e-01 9.31810319e-01 7.43504167e-01 -1.81912228e-01
5.24680674e-01 5.63771367e-01 3.37167680e-01 3.50869328e-01
-1.21520722e+00 -1.72307342e-01 2.02197030e-01 4.28950995e-01
8.56558740e-01 1.74042191e-02 -3.38474423e-01 7.39943385e-01
-3.80290151e-01 -2.79909670e-01 3.25845987e-01 8.50468874e-01
-6.82087719e-01 -1.53148520e+00 -3.68660271e-01 4.60015029e-01
-2.26483837e-01 -6.86566710e-01 -2.11100936e-01 3.97960722e-01
1.52316615e-01 9.68616843e-01 2.30249107e-01 -3.95825177e-01
5.37291430e-02 5.91338038e-01 4.68752205e-01 -1.08411658e+00
-7.34143615e-01 -9.51305702e-02 1.56243220e-01 -2.61837721e-01
-7.28347376e-02 -5.47122836e-01 -1.50668120e+00 -7.56946951e-02
-4.26891297e-01 3.61742139e-01 4.46106195e-01 9.71402764e-01
3.15355986e-01 6.25307977e-01 6.41316116e-01 -5.88629961e-01
-1.16351938e+00 -1.23594999e+00 -3.84273022e-01 4.61039186e-01
4.56310898e-01 -6.47955656e-01 -3.63873988e-01 -1.32528394e-01]
|
[10.877740859985352, 8.680668830871582]
|
77dabe9f-7ca5-4d19-8971-209c9a5a603f
|
deep-unsupervised-multi-view-detection-of
|
1807.09715
| null |
http://arxiv.org/abs/1807.09715v1
|
http://arxiv.org/pdf/1807.09715v1.pdf
|
Deep Unsupervised Multi-View Detection of Video Game Stream Highlights
|
We consider the problem of automatic highlight-detection in video game
streams. Currently, the vast majority of highlight-detection systems for games
are triggered by the occurrence of hard-coded game events (e.g., score change,
end-game), while most advanced tools and techniques are based on detection of
highlights via visual analysis of game footage. We argue that in the context of
game streaming, events that may constitute highlights are not only dependent on
game footage, but also on social signals that are conveyed by the streamer
during the play session (e.g., when interacting with viewers, or when
commenting and reacting to the game). In this light, we present a multi-view
unsupervised deep learning methodology for novelty-based highlight detection.
The method jointly analyses both game footage and social signals such as the
players facial expressions and speech, and shows promising results for
generating highlights on streams of popular games such as Player Unknown's
Battlegrounds.
|
['Charles Ringer', 'Mihalis A. Nicolaou']
|
2018-07-25
| null | null | null | null |
['highlight-detection']
|
['computer-vision']
|
[ 4.20712709e-01 -6.69418797e-02 3.77372891e-01 -5.14246859e-02
-7.59936690e-01 -7.44212866e-01 4.90816563e-01 8.33149552e-01
-3.56926888e-01 1.79836348e-01 4.55032259e-01 2.14481890e-01
2.81321436e-01 -8.43356371e-01 -4.06008631e-01 -5.38530409e-01
-5.32630146e-01 -2.62834907e-01 5.35374165e-01 -6.11034811e-01
2.99949110e-01 2.18732983e-01 -1.88151455e+00 7.68046439e-01
-2.18332633e-01 1.03525889e+00 -1.12603739e-01 1.25973368e+00
-8.34503099e-02 1.34442484e+00 -1.08914900e+00 -3.72684032e-01
5.12095653e-02 -7.00365841e-01 -1.34420276e-01 -1.07468970e-01
5.44026569e-02 -3.13534379e-01 -3.19651961e-01 9.41711426e-01
7.61874259e-01 2.87377924e-01 2.37881169e-01 -1.39588201e+00
5.34759045e-01 7.63049901e-01 -6.16854906e-01 5.30517399e-01
8.25403035e-01 1.28942607e-02 1.15863276e+00 -6.80677950e-01
8.44613194e-01 6.83602691e-01 4.71957773e-01 2.56262600e-01
-8.06618333e-01 -7.19715416e-01 9.09226984e-02 2.19696805e-01
-1.06306660e+00 -6.16623521e-01 1.18175387e+00 -5.57944119e-01
3.46409947e-01 5.60257733e-01 8.13205123e-01 1.28922796e+00
4.21409309e-02 8.23826551e-01 5.65267384e-01 -3.53279412e-01
4.87635970e-01 -1.97017595e-01 -5.14316320e-01 1.96843192e-01
-6.12049460e-01 4.75747101e-02 -1.30725873e+00 -4.50969607e-01
4.24941361e-01 -2.92722434e-01 -1.18335187e-01 1.13934070e-01
-1.13633990e+00 8.99060667e-01 -1.86817288e-01 2.76716501e-02
-7.50186443e-01 1.08770296e-01 9.03202057e-01 3.33090425e-01
9.90720689e-01 3.09573233e-01 1.34519786e-01 -1.09662843e+00
-1.29554355e+00 5.08433938e-01 6.96639180e-01 4.70408261e-01
3.09631884e-01 1.77631795e-01 -3.69784236e-01 5.70047200e-01
-3.74347121e-02 9.72111896e-02 1.97186619e-01 -5.03478527e-01
2.20609203e-01 3.51846695e-01 -5.82276732e-02 -1.40847719e+00
-5.36340475e-01 -2.72361904e-01 -1.74760520e-01 5.09360671e-01
3.34151834e-01 -5.41671395e-01 -1.98300824e-01 1.63687193e+00
4.39177096e-01 4.06513751e-01 -4.20612097e-01 9.55437064e-01
1.34005010e+00 5.32638848e-01 7.16934279e-02 -3.04212093e-01
1.49007940e+00 -1.86961129e-01 -6.92090929e-01 -9.27742478e-03
3.00188482e-01 -8.65589440e-01 7.11946547e-01 5.22055686e-01
-1.25507355e+00 -4.45128798e-01 -7.59389043e-01 2.89621025e-01
7.62318596e-02 -3.88095677e-01 2.87245721e-01 4.56128001e-01
-6.63582027e-01 4.99455482e-01 -3.26260209e-01 -3.65316540e-01
9.97536629e-02 -5.57778068e-02 -2.25398481e-01 6.63778603e-01
-9.74925518e-01 8.49211589e-02 -3.34880836e-02 -1.45911828e-01
-9.94806051e-01 -6.08684003e-01 -5.58629632e-01 2.40694135e-02
4.88376975e-01 1.50284037e-01 1.41616678e+00 -1.40315616e+00
-1.62035596e+00 1.11610973e+00 5.22044264e-02 -3.34030658e-01
8.15524578e-01 -3.24863821e-01 -5.27362525e-01 4.57769841e-01
8.88955817e-02 2.41420478e-01 1.02669799e+00 -9.23315644e-01
-9.34989154e-01 6.98358119e-02 5.75690508e-01 2.46120572e-01
-1.79249570e-01 9.01522815e-01 -3.50204617e-01 -6.09479189e-01
-2.91103244e-01 -5.79498768e-01 1.46274688e-02 -1.58198908e-01
-6.54478312e-01 1.96877867e-01 9.12391663e-01 -5.70815206e-01
1.24528086e+00 -2.60689688e+00 -7.84024894e-02 4.45474476e-01
5.04643083e-01 1.06950849e-01 4.32145633e-02 7.41172612e-01
-2.59113669e-01 -2.62738198e-01 3.40084463e-01 -3.42714399e-01
-7.35487193e-02 -3.88502210e-01 -4.81398165e-01 4.71161425e-01
4.35068980e-02 4.71203387e-01 -1.19346893e+00 -1.96175456e-01
1.83058321e-01 2.90563762e-01 -4.91716683e-01 3.52567524e-01
-1.66237682e-01 5.31849563e-01 -1.60341829e-01 4.37002778e-01
2.30074599e-01 2.98462451e-01 -1.38975322e-01 9.31292474e-02
-4.69727218e-01 3.59281957e-01 -1.17381299e+00 1.45109725e+00
-1.41132414e-01 1.26212513e+00 1.49767041e-01 -7.13465631e-01
1.03061497e+00 4.30613011e-01 6.29237950e-01 -6.50433362e-01
3.36894989e-01 -2.55950511e-01 -4.25269119e-02 -5.57319403e-01
9.37441230e-01 -7.50162452e-02 -2.95471013e-01 7.08453953e-01
-3.19947422e-01 1.53610289e-01 2.52708137e-01 2.88298696e-01
1.59746873e+00 6.97408840e-02 2.62717545e-01 4.45401847e-01
3.45516438e-03 -1.23385303e-01 2.48292804e-01 6.95541084e-01
-3.43705893e-01 8.93997371e-01 1.08818972e+00 -1.77286878e-01
-7.50994802e-01 -9.58132923e-01 7.29598463e-01 1.79102433e+00
-5.06548164e-03 -9.85831738e-01 -7.05649853e-01 -1.77758142e-01
-4.54190612e-01 2.45964646e-01 -5.87291121e-01 -2.80801624e-01
-4.61355090e-01 -2.94475883e-01 4.94574785e-01 4.14755680e-02
1.55594096e-01 -1.53299749e+00 -1.26335704e+00 6.55273259e-01
-3.67067575e-01 -1.18841505e+00 -2.99272954e-01 4.31438237e-02
-1.94777831e-01 -1.17024195e+00 -4.68522906e-01 -3.84522676e-01
4.13431935e-02 2.11329266e-01 1.17382443e+00 -8.75296369e-02
-4.75951761e-01 7.18850911e-01 -7.27075934e-01 -7.26669133e-01
-4.18026388e-01 -2.92888522e-01 -2.43017465e-01 4.92382586e-01
2.48474747e-01 -7.37017393e-01 -3.56481791e-01 -9.60864574e-02
-8.61721814e-01 1.60448134e-01 -3.29664275e-02 2.01952577e-01
3.36780101e-01 8.29976201e-02 1.82013169e-01 -7.10115790e-01
1.04092681e+00 -7.24680722e-01 -1.59102485e-01 -4.28419888e-01
5.64864516e-01 -9.00688767e-01 3.64888549e-01 -7.01754332e-01
-6.68967009e-01 -2.83170551e-01 -1.07460737e-03 -4.80603874e-01
-3.66836667e-01 6.15664124e-01 1.83544427e-01 2.17710435e-01
9.17342365e-01 8.47381912e-03 -2.36183494e-01 -4.53396030e-02
1.72810897e-01 5.66297412e-01 6.58756316e-01 -1.69019133e-01
6.92262471e-01 8.64982426e-01 -9.17271227e-02 -1.24508679e+00
-3.96368682e-01 -6.76451206e-01 -8.23027343e-02 -1.32843113e+00
5.39631128e-01 -1.13377547e+00 -1.02955258e+00 5.90071917e-01
-1.05966556e+00 -2.72973895e-01 -6.49460196e-01 3.15396875e-01
-4.58686471e-01 1.25491768e-01 -6.00867391e-01 -1.11657870e+00
-1.30902246e-01 -5.87843776e-01 1.12888730e+00 3.51612628e-01
-9.00564730e-01 -5.95843434e-01 2.61278987e-01 1.19341522e-01
2.45646209e-01 1.03361475e+00 3.35922778e-01 -5.65856338e-01
-1.24889620e-01 -5.89434206e-01 1.37948573e-01 -2.49118507e-01
2.17250675e-01 4.47907478e-01 -1.35647893e+00 4.07702550e-02
-2.29845867e-01 -6.15282916e-02 2.15434760e-01 5.47935843e-01
7.36306310e-01 9.31347813e-03 -1.42900168e-03 2.47853324e-01
7.69854426e-01 4.31980669e-01 5.56134462e-01 2.74317026e-01
3.85127991e-01 8.88217688e-01 6.68158352e-01 1.28347790e+00
5.46475835e-02 7.77853549e-01 8.02740097e-01 -3.61848146e-01
1.40313298e-01 -2.56645799e-01 6.94488764e-01 4.78771031e-01
-3.13493907e-01 -3.38715106e-01 -5.91012001e-01 4.85462725e-01
-1.87490392e+00 -1.43146098e+00 -5.05724430e-01 2.10817623e+00
4.47625250e-01 4.73611057e-01 7.92239964e-01 3.64529431e-01
9.73302484e-01 5.60888529e-01 -2.60880172e-01 -6.83242321e-01
-2.01943472e-01 3.64641368e-01 -3.94718386e-02 2.11359859e-01
-1.11842847e+00 6.68485582e-01 5.72986555e+00 7.06580520e-01
-1.27269924e+00 1.71547621e-01 4.76228476e-01 -9.91127014e-01
-2.18595698e-01 -2.63261408e-01 -7.30705336e-02 5.45879900e-01
7.09859908e-01 -1.56081304e-01 2.52929300e-01 7.75951564e-01
8.64089668e-01 -4.88349408e-01 -8.94355059e-01 1.19420683e+00
1.81586683e-01 -1.26322675e+00 -6.05858982e-01 8.38565454e-02
1.67527661e-01 -1.57769457e-01 4.32534963e-02 1.85248837e-01
1.44614540e-02 -7.60537565e-01 1.19210851e+00 2.29442477e-01
6.08028233e-01 -9.56951141e-01 3.46945584e-01 6.16749525e-02
-1.32643187e+00 1.88982889e-01 1.89190507e-01 -6.14484191e-01
3.60146552e-01 6.97297335e-01 -5.96589923e-01 -2.90461872e-02
6.68923199e-01 6.51831388e-01 -2.32781872e-01 1.25564408e+00
-4.60781187e-01 8.66552353e-01 -2.28890076e-01 -1.30282968e-01
2.63623912e-02 -7.07161874e-02 1.12910807e+00 1.41897416e+00
2.46957019e-01 5.55490293e-02 4.59968559e-02 6.46724224e-01
1.86569884e-01 3.26978922e-01 -6.98808074e-01 -1.08406261e-01
6.93731606e-02 1.42517686e+00 -1.07555425e+00 -9.73202381e-03
-2.52126575e-01 8.69269490e-01 -2.00550884e-01 3.26731235e-01
-9.80756283e-01 -4.94391352e-01 9.81105506e-01 3.42788935e-01
1.50502801e-01 -1.60187259e-01 1.19122237e-01 -7.29925931e-01
2.26013482e-01 -8.64666164e-01 2.14447409e-01 -9.52339709e-01
-8.88716757e-01 5.19330978e-01 -3.14660937e-01 -1.53542697e+00
-3.94393921e-01 -1.11772001e-01 -1.59078395e+00 4.11512911e-01
-8.09080362e-01 -5.37961364e-01 -4.65364546e-01 6.38427079e-01
3.69380563e-01 -5.41555844e-02 6.62472785e-01 3.16195071e-01
-2.12928981e-01 2.65305012e-01 -1.78405002e-01 3.39412123e-01
6.56056046e-01 -9.81404364e-01 4.71563339e-01 9.04294729e-01
5.04510641e-01 -2.21854001e-01 1.11503363e+00 -5.55441618e-01
-1.00987470e+00 -6.07478201e-01 6.77534580e-01 6.22964613e-02
7.57496953e-01 -8.25371265e-01 -4.69514191e-01 1.10634729e-01
-5.26324473e-02 -2.09683076e-01 1.10168338e+00 3.12162787e-01
-3.05921696e-02 1.27996117e-01 -6.95831239e-01 8.01423013e-01
8.30352962e-01 -9.59902048e-01 -2.70871937e-01 2.25704268e-01
1.11699350e-01 -6.61971509e-01 -1.88844472e-01 -2.06260204e-01
5.54215550e-01 -1.38363910e+00 5.90414464e-01 -6.54330313e-01
6.97618961e-01 -3.41423482e-01 -1.32575864e-03 -1.26169026e+00
5.78565896e-02 -1.20436740e+00 3.42652053e-02 1.47818053e+00
-4.22076322e-03 2.15811774e-01 9.46828485e-01 2.94035256e-01
2.12729841e-01 -1.07788950e-01 -1.09293079e+00 -1.01168111e-01
-8.08466971e-01 -1.19832218e+00 2.26654857e-01 9.00475025e-01
4.97178048e-01 1.20506056e-01 -7.01481402e-01 6.24591559e-02
2.77508378e-01 -1.38665929e-01 1.12659633e+00 -1.12587881e+00
-5.34798026e-01 -4.92059857e-01 -9.47643220e-01 -4.24418360e-01
-3.28773677e-01 -3.08964103e-01 2.28571936e-01 -1.02002978e+00
9.12719816e-02 2.14197159e-01 -1.61196977e-01 3.78120691e-02
-3.83415557e-02 4.09847766e-01 4.34677988e-01 -8.90709609e-02
-8.58285844e-01 1.85505852e-01 6.82836175e-01 9.28893760e-02
-4.07355726e-01 2.21550211e-01 -3.46379787e-01 7.56193221e-01
6.28968358e-01 -5.25657594e-01 -2.12363929e-01 3.31223667e-01
9.86859262e-01 1.87339187e-01 4.80234057e-01 -1.11845744e+00
2.73738921e-01 -2.16241903e-03 -5.93131632e-02 -3.79032910e-01
5.58128536e-01 -3.04181367e-01 1.04741327e-01 7.93737173e-02
-4.41678852e-01 4.01842818e-02 3.36780816e-01 6.09602571e-01
-5.42395353e-01 -3.48696229e-03 3.39582354e-01 -6.20487779e-02
-6.21174514e-01 1.23909920e-01 -1.23181105e+00 2.99373955e-01
1.00337446e+00 -4.57280666e-01 -1.88789755e-01 -1.18346393e+00
-9.25425708e-01 -2.17066109e-01 9.11998972e-02 5.45772552e-01
5.53674102e-01 -1.12259328e+00 -8.09618831e-01 9.66001824e-02
2.05361485e-01 -2.32063770e-01 3.90132219e-01 6.36716783e-01
-4.31005329e-01 -6.54069901e-01 -1.85792580e-01 -5.78866601e-01
-1.74857593e+00 -3.59725319e-02 5.98394200e-02 9.66976434e-02
-8.05473030e-01 1.14290202e+00 4.17402714e-01 5.04195809e-01
4.48942631e-01 -2.29425997e-01 -6.11826301e-01 8.30117464e-01
9.31007981e-01 2.85815269e-01 -1.28386971e-02 -7.85235345e-01
-3.57913971e-01 1.80446714e-01 1.28116518e-01 -4.31849748e-01
1.38543844e+00 -3.05319950e-02 2.39509791e-01 8.89793098e-01
8.11981797e-01 4.62083966e-01 -1.36856449e+00 -1.88738063e-01
-1.09552220e-01 -3.22810411e-01 -4.50995341e-02 -2.73788244e-01
-1.08331382e+00 8.90201569e-01 5.83210707e-01 7.29576111e-01
1.08544576e+00 -8.83262828e-02 6.24076486e-01 -1.64102897e-01
5.01043320e-01 -1.37712038e+00 3.60849917e-01 4.47298586e-01
8.24252248e-01 -8.16041589e-01 -2.03311324e-01 -2.17050940e-01
-6.84051335e-01 1.33384609e+00 1.76154047e-01 -6.28933012e-02
5.72446048e-01 5.00298381e-01 4.72662807e-01 -4.57216024e-01
-7.74266601e-01 -3.02273661e-01 -8.50075036e-02 6.38792694e-01
4.66387868e-01 2.44964167e-01 1.46637663e-01 6.08682096e-01
-5.29271603e-01 -2.55789757e-01 1.04375112e+00 1.06759453e+00
-3.53812307e-01 -4.82257932e-01 -5.31906843e-01 2.55668610e-01
-7.82972395e-01 -8.49890113e-02 -8.71602476e-01 4.80902940e-01
3.04545224e-01 1.35209978e+00 4.13068473e-01 -5.13907611e-01
4.81286824e-01 -5.29722795e-02 1.39360785e-01 -7.53886402e-01
-1.23234046e+00 3.00338626e-01 4.41206336e-01 -8.02836895e-01
-5.23881376e-01 -8.95197630e-01 -9.58883107e-01 -4.16163087e-01
-8.65043476e-02 1.63033187e-01 6.06133342e-01 9.25689995e-01
1.66781753e-01 7.85567760e-01 6.42630756e-01 -1.07055497e+00
6.07538283e-01 -7.66637444e-01 -8.31254721e-01 6.64358675e-01
3.23195040e-01 -4.61816549e-01 -3.78394961e-01 1.43443704e-01]
|
[10.148282051086426, 0.47637873888015747]
|
38e6e74f-4cfc-45f0-8410-bfbf80eefbff
|
label-embedding-by-johnson-lindenstrauss
|
2305.19470
| null |
https://arxiv.org/abs/2305.19470v2
|
https://arxiv.org/pdf/2305.19470v2.pdf
|
Label Embedding by Johnson-Lindenstrauss Matrices
|
We present a simple and scalable framework for extreme multiclass classification based on Johnson-Lindenstrauss matrices (JLMs). Using the columns of a JLM to embed the labels, a $C$-class classification problem is transformed into a regression problem with $\cO(\log C)$ output dimension. We derive an excess risk bound, revealing a tradeoff between computational efficiency and prediction accuracy, and further show that under the Massart noise condition, the penalty for dimension reduction vanishes. Our approach is easily parallelizable, and experimental results demonstrate its effectiveness and scalability in large-scale applications.
|
['Clayton Scott', 'Jianxin Zhang']
|
2023-05-31
| null | null | null | null |
['dimensionality-reduction']
|
['methodology']
|
[ 1.23136066e-01 -2.57135555e-02 -4.48899567e-01 -7.27443278e-01
-1.00523961e+00 -4.59564567e-01 -1.03316225e-01 6.93191886e-02
-3.40402424e-01 7.14415789e-01 -4.61144745e-01 -6.65932000e-01
-4.26584870e-01 -5.85921884e-01 -5.33226311e-01 -8.15917373e-01
-4.08844620e-01 4.23393697e-01 -1.75693631e-01 1.45298824e-01
3.13818246e-01 3.26493472e-01 -1.46883726e+00 4.50941056e-01
7.01505482e-01 1.66968513e+00 -4.09747809e-01 7.60151625e-01
1.18464492e-01 8.71549070e-01 -2.86561519e-01 -8.52511406e-01
7.46048152e-01 -3.56527507e-01 -9.39240873e-01 -4.80517559e-03
5.55723548e-01 6.48082271e-02 -2.34181359e-01 1.20461226e+00
2.00226605e-01 2.43477270e-01 1.03737652e+00 -1.53495729e+00
-6.28239930e-01 5.92793882e-01 -8.76660049e-01 -9.27367061e-03
1.13354325e-01 -1.81466147e-01 1.64489460e+00 -1.01228380e+00
1.90103024e-01 1.21464884e+00 9.86178815e-01 4.50586528e-01
-1.64145613e+00 -9.73422647e-01 1.41312391e-01 8.79523158e-02
-1.61944520e+00 -1.87343478e-01 4.23826158e-01 -4.18503672e-01
8.24002922e-01 7.20238268e-01 2.70826459e-01 4.93582696e-01
2.44031265e-01 7.24239171e-01 1.29379666e+00 -4.78762865e-01
4.40580636e-01 5.04773140e-01 4.16981757e-01 1.05246842e+00
2.79089451e-01 6.28909841e-02 -4.03815180e-01 -6.53378487e-01
1.07992172e-01 9.55143273e-02 7.60670528e-02 -6.25261009e-01
-2.43109271e-01 1.06552446e+00 3.96023959e-01 -1.87195539e-01
3.80567461e-01 4.37756956e-01 3.61437201e-01 6.83274209e-01
5.37488461e-01 5.66245206e-02 -4.13352847e-01 1.78603232e-01
-7.20269263e-01 -1.43454611e-01 1.05142474e+00 1.06721461e+00
6.84936404e-01 -2.47356385e-01 2.95366317e-01 1.03030562e+00
1.08240589e-01 4.52840239e-01 2.52653301e-01 -1.25217319e+00
4.13171679e-01 4.42300558e-01 -7.01342970e-02 -8.99561167e-01
-4.05126154e-01 -5.75077116e-01 -9.99032378e-01 2.62420595e-01
1.86950609e-01 1.38998017e-01 -4.41250294e-01 1.78097236e+00
1.36120737e-01 -4.04127017e-02 -6.97340742e-02 5.09982169e-01
9.32265446e-03 4.15410638e-01 9.31734294e-02 -4.22401279e-01
9.87866819e-01 -7.85948217e-01 -2.95691609e-01 -4.31029588e-01
1.10120451e+00 -2.03344464e-01 1.25287974e+00 4.94640142e-01
-1.03246069e+00 -8.44127685e-02 -1.18031263e+00 4.26018564e-03
-2.62654275e-01 -9.70428735e-02 9.80689704e-01 1.07834506e+00
-8.05514097e-01 7.14660764e-01 -5.54277301e-01 2.04976559e-01
5.57915211e-01 6.26791000e-01 -4.07300442e-01 -4.88827638e-02
-7.24531949e-01 4.80162352e-01 1.90992251e-01 -6.95388857e-03
-4.20134693e-01 -5.47427177e-01 -6.40129447e-01 1.22812659e-01
2.25202203e-01 -3.48973334e-01 9.62209165e-01 -6.37863755e-01
-1.10223162e+00 9.24549043e-01 2.66631935e-02 -4.43907350e-01
4.81658816e-01 3.68416868e-02 -2.43855178e-01 1.95744827e-01
-8.33827779e-02 1.29695639e-01 7.83340812e-01 -1.22534728e+00
-9.92597759e-01 -6.50494576e-01 -2.55788177e-01 -1.81747317e-01
-7.88954496e-01 -1.05457671e-01 -1.02491982e-01 -5.48859894e-01
3.87946397e-01 -1.18322337e+00 -5.22867858e-01 1.05995797e-01
-3.47471923e-01 -9.90471244e-02 4.28561211e-01 -3.90088707e-01
1.49108362e+00 -2.46744490e+00 -6.03803210e-02 8.14744711e-01
4.75832015e-01 -1.30405307e-01 -6.91164425e-03 2.14870766e-01
-5.14657386e-02 4.01686996e-01 -4.28765655e-01 -3.34804446e-01
2.53935099e-01 1.12791978e-01 -5.28070688e-01 6.58842385e-01
-1.65165395e-01 5.30071497e-01 -5.15274584e-01 -4.65230197e-01
-2.74733812e-01 -9.82828438e-02 -9.30730820e-01 -7.28399009e-02
4.68480885e-02 -2.13618204e-01 -2.01815963e-01 6.25204384e-01
6.88387096e-01 -7.64503837e-01 4.24277008e-01 2.79062718e-01
6.53560996e-01 -5.11226468e-02 -1.45521247e+00 1.08775198e+00
-6.15212321e-01 2.48233855e-01 2.34671697e-01 -1.37086487e+00
6.05844617e-01 -1.41194656e-01 4.42568898e-01 -1.00803636e-01
1.41042814e-01 4.76460397e-01 -2.82975703e-01 -2.88120080e-02
1.90620333e-01 -5.35562634e-01 -4.46779191e-01 6.51133537e-01
-1.20119371e-01 -3.68267782e-02 9.15547013e-02 1.26555502e-01
8.96299303e-01 -5.02995133e-01 1.92503452e-01 -4.14828390e-01
3.25379491e-01 -2.26428926e-01 6.43922508e-01 9.89248276e-01
-3.12619835e-01 1.04949526e-01 7.32556701e-01 -2.72486657e-01
-7.43489265e-01 -1.16604066e+00 -5.39343834e-01 1.39426458e+00
-5.26593551e-02 -4.66977626e-01 -6.63608611e-01 -9.43978488e-01
5.84998906e-01 6.11908495e-01 -7.33586729e-01 -3.98818046e-01
-1.77180588e-01 -1.23979378e+00 3.58106971e-01 6.64287686e-01
1.75175577e-01 -2.78747320e-01 -1.24254487e-01 -2.58314908e-01
2.48044163e-01 -9.55216348e-01 -4.75962758e-01 5.48253477e-01
-1.00571644e+00 -1.05074680e+00 -1.75617874e-01 -1.05034435e+00
5.78437388e-01 1.97507981e-02 8.95310521e-01 7.71023631e-02
-7.22246885e-01 2.27943882e-01 7.48982951e-02 -2.29746960e-02
-3.62860590e-01 -7.18766749e-02 3.69867057e-01 6.42246455e-02
3.82239044e-01 -7.37071395e-01 -7.28779972e-01 3.77209395e-01
-6.00031018e-01 -4.51138467e-01 5.74223280e-01 1.09769523e+00
6.93528354e-01 1.49787381e-01 6.26432955e-01 -1.23950219e+00
4.74692494e-01 -4.36270088e-01 -8.09336483e-01 4.47767705e-01
-1.36957538e+00 1.62728280e-01 8.48834753e-01 -5.63398182e-01
-4.61027324e-01 1.64364517e-01 5.20132966e-02 -3.04755807e-01
4.28621739e-01 2.74595410e-01 8.89982730e-02 -3.59756023e-01
7.77251601e-01 1.05806440e-01 -8.83957669e-02 -4.90533978e-01
3.97234917e-01 1.05242205e+00 1.27784297e-01 -7.87948966e-01
6.34680092e-01 4.56460625e-01 5.26511192e-01 -5.12283266e-01
-1.08692181e+00 -3.04144561e-01 -3.91201258e-01 2.05423057e-01
3.42406243e-01 -6.92578554e-01 -1.30608737e+00 -1.82502314e-01
-2.19562873e-01 -2.51179248e-01 -4.96258020e-01 3.74642104e-01
-8.46366346e-01 4.01773781e-01 -1.05538416e+00 -1.13505208e+00
-2.35525146e-01 -9.47363377e-01 6.42878175e-01 -3.71236384e-01
1.18599042e-01 -8.44155431e-01 -2.06225991e-01 3.91266197e-01
-6.19319156e-02 -1.26576692e-01 1.59121716e+00 -6.32739604e-01
-4.45593983e-01 -7.50149965e-01 -3.40185136e-01 6.54695451e-01
-4.20671344e-01 -3.82043421e-01 -9.37264383e-01 -6.44754827e-01
-2.08287816e-02 -7.15230942e-01 9.59620357e-01 5.73267974e-02
1.82385039e+00 -6.61962390e-01 -4.29592073e-01 8.40329826e-01
1.47147202e+00 -1.03319153e-01 9.48100835e-02 -9.64524373e-02
2.87135839e-01 2.97360659e-01 3.46290886e-01 5.51631987e-01
5.61488867e-02 1.89732105e-01 4.10555601e-02 1.36561602e-01
1.41821116e-01 -2.30018348e-01 2.06588760e-01 9.54308152e-01
3.65839332e-01 3.29198875e-02 -8.14717352e-01 7.94399530e-02
-1.60655630e+00 -7.54496753e-01 1.43205538e-01 2.46867609e+00
9.11160588e-01 2.97645599e-01 9.44495872e-02 4.73855495e-01
4.85260487e-01 -8.70475546e-02 -6.66678488e-01 -7.16548741e-01
-1.00758046e-01 4.47340399e-01 6.92825556e-01 6.09584630e-01
-1.09162712e+00 6.61433280e-01 8.22907162e+00 1.15202761e+00
-4.93486971e-01 1.10434987e-01 1.12081432e+00 -4.43292022e-01
-6.83815628e-02 -5.94602674e-02 -8.12743425e-01 2.74116397e-01
1.14379263e+00 -1.54534504e-01 7.45547235e-01 1.27522159e+00
-5.96391261e-01 1.33992910e-01 -1.51135302e+00 1.14486325e+00
-2.48829052e-02 -9.73444700e-01 -1.07862093e-01 2.29988366e-01
6.68127060e-01 -1.58879697e-01 6.59848988e-01 5.12589097e-01
6.52067602e-01 -9.72053349e-01 2.98428357e-01 -5.79324439e-02
1.25677538e+00 -9.72844064e-01 3.04261178e-01 5.13827801e-01
-1.21757305e+00 -9.44038689e-01 -6.25012338e-01 -8.82706046e-02
-4.85746533e-01 6.07463062e-01 -4.05912906e-01 -9.70324222e-03
7.50904858e-01 2.55022049e-01 -5.04548788e-01 5.60621381e-01
2.69876122e-01 4.57002580e-01 -4.90276247e-01 -1.62619874e-01
-1.03731684e-01 -4.23547179e-01 2.60890964e-02 1.36745656e+00
2.16378793e-01 3.02247673e-01 4.20292616e-01 5.05098879e-01
-3.81896406e-01 4.39154893e-01 -4.61078823e-01 2.59568132e-02
3.69695038e-01 9.01282728e-01 -6.26325786e-01 -3.27885926e-01
-3.72556448e-01 9.14077997e-01 7.48274505e-01 2.68982440e-01
-5.00020862e-01 -6.00805640e-01 8.38614583e-01 -3.23588312e-01
2.29997024e-01 2.15000566e-02 -6.93872869e-01 -1.17826498e+00
1.40885115e-01 -7.92035878e-01 1.03037727e+00 -1.66105181e-01
-1.68360841e+00 4.16739196e-01 -3.33730251e-01 -1.16707337e+00
-1.40427068e-01 -1.08448195e+00 1.32570222e-01 5.78627825e-01
-1.04562759e+00 -6.21086955e-01 3.37544829e-01 3.96665752e-01
-5.19366302e-02 -1.17443241e-01 1.11988664e+00 3.00896913e-01
-8.08310509e-01 1.41984332e+00 7.22604275e-01 9.37568247e-02
2.99126536e-01 -1.41309154e+00 -1.12192124e-01 4.51342374e-01
2.39286214e-01 5.02277434e-01 4.96097088e-01 -2.98763603e-01
-1.50875020e+00 -9.36759889e-01 6.65392876e-01 -4.22529876e-01
8.70658934e-01 -5.63039362e-01 -5.40129542e-01 8.71652544e-01
-5.94868004e-01 4.26990032e-01 1.43715549e+00 6.21582329e-01
-1.02520335e+00 -5.89133203e-01 -1.36790359e+00 5.20413816e-01
1.30534613e+00 -8.93957555e-01 -1.02631733e-01 7.75791049e-01
3.62254500e-01 -4.38594026e-03 -1.13430130e+00 4.29910809e-01
8.45695674e-01 -8.18293095e-01 1.00855482e+00 -1.08610845e+00
4.17577773e-01 4.75998968e-01 -7.35679567e-01 -9.31986272e-01
-3.02561074e-01 -5.14580429e-01 -3.08770835e-01 6.01136565e-01
7.58752465e-01 -8.42238665e-01 9.76006985e-01 9.90441561e-01
3.87590736e-01 -9.61608946e-01 -1.14688802e+00 -1.07939935e+00
5.80862582e-01 -6.85814142e-01 3.21369499e-01 7.80585527e-01
6.16739333e-01 1.86253831e-01 -3.07350516e-01 2.01263368e-01
8.15854728e-01 6.28773808e-01 3.92328173e-01 -1.23989558e+00
-8.11333239e-01 -6.18403971e-01 -5.49079180e-01 -1.04432571e+00
4.50735271e-01 -1.51308489e+00 -6.95011169e-02 -6.95973694e-01
6.65769994e-01 -9.07350481e-01 -7.82751083e-01 3.46033663e-01
-9.77689549e-02 5.26165187e-01 1.82530999e-01 2.10944325e-01
-7.60233343e-01 2.28598520e-01 4.92505759e-01 -1.43313646e-01
3.86227034e-02 1.59882218e-01 -8.00104678e-01 7.61365771e-01
5.58690131e-01 -5.85741758e-01 -2.79962629e-01 -8.30756277e-02
3.24845910e-01 2.72969812e-01 1.73242651e-02 -8.87850881e-01
-1.95172094e-02 -2.98655927e-01 4.35000151e-01 -1.39486149e-01
5.88991761e-01 -8.31668198e-01 -2.43675649e-01 8.10218632e-01
-9.45565999e-01 -5.19883372e-02 -2.00754166e-01 9.56860185e-01
2.50428557e-01 -3.43776375e-01 9.84134555e-01 1.85889304e-01
2.04238538e-02 4.12603170e-01 -1.14855275e-01 3.27513695e-01
1.15474725e+00 3.28440696e-01 -2.34606296e-01 -3.48306745e-01
-9.38780069e-01 2.01168790e-01 2.14420736e-01 -1.22010484e-01
5.15383184e-01 -1.37847841e+00 -4.45290536e-01 4.08942014e-01
1.28846616e-01 -5.83007812e-01 4.25551869e-02 5.60994923e-01
-2.86371320e-01 1.34288490e-01 3.85502815e-01 -5.53954840e-01
-1.08524871e+00 8.53850842e-01 3.91527742e-01 -1.27156228e-01
-6.40676975e-01 1.26136065e+00 1.86072975e-01 -3.43114734e-01
6.45876825e-01 -1.89947116e-03 5.26593566e-01 -2.55941808e-01
7.12204695e-01 5.64820766e-01 2.25167051e-02 -1.50415078e-01
-3.79679978e-01 4.70585704e-01 -8.83995146e-02 -2.40100086e-01
1.14143944e+00 -2.83206645e-02 -1.41045392e-01 6.61205053e-01
1.77476645e+00 -5.52972779e-02 -1.01167834e+00 -3.55692059e-01
3.27192992e-01 -6.52298570e-01 -7.91686177e-02 -6.45250499e-01
-1.06964278e+00 9.69025731e-01 7.48030245e-01 3.73093992e-01
1.18069828e+00 1.15554772e-01 6.74216390e-01 9.28708732e-01
4.60669160e-01 -1.41489494e+00 -2.75879875e-02 1.99324533e-01
5.65825522e-01 -1.18335307e+00 1.69397533e-01 -7.38073587e-01
-4.58994001e-01 8.70823443e-01 2.83125311e-01 -1.66147202e-01
1.24315941e+00 4.99796748e-01 -1.26964509e-01 1.68636397e-01
-9.34494436e-01 2.30943978e-01 1.07272558e-01 2.31734976e-01
8.15935582e-02 4.31740820e-01 -1.79020047e-01 9.05056059e-01
-3.53696555e-01 -2.56964087e-01 1.23609684e-01 8.34623277e-01
-5.08273900e-01 -1.00543213e+00 -1.56680688e-01 1.06415784e+00
-5.17858028e-01 -1.17503181e-01 -1.68968111e-01 4.25476551e-01
-2.79397666e-01 1.08879495e+00 1.89692765e-01 -7.78661609e-01
-2.31410619e-02 5.13661861e-01 3.49309593e-01 -3.51952314e-01
-3.15791428e-01 -1.53806701e-01 -1.96247414e-01 -5.09851575e-01
2.52620995e-01 -5.79647899e-01 -1.09117961e+00 -6.61184430e-01
-3.31311882e-01 4.14634049e-01 5.21370947e-01 6.40564203e-01
4.99989778e-01 -6.65521771e-02 1.27665806e+00 -8.04891512e-02
-1.46894896e+00 -6.28237367e-01 -1.14879453e+00 4.20913666e-01
2.64040679e-01 -5.28618872e-01 -9.34974968e-01 -1.87894836e-01]
|
[8.015353202819824, 4.209837436676025]
|
65820ce6-b331-454d-9870-957aa66be29a
|
kinematic-3d-object-detection-in-monocular
|
2007.09548
| null |
https://arxiv.org/abs/2007.09548v1
|
https://arxiv.org/pdf/2007.09548v1.pdf
|
Kinematic 3D Object Detection in Monocular Video
|
Perceiving the physical world in 3D is fundamental for self-driving applications. Although temporal motion is an invaluable resource to human vision for detection, tracking, and depth perception, such features have not been thoroughly utilized in modern 3D object detectors. In this work, we propose a novel method for monocular video-based 3D object detection which carefully leverages kinematic motion to improve precision of 3D localization. Specifically, we first propose a novel decomposition of object orientation as well as a self-balancing 3D confidence. We show that both components are critical to enable our kinematic model to work effectively. Collectively, using only a single model, we efficiently leverage 3D kinematics from monocular videos to improve the overall localization precision in 3D object detection while also producing useful by-products of scene dynamics (ego-motion and per-object velocity). We achieve state-of-the-art performance on monocular 3D object detection and the Bird's Eye View tasks within the KITTI self-driving dataset.
|
['Gerard Pons-Moll', 'Xiaoming Liu', 'Garrick Brazil', 'Bernt Schiele']
|
2020-07-19
| null |
https://www.ecva.net/papers/eccv_2020/papers_ECCV/html/4241_ECCV_2020_paper.php
|
https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123680137.pdf
|
eccv-2020-8
|
['vehicle-pose-estimation']
|
['computer-vision']
|
[-3.83739650e-01 -6.83985531e-01 -1.43933237e-01 -2.80613810e-01
-3.55788589e-01 -8.86569440e-01 6.03441596e-01 -2.83052415e-01
-4.98835444e-01 6.51636049e-02 -1.28035069e-01 -3.18774134e-01
2.63791829e-01 -1.36369154e-01 -7.81663597e-01 -4.61702704e-01
-6.46737516e-02 -9.51956399e-03 8.29197347e-01 1.04332089e-01
4.58117515e-01 7.99851716e-01 -1.83385074e+00 -1.17639378e-01
4.25561547e-01 1.10740912e+00 3.88023883e-01 1.14816654e+00
3.35379452e-01 7.02807546e-01 -2.09175289e-01 1.99172869e-01
5.71937740e-01 -6.99848682e-02 -1.15287349e-01 3.03196490e-01
9.19578910e-01 -7.69656479e-01 -5.68109751e-01 9.08050656e-01
4.39131916e-01 -4.79425490e-02 6.46016419e-01 -1.22810471e+00
-1.75582871e-01 -2.00912699e-01 -7.98163831e-01 5.87027848e-01
5.65800548e-01 6.13279939e-01 6.95226550e-01 -1.07898366e+00
6.56934798e-01 1.24452627e+00 4.38240379e-01 3.42401445e-01
-9.20891345e-01 -5.51981807e-01 3.85965079e-01 1.76863045e-01
-1.22798085e+00 -7.16111183e-01 6.62837267e-01 -6.89022183e-01
1.26568365e+00 -1.22560784e-02 8.66089702e-01 6.72765613e-01
5.05401373e-01 7.45764434e-01 9.11373258e-01 -2.48444244e-01
1.12932019e-01 8.98857862e-02 6.08451515e-02 7.97366560e-01
5.16977131e-01 5.20016670e-01 -9.41884100e-01 3.33215773e-01
8.57983708e-01 -9.22762789e-03 -5.26405983e-02 -1.14286578e+00
-1.27685463e+00 4.10835326e-01 3.92551601e-01 -2.47354001e-01
-3.05910711e-04 5.13671994e-01 5.98878488e-02 6.93971589e-02
5.42923391e-01 1.86535791e-01 -2.73910254e-01 -4.23991591e-01
-4.38972533e-01 4.12590891e-01 2.94765919e-01 1.29118001e+00
5.97427249e-01 1.46839663e-01 1.03110172e-01 2.01426312e-01
5.54299891e-01 1.00672889e+00 2.51677670e-02 -1.23998463e+00
2.82058984e-01 7.67074466e-01 4.86668795e-01 -7.41717875e-01
-5.55416584e-01 -5.24754107e-01 -3.68622541e-02 6.03857815e-01
5.42945027e-01 2.21018046e-01 -9.28169906e-01 1.59161937e+00
7.86842346e-01 5.96678741e-02 -2.56987184e-01 1.42260003e+00
6.39637411e-01 1.86705068e-01 -2.70636439e-01 1.39961675e-01
1.19142318e+00 -7.20005214e-01 -1.79027811e-01 -5.04534662e-01
6.44020379e-01 -7.92880774e-01 5.95111012e-01 1.94203243e-01
-1.14028096e+00 -6.51197135e-01 -1.12216938e+00 -2.43182153e-01
-6.07155170e-03 1.82166353e-01 6.02234423e-01 7.08815038e-01
-8.82792115e-01 5.65421283e-02 -1.10772419e+00 -5.02579808e-01
2.87728250e-01 2.23262414e-01 -4.31712389e-01 -1.27377346e-01
-5.45682490e-01 1.20381308e+00 1.26029149e-01 -1.59616113e-01
-9.48401392e-01 -7.49043584e-01 -1.01663411e+00 -2.69491553e-01
4.47840363e-01 -9.60284054e-01 1.43718684e+00 -6.45418167e-02
-1.23160720e+00 1.20546305e+00 -4.60980624e-01 -5.53975880e-01
5.88646650e-01 -4.37486380e-01 -1.94382221e-02 3.33127797e-01
1.60449192e-01 8.23736787e-01 1.10632837e+00 -1.12843633e+00
-1.06253588e+00 -7.16633618e-01 5.02659157e-02 6.13753915e-01
8.77492204e-02 -1.89312980e-01 -5.83265781e-01 -2.43929792e-02
6.08016729e-01 -1.00995576e+00 -1.27579104e-02 5.71029902e-01
-4.01797518e-02 -1.14632487e-01 8.36559117e-01 -6.25785887e-02
6.68238759e-01 -2.10911465e+00 5.74013777e-03 -3.37013900e-01
5.14652789e-01 6.93238825e-02 1.78184941e-01 -8.23591873e-02
4.28525597e-01 -3.60574692e-01 2.56888747e-01 -5.30516624e-01
-1.37868181e-01 -1.44129351e-01 -2.68613636e-01 9.50113475e-01
3.47078085e-01 1.00297821e+00 -1.02429450e+00 -1.72187835e-01
7.58836031e-01 4.93849516e-01 -7.17175364e-01 6.42499849e-02
-6.51721954e-02 2.64951527e-01 -4.26332504e-01 9.42714632e-01
8.10843766e-01 -3.81896808e-03 -3.60510111e-01 -1.88800722e-01
-5.27271688e-01 3.03834051e-01 -1.19873941e+00 1.62691581e+00
-2.11883321e-01 7.98891246e-01 1.83467880e-01 -3.74009460e-01
7.27985859e-01 -2.22058818e-01 4.19503599e-01 -6.89014018e-01
2.47125387e-01 2.43115336e-01 -1.22988194e-01 -2.86539704e-01
7.85842299e-01 1.41418442e-01 1.34825200e-01 3.49797487e-01
-1.56300962e-02 -6.27085030e-01 1.49546489e-01 1.85731903e-01
9.43935335e-01 4.61661637e-01 4.19896662e-01 -1.95431307e-01
1.39825046e-01 2.75976866e-01 2.68777996e-01 9.00295973e-01
-6.47219002e-01 6.76197708e-01 9.03406441e-02 -2.95212895e-01
-1.29840350e+00 -1.29535663e+00 -7.94665068e-02 5.74803233e-01
7.25330710e-01 -1.55885786e-01 -3.11824471e-01 -5.19299626e-01
5.41438520e-01 4.04960573e-01 -3.55801702e-01 -1.79207534e-01
-4.16174144e-01 -2.96090901e-01 9.15743932e-02 5.95762014e-01
2.65266180e-01 -2.69486994e-01 -1.53388226e+00 2.08046600e-01
1.44448414e-01 -1.47073185e+00 -7.21813381e-01 2.75438368e-01
-7.54010677e-01 -1.11347938e+00 -4.54441994e-01 -2.63763189e-01
3.84388596e-01 1.57414532e+00 8.56214941e-01 -4.10923719e-01
-6.17155552e-01 6.73786640e-01 -1.72000363e-01 -4.83420581e-01
-9.41273570e-02 -3.24446440e-01 6.36734605e-01 -2.62564212e-01
5.43490052e-01 -3.15937221e-01 -8.34881008e-01 5.15033245e-01
-3.23882163e-01 1.05760738e-01 4.17509139e-01 4.89259750e-01
2.55269527e-01 -2.98903763e-01 -1.05837435e-01 -1.15419496e-02
-1.84153140e-01 -1.10066116e-01 -1.13262403e+00 -3.84183139e-01
-3.94080698e-01 -1.98800534e-01 1.36307627e-01 -5.35745621e-01
-8.32967222e-01 4.22576964e-01 3.89001727e-01 -8.50451589e-01
-1.47606194e-01 -1.61398947e-01 2.85446435e-01 -5.65989137e-01
7.59861708e-01 2.33080551e-01 1.33733660e-01 -2.29750961e-01
5.05222023e-01 5.40573537e-01 5.94116449e-01 -2.24046797e-01
9.05676842e-01 1.09269202e+00 2.66300321e-01 -9.05408978e-01
-9.55426276e-01 -1.07240379e+00 -8.98577213e-01 -4.60245669e-01
8.97643626e-01 -1.46923280e+00 -1.15459728e+00 4.61197555e-01
-1.14984798e+00 5.03496714e-02 -7.70859495e-02 8.48978937e-01
-6.41245186e-01 3.47208560e-01 -2.23196730e-01 -1.16526198e+00
6.84472173e-02 -1.18094552e+00 1.43222058e+00 2.44780257e-01
3.94254997e-02 -4.23569918e-01 -8.16522092e-02 1.65464535e-01
1.65040910e-01 -1.01581074e-01 2.56383955e-01 1.91294000e-01
-1.12825334e+00 -1.19814336e-01 -5.27668655e-01 6.61694109e-02
-1.90502722e-02 -2.39850935e-02 -1.22599900e+00 -2.94638783e-01
1.06710523e-01 -1.52028278e-01 1.11506736e+00 6.78776264e-01
3.67626309e-01 3.97004247e-01 -3.37344259e-01 8.94366443e-01
1.08742058e+00 2.60235310e-01 -3.38035077e-02 3.33033264e-01
7.29797482e-01 5.47135711e-01 8.25377405e-01 5.86785793e-01
5.51281631e-01 8.90190542e-01 6.12759769e-01 1.19490005e-01
-3.24809253e-01 -2.08348066e-01 4.57796127e-01 4.52318549e-01
4.07329947e-02 9.09770951e-02 -8.32124293e-01 5.36360681e-01
-1.74942803e+00 -8.80076408e-01 -1.66094065e-01 2.21251607e+00
2.83622444e-01 5.14746904e-01 2.54120350e-01 -4.41422723e-02
2.95069784e-01 6.59900978e-02 -9.94382322e-01 2.10027993e-01
-1.38648063e-01 -3.57843220e-01 9.84215736e-01 6.14316940e-01
-1.17948961e+00 1.05067682e+00 6.43832254e+00 1.86110899e-01
-1.19360268e+00 -1.09430686e-01 -9.38973576e-03 -5.80407143e-01
7.73665670e-04 9.57033485e-02 -1.49085093e+00 1.18499093e-01
3.31938028e-01 -1.25596240e-01 2.20977098e-01 1.05362618e+00
3.95183295e-01 -5.27774692e-01 -1.20970488e+00 1.30194008e+00
1.33385569e-01 -1.06744623e+00 -2.88008451e-01 1.34237856e-01
5.72976589e-01 2.84303278e-01 2.10331663e-01 -1.28900662e-01
2.39171609e-02 -5.47914803e-01 1.21011615e+00 8.54265541e-02
6.72331333e-01 -4.56064016e-01 1.92994907e-01 5.31041145e-01
-1.31973648e+00 -8.38893950e-02 -4.03488100e-01 -3.83964598e-01
3.18641335e-01 5.67305684e-01 -1.02561951e+00 2.87068546e-01
7.87105143e-01 9.29648519e-01 -6.20111465e-01 1.31835580e+00
-1.86971109e-02 7.77624100e-02 -5.97314954e-01 -7.87151828e-02
1.96622491e-01 1.44550622e-01 1.03868055e+00 8.86557996e-01
2.93334067e-01 7.16736838e-02 1.15715891e-01 8.32029343e-01
3.10967565e-01 -5.41945577e-01 -8.59288394e-01 1.59576789e-01
5.02767682e-01 1.02371299e+00 -5.40665269e-01 -8.88455287e-02
-5.37443876e-01 7.03291953e-01 1.18246272e-01 2.89927542e-01
-6.88971758e-01 -6.46004006e-02 1.26438010e+00 4.01725560e-01
5.81298769e-01 -1.01620007e+00 -2.27732152e-01 -1.51656866e+00
2.79967815e-01 -3.48296404e-01 -1.64599553e-01 -1.03830993e+00
-8.06356192e-01 2.38944933e-01 8.78664628e-02 -1.52138984e+00
-2.98848391e-01 -1.00779343e+00 -6.17835671e-02 7.58961439e-01
-1.84365845e+00 -1.14371848e+00 -5.57700455e-01 4.02845949e-01
5.15902758e-01 3.38903032e-02 1.36474907e-01 9.22886804e-02
-1.31826997e-01 8.81098136e-02 -2.08417356e-01 -4.84639227e-01
9.52338755e-01 -1.00124049e+00 8.66210878e-01 1.33691967e+00
2.62274116e-01 5.69031715e-01 7.62446702e-01 -6.50052130e-01
-2.13141656e+00 -7.46028841e-01 4.35479432e-01 -1.10217822e+00
4.93757814e-01 -7.23436475e-01 -5.10966599e-01 4.14882928e-01
-4.77930009e-01 2.13260069e-01 -1.04484372e-01 -2.34417856e-01
-5.32127619e-01 -1.07876807e-01 -7.80911028e-01 7.01018989e-01
1.45323408e+00 -6.04958236e-01 -4.84681904e-01 2.34277640e-02
7.97084868e-01 -8.97860229e-01 -2.78943956e-01 3.74240011e-01
9.45250452e-01 -1.14413142e+00 1.28086722e+00 -3.72080654e-01
-1.50795028e-01 -8.16809118e-01 -2.97471195e-01 -8.01854491e-01
-3.31397802e-01 -6.30475819e-01 -6.05702221e-01 5.13880074e-01
-1.25572056e-01 -5.01193702e-01 9.57122087e-01 3.14622819e-01
-1.82207808e-01 -2.75148124e-01 -1.07400453e+00 -9.76364136e-01
-5.63735247e-01 -7.76518524e-01 -2.19339374e-02 3.20350349e-01
-3.40718985e-01 1.57249987e-01 -4.02553201e-01 3.06040764e-01
1.10829246e+00 2.92762250e-01 1.19950855e+00 -1.04322863e+00
-8.54436159e-02 -4.42419916e-01 -8.60297322e-01 -1.86531818e+00
-2.83106416e-01 -5.50880671e-01 1.42938778e-01 -1.08001399e+00
1.54544279e-01 -2.30635747e-01 -5.80503419e-02 -1.71379775e-01
-1.12368189e-01 3.64013553e-01 3.54643613e-01 2.88285732e-01
-6.24580681e-01 4.04063702e-01 1.24184203e+00 3.05744171e-01
-2.87566364e-01 -5.76751493e-02 -4.87445503e-01 5.05421162e-01
3.27593744e-01 -3.24841946e-01 -3.70055348e-01 -6.12842500e-01
-1.79840624e-02 3.34151797e-02 8.32000792e-01 -9.69581246e-01
5.43761551e-01 -1.53215423e-01 6.54150903e-01 -1.25665879e+00
7.29844153e-01 -5.41534841e-01 -3.14776897e-01 5.79340458e-01
2.03669533e-01 -1.14628702e-01 3.13605815e-01 7.50891328e-01
1.21036567e-01 2.49396309e-01 9.25127804e-01 -2.15455174e-01
-1.21840072e+00 3.64055663e-01 -2.41822615e-01 7.06783384e-02
1.15255117e+00 -5.59872687e-01 -4.56670254e-01 -2.75765091e-01
-2.85803735e-01 3.03698897e-01 8.98691773e-01 6.25052571e-01
8.02117348e-01 -1.13353479e+00 -4.53026086e-01 6.02689624e-01
3.91755432e-01 -2.07737479e-02 2.69840717e-01 8.75486851e-01
-5.82005799e-01 9.50641572e-01 -1.41805902e-01 -1.35037184e+00
-1.21164107e+00 5.60753822e-01 3.42965424e-01 5.98682821e-01
-6.68083549e-01 9.18740273e-01 5.11369526e-01 -1.03870586e-01
2.85357118e-01 -6.70571208e-01 3.33956957e-01 -2.53081262e-01
6.03865385e-01 3.47699761e-01 -5.73461279e-02 -5.01777112e-01
-6.41339839e-01 1.13913846e+00 -7.54405186e-02 -2.59854287e-01
9.58955348e-01 -7.05114186e-01 4.38503295e-01 5.08084357e-01
1.03920043e+00 2.96111144e-02 -1.94567156e+00 -1.86715320e-01
-3.13931137e-01 -9.29488420e-01 3.12111169e-01 -3.25954288e-01
-4.28799629e-01 1.05076218e+00 6.56856179e-01 -1.08495146e-01
6.46330893e-01 1.97517648e-01 4.48619783e-01 4.36726689e-01
6.39041722e-01 -6.60361111e-01 2.29625255e-01 6.48670971e-01
4.36552107e-01 -1.70856261e+00 1.17917798e-01 -5.36398530e-01
-4.16162908e-01 1.01169670e+00 7.29649842e-01 -6.11305870e-02
4.51500595e-01 3.03082556e-01 -5.66682965e-02 -5.04551567e-02
-8.44516873e-01 -6.11591399e-01 5.42625368e-01 6.26335204e-01
-1.46923540e-02 -1.81161597e-01 3.78145158e-01 1.79743543e-02
1.55679122e-01 -3.07168841e-01 4.44915980e-01 1.03849113e+00
-8.16190422e-01 -4.78499919e-01 -2.47087866e-01 3.61635685e-02
-9.87109840e-02 1.18589103e-01 -2.53746301e-01 8.48945856e-01
-2.00922146e-01 8.82220984e-01 1.19546540e-01 -4.36158866e-01
4.40385938e-01 -2.90472895e-01 1.05908203e+00 -6.58595324e-01
2.92033553e-02 2.37371504e-01 -9.66153368e-02 -9.02542830e-01
-3.42223376e-01 -7.24447310e-01 -8.25341523e-01 -2.28173167e-01
-3.54628652e-01 -6.02711916e-01 1.08199108e+00 6.61223114e-01
5.26479244e-01 4.29507680e-02 5.15346944e-01 -1.31403422e+00
-6.00596011e-01 -5.07100463e-01 -5.11170089e-01 3.39181442e-03
7.94362783e-01 -1.14265478e+00 -7.04543412e-01 -1.19524263e-01]
|
[7.971288681030273, -2.448915719985962]
|
606bc448-37a9-426c-b033-ceb810ff172c
|
point-cloud-segmentation-using-sparse
|
2112.00289
| null |
https://arxiv.org/abs/2112.00289v2
|
https://arxiv.org/pdf/2112.00289v2.pdf
|
Point Cloud Segmentation Using Sparse Temporal Local Attention
|
Point clouds are a key modality used for perception in autonomous vehicles, providing the means for a robust geometric understanding of the surrounding environment. However despite the sensor outputs from autonomous vehicles being naturally temporal in nature, there is still limited exploration of exploiting point cloud sequences for 3D seman-tic segmentation. In this paper we propose a novel Sparse Temporal Local Attention (STELA) module which aggregates intermediate features from a local neighbourhood in previous point cloud frames to provide a rich temporal context to the decoder. Using the sparse local neighbourhood enables our approach to gather features more flexibly than those which directly match point features, and more efficiently than those which perform expensive global attention over the whole point cloud frame. We achieve a competitive mIoU of 64.3% on the SemanticKitti dataset, and demonstrate significant improvement over the single-frame baseline in our ablation studies.
|
['Sridha Sridharan', 'Clinton Fookes', 'Peyman Moghadam', 'Joshua Knights']
|
2021-12-01
| null | null | null | null |
['point-cloud-segmentation']
|
['computer-vision']
|
[ 1.99087262e-01 5.51916547e-02 -1.85270116e-01 -6.63495421e-01
-1.02967000e+00 -6.77104294e-01 9.05339479e-01 -1.60983298e-02
-5.31459153e-01 2.09284380e-01 -9.40501988e-02 -1.88078880e-01
2.85031080e-01 -7.72985101e-01 -1.17126524e+00 -5.60141861e-01
-1.18837237e-01 5.78461826e-01 6.94059968e-01 -2.40786374e-01
1.72004417e-01 7.68791676e-01 -1.74225426e+00 2.25193664e-01
6.59759104e-01 1.06212151e+00 7.19837844e-01 5.56989789e-01
-3.32377642e-01 2.34921977e-01 -2.72249699e-01 -4.34531718e-02
4.29402590e-01 8.80325735e-02 -6.41546249e-01 1.95576891e-01
8.62910748e-01 -4.50137943e-01 -4.73965555e-01 9.97472048e-01
-8.93331468e-02 4.41230744e-01 2.88126141e-01 -1.21345603e+00
-1.33300662e-01 1.44633710e-01 -6.64224207e-01 4.13928598e-01
2.85539091e-01 4.31328118e-01 9.75871146e-01 -7.89259076e-01
9.36580539e-01 1.52281749e+00 4.69080329e-01 4.10568953e-01
-8.91120434e-01 -6.63267255e-01 6.58346236e-01 4.00350958e-01
-1.08566356e+00 -6.11037970e-01 6.10925317e-01 -2.52796352e-01
1.39428866e+00 7.92381838e-02 7.91178644e-01 8.15597355e-01
1.33152425e-01 7.89967597e-01 6.21547401e-01 1.88982829e-01
1.12476557e-01 -3.68876874e-01 2.72883754e-02 5.83953083e-01
-1.87844664e-01 2.73251414e-01 -5.92537582e-01 1.44518539e-01
7.11697459e-01 2.52991825e-01 1.55347973e-01 -4.30258602e-01
-1.40748286e+00 6.49583936e-01 9.53519404e-01 7.16111958e-02
-4.90536124e-01 8.75575840e-01 8.53237137e-02 2.31449619e-01
5.17113864e-01 1.11718006e-01 -5.85468054e-01 -2.43029281e-01
-8.02689552e-01 2.80921102e-01 3.78510594e-01 1.31205308e+00
1.10532796e+00 -5.54176532e-02 1.02977194e-01 3.61758262e-01
5.45698225e-01 9.76174533e-01 -3.99173759e-02 -1.43636966e+00
6.22467816e-01 3.55320573e-01 2.11710587e-01 -8.85437071e-01
-2.27511853e-01 -6.88204542e-02 -2.51041353e-01 4.45573539e-01
8.19187239e-02 1.98512301e-01 -1.37879825e+00 1.57905841e+00
5.64308047e-01 7.47082770e-01 1.49999812e-01 9.42176878e-01
8.33173633e-01 1.00368488e+00 9.36049893e-02 1.58922240e-01
1.12401509e+00 -8.57473552e-01 -2.97205687e-01 -7.52689004e-01
4.46483105e-01 -4.78144944e-01 6.79540694e-01 -3.91304940e-02
-9.12005305e-01 -5.79774797e-01 -1.06478441e+00 -5.06578445e-01
-2.47899950e-01 -5.58418334e-01 8.49020004e-01 -1.80933310e-03
-1.28786445e+00 4.75970089e-01 -1.38104641e+00 -5.10309637e-01
7.14022696e-01 3.86333257e-01 -4.49606121e-01 -2.52391368e-01
-6.58269823e-01 8.65102410e-01 2.22760692e-01 -3.66084613e-02
-9.62319016e-01 -7.67700255e-01 -1.12706292e+00 -1.84987426e-01
3.40408713e-01 -7.80607700e-01 1.41007197e+00 -6.85039878e-01
-1.32450521e+00 8.59566689e-01 -8.30167532e-01 -8.94980073e-01
3.09152931e-01 -3.50372225e-01 1.04246147e-01 3.19394112e-01
3.49059224e-01 1.47591186e+00 7.19025075e-01 -1.31862128e+00
-1.04210281e+00 -4.09244120e-01 9.95760933e-02 4.71671939e-01
5.93948722e-01 -2.00718105e-01 -1.16402471e+00 -1.02277853e-01
6.21282816e-01 -1.13853717e+00 -5.82398713e-01 3.33845407e-01
8.04151893e-02 -3.25484395e-01 1.17659187e+00 -4.19785619e-01
1.47225872e-01 -2.31944418e+00 -7.76749058e-03 1.67648777e-01
1.71992779e-01 3.25136222e-02 -2.30724275e-01 8.99914056e-02
3.01510334e-01 6.20965362e-02 -3.29483151e-01 -7.56004095e-01
-1.29191101e-01 6.70008779e-01 -5.69924951e-01 3.95383596e-01
4.98844028e-01 1.35571814e+00 -1.10334194e+00 -3.26113105e-01
6.47182703e-01 4.53429103e-01 -5.68177760e-01 -1.24544486e-01
-7.58827448e-01 5.85759044e-01 -8.10129702e-01 6.95098996e-01
6.50008380e-01 -2.34596193e-01 -4.60442781e-01 3.63793224e-02
-2.80290335e-01 4.29420859e-01 -6.11869037e-01 2.29468703e+00
-4.03952539e-01 8.88052762e-01 -2.36345641e-02 -5.88231504e-01
7.56105840e-01 2.54687872e-02 6.89271092e-01 -8.11689854e-01
9.93580073e-02 -4.03349008e-03 -3.40512246e-01 -3.45848888e-01
6.17382109e-01 1.27974153e-01 -2.38864705e-01 -1.81236520e-01
-3.88274230e-02 -6.18616760e-01 -1.10539503e-03 4.34166938e-01
1.15451908e+00 3.50676626e-01 -2.02131167e-01 7.43115768e-02
1.02749191e-01 5.22470057e-01 6.53104246e-01 6.81045473e-01
-3.31790984e-01 8.84812891e-01 9.39758718e-02 -5.10232866e-01
-1.02159953e+00 -1.03776038e+00 -1.53583952e-03 6.18964612e-01
8.84404421e-01 -2.58194119e-01 -1.55221105e-01 -6.11319661e-01
1.61782920e-01 6.77295506e-01 -3.72551590e-01 5.28682396e-02
-6.64618790e-01 -4.44309674e-02 2.40951270e-01 7.29915082e-01
6.86265469e-01 -9.27429557e-01 -9.38859165e-01 2.67240703e-01
-3.62873644e-01 -1.80256891e+00 -4.28945631e-01 1.91559389e-01
-1.15633440e+00 -7.31700361e-01 -1.70823216e-01 -5.10406911e-01
3.52729678e-01 8.31341684e-01 1.20862174e+00 1.03931455e-02
1.37453780e-01 3.11775863e-01 -4.37267095e-01 -5.22455752e-01
-1.43752635e-01 -1.08000681e-01 -2.10332438e-01 -2.49940991e-01
5.49648345e-01 -6.00356519e-01 -4.84883636e-01 1.21979028e-01
-5.95450163e-01 2.71034181e-01 5.87367177e-01 3.41320097e-01
9.88792181e-01 -3.16577494e-01 9.34237093e-02 -4.91849124e-01
-4.17270780e-01 -4.96027589e-01 -8.28449965e-01 -4.25519973e-01
-4.25492525e-02 4.20469493e-02 -2.00367980e-02 -6.22840188e-02
-8.80390823e-01 4.23857033e-01 -3.37936759e-01 -7.99010515e-01
-3.86960626e-01 2.51931131e-01 -1.66824803e-01 -2.24638239e-01
2.46665493e-01 -4.94456291e-03 3.81075107e-02 -2.93174922e-01
5.67670345e-01 8.29164013e-02 8.32556844e-01 -4.57894653e-01
1.02946067e+00 1.00426304e+00 -3.69615736e-04 -8.56271327e-01
-6.33596659e-01 -8.80389035e-01 -7.25727320e-01 -1.56818733e-01
1.19137764e+00 -1.20320952e+00 -6.24226272e-01 1.37304962e-01
-1.41193163e+00 -3.95153522e-01 -3.60942215e-01 3.56783539e-01
-7.79700458e-01 3.38238418e-01 -2.14107499e-01 -5.23273587e-01
5.83454908e-04 -1.31766343e+00 1.61941576e+00 1.45869851e-01
-4.87324670e-02 -6.03792131e-01 -2.59318084e-01 2.28682488e-01
4.46759425e-02 4.77589309e-01 3.24038446e-01 -2.09743351e-01
-1.49445522e+00 -4.65644300e-02 -4.33761865e-01 -1.20523289e-01
-1.25481173e-01 -9.80606750e-02 -1.05633259e+00 -1.48072511e-01
-5.55209666e-02 4.43099625e-02 1.28542125e+00 4.45496589e-01
8.38411212e-01 9.59931538e-02 -8.21270347e-01 9.45989788e-01
1.44863975e+00 2.20118046e-01 5.72576225e-01 3.37142438e-01
8.83927345e-01 2.66265631e-01 9.39677477e-01 6.29297197e-02
7.27903068e-01 7.52679765e-01 9.13260102e-01 -8.72407779e-02
-1.97550327e-01 -2.56562948e-01 3.49421591e-01 4.13665503e-01
-4.08482701e-02 -2.31345057e-01 -9.97807920e-01 8.98241162e-01
-1.95442891e+00 -1.00941789e+00 -1.31297722e-01 1.83595550e+00
2.74259418e-01 2.67975718e-01 -3.10409814e-01 -2.91837007e-01
3.63039136e-01 4.16755885e-01 -8.17013144e-01 -1.24840766e-01
-1.74200624e-01 2.80597340e-02 9.98882353e-01 7.82613754e-01
-1.22347260e+00 1.43134987e+00 5.92481852e+00 5.05084753e-01
-9.59829569e-01 1.56296909e-01 2.47112915e-01 -1.93163261e-01
-4.33504879e-01 2.73962736e-01 -8.14276278e-01 2.34508097e-01
8.66677642e-01 1.24656402e-01 1.80030122e-01 7.11120367e-01
3.92725617e-01 -3.33110541e-01 -1.22580719e+00 1.18242323e+00
-3.09027076e-01 -1.47835541e+00 -1.91047452e-02 2.41423979e-01
7.67701447e-01 1.16168427e+00 7.80146867e-02 -2.15074848e-02
5.07062972e-01 -7.46692479e-01 1.00210285e+00 4.93193299e-01
5.83275735e-01 -7.35742450e-01 4.70055789e-01 4.17396665e-01
-1.31258321e+00 1.37260422e-01 -3.53986204e-01 6.61289245e-02
5.55338025e-01 4.62359786e-01 -1.02327836e+00 5.32992184e-01
1.05378747e+00 1.11556923e+00 -4.03037190e-01 1.12497997e+00
2.97185611e-02 4.30757135e-01 -1.03428066e+00 1.85558692e-01
7.97864854e-01 -1.33830890e-01 9.34503734e-01 1.09469414e+00
3.85691315e-01 3.51332188e-01 3.25479150e-01 7.23881423e-01
3.55519056e-02 -5.21233797e-01 -9.25026596e-01 1.27462983e-01
5.76573491e-01 8.58021677e-01 -7.69635677e-01 -4.42824066e-01
-5.88914633e-01 9.17087555e-01 1.01842964e-02 5.16245604e-01
-8.74903440e-01 8.73229280e-02 1.25035393e+00 -1.00243449e-01
7.51885295e-01 -9.31056082e-01 -4.44930553e-01 -1.07964802e+00
8.77139047e-02 -3.80773187e-01 3.01163178e-03 -8.96350801e-01
-8.30551744e-01 5.29968500e-01 1.20624363e-01 -1.35628581e+00
-4.28183407e-01 -1.75589249e-01 -3.78755450e-01 8.75586152e-01
-1.77559304e+00 -1.30528450e+00 -5.22896767e-01 5.38595557e-01
9.93014336e-01 3.39527696e-01 3.09658915e-01 -2.83346307e-02
1.66683331e-01 1.29493400e-02 -1.43986106e-01 -1.87374175e-01
1.16555110e-01 -1.06390393e+00 1.15294373e+00 1.06357050e+00
3.88930202e-01 3.43951762e-01 6.96103215e-01 -7.70618796e-01
-1.58950555e+00 -1.28878701e+00 7.45338976e-01 -8.56105268e-01
3.75596374e-01 -3.05708051e-01 -9.38181043e-01 7.44659424e-01
1.88197363e-02 1.60818085e-01 -1.48944438e-01 -2.55289346e-01
-3.17704409e-01 -4.68218848e-02 -1.05828094e+00 6.60513997e-01
1.39156830e+00 -4.86084431e-01 -6.02451801e-01 1.75156459e-01
1.33756971e+00 -7.76060939e-01 -5.60039937e-01 6.15367055e-01
1.69726267e-01 -6.54046774e-01 1.25771379e+00 -5.37091419e-02
6.70375600e-02 -7.61928201e-01 -4.41996425e-01 -8.99429142e-01
-1.63162231e-01 -5.82747281e-01 -1.26211450e-01 8.82363379e-01
3.57570529e-01 -4.72809285e-01 1.03598142e+00 6.32752776e-01
-6.51203990e-01 -3.06897223e-01 -1.06876791e+00 -7.20412076e-01
-1.81193441e-01 -1.10550880e+00 7.36277401e-01 5.78560114e-01
-6.34681463e-01 1.31592840e-01 1.34904936e-01 7.78416395e-01
7.36193538e-01 3.08376282e-01 9.00360286e-01 -1.18916547e+00
1.26676962e-01 -3.98678601e-01 -9.79846060e-01 -1.80334330e+00
2.81379580e-01 -9.54173863e-01 5.42401850e-01 -1.81504977e+00
-2.29358435e-01 -6.63341343e-01 9.31427926e-02 5.76846540e-01
2.92810313e-02 4.62349564e-01 3.77920508e-01 3.10387135e-01
-8.28550458e-01 6.04533970e-01 1.14499331e+00 -3.74453545e-01
-3.22383225e-01 -1.22200236e-01 -3.31330687e-01 6.82297409e-01
5.75828671e-01 -5.06331503e-01 -4.63384539e-01 -1.05630505e+00
-1.35139003e-01 4.56229746e-02 7.94183373e-01 -1.16080689e+00
4.17718768e-01 -3.86708409e-01 3.31568033e-01 -1.31191361e+00
9.48191166e-01 -9.64920998e-01 1.22575603e-01 5.38240336e-02
8.14231262e-02 5.69580644e-02 4.87514645e-01 9.02613044e-01
-2.56318867e-01 2.40522996e-01 5.71178854e-01 -2.11118191e-01
-1.55696809e+00 8.97548556e-01 -8.97521675e-02 -2.22437605e-01
9.68109667e-01 -5.16332269e-01 2.46019699e-02 -3.06477100e-01
-6.94117606e-01 6.55640185e-01 8.66383791e-01 8.62148345e-01
9.24223363e-01 -1.16490149e+00 -5.12987614e-01 3.12700331e-01
2.39573345e-01 8.33869636e-01 1.24552406e-01 6.69390380e-01
-5.46320617e-01 5.05771458e-01 -7.51817646e-03 -1.34721351e+00
-1.11684787e+00 3.86269689e-01 3.47277105e-01 5.34006834e-01
-1.25383186e+00 9.82093334e-01 4.84037399e-01 -1.50550485e-01
1.35948449e-01 -8.64732504e-01 1.03651561e-01 -3.27327579e-01
2.85729200e-01 -8.66688266e-02 2.22387329e-01 -1.06724823e+00
-5.22365749e-01 9.18276608e-01 1.21533692e-01 -5.94046593e-01
1.25915635e+00 -5.10465205e-01 1.07123815e-01 5.63045204e-01
1.32842839e+00 -4.28261995e-01 -2.00473309e+00 -3.80448103e-01
2.01027449e-02 -7.52039135e-01 2.63462663e-01 -3.33645910e-01
-1.00732791e+00 7.83296108e-01 5.36041915e-01 -5.98230474e-02
8.07442486e-01 5.42586565e-01 9.28181052e-01 5.93325377e-01
7.81647861e-01 -5.53995907e-01 -3.16149682e-01 9.76500511e-01
7.88171887e-01 -1.41543496e+00 -2.60484666e-01 -5.56411922e-01
-5.92306733e-01 7.38027930e-01 4.55046713e-01 -3.46216887e-01
5.12422144e-01 1.72001943e-01 1.69833615e-01 -3.47491980e-01
-1.02036822e+00 -7.59159982e-01 2.96180338e-01 8.78831625e-01
-2.26212934e-01 -1.73399448e-01 4.33625877e-01 -1.81499273e-02
-2.44359359e-01 -2.77546763e-01 1.67279974e-01 9.38973784e-01
-6.82516158e-01 -7.80346870e-01 -1.47307888e-01 3.24275613e-01
-1.73212141e-02 1.03487194e-01 -1.18494108e-01 7.30986297e-01
2.37797111e-01 1.14029968e+00 4.42091793e-01 -2.83641577e-01
2.99047261e-01 -1.65693521e-01 4.63171482e-01 -6.67376876e-01
-6.43376112e-02 1.24381259e-01 5.04547805e-02 -1.20985770e+00
-6.48214281e-01 -1.05300868e+00 -1.82078218e+00 -1.21875584e-01
-1.04992226e-01 -1.24984384e-01 9.38119531e-01 1.27805221e+00
4.77965921e-01 3.77072871e-01 4.48802382e-01 -1.52541184e+00
2.35477343e-01 -5.35077214e-01 7.16498271e-02 3.23945612e-01
8.77053976e-01 -6.61374390e-01 -1.52762309e-01 2.33496889e-01]
|
[7.988485813140869, -2.4000415802001953]
|
9c66101d-6b6b-4efe-a35b-40fcbd3cf0e4
|
cpm-a-large-scale-generative-chinese-pre
|
2012.00413
| null |
https://arxiv.org/abs/2012.00413v1
|
https://arxiv.org/pdf/2012.00413v1.pdf
|
CPM: A Large-scale Generative Chinese Pre-trained Language Model
|
Pre-trained Language Models (PLMs) have proven to be beneficial for various downstream NLP tasks. Recently, GPT-3, with 175 billion parameters and 570GB training data, drew a lot of attention due to the capacity of few-shot (even zero-shot) learning. However, applying GPT-3 to address Chinese NLP tasks is still challenging, as the training corpus of GPT-3 is primarily English, and the parameters are not publicly available. In this technical report, we release the Chinese Pre-trained Language Model (CPM) with generative pre-training on large-scale Chinese training data. To the best of our knowledge, CPM, with 2.6 billion parameters and 100GB Chinese training data, is the largest Chinese pre-trained language model, which could facilitate several downstream Chinese NLP tasks, such as conversation, essay generation, cloze test, and language understanding. Extensive experiments demonstrate that CPM achieves strong performance on many NLP tasks in the settings of few-shot (even zero-shot) learning. The code and parameters are available at https://github.com/TsinghuaAI/CPM-Generate.
|
['Maosong Sun', 'Xiaoyan Zhu', 'Juanzi Li', 'Jie Tang', 'Wentao Han', 'Minlie Huang', 'Zhiyuan Liu', 'Zhenbo Sun', 'Daixuan Li', 'Shengqi Chen', 'Huanqi Cao', 'Guoyang Zeng', 'Yanan Zheng', 'Xiaozhi Wang', 'Fanchao Qi', 'Jian Guan', 'Haozhe Ji', 'Yusheng Su', 'Yujia Qin', 'Deming Ye', 'Yuxian Gu', 'Pei Ke', 'Hao Zhou', 'Xu Han', 'Zhengyan Zhang']
|
2020-12-01
| null | null | null | null |
['cloze-test']
|
['natural-language-processing']
|
[-1.84326544e-01 1.46640018e-01 -3.51672053e-01 -1.71471685e-01
-1.29889107e+00 -3.90963227e-01 6.13807678e-01 -2.35006899e-01
-3.48255247e-01 9.47582483e-01 5.49156725e-01 -6.76885962e-01
4.91365463e-01 -7.60539770e-01 -4.01570052e-01 -5.85605741e-01
2.05248281e-01 7.25779653e-01 -8.30328390e-02 -3.41198146e-01
7.30397031e-02 -1.87317565e-01 -8.05903554e-01 4.14422959e-01
1.20237982e+00 3.50618035e-01 5.97928584e-01 8.82101476e-01
-4.39076096e-01 7.15501904e-01 -5.50664723e-01 -4.85438347e-01
-1.85569122e-01 -5.61521173e-01 -8.54780853e-01 -2.92716116e-01
-1.68968245e-01 -4.01680827e-01 -3.54194224e-01 8.13407958e-01
7.97360122e-01 2.03943670e-01 3.24840605e-01 -1.08700788e+00
-1.26603615e+00 1.28624570e+00 -1.96627036e-01 1.31265193e-01
7.82277659e-02 4.15900916e-01 1.17620802e+00 -1.23577678e+00
5.20406544e-01 1.31997836e+00 2.99617440e-01 9.36599135e-01
-7.42615342e-01 -8.62141073e-01 -1.66909490e-02 -2.79323701e-02
-1.22645831e+00 -5.78371823e-01 2.76056856e-01 -7.49240443e-02
1.41315079e+00 -1.56498924e-01 3.28407228e-01 1.58073938e+00
1.12587735e-01 1.29695058e+00 8.54122818e-01 -5.95090866e-01
1.98897824e-01 9.80008096e-02 1.75033420e-01 3.78995061e-01
-1.48147270e-01 -2.71357685e-01 -5.07659674e-01 -1.96615607e-01
5.05229175e-01 -1.36154890e-01 -3.20699841e-01 4.93652016e-01
-1.19334662e+00 1.12544358e+00 6.84634596e-02 5.67563117e-01
-1.02159411e-01 1.36598870e-01 3.21581662e-01 2.93029249e-01
8.99361074e-01 5.37105143e-01 -5.28681099e-01 -8.08052719e-01
-7.60246634e-01 4.77849953e-02 1.10408068e+00 1.55009878e+00
6.50673330e-01 2.53364384e-01 -4.43888336e-01 9.92215574e-01
-2.79135462e-02 5.02094507e-01 1.04179394e+00 -7.20131159e-01
8.77104461e-01 2.30000511e-01 -8.39863718e-02 -3.08855444e-01
8.33787583e-03 -1.38064474e-01 -8.80111814e-01 -7.20751047e-01
2.09271565e-01 -8.61897886e-01 -9.05955732e-01 1.57317400e+00
-7.04091489e-02 4.31403697e-01 5.56128085e-01 5.48348546e-01
1.10917974e+00 1.37702692e+00 1.29461035e-01 -9.88034531e-02
1.21565878e+00 -1.29658270e+00 -6.04965985e-01 -5.49691975e-01
1.00391436e+00 -8.63730669e-01 1.60579312e+00 5.91040514e-02
-1.07531977e+00 -5.29296577e-01 -5.47848046e-01 -4.25387174e-01
-2.64667362e-01 2.09421709e-01 6.16706491e-01 6.10809207e-01
-9.29673135e-01 3.33441228e-01 -8.08020771e-01 -4.61510211e-01
3.85167956e-01 -1.91012293e-01 2.09460873e-02 -4.80345845e-01
-1.66219151e+00 6.53942287e-01 4.47571605e-01 -1.03650494e-02
-8.17473352e-01 -7.74690866e-01 -7.93934226e-01 4.36564654e-01
3.55987728e-01 -4.91221219e-01 1.50584280e+00 -4.24813330e-01
-1.98248160e+00 5.40143192e-01 -3.16208392e-01 -4.61475700e-01
3.85365158e-01 -4.92325962e-01 -3.44951570e-01 -8.48864689e-02
1.24046661e-01 8.20750415e-01 4.17141885e-01 -6.28374279e-01
-3.05518627e-01 2.22684771e-01 -3.30947936e-01 9.64185223e-02
-6.36240363e-01 1.80325806e-01 -7.97196448e-01 -5.41862845e-01
-4.79545683e-01 -9.28766191e-01 -3.33175927e-01 -8.22217345e-01
-5.21269441e-01 -7.02500045e-01 6.42793477e-01 -6.79248571e-01
1.44688237e+00 -2.15439153e+00 -3.29782009e-01 -3.27896386e-01
-8.67929757e-02 5.69994986e-01 -5.86109281e-01 7.60361910e-01
4.08767641e-01 5.52080929e-01 -1.86511204e-01 -5.88039398e-01
1.07012942e-01 2.42960915e-01 -5.33996403e-01 -8.36376473e-02
2.93924481e-01 1.52952170e+00 -1.04131126e+00 -4.14827079e-01
-1.07335709e-01 3.91702294e-01 -4.12826031e-01 3.50687623e-01
-5.75277686e-01 2.48395696e-01 -6.70288622e-01 5.45784235e-01
4.82145637e-01 -6.16899431e-01 8.63309875e-02 6.54516041e-01
-1.39973670e-01 6.46980941e-01 -3.86637598e-01 1.73252261e+00
-5.64303458e-01 6.66210532e-01 -3.71079832e-01 -4.28357571e-01
9.73969519e-01 6.67305171e-01 3.09514301e-03 -6.63692772e-01
1.04443081e-01 2.63619572e-01 8.48568380e-02 -4.36532021e-01
7.24345267e-01 -1.06415480e-01 -5.00085831e-01 9.72433984e-01
2.99172282e-01 -2.36302346e-01 5.35322070e-01 6.87975287e-01
1.03455281e+00 -9.23946351e-02 2.27099895e-01 -5.82706928e-02
1.18935153e-01 8.94059166e-02 6.35493696e-01 8.80153000e-01
-1.27938092e-01 6.37007892e-01 6.45814002e-01 9.24874991e-02
-1.11996949e+00 -8.22255433e-01 2.32998550e-01 1.42091143e+00
-3.83775890e-01 -6.47362113e-01 -8.18719506e-01 -3.49435329e-01
-2.39628360e-01 1.20426297e+00 -1.00940019e-01 -1.87775925e-01
-7.80413628e-01 -8.23649585e-01 8.67936015e-01 5.38007021e-01
5.05129039e-01 -1.51129484e+00 1.90116819e-02 4.74070400e-01
-3.82547110e-01 -1.26583207e+00 -9.10629570e-01 -8.24755952e-02
-6.77812457e-01 -6.20747030e-01 -9.63553846e-01 -1.05116057e+00
4.11552221e-01 3.10326964e-01 1.17416584e+00 8.02390873e-02
-4.34160158e-02 6.63968101e-02 -7.31824994e-01 -4.41853404e-01
-7.61376441e-01 5.46013594e-01 -2.79390901e-01 -3.74888718e-01
6.62885427e-01 -3.83376569e-01 -1.79786608e-01 -3.93025875e-02
-5.29425263e-01 3.33272278e-01 7.47319400e-01 9.08202469e-01
2.24160105e-01 -3.28469396e-01 8.75688314e-01 -1.25291419e+00
1.16099739e+00 -6.90479577e-01 -3.44021857e-01 5.44358432e-01
-3.35608155e-01 -8.37279633e-02 7.26231873e-01 -5.82988977e-01
-1.64560544e+00 -5.53017795e-01 -3.33744377e-01 -2.61426955e-01
-1.14467464e-01 7.89914310e-01 -1.59969002e-01 5.17487109e-01
4.69436437e-01 4.01294708e-01 -3.24588954e-01 -6.69108272e-01
5.95926642e-01 1.05863798e+00 2.75011450e-01 -6.83078766e-01
4.75107640e-01 -1.74182802e-01 -7.93848455e-01 -8.40890348e-01
-9.29999053e-01 -5.01403987e-01 -4.20214981e-01 3.38581085e-01
8.08832049e-01 -1.31440330e+00 -4.24987674e-01 5.73497057e-01
-1.16754043e+00 -9.66022789e-01 -8.32488108e-03 5.23295105e-01
-3.29054564e-01 2.38821760e-01 -1.30253983e+00 -7.58353591e-01
-9.25091088e-01 -8.71425629e-01 7.82938242e-01 3.65043581e-01
-1.80685326e-01 -1.07374966e+00 1.44828826e-01 3.49531502e-01
5.91027498e-01 -4.24451083e-01 1.08149445e+00 -8.08546305e-01
-4.47851300e-01 -9.15905982e-02 -1.42294556e-01 4.47296411e-01
-9.88439023e-02 -9.13619697e-02 -8.85988414e-01 -2.81328499e-01
-8.49220455e-02 -7.70666003e-01 8.77425075e-01 3.00315976e-01
1.16932023e+00 -3.10006469e-01 -2.24789992e-01 4.79567468e-01
1.08568347e+00 6.99220970e-02 5.90930164e-01 -1.78530023e-01
7.31258035e-01 2.72486120e-01 5.92029393e-01 4.74247247e-01
4.21888679e-01 1.46702915e-01 -2.53314495e-01 1.95888683e-01
-2.20687449e-01 -7.16358602e-01 6.10811293e-01 1.50095057e+00
8.32924545e-02 -8.78599942e-01 -1.22643030e+00 5.80540240e-01
-1.78315485e+00 -8.45575213e-01 -7.54468367e-02 1.69309664e+00
1.21641695e+00 2.15825975e-01 -3.10221285e-01 -7.18399584e-01
7.85905063e-01 1.57813832e-01 -5.12765229e-01 -5.27491391e-01
-1.55788079e-01 3.40587884e-01 1.74421474e-01 5.07647455e-01
-8.10489833e-01 1.52965474e+00 5.56657600e+00 1.20819271e+00
-1.06677496e+00 5.16525865e-01 9.06235754e-01 -1.59447715e-01
-2.80395120e-01 1.97761685e-01 -1.34631217e+00 7.57166445e-01
1.36104739e+00 -8.74945581e-01 2.32939333e-01 1.02556670e+00
3.13700467e-01 9.04231742e-02 -8.86985779e-01 6.27341509e-01
1.38750732e-01 -1.38903701e+00 2.93690134e-02 6.69620037e-02
1.15793335e+00 6.05164349e-01 3.08184437e-02 1.09405315e+00
7.67068207e-01 -1.04534554e+00 3.35423917e-01 9.29076821e-02
1.04303217e+00 -8.15207779e-01 6.02754235e-01 8.86513948e-01
-1.02891290e+00 1.29374400e-01 -8.37870657e-01 -1.07709542e-01
5.91848612e-01 6.81540072e-01 -1.12853003e+00 1.18244834e-01
3.02966893e-01 8.60420763e-01 -3.29092085e-01 6.87142670e-01
-6.59379065e-01 1.20395422e+00 -1.11510105e-01 -5.50606012e-01
5.88335752e-01 -1.07580842e-02 1.92913473e-01 1.43726337e+00
6.73703492e-01 3.98731530e-01 2.62818068e-01 9.73632455e-01
-6.24894798e-01 2.42049158e-01 -3.31693500e-01 -5.15971720e-01
8.28318000e-01 1.15219498e+00 -3.62135947e-01 -8.06805313e-01
-5.15351474e-01 9.69463646e-01 4.59446222e-01 5.38607538e-01
-7.38252997e-01 -4.65326488e-01 5.97890854e-01 -2.41754130e-01
2.49273941e-01 -3.13962311e-01 -9.60046519e-03 -1.50639045e+00
-3.26959640e-01 -8.46058011e-01 2.18186706e-01 -8.42100203e-01
-1.60300970e+00 6.47569954e-01 -2.32184097e-01 -9.42427933e-01
-5.20405948e-01 -5.13136208e-01 -1.11286604e+00 1.22128677e+00
-1.48901331e+00 -1.16636121e+00 -3.57692949e-02 3.72584730e-01
1.07411659e+00 -2.94657499e-01 1.03610730e+00 -4.24853601e-02
-7.61848092e-01 6.13541782e-01 3.91407937e-01 4.46839392e-01
9.21135187e-01 -1.09263659e+00 1.10994804e+00 8.17186117e-01
1.31327495e-01 8.24411750e-01 3.10230166e-01 -8.02737355e-01
-1.25856531e+00 -1.27911544e+00 1.54365790e+00 -3.72344404e-01
8.68355930e-01 -7.57969141e-01 -1.11086249e+00 8.67301106e-01
5.30836761e-01 -2.20677719e-01 9.30897892e-01 1.75538987e-01
-1.36990130e-01 4.14073557e-01 -5.00200868e-01 8.06252539e-01
8.24278831e-01 -5.50524652e-01 -5.62828481e-01 5.63078165e-01
1.24759984e+00 -4.02880698e-01 -8.10090005e-01 -1.07711658e-01
1.56566396e-01 -4.66205269e-01 5.71053803e-01 -6.02856934e-01
7.14301169e-01 4.51023996e-01 3.32296908e-01 -1.56883001e+00
-3.31941992e-01 -9.28161263e-01 -1.62794873e-01 1.32387817e+00
8.95276904e-01 -6.24066591e-01 7.02517331e-01 6.97984040e-01
-3.77834380e-01 -7.90900469e-01 -6.34050488e-01 -7.70642340e-01
6.87947869e-01 -4.11707848e-01 6.11574173e-01 9.81739402e-01
3.20909828e-01 8.09823275e-01 -7.51825154e-01 -4.64412123e-01
1.13533229e-01 2.79996306e-01 8.28148544e-01 -7.90731192e-01
-5.36131144e-01 -3.37238729e-01 5.65688312e-01 -1.46239543e+00
4.14755195e-01 -9.82662022e-01 2.69801676e-01 -1.46175075e+00
3.33281577e-01 -5.46591103e-01 9.15449634e-02 6.82359457e-01
-5.75037003e-01 -1.74558163e-01 3.27852875e-01 1.76390558e-01
-6.09541595e-01 7.27071941e-01 1.39842272e+00 1.19571418e-01
-2.65833974e-01 1.16660252e-01 -8.92587721e-01 2.89261073e-01
1.18092799e+00 -4.47002798e-01 -4.07091022e-01 -7.59951293e-01
9.91780087e-02 5.63291088e-02 -2.32649833e-01 -4.55408871e-01
4.28330004e-01 -2.51665443e-01 8.44973028e-02 -4.39574599e-01
3.44954699e-01 1.66521579e-01 -3.51079106e-01 4.01739895e-01
-5.47009528e-01 -3.71834673e-02 6.92328885e-02 3.07048410e-01
-2.03331754e-01 -3.45544070e-01 4.23657864e-01 -5.79925418e-01
-8.73945415e-01 5.77812195e-01 -3.60596746e-01 5.13664007e-01
6.61026001e-01 2.54437268e-01 -7.32021213e-01 -6.33043587e-01
-4.98376936e-01 4.16155428e-01 1.60649642e-01 6.47429168e-01
6.61298752e-01 -1.09726489e+00 -1.20503390e+00 1.76391870e-01
1.10654630e-01 8.19521770e-02 4.74532098e-01 6.81806326e-01
-3.02636802e-01 7.99823225e-01 2.25624844e-01 -1.08376741e-01
-9.91548181e-01 4.10648137e-01 -1.90416172e-01 -5.44733047e-01
-7.68119872e-01 1.02166259e+00 1.73926055e-01 -4.68424022e-01
-4.28732000e-02 -2.46582821e-01 6.76347092e-02 -1.81491986e-01
7.71623909e-01 1.82966098e-01 -2.66273588e-01 -2.09712699e-01
3.64002138e-02 -2.68121045e-02 -2.85941780e-01 -1.96210667e-01
1.37557626e+00 -7.72956237e-02 -5.70344478e-02 4.95123714e-01
1.31015849e+00 -4.24926206e-02 -1.14381742e+00 -4.61472690e-01
4.72053373e-03 -2.76576787e-01 -1.89734936e-01 -6.58664703e-01
-6.43549204e-01 1.32645547e+00 -3.99180412e-01 -1.86696813e-01
6.71649933e-01 1.07667066e-01 1.46321952e+00 6.24259949e-01
3.98243845e-01 -1.28680968e+00 3.29562008e-01 1.14070594e+00
6.30812228e-01 -1.32904494e+00 -5.05675077e-01 -2.07281426e-01
-1.01211631e+00 9.29633737e-01 8.21755826e-01 4.80647795e-02
5.97120583e-01 4.07284588e-01 9.17996392e-02 2.37613738e-01
-1.42008996e+00 -9.46453065e-02 -1.25454918e-01 3.23143512e-01
8.77107799e-01 3.79432052e-01 -3.17525417e-01 9.00764763e-01
-4.51272070e-01 -7.13751744e-03 6.33924484e-01 8.23560476e-01
-6.42887890e-01 -1.24341178e+00 8.52667615e-02 4.22013342e-01
-4.24384445e-01 -8.65461230e-01 -1.60475910e-01 5.96010029e-01
-3.12285811e-01 9.96780038e-01 1.40071318e-01 2.49755047e-02
-2.14625418e-01 2.91345537e-01 -5.95712662e-02 -1.23264718e+00
-5.00962615e-01 3.23113292e-01 2.62023419e-01 -2.19661951e-01
1.22280158e-01 -3.44187617e-01 -1.30488479e+00 -6.80032372e-01
-1.90209568e-01 3.70639622e-01 3.86014283e-01 8.38745713e-01
4.46767926e-01 2.64105141e-01 4.68155593e-01 -4.74357873e-01
-5.96157610e-01 -1.49380469e+00 -4.49087769e-01 -1.16203809e-02
-2.30377093e-01 6.15513176e-02 -2.63987422e-01 -4.17019427e-03]
|
[11.513409614562988, 9.0104398727417]
|
22c6b52a-2585-4a72-92df-669cfc428464
|
learning-dynamic-relationships-for-3d-human
| null | null |
http://openaccess.thecvf.com/content_CVPR_2020/html/Cui_Learning_Dynamic_Relationships_for_3D_Human_Motion_Prediction_CVPR_2020_paper.html
|
http://openaccess.thecvf.com/content_CVPR_2020/papers/Cui_Learning_Dynamic_Relationships_for_3D_Human_Motion_Prediction_CVPR_2020_paper.pdf
|
Learning Dynamic Relationships for 3D Human Motion Prediction
|
3D human motion prediction, i.e., forecasting future sequences from given historical poses, is a fundamental task for action analysis, human-computer interaction, machine intelligence. Recently, the state-of-the-art method assumes that the whole human motion sequence involves a fully-connected graph formed by links between each joint pair. Although encouraging performance has been made, due to the neglect of the inherent and meaningful characteristics of the natural connectivity of human joints, unexpected results may be produced. Moreover, such a complicated topology greatly increases the training difficulty. To tackle these issues, we propose a deep generative model based on graph networks and adversarial learning. Specifically, the skeleton pose is represented as a novel dynamic graph, in which natural connectivities of the joint pairs are exploited explicitly, and the links of geometrically separated joints can also be learned implicitly. Notably, in the proposed model, the natural connection strength is adaptively learned, whereas, in previous schemes, it was constant. Our approach is evaluated on two representations (i.e., angle-based, position-based) from various large-scale 3D skeleton benchmarks (e.g., H3.6M, CMU, 3DPW MoCap). Extensive experiments demonstrate that our approach achieves significant improvements against existing baselines in accuracy and visualization. Code will be available at https://github.com/cuiqiongjie/LDRGCN.
|
[' Fei Yang', ' Huaijiang Sun', 'Qiongjie Cui']
|
2020-06-01
| null | null | null |
cvpr-2020-6
|
['action-analysis']
|
['computer-vision']
|
[-2.97258906e-02 1.78107724e-01 -3.12753379e-01 5.35863861e-02
-1.11434139e-01 -3.78771752e-01 5.50481617e-01 -2.76425332e-01
-1.89568713e-01 6.57798052e-01 5.05626023e-01 2.04899372e-03
8.41633976e-02 -7.09853947e-01 -8.56180787e-01 -6.81542575e-01
-2.23283380e-01 3.75718534e-01 5.35518587e-01 -3.71768326e-01
-1.82816312e-02 4.97256964e-01 -9.88449097e-01 -2.25482315e-01
6.36490583e-01 6.12848341e-01 -1.97581455e-01 5.15219986e-01
1.53867334e-01 8.02538216e-01 -3.44595253e-01 -4.01181221e-01
2.71743238e-01 -4.87372041e-01 -6.64348364e-01 7.46551380e-02
2.91593879e-01 -3.12907398e-01 -1.04894805e+00 9.29796815e-01
4.44535553e-01 3.49498689e-01 6.42120600e-01 -1.46406162e+00
-3.56591552e-01 2.92974353e-01 -1.01191330e+00 5.36597073e-02
4.49359208e-01 5.34452021e-01 8.94333959e-01 -5.30005872e-01
9.27981973e-01 1.27943635e+00 6.43077135e-01 6.64969742e-01
-1.21916246e+00 -6.35203540e-01 3.68896455e-01 3.84287417e-01
-1.31729937e+00 -7.39911124e-02 1.16672981e+00 -4.63429064e-01
8.04821253e-01 -8.32667798e-02 1.06988835e+00 1.52650845e+00
5.39205134e-01 8.90728772e-01 5.58369279e-01 -9.32943970e-02
3.52707244e-02 -7.20438838e-01 -1.37798145e-01 9.97350633e-01
1.31443292e-01 -3.48220542e-02 -4.76715088e-01 -1.17952242e-01
1.17490804e+00 1.29480198e-01 -3.91134143e-01 -9.25463855e-01
-1.30298078e+00 6.77141428e-01 5.79217315e-01 5.66037595e-02
-3.23521793e-01 4.91097659e-01 5.14276922e-01 -7.40930736e-02
3.19737971e-01 -1.29762813e-01 -2.46852845e-01 -2.76131302e-01
-6.43693566e-01 4.48066980e-01 5.84453702e-01 9.09380019e-01
4.62162495e-01 1.56325012e-01 -4.62565534e-02 5.41393757e-01
3.29654098e-01 3.17982882e-01 5.83012700e-01 -1.05741549e+00
6.04427278e-01 5.43022633e-01 -6.72495961e-02 -1.42437112e+00
-5.49702585e-01 -2.33273447e-01 -1.10293329e+00 1.58756465e-01
6.22657359e-01 -1.76671118e-01 -9.38476324e-01 1.94621718e+00
4.98178184e-01 4.53802139e-01 -3.00641984e-01 1.11087871e+00
5.17825842e-01 4.91268784e-01 9.80975553e-02 1.54007733e-01
1.03041887e+00 -1.13588953e+00 -5.18526375e-01 -2.12107584e-01
3.21085602e-01 -4.04444605e-01 8.99541259e-01 2.06992105e-01
-1.04117298e+00 -6.39558017e-01 -9.84969258e-01 -1.34262532e-01
9.26556736e-02 -6.70584813e-02 6.61347151e-01 1.95324570e-01
-6.87748015e-01 7.71086574e-01 -1.45107841e+00 -4.55782264e-01
3.43980730e-01 1.43337771e-01 -4.66803461e-01 -3.74964625e-02
-1.16580915e+00 5.84034860e-01 2.91830868e-01 2.03928679e-01
-8.44218075e-01 -3.35106105e-01 -9.14293885e-01 -1.66945294e-01
4.63309199e-01 -1.03000093e+00 9.46348727e-01 -6.75982118e-01
-1.53676140e+00 5.67772031e-01 1.34311885e-01 -4.94463682e-01
1.01770830e+00 -5.45833826e-01 -1.25968307e-01 4.42538172e-01
-1.04006268e-01 7.49584436e-01 6.97953939e-01 -1.03901923e+00
-2.62355238e-01 -3.33864182e-01 2.48519272e-01 3.02268714e-01
-1.34007290e-01 -4.94191587e-01 -7.81922042e-01 -1.02657855e+00
2.09353730e-01 -1.28315592e+00 -4.66130614e-01 4.80775297e-01
-5.17487109e-01 -1.74593002e-01 7.00449824e-01 -8.90109003e-01
1.20506585e+00 -2.02200222e+00 6.16186738e-01 2.46275336e-01
1.96128294e-01 1.63098961e-01 -2.71262787e-02 4.23308223e-01
1.85335409e-02 2.84333434e-03 -3.69683832e-01 -1.50974989e-01
-1.03231426e-02 3.34676236e-01 9.47238356e-02 6.64179564e-01
1.62066549e-01 1.14327860e+00 -1.03493369e+00 -5.31627417e-01
2.96506375e-01 6.84379280e-01 -5.71114242e-01 1.03980787e-01
-1.91988930e-01 7.15503037e-01 -7.47438788e-01 5.36089122e-01
2.82910556e-01 -2.16126397e-01 2.77871072e-01 -3.24053317e-01
3.22721720e-01 1.02641448e-01 -1.09640133e+00 2.33531308e+00
-9.25098732e-02 3.87065768e-01 -3.43300819e-01 -1.05200708e+00
7.45281696e-01 2.61586845e-01 8.27402353e-01 -3.77349705e-01
1.09720998e-01 -1.76082194e-01 1.09614454e-01 -4.32703197e-01
2.27476284e-01 2.34151974e-01 3.03923525e-02 2.24645615e-01
-8.91080201e-02 3.61012556e-02 1.05618492e-01 3.12188983e-01
1.20597780e+00 7.29427218e-01 2.79382497e-01 2.42775530e-02
3.83144051e-01 -8.01513195e-02 8.63527834e-01 2.50884622e-01
-4.63861555e-01 8.75877798e-01 6.77297413e-01 -4.10877436e-01
-1.02356803e+00 -1.28342092e+00 4.31912661e-01 4.24005479e-01
3.58652472e-01 -4.45653200e-01 -7.74664581e-01 -8.07267129e-01
2.69806311e-02 3.63343358e-01 -6.64413214e-01 -5.23523808e-01
-1.06474090e+00 -4.48528022e-01 5.84742308e-01 7.61869252e-01
5.98022878e-01 -9.37665761e-01 -7.11452007e-01 3.14854503e-01
-3.27834666e-01 -1.13485801e+00 -7.17552423e-01 -5.88916957e-01
-1.03625321e+00 -1.23643613e+00 -8.21245730e-01 -5.52667379e-01
4.83337611e-01 2.16604173e-01 1.01292455e+00 2.17133746e-01
-3.11952472e-01 5.20491719e-01 -3.41614932e-01 1.80549487e-01
-7.19027966e-03 1.66172888e-02 1.39780454e-02 8.21509119e-03
7.41376653e-02 -1.05344748e+00 -9.66193318e-01 3.33982706e-01
-7.83874094e-01 3.05230558e-01 5.99528849e-01 8.48067880e-01
7.42433727e-01 -5.11438996e-02 3.21714938e-01 -7.38465846e-01
1.75870404e-01 -5.30970454e-01 -6.41006306e-02 7.43480697e-02
-1.98526859e-01 7.71098658e-02 4.62039888e-01 -6.58998251e-01
-9.24406469e-01 3.58444691e-01 -2.05095872e-01 -7.79654801e-01
-1.52001113e-01 3.86351496e-01 -4.82025892e-01 4.15281877e-02
4.65437800e-01 1.81655914e-01 1.86100051e-01 -4.67852980e-01
4.62461174e-01 -9.09670144e-02 6.38056636e-01 -6.66041672e-01
1.02236509e+00 5.97673714e-01 2.57429302e-01 -6.63152277e-01
-4.54109877e-01 -2.00066254e-01 -8.91539037e-01 -3.61804307e-01
8.75938118e-01 -7.91083753e-01 -4.97244209e-01 6.23624146e-01
-1.11049104e+00 -4.51018870e-01 -1.92263156e-01 6.43946648e-01
-8.07076991e-01 8.54459465e-01 -8.00902426e-01 -4.79765445e-01
-1.93325594e-01 -9.00461793e-01 8.27014983e-01 9.49987918e-02
-4.79707062e-01 -1.02522826e+00 1.87132016e-01 3.58144820e-01
-4.84623499e-02 1.09864783e+00 8.95840585e-01 -2.78812259e-01
-5.67179561e-01 -2.01312885e-01 1.11220084e-01 1.21534638e-01
2.82531321e-01 1.79020897e-01 -3.12466770e-01 -2.77595520e-01
-3.79641742e-01 -2.63526529e-01 8.07418466e-01 3.15377623e-01
1.13494539e+00 -3.24254960e-01 -3.87431562e-01 5.38507462e-01
1.01709557e+00 -5.84651753e-02 7.76923478e-01 2.20956832e-01
1.16387749e+00 5.34858048e-01 5.38307726e-01 4.83350784e-01
4.18067306e-01 7.59569943e-01 5.53000331e-01 6.90451041e-02
-3.72484326e-01 -6.49503708e-01 3.29873532e-01 7.30134964e-01
-6.45894170e-01 -1.44533217e-01 -8.60413969e-01 2.97404915e-01
-2.10782599e+00 -9.35959697e-01 -1.43904850e-01 2.02313352e+00
6.65950000e-01 5.29860198e-01 3.26059043e-01 1.81746423e-01
5.87659836e-01 4.84631419e-01 -8.50122571e-01 1.91878304e-01
2.71524377e-02 1.27882585e-01 2.63107181e-01 2.70323128e-01
-9.96087611e-01 9.35199916e-01 4.80919695e+00 7.73583651e-01
-8.84307206e-01 -3.52729559e-02 3.95476490e-01 -2.43874669e-01
-1.51899412e-01 -4.94725853e-02 -3.49513263e-01 4.48909223e-01
4.30874020e-01 -9.78825167e-02 1.67259485e-01 7.11763084e-01
2.49452770e-01 7.33080655e-02 -9.96662736e-01 7.57295966e-01
-9.18027572e-03 -1.07862639e+00 1.81288004e-01 1.94000378e-02
5.88533580e-01 -2.04946935e-01 -9.40820202e-02 9.48750749e-02
1.87928781e-01 -8.04659188e-01 7.82080114e-01 7.36370325e-01
5.18899083e-01 -9.40429688e-01 4.12129790e-01 4.28365380e-01
-1.40777743e+00 2.77270079e-01 -1.05925404e-01 -3.72450240e-02
3.89787555e-01 3.70638072e-01 -2.87011564e-01 8.98576379e-01
6.17092609e-01 1.05681825e+00 -4.24878538e-01 9.70116079e-01
-5.32065928e-01 6.33860290e-01 -2.22560301e-01 1.80985749e-01
1.56567141e-01 -2.59763420e-01 7.13861346e-01 9.56323087e-01
1.99870303e-01 2.26538569e-01 3.20642799e-01 6.25836492e-01
1.40482755e-02 -6.53270781e-02 -6.04256928e-01 2.34488547e-02
4.47854474e-02 1.04447186e+00 -7.60717809e-01 -5.50507717e-02
-4.02357757e-01 1.10973632e+00 4.30770516e-01 5.66411972e-01
-1.05325747e+00 -4.89587039e-02 8.52242768e-01 2.84770638e-01
2.15161711e-01 -8.03991616e-01 8.65356401e-02 -1.32782030e+00
2.48234183e-01 -7.97084391e-01 4.00360942e-01 -5.04571915e-01
-1.17278063e+00 2.59086400e-01 5.96452095e-02 -1.34779859e+00
-4.17926013e-01 -3.76983374e-01 -7.50452459e-01 4.93920416e-01
-1.08030176e+00 -1.17738128e+00 -3.67675960e-01 7.23202944e-01
5.65273225e-01 1.02592379e-01 4.17549223e-01 2.26609215e-01
-6.58984125e-01 5.33517122e-01 -2.62758493e-01 5.17274082e-01
5.47459364e-01 -1.08737373e+00 7.14052618e-01 7.88366735e-01
3.30174774e-01 4.14725274e-01 5.78319311e-01 -9.01444733e-01
-1.45437360e+00 -1.11109948e+00 4.53261703e-01 -4.04596210e-01
9.29099321e-01 -2.41125181e-01 -1.07945061e+00 7.99504459e-01
-8.17947835e-02 2.25970075e-01 3.95233691e-01 -3.94866228e-01
-1.96145207e-01 1.46734163e-01 -7.49152482e-01 9.12323654e-01
1.69276571e+00 -7.90023282e-02 -4.77764279e-01 2.96203882e-01
6.39229596e-01 -6.70610607e-01 -8.47111046e-01 4.30175215e-01
8.32155943e-01 -8.93748224e-01 1.24165738e+00 -8.47194493e-01
6.36424422e-01 -4.30275381e-01 1.92045629e-01 -1.18012047e+00
-3.09959471e-01 -6.57573342e-01 -6.74319327e-01 8.89101505e-01
2.36025453e-02 -3.68936092e-01 1.14471459e+00 5.92817307e-01
-1.15661867e-01 -1.04049587e+00 -1.05672657e+00 -8.89122665e-01
8.18169340e-02 -1.40310228e-01 3.38493645e-01 8.33042085e-01
-2.85777986e-01 2.36783639e-01 -7.28500187e-01 1.46026939e-01
6.80087030e-01 -9.17221457e-02 1.08578134e+00 -1.00853598e+00
-6.20760262e-01 -4.40223962e-01 -8.03534031e-01 -1.47182024e+00
3.26184511e-01 -6.03606045e-01 -2.73843467e-01 -1.55457544e+00
-8.65147188e-02 -1.54073134e-01 -1.58721089e-01 4.50959086e-01
-2.23320141e-01 1.20921098e-01 2.73831993e-01 2.76629031e-01
-3.96701425e-01 9.12034214e-01 1.70230913e+00 -2.32596666e-01
-1.96432009e-01 1.20365314e-01 -1.76100899e-02 1.09774733e+00
8.98256481e-01 -3.54650885e-01 -5.19683063e-01 -4.00935173e-01
-5.21211959e-02 2.26963356e-01 6.63665533e-01 -1.09908712e+00
2.68211514e-01 -3.00435930e-01 4.72410738e-01 -4.42885995e-01
6.17350399e-01 -7.60608971e-01 4.13676530e-01 8.09465706e-01
-1.83975250e-01 7.56978616e-02 -5.32644540e-02 9.88466442e-01
-1.20834194e-01 2.10112378e-01 4.79611635e-01 -1.74550906e-01
-6.74782872e-01 7.04113126e-01 -7.52519295e-02 2.45167613e-01
9.01874304e-01 -2.73841828e-01 -9.53031296e-04 -4.25978690e-01
-7.26727307e-01 2.39395350e-01 5.15489578e-01 4.99885082e-01
6.98862076e-01 -1.51623321e+00 -6.02547407e-01 -2.64308244e-01
-4.94213179e-02 2.83759356e-01 3.69325101e-01 9.51846004e-01
-6.50972426e-01 -7.60864615e-02 -5.06857991e-01 -5.56762576e-01
-1.01256442e+00 4.76036638e-01 2.17654034e-01 -4.33056951e-01
-1.05390298e+00 5.34913182e-01 2.52842486e-01 -6.63127005e-02
3.01750928e-01 -2.31633455e-01 -1.58614174e-01 -3.61922503e-01
1.10348150e-01 6.00352883e-01 -3.58828962e-01 -8.45123529e-01
-3.10255796e-01 8.32009256e-01 1.35087356e-01 -4.84915962e-03
1.17983556e+00 9.29524824e-02 2.38931060e-01 4.48599100e-01
1.04391158e+00 -9.88163874e-02 -1.68032014e+00 -1.93697006e-01
-2.08837256e-01 -4.72391725e-01 -4.20169353e-01 -2.70917088e-01
-1.47513974e+00 9.94203031e-01 3.40557218e-01 -1.95724681e-01
9.96824682e-01 -1.50497302e-01 1.12411833e+00 2.48366058e-01
5.25139451e-01 -8.69780362e-01 4.48796004e-01 2.82651901e-01
9.00658190e-01 -1.10704422e+00 1.80224597e-01 -5.12981176e-01
-5.60111344e-01 1.18173933e+00 8.21348846e-01 -4.41986412e-01
6.69368267e-01 -2.24538386e-01 -2.12258548e-01 -4.83967103e-02
-4.18842852e-01 -9.75671783e-03 4.35098767e-01 5.65861404e-01
3.96690577e-01 -3.00830025e-02 -4.48485702e-01 4.81078357e-01
-1.70107260e-01 -1.82101980e-01 2.92864591e-01 9.90042925e-01
-1.52396768e-01 -1.05744863e+00 -1.04304977e-01 9.26593319e-02
-3.49070877e-01 2.80020654e-01 -4.87523377e-01 1.15958059e+00
3.51851098e-02 5.77603400e-01 -1.46910146e-01 -5.07231653e-01
5.72736859e-01 -1.32192850e-01 4.87213135e-01 -3.13049376e-01
-2.07538784e-01 5.89823388e-02 -6.01452403e-02 -9.20833230e-01
-5.33482373e-01 -7.65464067e-01 -1.54246259e+00 -2.93587565e-01
1.33946478e-01 -1.01935044e-01 2.84144729e-01 6.57516718e-01
3.72216940e-01 5.17114341e-01 3.34121972e-01 -1.01798153e+00
-4.91121590e-01 -8.32222402e-01 -4.96715397e-01 7.57048845e-01
1.05187282e-01 -1.06864262e+00 -2.32578933e-01 2.35987261e-01]
|
[7.4229302406311035, -0.21662938594818115]
|
089aae1e-8ff6-4de8-a712-da8c018b3585
|
h-denseunet-hybrid-densely-connected-unet-for
|
1709.07330
| null |
http://arxiv.org/abs/1709.07330v3
|
http://arxiv.org/pdf/1709.07330v3.pdf
|
H-DenseUNet: Hybrid Densely Connected UNet for Liver and Tumor Segmentation from CT Volumes
|
Liver cancer is one of the leading causes of cancer death. To assist doctors
in hepatocellular carcinoma diagnosis and treatment planning, an accurate and
automatic liver and tumor segmentation method is highly demanded in the
clinical practice. Recently, fully convolutional neural networks (FCNs),
including 2D and 3D FCNs, serve as the back-bone in many volumetric image
segmentation. However, 2D convolutions can not fully leverage the spatial
information along the third dimension while 3D convolutions suffer from high
computational cost and GPU memory consumption. To address these issues, we
propose a novel hybrid densely connected UNet (H-DenseUNet), which consists of
a 2D DenseUNet for efficiently extracting intra-slice features and a 3D
counterpart for hierarchically aggregating volumetric contexts under the spirit
of the auto-context algorithm for liver and tumor segmentation. We formulate
the learning process of H-DenseUNet in an end-to-end manner, where the
intra-slice representations and inter-slice features can be jointly optimized
through a hybrid feature fusion (HFF) layer. We extensively evaluated our
method on the dataset of MICCAI 2017 Liver Tumor Segmentation (LiTS) Challenge
and 3DIRCADb Dataset. Our method outperformed other state-of-the-arts on the
segmentation results of tumors and achieved very competitive performance for
liver segmentation even with a single model.
|
['Chi-Wing Fu', 'Xiaojuan Qi', 'Qi Dou', 'Hao Chen', 'Pheng Ann Heng', 'Xiaomeng Li']
|
2017-09-21
| null | null | null | null |
['liver-segmentation', 'automatic-liver-and-tumor-segmentation']
|
['medical', 'medical']
|
[-3.03047299e-01 -7.17777982e-02 -2.98653185e-01 -5.62212586e-01
-8.35702717e-01 -3.25555027e-01 5.51587343e-01 1.59122229e-01
-2.13576779e-01 2.82360852e-01 4.51553285e-01 -4.51369971e-01
1.12417586e-01 -8.50128055e-01 -4.21452403e-01 -9.30775583e-01
-3.09679180e-01 5.33751428e-01 8.54573324e-02 1.75257176e-01
-2.87110656e-01 6.50567234e-01 -5.93508720e-01 2.78389335e-01
1.14479578e+00 1.17078030e+00 1.11383311e-01 5.24724364e-01
-4.18723166e-01 4.95397359e-01 7.01229945e-02 3.01429313e-02
4.05890733e-01 -4.73177791e-01 -9.13930297e-01 3.64393771e-01
2.32850716e-01 -2.69505352e-01 -3.88456970e-01 1.03299177e+00
5.24875343e-01 -4.03894961e-01 6.93412781e-01 -7.57854044e-01
-3.96563321e-01 7.30190277e-01 -7.77693808e-01 3.93754274e-01
-2.52393246e-01 4.13727343e-01 3.79764199e-01 -1.12552190e+00
3.35028321e-01 8.21109235e-01 9.35472846e-01 3.78987938e-01
-1.08902013e+00 -4.54728812e-01 -1.29570141e-01 -1.76253110e-01
-1.48945546e+00 -8.59985407e-03 5.66678345e-01 -5.01800478e-01
7.09773123e-01 2.75309324e-01 1.26213658e+00 4.63358313e-01
4.59219545e-01 9.87750232e-01 1.07585013e+00 -1.17532648e-02
-1.72649965e-01 -2.49176994e-01 1.32354930e-01 1.19758046e+00
1.41411588e-01 1.64475650e-01 3.12767513e-02 -1.59787700e-01
1.10162246e+00 3.17438811e-01 -6.40297174e-01 -5.46831012e-01
-1.66600764e+00 9.73049343e-01 1.12853146e+00 3.99537504e-01
-6.01956367e-01 2.50518590e-01 5.22818685e-01 -1.56624556e-01
5.67702651e-01 2.70891227e-02 -4.80486274e-01 4.19123083e-01
-1.14585292e+00 -3.29724029e-02 8.47424626e-01 6.71282411e-01
4.82544929e-01 -1.46842539e-01 -7.34447479e-01 4.50754911e-01
4.41611320e-01 1.80822238e-01 6.83495998e-01 -2.63218403e-01
-5.23679517e-02 8.65154207e-01 -3.81524593e-01 -3.72658670e-01
-8.17603350e-01 -8.77460182e-01 -1.55585229e+00 -2.43610695e-01
5.18281996e-01 -6.12487309e-02 -1.38797891e+00 1.25394940e+00
7.06415415e-01 6.35071158e-01 -1.48613647e-01 1.29261839e+00
1.30142248e+00 3.06477636e-01 3.64826590e-01 -1.06379554e-01
1.69282722e+00 -1.35883915e+00 -4.90723044e-01 3.00266743e-01
1.19469857e+00 -6.54456913e-01 7.48312771e-01 -1.79294780e-01
-1.01756048e+00 -9.81375501e-02 -6.69956148e-01 -1.31499946e-01
-8.57105702e-02 2.40891382e-01 1.04406214e+00 4.81349140e-01
-1.23125029e+00 3.41489881e-01 -1.27288377e+00 -2.23717943e-01
9.40344989e-01 4.94309247e-01 -2.65213519e-01 -1.68092802e-01
-8.77010405e-01 7.24368215e-01 3.44762594e-01 2.21588075e-01
-1.14628971e+00 -1.39258432e+00 -1.05440629e+00 1.17824182e-01
2.43761972e-01 -1.10776842e+00 1.17673957e+00 -6.24166906e-01
-1.35169327e+00 8.70223641e-01 1.57042295e-02 -5.51030695e-01
5.90747058e-01 2.63434410e-01 2.05153033e-01 2.74695873e-01
6.40093014e-02 8.09979260e-01 2.81083554e-01 -7.64477253e-01
-4.04953122e-01 -5.16199350e-01 -5.46006083e-01 3.01909089e-01
6.37225658e-02 -3.21096539e-01 -6.34581387e-01 -7.55420029e-01
1.98876619e-01 -7.50117719e-01 -6.30654037e-01 3.53356272e-01
-5.79501510e-01 -2.24764138e-01 1.08210707e+00 -8.47213149e-01
9.00743365e-01 -1.79957819e+00 3.87064874e-01 1.58627182e-01
5.72564244e-01 1.69279903e-01 1.25766486e-01 -4.65534151e-01
-8.22939258e-03 1.42185077e-01 -4.30890501e-01 -3.19794357e-01
-1.20371893e-01 9.11648721e-02 3.03382456e-01 9.00869846e-01
5.73490709e-02 1.50589311e+00 -9.70946610e-01 -8.96453500e-01
4.52591628e-01 8.52215350e-01 -6.02543712e-01 2.62081116e-01
-2.01048981e-02 7.83529341e-01 -5.82855165e-01 9.51975465e-01
6.11157537e-01 -7.47717381e-01 -2.05606036e-02 -5.61726511e-01
-1.71623722e-01 -7.25739375e-02 -3.12809587e-01 2.05517220e+00
-3.63366544e-01 2.02405274e-01 2.11906016e-01 -9.16150451e-01
4.09945160e-01 5.26815772e-01 1.08962393e+00 -5.69114149e-01
3.48386854e-01 3.51609379e-01 4.43488732e-03 -4.71496791e-01
-2.38829359e-01 -1.93097055e-01 4.78473455e-02 1.03232563e-01
6.04770668e-02 -3.39800805e-01 -2.63240412e-02 1.65887162e-01
8.40738893e-01 -6.13253228e-02 4.68177259e-01 -8.98827195e-01
6.22566938e-01 1.93301395e-01 5.97089052e-01 2.76561469e-01
-3.47788334e-01 6.19077206e-01 5.17280102e-01 -8.86251628e-01
-7.99516261e-01 -8.55941653e-01 -3.58554721e-01 4.78006840e-01
1.69747576e-01 -1.62602499e-01 -7.68109620e-01 -1.19517350e+00
1.66756079e-01 7.40036666e-02 -8.40387762e-01 1.77897215e-01
-6.39814556e-01 -1.23622072e+00 4.67475533e-01 5.33946812e-01
7.69684196e-01 -5.36823928e-01 -6.95501804e-01 3.14284950e-01
-3.10263447e-02 -9.15022612e-01 -9.07526731e-01 2.75385171e-01
-9.71471667e-01 -1.33579946e+00 -1.21509647e+00 -9.35718834e-01
8.09922993e-01 3.90868694e-01 1.36471629e+00 4.83212352e-01
-7.62675345e-01 1.03358343e-01 7.12687522e-03 -1.99721053e-01
-1.72640473e-01 2.00060695e-01 -5.50676346e-01 -2.11112857e-01
1.62473544e-01 -3.42486709e-01 -1.14691055e+00 1.55752018e-01
-7.11589634e-01 5.77705920e-01 8.72732997e-01 1.18925095e+00
8.77682686e-01 -1.24010548e-01 2.90581971e-01 -8.96330237e-01
1.58508658e-01 -6.16124511e-01 -5.52715302e-01 2.66194433e-01
-3.82680923e-01 -8.24374110e-02 4.39223737e-01 -1.05876833e-01
-6.90161288e-01 3.85780096e-01 -1.85817644e-01 -5.13696969e-01
9.48901102e-03 7.85958827e-01 1.14616074e-01 -5.54554522e-01
2.68114567e-01 4.58620697e-01 1.82140529e-01 -3.49692762e-01
1.51296735e-01 1.75273269e-01 3.89512658e-01 -3.26638550e-01
5.32243907e-01 4.87408131e-01 3.10234904e-01 -4.42262948e-01
-8.01855743e-01 -6.02909327e-01 -6.48715675e-01 -5.13756461e-02
1.07141769e+00 -1.07384813e+00 -4.55069900e-01 7.48907447e-01
-9.08845961e-01 -5.54615915e-01 -3.53899002e-01 5.52383244e-01
-3.01029235e-01 3.63762945e-01 -1.00635672e+00 4.61183786e-02
-8.33551705e-01 -1.89570034e+00 1.21237504e+00 4.38078344e-01
4.06815171e-01 -1.27989149e+00 -2.17436880e-01 3.80360559e-02
8.46435428e-01 5.69382191e-01 1.06046915e+00 -5.91042459e-01
-7.33690202e-01 -2.32605333e-03 -7.63925314e-01 -1.05242230e-01
1.79919675e-01 -4.16039079e-01 -4.91820514e-01 -5.38017869e-01
-2.25653574e-01 -1.93348914e-01 1.11343563e+00 8.88587892e-01
1.48761499e+00 -3.27946424e-01 -5.67438662e-01 1.36735427e+00
1.61241877e+00 -2.72119194e-01 1.77422926e-01 -1.46205947e-01
9.10003126e-01 -1.96000971e-02 -7.51598105e-02 3.26700360e-01
6.99531555e-01 4.73402292e-01 5.90040624e-01 -8.32767069e-01
-4.53174293e-01 2.12872609e-01 -2.99799919e-01 7.06234276e-01
9.36922580e-02 2.94630416e-02 -1.21290457e+00 7.00455964e-01
-1.48275614e+00 -3.34179878e-01 -2.12862492e-01 1.76687753e+00
9.82702255e-01 -4.41877246e-01 -1.10994361e-01 -3.22011381e-01
4.40373152e-01 -2.81139668e-02 -5.41044652e-01 3.33720148e-01
1.27732500e-01 1.58419028e-01 6.21909678e-01 3.65017027e-01
-1.47580457e+00 5.56660354e-01 5.21098518e+00 8.18228364e-01
-1.41631246e+00 3.20381910e-01 1.23384678e+00 1.71479791e-01
-2.11972505e-01 -2.05946505e-01 -5.15456080e-01 3.62423509e-01
4.45384264e-01 1.30586969e-02 9.96928141e-02 7.00604260e-01
1.26578495e-01 -1.83318198e-01 -9.87098873e-01 8.82352769e-01
-1.90475792e-01 -1.71921134e+00 3.13987955e-02 2.98018038e-01
9.54078436e-01 5.35879791e-01 -8.57696086e-02 3.38193089e-01
2.47817054e-01 -1.28255868e+00 2.83034563e-01 4.01418298e-01
1.05941093e+00 -5.90218961e-01 1.05335021e+00 2.68923074e-01
-1.44400477e+00 2.28164941e-01 -2.64302474e-02 6.29772544e-01
3.61640789e-02 8.88064086e-01 -1.09433782e+00 7.16134667e-01
4.45343465e-01 8.74068737e-01 -5.74306130e-01 1.52443194e+00
2.01819167e-01 4.82571751e-01 -4.38172609e-01 3.26125920e-01
6.25470340e-01 -1.12637281e-01 4.15921330e-01 1.48345244e+00
5.06440103e-01 2.50569433e-01 6.31133676e-01 8.60027134e-01
-3.81539911e-01 1.68647673e-02 3.44271958e-03 1.56951651e-01
1.25112636e-02 1.69664752e+00 -1.11719739e+00 -5.45239508e-01
-5.15463412e-01 8.08744371e-01 5.88793829e-02 9.27314311e-02
-9.13513422e-01 1.34144470e-01 2.71936417e-01 1.26380116e-01
1.51954323e-01 3.66390385e-02 -3.37065399e-01 -1.42771149e+00
-4.91284966e-01 -5.15055537e-01 5.46961069e-01 -3.01776882e-02
-1.37550521e+00 6.27576947e-01 -2.53198534e-01 -1.02730715e+00
1.08369209e-01 -3.66086274e-01 -7.13780582e-01 9.94813800e-01
-2.10135794e+00 -1.54666007e+00 -8.78864110e-01 6.47540689e-01
6.18350804e-01 1.00635618e-01 7.23972380e-01 2.58270979e-01
-4.80217397e-01 4.19531524e-01 -2.52407521e-01 5.98417103e-01
2.82766789e-01 -1.47369075e+00 4.62994166e-02 5.16944766e-01
-3.08046103e-01 1.33343488e-01 1.30764827e-01 -5.74332297e-01
-1.69077742e+00 -1.58245337e+00 5.47022879e-01 5.92403673e-02
2.80900985e-01 1.80849340e-02 -8.05005670e-01 5.51423311e-01
3.22558999e-01 1.06446242e+00 7.19741642e-01 -4.53969240e-01
7.10390061e-02 3.03454578e-01 -1.41973984e+00 2.53008217e-01
7.56395340e-01 -4.57890630e-02 -1.68610215e-01 6.46293104e-01
7.72888541e-01 -1.10434914e+00 -1.31994808e+00 7.36517072e-01
2.41143838e-01 -8.31071734e-01 1.15560567e+00 -2.79728591e-01
3.05543602e-01 -3.48904371e-01 1.29606664e-01 -1.27894115e+00
-3.99351537e-01 -2.92603672e-01 -2.03289211e-01 3.25458676e-01
1.09721422e-01 -3.44771177e-01 8.34782779e-01 3.51539135e-01
-7.03881323e-01 -1.56340945e+00 -9.85073626e-01 -2.38311380e-01
3.33018959e-01 2.78282166e-03 6.93776906e-01 9.00434792e-01
-2.28688121e-01 -1.00663938e-01 2.20567673e-01 1.24550745e-01
9.47385728e-01 3.12566847e-01 2.89957702e-01 -1.04915917e+00
5.57039864e-02 -8.37378383e-01 -4.12293136e-01 -1.07735527e+00
-1.11529291e-01 -1.43084478e+00 -1.99423566e-01 -1.68779969e+00
7.36537099e-01 -5.99921167e-01 -2.80174464e-01 6.12241209e-01
-2.45093197e-01 5.10064840e-01 6.25312179e-02 2.10871175e-01
-5.21405518e-01 5.25154233e-01 1.86878192e+00 -4.56695229e-01
-1.18114002e-01 -1.57886416e-01 -4.50312316e-01 7.12840796e-01
4.41428453e-01 -1.02172345e-01 -5.37021644e-02 -4.99539047e-01
-4.25773919e-01 5.62167525e-01 7.32434332e-01 -9.27088976e-01
4.75792229e-01 -5.63815534e-02 9.93571043e-01 -9.42090213e-01
-1.59162313e-01 -9.02040362e-01 4.92679365e-02 7.42308617e-01
-1.20697901e-01 -2.45121732e-01 8.86377320e-02 3.77642870e-01
-3.29235137e-01 3.90118033e-01 1.05944359e+00 -5.02059460e-01
-3.33903909e-01 1.14161372e+00 -1.10039331e-01 -1.00977987e-01
1.25729334e+00 3.72782387e-02 -1.53509853e-02 2.54389614e-01
-7.03123987e-01 6.67686403e-01 2.82484233e-01 -7.18202963e-02
6.88241541e-01 -1.40274858e+00 -9.95741427e-01 4.06011462e-01
-2.25851312e-01 9.71283734e-01 5.54254949e-01 1.73833966e+00
-8.16223502e-01 6.98225379e-01 2.58012891e-01 -1.04877853e+00
-8.14225018e-01 3.47342551e-01 1.04429328e+00 -9.47893739e-01
-1.04900599e+00 1.04265261e+00 6.51739895e-01 -3.49899083e-01
2.34318376e-01 -6.93212867e-01 -2.08918881e-02 -2.81700045e-01
4.60040927e-01 -1.94017097e-01 2.57037878e-01 -5.91687500e-01
-3.45554918e-01 4.15922254e-01 -1.83449805e-01 5.82604349e-01
1.36051464e+00 7.56405741e-02 -3.64010543e-01 -3.01299363e-01
1.35989249e+00 -4.53717053e-01 -1.32387197e+00 -5.03711939e-01
-5.37341684e-02 -3.13239753e-01 7.45942473e-01 -1.01270568e+00
-1.88888383e+00 7.57749259e-01 5.58710814e-01 -1.87999815e-01
1.30301154e+00 2.06423476e-02 1.25487423e+00 -1.64683476e-01
2.37540931e-01 -2.57140011e-01 -1.89108416e-01 3.06657612e-01
6.80253625e-01 -1.31301010e+00 1.09063089e-01 -6.60665154e-01
-4.06624645e-01 1.39556777e+00 4.97201294e-01 -1.45273477e-01
9.38881993e-01 5.32167852e-01 9.85977938e-04 -4.38199371e-01
-5.11449814e-01 -9.42298323e-02 4.63810116e-01 9.99242663e-02
6.22399092e-01 3.86671960e-01 -2.04895332e-01 7.13472545e-01
3.26039940e-01 2.61462867e-01 3.65683921e-02 7.62044132e-01
-2.25076243e-01 -7.03699887e-01 -1.58133611e-01 6.83969855e-01
-5.09400606e-01 -2.88481265e-01 1.13967329e-01 9.62445617e-01
1.99925974e-01 5.37800230e-02 6.58637136e-02 4.38432172e-02
-9.48681086e-02 -1.22805074e-01 4.82549608e-01 -4.97534811e-01
-9.69041288e-01 5.38188100e-01 -5.18777907e-01 -6.20544791e-01
-3.91594112e-01 -5.47703922e-01 -1.50941265e+00 -9.22667906e-02
-2.71070272e-01 1.73252091e-01 6.81677341e-01 8.87920737e-01
2.29765773e-01 6.89911962e-01 7.02570975e-01 -1.04673553e+00
-6.43513620e-01 -9.56786990e-01 -3.89028519e-01 1.99228153e-01
4.25302774e-01 -4.02593195e-01 -1.10684372e-01 -9.82774571e-02]
|
[14.681790351867676, -2.531067371368408]
|
85d40717-f7c0-442a-8112-1f0a71e9210d
|
masked-autoencoders-as-the-unified-learners
|
2208.00231
| null |
https://arxiv.org/abs/2208.00231v1
|
https://arxiv.org/pdf/2208.00231v1.pdf
|
Masked Autoencoders As The Unified Learners For Pre-Trained Sentence Representation
|
Despite the progresses on pre-trained language models, there is a lack of unified frameworks for pre-trained sentence representation. As such, it calls for different pre-training methods for specific scenarios, and the pre-trained models are likely to be limited by their universality and representation quality. In this work, we extend the recently proposed MAE style pre-training strategy, RetroMAE, such that it may effectively support a wide variety of sentence representation tasks. The extended framework consists of two stages, with RetroMAE conducted throughout the process. The first stage performs RetroMAE over generic corpora, like Wikipedia, BookCorpus, etc., from which the base model is learned. The second stage takes place on domain-specific data, e.g., MS MARCO and NLI, where the base model is continuingly trained based on RetroMAE and contrastive learning. The pre-training outputs at the two stages may serve different applications, whose effectiveness are verified with comprehensive experiments. Concretely, the base model are proved to be effective for zero-shot retrieval, with remarkable performances achieved on BEIR benchmark. The continuingly pre-trained models further benefit more downstream tasks, including the domain-specific dense retrieval on MS MARCO, Natural Questions, and the sentence embeddings' quality for standard STS and transfer tasks in SentEval. The empirical insights of this work may inspire the future design of sentence representation pre-training. Our pre-trained models and source code will be released to the public communities.
|
['Samuel Yang', 'Alexander Liu']
|
2022-07-30
| null | null | null | null |
['natural-questions']
|
['miscellaneous']
|
[ 3.55103612e-01 -1.64509922e-01 -1.73600152e-01 -3.15258056e-01
-1.04385602e+00 -3.12645495e-01 8.01574588e-01 3.54524910e-01
-6.00670516e-01 6.30947948e-01 4.90765959e-01 -2.94616729e-01
-2.61128634e-01 -9.21179414e-01 -3.30003411e-01 -4.49038804e-01
9.75565240e-02 3.35654527e-01 2.85472989e-01 -9.11478162e-01
4.23832804e-01 1.69368550e-01 -1.35402441e+00 4.84809726e-01
9.07197416e-01 9.31369245e-01 6.21944010e-01 6.32884085e-01
-4.35053378e-01 5.59034646e-01 -6.00764930e-01 -6.84034705e-01
-1.29191175e-01 -2.07224473e-01 -8.40417504e-01 -2.79101372e-01
7.77945621e-04 -1.87228844e-01 -5.69168031e-01 9.04570282e-01
7.53583610e-01 4.31836724e-01 7.27505267e-01 -6.91067994e-01
-1.09409082e+00 6.34332001e-01 -2.84520209e-01 4.22082901e-01
3.87500614e-01 9.57071781e-02 1.39236152e+00 -1.18576825e+00
5.46294689e-01 1.33488655e+00 4.38003391e-01 7.18053281e-01
-7.52942383e-01 -4.47204828e-01 -1.81392822e-02 2.81283408e-01
-1.28803957e+00 -5.29826283e-01 6.09987736e-01 -1.26932427e-01
1.14621818e+00 4.79391187e-01 1.49638861e-01 1.45642459e+00
1.22245371e-01 1.05107236e+00 7.82336831e-01 -6.30590320e-01
1.31775206e-02 2.76954293e-01 5.56602895e-01 3.44728380e-01
7.89283514e-02 -1.79337710e-01 -5.54337561e-01 3.08268815e-02
3.96749675e-01 2.22391725e-01 -4.10482883e-01 -3.07588428e-02
-9.51316953e-01 9.77402508e-01 3.91234666e-01 7.34824717e-01
-2.31570110e-01 -2.05821022e-01 8.45924258e-01 4.99671102e-01
6.97977126e-01 5.50810218e-01 -4.51042712e-01 -2.00320065e-01
-9.16175902e-01 1.04398839e-01 5.61920941e-01 9.76174414e-01
6.21215820e-01 -1.92101389e-01 -7.22259104e-01 1.40571404e+00
1.85744464e-01 3.07669699e-01 1.00375581e+00 -3.36550474e-01
9.67321098e-01 4.94841397e-01 -1.41048953e-01 -8.09222102e-01
-1.37072638e-01 -5.51535130e-01 -9.46791887e-01 -6.55475318e-01
-4.46725935e-02 -7.69020915e-02 -7.71607816e-01 1.55768013e+00
-7.69187286e-02 -9.97155905e-03 5.22565782e-01 7.68089056e-01
1.08748937e+00 1.04881012e+00 8.56800303e-02 -9.90243480e-02
1.43084180e+00 -1.17314625e+00 -8.21161211e-01 -3.25537145e-01
8.82641673e-01 -8.53965878e-01 1.40655220e+00 1.07212648e-01
-1.01225197e+00 -6.49674475e-01 -1.01475501e+00 -3.42892319e-01
-6.31802559e-01 3.26638788e-01 5.63497484e-01 4.28410351e-01
-8.48061860e-01 5.79920888e-01 -4.46194321e-01 -7.44569063e-01
3.08109790e-01 6.68789744e-02 -1.52927697e-01 -3.83561224e-01
-1.75975347e+00 1.12918258e+00 5.90660214e-01 3.84115547e-01
-6.84204698e-01 -5.28142214e-01 -9.16938245e-01 3.36623549e-01
2.14644790e-01 -7.08831847e-01 1.23071885e+00 -5.42102754e-01
-1.60749996e+00 8.20927680e-01 -1.50793836e-01 -5.58766246e-01
1.80779412e-01 -4.14486438e-01 -6.40807033e-01 7.76289999e-02
4.02209386e-02 4.47028726e-01 5.79139829e-01 -1.04655802e+00
-3.11975896e-01 -6.95885867e-02 3.15041214e-01 4.01760131e-01
-9.42285657e-01 2.00465634e-01 -5.53466737e-01 -7.57918298e-01
-3.21414143e-01 -4.85089689e-01 -1.99747786e-01 -4.14290100e-01
-2.26961464e-01 -6.51667058e-01 5.77472389e-01 -5.96363068e-01
1.69410956e+00 -2.16904235e+00 1.04197547e-01 -8.39809105e-02
-2.65910745e-01 7.72597253e-01 -6.59612834e-01 9.61949110e-01
2.93076076e-02 8.88952687e-02 -3.57657462e-01 -5.03107190e-01
1.38182610e-01 1.55024573e-01 -7.46266246e-01 2.83946283e-02
4.68028694e-01 1.07935297e+00 -1.02252793e+00 -4.95678902e-01
2.58520693e-01 2.32732221e-01 -2.90720701e-01 4.83763874e-01
-5.58052622e-02 -5.49060442e-02 -7.07540989e-01 4.04102296e-01
4.26143587e-01 -1.14855811e-01 -1.29649520e-01 -1.24093726e-01
1.06561057e-01 6.61112845e-01 -7.21941411e-01 2.07836843e+00
-6.68360651e-01 4.33430046e-01 -1.59468517e-01 -1.19621503e+00
1.18554890e+00 4.21602219e-01 1.06916480e-01 -8.98855209e-01
1.42587334e-01 3.09839845e-01 -2.81179491e-02 -7.87274778e-01
9.17126656e-01 -3.03713173e-01 -2.49812976e-01 3.45786363e-01
3.19643289e-01 -2.18288526e-01 3.80363613e-01 3.68061334e-01
1.07624733e+00 -7.74835721e-02 2.92066306e-01 -1.08306944e-01
8.09914172e-01 -7.37462938e-02 2.03196466e-01 7.50898957e-01
-1.01778619e-01 7.34043241e-01 1.25502512e-01 -4.02596071e-02
-7.97885537e-01 -9.67126489e-01 -3.75550270e-01 1.32559144e+00
1.04395881e-01 -5.41307271e-01 -4.84469026e-01 -7.15755165e-01
-1.91077903e-01 8.92774999e-01 -4.07705158e-01 -4.91029114e-01
-5.11188745e-01 -6.07859731e-01 6.18560195e-01 4.37808216e-01
4.22018528e-01 -1.41803217e+00 -1.30828410e-01 2.84352630e-01
-1.13957532e-01 -8.97308946e-01 -6.42439425e-01 1.47672758e-01
-8.32795262e-01 -7.00503111e-01 -9.80875611e-01 -8.57678235e-01
2.83375472e-01 5.86677492e-01 1.06484556e+00 3.17766368e-01
-8.42148811e-02 3.69648993e-01 -8.80261123e-01 -2.33848751e-01
-2.62243539e-01 4.91276145e-01 -4.39425185e-02 -9.26482156e-02
4.86413389e-01 -4.13319945e-01 -5.10266125e-01 -4.44126390e-02
-1.18642497e+00 -2.47217342e-01 7.05649316e-01 1.13030922e+00
1.26846790e-01 -2.48975769e-01 1.11590803e+00 -9.64635074e-01
1.21945345e+00 -7.53741682e-01 2.35025492e-02 5.99147618e-01
-5.30789793e-01 4.32593748e-03 7.85118461e-01 -3.74689519e-01
-1.26479518e+00 -5.92640817e-01 -5.50183594e-01 -3.57499391e-01
5.43647334e-02 8.83803129e-01 -1.43226996e-01 4.10949260e-01
6.84578001e-01 5.03955007e-01 -1.13651685e-01 -6.28280759e-01
4.18132901e-01 1.10941827e+00 2.71019340e-01 -6.76111579e-01
7.81961083e-01 -1.53061733e-01 -5.36426306e-01 -9.28369224e-01
-1.03741574e+00 -6.30430937e-01 -3.34036052e-01 1.43377125e-01
7.01185405e-01 -8.04922521e-01 -1.22391365e-01 1.41683117e-01
-1.49966276e+00 -1.01897828e-02 -2.09893361e-01 4.51308310e-01
-2.27224827e-01 6.63664997e-01 -8.03570926e-01 -7.82812357e-01
-8.25045884e-01 -9.20329213e-01 9.84253943e-01 2.81236768e-01
-1.60670578e-01 -1.21537852e+00 2.58029252e-01 4.55480069e-01
5.71968317e-01 -6.43302858e-01 1.01868129e+00 -1.03510904e+00
-2.13694051e-01 -2.63387740e-01 -1.53789476e-01 7.44667709e-01
-3.80440131e-02 -2.04942599e-01 -1.01547098e+00 -5.57841122e-01
-4.96668667e-02 -7.93000579e-01 1.01472795e+00 -4.80008610e-02
1.27692044e+00 -7.88284540e-02 -1.33622423e-01 3.53176475e-01
1.24779332e+00 1.01856273e-02 8.14087808e-01 4.25893486e-01
1.87580094e-01 7.16511071e-01 9.39443946e-01 1.20122962e-01
2.51995891e-01 5.90021133e-01 2.17375681e-01 6.87187687e-02
-1.02920622e-01 -3.56864393e-01 5.37144542e-01 1.69544554e+00
9.87475961e-02 -4.32594299e-01 -6.17834091e-01 5.55521011e-01
-1.81092882e+00 -1.07732868e+00 1.59602299e-01 1.97140586e+00
1.00616038e+00 8.28664452e-02 -3.76461357e-01 -1.05106167e-01
6.28771901e-01 6.23404324e-01 -3.06315064e-01 -6.23883426e-01
-1.38865039e-01 6.26549959e-01 -1.69881321e-02 2.36648709e-01
-8.97458673e-01 1.07102597e+00 5.70795584e+00 1.37848616e+00
-9.29938138e-01 2.54623264e-01 3.75025064e-01 8.06864575e-02
-5.30448496e-01 -2.17610095e-02 -9.79926348e-01 3.68263781e-01
1.04482973e+00 -5.43764770e-01 1.54137760e-01 6.73024058e-01
-2.19445955e-02 3.28876317e-01 -9.23858047e-01 8.73217404e-01
2.16647387e-01 -1.38273370e+00 2.43852153e-01 -3.24543297e-01
4.45727587e-01 1.22523293e-01 -3.23349908e-02 1.25083911e+00
1.59340445e-02 -9.48547959e-01 2.96453267e-01 5.09488285e-01
7.43765593e-01 -6.28652632e-01 9.59342599e-01 7.16839612e-01
-1.11327970e+00 -2.64495108e-02 -8.82044613e-01 -1.08659379e-02
2.14616194e-01 4.22275364e-01 -6.02092922e-01 1.03701806e+00
2.30991095e-01 8.20364892e-01 -6.22914493e-01 8.93975437e-01
-2.20301270e-01 6.88229859e-01 7.09432736e-02 -4.91817534e-01
3.27648908e-01 -2.61844277e-01 3.42042238e-01 1.59776080e+00
4.13203001e-01 1.89950336e-02 -7.71907717e-02 6.01876438e-01
-2.50853956e-01 4.63030219e-01 -7.54408240e-01 -1.52260184e-01
5.73622465e-01 1.53058398e+00 -1.00973919e-01 -4.73636687e-01
-5.42102337e-01 1.00794256e+00 7.29708552e-01 5.14904261e-01
-6.33396387e-01 -7.24606335e-01 4.11021352e-01 -1.29043251e-01
1.70241788e-01 -8.95220116e-02 -1.18454136e-01 -1.57153332e+00
5.82719855e-02 -8.87031972e-01 4.29917067e-01 -6.85198128e-01
-1.74822748e+00 7.67455280e-01 -1.13871861e-02 -1.27382827e+00
-2.40264505e-01 -6.64128602e-01 -9.92566586e-01 1.07691193e+00
-1.71953619e+00 -9.11012709e-01 7.09751472e-02 3.45414311e-01
9.92999852e-01 -4.18266535e-01 1.13807809e+00 5.13775229e-01
-8.15499961e-01 7.62472630e-01 3.68305832e-01 2.25794762e-01
7.50746906e-01 -9.91522193e-01 1.82723135e-01 7.65465438e-01
3.68053079e-01 1.03628397e+00 4.45041180e-01 -3.27346385e-01
-1.44327331e+00 -1.14986730e+00 1.03841794e+00 -2.44810775e-01
9.40659046e-01 -4.09657896e-01 -1.24120069e+00 4.60317314e-01
5.12937069e-01 -1.69371411e-01 7.46051610e-01 4.13668573e-01
-1.94392741e-01 -2.44015411e-01 -7.70854712e-01 5.92779756e-01
8.77159059e-01 -7.51982450e-01 -1.22515178e+00 4.14691746e-01
9.61790740e-01 -2.08659530e-01 -9.38588798e-01 4.58026707e-01
1.54471055e-01 -5.56763470e-01 1.03770471e+00 -8.76255572e-01
7.68419206e-01 9.82551724e-02 -1.96656972e-01 -1.42467105e+00
-2.16488987e-01 -3.91895026e-01 -1.89538017e-01 1.61974037e+00
5.05496740e-01 -6.19873762e-01 3.62848133e-01 2.34642178e-01
-4.55687195e-01 -1.14625347e+00 -9.92517114e-01 -7.89737999e-01
2.84366935e-01 -5.24496436e-01 4.65295702e-01 9.35982585e-01
5.01362048e-02 8.94898236e-01 -2.40354255e-01 -1.98653907e-01
4.77710478e-02 7.60811716e-02 6.67716742e-01 -8.15471709e-01
-3.74970973e-01 -4.64417785e-01 -4.03559171e-02 -1.61414266e+00
2.20119938e-01 -1.31767523e+00 6.35640174e-02 -1.64150798e+00
3.77458632e-01 -4.20652539e-01 -5.09433270e-01 2.22521782e-01
-5.40619910e-01 -1.90907001e-01 1.77499518e-01 1.87119126e-01
-7.16658533e-01 1.15116453e+00 1.43198287e+00 -2.48778552e-01
3.85438614e-02 -5.68657219e-02 -7.24154651e-01 3.87110382e-01
7.87220120e-01 -4.15253878e-01 -5.89868248e-01 -5.90086341e-01
-1.20492624e-02 6.87928945e-02 7.04905391e-02 -6.95633113e-01
2.33560622e-01 -4.76831235e-02 -6.25396743e-02 -6.06967986e-01
5.04329145e-01 -4.98090833e-01 -4.89260077e-01 3.48284543e-01
-5.65370619e-01 -1.81640554e-02 2.24198699e-02 4.14724618e-01
-5.40117621e-01 -9.50651824e-01 4.90445048e-01 -8.90343785e-02
-6.80567384e-01 3.85417521e-01 -1.44916832e-01 3.06479216e-01
6.83109522e-01 7.31242672e-02 -4.50812399e-01 -2.77295142e-01
-3.92101556e-01 3.59554201e-01 -7.64074549e-02 6.81409478e-01
8.19086075e-01 -1.35686886e+00 -8.53500068e-01 3.08377016e-02
3.81691188e-01 -6.71455860e-02 3.81704301e-01 5.00740767e-01
-1.44332290e-01 6.07138932e-01 1.92047238e-01 -3.54862064e-01
-9.69951630e-01 6.20733738e-01 -4.37176935e-02 -7.15173662e-01
-5.02474606e-01 6.82020724e-01 7.58705288e-02 -6.02189004e-01
1.51705250e-01 -1.49797380e-01 -4.74319309e-01 8.60197917e-02
5.85086048e-01 1.65422320e-01 8.77017304e-02 -4.66302335e-01
-1.90176442e-01 3.88362885e-01 -5.10010600e-01 -4.84608635e-02
1.54226887e+00 -1.24475859e-01 -1.44461870e-01 4.75444973e-01
1.57674754e+00 -6.17394671e-02 -6.27740979e-01 -5.16833961e-01
2.09682390e-01 -3.02800328e-01 -1.25266567e-01 -5.91246247e-01
-7.98146486e-01 1.33896160e+00 2.50683486e-01 1.87064454e-01
1.17232704e+00 -1.96609218e-02 1.18014383e+00 8.07226479e-01
3.37528944e-01 -1.07350373e+00 2.49556541e-01 9.48758423e-01
1.07233787e+00 -1.10921371e+00 -1.74984440e-01 -1.13379471e-01
-7.32646525e-01 1.19969893e+00 6.10789180e-01 -2.20088243e-01
3.37577909e-01 -1.43961042e-01 -1.26811475e-01 -1.14833608e-01
-9.99118626e-01 -3.05179179e-01 5.05435765e-01 3.64820451e-01
7.95627832e-01 -8.77096727e-02 -8.18644166e-01 1.01510620e+00
-1.48800105e-01 -9.21694841e-03 1.72892421e-01 8.27719986e-01
-5.42363644e-01 -1.41290939e+00 -1.63361803e-02 6.08372748e-01
-2.97885418e-01 -4.91461486e-01 -1.76165670e-01 6.70119643e-01
-1.63548544e-01 9.34449494e-01 -2.45068505e-01 -3.39687437e-01
6.34331703e-01 2.48044267e-01 2.73850828e-01 -1.04288888e+00
-8.37875247e-01 -2.73349255e-01 3.76784682e-01 -9.81669798e-02
-1.28883451e-01 -2.82764226e-01 -1.02324259e+00 -7.41037950e-02
-4.83796537e-01 4.34009910e-01 2.77733535e-01 9.32802260e-01
3.68379802e-01 5.65166771e-01 7.42659092e-01 -5.58514595e-01
-1.15649605e+00 -1.47935629e+00 -4.10576999e-01 5.73609471e-01
-1.06960544e-02 -4.67532754e-01 -3.23530883e-01 -3.26983064e-01]
|
[10.969178199768066, 8.468058586120605]
|
b2ffbe97-29c8-4586-b3be-e6de61cf111b
|
balancing-exploration-and-exploitation
|
2306.01683
| null |
https://arxiv.org/abs/2306.01683v1
|
https://arxiv.org/pdf/2306.01683v1.pdf
|
Balancing Exploration and Exploitation: Disentangled $β$-CVAE in De Novo Drug Design
|
Deep generative models have recently emerged as a promising de novo drug design method. In this respect, deep generative conditional variational autoencoder (CVAE) models are a powerful approach for generating novel molecules with desired drug-like properties. However, molecular graph-based models with disentanglement and multivariate explicit latent conditioning have not been fully elucidated. To address this, we proposed a molecular-graph $\beta$-CVAE model for de novo drug design. Here, we empirically tuned the value of disentanglement and assessed its ability to generate molecules with optimised univariate- or-multivariate properties. In particular, we optimised the octanol-water partition coefficient (ClogP), molar refractivity (CMR), quantitative estimate of drug-likeness (QED), and synthetic accessibility score (SAS). Results suggest that a lower $\beta$ value increases the uniqueness of generated molecules (exploration). Univariate optimisation results showed our model generated molecular property averages of ClogP = 41.07% $\pm$ 0.01% and CMR 66.76% $\pm$ 0.01% by the Ghose filter. Multivariate property optimisation results showed that our model generated an average of 30.07% $\pm$ 0.01% molecules for both desired properties. Furthermore, our model improved the QED and SAS (exploitation) of molecules generated. Together, these results suggest that the $\beta$-CVAE could balance exploration and exploitation through disentanglement and is a promising model for de novo drug design, thus providing a basis for future studies.
|
['Bingquan Shen', 'De Tao Irwin Chin', 'Guang Jun Nicholas Ang']
|
2023-06-02
| null | null | null | null |
['disentanglement']
|
['methodology']
|
[ 5.13309166e-02 1.51678681e-01 -1.99411258e-01 1.51847214e-01
-4.84974384e-01 -5.98980963e-01 6.07249916e-01 4.79645431e-01
-3.01952809e-01 1.26625752e+00 -5.75949587e-02 -4.39849466e-01
-2.22880259e-01 -1.01849473e+00 -8.51002693e-01 -1.20611477e+00
-3.65825921e-01 2.75498480e-01 -2.97770679e-01 -2.69537330e-01
2.32347190e-01 7.19147742e-01 -1.22458375e+00 -3.32270563e-02
1.39115155e+00 5.89304447e-01 2.72983521e-01 4.13528025e-01
3.71417403e-01 2.88902432e-01 -6.10683441e-01 -3.95419627e-01
-1.07161470e-01 -7.37911582e-01 -2.39960134e-01 -4.81202990e-01
1.14979565e-01 3.54754962e-02 -4.55112988e-03 9.98809457e-01
7.68259168e-01 4.38047826e-01 1.18343973e+00 -7.41658151e-01
-8.22086155e-01 4.29191649e-01 -3.18605095e-01 8.54733586e-02
2.60408700e-01 4.80664760e-01 9.64339912e-01 -8.98602605e-01
6.29663765e-01 8.10174763e-01 2.10983366e-01 6.01849794e-01
-1.68855560e+00 -8.89941812e-01 -1.44272059e-01 -7.32267275e-02
-1.55075932e+00 -1.77328408e-01 6.08305693e-01 -7.41543353e-01
1.32925248e+00 3.28961343e-01 8.79272342e-01 9.07371283e-01
7.90172994e-01 1.95975736e-01 8.99865627e-01 -7.66459703e-02
5.79523027e-01 2.78016984e-01 -5.36618173e-01 6.65676951e-01
5.54580808e-01 3.53859246e-01 -2.89305180e-01 -3.12584639e-01
8.04922462e-01 -1.75666407e-01 -2.78342783e-01 -4.21398789e-01
-8.03667128e-01 1.41745114e+00 4.85519826e-01 1.37089998e-01
-5.42618096e-01 -2.61706114e-03 -6.29482865e-02 -1.65968820e-01
1.77668005e-01 1.26078951e+00 -4.31377441e-01 -2.27382872e-03
-7.69675672e-01 4.58084434e-01 7.08887875e-01 5.21349132e-01
6.27169251e-01 4.97976959e-01 -7.64246797e-03 5.76311827e-01
4.43155497e-01 4.73114252e-01 1.02228202e-01 -5.68333983e-01
7.87879303e-02 4.95189816e-01 -2.57966993e-03 -8.77472699e-01
-3.65783781e-01 -7.54564047e-01 -9.51251805e-01 1.13952249e-01
1.04564898e-01 -2.29762182e-01 -8.85453284e-01 1.89299953e+00
3.16132605e-01 -1.27895057e-01 2.39363268e-01 6.23485386e-01
9.63178933e-01 8.91567111e-01 6.65807247e-01 -6.01621985e-01
1.18431282e+00 -4.78820562e-01 -6.07100010e-01 1.31046131e-01
5.16547561e-01 -6.55451834e-01 6.66189194e-01 4.41537350e-01
-1.17466712e+00 -4.75468546e-01 -1.17103553e+00 4.47795719e-01
-2.28428051e-01 -8.26838985e-03 8.26605976e-01 1.02320683e+00
-6.48797929e-01 9.89385784e-01 -7.98353314e-01 2.89845556e-01
5.38754463e-01 7.12196171e-01 -3.20896089e-01 1.52677640e-01
-1.34643757e+00 8.00461471e-01 6.08821213e-01 4.71812598e-02
-1.30066466e+00 -9.82652247e-01 -9.18455362e-01 1.80182010e-02
3.53087783e-02 -1.02370000e+00 5.41132569e-01 -3.80683988e-01
-1.75019562e+00 2.72282898e-01 3.32988724e-02 -3.42248887e-01
8.39081854e-02 1.18893661e-01 -4.30639923e-01 7.80678466e-02
-8.62230062e-02 9.26351070e-01 5.69432199e-01 -1.14468443e+00
-8.28061625e-02 -3.28059793e-01 -1.73626885e-01 1.90651461e-01
5.67486994e-02 -4.93491739e-01 3.96622419e-01 -5.35108566e-01
-2.17649072e-01 -9.03306246e-01 -4.37207192e-01 -5.35811424e-01
-3.23318750e-01 2.98356954e-02 8.99242237e-02 -5.81740320e-01
1.31478465e+00 -1.64684725e+00 5.23664057e-01 5.40687442e-01
5.89087784e-01 4.65335697e-01 -5.19858152e-02 7.97556281e-01
-3.97674829e-01 5.53197443e-01 -2.22837418e-01 4.82025862e-01
-3.40074450e-01 -2.85019308e-01 1.51190355e-01 4.48218793e-01
3.84688616e-01 9.54028845e-01 -9.53727782e-01 -7.39729637e-03
2.45868519e-01 9.45306301e-01 -1.06715012e+00 8.84664133e-02
-6.18477464e-01 6.43511236e-01 -4.95638222e-01 5.40603936e-01
7.27313638e-01 -1.58324838e-01 5.14752328e-01 -1.79685250e-01
-3.06462735e-01 -4.29992564e-02 -8.79210114e-01 1.32387233e+00
6.13284297e-02 3.32788259e-01 -5.62308550e-01 -5.02046824e-01
1.08549380e+00 1.73063338e-01 5.93285978e-01 -8.89896989e-01
1.62426904e-01 1.62222177e-01 4.84097809e-01 -2.34018430e-01
3.76074284e-01 -7.75779486e-01 1.88406184e-01 -8.71785432e-02
6.61313310e-02 -2.58902848e-01 2.06446260e-01 6.43882230e-02
6.93096519e-01 1.24199532e-01 2.82854944e-01 -4.22121108e-01
4.44659412e-01 -1.51241064e-01 2.97995389e-01 4.27022368e-01
4.08383086e-03 3.29052895e-01 6.30429864e-01 -1.86833709e-01
-1.12627375e+00 -1.11770868e+00 -3.39685947e-01 5.27779281e-01
-4.10027467e-02 -6.17137671e-01 -7.59160340e-01 -1.06565163e-01
-1.77674606e-01 8.74409497e-01 -4.45965201e-01 -5.17577946e-01
-2.89033949e-01 -1.22995985e+00 3.29680234e-01 2.96185017e-01
6.32864907e-02 -9.04395521e-01 -9.64243561e-02 4.66801375e-01
2.50894666e-01 -4.96239841e-01 -2.16606215e-01 2.90909380e-01
-7.36453354e-01 -9.20982122e-01 -8.92109573e-01 -4.16832149e-01
7.04640329e-01 -3.17564040e-01 7.79905796e-01 -2.92362154e-01
-2.51095206e-01 -2.93329358e-01 -2.12303251e-01 -4.85555530e-01
-6.06656969e-01 -6.77501485e-02 1.32149369e-01 -3.89189541e-01
1.08500987e-01 -7.14238763e-01 -1.17514265e+00 2.04891503e-01
-9.63734925e-01 -1.94152132e-01 4.93148953e-01 7.99529672e-01
9.09807086e-01 3.26367319e-02 6.93870425e-01 -6.60947442e-01
8.95246625e-01 -5.30436575e-01 -6.22688234e-01 -1.50808275e-01
-8.85080695e-01 4.27441418e-01 6.11098945e-01 -6.10750794e-01
-7.98764586e-01 -2.11655557e-01 -5.10880649e-01 -9.63796675e-02
-4.83763404e-02 8.32236826e-01 -4.82313484e-01 1.89273104e-01
8.88960361e-01 1.70035943e-01 4.23418218e-03 -2.60187257e-02
3.07173878e-01 1.20708838e-01 -2.67326593e-01 -6.00283146e-01
2.65620887e-01 9.56197008e-02 2.03522384e-01 -9.89437163e-01
9.53478515e-02 6.43187985e-02 -1.52103737e-01 9.70221683e-02
1.07309020e+00 -1.06591475e+00 -1.28658986e+00 4.04016562e-02
-7.93261588e-01 -4.18485314e-01 -7.79987138e-04 8.41109216e-01
-4.47637349e-01 3.40026796e-01 -2.35105783e-01 -7.04837918e-01
-5.55407166e-01 -1.47461832e+00 6.00280285e-01 3.21695536e-01
-6.44391298e-01 -1.03682804e+00 3.61920863e-01 2.88210779e-01
3.20651531e-01 8.63813341e-01 1.23112643e+00 -4.59577173e-01
-6.74012899e-01 -9.07108411e-02 5.33954129e-02 1.03646122e-01
5.46198376e-02 7.19363764e-02 -6.68202162e-01 -4.82503891e-01
-3.83450627e-01 1.60722569e-01 7.80322313e-01 8.23817253e-01
7.26536572e-01 -4.31434035e-01 -2.73806721e-01 5.53707540e-01
1.38924301e+00 1.01731229e+00 9.48543429e-01 1.06229961e-01
6.37664139e-01 8.86181593e-02 2.12963268e-01 7.70160675e-01
-1.27956152e-01 5.85246623e-01 4.45960850e-01 -1.13349490e-01
1.05949849e-01 -5.21978438e-01 4.45311457e-01 4.42870021e-01
-3.48416418e-01 -6.02612078e-01 -7.19173789e-01 2.16953963e-01
-1.25389206e+00 -9.37026680e-01 -2.38187551e-01 2.17417908e+00
9.72533226e-01 1.44812226e-01 3.15321058e-01 -2.43933648e-01
4.43514109e-01 -1.77130252e-02 -6.84096515e-01 -7.42690384e-01
-1.82717875e-01 8.16303849e-01 3.26361895e-01 5.85019052e-01
-5.35215497e-01 8.86498749e-01 5.58467865e+00 9.56277549e-01
-1.12382329e+00 -3.49796981e-01 6.48413002e-01 -7.81719610e-02
-8.17650735e-01 2.31271923e-01 -8.58595133e-01 6.50854170e-01
1.07385755e+00 -1.15881212e-01 1.13036796e-01 5.60224593e-01
2.65797228e-01 -2.71776855e-01 -8.45291257e-01 8.29952657e-01
-1.25353813e-01 -1.66605139e+00 2.89473116e-01 5.57431757e-01
8.60268176e-01 -3.57760787e-01 3.07948530e-01 5.88630363e-02
1.65309906e-01 -1.43125546e+00 4.02300030e-01 5.92649162e-01
9.77913499e-01 -1.07221580e+00 5.36007822e-01 1.49322003e-01
-8.74815047e-01 2.49460161e-01 -2.01472938e-01 1.92721412e-01
1.98256802e-02 5.81273139e-01 -1.04037559e+00 4.25933748e-01
1.13711655e-01 4.02492076e-01 -3.09148371e-01 8.91868711e-01
-2.51910925e-01 3.26152444e-01 -1.37950748e-01 -4.64355648e-01
1.57324389e-01 -6.84283316e-01 4.57670957e-01 8.98997843e-01
3.11275035e-01 1.91803917e-01 -2.10946023e-01 1.23745012e+00
5.95097914e-02 1.80089161e-01 -3.30432415e-01 -6.36059999e-01
3.11741114e-01 5.56018651e-01 -6.30062163e-01 -7.02352226e-02
2.69666255e-01 7.07423985e-01 -1.06663123e-01 5.76872945e-01
-1.01944959e+00 -4.58939582e-01 8.60366166e-01 2.50570744e-01
3.60403925e-01 -1.89066976e-01 -1.07463058e-02 -1.03653860e+00
-4.90214854e-01 -9.97725129e-01 1.59061402e-01 -5.89151800e-01
-7.73564458e-01 5.85116565e-01 7.57600889e-02 -7.93372810e-01
1.92369074e-02 -4.58098620e-01 -2.16702685e-01 1.09625828e+00
-1.18742681e+00 -7.96092987e-01 2.59703338e-01 1.12812907e-01
2.04187825e-01 -2.79509515e-01 1.04245126e+00 1.27560854e-01
-7.51496434e-01 4.60017949e-01 4.38016355e-01 -5.25570095e-01
4.83922511e-01 -1.08587599e+00 -4.02192362e-02 4.88705039e-01
-3.28254066e-02 1.00408649e+00 1.10029399e+00 -9.25496280e-01
-1.23406076e+00 -7.84103811e-01 5.46509027e-01 -4.07411635e-01
3.35223585e-01 -2.09264562e-01 -6.55370414e-01 3.86777222e-02
7.41411895e-02 -6.83271170e-01 1.20612907e+00 -9.41830948e-02
-8.52962397e-03 1.98522165e-01 -1.11481762e+00 9.07117426e-01
6.32422388e-01 -3.66905212e-01 6.90404475e-02 1.92784801e-01
6.22306883e-01 -2.88060606e-01 -1.55002594e+00 5.27781010e-01
6.91405892e-01 -7.96900392e-01 1.11164796e+00 -7.15506256e-01
3.86883974e-01 -2.99404919e-01 -1.54522983e-02 -1.24580657e+00
-5.67619860e-01 -5.57976902e-01 -2.56255358e-01 6.27343833e-01
8.16712558e-01 -6.92182899e-01 8.55033457e-01 7.10228503e-01
-1.66757524e-01 -1.02986777e+00 -7.36945689e-01 -5.59210002e-01
4.26874638e-01 -1.98468000e-01 6.22949719e-01 6.76093161e-01
3.92974541e-02 4.79096591e-01 -2.79109806e-01 -3.46756801e-02
2.28529736e-01 -2.21960008e-01 4.39657956e-01 -1.01038027e+00
-3.51175696e-01 -6.08694017e-01 -2.98177272e-01 -6.81054413e-01
-1.11124873e-01 -9.79387879e-01 -6.42790198e-01 -1.43480849e+00
1.89961210e-01 -2.79348403e-01 -1.73363641e-01 1.24609634e-01
-6.86293021e-02 -1.89803436e-01 -5.96272051e-02 -5.61067909e-02
4.56565619e-02 8.31978142e-01 1.36493993e+00 -2.36316606e-01
-7.54494667e-01 -8.76907483e-02 -9.29662228e-01 5.33095412e-02
1.05575776e+00 -5.18683136e-01 -5.93843341e-01 2.33205035e-01
8.26346815e-01 3.05754721e-01 1.03781901e-01 -6.66966796e-01
-2.63426095e-01 -4.31465864e-01 7.29750991e-01 -2.84934938e-01
4.36607212e-01 -3.53632838e-01 8.35566223e-01 7.13088870e-01
-2.95662098e-02 -2.73329973e-01 4.55124319e-01 6.55743241e-01
7.68927950e-03 -2.58701853e-02 7.23519802e-01 -6.31724745e-02
-2.25253612e-01 4.88230735e-01 -6.46542132e-01 -3.94751698e-01
1.05808604e+00 -6.79778934e-01 2.38617733e-02 -1.99463621e-01
-9.28181648e-01 -1.61866844e-01 4.40653145e-01 1.57791704e-01
8.22850525e-01 -1.05088246e+00 -4.82717842e-01 3.68728548e-01
1.17055111e-01 -1.67643800e-01 6.50035620e-01 7.44342387e-01
-7.34570384e-01 5.36057651e-01 -1.89125508e-01 -4.21017081e-01
-1.05974317e+00 5.87370217e-01 4.41674352e-01 -1.54496178e-01
2.09574178e-01 9.29261863e-01 3.45737934e-01 -1.05490424e-01
-3.28237504e-01 -7.20629171e-02 -2.92412132e-01 2.51991123e-01
1.46470293e-01 3.95458192e-01 1.47048816e-01 -4.66165125e-01
-4.17656004e-01 3.81863654e-01 -1.62500456e-01 1.00793287e-01
1.54700017e+00 3.76791686e-01 7.54349455e-02 -2.53174812e-01
1.11684287e+00 4.24966998e-02 -1.31736147e+00 5.75181603e-01
-4.55484301e-01 -1.29365861e-01 1.54779196e-01 -9.76054549e-01
-6.71813309e-01 6.44615769e-01 6.33123875e-01 -9.89880562e-02
7.03454137e-01 8.56380817e-03 5.13125718e-01 -5.28163910e-02
-4.53031994e-02 -9.38397348e-01 1.77587241e-01 2.87607670e-01
9.17242765e-01 -9.22826648e-01 3.14277261e-01 -2.39926085e-01
-6.34191334e-01 9.02512550e-01 4.53202426e-01 1.74579978e-01
5.26508927e-01 -2.84003228e-01 -5.07043302e-01 -5.89242280e-01
-5.13289750e-01 -6.52268380e-02 4.35398072e-01 6.76695049e-01
8.00045371e-01 3.68162036e-01 -6.18750393e-01 4.48727578e-01
-2.38758251e-01 -2.92491436e-01 2.15404138e-01 7.19337702e-01
-2.38853127e-01 -1.35772502e+00 -6.36493564e-02 1.60381228e-01
-2.33661905e-01 -4.71223235e-01 -4.19601709e-01 7.66815007e-01
3.09216022e-01 9.45384204e-01 -2.83408403e-01 -2.76280254e-01
2.99561352e-01 -4.55973074e-02 6.50147736e-01 -5.15100598e-01
-5.27111888e-01 4.75754708e-01 5.17554544e-02 -9.38300882e-03
-2.02101931e-01 -2.37798885e-01 -1.18884611e+00 -4.41602468e-01
-8.25631559e-01 6.56957567e-01 7.05738485e-01 4.98851806e-01
7.81657636e-01 6.02708101e-01 5.66711128e-01 -5.03697157e-01
7.06100911e-02 -7.06071675e-01 -5.91638982e-01 -1.64615661e-02
9.96109992e-02 -5.45464337e-01 -1.78175271e-01 -4.37364765e-02]
|
[5.012134075164795, 5.684840202331543]
|
37c8e5d4-b713-409c-bb6d-b357627e7a4b
|
pd-morl-preference-driven-multi-objective
|
2208.07914
| null |
https://arxiv.org/abs/2208.07914v3
|
https://arxiv.org/pdf/2208.07914v3.pdf
|
PD-MORL: Preference-Driven Multi-Objective Reinforcement Learning Algorithm
|
Multi-objective reinforcement learning (MORL) approaches have emerged to tackle many real-world problems with multiple conflicting objectives by maximizing a joint objective function weighted by a preference vector. These approaches find fixed customized policies corresponding to preference vectors specified during training. However, the design constraints and objectives typically change dynamically in real-life scenarios. Furthermore, storing a policy for each potential preference is not scalable. Hence, obtaining a set of Pareto front solutions for the entire preference space in a given domain with a single training is critical. To this end, we propose a novel MORL algorithm that trains a single universal network to cover the entire preference space scalable to continuous robotic tasks. The proposed approach, Preference-Driven MORL (PD-MORL), utilizes the preferences as guidance to update the network parameters. It also employs a novel parallelization approach to increase sample efficiency. We show that PD-MORL achieves up to 25% larger hypervolume for challenging continuous control tasks and uses an order of magnitude fewer trainable parameters compared to prior approaches.
|
['Umit Y. Ogras', 'Suat Gumussoy', 'Toygun Basaklar']
|
2022-08-16
| null | null | null | null |
['multi-objective-reinforcement-learning']
|
['methodology']
|
[ 3.32557023e-01 -2.01323479e-01 -6.53261304e-01 -1.26431718e-01
-9.94523346e-01 -6.82947576e-01 5.51901385e-03 2.01413170e-01
-6.36056900e-01 1.26019144e+00 -7.79902115e-02 -1.47328064e-01
-9.06100333e-01 -4.64695096e-01 -7.49820471e-01 -8.65392804e-01
-3.31183672e-02 7.74540365e-01 6.40942752e-02 -1.63204014e-01
5.27840018e-01 3.59451294e-01 -1.68776083e+00 -2.92318258e-02
1.22335970e+00 1.06644213e+00 7.15976119e-01 2.79132307e-01
2.55344987e-01 5.31881042e-02 -3.54094744e-01 2.57780820e-01
3.59217316e-01 1.09710038e-01 -7.91142046e-01 -3.15137133e-02
1.51943609e-01 -1.04712613e-03 1.45229444e-01 1.05307281e+00
6.61403358e-01 7.47580469e-01 4.91975635e-01 -1.43409646e+00
-2.70714462e-01 7.83586144e-01 -5.30168891e-01 -2.91788578e-01
-6.09513223e-02 3.86876434e-01 1.07320464e+00 -3.91536295e-01
3.78714323e-01 1.22360122e+00 2.24725634e-01 7.55472958e-01
-1.18936598e+00 -3.34382772e-01 5.59883296e-01 2.24239036e-01
-1.06890738e+00 -1.09348660e-02 7.47643769e-01 -4.13369061e-03
9.46124792e-01 5.43122850e-02 7.29646683e-01 1.00077450e+00
2.99792349e-01 8.89506936e-01 9.50729311e-01 -1.69578940e-01
6.98236644e-01 -1.13496333e-01 -3.81151259e-01 5.74618280e-01
4.41049725e-01 7.04426924e-03 -4.80846822e-01 -2.62121648e-01
5.55540562e-01 -1.25386968e-01 -2.29882568e-01 -1.00756490e+00
-1.09603608e+00 1.03768599e+00 8.69844034e-02 -1.27931476e-01
-4.46333200e-01 3.22566152e-01 3.86449873e-01 2.99690664e-01
-1.02752574e-01 1.27319467e+00 -9.49537516e-01 -2.13131621e-01
-5.99806607e-01 5.28513312e-01 5.38598180e-01 9.40093875e-01
7.23103940e-01 2.49013215e-01 -2.50412613e-01 1.01164401e+00
3.30293328e-01 3.57465655e-01 4.48816359e-01 -1.49882197e+00
7.01106012e-01 6.28748178e-01 5.64702511e-01 -7.11010516e-01
-5.95863760e-01 -6.57917202e-01 -5.58864534e-01 3.78197879e-01
2.08155036e-01 -5.23943007e-01 -7.25137949e-01 1.97266328e+00
5.52509725e-01 -4.04771090e-01 2.54366755e-01 9.95420039e-01
7.41773620e-02 6.49035752e-01 -1.87683463e-01 -3.93052757e-01
8.76395047e-01 -1.07790613e+00 -3.44017714e-01 -6.37585461e-01
3.21679771e-01 -3.08369190e-01 1.38406372e+00 5.49521685e-01
-1.11890924e+00 -1.25358313e-01 -1.29452991e+00 7.28097141e-01
-1.19269103e-01 7.42875412e-02 5.02564549e-01 4.44724590e-01
-8.48165751e-01 8.12643111e-01 -5.04783154e-01 -6.10408336e-02
3.80950928e-01 8.68334293e-01 1.44338727e-01 -1.72054559e-01
-1.05476105e+00 9.81421947e-01 9.24807310e-01 -5.02132811e-02
-1.04359198e+00 -7.25050688e-01 -7.40637660e-01 1.39913067e-01
9.65827107e-01 -4.28725034e-01 1.37904310e+00 -7.83112824e-01
-1.94074595e+00 -7.68913841e-03 3.74244809e-01 -1.97999850e-01
3.47502410e-01 -9.82122868e-02 -2.19105519e-02 -2.33906284e-02
-2.53271282e-01 8.16072226e-01 9.29501176e-01 -1.40724623e+00
-8.66962075e-01 -8.38702247e-02 2.76954085e-01 6.23137534e-01
-5.66501617e-01 -4.21949297e-01 -2.17121392e-01 -3.74958456e-01
-2.78327733e-01 -9.64193821e-01 -7.42285609e-01 -3.71942192e-01
-9.15608555e-02 -3.93485397e-01 7.08260834e-01 -4.52142879e-02
1.26531136e+00 -1.67068088e+00 5.61004698e-01 3.68996859e-01
-8.36585164e-02 2.00059637e-01 -6.30065203e-01 2.90698379e-01
4.29019213e-01 1.76007990e-02 -4.24999475e-01 1.12865986e-02
3.69163901e-01 5.99564075e-01 -4.31410549e-03 3.01838875e-01
1.84251234e-01 7.00116396e-01 -1.20220208e+00 -4.94741201e-01
3.71035654e-03 1.02443948e-01 -9.46896434e-01 4.71670646e-03
-6.89715505e-01 2.15858728e-01 -8.26802671e-01 6.49283588e-01
3.46957922e-01 -2.15699717e-01 7.19414949e-01 3.60027142e-02
-1.32779807e-01 -1.19109996e-01 -1.38284516e+00 1.80325234e+00
-5.76041758e-01 2.31650263e-01 6.21237196e-02 -1.21784592e+00
1.01307476e+00 1.58275217e-01 8.22304010e-01 -5.55655360e-01
3.86973828e-01 4.34720784e-01 -1.02797978e-01 -3.64897430e-01
6.17445648e-01 1.90994516e-01 -2.49632061e-01 4.27429706e-01
9.82632786e-02 -2.02751473e-01 6.48290753e-01 -5.71218610e-01
9.51418340e-01 4.48368698e-01 3.86386305e-01 -5.06371915e-01
5.09031951e-01 2.26074159e-01 1.14205980e+00 8.35453033e-01
-2.51530290e-01 1.50020435e-01 4.33195293e-01 -4.69390750e-01
-8.85604084e-01 -8.08025897e-01 2.67285615e-01 1.17719150e+00
3.16120714e-01 1.62288379e-02 -4.53705817e-01 -7.08276212e-01
1.03739323e-02 9.07615125e-01 -2.49946415e-01 -2.41672948e-01
-8.73504937e-01 -7.70531952e-01 1.94777399e-02 3.15250218e-01
3.80756736e-01 -1.27200949e+00 -1.33949590e+00 3.69030893e-01
-6.95909485e-02 -6.77392304e-01 -7.44860828e-01 4.77470249e-01
-1.00243402e+00 -1.04072392e+00 -6.65499151e-01 -1.00163412e+00
7.69085467e-01 -1.34170264e-01 9.87236202e-01 -3.73650998e-01
-1.98040172e-01 3.18153828e-01 -1.89980119e-01 -1.74919471e-01
-9.44565609e-02 3.14609408e-01 3.75114769e-01 -2.97479182e-01
-6.90010414e-02 -5.68586707e-01 -6.12808585e-01 3.49025756e-01
-7.44382441e-01 -2.04239696e-01 7.37611830e-01 9.68865037e-01
9.10495162e-01 1.71697363e-01 8.54171216e-01 -1.76700175e-01
1.05923796e+00 -3.97628427e-01 -1.01888442e+00 6.65412188e-01
-8.72693419e-01 6.97723269e-01 8.66996646e-01 -9.51216459e-01
-9.50143456e-01 2.66059607e-01 5.13895690e-01 -5.83241522e-01
1.80407315e-01 6.64992630e-01 -8.45317468e-02 -4.99556288e-02
4.62980360e-01 3.66779566e-02 1.44034833e-01 -1.70493171e-01
2.71571547e-01 2.93394864e-01 2.88471192e-01 -1.30745232e+00
4.53604519e-01 -2.59039223e-01 1.72542334e-01 -3.49982530e-01
-8.94874990e-01 -2.53283232e-01 -2.28448689e-01 -2.89528489e-01
4.85470206e-01 -3.65345925e-01 -1.10171294e+00 9.34086815e-02
-9.15040970e-01 -6.41054571e-01 -4.70750690e-01 5.71779609e-01
-9.68458891e-01 -4.92722578e-02 6.13853969e-02 -9.53848898e-01
-4.04211015e-01 -1.41929722e+00 5.16148806e-01 5.35360992e-01
-1.57902434e-01 -7.93488145e-01 2.48348072e-01 -3.16033736e-02
4.59669232e-01 3.00687909e-01 1.13107562e+00 -2.98404008e-01
-3.12367111e-01 2.24131003e-01 1.28792495e-01 3.99033390e-02
1.31605387e-01 -1.45315751e-01 -2.21496880e-01 -8.08727205e-01
-1.83448732e-01 -7.95365036e-01 4.86214370e-01 5.69386303e-01
1.61010766e+00 -6.38372540e-01 -1.50012210e-01 4.54869568e-01
1.63248277e+00 6.76882744e-01 1.35160536e-01 6.59144104e-01
2.30333313e-01 5.45947373e-01 1.11256063e+00 8.33016634e-01
5.44538721e-02 4.30343419e-01 7.64136016e-01 4.71872568e-01
5.84685743e-01 -5.49733527e-02 4.08659220e-01 4.86943573e-01
-2.85654366e-02 -6.06936157e-01 -7.16988146e-01 6.79166377e-01
-2.37688017e+00 -7.50357151e-01 6.84525609e-01 2.25733638e+00
8.41599464e-01 3.45658474e-02 1.33968920e-01 -8.23481828e-02
8.11755657e-01 2.05367468e-02 -1.41740322e+00 -6.55514657e-01
1.15435280e-01 7.91920498e-02 8.37114275e-01 2.96197385e-01
-9.70947981e-01 5.97683489e-01 6.25166941e+00 9.40461516e-01
-1.13234639e+00 -2.55132049e-01 4.29912359e-01 -5.38741052e-01
-3.31001222e-01 -3.24100107e-01 -6.38866782e-01 1.31824851e-01
6.30784452e-01 -3.15109938e-01 9.36242521e-01 9.56681371e-01
2.85464734e-01 -2.78206095e-02 -8.91604781e-01 1.07402825e+00
-2.87997991e-01 -1.22585905e+00 -6.66625276e-02 -6.90760314e-02
1.14376974e+00 -1.61748845e-02 3.10696214e-01 3.18778157e-01
6.46656156e-01 -9.86706376e-01 6.57596350e-01 2.23717138e-01
8.01509738e-01 -1.28089106e+00 4.46555018e-01 3.07564437e-01
-1.11696136e+00 -7.87250578e-01 -5.58314800e-01 2.29450718e-01
2.84872279e-02 6.21541664e-02 -6.63827538e-01 5.18114567e-01
6.27991319e-01 4.48921174e-01 -1.73330326e-02 1.17212760e+00
-1.51145577e-01 2.51520246e-01 -4.12611902e-01 -5.36417246e-01
7.64859438e-01 -1.86295941e-01 6.88652515e-01 6.24787152e-01
6.27375841e-01 -2.95663893e-01 4.60337639e-01 7.57454336e-01
1.47327900e-01 2.59471498e-03 -3.41219574e-01 -3.26594085e-01
6.58017814e-01 1.29108191e+00 -5.95698535e-01 3.56403321e-01
-1.02160417e-03 5.48355281e-01 4.93106812e-01 3.68777186e-01
-1.00891030e+00 -5.13642430e-01 8.01415682e-01 -5.90650558e-01
4.45405662e-01 -2.70418912e-01 -1.82056233e-01 -5.99651933e-01
-9.38415900e-03 -9.82810438e-01 5.77460587e-01 -2.93243945e-01
-1.25782239e+00 3.55050117e-01 2.15910077e-01 -1.38081658e+00
-2.51719356e-01 -8.25137913e-01 -4.90961313e-01 5.22873700e-01
-1.47755933e+00 -4.94513065e-01 -1.09044882e-02 3.98310393e-01
8.89900148e-01 -6.05366409e-01 5.83156705e-01 -1.01380743e-01
-7.69791305e-01 5.11151493e-01 4.56444383e-01 -8.57607007e-01
5.73561251e-01 -1.21509957e+00 -3.69891882e-01 3.01500410e-01
-6.72868192e-01 3.29563111e-01 7.56633699e-01 -4.17914957e-01
-1.76008749e+00 -1.15269959e+00 2.94262648e-01 2.22675756e-01
5.46835005e-01 2.04763323e-01 -4.45929974e-01 8.93824995e-02
2.45908856e-01 -2.82134473e-01 3.76200140e-01 -8.15625712e-02
2.09591448e-01 -2.64690608e-01 -1.37089109e+00 8.95133734e-01
1.02472854e+00 3.15600187e-01 -3.44754308e-01 1.56742811e-01
7.69147635e-01 -4.62996185e-01 -8.86802733e-01 5.82304597e-01
5.29906273e-01 -2.86788613e-01 8.97825122e-01 -7.66391635e-01
5.32255054e-01 -4.25615579e-01 -2.61962563e-01 -1.84526825e+00
-3.22497815e-01 -9.57360685e-01 -3.87656182e-01 7.05302358e-01
6.73943460e-01 -7.19902635e-01 7.42501378e-01 5.17286599e-01
-2.49877945e-01 -1.37180912e+00 -9.56388950e-01 -9.52671468e-01
4.06117402e-02 -5.60448365e-03 8.43893707e-01 6.59907758e-01
9.84422266e-02 1.29837573e-01 -5.20553946e-01 1.43356651e-01
9.12931323e-01 3.12467515e-01 1.69151083e-01 -1.10486901e+00
-3.88588130e-01 -5.77494025e-01 4.68884200e-01 -8.96967471e-01
1.57350019e-01 -6.97482586e-01 5.75260341e-01 -1.65219271e+00
-4.34353463e-02 -8.04033339e-01 -6.95207834e-01 4.91228670e-01
3.22328024e-02 -4.63262409e-01 1.52944341e-01 -1.38764188e-01
-9.27598774e-01 9.06209290e-01 1.51458740e+00 -4.33759272e-01
-6.20512545e-01 -3.33574861e-02 -7.48072863e-01 6.09150589e-01
1.49842787e+00 -4.13703799e-01 -9.48456347e-01 -5.01375496e-01
4.46658343e-01 2.23033011e-01 -1.88541472e-01 -1.02912176e+00
2.05989569e-01 -1.15310061e+00 1.75461411e-01 -5.54967642e-01
2.60799974e-01 -7.71363080e-01 -8.77297670e-02 5.69102883e-01
-5.28002203e-01 2.81697184e-01 3.33014458e-01 6.78806424e-01
5.95049076e-02 -5.25710166e-01 7.87888765e-01 -2.04812735e-01
-7.61649489e-01 4.40612286e-01 -4.54794765e-01 1.96348533e-01
1.11895084e+00 -1.81078300e-01 -3.23992640e-01 2.71131136e-02
-3.29041392e-01 1.13412464e+00 1.68658748e-01 5.28372407e-01
7.74412632e-01 -1.27519727e+00 -3.62427860e-01 -2.68679798e-01
-1.29804671e-01 3.02415788e-01 1.55715764e-01 4.78692085e-01
-1.13819048e-01 5.65366089e-01 -6.07630074e-01 -2.82886595e-01
-1.08934629e+00 5.15806556e-01 3.69856447e-01 -6.34357214e-01
-1.52950540e-01 6.24723732e-01 -2.67999887e-01 -7.82616258e-01
4.20098543e-01 -1.71875224e-01 -3.83874655e-01 1.05012013e-02
1.32835269e-01 7.51675129e-01 -2.69871503e-01 1.27188683e-01
-2.84836084e-01 5.69707096e-01 1.44502014e-01 -1.63373798e-01
1.69486535e+00 3.65879312e-02 2.06269801e-01 2.10711539e-01
8.76877248e-01 -5.42921245e-01 -1.84773886e+00 7.84379989e-03
8.57108980e-02 -2.69738972e-01 1.13387913e-01 -9.22687352e-01
-1.12220371e+00 2.65151113e-01 4.79378045e-01 -2.92230576e-01
1.20826733e+00 -4.15020376e-01 5.39679945e-01 1.01638651e+00
5.82153916e-01 -1.88960600e+00 3.05956781e-01 7.23182797e-01
1.01791000e+00 -1.00612092e+00 -8.23945031e-02 1.42531320e-01
-8.23340535e-01 1.38660812e+00 1.08868790e+00 -2.27415785e-02
2.31440991e-01 -4.79804166e-02 -3.81720722e-01 5.63536733e-02
-9.84126747e-01 -4.25053686e-02 2.34209925e-01 6.22563720e-01
-7.33003914e-02 6.00527152e-02 -4.79863077e-01 3.42112988e-01
1.20568819e-01 -2.22255930e-01 2.13742271e-01 1.27234471e+00
-7.49905407e-01 -1.39497685e+00 -2.92771310e-01 6.50808096e-01
-1.72732845e-02 3.43120605e-01 1.91135257e-02 5.42708576e-01
-6.07904196e-02 9.67900813e-01 -7.03116804e-02 -2.66489267e-01
2.08679929e-01 -4.18927595e-02 7.41236567e-01 -2.86212474e-01
-4.30649757e-01 -1.83260351e-01 3.30292583e-02 -6.35974646e-01
-2.96651840e-01 -6.31250858e-01 -1.39249933e+00 5.59610687e-02
4.30288240e-02 2.44029999e-01 4.69377458e-01 9.18184102e-01
4.08598453e-01 7.62424290e-01 6.89151585e-01 -9.64803159e-01
-1.10271263e+00 -4.63314414e-01 -2.89425105e-01 -2.30780974e-01
1.98800772e-01 -1.09715748e+00 -6.17327541e-02 -5.73849916e-01]
|
[4.253900527954102, 2.359938144683838]
|
621e34d8-c232-4f36-9b94-9285f2fbdd93
|
segment-anything-model-for-medical-image
|
2304.10517
| null |
https://arxiv.org/abs/2304.10517v3
|
https://arxiv.org/pdf/2304.10517v3.pdf
|
Segment Anything Model for Medical Image Analysis: an Experimental Study
|
Training segmentation models for medical images continues to be challenging due to the limited availability of data annotations. Segment Anything Model (SAM) is a foundation model that is intended to segment user-defined objects of interest in an interactive manner. While the performance on natural images is impressive, medical image domains pose their own set of challenges. Here, we perform an extensive evaluation of SAM's ability to segment medical images on a collection of 19 medical imaging datasets from various modalities and anatomies. We report the following findings: (1) SAM's performance based on single prompts highly varies depending on the dataset and the task, from IoU=0.1135 for spine MRI to IoU=0.8650 for hip X-ray. (2) Segmentation performance appears to be better for well-circumscribed objects with prompts with less ambiguity and poorer in various other scenarios such as the segmentation of brain tumors. (3) SAM performs notably better with box prompts than with point prompts. (4) SAM outperforms similar methods RITM, SimpleClick, and FocalClick in almost all single-point prompt settings. (5) When multiple-point prompts are provided iteratively, SAM's performance generally improves only slightly while other methods' performance improves to the level that surpasses SAM's point-based performance. We also provide several illustrations for SAM's performance on all tested datasets, iterative segmentation, and SAM's behavior given prompt ambiguity. We conclude that SAM shows impressive zero-shot segmentation performance for certain medical imaging datasets, but moderate to poor performance for others. SAM has the potential to make a significant impact in automated medical image segmentation in medical imaging, but appropriate care needs to be applied when using it.
|
['Yixin Zhang', 'Nicholas Konz', 'Jichen Yang', 'Hanxue Gu', 'Haoyu Dong', 'Maciej A. Mazurowski']
|
2023-04-20
| null | null | null | null |
['zero-shot-segmentation', 'interactive-segmentation']
|
['computer-vision', 'computer-vision']
|
[ 3.57246727e-01 1.92577973e-01 -3.46619099e-01 -4.92416561e-01
-1.30845022e+00 -6.77771270e-01 3.70697290e-01 3.99286449e-01
-5.94866037e-01 4.38060552e-01 -2.56556012e-02 -7.14234710e-01
-1.84223726e-01 -9.44819227e-02 -3.03291559e-01 -6.41106129e-01
-1.17130809e-01 8.29577386e-01 6.61500812e-01 -1.32629514e-01
1.91349298e-01 4.76421118e-01 -9.22382832e-01 2.86934257e-01
8.48960638e-01 6.33302271e-01 3.98975670e-01 9.25598681e-01
-1.56084567e-01 5.34674823e-01 -6.28206313e-01 5.52161559e-02
2.10178450e-01 -2.79654652e-01 -1.22289336e+00 2.96414673e-01
4.25251901e-01 -3.30723226e-01 1.64036199e-01 7.23343253e-01
6.28731012e-01 2.37448583e-03 7.69291937e-01 -9.30364251e-01
2.40642335e-02 3.96004140e-01 -7.21010685e-01 5.99175751e-01
4.69328284e-01 4.65240359e-01 4.77835327e-01 -5.24917662e-01
8.62398565e-01 8.18065584e-01 8.66848171e-01 5.74959636e-01
-1.32312846e+00 -4.09472376e-01 -5.05506434e-02 -3.24767172e-01
-1.10448873e+00 -8.93098861e-02 4.08403203e-02 -6.75907195e-01
7.87813783e-01 5.66849947e-01 5.24863720e-01 4.77123737e-01
3.61276597e-01 8.48144889e-01 1.13920462e+00 -2.08219036e-01
5.58955818e-02 -9.36648026e-02 4.67881590e-01 7.06084192e-01
-7.74099082e-02 -2.52664357e-01 -1.59321770e-01 -3.76209885e-01
1.02472579e+00 -2.04385966e-01 -2.08167106e-01 -1.85587537e-02
-1.55840576e+00 6.57233238e-01 3.96350622e-01 5.12058675e-01
-3.25235277e-01 -2.73970701e-03 5.29971540e-01 6.78392872e-02
3.35471034e-01 6.57627642e-01 -3.38832319e-01 -2.23348692e-01
-1.26842117e+00 2.81104684e-01 6.26308024e-01 9.17234361e-01
2.54455596e-01 -2.72486478e-01 -3.43920618e-01 9.69657421e-01
-1.24654174e-01 -1.46264734e-03 5.63799441e-01 -8.80699933e-01
1.39018729e-01 4.05950099e-01 -9.83816013e-03 -5.29437363e-01
-1.10770619e+00 -3.41609567e-01 -6.17585540e-01 3.42043579e-01
8.26216161e-01 -2.79971093e-01 -1.58340168e+00 1.37159419e+00
3.08121443e-01 -5.53825200e-02 -2.98127979e-01 9.94300544e-01
1.33528769e+00 3.65274221e-01 5.86551487e-01 -2.11868390e-01
1.63938200e+00 -8.24855149e-01 -4.82898265e-01 -3.42137456e-01
7.97962368e-01 -9.17993844e-01 1.30795968e+00 3.24912578e-01
-1.26403236e+00 -3.15183491e-01 -7.55932510e-01 -8.93645268e-03
-9.98581946e-02 -6.03539422e-02 6.92055404e-01 6.27688169e-01
-1.15455401e+00 4.56463337e-01 -1.11609042e+00 -5.36862195e-01
6.32933676e-01 6.46444798e-01 -2.86099434e-01 -1.20029658e-01
-5.93275130e-01 9.77225184e-01 3.85687709e-01 -3.22728932e-01
-6.10986054e-01 -1.05294657e+00 -5.81319988e-01 -3.25812310e-01
4.69010860e-01 -8.00871670e-01 1.63911951e+00 -7.79431760e-01
-9.39257205e-01 1.19965219e+00 -5.43451048e-02 -3.99645835e-01
8.83318603e-01 3.51335257e-02 -2.29818970e-01 6.21287346e-01
3.52852553e-01 1.20334065e+00 3.61358941e-01 -1.28310454e+00
-6.73094809e-01 -2.27844253e-01 1.34907514e-01 4.77540970e-01
2.73816258e-01 2.76055843e-01 -7.32789278e-01 -6.17951751e-01
2.84721732e-01 -1.02692902e+00 -7.51280367e-01 2.31085792e-02
-5.59635699e-01 -2.32369393e-01 8.49034667e-01 -6.32531285e-01
1.13536298e+00 -2.00953865e+00 -3.08141917e-01 8.97704586e-02
3.37693244e-01 2.32653916e-01 1.80858001e-01 8.84721428e-02
-2.39388049e-01 2.07054615e-01 -6.18231773e-01 -2.56885663e-02
-4.53161478e-01 2.79939830e-01 3.26434851e-01 3.34902853e-01
-6.13829195e-02 1.03064573e+00 -8.61359537e-01 -9.68643129e-01
3.80426198e-01 1.62485272e-01 -5.24675727e-01 -8.82120337e-03
-3.52786072e-02 9.08907115e-01 -3.01821709e-01 8.97521377e-01
2.97042876e-01 -7.84578741e-01 -7.45605975e-02 -3.92670855e-02
-8.09093639e-02 -3.24550122e-02 -9.17247057e-01 1.78244340e+00
-1.38809472e-01 4.63103563e-01 2.83193022e-01 -6.62006915e-01
3.38772178e-01 5.77680349e-01 1.01861179e+00 -4.02226388e-01
-2.17935145e-02 3.28407437e-01 3.88176590e-01 -7.64668047e-01
2.84172267e-01 -3.43222618e-01 3.76105658e-03 4.73306328e-01
-2.26192579e-01 -5.83088756e-01 4.17942137e-01 5.23979008e-01
1.28483808e+00 -2.35499308e-01 4.34522152e-01 -5.24176955e-01
1.90542221e-01 5.29836357e-01 1.85632825e-01 9.47011113e-01
-3.15211624e-01 1.13005793e+00 4.33823228e-01 -3.73835087e-01
-6.95577323e-01 -1.09809923e+00 -6.09478235e-01 1.13018441e+00
4.27573234e-01 -1.99823871e-01 -9.93514597e-01 -1.00415885e+00
-2.98435748e-01 6.25805199e-01 -4.68318433e-01 4.13008988e-01
-5.68712056e-01 -9.33173418e-01 3.66386682e-01 5.64660251e-01
3.54994774e-01 -1.27875793e+00 -1.20461333e+00 3.14304113e-01
-3.57129127e-01 -1.15348113e+00 -5.58779478e-01 2.92115092e-01
-1.13133574e+00 -1.14443862e+00 -1.19159365e+00 -7.28316426e-01
1.01844037e+00 1.66221872e-01 1.25105822e+00 3.05604339e-01
-6.61893129e-01 6.81754053e-01 -3.05259794e-01 -4.43191558e-01
-4.43116367e-01 2.67679483e-01 -4.87289548e-01 -7.31502473e-01
-5.33874705e-02 -1.18840234e-02 -8.88896644e-01 5.89787006e-01
-1.16860628e+00 3.39739740e-01 5.87300062e-01 8.48251283e-01
7.31516182e-01 -2.94953704e-01 4.98422116e-01 -1.41327035e+00
5.88006496e-01 -4.38611180e-01 -5.01064723e-03 5.65937422e-02
-4.65736926e-01 -4.29372340e-01 1.17450446e-01 -5.09658873e-01
-9.43107605e-01 2.28114054e-01 -3.77779484e-01 -2.82515623e-02
-5.94869733e-01 4.69252765e-01 6.20383143e-01 -8.36928412e-02
9.37804461e-01 -2.43807375e-01 7.06151873e-02 -3.54034871e-01
2.70307641e-02 4.60693151e-01 8.64980459e-01 -5.47007024e-01
2.58105129e-01 5.21685302e-01 -2.49122038e-01 -8.37459326e-01
-7.17664599e-01 -7.19176888e-01 -6.35290384e-01 -3.26163024e-01
1.10534108e+00 -3.73578817e-01 -2.54721522e-01 2.58534104e-01
-7.36530483e-01 -6.60027921e-01 -2.62697428e-01 3.61844927e-01
-5.47890544e-01 3.55014443e-01 -8.67450058e-01 -3.61333400e-01
-4.22597080e-01 -1.73673570e+00 1.21494806e+00 3.85262042e-01
-8.21374118e-01 -1.08470643e+00 -4.10958856e-01 5.17641902e-01
2.40402311e-01 5.89458406e-01 1.02816451e+00 -8.52257550e-01
-3.21009398e-01 -1.58764303e-01 -2.58118302e-01 -7.34649003e-02
3.67168188e-01 -1.32859200e-01 -5.34120083e-01 -2.27056578e-01
-2.11094752e-01 -1.23925231e-01 5.26971400e-01 9.47712243e-01
1.27173877e+00 1.58186719e-01 -5.97331047e-01 4.41650033e-01
1.18386567e+00 5.44466317e-01 4.32717472e-01 3.62139463e-01
4.78997737e-01 5.33393145e-01 8.20692182e-01 3.68615054e-02
1.47006974e-01 4.04352278e-01 3.32519263e-01 -7.89114833e-01
-1.90183818e-01 3.50084722e-01 -3.40074241e-01 2.92344749e-01
-9.08005461e-02 8.84412900e-02 -1.41616762e+00 5.25523305e-01
-1.67836010e+00 -4.78775769e-01 -4.85214204e-01 1.89891958e+00
8.55943620e-01 4.41286117e-01 3.95835549e-01 -1.36078089e-01
5.21889210e-01 -6.72894120e-02 -6.93764806e-01 -2.70372659e-01
3.11998487e-01 2.44912669e-01 5.21368504e-01 3.12493116e-01
-1.13268018e+00 7.93154478e-01 7.87178755e+00 7.92908907e-01
-1.22245884e+00 1.92627728e-01 1.01891100e+00 -9.39753503e-02
1.99534521e-02 -2.23120630e-01 -3.69101465e-01 4.57550406e-01
5.78691483e-01 -3.10430937e-02 -1.49907902e-01 6.92782283e-01
2.02998951e-01 -7.39467442e-01 -1.04890823e+00 9.30980504e-01
-1.81080893e-01 -1.38608801e+00 -3.69305670e-01 -8.39155912e-02
7.05367744e-01 9.23494324e-02 3.30836140e-03 1.04022712e-01
2.62179792e-01 -1.30761540e+00 3.09844553e-01 1.36922210e-01
1.13476431e+00 -4.80930865e-01 6.61008298e-01 5.04925668e-01
-7.94579327e-01 3.00991982e-01 1.06151171e-01 3.24825078e-01
4.00555015e-01 3.32367331e-01 -1.40845156e+00 3.95663857e-01
6.22299910e-01 2.48708367e-01 -5.78296483e-01 1.30508971e+00
7.74890557e-02 7.12490737e-01 -4.92864579e-01 4.34982598e-01
5.17770827e-01 4.37072106e-02 4.90734726e-01 1.65842843e+00
-1.48132341e-02 6.48553789e-01 3.95841330e-01 3.27623904e-01
2.79100358e-01 3.22122902e-01 -2.82195061e-01 3.44320327e-01
2.07160428e-01 1.51164305e+00 -1.53173792e+00 -7.02475011e-01
-3.07547927e-01 7.50365317e-01 -3.06040436e-01 3.61232251e-01
-8.17022562e-01 -5.84812202e-02 6.82586282e-02 4.64115739e-01
-1.68449774e-01 -6.75661415e-02 -7.47851849e-01 -5.57553470e-01
-3.35090131e-01 -9.35935378e-01 8.51118028e-01 -8.77375960e-01
-8.57000530e-01 7.10246205e-01 3.93677533e-01 -1.17135561e+00
-3.71785104e-01 -4.13676232e-01 -6.97024643e-01 6.25712991e-01
-9.88014638e-01 -9.96776044e-01 -3.67065430e-01 5.94698489e-01
9.12760615e-01 3.37871522e-01 7.62296438e-01 2.08183348e-01
-2.21126184e-01 4.56701964e-01 -3.60794544e-01 1.02148108e-01
8.01613808e-01 -1.48246896e+00 2.67714322e-01 5.22152483e-01
-6.49594367e-02 5.35323143e-01 8.27005982e-01 -6.09835386e-01
-7.08336413e-01 -6.61700726e-01 2.73876905e-01 -5.25920808e-01
2.25727290e-01 3.80341172e-01 -9.75568295e-01 8.03593814e-01
1.98667467e-01 6.17827438e-02 6.90073967e-01 -4.78924327e-02
2.23046362e-01 4.00985926e-01 -1.51599324e+00 5.82344592e-01
7.65913069e-01 -4.78088995e-03 -6.03373289e-01 7.24157512e-01
3.71958345e-01 -1.24595284e+00 -1.11571693e+00 5.98136902e-01
5.05242646e-01 -1.05487740e+00 9.21184242e-01 -4.32453185e-01
4.73120570e-01 1.24699250e-02 3.53908896e-01 -1.14622724e+00
-2.16494948e-01 -4.81127352e-01 3.83310884e-01 6.86180413e-01
5.80240071e-01 -6.01157844e-01 9.31808352e-01 9.24206436e-01
-4.77175534e-01 -1.07036459e+00 -8.45218837e-01 -4.56372529e-01
1.19241215e-01 -4.96507198e-01 9.98363793e-02 1.02559400e+00
7.47267855e-03 1.30322978e-01 2.34010264e-01 -2.15948429e-02
3.59842926e-01 -5.99341989e-02 6.01500690e-01 -9.57892656e-01
-2.07880139e-01 -5.66955924e-01 -1.59105614e-01 -1.00036442e+00
-3.69582653e-01 -8.02073717e-01 2.98997434e-03 -1.99366844e+00
3.48348826e-01 -7.59584069e-01 -8.69986191e-02 6.10080540e-01
-4.21256512e-01 4.91678804e-01 1.74163014e-01 3.50191832e-01
-4.58707899e-01 -5.24894297e-01 1.63207424e+00 1.14140511e-02
-2.90146500e-01 2.49599338e-01 -7.28950560e-01 1.01098728e+00
7.29434431e-01 -2.92034835e-01 -3.73992234e-01 -2.70125508e-01
-3.86691779e-01 4.91090089e-01 2.14165017e-01 -9.68443692e-01
2.70319343e-01 -1.35703638e-01 4.72915769e-01 -7.83984900e-01
1.48785859e-01 -5.90791702e-01 1.28447160e-01 7.16556728e-01
-3.28768551e-01 9.94176641e-02 3.93070459e-01 1.20484270e-01
-7.44647384e-02 -3.46537530e-01 1.13315618e+00 -4.73775536e-01
-7.33598471e-01 3.29127550e-01 -5.49809039e-01 4.12235647e-01
1.12397468e+00 -5.71174860e-01 -5.96183073e-03 -2.54883230e-01
-1.31319833e+00 4.78625685e-01 4.39437121e-01 2.45564103e-01
4.29544866e-01 -9.02888656e-01 -6.56152904e-01 -4.52214703e-02
7.10741654e-02 4.72063273e-01 3.81491959e-01 1.30180967e+00
-8.04997146e-01 2.60220766e-01 -2.58104261e-02 -1.22236061e+00
-1.65227497e+00 8.51969495e-02 3.93205851e-01 -4.69956189e-01
-1.04357016e+00 8.77208829e-01 4.67596114e-01 -2.29684502e-01
7.79902935e-02 -5.58226585e-01 -2.71185022e-02 -4.97925617e-02
3.81018400e-01 3.64391029e-01 1.43572360e-01 -5.37636280e-01
-4.08875942e-01 4.78446722e-01 -5.27795672e-01 -1.36100635e-01
1.05499601e+00 4.26711589e-02 2.69813985e-01 2.97828525e-01
7.78647602e-01 -3.46401006e-01 -1.14339912e+00 5.12181446e-02
7.34346285e-02 -2.26080209e-01 -9.78035778e-02 -1.10334253e+00
-8.91154766e-01 5.99389076e-01 6.64661229e-01 2.91118473e-01
1.09842563e+00 3.50354850e-01 7.35233128e-01 -2.79313236e-01
4.24123228e-01 -9.66589391e-01 2.29547657e-02 1.66520089e-01
6.68735683e-01 -1.36492968e+00 1.64290532e-01 -5.62132001e-01
-9.54870641e-01 9.92108166e-01 5.70272982e-01 1.07631758e-01
4.96968895e-01 4.99858052e-01 3.77968341e-01 -5.61389625e-01
-3.34151149e-01 -7.02486262e-02 4.40247089e-01 5.40273845e-01
6.36218429e-01 1.98608309e-01 -4.82554525e-01 1.91688821e-01
-1.83209702e-01 -7.47472197e-02 4.37214434e-01 1.11257613e+00
-4.92661864e-01 -8.50628316e-01 -6.39944434e-01 9.35477793e-01
-8.44185054e-01 8.08324851e-03 -1.41677991e-01 1.14629734e+00
2.25980897e-02 7.87889361e-01 -5.29697351e-02 2.09536329e-01
2.78204352e-01 -3.49719860e-02 4.57585931e-01 -9.81078744e-01
-8.83401573e-01 4.57875878e-01 9.52579379e-02 -5.94926953e-01
-3.60886842e-01 -7.45786130e-01 -1.73954618e+00 5.38149215e-02
-1.55322954e-01 -9.80676562e-02 5.72501898e-01 9.72226858e-01
-1.16394252e-01 7.21166492e-01 -1.59384869e-02 -7.94120789e-01
-2.46228769e-01 -8.38565111e-01 -4.29794282e-01 5.17244697e-01
2.38580629e-01 -4.59821761e-01 -6.93381131e-02 2.75921613e-01]
|
[14.708962440490723, -2.3083293437957764]
|
4c4a83f8-df10-4c33-8d78-52f46ad4f36e
|
state-wise-constrained-policy-optimization
|
2306.12594
| null |
https://arxiv.org/abs/2306.12594v2
|
https://arxiv.org/pdf/2306.12594v2.pdf
|
State-wise Constrained Policy Optimization
|
Reinforcement Learning (RL) algorithms have shown tremendous success in simulation environments, but their application to real-world problems faces significant challenges, with safety being a major concern. In particular, enforcing state-wise constraints is essential for many challenging tasks such as autonomous driving and robot manipulation. However, existing safe RL algorithms under the framework of Constrained Markov Decision Process (CMDP) do not consider state-wise constraints. To address this gap, we propose State-wise Constrained Policy Optimization (SCPO), the first general-purpose policy search algorithm for state-wise constrained reinforcement learning. SCPO provides guarantees for state-wise constraint satisfaction in expectation. In particular, we introduce the framework of Maximum Markov Decision Process, and prove that the worst-case safety violation is bounded under SCPO. We demonstrate the effectiveness of our approach on training neural network policies for extensive robot locomotion tasks, where the agent must satisfy a variety of state-wise safety constraints. Our results show that SCPO significantly outperforms existing methods and can handle state-wise constraints in high-dimensional robotics tasks.
|
['Changliu Liu', 'Tianhao Wei', 'Yifan Sun', 'Rui Chen', 'WeiYe Zhao']
|
2023-06-21
| null | null | null | null |
['robot-manipulation']
|
['robots']
|
[ 1.32508725e-01 1.12046063e-01 -6.33283615e-01 -6.08402416e-02
-4.59608734e-01 -3.27296138e-01 4.36334342e-01 7.81823620e-02
-7.63303459e-01 1.20416927e+00 -2.10271850e-01 -5.55729270e-01
-4.46066350e-01 -5.23994923e-01 -8.41948986e-01 -8.51991296e-01
-5.84940374e-01 4.68705505e-01 3.64034623e-01 -4.72845823e-01
2.66274244e-01 5.62442422e-01 -1.41315138e+00 -5.96739292e-01
8.92438293e-01 8.75936389e-01 2.84453332e-01 3.59191716e-01
6.44094288e-01 7.21857548e-01 -1.81792960e-01 3.38976473e-01
2.93426603e-01 -2.03001931e-01 -8.15774202e-01 -9.32998285e-02
-2.20584467e-01 -4.25850362e-01 -2.63093174e-01 1.13430476e+00
4.62737411e-01 6.00780845e-01 4.30589259e-01 -1.87639213e+00
1.76894307e-01 4.98642027e-01 -4.72328722e-01 -2.12124914e-01
6.48138579e-03 5.52113712e-01 8.53367269e-01 -1.10614866e-01
4.24796432e-01 1.37039959e+00 1.98532164e-01 9.83866513e-01
-1.04016888e+00 -6.13779724e-01 6.73028231e-01 1.89202458e-01
-8.72381330e-01 -8.66631791e-02 3.52279335e-01 -2.34095752e-01
1.13886344e+00 -3.29814672e-01 5.38725197e-01 1.21517694e+00
6.60790741e-01 9.03298020e-01 1.08107555e+00 -1.50182098e-01
8.11356664e-01 -4.75474983e-01 -1.14206627e-01 7.13095546e-01
4.65843618e-01 5.93114972e-01 -3.49702358e-01 -1.13258474e-01
5.79607785e-01 -2.71037549e-01 -1.23819089e-04 -8.60505641e-01
-1.21529806e+00 8.96580637e-01 -2.10723691e-02 -3.02745372e-01
-3.36616158e-01 4.97649431e-01 6.32145941e-01 2.38339618e-01
-2.07727998e-01 6.31677449e-01 -6.03033543e-01 -4.14646477e-01
-2.63226599e-01 6.38432086e-01 8.75436902e-01 1.01250446e+00
4.09201533e-01 4.21793252e-01 -9.62439403e-02 4.94458199e-01
2.60234118e-01 5.47379076e-01 1.95644408e-01 -1.46357930e+00
4.63075787e-01 -3.49876494e-03 6.39297426e-01 -5.94719470e-01
-5.55051148e-01 -2.18879774e-01 -5.98980010e-01 6.98827088e-01
2.33621791e-01 -6.01876140e-01 -6.63592935e-01 2.20128345e+00
3.93785208e-01 5.42969890e-02 3.08740616e-01 8.10979784e-01
-3.67164731e-01 6.56258404e-01 4.95077483e-02 -6.42627597e-01
8.77220571e-01 -8.17872226e-01 -8.27408195e-01 -4.48685199e-01
4.73378837e-01 1.51018485e-01 9.08945620e-01 6.17821634e-01
-1.04997325e+00 1.14099728e-02 -1.27986801e+00 4.95034993e-01
1.80340499e-01 -4.02708590e-01 5.84529102e-01 2.06134111e-01
-6.40583992e-01 6.25818014e-01 -1.19720507e+00 -2.99757838e-01
2.99155444e-01 5.19619286e-01 -2.16461167e-01 5.31301014e-02
-1.24550021e+00 1.39184880e+00 6.02882564e-01 4.48900694e-03
-1.55587220e+00 -2.36955211e-01 -1.02338421e+00 -5.84693477e-02
1.18782508e+00 -3.50976706e-01 1.71487498e+00 -1.07076742e-01
-1.95525074e+00 2.65720814e-01 1.18878156e-01 -7.65740335e-01
7.13990271e-01 -4.59711850e-01 2.14768611e-02 5.58000207e-02
1.28335893e-01 4.03892845e-01 8.48863542e-01 -1.21779788e+00
-9.12897706e-01 -9.73359961e-03 2.74010241e-01 3.29025269e-01
-2.83949137e-01 -1.68937698e-01 -4.59452420e-02 -4.13255543e-02
-3.76272202e-01 -1.16765809e+00 -8.44690561e-01 -5.72599955e-02
-1.31328136e-01 -4.39051539e-01 7.74014294e-01 4.19598520e-02
1.00804377e+00 -1.89907968e+00 3.92694056e-01 -5.85873351e-02
-2.67599314e-01 2.93910325e-01 -7.63147771e-02 4.96586651e-01
4.26328540e-01 -2.02605486e-01 -5.60288072e-01 -2.47037485e-01
4.30414885e-01 8.62582326e-01 -6.28643632e-01 5.31874359e-01
2.72896260e-01 4.96939093e-01 -1.22315848e+00 -5.09734511e-01
1.70004457e-01 -1.39822394e-01 -7.73157954e-01 2.61914134e-01
-6.63752854e-01 4.63859439e-01 -6.80044532e-01 3.22901249e-01
2.72555083e-01 2.76579231e-01 3.13654214e-01 5.59587657e-01
-3.06380361e-01 -3.49830613e-02 -1.06880617e+00 1.42754471e+00
-5.68025768e-01 2.72150367e-01 4.07039762e-01 -1.10381067e+00
6.03181660e-01 1.04226708e-01 7.35603273e-01 -5.71595728e-01
3.19536567e-01 9.65564698e-02 7.68528134e-02 -4.11571622e-01
5.12871683e-01 -3.86087686e-01 -5.34038842e-01 3.37201387e-01
-2.48584583e-01 -5.43090582e-01 2.73033261e-01 2.40991055e-03
1.16873837e+00 5.06993115e-01 5.96013606e-01 -3.13688904e-01
3.63642424e-01 2.18931898e-01 1.10271406e+00 8.45625401e-01
-7.82240212e-01 -2.89285094e-01 9.49434936e-01 6.25466853e-02
-7.12717533e-01 -6.86128914e-01 1.59883618e-01 9.50605214e-01
4.63326186e-01 -9.04837176e-02 -5.74891031e-01 -6.43779218e-01
2.87343353e-01 9.50486422e-01 -5.28574169e-01 -4.85944182e-01
-6.89773321e-01 -2.58703679e-01 3.94387662e-01 4.30288911e-01
4.49285001e-01 -1.21682894e+00 -1.45619786e+00 4.47013259e-01
1.04329228e-01 -1.23443377e+00 -2.83585638e-01 6.23326361e-01
-6.25559390e-01 -9.97380197e-01 -2.63976276e-01 -6.34387195e-01
5.13713360e-01 9.36917439e-02 5.04890025e-01 -1.66805208e-01
-3.02639753e-02 4.21215177e-01 -5.65056205e-02 -3.78440112e-01
-3.22365403e-01 -1.24216177e-01 6.01828873e-01 -4.47431982e-01
-2.10728660e-01 -2.72427738e-01 -1.81458861e-01 6.18095517e-01
-7.37467945e-01 -1.60654783e-01 3.13382834e-01 1.01061511e+00
6.68748140e-01 3.99878770e-01 7.47615457e-01 -3.81575108e-01
8.76024127e-01 -3.45439523e-01 -1.16471744e+00 8.52807090e-02
-6.31850302e-01 3.87632817e-01 8.00976992e-01 -5.30722141e-01
-9.53871489e-01 1.59209713e-01 6.59618229e-02 -3.47388417e-01
-6.22311272e-02 5.15668273e-01 -3.24251622e-01 1.63044885e-01
3.19721371e-01 1.21906407e-01 2.41443932e-01 8.63618478e-02
4.74870652e-02 2.50751764e-01 5.14360070e-01 -1.19575536e+00
6.99533939e-01 3.18056405e-01 6.40074670e-01 -5.64801872e-01
-7.21025527e-01 -1.65747926e-01 -3.08153659e-01 -3.68006945e-01
7.22450137e-01 -5.60176790e-01 -1.33513975e+00 3.68866682e-01
-7.49992609e-01 -9.79436517e-01 -1.87417552e-01 4.53696966e-01
-1.39881551e+00 3.24244589e-01 -3.85641128e-01 -1.26317739e+00
2.06828360e-02 -1.33067572e+00 6.95475876e-01 1.97940290e-01
1.87774450e-02 -6.60877585e-01 1.56136617e-01 -2.24830091e-01
1.75021440e-01 5.26503503e-01 9.47347105e-01 -1.99196219e-01
-2.71787614e-01 9.71244723e-02 2.33487308e-01 1.73637033e-01
1.30546186e-02 -2.24106312e-01 -3.01290035e-01 -7.34305263e-01
-7.26933032e-02 -7.72802472e-01 5.49590588e-01 3.51489007e-01
1.07932913e+00 -3.46594810e-01 -3.60232681e-01 2.43716925e-01
1.41560721e+00 5.39337099e-01 1.86502501e-01 6.05798185e-01
2.54077852e-01 6.51939213e-01 1.46601415e+00 8.70845079e-01
2.61540055e-01 5.82005024e-01 1.10633683e+00 4.91500914e-01
6.02575898e-01 -3.47124428e-01 7.80663610e-01 1.50806651e-01
2.08686545e-01 -2.70903289e-01 -8.11138570e-01 6.07061327e-01
-2.40605330e+00 -8.40687513e-01 2.37985015e-01 2.34669304e+00
9.21182632e-01 5.30449688e-01 2.41218418e-01 6.57978430e-02
5.57893634e-01 -1.02451749e-01 -1.21354187e+00 -7.61223316e-01
3.99033099e-01 -1.45512111e-02 7.67164946e-01 6.50401711e-01
-1.06629074e+00 1.07019997e+00 6.29170609e+00 6.71025872e-01
-1.03640330e+00 -1.59271583e-01 -1.08609691e-01 -1.70642018e-01
2.92668581e-01 -2.99223065e-02 -8.44984770e-01 3.65984529e-01
7.01970100e-01 -4.05045748e-01 4.97805566e-01 1.24577820e+00
6.08628392e-01 -4.46794152e-01 -1.12186384e+00 5.66796541e-01
-4.48716015e-01 -6.32014096e-01 -4.74540502e-01 2.28550643e-01
7.29514599e-01 -8.37225467e-02 1.30812585e-01 5.91371357e-01
1.03177178e+00 -8.71054411e-01 8.53150845e-01 -1.99399702e-02
6.51801109e-01 -1.24875271e+00 4.40026581e-01 7.19893694e-01
-9.24638987e-01 -6.04197800e-01 -3.54046762e-01 -3.42818290e-01
5.22303462e-01 1.53852850e-01 -6.42494977e-01 3.76964867e-01
3.44960630e-01 6.84850276e-01 3.54701847e-01 9.52744603e-01
-6.93326771e-01 5.21028996e-01 -4.66323674e-01 -3.77042562e-01
6.24306619e-01 -1.23773851e-01 6.12287939e-01 7.23975718e-01
3.80760655e-02 -3.82278711e-02 5.36922216e-01 4.40121740e-01
3.81592005e-01 -4.51458454e-01 -7.06999600e-01 -1.00641400e-01
4.53376383e-01 7.95268595e-01 -6.44866884e-01 1.01422211e-02
7.03049749e-02 6.87533736e-01 4.30170685e-01 1.84634328e-01
-1.13724422e+00 -4.44804281e-01 1.21936941e+00 -4.87305820e-01
3.70176673e-01 -9.19407129e-01 -1.79186672e-01 -8.01835656e-01
-1.22802630e-01 -8.54807556e-01 3.06297302e-01 -3.76047581e-01
-8.79548371e-01 1.35434479e-01 2.70000428e-01 -1.23895562e+00
-4.63045895e-01 -6.56561792e-01 -4.72086847e-01 2.86785930e-01
-1.65291393e+00 -6.43907785e-01 1.70864448e-01 5.65966487e-01
6.57920241e-01 -3.85779627e-02 6.09025121e-01 -1.65046945e-01
-9.02524233e-01 2.18598679e-01 5.23518808e-02 -4.02213007e-01
6.30541861e-01 -1.10349429e+00 2.09449232e-02 9.55464482e-01
-7.59944677e-01 4.25331801e-01 1.06582439e+00 -7.07823694e-01
-1.82748401e+00 -1.07304549e+00 2.56523430e-01 -2.50987094e-02
8.56391072e-01 -8.10645670e-02 -5.58385730e-01 6.25026047e-01
-7.70479217e-02 -7.76542649e-02 -1.12708220e-02 -1.72467858e-01
5.12953997e-02 8.06743205e-02 -1.11587512e+00 9.49980319e-01
1.16402245e+00 -1.19493119e-02 -5.24235427e-01 2.13147685e-01
8.48238945e-01 -4.80956674e-01 -6.09679699e-01 5.45742512e-01
4.66320515e-01 -3.87598068e-01 6.45380497e-01 -1.05896425e+00
3.78562331e-01 -4.14200276e-01 -1.52767941e-01 -1.55794370e+00
-1.78517714e-01 -9.26090777e-01 -2.94658005e-01 6.18891358e-01
1.69481322e-01 -7.43099034e-01 5.92058241e-01 5.56530118e-01
-4.12694395e-01 -9.50912118e-01 -1.19528806e+00 -1.46620488e+00
3.91204864e-01 -4.35580581e-01 3.58731300e-01 6.50985658e-01
5.50433040e-01 -4.04861644e-02 -8.31792593e-01 4.07317311e-01
6.91710413e-01 -9.12566110e-02 6.74733281e-01 -6.96709454e-01
-3.27547461e-01 -4.57546443e-01 1.12963282e-01 -9.47349370e-01
7.14769423e-01 -4.29756612e-01 9.75147069e-01 -1.67569482e+00
-1.49125785e-01 -6.02890074e-01 -2.68321842e-01 7.70740807e-01
-5.45583889e-02 -5.67971528e-01 2.48114154e-01 -1.61612570e-01
-9.22354877e-01 9.68762338e-01 1.38152051e+00 -1.06339399e-02
-2.32883796e-01 7.65989199e-02 -3.80997986e-01 6.20413125e-01
1.21534228e+00 -4.11861122e-01 -6.82777584e-01 -2.37334639e-01
2.26471245e-01 4.46847469e-01 -1.16477743e-01 -1.12442231e+00
1.67029560e-01 -1.21947050e+00 -5.84961593e-01 -4.69809145e-01
3.81441891e-01 -9.22049463e-01 -2.66970396e-01 1.26961958e+00
-6.07309461e-01 -3.54475826e-02 2.01043636e-01 9.09859538e-01
1.69237301e-01 -2.39846870e-01 1.03785050e+00 1.74919575e-01
-1.02781558e+00 4.47642088e-01 -1.01152325e+00 1.96458519e-01
1.53791130e+00 3.06946665e-01 -1.56980351e-01 -3.19437176e-01
-4.22066420e-01 1.14880800e+00 3.91856194e-01 4.04509515e-01
6.71009004e-01 -9.19824362e-01 -2.68212616e-01 -1.00915454e-01
3.35187241e-02 1.72320694e-01 -1.72823623e-01 6.93708956e-01
-1.32194698e-01 4.58036214e-01 -5.90098560e-01 -4.15901929e-01
-1.05199742e+00 8.34215045e-01 2.95955002e-01 -3.84456843e-01
-5.23665726e-01 6.11762881e-01 -1.75151341e-02 -2.82979310e-01
4.89912599e-01 -5.17338455e-01 -2.78689452e-02 -2.61406600e-01
1.97589949e-01 3.32766593e-01 -3.21956336e-01 -1.88822299e-01
-3.71475220e-01 2.81183004e-01 3.96771282e-02 -4.91806895e-01
1.28654611e+00 -2.17375252e-02 2.14888230e-01 2.03463212e-01
5.48453510e-01 -6.01853609e-01 -1.90658879e+00 -7.41779851e-03
1.49592206e-01 -2.21530542e-01 -4.39974405e-02 -6.90345526e-01
-5.65996826e-01 8.39264512e-01 1.73446193e-01 -2.63668686e-01
7.55891621e-01 -4.93047893e-01 7.29014575e-01 9.65501070e-01
1.22853184e+00 -1.47259045e+00 2.54500180e-01 1.08271408e+00
6.23797238e-01 -9.81821477e-01 -7.91265741e-02 -1.17898606e-01
-9.11889493e-01 6.97152615e-01 1.04545164e+00 -2.54671067e-01
3.86923850e-01 3.32405150e-01 -4.20758128e-01 4.05052006e-01
-1.18618190e+00 -4.08342391e-01 -5.50489664e-01 7.62281418e-01
-4.49456871e-01 1.60671860e-01 -5.36172330e-01 3.36504817e-01
1.08440228e-01 8.71322304e-02 6.96779251e-01 1.64596510e+00
-8.52278054e-01 -1.14009047e+00 -2.00139552e-01 8.15787613e-02
-1.42769918e-01 5.00569940e-01 3.27346399e-02 7.45703697e-01
-2.51564264e-01 1.04180586e+00 -2.55487800e-01 -2.12351039e-01
2.56760418e-01 -1.24664553e-01 6.95603192e-01 -5.45394242e-01
-1.18546315e-01 -2.53389269e-01 2.53554791e-01 -7.81758368e-01
-2.26254448e-01 -6.96623862e-01 -1.77888703e+00 -2.01473132e-01
-1.56672701e-01 1.99072406e-01 7.18265057e-01 1.14171076e+00
8.92164484e-02 5.29099882e-01 7.54371345e-01 -7.30660319e-01
-1.29472804e+00 -4.28749114e-01 -4.91319388e-01 6.17086627e-02
6.28857970e-01 -1.24670887e+00 -2.96199530e-01 -3.91377270e-01]
|
[4.584259510040283, 2.018918991088867]
|
416f43ec-7ada-456f-9f69-372fac1896e7
|
trec-cast-2019-the-conversational-assistance
|
2003.13624
| null |
https://arxiv.org/abs/2003.13624v1
|
https://arxiv.org/pdf/2003.13624v1.pdf
|
TREC CAsT 2019: The Conversational Assistance Track Overview
|
The Conversational Assistance Track (CAsT) is a new track for TREC 2019 to facilitate Conversational Information Seeking (CIS) research and to create a large-scale reusable test collection for conversational search systems. The document corpus is 38,426,252 passages from the TREC Complex Answer Retrieval (CAR) and Microsoft MAchine Reading COmprehension (MARCO) datasets. Eighty information seeking dialogues (30 train, 50 test) are an average of 9 to 10 questions long. Relevance assessments are provided for 30 training topics and 20 test topics. This year 21 groups submitted a total of 65 runs using varying methods for conversational query understanding and ranking. Methods include traditional retrieval based methods, feature based learning-to-rank, neural models, and knowledge enhanced methods. A common theme through the runs is the use of BERT-based neural reranking methods. Leading methods also employed document expansion, conversational query expansion, and generative language models for conversational query rewriting (GPT-2). The results show a gap between automatic systems and those using the manually resolved utterances, with a 35% relative improvement of manual rewrites over the best automatic system.
|
['Jamie Callan', 'Chenyan Xiong', 'Jeffrey Dalton']
|
2020-03-30
| null | null | null | null |
['conversational-search']
|
['natural-language-processing']
|
[ 3.00061822e-01 4.68644798e-01 7.90875852e-02 -2.86241889e-01
-1.56750357e+00 -7.53440022e-01 1.28734529e+00 1.54118612e-01
-6.43956721e-01 9.21972990e-01 9.71244156e-01 -7.64202774e-01
-4.03059930e-01 -1.73235267e-01 -4.74988185e-02 1.30634019e-02
4.06405441e-02 1.13017845e+00 3.34467262e-01 -9.80424881e-01
1.09011614e+00 -8.77059475e-02 -1.40590668e+00 8.87010038e-01
1.03905547e+00 6.05378866e-01 1.80768877e-01 1.53856611e+00
-5.77870965e-01 1.07588625e+00 -9.06148732e-01 -4.74279732e-01
-3.56095195e-01 -3.90247524e-01 -1.71573377e+00 -7.15179861e-01
3.66841525e-01 -3.08518499e-01 -1.84890017e-01 2.44270459e-01
7.99693406e-01 4.69915807e-01 8.33562016e-01 -8.36610317e-01
-6.95329785e-01 4.80265141e-01 9.75648239e-02 5.41944742e-01
1.27018785e+00 -3.36791724e-01 1.10415125e+00 -1.01008976e+00
7.64224589e-01 1.72563350e+00 1.71217009e-01 8.07967603e-01
-6.14879191e-01 -2.64712185e-01 -2.74763793e-01 4.27577078e-01
-9.06015992e-01 -6.04058564e-01 1.83758184e-01 -3.72882754e-01
1.74148595e+00 7.20456660e-01 3.54262322e-01 9.40995753e-01
1.57292366e-01 9.34202909e-01 8.98757279e-01 -8.92809272e-01
-4.20918614e-02 4.20188338e-01 6.62414789e-01 4.51154381e-01
-5.11482537e-01 -7.67015144e-02 -5.37130237e-01 -7.64431477e-01
-9.75444689e-02 -2.29862064e-01 -2.90203542e-01 3.49117815e-01
-1.03951859e+00 9.80988860e-01 -2.50819996e-02 3.74498874e-01
-2.94726878e-01 -1.92947000e-01 5.83828390e-01 9.50874627e-01
3.58361304e-01 1.27599728e+00 -5.06349206e-01 -6.54629767e-01
-2.48527989e-01 6.86737239e-01 1.62207842e+00 8.93064439e-01
6.12130225e-01 -7.31215715e-01 -8.84437382e-01 1.33146560e+00
4.13669020e-01 5.97303391e-01 7.70702958e-01 -1.43875241e+00
7.77693093e-01 6.67034268e-01 2.47501418e-01 -1.04422629e+00
-6.32190332e-02 6.79003447e-02 -2.31021211e-01 -7.77666450e-01
9.05478895e-02 -1.97945133e-01 -5.09382367e-01 1.24290693e+00
-1.69397458e-01 -7.35112786e-01 4.34711546e-01 2.93141901e-01
1.09310544e+00 9.03297246e-01 3.22350971e-02 -1.80079743e-01
1.35873520e+00 -1.39639914e+00 -7.41742551e-01 8.20244178e-02
1.02563536e+00 -1.28893173e+00 9.12783921e-01 1.53889120e-01
-1.19160163e+00 -2.46892303e-01 -5.66959023e-01 -3.17223966e-01
-7.37938821e-01 -2.03130648e-01 4.55554575e-01 5.75020492e-01
-1.56932092e+00 -7.43056983e-02 -1.12492278e-01 -7.73661554e-01
-5.82195938e-01 1.82535961e-01 2.66197443e-01 -2.58604258e-01
-1.53968155e+00 1.31850457e+00 -9.89799201e-02 -1.37711376e-01
-7.55010426e-01 -3.61502916e-01 -5.38340926e-01 1.04677223e-01
2.36152455e-01 -5.50547123e-01 1.99217570e+00 -3.06954443e-01
-1.70761013e+00 7.37309515e-01 -5.53729057e-01 -4.52037424e-01
-9.26388800e-03 -5.26935279e-01 -2.39565000e-01 4.12901044e-01
3.32141906e-01 9.59061623e-01 2.80463308e-01 -8.58980119e-01
-8.36515367e-01 5.81761375e-02 3.36831868e-01 8.89199138e-01
2.98687667e-02 4.47151750e-01 -4.92580801e-01 -3.29551883e-02
-1.90845162e-01 -1.00481844e+00 1.01920933e-01 -8.79596710e-01
-2.37589300e-01 -1.26238263e+00 7.70189285e-01 -1.00530112e+00
1.42372024e+00 -1.37084329e+00 -1.74528472e-02 3.98437604e-02
1.39252514e-01 4.24852997e-01 -3.13462317e-01 1.22115898e+00
3.39254677e-01 3.80169332e-01 3.62758577e-01 -1.58696756e-01
1.14786759e-01 -1.03675388e-01 -4.24366325e-01 -6.33216321e-01
-2.24739969e-01 1.04850698e+00 -9.47883844e-01 -6.05662942e-01
-5.74538633e-02 9.78134125e-02 -4.61542696e-01 4.30489331e-01
-2.88958192e-01 1.23864777e-01 -7.42451787e-01 6.55579627e-01
-1.88479945e-01 -2.68446892e-01 -2.57162482e-01 4.59018111e-01
8.21765326e-03 8.31989050e-01 -2.33956784e-01 1.60307908e+00
-5.16243279e-01 9.23438072e-01 -1.51738286e-01 -3.66035879e-01
8.65513206e-01 7.38480747e-01 2.15482309e-01 -1.05119050e+00
-1.80592358e-01 3.41606051e-01 -1.36585042e-01 -8.09934556e-01
1.05143034e+00 5.45023501e-01 -2.66593099e-01 9.98459876e-01
-2.00531974e-01 -4.68893468e-01 4.19795156e-01 8.22717547e-01
1.42307544e+00 -3.99853140e-01 1.84517968e-02 -1.29862487e-01
8.59884679e-01 3.77082884e-01 -2.56563872e-01 1.29076719e+00
-2.23744780e-01 2.53436834e-01 3.23320210e-01 -1.00173593e-01
-6.75434053e-01 -5.34098804e-01 3.18041205e-01 1.74069309e+00
-2.91921437e-01 -5.45039296e-01 -9.34420764e-01 -5.52671671e-01
-2.41307527e-01 9.46093082e-01 -2.92285174e-01 -2.36608520e-01
-5.98693669e-01 -1.28454119e-01 7.13407576e-01 7.75522664e-02
5.44004500e-01 -1.49049306e+00 -1.36408195e-01 1.61561817e-01
-8.47720623e-01 -7.46149480e-01 -8.00275743e-01 -2.37977609e-01
-6.22238994e-01 -1.05983901e+00 -1.04430521e+00 -9.82467115e-01
1.71482652e-01 5.75899541e-01 1.41033292e+00 3.68650913e-01
-1.61924288e-01 1.27927351e+00 -7.22180307e-01 -4.04315770e-01
-6.51105046e-01 5.41033387e-01 -2.83242673e-01 -8.24645579e-01
7.75794089e-01 -3.05525027e-02 -6.48893178e-01 5.03702402e-01
-5.22597432e-01 -2.28649870e-01 6.17855489e-01 1.10040748e+00
-3.76185149e-01 -8.49070311e-01 9.53812122e-01 -5.52234232e-01
1.95658898e+00 -4.41685587e-01 -5.30562811e-02 9.92441416e-01
-1.04422367e+00 9.27458555e-02 -3.12624156e-01 -1.29680023e-01
-1.52393043e+00 -5.96588194e-01 6.28712103e-02 3.70943576e-01
7.37219006e-02 7.63775766e-01 6.61099792e-01 8.57309848e-02
1.10932076e+00 7.80942738e-02 2.57106237e-02 -2.66396940e-01
4.22353029e-01 1.21126640e+00 2.32456058e-01 -7.76156545e-01
2.06952915e-01 -4.90093142e-01 -9.37413990e-01 -9.22326386e-01
-6.23564482e-01 -1.15783429e+00 -2.05473229e-01 -5.37827253e-01
8.77989531e-01 -5.97893238e-01 -8.90657961e-01 2.02598602e-01
-1.51120257e+00 -2.00416028e-01 9.76241976e-02 3.92575085e-01
-3.02001029e-01 4.56235111e-01 -9.02838886e-01 -1.07628572e+00
-1.05474317e+00 -1.11941981e+00 1.08972669e+00 4.70950812e-01
-7.01229811e-01 -9.86470461e-01 6.38483703e-01 1.18072629e+00
8.71526599e-01 -6.16831958e-01 1.22618401e+00 -1.14648712e+00
-4.22825754e-01 -3.55381995e-01 -1.98966935e-01 2.10356295e-01
-9.48597714e-02 -3.41110140e-01 -7.85037041e-01 -2.83626974e-01
-2.61232734e-01 -9.61053729e-01 6.62313640e-01 -7.13045225e-02
4.62727398e-01 -3.82132769e-01 -2.81860262e-01 -6.94491386e-01
6.04757130e-01 8.42917860e-01 6.72735214e-01 2.27655679e-01
5.02287745e-02 1.09695315e+00 4.81261551e-01 -1.58268347e-01
6.65073991e-01 5.51939070e-01 -2.32177556e-01 5.40848374e-01
8.38000551e-02 -2.29564473e-01 3.53752404e-01 1.23711145e+00
-3.56371514e-03 -3.97128761e-01 -1.18375838e+00 5.03580332e-01
-1.72542310e+00 -9.81665373e-01 2.84194946e-01 1.83876455e+00
9.20396209e-01 -1.44981503e-01 -2.98230827e-01 -4.19544995e-01
5.46857476e-01 -2.30561003e-01 -4.75293696e-01 -8.36308658e-01
9.52863023e-02 2.02458903e-01 -8.37219208e-02 1.06742191e+00
-5.03768504e-01 8.64875913e-01 7.12466764e+00 6.01466119e-01
-4.39904571e-01 -1.83234334e-01 5.95574677e-01 3.42261940e-01
-4.68801230e-01 1.53111622e-01 -9.30178225e-01 -1.53880212e-02
1.27381694e+00 -5.70513725e-01 5.91442108e-01 7.37105191e-01
-1.28714100e-01 -4.66005296e-01 -9.61826861e-01 6.91165209e-01
5.09249985e-01 -1.29623795e+00 1.80602625e-01 -1.28761858e-01
7.41260886e-01 2.25113049e-01 -3.19105923e-01 1.34908199e+00
5.82054496e-01 -9.36711609e-01 -2.15615690e-01 9.60712433e-01
1.76505193e-01 -3.17418486e-01 9.82803345e-01 4.82917547e-01
-5.96687376e-01 -3.69095430e-02 -2.10048139e-01 5.86647838e-02
-7.62127787e-02 -3.65404665e-01 -1.51463878e+00 3.01841110e-01
6.56299353e-01 1.51730821e-01 -8.65423262e-01 8.45051944e-01
2.45538726e-01 4.95494366e-01 -1.46932393e-01 -9.92827594e-01
4.88906980e-01 9.36275721e-02 7.72301257e-01 1.22809458e+00
-4.27183621e-02 2.98320293e-01 7.05260262e-02 1.54938221e-01
-2.24131322e-03 3.80961388e-01 -6.26503050e-01 -1.53531581e-01
7.58741379e-01 9.90435898e-01 -1.57518908e-01 -5.08361697e-01
-9.07306895e-02 7.99707592e-01 1.46158561e-01 6.77037001e-01
1.11920074e-01 -7.88971186e-01 -1.93286359e-01 -4.16028023e-01
-2.62575895e-01 1.41720576e-02 3.93597096e-01 -8.54713023e-01
1.27060022e-02 -1.52499104e+00 4.57324445e-01 -9.53060865e-01
-1.32381916e+00 7.26616442e-01 4.26228702e-01 -6.15690470e-01
-1.33258808e+00 -2.44893759e-01 -7.42595911e-01 1.08148170e+00
-1.24979901e+00 -6.76409483e-01 -1.76917970e-01 4.68289316e-01
1.13073838e+00 -7.04896688e-01 1.18281019e+00 2.36011192e-01
-2.73433439e-02 4.53618407e-01 2.53601670e-01 -8.93388465e-02
9.78083909e-01 -1.35104048e+00 3.36593240e-02 -1.59113351e-02
-2.63955086e-01 1.29263139e+00 4.28032905e-01 -4.79729146e-01
-1.26938188e+00 -2.58775592e-01 1.68134427e+00 -9.51163709e-01
4.29496378e-01 1.25166113e-02 -8.55975986e-01 3.82520825e-01
1.00316060e+00 -1.32689083e+00 7.23910928e-01 4.91082191e-01
-2.60347631e-02 4.48128730e-02 -7.60854661e-01 6.79443538e-01
3.31652701e-01 -1.09049213e+00 -1.14149117e+00 8.00079584e-01
1.15673423e+00 -1.64924130e-01 -6.52084887e-01 2.96445966e-01
5.75374663e-01 -6.02196515e-01 8.30255508e-01 -7.00238168e-01
1.20109729e-01 2.74247706e-01 -1.95261598e-01 -9.91503894e-01
1.76875412e-01 -1.01147139e+00 -3.30560580e-02 1.09369373e+00
9.66423631e-01 -6.41916215e-01 5.09437799e-01 1.19855642e+00
-3.49263251e-01 -8.02461922e-01 -7.91010678e-01 -1.15552716e-01
2.06120014e-01 5.64741064e-03 2.13450179e-01 6.25663221e-01
5.60175002e-01 1.21329451e+00 6.12250939e-02 -7.31029391e-01
-1.11721374e-01 -3.95676643e-01 8.39053273e-01 -1.13335347e+00
-2.21510138e-02 -7.59614646e-01 2.34188095e-01 -1.59477353e+00
-4.84800152e-02 -6.41940653e-01 1.80897325e-01 -1.68961024e+00
2.60760367e-01 -9.39988643e-02 5.28607070e-02 9.49523374e-02
-2.14816645e-01 -5.88861942e-01 -4.99441847e-02 5.42548239e-01
-1.10698617e+00 5.80967784e-01 1.22945464e+00 -2.99628645e-01
-4.25184250e-01 3.64032447e-01 -8.88397694e-01 2.68812001e-01
6.83968723e-01 -1.16975859e-01 -7.70091951e-01 -3.60110790e-01
4.14893836e-01 6.89087451e-01 -2.41781585e-02 -4.42312062e-01
8.80503237e-01 2.39638224e-01 -2.22604319e-01 -1.01882446e+00
4.26741391e-01 -1.24995440e-01 -6.87473118e-01 2.95378149e-01
-1.42680824e+00 5.48731506e-01 -4.99828868e-02 5.45822203e-01
-4.81371909e-01 -5.54190695e-01 -5.15110567e-02 -3.63753438e-01
-4.23237801e-01 -3.65959197e-01 -9.40808237e-01 3.91308367e-01
2.51270920e-01 1.04302689e-01 -7.33170867e-01 -1.37872708e+00
-5.13751388e-01 8.64474833e-01 -4.79122132e-01 9.28252876e-01
7.22280562e-01 -9.14377809e-01 -8.58527660e-01 -4.64135617e-01
2.55775928e-01 -2.49734655e-01 5.35011701e-02 5.18226147e-01
-3.54082823e-01 1.62068880e+00 3.69693816e-01 -6.03547871e-01
-1.36557984e+00 -1.71093717e-01 3.76347184e-01 -7.89669275e-01
3.14828381e-02 7.78814673e-01 -2.88600057e-01 -1.02904510e+00
7.43755758e-01 5.23821563e-02 -8.79068971e-01 1.55279219e-01
8.29221606e-01 5.59136808e-01 1.64350852e-01 -4.10043634e-02
1.90949608e-02 2.49730393e-01 -7.17588902e-01 -8.32514584e-01
5.05263925e-01 -4.45895344e-01 -2.76378810e-01 4.17440534e-01
1.46763241e+00 -3.85519803e-01 -2.42132153e-02 -5.84010839e-01
6.01406991e-01 7.15759769e-02 -2.22847492e-01 -1.37397850e+00
6.01653010e-02 7.53680050e-01 5.94731212e-01 5.63599408e-01
6.76370978e-01 6.50214776e-02 7.08588839e-01 1.80599451e+00
2.23209456e-01 -1.30585682e+00 4.92018491e-01 1.24827242e+00
1.46787214e+00 -1.26629555e+00 -2.53073394e-01 2.30853111e-01
-7.48016357e-01 1.01393116e+00 6.37278557e-01 4.61301029e-01
5.79734504e-01 -7.36438453e-01 3.86186779e-01 -5.87837338e-01
-1.26756954e+00 7.45990500e-02 6.01304948e-01 3.13138187e-01
7.77978957e-01 -3.95926207e-01 -6.15404963e-01 -5.97287379e-02
-4.55304742e-01 -3.25413376e-01 7.52821565e-02 1.15038466e+00
-6.73633337e-01 -1.02882743e+00 -1.00817986e-01 5.59680820e-01
-3.41131002e-01 -4.72189128e-01 -9.81522083e-01 6.36766970e-01
-1.04321742e+00 1.81489956e+00 -2.38085866e-01 -3.27257752e-01
2.60967165e-01 5.22220850e-01 -1.73881695e-01 -6.61311090e-01
-1.02392316e+00 -9.48658586e-02 8.90503168e-01 -3.91553521e-01
-5.60441673e-01 -6.77048981e-01 -6.34266436e-01 -3.16839293e-02
-6.72052562e-01 1.22352433e+00 8.07660818e-01 9.77305353e-01
6.18302882e-01 -1.29773775e-02 7.80280709e-01 -2.51206517e-01
-7.08008707e-01 -1.70201671e+00 3.25218230e-01 6.22403659e-02
1.21233582e-01 -3.06447685e-01 -5.19597411e-01 -9.10326391e-02]
|
[12.123334884643555, 7.823229789733887]
|
c2c4edf3-0cbc-407a-955a-1d929eb10b4c
|
disaster-anomaly-detector-via-deeper-fcdds
|
2306.02517
| null |
https://arxiv.org/abs/2306.02517v2
|
https://arxiv.org/pdf/2306.02517v2.pdf
|
Disaster Anomaly Detector via Deeper FCDDs for Explainable Initial Responses
|
Extreme natural disasters can have devastating effects on both urban and rural areas. In any disaster event, an initial response is the key to rescue within 72 hours and prompt recovery. During the initial stage of disaster response, it is important to quickly assess the damage over a wide area and identify priority areas. Among machine learning algorithms, deep anomaly detection is effective in detecting devastation features that are different from everyday features. In addition, explainable computer vision applications should justify the initial responses. In this paper, we propose an anomaly detection application utilizing deeper fully convolutional data descriptions (FCDDs), that enables the localization of devastation features and visualization of damage-marked heatmaps. More specifically, we show numerous training and test results for a dataset AIDER with the four disaster categories: collapsed buildings, traffic incidents, fires, and flooded areas. We also implement ablation studies of anomalous class imbalance and the data scale competing against the normal class. Our experiments provide results of high accuracies over 95% for F1. Furthermore, we found that the deeper FCDD with a VGG16 backbone consistently outperformed other baselines CNN27, ResNet101, and Inceptionv3. This study presents a new solution that offers a disaster anomaly detection application for initial responses with higher accuracy and devastation explainability, providing a novel contribution to the prompt disaster recovery problem in the research area of anomaly scene understanding. Finally, we discuss future works to improve more robust, explainable applications for effective initial responses.
|
['Junichiro Fujii', 'Masahiro Okano', 'Takato Yasuno']
|
2023-06-05
| null | null | null | null |
['scene-understanding']
|
['computer-vision']
|
[-3.17550749e-02 -2.05024704e-03 2.77975649e-01 -2.71621048e-01
-1.17252059e-01 -2.16565326e-01 3.84168983e-01 9.17237759e-01
-3.42666626e-01 5.94305873e-01 5.26899576e-01 -5.32753646e-01
-2.26759613e-01 -1.27243292e+00 -4.06955212e-01 -6.68993831e-01
-6.81063831e-01 3.50764513e-01 -1.18728586e-01 -7.85961270e-01
3.05497885e-01 1.08205664e+00 -1.52886498e+00 3.85696918e-01
9.44744170e-01 8.83267760e-01 -1.81956097e-01 4.86401260e-01
-2.87956476e-01 7.34265625e-01 -7.44685888e-01 1.86868578e-01
2.02712148e-01 3.10274251e-02 -8.25101316e-01 -3.86946589e-01
1.22435167e-01 -7.68151581e-01 -3.35161805e-01 6.66861117e-01
7.70677745e-01 1.94670454e-01 8.13330293e-01 -1.61145020e+00
-7.31480181e-01 3.31876576e-01 -7.88557410e-01 9.77414191e-01
3.00488323e-01 3.21698785e-01 5.10155916e-01 -1.02228045e+00
-1.91169372e-03 1.16068482e+00 8.69538844e-01 3.52199793e-01
-7.41828561e-01 -8.32451046e-01 3.76656324e-01 6.05248570e-01
-1.35639262e+00 5.09077497e-03 5.66829681e-01 -4.00674999e-01
1.36167324e+00 3.04333776e-01 2.93748707e-01 1.07527709e+00
1.90054029e-01 4.89771485e-01 4.10784066e-01 -8.77661780e-02
3.38738114e-01 -6.99413717e-01 2.24281013e-01 3.64854068e-01
4.88214284e-01 1.34392589e-01 -3.17473501e-01 -6.54844940e-02
5.07022798e-01 7.33093560e-01 -4.32030350e-01 5.66795349e-01
-1.04222822e+00 8.53224576e-01 1.25118244e+00 3.19255769e-01
-9.20874774e-01 1.73697457e-01 6.19528651e-01 2.48006284e-01
7.47455239e-01 5.13303399e-01 -3.46634567e-01 1.42757833e-01
-7.08470404e-01 1.96205691e-01 3.16690028e-01 3.21683586e-01
7.39350021e-01 5.73081672e-01 -2.62300193e-01 4.02274698e-01
-6.79171756e-02 7.80780315e-01 1.41586512e-01 -3.93223315e-01
4.25153881e-01 9.16604638e-01 -3.18003520e-02 -1.55103147e+00
-1.08932221e+00 -4.73607570e-01 -1.64972508e+00 3.24730664e-01
2.17713967e-01 -1.29268602e-01 -1.08722782e+00 1.45949781e+00
8.56241062e-02 3.51596296e-01 -8.90032724e-02 9.49500382e-01
6.76416755e-01 6.68918610e-01 5.10027468e-01 3.36548656e-01
1.13798654e+00 -3.12624991e-01 -7.18290806e-01 -3.56088728e-01
8.58375251e-01 -2.47503757e-01 1.20655417e+00 1.50044858e-01
-3.93951923e-01 -3.04148555e-01 -7.67248750e-01 2.38676816e-01
-7.84459949e-01 -3.13201547e-01 7.16180623e-01 9.49219167e-02
-1.01247036e+00 6.62485480e-01 -6.12623572e-01 -8.47383499e-01
6.95431590e-01 1.39580891e-01 -5.44780195e-01 -1.85559504e-02
-1.42055655e+00 8.88258517e-01 5.19597650e-01 3.70949864e-01
-8.66185427e-01 -9.44602668e-01 -8.28410566e-01 3.77045125e-01
-2.61980981e-01 -3.89481306e-01 5.23054540e-01 -5.61915755e-01
-1.95572302e-01 7.60883272e-01 3.14638168e-01 -8.11404884e-01
1.40207291e-01 -5.28067231e-01 -5.12611032e-01 8.81920084e-02
2.85006583e-01 5.40959537e-01 4.25835431e-01 -9.54096437e-01
-6.79404140e-01 -4.75688964e-01 1.27262017e-03 2.43182443e-02
-5.68557084e-01 3.50772752e-04 6.29483581e-01 -8.27967107e-01
1.71774223e-01 -3.16371709e-01 -4.17103887e-01 -9.83847156e-02
-5.67116618e-01 -9.61542130e-04 1.07255650e+00 -8.63385618e-01
1.36394370e+00 -2.28529644e+00 -5.15817702e-01 2.71546632e-01
3.54675144e-01 2.48773500e-01 -1.65195480e-01 5.90722024e-01
-5.20418346e-01 5.45431077e-01 -6.73376083e-01 1.60574913e-01
-2.42195159e-01 3.02659333e-01 -1.02361417e+00 5.47386467e-01
3.78981620e-01 7.14585900e-01 -8.45922112e-01 5.71533665e-02
2.01504424e-01 3.10337842e-01 -4.05615062e-01 3.70051622e-01
2.86293954e-01 4.57095504e-01 -3.72990072e-01 9.04537797e-01
9.61539209e-01 2.87498742e-01 -6.75172091e-01 1.45965502e-01
-2.23685935e-01 -2.16303408e-01 -8.71762097e-01 1.07385051e+00
-2.33028755e-01 7.19615161e-01 -2.59307802e-01 -1.23758948e+00
1.23508060e+00 2.11560190e-01 5.15097618e-01 -9.20247018e-01
-2.00968087e-01 3.00702482e-01 -2.99975604e-01 -5.72471917e-01
5.60923934e-01 9.90322381e-02 -1.65436625e-01 6.25082493e-01
-4.80802387e-01 1.33965567e-01 -2.03649610e-01 2.27539867e-01
1.34806609e+00 -5.50924480e-01 2.92787194e-01 -3.70162874e-01
1.46202788e-01 1.36434481e-01 3.61937106e-01 8.81092429e-01
-4.65734035e-01 1.10362256e+00 3.37605923e-01 -1.51050866e+00
-8.99867713e-01 -1.17341757e+00 3.31798871e-03 9.64191675e-01
1.58366673e-02 1.54491067e-01 -5.51140487e-01 -7.67248273e-01
2.25426201e-02 1.02093315e+00 -8.07874620e-01 -4.68897104e-01
-7.05402553e-01 -1.25317419e+00 1.02973580e+00 8.91802609e-01
1.01589894e+00 -1.56074214e+00 -7.08963811e-01 6.27340972e-02
-6.06335104e-01 -7.25681305e-01 1.44688651e-01 2.76990026e-01
-7.65495837e-01 -1.22945642e+00 -4.22328800e-01 -5.27117193e-01
8.31815004e-01 6.30045295e-01 1.00001729e+00 8.03507984e-01
-4.57807928e-01 2.60049373e-01 -5.23640871e-01 -6.41132414e-01
1.33618072e-01 1.38018340e-01 2.99349129e-01 -3.48598957e-01
4.56505746e-01 -7.54428387e-01 -9.45474684e-01 1.66371033e-01
-1.12341475e+00 -4.26084816e-01 2.52702683e-01 5.14359355e-01
3.45670134e-01 1.90911755e-01 8.60377133e-01 -3.04788381e-01
6.54939532e-01 -1.20899713e+00 2.53599375e-01 -1.29072025e-01
-5.87704539e-01 -1.33054808e-01 7.66204357e-01 3.97500023e-02
-9.06670928e-01 -2.76546508e-01 -3.72451901e-01 7.75707662e-02
-1.02951181e+00 5.33050656e-01 4.39448729e-02 3.43066275e-01
1.25610757e+00 3.87390070e-02 -4.36213166e-01 -4.54632491e-01
-1.16277477e-02 5.82819164e-01 8.78743291e-01 -3.89007032e-01
1.05146682e+00 6.67059541e-01 -4.85195443e-02 -1.09753084e+00
-7.31114388e-01 -6.00656211e-01 -6.67758882e-01 -2.16418698e-01
8.18881035e-01 -8.37668419e-01 -2.75950104e-01 8.18987727e-01
-1.47848296e+00 -4.93378043e-01 -5.52256584e-01 2.00737029e-01
-5.43541349e-02 2.60214418e-01 -2.14597598e-01 -6.65149271e-01
-7.59426594e-01 -3.67860258e-01 9.27142143e-01 3.42084765e-01
-3.05167198e-01 -9.69796658e-01 1.70119911e-01 -2.27484018e-01
1.04181087e+00 8.97312462e-01 1.05520594e+00 -1.03338587e+00
-1.73921824e-01 -1.92890912e-01 -5.47085404e-01 -1.11615852e-01
2.23521009e-01 -1.82197541e-02 -1.04854488e+00 -1.76126093e-01
-2.68626183e-01 -2.94820294e-02 1.12123764e+00 3.75456333e-01
1.57303715e+00 -5.16815364e-01 -2.01525941e-01 8.44843984e-01
1.09905350e+00 5.25791906e-02 1.20852208e+00 9.98332083e-01
6.55687690e-01 6.85377717e-01 5.99210262e-01 7.59355783e-01
5.28572738e-01 1.69261530e-01 1.26941502e+00 -8.12736332e-01
-3.70629989e-02 3.35078351e-02 2.56162405e-01 -5.81561290e-02
-2.58076906e-01 -6.09892309e-01 -1.58028007e+00 9.63189781e-01
-1.72199106e+00 -1.40676081e+00 -7.14704096e-01 2.08249593e+00
-4.38290136e-03 -1.65200412e-01 -3.65128294e-02 5.50469756e-01
7.53498137e-01 2.32385963e-01 -6.18483543e-01 -5.48167109e-01
-4.86302286e-01 3.02982837e-01 3.50080639e-01 2.85865605e-01
-1.38587308e+00 8.47129881e-01 5.90628958e+00 4.72308457e-01
-1.11008060e+00 -1.52237296e-01 9.20678258e-01 1.53579921e-01
-3.25897247e-01 -2.28825286e-01 -3.27329040e-01 3.06661129e-01
9.13451552e-01 7.40001425e-02 2.09308699e-01 8.47993851e-01
5.46576619e-01 -2.20776238e-02 -5.71166873e-01 6.42629743e-01
-2.42761791e-01 -1.02217162e+00 2.61409640e-01 -1.25126362e-01
4.32987571e-01 1.40151694e-01 1.59672219e-02 3.48727375e-01
6.02681898e-02 -1.32433784e+00 2.96810478e-01 5.38063705e-01
6.48002982e-01 -8.96458566e-01 1.08011687e+00 2.61886358e-01
-1.25008464e+00 -4.73072708e-01 -3.99031848e-01 -5.23367763e-01
6.05093576e-02 7.22366512e-01 -9.15449679e-01 3.51946473e-01
1.17298508e+00 5.92381418e-01 -5.96041799e-01 1.11785066e+00
-5.13574600e-01 8.27304423e-01 -4.19451773e-01 5.30663490e-01
1.66679129e-01 4.09505755e-01 6.32791579e-01 1.39844823e+00
5.99258304e-01 4.00188595e-01 9.19762999e-02 6.37687087e-01
2.84922570e-01 -6.27936423e-02 -1.05397880e+00 5.27354896e-01
3.42127383e-01 1.04610658e+00 -8.39296043e-01 1.22810602e-01
1.42059222e-01 1.00435317e+00 1.47892490e-01 5.69696724e-01
-8.33624899e-01 -6.13300562e-01 9.37184334e-01 3.65250289e-01
-3.20888698e-01 -2.09317356e-01 -6.04432583e-01 -9.04931426e-01
-2.24145159e-01 -5.18195629e-01 7.83494174e-01 -7.15378702e-01
-1.23347080e+00 7.64812350e-01 1.11363716e-01 -1.13481987e+00
-1.26592308e-01 -4.06311333e-01 -1.51437545e+00 7.07318902e-01
-1.69810283e+00 -9.99367893e-01 -1.05464172e+00 7.90075302e-01
5.54936171e-01 -1.40778534e-02 1.06553864e+00 3.37618798e-01
-7.77867734e-01 3.77032369e-01 -2.99813360e-01 2.47377962e-01
4.54350650e-01 -1.31710398e+00 9.18199897e-01 1.32230163e+00
-3.55642766e-01 1.67435497e-01 6.40601635e-01 -8.01996291e-01
-3.91246229e-01 -1.47562480e+00 8.15921307e-01 -9.39011276e-02
3.50709796e-01 3.64072882e-02 -1.48553145e+00 4.62741554e-01
2.69812029e-02 2.13920735e-02 4.59086239e-01 2.75909938e-02
-1.53331548e-01 -4.35554273e-02 -1.44331837e+00 4.33435112e-01
1.10524213e+00 -2.55180031e-01 -3.21418583e-01 4.41889495e-01
6.92354500e-01 -4.31334823e-02 -3.23982000e-01 6.77798212e-01
5.20950928e-02 -1.13667750e+00 1.04759991e+00 -8.17512214e-01
5.03819108e-01 -2.19073176e-01 -2.48153657e-01 -1.51881337e+00
-6.27429962e-01 2.82891840e-02 -3.32597420e-02 1.09527993e+00
1.97619244e-01 -7.52590597e-01 4.02900487e-01 6.11599028e-01
-5.45185566e-01 -5.75768054e-01 -8.01630616e-01 -6.18101239e-01
1.93407893e-01 -6.58288598e-01 1.06129098e+00 1.05992079e+00
-3.52373421e-01 -2.58782923e-01 -2.19158709e-01 7.09244132e-01
4.87708569e-01 -1.75717250e-01 5.99400759e-01 -1.38610709e+00
7.98397481e-01 -4.54415292e-01 -6.39893949e-01 -1.21182159e-01
-4.85762320e-02 -6.84903204e-01 -1.49795249e-01 -1.77618587e+00
1.86464582e-02 -5.17279148e-01 -5.59050918e-01 1.16097772e+00
-2.95704216e-01 2.23585501e-01 -2.89871186e-01 2.20981866e-01
-9.64738727e-02 7.24060893e-01 4.52345282e-01 -3.76997054e-01
-1.33484900e-01 -1.77036542e-02 -6.76391482e-01 8.66305709e-01
1.47724998e+00 -6.79838300e-01 -1.29918039e-01 -7.28306353e-01
2.51632452e-01 -5.47798991e-01 8.24545324e-01 -1.40412796e+00
8.73246044e-02 -3.79056960e-01 4.31194037e-01 -7.35482574e-01
-3.92403901e-01 -7.53421605e-01 -1.15289822e-01 6.97970569e-01
-1.82254966e-02 5.97553849e-01 6.00871325e-01 3.89444888e-01
-1.53904870e-01 9.74073410e-02 5.48951685e-01 1.48051143e-01
-1.09679091e+00 6.30645931e-01 -5.90709507e-01 5.60634173e-02
1.02095604e+00 -1.85315102e-01 -7.77578950e-01 -5.69371462e-01
-4.92024630e-01 4.56938088e-01 2.92656153e-01 5.20381629e-01
1.04731774e+00 -1.28648257e+00 -1.12572026e+00 3.49012613e-01
1.49117291e-01 1.92146823e-01 5.88625133e-01 6.88336432e-01
-9.40594077e-01 -9.08484980e-02 -6.32835448e-01 -4.17553335e-01
-6.48719490e-01 4.25774306e-01 3.73347551e-01 -2.38972351e-01
-8.10463607e-01 6.32695138e-01 3.80117655e-01 -5.04214466e-01
1.73435986e-01 -2.96310842e-01 -4.33556676e-01 1.48932666e-01
8.38348567e-01 7.45987535e-01 3.17296505e-01 -6.69096291e-01
-6.83526754e-01 2.67496288e-01 2.14488611e-01 2.99638718e-01
1.57141411e+00 -1.13067098e-01 -4.07486446e-02 -1.80633679e-01
7.25639462e-01 -4.33804154e-01 -9.47230458e-01 1.90547571e-01
1.56670790e-02 -3.57156396e-01 2.04273202e-02 -8.51443768e-01
-1.40560234e+00 1.24785471e+00 1.05163538e+00 4.35730577e-01
1.49361205e+00 -2.56835043e-01 9.60957348e-01 5.27532995e-01
1.54868588e-01 -7.78832853e-01 2.53139824e-01 7.24612713e-01
1.46298933e+00 -1.45203161e+00 -2.50577152e-01 2.29430050e-01
-5.35220742e-01 1.33573937e+00 9.11574781e-01 -1.28758803e-01
6.25377774e-01 1.24195039e-01 1.83932915e-01 -7.66476870e-01
-1.22676358e-01 -2.90753543e-01 5.59533155e-03 1.04291975e+00
1.35026798e-01 1.86258212e-01 1.81455374e-01 5.69564641e-01
-6.96648061e-02 -6.26549721e-01 4.96370524e-01 6.46080613e-01
-8.75580728e-01 -3.61661345e-01 -6.15245461e-01 4.59483743e-01
-3.63516718e-01 -3.42713237e-01 -4.26899374e-01 8.21787238e-01
1.18936770e-01 1.13938725e+00 4.27281141e-01 -5.91615260e-01
5.97825229e-01 -1.81787640e-01 -4.76286381e-01 -3.49881291e-01
-6.96659446e-01 -9.36175942e-01 -3.02221000e-01 -7.76760161e-01
5.33259995e-02 -3.98665279e-01 -1.70257747e+00 -7.74617434e-01
-3.65759432e-02 -2.68054623e-02 5.34983754e-01 9.46681559e-01
5.54527164e-01 4.64937508e-01 7.90423214e-01 -9.57439363e-01
-6.97917584e-03 -9.98828530e-01 -5.28974891e-01 6.19356453e-01
5.98758638e-01 -6.41349256e-01 -6.95351303e-01 -4.78569925e-01]
|
[9.5518217086792, -1.3130160570144653]
|
2ac0ba19-30bc-4d74-a9ff-40cd590b25a5
|
virtuously-safe-reinforcement-learning
|
1805.11447
| null |
http://arxiv.org/abs/1805.11447v1
|
http://arxiv.org/pdf/1805.11447v1.pdf
|
Virtuously Safe Reinforcement Learning
|
We show that when a third party, the adversary, steps into the two-party
setting (agent and operator) of safely interruptible reinforcement learning, a
trade-off has to be made between the probability of following the optimal
policy in the limit, and the probability of escaping a dangerous situation
created by the adversary. So far, the work on safely interruptible agents has
assumed a perfect perception of the agent about its environment (no adversary),
and therefore implicitly set the second probability to zero, by explicitly
seeking a value of one for the first probability. We show that (1) agents can
be made both interruptible and adversary-resilient, and (2) the
interruptibility can be made safe in the sense that the agent itself will not
seek to avoid it. We also solve the problem that arises when the agent does not
go completely greedy, i.e. issues with safe exploration in the limit.
Resilience to perturbed perception, safe exploration in the limit, and safe
interruptibility are the three pillars of what we call \emph{virtuously safe
reinforcement learning}.
|
['Alexandre Maurer', 'Rachid Guerraoui', 'Henrik Aslund', 'El Mahdi El Mhamdi']
|
2018-05-29
| null | null | null | null |
['safe-exploration']
|
['robots']
|
[-3.38540152e-02 7.19392538e-01 -1.05744891e-01 1.65432602e-01
-5.26167870e-01 -1.16618705e+00 5.42429626e-01 6.28397986e-02
-9.72127080e-01 1.02259672e+00 -4.70533706e-02 -6.62482381e-01
-2.84274459e-01 -8.22101712e-01 -7.30655789e-01 -1.00089061e+00
-4.92991596e-01 5.37001729e-01 1.77618146e-01 -2.84137547e-01
1.26902431e-01 4.07991856e-01 -1.31017268e+00 -5.70238650e-01
6.55754864e-01 5.53296208e-01 -2.44813904e-01 9.17239904e-01
4.37961757e-01 1.33103180e+00 -6.59620166e-01 -2.37090066e-02
5.15939176e-01 -4.02256280e-01 -1.04951727e+00 6.49833307e-02
-4.52454239e-01 -8.43244970e-01 -2.43250504e-01 1.37073457e+00
1.65661588e-01 3.57701510e-01 1.34614900e-01 -1.52817547e+00
-2.36402273e-01 9.62009490e-01 -4.07772630e-01 -4.08826731e-02
5.19365549e-01 7.59571195e-01 5.62139630e-01 3.60007107e-01
6.12055600e-01 1.23513210e+00 4.61891331e-02 7.80010223e-01
-1.39329684e+00 -4.30332154e-01 4.72384423e-01 -2.65302658e-01
-1.21498311e+00 -3.70744109e-01 4.12513733e-01 -1.95647135e-01
7.12734997e-01 6.39083683e-01 5.24487913e-01 1.14830112e+00
3.43892872e-01 2.53869712e-01 1.65491557e+00 -3.83268803e-01
8.74461591e-01 1.95222020e-01 -5.66671081e-02 2.51844406e-01
4.39926803e-01 1.10510015e+00 -1.10787824e-01 -6.86185896e-01
6.36446238e-01 -2.51226425e-01 -4.10455614e-01 -4.60550517e-01
-1.17047405e+00 5.54362297e-01 -3.37873287e-02 7.49921650e-02
-6.53626919e-01 2.97190607e-01 4.66234237e-01 9.22250867e-01
-4.65488404e-01 7.52961576e-01 -1.72511250e-01 -4.69749838e-01
-1.49692416e-01 5.83001196e-01 1.16298378e+00 5.29193819e-01
3.92488152e-01 2.32814118e-01 2.27346882e-01 -1.74121335e-01
1.66860327e-01 4.94311780e-01 1.81500643e-01 -1.49342453e+00
1.72116071e-01 2.61920001e-02 1.07851338e+00 -3.13240260e-01
-3.51807535e-01 -1.66212097e-01 -1.06684819e-01 1.30490112e+00
5.59072852e-01 -7.40695298e-01 -4.24818248e-01 2.46577287e+00
5.13243735e-01 -1.96030930e-01 5.93396008e-01 9.56200480e-01
-2.54914194e-01 6.70323670e-01 1.76329017e-01 -7.00339198e-01
1.10889101e+00 -3.80746394e-01 -6.86453700e-01 -3.01717222e-01
4.14807498e-01 -3.18746001e-01 1.00688589e+00 5.35426497e-01
-1.36585343e+00 6.72504380e-02 -1.45505631e+00 5.21170437e-01
1.32605344e-01 -1.16103375e+00 5.31675518e-01 5.52247703e-01
-1.01124799e+00 7.37459183e-01 -1.12650871e+00 -1.18570164e-01
-1.74323827e-01 4.81721669e-01 -4.54342037e-01 4.73685116e-01
-1.02903056e+00 1.01989603e+00 4.52815175e-01 -1.47688583e-01
-1.64617825e+00 7.79443383e-02 -5.54525435e-01 2.17352167e-01
1.07581067e+00 -3.96554440e-01 1.50705361e+00 -1.29185379e+00
-1.59282589e+00 5.42729259e-01 4.96194631e-01 -5.38325667e-01
8.99434149e-01 6.02563880e-02 -3.21949311e-02 9.43963230e-02
1.80310816e-01 3.86463284e-01 6.42889261e-01 -1.43414855e+00
-6.98288858e-01 -5.36455929e-01 7.44485199e-01 6.05549335e-01
-1.82894319e-02 -1.17262051e-01 5.81328332e-01 -1.90217376e-01
-1.07393332e-01 -1.44787371e+00 -4.57837224e-01 -3.43580484e-01
-2.10250005e-01 -2.37529427e-01 3.90658349e-01 1.31929755e-01
5.91399610e-01 -2.35066319e+00 -1.56395242e-01 1.87395707e-01
1.69336364e-01 -4.42728661e-02 -2.03820795e-01 4.83974159e-01
-8.54802430e-02 2.70921320e-01 7.41127878e-02 -1.81177452e-01
3.17240834e-01 5.52040458e-01 -7.41167605e-01 7.89318383e-01
-5.11025250e-01 2.28124544e-01 -1.10741496e+00 -1.61471754e-01
-2.24942476e-01 -5.91750927e-02 -3.79651666e-01 5.29024005e-01
-1.19961306e-01 4.71982241e-01 -8.23640585e-01 7.92247579e-02
4.97351795e-01 4.02638853e-01 5.35271883e-01 7.44882166e-01
-5.90257108e-01 4.03458267e-01 -1.41452622e+00 8.29591453e-01
-6.15710765e-02 -2.68862909e-03 6.53012395e-01 -3.62842321e-01
1.45992041e-01 6.81598425e-01 1.31712586e-01 -5.31472385e-01
3.03599685e-01 2.51123995e-01 3.04825515e-01 -3.62562925e-01
2.60701925e-01 -5.23445070e-01 -1.59102395e-01 1.04296148e+00
-3.51158649e-01 -1.45870950e-02 -2.30458468e-01 1.89276606e-01
1.16061819e+00 2.36799359e-01 1.30800724e-01 -5.62911630e-01
1.78417400e-01 1.69975609e-02 8.21513593e-01 1.36887383e+00
-6.73531055e-01 -2.18701839e-01 9.53604579e-01 -4.91591513e-01
-7.56291747e-01 -1.13119352e+00 1.98296353e-01 9.75564241e-01
5.16631305e-01 5.75968362e-02 -9.68974113e-01 -6.32533908e-01
-5.91770560e-02 1.05849469e+00 -8.34994614e-01 -2.57567674e-01
-5.49077094e-01 -1.73950896e-01 4.95954514e-01 2.62892634e-01
5.64816415e-01 -1.21902621e+00 -1.41437531e+00 1.51680335e-01
4.26010750e-02 -5.56435347e-01 -5.77841997e-01 6.64812028e-01
-5.69944859e-01 -9.72937822e-01 3.72553885e-01 -1.58950478e-01
6.01582825e-01 2.79703557e-01 6.33288205e-01 3.54854196e-01
4.85141724e-01 6.54717505e-01 -2.16227267e-02 -4.90882665e-01
-6.40106916e-01 -4.87757802e-01 5.37735939e-01 -4.51453298e-01
6.87401518e-02 -6.63645327e-01 -4.92875308e-01 1.39223948e-01
-9.47154045e-01 -4.50404644e-01 -1.38988703e-01 6.63088620e-01
2.75334835e-01 4.02362883e-01 4.94809687e-01 -5.59415400e-01
7.35143423e-01 -3.81061167e-01 -1.14887834e+00 3.37898880e-02
-5.11381030e-01 1.96422547e-01 8.27393830e-01 -6.26115203e-01
-8.88776779e-01 -1.25765055e-01 1.41781494e-01 -8.57565030e-02
-2.37161919e-01 -4.42311913e-02 -3.81444216e-01 -1.71085969e-01
7.45300412e-01 9.35270935e-02 4.35776234e-01 -2.00033858e-01
1.70222685e-01 4.14105922e-01 4.27147746e-01 -1.04964447e+00
8.35556567e-01 5.42599797e-01 9.75750014e-02 -4.88289148e-01
-4.64016140e-01 4.47291046e-01 1.52419671e-01 -1.14826605e-01
5.06468832e-01 -6.82780266e-01 -1.69025540e+00 3.15236211e-01
-6.47162139e-01 -6.85277581e-01 -7.68098235e-01 3.61348003e-01
-1.00920999e+00 4.25846040e-01 -3.60849887e-01 -1.35373700e+00
1.45378590e-01 -1.39963269e+00 1.33941233e-01 3.96239191e-01
-2.50325173e-01 -5.62222183e-01 1.41078085e-01 6.26769960e-02
4.61185515e-01 4.96111363e-01 6.28038526e-01 -8.60203683e-01
-8.06099951e-01 -6.74132258e-02 7.10042536e-01 1.75488651e-01
1.79595593e-02 -2.68502027e-01 -8.90274405e-01 -6.41263127e-01
8.73965144e-01 -7.05226839e-01 1.25724986e-01 7.05926567e-02
6.69173717e-01 -1.24323261e+00 -2.10853294e-01 2.86872625e-01
1.14384496e+00 8.43233645e-01 3.29030931e-01 8.55821550e-01
9.87083539e-02 6.83816195e-01 7.37349153e-01 5.01220405e-01
4.06253517e-01 3.44131291e-01 9.57614839e-01 2.75606632e-01
7.57952273e-01 -3.93342197e-01 6.29897118e-01 -2.65987039e-01
-2.34166533e-02 -2.58446466e-02 -4.83355582e-01 4.19759244e-01
-1.94819391e+00 -1.31554627e+00 6.73926473e-01 2.75397134e+00
9.41939414e-01 6.49400592e-01 4.04531866e-01 2.65448764e-02
3.52311432e-01 6.01373613e-02 -8.03119659e-01 -9.84707296e-01
2.06606194e-01 -5.03741384e-01 6.76550984e-01 1.19237518e+00
-8.54969144e-01 6.31851912e-01 6.91469002e+00 3.80670458e-01
-8.96123946e-01 2.09408790e-01 5.86341977e-01 -3.38911593e-01
-4.70369428e-01 5.83080947e-01 -3.62957537e-01 3.58210981e-01
9.66338992e-01 -4.96880442e-01 1.11587942e+00 1.02943254e+00
1.94814399e-01 -4.21153843e-01 -1.34416401e+00 9.80906785e-02
-5.54196119e-01 -6.68689966e-01 -4.85034376e-01 3.51268649e-01
2.71865040e-01 -5.82595803e-02 -8.16674158e-03 4.14111495e-01
1.02544701e+00 -1.07599747e+00 1.29120755e+00 8.88154730e-02
3.54950577e-01 -1.26392090e+00 4.93216097e-01 1.06674778e+00
-4.90588099e-01 -5.14560699e-01 1.06723337e-02 -5.62414467e-01
-2.51118749e-01 -3.14495772e-01 -5.69610000e-01 -2.21018735e-02
5.17315626e-01 -6.28807664e-01 2.72132009e-01 3.58542055e-01
-2.70247489e-01 3.20019722e-01 -6.29489243e-01 3.45448256e-02
5.63599765e-01 -5.00169337e-01 8.82794619e-01 3.69315714e-01
-2.28076160e-01 5.65282404e-01 6.96885407e-01 7.97980547e-01
3.45794827e-01 -5.76150119e-01 -8.55949104e-01 1.21798195e-01
7.17084527e-01 8.91572654e-01 -5.85004270e-01 -8.55349079e-02
3.99848036e-02 4.63627666e-01 2.02462450e-01 4.32706565e-01
-6.70904100e-01 -2.82242507e-01 8.55713725e-01 -8.47324505e-02
-1.19429991e-01 -1.00613818e-01 -3.93142402e-02 -6.11897171e-01
-5.87438494e-02 -1.25836253e+00 5.13496220e-01 -3.94666582e-01
-8.76637459e-01 7.02349246e-01 -9.71243009e-02 -8.33980203e-01
-5.54171145e-01 -8.19875672e-02 -5.50103307e-01 5.96631587e-01
-1.22556102e+00 -5.53393066e-01 5.41014254e-01 6.03870988e-01
4.69941720e-02 2.39771158e-01 9.45003688e-01 -5.76331615e-01
-3.36840123e-01 3.73039961e-01 -2.86795702e-02 -2.86781013e-01
9.42874402e-02 -1.37727189e+00 -3.47795077e-02 8.73114705e-01
-3.34328264e-01 8.58077407e-01 1.21953118e+00 -4.77739155e-01
-1.75678265e+00 -6.34176672e-01 3.91257852e-01 -6.07772946e-01
7.53595293e-01 -2.42445737e-01 -8.16905022e-01 1.15388191e+00
2.39953429e-01 -1.72721848e-01 3.50217789e-01 -4.18346049e-03
-2.42664844e-01 -4.93309870e-02 -1.38910246e+00 1.09234893e+00
7.23644733e-01 -5.13243675e-01 -7.73541570e-01 3.38145465e-01
9.50952709e-01 -3.57559204e-01 -5.50767601e-01 8.13943446e-02
4.38608587e-01 -9.99425471e-01 5.77314496e-01 -8.20066452e-01
-5.35713255e-01 -6.52261734e-01 -1.57982424e-01 -1.12923312e+00
-2.53241926e-01 -1.55480719e+00 1.13843530e-01 7.07227051e-01
-2.81976517e-02 -9.07592952e-01 5.49366057e-01 1.05984926e+00
1.57003924e-01 -2.58668065e-01 -1.47346568e+00 -1.10687351e+00
4.69857603e-01 -1.16208322e-01 7.74156094e-01 7.02148139e-01
6.44220412e-01 9.48860659e-04 -4.01595116e-01 5.39606512e-01
6.81585789e-01 1.40656605e-01 5.75190663e-01 -5.67312658e-01
-6.68518245e-01 -2.59896129e-01 1.94732979e-01 -6.74152017e-01
8.36122781e-02 -1.33800939e-01 4.46615815e-01 -7.64294386e-01
1.68673009e-01 -5.69039166e-01 -3.17222804e-01 7.55027115e-01
-5.09220921e-02 -5.64194441e-01 5.13552964e-01 2.99214900e-01
-7.23710418e-01 4.48368132e-01 1.13746178e+00 2.30236351e-01
-4.32284325e-01 1.14037752e-01 -1.09355545e+00 9.43665683e-01
7.41354525e-01 -5.84233701e-01 -6.44037485e-01 -1.39763448e-02
2.72188455e-01 7.67434776e-01 3.65598112e-01 -5.43348312e-01
2.33964384e-01 -9.11792397e-01 -3.10145795e-01 1.61011055e-01
1.49008930e-01 -1.06855643e+00 4.94683444e-01 1.02732611e+00
-6.44743979e-01 2.63063848e-01 6.00386150e-02 5.53679883e-01
4.25505698e-01 -3.59224170e-01 8.97241354e-01 -2.59966910e-01
-4.92454208e-02 -1.18476771e-01 -9.39803720e-01 1.39347361e-02
1.42785943e+00 -4.61463742e-02 -5.58631778e-01 -5.76280117e-01
-8.30835760e-01 6.91054523e-01 9.98704433e-01 9.78964344e-02
1.42228007e-01 -9.09366190e-01 -3.74043971e-01 9.68761891e-02
-2.65728176e-01 -2.29430199e-01 2.03185141e-01 5.47235787e-01
-7.77005479e-02 2.22795848e-02 -2.83972561e-01 -5.25129102e-02
-1.09379101e+00 1.18939626e+00 7.69930243e-01 -7.31311589e-02
-7.40985572e-01 4.38281238e-01 3.55026871e-01 -2.11350452e-02
5.03507018e-01 1.48495197e-01 9.88060012e-02 -4.03204262e-01
7.59575069e-01 1.93792984e-01 -3.23355138e-01 -2.66965568e-01
-4.63289887e-01 -4.17432189e-02 -2.52607644e-01 -8.31852853e-01
1.05434752e+00 -2.53524721e-01 -2.02477843e-01 2.91272491e-01
4.14893240e-01 2.56240785e-01 -1.84955823e+00 1.50397047e-01
-2.39217669e-01 -4.78413224e-01 -3.92442159e-02 -8.51690590e-01
-4.01286751e-01 2.62323797e-01 4.54803050e-01 9.67091024e-01
9.92212832e-01 -4.07131046e-01 1.96962625e-01 6.62519515e-01
9.29188132e-01 -1.18334329e+00 -2.37513945e-01 3.71079117e-01
7.28376150e-01 -8.18623662e-01 -2.22360685e-01 2.83703446e-01
-9.27018404e-01 6.80560231e-01 6.57180667e-01 -2.31381342e-01
1.42113522e-01 5.83974719e-01 4.92565408e-02 1.41478078e-02
-1.04600453e+00 8.06972682e-02 -8.90918314e-01 7.60390460e-01
-5.11791468e-01 3.96928459e-01 -1.14372589e-01 4.25275296e-01
-1.95251882e-01 -7.26292282e-02 1.13554025e+00 1.38309062e+00
-8.78405213e-01 -1.05355453e+00 -7.82898068e-01 -2.05766708e-01
-6.93703771e-01 4.08228606e-01 -2.33361334e-01 1.01961422e+00
6.88517764e-02 1.26339746e+00 -5.26522137e-02 8.85318369e-02
2.89513052e-01 -1.14643410e-01 2.32447475e-01 -2.82585651e-01
-6.21224821e-01 1.22427583e-01 4.58566360e-02 -8.36283982e-01
-4.84229662e-02 -7.10779846e-01 -1.40500307e+00 -5.74152350e-01
-5.53932190e-02 4.99177784e-01 2.12684005e-01 8.64327669e-01
4.59218808e-02 5.12440018e-02 1.19086802e+00 -3.68465185e-01
-1.56326473e+00 -3.35792124e-01 -7.88062334e-01 8.64771530e-02
1.05827689e+00 -4.60568696e-01 -1.00880635e+00 -6.37855351e-01]
|
[4.315341949462891, 2.1090633869171143]
|
8ed7ec62-71f8-4cbf-81b8-23b0db191ec4
|
caulking-the-leakage-effect-in-meeg-source
|
1810.00786
| null |
http://arxiv.org/abs/1810.00786v1
|
http://arxiv.org/pdf/1810.00786v1.pdf
|
Caulking the Leakage Effect in MEEG Source Connectivity Analysis
|
Simplistic estimation of neural connectivity in MEEG sensor space is
impossible due to volume conduction. The only viable alternative is to carry
out connectivity estimation in source space. Among the neuroscience community
this is claimed to be impossible or misleading due to Leakage: linear mixing of
the reconstructed sources. To address this problematic we propose a novel
solution method that caulks the Leakage in MEEG source activity and
connectivity estimates: BC-VARETA. It is based on a joint estimation of source
activity and connectivity in the frequency domain representation of MEEG time
series. To achieve this, we go beyond current methods that assume a fixed
gaussian graphical model for source connectivity. In contrast we estimate this
graphical model in a Bayesian framework by placing priors on it, which allows
for highly optimized computations of the connectivity, via a new procedure
based on the local quadratic approximation under quite general prior models. A
further contribution of this paper is the rigorous definition of leakage via
the Spatial Dispersion Measure and Earth Movers Distance based on the geodesic
distances over the cortical manifold. Both measures are extended for the first
time to quantify Connectivity Leakage by defining them on the cartesian product
of cortical manifolds. Using these measures, we show that BC-VARETA outperforms
most state of the art inverse solvers by several orders of magnitude.
|
['Pedro A. Valdes-Sosa', 'Maria Luisa Bringas-Vega', 'Jorge Bosch-Bayard', 'Pedro A. Valdes-Hernandez', 'Eduardo Martinez-Montes', 'Eduardo Gonzalez-Moreira', 'Deirel Paz-Linares']
|
2018-09-28
| null | null | null | null |
['connectivity-estimation']
|
['graphs']
|
[ 6.25788271e-02 2.12887883e-01 6.58801556e-01 6.46976801e-03
-4.39792186e-01 -5.62000215e-01 5.24089456e-01 7.35445768e-02
-5.80843866e-01 9.17689919e-01 4.57154736e-02 5.20853139e-02
-6.59021318e-01 -6.73651040e-01 -7.96536863e-01 -8.10783207e-01
-4.05167699e-01 9.94318351e-02 6.69604167e-02 -3.35545726e-02
3.33829165e-01 5.67438006e-01 -1.02901244e+00 -4.58319873e-01
9.42228913e-01 9.95778859e-01 6.91524968e-02 2.92414546e-01
1.09375820e-01 3.30478460e-01 -5.38007021e-01 -2.58727819e-01
5.09097353e-02 -6.09053493e-01 -6.91122770e-01 -3.94183367e-01
1.51001245e-01 -1.28748596e-01 -8.11629444e-02 1.26420569e+00
3.55386227e-01 4.31754999e-02 9.75532949e-01 -1.13571310e+00
-7.91394934e-02 6.10309899e-01 -4.90400881e-01 5.04559457e-01
8.31936672e-02 -4.08178478e-01 6.64533734e-01 -8.69796813e-01
6.84388816e-01 9.06222284e-01 7.29966164e-01 7.69734457e-02
-1.81820977e+00 -5.16844273e-01 -7.52465725e-02 6.48926124e-02
-1.74272263e+00 -3.33099782e-01 8.78990471e-01 -7.13029921e-01
7.98936903e-01 2.68248618e-01 9.29077148e-01 9.14590478e-01
3.95245880e-01 3.03170979e-02 1.38375127e+00 -2.98667341e-01
4.53687817e-01 2.02270448e-01 1.84156999e-01 3.87914389e-01
4.99283463e-01 7.80881988e-03 -5.46294570e-01 -2.79380232e-01
8.24018359e-01 -5.19645333e-01 -6.66005433e-01 -5.49450696e-01
-1.19685888e+00 8.69452298e-01 4.90049362e-01 7.50412941e-01
-3.18147629e-01 3.08141232e-01 6.44429550e-02 6.90012006e-03
6.45864844e-01 4.72592264e-01 -1.06160447e-01 3.58059853e-02
-1.22817659e+00 3.50416839e-01 1.05820608e+00 5.47670960e-01
4.39016521e-01 -1.08759962e-01 2.43759543e-01 4.46314037e-01
7.99675584e-01 5.12446344e-01 3.47124375e-02 -9.38659191e-01
2.04136372e-01 9.58265066e-02 -1.94781676e-01 -1.37798989e+00
-5.99265575e-01 -8.27692211e-01 -9.37809646e-01 3.39263976e-01
7.18693197e-01 -2.10921884e-01 -3.52511138e-01 2.15993381e+00
1.60679370e-01 2.20998198e-01 -3.94544661e-01 7.56525218e-01
3.14961702e-01 5.39505303e-01 -7.76808038e-02 -4.37068105e-01
9.85653043e-01 -1.06587440e-01 -8.65366697e-01 1.07369497e-01
3.53004932e-01 -3.43575627e-01 3.16083372e-01 7.14768112e-01
-1.18993664e+00 1.31786779e-01 -1.26368940e+00 2.00811878e-01
-4.17537451e-01 -1.01643525e-01 2.72214800e-01 8.00932586e-01
-1.32388270e+00 9.26121175e-01 -1.08911037e+00 -1.02414526e-01
4.54240888e-01 2.47413069e-01 -3.73570800e-01 4.45102721e-01
-1.03672385e+00 1.29821634e+00 -6.21899143e-02 2.98844278e-01
-4.51382041e-01 -8.81907940e-01 -6.71622336e-01 -9.65986773e-03
-2.47815520e-01 -5.24409771e-01 3.93333137e-01 -7.38669395e-01
-1.54362381e+00 4.23629731e-01 1.35220557e-01 -4.66846913e-01
7.20236957e-01 1.18102357e-01 8.40188265e-02 3.01174313e-01
-4.15806361e-02 5.32529235e-01 8.01253498e-01 -1.15796578e+00
5.01747072e-01 -5.77626228e-01 -3.37829947e-01 -1.21958286e-01
-2.20389560e-01 -2.55191982e-01 7.63769820e-02 -6.79949999e-01
5.01294136e-01 -6.30913436e-01 -7.93814212e-02 1.85073361e-01
-5.53724349e-01 3.50928515e-01 2.02844203e-01 -1.04899669e+00
8.96740735e-01 -1.96320271e+00 5.90212286e-01 6.99374735e-01
4.73683029e-01 -2.85748869e-01 1.41649544e-01 2.84170508e-01
-2.47665599e-01 7.21532106e-02 -7.56686628e-01 -3.04280728e-01
-9.84689891e-02 -2.72373766e-01 -2.88070261e-01 1.25884116e+00
2.84410864e-02 5.77405453e-01 -6.82475924e-01 -3.65406394e-01
1.07775822e-01 1.13819587e+00 -6.04721665e-01 -2.95205891e-01
3.62590849e-01 7.36533999e-01 -2.29421318e-01 -1.11904696e-01
1.02850318e+00 -7.33968914e-02 8.48902240e-02 -5.23482561e-01
-2.82370508e-01 1.42698199e-01 -1.34774888e+00 2.02000427e+00
-3.58585209e-01 6.60495698e-01 3.86340380e-01 -1.34410727e+00
7.45905817e-01 2.42424861e-01 7.30643630e-01 -3.91227454e-01
4.08612221e-01 3.95762950e-01 1.40553594e-01 -4.01372323e-03
-1.93848804e-01 -2.44190559e-01 2.50214726e-01 2.47380793e-01
4.42299277e-01 -2.31873408e-01 -1.03474803e-01 2.71125883e-01
1.15944934e+00 3.75320911e-01 1.15600646e-01 -1.00945127e+00
3.84286076e-01 -4.55492169e-01 1.30441636e-01 3.43804061e-01
1.19605988e-01 5.37771583e-01 8.00056934e-01 3.51012558e-01
-5.84951520e-01 -1.44087863e+00 -7.33584642e-01 -8.03663433e-02
-7.31253922e-02 -2.78158933e-01 -1.25291741e+00 -9.48884040e-02
-2.47794330e-01 5.71491361e-01 -7.30075479e-01 -1.87503025e-01
-1.66219115e-01 -1.11370635e+00 5.32924354e-01 1.54131809e-02
4.11614269e-01 -3.48906100e-01 -9.14642334e-01 2.12339133e-01
-2.03072786e-01 -9.49880600e-01 -9.58217494e-03 3.59498590e-01
-1.12278199e+00 -8.17374468e-01 -1.00489676e+00 -1.53225631e-01
5.75814009e-01 -3.86876374e-01 6.99060380e-01 -2.88887411e-01
-4.83552605e-01 4.86075729e-01 2.91122012e-02 -2.34409571e-01
1.96001325e-02 -1.58702359e-01 1.41046904e-02 1.05658934e-01
-9.19623375e-02 -1.20847952e+00 -6.11579478e-01 -1.13985268e-02
-7.89533913e-01 -1.72937602e-01 2.39497006e-01 4.26400274e-01
4.94059563e-01 -6.44118860e-02 7.03479290e-01 -4.36211050e-01
4.88524944e-01 -8.93985927e-01 -8.80103409e-01 -1.04020149e-01
-5.94094634e-01 2.48014227e-01 1.78821236e-01 -2.25886747e-01
-7.21669614e-01 6.64288402e-02 -1.00556672e-01 -9.07288566e-02
3.40144575e-01 4.57177579e-01 -2.06896693e-01 -5.75729489e-01
7.46674299e-01 4.64216731e-02 -2.91909613e-02 -4.36305821e-01
4.41700071e-01 1.49323389e-01 3.63642633e-01 -2.81177163e-01
5.78085184e-01 6.99070573e-01 4.98761445e-01 -9.96404231e-01
-2.40023822e-01 -1.33309901e-01 -7.01739550e-01 -2.94412076e-01
8.71936202e-01 -6.49909198e-01 -5.63475251e-01 4.94185805e-01
-1.48625791e+00 -2.32823700e-01 -7.44721740e-02 7.97552884e-01
-8.08410108e-01 4.29638207e-01 -3.42979878e-01 -9.19431388e-01
-1.25314549e-01 -1.10189056e+00 7.61402369e-01 -2.69964665e-01
-1.86808199e-01 -1.32265913e+00 3.81467819e-01 -2.68242747e-01
6.45575225e-01 5.90706587e-01 5.75663745e-01 -1.97426260e-01
-5.12132704e-01 3.60185206e-02 -1.87384889e-01 3.54263008e-01
-2.15913430e-01 -2.03109294e-01 -1.03538179e+00 -4.96103875e-02
5.35656750e-01 4.31146473e-01 8.55153382e-01 9.21372116e-01
9.38202858e-01 -1.04326345e-01 -5.32843471e-01 7.35901237e-01
1.85516810e+00 -2.49333352e-01 6.83850706e-01 -1.44660473e-01
3.98583502e-01 8.37938607e-01 -1.46719784e-01 3.72383386e-01
1.51831970e-01 7.06348896e-01 6.01283848e-01 3.10359001e-01
-1.54891595e-01 1.19325645e-01 1.40627876e-01 8.52429569e-01
-2.56687462e-01 2.35305838e-02 -8.27532649e-01 6.31412566e-01
-1.57945776e+00 -7.04936862e-01 -4.89992887e-01 2.47278786e+00
7.37739027e-01 2.12038495e-02 3.21130306e-02 3.31592143e-01
4.55925733e-01 -3.78699332e-01 -1.79785848e-01 -4.68522459e-02
-1.20554417e-01 4.17871535e-01 7.66388059e-01 8.67537618e-01
-7.27600217e-01 9.58081186e-02 6.46614790e+00 5.86526394e-01
-9.74594295e-01 6.35605395e-01 3.66313756e-01 -1.45621255e-01
-3.38816792e-01 2.43968572e-02 -4.69173193e-01 6.83771551e-01
1.18932927e+00 -9.71549675e-02 5.97748280e-01 2.62828350e-01
6.57700524e-02 -5.69190681e-01 -9.23796475e-01 1.10897458e+00
1.66709691e-01 -1.19638360e+00 -5.01835108e-01 5.16260624e-01
3.96445304e-01 1.43273726e-01 -8.14084113e-02 -5.27317584e-01
-3.16712230e-01 -1.11054206e+00 9.14773464e-01 1.01897264e+00
5.21997750e-01 -5.32471180e-01 3.65339309e-01 3.11314225e-01
-9.76146936e-01 3.05746555e-01 -1.80623621e-01 7.42025077e-02
3.39899629e-01 1.15376413e+00 -1.84624642e-01 3.66667926e-01
5.73225200e-01 7.28834271e-01 -2.72629350e-01 1.38852906e+00
7.65415877e-02 5.04751265e-01 -8.53133857e-01 1.26803637e-01
-2.11766303e-01 -5.83937228e-01 9.64968801e-01 1.04577184e+00
7.45621741e-01 -1.54269487e-01 -6.63801908e-01 1.79542565e+00
3.37513685e-01 1.57206804e-01 -7.10981548e-01 2.94222265e-01
2.05501780e-01 1.36304510e+00 -9.54298675e-01 2.85919487e-01
-1.40626803e-01 8.89429033e-01 2.77099699e-01 5.12636125e-01
-7.57594228e-01 -5.79705477e-01 3.55590016e-01 3.41575742e-01
-4.60045505e-03 -5.26412129e-01 -4.40503687e-01 -1.09271002e+00
2.81192541e-01 -1.53294653e-01 -2.25840092e-01 -6.06779337e-01
-1.03967690e+00 6.30779743e-01 4.84002024e-01 -8.57757390e-01
-2.36135751e-01 -6.61452472e-01 -4.28125322e-01 1.14541054e+00
-1.38607872e+00 -6.60313606e-01 3.54660153e-02 7.27497816e-01
-3.29403609e-01 3.28033775e-01 7.40633607e-01 4.38077211e-01
-2.32526526e-01 2.32643694e-01 1.75488442e-02 -2.27600336e-01
2.56689876e-01 -1.30062068e+00 -3.96036170e-02 8.95881772e-01
8.70657936e-02 7.58439720e-01 9.68967021e-01 -5.31764030e-01
-1.27933812e+00 -6.67152166e-01 7.17480302e-01 -3.66708994e-01
7.61190295e-01 -5.87381363e-01 -8.63407016e-01 5.03326297e-01
9.74674076e-02 3.05349410e-01 3.77286643e-01 -3.59826505e-01
-1.44030049e-01 -7.01222941e-02 -1.23330033e+00 1.62327141e-01
9.28699255e-01 -4.23562020e-01 -4.15154934e-01 2.90047079e-01
1.88341588e-01 -1.35156987e-02 -1.08535194e+00 2.28097662e-01
2.97730118e-01 -9.00882661e-01 9.45550263e-01 2.80065000e-01
-6.64777607e-02 -2.90188849e-01 -1.80521592e-01 -1.51957536e+00
1.21338174e-01 -6.95580661e-01 9.40558873e-03 1.08406305e+00
3.11361521e-01 -9.59192216e-01 2.28126988e-01 2.57422477e-01
1.55992195e-01 -5.47162116e-01 -1.53739214e+00 -8.08953822e-01
4.82082039e-01 -6.36072159e-01 -2.09852327e-02 7.17903078e-01
5.57600439e-01 9.10413414e-02 1.40798673e-01 1.04512684e-01
1.16923928e+00 -5.22611260e-01 -1.03391446e-01 -1.53951049e+00
-2.93385684e-01 -5.79092264e-01 -6.20163858e-01 -7.27912605e-01
1.54653355e-01 -1.04407632e+00 4.69523184e-02 -1.44290233e+00
-8.92567858e-02 -3.84141028e-01 -1.31984400e-02 -4.72117811e-02
5.40800691e-01 3.93576533e-01 -1.50274947e-01 -1.06393538e-01
4.38901111e-02 4.36677784e-01 8.67847741e-01 1.43810049e-01
-3.81515957e-02 -1.99580252e-01 -2.65407324e-01 8.11424792e-01
6.20773494e-01 -4.84468043e-01 -4.51834261e-01 -1.67012781e-01
8.09901297e-01 8.90119374e-02 8.31109405e-01 -1.29659164e+00
3.49940568e-01 4.99325484e-01 3.17476958e-01 -1.63547009e-01
5.88887990e-01 -8.89546335e-01 4.31452096e-01 3.88292104e-01
-1.34339765e-01 -3.81974280e-01 2.75343716e-01 5.97162127e-01
-4.22783382e-02 -2.29334712e-01 8.83529961e-01 2.85764337e-02
1.30753040e-01 2.62258742e-02 -6.47934973e-01 1.62839834e-02
7.93235958e-01 -6.57985881e-02 -8.36278722e-02 -4.18170184e-01
-1.10829341e+00 -3.08445811e-01 1.19790815e-01 -2.71545529e-01
6.21511102e-01 -1.24129927e+00 -6.12903357e-01 5.78771830e-02
-4.03136283e-01 -3.51328433e-01 1.66083425e-01 1.67217839e+00
-4.00593817e-01 2.65315086e-01 -2.96418428e-01 -5.65615118e-01
-7.44071603e-01 2.97606587e-01 9.03654873e-01 -6.04644008e-02
-5.33983827e-01 8.01446199e-01 1.14235394e-01 -1.91403165e-01
4.16511446e-02 -5.28877378e-01 -1.91844374e-01 2.06055343e-01
3.64919156e-01 5.99577427e-01 1.85874596e-01 -7.69451499e-01
-6.33375704e-01 8.28853548e-01 6.69686019e-01 -6.58508897e-01
1.32134724e+00 -4.19121236e-01 -5.92717230e-01 5.43201685e-01
1.43513465e+00 2.16949701e-01 -1.25115216e+00 2.07859501e-01
-1.55788377e-01 -6.16587624e-02 5.87040842e-01 -8.21009457e-01
-1.24552798e+00 1.14815044e+00 8.23100924e-01 2.90323019e-01
1.02270198e+00 3.20715122e-02 2.01452985e-01 2.23001931e-02
5.07886827e-01 -9.19335544e-01 -4.06468600e-01 -9.15393326e-03
1.16924155e+00 -5.34566522e-01 1.56723723e-01 -7.30559111e-01
1.65286183e-01 1.01263714e+00 -9.60994288e-02 -5.27116835e-01
1.38851893e+00 4.38785851e-01 -3.41402978e-01 -3.04277331e-01
-8.08371454e-02 7.06247464e-02 4.07028973e-01 5.38317919e-01
4.65327412e-01 -5.11836186e-02 -6.20894313e-01 5.78046978e-01
-1.68418765e-01 -2.09861286e-02 5.90140164e-01 4.94692981e-01
-2.20272198e-01 -8.53228807e-01 -2.62678057e-01 2.28490651e-01
-6.43674135e-01 -2.29363903e-01 -1.17153957e-01 5.64423800e-01
8.76444355e-02 9.09016788e-01 5.18937409e-02 4.14827168e-02
6.90713897e-02 5.59601560e-02 9.23102915e-01 -1.99002400e-01
-2.47806624e-01 2.26049617e-01 -2.89583474e-01 -7.28441656e-01
-6.43104672e-01 -7.37871170e-01 -1.22254169e+00 3.55632231e-02
-5.77185154e-01 9.73932147e-02 1.28687501e+00 9.85639870e-01
1.87258333e-01 5.85307896e-01 1.47791684e-01 -1.12585533e+00
-3.44008982e-01 -8.85812700e-01 -9.92390513e-01 -4.12155204e-02
3.23331386e-01 -1.01063085e+00 -6.93922460e-01 -1.76824436e-01]
|
[7.034594535827637, 3.9103424549102783]
|
a1fa1eb5-d712-4551-a94b-1aa3629e17ee
|
multi-granularity-semantic-aware-graph-model
|
2205.02132
| null |
https://arxiv.org/abs/2205.02132v2
|
https://arxiv.org/pdf/2205.02132v2.pdf
|
Multi-Granularity Semantic Aware Graph Model for Reducing Position Bias in Emotion-Cause Pair Extraction
|
The Emotion-Cause Pair Extraction (ECPE) task aims to extract emotions and causes as pairs from documents. We observe that the relative distance distribution of emotions and causes is extremely imbalanced in the typical ECPE dataset. Existing methods have set a fixed size window to capture relations between neighboring clauses. However, they neglect the effective semantic connections between distant clauses, leading to poor generalization ability towards position-insensitive data. To alleviate the problem, we propose a novel Multi-Granularity Semantic Aware Graph model (MGSAG) to incorporate fine-grained and coarse-grained semantic features jointly, without regard to distance limitation. In particular, we first explore semantic dependencies between clauses and keywords extracted from the document that convey fine-grained semantic features, obtaining keywords enhanced clause representations. Besides, a clause graph is also established to model coarse-grained semantic relations between clauses. Experimental results indicate that MGSAG surpasses the existing state-of-the-art ECPE models. Especially, MGSAG outperforms other models significantly in the condition of position-insensitive data.
|
['Songlin Hu', 'Wei Zhou', 'Lingwei Wei', 'Qianwen Ma', 'Yinan Bao']
|
2022-05-04
| null | null | null | null |
['emotion-cause-pair-extraction']
|
['natural-language-processing']
|
[-5.96718118e-03 1.32139036e-02 -4.08854693e-01 -7.13291168e-01
-6.71205461e-01 -4.87472296e-01 5.35215616e-01 5.63380301e-01
-1.35590598e-01 6.32173657e-01 5.72975934e-01 1.89055026e-01
-5.96498668e-01 -1.00380325e+00 -3.88057947e-01 -5.04194617e-01
2.42151134e-02 2.66559094e-01 1.51772559e-01 -4.27255154e-01
3.08288991e-01 4.74653244e-02 -1.77215600e+00 5.53510845e-01
1.15737748e+00 1.03490961e+00 3.32795382e-02 -7.50803277e-02
-6.99413121e-01 7.24955201e-01 -6.97621346e-01 -5.51493585e-01
-4.13930297e-01 -4.86898452e-01 -9.00326788e-01 2.95788161e-02
-1.60847783e-01 2.97647685e-01 -1.22687615e-01 1.16676569e+00
2.17657715e-01 7.70825446e-02 6.66724265e-01 -1.58686090e+00
-8.27287972e-01 8.28008950e-01 -7.14430988e-01 8.77701715e-02
4.37796891e-01 -7.41087735e-01 1.56903028e+00 -7.41239488e-01
6.79632604e-01 1.55500650e+00 4.05469328e-01 3.35843384e-01
-7.93844461e-01 -7.86741614e-01 7.53973186e-01 4.57953751e-01
-1.41915560e+00 1.86609983e-01 1.06612515e+00 3.75290550e-02
1.12396801e+00 4.37259793e-01 5.70659220e-01 1.05924332e+00
9.12056565e-02 9.32968557e-01 9.39700305e-01 -2.24496692e-01
2.70891637e-01 7.84328505e-02 3.97146553e-01 4.43133026e-01
1.57692030e-01 -4.19818431e-01 -7.28148103e-01 -2.42210597e-01
3.97423893e-01 -7.69324303e-02 -4.23727632e-01 -1.56717539e-01
-1.05074573e+00 1.01249516e+00 6.23064160e-01 6.38483167e-01
-2.81334877e-01 -1.10589735e-01 5.71935177e-01 2.42827773e-01
6.21847689e-01 4.18885648e-01 -7.01760411e-01 -1.33378040e-02
-6.03778660e-01 2.62757212e-01 5.38802981e-01 1.12317622e+00
7.68996179e-01 -3.60544711e-01 -2.70166218e-01 9.55787420e-01
2.29900643e-01 -1.08065587e-02 5.52847266e-01 -3.50438774e-01
6.71519458e-01 1.20496213e+00 -3.30390543e-01 -1.73005903e+00
-4.57274169e-01 -5.97068727e-01 -7.36307681e-01 -7.22793221e-01
-2.90300310e-01 3.38185802e-02 -6.30940199e-01 1.78758276e+00
5.04780293e-01 -6.32765889e-02 2.14876145e-01 1.00850534e+00
1.22604811e+00 7.87507355e-01 2.67801732e-01 -3.17388713e-01
1.49405062e+00 -8.66200030e-01 -1.08016551e+00 -4.83182073e-01
8.29219580e-01 -5.86839557e-01 1.06662583e+00 2.15600789e-01
-5.50751209e-01 -1.98316291e-01 -9.67558742e-01 -1.84044689e-01
-7.15129256e-01 6.45329505e-02 1.02560008e+00 1.29333794e-01
-6.39425099e-01 3.42254788e-01 -1.76466957e-01 -2.46419013e-01
1.95005700e-01 1.93487704e-01 -3.48765701e-01 -1.39311984e-01
-1.86682177e+00 5.58721542e-01 7.38829255e-01 4.43948656e-02
3.49840671e-02 -6.62959218e-01 -1.01383352e+00 2.68290102e-01
6.56673729e-01 -4.20267582e-01 7.36202598e-01 -8.97973418e-01
-7.07824171e-01 8.61292005e-01 -2.97275484e-01 5.44579513e-02
-1.93038315e-01 -9.25790444e-02 -7.80854464e-01 4.42595929e-02
3.18490386e-01 6.13002360e-01 6.04762495e-01 -1.31234920e+00
-7.14542806e-01 -5.98822296e-01 2.09437862e-01 5.46494484e-01
-6.78572059e-01 5.29951453e-02 -7.08739042e-01 -8.58556569e-01
3.45183194e-01 -4.16362792e-01 6.71876147e-02 -5.20054042e-01
-6.29667103e-01 -7.64308929e-01 8.70218992e-01 -3.39562565e-01
1.61295474e+00 -2.09389663e+00 1.71714649e-01 2.33036757e-01
3.92603934e-01 -1.62563920e-01 -1.68122739e-01 4.10646230e-01
-2.34736994e-01 1.80904821e-01 3.09272595e-02 -1.52943423e-02
2.53489792e-01 3.51768821e-01 -1.89813703e-01 -1.27191380e-01
3.08274090e-01 9.22993302e-01 -1.03054059e+00 -7.70058513e-01
-1.80908471e-01 3.69463742e-01 -5.27263105e-01 7.40007162e-02
-2.08863884e-01 -1.43928275e-01 -1.00465870e+00 8.22245717e-01
7.67525315e-01 -4.04801607e-01 2.60909319e-01 -4.76540267e-01
3.56801391e-01 4.79701936e-01 -1.01152003e+00 1.59470642e+00
-3.54568124e-01 1.77775636e-01 -1.39885366e-01 -1.14375949e+00
1.33109522e+00 8.79397392e-02 4.86734688e-01 -8.86185408e-01
2.72083253e-01 1.57649606e-01 -3.12996447e-01 -3.90385687e-01
4.83124316e-01 -3.02207947e-01 -5.74009657e-01 -3.74180675e-02
1.21386284e-02 -1.85511932e-02 1.94661185e-01 4.92338538e-01
9.53523219e-01 -1.96572617e-01 3.76656950e-01 -3.72532815e-01
4.59954500e-01 1.10010663e-02 8.91287506e-01 8.92957374e-02
-7.71197975e-02 4.30362642e-01 9.27816093e-01 -1.08073711e-01
-2.18309760e-01 -8.65493119e-01 8.58386308e-02 1.04665887e+00
7.16975391e-01 -9.62463737e-01 -6.27707064e-01 -9.85435307e-01
4.01473492e-02 8.05721700e-01 -7.46192873e-01 -3.12076986e-01
-2.20632315e-01 -9.42384779e-01 2.40713581e-01 6.21718168e-01
6.03665709e-01 -9.18333471e-01 2.09689103e-02 2.29444057e-01
-6.53203964e-01 -1.11301672e+00 -2.90216327e-01 2.40884691e-01
-5.39072573e-01 -1.08257329e+00 -1.97551355e-01 -9.15800393e-01
5.08664370e-01 1.68165118e-01 1.39161801e+00 7.52075836e-02
-1.23687908e-01 -7.29832724e-02 -7.58044362e-01 -3.33087951e-01
3.55452895e-01 2.00125966e-02 -3.45037580e-01 -8.50084126e-02
1.12220144e+00 -4.55327272e-01 -5.33486247e-01 1.16210744e-01
-8.61078024e-01 7.43077919e-02 4.93681997e-01 7.42227256e-01
6.99989915e-01 5.51230252e-01 8.51498306e-01 -9.57235813e-01
9.04667914e-01 -6.98138118e-01 -9.89580080e-02 3.96455407e-01
-7.57463634e-01 2.41863932e-02 5.75887442e-01 -1.24186099e-01
-1.26391399e+00 -4.65170383e-01 -6.25016987e-02 -1.33314043e-01
-1.37292579e-01 7.57294416e-01 -6.72992051e-01 5.14075875e-01
4.82585169e-02 1.34157583e-01 -6.10620320e-01 -2.17408240e-01
3.47527653e-01 6.95613146e-01 3.94750565e-01 -6.83666289e-01
4.28345829e-01 3.24485719e-01 -1.06335267e-01 -3.48568678e-01
-1.30212653e+00 -6.40676618e-01 -3.31438154e-01 -4.03606854e-02
8.41459513e-01 -1.04478979e+00 -5.00074804e-01 1.05458289e-01
-1.17754650e+00 3.06475163e-01 1.92301795e-02 4.29677695e-01
-2.09508657e-01 3.58631015e-02 -7.75882304e-01 -4.70550269e-01
-1.23200588e-01 -6.53307498e-01 1.44567776e+00 2.92090058e-01
-4.76105094e-01 -1.06286597e+00 -1.77497536e-01 4.26348507e-01
-6.45718873e-02 4.99956846e-01 1.31695187e+00 -7.73279130e-01
-2.50343829e-01 -5.77034764e-02 -5.65378547e-01 -4.98623066e-02
4.57977861e-01 1.06897457e-02 -6.93502247e-01 1.49848372e-01
5.53036444e-02 -4.04451668e-01 9.50074315e-01 1.28949314e-01
1.28871453e+00 -4.60157603e-01 -6.38139129e-01 3.14854294e-01
1.45455492e+00 2.26184249e-01 4.90564525e-01 3.79145920e-01
7.34806478e-01 9.91442442e-01 1.29413009e+00 5.12963057e-01
7.50594735e-01 3.76816362e-01 5.18000841e-01 -3.23358148e-01
2.75599301e-01 -3.36263835e-01 -6.91168979e-02 9.96629834e-01
2.23609418e-01 -5.20187795e-01 -5.63165367e-01 7.91469395e-01
-1.92634439e+00 -6.25611901e-01 -3.56249690e-01 1.51606548e+00
9.33984876e-01 8.36246014e-02 -2.98077703e-01 2.88065225e-01
8.55904281e-01 4.32425290e-01 -2.46746778e-01 -4.45638686e-01
-2.12447420e-01 -6.48426265e-02 -1.00347191e-01 1.60191402e-01
-1.03820252e+00 1.04627466e+00 4.70586777e+00 1.41419792e+00
-7.94536650e-01 -2.96222251e-02 6.70663238e-01 -2.74045523e-02
-8.65388989e-01 -2.01128975e-01 -6.96789861e-01 4.71723080e-01
3.37794304e-01 -2.85982430e-01 7.55228475e-02 8.22233856e-01
-9.46373641e-02 1.70400202e-01 -9.31645274e-01 9.75299299e-01
2.25765392e-01 -9.47407663e-01 2.51649380e-01 -1.13490932e-01
7.59611785e-01 -5.48049688e-01 -2.18949467e-01 3.86303246e-01
2.38351692e-02 -8.92022491e-01 4.46698725e-01 2.16281831e-01
5.90812802e-01 -1.34530568e+00 1.04657543e+00 2.42233593e-02
-1.59598637e+00 7.19680190e-02 -4.66871142e-01 -5.65018207e-02
-2.35691667e-02 9.88740444e-01 -2.72610515e-01 9.63179231e-01
8.56370091e-01 8.93696427e-01 -5.66336513e-01 3.68521303e-01
-3.94695550e-01 3.65440696e-01 -8.62383395e-02 -4.07697231e-01
5.82401514e-01 -1.38445899e-01 2.11789459e-01 1.27071381e+00
4.32765216e-01 4.23930794e-01 5.60194217e-02 8.82015169e-01
-2.00288177e-01 4.09638464e-01 -3.24382097e-01 -1.62062258e-01
6.55724883e-01 1.42954803e+00 -9.51689422e-01 -2.93420196e-01
-4.57994163e-01 1.04200864e+00 5.85200489e-01 3.26749951e-01
-9.74144995e-01 -6.97834253e-01 7.91585445e-01 -2.98526466e-01
3.25225204e-01 3.92813206e-01 -2.41674006e-01 -1.11183846e+00
2.24899366e-01 -6.48826778e-01 7.11549819e-01 -8.24915707e-01
-1.50343966e+00 6.69959724e-01 -4.71372381e-02 -9.36364174e-01
-1.94119215e-01 -4.33820128e-01 -5.99658728e-01 4.33264881e-01
-1.64564931e+00 -1.12244558e+00 -3.46387804e-01 7.22387314e-01
5.58686674e-01 2.68689722e-01 7.58530736e-01 1.82573631e-01
-4.55592453e-01 4.38663244e-01 -2.57969379e-01 4.18125801e-02
6.19920135e-01 -1.30791557e+00 -1.25885278e-01 5.06088197e-01
5.76209575e-02 7.62348294e-01 6.80708230e-01 -8.22314501e-01
-9.94723976e-01 -1.19368589e+00 1.47175252e+00 -1.52963772e-01
5.73608279e-01 -4.69251215e-01 -1.02453589e+00 5.24547756e-01
2.32817799e-01 -6.03523888e-02 7.89649785e-01 4.94498551e-01
-4.78693455e-01 -1.98841453e-01 -1.05646408e+00 5.18706083e-01
1.19130266e+00 -4.85644758e-01 -9.77367640e-01 2.90322989e-01
1.05773902e+00 -9.29542333e-02 -8.12518120e-01 6.11177802e-01
8.92038271e-02 -1.01127887e+00 8.96715283e-01 -4.82985109e-01
8.00978184e-01 -2.01601148e-01 -2.01081455e-01 -1.45469570e+00
-3.97602916e-01 -1.61082745e-01 -1.25747129e-01 1.87886322e+00
3.90867054e-01 -4.96280581e-01 5.81570029e-01 4.30273861e-01
1.08854845e-01 -1.02360427e+00 -5.97797751e-01 -7.47364879e-01
-8.71854052e-02 -2.36705348e-01 1.06282020e+00 1.42587972e+00
4.28679585e-01 7.09926426e-01 1.47759533e-02 5.96462004e-02
3.06576669e-01 7.00867534e-01 3.26707065e-02 -1.25236654e+00
-6.47434667e-02 -4.03245836e-01 -3.21524203e-01 -6.88478112e-01
4.29758579e-01 -8.53687763e-01 1.23642581e-02 -1.78016222e+00
4.57488596e-01 -4.35588807e-01 -6.38031363e-01 5.00287831e-01
-7.45158434e-01 3.78160775e-02 -5.31268045e-02 -1.46446228e-01
-8.70225191e-01 9.44311023e-01 1.25893688e+00 -5.88276237e-02
4.50302809e-02 -3.98181468e-01 -1.11053455e+00 8.74994218e-01
7.44282901e-01 -5.23174167e-01 -9.07267153e-01 -2.00450212e-01
5.50035834e-01 -2.48100162e-01 2.29783952e-01 -5.26123047e-01
1.28524408e-01 -3.04640979e-01 3.24516058e-01 -7.23352551e-01
2.12645814e-01 -8.10185730e-01 -1.87650621e-01 -8.30550492e-02
-4.01580185e-01 3.29311676e-02 -1.56978305e-04 6.69874072e-01
-7.90970683e-01 -3.24393772e-02 4.02049534e-02 -1.35228634e-02
-9.49816883e-01 1.20753326e-01 -1.83601421e-03 3.38740826e-01
8.55146408e-01 2.91983429e-02 -4.73356038e-01 -2.38527641e-01
-3.65984470e-01 4.30560946e-01 1.40165165e-01 6.94406807e-01
6.98535860e-01 -1.60882366e+00 -4.47043151e-01 -1.12782225e-01
5.13832569e-01 7.57439956e-02 3.92243981e-01 5.11500299e-01
1.99071854e-01 6.31788850e-01 1.96246400e-01 -1.51723698e-01
-1.46764636e+00 8.20550144e-01 -1.62370846e-01 -6.02071643e-01
-2.99825221e-01 1.14618492e+00 5.45855105e-01 -5.53221524e-01
4.58976999e-02 -3.83249998e-01 -5.67587078e-01 4.61913526e-01
3.21681440e-01 3.69212329e-02 1.21930391e-01 -5.67170620e-01
-7.06805468e-01 6.18352234e-01 -7.67993107e-02 4.34020430e-01
1.32597888e+00 -3.57996762e-01 -4.41478759e-01 3.82888913e-01
1.42155004e+00 5.20999767e-02 -6.96202278e-01 -1.85882851e-01
1.58303618e-01 -3.68452221e-01 1.49552211e-01 -7.70511746e-01
-1.25219011e+00 6.46891356e-01 -1.26380831e-01 3.28884840e-01
1.55032563e+00 3.63246500e-01 8.58072996e-01 4.30842414e-02
3.08478653e-01 -1.29809666e+00 7.87227452e-02 2.71503568e-01
9.38619852e-01 -1.04019690e+00 -1.08939283e-01 -1.17815197e+00
-7.97849894e-01 9.28203821e-01 9.77590382e-01 8.13388377e-02
4.24864680e-01 2.87900209e-01 -1.61211878e-01 -6.91509187e-01
-9.22325969e-01 -3.70093554e-01 5.87335408e-01 4.72508162e-01
6.57598495e-01 1.23221785e-01 -7.69051135e-01 1.25797749e+00
-2.87684888e-01 -2.27595419e-01 -1.54415537e-02 7.50574410e-01
-3.08032483e-01 -9.93292630e-01 -1.38894677e-01 3.79404038e-01
-3.75257969e-01 -3.46420288e-01 -8.87913227e-01 9.56791937e-01
3.49950761e-01 1.13343859e+00 1.79527774e-01 -4.22712684e-01
1.41313836e-01 -1.17625609e-01 8.40677321e-02 -3.86918873e-01
-4.36559111e-01 1.69270501e-01 3.34704340e-01 -7.71023631e-01
-5.98570764e-01 -3.80949378e-01 -1.77915287e+00 -1.67791694e-01
-4.53703195e-01 4.17488068e-01 2.48726189e-01 1.15963292e+00
4.52034444e-01 1.04511321e+00 6.07426405e-01 -7.68699124e-02
-1.76218022e-02 -6.44854963e-01 -1.01474929e+00 6.42829716e-01
-1.36005497e-02 -6.66352570e-01 -5.03253162e-01 -3.51949006e-01]
|
[12.625280380249023, 6.2133049964904785]
|
00fbd2a3-6ac7-448f-9620-e6de0ad85a9f
|
modular-adaptation-for-cross-domain-few-shot
|
2104.00619
| null |
https://arxiv.org/abs/2104.00619v1
|
https://arxiv.org/pdf/2104.00619v1.pdf
|
Modular Adaptation for Cross-Domain Few-Shot Learning
|
Adapting pre-trained representations has become the go-to recipe for learning new downstream tasks with limited examples. While literature has demonstrated great successes via representation learning, in this work, we show that substantial performance improvement of downstream tasks can also be achieved by appropriate designs of the adaptation process. Specifically, we propose a modular adaptation method that selectively performs multiple state-of-the-art (SOTA) adaptation methods in sequence. As different downstream tasks may require different types of adaptation, our modular adaptation enables the dynamic configuration of the most suitable modules based on the downstream task. Moreover, as an extension to existing cross-domain 5-way k-shot benchmarks (e.g., miniImageNet -> CUB), we create a new high-way (~100) k-shot benchmark with data from 10 different datasets. This benchmark provides a diverse set of domains and allows the use of stronger representations learned from ImageNet. Experimental results show that by customizing adaptation process towards downstream tasks, our modular adaptation pipeline (MAP) improves 3.1% in 5-shot classification accuracy over baselines of finetuning and Prototypical Networks.
|
['Yi Yao', 'Ajay Divakaran', 'Nikoletta Basiou', 'Giedrius Buracas', 'Yunye Gong', 'Meng Ye', 'Xiao Lin']
|
2021-04-01
| null | null | null | null |
['cross-domain-few-shot', 'cross-domain-few-shot-learning']
|
['computer-vision', 'computer-vision']
|
[ 2.46386662e-01 -6.61527216e-02 -2.78767407e-01 -5.02390325e-01
-6.63007677e-01 -5.04277229e-01 8.69739056e-01 -1.97793320e-01
-7.31139362e-01 6.97504818e-01 5.33651650e-01 -5.78890443e-02
-1.06667569e-02 -7.34404445e-01 -6.62654042e-01 -5.52522063e-01
1.33568704e-01 4.00493860e-01 6.46089494e-01 -7.22556233e-01
1.46429956e-01 3.02595884e-01 -1.58059537e+00 5.87297559e-01
5.35396934e-01 7.35354424e-01 3.22377294e-01 6.04767084e-01
-6.28301576e-02 4.58076060e-01 -5.01614571e-01 -3.12785864e-01
4.74290639e-01 -5.01856744e-01 -9.57012951e-01 -3.20136428e-01
3.05851996e-01 -2.35093951e-01 -3.77320468e-01 5.06122112e-01
8.26969087e-01 5.57588995e-01 8.99230301e-01 -8.14471781e-01
-1.12492633e+00 8.01688254e-01 -2.97496170e-01 6.73949122e-01
-2.31067672e-01 5.58319747e-01 8.40931416e-01 -9.73027527e-01
7.16666579e-01 1.02529514e+00 8.11232030e-01 9.92079616e-01
-1.55676091e+00 -8.25275898e-01 2.91315675e-01 4.60233301e-01
-1.18575525e+00 -6.88948274e-01 2.64646590e-01 -4.05850649e-01
1.58826256e+00 -2.84996063e-01 3.32414150e-01 1.65912795e+00
9.85882282e-02 5.64655125e-01 8.98737192e-01 -4.98834908e-01
2.72120953e-01 1.32932380e-01 3.23811695e-02 2.25356534e-01
-9.14711040e-03 1.16123535e-01 -4.31633055e-01 4.20249924e-02
6.44582272e-01 6.92307949e-02 -1.44120589e-01 -4.46632177e-01
-1.18484020e+00 8.01934540e-01 4.78603572e-01 3.03709328e-01
-2.06175856e-02 1.83524042e-01 7.44465768e-01 6.39563918e-01
3.85993004e-01 7.32270896e-01 -8.47261012e-01 -1.72706112e-01
-6.78376675e-01 -4.24696580e-02 7.45565236e-01 1.08877063e+00
8.01308632e-01 7.46177286e-02 -5.95155895e-01 1.02542663e+00
-1.69002816e-01 -7.51741603e-02 1.23255622e+00 -8.29787314e-01
4.03382331e-01 4.02713925e-01 -2.02172592e-01 -1.35411337e-01
-1.93256855e-01 -4.97364402e-01 -6.09914362e-01 2.66105175e-01
2.35593572e-01 -2.47466728e-01 -1.31179547e+00 1.88965654e+00
8.99518952e-02 4.69257325e-01 2.89012909e-01 6.02988243e-01
8.25996697e-01 5.27332246e-01 5.82125068e-01 1.95995390e-01
1.06824505e+00 -1.28597653e+00 -1.50319681e-01 -3.43479604e-01
7.25007296e-01 -5.50930083e-01 1.23345399e+00 -1.75262317e-02
-7.51051426e-01 -8.77676964e-01 -1.19097078e+00 -1.18693389e-01
-7.72297800e-01 -2.23008081e-01 4.04622704e-01 2.83679605e-01
-1.10260820e+00 7.84727454e-01 -4.99935001e-01 -8.58240902e-01
4.15435135e-01 3.13040316e-01 -4.79332596e-01 -1.79180145e-01
-1.44268739e+00 1.31419957e+00 6.32191896e-01 -6.65831447e-01
-1.41348720e+00 -1.18036985e+00 -6.68190002e-01 2.26754934e-01
1.42668173e-01 -9.08037364e-01 1.44687068e+00 -8.29437017e-01
-1.75820923e+00 8.32550764e-01 2.54698724e-01 -5.40509045e-01
2.77715772e-01 -3.71833533e-01 -3.29721183e-01 2.42612604e-02
8.99789110e-02 9.44878161e-01 1.07092166e+00 -8.74537826e-01
-6.21406138e-01 -8.41218680e-02 2.38839105e-01 1.09987117e-01
-6.26674175e-01 -8.99713784e-02 -2.86831617e-01 -5.92052102e-01
-5.53039074e-01 -7.75356054e-01 -4.43491817e-01 -2.66895652e-01
2.68948883e-01 -2.04247102e-01 7.07530499e-01 -2.55346596e-01
1.16475511e+00 -2.23671627e+00 2.15758175e-01 -2.84698546e-01
-3.58872898e-02 5.28558791e-01 -6.19489908e-01 5.02539754e-01
-3.65475118e-01 4.79397960e-02 -2.27625072e-01 -3.67698610e-01
-1.58943206e-01 1.62382692e-01 -1.50191456e-01 1.38981700e-01
4.18083102e-01 1.05029535e+00 -9.74845648e-01 -2.21123621e-01
8.88543576e-02 3.80149126e-01 -6.44953132e-01 1.82459995e-01
4.90403771e-02 2.95445561e-01 -2.09943265e-01 3.41706038e-01
1.94780663e-01 -2.81166732e-01 -1.26759052e-01 -8.79040956e-02
1.10680908e-01 1.52378827e-01 -6.34911716e-01 2.21432424e+00
-7.61433780e-01 4.20015186e-01 -4.59305376e-01 -1.19336438e+00
1.03526437e+00 3.00952315e-01 2.52609849e-01 -6.58748507e-01
4.09393013e-03 -6.58663362e-02 1.69878408e-01 -4.53533202e-01
3.19666058e-01 -2.20933333e-01 -3.36400747e-01 4.25475746e-01
9.48140383e-01 1.07415095e-01 1.80423379e-01 1.62767097e-01
1.53491962e+00 4.31150258e-01 7.34339714e-01 -2.01380312e-01
2.58744270e-01 1.90334156e-01 5.67204535e-01 8.80491674e-01
-5.18474042e-01 6.85159564e-01 3.88774909e-02 -4.14155424e-01
-1.22319353e+00 -1.26797378e+00 -5.47377989e-02 1.91198421e+00
-2.91194022e-01 -3.29871476e-01 -5.15659809e-01 -8.85230482e-01
5.16624786e-02 7.89381504e-01 -9.90738451e-01 -7.64097750e-01
-4.88503218e-01 -7.96183288e-01 6.95628643e-01 8.78570139e-01
4.58082527e-01 -1.22383630e+00 -5.56082070e-01 3.69162917e-01
1.89247161e-01 -9.99688983e-01 -4.77949917e-01 6.50771141e-01
-9.67473149e-01 -7.61329055e-01 -7.86597729e-01 -9.15497363e-01
3.82571161e-01 5.71317911e-01 1.33916008e+00 -3.67768109e-01
-3.37161064e-01 5.36519587e-01 -6.52230501e-01 -1.58603802e-01
-4.33258772e-01 6.23592794e-01 4.12819870e-02 -1.94740847e-01
7.27999151e-01 -9.39414382e-01 -7.38364458e-01 3.84905875e-01
-8.73135924e-01 -2.01303855e-01 8.87953579e-01 9.92321193e-01
5.08426070e-01 -5.66235900e-01 1.05267370e+00 -1.10350680e+00
7.43378997e-01 -1.01474452e+00 -1.41762510e-01 2.85711914e-01
-7.45217919e-01 2.42274612e-01 8.68818879e-01 -7.76271880e-01
-1.59957159e+00 1.69134870e-01 4.16702665e-02 -4.87499207e-01
-5.30065358e-01 2.42260620e-01 2.73584444e-02 -1.07448511e-02
1.47442114e+00 2.59138495e-01 -1.65868610e-01 -6.18222833e-01
9.22701061e-01 5.44899464e-01 4.44415689e-01 -5.21965981e-01
6.66007936e-01 3.70478809e-01 -3.62573087e-01 -4.31125343e-01
-8.29230547e-01 -6.50458097e-01 -9.80863094e-01 1.34648815e-01
8.76197100e-01 -1.15030956e+00 -1.03490055e-02 3.08882833e-01
-7.88623393e-01 -7.86471903e-01 -5.84498882e-01 1.97527468e-01
-6.41006470e-01 -6.97175562e-02 -6.02486193e-01 -7.65479952e-02
-3.89765501e-01 -1.09219646e+00 7.52271533e-01 2.45188326e-01
-4.29299653e-01 -9.91789222e-01 4.00073409e-01 -6.36347895e-03
8.07597637e-01 -1.00629941e-01 8.23356628e-01 -9.32433009e-01
-7.64182657e-02 3.50212753e-02 -1.39435962e-01 4.90586787e-01
7.14839399e-02 -4.47903007e-01 -1.18078017e+00 -2.10910678e-01
-2.85968542e-01 -7.16941357e-01 1.37641478e+00 2.13094071e-01
1.00346899e+00 7.18244836e-02 -5.01266181e-01 9.38293219e-01
1.46923506e+00 9.76218656e-03 7.57928729e-01 7.04681635e-01
4.44596767e-01 4.56861109e-01 4.76458043e-01 3.68017644e-01
2.14650422e-01 6.84603989e-01 1.09214164e-01 2.65682936e-01
-5.90203822e-01 -2.83435851e-01 5.61800420e-01 6.23852909e-01
-1.45143569e-01 2.56937563e-01 -7.03789592e-01 7.46589601e-01
-1.82911205e+00 -9.97890949e-01 3.82858843e-01 2.01406121e+00
1.06717896e+00 2.63219047e-02 5.07792830e-02 -4.53077525e-01
6.43276513e-01 3.77823785e-02 -9.35742021e-01 -5.15468478e-01
9.01560485e-02 6.83397353e-01 4.74862456e-01 2.20381133e-02
-1.06625724e+00 1.36828065e+00 6.59952068e+00 1.03151071e+00
-8.69465649e-01 5.83874583e-01 4.35348481e-01 -4.93972063e-01
-4.13986668e-02 -3.32072563e-02 -1.24586213e+00 4.04092610e-01
1.27315819e+00 -3.00284743e-01 2.49314025e-01 1.08083212e+00
-5.86183667e-02 2.66417414e-01 -1.21485734e+00 6.07826352e-01
1.34800732e-01 -1.38004875e+00 1.46077231e-01 -2.40993291e-01
1.03648961e+00 6.02534235e-01 1.35102244e-02 1.13583326e+00
7.16880858e-01 -8.55768979e-01 3.85154992e-01 4.94702697e-01
1.01226509e+00 -6.72320068e-01 5.00744343e-01 1.80580601e-01
-1.06098390e+00 -4.37411636e-01 -6.68060720e-01 -3.47935073e-02
6.62643909e-02 1.54167861e-01 -9.05761898e-01 6.09630123e-02
9.90155995e-01 7.51226723e-01 -7.54541516e-01 1.08508074e+00
-3.76388818e-01 5.26700258e-01 1.53168768e-01 1.76934987e-01
1.92882672e-01 2.01127797e-01 3.43061328e-01 1.66056573e+00
3.54320168e-01 8.25809911e-02 -1.22225918e-01 5.63879311e-01
-2.93559462e-01 -8.01601112e-02 -9.02674198e-01 1.53680876e-01
5.41737914e-01 1.38855410e+00 -3.15549880e-01 -4.24357086e-01
-5.17105997e-01 1.26533282e+00 8.69979620e-01 5.03003061e-01
-6.89030349e-01 -4.54276592e-01 1.19105613e+00 3.78481448e-02
6.36975944e-01 -1.03058316e-01 -1.20608360e-01 -1.17220652e+00
-4.65170711e-01 -6.98940158e-01 6.31596088e-01 -4.81355518e-01
-1.74109268e+00 5.39017916e-01 8.86224732e-02 -1.26076138e+00
-3.62056792e-01 -5.65068722e-01 -7.92512298e-01 7.20942557e-01
-1.56008375e+00 -1.13040400e+00 -3.39889079e-01 7.59643376e-01
1.00429869e+00 -6.26362145e-01 1.15513182e+00 1.30136698e-01
-3.37802917e-01 9.10516262e-01 2.67446518e-01 4.36705947e-02
1.27888298e+00 -1.01187015e+00 7.52057433e-01 7.03406513e-01
-6.92903101e-02 4.96750921e-01 4.11883622e-01 -3.39819938e-01
-8.81353915e-01 -1.40695226e+00 4.63875979e-01 -6.97739899e-01
9.09445643e-01 -2.91252822e-01 -9.18347359e-01 9.46512461e-01
3.87755960e-01 1.96432054e-01 8.89976501e-01 1.46784276e-01
-7.12188184e-01 -2.36207768e-01 -1.00777423e+00 6.16072834e-01
1.51826751e+00 -3.24641049e-01 -8.81839812e-01 -9.98719502e-03
1.11024666e+00 3.99037674e-02 -1.01204085e+00 3.94638062e-01
4.61801350e-01 -8.69663596e-01 1.17168915e+00 -1.13983524e+00
7.16074109e-01 2.06034128e-02 -1.10843733e-01 -1.92185020e+00
-9.52681303e-01 -3.10386539e-01 -2.25850940e-01 1.28115380e+00
6.34678185e-01 -5.53461909e-01 5.40999115e-01 3.84148449e-01
-4.77290183e-01 -7.11056054e-01 -8.31419468e-01 -9.20751512e-01
3.54856998e-01 -2.55566120e-01 4.49903637e-01 9.68963742e-01
1.22852117e-01 7.51029849e-01 -2.89022148e-01 -3.60707939e-01
3.73803556e-01 -1.51802510e-01 8.49925458e-01 -1.05048335e+00
-7.16867805e-01 -6.63688719e-01 -4.14777040e-01 -9.40056264e-01
2.05135062e-01 -1.07600510e+00 6.52347654e-02 -1.27043366e+00
3.76916111e-01 -3.02026212e-01 -8.23607147e-01 6.97837532e-01
-2.72133857e-01 2.79764414e-01 1.86525941e-01 2.53001153e-01
-7.18841016e-01 6.08126163e-01 1.10252154e+00 -2.26745546e-01
-2.86213756e-01 -1.36172965e-01 -9.79778171e-01 4.25237089e-01
1.04452562e+00 -5.19738436e-01 -6.61808968e-01 -5.63509345e-01
4.87460027e-04 -5.52758098e-01 1.65633913e-02 -1.15911114e+00
1.33900076e-01 -6.95731491e-02 6.13612115e-01 1.41467586e-01
4.13922697e-01 -4.65192944e-01 -2.63603151e-01 3.49636316e-01
-5.59550762e-01 -5.73317781e-02 4.40666974e-01 8.43709588e-01
-2.75668763e-02 -4.89278942e-01 1.03461194e+00 -5.17167211e-01
-1.48417664e+00 2.91470319e-01 -2.91770488e-01 3.13181639e-01
1.22675824e+00 -1.92060307e-01 -5.76011181e-01 -1.37235090e-01
-9.71002579e-01 4.79795337e-02 3.01490128e-01 6.92473829e-01
6.16864085e-01 -1.29996085e+00 -7.74132073e-01 4.77736071e-02
6.74109161e-01 -3.77456427e-01 2.25892678e-01 6.29354477e-01
-1.83115760e-03 1.23196922e-01 -6.75542355e-01 -3.15090090e-01
-8.18621099e-01 5.51378965e-01 2.21904010e-01 -3.86423290e-01
-6.02393627e-01 1.19679248e+00 2.87683368e-01 -4.88518924e-01
-1.61264107e-01 5.46436384e-02 -4.59840298e-01 2.20373675e-01
6.77786648e-01 2.95711040e-01 1.36907220e-01 -2.96906680e-01
-1.29606545e-01 2.05631837e-01 -4.92390156e-01 -7.38964677e-02
1.67545521e+00 -1.27430871e-01 4.14795965e-01 4.59583402e-01
1.16702771e+00 -5.86991251e-01 -1.64609706e+00 -4.09334004e-01
-5.44193797e-02 -1.49674341e-01 -1.95320711e-01 -9.39392209e-01
-8.12824070e-01 8.76459062e-01 6.31569862e-01 -2.92879820e-01
1.10731363e+00 9.99024063e-02 7.23093390e-01 5.44076324e-01
4.29382116e-01 -1.33729041e+00 3.12841952e-01 8.14596832e-01
7.11391985e-01 -1.41306484e+00 -2.92777568e-01 2.73695886e-01
-7.79324234e-01 9.90842760e-01 1.12854850e+00 -3.96076113e-01
7.59782791e-01 -2.84952540e-02 2.97194370e-03 8.08014646e-02
-1.14883804e+00 -4.25277621e-01 1.29684910e-01 1.01408207e+00
4.08132941e-01 -9.54056382e-02 -1.77410483e-01 7.25812554e-01
4.51571345e-02 1.04250468e-01 4.00268853e-01 8.28157604e-01
-6.21160984e-01 -1.22401857e+00 2.63781976e-02 5.09554625e-01
-2.13675037e-01 -3.42455208e-01 -2.41266355e-01 8.66621375e-01
2.47234926e-01 6.94318175e-01 -9.15035829e-02 -4.89016920e-01
4.75810319e-01 4.08680052e-01 4.93917733e-01 -1.18279982e+00
-6.90937400e-01 -6.26741469e-01 7.86487535e-02 -6.62373066e-01
-1.77609697e-01 -4.28200692e-01 -9.09556329e-01 -1.79098118e-02
1.36994258e-01 -1.69416070e-01 3.80958885e-01 9.86771882e-01
8.11901987e-01 8.34260345e-01 1.88516840e-01 -1.15497947e+00
-8.68166089e-01 -1.46431923e+00 -3.09763402e-01 7.29739845e-01
-9.56128910e-02 -8.96104157e-01 -3.21603149e-01 2.91999817e-01]
|
[9.938730239868164, 2.9196720123291016]
|
2238f64d-5245-40ce-8c38-ec1e099c3389
|
slow-motion-matters-a-slow-motion-enhanced
|
2211.11324
| null |
https://arxiv.org/abs/2211.11324v1
|
https://arxiv.org/pdf/2211.11324v1.pdf
|
Slow Motion Matters: A Slow Motion Enhanced Network for Weakly Supervised Temporal Action Localization
|
Weakly supervised temporal action localization (WTAL) aims to localize actions in untrimmed videos with only weak supervision information (e.g. video-level labels). Most existing models handle all input videos with a fixed temporal scale. However, such models are not sensitive to actions whose pace of the movements is different from the ``normal" speed, especially slow-motion action instances, which complete the movements with a much slower speed than their counterparts with a normal speed. Here arises the slow-motion blurred issue: It is hard to explore salient slow-motion information from videos at ``normal" speed. In this paper, we propose a novel framework termed Slow Motion Enhanced Network (SMEN) to improve the ability of a WTAL network by compensating its sensitivity on slow-motion action segments. The proposed SMEN comprises a Mining module and a Localization module. The mining module generates mask to mine slow-motion-related features by utilizing the relationships between the normal motion and slow motion; while the localization module leverages the mined slow-motion features as complementary information to improve the temporal action localization results. Our proposed framework can be easily adapted by existing WTAL networks and enable them be more sensitive to slow-motion actions. Extensive experiments on three benchmarks are conducted, which demonstrate the high performance of our proposed framework.
|
['Dong Xu', 'Qian Yu', 'Rui Su', 'Weiqi Sun']
|
2022-11-21
| null | null | null | null |
['weakly-supervised-temporal-action', 'action-localization']
|
['computer-vision', 'computer-vision']
|
[ 4.11300540e-01 -2.53110796e-01 -9.34883475e-01 -1.15245335e-01
-3.26528281e-01 -3.57838959e-01 5.45852542e-01 -4.25647497e-01
-3.22445601e-01 4.37946618e-01 5.20698130e-01 6.91611245e-02
-5.47905453e-02 -3.19657594e-01 -6.26872718e-01 -8.96965742e-01
-4.68743503e-01 -3.32171917e-01 7.44404078e-01 6.02468150e-03
1.46307081e-01 7.16265887e-02 -1.31549621e+00 6.95914328e-01
5.43634832e-01 8.74215841e-01 4.63225394e-01 5.45688689e-01
1.25931188e-01 1.41265190e+00 -3.59067053e-01 3.70851368e-01
4.33541566e-01 -5.38155496e-01 -7.46684909e-01 1.68033585e-01
2.74230480e-01 -5.15666485e-01 -6.03655457e-01 9.58702207e-01
1.47206247e-01 3.44474375e-01 1.15940474e-01 -1.45973051e+00
-4.18736458e-01 6.17659748e-01 -8.56566131e-01 7.21141458e-01
5.43851137e-01 5.27934015e-01 8.23850155e-01 -7.48101473e-01
7.89416015e-01 1.31672251e+00 6.44388735e-01 5.59733808e-01
-9.21402395e-01 -6.66062057e-01 7.36772597e-01 5.27653992e-01
-1.16274035e+00 -4.55929786e-01 9.29331183e-01 -2.75507689e-01
7.76599526e-01 -8.05654526e-02 3.94844353e-01 1.43219864e+00
2.79425204e-01 1.23041642e+00 8.18161786e-01 2.51030009e-02
1.68463558e-01 -3.64109278e-01 -2.68257916e-01 6.68013215e-01
-1.72440648e-01 -6.56954125e-02 -8.29957664e-01 1.95701003e-01
8.59555602e-01 4.69882369e-01 -4.56735104e-01 -4.48649079e-01
-1.80211067e+00 5.00330448e-01 3.27448010e-01 4.66922104e-01
-3.48442525e-01 3.10691148e-01 6.57088041e-01 4.14554209e-01
4.01681721e-01 4.82164398e-02 -6.15001798e-01 -6.64496005e-01
-8.66508365e-01 -9.05162562e-03 4.14103687e-01 9.58736122e-01
5.79786241e-01 -5.49675710e-03 -3.57002318e-01 5.52358389e-01
-6.31149299e-03 2.35167518e-01 7.04871535e-01 -1.16259778e+00
8.16967368e-01 6.18501782e-01 2.22918898e-01 -1.12067449e+00
-3.22698087e-01 -1.89932615e-01 -8.69314134e-01 1.00007541e-02
4.93940502e-01 -1.04735844e-01 -7.97143638e-01 1.88811529e+00
3.63671392e-01 7.93887138e-01 -8.51540565e-02 1.18443370e+00
4.88718420e-01 6.40174747e-01 8.40537548e-02 -4.37044263e-01
1.08836329e+00 -1.37339163e+00 -8.69460464e-01 -2.89610744e-01
9.59506035e-01 -6.14476264e-01 1.22734892e+00 1.42160967e-01
-8.58417630e-01 -7.33669281e-01 -8.75073671e-01 2.40404695e-01
5.18242270e-02 1.78638533e-01 5.48504472e-01 5.52502684e-02
-7.83292532e-01 7.83961058e-01 -1.21219218e+00 -3.29610229e-01
4.89713371e-01 1.67240784e-01 -4.99315619e-01 -1.15669012e-01
-1.10069752e+00 5.28712451e-01 5.07819653e-01 1.65075645e-01
-1.09344578e+00 -4.72082376e-01 -1.01201689e+00 -2.82892674e-01
9.45199788e-01 -3.92626524e-01 1.21641922e+00 -1.32733655e+00
-1.31838548e+00 3.71865004e-01 -2.92177856e-01 -5.15388370e-01
7.33108222e-01 -4.33273315e-01 -5.14246821e-01 5.05695641e-01
3.59069020e-01 6.29545212e-01 1.08630097e+00 -7.28090346e-01
-9.49313164e-01 -7.19673559e-02 2.44573832e-01 3.23830426e-01
-3.77537102e-01 -1.69853438e-02 -6.51132703e-01 -9.83496904e-01
1.30996048e-01 -8.74746203e-01 -1.22545719e-01 1.85419649e-01
-2.56858677e-01 -2.17477188e-01 1.39917815e+00 -4.38151360e-01
1.46540463e+00 -2.27740717e+00 7.19090030e-02 -3.09909761e-01
1.19689047e-01 1.39335215e-01 -3.93354625e-01 3.43233913e-01
-2.16122687e-01 -2.23440960e-01 -6.91630738e-03 -1.98440701e-01
-2.29583964e-01 3.79858166e-01 -2.36632705e-01 6.10681236e-01
2.25655541e-01 9.18233514e-01 -1.39919877e+00 -5.78742623e-01
3.00127745e-01 1.46140844e-01 -3.13120097e-01 1.84089169e-01
-1.28839999e-01 6.63625836e-01 -6.50963843e-01 8.96426857e-01
3.08578938e-01 -3.37305605e-01 1.71775036e-02 -2.95962006e-01
-1.77249730e-01 1.50943503e-01 -1.25052953e+00 1.94395065e+00
-2.24871486e-01 5.47558665e-01 4.99014631e-02 -1.07759166e+00
4.57156628e-01 3.55969667e-01 9.65846062e-01 -6.25305772e-01
-1.70220017e-01 -8.51429403e-02 3.36864665e-02 -9.27781463e-01
2.32432857e-01 3.60179693e-02 3.94013785e-02 6.16312921e-01
-4.70056869e-02 6.75105095e-01 3.76275063e-01 1.78604156e-01
1.54709244e+00 6.82274282e-01 3.34218889e-01 1.54685825e-01
5.51679850e-01 -1.04238585e-01 1.15548217e+00 6.34528875e-01
-7.89561093e-01 4.69069004e-01 5.24902403e-01 -5.74672937e-01
-7.39829957e-01 -7.89534390e-01 3.91000450e-01 1.30954528e+00
5.62354922e-01 -5.35994053e-01 -5.95004559e-01 -1.22268200e+00
-2.80759215e-01 2.13828146e-01 -7.54856825e-01 -3.46222550e-01
-8.69718432e-01 -4.51231480e-01 3.26958358e-01 9.09577072e-01
7.99334407e-01 -1.29660630e+00 -8.42495799e-01 1.12353064e-01
-4.01541054e-01 -1.45935249e+00 -9.43746030e-01 2.23455913e-02
-9.42728579e-01 -1.11124671e+00 -7.12108970e-01 -7.49567330e-01
6.22170031e-01 7.01822281e-01 6.51460230e-01 -8.58298391e-02
3.02621815e-02 1.80861354e-01 -7.33525872e-01 1.37650266e-01
-9.06991437e-02 -1.26354307e-01 2.76314408e-01 3.94604951e-01
4.43831593e-01 -6.01351500e-01 -9.21693683e-01 7.67626345e-01
-9.81600523e-01 2.99116522e-01 7.85230339e-01 7.89763153e-01
6.06252968e-01 4.14326727e-01 4.17015791e-01 -5.63926101e-01
6.42718822e-02 -6.93620145e-01 -3.77454050e-03 -1.51247391e-02
-2.89511383e-01 -6.67895079e-02 8.39660048e-01 -9.68115687e-01
-9.60314572e-01 2.46387020e-01 2.73416042e-01 -8.70471716e-01
-2.72175699e-01 2.68600196e-01 -2.14445010e-01 1.56934723e-01
2.82649726e-01 4.69321996e-01 -1.52365759e-01 -5.01490355e-01
2.48408318e-01 3.48464847e-01 5.98123789e-01 -2.56462395e-01
8.83262992e-01 8.41741860e-01 -1.50607258e-01 -4.78782237e-01
-9.94060755e-01 -7.58757174e-01 -7.30652392e-01 -4.09011960e-01
9.65121686e-01 -9.26862657e-01 -4.57046211e-01 5.41334212e-01
-7.27157354e-01 -5.88497043e-01 -4.15920913e-01 7.13211060e-01
-7.84262121e-01 6.86494291e-01 -6.82619452e-01 -5.13620496e-01
-4.29927632e-02 -1.02391350e+00 1.06901288e+00 1.04309425e-01
-2.73719907e-01 -9.39609706e-01 -3.35546173e-02 3.19730282e-01
1.42700851e-01 4.10035253e-01 4.63381886e-01 -4.02820885e-01
-5.51067770e-01 -3.43014114e-02 -1.88011572e-01 3.38045150e-01
4.99889791e-01 -2.18251765e-01 -6.12529695e-01 -3.96345228e-01
4.95439060e-02 -2.83927321e-01 9.94645238e-01 4.39206392e-01
1.22352135e+00 -5.38017213e-01 -3.49476576e-01 6.48351252e-01
1.08141613e+00 1.24883845e-01 3.03194553e-01 5.20458341e-01
8.64652932e-01 3.92478675e-01 1.19061661e+00 4.12556410e-01
2.24384606e-01 8.76966417e-01 4.37092543e-01 1.07228030e-02
-2.41868258e-01 -4.33557063e-01 9.90809560e-01 6.80485606e-01
-2.81872451e-01 5.38156703e-02 -5.55956542e-01 6.67290092e-01
-2.39606690e+00 -1.41901922e+00 9.13949758e-02 1.88233364e+00
6.74321055e-01 3.34806144e-01 2.65718341e-01 1.79558530e-01
7.80377269e-01 7.42035508e-01 -7.90027916e-01 1.82027057e-01
-8.44470039e-02 -3.43182147e-01 4.12371397e-01 6.85926601e-02
-1.47771823e+00 8.78597856e-01 5.68849993e+00 9.33455586e-01
-1.04276824e+00 3.04993838e-01 3.48813534e-01 -4.61216509e-01
1.51953831e-01 6.20858371e-02 -4.99213248e-01 9.55014825e-01
6.85000718e-01 1.89791948e-01 1.34881958e-01 1.00823474e+00
9.36472476e-01 -1.90424711e-01 -1.17905927e+00 9.60570037e-01
1.16610583e-02 -1.23569441e+00 -9.04666334e-02 -2.13796914e-01
7.46696174e-01 -6.35213479e-02 -1.93858668e-02 3.59020144e-01
2.35723145e-02 -7.07819402e-01 7.45915771e-01 4.45707798e-01
6.68468356e-01 -6.12548232e-01 7.29863346e-01 5.72023869e-01
-1.67744505e+00 -4.59278613e-01 -1.20887287e-01 -2.25084379e-01
3.78110200e-01 3.17074925e-01 -4.36144024e-01 3.98053646e-01
8.70076537e-01 1.52731419e+00 -4.33084309e-01 6.48943007e-01
-2.44930506e-01 5.57365417e-01 -4.82601346e-03 3.61442864e-01
5.84577560e-01 -1.00081719e-01 7.23667681e-01 1.19314468e+00
1.64815262e-01 3.09702731e-03 5.75157642e-01 3.78504187e-01
1.06757224e-01 -1.28985927e-01 -6.40674233e-01 -7.94988871e-02
1.82669684e-01 1.10736203e+00 -8.72203648e-01 -3.16782415e-01
-7.44226396e-01 1.28797030e+00 4.78638001e-02 5.59127629e-01
-1.04045093e+00 -7.45367706e-02 6.70019329e-01 9.17373449e-02
4.22819614e-01 -2.22134039e-01 2.78356299e-02 -1.39379084e+00
3.12013924e-01 -8.69693696e-01 6.36911273e-01 -7.73162723e-01
-9.60373223e-01 2.40222469e-01 1.48872528e-02 -1.77501309e+00
-2.37147748e-01 -2.08124787e-01 -6.94965303e-01 2.72165418e-01
-1.47037947e+00 -1.06385410e+00 -4.51131552e-01 8.69829834e-01
1.20195198e+00 -6.82613850e-02 3.11010331e-01 2.38935456e-01
-7.39026725e-01 3.72915983e-01 -1.62630588e-01 2.21778095e-01
7.83602357e-01 -1.14140785e+00 3.71703729e-02 1.08166957e+00
8.65379423e-02 4.30771947e-01 5.23417354e-01 -7.42932498e-01
-1.41481400e+00 -1.43955600e+00 6.04492903e-01 -4.07626778e-01
8.34171474e-01 -1.95881352e-01 -8.05799186e-01 6.98258638e-01
-9.67896506e-02 5.39398909e-01 3.42662215e-01 -3.62768918e-01
-2.77470022e-01 -1.01436429e-01 -8.35878193e-01 5.69359064e-01
1.41653442e+00 -5.69676161e-01 -6.86042190e-01 3.26390088e-01
7.92958319e-01 -2.70452172e-01 -6.34077847e-01 5.45369446e-01
5.08993864e-01 -1.04574776e+00 8.83965611e-01 -5.68939388e-01
5.79231024e-01 -6.85531080e-01 3.16825435e-02 -9.20625031e-01
-3.94379228e-01 -9.54109967e-01 -8.87184918e-01 8.68164778e-01
-1.72194898e-01 -1.19746126e-01 9.07394648e-01 6.27426133e-02
-1.41395673e-01 -9.78050828e-01 -9.24957931e-01 -1.05888057e+00
-6.47674322e-01 -5.22961557e-01 2.67107576e-01 1.07771027e+00
-3.31797916e-03 2.24331114e-02 -7.65686929e-01 5.07572778e-02
2.92131662e-01 2.00712577e-01 6.61969960e-01 -5.54141343e-01
-3.47634226e-01 -2.83881664e-01 -6.22301877e-01 -1.52846253e+00
1.52844369e-01 -5.28761923e-01 2.07283437e-01 -1.15193510e+00
4.83294070e-01 -3.59812267e-02 -6.33142054e-01 7.72807956e-01
-2.63998300e-01 3.39037538e-01 9.27911252e-02 4.83356684e-01
-1.10977399e+00 6.03007376e-01 1.42085648e+00 -9.46795419e-02
-3.69075179e-01 3.82341444e-02 -2.57726848e-01 1.16753757e+00
5.71281910e-01 -5.25235772e-01 -8.12879264e-01 -3.52678686e-01
-1.88919887e-01 -7.34023459e-04 3.32790643e-01 -1.11137664e+00
2.87111074e-01 -4.74433213e-01 3.26379150e-01 -5.76485217e-01
1.35053024e-01 -7.07281649e-01 -9.60531682e-02 4.74925309e-01
-3.99294138e-01 1.38452590e-01 -1.44200981e-01 9.99890625e-01
-3.50046873e-01 1.11685775e-01 7.00083196e-01 -1.19424812e-01
-1.26233029e+00 5.58311462e-01 -3.47488046e-01 3.88298593e-02
1.48496306e+00 -5.42868614e-01 -9.92134511e-02 -4.82059240e-01
-6.68520749e-01 2.81865865e-01 5.57782888e-01 8.26150000e-01
7.53224790e-01 -1.46048844e+00 -3.25201184e-01 8.67263749e-02
2.15020299e-01 -5.04265837e-02 4.31686580e-01 1.45097220e+00
-2.14800149e-01 2.83139706e-01 -1.71829373e-01 -7.22235084e-01
-1.26488876e+00 7.32890546e-01 1.98797032e-01 -1.89816982e-01
-1.07381260e+00 6.76799119e-01 3.61850739e-01 2.03432202e-01
4.04470235e-01 -5.07576287e-01 -2.70152271e-01 -5.80801293e-02
8.17106128e-01 4.12995815e-01 -5.10576189e-01 -9.19380188e-01
-4.41766560e-01 5.56645155e-01 -5.14252745e-02 1.82451054e-01
1.34067643e+00 -3.67246538e-01 2.33850703e-01 5.99618137e-01
1.12296820e+00 -2.76310772e-01 -2.01274300e+00 -4.24997956e-01
8.72809365e-02 -7.14502990e-01 -2.66710967e-02 -3.90770048e-01
-1.31049633e+00 6.27232194e-01 5.06507635e-01 -8.73444155e-02
1.42086136e+00 -6.65157987e-03 1.11419797e+00 2.83570707e-01
3.72213840e-01 -1.33705795e+00 6.18094444e-01 3.11185628e-01
5.29384911e-01 -1.35648286e+00 -1.04585230e-01 -2.42321879e-01
-8.70183647e-01 1.00928235e+00 8.22757006e-01 2.66989730e-02
4.11384672e-01 2.06813321e-01 -7.67543763e-02 -1.47909746e-01
-9.53316867e-01 -1.38824999e-01 2.71325946e-01 5.94994545e-01
6.43837750e-02 -3.22754711e-01 -2.26370066e-01 3.98192763e-01
5.82806647e-01 1.89018101e-01 4.35986310e-01 1.18188035e+00
-3.71345162e-01 -8.35227668e-01 -1.60840064e-01 3.15715343e-01
-4.71934587e-01 2.06282571e-01 -1.82154372e-01 7.68810272e-01
5.45117676e-01 9.65192735e-01 1.72421932e-02 -5.54209173e-01
1.55329272e-01 -1.54840365e-01 4.74741980e-02 -5.95695078e-01
-1.11316048e-01 2.02895522e-01 -3.06469202e-02 -1.43907905e+00
-9.34172332e-01 -7.93848395e-01 -1.40429115e+00 -1.00278765e-01
1.47864819e-02 -1.21804178e-01 2.61137565e-03 1.03719127e+00
3.05216938e-01 4.37678695e-01 7.51069903e-01 -7.88160145e-01
-4.63223755e-01 -9.70914245e-01 -6.15621448e-01 8.10008705e-01
6.29604042e-01 -7.87716448e-01 -3.20693612e-01 3.72275472e-01]
|
[8.444452285766602, 0.6317089796066284]
|
4c1f777d-8989-4b30-af6d-2b8d6ee7e948
|
faithfulness-in-natural-language-generation-a
|
2203.05227
| null |
https://arxiv.org/abs/2203.05227v1
|
https://arxiv.org/pdf/2203.05227v1.pdf
|
Faithfulness in Natural Language Generation: A Systematic Survey of Analysis, Evaluation and Optimization Methods
|
Natural Language Generation (NLG) has made great progress in recent years due to the development of deep learning techniques such as pre-trained language models. This advancement has resulted in more fluent, coherent and even properties controllable (e.g. stylistic, sentiment, length etc.) generation, naturally leading to development in downstream tasks such as abstractive summarization, dialogue generation, machine translation, and data-to-text generation. However, the faithfulness problem that the generated text usually contains unfaithful or non-factual information has become the biggest challenge, which makes the performance of text generation unsatisfactory for practical applications in many real-world scenarios. Many studies on analysis, evaluation, and optimization methods for faithfulness problems have been proposed for various tasks, but have not been organized, compared and discussed in a combined manner. In this survey, we provide a systematic overview of the research progress on the faithfulness problem of NLG, including problem analysis, evaluation metrics and optimization methods. We organize the evaluation and optimization methods for different tasks into a unified taxonomy to facilitate comparison and learning across tasks. Several research trends are discussed further.
|
['Hua Wu', 'Xinyan Xiao', 'Jiachen Liu', 'Moye Chen', 'Wenhao Wu', 'Wei Li']
|
2022-03-10
| null | null | null | null |
['data-to-text-generation']
|
['natural-language-processing']
|
[ 2.84072220e-01 4.59135711e-01 -2.80167282e-01 -3.29083920e-01
-8.03224802e-01 -4.32337582e-01 1.08949304e+00 3.59662503e-01
2.38266364e-02 1.23609865e+00 8.64886343e-01 2.06816673e-01
2.00500488e-01 -7.28811681e-01 -2.63966918e-01 -5.55925369e-01
2.62863964e-01 6.39020264e-01 -3.98765862e-01 -7.12597787e-01
4.98709053e-01 1.04750492e-01 -1.29446411e+00 3.00812870e-01
1.12528491e+00 7.34737039e-01 4.71861102e-02 6.78574264e-01
-3.31531167e-01 8.11741889e-01 -1.09930182e+00 -8.10167491e-01
-5.61576858e-02 -8.81016731e-01 -1.11174464e+00 1.59537420e-01
2.73152113e-01 -6.99283928e-02 -1.91598199e-02 1.03519320e+00
8.23626339e-01 2.79812425e-01 8.50757003e-01 -1.21135998e+00
-8.76015484e-01 1.03620839e+00 -4.07189548e-01 -6.53372705e-02
5.49291253e-01 6.23951703e-02 1.17958796e+00 -6.75466359e-01
6.18384719e-01 1.33168840e+00 4.84399170e-01 9.60188448e-01
-9.26587522e-01 -1.62289366e-01 -3.68973315e-02 4.93594222e-02
-7.06117749e-01 -5.74208915e-01 7.80629754e-01 -1.26749992e-01
1.14185405e+00 3.46079171e-01 5.89461148e-01 1.45256674e+00
4.59521174e-01 9.37703609e-01 9.21416461e-01 -5.38879752e-01
1.88190341e-02 1.82453111e-01 -1.20642669e-01 5.46870828e-01
3.87996495e-01 -2.34782308e-01 -7.83788741e-01 1.42978638e-01
3.27225775e-01 -6.74519002e-01 -3.67586166e-01 2.24519953e-01
-1.33902895e+00 1.10646224e+00 2.65060097e-01 5.49977541e-01
-2.87462860e-01 -1.08915195e-01 6.53914452e-01 3.25039774e-01
6.97381914e-01 1.02801728e+00 -1.47883371e-01 -3.25488836e-01
-1.07756233e+00 7.58506179e-01 1.07390463e+00 1.01761949e+00
4.00682241e-01 6.16778970e-01 -6.46341622e-01 9.05354381e-01
-9.27958041e-02 3.22052240e-01 1.13940179e+00 -7.58773804e-01
7.48678625e-01 5.53934097e-01 1.01305835e-01 -1.13980007e+00
-4.52927917e-01 -2.15255350e-01 -1.61217809e+00 -2.41543949e-01
8.51973370e-02 -4.08889025e-01 -6.76364660e-01 1.79750752e+00
1.93592366e-02 -5.28828859e-01 3.45317721e-01 7.11875677e-01
1.16537130e+00 1.12590921e+00 -2.41990313e-01 -5.99865854e-01
1.04903007e+00 -1.06431663e+00 -1.01229513e+00 -2.47394994e-01
3.75901848e-01 -1.10106301e+00 1.16107178e+00 3.02879989e-01
-1.38595748e+00 -6.35160863e-01 -9.75174725e-01 -4.38093722e-01
-4.25124913e-01 -2.39328239e-02 6.51513577e-01 4.50309783e-01
-1.13096333e+00 7.97940731e-01 -4.29471761e-01 -2.45491132e-01
3.02539259e-01 7.69630224e-02 -1.70032293e-01 2.60299563e-01
-1.50230336e+00 1.20913696e+00 6.46989524e-01 4.38911542e-02
-6.43409967e-01 -4.27430332e-01 -9.21563387e-01 -1.51199907e-01
8.73841792e-02 -1.05149269e+00 1.46312594e+00 -9.88836527e-01
-1.90571225e+00 8.43272388e-01 -6.34834319e-02 -6.88747287e-01
5.73910713e-01 -2.58354068e-01 -2.13026315e-01 -3.27040851e-01
1.53687522e-01 8.41430128e-01 8.41406405e-01 -1.03542435e+00
-5.00280201e-01 -8.17394033e-02 -1.42423585e-02 5.35487592e-01
-3.05348933e-01 3.58523987e-02 1.76384494e-01 -9.42423284e-01
-3.69262815e-01 -7.21536875e-01 -3.17730486e-01 -8.62037778e-01
-8.53066683e-01 -6.00202978e-01 3.72289836e-01 -5.43558419e-01
1.44725204e+00 -1.43018413e+00 5.15813887e-01 -5.20371616e-01
4.15335000e-02 2.07989737e-01 -8.73332173e-02 8.67044091e-01
1.04413010e-01 3.64942580e-01 -3.51786733e-01 -5.85437179e-01
2.90484782e-02 -9.53912362e-03 -8.87342989e-01 -1.04483003e-02
3.38042945e-01 9.54487920e-01 -1.06573772e+00 -6.39057934e-01
6.19367436e-02 2.42111072e-01 -2.85987914e-01 4.90413874e-01
-6.02232456e-01 3.31404209e-01 -3.74221534e-01 3.10113668e-01
1.60065189e-01 -2.46824369e-01 -1.33932054e-01 -5.73016815e-02
-1.74837664e-01 5.90439320e-01 -6.20266616e-01 1.68379331e+00
-6.88481033e-01 8.46793354e-01 -4.41679180e-01 -9.26723421e-01
9.61180210e-01 5.80575824e-01 2.48587355e-01 -4.06647265e-01
8.18105489e-02 2.42947400e-01 -2.61697490e-02 -3.98407280e-01
1.25826478e+00 -4.38571990e-01 -4.70130175e-01 5.70902586e-01
4.82886881e-02 -8.97783577e-01 7.40971267e-01 4.06903803e-01
4.51769382e-01 -4.10191938e-02 6.61175787e-01 -1.88145712e-01
5.67488015e-01 2.01921940e-01 2.68469542e-01 5.83368540e-01
6.71363026e-02 7.72610784e-01 5.68589509e-01 -2.94694871e-01
-1.10954058e+00 -6.68223798e-01 1.46205679e-01 8.28588128e-01
-1.27536610e-01 -5.09290397e-01 -9.79625702e-01 -3.59561324e-01
-4.59598184e-01 1.02999735e+00 -5.49848676e-01 -3.28325272e-01
-7.53064513e-01 -1.14499915e+00 5.82257986e-01 2.83902079e-01
5.58395863e-01 -1.57981455e+00 -4.52247202e-01 4.00360197e-01
-7.26122916e-01 -9.93276060e-01 -4.96461242e-01 -2.55330294e-01
-1.09821773e+00 -4.34090734e-01 -8.63296270e-01 -6.80311501e-01
2.97884047e-01 3.28824669e-02 1.57349074e+00 -1.75659612e-01
3.25987600e-02 -4.21377346e-02 -4.94049728e-01 -7.69093156e-01
-8.39365005e-01 3.86492789e-01 2.25775801e-02 -1.52802393e-01
-5.99547438e-02 -3.90062839e-01 -3.27688575e-01 -2.84999222e-01
-9.63162124e-01 4.57678795e-01 4.36039478e-01 1.00619197e+00
3.40176493e-01 -1.47327468e-01 1.04460454e+00 -8.11864138e-01
1.55120957e+00 -3.20050657e-01 -2.59972334e-01 2.84230649e-01
-6.39904678e-01 3.09057027e-01 9.26105142e-01 -2.29422659e-01
-1.25381231e+00 -6.36341929e-01 -1.62349820e-01 2.84535825e-01
4.32410017e-02 6.56769633e-01 -1.56514019e-01 5.29615581e-01
8.58241796e-01 4.61768746e-01 -4.26192358e-02 -1.42225638e-01
6.69505835e-01 8.00069690e-01 3.95095557e-01 -4.67593282e-01
5.35562634e-01 7.73023516e-02 -1.29770294e-01 -9.12046432e-01
-1.31638503e+00 -5.86376041e-02 -3.05347919e-01 -9.44792181e-02
6.88432276e-01 -7.02101946e-01 -4.33468930e-02 5.66238225e-01
-1.51706493e+00 -2.99886018e-01 -4.83980685e-01 9.32145342e-02
-7.13788986e-01 4.13543582e-01 -6.92188144e-01 -6.31745458e-01
-1.21114981e+00 -9.47881103e-01 1.01480794e+00 4.68071342e-01
-7.59858549e-01 -1.23255825e+00 3.44039530e-01 4.80774611e-01
6.23865843e-01 5.27821124e-01 9.01448667e-01 -5.04235506e-01
-2.94069737e-01 -3.79736960e-01 8.61592218e-02 4.22388345e-01
2.53048152e-01 1.25028715e-01 -7.00850070e-01 -1.86015330e-02
1.52110875e-01 -6.76892579e-01 7.79017210e-01 4.41733181e-01
8.88001680e-01 -8.25124860e-01 2.55924668e-02 2.62269825e-01
1.05394876e+00 -1.56937853e-01 6.60030901e-01 4.78541814e-02
6.26225412e-01 8.27116013e-01 5.30109107e-01 4.78473186e-01
4.25870389e-01 5.72080970e-01 2.39199951e-01 1.04645535e-01
-2.47788727e-01 -3.77717763e-01 4.90250081e-01 1.24151468e+00
-2.90353417e-01 -8.20645154e-01 -4.66352135e-01 5.65793157e-01
-1.82779205e+00 -1.27732956e+00 -1.73058838e-01 1.94875693e+00
1.36683035e+00 -3.76804918e-03 1.42144142e-02 5.76535575e-02
6.24499500e-01 5.01935065e-01 -3.33426744e-01 -8.43379736e-01
-3.98162842e-01 1.58419862e-01 -1.33687466e-01 5.29256463e-01
-8.42953026e-01 1.17048442e+00 5.99794436e+00 9.75728631e-01
-1.22185183e+00 -1.49741814e-01 9.57139611e-01 -4.36450578e-02
-5.39802849e-01 -2.78515577e-01 -7.44772851e-01 4.03708220e-01
8.40422153e-01 -6.85859799e-01 2.62080371e-01 6.57099128e-01
5.56334138e-01 2.32074391e-02 -1.17397118e+00 9.55757260e-01
3.08528662e-01 -1.43473589e+00 6.28053725e-01 -2.82111585e-01
1.33723092e+00 -2.33102262e-01 1.29926251e-02 3.71961623e-01
2.38615930e-01 -1.13683808e+00 7.72754431e-01 3.47847581e-01
6.52400434e-01 -7.61830926e-01 8.35716009e-01 5.36939383e-01
-6.52823627e-01 2.22084761e-01 -5.33281803e-01 -1.78567976e-01
5.01350701e-01 9.15682137e-01 -8.34232032e-01 7.26858377e-01
2.59241581e-01 6.52354658e-01 -2.45313182e-01 5.81830680e-01
-4.38934118e-01 5.38183689e-01 1.20386310e-01 -5.95481575e-01
3.45265150e-01 -1.99204430e-01 7.16422081e-01 1.19888735e+00
3.27553093e-01 2.45282315e-02 1.19507490e-02 8.03640068e-01
-4.36301261e-01 4.15196687e-01 -5.19998729e-01 -3.84194970e-01
1.43929213e-01 1.32920396e+00 -4.22250390e-01 -6.24530375e-01
-5.19713014e-03 9.86122072e-01 3.58053863e-01 7.86773637e-02
-6.06281877e-01 -4.39599961e-01 3.75734836e-01 3.15153338e-02
-5.32522023e-01 -1.10846244e-01 -7.03329444e-01 -1.27679193e+00
5.69998659e-03 -1.18184602e+00 2.40764275e-01 -6.60159111e-01
-1.37870979e+00 9.41438854e-01 8.88385400e-02 -1.05737579e+00
-8.02306175e-01 -1.94263771e-01 -9.06719565e-01 9.08817172e-01
-1.24243736e+00 -8.57484698e-01 -2.29617089e-01 2.38828510e-01
1.04996252e+00 -4.60291594e-01 9.23643172e-01 -3.20825189e-01
-6.15189195e-01 3.37351561e-01 -9.70071927e-02 -1.69931114e-01
7.97204316e-01 -1.35193694e+00 6.73777580e-01 7.88413763e-01
8.86437669e-02 5.11956513e-01 9.34323072e-01 -4.80661035e-01
-9.75164175e-01 -1.01976657e+00 1.51831579e+00 -3.85728002e-01
5.35740435e-01 -1.08353898e-01 -6.11377835e-01 2.54841924e-01
9.60712731e-01 -9.43815172e-01 5.06584227e-01 -3.63201424e-02
2.99051739e-02 -9.83183384e-02 -7.80087709e-01 9.47760761e-01
6.78786516e-01 2.26054173e-02 -7.05901146e-01 6.57951772e-01
7.78800845e-01 -4.76466417e-01 -4.62127596e-01 2.86995411e-01
1.41632929e-01 -1.06952870e+00 5.41267931e-01 -7.06203759e-01
1.19640923e+00 7.54641071e-02 2.77594388e-01 -1.69704723e+00
-9.29744393e-02 -1.06315637e+00 -1.40721500e-01 1.26800358e+00
5.95698893e-01 -3.69573444e-01 5.08359313e-01 5.34879029e-01
-5.05423784e-01 -8.77300918e-01 -6.02989912e-01 -4.29152846e-01
3.26359570e-01 4.45391200e-02 5.02124250e-01 7.94098556e-01
3.10766131e-01 1.17471409e+00 -7.21857667e-01 -6.82926714e-01
2.98222870e-01 5.76337278e-01 7.96648681e-01 -9.63241696e-01
-6.64379969e-02 -9.56496477e-01 1.08355835e-01 -9.69767451e-01
1.69243455e-01 -8.93208444e-01 2.72992104e-01 -1.97444403e+00
2.44673744e-01 9.36370268e-02 4.26444203e-01 2.16239393e-01
-5.05239964e-01 2.54883636e-02 1.58426821e-01 1.98881656e-01
-4.11682010e-01 8.30060422e-01 1.58824790e+00 -1.64205492e-01
-3.17762047e-01 1.96255177e-01 -1.05793512e+00 4.77835923e-01
1.24747086e+00 -1.57921821e-01 -5.49784362e-01 -4.68369722e-01
6.01300836e-01 1.94619104e-01 -1.02460966e-01 -8.01139712e-01
-9.95900780e-02 -3.18719983e-01 1.31789610e-01 -4.37189907e-01
3.28933239e-01 1.95898628e-03 -1.80041730e-01 2.37497970e-01
-6.54621840e-01 3.48495811e-01 -7.59583339e-02 2.03131914e-01
-5.48643351e-01 -4.89159197e-01 7.53469825e-01 -4.46971655e-01
-1.84261173e-01 2.80725688e-01 -2.97622651e-01 5.35656393e-01
5.83846629e-01 2.32048929e-02 -2.08911791e-01 -8.99839163e-01
-1.50249869e-01 1.04199588e-01 1.74679577e-01 5.52977026e-01
5.70084214e-01 -1.17725372e+00 -1.41751337e+00 -3.51110965e-01
-6.05416372e-02 1.98326811e-01 2.69618630e-01 4.70776349e-01
-5.21268725e-01 5.78483224e-01 -1.67309120e-01 -1.70466572e-01
-8.80784214e-01 2.87153631e-01 1.64551109e-01 -8.50590646e-01
-2.95430243e-01 8.00285220e-01 -3.47610493e-03 -1.27784908e-01
-2.33966541e-02 -3.03848982e-01 -4.28394079e-01 2.37110510e-01
5.36493778e-01 4.10339624e-01 4.78092320e-02 -5.68744719e-01
1.45450249e-01 2.03480855e-01 -7.95224085e-02 -2.05181763e-01
1.03232658e+00 5.72680309e-02 -3.14673662e-01 4.25538361e-01
8.89821708e-01 -8.73989239e-02 -6.05755448e-01 9.48479995e-02
-6.15737326e-02 9.40867662e-02 -2.40725845e-01 -7.84878731e-01
-8.12363565e-01 1.02547991e+00 -2.63382196e-01 6.57487750e-01
9.35517669e-01 -1.99796200e-01 9.77923453e-01 4.22003090e-01
1.44376159e-01 -1.41187310e+00 4.18509156e-01 8.50245893e-01
1.57204187e+00 -1.19818199e+00 8.77089277e-02 -1.00413106e-01
-9.66263592e-01 1.17612839e+00 5.07456779e-01 4.62121367e-02
2.37724632e-01 5.83681390e-02 -2.94253305e-02 9.99450013e-02
-9.87971902e-01 6.23342283e-02 3.78631473e-01 6.02151215e-01
1.03328717e+00 1.13566272e-01 -6.86342716e-01 4.51431006e-01
-1.04868972e+00 -3.36295992e-01 8.44356298e-01 3.23111773e-01
-5.07331908e-01 -1.14013231e+00 -1.49169430e-01 5.13367772e-01
-6.30275190e-01 -2.92186558e-01 -8.04302514e-01 5.69546163e-01
-2.65476555e-01 1.22094274e+00 -1.65673912e-01 -1.69592500e-02
2.35834345e-01 -2.83943787e-02 5.54052770e-01 -8.83823395e-01
-7.66186237e-01 -1.35551438e-01 5.02411604e-01 -2.89782323e-03
-4.63056624e-01 -4.43976998e-01 -1.19889784e+00 -4.62036341e-01
-3.57701510e-01 3.97099674e-01 5.33595145e-01 8.23949993e-01
1.45964935e-01 4.23241198e-01 5.84110498e-01 -9.60438848e-01
-9.19385791e-01 -1.51352286e+00 -1.71688005e-01 3.96165460e-01
8.75541493e-02 -7.85018280e-02 -2.06469566e-01 2.81435519e-01]
|
[11.984149932861328, 9.187590599060059]
|
32b10fd1-eb17-4ceb-90da-8aa6c4cf6cf7
|
i-vise-interactive-video-surveillance-as-an
|
2003.04169
| null |
https://arxiv.org/abs/2003.04169v1
|
https://arxiv.org/pdf/2003.04169v1.pdf
|
I-ViSE: Interactive Video Surveillance as an Edge Service using Unsupervised Feature Queries
|
Situation AWareness (SAW) is essential for many mission critical applications. However, SAW is very challenging when trying to immediately identify objects of interest or zoom in on suspicious activities from thousands of video frames. This work aims at developing a queryable system to instantly select interesting content. While face recognition technology is mature, in many scenarios like public safety monitoring, the features of objects of interest may be much more complicated than face features. In addition, human operators may not be always able to provide a descriptive, simple, and accurate query. Actually, it is more often that there are only rough, general descriptions of certain suspicious objects or accidents. This paper proposes an Interactive Video Surveillance as an Edge service (I-ViSE) based on unsupervised feature queries. Adopting unsupervised methods that do not reveal any private information, the I-ViSE scheme utilizes general features of a human body and color of clothes. An I-ViSE prototype is built following the edge-fog computing paradigm and the experimental results verified the I-ViSE scheme meets the design goal of scene recognition in less than two seconds.
|
['Erik Blasch', 'Yu Chen', 'Seyed Yahya Nikouei', 'Alexander Aved']
|
2020-03-09
| null | null | null | null |
['scene-recognition']
|
['computer-vision']
|
[ 1.69441119e-01 -2.54610270e-01 8.87450799e-02 -6.15664124e-01
-9.90516245e-02 -2.05377400e-01 4.00514632e-01 6.74296692e-02
-1.71013981e-01 4.99542236e-01 -1.79880887e-01 -2.40669101e-02
-2.38823384e-01 -8.45555842e-01 -2.23098099e-01 -7.22961307e-01
-1.08194083e-01 1.02309436e-01 6.16181195e-01 -2.85231799e-01
2.73366779e-01 6.99228525e-01 -1.98923659e+00 3.65595341e-01
3.41495812e-01 1.53792405e+00 2.41922140e-01 5.75461864e-01
-1.26452819e-01 9.49964762e-01 -5.65808475e-01 -4.81975317e-01
3.50672394e-01 -3.65375906e-01 -3.54783356e-01 4.63048637e-01
2.94708610e-01 -6.94699645e-01 -2.54737943e-01 1.24777389e+00
4.26589072e-01 2.57953912e-01 1.68003842e-01 -1.67228842e+00
-4.39839840e-01 -1.04226433e-01 -2.94881761e-01 7.03365207e-01
6.16460443e-01 6.65384606e-02 4.59029257e-01 -7.54626811e-01
6.52842522e-01 8.12422812e-01 2.72260189e-01 5.54709613e-01
-2.43108898e-01 -3.80646020e-01 2.31262952e-01 7.63404787e-01
-1.61083484e+00 -6.28044486e-01 8.27389359e-01 -3.78793553e-02
6.25860274e-01 6.30132854e-01 7.93311596e-01 6.42229497e-01
3.12542588e-01 6.09984457e-01 8.53272676e-01 -1.94973186e-01
4.22762424e-01 7.65652478e-01 5.56227751e-02 9.65232909e-01
4.06848192e-01 -1.41147420e-01 -1.04389346e+00 -2.55401641e-01
4.86749262e-01 7.17333794e-01 -3.63416165e-01 -2.44002268e-01
-8.38275135e-01 4.43286240e-01 1.71887919e-01 3.82777750e-01
-6.77579999e-01 -2.39728585e-01 3.42797279e-01 4.18664187e-01
4.18309599e-01 -2.86712050e-01 6.84594065e-02 -3.20663422e-01
-9.63717401e-01 -3.02727342e-01 8.58520806e-01 1.20104361e+00
7.25682139e-01 8.03764537e-03 -1.26639232e-01 1.20712452e-01
1.20631501e-01 4.10674989e-01 1.58850074e-01 -8.15825701e-01
-2.35254034e-01 6.60719573e-01 1.38366401e-01 -1.48972452e+00
-8.44763592e-03 -1.02458544e-01 -4.82788354e-01 4.44218107e-02
-1.33187382e-03 -1.19233154e-01 -5.96921861e-01 1.03340352e+00
6.37881875e-01 3.12492013e-01 -4.80335020e-02 1.16790271e+00
1.11635506e+00 5.32670081e-01 -1.90731939e-02 -4.18964773e-01
1.71634817e+00 -5.52715123e-01 -8.18101823e-01 -2.06359521e-01
4.48251665e-02 -5.23601770e-01 6.43166006e-01 4.49177593e-01
-7.74353921e-01 -2.39761248e-01 -5.98998904e-01 4.61517483e-01
-6.27782822e-01 -8.64252597e-02 8.33528697e-01 1.00081587e+00
-1.19601083e+00 6.49555326e-02 -4.94934171e-01 -8.95696163e-01
4.90067929e-01 4.86838371e-01 -5.10375679e-01 -3.68114382e-01
-8.05475175e-01 4.16859627e-01 2.80946821e-01 2.40863621e-01
-1.00558150e+00 -2.25683481e-01 -7.23916292e-01 2.15321064e-01
7.85519361e-01 -4.55269277e-01 7.76745200e-01 -1.16679287e+00
-1.23328924e+00 9.06207383e-01 -4.46373671e-01 -2.92239487e-01
3.88975203e-01 5.76771125e-02 -1.03631330e+00 8.32639992e-01
-3.36850509e-02 1.51503548e-01 1.18621051e+00 -9.99302208e-01
-9.00939882e-01 -5.48965275e-01 2.57275343e-01 7.73353502e-02
-6.82885826e-01 5.05254865e-01 -6.07560217e-01 -4.66872565e-02
9.81476083e-02 -4.80089545e-01 -4.07543927e-02 3.79282057e-01
-1.19849943e-01 -2.99451239e-02 1.56685114e+00 -4.47210848e-01
1.07001650e+00 -2.27554297e+00 -8.69706929e-01 3.15047860e-01
2.27354988e-01 3.51226270e-01 3.11982036e-01 2.35480636e-01
3.11096162e-01 -2.91420579e-01 5.35855182e-02 6.94895983e-02
-3.54400039e-01 1.39982909e-01 -3.81831117e-02 5.09597242e-01
1.45213410e-01 5.16060829e-01 -1.08838224e+00 -8.13448131e-01
2.12542683e-01 5.76864183e-01 -5.00194967e-01 1.47714093e-01
9.67709571e-02 4.39297110e-01 -7.58528709e-01 1.28063750e+00
4.34884906e-01 -2.93348223e-01 -5.77118583e-02 -1.25761405e-01
6.88379034e-02 -2.81430066e-01 -1.22629833e+00 1.24341702e+00
-6.39887154e-02 7.96997547e-01 3.36233199e-01 -1.10110688e+00
8.44592869e-01 5.91643870e-01 8.65598798e-01 -4.56537455e-01
3.27845812e-01 -4.50322777e-02 -4.99498665e-01 -9.35790956e-01
5.82939029e-01 4.16965149e-02 1.42181590e-01 3.10088277e-01
-1.76901221e-01 3.25884074e-01 -1.88784453e-03 2.33923107e-01
1.28241301e+00 -2.24890292e-01 2.76333958e-01 -3.79645489e-02
5.38352489e-01 8.90318304e-02 7.07235634e-01 6.15367770e-01
-5.87620437e-01 4.13807899e-01 5.30217737e-02 -6.50940061e-01
-1.35647729e-01 -9.55985844e-01 1.32941931e-01 9.31844473e-01
6.24317825e-01 -4.42621171e-01 -7.03836262e-01 -7.33240604e-01
-3.61188501e-01 5.49988031e-01 -3.20023477e-01 -2.46447697e-01
1.20065324e-02 -4.57195640e-01 -7.61264116e-02 -9.80965197e-02
1.06689286e+00 -1.17002606e+00 -1.32276273e+00 7.32022598e-02
-1.32406875e-02 -1.24626172e+00 -5.55231690e-01 -4.38243151e-01
-5.60552001e-01 -1.10606980e+00 -3.29027742e-01 -7.43832171e-01
1.04142046e+00 8.82387817e-01 9.23889041e-01 2.87916005e-01
-6.61479115e-01 1.15518248e+00 -6.09886646e-01 -7.13096142e-01
8.98134112e-02 -6.12062156e-01 1.21198885e-01 8.58086824e-01
9.89355803e-01 -4.32096213e-01 -8.39913845e-01 3.67478758e-01
-9.39782798e-01 -5.35899282e-01 1.64282858e-01 3.47949803e-01
4.28831369e-01 4.85584646e-01 4.08532381e-01 -8.59217942e-01
4.53886122e-01 -6.89622343e-01 -3.79675388e-01 4.69329178e-01
-5.71753800e-01 -6.76860154e-01 2.81890839e-01 -2.85330594e-01
-1.17790234e+00 2.42400259e-01 3.09124023e-01 -6.55466557e-01
-3.65626544e-01 2.09813222e-01 -2.42552474e-01 -2.07097441e-01
4.16010946e-01 5.09776950e-01 -2.26817047e-03 -2.89048925e-02
-1.15677759e-01 1.05137181e+00 5.22668362e-01 -8.36468637e-02
4.71578270e-01 8.85055840e-01 -7.39976987e-02 -1.37828434e+00
-4.44524646e-01 -9.17177022e-01 -2.26789698e-01 -9.39843893e-01
6.90807700e-01 -7.45057225e-01 -9.90839899e-01 1.88568428e-01
-8.87359142e-01 3.42267573e-01 -3.34234715e-01 3.60467762e-01
-2.64777899e-01 4.83122736e-01 -6.00999668e-02 -1.22266603e+00
-3.99741441e-01 -8.30933988e-01 9.03685808e-01 5.15104115e-01
1.99521422e-01 -6.74601495e-01 -6.45418942e-01 3.61956239e-01
6.52546346e-01 1.92050636e-01 9.70674977e-02 -6.46528900e-01
-9.77711856e-01 -6.59737945e-01 -1.33128360e-01 1.12202102e-02
5.39488375e-01 -1.46488607e-01 -1.13478899e+00 -1.74021080e-01
4.86209065e-01 1.35310188e-01 3.31407696e-01 2.05581293e-01
1.00190485e+00 -4.96212304e-01 -3.49907815e-01 5.21267295e-01
1.45720863e+00 6.63709641e-01 7.82464683e-01 9.68551636e-02
2.52763897e-01 7.11926281e-01 9.21549499e-01 8.17597270e-01
3.19148093e-01 2.90574908e-01 7.02850342e-01 2.21288636e-01
3.44158679e-01 7.13975132e-02 6.27366602e-01 3.53628635e-01
-2.41555825e-01 -2.36061096e-01 -5.49513042e-01 6.07037246e-01
-1.62108088e+00 -1.26399302e+00 1.90049544e-01 2.09938502e+00
1.27568170e-01 -7.53274933e-02 -3.51608694e-02 -7.18421936e-02
9.57120419e-01 -1.73909299e-03 -6.81989670e-01 -3.10768545e-01
1.40123978e-01 -2.07092419e-01 3.90183985e-01 -1.68284521e-01
-9.31351244e-01 7.36903131e-01 5.75709581e+00 3.87601703e-01
-1.31268525e+00 3.92489672e-01 5.73024750e-01 -1.47465229e-01
-3.38691100e-02 -5.81264915e-03 -7.83943176e-01 5.77377439e-01
8.32289755e-01 -2.49578699e-01 4.52886969e-01 1.19291759e+00
2.42056146e-01 -4.76128370e-01 -7.51811266e-01 1.55950737e+00
4.80018407e-01 -1.17014325e+00 -5.27864844e-02 -7.61406049e-02
3.83860677e-01 -2.36011982e-01 -9.41789821e-02 -1.40068412e-01
-1.23518363e-01 -4.84392494e-01 6.96022987e-01 7.68278539e-01
7.77392387e-01 -4.89644974e-01 6.76475704e-01 3.71218532e-01
-1.17128468e+00 -3.06670099e-01 -4.16712612e-01 1.04640745e-01
1.64159507e-01 5.12085915e-01 -7.47797668e-01 4.35764045e-01
1.23043907e+00 3.96978766e-01 -4.98207778e-01 1.28572071e+00
2.71276772e-01 4.83994722e-01 -4.72153366e-01 -2.64590710e-01
5.85658476e-03 -2.32235640e-01 7.78111339e-01 9.25081670e-01
6.57410085e-01 5.94783962e-01 1.99293673e-01 4.17995632e-01
1.85851678e-01 2.31720224e-01 -1.01235580e+00 -5.85348271e-02
2.12916121e-01 1.39150393e+00 -1.08578694e+00 -3.04140061e-01
-6.66178226e-01 1.32793093e+00 -5.24701536e-01 2.55153567e-01
-7.09977686e-01 -4.56182837e-01 5.56973755e-01 4.36092079e-01
3.81704509e-01 2.30383333e-02 4.76673216e-01 -1.03982210e+00
1.00732580e-01 -5.98007500e-01 6.72353983e-01 -9.24893558e-01
-1.01654148e+00 8.26574922e-01 -6.80934191e-02 -1.40975237e+00
-1.33728057e-01 -3.97996575e-01 -8.32359135e-01 1.07526608e-01
-1.35813701e+00 -1.02670455e+00 -6.83974385e-01 1.24111998e+00
5.15501022e-01 -4.66548532e-01 7.03956544e-01 3.93905789e-01
-4.17165816e-01 3.45381856e-01 -1.14648990e-01 5.69863878e-02
4.73399639e-01 -5.16892731e-01 -3.80084306e-01 1.05061126e+00
2.83389390e-01 3.78986597e-01 6.68395460e-01 -6.86695814e-01
-1.68528140e+00 -9.78119373e-01 7.10847199e-01 -1.75301045e-01
2.64089406e-01 -1.21641144e-01 -5.93430400e-01 5.07709622e-01
1.24645412e-01 6.91865683e-01 5.67451119e-01 -6.01130486e-01
1.08719975e-01 -6.25680506e-01 -1.68763578e+00 4.76723790e-01
1.05065715e+00 -6.29675090e-01 -1.82129115e-01 5.41817367e-01
3.09760779e-01 6.51578903e-02 -4.78246480e-01 2.34906375e-02
2.23207802e-01 -1.37701869e+00 7.45907128e-01 -3.96522790e-01
-4.18920875e-01 -4.16506648e-01 -2.62181818e-01 -6.02674723e-01
1.09809555e-01 -9.43684042e-01 -1.69336088e-02 1.20683217e+00
-1.37300104e-01 -8.32271814e-01 1.12891448e+00 1.03104866e+00
4.46533933e-02 -4.07274574e-01 -1.09374809e+00 -7.30678260e-01
-1.39668024e+00 -4.30376828e-01 6.80085480e-01 7.76832581e-01
4.98510152e-03 -2.66772062e-01 -3.26860964e-01 5.08341432e-01
7.87802041e-01 1.46395892e-01 5.57707071e-01 -1.15786576e+00
-2.16031522e-02 1.10972859e-01 -1.13232768e+00 -4.67372179e-01
-3.38001549e-01 -3.64201307e-01 9.62850973e-02 -1.45729673e+00
1.15783930e-01 -1.92862600e-01 -5.23608148e-01 3.12429368e-01
1.34345487e-01 3.08825254e-01 1.35611609e-01 1.91358238e-01
-1.02252626e+00 2.85705090e-01 8.86379540e-01 -2.30445594e-01
-1.47803605e-01 2.84782499e-01 -5.66974282e-01 6.55853987e-01
8.44223142e-01 -4.43666279e-01 -6.57212794e-01 -7.59405568e-02
3.38370875e-02 1.81012258e-01 5.68172634e-01 -1.25712025e+00
8.76739323e-01 -3.15937281e-01 7.33531937e-02 -4.98807311e-01
6.22304559e-01 -1.38232005e+00 3.43662322e-01 3.01669270e-01
3.86934310e-01 -6.33683875e-02 -6.78557456e-02 8.26458275e-01
-3.37793380e-01 -3.29977214e-01 3.63258421e-01 -4.45247918e-01
-1.47992218e+00 5.65521121e-01 -4.76247758e-01 -4.36875910e-01
1.51735294e+00 -7.91116595e-01 -1.36244530e-02 -8.57946455e-01
-6.22905970e-01 -4.00651479e-03 4.26183313e-01 2.72831917e-01
1.09551132e+00 -9.97780025e-01 -3.28730553e-01 5.03036678e-01
3.71891230e-01 -2.74956524e-01 3.70126218e-01 7.42526352e-01
-5.13535559e-01 2.88611114e-01 -1.94136888e-01 -6.31168425e-01
-1.63713968e+00 8.07264984e-01 1.11324690e-01 4.91682768e-01
-6.93963468e-01 6.44802749e-01 -1.38894245e-01 2.88987309e-01
3.42627376e-01 2.76046574e-01 -2.86194801e-01 7.40389600e-02
9.47286189e-01 2.38579363e-01 5.69955073e-02 -9.21757460e-01
-5.19668043e-01 3.19006711e-01 1.42584845e-01 2.77337313e-01
1.42788637e+00 -5.09386778e-01 -2.46767953e-01 6.29974157e-02
8.77941728e-01 -7.88504630e-02 -1.22370923e+00 -3.53050649e-01
-7.34059885e-02 -9.14212763e-01 2.16424763e-01 -2.73844093e-01
-1.54900825e+00 5.84892869e-01 1.09638429e+00 5.53733051e-01
1.70758629e+00 -5.97866718e-03 7.69159794e-01 6.85960829e-01
1.02962339e+00 -1.19820476e+00 -9.26762074e-02 -1.47288218e-01
5.02668142e-01 -1.40570402e+00 1.34501923e-02 -6.22091770e-01
-7.53473341e-01 1.00585747e+00 4.13676172e-01 2.93729663e-01
1.08411527e+00 -2.50698533e-03 -8.72187093e-02 -6.99206591e-01
-6.08946800e-01 -4.26918983e-01 -4.72624488e-02 6.83127761e-01
-3.00526381e-01 -3.25244486e-01 1.75922543e-01 2.38860488e-01
3.37162882e-01 2.51991212e-01 3.29494089e-01 1.34978378e+00
-6.96830034e-01 -5.39295673e-01 -5.75974524e-01 6.00430846e-01
-6.87037289e-01 1.50725514e-01 -1.97935864e-01 4.67160076e-01
1.29659235e-01 1.35653877e+00 1.92395836e-01 -4.01173830e-01
1.24500573e-01 8.62297788e-02 5.86711355e-02 -4.89049435e-01
-3.58909816e-01 -3.50978047e-01 -9.50642973e-02 -7.72475779e-01
-6.35529578e-01 -6.10250950e-01 -1.10893619e+00 -3.26015055e-01
-2.84573317e-01 1.99392498e-01 8.01643312e-01 6.29993737e-01
4.93065685e-01 1.27683014e-01 8.34178984e-01 -4.28268790e-01
1.26864836e-01 -3.20948213e-01 -8.61055553e-01 3.93285692e-01
1.45994753e-01 -5.09942651e-01 -2.16999426e-01 4.77991700e-01]
|
[8.405241966247559, -1.0639246702194214]
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.