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
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
5e55ca74-9307-41f2-aba6-c56f13e23861
|
controllable-user-dialogue-act-augmentation
|
2207.12757
| null |
https://arxiv.org/abs/2207.12757v1
|
https://arxiv.org/pdf/2207.12757v1.pdf
|
Controllable User Dialogue Act Augmentation for Dialogue State Tracking
|
Prior work has demonstrated that data augmentation is useful for improving dialogue state tracking. However, there are many types of user utterances, while the prior method only considered the simplest one for augmentation, raising the concern about poor generalization capability. In order to better cover diverse dialogue acts and control the generation quality, this paper proposes controllable user dialogue act augmentation (CUDA-DST) to augment user utterances with diverse behaviors. With the augmented data, different state trackers gain improvement and show better robustness, achieving the state-of-the-art performance on MultiWOZ 2.1
|
['Yun-Nung Chen', 'Chao-Wei Huang', 'Ming-Hao Hsu', 'Chun-Mao Lai']
|
2022-07-26
| null |
https://aclanthology.org/2022.sigdial-1.5
|
https://aclanthology.org/2022.sigdial-1.5.pdf
|
sigdial-acl-2022-9
|
['dialogue-state-tracking']
|
['natural-language-processing']
|
[ 3.09270024e-02 5.22664070e-01 -3.99919122e-01 -2.76310295e-01
-4.73534346e-01 -4.31102455e-01 9.51341033e-01 -6.06680438e-02
-2.60419190e-01 9.35542166e-01 6.36092186e-01 -3.39652419e-01
4.64767992e-01 -3.58329326e-01 1.97368100e-01 -4.91421223e-01
7.98771381e-02 6.77866459e-01 1.18486114e-01 -1.20591569e+00
6.99566968e-04 2.33685970e-01 -1.06963706e+00 2.71512747e-01
1.10120904e+00 4.02030587e-01 -1.18197955e-01 9.37411308e-01
-2.16993809e-01 6.30578458e-01 -1.29091632e+00 -2.80041456e-01
2.75442302e-02 -6.96056545e-01 -1.09076941e+00 3.44556510e-01
2.41138652e-01 -6.73497796e-01 -7.97207117e-01 9.16012764e-01
6.94831014e-01 5.93668103e-01 4.49760437e-01 -1.25327742e+00
-4.82314318e-01 8.18366945e-01 -2.88716108e-01 2.74009585e-01
7.98407435e-01 3.73841107e-01 9.06472743e-01 -3.29893380e-01
4.28891391e-01 1.68125248e+00 2.13422671e-01 1.50555134e+00
-1.03673232e+00 -5.11974990e-01 4.22417998e-01 -2.56724149e-01
-6.89815104e-01 -7.59421170e-01 6.45559311e-01 2.30102148e-03
1.16165280e+00 5.78224361e-01 6.84322715e-01 1.55933344e+00
-3.53720903e-01 1.43595183e+00 1.03547275e+00 -5.93626678e-01
-1.63495421e-01 1.67513624e-01 6.18123591e-01 5.66690147e-01
-3.22737306e-01 -1.95842966e-01 -5.12826383e-01 -3.24948490e-01
8.01335692e-01 -3.99713278e-01 -2.39227116e-01 1.91257030e-01
-1.12523139e+00 1.07495534e+00 1.43942153e-02 3.13932657e-01
-2.20688581e-01 -3.26956600e-01 6.49333894e-01 2.85534918e-01
6.71432912e-01 9.68573749e-01 -6.00695968e-01 -9.73012269e-01
-1.20725222e-01 5.99275947e-01 8.64499986e-01 9.77062821e-01
2.02484593e-01 6.24032199e-01 -8.80660892e-01 1.20695901e+00
4.82692383e-02 4.16828305e-01 7.63914943e-01 -9.41755712e-01
7.01873899e-01 7.68833756e-01 1.96358994e-01 -2.20458522e-01
-6.07110858e-01 -1.00662179e-01 -9.03792918e-01 -1.07151046e-01
6.20172858e-01 -6.71410084e-01 -8.46442938e-01 2.01662779e+00
2.73657471e-01 -1.76996872e-01 4.07318681e-01 7.61430681e-01
8.22859943e-01 7.85018027e-01 1.44034505e-01 -2.97321737e-01
1.32825840e+00 -1.16185319e+00 -1.60874271e+00 -3.74471635e-01
9.14681196e-01 -5.73514760e-01 1.23053217e+00 3.88008267e-01
-1.20665407e+00 -5.23505628e-01 -8.78839612e-01 1.30962729e-01
-3.56907815e-01 -8.88206214e-02 1.04405153e+00 1.01888466e+00
-7.58566380e-01 3.29046994e-01 -8.29195440e-01 -3.95000100e-01
-1.23765633e-01 3.10854584e-01 -6.31241873e-02 5.37171900e-01
-1.82090664e+00 1.44126427e+00 4.94622380e-01 -2.00754896e-01
-4.35787112e-01 -2.41676033e-01 -9.27551389e-01 -9.55083668e-02
3.10412228e-01 -2.77917355e-01 1.98461413e+00 -9.90932360e-02
-2.36822510e+00 1.08416475e-01 -4.45125438e-02 -5.80677152e-01
4.13865179e-01 -4.32367027e-01 -4.41215187e-01 -3.27288389e-01
-4.98500675e-01 7.29380310e-01 4.52923119e-01 -9.77814496e-01
-5.40256739e-01 -2.00250298e-01 3.43724728e-01 6.69769287e-01
-8.25941622e-01 -1.48785785e-01 -3.78471494e-01 -5.45123994e-01
-1.49214372e-01 -9.74336147e-01 -3.45467895e-01 -9.71044660e-01
-5.74171007e-01 -7.59917140e-01 1.15501845e+00 -7.36698925e-01
1.66610587e+00 -1.68543053e+00 1.65929288e-01 -3.48078758e-01
1.11551449e-01 8.21005940e-01 -2.68612415e-01 5.79201341e-01
1.11634016e-01 1.19966261e-01 1.01576962e-01 -6.85341299e-01
9.67733935e-02 3.17189783e-01 -1.68397069e-01 1.19709961e-01
1.11012906e-01 8.18123341e-01 -1.03983736e+00 -3.40432748e-02
5.66046059e-01 1.52565360e-01 -5.69833815e-01 6.21128500e-01
-4.41711366e-01 6.43597603e-01 -4.22790796e-01 3.34311187e-01
1.76828533e-01 6.42344654e-02 -5.86267114e-02 1.04624450e-01
3.43924314e-02 7.05862820e-01 -9.82948005e-01 1.68238974e+00
-3.23217094e-01 6.75396919e-01 4.53686342e-02 -4.11567360e-01
1.02498579e+00 6.41955316e-01 4.75875914e-01 -3.81644934e-01
4.05486107e-01 -3.30826044e-01 5.85464478e-01 -3.95203382e-01
1.05766487e+00 3.93439829e-01 -3.09154421e-01 4.29242104e-01
9.66008455e-02 -4.00437653e-01 4.17030811e-01 3.60565454e-01
8.12613606e-01 -1.17009953e-01 3.99430901e-01 7.82053545e-02
3.70105624e-01 -1.17885418e-01 3.62619311e-01 7.28800476e-01
-6.33296549e-01 2.66694188e-01 4.55496609e-01 -5.96533678e-02
-9.92640555e-01 -6.12950146e-01 -1.67624447e-02 1.37152648e+00
-6.53271526e-02 -4.22928840e-01 -1.03622925e+00 -7.67668605e-01
-2.88542449e-01 1.00425255e+00 -4.89547342e-01 -4.30230439e-01
-5.22425830e-01 -7.95953870e-01 1.04090238e+00 4.18870717e-01
8.63791525e-01 -1.14555967e+00 -2.48317003e-01 4.09634978e-01
-5.81523776e-01 -1.09187043e+00 -7.27157176e-01 3.47579271e-02
-7.18197405e-01 -8.05909514e-01 -6.53949916e-01 -4.08379585e-01
2.91918039e-01 1.12367548e-01 6.72937274e-01 -1.44220069e-01
1.34472787e-01 2.37787619e-01 -6.23231053e-01 -4.53489006e-01
-1.22148883e+00 5.23735225e-01 4.81125772e-01 -2.49206707e-01
2.03099087e-01 5.42821325e-02 -8.23402256e-02 8.70806873e-02
-5.35328329e-01 1.98740080e-01 3.10808450e-01 1.22118437e+00
-5.21221399e-01 -4.15493965e-01 8.54836583e-01 -9.55764711e-01
1.61034918e+00 -1.55208334e-02 -1.26447409e-01 -2.15809606e-02
-6.64847076e-01 1.39636412e-01 3.67888749e-01 -8.19711924e-01
-1.50729668e+00 -1.62092209e-01 -4.90545541e-01 -8.51080492e-02
-4.92281139e-01 2.86329567e-01 -1.58097595e-01 1.98844805e-01
8.73538733e-01 2.61987984e-01 4.47231591e-01 -4.39491004e-01
6.06416941e-01 1.02460313e+00 5.09626091e-01 -4.49916899e-01
3.61952901e-01 -2.70733267e-01 -5.54805815e-01 -1.21838140e+00
-6.20786309e-01 -6.05642617e-01 -6.11395419e-01 -1.87954053e-01
7.67364323e-01 -6.17446899e-01 -7.63569415e-01 6.05771303e-01
-1.47482848e+00 -5.77009439e-01 -1.21473469e-01 3.83231372e-01
-4.64503348e-01 5.39487600e-01 -1.27031529e+00 -1.41199017e+00
-6.01843119e-01 -1.28757083e+00 6.96212649e-01 3.59713942e-01
-7.13829517e-01 -9.66549933e-01 3.58567268e-01 4.83656913e-01
5.04190743e-01 -7.15213045e-02 5.67022741e-01 -1.26939952e+00
1.25381574e-01 -3.68102312e-01 2.10146531e-01 3.20727676e-01
4.94110674e-01 -1.00647397e-01 -1.11564934e+00 -4.16057318e-01
-1.30703464e-01 -3.26791942e-01 2.51612723e-01 6.96324110e-02
7.49114871e-01 -6.72588110e-01 -9.39711928e-02 -8.89552012e-02
2.23033547e-01 7.45798886e-01 6.12922192e-01 1.10632382e-01
5.33062518e-01 6.82705283e-01 8.26463640e-01 6.60312116e-01
3.34001333e-01 8.78503501e-01 3.11026752e-01 -2.24133283e-01
1.95672605e-02 -9.59403217e-02 6.20972753e-01 1.03228831e+00
-7.18632042e-02 -6.20868623e-01 -6.47276223e-01 2.62356997e-01
-1.98569524e+00 -9.54644680e-01 -1.10222153e-01 1.83528864e+00
1.19315374e+00 3.03317785e-01 5.67375720e-01 1.43790811e-01
6.74163938e-01 4.85215932e-01 -4.55446154e-01 -5.25330782e-01
1.58268034e-01 -2.00951800e-01 2.12440073e-01 7.87786007e-01
-1.08660793e+00 1.31807220e+00 7.18940639e+00 6.62590086e-01
-6.30293548e-01 -3.21378466e-03 4.94430006e-01 1.59701228e-01
1.67017192e-01 -4.95291293e-01 -1.12449813e+00 2.88778126e-01
9.83425915e-01 -1.38947323e-01 3.49396676e-01 8.09075713e-01
4.40849066e-01 1.43441841e-01 -7.94190168e-01 6.35785937e-01
-1.00916773e-01 -9.65904951e-01 -7.71721452e-02 1.27586812e-01
7.26768315e-01 -2.86915183e-01 -4.75401320e-02 9.85012889e-01
6.64228141e-01 -5.41300416e-01 8.01642686e-02 9.71770659e-02
4.78077263e-01 -7.28099942e-01 7.78132439e-01 5.77477634e-01
-7.72808671e-01 2.47821763e-01 5.04819676e-02 -2.60868937e-01
3.25677693e-01 -2.48589605e-01 -1.48575616e+00 5.08540571e-02
6.69013858e-02 3.71972829e-01 -5.06556571e-01 7.45498538e-01
-1.48792624e-01 6.92184329e-01 -1.67747393e-01 -5.06239116e-01
3.55854779e-01 -2.18459383e-01 7.62282550e-01 1.41313779e+00
-1.74085617e-01 2.64152646e-01 4.83350635e-01 2.99310058e-01
2.43309587e-02 2.08000522e-02 -7.56645739e-01 -3.24582607e-01
7.23670065e-01 1.06852341e+00 -1.22132406e-01 -5.76302946e-01
-2.39586815e-01 1.12324345e+00 2.31658101e-01 2.96594143e-01
-6.11909389e-01 -2.52295613e-01 1.08441615e+00 -3.17945421e-01
-6.34143770e-01 -3.55772406e-01 -4.79612827e-01 -1.13438201e+00
-6.22597456e-01 -1.28001535e+00 3.08970988e-01 -3.25767577e-01
-9.97150600e-01 7.06055939e-01 2.02144623e-01 -1.02581239e+00
-8.03512454e-01 -4.34543520e-01 -5.32947659e-01 9.83510196e-01
-8.02582920e-01 -1.01050520e+00 -2.44642377e-01 4.38368231e-01
1.32683158e+00 -4.78073120e-01 1.20217192e+00 -7.79403374e-03
-8.38693202e-01 9.90516365e-01 -1.21833205e-01 2.43306354e-01
8.46698999e-01 -1.61051559e+00 7.91886926e-01 7.78907835e-01
-2.56422818e-01 7.24190891e-01 9.38114345e-01 -6.94137990e-01
-1.17927873e+00 -6.77619755e-01 4.83280063e-01 -6.32223964e-01
5.56428313e-01 -4.33221161e-01 -1.08544743e+00 6.63039625e-01
7.76003242e-01 -8.43999028e-01 5.29766023e-01 3.20107311e-01
5.04930764e-02 5.61675370e-01 -1.05767560e+00 1.24001515e+00
7.20479250e-01 -3.82062703e-01 -7.12995946e-01 2.26093948e-01
9.63579416e-01 -9.61006105e-01 -1.09680510e+00 4.06008065e-01
9.57609415e-02 -5.48618257e-01 7.08489954e-01 -9.12261128e-01
-1.03309616e-01 1.10775292e-01 2.78918833e-01 -1.83218050e+00
-2.06040427e-01 -1.12717342e+00 -7.16851652e-01 1.32846045e+00
5.37871718e-01 -6.43177509e-01 9.66286361e-01 9.86946166e-01
-4.86440748e-01 -5.48565745e-01 -6.33269846e-01 -6.78100348e-01
-6.24017827e-02 -6.92194924e-02 5.17289877e-01 1.06993699e+00
8.78510296e-01 1.00419855e+00 -8.50825310e-01 -1.65325642e-01
1.11449726e-01 -5.97820044e-01 1.03930390e+00 -9.76950049e-01
-6.15936108e-02 -7.15011835e-01 2.09322333e-01 -1.66573870e+00
9.26644579e-02 -2.98702389e-01 3.63546252e-01 -1.36148667e+00
-2.90561110e-01 -2.11124614e-01 3.02050024e-01 6.43300653e-01
-8.21573615e-01 -3.33927900e-01 3.29422712e-01 -1.09359562e-01
-4.92341518e-01 9.43103731e-01 1.58009243e+00 -2.54660547e-01
-7.26718247e-01 3.66926134e-01 -5.64237833e-01 5.13022006e-01
1.14268923e+00 1.94682688e-01 -5.96479237e-01 1.74399558e-02
-6.71427548e-01 4.95937765e-01 -4.80233371e-01 -7.33243763e-01
1.36882793e-02 -2.05103725e-01 -1.08953543e-01 -7.06954658e-01
8.68910968e-01 -2.89953649e-01 -7.10586667e-01 6.17862046e-01
-8.36956561e-01 1.07580170e-01 4.79692489e-01 3.53753179e-01
-9.83520523e-02 -2.93698788e-01 7.64179707e-01 -8.01463276e-02
-4.99059230e-01 2.90550977e-01 -9.49654996e-01 1.89025793e-02
6.60197914e-01 1.18248753e-01 -5.44595182e-01 -9.97834206e-01
-7.33240485e-01 5.10178328e-01 -1.17089497e-02 7.35837698e-01
4.29848731e-01 -1.26646459e+00 -7.25811362e-01 1.72886208e-01
6.87376484e-02 -1.67898774e-01 9.45981871e-03 3.66515130e-01
-1.05209857e-01 6.22340381e-01 -6.77024126e-02 -3.65860820e-01
-1.45901513e+00 2.15603858e-01 2.37616748e-01 -5.79584777e-01
-4.13230240e-01 6.82841122e-01 -2.36683443e-01 -6.77320302e-01
6.04412973e-01 -2.26118028e-01 -7.58503854e-01 -1.42593924e-02
5.70997179e-01 6.07279181e-01 -1.39919862e-01 -3.65318567e-01
2.96395779e-01 -4.67806697e-01 -5.23985982e-01 -3.79673779e-01
7.04430699e-01 -2.46278763e-01 4.17404473e-01 5.29134750e-01
6.68521583e-01 -1.65346205e-01 -1.27917755e+00 -2.81919152e-01
2.59958897e-02 -3.17297190e-01 -1.91925749e-01 -9.72918510e-01
-4.34643209e-01 7.42736578e-01 5.45630813e-01 9.53257442e-01
5.29937804e-01 -4.18242157e-01 9.40985441e-01 6.57459617e-01
3.10787559e-01 -1.20155942e+00 3.92012566e-01 1.29037440e+00
9.61550832e-01 -1.44542575e+00 -3.18686008e-01 -3.78482819e-01
-1.30990624e+00 1.02883661e+00 1.38736677e+00 3.63041222e-01
-1.53405778e-02 9.10520330e-02 3.39195848e-01 4.00277860e-02
-9.29835439e-01 -3.38064402e-01 1.62728623e-01 6.75783157e-01
7.94942737e-01 3.99523154e-02 -3.15774262e-01 3.03029239e-01
-2.74066389e-01 -6.09347165e-01 7.48295963e-01 6.82406962e-01
-4.76069272e-01 -1.41908205e+00 -3.17270100e-01 5.67487955e-01
-2.86496490e-01 7.80510381e-02 -6.91168129e-01 9.91140008e-01
-7.05376208e-01 1.39454114e+00 -1.36070505e-01 -6.21751964e-01
6.59641147e-01 4.72210079e-01 3.43129247e-01 -8.90917897e-01
-9.78222489e-01 2.11251155e-01 7.53214955e-01 -2.75480509e-01
-8.68656710e-02 -5.54054976e-01 -1.22395802e+00 -3.47866446e-01
-6.96894348e-01 4.93605077e-01 4.12663966e-01 8.56683850e-01
-3.82442051e-03 8.56532574e-01 7.30953097e-01 -7.37498760e-01
-9.14355814e-01 -2.06134701e+00 -2.73464739e-01 6.17159903e-01
3.72368753e-01 -5.81984758e-01 -2.26149917e-01 -3.31268519e-01]
|
[12.848149299621582, 7.906217098236084]
|
33abd778-ba59-47aa-8ab1-2f21cb25774c
|
a-high-accuracy-unsupervised-person-re
|
2205.03124
| null |
https://arxiv.org/abs/2205.03124v1
|
https://arxiv.org/pdf/2205.03124v1.pdf
|
A High-Accuracy Unsupervised Person Re-identification Method Using Auxiliary Information Mined from Datasets
|
Supervised person re-identification methods rely heavily on high-quality cross-camera training label. This significantly hinders the deployment of re-ID models in real-world applications. The unsupervised person re-ID methods can reduce the cost of data annotation, but their performance is still far lower than the supervised ones. In this paper, we make full use of the auxiliary information mined from the datasets for multi-modal feature learning, including camera information, temporal information and spatial information. By analyzing the style bias of cameras, the characteristics of pedestrians' motion trajectories and the positions of camera network, this paper designs three modules: Time-Overlapping Constraint (TOC), Spatio-Temporal Similarity (STS) and Same-Camera Penalty (SCP) to exploit the auxiliary information. Auxiliary information can improve the model performance and inference accuracy by constructing association constraints or fusing with visual features. In addition, this paper proposes three effective training tricks, including Restricted Label Smoothing Cross Entropy Loss (RLSCE), Weight Adaptive Triplet Loss (WATL) and Dynamic Training Iterations (DTI). The tricks achieve mAP of 72.4% and 81.1% on MARS and DukeMTMC-VideoReID, respectively. Combined with auxiliary information exploiting modules, our methods achieve mAP of 89.9% on DukeMTMC, where TOC, STS and SCP all contributed considerable performance improvements. The method proposed by this paper outperforms most existing unsupervised re-ID methods and narrows the gap between unsupervised and supervised re-ID methods. Our code is at https://github.com/tenghehan/AuxUSLReID.
|
['Guiguang Ding', 'Yuchen Guo', 'Tao He', 'Hehan Teng']
|
2022-05-06
| null | null | null | null |
['unsupervised-person-re-identification']
|
['computer-vision']
|
[-2.30484217e-01 -4.11567122e-01 -2.75202155e-01 -4.58572656e-01
-4.85218108e-01 -5.25531709e-01 7.65811741e-01 -6.39278069e-02
-7.43866265e-01 7.23340392e-01 2.91476756e-01 2.21484512e-01
-2.01474298e-02 -4.01023030e-01 -4.68214154e-01 -6.71786249e-01
8.92890617e-03 2.76425511e-01 2.63945848e-01 9.48499888e-02
1.64934248e-01 2.74257511e-01 -1.75879145e+00 7.25045651e-02
9.49150443e-01 6.74214005e-01 2.31297642e-01 5.44963956e-01
2.35986784e-01 4.84748036e-01 -2.83992827e-01 -7.79961586e-01
2.81676203e-01 -1.57769457e-01 -6.61541224e-01 1.59985274e-02
4.16092962e-01 -3.15435052e-01 -4.98789042e-01 9.41881955e-01
7.58127928e-01 3.87786239e-01 6.14556313e-01 -1.53233695e+00
-3.31850618e-01 1.66330725e-01 -7.90875494e-01 3.20514351e-01
4.13309544e-01 5.26316725e-02 6.45331502e-01 -9.04024839e-01
4.11365837e-01 1.10716653e+00 1.08162582e+00 5.29250026e-01
-9.88546252e-01 -7.79028714e-01 3.20334285e-01 6.43378854e-01
-1.61647069e+00 -5.57693720e-01 6.24432981e-01 -5.77509284e-01
7.92686105e-01 3.46191347e-01 5.96270919e-01 1.25197768e+00
-3.65243703e-01 7.55418658e-01 1.05058193e+00 -4.21033710e-01
-9.63292122e-02 5.47607720e-01 3.17887813e-01 6.95580006e-01
1.80824548e-01 2.07737327e-01 -4.59664881e-01 -1.04241155e-01
5.56098580e-01 2.82387167e-01 -9.39089730e-02 -1.24134049e-01
-1.17330611e+00 5.50988495e-01 2.60486484e-01 2.00284868e-01
9.78945941e-02 -1.84315026e-01 5.24246216e-01 -1.35281503e-01
5.35055161e-01 -2.13362593e-02 -2.65216142e-01 -2.92737007e-01
-9.14861441e-01 1.24961510e-01 4.35445040e-01 1.32557809e+00
8.11117530e-01 -2.54585922e-01 -2.02552319e-01 1.13530421e+00
2.67832369e-01 6.94284678e-01 4.44789469e-01 -9.00851130e-01
6.87296093e-01 6.70788288e-01 1.38911352e-01 -1.15002191e+00
-5.36795080e-01 -2.48020709e-01 -9.45515573e-01 -4.34374362e-01
4.15898323e-01 -3.09509069e-01 -5.82136452e-01 1.64909303e+00
3.37484509e-01 4.83216107e-01 -3.05562645e-01 9.64880764e-01
7.20131636e-01 4.02594209e-01 2.23924071e-01 -5.54850325e-02
1.38922405e+00 -1.03338587e+00 -5.95136762e-01 -1.61528438e-01
7.26338923e-01 -4.54680443e-01 8.68343353e-01 1.61059469e-01
-7.99863160e-01 -8.18696737e-01 -8.34553003e-01 7.12347254e-02
-5.43401182e-01 6.73971891e-01 5.28057933e-01 7.66874552e-01
-8.59363496e-01 3.53347778e-01 -6.83251500e-01 -7.94434309e-01
3.32197487e-01 4.54158485e-01 -5.65154672e-01 -1.78904548e-01
-1.09702909e+00 7.46002495e-01 3.50489408e-01 1.40449181e-01
-5.53457737e-01 -5.51230192e-01 -7.22248971e-01 -1.88013017e-01
2.67931193e-01 -5.30299008e-01 8.30451608e-01 -8.76919210e-01
-1.31549311e+00 9.68832612e-01 -2.96703517e-01 -3.11794609e-01
6.43132746e-01 -2.14274526e-01 -6.79916561e-01 8.94590393e-02
2.63447702e-01 6.53637767e-01 5.75254500e-01 -1.14196515e+00
-9.40131962e-01 -4.63938326e-01 -1.48861900e-01 2.42463022e-01
-6.57089770e-01 1.38600260e-01 -1.00213110e+00 -6.79579616e-01
-3.32628429e-01 -1.23395121e+00 4.50423621e-02 -2.81355590e-01
-4.20095086e-01 -3.34815592e-01 6.83293760e-01 -9.30345476e-01
1.27536523e+00 -2.12058306e+00 1.53353900e-01 1.81059256e-01
8.54562819e-02 4.71799105e-01 2.75519863e-02 2.41664916e-01
2.95778159e-02 2.88440417e-02 -8.48835707e-02 -7.87514925e-01
5.86180463e-02 8.95379335e-02 2.12793201e-01 6.14653885e-01
-2.25277707e-01 6.53342009e-01 -7.62611032e-01 -7.69610405e-01
5.22325635e-01 4.43072051e-01 -4.88488764e-01 1.45028278e-01
3.90858561e-01 5.72595477e-01 -1.90069124e-01 6.84018075e-01
7.19388068e-01 -5.00316508e-02 7.68207461e-02 -4.70689416e-01
-2.93408275e-01 -1.00028858e-01 -1.37247694e+00 1.60708058e+00
-1.76559716e-01 5.18941104e-01 -1.72365904e-01 -9.08734500e-01
6.83440268e-01 1.28231630e-01 5.71117699e-01 -7.07476199e-01
6.81665540e-02 -1.94696292e-01 -5.85320771e-01 -5.40124238e-01
4.77242440e-01 3.78259182e-01 -5.38551286e-02 1.83919489e-01
5.11153191e-02 8.39774609e-01 3.03490579e-01 2.01803207e-01
7.13007510e-01 1.09238103e-01 1.16050228e-01 -2.51518607e-01
7.82042384e-01 -2.47871608e-01 8.79071295e-01 6.77880108e-01
-5.80595374e-01 5.60853124e-01 3.84051315e-02 -3.07820469e-01
-1.18202293e+00 -9.23550904e-01 -1.93346560e-01 1.08349514e+00
4.06441897e-01 -6.50421798e-01 -8.11584592e-01 -7.81101406e-01
5.03923111e-02 4.34525669e-01 -4.97649223e-01 -1.57693684e-01
-5.46443939e-01 -1.04753709e+00 8.31369698e-01 6.33208513e-01
1.09059536e+00 -3.73744398e-01 -2.25960966e-02 -1.16021134e-01
-5.93088150e-01 -1.25027800e+00 -8.20522487e-01 -3.20789427e-01
-6.35350168e-01 -1.13542807e+00 -8.55355263e-01 -5.63530922e-01
7.60303140e-01 6.68552518e-01 8.31834912e-01 6.45124987e-02
-3.57839048e-01 7.57843375e-01 -5.33530176e-01 -1.41348898e-01
2.33684659e-01 1.36889026e-01 4.90210146e-01 2.79148668e-01
7.36563802e-01 -4.28801298e-01 -6.64097726e-01 7.61642158e-01
-2.99147427e-01 1.58680424e-01 3.59219849e-01 7.57338047e-01
4.91714627e-01 -5.30985650e-03 2.70052344e-01 -5.90577483e-01
3.22095118e-02 -4.86054599e-01 -4.68540341e-01 3.72652709e-01
-7.79762566e-01 -2.50420421e-01 3.61380130e-01 -5.40232122e-01
-1.36656463e+00 1.33690402e-01 9.68828946e-02 -3.01810116e-01
-3.23267102e-01 1.12851048e-02 -4.14538652e-01 3.36147062e-02
3.27760011e-01 1.72838807e-01 -2.30964243e-01 -6.93012834e-01
3.58176604e-02 8.52169931e-01 5.27636886e-01 -4.96351838e-01
7.50073850e-01 6.43221974e-01 -2.50937521e-01 -9.23122585e-01
-7.05926001e-01 -8.68511260e-01 -1.02000892e+00 -3.48423243e-01
9.81237590e-01 -1.27701235e+00 -8.09997201e-01 6.59420788e-01
-8.23998094e-01 -5.83966151e-02 3.21056470e-02 6.89843297e-01
-1.60148352e-01 7.32058287e-01 -4.91943628e-01 -9.59062159e-01
-6.62180632e-02 -8.14437568e-01 8.57731998e-01 3.44226837e-01
2.63266265e-02 -9.36123788e-01 -1.33677600e-02 6.72039688e-01
1.50083736e-01 6.89498708e-02 1.42177403e-01 -5.00970960e-01
-3.86552721e-01 -3.45199347e-01 -4.69134420e-01 2.50550926e-01
1.04326708e-02 -2.30258048e-01 -1.08992410e+00 -5.13058066e-01
-3.85473847e-01 4.17262800e-02 8.47033858e-01 3.67376298e-01
1.15147018e+00 -2.89361477e-01 -7.13320673e-01 9.14211154e-01
1.22072113e+00 4.08140160e-02 6.99927926e-01 5.13909280e-01
1.06835699e+00 5.69499135e-01 6.21455312e-01 7.44308352e-01
9.38317716e-01 1.03761721e+00 9.05016661e-02 1.62401423e-02
-7.59657994e-02 -3.22737813e-01 6.06472552e-01 7.39113808e-01
-6.85841024e-01 -1.51189134e-01 -8.47886741e-01 5.71503758e-01
-2.14949107e+00 -1.21630180e+00 -2.55694032e-01 2.29906940e+00
4.43953782e-01 -1.40796006e-01 5.65465152e-01 -1.52096106e-03
1.17115533e+00 -1.37080297e-01 -4.30672705e-01 1.97681412e-01
-2.08899587e-01 -6.04487479e-01 9.15968716e-01 3.56113940e-01
-1.52056599e+00 8.18799198e-01 5.26634121e+00 8.20458829e-01
-4.41452563e-01 4.09799010e-01 4.32675958e-01 -1.86931640e-01
3.38058949e-01 -1.06416956e-01 -1.41346622e+00 9.20018077e-01
1.07366729e+00 2.10600868e-01 5.26526511e-01 7.52729237e-01
2.13492006e-01 -2.21064776e-01 -1.09861588e+00 1.57706809e+00
4.21496779e-01 -9.73980606e-01 -3.19291174e-01 1.50978968e-01
6.05563402e-01 -5.51550910e-02 -9.55639258e-02 2.49640629e-01
2.57214069e-01 -6.69332922e-01 5.55603862e-01 7.49542356e-01
7.82328248e-01 -8.59307587e-01 9.20057774e-01 2.42238879e-01
-1.52482092e+00 -3.03530931e-01 -4.09253657e-01 7.61916190e-02
1.04382791e-01 5.01478612e-01 -4.24916506e-01 7.14675426e-01
1.14889860e+00 1.16063237e+00 -9.75295782e-01 9.96414840e-01
7.58598074e-02 5.42488098e-01 -4.52965111e-01 2.23098189e-01
-1.71458155e-01 -2.85481453e-01 5.22178710e-01 1.55161846e+00
2.03171045e-01 1.09582931e-01 3.24861437e-01 5.41605115e-01
1.69879466e-01 -1.27743304e-01 -3.41398388e-01 3.40331793e-01
6.34507477e-01 1.23882306e+00 -4.39639151e-01 -2.98149019e-01
-6.39588296e-01 1.27311516e+00 3.00975114e-01 3.81836861e-01
-1.17139268e+00 -1.97474733e-01 5.76692700e-01 1.31629393e-01
1.34241059e-01 -3.21413994e-01 -5.47780609e-03 -1.46936381e+00
1.39214694e-01 -5.78455389e-01 6.75827801e-01 -4.88318294e-01
-1.55507886e+00 3.04091781e-01 2.62497962e-01 -1.45174623e+00
-8.59934278e-03 -4.79264587e-01 -4.16432202e-01 7.35723674e-01
-1.32497883e+00 -1.55565464e+00 -6.81336403e-01 8.92201006e-01
3.53267342e-01 -4.99847651e-01 5.07983387e-01 9.17525947e-01
-1.16972566e+00 1.08743536e+00 2.62467623e-01 4.61831242e-01
1.03171933e+00 -1.14538610e+00 6.92283735e-02 9.32140768e-01
-1.49290010e-01 5.21143973e-01 4.15060282e-01 -6.06032610e-01
-1.13297248e+00 -1.27612114e+00 8.20115745e-01 -6.99024796e-01
3.19888979e-01 -4.77046221e-01 -5.76183617e-01 8.28881085e-01
-1.44632921e-01 -6.38606548e-02 8.38338435e-01 2.45694771e-01
-3.60976189e-01 -4.13600147e-01 -1.07543612e+00 5.44061184e-01
1.28007853e+00 -5.29489160e-01 -2.53010392e-01 3.59632313e-01
3.89245242e-01 -5.99616766e-03 -9.13426578e-01 1.55656025e-01
5.89888871e-01 -1.03191960e+00 1.27881336e+00 -3.98239732e-01
-1.32882535e-01 -4.53601450e-01 -2.57393986e-01 -9.85052943e-01
-5.82161367e-01 -3.08397114e-01 -6.45911992e-02 1.83923304e+00
8.06767121e-02 -6.60805106e-01 6.85315847e-01 9.94034111e-01
-9.28639900e-03 -1.71625495e-01 -9.85987067e-01 -9.25016224e-01
-2.84102291e-01 -2.87566692e-01 5.93684316e-01 1.04700243e+00
3.27306055e-02 1.29865482e-01 -7.62388468e-01 3.05376559e-01
9.09358978e-01 -3.76355946e-01 9.51694787e-01 -1.30434644e+00
-3.71847786e-02 -1.19716704e-01 -4.40089613e-01 -9.64566469e-01
2.17173830e-01 -8.49449754e-01 -3.77463967e-01 -1.22567451e+00
6.85091913e-01 -6.42488241e-01 -2.53116250e-01 4.55047429e-01
-1.99374497e-01 2.00642422e-01 3.33958536e-01 5.61267972e-01
-9.60413873e-01 5.67050397e-01 6.14850879e-01 -1.45641342e-01
-1.99203849e-01 1.86672143e-03 -4.23965365e-01 8.91361535e-01
9.03301895e-01 -2.96284974e-01 -2.25495234e-01 -4.10402209e-01
-8.46184939e-02 -3.56097102e-01 8.36404920e-01 -1.24947417e+00
5.22599280e-01 -1.73607748e-02 7.63677359e-01 -7.47391462e-01
4.40478891e-01 -8.76604021e-01 2.59359300e-01 2.01926544e-01
-1.27947584e-01 1.78950638e-01 6.74365908e-02 7.46881306e-01
2.01001372e-02 -8.93432423e-02 8.30721974e-01 -2.19112203e-01
-9.98022199e-01 3.76960546e-01 -1.50851190e-01 1.41767247e-04
1.04488063e+00 -4.36599642e-01 -3.29190284e-01 -1.51269451e-01
-6.65304542e-01 4.72626925e-01 5.05939305e-01 4.37504232e-01
4.31857109e-01 -1.56685352e+00 -5.98076642e-01 1.77421704e-01
3.11885089e-01 -3.46343428e-01 6.23150885e-01 1.04999781e+00
-6.90594614e-02 4.98223066e-01 -1.87850609e-01 -6.53028786e-01
-1.47807872e+00 5.80485165e-01 2.69336611e-01 -1.72111034e-01
-5.86652339e-01 7.05280423e-01 -1.81532297e-02 -5.00136852e-01
3.85930628e-01 2.05870345e-01 -3.45580608e-01 1.87977344e-01
6.69036746e-01 9.56126630e-01 -2.80174851e-01 -1.06225371e+00
-5.79840899e-01 8.20735931e-01 -2.53357410e-01 3.85164917e-02
1.11008811e+00 -6.16544604e-01 6.40607029e-02 2.49736637e-01
1.23647320e+00 -1.47590354e-01 -1.42394650e+00 -3.17825854e-01
4.85375002e-02 -5.83019733e-01 -8.31253976e-02 -6.66497707e-01
-8.96953881e-01 7.17663884e-01 9.68903959e-01 -1.69973135e-01
1.06474590e+00 -5.84709458e-02 9.17229295e-01 2.40747347e-01
4.85248208e-01 -1.60766566e+00 1.27334949e-02 3.82264465e-01
4.63025898e-01 -1.51827681e+00 2.45178953e-01 -4.01265711e-01
-8.56956661e-01 5.82257628e-01 7.71846235e-01 1.35181561e-01
6.78395629e-01 -1.17017046e-01 -3.69516164e-01 1.95854217e-01
-2.52927214e-01 -3.97584587e-01 4.44002837e-01 8.74953985e-01
8.92450511e-02 1.36011750e-01 -9.31900963e-02 8.75913799e-01
-4.13536243e-02 -1.43080294e-01 1.82597995e-01 5.98858237e-01
-2.06805423e-01 -9.78234470e-01 -4.72030312e-01 3.08750451e-01
-1.82888359e-01 4.36721742e-02 -2.78655857e-01 8.17422092e-01
6.43733323e-01 1.12071955e+00 1.00599304e-01 -7.07675815e-01
3.15948904e-01 -4.77198511e-02 3.41760755e-01 -2.09443137e-01
-4.42314863e-01 -2.04417873e-02 3.33076596e-01 -5.02217829e-01
-8.23658705e-01 -9.17041838e-01 -8.75272214e-01 -7.11586177e-01
-3.18734974e-01 1.39721332e-03 4.23287272e-01 8.30999851e-01
5.31640887e-01 1.56463295e-01 6.38554990e-01 -1.01820242e+00
-1.67336643e-01 -9.31291521e-01 -5.68334579e-01 5.77920198e-01
9.57408771e-02 -1.01835680e+00 -4.18011755e-01 3.74138147e-01]
|
[14.767950057983398, 1.0272873640060425]
|
186f380e-2b8b-4682-9a0b-7f1a06483ed3
|
a-probabilistic-constrained-clustering-for
|
1806.11078
| null |
http://arxiv.org/abs/1806.11078v1
|
http://arxiv.org/pdf/1806.11078v1.pdf
|
A probabilistic constrained clustering for transfer learning and image category discovery
|
Neural network-based clustering has recently gained popularity, and in
particular a constrained clustering formulation has been proposed to perform
transfer learning and image category discovery using deep learning. The core
idea is to formulate a clustering objective with pairwise constraints that can
be used to train a deep clustering network; therefore the cluster assignments
and their underlying feature representations are jointly optimized end-to-end.
In this work, we provide a novel clustering formulation to address scalability
issues of previous work in terms of optimizing deeper networks and larger
amounts of categories. The proposed objective directly minimizes the negative
log-likelihood of cluster assignment with respect to the pairwise constraints,
has no hyper-parameters, and demonstrates improved scalability and performance
on both supervised learning and unsupervised transfer learning.
|
['Joel Schlosser', 'Yen-Chang Hsu', 'Zhaoyang Lv', 'Zsolt Kira', 'Phillip Odom']
|
2018-06-28
| null | null | null | null |
['ecg-risk-stratification']
|
['medical']
|
[-1.84041589e-01 -1.97975993e-01 1.01360910e-01 -8.46508563e-01
-6.13760114e-01 -4.61332768e-01 2.71917313e-01 2.17987373e-01
-6.99346840e-01 1.19032390e-01 -1.17520697e-01 7.82539397e-02
-4.83264357e-01 -3.86971623e-01 -4.33147818e-01 -8.53553891e-01
-3.32400292e-01 7.89897203e-01 -2.40662411e-01 5.13554037e-01
1.73683122e-01 2.18399391e-01 -1.47462380e+00 1.83466926e-01
8.09187174e-01 1.00675356e+00 4.25540835e-01 2.00948462e-01
2.35086009e-01 6.18376136e-01 -4.08907026e-01 -9.82967019e-02
8.62407610e-02 -2.33914182e-01 -7.65286803e-01 4.18734699e-01
3.33889335e-01 2.49171797e-02 -2.24961013e-01 1.11761224e+00
4.98435766e-01 4.94487166e-01 9.65768516e-01 -1.39001381e+00
-7.72629976e-01 5.53951442e-01 -7.02810109e-01 -1.51316151e-01
-4.19852138e-01 -1.56625584e-01 1.11590827e+00 -8.43382001e-01
1.58715248e-01 1.20898211e+00 6.47945762e-01 4.53242034e-01
-1.45647562e+00 -7.57200420e-01 2.42544711e-01 4.21520889e-01
-1.88458669e+00 -2.82757357e-02 7.31201172e-01 -6.03193879e-01
7.87008524e-01 -3.09948534e-01 4.24941897e-01 6.43735766e-01
-3.20708871e-01 8.64821851e-01 6.01322293e-01 -3.74542981e-01
5.43648005e-01 7.25385845e-02 1.51852161e-01 7.16481388e-01
-1.25370845e-01 -2.41599321e-01 -1.94637090e-01 7.05927089e-02
6.36094868e-01 3.49447489e-01 1.49537817e-01 -8.25474501e-01
-1.27498245e+00 1.33978021e+00 9.29873228e-01 3.05510581e-01
-1.61396474e-01 4.71982449e-01 3.24766070e-01 -7.62045309e-02
2.83672661e-01 4.34341848e-01 -5.24985135e-01 2.19200119e-01
-1.04745257e+00 -5.21605648e-02 5.80937028e-01 9.85155404e-01
1.02508235e+00 -2.46532068e-01 6.89783692e-02 1.15994406e+00
5.00160456e-01 2.65968800e-01 5.20999253e-01 -1.19051862e+00
1.77042544e-01 5.85637033e-01 -2.62304991e-01 -1.21626759e+00
-5.95940292e-01 -4.76067901e-01 -1.20375168e+00 -6.41655028e-02
-2.24731080e-02 -2.85835147e-01 -1.00593030e+00 1.95701063e+00
1.16792351e-01 2.45938063e-01 8.76319502e-03 1.17176449e+00
4.53044027e-01 5.71663260e-01 1.73063219e-01 1.50087997e-01
9.41206336e-01 -9.05427277e-01 -6.08929038e-01 -2.34364212e-01
4.92674619e-01 -5.53410947e-01 9.08361197e-01 2.22120956e-01
-7.75050223e-01 -4.87125486e-01 -9.06935275e-01 8.92880745e-03
-2.90574402e-01 3.83997470e-01 8.17727566e-01 2.32311800e-01
-1.43820655e+00 4.10414547e-01 -1.18446326e+00 -4.00332838e-01
7.16802061e-01 9.15103376e-01 -9.02925059e-02 -1.79631010e-01
-7.68760085e-01 4.09843773e-01 7.98176110e-01 1.37108952e-01
-8.22894871e-01 -4.35875654e-01 -7.81227291e-01 5.07411599e-01
2.15906233e-01 -5.66859543e-01 8.95014048e-01 -1.00232971e+00
-1.44349539e+00 8.50248814e-01 -7.80948699e-02 -1.97346807e-01
2.35553104e-02 -6.36271783e-04 4.97823171e-02 2.75269657e-01
2.87571877e-01 1.21462643e+00 5.69825113e-01 -1.23060572e+00
-6.46620452e-01 -4.19826925e-01 -3.20857018e-01 2.52197981e-01
-8.39726746e-01 3.48354876e-02 -1.00127268e+00 -6.10657811e-01
4.04408574e-01 -1.03766990e+00 -5.11811674e-01 -1.03077196e-01
-4.94299948e-01 -5.49384892e-01 7.28673995e-01 -2.22587511e-01
8.80348802e-01 -2.10393238e+00 6.18553817e-01 5.06080270e-01
3.99282634e-01 2.17122827e-02 -2.47240946e-01 9.14563611e-02
1.28669348e-02 1.37773901e-01 -5.63269377e-01 -6.38615131e-01
2.66351163e-01 1.84953976e-02 2.03067124e-01 5.43155551e-01
2.98053294e-01 8.20954680e-01 -6.68993533e-01 -5.70752501e-01
3.58664364e-01 7.02542484e-01 -1.00958085e+00 3.45428586e-01
-7.13533610e-02 3.87060136e-01 -1.22057721e-01 4.58360970e-01
5.90523660e-01 -6.64151967e-01 4.63044047e-01 -1.84903875e-01
1.09469131e-01 -2.04329029e-01 -1.09540045e+00 1.90866721e+00
-3.33493352e-01 8.01700294e-01 2.74044812e-01 -1.50031984e+00
7.28531778e-01 9.79916528e-02 7.70150483e-01 -3.31477553e-01
2.83184856e-01 -2.06057429e-01 5.22964559e-02 -2.96978146e-01
4.20927480e-02 -5.30162193e-02 -2.06109300e-01 3.78314525e-01
3.08753103e-01 2.83609331e-01 8.25051367e-02 3.03701699e-01
8.02759707e-01 -3.43637139e-01 -1.84805214e-01 -4.12616462e-01
2.95655966e-01 1.69816576e-02 5.81904650e-01 6.67789996e-01
7.92565942e-02 7.51205742e-01 3.33818018e-01 -4.64119278e-02
-8.35577369e-01 -1.05197501e+00 -1.84831738e-01 1.28388023e+00
1.57249555e-01 -2.38486588e-01 -8.91695023e-01 -4.90511239e-01
4.51842062e-02 2.26635978e-01 -5.31570971e-01 -1.91650674e-01
-3.66327912e-01 -8.06682646e-01 2.95187891e-01 6.39708817e-01
5.38691342e-01 -9.83155847e-01 -3.18822086e-01 2.46857062e-01
-2.51985341e-01 -1.06800473e+00 -4.95830029e-01 5.41166723e-01
-7.95490563e-01 -1.11706960e+00 -4.88665313e-01 -1.45255446e+00
9.65852737e-01 2.61039257e-01 8.71155441e-01 1.96552306e-01
-4.10900533e-01 2.83518970e-01 -2.02937707e-01 -1.99621785e-02
1.93021774e-01 2.50287652e-01 1.40037924e-01 2.11733729e-01
7.35026836e-01 -6.68606818e-01 -7.35596180e-01 4.82482076e-01
-9.36872303e-01 -1.32191733e-01 7.02203572e-01 8.13345015e-01
6.85135365e-01 2.38522500e-01 6.24924660e-01 -6.40141904e-01
3.22544664e-01 -5.23842990e-01 -7.06794798e-01 3.33706886e-02
-4.65856314e-01 -1.09425150e-01 6.94709003e-01 -3.65571141e-01
-6.64575577e-01 3.81673187e-01 1.98259115e-01 -9.86933827e-01
-4.50930417e-01 8.55093241e-01 -4.16734457e-01 1.98896587e-01
2.10167050e-01 1.50590269e-02 5.65054789e-02 -5.83755732e-01
5.34081340e-01 8.12029481e-01 6.45742714e-01 -5.47075808e-01
7.21615553e-01 5.66680372e-01 -2.44850442e-01 -5.42579830e-01
-9.82246041e-01 -1.00438559e+00 -1.35757911e+00 2.63438420e-03
1.25860894e+00 -1.16426563e+00 -1.05891216e+00 3.76720548e-01
-9.12151456e-01 -6.06094062e-01 3.46605599e-01 7.22060323e-01
-5.86920679e-01 3.80707115e-01 -5.75653493e-01 -3.03187549e-01
-2.13413969e-01 -1.13571513e+00 8.84798884e-01 1.89526483e-01
-6.49710298e-02 -1.18479121e+00 -2.45005980e-01 1.70792922e-01
3.41444105e-01 2.58339047e-01 1.16751909e+00 -6.25405371e-01
-4.28557873e-01 -1.63052365e-01 -6.13371491e-01 4.71948624e-01
8.46747011e-02 -1.64418101e-01 -7.66669989e-01 -7.98277259e-01
-2.31523603e-01 -4.82927322e-01 9.95260954e-01 6.83084130e-01
1.65162706e+00 -3.21740657e-01 -4.51655626e-01 8.98519456e-01
1.66639042e+00 1.37617797e-01 2.02085584e-01 6.03578910e-02
1.10339224e+00 6.50651872e-01 2.42067948e-01 2.95282245e-01
3.31447542e-01 6.23380065e-01 3.61651033e-01 -2.33458921e-01
5.11856750e-02 1.76209733e-01 -5.74290305e-02 9.94056463e-01
5.02834678e-01 -1.34135634e-01 -1.18269086e+00 5.99990427e-01
-2.30960369e+00 -8.34709346e-01 2.15034619e-01 1.96830428e+00
6.86636269e-01 -1.71263784e-01 5.30270077e-02 -2.54577845e-01
1.05191767e+00 -3.64380658e-01 -7.90797174e-01 1.23292029e-01
1.65140659e-01 1.03998423e-01 2.09755898e-01 3.60417426e-01
-1.60820997e+00 1.09237015e+00 6.55897188e+00 6.68969750e-01
-1.10466492e+00 1.28549173e-01 6.64828360e-01 -3.85479540e-01
2.37862140e-01 -2.07457364e-01 -5.06541371e-01 4.74478424e-01
7.39638150e-01 1.92242786e-01 4.29755718e-01 9.69083786e-01
2.05655202e-01 2.39587530e-01 -1.39775157e+00 1.04666591e+00
1.52546272e-01 -1.29985905e+00 -1.31484091e-01 3.76960754e-01
8.78209829e-01 3.71331304e-01 3.24862987e-01 4.45230991e-01
7.04522431e-01 -1.18974948e+00 5.34369230e-01 1.60418853e-01
7.80068815e-01 -1.17696857e+00 7.28099644e-01 2.19413802e-01
-9.68277812e-01 -3.63768309e-01 -7.06275582e-01 -3.12010106e-02
-2.78283596e-01 5.87869942e-01 -7.85338163e-01 -2.25420669e-02
1.02615082e+00 1.03646851e+00 -5.89219928e-01 1.21040857e+00
9.51343402e-03 8.21881056e-01 -5.03016293e-01 -3.54156420e-02
6.12181544e-01 -5.28611839e-01 -1.04306400e-01 1.33126545e+00
8.96637887e-02 1.13171106e-02 8.21818173e-01 1.14528978e+00
-3.39938790e-01 -1.34369478e-01 -6.61430284e-02 1.52167588e-01
5.81269562e-01 1.38701320e+00 -1.17163968e+00 -1.82633966e-01
-1.90706745e-01 8.98479044e-01 8.76581848e-01 5.12789249e-01
-7.30404854e-01 -5.18656552e-01 7.92401552e-01 -3.29491019e-01
8.08265507e-01 -3.28927308e-01 -3.11738849e-01 -1.05825162e+00
-2.74911612e-01 -4.19687688e-01 5.97425997e-01 -2.90865362e-01
-1.51933563e+00 2.86464036e-01 2.15913896e-02 -1.00194073e+00
-1.39277473e-01 -5.53308308e-01 -8.00991297e-01 6.04186833e-01
-1.32629800e+00 -1.13432598e+00 -3.31478715e-01 7.79813945e-01
2.98340082e-01 -3.28914523e-01 7.05985785e-01 4.65170920e-01
-9.40285742e-01 7.19741940e-01 7.53672183e-01 6.09658360e-01
6.62096322e-01 -1.54269505e+00 -1.96535200e-01 7.54778802e-01
6.07759468e-02 7.12042212e-01 3.14731091e-01 -2.60840267e-01
-1.01371670e+00 -1.55651414e+00 5.05127668e-01 -3.55100483e-01
4.70061421e-01 -7.18808711e-01 -8.45996499e-01 4.89836186e-01
1.67538345e-01 -1.43484101e-01 1.12993503e+00 4.06584322e-01
-4.89100128e-01 -2.58224290e-02 -1.00962186e+00 2.50939310e-01
7.02217519e-01 -4.87810433e-01 -1.92314222e-01 6.52844250e-01
7.07213879e-01 6.74780384e-02 -9.82374310e-01 9.25222039e-02
3.12130433e-02 -6.61190033e-01 8.97863328e-01 -5.01066089e-01
3.38603377e-01 -4.45163608e-01 -2.81033814e-01 -1.33214867e+00
-8.22558641e-01 -3.12721521e-01 1.30191281e-01 1.16638589e+00
2.72455722e-01 -2.63416450e-02 1.04258883e+00 5.85410833e-01
-3.12048286e-01 -7.28092372e-01 -8.37351382e-01 -5.97671747e-01
2.72881031e-01 -1.59679860e-01 2.87340969e-01 1.29854035e+00
9.68469381e-02 5.07635951e-01 -1.13419831e-01 2.60082424e-01
9.80033815e-01 3.36414367e-01 4.42923397e-01 -1.42499352e+00
-1.07189417e-01 -6.55366004e-01 -4.80941623e-01 -8.83104324e-01
4.92396921e-01 -1.36063719e+00 4.42045063e-01 -1.59907806e+00
4.62719560e-01 -5.19317329e-01 -5.64864457e-01 7.38691390e-01
-8.89928788e-02 4.08740759e-01 3.01478207e-01 6.21440351e-01
-1.24919176e+00 7.29793787e-01 5.43195605e-01 -2.73648947e-01
-3.00453752e-01 -1.92222863e-01 -7.93293178e-01 7.13957310e-01
8.38641226e-01 -6.43089950e-01 -2.79107153e-01 -7.67755628e-01
-2.46865049e-01 -3.54770720e-01 3.47736686e-01 -1.14777684e+00
6.30731285e-01 1.01894319e-01 6.62552536e-01 -7.90121615e-01
1.98038355e-01 -9.15661812e-01 -2.60051966e-01 3.45047057e-01
-6.97573423e-01 -1.87199160e-01 -4.81836423e-02 8.01890433e-01
-4.14344281e-01 -5.47562689e-02 1.01950681e+00 -6.96157217e-02
-7.69014716e-01 5.64083457e-01 -4.94787544e-01 -1.44509375e-01
1.05298913e+00 -7.78160319e-02 2.54652679e-01 -2.13174984e-01
-1.02970243e+00 7.86131978e-01 4.42456573e-01 3.68497729e-01
6.33629918e-01 -1.47636437e+00 -6.44669712e-01 -9.72916707e-02
8.18601698e-02 4.24711347e-01 6.48890957e-02 7.27885544e-01
-4.19369370e-01 5.41607380e-01 -6.45229369e-02 -1.06042922e+00
-9.45030153e-01 7.55168021e-01 4.02017295e-01 -1.39931254e-02
-4.05488968e-01 9.60807025e-01 3.50736409e-01 -8.10018301e-01
8.34823668e-01 1.53003633e-01 -2.61877954e-01 1.42144382e-01
3.02591771e-01 7.43869618e-02 5.95296547e-02 -5.61089337e-01
-3.62574071e-01 4.97404218e-01 -2.65779853e-01 1.33362159e-01
1.58768916e+00 -2.97000855e-01 -3.64819139e-01 1.32716119e-01
2.03346825e+00 -7.96642363e-01 -1.36404884e+00 -5.14957607e-01
2.88290679e-01 -1.64289638e-01 3.14893425e-01 -7.61326253e-01
-1.48324907e+00 1.06642306e+00 8.73972237e-01 -1.92744657e-01
9.24763441e-01 2.62783796e-01 5.97996891e-01 8.07590365e-01
-5.28960861e-02 -1.31097639e+00 4.91965026e-01 4.89426672e-01
4.15371060e-01 -1.55861282e+00 -2.87204415e-01 -1.67629272e-01
-5.37893593e-01 9.42794085e-01 9.08687115e-01 -2.23217085e-01
1.00496173e+00 -6.19007386e-02 1.54421002e-01 -5.37572324e-01
-4.80882257e-01 -3.26635331e-01 3.23910624e-01 9.21101987e-01
4.38377172e-01 5.08325435e-02 3.33841324e-01 4.77900892e-01
1.94507409e-02 -4.71031398e-01 9.75612253e-02 4.92916882e-01
-5.22662163e-01 -6.99821651e-01 -2.70969987e-01 6.51483536e-01
-2.77867347e-01 -1.02154329e-01 -4.19988006e-01 5.11454225e-01
1.71234995e-01 1.19529414e+00 6.23897254e-01 -3.73801559e-01
-1.24423623e-01 -1.69808060e-01 5.11439741e-02 -8.86252761e-01
-4.05402452e-01 3.76433909e-01 -7.16196179e-01 -4.17356044e-01
-6.46961212e-01 -6.65247500e-01 -1.46974409e+00 -7.83187151e-02
-4.25611615e-01 3.64351779e-01 6.90943837e-01 8.64233375e-01
6.16122246e-01 3.32409412e-01 8.55710864e-01 -1.10376465e+00
-2.69429386e-01 -9.83121693e-01 -5.57554066e-01 7.35970676e-01
4.62435260e-02 -5.41948617e-01 -4.03686255e-01 7.37757161e-02]
|
[9.175588607788086, 3.2306647300720215]
|
4193eb94-d280-4ad8-94d2-d2808c001117
|
contextual-dynamic-prompting-for-response
|
2301.13268
| null |
https://arxiv.org/abs/2301.13268v2
|
https://arxiv.org/pdf/2301.13268v2.pdf
|
Contextual Dynamic Prompting for Response Generation in Task-oriented Dialog Systems
|
Response generation is one of the critical components in task-oriented dialog systems. Existing studies have shown that large pre-trained language models can be adapted to this task. The typical paradigm of adapting such extremely large language models would be by fine-tuning on the downstream tasks which is not only time-consuming but also involves significant resources and access to fine-tuning data. Prompting (Schick and Sch\"utze, 2020) has been an alternative to fine-tuning in many NLP tasks. In our work, we explore the idea of using prompting for response generation in task-oriented dialog systems. Specifically, we propose an approach that performs contextual dynamic prompting where the prompts are learnt from dialog contexts. We aim to distill useful prompting signals from the dialog context. On experiments with MultiWOZ 2.2 dataset (Zang et al., 2020), we show that contextual dynamic prompts improve response generation in terms of combined score (Mehri et al., 2019) by 3 absolute points, and a massive 20 points when dialog states are incorporated. Furthermore, human annotation on these conversations found that agents which incorporate context were preferred over agents with vanilla prefix-tuning.
|
['Rashmi Gangadharaiah', 'Chacha Chen', 'Narges Tabari', 'Sandesh Swamy']
|
2023-01-30
| null | null | null | null |
['response-generation']
|
['natural-language-processing']
|
[ 1.12542398e-01 5.22234857e-01 1.04484744e-01 -7.37343550e-01
-6.88441396e-01 -9.25510406e-01 1.05699217e+00 -7.71275386e-02
-6.78188980e-01 1.09062672e+00 8.62445414e-01 -3.57531637e-01
1.09834678e-01 -6.07627511e-01 7.57681997e-03 -2.22758099e-01
2.22298682e-01 9.39372897e-01 2.85716593e-01 -1.02633548e+00
3.97161067e-01 1.71131536e-01 -9.47181165e-01 5.49438715e-01
7.27819562e-01 3.57979596e-01 4.51028526e-01 9.68547761e-01
-4.70618486e-01 6.97121143e-01 -1.02412021e+00 -3.32996726e-01
1.75354376e-01 -4.84431714e-01 -1.38768625e+00 -7.55856261e-02
1.10175699e-01 -4.85586464e-01 -1.79554462e-01 5.71566641e-01
6.50613189e-01 7.13029504e-01 5.27966440e-01 -9.32504535e-01
-5.06587923e-01 1.16480494e+00 1.35165542e-01 1.45961896e-01
6.90864682e-01 5.17876148e-01 9.55613136e-01 -7.45581448e-01
5.40382326e-01 1.73712027e+00 2.92369932e-01 1.22970665e+00
-1.22403681e+00 -4.47799444e-01 2.31490165e-01 -1.05172671e-01
-9.27521706e-01 -6.51164591e-01 5.61006963e-01 -5.06848276e-01
1.26588213e+00 3.52487206e-01 4.00849991e-03 1.24803996e+00
-2.85897199e-02 5.46374142e-01 1.06119871e+00 -5.75542748e-01
1.10171139e-01 3.92077357e-01 2.14058325e-01 3.23879510e-01
-5.39206386e-01 2.33195931e-01 -4.93281126e-01 -3.38661939e-01
5.49458206e-01 -2.79422879e-01 -1.72791988e-01 2.88646221e-01
-1.30037117e+00 1.21089947e+00 3.59882712e-01 4.14190322e-01
-3.24121267e-01 -1.92854002e-01 3.65830839e-01 6.20012760e-01
4.27778929e-01 1.06897676e+00 -7.70282865e-01 -3.36059213e-01
-2.74432182e-01 5.35565495e-01 1.20738101e+00 9.53138232e-01
6.79347277e-01 -5.50628379e-02 -7.33002543e-01 1.15542650e+00
2.51999557e-01 1.92119777e-01 8.63971651e-01 -1.23323369e+00
7.79329240e-01 6.23629868e-01 4.18151021e-01 -2.74392784e-01
-6.25333846e-01 3.49074155e-01 -6.62041962e-01 -2.36014128e-02
8.21451306e-01 -8.13774645e-01 -6.10343218e-01 2.09877872e+00
3.20216656e-01 -5.55100143e-01 3.49510461e-01 7.79125035e-01
8.79653871e-01 7.35272706e-01 4.89476770e-01 -3.39676887e-01
1.38808239e+00 -1.10616016e+00 -8.18056703e-01 -3.49845052e-01
6.83099568e-01 -9.58428681e-01 1.55164099e+00 9.21636671e-02
-1.06856370e+00 -6.93854094e-01 -5.72940171e-01 -2.91090012e-02
-2.65259206e-01 -2.34994859e-01 7.22061694e-01 5.67327559e-01
-1.31200278e+00 2.67025769e-01 -2.05572844e-01 -7.47677445e-01
-4.56702143e-01 4.69275296e-01 -1.21978849e-01 3.81107539e-01
-1.65518630e+00 1.25870645e+00 4.30650473e-01 -2.38055706e-01
-6.49007261e-01 -3.99198472e-01 -6.34743929e-01 1.81319892e-01
5.84847212e-01 -6.66333556e-01 2.12244630e+00 -4.93236303e-01
-2.33626604e+00 4.60959494e-01 -1.53797388e-01 -4.98587728e-01
3.88679355e-01 -2.65761882e-01 -6.84409216e-02 -1.09380983e-01
-1.18317045e-01 9.27026808e-01 6.46307409e-01 -9.73333776e-01
-6.96173728e-01 -3.86220776e-02 6.68269157e-01 4.72792059e-01
-2.44636297e-01 2.91224629e-01 1.33714631e-01 -4.31804478e-01
-3.82109731e-01 -1.09728336e+00 -5.88158607e-01 -7.48864472e-01
-2.31609300e-01 -8.46980929e-01 6.00546837e-01 -4.37414140e-01
1.26980937e+00 -1.65924883e+00 6.41681850e-02 -1.97623104e-01
2.38068104e-01 3.93829942e-01 -2.84173369e-01 8.21146667e-01
2.14760065e-01 7.11018369e-02 -1.37127265e-01 -2.05383196e-01
2.77756721e-01 7.93217719e-02 -5.15427768e-01 -3.52823317e-01
1.63052395e-01 1.01057601e+00 -8.43671918e-01 -2.23014906e-01
3.68357152e-01 8.21062103e-02 -8.42039764e-01 8.23688030e-01
-7.12560475e-01 8.11141908e-01 -3.12128395e-01 7.41732940e-02
7.95210749e-02 -9.53015983e-02 3.94784451e-01 2.13490412e-01
-1.60860121e-01 6.89814150e-01 -8.20638657e-01 1.51600790e+00
-7.97569692e-01 4.20809090e-01 1.05469182e-01 -4.60091531e-01
1.01463735e+00 7.78784156e-01 9.71511081e-02 -4.91166830e-01
1.61013559e-01 -2.07348634e-02 2.98166722e-01 -6.18888199e-01
8.09444904e-01 -3.58905584e-01 -6.19608939e-01 7.44312108e-01
1.15454063e-01 -5.95344663e-01 3.03602278e-01 3.76514316e-01
9.90240097e-01 -2.16288924e-01 6.21875942e-01 -2.65323937e-01
7.62111604e-01 1.19188562e-01 1.51194289e-01 7.69485891e-01
-3.28506172e-01 2.74647176e-01 2.57770836e-01 -2.80749351e-01
-7.99407244e-01 -8.08272779e-01 2.83726960e-01 1.87978840e+00
-3.00445139e-01 -4.86302257e-01 -9.89899516e-01 -7.32791662e-01
-2.37535894e-01 1.02910805e+00 -4.29050565e-01 -9.13987234e-02
-9.93168414e-01 -4.69536483e-01 5.96967280e-01 4.06727225e-01
5.50576091e-01 -1.50807440e+00 -5.84312797e-01 4.96775120e-01
-5.90781927e-01 -1.04562712e+00 -8.31186295e-01 2.96321571e-01
-7.11532116e-01 -6.88479245e-01 -6.71896875e-01 -5.66364586e-01
3.85692805e-01 9.31245759e-02 1.20762587e+00 1.24243768e-02
2.08905756e-01 3.44404936e-01 -4.75846589e-01 -2.57543355e-01
-9.45272505e-01 3.54212880e-01 2.44693711e-01 -4.48803931e-01
4.49195832e-01 -2.60803819e-01 -3.38414639e-01 5.51316619e-01
-6.53424859e-01 2.08444446e-02 5.20938098e-01 1.07809412e+00
-2.59725928e-01 -5.74527621e-01 1.15917015e+00 -1.18875182e+00
1.54279637e+00 -2.43150309e-01 -4.91009086e-01 2.05973402e-01
-6.10351920e-01 4.03396308e-01 7.13449538e-01 -5.65303564e-01
-1.67789257e+00 3.40643898e-02 -4.26498115e-01 2.18727306e-01
-5.12821674e-01 2.81753659e-01 -8.42836201e-02 2.32470036e-01
1.02359700e+00 -1.89665988e-01 -2.95440078e-01 -4.58172888e-01
7.77007103e-01 9.63990510e-01 3.34743559e-01 -9.75328684e-01
6.33381665e-01 -2.53831893e-01 -5.37988007e-01 -6.75923765e-01
-6.77570224e-01 -4.79757845e-01 -6.65622234e-01 -9.08434615e-02
8.91071320e-01 -5.00604153e-01 -9.16960955e-01 9.67176780e-02
-1.42605269e+00 -1.01192653e+00 -2.06706390e-01 3.51069778e-01
-6.47884190e-01 2.15490356e-01 -8.15512300e-01 -8.62344027e-01
-5.18299639e-01 -1.06641388e+00 7.17589974e-01 3.03811014e-01
-9.89510357e-01 -1.16313076e+00 3.32842201e-01 4.56109107e-01
7.39497542e-01 -4.09663677e-01 9.69312072e-01 -1.28395855e+00
-1.80727214e-01 1.22275040e-01 -2.67841923e-03 4.90152463e-02
3.54396224e-01 -4.12267506e-01 -1.14770436e+00 -1.83951348e-01
2.26631165e-01 -5.70914626e-01 3.47973138e-01 1.13431737e-02
4.97703284e-01 -7.66924202e-01 -4.58783545e-02 -8.73576924e-02
6.56506002e-01 5.98509014e-01 1.98137581e-01 5.18630967e-02
3.44715238e-01 1.18545187e+00 8.44132006e-01 3.56744677e-01
4.76005405e-01 9.21654522e-01 -1.59068644e-01 2.83625782e-01
-1.24910340e-01 -3.49454761e-01 4.89422232e-01 8.25901270e-01
5.47514223e-02 -4.04184908e-01 -8.27746153e-01 4.41342652e-01
-1.90203607e+00 -9.25588071e-01 -1.11891993e-03 1.93270028e+00
1.41579187e+00 3.52752917e-02 1.89150423e-01 -4.48864967e-01
6.63439035e-01 -1.18899234e-02 -2.48453394e-01 -7.16635227e-01
2.58022517e-01 2.13279203e-01 -8.44066069e-02 1.23929894e+00
-7.33132422e-01 1.47821188e+00 6.29329729e+00 3.62165332e-01
-9.39844370e-01 2.17747271e-01 3.92363638e-01 2.48783957e-02
-1.59193680e-01 5.83869368e-02 -1.01536953e+00 3.65407974e-01
1.24845707e+00 -3.47527802e-01 5.66736281e-01 6.30923390e-01
4.18916553e-01 -7.50279799e-02 -1.42392516e+00 5.68462074e-01
-2.16717571e-01 -9.73171413e-01 -4.85874563e-02 -1.73564956e-01
6.50199115e-01 -2.55971044e-01 -2.35215008e-01 9.34328616e-01
9.87246811e-01 -9.70315218e-01 3.06824356e-01 9.08384994e-02
5.03938377e-01 -3.70724112e-01 7.09837139e-01 7.01990962e-01
-8.87394249e-01 -1.12610213e-01 -3.22746962e-01 -3.02962899e-01
5.02180338e-01 -4.95798001e-03 -1.74606121e+00 3.21590118e-02
2.92572975e-01 -1.70300767e-01 -2.98436642e-01 3.95469159e-01
-5.33007741e-01 5.82216322e-01 -1.41334891e-01 -3.03748876e-01
2.31948480e-01 -7.96108916e-02 4.40841079e-01 1.48885882e+00
8.97211432e-02 6.08484387e-01 4.21107084e-01 5.93990684e-01
-1.01165205e-01 1.01681195e-01 -6.51891291e-01 5.34365140e-02
6.06080055e-01 1.22905850e+00 -3.16022903e-01 -5.55929184e-01
-1.42281368e-01 8.12901914e-01 3.36455435e-01 3.37500453e-01
-3.43586296e-01 -3.51639211e-01 6.21253312e-01 -1.84906125e-01
-1.52658328e-01 -1.83020681e-01 -2.24477127e-01 -9.53476250e-01
-5.04237056e-01 -1.12096000e+00 4.40009445e-01 -5.74724138e-01
-1.36682737e+00 9.37882066e-01 2.65232563e-01 -7.95956075e-01
-1.07653093e+00 -5.43092668e-01 -6.67376399e-01 1.19698060e+00
-1.06584466e+00 -9.12519097e-01 -2.90003389e-01 7.28892028e-01
1.35990548e+00 -1.84602544e-01 1.16813803e+00 7.85252675e-02
-2.12418243e-01 5.32852173e-01 -4.64444220e-01 5.81147112e-02
1.33814430e+00 -1.41900814e+00 4.80587661e-01 4.88227069e-01
-1.18854158e-01 1.05671275e+00 9.40450251e-01 -7.12475121e-01
-9.92716372e-01 -7.72316694e-01 1.29091978e+00 -8.65433216e-01
6.64009869e-01 -3.65048915e-01 -9.64506209e-01 7.76538908e-01
8.85320306e-01 -9.07055020e-01 6.15803897e-01 2.60512322e-01
-9.33364406e-02 2.80492097e-01 -1.12101758e+00 8.55510950e-01
9.25371170e-01 -4.75948185e-01 -1.13291264e+00 4.36354071e-01
1.10095465e+00 -4.53373909e-01 -7.21266747e-01 2.47839227e-01
1.70252949e-01 -7.46452868e-01 7.44158387e-01 -8.31454217e-01
6.34277463e-02 6.12654798e-02 -9.46162865e-02 -1.77253640e+00
-1.93083018e-01 -1.12847888e+00 7.14727938e-02 1.37919414e+00
5.72695673e-01 -6.42419636e-01 4.16085839e-01 1.05471444e+00
-2.29486778e-01 -1.48276910e-01 -5.35530567e-01 -5.10753095e-01
3.79033893e-01 -1.39565125e-01 6.49942398e-01 7.68216670e-01
6.61673605e-01 1.21692348e+00 -3.84843022e-01 -1.78859815e-01
-4.92098890e-02 -6.66665509e-02 1.12042260e+00 -1.23011136e+00
-3.32415581e-01 -4.23532665e-01 5.23593247e-01 -1.46908021e+00
3.50923359e-01 -6.43050909e-01 6.07341945e-01 -1.44512665e+00
-1.24224022e-01 -4.67156231e-01 1.66195273e-01 4.72284943e-01
-5.56689560e-01 -3.61826092e-01 3.55863661e-01 1.75472721e-01
-3.61847758e-01 4.81026769e-01 1.21426988e+00 -1.22195035e-01
-6.40233755e-01 3.47797632e-01 -7.64611185e-01 7.04474926e-01
1.08346748e+00 -2.30372727e-01 -7.15752900e-01 -3.14414889e-01
-1.01308793e-01 3.51113915e-01 -5.32418452e-02 -4.37554747e-01
4.71261144e-01 -6.14703298e-01 -1.78495541e-01 -2.39921108e-01
6.27747715e-01 -5.76189280e-01 -2.40942225e-01 3.92699212e-01
-1.00407159e+00 1.63408756e-01 2.08765000e-01 2.87400991e-01
-1.51437297e-01 -5.34378648e-01 6.71483755e-01 -5.38876951e-01
-7.72088170e-01 -1.38969019e-01 -7.84333229e-01 2.52287865e-01
8.02400768e-01 2.66755611e-01 -6.14405096e-01 -7.87132144e-01
-6.20352447e-01 3.46463710e-01 4.96448774e-04 5.94228029e-01
4.00834590e-01 -1.03694987e+00 -7.52292573e-01 5.28401062e-02
6.59318967e-03 -1.05517760e-01 8.47257208e-03 4.69372272e-01
-9.93860513e-02 9.73626435e-01 -1.35326192e-01 -3.06729317e-01
-1.36157382e+00 4.90460664e-01 8.46052542e-02 -6.27625406e-01
-2.93202102e-01 9.62740660e-01 5.60436070e-01 -8.96338522e-01
2.25635305e-01 -3.81606609e-01 -4.12786782e-01 1.88571602e-01
5.20598114e-01 1.12226754e-01 -1.40041128e-01 -2.53108293e-01
-6.40753880e-02 8.01562145e-02 -2.94516265e-01 -7.47518003e-01
9.44331646e-01 -3.53628129e-01 9.05195400e-02 4.04241443e-01
6.62087381e-01 1.39117941e-01 -1.08779776e+00 -5.62836468e-01
3.90439689e-01 -3.46680969e-01 -6.43820047e-01 -1.20487094e+00
-3.37883413e-01 9.20470059e-01 1.45895138e-01 6.62956893e-01
7.57535517e-01 1.29159346e-01 7.50931978e-01 1.05525315e+00
5.74713409e-01 -1.10931849e+00 3.98466110e-01 1.17120266e+00
1.17172039e+00 -1.52155280e+00 -5.20333052e-01 -2.54830301e-01
-1.06548941e+00 1.09447873e+00 1.14784849e+00 3.28457803e-01
1.68569356e-01 1.24702811e-01 5.92210412e-01 -2.68913899e-02
-1.17656708e+00 -1.84904262e-01 -6.42614858e-03 6.10492229e-01
7.73568749e-01 1.37628645e-01 -3.84833038e-01 6.16194367e-01
-5.94413638e-01 -2.42723390e-01 5.06466568e-01 6.79543972e-01
-7.15992033e-01 -1.44807458e+00 -5.00573099e-01 1.62424177e-01
-1.85976297e-01 -2.45596379e-01 -8.58165681e-01 5.34992635e-01
-4.51765567e-01 1.59630466e+00 -3.62333506e-01 -4.01577771e-01
5.92264652e-01 5.90414882e-01 1.38159946e-01 -1.14986014e+00
-1.26882052e+00 -6.20289557e-02 5.88900268e-01 -3.05761427e-01
-1.84928373e-01 -3.23936075e-01 -1.27045906e+00 -3.14234614e-01
-2.91374862e-01 6.15528941e-01 3.34544659e-01 9.65460420e-01
1.32169768e-01 5.78579962e-01 7.04218745e-01 -8.53679538e-01
-1.09359324e+00 -1.66458607e+00 5.93785420e-02 3.67427617e-01
1.96960732e-01 -4.84371156e-01 -3.61243188e-01 2.73637325e-01]
|
[12.804956436157227, 8.045710563659668]
|
91d49b38-b6e8-4433-867d-d60cd73e7365
|
forward-stagewise-additive-model-for
|
1608.01874
| null |
http://arxiv.org/abs/1608.01874v1
|
http://arxiv.org/pdf/1608.01874v1.pdf
|
Forward Stagewise Additive Model for Collaborative Multiview Boosting
|
Multiview assisted learning has gained significant attention in recent years
in supervised learning genre. Availability of high performance computing
devices enables learning algorithms to search simultaneously over multiple
views or feature spaces to obtain an optimum classification performance. The
paper is a pioneering attempt of formulating a mathematical foundation for
realizing a multiview aided collaborative boosting architecture for multiclass
classification. Most of the present algorithms apply multiview learning
heuristically without exploring the fundamental mathematical changes imposed on
traditional boosting. Also, most of the algorithms are restricted to two class
or view setting. Our proposed mathematical framework enables collaborative
boosting across any finite dimensional view spaces for multiclass learning. The
boosting framework is based on forward stagewise additive model which minimizes
a novel exponential loss function. We show that the exponential loss function
essentially captures difficulty of a training sample space instead of the
traditional `1/0' loss. The new algorithm restricts a weak view from over
learning and thereby preventing overfitting. The model is inspired by our
earlier attempt on collaborative boosting which was devoid of mathematical
justification. The proposed algorithm is shown to converge much nearer to
global minimum in the exponential loss space and thus supersedes our previous
algorithm. The paper also presents analytical and numerical analysis of
convergence and margin bounds for multiview boosting algorithms and we show
that our proposed ensemble learning manifests lower error bound and higher
margin compared to our previous model. Also, the proposed model is compared
with traditional boosting and recent multiview boosting algorithms.
|
['Prabir Kumar Biswas', 'Avisek Lahiri', 'Biswajit Paria']
|
2016-08-05
| null | null | null | null |
['multiview-learning']
|
['computer-vision']
|
[ 1.57171622e-01 -1.47496611e-01 -5.53292155e-01 -7.92515218e-01
-1.06627512e+00 -6.53699040e-01 4.27846998e-01 3.31082731e-01
-2.91233808e-01 7.77868211e-01 -1.31894842e-01 -3.88723582e-01
-3.95434231e-01 -6.61895573e-01 -6.68525696e-01 -9.27224219e-01
2.04906263e-03 1.62177667e-01 -1.05312891e-01 -3.27922046e-01
2.70620078e-01 2.67367691e-01 -1.84088457e+00 3.56698185e-01
8.68052185e-01 1.26649868e+00 5.25785349e-02 8.71014953e-01
9.14258659e-02 6.58855259e-01 -4.26814169e-01 -7.41469145e-01
4.04472470e-01 -2.69320637e-01 -3.54021192e-01 -3.42149772e-02
9.21732426e-01 -1.14050500e-01 3.80490035e-01 8.26236248e-01
6.29667044e-01 1.66258574e-01 8.59917939e-01 -1.60728967e+00
-7.74328649e-01 3.42064649e-02 -7.33565271e-01 2.51073420e-01
1.72989979e-01 -7.87568748e-01 1.07397270e+00 -1.26796651e+00
3.10920358e-01 8.77826929e-01 9.84347343e-01 3.85677576e-01
-1.01788676e+00 -6.50830865e-01 4.88768846e-01 5.14651179e-01
-1.06195581e+00 -1.20717481e-01 1.07712483e+00 -3.95219058e-01
5.69836259e-01 4.97504681e-01 6.15714431e-01 6.61214173e-01
5.34374475e-01 8.91641676e-01 1.41771245e+00 -8.83378744e-01
2.75781035e-01 5.60394585e-01 4.65675592e-01 8.76119852e-01
7.14143366e-02 2.45126989e-02 -6.67580843e-01 -2.76990384e-01
2.69559294e-01 3.38871181e-01 -2.70318128e-02 -1.03259826e+00
-5.36970735e-01 1.00855422e+00 4.07065243e-01 1.23022482e-01
-1.51502958e-04 -3.06405634e-01 5.45223236e-01 5.23689330e-01
8.46701145e-01 -4.09297228e-01 -3.67914677e-01 4.09452707e-01
-7.50011563e-01 1.62790969e-01 4.48072374e-01 7.96032965e-01
6.24136031e-01 1.48947584e-02 4.69797850e-01 1.14024901e+00
2.60615230e-01 4.13140267e-01 2.97368258e-01 -8.13459337e-01
5.54143488e-01 3.90518248e-01 -1.41910312e-03 -8.51904273e-01
-1.88263223e-01 -8.21316481e-01 -8.62372160e-01 7.20058978e-01
3.20929497e-01 1.06858544e-01 -2.77907878e-01 1.45815921e+00
7.26443827e-01 -3.82970512e-01 -2.48235967e-02 6.08787417e-01
5.63393891e-01 3.72989088e-01 -7.26481155e-02 -6.09929025e-01
9.48716879e-01 -1.06949055e+00 -7.57300377e-01 2.09292039e-01
5.13725221e-01 -6.89947486e-01 8.32203329e-01 8.77862036e-01
-1.05855107e+00 -6.61915183e-01 -1.50608361e+00 2.70697773e-01
-4.37364846e-01 -3.07786584e-01 6.49846613e-01 1.15740919e+00
-9.45239127e-01 3.70758951e-01 -2.61910319e-01 -6.74997047e-02
3.48694265e-01 3.25894982e-01 -3.86300147e-01 -8.06321427e-02
-8.09219837e-01 8.36163998e-01 -3.54837552e-02 2.15476728e-03
-4.94197130e-01 -7.73689330e-01 -6.23303711e-01 -4.66455758e-01
1.60778295e-02 -8.57491970e-01 1.04182851e+00 -1.43450010e+00
-1.25005782e+00 7.95566261e-01 -3.58801186e-01 -3.66260111e-01
8.08158338e-01 -2.93057054e-01 -3.55408102e-01 -9.03305933e-02
1.63926966e-02 5.09991460e-02 1.12951517e+00 -1.51955926e+00
-7.64896870e-01 -1.03202999e+00 1.30449347e-02 5.32423437e-01
-4.30671722e-01 -3.28718007e-01 3.40374261e-01 -8.48239124e-01
3.18369627e-01 -3.93896818e-01 -1.96683511e-01 6.81166798e-02
4.37503785e-01 -2.91111171e-01 1.20327652e+00 -3.81418735e-01
9.05546308e-01 -1.85206223e+00 6.21411204e-02 3.74171808e-02
1.73394546e-01 -6.43559620e-02 3.94252807e-01 6.04282260e-01
-3.00907940e-01 -1.55710876e-01 -8.52298290e-02 -3.25132668e-01
-2.43970752e-01 9.96394902e-02 -4.61115122e-01 8.72961998e-01
-6.94283426e-01 5.04580677e-01 -6.18945599e-01 -6.33062065e-01
4.09023583e-01 3.96704376e-01 -5.68557620e-01 6.13830313e-02
3.91780466e-01 6.20101243e-02 -1.21505022e-01 7.82669663e-01
9.31054413e-01 -8.22128952e-02 2.79940009e-01 -1.45126283e-01
2.13614255e-02 -4.57322657e-01 -1.23068881e+00 1.41449523e+00
-9.59213436e-01 3.07145983e-01 3.59970719e-01 -1.63507521e+00
1.08301198e+00 3.49079639e-01 5.24718881e-01 -3.91316324e-01
-7.73919374e-02 1.28157035e-01 -6.26235783e-01 -3.31188366e-02
2.17146784e-01 -7.80575395e-01 2.61109889e-01 3.45163584e-01
2.34750628e-01 1.19884335e-01 -5.50302327e-01 1.06512286e-01
3.12473744e-01 1.03174858e-01 7.01012969e-01 -5.00159562e-01
8.20555806e-01 -3.26522261e-01 4.11835581e-01 9.37726080e-01
-4.86238062e-01 2.88938463e-01 -1.87742636e-01 -7.60321081e-01
-8.29607666e-01 -1.39717686e+00 -4.49997962e-01 1.61882269e+00
2.96345204e-01 -1.95098191e-01 -2.31548831e-01 -1.21200633e+00
1.63131833e-01 4.25564915e-01 -7.08197594e-01 2.47824818e-01
-3.33452225e-01 -8.22013378e-01 -3.77125442e-02 6.13903821e-01
4.48498458e-01 -1.91352367e-01 -2.66881973e-01 1.14399582e-01
-6.24091700e-02 -5.07369518e-01 -1.17393307e-01 3.47054571e-01
-1.37151885e+00 -1.03381133e+00 -6.71696842e-01 -8.35495293e-01
5.13814688e-01 5.86402059e-01 9.81133461e-01 -2.03277081e-01
-2.35191405e-01 9.58987892e-01 -4.72039521e-01 -8.60364318e-01
-1.72824368e-01 -3.02130193e-01 2.59483725e-01 8.87186229e-02
2.15662181e-01 -7.41294861e-01 -7.37337351e-01 4.28585798e-01
-4.78808612e-01 -2.13840052e-01 7.10322186e-02 1.14578354e+00
5.30086279e-01 -1.18234478e-01 1.26078427e+00 -8.02953303e-01
1.67788431e-01 -4.23204005e-01 -6.56128168e-01 3.77018899e-01
-1.05133593e+00 -2.42067635e-01 7.92122126e-01 -1.81881472e-01
-1.14725876e+00 -6.11448521e-03 -1.21006794e-01 -2.70146489e-01
1.53152436e-01 -6.43787161e-02 -9.21223760e-02 -2.89813042e-01
7.27245450e-01 2.72463977e-01 1.16141951e-02 -4.31107163e-01
4.66374934e-01 8.42283726e-01 1.87184125e-01 -5.64187527e-01
4.26037103e-01 7.54705310e-01 2.26249248e-01 -7.41332054e-01
-9.82690156e-01 -7.31355667e-01 -7.10723519e-01 -7.22429872e-01
4.91990030e-01 -9.32318389e-01 -9.29844022e-01 2.24232420e-01
-7.27969050e-01 3.40411931e-01 -7.05950037e-02 4.61549550e-01
-9.20076191e-01 5.24899900e-01 -2.46524453e-01 -1.38485384e+00
-6.26443565e-01 -7.98515081e-01 7.32452154e-01 -1.57550961e-01
2.29452446e-01 -1.45048726e+00 6.42831996e-02 8.72025967e-01
1.45960689e-01 3.30303192e-01 8.00937831e-01 -5.49918890e-01
-1.18996970e-01 -3.78658682e-01 1.81352973e-01 6.55263364e-01
1.02209263e-01 -3.99441004e-01 -1.39008510e+00 -6.42223835e-01
5.83086371e-01 -4.91271913e-01 8.02490771e-01 5.19595802e-01
1.17855048e+00 -1.70514211e-01 -2.40904406e-01 4.94457871e-01
1.87477624e+00 4.48401958e-01 6.12599328e-02 4.38684076e-01
4.08390522e-01 6.69583619e-01 7.46045172e-01 4.41244185e-01
3.03508312e-01 6.01141334e-01 6.10104144e-01 6.24152534e-02
3.40732813e-01 -3.16145569e-02 2.19422713e-01 1.00743401e+00
-2.08254397e-01 9.95400846e-02 -6.74730659e-01 4.49559182e-01
-1.87328660e+00 -1.19368052e+00 -1.25873461e-01 2.72799468e+00
3.49430829e-01 -8.91048536e-02 3.01983893e-01 4.46967691e-01
8.67708266e-01 1.80461347e-01 -4.15570974e-01 -8.34535062e-01
-2.01644879e-02 1.62205398e-01 4.42437261e-01 9.22763407e-01
-1.35368240e+00 1.12097368e-01 6.81967640e+00 8.78179312e-01
-8.53431582e-01 4.95269150e-01 7.33072817e-01 -7.24747851e-02
-8.19016248e-02 -1.54142380e-01 -9.09768999e-01 2.60926243e-02
5.40511310e-01 -1.27788290e-01 3.77665013e-02 1.30527020e+00
-2.89032668e-01 3.05865612e-03 -1.06898522e+00 1.18592727e+00
4.29521650e-01 -1.32760918e+00 7.39789978e-02 3.63986865e-02
8.28050137e-01 -2.94584095e-01 2.28566721e-01 3.27361166e-01
9.58901569e-02 -7.38104761e-01 6.47299230e-01 2.96856821e-01
5.10134697e-01 -1.24585760e+00 6.57193899e-01 7.41196334e-01
-1.27462971e+00 -4.51460838e-01 -2.17271551e-01 -2.29603887e-01
-2.92707801e-01 6.62017941e-01 -6.23399079e-01 1.05622089e+00
5.54502249e-01 7.38027334e-01 -3.93010050e-01 6.98214591e-01
4.46263075e-01 3.37315947e-01 6.46493733e-02 -7.70391757e-03
-5.95211983e-03 -3.65727127e-01 6.08861804e-01 1.05163336e+00
1.46025568e-01 -5.71938828e-02 1.61685348e-01 -3.36785242e-02
2.24819645e-01 5.79640806e-01 -9.88065720e-01 9.62361515e-01
2.39945307e-01 1.05854499e+00 -5.01318574e-01 -4.04218346e-01
-7.22710907e-01 8.01556051e-01 5.36241829e-01 1.18920997e-01
-7.57497847e-01 -1.32193834e-01 1.82155699e-01 -7.43272454e-02
3.65666658e-01 1.27248660e-01 -6.20785594e-01 -1.21640956e+00
2.37789214e-01 -8.35153282e-01 8.05048823e-01 -2.51152068e-01
-1.55655050e+00 2.66743422e-01 7.18889236e-02 -1.35424221e+00
-1.20059557e-01 -5.28819978e-01 -3.99338841e-01 8.23960066e-01
-1.06681728e+00 -1.32151282e+00 -6.69690818e-02 7.02103496e-01
8.44084084e-01 -4.13053244e-01 7.97286570e-01 2.31101751e-01
-4.07769978e-02 9.08429742e-01 5.30555785e-01 -4.96768266e-01
9.16678190e-01 -1.37116206e+00 -3.45871449e-01 4.51887071e-01
-4.08579148e-02 4.52757180e-01 7.37401783e-01 -3.09225112e-01
-1.38090527e+00 -8.69904041e-01 7.23772526e-01 -7.21045971e-01
3.56343806e-01 -4.31700170e-01 -6.59631431e-01 6.00582540e-01
2.57495731e-01 1.18669115e-01 1.21343505e+00 5.47898650e-01
-5.73211014e-01 -6.92212045e-01 -1.43544495e+00 -2.83264965e-02
7.73489892e-01 -4.59853858e-01 -6.31780148e-01 3.30826044e-01
1.41652212e-01 -1.19918697e-02 -1.00507426e+00 6.37686610e-01
1.20903361e+00 -1.38465035e+00 1.22440612e+00 -5.38114786e-01
2.39388973e-01 2.65857056e-02 -8.37493002e-01 -1.16516376e+00
-1.17407367e-01 -1.40934989e-01 -1.32225960e-01 1.06887960e+00
1.39418229e-01 -9.58178282e-01 9.68435228e-01 2.65650749e-02
-1.72521397e-02 -1.17744350e+00 -1.21018398e+00 -9.24801826e-01
3.88696849e-01 -5.37608206e-01 2.56257989e-02 1.13535798e+00
-5.05919047e-02 2.67386764e-01 -3.81409794e-01 8.73173624e-02
1.36126757e+00 4.32675213e-01 7.04441428e-01 -1.24986458e+00
-4.36749905e-01 -2.25261942e-01 -5.01919389e-01 -7.54680216e-01
3.09524965e-03 -1.09358072e+00 -5.25206327e-01 -1.23271620e+00
3.81149799e-01 -6.55385554e-01 -5.65933824e-01 9.34082270e-03
-2.37879440e-01 3.74298692e-01 3.02984864e-01 4.47715484e-02
-4.40947831e-01 4.80596364e-01 1.07764316e+00 -1.40707016e-01
-2.27970760e-02 3.62231433e-01 -6.17300451e-01 8.72121036e-01
7.00918257e-01 -3.00027788e-01 -7.43108749e-01 5.64426696e-03
2.62116760e-01 3.88840348e-01 4.62522388e-01 -7.70625234e-01
1.06297441e-01 2.52109040e-02 7.45794237e-01 -9.02285933e-01
4.65528220e-01 -1.03492081e+00 -5.09766825e-02 4.04618919e-01
-1.90558270e-01 1.98986769e-01 -1.71804741e-01 1.04613304e+00
-2.84705371e-01 -2.31134027e-01 1.04235494e+00 -5.29266037e-02
-4.46284384e-01 -9.11483392e-02 -2.04017669e-01 -1.37397051e-01
1.26217496e+00 -4.11500722e-01 -3.91401686e-02 -5.87779164e-01
-1.13861036e+00 -4.43143994e-02 2.38110140e-01 2.57373750e-01
5.62536240e-01 -1.51019371e+00 -5.36711991e-01 1.53728306e-01
4.60340194e-02 -5.00904441e-01 4.66116607e-01 6.93160832e-01
-1.87634364e-01 2.06511542e-01 -1.95536062e-01 -8.30047309e-01
-1.85213101e+00 8.38828802e-01 3.60436589e-01 -2.28460178e-01
-3.92144591e-01 8.57135117e-01 4.05416846e-01 -5.65718055e-01
2.55245447e-01 3.73113692e-01 -1.10700890e-01 3.75933558e-01
6.09071970e-01 8.25327933e-01 4.88117397e-01 -4.35188025e-01
-4.31532592e-01 6.95423484e-01 -2.48883605e-01 -1.28182516e-01
1.28065920e+00 -2.74188519e-01 1.57430004e-02 9.13118660e-01
1.48006594e+00 -2.23368574e-02 -9.02984321e-01 -9.32250917e-02
-4.01973456e-01 -5.34941792e-01 7.41777942e-02 -7.78152704e-01
-5.42286932e-01 8.49290907e-01 1.19523752e+00 2.11505711e-01
1.34738636e+00 -2.49666274e-01 1.69725433e-01 2.55373269e-01
7.23511398e-01 -1.14749098e+00 4.73684445e-02 1.71794295e-01
9.95798349e-01 -1.71377933e+00 3.70198995e-01 -3.32284421e-01
-4.30450290e-01 1.21174431e+00 4.81104195e-01 -2.40221143e-01
9.58528399e-01 2.07422659e-01 1.45268038e-01 1.22674458e-01
-8.19467902e-01 4.11678940e-01 3.40153098e-01 7.85193145e-01
5.06846845e-01 1.33940056e-01 -4.81598258e-01 5.07893503e-01
-3.69090252e-02 -1.35860071e-01 1.46203697e-01 1.14388478e+00
-6.08786106e-01 -1.19549155e+00 -6.92798495e-01 4.09101158e-01
-6.77877009e-01 1.85717359e-01 -1.17280316e-02 8.66852760e-01
2.66739309e-01 1.01283252e+00 -2.66551107e-01 -2.89010942e-01
1.21055588e-01 3.80860329e-01 8.16112757e-01 -1.32038817e-01
-5.45738637e-01 -3.86754014e-02 -2.55056918e-01 -1.75834280e-02
-6.95746005e-01 -8.10748041e-01 -5.14688730e-01 -1.27461895e-01
-5.50959766e-01 3.08761328e-01 9.83422339e-01 6.67054355e-01
-8.01926553e-02 3.28445584e-02 1.08797526e+00 -7.76811063e-01
-7.23983824e-01 -4.98297513e-01 -7.06624508e-01 7.53851533e-02
6.45348132e-01 -7.42876351e-01 -4.10594046e-01 1.24900695e-02]
|
[8.436976432800293, 4.437785625457764]
|
2b0065dc-74b8-4a94-b81f-6c1c955f4321
|
fusionnet-a-deep-fully-residual-convolutional
|
1612.05360
| null |
http://arxiv.org/abs/1612.05360v2
|
http://arxiv.org/pdf/1612.05360v2.pdf
|
FusionNet: A deep fully residual convolutional neural network for image segmentation in connectomics
|
Electron microscopic connectomics is an ambitious research direction with the
goal of studying comprehensive brain connectivity maps by using
high-throughput, nano-scale microscopy. One of the main challenges in
connectomics research is developing scalable image analysis algorithms that
require minimal user intervention. Recently, deep learning has drawn much
attention in computer vision because of its exceptional performance in image
classification tasks. For this reason, its application to connectomic analyses
holds great promise, as well. In this paper, we introduce a novel deep neural
network architecture, FusionNet, for the automatic segmentation of neuronal
structures in connectomics data. FusionNet leverages the latest advances in
machine learning, such as semantic segmentation and residual neural networks,
with the novel introduction of summation-based skip connections to allow a much
deeper network architecture for a more accurate segmentation. We demonstrate
the performance of the proposed method by comparing it with state-of-the-art
electron microscopy (EM) segmentation methods from the ISBI EM segmentation
challenge. We also show the segmentation results on two different tasks
including cell membrane and cell body segmentation and a statistical analysis
of cell morphology.
|
['Won-Ki Jeong', 'Tran Minh Quan', 'David G. C. Hildebrand']
|
2016-12-16
| null | null | null | null |
['brain-image-segmentation']
|
['medical']
|
[ 1.19712763e-01 -1.39896646e-01 3.10954094e-01 -3.99853736e-01
-3.49624217e-01 -2.88248122e-01 2.07020164e-01 2.24648416e-01
-1.04929721e+00 7.89862394e-01 -3.53972554e-01 -2.10745111e-01
9.58285555e-02 -5.19252121e-01 -6.17482066e-01 -8.84395182e-01
4.49583009e-02 8.30220282e-01 3.01041305e-01 2.91050728e-02
3.34957361e-01 6.81689501e-01 -1.03480148e+00 7.61258556e-03
8.18819940e-01 9.27669764e-01 5.84922969e-01 4.07194436e-01
-2.28468433e-01 2.53473163e-01 -1.66390017e-01 -3.13067615e-01
1.49278510e-02 -2.52427459e-01 -1.14307380e+00 -1.16567768e-01
4.09301460e-01 -1.12556390e-01 1.27377138e-01 1.30245030e+00
7.22760499e-01 -5.50333336e-02 5.79330146e-01 -7.34146476e-01
-3.95172387e-01 5.09448051e-01 -5.71198881e-01 7.34479785e-01
-2.62996227e-01 1.58173963e-02 7.98197746e-01 -4.40189481e-01
8.95682335e-01 9.92665708e-01 5.60532331e-01 6.23680592e-01
-1.67213297e+00 -6.58139348e-01 2.28922442e-02 4.57681865e-01
-1.20640635e+00 -1.89164415e-01 5.36868155e-01 -7.27415562e-01
1.15332699e+00 -2.80155450e-01 9.69746888e-01 6.80930734e-01
4.50570285e-01 6.30934894e-01 1.27564752e+00 6.54555857e-02
4.89890188e-01 -3.95385355e-01 2.33072579e-01 6.91735983e-01
1.06503449e-01 -5.66296339e-01 1.38053328e-01 1.76051795e-01
1.01760066e+00 2.49365032e-01 -1.08474508e-01 -1.23830684e-01
-1.46167898e+00 6.78931117e-01 6.84769213e-01 7.33205259e-01
-3.26432645e-01 2.02568665e-01 4.84320849e-01 -1.73394866e-02
5.98955929e-01 3.80795360e-01 -5.12677789e-01 2.62696594e-01
-1.22609520e+00 1.12290956e-01 3.92460048e-01 4.02094841e-01
8.09892893e-01 -1.12857804e-01 1.39132038e-01 8.88065338e-01
2.75499374e-01 1.31223753e-01 6.39048040e-01 -1.02539182e+00
1.57376140e-01 7.45765388e-01 -4.28915590e-01 -6.06341481e-01
-1.06001687e+00 -2.23379210e-01 -1.14828372e+00 3.89065295e-01
4.98032123e-01 -1.36775881e-01 -8.82063389e-01 1.74559784e+00
4.10654545e-01 2.90869717e-02 -3.64842951e-01 1.03797913e+00
6.77501023e-01 2.26002753e-01 1.45371258e-01 -6.86229318e-02
1.37752926e+00 -7.24008322e-01 -5.23641884e-01 3.07278447e-02
6.31666899e-01 -4.05699104e-01 5.70493579e-01 1.38390452e-01
-9.11201835e-01 -1.54433399e-01 -9.77494299e-01 -2.20084846e-01
-6.59287691e-01 -1.28504977e-01 7.63566375e-01 2.57886320e-01
-1.39773822e+00 8.48118365e-01 -1.34383118e+00 -7.85973430e-01
1.04686272e+00 8.39551568e-01 -5.97288430e-01 3.41641873e-01
-5.89612424e-01 7.21043110e-01 4.20104265e-01 -2.74154563e-02
-6.24152303e-01 -7.28287518e-01 -5.91087580e-01 2.08578408e-01
-1.08270854e-01 -9.50760305e-01 1.00376213e+00 -5.42091191e-01
-1.39106107e+00 1.33909583e+00 -1.50832176e-01 -6.91942573e-01
2.25491479e-01 1.46634057e-01 9.72424075e-02 6.65900886e-01
1.65336266e-01 1.33298576e+00 2.75927097e-01 -6.93401217e-01
-3.18236411e-01 -8.43665421e-01 -3.17102700e-01 -1.72336459e-01
-2.43981779e-01 1.11106418e-01 -1.42317638e-01 -3.09507310e-01
2.32019931e-01 -6.67028368e-01 -4.30088162e-01 -6.00217544e-02
-5.48219144e-01 -1.97986647e-01 7.89691567e-01 -6.10397756e-01
5.49007177e-01 -1.69644237e+00 4.60749775e-01 -9.87734739e-03
8.07349682e-01 1.59093738e-01 -1.49675517e-03 1.19151380e-02
-9.40356255e-02 2.45515957e-01 -3.15673649e-01 -4.58929926e-01
-3.79929215e-01 -1.54226840e-01 3.99052829e-01 7.32388735e-01
1.46812469e-01 1.15037441e+00 -5.57134271e-01 -6.16708934e-01
2.98012376e-01 4.76440519e-01 -5.35165966e-01 1.96158560e-03
-1.14270687e-01 9.66873944e-01 -2.15063900e-01 4.86020863e-01
6.82232499e-01 -6.00720286e-01 2.56122738e-01 -2.47733876e-01
-2.84709454e-01 -1.20507553e-01 -4.83819544e-01 1.94087064e+00
-1.38469547e-01 7.20174372e-01 5.89031935e-01 -1.44421124e+00
5.87593675e-01 3.38421494e-01 8.27413142e-01 -7.95183957e-01
5.54450810e-01 2.59874851e-01 2.98099369e-01 -3.04638594e-01
-2.64263600e-01 -4.19138640e-01 4.68612939e-01 4.79408354e-01
3.69241148e-01 -1.23245150e-01 5.18188298e-01 1.40178263e-01
1.10896170e+00 -1.11329049e-01 -7.31744757e-03 -7.33567238e-01
5.54796040e-01 -3.00212782e-02 5.72220027e-01 2.59957701e-01
-3.48376900e-01 5.77205300e-01 4.14479584e-01 -5.56761742e-01
-1.30781651e+00 -8.58833551e-01 -4.08858687e-01 7.15765834e-01
1.72962900e-02 7.95530155e-02 -1.35407770e+00 -3.74651164e-01
-2.43771061e-01 -1.56485140e-01 -5.97795844e-01 3.27643096e-01
-5.59315383e-01 -1.13632321e+00 4.92941260e-01 4.75011975e-01
6.68819726e-01 -1.32317364e+00 -6.47884548e-01 3.25025260e-01
5.42749418e-03 -1.40094328e+00 -1.05003670e-01 3.45024049e-01
-1.24064052e+00 -1.11597145e+00 -9.91116464e-01 -1.12241912e+00
9.93582606e-01 1.01071157e-01 7.53860116e-01 2.68715441e-01
-6.58350408e-01 -3.16619463e-02 1.28010094e-01 2.14851983e-02
-7.24891573e-03 4.27737176e-01 2.20101643e-02 -3.05190206e-01
3.42036396e-01 -1.07027566e+00 -8.58019114e-01 2.28747234e-01
-9.22322690e-01 2.93092400e-01 4.91096020e-01 8.48725796e-01
1.00653017e+00 -1.24637984e-01 6.99259162e-01 -9.26936626e-01
5.02993047e-01 -4.06199783e-01 -6.95985496e-01 -5.39470501e-02
-4.46732044e-01 -1.08549498e-01 9.05579627e-01 -4.36850451e-02
-6.19129419e-01 -9.66969207e-02 -6.83626950e-01 -1.79910064e-01
-3.69851768e-01 4.58217144e-01 -9.87207294e-02 -3.82245928e-01
1.64266661e-01 7.23797530e-02 3.25793982e-01 -4.47155893e-01
4.97361310e-02 2.50733346e-01 4.62485999e-01 -4.37609643e-01
1.76859930e-01 8.03929627e-01 2.96825320e-01 -8.49887788e-01
-6.40924990e-01 -6.71980381e-01 -1.02049804e+00 -4.67710532e-02
1.45889843e+00 -6.08762920e-01 -1.04808319e+00 6.59506500e-01
-1.09472287e+00 -5.48206568e-01 1.71930596e-01 3.73809516e-01
-7.03833699e-01 4.16270465e-01 -1.17942929e+00 -2.31470525e-01
-7.87720859e-01 -1.44918334e+00 1.01167595e+00 4.12397206e-01
-3.96451578e-02 -1.26692688e+00 2.06307456e-01 4.18708384e-01
3.99072707e-01 1.98987752e-01 1.24649715e+00 -6.41531110e-01
-4.46209520e-01 5.80150336e-02 -4.66315478e-01 2.04254836e-01
-8.00705031e-02 -1.26371667e-01 -7.43624389e-01 -2.83905953e-01
-1.58589140e-01 -3.14541668e-01 1.13630772e+00 7.78846800e-01
1.34993970e+00 1.81728542e-01 -5.03299773e-01 1.04433477e+00
1.48038614e+00 3.26482058e-02 4.75531727e-01 4.58156347e-01
8.79020035e-01 6.82592332e-01 -2.51850784e-01 -1.50449714e-02
3.33972663e-01 5.39505064e-01 4.73281920e-01 -4.82167959e-01
-5.24289608e-02 4.21040386e-01 -2.57770360e-01 9.06038225e-01
-2.31907502e-01 1.75096124e-01 -1.04098547e+00 4.74001676e-01
-1.86968553e+00 -9.76722777e-01 -1.43851444e-01 1.77006423e+00
6.73196912e-01 -1.04351752e-01 2.29047850e-01 -5.37802875e-02
9.34373975e-01 -2.55575746e-01 -8.32735777e-01 -1.76368013e-01
-1.87284589e-01 4.02187437e-01 2.81255096e-01 2.16213539e-01
-1.19158351e+00 1.01165211e+00 6.47554874e+00 7.24568903e-01
-1.20062888e+00 3.26706201e-01 1.07370877e+00 -5.33248410e-02
1.06996752e-01 -4.10357445e-01 -7.64080644e-01 4.86818194e-01
7.28756726e-01 2.50277877e-01 6.86106980e-01 3.66826653e-01
3.52726221e-01 -1.30707160e-01 -1.02676749e+00 1.03737247e+00
-3.47576171e-01 -1.79310644e+00 -6.91341460e-02 2.74073392e-01
7.37146974e-01 7.30006754e-01 -1.11133918e-01 -1.36270747e-01
-9.13929939e-03 -1.13462996e+00 3.89216572e-01 2.94454038e-01
6.54495418e-01 -8.19260180e-01 9.44739997e-01 3.54873449e-01
-1.13897467e+00 1.63612828e-01 -6.54176593e-01 -1.40917283e-02
4.29342568e-01 8.02616179e-01 -4.75336909e-01 1.83217525e-01
7.90836096e-01 8.35403383e-01 -4.70272154e-01 1.26305377e+00
7.10722804e-02 3.35278183e-01 -2.74119377e-01 1.15504406e-01
1.33250177e-01 -6.32608831e-01 3.36256474e-01 1.39322579e+00
8.88332352e-02 -1.27153113e-01 1.42096862e-01 1.40105736e+00
-4.20004725e-01 1.38231799e-01 -2.79646456e-01 -2.61441976e-01
1.29276216e-01 1.94659531e+00 -1.58720279e+00 -1.77675530e-01
-3.25255990e-01 6.69694722e-01 8.64514887e-01 2.85652310e-01
-4.13672656e-01 -2.44697064e-01 6.54931545e-01 3.67555097e-02
3.67130339e-01 -4.17953074e-01 -5.67222953e-01 -1.04954290e+00
-4.17495996e-01 -2.46203616e-01 1.10537566e-01 -4.25059140e-01
-1.45770073e+00 4.62672174e-01 -3.86552989e-01 -3.72288913e-01
1.61451548e-01 -8.51730824e-01 -8.45172465e-01 8.07818055e-01
-1.34287143e+00 -9.13422108e-01 -8.23604390e-02 4.22953278e-01
2.52614439e-01 3.24990153e-02 6.42889678e-01 4.91354078e-01
-9.34988320e-01 7.68900663e-02 1.26422241e-01 2.06057653e-01
2.16788411e-01 -1.47055054e+00 4.69540566e-01 6.18737936e-01
-7.00180680e-02 7.98510373e-01 2.98848480e-01 -6.09144807e-01
-1.11385071e+00 -1.10490215e+00 4.44170713e-01 -6.12010360e-02
8.24429572e-01 -4.58040595e-01 -7.88709104e-01 5.46980739e-01
2.06175014e-01 1.36384130e-01 7.66733348e-01 -1.15232207e-01
1.36256441e-01 1.28243193e-01 -1.43174863e+00 4.91329759e-01
8.94936979e-01 -3.92704844e-01 -3.14566493e-01 3.31707090e-01
5.89721382e-01 4.90024090e-02 -1.07148635e+00 2.02374905e-01
5.35688937e-01 -9.76975203e-01 1.00807905e+00 -3.76167059e-01
4.09099519e-01 -2.44386464e-01 1.99543804e-01 -1.29803002e+00
-4.33409393e-01 -2.56824076e-01 1.67878747e-01 1.04578018e+00
2.33923703e-01 -6.39308691e-01 1.00740194e+00 3.56648445e-01
-3.28965992e-01 -1.03933895e+00 -1.07661927e+00 -5.75963736e-01
4.34705079e-01 8.07721242e-02 3.24927419e-01 6.73360944e-01
1.50095731e-01 5.72857201e-01 3.33840162e-01 -4.48063791e-01
7.45520830e-01 7.53383562e-02 2.66472787e-01 -1.45921707e+00
1.15834936e-01 -8.83047879e-01 -7.39664257e-01 -8.48898232e-01
5.27549565e-01 -1.23290789e+00 -1.15381610e-02 -1.83599138e+00
6.32000744e-01 9.28532705e-02 -3.62008929e-01 2.89341718e-01
3.14686587e-03 6.54019713e-01 2.64820606e-02 1.25437036e-01
-8.60472739e-01 4.21557993e-01 1.54817259e+00 -2.36834496e-01
4.98988256e-02 -3.20381969e-01 -5.31451225e-01 7.90589690e-01
1.11383772e+00 -3.44839036e-01 -3.80539186e-02 -5.19640326e-01
5.73019385e-02 -3.31567675e-01 3.47925395e-01 -1.24045348e+00
4.10798579e-01 3.70843053e-01 6.33310318e-01 -5.42548299e-01
1.61094218e-01 -6.71054125e-01 -4.67179865e-02 4.68316495e-01
-1.64809629e-01 4.00681049e-02 1.29160896e-01 2.13743389e-01
-9.67693254e-02 3.67721915e-02 1.14777339e+00 -4.09255326e-01
-3.88399571e-01 7.40163922e-01 -7.43969202e-01 -1.21571831e-02
8.75294983e-01 -2.01341256e-01 -6.03249192e-01 2.45050281e-01
-9.19169128e-01 2.57898241e-01 6.59466028e-01 -3.37369651e-01
5.83481252e-01 -1.02058172e+00 -3.82501394e-01 -2.57893987e-02
-2.72161782e-01 8.59718174e-02 3.24568331e-01 1.44915938e+00
-7.48020649e-01 4.64471042e-01 -7.12440968e-01 -7.97807634e-01
-9.55269396e-01 3.42866272e-01 3.30024689e-01 -4.41980273e-01
-8.39330792e-01 7.72358179e-01 5.02725363e-01 -5.78985631e-01
-1.42966717e-01 -4.04024094e-01 -5.70300639e-01 -2.11628735e-01
5.07768214e-01 4.29165035e-01 1.39749557e-01 -5.36613107e-01
-3.76237839e-01 7.79764295e-01 -2.64352977e-01 8.54359791e-02
1.72531748e+00 -3.77690494e-01 -7.63061523e-01 3.03846061e-01
1.20826995e+00 -6.59544766e-01 -1.25475121e+00 1.83827013e-01
1.14095427e-01 1.52458772e-01 2.45051593e-01 -6.67850673e-01
-1.40616882e+00 1.25780034e+00 5.70003271e-01 1.97501510e-01
9.35672581e-01 6.21441267e-02 1.19705725e+00 5.42850196e-01
6.00122333e-01 -1.07941532e+00 -1.99796751e-01 6.70524359e-01
2.46789232e-01 -1.31494653e+00 -1.81953982e-01 -1.81873024e-01
-3.95286605e-02 1.09690952e+00 6.69675648e-01 -2.70574391e-01
5.99906385e-01 3.76467615e-01 -2.05116402e-02 -6.15729272e-01
-4.29051936e-01 -2.43984520e-01 8.38186219e-03 6.37020290e-01
6.18753791e-01 -3.78816798e-02 -3.29272687e-01 5.96984863e-01
-2.30382895e-03 -3.33820507e-02 2.47005776e-01 6.21875346e-01
-6.08201385e-01 -1.05423045e+00 1.18432254e-01 7.90113866e-01
-9.14026022e-01 -1.26124278e-01 -3.88006538e-01 5.17197490e-01
6.44332841e-02 5.24323940e-01 2.11569667e-01 -1.40582025e-01
-2.01067805e-01 -4.68566641e-02 7.23155618e-01 -7.35362470e-01
-6.07819736e-01 2.38749355e-01 -5.20245075e-01 -5.91338992e-01
-7.36126423e-01 -5.00197530e-01 -1.86348367e+00 -4.56290394e-01
-1.56311408e-01 7.72931054e-02 8.85718167e-01 1.28279889e+00
3.56110305e-01 6.58053458e-01 1.35952517e-01 -1.47097087e+00
2.42561787e-01 -1.00518072e+00 -8.45942557e-01 3.60345960e-01
8.88299048e-02 -6.30014896e-01 -9.53469053e-02 5.30281849e-02]
|
[14.31287670135498, -3.107102870941162]
|
b67ed874-c63e-46d9-8655-a0667d153ca8
|
open-world-semantic-segmentation-via
|
2207.08455
| null |
https://arxiv.org/abs/2207.08455v3
|
https://arxiv.org/pdf/2207.08455v3.pdf
|
Open-world Semantic Segmentation via Contrasting and Clustering Vision-Language Embedding
|
To bridge the gap between supervised semantic segmentation and real-world applications that acquires one model to recognize arbitrary new concepts, recent zero-shot segmentation attracts a lot of attention by exploring the relationships between unseen and seen object categories, yet requiring large amounts of densely-annotated data with diverse base classes. In this paper, we propose a new open-world semantic segmentation pipeline that makes the first attempt to learn to segment semantic objects of various open-world categories without any efforts on dense annotations, by purely exploiting the image-caption data that naturally exist on the Internet. Our method, Vision-language-driven Semantic Segmentation (ViL-Seg), employs an image and a text encoder to generate visual and text embeddings for the image-caption data, with two core components that endow its segmentation ability: First, the image encoder is jointly trained with a vision-based contrasting and a cross-modal contrasting, which encourage the visual embeddings to preserve both fine-grained semantics and high-level category information that are crucial for the segmentation task. Furthermore, an online clustering head is devised over the image encoder, which allows to dynamically segment the visual embeddings into distinct semantic groups such that they can be classified by comparing with various text embeddings to complete our segmentation pipeline. Experiments show that without using any data with dense annotations, our method can directly segment objects of arbitrary categories, outperforming zero-shot segmentation methods that require data labeling on three benchmark datasets.
|
['Xiaodan Liang', 'Hang Xu', 'Chunjing Xu', 'Jianhua Han', 'Youpeng Wen', 'Quande Liu']
|
2022-07-18
| null | null | null | null |
['zero-shot-segmentation', 'online-clustering']
|
['computer-vision', 'computer-vision']
|
[ 3.82111847e-01 3.71281445e-01 -2.56273597e-01 -5.45650303e-01
-6.33475602e-01 -7.35461712e-01 6.18699908e-01 1.16561249e-01
-4.67167467e-01 1.38348639e-01 3.54554062e-03 -1.76270291e-01
3.27924699e-01 -8.65398645e-01 -8.86123002e-01 -4.99805599e-01
4.75725055e-01 7.21239448e-01 8.71789157e-01 -1.35631487e-01
1.16789468e-01 -5.47806919e-02 -1.74081600e+00 2.18547717e-01
1.04936945e+00 1.30009329e+00 6.07201576e-01 3.83785784e-01
-7.19006658e-01 4.39987540e-01 -1.17637426e-01 -4.76507574e-01
3.30481678e-01 -3.46162200e-01 -1.22919166e+00 6.03553653e-01
2.63932586e-01 -1.19939111e-01 4.86287475e-02 1.21802163e+00
3.25870700e-02 9.31246504e-02 6.02249026e-01 -1.28615463e+00
-9.98132586e-01 4.92987216e-01 -4.16376919e-01 -2.23583400e-01
1.06559180e-01 4.01274860e-01 1.18529725e+00 -7.48440981e-01
7.14521945e-01 1.16196811e+00 5.70315242e-01 6.01675451e-01
-1.18068123e+00 -1.91654414e-01 3.51809770e-01 1.36519998e-01
-1.29847252e+00 -1.48090020e-01 9.00283158e-01 -6.80904865e-01
4.59753275e-01 1.44096483e-02 7.57573485e-01 1.10827303e+00
-5.65885544e-01 9.09438789e-01 8.17309916e-01 -2.94378817e-01
4.73509073e-01 2.53878176e-01 3.35218966e-01 5.67617953e-01
1.25276625e-01 -3.03317457e-01 -1.04462229e-01 4.13656563e-01
6.71109974e-01 3.51952672e-01 -2.39597112e-01 -9.74181056e-01
-1.25561965e+00 8.25904429e-01 7.70430624e-01 5.00585794e-01
-2.50527143e-01 -2.40131635e-02 4.05214608e-01 -1.16739824e-01
4.13639843e-01 4.33101088e-01 -5.29805720e-01 2.03746155e-01
-1.08881950e+00 -1.97338805e-01 6.65622830e-01 1.12363577e+00
1.30258870e+00 -2.51283050e-01 -2.32161358e-01 9.25225854e-01
4.52602118e-01 2.95743346e-01 6.88890874e-01 -9.24662769e-01
3.21525425e-01 1.10106242e+00 -1.46649733e-01 -6.56279564e-01
-6.71818620e-04 -1.14699602e-01 -4.90361780e-01 -1.29826069e-01
4.76685941e-01 2.41225719e-01 -1.55948484e+00 1.58510077e+00
4.47852671e-01 1.99888363e-01 8.26383233e-02 9.90441501e-01
8.55529070e-01 5.25345027e-01 2.52508014e-01 2.80312747e-01
1.65125096e+00 -1.39536154e+00 -3.62832278e-01 -6.23542666e-01
5.28262675e-01 -3.93993855e-01 1.44922435e+00 -1.28966168e-01
-8.65946651e-01 -8.09556901e-01 -1.03209865e+00 -4.92368609e-01
-8.99015248e-01 -3.67131859e-01 5.10301769e-01 4.16505992e-01
-1.03217232e+00 3.19654584e-01 -8.12840044e-01 -6.84888363e-01
9.00867820e-01 -5.53852171e-02 -1.76517934e-01 -2.53045321e-01
-1.07899463e+00 4.94507402e-01 9.72390592e-01 -1.61814421e-01
-1.03575349e+00 -5.43086410e-01 -1.18720591e+00 1.59083217e-01
6.52551234e-01 -6.72716975e-01 9.34436977e-01 -1.52960479e+00
-1.23549163e+00 1.29867077e+00 9.74887088e-02 -2.56381899e-01
4.45580065e-01 1.60793766e-01 -1.57100596e-02 4.40567762e-01
4.63558257e-01 1.33606422e+00 9.12157655e-01 -1.66397846e+00
-6.46618903e-01 -5.17329812e-01 1.13253094e-01 2.03537866e-01
-4.59732831e-01 -3.35826129e-01 -1.04714775e+00 -4.86663491e-01
1.45506889e-01 -7.92824626e-01 -3.92430753e-01 1.46926299e-01
-6.06148481e-01 -3.52205485e-01 1.03600514e+00 -4.90645945e-01
7.43906736e-01 -2.19681954e+00 2.05102250e-01 -2.57819835e-02
2.34181985e-01 2.73110598e-01 -3.68351460e-01 1.23935699e-01
1.89713001e-01 1.47869632e-01 -6.86558247e-01 -4.47613031e-01
1.41508862e-01 5.55948436e-01 -2.44558170e-01 9.89421681e-02
2.82404065e-01 1.44125414e+00 -1.06278598e+00 -8.03720355e-01
2.97997057e-01 1.07969046e-01 -6.78958535e-01 4.96668160e-01
-7.26072550e-01 4.88069087e-01 -4.73725617e-01 7.44151294e-01
5.16895235e-01 -5.49668074e-01 7.45773241e-02 -1.95989192e-01
2.57530175e-02 -1.58878297e-01 -8.14995289e-01 2.16224647e+00
-2.58275658e-01 2.99926847e-01 1.40345911e-03 -1.53667831e+00
7.48915792e-01 2.85120052e-03 3.84345323e-01 -5.35541415e-01
3.78354698e-01 2.55615860e-01 -3.36515874e-01 -6.27676129e-01
3.68984640e-01 -1.29847139e-01 -4.67563689e-01 2.97845393e-01
4.45206314e-01 -4.28350836e-01 3.83065790e-01 3.22277546e-01
7.58567154e-01 2.97579765e-01 -1.50189444e-01 -1.70454308e-01
4.58483756e-01 3.13914716e-01 5.65628469e-01 6.02559447e-01
-3.13241184e-01 9.42762792e-01 4.02353376e-01 -2.97087014e-01
-1.16323018e+00 -1.36633956e+00 -1.33935869e-01 1.23844528e+00
7.06476808e-01 -3.78118493e-02 -1.16506922e+00 -9.69611824e-01
-3.82181555e-02 5.94729960e-01 -6.80695295e-01 -6.81577697e-02
-1.63341254e-01 -2.43947536e-01 1.71241939e-01 6.24704838e-01
5.34957826e-01 -1.28606594e+00 -6.45919085e-01 -8.52719247e-02
-3.92194301e-01 -1.42504358e+00 -5.78774691e-01 3.18052053e-01
-6.42422736e-01 -1.27985644e+00 -8.12539339e-01 -1.30698860e+00
8.69158387e-01 3.77360493e-01 1.12661052e+00 3.90557200e-02
-4.11908090e-01 6.48685217e-01 -5.76105237e-01 -1.39405623e-01
-3.06852102e-01 1.99360594e-01 -5.48524618e-01 3.68221283e-01
5.37174821e-01 -5.06824315e-01 -6.73700750e-01 2.56334960e-01
-1.20782828e+00 3.43763262e-01 6.03920639e-01 6.76542222e-01
6.80392087e-01 -4.09788936e-01 4.75558043e-01 -9.25228894e-01
6.48181736e-02 -6.08951330e-01 -4.12278444e-01 3.20808887e-01
-3.22390616e-01 5.90306371e-02 5.92694402e-01 -4.15285856e-01
-8.94880950e-01 3.46942365e-01 -2.38097925e-02 -6.61398113e-01
-4.89283204e-01 3.13642204e-01 -4.42447811e-01 3.26974362e-01
3.14384043e-01 4.15955096e-01 7.58500844e-02 -4.63061929e-01
1.12069678e+00 8.99246216e-01 8.77264380e-01 -4.95166451e-01
8.60451281e-01 6.98259354e-01 -7.33831108e-01 -8.06611180e-01
-1.31768441e+00 -8.87219369e-01 -1.13071322e+00 6.65317103e-02
1.68558884e+00 -9.39934075e-01 -4.18709144e-02 4.14437741e-01
-1.06787765e+00 -4.86588925e-01 -7.66046584e-01 -2.48187617e-03
-7.83772349e-01 5.31947911e-01 -4.12931025e-01 -2.32916564e-01
-1.93503469e-01 -1.16798973e+00 1.43196130e+00 3.54732573e-01
1.44306766e-02 -1.00819588e+00 -2.13258982e-01 7.53528357e-01
1.19328566e-01 1.80130601e-01 8.36043894e-01 -9.44861412e-01
-8.71020377e-01 9.06413198e-02 -6.91713333e-01 6.69599354e-01
9.76738334e-02 -3.16638291e-01 -9.70113337e-01 -1.31934389e-01
-1.92856237e-01 -6.97406948e-01 1.00148094e+00 1.54063419e-01
1.49877059e+00 -5.44463247e-02 -5.36880255e-01 7.82283127e-01
1.52717805e+00 -7.37447143e-02 5.89437425e-01 1.79695696e-01
1.23086774e+00 7.14242697e-01 3.66784096e-01 4.27510329e-02
6.73395455e-01 2.77454257e-01 5.79617441e-01 -4.19237047e-01
-1.97138414e-01 -5.16305327e-01 -1.79121923e-02 8.15003216e-01
3.43971521e-01 -7.34795332e-02 -9.11113977e-01 1.04116380e+00
-1.90595818e+00 -6.57626927e-01 1.09446079e-01 1.94994116e+00
9.97991085e-01 1.52827948e-01 1.39975874e-02 -7.53854364e-02
9.18519557e-01 2.90409744e-01 -8.66466284e-01 -2.55514234e-01
2.10409582e-01 9.46702585e-02 4.35333550e-01 1.23135537e-01
-1.29042470e+00 1.40970552e+00 5.19251347e+00 7.98522651e-01
-9.83028471e-01 2.80833542e-01 6.63092375e-01 4.38094407e-01
-4.03720707e-01 2.21150860e-01 -4.83288080e-01 6.07867122e-01
6.42732859e-01 7.39511177e-02 1.76774502e-01 9.93127286e-01
-1.32578284e-01 -1.57332923e-02 -1.18291068e+00 9.20484662e-01
2.27930665e-01 -1.26152086e+00 2.89460421e-01 -2.34386381e-02
8.26020360e-01 1.90429047e-01 -1.35667711e-01 3.71475011e-01
5.58122694e-01 -7.91649163e-01 8.83320034e-01 3.74591142e-01
7.90119648e-01 -1.79677621e-01 4.62488472e-01 3.66693646e-01
-1.22931552e+00 -1.47222579e-01 -3.47304761e-01 2.35000342e-01
2.46317193e-01 4.04772192e-01 -5.75028241e-01 4.82260823e-01
6.94636703e-01 9.35233235e-01 -7.20172167e-01 8.34796846e-01
-2.79736459e-01 4.21396911e-01 -1.04936332e-01 2.04830110e-01
6.13376677e-01 -3.63168240e-01 1.26509085e-01 9.80562389e-01
3.38107720e-02 -1.50905341e-01 7.22765625e-01 1.25778735e+00
-2.05583632e-01 -3.52110416e-02 -4.25224870e-01 -2.68231392e-01
3.01100940e-01 1.32645404e+00 -1.25173748e+00 -5.55602133e-01
-5.62232137e-01 1.32127213e+00 2.91410893e-01 4.80878860e-01
-9.02147472e-01 -4.71566081e-01 5.04850745e-01 9.16914642e-02
5.65524817e-01 -1.17938668e-01 -3.13758701e-01 -1.25749683e+00
-1.06049985e-01 -3.70137990e-01 4.12263781e-01 -8.48339558e-01
-1.39379299e+00 4.64824557e-01 -1.70987546e-01 -1.07911539e+00
6.46943077e-02 -5.88066339e-01 -5.70338428e-01 5.55278063e-01
-1.70953381e+00 -1.67241549e+00 -4.67776060e-01 6.57259583e-01
9.28609133e-01 2.19373465e-01 4.86639529e-01 2.01181471e-01
-5.12063324e-01 2.50492394e-01 -1.04155682e-01 4.93460357e-01
4.15117383e-01 -1.33792341e+00 4.02633965e-01 8.39338005e-01
3.07762384e-01 2.80319542e-01 3.04653615e-01 -4.01174963e-01
-9.70525026e-01 -1.46551907e+00 5.27845204e-01 -5.68872869e-01
8.11846375e-01 -7.82023966e-01 -1.11452270e+00 6.64465785e-01
1.09480187e-01 4.07714695e-01 5.26026607e-01 -8.93332735e-02
-5.37768424e-01 1.14002384e-01 -9.18181539e-01 4.67730999e-01
1.31295478e+00 -6.41011953e-01 -1.06101382e+00 3.51661205e-01
1.21336865e+00 -1.26590401e-01 -5.75193226e-01 2.36005723e-01
1.59010306e-01 -7.67625868e-01 1.17286432e+00 -5.41498899e-01
5.62223375e-01 -3.61472964e-01 -9.78530422e-02 -1.11699474e+00
-2.59711742e-02 -2.38684006e-02 3.87376696e-01 1.46079230e+00
2.42083505e-01 -4.61204112e-01 7.76341021e-01 4.41724867e-01
-4.51445073e-01 -4.76129770e-01 -6.99669659e-01 -7.87747622e-01
1.35392025e-01 -5.85924923e-01 4.19153959e-01 1.12053907e+00
-2.51324683e-01 5.12311637e-01 1.76065192e-01 7.83050731e-02
5.96485257e-01 4.33961987e-01 7.19712377e-01 -1.45420611e+00
-4.99227259e-04 -4.52757418e-01 -5.63485920e-01 -1.21618581e+00
3.46447110e-01 -1.30693388e+00 4.40568596e-01 -1.87013388e+00
3.62655163e-01 -5.32973647e-01 -3.90662730e-01 7.13612854e-01
-2.21615225e-01 5.75401187e-01 1.17919385e-01 2.56626070e-01
-1.22436917e+00 8.02889407e-01 1.30933177e+00 -4.22288358e-01
-1.58521235e-01 -5.06517589e-01 -9.24219191e-01 9.44663286e-01
4.49672073e-01 -2.80891269e-01 -5.89729965e-01 -5.84897578e-01
-2.46467009e-01 -3.73068035e-01 6.00231647e-01 -9.99631286e-01
1.76421046e-01 -1.39277965e-01 6.57648072e-02 -4.51401442e-01
3.91102694e-02 -8.94187212e-01 -3.07064593e-01 1.84462994e-01
-3.06118280e-01 -6.94062769e-01 -1.40252069e-01 7.98623621e-01
-3.37198585e-01 -2.58231372e-01 1.07137811e+00 -2.27029935e-01
-1.27848041e+00 5.46819270e-01 -7.68188993e-03 5.29356778e-01
1.32046008e+00 -5.30337512e-01 -1.00013494e-01 9.33249071e-02
-8.87272358e-01 6.66479111e-01 8.79027367e-01 8.05050611e-01
3.75808984e-01 -1.08945644e+00 -2.25238904e-01 2.84955949e-01
6.28009200e-01 5.15081346e-01 3.13067853e-01 5.64090669e-01
-4.24309939e-01 1.79461181e-01 -1.47204503e-01 -1.05362988e+00
-6.40512943e-01 9.68316257e-01 1.68234885e-01 3.70107032e-02
-7.10909784e-01 7.93397307e-01 6.99481726e-01 -7.02512085e-01
7.89587647e-02 -5.39249241e-01 -1.52550772e-01 3.03357124e-01
1.95369959e-01 -1.46965191e-01 -3.16253990e-01 -6.97432399e-01
-5.60033210e-02 7.74587870e-01 6.44227862e-02 1.84965491e-01
1.25855887e+00 -5.16835153e-01 -1.99978724e-01 5.68670154e-01
1.42816544e+00 -5.71846783e-01 -1.68786168e+00 -4.23546314e-01
3.37311983e-01 -3.20768863e-01 -5.66757470e-02 -6.06016934e-01
-1.17565370e+00 1.09946680e+00 4.33806062e-01 3.93614322e-01
9.57041502e-01 5.44034958e-01 1.28570271e+00 2.66929958e-02
3.52958083e-01 -1.29232943e+00 4.24452990e-01 4.99863446e-01
4.66345906e-01 -1.56315887e+00 -5.72763383e-01 -6.01731598e-01
-7.92501032e-01 7.44964600e-01 6.05160773e-01 -8.20985511e-02
5.48269391e-01 -2.63588905e-01 1.79783672e-01 -2.25100681e-01
-2.00879216e-01 -9.08929825e-01 3.46157819e-01 8.49524617e-01
-1.49166808e-01 -1.66539401e-02 -1.12493373e-01 8.31151366e-01
4.38471809e-02 -1.23113871e-01 3.09398532e-01 6.63473248e-01
-7.70738959e-01 -1.01814938e+00 -7.25549310e-02 3.41450542e-01
-1.42480507e-02 -5.96373305e-02 -2.79077470e-01 5.89611650e-01
4.25343096e-01 7.85457969e-01 5.17759681e-01 -1.21174708e-01
1.82352722e-01 3.48557472e-01 6.92459196e-02 -1.01616824e+00
-7.96885043e-02 -8.82758573e-02 -3.98225367e-01 -6.00280106e-01
-4.32579398e-01 -4.46570814e-01 -1.49617386e+00 4.57669914e-01
-4.07073721e-02 1.49591357e-01 6.03942752e-01 1.17004001e+00
3.39982599e-01 5.21932602e-01 5.62329829e-01 -1.00332642e+00
-1.20617643e-01 -5.78210235e-01 -6.38226330e-01 9.79239762e-01
1.22695610e-01 -6.78776443e-01 -4.01456207e-01 5.07571161e-01]
|
[9.746145248413086, 0.9110327959060669]
|
2205bd3f-40f1-4bdd-b2d1-87fc6829be63
|
suggestion-mining-from-online-reviews-using
|
1904.09076
| null |
http://arxiv.org/abs/1904.09076v1
|
http://arxiv.org/pdf/1904.09076v1.pdf
|
Suggestion Mining from Online Reviews using ULMFiT
|
In this paper we present our approach and the system description for Sub Task
A of SemEval 2019 Task 9: Suggestion Mining from Online Reviews and Forums.
Given a sentence, the task asks to predict whether the sentence consists of a
suggestion or not. Our model is based on Universal Language Model Fine-tuning
for Text Classification. We apply various pre-processing techniques before
training the language and the classification model. We further provide detailed
analysis of the results obtained using the trained model. Our team ranked 10th
out of 34 participants, achieving an F1 score of 0.7011. We publicly share our
implementation at https://github.com/isarth/SemEval9_MIDAS
|
['Simra Shahid', 'Laiba Mehnaz', 'Debanjan Mahata', 'Rajiv Ratn Shah', 'Karan Uppal', 'Haimin Zhang', 'Yaman Kumar', 'Sarthak Anand', 'Kartik Aggarwal']
|
2019-04-19
| null | null | null | null |
['suggestion-mining']
|
['natural-language-processing']
|
[-1.78666249e-01 3.47281694e-01 -3.09853286e-01 -6.63529396e-01
-9.51807320e-01 -4.33751255e-01 8.12683403e-01 5.93286753e-01
-8.05121362e-01 5.41604877e-01 4.71828431e-01 -8.48036468e-01
2.39340335e-01 -2.84664243e-01 -4.10789251e-01 -4.03283797e-02
2.38443702e-01 4.28876370e-01 -2.33756080e-02 -2.80231982e-01
7.61539757e-01 -4.72622454e-01 -1.25764823e+00 1.04297864e+00
4.97167885e-01 8.37533355e-01 3.54746543e-02 1.10241246e+00
-1.26651585e-01 6.20271504e-01 -4.73419666e-01 -7.58485258e-01
-1.79071687e-02 -2.50539988e-01 -1.21831691e+00 -2.13882834e-01
5.79873085e-01 2.87289801e-03 1.05141863e-01 7.00495064e-01
2.57023752e-01 3.87627125e-01 8.26028943e-01 -7.96696126e-01
-5.80743253e-01 1.18537641e+00 -3.20502251e-01 2.87308425e-01
5.57561278e-01 -2.59218186e-01 1.22365344e+00 -1.55376673e+00
4.32709157e-01 1.05991626e+00 3.95551085e-01 7.73302257e-01
-1.21539497e+00 -5.52655280e-01 2.63726324e-01 -4.24988940e-02
-1.22406650e+00 -6.45426095e-01 3.87708068e-01 -4.27744925e-01
1.38616967e+00 4.00686115e-01 1.92109108e-01 1.16967523e+00
3.79469454e-01 8.44446242e-01 1.29020000e+00 -7.01621294e-01
1.39006048e-01 8.14439237e-01 8.03092301e-01 5.25334358e-01
3.86188948e-03 -6.06741548e-01 -8.16416860e-01 -3.95103306e-01
5.26417047e-02 -2.48731077e-01 9.25396457e-02 4.79155570e-01
-9.43207383e-01 9.77859616e-01 8.36080015e-02 1.88244462e-01
-2.67023027e-01 -2.84088701e-01 6.46724701e-01 5.36017239e-01
1.01059556e+00 5.18414915e-01 -8.57427299e-01 -2.39705980e-01
-9.19137776e-01 4.49149370e-01 1.32051599e+00 5.97272992e-01
5.09197712e-01 -5.50320089e-01 -2.56028146e-01 1.29149723e+00
3.68428648e-01 1.64125159e-01 8.80215168e-01 -7.79849112e-01
5.15746653e-01 4.68262821e-01 1.69678688e-01 -8.12541187e-01
-6.06299937e-01 -3.27954710e-01 -5.47630191e-01 -3.83017719e-01
2.03785643e-01 -6.06873572e-01 -4.44778532e-01 1.07437658e+00
-3.24222334e-02 -3.24595869e-01 -1.50975361e-01 3.98384184e-01
1.25335240e+00 4.54772383e-01 2.06487790e-01 -2.95781553e-01
1.36421800e+00 -1.31831801e+00 -4.87708896e-01 -4.54033881e-01
1.18970394e+00 -1.12589765e+00 1.21828091e+00 5.06650150e-01
-8.65345061e-01 -7.09973156e-01 -7.73204148e-01 -2.07428917e-01
-2.56800413e-01 5.99426448e-01 4.22856569e-01 4.72386897e-01
-1.13370526e+00 2.24911898e-01 -4.88731951e-01 -5.69094419e-01
-1.55464724e-01 3.37096661e-01 -2.24624500e-01 2.88615763e-01
-9.86388206e-01 6.52354360e-01 1.35321751e-01 -2.91535109e-01
-2.68863976e-01 -3.77864212e-01 -6.46376431e-01 -1.61532342e-01
2.81622648e-01 -4.27737474e-01 2.01923251e+00 -7.91428983e-01
-1.47166204e+00 1.13371181e+00 -6.13857031e-01 -6.67369604e-01
3.10236037e-01 -4.80043828e-01 -5.26870668e-01 -3.37125212e-01
1.97870284e-01 4.95551080e-01 5.28069973e-01 -6.91387892e-01
-9.03901398e-01 -2.79323608e-02 -5.25980024e-03 1.56766832e-01
-5.24807811e-01 5.69705009e-01 -3.70626181e-01 -4.02803898e-01
-2.60594726e-01 -9.72287953e-01 -3.99443656e-01 -7.81932294e-01
-7.01138854e-01 -8.79652381e-01 1.17627814e-01 -5.96629739e-01
1.50684917e+00 -1.92416775e+00 -6.60389841e-01 1.97450653e-01
2.20367670e-01 1.26694173e-01 -2.39731997e-01 7.99265742e-01
7.77211711e-02 5.45215666e-01 1.22426338e-01 -7.29534566e-01
3.73961963e-02 -5.26377976e-01 -3.89896244e-01 1.74742952e-01
-1.27025738e-01 7.92291880e-01 -7.68713355e-01 -2.33375192e-01
-9.04494897e-02 1.96679205e-01 -4.22771007e-01 2.79106855e-01
-2.17603996e-01 1.22715205e-01 -4.52553272e-01 1.57866806e-01
3.80276263e-01 -3.59925747e-01 -2.03548814e-03 1.89433873e-01
-1.67363763e-01 1.27225316e+00 -7.28321910e-01 1.33276761e+00
-5.51599383e-01 5.80923975e-01 -8.05484317e-03 -6.88220918e-01
1.03302419e+00 2.34014288e-01 2.08763376e-01 -5.75468540e-01
1.21297665e-01 2.93834358e-01 5.23467101e-02 -4.32031751e-01
8.17685366e-01 1.22326985e-01 -1.80049241e-01 9.50891197e-01
-1.38395488e-01 6.51839375e-02 4.66415703e-01 6.61008418e-01
1.08745658e+00 -1.65118933e-01 4.90416348e-01 -4.23972785e-01
8.25733483e-01 -1.18394502e-01 5.08110933e-02 1.18098521e+00
-4.11120765e-02 4.10550922e-01 5.03081918e-01 -2.59139806e-01
-7.12460160e-01 -5.81213593e-01 -8.98122042e-02 1.54224873e+00
-5.60590208e-01 -1.33766246e+00 -5.28126776e-01 -1.03887188e+00
-6.17089262e-03 1.12309325e+00 -7.24042416e-01 1.77783296e-01
-2.57852614e-01 -4.77807969e-01 2.00550199e-01 4.96965162e-02
1.60913140e-01 -1.05149770e+00 -1.28035724e-01 2.91411430e-02
-2.74461746e-01 -1.10649562e+00 -7.17000663e-01 2.10579678e-01
-7.29167938e-01 -9.62311804e-01 -3.40484351e-01 -7.94429421e-01
4.66662258e-01 3.71154189e-01 1.31624842e+00 3.59127223e-01
1.77450329e-01 4.99025732e-01 -6.17142677e-01 -6.05908096e-01
-5.35225511e-01 4.08656985e-01 1.86113104e-01 -1.07911490e-01
8.95345986e-01 -6.56366646e-02 -5.95930576e-01 1.66699156e-01
-2.30602309e-01 3.13772559e-01 3.28320980e-01 5.76547205e-01
2.90252477e-01 -3.56007010e-01 8.33673775e-01 -1.58030486e+00
1.35699749e+00 -6.58164918e-01 -1.01545505e-01 -1.48163848e-02
-1.06777573e+00 -2.19228804e-01 5.21774232e-01 3.27370432e-03
-8.52289140e-01 1.22748343e-02 -5.62873840e-01 5.42613268e-01
-3.19311291e-01 8.07115495e-01 6.97998524e-01 4.27459657e-01
1.02581000e+00 -7.76146278e-02 -2.22413570e-01 -5.28141439e-01
3.09366196e-01 1.24385679e+00 1.29744232e-01 -2.90040374e-01
3.26016963e-01 4.34479304e-02 -7.61159182e-01 -9.02104795e-01
-1.32464790e+00 -9.38071191e-01 -5.64854741e-01 -1.71590149e-01
3.58203173e-01 -8.49618793e-01 -6.82209790e-01 1.27364337e-01
-1.07276082e+00 -5.56052446e-01 3.59762348e-02 4.74079669e-01
-2.22376660e-01 1.25808135e-01 -7.53730953e-01 -1.03839695e+00
-9.73340094e-01 -6.84908628e-01 7.00730622e-01 1.43620834e-01
-8.17563295e-01 -1.17617261e+00 2.24364191e-01 7.80963778e-01
3.72624338e-01 -6.38554215e-01 4.28144038e-01 -1.45897102e+00
4.04729158e-01 -4.70148474e-01 -2.15595737e-02 3.32400948e-01
-1.60996243e-02 3.53903673e-03 -9.71900761e-01 -1.52427152e-01
2.72282995e-02 -7.24261761e-01 1.08172154e+00 2.61841029e-01
1.45931792e+00 -5.98863184e-01 -2.33229339e-01 -1.68457121e-01
8.40415776e-01 -4.83537644e-01 3.23399119e-02 4.70040202e-01
4.11369354e-01 1.01418841e+00 9.05449510e-01 5.46090901e-01
7.42217541e-01 6.02416217e-01 -1.24596238e-01 3.79512489e-01
1.91648722e-01 -3.65141124e-01 8.73495281e-01 1.02677667e+00
3.87643933e-01 -4.67325270e-01 -1.00805867e+00 4.62152660e-01
-1.90556693e+00 -6.52962685e-01 -5.09963214e-01 2.10775208e+00
9.75401819e-01 5.10651231e-01 2.98904300e-01 -1.94227457e-01
4.30243313e-01 5.88704310e-02 -1.40397623e-01 -1.12796974e+00
8.58135223e-02 2.25844011e-01 2.06535414e-01 1.00647926e+00
-1.21171701e+00 1.01606834e+00 5.89459085e+00 7.49981225e-01
-9.52422082e-01 1.94888085e-01 1.02966654e+00 -1.98404565e-01
-3.10976714e-01 -1.39599621e-01 -1.22908509e+00 3.16023320e-01
1.75983751e+00 -4.45159405e-01 -6.77425116e-02 6.59167707e-01
6.20766163e-01 -4.01894331e-01 -1.12766266e+00 5.83184779e-01
3.64560992e-01 -1.20492744e+00 -3.61473337e-02 -6.40312657e-02
6.72544718e-01 5.25016725e-01 1.30270436e-01 7.23933220e-01
2.65337884e-01 -1.03550923e+00 5.25088847e-01 2.41577223e-01
4.24242020e-01 -3.13580275e-01 8.46765578e-01 8.51496339e-01
-5.53066850e-01 1.58296943e-01 -3.15334052e-01 -5.14890432e-01
-1.52639924e-02 7.98072219e-01 -1.33956468e+00 1.91721544e-01
6.12160623e-01 9.49997783e-01 -8.48412216e-01 7.43516982e-01
-6.18128240e-01 1.27731133e+00 -1.49555698e-01 -4.46957856e-01
6.41165376e-02 -1.55045595e-02 3.03402960e-01 1.72110403e+00
-1.07512668e-01 -4.41927500e-02 3.29917490e-01 3.43971938e-01
-4.52889889e-01 8.83977354e-01 -3.29043955e-01 1.25971427e-02
3.70181739e-01 1.67183101e+00 -6.14931047e-01 -4.75436270e-01
-6.03152096e-01 7.02570498e-01 6.41717196e-01 2.38110393e-01
-3.73797983e-01 -1.84339538e-01 9.04957354e-02 3.36420149e-01
4.86224107e-02 -9.28774625e-02 -6.43975019e-01 -1.29141676e+00
9.23952926e-03 -9.76835370e-01 4.26571935e-01 -4.63949919e-01
-1.40549147e+00 6.77352726e-01 -3.64060551e-01 -7.09213555e-01
-4.91002500e-01 -7.26256728e-01 -9.68873501e-01 9.83109236e-01
-1.14575338e+00 -9.23788905e-01 1.30225897e-01 8.80240649e-02
1.00257325e+00 -2.94043690e-01 1.01250780e+00 3.14356610e-02
-5.11843324e-01 8.49201560e-01 -4.15734798e-02 -1.26599297e-01
1.38136947e+00 -1.35768020e+00 6.45866215e-01 6.35831773e-01
2.90202111e-01 1.03255689e+00 8.56923342e-01 -6.58661127e-01
-8.46598804e-01 -1.05306566e+00 1.85387504e+00 -8.08896303e-01
1.03798223e+00 -5.63717186e-01 -8.42670679e-01 6.50983512e-01
5.56135952e-01 -4.65975642e-01 1.19972837e+00 7.90109456e-01
-2.61642367e-01 2.07045630e-01 -5.44552982e-01 4.98608440e-01
5.21023691e-01 -7.40829289e-01 -6.77502692e-01 8.35850954e-01
6.63681388e-01 -1.98810577e-01 -7.50660717e-01 1.34986028e-01
6.24880314e-01 -8.28234017e-01 4.51248884e-01 -6.33840621e-01
8.35513055e-01 1.99512735e-01 1.32010013e-01 -1.04148197e+00
-3.00643623e-01 -4.23644185e-01 -1.05314761e-01 1.24343240e+00
1.27723539e+00 -5.66200972e-01 7.49015152e-01 8.30474794e-01
-5.86557649e-02 -1.12944770e+00 -4.40713525e-01 -2.84454882e-01
2.78212398e-01 -7.91524887e-01 -2.94206589e-01 6.32062733e-01
6.08055294e-01 9.13834572e-01 -2.09123239e-01 -6.10579729e-01
2.22776324e-01 4.98511679e-02 9.46236134e-01 -8.39476109e-01
-3.33625078e-01 -4.30604815e-01 4.72238868e-01 -1.28290164e+00
4.27189976e-01 -1.15186298e+00 8.37951675e-02 -1.60770750e+00
4.47271138e-01 -2.01557964e-01 -5.51923454e-01 6.39934838e-01
-2.68286824e-01 3.89717340e-01 -7.80937374e-02 3.29531044e-01
-1.03524876e+00 3.69491577e-01 6.09548986e-01 -5.03367148e-02
-2.36870408e-01 6.94324970e-01 -1.20821202e+00 7.63006628e-01
1.25567937e+00 -4.45406765e-01 -2.21243933e-01 -2.43595585e-01
4.86591578e-01 -4.51280147e-01 4.39685136e-02 -5.00509858e-01
4.46526140e-01 7.33700767e-02 6.02582246e-02 -5.94080210e-01
6.17056526e-02 -1.65057272e-01 -7.46409237e-01 4.21055913e-01
-1.05931938e+00 4.39639911e-02 1.56178638e-01 3.30801368e-01
-7.67719746e-02 -4.36105371e-01 2.53360868e-01 3.49170603e-02
-3.32941055e-01 -5.91976866e-02 -8.16679001e-01 -5.90477847e-02
3.58305097e-01 4.35202688e-01 -4.90779638e-01 -6.55784607e-01
-6.86560571e-01 4.94046062e-01 1.39534757e-01 7.75701761e-01
6.42226756e-01 -6.64009988e-01 -1.13185275e+00 -9.22530070e-02
3.52293402e-01 -4.85634834e-01 3.35658193e-02 1.06666744e+00
-1.26579210e-01 8.93077075e-01 3.86088371e-01 -1.67040437e-01
-1.52731526e+00 4.85772453e-02 8.35721046e-02 -4.10957754e-01
-4.39927071e-01 1.03793299e+00 2.68586669e-02 -7.09705472e-01
1.95742667e-01 -1.20738775e-01 -6.98551595e-01 -4.43030223e-02
7.94369459e-01 3.25932890e-01 2.69569129e-01 -4.23849285e-01
-4.19071138e-01 -7.77209178e-02 -7.23875999e-01 -5.00462413e-01
1.16834021e+00 -3.56217951e-01 -3.90841633e-01 9.16792035e-01
1.02381802e+00 4.33367252e-01 -3.97454172e-01 -4.76146460e-01
4.28642571e-01 5.92895085e-04 4.03085947e-02 -1.07332003e+00
-3.08213532e-01 5.85783124e-01 2.80011725e-02 5.13556898e-01
5.18840432e-01 1.79988757e-01 5.07889450e-01 7.36880124e-01
-5.50586171e-03 -1.35283422e+00 -1.35512114e-01 1.04254925e+00
1.03183615e+00 -1.66818440e+00 3.08005393e-01 -3.18157852e-01
-7.90089488e-01 1.08640230e+00 8.07167232e-01 -2.11434841e-01
9.85997856e-01 -3.27958986e-02 2.69572467e-01 -3.37767243e-01
-1.40260577e+00 2.42974862e-01 7.82386422e-01 -3.62190343e-02
1.52866578e+00 2.85225391e-01 -1.03021872e+00 1.23869526e+00
-6.41909659e-01 -2.17644885e-01 7.10005939e-01 6.58194184e-01
-6.04488730e-01 -1.31348705e+00 2.02487320e-01 9.97995198e-01
-7.97092021e-01 -3.16113323e-01 -9.18769896e-01 1.65727884e-01
-2.85864472e-01 1.53425729e+00 -2.00233847e-01 -6.35194957e-01
2.23305598e-01 2.11395964e-01 -2.01535881e-01 -1.25013137e+00
-1.11286104e+00 1.92178097e-02 7.56849408e-01 -3.90423626e-01
-3.65802705e-01 -9.52387691e-01 -1.33938849e+00 -2.27070138e-01
-2.59575188e-01 7.20277846e-01 7.34480023e-01 9.71301436e-01
5.29027820e-01 1.19407550e-01 6.30832195e-01 -3.44519854e-01
-6.32165134e-01 -1.53282547e+00 -3.13089639e-01 6.30374849e-02
1.02651857e-01 2.23695692e-02 -4.55858231e-01 7.68035576e-02]
|
[10.936917304992676, 7.489565849304199]
|
a6e3ce46-a7b1-4414-9854-ca8b701c4a18
|
intermediate-deep-feature-compression-the
|
1809.06196
| null |
http://arxiv.org/abs/1809.06196v1
|
http://arxiv.org/pdf/1809.06196v1.pdf
|
Intermediate Deep Feature Compression: the Next Battlefield of Intelligent Sensing
|
The recent advances of hardware technology have made the intelligent analysis
equipped at the front-end with deep learning more prevailing and practical. To
better enable the intelligent sensing at the front-end, instead of compressing
and transmitting visual signals or the ultimately utilized top-layer deep
learning features, we propose to compactly represent and convey the
intermediate-layer deep learning features of high generalization capability, to
facilitate the collaborating approach between front and cloud ends. This
strategy enables a good balance among the computational load, transmission load
and the generalization ability for cloud servers when deploying the deep neural
networks for large scale cloud based visual analysis. Moreover, the presented
strategy also makes the standardization of deep feature coding more feasible
and promising, as a series of tasks can simultaneously benefit from the
transmitted intermediate layers. We also present the results for evaluation of
lossless deep feature compression with four benchmark data compression methods,
which provides meaningful investigations and baselines for future research and
standardization activities.
|
['Ling-Yu Duan', 'Zhuo Chen', 'Shiqi Wang', 'Alex C. Kot', 'Weisi Lin']
|
2018-09-17
| null | null | null | null |
['feature-compression']
|
['computer-vision']
|
[ 1.66937828e-01 -8.31023678e-02 1.12484001e-01 -3.59779984e-01
-3.18858385e-01 -1.38298184e-01 3.46315920e-01 -7.19963312e-02
-5.94468057e-01 3.00919682e-01 -4.66136187e-02 -8.98859203e-02
-3.50998998e-01 -1.16089487e+00 -5.67254484e-01 -9.68380451e-01
-4.19798613e-01 6.19315356e-02 -1.19126923e-01 3.81307065e-01
-1.95013329e-01 8.36632848e-01 -2.17774677e+00 3.39762270e-01
6.34711683e-01 1.91550469e+00 6.97768688e-01 7.46151090e-01
-6.75833672e-02 9.19158459e-01 -4.81656313e-01 -6.77897096e-01
3.47195536e-01 2.22697735e-01 -3.58621806e-01 -1.35323152e-01
3.50342363e-01 -6.65408731e-01 -6.49778545e-01 1.08118415e+00
7.06202269e-01 -1.01769850e-01 3.28005552e-01 -1.39941299e+00
-4.59051013e-01 5.38921356e-01 -2.56378263e-01 1.86990604e-01
-8.48853737e-02 1.42312765e-01 8.30859840e-01 -7.99284041e-01
4.60605323e-01 8.81139159e-01 6.72412753e-01 3.62310946e-01
-4.00770336e-01 -8.78312767e-01 -4.68872720e-03 8.70563447e-01
-1.19333112e+00 -8.35621119e-01 1.02950370e+00 -1.05314128e-01
1.03442287e+00 4.51536179e-01 1.07896149e+00 7.95109749e-01
1.36332422e-01 8.35318685e-01 4.44820315e-01 -1.14043035e-01
3.56286764e-01 7.60301352e-02 -3.63527015e-02 8.18769872e-01
3.24288845e-01 -9.22401398e-02 -6.97008729e-01 5.81866503e-02
3.50272924e-01 6.45233393e-01 -2.13179011e-02 -1.60721451e-01
-7.24709392e-01 6.47625625e-01 5.59405088e-01 3.92381042e-01
-9.89212453e-01 4.05745596e-01 7.49595463e-01 3.92428428e-01
5.15010178e-01 -1.40123010e-01 -2.95537233e-01 -4.11002845e-01
-1.46204829e+00 -1.13328844e-01 5.47316253e-01 8.44380140e-01
6.23073578e-01 3.51864487e-01 -1.47301584e-01 5.68872273e-01
2.00823754e-01 8.22908103e-01 5.85512757e-01 -1.28358781e+00
5.38766801e-01 4.36077178e-01 -3.87968123e-01 -1.31804931e+00
-4.11592692e-01 -9.50698435e-01 -1.41750729e+00 2.10383117e-01
-1.54487446e-01 -8.70521814e-02 -4.81891900e-01 1.41708136e+00
1.27088293e-01 8.82729143e-03 3.55582684e-01 9.37869728e-01
8.44266057e-01 5.99431872e-01 -7.29532689e-02 -2.18812793e-01
1.36524570e+00 -8.79023671e-01 -8.62229109e-01 2.91491412e-02
4.89847541e-01 -5.61237514e-01 1.07441139e+00 5.77065587e-01
-1.35864544e+00 -7.92706251e-01 -1.18536878e+00 -2.91485101e-01
-3.56046736e-01 3.65917623e-01 8.30727756e-01 4.31578219e-01
-1.22933650e+00 6.50328517e-01 -1.01352358e+00 1.32061288e-01
8.85308206e-01 3.16819459e-01 -2.90641576e-01 -2.33302057e-01
-9.88955200e-01 2.35975608e-01 1.98511615e-01 2.54025012e-01
-7.47210979e-01 -7.85507143e-01 -4.31744307e-01 7.22693503e-01
-2.38245100e-01 -7.80681849e-01 8.50572407e-01 -9.59797740e-01
-1.25284708e+00 6.74574912e-01 -1.44991847e-02 -8.11102450e-01
4.00719255e-01 -1.37533054e-01 -2.56361306e-01 6.92023814e-01
-4.72755104e-01 6.78255975e-01 1.04283810e+00 -6.60756290e-01
-6.73383594e-01 -6.82558060e-01 -2.98892647e-01 7.58120343e-02
-1.29694283e+00 -4.41333264e-01 -5.41203380e-01 -4.97588038e-01
1.33330792e-01 -5.31103134e-01 1.33836120e-01 6.59331918e-01
-9.36165974e-02 -4.28473391e-02 1.35180616e+00 -7.42969751e-01
9.32973087e-01 -2.53448653e+00 -2.87924055e-02 1.95937350e-01
6.65336311e-01 5.28912365e-01 -1.71202675e-01 1.90371543e-01
3.10347706e-01 -1.80363566e-01 6.81601986e-02 -7.94090867e-01
1.69014573e-01 1.19768642e-01 -4.23785657e-01 4.09729749e-01
-2.32966796e-01 9.78130400e-01 -4.03525561e-01 -5.92498720e-01
2.88589567e-01 7.73553967e-01 -5.79783082e-01 4.02374983e-01
7.77603686e-02 -1.89898163e-01 -6.02464616e-01 7.48192847e-01
9.36027646e-01 -3.33438218e-01 -1.42524779e-01 -3.71001571e-01
1.23509392e-01 -8.72288123e-02 -7.01780558e-01 1.88148248e+00
-7.26942837e-01 8.67981493e-01 4.77787346e-01 -1.14554489e+00
8.93399954e-01 1.35164052e-01 7.77247548e-01 -1.11057353e+00
3.94688174e-02 -3.89209874e-02 -3.84223729e-01 -7.34741688e-01
5.75029731e-01 1.63359031e-01 1.36888951e-01 2.93232739e-01
1.82144478e-01 7.04773143e-02 -1.36771724e-01 2.28480041e-01
1.19856739e+00 -2.88425356e-01 -2.05831558e-01 1.74122378e-01
3.72130096e-01 -2.76805788e-01 2.91475534e-01 5.08194149e-01
-3.29705089e-01 1.44643530e-01 1.62262365e-01 -6.31804645e-01
-1.11756694e+00 -9.60426986e-01 -5.86781986e-02 1.20326316e+00
-1.02756619e-01 -4.27285552e-01 -6.09697104e-01 -2.36339092e-01
-3.84078734e-03 3.01654845e-01 -2.57194936e-01 -3.87284636e-01
-1.70864239e-01 -4.01392519e-01 6.44125164e-01 4.99835998e-01
9.74043548e-01 -1.00607765e+00 -1.27544403e+00 -2.90504079e-02
-2.57198781e-01 -1.29203820e+00 2.78549194e-01 3.99515837e-01
-1.03601420e+00 -6.54840291e-01 -7.35068202e-01 -4.62923765e-01
3.42054963e-01 4.44985032e-01 8.77017915e-01 2.90274084e-01
-2.93384939e-01 3.62294942e-01 -4.20557320e-01 -6.37165785e-01
2.94464617e-03 1.51018471e-01 -1.43952608e-01 -1.13578171e-01
2.64729172e-01 -1.04317045e+00 -7.92360544e-01 -3.95933300e-01
-1.13086152e+00 3.58671814e-01 7.09324360e-01 4.52431172e-01
7.60302305e-01 3.23098838e-01 3.35362047e-01 -3.20437819e-01
4.26079392e-01 -3.82810831e-01 -6.21341467e-01 1.58954188e-01
-5.23576438e-01 1.81688759e-02 1.18274748e+00 -1.34116858e-01
-7.61970222e-01 3.82801183e-02 -4.15233046e-01 -8.50167334e-01
1.09150991e-01 3.73156637e-01 -2.27415770e-01 -1.78521723e-01
2.28136316e-01 6.51084125e-01 3.56479622e-02 -4.32003826e-01
3.25654089e-01 9.56964433e-01 6.34409547e-01 -4.38239016e-02
6.07486188e-01 6.43605351e-01 1.62211642e-01 -9.14452672e-01
-2.17712522e-01 -1.77058935e-01 -2.40995675e-01 -4.06607836e-01
7.70285666e-01 -1.07730627e+00 -1.07322609e+00 6.36284292e-01
-1.11468494e+00 -1.01006530e-01 -5.24424076e-01 3.77487779e-01
-5.04623234e-01 3.82144034e-01 -5.29772162e-01 -6.93643153e-01
-1.10781205e+00 -1.12821293e+00 1.15180039e+00 1.28403649e-01
4.04616028e-01 -6.82124615e-01 -4.66663510e-01 3.10515493e-01
8.71972501e-01 2.38040723e-02 8.68697882e-01 -3.92831773e-01
-7.07208097e-01 -3.63158017e-01 -5.25808990e-01 4.85648274e-01
-3.54305923e-01 -1.17379993e-01 -1.37955916e+00 -4.36770141e-01
2.79405326e-01 -6.01046920e-01 1.01441097e+00 3.29135448e-01
1.98577189e+00 -5.02595246e-01 -7.73748532e-02 1.26682961e+00
1.37711668e+00 -2.07476407e-01 6.19754910e-01 1.66253477e-01
4.68456894e-01 4.28631455e-01 2.60982394e-01 1.08887291e+00
3.99847388e-01 4.91715193e-01 9.13587332e-01 -2.72858113e-01
-2.81204611e-01 7.39846528e-02 3.05863470e-01 1.29640520e+00
-1.45080406e-02 -1.77174255e-01 -5.24548352e-01 1.31973445e-01
-1.49282074e+00 -1.11421192e+00 5.58350146e-01 1.87218678e+00
5.50734043e-01 -1.77416220e-01 -2.55324930e-01 5.27454078e-01
2.45724261e-01 4.07661885e-01 -8.36122632e-01 -4.23186421e-01
-7.01065315e-03 1.19814605e-01 4.29079533e-01 -6.20368905e-02
-8.51805687e-01 4.98522520e-01 5.72709990e+00 1.06861520e+00
-1.57994068e+00 2.65134841e-01 6.04696453e-01 -7.38000214e-01
-3.86575341e-01 -7.60125756e-01 -4.95976001e-01 5.72583914e-01
1.23627317e+00 -3.59436311e-02 7.17931449e-01 1.11593330e+00
1.43132269e-01 1.78704068e-01 -8.98777902e-01 1.43848264e+00
-2.25960493e-01 -1.52413797e+00 1.28204495e-01 4.33900431e-02
2.67744482e-01 3.51152629e-01 3.73053223e-01 1.03095166e-01
-2.47320011e-01 -7.71323025e-01 9.80184078e-01 8.23476255e-01
9.89222586e-01 -9.34625268e-01 6.86815023e-01 4.98048306e-01
-1.24924958e+00 -5.94061196e-01 -8.30490649e-01 -2.85899434e-02
-2.22090170e-01 1.10616934e+00 -4.91135597e-01 5.65228462e-01
1.11095989e+00 7.49105871e-01 -5.36860645e-01 6.15536451e-01
3.19212049e-01 3.90227705e-01 -3.26488823e-01 5.58691770e-02
8.26671645e-02 1.56280156e-02 2.31124073e-01 1.17763543e+00
5.53769112e-01 -2.00094894e-01 -2.31033489e-01 7.13587582e-01
-3.43951851e-01 -4.38160114e-02 -6.38499022e-01 -1.97301209e-01
6.33021116e-01 1.49070895e+00 -4.58166540e-01 -4.46644604e-01
-3.63628209e-01 9.37691271e-01 4.40027088e-01 2.72548527e-01
-7.58296609e-01 -6.02935016e-01 9.46602404e-01 -1.99108303e-01
5.43492675e-01 -3.94057184e-01 -3.67877990e-01 -1.08567142e+00
4.68164504e-01 -7.48622954e-01 2.10955918e-01 -8.87313366e-01
-8.19003940e-01 7.60366321e-01 -3.27466756e-01 -1.03893828e+00
-3.13493431e-01 -4.06523615e-01 -3.10213745e-01 5.96655250e-01
-1.90138280e+00 -1.04847801e+00 -7.39169538e-01 9.62697148e-01
3.49110633e-01 -3.30170542e-01 7.32058287e-01 6.78299367e-01
-4.06432629e-01 8.46968412e-01 4.15870160e-01 -5.09444922e-02
5.22662979e-03 -5.48777044e-01 1.55330729e-02 9.86240804e-01
2.33379930e-01 1.04347967e-01 3.72542202e-01 -1.29685372e-01
-2.09372950e+00 -1.28311753e+00 4.74354804e-01 6.45729601e-01
3.83777797e-01 -4.98876572e-01 -6.75730467e-01 1.04419708e-01
-1.80941727e-02 2.32454672e-01 5.90474427e-01 -3.86056483e-01
-2.06144169e-01 -7.94041991e-01 -1.37291765e+00 1.62291363e-01
1.17525339e+00 -8.17910075e-01 6.43547401e-02 3.33197623e-01
8.84913087e-01 2.89200023e-02 -7.85846174e-01 2.55671620e-01
7.28820503e-01 -1.10806000e+00 1.13087213e+00 -1.16940103e-01
6.70319259e-01 2.57625908e-01 -5.75440645e-01 -8.66630614e-01
-2.07148999e-01 -2.93176383e-01 -5.77830553e-01 1.03201747e+00
-1.11183167e-01 -4.72088844e-01 1.03741026e+00 3.84385943e-01
-1.92900985e-01 -8.14072549e-01 -1.22623444e+00 -5.17430425e-01
-3.35009724e-01 -7.80936539e-01 9.09007311e-01 4.45624441e-01
-3.11157644e-01 -4.05487388e-01 -3.17885041e-01 6.68681860e-02
6.45367980e-01 3.46614569e-01 5.08723378e-01 -1.30786800e+00
-3.17799687e-01 -3.47575217e-01 -7.21215844e-01 -9.86791134e-01
1.98505342e-01 -1.11365449e+00 -3.15595597e-01 -1.28817713e+00
1.41322285e-01 -5.15984595e-01 -4.95586127e-01 5.70917010e-01
4.11129445e-01 4.38425064e-01 5.79620063e-01 3.49945068e-01
-7.80686140e-01 8.30420375e-01 1.24906707e+00 -1.85969666e-01
1.41021207e-01 -1.73856884e-01 -5.76163828e-01 4.97993916e-01
5.02932668e-01 -4.14878637e-01 -6.35540605e-01 -8.30631018e-01
2.47463703e-01 1.81632355e-01 4.99857157e-01 -1.53188086e+00
5.28143585e-01 2.07381442e-01 7.12338924e-01 -5.76134264e-01
6.41413212e-01 -1.48047149e+00 7.07008168e-02 6.90079033e-01
-3.50954533e-01 1.20400392e-01 -6.20883470e-03 3.42824161e-01
-3.09942484e-01 4.22722697e-02 6.05526924e-01 2.39972532e-01
-6.02059543e-01 6.10158443e-01 -2.14887410e-01 -4.82931495e-01
9.49273407e-01 -3.66953105e-01 -2.67696768e-01 -4.63832051e-01
-5.46716154e-01 -1.39409095e-01 2.06547648e-01 1.06159009e-01
1.02556527e+00 -1.18746400e+00 -4.10041690e-01 6.59416318e-01
-2.91401178e-01 -1.00089490e-01 4.75361735e-01 5.96105337e-01
-5.68802655e-01 3.90053868e-01 -6.42694831e-01 -4.54422921e-01
-1.33485079e+00 6.56308889e-01 1.68652013e-01 -3.14005613e-01
-8.70826125e-01 7.32282043e-01 -1.68788344e-01 1.61993578e-01
7.52423525e-01 -4.10330415e-01 -1.77000180e-01 6.09394088e-02
8.19733620e-01 5.82229316e-01 2.63866007e-01 -3.22076857e-01
-2.66080737e-01 3.31723511e-01 3.92568976e-01 3.46908182e-01
1.76651847e+00 -3.10621440e-01 -1.08911164e-01 5.57950065e-02
1.53724539e+00 -3.88054848e-01 -1.26090169e+00 -2.88306683e-01
-5.08077919e-01 -3.13215494e-01 6.75653338e-01 -4.79969293e-01
-1.98570693e+00 1.26608336e+00 1.03209984e+00 3.49804133e-01
1.83166850e+00 -2.27393031e-01 1.18514669e+00 6.94872618e-01
4.11144167e-01 -9.30512846e-01 -2.64983103e-02 1.41406342e-01
8.22850704e-01 -7.53007650e-01 2.75581889e-02 -5.70716262e-02
-2.35154331e-01 1.24375403e+00 1.33893549e-01 3.67754810e-02
6.83475494e-01 7.84080327e-01 -1.48629516e-01 -2.60303110e-01
-9.09165859e-01 2.94661168e-02 9.93390530e-02 7.44887888e-01
3.76357660e-02 1.01496629e-01 1.09416366e-01 7.79263079e-01
-4.07290787e-01 2.92069912e-01 -4.10828367e-02 6.57405972e-01
-4.71931249e-01 -6.81268334e-01 -1.85099378e-01 7.24413991e-01
-5.64957917e-01 -1.84821010e-01 1.83196455e-01 3.12526077e-01
2.99721152e-01 8.30040574e-01 3.80967051e-01 -7.00238466e-01
1.00990117e-01 -9.75505561e-02 1.20593764e-01 1.39811337e-01
-4.94497806e-01 -3.48092288e-01 -2.36218065e-01 -9.26050782e-01
-2.94675827e-01 -1.69204459e-01 -1.12837231e+00 -7.92281449e-01
5.16844876e-02 -3.85375917e-02 1.13369024e+00 7.82311559e-01
8.55383337e-01 6.12976849e-01 7.86416173e-01 -1.05326593e+00
-6.25672638e-01 -7.00927496e-01 -6.35938823e-01 2.42032498e-01
4.06049073e-01 -2.52591640e-01 -1.92677528e-01 8.65704790e-02]
|
[8.441634178161621, 2.859452247619629]
|
0971e9e0-90a7-4e9a-a2b2-c3361abfa893
|
cma-es-for-post-hoc-ensembling-in-automl-a
|
2307.00286
| null |
https://arxiv.org/abs/2307.00286v1
|
https://arxiv.org/pdf/2307.00286v1.pdf
|
CMA-ES for Post Hoc Ensembling in AutoML: A Great Success and Salvageable Failure
|
Many state-of-the-art automated machine learning (AutoML) systems use greedy ensemble selection (GES) by Caruana et al. (2004) to ensemble models found during model selection post hoc. Thereby, boosting predictive performance and likely following Auto-Sklearn 1's insight that alternatives, like stacking or gradient-free numerical optimization, overfit. Overfitting in Auto-Sklearn 1 is much more likely than in other AutoML systems because it uses only low-quality validation data for post hoc ensembling. Therefore, we were motivated to analyze whether Auto-Sklearn 1's insight holds true for systems with higher-quality validation data. Consequently, we compared the performance of covariance matrix adaptation evolution strategy (CMA-ES), state-of-the-art gradient-free numerical optimization, to GES on the 71 classification datasets from the AutoML benchmark for AutoGluon. We found that Auto-Sklearn's insight depends on the chosen metric. For the metric ROC AUC, CMA-ES overfits drastically and is outperformed by GES -- statistically significantly for multi-class classification. For the metric balanced accuracy, CMA-ES does not overfit and outperforms GES significantly. Motivated by the successful application of CMA-ES for balanced accuracy, we explored methods to stop CMA-ES from overfitting for ROC AUC. We propose a method to normalize the weights produced by CMA-ES, inspired by GES, that avoids overfitting for CMA-ES and makes CMA-ES perform better than or similar to GES for ROC AUC.
|
['Joeran Beel', 'Lennart Purucker']
|
2023-07-01
| null | null | null | null |
['model-selection', 'automl']
|
['methodology', 'methodology']
|
[-1.28464401e-01 5.94644882e-02 2.32379153e-01 -1.13137983e-01
-7.91792035e-01 -5.71042120e-01 5.61676145e-01 3.12405020e-01
-5.45882761e-01 9.64968979e-01 -2.80841827e-01 -5.20265877e-01
-5.12129724e-01 -6.80562019e-01 -6.11688375e-01 -7.96230912e-01
-1.95915475e-01 5.95030248e-01 3.28256078e-02 -3.83267552e-01
3.08867723e-01 3.92629892e-01 -1.68275857e+00 2.18575329e-01
1.13519573e+00 8.84251475e-01 -2.91620106e-01 5.94276130e-01
3.46588232e-02 4.35692728e-01 -3.27359825e-01 -6.47061348e-01
2.51147300e-01 -6.30767643e-01 -6.91498816e-01 -5.77691913e-01
3.74980181e-01 4.55545634e-01 3.90516996e-01 7.56857097e-01
4.61895972e-01 4.86016832e-03 9.98550713e-01 -1.42501557e+00
4.45268899e-02 9.04851913e-01 -1.67839050e-01 -1.44466966e-01
5.62174246e-02 4.31876868e-01 8.58629107e-01 -9.90275919e-01
5.05512059e-01 9.82389212e-01 1.38076317e+00 4.51886594e-01
-1.49250138e+00 -5.87848723e-01 -4.58208881e-02 6.37533888e-02
-1.58520341e+00 -1.63873583e-01 3.66436541e-01 -5.56034446e-01
1.02683008e+00 6.39593899e-01 8.22425008e-01 9.96089697e-01
4.69496697e-01 5.34194231e-01 1.33297348e+00 -6.69270873e-01
6.72509253e-01 4.28509921e-01 2.38969564e-01 6.49153054e-01
5.89873314e-01 3.68665278e-01 -5.23670137e-01 -3.08096975e-01
2.15841040e-01 -6.67236745e-01 -1.98061079e-01 -5.43000460e-01
-1.24688089e+00 6.65203810e-01 1.59904614e-01 4.86632705e-01
-4.47382540e-01 -2.35213013e-03 3.20759714e-01 3.88843000e-01
4.52851415e-01 1.30005121e+00 -8.02426040e-01 -3.41116309e-01
-1.18631792e+00 4.67439324e-01 9.64694381e-01 4.38452810e-01
5.50601184e-01 1.86319545e-01 -2.99515456e-01 7.12850809e-01
-1.12776019e-01 2.93664664e-01 7.86546707e-01 -6.69470608e-01
8.62902254e-02 7.54893005e-01 -4.30111587e-02 -5.19581497e-01
-7.59392202e-01 -1.04714513e+00 -1.02315927e+00 5.60886204e-01
6.57248676e-01 -3.27233344e-01 -6.13381445e-01 1.69882143e+00
1.78894013e-01 -1.41205460e-01 1.64184779e-01 6.78135574e-01
3.46231997e-01 1.28215104e-01 5.84514476e-02 -3.57898861e-01
9.70165133e-01 -7.81857371e-01 -2.07157359e-01 1.68665364e-01
1.20053744e+00 -4.51495707e-01 9.36231077e-01 7.64326453e-01
-7.34610677e-01 -5.31057417e-01 -1.26592410e+00 7.74973989e-01
-5.48888683e-01 2.42030874e-01 7.02301025e-01 9.13173020e-01
-1.02930522e+00 1.21200705e+00 -8.82872403e-01 -4.13322598e-01
2.76256025e-01 4.61773902e-01 -4.64013368e-01 4.00863498e-01
-1.05064940e+00 1.27948272e+00 7.41860330e-01 6.13889843e-02
-5.73704958e-01 -7.74103343e-01 -5.54726005e-01 5.60428351e-02
3.59510601e-01 -8.44291508e-01 9.95479524e-01 -1.06032908e+00
-1.57717252e+00 4.97248411e-01 2.84926668e-02 -8.50110173e-01
7.33155191e-01 -3.39117587e-01 -4.22264963e-01 -4.86209989e-01
-3.33185077e-01 3.75043988e-01 7.75332093e-01 -1.31021810e+00
-1.38919890e-01 -3.31639618e-01 -5.50727427e-01 -1.07605964e-01
-2.81824678e-01 -3.90056759e-01 2.04599589e-01 -5.55164039e-01
1.15029193e-01 -1.06541908e+00 -3.95291030e-01 -7.07268834e-01
-4.17144209e-01 -9.23896357e-02 4.31190103e-01 -3.70173544e-01
1.75609326e+00 -1.80704355e+00 1.60671458e-01 6.53284073e-01
5.22762015e-02 4.95333314e-01 -1.30931437e-01 4.77778614e-01
-4.76030022e-01 2.74771005e-01 -5.70154488e-01 -2.68208385e-01
-1.17134549e-01 -3.17409150e-02 -3.35390009e-02 2.44338393e-01
2.16974229e-01 7.74566352e-01 -7.69945025e-01 -3.85092586e-01
1.32339478e-01 2.28570357e-01 -5.24804235e-01 -1.12395555e-01
-1.98241353e-01 2.49254405e-01 -1.73288345e-01 5.73002279e-01
5.07975936e-01 -3.10296118e-01 1.39658287e-01 -2.88606495e-01
-2.13713259e-01 -1.90138727e-01 -1.31023026e+00 1.18706203e+00
-4.63999182e-01 2.47835249e-01 -3.45447034e-01 -8.94415319e-01
1.03390884e+00 1.30532533e-01 3.36336195e-01 -3.13214928e-01
1.83627859e-01 7.08447695e-01 2.67815948e-01 -1.36463940e-01
3.06891888e-01 -1.21433005e-01 1.25063151e-01 2.50181258e-01
3.64925921e-01 -1.55657545e-01 2.39864126e-01 1.97285101e-01
1.19651628e+00 5.52452505e-01 6.76776946e-01 -5.95299423e-01
9.00691152e-01 7.92529210e-02 4.71434236e-01 1.05170727e+00
1.70413911e-01 5.92863619e-01 2.64032483e-01 -3.62804770e-01
-9.86081481e-01 -7.80456185e-01 -3.25572103e-01 7.44609773e-01
-3.17543417e-01 -7.60955334e-01 -1.05916572e+00 -9.05669332e-01
2.09122360e-01 1.32912624e+00 -7.09818184e-01 -2.69521266e-01
-3.36326212e-01 -1.11832333e+00 5.50668359e-01 2.28387207e-01
2.52986491e-01 -1.00527811e+00 -5.54280698e-01 5.08243680e-01
1.98356792e-01 -6.69809520e-01 1.65505439e-01 5.48799396e-01
-1.09295619e+00 -9.75670457e-01 -6.30151927e-01 1.45972803e-01
5.09392142e-01 -5.36477864e-01 1.34392083e+00 2.31751904e-01
-2.22540975e-01 3.60844404e-01 -5.71621656e-01 -5.87923646e-01
-8.35047364e-01 5.44185400e-01 3.17218155e-01 2.56742816e-03
3.07940185e-01 -6.35597229e-01 -2.15307593e-01 4.91913110e-01
-5.58459103e-01 4.22047637e-03 6.42977595e-01 1.10985518e+00
4.26642478e-01 5.39723821e-02 6.02153957e-01 -6.95802748e-01
4.96460736e-01 -1.95430681e-01 -5.24878383e-01 4.51086432e-01
-1.25680721e+00 5.82518220e-01 7.87585437e-01 -3.04087222e-01
-5.22157907e-01 -2.58577596e-02 -1.85712755e-01 -6.02696121e-01
-1.87538508e-02 4.70132291e-01 6.64853528e-02 -2.84219265e-01
1.03896081e+00 1.77035909e-02 1.26290604e-01 -5.42460501e-01
2.60723885e-02 5.03392041e-01 2.30845720e-01 -6.99951470e-01
6.65140927e-01 -2.84518898e-02 2.68317789e-01 -6.89077914e-01
-7.91675866e-01 -1.24200545e-01 -4.78476852e-01 -2.94079542e-01
7.23882318e-01 -5.46219826e-01 -7.39358664e-01 6.30449891e-01
-7.68591583e-01 -3.85730118e-01 -4.64258850e-01 6.24584079e-01
-5.76179206e-01 2.82051768e-02 4.95630875e-02 -1.27752113e+00
-5.04348218e-01 -1.07125580e+00 7.75601268e-01 3.18815932e-02
-6.55448079e-01 -9.62036490e-01 2.62773186e-01 6.46762699e-02
5.72135925e-01 3.32316160e-01 7.26893485e-01 -1.12382174e+00
-1.33189142e-01 -2.32865721e-01 2.99349517e-01 4.83487666e-01
-2.19596982e-01 2.03695923e-01 -1.05206943e+00 -3.63743752e-01
-2.62794971e-01 -1.10107504e-01 9.38741505e-01 2.83840597e-01
1.27135336e+00 -2.12664992e-01 -3.69639516e-01 5.69591045e-01
1.49369383e+00 9.06068310e-02 5.41090846e-01 8.48076999e-01
3.80951971e-01 4.29348111e-01 5.98488927e-01 2.59734809e-01
-9.47591141e-02 7.65536666e-01 3.29181254e-01 7.40264356e-02
2.85797805e-01 -1.55239895e-01 4.20177788e-01 7.30059147e-01
-5.29843092e-01 -5.62263429e-02 -9.55278754e-01 1.85880721e-01
-1.89540732e+00 -8.50613117e-01 -2.98880398e-01 2.44525695e+00
6.45953238e-01 4.27362889e-01 3.19044501e-01 4.38921720e-01
4.12084430e-01 -3.72334272e-01 -5.90457320e-01 -5.46571672e-01
-3.46674293e-01 3.65013510e-01 5.73334396e-01 3.86451274e-01
-9.50065255e-01 6.13542914e-01 5.28804350e+00 1.24573314e+00
-8.49399567e-01 2.71331761e-02 7.25017190e-01 -1.33060634e-01
-1.90311134e-01 2.79483914e-01 -9.80240643e-01 5.01640737e-01
1.15571165e+00 -3.58523289e-03 4.94638681e-01 9.66229379e-01
-1.45968884e-01 -3.07522565e-01 -1.14178073e+00 8.23838651e-01
-9.34495106e-02 -1.29002309e+00 -5.57064302e-02 3.77720259e-02
8.26791883e-01 -6.50955513e-02 -2.70669430e-01 6.69462323e-01
3.40945989e-01 -1.18428934e+00 7.52591133e-01 7.79404104e-01
5.65301478e-01 -7.84145951e-01 1.05422330e+00 5.36450744e-01
-8.03172290e-01 -1.12073235e-01 -1.31834717e-02 2.34123096e-01
-9.57417265e-02 9.81632948e-01 -6.48643792e-01 1.05890727e+00
7.47086704e-01 3.54709327e-01 -9.82841909e-01 1.06300461e+00
9.81450379e-02 9.89027143e-01 -4.80728656e-01 -2.33028978e-01
2.18044013e-01 -3.21884841e-01 9.64909792e-01 1.29107165e+00
6.66758597e-01 -1.15835154e-02 -3.55363607e-01 6.30967677e-01
4.97743189e-01 3.42153847e-01 -4.66402978e-01 -5.71046732e-02
3.75335425e-01 1.25878513e+00 -5.58716953e-01 -3.75723153e-01
2.52790511e-01 6.69325352e-01 1.98342249e-01 1.16598576e-01
-6.34188473e-01 -2.58511484e-01 2.72798836e-01 2.11691678e-01
2.79944927e-01 9.11768153e-02 -6.67744160e-01 -1.04163361e+00
-2.64536858e-01 -1.34122884e+00 4.80055273e-01 -8.77180934e-01
-1.24947238e+00 9.81443644e-01 5.35889938e-02 -1.35023355e+00
-3.72533530e-01 -8.11235309e-01 -5.47033966e-01 7.59963751e-01
-8.20413351e-01 -8.09472382e-01 -3.03107589e-01 1.60492226e-01
1.56760484e-01 -5.38758039e-01 9.53119338e-01 -2.97502756e-01
-5.80915391e-01 8.50992620e-01 3.34649682e-01 -3.90918583e-01
6.56172276e-01 -1.32965016e+00 2.06924945e-01 6.14259422e-01
1.83273226e-01 6.22727215e-01 1.01700747e+00 -5.15374839e-01
-1.08172572e+00 -7.76009321e-01 6.73334539e-01 -6.57691300e-01
5.48081875e-01 -7.98722357e-02 -1.05048907e+00 2.83865094e-01
1.34341240e-01 -4.72130209e-01 5.48347950e-01 4.12878841e-01
-1.17641434e-01 -1.98553488e-01 -1.13054860e+00 6.11771822e-01
9.39712405e-01 3.00611500e-02 -3.93751025e-01 9.10959393e-02
1.99592799e-01 -1.64724335e-01 -1.09974456e+00 9.10780132e-01
7.17590570e-01 -1.28394246e+00 7.48013139e-01 -6.52102470e-01
3.87964040e-01 -2.08521560e-01 -7.53368512e-02 -1.69115174e+00
-1.66037932e-01 -6.86287522e-01 -1.10150300e-01 1.25949883e+00
7.33293712e-01 -1.10636866e+00 5.22074938e-01 3.75353217e-01
-3.33044007e-02 -1.18927729e+00 -9.01764870e-01 -1.20407832e+00
3.22759509e-01 -5.69466054e-01 7.19021142e-01 8.74505699e-01
-1.40921667e-01 1.55595332e-01 -1.74619883e-01 -2.26415128e-01
7.84437656e-01 -1.14415467e-01 1.03848720e+00 -1.49548614e+00
-6.34079397e-01 -8.04074764e-01 -3.05824548e-01 -4.89067622e-02
-1.87070458e-03 -1.04224539e+00 -1.85874701e-01 -9.76701796e-01
-7.90829211e-02 -5.83724141e-01 -4.34654266e-01 6.44815266e-01
-3.01348418e-01 1.41433179e-01 4.12843138e-01 8.92627090e-02
-3.73245209e-01 4.67468351e-01 9.62696552e-01 2.62070507e-01
-4.35710698e-01 -5.99759221e-02 -5.49917221e-01 7.20284402e-01
8.23037565e-01 -5.72696328e-01 7.80168399e-02 4.84449536e-01
5.62499046e-01 3.86867416e-03 3.98874313e-01 -1.48671937e+00
-1.64521649e-01 1.51089579e-01 5.85903704e-01 -5.30270219e-01
8.06969777e-02 -6.52689576e-01 5.76208770e-01 5.03276587e-01
-1.60290226e-01 6.13246895e-02 4.58336920e-01 1.51971862e-01
-3.71927805e-02 -6.34730637e-01 5.92695236e-01 -2.34994781e-03
-3.09325874e-01 -1.67518497e-01 -4.77900267e-01 -1.99487805e-02
8.37119401e-01 -4.41325456e-01 -5.49936220e-02 -1.99357331e-01
-9.16710496e-01 1.91483840e-01 5.17093897e-01 9.77029800e-02
7.71340057e-02 -1.02351928e+00 -8.90371382e-01 4.50553179e-01
-7.85869639e-03 -2.56985664e-01 5.75745888e-02 1.24836206e+00
-2.66748160e-01 4.07166719e-01 4.41859961e-02 -7.43166864e-01
-1.39414561e+00 2.68046916e-01 5.74493706e-01 -7.57677913e-01
-2.09130585e-01 8.44148874e-01 -3.77454340e-01 -7.86350608e-01
-1.95607841e-01 -2.23183855e-01 -6.75493926e-02 9.59962681e-02
1.61725748e-02 6.35353923e-01 5.98110557e-01 -2.42690951e-01
-5.08992195e-01 5.93064427e-01 8.70810263e-03 -1.82710662e-01
1.32353771e+00 3.88063520e-01 3.05631515e-02 5.79302073e-01
8.19299400e-01 -7.77198076e-02 -7.91411519e-01 3.29968959e-01
1.01921521e-01 -8.13172385e-02 2.20670225e-03 -1.31607938e+00
-5.92659354e-01 6.45886242e-01 6.38582289e-01 1.25825942e-01
1.03491580e+00 -4.50653076e-01 4.92544323e-02 5.27044117e-01
7.71846056e-01 -1.06252003e+00 -2.10212037e-01 6.21695578e-01
1.07530761e+00 -1.08281112e+00 3.78911257e-01 3.74248438e-02
-8.48222375e-01 1.18884492e+00 6.79568470e-01 -4.45857272e-02
6.32779181e-01 1.98809937e-01 -6.21949229e-03 9.07314289e-03
-9.55126166e-01 -4.42801341e-02 4.23711240e-01 1.55106500e-01
3.57531607e-01 -2.65129991e-02 -5.34070909e-01 7.48282671e-01
-7.37832367e-01 1.61199719e-01 1.36823431e-01 6.98174119e-01
-1.16924137e-01 -1.09653068e+00 -6.34819031e-01 7.45567083e-01
-7.00767338e-02 -6.84210584e-02 -4.73313898e-01 1.09047914e+00
2.22376868e-01 6.81150377e-01 -9.36293676e-02 -7.19055831e-01
2.11282447e-01 6.05823576e-01 5.61269939e-01 -8.58656317e-02
-1.16731310e+00 -8.23595077e-02 3.17985952e-01 -4.22676235e-01
-1.73709944e-01 -8.28424990e-01 -9.14057672e-01 -1.13721341e-01
-7.60731339e-01 5.86918950e-01 8.50959539e-01 1.07459533e+00
5.02980292e-01 5.71209311e-01 3.82799000e-01 -8.34868193e-01
-9.88343418e-01 -1.13052964e+00 -4.38627571e-01 1.36699617e-01
-7.52686933e-02 -8.42297316e-01 -8.76307905e-01 -5.77603400e-01]
|
[8.070235252380371, 4.28082799911499]
|
281b4767-56f0-4e06-9420-48260fbfa093
|
deepps2-revisiting-photometric-stereo-using
|
2207.02025
| null |
https://arxiv.org/abs/2207.02025v2
|
https://arxiv.org/pdf/2207.02025v2.pdf
|
DeepPS2: Revisiting Photometric Stereo Using Two Differently Illuminated Images
|
Photometric stereo, a problem of recovering 3D surface normals using images of an object captured under different lightings, has been of great interest and importance in computer vision research. Despite the success of existing traditional and deep learning-based methods, it is still challenging due to: (i) the requirement of three or more differently illuminated images, (ii) the inability to model unknown general reflectance, and (iii) the requirement of accurate 3D ground truth surface normals and known lighting information for training. In this work, we attempt to address an under-explored problem of photometric stereo using just two differently illuminated images, referred to as the PS2 problem. It is an intermediate case between a single image-based reconstruction method like Shape from Shading (SfS) and the traditional Photometric Stereo (PS), which requires three or more images. We propose an inverse rendering-based deep learning framework, called DeepPS2, that jointly performs surface normal, albedo, lighting estimation, and image relighting in a completely self-supervised manner with no requirement of ground truth data. We demonstrate how image relighting in conjunction with image reconstruction enhances the lighting estimation in a self-supervised setting.
|
['Shanmuganathan Raman', 'Ashish Tiwari']
|
2022-07-05
| null | null | null | null |
['lighting-estimation', 'image-relighting']
|
['computer-vision', 'computer-vision']
|
[ 7.51517415e-01 -2.15471938e-01 5.75382173e-01 -5.63827395e-01
-7.28141248e-01 -4.78151202e-01 6.58575296e-01 -3.35985124e-01
-1.11786939e-01 6.17715478e-01 -1.19326048e-01 -2.45788038e-01
1.15392089e-01 -7.85998046e-01 -9.24009323e-01 -8.09917390e-01
5.57465613e-01 5.90537727e-01 7.63836801e-02 -3.10228258e-01
3.84944260e-01 7.51993060e-01 -2.07037377e+00 -1.16594888e-01
9.00527060e-01 1.14900661e+00 2.84423381e-01 5.86848617e-01
-3.74988496e-01 5.38753033e-01 -2.38075033e-01 -2.43158907e-01
6.50871515e-01 -3.58545154e-01 -5.32481611e-01 4.90959525e-01
1.03888309e+00 -5.52971005e-01 -4.09840085e-02 9.32923913e-01
4.61041987e-01 1.33056447e-01 5.59272528e-01 -9.98648703e-01
-4.55001712e-01 -5.61540306e-01 -7.42033064e-01 -2.87195653e-01
4.01832670e-01 1.74107656e-01 6.91249788e-01 -7.77862310e-01
3.38393092e-01 1.15636015e+00 6.82374597e-01 3.53742987e-01
-1.49650562e+00 -2.74630159e-01 -2.67083853e-01 3.62536847e-03
-1.19535148e+00 -5.45604050e-01 1.19635534e+00 -5.65841138e-01
4.48104680e-01 2.71146327e-01 7.36232817e-01 7.74648428e-01
-1.81488886e-01 3.42461944e-01 1.85916948e+00 -5.93227446e-01
3.29511315e-01 1.77886903e-01 -1.82990059e-01 6.86278522e-01
-1.28072416e-02 2.94481039e-01 -5.99321902e-01 -9.53644961e-02
8.99076641e-01 1.80545136e-01 -5.97888708e-01 -4.89558458e-01
-8.11365187e-01 6.67669833e-01 3.14498514e-01 -2.01285437e-01
-3.48782927e-01 -1.03146778e-02 2.31810380e-02 2.71108598e-01
7.80970931e-01 9.84793082e-02 -4.27038580e-01 2.52261549e-01
-8.84937346e-01 2.99455430e-02 8.69527936e-01 6.61449909e-01
1.58561420e+00 2.68454224e-01 4.78518635e-01 9.40743208e-01
3.98116738e-01 8.62290740e-01 1.67141706e-01 -1.27655041e+00
1.77628160e-01 5.97166479e-01 3.82223934e-01 -9.16805565e-01
-1.98632125e-02 -2.21014887e-01 -6.67808592e-01 7.61744499e-01
2.89246678e-01 7.77374879e-02 -8.22474360e-01 1.63399637e+00
6.57002330e-01 1.58520132e-01 1.55263215e-01 1.00472105e+00
8.01254570e-01 5.42716801e-01 -6.93173110e-01 -1.43779650e-01
1.02359331e+00 -7.49754727e-01 -4.33191001e-01 -4.45774913e-01
1.03752308e-01 -1.02586520e+00 1.28398907e+00 5.92010200e-01
-1.00385451e+00 -4.18219537e-01 -9.08297122e-01 -4.83033746e-01
-2.50768572e-01 -8.70229304e-02 4.95911300e-01 7.10231185e-01
-1.33416915e+00 5.39884210e-01 -5.03021479e-01 -2.95197397e-01
1.90446720e-01 1.86188400e-01 -3.56014669e-01 -3.38754475e-01
-8.19243312e-01 7.34551251e-01 -2.43761674e-01 1.98925480e-01
-8.04463625e-01 -6.53156698e-01 -9.81224358e-01 -2.19797090e-01
2.55369931e-01 -6.36274934e-01 7.76605785e-01 -1.16095924e+00
-1.80852640e+00 1.27036011e+00 -4.07129586e-01 6.29421771e-02
6.44800901e-01 -2.56350130e-01 7.32738897e-02 1.17392190e-01
-1.57720372e-01 3.41026843e-01 1.01219273e+00 -1.94775200e+00
-2.17030138e-01 -8.06450963e-01 1.64210126e-01 5.48887849e-01
1.45033898e-03 -3.63916934e-01 -5.11021078e-01 -1.94481239e-01
4.01117265e-01 -7.95801342e-01 -4.07678969e-02 3.15592438e-01
-3.55207443e-01 1.36186615e-01 8.22379649e-01 -7.77846813e-01
2.94813305e-01 -1.95781589e+00 -8.98476969e-03 -6.09911866e-02
-2.04031076e-02 9.27040353e-02 -1.39055043e-01 3.68363082e-01
-3.82989943e-02 -3.84413689e-01 -4.48338270e-01 -7.12047338e-01
-2.41843984e-01 4.25453544e-01 -4.36459571e-01 7.40609348e-01
4.98598488e-03 5.13192058e-01 -7.93664694e-01 -3.97008777e-01
7.09569395e-01 9.50294554e-01 -2.49304622e-01 5.07068038e-01
-1.29884526e-01 9.88959968e-01 -8.36756378e-02 4.55653191e-01
1.05231869e+00 -1.74249206e-02 -1.07812546e-01 -4.67103064e-01
-4.46585506e-01 9.23302844e-02 -1.24758017e+00 1.56473494e+00
-9.15764332e-01 6.95815742e-01 2.75322109e-01 -9.20302093e-01
1.26796615e+00 1.29544437e-01 5.44384122e-01 -9.42562282e-01
1.46243861e-02 3.08722109e-01 -6.91450596e-01 -4.81845438e-01
2.95844406e-01 -4.69514757e-01 6.05327785e-01 5.77451646e-01
-3.94304186e-01 -8.13780844e-01 -2.95529902e-01 -2.25961193e-01
6.53021097e-01 6.00193202e-01 -5.58439096e-05 -1.25182733e-01
6.59896374e-01 -1.27001539e-01 6.30250752e-01 3.14685673e-01
2.44732544e-01 1.12116337e+00 -3.28739919e-02 -6.50446415e-01
-1.23263085e+00 -9.64116216e-01 -3.18800569e-01 6.25736892e-01
3.47399265e-01 2.10457385e-01 -7.04754472e-01 -1.13230295e-01
-1.02987066e-02 5.84896743e-01 -5.21098316e-01 2.92474955e-01
-5.92060566e-01 -4.26911324e-01 -8.10501501e-02 7.28569850e-02
7.47959554e-01 -7.83972979e-01 -6.77341104e-01 -2.32931882e-01
-2.94197261e-01 -1.15181756e+00 -2.20187828e-01 3.54855135e-02
-9.04129565e-01 -1.26242959e+00 -8.97912979e-01 -5.50750196e-01
8.14161897e-01 8.74805808e-01 1.21475112e+00 2.30774265e-02
-2.95359105e-01 7.73677170e-01 -7.38805905e-02 -3.43581647e-01
-3.35159153e-01 -5.08483887e-01 -2.61974782e-01 5.08731961e-01
3.81237194e-02 -8.66688728e-01 -8.52933228e-01 5.63009858e-01
-1.12872565e+00 3.46642584e-01 2.30876133e-01 5.96401989e-01
8.23251367e-01 -7.52007440e-02 -7.45986849e-02 -1.03637099e+00
-2.51247734e-02 -6.62338808e-02 -1.04581738e+00 2.28203952e-01
-5.21508276e-01 -1.25834838e-01 5.72891295e-01 3.74122523e-02
-1.40638816e+00 2.51413733e-01 -7.26316944e-02 -4.86296922e-01
-2.97586501e-01 2.97899786e-02 -2.00729951e-01 -5.33825219e-01
7.46637642e-01 5.04660130e-01 2.72920340e-01 -5.25999486e-01
1.20344564e-01 6.06590390e-01 5.73314130e-01 -3.83978754e-01
1.11971545e+00 1.16970515e+00 3.86076152e-01 -1.13793206e+00
-1.08571041e+00 -5.31196475e-01 -7.93328822e-01 -3.71483564e-01
8.08194637e-01 -1.00625873e+00 -7.37120211e-01 6.65108681e-01
-1.18002331e+00 -5.52018106e-01 -1.54274449e-01 1.64069757e-01
-8.52922678e-01 8.98387372e-01 -1.60483107e-01 -9.80634749e-01
-4.02188540e-01 -1.18882227e+00 1.54536760e+00 1.65977612e-01
2.78789908e-01 -1.02999020e+00 1.16120167e-01 9.50712383e-01
3.66776794e-01 4.88246411e-01 6.35985851e-01 3.89907032e-01
-8.70302022e-01 -2.16527041e-02 -5.00983059e-01 8.01548958e-01
4.22174513e-01 -2.23674208e-01 -1.39869988e+00 -2.82778651e-01
4.78500932e-01 -5.82301140e-01 5.78485966e-01 2.30390787e-01
1.12040877e+00 6.30520843e-03 3.12115252e-01 1.07973278e+00
1.85861540e+00 -1.19157016e-01 6.79170191e-01 3.07040364e-01
9.23423409e-01 1.01913881e+00 4.65376228e-01 4.34904933e-01
5.42245269e-01 7.68691480e-01 8.45507085e-01 -5.46261132e-01
-3.05206090e-01 2.11381968e-02 1.10362157e-01 6.16505027e-01
-2.68827707e-01 2.23677661e-02 -7.90506482e-01 3.28762412e-01
-1.50386763e+00 -5.83230376e-01 -4.41932887e-01 2.69144750e+00
8.27574551e-01 -4.63225663e-01 -4.39861596e-01 2.36297294e-01
5.07858157e-01 2.14957878e-01 -7.75229514e-01 -2.79761672e-01
-3.64055753e-01 2.54816208e-02 5.08982360e-01 7.44904339e-01
-7.00757086e-01 8.32929492e-01 5.50919533e+00 3.31212014e-01
-1.27816510e+00 2.08041519e-02 6.86367571e-01 3.31916839e-01
-7.15395033e-01 1.58360049e-01 -4.58287060e-01 3.23212922e-01
3.00238401e-01 3.73123676e-01 1.08400190e+00 5.16971231e-01
2.84525424e-01 -4.40473676e-01 -1.01177442e+00 1.39321423e+00
5.24125278e-01 -9.07474577e-01 -6.62430450e-02 1.17000118e-01
9.99916136e-01 3.23117614e-01 1.00894170e-02 -3.22938710e-01
2.23055497e-01 -6.82183564e-01 6.08232200e-01 5.69881499e-01
9.38272834e-01 -2.76789486e-01 2.63811141e-01 2.88753301e-01
-8.18767428e-01 2.41569608e-01 -4.66934949e-01 9.39003378e-02
2.59788424e-01 8.91528487e-01 -4.14638579e-01 6.10131741e-01
8.44705582e-01 6.13098264e-01 -1.66211948e-01 7.66975701e-01
-4.86597955e-01 3.48795891e-01 -4.34045076e-01 5.47109067e-01
-8.46849754e-02 -8.28030884e-01 4.22187895e-01 6.75725102e-01
2.63428748e-01 7.00768530e-02 8.76865443e-03 9.13032770e-01
9.69542488e-02 -5.79698049e-02 -7.15811372e-01 5.20593762e-01
5.71129322e-02 1.29254997e+00 -4.58862305e-01 -1.33565113e-01
-4.76324141e-01 1.11565793e+00 1.64977148e-01 6.39990151e-01
-5.05904615e-01 -1.07372524e-02 5.26734531e-01 2.44936615e-01
-2.95835063e-02 -4.35278356e-01 -3.95353496e-01 -1.20418036e+00
2.21419156e-01 -6.02788329e-01 -8.84875581e-02 -1.27348518e+00
-1.29656339e+00 4.15347099e-01 -3.41393501e-01 -1.20766115e+00
1.09255157e-01 -5.75145245e-01 -5.76692104e-01 1.23781371e+00
-2.32650042e+00 -1.16882968e+00 -9.03479934e-01 8.45071375e-01
5.72934747e-01 2.23655388e-01 6.94733918e-01 2.40931094e-01
-2.32295424e-01 -4.06816751e-02 3.32183838e-01 -3.74593854e-01
7.65583158e-01 -1.41103113e+00 1.88938364e-01 5.35233915e-01
-8.12772214e-02 2.02448606e-01 8.69291306e-01 -2.75191933e-01
-1.89406919e+00 -9.66622829e-01 6.42523110e-01 -2.69707203e-01
1.79976806e-01 -2.70286083e-01 -9.26622510e-01 4.25710112e-01
8.79800692e-02 1.69012845e-01 5.11455894e-01 -2.85812497e-01
-3.11125427e-01 -5.17436922e-01 -1.16933584e+00 5.59894800e-01
9.55612421e-01 -6.91226661e-01 -2.96845913e-01 6.81918323e-01
4.27298903e-01 -4.18108970e-01 -6.50643945e-01 2.93889880e-01
4.22003627e-01 -1.52872968e+00 1.09747779e+00 1.82327613e-01
4.45713073e-01 -4.10454780e-01 -4.33374286e-01 -1.25657165e+00
1.43164754e-01 -6.35594726e-01 3.29125166e-01 1.15485871e+00
-2.58633588e-02 -9.36565936e-01 8.48322213e-01 7.60410309e-01
-1.77136928e-01 -4.47039902e-01 -7.93231130e-01 -5.66002548e-01
-2.67884374e-01 -1.94869652e-01 5.73345542e-01 1.00410235e+00
-9.68344510e-01 2.45164648e-01 -5.50410151e-01 4.15855199e-01
1.10315621e+00 6.31052315e-01 1.25048423e+00 -1.47181213e+00
-3.39206576e-01 1.38885891e-02 -1.35888621e-01 -1.15103388e+00
2.28898451e-01 -5.09690225e-01 3.31373751e-01 -1.69082403e+00
7.11932033e-02 -5.48610330e-01 2.69531846e-01 2.20516533e-01
4.14628796e-02 5.73823929e-01 -1.98990121e-01 4.37222749e-01
-1.66515633e-01 8.11443210e-01 1.31819475e+00 -2.12454461e-02
-8.61427262e-02 -6.62519410e-02 -3.62498701e-01 7.64241874e-01
6.16699398e-01 -1.09491833e-01 -5.31383276e-01 -8.93274128e-01
4.35173154e-01 6.39054775e-02 5.30484915e-01 -7.10573792e-01
2.88424129e-03 -1.92468539e-01 2.84128994e-01 -6.68064594e-01
8.27421486e-01 -9.41855788e-01 2.11203396e-01 1.21132724e-01
1.12003215e-01 -2.34783754e-01 -1.36175781e-01 5.25746047e-01
-1.43786520e-01 -2.17118010e-01 8.56045723e-01 -3.05644363e-01
-4.97846663e-01 2.68110305e-01 1.57904863e-01 -1.83593214e-01
6.22368693e-01 -6.99457765e-01 -1.38761580e-01 -3.74974877e-01
-2.05170006e-01 -5.38037047e-02 1.04870629e+00 -2.13729423e-02
7.03678131e-01 -9.90684211e-01 -5.61266601e-01 3.94094735e-01
7.54227340e-02 4.20483321e-01 3.68518472e-01 6.50161326e-01
-8.79124463e-01 -4.81875762e-02 1.73108354e-02 -9.10715759e-01
-1.18070126e+00 2.46673152e-01 3.89309406e-01 1.39173388e-01
-8.18408966e-01 4.76373523e-01 6.12490594e-01 -8.53325546e-01
7.93444216e-02 1.64995387e-01 9.46812183e-02 -3.83801311e-01
2.79613256e-01 4.13423955e-01 2.82987982e-01 -7.72963643e-01
1.07532293e-01 1.16323316e+00 4.62406546e-01 -7.85290301e-02
1.46903861e+00 -5.30714750e-01 -3.77983123e-01 4.78570521e-01
1.21225119e+00 -1.91223308e-01 -1.74497592e+00 -5.19766510e-01
-3.97245049e-01 -9.48438644e-01 5.21235883e-01 -6.17069483e-01
-1.28329253e+00 1.17030394e+00 5.78508794e-01 4.00405303e-02
1.31063509e+00 -2.70200789e-01 9.24770772e-01 4.09495831e-01
5.61384261e-01 -9.45441365e-01 -1.36161111e-02 4.21298683e-01
9.89651501e-01 -1.64475942e+00 1.36084259e-01 -5.06531298e-01
-3.93314511e-01 1.21308482e+00 3.60595942e-01 -4.69658226e-02
5.54727197e-01 6.69502243e-02 4.25266773e-01 -2.59477824e-01
-2.50697523e-01 -2.40387186e-01 2.49944270e-01 6.17514789e-01
3.45883161e-01 -3.27770352e-01 1.26416504e-01 -4.43620563e-01
-6.00935705e-02 -7.66371787e-02 3.45106989e-01 8.32436621e-01
-4.18900967e-01 -1.09377766e+00 -6.78346634e-01 9.92224142e-02
-2.29803324e-02 2.50182580e-02 -3.98346007e-01 3.86253327e-01
-3.98309901e-02 8.56863260e-01 7.16148131e-03 2.38233302e-02
1.62924975e-01 -2.31680006e-01 5.90330720e-01 -5.43737173e-01
-1.13116764e-01 1.02800392e-01 -1.44596502e-01 -6.44427061e-01
-7.85762072e-01 -6.55628443e-01 -8.49571288e-01 -1.22178838e-01
-3.14827770e-01 -2.67668933e-01 1.17816162e+00 9.17255282e-01
2.25023136e-01 9.51452367e-03 1.21771467e+00 -1.30409467e+00
-3.05073649e-01 -6.51287019e-01 -7.32009888e-01 6.31278455e-01
6.24125719e-01 -6.67366087e-01 -6.35470271e-01 1.62504792e-01]
|
[9.842090606689453, -2.9600391387939453]
|
af7f9d6c-4180-4bd5-812f-3ff405a4c5b0
|
an-acceleration-method-based-on-deep-learning
|
2110.08679
| null |
https://arxiv.org/abs/2110.08679v1
|
https://arxiv.org/pdf/2110.08679v1.pdf
|
An Acceleration Method Based on Deep Learning and Multilinear Feature Space
|
Computer vision plays a crucial role in Advanced Assistance Systems. Most computer vision systems are based on Deep Convolutional Neural Networks (deep CNN) architectures. However, the high computational resource to run a CNN algorithm is demanding. Therefore, the methods to speed up computation have become a relevant research issue. Even though several works on architecture reduction found in the literature have not yet been achieved satisfactory results for embedded real-time system applications. This paper presents an alternative approach based on the Multilinear Feature Space (MFS) method resorting to transfer learning from large CNN architectures. The proposed method uses CNNs to generate feature maps, although it does not work as complexity reduction approach. After the training process, the generated features maps are used to create vector feature space. We use this new vector space to make projections of any new sample to classify them. Our method, named AMFC, uses the transfer learning from pre-trained CNN to reduce the classification time of new sample image, with minimal accuracy loss. Our method uses the VGG-16 model as the base CNN architecture for experiments; however, the method works with any similar CNN model. Using the well-known Vehicle Image Database and the German Traffic Sign Recognition Benchmark, we compared the classification time of the original VGG-16 model with the AMFC method, and our method is, on average, 17 times faster. The fast classification time reduces the computational and memory demands in embedded applications requiring a large CNN architecture.
|
['Michel Vinagreiro Edson Kitani Armando Lagana Leopoldo Yoshioka']
|
2021-10-16
| null | null | null | null |
['traffic-sign-recognition']
|
['computer-vision']
|
[ 9.63130221e-02 -2.35885158e-01 -1.00259520e-01 -3.19087982e-01
-1.30431220e-01 -1.71731338e-01 3.36548626e-01 -5.07803321e-01
-9.62267280e-01 5.61871529e-01 -4.39233303e-01 -6.69255674e-01
5.00119515e-02 -1.10543776e+00 -6.32098675e-01 -7.60813832e-01
3.11414182e-01 6.17367998e-02 3.76186073e-01 -3.04920197e-01
2.86912620e-01 7.86404788e-01 -1.75080371e+00 2.13471815e-01
7.47458577e-01 1.28896439e+00 2.51547515e-01 6.15350425e-01
-1.17312476e-01 5.88169634e-01 -5.09614170e-01 -2.31085822e-01
7.47245848e-01 -1.74209833e-01 -4.61224645e-01 5.01816254e-03
5.02031267e-01 -2.93239862e-01 -5.18899024e-01 1.07536423e+00
4.20858800e-01 1.30205914e-01 4.63569820e-01 -1.43060899e+00
-2.91757673e-01 1.07957557e-01 -1.86570436e-01 2.89674774e-02
-4.80058491e-01 1.12176187e-01 5.52794456e-01 -1.05016923e+00
4.54831421e-01 1.08436584e+00 8.85776579e-01 7.04518497e-01
-8.89978111e-01 -8.41916740e-01 -1.79217443e-01 7.36324072e-01
-1.32723618e+00 -1.74908951e-01 8.14044118e-01 -4.39461470e-01
1.07220578e+00 1.95224807e-01 9.52748060e-01 5.37031472e-01
2.37936333e-01 5.80818892e-01 9.00174081e-01 -5.70238233e-01
2.55893648e-01 2.73451835e-01 6.40815496e-01 9.60754216e-01
6.38170898e-01 1.17324412e-01 7.67824752e-03 2.72197388e-02
5.07306337e-01 2.05032066e-01 -1.59607172e-01 -3.12180579e-01
-9.74723220e-01 1.02434611e+00 7.88235188e-01 3.28829914e-01
-3.74174297e-01 1.83100626e-01 5.54925084e-01 4.88202810e-01
6.32308051e-02 7.88956359e-02 -4.82405365e-01 1.60305217e-01
-7.70332038e-01 2.34304070e-01 7.16702402e-01 8.37831199e-01
1.14123201e+00 3.34297776e-01 1.87848940e-01 7.22822249e-01
1.77215919e-01 5.51216662e-01 7.78273642e-01 -6.02648497e-01
6.50452375e-01 9.61333513e-01 -4.25802886e-01 -1.36066651e+00
-5.57046473e-01 -2.18772933e-01 -1.15611041e+00 6.42767966e-01
5.04988670e-01 -2.47432232e-01 -1.01678097e+00 1.22597992e+00
1.37873888e-01 4.46753241e-02 3.18109989e-01 8.67212415e-01
7.90218949e-01 7.12284207e-01 -2.97517598e-01 2.54170716e-01
1.24022126e+00 -1.16989255e+00 -4.27355438e-01 1.14068370e-02
1.10202229e+00 -6.27412379e-01 7.20955312e-01 4.45303321e-01
-3.87414604e-01 -8.59647632e-01 -1.37442780e+00 -6.86590839e-03
-7.10635781e-01 7.34986544e-01 5.61335564e-01 7.95079231e-01
-1.13533580e+00 6.50296807e-01 -7.48264432e-01 -3.95432621e-01
5.30245185e-01 6.52184963e-01 -5.25759995e-01 -1.17448784e-01
-8.80790412e-01 1.02058053e+00 7.41944909e-01 5.02718329e-01
-2.98953712e-01 -2.74914294e-01 -7.78688729e-01 4.64377813e-02
2.05729976e-01 -3.35336834e-01 8.93519461e-01 -1.13091373e+00
-1.64334726e+00 3.36089373e-01 -1.11365348e-01 -7.05849111e-01
3.95606458e-01 8.66873115e-02 -3.72134238e-01 -1.57086611e-01
-3.03183138e-01 7.67667115e-01 1.03376508e+00 -6.57753944e-01
-7.81863868e-01 -1.69154614e-01 -7.27786720e-02 -2.49323204e-01
-6.69666290e-01 -2.43713692e-01 -2.29246646e-01 -3.68952096e-01
2.46196732e-01 -1.28130651e+00 -3.46740782e-01 1.95173874e-01
-7.14254528e-02 -3.13322872e-01 1.34866798e+00 -4.83239084e-01
9.21958625e-01 -2.13091803e+00 -3.18344831e-01 4.44471121e-01
1.57739878e-01 1.05887771e+00 -3.00002128e-01 -5.16179204e-02
-1.23551443e-01 -1.57192186e-01 -9.46844071e-02 6.32633418e-02
-3.14371169e-01 2.46360227e-01 -5.29982857e-02 4.10365373e-01
3.49962682e-01 8.49246025e-01 -4.93928760e-01 -4.43156779e-01
4.01819050e-01 4.85310733e-01 -7.37656236e-01 -1.83095038e-02
2.75988907e-01 -1.37105882e-01 -2.96326816e-01 3.79710257e-01
9.88985956e-01 6.13740124e-02 1.02006726e-01 -6.84510350e-01
-2.57472068e-01 -2.89337307e-01 -1.08642197e+00 1.20151329e+00
-5.06686568e-01 1.15029716e+00 -3.03845704e-01 -1.53973305e+00
1.30493784e+00 1.21582523e-01 2.45575890e-01 -7.52165079e-01
3.80159408e-01 6.02892697e-01 4.62950557e-01 -4.45713282e-01
3.65486532e-01 1.68302909e-01 2.42841229e-01 8.40475112e-02
1.76975965e-01 2.10428149e-01 3.04669678e-01 -1.86646312e-01
8.47180068e-01 -1.12895615e-01 2.51935601e-01 -2.42251292e-01
9.16493595e-01 2.50682265e-01 5.88495612e-01 2.64742374e-01
-1.60426751e-01 1.95514411e-01 3.29826266e-01 -1.11516368e+00
-1.20614362e+00 -5.52106977e-01 -4.31461446e-02 4.89351243e-01
-1.28632203e-01 -2.06131324e-01 -7.35596716e-01 -7.31913209e-01
-7.12709874e-02 1.74363375e-01 -4.37325031e-01 -1.51400641e-01
-1.06927514e+00 -5.72504103e-01 6.73406601e-01 6.56963766e-01
9.66922283e-01 -1.18114710e+00 -9.33023393e-01 2.45491281e-01
1.45824179e-01 -1.22714353e+00 -1.25990033e-01 2.65239477e-02
-1.08660185e+00 -1.26816642e+00 -7.96958327e-01 -1.14938176e+00
9.40234482e-01 3.68386149e-01 4.71947581e-01 3.63444388e-01
-5.25255084e-01 -2.41732836e-01 -3.40573937e-01 -5.06184757e-01
-3.02873433e-01 1.27282307e-01 6.83453605e-02 2.84420431e-01
5.23032665e-01 -1.28401756e-01 -5.38865745e-01 3.08559895e-01
-7.16574073e-01 1.36097372e-01 8.84707093e-01 1.22766936e+00
2.07061559e-01 -8.78078490e-02 6.56935155e-01 -6.52740300e-01
4.25343990e-01 -2.63242740e-02 -1.01888919e+00 2.39599179e-02
-6.89522684e-01 2.00027749e-01 1.01748121e+00 -3.54860902e-01
-6.91865444e-01 4.42157209e-01 -2.20392019e-01 -4.87966180e-01
-3.89950275e-02 4.49490696e-01 -6.55557290e-02 -6.04508579e-01
6.84154630e-01 1.85822651e-01 4.60106760e-01 -3.69460583e-01
1.63913265e-01 9.13251519e-01 3.33825529e-01 7.90463313e-02
8.58947515e-01 3.20683390e-01 3.65889072e-01 -1.08931136e+00
-2.23385945e-01 -2.81139553e-01 -7.52771914e-01 -1.93090111e-01
7.31434464e-01 -7.18571663e-01 -8.22989464e-01 7.03805268e-01
-1.37052345e+00 -2.50351310e-01 1.10498197e-01 7.94560254e-01
-3.24667096e-01 3.70827287e-01 -3.36908489e-01 -4.24366117e-01
-5.54508984e-01 -1.31033611e+00 3.83069456e-01 7.10738152e-02
2.49052212e-01 -7.48448730e-01 -2.41312236e-01 2.65645469e-03
6.96446538e-01 2.17427909e-01 7.48784184e-01 -5.77656984e-01
-7.19132125e-01 -5.33360004e-01 -5.69574356e-01 8.40454638e-01
7.70763531e-02 -8.76253769e-02 -7.99123168e-01 -3.30005050e-01
-1.23677224e-01 -7.29029775e-02 1.14331222e+00 1.96433619e-01
1.27629197e+00 -3.97355884e-01 -3.21657985e-01 6.76735342e-01
1.60070157e+00 4.70354199e-01 8.35619807e-01 5.98026335e-01
8.12980056e-01 4.69705224e-01 5.86356819e-01 8.67007747e-02
1.32177725e-01 7.03241169e-01 2.19645739e-01 -2.22797558e-01
-1.37253433e-01 1.48822412e-01 2.57820845e-01 9.88444149e-01
-3.59461039e-01 3.48256677e-02 -9.13682938e-01 3.61114293e-01
-1.92406857e+00 -8.64463329e-01 -3.06389689e-01 2.01768970e+00
2.46589422e-01 -2.21912358e-02 -1.13048553e-01 5.73034286e-01
5.88445067e-01 -2.22666278e-01 -3.72307122e-01 -5.26187956e-01
5.17996512e-02 4.02220547e-01 7.67274797e-01 3.23285520e-01
-1.14923084e+00 9.16793227e-01 5.27955866e+00 6.84911549e-01
-1.73930943e+00 2.72131581e-02 3.05139840e-01 1.90916091e-01
4.50947881e-01 -2.67870814e-01 -9.69302773e-01 3.29503953e-01
1.06019354e+00 1.69800669e-02 2.43387535e-01 1.21189594e+00
-2.02656612e-02 1.38248071e-01 -9.53341842e-01 1.36678493e+00
2.11705133e-01 -1.35547626e+00 1.97957069e-01 1.06231757e-01
5.19752085e-01 3.57348584e-02 -1.33680657e-01 4.02496308e-01
-3.34245026e-01 -8.52928519e-01 5.16630828e-01 4.60450888e-01
8.39905024e-01 -9.47379887e-01 1.27010930e+00 2.95382619e-01
-1.31693232e+00 -2.66058713e-01 -8.07564676e-01 -1.64959103e-01
-1.03352919e-01 3.74356955e-01 -9.43969309e-01 3.12621534e-01
6.40137851e-01 6.01950049e-01 -6.92134500e-01 1.19178307e+00
1.39847234e-01 5.46512842e-01 -1.97006136e-01 -5.13610661e-01
4.21893358e-01 -1.92451596e-01 2.25072145e-01 1.16078973e+00
6.53979659e-01 -1.31220862e-01 -4.81682122e-02 4.92171347e-01
1.21643454e-01 2.41043970e-01 -7.86369264e-01 6.38625696e-02
3.90673988e-02 1.40455627e+00 -7.56015718e-01 -6.16246760e-01
-5.08107662e-01 7.74378181e-01 1.58282354e-01 2.05183476e-01
-7.39992857e-01 -9.89011705e-01 5.54133058e-01 -1.10913664e-01
4.87006068e-01 -2.85533398e-01 -3.25527936e-01 -1.13297784e+00
2.27885097e-01 -7.43319690e-01 -1.98309775e-02 -2.86517084e-01
-8.70088339e-01 9.55386817e-01 -1.96423307e-01 -1.75114357e+00
-3.78404677e-01 -1.26580203e+00 -5.55790901e-01 9.34608221e-01
-1.56060791e+00 -8.81375492e-01 -7.42440522e-01 7.86692619e-01
5.14493525e-01 -6.40152633e-01 7.85449266e-01 6.39544725e-01
-5.72871506e-01 7.98706472e-01 2.87454072e-02 4.86275375e-01
3.64016145e-01 -6.72791779e-01 5.00298738e-01 8.55245709e-01
-1.94630399e-01 3.13017726e-01 1.25494927e-01 -3.27422172e-01
-1.43107235e+00 -1.49621582e+00 9.01941419e-01 1.32392421e-01
3.27843159e-01 -3.23220462e-01 -8.41252148e-01 5.18848479e-01
-4.46049438e-04 4.27548140e-01 3.67503494e-01 -2.50489205e-01
-2.99544722e-01 -5.04737735e-01 -1.12359643e+00 5.35403013e-01
8.16875160e-01 -1.66143849e-01 -4.37277585e-01 -3.09752114e-02
4.78302717e-01 -1.40701145e-01 -3.31833243e-01 3.61987859e-01
8.29026818e-01 -7.17564821e-01 8.12287211e-01 -4.45447892e-01
2.25925878e-01 -5.03066361e-01 -1.64225802e-01 -1.29247117e+00
-3.22482646e-01 -1.18525755e-02 7.94162601e-02 6.58069491e-01
4.45767552e-01 -1.18555915e+00 1.03949165e+00 4.30758625e-01
-2.89071172e-01 -7.83186674e-01 -9.32413042e-01 -1.05368912e+00
-5.27760200e-02 -4.52485055e-01 5.79422653e-01 5.55651367e-01
-3.58018488e-01 1.35370880e-01 -1.90303877e-01 -8.26585814e-02
6.19268000e-01 -4.32594568e-02 1.06705534e+00 -1.48169482e+00
1.47991925e-01 -4.47287500e-01 -1.13137150e+00 -6.43160880e-01
8.89030620e-02 -9.58302557e-01 -7.37965689e-04 -1.31880474e+00
-1.79516301e-01 -5.39497614e-01 -2.11349204e-01 7.40618289e-01
2.54663557e-01 6.24822438e-01 6.22856259e-01 1.22827394e-02
-1.18214920e-01 4.72789884e-01 1.08894742e+00 -3.94834846e-01
-2.42085218e-01 5.35275154e-02 -2.57534444e-01 9.50286984e-01
1.00320923e+00 -3.79721284e-01 -2.49588117e-01 -2.83144563e-01
-1.15899876e-01 -4.00937229e-01 4.74928290e-01 -1.61763656e+00
4.81574774e-01 2.11561799e-01 4.58311319e-01 -8.57232630e-01
3.28538001e-01 -1.07401168e+00 1.74153652e-02 1.05529010e+00
1.55916989e-01 8.27857554e-02 3.24549913e-01 2.37533420e-01
-4.62610662e-01 -4.80912238e-01 8.68580937e-01 8.64262730e-02
-1.16155016e+00 3.46052349e-01 -6.20464742e-01 -6.45093083e-01
1.07686794e+00 -5.60917914e-01 -6.73857406e-02 4.02165242e-02
-4.77299601e-01 -1.45466506e-01 -1.03913657e-01 4.35561448e-01
9.16175306e-01 -1.67429864e+00 -4.61860836e-01 5.98932803e-01
7.91971851e-03 -1.37982532e-01 1.31935015e-01 6.02024198e-01
-9.39768612e-01 9.36221719e-01 -7.24246442e-01 -6.86276972e-01
-1.53842151e+00 4.94554490e-01 2.14174628e-01 -9.10510272e-02
-6.12118185e-01 4.03215677e-01 -1.08176187e-01 -5.72016716e-01
8.94076377e-02 -5.95134020e-01 -4.37724024e-01 -3.20537277e-02
6.29971802e-01 5.92501640e-01 4.51761395e-01 -6.68232977e-01
-3.20644647e-01 8.75890195e-01 4.50813510e-02 7.09635764e-02
1.34284282e+00 4.39606398e-01 -2.58003362e-02 -6.03114069e-02
1.79136264e+00 -5.60510278e-01 -7.99589038e-01 -2.40934327e-01
-3.34219709e-02 -4.28686947e-01 1.12890199e-01 -1.56159699e-01
-1.42607844e+00 1.13068330e+00 1.05603611e+00 -9.17826071e-02
1.13260186e+00 -5.77773571e-01 8.77841771e-01 1.11209512e+00
5.49569190e-01 -1.11291218e+00 -6.45791218e-02 8.97437513e-01
9.50250506e-01 -1.28993237e+00 -1.93168342e-01 -3.55974883e-01
-3.14484060e-01 1.67850816e+00 9.78758335e-01 -4.71129179e-01
1.00451159e+00 4.10572113e-03 1.98326007e-01 6.96039200e-02
-4.33110625e-01 -2.21248478e-01 3.25466424e-01 4.82414752e-01
9.07759070e-02 1.45483524e-01 -6.42284811e-01 3.64966333e-01
-3.97514343e-01 3.31194252e-01 5.16517282e-01 8.15568328e-01
-4.06191766e-01 -1.09295487e+00 -3.82368952e-01 6.45969272e-01
-8.61036684e-03 8.36774781e-02 -1.89522635e-02 8.01811516e-01
4.53710884e-01 6.89751863e-01 2.94384629e-01 -8.91032457e-01
4.64832515e-01 -6.23631068e-02 2.23397315e-01 -3.53603423e-01
-3.04251254e-01 -5.13708532e-01 -2.02999592e-01 -5.79960346e-01
-4.56794143e-01 -3.17084461e-01 -1.19676614e+00 -3.21851492e-01
-4.74352002e-01 -4.72594872e-02 1.09607077e+00 8.40336263e-01
2.89393902e-01 3.33702892e-01 6.43873572e-01 -1.12041414e+00
-4.83650595e-01 -1.03505814e+00 -2.52106756e-01 1.26219809e-01
2.42092565e-01 -6.23187602e-01 -1.37581289e-01 -6.28230581e-03]
|
[8.206982612609863, -0.6125100255012512]
|
eb7ac7df-5593-4200-aa0d-19d24321ff05
|
group-sparse-regularization-for-deep-neural
|
1607.00485
| null |
http://arxiv.org/abs/1607.00485v1
|
http://arxiv.org/pdf/1607.00485v1.pdf
|
Group Sparse Regularization for Deep Neural Networks
|
In this paper, we consider the joint task of simultaneously optimizing (i)
the weights of a deep neural network, (ii) the number of neurons for each
hidden layer, and (iii) the subset of active input features (i.e., feature
selection). While these problems are generally dealt with separately, we
present a simple regularized formulation allowing to solve all three of them in
parallel, using standard optimization routines. Specifically, we extend the
group Lasso penalty (originated in the linear regression literature) in order
to impose group-level sparsity on the network's connections, where each group
is defined as the set of outgoing weights from a unit. Depending on the
specific case, the weights can be related to an input variable, to a hidden
neuron, or to a bias unit, thus performing simultaneously all the
aforementioned tasks in order to obtain a compact network. We perform an
extensive experimental evaluation, by comparing with classical weight decay and
Lasso penalties. We show that a sparse version of the group Lasso penalty is
able to achieve competitive performances, while at the same time resulting in
extremely compact networks with a smaller number of input features. We evaluate
both on a toy dataset for handwritten digit recognition, and on multiple
realistic large-scale classification problems.
|
['Simone Scardapane', 'Amir Hussain', 'Danilo Comminiello', 'Aurelio Uncini']
|
2016-07-02
| null | null | null | null |
['handwritten-digit-recognition']
|
['computer-vision']
|
[ 3.31852198e-01 1.47593185e-01 -5.51523976e-02 -2.81495750e-01
-2.14464039e-01 -2.01464847e-01 3.51725847e-01 4.10737097e-01
-7.69940078e-01 6.92676008e-01 -2.74873376e-01 -1.46851480e-01
-3.27110976e-01 -6.84469044e-01 -7.10915506e-01 -1.01842713e+00
-1.45613536e-01 4.63776082e-01 -6.96065575e-02 2.40317687e-01
1.47040084e-01 8.40405822e-01 -1.35308421e+00 -3.34042385e-02
6.59778774e-01 1.35805643e+00 1.21152379e-01 1.82402790e-01
3.18623245e-01 7.10660934e-01 -4.04636711e-01 -2.75804456e-02
2.66192973e-01 -1.53456673e-01 -5.23153365e-01 4.62994456e-01
4.97971624e-01 4.38344553e-02 -2.92128146e-01 1.05063593e+00
3.41589093e-01 4.58098710e-01 5.97215593e-01 -1.01485407e+00
-3.41504782e-01 6.94704175e-01 -4.23016429e-01 4.61702310e-02
-3.86997104e-01 -1.47501990e-01 1.34593737e+00 -9.96581256e-01
5.13535559e-01 8.25017929e-01 4.38015610e-01 3.68954957e-01
-1.71711624e+00 -4.35303092e-01 4.43448365e-01 -2.21806131e-02
-1.31274879e+00 -5.13542533e-01 8.35749567e-01 -5.47242403e-01
9.18929160e-01 1.85782939e-01 5.50523579e-01 7.92205632e-01
-9.65459794e-02 7.01019585e-01 6.48376346e-01 -4.29427594e-01
5.22885680e-01 2.81963319e-01 6.67509556e-01 8.31757426e-01
2.13691473e-01 -2.26068899e-01 -3.48016053e-01 -2.73678124e-01
7.95246780e-01 7.75945485e-02 -4.55172718e-01 -4.83400613e-01
-1.12624896e+00 1.08402371e+00 6.47304833e-01 3.99571478e-01
-6.38792455e-01 1.45255089e-01 2.59008497e-01 3.26521873e-01
3.92074287e-01 4.15969491e-01 -3.60505521e-01 5.22498608e-01
-8.64030540e-01 1.28144771e-01 1.08131731e+00 5.16305506e-01
1.04031718e+00 3.78492147e-01 -1.60357192e-01 1.00012851e+00
1.34912983e-01 4.82968837e-02 3.99034500e-01 -7.69664884e-01
7.02536762e-01 5.93009114e-01 -1.64690360e-01 -1.16834784e+00
-5.80507278e-01 -7.99623728e-01 -1.30611169e+00 4.42619741e-01
6.12934113e-01 -2.60231405e-01 -8.63231838e-01 1.96083486e+00
-2.96121859e-03 3.95748839e-02 -1.40390009e-01 9.07375872e-01
3.64474177e-01 6.45205975e-01 -3.88595052e-02 -3.02350521e-01
1.05380821e+00 -1.01742494e+00 -4.04040575e-01 -5.11848271e-01
4.87428367e-01 -3.94877523e-01 7.26468444e-01 5.23743272e-01
-1.16475952e+00 -3.29139411e-01 -9.79737520e-01 4.54063937e-02
-2.09479555e-01 7.97504425e-01 5.31816900e-01 6.41189069e-02
-9.19972062e-01 8.80701542e-01 -8.56229722e-01 1.06602795e-01
3.08070868e-01 8.15099716e-01 -6.06377304e-01 1.04280464e-01
-8.61559451e-01 6.82081044e-01 4.34939504e-01 4.18215394e-01
-6.59940064e-01 -4.41713780e-01 -9.16847765e-01 5.15456438e-01
5.10820925e-01 -5.38944125e-01 5.76813936e-01 -1.25414002e+00
-1.24398005e+00 7.81654119e-01 -3.97854373e-02 -6.98517084e-01
5.95025897e-01 -9.59097221e-03 1.33882314e-01 -2.86331922e-02
-1.92845047e-01 5.82717419e-01 1.11192524e+00 -9.49654520e-01
-4.36131299e-01 -4.99034703e-01 3.20824310e-02 6.55960068e-02
-8.69771838e-01 -1.53053120e-01 -4.71801549e-01 -9.37198043e-01
2.09453195e-01 -1.01373601e+00 -4.85114902e-01 2.02697963e-01
-6.41958296e-01 -2.47932941e-01 5.10145187e-01 -5.48554599e-01
1.13437891e+00 -2.40276098e+00 8.57672632e-01 5.55472791e-01
4.51255113e-01 2.97012568e-01 -1.54448107e-01 6.73779175e-02
-4.13295358e-01 -1.37490600e-01 -6.59970880e-01 -8.31603646e-01
-1.20698310e-01 3.24541092e-01 -8.76875222e-02 7.56504416e-01
3.74919444e-01 5.04041731e-01 -4.77152616e-01 -1.11876152e-01
8.88220966e-02 6.26635313e-01 -6.59856081e-01 4.00237441e-02
1.51437476e-01 2.47342378e-01 -4.92182761e-01 2.63276577e-01
3.55173439e-01 -3.97054881e-01 1.12241395e-01 -1.86966598e-01
-1.30071983e-01 1.41882822e-01 -1.63895965e+00 1.22247875e+00
-5.80097973e-01 6.03545189e-01 4.53974426e-01 -1.51423240e+00
1.11981010e+00 2.35779077e-01 5.79235137e-01 -9.64173079e-02
1.96878999e-01 3.56449395e-01 1.01516463e-01 -1.02849938e-01
6.51920959e-02 3.31160910e-02 1.57373562e-01 4.50273424e-01
3.10266584e-01 6.00597680e-01 4.52591896e-01 -1.24250181e-01
9.98345852e-01 -5.60658634e-01 2.92132616e-01 -5.06669462e-01
6.97793782e-01 -4.22193289e-01 6.54557407e-01 6.08383715e-01
2.33705923e-01 5.26749671e-01 9.01865482e-01 -3.88095707e-01
-9.70166445e-01 -6.38929307e-01 -9.33901295e-02 8.89479816e-01
-3.34662735e-01 7.66617134e-02 -6.13776326e-01 -5.39786458e-01
2.16484159e-01 3.34900826e-01 -8.78995597e-01 -3.14618535e-02
-8.98240566e-01 -8.65075588e-01 1.85114697e-01 4.72219467e-01
2.44717970e-01 -1.12632537e+00 -5.70882380e-01 3.04486513e-01
3.68237615e-01 -9.37866092e-01 -3.70137811e-01 9.46223915e-01
-8.82420957e-01 -9.66851592e-01 -7.83841193e-01 -1.02782726e+00
9.92504537e-01 -1.90169901e-01 7.78130531e-01 1.39497727e-01
-2.59050697e-01 -1.32520795e-01 -2.93016024e-02 1.56414639e-02
1.06263636e-02 4.16484386e-01 1.39187783e-01 5.87300241e-01
-1.40408561e-01 -6.92776144e-01 -1.31893739e-01 2.10038409e-01
-1.13547981e+00 -9.75978896e-02 4.62123543e-01 1.08829439e+00
6.63140059e-01 -1.28256857e-01 4.72549260e-01 -1.03141415e+00
5.08720100e-01 -3.66978258e-01 -8.22063386e-01 1.48472860e-01
-4.09313172e-01 7.73279816e-02 9.35723066e-01 -7.35904276e-01
-4.34169829e-01 3.90095145e-01 -1.19653316e-02 -5.26861548e-01
1.13302700e-01 6.49116814e-01 -1.50752291e-01 -3.29062432e-01
4.51748997e-01 2.04383656e-01 -1.11431731e-02 -6.24353111e-01
-1.29507100e-02 2.86492974e-01 2.38234550e-01 -3.07367533e-01
6.60625517e-01 2.76000857e-01 1.74500048e-01 -8.04651022e-01
-6.61092460e-01 -2.82141685e-01 -8.02516222e-01 9.90595296e-02
5.30329823e-01 -6.23525560e-01 -5.80664396e-01 4.36245561e-01
-1.05146122e+00 -4.56414968e-01 -4.43780869e-01 6.31232977e-01
-4.20719743e-01 8.34340677e-02 -6.52097166e-01 -4.78858024e-01
-2.13511690e-01 -1.21822429e+00 5.41573822e-01 -6.11753166e-02
6.93762526e-02 -1.17851031e+00 -1.42738208e-01 -1.33665353e-01
2.09379092e-01 2.11219743e-01 9.98658776e-01 -9.19424832e-01
-2.84079045e-01 -4.69747841e-01 -2.36750543e-01 8.77236366e-01
-8.38845521e-02 -1.63611084e-01 -6.91749752e-01 -4.20420438e-01
2.95659363e-01 -2.82994062e-01 1.38289893e+00 4.03289557e-01
1.27006638e+00 -5.59189737e-01 -1.31107256e-01 7.50443518e-01
1.37889290e+00 -1.69438183e-01 2.94545531e-01 1.21845871e-01
7.86506057e-01 6.16894186e-01 1.71324015e-01 6.37583911e-01
-1.38110086e-01 7.48463154e-01 4.52752620e-01 -2.29907826e-01
1.15402997e-01 3.67090732e-01 1.30125329e-01 7.67880321e-01
-2.28084072e-01 -1.71663389e-01 -7.85287201e-01 4.24383193e-01
-2.05899262e+00 -4.79921103e-01 -1.45035200e-02 2.35639119e+00
6.50287688e-01 9.40096602e-02 6.10580528e-03 3.14722806e-01
7.77038455e-01 2.58460492e-01 -6.37738645e-01 -2.22741619e-01
-2.92484909e-01 1.58229053e-01 5.19478977e-01 6.39345288e-01
-1.23514295e+00 5.39677858e-01 5.93705893e+00 6.57772183e-01
-1.29891121e+00 1.47092296e-02 7.06702948e-01 -3.41002941e-01
-3.26917097e-02 -2.07214534e-01 -9.56337154e-01 2.08875760e-01
3.56140852e-01 1.50488257e-01 6.47547603e-01 7.38833070e-01
2.85854749e-02 1.48149714e-01 -1.28639901e+00 8.25030267e-01
5.98451458e-02 -1.13736808e+00 1.12136588e-01 2.48754531e-01
6.42235041e-01 6.77551851e-02 2.06544846e-02 -1.99663285e-02
-1.59141943e-01 -9.33147132e-01 7.07519889e-01 3.95284832e-01
4.34872270e-01 -7.77297258e-01 5.32457173e-01 4.36521888e-01
-9.51040983e-01 -2.81176031e-01 -4.95153815e-01 -9.32516083e-02
1.96183901e-02 1.06979799e+00 -1.44393399e-01 1.83095172e-01
3.78674686e-01 8.99848342e-01 -2.93739200e-01 1.04652846e+00
-2.09583923e-01 5.25833666e-01 -5.88888109e-01 3.10831759e-02
6.12264276e-01 -3.08402330e-01 5.76773226e-01 1.11337447e+00
1.30129501e-01 -7.64664561e-02 3.36347461e-01 9.71940339e-01
-3.97260159e-01 1.71887279e-01 -4.03540760e-01 1.04758769e-01
8.67475942e-02 1.40076995e+00 -6.68253005e-01 -3.33091289e-01
-2.93318421e-01 8.13177228e-01 8.26063335e-01 6.70135438e-01
-6.09808683e-01 -5.47099054e-01 6.53859735e-01 -8.37895367e-03
5.30000687e-01 -3.53057504e-01 -3.26075315e-01 -1.26969790e+00
3.69854689e-01 -5.45218050e-01 2.94906139e-01 1.16043035e-02
-1.10824263e+00 6.75752640e-01 -3.20537359e-01 -9.55904007e-01
-4.72467504e-02 -7.91320384e-01 -5.56638777e-01 1.05065346e+00
-1.59282541e+00 -6.09711528e-01 -5.15240170e-02 6.83388650e-01
2.80972540e-01 -2.02497035e-01 7.14154124e-01 5.84408283e-01
-1.15573680e+00 4.50542688e-01 1.79717124e-01 2.39292756e-01
2.85556704e-01 -1.06739616e+00 -8.78502950e-02 6.54335201e-01
1.39973179e-01 6.24350011e-01 4.52594966e-01 -3.04040045e-01
-1.05433762e+00 -1.10249257e+00 9.26694274e-01 4.13423955e-01
7.66949475e-01 -4.97029990e-01 -1.10036349e+00 8.51152480e-01
-2.08426327e-01 4.10507321e-01 5.02285480e-01 1.15514554e-01
-1.34393066e-01 -1.81441143e-01 -8.89795542e-01 2.22087696e-01
6.55823886e-01 -1.90523848e-01 -2.21453995e-01 3.69996697e-01
2.45116204e-01 -3.18878777e-02 -8.49915504e-01 3.54890108e-01
3.57057333e-01 -7.33970225e-01 9.41557050e-01 -6.70171142e-01
3.20699513e-01 -1.26109913e-01 -4.89531308e-02 -1.23462725e+00
-6.17588460e-01 -4.01266843e-01 -4.37856525e-01 1.10208559e+00
6.32992029e-01 -7.43454278e-01 7.80460596e-01 6.11600757e-01
-2.59698387e-02 -1.20490360e+00 -1.23568881e+00 -6.69796705e-01
-1.59370363e-01 -5.94684817e-02 -4.92553413e-02 8.93327296e-01
-2.31138989e-01 4.35355306e-01 -5.03561497e-01 6.61882758e-02
5.30632615e-01 3.22998092e-02 1.23554684e-01 -1.33858240e+00
-6.39148533e-01 -7.37028122e-01 -4.46114838e-01 -9.05274868e-01
2.76202440e-01 -9.44291651e-01 8.59452486e-02 -1.17890453e+00
-1.58225484e-02 -6.52339220e-01 -4.81186777e-01 8.06891322e-01
-2.04087235e-02 3.63451421e-01 3.18743467e-01 3.53407413e-01
-2.79782772e-01 5.02213657e-01 8.43143225e-01 -2.37028852e-01
-2.99037039e-01 2.36334473e-01 -3.32580835e-01 9.27594900e-01
6.33970022e-01 -5.55508494e-01 -6.93248212e-02 -7.64544427e-01
1.79576561e-01 -9.38692987e-02 3.78019154e-01 -1.02080572e+00
4.06265706e-01 2.99660843e-02 3.40612948e-01 4.28230055e-02
4.82138544e-01 -8.78759205e-01 -3.05227250e-01 6.13634884e-01
-5.77956200e-01 -3.08849007e-01 4.41703200e-02 3.22563678e-01
-4.10306156e-01 -7.04961717e-01 9.93216574e-01 1.11435493e-02
-3.27815086e-01 4.03190643e-01 -3.38051856e-01 -3.26724112e-01
9.92628515e-01 4.81716245e-02 8.17640871e-02 -4.20324467e-02
-1.23363757e+00 2.33466789e-01 1.30528405e-01 9.56471339e-02
5.76931715e-01 -1.29505336e+00 -7.17965841e-01 3.64227057e-01
-1.64066404e-01 1.09780237e-01 -1.40146449e-01 1.01501620e+00
-3.20713580e-01 3.82343352e-01 -1.89269572e-01 -4.18020129e-01
-1.22569144e+00 4.39238250e-01 4.34084177e-01 -6.01062715e-01
-5.89107335e-01 8.42045546e-01 2.07843110e-01 -2.13250041e-01
7.69580126e-01 -3.94454896e-01 -4.26278740e-01 4.22181726e-01
4.13995057e-01 3.66987288e-01 2.01596260e-01 -5.09139121e-01
-3.45128328e-01 5.00659645e-01 3.16128135e-02 1.20060906e-01
1.67360747e+00 2.29952082e-01 -3.97241682e-01 4.77279037e-01
1.54679990e+00 -2.04018056e-01 -1.41453159e+00 -5.74249744e-01
-1.04597792e-01 -5.86350299e-02 2.47531161e-01 -3.49048436e-01
-1.61790609e+00 8.79625440e-01 2.65100569e-01 3.34869444e-01
1.21999121e+00 -1.17563866e-01 4.27696407e-01 8.21066439e-01
2.10239962e-01 -1.02190590e+00 -1.38448581e-01 5.71101248e-01
1.01023734e+00 -9.71198678e-01 -3.27268764e-02 -3.13483626e-01
-3.82445186e-01 1.40133286e+00 2.99050301e-01 -5.84260941e-01
7.51286864e-01 2.55544037e-01 -2.61856228e-01 1.99102946e-02
-6.50617421e-01 -3.34166735e-02 4.38682646e-01 -9.55552459e-02
3.19217473e-01 -2.94481162e-02 -3.60307992e-01 4.30030793e-01
2.92388499e-01 -2.06198782e-01 1.96780249e-01 6.94058180e-01
-4.70702112e-01 -1.00421762e+00 -2.30923414e-01 7.47552931e-01
-3.86259198e-01 -4.69694771e-02 -3.18985671e-01 4.67713475e-01
6.97905570e-02 4.93374079e-01 2.09429920e-01 -5.04314490e-02
3.73363823e-01 -1.25805601e-01 2.97604412e-01 -8.59021723e-01
-7.31625736e-01 4.16006669e-02 -8.87904167e-02 -3.90920639e-01
-2.84848869e-01 -7.00468302e-01 -1.01022351e+00 1.06030907e-02
-4.33509916e-01 5.85578680e-02 6.87221885e-01 9.43444550e-01
-9.20733437e-03 6.18590832e-01 7.11143017e-01 -1.02931798e+00
-7.52966821e-01 -1.03831697e+00 -8.27439427e-01 3.04310173e-01
4.56589729e-01 -4.82271582e-01 -4.49886739e-01 -4.30138521e-02]
|
[8.520243644714355, 3.4809741973876953]
|
32bf93c7-d84f-4a17-9333-3b5661f8d3b0
|
clustering-label-inference-attack-against
|
2203.05222
| null |
https://arxiv.org/abs/2203.05222v1
|
https://arxiv.org/pdf/2203.05222v1.pdf
|
Clustering Label Inference Attack against Practical Split Learning
|
Split learning is deemed as a promising paradigm for privacy-preserving distributed learning, where the learning model can be cut into multiple portions to be trained at the participants collaboratively. The participants only exchange the intermediate learning results at the cut layer, including smashed data via forward-pass (i.e., features extracted from the raw data) and gradients during backward-propagation.Understanding the security performance of split learning is critical for various privacy-sensitive applications.With the emphasis on private labels, this paper proposes a passive clustering label inference attack for practical split learning. The adversary (either clients or servers) can accurately retrieve the private labels by collecting the exchanged gradients and smashed data.We mathematically analyse potential label leakages in split learning and propose the cosine and Euclidean similarity measurements for clustering attack. Experimental results validate that the proposed approach is scalable and robust under different settings (e.g., cut layer positions, epochs, and batch sizes) for practical split learning.The adversary can still achieve accurate predictions, even when differential privacy and gradient compression are adopted for label protections.
|
['Xinchen Lyu', 'Junlin Liu']
|
2022-03-10
| null | null | null | null |
['inference-attack']
|
['adversarial']
|
[ 2.71567941e-01 -6.69267103e-02 -3.25769633e-01 -7.95120239e-01
-1.23158908e+00 -1.42390454e+00 2.73727477e-01 4.20711070e-01
-6.74860716e-01 4.07670051e-01 -3.23225528e-01 -3.88660580e-01
-1.61594272e-01 -6.11293197e-01 -7.54364133e-01 -1.38416612e+00
-2.79947519e-01 2.90067941e-01 5.57535328e-02 5.24531424e-01
2.36570761e-01 7.62353778e-01 -1.00835955e+00 6.58465028e-01
3.64021063e-01 1.16322768e+00 -5.87918520e-01 5.71869671e-01
1.43223092e-01 7.05097556e-01 -4.96543497e-01 -1.11468697e+00
8.47928703e-01 -2.62881219e-01 -9.47692096e-01 -4.82694805e-01
3.84051532e-01 -5.78976274e-01 -2.24413589e-01 1.25829637e+00
5.32840669e-01 -1.71668887e-01 2.96185344e-01 -1.78826249e+00
-1.22906752e-01 8.93766224e-01 -3.83706212e-01 -3.79499257e-01
-4.86154296e-02 -4.21031239e-03 9.99636829e-01 -5.27157843e-01
5.17803252e-01 9.33795333e-01 7.80614734e-01 7.43254840e-01
-1.32354224e+00 -1.37314916e+00 9.45874900e-02 3.45906019e-01
-1.59443712e+00 -3.37395787e-01 6.54712141e-01 -9.44388807e-02
6.33991137e-02 1.02398860e+00 6.63633645e-02 1.07597888e+00
-1.18339332e-02 8.87435615e-01 1.43071651e+00 -1.63497820e-01
6.46525681e-01 9.22237158e-01 1.34761870e-01 4.16780859e-01
2.87036657e-01 1.44214153e-01 -8.23624313e-01 -9.64908838e-01
-1.09923095e-01 4.23431665e-01 -3.56149614e-01 -8.26131284e-01
-6.53265893e-01 8.09375942e-01 2.65939474e-01 -3.22861403e-01
2.33835056e-01 1.80884272e-01 7.30966687e-01 8.10907304e-01
5.66614568e-01 -2.33549312e-01 -8.08149457e-01 2.86447197e-01
-8.80026221e-01 2.27510482e-01 1.22649813e+00 1.13081217e+00
8.94577861e-01 -9.38431025e-01 -2.26821095e-01 3.16349775e-01
3.71899575e-01 3.88619035e-01 1.59649923e-01 -8.74151826e-01
6.43523693e-01 2.48539686e-01 7.56421164e-02 -9.73918319e-01
1.40124902e-01 -2.46201307e-01 -8.41905415e-01 2.32734904e-01
6.17219388e-01 -5.96382499e-01 -1.11210041e-01 1.81018651e+00
8.76035929e-01 2.74674565e-01 7.89660886e-02 8.19058001e-01
8.67461637e-02 3.79788905e-01 1.39474422e-01 -3.24822158e-01
1.14201260e+00 -8.54887128e-01 -5.38446128e-01 2.02225953e-01
1.22331834e+00 -3.61202121e-01 5.75938106e-01 3.59603673e-01
-8.87861133e-01 2.04839334e-01 -8.26769590e-01 -1.98639408e-01
-4.97652739e-01 -1.88186154e-01 4.89435077e-01 1.32487869e+00
-1.04962873e+00 6.45399511e-01 -9.21948314e-01 2.41241064e-02
8.98810685e-01 8.74060392e-01 -4.40654844e-01 -1.88991353e-01
-1.10404027e+00 5.49886711e-02 1.47645117e-03 -2.31803190e-02
-9.44619656e-01 -8.81760716e-01 -2.98720747e-01 1.66099072e-01
9.51018110e-02 -2.93696165e-01 1.24922216e+00 -6.54246688e-01
-1.19372010e+00 1.12334895e+00 1.26480192e-01 -6.56456530e-01
1.09570992e+00 1.04228541e-01 9.59998369e-02 1.14122033e-01
-5.00508189e-01 1.10485949e-01 8.19568276e-01 -1.41099954e+00
-1.13933849e+00 -1.01403606e+00 -2.13261515e-01 2.06623301e-01
-8.36592853e-01 1.09429277e-01 8.10810253e-02 -2.12440759e-01
1.62201047e-01 -1.22317851e+00 -1.63391814e-01 5.74802160e-01
-5.52369893e-01 -4.50807549e-02 1.42225742e+00 -5.51023662e-01
8.96189094e-01 -2.39486909e+00 -4.32251096e-01 7.81832457e-01
2.86090881e-01 2.69637436e-01 2.35333517e-01 6.40433967e-01
2.75222450e-01 2.39689589e-01 -2.65437812e-01 -9.88325536e-01
2.87372053e-01 -9.75395739e-02 -5.31054974e-01 1.21287584e+00
-6.77692711e-01 6.93374693e-01 -5.90983093e-01 -4.46368426e-01
-2.98229069e-01 4.61802036e-01 -4.82523948e-01 4.55143481e-01
-7.41348695e-03 4.10247564e-01 -5.16889989e-01 3.06730032e-01
1.37928247e+00 2.18725447e-02 5.71744144e-01 1.51616365e-01
3.16301733e-01 1.48799449e-01 -1.34900475e+00 1.54876280e+00
-3.16859514e-01 3.53287011e-01 8.22480500e-01 -6.84606254e-01
7.12514520e-01 4.90835577e-01 3.42098564e-01 -4.25598808e-02
-3.28691415e-02 1.84231311e-01 -7.47125626e-01 -1.98325008e-01
-1.44806772e-01 -6.81167394e-02 -1.36746541e-01 1.04581547e+00
-2.57679224e-01 4.65825081e-01 -9.25311208e-01 2.33844146e-01
1.07928681e+00 -4.51263279e-01 -2.99414694e-01 -1.06537372e-01
6.61024213e-01 -4.68043774e-01 6.19965732e-01 8.47528160e-01
-3.47014546e-01 2.13013619e-01 5.46293259e-01 -4.22878116e-01
-5.51243842e-01 -9.56603467e-01 -3.31749506e-02 1.58089209e+00
4.08242166e-01 -4.62600917e-01 -1.01331174e+00 -1.48668337e+00
3.55762303e-01 3.74151021e-01 -4.74165857e-01 -3.73850256e-01
-2.26190999e-01 -4.03610796e-01 1.12562370e+00 9.88971069e-03
6.19303226e-01 -5.66535413e-01 -3.68042201e-01 -2.89665669e-01
-1.19993873e-01 -6.79141760e-01 -8.19666207e-01 4.78163838e-01
-8.54618430e-01 -1.11469507e+00 -7.02119293e-03 -8.58433127e-01
8.29140484e-01 4.40409064e-01 2.51907021e-01 1.25607010e-02
-5.37741408e-02 4.19359177e-01 2.55148783e-02 -3.42777431e-01
-4.32504177e-01 2.18171701e-01 -2.46285439e-01 7.53984630e-01
6.27594531e-01 -6.59754097e-01 -9.04863536e-01 3.38029146e-01
-1.09173846e+00 -2.38854915e-01 2.08770201e-01 5.99477112e-01
3.89341742e-01 1.71747148e-01 1.80796236e-01 -1.69584811e+00
5.90961874e-01 -4.62548971e-01 -6.46971107e-01 6.27783537e-01
-9.88938272e-01 -5.34237884e-02 9.00647998e-01 -4.85052109e-01
-1.21956050e+00 2.71572649e-01 5.78986667e-02 -2.91888922e-01
-3.20565462e-01 -4.00490344e-01 -5.60654163e-01 -6.25644982e-01
4.62901920e-01 2.46558249e-01 4.47330214e-02 -6.76976740e-01
5.35042048e-01 1.17183602e+00 1.13513239e-01 -5.68244815e-01
9.56441581e-01 9.36391890e-01 1.51446655e-01 -1.15762353e-01
-5.24271488e-01 -6.41246438e-01 -5.81441879e-01 1.06372684e-01
3.62101436e-01 -7.78785169e-01 -1.34794760e+00 8.32188666e-01
-7.76908994e-01 -1.66261390e-01 -4.22095299e-01 1.99897423e-01
-3.34321588e-01 5.23665726e-01 -1.03738856e+00 -9.16704714e-01
-6.65610969e-01 -9.43400741e-01 7.24318802e-01 9.03173089e-02
2.06926271e-01 -1.14536679e+00 -8.06217715e-02 6.19158924e-01
2.57525563e-01 1.79577142e-01 8.48787606e-01 -1.50631499e+00
-7.33414888e-01 -5.30602932e-01 -2.21778397e-02 4.06691074e-01
-7.75737092e-02 -5.38292825e-01 -1.54509807e+00 -9.29591835e-01
5.30817568e-01 -6.07115746e-01 5.61733305e-01 -3.26509356e-01
1.65654516e+00 -1.16450417e+00 -4.19617802e-01 1.05616808e+00
1.43204653e+00 -1.81432992e-01 1.15066595e-01 -1.36545554e-01
6.85899198e-01 7.95084238e-01 3.91408294e-01 8.21703970e-01
1.56196043e-01 1.48469537e-01 5.64230800e-01 6.57206178e-02
5.07413626e-01 -5.91466069e-01 2.77027428e-01 4.13809657e-01
8.33404243e-01 -6.53751269e-02 -4.25881386e-01 2.00048923e-01
-1.67099667e+00 -8.33018601e-01 -8.73597246e-03 2.61764956e+00
1.26100302e+00 -2.60475725e-01 -1.33058414e-01 1.82001457e-01
7.19324827e-01 1.43885270e-01 -8.69652390e-01 -5.43987215e-01
1.52450114e-01 1.30247086e-01 1.22031796e+00 4.06478941e-01
-1.12541521e+00 6.67074561e-01 4.76543427e+00 9.98553157e-01
-1.02308917e+00 4.89506632e-01 9.97778714e-01 -3.71608585e-01
-3.06161195e-01 1.42965958e-01 -8.83792281e-01 5.47770798e-01
1.08871901e+00 -1.32724836e-01 5.99734187e-01 1.01381326e+00
-3.72824252e-01 3.13169539e-01 -1.44167960e+00 9.31151867e-01
-2.03585207e-01 -1.05267262e+00 -4.32546616e-01 3.23839635e-01
6.76507235e-01 -1.70828328e-01 4.48194891e-01 -1.97621565e-02
6.72644496e-01 -6.96221888e-01 4.71889466e-01 6.94075525e-02
9.52274323e-01 -1.24791467e+00 5.71377814e-01 8.20530713e-01
-7.16485143e-01 -4.91901129e-01 -4.04594064e-01 4.00154173e-01
-5.08011580e-01 3.22021753e-01 -8.86299670e-01 2.72953093e-01
8.28778923e-01 1.13269322e-01 -5.68089128e-01 8.00224245e-01
-1.46237865e-01 1.02049112e+00 -6.97832286e-01 3.18442471e-03
1.02803424e-01 -3.16433549e-01 1.72806546e-01 1.35488486e+00
-1.03153236e-01 -6.29902408e-02 -2.28607971e-02 5.09540617e-01
-7.41086364e-01 3.55368733e-01 -4.27057892e-01 5.33236325e-01
8.91273201e-01 1.32449806e+00 -3.20285052e-01 6.01886176e-02
-1.42849758e-01 1.37622535e+00 4.97125626e-01 5.07163346e-01
-5.21243155e-01 -4.06804949e-01 1.07805979e+00 -1.14063680e-01
2.17169002e-01 1.91078410e-01 -4.66205269e-01 -8.36008728e-01
7.07895607e-02 -8.58960450e-01 9.74305153e-01 1.04733959e-01
-1.40905821e+00 -8.43115002e-02 -2.87530988e-01 -7.88745403e-01
2.24611297e-01 2.29294021e-02 -5.33122957e-01 7.71729887e-01
-1.33893967e+00 -1.33187127e+00 2.81899087e-02 1.06753743e+00
-1.63584813e-01 1.57787696e-01 1.04428756e+00 1.24913976e-01
-4.44404811e-01 1.60368466e+00 1.02926564e+00 3.01736414e-01
9.41297293e-01 -1.05918908e+00 1.35621294e-01 6.38609231e-01
3.85520816e-01 5.33248365e-01 2.74529785e-01 -2.82470733e-01
-1.51699054e+00 -1.40010595e+00 8.57750893e-01 -1.87014744e-01
1.66681439e-01 -1.20062089e+00 -8.43301833e-01 9.13530886e-01
-1.40635654e-01 3.56546491e-01 1.40931153e+00 -1.57031089e-01
-9.25132692e-01 -6.55383289e-01 -2.10838366e+00 2.72404462e-01
7.50472248e-01 -1.04737329e+00 3.36496025e-01 5.61052859e-01
8.47490430e-01 -2.49832168e-01 -7.81806648e-01 -1.28316134e-01
5.93191326e-01 -1.00327289e+00 7.25934863e-01 -8.63379538e-01
-5.39782584e-01 -1.30507629e-02 -5.47034368e-02 -7.37696230e-01
3.54836620e-02 -1.07936871e+00 -3.52523744e-01 1.60001552e+00
3.97171676e-01 -8.70247483e-01 1.45129788e+00 1.28210068e+00
9.29308355e-01 -6.08571708e-01 -1.25349140e+00 -3.97758305e-01
2.82978296e-01 -2.15694949e-01 8.92249346e-01 1.18228781e+00
3.36436369e-02 -3.94430012e-01 -4.19666350e-01 5.87273657e-01
1.24900031e+00 1.91954643e-01 6.81652009e-01 -1.03650606e+00
1.04249949e-02 2.59349793e-01 -1.01469047e-01 -9.17409658e-01
4.27730262e-01 -1.22381949e+00 -2.54785568e-01 -6.47960424e-01
1.37720659e-01 -1.02018070e+00 -5.96743941e-01 5.82294405e-01
1.53300181e-01 4.51870039e-02 2.70376205e-01 4.59066480e-01
-6.21439457e-01 2.26069897e-01 6.28665030e-01 -8.48367959e-02
-7.20513687e-02 7.98225641e-01 -7.67952442e-01 2.69904107e-01
1.04776382e+00 -1.12921989e+00 -6.54071689e-01 -7.89418537e-03
8.22400749e-02 -1.18612744e-01 3.75014037e-01 -7.44700909e-01
1.01130140e+00 -2.17509568e-01 1.93247646e-01 -3.58640701e-01
-6.80477917e-02 -1.69246972e+00 2.80446053e-01 5.37816584e-01
-1.36682451e+00 -4.60822463e-01 -4.16583866e-01 9.54595447e-01
3.43487799e-01 -2.20943347e-01 8.70362699e-01 1.15861617e-01
1.11338288e-01 6.86251283e-01 9.35245771e-03 7.96642601e-02
1.55705583e+00 -4.06448841e-02 -1.52754188e-01 -3.69899601e-01
-6.16666615e-01 5.73068678e-01 6.84071004e-01 -8.84425268e-02
4.50375408e-01 -1.01652896e+00 -5.25508165e-01 5.97125590e-01
-1.26017585e-01 2.33518556e-01 2.81210005e-01 4.21710998e-01
-3.02300930e-01 1.41220957e-01 2.61182576e-01 -1.71647191e-01
-1.81433439e+00 8.55961084e-01 3.34064990e-01 -2.64087826e-01
-5.29637337e-01 1.16755414e+00 -1.25248861e-02 -6.96805477e-01
1.07024586e+00 1.09867930e-01 6.28263354e-01 8.06990638e-02
7.59543657e-01 5.98493040e-01 1.88083634e-01 -1.09855622e-01
-2.01849282e-01 4.98136021e-02 -3.27597797e-01 -9.19548422e-02
1.07118225e+00 -5.19607663e-01 -2.70516217e-01 1.41072795e-01
1.98981965e+00 2.69191563e-01 -1.37953901e+00 -6.51488543e-01
2.49144193e-02 -7.55342662e-01 -2.05991149e-01 -6.45229757e-01
-1.39085495e+00 9.80599582e-01 8.43533754e-01 9.78097692e-02
1.03104734e+00 -1.58471927e-01 1.13272262e+00 3.88193816e-01
6.74210072e-01 -1.07560515e+00 -6.28627121e-01 -3.22179079e-01
1.08050115e-01 -8.57450306e-01 -1.10327728e-01 -4.62067336e-01
-5.46446204e-01 7.13844061e-01 3.38152289e-01 2.90675789e-01
1.08988035e+00 3.72477204e-01 3.25372368e-01 1.88964948e-01
-9.56337214e-01 8.84324431e-01 -3.91432256e-01 7.18917966e-01
-3.45537633e-01 1.13218017e-01 -3.31129543e-02 6.92778587e-01
8.69047120e-02 -4.98657793e-01 1.42881691e-01 1.09377003e+00
-8.87238160e-02 -1.45422471e+00 -3.12538803e-01 3.14790845e-01
-9.78973389e-01 2.83410609e-01 -6.66045845e-01 9.60413292e-02
1.71628311e-01 8.85334373e-01 -2.63728648e-01 -4.14855510e-01
-4.64499965e-02 4.42237079e-01 -2.46340603e-01 -2.42937997e-01
-1.29697537e+00 -5.47731817e-01 -3.66342247e-01 -7.39425004e-01
8.13879594e-02 -6.11533821e-01 -1.20778668e+00 -7.29279101e-01
-5.21673381e-01 6.33125901e-01 1.00068510e+00 3.98750156e-01
5.27853847e-01 -6.17233217e-01 1.65075111e+00 2.89365696e-03
-1.39341533e+00 -3.35017979e-01 -1.06665754e+00 7.04867840e-01
4.69397902e-01 4.03250694e-01 -9.60301101e-01 1.61291286e-02]
|
[5.812312602996826, 6.752541542053223]
|
5ee42830-8d90-48b5-93db-48a182d2fb09
|
compositional-learning-of-image-text-query
|
2006.11149
| null |
https://arxiv.org/abs/2006.11149v3
|
https://arxiv.org/pdf/2006.11149v3.pdf
|
Compositional Learning of Image-Text Query for Image Retrieval
|
In this paper, we investigate the problem of retrieving images from a database based on a multi-modal (image-text) query. Specifically, the query text prompts some modification in the query image and the task is to retrieve images with the desired modifications. For instance, a user of an E-Commerce platform is interested in buying a dress, which should look similar to her friend's dress, but the dress should be of white color with a ribbon sash. In this case, we would like the algorithm to retrieve some dresses with desired modifications in the query dress. We propose an autoencoder based model, ComposeAE, to learn the composition of image and text query for retrieving images. We adopt a deep metric learning approach and learn a metric that pushes composition of source image and text query closer to the target images. We also propose a rotational symmetry constraint on the optimization problem. Our approach is able to outperform the state-of-the-art method TIRG \cite{TIRG} on three benchmark datasets, namely: MIT-States, Fashion200k and Fashion IQ. In order to ensure fair comparison, we introduce strong baselines by enhancing TIRG method. To ensure reproducibility of the results, we publish our code here: \url{https://github.com/ecom-research/ComposeAE}.
|
['Martin Kleinsteuber', 'Egor Labintcev', 'Muhammad Umer Anwaar']
|
2020-06-19
| null | null | null | null |
['multi-modal']
|
['miscellaneous']
|
[ 1.40926301e-01 -3.57705206e-01 9.27696079e-02 -6.12554967e-01
-7.56472707e-01 -6.57577515e-01 5.88903546e-01 3.54592353e-02
-5.03900051e-01 1.50789067e-01 1.19101271e-01 1.21811561e-01
-1.83788180e-01 -7.99266577e-01 -1.06754398e+00 -5.67199469e-01
5.00472009e-01 3.70568633e-01 -1.24368899e-01 -3.48193616e-01
1.62161544e-01 5.07202804e-01 -1.55829096e+00 4.98945981e-01
2.95509756e-01 1.16831875e+00 3.43243837e-01 4.91408974e-01
2.84651726e-01 4.08686996e-01 -3.20462286e-01 -1.00309289e+00
6.10095024e-01 -2.11194038e-01 -9.25138712e-01 4.06064868e-01
7.93250203e-01 -5.54436326e-01 -7.49129713e-01 1.08816314e+00
6.23399913e-01 3.63714188e-01 7.26969957e-01 -1.15011990e+00
-1.17471266e+00 3.74875396e-01 -5.83467066e-01 -4.80029620e-02
2.77039766e-01 8.43541622e-02 1.02119303e+00 -8.78906369e-01
8.44955206e-01 1.06389320e+00 2.32295856e-01 4.43683058e-01
-1.10993338e+00 -2.67078549e-01 7.93071315e-02 4.97550249e-01
-1.57185268e+00 -5.35997987e-01 9.55684543e-01 -2.09228352e-01
4.27301526e-01 5.24733007e-01 3.50548595e-01 1.06660557e+00
1.35895029e-01 9.26503778e-01 8.89767766e-01 -3.78539860e-01
6.08337373e-02 2.57968545e-01 -2.76935577e-01 7.73927212e-01
-1.79596409e-01 -5.89674115e-02 -5.44203520e-01 1.42244264e-01
4.64275450e-01 3.34983826e-01 -1.31942973e-01 -5.65123081e-01
-1.36997437e+00 8.07980239e-01 6.13046288e-01 3.54053319e-01
-5.51326275e-01 1.55555144e-01 1.25189811e-01 4.39908147e-01
2.76285410e-01 2.49578908e-01 -3.26573491e-01 2.07118332e-01
-9.18179810e-01 3.65918905e-01 5.88284612e-01 1.03261709e+00
6.95457876e-01 -4.91529167e-01 -1.58223659e-01 9.66643214e-01
3.32841843e-01 4.64504868e-01 3.64297003e-01 -1.03377116e+00
3.15016776e-01 4.31888074e-01 2.20435202e-01 -1.16818476e+00
-1.71038046e-01 -1.58015996e-01 -8.64574730e-01 -1.66088223e-01
2.31466725e-01 2.60897636e-01 -8.66075099e-01 1.42675149e+00
4.19518113e-01 -2.46008724e-01 -1.01197839e-01 1.30786359e+00
9.05020893e-01 7.78413534e-01 -2.27322385e-01 1.80407390e-01
1.44639492e+00 -1.08863425e+00 -4.27802444e-01 -9.13263112e-02
2.29727164e-01 -1.16513479e+00 1.15741611e+00 2.67765671e-01
-1.08852839e+00 -7.15225756e-01 -9.35570776e-01 -3.09248090e-01
-5.78549206e-01 4.94768649e-01 2.68891364e-01 3.08904588e-01
-9.99009907e-01 6.37624800e-01 -5.32781065e-01 -7.74194658e-01
6.57748990e-03 3.16540897e-01 -5.07483184e-01 -5.49692035e-01
-1.05782652e+00 7.19885290e-01 3.54595035e-01 2.65539587e-01
-7.67057657e-01 -4.06210840e-01 -7.60429978e-01 -9.35439765e-02
4.07423019e-01 -7.84918010e-01 1.09442449e+00 -1.15499985e+00
-1.36642790e+00 1.24506176e+00 1.48208454e-01 -2.26040468e-01
6.35370553e-01 -1.43320531e-01 -5.66050529e-01 3.11974168e-01
8.42786729e-02 9.74191487e-01 1.22025156e+00 -1.31689405e+00
-5.14808476e-01 -5.30330122e-01 3.52227420e-01 2.24971473e-01
-3.78545046e-01 3.54602970e-02 -1.11019719e+00 -8.54840219e-01
-2.81147175e-02 -1.35280609e+00 9.06182602e-02 9.27648172e-02
-3.94378841e-01 -9.73033085e-02 7.30431736e-01 -8.23789716e-01
1.02513599e+00 -2.42494607e+00 3.26572061e-01 2.34219283e-01
-3.28795463e-02 -4.02979851e-02 -5.00692666e-01 5.12305200e-01
1.15496710e-01 -9.99904796e-02 -7.35799521e-02 -3.65015864e-01
3.35574478e-01 1.54969096e-01 -4.58060298e-03 4.84631211e-01
-5.68746813e-02 9.04341221e-01 -5.94555855e-01 -4.12721694e-01
2.90646791e-01 4.72729236e-01 -4.10856664e-01 1.66114107e-01
-2.16087461e-01 7.72086531e-02 -2.79412419e-01 7.37415612e-01
7.74580956e-01 -3.94872963e-01 1.46920472e-01 -9.01832223e-01
1.45164132e-01 -2.48509333e-01 -1.26809204e+00 2.01649809e+00
-4.35179561e-01 3.85010034e-01 1.35064069e-02 -9.05664563e-01
6.66741669e-01 -1.19071826e-01 6.28724873e-01 -9.75580692e-01
1.88866451e-01 -1.57481304e-03 -4.32068974e-01 -4.97149348e-01
8.18293154e-01 2.37363905e-01 -1.83255538e-01 5.32285452e-01
5.18270815e-03 -1.35166541e-01 3.90183151e-01 3.20831418e-01
7.52603769e-01 7.52303824e-02 -1.58750460e-01 -5.89503273e-02
3.59406859e-01 -2.07279563e-01 7.23247528e-02 7.05421627e-01
2.30830479e-02 7.59173036e-01 2.97192503e-02 -5.68806291e-01
-1.20420039e+00 -1.16007710e+00 1.75183877e-01 1.31543863e+00
4.83056873e-01 -3.60692084e-01 -5.46984494e-01 -6.76229298e-01
6.31158575e-02 5.03389001e-01 -5.62168360e-01 -2.07085788e-01
-4.85185683e-01 -4.75918561e-01 7.76897296e-02 1.72949091e-01
6.51872575e-01 -9.77070332e-01 -3.72328013e-01 4.40240046e-03
-4.62114394e-01 -1.09444213e+00 -1.07484889e+00 -1.79478154e-01
-4.40105289e-01 -1.02217686e+00 -9.75550830e-01 -1.06246555e+00
6.88699901e-01 3.88954490e-01 1.05799615e+00 6.61431178e-02
-3.63278449e-01 7.35119581e-01 -5.01060963e-01 -2.11602561e-02
-3.69753182e-01 1.65522709e-01 -1.60914045e-02 4.31159437e-01
2.97370195e-01 -2.57832021e-01 -1.01005125e+00 4.91754055e-01
-1.49133408e+00 6.99555203e-02 7.01725185e-01 6.23202443e-01
8.63649309e-01 2.29685143e-01 5.25201745e-02 -4.89382297e-01
5.38067877e-01 -3.30535680e-01 -3.92546952e-01 4.57685530e-01
-4.19179559e-01 1.11042932e-01 4.72211659e-01 -4.54936504e-01
-6.40290976e-01 3.00581485e-01 -4.35146019e-02 -7.02005565e-01
-1.01868831e-01 3.93249184e-01 -2.12222680e-01 6.41199127e-02
4.57067430e-01 3.60481441e-01 -8.52110758e-02 -6.27472281e-01
5.71623325e-01 7.68010974e-01 4.95599806e-01 -4.75118846e-01
7.35410631e-01 5.95176995e-01 -2.06215054e-01 -5.65432608e-01
-6.53350592e-01 -5.33173263e-01 -6.39393270e-01 -3.08513463e-01
7.51055062e-01 -8.45166743e-01 -6.10273302e-01 4.70219880e-01
-9.25018072e-01 -2.19835624e-01 -9.80034843e-02 3.50192815e-01
-6.32841110e-01 3.71944934e-01 -6.05791032e-01 -1.32868066e-01
-3.05128783e-01 -1.15695190e+00 1.28957927e+00 1.74014792e-01
6.49033040e-02 -5.74652910e-01 -2.43530989e-01 6.83454514e-01
3.87211472e-01 1.29900739e-01 8.39752197e-01 -4.46099579e-01
-7.36707747e-01 -1.91039830e-01 -2.75945455e-01 3.73265207e-01
2.14610025e-01 -1.08687915e-01 -5.55917561e-01 -4.84333903e-01
-3.48162860e-01 -2.04919040e-01 8.60779881e-01 1.62313193e-01
1.35564530e+00 -4.30874735e-01 -9.06405896e-02 4.86777395e-01
1.54690063e+00 5.12388200e-02 6.32451415e-01 5.16171753e-01
5.93685031e-01 4.69393402e-01 6.65660083e-01 4.89812940e-01
6.69216573e-01 9.11816299e-01 3.65333349e-01 5.82969456e-04
-9.19611230e-02 -2.18396112e-01 2.71942586e-01 5.63753486e-01
1.74692541e-01 -3.73034209e-01 -5.23492694e-01 5.95613539e-01
-1.81473565e+00 -8.05193067e-01 2.37292722e-01 2.07666898e+00
6.87654436e-01 -2.33631983e-01 1.55879691e-01 -2.01497078e-01
6.42948985e-01 2.64389813e-01 -6.84796989e-01 -2.63131827e-01
1.38164153e-02 -3.27687450e-02 4.86373782e-01 2.11794674e-01
-1.27331567e+00 7.68585861e-01 4.96726131e+00 8.24992716e-01
-1.17979288e+00 9.04810950e-02 6.68288767e-01 -3.44181031e-01
-3.74934554e-01 -3.02479386e-01 -6.09219909e-01 5.29743433e-01
6.34134591e-01 6.27979934e-02 1.00287461e+00 5.85077763e-01
1.87678635e-02 -1.10704370e-01 -1.23334086e+00 1.29573739e+00
5.10281563e-01 -1.22888339e+00 2.29041621e-01 -1.13326177e-01
5.91302693e-01 -3.16565931e-02 4.18912321e-01 1.85080171e-01
1.01538189e-01 -6.68173730e-01 9.34877038e-01 8.21021020e-01
7.61599183e-01 -5.77781379e-01 4.82020855e-01 -1.97007246e-02
-1.00041616e+00 1.79213583e-02 -2.65059561e-01 5.69160938e-01
2.73005739e-02 3.93112034e-01 -4.71080095e-01 7.20130324e-01
1.03227592e+00 5.78072548e-01 -9.05935287e-01 8.14349055e-01
2.03377172e-01 6.65772632e-02 -4.21719402e-01 1.64248832e-02
1.18792802e-01 -3.29721510e-01 4.17726129e-01 1.02713227e+00
4.33082461e-01 -3.19081284e-02 1.09428100e-01 7.70513952e-01
-5.63077390e-01 2.80916244e-01 -4.61047322e-01 -1.06413566e-01
1.13463357e-01 1.33814538e+00 -5.03466010e-01 -2.24063918e-01
-4.78680283e-01 1.68069065e+00 -2.69269403e-02 4.38745558e-01
-9.28101003e-01 -3.55494916e-01 4.21242863e-01 1.83155194e-01
6.21115267e-01 -6.30098358e-02 4.57556754e-01 -1.09617162e+00
4.39992696e-01 -1.12922192e+00 5.05921721e-01 -1.16016221e+00
-1.32133687e+00 5.19382119e-01 -4.26992178e-02 -1.07499266e+00
1.08258031e-01 -3.86203021e-01 -9.22522545e-02 4.71406907e-01
-1.29726708e+00 -1.49744642e+00 -3.05770457e-01 8.16716611e-01
5.76964974e-01 -1.39812604e-01 4.75590646e-01 6.70380890e-01
-3.07076663e-01 5.80453396e-01 4.25938487e-01 1.69993088e-01
9.81963217e-01 -9.56982851e-01 1.52880698e-01 5.88507593e-01
4.19548631e-01 4.78388935e-01 6.14152431e-01 -2.93298066e-01
-1.83687198e+00 -1.31001449e+00 7.12454975e-01 -3.00317496e-01
5.24845719e-01 -2.35782370e-01 -4.17868197e-01 7.35357344e-01
4.24721032e-01 6.97863195e-03 4.78624344e-01 -2.13529512e-01
-2.38456309e-01 -4.50971425e-01 -1.06959176e+00 5.65574467e-01
8.40962052e-01 -5.94587684e-01 -3.43195796e-01 4.88058418e-01
4.87181515e-01 -4.28355217e-01 -9.62053239e-01 2.27476269e-01
8.08880568e-01 -9.01723325e-01 1.04088306e+00 -3.90881479e-01
3.64118099e-01 -4.52260375e-01 -5.92165291e-01 -1.32167792e+00
-3.72335017e-01 -2.75470912e-01 -3.46268117e-02 1.14504707e+00
1.38353691e-01 -1.61612883e-01 6.37611091e-01 6.17852092e-01
2.63266861e-01 -7.03796089e-01 -7.36440182e-01 -4.99289423e-01
-2.28255272e-01 -9.38174874e-02 7.98221052e-01 7.34065652e-01
-6.50990009e-01 1.63488477e-01 -6.06428385e-01 1.95918009e-01
5.44409096e-01 4.87366319e-01 8.35794151e-01 -7.99929857e-01
-3.04868519e-01 -2.31231064e-01 -2.32056707e-01 -1.06285584e+00
1.05567195e-01 -9.41387713e-01 -1.65027648e-01 -1.59625316e+00
3.96150827e-01 -1.27356738e-01 -4.37560320e-01 3.17975998e-01
7.43834972e-02 5.65152526e-01 6.81555510e-01 1.37765348e-01
-1.01455498e+00 5.34591258e-01 1.13665950e+00 -7.30028331e-01
-8.48907232e-02 -1.39472075e-02 -7.12422550e-01 3.88782352e-01
8.26584637e-01 -3.98059040e-01 -9.49866772e-02 -8.03982258e-01
4.07010108e-01 -7.45142326e-02 6.35137022e-01 -7.83456087e-01
2.16824159e-01 -3.69143486e-02 5.57263494e-01 -6.89895749e-01
5.23764431e-01 -1.12226927e+00 3.87349397e-01 1.02731921e-01
-4.21910554e-01 3.05910826e-01 -5.55479899e-02 4.54199046e-01
-1.71178535e-01 -2.39121825e-01 6.92416549e-01 -1.17802903e-01
-8.38849187e-01 5.85623980e-01 -1.76836655e-01 -3.52328628e-01
8.52837443e-01 5.72402664e-02 -1.19534761e-01 -5.03686070e-01
-7.36388087e-01 1.76056266e-01 7.03182757e-01 8.49060357e-01
7.47622073e-01 -1.68434417e+00 -6.89968705e-01 5.19272275e-02
4.01425540e-01 -3.77654314e-01 4.12064195e-01 7.88940012e-01
-5.06941497e-01 2.26213023e-01 -1.18101381e-01 -3.62970918e-01
-1.35778046e+00 9.13031280e-01 2.81736255e-01 -1.00316353e-01
-3.72097433e-01 5.11830211e-01 9.84609947e-02 -6.18418932e-01
2.13767350e-01 -1.61264211e-01 6.26974329e-02 3.46197449e-02
4.58975703e-01 1.55444965e-01 2.59537131e-01 -7.79626191e-01
-2.63498276e-01 5.16530395e-01 -2.90011346e-01 -2.93250605e-02
1.38003421e+00 -4.40996617e-01 -5.61919548e-02 9.65637639e-02
1.56332886e+00 -5.47772422e-02 -8.51455986e-01 -4.89794940e-01
-2.43991137e-01 -6.29721582e-01 1.02855682e-01 -9.23703432e-01
-1.41515398e+00 4.30551797e-01 1.11924398e+00 8.08040947e-02
1.41705215e+00 2.51825243e-01 9.95783508e-01 6.30100429e-01
2.45362863e-01 -1.29016185e+00 2.47377515e-01 1.69936255e-01
1.20221651e+00 -1.39540291e+00 6.58775717e-02 1.23000756e-01
-7.39837050e-01 1.05358005e+00 3.28264743e-01 -4.33056094e-02
6.26026571e-01 -3.33950788e-01 9.09201503e-02 -3.85379612e-01
-4.18016821e-01 -2.05870971e-01 5.30341089e-01 1.71571523e-01
8.00665691e-02 1.03284158e-01 -8.86908472e-02 1.51969418e-01
-1.48743555e-01 7.97916874e-02 1.31218061e-01 7.98407674e-01
-1.84943178e-03 -1.15361643e+00 -4.47632641e-01 4.49429423e-01
-4.55262274e-01 -5.00426888e-02 -4.03379232e-01 6.33739352e-01
3.51243652e-02 9.43934381e-01 2.49842182e-02 -3.56864274e-01
5.81938922e-01 -1.10277385e-01 6.44254208e-01 -1.18512161e-01
-3.84844393e-01 1.78738773e-01 -1.28111318e-01 -6.00930870e-01
-6.30527437e-01 -5.28590024e-01 -7.32800782e-01 -6.03635490e-01
-5.39454520e-02 -2.64944788e-02 8.37814629e-01 6.29147530e-01
5.39932609e-01 1.56527475e-01 7.74264455e-01 -6.79383516e-01
-2.96169519e-01 -7.48006821e-01 -7.09618568e-01 9.55777705e-01
1.28052443e-01 -4.10716593e-01 2.23381240e-02 4.48406041e-01]
|
[10.78461742401123, 1.1444071531295776]
|
74a19d10-a4d2-4a7a-bbd7-3eeb1ab92669
|
tormentor-deterministic-dynamic-path-data
|
2204.03776
| null |
https://arxiv.org/abs/2204.03776v1
|
https://arxiv.org/pdf/2204.03776v1.pdf
|
TorMentor: Deterministic dynamic-path, data augmentations with fractals
|
We propose the use of fractals as a means of efficient data augmentation. Specifically, we employ plasma fractals for adapting global image augmentation transformations into continuous local transforms. We formulate the diamond square algorithm as a cascade of simple convolution operations allowing efficient computation of plasma fractals on the GPU. We present the TorMentor image augmentation framework that is totally modular and deterministic across images and point-clouds. All image augmentation operations can be combined through pipelining and random branching to form flow networks of arbitrary width and depth. We demonstrate the efficiency of the proposed approach with experiments on document image segmentation (binarization) with the DIBCO datasets. The proposed approach demonstrates superior performance to traditional image augmentation techniques. Finally, we use extended synthetic binary text images in a self-supervision regiment and outperform the same model when trained with limited data and simple extensions.
|
['Mathias Seuret', 'Georg Vogeler', 'Jian Shi', 'Edgar Riba', 'Vincent Christlein', 'Anguelos Nicolaou']
|
2022-04-07
| null | null | null | null |
['activity-recognition-in-videos', 'image-augmentation']
|
['computer-vision', 'computer-vision']
|
[ 7.67702341e-01 1.36464834e-01 1.20473817e-01 -2.61202604e-01
-2.73343116e-01 -6.62516475e-01 1.18241453e+00 -4.44241166e-02
-4.74163592e-01 4.79288399e-01 -1.52333707e-01 -6.86349928e-01
3.65043730e-01 -1.01526940e+00 -9.62956250e-01 -8.57231736e-01
-7.39831254e-02 6.96070313e-01 2.04902202e-01 -3.35263908e-01
5.29927969e-01 1.14287877e+00 -1.28171003e+00 5.21845818e-01
1.06945884e+00 9.96649623e-01 -6.71158135e-02 1.39665735e+00
-3.90734106e-01 8.55253339e-01 -6.02810144e-01 -4.57443058e-01
7.13501811e-01 -7.10996836e-02 -1.01866686e+00 5.56277454e-01
4.71893102e-01 -4.75547403e-01 -5.95701456e-01 7.20107257e-01
3.44516307e-01 1.90128341e-01 7.36803353e-01 -1.26752794e+00
-7.00815558e-01 4.10021484e-01 -1.04387617e+00 6.71046257e-01
-3.92467797e-01 4.24125046e-01 4.20510054e-01 -6.29672229e-01
3.74574870e-01 1.14354300e+00 5.71075320e-01 3.79223228e-01
-1.36410296e+00 -5.22750318e-01 -2.10884828e-02 4.19949219e-02
-8.91962111e-01 -1.43990979e-01 6.16529405e-01 -3.56040567e-01
1.10502255e+00 3.36211801e-01 5.13926208e-01 9.37326550e-01
-1.90732721e-02 6.63804948e-01 1.16150999e+00 -6.09834075e-01
6.67405650e-02 1.06992818e-01 1.53172731e-01 9.92032766e-01
-1.21670768e-01 3.07957798e-01 -3.70022774e-01 -1.33677617e-01
1.22627759e+00 -1.47761256e-01 2.76694801e-02 2.60598725e-03
-1.23529160e+00 1.01047885e+00 6.74053967e-01 7.76318610e-02
-1.47258863e-01 4.32873487e-01 5.82741857e-01 1.72611296e-01
7.17282653e-01 4.73052531e-01 -3.20776641e-01 2.83059683e-02
-8.92877460e-01 1.35652035e-01 4.64629471e-01 1.07540941e+00
7.44712293e-01 4.07035798e-01 -2.42464572e-01 4.80859369e-01
-3.33736449e-01 3.59789610e-01 4.94518906e-01 -9.62656975e-01
3.01022261e-01 4.80841786e-01 2.36384198e-02 -7.17743874e-01
-3.98810178e-01 -4.43657726e-01 -1.10677874e+00 3.38995010e-01
2.16491029e-01 -1.91714466e-01 -1.39244962e+00 1.14322078e+00
6.72811031e-01 3.60727489e-01 2.55628794e-01 5.83655179e-01
4.67201054e-01 8.87190163e-01 2.27581896e-02 2.56647378e-01
1.46718645e+00 -1.79185236e+00 -3.31421256e-01 -6.04673140e-02
8.57836902e-01 -7.88113356e-01 1.36835730e+00 4.42146301e-01
-9.19585109e-01 -7.10768819e-01 -1.30277324e+00 -4.13094461e-01
-6.17847025e-01 -2.45108549e-02 1.27800834e+00 1.09358871e+00
-1.10306942e+00 7.75377333e-01 -1.02138150e+00 -2.08285242e-01
6.04007781e-01 5.23651659e-01 1.56791490e-02 2.85543352e-01
-5.02342463e-01 1.62738353e-01 5.44964373e-01 -2.12439179e-01
-7.57687151e-01 -1.01675165e+00 -6.16208792e-01 -8.65961239e-02
-2.71107882e-01 -7.77907014e-01 1.16152620e+00 -9.36866343e-01
-1.66346514e+00 7.44286656e-01 2.03344926e-01 -1.04445660e+00
5.59307933e-01 -1.10183783e-01 -1.90629467e-01 7.04821765e-01
6.75375666e-03 1.33949745e+00 1.14419484e+00 -1.05855167e+00
-7.48903036e-01 -1.98749870e-01 -2.78700516e-03 3.70250821e-01
-5.18117607e-01 -7.66485482e-02 -4.02039677e-01 -7.42260337e-01
-1.47726178e-01 -7.74377346e-01 -6.06090486e-01 -5.65744285e-03
-5.77982903e-01 3.05408925e-01 1.28320670e+00 -5.49671590e-01
5.45622110e-01 -1.81704211e+00 -2.65794069e-01 2.31071874e-01
1.13496892e-01 -6.18915111e-02 -1.04556359e-01 -9.81066301e-02
-1.15305722e-01 1.52950823e-01 -7.00463235e-01 -4.97247905e-01
-2.79626191e-01 2.44978502e-01 -6.52488470e-01 4.50359821e-01
4.56560671e-01 1.14665842e+00 -4.19228524e-01 -4.81756508e-01
4.39623147e-01 6.11422002e-01 -7.64481127e-01 -1.09138831e-01
-3.88914317e-01 7.42868483e-01 -2.76811808e-01 4.98241514e-01
1.11418724e+00 -3.57352734e-01 -4.17840153e-01 -2.11497948e-01
-2.09664360e-01 -1.67779639e-01 -6.77936137e-01 1.69840825e+00
-5.10137439e-01 6.74823999e-01 1.71554480e-02 -1.10172093e+00
8.67003143e-01 -1.60136506e-01 4.75065112e-01 -7.05519736e-01
3.17178905e-01 -5.73515445e-02 -1.35464206e-01 -1.51428148e-01
9.32429314e-01 -4.30457257e-02 1.48474514e-01 7.87790358e-01
7.18528852e-02 -5.65195918e-01 2.26604924e-01 3.98684829e-01
1.13089716e+00 2.61042919e-02 -3.34057927e-01 -3.37646604e-01
6.48700774e-01 4.69749659e-01 -1.94578394e-01 8.67608726e-01
2.40625113e-01 5.32756627e-01 4.57497448e-01 -6.17096663e-01
-1.77204490e+00 -6.84374332e-01 -2.89469808e-01 1.03215647e+00
-6.85975477e-02 -2.19424531e-01 -1.04866910e+00 -6.77503824e-01
-1.95556179e-01 4.89872336e-01 -7.89485991e-01 3.02289933e-01
-6.94526315e-01 -1.33693159e+00 9.97335911e-01 6.55310452e-01
1.14100683e+00 -1.05782819e+00 -6.00974441e-01 9.93196666e-02
1.00815997e-01 -1.53786898e+00 -5.60768187e-01 4.53497201e-01
-1.00087416e+00 -7.33312547e-01 -6.36596620e-01 -8.22962403e-01
7.49665201e-01 3.69285613e-01 8.70100677e-01 2.91234404e-01
-7.06347287e-01 4.16593015e-01 -2.28015795e-01 -2.38897815e-01
-5.28674483e-01 3.03482264e-01 -1.92675903e-01 1.49785072e-01
-2.88698543e-02 -7.46283352e-01 -8.64158392e-01 1.47467464e-01
-1.20324624e+00 3.87931645e-01 2.70250887e-01 9.18024182e-01
5.49834490e-01 8.54372606e-02 -5.06825447e-02 -9.14661229e-01
5.20340502e-01 1.25472486e-01 -7.28083074e-01 -1.04327358e-01
-3.29140276e-01 9.23531204e-02 7.56948352e-01 -4.29380298e-01
-1.43765259e+00 1.53079554e-01 -1.45888910e-01 -5.51595449e-01
-1.82547495e-01 -1.90933630e-01 3.74123514e-01 -8.34370553e-01
9.94221151e-01 2.48888820e-01 -1.49091771e-02 -1.96024984e-01
8.78219187e-01 7.27278829e-01 9.95861471e-01 -9.93598402e-01
9.16444659e-01 1.15304506e+00 2.12398246e-01 -1.35853279e+00
-4.15580302e-01 -2.82478958e-01 -9.60344732e-01 3.06572840e-02
1.01622987e+00 -8.30597281e-01 -6.76168740e-01 5.39693892e-01
-1.21918905e+00 -7.72566855e-01 -6.95963323e-01 -8.50470513e-02
-7.35899985e-01 6.61199987e-01 -1.01342225e+00 -5.03409207e-01
-7.13803053e-01 -1.05263042e+00 1.35046124e+00 4.09528762e-01
3.53677273e-01 -8.97293687e-01 -2.36951962e-01 3.85151356e-01
2.23719910e-01 4.23737586e-01 1.02625883e+00 -2.43463293e-01
-9.14636672e-01 2.16652408e-01 -6.17711306e-01 2.48688892e-01
-4.16968465e-02 2.50264406e-01 -1.09344900e+00 -2.29201511e-01
1.78452656e-01 -3.59295428e-01 1.05082846e+00 2.35596240e-01
1.63273704e+00 -3.00127298e-01 -3.58649999e-01 1.39255512e+00
1.38137925e+00 1.20219178e-01 7.22023070e-01 7.17724502e-01
8.32502306e-01 4.27931279e-01 2.83802330e-01 3.89927804e-01
-1.20637350e-01 2.42489442e-01 1.27549291e-01 -8.23931694e-01
-2.43914977e-01 1.14322260e-01 7.49351606e-02 5.28113008e-01
-7.31919929e-02 -1.00635260e-01 -9.94894505e-01 2.72934586e-01
-1.48546517e+00 -5.52173555e-01 -3.57232034e-01 1.55953944e+00
4.05052006e-01 1.14656180e-01 2.10181195e-02 1.90494761e-01
5.86863637e-01 1.11492492e-01 -5.74610531e-01 -9.36184645e-01
-4.04516131e-01 6.83891237e-01 9.26895499e-01 5.74835956e-01
-1.35310221e+00 1.34909809e+00 6.88505316e+00 9.70071375e-01
-1.21865630e+00 3.02114248e-01 1.28760934e+00 8.16681683e-02
-1.63252994e-01 -2.37702265e-01 -5.47005117e-01 1.03592925e-01
8.48756015e-01 2.09201217e-01 5.02938032e-01 5.64647079e-01
-2.45331958e-01 4.74333912e-02 -5.94927251e-01 7.17340231e-01
-8.21387246e-02 -1.76783466e+00 3.43469918e-01 7.87266940e-02
9.87070322e-01 2.68872142e-01 4.21465486e-01 -2.15726579e-03
4.24142092e-01 -9.15616393e-01 4.34316725e-01 -3.01591791e-02
1.07867801e+00 -9.90490913e-01 1.51052937e-01 7.69458786e-02
-1.03869188e+00 -2.49225482e-01 -5.19013822e-01 1.39948666e-01
3.17730345e-02 2.45282680e-01 -9.22719002e-01 4.46254998e-01
6.16719782e-01 1.43885598e-01 -7.63749063e-01 6.50039613e-01
5.21725565e-02 4.59302634e-01 -7.49103129e-01 2.96161622e-01
7.67824411e-01 -6.96512997e-01 2.46939212e-01 1.38112772e+00
1.93076760e-01 3.03372920e-01 1.15811769e-02 7.75847316e-01
-1.01513155e-01 2.10083246e-01 -5.98437846e-01 -4.80648093e-02
-1.39116302e-01 1.54737604e+00 -1.51275849e+00 -5.32132089e-01
-4.72298473e-01 1.36730409e+00 2.10721731e-01 3.05290699e-01
-9.13061440e-01 -3.58053118e-01 2.92124867e-01 -5.90380281e-02
4.63307053e-01 -3.25250298e-01 -8.44547510e-01 -8.84477377e-01
-3.50346357e-01 -8.15324366e-01 2.61174351e-01 -8.91250432e-01
-1.11773694e+00 9.78550375e-01 -2.90648162e-01 -7.51200140e-01
2.28538334e-01 -8.05560827e-01 -8.44107270e-01 5.70375502e-01
-1.44968677e+00 -1.50683844e+00 -6.44385159e-01 7.02058017e-01
4.96555775e-01 -3.20238262e-01 6.44636512e-01 2.71964096e-03
-3.69454354e-01 4.47341472e-01 2.62968540e-01 -5.85371628e-02
2.23846197e-01 -1.29879224e+00 1.28331494e+00 9.85343933e-01
2.37681851e-01 8.07289779e-02 5.83155096e-01 -8.34533513e-01
-8.01355422e-01 -1.31700897e+00 -2.30473727e-02 -3.28508168e-01
7.68243730e-01 -6.84726655e-01 -8.29257727e-01 7.48941481e-01
5.74026883e-01 -4.07988988e-02 4.29509908e-01 -3.32833320e-01
-1.86051562e-01 1.66064799e-01 -1.35770833e+00 8.10573518e-01
9.92571115e-01 -4.79047030e-01 2.15487778e-02 1.02280223e+00
1.24902499e+00 -6.82946920e-01 -6.61947310e-01 2.83073068e-01
-1.27200708e-01 -8.76473546e-01 1.16147435e+00 -5.90177774e-01
5.82861185e-01 -1.56553656e-01 2.97730770e-02 -7.53372908e-01
-1.23166569e-01 -1.06838310e+00 -7.00754598e-02 9.68932688e-01
2.56697774e-01 -3.78996372e-01 1.36197591e+00 1.95546255e-01
-1.93577781e-01 -6.48520470e-01 -7.39472151e-01 -5.32369852e-01
2.73495257e-01 -3.46805781e-01 9.28485930e-01 8.13642800e-01
-2.89203525e-01 1.48735315e-01 -1.30778164e-01 2.40517557e-01
7.40331888e-01 1.16284184e-01 9.28057313e-01 -7.23154306e-01
-3.37262243e-01 -2.29480609e-01 -3.22761893e-01 -1.25127959e+00
-2.89268475e-02 -1.09048581e+00 -2.75615156e-01 -9.64582920e-01
1.41295595e-02 -6.09292328e-01 2.24768966e-01 3.26568246e-01
5.31447418e-02 8.02141309e-01 2.46163271e-02 1.80566117e-01
2.28940006e-02 6.43615484e-01 1.54197764e+00 -3.87228310e-01
-3.89355451e-01 -1.90220818e-01 -6.42109156e-01 6.24608696e-01
1.09182024e+00 -1.33662432e-01 -3.91517818e-01 -3.95275503e-01
-1.97186425e-01 -2.15364948e-01 4.15714264e-01 -1.22263980e+00
2.64268905e-01 3.18889737e-01 7.25529671e-01 -7.77547538e-01
4.17840153e-01 -4.57331359e-01 -4.03251112e-01 3.65474701e-01
-5.17395698e-02 3.35400194e-01 6.85945153e-01 4.90325183e-01
2.25743935e-01 -2.11281046e-01 9.79355574e-01 -1.39900267e-01
-2.20829472e-01 5.21765888e-01 -4.06466991e-01 -3.18671703e-01
1.04206097e+00 -3.29801291e-01 -5.06754577e-01 -5.48377819e-02
-6.97221816e-01 -6.01191260e-02 5.54251134e-01 1.20919116e-01
3.73100013e-01 -1.08748984e+00 -3.81653398e-01 5.26508570e-01
-3.78014863e-01 2.53073812e-01 2.92431355e-01 6.42826974e-01
-1.28108692e+00 5.22873461e-01 -4.21254903e-01 -8.63459766e-01
-1.13775790e+00 7.79196978e-01 4.48458642e-01 -3.97518158e-01
-1.09165668e+00 9.30898964e-01 5.19012868e-01 -5.43543756e-01
-4.37104031e-02 -3.46764803e-01 1.79089364e-02 -3.09448063e-01
4.18333709e-01 2.13205963e-01 3.10944766e-01 -4.26722020e-01
2.79598832e-01 4.49269950e-01 -4.52736557e-01 -3.28733414e-01
1.27420270e+00 1.06317624e-02 -2.11129233e-01 -2.74122417e-01
1.04500270e+00 -3.03379953e-01 -1.51605999e+00 -2.18794812e-02
-2.58969218e-01 -2.25378618e-01 6.40772134e-02 -5.79144835e-01
-1.15708995e+00 1.13455904e+00 7.09482253e-01 1.31372124e-01
1.22850621e+00 -3.60302418e-01 7.37075448e-01 5.13319790e-01
-5.45696989e-02 -9.63005483e-01 2.15376228e-01 4.76089507e-01
7.10761964e-01 -8.67272675e-01 -3.93669158e-02 -7.99037516e-01
-5.52060485e-01 1.26532733e+00 7.64871716e-01 -2.64009297e-01
4.51062441e-01 8.04927230e-01 5.80683239e-02 -2.80839622e-01
-4.98786628e-01 -8.22298154e-02 -1.37781546e-01 7.66312897e-01
6.10591518e-03 -1.68518543e-01 2.24659126e-02 3.51488627e-02
-6.34500504e-01 -3.56661737e-01 7.50277340e-01 8.19492579e-01
-2.35156685e-01 -1.01542544e+00 -4.86329913e-01 5.88978171e-01
-3.46227705e-01 -3.85170996e-01 -2.15255022e-01 8.57570827e-01
8.15129355e-02 4.80385453e-01 7.20700085e-01 -1.56986699e-01
-1.42488450e-01 -1.42932637e-03 8.05369616e-01 -2.19823480e-01
-8.94155204e-01 6.14749826e-02 -1.15516394e-01 -2.70495027e-01
-3.13576907e-01 -4.45163369e-01 -1.38831902e+00 -5.08855283e-01
-1.95477247e-01 9.15648341e-02 7.52760887e-01 7.55115926e-01
3.04605126e-01 8.33111227e-01 7.37209201e-01 -1.09843957e+00
-1.45279303e-01 -8.11192870e-01 -5.11434257e-01 1.88828096e-01
1.41560286e-01 -3.06514740e-01 -2.09712401e-01 2.91579902e-01]
|
[9.671895980834961, 0.08167777955532074]
|
5d3fbfe2-d738-4578-af17-d308eb761a2b
|
2d-reconstruction-of-small-intestines
|
1803.05817
| null |
http://arxiv.org/abs/1803.05817v1
|
http://arxiv.org/pdf/1803.05817v1.pdf
|
2D Reconstruction of Small Intestine's Interior Wall
|
Examining and interpreting of a large number of wireless endoscopic images
from the gastrointestinal tract is a tiresome task for physicians. A practical
solution is to automatically construct a two dimensional representation of the
gastrointestinal tract for easy inspection. However, little has been done on
wireless endoscopic image stitching, let alone systematic investigation. The
proposed new wireless endoscopic image stitching method consists of two main
steps to improve the accuracy and efficiency of image registration. First, the
keypoints are extracted by Principle Component Analysis and Scale Invariant
Feature Transform (PCA-SIFT) algorithm and refined with Maximum Likelihood
Estimation SAmple Consensus (MLESAC) outlier removal to find the most reliable
keypoints. Second, the optimal transformation parameters obtained from first
step are fed to the Normalised Mutual Information (NMI) algorithm as an initial
solution. With modified Marquardt-Levenberg search strategy in a multiscale
framework, the NMI can find the optimal transformation parameters in the
shortest time. The proposed methodology has been tested on two different
datasets - one with real wireless endoscopic images and another with images
obtained from Micro-Ball (a new wireless cubic endoscopy system with six image
sensors). The results have demonstrated the accuracy and robustness of the
proposed methodology both visually and quantitatively.
|
['Xiang Xie', 'Rahman Attar', 'Zhihua Wang', 'Shigang Yue']
|
2018-03-15
| null | null | null | null |
['image-stitching']
|
['computer-vision']
|
[ 2.33503059e-01 -8.70785788e-02 2.64657110e-01 2.02367708e-01
-5.18593490e-01 -4.48009908e-01 1.47333398e-01 2.59361595e-01
-7.23088801e-01 5.78964129e-02 -1.79332122e-02 -2.54320838e-02
-5.09846210e-01 -2.45982513e-01 -4.53560919e-01 -9.00799394e-01
-2.72195011e-01 1.07206441e-01 1.88780248e-01 -5.55445366e-02
3.79520059e-01 3.03442866e-01 -1.19253099e+00 -3.53607208e-01
9.08093989e-01 8.18986654e-01 3.05949748e-01 8.14359963e-01
5.15294135e-01 1.24968357e-01 -2.99406797e-01 6.25349907e-03
6.64568484e-01 -3.32532614e-01 -3.71975273e-01 3.79342437e-01
-3.28500718e-02 -2.48683155e-01 9.64559242e-02 1.33015823e+00
5.72926819e-01 2.70248771e-01 5.12778819e-01 -8.71798277e-01
-6.95775226e-02 -7.88294598e-02 -8.39851201e-01 3.90747003e-02
5.75177968e-01 -6.04834259e-02 4.28309441e-01 -6.33871853e-01
7.90812373e-01 9.00154710e-01 8.62779438e-01 -7.79783651e-02
-1.08532250e+00 -3.29711080e-01 -5.50212145e-01 1.29344454e-02
-1.18988633e+00 1.29586637e-01 7.71471143e-01 -3.05289924e-01
5.19953191e-01 2.85433054e-01 9.19358552e-01 4.44711566e-01
6.34339392e-01 2.88946182e-01 9.01818991e-01 -6.27583444e-01
3.01002245e-02 4.00473207e-01 5.69587648e-02 8.14510882e-01
6.02603316e-01 3.14921170e-01 -1.32605821e-01 -3.92953217e-01
8.25106263e-01 3.69492143e-01 -3.67477119e-01 -7.10652292e-01
-1.62654805e+00 7.36707747e-01 3.98742616e-01 5.76185644e-01
-5.38477182e-01 -2.53353089e-01 4.51752603e-01 5.29821217e-01
-7.88279548e-02 3.28990191e-01 -6.94425479e-02 -1.08109892e-03
-6.34591460e-01 -2.84703314e-01 9.37789440e-01 8.21091831e-01
3.05983096e-01 -2.94515967e-01 6.81226611e-01 4.69859451e-01
5.95307946e-01 5.15689075e-01 1.13432527e+00 -4.58981484e-01
2.37846956e-01 4.75958973e-01 2.97770381e-01 -1.68324292e+00
-5.53324163e-01 -1.20931923e-01 -8.77571762e-01 2.56068379e-01
3.54545534e-01 -1.94689229e-01 -4.79813844e-01 8.96875262e-01
6.77500069e-01 3.40437740e-01 8.37095305e-02 1.02928090e+00
6.04993701e-01 3.01297456e-01 -2.58369327e-01 -2.45399386e-01
1.46875536e+00 -6.11001432e-01 -6.30274415e-01 1.16105467e-01
6.29473925e-01 -1.29938436e+00 5.60122192e-01 5.56465149e-01
-6.58345997e-01 -5.69139779e-01 -1.30820775e+00 3.30398858e-01
-1.05285086e-01 2.61332899e-01 2.12112471e-01 5.58664441e-01
-7.78065026e-01 5.07274151e-01 -1.19244647e+00 -6.05993271e-01
-3.89670879e-01 6.35077596e-01 -8.76789093e-01 1.97908834e-01
-7.36883581e-01 7.05877483e-01 3.55417252e-01 2.02501789e-01
-4.50052433e-02 -2.39964172e-01 -9.58667159e-01 -2.99260169e-01
-3.45142814e-03 -7.92046845e-01 7.65190363e-01 -8.79907906e-01
-1.60613191e+00 6.58225894e-01 5.10479771e-02 -4.09347475e-01
4.41349059e-01 -1.36288870e-02 -2.69413501e-01 4.42795306e-01
-5.22144064e-02 2.65778929e-01 1.15123665e+00 -1.08105958e+00
-4.76195306e-01 -5.64942420e-01 -6.27662182e-01 5.27565718e-01
-1.95379600e-01 -1.09152183e-01 -2.61640489e-01 -6.77841306e-01
9.11054134e-01 -1.27458417e+00 -2.90771335e-01 -5.92164509e-03
-1.07947148e-01 3.11682880e-01 6.55169606e-01 -9.28108513e-01
9.10256803e-01 -2.32389975e+00 1.31010309e-01 6.89780474e-01
-5.88634871e-02 1.04210399e-01 1.33807242e-01 3.46128553e-01
-3.10942560e-01 -4.05901313e-01 2.04901725e-01 -3.00719477e-02
-4.17301565e-01 -1.22027934e-01 1.72234565e-01 1.00939775e+00
-4.38447028e-01 2.71466345e-01 -9.07400012e-01 -6.58546269e-01
8.19467485e-01 5.10248661e-01 -3.93452018e-01 1.01982011e-03
6.13355517e-01 6.77866638e-01 -5.19320428e-01 4.61064607e-01
7.80742168e-01 1.32155478e-01 1.27486005e-01 -8.53993535e-01
-2.53989935e-01 -3.98611635e-01 -1.72850084e+00 1.71499169e+00
-4.19466883e-01 2.99993694e-01 1.68182701e-01 -1.00347733e+00
9.52827692e-01 5.73301017e-01 9.12004590e-01 -5.32836199e-01
4.05913025e-01 4.78518784e-01 -1.09695785e-01 -9.24925685e-01
3.28949600e-01 -5.61127951e-03 9.78194177e-02 1.94665283e-01
6.24327697e-02 -3.71673703e-01 -1.08717363e-02 -1.35880142e-01
6.51367247e-01 -1.03717577e-02 7.08244443e-01 -3.28314841e-01
6.06059849e-01 2.19239801e-01 1.99877650e-01 3.56497854e-01
-2.18404844e-01 5.87329626e-01 -1.12889841e-01 -6.15499556e-01
-1.01103222e+00 -6.14331901e-01 -2.08732054e-01 -7.25468174e-02
6.96382105e-01 1.24057859e-01 -6.48099601e-01 -3.18129629e-01
-1.38285952e-02 2.83096880e-02 -3.80851626e-01 2.65448615e-02
-3.88659507e-01 -6.13545477e-01 -2.69425958e-02 -2.97449887e-01
3.79791677e-01 -8.15548539e-01 -1.01654983e+00 3.71854335e-01
8.21260456e-03 -7.53162920e-01 -5.67389190e-01 -1.35189191e-01
-1.24780869e+00 -1.35084486e+00 -8.18766832e-01 -1.21878982e+00
1.01276994e+00 6.23994648e-01 2.25008920e-01 1.33910179e-01
-7.86132693e-01 4.90709156e-01 -3.84768754e-01 -1.24108776e-01
-5.20454288e-01 -2.78782994e-01 1.79427698e-01 3.34981792e-02
1.48169979e-01 -4.49102163e-01 -9.61785078e-01 7.85548031e-01
-9.28009450e-01 -2.76804954e-01 6.12128913e-01 9.01780188e-01
5.50012052e-01 3.44887525e-01 9.35555175e-02 -1.97857901e-01
7.64980555e-01 -2.15129644e-01 -7.43379176e-01 8.97511691e-02
-4.17743176e-01 -1.34428278e-01 4.68355924e-01 -6.13042057e-01
-6.49769008e-01 2.80426711e-01 -4.99974489e-02 -2.79688656e-01
-4.08022031e-02 4.78378952e-01 5.34632742e-01 -5.00214636e-01
7.16593385e-01 2.02236831e-01 6.76964879e-01 -2.69737661e-01
1.93661749e-01 8.65409791e-01 6.92452192e-01 -8.92047361e-02
6.54936731e-01 6.68204844e-01 -6.91782758e-02 -1.14461172e+00
7.40595385e-02 -1.16061175e+00 -6.50496602e-01 -1.46877572e-01
7.19143391e-01 -6.79082453e-01 -8.67818177e-01 5.16538858e-01
-7.66357422e-01 4.86945927e-01 -3.76730412e-02 1.25647128e+00
-4.57688242e-01 1.03836882e+00 -5.46418309e-01 -5.11865735e-01
-6.16991103e-01 -1.43471408e+00 6.87569439e-01 2.73721755e-01
-2.72024155e-01 -1.03577793e+00 3.44252318e-01 1.55814691e-02
1.72533303e-01 5.91231942e-01 3.16991866e-01 -3.95581335e-01
-1.53144494e-01 -8.22822392e-01 6.63212687e-02 1.82083279e-01
4.72346544e-01 -1.00099901e-03 -2.64094561e-01 -5.88017881e-01
8.18907082e-01 2.94120073e-01 2.57507801e-01 7.40118265e-01
1.64075539e-01 -2.73742557e-01 -5.49540818e-01 6.14182651e-01
1.57897246e+00 4.44035321e-01 3.86160254e-01 7.02571630e-01
2.02076390e-01 4.57475871e-01 1.07684326e+00 4.70659375e-01
-1.23989634e-01 6.44437850e-01 4.90345299e-01 -1.40571281e-01
1.70121774e-01 8.93241689e-02 2.16300815e-01 1.18720627e+00
5.61812557e-02 2.58545667e-01 -4.80754495e-01 5.41665554e-01
-1.71961606e+00 -7.10909307e-01 -9.06059220e-02 2.57062960e+00
3.22059691e-01 -1.71778798e-01 -1.00021765e-01 2.84896880e-01
8.35968852e-01 -4.48592186e-01 -3.11453998e-01 -1.96316719e-01
3.22625130e-01 -1.30192503e-01 6.42177343e-01 4.25596863e-01
-1.11880982e+00 1.32406667e-01 4.71325111e+00 5.82389951e-01
-1.07981324e+00 -2.27000833e-01 2.69787852e-02 5.06725669e-01
1.74158245e-01 -8.21308047e-02 -1.87307820e-01 4.95172471e-01
4.11334425e-01 -4.04464863e-02 2.76356965e-01 7.95261621e-01
3.50500375e-01 -5.88260889e-01 -4.63802159e-01 1.21153057e+00
-1.91664062e-02 -7.63783574e-01 -2.76797175e-01 9.70311314e-02
6.38847351e-01 -3.00046597e-02 8.50441158e-02 -5.46858251e-01
-3.13352287e-01 -3.01094413e-01 3.09796661e-01 4.76807386e-01
4.25012529e-01 -5.55326700e-01 1.00660765e+00 3.22591752e-01
-1.15777349e+00 -2.59925611e-03 -3.68579060e-01 5.50972879e-01
1.56437457e-01 3.92390668e-01 -9.88446414e-01 6.35518372e-01
6.91928267e-01 5.19503415e-01 -1.04616597e-01 1.54978621e+00
2.89909124e-01 1.20366225e-02 -7.97466755e-01 -1.11081608e-01
1.52226880e-01 -7.69770801e-01 9.51221645e-01 7.94189930e-01
8.23791862e-01 9.33041200e-02 -7.55115524e-02 2.29202390e-01
4.78675544e-01 4.98303860e-01 -5.98553896e-01 3.06809098e-01
1.89693287e-01 1.44222653e+00 -1.13754714e+00 -7.28273913e-02
-4.28535849e-01 1.01113224e+00 -4.63123620e-01 -4.90961596e-02
-5.22826791e-01 -6.09598875e-01 -3.40826958e-02 -3.06625403e-02
1.43278256e-01 -3.73005629e-01 1.14867970e-01 -1.04600954e+00
-7.20319897e-02 -9.26869929e-01 5.39270103e-01 -6.39231145e-01
-7.61393964e-01 6.11959040e-01 -9.05776471e-02 -1.93769753e+00
-3.72109264e-01 -5.96280217e-01 -5.05326986e-01 5.63455105e-01
-9.30784643e-01 -8.56881976e-01 -5.16361594e-01 7.82523453e-01
3.56619060e-01 2.00213995e-02 9.33731318e-01 6.77368864e-02
-2.38772795e-01 2.21620440e-01 6.31505549e-01 -1.59427568e-01
8.68828058e-01 -9.68283474e-01 -9.28300545e-02 6.61859930e-01
-8.34937990e-02 8.61288130e-01 9.95923340e-01 -7.18509912e-01
-1.54963434e+00 -2.66300321e-01 4.06785220e-01 1.06019080e-01
6.28920674e-01 3.59775603e-01 -5.34186184e-01 4.88261074e-01
2.35243991e-01 -7.38114938e-02 3.74804944e-01 -6.66020572e-01
3.08486104e-01 5.14201187e-02 -1.52925348e+00 6.13336146e-01
1.20102644e-01 3.29261720e-02 -6.50931656e-01 3.81004751e-01
4.05987948e-01 -6.85196280e-01 -1.32059753e+00 2.32247010e-01
8.17510784e-01 -8.58951926e-01 1.09289694e+00 3.03021520e-01
-3.81062597e-01 -5.01277208e-01 3.31035733e-01 -1.25166702e+00
9.08745304e-02 -9.61116016e-01 6.53849363e-01 5.09950399e-01
-8.53473507e-03 -9.51272905e-01 6.00636840e-01 4.66533512e-01
-9.09744762e-03 -3.19840819e-01 -9.71187115e-01 -4.81832951e-01
-7.67676830e-01 -4.24835971e-03 5.70521317e-02 7.27160513e-01
5.32110512e-01 -3.03020537e-01 -1.44259021e-01 4.41594929e-01
8.88049304e-01 3.13774645e-02 7.38399863e-01 -9.77817178e-01
-1.12932742e-01 4.29148525e-02 -7.80137897e-01 -6.69627786e-01
-7.08517194e-01 -4.67536986e-01 -1.04565822e-01 -1.11563969e+00
-6.28083423e-02 -2.19235912e-01 -1.03624620e-01 -7.18685091e-02
7.62634575e-02 7.09156767e-02 -5.09921871e-02 4.56839919e-01
-1.05176479e-01 1.05676847e-02 1.21822107e+00 2.27745190e-01
-6.17976665e-01 4.73567545e-01 -2.91215032e-02 6.81502223e-01
5.29889166e-01 -5.47235072e-01 -2.14688480e-01 1.08477451e-01
-2.76439283e-02 5.16294122e-01 2.36943990e-01 -1.03901076e+00
6.32103443e-01 4.44982469e-01 1.31414488e-01 -5.06583035e-01
5.15566266e-04 -1.38023853e+00 5.38544714e-01 1.02258444e+00
1.96938485e-01 3.57179523e-01 1.45990223e-01 6.34429216e-01
-4.78478700e-01 -6.75406933e-01 7.40757823e-01 -1.96072057e-01
-4.48753685e-01 -1.14187822e-01 -3.57874691e-01 -6.71530426e-01
1.13447917e+00 -6.99405015e-01 2.48731360e-01 -2.27547109e-01
-9.47571993e-01 -2.40883231e-01 5.54443419e-01 1.11798525e-01
8.03665102e-01 -1.08729911e+00 -3.60971481e-01 6.91304564e-01
-2.55502425e-02 2.42028516e-02 4.08304572e-01 1.47542143e+00
-1.10722935e+00 1.96084306e-01 -1.80429921e-01 -8.92827332e-01
-1.54631603e+00 6.13083243e-01 3.24949443e-01 -2.25456923e-01
-7.52649724e-01 3.46473694e-01 -3.79969746e-01 -3.25193077e-01
-1.90423187e-02 -5.86493075e-01 -2.29477882e-01 -5.46600409e-02
1.81212947e-01 4.16037351e-01 2.62085110e-01 -7.10072279e-01
-7.74116740e-02 1.14539003e+00 5.16215339e-02 -1.30714476e-01
1.21084154e+00 -5.01956165e-01 -1.25584811e-01 -3.00806731e-01
1.51711917e+00 5.05609810e-02 -8.94126296e-01 -6.85484931e-02
4.34932299e-02 -5.23564100e-01 8.12154114e-02 -2.11389676e-01
-7.59346128e-01 3.65164220e-01 1.04461074e+00 4.08363640e-01
1.07804334e+00 -6.06778562e-01 8.00991356e-01 2.55582631e-01
3.84601146e-01 -9.13757324e-01 5.14212437e-03 -2.01953575e-01
7.24133790e-01 -1.23560107e+00 8.96730796e-02 -3.12908620e-01
-4.91824359e-01 1.45775628e+00 -1.33669674e-01 -5.37486613e-01
9.14514124e-01 1.46732643e-01 1.63663834e-01 -8.05215761e-02
1.91026047e-01 2.10004434e-01 3.65383774e-01 3.24906886e-01
4.49228138e-02 -2.44635254e-01 -6.22813046e-01 8.97630900e-02
-2.48704590e-02 1.22249879e-01 3.15611929e-01 1.02273583e+00
-4.72881109e-01 -8.10872853e-01 -7.39670455e-01 1.51324943e-01
-5.43027461e-01 1.72133163e-01 4.24083292e-01 1.04672921e+00
-2.46669069e-01 1.04239929e+00 -3.09591204e-01 -3.37127507e-01
5.33535123e-01 -4.40685660e-01 4.71086860e-01 -8.82576033e-02
-5.99917233e-01 6.68562829e-01 -4.22571599e-01 -4.71190244e-01
-6.07070684e-01 -7.82996893e-01 -1.11240840e+00 1.98006257e-01
-8.49173248e-01 4.03443098e-01 1.25946879e+00 5.91912329e-01
1.23910278e-01 -9.90473405e-02 8.46472681e-01 -1.09429467e+00
-6.78672433e-01 -7.62990355e-01 -6.11431003e-01 7.12532282e-01
3.60405892e-01 -3.13365072e-01 -7.91026831e-01 1.99107334e-01]
|
[13.904006958007812, -3.15810489654541]
|
72e55249-eb12-4a2c-9dc7-fa8c8a4b6d44
|
a-relational-learning-perspective-to-multi
|
2103.06220
| null |
https://arxiv.org/abs/2103.06220v1
|
https://arxiv.org/pdf/2103.06220v1.pdf
|
A Relational-learning Perspective to Multi-label Chest X-ray Classification
|
Multi-label classification of chest X-ray images is frequently performed using discriminative approaches, i.e. learning to map an image directly to its binary labels. Such approaches make it challenging to incorporate auxiliary information such as annotation uncertainty or a dependency among the labels. Building towards this, we propose a novel knowledge graph reformulation of multi-label classification, which not only readily increases predictive performance of an encoder but also serves as a general framework for introducing new domain knowledge. Specifically, we construct a multi-modal knowledge graph out of the chest X-ray images and its labels and pose multi-label classification as a link prediction problem. Incorporating auxiliary information can then simply be achieved by adding additional nodes and relations among them. When tested on a publicly-available radiograph dataset (CheXpert), our relational-reformulation using a naive knowledge graph outperforms the state-of-art by achieving an area-under-ROC curve of 83.5%, an improvement of "sim 1" over a purely discriminative approach.
|
['Brandon Malone', 'Jens Kleesiek', 'Daniel Oñoro-Rubio', 'Anjany Sekuboyina']
|
2021-03-10
| null | null | null | null |
['multi-modal-knowledge-graph']
|
['knowledge-base']
|
[ 6.26819611e-01 5.38433850e-01 -6.43594444e-01 -6.79549813e-01
-1.34114599e+00 -6.88135505e-01 4.70724046e-01 7.67031252e-01
-2.64802992e-01 7.34348416e-01 9.41689163e-02 -2.93152601e-01
-4.59181488e-01 -6.68820262e-01 -7.13727355e-01 -6.95909441e-01
1.35850444e-01 8.37624907e-01 3.91307801e-01 2.34506413e-01
-2.29084775e-01 3.21714669e-01 -1.25494123e+00 3.93757373e-01
3.69441122e-01 1.12608290e+00 -3.07915527e-02 4.95018870e-01
1.72519684e-01 1.11704874e+00 -8.50104913e-02 -7.22772956e-01
-1.11410446e-01 -2.01842144e-01 -1.25563192e+00 2.08826542e-01
5.46015501e-01 -1.29985794e-01 -2.70395190e-01 9.10535574e-01
3.95843327e-01 -1.38001218e-01 8.10817838e-01 -9.31893170e-01
-5.32219708e-01 6.10158086e-01 -5.49616516e-01 5.09755798e-02
3.56916994e-01 -5.06319284e-01 1.53663492e+00 -3.94906610e-01
8.08846951e-01 8.70825052e-01 8.61928642e-01 3.44567209e-01
-1.60433352e+00 -3.19645941e-01 -3.89166437e-02 2.57162005e-01
-1.61778367e+00 -1.04056345e-02 5.53182423e-01 -5.19780993e-01
6.26321316e-01 4.05075878e-01 1.89682931e-01 8.68961513e-01
2.70279557e-01 6.71097934e-01 1.42263937e+00 -5.42062998e-01
-6.06067479e-02 2.82802105e-01 2.19662264e-01 1.22880602e+00
2.27750298e-02 -1.88346609e-01 -4.29407865e-01 -2.80210942e-01
3.59056205e-01 9.00995657e-02 -2.05533624e-01 -7.10505188e-01
-1.04145288e+00 8.32226813e-01 6.04263663e-01 2.36444205e-01
-2.10242182e-01 2.48068914e-01 3.68800998e-01 1.38165042e-01
5.22626579e-01 2.51587063e-01 -3.49267066e-01 2.40389749e-01
-9.66196954e-01 -1.72517896e-01 7.70317435e-01 7.45304167e-01
9.14350986e-01 -7.25952029e-01 -1.76033944e-01 9.05483186e-01
3.81332844e-01 1.30539820e-01 2.84235626e-01 -7.09739625e-01
2.32120141e-01 7.70768046e-01 -4.11311448e-01 -7.43408799e-01
-8.48061323e-01 -5.58231056e-01 -7.35114932e-01 -1.16876781e-01
3.51419508e-01 1.28429443e-01 -1.01446223e+00 1.67133605e+00
3.77234042e-01 1.42491907e-01 -1.87667996e-01 5.90546489e-01
8.12419295e-01 6.65837079e-02 2.25046962e-01 -3.85340840e-01
1.60756803e+00 -9.38502192e-01 -5.48089206e-01 -6.71116114e-02
8.72673512e-01 -5.85901499e-01 5.51593065e-01 2.80893832e-01
-7.29754806e-01 -3.44310373e-01 -9.92999554e-01 -5.60475513e-02
-3.09464693e-01 1.17451422e-01 6.79535329e-01 7.13653982e-01
-1.06779337e+00 5.99927068e-01 -9.25689042e-01 -1.64448529e-01
5.43133914e-01 6.27165318e-01 -7.60068119e-01 -4.86826450e-01
-8.78755569e-01 1.13655818e+00 5.53440571e-01 -4.97858793e-01
-7.32203543e-01 -7.10315764e-01 -1.02511740e+00 -1.76297158e-01
8.50026786e-01 -8.00564051e-01 1.11104143e+00 -5.25332808e-01
-1.05917537e+00 1.21080387e+00 3.09726242e-02 -2.48048916e-01
4.05213296e-01 4.56966907e-02 -3.77843022e-01 5.86908638e-01
2.32720152e-01 6.38655782e-01 6.17309749e-01 -1.31299603e+00
-6.17994070e-01 -4.03026938e-01 3.65864217e-01 1.68037161e-01
-2.38644108e-01 -1.75971806e-01 -5.46686172e-01 -5.64459562e-01
2.63031423e-01 -1.38371909e+00 -2.05334380e-01 -2.29113037e-03
-6.04154766e-01 -2.90302753e-01 5.90258062e-01 -5.24721146e-01
1.23519826e+00 -2.00991869e+00 3.05416137e-01 5.54745197e-01
5.48814774e-01 -1.94259435e-01 9.07184258e-02 2.44878411e-01
-2.70050257e-01 5.46666272e-02 -3.27309966e-01 -4.89371032e-01
-3.85247567e-03 5.89685857e-01 2.69097030e-01 6.31259799e-01
2.70563394e-01 1.05812931e+00 -9.92464364e-01 -9.47798431e-01
4.85824123e-02 4.81783241e-01 -4.85585868e-01 -1.63624492e-02
-8.56435448e-02 5.21388531e-01 -4.71497387e-01 7.45936871e-01
2.89816707e-01 -9.64868784e-01 5.27059555e-01 -4.58704293e-01
4.46325362e-01 1.20655023e-01 -1.09104478e+00 2.03618932e+00
-5.38762510e-01 2.74239063e-01 -2.35936388e-01 -1.04109216e+00
6.41430080e-01 4.21724111e-01 9.88114357e-01 -2.63977617e-01
3.98264192e-02 1.03712395e-01 -3.83701414e-01 -4.39224362e-01
1.54807493e-01 -5.79770088e-01 -1.03779413e-01 4.46651250e-01
4.33627099e-01 -1.23921424e-01 4.42292243e-02 2.42540643e-01
1.42955863e+00 1.32983401e-01 5.67676246e-01 -1.41887918e-01
6.13178849e-01 -1.52189836e-01 3.05982202e-01 7.11218655e-01
1.57803729e-01 6.17132545e-01 5.10699928e-01 -2.04647303e-01
-5.93962610e-01 -1.19466221e+00 -5.23052514e-01 1.09362483e+00
1.56446062e-02 -7.28239179e-01 -2.71010846e-01 -1.40752435e+00
1.79695830e-01 4.02067930e-01 -8.36350143e-01 -1.11286089e-01
-4.60304886e-01 -8.40511203e-01 5.62987864e-01 5.10478556e-01
6.67476356e-02 -3.20389897e-01 -3.34843189e-01 1.00987539e-01
-3.07864070e-01 -1.39120138e+00 -1.94045097e-01 7.94902205e-01
-7.38528490e-01 -1.30899477e+00 -4.58811104e-01 -6.50342643e-01
7.38616228e-01 -2.27961734e-01 1.30417061e+00 1.60679489e-01
-4.48122650e-01 7.63238966e-01 -3.75213921e-01 3.57890129e-02
-4.49772358e-01 2.20133439e-01 -2.89903671e-01 1.96625203e-01
9.75762233e-02 -4.68138844e-01 -5.15537739e-01 3.97059768e-01
-1.06850994e+00 4.36017662e-02 8.10827613e-01 1.10795665e+00
1.11025357e+00 2.48362888e-02 6.92987263e-01 -1.44471288e+00
1.63997069e-01 -5.06946445e-01 -2.37646446e-01 7.54927874e-01
-9.43483531e-01 3.14232290e-01 3.72472107e-01 -1.27211750e-01
-9.21624005e-01 5.92472017e-01 -1.43466622e-01 -1.67602286e-01
-1.49698988e-01 6.22280419e-01 1.92139000e-01 -4.54244822e-01
4.62373495e-01 -4.61338125e-02 -1.33317471e-01 -4.74788994e-01
7.20495164e-01 5.92165768e-01 4.68100339e-01 -5.79817235e-01
4.46945459e-01 4.97721016e-01 7.34728277e-01 -2.50124156e-01
-1.39299977e+00 -9.01280701e-01 -1.04371321e+00 -1.44021988e-01
1.12121201e+00 -7.46747673e-01 -6.34296834e-01 -2.26698607e-01
-7.30250061e-01 2.72241179e-02 -1.83968514e-01 6.33893669e-01
-6.93795562e-01 4.95050937e-01 -7.20149219e-01 -1.67251006e-01
8.41489583e-02 -1.29480720e+00 1.41998184e+00 -2.16495350e-01
-1.47517070e-01 -1.30304921e+00 1.72612727e-01 6.90745950e-01
-4.45233891e-04 3.40852618e-01 1.29208779e+00 -8.53963137e-01
-4.26076591e-01 -3.46917152e-01 -4.98120219e-01 3.90686005e-01
2.54118323e-01 -5.72819471e-01 -1.00325406e+00 -4.33154434e-01
-1.38413593e-01 -7.42600501e-01 9.13727999e-01 8.28273296e-02
1.12783945e+00 -8.10958538e-03 -7.44960964e-01 4.01081204e-01
1.69579947e+00 -2.45462134e-01 1.51767850e-01 1.02168068e-01
8.50943863e-01 4.43796217e-01 3.58219355e-01 2.70045221e-01
7.60001600e-01 9.05215979e-01 4.70150471e-01 -1.24186380e-02
-3.63880068e-01 -3.04976348e-02 -3.56985688e-01 7.92651534e-01
-7.92729184e-02 -1.87803656e-01 -1.11978376e+00 2.53341198e-01
-1.72852218e+00 -4.87852097e-01 -5.82739301e-02 1.89907825e+00
1.07358479e+00 6.40583858e-02 -1.22901320e-01 9.01865363e-02
5.58046877e-01 1.00535564e-01 -3.30339581e-01 1.46236151e-01
5.75567111e-02 5.10817885e-01 6.37951195e-01 4.43740338e-01
-1.51363099e+00 4.70680803e-01 6.05797529e+00 7.81151593e-01
-9.12891209e-01 4.86947000e-01 5.38678229e-01 1.66175812e-01
-1.42977521e-01 -5.29063605e-02 -8.16199183e-01 1.16080739e-01
9.65044856e-01 1.10944450e-01 8.70210901e-02 6.15132868e-01
-6.06740475e-01 -3.54361475e-01 -1.59380019e+00 9.52026367e-01
3.83090645e-01 -1.22002363e+00 1.98786315e-02 1.89337984e-01
7.81558454e-01 8.01274180e-02 -3.24114300e-02 1.84157357e-01
3.43691856e-01 -1.02735054e+00 3.40725332e-01 5.53646803e-01
1.03022563e+00 -4.69270438e-01 8.78826320e-01 1.67973816e-01
-1.17235255e+00 1.00557476e-01 9.40962061e-02 4.41129208e-01
2.55408138e-01 5.27960002e-01 -1.01528943e+00 1.06269097e+00
2.83354342e-01 7.42609978e-01 -8.94818127e-01 8.56157124e-01
-2.66845167e-01 5.25488436e-01 -2.82298326e-01 4.13111866e-01
2.26933032e-01 2.71235168e-01 9.95066203e-03 1.38793206e+00
-9.56669170e-03 2.14913160e-01 5.11783779e-01 2.03841016e-01
-3.60868722e-01 1.87479228e-01 -5.20328224e-01 2.86321312e-01
1.13398239e-01 1.46493208e+00 -9.29563940e-01 -3.67892534e-01
-7.63728201e-01 8.80654156e-01 5.21238565e-01 -9.58079100e-02
-8.86666417e-01 5.17615452e-02 -1.10743649e-01 -4.39707637e-02
2.96923816e-01 4.19408083e-02 -1.05452731e-01 -9.67958450e-01
1.74329225e-02 -5.50480843e-01 8.36049438e-01 -6.16168678e-01
-1.30360579e+00 5.01030028e-01 1.66580573e-01 -1.09861290e+00
-2.60864735e-01 -6.17946565e-01 2.32850611e-01 4.92894590e-01
-1.69905758e+00 -1.62430239e+00 -3.71788383e-01 5.65311909e-01
1.09469071e-01 1.41727373e-01 1.24203253e+00 5.85824072e-01
-2.35242203e-01 7.69453406e-01 1.05721645e-01 -2.16960721e-02
8.33188295e-01 -1.55327761e+00 -5.01382649e-01 3.95327598e-01
4.87077206e-01 5.97816668e-02 2.49399275e-01 -5.20253181e-01
-1.07910407e+00 -1.18310416e+00 8.65233660e-01 -8.24113667e-01
9.42939222e-01 -1.68250635e-01 -8.87551069e-01 8.73180509e-01
-5.66657893e-02 4.07935172e-01 1.29007387e+00 1.85132325e-01
-6.28976166e-01 -4.68555503e-02 -1.08717954e+00 3.85216586e-02
1.01060498e+00 -8.42176795e-01 -5.47565341e-01 7.29076266e-01
5.27266324e-01 -4.77312893e-01 -1.69539499e+00 6.44192159e-01
4.11389768e-01 -6.76456213e-01 1.13562512e+00 -4.73754734e-01
4.79609877e-01 -1.32103100e-01 -4.27504241e-01 -1.15005481e+00
-3.01441699e-01 -1.42514497e-01 -2.31864192e-02 1.07780921e+00
5.42480528e-01 -4.22939062e-01 7.55114794e-01 4.74676728e-01
-2.17438519e-01 -1.14643252e+00 -1.10551798e+00 -6.50296390e-01
-1.49780318e-01 -4.17228967e-01 8.49893689e-02 1.05976999e+00
6.40345663e-02 4.88460273e-01 -2.96844333e-01 2.74841011e-01
7.22702444e-01 1.15707465e-01 6.57706782e-02 -1.26770806e+00
-5.67482948e-01 -1.93271235e-01 -7.87213743e-01 -8.43400478e-01
3.08512926e-01 -1.61087382e+00 -2.28338704e-01 -1.61559761e+00
5.56573808e-01 -8.87638450e-01 -6.73309088e-01 9.76656854e-01
-9.21386033e-02 7.57246435e-01 9.07401666e-02 3.41933459e-01
-1.11763561e+00 1.48364335e-01 1.15466750e+00 -2.37179458e-01
2.84121513e-01 4.94966749e-03 -6.45706058e-01 7.13652849e-01
4.33487862e-01 -8.73480737e-01 -3.71574521e-01 -1.85626239e-01
4.17589128e-01 4.17205036e-01 4.26786840e-01 -9.21107590e-01
2.98470765e-01 2.01536581e-01 1.45719424e-01 -4.93528217e-01
4.89138395e-01 -9.63014781e-01 1.85960457e-01 3.28783065e-01
-4.84190524e-01 -4.31538969e-02 -4.56581935e-02 9.42902207e-01
-3.27767491e-01 -4.14452940e-01 6.90133512e-01 -1.72614276e-01
-5.67377090e-01 1.22968040e-01 1.55215830e-01 1.06916621e-01
1.23854208e+00 1.41703293e-01 -3.37106109e-01 6.60165995e-02
-1.23168731e+00 1.23878516e-01 2.29139671e-01 2.54327834e-01
3.59379083e-01 -1.29595041e+00 -5.20036578e-01 -1.67106260e-02
4.43235397e-01 -3.15952934e-02 1.21633314e-01 1.16683948e+00
-2.61210084e-01 5.03273845e-01 3.51711586e-02 -8.48525286e-01
-1.49777949e+00 5.31523347e-01 2.56201029e-01 -7.60477483e-01
-5.52561402e-01 8.28941107e-01 1.01044178e-01 -2.50836998e-01
1.52360573e-01 -2.04179451e-01 -1.75398052e-01 3.24825317e-01
2.29147095e-02 1.25087529e-01 4.62721974e-01 -8.01744044e-01
-6.38753295e-01 7.99395680e-01 -2.46698797e-01 1.87892973e-01
1.25938809e+00 -7.49676824e-02 -2.19980061e-01 5.59223592e-01
1.43819714e+00 -8.67572799e-02 -7.92998254e-01 -5.59189379e-01
3.15256178e-01 -2.51544267e-01 1.19950682e-01 -1.11637175e+00
-8.99353623e-01 6.79048538e-01 6.43661916e-01 2.34111726e-01
1.06605482e+00 7.06903100e-01 2.89688170e-01 4.98373866e-01
4.14004594e-01 -7.90216863e-01 3.30422163e-01 1.07269920e-01
5.02990663e-01 -1.57218122e+00 3.77764553e-01 -6.55768812e-01
-5.80425084e-01 1.17660093e+00 2.11611509e-01 2.72678852e-01
9.93550062e-01 6.21276423e-02 4.87541668e-02 -5.61871827e-01
-5.69625497e-01 -3.56772304e-01 6.79624856e-01 3.71923745e-01
5.02519250e-01 3.01796228e-01 -2.24164113e-01 2.81662196e-01
8.67060199e-02 5.10172136e-02 1.79506928e-01 9.38654602e-01
-1.80183873e-01 -1.43128991e+00 -4.26770747e-02 7.75574327e-01
-8.89195383e-01 -2.01017596e-02 -2.75451895e-02 8.08088958e-01
3.85793239e-01 8.15829813e-01 -2.12575048e-01 -5.06597579e-01
1.67218164e-01 2.15389043e-01 5.98164439e-01 -1.01097012e+00
-5.27308702e-01 -1.25260919e-01 3.49900693e-01 -4.36596751e-01
-8.82298112e-01 -6.81045473e-01 -1.13919210e+00 2.74476737e-01
-6.46774173e-01 -3.21881175e-02 6.11604691e-01 1.10767210e+00
1.18504616e-03 8.33590031e-01 4.09166127e-01 -3.93961430e-01
-5.17937362e-01 -6.01611257e-01 -5.49011886e-01 5.77899933e-01
2.89803147e-01 -9.88377631e-01 -5.21579608e-02 1.87282652e-01]
|
[9.501699447631836, 4.129260063171387]
|
1702d621-2e5a-49ca-a881-096d6ad59636
|
publicly-available-datasets-of-breast
|
2306.01546
| null |
https://arxiv.org/abs/2306.01546v1
|
https://arxiv.org/pdf/2306.01546v1.pdf
|
Publicly available datasets of breast histopathology H&E whole-slide images: A systematic review
|
Advancements in digital pathology and computing resources have made a significant impact in the field of computational pathology for breast cancer diagnosis and treatment. However, access to high-quality labeled histopathological images of breast cancer is a big challenge that limits the development of accurate and robust deep learning models. In this systematic review, we identified the publicly available datasets of breast H&E stained whole-slide images (WSI) that can be used to develop deep learning algorithms. We systematically searched nine scientific literature databases and nine research data repositories. We found twelve publicly available datasets, containing 5153 H&E WSIs of breast cancer. Moreover, we reported image metadata and characteristics for each dataset to assist researchers in selecting proper datasets for specific tasks in breast cancer computational pathology. In addition, we compiled a list of patch and private datasets that were used in the included articles as a supplementary resource for researchers. Notably, 22% of the included articles utilized multiple datasets, and only 12% of the articles used an external validation set, suggesting that the performance of other developed models may be susceptible to overestimation. The TCGA-BRCA was used in 47.4% of the selected studies. This dataset has a considerable selection bias that can impact the robustness and generalizability of the trained algorithms. There is also a lack of consistent metadata reporting of breast WSI datasets that can be an issue in developing accurate deep learning models, indicating the necessity of establishing explicit guidelines for documenting breast WSI dataset characteristics and metadata.
|
['Kajsa Møllersen', 'Lill-Tove Rasmussen Busund', 'Nikita Shvetsov', 'Lars Ailo Bongo', 'Masoud Tafavvoghi']
|
2023-06-02
| null | null | null | null |
['whole-slide-images', 'selection-bias']
|
['computer-vision', 'natural-language-processing']
|
[ 3.27851065e-02 -6.16221614e-02 -8.06733012e-01 -2.30246276e-01
-1.28250301e+00 -4.11014885e-01 9.66043994e-02 6.92286849e-01
-6.06188476e-01 6.20288670e-01 1.72989070e-01 -6.84388340e-01
-2.85632163e-01 -8.63759398e-01 -7.64221489e-01 -1.16440380e+00
2.25511685e-01 3.24908495e-01 -6.85777068e-02 4.44560170e-01
9.85093862e-02 6.54874146e-01 -1.02625430e+00 5.59647143e-01
5.85335135e-01 8.39442313e-01 4.24720079e-01 5.66561520e-01
-3.83415163e-01 8.17152441e-01 -6.21187747e-01 -4.27522391e-01
-1.42348483e-01 -6.46427497e-02 -7.54708290e-01 -3.31644177e-01
3.34615082e-01 -3.75991732e-01 -3.82668227e-01 7.86579788e-01
8.62832546e-01 -7.29898751e-01 6.43760443e-01 -9.67532992e-01
-5.78811705e-01 5.73079765e-01 -4.85618979e-01 4.50061202e-01
-3.90313029e-01 3.45221490e-01 6.11039460e-01 -5.26521146e-01
1.10463035e+00 5.78948855e-01 8.19891214e-01 4.94828135e-01
-8.48884344e-01 -1.05508590e+00 -6.86504126e-01 3.66512477e-01
-1.27229714e+00 -3.63979012e-01 3.96859437e-01 -5.93575180e-01
6.49057746e-01 2.22659111e-01 8.12461138e-01 9.19607103e-01
6.12903714e-01 3.98850828e-01 1.18465233e+00 -3.73593539e-01
3.71370092e-02 1.79356724e-01 2.48887427e-02 5.93507290e-01
9.30321336e-01 -2.06915841e-01 -3.36525708e-01 -3.79651904e-01
4.85829949e-01 6.52359724e-02 -2.31696405e-02 -6.43096492e-02
-1.15529346e+00 6.59640908e-01 5.64804137e-01 5.55298030e-01
-6.03632331e-02 -4.95802686e-02 9.22985673e-01 -2.47412231e-02
3.35526496e-01 4.91153710e-02 -2.80738920e-01 2.05612242e-01
-9.00328159e-01 -1.12534478e-01 3.06283206e-01 3.71390909e-01
3.18311810e-01 -4.19204265e-01 1.31320715e-01 7.86158681e-01
2.08795711e-01 5.09224772e-01 8.16740096e-01 -6.35623515e-01
1.49489760e-01 7.35138059e-01 -3.52434963e-01 -1.08584702e+00
-8.06532919e-01 -5.02723277e-01 -1.04171848e+00 -1.15800418e-01
4.96205807e-01 2.83813745e-01 -8.60612452e-01 1.40131295e+00
2.60344744e-01 -1.86884344e-01 1.08981296e-01 7.33467460e-01
1.38664138e+00 9.66157541e-02 5.35255253e-01 1.87608637e-02
1.45713699e+00 -2.72591233e-01 -8.06173980e-01 1.19678535e-01
1.22543085e+00 -6.08525097e-01 8.62926960e-01 1.01721361e-01
-7.70118833e-01 -1.06136814e-01 -9.00849521e-01 -2.30495751e-01
-6.64938211e-01 5.42403996e-01 8.06497455e-01 3.99129808e-01
-9.34330165e-01 1.44060582e-01 -1.15857327e+00 -7.63963997e-01
1.03213203e+00 3.14476341e-01 -8.34929526e-01 -1.88896313e-01
-9.08595860e-01 1.02350473e+00 3.48873615e-01 8.96220654e-02
-9.06429052e-01 -1.09287453e+00 -5.46434879e-01 -2.47423708e-01
7.23765418e-02 -5.88417590e-01 9.23185110e-01 -7.44987130e-01
-6.57031357e-01 1.55888736e+00 -1.97735384e-01 -2.98997343e-01
2.13046566e-01 5.46959996e-01 -4.42248911e-01 3.95077646e-01
2.39222288e-01 4.39348310e-01 6.26090448e-03 -9.01266336e-01
-8.78033102e-01 -7.63997614e-01 -4.43069607e-01 -2.15658948e-01
-4.85244542e-01 -9.82937962e-03 -5.32892823e-01 -5.41551590e-01
-1.20050155e-01 -7.91736543e-01 -2.47584105e-01 4.05830413e-01
-1.37484461e-01 -1.44911688e-02 5.86225927e-01 -8.20534587e-01
9.40896749e-01 -2.25234890e+00 -3.39402229e-01 -1.76283307e-02
3.67176473e-01 3.19796205e-01 -2.80202907e-02 6.27896860e-02
-3.97002734e-02 8.16847563e-01 3.50176096e-01 -9.71422642e-02
-4.08305585e-01 1.60197273e-01 2.92204946e-01 1.01971018e+00
2.47939020e-01 9.33423519e-01 -6.61012352e-01 -8.94345164e-01
1.50297746e-01 3.92017514e-01 -1.54750332e-01 8.48663319e-03
2.84722924e-01 1.46234214e-01 -2.91684568e-01 8.96047950e-01
6.06582642e-01 -5.18865347e-01 4.55817074e-01 -4.75100726e-01
3.93243283e-02 2.86230177e-01 -4.80283916e-01 1.37288356e+00
-4.21338558e-01 9.68150854e-01 1.95999697e-01 -7.91030824e-01
5.76216519e-01 2.73633271e-01 7.82615542e-01 -5.95474780e-01
2.66755521e-01 3.56889725e-01 2.31497318e-01 -5.36593020e-01
-1.46701140e-02 -1.41579822e-01 3.21746975e-01 2.12103158e-01
1.13383085e-02 1.08035505e-01 1.83231905e-01 4.61584926e-02
1.27218294e+00 -5.19701183e-01 2.60514617e-01 -1.84082642e-01
2.51611680e-01 3.94769132e-01 8.61927450e-01 6.18496358e-01
-2.86654502e-01 5.74158967e-01 7.01653183e-01 -4.46587652e-01
-9.69528258e-01 -7.25152671e-01 -8.95805538e-01 5.68457484e-01
-3.70493233e-01 -1.51243970e-01 -3.11204910e-01 -4.12823677e-01
1.69137090e-01 8.96233618e-02 -1.02773201e+00 -1.38541505e-01
-2.42720738e-01 -1.32535231e+00 8.82961035e-01 5.40632784e-01
2.89140314e-01 -6.55983150e-01 -6.85403407e-01 5.14839403e-02
-1.75136626e-01 -9.15425658e-01 1.93161517e-01 3.59523445e-01
-8.02309036e-01 -1.64313114e+00 -8.21593344e-01 -8.80063653e-01
9.55528319e-01 1.41951069e-01 7.78077781e-01 3.24313402e-01
-7.15294600e-01 -7.30967522e-02 -1.92053244e-01 -8.88381422e-01
-8.47810090e-01 3.00654948e-01 -3.52969974e-01 -4.46817547e-01
6.38012886e-01 2.35728741e-01 -5.43674529e-01 -3.24941203e-02
-1.07301092e+00 1.30326673e-02 1.10795569e+00 9.23832417e-01
9.04536307e-01 9.07174200e-02 5.45016766e-01 -9.30650473e-01
3.98056470e-02 -7.43545711e-01 -3.44373167e-01 1.74177855e-01
-4.83914584e-01 -4.94188696e-01 3.13910842e-01 -2.26126328e-01
-8.03608298e-01 -3.60302567e-01 -1.14717364e-01 1.20600834e-01
-2.91379601e-01 1.00949454e+00 4.44618836e-02 -2.57478982e-01
5.22995889e-01 -1.31335393e-01 5.99251151e-01 -3.05449277e-01
-6.35028601e-01 9.11413670e-01 1.85590282e-01 -1.23838253e-01
2.45279312e-01 6.58619165e-01 2.45344430e-01 -7.17357397e-01
-6.58503592e-01 -4.45882171e-01 -3.16305220e-01 -3.90539831e-03
7.44205475e-01 -9.85136390e-01 -6.74050510e-01 6.38588905e-01
-5.30096173e-01 -5.86046994e-01 1.08497508e-01 4.59383339e-01
3.99036147e-02 1.40169099e-01 -1.01279306e+00 -2.04858422e-01
-6.02694392e-01 -1.44550550e+00 9.58620489e-01 2.19886243e-01
-3.70857000e-01 -9.07950699e-01 6.72643110e-02 4.84699249e-01
5.43406904e-01 4.15182769e-01 1.37757719e+00 -6.50813878e-01
-5.81798218e-02 -5.80579340e-01 -4.18605357e-01 -5.94134182e-02
4.64682341e-01 6.25864446e-01 -8.69563997e-01 -1.91838399e-01
-2.46922314e-01 -3.04995060e-01 7.59971380e-01 6.76046848e-01
1.55884671e+00 -1.65652335e-01 -9.20061111e-01 6.88255370e-01
1.58393967e+00 3.97592455e-01 4.46531355e-01 7.05166996e-01
4.81730163e-01 5.65841973e-01 5.19085705e-01 2.44120598e-01
4.53613639e-01 2.87343502e-01 2.74168760e-01 -4.34996277e-01
-2.42301404e-01 5.94372973e-02 -2.19623625e-01 4.09552664e-01
1.04536109e-01 -1.36122003e-01 -1.54563117e+00 9.72251952e-01
-1.17519665e+00 -6.87576234e-01 -2.66545296e-01 1.87056708e+00
1.03404272e+00 1.13404142e-02 -2.00860143e-01 2.04434395e-01
6.25277638e-01 -1.48097575e-01 -4.74578589e-01 1.87542289e-02
-2.77514279e-01 1.08693354e-01 7.73172140e-01 -2.43800953e-01
-8.22495699e-01 3.93864393e-01 6.93978071e+00 7.22562969e-01
-1.64409733e+00 2.32217073e-01 1.06377983e+00 -3.12797666e-01
-9.44887996e-02 -2.96641767e-01 -7.15245247e-01 5.37418842e-01
1.08691227e+00 -2.34468803e-01 -5.61881304e-01 6.68251872e-01
4.41669464e-01 -4.08326000e-01 -7.85921752e-01 6.96005344e-01
-2.12533727e-01 -1.85299075e+00 -8.46390873e-02 4.97374684e-01
3.41721147e-01 2.39008874e-01 2.86570728e-01 -1.29863337e-01
-1.81622401e-01 -1.08076143e+00 1.07166789e-01 4.25770760e-01
1.31499708e+00 -4.89537060e-01 1.47612166e+00 -5.71994707e-02
-4.78730172e-01 8.57140124e-02 -4.79961753e-01 3.76265317e-01
-4.42515552e-01 7.71539807e-01 -1.21171832e+00 3.86446595e-01
9.52577531e-01 6.03822768e-01 -9.32182610e-01 1.06065595e+00
3.31360817e-01 1.04052126e+00 -1.91863611e-01 -1.66522488e-01
-4.71025668e-02 3.59833747e-01 -1.87786311e-01 1.28060174e+00
3.47206771e-01 2.26091314e-02 -3.94627512e-01 5.61578393e-01
-3.52432020e-02 3.08897048e-01 -3.08677197e-01 -5.00659525e-01
4.81481045e-01 1.60707331e+00 -8.55377376e-01 -7.65919983e-02
-7.93902218e-01 9.56739299e-03 1.38917014e-01 1.10034890e-01
-5.16271770e-01 -1.16677834e-02 4.33478206e-01 5.61246276e-01
-2.18080476e-01 2.56736249e-01 -4.92180765e-01 -6.71033382e-01
-3.60448271e-01 -9.22493517e-01 8.44352603e-01 -4.66835409e-01
-1.27844155e+00 1.30441770e-01 -1.89283088e-01 -9.42445695e-01
3.13620657e-01 -8.55068207e-01 -2.98039198e-01 7.65653193e-01
-1.46458006e+00 -1.27540755e+00 -4.58426476e-01 2.13451833e-01
-2.56591570e-02 -2.75600791e-01 9.89430487e-01 2.37427324e-01
-8.36136162e-01 7.20036089e-01 4.49209154e-01 4.10264820e-01
1.06298363e+00 -1.00478101e+00 -4.09076750e-01 2.03941703e-01
-6.36521518e-01 6.00818694e-01 2.80761003e-01 -7.32818604e-01
-1.48683679e+00 -1.20231605e+00 6.51627064e-01 -1.59160450e-01
6.92149878e-01 1.66324377e-01 -9.46121037e-01 7.37627506e-01
-9.50552523e-02 2.19593570e-01 1.56431007e+00 -1.13994129e-01
9.79411528e-02 -2.94007719e-01 -1.46089673e+00 4.72533405e-01
3.73133093e-01 -2.53872454e-01 -2.07336247e-02 1.95342794e-01
6.35004342e-02 -5.84568262e-01 -1.24162006e+00 5.07117569e-01
6.58063710e-01 -5.46892166e-01 6.30167067e-01 -3.51741165e-01
6.49500489e-01 -5.81779741e-02 1.78822517e-01 -8.94598007e-01
-4.29124922e-01 4.88761574e-01 2.99177349e-01 1.18052626e+00
3.86240244e-01 -6.66205585e-01 1.11265159e+00 7.60882139e-01
-1.26906916e-01 -1.07997715e+00 -8.46628845e-01 -6.87110424e-02
6.33049726e-01 -1.78223297e-01 6.34840250e-01 9.27043200e-01
-4.67412621e-02 -5.72689116e-01 4.44754601e-01 3.07511073e-02
5.27142763e-01 -1.94699928e-01 5.82616210e-01 -9.15154397e-01
2.14263558e-01 -6.25621259e-01 -6.36054814e-01 2.88810641e-01
2.23264351e-01 -1.23318160e+00 -4.57693458e-01 -1.78075027e+00
8.85831594e-01 -7.54832387e-01 -4.10533637e-01 7.26110637e-01
-8.17181319e-02 3.42854440e-01 -3.51567030e-01 2.89667159e-01
-1.31176682e-02 -1.66792393e-01 1.16349971e+00 -3.89640421e-01
2.29838312e-01 -3.45788509e-01 -9.52658534e-01 6.71039820e-01
9.61406410e-01 -5.95653296e-01 6.85602948e-02 -2.64774501e-01
1.59748808e-01 -1.14849769e-01 5.61042428e-01 -8.70156765e-01
2.46080860e-01 -4.07415688e-01 8.60491395e-01 -7.92617559e-01
-1.70987368e-01 -7.81366587e-01 5.54183483e-01 7.48515785e-01
-4.09898460e-01 -3.12371910e-01 4.67382133e-01 1.11917831e-01
-3.42939824e-01 -2.99353898e-01 7.60811508e-01 -1.91122100e-01
-4.36002493e-01 4.07042503e-01 -7.74159908e-01 -2.70750672e-01
1.04396081e+00 -3.32778126e-01 -5.15466511e-01 8.64266381e-02
-2.72792757e-01 1.03975289e-01 7.46962905e-01 5.63190356e-02
3.42293233e-01 -1.28096962e+00 -8.07939947e-01 -2.10752711e-01
3.15058321e-01 2.50669360e-01 6.07138336e-01 1.14219463e+00
-9.35655475e-01 6.90837145e-01 -3.76768947e-01 -4.84100431e-01
-1.32952774e+00 4.23167586e-01 3.48275036e-01 -3.33186656e-01
-5.91206431e-01 5.02207816e-01 4.45723347e-02 -1.91630468e-01
3.61940682e-01 -3.79921049e-01 -1.94449767e-01 2.04306304e-01
5.82454324e-01 2.21353263e-01 3.80907953e-01 -6.30135775e-01
-5.84628522e-01 1.83828980e-01 -4.11838651e-01 2.71243602e-01
1.34445226e+00 1.60597891e-01 -3.15898478e-01 5.17533302e-01
1.36293018e+00 -1.63637146e-01 -4.54218030e-01 3.14045250e-02
-1.41186610e-01 -2.11729854e-01 2.82555223e-01 -7.92781115e-01
-1.30894208e+00 6.90705657e-01 7.88048327e-01 -3.46371680e-01
1.16455138e+00 1.28339827e-01 5.50389528e-01 -6.88549131e-02
2.89682478e-01 -9.33929443e-01 -2.98355162e-01 1.89691561e-03
5.75295210e-01 -1.44804311e+00 1.03064805e-01 -1.82630450e-01
-3.14656913e-01 1.18821931e+00 5.29091835e-01 3.58426958e-01
6.81056559e-01 6.87018454e-01 4.03099149e-01 -2.62239486e-01
-7.98847377e-01 2.31290728e-01 -2.95934994e-02 6.45570338e-01
8.83381724e-01 -7.58822039e-02 -9.38244045e-01 1.05304182e+00
-8.70197192e-02 3.99879426e-01 5.93615711e-01 8.24639082e-01
-8.04209337e-02 -9.01589274e-01 -3.61276895e-01 1.02817523e+00
-1.09753919e+00 2.59645611e-01 -3.64083856e-01 1.07711232e+00
2.11519435e-01 5.82870662e-01 9.64193046e-02 5.36468774e-02
-1.25911579e-01 -1.85813248e-01 1.63178518e-01 -4.39072073e-01
-5.22386730e-01 -1.14916481e-01 -6.26537874e-02 -1.21118806e-01
-4.92958874e-01 -6.55667841e-01 -1.25284791e+00 -4.03612226e-01
-3.30254227e-01 -5.96084595e-02 7.74071693e-01 7.85894752e-01
3.55369031e-01 5.76329947e-01 7.77659118e-02 -2.15715006e-01
-8.71820450e-02 -9.21013951e-01 -5.88348210e-01 1.89517856e-01
2.30863869e-01 -5.26593924e-01 -3.56423408e-01 5.18426597e-02]
|
[15.143091201782227, -2.9933855533599854]
|
abe6c22a-f3d9-496f-8915-aeb915b37f4c
|
an-evidential-real-time-multi-mode-fault
|
2305.00169
| null |
https://arxiv.org/abs/2305.00169v2
|
https://arxiv.org/pdf/2305.00169v2.pdf
|
An Evidential Real-Time Multi-Mode Fault Diagnosis Approach Based on Broad Learning System
|
Fault diagnosis is a crucial area of research in industry. Industrial processes exhibit diverse operating conditions, where data often have non-Gaussian, multi-mode, and center-drift characteristics. Data-driven approaches are currently the main focus in the field, but continuous fault classification and parameter updates of fault classifiers pose challenges for multiple operating modes and real-time settings. Thus, a pressing issue is to achieve real-time multi-mode fault diagnosis in industrial systems. In this paper, a novel approach to achieve real-time multi-mode fault diagnosis is proposed for industrial applications, which addresses this critical research problem. Our approach uses an extended evidence reasoning (ER) algorithm to fuse information and merge outputs from different base classifiers. These base classifiers based on broad learning system (BLS) are trained to ensure maximum fault diagnosis accuracy. Furthermore, pseudo-label learning is used to update model parameters in real-time. The effectiveness of the proposed approach is demonstrated on the multi-mode Tennessee Eastman process dataset.
|
['Xiao He', 'Minyue Li', 'LiMin Wang', 'Zeyi Liu', 'Chen Li']
|
2023-04-29
| null | null | null | null |
['pseudo-label']
|
['miscellaneous']
|
[ 2.47101575e-01 -5.59893131e-01 -2.51253396e-01 -3.84249300e-01
-6.68672621e-01 -3.56492728e-01 3.48031931e-02 1.93055689e-01
4.72446889e-01 7.32486665e-01 -8.20629120e-01 -4.49023485e-01
-8.16687107e-01 -6.08178020e-01 -2.97821730e-01 -9.11553502e-01
5.69470506e-03 7.42733061e-01 2.08754957e-01 2.92448640e-01
6.55923903e-01 7.72112370e-01 -1.68660569e+00 4.65516716e-01
9.55437899e-01 1.04903424e+00 2.11176813e-01 7.23669767e-01
1.63259551e-01 8.70152295e-01 -9.28135872e-01 3.51850539e-01
4.58552539e-02 -2.60960817e-01 -7.52822876e-01 4.60654825e-01
-1.95592955e-01 -1.21216327e-02 -6.67401180e-02 1.04276454e+00
4.44533616e-01 1.29149526e-01 6.11192763e-01 -1.80846334e+00
-8.59198943e-02 5.28525412e-01 -3.72965991e-01 2.29350328e-01
3.88991833e-02 2.13075727e-01 4.58078265e-01 -6.81005836e-01
3.77617637e-03 1.25908840e+00 4.48499233e-01 2.96991378e-01
-9.65255380e-01 -8.48579168e-01 2.43249103e-01 6.02844894e-01
-1.16044319e+00 -2.42292341e-02 7.60281265e-01 -5.03495991e-01
1.12599218e+00 1.14989251e-01 3.85941565e-01 7.32461393e-01
1.13154244e+00 5.33092022e-01 1.27662718e+00 -3.81722987e-01
6.03750765e-01 -1.07509352e-01 2.09862217e-01 4.57041532e-01
6.33407652e-01 1.53356642e-01 -4.19434667e-01 -1.79044306e-01
7.34775484e-01 6.69028997e-01 -2.39068270e-01 -1.22441888e-01
-6.72842741e-01 6.26701772e-01 2.95564234e-02 1.65737957e-01
-3.83544087e-01 5.01827821e-02 5.94905019e-01 6.19409621e-01
2.97947258e-01 5.88941276e-01 -9.28098619e-01 -3.09741855e-01
-6.72147095e-01 2.88011789e-01 7.84666479e-01 8.14797997e-01
2.64477909e-01 1.50078014e-01 1.47719100e-01 7.47066736e-01
6.48161948e-01 3.37820470e-01 5.42595088e-01 -8.49286854e-01
2.38862529e-01 8.90072107e-01 -3.47802602e-02 -4.61369634e-01
-3.72284353e-01 -4.21601772e-01 -7.38844633e-01 6.27041638e-01
-2.17554756e-02 -6.53339550e-02 -1.08939934e+00 8.80379915e-01
2.48626828e-01 2.35416681e-01 2.86097974e-01 5.10580420e-01
-6.47968426e-02 4.20665741e-01 -6.79999590e-02 -5.27751446e-01
1.24527407e+00 -6.28881335e-01 -1.10589254e+00 -5.02986349e-02
5.08187532e-01 -6.07816339e-01 4.23031092e-01 1.16309285e+00
-7.36881614e-01 -5.94051480e-01 -1.27660489e+00 7.08996952e-01
-2.69295663e-01 2.25180462e-02 3.34510803e-01 4.26463217e-01
-2.25005507e-01 7.75913000e-01 -1.13319695e+00 -1.64492399e-01
2.82816976e-01 4.91136372e-01 -1.39144287e-01 -6.54641867e-01
-1.19613934e+00 1.12683749e+00 6.04316473e-01 3.12314957e-01
-1.09378982e+00 -5.95917702e-01 -6.15749478e-01 -3.31998706e-01
5.65037847e-01 -2.89491177e-01 1.48020017e+00 -3.83976996e-01
-1.38083637e+00 -7.68341050e-02 1.87524348e-01 -2.01152012e-01
2.99383700e-01 -3.45599681e-01 -1.07399750e+00 1.88282385e-01
-5.05586229e-02 -2.82254755e-01 9.74745572e-01 -1.58267868e+00
-1.22094321e+00 -5.03319442e-01 -2.86093801e-01 6.69988766e-02
6.98362812e-02 -3.00822288e-01 1.99278682e-01 -3.55413854e-01
5.43329418e-01 -9.04659092e-01 -7.42767900e-02 -4.92516994e-01
-2.97454566e-01 -3.79036397e-01 1.56183732e+00 -5.45234799e-01
1.34679568e+00 -1.97764194e+00 -2.12740377e-01 3.53889078e-01
-5.71705550e-02 1.89617313e-02 6.30930603e-01 3.93597066e-01
-3.24824005e-01 -2.99353689e-01 -3.76253158e-01 2.67494887e-01
-1.90930977e-01 5.81015766e-01 -8.32340494e-03 4.94934678e-01
4.76589739e-01 1.85638517e-01 -8.83572876e-01 -3.01074564e-01
6.40368164e-01 -9.90576670e-02 6.86455965e-02 2.92691499e-01
-7.00170025e-02 2.92716771e-01 -3.74983460e-01 1.29899228e+00
5.31804860e-01 -8.64839181e-02 2.53889471e-01 -4.71715212e-01
1.12369105e-01 -3.15731734e-01 -1.63795865e+00 1.21276832e+00
-4.75969046e-01 5.08724153e-02 6.67595863e-02 -1.22802830e+00
1.16997254e+00 8.62110972e-01 8.25661063e-01 -2.14932665e-01
2.67796397e-01 4.30580318e-01 1.91597626e-01 -8.50141048e-01
1.15369394e-01 -3.71325731e-01 -4.01236489e-02 2.42877021e-01
5.71585298e-02 -3.57390612e-01 2.12908983e-01 -4.58601832e-01
1.30461776e+00 1.67668432e-01 1.37020513e-01 -1.96797639e-01
4.63612348e-01 1.18576832e-01 8.49797785e-01 2.85095870e-01
-2.95223475e-01 1.91125035e-01 2.08423659e-01 -2.61274904e-01
-5.19384027e-01 -8.36398542e-01 -3.99224758e-01 3.60794365e-01
4.17248160e-01 1.03985652e-01 -3.34777653e-01 -9.67764556e-01
5.14414072e-01 9.39934552e-01 -2.16187716e-01 -8.72079134e-01
-2.82419235e-01 -1.08840668e+00 2.73490369e-01 7.01361179e-01
3.13272595e-01 -8.86384249e-01 -8.09993207e-01 6.82532668e-01
2.14654103e-01 -6.78678751e-01 2.64841467e-01 1.02502787e+00
-1.22292197e+00 -1.72707331e+00 -1.17011465e-01 -6.95111871e-01
7.40274191e-01 -9.10494849e-03 7.91510046e-01 -1.67643622e-01
-7.30301976e-01 1.92577213e-01 -3.70016575e-01 -6.39486372e-01
-5.73434949e-01 -2.69051939e-01 2.32712761e-01 -2.61689514e-01
3.91325563e-01 -3.90675105e-02 -2.42207512e-01 6.18140101e-01
-9.66532826e-01 -7.36141264e-01 8.70918274e-01 1.10415912e+00
5.10799229e-01 1.33511055e+00 1.22882533e+00 -8.72106194e-01
7.11519599e-01 -6.80334568e-01 -7.13938713e-01 4.85407203e-01
-1.49711573e+00 -2.41478652e-01 3.79206926e-01 -4.58141148e-01
-1.32517076e+00 1.28271550e-01 1.73650071e-01 -6.85308933e-01
-4.36676472e-01 7.06478655e-01 -5.20272374e-01 1.03491411e-01
4.24073845e-01 -2.32835531e-01 1.96581975e-01 -4.00404662e-01
-1.32499278e-01 1.28492093e+00 5.41550040e-01 -7.78598428e-01
4.19666052e-01 -1.09985128e-01 2.63852745e-01 -3.15453500e-01
-4.91791844e-01 -6.98645949e-01 -5.03230870e-01 -4.98695374e-01
4.27048326e-01 -9.09340262e-01 -8.00598919e-01 8.71859968e-01
-6.44922435e-01 -1.93186447e-01 -2.39249542e-01 6.58439159e-01
-4.73132640e-01 1.39903754e-01 -8.55928898e-01 -1.14627647e+00
-2.07953662e-01 -1.37131274e+00 8.72659028e-01 1.51652068e-01
-1.83730334e-01 -9.48226333e-01 -3.72686803e-01 3.35134685e-01
2.03798804e-02 2.99860358e-01 1.10580325e+00 -7.68852532e-01
-3.22272867e-01 -6.68568313e-01 2.19602436e-01 6.51595473e-01
9.94780898e-01 1.82665005e-01 -7.37777233e-01 -5.64072728e-01
4.54386324e-01 -2.48195976e-02 3.47969472e-01 3.72733921e-01
1.05056655e+00 3.11058193e-01 -6.07116759e-01 3.70826194e-04
1.52723515e+00 8.84478092e-01 2.73952663e-01 4.19725597e-01
7.82398760e-01 4.93493319e-01 1.53068936e+00 6.27281427e-01
8.54820088e-02 2.09709018e-01 6.74795330e-01 2.80378819e-01
2.39980847e-01 3.76800388e-01 4.53024089e-01 6.37814939e-01
3.05296004e-01 -3.60451162e-01 -8.37946355e-01 4.14772332e-01
-1.99543822e+00 -9.00008559e-01 -4.15330142e-01 2.12381124e+00
6.73663855e-01 4.51719016e-01 -3.99173349e-01 9.64052677e-01
9.58364904e-01 -8.26127231e-01 -9.80424047e-01 -9.73739624e-02
2.06384808e-01 -6.64642900e-02 5.90006828e-01 8.39223415e-02
-1.10395265e+00 1.89505383e-01 6.22933054e+00 3.55853021e-01
-1.19408286e+00 -1.73875168e-01 3.87296498e-01 1.23741105e-01
4.49862301e-01 -1.04603849e-01 -6.52756929e-01 5.05697489e-01
1.26184511e+00 1.43476307e-01 2.53758058e-02 9.02253628e-01
1.26972511e-01 -4.84678030e-01 -1.24570513e+00 1.00676560e+00
3.29793468e-02 -7.21522808e-01 -3.32454830e-01 1.10899404e-01
7.85086930e-01 -4.58942652e-01 -3.47845554e-01 1.82779744e-01
3.27587664e-01 -6.83745563e-01 5.65069020e-01 6.71105325e-01
5.53545177e-01 -1.04984748e+00 1.35437548e+00 8.53114128e-02
-1.10494411e+00 -8.75494421e-01 -5.69125079e-03 2.02504441e-01
3.98409724e-01 1.23717046e+00 -8.17187905e-01 1.20030570e+00
8.19109321e-01 7.66287982e-01 -3.04296672e-01 1.11964750e+00
2.51476150e-02 6.50435209e-01 5.03325947e-02 4.57886189e-01
-5.66965759e-01 1.58138782e-01 1.68943286e-01 6.97492540e-01
5.85654259e-01 -3.33886772e-01 6.16302609e-01 3.71690333e-01
4.74558741e-01 -6.46829724e-01 -4.35125232e-01 -3.09587978e-02
6.66649461e-01 1.14661825e+00 -1.01576602e+00 -3.63322139e-01
-2.21614987e-01 8.17660630e-01 -4.16205019e-01 2.98051685e-02
-7.06270993e-01 -4.78944063e-01 7.28879750e-01 -8.44662413e-02
1.64432824e-01 -1.52101174e-01 -4.05977517e-01 -5.28481603e-01
-2.66394883e-01 -1.05212891e+00 6.03646696e-01 -4.64458555e-01
-1.89121592e+00 9.78971943e-02 2.14271024e-01 -1.43788266e+00
-2.36030117e-01 -8.77283216e-01 -3.83370638e-01 7.39603817e-01
-1.42255771e+00 -8.94633174e-01 -3.14554363e-01 5.91157734e-01
1.00525248e+00 -3.43396515e-01 7.12587237e-01 4.98997182e-01
-1.03115702e+00 7.27855340e-02 1.20646268e-01 -4.27466482e-01
9.75342333e-01 -1.56099486e+00 -2.53442585e-01 8.02670717e-01
-6.48906946e-01 3.10495555e-01 8.35862696e-01 -1.31344938e+00
-1.77882111e+00 -1.51238835e+00 2.63053417e-01 -2.37879843e-01
4.28974599e-01 3.22012275e-01 -1.19426727e+00 5.00301719e-01
-1.67636752e-01 1.55710384e-01 6.76170349e-01 -5.91878034e-02
2.52977967e-01 -3.62952769e-01 -1.53060341e+00 -7.54539371e-02
3.74386966e-01 -3.16648155e-01 -6.40497684e-01 2.88147599e-01
1.54858634e-01 -5.35227656e-01 -1.45276272e+00 7.23250866e-01
1.57342404e-01 -3.83813679e-01 5.33958793e-01 -1.04125522e-01
-1.12373412e-01 -8.58278453e-01 -1.14085190e-01 -1.57771540e+00
-3.48381490e-01 -1.25453264e-01 -5.30301392e-01 1.29283130e+00
1.07614368e-01 -8.06890607e-01 3.38275015e-01 3.97429079e-01
-7.45159209e-01 -7.39257336e-01 -6.77299678e-01 -9.08395588e-01
-5.07318437e-01 -3.59687239e-01 6.02281809e-01 8.95086050e-01
2.49013990e-01 2.02676535e-01 8.95299092e-02 6.05607748e-01
6.52385712e-01 7.73845464e-02 1.43820226e-01 -1.66929638e+00
-1.45847965e-02 -3.01913880e-02 -5.13162851e-01 -1.10615149e-01
1.07175611e-01 -6.28124714e-01 7.28742063e-01 -1.63453984e+00
-2.01790884e-01 -6.15814805e-01 -8.48964036e-01 5.55447817e-01
-1.96999341e-01 -9.62106213e-02 -4.30526048e-01 2.45328516e-01
-2.75034815e-01 3.20648789e-01 9.85540807e-01 -2.69196063e-01
-5.81477284e-02 2.97503829e-01 -3.18296105e-01 2.92117029e-01
9.98335660e-01 -4.84489471e-01 -8.80953670e-01 1.16522767e-01
-2.24754393e-01 1.33947834e-01 2.24377126e-01 -1.38170576e+00
2.94956863e-01 -2.94905394e-01 6.75990403e-01 -7.44289041e-01
-6.41188174e-02 -1.34347498e+00 5.55450976e-01 9.19419229e-01
1.48528084e-01 5.22731245e-01 1.08696945e-01 7.13941038e-01
-3.52285713e-01 -4.54629779e-01 6.32076800e-01 -1.16524138e-01
-7.27007866e-01 1.78055376e-01 -3.78989577e-01 -5.96112669e-01
1.57271457e+00 -2.55862266e-01 -2.52992898e-01 5.11955619e-01
-6.53468132e-01 3.78176004e-01 1.81885257e-01 6.83271646e-01
7.07338810e-01 -1.18412161e+00 -3.65050226e-01 3.94830734e-01
2.97376037e-01 3.63710314e-01 4.29585367e-01 7.77525961e-01
-3.61354351e-01 2.06209674e-01 -1.96098328e-01 -8.60118330e-01
-1.10730195e+00 5.22864521e-01 4.50272441e-01 -5.27464934e-02
-6.57871604e-01 4.92499739e-01 -6.57129765e-01 -1.58146903e-01
1.31663457e-01 -4.68120843e-01 5.19214645e-02 4.97609042e-02
3.62257838e-01 7.85030842e-01 8.56917560e-01 -5.72044961e-02
-3.25400114e-01 1.77631870e-01 -8.60878155e-02 1.20622493e-01
1.06654286e+00 -1.18231528e-01 -1.06223136e-01 1.01823413e+00
5.76632679e-01 -8.36789608e-01 -1.54689026e+00 1.71449065e-01
3.69172961e-01 -3.90934408e-01 2.71698952e-01 -1.21208107e+00
-1.03603995e+00 5.63875854e-01 1.12182963e+00 3.32506299e-01
1.12184954e+00 -2.50892907e-01 3.84984493e-01 1.48120478e-01
4.29676890e-01 -1.45702767e+00 5.88607900e-02 1.58617392e-01
6.68428302e-01 -1.19150722e+00 1.30359650e-01 -3.40394706e-01
-4.85543877e-01 1.40565562e+00 1.00794733e+00 1.29967138e-01
8.31175685e-01 7.05054045e-01 1.91232041e-01 -3.60021919e-01
-9.74393964e-01 3.68468851e-01 -3.05219620e-01 7.01234937e-01
2.34343380e-01 7.45897889e-02 8.74463171e-02 3.98572832e-01
4.25935924e-01 1.62461340e-01 3.03186774e-01 1.80118334e+00
-5.48658013e-01 -1.32237637e+00 -8.89914393e-01 8.61389637e-01
-4.66379166e-01 6.25096440e-01 3.01310509e-01 6.12003207e-01
2.98025310e-01 1.51089954e+00 5.28802499e-02 -5.51172197e-01
5.78561187e-01 3.79651546e-01 5.95597148e-01 -5.94558537e-01
-3.92081141e-01 1.09609529e-01 -1.30552247e-01 -2.09752396e-01
-7.71968216e-02 -9.37771678e-01 -1.68842328e+00 2.01121494e-01
-9.89796519e-01 2.44666860e-01 9.20093477e-01 1.20304203e+00
1.63812637e-01 1.51854217e+00 9.71518397e-01 -5.90382576e-01
-9.55858946e-01 -1.23420560e+00 -1.04749823e+00 2.78338373e-01
1.91130042e-01 -1.26528883e+00 -3.48165810e-01 3.18585038e-01]
|
[6.882375717163086, 2.398475170135498]
|
93a8996f-b0eb-4b82-8e53-389e62645626
|
an-acne-grading-framework-on-face-images-via
| null | null |
https://ieeexplore.ieee.org/document/9669431/
|
https://ieeexplore.ieee.org/document/9669431
|
An Acne Grading Framework on Face Images via Skin Attention and SFNet
|
Severity level grading is a vitally important step to make correct diagnoses and personalized treatment schemes for acne, which is mainly carried out in two ways: criterion-based lesion counting and experience-based global estimation. In this paper, the global estimation of acne severity grading is studied by Convolutional Neural Networks (CNNs) and a unified acne grading framework that can diagnose referring to different grading criteria is proposed. Firstly, an adaptive image preprocessing method that can efficiently reduce the background noise and emphasize the skin information is proposed. Next, an innovative CNN structure SFNet, which fuses local skin features with global features to effectively enhance the perception of color gaps between skin and lesion, is presented. The proposed framework is verified on two datasets with different acne grading criteria. Experimental results show that the accuracy of the proposed framework reaches 84.52% exceeding the state-of-the-art method by 1.7% and reaches the diagnostic level of a professional dermatologist.
|
['Jingchi Jiang', 'Xue Cheng', 'Haiyan You', 'Zhaoyang Ma', 'Yi Guan', 'Yi Lin']
|
2022-01-14
| null | null | null |
ieee-international-conference-on-5
|
['acne-severity-grading']
|
['medical']
|
[ 2.22811550e-01 -4.72759426e-01 -1.81957528e-01 -1.73903942e-01
-6.38665199e-01 -3.24254990e-01 2.04868987e-01 2.71150798e-01
-3.56229514e-01 5.19705951e-01 2.05484666e-02 1.77020460e-01
-3.39875042e-01 -9.38087821e-01 2.62031138e-01 -9.21958089e-01
3.19656909e-01 -1.15859985e-01 1.08586289e-01 -3.67897265e-02
2.33815253e-01 6.31549060e-01 -1.56839442e+00 4.59642142e-01
1.40231526e+00 1.41669083e+00 2.28634458e-02 8.18384171e-01
-2.02969179e-01 5.99944234e-01 -7.20903099e-01 -2.11306468e-01
1.08386196e-01 -3.33650023e-01 -5.40135682e-01 1.68077961e-01
8.31036866e-01 -4.70993280e-01 1.64420961e-03 1.28946948e+00
8.02080333e-01 -1.93691462e-01 6.19487345e-01 -8.05400610e-01
-6.92742050e-01 -1.27764314e-01 -6.50935888e-01 1.01129167e-01
-1.88094750e-01 2.37914771e-01 6.33228660e-01 -4.71478462e-01
4.31403816e-01 7.75927722e-01 6.98162258e-01 4.60375130e-01
-6.28058374e-01 -4.25642520e-01 -5.42283915e-02 5.52467704e-01
-1.35827947e+00 2.18545362e-01 6.04006052e-01 -5.14123917e-01
3.38041812e-01 4.32457000e-01 1.05611157e+00 4.32802022e-01
3.38478327e-01 4.67361927e-01 1.66008806e+00 -2.50534803e-01
2.62496918e-01 -8.96426290e-02 -1.19525522e-01 8.65321696e-01
2.46718079e-01 -2.50615746e-01 -5.67164943e-02 3.02322600e-02
9.90425467e-01 1.62677705e-01 -3.58867496e-01 8.60681012e-02
-7.03206241e-01 5.44339359e-01 8.05546343e-01 4.29918408e-01
-6.97408915e-01 -5.22537716e-02 3.58889520e-01 -2.29072034e-01
4.32138503e-01 2.18681067e-01 -2.58225381e-01 1.88257575e-01
-8.05084050e-01 4.12005559e-03 4.07873988e-01 1.35815918e-01
2.82755703e-01 -2.61777136e-02 -7.29593992e-01 1.22515023e+00
3.15321283e-03 3.39677155e-01 3.46481711e-01 -7.85996556e-01
-7.32283592e-02 1.10145390e+00 -6.84336424e-02 -9.79462147e-01
-3.37725610e-01 -5.13054788e-01 -1.20920181e+00 6.82409048e-01
3.52120578e-01 -2.28523210e-01 -1.40031445e+00 1.19239926e+00
4.13696438e-01 -6.72508776e-02 -3.67200196e-01 1.10000408e+00
1.09220827e+00 3.40046287e-01 5.46149552e-01 2.30532177e-02
1.52124953e+00 -1.05645096e+00 -9.73903239e-01 1.58288524e-01
3.02844375e-01 -8.22397172e-01 8.70782375e-01 7.01006174e-01
-1.14948058e+00 -5.98848581e-01 -1.13645470e+00 -1.62263978e-02
-4.98237461e-01 9.48968947e-01 6.57425761e-01 6.29287362e-01
-1.21761799e+00 3.92550081e-01 -5.35800457e-01 -5.98244846e-01
5.44072807e-01 1.37943745e-01 -3.69625390e-01 -2.59028286e-01
-1.05039942e+00 9.56387818e-01 2.99847603e-01 3.78058612e-01
-5.55926442e-01 -8.58828008e-01 -5.90796828e-01 6.75929859e-02
1.08945243e-01 -7.65727997e-01 1.01166666e+00 -9.92203057e-01
-1.50971687e+00 6.32446766e-01 -1.05928935e-01 3.69261717e-03
5.69114804e-01 4.72539701e-02 -6.80890679e-01 6.94351375e-01
-3.05739075e-01 3.77639472e-01 6.67469561e-01 -1.09168589e+00
-9.66321349e-01 -6.18307531e-01 -6.23200163e-02 3.26463908e-01
-4.22381490e-01 -6.73044547e-02 -4.33604389e-01 -4.42522854e-01
-2.32817486e-01 -5.23280442e-01 -3.92725915e-01 6.33332253e-01
-2.24910870e-01 -2.67189056e-01 6.74622297e-01 -1.26796985e+00
1.48084641e+00 -1.68533933e+00 -2.89848477e-01 5.38707316e-01
5.76594412e-01 8.28341126e-01 -3.33191842e-01 2.30701521e-01
-8.34659711e-02 1.31585807e-01 -1.59409434e-01 1.53824374e-01
-2.78067738e-01 -2.98439384e-01 6.38627946e-01 3.22257400e-01
2.81064451e-01 7.54578173e-01 -7.03214645e-01 -4.71667379e-01
6.03638053e-01 7.81019509e-01 -1.78215101e-01 9.02125388e-02
1.18577562e-01 2.04864636e-01 -2.55786628e-01 1.04135752e+00
1.16276824e+00 -3.62018943e-01 3.63086164e-01 -6.81270719e-01
-2.31706768e-01 -3.41404885e-01 -1.13342869e+00 1.28790498e+00
-7.05903828e-01 4.39339727e-01 2.85104156e-01 -4.63182569e-01
7.77358413e-01 4.06211853e-01 4.53802168e-01 -6.62174463e-01
5.12867928e-01 2.10709408e-01 1.32990792e-01 -7.34588563e-01
6.91161007e-02 2.20544845e-01 5.64958215e-01 -2.10677862e-01
1.72574201e-03 -1.74731910e-02 3.17898721e-01 -1.12817690e-01
9.68726754e-01 -3.30516279e-01 6.17226899e-01 -1.47240505e-01
8.98163140e-01 -1.12368248e-01 4.94517207e-01 4.09556419e-01
-5.22686899e-01 4.44639474e-01 4.14544880e-01 -4.64428365e-01
-9.01254773e-01 -9.20329154e-01 -1.26602978e-01 4.72084910e-01
3.00052673e-01 -1.65588394e-01 -1.07168055e+00 -6.97461963e-01
-1.66015506e-01 1.65092513e-01 -8.62577975e-01 6.69343621e-02
5.82821555e-02 -7.73391187e-01 1.44826576e-01 7.71097243e-01
1.04258466e+00 -8.61023962e-01 -8.46358165e-02 1.44422601e-03
-1.45015761e-01 -7.02156544e-01 -1.70375034e-01 -4.46288675e-01
-5.99603415e-01 -1.33895648e+00 -1.04119349e+00 -8.40561152e-01
9.69591916e-01 2.44711474e-01 6.17625058e-01 3.99306297e-01
-1.10855317e+00 2.29941934e-01 -1.47780553e-01 -1.88979983e-01
-4.51431572e-02 -1.84568450e-01 -3.84176433e-01 2.41273195e-01
4.54760283e-01 -2.82146722e-01 -1.30741429e+00 -7.91150331e-02
-8.18687558e-01 -2.16776386e-01 1.21766782e+00 7.48092711e-01
7.40889788e-01 2.80443639e-01 4.14536923e-01 -7.98781276e-01
8.43516946e-01 -1.41819209e-01 -3.93734932e-01 4.01914656e-01
-5.65209925e-01 -6.12963259e-01 5.83357215e-01 -2.49302894e-01
-1.37709785e+00 -8.12820196e-02 -2.65037805e-01 -1.50006592e-01
-5.33159733e-01 4.03541714e-01 5.65312244e-02 -6.93010032e-01
5.72968364e-01 -2.77341288e-02 6.47210330e-02 -3.62806439e-01
2.34580606e-01 8.44997346e-01 5.40833890e-01 -1.42983451e-01
4.21705514e-01 4.62316602e-01 2.34342948e-01 -6.86491072e-01
-8.20632875e-01 -7.17698157e-01 -4.12591875e-01 -4.96810913e-01
1.09083700e+00 -8.51586223e-01 -9.38622952e-01 1.06233823e+00
-9.97372866e-01 -7.65376240e-02 6.45592138e-02 2.49439836e-01
-1.46530926e-01 5.31596720e-01 -8.35007548e-01 -7.16280222e-01
-6.85961545e-01 -8.89670193e-01 8.40182841e-01 8.30851853e-01
1.82482108e-01 -1.15721428e+00 1.27239197e-01 5.94482780e-01
6.96303189e-01 4.88500267e-01 9.45563555e-01 -1.14566669e-01
-3.12352747e-01 -5.41551709e-01 -8.51446152e-01 7.78321803e-01
4.70428467e-01 4.26088899e-01 -9.89573240e-01 -1.75612360e-01
-4.34705615e-01 -1.43224835e-01 9.00039673e-01 7.21240461e-01
1.72324431e+00 -1.99474648e-01 -8.42270628e-02 5.05242109e-01
2.12859130e+00 3.52585346e-01 9.81063128e-01 4.75604879e-03
5.16300261e-01 5.91077328e-01 5.35193264e-01 4.47840124e-01
1.62299156e-01 1.20607754e-02 6.11589432e-01 -8.61792564e-01
-5.09416044e-01 1.81670282e-02 -4.50808644e-01 8.02105904e-01
-6.02675080e-01 -2.62674332e-01 -4.98843193e-01 7.23531783e-01
-1.57480025e+00 -7.86405623e-01 -2.58920074e-01 2.02279687e+00
5.80489397e-01 -3.23757648e-01 -1.67850360e-01 9.22926590e-02
1.09784937e+00 -1.43065929e-01 -4.45820004e-01 -5.16408086e-01
4.38296422e-02 7.63132155e-01 5.46157479e-01 2.77292818e-01
-1.26985776e+00 5.92516065e-01 6.16692162e+00 1.28582668e+00
-1.36428094e+00 6.82100803e-02 7.40526199e-01 3.16767037e-01
1.60332248e-01 -4.90740716e-01 -2.44253919e-01 4.73628461e-01
8.79575983e-02 1.45438641e-01 3.88695031e-01 6.14232540e-01
3.18325400e-01 -2.61531681e-01 -2.94567585e-01 1.05741632e+00
1.25707984e-01 -1.28190958e+00 2.54756510e-01 -7.81008676e-02
9.33489978e-01 -4.83629882e-01 2.71876723e-01 -1.70811370e-01
1.46954760e-01 -1.06916118e+00 -2.44356453e-01 8.82990181e-01
1.18199837e+00 -9.27981615e-01 1.19446826e+00 -2.55251378e-01
-1.32945812e+00 -1.74389929e-01 -1.39288738e-01 2.17979476e-01
-1.48624018e-01 7.11987019e-01 -6.81017816e-01 5.99659622e-01
6.23154521e-01 4.28829819e-01 -8.25424552e-01 1.69582033e+00
-4.78811264e-01 4.48038280e-01 -3.23254615e-02 -1.83985129e-01
2.65968591e-01 -2.89606988e-01 2.38673761e-01 1.10354137e+00
4.81835991e-01 1.39860049e-01 -9.85836834e-02 7.02724397e-01
1.28674179e-01 4.87093538e-01 2.24981806e-03 -1.47600353e-01
2.63214380e-01 2.13581538e+00 -4.29711729e-01 -1.56126752e-01
-3.10239911e-01 8.43013644e-01 4.80978005e-02 3.82070512e-01
-4.95859355e-01 -9.01863456e-01 4.02687162e-01 9.49855968e-02
3.72432582e-02 2.87781179e-01 -2.13371351e-01 -7.54237592e-01
-8.35902691e-02 -6.09860182e-01 4.76931959e-01 -1.05078864e+00
-1.47341454e+00 4.54241365e-01 -6.30411446e-01 -1.23713303e+00
2.57428169e-01 -9.89924848e-01 -1.08131933e+00 9.92155433e-01
-1.68041110e+00 -1.64220762e+00 -1.13859987e+00 4.79187816e-01
3.42395723e-01 4.94012646e-02 7.07510173e-01 3.49883020e-01
-4.48234528e-01 4.39095855e-01 -2.61964686e-02 6.48040995e-02
8.35567415e-01 -1.50632894e+00 -4.62590069e-01 7.28480458e-01
-9.06028152e-01 4.90353763e-01 9.49543193e-02 -6.44725800e-01
-6.52391076e-01 -1.15380299e+00 6.52707815e-01 3.10150295e-01
5.80303907e-01 3.00322592e-01 -6.93239331e-01 -1.98120266e-01
5.03977060e-01 1.94991399e-02 8.65429699e-01 -1.44545421e-01
-2.03888834e-01 -3.64293158e-01 -1.59279943e+00 7.25923955e-01
5.57074547e-01 -2.36585140e-01 9.92656276e-02 3.13624114e-01
7.81066790e-02 -2.89443910e-01 -1.20274782e+00 7.50426888e-01
7.36139596e-01 -8.80102634e-01 7.29167819e-01 -2.79283077e-01
6.73762918e-01 -3.85085911e-01 1.05550528e-01 -1.41956377e+00
-5.84622443e-01 -7.12603331e-02 1.62388548e-01 1.06313097e+00
-8.17485824e-02 -5.42018473e-01 6.67955399e-01 1.45892933e-01
-1.78025607e-02 -1.16819429e+00 -4.99675453e-01 -1.04906090e-01
-1.57148600e-01 3.22234869e-01 3.99156064e-01 9.12731171e-01
-4.90419775e-01 -1.99796781e-01 2.45013461e-02 1.94263592e-01
6.62897408e-01 -2.26165965e-01 2.45645508e-01 -1.32958055e+00
1.91451758e-01 -6.19134665e-01 -7.25617051e-01 -2.58189052e-01
-4.97279137e-01 -4.96890604e-01 -2.37649485e-01 -2.10466099e+00
4.06220496e-01 -9.30877179e-02 -8.65456760e-01 4.98966426e-01
-3.54780167e-01 6.59280241e-01 -3.49431895e-02 -4.17207927e-01
-3.47664148e-01 8.75753239e-02 1.88054299e+00 -3.61629814e-01
-5.67293800e-02 -8.03046227e-02 -7.01749682e-01 7.86108732e-01
9.94749844e-01 2.95679420e-01 -1.91775128e-01 8.48423466e-02
8.51493105e-02 -2.21922204e-01 5.66489995e-01 -1.35511959e+00
4.23102707e-01 -2.99362838e-01 8.80066991e-01 -8.08374524e-01
2.46689036e-01 -6.35820746e-01 -1.19647644e-01 6.17223263e-01
-3.06667358e-01 -4.69063699e-01 1.43264398e-01 4.35201526e-01
-4.43117708e-01 -1.96837723e-01 1.23416018e+00 -2.26190940e-01
-9.31688488e-01 6.57185793e-01 -4.07232851e-01 -4.16807353e-01
1.11637282e+00 -1.21177576e-01 -7.12648392e-01 -3.15124243e-01
-8.23519528e-01 1.03620149e-01 1.45058066e-01 -1.92297958e-02
6.82529986e-01 -1.46743631e+00 -7.75204480e-01 -6.88205138e-02
3.78329784e-01 -2.95018882e-01 1.10407031e+00 1.06531084e+00
-1.02003157e+00 2.29949132e-01 -6.20940149e-01 -4.84795809e-01
-1.54303074e+00 1.74612328e-02 5.55215538e-01 -2.70064473e-01
1.07331410e-01 6.59671545e-01 -1.61577556e-02 -2.51921207e-01
-8.05329531e-04 -1.87602013e-01 -7.51573265e-01 -4.90828753e-02
6.63657308e-01 7.40463793e-01 4.46053058e-01 -2.95669794e-01
-2.93025877e-02 1.00139284e+00 -2.82856226e-01 3.42263550e-01
9.51614916e-01 4.28522564e-02 -5.93075395e-01 -2.76063085e-01
8.92826617e-01 1.91566199e-01 -9.07893717e-01 4.49955314e-02
-7.22354412e-01 -6.53699517e-01 5.69106281e-01 -1.71152711e+00
-1.42166555e+00 1.13421130e+00 1.43582416e+00 9.78613123e-02
1.63345850e+00 -5.30052900e-01 9.57420945e-01 5.48526645e-02
-1.97798088e-02 -1.33406472e+00 5.06801717e-02 -5.79822548e-02
7.59132504e-01 -1.04006624e+00 -4.59218249e-02 -6.33494794e-01
-4.97610807e-01 1.28501832e+00 1.07314968e+00 -3.04726511e-01
3.75697553e-01 -1.16503380e-01 4.43875581e-01 -1.01329580e-01
-3.80382836e-01 -5.36887109e-01 6.90362930e-01 7.76194870e-01
2.03153580e-01 2.14031696e-01 -8.13789308e-01 6.16020977e-01
5.40706158e-01 3.15523088e-01 2.46568590e-01 5.79682112e-01
-6.07700706e-01 -7.85097241e-01 -1.02582648e-01 8.38095903e-01
-7.16626346e-01 -2.26942860e-02 -6.63072765e-01 8.28597486e-01
6.95722878e-01 8.44372153e-01 5.08832633e-02 -3.39374304e-01
-4.42543067e-02 -3.61214429e-01 5.70155084e-01 -2.58846909e-01
-5.73895276e-01 1.24204159e-01 4.48152088e-02 -5.65342724e-01
-5.32444000e-01 8.02691281e-02 -7.77531564e-01 -1.62519440e-01
-3.10301274e-01 -3.38146538e-01 9.10709739e-01 6.09898508e-01
2.33324468e-01 7.62323737e-01 6.62220299e-01 -3.46114337e-01
-2.56152332e-01 -9.24721360e-01 -8.61200392e-01 2.73352861e-01
2.56061077e-01 -4.76366609e-01 -3.29604179e-01 -1.10806122e-01]
|
[15.695219039916992, -3.0297603607177734]
|
974c4319-0b58-4c5e-a68d-6f3153461f63
|
leveraging-context-to-support-automated-food
|
1510.02078
| null |
http://arxiv.org/abs/1510.02078v1
|
http://arxiv.org/pdf/1510.02078v1.pdf
|
Leveraging Context to Support Automated Food Recognition in Restaurants
|
The pervasiveness of mobile cameras has resulted in a dramatic increase in
food photos, which are pictures reflecting what people eat. In this paper, we
study how taking pictures of what we eat in restaurants can be used for the
purpose of automating food journaling. We propose to leverage the context of
where the picture was taken, with additional information about the restaurant,
available online, coupled with state-of-the-art computer vision techniques to
recognize the food being consumed. To this end, we demonstrate image-based
recognition of foods eaten in restaurants by training a classifier with images
from restaurant's online menu databases. We evaluate the performance of our
system in unconstrained, real-world settings with food images taken in 10
restaurants across 5 different types of food (American, Indian, Italian,
Mexican and Thai).
|
['Gregory Abowd', 'Vinay Bettadapura', 'Irfan Essa', 'Edison Thomaz', 'Aman Parnami']
|
2015-10-07
| null | null | null | null |
['food-recognition']
|
['computer-vision']
|
[ 4.07374114e-01 -5.86514831e-01 -3.04736942e-01 -5.54886580e-01
-6.69231474e-01 -9.96521354e-01 1.36003464e-01 7.70805478e-01
-4.09850627e-01 -4.71374057e-02 5.51178336e-01 -8.01252052e-02
6.06050134e-01 -9.90842044e-01 -1.08260441e+00 -3.92484337e-01
-1.67618934e-02 -2.86135197e-01 -1.75709248e-01 -8.46687704e-02
8.17227066e-02 -9.30495858e-02 -1.34502578e+00 9.28062320e-01
2.58202344e-01 8.58782887e-01 4.46664333e-01 1.04096353e+00
3.27593982e-01 6.28255069e-01 -5.57102971e-02 -3.78232896e-01
4.73188877e-01 -2.49426633e-01 -8.76735672e-02 7.39412844e-01
8.11666131e-01 -6.15224898e-01 -5.57170771e-02 1.21822011e+00
1.20294973e-01 -8.25364813e-02 3.83132130e-01 -7.79870331e-01
-1.19726181e+00 6.91454113e-01 -4.66667444e-01 3.34510565e-01
9.67495620e-01 5.76868534e-01 8.80995810e-01 -6.32204413e-01
2.68983513e-01 1.26483738e+00 7.95825183e-01 6.49581328e-02
-1.31732774e+00 -2.68544048e-01 2.97129095e-01 8.95495713e-02
-1.13731921e+00 -5.07424474e-01 5.64485431e-01 -4.27903026e-01
9.12480712e-01 2.34577626e-01 1.32534719e+00 1.08229339e+00
8.85979459e-02 9.97817099e-01 1.12919557e+00 -7.29347408e-01
3.42500418e-01 1.94887787e-01 1.77955672e-01 6.55099213e-01
6.00241542e-01 -2.95830369e-01 -1.87683195e-01 1.25773042e-01
6.19015753e-01 6.56333804e-01 -1.76754802e-01 -4.52934951e-01
-1.46030486e+00 9.98546362e-01 5.92073560e-01 9.07068700e-03
-8.28240991e-01 -1.72317445e-01 3.28152925e-01 1.20969132e-01
4.22739059e-01 -2.29124445e-02 -4.66395348e-01 3.08631957e-01
-4.85740125e-01 5.38034998e-02 1.05181241e+00 1.00112677e+00
6.66375995e-01 -4.01777416e-01 4.78205413e-01 7.47619271e-01
5.38014352e-01 8.93483579e-01 3.50248098e-01 -7.17989564e-01
5.42628050e-01 6.17265940e-01 5.50312340e-01 -1.27865756e+00
-3.46210122e-01 5.56612015e-01 -3.88449550e-01 -1.86971575e-01
7.63057411e-01 -6.20851927e-02 -6.98661804e-01 1.22984731e+00
5.23302615e-01 -3.79718155e-01 4.69431020e-02 9.68341053e-01
7.35372484e-01 6.76154196e-01 2.65658617e-01 1.54484287e-01
2.02438045e+00 -1.01509297e+00 -5.54280162e-01 -3.78088057e-01
1.62086651e-01 -9.54588771e-01 1.21452355e+00 6.10199749e-01
-7.77532160e-01 -6.79228842e-01 -1.08443439e+00 -2.05474317e-01
-6.80029392e-01 1.64515495e-01 4.42803949e-01 8.39959979e-01
-6.38798952e-01 1.40229017e-01 -8.43481958e-01 -9.63979542e-01
9.90102515e-02 5.89390546e-02 -4.01018023e-01 -6.02074504e-01
-5.11817813e-01 4.92254734e-01 3.12767059e-01 -6.15464672e-02
-6.46491289e-01 -4.12757456e-01 -1.22503948e+00 -8.78131315e-02
5.69897890e-01 -5.60334504e-01 1.61667478e+00 -1.34289515e+00
-1.20258892e+00 1.15670562e+00 2.11385980e-01 -4.11406249e-01
1.77629098e-01 -6.44208193e-01 -7.45399177e-01 4.05137777e-01
3.36319149e-01 7.98726082e-01 4.91850019e-01 -9.87749875e-01
-9.74498272e-01 -7.04799354e-01 4.36168045e-01 1.71206668e-01
2.01872706e-01 -1.48931623e-01 -1.80961356e-01 -3.69902492e-01
-1.13600090e-01 -9.59882855e-01 -1.98928311e-01 1.76736340e-01
-3.59215438e-01 2.06279397e-01 2.66589403e-01 -8.67400944e-01
7.88047254e-01 -2.50311661e+00 -6.14567816e-01 -1.39698675e-02
-4.13075358e-01 1.48064762e-01 1.05013959e-01 4.71733898e-01
3.25174540e-01 -1.80070028e-01 1.64749220e-01 2.94970304e-01
1.51437089e-01 3.24783921e-01 2.28190035e-01 5.04857421e-01
-1.66010603e-01 7.41219342e-01 -1.29024410e+00 -2.23653451e-01
5.94077170e-01 5.79694152e-01 -6.11944199e-01 -1.54966831e-01
-5.29621601e-01 2.07523078e-01 -3.70682567e-01 9.72203255e-01
5.86706519e-01 -1.04437202e-01 6.25317812e-01 -6.56354129e-01
-5.56692839e-01 2.79228032e-01 -1.24695647e+00 1.76749766e+00
-1.75072983e-01 -3.61640640e-02 4.47622716e-01 -5.48846006e-01
4.46165502e-01 2.12576225e-01 2.69152462e-01 -6.55982852e-01
-3.47466674e-03 -8.49081799e-02 -4.19169962e-01 -1.16611648e+00
3.68389726e-01 1.34875268e-01 -2.47474402e-01 4.38592613e-01
-3.79888564e-01 3.57386827e-01 4.57680792e-01 -5.23275614e-01
7.72411942e-01 3.42184454e-01 9.53213155e-01 -3.57700169e-01
1.50223225e-01 4.80622649e-01 7.75189549e-02 4.91739601e-01
-6.58088446e-01 5.19092143e-01 -6.29907072e-01 -9.77363944e-01
-1.20207274e+00 -9.81599629e-01 2.93409884e-01 1.34457743e+00
4.00538780e-02 -1.60963789e-01 -9.71573770e-01 -6.41327322e-01
1.94067478e-01 6.72523201e-01 -5.10499954e-01 2.02423304e-01
-8.16592157e-01 -7.16845989e-01 -1.39606684e-01 4.09333348e-01
8.14734399e-01 -1.06578124e+00 -1.32250845e+00 4.71358806e-01
-4.99019116e-01 -1.16196036e+00 -1.19985366e+00 1.71406828e-02
-4.75865543e-01 -1.48868322e+00 -5.81368148e-01 -9.26540613e-01
4.17522490e-01 1.15875590e+00 1.31010234e+00 -4.34210032e-01
-4.95541692e-01 1.06050873e+00 -3.88789088e-01 -6.17347062e-01
-4.22813684e-01 -1.55744031e-01 -1.30937368e-01 3.51633847e-01
1.33410561e+00 -2.07936436e-01 -1.38731730e+00 2.54599988e-01
-8.89177024e-01 -4.27028500e-02 3.33762527e-01 7.73983374e-02
7.37643778e-01 -7.33877420e-02 1.44804806e-01 -8.16895962e-01
2.24664826e-02 -7.99996793e-01 -4.07700568e-01 2.12879717e-01
-1.07641198e-01 -3.83309364e-01 6.23619437e-01 -6.16245329e-01
-6.83901787e-01 7.03972340e-01 -3.44221331e-02 2.11061969e-01
-8.49135101e-01 2.90840685e-01 -9.47343260e-02 5.61300099e-01
4.83189434e-01 -9.67934132e-02 -2.40211457e-01 -7.07936227e-01
5.73854804e-01 7.20202327e-01 4.71451789e-01 -5.68418205e-03
-4.06953692e-02 6.47619069e-01 -5.60728848e-01 -9.35089409e-01
-9.82552111e-01 -9.97210801e-01 -6.96303666e-01 -2.82422304e-01
1.22012162e+00 -1.21468937e+00 -9.06120300e-01 2.90568054e-01
-5.33557057e-01 -2.53278613e-01 -7.30028227e-02 7.56868422e-01
-4.10210520e-01 8.29149261e-02 -5.94675481e-01 -8.06435883e-01
-3.34496140e-01 -8.61712456e-01 1.02991617e+00 9.90916863e-02
-2.55395412e-01 -8.49737525e-01 3.03581022e-02 5.08943081e-01
3.40150706e-02 5.84673703e-01 4.75571722e-01 -3.09534580e-01
-4.77481693e-01 -8.51259008e-02 -1.73684925e-01 1.31723747e-01
7.47429848e-01 -8.85490105e-02 -8.58059883e-01 -1.39303848e-01
5.28100580e-02 3.52310203e-02 6.34780884e-01 7.75848925e-01
1.64374143e-01 -7.17788041e-01 -2.03950614e-01 2.22624078e-01
1.85625684e+00 5.03600001e-01 2.34648854e-01 5.82193732e-01
6.39079869e-01 5.56397855e-01 5.61458826e-01 3.63168955e-01
9.65904236e-01 6.69282496e-01 3.21287900e-01 -1.72582176e-02
-2.00348929e-01 -5.55537164e-01 9.16054249e-01 6.10097706e-01
3.37036461e-01 -1.07635766e-01 -3.47166568e-01 7.19794035e-01
-1.72956467e+00 -1.09349024e+00 -1.49490237e-01 2.11216640e+00
6.24629140e-01 -3.77893567e-01 8.97183955e-01 5.85962310e-02
9.49938774e-01 -1.65246516e-01 -6.25146747e-01 -2.84557283e-01
2.27092475e-01 -3.62092823e-01 7.39094496e-01 1.55597582e-01
-1.64844561e+00 2.19700292e-01 6.87767839e+00 -2.17293754e-01
-1.02306426e+00 7.20211565e-02 8.01793158e-01 -6.29380019e-03
3.57745439e-01 -7.34460771e-01 -6.53210878e-01 4.90376472e-01
1.07205880e+00 4.20358092e-01 8.24227095e-01 8.94009054e-01
4.54824299e-01 -4.45607483e-01 -1.39941847e+00 8.37774217e-01
3.56862247e-01 -8.19460452e-01 -2.62424290e-01 -2.04428643e-01
5.09414315e-01 -7.74498284e-03 1.41311735e-01 -8.01829696e-02
5.30408919e-01 -3.74691695e-01 1.14575112e+00 1.78519621e-01
2.81786293e-01 -4.15589005e-01 1.13765225e-01 2.94005305e-01
-1.52695239e+00 -3.08800250e-01 -3.13563943e-01 -2.13600785e-01
6.58181757e-02 4.15284991e-01 -9.06402767e-01 2.05576122e-01
1.12750304e+00 8.89774024e-01 -4.55820590e-01 9.41040337e-01
-5.04034758e-02 4.23750579e-01 -6.30730867e-01 -9.30127650e-02
4.03315365e-01 -4.00378466e-01 -6.44736411e-03 1.53475499e+00
5.41363120e-01 4.30816293e-01 6.60637796e-01 5.26761234e-01
9.02876779e-02 3.37562621e-01 -9.04666245e-01 2.72312164e-02
-1.12021074e-01 1.26224637e+00 -1.04210210e+00 -2.20905885e-01
-9.73287404e-01 9.47125256e-01 -1.70457721e-01 1.81295760e-02
-5.06587744e-01 5.72341800e-01 2.83772469e-01 1.58770934e-01
8.66802692e-01 -1.95005342e-01 3.82024884e-01 -1.20072341e+00
-5.40811755e-02 -9.91841555e-01 6.74593747e-01 -8.27706456e-01
-1.52752614e+00 -1.13630697e-01 9.07611027e-02 -8.95585716e-01
1.64809749e-01 -7.23047078e-01 -1.48073718e-01 9.18605998e-02
-1.50667131e+00 -1.24975455e+00 -3.38119745e-01 5.75255454e-01
9.64715600e-01 4.19300348e-01 9.72286463e-01 5.74599624e-01
-4.35954809e-01 5.97066246e-02 2.07266659e-01 2.94817477e-01
5.94535768e-01 -1.27153635e+00 5.10921717e-01 4.06292766e-01
6.15638375e-01 2.71223307e-01 7.12795138e-01 -5.51407695e-01
-2.02108049e+00 -1.19537723e+00 7.27020144e-01 -5.19764125e-01
3.80737066e-01 -3.08677256e-01 -2.39853755e-01 9.95160460e-01
3.48938227e-01 -2.35706300e-01 9.73468900e-01 -2.61096686e-01
-5.04437625e-01 -2.35539645e-01 -1.52556515e+00 5.07961214e-01
7.80274332e-01 -2.77003199e-01 -5.89234293e-01 4.37720090e-01
2.77164370e-01 -9.44510996e-02 -7.52888680e-01 -3.03814083e-01
1.02514863e+00 -8.37043464e-01 1.27555823e+00 -1.39302060e-01
2.77399570e-01 -2.28037760e-01 -6.63670480e-01 -1.35119784e+00
-4.58750188e-01 -4.61728275e-02 4.09867197e-01 8.47562790e-01
1.94902629e-01 -3.44927609e-01 5.15193462e-01 5.98518133e-01
-3.39130834e-02 1.40068889e-01 -2.80188113e-01 -1.85891345e-01
-5.68525851e-01 8.82316902e-02 7.19974577e-01 7.98531234e-01
1.02925211e-01 3.27366978e-01 -2.23329380e-01 3.25873524e-01
4.01175201e-01 3.24899912e-01 7.76446998e-01 -7.52208948e-01
-3.51176351e-01 9.08700228e-02 -7.51324072e-02 -1.07529461e+00
-6.87951326e-01 -6.50095522e-01 2.79733509e-01 -1.48648429e+00
4.52725053e-01 2.44320929e-02 -2.54490137e-01 4.16117996e-01
-1.82623833e-01 5.38440228e-01 4.40168321e-01 7.83846900e-02
-7.47459292e-01 -4.96545047e-01 1.08093464e+00 -4.92970049e-01
-4.35905345e-02 -8.96129236e-02 -6.97855353e-01 8.58102798e-01
7.68604398e-01 -3.14758420e-01 -7.64582157e-02 -5.88050008e-01
2.89141715e-01 -3.12356442e-01 3.16811562e-01 -1.07144475e+00
-4.29227650e-02 -1.42595381e-01 4.66230005e-01 -4.62301552e-01
1.86396405e-01 -1.35372305e+00 6.30211294e-01 6.20593369e-01
-3.95388216e-01 5.70802331e-01 -1.13209682e-02 8.60475242e-01
7.62494802e-02 -1.44853622e-01 5.00499666e-01 -1.00512040e+00
-7.74941564e-01 -2.23899171e-01 -5.38780868e-01 -3.79434794e-01
1.07860470e+00 -2.47311041e-01 -1.29623204e-01 -9.60809067e-02
-8.72610748e-01 -1.83080927e-01 6.15160406e-01 2.46253133e-01
3.13605875e-01 -1.34823287e+00 -6.14507198e-01 4.97651666e-01
2.78405577e-01 -5.83274722e-01 1.37294248e-01 5.52740097e-01
-7.84486234e-01 4.52324688e-01 -2.65485108e-01 -6.93223774e-01
-1.30280650e+00 1.32690096e+00 -1.38359100e-01 -2.35939361e-02
-9.53999698e-01 3.66996862e-02 3.78285676e-01 -1.30734414e-01
1.36582449e-01 -1.05374324e+00 -2.37940326e-01 -3.30639584e-03
9.38570917e-01 3.02313358e-01 -3.33139375e-02 -6.16024494e-01
-1.61730364e-01 5.56490302e-01 2.83180689e-03 2.89103597e-01
1.36471653e+00 -7.64265060e-01 5.85470378e-01 9.10188377e-01
9.48462963e-01 -1.04108658e-02 -1.32613897e+00 -6.16009422e-02
-1.08408377e-01 -4.49893624e-01 -1.71330452e-01 -1.08682990e+00
-9.91477251e-01 5.38541079e-01 1.01185775e+00 7.92711854e-01
1.12207627e+00 -2.14052036e-01 1.25454950e+00 2.62079746e-01
6.70129001e-01 -1.00990164e+00 -1.53505921e-01 -1.41792923e-01
3.42347711e-01 -1.69904661e+00 -1.72830671e-01 -6.21576488e-01
-5.80643833e-01 1.16561079e+00 -1.87803760e-01 -4.10483867e-01
7.49811411e-01 -6.80723488e-02 3.75437498e-01 -3.21612358e-01
-3.04133713e-01 -2.43559361e-01 3.13427225e-02 6.89228714e-01
3.76708835e-01 6.25309467e-01 1.64242059e-01 3.04616898e-01
4.05633338e-02 3.79939079e-01 2.91828215e-01 1.26304519e+00
-6.87531948e-01 -7.16847658e-01 -8.43058169e-01 3.35948169e-02
-6.78188145e-01 -1.10502683e-01 -1.31443411e-01 7.60658801e-01
4.45798010e-01 1.46904552e+00 4.87921424e-02 1.64910659e-01
6.89340591e-01 1.12647809e-01 7.38011062e-01 -6.39951527e-01
-1.06388152e+00 5.18369555e-01 4.36123848e-01 -3.32690537e-01
-1.10768843e+00 -1.02989793e+00 -6.15260065e-01 -3.92169446e-01
1.48291022e-01 -3.37216556e-01 6.29150867e-01 5.83819449e-01
1.81516349e-01 -4.51152287e-02 5.27993262e-01 -9.97169197e-01
-3.30833197e-01 -6.84585094e-01 -5.57411015e-01 9.53985691e-01
4.14699793e-01 -2.13675827e-01 -5.18939979e-02 8.81563067e-01]
|
[11.562614440917969, 4.414370059967041]
|
871a3f01-07e6-4808-9938-82e8f9113679
|
laser-neuro-symbolic-learning-of-semantic
|
2304.07647
| null |
https://arxiv.org/abs/2304.07647v1
|
https://arxiv.org/pdf/2304.07647v1.pdf
|
LASER: Neuro-Symbolic Learning of Semantic Video Representations
|
Modern AI applications involving video, such as video-text alignment, video search, and video captioning, benefit from a fine-grained understanding of video semantics. Existing approaches for video understanding are either data-hungry and need low-level annotation, or are based on general embeddings that are uninterpretable and can miss important details. We propose LASER, a neuro-symbolic approach that learns semantic video representations by leveraging logic specifications that can capture rich spatial and temporal properties in video data. In particular, we formulate the problem in terms of alignment between raw videos and specifications. The alignment process efficiently trains low-level perception models to extract a fine-grained video representation that conforms to the desired high-level specification. Our pipeline can be trained end-to-end and can incorporate contrastive and semantic loss functions derived from specifications. We evaluate our method on two datasets with rich spatial and temporal specifications: 20BN-Something-Something and MUGEN. We demonstrate that our method not only learns fine-grained video semantics but also outperforms existing baselines on downstream tasks such as video retrieval.
|
['Ser-Nam Lim', 'Mayur Naik', 'David Jacobs', 'Ziyang Li', 'Jiani Huang']
|
2023-04-15
| null | null | null | null |
['video-captioning', 'video-retrieval', 'video-understanding']
|
['computer-vision', 'computer-vision', 'computer-vision']
|
[ 2.62780994e-01 -1.89630941e-01 -6.37880206e-01 -7.66825736e-01
-7.36415446e-01 -7.24970579e-01 6.47923827e-01 -3.25771905e-02
-2.11270988e-01 2.64123827e-01 8.38655710e-01 -1.03569277e-01
9.45473537e-02 -3.93541604e-01 -1.34483480e+00 -9.09042805e-02
-1.71401888e-01 4.19131935e-01 2.58038312e-01 -1.36415418e-02
1.00025788e-01 -7.56204426e-02 -1.70160007e+00 8.89824986e-01
3.69592726e-01 1.30597508e+00 3.25488418e-01 7.74727404e-01
1.51421979e-01 1.14305425e+00 -1.29085332e-01 -3.04233730e-01
2.66598724e-02 -1.87802240e-01 -8.55331898e-01 2.13161170e-01
9.62678909e-01 -6.52192175e-01 -7.18660235e-01 7.76903212e-01
-1.21273778e-01 1.00589067e-01 5.36165893e-01 -1.43609154e+00
-9.83929098e-01 7.12784708e-01 -1.15529768e-01 2.97343284e-01
7.05643713e-01 5.06868780e-01 1.39049149e+00 -7.15790987e-01
8.28301847e-01 1.52722371e+00 5.51078737e-01 7.31692731e-01
-1.23361027e+00 -5.60028672e-01 6.91727579e-01 4.67377126e-01
-1.25623906e+00 -7.35244155e-01 5.46988130e-01 -7.04618752e-01
1.18771422e+00 5.94699606e-02 7.63071656e-01 1.61484516e+00
-1.53978914e-01 9.72756982e-01 4.60012823e-01 1.64963171e-01
9.65212360e-02 -2.36459047e-01 7.48147070e-02 9.52697217e-01
-1.40344361e-02 -2.10983101e-02 -1.02718985e+00 1.73882276e-01
9.53109801e-01 4.66780439e-02 -3.86586756e-01 -6.88822806e-01
-1.65857720e+00 6.54960454e-01 3.17096174e-01 -2.30986997e-02
-3.45389545e-01 9.30057287e-01 7.88094878e-01 2.56824374e-01
2.03230321e-01 4.72671986e-01 -5.39391935e-01 -4.05405492e-01
-9.02679980e-01 4.62333024e-01 6.49895430e-01 1.39915717e+00
6.39121473e-01 7.24431202e-02 -5.38374901e-01 3.46334189e-01
2.86366642e-01 4.09277737e-01 5.14804661e-01 -1.29456365e+00
5.49178243e-01 1.72245726e-01 1.42855152e-01 -9.80964839e-01
-6.74828663e-02 1.55567795e-01 -2.08815292e-01 -4.39444333e-01
9.47278067e-02 2.53776848e-01 -1.21003199e+00 2.09801435e+00
-1.86029613e-01 9.55054283e-01 1.34092465e-01 1.32293940e+00
9.41362143e-01 6.64373994e-01 3.18290323e-01 2.52553195e-01
1.35083020e+00 -1.12531102e+00 -6.21960461e-01 -4.31212455e-01
5.71699560e-01 -1.19708441e-01 1.39931357e+00 7.23863542e-02
-1.22923946e+00 -5.85522175e-01 -9.23932612e-01 -4.62419361e-01
-1.50379971e-01 -3.41014683e-01 6.54519975e-01 -1.41998883e-02
-1.33326197e+00 4.63255554e-01 -1.01871634e+00 -4.06034470e-01
5.81397772e-01 2.77976424e-01 -5.52998245e-01 -3.24461132e-01
-1.13080525e+00 5.03037393e-01 6.44631684e-01 -1.03411019e-01
-1.36854637e+00 -8.89267147e-01 -1.51627755e+00 2.36021861e-01
7.92251587e-01 -9.96325970e-01 1.43629992e+00 -1.38012075e+00
-1.26357353e+00 8.05088580e-01 -3.30271542e-01 -6.51349962e-01
7.98070878e-02 -4.77479517e-01 -2.74706721e-01 5.59361160e-01
2.80247837e-01 1.00280762e+00 1.09051871e+00 -8.56585383e-01
-4.95716214e-01 -1.12269551e-01 4.47999418e-01 1.51571453e-01
-1.16906755e-01 5.37434891e-02 -9.59556341e-01 -8.47721517e-01
-1.83315873e-01 -7.99192309e-01 1.40866995e-01 2.25820884e-01
-1.31445035e-01 -1.87765390e-01 9.38697398e-01 -6.54306233e-01
1.03933597e+00 -2.21522450e+00 6.35027826e-01 -2.15394273e-02
1.97878957e-01 -6.95867836e-02 -4.07429338e-01 1.51974961e-01
-9.89330634e-02 2.98506051e-01 1.58921883e-01 -1.27556130e-01
4.00420189e-01 2.32671589e-01 -7.05680132e-01 2.50292838e-01
6.38564587e-01 1.32223666e+00 -1.17222822e+00 -4.28573519e-01
2.40523472e-01 3.99993479e-01 -1.03001714e+00 5.12264907e-01
-1.02806413e+00 7.06290826e-02 -5.12685001e-01 8.68767917e-01
-2.44178157e-02 -4.48970407e-01 1.43142566e-01 -7.53848374e-01
3.48101944e-01 3.01570624e-01 -6.28082573e-01 2.42829347e+00
-3.39216024e-01 8.79680872e-01 -2.21354321e-01 -1.18826568e+00
2.75214881e-01 4.21779841e-01 3.62364829e-01 -6.31693542e-01
-1.52074605e-01 -2.74644028e-02 -5.71200907e-01 -9.43098366e-01
4.21723157e-01 8.73634592e-02 -5.75085878e-01 2.34665558e-01
3.59572411e-01 -1.32766739e-01 1.14518940e-01 4.14923966e-01
1.18649852e+00 8.45756471e-01 1.07207134e-01 -1.05839349e-01
9.84677151e-02 1.97804391e-01 5.85291743e-01 7.40252554e-01
-7.01487809e-02 4.77781713e-01 7.47192025e-01 -6.36932313e-01
-1.12112093e+00 -1.14255917e+00 5.19124866e-01 1.48985326e+00
3.77384245e-01 -9.01504397e-01 -6.86137140e-01 -4.15590137e-01
8.21325034e-02 5.43297529e-01 -4.27553743e-01 -3.34553033e-01
-6.36759043e-01 3.12462807e-01 5.97382545e-01 9.05514419e-01
3.48120123e-01 -1.01786852e+00 -6.69981956e-01 1.64838359e-01
-4.38608527e-01 -1.93217719e+00 -8.00922692e-01 -2.46008158e-01
-7.03271627e-01 -1.06711674e+00 -1.82829112e-01 -6.75436199e-01
4.45155382e-01 2.60572195e-01 1.61561418e+00 1.58737481e-01
-1.53284863e-01 8.04403126e-01 -4.05027002e-01 3.19804094e-04
-8.00956562e-02 -2.25226849e-01 2.13491857e-01 -9.89232138e-02
5.83165526e-01 -3.05491596e-01 -4.32204783e-01 1.74080774e-01
-1.01434553e+00 3.12279671e-01 4.10091043e-01 6.89747334e-01
6.15786314e-01 -4.01844203e-01 3.69665027e-01 -7.90518939e-01
3.14996839e-01 -5.38221836e-01 -5.53982317e-01 3.85230035e-01
7.79241882e-03 3.82841259e-01 5.20426989e-01 -5.06394267e-01
-7.82507300e-01 1.32475600e-01 3.62068951e-01 -1.34057724e+00
-2.36118644e-01 5.96202612e-01 -2.82603621e-01 4.39923018e-01
2.78732121e-01 1.29268646e-01 -1.74140662e-01 -6.36756867e-02
6.73557937e-01 3.67679864e-01 1.03545344e+00 -9.86471713e-01
6.93207324e-01 4.63302821e-01 -3.41247767e-01 -7.12270916e-01
-1.02158546e+00 -3.52542579e-01 -4.54737782e-01 -3.87923867e-02
1.21093976e+00 -1.50483263e+00 -8.10908258e-01 1.32999614e-01
-1.16389978e+00 -5.70684731e-01 -7.57940784e-02 4.85762179e-01
-1.26715684e+00 1.89375207e-01 -5.64095199e-01 -2.07643092e-01
3.29951271e-02 -1.36556029e+00 1.52830923e+00 -5.36882952e-02
-6.88332498e-01 -7.75167286e-01 -4.99070048e-01 4.05231595e-01
2.31023759e-01 2.11017385e-01 9.55753922e-01 -3.82583946e-01
-1.16984725e+00 1.94563001e-01 -2.86181569e-01 1.52543327e-02
-1.61790296e-01 2.37651579e-02 -7.60903835e-01 -2.06602737e-01
-3.54103386e-01 -8.51175964e-01 1.02648818e+00 3.32606673e-01
1.66859877e+00 -7.25855827e-01 -2.89438605e-01 9.70142424e-01
1.29322946e+00 -1.99847460e-01 4.76753324e-01 1.86076760e-01
9.96093333e-01 4.34423238e-01 7.03073680e-01 2.62022406e-01
6.72062457e-01 7.78835475e-01 6.61157072e-01 3.70165616e-01
-7.67061263e-02 -6.46246791e-01 5.96439481e-01 3.91036838e-01
2.04767421e-01 -3.04571480e-01 -6.45986080e-01 6.85115695e-01
-2.36912417e+00 -1.38658035e+00 5.98834336e-01 1.75315738e+00
8.65912139e-01 1.83167756e-01 1.76795244e-01 -4.22550738e-01
4.31090832e-01 5.07731318e-01 -7.31505275e-01 -4.10693556e-01
3.12626660e-02 -2.59703547e-02 4.51547652e-01 4.01175708e-01
-1.26381683e+00 1.41703141e+00 5.89888334e+00 3.91863376e-01
-1.15494239e+00 -2.49059182e-02 4.08374488e-01 -4.50545788e-01
-6.63335025e-01 -6.47637323e-02 -4.41505045e-01 5.04657269e-01
1.04512358e+00 9.14499164e-02 6.13081574e-01 6.96151257e-01
2.54462451e-01 2.27143511e-01 -1.97136784e+00 1.21486461e+00
2.84737080e-01 -1.85411429e+00 3.73746216e-01 -3.21228921e-01
5.15728891e-01 1.25873443e-02 -3.40342186e-02 3.98893625e-01
2.65069962e-01 -1.38111460e+00 1.30482078e+00 6.54589176e-01
1.12413490e+00 -1.93277076e-01 2.53527492e-01 -1.85121391e-02
-1.24482393e+00 -2.23284528e-01 -1.24445662e-01 -1.24470433e-02
2.74014324e-01 -1.02366500e-01 -6.74112916e-01 7.65784010e-02
7.25583434e-01 1.36502659e+00 -2.19419450e-01 6.30271256e-01
-1.06416494e-01 4.33865279e-01 -1.27141446e-01 2.29264628e-02
5.09285569e-01 1.81452185e-01 3.71100932e-01 1.29334855e+00
2.17795923e-01 1.35068744e-01 4.25016999e-01 9.66307521e-01
-3.60634387e-01 -5.81031621e-01 -8.43825281e-01 -5.95119357e-01
5.51148772e-01 6.04192793e-01 -2.41943628e-01 -6.27509356e-01
-7.21799254e-01 1.16436195e+00 2.86879659e-01 6.87363625e-01
-1.17263401e+00 1.62220508e-01 1.33046818e+00 1.15593396e-01
7.43749976e-01 -2.71227092e-01 3.01092807e-02 -1.56134629e+00
2.35659048e-01 -1.07627690e+00 3.69095922e-01 -1.25893605e+00
-8.49137366e-01 3.71642530e-01 4.06331867e-01 -9.77926731e-01
-8.48901272e-01 -7.65585661e-01 -2.92081326e-01 3.77014577e-01
-1.34066415e+00 -1.37159586e+00 -4.43103492e-01 7.38681138e-01
1.05330288e+00 -1.41268656e-01 8.43481302e-01 1.48660794e-01
-2.21908778e-01 3.91801000e-01 -5.05387008e-01 2.55846411e-01
5.11655569e-01 -9.21151936e-01 4.65264350e-01 7.85661876e-01
3.50523770e-01 5.56952775e-01 8.66190076e-01 -4.70277280e-01
-1.81423533e+00 -1.23840952e+00 6.97776854e-01 -7.73549914e-01
8.33741784e-01 -6.15551233e-01 -6.82338476e-01 1.28524292e+00
7.67979622e-02 2.34895796e-01 4.83783424e-01 6.46540076e-02
-1.16408825e+00 -1.16279788e-01 -7.06727564e-01 7.95883954e-01
1.42418683e+00 -1.25233173e+00 -8.62252176e-01 3.50901186e-01
1.33176076e+00 -4.74610955e-01 -7.95064211e-01 4.11225945e-01
7.69591749e-01 -7.89120257e-01 1.22376692e+00 -1.30129302e+00
6.48445189e-01 -3.11357707e-01 -5.07322431e-01 -9.98100877e-01
-1.78981021e-01 -6.29534662e-01 -4.29685533e-01 8.15471590e-01
2.02425897e-01 2.25370854e-01 7.08490074e-01 9.51485932e-01
-2.60513663e-01 -5.99450827e-01 -5.55811167e-01 -7.06184626e-01
-4.42457676e-01 -8.13928127e-01 6.47219896e-01 7.71169960e-01
1.56423122e-01 3.22912544e-01 -4.01527852e-01 2.53581703e-01
4.40399945e-01 2.71082133e-01 7.21493065e-01 -8.55192006e-01
-2.92305380e-01 -5.08687675e-01 -8.57841730e-01 -1.35810924e+00
7.92558789e-01 -5.92566490e-01 1.64046794e-01 -1.43097198e+00
2.56130815e-01 9.83961970e-02 -2.24621922e-01 5.56430995e-01
2.58916706e-01 2.09163547e-01 1.47928342e-01 3.65182236e-02
-1.15507734e+00 3.36658597e-01 8.36164534e-01 -3.64384890e-01
2.38919511e-01 -7.17842698e-01 -7.13854790e-01 7.72589445e-01
4.02642339e-01 -6.65584877e-02 -7.41972446e-01 -1.16166472e+00
3.27977687e-01 2.44344443e-01 9.12030578e-01 -7.56801605e-01
1.38024971e-01 -4.44773763e-01 1.54652476e-01 -3.15218180e-01
7.17132032e-01 -8.80962074e-01 7.23195076e-02 3.77532206e-02
-7.75043488e-01 1.81582302e-01 9.38499644e-02 8.42269540e-01
-3.78800720e-01 2.64427751e-01 2.96591192e-01 -2.76138812e-01
-1.52184737e+00 6.41867816e-01 -7.99428448e-02 2.39901856e-01
1.01886296e+00 -3.16169828e-01 -4.32566315e-01 -6.03141606e-01
-6.39108121e-01 4.39259231e-01 8.14144373e-01 8.84141028e-01
8.91408145e-01 -1.41857803e+00 -3.81393373e-01 1.39973953e-01
4.98193055e-01 6.83447346e-02 3.54586802e-02 5.49806833e-01
-4.90981042e-01 7.00150132e-01 -3.24679434e-01 -7.62016475e-01
-9.77533221e-01 7.75380969e-01 1.70673177e-01 3.11976612e-01
-6.67813540e-01 1.01648510e+00 5.58693290e-01 2.94138566e-02
5.97564995e-01 -9.35228288e-01 2.25982741e-01 -3.40338141e-01
5.15979588e-01 -2.57561952e-01 -5.32767892e-01 -8.12194288e-01
-5.47002554e-01 6.19971812e-01 2.37148646e-02 -6.78263977e-03
1.26287866e+00 -1.47865087e-01 -2.21279822e-02 3.12602282e-01
1.49713838e+00 -6.23483181e-01 -1.72749829e+00 -3.71812552e-01
2.38039896e-01 -6.16443694e-01 -1.62641481e-01 -4.98476058e-01
-8.57476711e-01 7.17481017e-01 -7.13047013e-03 -2.90987968e-01
1.09691131e+00 4.08670396e-01 8.42536330e-01 6.15721107e-01
2.29575738e-01 -9.42638457e-01 5.59606433e-01 5.13731837e-01
7.82946050e-01 -1.27793825e+00 -3.52206975e-01 -2.92586714e-01
-8.36084008e-01 1.03369856e+00 8.34861159e-01 -2.55053878e-01
3.26488972e-01 1.31286472e-01 -2.65889198e-01 -3.45739633e-01
-1.33970785e+00 -1.82485253e-01 5.32475829e-01 6.29880786e-01
2.72734314e-01 -8.07311907e-02 3.30337703e-01 7.55236745e-01
5.58417179e-02 3.45034450e-01 3.74830484e-01 7.06014276e-01
-3.58348072e-01 -5.87452352e-01 1.73698425e-01 4.34025407e-01
-3.33980203e-01 -2.07159087e-01 -1.54232562e-01 5.75287163e-01
-8.05577263e-02 5.52307606e-01 4.69748318e-01 -4.77384418e-01
1.11895293e-01 2.90823784e-02 4.44561213e-01 -7.26336420e-01
5.23982234e-02 -3.86195723e-04 3.51822317e-01 -1.39024091e+00
-7.14385390e-01 -5.38074315e-01 -1.38594759e+00 -9.67924297e-02
3.80719662e-01 -7.26840347e-02 4.12150413e-01 1.13935733e+00
4.55485374e-01 3.84828448e-01 2.59015393e-02 -9.15032327e-01
-3.66206229e-01 -4.56098557e-01 -9.87809598e-02 8.76509666e-01
6.51740074e-01 -6.65999949e-01 2.35881805e-02 6.28878057e-01]
|
[9.967022895812988, 0.8584257960319519]
|
9bb265ca-5994-4dec-9036-f032ba8d982c
|
bilingual-gan-a-step-towards-parallel-text
|
1904.04742
| null |
https://arxiv.org/abs/1904.04742v2
|
https://arxiv.org/pdf/1904.04742v2.pdf
|
Bilingual-GAN: A Step Towards Parallel Text Generation
|
Latent space based GAN methods and attention based sequence to sequence models have achieved impressive results in text generation and unsupervised machine translation respectively. Leveraging the two domains, we propose an adversarial latent space based model capable of generating parallel sentences in two languages concurrently and translating bidirectionally. The bilingual generation goal is achieved by sampling from the latent space that is shared between both languages. First two denoising autoencoders are trained, with shared encoders and back-translation to enforce a shared latent state between the two languages. The decoder is shared for the two translation directions. Next, a GAN is trained to generate synthetic "code" mimicking the languages' shared latent space. This code is then fed into the decoder to generate text in either language. We perform our experiments on Europarl and Multi30k datasets, on the English-French language pair, and document our performance using both supervised and unsupervised machine translation.
|
['Mehdi Rezagholizadeh', 'Alan Do-Omri', 'Qun Liu', 'Ahmad Rashid', 'Md. Akmal Haidar']
|
2019-04-09
|
bilingual-gan-a-step-towards-parallel-text-1
|
https://aclanthology.org/W19-2307
|
https://aclanthology.org/W19-2307.pdf
|
ws-2019-6
|
['unsupervised-machine-translation']
|
['natural-language-processing']
|
[ 5.56655049e-01 4.33928847e-01 -2.76021183e-01 -2.08103180e-01
-1.22500563e+00 -8.51789176e-01 1.24967217e+00 -7.83822894e-01
1.44881696e-01 1.24016225e+00 7.21392453e-01 -4.86746401e-01
7.27008700e-01 -8.53357196e-01 -9.58598018e-01 -6.55476511e-01
6.05855107e-01 9.42401648e-01 -7.20534086e-01 -2.47615099e-01
-2.16026917e-01 -9.84886289e-02 -6.73011482e-01 4.99533594e-01
1.05622339e+00 1.51823521e-01 1.83881953e-01 7.96888590e-01
-2.19384223e-01 8.25784445e-01 -4.87485290e-01 -7.14323580e-01
5.51919460e-01 -1.29280305e+00 -8.36404502e-01 3.03686298e-02
-2.62345616e-02 -3.37488621e-01 -3.23319227e-01 9.81844068e-01
4.67724472e-01 -3.16406250e-01 7.23505318e-01 -1.19449604e+00
-1.06538677e+00 8.84843946e-01 -2.13344827e-01 -3.16819131e-01
4.27883357e-01 5.68149209e-01 8.37059200e-01 -7.29600847e-01
9.47596788e-01 1.16298330e+00 1.99021623e-01 1.21235728e+00
-1.60273290e+00 -7.46374428e-01 -2.70628780e-01 -5.92409134e-01
-9.01459098e-01 -7.37649083e-01 7.88016021e-01 -5.73906839e-01
1.03350282e+00 1.07934959e-01 6.40300989e-01 1.92128432e+00
4.22202051e-01 6.63211942e-01 1.31918621e+00 -6.95404053e-01
2.19847798e-01 1.40052542e-01 -7.46181011e-01 3.34291518e-01
-2.43118331e-01 3.02483618e-01 -6.86993718e-01 -1.44624159e-01
7.90280879e-01 -3.38394254e-01 -4.86892611e-02 -2.07273468e-01
-1.57529688e+00 1.12801373e+00 3.53635624e-02 2.43744403e-01
-4.59177285e-01 3.92279744e-01 4.16928858e-01 7.62802899e-01
7.90333867e-01 5.33848763e-01 -2.76831269e-01 -3.00902873e-01
-1.08011138e+00 1.34437576e-01 8.39365721e-01 1.31972861e+00
5.65930605e-01 4.39712286e-01 -5.41680098e-01 7.04019845e-01
2.99162090e-01 8.52947533e-01 8.06558073e-01 -6.12780392e-01
9.98585284e-01 2.57573575e-01 6.86827525e-02 -4.05734897e-01
5.07556260e-01 -3.10227752e-01 -1.07609010e+00 -3.72460373e-02
4.24845964e-02 -4.13354725e-01 -1.08822334e+00 1.98200572e+00
-1.00507796e-01 4.69446694e-03 6.52408183e-01 5.37635446e-01
1.68897852e-01 9.59517002e-01 -1.85570225e-01 -2.59676009e-01
1.06250989e+00 -1.35966980e+00 -7.95976222e-01 -4.46982533e-01
6.45445406e-01 -9.74754572e-01 9.96974468e-01 -1.96852610e-01
-1.43231571e+00 -4.62092608e-01 -9.55744386e-01 -1.28381535e-01
-8.75567347e-02 3.09901506e-01 3.25125694e-01 4.13210034e-01
-1.31745338e+00 2.93628693e-01 -1.01774406e+00 -2.06588089e-01
1.91924676e-01 1.94026649e-01 -3.28374416e-01 1.25744507e-01
-1.36518645e+00 7.90114999e-01 2.58261532e-01 6.57957094e-03
-1.49901772e+00 -2.55778581e-01 -8.79548788e-01 -1.05136648e-01
-4.07827258e-01 -1.33319664e+00 1.24180388e+00 -1.65278244e+00
-2.12144113e+00 9.57493961e-01 -3.21269363e-01 -8.01035702e-01
1.02684593e+00 -1.48478493e-01 -4.68263179e-01 -2.60958821e-01
6.73792839e-01 8.49029362e-01 1.04019392e+00 -1.02778161e+00
-1.77724123e-01 1.12535380e-01 -4.00663316e-01 4.05320168e-01
3.07137053e-02 2.43350357e-01 -2.85835236e-01 -9.62263167e-01
-1.67812854e-01 -1.21311367e+00 -2.36069232e-01 -5.63009501e-01
-7.98120499e-01 2.19626293e-01 4.60137159e-01 -1.10298538e+00
6.77043676e-01 -1.85927975e+00 7.89265156e-01 3.43962386e-02
-2.47768655e-01 -8.45356435e-02 -5.24416327e-01 7.24800348e-01
-2.53999323e-01 2.49578971e-02 -4.22737509e-01 -6.82615340e-01
-2.54811682e-02 2.92023599e-01 -8.65850747e-01 3.30619328e-02
2.94001222e-01 1.44962418e+00 -8.94030631e-01 -7.78070837e-02
-2.30219409e-01 3.62009227e-01 -4.89223033e-01 6.46763444e-01
-5.87572455e-01 1.01255500e+00 -1.92893162e-01 1.21061280e-01
3.18453282e-01 -1.38856798e-01 5.53558171e-01 4.31631982e-01
5.86325042e-02 7.28728414e-01 -3.50888789e-01 2.10069871e+00
-7.00626850e-01 6.51150763e-01 -2.52379090e-01 -7.34072745e-01
1.10530269e+00 8.80443215e-01 -9.22590494e-02 -7.04927325e-01
-1.23117067e-01 4.15116787e-01 -2.17365056e-01 -1.33547664e-01
2.21173093e-01 -4.09144759e-01 -2.71019757e-01 8.78358603e-01
2.77687341e-01 -3.80011469e-01 8.38571936e-02 2.80008703e-01
8.76646519e-01 5.29421806e-01 6.14871979e-02 -2.41941303e-01
4.35518622e-01 4.02310379e-02 3.50807786e-01 5.27326167e-01
3.29859465e-01 5.98272681e-01 5.95385432e-01 -4.02490973e-01
-1.76509213e+00 -1.24481869e+00 7.10165441e-01 7.51832545e-01
-4.86843497e-01 -3.67828041e-01 -9.05430079e-01 -8.57819796e-01
-4.46854740e-01 1.19812703e+00 -6.94294453e-01 -3.18027794e-01
-8.39301467e-01 -3.74782264e-01 8.21475089e-01 4.97441590e-01
3.72924685e-01 -1.15722573e+00 1.91626579e-01 2.73231745e-01
-5.83443522e-01 -8.80966663e-01 -8.71021748e-01 5.80662116e-02
-8.35591435e-01 -3.90156388e-01 -1.01038587e+00 -9.77230012e-01
8.52078557e-01 -3.32924306e-01 1.35248435e+00 -4.67884183e-01
3.28975856e-01 -3.43385935e-02 -1.17341084e-02 -1.40217885e-01
-1.43406153e+00 3.77128690e-01 2.13568747e-01 1.65218696e-01
1.23567976e-01 -7.15327024e-01 -3.85606378e-01 3.73952277e-02
-7.73924112e-01 1.02516615e+00 4.98739392e-01 1.14358962e+00
3.39988887e-01 -6.36572003e-01 5.41065156e-01 -7.91895449e-01
7.86415756e-01 -4.02416676e-01 -6.24738872e-01 3.01735610e-01
-4.26875561e-01 5.52626491e-01 9.12107944e-01 -2.93933868e-01
-1.16503870e+00 -2.51360070e-02 -1.22329541e-01 -2.26567283e-01
-5.67436740e-02 2.93364376e-01 -3.36850137e-01 6.85763001e-01
5.66253603e-01 8.22577059e-01 6.66389018e-02 -3.22310209e-01
6.82349265e-01 7.23309696e-01 5.02722979e-01 -5.95546782e-01
1.16608775e+00 2.76857972e-01 -4.31429178e-01 -1.41294211e-01
-4.73773986e-01 2.75662005e-01 -6.85245037e-01 2.78524339e-01
1.27522123e+00 -1.35763049e+00 1.82885647e-01 4.03845698e-01
-1.53435373e+00 -7.24320829e-01 -6.24731779e-01 4.27779555e-01
-1.04390764e+00 -1.52485475e-01 -7.66913891e-01 -3.56751293e-01
-7.22889304e-01 -1.30456173e+00 1.24523640e+00 -1.20904326e-01
-3.53312403e-01 -1.15227532e+00 6.64377630e-01 3.20602328e-01
5.33134818e-01 6.42961636e-02 9.63173509e-01 -5.30404031e-01
-6.81803226e-01 -1.61515139e-02 2.28470609e-01 4.28546280e-01
3.89656901e-01 -3.47563893e-01 -6.42834365e-01 -5.61868429e-01
1.22965537e-01 -3.79661798e-01 5.33827841e-01 3.46246525e-03
3.71675104e-01 -7.11117089e-01 -2.68739939e-01 8.61591876e-01
1.20852029e+00 1.35409549e-01 7.43331969e-01 1.02506742e-01
7.06138372e-01 4.12302971e-01 -7.33064786e-02 -1.56764731e-01
2.22307354e-01 6.79474592e-01 -1.43210992e-01 -1.81629524e-01
-1.65576667e-01 -9.35028255e-01 9.51359689e-01 1.49860394e+00
1.19286463e-01 -3.32705736e-01 -8.83473635e-01 5.94362795e-01
-1.71696854e+00 -1.01383018e+00 2.67906010e-01 1.90118611e+00
1.20947266e+00 -4.14876528e-02 -1.51116207e-01 -4.96930957e-01
6.86587334e-01 6.29182011e-02 -2.79836416e-01 -6.09002650e-01
-2.92688191e-01 3.83114308e-01 3.45381469e-01 7.34340012e-01
-6.69898927e-01 1.36554313e+00 6.32554483e+00 6.74980223e-01
-1.28341556e+00 4.58298534e-01 8.33268881e-01 -1.61772713e-01
-9.29518282e-01 4.38777119e-01 -5.54405391e-01 7.07233310e-01
1.21388125e+00 -3.22193861e-01 8.40475321e-01 4.84019697e-01
3.21788967e-01 6.66939139e-01 -1.15226591e+00 6.57657862e-01
8.29634815e-02 -1.61166537e+00 4.86770123e-01 4.13616002e-02
1.45193446e+00 1.24845110e-01 1.18831843e-01 3.20412695e-01
1.02087212e+00 -1.22613561e+00 8.08355749e-01 5.93119621e-01
1.42663753e+00 -6.18192375e-01 3.90199721e-01 5.77688754e-01
-7.29362845e-01 3.88547033e-01 -1.00174516e-01 8.28621835e-02
3.86680275e-01 2.39482045e-01 -9.01827276e-01 6.28614962e-01
2.48783920e-02 5.80025494e-01 -1.96912989e-01 -1.17894918e-01
-8.22425783e-01 8.75026822e-01 4.21917252e-02 3.71053934e-01
2.73785025e-01 -6.42218649e-01 6.91373527e-01 9.70392406e-01
5.78673661e-01 -4.74911422e-01 -3.70058231e-02 1.33283305e+00
-4.20562327e-01 -4.45970185e-02 -1.03340495e+00 -5.24214923e-01
4.64411639e-02 9.34271872e-01 -3.98026884e-01 -7.25481808e-01
-3.07634532e-01 1.99297392e+00 2.38600701e-01 7.14541674e-01
-8.17836225e-01 -8.16928372e-02 4.76471841e-01 -2.09903881e-01
-3.46336775e-02 -4.13029879e-01 -3.59056592e-01 -1.75454187e+00
-7.93104097e-02 -1.29029489e+00 -1.52528644e-01 -9.71304953e-01
-1.10723746e+00 9.80663776e-01 -4.15047199e-01 -1.22773933e+00
-1.01248074e+00 1.93012673e-02 -7.90311933e-01 1.75461304e+00
-9.75087285e-01 -1.72781742e+00 2.54782051e-01 4.50297415e-01
7.67116666e-01 -9.04116452e-01 1.20493150e+00 -4.53964435e-03
-2.63527364e-01 6.59622967e-01 5.41137874e-01 3.66433412e-01
6.04496658e-01 -1.30055153e+00 1.39430559e+00 1.06551731e+00
3.55175972e-01 7.46334016e-01 5.12650013e-01 -1.05387974e+00
-1.16127288e+00 -1.40087390e+00 1.46894932e+00 -6.59164071e-01
3.87155056e-01 -9.92768168e-01 -3.43969613e-01 1.17710507e+00
8.57022107e-01 -4.46093291e-01 6.39491022e-01 -4.70355183e-01
-4.18715805e-01 4.93702233e-01 -7.67636240e-01 9.82925177e-01
9.41428781e-01 -9.30699587e-01 -4.47409958e-01 6.67426467e-01
1.07767916e+00 -3.81676465e-01 -5.06319761e-01 -8.77084658e-02
3.73064697e-01 -4.26369309e-01 6.01284742e-01 -1.05889904e+00
1.15408623e+00 -1.68525681e-01 -1.61836624e-01 -1.76785433e+00
-1.31392360e-01 -1.20394826e+00 -3.56045291e-02 1.29241621e+00
7.44974077e-01 -6.62400365e-01 9.06480670e-01 5.73895536e-02
3.42378207e-02 -3.89198095e-01 -9.20256555e-01 -7.34994769e-01
5.35651982e-01 -3.11755352e-02 8.59196782e-01 1.04614341e+00
-3.78658324e-01 9.12852049e-01 -9.13715065e-01 -2.45221317e-01
3.12064826e-01 2.15636015e-01 1.14811707e+00 -3.60246927e-01
-7.39830077e-01 -1.46196857e-01 2.57024705e-01 -1.12759352e+00
4.66724902e-01 -1.45588434e+00 3.55063416e-02 -1.46514654e+00
2.44634435e-01 -5.69909848e-02 2.08411828e-01 4.11888301e-01
-7.61559755e-02 3.97540540e-01 4.78239283e-02 6.05073512e-01
4.19121720e-02 8.73897851e-01 1.17792976e+00 -3.26425731e-01
-1.47471875e-01 -4.69606072e-02 -6.78741753e-01 3.02095830e-01
8.80948901e-01 -5.30379117e-01 -3.91044527e-01 -8.55876029e-01
4.01407599e-01 4.53549385e-01 1.25834018e-01 -6.79562271e-01
-1.06192388e-01 -1.56022489e-01 3.75601381e-01 -1.36939526e-01
1.10343464e-01 -4.95555371e-01 5.07370353e-01 5.79053283e-01
-8.31332862e-01 3.46043378e-01 -1.29242465e-01 4.14503038e-01
-2.36242712e-01 1.65326953e-01 6.60395563e-01 -2.59198397e-01
2.05416411e-01 2.06492484e-01 -3.85639995e-01 1.04533352e-01
7.98001289e-01 9.63355675e-02 9.42725390e-02 -7.02811956e-01
-8.79919887e-01 4.79931533e-02 7.55915225e-01 7.12139428e-01
3.14152211e-01 -1.82849562e+00 -1.25899971e+00 7.59393215e-01
-4.33198065e-02 -3.40661407e-01 -6.29637241e-02 4.69055474e-01
-6.47864878e-01 5.71944356e-01 -2.95747668e-01 -3.99713457e-01
-8.33160818e-01 4.53948230e-01 3.38717103e-01 -7.70951927e-01
-4.06245053e-01 6.47381723e-01 1.25596106e-01 -9.25169647e-01
-3.70686799e-01 3.07911217e-01 4.16660845e-01 -4.67979103e-01
2.36739248e-01 -1.25257388e-01 -2.50148028e-01 -6.91829622e-01
1.25727683e-01 1.49282128e-01 -1.25497371e-01 -8.84627283e-01
1.17809892e+00 -1.00278035e-01 -4.48367774e-01 3.96470755e-01
1.18541110e+00 3.46180260e-01 -1.17355812e+00 -1.40436813e-01
-2.30243176e-01 -9.71195102e-02 -6.78562224e-01 -9.03028905e-01
-7.31471896e-01 9.76233363e-01 3.40811729e-01 -1.19760521e-01
8.65289986e-01 -5.98357841e-02 9.79770362e-01 -1.02198273e-01
2.56677449e-01 -9.36629891e-01 3.25101987e-02 7.53654540e-01
1.01603985e+00 -8.90474200e-01 -6.01984203e-01 2.66461931e-02
-8.01994324e-01 8.31449330e-01 4.41976607e-01 -1.94393620e-01
-5.81084862e-02 3.16063404e-01 3.43760222e-01 2.28230909e-01
-1.11266732e+00 4.73435193e-01 1.44620046e-01 5.93043327e-01
7.26440668e-01 3.34724545e-01 -2.45778069e-01 2.13926375e-01
-6.67425394e-01 4.50032949e-02 3.15544605e-01 5.89121699e-01
3.70097101e-01 -1.73299170e+00 -2.40764886e-01 2.36827508e-01
-4.20092523e-01 -5.44669271e-01 -5.87102711e-01 1.44512296e-01
-5.50990477e-02 5.85438073e-01 2.53806949e-01 -2.25970209e-01
-2.20856547e-01 5.80879986e-01 3.74026805e-01 -7.47000813e-01
-7.10994840e-01 5.02841882e-02 6.59490228e-02 -2.77412713e-01
5.22110015e-02 -5.71393371e-01 -6.93383515e-01 -1.68369859e-01
-1.35341346e-01 4.81308728e-01 7.05211699e-01 8.19945812e-01
4.72848922e-01 5.04824460e-01 7.75869370e-01 -4.34283465e-01
-7.33464897e-01 -1.14917707e+00 -1.01644052e-02 5.59603631e-01
1.36566788e-01 2.55207390e-01 -1.37103587e-01 5.12933671e-01]
|
[11.69420051574707, 9.904844284057617]
|
71467a3e-1768-4dc6-818c-9e222308d76e
|
negbert-a-transfer-learning-approach-for
|
1911.04211
| null |
https://arxiv.org/abs/1911.04211v4
|
https://arxiv.org/pdf/1911.04211v4.pdf
|
NegBERT: A Transfer Learning Approach for Negation Detection and Scope Resolution
|
Negation is an important characteristic of language, and a major component of information extraction from text. This subtask is of considerable importance to the biomedical domain. Over the years, multiple approaches have been explored to address this problem: Rule-based systems, Machine Learning classifiers, Conditional Random Field Models, CNNs and more recently BiLSTMs. In this paper, we look at applying Transfer Learning to this problem. First, we extensively review previous literature addressing Negation Detection and Scope Resolution across the 3 datasets that have gained popularity over the years: the BioScope Corpus, the Sherlock dataset, and the SFU Review Corpus. We then explore the decision choices involved with using BERT, a popular transfer learning model, for this task, and report state-of-the-art results for scope resolution across all 3 datasets. Our model, referred to as NegBERT, achieves a token level F1 score on scope resolution of 92.36 on the Sherlock dataset, 95.68 on the BioScope Abstracts subcorpus, 91.24 on the BioScope Full Papers subcorpus, 90.95 on the SFU Review Corpus, outperforming the previous state-of-the-art systems by a significant margin. We also analyze the model's generalizability to datasets on which it is not trained.
|
['Suraj Sawant', 'Aditya Khandelwal']
|
2019-11-11
|
negbert-a-transfer-learning-approach-for-1
|
https://aclanthology.org/2020.lrec-1.704
|
https://aclanthology.org/2020.lrec-1.704.pdf
|
lrec-2020-5
|
['negation-and-speculation-cue-detection', 'negation-detection', 'negation-scope-resolution']
|
['natural-language-processing', 'natural-language-processing', 'natural-language-processing']
|
[ 3.07973117e-01 1.59808099e-01 -7.69405246e-01 -2.71182090e-01
-1.28866374e+00 -3.86898071e-01 5.11076391e-01 4.31873858e-01
-9.51456368e-01 1.31816006e+00 9.41678360e-02 -5.23385525e-01
-3.94278355e-02 -5.68826616e-01 -7.40952075e-01 -3.26702923e-01
1.43236622e-01 1.98667571e-01 2.97708094e-01 -4.24958020e-01
4.00356859e-01 1.25043318e-01 -1.08786690e+00 7.61065364e-01
8.68572593e-01 1.00271642e+00 -1.43449575e-01 6.73348784e-01
-1.09583929e-01 9.32708144e-01 -8.61635506e-01 -5.46478331e-01
-4.34035808e-01 -2.40929052e-01 -8.78247619e-01 -7.30763257e-01
5.05512118e-01 -2.67941691e-02 -5.81276901e-02 1.00648820e+00
8.14611733e-01 -1.68664530e-01 7.64036119e-01 -8.14337969e-01
-5.35214305e-01 8.86065900e-01 -3.34711164e-01 7.74153173e-01
4.16799426e-01 -2.29492351e-01 1.16233265e+00 -7.43933856e-01
8.27029288e-01 1.13295269e+00 9.11350489e-01 8.42295766e-01
-9.14550126e-01 -9.70796347e-01 9.28744152e-02 4.49635774e-01
-1.16625643e+00 -3.54153633e-01 3.76786113e-01 -4.32818174e-01
1.68216193e+00 -1.06306739e-01 4.68845963e-01 1.37392139e+00
9.73383129e-01 8.79887938e-01 1.00609064e+00 -7.09620476e-01
2.27640912e-01 -3.53192128e-02 2.65679330e-01 5.02643168e-01
6.07603252e-01 -1.70374021e-01 -9.30072725e-01 -1.43089116e-01
2.04682499e-01 -4.37030852e-01 -4.48038310e-01 4.18800265e-01
-1.08949351e+00 7.21285999e-01 4.79618423e-02 6.45105481e-01
-3.05829495e-01 1.55089349e-01 8.68661284e-01 3.71195227e-01
4.44812894e-01 3.48913729e-01 -9.95254815e-01 -3.26931357e-01
-1.06977296e+00 3.97704363e-01 1.06492865e+00 8.25892091e-01
1.14749573e-01 -2.54410535e-01 -2.85659879e-01 7.67176390e-01
6.41753078e-02 3.06555450e-01 6.90596044e-01 -4.92185175e-01
7.24086761e-01 4.52337414e-01 -2.37365723e-01 -5.61680913e-01
-7.73692846e-01 -3.93031299e-01 -7.55766392e-01 -2.64497072e-01
3.05843174e-01 -3.47438157e-01 -9.45839405e-01 1.76149356e+00
-2.67813385e-01 -3.01317245e-01 2.02495500e-01 2.71265298e-01
1.32957661e+00 2.80107796e-01 3.88462305e-01 -1.78199336e-01
1.68147850e+00 -6.25420392e-01 -1.26860046e+00 -2.43751809e-01
9.58967805e-01 -6.11155927e-01 4.67227042e-01 1.00564885e+00
-1.08927143e+00 -3.44907761e-01 -1.36112535e+00 -2.29803830e-01
-5.91122985e-01 2.72353977e-01 3.17392260e-01 7.25735664e-01
-9.43487346e-01 6.84337795e-01 -8.08812499e-01 -5.02552569e-01
4.88423496e-01 4.50136006e-01 -2.90449440e-01 -2.22554654e-02
-1.90899634e+00 1.49992537e+00 5.08425891e-01 -1.44506305e-01
-5.62015355e-01 -7.45477021e-01 -7.92497396e-01 7.89104626e-02
3.77557099e-01 -4.85720634e-01 1.52297366e+00 -4.95253891e-01
-1.09774911e+00 1.19482076e+00 -4.87485304e-02 -8.30656111e-01
3.06103319e-01 -2.73045868e-01 -7.24771798e-01 8.97601768e-02
1.86183140e-01 5.28825164e-01 2.58072197e-01 -4.10851300e-01
-8.73981297e-01 -7.30258375e-02 -1.19291715e-01 -5.90542890e-02
-2.54613876e-01 8.15398470e-02 -2.77754784e-01 -6.49387002e-01
-4.54278320e-01 -6.98722839e-01 1.62719235e-01 -3.40397388e-01
-4.34530288e-01 -8.16161513e-01 3.63159150e-01 -7.29660511e-01
1.44300878e+00 -1.75748348e+00 -1.41166568e-01 -1.52601197e-01
7.65002891e-02 4.84766454e-01 2.18530461e-01 4.54511046e-01
-2.87896067e-01 4.17281210e-01 -2.24230126e-01 -7.91671202e-02
-2.20516071e-01 1.14579700e-01 -2.02541314e-02 3.22561622e-01
5.07679045e-01 9.98954654e-01 -9.93464231e-01 -6.32965446e-01
-2.49377668e-01 3.24107438e-01 -3.57794076e-01 -2.46263042e-01
-1.70419440e-01 5.45842946e-02 -3.23566914e-01 8.31986129e-01
2.25800157e-01 -1.80472136e-01 3.40332627e-01 -3.35538149e-01
-6.16026409e-02 7.38101304e-01 -5.83114088e-01 1.67724705e+00
-1.04141004e-01 7.37292945e-01 -7.57904947e-02 -1.16769671e+00
9.00768638e-01 7.27140129e-01 1.90940514e-01 -8.15514505e-01
1.92342117e-01 5.86017430e-01 3.09907466e-01 -5.43985665e-01
3.07825059e-01 -4.73825544e-01 -1.65173113e-01 2.59687956e-02
4.05370504e-01 4.08083610e-02 4.65370208e-01 1.06660530e-01
1.44496024e+00 8.75765607e-02 7.88202286e-01 -4.10324872e-01
7.38456130e-01 -4.91793901e-02 7.55214632e-01 8.40701520e-01
-5.22731781e-01 4.12011892e-01 8.32901597e-01 -1.56064764e-01
-5.48669875e-01 -5.96977174e-01 -5.31089306e-01 8.14451218e-01
-4.99380261e-01 -4.11950797e-01 -7.73360014e-01 -7.87848890e-01
9.22910199e-02 7.78337121e-01 -9.05283272e-01 -4.73469734e-01
-5.58364928e-01 -9.70287502e-01 1.06468034e+00 6.43570602e-01
4.80936348e-01 -1.39662957e+00 -6.30993247e-01 3.45752567e-01
-4.60167825e-01 -1.26465893e+00 -1.52246743e-01 6.83833003e-01
-7.83950627e-01 -1.18891597e+00 -6.54260397e-01 -7.45433986e-01
-8.18076879e-02 -6.84907079e-01 1.32651377e+00 -1.31550312e-01
-2.76525408e-01 2.25030392e-01 -2.58205056e-01 -8.72488379e-01
-4.03356969e-01 4.64334458e-01 -9.36193988e-02 -4.75465983e-01
8.69663358e-01 -1.52515888e-01 -4.15698230e-01 -2.65444696e-01
-7.99616098e-01 -3.61377746e-01 8.34879577e-01 1.12617362e+00
6.02407932e-01 -4.30921465e-01 1.19855452e+00 -1.26357377e+00
7.75371611e-01 -5.91505349e-01 -9.22183767e-02 1.39054194e-01
-8.71944785e-01 -1.46630064e-01 3.64946425e-01 -3.47759604e-01
-8.49245489e-01 -4.22730923e-01 -3.56782168e-01 6.07296117e-02
9.94822290e-03 9.79460835e-01 -9.50505063e-02 1.40928492e-01
9.08619642e-01 3.47598046e-02 -9.38563794e-02 -2.47033522e-01
-2.13190820e-02 9.06432033e-01 4.83750463e-01 -1.81328744e-01
9.14695039e-02 2.40854979e-01 -3.07068005e-02 -8.19529951e-01
-1.11509168e+00 -4.01312292e-01 -5.47631681e-01 9.23604816e-02
1.00136137e+00 -8.93866062e-01 -5.26278198e-01 3.63035709e-01
-1.14727414e+00 -3.58047724e-01 4.93426574e-03 5.30262053e-01
-4.03822333e-01 2.84523726e-01 -8.71588469e-01 -5.55635154e-01
-8.46251607e-01 -9.79625285e-01 8.36661041e-01 1.81568060e-02
-7.64524937e-01 -1.08838010e+00 1.05610386e-01 2.31043220e-01
2.27915719e-01 4.58121628e-01 1.17434537e+00 -1.17023659e+00
1.49214953e-01 -3.12491804e-01 -4.24233526e-02 4.56156582e-01
1.86051950e-01 -1.78606272e-01 -9.54517782e-01 -1.03700653e-01
-6.43606111e-02 -5.53603292e-01 1.26687241e+00 6.48818076e-01
9.46883082e-01 2.19667628e-01 -7.25157559e-01 1.11680932e-01
1.04086924e+00 4.77138966e-01 6.60813034e-01 5.27257085e-01
2.01030880e-01 2.90592670e-01 6.50296390e-01 3.93686257e-02
4.02663022e-01 2.83350199e-01 1.34849876e-01 -6.70653358e-02
-1.26831234e-01 6.23937547e-02 2.03258589e-01 8.84732366e-01
2.50105649e-01 -3.94469261e-01 -1.06809175e+00 7.48590052e-01
-1.65769768e+00 -6.75939381e-01 -2.42252275e-02 1.84723330e+00
1.19222796e+00 6.59072995e-01 -1.42166644e-01 4.62590069e-01
6.22477293e-01 -1.65048391e-01 -4.10914600e-01 -8.00990522e-01
-2.54020542e-01 4.26716566e-01 4.54652905e-01 2.00060859e-01
-1.13185227e+00 8.21561158e-01 6.79756832e+00 8.53230059e-01
-1.13211715e+00 1.12224564e-01 5.90578318e-01 -1.10066071e-01
1.02849185e-01 -3.11241627e-01 -1.20443785e+00 3.37204158e-01
1.48052096e+00 -6.89432099e-02 -1.96107313e-01 3.88000309e-01
1.05485007e-01 -2.83035576e-01 -1.30335271e+00 6.51330650e-01
3.62046897e-01 -1.32342863e+00 -6.60739616e-02 -1.29018873e-01
6.05306447e-01 2.87675261e-01 -1.74072102e-01 7.85285354e-01
-9.70191956e-02 -1.18762612e+00 6.14669025e-01 4.30023104e-01
1.08780706e+00 -6.67828679e-01 1.19233036e+00 3.00088465e-01
-5.98870397e-01 4.76688109e-02 -1.95707783e-01 -1.08754590e-01
-5.47592975e-02 8.28052938e-01 -7.61609912e-01 5.47992527e-01
9.13068771e-01 8.76954496e-01 -3.69418770e-01 7.55918086e-01
-4.67860907e-01 1.03738487e+00 -9.26547721e-02 -4.99553382e-01
1.37882337e-01 5.70905745e-01 2.94413537e-01 1.72065103e+00
3.92730311e-02 5.91142252e-02 -1.82313368e-01 7.63311088e-01
-4.76993293e-01 1.94023311e-01 -3.85026813e-01 -2.28437811e-01
4.43259984e-01 1.04433465e+00 -5.41579247e-01 -5.59084475e-01
-6.25555813e-01 4.51565087e-01 3.17919880e-01 2.38903672e-01
-7.76994288e-01 -8.53074670e-01 6.70423061e-02 -3.00565809e-01
5.21877408e-01 3.76862615e-01 -4.13964212e-01 -1.00930166e+00
-4.19388786e-02 -1.06586015e+00 6.67453110e-01 -6.69439137e-01
-1.27569902e+00 5.12476921e-01 5.95648549e-02 -9.13659632e-01
-2.34631151e-01 -9.10193026e-01 -3.44879657e-01 9.90790427e-01
-1.89409053e+00 -8.36952925e-01 1.61215395e-01 2.91452736e-01
3.67320865e-01 -2.08011016e-01 1.03223693e+00 4.82672095e-01
-4.23699260e-01 8.05738807e-01 -5.84526127e-03 3.11735958e-01
1.19815087e+00 -1.24352455e+00 2.84032792e-01 4.42996800e-01
-3.84202272e-01 8.23163629e-01 3.68826509e-01 -9.17682111e-01
-1.05407143e+00 -9.55253661e-01 1.50459957e+00 -4.29968178e-01
6.48613214e-01 -2.24275902e-01 -1.04827774e+00 7.77585864e-01
2.41401494e-01 -1.82126492e-01 8.34571421e-01 2.05142066e-01
-2.50292391e-01 1.38786510e-02 -1.34100878e+00 3.58772516e-01
7.37941623e-01 -5.10746598e-01 -1.10255551e+00 2.49585465e-01
4.95452523e-01 -6.52800679e-01 -1.30099392e+00 6.65425837e-01
6.11371994e-01 -5.32660484e-01 5.73264062e-01 -5.47235787e-01
7.26580858e-01 1.29116997e-01 9.69960913e-02 -1.14434147e+00
-3.14584315e-01 -4.38351244e-01 -1.85469449e-01 1.07204807e+00
8.78592491e-01 -6.29117072e-01 6.02007031e-01 2.61935264e-01
-3.04144591e-01 -1.15003681e+00 -1.20129883e+00 -4.27138180e-01
6.10561013e-01 -4.68834072e-01 1.53556252e-02 9.23361540e-01
4.11518872e-01 5.67174971e-01 8.54492486e-02 -3.45247388e-01
2.42046461e-01 -3.30588460e-01 3.51172835e-02 -1.24216759e+00
2.53419038e-02 -5.50479054e-01 -3.03722024e-01 -5.79555511e-01
1.98914275e-01 -1.13910961e+00 -2.51055863e-02 -1.83416080e+00
3.13292533e-01 1.58939824e-01 -5.62981963e-01 8.88206422e-01
-3.09697688e-01 9.89244282e-02 -2.72591740e-01 -7.98196942e-02
-4.64076817e-01 1.53737143e-01 9.84088838e-01 -3.30337048e-01
-3.00909448e-02 -2.62643099e-01 -9.93876755e-01 5.62380433e-01
8.84576917e-01 -4.72993433e-01 5.11807390e-02 -1.26060084e-01
4.28901404e-01 -6.82110190e-02 -3.42104919e-02 -8.41666341e-01
2.33548537e-01 2.87858903e-01 6.47785127e-01 -8.44314396e-01
6.48939684e-02 -7.23124593e-02 -4.47157919e-01 8.64650905e-01
-5.95257998e-01 1.05609566e-01 7.46333063e-01 3.91581863e-01
-2.13200107e-01 -4.04492408e-01 6.66356325e-01 -2.12069228e-01
-4.55955595e-01 -1.74653023e-01 -9.34185743e-01 5.34005046e-01
5.47671735e-01 9.54866409e-02 -4.38092232e-01 -1.41405851e-01
-6.13459885e-01 4.07346666e-01 -1.92326680e-01 5.64641118e-01
4.63072002e-01 -8.81700873e-01 -7.75095642e-01 -1.21741608e-01
1.87314510e-01 -1.75172523e-01 -3.12725827e-02 1.13875151e+00
-1.69849440e-01 1.09695256e+00 -5.15237525e-02 -5.22904873e-01
-1.21899736e+00 1.90151513e-01 3.86652738e-01 -6.93898320e-01
-4.33836758e-01 7.64569283e-01 -1.08624265e-01 -4.69017565e-01
3.23514402e-01 -8.41814637e-01 -6.42756343e-01 2.90670365e-01
5.23292661e-01 3.50425690e-01 8.07689130e-01 -1.32986844e-01
-6.41004324e-01 3.13974589e-01 -3.69256288e-01 -6.97109625e-02
1.01052833e+00 3.41642708e-01 -1.20196797e-01 7.27771461e-01
1.11428177e+00 -1.70121506e-01 -3.38992208e-01 -1.97046667e-01
3.77764612e-01 5.43885231e-01 2.47544125e-01 -1.38730240e+00
-6.69951558e-01 8.95722806e-01 4.31667060e-01 -4.87810373e-02
9.28870261e-01 -3.09284657e-01 7.29498625e-01 4.62781966e-01
2.40300652e-02 -1.24361658e+00 -1.29002005e-01 8.66708696e-01
7.05015779e-01 -1.15586197e+00 1.31694660e-01 -1.82680517e-01
-4.59499002e-01 1.12713063e+00 5.28079093e-01 3.56828012e-02
5.96249878e-01 3.65148455e-01 -3.96562461e-03 -1.87003911e-01
-1.04929364e+00 1.71320572e-01 3.68826866e-01 4.05120730e-01
9.64189470e-01 -1.12146884e-01 -8.99286985e-01 1.12136102e+00
-3.60074304e-02 5.32210648e-01 6.19012058e-01 1.17608142e+00
-3.93346876e-01 -1.10610259e+00 -2.17070282e-01 9.18772161e-01
-1.24952698e+00 -5.05079687e-01 -5.11922956e-01 1.02069545e+00
2.31869519e-01 8.61944735e-01 -2.30233878e-01 -1.97407663e-01
6.27986133e-01 5.07825375e-01 5.12315869e-01 -7.65147746e-01
-8.77100885e-01 1.13118999e-01 4.58280832e-01 -3.00793797e-01
-4.51648027e-01 -6.91768169e-01 -1.49801230e+00 1.42277420e-01
-4.98089761e-01 1.55782253e-01 4.68324393e-01 1.18897367e+00
2.04971790e-01 9.23754334e-01 -3.70799273e-01 -4.21094447e-01
-6.57313108e-01 -1.40310359e+00 -4.28879708e-01 1.98434782e-03
4.23512161e-01 -6.83422804e-01 -2.78608650e-01 2.61767744e-03]
|
[8.560348510742188, 8.75224494934082]
|
8df82a06-717b-4138-9870-7816f4ddc8d2
|
multiple-combined-constraints-for-image
|
1809.06706
| null |
http://arxiv.org/abs/1809.06706v1
|
http://arxiv.org/pdf/1809.06706v1.pdf
|
Multiple Combined Constraints for Image Stitching
|
Several approaches to image stitching use different constraints to estimate
the motion model between image pairs. These constraints can be roughly divided
into two categories: geometric constraints and photometric constraints. In this
paper, geometric and photometric constraints are combined to improve the
alignment quality, which is based on the observation that these two kinds of
constraints are complementary. On the one hand, geometric constraints (e.g.,
point and line correspondences) are usually spatially biased and are
insufficient in some extreme scenes, while photometric constraints are always
evenly and densely distributed. On the other hand, photometric constraints are
sensitive to displacements and are not suitable for images with large
parallaxes, while geometric constraints are usually imposed by feature matching
and are more robust to handle parallaxes. The proposed method therefore
combines them together in an efficient mesh-based image warping framework. It
achieves better alignment quality than methods only with geometric constraints,
and can handle larger parallax than photometric-constraint-based method.
Experimental results on various images illustrate that the proposed method
outperforms representative state-of-the-art image stitching methods reported in
the literature.
|
['Li Li', 'Kai Chen', 'Jian Yao', 'Jingmin Tu', 'Binbin Xiang']
|
2018-09-18
| null | null | null | null |
['image-stitching']
|
['computer-vision']
|
[ 3.50278616e-01 -4.29171681e-01 -2.20028639e-01 -6.17827401e-02
-1.31549478e-01 -5.23950040e-01 5.52966952e-01 -3.46522033e-02
-3.69074196e-01 4.86722410e-01 -5.10014556e-02 1.34419113e-01
-2.60357589e-01 -5.38263738e-01 -4.23434407e-01 -1.04665160e+00
5.13510287e-01 3.84062976e-01 7.14834571e-01 -3.66816461e-01
6.09526396e-01 5.16714871e-01 -1.58471739e+00 -2.14080036e-01
1.10307539e+00 8.85173857e-01 4.72382396e-01 4.29036990e-02
-1.12799697e-01 2.22013727e-01 -3.81119937e-01 -9.71444547e-02
3.72478992e-01 -3.50601047e-01 -2.60606617e-01 5.19883275e-01
5.40084124e-01 -3.23275447e-01 -2.70891130e-01 1.25385380e+00
3.62694532e-01 5.12058474e-02 5.34495413e-01 -1.14843249e+00
-3.10228884e-01 -2.69272663e-02 -1.21166968e+00 -7.39884824e-02
3.94313037e-01 3.22035351e-03 5.07694662e-01 -8.03280532e-01
5.23052335e-01 1.03066695e+00 6.65209234e-01 1.15587354e-01
-1.10587442e+00 -6.37880743e-01 6.29712641e-02 1.96957946e-01
-1.38553178e+00 -4.86739099e-01 1.32718623e+00 -3.62630934e-01
3.27926040e-01 4.25200820e-01 4.86966461e-01 5.65926313e-01
3.11148316e-01 4.62772965e-01 1.12010515e+00 -5.97604811e-01
-1.93284254e-03 4.47661020e-02 -2.21013706e-02 5.16888499e-01
3.48406225e-01 3.13147217e-01 -3.46114814e-01 -3.29046041e-01
1.19681573e+00 2.46701002e-01 -7.34387398e-01 -8.15821528e-01
-1.52275860e+00 7.05126762e-01 3.21911782e-01 5.71142972e-01
-5.86305931e-02 -9.85818729e-03 3.04416716e-01 8.54515657e-02
3.21536094e-01 2.76297003e-01 4.08038013e-02 1.47304252e-01
-8.95243824e-01 1.12633817e-01 3.40532631e-01 1.16597700e+00
9.07970905e-01 5.69093265e-02 2.63071716e-01 1.02549040e+00
2.11806774e-01 5.77760875e-01 4.26219225e-01 -7.63548613e-01
5.35351574e-01 2.80085325e-01 2.67748535e-01 -1.60006344e+00
-1.70131102e-01 -8.74152258e-02 -9.82052565e-01 2.88931638e-01
4.27010447e-01 1.33427650e-01 -8.00982594e-01 1.30795312e+00
3.58943641e-01 3.51772249e-01 -2.09544525e-01 9.75068808e-01
7.46422350e-01 5.28016150e-01 -6.01105332e-01 -5.82303882e-01
1.29220605e+00 -9.24503148e-01 -1.16678452e+00 -1.27673283e-01
2.94772923e-01 -1.45801997e+00 7.39254236e-01 2.76866198e-01
-1.03504658e+00 -7.53901303e-01 -1.18582642e+00 2.26468146e-01
1.45142584e-03 7.92280361e-02 2.46469274e-01 5.70441186e-01
-9.17986691e-01 4.55828696e-01 -6.75119340e-01 -2.25663811e-01
-2.43978143e-01 2.59194732e-01 -1.20279096e-01 -1.63678363e-01
-8.70529294e-01 8.28932762e-01 3.80973846e-01 1.67228192e-01
-1.62102818e-01 -1.96216851e-01 -8.14461291e-01 -7.98536316e-02
4.93382484e-01 -5.40446460e-01 7.22714841e-01 -9.83101785e-01
-1.50539422e+00 6.65320516e-01 -2.89730877e-01 3.60969491e-02
5.90671122e-01 -1.19336704e-02 -2.31168717e-01 2.82985628e-01
-3.10870945e-01 5.30867159e-01 1.14838862e+00 -1.74346602e+00
-4.18690890e-01 -1.15358301e-01 -1.77797079e-01 2.74854630e-01
-3.95489305e-01 3.11467284e-03 -6.98116183e-01 -8.61956596e-01
5.19021749e-01 -1.05989754e+00 -2.55253613e-01 6.54352978e-02
-1.97861224e-01 -8.37691501e-03 1.46762252e+00 -4.89496440e-01
1.27367687e+00 -2.15931010e+00 2.48063356e-01 2.94772923e-01
-3.44263986e-02 4.72896397e-01 -9.20439214e-02 6.40145659e-01
-1.33501425e-01 -3.24488133e-01 -2.33604163e-01 -3.30046445e-01
-4.34757590e-01 2.64073342e-01 -1.89648926e-01 7.54705310e-01
-1.91922173e-01 3.78538758e-01 -7.61133850e-01 -6.59784794e-01
6.92380667e-01 4.43193853e-01 -2.98726320e-01 -2.96897143e-02
-1.30321393e-02 6.53459847e-01 -3.79752904e-01 4.38952029e-01
1.12445998e+00 1.60633832e-01 -2.53139883e-02 -4.86772388e-01
-4.23186958e-01 -2.10898980e-01 -1.28409946e+00 1.69552016e+00
-2.83384532e-01 5.40373325e-01 1.31898403e-01 -8.60101998e-01
1.15489781e+00 4.14559633e-01 6.52595162e-01 -2.40339965e-01
4.07166123e-01 4.53594625e-01 -9.33657289e-02 -4.31611270e-01
5.90060353e-01 9.98265371e-02 2.04045147e-01 1.07404977e-01
-4.28914934e-01 -5.55584908e-01 1.46939903e-01 -2.68730253e-01
4.75643575e-01 2.76746988e-01 3.81510407e-01 -4.54372406e-01
9.33894277e-01 -4.14192751e-02 9.78353977e-01 3.97687554e-01
3.74487564e-02 8.69914532e-01 2.07531229e-02 -5.14933586e-01
-1.26677608e+00 -5.89291990e-01 -1.84001535e-01 1.16087720e-01
1.02814877e+00 -4.24025267e-01 -6.76765978e-01 -3.74379933e-01
-2.54470497e-01 3.33896488e-01 -2.20891804e-01 2.14327827e-01
-9.20217752e-01 -4.10355568e-01 6.06736802e-02 2.77525425e-01
8.37314546e-01 -5.76716185e-01 -6.57947958e-01 1.70965746e-01
-3.12069863e-01 -1.17058778e+00 -9.07460988e-01 -5.21662831e-01
-1.04706919e+00 -1.12011230e+00 -1.08241069e+00 -1.01721025e+00
1.13678408e+00 1.04372561e+00 7.89079368e-01 5.42649031e-01
-3.88273969e-02 1.37056634e-01 -5.87595046e-01 -1.41050547e-01
-2.90655196e-01 -3.36268514e-01 -3.92044969e-02 2.53125846e-01
1.17052970e-02 -6.51006877e-01 -6.94350302e-01 1.10853899e+00
-1.20894480e+00 2.84288108e-01 4.59826350e-01 1.08439541e+00
4.92181063e-01 3.27082068e-01 -5.51305637e-02 -4.94895607e-01
1.36978269e-01 2.09043548e-01 -7.96543181e-01 2.90740371e-01
-4.02622789e-01 -1.40740380e-01 4.81136888e-01 -7.42065310e-01
-1.16517317e+00 2.50577778e-01 1.89467400e-01 -8.12785327e-01
-1.12069644e-01 3.57811689e-01 -1.60887390e-01 -7.04584181e-01
3.03865284e-01 3.82038325e-01 2.02686816e-01 -4.41537946e-01
-1.20095596e-01 5.30279756e-01 5.46287060e-01 -5.03364623e-01
1.17381263e+00 7.58288622e-01 2.87164330e-01 -1.14785695e+00
-1.61245048e-01 -6.43543601e-01 -8.11736166e-01 -3.92855495e-01
7.08840311e-01 -5.67082524e-01 -3.60770851e-01 8.15080106e-01
-1.23585689e+00 1.66653812e-01 1.51620507e-01 7.36604691e-01
-6.08094752e-01 1.23805618e+00 -4.13354725e-01 -6.71882272e-01
-5.35992980e-02 -1.45244300e+00 1.04717171e+00 3.08764696e-01
1.45509699e-02 -1.03382099e+00 -3.90780084e-02 1.94681525e-01
4.19552207e-01 4.49228525e-01 6.39388978e-01 1.32468134e-01
-6.94365382e-01 -2.62874186e-01 -1.57380253e-01 2.00379476e-01
4.57500160e-01 2.60321856e-01 -5.91078758e-01 -4.94043887e-01
3.64858449e-01 1.60246685e-01 5.20260334e-01 3.61689150e-01
8.78351808e-01 -1.43104196e-01 -5.70685208e-01 5.20043790e-01
1.60880053e+00 3.84118289e-01 1.02883267e+00 4.31418002e-01
6.87674344e-01 7.05731928e-01 1.21401036e+00 3.24905127e-01
8.07192177e-03 1.32282686e+00 4.45529103e-01 -4.72630203e-01
-6.42684326e-02 7.15628341e-02 1.78564563e-01 7.94512928e-01
-3.51765156e-01 -2.02120826e-01 -6.75857544e-01 2.53973842e-01
-2.12351894e+00 -8.71906698e-01 -4.47132677e-01 2.55797911e+00
6.36946499e-01 -1.40602380e-01 -8.69815797e-02 3.16482395e-01
1.10866642e+00 3.90799731e-01 -1.47303820e-01 -2.77474895e-02
-1.96843058e-01 -3.38580996e-01 8.08007598e-01 5.65169394e-01
-9.53515947e-01 7.65404046e-01 5.31876326e+00 1.17381918e+00
-1.13334537e+00 -7.91854635e-02 5.05474173e-02 3.13281357e-01
-3.57904613e-01 3.70062083e-01 -6.04163945e-01 6.80166066e-01
-8.28476436e-03 1.24647446e-01 1.70476183e-01 3.49972129e-01
1.13384664e-01 -3.12726438e-01 -6.89668775e-01 1.31955659e+00
3.84424031e-01 -1.09292436e+00 -2.37251282e-01 2.08547965e-01
9.08467114e-01 -3.93252909e-01 -2.33699307e-02 -5.78961849e-01
-1.80763781e-01 -6.02748930e-01 6.87817335e-01 2.77294219e-01
7.27504611e-01 -8.31240714e-01 9.09885526e-01 4.04207468e-01
-1.30405247e+00 2.78587312e-01 -4.96233821e-01 1.99154034e-01
5.09534299e-01 7.45648205e-01 -1.53892919e-01 9.17418420e-01
4.99071509e-01 7.42743611e-01 -1.68303818e-01 1.39842343e+00
-2.03451648e-01 -1.65047236e-02 -4.73393917e-01 2.84600466e-01
1.52575374e-01 -7.77923763e-01 8.17882419e-01 7.11520910e-01
5.84513187e-01 2.13328809e-01 3.20845634e-01 5.75560033e-01
6.84526980e-01 3.67737561e-02 -7.00303018e-01 3.85550708e-01
4.40603465e-01 1.12302208e+00 -9.19657886e-01 -3.10637683e-01
-5.24174392e-01 1.00178528e+00 -3.47605467e-01 2.86241859e-01
-8.32800925e-01 -3.45798463e-01 3.53815794e-01 1.78224713e-01
3.30920815e-01 -6.62443876e-01 -1.63661957e-01 -1.35150254e+00
4.46291501e-03 -8.88573408e-01 6.59189001e-02 -6.72801614e-01
-9.45170760e-01 5.27291000e-01 3.21310461e-01 -2.04449224e+00
-1.31133914e-01 -4.24013197e-01 -6.09827220e-01 7.30109751e-01
-1.61406577e+00 -1.18701196e+00 -5.67212105e-01 6.92408442e-01
6.48591995e-01 1.61301702e-01 5.50705791e-01 2.71108329e-01
-4.34270620e-01 2.91214406e-01 1.14021100e-01 -1.75685018e-01
1.02684510e+00 -6.67590976e-01 -7.91315287e-02 1.00029099e+00
-1.15837477e-01 6.47768557e-01 9.24079299e-01 -7.13772237e-01
-1.33943319e+00 -6.03586614e-01 7.37082422e-01 1.03369050e-01
2.98316509e-01 -8.02585557e-02 -1.00937688e+00 2.71066844e-01
4.60071295e-01 -1.32450610e-01 2.82237738e-01 -3.99663568e-01
-2.07358673e-01 -6.24490045e-02 -1.06278431e+00 5.21454990e-01
8.91196191e-01 -6.86071441e-02 -4.50658619e-01 2.47470140e-01
2.90236264e-01 -7.24104643e-01 -8.10475469e-01 6.15227759e-01
4.52690572e-01 -1.27757740e+00 1.04211104e+00 4.74629372e-01
3.26840729e-02 -9.00579214e-01 1.71870530e-01 -1.18903530e+00
-3.29501450e-01 -8.17107320e-01 3.58359009e-01 1.38209617e+00
-1.28671631e-01 -1.00755823e+00 5.41508675e-01 2.36754894e-01
-1.74680531e-01 -4.47798461e-01 -8.92773330e-01 -1.07999754e+00
-3.19840968e-01 2.90839612e-01 5.44839144e-01 1.26876664e+00
-6.79753423e-02 -2.39179265e-02 -7.32092738e-01 2.27028966e-01
7.20513582e-01 3.79073799e-01 8.61409545e-01 -1.17106938e+00
-2.40004048e-01 -5.43959677e-01 -5.29759347e-01 -1.19620132e+00
-5.14791384e-02 -8.13386664e-02 1.38083309e-01 -1.11848652e+00
-5.02814446e-03 -8.40378106e-01 2.51401812e-01 1.55233741e-01
-1.92845702e-01 3.96368682e-01 1.69687703e-01 6.96746290e-01
2.35235095e-01 5.69691062e-01 1.31869018e+00 -4.99245450e-02
-2.52583981e-01 -2.17727218e-02 -4.72676642e-02 9.42071974e-01
7.46161461e-01 -1.39661908e-01 -4.49049622e-01 -6.03285551e-01
-4.81556952e-02 2.41029650e-01 9.29970369e-02 -9.65976536e-01
3.24669600e-01 -4.66931969e-01 1.36989474e-01 -9.08911765e-01
5.66053748e-01 -1.30623746e+00 5.62882185e-01 4.91860867e-01
1.94910452e-01 1.83014184e-01 1.59214437e-02 4.84811485e-01
-5.27409971e-01 -4.57077175e-01 1.06213367e+00 3.64256240e-02
-5.82331896e-01 1.97756410e-01 4.74250168e-02 -6.64003611e-01
1.19132519e+00 -7.38496363e-01 -2.18074322e-01 -5.04433632e-01
-1.25294060e-01 1.57048588e-03 1.05156314e+00 4.27025646e-01
7.38343477e-01 -1.49889863e+00 -5.73263645e-01 3.84011805e-01
9.31858271e-02 1.17419258e-01 1.93584993e-01 1.24118745e+00
-6.93502665e-01 3.11993390e-01 -2.12562710e-01 -9.71821725e-01
-1.75835323e+00 7.87322581e-01 1.35255642e-02 -5.77179901e-02
-5.45453787e-01 5.08808315e-01 4.35674131e-01 1.89512763e-02
1.96770146e-01 -2.21895307e-01 -1.64272010e-01 -1.76402703e-01
4.67773259e-01 3.22091490e-01 -1.40634060e-01 -1.03161204e+00
-2.43427426e-01 1.54628134e+00 6.37070984e-02 -6.01366684e-02
1.05986345e+00 -4.53843117e-01 -1.31561115e-01 3.81633230e-02
9.39606190e-01 3.04686755e-01 -1.19193387e+00 -5.26043653e-01
-3.30993414e-01 -1.18132627e+00 3.04013062e-02 -4.05725464e-02
-1.32249570e+00 8.17694008e-01 3.26054692e-01 2.37169042e-01
1.31289542e+00 -4.71122473e-01 8.96029055e-01 -1.90294296e-01
6.13310456e-01 -1.06605887e+00 2.14451820e-01 3.16743821e-01
9.74661410e-01 -8.22780132e-01 3.44195247e-01 -1.11340761e+00
-3.86671692e-01 1.41540813e+00 5.27024150e-01 -2.35051572e-01
3.02179754e-01 2.14416772e-01 -3.07884887e-02 1.53643623e-01
-7.77280703e-02 -1.32355407e-01 4.40824449e-01 5.82643688e-01
2.80965090e-01 -3.97667527e-01 -7.33824313e-01 -4.76322532e-01
1.91603675e-01 -2.02729449e-01 4.46560413e-01 1.04699230e+00
-5.19039333e-01 -1.47635710e+00 -1.03273070e+00 -1.34004414e-01
-3.80237065e-02 2.00641632e-01 -1.44872546e-01 7.88060546e-01
9.97385830e-02 1.11880374e+00 -1.71446111e-02 -3.94360155e-01
2.29690775e-01 -5.36692321e-01 6.98367178e-01 -1.41062751e-01
-1.92730054e-01 7.95025051e-01 1.19367661e-02 -4.08638805e-01
-8.84227931e-01 -7.31325388e-01 -9.42039669e-01 -3.61169904e-01
-9.38836813e-01 -6.47361949e-03 6.02296948e-01 7.55105913e-01
1.90428555e-01 4.95117791e-02 8.84650826e-01 -1.31131351e+00
-2.91412294e-01 -7.50054002e-01 -4.67267543e-01 3.69170994e-01
2.92326659e-01 -9.40505862e-01 -5.05003452e-01 1.72676861e-01]
|
[9.289320945739746, -2.371074676513672]
|
213bf32e-54ed-44f6-9ae8-b94311266a83
|
seqtrack-sequence-to-sequence-learning-for
|
2304.14394
| null |
https://arxiv.org/abs/2304.14394v1
|
https://arxiv.org/pdf/2304.14394v1.pdf
|
SeqTrack: Sequence to Sequence Learning for Visual Object Tracking
|
In this paper, we present a new sequence-to-sequence learning framework for visual tracking, dubbed SeqTrack. It casts visual tracking as a sequence generation problem, which predicts object bounding boxes in an autoregressive fashion. This is different from prior Siamese trackers and transformer trackers, which rely on designing complicated head networks, such as classification and regression heads. SeqTrack only adopts a simple encoder-decoder transformer architecture. The encoder extracts visual features with a bidirectional transformer, while the decoder generates a sequence of bounding box values autoregressively with a causal transformer. The loss function is a plain cross-entropy. Such a sequence learning paradigm not only simplifies tracking framework, but also achieves competitive performance on benchmarks. For instance, SeqTrack gets 72.5% AUC on LaSOT, establishing a new state-of-the-art performance. Code and models are available at here.
|
['Han Hu', 'Huchuan Lu', 'Dong Wang', 'Houwen Peng', 'Xin Chen']
|
2023-04-27
| null |
http://openaccess.thecvf.com//content/CVPR2023/html/Chen_SeqTrack_Sequence_to_Sequence_Learning_for_Visual_Object_Tracking_CVPR_2023_paper.html
|
http://openaccess.thecvf.com//content/CVPR2023/papers/Chen_SeqTrack_Sequence_to_Sequence_Learning_for_Visual_Object_Tracking_CVPR_2023_paper.pdf
|
cvpr-2023-1
|
['visual-tracking', 'visual-object-tracking']
|
['computer-vision', 'computer-vision']
|
[ 1.10326663e-01 5.07710017e-02 -4.04933095e-01 -2.74935126e-01
-8.31691682e-01 -7.13790357e-01 9.25944507e-01 -5.81090569e-01
-2.32660294e-01 6.45711780e-01 1.58054799e-01 -2.73121059e-01
6.63719296e-01 -3.16277266e-01 -1.21049356e+00 -7.51557648e-01
-2.90212892e-02 3.36281687e-01 6.16356730e-01 1.04098767e-01
-1.19721442e-01 1.20849855e-01 -1.25910997e+00 2.19260022e-01
4.77240920e-01 1.41833889e+00 1.39822602e-01 9.93741333e-01
-9.26313624e-02 1.30819523e+00 -4.53940243e-01 -8.56425405e-01
3.45553935e-01 -2.60786176e-01 -3.33943337e-01 -2.33106613e-01
7.88341701e-01 -4.32561994e-01 -8.19625199e-01 1.08731997e+00
2.96135277e-01 -3.12694371e-01 7.12604940e-01 -1.67679799e+00
-1.00465882e+00 5.41486382e-01 -5.47103465e-01 4.51793447e-02
1.68689623e-01 7.42776811e-01 1.16039598e+00 -8.41702938e-01
6.99199736e-01 1.32565665e+00 1.09073007e+00 8.32406938e-01
-1.28286457e+00 -1.03445661e+00 3.23613018e-01 8.56415555e-02
-1.09161568e+00 -4.07334149e-01 4.01740223e-01 -1.02711487e+00
8.43995512e-01 -1.19647786e-01 8.81369293e-01 1.63520098e+00
4.27853346e-01 1.16377556e+00 6.11892223e-01 4.71314639e-02
-8.61362368e-02 -6.77565262e-02 -2.10513007e-02 9.53121603e-01
2.50546217e-01 7.11931944e-01 -7.18430579e-01 1.74509659e-01
5.70121765e-01 1.41353801e-01 7.29258591e-03 -7.70592511e-01
-1.25518060e+00 7.80949891e-01 6.33359194e-01 -2.97882348e-01
5.30545693e-03 6.57242358e-01 6.47448778e-01 3.68510559e-03
2.08648875e-01 -4.87055145e-02 -2.30614901e-01 -1.45436972e-01
-9.71363068e-01 3.72718424e-01 5.78175247e-01 1.42626679e+00
4.07994330e-01 4.57862973e-01 -6.70048356e-01 9.61644202e-02
8.06386471e-01 1.00818431e+00 1.36367708e-01 -7.37161934e-01
4.59999621e-01 3.36092621e-01 6.97440654e-03 -4.93613899e-01
-7.66479671e-02 -4.50002998e-01 -7.00394452e-01 4.45158541e-01
4.48489875e-01 -3.57219696e-01 -1.19454396e+00 1.76260424e+00
2.50162125e-01 4.24903452e-01 -2.66036630e-01 1.03988767e+00
8.74553502e-01 5.83146989e-01 4.31470335e-01 1.43974647e-01
1.49274838e+00 -1.34705782e+00 -6.92620754e-01 -2.73580283e-01
2.39549577e-01 -5.12364030e-01 7.70759583e-01 1.11559995e-01
-1.09308839e+00 -7.23002017e-01 -9.37374532e-01 -3.42078090e-01
-2.97597766e-01 2.46347338e-01 4.60136235e-01 5.26236057e-01
-1.23223102e+00 3.13434690e-01 -1.18367910e+00 -1.42639309e-01
6.53778791e-01 1.29927263e-01 -1.52444378e-01 5.62542796e-01
-1.08780873e+00 8.28081250e-01 3.34071845e-01 1.05744917e-02
-1.60562980e+00 -1.02387059e+00 -1.15775311e+00 2.02216934e-02
4.11497951e-01 -8.65363538e-01 1.68570101e+00 -7.59709954e-01
-1.55371749e+00 7.64068186e-01 -3.75332057e-01 -9.63424742e-01
1.02038050e+00 -5.66940606e-01 -3.04513216e-01 -3.58658940e-01
1.74142152e-01 8.94986033e-01 1.26424074e+00 -1.03750944e+00
-7.83957243e-01 -6.64670765e-02 -3.85578424e-01 -2.64841527e-01
1.86171327e-02 1.81739300e-01 -7.63676286e-01 -6.84252083e-01
-5.82746565e-01 -1.06570017e+00 -1.16686717e-01 5.91612995e-01
-6.00340426e-01 -2.60042191e-01 9.52372909e-01 -6.91289008e-01
1.14948285e+00 -1.99210370e+00 -1.68034900e-02 -2.19968379e-01
6.51593864e-01 4.40545499e-01 -2.01330017e-02 -3.94235365e-02
5.54347448e-02 -1.09313309e-01 -3.97758111e-02 -4.30906028e-01
4.38853204e-01 -2.76383817e-01 -7.12381303e-01 5.57679832e-01
3.43004465e-01 1.38532531e+00 -9.33009565e-01 -6.92274570e-01
2.51897424e-01 6.44588888e-01 -5.59635162e-01 3.22999030e-01
-5.97550988e-01 3.15845788e-01 -3.23791891e-01 6.77698970e-01
4.63374287e-01 -7.28027403e-01 3.89498360e-02 -1.83652088e-01
-2.80817986e-01 1.97371751e-01 -4.73586500e-01 1.43484819e+00
-1.06281534e-01 1.15033650e+00 -2.91938901e-01 -4.84749228e-01
9.48005617e-01 4.74665649e-02 4.33688849e-01 -6.20649338e-01
1.02506720e-01 -2.10688978e-01 -2.43204013e-01 -2.44167492e-01
5.47456980e-01 1.72293052e-01 -1.06052525e-01 -6.10598475e-02
2.35263884e-01 4.09479350e-01 1.48656830e-01 2.77293772e-01
1.19847226e+00 9.35489297e-01 2.24716350e-01 1.35100245e-01
3.71190995e-01 1.18971780e-01 6.99727654e-01 8.09238791e-01
-4.50505137e-01 4.01933193e-01 6.35604322e-01 -3.27134460e-01
-1.44989347e+00 -1.52686059e+00 1.49202660e-01 1.11952603e+00
1.07476592e-01 -5.88569582e-01 -5.15928864e-01 -8.90292287e-01
3.77429634e-01 4.10120547e-01 -6.91389740e-01 -1.58070251e-01
-6.05514348e-01 -1.35755405e-01 8.18609476e-01 9.97337699e-01
3.51856798e-01 -9.71620500e-01 -5.78748822e-01 6.79702014e-02
6.75571412e-02 -1.26191723e+00 -9.58548307e-01 1.25878826e-01
-5.70653379e-01 -9.64790106e-01 -7.89920866e-01 -6.62214637e-01
2.48452887e-01 -1.38018951e-01 1.21445990e+00 -3.58764470e-01
-3.21001858e-01 -1.10165805e-01 -2.15156600e-02 -4.67797011e-01
-3.81223708e-01 1.18070945e-01 -1.04635395e-01 -1.47468030e-01
4.00973260e-01 -8.74876007e-02 -6.63625181e-01 2.62138933e-01
-2.49391705e-01 8.70569125e-02 6.29673004e-01 7.38890588e-01
6.20892763e-01 -9.25848126e-01 1.93318561e-01 -5.17905295e-01
2.05764845e-02 -3.29610020e-01 -1.32895231e+00 2.73134708e-01
-4.52976018e-01 2.63802290e-01 7.27700293e-01 -4.48360831e-01
-1.02294815e+00 5.65795124e-01 -1.67499244e-01 -1.06451142e+00
1.55769542e-01 -2.52752334e-01 -4.82795648e-02 -1.35604694e-01
5.29880881e-01 5.57698369e-01 7.25843832e-02 -4.29933667e-01
5.65659165e-01 2.32152447e-01 9.55440104e-01 -1.53486267e-01
1.24289370e+00 4.53469813e-01 -1.37241203e-02 -3.50606531e-01
-9.51567829e-01 -4.00925964e-01 -5.52826643e-01 -4.95705932e-01
1.18305838e+00 -1.20024180e+00 -1.53098619e+00 3.90121788e-01
-1.14887214e+00 -4.55843389e-01 -2.31107026e-01 3.53257716e-01
-7.81394541e-01 1.41094863e-01 -6.42827153e-01 -8.44664514e-01
-5.06728113e-01 -9.83581722e-01 1.41435111e+00 3.00518364e-01
-7.77974427e-02 -8.23025703e-01 1.79181248e-01 3.07978950e-02
3.38603467e-01 3.51801276e-01 1.76760063e-01 -5.39196789e-01
-1.15914607e+00 2.38386765e-02 -4.35092241e-01 9.98020396e-02
-3.45157236e-01 1.89565241e-01 -1.00988901e+00 -3.86112005e-01
-5.36551893e-01 -3.21984828e-01 1.06943703e+00 4.99654561e-01
1.14004827e+00 -3.48778188e-01 -8.18648458e-01 1.12888229e+00
1.21838462e+00 2.62986451e-01 4.45863515e-01 2.15264514e-01
9.08797860e-01 -4.57191579e-02 5.80935001e-01 4.41353947e-01
4.64645058e-01 9.39710855e-01 4.13867176e-01 3.70801240e-02
-4.60668534e-01 -9.27506506e-01 8.63255918e-01 4.05161530e-01
2.31775671e-01 -2.68723667e-01 -9.77539003e-01 3.46475959e-01
-1.93268573e+00 -1.19718063e+00 -1.36660680e-01 1.87523496e+00
7.13245630e-01 2.75374204e-01 5.13514876e-01 -5.88674903e-01
7.10266769e-01 3.83540630e-01 -8.88627172e-01 7.89554268e-02
1.01357110e-01 -1.89918429e-01 8.32717896e-01 3.12674701e-01
-1.50126040e+00 1.20805919e+00 6.82624912e+00 6.11744583e-01
-1.07517600e+00 1.90700144e-01 1.89346179e-01 -2.30589256e-01
1.40379906e-01 -9.11178142e-02 -1.49584568e+00 9.36537921e-01
1.07921648e+00 -1.53907537e-01 5.59205748e-02 1.23955929e+00
-2.83926781e-02 4.39429820e-01 -1.19086182e+00 9.72004473e-01
8.28374848e-02 -1.57731760e+00 -1.78981223e-03 7.72853717e-02
3.55888844e-01 4.50892508e-01 2.34558925e-01 6.22538447e-01
9.29465055e-01 -9.19912815e-01 1.19212270e+00 7.70467997e-01
9.47524846e-01 -2.78113723e-01 2.16498911e-01 4.65013571e-02
-1.71145296e+00 -2.62262188e-02 -2.22073287e-01 3.94300252e-01
5.00504136e-01 1.69017136e-01 -9.02382076e-01 2.68711060e-01
8.76743853e-01 1.09464264e+00 -5.88963985e-01 1.47482502e+00
-3.59470576e-01 7.04827189e-01 -1.46637345e-02 -1.53934389e-01
2.44090080e-01 9.36375931e-02 7.13567436e-01 1.58648515e+00
1.69323906e-01 -5.60543120e-01 3.33685398e-01 1.04462314e+00
-2.76028812e-01 -3.88899982e-01 -7.50822544e-01 -1.24575207e-02
5.14448225e-01 1.20016026e+00 -3.13743323e-01 -5.80063462e-01
-5.81117451e-01 6.61740184e-01 3.27220947e-01 2.65587181e-01
-1.60550809e+00 -2.97010899e-01 8.92395616e-01 6.35852143e-02
8.45690608e-01 -1.74966827e-01 -2.19029620e-01 -1.26784503e+00
-1.48850307e-01 -7.32593000e-01 2.63208449e-01 -8.44881892e-01
-1.19426787e+00 5.79592586e-01 -2.64687121e-01 -1.67014563e+00
-4.62131113e-01 -7.68059611e-01 -4.52915370e-01 6.43940210e-01
-1.45606041e+00 -1.35504782e+00 -3.89818788e-01 4.21889991e-01
6.26791060e-01 -2.20876530e-01 1.37051567e-01 2.53526419e-01
-4.81269240e-01 9.29093659e-01 6.81793615e-02 6.48069918e-01
7.18374312e-01 -1.36411047e+00 1.09640861e+00 9.29961622e-01
1.70194283e-01 3.55542421e-01 7.87187874e-01 -9.25452530e-01
-1.48178267e+00 -1.54054809e+00 8.38323712e-01 -8.66820097e-01
1.06216288e+00 -8.03504586e-01 -7.79691994e-01 1.14020288e+00
3.45411837e-01 3.62650096e-01 1.19966574e-01 -9.96406749e-02
-6.67768717e-01 -2.64231544e-02 -5.98047793e-01 6.46634877e-01
1.49789262e+00 -5.70620775e-01 -5.46335340e-01 2.27903679e-01
8.17160904e-01 -7.25862384e-01 -7.45606005e-01 3.32530767e-01
9.45097864e-01 -7.05575585e-01 1.12426472e+00 -8.91521811e-01
3.60168040e-01 -5.96107960e-01 2.10672423e-01 -9.07926202e-01
-5.25879681e-01 -9.38991606e-01 -7.79890954e-01 9.88944471e-01
4.17373776e-01 -3.67546558e-01 1.04581058e+00 2.24824250e-01
-2.06497833e-01 -6.44317389e-01 -6.25092506e-01 -1.25546348e+00
-1.50591448e-01 -2.22086459e-01 3.78609180e-01 4.56431150e-01
-3.34249645e-01 3.96286100e-01 -6.95721745e-01 -2.00728551e-02
1.13767695e+00 8.65483359e-02 9.22066391e-01 -1.18491340e+00
-2.17268527e-01 -5.74382126e-01 -4.64847237e-01 -1.54979253e+00
4.06410426e-01 -9.87796247e-01 4.78339672e-01 -1.06183791e+00
3.34885329e-01 -5.18329293e-02 -3.69014144e-02 5.52488804e-01
-1.50763363e-01 2.01837748e-01 5.84579349e-01 9.88156125e-02
-1.04623222e+00 8.43219697e-01 1.07294261e+00 -2.74799317e-01
1.87373221e-01 3.45575102e-02 -2.81047821e-01 8.19411576e-01
4.20157105e-01 -5.69256008e-01 -1.14393681e-01 -2.25036234e-01
-1.02810286e-01 8.15805718e-02 7.02032208e-01 -1.03292632e+00
5.96621633e-01 5.88577576e-02 7.77545691e-01 -9.69074190e-01
3.84789050e-01 -6.24271572e-01 2.77726352e-01 6.74553216e-01
-4.31303412e-01 2.24721551e-01 9.39683616e-02 8.30666184e-01
-6.70082867e-02 3.46435845e-01 7.51305163e-01 2.15384841e-01
-9.81599867e-01 6.17302775e-01 -8.39383155e-02 2.92538792e-01
1.20660996e+00 -3.31203729e-01 -6.07747674e-01 -2.28708655e-01
-6.27896369e-01 5.79581618e-01 4.16527480e-01 7.43572831e-01
4.70135868e-01 -1.51820648e+00 -6.56153798e-01 8.00109208e-02
8.32387730e-02 -2.85057098e-01 -3.95687334e-02 8.01775098e-01
-1.55171767e-01 7.42623746e-01 -2.45669231e-01 -1.02136779e+00
-1.37988210e+00 7.89569438e-01 5.14305472e-01 -2.49018937e-01
-9.11865234e-01 9.22622859e-01 6.54646933e-01 -1.18176714e-01
5.90255678e-01 -1.62753791e-01 7.02158511e-02 -5.18296659e-02
6.70296907e-01 2.26424173e-01 -4.89841551e-01 -6.60333693e-01
-5.04963934e-01 5.59971988e-01 -1.76075801e-01 1.01959385e-01
8.76342177e-01 9.92611349e-02 4.89027232e-01 2.85342157e-01
1.04542577e+00 -3.82003456e-01 -2.15162110e+00 -2.77670413e-01
1.99338615e-01 -4.93887842e-01 -2.73662567e-01 -7.36742377e-01
-1.15620255e+00 8.81796837e-01 5.62775552e-01 2.43111746e-03
6.89103901e-01 1.76215023e-01 7.86315084e-01 2.95574576e-01
9.70625356e-02 -7.29458392e-01 9.06308293e-02 7.98117518e-01
7.56402016e-01 -1.16895390e+00 -2.60408491e-01 -1.88327640e-01
-8.48500371e-01 7.65292943e-01 1.01565301e+00 -1.50145113e-01
5.18824339e-01 8.16704631e-01 -8.37226734e-02 -7.03957379e-02
-1.16328204e+00 -4.60280240e-01 4.68970478e-01 7.36558378e-01
5.22341728e-01 -1.13095582e-01 3.29858303e-01 3.59027267e-01
-2.54159808e-01 2.84402400e-01 -1.69808567e-02 4.86670345e-01
-4.31125820e-01 -7.47656941e-01 -2.27902144e-01 3.82536232e-01
-2.79402971e-01 -1.18223943e-01 -2.49987617e-01 8.40689957e-01
-1.17405489e-01 4.10936832e-01 2.62384534e-01 -4.71449703e-01
3.28489482e-01 2.13171855e-01 2.94606447e-01 -2.06157893e-01
-5.76155007e-01 -6.57871142e-02 -4.45545092e-02 -9.35171366e-01
9.80669353e-03 -7.43804097e-01 -1.17949080e+00 -4.01092499e-01
-7.83554837e-02 -1.20950349e-01 5.21470726e-01 6.51049137e-01
3.50998044e-01 7.68813252e-01 1.87871277e-01 -8.66984308e-01
-8.30962956e-01 -7.67472506e-01 -4.55386043e-02 2.05371723e-01
9.06204581e-01 -8.74941707e-01 9.63946059e-02 4.44801360e-01]
|
[6.289892673492432, -2.100856304168701]
|
b4a6e091-0a30-473d-83c1-fdc639e3df8d
|
neuromorphic-bayesian-optimization-in-lava
|
2305.11060
| null |
https://arxiv.org/abs/2305.11060v1
|
https://arxiv.org/pdf/2305.11060v1.pdf
|
Neuromorphic Bayesian Optimization in Lava
|
The ever-increasing demands of computationally expensive and high-dimensional problems require novel optimization methods to find near-optimal solutions in a reasonable amount of time. Bayesian Optimization (BO) stands as one of the best methodologies for learning the underlying relationships within multi-variate problems. This allows users to optimize time consuming and computationally expensive black-box functions in feasible time frames. Existing BO implementations use traditional von-Neumann architectures, in which data and memory are separate. In this work, we introduce Lava Bayesian Optimization (LavaBO) as a contribution to the open-source Lava Software Framework. LavaBO is the first step towards developing a BO system compatible with heterogeneous, fine-grained parallel, in-memory neuromorphic computing architectures (e.g., Intel's Loihi platform). We evaluate the algorithmic performance of the LavaBO system on multiple problems such as training state-of-the-art spiking neural network through back-propagation and evolutionary learning. Compared to traditional algorithms (such as grid and random search), we highlight the ability of LavaBO to explore the parameter search space with fewer expensive function evaluations, while discovering the optimal solutions.
|
['Maryam Parsa', 'Sumedh R. Risbud', 'Shay Snyder']
|
2023-05-18
| null | null | null | null |
['bayesian-optimization']
|
['methodology']
|
[-1.38622686e-01 -4.63017195e-01 3.16414118e-01 -2.50908524e-01
-5.34113526e-01 -4.04211104e-01 2.24026024e-01 2.10833699e-02
-9.94245946e-01 1.10397148e+00 -6.01301193e-01 -4.90352139e-02
-4.39437270e-01 -8.41086328e-01 -7.85585105e-01 -9.96379077e-01
-2.35173032e-01 9.11939621e-01 5.00204027e-01 8.81003402e-03
4.53358918e-01 6.73106015e-01 -1.94275188e+00 1.96069524e-01
5.64636230e-01 1.21312988e+00 4.96652871e-01 5.93097210e-01
-4.67123017e-02 1.14331171e-01 -5.36613703e-01 -1.24070056e-01
7.59205893e-02 -8.41097757e-02 -4.69463080e-01 -7.27939308e-01
4.26042974e-02 2.85218060e-01 -1.58832688e-02 8.95838499e-01
7.94252217e-01 2.87329942e-01 4.74204063e-01 -1.05625546e+00
2.89635230e-02 5.42144537e-01 -1.84787028e-02 5.08839548e-01
-2.63917357e-01 4.18856919e-01 5.90826273e-01 -9.35044944e-01
5.47352910e-01 1.25628912e+00 9.52598214e-01 5.35031736e-01
-1.74804091e+00 -7.30158985e-01 5.06063923e-02 4.80231881e-01
-1.60766733e+00 -5.38946033e-01 4.48301822e-01 -4.01027471e-01
1.60197985e+00 2.50612289e-01 1.08556235e+00 9.16233301e-01
6.62466586e-01 5.62218904e-01 1.24405146e+00 -2.77443677e-01
8.76239896e-01 -1.61279380e-01 2.91267306e-01 6.39374197e-01
2.95610547e-01 2.61290729e-01 -1.15626562e+00 -4.76412326e-01
3.95024031e-01 -5.26485205e-01 2.92229876e-02 -2.20182955e-01
-8.66664350e-01 7.78379023e-01 1.07086897e-01 2.17118561e-01
-4.37983900e-01 5.33785105e-01 8.20174664e-02 -8.67991373e-02
2.41272748e-01 9.49105084e-01 -6.70661449e-01 -4.84929323e-01
-1.21674907e+00 6.78057194e-01 1.11838090e+00 4.51636583e-01
8.49457681e-01 2.14392886e-01 -5.47178537e-02 7.33900666e-01
7.17061162e-01 1.90246284e-01 6.16226256e-01 -9.91149008e-01
-2.52926886e-01 1.38598800e-01 -1.18709497e-01 -6.43016696e-01
-6.21062458e-01 -7.16131747e-01 -5.09781659e-01 5.99526048e-01
4.62925524e-01 -1.91396192e-01 -8.00176859e-01 1.54031527e+00
2.98729718e-01 1.44766524e-01 -4.53680083e-02 6.52483225e-01
4.64024633e-01 7.39519954e-01 -1.74868464e-01 -2.63705552e-01
1.22918057e+00 -8.46380293e-01 -6.16903245e-01 -3.83029848e-01
2.39792064e-01 -4.91801411e-01 4.76763874e-01 8.26901436e-01
-1.20554638e+00 -2.60250568e-01 -1.33229816e+00 1.71189219e-01
-6.07531071e-01 -2.35943764e-01 8.71812642e-01 1.09723067e+00
-1.08514833e+00 1.00676894e+00 -1.47863543e+00 -7.83767272e-03
9.64440405e-01 8.31685960e-01 1.71552226e-01 1.74646392e-01
-7.76019275e-01 1.13754570e+00 7.04117239e-01 1.93925694e-01
-1.04949713e+00 -8.46459568e-01 -2.84895360e-01 -2.60953084e-02
2.03183502e-01 -5.78457952e-01 8.65281761e-01 -4.98912901e-01
-1.96010947e+00 6.89898551e-01 -2.31031507e-01 -6.54874265e-01
1.88610077e-01 -3.82294543e-02 3.13553773e-02 -3.57951581e-01
-5.20488203e-01 8.80688787e-01 7.04448044e-01 -8.14432561e-01
-2.34746471e-01 -6.14178002e-01 -4.86772716e-01 -2.21627295e-01
-3.28421861e-01 7.35868067e-02 -3.62722456e-01 -4.15980250e-01
1.26260325e-01 -9.79855239e-01 -3.04927915e-01 5.78007065e-02
1.20779403e-01 -1.39167652e-01 6.84943020e-01 -3.64011019e-01
1.12107050e+00 -1.80856800e+00 5.61058223e-01 3.96062702e-01
1.05087861e-01 2.22987205e-01 2.53736615e-01 1.25292540e-01
2.78536201e-01 -1.43543869e-01 -4.34758931e-01 -2.60959744e-01
6.73960224e-02 4.27443147e-01 -5.76204993e-02 4.90016371e-01
1.16046816e-01 8.52883697e-01 -5.69905639e-01 -2.67822474e-01
-5.38645722e-02 6.49924755e-01 -8.56846988e-01 1.03391714e-01
-5.09532630e-01 4.72723335e-01 -2.55692601e-01 7.18939066e-01
4.29567516e-01 -1.92431286e-01 8.36414546e-02 -4.98959310e-02
-5.20357728e-01 4.68398817e-02 -1.31618142e+00 1.85783017e+00
-3.29499274e-01 8.76080453e-01 3.50558087e-02 -1.00303268e+00
1.22600436e+00 -5.83087504e-02 4.74318624e-01 -5.56476593e-01
3.17686677e-01 4.72504169e-01 1.86753690e-01 -2.20162705e-01
2.42362559e-01 1.59548745e-01 3.14115107e-01 3.54946166e-01
5.26218176e-01 -3.49531859e-01 4.28610295e-01 -5.49099803e-01
1.04183042e+00 4.41374719e-01 1.74017981e-01 -7.77365804e-01
2.30038628e-01 8.27103294e-03 7.25329459e-01 9.49131966e-01
-1.31101340e-01 3.74568403e-01 2.45987266e-01 -6.14940763e-01
-8.86479914e-01 -9.37677860e-01 -6.61529601e-01 9.74278212e-01
-1.54389396e-01 -3.79581273e-01 -8.53831470e-01 3.09908926e-01
-6.71853945e-02 6.16420925e-01 -3.87174815e-01 3.97426449e-02
-4.93309885e-01 -1.71972227e+00 5.12553930e-01 1.84067279e-01
4.22345638e-01 -1.10568380e+00 -1.42126000e+00 6.21851504e-01
3.66682231e-01 -6.32292688e-01 3.42968762e-01 9.08681512e-01
-1.09231973e+00 -5.25950670e-01 -4.33607072e-01 -3.54751378e-01
1.86918736e-01 -6.43129885e-01 1.10640264e+00 -3.07427645e-01
-9.60613906e-01 -2.08580211e-01 1.23313643e-01 -6.55818224e-01
5.00561222e-02 1.32399455e-01 1.81095470e-02 -3.63709450e-01
4.25058424e-01 -9.34836328e-01 -5.43408453e-01 2.20504567e-01
-6.34605765e-01 1.18868239e-02 3.13378185e-01 1.12958515e+00
1.08067715e+00 8.02408084e-02 2.98597813e-01 -5.22811770e-01
7.29914248e-01 -4.04544652e-01 -1.32209551e+00 3.29251230e-01
-8.65378916e-01 5.26527941e-01 3.10642451e-01 -6.29509330e-01
-7.03397334e-01 1.33924887e-01 -2.54587382e-01 -2.72371441e-01
-1.77526344e-02 6.36117756e-01 1.53327748e-01 -5.29926240e-01
8.60736966e-01 1.68161333e-01 -7.10377544e-02 -3.44657034e-01
-1.24524102e-01 2.99266934e-01 3.78139734e-01 -8.81188393e-01
-6.39371946e-02 4.10499990e-01 2.52614796e-01 -5.77455103e-01
-4.03521568e-01 -5.58841154e-02 -4.05288279e-01 -5.27987599e-01
7.87951529e-01 -4.12157297e-01 -1.07327986e+00 8.30596507e-01
-1.14245343e+00 -5.57982504e-01 -9.95489508e-02 4.84063178e-01
-7.83518791e-01 -2.52555817e-01 -2.03625843e-01 -1.04963672e+00
-4.07723486e-01 -1.45973659e+00 6.77179754e-01 5.74025512e-01
-4.06540394e-01 -6.93421960e-01 4.10727769e-01 1.55037418e-01
7.44384706e-01 1.52111754e-01 8.64611864e-01 -4.24016923e-01
-5.63159585e-01 4.65047508e-02 1.92685977e-01 -1.31438643e-01
-6.74075902e-01 2.74507999e-01 -1.19412637e+00 -2.76121616e-01
2.00754389e-01 -4.88050103e-01 8.67539406e-01 7.49597192e-01
1.37478805e+00 1.87342361e-01 -3.40770900e-01 1.10424280e+00
1.47654343e+00 4.60599393e-01 5.23719668e-01 6.88482940e-01
1.24662735e-01 5.92573345e-01 2.58131385e-01 6.72998071e-01
6.20915927e-02 8.62849236e-01 5.57382405e-01 5.77076256e-01
1.23508625e-01 2.99151868e-01 1.70512125e-01 8.25584829e-01
-2.17536502e-02 -5.59367007e-04 -1.29965091e+00 3.03135037e-01
-2.12243080e+00 -6.75104380e-01 -9.36570577e-04 2.09337902e+00
1.11435413e+00 1.38590619e-01 -9.81016755e-02 -7.36034475e-03
3.93262446e-01 -2.84398913e-01 -1.01202238e+00 -4.72051352e-01
-2.41183370e-01 6.91818058e-01 5.12359798e-01 2.78627872e-01
-9.18853998e-01 8.12167525e-01 6.74252510e+00 1.09685278e+00
-1.18092918e+00 3.07600707e-01 6.02257907e-01 -6.46195054e-01
9.47371200e-02 -1.58666477e-01 -1.20513678e+00 6.43489242e-01
1.48369956e+00 8.07960238e-03 1.18556857e+00 8.04709911e-01
-1.95517782e-02 -3.44223380e-01 -9.98906195e-01 1.25095820e+00
-1.23006389e-01 -1.88223171e+00 -7.30753779e-01 -5.22704422e-02
7.95325696e-01 4.87986267e-01 6.35759756e-02 1.67432413e-01
3.71116340e-01 -1.30995190e+00 9.92753446e-01 8.28967392e-01
2.00817153e-01 -8.27249408e-01 6.32955074e-01 3.04994166e-01
-6.94523931e-01 -1.37916401e-01 -5.22066534e-01 -4.79632616e-02
4.30876054e-02 7.29976952e-01 -3.86282504e-01 -6.98191151e-02
1.32027435e+00 3.45678300e-01 -4.28612709e-01 1.46889973e+00
1.21524252e-01 4.58934695e-01 -7.66526222e-01 -5.40029347e-01
-2.92645511e-03 -3.16146642e-01 5.41832626e-01 1.05023694e+00
4.14521575e-01 -1.01137616e-01 -2.33725652e-01 1.33169305e+00
2.05428749e-01 -2.59939253e-01 -2.04168275e-01 -1.26594931e-01
6.41763151e-01 1.07558393e+00 -9.99885619e-01 1.03679739e-01
1.10166363e-01 2.99764663e-01 5.96027792e-01 1.79739147e-01
-5.66351593e-01 -4.30846751e-01 6.04463339e-01 -3.55356991e-01
4.81546879e-01 -4.89865839e-01 -8.50738525e-01 -7.43191004e-01
-2.32574359e-01 -9.51936007e-01 2.36697376e-01 -7.81677127e-01
-1.00542104e+00 6.63249433e-01 4.63101491e-02 -4.16060567e-01
-2.59979934e-01 -9.51656520e-01 -3.41028810e-01 9.21012700e-01
-1.33993876e+00 -5.28521478e-01 -1.25558972e-01 4.52712417e-01
2.92638987e-01 -4.05691028e-01 1.20854425e+00 6.72405362e-02
-6.77703083e-01 3.84500653e-01 5.85592091e-01 -6.78860426e-01
2.70466626e-01 -9.60012734e-01 1.57219902e-01 6.04475558e-01
1.98555753e-01 6.10996723e-01 8.93459797e-01 -4.41109508e-01
-1.83681285e+00 -4.24037129e-01 3.11249882e-01 -9.17229429e-02
7.28896797e-01 -5.35969257e-01 -1.03488398e+00 1.06476560e-01
1.40275098e-02 6.64585903e-02 8.04343581e-01 2.62926757e-01
-1.03371866e-01 -3.24957073e-01 -1.14759290e+00 5.34609914e-01
7.41244495e-01 -2.13465601e-01 -3.45287770e-01 2.34611496e-01
1.78389810e-02 -4.21117455e-01 -1.00052369e+00 4.05747861e-01
9.56165969e-01 -1.00659847e+00 9.50874984e-01 -2.18212008e-01
-5.34939207e-02 -2.56961614e-01 -1.66531861e-01 -1.19296169e+00
1.76607296e-02 -8.86077225e-01 -4.76467431e-01 7.00660467e-01
3.39882195e-01 -8.03398728e-01 8.95908654e-01 7.42393613e-01
-1.51210174e-01 -1.18598509e+00 -1.42236531e+00 -8.30194831e-01
5.51473573e-02 -7.27064133e-01 4.38423365e-01 3.00570756e-01
-1.43149719e-01 -2.38982186e-01 5.71973585e-02 2.27300394e-02
7.63206422e-01 3.62378433e-02 4.49328572e-01 -1.28808165e+00
-8.08109879e-01 -8.10036302e-01 -5.54186106e-01 -4.36638325e-01
2.02743039e-01 -7.36318409e-01 3.73408288e-01 -9.69378352e-01
5.68642765e-02 -5.63224256e-01 -3.10982823e-01 3.25612813e-01
1.10065363e-01 1.15753010e-01 -1.53475329e-01 2.37809494e-01
-3.82680178e-01 5.92521071e-01 6.18116677e-01 -4.03122827e-02
-1.65121183e-01 -2.21467838e-01 -8.12480003e-02 7.11330354e-01
7.98226893e-01 -9.89288449e-01 -1.16327263e-01 -7.13383555e-01
5.17598867e-01 -3.36059928e-01 3.06146920e-01 -1.36897004e+00
7.35535860e-01 -4.70010117e-02 3.10940862e-01 -5.44634879e-01
7.58474410e-01 -4.77881521e-01 5.66278458e-01 6.89619839e-01
-1.44518673e-01 3.46481875e-02 6.79166377e-01 3.42272669e-01
-7.39867687e-02 -7.68531322e-01 9.99162018e-01 -1.13121718e-01
-5.92968225e-01 1.00051731e-01 -7.53600419e-01 -7.26428106e-02
1.05911815e+00 -2.85059005e-01 -3.76279414e-01 3.20277035e-01
-7.10033536e-01 1.58270985e-01 2.15463027e-01 -6.48251995e-02
5.90489745e-01 -7.62411833e-01 -4.99664873e-01 2.95486480e-01
-2.45750561e-01 4.98177148e-02 1.26619831e-01 6.73465788e-01
-7.87595451e-01 4.26756322e-01 -7.17366636e-01 -8.79240394e-01
-1.15991127e+00 8.79599527e-02 5.90448797e-01 -3.39105546e-01
-2.08230272e-01 1.41548526e+00 -5.68196297e-01 -3.13974202e-01
2.87482798e-01 3.35310996e-02 3.91694866e-02 4.55208980e-02
3.96606952e-01 5.16095519e-01 3.76351088e-01 5.09139672e-02
-4.70528841e-01 3.66630197e-01 9.37472582e-02 -6.24640048e-01
1.89395154e+00 3.88697207e-01 -4.56009239e-01 7.88531899e-01
7.66501427e-01 -6.96726084e-01 -1.40302157e+00 8.89537185e-02
2.37538591e-01 -2.55727828e-01 7.03368187e-01 -7.81167924e-01
-8.47757161e-01 9.71126080e-01 1.07116067e+00 -1.47840664e-01
1.00165784e+00 -3.01634312e-01 2.67927498e-01 7.82588542e-01
6.24634206e-01 -1.34985602e+00 -1.47894919e-01 6.48620069e-01
7.99502671e-01 -8.36192846e-01 2.31263861e-01 1.39613539e-01
1.02244578e-02 1.39805365e+00 5.75400770e-01 -2.10408315e-01
7.74748325e-01 9.12019074e-01 -3.38791549e-01 -3.37875456e-01
-1.14423108e+00 6.90888688e-02 1.32587388e-01 5.79444408e-01
1.99415296e-01 -1.20097771e-01 -2.54480720e-01 5.57476878e-01
-3.33964437e-01 3.64839323e-02 2.46987119e-02 1.14026344e+00
-4.09487724e-01 -1.17363524e+00 -5.69845259e-01 5.28530478e-01
-5.08531272e-01 -2.41898134e-01 1.25747457e-01 3.30607176e-01
1.51996270e-01 6.71257973e-01 7.06242546e-02 -1.29285544e-01
-1.33181080e-01 3.12494665e-01 8.60688567e-01 -3.48136872e-01
-9.61418092e-01 7.21933171e-02 -4.19524834e-02 -7.12026060e-01
-2.50430465e-01 -9.65763927e-01 -1.10830200e+00 -1.10184081e-01
-3.24681133e-01 -1.01465508e-01 1.48105133e+00 1.06919599e+00
4.41327006e-01 7.87117660e-01 -1.33410737e-01 -1.26977742e+00
-4.35536414e-01 -7.32100546e-01 -2.98372239e-01 -4.36820865e-01
-2.45788530e-01 -8.74880254e-01 -1.42741418e-02 -1.64880887e-01]
|
[7.871139049530029, 3.109462022781372]
|
8c07f702-32bc-4dbf-91b4-40c81524c1ef
|
similarity-preserving-unsupervised-feature
| null | null |
https://ieeexplore.ieee.org/abstract/document/9345884
|
https://ieeexplore.ieee.org/abstract/document/9345884
|
Similarity Preserving Unsupervised Feature Selection based on Sparse Learning
|
Various feature selection methods have been recently proposed on different applications to reduce the computational burden of machine learning algorithms as well as the complexity of learned models. Preserving sample similarities and selecting discriminative features are two major factors should be satisfied, especially by unsupervised feature selection methods. This paper aims to propose a novel unsupervised feature selection approach which employs an ℓ 2,1 -norm regularization model to preserve global and local similarities by minimizing an objective function. Cluster analysis is also incorporated in this framework to take the inherent structure of the data into account. The experimental results show the strength of the proposed approach as compared with the earlier well-known methods on a variety of standard datasets.
|
['Sasan H. Alizadeh', 'Mehdi Ghatee', 'Mohsen Ghassemi Parsa', 'Hadi Zare']
|
2020-12-15
| null | null | null |
10th-international-symposium-on
|
['sparse-learning']
|
['methodology']
|
[ 2.66424805e-01 -4.54380512e-01 -5.89128435e-02 -6.28584683e-01
-4.11510170e-01 -1.44386664e-02 4.06658977e-01 3.20412368e-01
-6.44675195e-01 7.51984477e-01 9.88370255e-02 5.23502171e-01
-7.21502900e-01 -6.31292939e-01 1.21297516e-01 -9.20232952e-01
-1.07767425e-01 2.13334233e-01 1.76720828e-01 1.04051217e-01
7.13989437e-01 6.38841212e-01 -1.89312577e+00 -2.17633426e-01
8.04085076e-01 8.56139660e-01 2.99956650e-01 1.34569138e-01
6.77661598e-02 3.53512287e-01 -2.39263192e-01 1.34725720e-01
4.74331528e-01 -4.38909501e-01 -5.73866129e-01 6.76345825e-01
1.33082032e-01 1.97592393e-01 6.50797337e-02 1.14128137e+00
4.59803492e-01 7.56696343e-01 8.28593016e-01 -9.29493785e-01
-1.75863564e-01 1.80993572e-01 -7.03979254e-01 1.86464295e-01
9.95136127e-02 -1.69688404e-01 1.02732480e+00 -1.09453964e+00
5.00136673e-01 8.86511922e-01 2.83468306e-01 -9.15451050e-02
-1.24578083e+00 -4.15787399e-01 -8.88338089e-02 4.19058710e-01
-1.69919658e+00 -3.04122716e-01 9.27266300e-01 -3.64625931e-01
7.33304083e-01 3.83195013e-01 3.71270508e-01 1.26790807e-01
2.24056616e-01 5.18654108e-01 1.07115841e+00 -6.13353610e-01
2.87992775e-01 4.21934187e-01 5.15223324e-01 6.92838907e-01
3.98979187e-01 -9.57866013e-02 -4.45244074e-01 -4.01478052e-01
2.65361547e-01 1.97097391e-01 -5.26641048e-02 -7.67965496e-01
-7.53989458e-01 1.10675085e+00 6.82461783e-02 5.19870281e-01
-4.66273069e-01 -3.97327542e-01 5.87201834e-01 2.51269072e-01
3.08565617e-01 3.21174294e-01 -2.42621511e-01 1.39992088e-01
-1.00846827e+00 5.73739484e-02 3.36438090e-01 7.05577135e-01
1.04029548e+00 3.24902609e-02 1.93483695e-01 8.59552443e-01
4.55028445e-01 3.15024666e-02 8.14385533e-01 -4.18078899e-01
1.28337160e-01 1.06795704e+00 -1.54944092e-01 -1.35755706e+00
-4.25752550e-01 -6.70280814e-01 -7.56117284e-01 3.42436135e-01
-3.04922182e-02 1.39229909e-01 -5.57398736e-01 1.19145596e+00
5.79636633e-01 -4.75290883e-03 -1.06014192e-01 7.55845368e-01
6.02081895e-01 4.59926218e-01 -7.28659006e-03 -5.61281741e-01
8.86473358e-01 -5.97248554e-01 -6.69979095e-01 2.55287468e-01
4.78400648e-01 -1.06345057e+00 7.56389856e-01 5.29629230e-01
-6.19193971e-01 -6.75696075e-01 -1.05003154e+00 3.95149499e-01
-3.53366554e-01 4.38558102e-01 6.59563303e-01 6.69273376e-01
-4.95586812e-01 4.66567695e-01 -7.49382257e-01 -4.63468313e-01
1.62657976e-01 8.93120289e-01 -5.79168081e-01 1.36005610e-01
-8.36105108e-01 6.11594200e-01 6.29316807e-01 2.93032676e-01
-3.74173820e-01 -2.10063502e-01 -6.63658977e-01 7.96868429e-02
2.98904926e-01 -2.83146322e-01 5.12561917e-01 -1.12022579e+00
-1.35972273e+00 4.05657113e-01 -2.07616001e-01 -2.69011617e-01
3.14771354e-01 -1.97142020e-01 -4.10167903e-01 6.82744533e-02
-6.48920164e-02 -6.16819225e-02 7.16047347e-01 -8.50448668e-01
-7.12779582e-01 -3.48000169e-01 -6.44893646e-01 3.99384141e-01
-6.41951859e-01 2.78796732e-01 -1.40208632e-01 -6.63264155e-01
4.58957881e-01 -7.12799251e-01 -3.61092478e-01 -2.18947902e-01
-2.27250248e-01 -2.26204634e-01 1.00513613e+00 -3.51533324e-01
1.33241248e+00 -2.03726482e+00 2.65758991e-01 9.30237830e-01
-1.09628946e-01 2.92239785e-01 9.12404731e-02 6.73787653e-01
3.07499785e-02 -3.32473457e-01 -4.18542713e-01 -1.96746454e-01
-2.43447646e-01 -5.37766255e-02 2.67489105e-01 9.43506598e-01
1.78858548e-01 -7.32626021e-02 -4.60272670e-01 -5.07164240e-01
4.91127491e-01 5.66471696e-01 -3.57628047e-01 9.96237323e-02
3.43710065e-01 3.46865058e-01 -6.40634835e-01 3.14821661e-01
6.97244585e-01 2.55391896e-01 2.35236194e-02 -1.08923346e-01
-2.23208368e-01 -1.13318130e-01 -1.81837845e+00 1.26154149e+00
-3.80219072e-02 3.27590823e-01 2.56560519e-02 -1.47905755e+00
1.24864984e+00 2.76443243e-01 7.82372534e-01 -3.07904005e-01
4.06272411e-01 3.30742300e-01 3.63591034e-03 -5.10616481e-01
3.69850278e-01 -2.30524242e-01 6.82225227e-02 2.62589425e-01
1.06066428e-01 1.90891668e-01 2.54609495e-01 -1.58433065e-01
6.18041158e-01 -1.99836731e-01 8.55682194e-01 -6.74267411e-01
1.00142860e+00 -6.46014735e-02 8.08870971e-01 3.60955685e-01
-6.85770437e-02 4.90180165e-01 -7.68440962e-02 -4.28313851e-01
-5.90244472e-01 -3.97047102e-01 -3.05619955e-01 7.31512487e-01
1.10326491e-01 -2.61702985e-01 -3.48552942e-01 -7.70388424e-01
-1.80105846e-02 4.05566186e-01 -5.25683820e-01 -1.90853551e-01
-4.60263491e-01 -8.26659679e-01 -1.68421604e-02 1.60087675e-01
4.35160130e-01 -8.79903257e-01 -6.76405907e-01 2.67650187e-01
5.67044914e-01 -5.88689983e-01 -2.42177621e-01 2.91289747e-01
-1.14039564e+00 -1.09237540e+00 -3.05643588e-01 -1.04382646e+00
8.96895885e-01 4.30649370e-01 5.56122303e-01 1.86030403e-01
-6.23008490e-01 2.41046175e-02 -6.16864622e-01 -1.33034557e-01
1.57881260e-01 2.34018743e-01 1.63461357e-01 5.73076308e-01
6.55855000e-01 -5.66850483e-01 -3.67533296e-01 4.14660096e-01
-9.87678528e-01 -3.54297042e-01 6.81159496e-01 1.07479286e+00
7.33714163e-01 6.36232793e-01 7.11180806e-01 -1.06957150e+00
6.66724920e-01 -4.49081331e-01 -7.41060436e-01 2.29202822e-01
-7.84003258e-01 1.29974559e-01 7.35917628e-01 -9.30103064e-02
-1.17183185e+00 5.40125787e-01 2.05069125e-01 1.01132644e-03
-1.96042031e-01 6.43980443e-01 -3.98098111e-01 -4.80849475e-01
3.35738868e-01 6.02500856e-01 1.60962656e-01 -5.95290661e-01
-1.48346946e-01 6.16143882e-01 -3.47180963e-02 -3.30795765e-01
8.03520083e-01 3.76398355e-01 3.70211601e-01 -1.05027366e+00
-2.24901512e-01 -9.25960660e-01 -9.26791251e-01 1.17177732e-01
4.06851888e-01 -4.72936153e-01 -3.68949324e-01 2.27660134e-01
-3.86350989e-01 4.94846463e-01 -7.60165751e-02 9.98883784e-01
-3.97915363e-01 6.30099893e-01 -5.31206792e-03 -8.82208765e-01
-4.30570811e-01 -1.03156483e+00 4.00002927e-01 3.00112545e-01
-1.80641025e-01 -9.97475147e-01 2.01254100e-01 -6.45423010e-02
4.49500918e-01 2.82765418e-01 7.43247926e-01 -9.25329089e-01
-1.26536712e-01 -5.32517612e-01 8.73476937e-02 5.22964835e-01
6.39584243e-01 1.55442730e-01 -4.98874664e-01 -4.63650078e-01
4.09799635e-01 -1.97781503e-01 8.85107934e-01 2.16731057e-01
8.71167123e-01 -1.14939563e-01 -2.95020491e-01 4.01230097e-01
1.62226677e+00 3.45847815e-01 3.89479339e-01 2.90861487e-01
4.09931540e-01 6.85861349e-01 1.03165781e+00 6.76850259e-01
-2.44319201e-01 5.00647724e-01 1.43569618e-01 -2.17986945e-02
2.78948188e-01 2.14865446e-01 5.74753433e-02 1.05564404e+00
9.67175327e-03 2.34301329e-01 -5.74727476e-01 5.54891527e-01
-1.99415076e+00 -1.00754082e+00 -1.67985857e-01 2.38229465e+00
4.83760118e-01 -3.73282935e-03 7.73324892e-02 5.73096812e-01
7.65700340e-01 -3.46189365e-02 -2.58290887e-01 -3.57786655e-01
-2.09941387e-01 4.46369648e-01 3.04630220e-01 5.00737786e-01
-1.38764632e+00 5.31216502e-01 5.94476223e+00 8.51224899e-01
-1.13863254e+00 -2.77201116e-01 2.84705281e-01 -1.08791128e-01
3.86160575e-02 1.69964552e-01 -7.00988233e-01 3.97382051e-01
3.61571640e-01 -2.33560383e-01 3.78149468e-03 8.04487348e-01
3.83842885e-01 -3.24880034e-01 -6.84442699e-01 9.79017675e-01
2.38327578e-01 -7.83237815e-01 -3.06207910e-02 -1.41228884e-01
7.00977445e-01 -3.70425880e-01 2.86922250e-02 -1.37681991e-01
-3.89440835e-01 -7.81426430e-01 1.64686188e-01 5.35757363e-01
1.98309898e-01 -1.20739484e+00 9.79508400e-01 1.57433584e-01
-1.45279181e+00 -1.86767876e-02 -5.83667755e-01 -1.45248145e-01
-8.46549273e-02 8.24707031e-01 -7.29779303e-01 6.97118819e-01
3.34731579e-01 6.72438741e-01 -7.63324082e-01 1.52082896e+00
1.57227799e-01 4.96967137e-01 -5.41298449e-01 -2.58901030e-01
2.22243369e-01 -4.80102450e-01 6.27439916e-01 1.07953691e+00
3.65849942e-01 4.26293723e-02 3.93100560e-01 3.52504730e-01
2.53009826e-01 9.04595375e-01 -4.44865137e-01 6.20715618e-02
2.87289768e-01 1.34798229e+00 -9.21991944e-01 -1.20377079e-01
-5.43866217e-01 8.10997963e-01 1.28661305e-01 -9.57159102e-02
-2.80065715e-01 -8.26302826e-01 2.65029103e-01 2.05027208e-01
3.24012250e-01 -3.06556404e-01 -1.53061345e-01 -9.89839256e-01
5.14072143e-02 -7.44127333e-01 6.00487888e-01 1.84622154e-01
-1.18512583e+00 6.09921694e-01 2.91348957e-02 -1.51534593e+00
-7.56395087e-02 -4.49483126e-01 -6.23967111e-01 7.09684610e-01
-1.36032903e+00 -8.59706402e-01 -9.52955857e-02 7.71803796e-01
7.97141492e-01 -5.94615281e-01 7.53315091e-01 3.61073762e-01
-5.36412001e-01 4.81557935e-01 5.71706474e-01 -3.51044863e-01
8.57882738e-01 -8.45287442e-01 -4.85359818e-01 9.74101305e-01
1.82429537e-01 7.93770790e-01 6.70785129e-01 -6.67724550e-01
-1.29983854e+00 -7.39998400e-01 9.81680870e-01 4.06946331e-01
2.00233921e-01 5.36887906e-04 -8.98036063e-01 3.43152344e-01
1.85219973e-01 1.18225425e-01 1.10269165e+00 7.45661259e-02
6.53748661e-02 -3.20692211e-01 -1.39113450e+00 2.18888491e-01
5.21250665e-01 -1.28259867e-01 -4.32512105e-01 1.35181382e-01
-2.98145115e-01 2.31524989e-01 -7.91671634e-01 4.86559480e-01
4.82490838e-01 -1.01661527e+00 5.13742864e-01 -3.45595568e-01
-3.29926312e-01 -6.49554312e-01 -1.54086605e-01 -1.06717908e+00
-5.27345717e-01 -4.33280885e-01 2.67853171e-01 1.33427179e+00
2.55421311e-01 -6.47865236e-01 8.70801628e-01 5.84453166e-01
2.66981006e-01 -7.22695947e-01 -8.50875676e-01 -6.22690141e-01
-5.66176534e-01 7.08418116e-02 2.94274598e-01 9.03698623e-01
-3.24140675e-02 3.08304846e-01 -4.52478737e-01 -1.21264890e-01
7.58655667e-01 2.12895676e-01 5.30242562e-01 -1.44583547e+00
-2.33142197e-01 -3.68859887e-01 -9.30866838e-01 -1.03661425e-01
-5.03883511e-02 -6.90172374e-01 -1.85466111e-01 -1.08950293e+00
3.35566968e-01 -3.62416476e-01 -7.12043226e-01 1.85491607e-01
-2.27846086e-01 1.01722382e-01 -4.49570827e-02 3.74625683e-01
-4.48960036e-01 6.85317278e-01 6.18524849e-01 1.34154558e-01
-4.52649593e-01 2.65622735e-01 -4.99845266e-01 7.02254534e-01
9.04831648e-01 -6.18017673e-01 -6.57822132e-01 -1.14361078e-01
-2.46278390e-01 -4.81136858e-01 -3.06058735e-01 -1.15174067e+00
3.44841808e-01 -3.54991823e-01 5.68358958e-01 -3.60335737e-01
1.40575111e-01 -1.20029283e+00 3.94820422e-01 4.80914056e-01
-4.98098433e-01 9.19227675e-02 -8.72738287e-02 4.82335240e-01
-6.69072747e-01 -5.44315577e-01 9.87062156e-01 -8.77648294e-02
-8.80274892e-01 5.74834272e-02 -2.95624137e-01 -5.43318570e-01
1.31351614e+00 -4.40959871e-01 3.52460027e-01 6.60844753e-03
-6.23458982e-01 1.94077194e-02 2.48972371e-01 9.43973139e-02
8.39524329e-01 -1.26454556e+00 -6.75138712e-01 4.60379213e-01
1.81264669e-01 -3.95573348e-01 2.15887010e-01 8.75506222e-01
-4.71564919e-01 3.30385983e-01 -5.14117956e-01 -3.01071525e-01
-1.73020840e+00 4.38788474e-01 -8.57118145e-02 -3.07657093e-01
-4.57339704e-01 6.26082718e-01 -2.06863374e-01 -2.06856579e-01
2.55152673e-01 1.93494648e-01 -6.59529507e-01 2.38963246e-01
4.26101118e-01 5.32442033e-01 2.18304127e-01 -8.48372161e-01
-6.72133625e-01 8.45846534e-01 -2.93681294e-01 1.58935457e-01
1.55102909e+00 -8.33557621e-02 -3.35566610e-01 2.78531611e-01
1.42987633e+00 2.08054781e-01 -7.86082089e-01 -3.24857593e-01
3.63636136e-01 -7.37662554e-01 1.51392996e-01 -2.70347118e-01
-1.03667343e+00 6.26925409e-01 8.52177978e-01 -4.95131081e-03
1.28682506e+00 -5.87244272e-01 3.93253505e-01 3.85714859e-01
2.90803283e-01 -1.39327347e+00 -2.96601087e-01 1.43181741e-01
7.00452507e-01 -1.34472859e+00 4.58276153e-01 -4.93076265e-01
-6.45695448e-01 1.41287851e+00 4.35381055e-01 -6.40136957e-01
8.32960427e-01 -1.14527255e-01 -2.28515610e-01 -1.53888541e-03
-4.50905323e-01 -2.88706064e-01 5.05222678e-01 2.45872378e-01
6.19875371e-01 -1.13479204e-01 -1.17817616e+00 5.03938079e-01
2.90669352e-01 1.84779298e-02 9.12915394e-02 1.27681696e+00
-6.97810888e-01 -1.38158166e+00 -3.10964525e-01 5.15231133e-01
-5.29508650e-01 2.89713413e-01 -3.75113517e-01 7.29428947e-01
1.55892223e-01 8.54669511e-01 -3.08786273e-01 -5.12869179e-01
1.75647497e-01 -7.61275664e-02 2.07818538e-01 -6.26888454e-01
-8.88906538e-01 3.79638046e-01 -1.39284030e-01 -2.97823310e-01
-7.64245450e-01 -6.89762175e-01 -1.17988551e+00 1.51578903e-01
-7.30002224e-01 6.15953445e-01 6.99259758e-01 8.50032926e-01
4.20107931e-01 1.64652050e-01 9.79576111e-01 -7.32583225e-01
-6.59383118e-01 -8.98811042e-01 -8.48089397e-01 4.62127119e-01
-8.34061578e-02 -8.38734508e-01 -4.41750348e-01 2.31501404e-02]
|
[8.210809707641602, 4.1976494789123535]
|
57cb3c06-489a-4cba-ad14-c83c8f8bcad7
|
hierarchical-aggregation-for-3d-instance
|
2108.02350
| null |
https://arxiv.org/abs/2108.02350v1
|
https://arxiv.org/pdf/2108.02350v1.pdf
|
Hierarchical Aggregation for 3D Instance Segmentation
|
Instance segmentation on point clouds is a fundamental task in 3D scene perception. In this work, we propose a concise clustering-based framework named HAIS, which makes full use of spatial relation of points and point sets. Considering clustering-based methods may result in over-segmentation or under-segmentation, we introduce the hierarchical aggregation to progressively generate instance proposals, i.e., point aggregation for preliminarily clustering points to sets and set aggregation for generating complete instances from sets. Once the complete 3D instances are obtained, a sub-network of intra-instance prediction is adopted for noisy points filtering and mask quality scoring. HAIS is fast (only 410ms per frame) and does not require non-maximum suppression. It ranks 1st on the ScanNet v2 benchmark, achieving the highest 69.9% AP50 and surpassing previous state-of-the-art (SOTA) methods by a large margin. Besides, the SOTA results on the S3DIS dataset validate the good generalization ability. Code will be available at https://github.com/hustvl/HAIS.
|
['Xinggang Wang', 'Wenyu Liu', 'Qian Zhang', 'Jiemin Fang', 'Shaoyu Chen']
|
2021-08-05
| null |
http://openaccess.thecvf.com//content/ICCV2021/html/Chen_Hierarchical_Aggregation_for_3D_Instance_Segmentation_ICCV_2021_paper.html
|
http://openaccess.thecvf.com//content/ICCV2021/papers/Chen_Hierarchical_Aggregation_for_3D_Instance_Segmentation_ICCV_2021_paper.pdf
|
iccv-2021-1
|
['3d-instance-segmentation-1']
|
['computer-vision']
|
[ 1.09290443e-01 3.28083187e-02 6.87213540e-02 -4.39190984e-01
-8.36386919e-01 -4.52558368e-01 5.02008915e-01 2.55161107e-01
-2.60159880e-01 2.93256611e-01 -4.48158145e-01 -1.64754272e-01
-1.57482460e-01 -8.44263196e-01 -8.49412560e-01 -6.51276648e-01
-2.22846419e-02 7.50423789e-01 7.76953578e-01 5.13694957e-02
3.72215003e-01 8.81114960e-01 -1.91661572e+00 2.37252519e-01
1.05133927e+00 1.23926890e+00 4.85793591e-01 2.16287807e-01
-4.46066320e-01 6.81434050e-02 -5.15201449e-01 -2.10151404e-01
4.72016186e-01 2.56088506e-02 -6.50570154e-01 4.72330421e-01
4.45099115e-01 -1.73872188e-02 2.20036432e-01 9.33688641e-01
3.80509555e-01 1.92859098e-01 5.75977147e-01 -1.15636230e+00
-8.13709050e-02 3.75775009e-01 -7.79412568e-01 8.01912770e-02
9.66834724e-02 1.37401149e-01 9.58480239e-01 -1.29064131e+00
4.88427848e-01 1.08815789e+00 5.94239771e-01 2.74941683e-01
-1.25988972e+00 -6.09623134e-01 2.98363626e-01 2.56915390e-01
-1.75302088e+00 -1.73926294e-01 6.93614960e-01 -5.18891871e-01
8.31797242e-01 5.54473877e-01 7.46129930e-01 4.15656358e-01
-3.81007969e-01 8.35413575e-01 1.12089515e+00 -8.90662372e-02
5.04548967e-01 -1.50653288e-01 1.37084484e-01 3.25471371e-01
3.36614221e-01 -6.78882897e-02 -3.76723170e-01 -7.12444782e-02
8.04836571e-01 -9.29688278e-04 -6.70699924e-02 -4.35231179e-01
-1.38916945e+00 5.60639024e-01 7.96235144e-01 1.48787379e-01
-7.19397426e-01 -1.28181875e-01 1.33931369e-01 -8.98347721e-02
4.68303353e-01 2.69377887e-01 -3.18703830e-01 1.71399623e-01
-1.16918540e+00 2.93150097e-01 3.18318486e-01 1.18869054e+00
9.16343987e-01 -2.83690095e-01 -1.57048523e-01 8.83005857e-01
9.17423368e-02 4.00583923e-01 -9.05721262e-02 -9.72949564e-01
4.61796671e-01 9.52948034e-01 -1.04225357e-03 -8.14622223e-01
-4.81209159e-01 -6.23916745e-01 -8.70602012e-01 3.07858050e-01
1.97267845e-01 3.69521827e-01 -1.21881187e+00 1.03453803e+00
7.19652891e-01 5.29897094e-01 -2.37201661e-01 1.00126803e+00
1.09861481e+00 6.51315391e-01 -6.22139014e-02 -1.21717326e-01
1.15172470e+00 -6.78497851e-01 -1.83336645e-01 -9.43179354e-02
2.86259145e-01 -6.50802672e-01 9.23769772e-01 5.44952929e-01
-1.22229910e+00 -8.58009458e-01 -8.11072230e-01 1.11779630e-01
-1.65572122e-01 5.09903617e-02 3.76140475e-01 3.30650091e-01
-8.98491383e-01 6.05174899e-01 -1.03407848e+00 -2.53242821e-01
9.08762038e-01 4.22573119e-01 -1.67752534e-01 -7.56710246e-02
-5.12121797e-01 3.53126198e-01 5.67201555e-01 5.59520945e-02
-4.85035658e-01 -9.50955451e-01 -5.54562390e-01 -7.87529573e-02
5.70644617e-01 -5.29094040e-01 9.91025925e-01 -3.93188953e-01
-1.09451759e+00 8.38086784e-01 -2.82582045e-01 -5.92294157e-01
5.37785470e-01 -1.39248133e-01 -1.20521188e-01 2.28710398e-01
1.76727161e-01 1.04154050e+00 6.18046403e-01 -1.68314815e+00
-1.02213562e+00 -5.31426966e-01 -9.54620987e-02 2.83012331e-01
1.18345112e-01 -3.87925506e-01 -1.02710903e+00 -3.77936959e-01
9.02953982e-01 -9.99102712e-01 -5.20144284e-01 -1.60938352e-01
-7.19612598e-01 -4.27984357e-01 5.99852860e-01 -2.75757015e-01
1.19827640e+00 -2.26403260e+00 -8.37316066e-02 6.35531425e-01
2.88332909e-01 2.34012857e-01 9.11209434e-02 2.24460080e-01
2.06089433e-04 1.61301747e-01 -6.34086728e-01 -4.85629976e-01
2.41111014e-02 1.56055227e-01 -1.10015325e-01 4.91012901e-01
3.22446346e-01 7.24780440e-01 -7.43462980e-01 -5.76736510e-01
7.60065079e-01 3.56702417e-01 -6.11058891e-01 -1.01912737e-01
-3.86743367e-01 5.42137980e-01 -3.99862498e-01 8.35223317e-01
1.18634248e+00 -4.05654997e-01 -2.83822268e-01 -3.48911762e-01
-3.19945812e-01 2.58338481e-01 -1.53027415e+00 1.72919345e+00
-8.32625628e-02 2.07206592e-01 -1.36383891e-01 -7.93161035e-01
1.09635806e+00 -7.60462927e-03 8.63564551e-01 -5.76814771e-01
-5.51704578e-02 4.03357238e-01 -3.11757475e-02 -1.31332442e-01
4.16031152e-01 2.09320173e-01 7.95950964e-02 -8.69761631e-02
-1.71269163e-01 -5.12951434e-01 3.81956607e-01 1.14629202e-01
8.50922525e-01 6.37621135e-02 9.11375135e-02 -2.74876088e-01
4.86300468e-01 2.84063250e-01 5.73087692e-01 7.50362813e-01
-3.48245725e-02 1.04835665e+00 1.38190567e-01 -2.56890535e-01
-9.29763615e-01 -1.19792986e+00 -4.66087312e-01 5.20932674e-01
5.53873062e-01 -3.09566587e-01 -7.28275239e-01 -5.89317083e-01
1.75662607e-01 8.67013812e-01 -3.35283041e-01 2.53713161e-01
-4.01693046e-01 -5.26005626e-01 9.29657072e-02 4.25717026e-01
5.21721900e-01 -1.02318919e+00 -5.28772473e-01 2.88602740e-01
-1.32461682e-01 -1.24650526e+00 -5.26426956e-02 1.60067633e-01
-1.06520486e+00 -9.83352482e-01 -5.07762492e-01 -5.29178083e-01
7.65302062e-01 4.74487692e-01 1.28147256e+00 1.66241214e-01
-1.39361992e-01 1.95021480e-02 -4.99200702e-01 -4.63018417e-01
-2.17061080e-02 2.44609982e-01 -1.58251539e-01 -7.58734122e-02
4.77197707e-01 -6.00658655e-01 -7.69902408e-01 6.43373072e-01
-8.00120950e-01 1.76258743e-01 5.94099641e-01 3.74882460e-01
1.30870473e+00 1.68204695e-01 3.24189425e-01 -7.72909224e-01
3.84504497e-02 -4.80255187e-01 -7.95117497e-01 -2.11383954e-01
-3.94148886e-01 -4.03143942e-01 1.91813603e-01 -4.57032882e-02
-7.27843463e-01 4.02659595e-01 -2.84768581e-01 -6.76250935e-01
-6.15364254e-01 1.87080845e-01 -1.19111545e-01 3.65140587e-02
6.26951575e-01 1.91220313e-01 -2.10156679e-01 -5.85402906e-01
2.42426410e-01 5.29891431e-01 4.70121562e-01 -4.19156015e-01
9.58118618e-01 7.79153943e-01 5.79633750e-02 -9.22944427e-01
-7.38672674e-01 -8.51473629e-01 -9.12182152e-01 -3.15429419e-01
9.44463491e-01 -8.33347797e-01 -5.27385950e-01 3.17224890e-01
-1.21416128e+00 -1.41096577e-01 -4.09569681e-01 3.36746156e-01
-5.20652056e-01 2.57165015e-01 -2.02197179e-01 -7.96658695e-01
-1.15702316e-01 -1.24456143e+00 1.19633698e+00 7.82518312e-02
-9.76891816e-03 -4.14184958e-01 -3.13388288e-01 4.08067316e-01
-7.88116306e-02 4.61056083e-01 4.21699017e-01 -6.71125591e-01
-1.01231074e+00 -1.00195706e-01 -3.10784698e-01 3.85621309e-01
4.96742129e-02 9.84749123e-02 -8.94954085e-01 -7.25035742e-02
-1.61856011e-01 1.84133634e-01 8.12274098e-01 5.83141685e-01
1.58457255e+00 1.21431209e-01 -4.39677268e-01 5.64750731e-01
1.46206069e+00 2.06404567e-01 6.88577056e-01 2.44498238e-01
7.85234749e-01 5.89681685e-01 9.01451945e-01 6.37042463e-01
4.44809526e-01 7.63835788e-01 8.17390740e-01 -3.22068006e-01
-7.97730461e-02 2.99073458e-02 -2.02365085e-01 5.80209792e-01
-2.66685158e-01 -2.04540104e-01 -1.20695901e+00 6.30253017e-01
-1.71610761e+00 -8.32590759e-01 -7.36074746e-01 2.26529455e+00
4.03114378e-01 4.75490332e-01 4.13094521e-01 3.52999657e-01
9.24079955e-01 -6.93357661e-02 -5.93766391e-01 1.96072072e-01
-1.97758842e-02 3.11508507e-01 6.50260091e-01 3.95110905e-01
-1.23444557e+00 1.03622603e+00 4.71553087e+00 1.09371388e+00
-7.26882815e-01 1.10333785e-02 7.40570128e-01 -1.02421254e-01
-1.14511535e-01 -1.57387719e-01 -8.89001191e-01 6.34884894e-01
4.31103706e-01 2.14664280e-01 2.21724361e-02 7.71870852e-01
3.27837020e-01 -1.55381158e-01 -9.48663592e-01 1.09962726e+00
-2.97469974e-01 -1.35062325e+00 9.18971375e-02 1.92975819e-01
7.83823490e-01 3.88183594e-01 -1.33588716e-01 -2.91534495e-02
7.21125677e-02 -8.28060627e-01 9.01430428e-01 3.94719362e-01
5.28515220e-01 -8.76722455e-01 6.37675464e-01 5.22425532e-01
-1.33964837e+00 8.61058384e-02 -4.97674257e-01 2.89529353e-01
3.61483097e-01 1.12844253e+00 -8.32056820e-01 8.43205869e-01
9.75156009e-01 6.87405765e-01 -5.04455984e-01 1.60900462e+00
-2.78326981e-02 6.95434809e-01 -8.27120841e-01 1.85571939e-01
3.99809003e-01 -3.82270575e-01 7.21055448e-01 1.19276738e+00
5.24798691e-01 2.98725188e-01 4.57644671e-01 9.42515433e-01
2.33045891e-01 1.26416028e-01 -2.75905490e-01 4.72637802e-01
7.26982474e-01 1.22723281e+00 -1.16278958e+00 -4.34728622e-01
-2.83495128e-01 7.47245669e-01 3.66181205e-03 9.79846716e-02
-9.14852500e-01 -9.59751755e-02 6.78267419e-01 4.19603199e-01
6.38138533e-01 -3.16431195e-01 -7.40907431e-01 -7.08196461e-01
2.25225642e-01 -4.28442746e-01 3.13659996e-01 -6.62457705e-01
-1.19639277e+00 7.68545210e-01 1.39566317e-01 -1.74219024e+00
2.72623986e-01 -3.24756563e-01 -4.68965560e-01 5.66450834e-01
-1.48392642e+00 -8.57190549e-01 -7.27231383e-01 4.76617068e-01
7.81315565e-01 2.66333282e-01 2.13439256e-01 3.99068564e-01
-4.63795483e-01 1.94290757e-01 -2.22019136e-01 -2.23671362e-01
1.36359498e-01 -1.21755421e+00 6.43740714e-01 8.06068540e-01
1.51582032e-01 3.29057842e-01 5.33765495e-01 -7.33031809e-01
-9.19611752e-01 -1.30485773e+00 5.47264338e-01 -4.66985643e-01
3.39963496e-01 -4.21251774e-01 -1.14560986e+00 2.20376611e-01
-2.48132795e-01 8.02107677e-02 4.30932224e-01 -2.65441209e-01
4.13227752e-02 -1.02328613e-01 -1.24986386e+00 4.63245720e-01
1.38604522e+00 7.27200061e-02 -3.03448915e-01 3.98572236e-01
7.85747945e-01 -6.92485869e-01 -9.56547260e-01 7.79030204e-01
-2.09917352e-02 -1.27987969e+00 1.24304497e+00 8.77913013e-02
3.84245306e-01 -8.28477323e-01 -2.47848183e-01 -9.49812293e-01
-4.87086058e-01 -3.30247462e-01 7.66555313e-03 1.17842042e+00
5.00283778e-01 -4.33225453e-01 9.87477720e-01 3.81303072e-01
-5.75385332e-01 -8.03257763e-01 -1.19537830e+00 -1.08921838e+00
-2.16738626e-01 -8.52403522e-01 9.87685025e-01 7.48350561e-01
-5.39672673e-01 -3.07191573e-02 2.61055917e-01 6.53112829e-01
9.99181151e-01 2.29778573e-01 9.49147165e-01 -1.50427592e+00
1.32206276e-01 -6.67784572e-01 -5.83149076e-01 -1.18050659e+00
-2.20695093e-01 -9.87730980e-01 3.72520834e-03 -1.85152245e+00
-1.77574888e-01 -8.77457857e-01 -2.09593967e-01 4.00775373e-01
-1.11493364e-01 5.45232296e-01 3.21217746e-01 5.51911950e-01
-8.03932309e-01 4.90213931e-01 1.20231128e+00 7.69385695e-02
-5.32246947e-01 4.56389308e-01 -4.24947411e-01 7.18899190e-01
1.07306445e+00 -3.97299409e-01 -2.61914432e-01 -2.13410780e-01
-1.20827235e-01 -2.41488397e-01 4.85901207e-01 -1.46735835e+00
8.49022642e-02 -1.48682341e-01 2.55252540e-01 -1.46041965e+00
5.07219076e-01 -1.05594873e+00 4.59169537e-01 3.52302879e-01
6.00073300e-02 -1.53589576e-01 2.78545409e-01 4.07815695e-01
-1.88056961e-01 -4.72869240e-02 8.92047942e-01 -1.38660863e-01
-8.79419029e-01 6.12036824e-01 1.11229658e-01 -2.24741161e-01
1.27811801e+00 -7.60798037e-01 -6.03846014e-02 1.46273389e-01
-8.92420888e-01 4.00097787e-01 6.02948308e-01 2.46423170e-01
7.01247692e-01 -1.22457349e+00 -8.01706791e-01 1.87246725e-01
2.98213303e-01 8.02994490e-01 5.27439415e-01 8.80767226e-01
-5.20198405e-01 1.87582657e-01 2.13018045e-01 -1.34868181e+00
-1.19150043e+00 3.19058418e-01 9.06153172e-02 1.82332397e-01
-8.85767281e-01 8.63571167e-01 3.40962708e-01 -4.07260090e-01
1.06323235e-01 -5.07935166e-01 -9.99875888e-02 -8.80913362e-02
2.02157080e-01 5.06456375e-01 4.21233058e-01 -6.30719781e-01
-5.63076675e-01 7.17348039e-01 1.66832402e-01 6.39068708e-02
1.38418829e+00 -6.33190274e-02 -7.06115589e-02 3.63803208e-01
7.72319794e-01 -1.90785825e-01 -1.36722255e+00 -2.99271852e-01
8.13773647e-02 -6.12650394e-01 3.47287487e-03 -6.78478241e-01
-1.11578321e+00 6.92047536e-01 6.22225702e-01 5.19243777e-01
1.09077072e+00 4.58622128e-01 6.87927723e-01 1.79767758e-01
4.47797060e-01 -9.87084806e-01 -2.30851293e-01 3.67322832e-01
9.44331408e-01 -1.17682922e+00 9.50580239e-02 -8.82054508e-01
-6.05844080e-01 7.30310738e-01 7.25316942e-01 -2.13767335e-01
6.74677670e-01 9.50350687e-02 -9.89237279e-02 -4.21093196e-01
-4.03868169e-01 -5.47838390e-01 5.08747339e-01 5.61829627e-01
9.29499883e-03 1.95341840e-01 -1.77272961e-01 3.02323848e-01
-4.27660406e-01 -1.82477877e-01 1.84972271e-01 7.05838144e-01
-7.28804827e-01 -7.64936686e-01 -5.89803636e-01 6.56865418e-01
-9.67155844e-02 7.32730851e-02 -2.42952779e-01 7.61411846e-01
3.15055043e-01 9.15198565e-01 4.13778722e-01 -3.90864283e-01
6.91348612e-01 -2.24886522e-01 1.76496983e-01 -7.16730654e-01
-4.75276291e-01 2.59737879e-01 -1.51966080e-01 -8.58970165e-01
-6.50913417e-01 -8.22573185e-01 -1.50114000e+00 -1.81717947e-01
-4.72908378e-01 2.96748150e-03 7.38871336e-01 6.58147097e-01
6.50234461e-01 4.22019303e-01 7.37169027e-01 -1.29520833e+00
-4.19176668e-02 -7.48479307e-01 -4.50462013e-01 4.43839133e-01
3.30987126e-02 -7.49513447e-01 -4.38090205e-01 -1.22477777e-01]
|
[8.051158905029297, -3.047004461288452]
|
efd71191-cac2-4309-b6d5-30129666b655
|
dynamite-dynamic-query-bootstrapping-for
|
2304.06668
| null |
https://arxiv.org/abs/2304.06668v1
|
https://arxiv.org/pdf/2304.06668v1.pdf
|
DynaMITe: Dynamic Query Bootstrapping for Multi-object Interactive Segmentation Transformer
|
Most state-of-the-art instance segmentation methods rely on large amounts of pixel-precise ground-truth annotations for training, which are expensive to create. Interactive segmentation networks help generate such annotations based on an image and the corresponding user interactions such as clicks. Existing methods for this task can only process a single instance at a time and each user interaction requires a full forward pass through the entire deep network. We introduce a more efficient approach, called DynaMITe, in which we represent user interactions as spatio-temporal queries to a Transformer decoder with a potential to segment multiple object instances in a single iteration. Our architecture also alleviates any need to re-compute image features during refinement, and requires fewer interactions for segmenting multiple instances in a single image when compared to other methods. DynaMITe achieves state-of-the-art results on multiple existing interactive segmentation benchmarks, and also on the new multi-instance benchmark that we propose in this paper.
|
['Bastian Leibe', 'Alexander Hermans', 'Sabarinath Mahadevan', 'Amit Kumar Rana']
|
2023-04-13
| null | null | null | null |
['interactive-segmentation']
|
['computer-vision']
|
[ 5.18544197e-01 3.49066943e-01 -1.59812361e-01 -6.16335213e-01
-1.15251839e+00 -7.13882446e-01 4.18626070e-01 1.60167113e-01
-7.16926396e-01 5.80384791e-01 -3.45249921e-01 -2.35610127e-01
2.85089582e-01 -8.76238048e-01 -1.07231092e+00 -2.68540591e-01
1.68142006e-01 1.03411317e+00 1.08357799e+00 1.57254651e-01
1.06724419e-01 2.78350711e-01 -1.41598618e+00 5.40942252e-01
8.64115894e-01 1.02885187e+00 4.77448076e-01 7.68554747e-01
-3.41246277e-01 3.88584048e-01 -5.33622682e-01 -4.13511276e-01
2.99548835e-01 -3.10744673e-01 -1.19746399e+00 3.07252705e-01
6.32587492e-01 -4.57370758e-01 4.81153280e-03 8.15544188e-01
2.58611083e-01 1.88271835e-01 3.48759800e-01 -1.16376793e+00
-2.43907735e-01 7.80137956e-01 -6.96968973e-01 4.43543680e-03
1.62120879e-01 3.46678108e-01 9.39741850e-01 -7.33877838e-01
9.06827927e-01 9.04319286e-01 5.95138490e-01 4.25743848e-01
-1.48565853e+00 -4.81832415e-01 4.06975180e-01 2.24868767e-03
-1.22163713e+00 -1.17043488e-01 3.68461758e-01 -4.18426961e-01
7.52409160e-01 2.79415816e-01 1.00681078e+00 5.41853011e-01
-3.36665154e-01 1.15585482e+00 7.79644251e-01 -1.51960179e-01
1.50750682e-01 -1.93072006e-01 2.32100207e-02 9.18686628e-01
-2.01035216e-01 -3.02013516e-01 -3.57158333e-01 1.94988087e-01
1.17982686e+00 -9.36643407e-02 -6.17817678e-02 -4.38101858e-01
-1.38042676e+00 6.55078173e-01 6.08781636e-01 1.17299236e-01
-4.14477617e-01 4.36447710e-01 2.97943741e-01 -8.19182321e-02
5.49169481e-01 3.95408630e-01 -6.59717321e-01 -2.07465574e-01
-1.57540560e+00 3.14062357e-01 7.26362944e-01 9.53173339e-01
1.07394969e+00 -4.97579902e-01 -4.96226996e-01 5.53619862e-01
1.05469175e-01 9.05404389e-02 3.51675190e-02 -1.16941416e+00
4.81459051e-01 7.05799878e-01 2.15698168e-01 -4.65418488e-01
-3.80766273e-01 -4.19193596e-01 -4.61624056e-01 2.37886906e-01
6.46469355e-01 -1.79437160e-01 -1.39640522e+00 1.40978706e+00
5.93638718e-01 4.74902630e-01 -3.76168340e-01 8.72491062e-01
6.95094109e-01 7.31169760e-01 9.12352204e-02 1.74611285e-01
1.19462931e+00 -1.49974692e+00 -3.52314919e-01 -4.67300653e-01
5.39020240e-01 -6.70862019e-01 1.35739315e+00 4.10601825e-01
-1.39813614e+00 -4.80170041e-01 -7.06782162e-01 -3.29506069e-01
-3.70062470e-01 1.75896570e-01 7.34742463e-01 2.56551772e-01
-1.10451305e+00 7.11733401e-01 -1.16492546e+00 -8.22829679e-02
7.34065115e-01 5.61380863e-01 -1.27279714e-01 1.40618488e-01
-7.47531712e-01 5.33738375e-01 5.27377427e-01 1.51778787e-01
-1.01684368e+00 -1.01889086e+00 -7.56723881e-01 4.36139904e-04
7.61409700e-01 -6.39084995e-01 1.68984950e+00 -1.20904815e+00
-1.38637292e+00 9.58602607e-01 -3.30405027e-01 -5.39075553e-01
8.72532368e-01 -5.12288034e-01 2.15303853e-01 2.11530089e-01
2.79389620e-01 1.32786095e+00 5.52852809e-01 -1.37229955e+00
-9.96312559e-01 -2.39479706e-01 4.14126277e-01 2.50299454e-01
2.00274363e-01 -1.17691100e-01 -1.46204531e+00 -4.00095612e-01
2.02028468e-01 -9.99428749e-01 -6.20178699e-01 1.95686355e-01
-9.11885679e-01 -2.71325082e-01 1.03142846e+00 -4.66431171e-01
9.68973339e-01 -1.81336653e+00 2.11502627e-01 2.59867936e-01
1.83543995e-01 2.94665784e-01 -1.09179549e-01 -2.51978114e-02
2.45569527e-01 3.41034293e-01 -5.90754986e-01 -7.11058617e-01
-1.45326043e-02 1.96913958e-01 -7.16364905e-02 1.55535683e-01
-1.33960247e-02 1.05574560e+00 -9.85062599e-01 -6.99435711e-01
3.18717927e-01 4.80161399e-01 -6.03834569e-01 1.79547802e-01
-9.29783523e-01 6.63606465e-01 -3.63911390e-01 3.41220111e-01
5.14942348e-01 -6.41047776e-01 -9.28136334e-02 -2.37232611e-01
-3.13069135e-01 3.31270397e-01 -1.19616663e+00 2.24828076e+00
-5.23966491e-01 6.69088364e-01 -1.49135729e-02 -8.58338773e-01
2.60296673e-01 2.13598669e-01 4.79362160e-01 -6.24605358e-01
-1.24503881e-01 1.56702042e-01 -3.62399220e-01 -1.81798771e-01
4.33510065e-01 4.84033018e-01 5.33011407e-02 6.79405153e-01
-6.81816712e-02 -1.26648739e-01 7.22075224e-01 3.87132645e-01
9.64436054e-01 5.30143738e-01 -1.58612713e-01 -4.61866846e-03
1.80500969e-01 2.53096849e-01 5.45711935e-01 9.60673511e-01
2.55275398e-01 8.67723942e-01 6.19628131e-01 -5.03687561e-01
-1.07704127e+00 -9.90529478e-01 1.50129972e-02 1.22180188e+00
5.38168848e-01 -4.28744942e-01 -1.16890311e+00 -9.34902370e-01
-4.15889412e-01 6.26885414e-01 -5.52182913e-01 5.55672944e-01
-6.42721236e-01 -4.07077461e-01 4.15577143e-01 6.97454512e-01
7.36528814e-01 -1.28257155e+00 -9.44751084e-01 3.24290186e-01
-3.09025466e-01 -1.25402403e+00 -6.18021250e-01 1.25362292e-01
-9.70667839e-01 -1.13160670e+00 -7.86308527e-01 -7.27092505e-01
9.99599159e-01 -9.68413949e-02 1.39281213e+00 2.68853337e-01
-4.58402961e-01 1.36823254e-02 -3.13651077e-02 -1.89754203e-01
-6.82977587e-02 5.40034473e-01 -8.23059261e-01 -4.55635153e-02
-3.01570315e-02 -3.35668653e-01 -8.86496544e-01 4.84474421e-01
-1.00742531e+00 7.71584928e-01 5.06416142e-01 4.91894007e-01
1.17714262e+00 -2.41708115e-01 1.98511168e-01 -1.45904315e+00
2.04371586e-01 -1.71322033e-01 -9.65461910e-01 2.78966188e-01
-2.19286680e-01 5.27744554e-02 2.86491841e-01 -2.10432902e-01
-1.04956675e+00 5.84443867e-01 -1.53245136e-01 -2.19644591e-01
-4.37162668e-01 4.32941526e-01 1.07686996e-01 9.12266374e-02
5.62530339e-01 5.82464337e-02 -3.61698002e-01 -5.01918733e-01
6.00550830e-01 1.23061270e-01 6.78094447e-01 -4.91392225e-01
6.76816165e-01 4.31254596e-01 -3.29755396e-01 -5.02392054e-01
-1.16867435e+00 -5.17185748e-01 -9.38888609e-01 -1.98867917e-01
1.24978685e+00 -6.51754558e-01 -7.03354537e-01 6.14837706e-01
-1.40860593e+00 -1.08129346e+00 -2.88548678e-01 2.36021671e-02
-5.66886425e-01 -1.57202575e-02 -5.86638749e-01 -2.99435467e-01
-1.80200979e-01 -1.47750032e+00 1.40252173e+00 3.43610257e-01
-1.37445346e-01 -7.99658775e-01 -2.65275925e-01 5.39193571e-01
2.84408450e-01 3.09958339e-01 5.36956728e-01 -4.30504084e-01
-1.27630484e+00 -1.66411340e-01 -4.58759636e-01 -5.78127764e-02
-5.13567626e-02 7.18511790e-02 -8.24004650e-01 7.23519921e-02
-6.48662567e-01 -2.83730865e-01 8.32974255e-01 5.09399593e-01
1.74686527e+00 -1.85671553e-01 -6.51201725e-01 6.81232154e-01
1.30093014e+00 1.94423810e-01 6.60033464e-01 1.57132372e-01
9.56157386e-01 2.47860268e-01 6.54513419e-01 1.77485988e-01
4.78779644e-01 7.86005497e-01 5.32559216e-01 -7.49429464e-01
-8.51228237e-02 -1.59772754e-01 -2.68092066e-01 9.52529088e-02
-1.73461080e-01 -4.07993585e-01 -9.72757101e-01 7.82699108e-01
-2.05173469e+00 -6.60760522e-01 -3.53039384e-01 2.24085164e+00
1.07117271e+00 2.89740652e-01 1.59793466e-01 -2.93019656e-02
6.84714615e-01 9.75990966e-02 -6.91150486e-01 1.01229988e-01
3.72195244e-01 5.06125689e-01 6.51942074e-01 8.07151794e-01
-1.25112128e+00 1.40570498e+00 6.18664026e+00 7.11086929e-01
-1.12678301e+00 1.40471593e-01 1.02869284e+00 -2.69725502e-01
-2.72419363e-01 8.95263627e-02 -7.18205452e-01 3.06028515e-01
5.91204882e-01 2.95890689e-01 3.82938832e-01 8.81176949e-01
3.50789070e-01 -7.03267157e-01 -1.32189608e+00 7.60736167e-01
-2.99989074e-01 -1.67141974e+00 -1.80069491e-01 -7.43214265e-02
8.77738535e-01 2.48631403e-01 -1.83254182e-01 -1.02103941e-01
4.68505293e-01 -1.00873721e+00 7.16799259e-01 4.22183752e-01
8.18248272e-01 -6.66836083e-01 3.69635731e-01 4.24404532e-01
-1.18788540e+00 4.47966963e-01 1.46949902e-01 2.49128342e-01
6.27110124e-01 6.72419786e-01 -8.47485960e-01 1.60388008e-01
7.64849961e-01 4.02111918e-01 -5.81403077e-01 1.32012033e+00
-4.58410740e-01 6.79534435e-01 -7.50402868e-01 2.17498541e-01
5.79266250e-01 -9.75259319e-02 2.07396492e-01 1.19153965e+00
1.24506891e-01 -1.56134432e-02 5.56899607e-01 1.05584848e+00
-1.64417267e-01 -1.71886563e-01 -8.67179632e-02 1.00055575e-01
4.42186326e-01 1.21655524e+00 -1.42833197e+00 -7.38004684e-01
-3.64922047e-01 1.41178226e+00 2.74391174e-01 4.60306317e-01
-1.15601599e+00 -4.44177389e-01 2.79774904e-01 4.52675194e-01
5.81732690e-01 -2.81281322e-01 -4.51602370e-01 -8.01044643e-01
7.23152533e-02 -5.37049055e-01 1.75580814e-01 -7.57017076e-01
-5.85696757e-01 5.18294811e-01 1.05478168e-02 -6.52499020e-01
-2.11297974e-01 -1.48348272e-01 -5.82306623e-01 7.67385483e-01
-1.40337503e+00 -1.15002584e+00 -5.67147493e-01 6.57990992e-01
8.86007130e-01 6.33280098e-01 5.73631883e-01 4.60748464e-01
-5.47159910e-01 3.50492567e-01 -5.11220515e-01 2.81918734e-01
4.27151471e-01 -1.51241970e+00 9.07713234e-01 7.73044169e-01
4.14033145e-01 2.87861586e-01 5.36705911e-01 -6.46084070e-01
-8.48399341e-01 -1.05124986e+00 7.97842801e-01 -2.94669300e-01
2.03009203e-01 -4.20664072e-01 -7.71486342e-01 9.99264419e-01
1.75259069e-01 3.00256640e-01 2.69331604e-01 1.38099909e-01
9.89414975e-02 1.04695044e-01 -9.24247086e-01 6.28205478e-01
1.16587102e+00 -3.89679760e-01 -1.14634715e-01 5.35613894e-01
7.65591860e-01 -1.02323210e+00 -5.59736431e-01 1.83649719e-01
3.48181963e-01 -1.07724476e+00 9.83049572e-01 -3.07664365e-01
4.73763674e-01 -5.45654714e-01 4.45555478e-01 -1.03070712e+00
-1.28200604e-02 -7.63343573e-01 2.48964150e-02 9.62995052e-01
9.18782294e-01 -1.81081489e-01 1.23535407e+00 9.11419690e-01
-2.18177080e-01 -9.52583015e-01 -4.91230696e-01 -3.29184562e-01
-5.33469141e-01 -6.43261552e-01 6.01721227e-01 6.34211183e-01
-4.67562228e-01 1.98554754e-01 -3.34852226e-02 3.12026054e-01
6.81954086e-01 2.09902599e-01 9.44711387e-01 -1.26393640e+00
-3.53505760e-01 -2.98623651e-01 -6.02025688e-02 -1.49555719e+00
-5.72766960e-02 -6.95835888e-01 4.08495992e-01 -2.00727391e+00
1.14858747e-01 -6.80941045e-01 1.46242708e-01 7.60721385e-01
-2.62597293e-01 6.31178319e-01 2.11963162e-01 -2.23340541e-02
-1.02777755e+00 9.21456609e-03 1.45898378e+00 -5.71635179e-02
-4.50491816e-01 1.92534044e-01 -3.21549624e-01 9.41607654e-01
5.75750053e-01 -5.14069974e-01 -6.76788509e-01 -7.65051603e-01
1.65452182e-01 3.14714015e-01 5.34081280e-01 -1.10764718e+00
3.71797800e-01 -2.17332423e-01 3.99497300e-01 -7.88317382e-01
4.22414273e-01 -7.13392019e-01 2.50375599e-01 2.03247562e-01
-4.16779906e-01 -1.72794297e-01 2.42917851e-01 4.26108509e-01
-2.37431481e-01 -2.83880800e-01 8.39737356e-01 -4.39522177e-01
-8.68811786e-01 6.48890615e-01 -1.54014349e-01 3.27788442e-01
1.13043249e+00 -4.12505388e-01 7.33302021e-03 -1.78040132e-01
-1.10188818e+00 3.61372948e-01 4.36188191e-01 2.46418312e-01
2.63934523e-01 -8.68626297e-01 -3.12429667e-01 -4.36237045e-02
-3.50386471e-01 8.83196414e-01 2.20279932e-01 6.95422649e-01
-8.69295776e-01 1.86807409e-01 1.48543820e-01 -8.87564003e-01
-1.30120599e+00 5.97446673e-02 4.36543435e-01 -4.70464349e-01
-8.74943316e-01 1.20847368e+00 5.24453461e-01 -3.15913528e-01
3.32684964e-01 -3.92873615e-01 2.63449043e-01 2.81831995e-02
2.62736589e-01 2.08875895e-01 3.08865756e-02 -2.59382695e-01
-1.16043799e-01 4.84857708e-01 -2.61009723e-01 -5.36773741e-01
1.12969995e+00 1.10119879e-01 -1.29304575e-02 3.62012893e-01
9.23079491e-01 -4.15687561e-01 -1.85800922e+00 -1.34722218e-01
4.30023707e-02 -4.39413190e-01 1.09571554e-01 -1.07000780e+00
-1.45976305e+00 8.80101681e-01 3.11650693e-01 2.59874314e-01
9.20137405e-01 1.47122622e-01 1.07806385e+00 2.94621050e-01
2.56054848e-01 -1.26278687e+00 1.40875801e-01 4.61283922e-01
5.79955876e-01 -1.18781507e+00 -1.68376699e-01 -8.04389656e-01
-5.49139559e-01 8.35381925e-01 8.73130977e-01 1.91889424e-02
4.82617140e-01 3.73110384e-01 9.43432525e-02 -3.24788272e-01
-3.98499638e-01 -1.84758142e-01 2.79312611e-01 3.32439393e-01
5.55113971e-01 -4.66810279e-02 -1.83822557e-01 1.65073588e-01
-5.76798953e-02 1.38779417e-01 1.74483672e-01 9.21545506e-01
-3.93034369e-01 -1.26071846e+00 4.65948693e-02 5.48442721e-01
-5.49377978e-01 -1.88646391e-01 -2.95323253e-01 7.65295506e-01
8.22224915e-02 5.58787584e-01 4.16231185e-01 1.62213564e-01
2.33961523e-01 2.16641668e-02 2.92968750e-01 -9.04798329e-01
-6.86909914e-01 1.89159155e-01 9.78911147e-02 -1.13479233e+00
-3.66492063e-01 -4.94825363e-01 -1.73013449e+00 5.42679615e-02
-2.53115326e-01 -4.13640030e-02 7.21100509e-01 1.05156803e+00
4.93229896e-01 7.29196370e-01 1.72796641e-02 -1.35674191e+00
2.12187096e-01 -6.79197013e-01 -1.68871984e-01 3.58852178e-01
1.43571183e-01 -2.78333217e-01 1.20477654e-01 4.52505618e-01]
|
[9.428855895996094, 0.09840928018093109]
|
01a6263f-8465-45a2-8752-a7c0ecbb66cd
|
lvp-m3-language-aware-visual-prompt-for
|
2210.15461
| null |
https://arxiv.org/abs/2210.15461v2
|
https://arxiv.org/pdf/2210.15461v2.pdf
|
LVP-M3: Language-aware Visual Prompt for Multilingual Multimodal Machine Translation
|
Multimodal Machine Translation (MMT) focuses on enhancing text-only translation with visual features, which has attracted considerable attention from both natural language processing and computer vision communities. Recent advances still struggle to train a separate model for each language pair, which is costly and unaffordable when the number of languages increases in the real world. In other words, the multilingual multimodal machine translation (Multilingual MMT) task has not been investigated, which aims to handle the aforementioned issues by providing a shared semantic space for multiple languages. Besides, the image modality has no language boundaries, which is superior to bridging the semantic gap between languages. To this end, we first propose the Multilingual MMT task by establishing two new Multilingual MMT benchmark datasets covering seven languages. Then, an effective baseline LVP-M3 using visual prompts is proposed to support translations between different languages, which includes three stages (token encoding, language-aware visual prompt generation, and language translation). Extensive experimental results on our constructed benchmark datasets demonstrate the effectiveness of LVP-M3 method for Multilingual MMT.
|
['Zheng Cui', 'Furu Wei', 'Dongdong Zhang', 'Zhoujun Li', 'Jian Yang', 'Haoyang Huang', 'Jiaheng Liu', 'Hongcheng Guo']
|
2022-10-19
| null | null | null | null |
['multimodal-machine-translation']
|
['natural-language-processing']
|
[ 9.39583108e-02 -4.13758159e-01 -2.31445625e-01 -1.66030422e-01
-1.06565166e+00 -5.55244565e-01 8.11986685e-01 -6.37470633e-02
-3.97527337e-01 6.97168589e-01 1.26434401e-01 -5.70510745e-01
4.97249871e-01 -4.31108087e-01 -6.88061714e-01 -5.63341320e-01
6.95591748e-01 4.02772278e-01 -1.34291381e-01 -2.48671070e-01
1.18358508e-02 -1.23052411e-01 -8.50068569e-01 4.52284187e-01
1.26465762e+00 8.38378489e-01 5.34202576e-01 1.22771993e-01
-6.57735944e-01 3.00636768e-01 -4.05555516e-01 -7.23001778e-01
5.38937226e-02 -7.09031463e-01 -6.52478576e-01 2.40317419e-01
5.53666055e-01 -8.31172839e-02 -1.77349523e-01 1.06709254e+00
6.44754767e-01 -2.78358400e-01 4.51931179e-01 -1.56547618e+00
-1.18388045e+00 6.23651206e-01 -9.31798458e-01 -3.15507621e-01
4.52029645e-01 3.10561776e-01 8.79164398e-01 -1.32929635e+00
8.26911032e-01 1.60177219e+00 2.23531097e-01 4.32368398e-01
-1.06763577e+00 -7.11693823e-01 2.78467149e-01 2.44589910e-01
-1.35195839e+00 -2.80010968e-01 6.98267698e-01 -3.89873594e-01
6.51635587e-01 1.03329018e-01 5.07967770e-01 1.27587283e+00
3.73058431e-02 1.08566380e+00 1.39975250e+00 -4.97459263e-01
-2.18887284e-01 5.40201485e-01 -4.78069216e-01 5.57793498e-01
-5.12346886e-02 -2.25482538e-01 -5.61921775e-01 2.83554584e-01
6.77023530e-01 -1.84327483e-01 -4.70628768e-01 -4.52413678e-01
-1.94632804e+00 6.55178189e-01 2.83876449e-01 4.03977066e-01
-1.14161164e-01 -1.98134959e-01 7.69356370e-01 4.43794250e-01
3.54064316e-01 1.54711515e-01 -1.68947101e-01 7.99747780e-02
-7.54368246e-01 -1.29282221e-01 2.79200554e-01 1.26020384e+00
8.41133535e-01 2.89269444e-02 -3.67592126e-01 9.60959971e-01
4.52831507e-01 1.00568497e+00 4.73833919e-01 -5.08397579e-01
1.19975853e+00 7.27784038e-01 -3.40972398e-03 -1.09321499e+00
7.18599409e-02 -3.26290965e-01 -1.06402266e+00 -2.90276200e-01
1.25244781e-01 -5.48749454e-02 -6.11250341e-01 1.81115782e+00
2.21482143e-01 -2.16793925e-01 3.74481887e-01 1.08723462e+00
1.03161550e+00 1.02261460e+00 1.46403864e-01 -3.92268956e-01
1.33507192e+00 -1.27754056e+00 -8.67815197e-01 -4.12632495e-01
6.24748230e-01 -1.45445538e+00 1.45740390e+00 -1.91151038e-01
-9.33500588e-01 -5.95661044e-01 -8.21877003e-01 -3.45720023e-01
-2.78061628e-01 4.14326757e-01 4.90192235e-01 2.01389641e-01
-1.06438637e+00 -4.83490705e-01 -3.80903661e-01 -5.86868823e-01
2.98735946e-01 3.47526148e-02 -6.31928563e-01 -5.66038787e-01
-1.42987955e+00 1.00168002e+00 3.45472127e-01 3.87885004e-01
-7.29388058e-01 -2.35303253e-01 -1.07019210e+00 -2.42153510e-01
3.32223624e-01 -7.95042694e-01 9.90723729e-01 -1.40051484e+00
-1.29484773e+00 1.00739181e+00 -4.32657659e-01 9.41819549e-02
8.45233321e-01 1.08023219e-01 -4.37598974e-01 7.12942332e-02
4.47206080e-01 1.03228736e+00 7.76445210e-01 -1.46540654e+00
-6.30045831e-01 -2.51875728e-01 -1.08104482e-01 5.88475943e-01
-5.21428943e-01 1.75489649e-01 -9.06605721e-01 -7.23115683e-01
-7.03956559e-02 -8.86215329e-01 4.18441184e-02 -1.16648283e-02
-4.29270655e-01 -1.44044325e-01 8.31382751e-01 -9.31324840e-01
9.12263155e-01 -2.18145251e+00 5.14464140e-01 -1.79163232e-01
-2.18555313e-02 6.35440350e-02 -5.74948967e-01 5.39553821e-01
2.02407911e-01 4.74194624e-02 -2.51760781e-01 -4.85583961e-01
1.52254805e-01 1.40407786e-01 -2.51763552e-01 1.10081822e-01
1.65102810e-01 1.26690924e+00 -7.74942636e-01 -9.81676280e-01
8.91197100e-02 4.94309515e-01 -1.09997116e-01 2.49537110e-01
-1.31679937e-01 7.37951398e-01 -4.32816029e-01 8.54749084e-01
9.18492913e-01 -2.12308392e-01 2.15196699e-01 -3.88244361e-01
-2.75959402e-01 -2.03343123e-01 -7.11042404e-01 2.23292971e+00
-6.33639932e-01 5.46850264e-01 -3.24161574e-02 -7.76129067e-01
9.37710345e-01 3.92701179e-01 1.90922022e-01 -1.16388118e+00
9.72538590e-02 5.64460218e-01 -2.65942484e-01 -6.34000957e-01
4.65800315e-01 -2.17098892e-01 -2.92313933e-01 3.55898261e-01
6.67634830e-02 -3.95067334e-02 1.89558610e-01 2.76595920e-01
3.27573061e-01 3.09489101e-01 -4.12074067e-02 -7.30952322e-02
6.55408382e-01 3.00547570e-01 5.92099726e-01 1.49891779e-01
-2.01814026e-01 4.62685525e-01 2.38414064e-01 -1.69500500e-01
-1.00179720e+00 -1.01351249e+00 1.99946031e-01 8.37663412e-01
5.79837501e-01 -1.65451765e-01 -7.29689717e-01 -6.81616008e-01
-2.72018135e-01 4.17841285e-01 -2.70181745e-01 -1.14081293e-01
-6.19280696e-01 -5.12522817e-01 3.35246444e-01 3.35632265e-01
8.74025524e-01 -9.84734356e-01 -1.47486269e-01 4.27281624e-03
-9.60383356e-01 -1.55882943e+00 -1.03213620e+00 -3.92418742e-01
-6.12064958e-01 -7.65661299e-01 -1.20325053e+00 -1.39955831e+00
7.66657710e-01 6.99799061e-01 9.67846990e-01 -1.87550172e-01
4.12737280e-02 3.95497888e-01 -3.27929735e-01 -7.98481405e-02
-4.41030085e-01 1.12400994e-01 -1.30430996e-01 2.37204686e-01
3.21022451e-01 -1.74794510e-01 -4.81224597e-01 4.05797601e-01
-8.68433654e-01 7.61475503e-01 9.46354210e-01 8.26422453e-01
5.85522890e-01 -5.96857131e-01 5.44592738e-01 -2.25188494e-01
6.98957980e-01 -2.97875553e-01 -4.51921105e-01 9.77819622e-01
-2.45696500e-01 -7.20463470e-02 4.70889449e-01 -6.82747185e-01
-1.19394934e+00 -2.12222859e-01 1.65969640e-01 -4.68259782e-01
1.48470819e-01 7.60715187e-01 -5.84915876e-01 4.45600152e-02
-5.17570600e-02 7.23044753e-01 -9.75170135e-02 -2.10710272e-01
6.19462729e-01 8.06910038e-01 5.96053541e-01 -6.79761291e-01
8.83096099e-01 1.66382164e-01 -2.55600482e-01 -5.25581062e-01
-3.50930154e-01 -1.42232016e-01 -6.29399598e-01 -2.64835268e-01
1.04608786e+00 -1.09665740e+00 -3.36836725e-01 5.29943168e-01
-1.52362525e+00 5.23347780e-02 1.29810691e-01 6.12872541e-01
-3.54612559e-01 4.56019223e-01 -5.14069617e-01 -4.48451608e-01
-4.46698934e-01 -1.56459403e+00 1.23053277e+00 1.65474698e-01
1.98884606e-01 -8.98016453e-01 2.42105052e-02 6.96456492e-01
3.56170058e-01 4.30789106e-02 1.08726954e+00 -1.76595286e-01
-7.99307883e-01 2.51078129e-01 -7.89875627e-01 2.66296178e-01
1.54566050e-01 -2.54319489e-01 -4.96452749e-01 -3.94442320e-01
-3.08480591e-01 -6.04392052e-01 7.34425843e-01 5.67327859e-03
4.54807103e-01 -1.16801858e-01 -1.60521045e-01 3.99317056e-01
1.27086937e+00 1.77027866e-01 3.98084223e-01 3.98817152e-01
9.67866659e-01 7.86637604e-01 8.32461357e-01 -1.93210319e-02
8.97241831e-01 7.49166250e-01 2.59609848e-01 -6.14185631e-01
-1.95241436e-01 -5.13827384e-01 5.25615692e-01 1.38627338e+00
1.97012588e-01 -2.34234631e-01 -9.59090650e-01 6.48575246e-01
-2.02498436e+00 -5.42551160e-01 2.18148921e-02 2.06435490e+00
9.43366289e-01 -3.56358647e-01 -2.61848241e-01 -3.87888730e-01
9.06507134e-01 4.42525409e-02 -4.90701616e-01 -2.26918235e-01
-5.88496327e-01 -4.59842652e-01 -7.42920395e-03 3.03313375e-01
-7.19385862e-01 1.17497706e+00 4.71635580e+00 1.01956880e+00
-1.41661954e+00 4.77400005e-01 5.69341600e-01 3.75060618e-01
-6.50812209e-01 1.21092297e-01 -5.15553713e-01 4.45064962e-01
2.96714842e-01 -2.17608139e-01 3.36791158e-01 3.35110992e-01
3.37992817e-01 1.47771135e-01 -9.62706149e-01 1.37016094e+00
3.16052616e-01 -1.06657851e+00 5.51424682e-01 1.11683412e-02
8.37611794e-01 -1.18985839e-01 2.80229092e-01 4.98537630e-01
-1.66279823e-01 -8.59656751e-01 9.49512303e-01 3.09098661e-01
1.15897381e+00 -7.27868378e-01 6.24284685e-01 3.06842357e-01
-1.38296211e+00 3.22491109e-01 -3.13282281e-01 4.43645239e-01
3.58537763e-01 3.29888672e-01 -5.58720827e-01 9.01571333e-01
5.17043650e-01 7.69932508e-01 -6.21502578e-01 7.01274991e-01
-1.73013255e-01 1.24928966e-01 6.72672391e-02 1.72004044e-01
4.65148300e-01 -3.39292020e-01 3.16171855e-01 1.19686997e+00
6.76272631e-01 -5.30534983e-01 6.38664424e-01 7.60672629e-01
-3.35191607e-01 7.18321085e-01 -6.37054980e-01 -2.57018059e-01
2.92621702e-01 1.24110031e+00 -4.69134301e-01 -2.85513341e-01
-8.65291715e-01 1.31028152e+00 2.30344787e-01 6.97211742e-01
-8.91165376e-01 -1.47895679e-01 2.58012027e-01 -2.44834214e-01
-1.51428431e-01 -2.53965765e-01 -2.18968689e-02 -1.71695209e+00
2.94015229e-01 -1.21800208e+00 8.54370371e-02 -9.55701411e-01
-1.27454185e+00 8.21165204e-01 -1.89299509e-01 -1.50358558e+00
-1.32298889e-02 -4.28561717e-01 -2.08548933e-01 1.06699514e+00
-1.71997404e+00 -1.93266332e+00 -2.75036961e-01 7.99089670e-01
7.48877108e-01 -1.79184884e-01 4.90353018e-01 5.92192292e-01
-6.03299499e-01 6.98040366e-01 -7.11533707e-03 1.37199983e-01
1.24347281e+00 -7.82933593e-01 1.72956765e-01 8.76476943e-01
9.49022621e-02 5.38566470e-01 4.12784606e-01 -6.31449342e-01
-1.72321188e+00 -1.25543106e+00 1.35316432e+00 -6.56445697e-02
6.43508136e-01 -4.54496980e-01 -9.15782511e-01 6.66018486e-01
9.77575481e-01 -1.77020475e-01 5.58027327e-01 -3.06750834e-01
-5.23647189e-01 -6.38649613e-02 -6.54037178e-01 8.67821813e-01
8.57677281e-01 -5.52995324e-01 -3.01334083e-01 4.34202731e-01
9.06701982e-01 -2.78300852e-01 -6.52487040e-01 3.40488672e-01
3.27457160e-01 -5.32175243e-01 8.06518614e-01 -3.60935450e-01
5.96765876e-01 -4.90838468e-01 -3.73743385e-01 -1.41995692e+00
2.46922627e-01 -4.64042068e-01 5.25333524e-01 1.52240920e+00
5.41201055e-01 -6.85914755e-01 5.44861592e-02 4.10776101e-02
-4.03000116e-02 -4.97731447e-01 -9.29926813e-01 -6.98801696e-01
2.37859190e-01 -1.87143624e-01 5.98561406e-01 1.33801425e+00
-2.16687843e-02 7.91733086e-01 -7.16497481e-01 -1.21465484e-02
4.72120315e-01 5.69441974e-01 7.88462162e-01 -6.89874470e-01
-7.80657679e-02 -5.11535823e-01 -4.32766303e-02 -1.18982780e+00
2.13676766e-01 -1.09509742e+00 3.90190594e-02 -1.78164601e+00
7.64367759e-01 -2.79926449e-01 -1.09388955e-01 5.24021924e-01
-2.80018359e-01 2.15650886e-01 3.13780159e-01 4.19139683e-01
-7.87871718e-01 9.80912745e-01 1.76593161e+00 -6.39253497e-01
1.19932339e-01 -6.76741779e-01 -6.38170421e-01 3.67122144e-01
6.32019997e-01 -9.91972536e-02 -4.55715865e-01 -1.06953824e+00
2.40333736e-01 3.97059590e-01 2.68086463e-01 -4.67417061e-01
1.52457163e-01 -4.23302740e-01 2.49721557e-01 -5.64968646e-01
8.84258598e-02 -8.18503141e-01 -4.59885225e-02 2.28000432e-01
-2.87360847e-01 5.80477536e-01 1.73499599e-01 3.71966571e-01
-6.77626848e-01 2.30630025e-01 6.10708475e-01 -6.25571832e-02
-7.29038358e-01 4.25443500e-01 -8.70234519e-02 1.42237380e-01
1.04358637e+00 4.58906405e-02 -5.61212957e-01 -2.58284390e-01
-3.33765060e-01 6.84860349e-01 5.80105126e-01 8.38603675e-01
7.88367450e-01 -1.84844851e+00 -1.04895663e+00 -5.53321168e-02
6.50576293e-01 -2.66630322e-01 4.07744497e-01 1.24073696e+00
-3.56829911e-01 4.99979645e-01 -2.83584952e-01 -8.13141704e-01
-1.33655357e+00 6.97107553e-01 1.08503476e-01 -2.62072116e-01
-2.26554185e-01 4.67337251e-01 5.53822696e-01 -7.23415196e-01
3.55595648e-02 1.74693409e-02 -9.11669284e-02 4.93059531e-02
2.51150757e-01 -1.38443202e-01 -1.40034258e-01 -1.09723878e+00
-3.61022681e-01 7.97941625e-01 -1.01972960e-01 -3.92808050e-01
7.30473518e-01 -6.60988212e-01 -4.24138814e-01 4.14847076e-01
1.25144804e+00 -6.70715347e-02 -7.75966108e-01 -5.69390655e-01
-1.55688226e-01 -4.39717382e-01 -2.04659566e-01 -8.77987146e-01
-1.03949594e+00 1.37711620e+00 5.37696898e-01 -5.09719789e-01
1.21897650e+00 -2.40876507e-02 1.18882787e+00 1.63545310e-01
4.33089584e-01 -9.54618156e-01 1.66716084e-01 3.88845503e-01
1.13193798e+00 -1.59011340e+00 -4.73200411e-01 -2.69142926e-01
-9.42889750e-01 9.18560863e-01 8.71297181e-01 7.27914095e-01
-4.93098050e-02 -1.99139550e-01 5.51081955e-01 2.35975519e-01
-5.54095626e-01 -7.91770518e-02 4.26446259e-01 4.50009376e-01
4.89543617e-01 1.40348986e-01 -5.76032817e-01 3.61176431e-01
2.04549536e-01 -2.54569918e-01 2.84859389e-02 8.93495381e-01
-2.21268699e-01 -1.47187185e+00 -4.62920427e-01 -2.65722632e-01
-3.27524766e-02 -3.46971780e-01 -5.15697479e-01 7.62509704e-01
2.55168557e-01 9.26729977e-01 -3.53367567e-01 -3.50179732e-01
2.25357607e-01 -7.84975141e-02 5.09793460e-01 -3.37922782e-01
-2.62957394e-01 4.76230532e-01 -2.90764362e-01 -2.88018435e-01
-6.00955248e-01 -5.82780719e-01 -9.71994340e-01 -9.90659446e-02
-1.53943570e-02 1.69686422e-01 7.28624344e-01 9.52668190e-01
3.12305063e-01 2.89200097e-01 5.53345740e-01 -5.37342727e-01
-1.26812801e-01 -8.81590724e-01 -5.40427938e-02 3.09531420e-01
4.55363803e-02 -3.35326612e-01 3.38442065e-02 1.82253450e-01]
|
[11.460238456726074, 1.5413928031921387]
|
93ea00fb-73d0-441c-8a9d-f416072cf9c4
|
pefll-a-lifelong-learning-approach-to
|
2306.05515
| null |
https://arxiv.org/abs/2306.05515v1
|
https://arxiv.org/pdf/2306.05515v1.pdf
|
PeFLL: A Lifelong Learning Approach to Personalized Federated Learning
|
Personalized federated learning (pFL) has emerged as a popular approach to dealing with the challenge of statistical heterogeneity between the data distributions of the participating clients. Instead of learning a single global model, pFL aims to learn an individual model for each client while still making use of the data available at other clients. In this work, we present PeFLL, a new pFL approach rooted in lifelong learning that performs well not only on clients present during its training phase, but also on any that may emerge in the future. PeFLL learns to output client specific models by jointly training an embedding network and a hypernetwork. The embedding network learns to represent clients in a latent descriptor space in a way that reflects their similarity to each other. The hypernetwork learns a mapping from this latent space to the space of possible client models. We demonstrate experimentally that PeFLL produces models of superior accuracy compared to previous methods, especially for clients not seen during training, and that it scales well to large numbers of clients. Moreover, generating a personalized model for a new client is efficient as no additional fine-tuning or optimization is required by either the client or the server. We also present theoretical results supporting PeFLL in the form of a new PAC-Bayesian generalization bound for lifelong learning and we prove the convergence of our proposed optimization procedure.
|
['Christoph H. Lampert', 'Hossein Zakerinia', 'Jonathan Scott']
|
2023-06-08
| null | null | null | null |
['personalized-federated-learning']
|
['methodology']
|
[-4.30263370e-01 -1.01169292e-02 -4.27944511e-01 -5.33776999e-01
-9.78053629e-01 -4.65614200e-01 5.46599150e-01 -9.06439647e-02
-1.58524379e-01 4.53626841e-01 1.36334136e-01 1.12665638e-01
-4.44660932e-01 -8.11627567e-01 -7.48976707e-01 -9.53564286e-01
-2.39480972e-01 1.28581595e+00 1.81480959e-01 3.74842703e-01
-1.32912993e-01 8.54390144e-01 -1.45599246e+00 4.95631367e-01
3.44237089e-01 9.77553010e-01 5.32512441e-02 5.94647646e-01
-2.09732860e-01 7.75856972e-01 -3.33604574e-01 -4.32483882e-01
3.14652354e-01 -1.23681314e-01 -6.68125570e-01 2.06005648e-01
2.20406622e-01 -5.15764773e-01 -6.40475392e-01 5.38749754e-01
4.84962881e-01 1.05444506e-01 6.24775112e-01 -1.55954933e+00
-3.26031208e-01 7.94611573e-01 -4.17137593e-02 -7.54928067e-02
4.58467342e-02 3.10992479e-01 1.16514397e+00 -9.63530242e-01
5.67120969e-01 1.23318672e+00 5.42475164e-01 4.92444277e-01
-1.49555182e+00 -4.82315928e-01 2.84819931e-01 5.22856772e-01
-1.28744924e+00 -4.13380325e-01 6.58693135e-01 -2.07650200e-01
7.23894238e-01 1.91516355e-01 2.34801382e-01 1.30628240e+00
-1.05725884e-01 9.88117814e-01 7.79071927e-01 -2.89520264e-01
6.43757463e-01 5.99687517e-01 9.08147544e-02 3.91271770e-01
5.53621799e-02 1.08599864e-01 -8.57903421e-01 -8.33786845e-01
3.89065504e-01 4.04885978e-01 -6.74286634e-02 -9.74811554e-01
-7.03920722e-01 8.98282647e-01 1.35456592e-01 1.52867436e-01
-7.72690892e-01 1.90609783e-01 1.71462744e-01 2.63575286e-01
3.02496761e-01 -9.81657729e-02 -5.83400786e-01 -1.19636834e-01
-7.89212227e-01 1.55092075e-01 1.38042367e+00 8.07644486e-01
1.12066472e+00 -2.45481431e-01 -1.57042801e-01 7.57913709e-01
4.27308440e-01 3.66089672e-01 4.36734438e-01 -1.20711172e+00
4.28409040e-01 5.54345071e-01 4.63623218e-02 -4.30883408e-01
1.93483457e-02 -5.81321180e-01 -3.60864252e-01 1.44516364e-01
2.90766716e-01 -2.21790433e-01 -1.85309887e-01 1.72800446e+00
5.45661211e-01 3.11540127e-01 1.21319734e-01 4.19243753e-01
-3.12726721e-02 7.26953685e-01 -2.61741281e-01 -2.09296644e-01
6.92868114e-01 -1.03114212e+00 -2.70547897e-01 8.02854747e-02
3.73805135e-01 -4.79910970e-01 7.47312903e-01 5.75919986e-01
-9.78710592e-01 4.98572662e-02 -3.74379635e-01 5.16502798e-01
-5.16103879e-02 -3.24613601e-01 4.35985923e-01 5.46878934e-01
-1.12812757e+00 6.78036690e-01 -7.28694618e-01 -5.27644455e-01
6.17936492e-01 5.09682178e-01 -3.68049175e-01 -2.92876959e-01
-6.57474399e-01 5.89755177e-01 1.86174199e-01 -3.54736775e-01
-1.23896801e+00 -6.81982279e-01 -2.50747621e-01 6.29474759e-01
6.02849066e-01 -4.95296985e-01 1.49154150e+00 -8.36153448e-01
-1.59014034e+00 3.78124386e-01 -2.33837023e-01 -3.80640835e-01
7.08821118e-01 7.29047582e-02 -7.07865879e-02 -1.01520486e-01
-2.73562282e-01 2.50801388e-02 8.41171980e-01 -1.50834930e+00
-9.36897099e-01 -6.79450929e-01 3.95014230e-03 -1.21642025e-02
-6.63380802e-01 -1.02055207e-01 -6.79476500e-01 -2.17119396e-01
-1.02935787e-02 -9.26771343e-01 -2.30632678e-01 2.92005777e-01
-1.10917568e-01 -7.50547230e-01 1.01229703e+00 -8.20065811e-02
8.92297506e-01 -2.17541766e+00 -3.06541789e-02 6.17433488e-01
2.29136333e-01 7.15478361e-02 -4.11471158e-01 1.03929806e+00
3.17238659e-01 -1.94688827e-01 2.12265953e-01 -6.54425323e-01
5.20104825e-01 5.95082283e-01 -3.47289681e-01 5.31157732e-01
-3.64477724e-01 6.61634147e-01 -6.62967801e-01 -5.03626645e-01
3.19580398e-02 5.90763211e-01 -8.23751390e-01 6.95477307e-01
-5.34411371e-01 1.50676057e-01 -4.55180526e-01 4.24391657e-01
6.61510468e-01 -5.80756724e-01 6.39820933e-01 -6.04291959e-03
2.03518778e-01 1.37775004e-01 -1.26035631e+00 1.11733246e+00
-9.21837151e-01 2.58357316e-01 1.84029162e-01 -1.12971187e+00
8.08943748e-01 5.18746018e-01 8.94739509e-01 -1.61417305e-01
-1.71925485e-01 2.28622258e-01 -4.36355531e-01 -4.52352762e-01
-2.27530226e-01 -3.52318883e-02 1.97010458e-01 1.12610197e+00
3.98887932e-01 7.47746706e-01 -7.64674321e-02 3.50908674e-02
1.11565745e+00 -2.74531841e-01 -1.98735848e-01 1.46800473e-01
3.89468640e-01 -5.18807173e-01 5.82661211e-01 1.07845306e+00
3.70720611e-03 2.16049030e-01 6.31829560e-01 -6.81083679e-01
-7.70860195e-01 -1.24191749e+00 1.02963686e-01 1.62523985e+00
-3.46389785e-02 -3.53436112e-01 -4.74924952e-01 -9.76237893e-01
4.06040341e-01 8.09483230e-01 -5.58774114e-01 -9.58534107e-02
-4.55128402e-01 -4.19238478e-01 1.18324541e-01 1.30575165e-01
5.38741983e-02 -1.01202703e+00 -3.25422466e-01 5.44768274e-01
1.90489832e-02 -9.44267452e-01 -5.20865440e-01 3.29187632e-01
-9.13217008e-01 -1.04948187e+00 -4.43408519e-01 -4.74404871e-01
4.60663348e-01 1.61508620e-01 9.63556290e-01 -3.02990258e-01
-1.05649225e-01 7.28018761e-01 -2.05069646e-01 -1.44946240e-02
-4.64508772e-01 3.53211403e-01 1.33749899e-02 6.47051811e-01
4.69573408e-01 -9.25564587e-01 -5.29027045e-01 4.28048462e-01
-8.99948478e-01 -4.63530451e-01 5.07336676e-01 8.81128609e-01
2.74774849e-01 8.77750367e-02 4.93201047e-01 -1.05870485e+00
4.74141389e-01 -9.58625019e-01 -5.78362286e-01 6.44196689e-01
-9.69102740e-01 3.22707325e-01 8.26265395e-01 -4.99177754e-01
-1.01456630e+00 -1.25666251e-02 1.46673903e-01 -6.25043452e-01
-1.41114578e-01 1.66176394e-01 -3.19179416e-01 1.77262038e-01
5.14370501e-01 3.66900384e-01 2.09418625e-01 -1.10057342e+00
4.25597280e-01 9.57309663e-01 2.42183894e-01 -8.09694052e-01
7.84430265e-01 4.65032101e-01 -3.02186042e-01 -4.43446636e-01
-6.20934606e-01 -5.28277576e-01 -4.25080389e-01 -2.75929838e-01
8.22862312e-02 -5.68786979e-01 -1.19459534e+00 2.90114909e-01
-1.04843986e+00 -3.75510991e-01 -6.14462554e-01 4.98161227e-01
-6.85185313e-01 1.88190162e-01 -2.77932853e-01 -9.27999377e-01
-2.66973644e-01 -1.02788818e+00 6.28677428e-01 1.29592329e-01
3.43662500e-02 -1.14509296e+00 4.18134719e-01 2.64494658e-01
6.99531317e-01 -2.17323169e-01 9.65503037e-01 -1.16595685e+00
-9.42548513e-01 -5.12112617e-01 -8.43932778e-02 2.99767017e-01
7.87150711e-02 -4.03100513e-02 -1.03583860e+00 -6.37921870e-01
1.41484765e-02 -3.56388152e-01 5.02906799e-01 -1.39224576e-02
1.16921699e+00 -9.10483062e-01 -2.31605947e-01 5.57952166e-01
1.56460905e+00 6.67527467e-02 1.28068000e-01 2.68323869e-01
3.81234676e-01 8.30394864e-01 -5.61181270e-02 1.14355040e+00
5.23796558e-01 8.96656692e-01 7.55345404e-01 4.17583853e-01
1.57669425e-01 -2.96012968e-01 5.61950147e-01 3.68885994e-01
4.36780393e-01 -3.50541592e-01 -7.82609940e-01 5.95673442e-01
-2.08290339e+00 -9.90029693e-01 4.52695280e-01 2.41594481e+00
6.60217106e-01 -3.50730211e-01 3.47835302e-01 -3.20624560e-01
6.23797953e-01 -5.59677221e-02 -1.06075060e+00 -4.16566461e-01
8.77699163e-03 3.34905028e-01 4.18839872e-01 4.06297505e-01
-6.61547005e-01 6.00973964e-01 6.11616325e+00 7.02239275e-01
-1.18076122e+00 3.51979792e-01 4.38972145e-01 -2.88194984e-01
-4.62983668e-01 1.70458212e-01 -8.10767353e-01 5.02561331e-01
1.13802409e+00 -4.38085586e-01 8.67954016e-01 1.32686317e+00
4.45415750e-02 2.78800488e-01 -1.40551078e+00 8.32662642e-01
-5.41762002e-02 -1.40234888e+00 1.36698857e-01 3.74987453e-01
7.11886346e-01 4.60409075e-01 2.39948064e-01 3.54215980e-01
6.94628954e-01 -8.19664001e-01 3.41727614e-01 6.89770997e-01
2.03893900e-01 -8.60145569e-01 6.34896517e-01 8.00835669e-01
-8.13033760e-01 -5.49195528e-01 -2.93387324e-01 5.58456719e-01
-4.31644432e-02 3.06280941e-01 -1.03623533e+00 2.63852388e-01
6.73418581e-01 2.24469930e-01 -2.78805226e-01 1.15619493e+00
2.03422606e-01 6.94045722e-01 -6.24822259e-01 9.95841175e-02
1.91436216e-01 4.64047827e-02 3.99209410e-01 1.03865170e+00
4.80352104e-01 -2.81550199e-01 3.34565252e-01 6.47434533e-01
-2.96290874e-01 3.20428401e-01 -4.25257236e-01 1.17412070e-02
6.48389101e-01 1.01492369e+00 7.18217939e-02 -3.84297967e-01
-3.86266828e-01 9.02795553e-01 6.33279741e-01 5.37774920e-01
-4.75449532e-01 3.02593708e-02 8.69058669e-01 1.62003040e-01
6.57907307e-01 1.27246439e-01 2.63531059e-01 -1.14231408e+00
1.54370531e-01 -8.38362157e-01 6.83919668e-01 -1.85844690e-01
-1.78083551e+00 7.26624370e-01 -1.97809562e-01 -9.37190831e-01
-7.07375407e-01 -2.76192307e-01 -7.12209761e-01 8.42936456e-01
-1.39744055e+00 -1.09850121e+00 -1.16927385e-01 8.43430161e-01
4.01734829e-01 -5.66605687e-01 1.00393617e+00 1.64405137e-01
-5.60783982e-01 9.00684595e-01 8.03759158e-01 -3.21519881e-01
6.45797372e-01 -1.04209507e+00 -7.28991255e-02 1.74137846e-01
3.60491484e-01 5.98067284e-01 6.13652289e-01 -6.93839565e-02
-1.53586924e+00 -1.10756171e+00 1.03946531e+00 -4.64460216e-02
6.19112790e-01 -2.05898926e-01 -1.01116526e+00 7.06271231e-01
4.10083123e-02 2.72144347e-01 9.66900289e-01 2.83780605e-01
-7.50911534e-01 -7.48683512e-01 -1.32323194e+00 2.90140390e-01
6.33695424e-01 -6.32221520e-01 -4.52071317e-02 6.07345879e-01
2.28954896e-01 2.12296307e-01 -8.56199682e-01 -4.28715721e-02
7.17408061e-01 -1.29401398e+00 8.20996106e-01 -8.96524131e-01
-2.90064901e-01 2.19676450e-01 -4.88173932e-01 -1.04560590e+00
-1.69264540e-01 -8.95820558e-01 -7.07897007e-01 1.19295800e+00
1.29761621e-01 -9.48531866e-01 1.12555993e+00 8.83647561e-01
4.51828182e-01 -1.04278564e+00 -1.22209477e+00 -9.88258958e-01
2.09702495e-02 -4.34026927e-01 1.05714953e+00 3.95790935e-01
-1.31088778e-01 -1.10770039e-01 -3.46799612e-01 3.74635518e-01
1.14159608e+00 3.82779509e-01 1.15751553e+00 -1.28763628e+00
-7.40922928e-01 -2.02170521e-01 -7.48324767e-02 -1.00700557e+00
4.10130799e-01 -9.13410544e-01 -1.86941922e-01 -1.29051507e+00
5.23209035e-01 -7.53044486e-01 -6.53123617e-01 4.35363829e-01
3.31634074e-01 -1.62773609e-01 4.37811017e-01 6.90008938e-01
-8.46567929e-01 6.64449751e-01 6.00390255e-01 6.91669211e-02
-2.25205198e-01 6.05556190e-01 -5.33848345e-01 4.86818910e-01
7.60275960e-01 -7.02135265e-01 -3.64682943e-01 -1.82855785e-01
-6.60547242e-02 2.33235911e-01 3.45467210e-01 -6.57006323e-01
3.97395521e-01 -3.33294809e-01 -8.13117698e-02 -5.38283229e-01
3.18622410e-01 -1.19382334e+00 5.04931211e-01 2.43098691e-01
-4.26391721e-01 -2.68462330e-01 -4.39041078e-01 6.73060596e-01
1.58468522e-02 -2.61669874e-01 7.54115164e-01 1.04065485e-01
-9.75139067e-02 7.09960818e-01 -4.13618445e-01 -1.48990199e-01
1.06202841e+00 6.40729740e-02 -1.28834322e-01 -7.84394145e-01
-9.72780824e-01 3.61577213e-01 5.24780929e-01 2.63702929e-01
4.30380344e-01 -1.28846693e+00 -5.42011082e-01 1.89528644e-01
1.09791286e-01 -2.92760551e-01 2.88316369e-01 5.77887058e-01
-2.19382674e-01 4.31312203e-01 1.28208727e-01 -6.33618891e-01
-1.02227938e+00 5.67379534e-01 3.43963921e-01 -5.92570662e-01
-5.30574441e-01 5.64349711e-01 2.53623605e-01 -6.75682724e-01
7.35455334e-01 3.02274525e-01 1.02743596e-01 6.51981980e-02
4.79391038e-01 5.69254160e-01 -3.84038091e-02 -5.59702992e-01
-3.22619200e-01 1.78272992e-01 -3.66443723e-01 -1.95309415e-01
1.84464025e+00 -7.00998828e-02 -1.31411806e-01 5.56950033e-01
1.60847998e+00 -1.86312333e-01 -1.55954969e+00 -9.93785143e-01
2.20765993e-01 -6.41892254e-01 -7.76561573e-02 -6.30865633e-01
-1.26563787e+00 6.74444914e-01 7.33214438e-01 1.40989587e-01
7.75817513e-01 4.21559423e-01 7.13476300e-01 6.63801968e-01
6.10223293e-01 -8.81539762e-01 3.92817408e-01 3.56225342e-01
7.25756228e-01 -9.69764054e-01 -3.01925272e-01 2.23992959e-01
-4.66521174e-01 1.34574413e+00 2.91437089e-01 -1.10053688e-01
8.50138962e-01 -1.77631266e-02 5.41088060e-02 -1.28832817e-01
-1.43302381e+00 2.14575958e-02 -1.71200216e-01 6.15651011e-01
-1.75875872e-01 -3.71779390e-02 3.37142736e-01 4.09352362e-01
1.96877897e-01 -4.81977947e-02 2.70680755e-01 8.32537830e-01
-3.17164570e-01 -1.65603173e+00 -3.76883447e-01 4.42221999e-01
-3.44720364e-01 4.60872829e-01 -1.28897056e-01 3.01615804e-01
1.85830016e-02 8.23191464e-01 5.11426805e-03 -1.18904933e-01
2.29150355e-01 3.15903962e-01 2.10939869e-01 -6.91226125e-01
-4.78128672e-01 1.21902324e-01 -3.09990376e-01 -7.75231898e-01
9.69307125e-02 -9.07899022e-01 -8.32864225e-01 -4.51906770e-01
-3.21478426e-01 4.05624241e-01 8.32323194e-01 7.32908666e-01
5.29562950e-01 -8.00965875e-02 1.57233751e+00 -8.55195582e-01
-1.23330450e+00 -7.02349663e-01 -7.99359083e-01 1.63412929e-01
2.63349026e-01 -5.05443692e-01 -5.88190615e-01 -1.65556297e-01]
|
[5.779453754425049, 6.280388832092285]
|
fc0a0d78-eb70-4a9c-911c-35cdf5cbc965
|
self-supervised-learning-of-a-biologically
|
2006.16976
| null |
https://arxiv.org/abs/2006.16976v1
|
https://arxiv.org/pdf/2006.16976v1.pdf
|
Self-Supervised Learning of a Biologically-Inspired Visual Texture Model
|
We develop a model for representing visual texture in a low-dimensional feature space, along with a novel self-supervised learning objective that is used to train it on an unlabeled database of texture images. Inspired by the architecture of primate visual cortex, the model uses a first stage of oriented linear filters (corresponding to cortical area V1), consisting of both rectified units (simple cells) and pooled phase-invariant units (complex cells). These responses are processed by a second stage (analogous to cortical area V2) consisting of convolutional filters followed by half-wave rectification and pooling to generate V2 'complex cell' responses. The second stage filters are trained on a set of unlabeled homogeneous texture images, using a novel contrastive objective that maximizes the distance between the distribution of V2 responses to individual images and the distribution of responses across all images. When evaluated on texture classification, the trained model achieves substantially greater data-efficiency than a variety of deep hierarchical model architectures. Moreover, we show that the learned model exhibits stronger representational similarity to texture responses of neural populations recorded in primate V2 than pre-trained deep CNNs.
|
['Nikhil Parthasarathy', 'Eero P. Simoncelli']
|
2020-06-30
| null | null | null | null |
['texture-classification']
|
['computer-vision']
|
[ 5.88723958e-01 3.13307405e-01 1.27735496e-01 -6.09350860e-01
-6.13164127e-01 -3.48463207e-01 7.71978140e-01 -1.58378020e-01
-4.36466157e-01 4.18950915e-01 2.63196468e-01 3.12033564e-01
6.14950806e-02 -8.75856161e-01 -8.11414421e-01 -1.02216113e+00
-7.85192624e-02 2.53900766e-01 4.51386631e-01 -1.72403425e-01
3.54039401e-01 6.02567792e-01 -1.54113734e+00 1.02055371e+00
1.79104999e-01 1.49193478e+00 2.50870883e-01 6.13558412e-01
1.44117624e-01 1.01297820e+00 -2.52471447e-01 1.81041434e-01
1.19866848e-01 -3.55833262e-01 -1.02848816e+00 3.17172557e-01
6.44735277e-01 4.71258946e-02 -5.20289600e-01 9.32261646e-01
2.47601867e-01 1.78485617e-01 1.05691350e+00 -5.09807408e-01
-1.06022501e+00 6.61799610e-02 -3.25166672e-01 4.71490294e-01
-5.82183599e-02 -3.13069834e-03 1.01852643e+00 -1.00952482e+00
9.63300824e-01 1.32134295e+00 4.14430022e-01 5.95037878e-01
-1.98104060e+00 7.20063224e-02 2.00058799e-02 -7.58017823e-02
-1.28656507e+00 -6.46393418e-01 3.92953128e-01 -4.82224554e-01
1.24643779e+00 1.04397848e-01 4.70402092e-01 1.07434916e+00
6.43344462e-01 7.22198904e-01 1.34697258e+00 -3.65056306e-01
3.71250689e-01 1.22791439e-01 -7.31159002e-02 6.82755888e-01
-2.28762358e-01 9.49436203e-02 -4.82274354e-01 3.10468916e-02
1.25071120e+00 -1.27425462e-01 -1.76313251e-01 -4.11228538e-01
-1.04205132e+00 7.18352377e-01 8.35715473e-01 2.72001654e-01
-5.25632083e-01 3.84751670e-02 4.71300855e-02 5.00006020e-01
5.44447243e-01 4.56734657e-01 -3.27377856e-01 7.27831602e-01
-7.32215703e-01 -5.01799695e-02 6.27669573e-01 5.83001018e-01
1.14570403e+00 2.02682927e-01 -1.79683656e-01 1.13673258e+00
1.62435696e-01 4.03371662e-01 6.83845997e-01 -8.03161502e-01
-9.64425579e-02 7.29734004e-01 -3.94548774e-01 -8.46895099e-01
-3.96500826e-01 -1.81748822e-01 -1.05292499e+00 4.99688148e-01
5.13697505e-01 3.37202787e-01 -9.47967470e-01 1.60954368e+00
-3.36173058e-01 -5.19380450e-01 5.17877229e-02 1.01402438e+00
6.75315201e-01 5.53533614e-01 1.05027854e-01 6.53379112e-02
1.28296030e+00 -6.30847633e-01 -1.58605754e-01 -3.60780805e-01
2.45065778e-01 -6.17993057e-01 8.61385643e-01 2.23003089e-01
-1.23507810e+00 -7.61368692e-01 -9.38652754e-01 -2.50925750e-01
-4.21984553e-01 2.21843481e-01 4.53459620e-01 1.36546150e-01
-1.38559616e+00 5.26076376e-01 -5.96039832e-01 -7.12577581e-01
8.89751494e-01 5.29070914e-01 -8.09011817e-01 3.55324000e-02
-5.88963509e-01 9.56735671e-01 -5.82128437e-03 4.27936204e-02
-1.15093303e+00 -2.68845737e-01 -8.80508602e-01 2.65107512e-01
-1.78836197e-01 -5.65542698e-01 1.01767564e+00 -1.54140079e+00
-1.45965767e+00 1.48399138e+00 -3.66074860e-01 -3.15099239e-01
-4.98319827e-02 3.48155171e-01 -1.75784677e-01 4.92305338e-01
-1.03798077e-01 1.15121901e+00 9.86421824e-01 -1.35512924e+00
-5.28411806e-01 -4.05874670e-01 -3.45062912e-01 7.76131377e-02
-7.02754557e-02 -1.23897605e-01 -1.51818961e-01 -6.42743111e-01
5.02355993e-01 -7.92510748e-01 -2.76523381e-01 1.09669911e-02
-1.88059151e-01 -1.82569414e-01 5.93666971e-01 -3.89295042e-01
6.36823416e-01 -2.34542108e+00 3.49466980e-01 5.47635257e-01
3.88720185e-01 -2.66033620e-01 -5.81140518e-01 2.42339253e-01
-1.92366153e-01 -2.56722029e-02 -3.56857300e-01 -4.44568582e-02
-1.53968349e-01 2.37208039e-01 -4.89438355e-01 5.75719833e-01
9.33609903e-01 1.04429352e+00 -4.96012539e-01 -8.80567580e-02
-2.81930268e-02 4.59055543e-01 -7.14686692e-01 1.71137482e-01
-3.52702200e-01 4.68759775e-01 -2.57810324e-01 6.40643001e-01
4.63014871e-01 -5.23518562e-01 2.86765158e-01 -2.18094066e-01
-1.51707381e-01 1.40158743e-01 -6.01098359e-01 1.36688280e+00
-2.87301362e-01 1.04138005e+00 -8.38505998e-02 -1.43280578e+00
1.26659894e+00 1.91875488e-01 2.03745246e-01 -1.05470145e+00
3.86027366e-01 1.66728288e-01 2.13109389e-01 -5.61417222e-01
1.47804609e-02 2.26176642e-02 -7.26500303e-02 4.52717751e-01
8.50490272e-01 -2.07443774e-01 -6.80624740e-03 -6.38237894e-02
1.19323647e+00 1.44015789e-01 1.03574991e-01 -7.86496997e-01
5.30304968e-01 -8.03006161e-03 2.47430056e-01 7.84206986e-01
4.94208522e-02 8.64995182e-01 7.04241633e-01 -9.22425091e-01
-1.33964634e+00 -1.29247499e+00 -3.86135578e-01 1.25276959e+00
-3.13208513e-02 2.26866722e-01 -7.01625645e-01 -1.81332678e-01
1.13996388e-02 -1.27214149e-01 -1.31840384e+00 -8.10137242e-02
-4.05256569e-01 -6.85643733e-01 2.86989868e-01 5.80609560e-01
6.05180562e-01 -1.59438908e+00 -7.53611982e-01 1.46972641e-01
2.11160351e-03 -8.45128000e-01 -1.37869149e-01 7.77442455e-01
-8.23335230e-01 -1.00444615e+00 -7.05252647e-01 -1.42950821e+00
1.12287307e+00 3.03815715e-02 1.08319318e+00 -1.08575746e-01
-4.56780136e-01 3.17629278e-01 -7.78644010e-02 -2.02566698e-01
2.95983739e-02 -9.39526781e-02 -1.32848963e-01 4.55943018e-01
5.25135458e-01 -3.53464782e-01 -6.43023968e-01 4.06763524e-01
-9.09420669e-01 -3.95892970e-02 6.95997298e-01 1.22608089e+00
9.45912302e-01 -3.41476053e-01 4.66598988e-01 -8.03577423e-01
3.61175597e-01 -3.26322258e-01 -4.33802903e-01 2.16789410e-01
1.03895947e-01 2.34072596e-01 5.79553843e-01 -5.98997891e-01
-1.20834482e+00 2.34090209e-01 9.53758433e-02 6.23824187e-02
-3.79098475e-01 2.13558242e-01 2.17058271e-01 -4.21741188e-01
1.17611599e+00 4.16600227e-01 1.71000808e-01 -2.71527559e-01
1.67322814e-01 5.05081236e-01 8.21828246e-01 -5.06912649e-01
3.06067646e-01 7.97774136e-01 -1.62291050e-01 -1.14154208e+00
-7.32492089e-01 -1.20671354e-01 -6.99574172e-01 -6.74246028e-02
1.03709960e+00 -7.19283700e-01 -7.31983542e-01 6.63482308e-01
-9.25022721e-01 -7.69664526e-01 -5.71280718e-01 4.20402825e-01
-1.05212486e+00 -1.99663818e-01 -9.38134313e-01 -4.88552898e-01
-5.86157627e-02 -9.31297362e-01 1.07887077e+00 2.12269410e-01
-1.29657686e-01 -8.58006239e-01 1.64311901e-01 -2.72676706e-01
6.49025798e-01 -1.55150816e-02 1.25789475e+00 -4.45480227e-01
-5.58795512e-01 -4.04965915e-02 -5.21628380e-01 4.23161000e-01
3.29172565e-03 -1.07063375e-01 -1.37033689e+00 -2.19750643e-01
6.84640929e-02 -7.79564738e-01 1.35764074e+00 6.73306346e-01
1.28478336e+00 -2.85222054e-01 -2.28915781e-01 8.42727721e-01
1.51049161e+00 1.70735344e-01 9.11481857e-01 1.79991871e-01
2.09425315e-01 9.44968760e-01 -4.18875754e-01 -1.17326096e-01
7.12101683e-02 1.67034030e-01 1.43903881e-01 -4.69669729e-01
-6.00155219e-02 9.95254070e-02 -4.21256907e-02 4.95145231e-01
-4.08024639e-01 1.04109151e-02 -6.10390902e-01 5.48148394e-01
-1.71259797e+00 -1.07460487e+00 2.91991293e-01 2.15990472e+00
6.16953492e-01 2.48420745e-01 2.60603875e-02 5.32252118e-02
4.81535345e-01 -1.46025196e-02 -4.28598404e-01 -6.70626938e-01
-7.71077514e-01 8.26547444e-01 2.98368692e-01 3.71175945e-01
-1.21500552e+00 1.01285005e+00 7.52689695e+00 4.35110837e-01
-1.34818959e+00 -3.12026620e-01 9.20404553e-01 2.22798675e-01
-2.24769011e-01 -2.03738883e-01 -2.44796053e-01 -2.66439896e-02
6.61637485e-01 3.21694911e-01 7.42373347e-01 5.33132374e-01
-2.07210302e-01 -2.81957537e-01 -1.15432096e+00 6.42606378e-01
5.23304157e-02 -1.55691326e+00 1.07095391e-01 6.02876581e-02
7.85351276e-01 2.27474824e-01 3.20306450e-01 3.06631625e-01
3.15655291e-01 -1.26885271e+00 4.65397269e-01 7.86992371e-01
9.09866929e-01 -3.33644748e-01 4.63698596e-01 -3.07454541e-03
-9.64982212e-01 -3.36668730e-01 -1.00242889e+00 -1.42282426e-01
-4.24305290e-01 1.34841576e-01 -3.10703218e-01 -1.42945722e-01
1.03698838e+00 6.05247796e-01 -5.13056099e-01 7.84946799e-01
1.09070726e-02 2.85843730e-01 -9.11024511e-02 6.06697574e-02
2.21366212e-01 1.00491077e-01 -1.89875718e-03 1.30464363e+00
-1.12173110e-01 3.23543370e-01 -2.12654874e-01 8.49851668e-01
-8.97669420e-02 2.38821015e-01 -8.79451513e-01 1.86030120e-01
1.15501106e-01 1.53717864e+00 -8.17610562e-01 -4.86897886e-01
-4.38387901e-01 9.10690904e-01 7.20078766e-01 8.05851758e-01
-5.60453581e-03 -2.71595299e-01 2.44702742e-01 2.68592536e-01
5.56433678e-01 5.22906296e-02 -3.50602448e-01 -1.01598704e+00
-3.63122165e-01 -6.78108573e-01 1.67445913e-01 -1.02276850e+00
-1.64402735e+00 1.06181264e+00 -4.13387746e-01 -1.03132701e+00
-1.56760812e-02 -1.10762441e+00 -4.96386439e-01 1.10653305e+00
-1.14499104e+00 -1.07029378e+00 -2.74365276e-01 7.63348281e-01
2.58641660e-01 -4.94695783e-01 1.17644572e+00 -1.37308896e-01
-2.01557502e-01 3.76839936e-01 -7.30068609e-02 2.38568202e-01
5.56536555e-01 -1.19056344e+00 3.34504247e-01 2.54050344e-01
-9.22087021e-03 4.66056466e-01 4.02714051e-02 -2.31913373e-01
-1.12028182e+00 -9.62647259e-01 8.87616813e-01 -1.94114402e-01
4.47328657e-01 -5.47042251e-01 -1.13312924e+00 7.43044138e-01
2.75286704e-01 5.37166119e-01 5.74253380e-01 -1.02044240e-01
-6.51906490e-01 -5.22775427e-02 -1.03700280e+00 4.19504762e-01
8.11302662e-01 -7.92032719e-01 -6.72740042e-01 1.02602586e-01
-1.04659937e-01 -8.68725777e-02 -8.91937256e-01 4.37278241e-01
8.86753798e-01 -8.58496189e-01 6.74142420e-01 -8.84260178e-01
6.14688516e-01 -9.22827423e-02 -4.35424328e-01 -1.19770527e+00
-1.00519550e+00 -4.75588664e-02 5.07427275e-01 3.53480518e-01
3.97862554e-01 -6.94888294e-01 6.17282033e-01 5.75721040e-02
-1.46026844e-02 -6.26240492e-01 -7.72574067e-01 -2.46054739e-01
6.26046434e-02 3.55556697e-01 -2.54071355e-01 8.02487075e-01
3.22035924e-02 4.08873022e-01 -1.04592210e-02 -1.26667172e-01
5.57374895e-01 1.32225350e-01 3.90259415e-01 -1.28894114e+00
-1.06415972e-01 -7.56629109e-01 -6.89762056e-01 -1.09024966e+00
1.11565515e-01 -1.01696134e+00 2.86807686e-01 -1.27565312e+00
3.25251877e-01 -2.04734579e-01 -4.42829520e-01 6.69438660e-01
2.95151293e-01 7.75139570e-01 -2.17288256e-01 4.23523217e-01
-3.34478110e-01 2.72770733e-01 1.42743802e+00 -3.22028905e-01
-8.71516764e-02 -3.08818936e-01 -4.85647440e-01 7.23449349e-01
5.98097086e-01 -2.98095137e-01 -6.33743182e-02 -5.01176715e-01
1.70736033e-02 -7.70458207e-02 4.94823605e-01 -8.54952335e-01
1.98774785e-01 9.91149247e-03 1.06051552e+00 -5.63531891e-02
1.41271934e-01 -3.64434779e-01 -1.75754428e-01 2.69086808e-01
-6.15409255e-01 -1.08745322e-01 2.49318987e-01 2.62172580e-01
-4.19796526e-01 1.69812113e-01 1.09586608e+00 -3.94625753e-01
-8.85798454e-01 1.86545387e-01 -1.15734363e+00 -2.40528658e-01
6.12370908e-01 -3.64765346e-01 -6.27138495e-01 -1.82063371e-01
-1.13587308e+00 -2.88298756e-01 4.96252596e-01 2.30950966e-01
7.81515718e-01 -1.34232843e+00 -7.31449246e-01 8.01616132e-01
1.99639633e-01 -2.02027172e-01 4.57509190e-01 6.16305351e-01
-5.96494913e-01 3.34426999e-01 -1.08422828e+00 -8.62375200e-01
-5.63543618e-01 3.95713896e-01 6.31717682e-01 7.95033574e-02
-4.69232172e-01 1.02726352e+00 9.31952059e-01 -3.80193681e-01
4.40353267e-02 -2.29817331e-01 -4.31771457e-01 -2.00664550e-01
4.24758255e-01 -2.53747523e-01 1.25844926e-01 -7.64333487e-01
-2.52859890e-01 6.76336169e-01 -2.80202031e-02 -8.27803388e-02
1.36983788e+00 1.57985389e-01 -3.68749738e-01 4.35291320e-01
1.28611350e+00 -3.01906347e-01 -1.48495030e+00 -5.44461846e-01
1.09580897e-01 1.74891092e-02 -1.91104844e-01 -5.43725789e-01
-1.06621158e+00 9.90795791e-01 5.85564971e-01 2.16366157e-01
1.33319771e+00 6.29038438e-02 4.10159081e-02 5.49020767e-01
2.93757200e-01 -1.06834233e+00 4.63198811e-01 6.86017692e-01
9.71742868e-01 -1.23396873e+00 -4.88083482e-01 -1.56341881e-01
-5.86358607e-01 1.20442474e+00 6.07169092e-01 -9.51689899e-01
7.16363370e-01 3.92113596e-01 2.43951753e-01 -4.45626676e-01
-1.21241260e+00 -3.58540535e-01 4.67542291e-01 7.66291440e-01
6.23744011e-01 4.22604382e-03 8.89868960e-02 3.32127303e-01
1.26367614e-01 -1.61757395e-01 2.98865229e-01 8.20511997e-01
-6.57124221e-01 -5.70467472e-01 -2.13924125e-02 6.45803988e-01
-3.72767299e-01 1.73176862e-02 -3.66899669e-01 5.11361778e-01
1.16596585e-02 6.41150475e-01 7.98426390e-01 -4.05948967e-01
3.78702968e-01 -8.00336674e-02 9.77119327e-01 -7.44622588e-01
-5.95092237e-01 3.99802476e-01 -2.64743358e-01 -4.83099312e-01
-4.75092292e-01 -2.18589380e-01 -8.01089764e-01 -1.87720750e-02
1.93923682e-01 -2.45395243e-01 3.50289822e-01 7.15068281e-01
2.62063406e-02 4.22329128e-01 7.42225289e-01 -1.22281814e+00
-3.34565282e-01 -1.14271533e+00 -8.03662002e-01 6.02943599e-01
3.65007877e-01 -5.20654917e-01 -1.26920044e-01 4.62274849e-01]
|
[9.554303169250488, 2.3968663215637207]
|
ee48c168-f625-4782-8720-f1695ca2b004
|
line-as-a-visual-sentence-context-aware-line
|
2109.04753
| null |
https://arxiv.org/abs/2109.04753v1
|
https://arxiv.org/pdf/2109.04753v1.pdf
|
Line as a Visual Sentence: Context-aware Line Descriptor for Visual Localization
|
Along with feature points for image matching, line features provide additional constraints to solve visual geometric problems in robotics and computer vision (CV). Although recent convolutional neural network (CNN)-based line descriptors are promising for viewpoint changes or dynamic environments, we claim that the CNN architecture has innate disadvantages to abstract variable line length into the fixed-dimensional descriptor. In this paper, we effectively introduce Line-Transformers dealing with variable lines. Inspired by natural language processing (NLP) tasks where sentences can be understood and abstracted well in neural nets, we view a line segment as a sentence that contains points (words). By attending to well-describable points on aline dynamically, our descriptor performs excellently on variable line length. We also propose line signature networks sharing the line's geometric attributes to neighborhoods. Performing as group descriptors, the networks enhance line descriptors by understanding lines' relative geometries. Finally, we present the proposed line descriptor and matching in a Point and Line Localization (PL-Loc). We show that the visual localization with feature points can be improved using our line features. We validate the proposed method for homography estimation and visual localization.
|
['Ayoung Kim', 'Sungho Yoon']
|
2021-09-10
| null | null | null | null |
['homography-estimation']
|
['computer-vision']
|
[-1.71649307e-01 -2.05497637e-01 -3.25429916e-01 -5.95908523e-01
-2.64446437e-01 -7.25574017e-01 7.79306293e-01 3.40643525e-01
-4.15267020e-01 3.02926689e-01 -2.89268084e-02 -4.53549586e-02
-1.65676340e-01 -1.06449914e+00 -1.00485456e+00 -1.04134277e-01
-4.07966487e-02 1.96977943e-01 1.16728783e-01 -5.45158446e-01
6.52756631e-01 1.40374303e+00 -1.28870666e+00 1.47810638e-01
5.32836556e-01 1.10045767e+00 -1.46546364e-01 4.79328275e-01
-4.36196625e-01 3.95169497e-01 -5.02774835e-01 -2.76936710e-01
5.04225850e-01 -2.67109089e-02 -6.41265273e-01 2.95303553e-01
1.32426679e+00 -3.66274059e-01 -7.17355251e-01 1.11095881e+00
3.06015968e-01 9.72566530e-02 7.29535460e-01 -1.57612193e+00
-1.12848926e+00 3.19498420e-01 -7.51836181e-01 -2.26772428e-01
6.67051494e-01 -1.93053707e-01 8.97793055e-01 -1.24494278e+00
9.17366505e-01 1.28237891e+00 1.21498370e+00 1.14542268e-01
-8.55008662e-01 -1.23590387e-01 -1.13970771e-01 3.10216367e-01
-1.53508401e+00 -1.81550160e-01 1.29041529e+00 -3.82175446e-01
1.14677703e+00 7.10785836e-02 9.05564427e-01 8.33329797e-01
3.81582975e-01 7.18833804e-01 2.46474802e-01 -5.62834322e-01
-4.81995009e-02 -2.11453140e-01 2.22281262e-01 1.01465499e+00
2.80779332e-01 -1.34901758e-02 -5.67864358e-01 3.32173020e-01
1.33350611e+00 4.00454342e-01 -4.36905563e-01 -1.06062758e+00
-1.64638889e+00 8.13533604e-01 1.03537095e+00 2.17235461e-01
-8.03225487e-02 3.18998218e-01 4.96856153e-01 2.16675073e-01
7.23370537e-02 6.56520963e-01 -8.71558860e-02 2.17423603e-01
-7.65752971e-01 1.26721025e-01 6.99176729e-01 1.77841091e+00
1.16364920e+00 5.36722317e-02 -8.98445770e-02 9.51115191e-01
6.13056533e-02 9.30347443e-01 2.43975744e-01 -1.09084284e+00
6.25795722e-01 5.92507422e-01 3.74255031e-02 -1.76016271e+00
-8.71672153e-01 -4.22762364e-01 -1.15870380e+00 1.51452184e-01
3.26725990e-01 2.78816760e-01 -5.73734820e-01 1.24734676e+00
-3.35682958e-01 -3.62825990e-01 -5.02650738e-02 7.27688611e-01
9.18674231e-01 4.28829134e-01 -4.47768301e-01 2.82458991e-01
1.35983753e+00 -1.24460292e+00 -5.71270108e-01 -1.83768466e-01
9.21331346e-01 -8.15333247e-01 1.00897610e+00 1.34923354e-01
-1.00764668e+00 -9.12287891e-01 -1.39473546e+00 -6.30355716e-01
-7.83516169e-01 3.02044153e-01 6.42145097e-01 1.29506558e-01
-1.24433780e+00 5.61294556e-01 -3.87773544e-01 -7.92569458e-01
4.20187593e-01 3.63742232e-01 -7.54191220e-01 1.34456426e-01
-8.04920971e-01 8.58443677e-01 1.79758638e-01 3.75562876e-01
7.36812949e-02 -5.04449487e-01 -1.35546255e+00 9.82756764e-02
1.07786521e-01 -9.90796030e-01 8.95174861e-01 -5.02270162e-01
-1.16713810e+00 1.00205767e+00 -1.75192326e-01 -6.53877258e-01
7.09168673e-01 -2.64243186e-01 -1.74535230e-01 5.62481999e-01
1.50811553e-01 1.08553839e+00 5.74779868e-01 -1.35252202e+00
-5.15060008e-01 -3.14157605e-01 3.07170600e-01 2.25745723e-01
-2.88117193e-02 -3.84949595e-01 -2.51975089e-01 -3.32551330e-01
6.15994930e-01 -6.84810519e-01 3.54764163e-02 9.01704013e-01
-5.57225049e-01 -4.82709855e-01 1.15635681e+00 -3.31954807e-02
7.40002930e-01 -1.98974562e+00 -3.76859009e-01 3.34369987e-01
3.79860520e-01 7.76098371e-02 -1.79111838e-01 7.71025717e-01
-9.86390263e-02 6.49537891e-02 -2.94892788e-02 -2.40901932e-01
5.33531122e-02 9.38965529e-02 -3.40071708e-01 8.49098086e-01
3.28955472e-01 1.17626774e+00 -8.71493578e-01 -5.82256079e-01
6.94236636e-01 4.35511798e-01 -5.03311992e-01 -2.09507644e-01
1.21979415e-01 -1.18857808e-02 -2.10364476e-01 4.88394052e-01
8.75985861e-01 -5.16374521e-02 -6.11253858e-01 -9.00227070e-01
-4.17392999e-01 -5.55990696e-01 -9.05562222e-01 2.22114897e+00
-4.65295911e-01 1.25077271e+00 -6.97160900e-01 -7.52169788e-01
1.61932600e+00 -3.52102220e-01 4.72636610e-01 -7.18586624e-01
7.12382197e-02 2.18093574e-01 -3.78251374e-01 -5.96310079e-01
8.71061206e-01 7.01492548e-01 -1.76537678e-01 -1.97111011e-01
4.79796836e-05 -8.53511393e-01 6.26570284e-02 4.08416726e-02
6.30959213e-01 3.16996992e-01 5.99238098e-01 -1.90149039e-01
7.61986673e-01 1.57040969e-01 6.34886324e-02 9.67445254e-01
-3.05826873e-01 8.63438308e-01 2.40494758e-01 -9.02952552e-01
-1.61849928e+00 -1.13325119e+00 -2.42134690e-01 3.82500827e-01
6.05241239e-01 -3.54240924e-01 -3.67238909e-01 -1.99555099e-01
2.05149837e-02 1.69788823e-01 -4.31670368e-01 -4.98326160e-02
-8.33471656e-01 4.51420695e-01 7.30071247e-01 8.24179053e-01
1.13507104e+00 -7.51465917e-01 -5.57271183e-01 -6.67521805e-02
2.01761678e-01 -1.36317945e+00 -7.39431202e-01 5.82689345e-02
-3.60981524e-01 -9.85566497e-01 -7.79151917e-01 -1.39204681e+00
8.16389203e-01 7.37531006e-01 8.79427016e-01 -2.05591377e-02
-1.40532985e-01 7.19679296e-01 -2.99738526e-01 -3.56287450e-01
1.09522641e-01 -2.95079406e-02 5.82032651e-02 -3.48456740e-01
5.66178858e-01 -3.20580870e-01 -7.26697385e-01 1.04006715e-01
-6.40499771e-01 2.66727526e-02 1.72611877e-01 8.18680584e-01
4.97202784e-01 -5.95386088e-01 -6.93223029e-02 -3.99994075e-01
5.65957904e-01 9.55609977e-02 -6.56850457e-01 4.53267783e-01
-1.68734327e-01 5.96916154e-02 8.46225739e-01 -2.19588265e-01
-5.06102681e-01 2.01037928e-01 1.34802023e-02 -5.37743747e-01
-3.75715554e-01 2.47685209e-01 1.23484626e-01 -7.20144868e-01
8.06847930e-01 2.41412595e-01 -1.21800728e-01 4.47918549e-02
8.01514864e-01 3.78027707e-01 8.14781129e-01 -4.27068979e-01
1.08451319e+00 8.37814748e-01 6.44563258e-01 -1.15103352e+00
-7.64053822e-01 -6.97619975e-01 -1.44063365e+00 -2.99952209e-01
8.76286209e-01 -7.52383292e-01 -1.07048678e+00 5.33633351e-01
-1.85004520e+00 2.30132774e-01 -3.25231135e-01 3.59009057e-01
-1.08489943e+00 9.21478391e-01 -4.10524398e-01 -3.97374272e-01
-1.19919270e-01 -9.08014238e-01 1.38882995e+00 3.15459430e-01
-9.57843214e-02 -1.28798175e+00 -5.77974394e-02 -3.33391488e-01
1.13355979e-01 4.30983782e-01 7.59759426e-01 -5.66415370e-01
-6.33818030e-01 -3.84771883e-01 -6.76235139e-01 1.98434532e-01
-3.49422134e-02 1.82600588e-01 -6.56642795e-01 -6.89332038e-02
-3.00809354e-01 -1.43303424e-01 6.33204758e-01 2.04376653e-01
1.32946897e+00 -6.64481521e-02 -3.40937644e-01 1.12725198e+00
1.68041205e+00 2.19578698e-01 5.70475340e-01 7.49173284e-01
9.66064692e-01 6.26717269e-01 3.05434257e-01 4.81942803e-01
5.24798155e-01 6.03787243e-01 3.12238485e-01 -4.91263092e-01
-2.26285502e-01 -5.42321563e-01 -6.97370693e-02 7.81330228e-01
1.28489360e-01 -2.03023374e-01 -1.00716782e+00 2.58267581e-01
-1.82039630e+00 -1.18110442e+00 -5.82641840e-01 1.86133969e+00
2.14254573e-01 6.65597850e-04 -1.36855215e-01 1.20218424e-02
5.98104835e-01 2.07171217e-01 -3.65577132e-01 -6.41771317e-01
-8.34201932e-01 -2.09145218e-01 8.69209528e-01 4.04666036e-01
-1.20264518e+00 1.13031459e+00 6.11312008e+00 5.76277316e-01
-1.21278465e+00 -6.12465143e-01 8.40386525e-02 8.25397491e-01
-2.81043261e-01 -5.00368811e-02 -7.10367262e-01 -2.16605872e-01
2.10399311e-02 -2.51426935e-01 1.60352483e-01 9.79848385e-01
1.67121530e-01 1.38021678e-01 -1.40327740e+00 1.62372291e+00
6.75962806e-01 -1.68371820e+00 5.44801235e-01 -2.65070945e-01
7.32691646e-01 1.25956992e-02 1.40136942e-01 -1.04101904e-01
-3.41407567e-01 -8.55190814e-01 9.03541803e-01 9.26836193e-01
8.02619040e-01 -5.77461958e-01 6.41375899e-01 6.53339876e-03
-1.40696561e+00 4.80725430e-02 -8.58445406e-01 4.46678922e-02
-7.92808533e-02 -4.41892557e-02 -9.03596044e-01 8.18747818e-01
3.54706198e-01 1.20991409e+00 -9.01936948e-01 1.51410913e+00
6.17922172e-02 -3.97831857e-01 -4.26351905e-01 -9.33075547e-02
7.12284744e-01 -4.73190963e-01 3.33282173e-01 1.20124424e+00
4.69475389e-01 -4.08484548e-01 1.88380376e-01 1.07467020e+00
1.04271904e-01 2.83005416e-01 -1.45248950e+00 3.99673581e-01
6.65708899e-01 1.09452951e+00 -7.62960374e-01 -1.36466071e-01
-6.88476682e-01 1.10196662e+00 3.76832068e-01 6.04590595e-01
-4.71145958e-01 -1.04936624e+00 2.88808107e-01 -1.72698081e-01
1.26926437e-01 -9.75823998e-01 -4.55375433e-01 -1.10176945e+00
4.18345854e-02 -1.04235619e-01 -2.69969761e-01 -1.37754643e+00
-1.36191702e+00 6.20912492e-01 4.53080833e-02 -1.86559403e+00
-1.93886474e-01 -1.07135427e+00 -5.45596361e-01 4.21165735e-01
-1.61018455e+00 -1.62289166e+00 -7.03500211e-01 8.38741958e-01
6.84829354e-01 -1.40820846e-01 6.27660990e-01 6.60282269e-04
1.14794061e-01 6.15664899e-01 2.66813397e-01 6.20477200e-01
6.11090958e-01 -1.29062939e+00 5.08524418e-01 4.62777138e-01
3.65376294e-01 8.75225484e-01 5.51620662e-01 -3.40159655e-01
-1.29724360e+00 -8.91658247e-01 8.50070357e-01 -2.39517376e-01
7.91286528e-01 -4.55694795e-01 -7.09604859e-01 7.29172409e-01
2.24847108e-01 1.01216018e-01 2.55639195e-01 -3.60681862e-01
-6.56048656e-01 -1.85687318e-01 -9.00682271e-01 9.54357624e-01
1.47242117e+00 -8.19074988e-01 -6.54890776e-01 5.55591941e-01
7.97735155e-01 -4.04188395e-01 -8.21869493e-01 2.67227113e-01
5.01863599e-01 -1.09103930e+00 1.30176067e+00 -3.19460481e-01
2.87665993e-01 -3.56477618e-01 -1.49261981e-01 -1.22001982e+00
-2.57954985e-01 -3.55488151e-01 5.23702621e-01 9.11206067e-01
2.79852729e-02 -5.66956878e-01 7.11457849e-01 3.18390280e-02
-5.12391031e-01 -4.93848622e-01 -6.89743936e-01 -9.86700594e-01
1.07398987e-01 -3.49038303e-01 7.17707455e-01 9.54507709e-01
-5.27068712e-02 3.85417268e-02 -2.28482902e-01 1.37096375e-01
6.43341839e-01 1.32032365e-01 9.34543669e-01 -1.22635436e+00
5.13994992e-01 -7.14765728e-01 -1.09738588e+00 -1.69214344e+00
4.18902338e-01 -9.95651722e-01 -1.19811758e-01 -1.60773683e+00
-2.73479134e-01 -5.88601790e-02 3.33275706e-01 1.68982029e-01
6.37879074e-01 2.54679650e-01 5.36292315e-01 3.35510194e-01
-6.39780581e-01 7.23239243e-01 1.40299988e+00 -4.64715660e-01
6.37671500e-02 -2.68683076e-01 1.77794740e-01 9.82915461e-01
7.02201605e-01 1.93270355e-01 -2.60008574e-01 -8.09899211e-01
4.20392454e-01 -7.63031170e-02 7.11602747e-01 -1.24931860e+00
8.92799616e-01 4.45168912e-02 6.72172904e-01 -1.16736424e+00
5.05585968e-01 -1.10049236e+00 -4.01037604e-01 3.42300773e-01
-5.50395846e-01 5.81537604e-01 -5.08725159e-02 5.02974153e-01
-3.40862066e-01 -3.95793825e-01 4.56389040e-01 -1.35285810e-01
-1.03360081e+00 6.02120101e-01 -9.32286382e-02 -4.08273190e-01
9.88790572e-01 -9.88431871e-01 -7.25181162e-01 -6.11070156e-01
-5.07578015e-01 2.12949842e-01 7.70456493e-01 5.51140845e-01
1.18591726e+00 -1.68309152e+00 -2.97338545e-01 4.49406534e-01
7.01710820e-01 2.10381284e-01 1.03558719e-01 5.01705647e-01
-1.56396461e+00 8.37172389e-01 -6.04540050e-01 -9.12701309e-01
-9.12799299e-01 7.25661993e-01 5.56580842e-01 6.03758574e-01
-7.38035262e-01 6.58508062e-01 1.65341213e-01 -4.60837871e-01
3.79217148e-01 -6.22385800e-01 -2.67135590e-01 -1.11377016e-01
-5.81678860e-02 2.97645301e-01 -2.98746318e-01 -9.43070889e-01
-2.15963140e-01 1.55889094e+00 2.04828992e-01 3.25355977e-01
1.25066853e+00 -4.08372194e-01 -2.33797550e-01 6.03781462e-01
1.74074769e+00 8.48163068e-02 -9.54018176e-01 -4.08303231e-01
2.09780589e-01 -3.06956530e-01 -4.49507624e-01 -9.69157219e-02
-5.82026362e-01 1.06238258e+00 3.72181952e-01 1.96051702e-01
4.76979017e-01 -1.58994377e-01 7.49144316e-01 1.06046104e+00
6.74429834e-01 -9.05128062e-01 2.28866637e-01 8.27124894e-01
1.34632695e+00 -1.36490476e+00 -2.85680741e-02 -6.79513991e-01
-2.90779203e-01 2.14870691e+00 7.04494596e-01 -5.76051652e-01
5.06097734e-01 -7.96365142e-02 1.21357247e-01 -2.86146402e-01
1.00166919e-02 -1.59576327e-01 5.04694402e-01 1.05459726e+00
7.94236585e-02 -1.95125118e-01 3.76862735e-02 -3.10628861e-01
-6.96298718e-01 -4.07802373e-01 7.08614767e-01 7.01513529e-01
-5.99860907e-01 -7.71209300e-01 -4.48617369e-01 8.53545666e-02
4.64451462e-01 -8.86101276e-02 -3.71380717e-01 1.20439255e+00
8.91486555e-02 6.06167853e-01 5.21090150e-01 -2.55917490e-01
5.52823424e-01 -4.89450872e-01 5.56989849e-01 -3.18264455e-01
-2.04496950e-01 -3.63350898e-01 -2.37254605e-01 -6.56591475e-01
-6.19033933e-01 -2.67647266e-01 -1.32066953e+00 -2.82406390e-01
-1.26032650e-01 -1.97026819e-01 7.83759177e-01 6.85490668e-01
5.95284253e-02 1.20211266e-01 6.70845449e-01 -8.77851546e-01
-4.70108211e-01 -5.00169337e-01 -5.71587324e-01 5.64506531e-01
5.76633871e-01 -5.84611654e-01 -1.60628185e-01 -6.16590418e-02]
|
[8.160331726074219, -2.036993980407715]
|
3fbeceb6-4427-439a-a460-b2429ff00001
|
leveraging-uni-modal-self-supervised-learning
| null | null |
https://openreview.net/forum?id=sFZPi0Dy6XV
|
https://openreview.net/pdf?id=sFZPi0Dy6XV
|
Leveraging Uni-Modal Self-Supervised Learning for Multimodal Audio-visual Speech Recognition
|
Training Transformer-based models demands a large amount of data, while obtaining parallel aligned and labelled data in multimodality is rather cost-demanding, especially for audio-visual speech recognition (AVSR). Thus it makes a lot of sense to make use of unlabelled uni-modal data. On the other side, although the effectiveness of large-scale self-supervised learning is well established in both audio and visual modalities, how to integrate those pre-trained models into a multimodal scenario remains underexplored. In this work, we successfully leverage uni-modal self-supervised learning to promote the multimodal AVSR. In particular, we first train audio and visual encoders on a large-scale uni-modal dataset, then we integrate components of both encoders into a larger multimodal framework which learns to recognize paired audio-visual data into characters through a combination of CTC and seq2seq decoding. We show that both components inherited from uni-modal self-supervised learning cooperate well, resulting in that the multimodal framework yields competitive results through fine-tuning. Our model is experimentally validated on both word-level and sentence-level AVSR tasks. Especially, even without an external language model, our proposed model raises the state-of-the-art performances on the widely accepted Lip Reading Sentences 2 (LRS2) dataset by a large margin, with a relative improvement of 30%.
|
['Anonymous']
|
2021-11-16
| null | null | null |
acl-arr-november-2021-11
|
['audio-visual-speech-recognition']
|
['speech']
|
[ 5.54299235e-01 1.88958675e-01 -3.40193987e-01 -3.34255069e-01
-1.66701722e+00 -4.36263174e-01 5.55989206e-01 -1.88727319e-01
-4.49855953e-01 6.16081059e-01 4.49673921e-01 -2.77495503e-01
3.79411697e-01 -4.49323654e-01 -9.58441675e-01 -8.51739764e-01
4.35829163e-01 3.56668293e-01 1.30528390e-01 -2.21330360e-01
-1.42243892e-01 4.37791385e-02 -1.77454007e+00 5.70756018e-01
1.01140058e+00 1.16462195e+00 4.91243482e-01 7.14627683e-01
-1.98446110e-01 6.70197189e-01 -2.38496020e-01 -6.09678745e-01
-2.55629182e-01 -4.66452569e-01 -8.11768234e-01 1.38150111e-01
4.84407485e-01 -1.41145855e-01 -2.55945712e-01 8.33502352e-01
1.01372850e+00 -7.99781904e-02 6.41103923e-01 -1.17450988e+00
-6.05934918e-01 7.28112757e-01 -5.72635889e-01 -3.78943682e-01
3.99191052e-01 3.80296141e-01 1.34374034e+00 -9.43243444e-01
4.19489801e-01 1.22119510e+00 3.89473766e-01 7.57377923e-01
-1.19927311e+00 -5.70026398e-01 -7.50842616e-02 2.87403464e-01
-1.21381211e+00 -1.01621556e+00 8.19159508e-01 -1.24816604e-01
9.06969666e-01 1.91555128e-01 4.33399767e-01 1.64253247e+00
-2.04256013e-01 1.36671507e+00 1.18468904e+00 -5.06658852e-01
5.84656745e-03 3.10457587e-01 -3.87822807e-01 4.51805472e-01
-5.71439028e-01 6.24682009e-02 -9.99022782e-01 2.03969061e-01
2.13288724e-01 -3.75056982e-01 -3.70577574e-01 -3.78695607e-01
-1.34787774e+00 7.79533982e-01 1.99316591e-01 3.66125911e-01
-1.47288814e-01 -4.38716747e-02 6.56923473e-01 3.18924248e-01
3.44255954e-01 -3.10623869e-02 -3.88786107e-01 -4.08780336e-01
-1.01193476e+00 -4.20665056e-01 3.10671031e-01 8.17356348e-01
5.79845965e-01 2.32262954e-01 -2.57982820e-01 1.35075080e+00
4.31473553e-01 1.04210186e+00 5.85207522e-01 -5.05194187e-01
7.42560983e-01 3.09948832e-01 -5.47533453e-01 -4.55896437e-01
-2.45762333e-01 -3.82927388e-01 -1.13144553e+00 -5.08631095e-02
3.12891841e-01 3.49986069e-02 -7.66787171e-01 2.09505057e+00
2.82065868e-02 -6.05187789e-02 4.96780068e-01 7.54760385e-01
1.05313230e+00 7.64166236e-01 9.56572294e-02 -3.41970772e-01
1.34035492e+00 -1.13128912e+00 -7.01753676e-01 -1.41646072e-01
4.83954877e-01 -8.15327466e-01 1.42563367e+00 3.45079809e-01
-1.20432377e+00 -7.93265283e-01 -7.41952121e-01 -2.11824194e-01
-1.13462344e-01 5.19250214e-01 2.57648647e-01 5.55040181e-01
-1.15629959e+00 4.64636385e-02 -5.82078576e-01 -3.12248290e-01
4.29868728e-01 3.53888780e-01 -6.68615282e-01 1.73840416e-03
-1.21473610e+00 7.38351941e-01 2.11066842e-01 1.36518523e-01
-1.08475447e+00 -4.26472604e-01 -1.09568620e+00 -4.01763283e-02
4.08992946e-01 -5.72061718e-01 1.19397569e+00 -1.00864947e+00
-1.90385890e+00 9.22412872e-01 -5.75070798e-01 -3.87067795e-01
3.77929807e-01 2.66615152e-02 -4.81258482e-01 3.21300864e-01
-8.46416503e-02 1.15822351e+00 1.07919943e+00 -1.20341444e+00
-6.26070976e-01 -3.55127484e-01 -2.22839743e-01 3.21958184e-01
-6.51075840e-01 -7.79029205e-02 -5.36599398e-01 -4.05010730e-01
-2.07656533e-01 -7.73197114e-01 3.32312882e-01 -2.81084239e-01
-4.55503166e-01 -4.39065725e-01 5.77533424e-01 -6.27169609e-01
9.01713848e-01 -2.31673241e+00 4.68676955e-01 -2.26038054e-01
-1.03690535e-01 3.20621729e-01 -5.67809284e-01 5.05948305e-01
-3.32810581e-02 -1.21673279e-01 -2.69241005e-01 -7.91250587e-01
1.92444623e-01 1.25866234e-02 -4.54733312e-01 2.55699486e-01
4.37414110e-01 1.22459257e+00 -7.30579853e-01 -6.20862842e-01
2.39423156e-01 7.41651714e-01 -4.07435566e-01 4.22993183e-01
-2.26417974e-01 6.60842180e-01 -2.75808144e-02 8.02085757e-01
6.25705361e-01 -7.71219283e-02 8.65619183e-02 -3.05426061e-01
3.22664231e-02 3.66774261e-01 -7.64708042e-01 2.22550583e+00
-7.98523426e-01 6.47070110e-01 -5.29542156e-02 -1.14894485e+00
9.17053759e-01 5.87514579e-01 2.12626919e-01 -1.01527357e+00
-1.67599842e-02 4.15347904e-01 -1.54305220e-01 -6.84696019e-01
3.53638470e-01 -4.44677353e-01 -1.19797662e-01 2.30260536e-01
5.30390561e-01 -1.41496882e-01 -4.97478843e-02 1.81519851e-01
7.16797113e-01 1.43978208e-01 -2.31240969e-02 2.68200457e-01
9.67431903e-01 -3.35124999e-01 7.67122880e-02 3.89855474e-01
2.63444874e-02 7.73499727e-01 4.55244213e-01 3.98848593e-01
-8.03654194e-01 -1.13797820e+00 -1.56944558e-01 1.34577847e+00
-6.65964931e-02 -3.14680696e-01 -6.03149414e-01 -7.03150690e-01
-2.69488961e-01 6.24603808e-01 -4.31677818e-01 -1.61664844e-01
-1.65271044e-01 -4.20550883e-01 9.47318673e-01 5.76877415e-01
4.50500607e-01 -1.20291138e+00 -1.83316633e-01 -6.55417219e-02
-5.37734449e-01 -1.40572906e+00 -4.99517202e-01 2.24123836e-01
-5.49153626e-01 -8.17401648e-01 -1.27521873e+00 -9.43307996e-01
1.71115980e-01 1.69072479e-01 9.49263096e-01 -3.81156713e-01
2.38303859e-02 4.84504372e-01 -4.67504114e-01 -1.01527654e-01
-6.18619859e-01 4.27665830e-01 -9.71577764e-02 3.94246727e-01
1.27678245e-01 -5.16504407e-01 -2.21985027e-01 2.67806202e-01
-7.28831232e-01 1.99154377e-01 8.42800438e-01 1.18981123e+00
4.92346674e-01 -3.18537444e-01 9.55999136e-01 -2.56811827e-01
3.52610856e-01 -1.21969894e-01 -3.32505405e-01 4.19841647e-01
-2.13706210e-01 8.65160860e-03 5.16697764e-01 -5.86142361e-01
-1.07528555e+00 1.69262961e-01 -7.56526709e-01 -4.20402706e-01
-3.28356624e-01 3.47724259e-01 -6.20244682e-01 2.15708181e-01
2.44838849e-01 7.15583086e-01 2.93810666e-01 -5.13966322e-01
6.81050181e-01 1.19798207e+00 7.92468309e-01 -4.72847104e-01
6.67199194e-01 2.39764020e-01 -2.94893295e-01 -9.59401309e-01
-5.52782357e-01 -3.98787916e-01 -4.87345964e-01 -1.73691452e-01
1.05842626e+00 -1.21542847e+00 -9.24097955e-01 6.62998319e-01
-1.16516614e+00 -4.20052379e-01 -2.80584335e-01 5.17433763e-01
-7.82021105e-01 4.42328185e-01 -4.76988643e-01 -1.03039372e+00
-3.04169118e-01 -1.53445709e+00 1.59478521e+00 3.06273834e-03
1.12181544e-01 -8.44408989e-01 1.91575378e-01 8.78908873e-01
3.71706069e-01 -5.03983080e-01 7.26909637e-01 -6.54697001e-01
-4.67715710e-01 1.25319719e-01 -3.50444317e-01 5.80970883e-01
7.41429105e-02 -3.13516140e-01 -1.49787533e+00 -2.57422179e-01
-2.90484190e-01 -1.09892237e+00 1.06709492e+00 1.00731030e-01
9.44599748e-01 8.38723779e-02 -5.09669222e-02 4.03201163e-01
1.06101298e+00 -1.10751733e-01 8.17562401e-01 -7.89745152e-02
6.84193015e-01 7.88493395e-01 5.56794822e-01 1.06706403e-01
6.45525277e-01 9.74698961e-01 5.37280500e-01 -2.34279931e-01
-4.79655325e-01 -5.54340303e-01 7.62553811e-01 1.21400750e+00
1.33711681e-01 -3.15191269e-01 -7.24090755e-01 6.46441221e-01
-1.59410715e+00 -9.19545770e-01 8.81377980e-02 2.14019036e+00
1.19346583e+00 -2.68419951e-01 3.29230934e-01 2.46072397e-01
5.39929628e-01 2.91786343e-01 -4.45566177e-01 -1.82905048e-01
-5.98409653e-01 2.15103999e-01 5.65782562e-03 4.91337329e-01
-1.05485654e+00 1.12731910e+00 5.39767170e+00 1.41414595e+00
-1.44922435e+00 3.46208572e-01 5.98655164e-01 -8.88356660e-03
-4.88334328e-01 -2.54098654e-01 -8.37090254e-01 3.86566669e-01
1.14687979e+00 3.96019220e-01 3.52176517e-01 5.09331346e-01
7.95885250e-02 1.89855900e-02 -1.00677156e+00 1.30233693e+00
4.18370634e-01 -9.63471353e-01 3.16660404e-02 9.87808108e-02
4.77266610e-01 1.47252336e-01 3.15415442e-01 4.93817180e-01
-2.05660567e-01 -1.21921086e+00 7.94017732e-01 2.49473602e-01
1.26354027e+00 -6.73919082e-01 7.19880879e-01 3.61504763e-01
-1.23724449e+00 7.03405440e-02 -1.64434835e-01 4.99261528e-01
4.00380850e-01 3.49845171e-01 -7.12422132e-01 7.61954427e-01
4.89426583e-01 6.48854673e-01 -4.83752698e-01 7.34728098e-01
-3.09169292e-01 6.93963587e-01 -7.98400342e-02 1.19081110e-01
1.76004529e-01 1.19683631e-01 4.28547889e-01 1.24119616e+00
2.05714568e-01 -4.72139478e-01 -2.69467086e-01 5.91891825e-01
-1.67365238e-01 3.74170929e-01 -6.02670789e-01 -3.14171344e-01
1.07805692e-01 1.24478972e+00 -4.67582867e-02 -2.27846548e-01
-4.71822828e-01 1.11240697e+00 3.26404184e-01 2.99438000e-01
-8.16120207e-01 -1.69090539e-01 3.58169943e-01 -2.33447537e-01
4.59633082e-01 5.66388816e-02 -1.87013745e-01 -1.35303342e+00
7.89917037e-02 -1.23011982e+00 1.70296818e-01 -9.14544404e-01
-1.28177142e+00 8.73811126e-01 -4.46944654e-01 -1.35812378e+00
-5.60144961e-01 -5.34518540e-01 -2.44373024e-01 6.95139587e-01
-1.96060860e+00 -1.60168505e+00 -1.41692817e-01 8.54183078e-01
7.03004241e-01 -4.79246944e-01 9.14407313e-01 4.04353917e-01
-3.79949182e-01 1.04727829e+00 6.32305804e-04 1.61792338e-01
1.06803918e+00 -1.01528299e+00 -2.80662123e-02 5.24898231e-01
4.66673493e-01 2.66033947e-01 2.89113283e-01 -2.96698958e-01
-1.68937898e+00 -8.34376574e-01 1.02450192e+00 -1.72276229e-01
7.39632785e-01 -6.25553071e-01 -8.61290276e-01 2.51046449e-01
6.54635787e-01 -6.77137300e-02 7.53100574e-01 6.03783689e-02
-6.75464630e-01 -2.65252978e-01 -7.98539639e-01 4.94805098e-01
1.05590105e+00 -1.12000144e+00 -6.14606261e-01 2.13391930e-01
8.01409960e-01 -1.67216793e-01 -7.57304668e-01 5.56669414e-01
5.86641908e-01 -9.03512597e-01 9.89305437e-01 -4.78913575e-01
5.63407123e-01 -1.50271103e-01 -5.51404655e-01 -1.31108415e+00
4.63078618e-01 -6.48647308e-01 1.43281519e-01 1.56431544e+00
6.77122235e-01 -5.86924493e-01 4.26191211e-01 -1.21196307e-01
-2.89638340e-01 -7.01664209e-01 -1.22774458e+00 -7.42887616e-01
1.63331479e-01 -6.91043794e-01 2.71525979e-01 6.64587200e-01
3.19393396e-01 8.00173581e-01 -6.85285747e-01 -6.62066042e-02
4.99932826e-01 3.06007806e-02 9.77589369e-01 -9.15802658e-01
-4.83932495e-01 -4.09604311e-01 -2.26910338e-01 -1.44859028e+00
6.18753254e-01 -1.04245579e+00 2.88535655e-01 -1.26027346e+00
3.92210543e-01 -2.34863669e-01 -1.70672819e-01 6.67352915e-01
-1.04776077e-01 5.64542830e-01 3.29962224e-01 1.17371269e-02
-7.75510788e-01 1.16739857e+00 1.17195559e+00 -5.07037401e-01
-8.78622979e-02 -3.49214077e-02 -5.87969363e-01 3.74945283e-01
5.08891165e-01 -6.43598735e-02 -4.51235801e-01 -4.47234184e-01
-1.04513010e-02 3.00857365e-01 3.93976152e-01 -7.43875325e-01
2.87618905e-01 2.55225837e-01 -7.29613518e-03 -6.54965639e-01
7.53118038e-01 -7.21106887e-01 -3.75964910e-01 -8.05830397e-03
-5.93910933e-01 -3.59381884e-01 3.42662603e-01 5.40854812e-01
-6.63040042e-01 -4.94150072e-02 7.33873129e-01 3.15294623e-01
-4.59584832e-01 3.79095227e-02 -2.94496894e-01 2.23696098e-01
6.24810517e-01 1.61014665e-02 -3.22666407e-01 -6.45857871e-01
-6.20443106e-01 7.69430920e-02 2.63113201e-01 5.09751260e-01
6.71651304e-01 -1.39014745e+00 -7.72727132e-01 4.03602421e-01
3.97757053e-01 -2.75272340e-01 5.82051992e-01 1.26793492e+00
2.81489640e-01 6.96093142e-01 -1.43073006e-02 -1.05998814e+00
-1.38269794e+00 4.22739089e-01 5.06436117e-02 -8.17624032e-02
-3.67912889e-01 8.11937213e-01 2.44045421e-01 -6.23328328e-01
6.23689830e-01 2.02469267e-02 -2.09163606e-01 3.49967450e-01
5.05371988e-01 2.43982039e-02 6.51833937e-02 -1.06245804e+00
-3.71172160e-01 7.61236489e-01 2.46236563e-01 -5.02189279e-01
1.18212652e+00 -2.93093681e-01 3.47165838e-02 7.08082378e-01
1.53228390e+00 2.29246527e-01 -1.09147358e+00 -3.32556516e-01
-3.41400653e-01 -5.00564016e-02 -8.50981940e-03 -8.30323815e-01
-1.14943922e+00 1.60364866e+00 5.56520998e-01 -7.24491477e-02
1.36240792e+00 3.94070387e-01 1.01616299e+00 2.05484271e-01
1.94909438e-01 -9.56627250e-01 3.99894506e-01 3.88272345e-01
9.90142405e-01 -1.51830840e+00 -6.44984961e-01 -3.12652767e-01
-1.28775942e+00 9.67554092e-01 2.62378454e-01 4.90238667e-01
1.64064199e-01 2.72176117e-01 2.26964936e-01 2.99568564e-01
-7.87536263e-01 -6.06030583e-01 5.41733205e-01 7.16030598e-01
4.13902849e-01 -9.70559195e-02 8.09805766e-02 6.52603805e-01
-1.10486388e-01 -1.53086684e-03 -7.83515200e-02 3.54961276e-01
-2.29555771e-01 -1.35784137e+00 -2.36923829e-01 -8.32632035e-02
-2.56240934e-01 -4.28875834e-01 -2.39909992e-01 4.69410628e-01
-1.46924406e-01 1.19479859e+00 -7.19555020e-02 -5.48995614e-01
1.49740353e-01 3.80159616e-01 4.25330937e-01 -2.46918768e-01
-4.14841563e-01 5.01713276e-01 1.57911122e-01 -4.63276058e-01
-5.98101437e-01 -6.34135783e-01 -1.08112824e+00 1.26762033e-01
-3.05458993e-01 -1.37864336e-01 7.41479754e-01 1.03889132e+00
3.41622651e-01 3.66624027e-01 6.69080257e-01 -7.73396969e-01
-6.44952238e-01 -1.13696051e+00 -3.44279200e-01 2.49536172e-01
4.22236890e-01 -5.14178514e-01 -4.45859224e-01 -7.18331486e-02]
|
[14.206060409545898, 5.145693302154541]
|
99595eb0-7f50-4509-987d-710938a5e9cf
|
a-unified-semantic-embedding-relating
|
1411.5879
| null |
http://arxiv.org/abs/1411.5879v2
|
http://arxiv.org/pdf/1411.5879v2.pdf
|
A Unified Semantic Embedding: Relating Taxonomies and Attributes
|
We propose a method that learns a discriminative yet semantic space for
object categorization, where we also embed auxiliary semantic entities such as
supercategories and attributes. Contrary to prior work which only utilized them
as side information, we explicitly embed the semantic entities into the same
space where we embed categories, which enables us to represent a category as
their linear combination. By exploiting such a unified model for semantics, we
enforce each category to be represented by a supercategory + sparse combination
of attributes, with an additional exclusive regularization to learn
discriminative composition.
|
['Sung Ju Hwang', 'Leonid Sigal']
|
2014-11-18
|
a-unified-semantic-embedding-relating-1
|
http://papers.nips.cc/paper/5289-a-unified-semantic-embedding-relating-taxonomies-and-attributes
|
http://papers.nips.cc/paper/5289-a-unified-semantic-embedding-relating-taxonomies-and-attributes.pdf
|
neurips-2014-12
|
['object-categorization']
|
['computer-vision']
|
[ 8.08103681e-02 4.60670531e-01 -4.44358677e-01 -8.68307948e-01
-9.65800211e-02 -8.58354390e-01 7.83021808e-01 1.26813084e-01
-2.46716097e-01 3.53260577e-01 6.95381045e-01 2.10514084e-01
4.95017506e-02 -9.11860585e-01 -7.15871871e-01 -6.02382898e-01
2.06652984e-01 2.61212498e-01 -1.28033757e-01 5.27231097e-02
-5.72851375e-02 -2.94376370e-02 -1.74175394e+00 3.33111316e-01
7.71753550e-01 1.14071381e+00 -5.16707748e-02 -4.71416786e-02
-3.27046812e-01 9.10764575e-01 -1.93073943e-01 -5.37372708e-01
2.72485793e-01 -1.11215748e-01 -8.22119892e-01 2.89584011e-01
5.93474507e-01 -2.47375116e-01 -6.93081379e-01 1.05852973e+00
-8.79640430e-02 2.55778074e-01 1.07025552e+00 -1.44047666e+00
-1.19093192e+00 6.98503017e-01 -1.70558497e-01 -3.98995697e-01
3.63585621e-01 -2.12308675e-01 1.75057006e+00 -9.85763073e-01
6.10141039e-01 1.37087119e+00 5.78942418e-01 6.04465365e-01
-1.73916972e+00 -5.25927484e-01 6.43284261e-01 3.80018651e-01
-1.53109050e+00 -1.40361354e-01 9.43246067e-01 -5.88283837e-01
7.26689637e-01 1.00322068e-01 6.90978765e-01 1.22605956e+00
-3.31573486e-01 9.86540973e-01 6.95594668e-01 -2.28009582e-01
4.57145333e-01 4.57744926e-01 6.26921773e-01 5.04721045e-01
2.57678181e-01 -2.66325146e-01 -2.62012601e-01 -1.61517769e-01
4.37931210e-01 5.74716330e-01 1.82155430e-01 -1.04219568e+00
-1.13538277e+00 1.36500263e+00 9.97308254e-01 3.20569992e-01
-2.37111151e-01 4.40854192e-01 4.69512284e-01 1.07333869e-01
4.65642810e-01 3.85580778e-01 -5.11735678e-01 4.98765588e-01
-4.00019020e-01 1.65588617e-01 7.76050091e-01 1.21771657e+00
1.39795983e+00 8.98343232e-03 4.10954002e-03 1.03390491e+00
4.79787618e-01 4.03800607e-01 4.16902304e-01 -1.03207302e+00
1.09659947e-01 1.08170724e+00 -1.39121428e-01 -9.58026946e-01
-1.41753659e-01 -3.44020516e-01 -7.15048134e-01 -3.44092846e-01
-1.12417728e-01 2.69397944e-01 -1.18770313e+00 2.15940166e+00
2.35225007e-01 3.92028868e-01 8.09382349e-02 9.60834265e-01
9.20496166e-01 3.79449189e-01 5.47674239e-01 5.80792725e-01
1.45250106e+00 -9.12531734e-01 -5.63261390e-01 -3.38027298e-01
6.56835079e-01 -1.74899697e-01 1.05365121e+00 -1.06974117e-01
-5.11372328e-01 -4.26177174e-01 -8.27130675e-01 -5.50925434e-01
-7.73847342e-01 1.56701252e-01 1.31379783e+00 3.08521420e-01
-9.59163725e-01 2.45976478e-01 -6.88002944e-01 -3.02942455e-01
5.45327663e-01 2.32016072e-01 -8.46045136e-01 -1.18102022e-01
-1.36278343e+00 6.94903553e-01 5.49033642e-01 -4.04665530e-01
-9.23072815e-01 -8.64167213e-01 -1.54614937e+00 2.56423712e-01
2.30696365e-01 -1.09485638e+00 8.47790360e-01 -1.14537656e+00
-1.08664072e+00 1.26147783e+00 -3.05663854e-01 -3.01498830e-01
-2.11382166e-01 -5.64831197e-02 -2.79247880e-01 1.06874898e-01
4.81100112e-01 7.81975925e-01 8.64253759e-01 -1.52857125e+00
-5.38026214e-01 -5.98541558e-01 6.50309145e-01 2.21461311e-01
-8.51498425e-01 -1.95166171e-01 -2.77737141e-01 -9.53479767e-01
5.10386169e-01 -8.95696044e-01 -3.41751099e-01 9.84397009e-02
-7.07544267e-01 -4.88166451e-01 1.05857778e+00 -5.40990889e-01
1.20934463e+00 -2.36200690e+00 7.31206954e-01 2.92399824e-01
6.55432582e-01 -6.60962403e-01 -2.34241679e-01 1.15562260e-01
-5.13516925e-02 3.81625853e-02 -5.64201951e-01 -6.95629597e-01
5.91614604e-01 7.86057830e-01 -6.09065056e-01 3.99713188e-01
2.98841447e-01 1.00013959e+00 -1.02942264e+00 -2.15518862e-01
7.22819939e-02 6.55513108e-01 -1.11434233e+00 2.76528835e-01
-1.79977909e-01 5.15398895e-03 -7.76964724e-01 7.24991441e-01
6.61275387e-01 -3.41084361e-01 2.97323793e-01 -3.52529019e-01
2.82600880e-01 4.32872236e-01 -1.11747956e+00 1.93271470e+00
-6.92263603e-01 1.55579776e-01 1.06814943e-01 -1.74926317e+00
6.88708901e-01 2.10053846e-01 5.23358047e-01 -1.43475413e-01
-1.40756443e-01 1.25704244e-01 -4.94857132e-01 -1.17369089e-02
3.93889874e-01 -3.88871968e-01 -6.20001614e-01 3.14397424e-01
7.28732467e-01 -1.06750101e-01 -1.32210612e-01 6.11299753e-01
8.86378884e-01 1.11284867e-01 1.26213804e-01 -2.84211963e-01
3.97017956e-01 -1.32372409e-01 5.76120973e-01 5.89399636e-01
2.59852499e-01 3.59769344e-01 5.36359131e-01 -3.79162759e-01
-1.04288805e+00 -1.44416893e+00 -5.01482606e-01 1.28314328e+00
3.94159168e-01 -5.08858919e-01 -4.58918154e-01 -1.14562988e+00
7.28745341e-01 6.64797366e-01 -9.10287380e-01 -4.44134802e-01
-1.54153481e-01 -3.50027978e-01 1.53470010e-01 9.20839489e-01
8.47142711e-02 -6.09971941e-01 7.53035098e-02 9.25862119e-02
1.07934915e-01 -1.03291571e+00 -4.49060798e-01 6.22673869e-01
-6.44069433e-01 -9.94928598e-01 -3.56574774e-01 -7.53893852e-01
8.15015256e-01 1.16453245e-01 1.22966301e+00 -9.02866498e-02
-3.26103330e-01 8.26369524e-01 -3.32211614e-01 3.05446256e-02
-2.40244363e-02 -1.98973387e-01 2.52610743e-01 2.75779843e-01
7.76610613e-01 -8.15759361e-01 -7.57450879e-01 4.90945764e-02
-9.16554570e-01 -9.35296640e-02 2.59771794e-01 1.10269225e+00
5.53321064e-01 -2.94418097e-01 6.58619404e-01 -1.21397281e+00
-1.49093166e-01 -1.14779377e+00 -3.21468413e-01 3.55297662e-02
-4.37605441e-01 2.34608456e-01 7.52643704e-01 -3.88736159e-01
-9.09841418e-01 1.38370171e-01 8.79433751e-02 -6.12348020e-01
-3.34268272e-01 4.96374458e-01 -6.16779268e-01 8.33484456e-02
3.05715024e-01 1.62865296e-01 -2.22546067e-02 -6.96987212e-01
1.00973010e+00 6.73822999e-01 3.29618216e-01 -7.37020373e-01
9.42519844e-01 6.79563046e-01 -1.60252720e-01 -4.22482193e-01
-1.24770892e+00 -8.18096936e-01 -8.16615641e-01 4.33753341e-01
1.22929180e+00 -1.35609651e+00 -3.62155020e-01 -2.96544451e-02
-1.00329494e+00 -7.81261362e-03 -5.29309571e-01 5.25339782e-01
-6.18645966e-01 3.64855707e-01 -5.30253053e-01 -1.56132758e-01
3.77125144e-01 -7.80887544e-01 1.40520394e+00 -1.81593806e-01
-2.42078900e-01 -1.38992822e+00 -2.41190940e-01 2.02406541e-01
3.67902339e-01 -1.28999697e-02 1.05204284e+00 -9.97766197e-01
-4.95734811e-01 -1.00135036e-01 -4.15075600e-01 5.70711255e-01
2.92962134e-01 -4.73970234e-01 -9.33229029e-01 -1.85082436e-01
-7.31893331e-02 -4.02316749e-01 1.29726768e+00 3.18843080e-03
1.49175799e+00 -5.53172648e-01 -5.65725267e-01 9.89853263e-01
1.51281548e+00 -2.99928457e-01 1.75401330e-01 1.13572754e-01
1.13399863e+00 5.60983360e-01 1.62246943e-01 5.13170719e-01
7.32305050e-01 7.11386204e-01 3.85123998e-01 -2.05490559e-01
-1.26172509e-02 -5.21869004e-01 -7.97323603e-03 6.08110845e-01
1.57433823e-01 2.70674855e-01 -6.85554504e-01 6.19587183e-01
-1.77354836e+00 -9.64576721e-01 3.92555356e-01 1.83344018e+00
8.71822000e-01 -5.01263857e-01 -3.59721221e-02 -1.77389607e-01
5.80373168e-01 2.94881493e-01 -6.86938286e-01 1.06230658e-03
-2.41669580e-01 1.79944098e-01 2.61409014e-01 2.84277260e-01
-1.48827088e+00 1.14337385e+00 6.91354036e+00 5.57871699e-01
-1.00613964e+00 3.04279715e-01 2.18761474e-01 1.65288821e-01
-1.10241771e+00 3.52586925e-01 -6.74426019e-01 6.11490190e-01
4.57015634e-01 -3.37951809e-01 2.81046093e-01 1.24883330e+00
-4.89824414e-01 4.74287361e-01 -1.47730744e+00 6.57238245e-01
1.49287939e-01 -1.02138770e+00 4.11720037e-01 1.08464137e-01
7.44392514e-01 -2.91483790e-01 1.32675946e-01 6.96862578e-01
8.10007155e-01 -9.44214702e-01 9.17350054e-01 3.25677991e-01
8.57546687e-01 -3.24478120e-01 4.05405432e-01 -1.51643723e-01
-1.26174605e+00 -5.46399117e-01 -7.32608199e-01 2.01388374e-02
-9.38524604e-02 4.56615239e-01 -4.75836486e-01 5.97092092e-01
6.12839580e-01 1.41382647e+00 -4.00743693e-01 4.32809353e-01
-4.53001201e-01 6.42944336e-01 -2.38939747e-01 4.33736593e-01
2.93683380e-01 -4.72808719e-01 3.21755826e-01 1.30779743e+00
1.90268531e-01 2.19730839e-01 5.52188098e-01 1.16794384e+00
-4.11890417e-01 2.54924106e-03 -8.31695318e-01 -6.96420595e-02
5.91165066e-01 1.15141904e+00 -1.84322953e-01 -6.49120986e-01
-1.01214409e+00 1.08520973e+00 5.59384882e-01 6.72678113e-01
-5.95666528e-01 -2.07909331e-01 1.17740381e+00 -5.45341484e-02
3.49680662e-01 -8.09117779e-02 -4.20393854e-01 -1.72796834e+00
-1.44726306e-01 -3.64377141e-01 7.71867275e-01 -4.66554791e-01
-1.71913064e+00 1.66423857e-01 1.21109493e-01 -1.24389327e+00
-8.72636214e-02 -7.91974723e-01 -3.50538731e-01 8.35596025e-01
-1.52092314e+00 -1.66058469e+00 -8.71648341e-02 7.33332098e-01
2.75105387e-01 -5.41265070e-01 1.01870680e+00 1.87665075e-01
-2.33747944e-01 6.38930440e-01 6.60893321e-02 7.27256015e-02
5.76031804e-01 -1.67750382e+00 -2.12403730e-01 4.10263628e-01
2.46644795e-01 1.09949517e+00 4.67337906e-01 -2.78828651e-01
-1.39876461e+00 -1.27964115e+00 8.61208916e-01 -7.94190943e-01
1.11780179e+00 -7.98692644e-01 -9.08364594e-01 1.17480803e+00
1.96947642e-02 4.51147258e-01 1.07628489e+00 6.17068768e-01
-1.23858666e+00 5.17766457e-04 -1.03567541e+00 3.07836115e-01
1.44964552e+00 -1.11615908e+00 -1.01565182e+00 3.27814251e-01
1.22531831e+00 2.09711835e-01 -9.57261920e-01 3.86723846e-01
2.95620203e-01 -4.29416239e-01 1.37596500e+00 -1.00988984e+00
5.43443799e-01 -3.90878350e-01 -7.06890523e-01 -1.40696764e+00
-5.75812817e-01 4.08506155e-01 -4.31437552e-01 1.05624115e+00
2.21737236e-01 -7.73397148e-01 6.98952317e-01 6.20489180e-01
-3.36923808e-01 -5.65456629e-01 -8.24840069e-01 -7.52073169e-01
1.48468509e-01 -7.32612386e-02 7.31395483e-01 1.35894549e+00
3.49057429e-02 3.54492754e-01 -1.67028934e-01 1.23526610e-01
8.00224721e-01 5.45393884e-01 2.94885933e-01 -1.33412838e+00
-2.24083409e-01 -4.43185985e-01 -1.06202066e+00 -9.83959854e-01
1.09727561e+00 -1.46183169e+00 -1.63979337e-01 -1.45857978e+00
7.18235612e-01 -8.27003121e-01 -7.69613206e-01 9.59730744e-01
-3.28957021e-01 4.86946523e-01 1.20608583e-01 1.54448763e-01
-7.54745245e-01 8.80835891e-01 7.55126715e-01 -4.38847154e-01
2.52823293e-01 -2.96507627e-01 -1.27249825e+00 6.07455254e-01
1.30000278e-01 -3.17240626e-01 -4.02363330e-01 -4.99180555e-01
9.27470252e-02 -4.49833453e-01 6.60520494e-01 -5.26941240e-01
7.82296583e-02 -1.76242933e-01 2.80141920e-01 -8.51999521e-02
4.40586716e-01 -1.09331250e+00 6.98912293e-02 4.13136631e-02
-4.10880744e-01 -7.51516402e-01 -1.41802803e-01 8.38936627e-01
-8.15015256e-01 -1.22194469e-01 4.49959874e-01 -2.61276122e-02
-1.31996703e+00 5.29141426e-01 7.17489347e-02 -2.17818096e-01
1.01556396e+00 -2.22525775e-01 -2.31501963e-02 -1.58782855e-01
-1.26267087e+00 4.22512680e-01 7.31020451e-01 5.31776726e-01
4.34404194e-01 -1.71967649e+00 -3.59432787e-01 2.46536329e-01
5.17076194e-01 -2.30176985e-01 3.85271966e-01 2.93082327e-01
3.66780043e-01 4.08282697e-01 -1.43733010e-01 -6.37704968e-01
-6.95324004e-01 1.09638655e+00 1.27753485e-02 1.09838754e-01
-6.70807302e-01 8.91985774e-01 1.10482454e+00 -9.69121397e-01
2.19618946e-01 2.19239041e-01 -3.65994513e-01 2.76639551e-01
5.70393763e-02 -1.21532947e-01 -2.41398349e-01 -7.12629378e-01
-5.91201425e-01 6.63847387e-01 1.51123643e-01 2.55729645e-01
1.43724072e+00 -1.85780793e-01 -3.76032352e-01 5.38878679e-01
1.54133809e+00 3.86265991e-03 -1.16693175e+00 -5.75651407e-01
2.20309854e-01 -5.66929460e-01 -1.21188886e-01 -5.82554817e-01
-8.85487974e-01 8.01123917e-01 6.36151731e-02 -1.11282971e-02
1.04968250e+00 6.57625258e-01 3.08213264e-01 3.19211125e-01
4.49895382e-01 -8.92395854e-01 2.01362506e-01 6.67950511e-01
9.11684513e-01 -1.31343651e+00 -2.17227727e-01 -7.69407570e-01
-8.28474641e-01 7.63639271e-01 5.82904518e-01 -3.89480084e-01
9.19378877e-01 -3.01539022e-02 -2.21378088e-01 -1.18626885e-01
-7.72357285e-01 -4.18167174e-01 5.49372017e-01 5.89221895e-01
4.71271336e-01 3.72843087e-01 -2.93028861e-01 1.02441204e+00
-4.91441637e-02 -3.59720975e-01 1.07628733e-01 6.73324883e-01
-3.13460141e-01 -1.21194029e+00 1.19494565e-01 6.49772644e-01
-1.43445507e-01 -6.91626295e-02 -1.27218276e-01 4.56654459e-01
2.09706843e-01 4.92652833e-01 4.17028755e-01 -3.75533402e-01
3.02954763e-01 2.68562257e-01 1.71836048e-01 -1.24679995e+00
-1.71769992e-01 -4.68684167e-01 -1.88554123e-01 -7.28440464e-01
-2.22583786e-01 -5.37835240e-01 -1.30122280e+00 1.44800767e-01
1.76193751e-02 3.07240695e-01 5.93539536e-01 7.53847659e-01
3.94889116e-01 2.66926378e-01 8.68972421e-01 -6.64819658e-01
-7.03969419e-01 -7.29678631e-01 -9.78790700e-01 1.19390619e+00
3.68330359e-01 -1.23692167e+00 -8.64212751e-01 2.57136106e-01]
|
[10.084138870239258, 2.2501728534698486]
|
b1170f1e-7cd5-48c0-aba4-5832c102788b
|
instance-embedding-transfer-to-unsupervised
|
1801.00908
| null |
http://arxiv.org/abs/1801.00908v2
|
http://arxiv.org/pdf/1801.00908v2.pdf
|
Instance Embedding Transfer to Unsupervised Video Object Segmentation
|
We propose a method for unsupervised video object segmentation by
transferring the knowledge encapsulated in image-based instance embedding
networks. The instance embedding network produces an embedding vector for each
pixel that enables identifying all pixels belonging to the same object. Though
trained on static images, the instance embeddings are stable over consecutive
video frames, which allows us to link objects together over time. Thus, we
adapt the instance networks trained on static images to video object
segmentation and incorporate the embeddings with objectness and optical flow
features, without model retraining or online fine-tuning. The proposed method
outperforms state-of-the-art unsupervised segmentation methods in the DAVIS
dataset and the FBMS dataset.
|
['C. -C. Jay Kuo', 'Alireza Fathi', 'Siyang Li', 'Qin Huang', 'Bryan Seybold', 'Alexey Vorobyov']
|
2018-01-03
|
instance-embedding-transfer-to-unsupervised-1
|
http://openaccess.thecvf.com/content_cvpr_2018/html/Li_Instance_Embedding_Transfer_CVPR_2018_paper.html
|
http://openaccess.thecvf.com/content_cvpr_2018/papers/Li_Instance_Embedding_Transfer_CVPR_2018_paper.pdf
|
cvpr-2018-6
|
['unsupervised-video-object-segmentation']
|
['computer-vision']
|
[ 1.47284836e-01 1.92019522e-01 -5.94182432e-01 -3.45658481e-01
-1.74539238e-01 -5.70728362e-01 3.56782049e-01 -8.98893643e-03
-6.76673591e-01 3.79374802e-01 -2.06711069e-01 2.07968429e-01
-5.36781885e-02 -9.31335449e-01 -1.02756357e+00 -6.68593824e-01
-3.09474468e-01 3.73296052e-01 7.72929192e-01 4.16817516e-01
1.20821960e-01 5.68681896e-01 -1.41143823e+00 1.55257173e-02
6.51813686e-01 1.16645598e+00 3.05662155e-01 8.31775963e-01
-4.33136791e-01 1.02492583e+00 -4.35188532e-01 -2.42521867e-01
4.32902038e-01 -1.68632314e-01 -1.05482292e+00 6.83336973e-01
8.37400258e-01 -7.63692141e-01 -1.00773013e+00 1.02721155e+00
-1.20608211e-01 4.91742343e-01 6.62159622e-01 -1.42543864e+00
-1.02025449e+00 5.09283423e-01 -2.16395840e-01 4.89064902e-01
-9.55343321e-02 2.15452939e-01 1.02350998e+00 -6.87216818e-01
9.74780679e-01 1.09687448e+00 4.38025057e-01 5.98924935e-01
-1.30587137e+00 -1.67684510e-01 5.15055716e-01 5.39123833e-01
-1.10566938e+00 -3.20533365e-01 7.90782928e-01 -7.36879528e-01
7.05009997e-01 -1.59191251e-01 1.03978992e+00 7.54442573e-01
-2.88487703e-01 1.30410564e+00 4.99170005e-01 -6.99957134e-03
2.44373918e-01 9.90522057e-02 4.24304336e-01 8.24332654e-01
2.50835955e-01 6.28407151e-02 -2.42876127e-01 4.06204522e-01
1.28246498e+00 3.50054055e-01 -2.35444129e-01 -7.03403473e-01
-1.41126609e+00 7.69985437e-01 6.57231629e-01 1.83327109e-01
-3.17339838e-01 5.28455317e-01 5.15275300e-01 9.15458649e-02
3.07000190e-01 3.11120838e-01 -5.60412705e-01 -9.06790718e-02
-1.08744967e+00 -5.68542928e-02 7.52646208e-01 1.25204444e+00
1.26643038e+00 1.02359861e-01 -3.51207256e-01 5.89749336e-01
2.14884892e-01 3.11422050e-01 4.99061704e-01 -1.50983143e+00
1.34406000e-01 7.48302281e-01 9.16167647e-02 -1.08221161e+00
-9.79462862e-02 -5.20336740e-02 -4.35146391e-01 -3.47797908e-02
3.88934016e-01 7.17127100e-02 -1.18199205e+00 1.47402942e+00
3.47811580e-01 8.32518876e-01 1.47371665e-01 1.04506409e+00
1.05593872e+00 7.42340863e-01 -3.66264731e-02 -1.06267989e-01
9.50649261e-01 -1.62120914e+00 -8.20560515e-01 -7.89794624e-02
3.76299918e-01 -2.57100075e-01 7.15370715e-01 2.27909163e-01
-1.00531626e+00 -1.01790416e+00 -1.01750505e+00 -1.57981351e-01
-5.01682818e-01 -7.45598227e-02 6.97252989e-01 4.49537903e-01
-9.92667079e-01 7.88163900e-01 -1.01624179e+00 -2.80233324e-01
7.83535004e-01 5.34881711e-01 -3.87838602e-01 6.33690804e-02
-9.87617552e-01 3.45080614e-01 7.60950208e-01 2.63399571e-01
-1.29687822e+00 -5.67447186e-01 -1.06653547e+00 -1.58488993e-02
4.80758399e-01 -5.34040391e-01 8.92824650e-01 -1.40204442e+00
-1.61538398e+00 6.49963081e-01 -1.28405258e-01 -7.40010619e-01
5.48744380e-01 -3.19068223e-01 -9.75795016e-02 8.78033400e-01
1.89732742e-02 1.26400316e+00 1.21895206e+00 -1.45159137e+00
-5.85048795e-01 -4.24132645e-02 4.60299492e-01 1.46555004e-03
-5.16487896e-01 -3.46595883e-01 -1.01180768e+00 -5.77123582e-01
6.68040290e-02 -7.96620190e-01 -2.28943393e-01 1.36355579e-01
-4.11066085e-01 -1.00732818e-01 1.31376731e+00 -6.12155199e-01
1.19769800e+00 -2.12566590e+00 5.43941975e-01 -6.99715018e-02
1.94789261e-01 3.45960915e-01 -4.02668387e-01 -1.68902636e-01
1.10217191e-01 2.85175592e-01 -3.48916352e-01 -2.14013785e-01
-1.69464558e-01 7.55414665e-01 -1.49925062e-02 4.84292805e-01
3.77149910e-01 1.17324948e+00 -1.13336551e+00 -9.24530804e-01
5.50777555e-01 3.67575616e-01 -7.37799108e-01 3.85208756e-01
-4.89233822e-01 5.27180076e-01 -4.88207996e-01 5.40957749e-01
5.67846477e-01 -3.73734325e-01 4.40679230e-02 -3.51216644e-01
6.99408650e-02 -2.80558318e-01 -1.23325622e+00 1.83648205e+00
3.16898823e-02 9.19486344e-01 -2.37230450e-01 -1.43451869e+00
5.67295730e-01 2.39604861e-01 9.66041744e-01 -2.99425334e-01
1.48869202e-01 -1.72710940e-01 -2.17801288e-01 -9.35579121e-01
5.70169330e-01 4.55232650e-01 2.46263206e-01 3.17091167e-01
5.49103498e-01 -1.24151587e-01 6.95079267e-01 1.94821686e-01
7.65286088e-01 3.60654682e-01 -2.71903008e-01 5.37100770e-02
6.44677043e-01 -3.27362232e-02 6.90972447e-01 6.82091177e-01
-5.52985489e-01 6.11491263e-01 3.71457487e-01 -6.37350082e-01
-9.44851160e-01 -1.22869813e+00 -2.42243886e-01 9.74849045e-01
6.94429755e-01 -2.57478446e-01 -9.47137594e-01 -1.02305472e+00
1.27342254e-01 2.10646223e-02 -8.16759348e-01 1.67335961e-02
-6.38183415e-01 -1.61079094e-01 1.73013955e-01 8.38708818e-01
6.34914458e-01 -1.23575449e+00 -5.99458873e-01 3.48949283e-01
-1.95246473e-01 -1.57353616e+00 -6.18489087e-01 2.93406807e-02
-1.31757247e+00 -1.32575309e+00 -7.01943815e-01 -1.17758751e+00
8.78126681e-01 1.45726755e-01 9.23906922e-01 1.92281514e-01
-3.98769617e-01 8.56919050e-01 -3.50136071e-01 3.71931672e-01
-1.77559823e-01 1.19898714e-01 -1.72312111e-02 2.89182872e-01
2.27771640e-01 -2.52033085e-01 -7.90762484e-01 3.87020886e-01
-1.24899018e+00 -2.09416613e-01 2.84251034e-01 7.19530582e-01
1.03906429e+00 -5.91842793e-02 2.71802127e-01 -9.11510587e-01
-5.22868447e-02 -1.46925732e-01 -5.95136285e-01 2.48253703e-01
-1.81352407e-01 -5.35752922e-02 4.99941021e-01 -7.65442789e-01
-6.30346239e-01 2.78383344e-01 3.60528052e-01 -9.81114924e-01
-2.52076864e-01 1.93484738e-01 1.89620815e-02 -1.65801704e-01
2.67463952e-01 2.67037868e-01 8.06874633e-02 -2.38163918e-01
7.74401486e-01 4.45586532e-01 6.53560519e-01 -4.80290741e-01
7.60756254e-01 6.38928294e-01 -3.77118528e-01 -1.04198158e+00
-8.45963240e-01 -6.34057999e-01 -1.26176620e+00 -4.24784094e-01
1.40433145e+00 -7.93900669e-01 -4.82534528e-01 5.34768105e-01
-1.20872223e+00 -6.58398986e-01 -7.15078712e-01 5.20709336e-01
-9.37325418e-01 5.03632486e-01 -9.27855968e-01 -3.87434185e-01
9.74892527e-02 -1.38382077e+00 9.27470982e-01 5.09012759e-01
1.20862886e-01 -1.21912336e+00 -9.65770930e-02 2.77149916e-01
-6.40389919e-02 7.58592412e-02 4.67932075e-01 -4.60606009e-01
-1.18873858e+00 -7.50392079e-02 -3.61158550e-01 8.89196038e-01
3.60464305e-01 5.41523874e-01 -7.22393692e-01 -2.21931294e-01
-3.08198363e-01 -2.14352578e-01 1.06224656e+00 7.55770326e-01
1.59270871e+00 -3.97969127e-01 -4.21030432e-01 7.87409604e-01
1.42189312e+00 1.75199926e-01 7.64276505e-01 4.56927985e-01
1.11906326e+00 4.58018601e-01 3.82905155e-01 2.12321747e-02
2.15016723e-01 3.72215748e-01 5.69852054e-01 -1.25170141e-01
-7.41778910e-02 -1.13312744e-01 4.26704377e-01 9.15106952e-01
-1.24741606e-01 -6.62331879e-02 -5.05701840e-01 9.03798521e-01
-2.02823186e+00 -1.03513324e+00 1.56070337e-01 1.82570326e+00
6.08061850e-01 2.15880826e-01 1.71205759e-01 -8.89048204e-02
8.92193258e-01 4.03515935e-01 -7.27341294e-01 -2.42475808e-01
7.40132630e-02 7.72948265e-02 6.42019331e-01 4.76578146e-01
-1.48695016e+00 1.39922428e+00 7.09825611e+00 3.82079542e-01
-1.05315626e+00 1.54944643e-01 6.57579005e-01 1.32786512e-01
-8.62751603e-02 2.17287871e-03 -6.07343614e-01 4.14983362e-01
6.13790512e-01 2.00225599e-02 4.71075177e-01 9.14740801e-01
8.97500813e-02 -2.64176037e-02 -1.37468112e+00 6.75891578e-01
1.07882880e-01 -1.51264453e+00 2.93775618e-01 -7.00363368e-02
1.07827044e+00 -7.18989521e-02 4.26163189e-02 1.65600359e-01
-3.58446017e-02 -8.06423604e-01 6.73795164e-01 6.02108836e-01
4.71458733e-01 -4.28676277e-01 5.25287747e-01 -2.80798320e-02
-1.27355456e+00 -9.93195698e-02 -5.20171344e-01 3.10937345e-01
1.86563119e-01 1.94531977e-01 -4.82409358e-01 2.61690140e-01
8.12350988e-01 1.34323788e+00 -5.60863197e-01 1.07523227e+00
-2.58738816e-01 5.40198326e-01 -1.29430622e-01 3.37178975e-01
4.99020904e-01 -5.68244398e-01 5.35788953e-01 1.08170927e+00
-1.54073074e-01 -4.44203429e-02 4.96224493e-01 9.79922414e-01
-2.36038193e-01 -1.13352537e-01 -4.08707678e-01 -3.84096086e-01
1.96683004e-01 1.24600434e+00 -1.00352740e+00 -7.03545392e-01
-5.00831425e-01 1.21015358e+00 1.63697720e-01 8.57673585e-01
-1.00050569e+00 -2.56416529e-01 7.63752103e-01 -1.20642640e-01
9.34003234e-01 -4.32246566e-01 1.09259114e-01 -1.18466556e+00
-2.18909681e-01 -3.46281141e-01 2.47557566e-01 -5.39872706e-01
-1.16116345e+00 3.86268586e-01 -2.29070336e-02 -1.25488210e+00
-2.16731932e-02 -9.54960644e-01 -4.96219605e-01 7.15379976e-03
-1.64542997e+00 -8.68263185e-01 -3.63097280e-01 6.33184195e-01
7.52103984e-01 5.95738403e-02 3.54879797e-01 2.25980878e-01
-7.23989904e-01 3.89932722e-01 1.48364782e-01 7.54737556e-01
3.69529605e-01 -1.26889420e+00 1.81834340e-01 8.16030681e-01
4.98061955e-01 5.42625666e-01 2.95389503e-01 -4.42136019e-01
-1.31522679e+00 -1.39217114e+00 2.04449505e-01 -4.01971161e-01
6.18193924e-01 -1.68729827e-01 -9.78266895e-01 9.44471419e-01
2.82540917e-01 6.49865746e-01 4.77343589e-01 -2.24419579e-01
-2.07181215e-01 -1.51867762e-01 -9.83005583e-01 4.31962103e-01
1.09549713e+00 -5.40636122e-01 -6.72035456e-01 2.78569877e-01
1.11837196e+00 -3.61859202e-01 -1.22445059e+00 3.18878531e-01
3.66802067e-01 -7.31413841e-01 1.17260492e+00 -8.76426518e-01
5.54179192e-01 -4.65123683e-01 -7.41024911e-02 -9.67661142e-01
-2.87550837e-01 -4.57326353e-01 -5.67277133e-01 1.06813574e+00
1.92430332e-01 -4.10385370e-01 9.16123331e-01 5.45327544e-01
-1.02526277e-01 -6.13143563e-01 -6.80309534e-01 -8.34977567e-01
-1.05695561e-01 -3.72762591e-01 4.69805986e-01 8.59432817e-01
-5.44426203e-01 -3.13606262e-01 -8.12809449e-03 2.33563989e-01
7.38123715e-01 2.17485577e-01 7.93019414e-01 -1.10393918e+00
-1.81172997e-01 -4.61279482e-01 -9.60865676e-01 -1.44376314e+00
4.97320294e-01 -8.20995808e-01 7.04797581e-02 -1.61740887e+00
2.12503627e-01 -3.20726335e-01 -5.80562413e-01 4.02550787e-01
-1.27162382e-01 5.24772584e-01 2.65213460e-01 2.47421831e-01
-1.03946364e+00 5.34546137e-01 1.57307053e+00 -7.21660435e-01
-3.33543986e-01 -2.71548748e-01 -1.59933344e-01 6.46176875e-01
4.82947737e-01 -2.93208301e-01 -3.56199294e-01 -6.61283970e-01
-2.95353919e-01 -1.46413282e-01 4.42957520e-01 -1.01996565e+00
1.98533177e-01 -2.66728014e-01 3.94012332e-01 -5.51031590e-01
1.61887497e-01 -9.54745710e-01 -5.76163307e-02 3.66517723e-01
-3.45150650e-01 -4.17950571e-01 1.53653517e-01 8.23767662e-01
-4.84341383e-01 -4.73776370e-01 6.57624006e-01 -1.77746281e-01
-1.30134046e+00 8.73835683e-01 -3.54239583e-01 2.09143415e-01
1.25640094e+00 -7.00611889e-01 -1.01410888e-01 3.33167054e-02
-1.13329530e+00 3.09932917e-01 6.34679377e-01 6.43624961e-01
7.37693250e-01 -1.33339906e+00 -1.36992767e-01 1.92363933e-01
5.26672006e-02 4.45368469e-01 2.87963986e-01 6.61483049e-01
-7.77083218e-01 6.11263998e-02 -3.83700371e-01 -1.14006031e+00
-7.87595272e-01 7.89307237e-01 4.22711253e-01 2.26800457e-01
-6.80159152e-01 8.37802887e-01 2.51787513e-01 -2.97470391e-01
4.11081254e-01 -5.32644033e-01 -3.42699707e-01 2.72127002e-01
2.95944452e-01 4.25170839e-01 -4.62196171e-01 -7.47236550e-01
-1.67262182e-01 8.15131068e-01 -9.75529626e-02 1.60395458e-01
1.26810765e+00 -1.87133074e-01 -2.22434700e-01 6.74753070e-01
1.65600204e+00 -5.95354378e-01 -1.82798111e+00 -2.09679082e-01
-1.29036263e-01 -6.28031850e-01 1.05920151e-01 -6.54990822e-02
-1.68727899e+00 7.69149184e-01 6.27930582e-01 2.16105312e-01
8.65619600e-01 2.72472575e-02 9.76463616e-01 4.74216938e-01
2.18694046e-01 -1.44900298e+00 5.19241095e-01 5.22487760e-01
2.99518883e-01 -1.25702095e+00 -1.57814607e-01 -4.09655213e-01
-6.19142950e-01 1.31425965e+00 7.39742637e-01 -4.62520927e-01
7.57594466e-01 -4.99015376e-02 1.72817558e-02 -1.85903888e-02
-2.37852350e-01 -3.36841732e-01 3.28325927e-01 6.06681764e-01
5.88824879e-03 -9.46135968e-02 7.02241585e-02 2.39363223e-01
4.00826246e-01 1.60638139e-01 4.45647329e-01 8.00042987e-01
-4.41055268e-01 -8.66082907e-01 -2.74593513e-02 5.05080938e-01
-1.52957588e-01 2.58952618e-01 -1.30504802e-01 9.32358921e-01
1.85595155e-01 6.17131650e-01 5.76830983e-01 -3.00662518e-01
1.08951814e-01 -5.77139668e-03 7.23360658e-01 -6.06167257e-01
-3.04957807e-01 -6.29089028e-02 -4.71913844e-01 -7.37342536e-01
-1.02880085e+00 -6.17245734e-01 -1.61215162e+00 4.37498689e-02
-4.21291262e-01 -4.83887382e-02 2.56972283e-01 8.18756938e-01
1.32432371e-01 6.96910560e-01 7.24783063e-01 -1.24676931e+00
9.16757137e-02 -6.20247066e-01 -6.00658655e-01 6.67814612e-01
4.45532531e-01 -7.83333004e-01 -3.00532699e-01 7.03206241e-01]
|
[9.153938293457031, -0.19848661124706268]
|
a4832aeb-f4c2-479c-9a5f-03a22f73baef
|
deep-attentive-sentence-ordering-network
| null | null |
https://aclanthology.org/D18-1465
|
https://aclanthology.org/D18-1465.pdf
|
Deep Attentive Sentence Ordering Network
|
In this paper, we propose a novel deep attentive sentence ordering network (referred as ATTOrderNet) which integrates self-attention mechanism with LSTMs in the encoding of input sentences. It enables us to capture global dependencies among sentences regardless of their input order and obtains a reliable representation of the sentence set. With this representation, a pointer network is exploited to generate an ordered sequence. The proposed model is evaluated on Sentence Ordering and Order Discrimination tasks. The extensive experimental results demonstrate its effectiveness and superiority to the state-of-the-art methods.
|
['Baiyun Cui', 'Zhongfei Zhang', 'Yingming Li', 'Ming Chen']
|
2018-10-01
| null | null | null |
emnlp-2018-10
|
['sentence-ordering', 'concept-to-text-generation']
|
['natural-language-processing', 'natural-language-processing']
|
[ 3.57145160e-01 -1.33350134e-01 -3.54281291e-02 -7.23865509e-01
-1.08456142e-01 -2.09793001e-01 3.51466209e-01 1.78323716e-01
-5.82072675e-01 7.17891276e-01 5.58445573e-01 -4.02252704e-01
-5.18642087e-03 -8.02632034e-01 -5.28457463e-01 -2.91296422e-01
-7.50268102e-02 3.79153520e-01 3.96490425e-01 -4.78065819e-01
6.61113739e-01 2.69469291e-01 -1.48415971e+00 6.77054822e-01
1.09728563e+00 8.58163059e-01 6.11033142e-01 8.95684838e-01
-3.26412141e-01 1.23141503e+00 -9.85619962e-01 -1.65645853e-01
-2.95934528e-01 -7.57683694e-01 -1.21418941e+00 -1.25877142e-01
2.59801298e-01 -6.49706006e-01 -6.21760249e-01 1.08424640e+00
5.79749227e-01 5.36811233e-01 4.79033768e-01 -4.01082367e-01
-1.15018976e+00 9.06357646e-01 -4.54499722e-02 9.79113877e-01
7.26976514e-01 1.48935229e-01 1.12470174e+00 -5.53528309e-01
4.18680370e-01 1.30342484e+00 8.64151195e-02 5.07101834e-01
-9.07996595e-01 -3.13483775e-01 3.96575451e-01 7.32974172e-01
-1.04671955e+00 -5.63800156e-01 9.29582059e-01 -3.92351858e-02
1.55465317e+00 4.36685145e-01 6.58965290e-01 9.26864326e-01
7.96889305e-01 1.06294525e+00 9.13591683e-01 -5.42460084e-01
1.53208762e-01 -4.52646196e-01 7.76020706e-01 6.69756472e-01
-1.95685625e-01 -1.37718335e-01 -6.25261962e-01 2.79016674e-01
6.93263471e-01 1.12443103e-03 -2.08074793e-01 3.30174863e-01
-8.50431323e-01 5.11711180e-01 9.02853668e-01 6.73812687e-01
-6.39239192e-01 -1.50677279e-01 6.84770167e-01 4.54636157e-01
5.17346203e-01 3.31398100e-01 -1.13352060e-01 5.08843064e-02
-6.74635351e-01 -9.91890579e-02 5.20238101e-01 7.61871517e-01
2.88757592e-01 2.28089437e-01 -7.73784816e-01 7.21672118e-01
2.26827994e-01 1.06012277e-01 7.85043597e-01 -4.46524620e-01
7.57699549e-01 5.91235816e-01 -1.72276303e-01 -1.22610879e+00
-4.27095950e-01 -5.57388902e-01 -1.16202140e+00 -5.39517105e-01
-3.99906486e-01 9.39858779e-02 -8.30847085e-01 1.51879942e+00
-1.63496047e-01 2.04380155e-01 2.27137014e-01 1.01354289e+00
1.25757229e+00 9.85608578e-01 2.19993085e-01 -4.41092491e-01
1.31894076e+00 -9.69729364e-01 -9.42953408e-01 -5.22660255e-01
1.02695838e-01 -4.79471147e-01 1.23528016e+00 1.16072096e-01
-1.31759906e+00 -1.04067194e+00 -1.10772598e+00 -2.53369033e-01
7.66574172e-03 -2.56383885e-02 5.94132841e-01 4.80836071e-02
-1.35408890e+00 6.75221503e-01 -6.19234622e-01 -4.32833999e-01
1.79076806e-01 6.41956151e-01 -8.54133219e-02 1.96887806e-01
-1.88256013e+00 9.40524161e-01 9.21632707e-01 5.60298681e-01
-7.60264933e-01 -5.69232367e-02 -9.16074693e-01 4.97431219e-01
-3.20726857e-02 -9.09127891e-01 1.40990674e+00 -7.30301201e-01
-1.76220584e+00 6.50508225e-01 -5.45442820e-01 -5.56377530e-01
-1.92926094e-01 -2.77670175e-01 -4.36388493e-01 3.50062311e-01
-2.20773965e-02 4.37432140e-01 6.36332810e-01 -6.45939827e-01
-3.67268860e-01 -2.34823361e-01 2.69640416e-01 6.41408503e-01
-5.69051445e-01 3.04587901e-01 -1.60533875e-01 -4.56012547e-01
1.51796818e-01 -3.67625624e-01 -2.05036953e-01 -9.90002632e-01
-6.98768675e-01 -6.23112500e-01 5.09091079e-01 -7.30741560e-01
1.74608576e+00 -1.99499261e+00 2.96107143e-01 -3.07851315e-01
1.51534349e-01 2.96499252e-01 -3.34787130e-01 6.62226319e-01
-2.11860240e-02 2.46605888e-01 -1.78879425e-01 -2.55165994e-01
-9.59335640e-02 2.95202017e-01 -3.49945694e-01 1.06071644e-01
2.54121333e-01 1.05341780e+00 -8.99400830e-01 -6.97471738e-01
1.28134757e-01 4.42199446e-02 -2.48904154e-01 7.46923745e-01
-2.53799945e-01 3.75826627e-01 -2.91090697e-01 9.70315784e-02
5.73289812e-01 -3.43423367e-01 3.73852253e-01 -3.60666029e-02
2.61326414e-02 9.79970038e-01 -4.56613630e-01 1.63028634e+00
-2.74292707e-01 4.87543643e-01 -5.01573145e-01 -8.85032594e-01
1.10833204e+00 2.48073220e-01 -3.25825125e-01 -1.05496788e+00
3.98397148e-01 -1.27288446e-01 5.06010234e-01 -9.47032928e-01
6.49458468e-01 -7.26164952e-02 -2.00936005e-01 4.86983865e-01
2.15814680e-01 2.41152659e-01 5.87573528e-01 3.46147358e-01
1.04643655e+00 -2.23376140e-01 5.90711594e-01 -4.13584709e-01
1.02689803e+00 -4.83460963e-01 2.77390152e-01 7.40846813e-01
-1.44286990e-01 1.86787099e-01 5.04705608e-01 -7.43446827e-01
-8.27948153e-01 -1.08283043e+00 1.84892699e-01 1.39658821e+00
2.13034704e-01 -1.31643042e-01 -7.09345877e-01 -7.20583618e-01
-5.39648712e-01 1.00075901e+00 -5.38028598e-01 -4.17350262e-01
-9.58236039e-01 -3.87580365e-01 2.83816546e-01 6.89868450e-01
7.63458252e-01 -1.82073724e+00 -5.15531957e-01 3.50370377e-01
-4.92747307e-01 -9.48499560e-01 -5.79096198e-01 7.52768144e-02
-8.92120004e-01 -7.03586996e-01 -2.67880112e-01 -1.26078594e+00
8.44163895e-01 3.77883501e-02 1.25362945e+00 1.04768381e-01
2.60378301e-01 -3.41080993e-01 -4.90258783e-01 1.32126659e-01
-3.08247060e-01 4.76900190e-01 -5.17357402e-02 2.25575417e-02
4.54613000e-01 -5.30528009e-01 -7.13686466e-01 -2.26462245e-01
-9.60784972e-01 1.70101617e-02 7.39944160e-01 8.96167397e-01
1.95829064e-01 -3.04528605e-02 5.84384263e-01 -9.54877198e-01
1.40312958e+00 -2.90235668e-01 -2.29370907e-01 3.42276543e-01
-1.94330767e-01 2.79905319e-01 1.12029529e+00 -1.37561679e-01
-1.20584571e+00 -3.32353443e-01 -4.65650976e-01 9.68635455e-02
-3.17700386e-01 7.99936950e-01 -4.94951606e-02 3.60783249e-01
2.32205465e-01 8.59995544e-01 -2.00460836e-01 -4.82814014e-01
3.25604193e-02 7.20766068e-01 6.22923195e-01 -2.07269177e-01
1.29420131e-01 -1.53672740e-01 -2.20157981e-01 -6.73544407e-01
-1.08911729e+00 -1.14880152e-01 -8.73376548e-01 -2.29343399e-01
8.38545918e-01 -3.92924100e-01 -8.40371788e-01 4.67277676e-01
-1.90757155e+00 -9.30947885e-02 1.62645027e-01 1.82078943e-01
-3.10106963e-01 5.04817307e-01 -1.24891937e+00 -8.12408626e-01
-8.38305175e-01 -1.02360427e+00 6.90754235e-01 3.98660690e-01
-2.60922194e-01 -1.07118952e+00 -1.41140208e-01 -1.11280642e-01
4.40155745e-01 -2.29062170e-01 1.02152145e+00 -8.89444888e-01
-4.28560793e-01 -7.98353031e-02 -1.58284262e-01 2.31307715e-01
7.02396557e-02 -1.38016164e-01 -5.95671833e-01 -7.93937445e-02
3.21165651e-01 -2.38363042e-01 1.08853006e+00 2.66338587e-01
1.28015697e+00 -7.27053642e-01 2.13271212e-02 2.60066360e-01
1.14878976e+00 5.21516263e-01 8.39301646e-01 2.14207172e-01
4.26934272e-01 5.57676554e-01 3.97006512e-01 3.57656568e-01
5.63261926e-01 2.77213603e-01 3.98482591e-01 5.84112033e-02
1.45530805e-01 -2.71129787e-01 1.46716610e-01 1.70033419e+00
2.44946241e-01 -5.90998769e-01 -7.77566850e-01 3.24717790e-01
-1.70648634e+00 -1.23990655e+00 -1.48598582e-01 1.67901480e+00
9.66315866e-01 5.96959054e-01 -2.07298279e-01 1.91595793e-01
9.91265416e-01 6.81593418e-01 -4.16247219e-01 -1.04816520e+00
6.16874453e-03 2.55290836e-01 -1.71299115e-01 7.03766227e-01
-1.13420427e+00 1.07068586e+00 7.33050871e+00 5.65684080e-01
-1.16713035e+00 -5.57076968e-02 8.17905486e-01 2.76289493e-01
-2.77259886e-01 -3.89755756e-01 -7.14626193e-01 6.71316624e-01
1.04411280e+00 -3.75743449e-01 3.90221149e-01 4.94347274e-01
2.76812404e-01 9.05763060e-02 -9.75092232e-01 5.54561734e-01
4.08476323e-01 -1.17832184e+00 3.15087527e-01 -7.17707992e-01
3.99137735e-01 -2.46979266e-01 -3.78862098e-02 4.48646158e-01
1.74514607e-01 -9.49334443e-01 5.10158122e-01 7.88366735e-01
6.78201735e-01 -7.54165113e-01 1.04307461e+00 4.92109925e-01
-1.18161881e+00 -2.34684125e-01 -6.06560707e-01 -8.36475968e-01
3.25296640e-01 4.01006848e-01 -7.09936023e-01 6.65401936e-01
3.85015935e-01 7.97545433e-01 -6.99346006e-01 8.16106617e-01
-7.18209803e-01 5.23090482e-01 1.63514167e-01 -6.87059462e-01
2.52251655e-01 2.95568164e-02 3.22394341e-01 1.25501347e+00
-1.78012289e-02 4.26912576e-01 1.72404096e-01 6.61567211e-01
-2.45064780e-01 1.05737165e-01 -5.52344382e-01 9.27838907e-02
6.58118784e-01 9.86767709e-01 -5.23295343e-01 -5.87881207e-01
7.15873092e-02 1.13268220e+00 8.38665426e-01 3.02289277e-01
-7.06301749e-01 -7.29060054e-01 2.10587382e-01 -3.96495759e-01
3.13292980e-01 -2.15031818e-01 -2.58017987e-01 -1.00168931e+00
4.63746376e-02 -6.30111814e-01 3.81695092e-01 -7.99532056e-01
-1.34511292e+00 1.34307170e+00 -1.41495377e-01 -1.11470008e+00
-5.12447953e-01 -3.24146599e-01 -9.62745786e-01 8.79523575e-01
-1.37676299e+00 -6.40112698e-01 -1.26601234e-01 2.58831829e-01
8.39782596e-01 -3.71719226e-02 8.61581504e-01 8.45231339e-02
-7.42604256e-01 5.21331012e-01 -1.66342124e-01 2.19249889e-01
7.71366283e-02 -1.11667418e+00 7.45196521e-01 1.06381595e+00
-1.26060858e-01 1.15519500e+00 6.04516685e-01 -5.48237443e-01
-1.05227613e+00 -8.79976571e-01 1.49146867e+00 1.88624591e-01
5.87737262e-01 -3.73518497e-01 -8.42159629e-01 5.11794269e-01
1.02578890e+00 -3.58782977e-01 6.26327395e-01 -3.19574960e-02
3.00653698e-03 -3.81187052e-01 -6.71724737e-01 6.30297661e-01
1.12899470e+00 -5.54794848e-01 -1.14413607e+00 1.22649394e-01
1.25052071e+00 -5.50323546e-01 -5.81548274e-01 4.26071852e-01
2.52126902e-01 -1.11354077e+00 6.61236882e-01 -7.51067221e-01
8.92202020e-01 -1.13031030e-01 1.78031608e-01 -1.31130874e+00
-1.01242518e+00 -3.00935268e-01 -3.57725412e-01 1.23344040e+00
2.66324282e-01 -6.42970145e-01 4.12260205e-01 2.54242033e-01
-4.51106697e-01 -9.03319478e-01 -8.86466682e-01 -7.23268986e-01
-2.72601813e-01 8.51712599e-02 8.35939467e-01 5.12871623e-01
1.54461220e-01 1.04703176e+00 -2.36140624e-01 -1.82569146e-01
2.46455327e-01 2.19261333e-01 7.34430403e-02 -8.47566724e-01
-3.79545897e-01 -5.23327172e-01 -2.60577887e-01 -1.68029654e+00
5.75424671e-01 -8.28933239e-01 4.60256904e-01 -1.97728598e+00
2.68750131e-01 3.09079915e-01 -7.77013242e-01 2.35492453e-01
-5.93603671e-01 -1.83363948e-02 3.43988270e-01 2.07648441e-01
-1.18935287e+00 8.89999926e-01 1.37811744e+00 -4.14899826e-01
-3.20004039e-02 -1.25297323e-01 -6.23561740e-01 3.76864344e-01
1.14451408e+00 -2.78548867e-01 -4.41057265e-01 -1.01728511e+00
-5.66490404e-02 2.37497851e-01 7.10285157e-02 -1.06240523e+00
4.66670841e-01 -6.90747648e-02 4.41049069e-01 -1.02570701e+00
2.37165093e-02 -3.51948202e-01 -3.87922764e-01 7.51653731e-01
-1.06676328e+00 4.06337738e-01 9.69952494e-02 2.24843398e-01
-4.72193658e-01 -5.31306803e-01 4.22503293e-01 -2.83292621e-01
-8.41432691e-01 3.45869586e-02 -5.12769401e-01 2.57033017e-02
7.26263523e-01 1.11038692e-01 -6.15073323e-01 -4.01135832e-01
-5.82676470e-01 2.20530376e-01 -1.23198681e-01 4.48335856e-01
1.14715850e+00 -1.33387733e+00 -7.60575294e-01 3.40072304e-01
-2.74603903e-01 -1.91418260e-01 4.46754903e-01 4.99499083e-01
-6.64307117e-01 8.94795001e-01 -2.63741493e-01 -3.79645020e-01
-1.32064915e+00 7.15799272e-01 1.80623725e-01 -5.23799360e-01
-2.87526190e-01 1.03225029e+00 5.69472052e-02 -2.39404336e-01
2.94531912e-01 -1.96071699e-01 -8.60889733e-01 -2.59928316e-01
6.79521620e-01 -1.94941051e-02 -8.61855969e-02 -5.02933204e-01
-5.33540666e-01 2.09141880e-01 -4.53784764e-01 1.77779123e-01
1.13607240e+00 -1.58720344e-01 -8.84161413e-01 5.25042355e-01
1.15868700e+00 -4.31321204e-01 -8.59350502e-01 -7.40976483e-02
1.44255638e-01 -3.31565708e-01 -4.13247108e-01 -5.82359374e-01
-7.15914309e-01 1.07945132e+00 8.40698853e-02 7.20328748e-01
1.32015312e+00 -1.41166791e-01 9.90391314e-01 4.02929097e-01
1.08781226e-01 -6.98975921e-01 3.33123416e-01 1.17018342e+00
1.18773401e+00 -9.33518410e-01 -1.42806128e-01 -3.04454744e-01
-6.43104911e-01 1.19086754e+00 1.13005912e+00 -4.77641433e-01
1.85118988e-01 3.31381299e-02 -1.78936839e-01 4.00809618e-03
-1.05852401e+00 -2.63647437e-01 2.73761123e-01 3.63243282e-01
7.79238164e-01 -9.59590748e-02 -7.88510203e-01 7.36831248e-01
-3.41809422e-01 -1.13045506e-01 3.75356585e-01 7.22699285e-01
-6.92287564e-01 -1.02015376e+00 -1.13306902e-01 6.31092191e-01
-3.38850021e-01 -6.89432025e-01 -4.88742143e-01 1.79622069e-01
-1.59888029e-01 1.06320822e+00 4.29645807e-01 -7.00788438e-01
2.90358961e-01 -3.34739685e-04 3.37684155e-01 -7.69527495e-01
-8.24276328e-01 -1.67804524e-01 1.17060587e-01 -2.28805706e-01
-1.97297081e-01 -3.85050863e-01 -1.33292115e+00 -2.08592653e-01
-5.57123385e-02 2.86972076e-01 1.67792551e-02 9.91235733e-01
5.74397981e-01 1.24281061e+00 8.85845542e-01 -7.88287282e-01
-6.28429532e-01 -1.37940121e+00 -2.20546961e-01 3.14491600e-01
3.96898985e-01 -1.24388374e-01 -6.17997311e-02 -7.33153969e-02]
|
[11.167977333068848, 8.596162796020508]
|
07682c18-c8e4-4f95-8aaf-e9e71c30380e
|
deepbillboard-systematic-physical-world
|
1812.10812
| null |
http://arxiv.org/abs/1812.10812v1
|
http://arxiv.org/pdf/1812.10812v1.pdf
|
DeepBillboard: Systematic Physical-World Testing of Autonomous Driving Systems
|
Deep Neural Networks (DNNs) have been widely applied in many autonomous
systems such as autonomous driving. Recently, DNN testing has been intensively
studied to automatically generate adversarial examples, which inject
small-magnitude perturbations into inputs to test DNNs under extreme
situations. While existing testing techniques prove to be effective, they
mostly focus on generating digital adversarial perturbations (particularly for
autonomous driving), e.g., changing image pixels, which may never happen in
physical world. There is a critical missing piece in the literature on
autonomous driving testing: understanding and exploiting both digital and
physical adversarial perturbation generation for impacting steering decisions.
In this paper, we present DeepBillboard, a systematic physical-world testing
approach targeting at a common and practical driving scenario: drive-by
billboards. DeepBillboard is capable of generating a robust and resilient
printable adversarial billboard, which works under dynamic changing driving
conditions including viewing angle, distance, and lighting. The objective is to
maximize the possibility, degree, and duration of the steering-angle errors of
an autonomous vehicle driving by the generated adversarial billboard. We have
extensively evaluated the efficacy and robustness of DeepBillboard through
conducting both digital and physical-world experiments. Results show that
DeepBillboard is effective for various steering models and scenes. Furthermore,
DeepBillboard is sufficiently robust and resilient for generating
physical-world adversarial billboard tests for real-world driving under various
weather conditions. To the best of our knowledge, this is the first study
demonstrating the possibility of generating realistic and continuous
physical-world tests for practical autonomous driving systems.
|
['Cong Liu', 'Yuankun Zhu', 'Yuqun Zhang', 'Lingming Zhang', 'Husheng Zhou', 'Wei Li', 'Bei Yu']
|
2018-12-27
| null | null | null | null |
['dnn-testing']
|
['adversarial']
|
[ 4.60982285e-02 2.04924028e-02 3.19797635e-01 -2.62207121e-01
-2.89214939e-01 -8.60133708e-01 5.60307205e-01 -5.91361344e-01
-1.76737204e-01 8.94227266e-01 -6.19738996e-01 -7.23818004e-01
6.45465478e-02 -1.05672526e+00 -1.33321965e+00 -5.90408862e-01
-1.89058244e-01 2.01975971e-01 4.73319948e-01 -6.03714466e-01
-1.52119651e-01 8.80961239e-01 -1.94533885e+00 -2.35925898e-01
9.28910792e-01 7.21851528e-01 -1.06756300e-01 1.06319702e+00
5.85045338e-01 4.90391582e-01 -1.12789583e+00 -4.20378357e-01
7.36540556e-01 -2.36078873e-01 6.66848347e-02 -3.17850471e-01
6.11813366e-01 -7.67925620e-01 -1.00231385e+00 1.22166753e+00
6.11677885e-01 -4.70561050e-02 3.04814965e-01 -2.01062584e+00
-3.82506102e-01 2.18996197e-01 -1.06615934e-03 7.91980922e-02
1.30243674e-01 1.08362532e+00 1.45215333e-01 -2.76240647e-01
3.11156482e-01 1.30580080e+00 5.29641628e-01 7.97352970e-01
-7.85370052e-01 -1.20162213e+00 -2.05012709e-01 3.35559905e-01
-1.11041260e+00 -2.82119542e-01 7.49141216e-01 -3.68302345e-01
7.31980562e-01 3.10086548e-01 5.71321547e-01 1.56619287e+00
9.63396788e-01 4.72330093e-01 1.14919329e+00 -1.80887803e-02
4.22708482e-01 9.67866033e-02 -2.06737548e-01 3.54706883e-01
5.99959493e-01 1.11951542e+00 -1.51374474e-01 5.66692017e-02
3.77050489e-01 -3.99707496e-01 -2.73315702e-02 -1.94931164e-01
-8.47548425e-01 5.19364238e-01 2.76211232e-01 -5.66828907e-01
-6.77054226e-02 6.88091278e-01 3.81122649e-01 5.61672509e-01
-1.98822826e-01 6.18658125e-01 -3.18396002e-01 -2.74574846e-01
-2.86308706e-01 7.94603109e-01 7.81913698e-01 1.22022581e+00
6.45300090e-01 7.08511949e-01 -1.35980710e-01 4.14428174e-01
8.83933976e-02 1.30059171e+00 2.15087757e-01 -1.00583172e+00
3.80090028e-01 1.55891597e-01 2.50648588e-01 -1.04588330e+00
-3.82054567e-01 -1.02700926e-01 -4.11024094e-01 1.07302177e+00
1.18277214e-01 -8.69526505e-01 -1.18479764e+00 1.78317356e+00
2.62303233e-01 3.84463519e-01 3.99625003e-01 8.19398642e-01
5.37186742e-01 5.05557358e-01 -2.08312616e-01 4.54736501e-01
7.63047278e-01 -4.32841331e-01 -5.95449209e-01 -5.45862913e-01
3.83503258e-01 -6.12361908e-01 1.07628787e+00 4.39580619e-01
-7.62882531e-01 -7.29690254e-01 -1.68388343e+00 6.44078374e-01
-5.62724233e-01 -4.29334939e-01 2.91796297e-01 1.13282204e+00
-8.27284515e-01 3.04654390e-01 -6.52662098e-01 1.21117190e-01
1.10894389e-01 2.46799588e-01 -2.23094776e-01 -3.55311483e-01
-1.90508270e+00 1.13229167e+00 8.53670165e-02 2.39668608e-01
-1.81084621e+00 -7.74658322e-01 -9.74384069e-01 -5.85394144e-01
3.37970585e-01 -2.96530187e-01 1.32506788e+00 -5.95703185e-01
-1.51616824e+00 4.09221947e-01 3.26544881e-01 -8.45040023e-01
8.90254021e-01 -3.35042477e-01 -1.05737126e+00 -2.21268997e-01
-6.18987046e-02 8.47817600e-01 8.41630220e-01 -1.38397324e+00
-1.59693390e-01 -7.22576678e-02 3.48642379e-01 -5.15796170e-02
-6.63829446e-02 -3.44494075e-01 -2.11847760e-02 -5.84063172e-01
-4.58730459e-01 -1.44065082e+00 -1.76694766e-01 6.32245392e-02
-6.72727644e-01 3.54670793e-01 1.40736914e+00 -1.62561908e-01
4.18490380e-01 -2.29217482e+00 -4.95237052e-01 1.70080140e-01
-1.37032270e-01 6.70900047e-01 -3.00926208e-01 3.75962406e-01
5.12757674e-02 1.63826898e-01 -1.46114081e-01 3.56804192e-01
3.53539526e-01 5.14166832e-01 -7.54623950e-01 4.33466434e-01
4.48388755e-01 1.02942753e+00 -7.45098114e-01 6.01356775e-02
3.44062716e-01 2.60013193e-01 -4.65037376e-01 1.90013915e-01
-3.26818377e-01 1.78424031e-01 -2.77074695e-01 6.31756008e-01
9.19657886e-01 9.05069709e-01 -4.96117324e-01 2.40380958e-01
9.78353694e-02 -1.72967657e-01 -1.10092330e+00 6.71734571e-01
-2.88825840e-01 1.20021141e+00 -4.79601286e-02 -3.04773062e-01
1.10644913e+00 -9.18008089e-02 -1.36137724e-01 -1.06348860e+00
1.46244198e-01 1.97420314e-01 3.52926940e-01 -6.06992185e-01
6.15257740e-01 -8.19164515e-02 -5.53303421e-01 -7.01828906e-03
-4.44033444e-01 -9.15785670e-01 8.73433277e-02 2.60563586e-02
1.44764018e+00 -3.56732048e-02 -4.90433097e-01 -1.02311805e-01
2.07272381e-01 1.39882326e-01 5.03735662e-01 8.71322513e-01
-6.04700089e-01 4.47657526e-01 6.40468657e-01 -3.67881238e-01
-1.14267850e+00 -1.39485502e+00 -1.06873766e-01 3.11784804e-01
5.75937450e-01 2.58059740e-01 -8.89413536e-01 -4.18877602e-01
4.67428684e-01 1.19311559e+00 -5.94179094e-01 -9.20850992e-01
-4.95147198e-01 -4.02631938e-01 1.35138857e+00 5.14004648e-01
7.98079431e-01 -1.17010438e+00 -6.24388278e-01 -2.61425078e-02
4.35580462e-01 -1.37096500e+00 -1.66670084e-01 5.05975001e-02
-1.64512187e-01 -1.06117582e+00 -3.71502973e-02 -4.37824219e-01
5.26449978e-01 1.32721215e-01 8.52888763e-01 -1.10359877e-01
-4.43245679e-01 2.03406572e-01 -7.02312961e-02 -9.45508003e-01
-9.72631216e-01 -5.13889194e-01 4.59893584e-01 -2.64599204e-01
1.20506264e-01 -3.55670363e-01 -4.49939936e-01 9.65112388e-01
-1.16094685e+00 -2.89164335e-01 5.88440776e-01 6.16672099e-01
2.89708078e-01 3.85421515e-01 6.58039927e-01 -6.15566492e-01
7.97416627e-01 -4.57442313e-01 -1.01973534e+00 -2.31022596e-01
-3.41825008e-01 -1.69203818e-01 1.07561541e+00 -6.03594542e-01
-5.80460250e-01 -2.54760146e-01 -2.60071874e-01 -6.13484561e-01
-4.87058431e-01 6.49617836e-02 -5.60670972e-01 -4.91366595e-01
9.51215267e-01 9.09046829e-02 3.95932607e-02 4.39332187e-01
2.49135002e-01 4.32386309e-01 7.47343421e-01 -4.89942491e-01
1.49246240e+00 2.93421626e-01 2.27146804e-01 -7.31049836e-01
-9.86933708e-02 4.69481498e-01 -1.29527114e-02 -6.38703465e-01
5.37029862e-01 -7.61177242e-01 -8.36471260e-01 1.11837363e+00
-7.54844189e-01 -9.20806885e-01 -2.16053173e-01 3.06180239e-01
-5.69956958e-01 1.29743040e-01 -2.49782726e-01 -5.52903950e-01
2.67906368e-01 -1.52641523e+00 6.62469387e-01 1.92670986e-01
-5.29726408e-02 -5.19164443e-01 9.39173326e-02 1.71188280e-01
5.21348059e-01 7.35214531e-01 5.45373976e-01 -2.85075128e-01
-8.00057054e-01 -6.58393085e-01 1.63350806e-01 6.66094363e-01
-2.42271557e-01 3.32000256e-01 -9.88431215e-01 -3.48882854e-01
-3.01224470e-01 -5.50759852e-01 3.06596279e-01 4.57177684e-02
1.09220827e+00 -1.67477950e-01 -7.75195584e-02 6.70639932e-01
1.15607619e+00 5.79155743e-01 1.10523117e+00 4.65591103e-01
5.55727661e-01 3.27491552e-01 8.14766705e-01 8.21925849e-02
-7.85461068e-02 6.17710710e-01 1.06256318e+00 -1.80086732e-01
-1.69662997e-01 -3.22810829e-01 8.31465840e-01 -7.17328116e-02
6.83737099e-01 -6.45677984e-01 -8.75330567e-01 4.16857243e-01
-1.18098855e+00 -8.43882143e-01 -6.92868009e-02 2.09994006e+00
5.39730847e-01 8.53943169e-01 -2.18959510e-01 2.00078562e-01
7.78450251e-01 -3.27563286e-02 -1.15787053e+00 -9.10363197e-01
-3.07342410e-01 3.47465761e-02 1.03971338e+00 3.63807738e-01
-9.52396333e-01 7.82468140e-01 6.54068804e+00 5.24137557e-01
-1.17859375e+00 -3.32197994e-01 4.70783353e-01 -2.43733063e-01
-4.30485338e-01 -2.13374570e-01 -8.93711090e-01 6.52092278e-01
1.22059596e+00 -1.97428018e-01 3.33676696e-01 9.48180795e-01
4.80001122e-01 -1.73151400e-02 -9.02367949e-01 4.14924383e-01
2.06290018e-02 -9.36881602e-01 -4.75501679e-02 9.04657766e-02
9.06698823e-01 3.16124767e-01 3.91422153e-01 5.52975655e-01
8.09338987e-01 -1.21148336e+00 8.83511484e-01 3.46567661e-01
9.46459949e-01 -1.13559008e+00 9.13995445e-01 1.78612396e-01
-5.16199946e-01 -1.50978059e-01 -2.49579921e-01 -3.26652117e-02
6.28322288e-02 5.20946205e-01 -8.80220056e-01 5.13311438e-02
7.10454702e-01 1.20691925e-01 -4.79977310e-01 8.96018744e-01
-5.83227932e-01 7.75718927e-01 -2.80734479e-01 -2.72071362e-01
4.04385865e-01 2.91983336e-01 9.40231025e-01 9.02239799e-01
3.01641583e-01 -3.44298691e-01 -6.18400462e-02 8.64101648e-01
-2.38339435e-02 -7.35586941e-01 -1.31905651e+00 1.52652442e-01
6.44908905e-01 8.69164169e-01 -1.31261751e-01 1.72385015e-02
7.81624243e-02 5.31309843e-01 -2.94912845e-01 5.38534045e-01
-1.56081808e+00 -6.95373237e-01 1.34301591e+00 1.99963465e-01
1.44814089e-01 -5.21128714e-01 -3.76327097e-01 -4.44298714e-01
2.18100011e-01 -9.94623601e-01 -3.51507336e-01 -1.06158996e+00
-8.96731079e-01 6.20776355e-01 1.91952094e-01 -1.42572546e+00
-2.81951755e-01 -8.63375247e-01 -8.03440928e-01 7.91588247e-01
-1.53505957e+00 -7.42042780e-01 -6.52397335e-01 5.98350048e-01
4.53695595e-01 -4.47307914e-01 4.97656137e-01 1.52779266e-01
-8.33138645e-01 1.05031538e+00 6.65259957e-02 -5.81494123e-02
7.18780160e-01 -9.33662117e-01 9.71475184e-01 1.32742929e+00
-6.37625575e-01 2.97684997e-01 1.24396574e+00 -7.40131676e-01
-1.70938933e+00 -1.57593763e+00 -8.82851258e-02 -3.18097949e-01
7.64766574e-01 -5.99439085e-01 -6.62680507e-01 3.58100444e-01
1.62212431e-01 3.02728526e-02 1.06536590e-01 -7.00926006e-01
-2.37570941e-01 -3.93851191e-01 -1.42227030e+00 1.11571848e+00
9.01119292e-01 -3.49048346e-01 -1.07729405e-01 3.19127381e-01
9.99651611e-01 -7.91509748e-01 -5.01309216e-01 6.61316752e-01
4.17726517e-01 -9.17353451e-01 8.27297211e-01 -3.89604777e-01
4.72393841e-01 -6.96842372e-01 -1.46306425e-01 -1.64423811e+00
2.39108741e-01 -9.32934582e-01 1.18300512e-01 8.46640766e-01
5.63514471e-01 -1.07773042e+00 6.17914796e-01 5.61862648e-01
-7.37824976e-01 -4.58190411e-01 -9.18995559e-01 -1.23457491e+00
2.36911565e-01 -9.19055164e-01 7.72240162e-01 3.98883194e-01
-4.56434697e-01 -3.97859573e-01 -3.66590977e-01 7.25892663e-01
3.91818106e-01 -5.85058749e-01 1.36389923e+00 -5.90439677e-01
-1.74183100e-01 -1.06562696e-01 -1.10500526e+00 -5.66641808e-01
3.03977966e-01 -3.52464318e-01 5.53600371e-01 -9.66610849e-01
-5.52956879e-01 -5.42797863e-01 -6.55645388e-04 3.20747554e-01
-1.41785681e-01 3.77482235e-01 -4.44393158e-02 -2.37861291e-01
-5.41602820e-02 4.56537992e-01 1.24161184e+00 -4.24093187e-01
2.25711301e-01 2.47899871e-02 -4.26453322e-01 4.18561637e-01
1.15012753e+00 -4.24925476e-01 -7.85878062e-01 -3.69100273e-01
2.54396677e-01 -1.95709318e-01 8.52733493e-01 -1.56025743e+00
1.27008474e-02 -4.12394553e-01 1.59502044e-01 -4.92202133e-01
3.21001887e-01 -6.27321899e-01 3.38685453e-01 6.60966814e-01
1.03933096e-01 3.12689334e-01 1.05692279e+00 5.68672061e-01
-9.20868218e-02 1.47858873e-01 8.50394785e-01 3.37225169e-01
-1.14575434e+00 2.56003112e-01 -6.73770845e-01 1.17589302e-01
1.63328838e+00 -3.92435312e-01 -7.50011623e-01 -4.48694050e-01
-1.64532363e-01 2.92174876e-01 6.60540640e-01 7.53790915e-01
8.93898249e-01 -1.19822156e+00 -6.91611767e-01 6.98643327e-01
1.84021547e-01 -2.03734845e-01 4.02051270e-01 1.81776240e-01
-9.36887264e-01 2.33757928e-01 -5.04769444e-01 -5.99406660e-01
-8.69944334e-01 6.12673998e-01 7.18382537e-01 3.63620877e-01
-3.99809510e-01 5.75318456e-01 1.97665051e-01 -6.31323278e-01
2.31966395e-02 -2.00998873e-01 3.45311642e-01 -7.98504770e-01
2.90806413e-01 2.26679415e-01 3.14752281e-01 -3.42878342e-01
-2.91941792e-01 2.76686102e-01 2.59706289e-01 -1.26123652e-01
5.81689119e-01 3.00962538e-01 3.66016954e-01 1.25645667e-01
9.08890426e-01 5.04221208e-02 -1.69201517e+00 6.27241790e-01
-7.09598958e-01 -4.12073553e-01 -1.33742481e-01 -8.80765617e-01
-1.10392642e+00 6.42924070e-01 7.40352154e-01 1.83993533e-01
8.88936639e-01 -5.54371476e-01 1.02442575e+00 5.35519600e-01
4.61713135e-01 -9.35993850e-01 4.30498049e-02 5.81339002e-01
9.23308611e-01 -1.07014537e+00 -3.81652594e-01 -1.61572322e-01
-5.67522585e-01 7.50716090e-01 1.19168735e+00 -3.96665365e-01
5.40153623e-01 8.65213335e-01 4.52734977e-01 -7.00387545e-03
-7.04511702e-01 2.64589041e-01 -1.04635373e-01 9.44528520e-01
-4.53507990e-01 4.26281601e-01 3.07037026e-01 1.36926249e-02
-6.94362640e-01 -2.59144843e-01 8.85409772e-01 8.64350200e-01
-3.14247549e-01 -7.41994202e-01 -6.79752171e-01 2.57855773e-01
-9.82183497e-03 2.04415366e-01 -4.46370780e-01 1.11381876e+00
3.26484412e-01 1.24394619e+00 -2.57209875e-02 -9.25952137e-01
7.78366446e-01 -2.62770355e-01 5.88366985e-02 -7.26271048e-02
-2.32037783e-01 -8.34294140e-01 2.86568940e-01 -6.65073454e-01
3.39059561e-01 -5.06048322e-01 -1.17851281e+00 -6.61072969e-01
-9.72194001e-02 -2.25542426e-01 8.92853498e-01 7.81122088e-01
3.76857996e-01 8.84052217e-01 8.99662793e-01 -6.96778595e-01
-6.21888340e-01 -6.08634830e-01 -4.08873558e-01 4.07232404e-01
4.32419062e-01 -9.06527221e-01 -6.51071668e-01 -3.76406878e-01]
|
[5.352816581726074, 7.820037364959717]
|
d81eb498-d40c-46a0-aa1d-55c004db6096
|
multi-task-end-to-end-training-improves
| null | null |
https://openreview.net/forum?id=D5u046Zw_2F
|
https://openreview.net/pdf?id=D5u046Zw_2F
|
Multi-Task End-to-End Training Improves Conversational Recommendation
|
In this paper, we analyze the performance of a multitask end-to-end transformer model on the task of conversational recommendations, which aim to provide recommendations based on a user’s explicit preferences expressed in dialogue. While previous works in this area adopt complex multi-component approaches where the dialogue generation and entity recommendation tasks are handled by separate components, we show that a unified transformer model, based on the T5 text-to-text transformer model, can perform competitively in both recommending relevant items and generating conversation dialogue. We fine-tune our model on the ReDIAL conversational movie recommendation dataset, and create additional training tasks derived from MovieLens (such as the prediction of movie attributes and related movies based on an input movie), in a multitask learning setting. Using a series of probe studies, we demonstrate that the learned knowledge in the additional tasks is transferred to the conversational setting, where each task leads to a $9\% - 52\%$ increase in its related probe score.
|
['Anonymous']
|
2021-10-16
| null | null | null |
acl-arr-october-2021-10
|
['movie-recommendation']
|
['miscellaneous']
|
[ 3.70882720e-01 2.81549394e-01 1.77857019e-02 -8.00946295e-01
-1.02650332e+00 -7.26019859e-01 7.02452958e-01 -1.50409713e-01
-3.17960948e-01 6.61996245e-01 8.18716168e-01 -3.32026809e-01
-1.36643514e-01 -5.42919099e-01 -3.80648494e-01 -3.80389154e-01
1.29721895e-01 9.81661618e-01 9.72013399e-02 -5.71949303e-01
3.25575143e-01 -2.90071398e-01 -1.29614460e+00 1.11693382e+00
6.39931321e-01 1.03746045e+00 1.78516790e-01 8.51640701e-01
-1.16452113e-01 9.29487646e-01 -4.01154578e-01 -9.85707700e-01
1.31976530e-01 -4.75331724e-01 -1.11845899e+00 1.36296362e-01
4.11318570e-01 -5.21124423e-01 5.16559742e-02 3.55264157e-01
6.14411533e-01 8.68421614e-01 9.98631954e-01 -7.88538933e-01
-6.36308253e-01 9.26861644e-01 1.73367992e-01 4.63751554e-02
5.33470452e-01 -4.37386811e-01 1.64095771e+00 -1.17695856e+00
3.68933678e-01 1.24021864e+00 4.75450605e-01 7.78135419e-01
-1.34237742e+00 -3.38450104e-01 3.85403454e-01 -1.86202213e-01
-4.63092804e-01 -6.02950931e-01 4.89905447e-01 -4.22550470e-01
1.12812185e+00 4.84031826e-01 9.55276117e-02 1.37706912e+00
-9.76202544e-03 9.46507454e-01 8.55310261e-01 -2.00761780e-01
8.93345773e-02 5.79649448e-01 2.27333099e-01 2.51953244e-01
-5.18768609e-01 -2.41048381e-01 -7.30787396e-01 -3.86683792e-01
3.78269851e-01 8.37866291e-02 -1.32330194e-01 -1.68864697e-01
-1.04608035e+00 1.21461296e+00 1.73646528e-02 1.28634691e-01
-2.70051986e-01 -3.38814825e-01 3.01813871e-01 7.29500532e-01
1.04527259e+00 8.60499918e-01 -7.69047797e-01 -2.66702324e-01
-5.49433649e-01 5.27744532e-01 1.18450356e+00 8.93829226e-01
2.72443056e-01 -3.66411895e-01 -6.67863727e-01 1.42275536e+00
3.59769136e-01 2.69590348e-01 5.34905791e-01 -1.11241937e+00
6.20585442e-01 7.08714053e-02 5.89551985e-01 -6.40286028e-01
-4.63867038e-01 -4.08157229e-01 -3.06650609e-01 -4.67055619e-01
5.72238386e-01 -6.24172688e-01 -9.82401967e-02 1.87981451e+00
3.73953283e-01 -1.63232878e-01 2.25883454e-01 7.41854310e-01
1.05592620e+00 6.27869308e-01 -8.94449130e-02 -3.94037038e-01
1.32982230e+00 -1.46502388e+00 -5.49549162e-01 -1.61052421e-01
6.42532945e-01 -9.74238217e-01 1.26908183e+00 5.98239541e-01
-1.26126099e+00 -6.10400200e-01 -5.46298683e-01 -1.99696988e-01
-2.68997736e-02 1.83846191e-01 5.70043385e-01 3.78455639e-01
-1.01959634e+00 5.77619851e-01 -1.97161362e-01 -4.46564198e-01
-2.00500652e-01 2.31338665e-01 2.30398867e-02 9.57508832e-02
-1.28991568e+00 8.77351046e-01 -4.17781711e-01 5.01210941e-03
-8.20921361e-01 -4.19530541e-01 -5.52266955e-01 1.23492964e-01
4.47120160e-01 -7.88227260e-01 1.98094738e+00 -7.96649992e-01
-2.03439546e+00 4.01127070e-01 -1.55972078e-01 -1.94529712e-01
3.17398190e-01 -3.07258725e-01 -2.03519031e-01 -2.30883285e-01
-5.72633483e-02 3.98739278e-01 6.99054778e-01 -8.82473707e-01
-1.04821944e+00 -1.37630075e-01 5.67343712e-01 6.28759205e-01
-6.31607354e-01 1.74791396e-01 -1.74465328e-01 -6.55215859e-01
-3.83901060e-01 -1.12236106e+00 -2.91606963e-01 -6.47246897e-01
-2.61184663e-01 -5.99007785e-01 -2.01068576e-02 -4.28349763e-01
1.09153426e+00 -1.77464902e+00 4.11889970e-01 6.20978139e-02
1.26969472e-01 -1.39618367e-01 -3.98588538e-01 6.95461690e-01
3.84111196e-01 -1.94735885e-01 3.27658564e-01 -9.05895352e-01
2.06097662e-01 -1.49099857e-01 -6.39480710e-01 -1.21119492e-01
-1.99043199e-01 6.89976811e-01 -8.18176150e-01 -3.66300605e-02
-2.79667825e-01 1.60657749e-01 -9.89298761e-01 7.58143306e-01
-4.72806573e-01 4.75180715e-01 -6.43647552e-01 -1.13483213e-01
-2.29031406e-02 -4.57991004e-01 4.43503320e-01 -1.21658400e-01
2.76652277e-01 9.67719376e-01 -7.20727563e-01 1.60963619e+00
-9.15412486e-01 2.01585680e-01 1.29954368e-01 -7.25959539e-01
8.96306574e-01 6.43871248e-01 3.92724425e-01 -4.86490518e-01
3.80227976e-02 -1.46889925e-01 7.26887807e-02 -5.07798553e-01
9.42717850e-01 -2.61044174e-01 -2.71699101e-01 1.23090684e+00
2.99914896e-01 -3.71234156e-02 7.70915970e-02 5.67329645e-01
9.80720162e-01 1.79414287e-01 -1.19323745e-01 -1.52007788e-01
3.18420261e-01 -2.41324544e-01 5.13936542e-02 9.49205935e-01
3.90020847e-01 2.98026115e-01 2.91470766e-01 -2.54900008e-01
-7.54423141e-01 -7.75180638e-01 -3.81290726e-02 2.24183822e+00
-2.79815018e-01 -5.68796694e-01 -5.46932578e-01 -9.97735798e-01
-8.83918777e-02 1.06416917e+00 -6.43438935e-01 -7.26963878e-02
-3.05072904e-01 -8.18374991e-01 -4.72002337e-03 4.11942691e-01
-1.35704994e-01 -1.14864874e+00 -1.16834119e-01 4.82610732e-01
-6.70712650e-01 -1.07383502e+00 -9.67017114e-01 -2.38984264e-02
-7.77519882e-01 -6.29518926e-01 -5.99293888e-01 -6.76076651e-01
4.90956277e-01 4.05963182e-01 1.38649333e+00 -2.09425583e-01
4.65791941e-01 6.39655411e-01 -6.58256471e-01 -6.67838752e-02
-5.04605114e-01 2.10415483e-01 5.20691834e-02 3.44433308e-01
2.74097353e-01 -7.12409914e-01 -5.74467242e-01 8.77445877e-01
-4.82207745e-01 3.81208330e-01 2.44383767e-01 9.72697556e-01
2.44507827e-02 -5.90141714e-01 1.23504901e+00 -1.38838124e+00
1.32257974e+00 -6.14608228e-01 1.45798307e-02 2.14403003e-01
-8.15715849e-01 -3.64614688e-02 6.94290459e-01 -4.96090680e-01
-1.45998669e+00 -1.81559965e-01 -3.57820243e-01 2.34550655e-01
6.25745729e-02 5.90358019e-01 1.71155259e-01 3.49048883e-01
8.46557021e-01 1.18403241e-01 -1.59537762e-01 -7.46054947e-01
7.29197383e-01 9.83847857e-01 1.60727635e-01 -9.17371511e-01
2.69428760e-01 -1.51877217e-02 -6.52107358e-01 -1.81756884e-01
-1.45016003e+00 -4.67550039e-01 -2.74132967e-01 -3.98182571e-01
6.59957945e-01 -9.02913213e-01 -8.40170264e-01 -2.42827479e-02
-1.03127873e+00 -6.42901957e-01 -2.08338767e-01 5.99719763e-01
-7.07983375e-01 1.85844511e-01 -1.02131391e+00 -8.38921428e-01
-5.77853739e-01 -1.07941914e+00 1.01173866e+00 -1.17893718e-01
-4.41371739e-01 -1.14693320e+00 3.08255106e-01 8.67436528e-01
6.90980613e-01 -7.18841851e-01 8.38637054e-01 -1.44473588e+00
-1.38436835e-02 -1.52520701e-01 1.51142731e-01 2.66252100e-01
4.41683717e-02 -4.92030919e-01 -1.04126549e+00 -3.35097611e-01
1.03732981e-01 -8.06042135e-01 7.20045447e-01 8.65048692e-02
8.23506176e-01 -5.78359902e-01 -8.89513865e-02 -1.84771512e-02
4.15293396e-01 1.15877725e-01 2.20397055e-01 -2.07954481e-01
4.12201971e-01 9.91874158e-01 8.29234004e-01 5.30432999e-01
8.50029588e-01 1.08432746e+00 4.15145420e-02 2.33493060e-01
1.87889427e-01 -3.52703810e-01 5.94908059e-01 1.07596970e+00
-1.28045291e-01 -6.05126381e-01 -1.77587822e-01 2.88942426e-01
-2.14793801e+00 -8.92884493e-01 1.25438824e-01 2.23946452e+00
1.13694882e+00 1.21939130e-01 5.56727886e-01 -4.90775615e-01
5.16335666e-01 -5.31373210e-02 -3.59861046e-01 -6.02212846e-01
3.16212565e-01 1.04153432e-01 -2.09965810e-01 7.19227314e-01
-9.11384344e-01 8.69462848e-01 6.44046926e+00 6.85675621e-01
-9.38059747e-01 3.22651535e-01 7.89242804e-01 -4.48735625e-01
-6.28612936e-01 -2.76891947e-01 -9.10706580e-01 3.38952452e-01
1.11648488e+00 -1.34022281e-01 7.78920889e-01 6.96253777e-01
2.45129451e-01 2.73959219e-01 -1.54803169e+00 4.79855537e-01
3.37534487e-01 -1.06838012e+00 1.09005272e-01 7.26347864e-02
6.65504217e-01 -5.04154190e-02 1.54856980e-01 7.64956892e-01
8.10476601e-01 -7.79806316e-01 5.08962393e-01 3.40500504e-01
5.20943582e-01 -3.35374296e-01 5.99098802e-01 5.58357179e-01
-6.95288479e-01 -1.84510902e-01 -3.62639636e-01 -1.45511538e-01
3.80969405e-01 2.89143205e-01 -1.16938233e+00 3.14822584e-01
4.40368056e-01 6.95459545e-01 -1.54331729e-01 4.32925224e-01
-1.05402231e-01 7.62999058e-01 1.16002962e-01 -2.78196752e-01
-1.59532987e-02 -3.57899934e-01 4.16172981e-01 1.31626761e+00
3.55924517e-01 2.69601166e-01 3.46765190e-01 5.02067208e-01
-4.05261725e-01 6.12767339e-01 -2.50639111e-01 1.06045194e-01
2.75699645e-01 1.80380511e+00 -2.60487467e-01 -3.31147522e-01
-5.49713016e-01 8.32569599e-01 6.00860953e-01 3.23510408e-01
-4.91292447e-01 4.21168506e-02 4.38539326e-01 -7.98102990e-02
6.00312114e-01 1.45104080e-01 1.12600520e-01 -1.25895703e+00
-3.29427361e-01 -1.06241643e+00 4.01677430e-01 -7.72113979e-01
-1.79618430e+00 8.12793553e-01 -2.13760465e-01 -1.09205484e+00
-8.26545060e-01 -3.67232919e-01 -6.52237058e-01 8.07522297e-01
-1.07817364e+00 -1.03929424e+00 7.24708289e-02 6.71696484e-01
8.87759328e-01 -3.80732208e-01 1.20426512e+00 4.13313866e-01
-2.97163546e-01 8.21487725e-01 2.70428568e-01 -1.91127807e-01
1.36537921e+00 -1.33345211e+00 3.11325729e-01 2.81570792e-01
3.29707533e-01 6.58328056e-01 7.69720078e-01 -3.11713964e-01
-1.16298580e+00 -1.04076993e+00 1.07072937e+00 -8.42809021e-01
6.05969489e-01 -6.98103964e-01 -4.80716079e-01 9.24770832e-01
3.40208411e-01 -4.64053929e-01 1.25718701e+00 9.32498872e-01
-3.38886529e-01 -1.18750772e-02 -1.03978920e+00 5.04091561e-01
1.10978401e+00 -5.53220749e-01 -6.18838131e-01 5.98644376e-01
8.22474897e-01 -3.93027902e-01 -1.18791401e+00 2.07280099e-01
7.77370274e-01 -6.84738100e-01 8.01385880e-01 -1.36812520e+00
8.71381164e-01 3.63638401e-01 -2.78348744e-01 -1.66203594e+00
-5.71834385e-01 -7.54949331e-01 -3.91199999e-02 1.25134838e+00
9.84497011e-01 -3.93079400e-01 4.61224526e-01 8.55523109e-01
-4.09576893e-01 -7.40388453e-01 -6.18182778e-01 -9.11569595e-02
4.43523861e-02 -3.07075679e-01 3.31932038e-01 7.71623790e-01
5.99023223e-01 1.32021809e+00 -9.51182544e-01 -4.13146168e-01
3.54487114e-02 5.39907575e-01 8.86978686e-01 -1.29773307e+00
-9.91988897e-01 -2.29064018e-01 7.57946432e-01 -1.73736930e+00
9.95258614e-02 -9.22539413e-01 3.56188029e-01 -1.37976396e+00
3.46454263e-01 -5.65748990e-01 -3.39526623e-01 1.40026957e-01
-3.05368632e-01 2.34220654e-01 1.45794630e-01 1.61899745e-01
-1.15736139e+00 5.61922371e-01 1.33074284e+00 1.28705531e-01
-1.90699667e-01 7.45823264e-01 -1.18970835e+00 4.17172909e-01
3.73839289e-01 -4.08059210e-01 -7.45239139e-01 -4.17317331e-01
6.41310215e-01 5.10427892e-01 -3.59626621e-01 -2.19082624e-01
1.36427388e-01 -3.37360986e-02 -6.73010945e-02 -6.98445812e-02
7.64640152e-01 -5.02903283e-01 -1.84969783e-01 -3.38590264e-01
-1.38848221e+00 -1.39094340e-02 -1.56819090e-01 7.30806828e-01
1.41472653e-01 -2.44910643e-01 2.21100390e-01 -7.60455653e-02
1.03509091e-01 2.38745093e-01 -6.82595193e-01 1.95626125e-01
4.54841018e-01 3.34102690e-01 -4.80357856e-01 -1.05317700e+00
-1.15882170e+00 2.32090384e-01 -1.05187029e-01 6.45469129e-01
3.08431596e-01 -1.21964192e+00 -9.58539128e-01 -2.16940537e-01
1.57507494e-01 -4.08810705e-01 3.85045528e-01 8.57529938e-01
6.62023306e-01 4.59526211e-01 1.72815192e-02 -2.03205258e-01
-1.28556478e+00 4.38711077e-01 2.65985578e-01 -6.51755691e-01
-1.80901796e-01 1.13310134e+00 4.80911553e-01 -8.71129811e-01
4.18336898e-01 -1.49195552e-01 -8.55906188e-01 2.17906922e-01
6.25331521e-01 1.66559681e-01 9.52321291e-02 -3.00457388e-01
1.94012627e-01 -9.37505160e-03 -4.96921718e-01 -4.66798037e-01
1.39605093e+00 -3.65102530e-01 4.04203385e-02 5.42778552e-01
7.27008402e-01 4.35778826e-01 -1.00508928e+00 -7.17466950e-01
-2.47980326e-01 -3.29245627e-01 -2.08316952e-01 -1.25759149e+00
-8.19463193e-01 5.66214979e-01 1.18714549e-01 6.22747004e-01
6.94546163e-01 1.30520806e-01 8.68895590e-01 7.04470456e-01
2.75441557e-01 -9.68547344e-01 3.79616708e-01 6.56878293e-01
9.67282951e-01 -1.19325733e+00 -3.95904988e-01 -2.97751278e-01
-1.11760211e+00 9.71499860e-01 6.51052654e-01 1.44597650e-01
6.75905228e-01 -7.52867386e-02 3.31784636e-02 -7.51039982e-02
-1.49965405e+00 1.52142033e-01 5.56852341e-01 1.81726247e-01
9.45864856e-01 -1.25694675e-02 -3.42664093e-01 1.28204238e+00
-3.09998512e-01 -2.58424915e-02 4.59544420e-01 3.27857643e-01
-4.43396360e-01 -1.31527412e+00 2.37380147e-01 9.33927238e-01
-5.29706359e-01 -3.65782797e-01 -4.63771701e-01 -2.04838112e-01
-3.71044874e-01 1.49109900e+00 2.77905576e-02 -8.06493282e-01
4.32270050e-01 1.81252763e-01 2.52370656e-01 -9.48806882e-01
-1.14630508e+00 2.29681477e-01 1.04418659e+00 -3.34608197e-01
-3.95706981e-01 -6.80889666e-01 -7.18814969e-01 -1.49884939e-01
-4.38612103e-01 7.93294072e-01 3.84168893e-01 1.14696372e+00
5.10439038e-01 4.08923000e-01 1.01764393e+00 -8.78276646e-01
-1.01293957e+00 -1.43363750e+00 -4.92095947e-01 5.89887142e-01
3.15946178e-03 -5.37622929e-01 -2.87011862e-01 -1.03718340e-01]
|
[12.401700019836426, 7.569725036621094]
|
32c954f3-685a-4320-9a14-b672a22ce000
|
robust-learning-protocol-for-federated-tumor
|
2212.08290
| null |
https://arxiv.org/abs/2212.08290v1
|
https://arxiv.org/pdf/2212.08290v1.pdf
|
Robust Learning Protocol for Federated Tumor Segmentation Challenge
|
In this work, we devise robust and efficient learning protocols for orchestrating a Federated Learning (FL) process for the Federated Tumor Segmentation Challenge (FeTS 2022). Enabling FL for FeTS setup is challenging mainly due to data heterogeneity among collaborators and communication cost of training. To tackle these challenges, we propose Robust Learning Protocol (RoLePRO) which is a combination of server-side adaptive optimisation (e.g., server-side Adam) and judicious parameter (weights) aggregation schemes (e.g., adaptive weighted aggregation). RoLePRO takes a two-phase approach, where the first phase consists of vanilla Federated Averaging, while the second phase consists of a judicious aggregation scheme that uses a sophisticated reweighting, all in the presence of an adaptive optimisation algorithm at the server. We draw insights from extensive experimentation to tune learning rates for the two phases.
|
['Stefano Braghin', 'Jonathan P. Epperlein', 'Swanand Kadhe', 'Giulio Zizzo', 'Ambrish Rawat']
|
2022-12-16
| null | null | null | null |
['tumor-segmentation']
|
['computer-vision']
|
[ 2.28511482e-01 2.59077605e-02 8.35667625e-02 -2.47858018e-01
-1.17722058e+00 -6.26920164e-01 7.17111290e-01 4.92681623e-01
-7.09063649e-01 6.44882143e-01 1.31199628e-01 -6.90657139e-01
-2.95416862e-01 -6.16020024e-01 -4.49898094e-01 -1.15347171e+00
-1.55776367e-01 7.19536602e-01 3.79997879e-01 1.63922414e-01
-2.00485643e-02 4.96269345e-01 -9.63154674e-01 2.53071547e-01
5.84073305e-01 1.00838304e+00 -9.85111147e-02 1.29351473e+00
-3.73475820e-01 9.67479289e-01 -5.71087301e-01 -4.88393396e-01
3.83732200e-01 -1.64883435e-01 -1.02811289e+00 8.25430378e-02
-1.07852906e-01 -4.77258042e-02 3.62036526e-01 6.59520209e-01
8.98390532e-01 1.29807368e-01 3.85221429e-02 -1.40058863e+00
4.91852880e-01 8.37607861e-01 -3.28938007e-01 8.38923976e-02
-6.78758174e-02 6.65096343e-01 8.20544124e-01 -4.40947115e-01
6.92978084e-01 5.51790774e-01 7.32910991e-01 5.85662425e-01
-1.17558718e+00 -3.96767259e-01 -3.41721736e-02 7.88080990e-02
-9.48654532e-01 -6.42892480e-01 5.44725895e-01 -3.21541280e-01
5.36441803e-01 6.82426274e-01 6.34827554e-01 8.71922076e-01
-5.36323190e-02 8.68490696e-01 1.26831925e+00 -3.92639816e-01
9.70259845e-01 1.04608215e-01 -1.80786159e-02 4.31187540e-01
7.03812018e-03 -3.91317844e-01 -5.44454455e-01 -7.62876093e-01
3.17540288e-01 4.36124951e-02 -1.29324511e-01 -4.23249871e-01
-1.16094244e+00 6.36763036e-01 9.15431157e-02 2.80136973e-01
-8.65112722e-01 2.47811660e-01 6.52502537e-01 4.33000535e-01
5.03355801e-01 3.18430513e-01 -7.37889051e-01 -3.59708995e-01
-1.28161693e+00 1.52954295e-01 1.11887777e+00 3.07243377e-01
6.49184287e-01 -5.44326544e-01 -3.99612546e-01 5.80068231e-01
3.19460064e-01 -1.55986175e-01 5.51236689e-01 -1.11091554e+00
1.97241277e-01 5.26905417e-01 -7.13818446e-02 -2.04405949e-01
-5.64220011e-01 -4.34049338e-01 -8.11353981e-01 2.51895249e-01
7.80274868e-01 -6.83278024e-01 -6.05761826e-01 1.74420464e+00
1.16236997e+00 6.64814532e-01 8.25835913e-02 9.32466626e-01
4.57639635e-01 7.10416585e-02 3.88000399e-01 -1.99102759e-01
1.48535991e+00 -1.22598338e+00 -3.70172143e-01 4.86098409e-01
9.93113518e-01 -6.49724305e-01 4.89768535e-01 2.32485533e-01
-1.27853525e+00 2.51182288e-01 -8.78149211e-01 2.18401358e-01
-4.51316744e-01 -3.91013861e-01 6.40107512e-01 6.48966908e-01
-1.25530100e+00 8.22863698e-01 -1.09464848e+00 -4.25200164e-01
5.88308156e-01 6.18622363e-01 -3.62809867e-01 1.53249443e-01
-7.55372941e-01 3.30586761e-01 5.12164943e-02 -2.69925416e-01
-8.58285367e-01 -1.03057587e+00 -2.32860297e-01 -5.44552766e-02
5.27999997e-01 -1.07721829e+00 1.35140145e+00 -1.07372630e+00
-1.72099209e+00 6.64710104e-01 2.66403824e-01 -7.28269815e-01
1.04328823e+00 2.77625859e-01 9.29489434e-02 2.62720466e-01
-3.79650086e-01 6.64476305e-02 6.98639512e-01 -9.12498415e-01
-7.83914506e-01 -4.90762502e-01 1.32770941e-01 1.07938148e-01
-3.67893249e-01 6.49379417e-02 -4.13938433e-01 -1.57474190e-01
-2.70104766e-01 -9.41741705e-01 -7.11288035e-01 -7.77295753e-02
-4.41505849e-01 -2.39009589e-01 8.06717336e-01 -5.21037519e-01
1.06105900e+00 -1.84107018e+00 8.56277347e-02 5.62630713e-01
6.49500847e-01 4.04194027e-01 -2.46206090e-01 4.96308297e-01
1.34346887e-01 -7.60432482e-02 -1.67488441e-01 -8.03445101e-01
-3.39675546e-02 2.07971424e-01 1.63743645e-01 4.66991931e-01
-7.39562884e-02 6.62253320e-01 -1.01759601e+00 -7.41920352e-01
-1.41707003e-01 3.58150661e-01 -5.00282884e-01 3.74592513e-01
-6.35423183e-01 5.88459313e-01 -8.29201698e-01 6.90569699e-01
4.74927723e-01 -5.07483840e-01 6.09974682e-01 -8.21954831e-02
-2.20538408e-01 1.85818985e-01 -1.18657935e+00 1.71083307e+00
-5.32140255e-01 -7.57623091e-02 8.08917463e-01 -7.19238102e-01
2.32385486e-01 6.21170700e-01 1.14658856e+00 -7.03737438e-02
4.60424632e-01 3.30339998e-01 -2.75415003e-01 -3.80489171e-01
2.37039164e-01 2.39253700e-01 2.29442015e-01 1.16887605e+00
2.91616768e-01 3.01900804e-01 4.57744859e-02 4.96343821e-01
1.99145043e+00 -1.92254428e-02 2.76439399e-01 1.77524537e-02
6.22592866e-01 1.03691369e-01 5.74236989e-01 8.80865633e-01
-4.47515190e-01 2.16625333e-01 6.30116642e-01 -6.95462704e-01
-7.67557025e-01 -4.82969612e-01 3.69657546e-01 1.28648579e+00
-3.27993959e-01 -5.90791643e-01 -1.04587328e+00 -1.29311860e+00
-1.52317956e-01 2.48875558e-01 -6.18285954e-01 1.29627541e-01
-3.39025348e-01 -1.12418020e+00 6.06335402e-01 2.08415270e-01
3.89791012e-01 -7.74678409e-01 -9.63451624e-01 4.38538730e-01
-1.31139066e-02 -9.31950629e-01 -4.88387883e-01 5.18220305e-01
-6.74228132e-01 -1.28594232e+00 -6.10760689e-01 2.97714304e-02
5.44944823e-01 1.47671644e-02 9.99550223e-01 3.24564964e-01
-3.24020624e-01 6.50735259e-01 -3.91009510e-01 -3.53581369e-01
-3.37089747e-01 4.13817734e-01 -4.51250970e-01 4.67456013e-01
1.06975734e-01 -7.00990617e-01 -1.01157153e+00 6.05545454e-02
-1.11991286e+00 1.56446114e-01 7.84437239e-01 9.04640019e-01
8.20070505e-01 -1.65072724e-01 2.72455931e-01 -1.38876879e+00
5.97902894e-01 -8.29700768e-01 -6.30412519e-01 5.14256358e-01
-7.77396977e-01 -2.61044316e-02 7.19382465e-01 -4.15349782e-01
-8.44116449e-01 5.36384284e-01 -1.80552170e-01 -3.12831849e-01
-1.16235346e-01 3.05770874e-01 3.34129087e-03 -4.98482615e-01
5.77024698e-01 1.29229099e-01 2.23268524e-01 -4.87793893e-01
4.18920100e-01 8.38059902e-01 7.57911056e-02 -6.00676894e-01
6.52543783e-01 4.41862941e-01 4.44941409e-02 -3.62054199e-01
-6.61128402e-01 -4.82393771e-01 -2.59689033e-01 -2.00293228e-01
6.37437761e-01 -6.36575520e-01 -9.86446679e-01 4.38348740e-01
-7.97364116e-01 -8.24921966e-01 -6.13203228e-01 2.78899550e-01
-6.12827420e-01 7.31661543e-03 -7.83214033e-01 -5.35939753e-01
-1.05961800e+00 -1.12068355e+00 9.22367454e-01 2.54242122e-01
-1.71398506e-01 -9.63697731e-01 4.58811790e-01 8.89227867e-01
1.00225711e+00 4.97389913e-01 3.81040275e-01 -1.39220679e+00
-5.53321481e-01 -2.40134448e-01 -1.36072049e-02 -1.25341401e-01
-3.01688105e-01 1.08931353e-02 -1.08806992e+00 -4.31413919e-01
-1.76468432e-01 -3.15395176e-01 4.98412222e-01 -7.41265118e-02
1.22708094e+00 -6.31734312e-01 -2.67787457e-01 7.61752248e-01
1.34041572e+00 -4.04075503e-01 1.32138774e-01 3.69937092e-01
3.89229029e-01 2.81666934e-01 1.55293509e-01 8.20815802e-01
6.25695229e-01 5.86588323e-01 5.16448081e-01 -1.61186561e-01
-9.06632021e-02 2.94110924e-01 3.27555165e-02 6.67190969e-01
-2.62143314e-01 5.11385538e-02 -9.34799254e-01 3.76217961e-01
-2.05033875e+00 -7.43589818e-01 2.50812292e-01 2.09318304e+00
1.07041168e+00 -1.94196671e-01 4.91017729e-01 1.22029848e-01
2.74770081e-01 1.54563129e-01 -7.03956664e-01 -3.40755165e-01
1.67081311e-01 3.13457876e-01 6.68102801e-01 8.10006186e-02
-8.02916586e-01 6.77157938e-01 4.98367071e+00 1.02534425e+00
-1.30819738e+00 6.71226680e-01 5.17599404e-01 -3.79403383e-01
-2.81219810e-01 1.32557169e-01 -2.61864781e-01 5.63363731e-01
1.27886343e+00 -4.70372811e-02 6.71052217e-01 7.02903390e-01
7.56548671e-03 -1.30403833e-02 -5.57104051e-01 5.78569174e-01
-5.43542087e-01 -1.76483822e+00 -3.62078607e-01 2.29439884e-01
5.75324416e-01 6.23455465e-01 -3.90282929e-01 -1.94965571e-01
1.04607666e+00 -4.50529367e-01 3.96898657e-01 5.16280949e-01
4.05563593e-01 -7.00770795e-01 7.20137119e-01 3.72212797e-01
-9.37436283e-01 -2.66385972e-01 2.24505380e-01 6.41195297e-01
7.35714883e-02 9.91488576e-01 -8.41269314e-01 6.76918924e-01
5.13875902e-01 7.18351826e-02 -2.72025645e-01 1.14461017e+00
1.98673859e-01 9.37241673e-01 -4.50223982e-01 2.30902255e-01
3.01603824e-01 3.26804779e-02 6.82174325e-01 1.24448001e+00
-9.56903026e-02 2.67584901e-02 1.60067454e-01 2.46309564e-01
-3.96444321e-01 3.38453710e-01 1.84336007e-01 2.70771533e-02
5.47374189e-01 2.14925504e+00 -6.37023330e-01 -3.39334905e-01
-2.50955045e-01 8.17782223e-01 5.84265053e-01 1.58872046e-02
-5.19788623e-01 -1.77795455e-01 6.15202308e-01 1.10763028e-01
1.60790533e-01 3.06105753e-03 -7.48123601e-02 -8.48804593e-01
-2.35258594e-01 -8.86671245e-01 9.53657329e-01 -6.26009405e-02
-1.23731482e+00 5.06649554e-01 -6.52600765e-01 -7.32540190e-01
-4.16289717e-02 1.69144288e-01 -1.01580346e+00 4.70389605e-01
-1.66151500e+00 -1.33926320e+00 -2.29597121e-01 9.33595657e-01
5.44559695e-02 -2.57774070e-02 9.16299701e-01 2.72733778e-01
-9.04981077e-01 8.21190357e-01 -2.45758686e-02 -1.63143009e-01
5.86202681e-01 -1.43387187e+00 -6.12580441e-02 5.47497272e-01
-2.37285972e-01 2.06751019e-01 5.92563510e-01 -3.19722444e-01
-1.76880920e+00 -1.28655612e+00 8.16568851e-01 -8.39437246e-02
8.24083269e-01 -2.99607575e-01 -5.97695827e-01 4.21660542e-01
8.75222683e-02 5.92381656e-01 1.20459402e+00 1.39754349e-02
-3.68156582e-02 -4.28786576e-01 -1.59138036e+00 5.13774753e-01
5.38785398e-01 -1.34016857e-01 3.64836812e-01 6.86204255e-01
6.44940257e-01 -4.37877953e-01 -1.25988233e+00 5.82014397e-02
5.11240661e-01 -7.41722643e-01 5.35290360e-01 -7.00055838e-01
-1.30504623e-01 -2.98849314e-01 2.44693235e-02 -1.17775476e+00
3.42280269e-02 -1.46558166e+00 -4.26697940e-01 1.20007455e+00
3.40985328e-01 -8.04406822e-01 9.35127079e-01 9.16966498e-01
1.52366653e-01 -1.18166995e+00 -1.21053672e+00 -3.55499357e-01
-2.57885665e-01 -1.99870378e-01 9.26869571e-01 8.27784657e-01
3.05955298e-02 -1.33425564e-01 -1.20674133e-01 4.05844562e-02
8.72819901e-01 3.30337472e-02 9.66878593e-01 -9.90956485e-01
-7.64769197e-01 -3.71542513e-01 -2.46202275e-01 -9.50813815e-02
-1.09430514e-01 -7.18949795e-01 -1.57703698e-01 -1.02777886e+00
1.38938814e-01 -8.41416180e-01 -5.72505355e-01 8.93890560e-01
-1.63854852e-01 -7.01010274e-03 2.42282078e-01 3.21803153e-01
-1.15626967e+00 -3.78200077e-02 8.21012497e-01 1.22785069e-01
-1.84861019e-01 2.87360430e-01 -5.59235811e-01 4.26325202e-01
9.45710897e-01 -6.53059483e-01 -2.09418431e-01 -3.17756623e-01
9.02410820e-02 2.99177378e-01 3.12874079e-01 -9.82424378e-01
9.25949037e-01 -7.55316243e-02 1.25333562e-01 9.27819032e-03
3.04282326e-02 -8.81076694e-01 5.11060059e-01 5.61971605e-01
-4.86831844e-01 -8.44509751e-02 -2.39852428e-01 5.72326422e-01
2.27154762e-01 1.30663037e-01 8.03437293e-01 -2.10455373e-01
1.40610561e-01 6.45065248e-01 -3.02710891e-01 -1.59572568e-02
1.31517160e+00 2.01042846e-01 -2.12457374e-01 -9.01071727e-02
-8.01528335e-01 5.99398255e-01 4.34777439e-01 -2.63054639e-01
-1.23474285e-01 -6.81048632e-01 -7.55575657e-01 -7.78274462e-02
-1.33208066e-01 7.96770155e-02 2.94701487e-01 1.52948260e+00
-2.40029305e-01 -1.86281696e-01 1.93074539e-01 -2.54711449e-01
-1.32746637e+00 2.95795083e-01 5.58672965e-01 -1.12159145e+00
-4.88705873e-01 7.95346200e-01 -7.88205147e-01 -5.26714861e-01
4.60076094e-01 4.36708301e-01 4.43256497e-02 1.36825696e-01
7.67910004e-01 6.28646672e-01 5.54505885e-01 -2.63994724e-01
-3.83171946e-01 3.44930664e-02 -1.68619141e-01 -9.16170850e-02
1.59130061e+00 9.39283520e-02 -1.31317481e-01 -2.55604777e-02
7.89759815e-01 -5.67451902e-02 -1.29686844e+00 -4.85689998e-01
1.55294463e-01 6.65115863e-02 4.13075775e-01 -1.19229496e+00
-1.51811194e+00 2.20836282e-01 4.65824723e-01 1.21894889e-01
1.30456090e+00 -1.74849749e-01 1.16674268e+00 1.56958580e-01
3.07150364e-01 -1.25948763e+00 -5.37516892e-01 -2.98986547e-02
1.18676223e-01 -9.40846741e-01 -1.31417528e-01 4.65831123e-02
-4.61341649e-01 1.06041336e+00 2.30981082e-01 4.37159911e-02
6.93360746e-01 6.15921795e-01 2.95762241e-01 -2.81605899e-01
-1.56845260e+00 -1.68008104e-01 -1.84150234e-01 4.31977212e-02
1.04303055e-01 2.87209153e-01 -4.32548910e-01 6.79687679e-01
2.55254805e-01 3.48759979e-01 9.37260911e-02 1.18624306e+00
-1.08530074e-01 -1.40469074e+00 -3.22109789e-01 4.74334210e-01
-7.89190590e-01 1.54847041e-01 -3.76246631e-01 1.63177431e-01
3.07547778e-01 9.19991255e-01 -3.20520490e-01 -3.24455112e-01
1.52335510e-01 1.17519736e-01 1.72659844e-01 -1.37611553e-01
-1.50564778e+00 -3.25923972e-02 -1.53105659e-02 -1.10050011e+00
-3.25565100e-01 -7.66705394e-01 -1.19098961e+00 -2.53942907e-01
-1.46350801e-01 4.75569934e-01 1.27252448e+00 9.77064788e-01
6.95491672e-01 4.32873130e-01 9.46346581e-01 -7.56737471e-01
-8.36690366e-01 -4.94678229e-01 -3.58772725e-01 2.92974889e-01
2.72078246e-01 -5.74406385e-02 -4.72879797e-01 -3.00168514e-01]
|
[6.071413040161133, 6.401844501495361]
|
c9a26ff6-05a1-4e4c-88fa-ecb3c2adad0b
|
global-attention-decoder-for-chinese-spelling
| null | null |
https://aclanthology.org/2021.findings-acl.122
|
https://aclanthology.org/2021.findings-acl.122.pdf
|
Global Attention Decoder for Chinese Spelling Error Correction
| null |
['Guotong Xie', 'Wei Zhu', 'Keqiang Wang', 'Yuan Ni', 'Zhao Guo']
| null | null | null | null |
findings-acl-2021-8
|
['csc']
|
['natural-language-processing']
|
[-8.63703638e-02 1.71006292e-01 -6.22772932e-01 -4.08054382e-01
-8.41685571e-03 -9.08429027e-01 6.55310392e-01 -6.53472245e-01
-2.85945535e-01 1.06888819e+00 -4.63127941e-02 -1.01159286e+00
-3.91567826e-01 -9.63214397e-01 -4.95059669e-01 -6.31337762e-01
-9.79754329e-01 7.25764990e-01 3.30370307e-01 -6.93831444e-01
7.03166842e-01 7.88774848e-01 -1.68942046e+00 7.18545914e-01
7.04417467e-01 8.52217197e-01 2.49141872e-01 1.14950800e+00
-1.95044339e-01 1.55633950e+00 -7.48382092e-01 -5.46825826e-01
3.13719302e-01 -1.23176083e-01 -7.22945035e-01 -1.01074085e-01
9.28529128e-02 -8.59008506e-02 -2.09758401e-01 9.22211111e-01
5.37373662e-01 4.49454933e-02 1.08379531e+00 -1.42548037e+00
-5.91619551e-01 6.10313773e-01 -4.01565880e-02 1.21627934e-01
1.03678203e+00 -5.39447069e-01 1.19919395e+00 -1.13026452e+00
7.20913768e-01 1.26888943e+00 8.66221786e-01 5.44149756e-01
-1.22286928e+00 -1.94712028e-01 -3.26822817e-01 -9.51717794e-02
-1.46558487e+00 -3.25250506e-01 4.25783843e-02 -2.08119690e-01
1.66093647e+00 1.26596653e+00 1.20609856e+00 1.01401424e+00
1.26658809e+00 8.34431887e-01 1.04267764e+00 -5.13792276e-01
3.35295945e-01 3.66983831e-01 1.54683650e-01 6.33519173e-01
8.40953708e-01 5.26628852e-01 -7.06372619e-01 -9.13127720e-01
9.33553874e-01 -2.94925272e-01 1.71355158e-01 -5.05680561e-01
-9.05919552e-01 6.91228509e-01 1.78732842e-01 3.83959889e-01
-1.39880210e-01 9.89067405e-02 1.26390755e-01 5.30987144e-01
-2.58292928e-02 6.47037446e-01 -9.11868811e-01 -1.33165747e-01
-8.71728659e-01 5.10332465e-01 1.25398111e+00 1.52653182e+00
1.24482810e-01 2.94908643e-01 -9.34252143e-02 3.17179203e-01
8.92314315e-01 1.01808000e+00 4.28362608e-01 -1.36146402e+00
-6.87414408e-02 1.72361732e-01 5.01781464e-01 -8.52631688e-01
-6.33224547e-01 -9.64177120e-03 -8.93263519e-01 4.49267089e-01
3.49161088e-01 4.57367361e-01 -8.02827001e-01 5.07305264e-01
4.33481112e-02 -2.34125629e-01 4.53833073e-01 5.55570945e-02
4.99930978e-01 3.76208365e-01 -1.34477139e-01 -5.73289394e-01
1.06082785e+00 -1.36716676e+00 -1.35299087e+00 2.33215362e-01
9.05734658e-01 -1.07320261e+00 4.35900748e-01 5.33875942e-01
-1.55548143e+00 -1.37560293e-01 -1.08699942e+00 1.78573877e-01
-7.27255583e-01 -3.14239264e-01 8.57801437e-01 1.43120694e+00
-1.60129595e+00 9.73287821e-01 -4.91727620e-01 6.59165755e-02
1.20568443e-02 8.24621081e-01 -2.64718989e-03 4.62812334e-01
-1.33193445e+00 1.08501506e+00 2.22979754e-01 -1.21242590e-01
-1.65216476e-01 -2.13068098e-01 -8.23704481e-01 -5.63443303e-01
-4.78693932e-01 -5.29636025e-01 1.44139910e+00 -2.59346128e-01
-1.65295815e+00 9.71794367e-01 -1.42069459e-01 -1.97814897e-01
6.14786744e-01 -1.28011424e-02 -8.31891418e-01 2.42498964e-01
-1.89849049e-01 5.76383233e-01 9.28263724e-01 -1.35132408e+00
-7.59897232e-01 -1.67359829e-01 -1.23336017e-01 2.66287565e-01
-1.25510961e-01 1.89734384e-01 2.11616129e-01 -1.12999000e-01
3.27147305e-01 -7.20919967e-01 -2.53068686e-01 -5.32041907e-01
-1.46512717e-01 -7.10518599e-01 7.70373225e-01 -4.51523662e-01
1.83705616e+00 -1.67618537e+00 -2.06720144e-01 3.98590982e-01
3.57815564e-01 -1.24705513e-03 2.40583986e-01 1.08380008e+00
-2.76906848e-01 7.32199550e-01 4.11965609e-01 -1.10722095e-01
2.24991128e-01 4.85861301e-01 -4.16602850e-01 3.05609167e-01
-1.29282743e-01 1.15307164e+00 -1.16605783e+00 -5.23096442e-01
4.11106765e-01 9.42391157e-02 -4.58732933e-01 4.79237735e-01
2.99364805e-01 1.47170946e-01 -3.56553018e-01 1.39399457e+00
1.15709066e+00 -1.31984919e-01 1.45911396e-01 5.30878425e-01
-3.79135728e-01 3.55090618e-01 -6.96863770e-01 1.05554795e+00
7.08333924e-02 5.00173986e-01 1.02364108e-01 -7.94621468e-01
3.33247900e-01 8.61540735e-01 4.37155962e-01 -9.67555881e-01
-2.26398129e-02 6.92409754e-01 1.12803578e-01 -6.07703328e-01
7.58228302e-01 3.94563079e-02 -3.64872098e-01 6.40070081e-01
-2.37588286e-01 -6.59476995e-01 9.64643434e-02 3.08100313e-01
6.22585893e-01 -5.18246442e-02 5.62923312e-01 -1.05613089e+00
7.86340594e-01 -1.82965681e-01 -1.81299388e-01 1.03415680e+00
-3.09923887e-01 3.19085121e-01 2.99841821e-01 -6.65102363e-01
-6.45341039e-01 -1.12307119e+00 -4.89381433e-01 1.30636716e+00
3.24267983e-01 -4.39044595e-01 -9.54439282e-01 -2.49762803e-01
1.77620783e-01 6.89606130e-01 -5.90509653e-01 3.84124845e-01
-5.03739953e-01 -8.32535863e-01 7.39044368e-01 3.45434904e-01
-5.07752821e-02 -1.33414865e+00 -6.58416986e-01 1.25490099e-01
-2.20292807e-01 -6.63697243e-01 -6.23428151e-02 4.48765576e-01
-1.35989368e+00 -5.18594682e-01 -6.66252747e-02 -8.20914626e-01
5.87345481e-01 2.46782884e-01 1.27047324e+00 5.39230824e-01
-2.31483161e-01 4.26904231e-01 -1.21292919e-01 -4.95818377e-01
-4.59671497e-01 -8.00336525e-02 5.28869390e-01 -5.87835789e-01
5.19427478e-01 -2.50617653e-01 -7.29350567e-01 5.37953973e-01
-6.88540697e-01 1.62748516e-01 1.79803044e-01 1.04410267e+00
1.35816500e-01 -9.34035778e-02 1.22507080e-01 -6.38007045e-01
8.72274399e-01 -1.69219792e-01 -3.78732830e-01 5.77745810e-02
-6.77108407e-01 -3.74140263e-01 3.21430594e-01 -3.25342178e-01
-1.01981449e+00 -4.87835288e-01 -9.82677937e-02 2.45538145e-01
1.11353043e-02 -1.46784872e-01 6.47139177e-02 -5.24923325e-01
8.02199244e-01 9.25758183e-02 1.99174434e-02 -6.80815242e-03
3.01039815e-01 7.09525108e-01 -6.82967342e-03 -6.68678164e-01
8.44880998e-01 4.91470337e-01 7.98524171e-02 -9.57177758e-01
-1.52186140e-01 -2.60129690e-01 -9.51962709e-01 -6.54426932e-01
6.56643391e-01 -6.78531289e-01 -9.10833478e-01 3.91110867e-01
-9.38691139e-01 -3.38627815e-01 -3.91645581e-01 4.25431967e-01
-1.01278400e+00 2.75717527e-02 -3.90154392e-01 -1.27895141e+00
-5.10977268e-01 -1.02017939e+00 9.43384409e-01 5.30070923e-02
-5.10597289e-01 -1.26927447e+00 5.87685481e-02 2.71537274e-01
1.81734428e-01 -1.73075795e-01 6.90226793e-01 -2.38256708e-01
-4.24233019e-01 -1.53791070e-01 2.34436691e-02 -1.39755070e-01
1.70832314e-02 4.95917559e-01 -9.81751978e-01 -5.31145096e-01
6.65065646e-02 -1.92070693e-01 -1.08835101e-01 6.52520418e-01
5.91872573e-01 -2.29931593e-01 -8.56000841e-01 5.40386558e-01
1.38545322e+00 3.85070026e-01 5.32770038e-01 7.28214979e-01
1.41836226e-01 5.53460240e-01 9.17806149e-01 4.63203549e-01
1.30579369e-02 3.28798652e-01 2.40537539e-01 1.49327129e-01
1.11720070e-01 -1.54819340e-01 3.77893507e-01 1.16112018e+00
-8.18235934e-01 -2.69281328e-01 -5.07867396e-01 4.42987174e-01
-1.72482407e+00 -1.40330648e+00 -4.32368398e-01 6.90478683e-01
6.25676990e-01 1.56016424e-01 -1.48347050e-01 3.35214496e-01
4.99015123e-01 -2.03574806e-01 -1.19133167e-01 -1.06291151e+00
-1.43546045e-01 3.15233678e-01 7.37729073e-01 1.00061214e+00
-7.20721722e-01 1.03317809e+00 1.29781246e+01 1.02230716e+00
2.21112028e-01 1.03134915e-01 5.16071796e-01 3.48020852e-01
-4.36954498e-01 -4.56139445e-02 -1.04416132e+00 2.72933897e-02
1.38140702e+00 -4.30666685e-01 6.85999811e-01 5.44219851e-01
3.44648361e-01 -4.23268199e-01 -1.26188684e+00 5.26221812e-01
9.73738134e-02 -1.40886843e+00 -2.83300440e-04 6.85225725e-01
7.73699820e-01 -5.08050561e-01 6.22419357e-01 3.24184299e-01
6.09259963e-01 -1.14389277e+00 8.60300779e-01 2.53660440e-01
1.03040910e+00 -6.05088234e-01 5.67372203e-01 1.68872893e-01
-1.14389896e+00 -2.20873043e-01 -8.77727985e-01 -1.00755692e+00
3.93533185e-02 -1.81779593e-01 -4.29956943e-01 3.48861217e-01
9.58353162e-01 2.99398601e-01 -3.93658698e-01 9.95779395e-01
-4.78694476e-02 1.04875881e-02 -2.71853864e-01 -4.48467314e-01
4.83122796e-01 -3.54241252e-01 4.66730654e-01 1.00164843e+00
2.48499006e-01 3.51035744e-01 -9.84472036e-02 4.01770771e-01
5.45058846e-01 3.29446048e-02 -1.19659424e+00 -1.78908288e-01
2.83276141e-01 9.16795909e-01 -4.83487815e-01 -4.22520459e-01
-2.00212970e-01 8.62069130e-01 -3.55488248e-02 5.01107454e-01
-6.11489356e-01 -4.35615242e-01 9.72222984e-01 -1.27327025e-01
-1.14700586e-01 -3.48497719e-01 -6.23769283e-01 -7.30352640e-01
-5.89872956e-01 -4.54965204e-01 5.93606755e-02 -5.53365827e-01
-1.39813089e+00 5.79277515e-01 -2.27688253e-02 -1.40553558e+00
-6.99901402e-01 -1.27676582e+00 -4.76714373e-01 4.92853165e-01
-1.11898029e+00 -1.10984349e+00 2.50124663e-01 4.52870727e-01
1.64141744e-01 -5.34416080e-01 1.39563632e+00 3.57715860e-02
1.00637585e-01 9.24474537e-01 6.69434488e-01 -7.37814724e-01
5.56605101e-01 -1.27867436e+00 5.68737745e-01 -1.37897313e-01
-4.31265175e-01 9.05828118e-01 6.28349900e-01 -5.39804697e-01
-1.41196322e+00 -3.66917729e-01 1.08350635e+00 -9.83769417e-01
6.55218959e-01 -3.86345625e-01 4.23767231e-02 7.88592756e-01
7.15902448e-01 -6.18741751e-01 8.21781039e-01 -1.83753878e-01
1.80774391e-01 5.75296998e-01 -1.39248300e+00 6.12354755e-01
1.66275799e+00 -4.63594139e-01 -6.25784039e-01 7.60327101e-01
8.13696027e-01 -6.94087505e-01 -1.30082703e+00 3.34633321e-01
8.65424156e-01 -8.75409484e-01 1.61978090e+00 -1.32660246e+00
-4.43697497e-02 2.81152606e-01 -2.61993498e-01 -9.32519078e-01
-5.96193194e-01 -1.23518765e+00 -5.33532679e-01 -5.83747849e-02
5.96577883e-01 -1.13057327e+00 3.42365682e-01 8.74560475e-01
-2.82833427e-01 -6.42737269e-01 -1.06996536e+00 -1.32016802e+00
-3.35779637e-02 -1.45572275e-01 4.90409225e-01 7.63798356e-01
6.83744550e-01 1.09839931e-01 -6.36873543e-02 -1.00294888e-01
5.33176839e-01 9.55312885e-03 4.41501856e-01 -1.34294486e+00
3.86843324e-01 -5.75816095e-01 -3.07655483e-01 -9.45992947e-01
-8.85957032e-02 -8.12076271e-01 -6.53862000e-01 -1.28511906e+00
-8.32044985e-03 -1.91056758e-01 -1.11109078e-01 -1.65725678e-01
3.67937148e-01 2.13746816e-01 1.20859891e-02 1.03788137e-01
-3.71160030e-01 6.18435517e-02 1.29639816e+00 6.91750320e-05
-1.62315920e-01 4.85058486e-01 -4.81304944e-01 7.84440815e-01
8.58408585e-02 -2.99253196e-01 -6.78878546e-01 6.11881316e-02
6.69384480e-01 4.61409837e-02 3.24159935e-02 -7.42885649e-01
5.37211418e-01 -3.75702560e-01 4.78586555e-01 -1.32223868e+00
1.30741090e-01 -9.61415648e-01 6.95283338e-02 9.40189242e-01
2.74610907e-01 1.20424610e-02 8.66204947e-02 5.39500564e-02
-1.44871444e-01 -5.70943117e-01 9.21121240e-01 -4.07591403e-01
-4.92852688e-01 -5.20386267e-03 -1.03226590e+00 8.97834301e-02
9.92593169e-01 -7.84614205e-01 -3.59281451e-01 -4.20183957e-01
-8.29068601e-01 -1.95836127e-02 6.50830388e-01 3.03609259e-02
7.30431557e-01 -1.51530886e+00 -2.30721906e-01 7.18729138e-01
-3.22939813e-01 -3.74400020e-01 -1.70157343e-01 6.58265352e-01
-1.32361674e+00 1.02442718e+00 -5.41665435e-01 -4.55340147e-01
-1.14228773e+00 4.78126436e-01 4.28307921e-01 -2.41845414e-01
-2.15481281e-01 1.11954463e+00 2.71224789e-02 -8.23025763e-01
1.85185194e-01 -9.21545625e-02 -7.54407048e-01 3.55081353e-03
6.88606799e-01 1.05194807e+00 -2.91290224e-01 -6.04341030e-01
-4.56784427e-01 6.65885091e-01 2.32151806e-01 -2.87484169e-01
9.17833567e-01 -2.19243199e-01 -9.89108324e-01 4.28274393e-01
8.38715494e-01 -1.36269778e-01 -3.69319022e-02 4.16855574e-01
1.25943512e-01 -8.02164078e-01 -4.32406247e-01 -3.55811834e-01
-1.85641110e-01 5.54822803e-01 5.34874737e-01 8.99602413e-01
8.57008278e-01 -3.02566767e-01 8.18335712e-01 9.66778398e-01
5.72402716e-01 -1.68019545e+00 -2.13140488e-01 6.89524531e-01
9.12339568e-01 -9.28350806e-01 5.44190466e-01 -7.27165341e-01
-4.14997995e-01 1.32979155e+00 4.68304873e-01 -1.55325383e-01
1.27306652e+00 5.74917436e-01 1.14069022e-02 -3.70670199e-01
-9.44949508e-01 1.12705544e-01 3.75366658e-01 1.11147714e+00
5.06513238e-01 5.07374525e-01 -9.66500878e-01 3.21953118e-01
-7.28706717e-01 -2.34555230e-01 4.90474731e-01 1.41972518e+00
-6.43810987e-01 -1.20391917e+00 -7.23931909e-01 4.61561680e-01
-5.49773455e-01 -1.16372630e-01 -5.28106689e-01 8.46754074e-01
-5.76629937e-02 1.49448860e+00 -1.97535474e-03 -5.03491640e-01
4.11356747e-01 1.41089618e-01 7.21762300e-01 -1.23501487e-01
-9.04846430e-01 4.15413082e-01 3.84890139e-01 -1.23056793e+00
-8.58632207e-01 -1.05834293e+00 -1.40667629e+00 -1.19437599e+00
-5.12782812e-01 1.89310342e-01 3.83317530e-01 3.90289724e-01
-2.06836104e-01 2.85260603e-02 9.80917513e-01 -1.07949340e+00
-3.90341938e-01 -9.39418614e-01 -1.00026262e+00 -7.84516707e-02
2.89751232e-01 -8.00943017e-01 -7.83523321e-01 2.83909619e-01]
|
[-7.444065570831299, 3.794548273086548]
|
d5e5b2a4-9e3a-4d33-977b-74159a7066f2
|
causal-effect-estimation-with-global
|
2209.08885
| null |
https://arxiv.org/abs/2209.08885v2
|
https://arxiv.org/pdf/2209.08885v2.pdf
|
Causal Effect Estimation with Global Probabilistic Forecasting: A Case Study of the Impact of Covid-19 Lockdowns on Energy Demand
|
The electricity industry is heavily implementing smart grid technologies to improve reliability, availability, security, and efficiency. This implementation needs technological advancements, the development of standards and regulations, as well as testing and planning. Smart grid load forecasting and management are critical for reducing demand volatility and improving the market mechanism that connects generators, distributors, and retailers. During policy implementations or external interventions, it is necessary to analyse the uncertainty of their impact on the electricity demand to enable a more accurate response of the system to fluctuating demand. This paper analyses the uncertainties of external intervention impacts on electricity demand. It implements a framework that combines probabilistic and global forecasting models using a deep learning approach to estimate the causal impact distribution of an intervention. The causal effect is assessed by predicting the counterfactual distribution outcome for the affected instances and then contrasting it to the real outcomes. We consider the impact of Covid-19 lockdowns on energy usage as a case study to evaluate the non-uniform effect of this intervention on the electricity demand distribution. We could show that during the initial lockdowns in Australia and some European countries, there was often a more significant decrease in the troughs than in the peaks, while the mean remained almost unaffected.
|
['Christoph Bergmeir', 'Angela Dieyu Weng', 'Priscila Grecov', 'Ankitha Nandipura Prasanna']
|
2022-09-19
| null | null | null | null |
['load-forecasting']
|
['miscellaneous']
|
[-1.55008659e-01 -2.69218653e-01 3.74078751e-02 -3.24017256e-02
-2.24751770e-01 -7.57799208e-01 8.40977490e-01 4.01994765e-01
4.90059927e-02 9.38302100e-01 5.60794115e-01 -6.46393359e-01
-4.56760675e-01 -1.18452191e+00 -1.50745198e-01 -1.08487463e+00
-2.38180712e-01 6.94194496e-01 -3.95489126e-01 4.03286517e-02
2.02055067e-01 5.02085447e-01 -1.10206819e+00 -8.36749747e-03
6.26036108e-01 9.38125134e-01 6.62611648e-02 2.50327855e-01
1.87904797e-02 3.47530872e-01 -7.88828790e-01 2.47127116e-01
3.48240077e-01 -8.74896497e-02 -3.76425773e-01 -4.07723010e-01
-8.09652627e-01 -6.42199039e-01 1.08490624e-01 1.00014758e+00
7.67532468e-01 -1.04622863e-01 8.64431798e-01 -1.32361770e+00
-2.83228904e-01 1.01371264e+00 -6.70455635e-01 3.25524628e-01
1.51581869e-01 3.69914591e-01 7.57664025e-01 -6.15118109e-02
-1.99997872e-01 9.30357039e-01 4.68245655e-01 -4.69209075e-01
-1.56962156e+00 -7.89301157e-01 -1.12395361e-01 3.40903968e-01
-1.11551762e+00 2.80777868e-02 7.24452972e-01 -6.35037899e-01
1.35574150e+00 7.09500313e-02 7.61500776e-01 6.75370514e-01
6.17896020e-01 1.30339742e-01 1.05245566e+00 -5.14792085e-01
5.11898339e-01 -1.12555735e-01 -3.04058939e-01 -6.17610753e-01
1.88019678e-01 5.70889473e-01 6.59346655e-02 -1.84320018e-01
2.58085802e-02 -1.53128818e-01 -2.83481300e-01 -4.82873945e-03
-6.93928480e-01 8.44798625e-01 2.71759540e-01 6.69327259e-01
-8.83911371e-01 -1.29358107e-02 5.72708249e-01 1.63234118e-02
3.83780032e-01 1.22997060e-01 -7.82885194e-01 -2.82914430e-01
-1.02666497e+00 1.12059839e-01 8.16402793e-01 1.93928450e-01
1.69042557e-01 5.45650303e-01 -8.92704204e-02 2.61129159e-02
2.50196964e-01 9.29320157e-01 4.77256387e-01 -6.17878914e-01
2.02264071e-01 1.30690873e-01 5.82890868e-01 -7.98580348e-01
-7.24820852e-01 -4.10181195e-01 -9.29221153e-01 3.56431425e-01
2.94336170e-01 -5.45361996e-01 -4.12202626e-01 1.47215044e+00
-3.70734073e-02 3.89790162e-02 -1.21918522e-01 2.13932350e-01
-2.80920386e-01 9.67540026e-01 2.87892759e-01 -5.29680192e-01
9.18948710e-01 2.60138899e-01 -9.30039287e-01 3.30119222e-01
7.45932817e-01 -6.86366856e-01 3.19659442e-01 3.45785499e-01
-9.29492414e-01 -2.23033220e-01 -8.55135679e-01 9.38651264e-01
-5.91604054e-01 -2.99747646e-01 9.01719853e-02 6.32601142e-01
-5.93841493e-01 7.74472535e-01 -6.74199343e-01 2.76807323e-02
1.64232537e-01 1.06530420e-01 1.12525582e-01 4.88336414e-01
-1.49929202e+00 1.45252633e+00 6.52551353e-01 2.24756673e-01
-5.41090608e-01 -9.92750585e-01 -6.55484080e-01 7.91916072e-01
6.17259294e-02 -1.52506813e-01 1.26367784e+00 -5.00251651e-01
-1.23172987e+00 -1.38923928e-01 4.65188950e-01 -8.83829832e-01
4.39681888e-01 3.14055830e-01 -6.51401341e-01 -5.10112226e-01
4.04306874e-02 -3.21906894e-01 3.60361934e-01 -9.37363446e-01
-8.58310521e-01 -5.70659697e-01 -5.69047034e-01 -6.17295466e-02
2.14221075e-01 -1.49994388e-01 8.48578155e-01 -4.05662715e-01
-5.72591543e-01 -7.08151996e-01 -1.83160350e-01 -1.31371975e+00
-3.24316442e-01 -4.39097524e-01 9.43529069e-01 -9.27976668e-01
1.19571674e+00 -1.97230387e+00 -3.43558192e-01 6.60560727e-01
-4.65226591e-01 2.53055021e-02 4.37539071e-01 9.95156348e-01
-5.61420739e-01 4.82542999e-02 -2.44878709e-01 3.95333469e-01
5.74917853e-01 2.86920279e-01 -6.73801005e-01 5.92440605e-01
-4.24998477e-02 5.67770779e-01 -7.15855479e-01 6.39862418e-01
8.27926993e-01 2.60410994e-01 1.77694976e-01 1.24903079e-02
-3.30378920e-01 2.54457504e-01 -8.57973844e-02 3.20976898e-02
7.00336516e-01 6.70673549e-02 4.95098412e-01 -1.55820653e-01
-3.58027309e-01 2.39704475e-01 -1.32013893e+00 6.95704103e-01
-6.41898572e-01 4.99924690e-01 -1.54767811e-01 -1.42441154e+00
7.12558150e-01 4.29284960e-01 7.78296351e-01 -1.16492903e+00
3.57035734e-02 2.17344642e-01 3.07620555e-01 -2.68265426e-01
1.06007449e-01 -4.32564586e-01 -1.12048961e-01 9.83142793e-01
-3.85787159e-01 -3.59465182e-01 1.91012681e-01 -6.64371103e-02
6.29537523e-01 -1.05122529e-01 4.42080438e-01 -6.12141609e-01
2.36806512e-01 -4.31910366e-01 4.48117226e-01 1.58410802e-01
1.46228984e-01 -1.23779409e-01 6.95188582e-01 -3.64238113e-01
-9.44248557e-01 -1.00537467e+00 -4.21415597e-01 3.96708310e-01
-4.11312878e-01 2.16991112e-01 -3.93146306e-01 -4.59910423e-01
5.39805174e-01 2.00570321e+00 -4.69874889e-01 -1.81184247e-01
-2.08638608e-01 -1.28558278e+00 -1.04614802e-01 5.33810079e-01
2.26611376e-01 -1.01251686e+00 -1.03539979e+00 5.61585724e-01
1.96013793e-01 -5.66688299e-01 -1.27257705e-01 5.26313484e-01
-2.79677749e-01 -1.00433910e+00 -4.60440099e-01 1.13564238e-01
3.32916766e-01 -5.23378491e-01 1.19326794e+00 -5.33804953e-01
-1.79806389e-02 2.15821996e-01 2.57125288e-01 -7.76279211e-01
-3.58713299e-01 -1.46672323e-01 1.90360501e-01 -5.32331288e-01
3.53189439e-01 -8.18733811e-01 -6.33860946e-01 8.70515183e-02
-7.79815197e-01 -3.46277833e-01 1.70739934e-01 5.81795454e-01
6.09030910e-02 1.06606352e+00 1.30429637e+00 -4.47409511e-01
8.69822383e-01 -6.93522036e-01 -1.22325194e+00 1.48859397e-01
-1.21394253e+00 3.73393968e-02 6.89884603e-01 1.36267707e-01
-1.22656071e+00 -1.45967677e-01 8.77854377e-02 2.48649657e-01
-3.52787942e-01 8.50032926e-01 -1.94146916e-01 7.14138746e-01
5.59602231e-02 8.90232548e-02 -1.99512497e-01 -4.92949754e-01
3.40735465e-01 7.01873839e-01 3.60022128e-01 -3.06281477e-01
7.56801069e-01 1.98771209e-01 8.35301541e-03 -5.03041625e-01
-2.44073808e-01 2.72805877e-02 -4.76468384e-01 -2.38080680e-01
4.21404511e-01 -7.89082110e-01 -9.18214917e-01 5.48740089e-01
-7.47420788e-01 -5.44764042e-01 -5.87533414e-01 5.23348272e-01
-3.92161697e-01 -1.05637707e-01 1.50545780e-02 -9.91837800e-01
-2.72510231e-01 -1.16312253e+00 2.96190858e-01 1.54259726e-01
-2.45578736e-01 -1.25953555e+00 3.60021353e-01 -2.97591984e-01
9.06471193e-01 4.04455423e-01 1.33646703e+00 -7.97633052e-01
-2.24082153e-02 -2.56171972e-01 -7.04851151e-02 4.83870178e-01
6.43937767e-01 1.15851589e-01 -6.71969831e-01 -5.23824632e-01
5.48657328e-02 3.07160795e-01 5.17490096e-02 8.46171618e-01
8.31572235e-01 -5.31531215e-01 -1.18065819e-01 -6.35444932e-03
1.54005730e+00 8.60052884e-01 5.74694037e-01 5.22707462e-01
-5.22575108e-03 5.93386889e-01 1.59806862e-01 9.02368307e-01
4.77939785e-01 4.50408578e-01 5.35532832e-01 2.43080676e-01
5.84556818e-01 1.15287654e-01 2.35339299e-01 3.20314199e-01
2.97538519e-01 -2.93048352e-01 -1.03614974e+00 7.58333385e-01
-1.46588099e+00 -1.18129528e+00 -2.40300409e-02 2.35241723e+00
4.78169888e-01 2.52907306e-01 2.30973199e-01 2.81351447e-01
5.01206994e-01 7.84715861e-02 -5.03750741e-01 -8.99408698e-01
-1.21117584e-01 1.71266153e-01 6.84505761e-01 4.20740664e-01
-5.94398022e-01 -5.29298149e-02 6.20287037e+00 5.35937846e-01
-1.10296285e+00 -2.68176883e-01 1.10907030e+00 2.74065603e-02
-3.91607583e-01 -1.54955447e-01 -3.82200509e-01 9.48566616e-01
1.56780779e+00 -1.10756016e+00 4.86572176e-01 2.64326155e-01
1.07528806e+00 -5.68907320e-01 -8.43359351e-01 2.38920063e-01
-6.08506620e-01 -1.21631384e+00 -3.60930115e-01 2.72489130e-01
9.93009746e-01 2.80598253e-01 -1.60423562e-01 2.05013499e-01
6.93919837e-01 -1.02024150e+00 6.02544427e-01 6.90180421e-01
2.84653276e-01 -1.30814469e+00 1.24505842e+00 3.75744313e-01
-9.34360087e-01 -4.74319458e-01 1.28894538e-01 -3.21953595e-01
6.44245148e-01 9.79679167e-01 -7.80767083e-01 6.87883735e-01
7.57814586e-01 3.15053582e-01 1.55428961e-01 6.90268219e-01
-1.75698608e-01 8.70537400e-01 -6.12176597e-01 3.04361820e-01
1.06562831e-01 -4.62220252e-01 2.60557264e-01 8.16349089e-01
4.30728555e-01 -6.04297519e-02 -1.81651190e-02 7.40375578e-01
1.06784314e-01 -1.95592225e-01 -7.34605014e-01 1.18222390e-03
5.63749373e-01 9.17289853e-01 -4.25574332e-01 -1.78674817e-01
-4.27791655e-01 1.63455889e-01 -5.42649746e-01 5.51540256e-01
-6.03101194e-01 -1.77619189e-01 5.20282626e-01 1.19325332e-01
2.72724658e-01 9.17891860e-02 -5.61908662e-01 -4.22896266e-01
-3.19653749e-01 -7.19761908e-01 4.21397954e-01 -6.43022001e-01
-1.54970276e+00 -2.34515101e-01 3.84696156e-01 -6.27848566e-01
-1.05135381e+00 -5.55168316e-02 -1.16493702e+00 1.60001659e+00
-1.45440698e+00 -5.41373730e-01 5.24362206e-01 2.46536314e-01
1.31768093e-01 -3.33435200e-02 8.60417485e-01 7.00741913e-03
-5.78652442e-01 -1.48840278e-01 8.25026751e-01 -1.33805946e-01
-4.24337164e-02 -1.49423671e+00 1.36653557e-01 7.23055720e-01
-3.05636555e-01 -1.20143659e-01 8.39701414e-01 -7.22994924e-01
-1.00088954e+00 -7.07605183e-01 5.53112507e-01 9.33222696e-02
1.10256004e+00 2.15638146e-01 -7.60367215e-01 6.15530550e-01
9.52143550e-01 -6.34346008e-01 6.08540058e-01 -2.06273809e-01
1.44521788e-01 -2.74073809e-01 -1.41111481e+00 2.91367382e-01
-2.43053377e-01 -4.56621826e-01 -5.77301085e-01 1.86328679e-01
2.43231282e-02 2.84006417e-01 -1.12684655e+00 5.00993252e-01
6.12809956e-01 -8.44566464e-01 4.95997995e-01 -2.82146245e-01
-1.85996607e-01 -2.54876584e-01 -3.61754239e-01 -2.05110645e+00
-3.11414123e-01 -5.84814250e-01 -1.62789673e-01 1.35472429e+00
3.80586535e-01 -1.04257405e+00 1.62206799e-01 8.67789030e-01
4.17932302e-01 -5.49732745e-01 -1.23661256e+00 -6.28756940e-01
5.10071933e-01 -4.53341633e-01 1.46884942e+00 9.57820833e-01
1.17898688e-01 -5.21043204e-02 6.78394437e-02 6.47049725e-01
7.11093843e-01 4.28038061e-01 3.33449602e-01 -8.21902931e-01
1.45340994e-01 -7.44361699e-01 -1.14905506e-01 1.88976169e-01
-1.43726477e-02 -3.70696574e-01 -2.85092533e-01 -1.73871148e+00
-1.60651617e-02 -5.00182025e-02 -6.06679976e-01 4.25497949e-01
1.53113469e-01 -3.46034169e-01 2.48890951e-01 -1.27442092e-01
4.49304610e-01 7.65272796e-01 3.46941411e-01 -2.57399112e-01
-1.76199928e-01 4.24190432e-01 -3.35690320e-01 6.27866626e-01
1.33596575e+00 -2.09184021e-01 -3.77169102e-01 -5.92271462e-02
4.36253399e-01 2.13650659e-01 -3.48075517e-02 -7.88641393e-01
-6.57364726e-02 -3.53122890e-01 5.84538043e-01 -1.06028938e+00
-5.51872373e-01 -1.24003232e+00 9.82016027e-01 8.75780761e-01
-6.67753741e-02 4.36063349e-01 5.27311265e-01 4.34596568e-01
2.34529860e-02 -1.13878429e-01 5.53858280e-01 2.60447506e-02
-2.71825731e-01 2.61735730e-03 -7.82789886e-01 -2.69140333e-01
1.35558307e+00 3.72863114e-01 -3.28526407e-01 -6.06070220e-01
-7.36461401e-01 6.51305854e-01 2.05593288e-01 3.44598800e-01
-1.11791708e-01 -1.05975652e+00 -9.15728450e-01 1.32642105e-01
-5.80073416e-01 -4.14920568e-01 3.28374505e-01 5.41980803e-01
-1.00282542e-01 6.39210165e-01 1.33363664e-01 -3.83381695e-01
-4.79985714e-01 4.57961470e-01 6.34103358e-01 -5.87354898e-01
-3.53386849e-01 -2.83239322e-04 -2.76714899e-02 -1.41346231e-01
5.40704839e-02 -3.38605374e-01 -2.60989487e-01 6.44672096e-01
5.49047112e-01 7.59914339e-01 3.70350003e-01 -4.27976817e-01
-1.91872552e-01 -1.78880710e-02 2.61615485e-01 7.54113570e-02
1.59372556e+00 -3.27453077e-01 -1.95332542e-01 5.89465022e-01
7.11229563e-01 -3.23707581e-01 -1.22743762e+00 2.32501924e-01
4.05396372e-01 -2.64977634e-01 5.32178640e-01 -1.47570312e+00
-1.29945707e+00 5.83882809e-01 8.46037805e-01 1.12185550e+00
1.28116643e+00 -4.47311014e-01 4.71417606e-01 -1.99471265e-01
2.87882090e-01 -1.12986434e+00 -9.14977252e-01 7.99358860e-02
8.74354184e-01 -6.94692850e-01 1.34030074e-01 6.77577853e-01
-3.35188687e-01 8.96099508e-01 -2.19952464e-01 1.54047087e-01
1.17160594e+00 6.14837646e-01 -2.78972685e-01 9.07187117e-04
-7.49209404e-01 2.34854028e-01 -1.60804674e-01 5.67135632e-01
1.11279584e-01 7.85226345e-01 -1.84014082e-01 5.58266103e-01
2.13519223e-02 4.39547896e-02 6.20061159e-01 5.58817685e-01
1.01782300e-01 -7.85490930e-01 -3.02048087e-01 7.21197665e-01
-6.02634966e-01 -9.03157592e-02 3.31384569e-01 7.50066936e-01
1.98670715e-01 8.62330973e-01 4.91250217e-01 1.71537071e-01
4.70569164e-01 2.87033051e-01 -1.10173881e-01 -1.63707495e-01
-3.84154379e-01 1.18935853e-01 -4.73854914e-02 -2.61137187e-01
-2.41193119e-02 -1.10808170e+00 -1.33334589e+00 -5.15908003e-01
-4.79704499e-01 2.87778169e-01 1.03021765e+00 1.27235103e+00
1.99864447e-01 7.14114845e-01 1.21158862e+00 -9.39573884e-01
-1.05802417e+00 -1.12734270e+00 -1.04422784e+00 -4.26659323e-02
1.05372138e-01 -4.38401133e-01 -8.64184082e-01 -4.82527763e-01]
|
[6.089128494262695, 2.8376734256744385]
|
85b780c6-41f7-40cc-ac87-8ba4e4bb5421
|
loop-closure-detection-with-rgb-d-feature
|
1811.09938
| null |
http://arxiv.org/abs/1811.09938v1
|
http://arxiv.org/pdf/1811.09938v1.pdf
|
Loop Closure Detection with RGB-D Feature Pyramid Siamese Networks
|
In visual Simultaneous Localization And Mapping (SLAM), detecting loop
closures has been an important but difficult task. Currently, most solutions
are based on the bag-of-words approach. Yet the possibility of deep neural
network application to this task has not been fully explored due to the lack of
appropriate architecture design and of sufficient training data. In this paper
we demonstrate the applicability of deep neural networks by addressing both
issues. Specifically we show that a feature pyramid Siamese neural network can
achieve state-of-the-art performance on pairwise loop closure detection. The
network is trained and tested on large-scale RGB-D datasets with a novel
automatic loop closure labeling algorithm. Each image pair is labelled by how
much the images overlap, allowing loop closure to be computed directly rather
than by labor intensive manual labeling. We present an algorithm to adopt any
large-scale generic RGB-D dataset for use in training deep loop-closure
networks. We show for the first time that deep neural networks are capable of
detecting loop closures, and we provide a method for generating large-scale
datasets for use in evaluating and training loop closure detectors.
|
['Alexander Mai', 'Joseph Menke', 'Zhang Qianhao', 'Allen Yang']
|
2018-11-25
| null | null | null | null |
['loop-closure-detection']
|
['computer-vision']
|
[ 2.12898150e-01 -2.96919733e-01 5.79610802e-02 -4.59876329e-01
-8.99025917e-01 -6.20803893e-01 6.87332511e-01 4.42850530e-01
-7.60246396e-01 3.82555753e-01 -1.11988530e-01 -5.53076506e-01
4.20254916e-02 -5.39990902e-01 -8.83363366e-01 -1.26959279e-01
-4.15321797e-01 5.69623232e-01 2.36779362e-01 -2.71801293e-01
3.26094210e-01 9.66580689e-01 -1.76477289e+00 -4.07349579e-02
3.96623105e-01 1.09254897e+00 3.13322335e-01 6.29389346e-01
-2.66143978e-02 4.98260468e-01 -4.60000068e-01 3.67783278e-01
6.26786232e-01 -3.92506003e-01 -6.69761419e-01 -1.33037969e-01
9.89360392e-01 -4.29608047e-01 -3.45823020e-01 8.20511758e-01
5.05372763e-01 2.15474844e-01 5.48914015e-01 -1.19244218e+00
-2.75186032e-01 -1.76046357e-01 -4.49922681e-01 1.63348123e-01
4.92668182e-01 -1.62332341e-01 1.05525613e+00 -1.12668121e+00
7.98919022e-01 9.39843833e-01 1.00465345e+00 1.64151445e-01
-1.28688562e+00 -5.05222559e-01 -2.89160579e-01 -1.13306612e-01
-1.50246227e+00 -4.65617567e-01 7.32861340e-01 -4.19962138e-01
1.54952967e+00 5.13682924e-02 9.45706546e-01 6.80542886e-01
2.24613324e-01 7.39809394e-01 9.97426510e-01 -7.94213116e-01
1.03635743e-01 -3.14794004e-01 -1.18756600e-01 1.08041310e+00
2.67898530e-01 2.10408121e-01 -4.96621490e-01 4.89178021e-03
9.02481258e-01 -8.11370313e-02 -2.21371889e-01 -1.16074550e+00
-1.47522676e+00 9.75572765e-01 1.08289838e+00 4.10348862e-01
-5.68618961e-02 6.00300431e-01 3.92275572e-01 1.64271697e-01
2.82082170e-01 6.53953493e-01 -3.29995573e-01 5.36939763e-02
-1.18981004e+00 4.12515730e-01 4.70904619e-01 9.01340783e-01
1.10655177e+00 -5.48647344e-01 3.29670757e-01 5.42706370e-01
3.94234985e-01 2.45263338e-01 2.77179003e-01 -1.02892017e+00
3.04662168e-01 6.41302586e-01 1.12298988e-01 -1.13720787e+00
-8.52378130e-01 -1.55699536e-01 -6.47530377e-01 6.94299519e-01
3.67305726e-01 3.04121166e-01 -1.07608318e+00 1.50627422e+00
9.11212340e-02 -2.66533196e-01 -3.01463567e-02 8.98298442e-01
7.31316984e-01 3.14649433e-01 -3.04052681e-01 2.52592891e-01
1.04285383e+00 -8.51800859e-01 -4.77172077e-01 -7.31188059e-01
9.74375069e-01 -9.11762953e-01 9.02275264e-01 1.38305426e-01
-6.04740500e-01 -6.32681191e-01 -1.53577089e+00 -5.00351846e-01
-6.46344185e-01 3.38757575e-01 8.09680164e-01 2.49702349e-01
-1.26999176e+00 6.46824360e-01 -9.86447334e-01 -1.02576220e+00
2.82851160e-01 4.77659285e-01 -9.92221236e-01 -2.73310598e-02
-9.12285507e-01 1.11878312e+00 5.51236212e-01 3.97819877e-01
-6.00725472e-01 -1.48442769e-02 -1.35908318e+00 -3.62599909e-01
-3.42257544e-02 -5.33123016e-01 1.10889030e+00 -7.62129605e-01
-6.96642160e-01 1.34684002e+00 -1.49828434e-01 -7.06094205e-01
5.85001588e-01 -1.67421252e-01 6.21000119e-02 1.74934044e-01
4.05395657e-01 1.33805060e+00 5.06004155e-01 -1.43340766e+00
-6.99748456e-01 -3.15580904e-01 1.13846986e-02 3.19362819e-01
1.95355967e-01 -5.17281555e-02 -2.77881086e-01 -2.29803488e-01
7.46040225e-01 -1.17076492e+00 -2.18034387e-01 5.32621562e-01
-2.04391822e-01 -1.40842959e-01 8.98359656e-01 -4.80862617e-01
6.67547464e-01 -1.97445846e+00 -3.02851260e-01 2.05732569e-01
1.26852691e-01 2.37188905e-01 -2.12846801e-01 5.87579429e-01
-1.23545490e-01 -9.84834880e-03 -4.15780276e-01 -6.33871496e-01
9.78170545e-04 5.02291203e-01 -1.44382864e-01 9.16791856e-01
1.02432571e-01 8.30847979e-01 -8.67979050e-01 -5.02477944e-01
5.95461071e-01 3.44238549e-01 -3.49904388e-01 2.17199206e-01
-5.69415418e-03 2.80505002e-01 1.76485673e-01 6.56019866e-01
6.18652523e-01 -8.54048729e-02 -1.07474945e-01 -1.70103237e-01
-4.08424556e-01 2.63350487e-01 -1.10078609e+00 2.44292283e+00
-4.91653383e-01 1.34885573e+00 -1.93567485e-01 -7.53487587e-01
1.14020896e+00 -2.18509331e-01 2.97915816e-01 -8.86194468e-01
7.22125694e-02 6.36151552e-01 -1.42553255e-01 -2.85691589e-01
7.04212785e-01 5.47820702e-02 -2.79866099e-01 4.77648824e-01
1.51848182e-01 -3.95395964e-01 2.22449020e-01 9.63525623e-02
1.13183987e+00 3.95683825e-01 4.60914910e-01 -2.53889173e-01
1.62551373e-01 4.36838984e-01 7.14840516e-02 9.23283756e-01
-6.50406539e-01 8.54413688e-01 1.70041665e-01 -9.41590846e-01
-1.09263539e+00 -9.38868344e-01 -1.39452338e-01 7.72441149e-01
4.29537833e-01 -3.04708481e-01 -4.89012301e-01 -5.05011022e-01
2.89710820e-01 2.14731380e-01 -7.72951901e-01 8.01948234e-02
-5.20354509e-01 -9.23050046e-02 7.15405166e-01 4.96303767e-01
8.14891398e-01 -1.07621050e+00 -1.19909680e+00 2.89387200e-02
-2.57215530e-01 -1.00470710e+00 -1.30617723e-01 6.65362120e-01
-7.19688654e-01 -1.25223863e+00 -4.95461494e-01 -1.15692425e+00
6.72325253e-01 5.04397213e-01 9.68668401e-01 2.23781820e-02
-6.75868750e-01 2.20979765e-01 -1.60472661e-01 -2.44426042e-01
-2.67648548e-01 7.17516523e-03 1.94129914e-01 -6.44058824e-01
5.29651880e-01 -4.51340795e-01 -6.29911900e-01 1.44673195e-02
-6.29497170e-01 -7.71976262e-02 7.38231599e-01 7.34114051e-01
7.00517833e-01 -4.27630395e-01 -1.13355316e-01 -1.52826548e-01
6.14637375e-01 2.07917944e-01 -7.39070952e-01 1.00023687e-01
-4.33647364e-01 1.08506717e-01 -8.02902877e-02 8.56850576e-03
-3.79413426e-01 6.53151572e-01 -3.14537644e-01 -4.99746382e-01
-3.91192496e-01 4.28069323e-01 4.24700022e-01 -8.49722147e-01
1.01344538e+00 6.41936660e-02 2.19906017e-01 -2.03491241e-01
4.64452267e-01 6.04316294e-01 6.53087735e-01 -1.73366040e-01
5.54171264e-01 5.95817387e-01 2.77476907e-01 -8.21994662e-01
-5.76511979e-01 -8.37088585e-01 -1.04687333e+00 -3.71276997e-02
8.68694127e-01 -1.05482876e+00 -5.49809575e-01 3.69800687e-01
-1.43539107e+00 -3.14976692e-01 -1.09048799e-01 4.37516630e-01
-7.46954143e-01 4.29277807e-01 -3.62982810e-01 -5.68817496e-01
-1.04689747e-01 -1.10036635e+00 1.25828493e+00 -3.32810059e-02
-2.46357396e-01 -7.90174127e-01 4.99876410e-01 5.85421845e-02
2.89390415e-01 4.78518099e-01 4.15297717e-01 -4.64620560e-01
-6.99648440e-01 -5.77425122e-01 -5.02058983e-01 2.55667776e-01
2.43698824e-02 -1.53757662e-01 -9.84222889e-01 -3.09861839e-01
-3.37477773e-01 -5.77582836e-01 9.96163845e-01 2.26635978e-01
8.31657350e-01 2.93296099e-01 -4.37060505e-01 7.67477691e-01
1.63425875e+00 -2.26365000e-01 5.21543264e-01 6.73214078e-01
7.12083578e-01 5.70384085e-01 5.94097674e-01 1.16908543e-01
3.17542583e-01 6.19266748e-01 6.62715852e-01 -3.86197358e-01
1.12979310e-02 -4.41172928e-01 -1.06454805e-01 2.33070925e-01
2.24124640e-01 1.44880246e-02 -1.38444459e+00 7.77835786e-01
-1.92170799e+00 -5.99601328e-01 -8.30730647e-02 2.05778432e+00
3.36156547e-01 1.78751469e-01 -1.80436373e-01 1.39927521e-01
4.84778315e-01 1.71737060e-01 -1.19193308e-01 -4.26945716e-01
-7.01145455e-02 7.13908896e-02 8.58193934e-01 6.38824046e-01
-1.55270934e+00 1.03327560e+00 6.11904049e+00 3.92665982e-01
-9.43419158e-01 -1.07825875e-01 7.55451620e-02 2.27888808e-01
1.02591030e-01 2.25108430e-01 -5.39700031e-01 -2.98213750e-01
3.67254794e-01 5.05490363e-01 8.34205225e-02 7.84054518e-01
9.91906077e-02 -7.72670805e-01 -1.14023602e+00 1.34053743e+00
1.78281710e-01 -1.27479315e+00 -2.54996300e-01 8.48548040e-02
6.47895336e-01 5.58769584e-01 -4.35556769e-01 7.13505745e-02
-3.16892415e-02 -9.29880023e-01 7.16466069e-01 1.84099644e-01
8.12176764e-01 -6.24471009e-01 7.24355996e-01 2.01795429e-01
-1.34554327e+00 3.51229087e-02 -3.81379008e-01 -2.54489779e-01
-4.92692739e-02 3.01333785e-01 -8.72592866e-01 1.82603613e-01
7.40430057e-01 6.18185163e-01 -8.48936439e-01 1.25982249e+00
-2.73048997e-01 -1.81595728e-01 -7.61330366e-01 -6.97986856e-02
5.74716449e-01 3.07361539e-02 1.48221597e-01 1.29684794e+00
4.38152224e-01 -4.67394263e-01 3.18106294e-01 8.77101660e-01
2.08066449e-01 9.68431979e-02 -1.30547047e+00 7.08034039e-02
3.63364697e-01 1.14640260e+00 -1.03124154e+00 8.42807218e-02
-3.51739347e-01 1.15797830e+00 6.90004408e-01 2.26784408e-01
-4.59746063e-01 -7.62458861e-01 7.05346942e-01 1.81984752e-02
2.27743447e-01 -7.67062008e-01 -1.68338090e-01 -8.90432239e-01
6.23781346e-02 -3.37342709e-01 1.07862718e-01 -9.54581201e-01
-8.15216064e-01 5.69992423e-01 -1.42534196e-01 -1.28915226e+00
-4.29590583e-01 -7.93600500e-01 -3.49598318e-01 8.57669711e-01
-1.46717656e+00 -1.34290671e+00 -7.04697847e-01 5.07905781e-01
8.17514732e-02 1.77919030e-01 1.07676053e+00 1.11083828e-01
3.81462686e-02 2.29781389e-01 -9.08149220e-03 3.97673428e-01
8.02842200e-01 -1.19214308e+00 7.66990483e-01 1.01079178e+00
5.96011043e-01 8.04260373e-01 9.16015446e-01 -4.62248325e-01
-1.13369620e+00 -8.37556362e-01 1.28357124e+00 -4.31185842e-01
5.96634448e-01 -8.24291587e-01 -5.16355932e-01 8.64041626e-01
7.86628351e-02 4.94410008e-01 4.64673311e-01 -4.70163636e-02
-3.88767719e-01 -1.82351773e-03 -1.00508487e+00 3.70766133e-01
1.30863595e+00 -9.41383660e-01 -7.71034598e-01 4.73915607e-01
4.90868777e-01 -7.89276659e-01 -4.50600386e-01 4.88465667e-01
4.78741407e-01 -1.50221395e+00 9.31339443e-01 -4.69245277e-02
2.14919031e-01 -5.31842828e-01 -3.37786287e-01 -1.05408072e+00
-4.86971065e-02 -2.40229681e-01 1.97575539e-01 6.75044715e-01
1.79613680e-01 -5.82480431e-01 8.23886037e-01 6.78780973e-02
-2.32882544e-01 -5.37124693e-01 -1.05248868e+00 -6.86241210e-01
-2.56268978e-01 -4.94987577e-01 1.21939845e-01 6.47448957e-01
-1.81449667e-01 7.56922141e-02 -5.72859943e-02 1.39760152e-01
5.91401398e-01 4.05098855e-01 1.00430179e+00 -1.08565748e+00
2.02805936e-01 -4.13753390e-01 -9.40552115e-01 -1.16956306e+00
1.48719355e-01 -9.05315936e-01 5.47595143e-01 -2.02645278e+00
-2.43177846e-01 -5.02848625e-01 7.88211226e-02 5.42359948e-01
4.84860063e-01 7.68483460e-01 8.90542343e-02 3.09651792e-01
-6.32641315e-01 4.16561514e-01 7.71772861e-01 -5.20682074e-02
-9.44777131e-02 -3.39913726e-01 -1.15007408e-01 5.90859294e-01
6.98363602e-01 -3.10573667e-01 -2.48276100e-01 -3.50448906e-01
3.51294249e-01 -2.74187177e-01 7.02780366e-01 -1.48553121e+00
3.79282087e-01 4.13689762e-01 5.66226125e-01 -8.44425499e-01
3.75677437e-01 -9.17623758e-01 -8.71779919e-02 7.23951280e-01
-2.56929547e-01 2.56607741e-01 4.84656394e-01 5.69546342e-01
-3.60961169e-01 -8.80529508e-02 5.82841694e-01 -3.16770881e-01
-1.36807978e+00 2.12612301e-01 -2.96597749e-01 -2.68586874e-01
9.70687270e-01 -3.91351283e-01 1.78083871e-02 -3.12391579e-01
-5.80071211e-01 1.33681595e-01 8.56793225e-01 5.30461788e-01
8.28368723e-01 -1.42121351e+00 -2.62713343e-01 6.61483824e-01
8.30970824e-01 1.79166615e-01 -1.86218128e-01 6.99438989e-01
-1.35728776e+00 6.17036283e-01 -3.54034990e-01 -9.81019676e-01
-1.22803998e+00 6.26874387e-01 6.76698029e-01 9.29330196e-03
-5.81665993e-01 1.01995492e+00 -1.09623544e-01 -8.08196127e-01
3.68468076e-01 -5.36021173e-01 1.46372199e-01 7.06372708e-02
2.13158801e-01 -1.27343789e-01 3.12507123e-01 -8.52703631e-01
-6.83781922e-01 8.46743524e-01 3.44270885e-01 -1.51648521e-01
9.85814035e-01 -1.33431956e-01 -3.02227587e-01 3.99786592e-01
1.55612755e+00 -4.58861232e-01 -1.21822917e+00 2.60455962e-02
-1.34443520e-02 -4.42816377e-01 1.33704469e-02 -4.22277063e-01
-5.44310391e-01 1.11572456e+00 1.03114188e+00 -5.19013442e-02
8.02172661e-01 -3.71239968e-02 5.68875134e-01 8.82590652e-01
7.13337302e-01 -1.05668068e+00 2.40776595e-02 7.36244380e-01
9.03640568e-01 -1.76728284e+00 6.64110435e-03 7.64851198e-02
-8.97356495e-02 1.31419194e+00 3.59653115e-01 -4.65985149e-01
5.24937034e-01 2.07191393e-01 4.50121701e-01 -3.35434496e-01
-1.11472949e-01 -5.92493117e-01 3.01922143e-01 5.74745893e-01
3.94322366e-01 -8.91934186e-02 -2.05064699e-01 -4.82266128e-01
-2.99618971e-02 -6.96591511e-02 2.46468872e-01 1.31310284e+00
-7.52908170e-01 -1.01096904e+00 -3.90018821e-01 2.79484868e-01
-4.32557836e-02 -4.97112274e-02 -5.98060906e-01 1.17322338e+00
3.53867263e-01 6.94452167e-01 2.38602161e-01 -3.88855934e-01
2.29109347e-01 -1.29597738e-01 8.61573935e-01 -4.42406446e-01
-3.30542177e-01 -2.38180593e-01 4.91127111e-02 -8.32104087e-01
-5.93949616e-01 -6.27572596e-01 -1.33972621e+00 -9.74928960e-02
-2.56085604e-01 -1.50241360e-01 1.08981216e+00 8.97250533e-01
2.71762997e-01 4.29666117e-02 2.04042092e-01 -1.48649192e+00
-1.46214530e-01 -8.49743903e-01 -4.95234698e-01 4.50754702e-01
7.63555825e-01 -7.70245194e-01 -4.02872801e-01 -4.17349100e-01]
|
[7.563068866729736, -2.085418462753296]
|
577c12ec-4c72-4ae4-9b14-c8795188fd05
|
augmenting-autotelic-agents-with-large
|
2305.12487
| null |
https://arxiv.org/abs/2305.12487v1
|
https://arxiv.org/pdf/2305.12487v1.pdf
|
Augmenting Autotelic Agents with Large Language Models
|
Humans learn to master open-ended repertoires of skills by imagining and practicing their own goals. This autotelic learning process, literally the pursuit of self-generated (auto) goals (telos), becomes more and more open-ended as the goals become more diverse, abstract and creative. The resulting exploration of the space of possible skills is supported by an inter-individual exploration: goal representations are culturally evolved and transmitted across individuals, in particular using language. Current artificial agents mostly rely on predefined goal representations corresponding to goal spaces that are either bounded (e.g. list of instructions), or unbounded (e.g. the space of possible visual inputs) but are rarely endowed with the ability to reshape their goal representations, to form new abstractions or to imagine creative goals. In this paper, we introduce a language model augmented autotelic agent (LMA3) that leverages a pretrained language model (LM) to support the representation, generation and learning of diverse, abstract, human-relevant goals. The LM is used as an imperfect model of human cultural transmission; an attempt to capture aspects of humans' common-sense, intuitive physics and overall interests. Specifically, it supports three key components of the autotelic architecture: 1)~a relabeler that describes the goals achieved in the agent's trajectories, 2)~a goal generator that suggests new high-level goals along with their decomposition into subgoals the agent already masters, and 3)~reward functions for each of these goals. Without relying on any hand-coded goal representations, reward functions or curriculum, we show that LMA3 agents learn to master a large diversity of skills in a task-agnostic text-based environment.
|
['Marc-Alexandre Côté', 'Xingdi Yuan', 'Pierre-Yves Oudeyer', 'Laetitia Teodorescu', 'Cédric Colas']
|
2023-05-21
| null | null | null | null |
['common-sense-reasoning']
|
['reasoning']
|
[ 8.83644521e-02 3.83383274e-01 4.02015030e-01 -5.94953075e-03
-1.34001166e-01 -7.60429978e-01 9.74174500e-01 8.71373340e-02
-2.34163314e-01 8.08313251e-01 3.62611920e-01 -1.19856857e-01
-3.27081949e-01 -9.05954540e-01 -6.00017250e-01 -4.04746950e-01
-2.37885639e-01 8.16240609e-01 -3.20759654e-01 -8.64026487e-01
2.19890907e-01 2.40689516e-01 -1.89495182e+00 1.71703491e-02
1.18893647e+00 2.76987046e-01 5.22184968e-01 7.24038243e-01
-2.10790053e-01 1.12333071e+00 -5.99992931e-01 -7.44413286e-02
1.34196609e-01 -9.20489669e-01 -9.57506835e-01 2.61021435e-01
-2.92510982e-03 5.96257821e-02 -1.76921710e-01 9.68231261e-01
1.46543384e-02 5.00334561e-01 6.88445508e-01 -1.17328942e+00
-1.10568166e+00 8.26564491e-01 1.47689283e-01 -2.20999062e-01
6.89401686e-01 7.45275617e-01 8.07816446e-01 -3.56475651e-01
7.47469783e-01 1.32888544e+00 3.33521038e-01 1.08709896e+00
-1.41605616e+00 -4.50580448e-01 1.96508184e-01 -2.66560912e-01
-1.19100702e+00 -7.95480683e-02 6.89478636e-01 -6.89562321e-01
1.03988302e+00 9.13985893e-02 1.41740668e+00 1.32677245e+00
3.75932992e-01 4.76864964e-01 1.47852004e+00 -5.34817636e-01
6.25245333e-01 2.89079577e-01 -1.95614755e-01 9.27886903e-01
2.21948504e-01 6.94275618e-01 -9.79895175e-01 4.49544564e-02
1.15572608e+00 -2.61850446e-01 -1.98256671e-01 -4.52236384e-01
-1.30601132e+00 7.36388147e-01 3.90994012e-01 4.14136589e-01
-6.39073253e-01 3.95420045e-01 -1.05150163e-01 7.99469054e-01
-4.83970493e-02 1.40169930e+00 -3.10874194e-01 -1.31335825e-01
-4.63210970e-01 5.64438105e-01 5.95181346e-01 8.45821440e-01
9.48991358e-01 5.02510011e-01 -2.18679272e-02 5.37899494e-01
3.33790958e-01 5.73393703e-01 7.81250238e-01 -1.26848984e+00
-8.12637731e-02 8.31989586e-01 1.48863927e-01 -6.17219031e-01
-4.34181750e-01 -8.15189838e-01 -2.52886027e-01 7.64444470e-01
2.43525892e-01 -3.39723229e-01 -9.91527021e-01 2.37781024e+00
1.27729386e-01 -8.13947618e-02 5.40149093e-01 7.73143649e-01
5.31804502e-01 7.92114794e-01 3.46037000e-01 2.07429808e-02
9.78272319e-01 -6.98757052e-01 -2.68859953e-01 -6.51558220e-01
7.04203069e-01 -1.51731849e-01 1.45692086e+00 4.73806292e-01
-1.45016229e+00 -5.46651661e-01 -9.55873668e-01 1.48988917e-01
-5.44289887e-01 -5.91517806e-01 1.00560164e+00 5.48383534e-01
-1.52858961e+00 6.42117023e-01 -5.21961510e-01 -6.07935667e-01
1.63769290e-01 2.07240775e-01 -2.33423516e-01 1.52966365e-01
-1.13716257e+00 1.02485549e+00 7.39550531e-01 -6.69846833e-01
-1.34769273e+00 -5.34588397e-01 -9.92035866e-01 1.03318818e-01
3.66769314e-01 -1.32558763e+00 1.18498313e+00 -1.27827096e+00
-1.72597671e+00 9.91530895e-01 4.32764500e-01 -2.33301774e-01
8.31009150e-02 1.58602104e-01 -2.40245461e-01 -3.43835466e-02
4.98512574e-02 1.07796395e+00 7.03573465e-01 -1.48523724e+00
-5.36852777e-01 -1.80111423e-01 4.60106224e-01 7.86337078e-01
-2.70434916e-01 -2.14083776e-01 4.98851612e-02 -7.98792720e-01
-1.29626259e-01 -1.00067556e+00 -4.19831663e-01 -2.92429864e-01
-9.13804173e-02 -2.19190136e-01 -5.07421046e-02 -2.89258897e-01
9.66828525e-01 -1.93293560e+00 8.22900236e-01 3.28694105e-01
4.66694921e-01 -1.59893051e-01 -4.27260280e-01 7.39123881e-01
1.52806073e-01 1.69419006e-01 -3.37419286e-02 -1.27170011e-01
3.03680152e-01 2.55201578e-01 -1.64063275e-01 -1.75952271e-01
3.60643095e-03 1.11459768e+00 -1.28842795e+00 -3.18288535e-01
5.85406870e-02 1.47073433e-01 -7.60656893e-01 3.47536981e-01
-7.46011913e-01 8.15985382e-01 -6.35332584e-01 2.29074597e-01
-6.75520748e-02 -3.86212230e-01 2.69008547e-01 7.22867012e-01
-1.60063595e-01 2.07362220e-01 -8.46743584e-01 1.96936584e+00
-6.16193473e-01 2.62744188e-01 -1.50313318e-01 -3.90825212e-01
1.06530952e+00 1.36007383e-01 2.42866889e-01 -5.39030015e-01
2.46587828e-01 1.45867452e-01 3.26183468e-01 -3.64605933e-01
4.22084510e-01 -4.53606576e-01 -2.92957425e-01 8.39102209e-01
2.39735350e-01 -7.71340668e-01 3.38693887e-01 3.93787473e-01
1.00319767e+00 5.61779380e-01 3.41490895e-01 -6.38320088e-01
3.17538887e-01 3.99164587e-01 3.08245152e-01 9.86153960e-01
1.79570451e-01 1.84529945e-01 1.32101759e-01 -4.13553417e-01
-8.93871665e-01 -1.45053732e+00 5.65494776e-01 1.30727518e+00
-3.19154561e-02 -3.34230006e-01 -6.72930837e-01 -9.55768451e-02
-2.43356094e-01 1.28628612e+00 -7.24833906e-01 -4.64357138e-01
-5.09870946e-01 3.32690775e-02 3.18147570e-01 2.39609703e-01
4.22422349e-01 -1.82055736e+00 -1.31646800e+00 3.37822139e-01
2.97914911e-02 -3.58299047e-01 -2.91657150e-01 2.92305171e-01
-7.32719660e-01 -7.86949754e-01 -4.74239826e-01 -8.98838937e-01
7.18613446e-01 -8.22708290e-03 1.46937966e+00 3.46362561e-01
-2.52285451e-01 1.05269396e+00 -3.72133017e-01 -6.14406943e-01
-7.24082947e-01 -4.21562850e-01 2.89271235e-01 -5.39086819e-01
1.04297914e-01 -8.76110554e-01 -3.22512448e-01 -1.69148892e-01
-8.72538328e-01 5.78717530e-01 5.18125653e-01 8.43061090e-01
1.97511524e-01 4.64162156e-02 6.25470340e-01 -4.85832959e-01
1.24081302e+00 -5.61183929e-01 -3.87544990e-01 3.33731264e-01
-5.28595150e-01 4.16066438e-01 7.41504788e-01 -6.35313213e-01
-1.17161727e+00 -3.30091238e-01 2.78852731e-01 -1.57073006e-01
-2.96187073e-01 7.45330036e-01 2.26249754e-01 -2.54381876e-02
1.17557430e+00 7.43369162e-01 2.56493360e-01 -1.46249181e-03
7.06677675e-01 8.36582184e-02 5.81474125e-01 -1.25267792e+00
9.11461532e-01 -1.37817487e-01 -1.36347353e-01 -7.49904394e-01
-5.35665691e-01 2.06513584e-01 -2.13147581e-01 -4.79423255e-01
9.37841296e-01 -7.49829948e-01 -8.99218559e-01 2.94744819e-01
-8.44533503e-01 -8.79071832e-01 -9.54501271e-01 5.75815916e-01
-1.11344111e+00 -2.82024220e-02 -4.51097071e-01 -7.94898808e-01
-1.64089024e-01 -1.06034422e+00 5.51729620e-01 5.68010032e-01
-7.65757740e-01 -1.11707020e+00 1.31850913e-01 7.66780674e-02
5.88248074e-01 3.70808095e-01 1.32501101e+00 -5.48609495e-01
-5.94079792e-01 3.05373192e-01 4.78972971e-01 -1.05934821e-01
-3.45051549e-02 -5.14049232e-01 -3.32312882e-01 -4.01055574e-01
1.25762984e-01 -1.14392126e+00 2.88865834e-01 9.91504490e-02
7.34465241e-01 -3.63244236e-01 3.59343016e-03 4.97284114e-01
1.14476466e+00 5.12246430e-01 4.09550190e-01 4.42797780e-01
4.03279252e-02 9.34702337e-01 3.51684660e-01 4.89587367e-01
5.67348540e-01 2.67453045e-01 3.06533992e-01 2.48321474e-01
-1.27404585e-01 -5.17667115e-01 5.09498954e-01 6.52533650e-01
-3.39063972e-01 -2.80499905e-01 -1.01791704e+00 5.22379041e-01
-1.65370643e+00 -1.00740492e+00 5.29122949e-01 2.13862085e+00
1.17588162e+00 9.22074020e-02 4.20925133e-02 -4.07068342e-01
2.49306858e-01 -8.84431675e-02 -8.97431016e-01 -6.60115957e-01
-2.14510038e-02 4.84160006e-01 -2.42478535e-01 5.86920917e-01
-3.40522468e-01 1.41823590e+00 6.86262178e+00 4.08000499e-01
-7.62193561e-01 -2.29415208e-01 4.16514486e-01 -2.14782476e-01
-7.33291388e-01 -1.68120474e-01 -2.58930475e-01 6.45083115e-02
6.51018262e-01 -6.64674401e-01 1.29122174e+00 5.08438766e-01
-7.67496228e-02 7.95169994e-02 -1.24795759e+00 8.26398075e-01
-4.30951332e-04 -1.34548855e+00 3.60873699e-01 1.56615064e-01
8.08867872e-01 -3.49256516e-01 4.37088490e-01 8.15098226e-01
1.38024068e+00 -1.49158013e+00 1.04184747e+00 6.17111325e-01
7.84043431e-01 -6.66202545e-01 2.11255942e-02 7.00198948e-01
-7.85012662e-01 -3.59916151e-01 -1.35040835e-01 -4.84727979e-01
1.22477086e-02 -4.88371164e-01 -5.65513074e-01 1.58296004e-01
3.92291546e-01 2.49943003e-01 -4.25463527e-01 3.71982485e-01
-4.34219360e-01 2.40193978e-01 -1.38578430e-01 -4.71906543e-01
4.76622880e-01 -3.77764761e-01 7.39417017e-01 6.91035211e-01
6.29057229e-01 5.60491562e-01 4.05845404e-01 1.36913180e+00
3.22145760e-01 -3.52172996e-03 -8.06693256e-01 -5.55802226e-01
3.78601074e-01 8.18133295e-01 -4.65434074e-01 -2.39515543e-01
-9.69863012e-02 7.29609787e-01 4.65365976e-01 5.67893922e-01
-4.95661080e-01 1.11862589e-02 8.26809227e-01 9.24933925e-02
-2.80789107e-01 -4.90496218e-01 -2.64619797e-01 -1.00891554e+00
-7.50203490e-01 -1.36563694e+00 1.97024882e-01 -1.12774670e+00
-1.14905536e+00 6.81990266e-01 1.10738859e-01 -6.17356658e-01
-6.55835867e-01 -4.88880843e-01 -6.69926584e-01 1.09935188e+00
-8.58509600e-01 -1.28075516e+00 -2.47753009e-01 8.28894913e-01
6.41736031e-01 -7.31109023e-01 9.79673088e-01 -7.92253017e-01
-7.04453560e-03 2.92906821e-01 -3.46907794e-01 -3.49040985e-01
1.29100725e-01 -1.36950338e+00 3.66961420e-01 5.37168682e-01
1.73675671e-01 1.09166086e+00 1.01820898e+00 -9.08699989e-01
-1.32127297e+00 -6.50215089e-01 6.02219462e-01 -4.52346146e-01
7.25876212e-01 -3.56915057e-01 -6.37922525e-01 8.09758306e-01
4.37566847e-01 -8.23553920e-01 7.33453274e-01 1.95404246e-01
-1.27419844e-01 4.71132487e-01 -1.10723865e+00 1.24471724e+00
1.51143718e+00 -2.84878224e-01 -1.04144728e+00 3.51901770e-01
9.65898037e-01 -1.88651204e-01 -6.25763118e-01 -1.65303618e-01
3.94332200e-01 -9.53387856e-01 1.05878580e+00 -1.01309729e+00
5.61926723e-01 -1.78684950e-01 -8.08290094e-02 -1.81838548e+00
-9.20781851e-01 -8.83879244e-01 1.20431237e-01 7.17869401e-01
2.18292966e-01 -7.83139706e-01 5.96211791e-01 4.70952600e-01
-4.80285138e-01 -6.31710529e-01 -5.68414450e-01 -7.40592182e-01
3.77715468e-01 -2.91383505e-01 8.07170510e-01 9.91558135e-01
4.96218830e-01 3.79679769e-01 -1.44567877e-01 -2.27421746e-01
6.22937143e-01 8.03419948e-02 7.27507710e-01 -1.35984457e+00
-6.73472643e-01 -8.08506906e-01 -9.23227146e-02 -8.41091096e-01
1.33268327e-01 -1.11650550e+00 3.21507826e-02 -1.49152899e+00
1.66362303e-03 -6.01341307e-01 -1.83436513e-01 4.14878726e-01
1.02517396e-01 -3.26378703e-01 4.29625213e-01 1.63597330e-01
-5.30776918e-01 6.46591902e-01 1.46705186e+00 1.55634955e-01
-5.55310786e-01 -4.54181910e-01 -1.23909688e+00 9.14118350e-01
1.02404058e+00 -1.68317512e-01 -9.75281894e-01 -3.83427590e-01
6.89338386e-01 1.98609143e-01 3.26924771e-01 -1.29070973e+00
1.77847803e-01 -8.75999749e-01 3.51231545e-01 5.64024150e-02
3.39534819e-01 -5.27700722e-01 4.21177983e-01 7.37566352e-01
-6.51231349e-01 3.23426485e-01 1.69841170e-01 5.80208957e-01
3.02738965e-01 -1.33874625e-01 5.57253838e-01 -8.38389635e-01
-8.45207870e-01 -1.25520945e-01 -7.88427591e-01 1.34751573e-01
1.24511480e+00 -5.15843570e-01 -3.93308312e-01 -6.34261847e-01
-9.55533028e-01 5.50042391e-01 8.05455685e-01 5.18755376e-01
7.80708432e-01 -1.20593154e+00 -8.45274150e-01 1.43548474e-01
3.09230417e-01 -1.08491756e-01 1.10867478e-01 -6.19559474e-02
-5.57551861e-01 2.88945913e-01 -7.25933313e-01 -2.09381729e-01
-6.10115349e-01 6.85469508e-01 3.47929239e-01 -4.18837517e-01
-5.89021742e-01 1.06011403e+00 7.14690030e-01 -5.17456234e-01
-1.45877808e-01 -8.12214613e-03 -4.78471518e-01 -5.25022924e-01
3.42864156e-01 7.21775144e-02 -8.24907422e-01 -4.39138800e-01
6.36299923e-02 4.10930872e-01 1.04106322e-01 -5.55020869e-01
1.22572815e+00 -7.77502507e-02 -6.95145279e-02 5.02662599e-01
2.14195400e-01 -2.72004992e-01 -1.31159151e+00 -1.41941130e-01
-3.65382314e-01 -2.51190662e-01 -3.17914635e-01 -1.28160357e+00
-3.43352616e-01 5.90073586e-01 1.89894065e-01 9.24171358e-02
1.00681722e+00 1.31215081e-01 3.61312389e-01 5.20060301e-01
9.42457438e-01 -1.13098288e+00 9.32046056e-01 7.23052382e-01
1.28929234e+00 -6.46545053e-01 -2.98483759e-01 7.28351399e-02
-9.83802319e-01 9.73262370e-01 9.75987673e-01 4.28130329e-02
-6.02532085e-03 -3.09923980e-02 2.13858727e-02 -5.42154372e-01
-1.06885421e+00 -3.87250185e-01 1.54958934e-01 1.15351748e+00
3.28520387e-01 1.53565288e-01 -2.01588675e-01 4.10137266e-01
-8.87192667e-01 -8.94256234e-02 5.78797162e-01 1.01466787e+00
-7.98057437e-01 -9.26149368e-01 -3.83234173e-01 2.10134283e-01
2.62151510e-01 -1.25195965e-01 -5.05386293e-01 6.78961694e-01
4.08504635e-01 1.02433968e+00 -8.60261843e-02 -3.62437189e-01
-6.47792146e-02 2.08669439e-01 8.41411769e-01 -9.97869968e-01
-8.28140199e-01 -4.50914830e-01 -3.92745137e-02 -3.66431564e-01
1.11520551e-02 -6.98524833e-01 -1.49354494e+00 -3.95565182e-01
4.23296690e-01 1.29249483e-01 3.54428887e-01 7.17046142e-01
1.83133617e-01 6.88344538e-01 1.42472520e-01 -6.65666878e-01
-6.23882174e-01 -6.61570191e-01 -5.32046080e-01 6.86645806e-01
-9.07887742e-02 -6.54171109e-01 -4.97126728e-02 8.22190866e-02]
|
[4.20241117477417, 1.3791775703430176]
|
fe158c80-f377-40e7-bab6-5347a5792458
|
rebooting-acgan-auxiliary-classifier-gans
|
2111.01118
| null |
https://arxiv.org/abs/2111.01118v1
|
https://arxiv.org/pdf/2111.01118v1.pdf
|
Rebooting ACGAN: Auxiliary Classifier GANs with Stable Training
|
Conditional Generative Adversarial Networks (cGAN) generate realistic images by incorporating class information into GAN. While one of the most popular cGANs is an auxiliary classifier GAN with softmax cross-entropy loss (ACGAN), it is widely known that training ACGAN is challenging as the number of classes in the dataset increases. ACGAN also tends to generate easily classifiable samples with a lack of diversity. In this paper, we introduce two cures for ACGAN. First, we identify that gradient exploding in the classifier can cause an undesirable collapse in early training, and projecting input vectors onto a unit hypersphere can resolve the problem. Second, we propose the Data-to-Data Cross-Entropy loss (D2D-CE) to exploit relational information in the class-labeled dataset. On this foundation, we propose the Rebooted Auxiliary Classifier Generative Adversarial Network (ReACGAN). The experimental results show that ReACGAN achieves state-of-the-art generation results on CIFAR10, Tiny-ImageNet, CUB200, and ImageNet datasets. We also verify that ReACGAN benefits from differentiable augmentations and that D2D-CE harmonizes with StyleGAN2 architecture. Model weights and a software package that provides implementations of representative cGANs and all experiments in our paper are available at https://github.com/POSTECH-CVLab/PyTorch-StudioGAN.
|
['Jaesik Park', 'Minsu Cho', 'Woohyeon Shim', 'Minguk Kang']
|
2021-11-01
| null |
http://proceedings.neurips.cc/paper/2021/hash/c5ab6cebaca97f7171139e4d414ff5a6-Abstract.html
|
http://proceedings.neurips.cc/paper/2021/file/c5ab6cebaca97f7171139e4d414ff5a6-Paper.pdf
|
neurips-2021-12
|
['conditional-image-generation']
|
['computer-vision']
|
[ 1.16004199e-01 3.61706376e-01 1.02688223e-01 -2.28643164e-01
-6.12265229e-01 -5.15788019e-01 6.89458251e-01 -8.41655552e-01
-8.08421001e-02 9.83023763e-01 8.65447521e-02 -1.81083545e-01
2.38520786e-01 -1.10953736e+00 -9.08124983e-01 -8.71346533e-01
2.42154121e-01 3.26808512e-01 -2.93794960e-01 -3.24514478e-01
-2.84179986e-01 3.56938243e-01 -1.12125099e+00 1.47916481e-01
1.06781960e+00 1.04886067e+00 -1.25977054e-01 6.20258331e-01
2.63100326e-01 9.29038823e-01 -8.21831286e-01 -8.60936344e-01
6.69166625e-01 -9.41915751e-01 -6.93483353e-01 -2.08516449e-01
2.35127658e-01 -2.68240809e-01 -4.98273164e-01 1.07996941e+00
6.16121829e-01 9.29097906e-02 7.13175952e-01 -1.74414670e+00
-1.06964147e+00 6.69011354e-01 -3.53171080e-01 -1.36333168e-01
-2.28652433e-01 4.15514052e-01 5.64546943e-01 -8.61561537e-01
5.82919359e-01 1.32260251e+00 6.38136864e-01 9.69011426e-01
-9.62564826e-01 -1.08069289e+00 -1.24279307e-02 -6.03808574e-02
-1.26677513e+00 -1.77725613e-01 9.57636416e-01 -3.28463197e-01
3.76385272e-01 3.98555160e-01 7.25644648e-01 1.56337678e+00
1.01790711e-01 8.31339240e-01 1.15155947e+00 -1.74151808e-01
1.28251463e-01 1.52980372e-01 -3.51833522e-01 6.31788492e-01
1.13128640e-01 4.98573422e-01 -1.31872684e-01 1.99347928e-01
1.01025712e+00 5.42294122e-02 -4.29203510e-01 -1.25232622e-01
-8.90984774e-01 1.07145524e+00 7.97550440e-01 -2.25273035e-02
-2.44399533e-01 2.71814018e-01 2.15473861e-01 4.24818814e-01
6.04258955e-01 4.30209368e-01 -9.49020684e-02 -3.31803262e-02
-5.18916607e-01 1.92266151e-01 6.53320849e-01 1.20517921e+00
7.02310681e-01 6.65501595e-01 -2.82908827e-01 7.66736925e-01
1.00264117e-01 5.60638607e-01 6.79573774e-01 -8.12600195e-01
2.69835234e-01 4.34981972e-01 -5.39347231e-01 -7.84119189e-01
9.81795788e-02 -7.24854946e-01 -1.42002869e+00 4.64003384e-01
1.74347863e-01 -5.63490987e-01 -1.24994278e+00 1.93475652e+00
9.58946794e-02 2.56158769e-01 3.37966472e-01 8.23913455e-01
9.82388258e-01 6.31866932e-01 -2.03742739e-02 2.07230031e-01
8.60028982e-01 -1.19186842e+00 -6.39562786e-01 -3.56369793e-01
1.98560402e-01 -5.49623430e-01 1.01332545e+00 1.72688693e-01
-1.03036106e+00 -8.01241755e-01 -1.06800365e+00 -1.14786094e-02
-3.94105792e-01 1.61124140e-01 6.30926967e-01 5.70311189e-01
-8.97250950e-01 6.35042548e-01 -8.70155334e-01 6.73250034e-02
8.24343324e-01 1.26650423e-01 -2.66195774e-01 -8.55763927e-02
-1.34569669e+00 7.31768012e-01 4.80287373e-01 1.33853987e-01
-1.25846183e+00 -9.16624606e-01 -9.09726381e-01 -5.46301492e-02
4.49464507e-02 -7.63118088e-01 8.66837919e-01 -1.36716914e+00
-1.65561593e+00 5.74314892e-01 3.21824461e-01 -5.32196283e-01
9.97528195e-01 -1.77548096e-01 -4.11905557e-01 -6.88477904e-02
-8.21518600e-02 1.07279086e+00 9.26860929e-01 -1.28712797e+00
-1.28792793e-01 -4.29115184e-02 -5.47059476e-02 1.29323319e-01
-2.39221618e-01 -4.25105602e-01 -1.39357373e-01 -1.09750736e+00
-2.88391322e-01 -1.05970085e+00 -2.02461585e-01 -2.31727466e-01
-7.46282041e-01 -2.22920720e-02 1.04518628e+00 -6.18644238e-01
8.48266423e-01 -2.10522532e+00 2.51987278e-01 5.49170338e-02
2.10035980e-01 4.41439360e-01 -2.91241288e-01 2.04817221e-01
-3.70693296e-01 4.36773151e-01 -5.05262673e-01 -4.08232778e-01
-2.13362854e-02 2.06453428e-01 -6.23282135e-01 2.44606882e-01
4.18101341e-01 1.25865448e+00 -8.07613194e-01 -1.76071167e-01
2.17316031e-01 8.76863539e-01 -5.55989981e-01 2.61011869e-01
-1.11285262e-01 6.22427404e-01 -2.41886437e-01 6.55577421e-01
8.01771045e-01 -1.66544527e-01 -1.79705799e-01 -1.38194561e-01
3.92244667e-01 -4.73732725e-02 -8.03434134e-01 1.52443433e+00
-3.27258021e-01 6.30981028e-01 -3.64511997e-01 -9.01491761e-01
9.23341513e-01 1.87115982e-01 2.17826478e-02 -4.23229843e-01
1.43162027e-01 6.17376976e-02 4.61192876e-02 9.45185721e-02
2.15973020e-01 -1.07721306e-01 3.02561484e-02 1.73654720e-01
3.17852706e-01 -2.85309851e-01 9.91228819e-02 1.17778957e-01
7.41669595e-01 2.76156455e-01 3.45341675e-02 -1.47528276e-01
2.83238471e-01 -2.37712502e-01 7.12336957e-01 5.00687599e-01
6.18424453e-02 9.71165180e-01 5.75703979e-01 -3.55505943e-01
-1.14231825e+00 -1.05661690e+00 -9.37873945e-02 5.25197446e-01
-1.21208556e-01 -2.41836444e-01 -1.07772815e+00 -8.38254452e-01
-1.88413680e-01 9.02961195e-01 -7.99953461e-01 -4.52267081e-01
-4.04661059e-01 -9.79448736e-01 7.86652207e-01 7.13557005e-01
1.12476528e+00 -1.20695925e+00 -1.07939668e-01 -9.24830288e-02
-2.49614362e-02 -9.35345829e-01 -4.41111267e-01 -1.35292783e-01
-6.77527726e-01 -1.01812577e+00 -8.95304799e-01 -7.04322755e-01
9.72214520e-01 -2.63421804e-01 1.21459234e+00 -7.54116103e-02
-2.35526934e-01 1.54379055e-01 -4.30656970e-01 -7.20123470e-01
-6.72668397e-01 1.78396538e-01 -1.59066796e-01 -1.46627292e-01
1.98617592e-01 -7.23912597e-01 -7.40077853e-01 3.20630759e-01
-8.40520561e-01 3.75188529e-01 5.43912709e-01 1.04290462e+00
7.54904985e-01 5.57951927e-02 6.92799449e-01 -1.03975642e+00
5.42347908e-01 -6.00674450e-01 -6.18586123e-01 2.57004239e-02
-7.95469761e-01 -1.22842394e-01 9.41766202e-01 -5.17177403e-01
-1.05815625e+00 -1.08588099e-01 -4.32210267e-01 -9.62755620e-01
-5.13528548e-02 3.37120354e-01 -4.05774057e-01 1.18093275e-01
6.98852479e-01 3.84927005e-01 9.42158625e-02 -1.79309472e-01
4.47452784e-01 3.77068907e-01 5.30894637e-01 -4.74346787e-01
1.10022402e+00 4.23912138e-01 -1.29757011e-02 -5.94552636e-01
-8.33942831e-01 2.61432678e-01 -2.38242239e-01 -6.15595654e-02
8.65602732e-01 -1.06568885e+00 -4.37130779e-01 8.49104404e-01
-8.22125137e-01 -7.77501285e-01 -6.81317389e-01 2.81155646e-01
-5.58535695e-01 -8.63804743e-02 -5.64587891e-01 -5.28311014e-01
-6.78950429e-01 -1.12053931e+00 6.25996530e-01 3.61324161e-01
1.44496769e-01 -1.16294098e+00 -3.09749711e-02 2.77276039e-01
5.13093054e-01 8.82791042e-01 4.75360662e-01 -5.37801564e-01
-6.11874223e-01 -2.15113387e-01 -1.16393408e-02 1.03397655e+00
1.91475585e-01 -2.54952386e-02 -1.19157434e+00 -4.36956882e-01
5.67292273e-02 -5.50593257e-01 9.44852352e-01 3.52136135e-01
1.57627058e+00 -6.12403572e-01 -5.81237338e-02 1.18954015e+00
1.33204901e+00 3.32758307e-01 1.13001704e+00 2.20193435e-02
1.06540358e+00 1.98370099e-01 2.85856575e-01 1.16148911e-01
1.53543308e-01 2.58342743e-01 6.67015374e-01 -4.84954268e-01
-5.46749473e-01 -5.48843861e-01 3.81097317e-01 7.66201854e-01
-2.38342881e-01 -5.67435443e-01 -6.92745090e-01 4.06388044e-01
-1.52896237e+00 -1.01788616e+00 1.12671003e-01 1.90105200e+00
9.52909708e-01 6.41462728e-02 -1.14634320e-01 5.14575057e-02
6.59489453e-01 1.09999739e-01 -7.51273274e-01 -7.72593096e-02
-4.10609007e-01 4.97060508e-01 5.66574395e-01 5.20317793e-01
-1.03523040e+00 9.10747409e-01 5.20906544e+00 9.02356029e-01
-1.22758818e+00 1.95786387e-01 1.25706112e+00 -3.61485668e-02
-2.73867995e-01 -1.57337889e-01 -7.14699686e-01 7.70653009e-01
6.58449650e-01 -2.22483665e-01 6.10581756e-01 1.07408440e+00
-4.42610502e-01 4.99444395e-01 -1.10452759e+00 8.90398800e-01
1.70690343e-01 -1.37654376e+00 2.03173891e-01 1.60577565e-01
1.06910884e+00 1.40968800e-01 4.41864312e-01 5.16492784e-01
7.69747555e-01 -1.36925769e+00 5.48431098e-01 4.10557270e-01
1.31063879e+00 -9.77346778e-01 7.90341198e-01 8.73265564e-02
-8.60389173e-01 8.49088877e-02 -3.93811166e-01 2.25242406e-01
-1.37154952e-01 5.29763997e-01 -6.66552782e-01 6.16181374e-01
6.38684869e-01 7.81287968e-01 -5.55732250e-01 6.37297809e-01
-6.74800396e-01 7.91935325e-01 -1.38435230e-01 3.20873290e-01
1.93207800e-01 -4.53297168e-01 6.03326499e-01 8.93566251e-01
4.97279257e-01 9.44968760e-02 -1.64233774e-01 1.21419752e+00
-5.21420002e-01 -4.14893538e-01 -6.07914925e-01 -1.19861931e-01
4.44391727e-01 1.21027422e+00 -4.46054250e-01 -2.90976465e-01
-4.48952354e-02 1.13541067e+00 1.26326054e-01 4.91788208e-01
-9.75285470e-01 -4.87381697e-01 7.76894808e-01 -2.91140657e-02
1.80023953e-01 2.99368709e-01 -2.16183588e-01 -1.23144782e+00
-3.58462706e-02 -1.02679121e+00 2.56460607e-01 -8.15621138e-01
-1.44057274e+00 9.36167359e-01 -1.80193648e-01 -1.28335428e+00
-3.88128579e-01 -5.07799864e-01 -9.58828688e-01 9.74697888e-01
-1.49009013e+00 -1.34757221e+00 -6.89213336e-01 6.63902164e-01
5.30446351e-01 -4.75056767e-01 8.32243860e-01 3.52959424e-01
-6.38645291e-01 1.03057003e+00 4.79408354e-02 6.70598269e-01
5.48570633e-01 -1.37501490e+00 6.65013194e-01 1.01635981e+00
-2.24624593e-02 3.23304087e-01 4.74316388e-01 -6.24261737e-01
-1.10996413e+00 -1.60491943e+00 2.41680503e-01 -4.07694131e-01
2.55994201e-01 -5.34788013e-01 -7.31301606e-01 1.04907835e+00
4.80237931e-01 3.01380366e-01 5.55863380e-01 -3.96844089e-01
-2.61231989e-01 -2.29359254e-01 -1.20169437e+00 5.70624590e-01
1.12611639e+00 -2.33808547e-01 -6.32373989e-02 4.32154953e-01
8.96051526e-01 -7.61015475e-01 -9.13808346e-01 5.34005940e-01
2.60762900e-01 -8.75655770e-01 9.70316112e-01 -5.16044319e-01
9.13090527e-01 -1.66325510e-01 1.48337055e-02 -1.80104637e+00
-1.62842527e-01 -7.54079640e-01 -6.03965782e-02 1.39387512e+00
3.08082908e-01 -9.22861874e-01 7.63066530e-01 2.47574612e-01
-3.84862095e-01 -9.00965333e-01 -6.83201849e-01 -8.88869166e-01
5.57252526e-01 -1.08568445e-01 8.79451931e-01 1.08621907e+00
-5.62345564e-01 2.44246766e-01 -4.96862352e-01 -9.89629179e-02
7.13744104e-01 -1.48967519e-01 9.32465613e-01 -9.65001464e-01
-2.42147028e-01 -4.47522432e-01 -3.31983864e-01 -7.00840235e-01
2.35514432e-01 -1.06626606e+00 -2.04116628e-01 -1.31534100e+00
-8.97158310e-02 -6.94435954e-01 -2.25624189e-01 6.57508194e-01
-2.48822987e-01 5.13596773e-01 3.56847495e-01 1.00416467e-01
3.54986228e-02 8.51716995e-01 1.56723261e+00 -2.27832988e-01
1.25116324e-02 -2.48956122e-02 -8.46386313e-01 6.28031552e-01
1.07400811e+00 -4.04413790e-01 -8.00013661e-01 -2.89288282e-01
3.41859497e-02 -2.01673448e-01 5.96975565e-01 -1.12295938e+00
-1.20763108e-01 5.30597195e-03 7.59702981e-01 -3.03160936e-01
3.68115276e-01 -5.54929614e-01 5.01805007e-01 5.35745859e-01
-2.07385510e-01 -1.72129095e-01 2.04989806e-01 3.57083291e-01
-2.84157664e-01 4.99695465e-02 1.05297530e+00 -1.09845638e-01
-2.53364503e-01 6.58468902e-01 1.96836740e-01 4.50235367e-01
1.03860641e+00 5.14284968e-02 -6.66686416e-01 -5.73824644e-01
-4.68179137e-01 1.37422740e-01 5.34996212e-01 5.17377615e-01
7.01369643e-01 -1.73403895e+00 -1.02701294e+00 4.77655083e-01
-1.79607049e-01 4.66803342e-01 2.66997159e-01 3.54676545e-01
-6.00562811e-01 9.00528952e-02 -3.33565772e-01 -3.65123957e-01
-9.06793058e-01 5.01002729e-01 5.67009091e-01 -2.41489530e-01
-5.86215258e-01 1.18057179e+00 5.96457005e-01 -5.33357680e-01
4.75850627e-02 -3.91790979e-02 7.26009905e-02 -1.71244547e-01
3.22121114e-01 3.38658571e-01 -1.04200810e-01 -3.85666758e-01
2.37278268e-03 8.70570093e-02 -5.81451505e-02 1.89585388e-01
1.31370628e+00 4.16323334e-01 2.26893630e-02 4.58047949e-02
1.17719865e+00 -1.60361096e-01 -1.72826183e+00 5.46851419e-02
-9.01483417e-01 -3.13531131e-01 -3.72803807e-02 -8.89168918e-01
-1.75659716e+00 8.77580583e-01 6.65759623e-01 1.20757535e-01
1.37957311e+00 -2.37405539e-01 7.84198642e-01 1.23935379e-02
-9.88492742e-02 -7.42407084e-01 1.47771120e-01 3.44477922e-01
1.27217185e+00 -1.24173594e+00 -8.62936601e-02 -3.67974490e-01
-9.69595969e-01 7.63762414e-01 9.73734081e-01 -3.37309748e-01
7.04116881e-01 2.42770880e-01 1.48797825e-01 1.05590736e-02
-6.15272641e-01 1.45015389e-01 1.91533729e-01 7.24975646e-01
1.39983088e-01 1.24423839e-01 -4.84985970e-02 5.55136263e-01
-8.37894559e-01 -1.22805730e-01 5.38170278e-01 6.03816211e-01
3.35994363e-01 -1.01159298e+00 -1.20438427e-01 4.14574832e-01
-5.60767353e-01 -3.38825881e-01 -3.82080376e-01 9.59834099e-01
2.12944314e-01 5.93233883e-01 2.93222845e-01 -4.62997705e-01
3.90856080e-02 -4.95931096e-02 3.74819666e-01 -2.98756003e-01
-4.37794030e-01 -1.16100833e-01 -2.18151331e-01 -3.95724833e-01
-1.86455712e-01 -3.57695192e-01 -1.02803159e+00 -4.34571058e-01
-2.80751109e-01 1.72406450e-01 6.43411398e-01 4.73008811e-01
4.69058812e-01 9.35216367e-01 7.23621905e-01 -5.80245614e-01
-4.71897990e-01 -1.23802471e+00 -3.22176158e-01 4.45695013e-01
2.56574869e-01 -4.54732269e-01 -6.62310421e-01 1.38730288e-01]
|
[11.633837699890137, -0.27682408690452576]
|
6aa14bfc-e9e6-4516-bc61-d6b42bcc5353
|
refocusing-is-key-to-transfer-learning
|
2305.15542
| null |
https://arxiv.org/abs/2305.15542v2
|
https://arxiv.org/pdf/2305.15542v2.pdf
|
TOAST: Transfer Learning via Attention Steering
|
Transfer learning involves adapting a pre-trained model to novel downstream tasks. However, we observe that current transfer learning methods often fail to focus on task-relevant features. In this work, we explore refocusing model attention for transfer learning. We introduce Top-Down Attention Steering (TOAST), a novel transfer learning algorithm that keeps the pre-trained backbone frozen, selects task-relevant features in the output, and feeds those features back to the model to steer the attention to the task-specific features. By refocusing the attention only, TOAST achieves state-of-the-art results on a number of transfer learning benchmarks, while having a small number of tunable parameters. Compared to fully fine-tuning, LoRA, and prompt tuning, TOAST substantially improves performance across a range of fine-grained visual classification datasets (e.g., 81.1% -> 86.2% on FGVC). TOAST also outperforms the fully fine-tuned Alpaca and Vicuna models on instruction-following language generation. Code is available at https://github.com/bfshi/TOAST.
|
['Xin Wang', 'Trevor Darrell', 'Siyu Gai', 'Baifeng Shi']
|
2023-05-24
| null | null | null | null |
['fine-grained-image-classification', 'instruction-following']
|
['computer-vision', 'natural-language-processing']
|
[ 1.60254419e-01 -2.77753383e-01 -4.88897592e-01 -5.39246500e-01
-8.72299373e-01 -7.06818938e-01 6.87425792e-01 -1.61927149e-01
-3.72398138e-01 8.65260065e-01 3.42513293e-01 -5.40991127e-01
1.66090220e-01 -7.00389266e-01 -1.04859757e+00 -3.35871458e-01
2.28459999e-01 2.88190216e-01 4.16672140e-01 -4.54718083e-01
3.71359944e-01 2.34842598e-01 -1.34667861e+00 6.77121162e-01
7.46987462e-01 7.50670612e-01 4.96340662e-01 7.91261673e-01
2.51050778e-02 7.25716114e-01 -3.83307487e-01 -1.18423834e-01
-8.21624249e-02 -3.47528398e-01 -1.13133502e+00 -4.33654159e-01
7.15846777e-01 -5.00784695e-01 -3.71168226e-01 5.85338831e-01
5.15221775e-01 2.31335506e-01 7.27656782e-01 -1.19275880e+00
-1.18834555e+00 4.79137570e-01 -4.82199192e-01 5.49070477e-01
-7.71548375e-02 4.28407252e-01 9.11628246e-01 -1.17607355e+00
3.14083904e-01 1.24158657e+00 4.77459431e-01 8.23949814e-01
-1.31074846e+00 -9.58782852e-01 4.05813754e-01 2.36853242e-01
-1.05567217e+00 -6.03392005e-01 3.37541372e-01 -5.92559099e-01
1.20243096e+00 -6.20495081e-02 1.45997003e-01 1.17007661e+00
3.50213587e-01 6.53669357e-01 1.06208551e+00 -5.28468668e-01
-1.25361502e-01 -4.41110060e-02 1.37047738e-01 8.79069746e-01
1.24030868e-02 8.72120485e-02 -8.18669617e-01 7.47479647e-02
6.44403696e-01 -1.01847753e-01 -2.99033135e-01 -3.67152691e-01
-1.40985084e+00 9.09063458e-01 7.78976440e-01 7.45781884e-02
-2.32938200e-01 5.46342373e-01 4.85170156e-01 4.75368708e-01
4.99285966e-01 6.40652180e-01 -8.60069990e-01 -2.91072220e-01
-5.21480322e-01 -5.22221252e-02 3.96051198e-01 1.18360817e+00
1.10594738e+00 1.91956535e-01 -9.39061463e-01 7.63977230e-01
1.93918064e-01 3.74482095e-01 6.27326488e-01 -9.87746119e-01
5.51581442e-01 2.12178588e-01 -1.55345008e-01 -2.12972492e-01
-6.58486634e-02 -5.95398843e-01 -5.10380983e-01 2.84983695e-01
2.36164227e-01 -3.14689696e-01 -1.28965938e+00 1.89151859e+00
-1.61476806e-02 2.19591245e-01 -9.11602676e-02 7.52892673e-01
8.54933202e-01 8.31959188e-01 3.75716865e-01 3.80108714e-01
1.13824511e+00 -1.65345097e+00 -2.55621612e-01 -4.67714757e-01
7.28202343e-01 -7.87444830e-01 1.60352349e+00 2.95000523e-01
-1.01892078e+00 -9.16139603e-01 -8.56483400e-01 -4.10646737e-01
-1.86867684e-01 2.07371116e-01 4.89536375e-01 2.41457447e-01
-1.39856684e+00 5.91409564e-01 -8.52813125e-01 -3.98573250e-01
7.41303146e-01 3.88676882e-01 -1.95371479e-01 -3.34186912e-01
-9.83155131e-01 7.55498886e-01 1.54148892e-01 -3.45724314e-01
-1.36501253e+00 -1.14429843e+00 -6.60806656e-01 2.97631621e-01
1.84666142e-01 -1.00557709e+00 1.72811413e+00 -7.83893287e-01
-1.81494319e+00 7.56160140e-01 -2.77870238e-01 -2.54685700e-01
3.39248866e-01 -6.91984415e-01 3.79245020e-02 -2.01265067e-01
1.14636041e-01 1.13130796e+00 9.12151098e-01 -1.04723370e+00
-6.55386925e-01 1.03850342e-01 -6.63944380e-03 2.02458054e-01
-5.20878792e-01 -6.15131259e-02 -4.54074889e-01 -5.78623056e-01
-6.04699671e-01 -1.00476575e+00 2.86603682e-02 -5.66041134e-02
-8.55591670e-02 -5.07983565e-01 7.88919747e-01 -2.32375950e-01
1.11981726e+00 -2.17201185e+00 2.41699845e-01 -3.32761884e-01
1.20028898e-01 4.38314915e-01 -5.33909976e-01 3.91725779e-01
1.47115067e-02 1.19433984e-01 2.32637394e-02 -3.15138996e-01
-6.31343722e-02 -1.32827824e-02 -3.79549474e-01 6.16978705e-02
4.27005142e-01 1.21554267e+00 -9.95344937e-01 -1.49927035e-01
1.16279230e-01 4.53018844e-01 -9.88008797e-01 4.82229114e-01
-3.34557623e-01 5.68542242e-01 -5.78405380e-01 2.76727974e-01
2.24212423e-01 -4.87964928e-01 -2.20566005e-01 -9.32588354e-02
-4.68834117e-02 6.47316039e-01 -2.95723498e-01 1.99781263e+00
-9.22886670e-01 7.58686125e-01 -1.25986978e-01 -8.04831088e-01
1.00947332e+00 1.14231538e-02 -1.49402753e-01 -7.66401708e-01
-5.61046600e-02 1.26828179e-01 1.04387790e-01 -3.42364669e-01
4.62214977e-01 1.49928004e-01 -5.81411719e-02 5.24431825e-01
4.81428713e-01 -5.04566878e-02 -8.44591558e-02 2.08233640e-01
1.09119451e+00 5.72182655e-01 1.54586136e-01 -4.58770066e-01
4.05709505e-01 2.07508728e-02 3.12750280e-01 6.97187245e-01
-1.93254396e-01 5.49289525e-01 8.97504240e-02 -2.83737421e-01
-7.21124947e-01 -1.11169684e+00 2.91695029e-01 2.33143258e+00
-1.98418483e-01 -4.18594599e-01 -6.16580129e-01 -8.46994102e-01
2.17600614e-01 9.03649986e-01 -8.87391269e-01 -6.35933161e-01
-5.69719374e-01 -3.64745259e-01 3.09820265e-01 7.21769691e-01
4.37661648e-01 -1.26786971e+00 -5.50171554e-01 1.76311165e-01
-2.01243293e-02 -7.73336232e-01 -9.68926072e-01 4.64903802e-01
-9.75565851e-01 -5.64860761e-01 -6.95257604e-01 -9.20573294e-01
5.68980157e-01 4.53763515e-01 1.49357224e+00 9.92528349e-02
1.01800617e-02 1.31431058e-01 -3.07427853e-01 -2.73773789e-01
-3.50186139e-01 7.39173591e-01 -3.19504052e-01 -1.69474766e-01
3.61244142e-01 -3.38639110e-01 -6.16863668e-01 3.26088279e-01
-4.16944712e-01 1.66974545e-01 6.89943910e-01 1.17173076e+00
4.50464368e-01 -8.28407705e-01 9.96457994e-01 -9.82955873e-01
6.18268907e-01 -5.84432483e-01 -3.41645390e-01 2.26247653e-01
-6.37833476e-01 3.77899289e-01 7.21513927e-01 -5.03262520e-01
-1.16591775e+00 -1.48493960e-01 1.60356417e-01 -5.78167200e-01
-1.60733387e-01 3.20166647e-01 1.33997038e-01 1.11465968e-01
8.45085442e-01 1.94977030e-01 -1.66361436e-01 -5.50600767e-01
5.38892269e-01 6.48277044e-01 5.71835220e-01 -9.08513725e-01
5.91481388e-01 8.48928690e-02 -3.70026201e-01 -2.40846694e-01
-9.91934121e-01 -2.58452207e-01 -6.28911734e-01 1.09749444e-01
8.15600395e-01 -1.08244300e+00 -3.50544393e-01 4.03740555e-01
-8.85110438e-01 -1.29991376e+00 -2.16287479e-01 4.23659623e-01
-7.42030740e-01 -3.82298499e-01 -8.29441488e-01 2.23345179e-02
-4.97735023e-01 -1.30651629e+00 1.03233945e+00 9.53611135e-02
-3.96248549e-01 -8.70665908e-01 1.04292028e-01 2.59117842e-01
9.41659868e-01 -3.41008216e-01 1.12444568e+00 -4.53366727e-01
-5.29540777e-01 4.85944122e-01 -4.12844688e-01 2.76197284e-01
3.29992235e-01 2.15910822e-02 -1.12931263e+00 -6.63283467e-01
-5.98721862e-01 -9.41293180e-01 1.25074208e+00 3.78827989e-01
1.42714977e+00 -6.69748858e-02 -5.02766252e-01 9.65896785e-01
9.90835965e-01 1.50792897e-01 3.19837332e-01 4.51029658e-01
6.81729317e-01 3.32308710e-01 6.48207903e-01 1.81075394e-01
5.01801908e-01 7.13517785e-01 3.38459909e-01 -3.22024599e-02
-5.73559403e-01 -4.12817299e-01 5.29561698e-01 7.68012762e-01
7.48983771e-02 -2.13343680e-01 -9.29162383e-01 7.29911327e-01
-1.70009875e+00 -5.60121715e-01 1.46254048e-01 2.02638268e+00
1.28672612e+00 2.82939464e-01 -8.17067474e-02 -5.71551442e-01
6.30025208e-01 1.64064742e-03 -8.01460385e-01 -7.22084582e-01
4.56515789e-01 5.91166973e-01 3.85321528e-01 6.29442155e-01
-9.92237985e-01 1.42672443e+00 5.54036140e+00 8.03812981e-01
-1.43467855e+00 3.16041142e-01 7.99654603e-01 -2.95706183e-01
-3.40496421e-01 -1.31477162e-01 -1.01692951e+00 2.22068310e-01
1.11677098e+00 -3.87452334e-01 4.85539496e-01 7.01184392e-01
2.86016818e-02 2.78994858e-01 -1.38915026e+00 4.90088254e-01
1.71484277e-02 -1.29367304e+00 2.51151919e-01 -1.61320418e-01
1.08911455e+00 5.95093071e-01 3.67992491e-01 9.83355880e-01
7.05555737e-01 -1.32566452e+00 6.76015496e-01 2.92450875e-01
1.20956779e+00 -6.98847592e-01 2.99216062e-01 1.35700300e-01
-9.51080263e-01 -1.26704052e-02 -3.16511542e-01 -2.13591516e-01
-2.20447317e-01 8.51762742e-02 -9.08658147e-01 6.12200238e-02
8.84556174e-01 8.03020418e-01 -7.34554589e-01 9.24705207e-01
-4.01100487e-01 1.04846454e+00 9.21884999e-02 -1.82862976e-04
3.66957515e-01 3.54619086e-01 -5.60814217e-02 1.49835181e+00
4.19339657e-01 -5.30166291e-02 1.08613528e-01 8.28969955e-01
-5.33390582e-01 -1.24009237e-01 -3.83250445e-01 2.33327523e-01
5.60183167e-01 1.21291554e+00 -2.46295601e-01 -4.87269610e-01
-4.76879448e-01 1.04930818e+00 8.20977509e-01 5.66684783e-01
-7.99526870e-01 -4.25737113e-01 7.99470305e-01 1.05768546e-01
6.38809800e-01 -2.97483318e-02 -1.31807193e-01 -1.02966177e+00
-4.29731637e-01 -8.49650264e-01 2.82905996e-01 -9.13687110e-01
-1.21261609e+00 7.74818718e-01 -5.68051748e-02 -9.77751195e-01
-3.20414752e-01 -6.39357984e-01 -8.30416024e-01 1.16594946e+00
-1.85674608e+00 -1.11150110e+00 -5.31869292e-01 6.74776733e-01
8.54220748e-01 -3.50191623e-01 9.19517577e-01 9.43952873e-02
-3.63469839e-01 9.24847901e-01 6.71972930e-02 -7.91620240e-02
1.35452676e+00 -1.28126001e+00 8.34254086e-01 3.92280638e-01
-1.12300158e-01 5.54634690e-01 3.13304454e-01 -4.31554347e-01
-1.16830540e+00 -1.47621548e+00 7.61546969e-01 -4.66433525e-01
6.84776366e-01 -4.73615587e-01 -1.10194039e+00 1.16654253e+00
6.11906886e-01 4.16777551e-01 5.46938181e-01 1.67250603e-01
-7.04525828e-01 -1.35443017e-01 -6.88697159e-01 4.77135390e-01
1.14911962e+00 -4.94274080e-01 -4.88758117e-01 1.99191689e-01
9.93879557e-01 -3.55271220e-01 -7.34514177e-01 3.18013519e-01
4.83633935e-01 -6.91921175e-01 9.46165919e-01 -1.00668144e+00
7.02820122e-01 1.26330853e-02 -1.01709977e-01 -1.93633819e+00
-9.69381690e-01 -5.44144571e-01 -7.78608546e-02 1.12503135e+00
5.69862604e-01 -5.75013697e-01 7.27926910e-01 1.91456899e-02
-6.27516031e-01 -9.21675086e-01 -5.34408391e-01 -6.93572164e-01
7.19673216e-01 7.24724121e-03 5.25702655e-01 8.64278555e-01
-2.36124188e-01 7.36286283e-01 -2.87590474e-01 -2.71064341e-01
3.44843864e-01 3.09016436e-01 8.82801890e-01 -9.75435019e-01
-5.12460113e-01 -5.50942302e-01 2.24390060e-01 -1.36184406e+00
3.39721203e-01 -1.10897517e+00 1.26356199e-01 -1.47993076e+00
3.19889843e-01 -7.05058813e-01 -7.07727671e-01 8.92281055e-01
-5.61437488e-01 4.01570052e-01 3.04664403e-01 9.39968526e-02
-5.84381640e-01 6.95081413e-01 1.51733446e+00 -1.81729242e-01
-1.70032203e-01 -6.63022846e-02 -9.24154997e-01 4.25028920e-01
1.21429741e+00 -4.85827059e-01 -5.47839880e-01 -9.69986737e-01
-1.98934346e-01 -2.55161315e-01 1.48208961e-01 -8.66936326e-01
5.80977201e-02 -3.68030965e-01 5.02009749e-01 -1.02339454e-01
1.67699948e-01 -2.38479078e-01 -4.76678044e-01 5.46271801e-01
-8.41551542e-01 1.47074118e-01 6.79743648e-01 3.57483834e-01
-1.19650424e-01 -1.47523418e-01 7.81506538e-01 8.05004090e-02
-9.65008616e-01 2.29279757e-01 -3.25211018e-01 4.52718556e-01
8.35891187e-01 2.14762658e-01 -8.69740963e-01 -3.30465615e-01
-5.14451206e-01 4.30256575e-01 3.16603839e-01 8.06461155e-01
4.89089012e-01 -1.34424222e+00 -1.09205103e+00 3.02540779e-01
3.88346851e-01 7.07157999e-02 1.82759970e-01 5.33544004e-01
-2.40322992e-01 6.76381588e-01 -5.44059336e-01 -6.58480346e-01
-1.10960710e+00 4.72249717e-01 3.25985909e-01 -2.06418484e-01
-2.07498431e-01 1.25934803e+00 7.07739174e-01 -5.69854498e-01
5.84924780e-02 -5.48891902e-01 -1.70898270e-02 -2.78632134e-01
4.11192924e-01 2.46896848e-01 2.08796024e-01 -3.27121764e-01
-3.47313941e-01 4.98418242e-01 -4.52814609e-01 1.76221997e-01
1.18761015e+00 3.91657650e-03 2.65037984e-01 4.39135432e-01
1.35851359e+00 -1.11230657e-01 -1.80812573e+00 -4.33964342e-01
-2.88550556e-01 -2.15377152e-01 1.97907865e-01 -1.17765141e+00
-1.06822360e+00 1.27025163e+00 3.54493439e-01 -4.55050945e-01
1.05549526e+00 5.89916781e-02 6.50652230e-01 5.17217994e-01
3.52350831e-01 -5.56968093e-01 5.33347189e-01 8.56590748e-01
1.00303090e+00 -1.34894919e+00 -2.46552631e-01 -1.64203018e-01
-5.59848547e-01 9.22006071e-01 1.20431757e+00 -1.73322394e-01
4.64123160e-01 1.89106926e-01 -3.49776670e-02 2.16949075e-01
-1.49017000e+00 1.02099194e-03 4.40702766e-01 6.00827515e-01
9.08417523e-01 -3.37308720e-02 2.26877913e-01 3.62886280e-01
-1.70054555e-01 3.57684612e-01 4.49998647e-01 9.30490911e-01
-5.97620845e-01 -1.08205163e+00 -9.43964794e-02 6.76551521e-01
-3.37438732e-01 -5.15852928e-01 -1.51467830e-01 7.10553706e-01
-1.45335659e-01 7.43847847e-01 3.00694108e-01 -3.29450339e-01
4.04101193e-01 1.25279158e-01 5.63838363e-01 -1.11236358e+00
-8.07309687e-01 -1.57956243e-01 -4.90420535e-02 -6.25860751e-01
-2.09592432e-02 -4.42760497e-01 -1.18441117e+00 -2.74524033e-01
-5.37092276e-02 1.20868094e-01 2.12004051e-01 6.06286705e-01
7.89020002e-01 9.57839549e-01 4.19300735e-01 -8.83507431e-01
-7.85260797e-01 -1.26484358e+00 1.14183128e-01 3.28558087e-01
5.58146298e-01 -7.31812954e-01 -8.69340375e-02 2.51169533e-01]
|
[10.673981666564941, 8.222457885742188]
|
fdd619af-cc61-463e-9508-767da2d706bb
|
drugehrqa-a-question-answering-dataset-on
| null | null |
https://openreview.net/forum?id=HBYJawngw5
|
https://openreview.net/pdf?id=HBYJawngw5
|
DrugEHRQA: A Question Answering Dataset on Structured and Unstructured Electronic Health Records For Medicine Related Queries
|
This paper develops the first question answering dataset (DrugEHRQA) containing question-answer pairs from both structured tables and unstructured notes from a publicly available Electronic Health Record (EHR). EHRs contain patient records, stored in structured tables as well as unstructured clinical notes. The information in structured and unstructured EHR records is not strictly disjoint: information may be duplicated, contradictory, or provide additional context between these sources. This presents a rich opportunity to study question answering (QA) models that combine reasoning over both structured and unstructured data. Additionally, we propose a novel methodology that automatically generates a large QA dataset by retrieving answers from both structured and unstructured EHR records. The automatically-generated dataset has medication-related queries, containing over 70,000 question-answer pairs. Our dataset is validated for both individual modalities using state-of-the-art QA models. In order to address the problem arising from complex, nested queries, this is the first time Relation-Aware Schema Encoding and Linking for Text-to-SQL Parsers (RAT-SQL) has been used for EHR data. Finally, we introduce a rule-based method to obtain multi-modal answers, combining the answers from the different modalities. Our goal is to provide a benchmark dataset for multi-modal QA systems, and to open up new avenues of research in improving question answering over EHR structured data by using context from unstructured clinical data.
|
['Anonymous']
|
2021-11-16
| null | null | null |
acl-arr-november-2021-11
|
['text-to-sql']
|
['computer-code']
|
[ 1.12071998e-01 6.00713789e-01 -9.76922289e-02 -5.32840073e-01
-1.44399512e+00 -8.00176561e-01 1.31109670e-01 1.29215682e+00
-3.94700281e-02 9.01151180e-01 6.59695745e-01 -6.04666293e-01
-7.07363665e-01 -1.35354531e+00 -5.17585814e-01 2.11931124e-01
4.31319624e-02 1.03269577e+00 4.63998765e-01 -5.59482098e-01
-2.28756964e-01 3.60123254e-02 -1.42830276e+00 1.17300451e+00
1.19151115e+00 9.79089677e-01 -2.47571409e-01 6.64837301e-01
-7.39205182e-01 1.63682997e+00 -5.01093507e-01 -8.69871736e-01
-1.33657277e-01 -5.98375201e-01 -1.44274104e+00 -5.26008308e-01
4.50003713e-01 1.20542571e-01 2.08823770e-01 7.78987885e-01
6.24904335e-01 -5.29500306e-01 1.57664120e-01 -9.74805832e-01
-6.64831281e-01 6.31463528e-01 5.52350700e-01 -2.05683131e-02
1.43549120e+00 2.70951372e-02 9.64933693e-01 -3.17560196e-01
1.29552913e+00 1.01842833e+00 7.25312531e-01 5.92384934e-01
-1.05803895e+00 -1.05692483e-01 -6.42879605e-01 3.25033277e-01
-1.05023181e+00 -2.43431091e-01 2.59045720e-01 -3.81216288e-01
1.19228137e+00 8.61586630e-01 3.81471783e-01 5.56952655e-01
3.33920091e-01 4.88377035e-01 1.08162963e+00 -3.98598552e-01
3.96595597e-01 3.65116388e-01 5.50243199e-01 7.23891020e-01
4.92525369e-01 -2.98917294e-01 -2.75732845e-01 -8.46496940e-01
1.06885374e-01 -3.92940603e-02 -2.21331432e-01 -1.78749979e-01
-1.00805032e+00 6.71373785e-01 1.64191619e-01 3.14429820e-01
-6.05696797e-01 -5.26513875e-01 6.26723230e-01 6.47584736e-01
-3.33894402e-01 8.44666719e-01 -1.00806510e+00 3.34778726e-02
-4.63115752e-01 8.16321313e-01 1.53993964e+00 1.33578980e+00
5.70184648e-01 -9.05033290e-01 -5.69766521e-01 4.92646664e-01
2.94978291e-01 4.39554036e-01 3.77940595e-01 -1.03061461e+00
8.90033305e-01 1.35131252e+00 3.01808357e-01 -8.76276910e-01
-6.67021692e-01 1.75623924e-01 -6.13502145e-01 -7.71740258e-01
4.88995433e-01 3.46562197e-03 -6.10077977e-01 1.41730583e+00
5.69246531e-01 -4.81958628e-01 6.94047630e-01 3.79525244e-01
1.87563276e+00 2.30588540e-01 6.69575810e-01 -4.33180988e-01
2.17870045e+00 -3.40898842e-01 -1.31626725e+00 3.07446629e-01
1.14880526e+00 -6.22618198e-01 6.98593080e-01 1.92454398e-01
-1.31558359e+00 -2.89708316e-01 -4.44556504e-01 -4.42634046e-01
-8.04276049e-01 -3.99257571e-01 1.91633999e-01 6.49280131e-01
-7.96679914e-01 6.80834353e-02 -4.21885133e-01 -4.53559577e-01
2.50211447e-01 1.62733719e-01 -5.04635334e-01 -4.39683408e-01
-1.63767111e+00 9.03600693e-01 5.35829782e-01 -5.04535735e-01
-1.40132308e-01 -1.31458139e+00 -1.16755414e+00 8.28300044e-02
7.10380912e-01 -1.41808975e+00 1.17305577e+00 -7.49522224e-02
-8.14795136e-01 8.04327071e-01 -2.35548437e-01 -5.47185183e-01
1.71223983e-01 1.15839921e-01 -9.63881075e-01 3.09775412e-01
3.65741253e-01 2.46175125e-01 -1.75988242e-01 -1.02969229e+00
-4.61500019e-01 -9.74103391e-01 2.24131271e-01 -2.01170459e-01
2.95512557e-01 1.61297038e-01 -1.40922904e-01 -2.93064594e-01
-1.50753766e-01 -2.25136921e-01 -2.13774770e-01 -3.28154981e-01
-5.26046097e-01 -3.59842926e-01 1.98648810e-01 -9.72058713e-01
1.77934325e+00 -1.48685336e+00 -2.58443326e-01 1.36933491e-01
3.79098147e-01 2.28924930e-01 5.52026555e-02 1.08033597e+00
-1.96248278e-01 3.07874292e-01 -5.94510317e-01 3.45960796e-01
-4.15081680e-02 5.39099753e-01 -4.44380611e-01 -5.56473792e-01
4.46752757e-01 1.27186072e+00 -9.03477490e-01 -1.10202038e+00
-3.72387975e-01 1.25799850e-01 -7.25807905e-01 5.08268178e-01
-7.78811753e-01 3.48143637e-01 -5.83051145e-01 9.47223246e-01
6.46958768e-01 -4.41561431e-01 4.32222903e-01 -3.21907759e-01
2.17812479e-01 5.89346647e-01 -1.26932359e+00 1.56085026e+00
-1.07313231e-01 -4.73745435e-01 -1.86950609e-01 -5.85872352e-01
7.71327198e-01 7.98577130e-01 8.60310853e-01 -9.49724078e-01
-2.21268699e-01 4.04052526e-01 -4.21129346e-01 -1.45357239e+00
4.54217136e-01 -2.76521206e-01 -3.31465483e-01 6.95597827e-02
4.30242270e-02 -3.06870878e-01 5.12608945e-01 2.49767318e-01
1.54221237e+00 -6.72509670e-02 8.20137382e-01 -8.37626234e-02
1.03889799e+00 7.33294129e-01 4.16044474e-01 6.98937595e-01
1.12775385e-01 3.39914978e-01 5.97145259e-01 -6.12245142e-01
-6.19781792e-01 -1.01132262e+00 -7.38927424e-01 3.54301095e-01
-2.84262598e-01 -9.46347296e-01 -6.46763146e-01 -7.90981174e-01
1.60728261e-01 5.94285071e-01 -6.01492584e-01 1.68253943e-01
-6.22089088e-01 -3.92233521e-01 8.40889633e-01 4.58766192e-01
1.65126026e-01 -1.18062067e+00 -8.47690463e-01 8.12327623e-01
-7.03874946e-01 -1.24903691e+00 9.75721478e-02 1.02720760e-01
-1.12207294e+00 -1.89758360e+00 -2.36592054e-01 -5.94637394e-01
1.94615617e-01 -6.60605013e-01 1.94260347e+00 1.11934148e-01
-2.62994140e-01 7.13121891e-01 -4.65218216e-01 -5.72128236e-01
-7.53602862e-01 -1.35093872e-02 -6.90706730e-01 -4.48845744e-01
7.88529813e-01 5.10550924e-02 -5.14442742e-01 2.22924143e-01
-1.67051208e+00 -4.92690116e-01 8.60669650e-03 5.88261664e-01
1.04635262e+00 -2.69008487e-01 9.45128083e-01 -1.55930912e+00
9.37386632e-01 -9.01201904e-01 -4.59400117e-01 9.27269995e-01
-6.13019168e-01 4.14099395e-01 8.64743590e-01 4.15365279e-01
-9.63131309e-01 5.74937463e-02 -8.52969527e-01 3.58041197e-01
-5.49825549e-01 9.61111903e-01 -2.73460954e-01 5.95962346e-01
9.61263299e-01 -8.70084465e-02 -8.30243751e-02 -6.29476011e-01
5.67935467e-01 6.63184643e-01 5.27499914e-01 -6.84071958e-01
1.51658654e-01 2.72816002e-01 1.36783376e-01 -2.12448627e-01
-6.98153377e-01 -7.55116999e-01 -4.29707646e-01 3.70418549e-01
1.10884702e+00 -6.13280714e-01 -1.00978589e+00 -2.87959129e-01
-1.15380144e+00 1.61120683e-01 -1.02789974e+00 8.41037333e-02
-6.51055753e-01 4.13024783e-01 -6.56440556e-01 -5.35556674e-01
-5.80966294e-01 -9.10007596e-01 1.19354725e+00 -1.42775327e-01
-4.74021941e-01 -9.59769487e-01 6.16966724e-01 7.91602731e-01
2.04341218e-01 4.78935212e-01 1.57766175e+00 -1.04839778e+00
-4.78414446e-01 -8.97713825e-02 -1.04835749e-01 -3.06874752e-01
3.51309180e-01 -4.06709075e-01 -6.17425978e-01 2.92851895e-01
1.88815922e-01 -4.52979922e-01 2.88197696e-01 -1.25070028e-02
1.02926016e+00 -4.64884639e-01 -1.35596961e-01 -2.18490083e-02
1.66996908e+00 2.66651452e-01 9.83587503e-01 3.13386649e-01
2.69666731e-01 1.02386522e+00 5.59865713e-01 4.06375587e-01
1.06020451e+00 4.45066601e-01 2.42872566e-01 1.47870749e-01
1.72775805e-01 -2.24085689e-01 -3.18909317e-01 7.34550595e-01
2.23515972e-01 -7.58736059e-02 -1.22819984e+00 5.56116402e-01
-1.70605314e+00 -8.97284329e-01 -8.35652649e-01 2.09082580e+00
1.47565222e+00 -4.29758519e-01 1.42832309e-01 2.51268893e-01
-3.18342671e-02 -5.57318568e-01 -3.77430886e-01 -4.28105921e-01
-9.79093090e-02 7.67632365e-01 4.10130583e-02 4.92208034e-01
-7.86535740e-01 2.42075965e-01 6.16123962e+00 2.60945290e-01
-2.14029163e-01 3.76947336e-02 1.53436691e-01 4.54014957e-01
-8.99470866e-01 -1.28832757e-01 -8.18871140e-01 2.40123540e-01
1.49765241e+00 -2.60416865e-02 -1.16507642e-01 5.62465191e-01
-1.41581953e-01 -1.66820452e-01 -1.35482979e+00 6.92091703e-01
-7.97811821e-02 -1.78040814e+00 2.61673301e-01 -1.59838963e-02
4.42612559e-01 -2.97491670e-01 -3.99296016e-01 3.08202177e-01
3.89442533e-01 -8.68936360e-01 1.19926997e-01 1.14459872e+00
6.73075616e-01 -1.95982695e-01 1.24475825e+00 3.26884896e-01
-1.20378423e+00 -2.45222673e-01 -1.36298127e-02 5.03576100e-01
1.19736843e-01 6.33590102e-01 -8.47372591e-01 1.59019995e+00
8.28432560e-01 3.71586084e-01 -8.63493264e-01 1.06425595e+00
2.30982989e-01 1.79798782e-01 -1.97964802e-01 3.45559448e-01
-2.84997225e-01 1.24127477e-01 9.35472399e-02 1.27138627e+00
2.55195111e-01 6.05728626e-01 -1.14580259e-01 7.49219060e-01
-9.49418545e-02 6.02853894e-01 -8.68484259e-01 7.52657279e-02
2.70674407e-01 8.03124785e-01 -1.45719215e-01 -7.86279440e-01
-5.01314342e-01 1.61610946e-01 5.65964803e-02 1.13017462e-01
-5.02923429e-01 -2.76581943e-01 1.56017616e-01 4.67402220e-01
-4.23403420e-02 4.28012431e-01 -3.64073455e-01 -1.28517151e+00
2.51877725e-01 -1.46170473e+00 1.52838576e+00 -7.89148331e-01
-1.53721690e+00 8.17637801e-01 1.92695379e-01 -1.09576178e+00
-7.12789595e-01 -3.17834020e-01 3.28616887e-01 8.34699333e-01
-1.64403009e+00 -8.54605913e-01 -2.88099885e-01 1.03826821e+00
2.56292932e-02 1.48379087e-01 1.47768939e+00 8.77338052e-01
-1.55119027e-03 3.56563359e-01 -3.63783330e-01 1.62385702e-01
7.51129270e-01 -1.51178479e+00 -9.02295336e-02 3.43372636e-02
-1.69082180e-01 9.70356703e-01 3.47432792e-01 -8.68875086e-01
-1.77796352e+00 -1.09599698e+00 1.68263626e+00 -1.25017142e+00
2.76628286e-01 4.70875166e-02 -1.47718215e+00 5.51000416e-01
3.18367958e-01 -1.94181874e-02 1.39325094e+00 -1.01404391e-01
-4.64908630e-01 -1.70792446e-01 -1.72218728e+00 7.60947764e-02
6.81533039e-01 -6.43810868e-01 -1.28861022e+00 3.89357448e-01
9.13618505e-01 -4.78644550e-01 -1.84674633e+00 8.23342443e-01
1.16522193e-01 -7.33907759e-01 1.15642810e+00 -1.23126709e+00
5.34541547e-01 -5.88367879e-01 -3.95727038e-01 -6.56719148e-01
2.20717996e-01 -3.11452657e-01 -4.94924307e-01 1.05177319e+00
6.37785554e-01 -6.85323954e-01 5.35016894e-01 1.15928888e+00
4.38625030e-02 -9.81685102e-01 -9.13491130e-01 -3.29888523e-01
6.36860207e-02 -3.90164554e-01 1.15749705e+00 1.00986516e+00
3.70258868e-01 3.84919852e-01 4.39646602e-01 1.85327232e-01
2.34151870e-01 3.65521222e-01 4.37725216e-01 -1.44359863e+00
-2.45159864e-01 1.75242752e-01 -2.51928300e-01 -4.85484481e-01
-6.55963898e-01 -9.47994471e-01 -3.67249459e-01 -2.25568271e+00
6.19925894e-02 -5.05447567e-01 1.39098287e-01 6.46525204e-01
-2.00671837e-01 -1.72900438e-01 -5.13013601e-02 4.21326645e-02
-7.60857940e-01 4.60806936e-02 1.27483845e+00 -2.21103370e-01
-1.56484693e-01 -2.15110317e-01 -7.26274133e-01 2.87300795e-01
4.32507813e-01 -7.94355631e-01 -4.96272415e-01 -1.66844591e-01
9.52002883e-01 9.55472112e-01 2.05404386e-01 -7.09605992e-01
4.77495998e-01 -6.42440394e-02 -1.09790169e-01 -8.15368891e-01
-9.66814682e-02 -1.30468631e+00 4.66990501e-01 6.41095281e-01
-5.66806436e-01 3.39591622e-01 2.41477400e-01 3.18016082e-01
-7.31983602e-01 -3.66054595e-01 2.27445066e-01 -4.87865150e-01
-1.53119177e-01 2.86072437e-02 -2.34068140e-01 8.48620713e-01
7.97866523e-01 8.27418938e-02 -6.27725422e-01 -4.27332893e-02
-1.06506383e+00 6.26187325e-01 -5.70221916e-02 2.84606874e-01
6.21456265e-01 -1.10192156e+00 -7.93845773e-01 1.19149014e-01
6.66012287e-01 1.28036439e-01 4.10564184e-01 6.71510637e-01
-6.14517152e-01 9.15499628e-01 3.40195862e-03 -4.62435901e-01
-1.22429121e+00 1.07102275e+00 3.63615632e-01 -9.22859907e-01
-3.15855175e-01 3.09208427e-02 -4.41804767e-01 -1.05654311e+00
2.73704366e-03 -8.08082640e-01 -6.27194345e-01 1.99177861e-01
6.70034707e-01 1.52261660e-01 7.09981382e-01 -2.74552852e-01
-4.85321820e-01 3.52990746e-01 1.30902901e-01 1.92042619e-01
1.15144968e+00 1.27404213e-01 -5.23665547e-01 1.12081423e-01
9.81317818e-01 7.77187273e-02 1.34351552e-01 -2.61326611e-01
5.17707169e-01 -1.50590390e-01 -7.52408564e-01 -1.28243136e+00
-4.17658478e-01 5.25220335e-01 4.14230227e-01 7.72252023e-01
1.23404813e+00 2.91190267e-01 8.82266939e-01 6.41231537e-01
3.90040636e-01 -7.39807189e-01 -3.08143497e-01 4.94718939e-01
9.97876287e-01 -1.17326534e+00 -2.26912484e-01 -5.76811850e-01
-5.36006689e-01 1.08249605e+00 3.23581219e-01 6.75573170e-01
6.62828028e-01 4.71033484e-01 3.41640174e-01 -9.53484058e-01
-9.04042423e-01 -4.50248748e-01 3.23314130e-01 7.08640635e-01
6.38930678e-01 -9.39421877e-02 -5.91243863e-01 9.50079322e-01
-1.25729904e-01 8.48242700e-01 2.59934783e-01 1.34317315e+00
-4.80399504e-02 -1.76920307e+00 -6.25454783e-01 7.89090216e-01
-8.23322475e-01 -2.60333955e-01 -4.13333893e-01 6.29375160e-01
3.12798917e-01 1.25656664e+00 -3.46509397e-01 1.71801914e-02
1.09590840e+00 5.40048659e-01 3.59516740e-01 -7.68877685e-01
-1.40568757e+00 -5.71815908e-01 5.03778994e-01 -5.97204983e-01
-5.13831973e-01 -4.66136903e-01 -1.60336006e+00 1.54505685e-01
1.80014998e-01 5.48774660e-01 2.63360530e-01 7.45580316e-01
7.66873896e-01 5.95161796e-01 -1.76776741e-02 9.34944987e-01
-4.75650340e-01 -4.67006922e-01 2.22585555e-02 6.99506700e-01
2.22925201e-01 -1.70356743e-02 4.23084199e-01 2.98856080e-01]
|
[8.753884315490723, 8.50833511352539]
|
23fcb4de-7b82-4c76-8884-51bb260ed077
|
quantum-neural-network-for-quantum-neural
|
2305.08544
| null |
https://arxiv.org/abs/2305.08544v1
|
https://arxiv.org/pdf/2305.08544v1.pdf
|
Quantum Neural Network for Quantum Neural Computing
|
Neural networks have achieved impressive breakthroughs in both industry and academia. How to effectively develop neural networks on quantum computing devices is a challenging open problem. Here, we propose a new quantum neural network model for quantum neural computing using (classically-controlled) single-qubit operations and measurements on real-world quantum systems with naturally occurring environment-induced decoherence, which greatly reduces the difficulties of physical implementations. Our model circumvents the problem that the state-space size grows exponentially with the number of neurons, thereby greatly reducing memory requirements and allowing for fast optimization with traditional optimization algorithms. We benchmark our model for handwritten digit recognition and other nonlinear classification tasks. The results show that our model has an amazing nonlinear classification ability and robustness to noise. Furthermore, our model allows quantum computing to be applied in a wider context and inspires the earlier development of a quantum neural computer than standard quantum computers.
|
['Zeng-Bing Chen', 'Tong-Kai Xu', 'Chen-Long Li', 'Hua-Lei Yin', 'Zhi-Ping Liu', 'Min-Gang Zhou']
|
2023-05-15
| null | null | null | null |
['handwritten-digit-recognition']
|
['computer-vision']
|
[ 2.82732874e-01 -3.79933953e-01 2.97264792e-02 -2.53185660e-01
-2.52913594e-01 -4.30710346e-01 4.67312962e-01 -1.56544104e-01
-7.89742291e-01 7.55513608e-01 -4.59575742e-01 -5.29337287e-01
-1.93797182e-02 -1.29486501e+00 -6.45058334e-01 -1.15907538e+00
1.08788818e-01 1.82923153e-01 1.57164350e-01 -4.83510107e-01
4.66640472e-01 3.43477875e-01 -1.40872598e+00 -1.39585093e-01
7.40651488e-01 7.91731775e-01 -9.84693691e-02 7.74546087e-01
3.83985132e-01 8.23328733e-01 -4.85614479e-01 -2.83545017e-01
4.05133128e-01 -6.60555720e-01 -5.41342914e-01 -4.77484137e-01
4.19471860e-01 -2.56783843e-01 -1.18204832e+00 1.72970676e+00
4.95985180e-01 2.87990481e-01 5.17235160e-01 -6.76727533e-01
-1.26480389e+00 4.75871354e-01 3.17769229e-01 2.09202990e-01
-2.91574031e-01 1.74852133e-01 1.13456261e+00 -1.76910862e-01
4.22458351e-01 8.44996452e-01 4.50179011e-01 9.06969666e-01
-1.32772923e+00 -7.60347664e-01 -8.74828279e-01 7.32294261e-01
-1.64290929e+00 -4.43553805e-01 4.08517510e-01 5.28064333e-02
1.29363298e+00 -2.00488772e-02 6.44639909e-01 6.30909085e-01
5.54514050e-01 2.19604790e-01 1.04797852e+00 -5.84059834e-01
8.01995933e-01 -4.61331792e-02 3.46666485e-01 1.09620368e+00
2.60720015e-01 6.72352672e-01 -6.19839787e-01 1.67018145e-01
9.23980117e-01 -1.47396103e-01 -1.25459060e-01 -4.07520346e-02
-1.32848036e+00 9.47460473e-01 7.57278442e-01 3.45691085e-01
-2.45512202e-01 7.31671154e-01 1.11314379e-01 5.56031346e-01
-4.65990394e-01 7.79820204e-01 -5.11388993e-03 -1.92797571e-01
-5.63789546e-01 3.36911343e-02 8.97529185e-01 7.66893148e-01
8.43656361e-01 1.92369714e-01 1.97321773e-01 3.64404798e-01
-1.74524188e-02 9.84847665e-01 3.86791766e-01 -1.24947894e+00
1.17220050e-02 7.39196911e-02 -5.31044938e-02 -5.76378107e-01
-2.60252327e-01 -1.83737025e-01 -1.30384278e+00 2.26792723e-01
4.21672136e-01 -7.48086423e-02 -9.59161937e-01 1.74620390e+00
-1.54146075e-01 1.63930997e-01 4.69629109e-01 9.37381208e-01
5.24985909e-01 8.19266379e-01 -4.11165178e-01 -4.95260060e-02
1.25294960e+00 -7.86507070e-01 -8.74746144e-01 3.28508355e-02
7.75627196e-01 -3.00647765e-01 4.89723533e-01 5.21037102e-01
-7.84146845e-01 -2.17622310e-01 -1.56144726e+00 -1.75229847e-01
-4.58350450e-01 -2.68806189e-01 1.22316241e+00 1.17662644e+00
-1.09705448e+00 1.21097016e+00 -1.10069132e+00 -1.18189484e-01
2.12533846e-01 9.17294204e-01 -5.59495687e-01 -9.76385176e-02
-1.58468497e+00 1.13790309e+00 7.68098354e-01 2.84819812e-01
-3.93949479e-01 4.29137945e-02 -6.11731410e-01 -5.52507862e-02
-6.94928244e-02 -8.28612089e-01 1.09675324e+00 -3.33593130e-01
-2.24217367e+00 4.25818801e-01 -1.79110825e-01 -4.29157764e-01
-4.01711851e-01 3.37523043e-01 -3.57771575e-01 6.74689561e-02
-3.87923092e-01 3.88862193e-01 4.74651635e-01 -1.27041340e-01
-2.30613410e-01 -5.70911050e-01 1.20337680e-02 -2.10289042e-02
-6.01832449e-01 -3.98886740e-01 -5.96453734e-02 1.53713733e-01
5.37797689e-01 -1.25244725e+00 -4.21553046e-01 -3.26580673e-01
-2.34208539e-01 -2.13037178e-01 2.00531095e-01 2.00105608e-01
9.36653376e-01 -1.91442382e+00 2.99279094e-01 2.30980232e-01
1.63380176e-01 3.51410240e-01 -3.50085288e-01 3.25604707e-01
2.45569482e-01 -3.89130004e-02 -1.72027349e-01 1.89179152e-01
1.45697042e-01 5.28590381e-01 -1.85392275e-01 5.65449238e-01
2.67366379e-01 1.22822487e+00 -9.30800617e-01 -1.47154823e-01
1.30061686e-01 3.98717135e-01 -6.89538062e-01 -1.93576157e-01
3.19600731e-01 4.02215719e-01 -5.91068268e-01 4.93639261e-01
5.22289991e-01 -4.26332086e-01 7.22472742e-02 -1.05633095e-01
-2.33280092e-01 4.83622819e-01 -1.16784930e+00 1.58347631e+00
-2.61930019e-01 7.94372916e-01 -2.29145169e-01 -1.17926311e+00
7.14778364e-01 1.36758462e-01 -7.21467137e-02 -1.20221663e+00
4.63968217e-01 4.32190537e-01 5.76856315e-01 -3.77924085e-01
4.78636026e-01 -5.82246304e-01 -1.38839349e-01 5.89430332e-01
5.40390491e-01 -5.75065851e-01 2.28379950e-01 -5.99180907e-02
1.26856589e+00 -3.71790111e-01 1.22929290e-01 -1.70458034e-01
3.15885991e-01 -1.62532985e-01 4.36074138e-01 1.20451224e+00
-6.39542043e-01 2.32532308e-01 2.04598859e-01 -5.48701584e-01
-1.20130074e+00 -1.16501832e+00 -5.44082820e-01 7.35885561e-01
4.09064144e-01 -2.44229496e-01 -7.84140706e-01 3.03300321e-01
-1.93511799e-01 4.68726575e-01 -4.51135695e-01 -5.61726749e-01
-5.12982726e-01 -1.21194708e+00 1.08432806e+00 3.11783522e-01
6.35125577e-01 -9.76081192e-01 -4.82041240e-01 2.92320877e-01
2.51842648e-01 -1.23689723e+00 5.08304015e-02 7.60477602e-01
-1.13154185e+00 -5.44212162e-01 -2.88026154e-01 -7.31932998e-01
4.38150346e-01 -9.54893529e-02 6.37030721e-01 -1.37418240e-01
-3.80621940e-01 -2.34636366e-01 9.90124419e-03 -1.26232207e-01
-6.39871418e-01 1.18247785e-01 6.07084811e-01 -3.30490440e-01
7.68440306e-01 -6.46264493e-01 -4.84176755e-01 -9.52204838e-02
-8.74364316e-01 -2.37516910e-01 6.69039488e-01 1.32594752e+00
4.34471726e-01 2.23405153e-01 3.12866241e-01 -4.02128279e-01
6.90201998e-01 6.93496987e-02 -9.49853003e-01 2.77890205e-01
-5.93047023e-01 8.76940727e-01 8.69036019e-01 -3.16644549e-01
-5.21983802e-01 2.59749126e-02 -1.65725932e-01 2.25096107e-01
-1.46799535e-02 5.34366488e-01 3.99743378e-01 -7.41916001e-01
9.72844720e-01 5.20458400e-01 -3.01975831e-02 1.54986441e-01
4.53315049e-01 8.01499069e-01 6.33034170e-01 -3.68624389e-01
8.50054145e-01 4.12263840e-01 6.63491249e-01 -1.16659665e+00
-5.39081216e-01 -1.07203471e-03 -8.21437120e-01 2.68842727e-01
9.19045269e-01 -6.33565664e-01 -1.29190552e+00 6.48821056e-01
-1.31807470e+00 -1.03528969e-01 -4.13532406e-02 9.42842841e-01
-3.98079515e-01 3.94559741e-01 -1.22291410e+00 -9.47992325e-01
-2.86499143e-01 -1.31645489e+00 5.42812586e-01 5.91601133e-01
3.45000625e-01 -9.38003719e-01 2.10527316e-01 2.09291518e-01
6.13010883e-01 -1.83691174e-01 8.22071433e-01 -9.43665877e-02
-1.10965133e+00 -4.74929571e-01 -2.92658120e-01 5.38894951e-01
-2.46690854e-01 -1.28288940e-01 -9.66104269e-01 -1.11402683e-01
7.74652883e-02 -6.98024809e-01 1.02912688e+00 8.65621865e-03
7.16728270e-01 1.48710191e-01 1.44737229e-01 7.22415805e-01
1.36783266e+00 4.54203606e-01 7.92080224e-01 1.29706606e-01
6.10421658e-01 4.15240452e-02 -2.41219178e-01 1.40575320e-01
1.66549519e-01 3.92224401e-01 3.56230557e-01 5.20588815e-01
4.14849669e-01 9.86301303e-02 3.92717212e-01 1.58812249e+00
-4.99906421e-01 1.22573227e-01 -8.83496284e-01 5.05535118e-02
-1.61234426e+00 -1.13289857e+00 -1.98517010e-01 2.37176681e+00
8.32104504e-01 9.30925235e-02 -6.47521377e-01 4.63567302e-02
4.97402012e-01 -1.50052786e-01 -7.17536867e-01 -6.71902239e-01
-3.52183670e-01 1.01789200e+00 7.78530777e-01 3.56057316e-01
-1.16373575e+00 1.10539770e+00 7.20926094e+00 5.67644954e-01
-1.25962365e+00 2.39081755e-01 5.32049127e-02 1.03199735e-01
1.29315391e-01 1.30902335e-01 -5.97007573e-01 8.11781883e-02
1.46333885e+00 -1.11035064e-01 1.11933041e+00 5.81215084e-01
-4.13035512e-01 -7.21914098e-02 -1.22310638e+00 1.32451403e+00
-1.97863817e-01 -1.54295802e+00 -2.03427821e-01 2.85649568e-01
8.56336653e-01 7.15728581e-01 6.53804764e-02 4.45447326e-01
-3.51738334e-02 -1.22160518e+00 4.80348200e-01 3.56845170e-01
7.50121891e-01 -6.77685380e-01 9.09172654e-01 4.47586656e-01
-6.95328414e-01 -2.74879277e-01 -8.63510966e-01 -7.06795156e-01
-3.44934640e-03 3.65714699e-01 -2.61728734e-01 1.18704155e-01
2.29575902e-01 4.24679339e-01 -2.66309768e-01 8.40760648e-01
-1.94981247e-01 4.55274612e-01 -5.36692083e-01 -8.89285743e-01
4.11505222e-01 -8.05744529e-01 7.42946267e-02 7.97796130e-01
3.73821050e-01 5.09995341e-01 -4.49846625e-01 1.19930553e+00
-2.18138054e-01 -2.49299020e-01 -6.69388652e-01 -5.99612474e-01
4.53724712e-01 1.00513589e+00 -5.86080253e-01 -3.62243533e-01
-8.96919817e-02 1.18001366e+00 4.13151026e-01 2.63681531e-01
-6.00597143e-01 -1.06695795e+00 4.19168413e-01 -7.53632188e-01
1.52818680e-01 -6.26961768e-01 -4.56602126e-01 -1.63456726e+00
-1.67742953e-01 -3.47177625e-01 -4.19976860e-01 -2.86534816e-01
-1.02953839e+00 4.10125762e-01 -6.78426743e-01 -9.08898771e-01
-7.55393282e-02 -1.27190077e+00 -4.33569014e-01 8.18434000e-01
-1.40485656e+00 -3.07939261e-01 1.61857516e-01 3.78433734e-01
-6.52172804e-01 -1.68751001e-01 1.47189116e+00 4.45632100e-01
-7.91545510e-01 5.33326626e-01 8.01606059e-01 3.80378723e-01
2.59665847e-01 -1.17625833e+00 5.40013850e-01 8.69017541e-01
4.49176490e-01 1.24097061e+00 5.68809748e-01 4.54602279e-02
-2.25152564e+00 -5.39890230e-01 7.72440314e-01 -5.22880137e-01
1.00596523e+00 -4.78914946e-01 -7.45321929e-01 1.23769589e-01
2.03485638e-02 1.92901567e-01 8.06480467e-01 2.34636530e-01
-6.04828179e-01 -1.54047370e-01 -8.12627256e-01 5.30012012e-01
8.63711834e-01 -1.28028941e+00 -6.52236283e-01 5.74799776e-01
3.18417341e-01 -2.68515587e-01 -9.80051339e-01 6.67783767e-02
9.68998313e-01 -9.11405683e-01 6.39956355e-01 -6.29112542e-01
2.06227183e-01 -2.14017466e-01 -3.09279293e-01 -1.11248815e+00
-4.22356039e-01 -7.67190099e-01 -1.31761298e-01 1.40059948e-01
3.01597625e-01 -1.01244509e+00 8.64360869e-01 7.40112364e-01
6.91219270e-02 -4.06145960e-01 -1.42760706e+00 -8.95742714e-01
4.51171905e-01 -4.76801068e-01 2.05660805e-01 9.05409396e-01
4.25987840e-01 5.45499027e-01 -2.20520571e-01 2.38916755e-01
9.55918014e-01 7.30000585e-02 2.56863356e-01 -1.09742582e+00
-4.97305483e-01 -5.83497643e-01 -1.05745399e+00 -1.21011603e+00
1.58499792e-01 -1.02107859e+00 2.42316321e-01 -9.18864131e-01
2.53967434e-01 -2.37892300e-01 -7.53798664e-01 3.05808991e-01
1.09384127e-01 6.54992938e-01 2.55272575e-02 1.48118272e-01
-6.21763289e-01 7.58246124e-01 1.26584840e+00 -1.72245994e-01
-3.01912948e-02 -3.63580734e-01 -2.03744322e-01 3.18739384e-01
6.11562192e-01 -6.29838824e-01 2.13374775e-02 -5.56796253e-01
5.16623676e-01 4.17955630e-02 2.62674540e-01 -1.33124840e+00
6.62957668e-01 -2.26385868e-03 5.53679839e-02 1.35164401e-02
5.23454905e-01 -2.00038254e-01 -2.59059936e-01 9.18921411e-01
-1.26634613e-01 -1.73459545e-01 -2.85786450e-01 3.98511320e-01
-3.14861983e-01 -4.81849074e-01 9.66072261e-01 -3.55604813e-02
-6.66049421e-01 3.64120156e-01 -4.43940789e-01 -3.64136040e-01
6.35068536e-01 -9.72259417e-02 -6.37332380e-01 -1.36396572e-01
-6.14211679e-01 -1.44905195e-01 3.62233013e-01 3.05175986e-02
5.00646412e-01 -1.30488884e+00 -2.72112787e-01 4.86273468e-01
1.07637562e-01 -4.26842779e-01 2.53798842e-01 7.12714076e-01
-9.24910307e-01 8.68116677e-01 -5.10183752e-01 -5.35992444e-01
-6.05002224e-01 2.20771238e-01 5.09411395e-01 1.02378771e-01
-1.71965718e-01 1.01044083e+00 -3.19726259e-01 -7.04255164e-01
-7.06883445e-02 -3.88680398e-01 2.74424702e-01 -5.47719836e-01
6.22034013e-01 1.75726414e-01 1.02404922e-01 -5.99226058e-01
-2.78529495e-01 5.76144576e-01 5.45771420e-02 -3.55324626e-01
1.22827137e+00 3.65590602e-01 -5.81642807e-01 5.11873960e-01
1.38520038e+00 -6.89166903e-01 -6.65325463e-01 -3.43118429e-01
-4.97561023e-02 1.49972752e-01 4.73424345e-01 -3.30683112e-01
-7.40548074e-01 1.39227223e+00 7.97520220e-01 2.32862279e-01
7.55510330e-01 -3.42046499e-01 9.35491145e-01 1.74200356e+00
8.99384737e-01 -1.31554067e+00 -2.54052013e-01 1.10529923e+00
-4.47044289e-03 -1.33316624e+00 -1.06949359e-01 2.76916474e-01
-2.16332078e-02 1.60498059e+00 1.79627687e-01 -3.83853525e-01
6.43022001e-01 5.93171902e-02 -6.22670203e-02 -7.13738874e-02
-5.73756874e-01 -1.29968613e-01 1.58261791e-01 3.87709439e-01
4.23652828e-01 4.52818304e-01 -1.93746418e-01 1.89589947e-01
-3.95503700e-01 1.40864238e-01 8.33111286e-01 1.03734660e+00
-5.89977562e-01 -1.40826631e+00 -2.18412250e-01 3.49451333e-01
-2.16003209e-01 -3.79470736e-01 -3.26084457e-02 1.93904161e-01
7.26788566e-02 7.73472726e-01 -9.68980268e-02 -6.94454193e-01
1.33682070e-02 1.53618366e-01 1.04246676e+00 -5.46849489e-01
-8.37828070e-02 -5.68072319e-01 -3.29900503e-01 -4.84315723e-01
-3.57581049e-01 -4.25845146e-01 -1.48193192e+00 -5.34583032e-01
-8.94129515e-01 1.27514988e-01 1.13490665e+00 1.09714735e+00
4.72856671e-01 3.12680632e-01 4.07581538e-01 -7.38174617e-01
-1.20687079e+00 -7.54167736e-01 -8.30753028e-01 5.06196916e-03
4.23948675e-01 -4.71425861e-01 -2.15469107e-01 -2.60725051e-01]
|
[5.564149379730225, 4.947953701019287]
|
a269aef1-cdb2-4762-b90f-3c3222517fc6
|
two-staged-acoustic-modeling-adaption-for
|
1908.06709
| null |
https://arxiv.org/abs/1908.06709v1
|
https://arxiv.org/pdf/1908.06709v1.pdf
|
Two-Staged Acoustic Modeling Adaption for Robust Speech Recognition by the Example of German Oral History Interviews
|
In automatic speech recognition, often little training data is available for specific challenging tasks, but training of state-of-the-art automatic speech recognition systems requires large amounts of annotated speech. To address this issue, we propose a two-staged approach to acoustic modeling that combines noise and reverberation data augmentation with transfer learning to robustly address challenges such as difficult acoustic recording conditions, spontaneous speech, and speech of elderly people. We evaluate our approach using the example of German oral history interviews, where a relative average reduction of the word error rate by 19.3% is achieved.
|
['Joachim köhler', 'Sven Behnke', 'Christoph Schmidt', 'Michael Gref']
|
2019-08-19
| null | null | null | null |
['robust-speech-recognition']
|
['speech']
|
[ 3.11087221e-01 3.71977031e-01 4.09360021e-01 -7.11288750e-01
-1.46192479e+00 -1.36101916e-01 3.44451934e-01 -2.04109903e-02
-6.57223821e-01 7.02338278e-01 5.88768840e-01 -5.97432613e-01
4.40082431e-01 -1.66806579e-01 -4.00434732e-01 -2.95466989e-01
6.83004260e-02 3.89035910e-01 5.86494477e-03 -3.95725578e-01
-3.51783603e-01 2.54791796e-01 -1.36756575e+00 5.15362918e-01
7.13741541e-01 7.79899776e-01 3.92400175e-01 9.85285759e-01
-1.66820705e-01 4.96386439e-01 -1.12632608e+00 -1.32992268e-01
-2.35820293e-01 -2.13863030e-01 -8.36428225e-01 2.73492575e-01
1.97086751e-01 -3.56092364e-01 -4.69376892e-01 8.28963637e-01
1.10867178e+00 4.75997269e-01 3.65058631e-01 -5.07906139e-01
-4.95211512e-01 8.07362258e-01 2.90261984e-01 4.55450386e-01
4.07947898e-01 -7.50130713e-02 6.52449906e-01 -1.09716356e+00
1.32728770e-01 1.36213565e+00 5.86409390e-01 9.70870852e-01
-1.24010408e+00 -4.99712914e-01 3.14880759e-01 2.93393195e-01
-1.22139192e+00 -1.14369035e+00 6.14032984e-01 -2.10967347e-01
1.40460360e+00 2.73762554e-01 2.97994554e-01 1.58438182e+00
-5.50483763e-01 8.93579543e-01 9.10039544e-01 -7.35616207e-01
3.89067352e-01 1.72681853e-01 2.84636348e-01 5.40763326e-02
-4.88995969e-01 8.21517706e-02 -6.24007165e-01 -2.50645339e-01
1.60126731e-01 -5.71247160e-01 -3.43777657e-01 4.92134839e-01
-7.83510923e-01 5.74512303e-01 -7.04090521e-02 5.46716392e-01
-3.95730019e-01 -2.54052788e-01 5.87103844e-01 4.28750217e-01
7.08209276e-01 1.77708343e-01 -8.27262044e-01 -5.70675671e-01
-8.84737551e-01 -6.62119910e-02 9.03849542e-01 8.13307643e-01
2.19733581e-01 6.17117345e-01 1.01470515e-01 1.72509992e+00
3.96318674e-01 7.03872442e-01 7.83749640e-01 -5.23301661e-01
5.88391483e-01 -8.65820050e-02 -1.08776428e-01 -1.73572034e-01
-3.71290088e-01 -5.36321938e-01 -5.32690644e-01 -1.20785274e-01
2.38084555e-01 -3.79861087e-01 -1.27081406e+00 1.62350333e+00
1.00111470e-01 -1.88330598e-02 4.55614984e-01 5.21381080e-01
8.10653508e-01 8.89184415e-01 1.61981821e-01 -5.43670535e-01
1.07719004e+00 -1.08076537e+00 -1.32903099e+00 -8.47240150e-01
5.80568075e-01 -8.02003503e-01 1.29360163e+00 4.41492349e-01
-1.16150904e+00 -6.31373405e-01 -9.95324373e-01 4.05526385e-02
-2.49339476e-01 1.70908555e-01 -5.44179492e-02 1.11063778e+00
-9.95450318e-01 7.94656351e-02 -8.25550258e-01 -2.61970818e-01
9.65432450e-02 4.05875206e-01 -3.15976769e-01 -1.50820330e-01
-1.23932743e+00 9.30339098e-01 2.61514395e-01 8.16241652e-02
-7.39978969e-01 -6.25643313e-01 -8.34451497e-01 2.42631268e-02
3.24020743e-01 -1.34029284e-01 1.87098110e+00 -8.57760727e-01
-1.91251695e+00 6.25558436e-01 -2.16697633e-01 -5.17738402e-01
2.33818278e-01 -7.47242212e-01 -1.02749109e+00 -1.45870551e-01
-4.07136172e-01 5.84121570e-02 6.82026863e-01 -1.04539120e+00
-2.63344854e-01 -4.01885837e-01 -6.31313980e-01 1.70814574e-01
-6.61607385e-01 4.23410892e-01 -1.24499381e-01 -8.68790805e-01
-5.29314652e-02 -6.22798562e-01 -2.22449049e-01 -5.02266586e-01
-6.07976690e-02 -1.82370260e-01 7.30629981e-01 -1.25167239e+00
1.31876326e+00 -2.29683304e+00 -1.16321683e-01 1.96022421e-01
-5.11343122e-01 6.58974767e-01 -1.53204620e-01 1.85900584e-01
8.82753730e-02 -4.53696996e-02 -3.57032001e-01 -8.90034020e-01
-1.42106190e-01 4.79722798e-01 -1.87402323e-01 7.53361434e-02
2.13492736e-01 3.29887420e-01 -7.97075927e-01 -2.93973908e-02
2.38040522e-01 8.06297123e-01 -3.51393819e-01 5.85440993e-01
2.10266374e-02 4.90472019e-01 -2.66998224e-02 5.21205783e-01
1.35822535e-01 4.22060937e-01 1.84937671e-01 1.32995158e-01
-2.20767464e-02 1.04423594e+00 -1.10078239e+00 1.52687931e+00
-7.27785826e-01 6.71914876e-01 5.97264111e-01 -9.96893167e-01
9.20276165e-01 8.06187630e-01 4.11980823e-02 -6.32882595e-01
1.88090548e-01 5.95192432e-01 9.15474221e-02 -5.68974435e-01
2.83439279e-01 -3.11895102e-01 1.75897464e-01 6.97816089e-02
3.07201028e-01 -1.48761064e-01 -3.70073020e-01 -3.18348080e-01
1.20028460e+00 -4.63720620e-01 1.27902761e-01 -1.64674744e-01
5.99814236e-01 -5.32648087e-01 3.29769284e-01 6.52862072e-01
-3.35139811e-01 8.17106843e-01 -3.15672487e-01 -4.97974716e-02
-1.04800284e+00 -9.16503787e-01 -8.06800425e-02 1.38332999e+00
-8.85735631e-01 -4.64948624e-01 -1.12347567e+00 -3.13255310e-01
-4.73553687e-01 9.02827978e-01 -2.07371935e-01 -1.38609260e-01
-8.23588550e-01 -5.18043816e-01 8.24375451e-01 8.04969072e-01
1.28991559e-01 -1.27343440e+00 1.25597075e-01 5.31205416e-01
-3.71594906e-01 -1.56472600e+00 -4.52904642e-01 3.54961485e-01
-6.13599360e-01 -5.18222630e-01 -8.32735538e-01 -1.06971371e+00
2.74089843e-01 -2.89147647e-05 8.72605026e-01 -7.89307356e-02
-9.61122215e-02 4.98838603e-01 -5.70364058e-01 -5.98643601e-01
-1.07969308e+00 1.39489219e-01 4.63396400e-01 1.63160916e-02
2.95319885e-01 -4.96547997e-01 -1.36693075e-01 2.78802574e-01
-6.03335440e-01 -4.47400182e-01 5.55171907e-01 1.15415609e+00
3.32163334e-01 -2.64041871e-01 1.01648533e+00 -4.29076761e-01
7.88587809e-01 -2.73814023e-01 -3.30054909e-02 1.85364649e-01
-2.41840422e-01 -2.51639366e-01 3.05103809e-01 -8.89541864e-01
-1.14462900e+00 4.07207310e-02 -9.52242374e-01 -6.86781481e-02
-4.27535892e-01 6.69275463e-01 -4.11383569e-01 3.46385688e-01
9.07605350e-01 2.58900046e-01 1.47428632e-01 -7.64734268e-01
2.28582874e-01 1.46421874e+00 8.28033686e-01 -3.37168843e-01
3.63611817e-01 -1.82230875e-01 -8.77053440e-01 -1.60093081e+00
-6.75741196e-01 -5.62217832e-01 -3.83354634e-01 -1.67985618e-01
6.71763599e-01 -9.73512769e-01 -9.68948007e-02 5.72353303e-01
-1.23439848e+00 -6.83048010e-01 -3.15682858e-01 7.56289840e-01
-6.00715578e-01 2.71548092e-01 -6.01474822e-01 -1.47048342e+00
-5.31019270e-01 -1.12518859e+00 8.84113491e-01 -3.76772463e-01
-5.20814359e-01 -6.35904491e-01 -4.97366935e-02 6.38646483e-01
5.91341615e-01 -5.69675088e-01 6.78092718e-01 -1.15791357e+00
2.60212332e-01 -2.84014225e-01 3.99099261e-01 1.17904699e+00
4.75462973e-01 -5.28619885e-01 -1.49947298e+00 -3.03067893e-01
2.46823996e-01 -4.82914805e-01 3.95203114e-01 3.52120101e-01
9.06867027e-01 -2.46870279e-01 6.89699501e-02 -1.71357542e-01
4.35315609e-01 5.02932906e-01 6.76809251e-01 9.93322730e-02
4.27768767e-01 7.88559020e-01 3.18464309e-01 2.70875245e-01
1.72828197e-01 7.39615679e-01 -2.31679678e-01 9.37290043e-02
-5.73999345e-01 -7.15926141e-02 6.20555222e-01 1.47490287e+00
2.44238108e-01 -2.86786824e-01 -1.13127267e+00 8.14528763e-01
-1.51437151e+00 -6.14002526e-01 2.37258077e-01 2.08127689e+00
1.07204604e+00 2.68192977e-01 1.95617944e-01 6.48596883e-01
6.65865600e-01 -1.10751167e-01 -1.40838146e-01 -2.71492958e-01
-1.01795197e-01 3.24517161e-01 -2.11996194e-02 8.26029181e-01
-9.42142904e-01 8.54426384e-01 7.77500486e+00 7.30916619e-01
-8.75805140e-01 2.88226485e-01 6.13439620e-01 -8.35238695e-02
-6.51064590e-02 -6.59046531e-01 -6.42826438e-01 2.89200574e-01
1.65238881e+00 3.11344117e-01 3.91315371e-01 8.38892817e-01
4.01941687e-01 3.41793835e-01 -1.02942383e+00 1.05887389e+00
2.60296643e-01 -6.19542301e-01 -2.66250998e-01 -1.44101068e-01
3.41271281e-01 2.08674371e-01 2.12738104e-02 5.38083494e-01
-2.30782680e-04 -1.01124644e+00 7.62552261e-01 1.02787524e-01
6.48553431e-01 -5.66609383e-01 7.07011282e-01 4.13462669e-01
-7.79216409e-01 -2.44726703e-01 4.74633500e-02 -2.80064233e-02
4.44788307e-01 4.46122378e-01 -1.13846660e+00 7.86661431e-02
7.65168548e-01 3.08258291e-02 1.47674317e-02 8.75871181e-01
-2.33396515e-01 1.27280509e+00 -5.71226835e-01 -1.30868420e-01
-8.62572044e-02 3.56551737e-01 5.45896411e-01 1.50733316e+00
4.59588885e-01 4.30498153e-01 1.73848256e-01 2.15382323e-01
-2.08200812e-02 3.99568021e-01 -4.14209247e-01 -9.13617387e-02
6.46462679e-01 5.13851225e-01 -9.07900855e-02 -3.63953948e-01
-4.80525851e-01 8.26989710e-01 2.11561412e-01 5.01573563e-01
-1.76724628e-01 -5.45245148e-02 6.85437560e-01 1.50751069e-01
2.94878304e-01 -5.92671633e-01 -3.48755151e-01 -8.23798358e-01
2.62610406e-01 -1.25458324e+00 2.92319730e-02 -6.46732748e-01
-1.21984291e+00 1.00132811e+00 -2.02718586e-01 -6.76420689e-01
-6.51353657e-01 -7.41468012e-01 -3.76762003e-01 9.59267855e-01
-1.31913900e+00 -9.90959048e-01 7.11834282e-02 4.10129279e-01
1.28641212e+00 -6.24119699e-01 1.34579861e+00 7.34795451e-01
-3.98260295e-01 8.22270036e-01 6.81878924e-02 6.80814087e-02
6.07460558e-01 -1.08975112e+00 8.50763619e-01 7.42187619e-01
1.59463167e-01 3.53072405e-01 8.81541848e-01 -5.35974920e-01
-9.99891400e-01 -7.71951616e-01 1.05190563e+00 -2.64833272e-01
8.15826237e-01 -6.38641477e-01 -1.24859762e+00 5.17576635e-01
3.59775238e-02 -1.60629041e-02 9.65196908e-01 4.83716786e-01
-3.01837146e-01 -8.15487653e-02 -8.49468529e-01 5.99534690e-01
9.08513129e-01 -9.94871378e-01 -8.34683836e-01 3.52821082e-01
8.47641766e-01 -2.66081840e-01 -8.03787351e-01 3.83658350e-01
3.41643065e-01 -3.10373902e-01 8.96488547e-01 -6.37133956e-01
-3.49542499e-01 1.50543243e-01 -5.51291943e-01 -1.68380296e+00
9.36822444e-02 -9.48146164e-01 -3.20412368e-01 1.38868165e+00
7.89165676e-01 -3.75381649e-01 4.75164384e-01 7.57935047e-01
-7.13919759e-01 -2.83964008e-01 -1.24719727e+00 -9.04262722e-01
9.46974233e-02 -9.67726052e-01 1.19849868e-01 5.38063824e-01
2.10583985e-01 4.51834381e-01 -4.94633526e-01 2.51611114e-01
1.67185679e-01 -1.11367154e+00 4.32691365e-01 -1.06540191e+00
-3.28893334e-01 -1.79969504e-01 -3.72684121e-01 -9.75995064e-01
3.63583088e-01 -2.96561778e-01 7.17144728e-01 -1.49712157e+00
-4.50391740e-01 -1.64356187e-01 -1.80685416e-01 4.86558229e-01
-1.76023185e-01 -1.77622139e-01 -1.22603416e-01 -3.15368652e-01
-5.98738678e-02 8.34842026e-01 5.14889538e-01 -3.90311867e-01
-4.72425342e-01 3.32543522e-01 -3.35808605e-01 6.31194413e-01
8.30268741e-01 -1.66309923e-01 -5.10955870e-01 -5.88735104e-01
-4.19165999e-01 2.14888132e-03 -4.79003601e-02 -1.09377599e+00
1.05385855e-01 3.01391721e-01 -5.04420623e-02 -2.44297028e-01
8.97493958e-01 -6.63271546e-01 -1.91102967e-01 2.63331890e-01
-4.70784038e-01 -2.39559084e-01 5.75426459e-01 4.40820515e-01
-3.62966061e-01 -2.54734069e-01 6.97351873e-01 4.39480916e-02
-3.36333066e-01 -2.57348120e-01 -1.02078986e+00 3.66877355e-02
3.02879155e-01 8.66497681e-02 -9.07144323e-02 -6.34150326e-01
-1.28819668e+00 -1.16933964e-01 -3.65960002e-01 8.15756023e-01
6.97429359e-01 -1.22129345e+00 -1.03355992e+00 5.22739530e-01
6.28668144e-02 -1.30563229e-01 2.09966004e-01 4.30926919e-01
1.27927780e-01 2.00035214e-01 3.23958784e-01 -2.33768821e-01
-1.64893365e+00 1.46169931e-01 4.79305834e-01 1.60089374e-01
-6.23216748e-01 9.44253087e-01 -1.92070872e-01 -5.71556926e-01
8.37363839e-01 -3.76377076e-01 -1.86574340e-01 -7.72127882e-02
9.48626816e-01 5.09551167e-01 6.77007198e-01 -7.97924459e-01
-2.76499510e-01 -6.27898425e-02 -2.48145252e-01 -7.39991248e-01
1.27831769e+00 -1.86415181e-01 4.60777760e-01 7.98755169e-01
1.14094424e+00 -1.76244855e-01 -9.04731512e-01 -5.42308509e-01
2.14647591e-01 -5.97853621e-04 2.74865329e-01 -1.06122041e+00
-4.67019141e-01 9.36221182e-01 9.17969227e-01 4.22972858e-01
1.03373682e+00 9.95869860e-02 9.62311745e-01 7.55646765e-01
5.66113666e-02 -1.49562073e+00 1.46979645e-01 7.99195111e-01
1.21478045e+00 -1.28446686e+00 -5.85174739e-01 -5.25486350e-01
-5.66547573e-01 8.67440045e-01 3.91939044e-01 3.83058041e-01
8.37593734e-01 5.76757669e-01 5.59058309e-01 2.40153685e-01
-4.92445290e-01 -1.99083596e-01 3.89993608e-01 7.88982987e-01
5.29120624e-01 8.08764249e-02 -3.84444818e-02 8.79440665e-01
-4.96450901e-01 -4.64655638e-01 2.15620920e-01 1.04295528e+00
-7.12500989e-01 -1.18790901e+00 -5.31692743e-01 1.46183103e-01
-7.02287138e-01 -3.97773236e-01 -3.92327785e-01 2.52182662e-01
-4.25092041e-01 1.48149955e+00 -4.10288870e-02 -3.79434854e-01
7.97771811e-01 6.72229826e-01 1.82142749e-01 -9.38367307e-01
-5.62135100e-01 5.73227704e-01 6.60686553e-01 -7.12059066e-02
-4.67151284e-01 -7.90441155e-01 -1.24463356e+00 4.37515527e-01
-4.02820885e-01 2.16516823e-01 1.03716552e+00 1.24924338e+00
5.49806952e-02 9.59383845e-01 3.97353292e-01 -7.52854347e-01
-8.19575489e-01 -1.63678181e+00 -3.52472782e-01 1.09520771e-01
7.40014017e-01 -2.95627117e-01 -4.73727643e-01 1.87967286e-01]
|
[14.499161720275879, 6.449317932128906]
|
832fa60b-9aa6-49bc-a9bf-4412d6935a18
|
improving-deep-image-matting-via-local
|
2112.13809
| null |
https://arxiv.org/abs/2112.13809v2
|
https://arxiv.org/pdf/2112.13809v2.pdf
|
Improving Deep Image Matting via Local Smoothness Assumption
|
Natural image matting is a fundamental and challenging computer vision task. Conventionally, the problem is formulated as an underconstrained problem. Since the problem is ill-posed, further assumptions on the data distribution are required to make the problem well-posed. For classical matting methods, a commonly adopted assumption is the local smoothness assumption on foreground and background colors. However, the use of such assumptions was not systematically considered for deep learning based matting methods. In this work, we consider two local smoothness assumptions which can help improving deep image matting models. Based on the local smoothness assumptions, we propose three techniques, i.e., training set refinement, color augmentation and backpropagating refinement, which can improve the performance of the deep image matting model significantly. We conduct experiments to examine the effectiveness of the proposed algorithm. The experimental results show that the proposed method has favorable performance compared with existing matting methods.
|
['Dezhen Qi', 'Jiacheng Han', 'Jun Xie', 'Rui Wang']
|
2021-12-27
| null | null | null | null |
['image-matting']
|
['computer-vision']
|
[ 3.51443172e-01 -1.84244409e-01 -1.36394605e-01 -3.03295076e-01
-3.64061475e-01 -2.32545529e-02 3.79310936e-01 -4.06226248e-01
-2.42488325e-01 6.23905241e-01 -6.19025230e-02 -3.24843079e-01
1.83782727e-01 -5.91876805e-01 -7.34908581e-01 -9.48483527e-01
5.49736917e-01 1.54652715e-01 1.25125438e-01 1.00485139e-01
2.09366605e-01 1.52571216e-01 -1.19319844e+00 1.12317190e-01
1.30608273e+00 8.89577806e-01 3.12326044e-01 3.11935931e-01
-5.32426894e-01 9.86487985e-01 -3.59925598e-01 -2.70368785e-01
3.88693243e-01 -6.47414148e-01 -5.82399487e-01 7.65117645e-01
4.04091865e-01 -4.94563431e-01 -3.40269297e-01 1.46599782e+00
-2.19813902e-02 1.24647759e-01 4.41172808e-01 -1.23326266e+00
-5.76128602e-01 4.07539606e-01 -1.05596757e+00 2.10023522e-01
-1.42509311e-01 1.09971151e-01 6.41311288e-01 -9.71002936e-01
1.26936391e-01 1.20774591e+00 4.80457753e-01 3.32930267e-01
-9.43774819e-01 -7.17113316e-01 6.18600070e-01 2.89148092e-01
-1.34309232e+00 -3.60178947e-01 9.89078164e-01 -3.91861379e-01
6.19918108e-02 1.88183472e-01 5.73820233e-01 4.13967371e-01
1.47809789e-01 1.06607509e+00 1.60850263e+00 -4.00636494e-01
1.54785916e-01 9.26576927e-02 -4.35106903e-02 8.77151728e-01
4.64540511e-01 -3.16989124e-01 -4.24115062e-01 -6.38784021e-02
1.14152658e+00 3.32462877e-01 -3.63800287e-01 -2.66574323e-01
-1.11101043e+00 5.56200206e-01 5.54073036e-01 7.59601817e-02
-4.43415076e-01 2.19096407e-01 -4.55432339e-03 7.72330537e-02
7.85591543e-01 -2.48518199e-01 -2.38238461e-02 2.27615044e-01
-1.26832020e+00 1.06983341e-01 3.40555102e-01 9.40890133e-01
8.89268517e-01 5.13973296e-01 8.75583217e-02 7.58784235e-01
6.62612617e-01 5.55439234e-01 1.70785263e-01 -5.84894657e-01
5.19002140e-01 5.98907530e-01 1.65958360e-01 -1.22359753e+00
8.87549520e-02 -3.91928375e-01 -1.15053165e+00 2.98468471e-01
3.43837172e-01 -2.07371920e-01 -1.41115403e+00 1.48560131e+00
5.74027121e-01 7.75954366e-01 -2.19947711e-01 1.12201393e+00
7.16931224e-01 8.43272269e-01 -1.25176787e-01 -4.18489665e-01
8.49012136e-01 -1.02827871e+00 -1.14048529e+00 -2.13722929e-01
8.52143615e-02 -1.02647054e+00 9.20888007e-01 4.67039943e-01
-1.17827868e+00 -5.26821375e-01 -1.15906477e+00 1.58436790e-01
1.08529612e-01 -3.73294130e-02 7.48382807e-01 6.45654202e-01
-7.10024059e-01 1.77998871e-01 -9.88634765e-01 6.32993691e-03
4.90101367e-01 2.91427583e-01 -1.14994191e-01 -3.13875377e-01
-7.74006903e-01 6.39786303e-01 3.67920637e-01 6.70183003e-01
-1.06006336e+00 -2.33113140e-01 -7.21392453e-01 -1.55555129e-01
4.46204573e-01 -5.04510045e-01 1.10877192e+00 -1.23167813e+00
-1.48061299e+00 8.17620456e-01 -4.43914175e-01 -2.92597920e-01
7.44111180e-01 -4.80761915e-01 -1.19965181e-01 -6.26364052e-02
-1.44515023e-01 2.34805897e-01 1.16350675e+00 -1.58937025e+00
-5.39876759e-01 -1.84986025e-01 5.42208739e-02 4.47518647e-01
-2.91869193e-01 -5.58067337e-02 -6.58800542e-01 -9.51014459e-01
4.19144750e-01 -7.89561749e-01 -4.00473773e-01 9.86729115e-02
-5.30526340e-01 3.01866710e-01 8.72598290e-01 -8.88263881e-01
9.83963490e-01 -2.12277365e+00 1.41409963e-01 2.59179860e-01
2.94442594e-01 2.70175070e-01 1.18829429e-01 5.73481172e-02
6.52568117e-02 -7.17291757e-02 -5.98108172e-01 -3.86866778e-01
-1.97333276e-01 3.14123124e-01 -2.48928964e-01 7.63921499e-01
6.87653571e-02 6.37097478e-01 -5.58035076e-01 -5.54491520e-01
3.97560626e-01 4.01814997e-01 -3.42417508e-01 5.39971888e-01
-3.57738376e-01 5.10673881e-01 -4.52390015e-01 5.54382682e-01
1.28157222e+00 -1.80841118e-01 -8.40369612e-02 -2.72263229e-01
-4.11836989e-02 2.03399555e-04 -1.51456356e+00 1.45034492e+00
-2.22408757e-01 3.82153183e-01 4.29044545e-01 -1.05490243e+00
9.94827747e-01 1.13870300e-01 4.38363463e-01 -2.72346199e-01
3.35236341e-01 4.86798398e-02 1.02162361e-01 -5.10869384e-01
1.60683006e-01 -5.15520334e-01 6.42249703e-01 3.16872418e-01
-2.81292528e-01 2.08687000e-02 -2.21914142e-01 7.52607733e-02
6.38771415e-01 1.50663286e-01 -1.73251405e-02 -3.45580935e-01
5.67567289e-01 1.09948240e-01 9.60406780e-01 4.39573675e-01
-1.25212679e-02 6.60249054e-01 1.74755767e-01 -3.12883496e-01
-7.87355363e-01 -9.89857733e-01 1.32098004e-01 7.22052634e-01
7.17332423e-01 -2.20222417e-02 -9.32835758e-01 -6.25812709e-01
-2.57121116e-01 3.63049626e-01 -5.92700243e-01 -5.93802296e-02
-7.25445926e-01 -1.11679912e+00 1.38476297e-01 4.45106894e-01
1.19734478e+00 -8.29959452e-01 -1.34024590e-01 -2.78198663e-02
-2.27156594e-01 -1.10360622e+00 -6.01884782e-01 -1.78217471e-01
-1.14895427e+00 -9.63652968e-01 -1.07711208e+00 -9.79728043e-01
1.12305999e+00 7.90653706e-01 7.24899769e-01 6.31016910e-01
2.77489014e-02 3.05043980e-02 -3.08166891e-01 -3.54581863e-01
-1.36148974e-01 -3.15658599e-01 -7.23291263e-02 4.33166832e-01
3.05497972e-03 -4.25456792e-01 -6.50730729e-01 3.28958631e-01
-1.16929960e+00 6.79816723e-01 1.02791512e+00 8.29603434e-01
7.17817247e-01 3.13640505e-01 3.32217991e-01 -1.28099954e+00
3.68410349e-01 -3.35323900e-01 -5.08695126e-01 2.75595754e-01
-4.73431200e-01 3.02541144e-02 4.50856209e-01 -5.32207072e-01
-1.44384801e+00 7.20233992e-02 -1.15684476e-02 -6.90194905e-01
-1.65404484e-01 7.18468845e-01 -5.54420114e-01 -4.10303682e-01
-1.03331301e-02 6.33272946e-01 3.67147550e-02 -6.19899273e-01
2.92122275e-01 3.79847825e-01 4.75633919e-01 -7.26589978e-01
1.16966772e+00 5.50494075e-01 -1.67807907e-01 -8.04473221e-01
-8.57109725e-01 -3.88468355e-01 -4.70822036e-01 -3.14444542e-01
8.58212948e-01 -8.76645148e-01 -2.91257113e-01 7.78455496e-01
-8.92146409e-01 -4.41077769e-01 3.23610514e-01 2.95339078e-01
-3.43513995e-01 6.34814858e-01 -5.99934757e-01 -9.85122859e-01
-2.54295766e-01 -1.22657251e+00 6.08844221e-01 4.21738595e-01
4.82980192e-01 -1.09306800e+00 -1.72043055e-01 4.04133081e-01
4.47526425e-01 5.14975727e-01 7.25349188e-01 1.38690760e-02
-9.60515380e-01 5.28492555e-02 -4.27579701e-01 4.02466863e-01
4.88987386e-01 -8.36473703e-02 -8.53534460e-01 -1.69497371e-01
3.62318009e-01 -7.15660155e-02 1.04651499e+00 4.33608294e-01
1.37268603e+00 -4.19251740e-01 -6.50467947e-02 8.76819789e-01
1.44249356e+00 2.13203222e-01 7.49042869e-01 2.95844704e-01
1.00367820e+00 2.45648474e-01 7.83716261e-01 4.37606901e-01
2.14896530e-01 3.31842989e-01 6.17802501e-01 -6.51351094e-01
2.87252609e-02 -5.73560931e-02 1.75951317e-01 1.08336926e+00
-2.73138225e-01 -3.89816724e-02 -6.31971955e-01 3.52109700e-01
-2.12366056e+00 -7.46959448e-01 -3.36406738e-01 2.12636781e+00
9.39137816e-01 3.61069679e-01 -7.31227919e-02 1.62251666e-01
8.18529725e-01 2.07174793e-01 -6.48688614e-01 -2.52990872e-02
-3.04130521e-02 9.08673108e-02 3.94746363e-01 4.88306224e-01
-1.14235210e+00 9.89683509e-01 6.01039505e+00 7.21750617e-01
-1.17377746e+00 5.03408611e-02 8.26846898e-01 2.53398091e-01
-3.82242709e-01 6.67083189e-02 -4.83720362e-01 5.85549891e-01
-6.27477169e-02 1.43630644e-02 4.58375841e-01 6.44789934e-01
3.65966380e-01 -2.10075468e-01 -7.84200072e-01 1.13721311e+00
1.35678500e-01 -1.16067493e+00 1.91575006e-01 -6.00498244e-02
9.07473683e-01 -5.03275096e-01 2.62810200e-01 1.53363105e-02
2.16446877e-01 -1.09127128e+00 7.78943121e-01 5.75591028e-01
5.41035831e-01 -8.10924590e-01 7.60925770e-01 4.20293510e-01
-1.13510096e+00 4.12420899e-01 -3.21502060e-01 -2.36957282e-01
1.87912479e-01 8.21841836e-01 -3.70992690e-01 5.33842325e-01
5.14930904e-01 8.50870788e-01 -3.87576818e-01 1.35878026e+00
-2.87084013e-01 8.94898355e-01 -2.35761687e-01 3.01802576e-01
3.20326388e-01 -5.03628075e-01 2.60590136e-01 1.04928195e+00
4.31716954e-03 2.54634947e-01 4.00031835e-01 9.16041315e-01
-9.74251255e-02 5.69300912e-02 -3.21557790e-01 1.96167096e-01
1.22906677e-01 1.23421681e+00 -8.30460489e-01 -3.56628597e-01
-5.09300828e-01 1.07603824e+00 1.14411540e-01 6.67860448e-01
-1.04912746e+00 -6.93164915e-02 5.65956533e-01 1.92929849e-01
1.16427653e-01 -4.77868110e-01 -6.04730546e-01 -1.39059222e+00
1.08609177e-01 -1.06499815e+00 2.52491027e-01 -6.15024626e-01
-1.15487790e+00 3.27710778e-01 7.11937994e-02 -1.12516475e+00
5.21166503e-01 -5.31122386e-01 -1.03011191e+00 9.51134324e-01
-1.60455120e+00 -1.18503249e+00 -7.29435623e-01 7.54608214e-01
6.72324181e-01 8.42520148e-02 1.94864228e-01 3.87632906e-01
-8.58181238e-01 4.01322752e-01 8.24674591e-02 1.13706596e-01
5.70948541e-01 -1.24755740e+00 -2.10952945e-02 1.41944075e+00
3.86677943e-02 7.93827891e-01 8.37718368e-01 -7.92823970e-01
-1.35550559e+00 -1.23810577e+00 7.74126872e-02 2.97516435e-01
3.94443125e-01 -3.12476307e-01 -1.20076346e+00 7.55070865e-01
5.42578101e-01 -6.06984980e-02 4.43990946e-01 -1.75727472e-01
-5.68906963e-02 -9.01210234e-02 -9.94291663e-01 5.95724344e-01
7.25208342e-01 -5.70435114e-02 -4.88828421e-01 2.91995555e-01
4.33631986e-01 -5.71823835e-01 -4.33562726e-01 4.39227611e-01
2.29684517e-01 -7.56425381e-01 9.45800960e-01 -3.78934681e-01
3.24745953e-01 -6.53002441e-01 -1.00618731e-02 -1.13356149e+00
-2.76303291e-01 -5.66278696e-01 -2.70129502e-01 1.39339077e+00
8.19335282e-02 -5.54380178e-01 9.71841156e-01 7.47876346e-01
-1.14131056e-01 -8.16017628e-01 -5.43870449e-01 -5.10449350e-01
-1.23932794e-01 -2.21326351e-02 3.92779797e-01 1.10090077e+00
-4.57226783e-01 5.83386936e-05 -8.22878957e-01 2.68880934e-01
9.20503855e-01 2.11554155e-01 9.36626792e-01 -8.40391934e-01
-3.52212399e-01 -3.43897611e-01 -1.40310869e-01 -1.34851611e+00
-7.83169866e-02 -5.86327791e-01 4.46438164e-01 -1.77784300e+00
4.48064089e-01 -6.36747301e-01 -4.64856446e-01 3.49285215e-01
-8.07612896e-01 1.54597193e-01 8.04686099e-02 3.44114035e-01
-4.84544516e-01 8.20753098e-01 1.48331714e+00 -3.90984565e-01
-8.84052366e-03 5.84137291e-02 -5.80190182e-01 9.69844043e-01
9.86299694e-01 -3.72577906e-01 -4.78160292e-01 -8.08522403e-01
-1.35719433e-01 -1.69573903e-01 2.44788259e-01 -8.74461293e-01
5.31136841e-02 -7.79211283e-01 4.90587562e-01 -5.77412724e-01
3.92438918e-01 -7.99763322e-01 8.99240226e-02 4.26566899e-01
-1.32493585e-01 2.40000244e-02 2.65723556e-01 5.80609024e-01
-2.60770351e-01 -2.14246973e-01 1.03330386e+00 -2.23361969e-01
-7.19190776e-01 5.08366227e-01 -2.36050114e-01 7.03511462e-02
9.85719264e-01 -1.14849970e-01 4.30141203e-02 -3.82780254e-01
-4.94400889e-01 1.81763768e-01 5.41084886e-01 7.13723004e-02
1.02990866e+00 -1.36791813e+00 -6.92208350e-01 1.61222205e-01
-2.45237395e-01 3.41637701e-01 1.48154691e-01 9.94214773e-01
-8.35646749e-01 -2.66837090e-01 -1.08128436e-01 -5.21543562e-01
-1.27926397e+00 5.77011287e-01 2.63553292e-01 -2.31850590e-03
-7.32209921e-01 8.96000326e-01 5.64644217e-01 5.58754243e-02
3.84914070e-01 -4.11548853e-01 -2.88950466e-02 -6.29118145e-01
5.62102377e-01 1.64937243e-01 -2.21300408e-01 -4.99575913e-01
-2.18974367e-01 5.03409147e-01 -3.07947069e-01 -3.30929048e-02
1.16906679e+00 -2.85354227e-01 -3.91791314e-01 3.42328697e-01
6.21304393e-01 -1.19487077e-01 -1.24587595e+00 -4.43240494e-01
-2.50838727e-01 -8.93150926e-01 2.85765320e-01 -3.50466400e-01
-1.41851246e+00 1.07163143e+00 4.67994034e-01 3.02092116e-02
1.26826549e+00 -4.07976985e-01 9.88451064e-01 7.32136145e-02
2.61270821e-01 -8.79036367e-01 2.35764340e-01 2.67326385e-01
7.44112790e-01 -1.34291065e+00 2.76274532e-01 -7.32519984e-01
-5.15026689e-01 9.28465962e-01 1.02230144e+00 -3.65349114e-01
6.17474616e-01 2.17445761e-01 2.39149928e-01 1.41194919e-02
-2.49185294e-01 -1.80888459e-01 3.81950140e-01 2.31144324e-01
3.41267526e-01 -1.03574149e-01 -1.73457503e-01 2.78604805e-01
3.25251013e-01 2.23062430e-02 3.95060718e-01 1.00650942e+00
-7.09796965e-01 -8.70993137e-01 -7.63615429e-01 4.61620986e-01
-5.37065864e-01 -7.35233277e-02 -3.09649646e-01 5.11813462e-01
1.69361219e-01 9.75277305e-01 -3.06288660e-01 -2.73460895e-01
-6.42289817e-02 -2.73956329e-01 5.79798877e-01 -6.25944078e-01
-1.09381676e-01 3.39416355e-01 -3.76841426e-01 -1.81367189e-01
-7.12643743e-01 -4.77503747e-01 -1.23565245e+00 -4.12590414e-01
-7.49157369e-01 -1.20762428e-02 3.54538888e-01 1.24037290e+00
-1.60933718e-01 6.11726224e-01 4.93427694e-01 -6.94559455e-01
-2.24766150e-01 -8.60134602e-01 -5.27780354e-01 4.41929430e-01
3.15813929e-01 -8.29819739e-01 -1.98487431e-01 5.27488112e-01]
|
[10.654693603515625, -0.9530157446861267]
|
aa71d913-c0ea-4065-a1ef-5e00b1a5dd49
|
a-pragmatic-machine-learning-approach-to
|
2202.06590
| null |
https://arxiv.org/abs/2202.06590v1
|
https://arxiv.org/pdf/2202.06590v1.pdf
|
A Pragmatic Machine Learning Approach to Quantify Tumor Infiltrating Lymphocytes in Whole Slide Images
|
Increased levels of tumor infiltrating lymphocytes (TILs) in cancer tissue indicate favourable outcomes in many types of cancer. Manual quantification of immune cells is inaccurate and time consuming for pathologists. Our aim is to leverage a computational solution to automatically quantify TILs in whole slide images (WSIs) of standard diagnostic haematoxylin and eosin stained sections (H&E slides) from lung cancer patients. Our approach is to transfer an open source machine learning method for segmentation and classification of nuclei in H&E slides trained on public data to TIL quantification without manual labeling of our data. Our results show that additional augmentation improves model transferability when training on few samples/limited tissue types. Models trained with sufficient samples/tissue types do not benefit from our additional augmentation policy. Further, the resulting TIL quantification correlates to patient prognosis and compares favorably to the current state-of-the-art method for immune cell detection in non-small lung cancer (current standard CD8 cells in DAB stained TMAs HR 0.34 95% CI 0.17-0.68 vs TILs in HE WSIs: HoVer-Net PanNuke Aug Model HR 0.30 95% CI 0.15-0.60, HoVer-Net MoNuSAC Aug model HR 0.27 95% CI 0.14-0.53). Moreover, we implemented a cloud based system to train, deploy and visually inspect machine learning based annotation for H&E slides. Our pragmatic approach bridges the gap between machine learning research, translational clinical research and clinical implementation. However, validation in prospective studies is needed to assert that the method works in a clinical setting.
|
['Thomas K. Kilvaer', 'Lars Ailo Bongo', 'Ruth Schwienbacher', 'Lill-Tove Rasmussen Busund', 'Kajsa Møllersen', 'Edvard Pedersen', 'Morten Grønnesby', 'Nikita Shvetsov']
|
2022-02-14
| null | null | null | null |
['cell-detection']
|
['computer-vision']
|
[ 6.74662665e-02 2.21706316e-01 -3.83287400e-01 1.63376063e-01
-1.26779616e+00 -7.35870063e-01 3.70803595e-01 6.53966248e-01
-8.25994074e-01 6.61807835e-01 -9.14979950e-02 -8.65489960e-01
3.78535032e-01 -8.72065008e-01 -1.42737344e-01 -1.19207132e+00
2.72351533e-01 1.06483388e+00 1.79504365e-01 1.66009814e-01
-1.73995316e-01 8.31327677e-01 -7.27213204e-01 3.51881981e-01
2.59606808e-01 7.56831944e-01 6.92096446e-03 1.42630494e+00
-1.54076427e-01 6.64364755e-01 -2.21801758e-01 -2.31595207e-02
9.89515781e-02 -2.42070645e-01 -7.97865629e-01 -1.02775730e-01
4.14004564e-01 -9.47156474e-02 1.57028675e-01 6.11037970e-01
5.99152327e-01 -5.54388583e-01 1.05957437e+00 -1.07093787e+00
1.33223683e-01 7.06068650e-02 -6.88937366e-01 3.79598707e-01
-4.16148663e-01 5.06437957e-01 7.62057483e-01 -6.49149179e-01
9.46652889e-01 4.64763403e-01 1.03704143e+00 6.26595855e-01
-1.25396001e+00 -5.91162324e-01 -6.18542433e-01 -9.08295661e-02
-1.56297171e+00 2.42487080e-02 -2.03820422e-01 -7.71453142e-01
9.66440320e-01 7.31297374e-01 8.79024088e-01 3.77722710e-01
5.09459555e-01 4.68067110e-01 1.40474141e+00 -4.82678980e-01
2.43818000e-01 6.06100976e-01 8.40447247e-02 8.14605176e-01
3.60071063e-01 -1.81252673e-01 -7.73915788e-03 -1.90270036e-01
6.19830489e-01 2.34949261e-01 -9.28967744e-02 2.04422176e-02
-1.23439932e+00 8.65251839e-01 4.02310818e-01 6.22893989e-01
-1.65137738e-01 1.21736310e-01 7.26779163e-01 6.04793467e-02
5.46410680e-01 7.55844936e-02 -2.68178970e-01 1.41007915e-01
-1.17617035e+00 -1.62003279e-01 4.51668888e-01 1.21978745e-01
4.22677308e-01 -5.09833992e-01 -1.99362382e-01 6.07579470e-01
2.14635819e-01 5.23419678e-01 6.90273106e-01 -5.11767924e-01
-3.27740341e-01 7.09475935e-01 -6.43299893e-02 -3.00952703e-01
-7.13394165e-01 -4.49634761e-01 -9.79542732e-01 2.67127186e-01
7.35182047e-01 4.46673669e-02 -8.04456353e-01 1.06503737e+00
5.02627611e-01 -5.93243428e-02 -1.19863212e-01 8.05982947e-01
5.14212668e-01 1.09111376e-01 4.27280217e-01 -6.75563663e-02
1.70615578e+00 -5.94751596e-01 -4.59586948e-01 1.17388539e-01
1.50878978e+00 -7.52259851e-01 1.27269375e+00 1.52585074e-01
-7.69472241e-01 9.56622139e-02 -7.52566874e-01 -4.22368571e-03
-5.87726772e-01 3.85678351e-01 7.01973617e-01 9.55842912e-01
-1.40424621e+00 2.47851506e-01 -1.08610356e+00 -8.47996533e-01
1.04757142e+00 6.81224644e-01 -5.84985793e-01 3.57001796e-02
-4.16020095e-01 9.41402972e-01 1.85190991e-01 -1.17119707e-01
-7.43389308e-01 -1.03303242e+00 -3.27450275e-01 -4.17153716e-01
-1.19087450e-01 -9.14558530e-01 1.30206764e+00 -7.39173412e-01
-1.01269019e+00 1.61406040e+00 -2.14573339e-01 -3.56504560e-01
5.63969612e-01 5.66616654e-01 6.73075542e-02 3.70271891e-01
4.01853099e-02 8.26324105e-01 2.98247598e-02 -9.02771175e-01
-9.60795164e-01 -4.47660536e-01 -5.19361198e-01 -9.39493701e-02
-3.96450579e-01 -5.22022173e-02 -7.54682943e-02 -3.45684260e-01
-5.13010800e-01 -1.08220840e+00 -4.65722948e-01 4.07292247e-01
-1.90444529e-01 -1.41756549e-01 8.48077655e-01 -5.88884652e-01
9.13864791e-01 -1.78228426e+00 -4.64191318e-01 4.14886981e-01
3.84571016e-01 2.86928326e-01 1.07406527e-01 -6.95797801e-02
5.08367345e-02 5.84756613e-01 2.18070075e-02 -1.23952635e-01
-4.23598886e-02 8.22489485e-02 3.60972613e-01 1.00488007e+00
6.87370449e-02 1.17042029e+00 -6.79698706e-01 -1.09124887e+00
2.53735155e-01 5.66092849e-01 -3.95907283e-01 1.20359115e-01
3.27719599e-02 2.71427721e-01 -7.27867559e-02 8.14284623e-01
4.55482304e-01 -5.99037588e-01 3.43264937e-01 -1.38970792e-01
-8.09872821e-02 -1.04841426e-01 -5.02752364e-01 1.15792048e+00
-4.45444077e-01 8.25959444e-01 3.19211751e-01 -5.50872505e-01
3.40189040e-01 5.15708983e-01 5.12566030e-01 -1.81921035e-01
6.37255430e-01 2.16902003e-01 7.85711631e-02 -3.53211641e-01
-1.44968152e-01 -8.64162266e-01 2.45911896e-01 5.60141444e-01
-1.70492962e-01 -3.94610912e-01 2.00075567e-01 1.98494092e-01
1.31922972e+00 -5.12676656e-01 4.82947618e-01 -4.34012681e-01
5.97229600e-01 6.25864923e-01 9.91361588e-02 4.07221526e-01
-4.33888614e-01 4.85143244e-01 6.85004711e-01 -2.85303801e-01
-1.04604948e+00 -8.09949160e-01 -4.94947970e-01 8.57911229e-01
-6.36015415e-01 -9.04984251e-02 -8.06302905e-01 -8.64125490e-01
-1.86881840e-01 2.71636009e-01 -9.10317600e-01 4.50914204e-01
-2.95798659e-01 -1.19480610e+00 8.97960365e-01 5.84881783e-01
2.06570532e-02 -7.16816545e-01 -8.23856652e-01 1.05211446e-02
-2.47363411e-02 -7.79477596e-01 -2.27105722e-01 3.54886621e-01
-8.72550726e-01 -1.32886231e+00 -1.03378296e+00 -6.92935944e-01
1.04960966e+00 -3.06560602e-02 1.04823256e+00 5.07330418e-01
-1.01538610e+00 3.58354121e-01 -1.37465239e-01 -7.47462273e-01
-6.26610696e-01 1.14499077e-01 -3.18332732e-01 -3.40478003e-01
6.12200975e-01 1.36452382e-02 -8.99799287e-01 2.81523734e-01
-7.37555325e-01 1.41318247e-01 9.69340086e-01 9.64589298e-01
1.03347504e+00 -1.65559188e-01 1.16093919e-01 -1.29245687e+00
-2.85159964e-02 -4.05773133e-01 -3.13511550e-01 -2.35191658e-02
-5.85294545e-01 -4.90166128e-01 6.17336512e-01 -1.89173400e-01
-6.58030093e-01 1.11668734e-02 -2.09162518e-01 -2.43584923e-02
-4.46201622e-01 4.83497560e-01 5.37957728e-01 -1.21000782e-01
7.11368024e-01 -3.89398098e-01 4.25213307e-01 1.67131096e-01
-2.28689983e-01 7.40946233e-01 2.82697499e-01 -5.87641820e-03
6.25882804e-01 1.02173185e+00 4.35157835e-01 -6.81056201e-01
-5.60491502e-01 -1.02847421e+00 -4.01288152e-01 -6.23856746e-02
9.74084735e-01 -8.02702010e-01 -8.91694784e-01 2.05364257e-01
-5.38604200e-01 -1.03493071e+00 -2.84750462e-01 4.86173749e-01
-4.59726393e-01 2.24132240e-02 -1.16517115e+00 -5.40192664e-01
-8.42905581e-01 -7.33465970e-01 1.19358563e+00 2.01287955e-01
-6.47285700e-01 -1.38298261e+00 5.28096795e-01 4.69927579e-01
6.12588882e-01 5.39392173e-01 9.34755623e-01 -7.60994613e-01
-1.35032997e-01 -6.38180017e-01 -3.93152744e-01 3.27145085e-02
1.02591380e-01 4.72781509e-01 -1.06230307e+00 -2.91194767e-01
-3.50597054e-01 -1.98980913e-01 6.40206635e-01 5.31711221e-01
1.01451635e+00 -3.92063588e-01 -8.64447773e-01 4.42093074e-01
1.72070789e+00 -1.73296958e-01 5.67929327e-01 4.82649654e-01
5.36897063e-01 7.20719755e-01 6.62741363e-01 1.62937984e-01
1.78118408e-01 1.24896958e-01 2.85551786e-01 -8.79163980e-01
-2.14040533e-01 2.11919129e-01 1.23050213e-02 1.52701139e-01
-5.07115349e-02 -2.55946815e-01 -1.42943752e+00 8.78148973e-01
-1.15660083e+00 -6.26126528e-01 -4.80224699e-01 1.84074271e+00
9.80082452e-01 8.69599953e-02 2.67902136e-01 1.96552113e-01
5.85483968e-01 -5.75185359e-01 -1.60281017e-01 -3.54998291e-01
4.51327413e-02 5.84758282e-01 7.45616019e-01 4.26241130e-01
-7.61388779e-01 5.49936533e-01 5.74482059e+00 9.50736940e-01
-1.45174599e+00 3.72985512e-01 1.24513733e+00 -3.05012494e-01
6.29671738e-02 -1.40733600e-01 -6.00722790e-01 3.57165262e-02
1.06411493e+00 -1.60282981e-02 -3.28215331e-01 4.86033231e-01
3.98375452e-01 -5.75479865e-01 -1.03697610e+00 5.60411215e-01
-2.44757235e-01 -1.57579195e+00 -3.26533586e-01 5.93786478e-01
5.19404113e-01 2.35216156e-01 8.01489055e-02 1.88894928e-01
2.13761836e-01 -1.33582640e+00 2.64459886e-02 3.30616564e-01
1.36896706e+00 -6.00069046e-01 1.69570506e+00 2.60561466e-01
-8.23163033e-01 6.16640925e-01 -1.33065164e-01 2.61897713e-01
-3.10379654e-01 5.53337514e-01 -1.97546005e+00 -1.74938459e-02
5.57093501e-01 2.47504845e-01 -8.38770986e-01 6.72776759e-01
2.32494131e-01 8.69969547e-01 -4.23447400e-01 -3.29778463e-01
3.20385426e-01 3.33908856e-01 -1.12145633e-01 1.73061490e+00
2.26841956e-01 1.43329844e-01 -3.01215291e-01 3.05265039e-01
2.99869746e-01 4.14265484e-01 -2.17819855e-01 -2.03715041e-01
7.60560483e-02 2.08547115e+00 -1.43649113e+00 -3.91924113e-01
-2.02623561e-01 4.33110595e-01 1.57534312e-02 -3.47425193e-02
-7.47935057e-01 -1.23544395e-01 5.69036119e-02 9.24071372e-01
2.23656781e-02 5.45588851e-01 -6.32019699e-01 -5.07446766e-01
-7.75918901e-01 -6.19323015e-01 8.03457499e-01 -4.85825211e-01
-1.29469121e+00 2.63183266e-01 -4.23371911e-01 -1.13273370e+00
-9.45442636e-03 -9.87585187e-01 -7.85303354e-01 8.11643481e-01
-1.55576205e+00 -1.39814532e+00 -4.99115497e-01 2.74759293e-01
4.24712226e-02 2.79032588e-01 1.14801133e+00 -7.84900486e-02
-4.96869475e-01 7.67577946e-01 -1.76982448e-01 1.90819010e-01
9.01252091e-01 -1.56235147e+00 -4.33421075e-01 2.19428703e-01
-6.26825750e-01 4.21244353e-01 5.58969021e-01 -3.89058650e-01
-1.16505527e+00 -1.29384625e+00 7.70286679e-01 -5.86118937e-01
8.14486027e-01 -1.51532209e-02 -5.49689412e-01 6.43865943e-01
4.09070045e-01 4.25880104e-01 1.59962332e+00 -1.59974173e-01
-1.17806057e-02 1.38943031e-01 -1.51977885e+00 5.96850574e-01
2.73111705e-02 -3.45565170e-01 9.90095511e-02 6.77453578e-01
-2.36532286e-01 -4.26932752e-01 -1.30971134e+00 2.20121741e-01
6.30428016e-01 -9.56100762e-01 6.63957417e-01 -2.43180439e-01
1.54107004e-01 -4.28463787e-01 1.52159870e-01 -8.04951012e-01
-1.09757133e-01 -2.04331949e-01 6.68670535e-01 8.36120307e-01
5.91290593e-01 -7.99472630e-01 1.36275125e+00 4.86555964e-01
-1.07637167e-01 -9.77465987e-01 -9.13417220e-01 -2.65881270e-01
5.69101512e-01 -6.31304085e-02 5.27548976e-02 9.53796744e-01
3.98262352e-01 -1.39707858e-02 7.74263144e-01 -1.39904335e-01
5.28703809e-01 -4.13998544e-01 8.48978817e-01 -9.50708330e-01
-1.90186501e-01 -8.13916624e-01 -5.10199845e-01 1.34722248e-01
-8.19455553e-03 -1.01086688e+00 -2.70216852e-01 -1.70473909e+00
8.13532829e-01 -5.60134232e-01 -3.54898781e-01 8.50183725e-01
-1.32968500e-01 7.50072181e-01 -2.65564501e-01 2.56176621e-01
-3.40781480e-01 -3.56999278e-01 1.12771726e+00 -3.16174388e-01
1.72025681e-01 -3.50966543e-01 -5.61141074e-01 7.73640156e-01
1.05349910e+00 -6.83824241e-01 9.42547768e-02 5.06230593e-01
1.45323515e-01 -7.65095502e-02 7.10532486e-01 -8.66688251e-01
1.76635355e-01 -2.39119619e-01 5.94701111e-01 -8.52370381e-01
-1.65941074e-01 -7.32011378e-01 3.06450635e-01 1.01500237e+00
-1.24104366e-01 -2.12782532e-01 4.57309395e-01 1.16997905e-01
6.76202849e-02 -1.27270713e-01 9.85174835e-01 -3.30948561e-01
1.49027973e-01 1.76341012e-01 -8.63195300e-01 -3.20120692e-01
1.19772208e+00 -6.20796502e-01 -6.08468354e-01 1.06700741e-01
-6.25139296e-01 1.89164400e-01 8.74973893e-01 -5.26779890e-01
-5.20335548e-02 -8.78786922e-01 -8.01842034e-01 -3.26528311e-01
2.67380625e-01 9.95022878e-02 4.20807332e-01 1.55727577e+00
-1.18811107e+00 6.71427131e-01 -6.63248077e-02 -8.25763106e-01
-1.56079042e+00 2.71491259e-01 5.95713675e-01 -9.50065911e-01
-2.79810220e-01 9.94098961e-01 3.19609761e-01 -3.68143439e-01
-1.32397026e-01 -3.17990899e-01 8.00100267e-02 1.98160499e-01
4.08861816e-01 4.52733487e-01 3.35317522e-01 -4.67852473e-01
-4.85702097e-01 3.57796043e-01 -4.18228418e-01 -1.46096841e-01
1.01547265e+00 2.16334730e-01 -4.50176209e-01 5.24925172e-01
1.28532231e+00 1.07074462e-01 -6.75538361e-01 3.12402248e-01
-7.70650506e-02 -2.69001536e-02 2.51114368e-01 -9.37985003e-01
-9.11788583e-01 7.62981534e-01 9.60034013e-01 -1.25857368e-01
1.04397810e+00 5.40320799e-02 5.50923526e-01 -2.84957029e-02
-5.82024977e-02 -1.02554393e+00 -1.04483746e-01 4.91730869e-02
3.29762995e-01 -1.22327113e+00 1.29906550e-01 -4.40734595e-01
-5.16384304e-01 1.20038998e+00 4.09299463e-01 -3.54848728e-02
4.64644462e-01 9.16618645e-01 5.98070264e-01 -3.36303473e-01
-9.90180910e-01 -2.26480946e-01 -1.68714121e-01 6.07988298e-01
9.42916870e-01 3.77272576e-01 -3.56700510e-01 4.48568195e-01
-3.93783659e-01 1.71233907e-01 5.99717259e-01 9.70531464e-01
-3.57153982e-01 -9.68031645e-01 -5.19584000e-01 6.86217248e-01
-9.77289557e-01 1.96700394e-02 -6.84907973e-01 1.27630246e+00
4.36377712e-02 6.47251308e-01 2.07981497e-01 2.12979957e-01
2.32713353e-02 3.27312686e-02 3.57384771e-01 -7.76262283e-01
-1.13584805e+00 5.65049469e-01 4.53664064e-02 -4.05044705e-02
-4.96447533e-01 -7.41663218e-01 -1.61735559e+00 -4.24744666e-01
-4.82451439e-01 6.19855523e-02 6.19149566e-01 5.70431948e-01
-3.45960073e-02 6.81797981e-01 3.07980090e-01 -3.54213148e-01
1.14805207e-01 -9.41042483e-01 -7.82207370e-01 -3.49591263e-02
4.24038738e-01 -1.94545954e-01 -8.26964080e-01 3.22728544e-01]
|
[15.092957496643066, -3.1036362648010254]
|
7cf922f6-5358-4d12-a5d0-834b418cdb34
|
better-computer-go-player-with-neural-network
|
1511.06410
| null |
http://arxiv.org/abs/1511.06410v3
|
http://arxiv.org/pdf/1511.06410v3.pdf
|
Better Computer Go Player with Neural Network and Long-term Prediction
|
Competing with top human players in the ancient game of Go has been a
long-term goal of artificial intelligence. Go's high branching factor makes
traditional search techniques ineffective, even on leading-edge hardware, and
Go's evaluation function could change drastically with one stone change. Recent
works [Maddison et al. (2015); Clark & Storkey (2015)] show that search is not
strictly necessary for machine Go players. A pure pattern-matching approach,
based on a Deep Convolutional Neural Network (DCNN) that predicts the next
move, can perform as well as Monte Carlo Tree Search (MCTS)-based open source
Go engines such as Pachi [Baudis & Gailly (2012)] if its search budget is
limited. We extend this idea in our bot named darkforest, which relies on a
DCNN designed for long-term predictions. Darkforest substantially improves the
win rate for pattern-matching approaches against MCTS-based approaches, even
with looser search budgets. Against human players, the newest versions,
darkfores2, achieve a stable 3d level on KGS Go Server as a ranked bot, a
substantial improvement upon the estimated 4k-5k ranks for DCNN reported in
Clark & Storkey (2015) based on games against other machine players. Adding
MCTS to darkfores2 creates a much stronger player named darkfmcts3: with 5000
rollouts, it beats Pachi with 10k rollouts in all 250 games; with 75k rollouts
it achieves a stable 5d level in KGS server, on par with state-of-the-art Go
AIs (e.g., Zen, DolBaram, CrazyStone) except for AlphaGo [Silver et al.
(2016)]; with 110k rollouts, it won the 3rd place in January KGS Go Tournament.
|
['Yuandong Tian', 'Yan Zhu']
|
2015-11-19
| null | null | null | null |
['game-of-go']
|
['playing-games']
|
[-4.59856689e-01 1.94986556e-02 -3.38070810e-01 2.94679523e-01
-5.94615936e-01 -7.97536910e-01 5.35356641e-01 -3.78491610e-01
-7.34837830e-01 5.59585273e-01 -2.24314809e-01 -8.27582240e-01
-3.82377356e-01 -1.13529313e+00 -8.39361608e-01 -3.22692364e-01
-3.04244906e-01 9.82179463e-01 1.04188573e+00 -8.84647250e-01
3.05405945e-01 2.24036440e-01 -1.62226129e+00 3.15939546e-01
3.84901583e-01 1.21926141e+00 8.33239853e-02 1.13940680e+00
3.26467127e-01 1.33452976e+00 -4.89683390e-01 -7.14950860e-01
9.59688723e-01 -1.41758561e-01 -1.15548623e+00 -1.10081637e+00
3.27220947e-01 -3.49734664e-01 -6.68556809e-01 7.84599423e-01
6.38506293e-01 8.26205090e-02 2.26192459e-01 -1.29816532e+00
2.09449362e-02 1.10555255e+00 -6.09123111e-01 4.34009969e-01
-1.16234533e-01 8.83953393e-01 1.42951775e+00 -3.79357159e-01
5.75731874e-01 8.35982025e-01 1.09452283e+00 5.33561587e-01
-8.35194290e-01 -8.41804445e-01 -3.63521636e-01 5.44646800e-01
-1.17382920e+00 -3.32107954e-02 1.47093892e-01 -1.44018814e-01
1.29821193e+00 4.19572234e-01 1.03678143e+00 9.81880784e-01
3.40374857e-01 9.24611866e-01 9.20910954e-01 -2.28296325e-01
4.45158333e-01 -1.00453460e+00 -1.73114002e-01 9.66183901e-01
3.46975744e-01 5.66303730e-01 -7.69072592e-01 -3.16962928e-01
8.49053621e-01 -3.43763292e-01 3.20954323e-01 -1.49387375e-01
-1.00487351e+00 9.43427980e-01 7.17307150e-01 2.29745626e-01
-3.80340189e-01 1.01353407e+00 5.25037348e-01 2.67918438e-01
2.91721094e-02 8.93058360e-01 -6.05459154e-01 -1.00799799e+00
-1.16654038e+00 9.09787595e-01 1.04971886e+00 6.72704518e-01
5.75054228e-01 2.37458590e-02 -2.07463816e-01 9.00730044e-02
-1.60960019e-01 1.02592826e-01 6.75103605e-01 -1.26593757e+00
3.54795307e-01 4.03297514e-01 -1.50841586e-02 -8.70663822e-01
-6.90944135e-01 -6.98068380e-01 -4.99580711e-01 7.99934626e-01
8.01536679e-01 -3.12145025e-01 -8.01843584e-01 1.51635289e+00
1.86467513e-01 1.01029076e-01 -2.80856401e-01 9.11315918e-01
6.33059740e-01 3.27157587e-01 -1.11238092e-01 7.02074766e-01
1.26378119e+00 -1.43220794e+00 3.84459347e-01 -6.00167751e-01
7.95166790e-01 -3.24452847e-01 1.05086720e+00 7.84891427e-01
-1.15292740e+00 -1.99635386e-01 -1.05894721e+00 1.14790685e-01
-1.21470667e-01 -3.17522258e-01 1.38806105e+00 8.58686924e-01
-1.30778730e+00 1.00938666e+00 -1.06427717e+00 -1.65836304e-01
4.34605718e-01 6.61621153e-01 -3.56352851e-02 2.33013377e-01
-1.33197510e+00 8.80298555e-01 4.97744530e-01 -2.37972304e-01
-1.31792521e+00 -5.63241661e-01 -1.79640576e-01 2.23451719e-01
9.91794109e-01 -6.92013741e-01 1.75658333e+00 -4.93496001e-01
-1.56186283e+00 9.59211171e-01 5.88606715e-01 -1.07412517e+00
1.01434255e+00 -8.28208923e-02 -9.42368060e-03 -8.11223388e-02
1.74266055e-01 6.65697038e-01 4.07196432e-01 -4.94493514e-01
-1.02390432e+00 -1.94700971e-01 4.97426182e-01 2.20664322e-01
2.09633797e-01 9.67011750e-02 -4.74653244e-01 -1.45622984e-01
-1.51383713e-01 -9.73525465e-01 -5.68222582e-01 -3.64564955e-01
-3.66124898e-01 -4.24645305e-01 1.48290768e-01 -3.75949115e-01
1.16986001e+00 -1.59159517e+00 7.55913928e-02 1.54365674e-01
6.36074901e-01 3.25059146e-01 -3.52270454e-02 4.19644058e-01
2.19156146e-01 2.37231791e-01 3.10841709e-01 2.03279525e-01
3.36591095e-01 1.21855982e-01 6.32405281e-02 1.55078471e-01
-2.40288466e-01 1.43567300e+00 -9.96261954e-01 -4.80949581e-02
-8.15488473e-02 -4.09805328e-01 -1.13929939e+00 -4.80305791e-01
-3.85672718e-01 -4.13304083e-02 -3.35912883e-01 6.79790556e-01
9.96687338e-02 -3.97878140e-01 -4.09408994e-02 4.47517693e-01
-2.93173850e-01 5.33622384e-01 -9.67505932e-01 1.57136142e+00
-1.36137679e-01 4.73619163e-01 6.38260394e-02 -7.57591069e-01
6.04929864e-01 -1.87663227e-01 3.04591835e-01 -7.75392056e-01
3.98262322e-01 5.03993690e-01 4.36353743e-01 1.61542930e-02
8.95277798e-01 1.94851145e-01 -4.47018892e-01 2.54897773e-01
7.14562014e-02 -4.02069002e-01 5.18067062e-01 3.40376794e-01
2.05433226e+00 1.88271746e-01 -1.16474535e-02 -2.72377163e-01
-1.12011477e-01 6.43461645e-01 4.02660757e-01 1.51213229e+00
-3.72112691e-01 3.17044228e-01 7.97330081e-01 -9.23603356e-01
-1.03513777e+00 -7.03187943e-01 3.78164023e-01 1.69421399e+00
1.55513838e-01 -8.50832105e-01 -8.86509955e-01 -6.95359051e-01
6.29541604e-03 3.31862450e-01 -5.95236242e-01 -2.25386307e-01
-6.53485477e-01 -7.19137967e-01 1.39568603e+00 3.49287987e-01
9.35619593e-01 -1.31181300e+00 -1.13038933e+00 3.99659544e-01
-7.83901885e-02 -6.18857086e-01 -7.07624108e-02 6.43084347e-01
-3.90150607e-01 -1.27146542e+00 -4.01266962e-01 -2.93423027e-01
-3.97618413e-01 -6.48458153e-02 1.35044467e+00 2.45420367e-01
-4.61840957e-01 -3.01574677e-01 -4.10147250e-01 -2.57159233e-01
-2.26067454e-01 7.07967937e-01 1.72742814e-01 -9.19662654e-01
2.00233400e-01 -6.99646294e-01 -7.17704356e-01 4.00498509e-01
-4.62095559e-01 1.36341602e-01 6.13323152e-01 1.05355060e+00
1.38841858e-02 1.78809389e-01 -4.63526137e-02 -5.70105672e-01
5.00234663e-01 -4.77007776e-01 -9.26407456e-01 -1.21972293e-01
-5.65814257e-01 -2.54721995e-02 6.66707218e-01 -2.81515658e-01
-2.30217636e-01 -2.99628466e-01 -2.69153118e-01 -4.20166582e-01
3.22228521e-01 4.81267929e-01 4.74591345e-01 -5.51127374e-01
1.31001937e+00 8.91670212e-02 -4.06748384e-01 -4.15929317e-01
2.81148702e-01 2.71999657e-01 8.19325686e-01 -8.52399826e-01
7.69087911e-01 1.74740419e-01 2.11574212e-01 1.24091864e-01
-6.29617751e-01 -1.66929394e-01 -1.23867124e-01 -2.26025760e-01
6.05190694e-01 -7.67897010e-01 -1.70612586e+00 8.38919699e-01
-8.40195596e-01 -1.05566490e+00 -4.23545897e-01 6.16071932e-02
-6.82027161e-01 5.55903018e-02 -9.81311560e-01 -4.93477672e-01
-3.14282954e-01 -1.10077298e+00 8.01713467e-01 2.93131679e-01
-2.51084208e-01 -4.77055401e-01 2.08371773e-01 6.10367954e-01
6.46668732e-01 7.37120435e-02 6.58264816e-01 -7.70439506e-01
-6.22245193e-01 -2.78078407e-01 -1.86879039e-01 -1.63541123e-01
-5.18796206e-01 -1.79590598e-01 -5.22484422e-01 -1.08429901e-01
-5.72418511e-01 -6.45298898e-01 9.32433784e-01 2.92262435e-01
9.38411891e-01 -2.08432212e-01 -3.15560490e-01 6.41356289e-01
1.24536955e+00 1.26998603e-01 6.91131353e-01 1.08166921e+00
4.96565700e-01 6.51409552e-02 4.24929291e-01 3.93175930e-01
2.57665664e-01 8.33229542e-01 1.12810266e+00 3.13182205e-01
-9.49819833e-02 -4.89520639e-01 2.47741863e-01 2.81272501e-01
-7.11835980e-01 -4.26119268e-01 -1.36534882e+00 3.58963042e-01
-2.03879976e+00 -1.03355241e+00 -2.76007593e-01 1.97014737e+00
7.33395040e-01 7.80903161e-01 5.11361361e-01 5.70107326e-02
5.43082476e-01 4.00776304e-02 -6.62391841e-01 -5.10621130e-01
4.01016884e-02 8.47034395e-01 1.18665540e+00 3.05444419e-01
-9.16572332e-01 1.67659163e+00 6.65340757e+00 1.76187015e+00
-9.70004320e-01 1.90715566e-01 5.54988086e-01 -5.62667787e-01
2.26976916e-01 2.68169343e-01 -8.69593680e-01 6.29947424e-01
9.84887660e-01 -3.25732112e-01 9.79017913e-01 1.27728891e+00
-3.22346151e-01 -3.59689951e-01 -7.31685817e-01 9.42677319e-01
-3.31532717e-01 -1.81836867e+00 -5.39744020e-01 5.44596970e-01
7.12298274e-01 8.54619443e-01 -9.88667607e-02 8.43346179e-01
1.68287945e+00 -1.36607504e+00 1.10089064e+00 2.45743319e-02
6.62019253e-01 -8.62157881e-01 6.96357489e-01 6.58582151e-01
-8.79011631e-01 -1.67679161e-01 -3.87130141e-01 -8.07267666e-01
-5.44736832e-02 1.63348496e-01 -6.30734324e-01 4.93250698e-01
1.12447989e+00 5.89731187e-02 -5.74335635e-01 1.19339907e+00
-3.94072652e-01 8.96716714e-01 -7.53412902e-01 -3.84428382e-01
9.61989760e-01 3.26381266e-01 5.97608805e-01 5.56840420e-01
1.02624722e-01 5.89330792e-02 2.95290083e-01 6.30239308e-01
-3.36776078e-01 -3.74231428e-01 -2.08549947e-01 -9.34422109e-03
4.37524319e-01 1.21215236e+00 -8.16416800e-01 -1.91360667e-01
2.90911555e-01 8.66044104e-01 4.13815767e-01 -3.67005110e-01
-9.60930228e-01 -3.66902828e-01 5.48161864e-01 1.86449960e-01
4.47212487e-01 -1.78997308e-01 -3.12896520e-01 -8.32493246e-01
-2.67237097e-01 -1.12249839e+00 4.73608613e-01 -7.16930091e-01
-9.08123970e-01 5.90150058e-01 -2.50038058e-01 -8.65509570e-01
-5.67065597e-01 -8.58619571e-01 -8.49374056e-01 6.30902588e-01
-9.16015327e-01 -1.13341320e+00 -1.19586261e-02 4.46823955e-01
3.20890516e-01 -4.19747174e-01 5.28704643e-01 1.46718025e-01
-2.97455847e-01 8.05267751e-01 2.08993673e-01 5.69211170e-02
1.10494435e-01 -1.07563245e+00 1.07500279e+00 6.54842317e-01
2.24385381e-01 1.74065322e-01 7.20057607e-01 -7.13663042e-01
-1.42736721e+00 -5.89210272e-01 3.99409503e-01 -4.31440711e-01
1.09943318e+00 -1.97584480e-01 -1.98949769e-01 6.82780981e-01
-1.68598011e-01 6.00239225e-02 -1.54722884e-01 3.36701155e-01
-3.95124733e-01 2.12800100e-01 -1.01385534e+00 9.03348148e-01
1.47886682e+00 -1.37202129e-01 -1.25500798e-01 2.42901906e-01
5.99430859e-01 -9.05640125e-01 -2.58871555e-01 3.24024916e-01
8.12104225e-01 -1.36615860e+00 7.50429690e-01 -9.01009977e-01
5.67306459e-01 -8.35731775e-02 -6.88041514e-03 -1.18385625e+00
-5.99109352e-01 -1.04789972e+00 6.50513545e-02 3.66447657e-01
2.60823667e-01 -6.16970420e-01 1.53796721e+00 2.56891221e-01
-2.79576778e-01 -9.28564250e-01 -1.24579942e+00 -1.25245249e+00
3.70249778e-01 -6.56849802e-01 5.91869354e-01 6.34645998e-01
1.92293689e-01 2.98625547e-02 -6.30237877e-01 -1.99887410e-01
3.00623298e-01 -2.38832712e-01 1.04181623e+00 -1.15381801e+00
-1.15992475e+00 -9.47173059e-01 -8.31844926e-01 -1.02637911e+00
-2.62879044e-01 -9.71674144e-01 -1.56400464e-02 -1.21714032e+00
2.15784356e-01 -6.32979333e-01 -4.07612287e-02 1.01909280e+00
1.54693663e-01 6.71826899e-01 4.42490011e-01 3.70357424e-01
-8.23172629e-01 1.73005372e-01 9.82722580e-01 -9.57882777e-03
1.63052991e-01 -6.27474338e-02 -7.53594160e-01 9.48793769e-01
7.40325868e-01 -5.13444602e-01 2.20045120e-01 -3.46373558e-01
8.66462708e-01 -4.15171348e-02 5.61762929e-01 -1.58933878e+00
6.87509954e-01 -2.73483783e-01 -6.23467527e-02 -3.38126481e-01
2.76318282e-01 -2.61842579e-01 2.19031051e-01 1.13352191e+00
4.18945327e-02 1.32599726e-01 3.15900743e-01 1.90303087e-01
2.05389768e-01 -4.19255465e-01 4.45993751e-01 -4.87115353e-01
-8.54320586e-01 3.52809668e-01 -6.71403587e-01 2.65340865e-01
8.97341728e-01 -4.86347139e-01 -6.68874443e-01 -5.24562776e-01
-5.26188076e-01 3.71879309e-01 4.00474936e-01 1.79521859e-01
-1.49770319e-01 -9.33457255e-01 -4.85704452e-01 -3.16856235e-01
-2.39070922e-01 1.05422795e-01 1.97660387e-01 6.47927523e-01
-1.22575128e+00 2.55409062e-01 -4.22844321e-01 -1.16519384e-01
-1.08414030e+00 1.14157096e-01 4.76272851e-01 -1.10375285e+00
-5.27599096e-01 1.37862682e+00 -1.73421770e-01 -5.89983404e-01
-2.38720160e-02 -7.35510141e-02 3.90367776e-01 -3.86741817e-01
3.61556739e-01 5.69150388e-01 3.42071414e-01 -8.65884796e-02
-4.31990206e-01 2.51868237e-02 -4.81212623e-02 -3.95187855e-01
1.38646257e+00 6.67772055e-01 1.47915393e-01 -4.97990064e-02
3.54589999e-01 6.10877387e-02 -1.12328649e+00 7.71727413e-02
1.05013035e-01 -3.89000565e-01 8.64718333e-02 -1.14745510e+00
-9.55358207e-01 7.19594717e-01 2.71047652e-01 2.98254848e-01
5.86413085e-01 6.56718984e-02 9.50068533e-01 6.75988793e-01
1.25163901e+00 -9.06437814e-01 1.50201261e-01 1.13188171e+00
3.55211258e-01 -7.96292543e-01 -2.44099900e-01 2.20281910e-02
-4.75376070e-01 9.96636450e-01 8.41419637e-01 -4.97494876e-01
1.81095898e-02 2.28875533e-01 -4.86396343e-01 -5.81917644e-01
-1.03307235e+00 -4.10404176e-01 -3.02582234e-01 3.08793843e-01
-2.32256725e-01 2.49860108e-01 -2.30124265e-01 8.78457725e-01
-1.05674958e+00 1.81636423e-01 3.37153167e-01 1.00969982e+00
-6.55695796e-01 -9.72249985e-01 -2.22857565e-01 4.85901326e-01
-4.86019164e-01 -5.24703681e-01 -5.34860611e-01 9.96283531e-01
4.06511098e-01 7.26161122e-01 -1.98495165e-02 -8.67033482e-01
1.79332048e-01 -1.26903370e-01 4.79163438e-01 -5.34187198e-01
-1.27419889e+00 -4.80367512e-01 4.65723187e-01 -9.62335765e-01
4.52652961e-01 -2.76388437e-01 -1.11995888e+00 -1.28697836e+00
-4.30220932e-01 2.39508182e-01 4.79226470e-01 8.39581490e-01
3.35733771e-01 4.92066532e-01 1.44519538e-01 -9.83189940e-01
-6.77217543e-01 -7.98509061e-01 -6.84509695e-01 -2.61910617e-01
-4.61892247e-01 -5.83571017e-01 -3.25761318e-01 -8.74809027e-01]
|
[3.4663331508636475, 1.4182074069976807]
|
efa92187-675d-4540-9b8b-71adb6dda85d
|
octet-object-aware-counterfactual
|
2211.12380
| null |
https://arxiv.org/abs/2211.12380v2
|
https://arxiv.org/pdf/2211.12380v2.pdf
|
OCTET: Object-aware Counterfactual Explanations
|
Nowadays, deep vision models are being widely deployed in safety-critical applications, e.g., autonomous driving, and explainability of such models is becoming a pressing concern. Among explanation methods, counterfactual explanations aim to find minimal and interpretable changes to the input image that would also change the output of the model to be explained. Such explanations point end-users at the main factors that impact the decision of the model. However, previous methods struggle to explain decision models trained on images with many objects, e.g., urban scenes, which are more difficult to work with but also arguably more critical to explain. In this work, we propose to tackle this issue with an object-centric framework for counterfactual explanation generation. Our method, inspired by recent generative modeling works, encodes the query image into a latent space that is structured in a way to ease object-level manipulations. Doing so, it provides the end-user with control over which search directions (e.g., spatial displacement of objects, style modification, etc.) are to be explored during the counterfactual generation. We conduct a set of experiments on counterfactual explanation benchmarks for driving scenes, and we show that our method can be adapted beyond classification, e.g., to explain semantic segmentation models. To complete our analysis, we design and run a user study that measures the usefulness of counterfactual explanations in understanding a decision model. Code is available at https://github.com/valeoai/OCTET.
|
['Matthieu Cord', 'Patrick Pérez', 'Hédi Ben-Younes', 'Éloi Zablocki', 'Mickaël Chen', 'Mehdi Zemni']
|
2022-11-22
| null |
http://openaccess.thecvf.com//content/CVPR2023/html/Zemni_OCTET_Object-Aware_Counterfactual_Explanations_CVPR_2023_paper.html
|
http://openaccess.thecvf.com//content/CVPR2023/papers/Zemni_OCTET_Object-Aware_Counterfactual_Explanations_CVPR_2023_paper.pdf
|
cvpr-2023-1
|
['counterfactual-explanation', 'explanation-generation']
|
['miscellaneous', 'natural-language-processing']
|
[ 3.97968113e-01 4.61602271e-01 -1.22650199e-01 -5.25123358e-01
-1.81255102e-01 -6.32891238e-01 8.39333057e-01 -2.84562707e-02
-1.18373185e-01 5.64913750e-01 1.57356650e-01 -7.85642505e-01
-5.20592667e-02 -7.86851108e-01 -1.08307803e+00 -3.92156303e-01
3.74987751e-01 5.39240718e-01 6.08769618e-03 -1.13774225e-01
5.59389234e-01 3.30068380e-01 -1.90379727e+00 2.87873387e-01
1.06784832e+00 5.32509446e-01 4.82072324e-01 5.27289510e-01
-6.59073740e-02 3.75264376e-01 -4.74498212e-01 -4.73556280e-01
9.06531140e-02 -4.83893335e-01 -9.02285039e-01 4.06928897e-01
4.05737042e-01 -2.11255625e-01 -1.85406238e-01 1.06438148e+00
8.73525590e-02 3.00955594e-01 7.31356859e-01 -1.59210968e+00
-7.09786236e-01 7.19115973e-01 -2.81550795e-01 2.16232538e-01
3.71807138e-03 6.90065682e-01 8.17895651e-01 -7.88929701e-01
6.39633060e-01 1.45502269e+00 9.09173395e-03 7.78385520e-01
-1.36758614e+00 -7.09953725e-01 5.06487012e-01 6.20478928e-01
-8.41056287e-01 -3.96353304e-01 9.92707193e-01 -4.44316775e-01
7.16787517e-01 5.60593605e-01 6.85108125e-01 1.20545971e+00
3.67912263e-01 8.90421927e-01 1.09427655e+00 -2.20794916e-01
5.30436456e-01 2.13121235e-01 -3.48517671e-02 4.46263552e-01
5.50445557e-01 2.67668426e-01 -3.57710749e-01 1.63491711e-01
5.97674191e-01 1.16297871e-01 -3.23745161e-01 -7.82087743e-01
-1.06303585e+00 1.11340678e+00 5.32891631e-01 2.39716381e-01
-4.75356668e-01 4.30650204e-01 3.36381122e-02 -8.63355100e-02
4.11527693e-01 8.35277438e-01 -3.76813203e-01 -1.38705611e-01
-6.85876012e-01 6.55255079e-01 5.29291034e-01 8.00488591e-01
7.67744362e-01 -1.31260946e-01 -3.85373592e-01 5.56219101e-01
3.19627047e-01 3.03219438e-01 5.29217124e-01 -1.11509335e+00
4.19820875e-01 5.11568606e-01 1.57922477e-01 -8.70056748e-01
-3.37705702e-01 -4.47266459e-01 -5.51739275e-01 3.83790016e-01
3.52709144e-01 -6.61836267e-02 -1.13891089e+00 1.76371670e+00
2.36344874e-01 2.42113575e-01 9.78977140e-03 1.29726517e+00
4.97976005e-01 4.54960734e-01 2.07893759e-01 -2.69724969e-02
1.30669475e+00 -1.15256774e+00 -6.64473593e-01 -7.85735905e-01
6.54224217e-01 -7.04700589e-01 1.25560510e+00 1.85108200e-01
-9.75053012e-01 -7.39238977e-01 -9.23232615e-01 7.52004460e-02
-4.39526349e-01 -2.55133182e-01 8.59321177e-01 5.50635695e-01
-7.29690433e-01 5.61350882e-01 -6.99649036e-01 -3.13192457e-01
5.85430026e-01 2.10629836e-01 -6.39497116e-02 -2.68490892e-02
-1.12169027e+00 8.84987533e-01 5.26940346e-01 5.75159267e-02
-1.03283656e+00 -5.94084740e-01 -1.04499185e+00 2.40035668e-01
6.91336632e-01 -1.00877380e+00 1.46724212e+00 -1.12014544e+00
-1.15964925e+00 6.30139112e-01 -3.64082217e-01 -7.52038836e-01
6.49167657e-01 -2.15871692e-01 -2.17012246e-03 -2.72307038e-01
4.77321535e-01 1.15568805e+00 8.89632761e-01 -1.57481778e+00
-6.56994581e-01 -4.30733681e-01 5.20481169e-01 2.93950349e-01
2.25812092e-01 -4.49213833e-01 -4.41915989e-01 -5.33274174e-01
9.06910375e-02 -1.24096000e+00 -5.30073643e-01 -2.18962088e-01
-7.54020691e-01 -9.43469480e-02 9.58907545e-01 -2.44522393e-01
8.37461352e-01 -2.07765126e+00 5.74698523e-02 -6.48691803e-02
3.47849309e-01 6.04032949e-02 -8.80933478e-02 -5.51827997e-02
-2.78230190e-01 5.58337033e-01 -5.04150271e-01 -4.61265624e-01
2.24296153e-01 2.01782003e-01 -4.68026787e-01 1.35774970e-01
2.40114301e-01 1.09055805e+00 -7.92466402e-01 -2.05586046e-01
5.33999324e-01 1.95326284e-01 -7.71590114e-01 1.71980128e-01
-6.23768687e-01 6.26074910e-01 -5.11902452e-01 1.83184132e-01
6.45027399e-01 -1.92623049e-01 -2.91945529e-03 1.63504854e-01
-1.30163610e-01 3.15591723e-01 -1.00114834e+00 1.57046890e+00
-6.24116540e-01 8.82992089e-01 -5.72018266e-01 -8.74747455e-01
5.78465104e-01 2.37303507e-03 -3.57743725e-03 -6.41047001e-01
1.19187668e-01 3.94279622e-02 3.28176200e-01 -4.94709432e-01
6.28211558e-01 -1.08728394e-01 -7.56101310e-02 4.30608153e-01
-4.85394716e-01 -3.78502160e-01 1.39510751e-01 1.33613855e-01
6.56308472e-01 1.20785214e-01 4.03603792e-01 -5.98293096e-02
1.86304912e-01 4.13137138e-01 4.42069292e-01 9.70373988e-01
-1.27318829e-01 7.20436513e-01 5.26209474e-01 -6.24700367e-01
-8.47189426e-01 -8.69018078e-01 7.87952170e-02 6.88086927e-01
5.91585517e-01 -5.32463677e-02 -9.58660662e-01 -8.93456697e-01
-3.31531316e-02 1.61425757e+00 -6.94083691e-01 -5.24553418e-01
-3.93630952e-01 -5.31212091e-01 2.87503917e-02 3.84535342e-01
5.41362166e-01 -1.37276816e+00 -1.17150056e+00 9.61261243e-02
-3.30097646e-01 -8.63920927e-01 -4.70874250e-01 -1.01320513e-01
-9.09588993e-01 -1.04729033e+00 -3.36449325e-01 -2.15316534e-01
6.96538448e-01 5.55263698e-01 1.16874027e+00 1.61504164e-01
-1.59214273e-01 2.83582598e-01 -1.57478571e-01 -8.02398026e-01
-3.73170882e-01 -5.43047488e-02 -1.67737380e-01 5.71109168e-02
3.45254362e-01 -3.26087803e-01 -8.76727819e-01 4.28456068e-01
-1.08933926e+00 7.85009503e-01 6.57813251e-01 7.42515087e-01
6.91999137e-01 4.44858633e-02 3.44471246e-01 -1.12379992e+00
6.11874104e-01 -4.78123188e-01 -5.74435711e-01 -7.79346144e-03
-7.90199697e-01 3.48673671e-01 5.87196350e-01 -4.16842729e-01
-1.12720525e+00 3.41885872e-02 1.25355691e-01 -5.20463884e-01
-6.02931559e-01 3.76430482e-01 -3.76772553e-01 4.83128756e-01
5.43476820e-01 1.88960046e-01 -2.70183682e-01 -1.62330464e-01
6.85850501e-01 3.57946873e-01 3.04733604e-01 -3.21733534e-01
7.64510989e-01 5.50420642e-01 -1.69373319e-01 -3.80006582e-01
-6.52182162e-01 -1.03338256e-01 -3.97412032e-01 -1.74794286e-01
9.41288769e-01 -3.81683558e-01 -5.08203685e-01 -1.73017997e-02
-1.36109221e+00 -4.88707721e-01 -3.10141325e-01 4.10681784e-01
-8.28705132e-01 -9.47508365e-02 1.78095043e-01 -7.25605249e-01
2.12274283e-01 -1.49489665e+00 1.07421219e+00 4.08818901e-01
-5.75399220e-01 -8.25013399e-01 -2.13321701e-01 7.38235354e-01
2.90715724e-01 1.00703306e-01 9.90378499e-01 -5.91704428e-01
-9.86291826e-01 2.33134747e-01 -1.84465677e-01 -1.62548766e-01
7.94322714e-02 -2.02442199e-01 -1.11509228e+00 2.67700460e-02
-2.46130638e-02 2.27009267e-01 9.49316084e-01 6.59143627e-01
1.51770985e+00 -5.62348604e-01 -6.00924432e-01 3.75204593e-01
1.07367110e+00 4.17498738e-01 7.84888685e-01 4.55816507e-01
7.37851858e-01 8.04568887e-01 9.42588627e-01 1.75715491e-01
5.52129388e-01 9.63569641e-01 9.43812549e-01 -2.28878036e-01
-1.85574181e-02 -3.03178817e-01 1.46335885e-02 -1.76422194e-01
3.96147966e-02 -3.89333457e-01 -8.10697198e-01 7.56340384e-01
-2.11865234e+00 -9.65241134e-01 -2.07043067e-01 2.03544378e+00
4.16682810e-01 3.58759493e-01 -2.18251184e-01 1.47547070e-02
6.25766635e-01 1.97904423e-01 -9.38413084e-01 -6.63193166e-01
2.59621263e-01 -1.03234492e-01 3.88959140e-01 5.62984943e-01
-9.06817675e-01 1.13506508e+00 4.54985952e+00 6.40168786e-01
-1.16570854e+00 7.20131025e-02 9.27383542e-01 -1.48595348e-01
-7.82337904e-01 3.55850995e-01 -4.49594945e-01 5.21313787e-01
6.84288800e-01 -2.70751119e-01 4.02007222e-01 9.03921485e-01
6.92503810e-01 -1.63349032e-01 -1.32717884e+00 8.55534971e-01
-1.78506702e-01 -1.52589524e+00 3.37570399e-01 2.37985075e-01
6.99742615e-01 -3.23342770e-01 3.61873299e-01 2.43707523e-01
1.61852151e-01 -1.14217079e+00 1.12429345e+00 2.93905735e-01
3.77346426e-01 -7.57409036e-01 6.66783631e-01 4.92750168e-01
-7.01961637e-01 -7.74529278e-02 -2.43122399e-01 -2.66926408e-01
1.37105078e-01 5.13016164e-01 -1.02382243e+00 3.13456744e-01
5.53660512e-01 3.29872310e-01 -5.33006489e-01 7.79564679e-01
-6.83236361e-01 6.26811087e-01 1.41962290e-01 -1.16975617e-03
2.86312282e-01 2.72551775e-02 6.92779541e-01 8.45259309e-01
3.00964683e-01 1.77290272e-02 -1.10847764e-01 1.47360122e+00
7.38662258e-02 -1.50439665e-01 -9.61998284e-01 7.82731846e-02
2.85000771e-01 9.31151390e-01 -8.85792196e-01 -4.07800913e-01
-7.28662014e-02 1.06412733e+00 1.08714938e-01 5.30112445e-01
-1.19160748e+00 -4.06806059e-02 9.57004845e-01 2.77958065e-01
2.19389752e-01 -3.11480239e-02 -7.30471849e-01 -1.03031015e+00
1.04125120e-01 -9.03370440e-01 1.03686407e-01 -9.95924890e-01
-7.10023940e-01 4.91260946e-01 2.05768108e-01 -1.03697014e+00
-4.63263869e-01 -3.82078260e-01 -7.52693415e-01 8.02281082e-01
-1.42420053e+00 -9.11093831e-01 -4.11086410e-01 1.41821235e-01
1.20572770e+00 2.45426700e-01 3.66728485e-01 -1.66311666e-01
-5.31105042e-01 2.22954601e-01 -5.14466882e-01 -2.92247236e-01
2.62294352e-01 -1.22844362e+00 7.87189841e-01 1.02766180e+00
5.06404281e-01 7.68782318e-01 1.17674434e+00 -6.04177415e-01
-1.08332288e+00 -1.15775645e+00 8.77459943e-01 -5.48193932e-01
1.50943860e-01 -3.09158444e-01 -8.56925011e-01 7.38317013e-01
1.63570300e-01 -3.56927305e-01 2.81041235e-01 1.18401378e-01
-3.89834605e-02 1.78648591e-01 -9.85532939e-01 1.13813508e+00
1.21605802e+00 -1.67250514e-01 -6.40143156e-01 1.69569612e-01
8.60207438e-01 -5.31698585e-01 2.92296056e-02 2.91944146e-01
4.85630035e-01 -1.25594747e+00 8.07086527e-01 -7.42825687e-01
7.80954421e-01 -5.18018842e-01 1.21790186e-01 -1.64612830e+00
-3.51109296e-01 -4.36119825e-01 8.16686600e-02 8.11807692e-01
7.07806170e-01 -8.70838523e-01 7.72985816e-01 8.91521215e-01
-3.41935873e-01 -7.18210697e-01 -7.70096958e-01 -5.55679560e-01
-2.31165826e-01 -7.66903639e-01 9.85867739e-01 7.29175448e-01
-3.96914005e-01 3.21695328e-01 -1.22577436e-01 4.18872774e-01
4.93105948e-01 5.63276649e-01 1.05992663e+00 -9.56590176e-01
-3.00514191e-01 -6.23168170e-01 -2.91751325e-01 -1.17687523e+00
2.79725939e-01 -6.88338399e-01 2.02791885e-01 -1.74669552e+00
2.01386526e-01 -3.08478266e-01 8.68491605e-02 4.99478996e-01
-3.84814382e-01 8.54542665e-03 4.49205875e-01 2.16372237e-01
-3.84073168e-01 6.47771358e-01 1.42769456e+00 -3.62507820e-01
-2.68448561e-01 1.03624567e-01 -1.00425541e+00 8.70867133e-01
1.04096341e+00 -5.97287774e-01 -6.71769023e-01 -6.09019637e-01
-1.13673873e-01 8.16417206e-03 7.69587636e-01 -7.38406718e-01
-2.42424719e-02 -5.76489329e-01 2.50961781e-01 -2.55584896e-01
2.64462203e-01 -8.01766753e-01 3.54061335e-01 6.64218605e-01
-4.32346433e-01 3.71397547e-02 3.30147058e-01 6.12182438e-01
-1.93014368e-01 -1.70443326e-01 5.35342574e-01 -3.24601866e-02
-9.97309387e-01 2.56911635e-01 -4.08946902e-01 -2.86258817e-01
1.08313715e+00 -4.40923125e-01 -3.23657781e-01 -6.10792160e-01
-6.81397259e-01 3.17460179e-01 5.38458407e-01 8.14765155e-01
6.80234253e-01 -1.13392448e+00 -5.53828597e-01 1.02828123e-01
3.80990595e-01 1.72863707e-01 3.39334965e-01 6.39128387e-01
-2.06394672e-01 6.12927198e-01 -1.53552800e-01 -5.99719048e-01
-1.09848619e+00 6.38838708e-01 3.79374951e-01 -2.47693688e-01
-5.32592893e-01 5.10940552e-01 9.37192261e-01 -2.61303008e-01
-2.76621431e-01 -6.01538599e-01 -2.79601723e-01 -2.53494591e-01
2.81102419e-01 1.01354711e-01 -1.45898446e-01 -3.43959540e-01
-2.15583950e-01 2.58191347e-01 -1.26670087e-02 -2.38409877e-01
9.91680384e-01 -2.63440788e-01 3.05682689e-01 3.18767130e-01
6.86864018e-01 -2.59307802e-01 -1.59198678e+00 2.17501089e-01
-5.60888462e-02 -7.14199901e-01 -1.21229127e-01 -7.99327314e-01
-9.32565510e-01 9.06856358e-01 5.60790062e-01 3.25270742e-01
9.72523153e-01 2.99822718e-01 5.32170892e-01 1.69339001e-01
2.96800137e-01 -7.65360117e-01 -8.84361491e-02 1.55178085e-01
1.14862514e+00 -1.47606194e+00 -3.47568631e-01 -4.76925492e-01
-7.60887980e-01 7.70428896e-01 7.27675498e-01 2.15435669e-01
1.79567680e-01 -2.48607203e-01 7.92421177e-02 -4.90236133e-01
-8.42227817e-01 -3.02805513e-01 4.37979102e-01 4.49030191e-01
1.94550246e-01 4.30579096e-01 -3.43072772e-01 6.54362381e-01
-4.88140494e-01 -8.65987316e-02 6.07206285e-01 4.50735033e-01
-3.08335841e-01 -8.19379449e-01 -4.21055406e-01 4.41948831e-01
-4.03451882e-02 6.08599621e-05 -4.76933300e-01 8.34901631e-01
4.10709232e-01 1.18692064e+00 1.40384734e-01 -1.92731753e-01
4.42388505e-01 -7.97814354e-02 1.89665630e-01 -8.10353041e-01
-8.23083818e-02 -2.83079237e-01 3.57878879e-02 -7.58754909e-01
-2.71582246e-01 -7.29629815e-01 -1.34144568e+00 -1.14363015e-01
-1.51794270e-01 1.74953640e-01 8.03207338e-01 1.23130679e+00
4.39712971e-01 7.84450054e-01 4.31614041e-01 -7.48024702e-01
-2.36224577e-01 -8.27458322e-01 -1.67984933e-01 6.08390450e-01
1.72066301e-01 -8.45613241e-01 -3.42846572e-01 2.46379301e-01]
|
[8.930286407470703, 5.3797125816345215]
|
ed2d7469-f900-43e4-afbd-ff51ef7acc03
|
dvhn-a-deep-hashing-framework-for-large-scale
|
2112.04937
| null |
https://arxiv.org/abs/2112.04937v1
|
https://arxiv.org/pdf/2112.04937v1.pdf
|
DVHN: A Deep Hashing Framework for Large-scale Vehicle Re-identification
|
In this paper, we make the very first attempt to investigate the integration of deep hash learning with vehicle re-identification. We propose a deep hash-based vehicle re-identification framework, dubbed DVHN, which substantially reduces memory usage and promotes retrieval efficiency while reserving nearest neighbor search accuracy. Concretely,~DVHN directly learns discrete compact binary hash codes for each image by jointly optimizing the feature learning network and the hash code generating module. Specifically, we directly constrain the output from the convolutional neural network to be discrete binary codes and ensure the learned binary codes are optimal for classification. To optimize the deep discrete hashing framework, we further propose an alternating minimization method for learning binary similarity-preserved hashing codes. Extensive experiments on two widely-studied vehicle re-identification datasets- \textbf{VehicleID} and \textbf{VeRi}-~have demonstrated the superiority of our method against the state-of-the-art deep hash methods. \textbf{DVHN} of $2048$ bits can achieve 13.94\% and 10.21\% accuracy improvement in terms of \textbf{mAP} and \textbf{Rank@1} for \textbf{VehicleID (800)} dataset. For \textbf{VeRi}, we achieve 35.45\% and 32.72\% performance gains for \textbf{Rank@1} and \textbf{mAP}, respectively.
|
['Zhengwei Qi', 'Kaicheng Guo', 'Chenggang Wu', 'Fangxin Liu', 'Sheng Zhang', 'Yongbiao Chen']
|
2021-12-09
| null | null | null | null |
['2048']
|
['playing-games']
|
[-3.70096833e-01 -1.55146420e-01 -5.85055172e-01 -6.42552257e-01
-1.04084289e+00 -4.91030127e-01 3.24721396e-01 2.26985529e-01
-7.16895401e-01 3.97673309e-01 -3.82090330e-01 -4.68042165e-01
-4.00239136e-03 -1.02846301e+00 -1.12385249e+00 -7.19734192e-01
-2.99295813e-01 4.49631423e-01 -9.51700751e-03 -8.89343321e-02
1.63849160e-01 5.35297513e-01 -1.74678349e+00 -1.66942522e-01
3.12647432e-01 1.35933530e+00 2.37824433e-02 4.97606248e-01
3.12161803e-01 4.44769651e-01 -3.36883157e-01 -7.01181173e-01
5.39385259e-01 1.66319653e-01 -5.44074595e-01 -6.52696609e-01
8.17311168e-01 -1.15724194e+00 -1.55007422e+00 1.08211458e+00
4.88127619e-01 7.19494559e-03 7.46460319e-01 -1.59969151e+00
-8.69138360e-01 4.82410043e-01 -5.10977089e-01 4.77074414e-01
-4.04572114e-02 1.65406182e-01 1.07809210e+00 -1.00000775e+00
2.44530246e-01 1.10251307e+00 1.00293791e+00 4.44112211e-01
-1.13753986e+00 -1.71100259e+00 -5.39988518e-01 7.15331435e-01
-2.24924541e+00 -5.87172568e-01 3.68552655e-01 -3.03737819e-01
1.08332705e+00 4.10222895e-02 3.44483733e-01 5.37889183e-01
2.63305157e-01 6.82356656e-01 2.77636945e-01 1.82119474e-01
-1.14135511e-01 2.18258295e-02 3.75624686e-01 8.64058673e-01
6.48763418e-01 4.99490052e-01 -3.66068631e-01 -8.80178511e-02
4.37322199e-01 3.04540902e-01 1.40827626e-01 -5.69882169e-02
-1.05684328e+00 1.30539536e+00 6.72921121e-01 -3.17986250e-01
4.05323915e-02 8.75056267e-01 5.49310386e-01 2.96533018e-01
-2.48516977e-01 -2.98983425e-01 1.84333548e-01 2.04879791e-01
-1.05208814e+00 6.24786675e-01 5.11952162e-01 1.30647087e+00
1.38796544e+00 1.40701775e-02 -9.50204879e-02 6.72344625e-01
6.68811262e-01 1.40496349e+00 1.41334236e-01 -9.57695842e-01
5.35605609e-01 8.59894454e-02 -1.03320040e-01 -1.33111703e+00
-2.48200446e-01 2.04356462e-02 -1.13111711e+00 -1.41562596e-01
-9.32254568e-02 1.44251570e-01 -9.58253622e-01 1.76072299e+00
1.08947359e-01 2.88012028e-01 1.12326518e-01 7.04599142e-01
1.13261485e+00 9.83475506e-01 7.13634938e-02 1.99799299e-01
1.17408144e+00 -6.62937343e-01 -4.10977036e-01 5.73598333e-02
7.89586186e-01 -5.65206230e-01 4.73059952e-01 -1.21120989e-01
-8.74627948e-01 -8.57729137e-01 -1.45903862e+00 -2.29375929e-01
-3.44005495e-01 -2.86465921e-02 2.42634669e-01 6.67258799e-01
-1.25824201e+00 2.88432777e-01 -4.87151325e-01 -5.30218007e-03
2.70783931e-01 9.27021146e-01 -3.08245152e-01 -3.03775817e-01
-1.53581762e+00 3.83647621e-01 4.01593536e-01 -1.46605954e-01
-1.33949566e+00 -4.58137542e-01 -1.16311455e+00 1.50698930e-01
-3.08375150e-01 -2.69735068e-01 1.09494269e+00 5.90633601e-02
-7.27200449e-01 9.52972114e-01 -1.63767278e-01 -9.58204567e-01
-1.92984007e-02 1.24571085e-01 -6.40543759e-01 2.80826241e-01
3.15324575e-01 1.23398471e+00 8.83570969e-01 -9.93580043e-01
-7.76077032e-01 -2.10940093e-01 -2.26744905e-01 -2.91231900e-01
-5.67407310e-01 -2.74151295e-01 -3.52874130e-01 -6.32309556e-01
-1.78764999e-01 -1.24314582e+00 2.02336773e-01 6.08354509e-02
-2.35518575e-01 -4.08514082e-01 1.21351707e+00 -4.73802626e-01
1.50823522e+00 -2.42673445e+00 -4.66791034e-01 5.64244807e-01
5.04135609e-01 4.36863095e-01 -2.79726267e-01 3.44725043e-01
1.50881231e-01 1.18801691e-01 3.71420532e-02 -4.52907830e-01
4.11094129e-01 2.78466821e-01 -4.95892793e-01 1.01923037e+00
-2.62107879e-01 1.01512837e+00 -4.91068661e-01 -5.68887711e-01
2.19639957e-01 6.93405807e-01 -5.84709167e-01 9.80110466e-03
3.21810573e-01 -4.01521891e-01 -3.17012638e-01 6.50353432e-01
1.18608046e+00 -2.65170872e-01 -2.05510899e-01 -3.99625331e-01
-6.28462285e-02 9.77586769e-03 -9.28992271e-01 1.10855949e+00
5.02747595e-02 9.08235610e-01 -1.86413899e-01 -9.75745916e-01
1.09831071e+00 2.16412283e-02 5.90759456e-01 -1.23408306e+00
2.30489552e-01 2.78096527e-01 -4.67397690e-01 -1.19731158e-01
8.58082116e-01 1.33565247e-01 -5.44957161e-01 4.21265870e-01
-1.06295580e-02 3.61050874e-01 -5.64046465e-02 1.40637055e-01
9.41051304e-01 -8.77067566e-01 -1.97511017e-01 -3.60568792e-01
5.77144086e-01 -1.35961100e-01 2.87608653e-01 9.33170676e-01
-3.93638492e-01 3.13136995e-01 -7.82628357e-02 -6.57739162e-01
-1.36851132e+00 -1.07054043e+00 -5.23046911e-01 1.05545115e+00
8.63563955e-01 -3.76566410e-01 -6.64613485e-01 -4.04162616e-01
6.60110235e-01 3.56025934e-01 -4.69377548e-01 -5.55597723e-01
-8.84978712e-01 -6.09055161e-01 1.45589817e+00 6.17770135e-01
8.98235440e-01 -5.42402387e-01 -3.43647778e-01 3.95330936e-02
-1.39474288e-01 -1.06628990e+00 -7.62121856e-01 1.49321146e-02
-4.14308816e-01 -8.16848218e-01 -5.44036448e-01 -1.30543721e+00
2.36368328e-01 6.29063308e-01 7.39793599e-01 4.44361031e-01
-2.66654342e-01 1.00222103e-01 -2.22832531e-01 -7.59581104e-03
-2.08908111e-01 1.53096899e-01 3.68700534e-01 -1.89562276e-01
7.88409948e-01 -1.67110711e-01 -1.09626448e+00 6.29812360e-01
-8.86420667e-01 -4.91632462e-01 6.12812877e-01 8.71122062e-01
8.54021311e-01 3.98949310e-02 4.95066762e-01 1.05459839e-02
-2.50346884e-02 -5.17642140e-01 -7.04110026e-01 -1.55528694e-01
-9.01727557e-01 4.10519540e-02 4.69941437e-01 -1.94922835e-01
-7.89036900e-02 3.64509621e-03 -3.84232253e-01 -5.60387015e-01
1.90454215e-01 1.56977996e-01 6.33576810e-02 -5.55764258e-01
3.66354764e-01 6.33003771e-01 5.86905740e-02 -2.03392997e-01
1.60135761e-01 9.32626367e-01 7.51825213e-01 -1.40816301e-01
1.19659591e+00 6.16470635e-01 -1.25426784e-01 -6.62583530e-01
-1.13248010e-03 -4.98883665e-01 -1.07188821e-01 2.95291748e-02
8.71790767e-01 -1.37048066e+00 -1.37504697e+00 5.41804910e-01
-9.50830042e-01 -2.45956272e-01 1.57430768e-01 3.74906808e-01
-4.94466722e-01 6.60590947e-01 -7.80513763e-01 -3.52432936e-01
-7.18087435e-01 -1.38226950e+00 1.44534791e+00 1.06838927e-01
1.07501604e-01 -4.53612775e-01 -9.27566439e-02 2.27937877e-01
3.79061788e-01 -9.42963064e-02 1.00261486e+00 -5.99301875e-01
-8.57053399e-01 -5.12381256e-01 -7.32236087e-01 3.08033168e-01
-3.33887666e-01 -3.67822766e-01 -8.84546041e-01 -8.72087359e-01
-6.10942125e-01 -4.17986482e-01 9.18145597e-01 1.37981817e-01
1.28388190e+00 -5.84995925e-01 -5.21236181e-01 1.02504051e+00
1.43532598e+00 3.92502308e-01 8.59456718e-01 4.11035955e-01
7.41913974e-01 -1.10714234e-01 6.15604937e-01 7.98903823e-01
1.01713562e+00 8.40723217e-01 6.17301285e-01 8.37185979e-02
-1.06669299e-01 -3.72877061e-01 2.93103427e-01 5.77667832e-01
6.17249429e-01 -2.66286641e-01 -9.42310631e-01 6.42560959e-01
-1.51213765e+00 -1.07819009e+00 1.92205906e-01 2.16575289e+00
5.88859081e-01 -4.66477536e-02 1.28301471e-01 1.40490815e-01
9.71142232e-01 2.59213179e-01 -6.61452055e-01 -2.35244960e-01
4.06739600e-02 -2.52876971e-02 1.21563947e+00 4.89133239e-01
-1.48016191e+00 8.01612139e-01 5.36828661e+00 1.00051558e+00
-1.18748999e+00 -1.06935367e-01 7.20958948e-01 1.33058086e-01
-4.17983502e-01 -3.62759590e-01 -1.46205139e+00 7.31926560e-01
1.29988313e+00 -3.06597710e-01 4.58455116e-01 9.70376074e-01
-3.45892429e-01 3.23645681e-01 -1.17941236e+00 1.56164289e+00
2.21214443e-01 -1.58143413e+00 3.95254254e-01 3.91008794e-01
5.12249231e-01 4.20964271e-01 4.77808416e-01 5.68795502e-01
1.45286933e-01 -1.21720815e+00 9.54114139e-01 1.76048443e-01
1.52937567e+00 -1.17005444e+00 7.21275806e-01 -4.26879935e-02
-1.86006081e+00 -2.73371547e-01 -5.80585778e-01 4.97827262e-01
1.01376148e-02 1.33842811e-01 -7.87315726e-01 8.54245499e-02
1.26297569e+00 6.90468132e-01 -4.56873357e-01 8.56442809e-01
6.28782332e-01 2.50626802e-01 -3.83412123e-01 -1.66875646e-02
3.88051838e-01 3.99717212e-01 2.29107276e-01 1.42780757e+00
5.28712094e-01 1.74319774e-01 2.31910214e-01 5.95649004e-01
-3.82935196e-01 -2.29080066e-01 -6.99691057e-01 -3.45884007e-03
1.00036955e+00 7.88239777e-01 -3.02201360e-01 -3.90576929e-01
-1.59935355e-01 6.85170293e-01 -1.96321178e-02 3.60745758e-01
-1.41466808e+00 -9.44222808e-01 1.07008815e+00 2.44641975e-01
8.36469233e-01 -2.90268451e-01 -9.76505876e-02 -6.43566847e-01
-2.52289832e-01 -6.28740668e-01 3.76346380e-01 -1.75671369e-01
-9.36142445e-01 6.21329010e-01 8.06362927e-02 -1.30384314e+00
-2.14236259e-01 -4.36934382e-01 -3.33529413e-02 5.02341390e-01
-1.72348845e+00 -9.91488814e-01 -5.01941621e-01 8.59823763e-01
7.20577985e-02 -3.58393133e-01 5.05279124e-01 9.99211967e-01
-4.72766876e-01 1.66991031e+00 7.21580148e-01 4.12551790e-01
6.14497781e-01 -5.83868027e-01 7.52180099e-01 3.98078322e-01
-2.35534772e-01 5.55654287e-01 3.07489783e-01 -5.62708855e-01
-2.07515836e+00 -1.47445846e+00 9.48733151e-01 -2.50527132e-02
4.49713796e-01 -4.44743454e-01 -7.92810619e-01 6.12804651e-01
-1.78417563e-01 1.61657184e-01 6.60128415e-01 -7.67789245e-01
-7.42837012e-01 -7.72335172e-01 -1.39183128e+00 1.87055692e-01
8.33285034e-01 -9.07599986e-01 -1.93746999e-01 8.16536322e-02
8.60515475e-01 -2.33639196e-01 -1.03697872e+00 8.05287302e-01
8.18914950e-01 -6.54529750e-01 1.48969924e+00 9.27063078e-02
-2.05519542e-01 -4.40697879e-01 -6.16315246e-01 -2.30050877e-01
-5.67422628e-01 -3.23809952e-01 -3.25553596e-01 9.24992561e-01
1.62533596e-01 -6.17430329e-01 9.08011794e-01 3.31544101e-01
-4.05076891e-02 -4.41771209e-01 -1.36829984e+00 -6.95142806e-01
1.05146565e-01 -3.91344786e-01 9.11504984e-01 5.57801306e-01
-2.94454783e-01 -1.91739783e-01 -5.17480135e-01 4.23128396e-01
1.00450909e+00 1.10938698e-01 9.44551945e-01 -9.99017179e-01
1.02547817e-01 -3.26727748e-01 -8.03166389e-01 -1.49728894e+00
3.63180190e-01 -9.75005984e-01 2.44211197e-01 -9.91573215e-01
4.80836093e-01 -8.79587233e-01 -4.83068466e-01 5.80069721e-01
3.82557899e-01 8.78326535e-01 1.46605238e-01 5.26333094e-01
-8.71598721e-01 6.22547448e-01 5.55135667e-01 -6.07390106e-01
4.48503941e-01 -3.65667969e-01 -4.30900514e-01 6.08610585e-02
6.01921082e-01 -5.51636636e-01 -3.77748400e-01 -7.37807989e-01
8.37513432e-02 6.91993386e-02 7.14012921e-01 -1.10587013e+00
5.31513035e-01 3.03374946e-01 3.63205254e-01 -1.02904487e+00
3.07114363e-01 -7.43614614e-01 2.79613703e-01 6.83663905e-01
-1.98670864e-01 4.24703628e-01 3.52935582e-01 7.53812730e-01
-2.80065477e-01 1.59963146e-02 8.76086771e-01 4.71159846e-01
-9.19585884e-01 7.16087997e-01 -3.66287917e-01 -2.31281787e-01
1.09175563e+00 -4.33218777e-01 -5.10502100e-01 -5.61719298e-01
-4.95820269e-02 5.01335919e-01 4.70035285e-01 3.63933563e-01
1.12540579e+00 -1.77829158e+00 -5.91028273e-01 6.33750677e-01
3.38706017e-01 -2.01083258e-01 4.07102555e-01 4.37649310e-01
-7.84382761e-01 9.16326404e-01 -8.46376121e-02 -6.59247577e-01
-1.09829593e+00 5.96006215e-01 1.84666961e-01 2.59475976e-01
-5.66357791e-01 8.21673691e-01 2.89538652e-01 -1.69674397e-01
6.71759367e-01 -1.01724155e-01 1.60684884e-01 -1.44191846e-01
7.41366506e-01 6.93695962e-01 2.48138979e-02 -1.22461963e+00
-5.16129553e-01 7.33885705e-01 -4.17789966e-01 2.37062156e-01
8.94374490e-01 -3.57127517e-01 1.27978876e-01 -4.88670945e-01
2.14008689e+00 -6.50784016e-01 -1.16627371e+00 -1.98815033e-01
-2.73670256e-01 -4.51271445e-01 2.19789445e-02 -1.48068041e-01
-1.26149845e+00 7.36479819e-01 1.21884918e+00 -1.13392420e-01
7.95316696e-01 8.66944790e-02 1.78610969e+00 8.51423860e-01
5.60013473e-01 -8.93669069e-01 2.39802971e-02 6.52887106e-01
3.67014408e-01 -1.28699136e+00 -1.67851478e-01 2.02255204e-01
-1.30012274e-01 8.95870566e-01 5.63849330e-01 -1.89928532e-01
8.42207432e-01 1.64768234e-01 -2.45812714e-01 -2.63996065e-01
-1.72992915e-01 1.52858570e-01 1.96223840e-01 2.20187470e-01
-2.92062551e-01 3.14295962e-02 3.34460214e-02 3.93433690e-01
-3.27690512e-01 -1.23517998e-01 -1.11885354e-01 8.73878539e-01
-8.40408921e-01 -7.64234602e-01 -2.94565380e-01 3.71568650e-01
-2.25562736e-01 -1.73361450e-01 1.87980413e-01 7.12834179e-01
1.18382528e-01 6.97399974e-01 3.00495058e-01 -1.14728105e+00
2.48074327e-02 -3.75390261e-01 4.89675626e-02 2.69856136e-02
-1.47736162e-01 -2.78948218e-01 -2.95829475e-01 -5.35829127e-01
1.57208979e-01 -4.81852204e-01 -1.69248307e+00 -1.07504308e+00
-2.04924673e-01 1.85852513e-01 4.79980111e-01 5.02475977e-01
4.36692089e-01 -2.78286129e-01 9.54232097e-01 -9.32123899e-01
-6.99835241e-01 -2.55894542e-01 -5.08364201e-01 2.47742116e-01
8.49936426e-01 -6.59847856e-01 -3.82219583e-01 -1.42784998e-01]
|
[11.40276050567627, 0.9134751558303833]
|
904a0b91-4d3d-4b4b-93c9-e70ef37d9542
|
dae-talker-high-fidelity-speech-driven
|
2303.17550
| null |
https://arxiv.org/abs/2303.17550v2
|
https://arxiv.org/pdf/2303.17550v2.pdf
|
DAE-Talker: High Fidelity Speech-Driven Talking Face Generation with Diffusion Autoencoder
|
While recent research has made significant progress in speech-driven talking face generation, the quality of the generated video still lags behind that of real recordings. One reason for this is the use of handcrafted intermediate representations like facial landmarks and 3DMM coefficients, which are designed based on human knowledge and are insufficient to precisely describe facial movements. Additionally, these methods require an external pretrained model for extracting these representations, whose performance sets an upper bound on talking face generation. To address these limitations, we propose a novel method called DAE-Talker that leverages data-driven latent representations obtained from a diffusion autoencoder (DAE). DAE contains an image encoder that encodes an image into a latent vector and a DDIM image decoder that reconstructs the image from it. We train our DAE on talking face video frames and then extract their latent representations as the training target for a Conformer-based speech2latent model. This allows DAE-Talker to synthesize full video frames and produce natural head movements that align with the content of speech, rather than relying on a predetermined head pose from a template video. We also introduce pose modelling in speech2latent for pose controllability. Additionally, we propose a novel method for generating continuous video frames with the DDIM image decoder trained on individual frames, eliminating the need for modelling the joint distribution of consecutive frames directly. Our experiments show that DAE-Talker outperforms existing popular methods in lip-sync, video fidelity, and pose naturalness. We also conduct ablation studies to analyze the effectiveness of the proposed techniques and demonstrate the pose controllability of DAE-Talker.
|
['Jiang Bian', 'Sheng Zhao', 'Kai Yu', 'Xie Chen', 'Xu Tan', 'Tianyu He', 'Qi Chen', 'Chenpng Du']
|
2023-03-30
| null | null | null | null |
['talking-face-generation', 'face-generation']
|
['computer-vision', 'computer-vision']
|
[ 1.33879617e-01 4.08333719e-01 -1.63668245e-01 -4.25030142e-01
-8.37751865e-01 -4.02557194e-01 6.72224045e-01 -9.05842066e-01
5.81953563e-02 4.27897602e-01 5.97951055e-01 7.51598999e-02
3.41885209e-01 -4.24301744e-01 -9.80143428e-01 -7.65065968e-01
2.41557792e-01 1.35601878e-01 -1.12980701e-01 -6.92704367e-03
-1.22495070e-01 4.60353136e-01 -1.78934276e+00 2.32959390e-01
5.13640463e-01 9.78009224e-01 3.49175662e-01 7.29974985e-01
4.69428450e-02 6.54110670e-01 -6.94576144e-01 -4.06411886e-01
2.71021515e-01 -6.94184303e-01 -3.38286877e-01 5.18792689e-01
5.31682193e-01 -6.75622106e-01 -5.14711201e-01 7.45121956e-01
7.40086615e-01 -9.55206603e-02 6.52838469e-01 -1.38363504e+00
-3.78958464e-01 4.97775882e-01 -4.65568006e-01 -2.73051441e-01
6.03556275e-01 3.44869196e-01 5.74248493e-01 -1.04467869e+00
7.37708151e-01 1.54038215e+00 5.12439728e-01 1.05484784e+00
-1.17101753e+00 -8.67303729e-01 5.64490557e-02 -8.42111707e-02
-1.60682952e+00 -1.41370082e+00 9.16364253e-01 -3.32557321e-01
5.83216250e-01 1.14589028e-01 6.13448083e-01 1.54932392e+00
-1.30556464e-01 7.85312295e-01 6.31729245e-01 -3.95583957e-01
1.79004148e-01 4.11417000e-02 -6.12505913e-01 7.62215853e-01
-2.37115264e-01 2.69660383e-01 -9.55661833e-01 2.50534564e-02
8.76687706e-01 -3.04522634e-01 -6.36328638e-01 -3.79199982e-01
-1.03774726e+00 7.30736315e-01 5.90159856e-02 9.36867148e-02
-4.36405212e-01 2.44658485e-01 1.44897789e-01 5.76692186e-02
3.48865509e-01 -2.00893030e-01 2.10349299e-02 -1.70845196e-01
-1.30096161e+00 1.75211564e-01 6.93686843e-01 1.24231267e+00
5.13473749e-01 4.64293689e-01 -1.34029418e-01 7.72873342e-01
4.60940808e-01 8.39940131e-01 5.51626921e-01 -1.19423628e+00
4.38681483e-01 3.55859250e-02 4.46713576e-03 -1.07128429e+00
8.13538302e-03 3.36645618e-02 -5.50277591e-01 5.79087324e-02
3.90102752e-02 -3.11677128e-01 -9.92937982e-01 2.06268311e+00
3.94842476e-01 2.06340045e-01 8.83542225e-02 8.98896635e-01
7.08924711e-01 8.33260775e-01 -2.04154074e-01 -5.36677241e-01
9.02919292e-01 -9.10435140e-01 -9.95411754e-01 -7.12696537e-02
1.82660311e-01 -7.47881830e-01 8.83366942e-01 1.98439285e-01
-1.34675097e+00 -5.90215206e-01 -8.53889167e-01 7.09311366e-02
2.21536726e-01 2.76246130e-01 2.74188429e-01 7.48882234e-01
-1.25157428e+00 1.10247776e-01 -8.16452503e-01 -2.33770609e-01
1.49935469e-01 4.37553763e-01 -5.11132061e-01 1.31906047e-01
-9.43421423e-01 6.47481740e-01 6.00222461e-02 1.11582026e-01
-1.26052809e+00 -5.48638880e-01 -1.20640242e+00 -3.28601561e-02
1.92145854e-01 -6.28104091e-01 1.28351045e+00 -1.28264046e+00
-2.21967721e+00 6.66262031e-01 -5.89808345e-01 -3.99332315e-01
6.17570937e-01 -1.05881013e-01 -4.59038019e-01 4.45961714e-01
-4.28417251e-02 1.15608203e+00 1.43410325e+00 -1.42529619e+00
-2.24582851e-01 -1.50989983e-02 -7.85696879e-02 2.49921486e-01
-4.18893248e-01 -8.95477757e-02 -9.29216743e-01 -7.60881722e-01
-1.27606690e-02 -1.17865419e+00 2.43744537e-01 1.49998456e-01
-4.01831627e-01 1.55876726e-01 1.21199405e+00 -6.45338774e-01
1.08962750e+00 -2.20316100e+00 2.69207656e-01 4.83934721e-03
-3.34620066e-02 2.39923239e-01 -2.93924749e-01 2.44211331e-01
-1.89550936e-01 -6.88003078e-02 -9.81881991e-02 -6.98940396e-01
-1.22300617e-01 1.78472713e-01 -4.17679876e-01 5.01389980e-01
8.45992938e-02 6.92199588e-01 -5.97119391e-01 -5.65367341e-01
3.05125475e-01 9.88586426e-01 -9.64072645e-01 4.20830607e-01
-2.26472110e-01 7.13146567e-01 -2.01334551e-01 5.18280029e-01
6.07275784e-01 8.97205845e-02 2.39383802e-01 -5.07765293e-01
-1.92302074e-02 1.51165575e-01 -9.74668384e-01 1.83151913e+00
-7.85316169e-01 5.75832486e-01 3.95428807e-01 -6.52674139e-01
8.03378284e-01 6.99693680e-01 5.26971936e-01 -3.57485563e-01
2.60168761e-01 8.57747197e-02 -2.90372342e-01 -6.76984429e-01
1.19229473e-01 -2.64948756e-01 1.55159220e-01 2.93390423e-01
2.85616517e-01 -2.95980096e-01 -8.00453797e-02 -1.45162810e-02
6.83150291e-01 2.12943554e-01 -7.53868669e-02 5.20551428e-02
6.28010452e-01 -6.81347668e-01 5.32937944e-01 1.40924662e-01
2.25714091e-02 1.02171648e+00 2.24671692e-01 -7.22075626e-02
-9.05347109e-01 -1.06913435e+00 1.04931779e-01 6.70665205e-01
-1.57049671e-02 -5.52014410e-01 -1.19335103e+00 -5.95619857e-01
-3.96351218e-01 7.36366868e-01 -4.80796874e-01 -1.47579148e-01
-6.34143710e-01 -2.10290879e-01 6.17297709e-01 3.06733757e-01
4.88317907e-01 -8.89738083e-01 -3.84698719e-01 8.28582868e-02
-5.66385984e-01 -1.44203568e+00 -9.97736752e-01 -4.90340263e-01
-4.57997173e-01 -7.61450887e-01 -1.06085575e+00 -7.20959127e-01
9.33624566e-01 2.08810389e-01 7.28293300e-01 -2.54760742e-01
-1.60092518e-01 5.63419402e-01 -1.43551171e-01 -1.99048132e-01
-7.07396328e-01 -1.93761930e-01 5.57523549e-01 5.87277591e-01
-4.96786721e-02 -7.12867916e-01 -7.47515500e-01 4.12791848e-01
-9.32645500e-01 3.39794159e-01 4.00235742e-01 7.62436986e-01
4.04935420e-01 -1.64111495e-01 4.67476785e-01 -1.97986379e-01
4.37144428e-01 -2.87468940e-01 -5.92975020e-01 7.42910989e-03
-1.60385400e-01 1.31161317e-01 4.88518596e-01 -5.66532373e-01
-1.23712492e+00 3.65904301e-01 -4.18189049e-01 -1.03674948e+00
-3.27591002e-02 -5.16183525e-02 -6.12738967e-01 4.67022173e-02
4.28745568e-01 4.36600983e-01 3.80637735e-01 -1.98582873e-01
5.45037925e-01 8.97120595e-01 6.80119395e-01 -5.70225179e-01
7.68533170e-01 5.86765409e-01 -2.81934232e-01 -1.07810867e+00
-3.28505039e-01 9.54828486e-02 -3.52096885e-01 -3.40361714e-01
9.52686965e-01 -1.06923485e+00 -9.66851473e-01 4.22473818e-01
-1.22778308e+00 -2.05428392e-01 -7.94404596e-02 5.35108209e-01
-9.22720492e-01 2.13031113e-01 -4.94763315e-01 -8.34784269e-01
-2.25783914e-01 -1.54778814e+00 1.39895403e+00 -1.09515851e-02
-3.00240397e-01 -7.17878401e-01 -2.05698997e-01 5.60089946e-01
2.97531128e-01 1.24303505e-01 4.81545329e-01 3.06565799e-02
-6.29896581e-01 -3.64583805e-02 2.77482778e-01 4.32895184e-01
4.30626392e-01 6.33165315e-02 -1.10852194e+00 -4.30345953e-01
1.00401811e-01 -3.23337078e-01 5.09637058e-01 5.53167284e-01
9.50965464e-01 -7.24168897e-01 -3.32951844e-01 7.63672173e-01
8.60855222e-01 2.17245519e-01 5.16237915e-01 -1.71762437e-01
6.31198108e-01 7.87382841e-01 2.37390414e-01 4.17600065e-01
4.14377421e-01 1.08558953e+00 1.79540247e-01 6.39806613e-02
-6.08463049e-01 -6.84234321e-01 8.77536356e-01 9.91805732e-01
-2.99490001e-02 -3.15200478e-01 -5.91461837e-01 5.29171050e-01
-1.41119087e+00 -1.03514385e+00 5.74739695e-01 2.12699127e+00
8.41424882e-01 -1.63828418e-01 9.15082544e-02 1.04472801e-01
8.23711038e-01 1.60129011e-01 -5.05239546e-01 -1.73088193e-01
3.03561147e-02 -1.54562416e-02 2.40069345e-01 6.98552608e-01
-7.33428121e-01 1.09709644e+00 6.44064617e+00 7.35331118e-01
-1.61343086e+00 1.98435746e-02 5.22941470e-01 -3.32211852e-01
-3.73273522e-01 -1.49508044e-01 -9.16391671e-01 4.95667189e-01
1.06472552e+00 -1.62448555e-01 4.29778308e-01 7.49756157e-01
7.64900625e-01 2.51376867e-01 -1.14286625e+00 1.36230755e+00
4.88713592e-01 -1.34961605e+00 4.24489349e-01 1.63032025e-01
6.35849118e-01 -4.32855636e-01 3.84794772e-01 1.27080521e-02
-4.99577895e-02 -1.01352668e+00 1.06420994e+00 3.84670913e-01
1.20647037e+00 -6.60606861e-01 8.07501078e-02 2.81937927e-01
-1.09534407e+00 -1.78585539e-03 -7.17139337e-03 4.46334898e-01
4.11540985e-01 2.36750931e-01 -1.13853836e+00 2.81681448e-01
4.11606878e-01 3.60779494e-01 -5.19733876e-02 4.00068253e-01
-2.56619602e-01 5.65095067e-01 -3.59524965e-01 5.20702720e-01
-5.16518094e-02 1.52828157e-01 6.70898974e-01 9.47277725e-01
5.91144562e-01 3.03032566e-02 -2.21878057e-03 8.62828910e-01
-2.05319762e-01 -5.88555671e-02 -7.69545257e-01 -1.17066577e-01
4.87912476e-01 9.93939459e-01 -2.65816838e-01 -1.61929071e-01
-4.49150383e-01 1.12337291e+00 -2.33825222e-01 4.51365203e-01
-9.50350165e-01 3.48147191e-02 8.79095793e-01 4.47689861e-01
3.84007633e-01 -2.60225236e-01 2.63218015e-01 -1.27101696e+00
2.01812722e-02 -1.23259544e+00 -2.12488309e-01 -8.33349824e-01
-7.38151968e-01 9.49162304e-01 1.20256856e-01 -1.32697880e+00
-8.82069051e-01 -2.17408419e-01 -4.20161337e-01 6.53141141e-01
-1.25039482e+00 -1.30760860e+00 -1.59822360e-01 9.77290869e-01
1.00395572e+00 -2.93421477e-01 7.56883979e-01 2.54251927e-01
-5.90176582e-01 1.00540733e+00 -3.48128825e-01 1.14071026e-01
6.01314425e-01 -4.32070166e-01 3.33415121e-01 7.24466145e-01
8.68669897e-02 6.48310125e-01 8.45040262e-01 -4.41048741e-01
-1.53668678e+00 -1.01695359e+00 6.09380841e-01 -1.31593734e-01
2.14908972e-01 -6.27343476e-01 -4.50550884e-01 6.95295155e-01
1.27856389e-01 8.35147426e-02 4.53097880e-01 -5.72166264e-01
-1.79872319e-01 -2.73885876e-01 -1.06765544e+00 7.65239716e-01
1.11194658e+00 -6.32391274e-01 -3.50323319e-01 8.24682936e-02
6.99993491e-01 -5.15435636e-01 -5.13247907e-01 4.13193285e-01
6.10831499e-01 -9.62322712e-01 9.01878834e-01 4.66190390e-02
3.00675005e-01 -2.77151883e-01 -1.97647363e-01 -1.18847990e+00
2.95127064e-01 -1.09788239e+00 -1.95724607e-01 1.38400924e+00
2.07509443e-01 -2.84218192e-01 9.30503607e-01 5.90444207e-01
3.72378677e-02 -5.00817001e-01 -9.53112543e-01 -7.64310658e-01
-2.16115952e-01 -4.01625454e-01 7.30880618e-01 6.91194177e-01
-2.25325033e-01 2.52029747e-01 -7.42598712e-01 2.94958085e-01
5.74671924e-01 -8.11236352e-02 1.05665493e+00 -5.95991492e-01
-2.45173961e-01 -2.88485158e-02 -3.81327927e-01 -1.36416841e+00
6.53484881e-01 -6.10148132e-01 2.79472381e-01 -1.06783092e+00
-1.46256790e-01 -9.16404799e-02 4.23520148e-01 3.36051941e-01
1.86846867e-01 7.87529647e-02 2.63464749e-01 2.91430771e-01
3.15571699e-04 9.15902913e-01 1.27741778e+00 -1.23260990e-01
-3.18079263e-01 -9.43951160e-02 -3.89043212e-01 9.27402377e-01
4.54033047e-01 -2.25578189e-01 -8.94464552e-01 -5.27822196e-01
-3.75828087e-01 6.01399779e-01 1.97153494e-01 -1.14913642e+00
2.77858913e-01 -9.59781930e-02 3.40861470e-01 -2.38297448e-01
8.65617096e-01 -7.73926497e-01 3.01674247e-01 8.00966173e-02
-3.37374061e-01 -6.62708879e-02 1.39386788e-01 3.91938984e-01
-3.33667845e-01 1.95519939e-01 8.32283378e-01 1.23891905e-01
-3.90016764e-01 5.13646126e-01 -3.94716561e-01 -2.42534876e-01
9.15061951e-01 -3.23130369e-01 2.14148328e-01 -1.07932925e+00
-7.57244051e-01 -2.32094333e-01 4.82342511e-01 7.21351862e-01
8.27850223e-01 -1.41632140e+00 -6.20829523e-01 6.45408273e-01
-1.58579305e-01 -1.41010404e-01 2.45112881e-01 5.21807075e-01
-3.01567495e-01 4.57541019e-01 -5.57919294e-02 -7.67230332e-01
-1.34754646e+00 5.99558175e-01 3.06962341e-01 4.12520230e-01
-7.43354499e-01 7.53208399e-01 6.79554403e-01 -2.64229059e-01
2.98038691e-01 -1.29960580e-02 2.56224163e-02 -1.18925124e-01
5.81594586e-01 -1.12020507e-01 -3.43020484e-02 -1.23732626e+00
-2.99120188e-01 7.10323274e-01 1.80444270e-01 -5.73729634e-01
1.14539957e+00 -3.58064026e-01 3.48862708e-01 7.25393072e-02
1.41099536e+00 3.38622004e-01 -1.69795549e+00 1.36815235e-01
-5.30206144e-01 -4.60686535e-01 9.69225690e-02 -3.55535656e-01
-1.41755140e+00 8.50390196e-01 5.73811650e-01 -5.34112811e-01
1.23770511e+00 -1.22285955e-01 1.00872362e+00 5.43579645e-02
3.98160547e-01 -7.71244109e-01 3.32959026e-01 1.01814054e-01
1.27145207e+00 -9.04168129e-01 -3.93925935e-01 -5.96219242e-01
-7.24040508e-01 1.08325672e+00 4.55908597e-01 3.89340430e-01
6.38684809e-01 5.04766643e-01 2.80646145e-01 1.49538830e-01
-6.99455261e-01 -6.46225885e-02 1.57515064e-01 7.11747885e-01
3.84405226e-01 -2.46119678e-01 7.99173564e-02 3.47242236e-01
-5.86291552e-01 2.33683795e-01 2.95241356e-01 6.86399877e-01
-1.33777097e-01 -1.06567836e+00 -5.55397451e-01 -2.84980327e-01
-3.39799017e-01 -4.72397283e-02 -1.54946864e-01 4.95928913e-01
3.09972726e-02 1.14133632e+00 2.30694897e-02 -4.87888336e-01
1.56726241e-01 1.19512267e-01 5.49706042e-01 -3.36920291e-01
6.96726292e-02 3.37458581e-01 -1.44918112e-03 -5.94392896e-01
-3.57289553e-01 -3.92963409e-01 -1.11244559e+00 -3.54990363e-01
-2.79904574e-01 1.33427724e-01 8.10194969e-01 8.08898747e-01
5.91581285e-01 2.89926797e-01 8.82039726e-01 -1.25920665e+00
-2.74144828e-01 -7.88734794e-01 -2.05084249e-01 4.27523077e-01
5.31608999e-01 -8.08187842e-01 -4.50710893e-01 5.81103981e-01]
|
[13.246081352233887, -0.41128629446029663]
|
af184747-610e-4038-acd2-791ec6cb124c
|
is-the-computation-of-abstract-sameness
|
2205.06149
| null |
https://arxiv.org/abs/2205.06149v1
|
https://arxiv.org/pdf/2205.06149v1.pdf
|
Is the Computation of Abstract Sameness Relations Human-Like in Neural Language Models?
|
In recent years, deep neural language models have made strong progress in various NLP tasks. This work explores one facet of the question whether state-of-the-art NLP models exhibit elementary mechanisms known from human cognition. The exploration is focused on a relatively primitive mechanism for which there is a lot of evidence from various psycholinguistic experiments with infants. The computation of "abstract sameness relations" is assumed to play an important role in human language acquisition and processing, especially in learning more complex grammar rules. In order to investigate this mechanism in BERT and other pre-trained language models (PLMs), the experiment designs from studies with infants were taken as the starting point. On this basis, we designed experimental settings in which each element from the original studies was mapped to a component of language models. Even though the task in our experiments was relatively simple, the results suggest that the cognitive faculty of computing abstract sameness relations is stronger in infants than in all investigated PLMs.
|
['Benjamin Roth', 'Lukas Thoma']
|
2022-05-12
| null | null | null | null |
['language-acquisition']
|
['natural-language-processing']
|
[-1.32378891e-01 4.78025883e-01 1.58674106e-01 -4.74302262e-01
4.67634678e-01 -3.82627547e-01 7.85775840e-01 5.46977520e-01
-6.23531520e-01 2.14220256e-01 9.85215008e-02 -5.60910642e-01
-2.26802900e-01 -9.34263289e-01 -9.52570140e-01 -3.60960960e-01
7.71434978e-03 4.83226359e-01 2.70756602e-01 -4.32770520e-01
7.15777516e-01 6.88758492e-01 -1.56492019e+00 4.09693241e-01
9.03090179e-01 2.62844086e-01 7.17030942e-01 3.75741512e-01
-4.76718009e-01 7.25665927e-01 -2.45256826e-01 -6.71756923e-01
1.86186712e-02 -6.42079353e-01 -1.02018011e+00 -5.30387044e-01
1.90229371e-01 -2.45904550e-01 -1.22059174e-01 1.18688846e+00
2.77146727e-01 1.05003916e-01 6.37826860e-01 -7.20504940e-01
-1.10374200e+00 1.01672387e+00 -5.77429682e-02 3.83523434e-01
5.66707730e-01 -1.60586182e-02 9.13939357e-01 -9.70009685e-01
3.66768658e-01 1.55810869e+00 3.35428149e-01 5.87794423e-01
-9.72463131e-01 -4.98131603e-01 3.16386282e-01 3.32317084e-01
-1.15615869e+00 -1.03906542e-01 5.92223167e-01 -4.77933794e-01
1.24523413e+00 -1.76475391e-01 1.03355205e+00 7.63004661e-01
4.74853218e-01 5.03754675e-01 1.57747746e+00 -9.77124691e-01
-1.22327320e-01 5.84672511e-01 5.26958525e-01 8.19402039e-01
6.22789800e-01 4.23378646e-01 -8.49189281e-01 3.71028334e-01
1.02977383e+00 -4.60314482e-01 -2.12125406e-01 7.13669062e-02
-1.07346046e+00 6.74195945e-01 2.54849166e-01 1.18219185e+00
-2.83880383e-01 -1.25289306e-01 1.90158457e-01 4.97775912e-01
-8.45395327e-02 6.37163460e-01 -6.35373533e-01 7.41090477e-02
-6.75794065e-01 -6.27529621e-02 9.06135798e-01 8.52153420e-01
7.80308425e-01 -7.57528543e-02 8.86265710e-02 7.57169783e-01
6.61916077e-01 2.08131745e-01 6.79986656e-01 -4.05655831e-01
2.97033191e-01 7.54094183e-01 -4.65812117e-01 -9.11791921e-01
-5.59795320e-01 -2.03022316e-01 -3.40791196e-01 5.88190965e-02
5.84218025e-01 1.48978487e-01 -6.29428923e-01 2.04869413e+00
1.02589376e-01 -2.30113983e-01 -1.29001234e-02 5.93628526e-01
9.54926074e-01 7.82202363e-01 6.23978972e-01 -3.64195317e-01
1.54576492e+00 -4.64396238e-01 -5.73394716e-01 -1.05228089e-01
7.60768414e-01 -5.17552316e-01 1.57019162e+00 5.11998117e-01
-1.25486481e+00 -8.77065659e-01 -1.10557783e+00 -3.59337240e-01
-6.30643845e-01 -1.15094386e-01 1.15077531e+00 8.48142385e-01
-1.37133908e+00 5.92572868e-01 -6.71179235e-01 -8.21806908e-01
-4.70150821e-03 2.00444892e-01 -4.20953244e-01 1.93102397e-02
-1.32757246e+00 1.33943868e+00 8.61772478e-01 1.80393532e-01
-9.54892576e-01 -2.95609027e-01 -6.45432115e-01 2.16915905e-01
2.57591158e-01 -4.85271275e-01 1.16327870e+00 -9.34647679e-01
-1.47715569e+00 1.50064421e+00 -1.41537815e-01 -7.30290115e-02
-2.49911219e-01 -3.92447591e-01 -3.57907772e-01 1.35986805e-01
-2.11394757e-01 5.05971670e-01 4.25390691e-01 -1.11962092e+00
-2.62875170e-01 -4.89794016e-01 2.55840003e-01 1.72583148e-01
-9.96631291e-03 6.66716933e-01 2.05827445e-01 -6.05143249e-01
4.99774516e-01 -6.15161121e-01 2.27029726e-01 -2.41270795e-01
2.24326044e-01 -7.77726054e-01 -3.87439251e-01 -5.70564866e-01
1.12850070e+00 -1.82224834e+00 9.00670290e-02 -9.57091153e-02
2.30549201e-01 1.68312147e-01 -3.08609847e-03 7.42488027e-01
-1.75472438e-01 3.65986615e-01 1.24021798e-01 2.50570513e-02
2.54925907e-01 2.90313542e-01 -2.34486178e-01 1.93347588e-01
2.72577051e-02 1.00233340e+00 -6.08485222e-01 -5.92373788e-01
-2.37526789e-01 2.81567015e-02 -5.97422123e-01 3.70969415e-01
-4.30507839e-01 3.43752474e-01 -2.76702911e-01 4.83763427e-01
5.24213612e-01 5.44329397e-02 2.79434294e-01 2.89493263e-01
-4.78889197e-01 7.53971517e-01 -7.33803332e-01 1.71870995e+00
-1.18587688e-01 6.93087280e-01 -1.16335250e-01 -1.19622052e+00
7.22483754e-01 6.18258059e-01 -4.62581158e-01 -8.13711584e-01
4.37008053e-01 1.98590323e-01 1.08981812e+00 -8.04688096e-01
1.18170038e-01 -4.95773226e-01 3.03755701e-01 6.50603890e-01
5.14303207e-01 -3.25392514e-01 2.86950141e-01 5.15989549e-02
5.95748603e-01 3.53303194e-01 6.20352328e-01 -8.17538798e-01
6.65845931e-01 -4.48350072e-01 4.16316152e-01 7.78769135e-01
-1.40853589e-02 -4.16822731e-03 5.97650588e-01 -5.07405818e-01
-7.38226354e-01 -1.06089997e+00 -2.30963960e-01 1.43807340e+00
-2.83861637e-01 -1.56577393e-01 -1.03953874e+00 -6.98266625e-02
-6.73295379e-01 1.08631468e+00 -5.98430276e-01 -7.55359158e-02
-7.37871885e-01 -6.42053962e-01 5.07206023e-01 3.72489154e-01
3.89784575e-01 -1.93801320e+00 -8.77363563e-01 3.35338444e-01
3.92473072e-01 -1.02296495e+00 2.45748311e-01 1.53025882e-02
-9.84635472e-01 -8.55908632e-01 -2.44048029e-01 -1.20648634e+00
5.37405908e-01 -6.70100898e-02 1.20012391e+00 6.61969125e-01
2.54169479e-02 3.44832093e-01 -4.97860700e-01 -1.04819727e+00
-6.08692348e-01 -1.18074350e-01 1.80907816e-01 -4.63086993e-01
8.70131671e-01 -8.56560588e-01 -1.53248325e-01 -2.73795128e-01
-9.19458270e-01 2.03346923e-01 6.84225857e-01 1.96265444e-01
8.16787109e-02 -1.07117563e-01 5.71218789e-01 -5.81373811e-01
9.18593824e-01 -3.49914163e-01 -7.06996620e-01 3.65783811e-01
-4.74223435e-01 2.33682573e-01 5.98193526e-01 -5.23827553e-01
-1.14022028e+00 -6.16747797e-01 -3.54664505e-01 5.34952939e-01
-7.98305154e-01 7.52460837e-01 -3.81033838e-01 -5.83649501e-02
4.38078552e-01 5.32281935e-01 -1.40929744e-01 -5.25022686e-01
1.56494692e-01 1.62328273e-01 1.13628559e-01 -1.31348026e+00
4.76303339e-01 -1.43054813e-01 2.59820987e-02 -1.15862215e+00
-7.73723304e-01 1.82772711e-01 -8.57387900e-01 1.35534094e-03
9.84893203e-01 -6.88555777e-01 -7.52095878e-01 5.24382532e-01
-1.51503813e+00 -4.72495407e-01 1.31709009e-01 9.74048316e-01
-4.94762063e-01 2.23773047e-01 -8.17622662e-01 -7.35326350e-01
-1.91696584e-01 -1.00201833e+00 2.83683807e-01 4.15499687e-01
-3.36069465e-01 -1.12911916e+00 1.37669295e-01 -1.36713922e-01
2.05745220e-01 -1.87906474e-01 1.74301982e+00 -1.01526403e+00
-4.95806992e-01 3.64113778e-01 -1.19916186e-01 1.71325728e-01
-3.44761163e-01 1.65844992e-01 -8.63694310e-01 2.31779620e-01
8.08815718e-01 -3.01992178e-01 4.70754474e-01 -1.78849455e-02
7.55826414e-01 -6.18872866e-02 -5.12313358e-02 3.44415635e-01
1.55177605e+00 5.75794756e-01 6.46968186e-01 1.47249941e-02
3.36818695e-01 1.26437581e+00 3.37299079e-01 -1.74473286e-01
5.67610681e-01 2.05946028e-01 -4.52195555e-02 4.26486224e-01
2.29742192e-02 -3.55695248e-01 4.06997442e-01 1.28317356e+00
-1.97304025e-01 -5.56793064e-02 -1.19668305e+00 6.52234077e-01
-1.38954484e+00 -8.42228413e-01 -2.06427589e-01 2.03191829e+00
1.07339644e+00 3.45795214e-01 -2.83112317e-01 1.45096049e-01
5.10320663e-01 -5.87245487e-02 1.09972008e-01 -1.06012321e+00
-8.81488845e-02 5.86042464e-01 -5.37274778e-01 4.70395863e-01
-3.34381938e-01 1.32418120e+00 6.32879066e+00 2.69220144e-01
-1.09091330e+00 2.98738908e-02 1.78444058e-01 3.67549628e-01
-2.65410274e-01 -4.41353433e-02 -8.37674201e-01 1.47837028e-01
1.09196985e+00 -9.09794569e-02 6.46999776e-01 4.80093628e-01
1.60317600e-01 -3.94793421e-01 -1.69772279e+00 6.21695876e-01
4.07542326e-02 -7.58955598e-01 3.64456087e-01 -1.07930772e-01
1.95572600e-01 -7.86488354e-02 -2.62174189e-01 4.32876348e-01
-9.40634236e-02 -1.24531114e+00 8.46849680e-01 5.33716857e-01
4.37751949e-01 -3.23946476e-01 4.14958656e-01 9.94657040e-01
-8.16775799e-01 1.98291600e-01 -7.33198285e-01 -8.69233251e-01
1.10286810e-01 2.60236323e-01 -5.06339967e-01 7.75964260e-02
5.72301924e-01 -2.16673762e-02 -5.90601921e-01 7.53685772e-01
-9.23517108e-01 8.54972363e-01 -3.59854639e-01 -6.52557552e-01
9.95141417e-02 -2.71898389e-01 1.56622484e-01 1.21139145e+00
2.09965423e-01 7.96683967e-01 -4.78614956e-01 1.25696516e+00
1.73567533e-01 7.01790929e-01 -8.41742933e-01 -2.79568195e-01
4.59446907e-01 1.03789485e+00 -1.09796798e+00 -2.12313011e-01
-7.32294500e-01 4.71035570e-01 6.84145987e-01 6.67460859e-02
-5.92823029e-01 9.22228917e-02 1.66354597e-01 2.38241330e-01
-6.12489954e-02 -6.85090244e-01 -3.87551963e-01 -1.09487998e+00
-1.22878119e-01 -7.11624146e-01 -1.88325047e-01 -7.86295891e-01
-9.56382751e-01 4.92215127e-01 4.44843173e-01 -1.63398787e-01
-2.48052590e-02 -1.15871811e+00 -7.27286458e-01 1.10965931e+00
-9.91743624e-01 -8.97220194e-01 -1.05026495e-02 6.02238715e-01
5.26303291e-01 -9.70986262e-02 1.05162990e+00 -7.94742256e-02
-4.22129631e-01 3.80412489e-01 -6.83754027e-01 1.70578510e-01
1.54908285e-01 -1.06030571e+00 4.56212014e-01 1.09474218e+00
5.67862689e-01 1.41472554e+00 7.13817894e-01 -6.77885354e-01
-1.13211167e+00 1.05831958e-02 1.46122849e+00 -4.86426681e-01
6.49372160e-01 -7.11160660e-01 -1.26478136e+00 7.45785177e-01
7.60008216e-01 -3.49429548e-01 7.49998927e-01 -1.32433046e-02
-1.64632797e-01 2.39760384e-01 -9.66360092e-01 6.89729035e-01
1.36222875e+00 -5.83928287e-01 -1.65903592e+00 1.28769398e-01
8.72780442e-01 7.87118152e-02 -7.00405717e-01 2.83314317e-01
5.00097513e-01 -1.12016630e+00 4.95565176e-01 -7.72208333e-01
4.08236951e-01 -1.12170130e-01 1.92214884e-02 -1.13327491e+00
-4.31841522e-01 -3.69215608e-01 3.53303403e-02 1.29996896e+00
4.64002907e-01 -9.65318620e-01 2.92533427e-01 7.04176843e-01
-2.41621301e-01 -6.77275062e-01 -6.97515309e-01 -3.98546308e-01
6.03137553e-01 -5.62630534e-01 4.47385073e-01 8.88194084e-01
1.60387054e-01 6.07128203e-01 4.60629255e-01 9.23978612e-02
3.54108959e-01 -5.57003990e-02 4.13490981e-01 -1.64056361e+00
-3.90162796e-01 -5.30082643e-01 1.00030385e-01 -8.41746330e-01
3.82775575e-01 -1.04199171e+00 -9.16861892e-02 -1.50323582e+00
1.11328280e-02 -2.49335870e-01 -1.64526597e-01 2.85580337e-01
-1.26508474e-02 -5.84190011e-01 2.98117518e-01 -2.25335941e-01
-1.29841924e-01 3.45686853e-01 1.11790907e+00 5.95110238e-01
-1.10884309e-01 -1.11071922e-01 -8.71004760e-01 1.29073405e+00
9.64286268e-01 -6.01906657e-01 -6.47998393e-01 -6.77619874e-01
7.52318084e-01 -8.50545019e-02 9.11383405e-02 -9.55068052e-01
1.73610076e-01 -2.08507538e-01 1.61975548e-01 -2.82529145e-01
-1.79586962e-01 -8.85053694e-01 -1.36764005e-01 5.40826261e-01
-1.78493872e-01 4.16787803e-01 3.89895916e-01 -1.61485821e-01
-4.59207632e-02 -7.76043653e-01 7.51244724e-01 -5.91832578e-01
-7.66107202e-01 -1.02312148e-01 -6.51354849e-01 1.13195091e-01
6.40630484e-01 -3.08711171e-01 -1.31996706e-01 1.55233949e-01
-7.06612706e-01 -2.56763458e-01 2.49063820e-01 4.50315416e-01
5.43166339e-01 -9.30614412e-01 -5.80746591e-01 2.24563286e-01
-9.38715115e-02 -2.18176603e-01 -1.51020497e-01 7.30250299e-01
-1.00800717e+00 7.49365985e-01 -4.37915534e-01 4.12962325e-02
-8.19589317e-01 1.06755340e+00 2.82147527e-01 -4.53303866e-02
-3.60209256e-01 1.06879985e+00 8.29102933e-01 -2.94921130e-01
7.81894624e-02 -8.53556573e-01 -6.40975475e-01 5.50456718e-02
6.46825612e-01 1.44526064e-01 -2.54989922e-01 -6.41685784e-01
-1.67622462e-01 7.00986207e-01 -1.11599140e-01 -1.79986164e-01
1.23579311e+00 2.46657636e-02 -8.28222573e-01 8.52152646e-01
6.74809754e-01 3.20149213e-01 -5.38812160e-01 -2.14688778e-01
3.22250456e-01 5.71479686e-02 -6.39250934e-01 -6.53011739e-01
-4.73712265e-01 1.32969189e+00 1.16222605e-01 3.40066731e-01
9.64347601e-01 1.74365580e-01 8.01905617e-02 5.25865197e-01
8.70778382e-01 -1.09590805e+00 -1.44378379e-01 8.78415644e-01
1.19359338e+00 -9.85873461e-01 -1.65553808e-01 -3.98946136e-01
-2.15584308e-01 1.15491140e+00 9.79241431e-01 -3.14108133e-01
8.63279641e-01 2.73454607e-01 -3.53480875e-01 -4.32352096e-01
-8.56243432e-01 -1.37887970e-01 2.12510988e-01 3.85753095e-01
1.06716144e+00 -8.49685445e-02 -1.14797699e+00 1.00373101e+00
-6.35139823e-01 -9.32276547e-02 4.71907228e-01 7.45237410e-01
-7.99395561e-01 -1.02049661e+00 -3.76151264e-01 1.73958436e-01
-7.54815936e-01 -6.07278585e-01 -4.74637955e-01 1.19591844e+00
5.25480926e-01 6.92599595e-01 -1.67590767e-01 9.81321465e-03
2.27867201e-01 3.91535759e-01 1.02151096e+00 -1.09126091e+00
-8.22047353e-01 -3.23767930e-01 -1.13079824e-01 -3.85360152e-01
-5.04512608e-01 -7.29232311e-01 -1.73838139e+00 -1.52821779e-01
-1.14142030e-01 2.23432258e-01 4.93472368e-01 1.33286357e+00
-3.07269424e-01 2.90328354e-01 -3.53432655e-01 -5.80242753e-01
-1.90490112e-01 -1.24759030e+00 -6.88147664e-01 1.58534572e-02
-1.58769473e-01 -5.20115614e-01 -3.24792653e-01 -1.18290931e-01]
|
[10.265765190124512, 8.811880111694336]
|
17f2ef62-959f-4185-849e-8bdd6ffdad66
|
temporal-superimposed-crossover-module-for
|
2211.03387
| null |
https://arxiv.org/abs/2211.03387v3
|
https://arxiv.org/pdf/2211.03387v3.pdf
|
Temporal superimposed crossover module for effective continuous sign language
|
The ultimate goal of continuous sign language recognition(CSLR) is to facilitate the communication between special people and normal people, which requires a certain degree of real-time and deploy-ability of the model. However, in the previous research on CSLR, little attention has been paid to the real-time and deploy-ability. In order to improve the real-time and deploy-ability of the model, this paper proposes a zero parameter, zero computation temporal superposition crossover module(TSCM), and combines it with 2D convolution to form a "TSCM+2D convolution" hybrid convolution, which enables 2D convolution to have strong spatial-temporal modelling capability with zero parameter increase and lower deployment cost compared with other spatial-temporal convolutions. The overall CSLR model based on TSCM is built on the improved ResBlockT network in this paper. The hybrid convolution of "TSCM+2D convolution" is applied to the ResBlock of the ResNet network to form the new ResBlockT, and random gradient stop and multi-level CTC loss are introduced to train the model, which reduces the final recognition WER while reducing the training memory usage, and extends the ResNet network from image classification task to video recognition task. In addition, this study is the first in CSLR to use only 2D convolution extraction of sign language video temporal-spatial features for end-to-end learning for recognition. Experiments on two large-scale continuous sign language datasets demonstrate the effectiveness of the proposed method and achieve highly competitive results.
|
['Quan Gan', 'Fei Yuan', 'Jing Li', 'Qidan Zhu']
|
2022-11-07
| null | null | null | null |
['sign-language-recognition', 'video-recognition']
|
['computer-vision', 'computer-vision']
|
[ 1.31322041e-01 -5.08224070e-01 1.23905338e-01 -3.03255558e-01
-5.17701328e-01 -1.51840985e-01 3.78181368e-01 -1.15051806e+00
-8.65776718e-01 2.81212121e-01 1.70227766e-01 -4.87059176e-01
1.11886352e-01 -5.39464056e-01 -4.45846766e-01 -7.31300235e-01
1.53769627e-01 -2.70313680e-01 5.31798899e-01 -9.48598310e-02
-4.82004024e-02 5.54714501e-01 -1.49535310e+00 3.09406102e-01
9.21004057e-01 1.10270274e+00 5.18218994e-01 6.51165545e-01
-3.09612960e-01 9.00319576e-01 -5.25717854e-01 -2.25145310e-01
4.03263390e-01 -3.69183421e-01 -2.99020231e-01 -1.22206777e-01
5.73290348e-01 -6.02114975e-01 -8.15030098e-01 8.49976063e-01
1.09099793e+00 2.63725907e-01 2.99232841e-01 -1.08993602e+00
-6.41512394e-01 3.10859770e-01 -3.24967176e-01 6.24467246e-02
-2.23395433e-02 4.74350750e-01 5.27886987e-01 -7.44699121e-01
2.47981265e-01 1.37259448e+00 7.03478694e-01 8.48977149e-01
-5.56412220e-01 -1.05103791e+00 3.86569291e-01 4.40816611e-01
-1.38100147e+00 -3.52723032e-01 5.75893998e-01 -1.80191174e-01
1.22879767e+00 2.12999165e-01 8.32672536e-01 8.65733683e-01
-2.57743686e-01 1.14485919e+00 1.18207777e+00 -4.48634475e-01
-1.40172586e-01 -2.13704064e-01 2.52333760e-01 8.48028779e-01
-2.70698190e-01 2.71598756e-01 -2.25562111e-01 4.83531833e-01
9.56465960e-01 1.25445649e-01 -3.06352556e-01 2.80335903e-01
-8.91679943e-01 3.89593124e-01 6.32894635e-01 4.70753878e-01
-1.94285929e-01 3.84311616e-01 5.25198042e-01 4.09952015e-01
1.29757240e-01 -2.66224176e-01 -3.58288467e-01 -4.54476237e-01
-7.74070084e-01 -2.62090683e-01 3.57685864e-01 9.36457396e-01
7.17365816e-02 3.51813674e-01 -2.92471647e-01 1.13099778e+00
2.04058304e-01 8.18958819e-01 9.06796157e-01 -5.03338993e-01
6.07821763e-01 6.81767821e-01 -4.82815862e-01 -4.33899164e-01
-3.16713393e-01 -5.37732780e-01 -9.74760175e-01 3.72505069e-01
8.82337615e-02 -7.83889666e-02 -1.48033285e+00 1.66884506e+00
-8.15613344e-02 5.07068634e-01 6.57423362e-02 1.21006823e+00
1.11887681e+00 5.78648865e-01 1.67309552e-01 1.53107103e-02
1.16482902e+00 -1.18411434e+00 -5.29398322e-01 1.03421278e-01
7.76995420e-01 -6.79034650e-01 1.25634134e+00 3.47136617e-01
-8.80044878e-01 -8.80879164e-01 -8.34350765e-01 -2.51491427e-01
-3.15178931e-01 6.82536781e-01 6.39232218e-01 7.49191761e-01
-9.33981299e-01 1.03814684e-01 -7.85070777e-01 -2.96744555e-01
3.29437613e-01 5.37285149e-01 -2.48291478e-01 -3.14752847e-01
-1.09411895e+00 9.96647120e-01 2.77380913e-01 6.67998970e-01
-3.97412807e-01 -3.27560067e-01 -6.07995391e-01 -1.64530829e-01
1.75689295e-01 -4.53099489e-01 1.16123712e+00 -9.97170746e-01
-1.86268628e+00 5.26270270e-01 -2.11064890e-01 -2.99564868e-01
8.13519359e-01 -1.21943198e-01 -6.43042326e-01 -9.50124338e-02
-4.32110906e-01 6.79805756e-01 7.36359060e-01 -7.11972535e-01
-8.52437377e-01 -5.24312668e-02 9.97971296e-02 2.42778882e-01
-2.95113355e-01 3.06760520e-01 -7.29336798e-01 -7.69892216e-01
2.48259902e-01 -8.66993546e-01 5.17010652e-02 -1.31289416e-04
4.37610447e-02 -3.77273589e-01 1.23973906e+00 -9.06628728e-01
1.25277174e+00 -2.31462789e+00 -1.91689000e-01 1.70294791e-01
-1.29095331e-01 9.72259343e-01 -5.01743019e-01 -1.05053417e-01
3.94505598e-02 -1.20966107e-01 -1.95350125e-01 -3.13674808e-01
6.12065569e-03 3.12557638e-01 -2.64584392e-01 2.75856614e-01
2.97703892e-02 1.06402624e+00 -5.57020664e-01 -3.55294615e-01
5.49467564e-01 6.41855955e-01 -4.51541543e-01 1.36458322e-01
1.29811898e-01 3.09905142e-01 -4.19719577e-01 6.45995140e-01
8.52446258e-01 4.39861305e-02 -1.79801747e-01 -2.73948669e-01
-3.43796700e-01 2.82907277e-01 -1.07420719e+00 1.35449684e+00
-7.45369375e-01 6.95103049e-01 -7.90583715e-02 -8.68798852e-01
9.16283786e-01 2.98490107e-01 3.10575336e-01 -1.14723063e+00
3.01303297e-01 3.23710173e-01 4.95683439e-02 -7.65698731e-01
1.40841424e-01 1.22532226e-01 1.31641030e-01 2.35235766e-02
-1.89335540e-01 1.55628100e-01 -2.52480544e-02 -1.95856839e-01
9.91909504e-01 3.66388291e-01 -1.32212266e-01 2.29850456e-01
7.81810820e-01 -4.12782937e-01 5.18472433e-01 5.93729079e-01
-1.20916709e-01 4.15534824e-01 1.32206008e-02 -1.96700215e-01
-8.73487890e-01 -9.28537309e-01 -6.03306689e-05 8.12976122e-01
1.54359654e-01 -1.95708312e-02 -5.54061294e-01 -8.26267958e-01
-2.06132635e-01 4.72910523e-01 -1.15071304e-01 -9.36934948e-02
-1.06223285e+00 -5.25476515e-01 9.53303754e-01 7.12508380e-01
1.50979924e+00 -1.12448728e+00 -5.16529799e-01 1.07103691e-01
-2.16543928e-01 -1.20715535e+00 -9.21682417e-01 -2.25690931e-01
-6.06039882e-01 -8.89238834e-01 -9.23508883e-01 -1.30867887e+00
6.48821592e-01 3.95512730e-01 1.51575372e-01 2.25496963e-01
-3.34499031e-01 2.45538682e-01 -6.45608068e-01 -2.54971206e-01
3.55227552e-02 -1.73756987e-01 -6.80896565e-02 1.17320389e-01
4.49317396e-01 -5.25812685e-01 -4.97767299e-01 4.39025581e-01
-7.92137086e-01 2.24378988e-01 9.20449853e-01 1.18213940e+00
1.92486539e-01 1.49772778e-01 2.63030708e-01 2.25268025e-03
6.43405378e-01 2.96797931e-01 -6.02614343e-01 4.28028524e-01
-3.91055584e-01 -1.13350898e-01 7.33066678e-01 -9.50993299e-01
-1.11479342e+00 6.93687052e-02 -5.02216339e-01 -5.67852378e-01
3.16853411e-02 3.68461162e-01 -2.99692988e-01 -4.96190608e-01
1.27763689e-01 8.22533906e-01 1.83854625e-01 -6.13956571e-01
3.07650328e-01 1.11044705e+00 6.36572659e-01 -1.37156948e-01
6.52825534e-01 2.13219136e-01 -1.33317962e-01 -1.08608961e+00
-3.59961390e-01 -3.83705735e-01 -4.20708776e-01 -3.06320250e-01
8.26689661e-01 -8.95267487e-01 -1.00870347e+00 1.16923726e+00
-1.03032207e+00 -5.25336802e-01 -1.46354005e-01 1.05836844e+00
-2.79073715e-01 5.57277381e-01 -5.16520321e-01 -9.45417643e-01
-3.88930351e-01 -1.33519340e+00 8.53018105e-01 3.01350445e-01
4.89806205e-01 -5.42344332e-01 -5.49321532e-01 3.44151288e-01
7.01491058e-01 -3.82890135e-01 5.62158227e-01 -2.05931157e-01
-8.89839053e-01 -2.49882057e-01 -6.72148883e-01 1.00764263e+00
-1.43173262e-02 -4.09174621e-01 -9.46601748e-01 -3.93132180e-01
-2.25858629e-01 -2.54309148e-01 1.20317721e+00 5.22951603e-01
1.17810130e+00 -2.60086805e-01 -9.48098972e-02 8.99884999e-01
1.18986619e+00 6.23750329e-01 1.09150219e+00 1.63405359e-01
6.96415246e-01 1.91537187e-01 4.58161920e-01 -4.60142717e-02
3.87708485e-01 9.48329031e-01 2.83769239e-02 -4.35948491e-01
-7.48078823e-01 -2.78728545e-01 6.39852762e-01 1.05138886e+00
-3.15681517e-01 -1.30413607e-01 -4.76050735e-01 2.42330149e-01
-1.78202963e+00 -1.13929081e+00 -2.07923036e-02 2.10119939e+00
5.12617409e-01 1.10093979e-02 -5.44128045e-02 8.46305937e-02
6.97873533e-01 8.24406147e-02 -4.52139974e-01 -1.69994473e-01
-2.91148812e-01 2.86112875e-01 6.81312442e-01 5.55439115e-01
-1.02484465e+00 1.17911077e+00 5.39768600e+00 1.32918131e+00
-1.76361144e+00 2.40513667e-01 5.72439432e-02 -1.56072870e-01
2.88953811e-01 -1.08584039e-01 -1.00328803e+00 5.91081023e-01
5.17783701e-01 4.86212254e-01 5.91751933e-01 7.46144891e-01
4.77083027e-01 1.11522801e-01 -6.30997300e-01 1.33528590e+00
1.98195100e-01 -8.34644556e-01 -2.01187376e-03 9.05517209e-03
2.69224107e-01 1.32841855e-01 -4.94468138e-02 6.77911401e-01
1.33741811e-01 -9.36504543e-01 6.78706169e-01 5.94581842e-01
1.31627333e+00 -5.20158529e-01 8.71385753e-01 3.37823212e-01
-1.67188501e+00 -4.09965754e-01 -2.07097307e-01 -5.84772043e-02
4.17258173e-01 1.90411463e-01 -5.83072126e-01 2.29822069e-01
6.34730995e-01 6.03835285e-01 -3.46181214e-01 1.27787244e+00
-3.97610068e-01 5.98552704e-01 -5.57826579e-01 -4.65317011e-01
4.35985744e-01 -1.74317420e-01 4.94597167e-01 1.42184150e+00
4.25330967e-01 3.53325993e-01 1.68605193e-01 4.19201851e-01
1.03400595e-01 -1.62118450e-02 -2.77060777e-01 1.50059268e-01
2.67927349e-01 7.66031742e-01 -2.65590400e-01 -4.30225998e-01
-4.19486612e-01 1.14248574e+00 5.11103217e-03 4.43203330e-01
-8.49341214e-01 -6.87277913e-01 4.33889925e-01 -1.79442942e-01
4.69522983e-01 -4.58584547e-01 -1.29154578e-01 -1.21177876e+00
5.14976501e-01 -7.62244105e-01 1.96322158e-01 -8.11862946e-01
-1.17804348e+00 6.67880893e-01 -2.44795352e-01 -1.47915626e+00
5.08276969e-02 -9.35000598e-01 -5.25969267e-01 1.20989299e+00
-1.78847456e+00 -1.71833932e+00 -4.87679869e-01 8.98848236e-01
5.79385221e-01 -5.30182004e-01 5.42174935e-01 6.53161407e-01
-5.25452912e-01 1.10643017e+00 -2.44762320e-02 4.71954644e-01
3.97438973e-01 -5.86818755e-01 3.40702683e-01 1.08768785e+00
-3.08874547e-01 4.95067656e-01 2.80324500e-02 -6.47140384e-01
-1.24324775e+00 -1.22175682e+00 1.02608550e+00 1.48566365e-01
2.66667128e-01 -2.88658649e-01 -5.56507289e-01 3.09174508e-01
-3.03520650e-01 -1.05402768e-02 9.41353366e-02 -3.34557176e-01
-4.54865694e-01 -3.89858752e-01 -1.10947704e+00 8.72350872e-01
1.54851055e+00 -6.56555057e-01 -5.59958756e-01 1.69547722e-01
7.72602379e-01 -3.71393055e-01 -4.39251721e-01 5.78348339e-01
9.61751461e-01 -5.79423189e-01 9.34829116e-01 -3.57666582e-01
-4.58118059e-02 -5.05790830e-01 -2.31731310e-01 -8.88585627e-01
-1.63111120e-01 -3.11779976e-01 1.67792171e-01 9.89277124e-01
2.60518312e-01 -1.05506277e+00 6.22334599e-01 4.35920656e-01
-4.94235933e-01 -7.04507232e-01 -1.29956484e+00 -1.12227917e+00
-2.54073590e-01 -7.72865236e-01 4.49922234e-01 5.27652621e-01
-2.69949675e-01 -1.31960530e-02 -4.59514052e-01 7.75654912e-02
3.16646397e-01 -2.03092635e-01 7.60699213e-01 -7.17861772e-01
-3.54637295e-01 -6.92050457e-01 -6.30461633e-01 -1.87393391e+00
1.45109079e-03 -8.98953319e-01 -9.17793077e-04 -1.55125010e+00
-1.13476664e-01 -6.36859417e-01 -1.71084821e-01 7.23477721e-01
1.34498298e-01 2.78538316e-01 4.90057170e-01 2.34905764e-01
-2.74468780e-01 6.20241582e-01 1.42766416e+00 -2.78171897e-01
-4.24724013e-01 2.03325942e-01 -1.13375582e-01 5.57789862e-01
4.26490635e-01 -1.28880758e-02 -3.87433112e-01 -5.68523109e-01
-5.60473323e-01 -1.59635037e-01 6.67489111e-01 -1.11460948e+00
4.83775973e-01 3.60209011e-02 9.18850228e-02 -7.33130991e-01
6.35472655e-01 -9.11038280e-01 -2.26299658e-01 6.32483423e-01
-1.37822926e-01 -1.81110010e-01 1.64112881e-01 1.31680056e-01
-2.90612638e-01 8.72323569e-03 8.70345354e-01 -1.01117648e-01
-1.03455961e+00 3.95941466e-01 -1.41451538e-01 -4.01389331e-01
8.47156048e-01 -6.63202524e-01 -2.62605876e-01 -2.81475842e-01
-5.84358215e-01 2.02173308e-01 -1.45777628e-01 5.89048624e-01
9.71161366e-01 -1.41380608e+00 -5.82657576e-01 5.31844497e-01
-2.55598068e-01 -1.54543087e-01 5.79376459e-01 9.53991234e-01
-6.35527015e-01 6.76690042e-01 -5.85724786e-02 -4.76510733e-01
-1.54979849e+00 1.55923590e-01 6.59551382e-01 -8.71148854e-02
-9.77759898e-01 9.47061539e-01 -2.16802880e-02 -4.85482395e-01
7.03164876e-01 -3.79477262e-01 -1.67658716e-01 -3.26288134e-01
5.62169552e-01 3.68260205e-01 -6.14999719e-02 -6.34263098e-01
-2.46438667e-01 9.73350286e-01 -5.71169443e-02 -2.79133141e-01
1.11566281e+00 -8.09907913e-04 1.60352588e-02 -2.07818169e-02
1.23972654e+00 -2.72791713e-01 -1.20904624e+00 -3.57804507e-01
-4.99786526e-01 -4.92070735e-01 2.69170612e-01 -1.14004147e+00
-1.23140383e+00 1.03021002e+00 1.06969428e+00 -4.45437372e-01
1.58040476e+00 -4.59331930e-01 1.15911269e+00 4.26284939e-01
4.29386258e-01 -1.17303550e+00 -1.11387253e-01 7.61086762e-01
9.70263362e-01 -9.06129181e-01 -4.26623434e-01 -2.06001639e-01
-5.07443249e-01 1.07525039e+00 8.21353614e-01 -1.08327262e-01
7.06742346e-01 3.81661057e-01 2.18367428e-01 2.87047744e-01
-2.86131263e-01 -4.90456223e-01 3.57924908e-01 5.63365400e-01
1.04445226e-01 2.21918877e-02 -7.05575049e-01 5.79330981e-01
-2.67279804e-01 4.80185270e-01 -9.85335279e-03 8.45449924e-01
-2.72477925e-01 -1.14421105e+00 -1.11180112e-01 4.15722698e-01
-5.84526807e-02 -2.55040973e-01 3.81794609e-02 6.42292023e-01
5.41416824e-01 8.45966935e-01 8.52914003e-04 -7.97245204e-01
6.46391571e-01 4.75875242e-03 3.77879292e-01 -3.90981175e-02
-4.82241124e-01 2.06476420e-01 1.35394722e-01 -4.40681159e-01
-3.66454631e-01 -4.06623125e-01 -1.39859617e+00 -2.63973355e-01
-5.17826676e-01 -2.43836388e-01 8.93923283e-01 1.08879042e+00
2.31882200e-01 6.17039323e-01 5.07364511e-01 -6.02201283e-01
-7.28882372e-01 -1.13991165e+00 -4.63296890e-01 9.17427912e-02
2.87347376e-01 -4.86422300e-01 -2.03849509e-01 -1.51774377e-01]
|
[9.246593475341797, -6.5054731369018555]
|
21d825be-c76a-4382-9bed-e26716a881b7
|
unsupervised-domain-agnostic-fake-news
|
2305.11349
| null |
https://arxiv.org/abs/2305.11349v1
|
https://arxiv.org/pdf/2305.11349v1.pdf
|
Unsupervised Domain-agnostic Fake News Detection using Multi-modal Weak Signals
|
The emergence of social media as one of the main platforms for people to access news has enabled the wide dissemination of fake news. This has motivated numerous studies on automating fake news detection. Although there have been limited attempts at unsupervised fake news detection, their performance suffers due to not exploiting the knowledge from various modalities related to news records and due to the presence of various latent biases in the existing news datasets. To address these limitations, this work proposes an effective framework for unsupervised fake news detection, which first embeds the knowledge available in four modalities in news records and then proposes a novel noise-robust self-supervised learning technique to identify the veracity of news records from the multi-modal embeddings. Also, we propose a novel technique to construct news datasets minimizing the latent biases in existing news datasets. Following the proposed approach for dataset construction, we produce a Large-scale Unlabelled News Dataset consisting 419,351 news articles related to COVID-19, acronymed as LUND-COVID. We trained the proposed unsupervised framework using LUND-COVID to exploit the potential of large datasets, and evaluate it using a set of existing labelled datasets. Our results show that the proposed unsupervised framework largely outperforms existing unsupervised baselines for different tasks such as multi-modal fake news detection, fake news early detection and few-shot fake news detection, while yielding notable improvements for unseen domains during training.
|
['Christopher Leckie', 'Shanika Karunasekera', 'Ling Luo', 'Amila Silva']
|
2023-05-18
| null | null | null | null |
['fake-news-detection']
|
['natural-language-processing']
|
[-6.87743947e-02 3.15631106e-02 -5.50586998e-01 -8.72645676e-02
-8.02040100e-01 -4.45270211e-01 1.14496529e+00 3.33748907e-01
-4.36211169e-01 6.44445240e-01 7.70424545e-01 1.26151368e-01
2.30784789e-01 -7.86039412e-01 -7.19494939e-01 -3.37668538e-01
3.88553113e-01 1.92509204e-01 1.99012488e-01 -3.82428229e-01
6.42072916e-01 6.59082159e-02 -1.54762983e+00 5.40646434e-01
7.04622805e-01 6.87357485e-01 -4.84564155e-01 2.58768439e-01
-4.50891927e-02 1.00079405e+00 -8.94226491e-01 -7.46046305e-01
5.71916252e-02 -6.25600934e-01 -5.49834132e-01 7.84264430e-02
3.99664730e-01 -4.10458922e-01 -7.34375834e-01 1.19161880e+00
4.30417508e-01 -6.56859055e-02 7.32379675e-01 -1.27794707e+00
-1.04491580e+00 6.39908791e-01 -4.27182078e-01 4.34124678e-01
3.58382881e-01 -2.81493723e-01 8.06377172e-01 -9.23831403e-01
9.98257756e-01 1.05517936e+00 7.74368048e-01 4.19450641e-01
-8.88935745e-01 -4.03056741e-01 -3.63554716e-01 -2.44213431e-03
-1.14468336e+00 -6.78778648e-01 8.80492508e-01 -6.12146378e-01
7.02794492e-01 1.86915949e-01 4.77408350e-01 1.75688684e+00
1.87361136e-01 8.57065618e-01 1.47130144e+00 -4.96524543e-01
1.50206238e-01 7.23939598e-01 3.96326780e-01 6.37171030e-01
6.12772703e-01 2.42300957e-01 -7.53714561e-01 -8.09409380e-01
2.97982424e-01 1.89334124e-01 -3.54845002e-02 -2.21121833e-02
-1.34611070e+00 1.19997549e+00 1.85458139e-01 3.87743115e-01
-2.34319106e-01 -1.18973628e-01 8.72434318e-01 6.27423465e-01
1.03070414e+00 7.05234110e-01 -2.35487223e-01 -9.53364447e-02
-8.93777609e-01 3.45561773e-01 8.36095095e-01 6.66529417e-01
3.84346157e-01 -2.11972818e-02 -8.00658762e-02 9.17544484e-01
3.24326098e-01 6.50540590e-01 9.32453275e-01 -2.45438278e-01
7.11684346e-01 6.32281899e-01 5.60816467e-01 -1.85827744e+00
-1.70284927e-01 -3.16306949e-01 -6.05233908e-01 -4.18283939e-01
2.20969424e-01 1.36710540e-03 -7.92820692e-01 1.21589279e+00
4.14990962e-01 3.03213745e-01 3.42563391e-01 9.00719643e-01
1.00917184e+00 7.12460518e-01 -1.57709286e-01 -2.95447886e-01
1.34777236e+00 -1.04408371e+00 -1.10842323e+00 -2.08922461e-01
8.12622786e-01 -9.71780062e-01 8.28048944e-01 1.97032526e-01
-1.98866069e-01 -1.01718605e-02 -9.64972258e-01 5.64045366e-03
-9.02274549e-01 2.55926877e-01 5.31513572e-01 8.63485634e-01
-2.70140529e-01 1.82675377e-01 -4.44002777e-01 -3.61631751e-01
6.05328023e-01 -3.14907193e-01 -5.07757902e-01 -1.12695664e-01
-1.45426857e+00 9.81255114e-01 6.19521677e-01 -1.74012318e-01
-9.70622778e-01 -4.77526709e-02 -6.12106621e-01 -5.07641792e-01
4.87178057e-01 -2.31709585e-01 6.79711878e-01 -1.10171711e+00
-1.36290085e+00 9.49856460e-01 1.93675920e-01 -5.53067327e-01
6.95384741e-01 -4.22605008e-01 -1.09433460e+00 1.82041615e-01
4.03761178e-01 6.69411942e-02 1.42841041e+00 -1.22699904e+00
-3.11410159e-01 -2.39271343e-01 -2.49827385e-01 -2.90673710e-02
-7.78814614e-01 2.64559686e-01 1.04737943e-02 -1.23484433e+00
1.17487684e-01 -9.22843277e-01 2.77092665e-01 -4.56475258e-01
-6.65104628e-01 6.86398223e-02 1.22338736e+00 -7.41858780e-01
1.13530982e+00 -2.16814899e+00 -2.42659360e-01 -8.41974020e-02
2.94975370e-01 3.93333584e-01 1.14272170e-01 5.99468768e-01
3.31106961e-01 -2.02423427e-02 -8.70169103e-02 -4.26225394e-01
-1.33454323e-01 1.92805156e-01 -7.73742080e-01 9.71171081e-01
-9.50821489e-02 7.21237898e-01 -1.15602446e+00 -3.77576679e-01
-8.90874565e-02 3.02207112e-01 -4.07186180e-01 -2.76210234e-02
-1.70235872e-01 3.53620380e-01 -4.91954386e-01 8.46175909e-01
5.70645034e-01 -1.94996864e-01 -1.82266533e-01 -1.88554153e-02
8.43346119e-02 3.87025833e-01 -8.38035703e-01 1.10246241e+00
-3.31258029e-02 7.15084076e-01 -4.95027483e-01 -1.20751822e+00
8.25440884e-01 5.14402628e-01 2.17619911e-01 -4.85400558e-01
4.49195325e-01 5.34668744e-01 -6.08196199e-01 -7.72843242e-01
7.89470434e-01 -2.46158063e-01 -5.05064845e-01 6.97475672e-01
2.45361015e-01 2.03949273e-01 -2.60322928e-01 4.33464736e-01
1.00484705e+00 -2.74520248e-01 3.00326943e-01 1.47299662e-01
2.87394643e-01 5.25383234e-01 2.11818159e-01 1.08813894e+00
-3.37518096e-01 2.45685309e-01 3.40410531e-01 -5.41998327e-01
-1.11612201e+00 -5.35998762e-01 -2.15668023e-01 1.01237345e+00
3.79374027e-01 -4.04884934e-01 -3.18625033e-01 -1.17288017e+00
1.38457909e-01 6.43613517e-01 -8.26279521e-01 -2.27233469e-01
-1.64703399e-01 -1.12579310e+00 1.11078405e+00 -3.20187032e-01
6.10056043e-01 -9.20886993e-01 -8.65760297e-02 2.79218495e-01
-4.60066319e-01 -1.25579238e+00 -1.71172619e-01 -2.37981737e-01
-7.00932980e-01 -1.02697563e+00 -3.14671189e-01 -6.34184182e-01
6.77056074e-01 5.78347743e-01 6.70801401e-01 3.77278067e-02
-5.28528169e-02 2.19196260e-01 -7.58962393e-01 -2.76257485e-01
-8.06081533e-01 -1.83278859e-01 4.46054518e-01 2.03542769e-01
6.35596752e-01 6.01279736e-02 -1.34716317e-01 3.26130241e-01
-1.14290154e+00 -1.47942916e-01 4.33133990e-01 1.17877328e+00
2.19295546e-01 1.34093598e-01 1.03257978e+00 -1.33463478e+00
8.21127057e-01 -1.13265145e+00 -3.66222501e-01 -6.53956756e-02
-6.76627994e-01 -2.10476056e-01 6.63396597e-01 -5.96180201e-01
-8.72575521e-01 -4.53753471e-01 1.73761010e-01 -1.13592833e-01
-1.63635552e-01 7.43000269e-01 4.37365860e-01 -6.48629516e-02
9.87379491e-01 3.31836045e-01 -4.44939733e-03 -4.45663184e-01
5.96709907e-01 1.22802210e+00 2.63887197e-01 -1.11239485e-01
9.24239695e-01 7.71005630e-01 -7.34566391e-01 -8.87054324e-01
-1.37892711e+00 -7.19379723e-01 -1.80623144e-01 1.61836222e-02
5.90896010e-01 -1.16900587e+00 9.68045667e-02 7.92890728e-01
-1.17177856e+00 5.37478447e-01 -1.12105692e-02 6.91188574e-01
-1.38357952e-01 5.99575341e-01 -7.76630700e-01 -8.03545713e-01
-1.40946507e-01 -7.80792117e-01 8.41095209e-01 -4.20771599e-01
1.46919303e-02 -9.32137728e-01 3.39021206e-01 8.16577256e-01
4.04863566e-01 4.51036125e-01 3.92582953e-01 -1.29489851e+00
-1.73210323e-01 -8.60497177e-01 -2.02201098e-01 4.97194827e-01
2.46327505e-01 -3.13161105e-01 -1.00824118e+00 -3.05731416e-01
2.68854409e-01 -7.41723776e-01 9.44448590e-01 -1.98210686e-01
4.56449866e-01 -9.73316848e-01 -3.40193629e-01 8.66551548e-02
1.32667744e+00 -4.21455979e-01 4.11922097e-01 7.62961924e-01
8.86610508e-01 4.73723948e-01 5.61435997e-01 5.38597226e-01
2.77214617e-01 5.63859403e-01 1.74528942e-01 2.98590302e-01
1.91327274e-01 -3.86677980e-01 6.65123701e-01 1.21953559e+00
3.15511584e-01 -4.51451838e-01 -6.50293708e-01 7.70679295e-01
-1.79759681e+00 -1.29949152e+00 -2.70791411e-01 1.95374131e+00
7.43063033e-01 2.25690097e-01 1.78735599e-01 8.76732767e-02
8.80626917e-01 3.94731998e-01 -9.51017532e-03 -2.07904473e-01
-4.58157539e-01 -3.82471710e-01 7.31556654e-01 3.31399173e-01
-1.54388714e+00 1.10769486e+00 6.31558466e+00 8.91202509e-01
-1.10966098e+00 7.47229278e-01 1.63469329e-01 1.03199691e-01
-2.71542668e-01 -7.07047526e-03 -6.32237673e-01 8.89659464e-01
8.71185601e-01 5.73382638e-02 2.64279932e-01 9.94192004e-01
2.91444242e-01 1.52204884e-02 -4.40735042e-01 8.09738338e-01
7.39879906e-01 -1.56599498e+00 1.21224798e-01 -2.66938843e-02
1.09873915e+00 2.94807971e-01 3.14325243e-02 3.66498321e-01
1.48104846e-01 -6.80428624e-01 7.39528656e-01 2.42029965e-01
5.03615797e-01 -5.16865849e-01 1.06500709e+00 7.57898331e-01
-1.59878865e-01 -1.06870145e-01 -4.36809391e-01 -1.29321128e-01
7.56806433e-02 8.86934578e-01 -9.28274989e-01 3.41658801e-01
3.49016815e-01 9.51351047e-01 -6.03494942e-01 7.91249990e-01
-1.34726390e-01 8.11836958e-01 -3.37809682e-01 -1.52168140e-01
2.36850575e-01 1.44174844e-01 8.51372361e-01 1.29111505e+00
1.48898154e-01 -2.26505816e-01 7.01833703e-03 4.88719612e-01
-3.91678780e-01 2.17934325e-01 -9.60122406e-01 -6.56901121e-01
5.02178848e-01 8.53140950e-01 -7.19194710e-01 -7.71258354e-01
-5.50612211e-01 1.16287398e+00 2.10806489e-01 5.89024760e-02
-9.33815658e-01 -2.56836057e-01 1.22374736e-01 1.21244803e-01
1.30413026e-01 -4.35135216e-02 -1.86140966e-02 -1.81692755e+00
-2.70590875e-02 -1.11387408e+00 4.41803932e-01 -3.18496168e-01
-1.78916597e+00 6.27391100e-01 -4.88301739e-02 -1.51396918e+00
6.92177713e-02 -2.28356704e-01 -6.84450492e-02 1.34999678e-01
-1.46487474e+00 -1.14484715e+00 -1.50129676e-01 5.57244062e-01
3.72098744e-01 -6.47619367e-01 8.41076195e-01 4.12993282e-01
-4.90330219e-01 4.83827531e-01 6.78236127e-01 2.48172760e-01
1.11262810e+00 -7.40797162e-01 2.86393702e-01 9.42393422e-01
3.68343204e-01 6.55466020e-01 9.49425459e-01 -1.07305038e+00
-1.31763601e+00 -1.04040611e+00 1.04413474e+00 -7.85250485e-01
1.07556784e+00 -3.78627896e-01 -8.38952720e-01 6.47781909e-01
-1.34355277e-01 2.97588050e-01 7.83021331e-01 -3.53963315e-01
-7.77444601e-01 4.29572970e-01 -1.50908232e+00 1.97977126e-01
6.51667535e-01 -8.79396975e-01 -1.13538110e+00 8.24908674e-01
6.24518812e-01 -3.30834448e-01 -5.33702850e-01 -7.61332214e-02
3.95515591e-01 -7.46379912e-01 7.45671690e-01 -7.98936605e-01
6.62860990e-01 -3.27704012e-01 3.84941250e-02 -1.27957428e+00
-1.81411244e-02 -3.91984820e-01 -4.71854538e-01 1.10978389e+00
2.23019332e-01 -9.39320862e-01 6.14001155e-01 -7.13560060e-02
9.77508724e-02 -1.74542293e-01 -9.91477430e-01 -7.09735036e-01
-3.88494670e-01 -3.06996405e-01 1.04040891e-01 1.84288406e+00
7.82583654e-02 1.66335449e-01 -1.20129919e+00 4.45311189e-01
7.10582316e-01 4.62843925e-02 9.36977804e-01 -1.01266372e+00
-2.39350557e-01 9.09239575e-02 -6.44632578e-01 -9.98583078e-01
6.33798689e-02 -7.41622448e-01 -1.06634468e-01 -1.08285928e+00
1.92232683e-01 -4.14240748e-01 -6.61804006e-02 2.54780620e-01
-1.19719230e-01 7.10408747e-01 -2.94400126e-01 9.13571477e-01
-6.81991518e-01 5.59535325e-01 9.57316697e-01 -1.34513006e-01
-7.83874094e-02 -2.52495259e-01 -5.41362405e-01 9.12633657e-01
4.99988616e-01 -1.17677450e+00 -2.58854851e-02 -2.87619740e-01
3.48067433e-01 -9.36357081e-02 6.35262787e-01 -7.06054747e-01
1.10560507e-01 -1.24350816e-01 8.60048756e-02 -4.94251668e-01
1.84180170e-01 -6.74277306e-01 -1.30768582e-01 3.27743202e-01
-3.64292622e-01 -1.63982779e-01 -1.85315967e-01 1.12951577e+00
-4.39078450e-01 -1.81212783e-01 7.15149164e-01 -1.24954373e-01
-5.83448172e-01 -1.76391706e-01 -6.65113449e-01 3.18778127e-01
9.47749972e-01 -8.89411345e-02 -9.85682368e-01 -2.90682167e-01
-5.60144722e-01 -3.47794652e-01 5.27047038e-01 8.24720919e-01
6.15171313e-01 -1.43954229e+00 -7.35351264e-01 1.17330588e-01
6.57184303e-01 -6.25413597e-01 1.91050041e-02 8.96079898e-01
-6.05253756e-01 1.32706434e-01 -1.02299057e-01 -1.58718139e-01
-9.19807196e-01 7.23862588e-01 -1.67377874e-01 -8.41002166e-02
-7.12859273e-01 6.62238002e-01 -5.99954605e-01 -4.33772117e-01
-9.69447047e-02 1.77069366e-01 -3.00518841e-01 2.61505574e-01
5.78485608e-01 3.09936881e-01 2.89217029e-02 -1.30137658e+00
-3.15161645e-01 -2.87632681e-02 -2.96324700e-01 1.23002296e-02
1.44939208e+00 -2.22469389e-01 -2.83573568e-01 5.26697278e-01
1.24729097e+00 6.01754308e-01 -4.07801419e-01 -5.20140946e-01
2.36167163e-01 -9.34107482e-01 2.57028371e-01 -7.52152681e-01
-5.96637428e-01 6.90301955e-02 4.35986221e-01 5.04533648e-01
3.46606642e-01 -6.11322969e-02 1.11231565e+00 4.39704686e-01
4.53407079e-01 -1.24834275e+00 4.05569315e-01 5.81250310e-01
5.36306500e-01 -1.68204689e+00 2.14794263e-01 -3.32076669e-01
-7.86730766e-01 8.67079437e-01 2.29148328e-01 -4.09908175e-01
4.95814055e-01 -2.09819928e-01 2.46612549e-01 -6.90957785e-01
-2.73339808e-01 2.28147373e-01 2.33125374e-01 1.70873508e-01
7.61283860e-02 1.66925006e-02 -6.17221832e-01 4.67298836e-01
3.31835705e-03 -5.40015846e-02 8.57463896e-01 1.11076212e+00
-4.91971374e-01 -8.37572873e-01 -5.79733729e-01 6.60517931e-01
-8.66495967e-01 -1.12503186e-01 -5.14194727e-01 5.22469103e-01
2.42854729e-01 9.83512163e-01 -4.34039116e-01 -4.14715976e-01
3.90579225e-03 -1.20749911e-02 -6.96708262e-02 -7.39253819e-01
-4.37768936e-01 -2.23368466e-01 3.68034363e-01 -1.89225376e-01
-8.48523140e-01 -3.49552929e-01 -5.18176556e-01 -2.53878653e-01
-9.18130636e-01 3.56840074e-01 6.03255272e-01 1.32302177e+00
3.92360687e-01 -2.21269682e-01 7.49970198e-01 -5.32384992e-01
-6.79741204e-01 -1.09410417e+00 -4.61302400e-01 8.29910576e-01
6.01463318e-01 -9.46812510e-01 -9.40928161e-01 3.30556184e-02]
|
[8.17074966430664, 10.300657272338867]
|
26facaae-c461-40eb-8c38-cec1efd7d1ab
|
watch-the-neighbors-a-unified-k-nearest
|
2210.08909
| null |
https://arxiv.org/abs/2210.08909v1
|
https://arxiv.org/pdf/2210.08909v1.pdf
|
Watch the Neighbors: A Unified K-Nearest Neighbor Contrastive Learning Framework for OOD Intent Discovery
|
Discovering out-of-domain (OOD) intent is important for developing new skills in task-oriented dialogue systems. The key challenges lie in how to transfer prior in-domain (IND) knowledge to OOD clustering, as well as jointly learn OOD representations and cluster assignments. Previous methods suffer from in-domain overfitting problem, and there is a natural gap between representation learning and clustering objectives. In this paper, we propose a unified K-nearest neighbor contrastive learning framework to discover OOD intents. Specifically, for IND pre-training stage, we propose a KCL objective to learn inter-class discriminative features, while maintaining intra-class diversity, which alleviates the in-domain overfitting problem. For OOD clustering stage, we propose a KCC method to form compact clusters by mining true hard negative samples, which bridges the gap between clustering and representation learning. Extensive experiments on three benchmark datasets show that our method achieves substantial improvements over the state-of-the-art methods.
|
['Weiran Xu', 'Wei Wu', 'Jingang Wang', 'Yanan Wu', 'Pei Wang', 'Keqing He', 'Yutao Mou']
|
2022-10-17
| null | null | null | null |
['intent-discovery', 'task-oriented-dialogue-systems']
|
['natural-language-processing', 'natural-language-processing']
|
[ 2.00724110e-01 2.10446611e-01 -4.96117771e-01 -7.29380667e-01
-7.73553848e-01 -5.35004258e-01 4.23198164e-01 1.64391503e-01
-2.23302498e-01 3.05536330e-01 5.15816689e-01 -1.50346100e-01
-2.82523632e-01 -2.83940852e-01 -7.83587694e-02 -4.93803531e-01
1.84518829e-01 8.41335952e-01 1.20981544e-01 -1.58498645e-01
5.21599174e-01 4.38965261e-02 -1.50947666e+00 6.93034232e-01
1.49041009e+00 8.63362670e-01 2.64808029e-01 3.51685554e-01
-4.89184439e-01 9.96898651e-01 -8.93509686e-01 -1.96116447e-01
-1.39241710e-01 -4.11519080e-01 -1.14314532e+00 6.23208761e-01
4.35724072e-02 -1.61055148e-01 -4.34624344e-01 8.32733154e-01
5.90753019e-01 3.99386257e-01 1.21258616e+00 -1.35721922e+00
-8.59893560e-01 3.80384505e-01 -8.82747173e-01 -2.23599717e-01
1.70863718e-01 -2.51659989e-01 9.71542656e-01 -8.34422350e-01
4.25336421e-01 1.40623975e+00 4.26733583e-01 9.11965370e-01
-1.21134853e+00 -5.79168022e-01 3.09305042e-01 3.16179954e-02
-1.23627365e+00 -3.40476424e-01 9.24798906e-01 -6.00202084e-01
8.02050948e-01 1.66632801e-01 3.47593725e-01 9.39654112e-01
-3.52167904e-01 1.39366233e+00 1.08277512e+00 -6.26669168e-01
4.06070888e-01 4.71583962e-01 4.29442227e-01 4.79198664e-01
-1.26833901e-01 -5.78219950e-01 -5.91316164e-01 -4.14597660e-01
5.03239512e-01 2.85176367e-01 -1.23669572e-01 -7.28389919e-01
-7.23356068e-01 1.21348023e+00 1.12429842e-01 2.49552459e-01
1.48438916e-01 -5.16312778e-01 5.61502278e-01 4.08538193e-01
6.59195423e-01 3.66836458e-01 -5.50914943e-01 -3.94636571e-01
-4.69860435e-01 1.60471857e-01 8.99972320e-01 1.10219884e+00
8.22020888e-01 -5.81369221e-01 -2.69503176e-01 1.48423314e+00
3.07066262e-01 -1.14688843e-01 9.06177819e-01 -9.63682175e-01
9.28591371e-01 1.13879848e+00 -2.01317132e-01 -9.64478791e-01
-2.11813599e-01 -1.77707881e-01 -1.01539850e+00 -2.97627628e-01
1.67835072e-01 -2.40674883e-01 -7.19884276e-01 1.47210181e+00
5.46796858e-01 5.07733859e-02 4.11646336e-01 8.67726982e-01
7.98108697e-01 6.26207054e-01 -1.67412475e-01 -3.60433787e-01
1.26832545e+00 -1.04446971e+00 -8.59174609e-01 -4.52550799e-01
1.14459074e+00 -5.60483277e-01 1.20913005e+00 4.03463274e-01
-6.53417408e-01 -5.47990382e-01 -7.21244156e-01 -1.86804861e-01
-1.07771866e-01 1.87784895e-01 7.02077806e-01 5.31243682e-01
-5.59256613e-01 1.95223883e-01 -3.83504570e-01 -1.15610950e-01
4.83777225e-01 4.47180003e-01 -3.21340561e-01 -4.44396228e-01
-1.08513010e+00 4.57168043e-01 6.21550858e-01 -2.96054244e-01
-6.56441689e-01 -6.71468973e-01 -9.40225899e-01 -2.42427379e-01
3.84450942e-01 -4.25931588e-02 1.24065030e+00 -7.44809628e-01
-1.50931144e+00 1.16103768e+00 -3.23789477e-01 -1.86499342e-01
3.06573778e-01 -2.49275282e-01 -1.88355803e-01 -9.27691609e-02
2.21208900e-01 6.09658301e-01 7.99055338e-01 -1.35541809e+00
-7.42725909e-01 -6.52608037e-01 -1.86322287e-01 7.52044678e-01
-9.51178491e-01 -2.53116876e-01 -8.98515224e-01 -5.79643905e-01
3.71417642e-01 -8.50180209e-01 -2.50983298e-01 -4.03891414e-01
-3.84127766e-01 -1.05210853e+00 1.12195218e+00 -2.97322303e-01
1.42047000e+00 -2.35874891e+00 2.23346874e-01 1.39123827e-01
5.94309151e-01 3.54030520e-01 4.11139913e-02 2.43924439e-01
1.04590639e-01 -1.98336273e-01 -1.71835482e-01 -5.38098514e-01
2.56703198e-01 2.55630255e-01 -3.13483149e-01 2.26388231e-01
3.38870972e-01 4.39391851e-01 -8.77390265e-01 -7.22950220e-01
4.88249771e-02 1.21394642e-01 -7.18570828e-01 6.71012521e-01
-1.76569298e-01 2.89436787e-01 -6.67798877e-01 4.72608596e-01
7.55773544e-01 -2.35975489e-01 5.68037868e-01 8.52245465e-02
1.43164918e-01 5.03526747e-01 -1.18126690e+00 1.97480977e+00
-4.26285833e-01 4.75858241e-01 -7.70657184e-03 -1.64612639e+00
1.31607008e+00 1.42944559e-01 2.31739461e-01 -6.95368290e-01
6.95146387e-04 1.68307036e-01 -1.49840057e-01 -5.94397247e-01
5.96830845e-01 -1.94534957e-02 -4.98813719e-01 5.57306170e-01
2.26428211e-01 -7.06350356e-02 -7.46755004e-02 3.05826306e-01
7.66935885e-01 -3.27786595e-01 3.08743805e-01 -2.40164608e-01
3.86150181e-01 1.79084659e-01 7.56161928e-01 5.84399819e-01
-2.40603924e-01 7.49911427e-01 1.01667750e+00 -1.80349335e-01
-5.25926232e-01 -8.04667234e-01 -1.13466740e-01 1.41789031e+00
1.06593616e-01 -4.39031959e-01 -8.45994771e-01 -1.17374015e+00
4.68393043e-02 4.77638811e-01 -6.49510860e-01 -5.00411093e-01
-4.79441941e-01 -5.80041230e-01 4.41441208e-01 4.76793110e-01
3.70744973e-01 -7.08270609e-01 1.29539102e-01 1.87468931e-01
-3.19439679e-01 -9.08790648e-01 -5.92273712e-01 5.56722343e-01
-8.95572245e-01 -8.54161680e-01 -8.13360214e-01 -1.42878997e+00
8.16864371e-01 7.33240962e-01 1.01959383e+00 -8.26729536e-02
-2.85155267e-01 1.91056907e-01 -5.36700606e-01 -3.69290501e-01
-2.22334161e-01 3.30147773e-01 3.60479861e-01 -7.19321221e-02
1.07553208e+00 -3.45612347e-01 -3.29863936e-01 3.39580357e-01
-6.36245370e-01 -2.17928424e-01 6.15206957e-01 1.12447202e+00
6.26144111e-01 4.33761060e-01 7.52377391e-01 -1.36368382e+00
1.10728121e+00 -7.58427501e-01 -1.86515078e-01 1.90115586e-01
-5.85859299e-01 4.32338342e-02 6.65311635e-01 -4.99617547e-01
-1.14585638e+00 2.35436872e-01 2.52125055e-01 -7.20724404e-01
-3.91991913e-01 4.51354295e-01 -4.54322696e-01 4.41901386e-01
4.75015730e-01 4.08827692e-01 1.93513304e-01 -6.74447775e-01
2.96097845e-01 1.49511194e+00 3.83732468e-01 -8.67546320e-01
4.79131490e-01 2.13223457e-01 -7.23011196e-01 -1.01716352e+00
-1.24297309e+00 -1.12756562e+00 -1.01225364e+00 2.11492151e-01
8.03205311e-01 -1.29203188e+00 -5.25111616e-01 3.09841573e-01
-9.27444994e-01 -3.55121523e-01 -1.68513075e-01 4.32359338e-01
-5.25031209e-01 4.97402787e-01 -3.24346721e-01 -1.00037265e+00
5.64738400e-02 -1.10536468e+00 9.78280067e-01 4.19351727e-01
-3.46082091e-01 -1.11772597e+00 6.76383823e-02 8.98046017e-01
-3.15884858e-01 -3.65059793e-01 8.91596019e-01 -1.25595510e+00
9.98317003e-02 -6.29094988e-02 -2.91043371e-01 4.78294939e-01
3.38616669e-01 -6.50167525e-01 -9.32740033e-01 -3.65953356e-01
9.29003730e-02 -9.73411322e-01 8.20052862e-01 5.65679260e-02
1.47937191e+00 -3.87672454e-01 -4.45635110e-01 2.14080796e-01
7.16805279e-01 1.30506635e-01 3.95328641e-01 7.69695565e-02
7.35133648e-01 1.14019966e+00 1.30889261e+00 6.47956967e-01
6.82593882e-01 7.30321586e-01 -8.51824880e-02 -1.49544030e-01
1.57867312e-01 -3.41003686e-01 2.78932661e-01 1.03836238e+00
5.05807459e-01 -6.95759356e-02 -9.23054338e-01 7.47011542e-01
-1.99629128e+00 -6.95131183e-01 1.11600846e-01 1.94511271e+00
1.38003767e+00 -5.08038269e-04 3.70398730e-01 1.48062944e-01
8.19022059e-01 4.61960956e-02 -6.49469554e-01 -2.98211962e-01
3.71930808e-01 -1.01019993e-01 -9.51224193e-02 2.91893691e-01
-1.35472715e+00 1.08863842e+00 5.18786573e+00 1.27403224e+00
-6.32809639e-01 3.22124325e-02 6.55940831e-01 2.63241678e-01
-2.26635903e-01 -1.46530718e-01 -1.11337268e+00 4.87440258e-01
3.81934792e-01 1.76992744e-01 1.30477399e-01 1.15231979e+00
-4.76930588e-02 3.07567775e-01 -9.33245599e-01 1.06478703e+00
3.64940256e-01 -1.02204692e+00 -1.04871199e-01 1.94313839e-01
9.35526490e-01 -3.88880461e-01 8.85012969e-02 6.64019704e-01
6.14920378e-01 -1.02932107e+00 -1.28919976e-02 3.96559201e-02
7.41399527e-01 -1.11490464e+00 6.81669474e-01 7.53618300e-01
-1.06963873e+00 -5.62951900e-02 -7.82297015e-01 -7.33487727e-03
-5.93907535e-01 6.61781251e-01 -1.12639630e+00 2.93183029e-01
8.08501363e-01 9.85946894e-01 -3.00507933e-01 5.10718822e-01
6.66668862e-02 5.09925067e-01 -1.24918418e-02 -2.14053720e-01
4.07198668e-01 -3.01298738e-01 5.97909950e-02 1.26515341e+00
-1.29608348e-01 2.40123689e-01 5.51222682e-01 8.13136816e-01
-2.01013803e-01 1.30841862e-02 -6.71428144e-01 -2.18480945e-01
9.64498460e-01 9.85200405e-01 -3.79679531e-01 -1.30886421e-01
-4.43331450e-01 1.26372778e+00 7.58177221e-01 1.19940870e-01
-4.39263999e-01 -8.10430825e-01 9.23806489e-01 -1.52260706e-01
2.68345058e-01 -1.07486479e-01 -2.12949589e-01 -1.11038828e+00
-1.55816153e-02 -1.20636284e+00 6.57785654e-01 3.20188142e-02
-1.72239494e+00 -4.10796404e-02 -2.21584156e-01 -1.56544924e+00
-1.19180523e-01 -4.40603733e-01 -6.72476053e-01 6.13383472e-01
-1.56043112e+00 -9.08470094e-01 -2.14903906e-01 7.23456025e-01
9.46581602e-01 -5.41842818e-01 7.48247445e-01 1.47370532e-01
-5.28603852e-01 9.41011906e-01 6.11949205e-01 3.88642102e-01
1.02369165e+00 -1.41227794e+00 -3.09578218e-02 1.46339089e-01
-6.33120537e-02 8.15983057e-01 1.13175869e-01 -5.54849207e-01
-1.39805865e+00 -1.24219346e+00 9.40445185e-01 -3.22071761e-01
3.56433898e-01 -6.98863685e-01 -1.01684713e+00 3.87231469e-01
-1.05728559e-01 -1.16510026e-01 1.30796897e+00 5.46366274e-01
-3.68934810e-01 1.54511914e-01 -9.84659553e-01 3.10513794e-01
9.85369623e-01 -5.75845838e-01 -1.07236576e+00 6.28342152e-01
9.71235156e-01 -4.17981833e-01 -1.05912936e+00 1.54038134e-03
1.85446009e-01 -8.54002774e-01 1.00792813e+00 -7.70225048e-01
3.99400979e-01 -8.82470906e-02 -1.16795093e-01 -1.19505167e+00
-1.58896595e-01 -5.84819913e-01 -3.36365998e-01 1.68884778e+00
1.59657776e-01 -3.20342034e-01 1.06354380e+00 4.41444814e-01
-3.73456806e-01 -8.95693183e-01 -7.14712143e-01 -6.43926620e-01
3.00337970e-01 -1.61957458e-01 2.65235811e-01 1.51401556e+00
5.46837747e-01 7.58295357e-01 -3.79508257e-01 6.92253932e-02
5.99683344e-01 3.78715336e-01 1.01610625e+00 -1.55169284e+00
-1.88265115e-01 1.05915871e-02 -2.34754175e-01 -1.82503247e+00
5.55168808e-01 -8.40908766e-01 2.10509688e-01 -1.35278046e+00
4.55293775e-01 -8.85245204e-01 -3.38186845e-02 4.34909493e-01
-4.08272803e-01 -8.27471912e-02 -5.44938520e-02 5.22775352e-01
-1.29564261e+00 8.18300128e-01 1.13182616e+00 -1.56295642e-01
-5.81783593e-01 7.93136805e-02 -1.07057345e+00 7.98173130e-01
6.57220185e-01 -7.57947028e-01 -8.31558228e-01 -3.65263164e-01
-4.83217806e-01 -3.72278132e-02 -2.22748309e-01 -7.44564414e-01
3.61259043e-01 -2.80406684e-01 3.99804264e-01 -8.92461896e-01
3.79449695e-01 -5.65949142e-01 -8.54310989e-01 4.02835876e-01
-7.30936766e-01 -6.25332892e-01 -2.59939462e-01 8.78819287e-01
-4.95544434e-01 -3.88976872e-01 7.12641537e-01 -1.02807879e-01
-8.08735192e-01 1.13871545e-01 -3.52841765e-01 7.66025841e-01
9.82564688e-01 -2.41795495e-01 -2.80622989e-01 -2.28372589e-01
-4.76403177e-01 7.41222203e-01 3.47757012e-01 6.73565090e-01
7.12560415e-01 -1.24771535e+00 -5.01248598e-01 2.92614609e-01
3.84407669e-01 5.44304609e-01 4.08257514e-01 5.96591115e-01
-2.43861806e-02 4.87993658e-01 2.94642061e-01 -8.03722680e-01
-1.31554949e+00 3.36063325e-01 -5.84674701e-02 -4.58990306e-01
-3.59712273e-01 1.12040687e+00 4.72152174e-01 -1.32948756e+00
7.73561835e-01 1.91657171e-01 -5.09037495e-01 2.65883923e-01
5.01608431e-01 2.35883951e-01 -8.88769180e-02 -2.74414808e-01
-3.11975688e-01 5.02625406e-01 -6.81021988e-01 2.86089450e-01
1.27868986e+00 -3.27386439e-01 5.24356030e-02 5.39168239e-01
1.50453317e+00 -2.25134343e-01 -1.24570858e+00 -4.42100048e-01
5.81256412e-02 -4.11943406e-01 -1.12305343e-01 -5.13802290e-01
-7.42641628e-01 1.25368583e+00 4.46714669e-01 2.60988832e-01
8.64442468e-01 3.34783465e-01 8.52338731e-01 7.15653002e-01
-9.28828865e-02 -1.55175388e+00 6.90646231e-01 7.50968218e-01
4.95132029e-01 -1.53352094e+00 -4.64085080e-02 -5.06838799e-01
-1.35515666e+00 9.43011642e-01 1.13060725e+00 3.55808362e-02
5.08401215e-01 -3.70687842e-02 -3.75845060e-02 -1.85410544e-01
-7.43371844e-01 -7.46250600e-02 3.51751089e-01 9.83528018e-01
6.42434239e-01 5.65018989e-02 -1.39119908e-01 1.10215771e+00
5.86235262e-02 -3.86003762e-01 1.91016436e-01 1.09166944e+00
-6.63629413e-01 -1.25088215e+00 -1.94576353e-01 6.06784642e-01
-1.23039939e-01 9.57131013e-02 -9.67126250e-01 6.42791688e-01
6.68445928e-03 1.08702195e+00 2.57571340e-01 -6.16419196e-01
3.39630306e-01 4.52716351e-02 3.53056230e-02 -1.08564889e+00
-3.04750621e-01 1.24538645e-01 5.91301769e-02 -2.16461375e-01
-2.99907565e-01 -6.83197200e-01 -1.50779712e+00 -2.58154087e-02
-6.25092924e-01 6.91787422e-01 1.95957676e-01 8.91823769e-01
6.51642144e-01 3.65058124e-01 1.11643982e+00 -4.26755339e-01
-8.46523941e-01 -9.50648189e-01 -7.82043457e-01 5.03188431e-01
8.26503560e-02 -7.57800937e-01 -4.47717905e-01 -6.54575378e-02]
|
[12.42894172668457, 7.53891658782959]
|
3b4d2c85-52f4-4d09-aee3-273780591558
|
upsampling-layers-for-music-source-separation
|
2111.11773
| null |
https://arxiv.org/abs/2111.11773v1
|
https://arxiv.org/pdf/2111.11773v1.pdf
|
Upsampling layers for music source separation
|
Upsampling artifacts are caused by problematic upsampling layers and due to spectral replicas that emerge while upsampling. Also, depending on the used upsampling layer, such artifacts can either be tonal artifacts (additive high-frequency noise) or filtering artifacts (substractive, attenuating some bands). In this work we investigate the practical implications of having upsampling artifacts in the resulting audio, by studying how different artifacts interact and assessing their impact on the models' performance. To that end, we benchmark a large set of upsampling layers for music source separation: different transposed and subpixel convolution setups, different interpolation upsamplers (including two novel layers based on stretch and sinc interpolation), and different wavelet-based upsamplers (including a novel learnable wavelet layer). Our results show that filtering artifacts, associated with interpolation upsamplers, are perceptually preferrable, even if they tend to achieve worse objective scores.
|
['Davide Scaini', 'Daniel Arteaga', 'Giulio Cengarle', 'Santiago Pascual', 'Joan Serrà', 'Jordi Pons']
|
2021-11-23
| null | null | null | null |
['music-source-separation']
|
['music']
|
[ 3.04994822e-01 -1.71917617e-01 1.58093408e-01 1.46930501e-01
-8.33709896e-01 -5.75550854e-01 4.17526335e-01 -1.07644238e-01
-1.90631568e-01 7.29890168e-01 6.28131807e-01 5.48576936e-02
-2.81702161e-01 -5.60561478e-01 -8.09899449e-01 -5.60935438e-01
-2.88396925e-01 -2.31935516e-01 2.52288371e-01 -1.25795707e-01
1.00363128e-01 1.84593216e-01 -1.88837564e+00 7.51232207e-01
1.05360568e+00 7.53703177e-01 8.74988288e-02 7.90344238e-01
1.16371907e-01 6.06077015e-01 -1.03920496e+00 -2.40409330e-01
5.03244221e-01 -4.93561029e-01 -4.09398556e-01 -2.08746530e-02
7.14821756e-01 -3.36210072e-01 1.82653934e-01 1.06806648e+00
5.34229517e-01 3.80772240e-02 3.96575063e-01 -9.00949955e-01
-4.38894063e-01 1.21307313e+00 -4.12890881e-01 4.82589565e-02
3.31673205e-01 2.42262572e-01 8.89157116e-01 -6.83669806e-01
4.12673563e-01 1.36497414e+00 1.26175952e+00 6.13055713e-02
-1.81351471e+00 -7.44323730e-01 -2.81592309e-01 4.93269473e-01
-1.11730099e+00 -5.75805366e-01 8.79707456e-01 -4.13068056e-01
5.76363444e-01 6.44699156e-01 6.98780179e-01 1.35918486e+00
3.34404111e-02 2.24226162e-01 1.56162596e+00 -4.64312494e-01
3.46066296e-01 2.26540029e-01 -1.55470237e-01 -1.80043265e-01
3.62564683e-01 2.98939973e-01 -6.17906749e-01 -1.48569599e-01
9.57846761e-01 -5.03907382e-01 -6.50310576e-01 9.22628269e-02
-1.11627531e+00 4.13588226e-01 3.38368714e-01 4.47518677e-01
-4.30847377e-01 4.43670064e-01 2.55959660e-01 5.79023540e-01
3.84200007e-01 1.01853645e+00 -5.48412919e-01 -1.19158499e-01
-1.54602253e+00 3.22743833e-01 5.18955410e-01 5.25342524e-01
4.71333355e-01 3.24112087e-01 -5.04239678e-01 9.31665063e-01
-1.35200724e-01 6.73235906e-03 6.51472092e-01 -1.13376546e+00
2.64614075e-01 -1.05788209e-01 3.20504040e-01 -7.66240954e-01
-6.13901198e-01 -8.34358871e-01 -7.23245263e-01 6.65997744e-01
7.60739625e-01 -3.34930807e-01 -5.73304534e-01 1.64004183e+00
-1.27767533e-01 5.97100139e-01 -2.67030597e-01 1.11221278e+00
4.19660807e-01 4.33197170e-01 -4.07962762e-02 -1.61146969e-01
1.64573312e+00 -9.63635802e-01 -8.67512465e-01 3.84074561e-02
1.73866913e-01 -1.38923764e+00 1.39455998e+00 1.00188112e+00
-1.34134567e+00 -1.05645549e+00 -1.33715808e+00 -2.59053171e-01
-1.32735178e-01 3.65422666e-01 3.69882047e-01 8.21449161e-01
-1.12863922e+00 1.44377506e+00 -4.36594307e-01 -5.94085176e-03
3.00913714e-02 -4.54600975e-02 2.28636041e-01 4.37044382e-01
-1.21008670e+00 9.06029880e-01 1.40532926e-01 -1.96842492e-01
-5.50100148e-01 -1.17560148e+00 -3.06093305e-01 4.36568409e-01
-2.31822208e-02 -7.30536342e-01 1.09975553e+00 -1.55855572e+00
-1.42340541e+00 3.76485854e-01 1.56476825e-01 -7.77998865e-01
9.59069073e-01 -5.94175279e-01 -6.78148031e-01 -2.14329716e-02
-4.47498649e-01 2.87263274e-01 1.49073827e+00 -1.29350567e+00
-1.50402576e-01 1.56669971e-02 -7.57602751e-02 1.99371600e-03
-1.62891641e-01 -1.07339770e-01 3.18733722e-01 -1.36513257e+00
1.09389178e-01 -7.19031453e-01 -9.47928429e-02 -1.28927305e-01
-4.17416364e-01 5.55383742e-01 6.18241608e-01 -1.15371692e+00
1.28426921e+00 -2.43607426e+00 -2.98529472e-02 -1.04754247e-01
-8.55020583e-02 2.31981322e-01 -3.60142767e-01 3.91463488e-01
-4.47201967e-01 1.51749715e-01 -8.77844617e-02 -4.46375906e-01
-8.10808092e-02 -1.73999161e-01 -4.11595821e-01 1.23232514e-01
3.60285342e-01 2.20867217e-01 -7.81419575e-01 4.71325032e-02
3.02056044e-01 8.21707547e-01 -7.92027414e-01 -3.40210199e-01
-1.44522294e-01 6.83050454e-01 1.75280720e-01 2.30328113e-01
9.37882662e-01 1.50427818e-01 -6.27512708e-02 -5.99707425e-01
-3.99277776e-01 7.19103456e-01 -1.50554800e+00 1.59395647e+00
-8.12422991e-01 1.02145672e+00 2.51495093e-01 -3.00969481e-01
7.15293586e-01 6.03545129e-01 7.81153515e-02 -3.81774098e-01
-7.12418780e-02 4.64316368e-01 2.17508763e-01 -4.86395240e-01
7.46913314e-01 -2.16040820e-01 6.73604429e-01 6.88589960e-02
9.35791805e-02 -3.18712056e-01 -1.02336495e-03 -5.19263685e-01
9.65884209e-01 2.31004521e-01 9.67047215e-02 -5.33186376e-01
3.08703184e-01 -4.37987715e-01 3.27276826e-01 6.79302454e-01
-1.99493556e-03 1.22446835e+00 5.16272008e-01 -1.28681749e-01
-1.13602996e+00 -1.09536636e+00 -3.35787237e-01 8.77952874e-01
-1.74927413e-01 -4.97737676e-01 -8.86042237e-01 7.81384781e-02
-1.47771806e-01 8.30911577e-01 -3.94635260e-01 -2.91962206e-01
-4.84471709e-01 -5.62652409e-01 7.46937037e-01 3.49017173e-01
4.35900539e-01 -6.90349340e-01 -7.93524921e-01 2.89923340e-01
-4.04354274e-01 -8.16697061e-01 -5.25056064e-01 2.11022124e-01
-1.17933285e+00 -8.34333837e-01 -8.73180926e-01 -2.80186445e-01
2.34738708e-01 1.28058314e-01 1.18615532e+00 -9.78661403e-02
-1.50886580e-01 -1.11151777e-01 -4.56057519e-01 -4.35800374e-01
-5.30309200e-01 -2.37645611e-01 1.30960807e-01 1.87308505e-01
-2.29379147e-01 -9.96715546e-01 -8.74333501e-01 1.81983486e-01
-9.26808715e-01 1.83146209e-01 6.87459230e-01 5.42483330e-01
1.58650070e-01 3.08258444e-01 4.87877935e-01 -6.67368531e-01
8.47261131e-01 -1.73453763e-01 -5.90376079e-01 -2.35948250e-01
-5.35419703e-01 6.61836043e-02 9.42643583e-01 -8.58466983e-01
-9.77406979e-01 -3.54297936e-01 -1.96783140e-01 -6.76963151e-01
-1.80748612e-01 1.24501571e-01 1.88948706e-01 2.67670471e-02
1.07083929e+00 -1.50013983e-01 -3.90877366e-01 -9.32462335e-01
2.51562864e-01 4.30212319e-01 4.06661540e-01 -4.63168293e-01
8.88755143e-01 4.54379976e-01 -1.61129877e-01 -8.71435046e-01
-4.39969033e-01 -3.30339335e-02 -1.66667551e-01 -2.49082938e-01
7.25326002e-01 -9.23755586e-01 -1.57928810e-01 3.95948172e-01
-1.15195978e+00 -3.44157904e-01 -7.35006034e-01 8.21387589e-01
-4.28908676e-01 9.65442955e-02 -9.16477203e-01 -9.57960844e-01
-8.55184719e-03 -1.24891865e+00 7.84895480e-01 2.56926477e-01
-6.31404459e-01 -5.85541725e-01 -9.76333842e-02 1.97693869e-01
8.40114474e-01 3.65166396e-01 8.22264194e-01 -2.77507603e-01
-3.70824158e-01 3.81768763e-01 -9.08073876e-03 6.12700164e-01
1.81959182e-01 2.37743273e-01 -1.64656806e+00 -1.40940845e-01
2.19340175e-01 1.83186918e-01 1.00371480e+00 8.19179177e-01
1.07918739e+00 -6.50524199e-01 2.53439188e-01 6.12423480e-01
1.27809811e+00 -1.10639051e-01 1.01719022e+00 3.74340862e-01
2.75455147e-01 7.24764705e-01 2.32969254e-01 5.53508878e-01
-4.23306763e-01 7.16911256e-01 3.18265140e-01 -3.23962033e-01
-6.98676407e-01 -2.14147910e-01 5.65672696e-01 6.06159091e-01
-5.39232731e-01 2.42825776e-01 -2.84925461e-01 3.64901572e-01
-1.32010651e+00 -1.00008297e+00 -6.45114541e-01 2.60201907e+00
1.17759717e+00 1.24700233e-01 3.43258202e-01 7.24821150e-01
7.17816889e-01 2.39118412e-01 -1.65189788e-01 -5.10873675e-01
-2.57676214e-01 7.34314859e-01 6.42486989e-01 5.48052907e-01
-7.91878760e-01 2.82117844e-01 6.47709608e+00 8.74646544e-01
-1.41179621e+00 1.74195796e-01 4.46882755e-01 -4.68906432e-01
-5.19737184e-01 8.46462175e-02 -1.32697850e-01 6.57672286e-01
8.40918958e-01 1.11253053e-01 8.43868971e-01 4.70835894e-01
6.37025297e-01 3.01172994e-02 -1.02403569e+00 7.54561007e-01
-3.92723769e-01 -1.14204490e+00 -1.21299751e-01 -2.54056573e-01
5.93958616e-01 -1.76882058e-01 3.43332112e-01 1.32869542e-01
-7.41531234e-03 -9.70188022e-01 1.35913074e+00 4.87738729e-01
5.61310828e-01 -5.45694470e-01 5.05505681e-01 -7.03348815e-02
-9.41303849e-01 -9.18191671e-02 -1.09676823e-01 -3.66227120e-01
-1.12984784e-01 1.13732147e+00 -6.97625756e-01 3.45406860e-01
7.54669487e-01 1.89021125e-01 -4.61791039e-01 1.21032095e+00
-3.85999829e-01 8.32305610e-01 -2.21223205e-01 4.43164676e-01
-2.99425662e-01 -2.57832587e-01 8.86389315e-01 1.34662628e+00
6.34005904e-01 -2.84823507e-01 -5.50275564e-01 1.15241432e+00
3.27610746e-02 -2.16986239e-01 -6.99442476e-02 2.64909744e-01
6.08005524e-01 1.19393873e+00 -5.47293067e-01 -1.70544669e-01
-2.60211319e-01 8.43331218e-01 -4.68302697e-01 6.34651184e-01
-1.05051267e+00 -4.23008174e-01 9.80072260e-01 5.70300221e-01
1.77195668e-01 8.59756246e-02 -5.26279688e-01 -1.07891572e+00
1.59608692e-01 -9.86161470e-01 2.70538311e-02 -1.04059720e+00
-1.04844320e+00 4.54623163e-01 -2.95148790e-01 -1.63001192e+00
3.25486846e-02 -3.69428784e-01 -7.57710099e-01 8.97270977e-01
-1.67886126e+00 -6.60090148e-01 -2.71031737e-01 3.26914221e-01
5.37878096e-01 4.86470461e-01 5.60805738e-01 6.18594170e-01
1.67487841e-02 4.60395783e-01 5.24972044e-02 -3.12091112e-01
1.04456246e+00 -1.14850187e+00 2.45457888e-01 8.08174074e-01
2.99559146e-01 6.34391010e-01 1.34601104e+00 -1.82119280e-01
-9.96858120e-01 -1.08065927e+00 6.18276656e-01 -2.99685765e-02
5.37511826e-01 -2.67019391e-01 -1.12893498e+00 3.32646161e-01
2.37778515e-01 -2.21075088e-01 4.91948843e-01 1.66982457e-01
-5.90668380e-01 -1.36059687e-01 -1.23158717e+00 7.48195708e-01
7.36080468e-01 -4.67321157e-01 -5.21169960e-01 3.25111821e-02
8.34313333e-01 -3.94006550e-01 -7.91085422e-01 2.34270111e-01
7.11075366e-01 -1.63967526e+00 1.31381118e+00 -1.14969686e-01
8.26519489e-01 -4.03584421e-01 1.35786727e-01 -1.69652057e+00
-5.70953667e-01 -9.07437325e-01 3.73294465e-02 1.29278040e+00
4.96781558e-01 -7.77577043e-01 2.05559999e-01 2.02968549e-02
-2.26261511e-01 -3.16957116e-01 -8.51020217e-01 -1.10796309e+00
8.15532207e-02 -4.24434841e-01 6.21646345e-01 1.13198221e+00
3.99751179e-02 -9.19221193e-02 -4.69779283e-01 2.95279831e-01
3.56500119e-01 -8.09282213e-02 4.65535432e-01 -1.19032657e+00
-1.00188947e+00 -6.72570527e-01 -8.95339176e-02 -5.75347006e-01
-6.22471094e-01 -3.45322222e-01 -5.00176698e-02 -8.69736612e-01
-4.45580125e-01 -3.39937299e-01 -4.28242862e-01 1.07578076e-01
-7.37628862e-02 4.44835395e-01 5.34306645e-01 1.88322350e-01
3.03609133e-01 1.14551887e-01 1.15518320e+00 -5.72137870e-02
-4.49567795e-01 -2.15516221e-02 -5.72539032e-01 1.25537467e+00
7.03921914e-01 -2.86444485e-01 -2.56308168e-01 -4.37216967e-01
3.28186959e-01 4.69371155e-02 6.86041772e-01 -1.62879980e+00
-2.39351705e-01 8.86297598e-02 4.17083263e-01 -1.14585876e-01
5.50441861e-01 -7.04659045e-01 6.96098268e-01 5.81578970e-01
-5.22434771e-01 -3.54703665e-01 5.17791390e-01 2.05304444e-01
-2.33167678e-01 -2.50635684e-01 9.78044271e-01 5.97662292e-02
-5.51346987e-02 -5.73949873e-01 -3.94572079e-01 -1.48086667e-01
5.07027745e-01 -1.56879723e-01 -2.40749553e-01 -4.41028684e-01
-8.16897750e-01 -4.82972383e-01 4.94472355e-01 3.51997465e-01
2.54753888e-01 -1.04483306e+00 -8.60984862e-01 1.36437699e-01
-3.41980547e-01 -4.87615615e-01 3.96493882e-01 1.07551682e+00
-3.52320075e-01 -2.04746928e-02 -2.78708935e-01 -1.56363621e-01
-1.19032407e+00 3.57323378e-01 2.83289075e-01 -1.89143434e-01
-6.10352278e-01 9.36097860e-01 3.56919616e-02 1.45283967e-01
2.76743889e-01 -1.13432515e+00 2.39818990e-02 3.39315116e-01
5.34972429e-01 7.22251713e-01 3.37470800e-01 -7.11126626e-02
-1.65364072e-02 4.45304185e-01 4.34210986e-01 -2.23778710e-01
1.03371918e+00 1.62838027e-01 1.01644576e-01 7.39644170e-01
8.44067812e-01 7.45194435e-01 -1.22771692e+00 1.21506438e-01
-2.25498781e-01 -5.19218087e-01 1.14522927e-01 -7.94561684e-01
-8.90295863e-01 6.95265114e-01 6.57997608e-01 8.06669235e-01
1.50884068e+00 -4.33240414e-01 7.66700447e-01 -2.84005910e-01
4.62438822e-01 -1.10382843e+00 -1.75584927e-01 -2.00479236e-02
1.19493830e+00 -5.12432337e-01 5.12069725e-02 -5.75807810e-01
-1.29224762e-01 1.20984185e+00 -2.12639868e-02 -5.16480148e-01
2.86772162e-01 4.34963614e-01 1.90588549e-01 4.42367822e-01
-5.16413212e-01 -2.33924165e-01 2.34827489e-01 5.66351116e-01
7.60493279e-01 9.71240774e-02 -5.58198869e-01 8.00000072e-01
-8.32350433e-01 1.42402694e-01 8.36433589e-01 2.67890126e-01
-2.20420793e-01 -9.20841634e-01 -1.06169999e+00 3.92244905e-01
-4.91873205e-01 -4.31990176e-01 -3.02358419e-01 4.89030838e-01
6.82873011e-01 1.04379153e+00 5.74175939e-02 -3.37947667e-01
4.05107707e-01 7.19966367e-02 6.08059943e-01 -2.03543335e-01
-1.08772993e+00 4.15287048e-01 2.45534450e-01 -5.76787353e-01
-2.57114977e-01 -7.31725574e-01 -9.27039325e-01 -1.51235655e-01
-6.36449933e-01 -1.27428949e-01 6.59250855e-01 5.16301095e-01
4.11605209e-01 1.24581802e+00 2.98701823e-01 -1.04508567e+00
-6.35149240e-01 -1.28086448e+00 -8.51216495e-01 5.48004389e-01
6.55576885e-01 -3.80731106e-01 -7.99354374e-01 4.44087327e-01]
|
[15.470831871032715, 5.801798343658447]
|
10371563-ba38-4c58-8726-c56d99c5e463
|
infusing-future-information-into-monotonic
|
2109.03121
| null |
https://arxiv.org/abs/2109.03121v1
|
https://arxiv.org/pdf/2109.03121v1.pdf
|
Infusing Future Information into Monotonic Attention Through Language Models
|
Simultaneous neural machine translation(SNMT) models start emitting the target sequence before they have processed the source sequence. The recent adaptive policies for SNMT use monotonic attention to perform read/write decisions based on the partial source and target sequences. The lack of sufficient information might cause the monotonic attention to take poor read/write decisions, which in turn negatively affects the performance of the SNMT model. On the other hand, human translators make better read/write decisions since they can anticipate the immediate future words using linguistic information and domain knowledge.Motivated by human translators, in this work, we propose a framework to aid monotonic attention with an external language model to improve its decisions.We conduct experiments on the MuST-C English-German and English-French speech-to-text translation tasks to show the effectiveness of the proposed framework.The proposed SNMT method improves the quality-latency trade-off over the state-of-the-art monotonic multihead attention.
|
['Sangha Kim', 'Nikhil Kumar Lakumarapu', 'Beomseok Lee', 'Sathish Indurthi', 'Mohd Abbas Zaidi']
|
2021-09-07
|
infusing-future-information-into-monotonic-1
|
https://openreview.net/forum?id=lgGKToqwtwG
|
https://openreview.net/pdf?id=lgGKToqwtwG
| null |
['speech-to-text-translation']
|
['natural-language-processing']
|
[ 3.05288702e-01 -2.76091937e-02 -4.64012772e-01 -4.45715874e-01
-9.03553903e-01 -4.69180733e-01 6.32772505e-01 -2.29851276e-01
-6.12275481e-01 8.42205942e-01 3.92493546e-01 -7.59295344e-01
5.33114672e-01 -4.54871088e-01 -7.79903233e-01 -5.14444232e-01
6.54187083e-01 6.32472396e-01 5.18937521e-02 -1.92228243e-01
1.65388882e-01 7.80233592e-02 -1.01219213e+00 5.03112078e-01
1.16525674e+00 6.33600533e-01 9.02418375e-01 5.40309131e-01
-3.24471116e-01 7.50267565e-01 -3.51039320e-01 -4.79212433e-01
1.02557741e-01 -7.58702993e-01 -9.02590871e-01 -2.98226207e-01
5.47667369e-02 -3.12510431e-01 -1.62891120e-01 9.92991030e-01
6.60020888e-01 1.73849255e-01 4.23469275e-01 -9.00324523e-01
-1.02157640e+00 9.84802485e-01 -5.09339035e-01 5.68247557e-01
2.08075419e-01 4.77890402e-01 9.70230281e-01 -1.10987067e+00
4.80954409e-01 1.37621665e+00 4.54717390e-02 7.18242884e-01
-8.22100043e-01 -5.24085820e-01 6.64672673e-01 4.12191182e-01
-1.01023257e+00 -7.70483255e-01 4.49739546e-01 -7.12671950e-02
1.49195087e+00 1.14673860e-01 1.09988675e-01 1.22681689e+00
5.77653408e-01 9.53819692e-01 1.07800889e+00 -6.21666193e-01
-2.90631521e-02 9.20205563e-02 1.87053382e-01 5.48257351e-01
-1.65374130e-01 1.89565808e-01 -7.44537115e-01 1.39457017e-01
4.06755060e-01 -1.04427963e-01 -3.84948641e-01 3.06116670e-01
-1.47285974e+00 5.23725986e-01 3.47950399e-01 3.05738539e-01
-7.04933703e-01 5.17056473e-02 4.70213890e-01 6.64505720e-01
4.49055254e-01 2.41640106e-01 -5.80253959e-01 -3.02566677e-01
-8.09704900e-01 -3.74880135e-01 4.93940860e-01 1.13913846e+00
5.97460866e-01 6.95278868e-02 -4.22976404e-01 6.51501298e-01
3.21409792e-01 8.83523822e-01 7.96764433e-01 -2.80033529e-01
9.39122915e-01 1.53873801e-01 1.18920378e-01 -4.18452889e-01
-1.16531819e-01 -5.85261345e-01 -6.89033687e-01 -1.80427298e-01
1.31984964e-01 -2.54671782e-01 -1.05784547e+00 1.80500686e+00
-1.02426559e-01 -1.26410753e-01 2.06867546e-01 1.15671682e+00
3.46862674e-01 1.10888243e+00 -9.79640484e-02 -6.74847305e-01
1.04219675e+00 -1.53968120e+00 -9.72271085e-01 -6.21199787e-01
4.85671222e-01 -1.11153233e+00 1.42280495e+00 1.08923607e-01
-1.24764764e+00 -8.73449385e-01 -8.55350137e-01 -6.81804791e-02
1.88670501e-01 2.37953007e-01 9.67446193e-02 1.41468793e-01
-1.03721976e+00 3.83174837e-01 -1.14841712e+00 -4.12445784e-01
2.74271835e-02 4.25323963e-01 5.47228903e-02 -1.73907995e-01
-1.31616652e+00 1.13299954e+00 9.40457955e-02 4.20299232e-01
-1.07272053e+00 -1.72557440e-02 -4.67221469e-01 1.36953652e-01
4.15973485e-01 -7.56127119e-01 1.67282522e+00 -1.50465679e+00
-1.69633460e+00 4.02766377e-01 -8.57771873e-01 -5.25686085e-01
4.96319532e-01 -5.68010926e-01 -3.36601019e-01 -1.86982200e-01
1.23524219e-01 6.31193638e-01 9.31738377e-01 -6.53090656e-01
-7.78790891e-01 -2.81781018e-01 -1.35198057e-01 5.54187715e-01
-1.79279193e-01 1.91876769e-01 -4.75990564e-01 -5.30029118e-01
-2.07189202e-01 -9.75130975e-01 -4.49926071e-02 -3.95148098e-01
-3.74543726e-01 -2.32211962e-01 6.99554622e-01 -7.61695743e-01
1.38510334e+00 -1.88172281e+00 4.94236082e-01 -3.40207458e-01
-3.43907446e-01 1.84466437e-01 -5.06989062e-01 3.92548829e-01
1.95203692e-01 1.25157952e-01 -7.64304847e-02 -2.91413933e-01
-3.51206452e-01 2.72280455e-01 -5.08375049e-01 2.97285825e-01
1.77741483e-01 1.16573381e+00 -8.05859208e-01 -3.62783939e-01
-1.07185580e-01 4.49481048e-02 -1.92198455e-01 4.80689108e-01
-5.08502722e-01 4.99473602e-01 -4.16576207e-01 4.60006237e-01
2.93956161e-01 -1.91250965e-01 4.64005247e-02 2.61358857e-01
-1.73144341e-01 9.38420951e-01 -3.42290789e-01 1.71641374e+00
-8.64584267e-01 7.02238560e-01 -4.32129353e-02 -6.33815348e-01
8.16480219e-01 5.22399604e-01 -2.31699347e-01 -1.19881070e+00
6.14347160e-02 4.58296716e-01 4.40238923e-01 -4.29722786e-01
3.29649538e-01 -1.11633614e-01 1.47477970e-01 8.89288247e-01
-1.51069194e-01 5.30848861e-01 -1.11711405e-01 -1.39984228e-02
7.00302958e-01 4.02888596e-01 4.12795544e-01 -2.57388711e-01
6.85341239e-01 2.86308266e-02 5.97305536e-01 6.70760274e-01
-1.65317148e-01 3.38591516e-01 2.70369351e-01 -4.71112728e-01
-9.35505331e-01 -6.57621980e-01 6.16803944e-01 1.56885672e+00
8.68450552e-02 5.62542044e-02 -7.38466620e-01 -9.56916511e-01
-5.01587331e-01 1.07867074e+00 -3.60701799e-01 -2.50721961e-01
-9.55199897e-01 -5.10164380e-01 3.56770515e-01 5.60959041e-01
3.13443571e-01 -1.53098488e+00 -6.22901320e-01 5.06269455e-01
-6.00290477e-01 -1.00049305e+00 -1.12596262e+00 2.84106791e-01
-8.39552581e-01 -4.35990959e-01 -7.45130002e-01 -9.40022945e-01
5.33556759e-01 2.44267344e-01 1.02028310e+00 7.33441785e-02
6.35476470e-01 -3.77938986e-01 -4.42704201e-01 -3.38523626e-01
-7.44025886e-01 6.27539933e-01 3.37104827e-01 -1.93915667e-03
3.71511698e-01 -2.34592870e-01 -3.14019859e-01 4.33884889e-01
-4.94458079e-01 4.34019089e-01 9.05061543e-01 9.65449393e-01
5.38216531e-01 -6.14349782e-01 8.00286889e-01 -6.58381641e-01
8.96037459e-01 -3.02985847e-01 -5.09608269e-01 5.19341111e-01
-8.28018010e-01 5.34332752e-01 1.08501244e+00 -7.00948238e-01
-1.17864347e+00 -2.66326427e-01 -1.25429124e-01 -3.66282821e-01
1.36736542e-01 6.43729866e-01 -3.12964141e-01 4.79509234e-01
4.46658254e-01 6.36924446e-01 -1.63119063e-01 -6.24720454e-01
2.04380140e-01 9.01473343e-01 4.19418693e-01 -4.14089769e-01
4.21629518e-01 -1.87384129e-01 -4.91564929e-01 -2.69539803e-01
-6.21756136e-01 -1.47044018e-01 -6.28479660e-01 7.10517764e-02
7.58071899e-01 -7.73149848e-01 -3.09302986e-01 4.27394807e-01
-1.57440591e+00 -5.57044327e-01 3.07574481e-01 5.40473998e-01
-4.81204063e-01 2.11094081e-01 -7.84079075e-01 -7.52511859e-01
-7.29092002e-01 -1.58716404e+00 9.58000124e-01 5.52655049e-02
-2.75222570e-01 -7.07130015e-01 -1.95792258e-01 1.43647537e-01
6.10201001e-01 -6.88022256e-01 1.20177662e+00 -7.92000473e-01
-7.29264081e-01 3.98948014e-01 -8.10267106e-02 1.97914824e-01
1.58403784e-01 -3.47740918e-01 -8.07960451e-01 -3.80620241e-01
7.22585618e-02 -8.84924456e-02 8.29571545e-01 2.78208435e-01
3.70107800e-01 -6.06525481e-01 -3.06359708e-01 4.15839106e-01
1.16431403e+00 6.12524152e-01 3.43602031e-01 1.71148896e-01
7.53722191e-01 3.64386141e-01 8.91756237e-01 1.92962885e-02
3.05924982e-01 8.76446068e-01 2.53552377e-01 2.39095255e-03
-1.73540667e-01 -3.44925433e-01 1.08403587e+00 1.29754591e+00
2.37310633e-01 -7.37679660e-01 -9.75001276e-01 5.92400372e-01
-2.02381063e+00 -5.95519960e-01 8.55171978e-02 2.11109924e+00
8.44862163e-01 4.02198762e-01 -2.38899857e-01 -4.08383369e-01
8.37203205e-01 8.95150900e-02 -7.27555335e-01 -9.64260936e-01
7.90716112e-02 3.92279737e-02 3.76679361e-01 9.53136206e-01
-5.55509567e-01 1.40509999e+00 5.30311489e+00 7.73401201e-01
-1.62987065e+00 3.69205505e-01 6.48412049e-01 -1.43261164e-01
-4.10994470e-01 -8.03561658e-02 -8.51778567e-01 5.37124693e-01
1.32964802e+00 -4.94404882e-01 7.54411817e-01 4.89603281e-01
4.75346178e-01 7.05948845e-03 -1.37829340e+00 6.61282241e-01
-9.71705653e-03 -9.46472406e-01 2.32346013e-01 -2.00948436e-02
6.45778239e-01 3.89725685e-01 9.27830637e-02 4.23187345e-01
1.63697407e-01 -9.20899630e-01 8.82000983e-01 5.57459176e-01
8.64939690e-01 -8.65156889e-01 1.00189996e+00 9.51940775e-01
-9.12670732e-01 -1.48322824e-02 -3.67419034e-01 -1.80701032e-01
3.50456119e-01 1.58906743e-01 -1.07142663e+00 5.20792961e-01
2.05091238e-01 5.07443964e-01 -3.83362770e-01 4.88083750e-01
-4.96228963e-01 8.87103379e-01 -4.19127829e-02 -2.64516592e-01
5.51627755e-01 -5.82142510e-02 5.35480022e-01 1.14472139e+00
5.73634982e-01 -9.87712741e-02 1.81051016e-01 6.45042837e-01
-2.65604347e-01 3.35841984e-01 -4.54473495e-01 -2.50847280e-01
4.15150106e-01 6.69356108e-01 -3.76487136e-01 -4.65176344e-01
-5.41371644e-01 1.38460636e+00 4.82603073e-01 5.93430758e-01
-7.49996185e-01 -5.84583580e-02 5.34273565e-01 -3.72056179e-02
1.70989856e-01 -9.43700522e-02 -5.19108176e-01 -1.09213972e+00
2.21435755e-01 -1.19054341e+00 5.70721738e-02 -8.20592940e-01
-9.83922541e-01 1.21016872e+00 -4.43500936e-01 -1.05178452e+00
-5.97223282e-01 -1.07495636e-01 -5.11592209e-01 1.32546353e+00
-1.53060257e+00 -9.85619009e-01 3.72355580e-01 1.82348415e-01
1.29280138e+00 -3.01894099e-01 7.66933024e-01 7.95649290e-02
-5.64445436e-01 6.63139403e-01 -6.10060357e-02 -5.59113733e-02
9.87154305e-01 -8.06854069e-01 1.01185727e+00 1.27039433e+00
1.64318129e-01 7.13694572e-01 6.12895250e-01 -8.29867542e-01
-1.31017518e+00 -1.03172123e+00 1.47142696e+00 -3.04520130e-01
2.95002341e-01 -2.40283608e-01 -9.45242763e-01 7.87470102e-01
8.29823256e-01 -3.42769772e-01 2.08301857e-01 -1.18213310e-03
-6.89087734e-02 -6.70660883e-02 -7.13173330e-01 8.58809233e-01
9.89035666e-01 -5.77741921e-01 -6.81019247e-01 1.45123407e-01
1.10332286e+00 -2.40605667e-01 -7.19158053e-02 1.72031820e-01
4.60638702e-01 -6.14877462e-01 2.93165326e-01 -5.71876466e-01
4.89857763e-01 -2.80906409e-01 -1.42464101e-01 -1.60696101e+00
-4.17080849e-01 -8.52744460e-01 -6.36603609e-02 9.05268192e-01
9.08784926e-01 -5.20195842e-01 1.57341897e-01 2.02618062e-01
-3.69494796e-01 -8.83245051e-01 -1.03823256e+00 -6.38221323e-01
1.36209831e-01 -2.97228426e-01 6.94471002e-01 5.44604778e-01
3.48334573e-02 8.71140838e-01 -6.66024148e-01 3.55135240e-02
4.40998748e-02 3.55022490e-01 5.04493058e-01 -6.11040175e-01
-4.04616356e-01 -4.16283548e-01 2.94109285e-01 -1.63886535e+00
2.51215786e-01 -9.66437399e-01 4.01250839e-01 -1.58237696e+00
2.10904121e-01 -9.21815075e-03 -4.24313664e-01 5.43587446e-01
-5.58397830e-01 -2.64872134e-01 3.29974651e-01 2.84878314e-01
-6.36081517e-01 7.31739938e-01 1.40515804e+00 -1.74333125e-01
-2.84495592e-01 4.73570302e-02 -6.30550981e-01 2.62607247e-01
6.56880140e-01 -5.84931433e-01 -4.83668864e-01 -1.13612795e+00
9.74415243e-02 3.81786138e-01 -3.03807855e-01 -6.26771688e-01
3.67001921e-01 -3.75511289e-01 8.74850824e-02 -5.47070324e-01
3.24334949e-02 -6.93300247e-01 -2.52613008e-01 4.67671782e-01
-7.15544343e-01 8.25742304e-01 1.36364654e-01 4.26523775e-01
-1.83432490e-01 9.42893978e-03 6.54901803e-01 -2.25684419e-01
-4.60371345e-01 2.35495538e-01 -7.09921002e-01 -1.47523984e-01
6.84628665e-01 1.52001724e-01 -2.95689195e-01 -6.25583112e-01
-5.80257237e-01 3.94333899e-01 4.77119476e-01 8.97521675e-01
6.76886201e-01 -1.19547379e+00 -9.12531316e-01 3.93127054e-01
-9.74519644e-03 -3.59469146e-01 -6.15657791e-02 9.58119035e-01
-2.12268889e-01 8.15694034e-01 -2.72543252e-01 -4.08143044e-01
-1.36641812e+00 8.39347482e-01 2.97577232e-01 -3.96198452e-01
-2.80479580e-01 8.16956818e-01 3.12605441e-01 -2.46019155e-01
1.68670580e-01 -4.82675046e-01 1.80058610e-02 -1.89685613e-01
4.99014705e-01 1.57588542e-01 1.94768295e-01 -6.28016174e-01
-4.74407166e-01 2.64613003e-01 -4.93964523e-01 -5.93354285e-01
9.20012236e-01 -5.85043430e-01 -2.18854863e-02 5.20045817e-01
9.21551585e-01 -2.59652901e-02 -9.44431663e-01 -6.35985374e-01
9.65681598e-02 -2.53144979e-01 4.92938459e-02 -1.14924514e+00
-7.83151865e-01 1.27808833e+00 3.63275588e-01 -4.41117287e-01
1.13220501e+00 -2.51655132e-01 1.22107685e+00 5.16457498e-01
5.64383507e-01 -1.08293974e+00 -1.09959878e-01 8.94965529e-01
8.85308444e-01 -1.20067430e+00 -5.30747771e-01 -8.08223039e-02
-9.53141451e-01 9.98917937e-01 8.74551773e-01 3.66061389e-01
5.13415299e-02 2.22949967e-01 4.77903485e-01 3.62710029e-01
-1.55202591e+00 -9.92687941e-02 3.50540698e-01 2.72379905e-01
6.54302537e-01 1.65389970e-01 -4.82783973e-01 5.75841486e-01
-1.28439385e-02 4.01281565e-03 3.86723757e-01 7.14968622e-01
-4.89793152e-01 -1.37428164e+00 -3.02408725e-01 2.52162814e-01
-5.39485812e-01 -4.82453614e-01 -6.21207595e-01 2.74743527e-01
-1.73845872e-01 1.09523237e+00 -7.98915252e-02 -3.02556485e-01
1.72719985e-01 5.08790970e-01 2.92813331e-01 -6.49934173e-01
-7.67208993e-01 4.50531244e-01 1.36551350e-01 -3.68101060e-01
-2.51576491e-02 -6.34320140e-01 -1.29683697e+00 -1.86685100e-01
-3.67255509e-01 1.19689062e-01 3.69420052e-01 1.16642463e+00
5.16661048e-01 5.27726233e-01 6.67501926e-01 -3.17858368e-01
-7.36894667e-01 -1.36964273e+00 2.57623166e-01 3.21166217e-03
6.09166682e-01 -1.04455546e-01 1.00151218e-01 -3.47687975e-02]
|
[11.823163986206055, 9.984027862548828]
|
4d8a3e1d-2ad5-4071-b6b6-2f40cfefef72
|
speech-text-dialog-pre-training-for-spoken
|
2305.11579
| null |
https://arxiv.org/abs/2305.11579v2
|
https://arxiv.org/pdf/2305.11579v2.pdf
|
Speech-Text Dialog Pre-training for Spoken Dialog Understanding with Explicit Cross-Modal Alignment
|
Recently, speech-text pre-training methods have shown remarkable success in many speech and natural language processing tasks. However, most previous pre-trained models are usually tailored for one or two specific tasks, but fail to conquer a wide range of speech-text tasks. In addition, existing speech-text pre-training methods fail to explore the contextual information within a dialogue to enrich utterance representations. In this paper, we propose Speech-text dialog Pre-training for spoken dialog understanding with ExpliCiT cRoss-Modal Alignment (SPECTRA), which is the first-ever speech-text dialog pre-training model. Concretely, to consider the temporality of speech modality, we design a novel temporal position prediction task to capture the speech-text alignment. This pre-training task aims to predict the start and end time of each textual word in the corresponding speech waveform. In addition, to learn the characteristics of spoken dialogs, we generalize a response selection task from textual dialog pre-training to speech-text dialog pre-training scenarios. Experimental results on four different downstream speech-text tasks demonstrate the superiority of SPECTRA in learning speech-text alignment and multi-turn dialog context.
|
['Yongbin Li', 'Fei Huang', 'Chao Wang', 'Wentao Ma', 'Yuchuan Wu', 'Min Yang', 'Ting-En Lin', 'Haoyu Gao', 'Tianshu Yu']
|
2023-05-19
| null | null | null | null |
['multimodal-sentiment-analysis', 'multimodal-intent-recognition', 'emotion-recognition-in-conversation', 'multimodal-sentiment-analysis']
|
['computer-vision', 'miscellaneous', 'natural-language-processing', 'natural-language-processing']
|
[ 5.41193426e-01 2.63224989e-01 -2.31947780e-01 -9.49285448e-01
-6.07863367e-01 -4.32068080e-01 7.85212994e-01 1.70372918e-01
-2.26573452e-01 4.64913636e-01 7.71714866e-01 -7.30675340e-01
3.08124218e-02 -1.75858229e-01 -1.08416900e-01 -2.92683899e-01
2.65441984e-01 7.35195875e-01 1.51376233e-01 -7.15243697e-01
9.87565666e-02 6.14969209e-02 -1.03176689e+00 6.48807585e-01
7.49930918e-01 6.70821369e-01 6.80999279e-01 7.10102260e-01
-5.85265219e-01 7.38146245e-01 -6.75948083e-01 -6.45257905e-02
-2.50331610e-01 -8.22505295e-01 -1.09506023e+00 3.30652833e-01
-1.50250830e-02 -2.56089061e-01 -4.69810128e-01 6.88269854e-01
5.26004255e-01 5.08511126e-01 4.91840065e-01 -1.09144115e+00
-3.18439186e-01 1.16024137e+00 -9.41344425e-02 4.70752656e-01
6.03586495e-01 -7.33547956e-02 1.13362610e+00 -9.36995447e-01
2.63923675e-01 1.59773886e+00 3.85277182e-01 8.82208526e-01
-9.77814496e-01 -4.04928654e-01 6.32155001e-01 1.68206617e-01
-7.36434937e-01 -7.42438495e-01 8.45878839e-01 -4.28676814e-01
1.14276576e+00 3.21734309e-01 1.95554748e-01 1.24869990e+00
2.46412512e-02 1.00976908e+00 8.34090590e-01 -5.61896324e-01
-9.29511264e-02 1.01487905e-01 3.08465332e-01 1.61882788e-01
-1.05238044e+00 8.65850449e-02 -5.86516023e-01 -2.40532123e-02
3.17748874e-01 -8.86630267e-02 -1.55696064e-01 1.73238784e-01
-1.29851651e+00 8.82933915e-01 -1.43786930e-02 4.78176057e-01
-1.88921362e-01 -5.45728922e-01 6.98520660e-01 5.43766737e-01
4.22503531e-01 1.94773808e-01 -6.16425991e-01 -9.78099778e-02
-5.29992104e-01 4.48467135e-02 9.84585047e-01 9.43325698e-01
5.48200190e-01 2.20926553e-01 -5.61919272e-01 1.19150388e+00
3.18529576e-01 4.28470105e-01 7.98956931e-01 -5.01772344e-01
7.85678387e-01 4.13362950e-01 -1.98699236e-01 -5.03613472e-01
-6.28989518e-01 1.95928067e-01 -7.83407688e-01 -4.81831133e-01
3.40513378e-01 -4.47526544e-01 -7.21551180e-01 1.74456275e+00
3.78143251e-01 -5.44908717e-02 5.58111548e-01 7.05282927e-01
9.04533148e-01 1.18263376e+00 1.32663324e-01 -7.76243925e-01
1.53466237e+00 -1.23999989e+00 -1.11700583e+00 -6.93648517e-01
5.23710966e-01 -1.03724277e+00 1.30734622e+00 9.26727988e-03
-9.15635347e-01 -7.29867995e-01 -6.16085768e-01 9.68475826e-03
-1.82265297e-01 1.59384891e-01 1.66148871e-01 2.98574090e-01
-5.19738078e-01 5.96321784e-02 -4.89133567e-01 -3.86014909e-01
-4.39534217e-01 2.81548053e-01 5.27345464e-02 3.31974953e-01
-1.59288955e+00 1.00445545e+00 3.60281318e-01 3.53757888e-02
-7.27047563e-01 -5.47445536e-01 -8.81119013e-01 1.11584842e-01
7.18821764e-01 -3.66031617e-01 1.96237314e+00 -9.59503770e-01
-2.14708591e+00 6.04592741e-01 -5.50699472e-01 -4.22600865e-01
6.53911233e-02 -1.85594231e-01 -6.00189328e-01 -9.56786796e-02
-1.84402570e-01 4.52659816e-01 9.00679350e-01 -1.10427439e+00
-7.50910997e-01 -2.23460495e-01 -5.98591231e-02 6.73349321e-01
-3.17168772e-01 2.62404472e-01 -5.90231046e-02 -7.10820675e-01
4.11121408e-03 -8.83624732e-01 -2.38841012e-01 -8.32326710e-01
-4.86785471e-01 -6.18242323e-01 9.34219778e-01 -5.14757097e-01
1.45118701e+00 -2.21630049e+00 2.93875754e-01 -2.35224262e-01
-2.32599691e-01 2.62419939e-01 -2.13234484e-01 9.52176511e-01
-4.17406261e-02 -3.11898798e-01 -1.52192771e-01 -7.34347641e-01
2.42605850e-01 5.47191441e-01 -8.84538293e-01 -2.22969502e-02
8.25351924e-02 6.74585938e-01 -7.58709908e-01 -4.44207519e-01
3.89632583e-01 2.13593822e-02 -3.09878856e-01 7.02608526e-01
-7.23685026e-01 1.07482290e+00 -4.24723893e-01 9.77417678e-02
8.18450600e-02 -5.11054657e-02 4.12728995e-01 2.09552720e-02
-1.93507999e-01 9.33220446e-01 -6.46146476e-01 1.69777799e+00
-7.06347704e-01 5.61139882e-01 2.18126088e-01 -1.08707333e+00
9.98285353e-01 7.15338051e-01 5.02169669e-01 -7.50396729e-01
3.00220996e-01 3.73618267e-02 4.81280237e-01 -6.99576855e-01
7.09840417e-01 -2.86813140e-01 -2.91870028e-01 6.14485145e-01
5.71153611e-02 -3.69859606e-01 6.36517555e-02 1.22935489e-01
4.96223330e-01 -4.16166306e-01 3.61430079e-01 6.37654960e-02
8.53221416e-01 -2.10409522e-01 3.34271967e-01 3.10650706e-01
-1.49014905e-01 4.53984201e-01 2.73928612e-01 -8.81047472e-02
-8.89985740e-01 -6.98176265e-01 -3.45912129e-02 1.92781246e+00
1.69936880e-01 -5.58776975e-01 -6.33334517e-01 -6.17961168e-01
-3.77114743e-01 9.49394107e-01 -1.95089921e-01 -1.33119881e-01
-9.76574898e-01 -3.06565434e-01 5.54084301e-01 4.07120764e-01
1.83341950e-01 -1.38908446e+00 -1.74357742e-01 3.74226928e-01
-5.07504821e-01 -1.52364385e+00 -1.06803012e+00 2.61484146e-01
-5.39212286e-01 -6.92265511e-01 -5.46223700e-01 -1.25722826e+00
3.46028507e-01 1.91108570e-01 7.31102824e-01 -2.27429569e-01
1.81773558e-01 4.03103679e-01 -6.30226791e-01 -3.72589231e-01
-9.65971351e-01 2.43431076e-01 1.83669508e-01 9.06864256e-02
3.58609349e-01 -2.78016210e-01 -2.20293194e-01 7.32608378e-01
-7.32793450e-01 9.46350619e-02 5.30215561e-01 1.11711931e+00
1.81623951e-01 -2.40987971e-01 1.03232598e+00 -6.02577865e-01
1.10589933e+00 -4.08991784e-01 -1.69581175e-01 4.50282007e-01
-3.23750287e-01 3.00651602e-02 8.15628827e-01 -9.42847669e-01
-1.37465036e+00 1.62706196e-01 -5.39188802e-01 -2.58318484e-01
-4.10765380e-01 8.81746113e-01 -2.81330347e-01 7.76586652e-01
3.52951169e-01 5.28710604e-01 1.47443369e-01 -3.40784371e-01
4.47898537e-01 8.90649557e-01 6.42210901e-01 -7.30497897e-01
3.99318486e-01 -1.25555262e-01 -4.87579793e-01 -1.25670457e+00
-9.28778946e-01 -7.87722170e-01 -7.95476258e-01 -1.74625106e-02
9.69906867e-01 -4.84185308e-01 -7.48563707e-01 2.91507930e-01
-1.45637357e+00 -6.54440284e-01 -4.32165042e-02 6.09450996e-01
-5.88691831e-01 6.33757591e-01 -4.62596774e-01 -1.10658157e+00
-3.00152987e-01 -1.23456681e+00 1.05954373e+00 8.67143199e-02
-4.58540410e-01 -1.10067797e+00 1.14734970e-01 3.50925654e-01
3.55643123e-01 -5.89444578e-01 8.75558376e-01 -1.43140197e+00
1.42794121e-02 2.03823388e-01 2.24920854e-01 2.40703478e-01
4.69925463e-01 -2.95504153e-01 -9.62267578e-01 -1.41276062e-01
3.56679678e-01 -4.54635113e-01 3.85665506e-01 3.71565491e-01
6.75403655e-01 -6.75803304e-01 -2.86327869e-01 4.99874130e-02
2.42333949e-01 7.50863075e-01 2.07202941e-01 -1.10284247e-01
3.94708186e-01 1.16528594e+00 8.89657795e-01 4.04990822e-01
4.81273264e-01 8.81547332e-01 1.16215765e-01 2.83089206e-02
1.29736707e-01 -2.47810051e-01 5.77658713e-01 1.37037325e+00
6.56703234e-01 -5.27187884e-01 -9.40925181e-01 4.65736598e-01
-1.95419300e+00 -9.09221470e-01 -3.21159214e-02 1.86492813e+00
1.19791317e+00 1.64117545e-01 3.55326623e-01 -1.54991567e-01
7.37647891e-01 3.68753493e-01 -4.68225092e-01 -2.66853392e-01
8.71215835e-02 -1.09001622e-01 -2.24978477e-01 8.16806197e-01
-1.01573861e+00 1.23046005e+00 5.58951569e+00 6.86737537e-01
-1.23921001e+00 1.07428521e-01 3.63297731e-01 3.27328354e-01
-2.24500135e-01 8.00896958e-02 -1.05934691e+00 3.63914430e-01
1.22913146e+00 -4.20003414e-01 2.47456238e-01 7.19657958e-01
3.96865606e-01 1.89651906e-01 -1.53975844e+00 7.89871275e-01
4.78860270e-03 -9.33336020e-01 8.93768221e-02 -3.44260097e-01
3.75947267e-01 -2.80335635e-01 6.60496438e-03 7.01338410e-01
2.55594432e-01 -1.07406759e+00 4.61078912e-01 5.01899086e-02
5.29370546e-01 -4.31818515e-01 5.60579360e-01 8.07107508e-01
-1.30809104e+00 -1.31800666e-01 -8.98755416e-02 4.30101119e-02
8.18632483e-01 4.84153852e-02 -1.46433604e+00 5.13861358e-01
9.25009772e-02 3.93371046e-01 2.85644326e-02 4.94468302e-01
-1.05045021e-01 6.67656600e-01 -2.26036817e-01 -2.92464167e-01
4.39596534e-01 -4.12973091e-02 7.12622941e-01 1.45621777e+00
1.96208730e-01 5.32082975e-01 7.33362317e-01 3.45716208e-01
7.99854696e-02 3.06781709e-01 -4.46023941e-01 -3.99583727e-01
7.95938671e-01 8.29259157e-01 -4.03933823e-01 -2.50739962e-01
-4.63394552e-01 8.31415892e-01 -8.68789703e-02 3.45630735e-01
-6.17790639e-01 -2.53493726e-01 6.46568358e-01 -2.04647556e-01
2.65039913e-02 -3.95257354e-01 7.09794869e-04 -9.38642383e-01
-2.22554028e-01 -1.14183128e+00 2.92998433e-01 -4.76664871e-01
-1.45269001e+00 1.00108814e+00 2.40779921e-01 -1.08748329e+00
-8.43613684e-01 -5.10841191e-01 -9.03889537e-01 9.55099583e-01
-1.18672740e+00 -1.19520152e+00 2.02719457e-02 7.00156689e-01
1.56938195e+00 -4.81148452e-01 7.86676109e-01 1.01378441e-01
-5.06365955e-01 5.13193548e-01 -1.68423861e-01 2.18438923e-01
9.64765370e-01 -1.12425447e+00 4.87175286e-01 5.76287508e-01
1.17004424e-01 7.30597138e-01 8.14017475e-01 -6.18843198e-01
-1.21642518e+00 -7.54756689e-01 1.05149567e+00 -2.80258060e-01
9.33061361e-01 -4.81781512e-01 -1.13931990e+00 9.46484268e-01
8.20746839e-01 -6.60604835e-01 6.86461389e-01 2.50479996e-01
4.50629070e-02 -1.43953050e-02 -3.73739630e-01 7.73554981e-01
6.42744660e-01 -9.08510625e-01 -1.18965352e+00 4.58331198e-01
1.18673539e+00 -8.02415371e-01 -6.26585841e-01 5.07756233e-01
1.06800802e-01 -4.99115288e-01 9.10363138e-01 -8.28628719e-01
2.93568730e-01 2.55857483e-02 -2.54569501e-01 -1.54260075e+00
1.82756454e-01 -1.15533841e+00 1.18451506e-01 1.46303368e+00
4.56001282e-01 -3.82872045e-01 4.31033254e-01 2.55910456e-01
-7.58646846e-01 -4.75929856e-01 -8.77146423e-01 -5.26333094e-01
9.48489532e-02 -3.56424123e-01 4.57585365e-01 9.71287847e-01
7.30214715e-01 1.23285997e+00 -5.68929553e-01 2.33624384e-01
-7.29824677e-02 1.96405083e-01 1.00298262e+00 -1.04253650e+00
-3.27876002e-01 -5.83808482e-01 4.17700708e-01 -2.02962232e+00
6.45281732e-01 -7.54370987e-01 7.08564937e-01 -1.38091707e+00
-3.31935138e-01 -3.14190954e-01 3.15404475e-01 2.84180760e-01
-3.50680202e-01 -9.42659140e-01 7.80472904e-02 2.06685618e-01
-3.52980822e-01 1.02243221e+00 1.28164566e+00 -2.22365513e-01
-6.45076811e-01 5.36025047e-01 -1.78031370e-01 5.64094067e-01
7.28082418e-01 -2.15405226e-01 -9.50499415e-01 -1.75058395e-01
-3.18097562e-01 7.95618415e-01 -1.08130381e-01 -4.89617944e-01
6.89264178e-01 -6.26377761e-01 -3.82242441e-01 -1.03142869e+00
6.29281521e-01 -7.96177208e-01 -4.14212137e-01 1.15646750e-01
-9.21267748e-01 -1.94323864e-02 3.45184922e-01 5.44338167e-01
-4.89216596e-01 -5.25183618e-01 6.46375835e-01 -7.70697836e-03
-7.28901744e-01 1.63025066e-01 -8.12863111e-01 2.07454070e-01
8.96818101e-01 1.05348498e-01 -3.99312556e-01 -7.12701976e-01
-1.04974234e+00 6.44519091e-01 -2.98189193e-01 7.47478306e-01
5.83609104e-01 -9.31316674e-01 -6.17430925e-01 1.84296489e-01
1.27421161e-02 9.38160121e-02 3.45768213e-01 8.41792166e-01
3.41280311e-01 6.90673709e-01 5.48385531e-02 -8.16526651e-01
-1.57056737e+00 5.53828001e-01 3.27428848e-01 -3.06372553e-01
-6.47565782e-01 7.16500998e-01 6.05981231e-01 -7.58054197e-01
7.26517379e-01 -5.19376993e-01 -3.79313290e-01 7.39677846e-02
4.73856121e-01 -1.30005002e-01 -1.25351161e-01 -7.38854110e-01
-1.26576677e-01 1.58837497e-01 -2.74476588e-01 -5.13227284e-01
9.38412011e-01 -6.65806651e-01 1.53914139e-01 8.48402560e-01
8.80467296e-01 1.02144361e-01 -1.10440278e+00 -6.29672825e-01
4.40914273e-01 8.40616524e-02 -4.61344063e-01 -6.44857764e-01
-4.78080601e-01 1.15335512e+00 -3.32880276e-03 5.96093953e-01
8.21562290e-01 1.60712585e-01 1.03343451e+00 7.72841513e-01
-9.82110798e-02 -1.06060743e+00 6.67843640e-01 1.37580872e+00
1.15263319e+00 -1.25312448e+00 -4.54682887e-01 -4.32312787e-01
-1.29055011e+00 1.34886253e+00 1.02518880e+00 5.50279498e-01
6.42557681e-01 2.47878730e-01 3.30490649e-01 -6.51841760e-02
-9.19565439e-01 -2.32163295e-01 3.74305069e-01 4.41374749e-01
5.87524831e-01 -2.90838897e-01 7.20527321e-02 6.44087553e-01
-5.03139436e-01 -5.56971788e-01 2.41304800e-01 8.22238922e-01
-7.01593578e-01 -1.30098319e+00 -2.71642625e-01 -5.60631417e-02
-1.85410157e-01 -2.14713931e-01 -7.83516943e-01 5.22744477e-01
-5.11567056e-01 1.32987571e+00 1.27064228e-01 -5.53258240e-01
5.25478840e-01 5.57911634e-01 1.60125662e-02 -9.44899619e-01
-7.89105177e-01 5.92367768e-01 3.49562675e-01 1.60364117e-02
-4.69451308e-01 -4.70584542e-01 -1.60699189e+00 -9.06252265e-02
-3.59601289e-01 5.87999761e-01 4.23850477e-01 1.43627644e+00
1.21892067e-02 7.81530440e-01 7.89929390e-01 -8.06463718e-01
-9.29020405e-01 -1.44208717e+00 -1.47419944e-01 2.88910359e-01
5.36118686e-01 -4.86877382e-01 -2.35715613e-01 2.86036283e-01]
|
[12.784989356994629, 7.76558780670166]
|
58cc8166-ae91-4207-9133-89f5a092d839
|
multi-input-architecture-and-disentangled
|
2111.01710
| null |
https://arxiv.org/abs/2111.01710v1
|
https://arxiv.org/pdf/2111.01710v1.pdf
|
Multi-input Architecture and Disentangled Representation Learning for Multi-dimensional Modeling of Music Similarity
|
In the context of music information retrieval, similarity-based approaches are useful for a variety of tasks that benefit from a query-by-example scenario. Music however, naturally decomposes into a set of semantically meaningful factors of variation. Current representation learning strategies pursue the disentanglement of such factors from deep representations, resulting in highly interpretable models. This allows the modeling of music similarity perception, which is highly subjective and multi-dimensional. While the focus of prior work is on metadata driven notions of similarity, we suggest to directly model the human notion of multi-dimensional music similarity. To achieve this, we propose a multi-input deep neural network architecture, which simultaneously processes mel-spectrogram, CENS-chromagram and tempogram in order to extract informative features for the different disentangled musical dimensions: genre, mood, instrument, era, tempo, and key. We evaluated the proposed music similarity approach using a triplet prediction task and found that the proposed multi-input architecture outperforms a state of the art method. Furthermore, we present a novel multi-dimensional analysis in order to evaluate the influence of each disentangled dimension on the perception of music similarity.
|
['Hanna Lukashevich', 'Jakob Abeßer', 'Sebastian Ribecky']
|
2021-11-02
| null | null | null | null |
['music-information-retrieval']
|
['music']
|
[ 2.84955978e-01 -3.27230096e-01 -5.64442724e-02 -1.43381327e-01
-8.44680667e-01 -7.56488383e-01 8.61713171e-01 3.50138307e-01
-2.59491056e-01 -4.45144325e-02 7.61794388e-01 1.50718525e-01
-7.58794606e-01 -5.04599512e-01 -2.00140193e-01 -5.25997996e-01
9.65455994e-02 3.10208082e-01 -4.19636816e-01 -2.91065216e-01
5.07094204e-01 2.72050947e-01 -1.95246005e+00 4.58922178e-01
4.28978175e-01 1.09982121e+00 -2.12617647e-02 5.50863981e-01
1.97397783e-01 4.19439256e-01 -5.95357358e-01 -3.77125829e-01
4.55834717e-01 -5.98768353e-01 -5.76672196e-01 -2.03024492e-01
6.08895004e-01 -2.61463635e-02 -1.73380747e-01 9.18523550e-01
6.53339684e-01 3.07451010e-01 8.20681453e-01 -1.01482987e+00
-7.51836658e-01 9.36748564e-01 -2.20790640e-01 1.47459805e-01
5.30563116e-01 -8.06660429e-02 1.83084524e+00 -7.33889163e-01
2.01752484e-01 1.09462178e+00 5.14684498e-01 4.55793999e-02
-1.41697264e+00 -6.31299257e-01 -2.60675162e-01 3.96445006e-01
-1.29773891e+00 -4.24680799e-01 1.15179324e+00 -6.42485142e-01
5.42563796e-01 7.40835428e-01 7.52848685e-01 1.20700097e+00
-5.07962331e-02 7.21048534e-01 1.09683573e+00 -5.29383361e-01
1.18189998e-01 -1.21779494e-01 4.07916084e-02 1.78799510e-01
5.99882789e-02 2.61646509e-01 -9.34750855e-01 -2.22185478e-01
7.42425323e-01 7.96445608e-02 -2.36341968e-01 -5.36710739e-01
-1.67403531e+00 7.00807035e-01 4.52688813e-01 5.99083245e-01
-4.01927799e-01 4.15353402e-02 5.37957132e-01 2.92310357e-01
8.11391473e-02 1.11164749e+00 -2.16273457e-01 -2.15936884e-01
-1.11965263e+00 4.57690597e-01 6.13180816e-01 4.40475136e-01
4.63948786e-01 1.58113554e-01 -5.06623566e-01 8.03602338e-01
4.11611170e-01 2.30120510e-01 9.44299340e-01 -1.07563698e+00
1.66753799e-01 4.92450505e-01 -7.41572753e-02 -1.23671973e+00
-5.06723642e-01 -7.44971335e-01 -8.65985394e-01 1.85276181e-01
3.51020545e-01 4.71081942e-01 -3.30856323e-01 1.99774218e+00
-1.29775003e-01 1.77955478e-01 -8.34109187e-02 1.16140592e+00
6.26423120e-01 1.47764847e-01 -9.55041051e-02 -6.07381314e-02
1.39875972e+00 -6.49096847e-01 -6.28494680e-01 1.06298581e-01
2.18191832e-01 -8.82462323e-01 1.26863205e+00 5.39627314e-01
-1.14541137e+00 -9.14965332e-01 -1.32802749e+00 -2.49858662e-01
-3.59801501e-01 2.60693908e-01 6.90321386e-01 4.03864384e-01
-6.67917550e-01 9.48356807e-01 -4.08580095e-01 -1.51955992e-01
-5.15431073e-03 2.89780170e-01 -2.32969314e-01 4.77183521e-01
-1.21894383e+00 7.76432633e-01 4.91451979e-01 -2.02481896e-02
-4.95527625e-01 -5.87624490e-01 -6.64404154e-01 4.41384524e-01
4.90048937e-02 -1.00244689e+00 1.08620417e+00 -8.93158078e-01
-1.49526894e+00 8.01840842e-01 2.09734842e-01 -1.79045260e-01
2.29366440e-02 -3.24081331e-01 -4.43471104e-01 -6.00397736e-02
-5.26215099e-02 2.67462373e-01 9.91121471e-01 -9.76813018e-01
-2.01529086e-01 -5.92189670e-01 1.17695153e-01 4.29000497e-01
-4.83224213e-01 -6.73117563e-02 -3.85292247e-02 -1.06616092e+00
3.27945590e-01 -1.03518999e+00 1.70291871e-01 -1.84797391e-01
-4.14559960e-01 -2.92033181e-02 1.79547875e-04 -4.67158139e-01
1.42370331e+00 -2.32822323e+00 7.06491649e-01 1.14296615e-01
1.96509123e-01 7.49395054e-04 -1.88472226e-01 5.06484210e-01
-2.97548145e-01 -7.29552433e-02 9.88828912e-02 -3.90151858e-01
5.33104420e-01 -1.43708453e-01 -5.12098610e-01 2.65131384e-01
-1.37844846e-01 7.34541237e-01 -7.42027700e-01 -1.83682457e-01
1.22497864e-01 5.02666771e-01 -7.70202100e-01 2.40277752e-01
-1.16434664e-01 4.02036697e-01 -2.45290026e-01 4.70769852e-01
1.71191394e-01 -8.78194571e-02 1.91471845e-01 -5.53947151e-01
2.93208417e-02 5.83483934e-01 -1.24681818e+00 2.35898089e+00
-4.42443907e-01 6.04176462e-01 -3.84596348e-01 -6.39136672e-01
1.10224307e+00 4.41759348e-01 5.95910907e-01 -6.16968691e-01
2.02508017e-01 8.20782855e-02 3.13305140e-01 -2.85978436e-01
9.16345894e-01 -3.12756449e-01 -3.69176745e-01 7.53175318e-01
1.65432170e-01 -1.07958009e-02 -7.97119960e-02 -7.66324475e-02
7.52431214e-01 2.05803052e-01 5.89745522e-01 -9.12656561e-02
3.78969610e-01 -5.70876837e-01 3.57683867e-01 4.81308669e-01
-3.90605405e-02 7.47515082e-01 3.00988048e-01 -3.79795879e-01
-8.42380524e-01 -1.20811701e+00 -2.77437605e-02 1.40444613e+00
-1.51846772e-02 -7.66606748e-01 -2.88983107e-01 -3.88500243e-02
1.97072253e-02 7.55778015e-01 -5.71539521e-01 -3.43702465e-01
-2.73029327e-01 -3.88802022e-01 7.25188971e-01 2.99633920e-01
-1.03319801e-01 -1.10401237e+00 -6.13800943e-01 1.41774788e-01
-1.83438256e-01 -7.24443614e-01 -4.73575443e-01 1.79438770e-01
-6.61652625e-01 -9.20484841e-01 -4.52642918e-01 -3.87150526e-01
-3.33767354e-01 1.14202693e-01 1.22485864e+00 -3.63030761e-01
-3.12480003e-01 3.56034547e-01 -2.66537309e-01 -3.60028088e-01
-2.66292721e-01 8.38824138e-02 2.78825074e-01 2.27367789e-01
5.06561756e-01 -1.07048965e+00 -6.87221348e-01 5.13311438e-02
-1.03811383e+00 3.93501855e-02 6.88572407e-01 6.75079823e-01
5.90747952e-01 -3.74101222e-01 6.98616087e-01 -3.07315379e-01
1.06934643e+00 -5.46045721e-01 -2.52811108e-02 1.25350773e-01
-5.47903240e-01 3.67182583e-01 5.19191265e-01 -6.82505548e-01
-5.57085752e-01 -1.06082730e-01 1.96769819e-01 -6.95769072e-01
-2.98851252e-01 6.62501991e-01 -2.45977774e-01 3.92442584e-01
8.40890527e-01 1.84342802e-01 -1.53947219e-01 -7.74436295e-01
7.82958448e-01 6.40986502e-01 8.01093102e-01 -6.55922413e-01
5.67939878e-01 2.28131980e-01 8.09364021e-02 -4.88015771e-01
-9.21844363e-01 -6.44127786e-01 -7.52147555e-01 2.03507356e-02
6.83955550e-01 -9.56320941e-01 -8.49802434e-01 -2.55234726e-02
-9.73403394e-01 4.11423504e-01 -5.25992811e-01 7.65282452e-01
-9.65118527e-01 4.40954864e-01 -4.20105755e-01 -5.65077722e-01
-3.65873754e-01 -9.23384070e-01 1.09139466e+00 -1.18411988e-01
-9.70879912e-01 -7.23831177e-01 4.76724148e-01 5.06813228e-01
3.33056718e-01 3.20200652e-01 1.18801415e+00 -1.05148375e+00
-5.36084354e-01 -1.29270807e-01 4.78834696e-02 5.91226034e-02
1.45535544e-01 -4.51638907e-01 -1.19590306e+00 -2.80169975e-02
1.34586513e-01 -3.97708625e-01 8.31018507e-01 -5.32632358e-02
1.10962021e+00 -2.18304679e-01 4.01371837e-01 6.84235692e-01
1.15269709e+00 -1.29638210e-01 3.68263602e-01 3.60347301e-01
7.44609118e-01 4.34825778e-01 3.73022437e-01 6.57348096e-01
3.04141551e-01 1.11360431e+00 3.83791983e-01 2.29074419e-01
-8.91317949e-02 -2.61459678e-01 2.59979129e-01 1.17612374e+00
-1.74694911e-01 9.33462828e-02 -5.61839342e-01 3.85053009e-01
-1.79800856e+00 -1.25261581e+00 2.44498268e-01 2.26775455e+00
7.35628963e-01 -1.58584177e-01 2.66495913e-01 5.84047616e-01
2.91542202e-01 4.00319427e-01 -5.20777822e-01 -5.09476900e-01
-1.59916103e-01 3.58252883e-01 -2.49914035e-01 1.11984141e-01
-9.90030468e-01 6.22602940e-01 6.10001135e+00 6.92489386e-01
-1.05267739e+00 -3.09731662e-01 -2.45060772e-01 -4.15621340e-01
-6.04828119e-01 -1.27441138e-01 -1.15715146e-01 1.50435522e-01
7.99601316e-01 -2.49022529e-01 7.86040068e-01 5.49717009e-01
1.27940893e-01 4.00922060e-01 -1.72082961e+00 1.48048735e+00
3.02175820e-01 -1.01870084e+00 4.96565610e-01 1.18954532e-01
3.52196842e-01 -2.33881131e-01 4.12247032e-01 3.18712175e-01
-3.35925549e-01 -1.13865244e+00 8.45634460e-01 9.76954460e-01
5.20804346e-01 -6.48798227e-01 1.64788470e-01 2.21262589e-01
-1.00820637e+00 -9.18340981e-02 -4.01293449e-02 -3.89236450e-01
-1.78497568e-01 2.22316399e-01 -6.38340950e-01 6.96739197e-01
2.66140968e-01 8.19809616e-01 -7.20634043e-01 9.21483338e-01
1.45905450e-01 2.02809706e-01 7.28000477e-02 1.94008783e-01
-1.91726372e-01 -2.77109444e-01 6.98409736e-01 1.03831863e+00
5.57798207e-01 -1.56239882e-01 5.58895478e-03 1.20991921e+00
3.61284465e-02 3.35631460e-01 -5.12406766e-01 -3.24559063e-01
4.15208489e-01 1.14459074e+00 -2.92550296e-01 -3.96326222e-02
-2.01667428e-01 8.60657632e-01 2.43331984e-01 2.45983526e-01
-3.76923054e-01 -2.04466656e-01 1.12399960e+00 -1.74667820e-01
2.35482171e-01 -3.55973721e-01 -4.48525131e-01 -1.32921255e+00
-1.10082433e-01 -1.17598128e+00 3.49846870e-01 -8.54593337e-01
-1.37128925e+00 5.86847842e-01 -1.68969572e-01 -1.52158713e+00
-8.02896976e-01 -4.08144325e-01 -4.40213859e-01 1.13293207e+00
-1.02919197e+00 -1.22899806e+00 -9.67812389e-02 6.71059072e-01
3.26192647e-01 -5.31962931e-01 1.44326150e+00 2.20568359e-01
-1.43411249e-01 7.07201362e-01 1.34686500e-01 -5.16029857e-02
9.89398003e-01 -1.28447354e+00 1.31551206e-01 4.58380848e-01
1.09584486e+00 8.12918723e-01 6.88400924e-01 -1.57389045e-01
-1.38477743e+00 -5.36702752e-01 8.66696298e-01 -5.38792610e-01
7.70117939e-01 -1.59895584e-01 -7.13032961e-01 3.46561223e-01
1.23077601e-01 -3.89500946e-01 1.44328642e+00 6.19940281e-01
-8.61869931e-01 -1.10577019e-02 -5.79304695e-01 6.75713897e-01
8.53095174e-01 -1.19097877e+00 -1.04047453e+00 -1.81277871e-01
5.51215589e-01 -3.86217982e-02 -1.17771518e+00 2.74310023e-01
1.09891415e+00 -1.19427001e+00 1.20378351e+00 -6.79509461e-01
4.41219002e-01 -3.95559341e-01 -6.52988017e-01 -1.49212909e+00
-7.63928175e-01 -5.18433034e-01 -2.41383791e-01 9.43200171e-01
4.57443744e-02 1.05602019e-01 4.49706793e-01 2.69057423e-01
4.94129062e-02 -5.76975822e-01 -6.66866601e-01 -4.94840503e-01
-1.30041853e-01 -7.61213541e-01 8.66888404e-01 1.09185445e+00
3.35119039e-01 8.42879057e-01 -5.55578947e-01 -5.25149964e-02
5.02424419e-01 6.52326286e-01 6.56518996e-01 -1.57420766e+00
-9.81403232e-01 -8.43521655e-01 -8.03272724e-01 -7.59176195e-01
9.20926705e-02 -1.32065058e+00 -4.01570022e-01 -1.05853760e+00
2.29291335e-01 -5.07257767e-02 -8.76787424e-01 6.96700886e-02
-7.82942846e-02 2.56741256e-01 6.52160585e-01 4.43632752e-01
-5.82767308e-01 6.46244764e-01 1.12920618e+00 -2.25393653e-01
-2.18703389e-01 7.77201308e-03 -9.65975463e-01 7.43556201e-01
6.73855424e-01 -1.49429217e-01 -4.75806177e-01 -3.74033719e-01
5.22646129e-01 2.40452945e-01 3.99029851e-01 -1.03683031e+00
1.41293079e-01 5.50176799e-02 3.62651169e-01 -3.49533498e-01
8.18582952e-01 -6.90108418e-01 2.37185553e-01 -4.19707783e-02
-8.70581925e-01 3.39332372e-02 -2.64009312e-02 6.45318329e-01
-4.84027028e-01 -9.35101137e-02 3.20044696e-01 1.82434693e-02
-3.03131938e-01 6.77020336e-03 2.60280669e-02 -1.71674404e-03
3.32450092e-01 -1.59813821e-01 8.66985843e-02 -3.71209443e-01
-1.03099179e+00 -5.74523687e-01 3.12776327e-01 7.65351653e-01
4.31132376e-01 -1.71901822e+00 -7.12742865e-01 2.28804052e-01
3.53808194e-01 -6.32095456e-01 1.09693505e-01 4.58923668e-01
8.15865919e-02 3.51105630e-01 -5.34752250e-01 -4.92617577e-01
-1.18172157e+00 7.45140374e-01 6.43882006e-02 -2.06522316e-01
-3.25341702e-01 3.41273963e-01 1.86947316e-01 -2.92016685e-01
3.72986585e-01 -4.46822673e-01 -4.73948687e-01 3.77505213e-01
4.97996062e-01 3.46002489e-01 -7.67488554e-02 -8.62875462e-01
-1.52433952e-02 6.77510798e-01 2.62431562e-01 -4.24920499e-01
1.22568238e+00 -2.40631588e-02 1.35182729e-02 1.06956148e+00
1.22845805e+00 1.56051531e-01 -6.27274871e-01 -3.88785660e-01
2.20220461e-01 -4.97445911e-01 -7.98383914e-03 -8.78326833e-01
-5.81253171e-01 9.25951183e-01 6.28151953e-01 3.52168620e-01
1.18592834e+00 -9.42011848e-02 5.27907014e-01 4.03080523e-01
1.44057289e-01 -7.48174012e-01 1.81401566e-01 4.43979234e-01
1.17623413e+00 -1.08222055e+00 -1.39043108e-02 1.79216012e-01
-5.30458689e-01 1.13391328e+00 1.97059378e-01 -1.28515124e-01
6.05699778e-01 -3.52271557e-01 -2.19303463e-03 -2.20229179e-01
-6.62148416e-01 -3.04830194e-01 1.11683989e+00 2.01301515e-01
8.29268098e-01 4.00060534e-01 -2.12102592e-01 1.15241945e+00
-9.20410812e-01 -2.66854078e-01 2.45793939e-01 3.78861755e-01
-2.08325207e-01 -1.16095161e+00 -4.79769379e-01 2.61120260e-01
-4.17819589e-01 -1.82838961e-01 -5.95747471e-01 4.79996890e-01
2.73516893e-01 7.40654051e-01 -2.62734038e-03 -6.21977448e-01
2.56930709e-01 3.17025661e-01 7.18023777e-01 -4.65847701e-01
-7.72276521e-01 2.60047644e-01 -8.64014998e-02 -5.39224684e-01
-5.44242740e-01 -6.46158278e-01 -7.80813158e-01 1.28110945e-01
6.61299154e-02 -1.31180054e-02 8.73457849e-01 8.81676733e-01
4.17382836e-01 6.24486685e-01 5.50313234e-01 -1.13322139e+00
-7.81940579e-01 -1.17102230e+00 -8.92406166e-01 9.39612687e-01
2.97103047e-01 -6.17373824e-01 -3.29378337e-01 -7.11504892e-02]
|
[15.873465538024902, 5.3262529373168945]
|
edd582b8-fe00-4998-827a-8b86b0a6ef3e
|
masked-autoencoders-for-egocentric-video
|
2211.15286
| null |
https://arxiv.org/abs/2211.15286v1
|
https://arxiv.org/pdf/2211.15286v1.pdf
|
Masked Autoencoders for Egocentric Video Understanding @ Ego4D Challenge 2022
|
In this report, we present our approach and empirical results of applying masked autoencoders in two egocentric video understanding tasks, namely, Object State Change Classification and PNR Temporal Localization, of Ego4D Challenge 2022. As team TheSSVL, we ranked 2nd place in both tasks. Our code will be made available.
|
['Kui Ren', 'Ashish Kapoor', 'Sai Vemprala', 'Zhongjie Ba', 'Shuang Ma', 'Jiachen Lei']
|
2022-11-18
| null | null | null | null |
['video-understanding']
|
['computer-vision']
|
[-3.97656143e-01 -1.98329277e-02 -2.21932366e-01 -3.28461677e-01
4.71713245e-02 -5.20194292e-01 8.83637667e-01 -5.90552926e-01
-4.42052096e-01 6.25857055e-01 6.44507229e-01 1.17160656e-01
2.11293310e-01 -1.45809487e-01 -9.78108525e-01 -1.86446235e-01
-6.37419522e-01 1.70781404e-01 1.34273618e-01 6.84552416e-02
1.07916333e-01 3.91856492e-01 -1.81952560e+00 4.38288808e-01
4.14093677e-03 1.03130317e+00 8.38737488e-02 9.13903296e-01
4.66226876e-01 1.40095913e+00 -2.82125622e-01 -2.25794047e-01
2.50424325e-01 -1.41031042e-01 -1.16505277e+00 8.01032111e-02
6.78355813e-01 -8.50690186e-01 -1.34133458e+00 1.05747962e+00
4.57308680e-01 6.10908508e-01 6.87333584e-01 -1.41390705e+00
-4.15931851e-01 5.85521460e-01 -2.78172195e-01 1.08571923e+00
5.80638885e-01 2.43794590e-01 7.06807077e-01 -1.08952379e+00
1.00617182e+00 1.20890546e+00 4.72854227e-01 8.18445265e-01
-8.07339072e-01 -5.44255435e-01 3.63319010e-01 1.06396925e+00
-1.34656906e+00 -1.24376142e+00 5.92999399e-01 -7.26717770e-01
1.58169353e+00 -3.98968965e-01 7.47891903e-01 1.71492970e+00
1.51173353e-01 1.53937137e+00 7.17178583e-01 2.46726766e-01
4.00100380e-01 -1.01276755e-01 4.86021414e-02 4.12276626e-01
-2.23295346e-01 3.65075022e-01 -8.57644558e-01 1.07239999e-01
7.33664572e-01 -3.23839217e-01 -3.75419289e-01 -4.92957175e-01
-1.46262980e+00 4.89647567e-01 3.10073137e-01 1.85553432e-01
-9.86533463e-01 7.71503031e-01 6.55844510e-01 4.49588835e-01
5.96993625e-01 1.84486695e-02 -5.71188748e-01 -7.83415854e-01
-6.58743560e-01 2.75643349e-01 4.89663512e-01 1.25627184e+00
3.62476707e-01 4.78090912e-01 1.49832845e-01 3.19140702e-01
-5.88034885e-03 3.39809805e-01 6.71493471e-01 -1.50894821e+00
5.70032895e-02 -7.48546720e-02 1.35188445e-01 -9.42408919e-01
-5.03617704e-01 1.06477931e-01 -3.22121859e-01 7.12642958e-03
-1.93538830e-01 -3.65952879e-01 -9.49919522e-01 1.91597009e+00
1.13464661e-01 1.02611291e+00 5.27302623e-01 1.06356645e+00
1.40741873e+00 5.53944170e-01 2.45270073e-01 1.31300554e-01
1.25625300e+00 -1.00199592e+00 -9.65540767e-01 -2.77584016e-01
4.48893696e-01 -1.72415882e-01 -9.33294836e-03 1.88720614e-01
-1.42288470e+00 -6.48435593e-01 -1.03385615e+00 4.37075272e-02
-2.78201282e-01 -5.12151718e-02 8.45972538e-01 1.60655096e-01
-1.63073766e+00 7.58506894e-01 -1.37470424e+00 -1.05211234e+00
4.52776581e-01 2.77204216e-01 -6.71104312e-01 2.07767874e-01
-1.38294077e+00 8.55472744e-01 6.37753904e-01 -1.69959873e-01
-1.57893991e+00 -5.06589055e-01 -1.12210917e+00 -1.20936632e-01
7.01525658e-02 -5.71969986e-01 1.67631400e+00 -8.31061125e-01
-1.56067908e+00 1.02754784e+00 -9.13814902e-02 -1.05019522e+00
2.39043579e-01 -4.14813042e-01 -6.59992576e-01 3.80582601e-01
6.96266517e-02 1.14836180e+00 8.11511695e-01 -5.95128000e-01
-7.28806317e-01 -3.22961211e-01 2.90191919e-01 5.20312607e-01
-1.02966547e-01 7.90477246e-02 -3.76770228e-01 -3.28861296e-01
1.34396210e-01 -1.00837970e+00 -1.46332686e-03 -4.38132912e-01
1.18946157e-01 -5.84175646e-01 9.23374236e-01 -6.75058782e-01
4.13519979e-01 -2.16072774e+00 6.77908599e-01 -4.22425359e-01
3.48287791e-01 -8.84877220e-02 -2.36434609e-01 1.25858605e-01
-6.64034307e-01 -3.06019992e-01 5.45862198e-01 -4.85297203e-01
1.66083816e-02 -6.93520382e-02 -4.36207980e-01 8.21011782e-01
1.91063792e-01 9.66155231e-01 -1.10474885e+00 -5.16352132e-02
4.36959296e-01 2.25774840e-01 -5.17267048e-01 1.73006058e-01
2.17564628e-02 3.39248657e-01 -3.66450936e-01 6.69085383e-01
3.98746580e-01 -2.35736623e-01 -1.41231567e-01 -6.04533792e-01
-1.49548501e-01 3.23909223e-01 -9.24587548e-01 2.18229914e+00
-1.82154357e-01 1.38464224e+00 -7.35455528e-02 -7.67613947e-01
3.08831394e-01 8.02556992e-01 9.00667191e-01 -5.62294126e-01
4.23443943e-01 -4.31305617e-01 -2.30266199e-01 -7.17501819e-01
6.69843554e-01 3.93131882e-01 4.15761583e-02 2.59562075e-01
8.05104434e-01 1.78186417e-01 2.73853689e-01 5.30631959e-01
1.42249489e+00 5.02216339e-01 3.67093712e-01 -3.41986746e-01
4.77157921e-01 -1.63580347e-02 2.72334576e-01 8.96546304e-01
-1.05823123e+00 5.51557541e-01 3.95869315e-01 -7.58982003e-01
-8.83974373e-01 -1.15242684e+00 1.19209647e-01 9.07860696e-01
1.55073851e-01 -5.20683706e-01 -7.43318081e-01 -7.66982794e-01
-2.02902555e-01 7.55273044e-01 -7.91250408e-01 -2.75188237e-01
-3.40162784e-01 -1.54343590e-01 7.11369991e-01 8.17433536e-01
5.28734624e-01 -1.33546734e+00 -4.15724069e-01 4.65027280e-02
-3.41797054e-01 -1.65083110e+00 -2.81087279e-01 5.15053794e-02
-7.76397824e-01 -1.19699860e+00 -7.00180292e-01 -6.90319896e-01
-2.62271594e-02 5.53380787e-01 1.14001691e+00 -6.13046765e-01
-1.50323614e-01 1.19551814e+00 -3.13436747e-01 -2.86636770e-01
7.47083724e-02 5.18281385e-02 1.03572166e+00 -6.45231456e-02
8.08383644e-01 -7.62962878e-01 -6.18903697e-01 1.71455517e-01
-3.69555473e-01 -1.22685611e-01 1.31816804e-01 3.07051271e-01
2.51413703e-01 -1.63045689e-01 3.06368142e-01 -8.99275467e-02
1.57410219e-01 -8.35476637e-01 -5.55594325e-01 -4.26426888e-01
-1.05059817e-01 -2.87124455e-01 -6.67143688e-02 -1.95100680e-01
-9.11897600e-01 3.21977168e-01 -7.82868862e-02 -1.21864808e+00
-5.94367027e-01 2.14967668e-01 1.17942549e-01 -4.57221176e-04
7.30322480e-01 4.71986592e-01 -2.37900108e-01 -1.65657580e-01
4.83788431e-01 3.41766536e-01 7.53429055e-01 -2.04230309e-01
4.92642224e-01 8.38159144e-01 -4.35671717e-01 -1.01370597e+00
-5.08135855e-01 -5.61897516e-01 -4.46775466e-01 -5.90795875e-01
1.29932928e+00 -1.89682996e+00 -9.24150646e-01 6.13075793e-01
-1.31398320e+00 -3.91735077e-01 -3.26084644e-01 8.68276477e-01
-1.30297887e+00 2.78716475e-01 -6.90911233e-01 -3.86478752e-01
-1.30695835e-01 -8.76882553e-01 7.54760921e-01 5.19812405e-01
-2.37324476e-01 -8.22805762e-01 5.34653187e-01 1.31324083e-01
3.44431072e-01 -3.10311615e-01 -1.43414997e-02 -8.06865931e-01
-7.09861994e-01 -2.56789714e-01 -1.77912429e-01 2.03316778e-01
-1.45933479e-01 -4.64356244e-01 -1.28308630e+00 -4.72353160e-01
6.09995681e-04 -6.65617883e-01 9.66743886e-01 8.76497686e-01
1.34062850e+00 1.31304264e-01 -4.74993676e-01 7.58684158e-01
8.73769760e-01 1.52832896e-01 8.22763085e-01 2.18104824e-01
4.57299292e-01 5.34505606e-01 4.04102236e-01 6.29585505e-01
6.67835712e-01 5.68640172e-01 5.03398478e-01 4.87381965e-01
-2.61915624e-01 -1.25917181e-01 6.77899301e-01 5.57094276e-01
-5.52404225e-01 -5.79294801e-01 -7.34197140e-01 8.86332035e-01
-1.93462384e+00 -1.50609779e+00 2.06489637e-01 1.42958295e+00
5.94636053e-02 -5.09120263e-02 1.31026924e-01 -6.64654076e-01
6.28821790e-01 6.12513900e-01 -6.12973034e-01 1.69129726e-02
-7.77373835e-02 -1.85552284e-01 4.33284193e-01 3.30214679e-01
-1.73617375e+00 1.45325840e+00 7.47162771e+00 2.03619674e-01
-8.05633307e-01 3.82954121e-01 3.71176191e-02 -5.59620202e-01
5.13844907e-01 -1.37547895e-01 -5.64227760e-01 2.95897037e-01
1.36311078e+00 -2.48607054e-01 6.84854269e-01 1.23725820e+00
1.68529078e-01 -3.56921218e-02 -1.38981593e+00 1.47034740e+00
1.66490883e-01 -1.40858686e+00 -2.59858996e-01 -7.60557503e-02
7.02974558e-01 9.62727666e-01 2.37145592e-02 7.23745823e-01
4.54338878e-01 -8.46800923e-01 5.79483926e-01 5.66296101e-01
4.24740285e-01 -4.70249951e-01 6.75865531e-01 -1.70909643e-01
-1.23720133e+00 1.13676041e-02 -2.21919179e-01 -1.30709559e-01
3.34417492e-01 -2.62081563e-01 -7.43719816e-01 2.33543098e-01
1.32287359e+00 1.59425342e+00 -2.18603194e-01 8.47597659e-01
-2.10540637e-01 5.62981248e-01 -1.17747918e-01 9.82827172e-02
1.62348762e-01 1.87293932e-01 1.24888813e+00 1.20148277e+00
1.09896027e-02 5.44050157e-01 -5.32373674e-02 7.21602738e-01
-9.17815566e-02 -5.10703027e-01 -9.40423846e-01 -3.50913286e-01
4.49616052e-02 1.09618068e+00 -2.51717567e-01 -6.78978622e-01
-3.71645540e-01 1.39388525e+00 1.16292261e-01 7.14773893e-01
-1.05410802e+00 -3.12036842e-01 1.40506816e+00 -1.87047511e-01
6.03085876e-01 -2.11464942e-01 4.42616940e-01 -1.69205570e+00
-3.38593423e-01 -6.15115464e-01 3.82015228e-01 -1.19064438e+00
-9.28723097e-01 7.92045951e-01 1.96730241e-01 -1.18955219e+00
-6.09653234e-01 -6.16704583e-01 -5.37395418e-01 3.84008050e-01
-1.07061088e+00 -7.17539966e-01 -5.06659806e-01 7.94436276e-01
9.17834282e-01 -6.68961585e-01 6.49315953e-01 5.11200011e-01
-6.07564211e-01 3.21129233e-01 -2.36846562e-02 1.88193604e-01
5.66721320e-01 -1.04688072e+00 9.28417265e-01 8.40673387e-01
1.07648619e-01 4.04634833e-01 1.14553916e+00 -6.02888107e-01
-1.46654224e+00 -9.21143770e-01 5.05252779e-01 -6.53190851e-01
9.99353409e-01 -1.77713007e-01 -4.66722906e-01 1.59113681e+00
5.49261332e-01 -1.08629845e-01 2.69831955e-01 1.25379086e-01
6.00747094e-02 4.13355500e-01 -8.49465191e-01 7.73226380e-01
1.49616468e+00 -6.21776879e-01 -8.33892703e-01 5.75817704e-01
7.34775186e-01 -7.06343174e-01 -7.98044503e-01 3.79514515e-01
3.59119028e-01 -7.51170039e-01 1.00789547e+00 -1.25645244e+00
5.25221527e-01 -4.20303017e-01 -4.77452606e-01 -1.40888751e+00
-7.39794970e-01 -5.71806729e-01 -8.86012077e-01 6.71034455e-01
-1.42306700e-01 -1.99887186e-01 1.24238634e+00 1.36850253e-01
-2.43507355e-01 2.08886057e-01 -1.28457940e+00 -7.53132641e-01
-5.35540283e-01 -5.09923697e-01 9.32519510e-02 7.54052520e-01
-3.40517163e-02 2.94416517e-01 -5.45860529e-01 3.64266336e-01
8.01487267e-01 -5.37600219e-01 8.41492832e-01 -7.45576203e-01
-6.16207458e-02 -2.95965374e-01 -1.27391076e+00 -1.36115861e+00
6.21067524e-01 -6.62840605e-01 -8.69542807e-02 -1.03729177e+00
1.44260585e-01 6.43905818e-01 -5.61532140e-01 3.38923961e-01
3.43714505e-01 2.09248587e-01 8.56853873e-02 -7.92442560e-02
-1.19037199e+00 9.61499870e-01 4.44805712e-01 -3.65014941e-01
-1.65115818e-02 -2.15541068e-02 -2.06170008e-01 8.61928165e-01
6.43241763e-01 -3.98823380e-01 -4.72590059e-01 -4.45691228e-01
-2.06440285e-01 -4.50272491e-04 6.72183990e-01 -1.53946745e+00
3.11654776e-01 1.12767117e-02 8.11838806e-01 -1.08071196e+00
8.78340721e-01 -6.66036963e-01 9.88909528e-02 4.74981666e-01
-1.17089644e-01 1.91334099e-01 6.17519021e-01 7.91107595e-01
-1.11494429e-01 8.47957432e-02 3.50388229e-01 -2.62639135e-01
-1.67095518e+00 7.01104343e-01 -1.03811526e+00 -1.34585693e-01
1.04194403e+00 -6.18045554e-02 -2.80970544e-01 -9.41655338e-01
-1.18757868e+00 4.94758040e-01 3.80792588e-01 1.05093968e+00
7.77382553e-01 -1.34693539e+00 -7.51137435e-01 -8.33713412e-02
1.92430437e-01 -7.32226193e-01 8.35076451e-01 9.22162771e-01
-3.01071644e-01 5.96713424e-01 -5.78917921e-01 -8.49304974e-01
-9.98842776e-01 7.55437434e-01 6.08855724e-01 4.86127764e-01
-6.79370284e-01 8.85136902e-01 4.18533504e-01 -2.09943056e-01
3.16970527e-01 2.25295246e-01 -5.43368399e-01 6.33481424e-03
1.05157948e+00 4.38102633e-01 -2.48942286e-01 -6.21672273e-01
-7.03069806e-01 1.09732732e-01 -2.90127695e-01 -1.38768479e-01
1.39541912e+00 -4.12241906e-01 1.05819605e-01 4.46392626e-01
1.15119886e+00 -8.17045033e-01 -1.54217994e+00 -2.75554329e-01
-3.20671320e-01 -5.22244647e-02 5.26481211e-01 -3.07556361e-01
-1.15951872e+00 5.68638980e-01 1.21581650e+00 -3.49200964e-02
9.54850793e-01 3.47178042e-01 4.80411083e-01 8.13977540e-01
3.88529450e-01 -9.85864401e-01 -2.38738693e-02 9.45145786e-01
8.46830726e-01 -1.55832458e+00 1.27366900e-01 1.02771960e-01
-7.89359093e-01 8.52690578e-01 1.00301659e+00 -3.77134204e-01
7.51901448e-01 -2.22147092e-01 -1.88848138e-01 -5.14695346e-01
-1.11700332e+00 -3.61438572e-01 2.43930906e-01 8.42702508e-01
1.40334055e-01 -1.37327641e-01 3.79050881e-01 2.98016608e-01
-1.00270044e-02 2.33842269e-01 6.53284371e-01 8.69105995e-01
-2.06692144e-01 -1.02788329e-01 1.78840950e-01 -1.83656684e-03
-3.02821398e-01 -1.21511653e-01 -2.14670107e-01 7.45111048e-01
-3.34394276e-01 7.33845115e-01 4.68843281e-01 -7.64648795e-01
3.71217877e-01 5.92841953e-02 6.48404956e-01 -1.84369460e-01
-2.17816476e-02 -3.73420507e-01 1.96699306e-01 -1.30259502e+00
-5.10218740e-01 -1.19814682e+00 -8.80902290e-01 -6.38957560e-01
1.12451158e-01 -1.51128858e-01 8.06699812e-01 7.82550991e-01
6.50935769e-01 6.01094842e-01 3.91905725e-01 -1.30885625e+00
-2.76899248e-01 -1.15017056e+00 -4.52956289e-01 1.46187738e-01
5.22971213e-01 -9.95015025e-01 -3.66459221e-01 2.76917875e-01]
|
[8.372199058532715, 0.554556131362915]
|
5c7ddf51-c452-428a-90f2-82d1cf4e95b5
|
mobile-mapping-mesh-change-detection-and
|
2303.07182
| null |
https://arxiv.org/abs/2303.07182v1
|
https://arxiv.org/pdf/2303.07182v1.pdf
|
Mobile Mapping Mesh Change Detection and Update
|
Mobile mapping, in particular, Mobile Lidar Scanning (MLS) is increasingly widespread to monitor and map urban scenes at city scale with unprecedented resolution and accuracy. The resulting point cloud sampling of the scene geometry can be meshed in order to create a continuous representation for different applications: visualization, simulation, navigation, etc. Because of the highly dynamic nature of these urban scenes, long term mapping should rely on frequent map updates. A trivial solution is to simply replace old data with newer data each time a new acquisition is made. However it has two drawbacks: 1) the old data may be of higher quality (resolution, precision) than the new and 2) the coverage of the scene might be different in various acquisitions, including varying occlusions. In this paper, we propose a fully automatic pipeline to address these two issues by formulating the problem of merging meshes with different quality, coverage and acquisition time. Our method is based on a combined distance and visibility based change detection, a time series analysis to assess the sustainability of changes, a mesh mosaicking based on a global boolean optimization and finally a stitching of the resulting mesh pieces boundaries with triangle strips. Finally, our method is demonstrated on Robotcar and Stereopolis datasets.
|
['Cédric Demonceaux', 'Bruno Vallet', 'Teng Wu']
|
2023-03-13
| null | null | null | null |
['change-detection']
|
['computer-vision']
|
[ 3.39725375e-01 -3.88235271e-01 4.44216996e-01 -3.34804803e-01
-3.23756248e-01 -5.57616591e-01 7.42963970e-01 5.32754004e-01
-4.13533807e-01 1.03239238e+00 -4.07934427e-01 -2.14347303e-01
-2.76413560e-01 -1.36567044e+00 -5.68291545e-01 -4.69570339e-01
2.57385820e-02 1.05014241e+00 8.00747037e-01 -5.45489490e-01
3.19277167e-01 1.01979113e+00 -2.07788944e+00 -3.48514438e-01
1.20370924e+00 8.50052118e-01 5.64351618e-01 4.38130468e-01
-4.98196572e-01 -1.69746175e-01 -3.18860114e-01 4.87140492e-02
3.36216778e-01 1.21084243e-01 -4.56350327e-01 1.50463879e-01
3.81883264e-01 -4.82974276e-02 4.81838793e-01 1.11084700e+00
1.71182618e-01 6.77075833e-02 2.91831255e-01 -1.18633199e+00
3.68080020e-01 -1.19188890e-01 -7.39883304e-01 -1.25298783e-01
4.75277066e-01 1.68592364e-01 3.52446139e-01 -9.11616027e-01
9.56884563e-01 1.04597485e+00 7.83277750e-01 -3.48587155e-01
-1.58559370e+00 -6.16872489e-01 8.36462751e-02 4.48661387e-01
-1.55339050e+00 -3.43016177e-01 6.58287764e-01 -7.32720792e-01
4.54672784e-01 5.45507848e-01 1.16401422e+00 5.37182868e-01
-3.53606418e-02 -2.90684819e-01 1.22644007e+00 -1.65857226e-01
4.33347702e-01 2.02496320e-01 -2.07338497e-01 3.13458711e-01
4.04686838e-01 -8.00232217e-02 -2.03765869e-01 -4.14593294e-02
6.61041796e-01 3.45938541e-02 -2.17737943e-01 -6.35407209e-01
-1.06377554e+00 6.25076354e-01 3.85766625e-01 4.00431544e-01
-5.39158165e-01 -6.75312355e-02 1.01457313e-01 3.72957587e-01
4.56833750e-01 2.44856060e-01 -3.61134529e-01 -1.60285980e-01
-1.09379208e+00 4.43575591e-01 4.98114765e-01 7.24050403e-01
1.39009261e+00 -2.43585840e-01 4.10939813e-01 5.48025012e-01
2.93246865e-01 6.34974360e-01 7.73644494e-03 -8.18290114e-01
4.22420502e-01 8.09074521e-01 2.35139474e-01 -1.26373994e+00
-6.02882862e-01 -2.83711255e-01 -8.87692988e-01 7.25011289e-01
2.74014324e-01 2.80878246e-01 -7.20580161e-01 1.27489865e+00
8.15311432e-01 2.54403293e-01 -4.58302766e-01 5.39128244e-01
3.36653650e-01 4.91940230e-01 -2.47049972e-01 -3.60550165e-01
1.18276906e+00 -1.69369802e-01 -7.02755749e-01 -1.23657502e-01
3.30258399e-01 -7.92041779e-01 8.40555966e-01 3.18720847e-01
-7.80433834e-01 -4.96574789e-01 -1.10665965e+00 1.67571440e-01
-5.85162640e-01 -2.97581613e-01 1.93377152e-01 4.31167781e-01
-1.13723385e+00 6.38171852e-01 -6.81286037e-01 -5.99933386e-01
2.38450855e-01 2.29328856e-01 -2.51011550e-01 -4.16594669e-02
-9.52479899e-01 1.05513811e+00 2.29106233e-01 1.09289065e-01
-8.49217735e-03 -8.23628545e-01 -7.84891069e-01 -1.18862025e-01
3.10848594e-01 -5.79129100e-01 5.29674888e-01 -6.02791667e-01
-1.38187456e+00 7.73826599e-01 -2.44132802e-01 -3.82213056e-01
8.46139431e-01 2.32209414e-01 -4.22277957e-01 -3.96366298e-01
4.08265382e-01 3.11722308e-01 6.07648075e-01 -1.39864719e+00
-8.39234650e-01 -6.60623133e-01 -1.30497098e-01 3.03478479e-01
3.51834148e-01 -5.08117676e-01 -3.31700355e-01 -2.38277495e-01
5.69477975e-01 -8.81510675e-01 -4.23317879e-01 1.74459502e-01
-4.92477752e-02 3.66361529e-01 1.07396352e+00 -5.72929740e-01
1.14710224e+00 -2.21488142e+00 1.03216700e-01 5.63388586e-01
-4.06717882e-02 1.61572229e-02 2.53679395e-01 4.91754413e-01
2.50516802e-01 5.27909435e-02 -6.68777943e-01 -3.73259813e-01
-3.33227962e-01 2.63480216e-01 -1.45818993e-01 5.12080967e-01
-7.51116592e-03 4.49244052e-01 -7.63785720e-01 -4.79439765e-01
6.67145014e-01 3.99085879e-01 -1.92830458e-01 -1.95947707e-01
-3.32086593e-01 9.32024419e-01 -1.60585865e-01 5.72716892e-01
1.28429031e+00 2.59739220e-01 9.07510370e-02 -4.01227102e-02
-9.84851718e-01 7.02917352e-02 -1.80901587e+00 1.79894674e+00
-4.38299179e-01 7.36117005e-01 3.51271868e-01 -6.36725366e-01
1.04767239e+00 1.08007034e-02 5.76072812e-01 -8.07384253e-01
-1.05416209e-01 4.91145968e-01 -4.23752874e-01 -3.57800394e-01
9.33358073e-01 7.02966377e-02 9.04195309e-02 9.42917913e-02
-7.74871409e-01 -7.21718550e-01 1.66934326e-01 -2.35198945e-01
7.54524648e-01 7.33389184e-02 1.85812578e-01 -3.36648762e-01
6.99942350e-01 4.17289108e-01 4.53968734e-01 3.24196756e-01
1.93997040e-01 5.58997571e-01 1.95978507e-01 -5.78596771e-01
-1.05441606e+00 -8.91003966e-01 -5.70690572e-01 1.04023904e-01
4.74298418e-01 8.16324085e-04 -3.19170624e-01 -5.32256179e-02
2.02689201e-01 6.53927445e-01 -3.79657507e-01 3.42057377e-01
-6.96123064e-01 -4.87785071e-01 -1.90175138e-02 -2.33158439e-01
6.31414175e-01 -7.05047131e-01 -1.17882419e+00 4.83966082e-01
-2.04432428e-01 -9.31304038e-01 1.27027899e-01 -1.04533367e-01
-1.10502553e+00 -1.11209536e+00 -1.17408484e-01 -3.01739454e-01
4.36298311e-01 4.89433289e-01 9.83456910e-01 1.49581298e-01
-2.82744318e-01 1.08474448e-01 -1.87852189e-01 -3.37736249e-01
-3.24584454e-01 -4.92179878e-02 -1.69334307e-01 1.22248419e-01
-3.17761689e-01 -1.08491135e+00 -4.12424415e-01 4.06818271e-01
-8.52862298e-01 3.44923884e-01 3.46716702e-01 3.17960143e-01
9.84006166e-01 3.53400469e-01 2.26051539e-01 -7.23347485e-01
3.17472070e-01 -6.23351038e-01 -1.08576584e+00 5.98215908e-02
-6.09141588e-01 -2.54930913e-01 1.02147870e-01 -1.48263559e-01
-7.65926719e-01 2.25112900e-01 -4.34987880e-02 -1.95365280e-01
-1.51142731e-01 6.83386981e-01 -1.36028469e-01 -1.12156853e-01
5.07192969e-01 2.09474228e-02 1.26175255e-01 -5.39718151e-01
3.38886380e-01 4.08003360e-01 4.00791466e-01 -1.50781512e-01
1.09321213e+00 9.50524867e-01 3.30738574e-01 -1.11005235e+00
1.69153735e-01 -4.78338391e-01 -1.01618409e+00 -5.87587535e-01
7.53971636e-01 -6.24456704e-01 -2.75870234e-01 5.08364618e-01
-1.31256306e+00 -2.25849852e-01 -4.40608472e-01 2.55461276e-01
-2.67898798e-01 3.33253831e-01 1.83057934e-01 -7.84583926e-01
2.11520474e-02 -1.23250163e+00 9.57786679e-01 1.40485644e-01
-1.01748176e-01 -7.12954342e-01 3.61808270e-01 2.11955905e-02
6.35478616e-01 9.29998577e-01 7.39219904e-01 3.39358330e-01
-9.70313609e-01 -1.27974108e-01 -2.37308711e-01 -4.55556929e-01
3.36138457e-01 2.38030568e-01 -5.79301298e-01 -1.79648712e-01
-5.89151541e-03 4.72279578e-01 3.59162927e-01 3.40705246e-01
6.21322393e-01 2.35712249e-02 -4.96068031e-01 5.81836939e-01
1.82155824e+00 2.29120910e-01 8.09515536e-01 7.27730393e-01
4.49848771e-01 7.59434104e-01 9.51765418e-01 4.86546755e-01
5.68328500e-01 1.31187427e+00 9.03065443e-01 9.64805260e-02
-6.02641031e-02 1.60919100e-01 -9.32897478e-02 4.53949213e-01
-1.34386644e-01 2.09641799e-01 -1.06411219e+00 4.73106563e-01
-1.83839810e+00 -8.96587253e-01 -9.96647477e-01 2.73342705e+00
3.48046124e-01 1.85958490e-01 -9.40954983e-02 3.36097598e-01
8.00044298e-01 1.15511909e-01 -3.59200269e-01 -3.12670320e-01
-1.10604346e-01 2.07086772e-01 7.88595200e-01 8.73872876e-01
-7.04517841e-01 6.50624335e-01 5.05862284e+00 4.56483334e-01
-1.41596889e+00 2.72559136e-01 -1.01138525e-01 1.20554507e-01
-5.80031872e-01 3.39921504e-01 -5.48071086e-01 6.82548523e-01
7.29064584e-01 -1.06086224e-01 3.88615459e-01 4.72438514e-01
5.13042808e-01 -7.23640919e-01 -4.89959031e-01 1.07854342e+00
-3.10089916e-01 -1.54340160e+00 -2.68395960e-01 4.31955904e-01
5.99063516e-01 2.45813638e-01 -3.36822152e-01 -6.35450706e-02
-4.89236042e-02 -6.88704908e-01 1.02958548e+00 7.93966711e-01
1.02082872e+00 -5.47509849e-01 4.04936820e-01 5.29968739e-01
-1.61811662e+00 1.66163415e-01 -7.80574381e-02 -2.14436159e-01
6.35403037e-01 1.04458380e+00 -8.49271178e-01 1.05766201e+00
6.93977594e-01 5.20420313e-01 -6.51845157e-01 1.30594289e+00
1.36337221e-01 -1.28262520e-01 -7.05860138e-01 2.82581151e-01
-8.88580382e-02 -7.82745004e-01 9.07991827e-01 8.01649392e-01
6.50549591e-01 -1.40818849e-01 1.83047831e-01 9.07082379e-01
5.11459768e-01 1.16166010e-01 -8.11847687e-01 6.72318757e-01
7.80096829e-01 1.13466597e+00 -8.83621752e-01 -1.93680272e-01
-1.10945687e-01 7.08810389e-01 1.04751594e-01 4.59547266e-02
-7.15737522e-01 -1.77941248e-02 6.61533237e-01 6.78556085e-01
1.60816237e-01 -6.22369587e-01 -4.82071936e-01 -8.62132370e-01
1.91601381e-01 -3.13183576e-01 -3.21200453e-02 -6.65858746e-01
-5.34448087e-01 5.11040926e-01 1.09866925e-01 -1.52781653e+00
-1.51360452e-01 2.25811396e-02 -4.90687072e-01 9.81258154e-01
-1.60000908e+00 -9.09840882e-01 -8.88602853e-01 4.17895645e-01
5.34928739e-01 3.57751310e-01 6.30205095e-01 7.03446269e-01
-2.99127251e-01 -4.14342582e-01 1.00257002e-01 -6.60874188e-01
3.36543739e-01 -1.11729312e+00 3.19195539e-01 8.43828857e-01
-1.09315947e-01 -2.63754819e-02 1.03374994e+00 -9.68827546e-01
-1.11989164e+00 -1.00209785e+00 8.17527771e-01 -9.20678228e-02
4.03999954e-01 -1.99144632e-01 -1.13077891e+00 3.79536778e-01
-2.69960225e-01 -1.99675649e-01 -2.19262894e-02 -7.17263818e-02
1.75470546e-01 -3.49856555e-01 -1.37730885e+00 4.46848422e-01
9.43662405e-01 -8.35375190e-02 -2.38982394e-01 6.01902083e-02
3.96420389e-01 -5.41641355e-01 -7.41317987e-01 6.89242125e-01
6.41346455e-01 -1.32369173e+00 7.28482425e-01 2.59388030e-01
-1.35384962e-01 -8.10023189e-01 -1.70310006e-01 -1.21377790e+00
-1.76152065e-01 -1.94017351e-01 4.31760401e-01 1.17134559e+00
1.24375008e-01 -8.62363756e-01 5.66011608e-01 3.99720848e-01
-9.71996561e-02 -5.37553132e-01 -1.57429397e+00 -6.65625036e-01
-4.64257747e-01 -5.50573289e-01 1.02106714e+00 1.05356967e+00
-5.60793936e-01 3.80745344e-02 2.41993237e-02 4.94243562e-01
5.32573104e-01 2.14302003e-01 1.24236763e+00 -2.06961751e+00
3.20400059e-01 -5.93898773e-01 -6.58727825e-01 -5.08605480e-01
-3.99927437e-01 -5.66947222e-01 -1.60116971e-01 -1.79409766e+00
-4.57290351e-01 -1.17573988e+00 5.25313735e-01 -1.30932912e-01
3.01130652e-01 1.15023427e-01 -6.53320625e-02 5.80707669e-01
5.92781492e-02 3.90881628e-01 7.53768325e-01 1.60550438e-02
-6.20593429e-01 -8.23915005e-02 2.18472809e-01 6.19514585e-01
6.05944514e-01 -4.24474388e-01 -2.50231445e-01 -5.95528543e-01
6.31457925e-01 9.81229767e-02 3.70997071e-01 -1.35351491e+00
2.61205226e-01 -3.46496463e-01 -9.33023617e-02 -1.18719411e+00
6.10942304e-01 -1.30688655e+00 1.17221808e+00 6.01150513e-01
6.04707181e-01 2.79659867e-01 2.19547287e-01 5.46993375e-01
-9.53793824e-02 -2.10265756e-01 9.19541836e-01 -1.25005618e-01
-6.57852888e-01 3.70915741e-01 -2.37733066e-01 -5.33544421e-01
1.26132917e+00 -7.13570356e-01 1.32153079e-01 -1.23559646e-01
-6.09157503e-01 3.29047054e-01 1.14084947e+00 1.70429826e-01
4.77783322e-01 -1.33280694e+00 -6.70229137e-01 3.73015195e-01
1.39122561e-01 5.31119287e-01 3.23575199e-01 9.83751655e-01
-8.92077267e-01 2.22217198e-02 -2.41758198e-01 -9.67986524e-01
-1.27072024e+00 2.12080583e-01 2.80570418e-01 -3.26846749e-01
-6.69557512e-01 2.64482647e-01 -4.45280820e-01 -4.49687898e-01
-2.69760489e-01 -4.38138515e-01 -3.86180580e-01 5.36991537e-01
1.52586535e-01 6.35669351e-01 3.88475716e-01 -8.01485837e-01
-3.62122983e-01 9.92273629e-01 6.37440562e-01 -3.77504587e-01
1.37866795e+00 -5.94637990e-01 -3.98379475e-01 8.14906478e-01
6.96565688e-01 3.16201419e-01 -1.00487804e+00 -9.89609510e-02
1.93513006e-01 -8.17732930e-01 -7.91271105e-02 -2.69021273e-01
-8.48643959e-01 5.07210374e-01 7.37701237e-01 4.77208048e-01
9.83069539e-01 -1.75889879e-01 6.32543147e-01 -1.26937162e-02
8.68474066e-01 -1.06830740e+00 -6.21948123e-01 4.58655953e-01
1.07357264e+00 -1.02363193e+00 2.93824583e-01 -6.06597424e-01
-1.32265493e-01 1.12774801e+00 1.45510701e-03 8.05479661e-02
7.69390106e-01 2.32940558e-02 -1.48504853e-01 -4.19820786e-01
-2.10126594e-01 -4.93959308e-01 -1.04283012e-01 4.21054244e-01
-2.81109899e-01 1.24369055e-01 -4.99763250e-01 -2.38370359e-01
-4.75328863e-01 -2.14230246e-03 5.71169138e-01 9.03189063e-01
-6.64846480e-01 -1.18436015e+00 -8.26719582e-01 3.85472447e-01
3.62167418e-01 2.48533413e-01 -6.59421459e-03 8.21537018e-01
5.68102062e-01 7.45901644e-01 3.03236067e-01 -2.37302467e-01
7.39373326e-01 -4.45301235e-02 3.56708229e-01 -6.03599668e-01
-1.62042320e-01 -1.63757727e-01 1.37794018e-02 -5.42698503e-01
-3.65225971e-01 -1.18898690e+00 -1.16900277e+00 -3.24539691e-01
-3.22424680e-01 -4.32703644e-02 1.35221589e+00 8.06996882e-01
3.92473966e-01 2.38991767e-01 7.37708569e-01 -1.22156155e+00
1.29836991e-01 -7.80378282e-01 -5.01177907e-01 4.64379877e-01
2.03173950e-01 -1.07951581e+00 -2.43089527e-01 -2.43193448e-01]
|
[8.370725631713867, -2.6152114868164062]
|
594f4953-45cc-4a06-816c-bd8c53c59cb5
|
generalization-in-transfer-learning
|
1909.01331
| null |
https://arxiv.org/abs/1909.01331v2
|
https://arxiv.org/pdf/1909.01331v2.pdf
|
Generalization in Transfer Learning
|
Agents trained with deep reinforcement learning algorithms are capable of performing highly complex tasks including locomotion in continuous environments. We investigate transferring the learning acquired in one task to a set of previously unseen tasks. Generalization and overfitting in deep reinforcement learning are not commonly addressed in current transfer learning research. Conducting a comparative analysis without an intermediate regularization step results in underperforming benchmarks and inaccurate algorithm comparisons due to rudimentary assessments. In this study, we propose regularization techniques in deep reinforcement learning for continuous control through the application of sample elimination, early stopping and maximum entropy regularized adversarial learning. First, the importance of the inclusion of training iteration number to the hyperparameters in deep transfer reinforcement learning will be discussed. Because source task performance is not indicative of the generalization capacity of the algorithm, we start by acknowledging the training iteration number as a hyperparameter. In line with this, we introduce an additional step of resorting to earlier snapshots of policy parameters to prevent overfitting to the source task. Then, to generate robust policies, we discard the samples that lead to overfitting via a method we call strict clipping. Furthermore, we increase the generalization capacity in widely used transfer learning benchmarks by using maximum entropy regularization, different critic methods, and curriculum learning in an adversarial setup. Subsequently, we propose maximum entropy adversarial reinforcement learning to increase the domain randomization. Finally, we evaluate the robustness of these methods on simulated robots in target environments where the morphology of the robot, gravity, and tangential friction coefficient of the environment are altered.
|
['Suzan Ece Ada', 'Emre Ugur', 'H. Levent Akin']
|
2019-09-03
| null | null | null | null |
['transfer-reinforcement-learning']
|
['methodology']
|
[ 4.53394443e-01 2.58020103e-01 1.55495107e-01 1.94978192e-01
-5.63869655e-01 -5.31187713e-01 6.16967976e-01 1.32060468e-01
-9.11392987e-01 1.27930522e+00 -1.67527169e-01 -8.13714862e-02
-2.23432511e-01 -7.05192566e-01 -1.11472058e+00 -9.04882133e-01
-1.20440431e-01 3.60006034e-01 1.30103588e-01 -4.81946141e-01
1.38707787e-01 6.16975129e-01 -1.36913693e+00 -2.26249576e-01
9.10945296e-01 6.02907121e-01 1.58327352e-02 4.80974406e-01
3.92026335e-01 6.21491015e-01 -7.72847533e-01 -3.20426598e-02
4.44956392e-01 -4.08950567e-01 -6.67382419e-01 1.04507096e-01
5.77899702e-02 -4.12857473e-01 5.91180697e-02 9.09462690e-01
4.90195066e-01 6.04412556e-01 8.62301588e-01 -1.12440431e+00
-1.21364355e-01 4.46861833e-01 -1.59334630e-01 -4.71581742e-02
1.12099096e-01 4.73547280e-01 4.96851176e-01 -3.45487505e-01
6.26018703e-01 1.02920413e+00 6.38931811e-01 5.99900723e-01
-1.39028645e+00 -3.75762075e-01 9.69865173e-02 -1.97756574e-01
-7.59237945e-01 -1.86298475e-01 7.46970057e-01 -5.19992352e-01
6.57457590e-01 -1.53966501e-01 5.26307881e-01 1.47517169e+00
1.70431256e-01 2.34599367e-01 1.13773620e+00 -4.29590583e-01
7.82992363e-01 2.15044528e-01 -4.46854740e-01 4.56154168e-01
3.20595026e-01 6.04522884e-01 9.95437056e-02 -1.37129515e-01
6.22893453e-01 -3.51925999e-01 -1.28478393e-01 -6.58504725e-01
-9.26690936e-01 8.77153993e-01 5.36142766e-01 2.41274983e-01
-4.56872374e-01 3.40918541e-01 7.00000882e-01 5.01345813e-01
1.60714105e-01 1.05139136e+00 -4.92977947e-01 -1.12836637e-01
-3.11809331e-01 4.97156352e-01 7.30630159e-01 4.87621754e-01
7.14724481e-01 4.57069755e-01 -2.72045508e-02 6.63653791e-01
-2.28217527e-01 3.24015528e-01 6.78409278e-01 -1.18120039e+00
3.12378794e-01 4.97490972e-01 3.21051329e-01 -6.70993090e-01
-3.99103373e-01 -6.27698421e-01 -5.11963308e-01 9.46965456e-01
6.17577016e-01 -5.11126578e-01 -7.44189322e-01 1.77440441e+00
3.73106569e-01 -4.55265827e-02 3.98361921e-01 8.32228839e-01
-1.39779255e-01 3.47898096e-01 3.07882816e-01 -1.30193755e-01
7.50846744e-01 -7.76665628e-01 -4.22480285e-01 -2.07539603e-01
6.56027377e-01 -3.01961333e-01 1.31817889e+00 4.28085715e-01
-8.86521876e-01 -5.06119967e-01 -1.16269267e+00 3.88430059e-01
-5.73512256e-01 -3.63851059e-03 3.15002531e-01 3.36696684e-01
-4.94914293e-01 1.17226684e+00 -9.40905333e-01 -2.32056662e-01
2.52525747e-01 4.99733359e-01 -2.96754539e-01 3.00841540e-01
-1.19402397e+00 1.19407713e+00 7.09138811e-01 -1.35930344e-01
-1.02874529e+00 -5.74550033e-01 -7.10636318e-01 -5.88857802e-03
4.69518274e-01 -6.65647328e-01 9.22990918e-01 -1.28699219e+00
-2.01455951e+00 3.77687365e-01 6.35240078e-01 -7.56668031e-01
9.64890659e-01 -3.51170897e-01 7.55564570e-02 9.10378620e-02
-1.44862682e-01 6.81601048e-01 1.04145503e+00 -1.45473063e+00
-2.22476259e-01 -1.55040935e-01 2.85198659e-01 2.66117901e-01
-3.35231274e-01 -4.63957846e-01 2.79518485e-01 -6.81685805e-01
-3.94284576e-01 -1.16645873e+00 -3.67855430e-01 -1.81649938e-01
6.12800866e-02 2.26947665e-01 5.97745597e-01 -4.47251827e-01
6.33712649e-01 -2.08750963e+00 4.82098877e-01 2.42537647e-01
-3.41741443e-01 3.25805247e-01 -2.19449952e-01 3.96982938e-01
-6.50042742e-02 -1.13054425e-01 -6.21232808e-01 3.67321558e-02
2.38130745e-02 4.08876330e-01 -3.31132263e-01 4.19920534e-01
3.64493221e-01 6.00794435e-01 -1.05357444e+00 -9.97599140e-02
2.66359180e-01 3.95225167e-01 -8.15912724e-01 1.45148873e-01
-4.02497619e-01 7.88647771e-01 -5.76250553e-01 1.15731515e-01
1.12773716e-01 2.68038958e-01 7.71621754e-03 1.40598074e-01
1.06979988e-03 -8.30016285e-02 -9.72695112e-01 1.45621657e+00
-7.14547992e-01 2.82372653e-01 -6.42539635e-02 -1.28516102e+00
1.03510010e+00 1.17531352e-01 4.63527173e-01 -5.38972974e-01
3.61809283e-01 2.96813011e-01 1.64913610e-01 -4.45647001e-01
3.10392976e-01 -1.29321501e-01 3.02342921e-02 6.19565956e-02
1.72583103e-01 -5.01161397e-01 2.04812497e-01 -3.47431034e-01
1.13812828e+00 5.93273759e-01 1.70704409e-01 -2.35986412e-01
5.56467056e-01 3.79937589e-02 4.06230718e-01 5.90997100e-01
-2.70134062e-01 2.60158181e-01 5.45880079e-01 -2.37969995e-01
-1.27805531e+00 -9.19238448e-01 -2.81004906e-02 1.13311613e+00
-1.58497810e-01 6.66183457e-02 -9.06177104e-01 -7.99771070e-01
1.15006760e-01 8.05112958e-01 -8.92356455e-01 -6.85921848e-01
-8.67157996e-01 -7.71085083e-01 6.25571311e-01 3.95679981e-01
4.45736945e-01 -1.44970000e+00 -1.10139883e+00 3.30869973e-01
2.18494192e-01 -9.41815019e-01 9.25026685e-02 6.65870428e-01
-8.71720791e-01 -1.06446791e+00 -7.59886146e-01 -5.80312908e-01
6.45497561e-01 -6.98488832e-01 6.83661640e-01 -1.23124324e-01
-1.18968576e-01 3.44829649e-01 -3.74493688e-01 -2.54024059e-01
-8.29234421e-01 2.88224608e-01 1.74670458e-01 -4.11534220e-01
-3.58194053e-01 -7.30568647e-01 -5.12909412e-01 2.37704098e-01
-8.93581569e-01 -4.41138476e-01 5.29697597e-01 1.13871145e+00
2.90426314e-01 -3.50285843e-02 8.14191043e-01 -6.82174563e-01
7.79075205e-01 -4.09246922e-01 -7.52277195e-01 -2.97720078e-02
-5.03706455e-01 4.44643587e-01 1.15833509e+00 -8.27866137e-01
-9.30507660e-01 1.18848853e-01 -5.76278046e-02 -4.27481562e-01
-3.12833726e-01 2.59519279e-01 2.17172410e-02 -1.81164503e-01
1.01553380e+00 5.44241108e-02 3.43104452e-01 -1.61373671e-02
1.07100435e-01 2.07198206e-02 3.26541126e-01 -8.53721201e-01
8.31560552e-01 2.74705231e-01 2.20836312e-01 -6.05936468e-01
-4.32506502e-01 2.40018472e-01 -3.85442138e-01 -1.78847939e-01
6.38302743e-01 -3.27921599e-01 -7.65904486e-01 3.56568277e-01
-6.94441319e-01 -9.20851052e-01 -6.93113804e-01 5.57778835e-01
-1.13662994e+00 2.52316177e-01 -3.97816509e-01 -6.69805050e-01
-1.31707996e-01 -1.20614624e+00 5.93906224e-01 8.51095244e-02
-7.06575885e-02 -1.11087823e+00 2.31724367e-01 -1.12707994e-03
3.92399430e-01 7.04554617e-01 8.10868263e-01 -7.29552746e-01
3.40085104e-02 4.79642041e-02 4.08120960e-01 6.30050778e-01
5.37704788e-02 -2.33025905e-02 -7.89148510e-01 -5.70004582e-01
-9.76417866e-03 -6.81796491e-01 7.61386156e-01 1.87715709e-01
9.11310792e-01 -3.25960964e-01 -3.50067243e-02 5.01354992e-01
1.27413654e+00 3.19131702e-01 5.04133701e-01 1.01292670e+00
2.97528535e-01 7.49744594e-01 6.68200016e-01 4.30226386e-01
-2.76022822e-01 5.66230893e-01 6.81523204e-01 1.57418922e-01
2.28263661e-01 -2.16181204e-01 5.79011142e-01 1.35749161e-01
-1.46222398e-01 5.25252447e-02 -6.87349856e-01 3.46498430e-01
-1.65067375e+00 -8.03868473e-01 4.63905632e-01 2.23454881e+00
6.47724450e-01 5.02625942e-01 3.66824687e-01 2.86728859e-01
5.69670022e-01 -2.83722878e-01 -7.28261113e-01 -6.20000124e-01
1.35895982e-01 8.26602727e-02 6.56456470e-01 4.40722674e-01
-1.04480088e+00 8.76228809e-01 5.05093861e+00 5.55506229e-01
-1.27182424e+00 -3.24898809e-01 3.90281975e-01 7.74542093e-02
1.12750955e-01 -6.07283302e-02 -3.72879237e-01 4.68496829e-01
8.69502425e-01 1.93595409e-01 8.16961765e-01 9.50087786e-01
2.39435598e-01 -7.05253184e-02 -9.96507168e-01 2.31687441e-01
-3.30184370e-01 -6.79698229e-01 -2.05933854e-01 -1.93539020e-02
6.85751975e-01 -1.29729316e-01 2.05188021e-01 6.21074498e-01
4.01323706e-01 -8.41128826e-01 5.74476421e-01 3.59238952e-01
3.94984365e-01 -8.34200382e-01 7.13076830e-01 4.05784637e-01
-5.38268209e-01 -4.37255204e-01 -2.62250751e-01 -6.10975586e-02
-9.90252644e-02 3.45667675e-02 -8.81401956e-01 4.78145033e-01
3.77139807e-01 3.23610604e-01 -2.72614688e-01 7.80502737e-01
-2.00737432e-01 4.96357322e-01 -3.29386652e-01 -3.56251821e-02
4.50831562e-01 -2.41629645e-01 6.82861030e-01 8.09918642e-01
2.05189973e-01 -2.35240430e-01 -3.76191624e-02 8.52762520e-01
1.40131414e-01 -5.52338921e-03 -8.46906066e-01 1.99688882e-01
2.32224822e-01 9.07387972e-01 -6.15768075e-01 4.04618122e-02
7.69986212e-02 8.39072049e-01 5.24834991e-01 5.77628076e-01
-9.07775104e-01 -3.01650017e-01 3.38133901e-01 -4.78636511e-02
4.02302593e-01 -1.35951877e-01 -1.01512529e-01 -7.55174398e-01
-6.48982301e-02 -9.21164691e-01 1.54285252e-01 -3.07653874e-01
-1.02839947e+00 5.29775441e-01 7.80845508e-02 -1.31433523e+00
-6.45287335e-01 -5.55724919e-01 -7.00112641e-01 4.72163558e-01
-1.51803231e+00 -6.09686673e-01 -1.23235926e-01 4.29911673e-01
3.70523691e-01 -2.33158693e-01 7.84814954e-01 -4.25818041e-02
-5.59116244e-01 5.65594137e-01 3.20921034e-01 -2.95426883e-02
7.08407640e-01 -1.25859463e+00 -1.47833554e-02 4.46974039e-01
-5.17594993e-01 4.04478520e-01 1.06871796e+00 -4.56739575e-01
-8.58470201e-01 -1.15522265e+00 -4.24920730e-02 3.04935640e-03
7.60156751e-01 -1.83159426e-01 -1.06833351e+00 5.14447212e-01
-2.39292253e-03 2.00524628e-02 1.14596218e-01 -1.89751521e-01
1.26003519e-01 -1.61045671e-01 -1.30336237e+00 6.81643546e-01
5.49779534e-01 -8.86032060e-02 -6.46857202e-01 1.09596528e-01
5.63287497e-01 -2.92191416e-01 -9.49868441e-01 7.15973318e-01
4.68352169e-01 -6.50140941e-01 8.87111664e-01 -7.41000593e-01
5.59588194e-01 -4.73646037e-02 1.06285788e-01 -1.74670708e+00
3.54838595e-02 -4.90250230e-01 1.47460461e-01 1.00981486e+00
3.16520780e-01 -6.49959207e-01 7.92023420e-01 2.27359816e-01
-2.19830945e-01 -7.92762756e-01 -8.02445590e-01 -9.05957460e-01
6.91787601e-01 1.58202022e-01 1.79885447e-01 8.77827287e-01
3.53772007e-02 -1.20339673e-02 -2.34138042e-01 3.46395187e-02
5.05527496e-01 -3.01744640e-01 8.78704965e-01 -9.48728085e-01
-3.64876539e-01 -3.58924270e-01 -1.16509564e-01 -3.45157385e-01
5.85541070e-01 -4.78993982e-01 2.17990845e-01 -1.02830100e+00
-4.20167774e-01 -4.14738894e-01 -2.54912823e-01 3.92623872e-01
-1.12232007e-01 -1.26357570e-01 1.61274806e-01 1.07117139e-01
-2.09027797e-01 9.15000796e-01 1.35622346e+00 -9.82785523e-02
-4.63984400e-01 9.56790149e-02 5.18067963e-02 8.45142782e-01
1.20117605e+00 -4.28065538e-01 -4.45486575e-01 -5.45062907e-02
4.95884158e-02 9.27209482e-02 5.30387402e-01 -1.28996027e+00
-2.24897489e-01 -1.34149626e-01 3.24432731e-01 3.23297292e-01
2.60479122e-01 -1.01369321e+00 -2.56001204e-01 8.94353569e-01
-5.38283169e-01 -8.81966129e-02 4.22423571e-01 5.25911510e-01
-4.02932502e-02 -5.01224995e-01 9.84439373e-01 -1.30810812e-01
-5.01450896e-01 -2.74019837e-01 -6.47944272e-01 1.80649400e-01
1.12622356e+00 -2.47966364e-01 -4.07426171e-02 -1.17185943e-01
-9.10131156e-01 2.08112121e-01 5.87645292e-01 2.47937649e-01
2.30594352e-01 -7.95748472e-01 -4.70421225e-01 1.18469052e-01
-1.43989235e-01 -1.69256672e-01 -5.81198446e-02 5.64911544e-01
-5.07497489e-01 3.67492661e-02 -7.79218853e-01 -2.56900370e-01
-7.09447265e-01 8.10558796e-01 6.17777765e-01 -3.41583252e-01
-4.52335447e-01 1.68825775e-01 -1.73505489e-02 -5.96325994e-01
3.21472079e-01 -4.04090673e-01 -2.78899282e-01 -1.56899229e-01
-2.39862084e-01 5.55596590e-01 1.37013406e-01 -5.91940582e-02
-1.13690838e-01 3.75216633e-01 6.45859465e-02 -1.36462927e-01
1.22899926e+00 1.27462342e-01 4.44224268e-01 2.68313915e-01
9.96129572e-01 -1.83278561e-01 -1.86857450e+00 3.74374628e-01
1.78274866e-02 1.27175927e-01 -2.25772709e-01 -7.17989087e-01
-7.33846843e-01 6.19602203e-01 7.00658202e-01 9.32948813e-02
9.88505900e-01 -5.05742192e-01 2.83090144e-01 6.56469703e-01
2.46815786e-01 -1.36590075e+00 3.75504613e-01 6.90463722e-01
1.03538585e+00 -1.28361511e+00 -8.64936784e-02 7.53646791e-02
-5.79485357e-01 1.09672582e+00 8.94642353e-01 -6.24272525e-01
2.20224231e-01 3.37675959e-01 6.16156124e-03 2.54239529e-01
-5.35092413e-01 -7.81486183e-02 -7.86947832e-02 5.88749409e-01
-3.59985232e-02 -2.66795844e-01 -4.88134325e-01 1.13576904e-01
-2.17012852e-01 4.23150632e-04 5.03551483e-01 9.95321393e-01
-4.73890305e-01 -9.87866759e-01 -3.37679684e-01 -4.08809893e-02
-4.16195154e-01 2.11597398e-01 -1.65624812e-01 1.19259071e+00
3.13684456e-02 5.25198996e-01 -1.47243321e-01 -1.23982832e-01
3.70386243e-01 5.39436713e-02 5.58796465e-01 -3.80235285e-01
-7.65153348e-01 -2.60512739e-01 -1.16342731e-01 -2.40731061e-01
-2.05507740e-01 -5.20820677e-01 -1.33681083e+00 1.30274877e-01
-2.22343639e-01 2.76689589e-01 6.51960552e-01 9.25205827e-01
7.47798532e-02 8.50776672e-01 5.03069103e-01 -1.01867318e+00
-1.22333086e+00 -9.99134660e-01 -1.90458044e-01 5.67052960e-01
4.34097648e-01 -1.04418635e+00 -6.31655037e-01 -6.56902269e-02]
|
[4.297308921813965, 1.88939368724823]
|
38e96b5c-aa6d-4d6d-be81-bf05e2db38b7
|
towards-responsible-ai-for-financial
|
2206.02419
| null |
https://arxiv.org/abs/2206.02419v1
|
https://arxiv.org/pdf/2206.02419v1.pdf
|
Towards Responsible AI for Financial Transactions
|
The application of AI in finance is increasingly dependent on the principles of responsible AI. These principles - explainability, fairness, privacy, accountability, transparency and soundness form the basis for trust in future AI systems. In this study, we address the first principle by providing an explanation for a deep neural network that is trained on a mixture of numerical, categorical and textual inputs for financial transaction classification. The explanation is achieved through (1) a feature importance analysis using Shapley additive explanations (SHAP) and (2) a hybrid approach of text clustering and decision tree classifiers. We then test the robustness of the model by exposing it to a targeted evasion attack, leveraging the knowledge we gained about the model through the extracted explanation.
|
['Christian W. Omlin', 'Jan Erik Modal', 'Charl Maree']
|
2022-06-06
| null | null | null | null |
['text-clustering']
|
['natural-language-processing']
|
[-2.86737308e-02 9.45453227e-01 -2.01785371e-01 -6.43441677e-01
-5.22495061e-02 -5.60379744e-01 9.10574138e-01 2.99614906e-01
-9.49642807e-02 7.44010866e-01 2.91941077e-01 -7.44440436e-01
-2.89430618e-01 -5.63515902e-01 -7.34334648e-01 -3.87295932e-01
-6.16567880e-02 3.66060048e-01 -5.56236029e-01 7.49854976e-03
5.50630450e-01 4.62842882e-01 -1.27521443e+00 5.21379352e-01
8.13996673e-01 1.14644408e+00 -9.61320639e-01 6.64533019e-01
2.01787204e-01 1.44798398e+00 -4.42142814e-01 -1.29718053e+00
6.34826779e-01 -4.15568650e-01 -8.75798047e-01 -2.10272148e-01
-3.15135159e-02 -8.56314898e-01 -2.54558865e-03 1.06461620e+00
-1.41671658e-01 -2.74421692e-01 9.37675416e-01 -1.84617043e+00
-1.19150257e+00 1.24229681e+00 -2.46097535e-01 -2.24343956e-01
-9.30652097e-02 3.54578525e-01 1.38157582e+00 -6.39077544e-01
2.08744809e-01 1.15560579e+00 3.96152139e-01 6.98330879e-01
-1.15641272e+00 -5.79674363e-01 -9.36587974e-02 1.76778540e-01
-5.58342576e-01 -2.85616100e-01 5.79708159e-01 -4.45640296e-01
9.73608851e-01 5.28887391e-01 8.41559350e-01 1.13985920e+00
4.28407282e-01 4.94875789e-01 1.11184251e+00 -5.86231947e-01
6.45178974e-01 9.51335549e-01 4.48701560e-01 3.51372600e-01
1.02755213e+00 4.07240152e-01 -4.56939608e-01 -4.83294547e-01
4.09048915e-01 1.25783473e-01 1.73808143e-01 -4.01641339e-01
-7.49763072e-01 1.48806012e+00 5.43348372e-01 6.04662001e-02
-7.57432461e-01 4.60609436e-01 3.41113180e-01 4.75487590e-01
4.46661919e-01 8.80506217e-01 -7.27365017e-01 3.00408304e-01
-5.61889410e-01 2.15508491e-01 1.15256917e+00 3.32105577e-01
2.96408921e-01 3.48029941e-01 3.00428923e-02 -2.55205721e-01
9.23281312e-01 2.56951809e-01 4.80033666e-01 -1.25073695e+00
4.74627502e-02 7.59325802e-01 4.25321013e-02 -9.97709095e-01
5.14373928e-03 -4.16117907e-01 -5.20115733e-01 8.33681762e-01
3.30755353e-01 -2.86181629e-01 -5.66702485e-01 1.59020567e+00
6.99725673e-02 -5.33155918e-01 5.45614660e-01 8.70948315e-01
5.67806542e-01 1.42314956e-01 2.67251253e-01 9.98647287e-02
1.19148433e+00 -7.14732707e-01 -6.97811842e-01 -1.17395557e-01
5.85854173e-01 -5.98516017e-02 7.40924418e-01 5.85667372e-01
-1.00798357e+00 -3.89179811e-02 -1.22292089e+00 -7.23914877e-02
-4.69764024e-01 -1.55094594e-01 9.58387196e-01 1.02799821e+00
-1.00388074e+00 8.87283444e-01 -2.67592072e-01 -4.28456552e-02
8.15146327e-01 7.04207540e-01 -5.24010777e-01 4.41626370e-01
-1.27180612e+00 8.14943671e-01 2.75678039e-01 -1.92172959e-01
-5.67367971e-01 -2.59901196e-01 -4.95340556e-01 7.46746063e-01
-6.66091964e-02 -8.76676798e-01 9.67282474e-01 -1.65756118e+00
-1.40566874e+00 7.84528017e-01 5.70179880e-01 -1.17978358e+00
8.10048342e-01 -3.86962414e-01 8.44536647e-02 1.15610108e-01
-1.19236059e-01 4.98944283e-01 7.51546025e-01 -1.43290186e+00
-4.06530231e-01 -5.71955442e-01 -6.89554140e-02 -1.95174515e-01
-2.91437358e-01 -5.27556948e-02 7.56873906e-01 -4.59373623e-01
-2.38817751e-01 -7.19406724e-01 -1.76676348e-01 -1.25967383e-01
-7.31925905e-01 6.54328465e-02 5.25713325e-01 -7.86855102e-01
6.52489483e-01 -2.06214833e+00 -4.07622784e-01 6.56906247e-01
7.67228007e-01 2.50181369e-02 2.83592314e-01 2.62947142e-01
-4.40972567e-01 7.15661705e-01 4.43575643e-02 -3.22056800e-01
5.99857152e-01 -1.96593389e-01 -7.04439044e-01 3.55582088e-01
1.67604297e-01 1.07368183e+00 -3.77942055e-01 -9.47714001e-02
-3.22816637e-03 3.61792296e-01 -8.45142126e-01 2.50776321e-01
-2.08185032e-01 3.44263501e-02 -5.13808250e-01 3.61491084e-01
4.20401633e-01 -3.29388738e-01 2.99972743e-01 1.44256875e-01
1.19763553e-01 3.93739104e-01 -6.62817836e-01 7.18972802e-01
3.41872543e-01 6.49341941e-01 -3.67797077e-01 -8.83030891e-01
9.76311505e-01 5.01310825e-01 1.43540353e-01 -4.31553990e-01
6.26565456e-01 4.42607284e-01 3.41947675e-01 -4.05085802e-01
1.96932718e-01 -5.98283187e-02 -9.96373501e-03 1.23115540e+00
-3.58546138e-01 2.03455687e-01 -5.93494356e-01 5.52383721e-01
8.77443492e-01 -1.41407192e-01 6.50021017e-01 -3.70293975e-01
1.55722156e-01 7.04738200e-02 1.78548411e-01 9.54312146e-01
-4.88808662e-01 1.84293792e-01 8.82273972e-01 -1.03506672e+00
-1.14315832e+00 -7.24990189e-01 1.88781425e-01 5.73314905e-01
-4.76845562e-01 -3.56603898e-02 -1.13153470e+00 -1.05828750e+00
3.72348249e-01 1.52279913e+00 -1.22103405e+00 -5.93131363e-01
1.15433857e-01 -3.00170243e-01 4.57322985e-01 4.03628200e-01
4.89522964e-01 -1.08230472e+00 -1.41998899e+00 -1.62976399e-01
3.27615470e-01 -5.69335818e-01 1.01617567e-01 5.24598837e-01
-7.42714167e-01 -8.67395878e-01 5.34149036e-02 9.31261927e-02
6.52094722e-01 -1.81610554e-01 9.89185452e-01 6.49561346e-01
1.67519540e-01 3.98107886e-01 -2.34737948e-01 -9.20535326e-01
-6.75723493e-01 -2.32157186e-01 2.33750120e-01 8.39078873e-02
8.56735766e-01 -5.04147708e-01 -5.87197483e-01 -3.32859248e-01
-8.94051373e-01 6.74313232e-02 7.09298849e-01 7.47974396e-01
-1.98872313e-01 -5.20878024e-02 5.49113393e-01 -1.22837353e+00
8.54049504e-01 -6.25132442e-01 -3.38399708e-01 3.21668446e-01
-1.20910192e+00 1.73129871e-01 5.57674885e-01 -2.80396342e-01
-9.73715663e-01 4.45276126e-02 5.06402075e-01 -1.42893538e-01
-2.02441454e-01 4.66092467e-01 -1.95770845e-01 3.00908312e-02
8.88751388e-01 -3.18088382e-01 1.49546728e-01 2.61400011e-03
6.65038466e-01 8.40628982e-01 3.26633424e-01 -3.32010746e-01
8.42745423e-01 4.27076459e-01 -1.00406282e-01 -3.18004899e-02
-7.41189957e-01 6.64848983e-01 -5.13383150e-01 -1.63861752e-01
8.59739900e-01 -3.62967879e-01 -1.20550656e+00 -4.51545976e-02
-1.23043084e+00 -9.98803750e-02 -5.15051126e-01 7.32913554e-01
-4.08228546e-01 -1.57035962e-02 -5.82106352e-01 -1.31390142e+00
-5.96375108e-01 -8.43296289e-01 2.01796472e-01 2.62954444e-01
-6.72993302e-01 -8.73590291e-01 -1.58602118e-01 6.31253302e-01
4.70621794e-01 4.40187097e-01 1.22555768e+00 -1.82676899e+00
-7.15199590e-01 -6.78319216e-01 -2.28698522e-01 3.46953154e-01
-8.88300911e-02 2.92867154e-01 -1.27491474e+00 1.99927405e-01
5.13165116e-01 -4.95407522e-01 6.61244929e-01 4.25653249e-01
8.88155341e-01 -1.13069189e+00 2.35865608e-01 2.24372238e-01
1.12351120e+00 2.02100739e-01 5.25934339e-01 6.83695495e-01
3.91382068e-01 1.08498311e+00 1.65336043e-01 7.46530771e-01
4.54095602e-01 8.97288695e-02 8.52270305e-01 -1.61659658e-01
7.22567737e-01 -7.49848261e-02 3.01448166e-01 -3.08054835e-02
-3.25356163e-02 -4.30172384e-02 -9.34512913e-01 1.81657672e-01
-2.15251398e+00 -1.02538443e+00 -2.47646406e-01 2.03692460e+00
4.26691353e-01 3.93046230e-01 1.29699558e-01 2.67011583e-01
4.14921731e-01 -5.29648840e-01 -7.47236729e-01 -1.22318876e+00
-4.04386781e-03 -2.23366976e-01 5.92322469e-01 2.37548694e-01
-7.90587366e-01 6.58325911e-01 6.56628895e+00 6.04264773e-02
-7.11097896e-01 -3.33092302e-01 1.33478343e+00 7.84492642e-02
-8.77400815e-01 2.13062122e-01 -9.43124220e-02 8.59201998e-02
1.01065898e+00 -5.18630266e-01 4.73239571e-01 1.17668033e+00
-2.80428231e-02 9.63924378e-02 -1.36031818e+00 1.84661552e-01
-3.35452817e-02 -1.38109875e+00 3.70662242e-01 4.89062160e-01
5.30915558e-01 -4.20575023e-01 4.92048085e-01 -6.85840994e-02
9.31488872e-01 -1.54961300e+00 1.15385914e+00 5.07545412e-01
2.07708895e-01 -6.76404774e-01 1.09604645e+00 2.11940825e-01
-1.15748465e-01 -5.04500806e-01 -2.24284843e-01 -4.99348223e-01
-5.53528488e-01 1.88732594e-01 -7.91439593e-01 3.76637846e-01
5.63385725e-01 3.98419023e-01 -5.17901778e-01 3.52962047e-01
-4.48677987e-01 6.11514151e-01 4.54392061e-02 -1.19021423e-01
1.28415316e-01 -8.65779966e-02 2.05964789e-01 6.87463284e-01
1.19587064e-01 3.12009692e-01 -6.50461674e-01 1.50159252e+00
-9.84767899e-02 -1.95166878e-02 -8.72464120e-01 -2.56097227e-01
4.21050906e-01 1.13555932e+00 -7.25132763e-01 -4.74266261e-01
-3.12552691e-01 6.69800341e-01 1.45284683e-01 2.51331359e-01
-5.59009731e-01 -8.40811711e-03 4.46646124e-01 -2.31592089e-01
1.61430344e-01 4.14347589e-01 -1.19615579e+00 -9.37605023e-01
-2.52256006e-01 -1.16857243e+00 5.99747419e-01 -8.51241469e-01
-1.38945520e+00 6.31677687e-01 -2.67664164e-01 -6.40838981e-01
-4.45483983e-01 -5.54358125e-01 -8.86155605e-01 8.00149322e-01
-1.15131617e+00 -1.13519013e+00 6.89485222e-02 3.37071151e-01
-1.86370715e-01 -6.75585747e-01 9.55779195e-01 -4.53179747e-01
-5.14262617e-01 6.88897014e-01 -1.28506660e-01 1.73570186e-01
2.77447850e-01 -1.48359382e+00 4.69746292e-01 7.29227722e-01
2.62730509e-01 9.81369734e-01 9.89763618e-01 -6.38428271e-01
-9.42218781e-01 -5.66349804e-01 1.19355106e+00 -7.30367839e-01
6.16200864e-01 -1.70050025e-01 -9.88915086e-01 9.96387422e-01
5.17453015e-01 -3.79636288e-01 1.03244078e+00 -1.98986337e-01
-6.28942072e-01 1.25026628e-01 -1.64576375e+00 5.23302078e-01
3.65849614e-01 -3.74426097e-01 -6.99133337e-01 4.59061526e-02
9.19723511e-01 2.64243394e-01 -6.47850931e-01 2.82892436e-02
9.14068103e-01 -1.45319581e+00 6.43040121e-01 -1.21834445e+00
1.00814259e+00 1.20665818e-01 -3.21530819e-01 -1.02710807e+00
-3.59434932e-01 -4.55661029e-01 -1.31478027e-01 1.08461356e+00
6.85879469e-01 -1.03200173e+00 9.09595311e-01 1.74200070e+00
4.74290848e-01 -4.14019287e-01 -7.52045631e-01 -2.29281008e-01
2.96758056e-01 -1.92660034e-01 1.14155734e+00 1.22380197e+00
3.75504047e-01 5.97238354e-02 -4.10554886e-01 1.15741909e-01
9.74694490e-01 2.90255970e-03 8.04033279e-01 -1.58987939e+00
-2.77059495e-01 -5.89385211e-01 -2.74862945e-01 5.65963462e-02
2.33883664e-01 -6.27184093e-01 -3.58491421e-01 -1.17122328e+00
4.62366700e-01 1.98944300e-01 -2.00055569e-01 5.13332486e-01
3.28215538e-03 -3.49400192e-02 3.57326001e-01 5.45988977e-01
-2.93556064e-01 4.10908848e-01 4.32495624e-01 1.81521282e-01
1.96525708e-01 -2.24353716e-01 -1.45824194e+00 8.70623171e-01
1.04175973e+00 -8.13652337e-01 -1.50780410e-01 -2.42937386e-01
5.71072161e-01 -1.51486963e-01 8.18038344e-01 -4.45586324e-01
1.22673176e-01 -2.05316871e-01 4.63792354e-01 7.92096034e-02
-1.31089352e-02 -1.19733167e+00 2.44969219e-01 1.03527808e+00
-1.03241062e+00 -3.79710011e-02 -1.08553924e-01 4.71674711e-01
2.29751300e-02 -2.40756392e-01 4.24329698e-01 -1.86882123e-01
3.61792445e-02 -1.23408288e-01 -3.50783736e-01 -6.17877126e-01
9.71231163e-01 -3.68633896e-01 -5.80496728e-01 -7.03353643e-01
-5.47288954e-01 -1.21377520e-01 6.09975994e-01 1.12804100e-02
5.37984967e-01 -1.18381381e+00 -8.22492301e-01 3.35650802e-01
-9.66690555e-02 -6.50258124e-01 -2.12329105e-01 4.99329716e-01
-3.65808159e-01 5.75231671e-01 -5.69452107e-01 3.22368816e-02
-1.09558070e+00 5.27533412e-01 4.90257949e-01 -1.61654085e-01
-3.15538824e-01 5.92212796e-01 4.46556449e-01 -2.82464892e-01
2.30933994e-01 7.55032003e-02 -5.28317273e-01 -3.91654700e-01
3.38037550e-01 1.92005530e-01 -4.86524791e-01 -3.87151241e-01
-3.12935919e-01 -1.71817809e-01 -1.61629915e-01 -3.98699284e-01
1.59111476e+00 -2.10059900e-02 -1.29425809e-01 3.20949495e-01
6.03193343e-01 -1.52214900e-01 -1.10622549e+00 8.15625191e-02
3.58939976e-01 -6.60376072e-01 6.06263205e-02 -1.15860403e+00
-1.04918194e+00 8.97289455e-01 3.40790302e-01 9.18108582e-01
6.69715703e-01 -2.58358896e-01 1.52450204e-01 4.36399609e-01
-2.17532977e-01 -8.87050509e-01 -6.81009190e-03 -1.64700612e-01
9.35284972e-01 -1.31932223e+00 1.90442920e-01 1.08098723e-01
-1.15490353e+00 1.39459264e+00 3.82864386e-01 -5.41971475e-02
6.00489378e-01 -6.25108331e-02 3.35667163e-01 -6.35257065e-01
-1.16011429e+00 4.08679873e-01 3.23578596e-01 4.62773740e-01
4.48492140e-01 2.89399475e-01 -1.70507833e-01 1.44466269e+00
-4.51202244e-01 8.23240057e-02 8.89129221e-01 4.18973267e-01
-3.29775095e-01 -6.48147285e-01 -3.09000283e-01 5.39933205e-01
-7.48678625e-01 -2.22230196e-01 -1.37015975e+00 8.25934947e-01
-1.99296966e-01 1.06343126e+00 -1.51109502e-01 -5.79287112e-01
-5.38265221e-02 8.63191411e-02 -3.31718594e-01 -2.24651158e-01
-1.13575387e+00 -4.15616721e-01 1.42896205e-01 -4.74659055e-01
-1.96581244e-01 -8.13792169e-01 -1.12533903e+00 -8.53546858e-01
-2.54302680e-01 3.90855312e-01 7.98778117e-01 1.16100717e+00
4.38916951e-01 1.66317210e-01 6.97463453e-01 -1.99392363e-01
-1.01273108e+00 -7.37828791e-01 -6.80283427e-01 5.27567327e-01
2.60415375e-01 -1.32627442e-01 -1.00219584e+00 -4.70936038e-02]
|
[8.751476287841797, 5.7056474685668945]
|
2880d717-606c-42f4-8fac-cc03fa8dd3c3
|
learning-hierarchical-metrical-structure
|
2209.10259
| null |
https://arxiv.org/abs/2209.10259v1
|
https://arxiv.org/pdf/2209.10259v1.pdf
|
Learning Hierarchical Metrical Structure Beyond Measures
|
Music contains hierarchical structures beyond beats and measures. While hierarchical structure annotations are helpful for music information retrieval and computer musicology, such annotations are scarce in current digital music databases. In this paper, we explore a data-driven approach to automatically extract hierarchical metrical structures from scores. We propose a new model with a Temporal Convolutional Network-Conditional Random Field (TCN-CRF) architecture. Given a symbolic music score, our model takes in an arbitrary number of voices in a beat-quantized form, and predicts a 4-level hierarchical metrical structure from downbeat-level to section-level. We also annotate a dataset using RWC-POP MIDI files to facilitate training and evaluation. We show by experiments that the proposed method performs better than the rule-based approach under different orchestration settings. We also perform some simple musicological analysis on the model predictions. All demos, datasets and pre-trained models are publicly available on Github.
|
['Gus Xia', 'Yixiao Zhang', 'Daniel Chin', 'Junyan Jiang']
|
2022-09-21
| null | null | null | null |
['music-information-retrieval']
|
['music']
|
[ 2.05568761e-01 -4.80043665e-02 -3.08031023e-01 -3.32089126e-01
-9.36980426e-01 -8.63710165e-01 3.65723431e-01 -1.71025753e-01
4.35176156e-02 4.91539776e-01 7.61755407e-01 1.02732316e-01
-6.36878550e-01 -5.08394122e-01 -3.92838538e-01 -2.58061051e-01
-5.86152196e-01 6.54961467e-01 1.51009336e-01 2.22295616e-02
5.10995388e-01 1.38817310e-01 -1.55577493e+00 7.54401624e-01
3.78843725e-01 1.09841037e+00 5.37669798e-03 9.85530734e-01
9.24969688e-02 1.08777785e+00 -6.05914652e-01 -2.61996359e-01
2.99049109e-01 -6.77058816e-01 -1.17348540e+00 -8.64873901e-02
4.10965919e-01 -1.09011315e-01 -5.49835041e-02 6.60364866e-01
5.04164517e-01 4.72397469e-02 4.82673556e-01 -8.04699361e-01
-2.76537240e-01 1.49407506e+00 -3.51975225e-02 -2.43603319e-01
3.13489139e-01 -1.71509340e-01 1.86198771e+00 -4.88601536e-01
6.06485248e-01 9.28546131e-01 1.01954663e+00 2.72171170e-01
-1.40156507e+00 -7.56577075e-01 -3.19867402e-01 4.43811029e-01
-1.37949109e+00 -3.73734832e-01 1.04315639e+00 -6.00060403e-01
6.87914908e-01 4.04489815e-01 1.03512859e+00 9.17220294e-01
-2.84806818e-01 7.64596224e-01 1.01835775e+00 -4.63080704e-01
1.43734172e-01 -7.78884232e-01 -2.68504977e-01 5.15183508e-01
-5.59898674e-01 6.18048459e-02 -1.24899077e+00 -2.64535904e-01
8.25296283e-01 -4.46977526e-01 1.98290169e-01 1.63208663e-01
-1.49087262e+00 5.45738935e-01 1.98178068e-01 6.97962403e-01
-3.43680859e-01 5.30855775e-01 4.49844986e-01 1.47163957e-01
-2.23352890e-02 7.82847285e-01 -6.34226441e-01 -6.36044085e-01
-1.68676853e+00 4.82149273e-01 8.21241856e-01 5.55900335e-01
4.43251431e-01 8.03263783e-02 -2.62570381e-01 9.90840018e-01
1.24777846e-01 7.39007741e-02 6.21976256e-01 -1.48843634e+00
1.53548375e-01 3.35821062e-01 -9.99011174e-02 -8.45613360e-01
-4.92825449e-01 -7.09303379e-01 -6.89966261e-01 -1.36050880e-01
5.33475816e-01 2.85602242e-01 -4.67794865e-01 1.78970969e+00
-4.75252233e-02 6.07622087e-01 -3.74018013e-01 8.58845532e-01
7.37305880e-01 4.21055287e-01 -1.96783915e-01 -2.67604351e-01
1.28656161e+00 -7.01175034e-01 -6.60963655e-01 4.57619190e-01
5.19060850e-01 -1.05293882e+00 1.17435670e+00 1.08724427e+00
-1.16594720e+00 -7.62751877e-01 -1.06137657e+00 -1.28411010e-01
3.04353565e-01 4.35597658e-01 7.22653747e-01 1.90656409e-01
-7.21190989e-01 1.39733315e+00 -8.03740799e-01 -1.25870138e-01
2.19917729e-01 4.08316702e-01 7.74583872e-03 8.16640615e-01
-9.33977723e-01 3.64690125e-01 5.09864867e-01 -7.99839944e-02
-1.01163661e+00 -5.59740901e-01 -3.23646456e-01 -3.51724364e-02
3.32973115e-02 -3.57404977e-01 1.77882874e+00 -8.36076736e-01
-1.88338542e+00 8.72536004e-01 1.16271518e-01 -6.69612527e-01
1.56533703e-01 -2.91460633e-01 -2.79424459e-01 -5.44709265e-02
6.95834234e-02 6.38923943e-01 4.64710474e-01 -9.11725044e-01
-5.55441082e-01 -2.04012096e-01 4.85080630e-02 3.86479497e-02
-3.25498849e-01 2.15472355e-01 -3.28323215e-01 -9.76884842e-01
3.13179255e-01 -1.11877310e+00 -2.74148490e-03 -5.39280832e-01
-8.18637311e-01 -3.03416371e-01 1.38460353e-01 -7.96561658e-01
1.98908913e+00 -1.85113525e+00 3.51785809e-01 2.32238531e-01
-2.37116039e-01 -3.53406608e-01 3.69185992e-02 4.02729660e-01
2.08275780e-01 1.01569444e-01 -3.12530428e-01 -2.50677586e-01
3.85527939e-01 2.68463254e-01 -3.70081335e-01 5.17187603e-02
-2.50079453e-01 7.82732785e-01 -5.78792632e-01 -8.89741540e-01
-8.84493887e-02 2.97713280e-01 -1.01622903e+00 1.08521797e-01
-4.30021375e-01 8.44296694e-01 -2.38608167e-01 6.70732498e-01
-4.51749191e-02 6.36123586e-03 4.54677969e-01 -2.45359868e-01
-3.47437978e-01 1.03162086e+00 -1.28344393e+00 2.43486047e+00
-4.29196149e-01 5.70748091e-01 -4.16424543e-01 -6.84124529e-01
1.04714859e+00 4.29756194e-01 8.59872699e-01 -3.74260187e-01
3.01893670e-02 2.43834630e-01 4.32971448e-01 -1.98056728e-01
7.40508258e-01 -3.37616950e-01 -6.46698833e-01 4.33312476e-01
3.55997384e-01 -1.86511755e-01 3.44104290e-01 -8.97167251e-02
1.21986496e+00 7.38059521e-01 4.29951727e-01 -1.23166986e-01
3.64931285e-01 4.98703234e-02 1.04712403e+00 4.36621130e-01
1.81807846e-01 5.74165463e-01 4.44165349e-01 -5.35820663e-01
-1.06313598e+00 -9.21300650e-01 -3.08118403e-01 1.52162528e+00
-3.64206046e-01 -1.36281598e+00 -7.30786204e-01 -2.91745942e-02
-3.01731050e-01 4.20322061e-01 -4.43557531e-01 3.68448406e-01
-8.68666112e-01 -3.27675015e-01 1.16202807e+00 3.89000088e-01
2.59189874e-01 -1.56846952e+00 -5.13262868e-01 6.42954826e-01
-5.12566447e-01 -8.02428782e-01 -4.42817539e-01 2.94712305e-01
-9.70973790e-01 -9.89907205e-01 -4.95310724e-02 -7.85013437e-01
-5.27991593e-01 -7.69115806e-01 1.41242301e+00 -1.44460082e-01
-3.45187783e-01 -2.99356610e-01 -3.66946518e-01 -4.27364796e-01
-3.29550505e-01 4.82794136e-01 1.47276640e-01 3.31762154e-03
1.70993745e-01 -1.35841954e+00 -5.98363161e-01 3.07452381e-01
-4.76303190e-01 1.89370736e-01 2.61010677e-01 5.24022639e-01
9.15726185e-01 -1.15902416e-01 5.38927138e-01 -5.99075377e-01
4.25601572e-01 1.16397403e-01 -3.46589953e-01 -1.95545256e-01
-3.01390290e-01 1.83699310e-01 6.68099225e-01 -4.72586870e-01
-4.80040014e-01 4.89573985e-01 -2.32193738e-01 -3.37858111e-01
-2.14915723e-01 6.47909522e-01 -1.00357816e-01 6.11115038e-01
8.67464900e-01 1.45187965e-02 -8.13863218e-01 -1.05287266e+00
5.51266134e-01 7.73536503e-01 1.07093334e+00 -1.00444853e+00
7.29543209e-01 2.36841068e-01 9.96730030e-02 -4.57378119e-01
-1.27335429e+00 -3.99463534e-01 -1.25441158e+00 -4.41677332e-01
9.82894480e-01 -7.40319490e-01 -1.03355277e+00 2.36255601e-05
-9.77408469e-01 -4.22271401e-01 -5.59877217e-01 4.65926051e-01
-1.17449582e+00 1.67178750e-01 -9.18048561e-01 -7.83865809e-01
-4.25756484e-01 -5.39617002e-01 1.03456271e+00 -3.20288986e-02
-8.79443467e-01 -5.28758287e-01 6.07294083e-01 4.34693128e-01
-5.54574048e-03 5.01226664e-01 9.06857491e-01 -5.02952993e-01
-5.11334777e-01 1.62850156e-01 5.12220562e-01 1.30563498e-01
-6.72060698e-02 7.53575563e-02 -1.06722224e+00 3.30677122e-01
-4.31715697e-01 -6.63023829e-01 6.83436275e-01 3.48531634e-01
1.44306934e+00 -3.94503415e-01 2.46257782e-01 7.01714277e-01
1.04339409e+00 -3.72842476e-02 5.67157924e-01 4.19809550e-01
5.87198138e-01 4.41567093e-01 5.66110194e-01 7.55942404e-01
2.92532921e-01 1.01295722e+00 2.76143342e-01 4.25669670e-01
-3.26506197e-01 -6.83767259e-01 4.84167784e-01 1.61764622e+00
-6.54927075e-01 3.53276730e-01 -7.75787055e-01 7.05093384e-01
-1.96221769e+00 -1.35261989e+00 -2.66842037e-01 2.06137848e+00
1.29235446e+00 1.90525070e-01 5.82135677e-01 7.00976253e-01
5.31233668e-01 8.95197764e-02 -3.05652112e-01 -4.09116298e-01
-1.45912305e-01 7.18635261e-01 1.50670588e-03 2.37439021e-01
-1.24340308e+00 1.20603633e+00 6.23374081e+00 9.21570718e-01
-8.06958556e-01 8.08912590e-02 -9.78225321e-02 -4.48416531e-01
-6.61781430e-02 4.25877869e-01 -4.03437585e-01 2.97298968e-01
1.05426705e+00 1.30222559e-01 7.51778543e-01 7.15680480e-01
3.92989486e-01 4.04734105e-01 -1.21713400e+00 1.24816120e+00
-2.17563361e-01 -1.48860085e+00 -8.20812136e-02 1.81575179e-01
6.86037540e-01 8.50050002e-02 -2.39254057e-01 2.86327809e-01
3.79880726e-01 -1.16561139e+00 1.22383416e+00 8.34020257e-01
9.97789502e-01 -7.16962814e-01 2.40990043e-01 2.88338363e-01
-1.60489047e+00 -8.08037668e-02 -7.00619444e-02 -4.55202699e-01
2.65759051e-01 2.89451003e-01 -7.89904892e-01 5.44927895e-01
6.37218535e-01 9.87295568e-01 -5.61762631e-01 1.07397270e+00
-4.81563359e-01 1.30755830e+00 -2.67878592e-01 2.66031533e-01
-5.75856529e-02 -9.73099321e-02 4.32284415e-01 1.20825064e+00
5.64902782e-01 -6.02714829e-02 3.98274213e-01 8.02329242e-01
-9.14764628e-02 3.89114857e-01 -1.80224270e-01 -1.60749927e-01
4.07969445e-01 1.37016904e+00 -7.50395298e-01 -2.33268812e-01
4.06914651e-02 7.64099896e-01 2.01296300e-01 -2.27614209e-01
-8.10068727e-01 -2.30997458e-01 4.02795643e-01 3.06368411e-01
3.25875670e-01 -4.58577454e-01 -4.76267159e-01 -1.20445371e+00
-2.07502872e-01 -9.31715012e-01 5.31237602e-01 -8.81805182e-01
-1.31324577e+00 5.57643473e-01 -2.68108070e-01 -1.70763719e+00
-6.19178653e-01 -3.15835983e-01 -5.61006069e-01 2.70623565e-01
-8.72022033e-01 -1.28218412e+00 6.35382980e-02 5.56398749e-01
4.85452563e-01 -1.78070888e-01 1.37342751e+00 4.33692843e-01
9.91759822e-03 4.12064075e-01 -2.09896132e-01 5.10079324e-01
8.84492218e-01 -1.47835207e+00 1.09775662e-01 3.54010046e-01
1.19258165e+00 3.36918235e-01 5.53442895e-01 -5.02627194e-01
-6.81085706e-01 -9.37937617e-01 1.18219626e+00 -4.98723537e-01
9.21684921e-01 -3.10542554e-01 -5.44084668e-01 6.08410776e-01
2.57465422e-01 -4.70191956e-01 1.24588716e+00 8.06484103e-01
-5.52173138e-01 -1.04156062e-01 -3.67027938e-01 3.79101276e-01
1.53260863e+00 -7.97543347e-01 -9.31600213e-01 1.07721813e-01
3.59429657e-01 -1.10399798e-01 -1.37496221e+00 6.13628149e-01
1.03475666e+00 -9.51928973e-01 6.20489180e-01 -6.24192953e-01
6.56978548e-01 -7.07790434e-01 -5.23946404e-01 -1.00491202e+00
-5.41852117e-01 -1.05180395e+00 -2.55338997e-01 1.20428681e+00
4.30672675e-01 5.12108803e-01 7.84365237e-01 -4.51491863e-01
-2.88775057e-01 -3.97279412e-01 -8.86329412e-01 -7.50303209e-01
4.30517793e-02 -1.18392265e+00 5.58333874e-01 9.31832850e-01
3.06706667e-01 7.12899148e-01 -7.26244211e-01 -1.20746151e-01
6.00909173e-01 7.44394898e-01 8.62524450e-01 -1.75058866e+00
-9.52316523e-01 -5.38946629e-01 -5.38445711e-01 -8.47159803e-01
6.29827827e-02 -1.37027061e+00 -2.81303446e-03 -1.43169379e+00
1.88972667e-01 -3.22585225e-01 -7.55707085e-01 8.50976050e-01
5.02790391e-01 9.26080287e-01 3.84344041e-01 5.64698339e-01
-9.78487492e-01 3.39892507e-01 1.03896761e+00 -1.15054846e-01
-5.13330579e-01 5.96336760e-02 -2.94566751e-01 1.08473670e+00
1.04001355e+00 -8.04881155e-01 -1.35760820e-02 -9.33403149e-02
6.49616063e-01 2.41648480e-01 2.00312138e-01 -1.41678035e+00
1.46785319e-01 -2.19537392e-02 3.44007045e-01 -9.29340065e-01
3.45678151e-01 -1.22386895e-01 3.59401166e-01 2.32485071e-01
-7.18227088e-01 -1.42632455e-01 -7.96976611e-02 1.47264898e-01
-4.75222766e-01 -8.50912109e-02 5.41175544e-01 -1.94907654e-02
-2.03815892e-01 1.38599470e-01 -2.43181825e-01 -1.21791296e-01
1.52736172e-01 6.24296479e-02 3.60865653e-01 -4.05314952e-01
-1.42482483e+00 -3.90909731e-01 1.26297936e-01 4.50610578e-01
1.13402486e-01 -1.78041160e+00 -7.70792782e-01 -2.04966351e-01
2.44143888e-01 -3.41959327e-01 8.68607759e-02 7.36154556e-01
-3.91517878e-01 3.39912951e-01 -3.49096626e-01 -6.65937185e-01
-1.24929607e+00 1.08133867e-01 1.99491400e-02 -5.43648183e-01
-5.89477956e-01 5.95723152e-01 -2.18196079e-01 -6.66340768e-01
3.52242827e-01 -5.69764197e-01 -3.23022395e-01 1.24103799e-01
3.13429981e-01 1.81887269e-01 -1.44479290e-01 -8.69971812e-01
-1.72991365e-01 7.68871367e-01 5.79449356e-01 -6.11023605e-01
1.44792473e+00 1.80211157e-01 -1.64853588e-01 1.12069178e+00
5.82926512e-01 4.15356308e-01 -1.09704769e+00 -1.26743123e-01
6.65600419e-01 -8.48731026e-02 -6.60893917e-02 -9.15075719e-01
-4.61660564e-01 7.46847510e-01 3.02616388e-01 2.81874835e-01
1.18717575e+00 1.45225823e-01 7.93188810e-01 4.46365088e-01
3.98035496e-01 -1.26697695e+00 2.13252425e-01 7.01463819e-01
9.18602705e-01 -3.61959398e-01 -6.44751787e-02 7.74614587e-02
-5.06240368e-01 1.10224366e+00 1.08510479e-01 -3.53248447e-01
6.57031417e-01 2.12383598e-01 6.62821578e-03 -9.71121639e-02
-8.90516698e-01 -5.47638416e-01 6.48915350e-01 1.26830757e-01
9.19967175e-01 4.21974957e-01 -4.66471195e-01 1.32821798e+00
-1.42386270e+00 2.29732350e-01 3.45424622e-01 4.62998211e-01
-5.03991961e-01 -1.53415442e+00 -3.48127693e-01 7.65847564e-02
-9.22447264e-01 -3.60615134e-01 -7.10772336e-01 5.06824911e-01
8.48566771e-01 9.71964598e-01 2.07471102e-02 -8.59267771e-01
1.95077077e-01 4.98789519e-01 8.44703138e-01 -6.23658955e-01
-8.80745173e-01 7.10986912e-01 3.56787175e-01 -6.06778145e-01
-7.41443574e-01 -7.30563760e-01 -1.49379265e+00 -1.44358343e-02
-4.35984991e-02 2.98672259e-01 6.87539577e-01 7.38684535e-01
1.07647225e-01 6.09465599e-01 6.16905451e-01 -9.45133984e-01
-1.60363153e-01 -1.51096523e+00 -8.15933645e-01 4.55937386e-01
-2.17481315e-01 -3.58515263e-01 -6.74762055e-02 6.01409733e-01]
|
[15.916718482971191, 5.358638286590576]
|
96937517-a7cc-4993-b9e2-864664f9c66a
|
hq-50k-a-large-scale-high-quality-dataset-for
|
2306.05390
| null |
https://arxiv.org/abs/2306.05390v1
|
https://arxiv.org/pdf/2306.05390v1.pdf
|
HQ-50K: A Large-scale, High-quality Dataset for Image Restoration
|
This paper introduces a new large-scale image restoration dataset, called HQ-50K, which contains 50,000 high-quality images with rich texture details and semantic diversity. We analyze existing image restoration datasets from five different perspectives, including data scale, resolution, compression rates, texture details, and semantic coverage. However, we find that all of these datasets are deficient in some aspects. In contrast, HQ-50K considers all of these five aspects during the data curation process and meets all requirements. We also present a new Degradation-Aware Mixture of Expert (DAMoE) model, which enables a single model to handle multiple corruption types and unknown levels. Our extensive experiments demonstrate that HQ-50K consistently improves the performance on various image restoration tasks, such as super-resolution, denoising, dejpeg, and deraining. Furthermore, our proposed DAMoE, trained on our \dataset, outperforms existing state-of-the-art unified models designed for multiple restoration tasks and levels. The dataset and code are available at \url{https://github.com/littleYaang/HQ-50K}.
|
['Nenghai Yu', 'Gang Hua', 'Lu Yuan', 'Jianmin Bao', 'Qi Chu', 'Qiankun Liu', 'Zhentao Tan', 'Dongdong Chen', 'Qinhong Yang']
|
2023-06-08
| null | null | null | null |
['super-resolution', 'image-restoration']
|
['computer-vision', 'computer-vision']
|
[ 4.42846268e-02 -5.27903497e-01 -8.64232555e-02 -2.02687711e-01
-1.30217445e+00 -9.48114172e-02 1.54976681e-01 -2.47347057e-01
5.96418343e-02 7.22075284e-01 6.05789542e-01 1.19336545e-01
-2.25819752e-01 -7.52382576e-01 -5.03893375e-01 -9.25352037e-01
9.28583071e-02 -1.61449283e-01 3.93406451e-01 -3.34172368e-01
3.54678392e-01 2.80919999e-01 -1.63662863e+00 5.26541710e-01
1.30636275e+00 1.10347235e+00 5.08592606e-01 5.77948451e-01
3.71285766e-01 9.07274723e-01 -3.45007151e-01 -2.44028330e-01
3.45313132e-01 -3.60074461e-01 -5.72016656e-01 3.11250508e-01
6.74394727e-01 -7.17111230e-01 -6.17011786e-01 1.32614660e+00
9.38663900e-01 5.10300733e-02 2.65835792e-01 -8.24467778e-01
-1.25881016e+00 2.18753919e-01 -7.61604071e-01 5.77872813e-01
1.44783571e-01 2.50698745e-01 5.25402725e-01 -1.09937394e+00
6.78792059e-01 1.24940205e+00 7.31687725e-01 2.61488259e-01
-1.24951398e+00 -6.42580926e-01 -4.17111143e-02 5.29377103e-01
-1.44502473e+00 -8.82643282e-01 6.00024581e-01 -2.62926280e-01
6.00474000e-01 2.03544736e-01 1.99847922e-01 8.59885693e-01
3.42285633e-01 6.57953978e-01 1.64568818e+00 -2.28825361e-01
1.80110604e-01 -4.13385630e-01 8.84085521e-02 2.81989336e-01
1.53018653e-01 1.30785242e-01 -8.00743222e-01 9.82105583e-02
1.19803607e+00 -1.52419776e-01 -7.84196734e-01 5.17417118e-02
-1.25096071e+00 4.46340203e-01 3.06560189e-01 6.34104162e-02
-3.61141652e-01 -9.41184834e-02 3.61389160e-01 3.45385611e-01
8.14963520e-01 1.26747135e-03 -4.03123438e-01 2.55530119e-01
-7.75105178e-01 1.11292124e-01 1.60554752e-01 9.56277251e-01
6.30336940e-01 2.16607526e-01 -4.10386771e-01 1.45434809e+00
6.14697970e-02 6.88503623e-01 3.26060057e-01 -1.34050286e+00
4.64978516e-01 1.35740668e-01 2.00259581e-01 -9.11627889e-01
-1.53235078e-01 -5.22974133e-01 -1.46882498e+00 3.56637269e-01
-1.04593515e-01 3.61206263e-01 -1.08883178e+00 1.46242678e+00
3.73044968e-01 2.57282764e-01 7.54370019e-02 1.12417722e+00
1.12260115e+00 7.18680620e-01 -4.54006493e-02 -5.00274837e-01
1.32953620e+00 -1.15435004e+00 -1.05801964e+00 -1.24406762e-01
-1.22393735e-01 -1.06532073e+00 1.07403147e+00 7.29019344e-01
-1.26856685e+00 -8.83701921e-01 -9.69685197e-01 -4.65878695e-01
6.83776196e-03 2.06364602e-01 4.05436456e-01 3.64371300e-01
-1.45105743e+00 6.97112978e-01 -4.71368939e-01 -2.84834802e-01
6.08080983e-01 -2.31417447e-01 -3.70483011e-01 -8.45541358e-01
-9.58650768e-01 7.68537223e-01 6.26765564e-02 7.68227577e-02
-1.02479351e+00 -5.39065301e-01 -7.38514304e-01 -2.90848076e-01
3.39124739e-01 -7.26522624e-01 7.99003184e-01 -3.76003236e-01
-1.19509447e+00 6.88522398e-01 -2.19596103e-01 3.73301166e-03
3.66779953e-01 -2.24104330e-01 -8.30224752e-01 3.34230155e-01
2.28238702e-01 2.56209016e-01 1.00841749e+00 -1.74392307e+00
-5.69775045e-01 -4.74546105e-01 -8.90872851e-02 2.25812554e-01
-1.01575918e-01 2.35315770e-01 -9.34311569e-01 -1.06865978e+00
2.89562792e-01 -3.25342149e-01 -2.33421832e-01 1.22377239e-01
-4.08868790e-01 2.25948080e-01 4.75210756e-01 -1.37103081e+00
1.37655437e+00 -2.26196885e+00 2.72112042e-01 -2.42681146e-01
2.86495000e-01 1.60405308e-01 -3.98815125e-01 3.15503001e-01
7.97607452e-02 1.45128384e-01 -4.53809321e-01 -5.42803168e-01
-1.69402719e-01 3.36815625e-01 -1.17691882e-01 5.84758818e-01
-1.34293303e-01 5.71161449e-01 -7.41134703e-01 -6.33735836e-01
5.17411053e-01 7.41214216e-01 -3.41040730e-01 2.61546403e-01
2.59588212e-01 5.90054750e-01 -2.11604223e-01 1.11200190e+00
1.28116691e+00 -3.27468365e-01 -7.54952729e-02 -6.24448478e-01
4.88956869e-02 -2.85995424e-01 -1.40951788e+00 1.81410360e+00
-4.34720099e-01 2.52826542e-01 2.78019100e-01 -7.27415264e-01
7.08174646e-01 2.51803815e-01 4.76543307e-01 -1.23103082e+00
-3.06423604e-01 4.12286788e-01 -5.80297053e-01 -5.12987554e-01
6.88397586e-01 6.05458431e-02 3.75332981e-01 2.65087206e-02
8.04446712e-02 1.02688566e-01 3.89932901e-01 2.12035343e-01
9.41780627e-01 -5.03438860e-02 2.07335085e-01 -3.58363241e-01
4.13343430e-01 -2.26499587e-01 9.36919987e-01 7.73810267e-01
-3.30462068e-01 1.31644297e+00 1.28108695e-01 -3.13692153e-01
-1.35656261e+00 -1.34687936e+00 -4.26021844e-01 7.27557898e-01
6.80656552e-01 -4.02014256e-01 -5.97393811e-01 -1.92129277e-02
-2.01136291e-01 2.10240394e-01 -4.07617539e-01 1.74160287e-01
-4.35632586e-01 -1.18129373e+00 1.82584431e-02 2.38633826e-01
9.79060769e-01 -9.34259474e-01 1.75724283e-01 1.18291788e-01
-7.61154652e-01 -1.27073312e+00 -5.97517312e-01 -3.35202962e-01
-1.00855756e+00 -1.16804564e+00 -8.13442767e-01 -6.13759518e-01
5.16446292e-01 6.59287214e-01 1.44040084e+00 2.17902049e-01
-3.44066203e-01 3.17749560e-01 -6.90249026e-01 1.83118746e-01
-2.76520878e-01 -5.17942131e-01 -4.99342754e-02 -3.06654396e-03
-2.55509704e-01 -8.67783189e-01 -1.18001592e+00 5.21489739e-01
-1.21619010e+00 1.47878513e-01 6.93102300e-01 8.10462594e-01
1.26655114e+00 6.12816751e-01 6.35412812e-01 -5.47191799e-01
5.75182080e-01 -4.22724903e-01 -1.61875308e-01 3.44316483e-01
-8.17148089e-01 -3.66591483e-01 5.96980155e-01 -4.30014580e-01
-1.21337390e+00 -5.05485713e-01 -1.36302069e-01 -3.68566334e-01
-1.49883583e-01 1.65921763e-01 -4.43574756e-01 -1.48738578e-01
5.16213357e-01 6.29999161e-01 -1.87942445e-01 -1.15278554e+00
2.82182217e-01 6.90444708e-01 9.62637365e-01 -5.31128049e-01
7.76101112e-01 6.19675696e-01 -2.17967540e-01 -7.77072132e-01
-8.72653365e-01 -4.63661432e-01 -3.34238023e-01 -3.22391391e-01
5.80398440e-01 -1.56535304e+00 -1.21241979e-01 1.22108674e+00
-7.84639955e-01 -3.55734676e-01 -3.45949113e-01 2.05228537e-01
-5.77443421e-01 6.70229316e-01 -1.04607356e+00 -4.24899876e-01
-5.88367879e-01 -1.34909093e+00 1.18765295e+00 1.84601948e-01
7.03118026e-01 -5.36894441e-01 -2.39746898e-01 8.46562982e-01
6.93712234e-01 1.92270458e-01 7.22606361e-01 2.79047191e-01
-7.67211497e-01 1.65606141e-01 -6.96008325e-01 7.46243715e-01
1.79728985e-01 -4.96821702e-01 -7.58923233e-01 -6.96890295e-01
-2.43741106e-02 -3.46522272e-01 1.12336135e+00 6.52680814e-01
1.41663051e+00 -3.00582677e-01 1.46865815e-01 9.46262002e-01
1.85300672e+00 -1.57237977e-01 1.34660304e+00 6.33574963e-01
6.01525009e-01 1.32320285e-01 6.29055798e-01 4.62956965e-01
5.68940163e-01 7.54810393e-01 5.30343473e-01 -3.83793294e-01
-8.49624157e-01 7.65340179e-02 3.68615806e-01 1.02074420e+00
-3.86826187e-01 -2.90030748e-01 -4.55328912e-01 6.37116015e-01
-1.78981972e+00 -8.02170217e-01 -1.66395500e-01 1.98215473e+00
1.03762412e+00 -2.50191659e-01 -2.58791238e-01 1.30688816e-01
7.92463422e-01 4.21445847e-01 -7.65643477e-01 3.03472817e-01
-7.36322463e-01 1.04224652e-01 5.68738401e-01 5.43603897e-01
-1.11383700e+00 6.72892511e-01 6.38212013e+00 1.45566940e+00
-4.04047757e-01 5.01670718e-01 1.00089145e+00 7.39872735e-03
-1.75784171e-01 -1.67972013e-01 -5.35135984e-01 6.68917954e-01
4.83913660e-01 3.87461483e-02 7.31438696e-01 4.49711621e-01
5.09286284e-01 -3.23577940e-01 -4.44034457e-01 1.19523442e+00
3.14582527e-01 -1.24039030e+00 2.52275944e-01 -7.41708931e-03
1.02925801e+00 6.11309893e-02 9.62378606e-02 -2.39483323e-02
2.71577120e-01 -1.04018700e+00 5.70599854e-01 8.10003400e-01
1.17091990e+00 -4.84260947e-01 7.73413122e-01 -6.84546754e-02
-1.19453442e+00 -7.18885884e-02 -4.95453745e-01 3.70916784e-01
4.62706655e-01 1.02500558e+00 6.19357049e-01 1.12347674e+00
1.29407978e+00 1.00425947e+00 -7.18309343e-01 1.20995736e+00
-2.33803317e-01 4.06020314e-01 1.36562418e-02 1.20533490e+00
-4.28392947e-01 -2.69394904e-01 5.24202168e-01 9.65864897e-01
5.85086882e-01 4.02087420e-01 3.04585397e-01 4.60569948e-01
8.30932632e-02 -6.93701953e-02 -8.55170004e-03 6.06068134e-01
5.40769696e-01 1.08706415e+00 -4.75600541e-01 -3.73909265e-01
-4.70216721e-01 1.15646386e+00 -4.08981554e-02 7.11490452e-01
-6.32608533e-01 -1.97977331e-02 1.01205254e+00 1.61259830e-01
1.90788552e-01 -1.36204779e-01 -3.04309189e-01 -1.48946524e+00
3.89786959e-01 -1.23459792e+00 3.92641276e-01 -1.28999865e+00
-1.48500288e+00 6.92595243e-01 -2.13315681e-01 -1.38234591e+00
4.42261279e-01 -3.25911671e-01 -4.13485989e-02 9.18301642e-01
-1.99096835e+00 -1.30853140e+00 -6.49258852e-01 8.67708147e-01
7.78774440e-01 1.73564907e-02 3.72277260e-01 7.74695873e-01
-7.66439438e-01 2.64995575e-01 5.45300245e-01 -9.21754539e-02
1.00345910e+00 -9.75526869e-01 1.60382092e-01 1.32896769e+00
-3.64658177e-01 1.98476717e-01 7.78106928e-01 -5.96411526e-01
-1.39506364e+00 -1.14118338e+00 4.20228630e-01 -1.14714466e-01
2.82730311e-01 1.30184531e-01 -1.01566613e+00 3.03220272e-01
1.24262668e-01 4.49923784e-01 2.44940668e-01 -2.80436516e-01
-4.54485416e-01 -3.54944706e-01 -1.29635775e+00 3.39291781e-01
1.25907445e+00 -4.43039417e-01 -2.07745321e-02 3.80067170e-01
8.13986242e-01 -6.36113405e-01 -1.36605442e+00 6.16038144e-01
2.24413186e-01 -1.40792656e+00 1.59564257e+00 1.47746399e-03
8.71027827e-01 -6.34720802e-01 -6.98591352e-01 -1.29230332e+00
-6.59584463e-01 -1.57482907e-01 -3.06487173e-01 1.38624835e+00
-1.63254768e-01 -5.16799450e-01 6.58861995e-02 1.10653564e-01
-3.17605287e-01 -6.30606592e-01 -1.01248908e+00 -8.64624560e-01
-1.74203306e-01 -1.64549083e-01 6.55390382e-01 9.55395997e-01
-7.43232369e-01 -1.00669250e-01 -1.08005977e+00 5.94831467e-01
1.35097289e+00 2.07245767e-01 5.09061635e-01 -8.34403634e-01
-1.02078460e-01 -8.23713839e-02 -2.29219511e-01 -1.06614387e+00
-4.79038268e-01 -5.11867881e-01 -1.46102056e-01 -2.09724641e+00
6.68534338e-01 -2.80833066e-01 -4.32423979e-01 3.66287947e-01
-4.21295911e-01 6.99054122e-01 3.44524011e-02 6.80751979e-01
-7.79930770e-01 7.04171240e-01 1.65350258e+00 -2.06188746e-02
9.27230045e-02 -5.51057041e-01 -1.00460279e+00 4.74685401e-01
7.14592338e-01 -1.05222657e-01 -1.95920706e-01 -9.52860951e-01
-6.93223104e-02 2.07872704e-01 6.04348183e-01 -1.08266747e+00
1.64569899e-01 -3.63256425e-01 5.76891720e-01 -4.94188845e-01
5.10621190e-01 -5.56650877e-01 4.76831019e-01 6.37745783e-02
1.06381781e-01 -2.24101305e-01 -4.49506473e-03 6.95932627e-01
-5.43437243e-01 3.46364141e-01 1.21616936e+00 -2.64009833e-01
-1.16442347e+00 6.01686954e-01 9.40516889e-02 -1.42299440e-02
5.02176285e-01 -2.34210715e-01 -9.62058604e-01 -2.75538474e-01
-6.60493910e-01 2.88371176e-01 8.00561011e-01 3.72066438e-01
8.62695217e-01 -1.37896395e+00 -1.26738620e+00 -2.71930695e-02
2.63116956e-01 3.11850049e-02 1.20524538e+00 7.18932092e-01
-4.06297803e-01 -2.51500368e-01 -2.71328777e-01 -4.44699109e-01
-1.15708172e+00 5.51283777e-01 2.86114246e-01 -4.26106900e-01
-1.15486765e+00 5.42654037e-01 1.59897178e-01 -2.77141958e-01
2.18453221e-02 6.80019259e-02 -2.39113361e-01 -1.37342691e-01
9.83996511e-01 5.79909563e-01 6.84775040e-02 -7.90373981e-01
-9.97531712e-02 8.38284135e-01 8.00656229e-02 2.80536324e-01
1.67299879e+00 -8.86042118e-01 -3.07030410e-01 -4.50571142e-02
7.08480120e-01 -2.31269866e-01 -1.39265084e+00 -6.62094712e-01
-4.85175997e-01 -1.10728633e+00 4.26907629e-01 -1.05222857e+00
-1.53131473e+00 3.26909453e-01 1.02160382e+00 -1.89287528e-01
1.91740620e+00 -1.46987543e-01 9.21805084e-01 -2.43558750e-01
5.87991476e-01 -1.10247016e+00 1.77999854e-01 1.67203560e-01
1.22105432e+00 -1.53360248e+00 3.71900827e-01 -6.30032003e-01
-5.44216752e-01 7.22476661e-01 6.40687048e-01 4.48011830e-02
5.19368827e-01 2.01456159e-01 1.48841828e-01 -9.74501148e-02
-6.17262542e-01 -1.79473311e-01 1.25690103e-02 7.98830926e-01
1.28168195e-01 4.50455323e-02 -3.76740664e-01 4.91941959e-01
1.01202227e-01 1.92009076e-01 7.20436454e-01 7.10096836e-01
-5.29335797e-01 -9.23774004e-01 -6.48847222e-01 3.82889599e-01
-5.19348919e-01 -4.63143915e-01 4.22515482e-01 3.85461450e-01
2.68617064e-01 1.33441889e+00 -4.36513126e-01 -3.00005317e-01
4.11482304e-01 -6.16064429e-01 3.13341141e-01 -1.23079389e-01
1.05999680e-02 1.84708759e-01 -1.46083347e-02 -9.51758087e-01
-6.58032715e-01 -3.60541999e-01 -4.50279474e-01 -6.61046684e-01
-1.31219417e-01 -1.09781004e-01 3.58568132e-01 4.39584255e-01
5.13564944e-01 7.48783648e-01 6.56706989e-01 -9.86895084e-01
-3.15232813e-01 -1.16131461e+00 -1.13955569e+00 5.38196325e-01
5.55530369e-01 -5.33224165e-01 -4.77994174e-01 3.85210305e-01]
|
[11.208003997802734, -2.2024459838867188]
|
dc578e38-1926-416b-b3f0-37f37d099dfc
|
ps-arm-an-end-to-end-attention-aware-relation
|
2210.03433
| null |
https://arxiv.org/abs/2210.03433v1
|
https://arxiv.org/pdf/2210.03433v1.pdf
|
PS-ARM: An End-to-End Attention-aware Relation Mixer Network for Person Search
|
Person search is a challenging problem with various real-world applications, that aims at joint person detection and re-identification of a query person from uncropped gallery images. Although, the previous study focuses on rich feature information learning, it is still hard to retrieve the query person due to the occurrence of appearance deformations and background distractors. In this paper, we propose a novel attention-aware relation mixer (ARM) module for person search, which exploits the global relation between different local regions within RoI of a person and make it robust against various appearance deformations and occlusion. The proposed ARM is composed of a relation mixer block and a spatio-channel attention layer. The relation mixer block introduces a spatially attended spatial mixing and a channel-wise attended channel mixing for effectively capturing discriminative relation features within an RoI. These discriminative relation features are further enriched by introducing a spatio-channel attention where the foreground and background discriminability is empowered in a joint spatio-channel space. Our ARM module is generic and it does not rely on fine-grained supervision or topological assumptions, hence being easily integrated into any Faster R-CNN based person search methods. Comprehensive experiments are performed on two challenging benchmark datasets: CUHKSYSU and PRW. Our PS-ARM achieves state-of-the-art performance on both datasets. On the challenging PRW dataset, our PS-ARM achieves an absolute gain of 5 in the mAP score over SeqNet, while operating at a comparable speed.
|
['Fahad Shahbaz Khan', 'Rao Muhammad Anwer', 'Sanath Narayan', 'Hisham Cholakkal', 'Mustansar Fiaz']
|
2022-10-07
| null | null | null | null |
['person-search']
|
['computer-vision']
|
[ 9.62508842e-02 -3.71268958e-01 6.56542554e-02 -2.30439052e-01
-7.54006386e-01 -3.52306187e-01 8.50752532e-01 -6.76385537e-02
-7.68854320e-01 4.37748522e-01 2.77320653e-01 1.65720895e-01
-1.33762673e-01 -5.80553949e-01 -4.88901138e-01 -7.69872725e-01
1.07617453e-01 5.13544142e-01 3.48784149e-01 -1.20028391e-01
-2.38694936e-01 5.48878133e-01 -1.53062236e+00 1.04399174e-01
7.08480537e-01 1.14221752e+00 5.00134192e-02 5.61664045e-01
4.64885294e-01 3.59441251e-01 -5.79985261e-01 -6.25452518e-01
4.33442146e-01 -2.62867838e-01 -6.84653223e-01 1.92919254e-01
7.08177626e-01 -2.54312813e-01 -7.95565367e-01 8.57000172e-01
1.01254213e+00 3.73950452e-01 4.93194520e-01 -9.80363071e-01
-6.69406474e-01 5.31081902e-03 -9.04389560e-01 6.29684865e-01
6.41978025e-01 6.04386628e-01 9.59484816e-01 -7.84344435e-01
2.99675405e-01 1.46126747e+00 5.64482033e-01 4.64281440e-01
-1.20631921e+00 -7.20429242e-01 4.28546429e-01 4.79825497e-01
-1.67131996e+00 -2.35985711e-01 7.45275438e-01 -2.49014571e-01
8.07978272e-01 4.55979049e-01 7.62000024e-01 1.27618897e+00
-1.62968501e-01 9.75447178e-01 9.73300338e-01 -1.47935063e-01
-1.85502931e-01 6.46163002e-02 1.53329819e-01 7.46176541e-01
2.17017040e-01 2.02335045e-01 -5.68554342e-01 -5.46342097e-02
9.21380758e-01 2.80622005e-01 -3.61157537e-01 -3.12964976e-01
-1.15744531e+00 5.49148738e-01 1.00686204e+00 2.91340441e-01
-3.11430454e-01 5.83077110e-02 1.92060500e-01 -9.36188623e-02
2.42591038e-01 6.86963424e-02 -1.59790456e-01 2.75217801e-01
-9.27259386e-01 4.80339885e-01 4.12244558e-01 7.86225975e-01
3.88226479e-01 -3.74519199e-01 -9.12489831e-01 8.92183483e-01
2.78677166e-01 6.77742302e-01 4.61262882e-01 -2.94060588e-01
5.49652636e-01 9.12332773e-01 2.43518338e-01 -1.08396626e+00
-5.48403561e-01 -9.61360812e-01 -1.01119113e+00 -1.64184809e-01
4.71888572e-01 1.86142191e-01 -1.07737160e+00 1.55250716e+00
5.35897732e-01 2.01592848e-01 -4.13737744e-01 1.42780018e+00
1.08274400e+00 2.45443523e-01 2.04254016e-01 3.16507548e-01
1.98338568e+00 -1.37055862e+00 -2.81928390e-01 -4.36983585e-01
3.16154328e-05 -4.70734507e-01 9.68477845e-01 -9.63485893e-03
-1.04024732e+00 -9.09069359e-01 -8.26753497e-01 -3.05792272e-01
-4.68507409e-01 4.23561573e-01 4.27295625e-01 6.66343808e-01
-8.75967503e-01 5.42598031e-02 -4.98409569e-01 -5.38114250e-01
5.76923132e-01 5.77819407e-01 -6.21666431e-01 -1.02010064e-01
-1.13831615e+00 6.77272975e-01 8.00188258e-02 5.66802740e-01
-5.95734239e-01 -3.80197257e-01 -8.57614279e-01 1.03868753e-01
4.11800951e-01 -1.01302910e+00 6.97065294e-01 -6.92974508e-01
-1.24299359e+00 1.04392385e+00 -3.16015452e-01 -3.64136457e-01
8.42137337e-01 -2.72746861e-01 -4.73642617e-01 2.43207961e-01
2.03900382e-01 4.83290136e-01 9.08074856e-01 -1.07360363e+00
-6.04630768e-01 -8.69723856e-01 -3.20331343e-02 3.75719786e-01
-3.84378076e-01 3.50141436e-01 -1.30024326e+00 -9.20364499e-01
-1.36208579e-01 -9.01548028e-01 -2.49621928e-01 4.51129191e-02
-4.72949415e-01 -3.87982607e-01 5.95033407e-01 -9.17291999e-01
1.05130208e+00 -1.85905993e+00 3.10711086e-01 3.44244272e-01
2.12893531e-01 4.87547398e-01 -2.25005507e-01 7.91829079e-02
1.46386147e-01 -2.62877941e-01 -9.46725905e-03 -6.23470902e-01
1.33691922e-01 -2.71402478e-01 -3.53650264e-02 8.13692451e-01
2.54447013e-01 1.35853994e+00 -6.87939763e-01 -5.34589231e-01
3.44162405e-01 8.44314396e-01 -2.88794130e-01 1.69356421e-01
2.41021544e-01 6.90502405e-01 -5.16811669e-01 9.63892579e-01
6.47210181e-01 -3.88960183e-01 -1.85747698e-01 -9.68928561e-02
1.66831791e-01 -2.68091321e-01 -1.16576552e+00 1.68024814e+00
-2.24084079e-01 3.53662789e-01 4.80155312e-02 -8.94296885e-01
9.70047116e-01 2.14068182e-02 1.70057043e-01 -1.02430630e+00
1.12778202e-01 3.18472199e-02 -1.14105560e-01 -4.24622416e-01
3.94881785e-01 4.80175644e-01 -1.49847001e-01 1.96485519e-01
-5.99166639e-02 8.78624141e-01 4.95794639e-02 4.79249321e-02
1.01019704e+00 1.63319662e-01 7.14620128e-02 -3.86534214e-01
1.12368667e+00 -4.89259511e-01 4.93156135e-01 8.94798100e-01
-4.43557113e-01 6.94514275e-01 -1.30253388e-02 -8.47653747e-01
-7.50688434e-01 -1.09643567e+00 -6.45093694e-02 1.18510020e+00
7.15897918e-01 -2.98535705e-01 -5.45985878e-01 -7.57770658e-01
1.70907736e-01 -1.01208419e-01 -8.02488506e-01 -5.03627732e-02
-7.25008845e-01 -8.70901585e-01 6.54937744e-01 6.36243045e-01
1.08633935e+00 -1.13728440e+00 -6.70431137e-01 -8.39973241e-02
-3.66066396e-01 -1.34640479e+00 -9.57715392e-01 -3.44734907e-01
-2.40980819e-01 -1.14281976e+00 -1.24631870e+00 -8.60311687e-01
7.19515622e-01 2.78176546e-01 9.00779903e-01 1.55083403e-01
-8.42093587e-01 4.09778327e-01 -1.30380601e-01 1.93313062e-02
6.25227928e-01 2.14473560e-01 -3.85095216e-02 5.68636715e-01
6.24232411e-01 -1.74096063e-01 -1.28602159e+00 7.00310588e-01
-5.41906178e-01 -2.24177852e-01 6.49886966e-01 9.81304705e-01
5.84074676e-01 6.84017763e-02 2.03943849e-01 -2.41997153e-01
3.05442393e-01 -1.33767694e-01 -4.15918678e-01 3.61173451e-01
-1.38983265e-01 -1.78770199e-01 3.15524608e-01 -5.35420418e-01
-1.00951731e+00 1.61940038e-01 6.23264350e-03 -2.11524725e-01
-2.71885127e-01 -2.56824762e-01 -5.16533434e-01 -1.28193945e-01
4.14326489e-01 4.33139384e-01 -3.27089369e-01 -5.60814083e-01
2.56532013e-01 5.55493653e-01 7.86607325e-01 -5.50997615e-01
9.74144280e-01 7.02795088e-01 -1.34584069e-01 -7.96053290e-01
-5.75079322e-01 -8.75488579e-01 -6.86832309e-01 -2.13511184e-01
9.82300818e-01 -1.16417837e+00 -9.83889222e-01 5.97066462e-01
-9.02253628e-01 -1.71688348e-01 -4.14506495e-02 1.64109662e-01
-9.29940641e-02 3.64266634e-01 -4.12009448e-01 -8.26650262e-01
-6.31431043e-01 -1.21683288e+00 1.53297007e+00 5.71523488e-01
-8.01623911e-02 -7.14144945e-01 -1.80333242e-01 7.80341268e-01
3.15737844e-01 3.22469443e-01 3.14140230e-01 -6.52749419e-01
-7.48380065e-01 -5.91736436e-01 -6.90407395e-01 -1.40392199e-01
-5.11551350e-02 -7.98320651e-01 -1.11466551e+00 -6.16199195e-01
-5.31882644e-01 -1.01782389e-01 1.37907434e+00 1.91667303e-01
1.19448245e+00 -2.23769337e-01 -7.21427321e-01 6.82284832e-01
1.13346207e+00 -2.35531062e-01 5.79309344e-01 3.90306562e-01
8.56553972e-01 6.63834572e-01 2.87961721e-01 2.18505055e-01
5.08341908e-01 1.34218562e+00 1.22177109e-01 -4.08566296e-01
-4.32558060e-01 -2.14496434e-01 1.71536162e-01 -2.94211030e-01
-4.97278988e-01 -2.51985162e-01 -7.39402175e-01 6.06675506e-01
-2.08739614e+00 -1.06633079e+00 1.13471031e-01 2.28787017e+00
5.41841984e-01 -1.16611250e-01 5.52839756e-01 2.01224927e-02
7.88273871e-01 2.38237336e-01 -5.54455757e-01 3.65578026e-01
-3.35096538e-01 2.23012775e-01 3.93964916e-01 3.71540576e-01
-1.50117695e+00 7.94109523e-01 4.57905197e+00 9.02498066e-01
-6.77674353e-01 1.66129038e-01 7.18491852e-01 -3.12410653e-01
2.43110448e-01 -4.76480007e-01 -1.15070450e+00 5.38843930e-01
1.80444852e-01 3.50557119e-01 2.22800106e-01 4.84466344e-01
2.42883172e-02 -6.78144693e-02 -1.01617825e+00 1.35858810e+00
3.62362534e-01 -8.09894979e-01 -6.79519922e-02 2.43471503e-01
4.48511750e-01 -3.53378147e-01 4.01337534e-01 2.10849211e-01
-1.10471286e-01 -1.34213293e+00 5.78799367e-01 6.37664974e-01
6.57671452e-01 -8.85203362e-01 9.24530149e-01 1.18144833e-01
-1.69586372e+00 -3.09209347e-01 -1.64669573e-01 2.31280565e-01
8.05648640e-02 2.84697384e-01 -3.86868089e-01 7.18586087e-01
1.12794948e+00 5.07958889e-01 -9.94280457e-01 1.25305438e+00
-1.54538676e-01 -2.24211477e-02 -3.62006426e-01 1.17981449e-01
3.10959239e-02 1.07389852e-01 6.13657057e-01 1.56226432e+00
-1.18695699e-01 1.07200770e-02 3.58363122e-01 8.62967432e-01
1.40447775e-02 3.99626531e-02 -1.55320272e-01 6.10939622e-01
5.67532480e-02 1.36510611e+00 -7.47451663e-01 -2.31100544e-01
-3.32982421e-01 1.51730037e+00 2.40032405e-01 6.22393906e-01
-9.15659547e-01 -2.19055772e-01 7.09420681e-01 2.63056308e-01
6.13821566e-01 4.91218679e-02 1.33256940e-02 -1.30650783e+00
4.45302606e-01 -8.49165142e-01 6.94524169e-01 -2.92953998e-01
-1.43985415e+00 7.93507040e-01 -1.71700746e-01 -8.85909915e-01
9.98109207e-02 -5.06008863e-01 -5.03532767e-01 1.26665747e+00
-1.65766263e+00 -1.82116842e+00 -7.54618883e-01 1.03653312e+00
5.33722103e-01 -3.98269832e-01 6.91680074e-01 5.33828735e-01
-1.00291169e+00 1.15800226e+00 -4.45518792e-01 5.28464735e-01
7.13862360e-01 -1.09666109e+00 4.18946564e-01 1.03374481e+00
1.03533097e-01 7.88679719e-01 3.88532996e-01 -6.88334703e-01
-1.27862275e+00 -1.15374053e+00 8.95014584e-01 -6.20510757e-01
2.34162182e-01 -5.98041356e-01 -6.26313329e-01 4.01013553e-01
-8.63755774e-03 4.66708004e-01 4.18440908e-01 1.36398658e-01
-6.18111253e-01 -2.24968955e-01 -1.06733060e+00 5.19407749e-01
1.41825306e+00 -5.42112827e-01 -3.59979570e-01 3.48947346e-01
3.00277442e-01 -3.71216744e-01 -5.89987040e-01 3.69354159e-01
7.04277813e-01 -9.51374352e-01 1.66602552e+00 -4.02606666e-01
-6.66570812e-02 -5.47725499e-01 1.89305931e-01 -7.02909827e-01
-5.91419876e-01 -6.27104163e-01 -2.78548896e-01 1.18806720e+00
-1.62760410e-02 -6.39105439e-01 7.47583151e-01 6.02824569e-01
4.44340050e-01 -7.95200884e-01 -1.05494523e+00 -8.81100118e-01
-3.68592292e-01 -3.35726999e-02 5.78333080e-01 3.74254793e-01
-3.42084318e-01 3.41446131e-01 -5.45449615e-01 4.75937724e-01
9.46274400e-01 1.60372272e-01 9.10249650e-01 -1.13043189e+00
-5.39299607e-01 -5.81207275e-01 -6.25437737e-01 -1.31336832e+00
-4.44280282e-02 -7.52242088e-01 -1.90917373e-01 -1.47188330e+00
5.55665255e-01 -3.37767094e-01 -3.92385721e-01 3.89839232e-01
-5.26231945e-01 7.16257870e-01 2.35182628e-01 3.22947025e-01
-8.27663004e-01 7.16393530e-01 9.90925968e-01 -4.37966436e-01
-3.15182805e-01 1.89450860e-01 -5.66917777e-01 5.41519344e-01
4.55954164e-01 1.09862782e-01 -6.67184964e-02 -3.81270379e-01
-2.14018509e-01 -4.91791695e-01 1.23896646e+00 -1.06704974e+00
5.15786171e-01 4.02332723e-01 1.06181705e+00 -6.53513968e-01
6.98980153e-01 -6.77771807e-01 5.44990599e-02 4.83264297e-01
-1.29596785e-01 8.21471773e-03 1.17768645e-01 7.85754025e-01
-8.69447887e-02 3.88874471e-01 8.02161217e-01 -1.35348275e-01
-7.66421318e-01 6.36024594e-01 3.51457506e-01 -1.17727973e-01
1.03776777e+00 -2.69745737e-01 -2.12178364e-01 -1.56723201e-01
-6.63005292e-01 4.95764852e-01 2.43345544e-01 6.22710764e-01
6.24055028e-01 -1.32655263e+00 -8.34056735e-01 4.27945137e-01
2.46702090e-01 -1.18560329e-01 5.28653920e-01 8.47296357e-01
-1.25034019e-01 6.68760538e-01 -5.08560166e-02 -6.15201414e-01
-1.43429732e+00 5.47553599e-01 5.99205315e-01 -4.77653593e-01
-8.79751146e-01 1.12705493e+00 5.58007717e-01 -1.59528583e-01
5.13639629e-01 1.56938404e-01 -4.00638014e-01 -7.29383379e-02
9.03552413e-01 3.63881648e-01 -1.92343354e-01 -1.10025108e+00
-6.71613455e-01 9.35404181e-01 -2.38079322e-03 3.91203351e-02
9.66086745e-01 -1.85839221e-01 8.45511332e-02 -2.07641184e-01
9.05351162e-01 -1.11236393e-01 -1.32010162e+00 -5.62641144e-01
-1.95773870e-01 -6.64484978e-01 -2.66234875e-01 -8.56701553e-01
-1.10766912e+00 7.33003020e-01 8.65744412e-01 -8.27102885e-02
1.15149915e+00 3.51714104e-01 8.78905952e-01 5.89922704e-02
2.36909345e-01 -9.75244284e-01 2.35003233e-01 1.18500285e-01
1.03002203e+00 -1.36985326e+00 1.07233979e-01 -3.52561504e-01
-4.27177340e-01 6.92838192e-01 7.54082501e-01 -4.35624942e-02
3.47646713e-01 -3.41018677e-01 -2.79950827e-01 -2.13885278e-01
-1.20094739e-01 -6.68864071e-01 9.29476023e-01 6.21838927e-01
-3.66202705e-02 -4.81352769e-02 -4.69742110e-03 8.65148365e-01
-7.08364397e-02 -2.63692826e-01 -5.80744386e-01 4.65770155e-01
-1.04178928e-01 -9.37913716e-01 -5.79040945e-01 1.73179358e-01
-4.12950605e-01 5.64984195e-02 -4.50196177e-01 7.80879617e-01
5.23638248e-01 9.83305871e-01 6.64141849e-02 -2.45523423e-01
5.22708714e-01 -1.71110108e-01 4.44923133e-01 -2.15270430e-01
-8.18242073e-01 1.12543762e-01 -1.54325694e-01 -7.56782830e-01
-4.96365011e-01 -7.21399665e-01 -7.47480571e-01 -1.54562488e-01
-2.02137023e-01 -1.45608664e-01 1.07360579e-01 9.27893341e-01
4.02322382e-01 4.41392481e-01 2.87764877e-01 -8.70106459e-01
-1.84356943e-01 -9.42799747e-01 -3.03027600e-01 5.76402068e-01
4.59792793e-01 -7.77420998e-01 9.00225118e-02 -7.26154745e-02]
|
[14.793800354003906, 0.824307918548584]
|
100a0773-441e-427f-9b5e-fc3b5408ceee
|
a-perceptual-measure-for-evaluating-the
|
2202.12257
| null |
https://arxiv.org/abs/2202.12257v2
|
https://arxiv.org/pdf/2202.12257v2.pdf
|
A Perceptual Measure for Evaluating the Resynthesis of Automatic Music Transcriptions
|
This study focuses on the perception of music performances when contextual factors, such as room acoustics and instrument, change. We propose to distinguish the concept of "performance" from the one of "interpretation", which expresses the "artistic intention". Towards assessing this distinction, we carried out an experimental evaluation where 91 subjects were invited to listen to various audio recordings created by resynthesizing MIDI data obtained through Automatic Music Transcription (AMT) systems and a sensorized acoustic piano. During the resynthesis, we simulated different contexts and asked listeners to evaluate how much the interpretation changes when the context changes. Results show that: (1) MIDI format alone is not able to completely grasp the artistic intention of a music performance; (2) usual objective evaluation measures based on MIDI data present low correlations with the average subjective evaluation. To bridge this gap, we propose a novel measure which is meaningfully correlated with the outcome of the tests. In addition, we investigate multimodal machine learning by providing a new score-informed AMT method and propose an approximation algorithm for the $p$-dispersion problem.
|
['Stavros Ntalampiras', 'Federico Avanzini', 'Federico Simonetta']
|
2022-02-24
| null | null | null | null |
['music-transcription']
|
['music']
|
[ 3.63342762e-01 -7.21862763e-02 2.71664113e-01 -3.42894763e-01
-1.01344395e+00 -6.27978563e-01 4.19307888e-01 1.63040698e-01
-3.43414515e-01 5.35977781e-01 4.15779918e-01 3.10095400e-01
-3.85834754e-01 -3.99001211e-01 -5.30084789e-01 -6.89655542e-01
1.20406479e-01 2.72695124e-01 -1.99471503e-01 -1.59090593e-01
4.84811574e-01 1.62815884e-01 -2.03932524e+00 4.56819981e-01
5.89409173e-01 1.14918375e+00 1.51681617e-01 8.06910813e-01
-4.80690673e-02 4.35556501e-01 -1.17402077e+00 -1.70978785e-01
1.93774104e-01 -7.07062721e-01 -5.43396592e-01 -1.84306502e-02
3.33772629e-01 1.93882197e-01 3.92345428e-01 1.21797359e+00
7.22218335e-01 1.88182697e-01 7.08187819e-01 -9.43799257e-01
-2.78664917e-01 1.02210212e+00 -8.09493586e-02 -1.62686452e-01
8.98409724e-01 -3.20763178e-02 1.11043632e+00 -6.75919235e-01
2.80308664e-01 9.21951294e-01 7.12928951e-01 2.65302956e-01
-1.40861928e+00 -4.37270790e-01 -1.32180586e-01 3.83124232e-01
-1.53488946e+00 -4.81256425e-01 1.27957594e+00 -5.95280766e-01
3.59765917e-01 8.30504298e-01 6.67134821e-01 1.02263606e+00
-1.06842913e-01 4.36611921e-01 1.34187329e+00 -7.94865966e-01
4.55261022e-01 3.77372235e-01 -1.58762112e-01 1.72952227e-02
-3.95435154e-01 6.92272484e-02 -7.14111030e-01 -1.54931664e-01
9.65761095e-02 -6.48802757e-01 -4.99143541e-01 1.42500382e-02
-1.38865077e+00 3.83593321e-01 -1.25236630e-01 8.47554922e-01
-4.32147056e-01 1.35988474e-01 4.34665322e-01 5.19895375e-01
1.18985727e-01 8.33660603e-01 -1.01106957e-01 -6.08330667e-01
-9.80577052e-01 2.94832826e-01 8.69920850e-01 2.90650785e-01
2.98206776e-01 1.13735922e-01 -2.34762728e-01 1.03332055e+00
1.50663599e-01 6.32572651e-01 8.28760326e-01 -1.10446846e+00
1.18144557e-01 2.66849220e-01 4.75848317e-01 -1.18279791e+00
-3.77360582e-01 -5.93387544e-01 -4.03519750e-01 3.53837401e-01
4.87420678e-01 3.20791863e-02 -2.06645861e-01 1.92067862e+00
1.11937812e-02 1.02908246e-01 -8.47457126e-02 1.09816086e+00
6.08915865e-01 4.86455560e-01 -1.93496943e-01 -6.90211654e-01
1.24429619e+00 -4.00793046e-01 -1.12066185e+00 2.23564446e-01
1.66640654e-01 -1.13412154e+00 1.60688233e+00 1.00787234e+00
-1.29523194e+00 -1.16740394e+00 -1.26057065e+00 4.76042092e-01
-1.63350523e-01 2.13239044e-01 1.09238714e-01 9.88565266e-01
-8.53620410e-01 7.19779909e-01 -3.67117316e-01 2.86844615e-02
-4.75170583e-01 2.20810309e-01 -2.28740480e-02 6.53526843e-01
-9.98475432e-01 5.79174101e-01 2.15285972e-01 2.86011219e-01
-6.41390800e-01 -3.34609121e-01 -3.59776646e-01 1.79698065e-01
1.82991400e-01 -3.59778494e-01 1.34872699e+00 -1.41005898e+00
-1.86525595e+00 8.13878000e-01 -4.30540778e-02 -1.07000887e-01
4.71029907e-01 -1.61967725e-01 -7.61387825e-01 1.08510889e-01
-2.05877021e-01 1.73088312e-01 1.00395143e+00 -1.44982219e+00
-3.46409947e-01 -3.22891444e-01 -2.91739643e-01 3.24217469e-01
-2.74420023e-01 3.41046453e-02 -4.90817474e-03 -7.58478284e-01
2.14449108e-01 -9.29395676e-01 2.66793996e-01 -6.07515991e-01
-5.29456258e-01 -2.69538519e-04 2.45445684e-01 -5.86239278e-01
1.52026665e+00 -2.53216696e+00 3.03677529e-01 5.89010596e-01
-6.81670681e-02 2.12055236e-01 -3.84129100e-02 3.83573890e-01
-1.67269081e-01 -2.34485880e-01 -2.06158668e-01 -3.15506905e-01
4.04754281e-01 -3.92616652e-02 -3.57497931e-01 1.00660518e-01
-5.22466719e-01 3.23923498e-01 -5.21233439e-01 -3.81034821e-01
6.78732321e-02 3.20516258e-01 -5.09941220e-01 1.96381599e-01
-1.79397181e-01 5.48961043e-01 -4.56043221e-02 4.18649584e-01
3.95276159e-01 3.73243719e-01 1.67765707e-01 -5.66233039e-01
-2.11604655e-01 1.51149631e-01 -1.53780043e+00 1.60380781e+00
-2.92767733e-01 6.01705551e-01 1.43202931e-01 -9.63225365e-01
1.29823768e+00 5.35548866e-01 3.54579955e-01 -5.70043504e-01
2.64622569e-01 4.15428698e-01 1.08877018e-01 -8.28277171e-01
5.84069848e-01 -3.40587527e-01 -1.97914898e-01 3.75617206e-01
-2.79985487e-01 -4.28258508e-01 3.75934877e-02 -4.80826467e-01
7.87182510e-01 8.44411179e-02 2.90190548e-01 -8.69061947e-02
6.68539286e-01 -4.22286123e-01 1.94320008e-01 6.37963891e-01
-2.10515961e-01 6.14401102e-01 3.92769933e-01 -7.46454373e-02
-5.31634629e-01 -1.33364058e+00 1.09301105e-01 1.23384225e+00
-8.99946466e-02 -3.53030384e-01 -9.81525779e-01 -4.21742760e-02
-2.66739666e-01 9.82837796e-01 -4.99220788e-01 -1.47929981e-01
-4.01947588e-01 -6.07474148e-01 5.36988020e-01 2.53192484e-01
1.82664946e-01 -1.14527261e+00 -4.50670779e-01 2.33972430e-01
-7.31248319e-01 -7.62382627e-01 -5.90943098e-01 6.40888214e-02
-6.07128978e-01 -5.94128489e-01 -4.78087693e-01 -5.27745545e-01
7.79841766e-02 -2.89478332e-01 9.52874959e-01 -3.93856019e-01
-9.59336665e-03 6.94013357e-01 -5.74079692e-01 -5.76609910e-01
-7.03520715e-01 -1.22303531e-01 2.19662711e-01 5.00121534e-01
3.10027808e-01 -1.06399190e+00 -5.54309368e-01 4.09984529e-01
-8.45352888e-01 -3.08688462e-01 5.81159890e-01 3.33870143e-01
6.17063284e-01 6.94493800e-02 7.68911541e-01 -3.87394160e-01
1.18511128e+00 9.35924724e-02 -1.87283635e-01 1.84270348e-02
-7.03621805e-01 -7.15282932e-02 4.39549953e-01 -6.67591691e-01
-9.58834231e-01 -2.17845380e-01 -2.66975820e-01 -1.78271636e-01
-4.13803399e-01 4.63974744e-01 -3.92973423e-01 1.78788617e-01
8.26695621e-01 2.23333940e-01 1.12262871e-02 -6.00232780e-01
9.13003534e-02 9.45040464e-01 8.62180114e-01 -7.83167124e-01
6.22677147e-01 1.86127558e-01 -1.64447188e-01 -9.64048862e-01
-6.32549047e-01 -3.10528129e-01 -4.01175797e-01 -6.76507354e-01
8.38898361e-01 -2.84040481e-01 -1.03109407e+00 3.59467030e-01
-9.89196360e-01 -1.06059209e-01 -5.87421834e-01 1.03660047e+00
-8.90954375e-01 4.84924376e-01 -4.37414616e-01 -9.73163724e-01
-1.53350458e-01 -1.11527395e+00 9.13493037e-01 -2.14834996e-02
-6.86432302e-01 -6.40007377e-01 5.30348003e-01 5.75904787e-01
3.14426184e-01 2.04609051e-01 8.39949250e-01 -5.79005241e-01
-6.89553320e-02 -2.87196785e-01 2.88782716e-01 6.82601213e-01
1.36645019e-01 -1.76592469e-01 -1.35186040e+00 5.81065863e-02
3.81103784e-01 -2.06595436e-01 3.31677765e-01 4.11760658e-01
1.10234571e+00 -2.02255011e-01 3.71509105e-01 3.22493911e-01
1.14758849e+00 4.98791516e-01 7.43763328e-01 9.65905041e-02
1.48135871e-01 8.02991152e-01 6.78971052e-01 5.23135841e-01
-1.16779260e-01 1.07302547e+00 3.44664723e-01 2.92359740e-01
-1.43858120e-01 -2.71980524e-01 6.16804123e-01 1.31282759e+00
-4.85519171e-01 -1.65861815e-01 -5.63738585e-01 1.66014329e-01
-1.50131226e+00 -1.05241418e+00 1.58708412e-02 2.32351065e+00
7.66345918e-01 2.00064763e-01 1.32825494e-01 7.57321894e-01
6.49722636e-01 1.03691943e-01 -1.24820389e-01 -7.02808142e-01
-1.20057747e-01 2.58901626e-01 -1.94169626e-01 6.61238790e-01
-7.21069813e-01 2.82019913e-01 6.24895382e+00 1.00469232e+00
-1.19517827e+00 -7.77262496e-03 1.96535140e-01 -2.19314680e-01
-3.78306419e-01 -3.97087663e-01 -1.71684638e-01 4.41580564e-01
9.11918700e-01 6.35316744e-02 5.97138762e-01 5.79331100e-01
5.41276157e-01 -3.48529369e-02 -1.25667298e+00 1.13576949e+00
4.16582465e-01 -5.86764038e-01 -3.58446091e-02 2.48522242e-03
3.41990083e-01 -5.90409279e-01 3.74600112e-01 2.35537842e-01
-6.70552492e-01 -9.32980180e-01 1.03859890e+00 1.00204778e+00
6.33752882e-01 -4.62869048e-01 6.64811790e-01 1.99982330e-01
-1.06753361e+00 -5.14939278e-02 1.80881217e-01 -3.03376198e-01
4.45063971e-02 2.54122734e-01 -8.64545584e-01 3.52138817e-01
4.03823525e-01 -9.55600590e-02 -4.79807287e-01 9.28444803e-01
-8.91392399e-03 7.68628418e-01 -1.35463580e-01 -2.12699518e-01
-2.00727850e-01 -3.12002987e-01 9.86927867e-01 1.18194366e+00
7.21656978e-01 -1.36963472e-01 -9.73964259e-02 9.32128549e-01
2.56243169e-01 5.53284883e-01 -3.96755457e-01 -7.39412233e-02
3.34608108e-01 8.52714956e-01 -5.74463010e-01 -4.31616530e-02
-5.77343181e-02 9.24941838e-01 -4.84407455e-01 4.02299702e-01
-8.75575840e-01 -3.63061577e-01 3.02172124e-01 6.25438243e-02
3.78007889e-02 -7.68840089e-02 -3.62071007e-01 -7.87000477e-01
1.45202637e-01 -1.01341212e+00 -1.03279166e-02 -1.21628702e+00
-1.02145922e+00 7.22138107e-01 4.19194438e-02 -1.49873340e+00
-4.91447538e-01 -5.18396378e-01 -4.76602972e-01 7.07827270e-01
-7.50228226e-01 -6.28091753e-01 -2.43258670e-01 4.45726156e-01
3.91479820e-01 -1.02743737e-01 1.10787737e+00 5.20532131e-01
-8.90419036e-02 6.34770870e-01 -1.12270407e-01 -1.75530717e-01
8.96384060e-01 -1.25888920e+00 -5.50456941e-01 3.24583679e-01
4.85220581e-01 3.35157365e-01 1.34293246e+00 -1.71858996e-01
-8.90865147e-01 -4.26050603e-01 9.29042041e-01 -3.25202644e-01
4.09741879e-01 -5.25992587e-02 -7.34607816e-01 2.26246223e-01
5.92257828e-02 -6.52494669e-01 1.10881114e+00 3.15908074e-01
-1.70591339e-01 -6.13015771e-01 -8.55372310e-01 4.84989852e-01
8.42777193e-01 -7.72170544e-01 -1.07182825e+00 -1.20559104e-01
5.00781238e-01 -6.51920587e-03 -1.03558755e+00 4.48024780e-01
9.07962859e-01 -1.30192733e+00 9.22963262e-01 -1.08886406e-01
2.74504442e-02 -4.12704438e-01 -6.45412564e-01 -1.37641895e+00
-2.80122031e-02 -6.01586223e-01 2.78407604e-01 1.35016358e+00
4.61151034e-01 -4.04745668e-01 4.14570630e-01 2.34825805e-01
-1.87564597e-01 -2.73285002e-01 -9.79247212e-01 -6.75580204e-01
-3.65186989e-01 -9.26053226e-01 5.83273709e-01 8.12694907e-01
3.11997920e-01 2.65387803e-01 -3.85088503e-01 4.90092300e-02
4.13094521e-01 1.90534875e-01 7.21005201e-01 -1.23640263e+00
-8.01095605e-01 -7.24078834e-01 -4.90861475e-01 -6.13588750e-01
-8.95754918e-02 -5.23329854e-01 9.99207795e-02 -8.60285938e-01
-7.52389953e-02 2.23327465e-02 -6.68738604e-01 -1.58001289e-01
1.65392846e-01 3.67349267e-01 3.63934219e-01 1.95867360e-01
-4.93176013e-01 5.47500134e-01 9.98153508e-01 -1.29105806e-01
-5.61061978e-01 4.91862446e-01 -4.20367390e-01 9.71675038e-01
7.03387082e-01 -1.49702609e-01 -3.86334836e-01 -1.09830387e-02
5.31318784e-01 5.21130204e-01 2.65892267e-01 -1.38886738e+00
-5.14999330e-02 5.33167720e-02 3.06425057e-02 -3.78332496e-01
6.77914381e-01 -8.55517626e-01 4.18321669e-01 3.25282753e-01
-7.90970206e-01 -7.73996487e-02 1.80371676e-03 4.62757677e-01
-6.31884277e-01 -4.46800977e-01 5.27069271e-01 1.73020929e-01
-2.55371898e-01 -4.46937293e-01 -4.79547173e-01 -1.18000776e-01
4.95422632e-01 -3.63107562e-01 1.39156058e-01 -6.46842480e-01
-1.24748921e+00 -7.27806211e-01 1.14063226e-01 1.70910016e-01
3.87889922e-01 -1.36041164e+00 -6.09004200e-01 1.82764858e-01
1.07055247e-01 -7.78109789e-01 3.88942510e-01 9.29851234e-01
-2.86698669e-01 2.01358661e-01 -1.47902727e-01 -5.41157663e-01
-1.39548910e+00 3.73270959e-01 4.04985815e-01 7.33587742e-02
-9.41221192e-02 5.75445175e-01 2.53584087e-02 -1.37004927e-01
5.98990679e-01 -6.46026611e-01 -4.28540260e-01 3.70033830e-01
5.11726797e-01 5.34431696e-01 1.10397823e-01 -5.88722646e-01
8.41304809e-02 6.96279883e-01 6.98092580e-01 -7.35325336e-01
7.42597342e-01 -3.48518431e-01 1.74958333e-02 1.24226451e+00
1.01755905e+00 6.99636281e-01 -6.11360312e-01 -3.28619406e-02
-2.08627135e-02 -3.67209584e-01 -1.59211144e-01 -1.08128262e+00
-4.37252820e-01 8.13270926e-01 1.03762245e+00 6.40101910e-01
1.47015381e+00 -2.42174909e-01 5.48672318e-01 5.72583199e-01
3.07733566e-01 -1.47688293e+00 2.92596608e-01 1.51773158e-03
1.18607342e+00 -7.16686189e-01 -4.95236635e-01 -7.06121922e-02
-5.91391504e-01 1.02786148e+00 1.04023673e-01 1.89803794e-01
6.75181985e-01 1.16627142e-01 3.19114238e-01 -4.63420413e-02
-4.02976692e-01 -1.12588830e-01 5.11669815e-01 5.33064365e-01
4.46038902e-01 4.25041884e-01 -4.72170681e-01 9.53462422e-01
-8.56907785e-01 1.61342919e-04 3.23314577e-01 2.95281947e-01
-6.33254588e-01 -9.63885963e-01 -9.12610590e-01 -4.52379957e-02
-4.32354003e-01 1.60461590e-01 -5.00228703e-01 4.39515978e-01
4.83341783e-01 1.11907494e+00 -1.50546893e-01 -6.63838565e-01
6.63268209e-01 4.81802732e-01 6.18708789e-01 -2.57054240e-01
-5.02480924e-01 4.45212066e-01 1.75664172e-01 -5.47216415e-01
-7.11501479e-01 -7.40873873e-01 -7.59286642e-01 1.54569335e-02
-1.81822088e-02 5.94482005e-01 1.01373971e+00 8.80606890e-01
-8.55213255e-02 8.41363311e-01 6.81941867e-01 -8.78750801e-01
-7.34531224e-01 -1.24722052e+00 -8.69965076e-01 5.99280596e-01
1.27615497e-01 -3.91113847e-01 -7.40956128e-01 1.58350453e-01]
|
[15.725272178649902, 5.410780429840088]
|
2d8a1e0f-970e-47e0-a86c-df63352bad28
|
a-warm-start-and-a-clean-crawled-corpus-a
|
2201.05601
| null |
https://arxiv.org/abs/2201.05601v2
|
https://arxiv.org/pdf/2201.05601v2.pdf
|
A Warm Start and a Clean Crawled Corpus -- A Recipe for Good Language Models
|
We train several language models for Icelandic, including IceBERT, that achieve state-of-the-art performance in a variety of downstream tasks, including part-of-speech tagging, named entity recognition, grammatical error detection and constituency parsing. To train the models we introduce a new corpus of Icelandic text, the Icelandic Common Crawl Corpus (IC3), a collection of high quality texts found online by targeting the Icelandic top-level-domain (TLD). Several other public data sources are also collected for a total of 16GB of Icelandic text. To enhance the evaluation of model performance and to raise the bar in baselines for Icelandic, we translate and adapt the WinoGrande dataset for co-reference resolution. Through these efforts we demonstrate that a properly cleaned crawled corpus is sufficient to achieve state-of-the-art results in NLP applications for low to medium resource languages, by comparison with models trained on a curated corpus. We further show that initializing models using existing multilingual models can lead to state-of-the-art results for some downstream tasks.
|
['Svanhvít Lilja Ingólfsdóttir', 'Hafsteinn Einarsson', 'Vilhjálmur Þorsteinsson', 'Haukur Páll Jónsson', 'Pétur Orri Ragnarsson', 'Haukur Barri Símonarson', 'Vésteinn Snæbjarnarson']
|
2022-01-14
| null | null | null | null |
['grammatical-error-detection', 'constituency-parsing']
|
['natural-language-processing', 'natural-language-processing']
|
[-3.07298779e-01 1.82961524e-01 -2.82832831e-02 -3.69976163e-01
-1.78771448e+00 -1.07378411e+00 6.19305134e-01 3.57246757e-01
-7.63527572e-01 8.19805026e-01 6.37428880e-01 -5.76152861e-01
2.79772878e-01 -4.95514125e-01 -8.68016601e-01 -8.68949071e-02
-1.35174453e-01 9.55076039e-01 1.25626281e-01 -4.15425301e-01
-8.29029232e-02 1.64042920e-01 -9.91581559e-01 7.06703842e-01
1.04828691e+00 1.48352817e-01 3.84885937e-01 8.16349089e-01
-2.07640097e-01 5.99299371e-01 -6.73364699e-01 -6.41829729e-01
2.87473947e-01 -2.20509861e-02 -1.24699438e+00 -6.31673932e-01
6.13256395e-01 2.07679197e-01 -1.66305900e-01 8.39480519e-01
4.72568423e-01 -2.91379541e-01 3.55595052e-01 -6.03096485e-01
-3.10943514e-01 1.01469862e+00 1.02459043e-01 3.58448446e-01
3.33246469e-01 2.25414895e-02 1.35137236e+00 -8.92142713e-01
1.31500256e+00 1.15374053e+00 8.07595491e-01 4.24880981e-01
-8.98578465e-01 -6.95686519e-01 1.74129903e-01 -2.11982787e-01
-1.14893281e+00 -7.38788307e-01 1.20601475e-01 -2.86677122e-01
1.71112061e+00 8.50694925e-02 1.71606109e-01 1.18835270e+00
2.73446798e-01 7.04267025e-01 9.64877188e-01 -8.14419985e-01
-2.72803605e-01 -5.03364727e-02 2.10174873e-01 6.86643541e-01
5.21753013e-01 -3.05051729e-02 -6.07740879e-01 -2.10490003e-01
-4.52465229e-02 -6.90455437e-01 -1.33602545e-01 2.11408824e-01
-1.15070963e+00 6.80413127e-01 -2.16927141e-01 5.77246487e-01
-2.57263124e-01 -4.75159055e-03 7.51505196e-01 5.55152297e-01
9.94642079e-01 5.15081048e-01 -1.36121082e+00 -4.98233080e-01
-1.01693499e+00 2.44816527e-01 1.32745576e+00 1.09610999e+00
5.62286615e-01 -4.19438660e-01 2.42047772e-01 1.02813733e+00
3.81011486e-01 8.79015148e-01 3.35595936e-01 -7.72200644e-01
1.50590479e+00 4.14585263e-01 1.93269521e-01 -1.89549044e-01
-5.42587757e-01 -3.29769045e-01 -1.15143932e-01 -1.47634313e-01
6.88098073e-01 -5.40507019e-01 -1.03170180e+00 1.62159455e+00
2.82871634e-01 -5.58439076e-01 6.18708909e-01 4.39886898e-01
9.20059443e-01 8.56439650e-01 4.55420703e-01 8.22924227e-02
1.60595739e+00 -9.10691738e-01 -4.96026248e-01 -7.72818446e-01
1.29552150e+00 -1.09956658e+00 8.33416879e-01 1.25640735e-01
-9.46139276e-01 -3.02850455e-01 -9.89783764e-01 -2.88349390e-01
-6.29830897e-01 2.04537630e-01 3.41939211e-01 6.00651920e-01
-8.65207016e-01 5.54450929e-01 -1.27102184e+00 -8.18841636e-01
-1.93243489e-01 -5.23074791e-02 -8.67190301e-01 -2.44103789e-01
-1.45229352e+00 1.10228133e+00 5.77275217e-01 -3.07755053e-01
-7.68679023e-01 -8.29411685e-01 -1.15947783e+00 -1.84855089e-01
-3.95237729e-02 -6.76615685e-02 1.34161806e+00 -2.80656606e-01
-9.73750412e-01 1.44781411e+00 -1.13363806e-02 -4.08981532e-01
1.86825141e-01 -7.71090806e-01 -6.97075725e-01 4.25572246e-02
6.55731440e-01 4.83438373e-01 4.46040817e-02 -8.91690373e-01
-9.98606741e-01 -3.52800101e-01 -2.51184672e-01 2.10345834e-01
-3.73738185e-02 5.64897954e-01 -5.17223656e-01 -4.88019973e-01
2.01136284e-02 -1.03735983e+00 -1.97540611e-01 -9.26776052e-01
-1.66824237e-01 -1.69348821e-01 5.07790446e-01 -1.60305703e+00
1.21085930e+00 -1.90079308e+00 -2.60232955e-01 -1.36662498e-01
-3.10287207e-01 5.56145966e-01 -4.38182473e-01 9.52189922e-01
1.87010571e-01 5.30839562e-01 -3.42959225e-01 -5.10501802e-01
-4.71820906e-02 3.90574634e-01 1.06932729e-01 4.07753706e-01
5.59992671e-01 7.67758250e-01 -9.38399851e-01 -5.23314476e-01
-1.79556623e-01 2.99317449e-01 -3.69334161e-01 2.61624187e-01
-2.72195756e-01 4.16893244e-01 -2.87277162e-01 6.11662507e-01
6.15912855e-01 2.61949927e-01 5.70586920e-01 3.71301830e-01
-4.35884535e-01 1.37628889e+00 -7.23969579e-01 2.06691456e+00
-9.01664138e-01 4.34805840e-01 5.33989131e-01 -5.78853965e-01
8.71006191e-01 3.50040853e-01 4.50273082e-02 -6.36457086e-01
-3.48449916e-01 6.72482371e-01 -1.48293795e-02 -5.74227452e-01
8.41834366e-01 9.09268036e-02 -7.20934510e-01 2.66979337e-01
4.95717585e-01 -8.31468031e-02 5.11382997e-01 2.55783081e-01
1.48960173e+00 5.09164751e-01 1.90201879e-01 -7.09602177e-01
4.20358360e-01 6.71983242e-01 8.87074172e-01 6.22887254e-01
-3.83501463e-02 5.47538877e-01 2.24880874e-01 -3.88402432e-01
-1.31580639e+00 -7.85073817e-01 -2.59265631e-01 1.37200940e+00
-5.91113627e-01 -7.15251803e-01 -8.90145957e-01 -1.14413834e+00
-2.23165557e-01 7.55981863e-01 -2.74740368e-01 5.62071323e-01
-1.46395171e+00 -9.64317977e-01 1.29631567e+00 2.87286818e-01
5.27609229e-01 -1.08966017e+00 -3.22350711e-01 6.29600406e-01
-6.89295292e-01 -1.54906368e+00 -3.72197479e-01 6.01676464e-01
-6.31302714e-01 -1.14349186e+00 -5.67753077e-01 -1.16519165e+00
2.25155074e-02 -3.94620508e-01 1.80507684e+00 -1.46984383e-01
4.27051634e-02 2.44539589e-01 -6.91463888e-01 -2.27357075e-01
-1.21942890e+00 8.01713705e-01 -2.89512068e-01 -8.65448236e-01
5.80985546e-01 -2.14070410e-01 -4.02400903e-02 -1.32742897e-01
-5.79017341e-01 -2.76894361e-01 5.71300387e-01 6.70316577e-01
2.01499209e-01 -6.50643229e-01 4.65802848e-01 -1.24359000e+00
2.33970687e-01 -4.07995164e-01 -5.87415338e-01 3.61007929e-01
-3.62299919e-01 3.83352697e-01 8.87649953e-01 2.51145214e-01
-1.49886155e+00 3.77565324e-02 -7.07661808e-01 5.49840868e-01
-3.66241574e-01 5.97577155e-01 -3.33136678e-01 4.02821004e-01
7.05159128e-01 -4.76193018e-02 -4.03867990e-01 -9.97227967e-01
6.95462823e-01 1.12296832e+00 6.75393760e-01 -8.08253348e-01
4.65189427e-01 1.12990579e-02 -5.15823245e-01 -7.56129444e-01
-9.47249472e-01 -7.77076960e-01 -9.03216958e-01 4.04282689e-01
8.81308794e-01 -1.44476974e+00 -5.62149510e-02 2.99667478e-01
-1.36235428e+00 -6.17085755e-01 9.74655896e-02 3.64727050e-01
-1.63234487e-01 2.93031156e-01 -1.29552913e+00 -5.23242712e-01
-8.19454312e-01 -7.83513725e-01 1.25536454e+00 -2.52095491e-01
-1.80981308e-01 -1.13043177e+00 6.56951487e-01 4.18279916e-01
-1.04981773e-01 1.13205956e-02 1.09077585e+00 -1.13918924e+00
-6.54246360e-02 7.78317153e-02 -1.45517047e-02 3.41349453e-01
-2.59704351e-01 -2.68362790e-01 -6.34529412e-01 -4.12504822e-01
-3.13042670e-01 -3.03952008e-01 7.93685019e-01 -4.66727503e-02
-1.21394485e-01 -2.18749970e-01 -4.72128719e-01 4.25800413e-01
1.49863517e+00 9.47997347e-02 5.84515870e-01 9.19468701e-01
4.98818368e-01 6.81632280e-01 7.25971997e-01 8.05411413e-02
6.99301302e-01 3.19728881e-01 -9.21443552e-02 -2.38931701e-02
-4.37189996e-01 -5.22474647e-01 6.10937178e-01 1.23950875e+00
3.87359828e-01 -3.42746496e-01 -1.54991734e+00 1.11486554e+00
-1.71096814e+00 -5.31796157e-01 -3.50043416e-01 1.98953676e+00
1.12836921e+00 1.09135490e-02 -2.86598295e-01 -4.13969278e-01
7.66079605e-01 1.85851261e-01 1.75460547e-01 -4.74325597e-01
-3.22014779e-01 5.03774524e-01 9.26189661e-01 6.65018439e-01
-1.36262977e+00 1.51605403e+00 6.66354799e+00 6.99295044e-01
-7.47918308e-01 5.51728904e-01 5.09944320e-01 1.83566719e-01
-1.41743526e-01 2.13024586e-01 -1.61079502e+00 3.06728840e-01
1.75690269e+00 2.52671093e-01 1.84531257e-01 6.87466204e-01
2.64155567e-01 1.01170011e-01 -8.73782575e-01 4.62727934e-01
-5.14948741e-02 -1.25210667e+00 -4.03450131e-01 -1.01139396e-01
8.38856101e-01 1.05794179e+00 -6.66894138e-01 7.03367889e-01
9.95128572e-01 -7.22033262e-01 7.21004963e-01 -7.91236386e-02
9.40308332e-01 -6.77219570e-01 1.05472112e+00 4.09729779e-01
-1.09364331e+00 3.09829265e-01 -4.04019296e-01 1.75197572e-01
3.67286563e-01 5.98875701e-01 -9.03866827e-01 8.29983056e-01
9.13117886e-01 5.76971889e-01 -6.39732778e-01 5.88447094e-01
-5.39252758e-01 1.16524160e+00 -6.68727756e-01 -3.77387367e-03
5.85805237e-01 8.01015645e-02 6.24182582e-01 1.95328331e+00
1.69055194e-01 -1.10414274e-01 2.76637465e-01 2.47981086e-01
-4.47079599e-01 6.00710869e-01 -4.77052063e-01 1.11397738e-02
6.00270867e-01 1.26455057e+00 -3.49914312e-01 -4.57550257e-01
-6.73351228e-01 8.05165768e-01 8.39154482e-01 -2.82731671e-02
-3.62134457e-01 -6.92185640e-01 6.78766429e-01 1.64229259e-01
2.77730405e-01 -5.31845689e-01 -1.28632262e-01 -1.24032831e+00
1.74921930e-01 -1.13033760e+00 7.00112283e-01 -4.05707061e-01
-1.27700913e+00 8.08267534e-01 -1.54026700e-02 -7.61307478e-01
-5.53452790e-01 -8.15064788e-01 -3.02201569e-01 9.80577350e-01
-1.57047904e+00 -1.47251856e+00 2.99514055e-01 5.78853823e-02
6.03146255e-01 -1.74071595e-01 1.05290759e+00 5.93514919e-01
-2.26773545e-01 3.74822378e-01 4.32657480e-01 7.29930699e-01
9.91113424e-01 -1.26052988e+00 1.21310866e+00 1.24241257e+00
2.90085852e-01 4.50439334e-01 5.79493880e-01 -1.03122854e+00
-1.17510128e+00 -1.37676251e+00 1.76448452e+00 -7.14492679e-01
9.47820067e-01 -7.82563448e-01 -8.41531336e-01 9.54948962e-01
4.33037192e-01 -2.48431578e-01 4.96671438e-01 5.20644665e-01
-2.98875600e-01 2.58530259e-01 -1.03740644e+00 1.75611019e-01
1.24447000e+00 -3.81033808e-01 -1.00867093e+00 5.22238553e-01
8.52015555e-01 -4.86893505e-01 -1.13743126e+00 2.73097903e-01
3.60968232e-01 -6.16482854e-01 3.85803342e-01 -7.34955549e-01
3.20125431e-01 1.20974079e-01 -2.78985471e-01 -1.54106879e+00
-1.53181612e-01 -7.87609756e-01 6.13382578e-01 1.83765948e+00
1.05544639e+00 -6.59994841e-01 5.95484138e-01 4.21204716e-01
-7.20139265e-01 3.44711207e-02 -1.24072576e+00 -9.91970599e-01
7.61560798e-01 -4.42895979e-01 2.86634356e-01 8.31401706e-01
3.07927281e-01 6.19118214e-01 7.35594630e-02 3.09935868e-01
3.36752743e-01 -2.17362240e-01 5.86597919e-01 -1.09682775e+00
-1.50816828e-01 3.41691047e-01 -2.55539902e-02 -8.85424018e-01
4.81693953e-01 -1.21269548e+00 5.49224019e-01 -1.70432818e+00
-9.27837286e-03 -8.59818816e-01 3.57506007e-01 7.38299787e-01
-1.61784664e-01 1.08137473e-01 2.88516045e-01 2.91698784e-01
-6.17824733e-01 2.98063420e-02 5.34228086e-01 2.06398293e-01
-2.20671803e-01 -5.78819752e-01 -5.48685849e-01 6.20283246e-01
7.04904556e-01 -1.00974917e+00 5.44537723e-01 -8.31705749e-01
2.81535864e-01 5.81775308e-02 -3.48952711e-01 -9.86725450e-01
-6.50084615e-02 2.68657029e-01 1.01478606e-01 -6.96670651e-01
-5.96021563e-02 -3.43983710e-01 -1.26938252e-02 6.22055113e-01
-1.69225350e-01 4.13128793e-01 4.01017159e-01 9.93614197e-02
-2.94489831e-01 -4.89107043e-01 6.76786780e-01 -6.02616787e-01
-8.09468567e-01 -1.60168812e-01 -7.45204389e-01 8.85423303e-01
2.55928844e-01 6.34284854e-01 -7.11551547e-01 3.37920561e-02
-4.53328788e-01 1.71128720e-01 3.41741443e-01 4.68203068e-01
-3.69010329e-01 -7.13512063e-01 -1.34594357e+00 1.91056341e-01
2.91710466e-01 -1.96772188e-01 -3.40950608e-01 2.29786143e-01
-9.59132731e-01 9.39223468e-01 8.88867825e-02 -3.61131161e-01
-1.09033525e+00 8.51746723e-02 -2.26511359e-02 -1.22555566e+00
-7.14510262e-01 4.02754635e-01 -2.54400849e-01 -1.18450785e+00
-3.86305660e-01 -1.75861001e-01 5.81488013e-02 -1.75417542e-01
3.51325393e-01 1.16837472e-01 6.71901047e-01 -8.10509384e-01
-6.77109957e-01 3.32572937e-01 -1.67665720e-01 -5.01331568e-01
1.44987965e+00 -2.59341002e-01 8.62742588e-02 1.38770178e-01
1.14543939e+00 4.13875371e-01 -8.56482804e-01 2.32117996e-02
4.80367690e-01 1.10189974e-01 -2.07207710e-01 -1.33241928e+00
-4.21544552e-01 6.56239450e-01 2.32738599e-01 -1.50343046e-01
5.80676973e-01 2.92273641e-01 9.58460867e-01 6.02565527e-01
6.40439808e-01 -1.32147431e+00 -7.16527164e-01 1.19907844e+00
4.65438962e-01 -1.29815209e+00 -2.62892187e-01 -5.34196079e-01
-5.42926550e-01 8.69301140e-01 2.71192491e-01 -1.38447449e-01
6.02663875e-01 8.14669907e-01 5.63946605e-01 -4.62118946e-02
-7.55645633e-01 -3.83859545e-01 -2.87693124e-02 6.78755641e-01
9.20571744e-01 6.86368719e-02 -8.49918842e-01 7.32494771e-01
-4.64196175e-01 -4.99272585e-01 3.79032940e-01 1.00659955e+00
-4.14530516e-01 -1.58720267e+00 -3.00709099e-01 2.16993630e-01
-1.34456182e+00 -8.71708333e-01 -1.64542407e-01 1.33671701e+00
-1.03091396e-01 1.21138453e+00 5.50701208e-02 7.58442357e-02
3.69882315e-01 5.21812379e-01 2.09319383e-01 -9.39016759e-01
-1.12784648e+00 1.19050659e-01 1.13608837e+00 -2.99392164e-01
-2.03039363e-01 -7.10382819e-01 -1.29241705e+00 -1.27811313e-01
-2.38828436e-01 5.96690893e-01 9.85552490e-01 9.31439281e-01
5.60938478e-01 2.92125549e-02 4.39309478e-02 -5.98286688e-01
-2.87968248e-01 -1.41309166e+00 -3.52293283e-01 5.27355731e-01
-3.04560125e-01 1.62400659e-02 -3.53761941e-01 2.55714327e-01]
|
[10.408175468444824, 9.840657234191895]
|
ab86937f-09e6-465e-8a6e-20127d1f3de5
|
rgcl-wlv-at-semeval-2019-task-12-toponym
| null | null |
https://aclanthology.org/S19-2228
|
https://aclanthology.org/S19-2228.pdf
|
RGCL-WLV at SemEval-2019 Task 12: Toponym Detection
|
This article describes the system submitted by the RGCL-WLV team to the SemEval 2019 Task 12: Toponym resolution in scientific papers. The system detects toponyms using a bootstrapped machine learning (ML) approach which classifies names identified using gazetteers extracted from the GeoNames geographical database. The paper evaluates the performance of several ML classifiers, as well as how the gazetteers influence the accuracy of the system. Several runs were submitted. The highest precision achieved for one of the submissions was 89{\%}, albeit it at a relatively low recall of 49{\%}.
|
['Constantin Or{\\u{a}}san', 'Tharindu Ranasinghe', 'Pablo Calleja', 'Alistair Plum', 'Ruslan Mitkov']
|
2019-06-01
| null | null | null |
semeval-2019-6
|
['toponym-resolution']
|
['natural-language-processing']
|
[-3.41311663e-01 1.53134137e-01 -4.17268813e-01 -6.35993257e-02
-7.42375672e-01 -9.39363480e-01 1.17265654e+00 5.65811634e-01
-6.91174746e-01 1.19305158e+00 -8.88936073e-02 -2.80139714e-01
-4.94270205e-01 -6.21607661e-01 -4.46921587e-01 -3.23954731e-01
2.23543465e-01 9.04655516e-01 1.09127909e-02 -4.67139557e-02
8.22622478e-01 6.22986197e-01 -1.36251342e+00 3.45877707e-01
8.09001982e-01 4.13679600e-01 -3.86179015e-02 5.95173717e-01
-4.15832013e-01 4.90756750e-01 -1.25949788e+00 -6.59733832e-01
6.85255378e-02 1.26372397e-01 -9.23707485e-01 -1.09586096e+00
9.03636396e-01 5.91326773e-01 -4.78623286e-02 1.02527094e+00
2.33196616e-01 -2.70216286e-01 1.07088065e+00 -1.22291303e+00
-4.92522985e-01 1.28175604e+00 -4.13549006e-01 6.05725706e-01
7.46819019e-01 -6.09861314e-01 1.19508922e+00 -1.11591709e+00
1.15446341e+00 9.45682108e-01 7.01571584e-01 8.34091529e-02
-9.53881025e-01 -1.20711625e+00 -3.25775027e-01 2.09498569e-01
-2.22246265e+00 -4.00881499e-01 2.31917072e-02 -7.19569266e-01
1.00183666e+00 4.93484318e-01 1.71576038e-01 8.58188868e-01
1.19222403e-02 9.88323465e-02 1.14858115e+00 -8.32686663e-01
3.01034659e-01 5.27288616e-01 3.29166263e-01 2.76295811e-01
9.14360464e-01 -3.78706932e-01 -7.40165770e-01 -6.75073683e-01
4.56299663e-01 -5.79060614e-01 -1.06882602e-01 1.27103776e-01
-1.41920352e+00 5.32504678e-01 2.68670440e-01 7.70140886e-01
-2.27012843e-01 -5.23638606e-01 1.00601785e-01 3.83767247e-01
3.36705655e-01 1.38270509e+00 -7.47195959e-01 -7.96561167e-02
-1.55309236e+00 3.82297486e-01 1.19194639e+00 1.11526144e+00
4.82894838e-01 -4.74888116e-01 -8.75617936e-02 7.64741480e-01
1.00891322e-01 4.67020243e-01 3.56226891e-01 -7.32537091e-01
4.53377336e-01 5.39766610e-01 6.19751155e-01 -9.89832878e-01
-7.54623532e-01 -5.53877950e-01 -2.01398715e-01 -2.07875714e-01
7.12213814e-01 -7.56063238e-02 -7.10773706e-01 1.34549320e+00
-2.12968424e-01 3.00527867e-02 -1.52392611e-01 2.94858277e-01
1.45423710e+00 5.05254209e-01 4.36523944e-01 -2.79451638e-01
1.13667345e+00 -2.98459500e-01 -7.99349606e-01 2.86004066e-01
8.50194991e-01 -1.02782607e+00 2.28501931e-01 4.21191156e-01
-7.01165378e-01 -4.59466219e-01 -1.24688482e+00 4.03159082e-01
-1.11160314e+00 4.33286160e-01 1.59988344e-01 7.01966405e-01
-1.06430054e+00 7.44613051e-01 -1.61054105e-01 -7.08478034e-01
5.74256293e-02 4.81135994e-01 -4.22796488e-01 3.31042111e-01
-1.56086814e+00 1.34260714e+00 4.93804961e-01 -3.74830961e-01
5.68854213e-02 -9.50794280e-01 -3.06032002e-01 -2.55844921e-01
-7.27855489e-02 -3.36370856e-01 7.71508634e-01 -2.18306839e-01
-8.48004580e-01 1.33194888e+00 -1.65644631e-01 -2.64841974e-01
5.42835653e-01 -1.63886890e-01 -1.20476425e+00 -1.53834283e-01
6.83988810e-01 3.03014815e-01 1.59119040e-01 -8.70665133e-01
-8.53165627e-01 -3.92065108e-01 -5.36711454e-01 -7.16471598e-02
-7.80103132e-02 4.13510054e-01 -2.24416941e-01 -6.39514625e-01
3.90405171e-02 -9.32538331e-01 3.96722585e-01 -8.59678388e-01
-4.98655856e-01 -8.35668981e-01 2.02974290e-01 -7.20344305e-01
1.69705915e+00 -1.82915461e+00 -2.29439512e-02 5.72023749e-01
3.38218123e-01 3.45761925e-01 2.91430712e-01 7.12342203e-01
-3.76273811e-01 4.43064868e-01 1.88398108e-01 2.12100610e-01
-2.82891933e-02 -3.57722163e-01 -5.53299665e-01 4.45316046e-01
-2.63309270e-01 5.38520575e-01 -9.36427712e-01 -4.89499539e-01
1.91487357e-01 1.94807962e-01 2.51559049e-01 -2.70371735e-01
1.56808987e-01 5.00640385e-02 -4.06373799e-01 6.69203877e-01
4.59853917e-01 -1.54308975e-01 1.91500068e-01 9.00968537e-02
-4.19789076e-01 2.85248637e-01 -1.28578651e+00 1.28210640e+00
-2.45825514e-01 8.37274790e-01 -3.06020498e-01 -4.86086756e-01
1.23063552e+00 3.69217336e-01 3.40339750e-01 -3.57768089e-01
-3.40501130e-01 8.34248126e-01 -3.09558183e-01 -2.49900863e-01
5.62119722e-01 5.69662809e-01 -3.05999994e-01 1.09645620e-01
-1.36439002e-03 -5.10495342e-02 3.46278936e-01 4.69343603e-01
1.06821525e+00 2.08044305e-01 7.75823653e-01 -6.96012914e-01
7.82966256e-01 3.03178519e-01 3.89472336e-01 1.16612768e+00
1.37493294e-02 4.77655441e-01 1.67800426e-01 -7.26974964e-01
-8.80366504e-01 -9.87754345e-01 -8.39475989e-01 1.19229162e+00
-7.89496824e-02 -8.41002941e-01 -7.31940806e-01 -4.20140386e-01
3.78323019e-01 1.06841576e+00 -5.35856366e-01 1.11764096e-01
-3.40531945e-01 -7.09276676e-01 9.70419466e-01 3.31137865e-03
-8.24119821e-02 -1.15449929e+00 -4.14749503e-01 1.86077841e-02
-1.07769310e-01 -1.30000937e+00 8.14298317e-02 1.43273054e-02
-3.47150594e-01 -1.13474226e+00 -7.95791268e-01 -6.14123821e-01
2.23269686e-01 -3.69438350e-01 1.44543386e+00 -1.16108298e-01
-3.17431718e-01 -1.77087232e-01 -2.74099588e-01 -6.82346106e-01
-2.54672110e-01 8.25082719e-01 3.07252407e-01 -4.87423569e-01
1.27667236e+00 -1.91909656e-01 -1.23079233e-01 2.02292591e-01
-1.39001885e-03 -5.41021287e-01 6.23256981e-01 4.66186285e-01
2.33213216e-01 -4.63939697e-01 9.89743352e-01 -1.30614793e+00
6.75304115e-01 -7.05642879e-01 -7.15563774e-01 6.07149005e-01
-1.13525605e+00 -9.38528478e-02 4.70949054e-01 1.07367463e-01
-5.51455259e-01 5.23732007e-02 9.58033726e-02 -3.17684957e-03
-2.95293123e-01 3.43705654e-01 5.26133478e-02 -2.18958139e-01
1.05466628e+00 -1.64601073e-01 -7.44833052e-01 -7.60686934e-01
4.75077838e-01 1.26962137e+00 5.86578131e-01 -4.27484959e-01
6.67148709e-01 -1.94994450e-01 -8.46006721e-02 -9.63017642e-01
-6.81110740e-01 -7.12495565e-01 -8.54467809e-01 3.77576239e-02
5.87291062e-01 -1.11241567e+00 -6.58715725e-01 1.23836115e-01
-1.12795997e+00 4.59094614e-01 1.07905038e-01 7.19393611e-01
-4.54973988e-02 -1.68820769e-01 -2.36222342e-01 -5.45274734e-01
-4.77710426e-01 -8.22205842e-01 8.44332695e-01 4.81528759e-01
-1.15069854e+00 -7.02458143e-01 2.65972108e-01 1.56182960e-01
3.50345850e-01 3.55052233e-01 8.20440292e-01 -1.48081934e+00
2.91411936e-01 -5.22494733e-01 -2.45591760e-01 -6.31939411e-01
4.49666493e-02 4.19584751e-01 -1.09200788e+00 -1.37444779e-01
-7.61789739e-01 7.50567615e-02 7.12817669e-01 2.11855397e-01
8.39135349e-01 -8.46449882e-02 -9.71766353e-01 3.78056258e-01
1.22386813e+00 2.24259838e-01 5.45632020e-02 9.92226720e-01
6.84024155e-01 4.82884973e-01 3.67491901e-01 3.64616901e-01
9.39833447e-02 8.34400058e-01 -1.04357265e-01 3.60776216e-01
9.40910578e-02 -7.54591599e-02 -3.60948026e-01 4.25233126e-01
-7.71812201e-02 -5.16719595e-02 -1.46320117e+00 5.63553393e-01
-1.53148985e+00 -8.31501663e-01 -4.40890372e-01 2.17073464e+00
1.10371745e+00 1.93706825e-01 1.57095283e-01 -9.01102349e-02
1.01485038e+00 -1.08280955e-02 -5.57104535e-02 -6.49500966e-01
-3.19067955e-01 5.07504582e-01 1.00482595e+00 4.46258008e-01
-1.08646905e+00 1.05377984e+00 7.54693604e+00 7.32383430e-01
-7.65213788e-01 -2.67943263e-01 1.46624163e-01 -7.14706928e-02
7.79863000e-02 -9.74501893e-02 -1.09492111e+00 7.57242560e-01
1.39656699e+00 -5.78143299e-01 8.42630491e-02 7.49269605e-01
-3.40704441e-01 -4.72651273e-02 -9.65524971e-01 1.14210856e+00
1.09070726e-01 -1.61894643e+00 5.27249984e-02 -9.86257344e-02
6.18638813e-01 3.28616589e-01 -3.09610814e-01 1.69638231e-01
6.02071166e-01 -1.33460438e+00 6.38323784e-01 8.00587118e-01
9.94033575e-01 -8.40036750e-01 7.93992937e-01 2.89256662e-01
-7.94185340e-01 4.41290326e-02 -2.60975271e-01 3.59039456e-02
-5.17974615e-01 7.19126046e-01 -9.46330607e-01 7.09580183e-01
9.87010598e-01 7.66766250e-01 -1.13608873e+00 1.26219821e+00
-2.91678429e-01 5.33231795e-01 -4.40439254e-01 -3.45533788e-01
-1.31300703e-01 6.14720434e-02 6.64755464e-01 1.70524013e+00
1.03007033e-01 -1.17110424e-01 -1.68021291e-01 6.79750204e-01
-3.11639935e-01 2.45385513e-01 -6.97992504e-01 7.80953839e-02
1.43994188e+00 1.20130265e+00 -5.22747159e-01 -3.56853873e-01
-9.92534682e-02 3.65438521e-01 2.55052775e-01 3.90609562e-01
-3.88427675e-01 -1.07656360e+00 2.88733602e-01 4.28321511e-02
-2.42882892e-02 1.56094253e-01 -6.63501561e-01 -7.97138155e-01
-1.27372056e-01 -6.16564631e-01 9.21439767e-01 -6.59777880e-01
-1.10342741e+00 7.34819174e-01 2.44819909e-01 -1.01562905e+00
-2.94193178e-01 -6.30587935e-01 -2.34512538e-01 1.32228196e+00
-8.63566756e-01 -8.79897237e-01 -7.08108842e-02 1.74528405e-01
-1.11611903e-01 -7.55799294e-01 1.15078664e+00 2.44771674e-01
-6.36392832e-01 8.12012196e-01 5.45038164e-01 2.97449857e-01
1.44154036e+00 -1.56654191e+00 5.93211353e-01 3.50336105e-01
3.71956170e-01 1.09212780e+00 9.52051997e-01 -7.50725925e-01
-8.15654218e-01 -9.11944151e-01 1.72811103e+00 -1.17275274e+00
7.45869458e-01 -7.53809735e-02 -6.87878251e-01 6.04977906e-01
3.32004011e-01 -2.91744858e-01 7.04848409e-01 4.59092021e-01
-5.57191491e-01 4.53224704e-02 -1.09409595e+00 2.87316740e-01
5.97226441e-01 -5.59315920e-01 -1.05729890e+00 4.45541650e-01
2.37958983e-01 -3.39997828e-01 -1.22756696e+00 3.33301157e-01
6.50329053e-01 -1.27035186e-01 1.03112233e+00 -6.40276909e-01
1.49800584e-01 -3.73106539e-01 8.74064937e-02 -1.11709285e+00
-4.92962450e-01 -4.02954549e-01 2.87892390e-02 1.37861502e+00
9.50891137e-01 -6.38149500e-01 4.79329199e-01 5.92529058e-01
4.43302244e-01 -1.63425595e-01 -1.13477886e+00 -7.36854255e-01
4.92079139e-01 1.65603817e-01 5.59904218e-01 1.57556951e+00
3.97521019e-01 4.04619515e-01 9.21520637e-04 -7.06914291e-02
5.74486554e-01 -1.45500666e-03 4.78861094e-01 -1.82436049e+00
5.30421257e-01 -6.49713159e-01 -5.34490168e-01 1.47044927e-01
1.86028004e-01 -1.12429452e+00 -4.37320590e-01 -1.35526431e+00
-4.45345975e-02 -3.80305916e-01 -7.10853875e-01 5.12534738e-01
-1.65398091e-01 3.14793974e-01 -1.05550170e-01 8.46508503e-01
-6.09945357e-01 -5.04432261e-01 1.82519510e-01 -2.65661371e-03
-1.72735810e-01 -7.36209974e-02 -8.10132563e-01 5.49875081e-01
5.82490325e-01 -8.38375330e-01 3.35902840e-01 -1.24204859e-01
5.66049695e-01 -4.71360415e-01 -7.75605291e-02 -8.96942258e-01
5.85826218e-01 -9.51604173e-03 8.34701598e-01 -6.48631155e-01
-2.82887191e-01 -2.84195781e-01 3.97516012e-01 3.88055474e-01
-6.09465182e-01 4.27069604e-01 2.34511420e-01 2.44307205e-01
-7.47925639e-02 -1.37387395e-01 5.32581151e-01 -2.64973134e-01
-8.03798795e-01 -3.07263821e-01 -2.79370993e-01 2.48778403e-01
8.98845971e-01 -2.20673233e-02 -5.86833417e-01 2.88405597e-01
-6.96322262e-01 1.77404061e-01 5.62550068e-01 6.55973852e-01
-5.42983785e-02 -1.02438188e+00 -1.08378637e+00 -1.63061947e-01
6.79844499e-01 -7.01568782e-01 -5.42780817e-01 4.84483331e-01
-8.37289274e-01 9.63484645e-01 -2.75259256e-01 -3.06828052e-01
-1.08397031e+00 3.08613271e-01 2.84667969e-01 -1.73272938e-02
-4.75317180e-01 9.48460877e-01 -6.00595176e-01 -5.00853300e-01
3.65617275e-01 4.42280173e-01 -9.87875044e-01 5.08540332e-01
7.55481005e-01 6.46359384e-01 5.43182015e-01 -8.46859217e-01
-1.02019465e+00 7.45232046e-01 -3.73275667e-01 -2.08242968e-01
1.13875782e+00 2.36053512e-01 -2.69604772e-01 6.74226642e-01
8.70244861e-01 1.72555491e-01 1.71440005e-01 -3.52792889e-01
8.89218569e-01 -2.78050870e-01 2.28235144e-02 -1.29835331e+00
-4.04998362e-01 2.50169396e-01 5.01026332e-01 2.70859122e-01
3.67871404e-01 -1.12793192e-01 1.44443676e-01 5.87205529e-01
3.65068465e-01 -1.45879769e+00 -1.07440066e+00 3.97756696e-01
7.15842426e-01 -1.08645988e+00 4.94693607e-01 -1.40363351e-01
-2.36264169e-01 1.23690689e+00 2.72289783e-01 8.20261687e-02
7.50175893e-01 3.34393717e-02 1.54571936e-01 -3.06906819e-01
-4.07167643e-01 2.96105981e-01 5.19952357e-01 3.01396072e-01
7.17540562e-01 3.23675603e-01 -9.95917976e-01 6.60488904e-01
-5.66691101e-01 2.46249977e-02 5.19899249e-01 7.09481776e-01
-2.62492180e-01 -9.28070009e-01 -5.11068523e-01 6.43811703e-01
-8.97189200e-01 -3.00925970e-01 -1.32560265e+00 9.24156845e-01
2.23608270e-01 9.39120293e-01 -6.03320748e-02 -3.84295553e-01
2.96225548e-01 3.48551631e-01 6.22592792e-02 -7.79740632e-01
-1.01155245e+00 -4.59826618e-01 2.99094051e-01 -3.01462680e-01
-3.15761536e-01 -9.00482535e-01 -9.63516116e-01 -3.89552802e-01
-2.36387759e-01 8.38652611e-01 6.50289834e-01 1.05608165e+00
2.87934303e-01 1.80934042e-01 4.42189485e-01 -1.35600209e-01
-2.74079174e-01 -1.20605922e+00 -6.41912699e-01 3.56372565e-01
-5.97369708e-02 -6.53848827e-01 -5.10446489e-01 -1.99749604e-01]
|
[9.189214706420898, 8.951967239379883]
|
63abadf3-02bc-40f3-b752-ee1c36f00af4
|
enhancing-deep-neural-networks-testing-by
|
2112.01956
| null |
https://arxiv.org/abs/2112.01956v1
|
https://arxiv.org/pdf/2112.01956v1.pdf
|
Enhancing Deep Neural Networks Testing by Traversing Data Manifold
|
We develop DEEPTRAVERSAL, a feedback-driven framework to test DNNs. DEEPTRAVERSAL first launches an offline phase to map media data of various forms to manifolds. Then, in its online testing phase, DEEPTRAVERSAL traverses the prepared manifold space to maximize DNN coverage criteria and trigger prediction errors. In our evaluation, DNNs executing various tasks (e.g., classification, self-driving, machine translation) and media data of different types (image, audio, text) were used. DEEPTRAVERSAL exhibits better performance than prior methods with respect to popular DNN coverage criteria and it can discover a larger number and higher quality of error-triggering inputs. The tested DNN models, after being repaired with findings of DEEPTRAVERSAL, achieve better accuracy
|
['Shuai Wang', 'Qi Pang', 'Yuanyuan Yuan']
|
2021-12-03
| null | null | null | null |
['dnn-testing']
|
['adversarial']
|
[-1.52984455e-01 1.04703298e-02 -2.88518190e-01 -1.87810183e-01
-8.22531760e-01 -7.97313809e-01 5.04711688e-01 -1.61509305e-01
-2.97255397e-01 8.63864005e-01 -1.30025759e-01 -5.53485155e-01
-1.83823988e-01 -8.62793088e-01 -1.09322715e+00 -5.01356684e-02
1.39873400e-01 8.31277549e-01 5.03477216e-01 3.56047861e-02
1.73783958e-01 4.42474812e-01 -1.81077123e+00 1.86255366e-01
1.00187707e+00 1.10390866e+00 3.09618592e-01 7.70425320e-01
-1.09658511e-02 8.37744057e-01 -6.11065984e-01 -8.90067339e-01
2.51811624e-01 -2.26037964e-01 -1.04165447e+00 -4.23291922e-02
7.66216815e-01 -4.64572877e-01 -7.02876568e-01 1.32060266e+00
5.65823913e-01 5.84049970e-02 2.35168651e-01 -1.42510176e+00
-4.75380629e-01 9.64920342e-01 3.11629534e-01 5.39151609e-01
2.91753978e-01 5.49059153e-01 1.05713236e+00 -1.33672643e+00
1.03682613e+00 1.18675923e+00 4.21440125e-01 7.24429786e-01
-1.25484204e+00 -7.85359561e-01 -1.75543986e-02 4.65234846e-01
-9.95549083e-01 -7.69324601e-01 3.73948604e-01 -3.25593442e-01
1.14799178e+00 2.94636726e-01 7.52617121e-01 1.61495471e+00
1.03466570e-01 1.08334804e+00 6.54084802e-01 4.09741066e-02
4.37143952e-01 9.56043378e-02 -7.58181140e-02 6.36283755e-01
4.54243012e-02 3.47352684e-01 -6.22557640e-01 2.59566635e-01
7.75680840e-01 -3.82514000e-01 -1.90680828e-02 2.88598716e-01
-1.55394363e+00 3.31784099e-01 1.22868251e-02 8.66191909e-02
-3.61824751e-01 -1.37905348e-02 5.26752114e-01 8.34954321e-01
3.08485061e-01 8.39186966e-01 -7.59649932e-01 -8.50811243e-01
-8.89733374e-01 1.66629866e-01 5.55592835e-01 1.48793936e+00
6.48085773e-01 2.41027221e-01 -4.46464777e-01 6.83269560e-01
2.41041735e-01 4.07582223e-01 5.24552584e-01 -7.56720126e-01
1.00003731e+00 7.86214709e-01 -4.17722613e-01 -5.37034750e-01
-2.99711853e-01 -5.42885184e-01 -5.73014677e-01 -2.13053346e-01
2.35846311e-01 -3.86687368e-01 -1.10289502e+00 1.66997111e+00
3.42077166e-01 5.40780306e-01 2.30626881e-01 8.46733749e-01
8.82105350e-01 5.14662266e-01 -1.73926875e-01 1.17559150e-01
8.45022619e-01 -8.82101178e-01 -4.58884954e-01 2.82150228e-02
8.57150316e-01 -5.63893557e-01 1.39061856e+00 7.30002761e-01
-1.19856393e+00 -5.88580370e-01 -9.19555962e-01 1.72522172e-01
-1.46485209e-01 1.48965344e-01 3.64052653e-01 4.42009896e-01
-1.32778108e+00 9.38175857e-01 -7.00541019e-01 -2.18280748e-01
5.77205777e-01 2.19078481e-01 -3.15702170e-01 -3.36289674e-01
-1.26174593e+00 7.51771033e-01 5.24155080e-01 -2.61361241e-01
-1.76118922e+00 -9.65977490e-01 -5.85248888e-01 -2.38897488e-01
2.54151583e-01 -5.40634751e-01 1.25721133e+00 -6.90576553e-01
-1.51502299e+00 7.01855183e-01 2.29846135e-01 -6.09192252e-01
9.74924088e-01 -1.25370920e-01 -9.58300173e-01 5.83085828e-02
-5.84018119e-02 1.01970673e+00 7.09716201e-01 -6.46873653e-01
-7.52614617e-01 -8.31255317e-02 5.00199795e-02 -8.24026763e-02
-5.01561403e-01 -1.28559321e-01 -6.73739851e-01 -5.00591516e-01
1.26917079e-01 -9.57999051e-01 6.39975592e-02 -2.32856393e-01
-1.14145601e+00 -3.00375193e-01 7.87369370e-01 -2.68295109e-01
1.15072513e+00 -1.96879005e+00 3.85609388e-01 2.32241228e-01
3.99203658e-01 2.30221555e-01 -5.13857305e-01 2.58930147e-01
3.14619131e-02 3.22815716e-01 2.54510224e-01 -4.49908495e-01
3.81448045e-02 1.90423101e-01 -1.77052885e-01 1.28387019e-01
5.09019792e-01 9.08405304e-01 -9.91073608e-01 -5.08660197e-01
1.57179207e-01 -9.80939418e-02 -7.88901269e-01 2.28818178e-01
-6.21617436e-01 3.69567811e-01 -8.64714459e-02 9.94015813e-01
4.87447649e-01 -2.50634491e-01 -2.40355954e-01 -6.16395846e-03
-6.00637011e-02 5.91452122e-01 -9.65111434e-01 1.59167361e+00
-1.38622776e-01 1.07983851e+00 -4.63398248e-01 -6.89503551e-01
1.01303589e+00 3.19979817e-01 4.98167574e-01 -9.32798743e-01
1.22831807e-01 1.26121208e-01 1.27981171e-01 -6.19505167e-01
6.54498637e-01 5.87770045e-01 2.71235794e-01 1.44273534e-01
6.19390965e-01 2.68635243e-01 4.69625086e-01 1.47720084e-01
1.63824058e+00 -1.68428928e-01 -6.72496974e-01 -1.16057813e-01
2.24865317e-01 2.63073385e-01 6.67815268e-01 5.98365664e-01
-1.83030993e-01 3.85061651e-01 6.63488984e-01 -1.82161435e-01
-1.36625838e+00 -1.19923580e+00 -3.57643276e-01 9.52164292e-01
5.95742241e-02 -3.52869600e-01 -9.34554458e-01 -8.95022035e-01
-1.66401900e-02 9.87564385e-01 -3.68413657e-01 -4.61275637e-01
-1.56835616e-01 -3.66291702e-01 1.05636024e+00 4.10742104e-01
7.44949102e-01 -1.25408757e+00 -2.20971301e-01 1.07923262e-01
-2.04647645e-01 -1.25561917e+00 -3.18434745e-01 8.77534412e-03
-1.34023428e+00 -1.17514384e+00 -3.29494596e-01 -6.45867825e-01
6.12341940e-01 -2.15097323e-01 1.39936626e+00 -1.69374138e-01
-5.37302671e-03 1.92012906e-01 -4.41230237e-01 -2.56887883e-01
-6.90283895e-01 3.18349957e-01 1.81840464e-01 -1.78541288e-01
3.01936597e-01 -5.87290406e-01 -3.66504759e-01 7.36353636e-01
-9.57991600e-01 6.66009933e-02 6.61009014e-01 7.05690503e-01
7.83293426e-01 1.45220429e-01 5.38497508e-01 -6.66318357e-01
6.57766759e-01 -7.64596224e-01 -8.91018450e-01 1.42648384e-01
-6.46995544e-01 -1.28223961e-02 5.58105946e-01 -8.85404527e-01
-2.42372409e-01 -3.40678662e-01 -2.62012035e-01 -1.04023015e+00
-1.32528335e-01 5.65731347e-01 -3.61990571e-01 1.91315711e-01
9.26524282e-01 3.02676797e-01 -2.36225754e-01 -3.11327279e-01
-3.88358952e-03 1.52902901e-01 4.63766634e-01 -4.60308403e-01
7.01839328e-01 -2.44245857e-01 -2.76488304e-01 -5.06789982e-01
-4.98011649e-01 5.80734201e-02 -3.55390131e-01 -7.85489023e-01
6.50351822e-01 -8.32839191e-01 -6.30768120e-01 4.64365184e-01
-1.12000060e+00 -6.08276129e-01 -3.45348001e-01 6.87071323e-01
-3.66135955e-01 -3.55866730e-01 -8.18405330e-01 -5.54218769e-01
-2.05009103e-01 -1.36707723e+00 8.33747923e-01 -1.55204579e-01
-6.06752560e-02 -8.30581546e-01 9.23160277e-03 1.74579233e-01
1.14097618e-01 -8.62894431e-02 8.70933771e-01 -9.48887348e-01
-9.66733336e-01 -1.38089091e-01 -1.33382201e-01 4.91963387e-01
-3.59592170e-01 4.00373757e-01 -9.74594116e-01 -2.25089222e-01
-6.14711225e-01 -5.72051466e-01 6.13930941e-01 5.45548610e-02
1.45369172e+00 -4.10007834e-01 -1.53199926e-01 5.46922743e-01
1.17114711e+00 2.23064721e-01 1.02831721e+00 3.91420752e-01
6.68025017e-01 2.51328915e-01 7.10503340e-01 3.81660551e-01
4.10543442e-01 3.93268138e-01 9.46021378e-01 2.43907437e-01
4.07661088e-02 -5.14856339e-01 8.73520195e-01 1.07264733e+00
2.64165670e-01 -6.05184078e-01 -1.02543998e+00 6.51606560e-01
-1.58614457e+00 -8.99245322e-01 -1.55569807e-01 2.02707791e+00
6.34662867e-01 5.57692230e-01 2.46950328e-01 2.84114629e-01
7.97013223e-01 -4.72307295e-01 -1.17080379e+00 -5.52153364e-02
-4.45050389e-01 -1.34372920e-01 5.48171759e-01 4.66382504e-02
-7.89490402e-01 1.30201328e+00 6.81245661e+00 7.84313440e-01
-1.18562055e+00 2.10470363e-01 6.71571314e-01 -3.12721312e-01
-7.52134144e-01 -2.57541299e-01 -1.03474844e+00 6.93586707e-01
1.17541373e+00 1.47467226e-01 6.20972574e-01 9.16692078e-01
1.23053618e-01 3.32239151e-01 -1.52336609e+00 1.07691824e+00
-4.08653885e-01 -1.81237292e+00 1.75722212e-01 1.54802144e-01
8.46056879e-01 8.82998347e-01 1.93991765e-01 6.23613656e-01
6.64954722e-01 -9.45653558e-01 1.30610514e+00 6.27437413e-01
9.15883064e-01 -8.62760186e-01 6.78663552e-01 4.29836005e-01
-6.46114349e-01 -1.54926836e-01 -4.75627959e-01 4.13861990e-01
-3.77972536e-02 9.06239867e-01 -1.18616831e+00 2.97106415e-01
5.96709430e-01 8.21017623e-01 -5.81433356e-01 1.19413614e+00
-1.64768115e-01 9.63682353e-01 -2.23807275e-01 -1.24060296e-01
9.83600393e-02 -3.82170104e-03 9.63934660e-01 1.01867926e+00
6.02130353e-01 -3.86772811e-01 3.61029990e-02 1.15989149e+00
-5.26011527e-01 1.86439678e-01 -6.85320318e-01 -2.77016401e-01
9.72122669e-01 1.04042006e+00 -3.75255227e-01 -4.34437901e-01
-1.13573886e-01 8.12556267e-01 3.63366246e-01 3.54893446e-01
-1.09579813e+00 -3.79091091e-02 9.60329294e-01 1.67701378e-01
8.16750824e-02 7.75682330e-02 -1.89776316e-01 -1.28531563e+00
7.04136789e-02 -9.94785011e-01 2.99311191e-01 -8.42075825e-01
-1.04561234e+00 1.08203781e+00 -4.44039315e-01 -1.52846670e+00
-2.45881438e-01 -6.61170423e-01 -5.19660830e-01 3.85559708e-01
-1.04927850e+00 -5.71862698e-01 -1.31610751e-01 9.20288086e-01
6.55288398e-01 -6.52848780e-01 6.46512747e-01 8.08753133e-01
-1.16733861e+00 9.80312884e-01 -1.47914961e-01 1.65625095e-01
3.01885962e-01 -1.07708120e+00 5.88592768e-01 9.40046191e-01
2.66539186e-01 1.81176245e-01 5.74119687e-01 -9.56240714e-01
-1.68805826e+00 -1.70840478e+00 5.19249618e-01 -3.46929461e-01
7.57743299e-01 -4.37235534e-01 -5.29967785e-01 6.60334527e-01
-1.71483740e-01 -2.50535440e-02 3.22529167e-01 1.97559129e-02
-1.67329654e-01 -1.91703215e-01 -1.28504121e+00 7.63109446e-01
1.51690423e+00 -5.77092826e-01 -6.96134865e-02 3.54417086e-01
1.02880454e+00 -6.88762665e-01 -1.10769176e+00 2.94154614e-01
1.03956729e-01 -7.71704257e-01 5.00486016e-01 -8.92463028e-01
7.01726079e-01 -1.55134022e-01 -2.87212253e-01 -1.48373818e+00
5.96104749e-02 -7.42575228e-01 -3.54173541e-01 1.23527265e+00
1.15805244e+00 -6.15219057e-01 9.21232760e-01 2.88194388e-01
-7.11626172e-01 -7.22546697e-01 -1.14767492e+00 -9.07218993e-01
-3.49378198e-01 -1.11769557e+00 7.16134965e-01 1.05660880e+00
-1.46755293e-01 1.54176503e-01 -1.68493986e-01 1.47909731e-01
3.75888169e-01 -4.57915097e-01 8.44981015e-01 -1.42219150e+00
-6.81500230e-03 -4.82201248e-01 -7.25645125e-01 -1.14327312e+00
3.88325155e-02 -1.29673588e+00 -1.51929423e-01 -1.21179748e+00
-1.80508047e-01 -6.15048230e-01 -3.81797194e-01 4.63689715e-01
2.71330476e-01 3.58825997e-02 8.63931142e-03 1.71700343e-01
-9.06777561e-01 5.05387545e-01 1.46319091e+00 -2.10253671e-01
-1.05311155e-01 -3.32120880e-02 -3.41237783e-01 1.80983543e-01
7.75701761e-01 -4.27354753e-01 -5.73819876e-01 -7.04506338e-01
5.19712687e-01 7.63215050e-02 4.71282095e-01 -1.26098275e+00
2.51358271e-01 6.39158860e-02 3.74557018e-01 -1.00774860e+00
1.40288770e-01 -5.46888113e-01 1.86638802e-01 2.44459912e-01
-3.77251595e-01 6.51172817e-01 5.60043454e-01 2.86746919e-01
-2.31145307e-01 -8.76151770e-02 5.72822452e-01 1.80902764e-01
-7.68607318e-01 7.67286599e-01 -4.79549617e-01 3.68604898e-01
9.20613468e-01 -3.40989798e-01 -4.11369562e-01 -1.23295888e-01
-8.00910711e-01 4.42521930e-01 2.53866076e-01 7.39249349e-01
1.03863466e+00 -1.57238030e+00 -6.40538812e-01 2.38211319e-01
1.88143611e-01 1.44799083e-01 7.50668123e-02 9.61878300e-01
-5.06491125e-01 3.00520331e-01 -1.80182472e-01 -9.67966437e-01
-8.34406376e-01 3.32644880e-01 4.04296011e-01 -9.06629413e-02
-6.67924643e-01 9.17421758e-01 -4.86841857e-01 -7.00412810e-01
4.55583990e-01 -6.40307844e-01 7.49397203e-02 1.39150228e-02
3.00913244e-01 6.53195500e-01 4.58696395e-01 -1.56028986e-01
-2.17426792e-01 -1.52670994e-01 1.38706490e-02 -3.39627296e-01
1.38587177e+00 8.77831429e-02 1.63689211e-01 3.89143646e-01
1.00435293e+00 -2.64116466e-01 -1.35863495e+00 -1.31654099e-01
2.90085286e-01 -1.95499554e-01 1.76413089e-01 -8.50039899e-01
-1.39250624e+00 6.00139737e-01 5.75602055e-01 1.89652771e-01
9.64601874e-01 8.35077539e-02 7.76264250e-01 4.53079998e-01
3.43881160e-01 -1.45139480e+00 4.00061429e-01 8.69178891e-01
6.48320019e-01 -1.10012293e+00 -7.91550100e-01 -5.17450124e-02
-5.02736032e-01 1.11127949e+00 1.11785781e+00 1.05062596e-01
4.26343292e-01 2.84865469e-01 -1.41277209e-01 -2.55127668e-01
-1.48349869e+00 -3.02466433e-02 2.76956171e-01 3.94767851e-01
-5.94604388e-02 -2.11627018e-02 5.57548940e-01 1.81764394e-01
-5.84515035e-01 -3.73936594e-02 2.60899186e-01 2.93800741e-01
-2.94725001e-01 -7.38893569e-01 -7.26932511e-02 6.60313189e-01
-1.26433596e-01 -2.45129794e-01 -4.59270030e-01 8.01687181e-01
2.66475111e-01 8.65574539e-01 5.27279139e-01 -1.24937630e+00
5.89898765e-01 -1.28182605e-01 2.81858891e-01 -2.93646634e-01
-3.88679832e-01 -2.85058290e-01 4.29631382e-01 -8.10941696e-01
1.50939777e-01 -8.50763202e-01 -1.07646346e+00 -7.63467848e-01
-2.90811926e-01 -1.50560260e-01 7.80219793e-01 9.40826416e-01
4.86634851e-01 7.43086696e-01 7.01798737e-01 -2.16496974e-01
-2.45171219e-01 -1.27466059e+00 -7.39843845e-02 2.24206567e-01
1.91972241e-01 -6.28359079e-01 -8.04328173e-02 -2.83328205e-01]
|
[8.899528503417969, 3.4024550914764404]
|
cb44c868-fa3f-4e46-b51e-50028231abc4
|
syntf-synthetic-and-differentially-private
|
1805.00904
| null |
http://arxiv.org/abs/1805.00904v1
|
http://arxiv.org/pdf/1805.00904v1.pdf
|
SynTF: Synthetic and Differentially Private Term Frequency Vectors for Privacy-Preserving Text Mining
|
Text mining and information retrieval techniques have been developed to
assist us with analyzing, organizing and retrieving documents with the help of
computers. In many cases, it is desirable that the authors of such documents
remain anonymous: Search logs can reveal sensitive details about a user,
critical articles or messages about a company or government might have severe
or fatal consequences for a critic, and negative feedback in customer surveys
might negatively impact business relations if they are identified. Simply
removing personally identifying information from a document is, however,
insufficient to protect the writer's identity: Given some reference texts of
suspect authors, so-called authorship attribution methods can reidentfy the
author from the text itself.
One of the most prominent models to represent documents in many common text
mining and information retrieval tasks is the vector space model where each
document is represented as a vector, typically containing its term frequencies
or related quantities. We therefore propose an automated text anonymization
approach that produces synthetic term frequency vectors for the input documents
that can be used in lieu of the original vectors. We evaluate our method on an
exemplary text classification task and demonstrate that it only has a low
impact on its accuracy. In contrast, we show that our method strongly affects
authorship attribution techniques to the level that they become infeasible with
a much stronger decline in accuracy. Other than previous authorship obfuscation
methods, our approach is the first that fulfills differential privacy and hence
comes with a provable plausible deniability guarantee.
|
['Florian Kerschbaum', 'Benjamin Weggenmann']
|
2018-05-02
| null | null | null | null |
['text-anonymization']
|
['natural-language-processing']
|
[ 4.59256351e-01 2.08512932e-01 -2.96543330e-01 -8.65538046e-02
-5.22751212e-01 -1.08985877e+00 8.50965619e-01 7.02938616e-01
-5.80964744e-01 7.06081510e-01 7.03718662e-02 -5.66530406e-01
-2.50235140e-01 -5.59715033e-01 -5.06559730e-01 -5.69504440e-01
3.32722962e-01 5.84304571e-01 -1.59804255e-01 -9.65163019e-03
6.49694145e-01 8.25598538e-01 -1.18184435e+00 1.81917667e-01
7.09209263e-01 7.74796963e-01 -5.98739684e-01 3.77594203e-01
-2.80490100e-01 4.58866239e-01 -8.75210702e-01 -1.07957077e+00
5.64355910e-01 -1.84032023e-01 -9.07156110e-01 -1.78981900e-01
1.20646313e-01 -1.41448155e-01 -2.75220484e-01 1.30103624e+00
1.30263656e-01 -3.40977937e-01 8.27238917e-01 -1.50357115e+00
-7.03658998e-01 7.33721375e-01 -8.09232295e-01 -1.89061407e-02
4.14875358e-01 -2.41284981e-01 8.82452369e-01 -3.43174428e-01
6.97424591e-01 1.03215241e+00 6.71808720e-01 2.69062966e-01
-1.49704087e+00 -7.48582244e-01 -2.39111722e-01 -2.25029916e-01
-1.37313819e+00 -3.97480905e-01 7.96706855e-01 -4.96819675e-01
1.18554331e-01 8.15915108e-01 2.39074633e-01 1.29674697e+00
1.83492079e-01 5.78970492e-01 9.18478847e-01 -4.49391991e-01
2.00894699e-01 7.31732965e-01 4.50547040e-01 3.21065962e-01
1.02881086e+00 -2.30456352e-01 -4.24911350e-01 -1.12364566e+00
2.80971779e-03 1.40841126e-01 -2.94848979e-01 -5.92554688e-01
-1.03670633e+00 1.10261095e+00 -1.22258060e-01 3.71105969e-01
-2.00956732e-01 7.23555759e-02 7.04906344e-01 5.79091609e-01
3.76205623e-01 6.62237346e-01 -2.28062287e-01 1.90143749e-01
-8.41270864e-01 5.16027808e-01 1.03874791e+00 7.41991997e-01
4.99391317e-01 -4.63425457e-01 -1.45626351e-01 3.45644414e-01
1.22511730e-01 4.00098830e-01 8.24832797e-01 -6.15995705e-01
3.52299184e-01 6.06489718e-01 5.20091712e-01 -1.49822652e+00
1.37686029e-01 -2.28050888e-01 -6.45852327e-01 6.82614222e-02
5.68742394e-01 7.99098015e-02 -3.05837899e-01 1.56778347e+00
8.05473030e-02 -5.27453125e-01 1.30752325e-02 6.90313280e-01
1.63387001e-01 3.85109723e-01 -6.74421489e-02 -2.16588214e-01
1.55754042e+00 -1.18493937e-01 -9.06161487e-01 7.05402344e-03
7.28992045e-01 -7.57683277e-01 8.50341439e-01 3.83938491e-01
-7.38463223e-01 2.33512431e-01 -8.55830431e-01 -2.62991320e-02
-6.42805278e-01 -2.21777588e-01 6.68297410e-01 1.38309944e+00
-5.67962229e-01 6.50647461e-01 -4.20940697e-01 -3.31332296e-01
6.85006142e-01 3.81306976e-01 -6.30500138e-01 2.32928768e-01
-1.26517582e+00 5.41088343e-01 1.68368399e-01 -3.18447977e-01
-2.12920774e-02 -6.71779513e-01 -3.96131516e-01 2.26348877e-01
2.62817025e-01 -5.11247337e-01 8.74252796e-01 -9.82368827e-01
-7.19064772e-01 1.08878732e+00 -1.63770229e-01 -7.66334474e-01
8.83210719e-01 5.41882105e-02 -4.93965060e-01 2.21542977e-02
1.72840387e-01 7.71749541e-02 1.26524806e+00 -1.30112517e+00
-4.65984643e-01 -7.54796565e-01 -4.22299296e-01 -3.05144131e-01
-8.14081192e-01 2.15537980e-01 1.14489555e-01 -9.65941131e-01
4.44927663e-02 -8.50324810e-01 -1.24678258e-02 6.32840022e-02
-8.11178386e-01 -5.01899309e-02 1.07844460e+00 -7.25617290e-01
1.21530199e+00 -2.11351061e+00 -2.49401554e-01 7.07223475e-01
4.92566913e-01 3.11593473e-01 3.42136383e-01 6.34427547e-01
-7.07072616e-02 8.82179081e-01 -2.65616983e-01 -3.65829825e-01
1.55702651e-01 -9.10431221e-02 -8.71149421e-01 1.01574612e+00
-2.70449549e-01 7.47440577e-01 -5.61430633e-01 -1.53772935e-01
-2.18141735e-01 3.03461224e-01 -2.82428026e-01 -1.59707919e-01
1.19859792e-01 -1.73965506e-02 -5.92102289e-01 3.40503037e-01
6.91918850e-01 -1.76990345e-01 2.66617775e-01 1.12323605e-01
2.19545975e-01 1.02457814e-01 -1.17340481e+00 1.08799231e+00
-1.22751277e-02 8.90726745e-01 2.39312369e-02 -8.48090529e-01
8.67955685e-01 2.61171669e-01 3.58098835e-01 -3.37328732e-01
3.00134659e-01 2.15813264e-01 -2.59008467e-01 -3.05214524e-01
8.26025724e-01 1.07078217e-01 -2.37371832e-01 1.18705833e+00
-5.90855658e-01 6.05695724e-01 -1.15496896e-01 4.55040634e-01
1.06626582e+00 -5.68027258e-01 2.40374357e-01 -2.69091219e-01
4.11732405e-01 -2.10162342e-01 3.39334220e-01 9.87758994e-01
-5.32505289e-02 4.45105374e-01 8.72808933e-01 -2.60972887e-01
-1.10157096e+00 -6.26247108e-01 -2.96298265e-01 6.30183041e-01
-5.91181666e-02 -3.67158145e-01 -8.11326623e-01 -9.58324015e-01
6.31681323e-01 9.39351797e-01 -7.38104343e-01 -4.26907629e-01
-1.31254613e-01 -5.59944808e-01 9.87861156e-01 -7.92855471e-02
-8.43654722e-02 -5.01023233e-01 -6.61382675e-01 -8.26923102e-02
-3.92507128e-02 -6.99891686e-01 -6.70796990e-01 -2.14431081e-02
-5.55327296e-01 -1.16768837e+00 -7.04788625e-01 -9.51314941e-02
8.90680134e-01 2.38591671e-01 5.27682364e-01 9.99730676e-02
-2.68762648e-01 2.72162795e-01 -2.95732319e-01 -6.00590408e-01
-8.52917492e-01 9.46653336e-02 1.97699144e-01 5.56442916e-01
6.94968402e-01 -5.04494071e-01 -2.96120882e-01 1.48336634e-01
-1.22299433e+00 -6.39349401e-01 1.33767560e-01 6.65394545e-01
-2.23526116e-02 1.64238721e-01 4.86385494e-01 -1.66426277e+00
1.07787430e+00 -5.08300424e-01 -4.53414589e-01 3.08074385e-01
-1.10726953e+00 2.49619365e-01 6.74837530e-01 -5.88411629e-01
-6.71481192e-01 -1.29094392e-01 4.98935550e-01 -4.65153009e-01
-1.90731674e-01 2.55724758e-01 -2.95856327e-01 -2.04519451e-01
7.02762902e-01 1.84431866e-01 2.60220677e-01 -7.10005343e-01
4.34425533e-01 9.91250396e-01 3.54431987e-01 -3.37605417e-01
1.12271190e+00 7.44990408e-01 2.80952360e-02 -7.06522822e-01
-2.78824210e-01 -5.20397305e-01 -3.96809787e-01 2.13989630e-01
1.96479872e-01 -3.77044290e-01 -1.01747561e+00 2.37667128e-01
-1.14961946e+00 6.80139065e-01 -3.67054790e-01 1.10142983e-01
-1.75536945e-01 7.59918272e-01 -2.34831899e-01 -8.98796499e-01
-1.76202998e-01 -8.90427232e-01 6.28309429e-01 -2.93448269e-01
-8.78442824e-01 -8.91643107e-01 -1.11554623e-01 5.20735621e-01
3.47833931e-01 2.98435599e-01 1.32439363e+00 -1.34787261e+00
-1.47645384e-01 -1.07705820e+00 -6.30205050e-02 8.34618732e-02
1.83814317e-01 -5.82318045e-02 -1.05056238e+00 -4.08419490e-01
1.71422660e-01 7.35183284e-02 6.78231239e-01 -1.78977609e-01
1.30483902e+00 -1.03083885e+00 -5.05448759e-01 4.53147113e-01
1.00157726e+00 1.20037787e-01 5.85853398e-01 5.26405334e-01
6.79536462e-01 1.00939023e+00 4.66993824e-02 6.74597442e-01
-2.44527712e-01 6.71217442e-01 2.03424186e-01 1.97633952e-01
5.85851490e-01 -4.71624434e-01 -1.12251993e-02 3.00482586e-02
4.27640229e-01 -6.81779385e-01 -4.76016074e-01 3.78022611e-01
-1.71037674e+00 -1.06243300e+00 -2.75482386e-01 2.67698932e+00
6.60650611e-01 -8.89897943e-02 2.49000758e-01 5.16969442e-01
7.66967833e-01 8.59853402e-02 -4.79415625e-01 -7.40428746e-01
-1.56975150e-01 -1.76267967e-01 1.20375144e+00 2.25122094e-01
-8.05608273e-01 4.02000070e-01 5.43924570e+00 6.64370537e-01
-8.97617698e-01 -6.70654029e-02 7.18665600e-01 -1.14417650e-01
-7.26206124e-01 1.71701953e-01 -4.62039173e-01 7.40603089e-01
1.00218260e+00 -7.81527042e-01 3.27940285e-01 7.94586778e-01
1.20746113e-01 1.32668614e-01 -1.24384761e+00 9.91925538e-01
1.52829036e-01 -1.38258326e+00 3.37268829e-01 7.08427429e-01
3.53320777e-01 -7.92618155e-01 3.79931688e-01 -3.58104318e-01
2.96534926e-01 -1.10752046e+00 6.83014154e-01 4.24253494e-01
6.43130183e-01 -1.14401698e+00 6.35451019e-01 3.26068312e-01
-2.77129143e-01 -1.73424169e-01 -3.80377531e-01 2.25243166e-01
-2.24767029e-02 7.14418709e-01 -8.58183563e-01 4.49170262e-01
3.46828729e-01 2.69626915e-01 -5.76576889e-01 5.78763127e-01
8.06608647e-02 6.95750415e-01 -2.85351068e-01 -1.50813118e-01
3.95746641e-02 -2.99805880e-01 8.80276620e-01 9.55423474e-01
2.89522558e-01 -6.25865459e-02 -5.08118331e-01 9.70304608e-01
-6.91376865e-01 2.76025772e-01 -1.02118003e+00 -5.27611852e-01
7.04500914e-01 9.67010200e-01 -7.43574798e-01 -4.18046378e-02
-1.88767642e-01 1.17496598e+00 7.57766441e-02 2.28089526e-01
-2.33785495e-01 -5.32031298e-01 7.58746624e-01 4.89721179e-01
1.27069131e-01 2.30892435e-01 -5.04656792e-01 -1.00605428e+00
2.67600894e-01 -1.00350463e+00 4.28035319e-01 -3.81900162e-01
-1.32122731e+00 2.54257172e-01 -3.66936713e-01 -1.00452912e+00
-3.03313464e-01 -2.22069174e-01 -2.26739660e-01 1.11942697e+00
-9.62754905e-01 -7.00899124e-01 3.56572956e-01 4.77283806e-01
-2.06145391e-01 -4.72386092e-01 8.25553656e-01 3.39234695e-02
-3.83453637e-01 9.77038503e-01 6.75256670e-01 3.09507728e-01
7.15962887e-01 -1.09858859e+00 4.48723644e-01 8.26233208e-01
4.38441932e-01 1.15601206e+00 8.64576399e-01 -8.42504203e-01
-1.64355600e+00 -9.47121739e-01 1.49229383e+00 -8.54095221e-01
7.18660831e-01 -6.44680619e-01 -1.17616320e+00 7.24150717e-01
-8.17049816e-02 -3.50905180e-01 7.71395683e-01 4.90087792e-02
-5.87203264e-01 4.65494841e-02 -1.44289184e+00 6.01696551e-01
5.28782785e-01 -7.42371023e-01 -5.74607313e-01 5.36889911e-01
5.09725094e-01 2.77791232e-01 -4.99306262e-01 -4.37828839e-01
5.76493502e-01 -7.98990786e-01 8.59681726e-01 -1.10406554e+00
1.68501556e-01 -1.92201570e-01 -4.35271449e-02 -9.15848851e-01
-1.32877737e-01 -9.99361873e-01 -8.80290568e-02 1.57878900e+00
2.12497458e-01 -9.28831100e-01 9.32338417e-01 1.08519518e+00
8.54683280e-01 -7.34782815e-02 -9.19235468e-01 -8.66519988e-01
3.05864979e-02 -2.29000688e-01 8.32000077e-01 1.35226440e+00
3.95139456e-02 1.57832772e-01 -5.63389599e-01 1.72587693e-01
1.00710738e+00 8.70862082e-02 8.53549838e-01 -1.59932375e+00
7.20051080e-02 -4.08524096e-01 -5.35523117e-01 -4.24645305e-01
5.27471781e-01 -1.04808307e+00 -6.83109224e-01 -7.90722549e-01
1.91329435e-01 -5.78973591e-01 -7.53722712e-03 1.59524232e-01
1.57321617e-01 -1.34057440e-02 -1.49752777e-02 5.89575887e-01
1.02366492e-01 2.18914300e-01 3.91937941e-01 -2.82010376e-01
-6.32506981e-02 4.20305848e-01 -1.20779574e+00 4.91025120e-01
5.66553831e-01 -9.85978901e-01 -2.60266632e-01 5.38373664e-02
4.42438245e-01 -1.44991055e-01 5.97593248e-01 -3.60438466e-01
3.99189085e-01 7.36058056e-02 1.97098941e-01 -2.42075965e-01
-9.36203524e-02 -1.15603983e+00 4.16223168e-01 5.17679214e-01
-8.16886425e-01 1.80284455e-01 -2.16485634e-01 9.83088017e-01
6.07266128e-02 -6.03136122e-01 5.09928823e-01 1.34598780e-02
8.81384835e-02 1.26941487e-01 -6.64891899e-01 -1.32238194e-01
9.34533596e-01 -3.47650826e-01 -4.01886195e-01 -5.18755615e-01
-1.56130686e-01 -1.27430454e-01 9.17410851e-01 3.43539625e-01
2.21160948e-01 -9.56620574e-01 -6.76272631e-01 2.12805152e-01
2.49819309e-01 -6.90531433e-01 -2.60720938e-01 4.30049568e-01
-1.78767070e-01 5.69239378e-01 -3.26558538e-02 2.97456142e-03
-1.42133415e+00 9.71074760e-01 -8.39961097e-02 8.46822187e-02
-6.59488142e-01 4.39169914e-01 7.51065239e-02 6.51561131e-04
2.30753765e-01 9.43717808e-02 -3.56089398e-02 3.57277244e-01
7.27367520e-01 4.76380974e-01 3.31542820e-01 -7.94727921e-01
-3.61897826e-01 -9.93277207e-02 -5.78862786e-01 -1.86702326e-01
1.06415105e+00 -2.28962168e-01 -2.12474003e-01 4.06159788e-01
1.35576844e+00 5.25230050e-01 -5.02629220e-01 -2.00609371e-01
3.00965160e-01 -9.16268349e-01 -1.44955134e-02 -4.54364628e-01
-9.64523375e-01 6.83489799e-01 4.41062212e-01 7.31635630e-01
6.37582719e-01 -8.77710432e-02 7.04626024e-01 5.42561054e-01
1.92998201e-01 -8.50869656e-01 -5.88186145e-01 -2.09887013e-01
5.73101342e-01 -1.08879673e+00 3.12404096e-01 -2.79291302e-01
-5.98615646e-01 1.01860499e+00 -2.98424900e-01 2.98036814e-01
3.85743797e-01 1.62274495e-01 6.19224347e-02 -1.28552765e-01
-3.42021674e-01 6.75248742e-01 9.41318274e-02 5.92660904e-01
1.10753350e-01 1.56355230e-03 -6.36374533e-01 7.81662643e-01
-2.57725239e-01 -3.34981740e-01 8.84240210e-01 9.73595262e-01
-2.48989742e-02 -1.36846280e+00 -6.54151201e-01 7.36671984e-01
-9.49688554e-01 -1.07363366e-01 -9.23924685e-01 7.13728726e-01
-4.59436268e-01 7.62951910e-01 9.01930854e-02 -2.59667516e-01
1.40064806e-01 2.45274499e-01 -2.12631822e-01 -1.97700605e-01
-8.04333270e-01 -4.20822024e-01 8.48437920e-02 -3.19784790e-01
-1.02837265e-01 -1.13290453e+00 -7.78963566e-01 -9.34495628e-01
-3.27958405e-01 5.61936796e-01 7.90492713e-01 7.77282774e-01
5.26505232e-01 -1.57433704e-01 8.83443952e-01 -1.58095658e-01
-8.33912015e-01 -4.20334935e-01 -9.94585574e-01 7.71452308e-01
2.71332145e-01 -3.96088630e-01 -7.28158772e-01 2.12985307e-01]
|
[6.14246940612793, 7.005415439605713]
|
ccc57730-e0d8-4653-b67d-4fd718324cd7
|
palm-2-technical-report
| null | null |
https://ai.google/static/documents/palm2techreport.pdf
|
https://ai.google/static/documents/palm2techreport.pdf
|
PaLM 2 Technical Report
|
We introduce PaLM 2, a new state-of-the-art language model that has better multilingual and reasoning capabilities
and is more compute-efficient than its predecessor PaLM (Chowdhery et al., 2022). PaLM 2 is a Transformer-based
model trained using a mixture of objectives similar to UL2 (Tay et al., 2023). Through extensive evaluations on English
and multilingual language, and reasoning tasks, we demonstrate that PaLM 2 has significantly improved quality on
downstream tasks across different model sizes, while simultaneously exhibiting faster and more efficient inference
compared to PaLM. This improved efficiency enables broader deployment while also allowing the model to respond
faster, for a more natural pace of interaction. PaLM 2 demonstrates robust reasoning capabilities exemplified by large
improvements over PaLM on BIG-Bench and other reasoning tasks. PaLM 2 exhibits stable performance on a suite of
responsible AI evaluations, and enables inference-time control over toxicity without additional overhead or impact on
other capabilities. Overall, PaLM 2 achieves state-of-the-art performance across a diverse set of tasks and capabilities
|
['Google']
|
2023-05-10
| null | null | null | null |
['math-word-problem-solving', 'multi-task-language-understanding', 'movie-recommendation', 'sports-understanding', 'word-sense-disambiguation', 'multiple-choice-qa', 'sarcasm-detection', 'toxic-comment-classification', 'cross-lingual-question-answering', 'text-summarization', 'coreference-resolution', 'cross-lingual-transfer', 'disambiguation-q', 'ruin-names', 'snarks', 'formal-fallacies-syllogisms-negation', 'hyperbaton', 'penguins-in-a-table', 'temporal-sequences', 'arithmetic-reasoning', 'logical-reasoning', 'math-word-problem-solving', 'common-sense-reasoning', 'reasoning-about-colored-objects', 'causal-judgment', 'date-understanding', 'math-word-problem-solving']
|
['knowledge-base', 'methodology', 'miscellaneous', 'miscellaneous', 'natural-language-processing', 'natural-language-processing', 'natural-language-processing', 'natural-language-processing', 'natural-language-processing', 'natural-language-processing', 'natural-language-processing', 'natural-language-processing', 'natural-language-processing', 'natural-language-processing', 'natural-language-processing', 'natural-language-processing', 'natural-language-processing', 'reasoning', 'reasoning', 'reasoning', 'reasoning', 'reasoning', 'reasoning', 'reasoning', 'reasoning', 'reasoning', 'time-series']
|
[-3.99293095e-01 1.29679710e-01 -2.45092645e-01 -2.42460504e-01
-1.23294854e+00 -7.32487440e-01 9.05532837e-01 -1.07566059e-01
-3.39789003e-01 6.55567050e-01 3.06147695e-01 -6.90829098e-01
-1.50685206e-01 -6.94299698e-01 -6.41305447e-01 8.82339943e-03
-1.22755013e-01 1.14687145e+00 6.11885041e-02 -5.65449119e-01
-2.14612968e-02 3.72147173e-01 -1.34266067e+00 4.07533050e-01
6.88685894e-01 6.60249233e-01 -4.20445055e-02 7.37330854e-01
1.41907483e-01 1.47484219e+00 -3.63662183e-01 -6.93998814e-01
1.38854101e-01 2.10274443e-01 -1.08205235e+00 -6.61006093e-01
7.11338103e-01 -4.83000785e-01 -3.34232777e-01 6.21341944e-01
7.41425097e-01 -1.38400970e-02 3.51597339e-01 -1.52508378e+00
-6.91043079e-01 1.27242959e+00 -2.56017804e-01 1.49720892e-01
6.17208838e-01 5.24459958e-01 1.14641023e+00 -8.52303088e-01
4.96666163e-01 1.61204672e+00 1.03059936e+00 3.76285493e-01
-1.40565288e+00 -8.74325037e-01 3.33366245e-01 5.27288066e-03
-1.53224313e+00 -8.69972110e-01 9.01060775e-02 -1.93846226e-02
1.57815242e+00 1.17955580e-01 4.89538014e-01 1.17001569e+00
3.47182512e-01 1.37729049e+00 1.00475538e+00 -2.79820055e-01
3.48885693e-02 -1.26166955e-01 5.69227524e-02 1.15745676e+00
-2.14051992e-01 -2.69360900e-01 -9.77998316e-01 -1.76499292e-01
4.84770745e-01 -4.77119178e-01 6.45122081e-02 -1.71525180e-01
-1.48581183e+00 2.71326989e-01 9.60243866e-02 9.03654099e-02
-3.20008516e-01 6.79834783e-01 3.57309967e-01 3.29244405e-01
2.68050283e-01 7.16179013e-01 -6.35272086e-01 -6.46848798e-01
-1.10437155e+00 8.08866620e-01 1.08139753e+00 1.20765531e+00
2.00902909e-01 1.63898647e-01 -6.10389590e-01 6.37072682e-01
3.21077555e-01 9.69003022e-01 2.64065683e-01 -1.40818787e+00
5.86460233e-01 3.73456597e-01 3.77312712e-02 -5.83136022e-01
-4.96939689e-01 -6.05419695e-01 -4.84891176e-01 1.77659780e-01
5.26214361e-01 -5.94552513e-03 -7.15830147e-01 2.12414575e+00
-1.40637785e-01 -1.92183673e-01 3.53950709e-01 6.37130380e-01
5.89824140e-01 5.53878427e-01 9.76557210e-02 3.31289560e-01
1.43033087e+00 -1.28760588e+00 -5.48263252e-01 -6.58645689e-01
5.95177829e-01 -6.58536851e-01 1.28955948e+00 8.90035987e-01
-1.80705392e+00 -2.62092471e-01 -9.47951555e-01 -6.41598582e-01
-4.75736111e-01 1.86148612e-03 1.26476848e+00 4.23653275e-01
-1.44717968e+00 3.96740109e-01 -1.08882546e+00 -3.66286308e-01
1.86099306e-01 2.40637079e-01 -1.76753044e-01 -2.02269688e-01
-1.39266384e+00 1.41042686e+00 2.84607381e-01 7.39750192e-02
-1.09296048e+00 -1.03444397e+00 -8.64454031e-01 7.19904229e-02
4.43940014e-01 -8.66759896e-01 1.91201305e+00 -2.39999458e-01
-1.65944970e+00 7.95710385e-01 -2.41713300e-01 -7.40807950e-01
8.26653957e-01 -5.31872749e-01 -5.51771224e-01 -9.35137570e-02
3.05395126e-01 9.42960203e-01 4.95146215e-01 -6.75823748e-01
-5.95551729e-01 -1.78058803e-01 5.43229461e-01 3.90924037e-01
-1.87934205e-01 1.03753433e-02 -8.63034070e-01 -4.76061702e-01
-3.40473950e-01 -8.65322649e-01 2.47227877e-01 1.18679434e-01
-1.78720579e-01 -4.61352497e-01 5.92905819e-01 -8.10089111e-01
1.22372532e+00 -1.62412059e+00 4.56566215e-01 -1.26086876e-01
4.86252815e-01 2.01478034e-01 -3.02350014e-01 6.62696958e-01
2.77279019e-01 6.23955280e-02 -2.44734570e-01 -4.57305104e-01
5.01531363e-01 1.56823605e-01 -2.51641721e-01 1.59746841e-01
1.77813590e-01 1.24352205e+00 -1.21940887e+00 -7.49209166e-01
1.00521214e-01 3.05904329e-01 -7.24943578e-01 6.26287237e-02
-6.08007193e-01 5.54390289e-02 -3.56982559e-01 9.40697134e-01
2.08828479e-01 -3.99954796e-01 3.31957638e-01 -9.81444418e-02
-9.04913992e-02 3.99351805e-01 -9.54284668e-01 2.29348350e+00
-1.06835222e+00 8.21096063e-01 2.85849065e-01 -4.27383751e-01
4.14558351e-01 3.12714338e-01 5.47052436e-02 -9.23459649e-01
-8.68531689e-02 2.02220052e-01 8.76089931e-03 -3.53376120e-01
6.02707982e-01 3.84840518e-01 -3.72090667e-01 7.37400353e-01
-1.30782612e-02 -5.78918040e-01 6.70132101e-01 6.44589603e-01
1.18388748e+00 5.74511766e-01 7.09913075e-02 -6.18971944e-01
3.69797498e-01 1.15870476e-01 1.21944606e-01 9.69226480e-01
3.83232869e-02 -1.08130932e-01 2.57402539e-01 -2.22164601e-01
-8.62545848e-01 -1.40511715e+00 -1.23528309e-01 1.47385383e+00
-1.51353568e-01 -9.44622159e-01 -5.64874351e-01 -3.85378480e-01
2.62376219e-01 1.22515404e+00 -2.83239782e-01 -1.66212082e-01
-5.05553186e-01 -4.08228248e-01 1.46802592e+00 8.69623959e-01
9.68081951e-01 -9.43740547e-01 -6.06779754e-01 1.80802315e-01
-6.40767932e-01 -1.17331553e+00 -1.85036048e-01 -4.90108766e-02
-5.73116362e-01 -8.50489557e-01 -3.06427211e-01 -4.66633052e-01
1.17028624e-01 -2.61935234e-01 1.73835862e+00 -1.96019366e-01
-1.08099155e-01 4.90899056e-01 9.54024121e-02 -3.37907016e-01
-5.76144457e-01 3.40887845e-01 2.41674885e-01 -8.22002411e-01
1.38738170e-01 -4.01187479e-01 3.49687077e-02 -3.56443375e-02
-3.15876067e-01 4.05283719e-01 7.28719175e-01 8.04317355e-01
1.49722129e-01 8.56962353e-02 2.51383275e-01 -6.40150368e-01
9.73267078e-01 -3.11469525e-01 -6.39073074e-01 5.09613633e-01
-8.82701218e-01 5.07068574e-01 9.14291739e-01 -1.20391846e-01
-1.17341781e+00 -6.30101085e-01 1.46616520e-02 -2.05626622e-01
3.07008177e-01 5.84668517e-01 1.30907029e-01 2.11640567e-01
5.47449589e-01 2.92661905e-01 -7.76937604e-02 -4.00388658e-01
6.20847821e-01 6.04894578e-01 8.82381201e-01 -1.39481020e+00
5.60982227e-01 1.04129858e-01 -1.49974421e-01 -4.34347689e-01
-8.54177773e-01 9.80948359e-02 -2.53990859e-01 3.40748020e-02
5.87253988e-01 -1.28278267e+00 -1.18136442e+00 6.12643957e-01
-9.51389313e-01 -9.99811828e-01 2.47407243e-01 3.73521820e-02
-5.44783652e-01 -8.76094475e-02 -1.04118669e+00 -7.96267033e-01
-7.20944881e-01 -1.36930096e+00 1.41085386e+00 -1.56025812e-01
-5.64041674e-01 -1.14252293e+00 -2.33256072e-01 5.93136966e-01
6.43982410e-01 -2.07240969e-01 1.06182313e+00 -2.88774908e-01
-5.32122135e-01 -1.08865596e-01 -3.65444303e-01 -5.07559255e-02
-2.89480567e-01 6.76760972e-02 -8.21561158e-01 -4.00817275e-01
-7.81663001e-01 -9.15130913e-01 4.60440695e-01 -7.11395070e-02
1.07898569e+00 -9.99964587e-03 -4.32858586e-01 6.61589026e-01
8.85169864e-01 -2.05545411e-01 1.82428077e-01 3.72151911e-01
4.25404936e-01 8.03576410e-02 5.36405504e-01 9.56234112e-02
9.07549024e-01 7.06207514e-01 3.84928696e-02 1.02879770e-01
-1.19804420e-01 -5.12091160e-01 5.81143498e-01 7.50740349e-01
-8.49259496e-02 -3.86225015e-01 -1.43876171e+00 4.33625340e-01
-2.17735195e+00 -9.32912886e-01 1.75679773e-01 1.76097000e+00
1.15731263e+00 4.86609012e-01 -1.22819534e-02 -1.72818050e-01
2.61332393e-02 -6.62075952e-02 -7.89761841e-01 -4.92930233e-01
-5.97558506e-02 4.90853667e-01 4.53478813e-01 6.33121073e-01
-7.27758408e-01 1.43954301e+00 7.86023808e+00 8.32219541e-01
-8.06639254e-01 4.38935794e-02 2.29825243e-01 -3.71753365e-01
-3.32333595e-01 -1.41017780e-01 -8.09597909e-01 1.11848928e-01
1.10508513e+00 -4.68608499e-01 1.11800897e+00 5.85218966e-01
-7.85883591e-02 -1.47201896e-01 -1.52182090e+00 8.65047157e-01
2.28899702e-01 -1.38020658e+00 -2.79663899e-03 -3.38957340e-01
3.49580348e-01 3.77307475e-01 3.75030302e-02 9.87601876e-01
1.11452949e+00 -1.15333652e+00 1.29110491e+00 4.82255816e-01
9.66337860e-01 -6.49087012e-01 5.22859812e-01 4.74398404e-01
-1.22368860e+00 -1.23764284e-01 2.91916132e-01 -3.64279717e-01
1.80972800e-01 9.19051766e-02 -5.14232874e-01 5.13210654e-01
7.81472385e-01 7.43226230e-01 -6.98725224e-01 8.29767808e-02
-4.74706411e-01 4.68116134e-01 -4.87235337e-01 7.99122378e-02
3.86835784e-01 2.58421898e-01 3.43826264e-01 1.27414179e+00
-2.25940526e-01 -6.42527789e-02 3.87285739e-01 1.06672251e+00
-1.96365297e-01 -3.90500098e-01 -4.80479002e-01 -3.13170969e-01
9.14883494e-01 1.21651101e+00 -2.03923225e-01 -6.02946818e-01
-1.94536313e-01 8.14728856e-01 7.92366326e-01 2.28032246e-01
-1.20404601e+00 -1.68412507e-01 6.43442273e-01 -2.39831552e-01
-2.28458196e-01 -4.07401145e-01 3.33169885e-02 -1.21905446e+00
-1.64418951e-01 -1.37460387e+00 4.45843130e-01 -1.30159545e+00
-1.02283812e+00 4.91504043e-01 3.46397758e-01 -5.25819123e-01
-5.31804085e-01 -6.29241467e-01 -1.03628777e-01 1.04272377e+00
-1.34238732e+00 -1.62180841e+00 -1.32327959e-01 6.11446440e-01
5.83429694e-01 -2.88094372e-01 1.15493155e+00 3.37839574e-01
-6.64589465e-01 8.89704943e-01 -2.23467797e-01 1.65054157e-01
8.65260959e-01 -1.43812573e+00 8.16045284e-01 8.76919627e-01
-8.04818347e-02 1.01651585e+00 6.01944506e-01 -5.29688597e-01
-2.12836885e+00 -8.71096849e-01 8.00947547e-01 -9.16200757e-01
1.03215468e+00 -4.62814748e-01 -1.53868616e-01 1.43358922e+00
4.27836359e-01 -3.93811792e-01 1.64408818e-01 6.30976081e-01
-6.06510878e-01 -9.50604156e-02 -9.93642509e-01 1.17947102e+00
1.37607253e+00 -8.15208912e-01 -8.03939939e-01 4.99874502e-01
5.81636906e-01 -8.97956848e-01 -1.16998279e+00 2.30731875e-01
9.74952281e-01 -8.63223135e-01 1.16600978e+00 -6.60609663e-01
7.03698397e-01 -3.09593845e-02 -2.43949965e-01 -1.29490149e+00
-6.12375498e-01 -7.21918762e-01 -4.23099577e-01 1.06804895e+00
6.61418617e-01 -7.56093681e-01 2.31976777e-01 6.59576893e-01
-1.67953670e-01 -6.61766768e-01 -3.26188356e-01 -8.43058705e-01
2.38158837e-01 -7.89606035e-01 7.00647414e-01 7.67769992e-01
2.64429271e-01 5.62558115e-01 -1.15197450e-01 2.45972797e-01
7.57487118e-01 1.18380420e-01 8.71239245e-01 -8.92693400e-01
-5.57913482e-01 -6.57033026e-01 2.27052599e-01 -1.00899434e+00
5.95780551e-01 -1.42007208e+00 -4.26893532e-02 -1.63389885e+00
-4.71863244e-03 -5.09180605e-01 -1.81896493e-01 1.11176658e+00
3.14262807e-02 1.93110406e-01 3.37363362e-01 2.02187464e-01
-7.71208286e-01 2.56505072e-01 1.00645924e+00 -4.09812421e-01
2.27583468e-01 -4.11592484e-01 -7.40804315e-01 6.74132049e-01
5.17034769e-01 1.48425847e-01 -6.10300422e-01 -1.27236462e+00
5.47154009e-01 5.77490963e-02 2.34505042e-01 -1.13420200e+00
3.50232810e-01 9.70582291e-02 1.71886340e-01 -5.08465409e-01
5.70449114e-01 -4.68559951e-01 1.74826700e-02 4.35055941e-01
-6.14602149e-01 6.87474430e-01 7.52152920e-01 -7.05487980e-03
1.01332344e-01 3.16765428e-01 4.01434183e-01 -3.20953965e-01
-9.62760627e-01 -7.76602775e-02 -5.35662234e-01 4.49149877e-01
6.69804871e-01 2.87712246e-01 -5.53724289e-01 -3.98932219e-01
-4.91091400e-01 8.94828439e-01 2.45934412e-01 6.93189681e-01
7.72371963e-02 -1.14321530e+00 -8.05245876e-01 1.22125529e-01
2.11860195e-01 -3.00675314e-02 -1.85836717e-01 8.81804168e-01
-7.49675512e-01 7.97552884e-01 -1.87146470e-01 -4.92031634e-01
-9.57086086e-01 4.31021243e-01 4.79702055e-01 -6.57582581e-01
-6.83922112e-01 9.13267910e-01 -5.13740003e-01 -9.03245568e-01
2.66339540e-01 -3.93509418e-01 4.44690913e-01 -3.75410080e-01
5.53700209e-01 5.77469766e-01 9.40228775e-02 -1.18088368e-02
-5.93347311e-01 1.52666911e-01 -1.58314615e-01 -4.21618640e-01
1.02943265e+00 1.77450493e-01 -2.65452355e-01 5.69343090e-01
4.54470187e-01 2.40974464e-02 -8.90898645e-01 -1.39195517e-01
-1.69715047e-01 5.82997873e-02 3.66680712e-01 -1.56538820e+00
-8.10665607e-01 6.39827013e-01 -9.61919595e-03 -2.01902255e-01
7.89328516e-01 -3.74893188e-01 8.46105218e-01 8.61882508e-01
8.34467709e-01 -1.07543314e+00 -3.29506755e-01 6.93039000e-01
8.10723960e-01 -1.02099276e+00 1.35113075e-01 6.85303006e-03
-5.13351798e-01 1.00447464e+00 8.49863112e-01 3.29301149e-01
2.09769592e-01 9.74801719e-01 9.83123016e-03 -1.80795878e-01
-1.32054460e+00 1.41165322e-02 -5.24042286e-02 4.29145098e-01
7.23714948e-01 1.50940284e-01 1.16928197e-01 2.82965839e-01
-5.95241547e-01 3.97631615e-01 6.59826770e-02 1.04118645e+00
2.69251496e-01 -9.47347939e-01 -2.59097517e-01 2.73894519e-01
-1.89606383e-01 -6.43341064e-01 -2.71528482e-01 1.08656287e+00
-1.96175292e-01 1.01703048e+00 1.17252968e-01 -9.19138044e-02
2.67001837e-01 2.69507468e-01 8.77879202e-01 -3.19913894e-01
-7.51713216e-01 -3.16328228e-01 7.18854368e-01 -9.78442013e-01
-3.70267890e-02 -4.39032316e-01 -1.40423024e+00 -9.35719252e-01
1.06102481e-01 1.03962936e-01 4.15740758e-01 1.07582068e+00
5.85379303e-01 7.04156637e-01 -2.73102611e-01 -7.05467224e-01
-8.44476342e-01 -9.03928459e-01 -1.52734116e-01 2.00956352e-02
-3.24660577e-02 -5.55425465e-01 1.65235698e-01 -8.05944875e-02]
|
[9.73069953918457, 7.479613304138184]
|
d8d2abf1-49dc-4c14-89d2-5bfc83762e73
|
semeval-2021-task-10-source-free-domain
| null | null |
https://aclanthology.org/2021.semeval-1.42
|
https://aclanthology.org/2021.semeval-1.42.pdf
|
SemEval-2021 Task 10: Source-Free Domain Adaptation for Semantic Processing
|
This paper presents the Source-Free Domain Adaptation shared task held within SemEval-2021. The aim of the task was to explore adaptation of machine-learning models in the face of data sharing constraints. Specifically, we consider the scenario where annotations exist for a domain but cannot be shared. Instead, participants are provided with models trained on that (source) data. Participants also receive some labeled data from a new (development) domain on which to explore domain adaptation algorithms. Participants are then tested on data representing a new (target) domain. We explored this scenario with two different semantic tasks: negation detection (a text classification task) and time expression recognition (a sequence tagging task).
|
['Steven Bethard', 'Timothy Miller', '{\\"O}zlem Uzuner', 'Yiyun Zhao', 'Xin Su', 'Egoitz Laparra']
|
2021-08-01
| null | null | null |
semeval-2021
|
['source-free-domain-adaptation', 'negation-detection']
|
['computer-vision', 'natural-language-processing']
|
[ 6.21549904e-01 4.89451647e-01 -2.52736032e-01 -1.16807795e+00
-3.98417383e-01 -6.53410792e-01 6.90812171e-01 5.03519237e-01
-9.50879574e-01 1.12981391e+00 -1.31051034e-01 -5.13224788e-02
2.01443017e-01 -4.14584965e-01 -3.98998111e-01 -6.92723021e-02
2.39687189e-01 7.65015900e-01 1.50030896e-01 -2.62626439e-01
-6.30255118e-02 1.24642178e-01 -1.41233504e+00 7.59484112e-01
9.86574888e-01 6.05919898e-01 9.92517173e-02 4.81132179e-01
-1.75312176e-01 1.70248255e-01 -7.42480159e-01 -6.83392525e-01
1.27279073e-01 -4.06433761e-01 -1.10691416e+00 -2.83983707e-01
4.91418511e-01 -3.55204344e-02 3.07528794e-01 8.95424724e-01
3.98472667e-01 7.45438516e-01 5.29028177e-01 -1.65639210e+00
-6.53244138e-01 5.09074569e-01 1.32687762e-01 9.72924456e-02
6.61466599e-01 -1.54406399e-01 5.06076038e-01 -7.76188970e-01
1.09352350e+00 9.69875336e-01 3.47362280e-01 1.33911943e+00
-1.37915409e+00 -8.48604739e-01 4.61765021e-01 3.09319526e-01
-1.02206028e+00 -5.05785763e-01 4.11805689e-01 -3.35876137e-01
1.22131920e+00 -6.45621493e-02 2.55138248e-01 1.81170416e+00
-1.14779390e-01 4.80679870e-01 1.09993684e+00 -7.80902088e-01
5.74255526e-01 7.77656257e-01 4.33942527e-01 1.60971448e-01
1.55971423e-01 1.49338648e-01 -1.16585863e+00 -1.79219872e-01
-2.29857732e-02 -4.62557644e-01 7.40226433e-02 -5.25020301e-01
-8.55887830e-01 5.78807950e-01 -7.41303861e-02 6.92827165e-01
-7.22295269e-02 -4.60286289e-01 8.09556365e-01 8.79855990e-01
6.63379312e-01 4.67909992e-01 -1.33759248e+00 -2.49458179e-01
-5.15460312e-01 5.80518425e-01 1.15421724e+00 9.30885077e-01
7.27554977e-01 -2.47833684e-01 -1.31118044e-01 9.09315050e-01
2.98814982e-01 2.95126557e-01 9.00882840e-01 -6.45193458e-01
6.19248390e-01 5.98142445e-01 2.69603252e-01 -2.59051889e-01
-5.21025419e-01 5.56894392e-02 -4.13964719e-01 2.16407388e-01
6.43469155e-01 -7.08568037e-01 -7.40317166e-01 2.25178981e+00
3.97918671e-01 1.22022741e-01 4.40150887e-01 6.54453158e-01
5.83329082e-01 1.11625738e-01 7.12623060e-01 -2.44900271e-01
1.23203194e+00 -6.73266411e-01 -9.98611450e-01 -6.29856348e-01
1.25077164e+00 -4.49826926e-01 7.51275003e-01 5.37146986e-01
-9.02578712e-01 -6.82821393e-01 -1.09107971e+00 -3.03078201e-02
-8.80496323e-01 -2.90842205e-01 4.09436494e-01 9.19433892e-01
-8.02481949e-01 3.18414330e-01 -4.81674135e-01 -9.22211587e-01
9.14212391e-02 3.12042594e-01 -6.49671495e-01 -3.02950531e-01
-1.65822232e+00 1.20914018e+00 6.63663685e-01 -3.56070876e-01
-5.86726367e-01 -7.67366707e-01 -1.03102899e+00 -1.98920324e-01
1.11369386e-01 -5.80068409e-01 1.70395982e+00 -1.92474675e+00
-1.64718056e+00 1.45648313e+00 -5.73659763e-02 -5.38203537e-01
3.62903744e-01 -4.62085605e-01 -9.40995991e-01 -3.35607529e-01
1.82346478e-01 5.93247116e-01 6.35582864e-01 -6.83912456e-01
-5.80108702e-01 -5.56399763e-01 -1.19846426e-01 3.17194223e-01
-5.32882810e-01 2.74459124e-01 3.29613239e-01 -5.44233024e-01
-4.55678016e-01 -8.78045738e-01 1.04172044e-01 -1.63487047e-01
6.23847358e-02 -3.70244712e-01 6.17817402e-01 -7.90595353e-01
1.20642900e+00 -2.15629315e+00 1.43354878e-01 2.18346968e-01
-3.48898172e-01 4.28828090e-01 -4.15259242e-01 2.59833455e-01
-5.17743587e-01 2.80644238e-01 -2.78144300e-01 -4.78075147e-01
-2.72267517e-02 4.77147028e-02 -5.69101563e-03 8.11363477e-03
3.98346856e-02 4.65248823e-01 -7.39731252e-01 -2.66840518e-01
-2.65348375e-01 -8.98567066e-02 -5.12950599e-01 2.27748066e-01
-6.84099734e-01 5.77542424e-01 -5.35255730e-01 2.87136406e-01
5.36517501e-01 3.69824380e-01 5.56013286e-01 2.36233652e-01
3.20313163e-02 7.86002651e-02 -1.18134129e+00 2.24723077e+00
-4.64464724e-01 4.80552465e-01 4.93318699e-02 -1.02759767e+00
1.04895043e+00 6.13101065e-01 2.97754884e-01 -9.47429776e-01
-7.72795156e-02 2.73227215e-01 1.17343545e-01 -6.79974616e-01
2.80507267e-01 -2.53743023e-01 -3.84496123e-01 5.66810191e-01
4.04754162e-01 2.40092203e-01 6.52078390e-02 -5.78644276e-02
1.22456336e+00 5.27072787e-01 5.88846564e-01 -1.62626788e-01
5.55305719e-01 1.46463245e-01 6.73724473e-01 4.82901096e-01
-6.21696234e-01 -5.67144388e-03 1.43962726e-01 -5.28124630e-01
-1.03945005e+00 -7.98449516e-01 -9.35677513e-02 1.74809325e+00
-2.84867227e-01 -4.29727197e-01 -6.90363586e-01 -1.05935276e+00
8.68364796e-03 1.21690094e+00 -6.98282957e-01 -4.92580146e-01
-4.44279343e-01 -8.46622214e-02 5.00724733e-01 4.56301957e-01
5.84143341e-01 -1.16169202e+00 -6.66767418e-01 2.73854017e-01
-3.78414959e-01 -1.04856205e+00 -3.79910499e-01 3.97932053e-01
-7.20261812e-01 -9.15041268e-01 -3.98477226e-01 -8.69834483e-01
4.25901294e-01 -4.32430595e-01 1.07706511e+00 -2.69240767e-01
-7.25417510e-02 7.89735377e-01 -3.86644393e-01 -9.02405202e-01
-4.20193434e-01 1.74101919e-01 1.85164437e-01 -3.25815752e-02
1.17030072e+00 -2.09123537e-01 -5.58131747e-03 2.93811202e-01
-8.46374631e-01 -6.53601661e-02 8.88653323e-02 6.42177820e-01
1.39785469e-01 -1.91582873e-01 9.28574920e-01 -1.23380852e+00
7.51068890e-01 -7.25568831e-01 -3.12075555e-01 5.09811401e-01
-5.93066752e-01 -1.28009677e-01 2.24968195e-01 -7.30695844e-01
-1.77259314e+00 2.37869322e-01 2.38819480e-01 4.83504357e-03
-7.75141358e-01 7.72949994e-01 -4.74721313e-01 4.14232790e-01
1.15468836e+00 -4.08243895e-01 -5.68264797e-02 -4.64066803e-01
2.76057303e-01 6.29861534e-01 3.46860617e-01 -8.81801367e-01
4.35616374e-01 -2.93347776e-01 -3.48967046e-01 -5.90537071e-01
-8.19236994e-01 -3.84673506e-01 -1.08492005e+00 1.02626972e-01
1.10722005e+00 -1.00234354e+00 -2.82570153e-01 5.33580005e-01
-1.32481229e+00 -8.32025170e-01 -4.99505520e-01 4.69821811e-01
-5.87034404e-01 -2.34272271e-01 2.07194928e-02 -6.46193087e-01
7.02016130e-02 -2.30894223e-01 5.27266920e-01 2.32103378e-01
-9.63613212e-01 -1.47370648e+00 4.12292361e-01 3.98457885e-01
3.92453998e-01 3.31252933e-01 1.13056517e+00 -1.88275754e+00
1.54774889e-01 -1.53507888e-01 3.28513265e-01 4.01780248e-01
-8.04544017e-02 -3.22375119e-01 -1.15300941e+00 -2.63981342e-01
5.61645329e-02 -7.15158761e-01 2.31197491e-01 -2.05043405e-01
9.54474747e-01 -8.23630020e-02 -5.08590698e-01 5.20001762e-02
9.30168390e-01 3.11443597e-01 1.98585600e-01 4.42175984e-01
8.71649012e-02 1.04628694e+00 8.48381162e-01 4.31816548e-01
3.61044973e-01 9.48613763e-01 -2.21202418e-01 1.96615249e-01
1.82251468e-01 -1.34339258e-01 6.08758092e-01 -2.43959222e-02
4.06513244e-01 -5.63191772e-01 -1.21597207e+00 7.28263736e-01
-1.96749270e+00 -7.34547615e-01 1.16079450e-01 2.23221302e+00
1.13422704e+00 6.50175475e-03 8.82648304e-02 -3.86268139e-01
7.86210179e-01 -4.81559545e-01 -7.70517826e-01 -9.69092667e-01
-5.20295091e-02 7.02276587e-01 3.83484475e-02 4.47942168e-01
-1.02391291e+00 7.54759133e-01 6.10495710e+00 2.66266793e-01
-6.35211766e-01 3.78400356e-01 4.26638573e-01 -1.68421820e-01
-1.29580364e-01 4.51920405e-02 -9.09491897e-01 3.50525439e-01
1.61711931e+00 -6.48904979e-01 8.35094452e-02 6.76055253e-01
-2.57046223e-01 -1.50661573e-01 -1.86038816e+00 5.69239020e-01
2.18079403e-01 -5.47102273e-01 -2.34506279e-01 -1.89224243e-01
5.63172460e-01 -8.41428936e-02 -8.52292180e-02 7.86381423e-01
4.39560086e-01 -9.83049214e-01 2.67402977e-01 7.52428412e-01
7.59978712e-01 -3.83567929e-01 7.29180396e-01 6.79940224e-01
-6.39563560e-01 -5.58520071e-02 2.05582723e-01 -2.21658632e-01
-7.58890510e-02 2.72954136e-01 -9.61185157e-01 4.86508667e-01
6.95220411e-01 5.70836544e-01 -5.91373563e-01 7.38933921e-01
-2.66399086e-01 4.61293042e-01 -2.11748675e-01 2.01420352e-01
-3.45198363e-01 3.23121995e-02 4.57892269e-01 1.27786815e+00
4.06493992e-01 1.03724204e-01 2.78173983e-01 6.15676463e-01
-3.78288984e-01 2.41244674e-01 -9.33823526e-01 -2.52625030e-02
6.22986078e-01 9.69030559e-01 -2.79145509e-01 -3.21392536e-01
-7.32861817e-01 1.40556049e+00 3.48957241e-01 6.25708461e-01
-4.38694388e-01 -2.07882062e-01 7.92379618e-01 1.99603304e-01
-1.19298130e-01 7.37870065e-03 -1.94437772e-01 -1.08422601e+00
-2.33776607e-02 -9.60727990e-01 7.83428729e-01 -7.92708695e-01
-1.61584234e+00 7.97237903e-02 4.13269490e-01 -9.10338700e-01
-6.09403193e-01 -6.11926317e-01 -3.50531459e-01 1.26677382e+00
-1.18778062e+00 -1.05214715e+00 -2.93879002e-01 8.62049997e-01
6.20930374e-01 -6.50314450e-01 1.31762421e+00 3.46025318e-01
-3.12400341e-01 6.98466599e-01 -8.77666250e-02 -2.96974834e-02
1.71975780e+00 -1.12571955e+00 2.92694688e-01 5.14419079e-01
-1.97357893e-01 6.70466065e-01 6.79664850e-01 -1.09056401e+00
-6.11415207e-01 -1.19802272e+00 1.73852170e+00 -7.71667659e-01
3.60116303e-01 -4.27414089e-01 -1.31227660e+00 1.12888837e+00
5.74108422e-01 -1.69211607e-02 1.36434257e+00 3.95939052e-01
-3.41196984e-01 2.30389833e-01 -1.59893012e+00 2.30125830e-01
1.06867659e+00 -6.56960666e-01 -8.75229418e-01 3.31710726e-01
6.48118019e-01 -6.05495632e-01 -7.53136873e-01 2.47929730e-02
2.61786550e-01 -4.68666404e-01 4.25732762e-01 -1.25308502e+00
1.23447910e-01 2.10024610e-01 -1.58267930e-01 -1.63377559e+00
-6.91074599e-03 -4.82899040e-01 1.74625814e-02 1.52934790e+00
5.50128281e-01 -6.98264539e-01 5.59495091e-01 1.41013527e+00
3.98849603e-03 1.61479622e-01 -1.10636723e+00 -8.45742166e-01
4.87176180e-01 -4.16470379e-01 5.74141026e-01 1.38834858e+00
3.47165346e-01 6.36606872e-01 8.22358802e-02 -7.75048360e-02
1.71388820e-01 -5.77556372e-01 5.86045086e-01 -1.70090020e+00
1.15733936e-01 4.40384597e-02 -1.05188802e-01 -4.25679892e-01
6.96092546e-01 -1.12126160e+00 -1.38330027e-01 -1.10054946e+00
-7.41193667e-02 -3.67119104e-01 -3.97267312e-01 9.05605555e-01
5.15206531e-02 -4.20882523e-01 1.59938991e-01 -1.33379862e-01
-5.86619794e-01 3.60880792e-01 4.74710077e-01 7.78776854e-02
-3.42785805e-01 6.26641512e-02 -6.44216061e-01 7.54344165e-01
1.02794230e+00 -6.60609066e-01 -6.35976613e-01 -4.74448711e-01
4.67997640e-01 -2.83661354e-02 4.15963590e-01 -1.06688333e+00
3.43864828e-01 -2.68601656e-01 5.14916420e-01 3.40336212e-03
1.57938719e-01 -1.05627525e+00 6.39716983e-02 1.92355648e-01
-8.38852584e-01 -2.30545513e-02 7.83858478e-01 4.22911137e-01
-2.59639829e-01 -6.47341251e-01 6.79794431e-01 4.47233506e-02
-1.10679936e+00 -4.31064144e-02 -4.55711126e-01 1.84078649e-01
1.37849045e+00 -3.23233753e-01 -2.77474284e-01 1.80859808e-02
-1.42941630e+00 3.93822074e-01 4.28211778e-01 7.95932651e-01
4.76126522e-01 -1.37055779e+00 -7.41281211e-01 3.87002885e-01
6.12589359e-01 -2.06469893e-01 1.80882424e-01 2.19299808e-01
2.36713439e-01 3.22007269e-01 -7.18416035e-01 -1.17982984e-01
-1.51029336e+00 5.93153000e-01 3.63152176e-01 -3.81711185e-01
-8.12775716e-02 7.86371529e-01 3.15697603e-02 -8.64245474e-01
2.52212673e-01 -1.67097598e-02 -2.76555210e-01 2.52125591e-01
4.38763708e-01 1.46591425e-01 3.88783157e-01 -2.09792092e-01
-5.01943290e-01 -6.32208260e-03 -3.36986631e-01 -5.29723167e-01
1.04126596e+00 -1.44483700e-01 2.75205165e-01 8.28046799e-01
1.01565230e+00 -3.35251600e-01 -8.31041157e-01 -7.37764597e-01
4.20787930e-01 -7.32886791e-02 -4.25168306e-01 -1.65651453e+00
-2.84392804e-01 6.75082326e-01 1.12019515e+00 -1.61261469e-01
9.65785205e-01 -3.28154415e-01 3.65397364e-01 6.13188088e-01
2.61565953e-01 -1.72280705e+00 1.39414683e-01 7.90465474e-01
8.76469314e-01 -1.38304353e+00 -2.09331781e-01 -6.36316538e-02
-9.68079388e-01 1.04391563e+00 1.27081871e+00 4.75539178e-01
5.61313272e-01 5.06743602e-02 1.09211756e-02 -2.05957890e-03
-1.19672489e+00 1.65395886e-01 1.84981450e-01 1.22344148e+00
7.66009808e-01 -3.28892544e-02 -2.16078773e-01 1.16362345e+00
1.00857057e-01 4.55036968e-01 3.63997370e-01 1.18857563e+00
-1.97616816e-01 -1.75933254e+00 -6.94633648e-02 1.39541626e-01
-2.64781177e-01 4.51388359e-02 -8.00669372e-01 7.87617564e-01
6.13625050e-01 8.68078291e-01 1.65801480e-01 8.59774798e-02
7.00769246e-01 1.05731416e+00 4.05156940e-01 -1.28538477e+00
-1.02787137e+00 -6.58465028e-01 4.95435238e-01 -4.66343462e-01
-5.68018973e-01 -1.01789689e+00 -1.26116252e+00 1.42337337e-01
1.56862393e-01 1.89519122e-01 8.00972641e-01 9.32799518e-01
4.08723980e-01 4.80561733e-01 1.19836535e-02 -1.25487307e-02
-3.01398844e-01 -1.01282263e+00 -3.32315236e-01 8.00005019e-01
9.01970193e-02 -4.36106205e-01 5.91704920e-02 6.62073433e-01]
|
[10.626606941223145, 7.990034103393555]
|
3c7d14fe-b977-4d3c-945c-7cec56d2b441
|
offline-reinforcement-learning-via-high
|
2209.14548
| null |
https://arxiv.org/abs/2209.14548v2
|
https://arxiv.org/pdf/2209.14548v2.pdf
|
Offline Reinforcement Learning via High-Fidelity Generative Behavior Modeling
|
In offline reinforcement learning, weighted regression is a common method to ensure the learned policy stays close to the behavior policy and to prevent selecting out-of-sample actions. In this work, we show that due to the limited distributional expressivity of policy models, previous methods might still select unseen actions during training, which deviates from their initial motivation. To address this problem, we adopt a generative approach by decoupling the learned policy into two parts: an expressive generative behavior model and an action evaluation model. The key insight is that such decoupling avoids learning an explicitly parameterized policy model with a closed-form expression. Directly learning the behavior policy allows us to leverage existing advances in generative modeling, such as diffusion-based methods, to model diverse behaviors. As for action evaluation, we combine our method with an in-sample planning technique to further avoid selecting out-of-sample actions and increase computational efficiency. Experimental results on D4RL datasets show that our proposed method achieves competitive or superior performance compared with state-of-the-art offline RL methods, especially in complex tasks such as AntMaze. We also empirically demonstrate that our method can successfully learn from a heterogeneous dataset containing multiple distinctive but similarly successful strategies, whereas previous unimodal policies fail.
|
['Jun Zhu', 'Hang Su', 'Chengyang Ying', 'Cheng Lu', 'Huayu Chen']
|
2022-09-29
| null | null | null | null |
['d4rl']
|
['robots']
|
[-4.51015495e-03 1.52330801e-01 -6.82556629e-01 -1.23971090e-01
-8.36509883e-01 -7.36970663e-01 7.48617172e-01 -2.78254241e-01
-5.05975604e-01 9.60445344e-01 3.32452357e-01 -3.83062750e-01
-7.11060762e-02 -6.42404318e-01 -7.45998204e-01 -1.01016057e+00
-1.83410626e-02 7.18904912e-01 1.76207557e-01 -2.40167364e-01
2.22666144e-01 4.83060658e-01 -1.27433467e+00 6.51347041e-02
9.90186155e-01 6.31401777e-01 2.29363650e-01 4.48945373e-01
-1.38524398e-01 1.25185478e+00 -6.08214200e-01 1.44412994e-01
3.12651038e-01 -6.51113987e-01 -5.06936848e-01 2.73650706e-01
1.65819861e-02 -7.87471056e-01 -4.64942247e-01 9.33692694e-01
4.99046981e-01 4.71760124e-01 7.31997669e-01 -1.11140895e+00
-6.63545072e-01 8.36014748e-01 -4.85356867e-01 -1.50213018e-01
5.04382439e-02 4.60558236e-01 8.65205228e-01 -3.94343019e-01
6.18987441e-01 1.30367458e+00 1.61683589e-01 9.48887110e-01
-1.34297967e+00 -6.33874118e-01 7.09371626e-01 -8.31291825e-02
-9.84464109e-01 -3.59795958e-01 7.05616713e-01 -1.92958400e-01
8.02635252e-01 -8.73131305e-02 7.36198187e-01 1.57268083e+00
4.66754325e-02 1.45507669e+00 1.33140230e+00 -2.11177468e-01
6.17536724e-01 -1.31198227e-01 -3.88736755e-01 6.93432629e-01
-4.79115620e-02 3.81263733e-01 -3.30116391e-01 -3.85465026e-01
9.45529759e-01 1.31235123e-01 -2.62680858e-01 -6.60195470e-01
-9.86830950e-01 9.88061905e-01 2.20511585e-01 6.89448491e-02
-3.97016734e-01 5.46892822e-01 1.73015773e-01 2.48831585e-01
3.80276263e-01 4.48050886e-01 -3.54169935e-01 -5.67419231e-01
-7.57945418e-01 5.96160650e-01 7.66632080e-01 1.00606132e+00
6.81024075e-01 4.18477744e-01 -5.21567643e-01 8.69299948e-01
1.92770094e-01 5.41436791e-01 5.72652638e-01 -1.25621951e+00
4.89565492e-01 3.29777956e-01 5.08245647e-01 -3.16950500e-01
-1.39191598e-01 -4.03322756e-01 -4.46948290e-01 2.40651071e-01
4.96977717e-01 -3.61159801e-01 -1.02773559e+00 2.02304769e+00
3.57849181e-01 2.27468044e-01 1.52061805e-01 8.54199350e-01
-2.51650326e-02 5.25207639e-01 1.19764596e-01 -2.52327234e-01
6.60578132e-01 -1.12312770e+00 -4.90379065e-01 -3.53180051e-01
7.65476704e-01 -3.53518277e-01 1.25020242e+00 4.71931309e-01
-1.02523923e+00 -2.27022976e-01 -8.02094579e-01 4.16296721e-01
3.09942365e-02 2.68120736e-01 7.41000831e-01 3.31364006e-01
-7.88093209e-01 7.41078973e-01 -1.19718504e+00 -1.74596339e-01
5.63791275e-01 3.00379634e-01 5.21374345e-02 -9.46385860e-02
-8.72349143e-01 7.16509163e-01 4.02734429e-01 -1.58968762e-01
-1.59387982e+00 -5.45276940e-01 -7.00722516e-01 -6.56814054e-02
1.06863737e+00 -4.82327789e-01 1.72458375e+00 -1.05142415e+00
-2.18160796e+00 7.10629821e-02 9.38307960e-03 -5.44332266e-01
6.73570454e-01 -3.91720474e-01 -1.30221918e-01 6.00012131e-02
-4.12281118e-02 5.59828639e-01 1.00686479e+00 -1.42658627e+00
-7.16843307e-01 -1.98697075e-01 3.32463205e-01 3.54236245e-01
-2.66671509e-01 -3.63722205e-01 -5.41071773e-01 -6.36672974e-01
-2.73414910e-01 -1.22089934e+00 -5.55300653e-01 -4.17790651e-01
-3.88743579e-01 -3.15265000e-01 6.87303901e-01 -9.82706845e-02
1.28738713e+00 -1.85944521e+00 3.83409530e-01 2.28978917e-01
6.09529577e-02 2.66544610e-01 -1.99309200e-01 6.68020248e-01
3.99014086e-01 -7.71479905e-02 -2.05223531e-01 -3.80390555e-01
2.57849425e-01 6.61584675e-01 -7.41233647e-01 5.48380554e-01
3.63876345e-04 8.78145397e-01 -1.28470778e+00 -2.04677865e-01
1.27358034e-01 1.92924678e-01 -7.66389906e-01 3.45467836e-01
-8.08669209e-01 7.43217766e-01 -9.59833860e-01 5.92721164e-01
2.78042972e-01 -2.06382841e-01 6.10559940e-01 3.94526511e-01
6.66298494e-02 2.79807359e-01 -1.03252149e+00 1.72988594e+00
-5.42210221e-01 8.00606981e-02 -7.88686052e-02 -1.05560958e+00
6.75140202e-01 3.68744545e-02 6.55940771e-01 -6.86407030e-01
5.00569828e-02 1.16373710e-01 1.29673854e-01 -2.91900754e-01
3.63136828e-01 3.48323211e-02 -7.99319744e-02 7.87584245e-01
5.58966920e-02 -4.65453006e-02 2.33165845e-01 1.42920613e-01
1.09493697e+00 8.83389294e-01 2.65716493e-01 -1.58280522e-01
1.19100787e-01 -3.97531055e-02 6.41056001e-01 1.16923213e+00
-1.86662242e-01 1.97056025e-01 7.24811494e-01 -1.59544662e-01
-8.22019756e-01 -8.83416951e-01 3.61772090e-01 1.29506898e+00
9.74851996e-02 -4.66664344e-01 -6.91774666e-01 -1.19898105e+00
1.39928803e-01 1.00115633e+00 -7.16412365e-01 -2.02945977e-01
-8.44387829e-01 -5.60677707e-01 4.18960869e-01 7.20107675e-01
2.61724979e-01 -1.15906906e+00 -6.65160656e-01 4.61485296e-01
1.90680847e-01 -9.44131255e-01 -5.10794401e-01 3.25106621e-01
-8.32483888e-01 -9.85895395e-01 -7.63139665e-01 -2.81465441e-01
7.90716290e-01 1.61435544e-01 9.48761463e-01 -1.14241250e-01
2.06958100e-01 6.80068016e-01 -4.30309981e-01 -2.73952097e-01
-5.99663258e-01 1.11787036e-01 1.08726047e-01 -7.75351375e-02
1.74551696e-01 -5.33272088e-01 -7.19412982e-01 3.56867522e-01
-9.25995111e-01 -5.79703115e-02 6.77811384e-01 1.01971769e+00
7.21009851e-01 -1.41164705e-01 6.27408564e-01 -9.15293455e-01
8.64487410e-01 -5.46827734e-01 -8.31863999e-01 2.69152522e-01
-7.75092065e-01 5.41484296e-01 1.00233805e+00 -9.49499249e-01
-1.06037891e+00 -6.63154051e-02 1.06089905e-01 -7.31308699e-01
-2.50957068e-02 2.06833571e-01 -7.12434202e-03 2.44698554e-01
5.16085327e-01 4.89078403e-01 6.54968843e-02 -4.14560616e-01
6.10053539e-01 4.16165441e-01 1.50830626e-01 -1.20317101e+00
5.90425730e-01 5.58834732e-01 -2.03806579e-01 -5.29122531e-01
-7.78633177e-01 -1.61110938e-01 -2.97223646e-02 -1.91489413e-01
5.29583454e-01 -7.23134637e-01 -8.51896882e-01 1.89853191e-01
-6.70379877e-01 -1.08114898e+00 -4.71332729e-01 5.94763577e-01
-1.12441671e+00 6.13440722e-02 -5.30707479e-01 -1.05869973e+00
6.25185594e-02 -1.38878798e+00 1.09776640e+00 2.04514578e-01
-6.01513684e-02 -9.41641629e-01 2.79442728e-01 -3.62238549e-02
3.58566493e-01 -8.60091969e-02 8.09556067e-01 -7.90934920e-01
-7.42306590e-01 1.89859480e-01 3.68389547e-01 2.43975103e-01
1.80914447e-01 -1.45662352e-01 -5.62146604e-01 -5.20675898e-01
-6.46612346e-02 -8.04965675e-01 9.60763872e-01 3.16692591e-01
1.42929459e+00 -3.85016471e-01 -2.90091544e-01 5.78058660e-01
1.23024929e+00 4.07176405e-01 5.19917965e-01 2.99978495e-01
5.66311359e-01 2.93190539e-01 8.67537618e-01 7.16032565e-01
2.39686593e-01 7.66673565e-01 4.59860593e-01 3.10933173e-01
5.55498786e-02 -8.99382830e-01 8.19254160e-01 3.62436980e-01
-6.04267269e-02 -4.08489108e-01 -6.49291396e-01 4.02129531e-01
-2.36253762e+00 -1.09738111e+00 6.93535984e-01 2.26847148e+00
9.70750570e-01 1.24888942e-01 4.18791026e-01 -3.94602954e-01
2.24056065e-01 2.98505992e-01 -9.66110945e-01 -2.72164136e-01
2.30846003e-01 1.96284205e-01 7.67066181e-01 5.08489966e-01
-9.30646837e-01 1.36624277e+00 6.16681623e+00 1.26918089e+00
-1.12011230e+00 2.81402245e-02 4.88806963e-01 -3.99552733e-01
-5.03353477e-01 7.07367435e-02 -1.06049562e+00 4.78683978e-01
6.65350556e-01 1.53309673e-01 8.51186395e-01 1.02189243e+00
3.89575869e-01 -1.08199626e-01 -1.14389062e+00 6.22177362e-01
-1.12982817e-01 -1.25182831e+00 -3.25686634e-02 2.86213070e-01
8.60249698e-01 -2.28084950e-03 1.65597290e-01 8.07200909e-01
1.09359658e+00 -9.66758013e-01 7.56871462e-01 4.22480047e-01
4.12289917e-01 -8.56099725e-01 6.06585220e-02 7.18691885e-01
-8.85626197e-01 -3.41607362e-01 -4.34666336e-01 1.73611239e-01
7.43546784e-02 -3.66654582e-02 -7.90318191e-01 3.99733365e-01
3.78167361e-01 6.87218189e-01 -1.16000034e-01 6.34017348e-01
-6.51641369e-01 7.95984626e-01 -2.99045354e-01 -2.74057120e-01
7.63461649e-01 -4.09841359e-01 5.07506013e-01 7.51473010e-01
3.44864637e-01 -7.09264949e-02 6.61686301e-01 9.66555893e-01
8.66874307e-02 -4.73992452e-02 -8.03864360e-01 -3.73432070e-01
3.18033069e-01 8.87162149e-01 -5.59199333e-01 -2.19043359e-01
-3.17316920e-01 8.66832495e-01 7.59273648e-01 7.80287862e-01
-1.11223340e+00 1.90302953e-01 6.61174059e-01 -6.80452287e-02
6.95348620e-01 -5.37716031e-01 3.40604484e-01 -1.25193143e+00
-1.90462798e-01 -1.23981857e+00 1.85410619e-01 -3.01224858e-01
-1.12026978e+00 3.26284617e-01 2.21736386e-01 -1.29337442e+00
-7.54119039e-01 -3.56407851e-01 -4.63987052e-01 5.05116880e-01
-1.48395610e+00 -1.04199219e+00 3.19993049e-01 6.68449938e-01
7.85641491e-01 -3.56382191e-01 5.46533942e-01 -1.16248779e-01
-6.31180286e-01 4.41604525e-01 3.50609422e-01 -1.44921884e-01
6.94516420e-01 -1.33251047e+00 -7.22310767e-02 6.58536613e-01
2.90801644e-01 6.21890962e-01 5.16957700e-01 -7.42920280e-01
-1.64456522e+00 -1.03307474e+00 -3.42046767e-02 -1.03471570e-01
7.56535411e-01 -3.30643237e-01 -6.48732901e-01 8.22149873e-01
1.75889298e-01 -7.82194883e-02 5.00646412e-01 6.61852583e-02
-4.31806482e-02 3.82284597e-02 -7.50770271e-01 1.16146290e+00
1.15475702e+00 -1.45552382e-01 -3.75882328e-01 3.21000159e-01
5.90942383e-01 -5.19877315e-01 -5.01132011e-01 2.88352638e-01
4.94914323e-01 -9.16464865e-01 7.64248312e-01 -9.52189147e-01
3.08454037e-01 -1.50179833e-01 -1.43635869e-01 -1.62562430e+00
-9.85985696e-02 -1.07827246e+00 -6.59615993e-01 9.43291128e-01
2.33640283e-01 -6.45578146e-01 8.39771569e-01 5.08659601e-01
-1.23565145e-01 -1.21194994e+00 -5.86932600e-01 -1.02537489e+00
2.01267913e-01 -4.50396508e-01 5.77870727e-01 5.07669568e-01
-1.77507460e-01 1.08209036e-01 -7.83021986e-01 -1.13520436e-01
5.68385601e-01 3.99648428e-01 9.99189258e-01 -5.17059386e-01
-9.01315928e-01 -5.36314309e-01 3.03489715e-01 -1.86724043e+00
6.89751089e-01 -7.10875750e-01 1.58826768e-01 -1.31840134e+00
1.54615538e-02 -6.94129169e-01 -3.54871094e-01 6.34179294e-01
-2.18468178e-02 -4.14104223e-01 2.77012378e-01 2.77645677e-01
-7.14001238e-01 1.06710231e+00 1.69356513e+00 -7.55202174e-02
-5.79371452e-01 2.02141494e-01 -7.39728749e-01 7.45346665e-01
8.36995125e-01 -5.89070916e-01 -1.04866958e+00 -2.44572669e-01
3.32996179e-03 2.12948471e-01 1.32794261e-01 -5.66534042e-01
4.61643822e-02 -7.88070560e-01 1.84710070e-01 -3.05037171e-01
3.27656239e-01 -6.72305703e-01 -3.10370356e-01 4.78158265e-01
-5.33494115e-01 -1.78981200e-01 1.40529405e-02 9.11059320e-01
1.13664858e-01 -2.64165223e-01 5.70643783e-01 -1.27354562e-01
-6.67731464e-01 5.06687880e-01 -3.65862608e-01 2.22992241e-01
1.14302957e+00 8.78061503e-02 -4.29585516e-01 -6.52163029e-01
-6.26541853e-01 4.51876760e-01 5.24595559e-01 4.04258519e-01
5.50305903e-01 -1.31446195e+00 -4.26750004e-01 1.20197460e-01
-3.87202650e-02 -9.24523473e-02 3.39226685e-02 8.55820179e-01
-4.44994867e-02 2.63178229e-01 1.14561066e-01 -4.66722608e-01
-6.39929950e-01 6.27256572e-01 3.16543758e-01 -6.13026381e-01
-8.18377197e-01 4.13610846e-01 3.19778085e-01 -3.16078395e-01
3.57463807e-01 -3.34338248e-01 -4.10428308e-02 -2.22305223e-01
3.08113039e-01 1.61092803e-01 -4.77897525e-01 -1.22731782e-01
-5.25970049e-02 3.05468559e-01 -1.91984296e-01 -3.81559312e-01
1.29688799e+00 1.34876043e-01 4.91933614e-01 4.20293659e-01
7.87615776e-01 1.33370832e-01 -1.94164169e+00 -1.92484736e-01
-2.08649769e-01 -5.74824274e-01 -8.00597221e-02 -7.67130435e-01
-9.47784543e-01 5.09959161e-01 2.03531072e-01 1.28541678e-01
8.44887197e-01 -1.58904232e-02 6.96783185e-01 6.14441156e-01
6.09965026e-01 -1.42111063e+00 3.03896725e-01 4.99815255e-01
8.07348430e-01 -1.29806912e+00 -9.12810713e-02 1.48870289e-01
-1.03427875e+00 9.27400768e-01 7.59625971e-01 -4.06836063e-01
3.42189431e-01 2.93905765e-01 -1.35185316e-01 2.86551248e-02
-1.04088402e+00 -4.42546129e-01 1.03466369e-01 5.29305637e-01
1.17143892e-01 6.63877949e-02 -2.78621137e-01 3.89507294e-01
1.73404843e-01 7.53786275e-03 3.05731744e-01 1.24242532e+00
-4.32958931e-01 -1.54801285e+00 -1.71567053e-02 2.72529006e-01
-3.47061843e-01 1.59059748e-01 -1.89376950e-01 1.12525785e+00
-4.04658377e-01 7.70415664e-01 -1.19635843e-01 -2.09098563e-01
8.49420801e-02 5.73355285e-03 8.95031750e-01 -5.50909936e-01
-4.22272623e-01 4.45074648e-01 5.91292717e-02 -9.82339263e-01
-3.88144165e-01 -4.87850279e-01 -1.31524694e+00 -5.47829792e-02
-3.03637027e-03 7.50093460e-02 3.48151863e-01 1.03938997e+00
3.20054919e-01 5.47335982e-01 6.02830589e-01 -8.20393443e-01
-1.44603705e+00 -7.46432602e-01 -5.39140284e-01 3.16842467e-01
2.84492403e-01 -9.57478404e-01 -2.39130825e-01 -3.83101076e-01]
|
[4.079798221588135, 2.027352809906006]
|
e4b7df2e-d5e7-4da0-99a7-f92b84c057a2
|
a-comment-on-the-paper-prediction-of-kidney
|
1707.09869
| null |
http://arxiv.org/abs/1707.09869v1
|
http://arxiv.org/pdf/1707.09869v1.pdf
|
A comment on the paper Prediction of Kidney Function from Biopsy Images using Convolutional Neural Networks
|
This letter presente a comment on the paper Prediction of Kidney Function
from Biopsy Images using Convolutional Neural Networks by Ledbetter et al.
(2017)
|
['Washington LC dos-Santos', 'Luiz AR de Freitas', 'Angelo A Duarte']
|
2017-07-23
| null | null | null | null |
['kidney-function']
|
['medical']
|
[ 1.27503604e-01 5.84076643e-01 -1.89660653e-01 -8.83463204e-01
-4.66835164e-02 -1.69398069e-01 1.80356890e-01 2.15527058e-01
-5.53965807e-01 1.19790328e+00 2.69881755e-01 -5.78297794e-01
-2.68430024e-01 -9.00581300e-01 -6.60504878e-01 -5.45251608e-01
-4.59679306e-01 5.84995866e-01 -3.05290282e-01 3.31624210e-01
9.72808972e-02 9.38781619e-01 -6.98069334e-01 5.77549577e-01
7.90096998e-01 9.58979011e-01 1.70267999e-01 1.17152429e+00
8.55167732e-02 1.14994252e+00 -1.82244867e-01 -3.28501344e-01
3.49482149e-01 -3.03301662e-01 -7.75424957e-01 -1.03967302e-02
8.48522544e-01 -8.54981959e-01 -6.91007853e-01 7.03385770e-01
7.75056779e-01 -6.05266333e-01 7.17341185e-01 -2.47102574e-01
-1.06855094e+00 9.69832420e-01 4.11154658e-01 8.46447587e-01
-2.78065026e-01 2.42095768e-01 2.52206147e-01 -8.87258530e-01
6.37608230e-01 7.08286166e-01 9.46514249e-01 6.64721251e-01
-1.19757628e+00 -6.71769500e-01 -5.14599800e-01 4.66166347e-01
-8.31297040e-01 -1.65428892e-01 1.43267050e-01 -6.00299716e-01
9.26560163e-01 1.97994351e-01 1.11612725e+00 6.74714923e-01
5.96622705e-01 8.60305667e-01 1.33263052e+00 -4.40810829e-01
-1.70046538e-01 -3.07177693e-01 4.17327672e-01 9.46465313e-01
7.12391019e-01 4.19731975e-01 3.86022665e-02 -1.57487035e-01
1.17257547e+00 -1.46879554e-01 -1.35041364e-02 -2.78549403e-01
-1.01665735e+00 6.89526975e-01 6.91701651e-01 8.35851654e-02
-7.91394114e-01 3.38010430e-01 5.85708916e-01 2.41622776e-01
2.24175423e-01 2.01856971e-01 -6.18384898e-01 2.37806246e-01
-7.91650891e-01 4.51194882e-01 8.65894675e-01 5.93415797e-01
-5.36451414e-02 -4.51396313e-03 -2.88282067e-01 5.96840382e-01
3.85857940e-01 5.13911366e-01 1.72255591e-01 -6.86117172e-01
-1.21354207e-01 1.70163423e-01 -1.47109509e-01 -3.25151496e-02
-9.12169755e-01 -2.10304856e-01 -1.28011632e+00 2.99227118e-01
6.80286825e-01 -4.64650154e-01 -1.17909527e+00 5.38413584e-01
-1.47881553e-01 -3.29332411e-01 -1.27188176e-01 1.12707067e+00
1.48824656e+00 -1.42701879e-01 4.02189225e-01 2.85975486e-01
1.06864381e+00 -9.33609843e-01 -5.89452624e-01 3.60831499e-01
4.54572499e-01 -4.82805789e-01 -1.74071461e-01 4.11300302e-01
-1.23583937e+00 -3.47041100e-01 -8.46340120e-01 -1.52768999e-01
-1.44489810e-01 5.62353075e-01 1.16583669e+00 6.50926709e-01
-1.19440734e+00 7.89941669e-01 -1.07834768e+00 -7.13473082e-01
7.36718953e-01 3.45153570e-01 -6.42046869e-01 4.26135138e-02
-9.15140390e-01 1.51854384e+00 1.96785316e-01 5.63409865e-01
-4.79364306e-01 -8.92695546e-01 -5.52501500e-01 -1.57366544e-01
-4.53738689e-01 -9.75965738e-01 1.33559608e+00 -4.26391184e-01
-1.54373038e+00 1.46698856e+00 1.13272905e-01 -9.89384055e-01
6.78697407e-01 -1.45979434e-01 -2.89979935e-01 4.16218132e-01
-5.36400378e-01 5.36246121e-01 8.05075541e-02 -4.36752915e-01
-3.26880038e-01 -5.79812527e-01 -2.74787247e-01 -4.17839676e-01
4.61075276e-01 3.03163230e-01 4.46006358e-01 -5.94886124e-01
2.66507268e-01 -6.79088056e-01 -4.92244959e-01 3.71563494e-01
-2.23298728e-01 -1.36709571e-01 3.86051368e-03 -8.66950452e-01
6.08739257e-01 -1.24014091e+00 -2.18961611e-01 8.94507021e-02
6.81301892e-01 5.07670105e-01 3.77529860e-01 2.05880746e-01
-7.29818583e-01 2.56175041e-01 1.11702845e-01 4.01480287e-01
-1.82263806e-01 3.25828284e-01 1.49432987e-01 8.92475605e-01
1.69433564e-01 1.56448293e+00 -6.55757070e-01 -2.90969402e-01
6.96696997e-01 3.09027255e-01 -6.87699318e-02 1.23464748e-01
6.26507759e-01 6.11901522e-01 -9.13644433e-02 1.04636264e+00
1.02532113e+00 -4.23064321e-01 1.90691963e-01 -3.01967978e-01
-2.99955070e-01 1.46473393e-01 -5.01756907e-01 1.02748358e+00
1.38156369e-01 6.51613593e-01 3.14102352e-01 -1.06183195e+00
8.37810934e-01 6.51910901e-01 7.08497643e-01 -3.51151466e-01
3.62642109e-01 3.02179635e-01 3.56044412e-01 -6.49195313e-01
-3.51843357e-01 -6.89093709e-01 4.65946555e-01 1.70319557e-01
1.92546934e-01 7.21413195e-02 8.73626322e-02 -5.37340760e-01
1.15408039e+00 2.93433107e-02 7.61102736e-01 -7.40310311e-01
5.60009003e-01 -6.08140379e-02 1.31837189e-01 1.02083683e+00
-9.99417543e-01 7.27592468e-01 4.90496516e-01 -1.38096440e+00
-1.32917583e+00 -1.17788935e+00 -1.16616297e+00 1.28276467e-01
-4.08782452e-01 2.35301167e-01 -4.43642765e-01 -4.20817018e-01
7.84000039e-01 -1.66104957e-01 -1.55800152e+00 3.53957355e-01
-6.91135287e-01 -8.61338735e-01 9.07922029e-01 1.00839686e+00
9.90056694e-02 -1.34656966e+00 -5.78267634e-01 4.54458967e-02
2.13838607e-01 -4.38536972e-01 2.82372892e-01 7.43821383e-01
-1.29342222e+00 -1.43312383e+00 -1.42995906e+00 -9.62819576e-01
7.21281826e-01 -4.04370546e-01 1.05516183e+00 2.28792742e-01
-7.67723441e-01 -2.43979752e-01 -4.83438112e-02 -9.45190907e-01
-5.42927563e-01 9.08550620e-02 -1.86506003e-01 -6.43296421e-01
9.12945092e-01 -4.28635895e-01 -1.00820422e+00 -2.13509768e-01
-7.44136155e-01 6.09729104e-02 9.69993114e-01 7.87290454e-01
5.03775656e-01 -1.17015338e+00 7.26991773e-01 -1.02605534e+00
4.14882213e-01 -3.85664046e-01 -4.80935156e-01 1.14973672e-01
-1.09370792e+00 -1.59538448e-01 1.13145575e-01 3.49494778e-02
-5.15498042e-01 -8.44086185e-02 -3.68576556e-01 -1.44033700e-01
-3.64984334e-01 2.24239573e-01 8.77604246e-01 -5.44948936e-01
7.07666874e-01 5.06525040e-02 3.21207672e-01 -6.68740630e-01
-6.67602271e-02 5.91201782e-01 6.38978720e-01 -1.81589037e-01
1.61701798e-01 5.91440678e-01 2.65639096e-01 -3.48509908e-01
-4.13460642e-01 -3.56676221e-01 -1.04445827e+00 -3.43683720e-01
9.78083372e-01 -8.99096727e-01 -9.21157002e-01 5.09753823e-01
-8.87497544e-01 -3.75588000e-01 -3.96856725e-01 1.07569253e+00
-6.94651246e-01 2.01988459e-01 -1.18148112e+00 -4.94171411e-01
-8.99621665e-01 -9.25792933e-01 4.03005302e-01 2.63584822e-01
-1.37772575e-01 -9.24621642e-01 2.39779145e-01 2.09191635e-01
6.99760020e-01 4.95936036e-01 9.64877665e-01 -4.73543674e-01
-4.96165395e-01 -4.58760023e-01 -7.27233589e-01 3.18488002e-01
2.93112993e-02 -6.11546077e-02 -8.12720180e-01 -1.56629160e-01
-3.94851506e-01 -3.00770462e-01 1.13856220e+00 1.16204453e+00
1.50825822e+00 1.58781216e-01 -1.46560282e-01 7.02617764e-01
1.56726468e+00 3.41368169e-02 8.36164057e-01 6.55745745e-01
1.85298502e-01 3.02435488e-01 -1.43868491e-01 2.32198447e-01
2.28314951e-01 -3.37436199e-01 4.16478366e-01 -4.65337992e-01
-4.94497210e-01 4.22882468e-01 -7.29099333e-01 4.05460596e-01
-6.45268321e-01 1.45636946e-01 -9.79373932e-01 7.12909818e-01
-1.53971684e+00 -5.24490178e-01 -6.78070009e-01 1.75364947e+00
3.73509824e-01 -3.95201407e-02 1.26949579e-01 -4.20614988e-01
7.47772336e-01 -2.43980005e-01 -3.90432358e-01 -4.83438790e-01
-4.72233072e-02 8.24249506e-01 1.10123324e+00 3.37836921e-01
-9.08198237e-01 2.84881622e-01 9.22211742e+00 -3.51742566e-01
-1.04722428e+00 -1.10733435e-01 6.32755876e-01 1.71908364e-03
1.51375473e-01 -2.51310080e-01 -1.24190196e-01 2.21589148e-01
9.72442806e-01 -3.47550884e-02 1.77154481e-01 2.74574131e-01
2.08760723e-02 -1.27759308e-01 -8.79453123e-01 5.73523760e-01
-2.58411914e-02 -1.51609981e+00 -6.21452391e-01 9.47600156e-02
2.66305596e-01 8.32219839e-01 -2.19841182e-01 2.85803914e-01
2.42085934e-01 -1.51697707e+00 1.50695652e-01 1.18664622e+00
6.53172970e-01 -3.81881177e-01 1.32071614e+00 -1.71342418e-01
-6.41652048e-01 7.16410726e-02 -4.61698085e-01 -4.14562970e-01
-1.20076835e-02 8.46896648e-01 -1.27304983e+00 2.96514392e-01
7.98207998e-01 4.01587963e-01 -5.31562090e-01 1.50870287e+00
-9.86669958e-02 7.82154083e-01 1.29992291e-01 -1.70771241e-01
9.30392891e-02 -7.10498989e-02 1.24150209e-01 1.52961016e+00
-1.29455596e-01 5.83288372e-02 -1.07393831e-01 9.91633713e-01
4.70684692e-02 3.39924276e-01 -3.47589523e-01 -2.98761010e-01
-3.76375884e-01 1.28166902e+00 -8.32514644e-01 -3.79300356e-01
-5.45144498e-01 4.40001786e-01 1.94410682e-01 1.67429537e-01
-2.10198328e-01 -3.66688311e-01 3.74119788e-01 2.41582111e-01
2.03302801e-01 5.23446083e-01 -1.17450869e+00 -1.04807889e+00
6.49224818e-02 1.60661899e-02 3.68511170e-01 -8.69286895e-01
-1.57294416e+00 -5.85195720e-02 -2.94164181e-01 -9.35115457e-01
6.89437054e-03 -1.22118783e+00 -4.65949833e-01 1.52421939e+00
-1.68148232e+00 -1.21474791e+00 -1.20560899e-01 -2.67424345e-01
-4.77235094e-02 -1.43621147e-01 1.06975615e+00 2.30305538e-01
-8.65980983e-02 9.47160274e-02 2.87315041e-01 7.29244590e-01
9.27749515e-01 -1.81633902e+00 5.43978453e-01 -4.10149023e-02
-9.57835436e-01 6.93837285e-01 4.78173018e-01 -7.01372385e-01
-1.04424262e+00 -9.50529754e-01 1.53844833e+00 -4.26950336e-01
3.75665367e-01 2.47502953e-01 -5.68880677e-01 9.88536596e-01
7.15847790e-01 4.85939085e-01 9.45965052e-01 7.95698613e-02
1.66354895e-01 2.27313429e-01 -1.39388859e+00 6.47739768e-02
4.47852612e-01 -3.03875268e-01 -1.02118754e+00 3.38572472e-01
-2.24729121e-01 -8.04315805e-01 -1.38384330e+00 9.74214673e-01
1.22064769e+00 -9.08979535e-01 8.23352039e-01 -1.22886443e+00
7.82718003e-01 1.79730535e-01 4.50270295e-01 -1.03709757e+00
-5.99974275e-01 2.88021147e-01 -1.48031503e-01 -1.02515742e-02
1.18115321e-02 -5.26081860e-01 9.68322158e-01 5.10625601e-01
-2.47189730e-01 -1.20284510e+00 -8.52892756e-01 -4.02189672e-01
1.00416493e+00 1.12681679e-01 3.54754210e-01 8.24880660e-01
5.01221895e-01 -2.78973311e-01 1.74058661e-01 -6.89556673e-02
5.48968911e-01 1.19840726e-01 3.59095424e-01 -1.65763807e+00
2.37460300e-01 -4.78192836e-01 -8.80895615e-01 -3.24634492e-01
-1.08593991e-02 -1.44674659e+00 -8.16336274e-02 -2.04237914e+00
5.11345685e-01 -4.30958532e-02 -8.81795049e-01 4.45075035e-01
-5.10066003e-02 5.63028812e-01 2.34611318e-01 3.20710391e-02
-1.66361392e-01 -4.25250471e-01 1.44081187e+00 -1.85146570e-01
3.22525442e-01 2.32874304e-01 -7.32844234e-01 3.38900030e-01
6.39976978e-01 -2.53186494e-01 1.88068882e-01 -1.50653735e-01
8.76940563e-02 3.17911029e-01 7.69408822e-01 -8.22806001e-01
1.32980779e-01 7.26654660e-03 1.39182234e+00 -6.61593676e-01
-3.65171820e-01 -4.95549053e-01 7.64089376e-02 1.20778704e+00
-4.65326041e-01 -1.61150977e-01 2.17206284e-01 3.13056886e-01
1.37606442e-01 -3.24478447e-01 9.96156573e-01 -7.21192479e-01
-3.22151959e-01 4.21361923e-01 -8.65064621e-01 -4.75021601e-01
6.22199357e-01 -2.77600229e-01 -2.36461133e-01 -1.01908363e-01
-1.47139239e+00 4.16752815e-01 -5.93594313e-02 1.16950637e-02
5.12765288e-01 -1.24983120e+00 -1.26218021e+00 7.00741634e-02
1.63615212e-01 -1.98551297e-01 4.29911435e-01 1.41433632e+00
-1.21257389e+00 1.22763956e+00 -6.84073806e-01 -4.67502356e-01
-9.64026690e-01 2.46200293e-01 7.36741424e-01 -2.07636058e-01
-1.16418576e+00 7.66140938e-01 -3.32345366e-01 -8.02161753e-01
-3.07524800e-02 -8.03259373e-01 -3.06664407e-01 -4.38183039e-01
7.66146302e-01 3.92046630e-01 4.67566252e-01 -4.96207662e-02
-3.67234260e-01 -1.72055364e-01 -3.31372350e-01 4.50002134e-01
1.59696710e+00 3.74625102e-02 -4.20647949e-01 4.57607657e-01
9.72496092e-01 -5.90719342e-01 -6.96325839e-01 -2.75027901e-01
-7.60394707e-03 -2.76830792e-01 2.51851976e-01 -1.46252096e+00
-9.51421678e-01 5.38153172e-01 1.10021400e+00 5.55459410e-02
8.15597892e-01 -1.19200811e-01 9.17480946e-01 5.54404736e-01
-1.06244013e-02 -1.00075376e+00 -9.30265605e-01 5.62788785e-01
8.59174192e-01 -1.24172354e+00 3.45638752e-01 -2.66653031e-01
-2.07372755e-01 1.82511175e+00 4.71271813e-01 -4.70228404e-01
1.08049238e+00 1.07933305e-01 6.21670604e-01 -4.18447107e-01
-7.24545419e-01 -3.21638882e-01 2.27842599e-01 8.27190280e-01
8.75155747e-01 5.02849579e-01 -1.35274887e+00 4.80005264e-01
-1.03047360e-02 9.30462182e-01 6.29655004e-01 8.48381639e-01
-5.28717756e-01 -9.90955651e-01 -2.83520073e-01 1.30833673e+00
-6.41806364e-01 -1.51010290e-01 -6.08522832e-01 5.06686985e-01
2.99563378e-01 2.94938385e-01 1.32062316e-01 3.09411585e-01
2.72139907e-01 5.59778452e-01 8.54996920e-01 -5.04427552e-01
-7.95782626e-01 -3.77777778e-02 -2.00665146e-02 -1.04694210e-01
-6.28117561e-01 -5.92312157e-01 -1.06395698e+00 -2.26570845e-01
-6.68339506e-02 -7.86022246e-02 7.20521212e-01 1.15626442e+00
-1.46577135e-01 8.24154675e-01 2.12549031e-01 -3.94443035e-01
-4.50601906e-01 -1.62646246e+00 -1.25727928e+00 -1.04248092e-01
5.52561641e-01 -3.72523397e-01 -1.26987308e-01 1.95445701e-01]
|
[14.2943754196167, -2.482269763946533]
|
8589d6db-226a-4f7d-bd8f-61883f3e89e6
|
fast-blind-audio-copy-move-detection-and
|
2302.07584
| null |
https://arxiv.org/abs/2302.07584v1
|
https://arxiv.org/pdf/2302.07584v1.pdf
|
Fast Blind Audio Copy-Move Detection and Localization Using Local Feature Tensors in Noise
|
The increasing availability of audio editing software altering digital audios and their ease of use allows create forgeries at low cost. A copy-move forgery (CMF) is one of easiest and popular audio forgeries, which created by copying and pasting audio segments within the same audio, and potentially post-processing it. Three main approaches to audio copy-move detection exist nowadays: samples/frames comparison, acoustic features coherence searching and dynamic time warping. But these approaches will suffer from computational complexity and/or sensitive to noise and post-processing. In this paper, we propose a new local feature tensors-based copy-move detection algorithm that can be applied to transformed duplicates detection and localization problem to a special locality sensitive hash like procedure. The experimental results with massive online real-time audios datasets reveal that the proposed technique effectively determines and locating copy-move forgeries even on a forged speech segment are as short as fractional second. This method is also computational efficient and robust against the audios processed with severe nonlinear transformation, such as resampling, filtering, jsittering, compression and cropping, even contaminated with background noise and music. Hence, the proposed technique provides an efficient and reliable way of copy-move forgery detection that increases the credibility of audio in practical forensics applications
|
['Muyong Cao', 'Mingle Liu', 'Dong Yang']
|
2023-02-15
| null | null | null | null |
['dynamic-time-warping']
|
['time-series']
|
[ 5.14257073e-01 -6.62595987e-01 4.45058823e-01 3.56714696e-01
-1.21244395e+00 -8.39958310e-01 2.35916659e-01 4.08259064e-01
-2.23480687e-01 4.36416626e-01 1.22766078e-01 1.57615036e-01
-4.04476285e-01 -4.93526042e-01 -4.49018270e-01 -8.06920350e-01
-5.05030453e-01 -2.60000497e-01 6.58727288e-01 8.77755880e-03
7.45238721e-01 5.95799148e-01 -1.86851466e+00 3.73906702e-01
4.61840928e-01 1.03286946e+00 3.72312039e-01 9.86203194e-01
2.28962973e-01 4.13263738e-01 -1.05153167e+00 -2.88477898e-01
4.85643327e-01 -3.53344351e-01 -4.73491937e-01 5.81970289e-02
2.38524675e-02 -4.72981989e-01 -3.08579594e-01 1.24388885e+00
7.84173250e-01 2.57267475e-01 1.33260712e-01 -1.25593257e+00
-2.15204075e-01 6.47480428e-01 -8.37437332e-01 7.53003776e-01
7.55992174e-01 -5.18408157e-02 3.39478910e-01 -1.00678647e+00
5.04728913e-01 1.38836741e+00 9.97257113e-01 -2.65632570e-01
-5.98328769e-01 -9.80755091e-01 -7.53276408e-01 7.98651099e-01
-1.90849888e+00 -6.83483243e-01 1.07765043e+00 -4.22446057e-02
5.16162097e-01 7.72366166e-01 4.60092425e-01 5.02276182e-01
3.65027338e-01 4.38477933e-01 9.51040864e-01 -7.51614392e-01
-3.23061198e-02 -3.73295903e-01 -3.03814769e-01 1.52912661e-01
-7.89836198e-02 1.53462946e-01 -9.47415173e-01 -6.58128917e-01
3.27444196e-01 -5.40568074e-03 -6.82183266e-01 4.61291999e-01
-1.25758350e+00 3.69093895e-01 -2.84676224e-01 4.13033724e-01
-3.62650692e-01 1.67546719e-01 7.35522985e-01 7.85288572e-01
6.28069267e-02 2.18494713e-01 6.94000497e-02 -4.29275393e-01
-1.39017868e+00 5.06728530e-01 1.56215534e-01 6.37698412e-01
4.59846497e-01 4.49744388e-02 1.51853517e-01 7.10932732e-01
1.56547725e-01 2.90479958e-01 9.34234023e-01 -7.70536900e-01
5.64108670e-01 -1.05709463e-01 6.41280040e-02 -1.74581432e+00
-4.72832024e-02 -1.07149564e-01 -6.41192436e-01 8.38099495e-02
1.62315384e-01 2.10198373e-01 -2.18524590e-01 1.17983937e+00
6.42365277e-01 7.39336491e-01 -3.17449123e-01 7.27934241e-01
3.01896691e-01 8.61106336e-01 -3.99040520e-01 -3.76240790e-01
1.55517805e+00 -3.92474294e-01 -1.16184235e+00 3.86182934e-01
3.23052555e-01 -1.74016786e+00 6.87009454e-01 9.27279949e-01
-1.14267731e+00 -5.92534244e-01 -1.20262086e+00 2.50633180e-01
-2.80662298e-01 -4.53055603e-03 -3.47128026e-02 1.03938544e+00
-7.94978738e-01 9.46023226e-01 -4.70243126e-01 -4.13125008e-03
-1.49927028e-02 1.88671425e-01 -4.28407878e-01 3.64902131e-02
-1.22435021e+00 3.38951498e-01 3.99157763e-01 1.41201362e-01
-7.52224386e-01 -6.18869841e-01 -1.90479189e-01 2.27140393e-02
2.30587408e-01 7.53121376e-02 9.14720893e-01 -7.28291750e-01
-1.23734772e+00 5.57717562e-01 1.27944618e-01 -6.34817839e-01
5.47380924e-01 -9.73466784e-02 -1.05621672e+00 6.02612257e-01
-2.29319674e-03 -2.39776015e-01 1.76799142e+00 -5.62339008e-01
-4.23063368e-01 -5.11704326e-01 -7.30539858e-01 1.68084376e-03
-3.16642404e-01 7.26377964e-01 -2.72280276e-02 -1.40309060e+00
5.88668227e-01 -6.52623594e-01 4.94602054e-01 3.57584655e-02
-3.73254716e-01 1.46981031e-01 1.50647795e+00 -1.21301413e+00
1.49315381e+00 -2.61688757e+00 -3.28932881e-01 3.20667565e-01
-1.08227134e-01 3.49340051e-01 5.90426102e-02 9.66554999e-01
-2.42643118e-01 -1.93881039e-02 -2.55119145e-01 -2.07979884e-03
-3.51342231e-01 -2.10487917e-01 -7.83923268e-01 1.07876062e+00
-1.71541572e-01 1.42698467e-01 -6.71473563e-01 -5.27732253e-01
7.65540823e-02 4.35839534e-01 -2.29078099e-01 -2.66064703e-03
5.33052027e-01 1.17421746e-01 -1.01878308e-01 7.59683728e-01
1.04903972e+00 7.55955398e-01 -1.92225114e-01 -4.03803140e-01
-2.93230087e-01 1.59543231e-01 -2.02019215e+00 1.59683597e+00
-3.13680340e-03 7.78923094e-01 5.09131968e-01 -8.99885476e-01
1.06914997e+00 5.96043527e-01 2.69654304e-01 -3.64702553e-01
1.01210900e-01 4.50529724e-01 -3.19549799e-01 -8.75375807e-01
8.62469673e-01 1.73435975e-02 1.26657039e-01 7.21713424e-01
-2.48328641e-01 -5.31050237e-03 -2.71065205e-01 -3.09471786e-03
1.33843791e+00 -1.09568849e-01 2.96785861e-01 3.13581415e-02
7.90669143e-01 -6.71838820e-01 5.39883375e-01 3.19233388e-01
-3.34631205e-01 7.15814412e-01 -7.83979222e-02 2.32394282e-02
-8.84451032e-01 -7.39242375e-01 -4.65055332e-02 7.79430270e-01
9.14094970e-02 -5.21159768e-01 -7.60365784e-01 6.55730963e-02
-4.98650856e-02 1.51187465e-01 3.78269739e-02 -3.12022954e-01
-8.16693068e-01 -3.49137485e-01 1.36597764e+00 4.27582487e-02
5.31540692e-01 -8.14394951e-01 -5.85349381e-01 6.42339706e-01
-4.77270186e-01 -1.05856466e+00 -7.12282658e-01 -3.61798227e-01
-9.63621914e-01 -1.04215372e+00 -6.97241962e-01 -7.73533881e-01
5.48815504e-02 7.11235464e-01 3.15520853e-01 3.09196591e-01
-8.48035038e-01 2.39198595e-01 -6.94932401e-01 -1.27101988e-01
-6.07616186e-01 -6.53921962e-01 2.05162734e-01 4.81256366e-01
-5.32045849e-02 -8.92931163e-01 -6.40756726e-01 4.66423035e-01
-1.16572237e+00 -8.18657875e-01 2.50675410e-01 5.21677375e-01
2.65087903e-01 8.93587708e-01 6.55725360e-01 -1.69478104e-01
1.01077151e+00 -3.73584151e-01 -2.66793638e-01 -3.77910957e-02
-2.99564987e-01 -4.93992746e-01 6.50985837e-01 -4.93912995e-01
-8.44882131e-01 -1.97536215e-01 6.86999708e-02 -6.80540740e-01
-5.40646240e-02 2.54248977e-01 -1.20643340e-01 -3.71673197e-01
5.88864207e-01 6.39354110e-01 -8.49572662e-03 -1.01425934e+00
2.43478104e-01 1.10376418e+00 1.08232474e+00 -3.69264096e-01
1.11259246e+00 5.67385435e-01 6.91363662e-02 -1.13674235e+00
4.96944964e-01 -6.43010437e-01 -4.51874554e-01 -3.58035386e-01
3.59004080e-01 -7.67555714e-01 -6.21340692e-01 9.28756058e-01
-1.13001621e+00 5.83068788e-01 2.29516644e-02 4.50723916e-01
-2.51777083e-01 1.61579311e+00 -7.09667802e-01 -1.21326685e+00
-4.48300749e-01 -8.86995673e-01 9.02162135e-01 -3.02171797e-01
-2.07948267e-01 -3.25588554e-01 -6.66935593e-02 3.75096321e-01
2.78246075e-01 4.59932446e-01 6.02486312e-01 -3.30109626e-01
-6.13056719e-01 -6.66414976e-01 1.76788077e-01 2.06302494e-01
1.97092324e-01 1.27940491e-01 -1.07166910e+00 -4.67279881e-01
4.64013636e-01 6.97768927e-02 4.02702063e-01 -8.07807297e-02
1.12833846e+00 -5.67693233e-01 -1.03891075e-01 4.31016743e-01
1.23308003e+00 4.17171508e-01 9.40246522e-01 2.99123347e-01
1.62784666e-01 5.48578799e-01 9.90805745e-01 1.01480544e+00
-4.00776803e-01 8.98532152e-01 4.37111259e-01 5.21582127e-01
-2.27414817e-01 -1.43698245e-01 5.57068050e-01 1.36031806e+00
-3.09450296e-03 -1.93241820e-01 -4.92960870e-01 6.52267516e-01
-1.37735748e+00 -1.48493993e+00 -3.40844572e-01 2.63369560e+00
7.69528985e-01 -1.75354302e-01 2.16511920e-01 1.36695504e+00
1.28627431e+00 1.99617252e-01 8.51979828e-04 -3.28138262e-01
-1.77025229e-01 5.26521504e-01 5.18996418e-01 2.06995964e-01
-8.27627838e-01 4.00107145e-01 5.40117741e+00 1.58604860e+00
-1.00190997e+00 4.47421134e-01 -9.67182666e-02 -6.03461750e-02
1.27685353e-01 1.35785282e-01 -1.58493713e-01 8.91731203e-01
9.40554738e-01 -1.21080101e-01 5.75371265e-01 3.98362458e-01
6.59076273e-01 -1.71086311e-01 -4.37449932e-01 1.47071469e+00
2.02376813e-01 -1.09602499e+00 -2.62876630e-01 -6.85999766e-02
7.51920566e-02 -6.47339225e-01 -5.88601315e-03 -3.99000347e-01
-6.37651622e-01 -5.76562047e-01 9.55181241e-01 2.43194669e-01
6.34416521e-01 -1.07951367e+00 4.85476613e-01 2.61901438e-01
-1.57918668e+00 -5.07274307e-02 -4.41682458e-01 2.18219776e-03
2.68614471e-01 7.43160248e-01 -8.58450174e-01 5.62461555e-01
8.34569216e-01 3.03549230e-01 -2.66079605e-01 1.34915793e+00
3.23798209e-01 6.97970033e-01 -5.75818896e-01 4.10110414e-01
-6.34331480e-02 -1.49899796e-01 1.01440012e+00 1.25840771e+00
1.01322174e+00 -4.23410460e-02 -2.24143460e-01 5.01435220e-01
9.63965878e-02 4.71308321e-01 -3.85144442e-01 5.22166258e-03
1.04686105e+00 9.56240714e-01 -7.60288119e-01 -7.83984512e-02
1.24797016e-01 1.14014268e+00 -5.48008263e-01 -8.53794143e-02
-8.75738502e-01 -1.10202861e+00 3.32014859e-01 3.74489695e-01
3.89616549e-01 -3.77248794e-01 2.69013494e-01 -7.66538143e-01
4.68817294e-01 -1.10696983e+00 4.12298799e-01 -7.87042618e-01
-8.92550468e-01 1.31987184e-01 -1.67536169e-01 -1.91608322e+00
-1.74655498e-03 -6.51938990e-02 -7.58221865e-01 5.24672747e-01
-1.11123657e+00 -7.22941458e-01 -1.64597541e-01 9.67297077e-01
7.16714561e-01 -1.94063261e-01 5.37116051e-01 7.47477591e-01
-1.45663530e-01 7.23684490e-01 2.94275671e-01 -1.43355608e-01
9.14509952e-01 -5.45537353e-01 3.38481009e-01 1.13875365e+00
1.28899336e-01 6.87838554e-01 8.53607833e-01 -8.14487457e-01
-1.61913741e+00 -6.58714950e-01 8.31029773e-01 3.66304666e-01
7.30201244e-01 -1.08409978e-01 -1.04169977e+00 1.90628886e-01
1.31147400e-01 -2.49409407e-01 7.66183496e-01 -7.69557655e-01
-3.61455530e-01 -1.71205327e-01 -1.58899379e+00 3.30944359e-03
6.86653912e-01 -8.90217721e-01 -7.43217111e-01 4.60392475e-01
4.15419608e-01 -2.80209601e-01 -1.01770437e+00 -1.25817537e-01
5.93882859e-01 -1.25564826e+00 1.21052635e+00 1.32612795e-01
-1.94891647e-01 -7.43410587e-01 -2.64640391e-01 -6.56271160e-01
-2.24481846e-04 -1.69938207e+00 2.94028898e-03 1.63748193e+00
-3.84628952e-01 -5.17923832e-01 3.05986851e-01 -2.38271192e-01
-9.58582386e-02 5.38176578e-03 -1.53643084e+00 -1.02536309e+00
-6.43633902e-01 -6.85274184e-01 8.53354692e-01 1.34851813e+00
2.05138043e-01 -5.84774017e-01 -8.33564758e-01 4.80327696e-01
9.04736817e-01 -2.44667277e-01 7.30272770e-01 -9.13641095e-01
-6.73454821e-01 4.70746122e-02 -1.06910169e+00 -6.69058979e-01
-4.95190054e-01 -4.93792653e-01 -2.33387187e-01 -4.65031147e-01
-4.23549116e-01 -2.91511923e-01 -1.46307768e-02 -3.73029225e-02
5.86740933e-02 5.20893574e-01 2.26164475e-01 6.95223808e-01
1.70495331e-01 3.44069004e-01 8.56207192e-01 8.65240917e-02
-4.00599791e-03 1.80261970e-01 1.04943804e-01 5.25076389e-01
4.91611868e-01 -9.31011558e-01 -3.09474617e-01 1.23043932e-01
1.97939306e-01 3.31574231e-01 5.51505744e-01 -1.41015065e+00
3.31589788e-01 1.73760235e-01 -1.46963960e-02 -8.94172192e-01
4.54512089e-01 -8.33528876e-01 5.31738937e-01 5.49313068e-01
-6.38287291e-02 5.88901401e-01 1.17911033e-01 9.23675179e-01
-5.28751791e-01 -7.22822726e-01 8.00495327e-01 -8.55345353e-02
-4.74443287e-01 -2.66090721e-01 -5.90009868e-01 -3.54246914e-01
1.05308437e+00 -6.34502649e-01 -5.29429503e-02 -4.34649467e-01
-5.50366044e-01 -7.28561878e-01 2.04126254e-01 2.50650376e-01
8.37316573e-01 -1.18716049e+00 -6.56821072e-01 2.96690762e-01
-2.45533019e-01 -5.26805639e-01 5.80875039e-01 7.40725994e-01
-8.79436016e-01 -3.60609256e-02 -2.49882296e-01 -4.08151090e-01
-1.83606529e+00 6.94361329e-01 -1.62447020e-01 1.99591115e-01
-8.70296538e-01 7.25729823e-01 -6.62373602e-01 3.70618343e-01
1.14969842e-01 -3.79040480e-01 1.71549901e-01 2.42007121e-01
8.71104240e-01 1.14480615e+00 4.48758572e-01 -8.46617043e-01
-3.52079928e-01 7.69007862e-01 1.64117441e-01 -3.48817170e-01
8.92487288e-01 -5.34325898e-01 -2.42368400e-01 2.03875527e-01
1.41083479e+00 5.53633630e-01 -4.46801156e-01 -1.48930028e-01
2.66305506e-02 -1.10249400e+00 1.85241312e-01 -4.44290750e-02
-8.79336238e-01 8.02427351e-01 9.05520558e-01 4.99802083e-01
1.26644921e+00 -6.45081878e-01 1.38958991e+00 -3.69254574e-02
4.28699315e-01 -1.23207033e+00 1.13561429e-01 -2.20875740e-01
9.72073555e-01 -2.99872607e-01 4.24466819e-01 -4.20387775e-01
-3.59161124e-02 1.38443935e+00 -3.40441853e-01 -2.48071223e-01
6.45185709e-01 4.63450551e-01 -1.50207117e-01 4.87841889e-02
-3.34642678e-01 5.00249088e-01 -2.38305092e-01 7.52037823e-01
1.05556585e-01 -1.24821998e-01 -3.89586657e-01 2.94282496e-01
-5.34623563e-01 -1.51337966e-01 7.32979774e-01 1.23861194e+00
-7.11885452e-01 -1.11785650e+00 -1.45978880e+00 -6.68107858e-03
-7.61100411e-01 -4.28323671e-02 -2.66484320e-01 4.26106215e-01
1.60278291e-01 1.25851786e+00 -1.74799651e-01 -5.78454733e-01
8.87834951e-02 2.55004704e-01 3.09881479e-01 5.68401814e-02
-9.38007832e-01 4.67343837e-01 -2.05489293e-01 -5.80396354e-01
-3.88367087e-01 -9.08821881e-01 -1.17711806e+00 -5.30160069e-01
-8.37507904e-01 5.60888499e-02 8.92922938e-01 6.81090355e-01
4.50980008e-01 1.52734697e-01 9.19782579e-01 -7.58469880e-01
-7.97936440e-01 -9.36150014e-01 -1.01572597e+00 5.58748603e-01
4.08620507e-01 -5.48582017e-01 -7.41473675e-01 3.13583851e-01]
|
[12.357894897460938, 0.9749763011932373]
|
7708bad2-cfbf-4abf-853d-d5d5d0d337d6
|
lifelong-multi-agent-path-finding-in-large
|
2005.07371
| null |
https://arxiv.org/abs/2005.07371v2
|
https://arxiv.org/pdf/2005.07371v2.pdf
|
Lifelong Multi-Agent Path Finding in Large-Scale Warehouses
|
Multi-Agent Path Finding (MAPF) is the problem of moving a team of agents to their goal locations without collisions. In this paper, we study the lifelong variant of MAPF, where agents are constantly engaged with new goal locations, such as in large-scale automated warehouses. We propose a new framework Rolling-Horizon Collision Resolution (RHCR) for solving lifelong MAPF by decomposing the problem into a sequence of Windowed MAPF instances, where a Windowed MAPF solver resolves collisions among the paths of the agents only within a bounded time horizon and ignores collisions beyond it. RHCR is particularly well suited to generating pliable plans that adapt to continually arriving new goal locations. We empirically evaluate RHCR with a variety of MAPF solvers and show that it can produce high-quality solutions for up to 1,000 agents (= 38.9\% of the empty cells on the map) for simulated warehouse instances, significantly outperforming existing work.
|
['Sven Koenig', 'Andrew Tinka', 'T. K. Satish Kumar', 'Scott Kiesel', 'Joseph W. Durham', 'Jiaoyang Li']
|
2020-05-15
| null | null | null | null |
['multi-agent-path-finding']
|
['playing-games']
|
[ 4.63492982e-02 4.53476667e-01 2.78123140e-01 1.88390203e-02
-8.71508360e-01 -1.09144187e+00 3.41998696e-01 5.10731518e-01
-4.23876673e-01 1.32582581e+00 -6.78701177e-02 -2.25672871e-01
-8.63748908e-01 -1.14657545e+00 -7.82237411e-01 -5.09618580e-01
-8.41053784e-01 1.50934112e+00 5.89587927e-01 -7.32708275e-01
1.16145507e-01 4.31296587e-01 -1.15964699e+00 1.46312490e-01
7.21660137e-01 4.14657295e-01 5.23298383e-01 9.02712047e-01
-6.47158846e-02 2.70217240e-01 -8.21776867e-01 -3.31613943e-02
5.49660802e-01 -2.04942361e-01 -9.33266282e-01 2.96167016e-01
-5.07086217e-01 -2.50714540e-01 1.14329837e-01 6.32489145e-01
1.74071386e-01 3.57776761e-01 1.99017242e-01 -2.11015034e+00
1.47033274e-01 7.98920453e-01 -8.12948585e-01 2.07636014e-01
7.49861956e-01 1.53506652e-01 6.65090084e-01 -2.55714774e-01
9.26846385e-01 1.28375137e+00 4.83301461e-01 2.46460050e-01
-1.16902328e+00 -8.36940780e-02 5.65229595e-01 4.46136715e-03
-1.29416227e+00 -1.21875361e-01 1.68448240e-02 -4.20211442e-02
1.55375004e+00 3.80630940e-01 5.00509560e-01 2.61351228e-01
6.98680103e-01 2.50045359e-01 8.19777966e-01 -2.04576492e-01
5.71965396e-01 -4.13018733e-01 1.13875559e-02 4.32325512e-01
4.55975890e-01 7.88384397e-03 -3.76132816e-01 -4.74635571e-01
6.18136823e-01 -2.57779390e-01 5.00421040e-02 -3.32531571e-01
-1.55829740e+00 9.46398795e-01 -5.19253965e-03 -2.12447509e-01
-7.06576169e-01 1.41769931e-01 9.34462175e-02 3.99966896e-01
-8.87652189e-02 7.41808176e-01 -4.94945377e-01 -2.78937221e-01
-2.82706529e-01 1.07200897e+00 1.08792639e+00 1.45467448e+00
6.29734814e-01 -2.52327740e-01 2.06269789e-02 1.18407957e-01
-1.33198678e-01 5.36971033e-01 -4.18237299e-01 -1.56453013e+00
8.08413327e-01 4.67174947e-01 1.07585192e+00 -7.70492256e-01
-9.82188106e-01 5.53200115e-03 -2.00123981e-01 5.53272247e-01
5.02607107e-01 -5.49392879e-01 -6.39965653e-01 1.51366973e+00
6.85059488e-01 5.49369045e-02 3.17831248e-01 7.23276913e-01
1.19778804e-01 1.16183341e+00 -4.01392788e-01 -8.65464330e-01
1.11660993e+00 -1.25821972e+00 -6.65829480e-01 -3.08070838e-01
6.07984006e-01 -6.16895974e-01 4.12909627e-01 6.43739939e-01
-1.76135182e+00 1.49646640e-01 -8.57801378e-01 5.65774620e-01
-3.07939678e-01 -9.14578378e-01 6.45476937e-01 2.44336829e-01
-1.31650543e+00 2.17698634e-01 -9.56238151e-01 -3.04136664e-01
-2.44596407e-01 5.71503878e-01 -3.18154067e-01 -5.28255105e-01
-7.41168201e-01 9.02916849e-01 3.88366163e-01 1.25738353e-01
-1.04440093e+00 -4.66278613e-01 -7.97508836e-01 8.59622210e-02
1.31053615e+00 -5.84689319e-01 1.55892897e+00 -5.65356836e-02
-1.18505132e+00 8.99422616e-02 -1.73066288e-01 -3.68410110e-01
4.81346518e-01 2.73055613e-01 -1.91408560e-01 -1.75832883e-01
6.00710928e-01 5.02725303e-01 1.97202295e-01 -1.51760590e+00
-1.25428677e+00 -2.01840326e-01 5.89675725e-01 4.73260611e-01
4.86141890e-01 1.09352851e-02 -6.79055825e-02 4.14719731e-02
-1.26933940e-02 -1.10248172e+00 -1.03006327e+00 -7.01771259e-01
-3.25304836e-01 -2.94382304e-01 3.46422970e-01 -2.60144589e-03
9.39255953e-01 -1.63232362e+00 4.52224195e-01 4.48822379e-01
8.24678317e-02 -4.14640337e-01 -5.73451281e-01 1.20898342e+00
3.80630195e-01 -1.89451367e-01 -1.02725245e-01 -1.66652337e-01
4.29838777e-01 6.86106086e-01 -1.48917660e-01 2.23574221e-01
-1.60713077e-01 8.38387668e-01 -1.09848237e+00 -7.12024942e-02
-1.99510023e-01 -3.25647205e-01 -5.36897302e-01 3.75495069e-02
-6.99501216e-01 -1.36982799e-01 -4.33338016e-01 4.56545055e-01
8.61166000e-01 1.04121342e-01 4.76469427e-01 7.72234797e-01
-6.41650617e-01 -1.32932857e-01 -1.64948797e+00 1.70974040e+00
-6.00815304e-02 -8.36565048e-02 6.49694502e-01 -2.93134421e-01
6.70494974e-01 -1.40482351e-01 6.69561625e-01 -7.45790005e-01
-1.60251752e-01 7.93673769e-02 -1.22309439e-01 -9.99737680e-02
9.02191460e-01 1.50643280e-02 -6.45853400e-01 9.44346249e-01
-7.23368049e-01 -2.17902258e-01 8.51084530e-01 2.91833878e-01
1.69253194e+00 -3.40715051e-01 1.31896242e-01 -2.30175510e-01
7.36452267e-02 9.30351436e-01 8.68895173e-01 1.13382494e+00
-2.63682544e-01 1.41825661e-01 6.25934243e-01 -7.91404247e-01
-8.04125965e-01 -1.26198804e+00 6.39456689e-01 1.02887762e+00
7.64202714e-01 -5.47301948e-01 -7.55177021e-01 -3.06421816e-01
1.46250725e-01 9.03641105e-01 -4.00189012e-01 3.67007375e-01
-9.63772893e-01 -7.83568323e-01 -1.62291035e-01 2.03861505e-01
2.10838273e-01 -1.06743789e+00 -1.16767573e+00 8.70362222e-01
-2.11972296e-01 -1.18938220e+00 -6.13221347e-01 1.94113657e-01
-3.34356070e-01 -1.35574245e+00 -2.95577973e-01 -7.41371930e-01
8.07770848e-01 7.30518520e-01 9.53023553e-01 -1.74154580e-01
-2.66095579e-01 3.89808267e-01 -5.14328659e-01 -4.24381852e-01
-3.28783482e-01 1.38364688e-01 -3.78588066e-02 -6.30129576e-01
-1.31120145e-01 -1.91751823e-01 -1.40651017e-01 6.02416813e-01
-6.56123042e-01 1.00252651e-01 9.70941931e-02 4.48539734e-01
7.63154387e-01 1.11541986e+00 7.67124951e-01 -6.16237700e-01
1.15273547e+00 -6.57898724e-01 -1.00494397e+00 4.13152128e-01
-2.86515236e-01 -3.35361332e-01 5.49834669e-01 -1.97236881e-01
-8.23149204e-01 -7.41981855e-03 4.85073507e-01 2.90466815e-01
-6.30665943e-02 6.04345262e-01 1.04984932e-01 -9.28068161e-03
2.30362415e-01 -4.03323732e-02 7.62278810e-02 1.26664713e-01
3.20866466e-01 -1.92127347e-01 5.06049216e-01 -8.49626005e-01
7.95128107e-01 2.56680518e-01 2.23332673e-01 -1.40452236e-01
-6.93311989e-02 -1.09912679e-01 -7.08260238e-02 -2.36223817e-01
4.63407785e-01 -3.07908922e-01 -1.33993161e+00 2.32969493e-01
-1.16310978e+00 -9.63637173e-01 -4.59718138e-01 -1.34697724e-02
-1.00585604e+00 -1.54981881e-01 -4.80522752e-01 -1.03954291e+00
2.09940717e-01 -1.17920637e+00 8.24405849e-01 3.10236901e-01
-1.74376681e-01 -6.67929471e-01 3.44699323e-01 4.44432795e-02
4.34402615e-01 7.94527411e-01 8.91166091e-01 -4.49805290e-01
-9.60902572e-01 1.02411859e-01 1.15519673e-01 -1.06744623e+00
-9.35321152e-02 -3.61698538e-01 2.42343783e-01 -6.71743095e-01
-3.41089100e-01 1.32115660e-02 -5.46895638e-02 4.87082988e-01
1.43918768e-01 -5.09550333e-01 -7.77450502e-01 -1.06437169e-01
1.40640736e+00 9.37774360e-01 4.54500228e-01 9.76897001e-01
-2.09418863e-01 7.09674060e-01 1.37363124e+00 9.21698153e-01
1.20847201e+00 8.71660292e-01 6.44405067e-01 2.23564506e-01
5.89377522e-01 2.35140264e-01 3.08029711e-01 1.57099441e-01
-1.10219851e-01 -8.08357835e-01 -1.16491985e+00 6.93129420e-01
-2.40586758e+00 -7.85712898e-01 -1.04477711e-01 1.92340434e+00
4.10173118e-01 3.56823057e-01 5.82407534e-01 2.15867255e-02
6.93104863e-01 -2.77967662e-01 -5.64875424e-01 -8.13114643e-01
9.61672589e-02 -1.67425752e-01 6.95074916e-01 1.09273422e+00
-7.55509615e-01 8.44803035e-01 6.66225576e+00 2.17765689e-01
-5.47165386e-02 4.10650596e-02 2.67747372e-01 -4.41851646e-01
-1.46402508e-01 -9.39856917e-02 -8.21923971e-01 8.39544013e-02
9.50435102e-01 -6.44334555e-01 1.07883739e+00 4.77376163e-01
5.15259147e-01 -5.75016916e-01 -9.46153104e-01 3.82237613e-01
-2.28923529e-01 -1.36447871e+00 -5.76743484e-01 2.88491607e-01
9.95897651e-01 -3.09362471e-01 -1.88338503e-01 2.53178477e-01
1.03737140e+00 -9.50194538e-01 6.79456711e-01 2.59449393e-01
6.19004406e-02 -1.37196505e+00 6.93071663e-01 6.97663248e-01
-1.44014907e+00 -4.41901565e-01 -1.58296809e-01 -4.67100233e-01
1.04130960e+00 6.99348375e-02 -1.27128649e+00 9.02203679e-01
7.10860133e-01 -1.55600771e-01 2.09553167e-01 1.27483058e+00
3.25319290e-01 -4.79330093e-01 -7.05505371e-01 2.76508811e-03
6.61368847e-01 -2.74311662e-01 7.52471745e-01 7.08988190e-01
4.41837132e-01 6.23156428e-01 8.88552487e-01 5.81294000e-01
7.48796225e-01 -4.37118232e-01 -5.36288381e-01 1.99389055e-01
6.88042283e-01 1.12582576e+00 -1.26757920e+00 8.55576396e-02
4.45768097e-03 5.56490123e-01 1.97190478e-01 4.56468821e-01
-1.03359985e+00 -3.98740262e-01 8.29032302e-01 1.46501973e-01
1.59546316e-01 -6.72002316e-01 -7.59704188e-02 -2.36284047e-01
-8.77316445e-02 -7.17089176e-01 6.30627155e-01 -8.22845638e-01
-8.29168260e-01 7.48782337e-01 3.44182372e-01 -6.20878994e-01
-5.10002494e-01 -2.04691023e-01 -6.48202956e-01 5.73949635e-01
-1.58685338e+00 -7.62534797e-01 -3.69999081e-01 5.92383444e-01
8.35185289e-01 -1.07882269e-01 8.33174586e-01 -4.39347848e-02
-5.30819237e-01 1.42934844e-02 -1.19191021e-01 -7.93226838e-01
5.69961257e-02 -1.19756150e+00 7.18867481e-01 7.84506559e-01
-6.60930693e-01 3.47742468e-01 9.93317664e-01 -9.17013168e-01
-1.93358886e+00 -1.01230502e+00 7.07873166e-01 -1.00897163e-01
5.63175678e-01 -2.63411134e-01 -4.34707522e-01 9.50085282e-01
3.45904112e-01 -4.19326097e-01 1.79175198e-01 -2.01344892e-01
4.27730113e-01 -1.44095123e-02 -1.36565888e+00 6.65912449e-01
1.20153081e+00 6.08810067e-01 -2.60502100e-01 5.84784091e-01
8.79677296e-01 -9.06221986e-01 -6.65007234e-01 2.39933133e-01
2.04423089e-02 -8.15452278e-01 7.99225152e-01 -5.41969240e-01
-1.08766206e-01 -6.45576417e-01 -9.54195037e-02 -1.60831869e+00
-4.97464597e-01 -1.20933187e+00 2.02609062e-01 9.12214637e-01
5.57870328e-01 -9.44881797e-01 9.33273852e-01 8.89916658e-01
-4.73856956e-01 -6.99940741e-01 -1.16666150e+00 -9.95016456e-01
-1.36911064e-01 1.24464929e-03 1.12877059e+00 5.88652790e-01
3.38112235e-01 -1.27171174e-01 -4.76021655e-02 6.62288964e-01
7.88116455e-01 3.20822328e-01 8.69082034e-01 -7.85599351e-01
-2.24098668e-01 -2.86611408e-01 2.37936899e-01 -6.37987673e-01
3.03389654e-02 -4.84985799e-01 3.70933145e-01 -2.08444595e+00
-2.80078556e-02 -8.44253540e-01 3.28258604e-01 6.66035891e-01
3.97517622e-01 -3.67158443e-01 4.92423177e-01 -2.17422336e-01
-1.14407742e+00 1.34675711e-01 1.44289708e+00 -3.82998176e-02
-5.39853096e-01 3.44013199e-02 -5.93355000e-01 3.38306129e-01
7.84772813e-01 -4.28164065e-01 -7.09891379e-01 -5.10971308e-01
6.16613686e-01 9.73789573e-01 -1.86362013e-01 -7.31043875e-01
7.29913473e-01 -1.04587758e+00 -3.71385604e-01 -7.95703888e-01
5.56625605e-01 -8.96719277e-01 8.65582287e-01 6.77074432e-01
-8.65884572e-02 1.08177137e+00 3.76610786e-01 4.87301320e-01
1.45000368e-01 -2.77530342e-01 6.72377795e-02 -3.94665152e-01
-7.70206928e-01 4.47799526e-02 -7.67740190e-01 1.26712233e-01
2.02694964e+00 -3.36284548e-01 -9.43356156e-01 -4.76082087e-01
-8.22888970e-01 1.23454928e+00 4.07879651e-01 1.60303235e-01
8.34873974e-01 -9.44588542e-01 -7.70792067e-01 6.07122742e-02
-2.67876565e-01 4.13796395e-01 2.66160458e-01 6.53202593e-01
-7.02389359e-01 4.43234086e-01 -3.72455597e-01 -1.43583909e-01
-1.03459430e+00 8.42191637e-01 1.59836382e-01 -5.32115817e-01
-7.32351661e-01 6.52970970e-01 1.15581986e-03 -3.42748106e-01
1.37576252e-01 -1.93745047e-01 5.76233715e-02 -1.12428702e-01
7.43509114e-01 8.46959233e-01 -1.91632614e-01 -1.72840059e-01
-5.47890186e-01 3.06280494e-01 -1.47874728e-01 -4.47273552e-01
1.53291142e+00 -3.35697144e-01 -2.70060569e-01 -5.94917051e-02
3.92691344e-02 -2.46150553e-01 -1.24529386e+00 2.16074988e-01
1.07454188e-01 -4.87060189e-01 -4.51970518e-01 -9.60940838e-01
-7.03434348e-01 -2.57693172e-01 -2.87998497e-01 7.51739323e-01
1.03863692e+00 -3.44401151e-02 9.13994968e-01 5.02954781e-01
1.37645662e+00 -1.02620387e+00 1.11706425e-02 6.92491770e-01
8.38627875e-01 -5.83778977e-01 -5.91818132e-02 -7.12212205e-01
-8.15388858e-01 9.32874143e-01 6.85469568e-01 -1.78482056e-01
-2.74221953e-02 9.59436953e-01 -1.98541105e-01 -2.83760518e-01
-1.19461274e+00 -2.20985115e-01 -9.22734678e-01 9.84216690e-01
-7.51030803e-01 2.96748668e-01 -3.00506026e-01 5.43302476e-01
-4.67914492e-01 -2.02924654e-01 1.32316136e+00 1.50430691e+00
-7.75103748e-01 -1.09633386e+00 -7.78249204e-01 9.45774391e-02
3.35480273e-01 5.52905440e-01 -2.20333099e-01 9.73832071e-01
1.09090041e-02 1.39701021e+00 3.19918573e-01 -3.13717946e-02
7.39445210e-01 -3.74586254e-01 6.17206752e-01 -5.94071329e-01
-6.40870512e-01 1.83408082e-01 7.08993375e-01 -8.17105889e-01
-4.87876981e-02 -9.26320553e-01 -1.80871594e+00 -7.31907368e-01
7.39421370e-03 5.25322795e-01 3.39532048e-01 7.96973050e-01
3.24459612e-01 7.05981374e-01 5.74361205e-01 -9.63357091e-01
-3.11574817e-01 -8.02278444e-02 -5.69042683e-01 -3.01526636e-01
8.97315443e-02 -7.54705966e-01 8.20347965e-02 -5.55521727e-01]
|
[4.962496757507324, 1.7456916570663452]
|
3c1f2991-1594-45b8-a487-eb3d36fe6495
|
skip-clip-self-supervised-spatiotemporal
|
1910.12770
| null |
https://arxiv.org/abs/1910.12770v1
|
https://arxiv.org/pdf/1910.12770v1.pdf
|
Skip-Clip: Self-Supervised Spatiotemporal Representation Learning by Future Clip Order Ranking
|
Deep neural networks require collecting and annotating large amounts of data to train successfully. In order to alleviate the annotation bottleneck, we propose a novel self-supervised representation learning approach for spatiotemporal features extracted from videos. We introduce Skip-Clip, a method that utilizes temporal coherence in videos, by training a deep model for future clip order ranking conditioned on a context clip as a surrogate objective for video future prediction. We show that features learned using our method are generalizable and transfer strongly to downstream tasks. For action recognition on the UCF101 dataset, we obtain 51.8% improvement over random initialization and outperform models initialized using inflated ImageNet parameters. Skip-Clip also achieves results competitive with state-of-the-art self-supervision methods.
|
['Graham W. Taylor', 'Alaaeldin El-Nouby', 'Joshua M. Susskind', 'Shuangfei Zhai']
|
2019-10-28
| null | null | null | null |
['self-supervised-action-recognition']
|
['computer-vision']
|
[ 3.51806939e-01 7.60517716e-02 -6.50822341e-01 -8.07540417e-01
-8.57446432e-01 -4.48799491e-01 5.47207057e-01 -2.38756880e-01
-6.35505736e-01 6.42562807e-01 9.20175731e-01 3.60466957e-01
1.43690497e-01 -1.77349553e-01 -1.11786497e+00 -3.16656947e-01
-4.92745310e-01 1.76636800e-01 3.39121580e-01 2.88188040e-01
-8.17012321e-03 1.89693078e-01 -1.49574757e+00 1.07209790e+00
3.37835908e-01 1.27162039e+00 1.53277963e-01 6.05056047e-01
6.25857890e-01 1.32466149e+00 -2.82226384e-01 -9.82737243e-02
3.74142170e-01 -3.56775939e-01 -8.92802656e-01 1.28950626e-01
9.29720819e-01 -9.53699589e-01 -1.07253182e+00 4.70919490e-01
1.05888486e-01 4.48308289e-01 4.48105365e-01 -1.14960110e+00
-5.72822511e-01 5.70866227e-01 -1.68003991e-01 6.65744007e-01
2.75099963e-01 3.92013490e-01 1.18048906e+00 -8.08762372e-01
1.10269046e+00 9.56778407e-01 6.34855211e-01 7.29973376e-01
-1.30745912e+00 -5.99240243e-01 5.08023322e-01 6.68820322e-01
-1.17158544e+00 -7.97421813e-01 5.17705917e-01 -6.04267359e-01
1.40976560e+00 -5.19285016e-02 7.23964512e-01 1.59988832e+00
1.21326270e-02 1.29857922e+00 5.37549138e-01 -3.97794321e-02
-1.27817262e-02 -4.68457103e-01 -3.50976456e-03 7.57262588e-01
-3.68022561e-01 1.30561292e-01 -9.44843829e-01 2.24593028e-01
6.46779060e-01 1.10965364e-01 -3.24205279e-01 -3.43794733e-01
-1.43557048e+00 5.48641562e-01 4.20374870e-01 1.87492445e-01
-2.40871072e-01 6.27872646e-01 6.87173307e-01 2.19033629e-01
5.74879050e-01 5.62482476e-01 -7.75246918e-01 -5.29180169e-01
-1.08816326e+00 1.37195840e-01 4.26565319e-01 9.08266187e-01
4.05509204e-01 1.17660277e-02 -5.92952073e-01 4.26310569e-01
2.00632084e-02 8.38143006e-02 7.37841070e-01 -1.67618287e+00
5.59845448e-01 2.07949549e-01 8.19406062e-02 -6.72266841e-01
-3.81733090e-01 -3.72411102e-01 -3.93314451e-01 -1.48288518e-01
2.45610595e-01 -6.91527128e-02 -9.35642242e-01 2.01233244e+00
-1.97129190e-01 7.92405486e-01 2.90502328e-02 8.99864852e-01
5.35716593e-01 7.18015969e-01 3.10966581e-01 -2.12406412e-01
7.03684628e-01 -1.53324461e+00 -3.52678865e-01 -3.35320421e-02
7.90078163e-01 -1.29876375e-01 9.32202876e-01 4.24781233e-01
-9.03220654e-01 -7.99701393e-01 -8.48123908e-01 -1.70009941e-01
8.02250430e-02 5.48845232e-01 7.36451268e-01 -2.59813100e-01
-1.18455803e+00 9.93291199e-01 -1.37412596e+00 -3.30337733e-01
8.61122787e-01 3.93189579e-01 -6.97231114e-01 7.01444894e-02
-1.01741588e+00 6.28922343e-01 4.43885207e-01 -6.77810684e-02
-1.59731793e+00 -8.31948817e-01 -8.69156361e-01 1.76686212e-01
3.02601784e-01 -4.05135036e-01 1.44094157e+00 -1.45008850e+00
-1.46978545e+00 8.73234808e-01 -2.35239342e-01 -1.12697136e+00
3.86205167e-01 -8.37529480e-01 -3.23517799e-01 7.14464903e-01
2.82910943e-01 1.22298563e+00 7.37593114e-01 -5.96180260e-01
-6.82534039e-01 -2.69480012e-02 1.52919397e-01 1.24627903e-01
-7.07952142e-01 -9.49846860e-03 -7.46869445e-01 -6.98756099e-01
-4.83183503e-01 -9.70513046e-01 -3.12881827e-01 1.55025527e-01
-1.22857772e-01 -4.09549862e-01 1.01235068e+00 -6.26048684e-01
1.20076072e+00 -2.20552969e+00 2.10510790e-01 -2.22793043e-01
7.67659545e-02 3.27198029e-01 -4.97023731e-01 1.90938756e-01
-1.59373909e-01 -8.43324587e-02 9.49680582e-02 -5.11940241e-01
-5.62676676e-02 2.87025303e-01 -5.02560973e-01 3.91135514e-01
5.68100274e-01 8.49748194e-01 -1.07614303e+00 -5.79524994e-01
1.82756245e-01 4.20417577e-01 -8.90258372e-01 3.62801343e-01
-5.35246015e-01 6.37968421e-01 -4.12586451e-01 5.23395777e-01
-1.38767779e-01 -5.88364542e-01 3.25149119e-01 -4.40569133e-01
8.96176323e-02 4.13228750e-01 -4.52418536e-01 2.31441259e+00
-1.78589046e-01 1.10933137e+00 -4.76998687e-01 -1.29065275e+00
4.18685257e-01 5.09955108e-01 9.48469758e-01 -5.19496441e-01
-1.36024982e-01 -2.99649775e-01 -3.03554922e-01 -8.62871587e-01
2.14170843e-01 3.18469673e-01 -7.16618299e-02 2.64273286e-01
7.39224970e-01 5.30060172e-01 5.16936660e-01 4.53202605e-01
1.59298038e+00 1.01693785e+00 3.96796241e-02 -5.49524762e-02
2.65462101e-01 5.24996556e-02 9.22125936e-01 6.38448715e-01
-5.17504632e-01 6.20550692e-01 5.95630050e-01 -8.29722524e-01
-9.69526052e-01 -8.98336530e-01 3.81883420e-02 1.50327694e+00
-1.79949000e-01 -7.25327373e-01 -5.17329812e-01 -1.03860962e+00
-1.98090691e-02 3.28113079e-01 -8.17299962e-01 -1.60955042e-01
-8.52261066e-01 -2.63628781e-01 4.19031292e-01 1.03111684e+00
4.74186778e-01 -1.03838003e+00 -6.56417131e-01 3.03087890e-01
-3.00303072e-01 -1.61299670e+00 -5.21937668e-01 1.97247453e-02
-9.93839622e-01 -9.82220769e-01 -4.80661958e-01 -7.07841456e-01
5.11217415e-01 1.19891558e-02 1.38335347e+00 -1.25831336e-01
-2.32772931e-01 5.30194104e-01 -4.05065209e-01 1.91417158e-01
3.78682613e-02 9.21045616e-02 1.82078779e-01 8.32880661e-02
4.10830826e-01 -7.12704718e-01 -9.19031382e-01 1.59983978e-01
-6.79050148e-01 2.12236866e-01 4.88865316e-01 8.05463910e-01
6.69131100e-01 -3.80771399e-01 4.52709943e-01 -7.14721203e-01
-1.62797138e-01 -5.94066620e-01 -3.66605788e-01 1.19358294e-01
-1.63713560e-01 1.50747105e-01 7.77322114e-01 -4.71483737e-01
-1.09779608e+00 7.55745113e-01 1.31065667e-01 -1.05742908e+00
-3.58839214e-01 4.48707342e-01 6.54234216e-02 1.99207529e-01
6.00223958e-01 -4.52983752e-03 -8.53318051e-02 -4.07670289e-01
4.34184790e-01 1.75362378e-01 8.81736934e-01 -4.58835512e-01
3.56518209e-01 6.72156632e-01 -7.45351315e-02 -3.51963937e-01
-1.49814475e+00 -4.40797269e-01 -1.04386282e+00 -3.94462824e-01
1.08254659e+00 -1.49368715e+00 -7.16449738e-01 1.55120477e-01
-9.08330023e-01 -9.21038926e-01 -2.90827125e-01 8.15846503e-01
-9.43061292e-01 1.34921178e-01 -9.12028909e-01 -3.68901581e-01
1.26358401e-02 -7.92652667e-01 1.08821845e+00 2.24777088e-02
-3.24414134e-01 -7.79062212e-01 1.77517667e-01 5.01470089e-01
1.41876802e-01 3.04175079e-01 1.61404699e-01 -7.27362871e-01
-9.04107213e-01 -2.89069896e-04 -1.22956626e-01 5.62418818e-01
-1.19918317e-01 -6.84274035e-03 -8.92015457e-01 -2.95230567e-01
-3.82616669e-01 -1.05854082e+00 1.52663255e+00 3.64112347e-01
1.73971331e+00 -4.35250193e-01 -4.98344868e-01 7.92720795e-01
1.14125168e+00 -1.95483826e-02 5.93337774e-01 3.13738346e-01
6.26167893e-01 2.70762920e-01 8.08596253e-01 6.51428580e-01
3.82098109e-01 7.38727808e-01 4.03445214e-01 2.21045688e-01
-9.43549871e-02 -4.23395663e-01 6.58435762e-01 3.67412984e-01
-2.59954780e-01 -2.19500557e-01 -6.63036048e-01 7.65000343e-01
-2.16271424e+00 -1.36948800e+00 4.57762718e-01 1.71573174e+00
9.62273717e-01 2.55574614e-01 2.20550835e-01 -3.43112051e-01
3.66815716e-01 4.16725844e-01 -7.59101510e-01 2.33516157e-01
-6.00084774e-02 1.70629799e-01 4.80789542e-01 2.54013479e-01
-1.81032979e+00 1.20879233e+00 6.84826279e+00 4.82664436e-01
-1.09191751e+00 1.55616432e-01 8.00053596e-01 -6.21296942e-01
2.46944204e-01 1.88495517e-02 -7.18162417e-01 4.59010810e-01
1.15826809e+00 1.49405345e-01 1.47195190e-01 9.75034475e-01
3.92153561e-01 1.21959381e-01 -1.55565715e+00 9.21088934e-01
3.64977032e-01 -1.66715240e+00 -2.18029283e-02 -2.40553871e-01
1.03147721e+00 5.40978849e-01 6.96900487e-02 4.49819446e-01
3.85029465e-01 -9.79405284e-01 5.82086504e-01 8.52690935e-01
7.21064925e-01 -3.01623732e-01 2.93087542e-01 -5.33405393e-02
-1.08063829e+00 -3.65111083e-01 -2.95758188e-01 -3.36710960e-01
2.71790862e-01 2.22289339e-01 -6.84776723e-01 1.71167981e-02
7.77630389e-01 1.65774620e+00 -6.35862887e-01 9.51696515e-01
-2.39734858e-01 9.94783938e-01 -1.47488773e-01 4.25115466e-01
4.57354426e-01 4.15826976e-01 3.17917585e-01 1.35856926e+00
1.28826424e-01 1.83821633e-01 5.05680978e-01 4.79366183e-01
-3.55166227e-01 -3.18208605e-01 -6.63946629e-01 -3.33977729e-01
1.03080995e-01 1.08070040e+00 -3.97421718e-01 -6.87680483e-01
-5.41293800e-01 1.29578006e+00 6.03589296e-01 5.05301595e-01
-9.66628313e-01 2.00452525e-02 7.95908988e-01 -1.40227541e-01
7.36551762e-01 -3.43934983e-01 2.62292355e-01 -1.62092340e+00
1.11264817e-01 -6.03424966e-01 5.37899673e-01 -7.79093444e-01
-1.05065787e+00 5.69757044e-01 1.86907854e-02 -1.57724786e+00
-6.53469741e-01 -7.29632020e-01 -5.66488147e-01 -3.84613052e-02
-1.37993431e+00 -1.26398659e+00 -2.91689575e-01 7.39968896e-01
8.41550589e-01 -3.32111776e-01 7.28721440e-01 3.02611768e-01
-5.40582120e-01 5.26424468e-01 4.09051143e-02 4.45989341e-01
8.37753594e-01 -1.03938556e+00 3.19394320e-01 8.67092431e-01
5.85252881e-01 2.86006510e-01 4.99725103e-01 -4.03925866e-01
-1.20565093e+00 -1.39834225e+00 7.17621148e-01 -6.48843169e-01
8.66716802e-01 -1.91452503e-01 -5.51704586e-01 1.25990927e+00
2.85704792e-01 5.98600030e-01 7.91382134e-01 6.84258491e-02
-6.03673577e-01 -3.21523905e-01 -5.62318504e-01 4.05035257e-01
1.55252159e+00 -6.81989014e-01 -6.24662399e-01 7.06646860e-01
8.76143098e-01 -2.74757683e-01 -1.04067874e+00 5.71790576e-01
7.30492890e-01 -8.57686818e-01 8.60456944e-01 -1.25267160e+00
8.42889786e-01 2.10930258e-02 -1.25338972e-01 -9.75383639e-01
-5.08116186e-01 -7.60676384e-01 -5.29131174e-01 7.63670444e-01
4.75124478e-01 1.40589133e-01 1.14715505e+00 6.71343029e-01
-4.05343115e-01 -7.38226354e-01 -7.82829463e-01 -8.19922328e-01
-2.49047890e-01 -3.74480188e-01 5.20226061e-02 8.23170543e-01
2.39001244e-01 3.31493914e-01 -7.14048922e-01 -6.29120469e-02
4.56573278e-01 8.93192068e-02 6.48170471e-01 -8.96750629e-01
-5.96375823e-01 -2.09472507e-01 -6.85259521e-01 -1.44524610e+00
7.05549777e-01 -7.13971913e-01 1.46369431e-02 -1.26708817e+00
3.63825858e-01 -1.13212352e-03 -8.38679731e-01 7.64313042e-01
8.00736696e-02 6.53975904e-01 3.81510407e-01 4.08106923e-01
-1.65753758e+00 6.57677472e-01 9.45692062e-01 -1.29293069e-01
6.18657619e-02 -2.07327679e-01 -2.29884163e-01 8.42855096e-01
6.68152153e-01 -2.59528697e-01 -5.84475815e-01 -8.50507736e-01
-3.08048613e-02 1.72827810e-01 4.76009637e-01 -1.40428448e+00
2.15486616e-01 -1.40469402e-01 6.97947145e-01 -5.33169448e-01
6.22784734e-01 -6.41278386e-01 -2.95846909e-01 1.98945761e-01
-1.08630133e+00 -1.39701039e-01 1.12854183e-01 8.56013596e-01
-4.27909493e-01 5.50306216e-02 3.73106658e-01 -1.71165138e-01
-1.15078104e+00 5.74029088e-01 -4.31272238e-01 1.85502708e-01
1.10590589e+00 2.26992428e-01 -3.24024826e-01 -5.84544480e-01
-1.28018284e+00 3.95719498e-01 2.56255358e-01 5.13859332e-01
6.00520253e-01 -1.50529552e+00 -5.48767269e-01 -9.04914830e-03
1.21634752e-01 -4.13074225e-01 2.28793442e-01 7.73903072e-01
-2.75520414e-01 7.37278640e-01 -3.63212317e-01 -5.66413522e-01
-1.14220548e+00 5.36310077e-01 1.26572624e-01 -4.17119324e-01
-6.71889126e-01 1.04341257e+00 3.39674324e-01 1.09187171e-01
5.56401014e-01 -4.78659838e-01 -2.01916501e-01 -2.36152321e-01
5.91882110e-01 -7.84082338e-02 -2.96669424e-01 -6.10548437e-01
-3.79461974e-01 1.48524553e-01 -2.32911572e-01 -1.70656905e-01
1.58896577e+00 7.43884295e-02 3.12016815e-01 3.59403789e-01
1.66960180e+00 -5.53136051e-01 -2.19956040e+00 -2.96184868e-01
9.62756798e-02 -5.21066546e-01 -2.32453234e-02 -7.66526878e-01
-1.22267449e+00 6.59525096e-01 5.05845845e-01 -3.22542906e-01
1.20001686e+00 1.29107475e-01 8.13645542e-01 9.41337228e-01
1.89648792e-01 -1.32744741e+00 6.17752433e-01 6.00474000e-01
8.84384036e-01 -1.42457223e+00 -1.39261410e-01 9.12024677e-02
-9.59178925e-01 1.16126013e+00 1.01729059e+00 -5.36754727e-01
7.20036685e-01 7.67840222e-02 -2.22604871e-01 -9.79690179e-02
-1.42868805e+00 -1.89502746e-01 4.15921301e-01 3.96950334e-01
5.64231217e-01 -2.27095291e-01 -3.04716062e-02 5.68454504e-01
2.56290674e-01 6.08655989e-01 2.05652639e-01 9.66716826e-01
-2.38046333e-01 -8.69675517e-01 3.76962155e-01 5.45997679e-01
-6.19944394e-01 -8.58741254e-02 -1.57053858e-01 4.69565898e-01
1.26098812e-01 5.85805655e-01 3.94974470e-01 -5.92274070e-01
-6.18148502e-03 8.48888978e-02 3.67860347e-01 -5.78571737e-01
-2.47490540e-01 5.49072251e-02 3.84860069e-01 -1.37377667e+00
-8.91801953e-01 -8.04270089e-01 -1.13635540e+00 3.72253299e-01
1.75879985e-01 -5.83179407e-02 1.24652699e-01 9.79991794e-01
8.17689538e-01 4.06556875e-01 5.68213463e-01 -1.14143884e+00
-3.11651081e-01 -9.89406943e-01 -8.05905610e-02 7.69186676e-01
2.93629080e-01 -7.82162666e-01 -2.33140856e-01 7.55925357e-01]
|
[8.45142936706543, 0.6389753818511963]
|
33ed38cf-c875-4843-b7ac-d001795686b2
|
the-theory-behind-controllable-expressive
|
1910.06234
| null |
https://arxiv.org/abs/1910.06234v1
|
https://arxiv.org/pdf/1910.06234v1.pdf
|
The Theory behind Controllable Expressive Speech Synthesis: a Cross-disciplinary Approach
|
As part of the Human-Computer Interaction field, Expressive speech synthesis is a very rich domain as it requires knowledge in areas such as machine learning, signal processing, sociology, psychology. In this Chapter, we will focus mostly on the technical side. From the recording of expressive speech to its modeling, the reader will have an overview of the main paradigms used in this field, through some of the most prominent systems and methods. We explain how speech can be represented and encoded with audio features. We present a history of the main methods of Text-to-Speech synthesis: concatenative, parametric and statistical parametric speech synthesis. Finally, we focus on the last one, with the last techniques modeling Text-to-Speech synthesis as a sequence-to-sequence problem. This enables the use of Deep Learning blocks such as Convolutional and Recurrent Neural Networks as well as Attention Mechanism. The last part of the Chapter intends to assemble the different aspects of the theory and summarize the concepts.
|
['Thierry Dutoit', 'Noé Tits', 'Kevin El Haddad']
|
2019-10-14
| null | null | null | null |
['expressive-speech-synthesis']
|
['speech']
|
[ 4.35820371e-01 4.39204782e-01 -2.38962501e-01 -1.27501339e-01
-6.42007113e-01 -4.00320798e-01 8.27839315e-01 -3.82499307e-01
2.60001346e-02 7.01974988e-01 6.57618523e-01 -2.36548916e-01
4.43638302e-02 -4.39859271e-01 -5.28041422e-01 -6.26551867e-01
-9.62179061e-03 2.09638864e-01 -9.96351466e-02 -5.67471087e-01
1.57672301e-01 7.77141154e-01 -1.90624344e+00 5.58248103e-01
5.46996236e-01 9.90973353e-01 3.14973414e-01 1.13613212e+00
-4.52767462e-01 1.02425432e+00 -1.05183578e+00 -2.60990828e-01
-3.55972588e-01 -6.37856305e-01 -8.03371608e-01 1.17219286e-03
-1.03669770e-01 -2.22792223e-01 -5.58981299e-01 8.12207460e-01
8.50079000e-01 3.89758199e-01 8.65461528e-01 -1.22613430e+00
-6.15560055e-01 1.11993146e+00 2.68656880e-01 1.29485905e-01
7.35562384e-01 -1.20348826e-01 8.52723479e-01 -1.02038276e+00
4.52484369e-01 1.39356625e+00 3.75828415e-01 7.51103520e-01
-7.65955091e-01 -6.01722717e-01 -7.66240284e-02 2.46935442e-01
-1.08199787e+00 -9.70353723e-01 1.05237293e+00 -4.62961078e-01
1.35220373e+00 4.52649176e-01 7.51576245e-01 1.53061223e+00
-5.88154346e-02 1.22599101e+00 8.63994658e-01 -8.41456354e-01
-1.19406424e-01 1.88735709e-01 -1.41770914e-01 3.93233865e-01
-7.97862947e-01 3.67602587e-01 -7.09142506e-01 2.62248129e-01
7.60744512e-01 -5.58003783e-01 -2.04070851e-01 2.15550795e-01
-1.19660127e+00 8.04770947e-01 -9.08684824e-03 8.57279539e-01
-3.19401711e-01 1.58516899e-01 6.82656288e-01 4.06250387e-01
3.34072500e-01 3.06978017e-01 -1.72249019e-01 -4.30908114e-01
-1.17991436e+00 3.18601638e-01 9.04942751e-01 1.13745546e+00
4.30135466e-02 9.79215324e-01 -2.14875787e-01 1.21884251e+00
1.83483884e-01 5.14471710e-01 8.19000661e-01 -9.12230194e-01
2.63513535e-01 -2.93581158e-01 -1.75811842e-01 -6.39371336e-01
-4.99652565e-01 -2.05321684e-01 -9.51799214e-01 -6.46770597e-02
-1.34430438e-01 -3.55854422e-01 -5.27411103e-01 1.52998090e+00
-7.25442991e-02 3.01537365e-02 4.54402342e-03 4.80631113e-01
1.52800810e+00 1.22246492e+00 -2.60511786e-02 -7.07243562e-01
1.19026554e+00 -1.23833668e+00 -1.53495586e+00 1.33822626e-02
2.87754297e-01 -1.01861620e+00 9.46672022e-01 2.23463416e-01
-1.54411197e+00 -7.54127800e-01 -9.54475880e-01 -3.93365145e-01
-6.48786843e-01 3.96060169e-01 3.97321045e-01 6.21562898e-01
-1.06576598e+00 6.47001863e-01 -5.93387187e-01 -5.68137407e-01
-4.09393162e-02 3.57282221e-01 -2.52825022e-01 8.19360614e-01
-1.44096005e+00 9.63450909e-01 2.95848072e-01 -1.29789248e-01
-6.61713839e-01 -5.55329204e-01 -9.30206835e-01 1.76260829e-01
-3.14546376e-02 -5.87355554e-01 1.85013890e+00 -1.21436453e+00
-2.45409322e+00 8.14159214e-01 -3.00360024e-01 -6.19927347e-01
2.86549240e-01 -2.92635322e-01 -6.96017444e-01 9.06800702e-02
-4.15098339e-01 6.15911841e-01 1.02901065e+00 -9.19873118e-01
-5.87179303e-01 1.10914305e-01 -3.15859258e-01 9.55470800e-02
-3.30462791e-02 6.51799202e-01 8.41753185e-03 -1.10537791e+00
-3.22835624e-01 -6.82865560e-01 2.18508616e-01 -3.60665470e-01
-4.13581133e-01 -5.31782448e-01 7.37675190e-01 -6.55180454e-01
1.61796093e+00 -2.02764726e+00 4.34129119e-01 -2.76016355e-01
-1.05677955e-01 4.95664507e-01 3.81605625e-02 1.16410947e+00
-5.11770904e-01 1.90383419e-02 -5.41146919e-02 -6.15980685e-01
1.87030643e-01 7.12722540e-02 -8.46548140e-01 2.17487648e-01
1.32150903e-01 1.13440609e+00 -6.25082672e-01 -3.02242637e-01
5.58315873e-01 6.28981471e-01 -1.01006910e-01 3.79296124e-01
-1.05395414e-01 5.60442924e-01 -1.24854967e-01 5.06230950e-01
1.15896299e-01 3.13650012e-01 -8.79351348e-02 -9.93791521e-02
-5.13799787e-01 7.68202424e-01 -7.64418960e-01 1.55893862e+00
-8.38639259e-01 1.07482791e+00 2.23766446e-01 -1.40582287e+00
8.40500951e-01 9.56511199e-01 3.59300762e-01 -4.92356360e-01
3.83551449e-01 4.96216863e-01 -1.92393120e-02 -8.24018836e-01
6.69619501e-01 -2.51099676e-01 -5.76698743e-02 3.38382602e-01
4.02973324e-01 -4.69210714e-01 3.75763290e-02 -2.23400816e-01
5.87195516e-01 9.05378140e-04 8.02294850e-01 1.30433831e-02
7.90338576e-01 -5.46501875e-01 -1.12111755e-01 4.09061968e-01
-2.94111222e-02 5.36430418e-01 3.81244421e-01 -2.62380809e-01
-1.19087517e+00 -7.95840442e-01 -7.87432268e-02 1.46216846e+00
-6.10179663e-01 -2.96751559e-01 -9.85000312e-01 -3.81345786e-02
-4.49940473e-01 9.04274523e-01 -5.88790119e-01 -5.42154573e-02
-6.72292590e-01 -1.10783599e-01 8.54370594e-01 5.98747075e-01
2.62774616e-01 -1.85256517e+00 -1.98335439e-01 3.90861988e-01
-3.74536484e-01 -1.10782063e+00 -2.99853623e-01 2.65169293e-01
-7.27199018e-01 -1.94022119e-01 -1.04812181e+00 -1.05484498e+00
-1.87931016e-01 -1.52359277e-01 1.04979980e+00 -2.16522619e-01
-5.99946156e-02 2.53439873e-01 -4.88072217e-01 -6.64265215e-01
-1.01373959e+00 3.25569093e-01 1.08342387e-01 -1.19606517e-01
1.64125204e-01 -9.06716704e-01 1.24484990e-02 -2.61009723e-01
-7.83415496e-01 7.32756630e-02 5.94576299e-01 8.45634639e-01
1.70100644e-01 -4.21314150e-01 7.99330890e-01 -6.48730993e-01
1.03730452e+00 -2.05655947e-01 -1.91163167e-01 1.25470450e-02
2.00510606e-01 -1.23043336e-01 7.81556368e-01 -5.60209632e-01
-8.07031929e-01 1.42148912e-01 -8.72732818e-01 -3.85010898e-01
-3.66860211e-01 4.66977477e-01 -2.47872397e-01 2.57326305e-01
7.21817017e-01 6.70785785e-01 -1.09726787e-02 -3.99560302e-01
5.77404797e-01 1.30053067e+00 5.06649911e-01 -3.03692430e-01
3.34580481e-01 1.94245521e-02 -2.30906919e-01 -1.58072209e+00
-4.45868075e-01 -3.17037493e-01 -6.71865106e-01 -3.15248191e-01
7.30389595e-01 -5.46334624e-01 -7.57688403e-01 6.83028936e-01
-1.63033748e+00 -5.25443673e-01 -5.63894629e-01 4.66478974e-01
-1.23852289e+00 1.39344439e-01 -8.62414896e-01 -1.13226688e+00
-5.29608667e-01 -1.27202284e+00 1.32890415e+00 3.99149656e-02
-5.02899647e-01 -1.01066792e+00 -3.55925709e-02 1.76212698e-01
4.00329202e-01 -2.89211422e-02 8.92555654e-01 -7.91002870e-01
-3.74076962e-02 -1.16714641e-01 3.09672236e-01 4.27027166e-01
1.22290634e-01 2.71482021e-01 -1.50596333e+00 1.34523943e-01
-5.55926301e-02 -2.35123113e-01 5.85420251e-01 6.18790209e-01
1.19591975e+00 -4.06769186e-01 -2.03570127e-01 5.54762542e-01
5.46831548e-01 6.41562998e-01 6.55515015e-01 -2.05687389e-01
3.93228501e-01 9.28480625e-01 2.09804788e-01 4.57263470e-01
7.29342643e-03 7.70677209e-01 -5.86424991e-02 4.97127883e-02
-5.15522718e-01 -3.47949982e-01 5.56147277e-01 1.62498534e+00
-1.01098023e-01 -5.12505710e-01 -4.83185351e-01 4.59695220e-01
-1.56835258e+00 -1.29488540e+00 1.19645745e-01 1.88251197e+00
8.79920661e-01 -1.87039480e-01 4.12527591e-01 7.90560186e-01
7.43729413e-01 2.66791314e-01 -8.93655345e-02 -1.07900453e+00
-1.61689058e-01 4.50116962e-01 -1.70312777e-01 7.70011604e-01
-1.19706881e+00 1.30803275e+00 7.81296635e+00 1.14123631e+00
-1.33906674e+00 -1.03839755e-01 2.88639843e-01 9.96838585e-02
1.03631523e-02 -6.81675136e-01 -6.74214423e-01 4.22409028e-01
1.58533120e+00 -2.15596467e-01 8.06608677e-01 7.30815411e-01
5.67109704e-01 2.99093336e-01 -1.03055894e+00 1.10724926e+00
1.04372561e-01 -1.41094112e+00 1.17961667e-01 -2.62395531e-01
4.94579405e-01 -1.49788305e-01 2.56416857e-01 4.81194913e-01
-1.75334379e-01 -1.33510125e+00 9.55213249e-01 4.16308582e-01
1.06845617e+00 -8.33332777e-01 3.10631633e-01 4.43342090e-01
-1.24261129e+00 -1.51432782e-01 -5.17564714e-02 -7.75038376e-02
3.15444589e-01 2.23956376e-01 -5.87385595e-01 3.72410059e-01
4.29594636e-01 6.74512744e-01 2.87034005e-01 7.82356024e-01
-2.81779736e-01 6.08234167e-01 -1.70871273e-01 -4.91542011e-01
1.97918952e-01 2.05783211e-02 7.25563347e-01 1.69148099e+00
4.54724848e-01 1.55325294e-01 -2.02248976e-01 7.70061731e-01
2.01128218e-02 4.79751974e-01 -9.28125501e-01 -6.22826934e-01
3.92600387e-01 9.89827752e-01 -3.59422714e-01 -4.91048038e-01
-4.79270399e-01 7.40961015e-01 -2.21176594e-02 3.19833457e-01
-6.70958221e-01 -8.59813213e-01 6.23979390e-01 1.52810559e-01
1.40512705e-01 -3.18787992e-01 -2.54342407e-01 -8.02516878e-01
-3.95632297e-01 -1.14539802e+00 -2.82847762e-01 -1.07363236e+00
-9.22298968e-01 8.97334814e-01 1.74033225e-01 -1.04419494e+00
-1.00151789e+00 -6.04953408e-01 -5.78878105e-01 9.06222701e-01
-1.16326487e+00 -1.07694483e+00 -4.31263680e-03 4.04819876e-01
1.18004215e+00 -4.33803767e-01 1.15466213e+00 3.62488717e-01
-3.15615416e-01 5.85126579e-01 1.17346786e-01 1.97580829e-01
4.38262403e-01 -1.04275095e+00 6.31174803e-01 4.07918572e-01
2.66722828e-01 5.36596954e-01 9.36319649e-01 -1.85430393e-01
-9.96394515e-01 -6.55177355e-01 1.35892439e+00 -2.01865919e-02
8.42615545e-01 -5.28881729e-01 -5.95060587e-01 6.85222387e-01
6.73166454e-01 -3.17535102e-01 6.52682960e-01 -2.86252260e-01
9.92130339e-02 5.35065234e-02 -7.17756331e-01 8.40799391e-01
9.10864532e-01 -8.68843377e-01 -5.74390829e-01 4.90328074e-01
9.52928483e-01 -6.29051030e-01 -7.71179020e-01 3.88447531e-02
6.16581678e-01 -8.76125038e-01 9.29672062e-01 -6.49845958e-01
3.76393020e-01 3.22773457e-02 -1.15315750e-01 -1.36632729e+00
-3.11687887e-01 -1.32268083e+00 -2.13319778e-01 1.26339686e+00
3.35010827e-01 -4.18462634e-01 2.53210217e-01 -4.99938130e-02
-7.12022722e-01 -6.41605794e-01 -1.01876736e+00 -5.76234162e-01
3.06435227e-01 -6.59178734e-01 6.13178313e-01 5.58641970e-01
3.27364177e-01 5.73907375e-01 -6.45468950e-01 -2.91141748e-01
-1.38073832e-01 -1.22232765e-01 6.11077726e-01 -1.06864977e+00
-4.27347496e-02 -7.99294829e-01 -3.72594088e-01 -1.31511843e+00
5.74147105e-01 -7.93868423e-01 1.42664269e-01 -1.48215878e+00
-5.39408445e-01 3.25222224e-01 2.66001314e-01 1.41620427e-01
4.29125100e-01 7.66897690e-04 2.05469728e-01 -1.82510063e-01
6.33380637e-02 7.53965676e-01 1.26340699e+00 -1.33713037e-01
-3.42421591e-01 4.92765844e-01 -2.94409841e-01 5.97818136e-01
8.72829199e-01 -1.18387878e-01 -4.61139709e-01 1.80433842e-03
4.91842143e-02 3.92624199e-01 2.28671636e-02 -9.11887825e-01
2.68329531e-01 2.46548168e-02 4.14125770e-02 -6.46876574e-01
8.99635911e-01 -6.52992785e-01 -2.09879242e-02 2.45253503e-01
-6.90007269e-01 -1.33602750e-02 2.39268124e-01 9.83348563e-02
-5.93379855e-01 -4.81735110e-01 9.03494358e-01 -1.16599157e-01
-5.04606068e-01 -5.97063126e-03 -1.01633298e+00 -4.78285365e-02
8.38125944e-01 -1.88427776e-01 3.85092683e-02 -8.43965828e-01
-1.05225074e+00 -3.30674201e-01 -3.69926065e-01 6.06479049e-01
5.39200604e-01 -1.35528314e+00 -6.56250715e-01 3.64664555e-01
-1.73529297e-01 -5.90722382e-01 2.83817440e-01 7.72202730e-01
-4.61902887e-01 1.07177579e+00 -5.11419326e-02 -3.48116726e-01
-1.41839695e+00 6.91711187e-01 4.23980296e-01 1.70266524e-01
-4.63843524e-01 7.26235926e-01 9.52626020e-02 -4.20765549e-01
6.22794569e-01 -5.82483709e-01 -6.26048625e-01 1.42028242e-01
6.15450501e-01 4.42879021e-01 9.46103483e-02 -1.09648597e+00
-1.94321424e-01 5.71609259e-01 3.85114253e-01 -4.99776423e-01
1.09489322e+00 -1.63060620e-01 -7.01501295e-02 9.72732008e-01
1.30228639e+00 -9.76432860e-02 -6.30846739e-01 -5.72405159e-02
-1.22849233e-01 1.71192482e-01 1.17619336e-01 -6.10177517e-01
-8.59998524e-01 1.50694835e+00 1.67624146e-01 6.62825048e-01
1.10881507e+00 -8.84946343e-03 8.53957117e-01 3.13067794e-01
-8.18341151e-02 -1.20327973e+00 -3.55042890e-02 8.94368052e-01
1.48571336e+00 -7.40325630e-01 -3.86713296e-01 -3.08546156e-01
-6.96727693e-01 1.43236744e+00 -9.70755816e-02 -1.46420032e-01
9.18531597e-01 6.67311370e-01 -2.47340333e-02 3.67918998e-01
-8.84913087e-01 -2.19202787e-01 5.46263874e-01 8.85167420e-01
1.02176118e+00 4.86643463e-02 -1.18179761e-01 7.15154231e-01
-1.02108157e+00 5.83917275e-02 3.19144726e-01 4.94506866e-01
-6.00648165e-01 -1.19708204e+00 -3.50040734e-01 6.52508736e-02
-6.78268969e-01 -1.63533404e-01 -4.93299752e-01 6.51805460e-01
-1.58806816e-01 1.13928258e+00 1.76538959e-01 -3.44719201e-01
3.94709140e-01 4.24041122e-01 5.31260371e-01 -5.69437623e-01
-8.80863428e-01 4.30774927e-01 2.55948544e-01 -3.08374763e-01
-5.42962432e-01 -5.48307002e-01 -1.18341172e+00 -5.21326125e-01
-1.71734825e-01 2.87851654e-02 8.07436347e-01 9.54282939e-01
3.52732763e-02 1.00891137e+00 5.04027605e-01 -1.23063660e+00
-3.56825262e-01 -1.35298216e+00 -7.43921936e-01 -3.06136727e-01
6.13053560e-01 -4.44216430e-01 -3.09915751e-01 2.46703297e-01]
|
[14.807310104370117, 6.600111484527588]
|
45b93104-f275-46be-91f2-a973735b9f8a
|
3d-hierarchical-refinement-and-augmentation
|
2112.03045
| null |
https://arxiv.org/abs/2112.03045v2
|
https://arxiv.org/pdf/2112.03045v2.pdf
|
3D Hierarchical Refinement and Augmentation for Unsupervised Learning of Depth and Pose from Monocular Video
|
Depth and ego-motion estimations are essential for the localization and navigation of autonomous robots and autonomous driving. Recent studies make it possible to learn the per-pixel depth and ego-motion from the unlabeled monocular video. A novel unsupervised training framework is proposed with 3D hierarchical refinement and augmentation using explicit 3D geometry. In this framework, the depth and pose estimations are hierarchically and mutually coupled to refine the estimated pose layer by layer. The intermediate view image is proposed and synthesized by warping the pixels in an image with the estimated depth and coarse pose. Then, the residual pose transformation can be estimated from the new view image and the image of the adjacent frame to refine the coarse pose. The iterative refinement is implemented in a differentiable manner in this paper, making the whole framework optimized uniformly. Meanwhile, a new image augmentation method is proposed for the pose estimation by synthesizing a new view image, which creatively augments the pose in 3D space but gets a new augmented 2D image. The experiments on KITTI demonstrate that our depth estimation achieves state-of-the-art performance and even surpasses recent approaches that utilize other auxiliary tasks. Our visual odometry outperforms all recent unsupervised monocular learning-based methods and achieves competitive performance to the geometry-based method, ORB-SLAM2 with back-end optimization.
|
['Hesheng Wang', 'Zhe Liu', 'Wenhua Wu', 'Shijie Zhao', 'Jiquan Zhong', 'Guangming Wang']
|
2021-12-06
| null | null | null | null |
['image-augmentation']
|
['computer-vision']
|
[ 1.21100329e-01 1.65828526e-01 -2.97689825e-01 -4.71119523e-01
-3.83408606e-01 -3.39699686e-01 4.88373190e-01 -5.98664105e-01
-6.01505637e-01 7.71519840e-01 -8.96414518e-02 2.98004821e-02
3.07983696e-01 -7.78433740e-01 -8.85083854e-01 -9.39701617e-01
4.64303911e-01 6.51202857e-01 3.49732161e-01 -1.09412819e-01
4.19273853e-01 3.90494525e-01 -1.74238455e+00 -4.88138765e-01
8.25375676e-01 9.12743807e-01 6.92225575e-01 5.69532752e-01
-5.16599976e-02 6.30145967e-01 -1.43367029e-03 1.64248288e-01
4.79767472e-01 -2.00266451e-01 -4.28063333e-01 6.74693108e-01
7.24422693e-01 -6.36200070e-01 -5.70889711e-01 1.19841695e+00
2.13638678e-01 1.01030603e-01 2.84968644e-01 -1.07999551e+00
-2.13689357e-01 -7.23787919e-02 -8.34477365e-01 -1.83674648e-01
4.14531499e-01 1.53675914e-01 5.77587485e-01 -1.07740510e+00
1.02320647e+00 1.24251032e+00 2.54660159e-01 4.16089416e-01
-9.06512201e-01 -7.77096570e-01 4.49303269e-01 3.53002965e-01
-1.45850515e+00 -2.19292358e-01 8.57615829e-01 -3.82140726e-01
7.19124973e-01 -3.93479556e-01 7.86350369e-01 5.47525227e-01
3.74988556e-01 6.17613852e-01 1.00169742e+00 -2.44176507e-01
6.62322044e-02 -1.90878510e-01 -3.77855092e-01 1.25818145e+00
2.07071126e-01 3.25171083e-01 -6.54224455e-01 3.96208286e-01
1.08860004e+00 3.89716625e-01 -3.65591168e-01 -1.03346348e+00
-1.58032119e+00 6.95526719e-01 6.23870254e-01 -2.19985396e-01
-3.31373990e-01 3.24548215e-01 -1.70770615e-01 5.16224690e-02
4.30443794e-01 3.57727736e-01 -4.32582855e-01 6.56211600e-02
-6.57361686e-01 1.24829277e-01 4.60198879e-01 1.30978131e+00
1.68036139e+00 2.52894461e-01 3.49236459e-01 4.67284709e-01
5.14410138e-01 7.42305040e-01 4.27479893e-01 -1.50177777e+00
5.49328685e-01 8.01299334e-01 2.94900626e-01 -7.90092707e-01
-4.79197681e-01 -4.68510002e-01 -6.73616409e-01 4.99614954e-01
2.35587016e-01 -1.15459062e-01 -1.21492398e+00 1.50504613e+00
7.67725825e-01 3.42681825e-01 1.08642884e-01 1.05980611e+00
5.64022064e-01 5.59010684e-01 -5.47320306e-01 -1.42885834e-01
1.06137121e+00 -1.32981634e+00 -7.37069607e-01 -8.37348580e-01
4.67065513e-01 -5.99297166e-01 5.88068604e-01 4.32948500e-01
-9.48312998e-01 -7.02699959e-01 -1.42594874e+00 -4.15260613e-01
3.16889733e-02 2.07948968e-01 4.94632870e-01 2.56014466e-01
-8.97186100e-01 9.73134488e-02 -8.81720901e-01 -2.77515560e-01
5.47734573e-02 2.74366587e-01 -6.77456558e-01 -4.51769292e-01
-8.82622004e-01 9.09102082e-01 5.92109442e-01 6.80380017e-02
-1.16366374e+00 -3.47976983e-01 -1.46335149e+00 -5.65458298e-01
5.63252747e-01 -1.05050421e+00 9.67765510e-01 -5.90260088e-01
-1.77923381e+00 8.77664983e-01 -5.26152968e-01 -3.38186741e-01
5.81738412e-01 -2.03334555e-01 2.07969964e-01 2.07861781e-01
4.96657699e-01 1.22481894e+00 1.01449049e+00 -1.29325104e+00
-1.09800899e+00 -6.93003118e-01 1.19836532e-01 8.63862693e-01
1.80573016e-01 -1.03488517e+00 -9.90533888e-01 -7.88023695e-02
1.14310122e+00 -1.13142502e+00 -5.89958251e-01 1.53176263e-01
-1.18621752e-01 2.67395556e-01 7.69835353e-01 -3.70496213e-01
5.61310649e-01 -1.92438948e+00 6.59469664e-01 -1.01277091e-01
3.99495512e-01 -3.79495680e-01 8.45176131e-02 -6.82381317e-02
4.00691658e-01 -5.80165505e-01 -3.51327598e-01 -6.58284009e-01
-3.82343560e-01 5.41835189e-01 5.96061349e-02 9.67865467e-01
-7.53593594e-02 8.25107515e-01 -1.14887607e+00 -4.88870829e-01
6.90910578e-01 3.88818771e-01 -7.90299535e-01 2.36168072e-01
-1.79521099e-01 1.11550403e+00 -4.30684924e-01 6.25250041e-01
9.36079085e-01 -5.42504676e-02 -9.27937254e-02 -1.89881161e-01
-5.82007885e-01 2.44378746e-01 -1.25375378e+00 2.31735229e+00
-5.37127554e-01 4.86939996e-01 1.09604523e-01 -6.81034863e-01
1.11873293e+00 -1.31322131e-01 3.71876985e-01 -5.92143357e-01
1.95052952e-01 3.05937260e-01 -3.16964358e-01 -3.92982453e-01
7.16254711e-01 -6.54980689e-02 -9.38358977e-02 -7.36148059e-02
5.61627895e-02 -9.68523562e-01 -2.17836723e-02 -6.12544529e-02
6.12185538e-01 7.73719370e-01 3.51047128e-01 1.09184265e-01
9.24765348e-01 1.71623632e-01 7.56407738e-01 2.74273723e-01
7.99759757e-03 6.55249774e-01 5.81512675e-02 -4.80900198e-01
-1.27046990e+00 -1.06728709e+00 -2.85617504e-02 4.56908226e-01
9.53105152e-01 -1.33748464e-02 -4.93591934e-01 -3.72258931e-01
1.84218772e-02 1.98624760e-01 -5.35906374e-01 -1.23570608e-02
-5.38712502e-01 -2.96654195e-01 -6.61504222e-03 3.33199322e-01
9.45924640e-01 -5.89454114e-01 -6.80423439e-01 2.82354563e-01
-4.67406571e-01 -1.45724201e+00 -5.05648017e-01 2.42338777e-01
-1.01374722e+00 -8.50169718e-01 -5.78596890e-01 -1.04546034e+00
8.02675962e-01 6.57418311e-01 4.97459441e-01 -2.04112455e-01
1.11276403e-01 5.65243326e-02 -4.29070406e-02 -1.36415735e-02
2.80326791e-02 5.53252734e-02 2.54724890e-01 7.92298093e-02
1.15198366e-01 -6.67489529e-01 -7.55687833e-01 4.73006904e-01
-5.45379996e-01 5.18022716e-01 7.23956704e-01 8.84174109e-01
8.99516761e-01 -2.72686541e-01 -1.12219052e-02 -6.23255730e-01
-4.97044683e-01 -3.06284457e-01 -9.78396595e-01 -5.16242266e-01
-6.58596814e-01 2.96116084e-01 1.47627339e-01 -1.64771259e-01
-1.12056112e+00 7.96772361e-01 -1.39139816e-01 -6.97888017e-01
-1.61450934e-02 2.07648501e-01 -2.16611385e-01 -2.93088257e-01
3.98741424e-01 4.39585060e-01 2.40710229e-01 -1.34188130e-01
5.37953734e-01 4.74868685e-01 8.91587734e-01 -1.40974864e-01
1.20622492e+00 1.11368942e+00 9.68785509e-02 -7.17284679e-01
-8.72775674e-01 -6.51423812e-01 -1.10187733e+00 -1.72813922e-01
1.20206714e+00 -1.53988802e+00 -5.68025470e-01 5.89364886e-01
-1.21938860e+00 -1.93456799e-01 1.12697676e-01 9.47849631e-01
-7.93038428e-01 5.99546313e-01 -3.49513113e-01 -5.05563021e-01
3.70322354e-02 -1.43883526e+00 1.43108332e+00 1.38367966e-01
2.44337961e-01 -7.19207466e-01 3.57310884e-02 5.12936473e-01
-2.29555696e-01 1.28168046e-01 2.70957738e-01 3.90952706e-01
-1.19788051e+00 9.23312977e-02 -2.06971377e-01 1.78234920e-01
7.80804828e-02 -3.75449866e-01 -8.51656973e-01 -2.59583652e-01
1.75319746e-01 -5.58936261e-02 8.18175375e-01 3.50171894e-01
4.27992195e-01 1.30138099e-01 -3.37325037e-01 1.28659415e+00
1.51499200e+00 3.23329777e-01 5.66924453e-01 8.04550827e-01
1.19562888e+00 5.83215535e-01 1.04778755e+00 3.57696235e-01
7.89751828e-01 5.71887732e-01 9.97271597e-01 -1.78503413e-02
9.57682952e-02 -4.67431724e-01 3.87159795e-01 7.77206182e-01
-1.38649112e-02 2.67852366e-01 -7.24721491e-01 5.10752738e-01
-1.92089713e+00 -5.81355631e-01 -1.44178048e-01 2.12611222e+00
6.14921212e-01 1.45675182e-01 -4.76759344e-01 -8.48589987e-02
5.39136946e-01 3.61554056e-01 -8.46323073e-01 1.79487467e-01
-1.97628345e-02 -2.03855976e-01 9.52705443e-01 1.14788485e+00
-9.61738467e-01 1.30452538e+00 5.16421795e+00 2.51332074e-01
-1.12311697e+00 8.34033191e-02 2.10778445e-01 4.08171527e-02
-3.43029320e-01 1.63708627e-01 -1.18930876e+00 1.07837938e-01
2.10916415e-01 1.96808264e-01 4.12863553e-01 1.01711655e+00
1.73538014e-01 -4.70847249e-01 -9.49364066e-01 1.36353123e+00
3.28029245e-01 -1.21933413e+00 -5.11492230e-02 2.37652659e-01
1.23345172e+00 3.19731921e-01 -8.80929902e-02 1.19745150e-01
2.27440923e-01 -5.94056666e-01 9.07877684e-01 3.91090542e-01
7.95462489e-01 -7.32463539e-01 7.83596277e-01 6.82706237e-01
-1.39592874e+00 -1.00669973e-01 -4.84898329e-01 -5.20634949e-01
4.46492076e-01 3.36191177e-01 -7.41966248e-01 6.22920811e-01
6.81372881e-01 1.21039379e+00 -4.09168631e-01 6.80514216e-01
-5.69958210e-01 -3.90689760e-01 -2.69584566e-01 2.56464988e-01
5.47148168e-01 -5.54530203e-01 6.03177011e-01 3.51496488e-01
4.89369720e-01 -3.18508176e-03 3.12956184e-01 7.22569406e-01
3.84840183e-02 -1.92606151e-01 -7.91560292e-01 6.74225748e-01
5.27248621e-01 1.25727344e+00 -3.82199079e-01 -5.94247103e-01
-4.96321052e-01 1.26700389e+00 1.82437405e-01 3.62537980e-01
-6.53020501e-01 -1.11579210e-01 6.79383337e-01 2.42349040e-02
4.13840532e-01 -5.96638262e-01 -2.99347818e-01 -1.33608699e+00
1.22859254e-02 -4.61923748e-01 -1.84177786e-01 -1.02287745e+00
-5.08266687e-01 5.40471673e-01 -5.32887876e-02 -1.62714577e+00
-5.45828164e-01 -5.30738831e-01 -1.53066635e-01 9.12702024e-01
-1.93723822e+00 -8.82571161e-01 -9.65618253e-01 5.00402987e-01
7.86678672e-01 -1.69839915e-02 2.02652499e-01 2.40790412e-01
-1.16837174e-01 -5.41652180e-02 -9.20036882e-02 -2.43067637e-01
7.82606304e-01 -1.10834241e+00 4.79327351e-01 9.20165837e-01
-8.79898071e-02 1.91042691e-01 5.18338025e-01 -6.55530870e-01
-1.45553398e+00 -1.11129272e+00 6.97761595e-01 -5.82067549e-01
3.93477798e-01 -2.70658553e-01 -6.11962974e-01 8.67856383e-01
-1.90187749e-02 -6.84947148e-03 -3.56715649e-01 -6.20873332e-01
1.12451846e-02 -1.49067253e-01 -9.37158585e-01 6.76333845e-01
1.34190464e+00 -5.45132875e-01 -5.50212979e-01 3.04373242e-02
9.49635923e-01 -1.03232276e+00 -4.85367894e-01 4.76306319e-01
6.29474044e-01 -1.02054155e+00 9.58806574e-01 2.46578202e-01
3.01990837e-01 -8.73129666e-01 -2.22977892e-01 -1.08697104e+00
-7.85241947e-02 -5.63958704e-01 6.98674023e-02 7.36454070e-01
1.77364260e-01 -8.11789572e-01 1.16263795e+00 -1.74242537e-02
-4.25080121e-01 -5.77606976e-01 -9.08638000e-01 -3.60269845e-01
-3.50841880e-01 -4.46472734e-01 4.32578862e-01 6.87403083e-01
-3.67863774e-01 4.86678392e-01 -4.02261734e-01 6.54430449e-01
8.98270011e-01 1.52938768e-01 1.34051013e+00 -1.08075643e+00
2.05032364e-03 1.16557270e-01 -7.79708326e-01 -2.06154704e+00
1.78715274e-01 -6.02749050e-01 4.54565734e-01 -1.54485989e+00
-1.91971555e-01 -1.18698917e-01 2.00659677e-01 8.02353173e-02
-7.81046078e-02 5.33184052e-01 -2.88848262e-02 3.18410873e-01
-3.39898527e-01 7.47953415e-01 1.76819396e+00 -1.65952623e-01
-3.55236471e-01 -2.03916073e-01 -1.58870995e-01 1.04686439e+00
5.25511980e-01 -2.99954861e-01 -5.31875908e-01 -5.32861650e-01
7.67801255e-02 3.10657978e-01 3.34785074e-01 -1.12741756e+00
4.43245351e-01 -1.58124387e-01 4.38561946e-01 -1.21110559e+00
6.86394513e-01 -8.03766906e-01 -6.81380332e-02 5.92797756e-01
1.99274614e-01 8.94796327e-02 -2.45158136e-01 8.09165537e-01
-2.85564780e-01 -1.50832966e-01 7.92132497e-01 -3.65760654e-01
-1.32755661e+00 7.38212347e-01 -1.51045278e-01 -1.95720524e-01
1.04102647e+00 -6.18525088e-01 2.00426951e-02 -4.82892573e-01
-6.11909866e-01 4.32571620e-01 9.86538470e-01 4.61959898e-01
9.11783576e-01 -1.33683467e+00 -4.78200525e-01 6.33209109e-01
3.64406943e-01 8.11172903e-01 2.19073951e-01 1.01506746e+00
-1.02275288e+00 3.53400618e-01 -4.10646796e-01 -1.19316828e+00
-9.30572450e-01 4.17216361e-01 3.77027422e-01 2.32765172e-02
-6.68793917e-01 6.69421673e-01 8.74158323e-01 -8.47123623e-01
9.55547169e-02 -4.93343323e-01 -1.38179749e-01 -2.35822454e-01
3.14111143e-01 3.28011036e-01 -2.07331896e-01 -9.05886173e-01
-2.09437355e-01 1.24250519e+00 8.39782506e-02 -4.38290685e-01
1.13145459e+00 -8.61293137e-01 -1.38011158e-01 4.27203894e-01
1.40473592e+00 -2.40007918e-02 -1.98507321e+00 -3.87260735e-01
-3.83028388e-01 -6.11735404e-01 1.19765244e-01 4.99814525e-02
-1.03581524e+00 8.86878610e-01 5.21533608e-01 -7.14374244e-01
8.28396142e-01 -6.81455955e-02 6.73608661e-01 4.72068369e-01
8.25812697e-01 -8.94069612e-01 2.24997729e-01 8.70358407e-01
4.95294452e-01 -1.40258741e+00 1.65874064e-01 -5.61003745e-01
-4.79564220e-01 1.04746878e+00 9.89912391e-01 -3.17444980e-01
3.43650669e-01 8.56574923e-02 2.67178118e-01 7.30902003e-03
-4.69085366e-01 -3.99158120e-01 1.43762216e-01 6.51669502e-01
-9.84894708e-02 -3.27604383e-01 -4.36110161e-02 -2.51313806e-01
-1.99216008e-01 -2.01858968e-01 5.86498201e-01 7.81783640e-01
-7.02355623e-01 -9.19005573e-01 -5.00379860e-01 -1.64558321e-01
2.02440143e-01 6.79274788e-03 4.89486987e-03 8.67895842e-01
4.36337680e-01 6.25137866e-01 2.23569676e-01 -4.77161169e-01
1.89425111e-01 -1.94085494e-01 6.34832323e-01 -9.22875524e-01
2.81029046e-01 2.10076660e-01 -8.43574330e-02 -8.09123695e-01
-4.65635747e-01 -5.83535433e-01 -1.58901143e+00 1.68564282e-02
-2.20492154e-01 -8.88370946e-02 7.89476633e-01 1.06679904e+00
-4.75275218e-02 4.22048122e-01 7.32880652e-01 -1.55152905e+00
-6.23821765e-02 -7.31218219e-01 -4.76458520e-01 -3.19630206e-02
7.18929708e-01 -1.03294468e+00 -5.07431865e-01 1.13409147e-01]
|
[8.09899616241455, -2.2465481758117676]
|
4bf225cc-c5ec-490b-b8bc-a00620f0fc0c
|
networks-of-piecewise-linear-neural-mass
|
1801.08366
| null |
http://arxiv.org/abs/1801.08366v1
|
http://arxiv.org/pdf/1801.08366v1.pdf
|
Networks of piecewise linear neural mass models
|
Neural mass models are ubiquitous in large scale brain modelling. At the node
level they are written in terms of a set of ODEs with a nonlinearity that is
typically a sigmoidal shape. Using structural data from brain atlases they may
be connected into a network to investigate the emergence of functional dynamic
states, such as synchrony. With the simple restriction of the classic sigmoidal
nonlinearity to a piecewise linear caricature we show that the famous
Wilson-Cowan neural mass model can be analysed at both the node and network
level. The construction of periodic orbits at the node level is achieved by
patching together matrix exponential solutions, and stability is determined
using Floquet theory. For networks with interactions described by circulant
matrices, we show that the stability of the synchronous state can be determined
in terms of a low-dimensional Floquet problem parameterised by the eigenvalues
of the interaction matrix. This network Floquet problem is readily solved using
linear algebra, to predict the onset of spatio-temporal network patterns
arising from a synchronous instability. We consider the case of a discontinuous
choice for the node nonlinearity, namely the replacement of the sigmoid by a
Heaviside nonlinearity. This gives rise to a continuous-time switching network.
At the node level this allows for the existence of unstable sliding periodic
orbits, which we construct. The stability of a periodic orbit is now treated
with a modification of Floquet theory to treat the evolution of small
perturbations through switching manifolds via saltation matrices. At the
network level the stability analysis of the synchronous state is considerably
more challenging. Here we report on the use of ideas originally developed for
the study of Glass networks to treat the stability of periodic network states
in neural mass models with discontinuous interactions.
|
[]
|
2018-01-25
| null | null | null | null |
['caricature']
|
['computer-vision']
|
[-3.35130584e-03 3.74150783e-01 3.79721940e-01 4.07094002e-01
4.02916908e-01 -5.59729397e-01 8.52776825e-01 -1.17465835e-02
-2.97595173e-01 5.28859198e-01 -1.18074469e-01 -1.79527193e-01
-4.17890608e-01 -4.47840333e-01 -7.93332279e-01 -1.37597311e+00
-7.28080809e-01 3.10208917e-01 3.90108645e-01 -7.71074414e-01
-1.26536697e-01 4.74920899e-01 -1.33981502e+00 -1.21051110e-01
4.84157234e-01 7.10256279e-01 -5.99913150e-02 6.22817874e-01
1.87124252e-01 5.99918246e-01 -3.33704352e-01 9.90834385e-02
3.90475661e-01 -5.52299201e-01 -6.96208477e-01 1.13344051e-01
-2.09863827e-01 4.26529348e-01 -3.16960126e-01 9.47135985e-01
3.05692434e-01 8.06521848e-02 5.80752134e-01 -1.44994175e+00
-2.67111719e-01 6.11494780e-01 -2.71490008e-01 4.05684173e-01
8.32818523e-02 -1.47064179e-01 5.10371864e-01 -5.63650787e-01
8.24662089e-01 1.11700082e+00 8.09933126e-01 4.71042335e-01
-1.87859368e+00 -2.31594905e-01 -3.02930862e-01 -1.00343056e-01
-1.48169446e+00 -4.56128776e-01 6.49128079e-01 -8.58286560e-01
7.83800602e-01 1.95399106e-01 9.99445081e-01 7.07186818e-01
8.81275356e-01 1.01707593e-01 1.01051497e+00 -5.05686700e-01
3.37362051e-01 -1.51719868e-01 2.36123934e-01 8.44779670e-01
1.22260556e-01 2.51944810e-01 -1.36196062e-01 -3.47401619e-01
9.74750519e-01 -2.10052580e-01 -2.49042436e-01 -4.48378474e-01
-1.26738191e+00 7.65956104e-01 6.96163356e-01 9.09932613e-01
-2.51250058e-01 2.28806958e-01 2.43915886e-01 7.44467854e-01
3.68717462e-01 2.51856387e-01 1.14384882e-01 4.19398695e-01
-7.33532310e-01 1.25549838e-01 1.10237014e+00 5.23530245e-01
5.36069930e-01 6.64033219e-02 3.27532023e-01 3.72112542e-01
3.23340118e-01 3.58944237e-01 5.12434781e-01 -9.23647523e-01
-2.59452879e-01 4.93519574e-01 -2.89047062e-01 -6.63978577e-01
-8.71955812e-01 -5.30731618e-01 -1.15390885e+00 3.81637990e-01
7.08721220e-01 -2.10022509e-01 -4.73529816e-01 1.95391738e+00
2.25047350e-01 1.95528284e-01 1.75946057e-01 4.93063241e-01
2.72313088e-01 7.65807450e-01 -2.94031948e-01 -7.26165473e-01
1.25656259e+00 -1.94984615e-01 -6.44770980e-01 4.06353801e-01
7.90675879e-01 -3.03292304e-01 3.10597867e-01 1.37913570e-01
-1.38886106e+00 -3.64290439e-02 -9.84112680e-01 4.49369997e-01
-4.11032528e-01 -2.51293808e-01 -1.40644148e-01 1.16075657e-01
-1.62524736e+00 7.78278708e-01 -9.88392591e-01 -6.30421102e-01
-3.39287192e-01 6.56647801e-01 -3.24591666e-01 5.76817811e-01
-1.33539271e+00 1.15279889e+00 1.74969971e-01 4.37004596e-01
-4.42293406e-01 -4.92442250e-01 -6.77598536e-01 4.46600001e-03
-1.39354914e-01 -6.41076028e-01 8.30287337e-01 -1.12063682e+00
-1.21094501e+00 8.08602512e-01 7.19630271e-02 -3.47942889e-01
4.70951945e-01 8.05037796e-01 -2.11888582e-01 2.60719091e-01
-5.35361022e-02 4.88332123e-01 9.40848827e-01 -1.07321620e+00
2.74398208e-01 -1.68570846e-01 -7.56798461e-02 -1.06599160e-01
-1.62959546e-01 -3.11805081e-04 3.80788893e-01 -5.17162025e-01
2.40420043e-01 -1.39218533e+00 -9.00245383e-02 -2.06379443e-01
-6.07242510e-02 -2.37418875e-01 7.07719803e-01 -3.29990476e-01
1.04772580e+00 -2.17274761e+00 6.52043104e-01 5.39235651e-01
5.27069926e-01 2.67051323e-03 1.53830200e-02 7.90732145e-01
-4.73433584e-01 -1.06252529e-01 -6.04601204e-01 -7.98128992e-02
-2.57444143e-01 2.33036682e-01 -6.02370352e-02 9.86081183e-01
1.20767012e-01 8.28765869e-01 -5.93922615e-01 -2.78043658e-01
-3.79774332e-01 5.66786528e-01 -5.00808775e-01 -3.79812986e-01
2.03976616e-01 2.60298759e-01 -9.78367329e-02 1.49268089e-02
2.97315806e-01 -1.55143037e-01 -4.36327681e-02 2.50220478e-01
-5.80507278e-01 -3.13235909e-01 -1.11823356e+00 1.00232601e+00
-1.67580053e-01 7.35205829e-01 6.35820448e-01 -1.24054801e+00
7.95714438e-01 7.04865277e-01 6.02436066e-01 -4.49742861e-02
4.29888487e-01 3.82751197e-01 6.82001173e-01 -2.74973541e-01
-1.55148983e-01 -3.98281008e-01 1.26168326e-01 4.45067465e-01
2.17098489e-01 2.02307299e-01 2.88163573e-01 2.67426550e-01
1.35551083e+00 -4.43463713e-01 -9.36289057e-02 -1.24902868e+00
5.70030928e-01 -4.54434231e-02 7.90512711e-02 2.36543104e-01
-6.30871058e-02 2.02244267e-01 1.04715943e+00 -3.02000701e-01
-1.06487811e+00 -9.37494874e-01 -7.04333723e-01 5.44172704e-01
-7.91120306e-02 -7.61583745e-02 -1.03997374e+00 4.69212532e-01
-6.11005072e-03 -2.21889075e-02 -1.05443704e+00 -5.87313354e-01
-6.86470509e-01 -1.05668187e+00 5.17832100e-01 -1.59846663e-01
3.81311297e-01 -1.18531215e+00 -5.96468627e-01 3.76023680e-01
6.33462891e-02 -7.35809147e-01 -3.23324323e-01 4.26837176e-01
-1.10094786e+00 -7.67081857e-01 -9.12965059e-01 -1.01660264e+00
7.75537550e-01 -4.58052725e-01 4.31615502e-01 2.47955114e-01
-3.32559049e-01 4.44024950e-01 1.99487284e-01 -4.56425399e-02
-9.08789515e-01 7.87611902e-02 6.16565526e-01 3.06288511e-01
-2.57748574e-01 -9.43608046e-01 -4.61589754e-01 6.34922802e-01
-1.32677472e+00 -2.87308216e-01 1.18446305e-01 5.14461040e-01
-1.94846801e-02 -2.22598370e-02 8.79433215e-01 -1.26903862e-01
9.67040479e-01 -6.42509639e-01 -5.49637437e-01 9.11417753e-02
-3.84776890e-01 4.70205396e-01 5.69706857e-01 -8.51425469e-01
-3.91046405e-01 1.45487204e-01 3.33997577e-01 -2.12343261e-01
3.21354002e-01 7.06305087e-01 4.46880996e-01 -6.98528171e-01
8.84856105e-01 3.31651479e-01 6.91471279e-01 -1.87336162e-01
5.90061992e-02 3.75818968e-01 4.04132217e-01 -4.51420933e-01
6.44495487e-01 3.62542450e-01 5.77740788e-01 -1.19526517e+00
2.10459843e-01 -2.69463748e-01 -7.16834962e-01 -7.08852053e-01
7.16697216e-01 -4.06689256e-01 -1.00315309e+00 7.12282181e-01
-1.07621491e+00 -5.89441597e-01 -4.98125196e-01 1.93624631e-01
-7.85399377e-01 6.46973550e-02 -9.48503613e-01 -8.89544725e-01
6.19504936e-02 -8.58754635e-01 4.92502213e-01 -2.14129731e-01
-1.50316373e-01 -1.42413676e+00 5.48636496e-01 -3.95601332e-01
5.39245069e-01 4.51727539e-01 9.50850487e-01 -2.56724209e-01
-3.17580760e-01 -3.60255480e-01 4.60224122e-01 -7.45368227e-02
-3.06477427e-01 2.25831971e-01 -5.35807967e-01 -3.19230825e-01
6.41790867e-01 3.68564576e-01 7.41971016e-01 6.64962769e-01
-1.88893840e-01 -3.69628996e-01 -3.23379159e-01 2.68813491e-01
1.26017249e+00 2.02053800e-01 3.78158838e-01 1.84169918e-01
3.79035711e-01 8.69581103e-01 -5.45315444e-01 1.32600199e-02
4.94446643e-02 5.42078853e-01 2.37745479e-01 3.17974776e-01
2.81093448e-01 4.21066701e-01 5.81943870e-01 1.15912116e+00
-4.54821020e-01 3.70341122e-01 -1.26128209e+00 6.11647785e-01
-2.05347466e+00 -1.00657380e+00 -3.64296794e-01 2.22290945e+00
8.48661005e-01 1.30132169e-01 5.32573819e-01 1.91651329e-01
9.44758475e-01 -2.44082153e-01 -1.53256863e-01 -4.63092417e-01
-2.98265427e-01 -2.88135141e-01 5.15852034e-01 9.46133316e-01
-5.32364666e-01 3.23080450e-01 6.66069698e+00 2.94760466e-01
-1.10279369e+00 2.59430528e-01 1.88797712e-01 -1.29001394e-01
1.41713157e-01 7.19579756e-02 -4.57835555e-01 4.77225244e-01
1.21704245e+00 -3.29534948e-01 7.61235774e-01 1.37642786e-01
3.13647479e-01 -3.02477986e-01 -7.60021448e-01 5.87261260e-01
-1.25096321e-01 -1.21904969e+00 -4.73231226e-01 4.75658089e-01
8.60307813e-01 6.70889765e-02 -3.32005285e-02 -2.04704940e-01
5.21085076e-02 -7.21579015e-01 6.40788317e-01 7.08223403e-01
3.57923776e-01 -4.07937974e-01 2.96370715e-01 6.72958255e-01
-1.26715255e+00 -1.04612477e-01 -2.50964195e-01 -3.12965930e-01
1.88089788e-01 2.58031487e-01 -9.50958133e-02 -7.45345801e-02
1.90208688e-01 6.46374881e-01 -3.00423533e-01 1.12317312e+00
6.01949096e-01 3.57657969e-01 -8.52653027e-01 -1.28949463e-01
2.61429012e-01 -7.27043867e-01 9.07914579e-01 7.85374582e-01
3.46701056e-01 2.77523905e-01 -4.47749496e-01 1.03141487e+00
2.16932014e-01 -3.90728936e-02 -7.96226740e-01 2.08140656e-01
-2.48480737e-01 1.27107155e+00 -1.23044622e+00 -3.26395258e-02
3.80709134e-02 4.61751223e-01 1.40913829e-01 4.99136150e-01
-3.67105752e-01 -3.17554027e-01 4.61011320e-01 6.22539520e-01
2.89652258e-01 -4.54062223e-01 1.34914160e-01 -1.20871294e+00
4.30604592e-02 -1.18724652e-01 -2.10898429e-01 -6.69485509e-01
-8.77215981e-01 7.95010448e-01 4.11058575e-01 -9.49854672e-01
-4.36786473e-01 -4.83675390e-01 -9.97239113e-01 7.54743099e-01
-6.16519868e-01 -6.67793453e-01 3.91228974e-01 1.00098860e+00
-3.90371084e-01 -8.37311298e-02 5.66411138e-01 8.41956958e-03
-4.86572862e-01 -9.50892195e-02 5.03450453e-01 -1.02060266e-01
7.66230449e-02 -1.15463603e+00 1.19788133e-01 6.86545014e-01
-2.53524810e-01 6.79544449e-01 1.12124562e+00 -7.00272799e-01
-1.07418334e+00 -5.85090280e-01 1.18861842e+00 -1.56745881e-01
1.30342364e+00 -7.60053158e-01 -1.06671500e+00 5.94984114e-01
2.49984905e-01 3.30878973e-01 7.99679458e-02 -4.68336433e-01
3.58404547e-01 7.44281337e-02 -9.42131758e-01 5.23810089e-01
8.56687486e-01 -4.07644093e-01 -3.64641041e-01 4.64301825e-01
3.25964123e-01 1.42343193e-01 -9.54739809e-01 3.90260406e-02
4.43356812e-01 -5.95510364e-01 6.07874036e-01 -4.26839828e-01
6.82568625e-02 -2.87186116e-01 4.70750123e-01 -1.42310822e+00
-3.85301441e-01 -1.19463801e+00 9.65605397e-03 7.66446531e-01
2.20634997e-01 -1.01580524e+00 1.75826192e-01 5.13792753e-01
1.02361538e-01 -7.31519103e-01 -1.56136501e+00 -1.02748907e+00
3.27407211e-01 3.12067807e-01 -1.62604317e-01 8.40649605e-01
8.67472708e-01 3.29324335e-01 1.54908359e-01 -1.73001707e-01
6.71068609e-01 -5.18645108e-01 -1.85901433e-01 -1.59230280e+00
-2.11935882e-02 -6.93405271e-01 -8.44165206e-01 -4.40869361e-01
1.27838045e-01 -1.14589703e+00 1.36123970e-01 -1.07015812e+00
-4.23127890e-01 -1.82026282e-01 -1.19541876e-01 2.20973771e-02
6.40883565e-01 9.06832293e-02 1.07123740e-01 6.58638000e-01
3.43998708e-02 2.50611216e-01 1.11101198e+00 2.87005007e-01
-5.15056074e-01 7.31556639e-02 -4.62381616e-02 6.84808254e-01
7.26463795e-01 -6.24582827e-01 -2.52079010e-01 3.05671930e-01
8.42797995e-01 4.41261619e-01 6.01420045e-01 -1.16637969e+00
6.97408795e-01 4.49901968e-01 -3.58778387e-01 2.49870464e-01
2.64062583e-01 -7.15150356e-01 5.62792420e-01 9.33626592e-01
-3.75113666e-01 3.77036899e-01 1.27687693e-01 5.84847391e-01
-6.94122761e-02 -4.69982237e-01 8.34899426e-01 -6.51426730e-04
3.18289250e-01 1.23502910e-02 -1.18445432e+00 -1.51348054e-01
1.08898020e+00 -1.83823854e-01 -2.28343636e-01 -3.98880690e-01
-1.52770996e+00 1.14258036e-01 3.93326670e-01 -1.61493346e-01
2.49004707e-01 -1.52113402e+00 -5.18405318e-01 4.26576585e-01
-4.24953490e-01 -3.92366469e-01 -2.62661595e-02 1.60130787e+00
-5.57268083e-01 1.99744686e-01 -4.93230104e-01 -6.53049588e-01
-1.10701478e+00 5.69841504e-01 9.85718846e-01 -6.96848705e-02
-2.36692786e-01 3.98916245e-01 1.43684477e-01 -2.36852616e-02
-7.36596584e-02 -6.36188805e-01 -2.88758606e-01 4.54106510e-01
1.30156189e-01 2.43697897e-01 -4.30456698e-02 -1.13309157e+00
-2.25051478e-01 7.37384796e-01 6.18364930e-01 -4.58614171e-01
1.33750749e+00 -8.44077393e-02 -8.69346619e-01 8.78090799e-01
1.32383657e+00 -3.49723577e-01 -1.08506930e+00 -2.37846255e-01
-2.01126412e-02 5.11858940e-01 -1.02758348e-01 -9.32406113e-02
-9.85848844e-01 5.88537991e-01 3.64874929e-01 1.22819412e+00
8.30702722e-01 2.38046974e-01 3.66360545e-01 3.60299677e-01
2.41670869e-02 -7.48745799e-01 -3.01899880e-01 6.98050320e-01
1.25807285e+00 -7.17631280e-01 -4.77625132e-01 -1.14673480e-01
4.54897946e-03 1.40654147e+00 -8.43915418e-02 -9.67100978e-01
1.38674390e+00 5.55394113e-01 -3.01794797e-01 -3.81585807e-01
-8.41482222e-01 -5.34068793e-03 2.81262755e-01 2.12686270e-01
9.28590968e-02 -6.51065111e-02 -4.83923346e-01 1.55149549e-01
-1.95242643e-01 -1.87349990e-01 8.89290750e-01 8.12107325e-01
-5.64090312e-01 -8.06650698e-01 -5.12842298e-01 3.12619358e-01
-2.53987104e-01 -1.02505425e-03 -4.59947288e-01 7.75596619e-01
-5.10561652e-02 5.74039400e-01 4.02060032e-01 -1.45743847e-01
1.92858428e-01 5.98809421e-01 5.86271107e-01 -2.75523484e-01
-7.59967625e-01 9.36276391e-02 -3.83401245e-01 -2.56994776e-02
-6.33008361e-01 -1.05144846e+00 -1.13059473e+00 -6.25949085e-01
-3.42580557e-01 1.80289373e-01 5.94583869e-01 1.04446566e+00
6.71178848e-02 4.01029050e-01 2.84131199e-01 -1.09032691e+00
-3.54009390e-01 -8.99785221e-01 -1.05001676e+00 2.05728952e-02
8.37209165e-01 -5.41730762e-01 -7.17246890e-01 1.51070505e-01]
|
[6.261447906494141, 4.097131252288818]
|
22dab665-b10d-4db0-814d-27e3686598d7
|
using-kalman-filter-the-right-way-noise
|
2104.02372
| null |
https://arxiv.org/abs/2104.02372v4
|
https://arxiv.org/pdf/2104.02372v4.pdf
|
The Fragility of Noise Estimation in Kalman Filter: Optimization Can Handle Model-Misspecification
|
The Kalman Filter (KF) parameters are traditionally determined by noise estimation, since under the KF assumptions, the state prediction errors are minimized when the parameters correspond to the noise covariance. However, noise estimation remains the gold-standard regardless of the assumptions - even when it is not equivalent to errors minimization. We demonstrate that even seemingly simple problems may include multiple assumptions violations - which are sometimes hard to even notice. We show theoretically and empirically that even a minor violation may largely shift the optimal parameters. We propose a gradient-based method along with the Cholesky parameterization to explicitly optimize the state prediction errors. We show consistent improvement over noise estimation in tens of experiments in 3 different domains. Finally, we demonstrate that optimization makes the KF competitive with an LSTM model - even in non linear problems.
|
['Netanel Yannay', 'Shie Mannor', 'Ido Greenberg']
|
2021-04-06
| null | null | null | null |
['noise-estimation']
|
['medical']
|
[-1.03020512e-01 -1.05526365e-01 6.54120147e-02 -2.24289268e-01
-9.56674755e-01 -6.77860320e-01 6.42819345e-01 -4.95984644e-01
-5.91147602e-01 9.16456223e-01 2.91534096e-01 -5.31066895e-01
-1.89449698e-01 -2.64375687e-01 -8.48025978e-01 -8.49315286e-01
7.49449879e-02 1.68381661e-01 2.82631852e-02 -8.02853424e-03
1.83559924e-01 1.88457876e-01 -1.16589618e+00 -2.61374950e-01
7.99131632e-01 1.00172698e+00 5.11650331e-02 7.41466343e-01
2.92386621e-01 7.25228727e-01 -4.95463759e-01 -1.57806322e-01
5.27268827e-01 -2.02804834e-01 -5.35527706e-01 -3.42993081e-01
6.81289077e-01 -3.97390127e-01 -4.71163929e-01 1.41128802e+00
2.96343833e-01 3.96680444e-01 4.15168941e-01 -8.19747090e-01
-1.08224832e-01 6.55012608e-01 6.36782274e-02 2.02019528e-01
7.17185289e-02 2.67799228e-01 9.77192104e-01 -6.43956065e-01
4.04660165e-01 1.25079739e+00 1.15559030e+00 3.98760617e-01
-1.53259528e+00 -5.87529778e-01 2.68035829e-01 4.09217104e-02
-1.35052156e+00 -8.31397355e-01 1.69494897e-01 -3.78798217e-01
1.18516946e+00 -1.70549825e-02 4.82813895e-01 1.23754621e+00
5.58263779e-01 5.14753997e-01 1.43942428e+00 -3.91150951e-01
2.23605961e-01 7.74490833e-02 5.35629153e-01 4.98613417e-01
4.25741851e-01 5.34011006e-01 -4.51484919e-01 -3.77739131e-01
8.11586738e-01 -5.03085017e-01 -4.36113596e-01 -1.87665015e-01
-1.42464006e+00 4.82697666e-01 -1.12364911e-01 1.64437860e-01
-2.19867989e-01 6.54532313e-01 2.84557939e-01 6.92141414e-01
3.90123874e-01 5.37070870e-01 -9.69111919e-01 -5.20467341e-01
-1.02429557e+00 4.36314404e-01 1.19087279e+00 8.39741349e-01
7.31439948e-01 5.59535980e-01 1.82954177e-01 4.82480705e-01
4.84507471e-01 9.88989830e-01 3.71538579e-01 -1.21830630e+00
2.03384951e-01 -1.34478539e-01 6.07643366e-01 -7.43978381e-01
-8.09670806e-01 -9.65335488e-01 -8.05056572e-01 9.44089070e-02
8.80978465e-01 -6.84305787e-01 -8.75214398e-01 1.94493830e+00
-3.64970975e-02 5.12208700e-01 1.54372826e-01 7.20765114e-01
5.77534959e-02 5.29739499e-01 -1.52536839e-01 -5.88950634e-01
9.74990368e-01 -7.56497562e-01 -1.21326184e+00 -6.95850968e-01
8.46079528e-01 -8.07440579e-01 9.41220105e-01 7.05156326e-01
-8.43868434e-01 -3.15045893e-01 -9.67232704e-01 2.47571141e-01
1.91177074e-02 2.81223834e-01 3.55063587e-01 8.43535244e-01
-1.05094850e+00 1.14190328e+00 -1.17074656e+00 -3.44136149e-01
-3.51751179e-01 4.23774630e-01 -1.96670130e-01 6.28033817e-01
-1.33475900e+00 1.54415011e+00 3.32398951e-01 3.85715038e-01
-8.57231498e-01 -6.42494857e-01 -5.06408393e-01 9.51683335e-03
5.15673578e-01 -6.93540096e-01 1.65760326e+00 -6.84738398e-01
-1.89436793e+00 1.34511262e-01 -4.45813566e-01 -8.84907722e-01
5.85115314e-01 -9.02146161e-01 -5.49146056e-01 -4.22756016e-01
-4.39741433e-01 -9.55375470e-03 1.17302775e+00 -9.83355522e-01
-5.18495083e-01 9.81487054e-03 -1.56910256e-01 4.22722064e-02
-3.97734083e-02 -1.81506306e-01 -2.19454721e-01 -4.14215386e-01
4.02217388e-01 -1.26274586e+00 -4.96943384e-01 -4.94401455e-01
-5.42962849e-01 3.03509116e-01 5.04720390e-01 -8.06538105e-01
1.56398010e+00 -1.84543300e+00 -2.00916201e-01 5.51896513e-01
-9.09355953e-02 2.23317340e-01 1.33465141e-01 3.69245112e-01
7.88822770e-04 -1.31371887e-02 2.38814596e-02 -3.97901148e-01
2.76911795e-01 3.09690058e-01 -6.51766121e-01 8.45954895e-01
-1.20781079e-01 8.97460580e-01 -7.94300377e-01 1.21099636e-01
3.37871313e-01 2.86438704e-01 -3.88359010e-01 -2.14572996e-01
-8.40877965e-02 3.83352786e-01 -2.11935610e-01 -9.31885913e-02
5.69432139e-01 -3.57332826e-01 4.82742220e-01 -4.97690022e-01
-1.93867534e-01 6.94966435e-01 -1.65539932e+00 1.33894587e+00
-2.82356501e-01 7.85019040e-01 1.72614262e-01 -8.73751640e-01
4.47366059e-01 2.28990838e-01 6.84247762e-02 -3.94246161e-01
2.05181450e-01 2.99920559e-01 2.12104276e-01 -2.87590623e-01
5.06644607e-01 -2.97167212e-01 3.19365482e-03 2.58691639e-01
1.23337939e-01 -1.80461079e-01 -5.77800050e-02 5.18389530e-02
9.97790337e-01 2.11328655e-01 4.80731875e-01 -7.30779290e-01
1.25801951e-01 -6.49238974e-02 9.06598687e-01 1.37289977e+00
-3.28450739e-01 2.90072203e-01 4.84993160e-01 -1.77711993e-01
-9.86425996e-01 -9.50772882e-01 -2.77336419e-01 6.82516932e-01
-1.99746460e-01 -5.78440845e-01 -8.43425095e-01 -4.53343213e-01
2.59447638e-02 7.81545460e-01 -3.79269361e-01 -1.73458531e-01
-4.65233326e-01 -8.95038307e-01 7.53104746e-01 4.04285908e-01
3.79739374e-01 -2.81192929e-01 -3.04081202e-01 4.17760521e-01
3.57315391e-02 -1.22689462e+00 -4.40551370e-01 4.12236392e-01
-1.01888287e+00 -6.95553601e-01 -2.17831060e-01 -1.43697053e-01
3.21653038e-01 -1.18523668e-02 8.74334216e-01 -1.49101496e-01
4.64788139e-01 5.13791919e-01 1.06842473e-01 -2.84422964e-01
-5.92769325e-01 8.87880921e-02 8.00977826e-01 -3.42493534e-01
2.11993501e-01 -5.20696282e-01 -4.12500560e-01 2.97124624e-01
-3.95850778e-01 -2.32455969e-01 3.68521512e-01 9.90423977e-01
2.83163816e-01 -1.90457433e-01 3.78990263e-01 -8.37920785e-01
6.87246680e-01 -6.65658992e-03 -9.86750722e-01 1.51544258e-01
-9.02247250e-01 4.42007720e-01 5.39938033e-01 -5.91461420e-01
-1.07640016e+00 2.02293009e-01 -1.00622177e-01 -4.03452396e-01
-3.87431644e-02 6.17784977e-01 2.06437677e-01 -2.62699068e-01
8.79844368e-01 1.66525826e-01 -8.28416198e-02 -4.92657959e-01
4.17565972e-01 1.89329952e-01 5.01391113e-01 -4.50369447e-01
9.69221294e-01 2.79081672e-01 -3.02565303e-02 -8.91077816e-01
-1.14015448e+00 -1.80307359e-01 -4.33813512e-01 -8.30109864e-02
3.56359005e-01 -1.25556338e+00 -8.50516677e-01 4.21527207e-01
-1.00159132e+00 -5.31000733e-01 -2.68670380e-01 9.14453864e-01
-5.37942052e-01 5.67143083e-01 -8.03774834e-01 -1.09870410e+00
-1.28404155e-01 -1.10473001e+00 5.54721951e-01 7.64649361e-02
-3.56537759e-01 -1.19770396e+00 2.40696035e-02 -3.00733119e-01
5.98975778e-01 -3.52814108e-01 3.90959114e-01 -6.18448138e-01
-4.49779809e-01 -9.94459018e-02 -1.91272497e-02 6.71886683e-01
-4.80200946e-02 1.50918126e-01 -1.17690170e+00 -4.54731941e-01
3.08967829e-01 1.17555119e-01 9.79524732e-01 7.87119210e-01
4.95759100e-01 -3.56560379e-01 -3.01481158e-01 5.96255958e-01
1.31732178e+00 -2.20622141e-02 5.27423561e-01 3.76264453e-01
4.52914566e-01 9.00517851e-02 4.60636914e-01 5.57152331e-01
7.36236498e-02 5.03632009e-01 -4.01675403e-02 3.47558677e-01
1.89777955e-01 -2.13026837e-01 7.18706250e-01 1.24977851e+00
-1.52005898e-02 3.61580215e-02 -8.63151670e-01 2.06903726e-01
-1.93903852e+00 -9.00927544e-01 -4.01091725e-01 2.39434290e+00
8.77577424e-01 3.82288992e-01 -1.94542840e-01 -1.28767207e-01
2.61756033e-01 7.70992264e-02 -5.36979795e-01 -1.30683333e-01
-1.48952797e-01 -2.36962978e-02 1.07422340e+00 9.84840572e-01
-1.25789988e+00 1.10775399e+00 8.43386269e+00 7.70914316e-01
-9.56612587e-01 2.11922318e-01 -6.48679733e-02 -2.08088353e-01
-4.75905016e-02 3.11132103e-01 -1.26605546e+00 4.45274889e-01
1.44851840e+00 -2.38137335e-01 5.38225472e-01 7.57705331e-01
6.64778471e-01 -3.00650030e-01 -9.15493846e-01 8.61207187e-01
-5.33397913e-01 -1.06274939e+00 -4.46932852e-01 -3.21949720e-02
8.78995121e-01 4.07779753e-01 2.65257917e-02 3.47542852e-01
6.11225307e-01 -6.88306153e-01 8.84697795e-01 7.28143930e-01
2.55084217e-01 -4.30305302e-01 5.30396104e-01 5.69528282e-01
-8.38932872e-01 -2.28447482e-01 -6.21736050e-01 -3.24237049e-01
3.86140078e-01 1.06220257e+00 -5.82653761e-01 2.32915446e-01
4.88742292e-01 5.60730577e-01 -1.15185544e-01 1.15085196e+00
-2.71110326e-01 9.70702529e-01 -6.50545776e-01 1.61427528e-01
2.14232534e-01 -4.21529591e-01 6.93266273e-01 1.36297858e+00
5.51573396e-01 6.46098405e-02 -4.46524657e-02 6.16780400e-01
2.91489422e-01 -2.96620429e-01 -4.64310944e-01 -1.31187774e-03
4.55567598e-01 8.91641438e-01 -3.05026174e-01 -4.05018330e-01
-4.60968882e-01 6.38042629e-01 2.42411882e-01 7.66075075e-01
-7.34237075e-01 -9.43857580e-02 1.28236794e+00 -3.10989916e-01
2.45094657e-01 -7.06686616e-01 -1.83198139e-01 -1.55906272e+00
-9.16961506e-02 -1.09899592e+00 -9.15102381e-03 -5.38202047e-01
-1.11753201e+00 3.87524933e-01 -3.02297138e-02 -1.20917690e+00
-5.09979606e-01 -8.78281116e-01 -1.62823722e-01 7.82384038e-01
-1.31261373e+00 -4.98422205e-01 2.10509375e-01 2.54667222e-01
1.16692722e-01 2.30842501e-01 7.87902951e-01 2.30151176e-01
-8.03823471e-01 4.69117939e-01 8.61734569e-01 -1.76405385e-01
8.89141679e-01 -1.29218817e+00 5.88744640e-01 1.09896505e+00
5.15860394e-02 1.14367187e+00 1.40306044e+00 -7.77905524e-01
-1.48851466e+00 -6.63784564e-01 7.93876052e-01 -4.20121968e-01
1.14397454e+00 -3.04235131e-01 -9.10439789e-01 1.19785452e+00
8.87015983e-02 -8.17970261e-02 2.66696244e-01 2.86199957e-01
-2.46590182e-01 4.00754949e-03 -5.35467684e-01 6.10837579e-01
1.02191699e+00 -8.04474652e-01 -5.58317363e-01 4.74308819e-01
6.22016311e-01 -7.88000703e-01 -9.17574048e-01 4.01797116e-01
7.09587336e-01 -7.00919926e-01 7.48527169e-01 -8.61612678e-01
-4.27159965e-01 -4.64518726e-01 -3.46284509e-01 -1.53673053e+00
-4.10355419e-01 -1.15442848e+00 -3.23215574e-01 9.13681507e-01
5.94785392e-01 -9.21586633e-01 6.86284602e-01 8.81807923e-01
-8.06109793e-03 -4.78010103e-02 -9.47832942e-01 -1.14561379e+00
8.65259990e-02 -1.07299471e+00 6.11306652e-02 8.12693954e-01
-2.29939409e-02 1.86439708e-01 -7.61650741e-01 5.50206482e-01
6.17570698e-01 -3.03948313e-01 7.26439118e-01 -1.19521523e+00
-5.56276917e-01 -2.82059610e-01 -9.56341773e-02 -1.58011389e+00
2.16458961e-01 -2.73412496e-01 3.75423372e-01 -1.06675017e+00
-3.20211798e-01 -2.24199697e-01 -3.45812738e-01 3.28763187e-01
-1.30739331e-01 -2.78078228e-01 1.04050919e-01 7.08716065e-02
-4.68971670e-01 3.31608951e-01 8.74249816e-01 1.32930011e-01
-1.83120891e-01 2.37147927e-01 -4.60170656e-01 9.93589878e-01
7.09949195e-01 -5.21363139e-01 -3.18008959e-01 -4.88759935e-01
5.69447637e-01 -3.74587551e-02 4.04525936e-01 -1.27938724e+00
3.93392026e-01 -3.97925414e-02 2.14188293e-01 -4.80107218e-01
3.63370597e-01 -7.74437129e-01 4.51664150e-01 5.25126815e-01
-3.32935095e-01 5.35197631e-02 4.78751004e-01 7.31964290e-01
7.62009397e-02 -3.34789336e-01 7.57458031e-01 2.56911032e-02
-6.13752544e-01 -1.60250470e-01 -7.99640834e-01 3.33629549e-02
4.02884424e-01 1.59134373e-01 -4.87704694e-01 -8.37389350e-01
-1.11647880e+00 4.78906818e-02 2.86466300e-01 8.82760249e-03
1.53927833e-01 -1.03268409e+00 -3.70011181e-01 1.69891834e-01
-4.68815029e-01 -5.76215386e-01 2.08859831e-01 1.22378290e+00
-9.77297798e-02 6.52515709e-01 4.52626526e-01 -6.24662042e-01
-1.04526007e+00 1.39715150e-01 7.23502457e-01 -3.61249596e-01
-5.92896163e-01 7.19363213e-01 3.84734944e-02 -4.80959356e-01
3.95417988e-01 -9.42981362e-01 3.64721119e-01 -1.55128092e-01
5.14303863e-01 4.83361632e-01 2.85243094e-01 -3.98662388e-01
-2.72443801e-01 6.10892117e-01 4.85278144e-02 -4.09499109e-01
8.51996243e-01 -4.51643378e-01 1.00014806e-01 7.02739775e-01
9.57539201e-01 -2.72463504e-02 -1.45504236e+00 -4.93824154e-01
3.50404888e-01 -2.78174520e-01 3.20219129e-01 -7.74261832e-01
-6.74754620e-01 8.43589187e-01 7.00397968e-01 2.07407191e-01
7.49189496e-01 -5.93177676e-01 5.17145038e-01 9.70523477e-01
4.72581953e-01 -1.37013459e+00 -6.18207812e-01 1.11686349e+00
4.06403214e-01 -8.91534507e-01 3.46953757e-02 -1.92584127e-01
-4.44831431e-01 1.00342274e+00 3.16003442e-01 -1.26426190e-01
9.64613557e-01 6.79646790e-01 3.13550919e-01 3.62599224e-01
-1.03685212e+00 -2.01811463e-01 2.29368165e-01 3.73920500e-01
2.35253364e-01 -3.91009264e-02 -1.57205313e-01 5.91196418e-01
-3.16022456e-01 -4.94071320e-02 4.77146536e-01 6.11233294e-01
-7.07074940e-01 -1.02925944e+00 -4.45176691e-01 6.61123157e-01
-5.82368374e-01 -2.29956582e-01 1.11124299e-01 8.23706150e-01
-4.20105308e-01 1.05812144e+00 -2.14149207e-01 -5.31893253e-01
5.09409785e-01 3.87475789e-01 5.45752227e-01 -1.29691988e-01
-5.66413701e-01 3.53363067e-01 3.83060336e-01 -8.40165913e-01
-1.60950229e-01 -8.41940761e-01 -1.14699495e+00 -5.73887944e-01
-7.29842305e-01 1.28783688e-01 6.01799190e-01 1.31211269e+00
3.68905634e-01 3.05236042e-01 1.19803689e-01 -5.43195069e-01
-1.34952796e+00 -9.87285972e-01 -6.45873070e-01 -2.04741713e-02
5.59144616e-01 -6.04009271e-01 -7.52832711e-01 -6.53647631e-02]
|
[6.6999101638793945, 3.7128689289093018]
|
fe9b2153-3b8f-48b9-a33f-05a39b34a6de
|
person-search-new-paradigm-of-person-re
| null | null |
https://www.sciencedirect.com/science/article/pii/S0262885620301025
|
https://reader.elsevier.com/reader/sd/pii/S0262885620301025?token=F42274C35788A55FD4B3B3605EAC08CE31ADEA7600758EF9217B69A4EBFAF97C56F5B0BEDEB837CCF0AEDC4C4D2DB249
|
Person search: New paradigm of person re-identification: A survey and outlook of recent works
|
Person Search (PS) has become a major field because of its need in community and in the field of research among researchers. This task aims to find a probe person from whole scene which shows great significance in video surveillance field to track lost people, re-identification, and verification of person. In last few years, deep learning has played unremarkable role for the solution of re-identification problem. Deep learning shows incredible performance in person (re-ID) and search. Researchers experience more flexibility in proposing new methods and solve challenging issues such as low resolution, pose variation, background clutter, occlusion, viewpoints, and low illumination. Specially, convolutional neural network (CNN) achieves breakthrough performance and extracts useful patterns and characteristics. Development of new framework takes substantial efforts; hard work and computation cost are required to acquire excellent results. This survey paper includes brief discussion about feature representation learning and deep metric learning with novel loss functions. We thoroughly review datasets with performance analysis on existing datasets. Finally, we are reviewing current solutions for further consideration.
|
['Khawar Islam']
|
2020-06-30
| null | null | null |
image-and-vision-computing-2020-6
|
['person-search']
|
['computer-vision']
|
[-1.31304294e-01 -7.51780987e-01 1.76222101e-01 -5.71914136e-01
-5.35801589e-01 -5.42134941e-01 5.15072942e-01 6.33560307e-03
-5.68986356e-01 6.72980011e-01 2.98444688e-01 3.78552645e-01
-1.89468399e-01 -6.48324907e-01 -2.21881539e-01 -5.73517263e-01
-6.35121465e-02 3.37931246e-01 1.43931076e-01 -1.45894140e-01
2.06129164e-01 7.16737032e-01 -1.58082223e+00 -9.54403654e-02
5.54568112e-01 9.13497865e-01 3.45592164e-02 4.60982978e-01
1.60936370e-01 4.73434269e-01 -8.97725463e-01 -7.30776429e-01
5.02864957e-01 -1.58172280e-01 -7.69395471e-01 9.32373703e-02
7.65946209e-01 -3.64192218e-01 -8.37686658e-01 1.06421900e+00
1.05940068e+00 4.99348760e-01 4.07512128e-01 -1.38107550e+00
-7.25446701e-01 -6.22831173e-02 -7.73834825e-01 7.35960245e-01
5.91593385e-01 2.99873445e-02 4.16939735e-01 -6.47261918e-01
2.82282699e-02 1.34793985e+00 9.40323114e-01 6.09892488e-01
-5.46193004e-01 -6.33385241e-01 -3.06070987e-02 8.78738761e-01
-1.53582323e+00 -3.70920569e-01 5.20538926e-01 -3.34079355e-01
7.95495212e-01 4.56956238e-01 5.52875817e-01 1.13271332e+00
-1.52053013e-01 8.05985451e-01 6.60818994e-01 -2.25784808e-01
-2.75481284e-01 1.83525056e-01 3.20915490e-01 6.91382170e-01
5.37977695e-01 1.17467448e-01 -5.28477669e-01 -1.86673761e-03
5.80203652e-01 3.80349398e-01 -3.56221020e-01 -1.66005507e-01
-1.02684128e+00 4.48172271e-01 4.60772634e-01 1.99816033e-01
-1.77372977e-01 8.96821246e-02 4.76965874e-01 3.25800031e-01
-1.13118477e-01 2.73519367e-01 -3.51238221e-01 -2.20671296e-01
-8.04898202e-01 4.53449816e-01 4.03934747e-01 9.43023562e-01
3.12074274e-01 4.70120646e-02 -4.90032703e-01 1.06165111e+00
-1.59074098e-01 7.37781286e-01 5.09372115e-01 -8.95389676e-01
2.39309371e-01 5.80999911e-01 3.53032738e-01 -1.55975819e+00
-4.75316912e-01 -7.95624077e-01 -1.16382539e+00 -9.41254571e-02
2.99769402e-01 -1.16140544e-01 -6.81739450e-01 1.50899017e+00
3.03857893e-01 2.72809535e-01 -1.84116900e-01 1.10546172e+00
1.09033418e+00 5.33192754e-01 -1.15061074e-01 -3.71472649e-02
1.47692168e+00 -1.05691481e+00 -4.96265709e-01 -1.00978583e-01
-1.65377006e-01 -6.80483460e-01 4.83139992e-01 2.53851980e-01
-8.97140920e-01 -9.59879816e-01 -7.75216877e-01 2.76305969e-03
-4.10628468e-01 3.38965684e-01 5.26792526e-01 8.02959144e-01
-1.00700152e+00 3.89671445e-01 -5.03512740e-01 -9.45021033e-01
6.26432657e-01 6.64643049e-01 -4.78898942e-01 -2.08003417e-01
-9.90347266e-01 6.92873776e-01 -1.08461209e-01 4.34117764e-01
-9.26955223e-01 -2.36528754e-01 -5.53211987e-01 4.33381880e-03
2.61993438e-01 -9.16409373e-01 8.34997833e-01 -6.71769440e-01
-9.59369540e-01 1.08199859e+00 -2.40816787e-01 -4.54211950e-01
6.83661819e-01 -4.17910457e-01 -7.93796718e-01 -1.28692821e-01
2.83296615e-01 3.89440000e-01 8.75680089e-01 -7.93099284e-01
-8.97746980e-01 -6.81093872e-01 3.77812609e-02 1.89790443e-01
-5.00345230e-01 4.71049547e-01 -7.59876013e-01 -5.44053316e-01
1.40756993e-02 -7.51924515e-01 -7.23936930e-02 -3.83857302e-02
-1.44048154e-01 -6.00735605e-01 6.78559959e-01 -7.67290473e-01
1.07513642e+00 -1.75966775e+00 -9.81898680e-02 -1.34064704e-01
1.45467058e-01 5.72329283e-01 4.68886569e-02 2.53255755e-01
1.98637784e-01 -2.48838842e-01 1.15980580e-01 -3.13593745e-01
-8.76675546e-02 -2.11104438e-01 4.79016006e-02 6.91052020e-01
-1.03761300e-01 8.05476069e-01 -6.34917617e-01 -4.74615812e-01
3.36777747e-01 4.94204253e-01 -1.46233469e-01 2.61293143e-01
5.44437230e-01 4.63045388e-01 -4.58160400e-01 9.38947320e-01
9.85699654e-01 -2.95886099e-01 -4.42190796e-01 -4.78285730e-01
-2.25473613e-01 -4.42531973e-01 -1.31672883e+00 1.45180869e+00
1.14397548e-01 8.09558868e-01 6.32693432e-03 -1.16013026e+00
8.75694454e-01 7.67145753e-02 4.51226860e-01 -7.86882162e-01
3.35364312e-01 -7.07547143e-02 -3.60866636e-01 -8.91160607e-01
5.14628649e-01 3.68666768e-01 1.75320936e-04 1.79387093e-01
-1.81364611e-01 8.76259565e-01 6.66542500e-02 -9.29261893e-02
9.98675287e-01 -2.56645411e-01 1.57173634e-01 -2.00046882e-01
1.16572332e+00 -1.83968022e-01 7.30741203e-01 7.75839508e-01
-7.87442267e-01 5.94898999e-01 -1.99360490e-01 -1.17419684e+00
-8.77709687e-01 -8.93691719e-01 -9.78834927e-02 9.82131064e-01
5.48359573e-01 2.79542319e-02 -5.04334331e-01 -3.26692075e-01
8.31392109e-02 -1.47392243e-01 -5.82567811e-01 -1.12344310e-01
-9.01200950e-01 -9.52364206e-01 6.49285853e-01 5.53722739e-01
1.22931242e+00 -9.73624468e-01 -5.97303450e-01 -3.01700477e-02
-3.25907290e-01 -1.10533214e+00 -5.67770720e-01 -5.49876392e-01
-4.73571509e-01 -1.37633574e+00 -1.37070477e+00 -1.30358219e+00
5.87585926e-01 7.88639367e-01 8.84722769e-01 1.59738600e-01
-7.48843968e-01 6.01755977e-01 -2.17961654e-01 -2.94765592e-01
3.80437523e-01 -1.08294919e-01 3.90971392e-01 4.26350273e-02
8.70866299e-01 -1.54530793e-01 -1.09060073e+00 5.39601624e-01
-3.81382346e-01 -5.62368929e-01 1.26186118e-01 6.36918962e-01
1.42922819e-01 4.49907184e-01 4.59215015e-01 -3.22920643e-02
6.96646452e-01 -2.33133823e-01 -6.18887424e-01 5.14873087e-01
-3.81960273e-01 -3.04345012e-01 2.16444373e-01 -3.83066051e-02
-1.01395297e+00 -2.82040358e-01 1.83961131e-02 -1.23162262e-01
-2.54035830e-01 -1.49092600e-01 -1.14597365e-01 -3.41502726e-01
5.79470277e-01 4.21564788e-01 -1.38006523e-01 -7.25153804e-01
-2.06303537e-01 8.15989435e-01 8.37076306e-01 -4.07919884e-01
7.23840892e-01 6.30299091e-01 -3.66184972e-02 -7.52969086e-01
-6.67512953e-01 -7.09558487e-01 -4.21898097e-01 -3.03049445e-01
7.52146184e-01 -1.07431078e+00 -1.11691236e+00 8.65824282e-01
-1.22836637e+00 4.11919624e-01 2.31531784e-01 3.29101413e-01
1.02530755e-01 6.94771826e-01 -2.59502590e-01 -6.82160139e-01
-6.57213211e-01 -1.02283275e+00 8.68390083e-01 8.53217006e-01
1.69167548e-01 -6.38655066e-01 9.48483422e-02 6.53043270e-01
5.70684195e-01 2.84247220e-01 1.13200746e-01 -3.51327986e-01
-7.06571579e-01 -6.92400038e-01 -5.14554381e-01 2.48592243e-01
3.33146960e-01 -5.00218570e-01 -9.85421658e-01 -7.24790156e-01
-5.67662269e-02 -7.79741630e-02 1.00988841e+00 4.65674967e-01
1.35231805e+00 -1.50604755e-01 -6.08676791e-01 9.13228869e-01
1.33118021e+00 3.20465267e-01 3.97200912e-01 6.12511277e-01
7.11708188e-01 5.42063594e-01 3.16468745e-01 5.78033566e-01
5.75071812e-01 8.04883659e-01 6.63276315e-02 2.36298814e-02
-2.84992933e-01 9.99613106e-02 -6.94036037e-02 1.22655287e-01
-3.83701861e-01 -3.83257419e-01 -8.19374621e-01 6.01395190e-01
-1.76528406e+00 -1.34826636e+00 2.42918879e-02 2.13431287e+00
2.15574279e-01 -2.98973799e-01 3.39139372e-01 1.03298642e-01
1.07648849e+00 5.93981706e-02 -5.84765553e-01 2.59952277e-01
-4.73980993e-01 -1.97783619e-01 3.63978744e-01 2.74057567e-01
-1.34442902e+00 8.11117232e-01 5.83612919e+00 5.72611332e-01
-9.03050840e-01 4.29654643e-02 6.31377220e-01 -8.27593431e-02
3.78107876e-01 -5.62529206e-01 -1.15717804e+00 6.99247301e-01
2.16299281e-01 -8.56963545e-02 3.64172101e-01 7.65333176e-01
6.65716752e-02 -1.89115014e-02 -1.00187612e+00 2.00507188e+00
5.14423847e-01 -1.20599651e+00 -5.40504940e-02 -2.42427990e-01
6.91262960e-01 -1.49599969e-01 1.78210676e-01 2.06368819e-01
-1.34629920e-01 -1.08632505e+00 2.69639730e-01 6.87077582e-01
6.18720889e-01 -8.25495183e-01 1.18096125e+00 2.40016524e-02
-1.34283245e+00 -4.19816524e-01 -7.69830704e-01 -7.95155540e-02
2.76027650e-01 2.02407017e-01 -3.32001388e-01 2.95157045e-01
1.25905156e+00 8.40917945e-01 -6.65147245e-01 1.59880245e+00
3.93910080e-01 4.16321531e-02 -1.84756443e-01 7.84162432e-02
9.02631506e-03 -4.06516045e-02 6.52088225e-01 1.29958582e+00
2.73909777e-01 2.36113846e-01 2.84456611e-01 4.18575585e-01
4.78228554e-03 -1.83080956e-01 -3.27450037e-01 3.43090981e-01
3.62020880e-01 1.08687592e+00 -5.17076790e-01 -2.32532427e-01
-3.19846421e-01 1.25830805e+00 1.45998761e-01 4.04779255e-01
-7.32670486e-01 -2.99881041e-01 9.13938820e-01 8.49654749e-02
8.03229660e-02 -2.44067118e-01 -4.28541750e-03 -1.13402390e+00
1.48820534e-01 -8.84326696e-01 7.15702891e-01 -3.80024672e-01
-1.31760299e+00 7.43680656e-01 -1.15092814e-01 -1.10220623e+00
1.66076332e-01 -5.53685188e-01 -5.24101555e-01 8.58389258e-01
-1.43350983e+00 -1.04265988e+00 -1.07381880e+00 9.39630270e-01
8.27322066e-01 -8.68279576e-01 4.75531906e-01 7.55766749e-01
-8.63875329e-01 1.01323211e+00 7.17053115e-02 4.80150461e-01
8.44304144e-01 -7.85154879e-01 3.14769626e-01 9.09977794e-01
-2.16417924e-01 6.83344662e-01 6.21417403e-01 -4.85354483e-01
-1.35013866e+00 -9.04370010e-01 9.04890776e-01 -6.52012587e-01
-1.05999075e-01 -4.83689569e-02 -4.55520928e-01 2.95723736e-01
1.69207230e-01 3.78753245e-02 5.76375723e-01 -4.66440842e-02
-4.28798562e-03 -7.13720083e-01 -1.20520425e+00 2.79409349e-01
1.16378200e+00 -2.12199032e-01 -2.59822607e-01 3.36208373e-01
2.97344923e-01 -2.68053561e-01 -2.19858021e-01 3.88614058e-01
6.31021321e-01 -1.13246763e+00 1.38609397e+00 -5.12558997e-01
-4.18293685e-01 -3.39269251e-01 -7.83871859e-02 -6.92991853e-01
-7.27345884e-01 -7.07506597e-01 5.74292466e-02 1.17853975e+00
-3.20784390e-01 -4.59080637e-01 1.05840874e+00 7.52945364e-01
7.09601790e-02 -4.61342037e-01 -7.04826713e-01 -7.68265545e-01
-4.32299823e-01 1.94739718e-02 7.69064724e-01 6.97754860e-01
-5.39881229e-01 1.72716752e-01 -8.65159988e-01 3.72065902e-01
1.04223371e+00 1.52217358e-01 9.53217268e-01 -1.27835011e+00
7.22076744e-02 -2.66724497e-01 -8.40154827e-01 -1.06930828e+00
-2.19764709e-01 -4.41986233e-01 -3.24103326e-01 -1.63315558e+00
5.92614532e-01 -5.01188338e-01 -5.36223292e-01 1.98684543e-01
-2.25197241e-01 2.20496714e-01 8.32289159e-02 3.44486952e-01
-9.07066524e-01 6.06785297e-01 9.57317650e-01 -4.59004819e-01
-1.64346828e-04 3.42021167e-01 -6.44989133e-01 6.30510569e-01
8.71325493e-01 -2.32103229e-01 -2.16022879e-01 -1.05724299e+00
9.25901532e-02 -1.90943837e-01 8.43571723e-01 -1.48512757e+00
6.59183979e-01 6.45921826e-02 9.53345060e-01 -7.70038068e-01
5.19193530e-01 -8.97573233e-01 1.62810251e-01 4.26314622e-01
-1.10115796e-01 4.06666487e-01 4.30200025e-02 5.75178385e-01
-2.88314462e-01 -1.47120163e-01 7.24706471e-01 -4.67940748e-01
-1.13182259e+00 7.40630209e-01 1.76545426e-01 -3.99738625e-02
9.95457709e-01 -6.44071162e-01 -1.65009215e-01 -4.13408816e-01
-4.72917110e-01 4.85502630e-01 1.92434281e-01 6.81748152e-01
7.04227507e-01 -1.30197656e+00 -8.15969884e-01 2.87597179e-01
-1.56652197e-01 -2.04822868e-01 5.83072841e-01 4.02213842e-01
-6.05495393e-01 6.22010231e-01 -4.33178037e-01 -5.02458096e-01
-1.60656524e+00 4.84037250e-01 5.70169330e-01 1.56229123e-01
-4.86608803e-01 1.31839120e+00 1.06287666e-01 -2.96975896e-02
6.61622405e-01 2.56945282e-01 -4.18188214e-01 -1.10296860e-01
9.35506821e-01 1.05384672e+00 -2.08880335e-01 -7.65863836e-01
-7.37301648e-01 9.49395478e-01 -1.87276483e-01 4.46488857e-01
1.25019407e+00 -4.01741952e-01 3.74316685e-02 -3.61780912e-01
1.18785298e+00 -2.67542243e-01 -1.13022327e+00 -3.59349161e-01
-1.97712019e-01 -6.64336801e-01 -1.77240685e-01 -7.48949289e-01
-1.03430378e+00 8.03171158e-01 1.21923089e+00 -9.42389965e-02
1.02315545e+00 -1.23494603e-01 1.09936070e+00 5.89554071e-01
6.38128042e-01 -1.19020569e+00 3.75572145e-01 3.24731261e-01
9.31626320e-01 -1.79394972e+00 9.42205787e-02 -4.48513851e-02
-3.41864288e-01 9.55705702e-01 7.70960152e-01 -1.25023976e-01
7.18000472e-01 -2.44284272e-01 1.50932418e-03 -2.38724187e-01
2.45246049e-02 -1.15455061e-01 2.61930466e-01 9.33756411e-01
7.82351270e-02 -1.02401979e-01 3.59999649e-02 5.70339322e-01
-2.04220861e-01 -1.05765365e-01 -1.51337504e-01 7.47676015e-01
-6.35304987e-01 -9.22476172e-01 -5.72436750e-01 3.13443482e-01
-4.22431558e-01 1.75550103e-01 -1.98187098e-01 4.35903728e-01
4.42505717e-01 1.10531151e+00 3.09412200e-02 -3.60369533e-01
3.36946100e-01 -4.62079018e-01 5.91376901e-01 -3.27456109e-02
-3.67030919e-01 -3.92309517e-01 -2.30879739e-01 -4.71096158e-01
-5.49498320e-01 -6.50290430e-01 -6.19121909e-01 -6.17511213e-01
8.44617095e-03 1.93169802e-01 3.54061097e-01 7.16676116e-01
3.51517081e-01 1.91624999e-01 5.10572910e-01 -5.25860369e-01
-4.91988271e-01 -7.62234628e-01 -2.53195912e-01 4.65991080e-01
4.89367843e-01 -6.94518328e-01 7.15246610e-03 7.44675919e-02]
|
[14.706201553344727, 0.9139581322669983]
|
8b72549e-230e-4fa9-af46-54a346ece7cb
|
control-theoretically-explainable-application
|
2208.01291
| null |
https://arxiv.org/abs/2208.01291v2
|
https://arxiv.org/pdf/2208.01291v2.pdf
|
Control theoretically explainable application of autoencoder methods to fault detection in nonlinear dynamic systems
|
This paper is dedicated to control theoretically explainable application of autoencoders to optimal fault detection in nonlinear dynamic systems. Autoencoder-based learning is a standard machine learning method and widely applied for fault (anomaly) detection and classification. In the context of representation learning, the so-called latent (hidden) variable plays an important role towards an optimal fault detection. In ideal case, the latent variable should be a minimal sufficient statistic. The existing autoencoder-based fault detection schemes are mainly application-oriented, and few efforts have been devoted to optimal autoencoder-based fault detection and explainable applications. The main objective of our work is to establish a framework for learning autoencoder-based optimal fault detection in nonlinear dynamic systems. To this aim, a process model form for dynamic systems is firstly introduced with the aid of control theory, which also leads to a clear system interpretation of the latent variable. The major efforts are made on the development of a control theoretic solution to the optimal fault detection problem, in which an analog concept to minimal sufficient statistic, the so-called lossless information compression, is introduced and proven for dynamic systems and fault detection specifications. In particular, the existence conditions for such a latent variable are derived, based on which a loss function and further a learning algorithm are developed. This learning algorithm enables optimally training of autoencoders to achieve an optimal fault detection in nonlinear dynamic systems. A case study on three-tank system is given at the end of this paper to illustrate the capability of the proposed autoencoder-based fault detection and to explain the essential role of the latent variable in the proposed fault detection system.
|
['Ting Xue', 'Zhiwen Chen', 'Ketian Liang', 'Steven X. Ding', 'Linlin Li']
|
2022-08-02
| null | null | null | null |
['fault-detection']
|
['miscellaneous']
|
[-1.08471839e-03 1.35146916e-01 1.64809406e-01 2.48097658e-01
1.87018245e-01 8.27698410e-02 1.78524926e-01 -7.62550160e-02
2.43776113e-01 4.55434948e-01 -2.86411881e-01 -1.84516534e-01
-7.97986507e-01 -6.73127592e-01 -7.54833937e-01 -9.77534354e-01
-1.48845434e-01 3.05534333e-01 -3.37892383e-01 -3.17167222e-01
1.50371924e-01 6.52148724e-01 -1.81681323e+00 -2.61443734e-01
7.84813106e-01 1.04216588e+00 4.17621374e-01 6.02234364e-01
2.07928255e-01 6.59086049e-01 -7.85273731e-01 2.38349155e-01
1.32727623e-01 -7.77218163e-01 -6.89858794e-01 5.98970354e-01
-3.62079293e-01 -3.62027377e-01 -4.91334021e-01 1.02690530e+00
2.99331486e-01 3.18235248e-01 1.04805815e+00 -1.31907547e+00
-6.13285482e-01 3.14933598e-01 4.34299380e-01 3.11340779e-01
-3.84253003e-02 4.21314947e-02 6.72121406e-01 -8.53540421e-01
7.16769621e-02 9.08014417e-01 3.05641472e-01 3.90434176e-01
-8.03251326e-01 2.15377696e-02 -3.38151544e-01 6.50667369e-01
-1.28565061e+00 1.33362156e-03 1.00547063e+00 -6.81138813e-01
1.03694963e+00 1.82813540e-01 8.07971835e-01 5.83115280e-01
1.09300709e+00 5.12938559e-01 7.74798393e-01 -7.12459385e-01
5.92167139e-01 -1.30943675e-02 1.63005576e-01 7.02047408e-01
4.91046458e-01 3.76496851e-01 1.00680418e-01 1.11728303e-01
9.43298042e-01 5.34620106e-01 -5.05308628e-01 -1.61367372e-01
-5.33184469e-01 8.64930212e-01 3.06774795e-01 8.49199593e-01
-6.35528028e-01 -5.65433837e-02 3.64098787e-01 7.25713134e-01
3.61952960e-01 5.18273294e-01 -3.34577501e-01 1.67317823e-01
-4.61219221e-01 6.35371730e-02 9.08837676e-01 4.72383052e-01
5.31826615e-01 1.03875387e+00 2.03293562e-01 4.80859309e-01
3.06889623e-01 6.00673735e-01 1.03723192e+00 -7.55151153e-01
-8.79044384e-02 7.28979886e-01 -9.90932658e-02 -1.06538761e+00
-5.03808782e-02 -6.27745271e-01 -9.52003181e-01 5.17813265e-01
-2.70233393e-01 -1.77279755e-01 -6.46147311e-01 1.12286496e+00
5.75232916e-02 2.56021738e-01 6.46179497e-01 8.77291918e-01
6.61759463e-04 8.47558260e-01 -5.42999089e-01 -6.58502281e-01
1.09236860e+00 -4.79477435e-01 -1.26786137e+00 2.08436340e-01
5.01121163e-01 -3.55925560e-01 7.49779344e-01 4.72635210e-01
-7.43966877e-01 -5.38295865e-01 -1.20166695e+00 7.14602053e-01
-3.21318924e-01 1.19355418e-01 2.06563741e-01 2.87220359e-01
-7.48251379e-01 9.45420206e-01 -8.97110879e-01 -3.02567422e-01
-3.73784959e-01 3.08249503e-01 -4.31151688e-01 8.81750584e-02
-1.40617430e+00 1.20401132e+00 6.00327075e-01 4.90516990e-01
-1.25112295e+00 -3.55666339e-01 -9.22655404e-01 2.63791829e-01
3.36744905e-01 -6.53154910e-01 1.06516957e+00 -1.17508316e+00
-1.51554728e+00 -1.34686532e-03 5.11915013e-02 -8.39752376e-01
2.10029423e-01 -2.89922476e-01 -7.15055704e-01 5.23835480e-01
-2.27220029e-01 -4.62667912e-01 1.37803328e+00 -1.12817407e+00
-3.19047898e-01 -1.07694834e-01 -3.04151058e-01 4.14502546e-02
-6.48545206e-01 -4.22754228e-01 3.63133222e-01 -5.71576476e-01
3.02075177e-01 -4.37312067e-01 -6.96687680e-03 -4.63806599e-01
-1.79177836e-01 -2.41925970e-01 1.18840563e+00 -7.36501276e-01
1.13031793e+00 -1.94628000e+00 3.41370076e-01 3.48463297e-01
-3.05626988e-02 1.62082061e-01 4.03948307e-01 6.73559964e-01
-4.51339632e-01 -3.29512954e-01 -5.75429499e-01 1.79196134e-01
-2.21920609e-01 6.05977833e-01 -3.77072662e-01 6.02809310e-01
4.05366093e-01 1.49252519e-01 -4.14727628e-01 -1.00618809e-01
8.18554521e-01 4.99476284e-01 -3.91320735e-01 4.96104032e-01
3.05529386e-02 1.65523067e-01 -5.09758711e-01 4.86458272e-01
1.74034238e-01 1.01078175e-01 -1.34897962e-01 -2.96206504e-01
-1.49085924e-01 -4.86327797e-01 -1.17899024e+00 6.77264512e-01
-5.86651325e-01 5.29975176e-01 5.47910444e-02 -1.68514442e+00
1.27263999e+00 8.66577327e-01 8.78629208e-01 -3.31210434e-01
3.97090495e-01 2.88904190e-01 -6.40962347e-02 -9.28337574e-01
8.43541846e-02 -4.45231795e-01 1.33533105e-01 6.29473627e-02
2.07337990e-01 -1.82664797e-01 5.34742177e-02 -2.26154193e-01
1.05607915e+00 -4.68465418e-01 3.88636708e-01 -4.56827313e-01
1.03320670e+00 -4.92853820e-02 5.42016923e-01 2.78751224e-01
6.98856041e-02 1.57806203e-01 2.12524205e-01 -4.99612659e-01
-1.17916131e+00 -5.55515945e-01 -2.22578481e-01 -7.89136291e-02
1.64007261e-01 5.66467382e-02 -8.22590172e-01 -2.92270929e-01
2.37798899e-01 6.78355217e-01 -4.37203467e-01 -7.81707108e-01
-3.35840076e-01 -4.49946314e-01 3.13191950e-01 3.52302909e-01
5.44140399e-01 -1.00518465e+00 -6.17314160e-01 4.86279994e-01
1.68087542e-01 -4.30716366e-01 3.31098855e-01 5.59607089e-01
-1.21733069e+00 -1.32031286e+00 -5.90553761e-01 -7.33916759e-01
8.83131623e-01 -1.28619716e-01 5.03358305e-01 1.75350904e-01
-4.01553273e-01 7.70649910e-01 -4.26845878e-01 -3.10614944e-01
-7.74065197e-01 -5.04836023e-01 5.89235842e-01 1.24640681e-01
8.43393877e-02 -3.98996234e-01 -2.68948704e-01 3.53395015e-01
-1.28698361e+00 -5.84671140e-01 7.58676350e-01 1.18470216e+00
5.31937897e-01 1.13374102e+00 7.29804277e-01 -3.90320837e-01
5.95052123e-01 -6.04203999e-01 -7.26944208e-01 7.25302845e-02
-1.09957826e+00 2.34676689e-01 1.06671381e+00 -2.04895601e-01
-9.22046185e-01 -1.15220323e-01 -2.37628639e-01 -9.37274814e-01
-3.35897505e-01 7.02477336e-01 -4.29287285e-01 -6.64900318e-02
4.86714184e-01 6.17403984e-01 5.82809031e-01 -5.31932116e-01
-6.24550730e-02 6.31069005e-01 4.89365131e-01 -3.04936022e-01
7.26153076e-01 1.59943178e-01 2.70869285e-01 -1.08967686e+00
-1.46920189e-01 -3.62910628e-01 -3.78563702e-01 -4.61395442e-01
5.87624907e-01 -5.28213620e-01 -7.60161579e-01 5.61739624e-01
-8.70426059e-01 -1.19206242e-01 -6.86465979e-01 7.62404919e-01
-9.73085701e-01 4.19298559e-01 -9.07968938e-01 -1.01981056e+00
-4.50392067e-01 -1.17081702e+00 5.61915457e-01 -1.92404364e-03
2.93383390e-01 -1.27800584e+00 -3.03960163e-02 -1.39409170e-01
2.87036031e-01 2.10296601e-01 9.34884667e-01 -6.33892953e-01
-2.35755086e-01 -6.08388484e-01 6.45644844e-01 1.02569640e+00
2.15520009e-01 -2.71607172e-02 -5.65232813e-01 -4.10305172e-01
9.39124048e-01 2.48131305e-01 6.61266863e-01 4.81550425e-01
7.84850538e-01 -5.88575780e-01 -1.50415584e-01 2.85057634e-01
1.81512117e+00 4.88134325e-01 5.34075797e-01 3.20687205e-01
4.81636405e-01 4.86528158e-01 5.90545177e-01 6.23270035e-01
-3.41001868e-01 3.04048002e-01 7.17275918e-01 2.51749247e-01
2.66890973e-01 2.29870215e-01 5.84623337e-01 1.34853995e+00
4.00649570e-02 -2.57458508e-01 -5.51513612e-01 5.19381523e-01
-1.56043029e+00 -9.55088437e-01 -4.20061275e-02 2.00984836e+00
2.55735844e-01 -6.92457557e-02 -3.47115219e-01 1.18219292e+00
1.01740241e+00 -3.37641716e-01 -4.22364235e-01 -5.21673381e-01
-1.35198519e-01 -1.37109414e-01 1.97920918e-01 6.43790126e-01
-1.02286971e+00 3.77514035e-01 5.78232002e+00 6.42462373e-01
-1.23996913e+00 -8.80295187e-02 1.88512728e-02 6.08212233e-01
-8.55000243e-02 -5.32463789e-02 -5.49579442e-01 3.75079483e-01
1.15025640e+00 -3.44643831e-01 3.13091487e-01 1.17121148e+00
4.97455657e-01 8.62240642e-02 -7.96349049e-01 8.61127734e-01
2.69393682e-01 -7.21096456e-01 1.32303759e-01 4.92636450e-02
6.34956181e-01 -7.92534709e-01 4.81561124e-02 7.43150786e-02
-3.21454555e-01 -6.03683591e-01 5.80587626e-01 8.75494599e-01
3.08414280e-01 -1.09707344e+00 1.25835323e+00 3.47045779e-01
-8.91720414e-01 -7.20061243e-01 -6.41160429e-01 -3.55666965e-01
1.77898273e-01 1.05427325e+00 -6.81219697e-01 7.71853983e-01
3.96190047e-01 8.05252969e-01 -1.87135920e-01 1.01771140e+00
-3.03015023e-01 7.79366076e-01 -1.54851690e-01 6.49811253e-02
-4.32083905e-02 3.01652066e-02 8.88055384e-01 6.58363044e-01
6.99895680e-01 -3.92190889e-02 1.89452283e-02 8.22744787e-01
5.36713302e-01 3.55321653e-02 -9.33295548e-01 -1.06679656e-01
3.27400118e-01 8.47522616e-01 -5.32891810e-01 -3.31366152e-01
-7.17110708e-02 1.03802621e+00 -3.35048467e-01 3.03914130e-01
-4.51449186e-01 -4.63363826e-01 5.83961189e-01 4.28764522e-02
3.00126374e-01 -5.23772053e-02 4.77526076e-02 -9.79916215e-01
-1.35678932e-01 -6.62612081e-01 4.64015365e-01 -5.86475492e-01
-1.26837087e+00 5.71166515e-01 1.89622298e-01 -1.59372902e+00
-9.09983099e-01 -7.41431713e-01 -6.95548534e-01 6.53537452e-01
-1.05218852e+00 -5.29350638e-01 -1.39414459e-01 8.21622014e-01
6.52458668e-01 -5.97511590e-01 9.44874585e-01 9.76374969e-02
-8.30860198e-01 1.93016574e-01 7.02213764e-01 -2.52037108e-01
1.45669803e-01 -1.30032122e+00 -4.52002794e-01 1.18738854e+00
-1.69429988e-01 3.46539706e-01 1.17582762e+00 -9.51157629e-01
-1.65185952e+00 -1.00218582e+00 6.01760685e-01 5.32028079e-02
4.92154419e-01 3.80953431e-01 -1.10630774e+00 5.26379704e-01
-6.31993040e-02 -9.29525569e-02 1.84590310e-01 -5.46102703e-01
5.49852908e-01 -3.23600382e-01 -1.27412128e+00 2.18939334e-01
8.91381726e-02 -6.16027355e-01 -8.84822726e-01 2.86027551e-01
6.64224386e-01 -5.89933917e-02 -1.15705812e+00 4.50678438e-01
1.34492293e-01 -5.79298854e-01 6.64835393e-01 -3.50560904e-01
4.50554252e-01 -5.40038943e-01 1.23216242e-01 -1.42641866e+00
-3.09901327e-01 -3.03314954e-01 -6.55418515e-01 8.59908402e-01
-2.12862596e-01 -7.47732878e-01 4.42053735e-01 1.99193522e-01
-6.78246021e-01 -8.20453584e-01 -8.58517945e-01 -1.07026386e+00
-1.60244644e-01 -3.56085390e-01 1.86050460e-01 7.30411053e-01
1.39551759e-01 -1.53809279e-01 -3.73192102e-01 6.60707116e-01
5.01793742e-01 -2.29184300e-01 2.34569639e-01 -1.13130963e+00
-3.94223124e-01 -1.67896688e-01 -8.16576660e-01 -4.24742281e-01
2.62702256e-01 -5.71199596e-01 2.95265347e-01 -1.33409059e+00
-4.83559757e-01 2.56010264e-01 -4.60718274e-01 2.19831437e-01
6.19634520e-03 -2.83674181e-01 -1.96579531e-01 5.25805652e-01
4.76080440e-02 8.35031509e-01 9.42283511e-01 -1.89649209e-01
-1.63067073e-01 2.84173578e-01 -5.28047374e-03 5.35446882e-01
9.04749691e-01 -2.58676708e-01 -6.29635632e-01 -5.95016889e-02
-3.32355857e-01 5.02011478e-01 5.76156259e-01 -1.51623535e+00
1.75493896e-01 1.32963568e-01 2.61825681e-01 -4.04734969e-01
6.16965629e-02 -1.49817252e+00 2.38903269e-01 1.09832382e+00
-7.25130141e-02 1.98454827e-01 -6.44682199e-02 7.80444205e-01
-6.86343014e-01 -7.55919278e-01 4.64460731e-01 8.75075608e-02
-9.32060122e-01 3.41162235e-02 -9.70130563e-01 -6.46816850e-01
1.03876257e+00 -2.11323380e-01 1.13038875e-01 -3.80417645e-01
-9.26145613e-01 -2.03613788e-01 6.51653334e-02 1.51607215e-01
1.00241995e+00 -1.19858849e+00 -3.38807851e-01 5.39638281e-01
-1.67857140e-01 -2.74951011e-01 5.58192372e-01 6.82923079e-01
-6.53410375e-01 5.42487741e-01 -3.72164816e-01 -4.52214837e-01
-8.69986117e-01 8.29439640e-01 6.42902493e-01 7.81281963e-02
-7.78208256e-01 2.08235040e-01 -1.65509388e-01 2.46257409e-01
-9.34317987e-03 -3.42947632e-01 -3.21522951e-01 -4.43043679e-01
4.38310057e-01 5.62351108e-01 2.44777173e-01 -6.10501409e-01
-9.53913480e-03 4.70736295e-01 4.25737828e-01 7.41411140e-03
1.18283927e+00 -1.34040743e-01 -1.15624152e-01 5.63203752e-01
9.98862684e-01 -6.55836105e-01 -1.31979227e+00 1.02112606e-01
-1.50963560e-01 -6.15149066e-02 2.47939482e-01 -3.07191789e-01
-1.16965771e+00 8.18787456e-01 8.45011175e-01 6.16008997e-01
1.25491810e+00 -3.21458846e-01 6.19353890e-01 6.68726921e-01
2.52099633e-01 -1.02793062e+00 2.46380821e-01 5.90895891e-01
1.12244499e+00 -7.62957394e-01 -2.80860662e-01 -1.50591999e-01
-2.74192303e-01 1.61669922e+00 5.80413878e-01 -5.62915206e-01
7.12915957e-01 4.03204203e-01 -3.43205124e-01 -2.27132425e-01
-5.27752578e-01 -1.35067031e-01 2.04178274e-01 5.27115583e-01
1.13137163e-01 -1.31165804e-02 -4.76108819e-01 7.14691162e-01
2.72262376e-02 -1.16239525e-01 4.47593600e-01 9.71196711e-01
-8.24038208e-01 -7.20295072e-01 -8.29719067e-01 4.01338130e-01
-4.83981252e-01 3.31118345e-01 2.28023753e-01 8.51053238e-01
1.32481292e-01 1.01482618e+00 1.43724486e-01 -5.04244030e-01
4.35944796e-01 3.20446700e-01 1.65859461e-01 -3.29110384e-01
-1.41746178e-01 -3.99360955e-02 -5.08101523e-01 -2.05967784e-01
-1.56149864e-01 -3.40327948e-01 -1.41090667e+00 -1.50173128e-01
-7.44951725e-01 6.24390662e-01 7.38093257e-01 1.28107345e+00
-2.47466397e-02 9.29016769e-01 7.84657955e-01 -5.66677690e-01
-7.61779308e-01 -1.04519761e+00 -1.01897931e+00 3.33478361e-01
3.17881256e-01 -9.48359728e-01 -8.61864746e-01 1.62046954e-01]
|
[6.709981918334961, 2.4166641235351562]
|
0970fd9b-9394-4ca1-a8d7-a8f4940f70ca
|
hitea-hierarchical-temporal-aware-video
|
2212.14546
| null |
https://arxiv.org/abs/2212.14546v1
|
https://arxiv.org/pdf/2212.14546v1.pdf
|
HiTeA: Hierarchical Temporal-Aware Video-Language Pre-training
|
Video-language pre-training has advanced the performance of various downstream video-language tasks. However, most previous methods directly inherit or adapt typical image-language pre-training paradigms to video-language pre-training, thus not fully exploiting the unique characteristic of video, i.e., temporal. In this paper, we propose a Hierarchical Temporal-Aware video-language pre-training framework, HiTeA, with two novel pre-training tasks for modeling cross-modal alignment between moments and texts as well as the temporal relations of video-text pairs. Specifically, we propose a cross-modal moment exploration task to explore moments in videos, which results in detailed video moment representation. Besides, the inherent temporal relations are captured by aligning video-text pairs as a whole in different time resolutions with multi-modal temporal relation exploration task. Furthermore, we introduce the shuffling test to evaluate the temporal reliance of datasets and video-language pre-training models. We achieve state-of-the-art results on 15 well-established video-language understanding and generation tasks, especially on temporal-oriented datasets (e.g., SSv2-Template and SSv2-Label) with 8.6% and 11.1% improvement respectively. HiTeA also demonstrates strong generalization ability when directly transferred to downstream tasks in a zero-shot manner. Models and demo will be available on ModelScope.
|
['Fei Huang', 'Ji Zhang', 'Qi Qian', 'Haiyang Xu', 'Ming Yan', 'Guohai Xu', 'Qinghao Ye']
|
2022-12-30
| null | null | null | null |
['video-question-answering']
|
['computer-vision']
|
[ 1.48971871e-01 -5.08110344e-01 -2.20544651e-01 -3.45296919e-01
-8.65452051e-01 -5.33885181e-01 9.13610220e-01 -1.47238851e-01
-4.36053962e-01 3.97097439e-01 3.93674344e-01 -2.66469359e-01
-5.98962605e-02 -4.67298537e-01 -8.76728892e-01 -4.74385887e-01
-2.73798078e-01 1.77654698e-01 2.91773319e-01 -3.07521850e-01
1.12350598e-01 7.63675794e-02 -1.61441517e+00 8.79672825e-01
4.93392140e-01 1.00647604e+00 3.95873368e-01 8.32241774e-01
-2.82313257e-01 8.10421407e-01 -2.03436419e-01 -1.76732287e-01
1.06700294e-01 -4.21123058e-01 -7.64145613e-01 2.87404716e-01
4.87086356e-01 -3.32945615e-01 -7.17813849e-01 6.61294401e-01
3.27880651e-01 4.22570288e-01 4.53863740e-01 -1.36101651e+00
-6.84939802e-01 7.63395131e-01 -6.48910642e-01 5.37345052e-01
6.21286750e-01 3.91529173e-01 8.05135071e-01 -1.02503026e+00
9.75199997e-01 1.31999385e+00 5.14027536e-01 4.78954762e-01
-9.39367890e-01 -7.38721609e-01 4.67564374e-01 7.59874165e-01
-1.45992553e+00 -5.51199853e-01 8.35213900e-01 -6.65104806e-01
1.16468203e+00 1.38156125e-02 6.17333472e-01 1.56368339e+00
1.17622949e-01 1.02825558e+00 1.07549083e+00 -2.05761045e-01
-2.33459815e-01 -1.49376422e-01 -2.91634262e-01 5.59702873e-01
-5.47891319e-01 1.48222268e-01 -1.02542496e+00 5.10039985e-01
6.20932281e-01 -1.44826502e-01 -2.01079845e-01 -1.30951032e-01
-1.69186246e+00 4.61953253e-01 6.35740459e-02 4.79043245e-01
-2.35313073e-01 1.03378594e-01 9.10608470e-01 3.82613987e-01
3.74221474e-01 -1.08011467e-02 -3.12569559e-01 -3.73551071e-01
-1.23016441e+00 -1.76996738e-02 2.74305105e-01 1.26566732e+00
5.29874802e-01 1.92464843e-01 -6.14665031e-01 5.88039279e-01
2.98158437e-01 3.51575404e-01 7.36203134e-01 -7.15941608e-01
9.04245257e-01 3.54209125e-01 -3.56759667e-01 -9.86960232e-01
-2.34226674e-01 -2.50899613e-01 -8.49838793e-01 -3.73425305e-01
1.40502989e-01 -7.05391052e-04 -1.02873385e+00 1.96921682e+00
2.20500231e-01 8.00084174e-01 2.80396104e-01 8.17478776e-01
8.76836002e-01 1.08399141e+00 2.11608335e-01 -5.18997848e-01
1.39083385e+00 -1.25434887e+00 -7.64032304e-01 -1.79953314e-02
6.43679440e-01 -8.93917143e-01 1.25371766e+00 3.87348950e-01
-1.12481546e+00 -1.00868583e+00 -8.02637577e-01 -1.97664902e-01
-4.32883918e-01 1.32406041e-01 3.65380973e-01 -1.76127367e-02
-9.81106699e-01 5.31906903e-01 -7.81737685e-01 -5.22769988e-01
1.61402181e-01 1.03458300e-01 -4.56401289e-01 -1.29557133e-01
-1.49051404e+00 4.23560441e-01 7.44996071e-01 2.62971390e-02
-1.23836708e+00 -8.08243275e-01 -9.16281819e-01 -2.19610572e-01
6.54933035e-01 -6.01368070e-01 1.15646362e+00 -9.93019700e-01
-1.41280890e+00 7.86628246e-01 -2.50152975e-01 -5.98136842e-01
7.34027147e-01 -3.59945744e-01 -7.12924063e-01 4.45864111e-01
2.03799754e-01 1.07341862e+00 9.82452691e-01 -1.05663407e+00
-7.61151671e-01 6.91347383e-03 2.15746999e-01 4.14114684e-01
-6.13194525e-01 1.32114091e-03 -1.03617656e+00 -1.10191727e+00
-1.75866872e-01 -8.66472542e-01 1.21152669e-01 -1.69522449e-01
-3.60600948e-01 -1.88427612e-01 1.25490105e+00 -6.80433869e-01
1.62982988e+00 -2.13437200e+00 2.90475935e-01 -2.86196679e-01
-3.09560411e-02 1.71401277e-01 -4.99343187e-01 5.94093263e-01
-2.44979486e-01 1.08102448e-01 5.01412936e-02 -5.07639468e-01
-6.58209249e-02 1.05435044e-01 -3.17582190e-01 1.95260003e-01
2.05516875e-01 8.89312983e-01 -1.02108419e+00 -8.33268225e-01
4.66614574e-01 4.32620585e-01 -4.07447636e-01 2.04861924e-01
-3.78332049e-01 5.93225598e-01 -2.81017214e-01 5.36441863e-01
3.99290770e-01 -3.52131665e-01 -1.63147356e-02 -6.93748832e-01
-1.88434213e-01 1.11488827e-01 -9.65982020e-01 2.25328517e+00
-6.91080570e-01 8.03179264e-01 -3.49094272e-01 -9.60599244e-01
5.69005370e-01 5.60074210e-01 7.79724538e-01 -9.28851664e-01
-7.76882395e-02 -1.13984495e-01 -1.73386022e-01 -8.41892481e-01
6.75949037e-01 -2.12060586e-02 -8.79947692e-02 3.20745260e-01
2.60742068e-01 2.81950802e-01 7.62989581e-01 4.41497266e-01
6.83360636e-01 7.76298523e-01 1.07847437e-01 -8.09228644e-02
8.31651866e-01 -9.25798714e-02 1.33319393e-01 4.65894639e-01
-2.53093373e-02 6.58678472e-01 3.43205869e-01 -1.83901682e-01
-9.76843238e-01 -1.05862200e+00 1.93866983e-01 1.30130374e+00
2.07979515e-01 -8.94434214e-01 -4.05415893e-01 -6.87833667e-01
-3.67600322e-01 6.92379057e-01 -5.92748404e-01 -1.26109064e-01
-6.19084835e-01 -2.95737147e-01 5.63699424e-01 6.32648826e-01
5.35176277e-01 -9.63574648e-01 -4.20327425e-01 2.29231641e-01
-5.83083510e-01 -1.84545946e+00 -7.61973143e-01 -2.24269375e-01
-8.22875202e-01 -8.65771711e-01 -6.98336840e-01 -8.01598072e-01
3.22182208e-01 4.14208293e-01 1.09844494e+00 -1.48433536e-01
2.18531471e-02 6.03978813e-01 -6.51889741e-01 1.99833274e-01
-2.98426092e-01 1.02568641e-02 5.64096533e-02 2.25874573e-01
1.91354886e-01 -6.53227031e-01 -5.35155535e-01 4.74178642e-01
-1.00874746e+00 7.11304903e-01 6.55948758e-01 7.23579228e-01
7.39560306e-01 -5.19517548e-02 3.56195182e-01 -3.62838179e-01
1.88413247e-01 -6.91773534e-01 -2.51996964e-01 4.01496887e-01
-2.43162587e-01 1.02221258e-01 7.67811716e-01 -8.13821614e-01
-1.13747048e+00 4.31143381e-02 5.78168295e-02 -1.08709109e+00
-9.55932513e-02 7.34554887e-01 -4.26318757e-02 3.61665666e-01
3.13865513e-01 6.02096796e-01 -3.52271944e-01 -2.40076274e-01
4.62947369e-01 3.01159412e-01 7.43886232e-01 -7.77651489e-01
9.11162019e-01 5.37611723e-01 -3.05185653e-02 -8.88858914e-01
-7.61036575e-01 -7.39582539e-01 -7.08379745e-01 -4.45270807e-01
1.21067619e+00 -1.19737363e+00 -4.55609083e-01 4.31372285e-01
-1.13333237e+00 -4.71224755e-01 4.59073372e-02 5.68428457e-01
-7.36105084e-01 5.81550300e-01 -5.61858237e-01 -3.89336377e-01
-1.40722901e-01 -1.18646681e+00 1.26314688e+00 -7.24720955e-02
5.56318276e-02 -9.86909807e-01 -1.77115217e-01 4.58077520e-01
5.68737835e-02 1.93931848e-01 6.29259408e-01 -5.27910829e-01
-6.69642627e-01 3.02688122e-01 -3.24154168e-01 1.43385857e-01
3.91121432e-02 7.51324221e-02 -7.42518961e-01 -2.60860234e-01
-2.56975323e-01 -4.73987848e-01 8.16364944e-01 2.44998604e-01
1.24827492e+00 -2.68149406e-01 -2.62439549e-01 7.06008852e-01
1.22852504e+00 2.39734873e-01 5.57118475e-01 3.35363239e-01
9.40863371e-01 6.53445780e-01 9.71237421e-01 5.49064994e-01
4.23814476e-01 9.64790404e-01 1.00331061e-01 1.90684617e-01
-2.46961519e-01 -4.87417877e-01 7.23221540e-01 1.16755366e+00
-7.60524049e-02 -6.22076392e-01 -8.75152469e-01 6.06678665e-01
-1.95760667e+00 -1.16538703e+00 -9.58005115e-02 1.77985609e+00
7.60549486e-01 2.11095393e-01 6.79577813e-02 -1.28425136e-01
6.41138077e-01 5.01747847e-01 -4.18346465e-01 9.08570960e-02
-1.19569413e-01 -1.41440287e-01 2.22431093e-01 3.17178696e-01
-1.28852880e+00 1.15447807e+00 5.15498543e+00 1.18566287e+00
-1.23798084e+00 2.22270772e-01 6.78246737e-01 -3.68677139e-01
-2.38137528e-01 -7.93097615e-02 -8.26820910e-01 5.07010162e-01
1.03709435e+00 -3.61743003e-01 2.47773930e-01 5.23840606e-01
6.71508908e-01 8.71440843e-02 -1.38229382e+00 1.24264216e+00
2.19399720e-01 -1.47014356e+00 4.76399899e-01 -2.49308944e-01
7.97957897e-01 -1.07638337e-01 3.25953923e-02 6.06537759e-01
-3.71609628e-01 -9.20496881e-01 1.03171456e+00 4.25761700e-01
1.15726948e+00 -5.28304040e-01 4.54356849e-01 1.77913353e-01
-1.87908876e+00 -2.42433306e-02 2.63218462e-01 1.50879622e-01
7.46993303e-01 2.22797245e-01 -5.54286063e-01 9.83240485e-01
7.59618521e-01 1.21661532e+00 -6.20839417e-01 5.50295174e-01
4.77376133e-02 4.46960419e-01 -3.23691398e-01 3.71610940e-01
7.14391947e-01 -8.58661979e-02 6.06055021e-01 1.60630012e+00
6.09712422e-01 3.31460796e-02 3.84966612e-01 5.14395118e-01
3.96959707e-02 1.26527786e-01 -6.00481749e-01 -3.86983395e-01
2.63700694e-01 1.14021635e+00 -7.85259366e-01 -6.53531194e-01
-6.34692192e-01 1.02743888e+00 -1.00792736e-01 4.27414745e-01
-1.29757357e+00 9.23791975e-02 4.76638407e-01 1.25480086e-01
4.00379628e-01 -5.60881138e-01 3.20338190e-01 -1.33326864e+00
3.84485759e-02 -9.40583050e-01 4.70937937e-01 -9.70355570e-01
-1.04463196e+00 7.15831578e-01 5.21677554e-01 -1.63685262e+00
-6.17752016e-01 -5.75148523e-01 -4.56914544e-01 3.56303275e-01
-1.38984132e+00 -1.56897902e+00 -5.31202853e-01 9.94828820e-01
1.27658403e+00 -2.90721565e-01 2.41364971e-01 6.01461411e-01
-3.13136429e-01 6.36460721e-01 -1.94955438e-01 7.52481669e-02
8.61052632e-01 -8.50223899e-01 2.64580071e-01 1.13373256e+00
5.02161086e-01 3.34163100e-01 4.67523396e-01 -6.26813233e-01
-1.61645913e+00 -1.23796189e+00 6.67265534e-01 -2.57733405e-01
1.07007945e+00 -4.16713446e-01 -8.13706756e-01 7.28671134e-01
4.37004209e-01 -6.83198795e-02 3.22253108e-01 -2.48189241e-01
-3.21832776e-01 -5.73296733e-02 -4.07732666e-01 9.17610168e-01
1.49316669e+00 -8.87772918e-01 -4.36485857e-01 4.82187808e-01
1.03137469e+00 -6.46626830e-01 -8.37038040e-01 4.90172863e-01
4.85211730e-01 -9.53682482e-01 1.12944055e+00 -6.16845012e-01
8.45480382e-01 -3.59566092e-01 -3.19473743e-01 -8.17216337e-01
-8.07435289e-02 -9.05382216e-01 -3.36487770e-01 1.39383101e+00
2.26709068e-01 1.44433007e-01 5.52694738e-01 -2.15242058e-01
-3.52280051e-01 -7.47051418e-01 -8.20143759e-01 -8.88244927e-01
-3.35301518e-01 -8.88267994e-01 7.37338662e-02 9.39968050e-01
-1.87197506e-01 5.31124949e-01 -6.76137030e-01 6.66811913e-02
2.91726619e-01 3.98281217e-02 6.76335633e-01 -4.11803961e-01
-4.37960744e-01 -4.94513303e-01 -2.23815426e-01 -1.39463258e+00
3.22855860e-01 -8.85808408e-01 2.19418537e-02 -1.29753363e+00
1.62550241e-01 1.08869690e-02 -3.91878605e-01 3.26459557e-01
2.52048373e-02 2.35485256e-01 4.26047474e-01 2.19838038e-01
-9.82275486e-01 6.52108490e-01 1.32093251e+00 -1.85624346e-01
-2.51823276e-01 -3.51222306e-01 -5.94056658e-02 5.98665714e-01
5.15808761e-01 -2.09838256e-01 -9.79249656e-01 -5.85685790e-01
9.70952678e-03 3.04962039e-01 3.96404803e-01 -1.14925194e+00
2.49830782e-01 -3.28189611e-01 1.60523161e-01 -9.55786884e-01
4.89772767e-01 -6.88767731e-01 3.29521090e-01 1.99790686e-01
-4.61728394e-01 4.66521412e-01 5.23931265e-01 7.71302044e-01
-5.10884166e-01 -2.04656217e-02 5.58806956e-01 -1.36896819e-01
-1.44663262e+00 5.58717012e-01 -2.99650371e-01 1.68029040e-01
1.13163340e+00 -3.25108409e-01 -1.55468717e-01 -4.54249650e-01
-8.66247475e-01 3.75043899e-01 2.12591738e-01 8.54176879e-01
7.15399802e-01 -1.47972214e+00 -6.88995898e-01 -6.64871186e-03
3.62105459e-01 -1.59749389e-01 8.07251275e-01 1.19013035e+00
-2.66078740e-01 5.42342901e-01 -2.91376978e-01 -9.86559629e-01
-1.41192365e+00 8.37042153e-01 -1.32796299e-02 -4.39683974e-01
-7.05779612e-01 6.65043056e-01 6.09812319e-01 1.53211728e-01
1.87487349e-01 -4.08326119e-01 -1.66035607e-01 3.01671624e-01
4.59057719e-01 7.59441480e-02 -3.11421007e-01 -8.78301561e-01
-3.76736552e-01 7.80804098e-01 -1.97475776e-02 -3.06905687e-01
9.71938372e-01 -4.07294840e-01 2.91219890e-01 6.25654638e-01
1.48929441e+00 -2.45585814e-01 -1.39410973e+00 -2.95439452e-01
6.98550567e-02 -2.48872399e-01 -1.61582381e-01 -4.87751305e-01
-1.11491120e+00 9.42061782e-01 4.41922605e-01 -2.17932269e-01
1.26917422e+00 -5.87796196e-02 9.60867703e-01 3.09690535e-01
7.90953264e-02 -1.05382717e+00 6.77668869e-01 6.07467055e-01
9.75079536e-01 -1.27525032e+00 -2.54064929e-02 -2.13891119e-01
-8.20722759e-01 1.20868433e+00 8.53292525e-01 2.86095172e-01
5.04748702e-01 1.34613872e-01 -1.52113259e-01 4.46603633e-02
-1.12540352e+00 -1.88304618e-01 6.82163239e-01 4.58488345e-01
4.30059731e-01 -2.98134834e-01 -1.37547866e-01 3.63040388e-01
5.02616800e-02 2.42385656e-01 2.44903088e-01 7.49523938e-01
1.60818342e-02 -9.21870410e-01 -1.38497144e-01 5.99546321e-02
-2.35646829e-01 -2.30113834e-01 1.26579314e-01 1.03552902e+00
2.31903672e-01 7.66473949e-01 1.42621920e-01 -6.82708442e-01
6.20134436e-02 4.97029945e-02 4.36931998e-01 -3.99342507e-01
-2.82152444e-01 4.61676389e-01 2.16459006e-01 -8.22046757e-01
-8.07779610e-01 -6.34853661e-01 -1.26065671e+00 -2.30287224e-01
1.20944001e-01 2.78677349e-03 3.04606766e-01 1.05200315e+00
4.88279045e-01 7.62815535e-01 4.29990381e-01 -1.05585337e+00
-1.66308135e-01 -9.09922600e-01 -1.23474739e-01 6.94448590e-01
2.11613208e-01 -7.08458066e-01 -2.12689877e-01 6.75818384e-01]
|
[10.19348430633545, 0.8932616114616394]
|
cf826164-face-475e-8e01-df0d1892b3e7
|
alphadda-game-artificial-intelligence-with
|
2111.06266
| null |
https://arxiv.org/abs/2111.06266v4
|
https://arxiv.org/pdf/2111.06266v4.pdf
|
AlphaDDA: Strategies for Adjusting the Playing Strength of a Fully Trained AlphaZero System to a Suitable Human Training Partner
|
Artificial intelligence (AI) has achieved superhuman performance in board games such as Go, chess, and Othello (Reversi). In other words, the AI system surpasses the level of a strong human expert player in such games. In this context, it is difficult for a human player to enjoy playing the games with the AI. To keep human players entertained and immersed in a game, the AI is required to dynamically balance its skill with that of the human player. To address this issue, we propose AlphaDDA, an AlphaZero-based AI with dynamic difficulty adjustment (DDA). AlphaDDA consists of a deep neural network (DNN) and a Monte Carlo tree search, as in AlphaZero. AlphaDDA learns and plays a game the same way as AlphaZero, but can change its skills. AlphaDDA estimates the value of the game state from only the board state using the DNN. AlphaDDA changes a parameter dominantly controlling its skills according to the estimated value. Consequently, AlphaDDA adjusts its skills according to a game state. AlphaDDA can adjust its skill using only the state of a game without any prior knowledge regarding an opponent. In this study, AlphaDDA plays Connect4, Othello, and 6x6 Othello with other AI agents. Other AI agents are AlphaZero, Monte Carlo tree search, the minimax algorithm, and a random player. This study shows that AlphaDDA can balance its skill with that of the other AI agents, except for a random player. The DDA ability of AlphaDDA is based on an accurate estimation of the value from the state of a game. We believe that the AlphaDDA approach for DDA can be used for any game AI system if the DNN can accurately estimate the value of the game state and we know a parameter controlling the skills of the AI system.
|
['Kazuhisa Fujita']
|
2021-11-11
| null | null | null | null |
['board-games']
|
['playing-games']
|
[-4.94436681e-01 2.53459722e-01 4.95773405e-02 3.08845013e-01
-7.48819336e-02 -8.01916718e-01 3.15325037e-02 -2.81969339e-01
-9.15364563e-01 7.84309566e-01 -6.27857327e-01 -2.00610861e-01
-3.56484830e-01 -1.16606474e+00 -4.55887705e-01 -9.02209342e-01
3.55901830e-02 1.14815640e+00 6.11627221e-01 -6.59225762e-01
1.91850752e-01 1.87569618e-01 -1.29580975e+00 8.52173753e-03
8.65316629e-01 9.56348300e-01 1.86397567e-01 1.06210661e+00
2.78343320e-01 1.21637225e+00 -9.22186613e-01 -5.25802910e-01
8.04362357e-01 -6.01462960e-01 -6.52054012e-01 -4.62557763e-01
-3.08553338e-01 -5.48418999e-01 -4.72260892e-01 1.23744667e+00
4.97657269e-01 3.06171030e-01 5.46414077e-01 -1.35533261e+00
2.60124505e-01 8.26483071e-01 -3.50452483e-01 2.88432181e-01
1.15129218e-01 6.12057388e-01 8.43117177e-01 4.69053015e-02
2.73360521e-01 1.07702172e+00 4.67702240e-01 6.72891498e-01
-7.41101086e-01 -8.76520395e-01 1.71356469e-01 2.46377334e-01
-1.46451199e+00 1.30597919e-01 5.64353764e-01 -2.69541740e-01
7.99263358e-01 -1.00279320e-02 1.36557055e+00 7.29716003e-01
5.61568320e-01 8.27205896e-01 8.32526445e-01 -3.92623186e-01
8.97428095e-01 -3.01297814e-01 -2.55234122e-01 6.87372148e-01
2.41768509e-01 5.57523012e-01 -4.09656256e-01 -5.26336469e-02
1.13019764e+00 -5.17407358e-01 2.94742752e-02 -2.98336118e-01
-8.76787722e-01 8.42542887e-01 4.65496629e-01 2.23617077e-01
-7.19513118e-01 4.47874218e-01 2.78959453e-01 6.50953412e-01
-8.94688368e-02 1.28654766e+00 -4.47339386e-01 -7.03798056e-01
-7.33627081e-01 5.77946067e-01 1.07437599e+00 3.75057429e-01
4.82930154e-01 3.02565575e-01 6.16530254e-02 3.26137841e-01
1.36514634e-01 4.29630488e-01 6.52409852e-01 -1.48034561e+00
1.65897235e-01 6.55795634e-01 1.35064751e-01 -1.15066969e+00
-5.54390132e-01 -4.61342186e-01 -6.27552688e-01 1.09860861e+00
7.78388321e-01 -5.74917436e-01 -8.89220357e-01 2.01309657e+00
3.49276066e-01 1.46865308e-01 1.07446879e-01 9.71685112e-01
2.84821719e-01 7.25339592e-01 -2.48103604e-01 -2.30719987e-03
1.08221543e+00 -8.79032552e-01 -4.44548905e-01 -5.40018380e-01
4.10830528e-01 -1.04066402e-01 8.96482348e-01 1.05558395e+00
-1.48633707e+00 -4.39327538e-01 -1.19769192e+00 6.78755224e-01
-8.21763352e-02 -3.64600509e-01 7.50575423e-01 5.58315754e-01
-1.05896163e+00 5.92115581e-01 -1.01966763e+00 -1.35822266e-01
2.67214060e-01 9.06906605e-01 -1.96051528e-03 2.84415096e-01
-1.56727743e+00 1.16151595e+00 8.49426210e-01 7.96903223e-02
-1.29242492e+00 -5.18114626e-01 -5.21415353e-01 1.81630388e-01
8.21835876e-01 -8.67396414e-01 1.49971926e+00 -1.37431753e+00
-2.14329433e+00 6.61132097e-01 6.11715078e-01 -7.05646992e-01
8.75793219e-01 1.41908914e-01 1.99134961e-01 1.31199151e-01
-4.11010683e-02 6.19201243e-01 6.84413671e-01 -9.58926320e-01
-7.70382702e-01 -3.46326202e-01 7.29737878e-01 7.72457182e-01
9.13882330e-02 -4.84006763e-01 -4.28265035e-01 -2.68305451e-01
-1.11647872e-02 -1.03143239e+00 -4.35492575e-01 -9.96986777e-02
-1.20314531e-01 -2.93292671e-01 1.29151091e-01 8.64636153e-03
1.34176052e+00 -1.94182396e+00 3.45095158e-01 4.70947593e-01
5.55458903e-01 4.54378724e-01 -1.38658524e-01 1.10076405e-01
2.08506405e-01 -1.95403889e-01 1.17374815e-01 4.36247319e-01
-4.87121530e-02 3.09428811e-01 2.90748179e-01 1.63765207e-01
-4.67599869e-01 8.87632370e-01 -1.12127733e+00 -2.28463709e-01
4.42105383e-02 -1.51672348e-01 -9.49381828e-01 3.41252714e-01
-2.34470204e-01 8.28300267e-02 -7.15946317e-01 1.68900520e-01
2.84240723e-01 -6.93330392e-02 3.11677277e-01 1.86771914e-01
-9.83703211e-02 -9.23163071e-02 -1.37264919e+00 1.21400619e+00
-2.53958195e-01 3.33883613e-01 1.81776360e-02 -9.91855383e-01
8.10116947e-01 3.73214573e-01 4.28895444e-01 -7.50651360e-01
8.07955563e-01 1.57394350e-01 8.62385273e-01 -2.45833904e-01
8.82239938e-02 -2.86012292e-01 -1.94326699e-01 5.90713263e-01
-1.41392544e-01 -6.33204520e-01 4.18330193e-01 1.46972775e-01
1.46369100e+00 -2.51116455e-01 5.90787113e-01 -3.14153373e-01
3.13923031e-01 3.29348594e-01 5.91645241e-01 1.31748450e+00
-4.76690352e-01 2.04420358e-01 8.24230075e-01 -7.13321865e-01
-9.14396465e-01 -1.08861768e+00 4.34692532e-01 1.01527286e+00
4.01620984e-01 -9.23804045e-02 -1.05333483e+00 -3.87807757e-01
-2.46290177e-01 7.36796200e-01 -7.56312132e-01 -7.59223223e-01
-5.69251120e-01 -4.69580233e-01 5.51284909e-01 3.03637356e-01
9.72811580e-01 -1.37934446e+00 -1.01828015e+00 5.05044103e-01
-1.58304825e-01 -5.86726427e-01 -3.62337232e-01 3.17098737e-01
-5.47735155e-01 -1.01122487e+00 -4.36592460e-01 -5.92137635e-01
3.75204414e-01 -1.80093944e-01 1.09705853e+00 1.27206311e-01
2.06963062e-01 7.97847882e-02 -1.37736842e-01 -4.66848761e-01
-5.48222780e-01 4.16209370e-01 2.70923495e-01 -5.47022164e-01
2.62730211e-01 -6.92720950e-01 -5.55106759e-01 4.64272916e-01
-7.25133955e-01 2.72611290e-01 4.26927775e-01 7.32470572e-01
1.99078053e-01 8.12485933e-01 3.76165450e-01 -5.72278976e-01
8.87300909e-01 -2.77861446e-01 -9.10546541e-01 -1.23258680e-01
-5.03088117e-01 1.07009009e-01 7.68014669e-01 -9.57161844e-01
-7.49023020e-01 -1.13706052e-01 -3.44922468e-02 -5.63394248e-01
2.83265561e-01 5.14115095e-01 -3.68582964e-01 8.23629946e-02
8.18556011e-01 -7.31084496e-02 6.38513342e-02 1.04353890e-01
4.18370813e-02 3.89883906e-01 6.60213828e-01 -7.73061275e-01
7.48228908e-01 5.43464720e-02 -2.13909358e-01 -7.94968903e-02
-8.27078938e-01 1.53425664e-01 -3.87235761e-01 -6.65382922e-01
7.77216136e-01 -7.14083493e-01 -1.66966212e+00 1.00376105e+00
-8.43179345e-01 -8.37840855e-01 -6.50492728e-01 5.38858712e-01
-7.72266090e-01 -2.81016767e-01 -6.35612011e-01 -6.83899641e-01
-8.21932405e-02 -1.12247586e+00 1.80734605e-01 7.12761819e-01
-4.66808915e-01 -9.16702032e-01 2.30508417e-01 4.25420284e-01
2.32601985e-01 3.29006873e-02 6.39211237e-01 -5.71377456e-01
-4.49986815e-01 -3.59793782e-01 3.49687308e-01 1.65815890e-01
-1.40256166e-01 -1.56391695e-01 -4.78964150e-01 -3.51549357e-01
8.48068222e-02 -4.25837398e-01 3.25133443e-01 6.80792630e-01
7.91179180e-01 -3.85926247e-01 1.54844858e-02 4.13414896e-01
1.12491810e+00 9.38765824e-01 7.48906553e-01 9.19882894e-01
3.66338015e-01 1.86934948e-01 6.48288429e-01 5.76395988e-01
4.01566774e-01 5.40090203e-01 7.30640113e-01 2.79267371e-01
2.72635818e-01 -3.01406324e-01 4.21552539e-01 6.78441465e-01
-4.08425510e-01 -5.08760989e-01 -9.57204938e-01 2.27081299e-01
-1.90756762e+00 -9.79240954e-01 3.52908522e-01 2.20082307e+00
1.11263549e+00 7.01838195e-01 3.07658195e-01 2.60784358e-01
6.12312913e-01 -1.10555597e-01 -1.12841344e+00 -8.60708714e-01
2.46319696e-01 2.09752798e-01 5.14499426e-01 6.77944005e-01
-6.02716684e-01 1.34512126e+00 6.38100815e+00 1.07228804e+00
-8.67442310e-01 -2.79858768e-01 5.63318491e-01 -3.56945246e-01
6.06467528e-03 -8.43885243e-02 -4.84301358e-01 6.57919049e-01
7.39007890e-01 -5.46482861e-01 1.04559791e+00 8.58689368e-01
2.32301101e-01 -5.83816946e-01 -8.84421468e-01 7.08340526e-01
-1.87721267e-01 -8.87562752e-01 -1.77624241e-01 2.23275796e-01
7.52220988e-01 -1.98640913e-01 3.41587812e-02 6.91779673e-01
1.18648481e+00 -1.12396705e+00 8.00141156e-01 4.48817939e-01
2.54071891e-01 -1.14998543e+00 1.01468801e+00 7.70741165e-01
-8.63712549e-01 -3.33965212e-01 -2.11859971e-01 -6.19076908e-01
-1.56403854e-01 2.05279887e-01 -6.18441522e-01 -8.62691626e-02
5.01728714e-01 -5.60427159e-02 7.66706094e-03 9.76953804e-01
-6.41321361e-01 5.11920571e-01 -5.08269787e-01 -1.79057941e-01
4.98061359e-01 -2.78751701e-01 6.08639777e-01 2.49397010e-01
8.94230381e-02 7.52874136e-01 3.60742182e-01 7.39565313e-01
1.47517294e-01 -3.00029218e-01 -1.62034318e-01 -6.85953477e-04
5.80806732e-01 8.60717952e-01 -6.28891468e-01 -2.92022079e-01
2.96301246e-01 7.95174599e-01 2.70549506e-01 1.75335303e-01
-8.84079695e-01 -3.93567920e-01 8.33549619e-01 7.99014494e-02
-3.66288386e-02 -6.43003583e-02 -2.81085014e-01 -9.71362531e-01
-4.97125685e-01 -1.40970135e+00 4.51647401e-01 -1.02963555e+00
-9.44053471e-01 7.64537334e-01 -2.38812193e-02 -1.14546025e+00
-5.33557117e-01 -5.52016437e-01 -7.73625791e-01 6.13354504e-01
-8.47363532e-01 -3.01827818e-01 -1.92470655e-01 6.84292853e-01
4.44990188e-01 -4.53448027e-01 3.85719150e-01 -2.26420313e-01
-7.89612651e-01 5.59792757e-01 -9.37236920e-02 2.39282936e-01
1.32502049e-01 -1.13674819e+00 1.70132756e-01 5.92516303e-01
-2.78329790e-01 7.64023736e-02 8.50246727e-01 -5.65654278e-01
-1.05446780e+00 -2.58054107e-01 1.91376105e-01 -2.00203493e-01
8.43947053e-01 -7.63109177e-02 -6.87665522e-01 4.67742532e-01
-3.99485752e-02 -2.25195542e-01 -2.66778599e-02 -6.90717474e-02
3.87946248e-01 -2.60635257e-01 -1.14290917e+00 8.47656906e-01
8.78825247e-01 -1.16713636e-01 -6.09704971e-01 -1.94511771e-01
2.84435481e-01 -8.50462914e-01 -4.57530946e-01 1.70525804e-01
7.09272325e-01 -9.26088929e-01 8.08847606e-01 -5.57662487e-01
3.04376900e-01 -2.90049732e-01 3.43054444e-01 -1.78876543e+00
-5.26899397e-01 -5.69581509e-01 -1.33987918e-01 4.67296958e-01
1.86731935e-01 -7.52363443e-01 1.18477881e+00 9.63386834e-01
3.54530454e-01 -6.62764013e-01 -1.09426093e+00 -8.56896043e-01
4.76635516e-01 -4.09175426e-01 4.72232014e-01 6.83622837e-01
1.67961955e-01 1.80868134e-01 -3.26490939e-01 2.48363435e-01
3.76492083e-01 -1.70785725e-01 7.55739748e-01 -1.26237869e+00
-6.94523454e-01 -5.04415989e-01 -6.47814393e-01 -9.86315668e-01
1.04120195e-01 -4.48964566e-01 2.01509401e-01 -1.20449996e+00
1.38071239e-01 -4.71538275e-01 -2.98115134e-01 5.84240377e-01
-1.48610920e-01 -1.18589057e-02 4.88991290e-01 3.51130739e-02
-5.58274329e-01 2.77634531e-01 1.65148675e+00 -2.65127182e-01
-5.32810092e-01 2.28711993e-01 -7.90683210e-01 1.05774140e+00
9.98165429e-01 -7.36467779e-01 -6.41181648e-01 -3.05660456e-01
8.55664849e-01 3.26225281e-01 2.00162176e-02 -1.29747570e+00
6.39976978e-01 -4.43401009e-01 2.35307649e-01 5.83256856e-02
2.53556699e-01 -7.19586790e-01 1.54750988e-01 1.01961148e+00
-2.07462445e-01 -5.23058511e-02 2.56613642e-01 5.28534427e-02
-6.02667145e-02 -5.33630550e-01 1.03905904e+00 -4.83009994e-01
-5.62489748e-01 1.52315378e-01 -1.19518185e+00 3.72958422e-01
1.19064879e+00 -5.74101448e-01 -3.52517888e-02 -8.42676103e-01
-8.91307831e-01 6.19665086e-01 2.96738088e-01 -6.62397370e-02
2.86960185e-01 -1.00931299e+00 -4.38123822e-01 1.81126758e-01
-4.16495532e-01 1.86368853e-01 2.91223854e-01 4.03207481e-01
-8.03407967e-01 -1.27629355e-01 -5.37688553e-01 -1.74560338e-01
-1.02101433e+00 3.75546575e-01 1.13106155e+00 -6.31503284e-01
-2.34370619e-01 8.44120383e-01 5.18853605e-01 -4.93899941e-01
1.82837710e-01 -1.57744005e-01 -2.68056303e-01 -9.50550288e-03
6.15177512e-01 3.54358524e-01 -2.40612239e-01 -3.19307968e-02
-2.22603008e-01 3.72893244e-01 -2.65337769e-02 -3.26131642e-01
1.13078547e+00 1.78073123e-02 1.41585231e-01 3.15621376e-01
3.58512610e-01 -3.60437959e-01 -1.48472655e+00 1.02113731e-01
-6.08481824e-01 -3.83902103e-01 4.26539570e-01 -1.19919991e+00
-1.40505624e+00 5.15624523e-01 4.30343211e-01 1.51906043e-01
1.24259114e+00 -3.03204685e-01 5.97010136e-01 6.80584431e-01
8.12157273e-01 -1.30726290e+00 3.80962789e-01 8.18165302e-01
4.90463138e-01 -9.16533768e-01 -1.19888775e-01 2.79829353e-01
-9.44408417e-01 1.11628008e+00 1.25361156e+00 -3.89419466e-01
5.10970592e-01 5.92903852e-01 3.64999384e-01 -1.81779504e-01
-1.01386023e+00 -2.38735646e-01 -1.57604381e-01 5.56971550e-01
-3.73735279e-01 1.36078715e-01 -1.15828976e-01 7.77616203e-01
-6.48385763e-01 2.07310051e-01 6.79095447e-01 7.33802378e-01
-6.45815194e-01 -8.04128349e-01 -4.13475573e-01 1.62199050e-01
-1.26877189e-01 7.37062693e-02 -2.74133176e-01 9.28587317e-01
3.33398908e-01 8.86381805e-01 3.16026181e-01 -3.50226223e-01
3.16011339e-01 -4.52622801e-01 4.51882869e-01 -3.33091259e-01
-8.44244838e-01 -3.19651216e-01 -1.83443964e-01 -6.80024266e-01
6.65460601e-02 -3.42133045e-01 -1.46001065e+00 -9.71582949e-01
-3.63356292e-01 4.57500279e-01 1.62852854e-01 1.00952339e+00
-1.31496415e-01 5.68200231e-01 6.04936719e-01 -6.16462648e-01
-3.93592238e-01 -5.20939767e-01 -8.00789535e-01 -1.14884898e-01
-1.87369436e-01 -7.73574769e-01 -4.87962037e-01 -6.76898181e-01]
|
[3.5193121433258057, 1.5423507690429688]
|
f11220e4-b97c-41ab-b132-ddbbcf2c10e9
|
privacy-preserving-data-filtering-in
|
2205.11518
| null |
https://arxiv.org/abs/2205.11518v2
|
https://arxiv.org/pdf/2205.11518v2.pdf
|
A Practical Influence Approximation for Privacy-Preserving Data Filtering in Federated Learning
|
Federated Learning by nature is susceptible to low-quality, corrupted, or even malicious data that can severely degrade the quality of the learned model. Traditional techniques for data valuation cannot be applied as the data is never revealed. We present a novel technique for filtering, and scoring data based on a practical influence approximation (`lazy' influence) that can be implemented in a privacy-preserving manner. Each participant uses his own data to evaluate the influence of another participant's batch, and reports to the center an obfuscated score using differential privacy. Our technique allows for highly effective filtering of corrupted data in a variety of applications. Importantly, we show that most of the corrupted data can be filtered out (recall of $>90\%$, and even up to $100\%$), even under really strong privacy guarantees ($\varepsilon \leq 1$).
|
['Boi Faltings', 'Panayiotis Danassis', 'Ljubomir Rokvic']
|
2022-05-23
| null | null | null | null |
['influence-approximation']
|
['methodology']
|
[-3.76367345e-02 1.12472594e-01 -1.39974281e-01 -4.00860697e-01
-9.90958989e-01 -1.04835665e+00 2.43059576e-01 4.55451578e-01
-7.13139474e-01 1.07283115e+00 -3.85423563e-02 -2.27427647e-01
-1.32778749e-01 -8.49349916e-01 -1.05458021e+00 -8.80369127e-01
-5.11340320e-01 2.11136654e-01 8.53447467e-02 9.80062708e-02
-3.16435099e-03 2.71257222e-01 -1.39289391e+00 3.21941704e-01
4.05775726e-01 1.17367804e+00 -7.47558415e-01 5.87150514e-01
1.35996491e-01 9.59171295e-01 -9.21259522e-01 -1.12515640e+00
1.05539477e+00 -1.93263546e-01 -4.47753996e-01 -6.13812625e-01
2.73422420e-01 -7.48490512e-01 -1.31997019e-01 1.27127004e+00
1.76065192e-01 -6.54416978e-02 8.98471195e-03 -1.51411211e+00
-1.33057848e-01 1.06585371e+00 -2.96031266e-01 -1.74204841e-01
1.63673401e-01 3.92235130e-01 9.21068430e-01 -3.24354559e-01
7.77768016e-01 7.90309250e-01 5.73342502e-01 5.20643532e-01
-1.52716029e+00 -1.09818125e+00 -1.24663651e-01 -2.62838870e-01
-1.03191292e+00 -5.97759068e-01 4.43270564e-01 -9.25837159e-02
2.50991583e-01 8.11328888e-01 4.52781171e-01 7.76502371e-01
5.05170040e-02 6.78553581e-01 1.41602671e+00 -3.41568962e-02
5.60245395e-01 4.66787905e-01 7.96181038e-02 5.11897683e-01
8.19758534e-01 5.25373399e-01 -9.57908452e-01 -1.09335279e+00
8.14043209e-02 2.72751093e-01 -4.02302504e-01 -6.18839085e-01
-9.81225073e-01 7.74848282e-01 1.05972342e-01 -2.17068762e-01
-1.48595482e-01 4.92264181e-01 5.04317582e-01 1.02718151e+00
4.79259908e-01 2.44632840e-01 -8.61985266e-01 -1.87698796e-01
-8.84009242e-01 4.52782512e-01 1.08341396e+00 8.51111233e-01
8.20925653e-01 -2.03215122e-01 -1.69380710e-01 5.39297797e-02
-6.26784787e-02 4.47081268e-01 -5.23656048e-02 -1.38054287e+00
3.69974643e-01 3.34135503e-01 4.86337543e-01 -6.74953103e-01
2.75772393e-01 -2.19664291e-01 -8.03618550e-01 7.57736027e-01
6.61710680e-01 -5.00388324e-01 -2.42341906e-01 1.98072851e+00
5.00873327e-01 -3.41138333e-01 -7.90979937e-02 7.92817831e-01
-7.10703209e-02 6.34137392e-02 2.80450173e-02 -5.21874785e-01
1.02658951e+00 -3.65449250e-01 -5.99500954e-01 3.10669899e-01
4.96779889e-01 -2.48419285e-01 7.04058707e-01 6.49095476e-01
-1.23454082e+00 3.00648391e-01 -9.33460593e-01 2.04678446e-01
-1.89247176e-01 -7.96507061e-01 9.56713319e-01 1.04745233e+00
-1.05875480e+00 8.19053054e-01 -8.69438827e-01 4.67367589e-01
1.05681527e+00 5.41575313e-01 -7.60800719e-01 -1.93249330e-01
-1.15943015e+00 3.12311739e-01 -6.82953224e-02 -7.58888423e-01
-1.10268271e+00 -1.24753499e+00 -3.83157521e-01 1.62396938e-01
3.57498884e-01 -6.98099196e-01 1.07645023e+00 -7.65954375e-01
-8.50944042e-01 8.24950814e-01 1.08829357e-01 -8.61010849e-01
1.22522902e+00 3.43423523e-02 -1.28213361e-01 8.34288299e-02
-2.16862455e-01 -9.98325348e-02 8.65432739e-01 -1.18134272e+00
-8.56542349e-01 -9.41840947e-01 1.48033813e-01 -1.30070210e-01
-5.09921670e-01 7.63734281e-02 1.02328904e-01 -5.86667180e-01
-3.31330270e-01 -5.23432374e-01 -4.50100809e-01 4.98174697e-01
-7.52451420e-02 3.07897478e-01 7.19719291e-01 -4.30969566e-01
1.07025433e+00 -2.27381325e+00 -5.06789625e-01 4.79722649e-01
4.73488778e-01 1.69790357e-01 1.13575190e-01 2.77664632e-01
5.06032884e-01 6.31861329e-01 -3.98551613e-01 -4.25659925e-01
1.72963619e-01 -2.67779715e-02 -3.72622281e-01 8.53475749e-01
-5.15150189e-01 6.32446826e-01 -8.13971460e-01 -6.23790510e-02
-2.97669500e-01 3.36591095e-01 -6.44247949e-01 2.14612544e-01
-2.12697521e-01 6.04210421e-02 -2.95344055e-01 7.02495098e-01
9.92475569e-01 1.60842519e-02 3.58451337e-01 2.90329009e-01
1.28125250e-01 8.86086002e-02 -1.33977163e+00 1.33640087e+00
-1.34109542e-01 1.79399863e-01 8.87654245e-01 -3.81553262e-01
6.14839196e-01 3.27680975e-01 6.31601691e-01 -3.88952672e-01
6.29074648e-02 1.50337785e-01 -5.40377676e-01 9.71592292e-02
2.48246580e-01 -9.81886983e-02 -2.98223287e-01 1.32926524e+00
-1.94960982e-01 4.15323973e-01 -2.73077101e-01 3.46198350e-01
1.71142495e+00 -5.16199589e-01 8.79045054e-02 -1.70751244e-01
7.70902932e-02 -1.26459241e-01 8.25865805e-01 1.05090654e+00
-5.33888042e-01 7.04015717e-02 6.07296288e-01 -5.87988913e-01
-1.02529407e+00 -9.86825347e-01 4.10587192e-02 1.35559344e+00
1.07073329e-01 -3.41217250e-01 -8.67049277e-01 -1.17217052e+00
8.83885741e-01 5.18379629e-01 -5.79983413e-01 -1.66684270e-01
-1.05487719e-01 -7.53203511e-01 7.42719710e-01 7.94137493e-02
3.58342379e-01 -7.41749465e-01 -6.76380992e-01 5.19496873e-02
1.48648068e-01 -4.26244497e-01 -5.98244727e-01 2.09054902e-01
-9.01034832e-01 -1.22989690e+00 3.79005559e-02 3.02980930e-01
7.34988034e-01 1.94826826e-01 9.34789658e-01 2.04821050e-01
-1.09442152e-01 1.96932420e-01 1.41639709e-02 -5.58859169e-01
-4.39099699e-01 -3.10607761e-01 9.06883925e-02 1.80730075e-01
4.96666491e-01 -6.41434610e-01 -8.33777785e-01 7.21393600e-02
-1.05807865e+00 -8.29135716e-01 1.45337090e-01 6.40428782e-01
5.28188884e-01 3.43277082e-02 4.87908483e-01 -1.63536024e+00
6.89299285e-01 -4.27374631e-01 -8.41584384e-01 1.94463670e-01
-1.19252515e+00 1.22694887e-01 8.59727919e-01 -2.77370363e-01
-8.10020268e-01 1.03903273e-02 3.92931610e-01 -5.99194348e-01
4.20345999e-02 -7.77243823e-02 -2.42770314e-01 -4.43008304e-01
6.49632514e-01 6.35119528e-02 4.27130610e-01 -6.56191766e-01
6.69607997e-01 5.03645837e-01 4.06373858e-01 -5.13219237e-01
7.71859050e-01 9.28319156e-01 -7.96192326e-03 1.81565389e-01
-3.57868522e-01 7.84161910e-02 -1.77459627e-01 1.68544248e-01
-1.47467583e-01 -8.66156876e-01 -1.46723568e+00 3.24026644e-01
-5.38546324e-01 -2.09764421e-01 -7.89499640e-01 1.65406436e-01
-2.95643300e-01 3.03322405e-01 -6.28122747e-01 -1.14128673e+00
-7.70834208e-01 -6.26160264e-01 2.55500644e-01 -7.76809379e-02
-8.27986281e-03 -6.02267325e-01 -9.17250291e-02 4.52262372e-01
7.48217404e-01 6.10978305e-01 5.34307897e-01 -9.63592052e-01
-8.41692209e-01 -7.65947104e-01 6.52278289e-02 3.26655298e-01
-2.17806339e-01 -1.11984178e-01 -1.06633961e+00 -7.62298822e-01
2.44141206e-01 -3.39410216e-01 7.92650819e-01 -9.94054154e-02
1.32553542e+00 -1.08396351e+00 -3.19816694e-02 7.71955848e-01
1.49632823e+00 -1.44120753e-01 2.57927120e-01 1.29843384e-01
1.84624091e-01 3.51427585e-01 5.37238896e-01 1.19985318e+00
1.22226655e-01 -1.38042662e-02 7.00368106e-01 1.85370609e-01
3.27599257e-01 -3.25877070e-01 4.69506592e-01 1.70404539e-01
1.13048241e-01 1.21834569e-01 -2.27269292e-01 5.83707392e-01
-1.56046724e+00 -1.04916847e+00 -7.76015408e-03 2.92582774e+00
1.21402264e+00 9.07846838e-02 1.25374332e-01 9.10570100e-02
3.94339532e-01 1.41338229e-01 -8.06935191e-01 -5.81461549e-01
-1.80088416e-01 5.14277279e-01 1.34404349e+00 2.04266757e-01
-7.68222153e-01 4.60884482e-01 6.79430151e+00 6.14918172e-01
-8.41584504e-01 5.99714875e-01 8.00577104e-01 -8.33779633e-01
-8.00247550e-01 1.35065615e-01 -3.65352601e-01 7.62823701e-01
1.17427313e+00 -7.90041804e-01 8.66416156e-01 1.13438570e+00
-2.03431830e-01 -5.58080636e-02 -1.08879888e+00 6.80334270e-01
-2.69688487e-01 -1.39147234e+00 -3.20864797e-01 4.66571927e-01
7.08317637e-01 1.00273162e-01 1.33897975e-01 1.58583269e-01
1.11632955e+00 -7.36525655e-01 6.80555284e-01 5.63357294e-01
1.10024393e+00 -1.32335699e+00 5.35676003e-01 5.84961772e-01
-5.14263093e-01 -3.64480644e-01 -4.99211758e-01 5.38592748e-02
-3.28710228e-01 9.81509984e-01 -5.02571285e-01 4.28678334e-01
8.53777826e-01 -5.67632951e-02 -2.91517913e-01 6.65661752e-01
-3.72143202e-02 6.40723288e-01 -6.00135565e-01 2.14303583e-02
-2.95291811e-01 -4.71208291e-03 5.49362600e-01 8.32092464e-01
2.04716265e-01 7.10514113e-02 -6.74201325e-02 6.96265578e-01
-9.12117481e-01 -1.19898738e-02 -6.52882397e-01 1.46662012e-01
9.47570324e-01 1.02244043e+00 1.51026651e-01 -3.25826824e-01
2.54435511e-03 7.64227688e-01 3.56571615e-01 7.55828395e-02
-4.06942874e-01 -2.86632419e-01 1.24936390e+00 2.51071066e-01
2.77267456e-01 1.31848305e-01 -5.07772207e-01 -1.26514447e+00
2.07644582e-01 -1.10397685e+00 1.00153315e+00 1.93513278e-02
-1.47053647e+00 2.52724916e-01 -5.86105049e-01 -9.02125955e-01
-6.70404583e-02 1.30623236e-01 -2.70792425e-01 7.96644628e-01
-1.27075291e+00 -6.80421054e-01 1.54048562e-01 1.02809656e+00
-4.96101797e-01 -1.80297419e-01 9.67694640e-01 1.22551277e-01
-1.18802734e-01 1.35351062e+00 5.64629436e-01 -6.05869070e-02
8.85389924e-01 -1.19043100e+00 1.10769354e-01 8.87331009e-01
-4.83052321e-02 7.50148296e-01 5.47766685e-01 -4.91733342e-01
-1.77881956e+00 -1.26287532e+00 8.60119700e-01 -6.97905481e-01
3.31490606e-01 -5.07412553e-01 -8.61384690e-01 6.06967568e-01
-7.60766566e-02 6.17310047e-01 9.88225102e-01 -2.00710651e-02
-9.76353824e-01 -7.59296060e-01 -2.32816815e+00 2.69099146e-01
9.88748193e-01 -5.89855373e-01 1.27783176e-02 1.76498536e-02
8.25226605e-01 -1.61388651e-01 -1.05930400e+00 1.84564263e-01
7.51186430e-01 -1.25159383e+00 5.38265169e-01 -9.75357473e-01
-2.17876315e-01 -1.17561094e-01 -2.48554185e-01 -8.98012936e-01
-6.58583418e-02 -1.24651277e+00 -7.02408850e-01 1.16727483e+00
1.93227753e-01 -8.69294703e-01 1.04796576e+00 1.41052914e+00
7.91807413e-01 -3.41252655e-01 -1.22842395e+00 -5.28704464e-01
2.44195852e-02 -4.35469240e-01 1.24234676e+00 1.23325813e+00
6.09583147e-02 -8.22034419e-01 -3.90806109e-01 1.63542479e-01
1.53126490e+00 2.97815710e-01 9.09208119e-01 -1.10708511e+00
-2.92564124e-01 -1.76938698e-01 -2.74233162e-01 -2.54277766e-01
-1.83587462e-01 -8.21142018e-01 -4.16634023e-01 -5.83385646e-01
3.43762010e-01 -7.30968058e-01 -7.63828635e-01 7.98396707e-01
4.03174981e-02 2.62822688e-01 2.87307262e-01 3.30700338e-01
-5.34624755e-01 7.51549155e-02 7.32138276e-01 -3.66950929e-02
2.11700231e-01 3.21981221e-01 -1.12811506e+00 1.29038244e-01
7.10715115e-01 -9.04873252e-01 -1.56971261e-01 -1.01515427e-01
4.22649711e-01 1.71917185e-01 4.41200554e-01 -6.83153510e-01
4.46893454e-01 -3.60247910e-01 1.51246384e-01 -2.23209098e-01
-1.25508144e-01 -1.06220102e+00 7.32414722e-01 5.91065705e-01
-7.15502203e-01 -2.66457021e-01 -4.10404176e-01 8.95720184e-01
3.46403718e-02 1.39248028e-01 8.18368018e-01 -4.58523929e-01
2.08438873e-01 7.07288682e-01 9.16917622e-02 8.45482945e-02
1.25357914e+00 1.58614472e-01 -5.88539839e-01 -4.08862561e-01
-5.22537589e-01 3.62552971e-01 1.13338292e+00 -1.56923354e-01
2.75317430e-01 -1.09506249e+00 -6.56541049e-01 4.41270649e-01
-2.11340655e-02 -1.58811480e-01 2.72126496e-01 5.16012013e-01
-5.97017035e-02 -6.86791539e-02 -8.53313059e-02 3.11863534e-02
-1.31447709e+00 8.65371466e-01 2.29403585e-01 -1.68185964e-01
-4.96305585e-01 8.01353097e-01 -4.85394657e-01 -2.63369709e-01
6.78074479e-01 1.15712106e-01 7.80485928e-01 -1.37488707e-03
9.35859203e-01 4.60800916e-01 3.75343293e-01 -5.97179346e-02
-3.07629913e-01 -4.93285894e-01 -4.02352400e-02 -3.07767957e-01
1.44198954e+00 -1.75376803e-01 -2.72867292e-01 1.86242741e-02
1.27072728e+00 3.96389008e-01 -1.44641912e+00 -4.17406589e-01
-2.30248988e-01 -1.14295840e+00 6.34582937e-02 -1.14919925e+00
-1.39865696e+00 4.85645264e-01 5.31968892e-01 3.21937531e-01
1.16076970e+00 -3.98735374e-01 8.42311919e-01 1.06833294e-01
1.12083256e+00 -9.61467087e-01 -7.72787869e-01 -1.98075384e-01
2.89178342e-01 -1.00433958e+00 1.32522851e-01 1.77660480e-01
-5.01631439e-01 5.04520714e-01 1.92757517e-01 1.16761781e-01
7.55590200e-01 6.11699283e-01 5.58211394e-02 8.12295005e-02
-1.43382573e+00 4.12627786e-01 -5.77752352e-01 6.35625482e-01
-1.93164513e-01 3.45993698e-01 -4.47433650e-01 9.73347187e-01
1.31387180e-02 1.15304753e-01 6.74354196e-01 1.20368183e+00
-2.52005845e-01 -1.16214824e+00 -5.36878526e-01 6.98347509e-01
-1.13120043e+00 1.48875579e-01 -4.78880852e-01 2.65765429e-01
9.69075114e-02 7.55053461e-01 -1.32899612e-01 -2.91788399e-01
1.59673050e-01 2.25088581e-01 -3.34174260e-02 -3.76134105e-02
-1.34252882e+00 -3.01953971e-01 -1.13170974e-01 -1.04151058e+00
-1.09142944e-01 -7.13463008e-01 -1.03439629e+00 -1.16000569e+00
-3.09871174e-02 4.08676922e-01 8.66672099e-01 2.86916912e-01
8.07923019e-01 -3.20722431e-01 1.35614049e+00 1.44329667e-01
-1.11282337e+00 -3.97871524e-01 -9.52382445e-01 7.40711212e-01
4.25090253e-01 -1.00647509e-01 -9.26149130e-01 -1.11690499e-01]
|
[5.875688076019287, 6.675232887268066]
|
fb49cee3-0e77-4020-bd2d-1d92991a3dd3
|
towards-expert-level-medical-question
|
2305.09617
| null |
https://arxiv.org/abs/2305.09617v1
|
https://arxiv.org/pdf/2305.09617v1.pdf
|
Towards Expert-Level Medical Question Answering with Large Language Models
|
Recent artificial intelligence (AI) systems have reached milestones in "grand challenges" ranging from Go to protein-folding. The capability to retrieve medical knowledge, reason over it, and answer medical questions comparably to physicians has long been viewed as one such grand challenge. Large language models (LLMs) have catalyzed significant progress in medical question answering; Med-PaLM was the first model to exceed a "passing" score in US Medical Licensing Examination (USMLE) style questions with a score of 67.2% on the MedQA dataset. However, this and other prior work suggested significant room for improvement, especially when models' answers were compared to clinicians' answers. Here we present Med-PaLM 2, which bridges these gaps by leveraging a combination of base LLM improvements (PaLM 2), medical domain finetuning, and prompting strategies including a novel ensemble refinement approach. Med-PaLM 2 scored up to 86.5% on the MedQA dataset, improving upon Med-PaLM by over 19% and setting a new state-of-the-art. We also observed performance approaching or exceeding state-of-the-art across MedMCQA, PubMedQA, and MMLU clinical topics datasets. We performed detailed human evaluations on long-form questions along multiple axes relevant to clinical applications. In pairwise comparative ranking of 1066 consumer medical questions, physicians preferred Med-PaLM 2 answers to those produced by physicians on eight of nine axes pertaining to clinical utility (p < 0.001). We also observed significant improvements compared to Med-PaLM on every evaluation axis (p < 0.001) on newly introduced datasets of 240 long-form "adversarial" questions to probe LLM limitations. While further studies are necessary to validate the efficacy of these models in real-world settings, these results highlight rapid progress towards physician-level performance in medical question answering.
|
['Vivek Natarajan', 'Alan Karthikesalingam', 'Shekoofeh Azizi', 'Yossi Matias', 'Greg S. Corrado', 'Dale Webster', 'Joelle Barral', 'S. Sara Mahdavi', 'Christopher Semturs', 'Renee Wong', 'Yun Liu', 'Nenad Tomasev', 'Blaise Aguera y Arcas', 'Ewa Dominowska', 'Bradley Green', 'Sushant Prakash', 'Philip Mansfield', 'Sami Lachgar', 'Mohamed Amin', 'Amy Wang', 'Mike Schaekermann', 'Darlene Neal', 'Heather Cole-Lewis', 'Stephen Pfohl', 'Kevin Clark', 'Le Hou', 'Ellery Wulczyn', 'Rory Sayres', 'Juraj Gottweis', 'Tao Tu', 'Karan Singhal']
|
2023-05-16
| null | null | null | null |
['multiple-choice-qa', 'protein-folding']
|
['natural-language-processing', 'natural-language-processing']
|
[ 1.20274618e-01 3.22129905e-01 -1.31494954e-01 -4.09255028e-01
-1.72893524e+00 -7.86459088e-01 3.31772596e-01 6.21690452e-01
-6.17279112e-01 6.54787242e-01 4.66680944e-01 -8.33867908e-01
-7.04054534e-01 -4.78467762e-01 -5.26004374e-01 -1.79485474e-02
2.69600004e-01 1.02510297e+00 -2.57479587e-05 -5.15022933e-01
-3.45592462e-02 1.49363667e-01 -8.15544307e-01 7.87898421e-01
1.34512103e+00 6.08134270e-01 -3.06635678e-01 9.32185650e-01
5.13309650e-02 9.76546705e-01 -7.13270068e-01 -9.33654964e-01
-1.04643665e-01 -2.06838995e-01 -1.21360302e+00 -5.15608251e-01
7.51203716e-01 -2.07190245e-01 -1.75388962e-01 5.53111076e-01
9.14451241e-01 -3.03741675e-02 4.23043132e-01 -8.21041226e-01
-8.68505180e-01 3.33608866e-01 2.48285942e-02 3.77432972e-01
8.70072603e-01 6.30734324e-01 1.20423543e+00 -4.07242149e-01
8.33375454e-01 1.14581656e+00 7.82467961e-01 8.85732591e-01
-1.16529155e+00 -6.63120508e-01 -1.05828367e-01 1.34391621e-01
-9.29422081e-01 -1.82581350e-01 1.05163567e-01 -4.34230655e-01
1.18083310e+00 7.21099675e-01 1.27367303e-01 1.05211973e+00
5.67644179e-01 6.00131154e-01 1.15884292e+00 3.37993801e-02
2.82485485e-01 2.96150297e-01 3.88165623e-01 5.58048427e-01
-7.90892355e-03 -1.47947848e-01 -3.58748823e-01 -9.21131074e-01
1.09684609e-01 -2.77966619e-01 -3.09562773e-01 4.86366421e-01
-1.23049319e+00 8.49550426e-01 3.57023656e-01 2.48664811e-01
-5.60795784e-01 -1.73839316e-01 1.57777831e-01 3.25724810e-01
2.81580120e-01 1.22366118e+00 -8.38553250e-01 -2.45643318e-01
-8.57715249e-01 6.60803020e-01 1.18310678e+00 5.07644713e-01
1.21879149e-02 -6.42479062e-01 -5.98965645e-01 6.75697029e-01
7.82144442e-02 6.61617517e-01 4.15133864e-01 -1.19292140e+00
4.76939261e-01 6.13315344e-01 2.37881631e-01 -9.50292349e-01
-7.34984696e-01 -7.14946151e-01 -7.05314934e-01 -3.65872175e-01
5.44488132e-01 -1.94541097e-01 -9.59586680e-01 1.79403591e+00
2.12286204e-01 -1.73835173e-01 2.55418450e-01 7.95289099e-01
1.29024470e+00 5.19539058e-01 7.53263235e-01 1.63289219e-01
1.97509253e+00 -6.11576974e-01 -6.94718182e-01 -6.86742142e-02
7.73590744e-01 -7.67807424e-01 1.17027056e+00 6.71821892e-01
-1.16995037e+00 -3.76065969e-01 -6.09812260e-01 -1.43420368e-01
-2.52312213e-01 -2.75880665e-01 4.49492276e-01 6.57503605e-01
-1.08518970e+00 2.39750430e-01 -6.15876913e-01 -3.50217372e-01
5.07287383e-01 4.82840180e-01 -2.99486935e-01 -2.16963261e-01
-1.71729875e+00 1.13287103e+00 -1.04364552e-01 -4.58896816e-01
-8.58461082e-01 -1.53577876e+00 -4.85012591e-01 -3.66301760e-02
3.65922391e-01 -1.32747018e+00 1.40758765e+00 -2.95130372e-01
-1.04239058e+00 9.54912782e-01 -1.88182667e-01 -6.82638764e-01
5.32445550e-01 -3.84132415e-01 -1.01041710e+00 4.31285083e-01
1.69000909e-01 8.90359581e-01 1.93057001e-01 -8.27248931e-01
-5.29635191e-01 -2.66181856e-01 3.37572306e-01 -1.66759696e-02
-5.97325452e-02 7.63505101e-02 -2.21737564e-01 -5.46894014e-01
-4.16189432e-01 -9.19846594e-01 -6.15427852e-01 -1.75095975e-01
-2.23839238e-01 -5.68845332e-01 1.74082473e-01 -1.02734363e+00
1.48022139e+00 -1.72761834e+00 -2.45428547e-01 1.37001306e-01
6.45041168e-01 5.08960605e-01 -3.49818498e-01 4.22558844e-01
-2.17700139e-01 3.31659764e-01 -3.75374675e-01 6.06447086e-02
-1.61941051e-01 4.72317524e-02 -3.54294211e-01 -2.66887639e-02
5.18176913e-01 1.37612927e+00 -9.71571684e-01 -5.41919351e-01
-9.67244282e-02 3.66602242e-01 -1.04009140e+00 1.33249998e-01
-5.09168148e-01 6.13999844e-01 -6.26462519e-01 6.35577321e-01
3.58130574e-01 -8.15447450e-01 1.49236079e-02 -6.91292509e-02
5.99031806e-01 4.51332062e-01 -6.28689408e-01 1.66635466e+00
-2.91924596e-01 1.58753097e-02 -8.13189894e-02 -4.93276745e-01
6.03834212e-01 6.35379791e-01 6.63998067e-01 -6.96675956e-01
-1.68999791e-01 3.17482799e-01 3.31136405e-01 -9.36415374e-01
2.95522720e-01 -2.92937785e-01 -1.08227320e-02 2.13363364e-01
-3.27558033e-02 -3.10992509e-01 7.15713715e-03 6.89083695e-01
1.52233791e+00 -3.55299115e-01 8.72476771e-02 -4.74904954e-01
5.93316376e-01 4.52468663e-01 2.94789851e-01 9.90412951e-01
-2.94507742e-01 5.54618180e-01 4.26377743e-01 -3.08536559e-01
-4.45507377e-01 -1.10380077e+00 -4.73626792e-01 9.44550276e-01
-3.38355511e-01 -5.04938602e-01 -8.67822528e-01 -1.03193223e+00
1.76640853e-01 1.05932641e+00 -6.52648509e-01 -1.41449749e-01
-3.74591142e-01 -8.28689039e-01 1.11417603e+00 3.30905199e-01
7.50496164e-02 -1.01011884e+00 -6.13741398e-01 5.08882463e-01
-5.94483018e-01 -1.18737054e+00 -4.28263545e-01 -2.74339318e-01
-7.52751231e-01 -1.25832617e+00 -8.94599497e-01 -3.87479693e-01
3.02259147e-01 -5.27400732e-01 1.60532403e+00 -1.52151555e-01
-5.30491352e-01 8.37584436e-01 -1.52109712e-01 -5.92590988e-01
-7.17987061e-01 2.42236152e-01 -1.12222046e-01 -4.27263677e-01
6.75110757e-01 -6.81499466e-02 -1.01376688e+00 3.10394943e-01
-1.26208639e+00 -3.36457342e-01 6.96293592e-01 8.95989358e-01
5.42374253e-01 -5.24293602e-01 1.09827840e+00 -1.29219961e+00
1.12540674e+00 -8.58751953e-01 -7.56630301e-02 5.58938086e-01
-9.12945628e-01 8.92282724e-02 4.48170334e-01 -2.57199258e-01
-8.47076118e-01 -4.35876578e-01 -8.22883487e-01 -2.87308618e-02
-3.26986521e-01 5.95632434e-01 2.31411204e-01 2.65641689e-01
1.12696695e+00 -2.70182669e-01 8.14719200e-02 -5.06334245e-01
4.45186645e-01 7.49934793e-01 6.42228961e-01 -7.55912900e-01
4.59784538e-01 2.69303858e-01 -2.53101081e-01 -3.46463919e-01
-9.28850293e-01 -5.75901985e-01 1.36337414e-01 3.30835938e-01
1.15843797e+00 -7.85334408e-01 -1.18179452e+00 -6.81527257e-02
-9.90356266e-01 4.99179289e-02 -3.06084424e-01 4.16435987e-01
-1.40979275e-01 3.57257903e-01 -7.39474654e-01 -4.60741520e-01
-7.51872122e-01 -1.27398634e+00 1.01570559e+00 6.76771477e-02
-9.09308612e-01 -1.15507782e+00 3.19023222e-01 1.19931495e+00
5.51182032e-01 1.46581382e-01 1.32059717e+00 -1.31872392e+00
-1.34050563e-01 -1.87929407e-01 -1.88561007e-02 1.58394933e-01
8.61173868e-02 -5.51200569e-01 -9.97296810e-01 -2.01002523e-01
1.20111413e-01 -4.05430973e-01 6.30647361e-01 2.78613865e-01
1.28788972e+00 -1.56932354e-01 -3.67312610e-01 1.91936597e-01
1.01452398e+00 3.55437845e-01 5.26609540e-01 1.15432717e-01
1.62265196e-01 6.54292941e-01 4.45782691e-01 1.07816085e-01
6.20892286e-01 3.64964336e-01 1.14266820e-01 -2.37568781e-01
-7.13442490e-02 -2.20866591e-01 7.19771609e-02 7.05941379e-01
3.11191499e-01 -3.83642346e-01 -1.36529887e+00 5.66087902e-01
-1.39458680e+00 -6.09275043e-01 -1.18927144e-01 1.85434616e+00
1.18621743e+00 1.84162706e-01 -5.88331930e-02 -3.46917510e-01
4.81811427e-02 -2.84485698e-01 -9.47350800e-01 -5.50224483e-01
-1.65902168e-01 7.31220186e-01 1.20218270e-01 6.56253815e-01
-8.01310897e-01 5.68757951e-01 6.91664314e+00 9.39046502e-01
-7.79759288e-01 2.30956987e-01 1.14912641e+00 -3.62927886e-03
-8.78863573e-01 -2.95007408e-01 -6.62915289e-01 3.91211718e-01
1.36142945e+00 -5.31632788e-02 6.27272353e-02 5.23720086e-01
-6.15539849e-02 -4.58984869e-03 -1.16336381e+00 7.86172986e-01
9.00776461e-02 -1.56514013e+00 2.88787991e-01 5.32712191e-02
8.89733136e-01 9.65689421e-02 3.80941451e-01 5.11995912e-01
5.96352100e-01 -1.51605213e+00 -1.34672254e-01 7.06307769e-01
1.02330244e+00 -4.15192902e-01 9.20651972e-01 3.82106125e-01
-4.49974418e-01 3.00544016e-02 6.89519942e-02 3.16092879e-01
1.39493108e-01 5.26724935e-01 -1.23790181e+00 9.12477195e-01
6.99493170e-01 2.31753945e-01 -7.04659343e-01 6.52570069e-01
1.51402444e-01 1.02788353e+00 -3.35559338e-01 -2.43099593e-02
4.50258732e-01 4.52979803e-01 5.01361370e-01 1.50016820e+00
-9.48721990e-02 6.81920052e-01 -4.22697254e-02 7.43696630e-01
-2.85215259e-01 2.23922133e-01 -1.21723168e-01 -3.09464753e-01
2.27523163e-01 1.12725103e+00 -1.03758305e-01 -4.68847692e-01
-2.32248560e-01 6.12809658e-01 -3.37518230e-02 3.81590158e-01
-8.45837712e-01 -3.82315665e-01 6.26895547e-01 2.80266076e-01
-4.81566101e-01 4.34530377e-01 -4.01338637e-01 -9.10551846e-01
-2.09113315e-01 -1.77513945e+00 1.04767966e+00 -6.24245822e-01
-1.65676463e+00 7.84241378e-01 -1.31822288e-01 -8.92500877e-01
-4.38493431e-01 -6.89817727e-01 6.00727927e-03 1.06642449e+00
-1.32549405e+00 -9.23610985e-01 -1.81453731e-02 6.38405502e-01
3.04646969e-01 -1.76194593e-01 1.24618661e+00 4.57825035e-01
-9.92308408e-02 1.08453166e+00 -1.29411593e-02 -1.85859445e-02
1.14720368e+00 -1.13618648e+00 4.11645591e-01 1.78035691e-01
2.72786920e-03 1.00497031e+00 4.77605641e-01 -7.06163347e-01
-1.26119983e+00 -7.81385243e-01 1.06311059e+00 -1.51199055e+00
4.44616675e-01 1.76951557e-01 -1.08867002e+00 5.56461275e-01
3.31912816e-01 -4.94864374e-01 1.18452990e+00 1.59320280e-01
-4.01579082e-01 2.71869227e-02 -1.51940691e+00 4.49696481e-01
6.25584722e-01 -6.54491663e-01 -1.01194239e+00 5.02096176e-01
1.04099715e+00 -5.01944184e-01 -1.41922331e+00 8.39403510e-01
5.03876984e-01 -4.48903710e-01 1.10729361e+00 -1.38819897e+00
6.53533041e-01 -8.90109241e-02 1.88109768e-03 -1.11160374e+00
-3.22364628e-01 -8.31593275e-01 2.98629049e-02 6.87026739e-01
7.27364898e-01 -8.20088983e-01 6.42536640e-01 1.10892749e+00
3.16676758e-02 -1.40135038e+00 -8.26315105e-01 -1.30502269e-01
5.60388923e-01 -5.39949954e-01 5.21554530e-01 1.11154580e+00
1.39281340e-02 3.39020967e-01 1.38333082e-01 1.36394158e-01
1.72102824e-01 -2.46037185e-01 4.24081862e-01 -9.59057570e-01
-5.96651495e-01 -4.47321743e-01 -1.19054668e-01 -1.02625859e+00
-1.88509598e-01 -1.10971141e+00 -3.45187455e-01 -1.82004058e+00
3.87084007e-01 -3.42144102e-01 -6.62517428e-01 5.00481009e-01
-8.25232685e-01 1.06014227e-02 6.87037781e-02 -5.83455898e-02
-5.60294688e-01 -3.68979946e-02 1.31004512e+00 -3.35591853e-01
8.31064954e-02 -2.98420973e-02 -1.29650533e+00 3.97897840e-01
6.00040853e-01 -3.94737810e-01 -3.85217309e-01 -4.89537090e-01
2.78610915e-01 2.96528250e-01 2.58035809e-01 -8.58083189e-01
3.81418884e-01 -1.97090637e-02 2.24272847e-01 -4.42681074e-01
2.39300877e-01 -4.98031944e-01 3.58284973e-02 9.39833403e-01
-8.94482732e-01 2.54154533e-01 6.66116536e-01 4.31687891e-01
-1.22071490e-01 1.54148892e-01 5.36274910e-01 -1.74563199e-01
-3.29798162e-01 1.49587363e-01 -2.72887379e-01 7.87638187e-01
6.88824713e-01 2.81997651e-01 -5.12245119e-01 -5.49856842e-01
-9.71321285e-01 8.12144160e-01 -2.37328053e-01 5.51214814e-01
5.42389810e-01 -6.66199028e-01 -1.36718249e+00 -1.82143107e-01
2.42881194e-01 -1.52761385e-01 6.98786378e-01 8.16917837e-01
-5.89010179e-01 1.01257610e+00 3.13755453e-01 -5.65600514e-01
-1.06805801e+00 5.19075096e-01 5.00959933e-01 -9.51299548e-01
-9.92104113e-02 1.07970786e+00 2.73904383e-01 -8.62079561e-01
1.07218161e-01 -3.52470398e-01 -2.79355273e-02 -1.62399605e-01
6.70400500e-01 4.16592866e-01 3.41653883e-01 -9.84340981e-02
-6.15475714e-01 3.96654695e-01 -5.19843400e-01 -2.69075662e-01
1.05659568e+00 3.56256217e-01 2.92472821e-03 -7.41106570e-02
9.15103793e-01 2.06894755e-01 -3.95855159e-01 -1.90801397e-01
1.32288069e-01 1.09753378e-01 -3.02677184e-01 -1.95172167e+00
-5.19607961e-01 8.52831841e-01 7.70415306e-01 -1.92472741e-01
1.09388328e+00 1.68487370e-01 1.09564972e+00 5.57012379e-01
5.69456555e-02 -7.12701023e-01 8.20875689e-02 2.75398999e-01
8.67299676e-01 -1.31913090e+00 -1.80801317e-01 -7.84098208e-02
-8.48103225e-01 5.90974808e-01 6.33762538e-01 2.88379639e-01
6.61693215e-01 6.21139631e-02 4.73607570e-01 -4.69686300e-01
-9.89933193e-01 1.79749355e-01 7.91976035e-01 3.08854669e-01
6.48390830e-01 1.53513625e-01 -4.23300266e-01 9.15753901e-01
-2.68089622e-01 2.81849563e-01 -5.46288602e-02 5.81869006e-01
-2.76216329e-03 -1.18641698e+00 -3.96422654e-01 7.94001460e-01
-1.10504520e+00 -4.21553940e-01 -3.44371706e-01 6.27623498e-01
6.19830191e-02 1.35171652e+00 -3.05616617e-01 -4.32538331e-01
6.45494640e-01 3.02042276e-01 6.74721301e-02 -5.99960148e-01
-1.28401661e+00 -3.55530709e-01 3.03978711e-01 -6.73428774e-01
-5.04256040e-02 -4.00545925e-01 -1.36445844e+00 -3.11926216e-01
9.25101992e-03 5.99041045e-01 3.32795560e-01 7.92401314e-01
9.94836390e-01 7.23284662e-01 -2.46712323e-02 6.00813329e-01
-9.70579684e-01 -8.22859585e-01 2.06849724e-01 6.75872803e-01
3.40890706e-01 4.35584486e-02 -1.21546313e-02 -1.14625223e-01]
|
[8.78553295135498, 8.549151420593262]
|
9ac23656-43f4-41a9-be9e-abc2a1249f96
|
stpls3d-a-large-scale-synthetic-and-real
|
2203.09065
| null |
https://arxiv.org/abs/2203.09065v3
|
https://arxiv.org/pdf/2203.09065v3.pdf
|
STPLS3D: A Large-Scale Synthetic and Real Aerial Photogrammetry 3D Point Cloud Dataset
|
Although various 3D datasets with different functions and scales have been proposed recently, it remains challenging for individuals to complete the whole pipeline of large-scale data collection, sanitization, and annotation. Moreover, the created datasets usually suffer from extremely imbalanced class distribution or partial low-quality data samples. Motivated by this, we explore the procedurally synthetic 3D data generation paradigm to equip individuals with the full capability of creating large-scale annotated photogrammetry point clouds. Specifically, we introduce a synthetic aerial photogrammetry point clouds generation pipeline that takes full advantage of open geospatial data sources and off-the-shelf commercial packages. Unlike generating synthetic data in virtual games, where the simulated data usually have limited gaming environments created by artists, the proposed pipeline simulates the reconstruction process of the real environment by following the same UAV flight pattern on different synthetic terrain shapes and building densities, which ensure similar quality, noise pattern, and diversity with real data. In addition, the precise semantic and instance annotations can be generated fully automatically, avoiding the expensive and time-consuming manual annotation. Based on the proposed pipeline, we present a richly-annotated synthetic 3D aerial photogrammetry point cloud dataset, termed STPLS3D, with more than 16 $km^2$ of landscapes and up to 18 fine-grained semantic categories. For verification purposes, we also provide a parallel dataset collected from four areas in the real environment. Extensive experiments conducted on our datasets demonstrate the effectiveness and quality of the proposed synthetic dataset.
|
['Qingyong Hu', 'Fengbo Ren', 'Kyle McCullough', 'Hugues Thomas', 'Zifan Yu', 'Lucio Soibelman', 'Yu Hou', 'Andrew Feng', 'Meida Chen']
|
2022-03-17
| null | null | null | null |
['3d-instance-segmentation-1']
|
['computer-vision']
|
[ 2.45318115e-02 -5.29808328e-02 5.91245651e-01 -3.03467065e-01
-4.13076103e-01 -7.99434364e-01 5.17947018e-01 2.16652945e-01
1.26622079e-04 7.81918049e-01 -3.92193258e-01 -2.19106749e-02
-2.06252009e-01 -1.60827971e+00 -7.70197272e-01 -3.72729778e-01
-9.60119814e-02 7.66169310e-01 3.44848722e-01 -4.10584182e-01
8.52534622e-02 5.64361453e-01 -2.24561644e+00 -2.49706537e-01
1.39038932e+00 9.10477400e-01 3.73867214e-01 4.26170498e-01
-3.05222601e-01 -1.12065859e-01 -6.73404753e-01 -4.71537858e-01
8.58833671e-01 -6.82964101e-02 -3.06070209e-01 3.93251568e-01
6.72312498e-01 -3.39427918e-01 8.88860896e-02 1.26712298e+00
6.31177187e-01 6.66946843e-02 3.64588857e-01 -1.38871920e+00
-4.54121798e-01 1.41557932e-01 -6.28303051e-01 -5.97024679e-01
2.66999543e-01 3.76969010e-01 6.07166886e-01 -6.93220079e-01
5.02199531e-01 9.83507514e-01 9.60474133e-01 -1.94175690e-02
-1.09577012e+00 -8.70774388e-01 9.03217867e-02 -3.89233261e-01
-1.95119143e+00 9.47607458e-02 7.20976472e-01 -5.44808209e-01
4.58038330e-01 4.89411384e-01 1.39698350e+00 7.74924219e-01
-2.44176373e-01 2.57825911e-01 1.10404181e+00 -1.13903070e-02
3.97985220e-01 2.18104292e-02 -2.87874728e-01 6.07591391e-01
6.23532414e-01 1.03917167e-01 -1.49051800e-01 -2.77331322e-01
9.74243879e-01 1.42174512e-01 -1.82262301e-01 -4.61164296e-01
-1.24571311e+00 3.10255170e-01 4.98023093e-01 -2.46881217e-01
-5.27268350e-01 -4.14369069e-02 1.10498816e-01 -1.07750997e-01
5.31075120e-01 2.17987165e-01 -2.82273382e-01 -6.08513355e-02
-1.15710366e+00 6.87810957e-01 2.52459317e-01 1.55805457e+00
1.03318894e+00 1.70488715e-01 3.65584195e-01 7.10835099e-01
2.27274001e-01 9.87529278e-01 1.30885199e-01 -9.52055871e-01
5.92691183e-01 1.22513998e+00 4.67631012e-01 -1.39819002e+00
-1.62117332e-01 -4.67087686e-01 -1.01898158e+00 4.64668512e-01
6.59258813e-02 -9.82214510e-02 -8.59785080e-01 1.22260511e+00
7.43652046e-01 2.29950145e-01 -1.14216037e-01 1.00111914e+00
7.79583097e-01 5.01539171e-01 4.86269109e-02 2.39566952e-01
1.09176540e+00 -4.43250954e-01 -3.15607578e-01 -1.29627809e-01
2.88967043e-01 -5.26150286e-01 1.37528598e+00 3.22812021e-01
-7.96243608e-01 -6.27419889e-01 -1.10433447e+00 3.23181778e-01
-4.24859196e-01 3.86378080e-01 8.29425514e-01 7.09594905e-01
-8.33633840e-01 5.82604885e-01 -6.26665831e-01 -4.63562816e-01
6.71543717e-01 -1.05339494e-02 -3.79564911e-01 8.50981623e-02
-1.07297552e+00 3.57693702e-01 4.19616312e-01 2.88422346e-01
-8.07143331e-01 -8.47598433e-01 -7.42537379e-01 -1.85554624e-01
1.98109850e-01 -7.64652014e-01 6.47934437e-01 -7.30006874e-01
-1.08868527e+00 1.03758812e+00 4.11004066e-01 -2.34358385e-02
7.95759618e-01 -1.13281474e-01 -3.53014767e-01 -1.41348779e-01
4.53652173e-01 5.38145721e-01 5.00759482e-01 -1.50835943e+00
-7.56474733e-01 -6.55135095e-01 2.62536764e-01 4.25723791e-01
-6.84260428e-02 -3.40947181e-01 -1.67930648e-01 -8.14138353e-01
1.79852709e-01 -8.60173643e-01 -4.31640983e-01 2.75089085e-01
-3.19849581e-01 4.24681425e-01 3.94932866e-01 -5.42087853e-01
1.01216161e+00 -2.21087337e+00 -1.81755811e-01 3.73903006e-01
-6.61241338e-02 2.22204477e-01 6.68638991e-03 5.07808566e-01
6.22620471e-02 5.45062125e-01 -5.50816536e-01 -1.21215448e-01
1.14913069e-01 2.46066421e-01 -3.77130926e-01 2.74449587e-01
6.12978712e-02 4.37470585e-01 -1.01563108e+00 -5.84324300e-01
5.04810333e-01 1.80650219e-01 -3.92979503e-01 8.33290666e-02
-2.87869066e-01 3.68346274e-01 -6.01091325e-01 1.14251530e+00
1.15093684e+00 1.76509082e-01 -5.83181866e-02 2.59333421e-02
-2.23189339e-01 -3.09966505e-01 -1.73195434e+00 1.94613743e+00
-2.59776592e-01 8.45403597e-02 2.90193349e-01 -4.89182562e-01
1.38656700e+00 4.55061235e-02 4.00714129e-01 -2.80224741e-01
-2.73007471e-02 2.82264620e-01 -5.54371893e-01 -2.33525544e-01
8.56217206e-01 -6.79227412e-02 -2.42016882e-01 1.44662142e-01
-3.42300653e-01 -9.18972433e-01 -2.76123849e-03 -1.17437065e-01
6.82423353e-01 6.07339680e-01 2.62084037e-01 -2.99128264e-01
1.64844334e-01 7.92564452e-01 5.50432205e-01 3.94340515e-01
-9.05870050e-02 8.25398088e-01 2.93180775e-02 -6.37911201e-01
-1.24205506e+00 -1.11238587e+00 -2.45406851e-01 2.97042489e-01
5.39247751e-01 -4.79678035e-01 -7.16339171e-01 -2.68307030e-01
1.70650482e-01 5.03582239e-01 -4.12917644e-01 3.80523771e-01
-6.42193854e-02 -9.81508732e-01 8.24390173e-01 1.96509138e-01
9.76828456e-01 -7.39054084e-01 -7.73402393e-01 -5.46321198e-02
-9.56807584e-02 -7.58881271e-01 1.39899865e-01 -5.75703740e-01
-8.11356723e-01 -1.19511592e+00 -2.92982519e-01 -5.63104451e-01
8.68759632e-01 4.65145588e-01 1.14685595e+00 2.55524337e-01
-4.17125463e-01 1.17404386e-01 -6.07639670e-01 -6.17296934e-01
-3.61186117e-02 -8.34383518e-02 1.76984176e-01 -1.72423691e-01
8.04212466e-02 -9.84572768e-01 -5.76700926e-01 5.98973095e-01
-1.04736233e+00 4.31720555e-01 3.28029096e-01 4.47946548e-01
1.14116609e+00 7.60671020e-01 3.56648326e-01 -6.05972588e-01
3.00534099e-01 -4.79011983e-01 -9.25159216e-01 1.47887036e-01
-1.34770215e-01 -5.70120990e-01 6.49952590e-01 -7.58213997e-02
-9.92489755e-01 3.27431262e-01 -7.63073843e-03 -4.92906153e-01
-7.24112689e-01 3.63050133e-01 -5.88238955e-01 -1.07244544e-01
8.53536546e-01 1.30378351e-01 -2.46169388e-01 -2.59180397e-01
4.91133243e-01 8.60702991e-01 6.83573663e-01 -7.72962511e-01
1.39848948e+00 6.33596659e-01 6.90222159e-03 -8.25596452e-01
-3.54420960e-01 -6.13927469e-02 -9.01085138e-01 -2.40475744e-01
6.29485667e-01 -1.22599697e+00 -3.42250347e-01 8.28149319e-01
-8.62420976e-01 -4.15588647e-01 -6.85552180e-01 2.27863863e-01
-4.50983882e-01 3.65405560e-01 2.12892652e-01 -8.29146564e-01
-2.49089912e-01 -9.65959549e-01 1.32500958e+00 2.36303091e-01
1.44283712e-01 -4.81829882e-01 -3.04304734e-02 2.27078408e-01
2.46540885e-02 9.63453710e-01 6.12050593e-01 1.17958970e-02
-8.18458259e-01 -3.32640231e-01 -2.22203076e-01 1.34537265e-01
2.11302295e-01 4.38860089e-01 -8.04659963e-01 5.81279323e-02
-3.15325767e-01 -1.62483275e-01 1.50698662e-01 1.43243568e-02
1.29071367e+00 -6.13411795e-03 -1.73225209e-01 8.62275362e-01
1.64005888e+00 -7.18937889e-02 7.14629829e-01 2.60665089e-01
6.86094403e-01 8.51564109e-01 1.27615988e+00 7.43075728e-01
5.28706789e-01 4.79226291e-01 9.58847642e-01 -1.63906366e-01
1.49711788e-01 -6.86676562e-01 -1.19084679e-01 3.26220334e-01
-4.63101953e-01 -1.38637647e-01 -1.12009966e+00 6.36840999e-01
-1.69801903e+00 -9.83771801e-01 -5.69630027e-01 2.54857278e+00
4.58158612e-01 -1.69020489e-01 -1.35977879e-01 2.33154118e-01
8.03000152e-01 2.23481074e-01 -3.69938880e-01 2.46761099e-01
-3.48252684e-01 2.48378098e-01 7.15814948e-01 1.78455472e-01
-1.08415782e+00 1.01023293e+00 5.40968418e+00 9.37185049e-01
-7.74420798e-01 -2.04477236e-01 1.45856246e-01 -3.50553952e-02
-4.87140119e-01 1.65853173e-01 -6.07033908e-01 5.98833084e-01
5.19859672e-01 -1.60353318e-01 4.87566471e-01 1.08460093e+00
4.11757767e-01 -1.67920828e-01 -4.18962717e-01 1.21120036e+00
-1.93475768e-01 -1.29182446e+00 1.49121299e-01 2.49239355e-01
8.19645345e-01 -8.84792879e-02 -2.76091099e-01 -3.46030556e-02
6.97932005e-01 -8.19448590e-01 9.31899905e-01 5.21507084e-01
1.28790200e+00 -7.44996011e-01 6.00460351e-01 6.48819685e-01
-1.56584954e+00 5.95536344e-02 -6.65904641e-01 -3.06849927e-01
2.27459133e-01 7.40774810e-01 -7.14165032e-01 1.03547418e+00
1.05842340e+00 6.43426955e-01 -7.74132013e-01 1.15808761e+00
-3.93675089e-01 2.08351612e-01 -6.23112917e-01 1.70445964e-01
-7.03861117e-02 -7.70587802e-01 5.08596361e-01 4.50303018e-01
9.14325297e-01 3.08944851e-01 2.91867465e-01 9.24303114e-01
1.29654691e-01 1.84884012e-01 -1.04395068e+00 2.04864107e-02
9.65408444e-01 1.36265731e+00 -6.40145242e-01 -1.35781333e-01
-3.33059616e-02 8.19428504e-01 8.04003561e-04 -2.15496123e-03
-9.18487728e-01 -4.80951399e-01 7.57574201e-01 5.76082885e-01
-2.29632765e-01 -4.74254787e-01 -4.86723721e-01 -1.17166281e+00
1.96136996e-01 -6.24891162e-01 -4.51355949e-02 -1.19100559e+00
-1.16571808e+00 6.11709416e-01 1.75530732e-01 -1.79614711e+00
7.88633600e-02 -3.11846226e-01 -6.10472739e-01 9.95742083e-01
-1.26825500e+00 -1.68815422e+00 -1.33687794e+00 4.76615071e-01
4.94951427e-01 -1.45715371e-01 9.70307171e-01 2.53367305e-01
-4.10538048e-01 -2.57326961e-02 -1.55015551e-02 -1.19140029e-01
4.53168243e-01 -1.24162257e+00 8.00840020e-01 9.72368002e-01
-1.55314222e-01 3.17994952e-01 3.92839760e-01 -1.10907459e+00
-1.17949128e+00 -1.70816255e+00 1.77827522e-01 -4.23898995e-01
3.86592895e-01 -3.75311911e-01 -6.44748688e-01 5.50955474e-01
-3.75885695e-01 4.95365411e-02 5.53061068e-01 -2.94866979e-01
2.61404037e-01 -2.33837619e-01 -1.49888086e+00 4.72811282e-01
1.65057683e+00 -6.63977787e-02 -3.64099294e-01 5.39396286e-01
6.07014537e-01 -5.27756333e-01 -9.94483292e-01 6.27100348e-01
4.00206715e-01 -1.23338568e+00 1.05938327e+00 -4.39643450e-02
2.91067421e-01 -8.65891814e-01 -4.60191309e-01 -1.50118685e+00
-1.30921602e-01 -2.59648830e-01 6.74144149e-01 1.43327117e+00
-4.67192121e-02 -5.43809175e-01 8.54568064e-01 5.67302823e-01
-4.83445317e-01 -3.55404317e-01 -6.14214420e-01 -5.51677048e-01
-2.74443299e-01 -6.89537227e-01 1.47244859e+00 1.00480556e+00
-5.60944617e-01 -3.51041764e-01 -2.00248539e-01 7.57705688e-01
6.94728971e-01 4.32798356e-01 1.56413019e+00 -1.72797763e+00
-1.50862765e-02 -2.16328856e-02 -6.22182429e-01 -5.75506985e-01
-1.37231676e-02 -7.30678260e-01 -1.51685566e-01 -1.67547572e+00
-4.14849162e-01 -1.27386022e+00 7.77521372e-01 2.91719735e-01
-4.29360047e-02 4.02273715e-01 -2.58051276e-01 2.99397260e-01
-6.32049069e-02 8.68251026e-01 1.06010389e+00 1.23086050e-02
-2.77083218e-01 -1.09149061e-01 -4.71441984e-01 9.47888851e-01
7.16683269e-01 -1.48867831e-01 -6.40896380e-01 -8.25594962e-01
3.86112303e-01 -1.86764643e-01 6.69654906e-01 -1.35273838e+00
-1.54987440e-01 -6.14803433e-01 4.05894995e-01 -9.56739426e-01
4.11112756e-01 -1.15066266e+00 1.05875552e+00 5.50820976e-02
5.33806920e-01 2.22719880e-03 1.75272778e-01 3.36599201e-01
-1.23399600e-01 -4.83563729e-02 5.92080057e-01 -5.30472398e-01
-7.65232384e-01 6.87998116e-01 2.57101178e-01 -1.47852227e-01
1.41886258e+00 -6.20105624e-01 -2.77645797e-01 8.15470591e-02
-5.06151974e-01 3.49997520e-01 1.50683355e+00 1.22174732e-01
5.72740316e-01 -1.36329734e+00 -7.06260085e-01 6.75752461e-01
3.64669681e-01 9.24894869e-01 6.49745047e-01 1.31662369e-01
-1.22841036e+00 -2.37525895e-01 -5.80654502e-01 -5.69424570e-01
-9.02263343e-01 2.29835406e-01 3.49579901e-01 1.94590449e-01
-5.78559577e-01 5.66179693e-01 -5.86949624e-02 -1.12412155e+00
-3.67088467e-01 -4.09924239e-01 4.79355454e-02 5.99321462e-02
3.49331945e-01 3.68752569e-01 2.10277259e-01 -8.16093206e-01
-2.10577965e-01 7.98126161e-01 1.05188978e+00 2.54744478e-03
1.37744522e+00 -1.48361951e-01 -1.02855340e-01 1.65872470e-01
1.53899401e-01 2.94843227e-01 -1.23707426e+00 1.72088400e-01
-5.35599768e-01 -1.24585378e+00 -1.56264856e-01 -5.34364402e-01
-9.35490549e-01 8.15928817e-01 5.85262418e-01 2.58500725e-01
1.09009981e+00 -4.45034504e-01 5.88343382e-01 6.38647005e-02
1.23455048e+00 -1.00612569e+00 -6.04567409e-01 -9.84708518e-02
1.00513053e+00 -1.05465329e+00 1.19586349e-01 -7.72246301e-01
-6.55398250e-01 7.63339877e-01 7.45976746e-01 -2.06364170e-01
4.90415782e-01 8.61629173e-02 -7.99851567e-02 -3.50878656e-01
-1.00437529e-01 -1.94539756e-01 -3.04425657e-01 1.22140348e+00
-2.76069403e-01 4.53468323e-01 8.52370039e-02 7.34415889e-01
-8.26173544e-01 1.19071260e-01 7.24987030e-01 7.63177335e-01
-4.03550029e-01 -1.01116419e+00 -7.76897967e-01 5.05862772e-01
2.37499505e-01 7.48528987e-02 -3.17986786e-01 8.94273520e-01
5.97073555e-01 7.16534495e-01 1.43327668e-01 -5.20768821e-01
6.69681609e-01 -3.49873841e-01 7.84042776e-02 -7.57694066e-01
-3.41193020e-01 -2.71429777e-01 8.36365521e-02 -4.88567173e-01
-5.45316398e-01 -6.33029580e-01 -1.08200264e+00 -5.13762712e-01
-1.68924898e-01 1.43477619e-01 7.92931080e-01 2.85486013e-01
5.64105749e-01 2.42508754e-01 5.59184551e-01 -1.24393523e+00
-2.01505110e-01 -9.70328808e-01 -9.91382003e-01 4.28298354e-01
-3.23638409e-01 -1.00439739e+00 -1.61603451e-01 1.32961972e-02]
|
[8.411576271057129, -2.6018614768981934]
|
3ccc7918-4eff-4f7d-9722-363d5a725fa8
|
user-modeling-for-task-oriented-dialogues
|
1811.04369
| null |
http://arxiv.org/abs/1811.04369v1
|
http://arxiv.org/pdf/1811.04369v1.pdf
|
User Modeling for Task Oriented Dialogues
|
We introduce end-to-end neural network based models for simulating users of
task-oriented dialogue systems. User simulation in dialogue systems is crucial
from two different perspectives: (i) automatic evaluation of different dialogue
models, and (ii) training task-oriented dialogue systems. We design a
hierarchical sequence-to-sequence model that first encodes the initial user
goal and system turns into fixed length representations using Recurrent Neural
Networks (RNN). It then encodes the dialogue history using another RNN layer.
At each turn, user responses are decoded from the hidden representations of the
dialogue level RNN. This hierarchical user simulator (HUS) approach allows the
model to capture undiscovered parts of the user goal without the need of an
explicit dialogue state tracking. We further develop several variants by
utilizing a latent variable model to inject random variations into user
responses to promote diversity in simulated user responses and a novel goal
regularization mechanism to penalize divergence of user responses from the
initial user goal. We evaluate the proposed models on movie ticket booking
domain by systematically interacting each user simulator with various dialogue
system policies trained with different objectives and users.
|
['Dilek Hakkani-Tur', 'Izzeddin Gur', 'Pararth Shah', 'Gokhan Tur']
|
2018-11-11
| null | null | null | null |
['user-simulation']
|
['natural-language-processing']
|
[ 2.77445585e-01 5.34127474e-01 2.14959294e-01 -4.51176882e-01
-5.91576934e-01 -7.18671560e-01 8.23920131e-01 -2.64054894e-01
-5.02272606e-01 7.98296213e-01 4.93029028e-01 -3.75129193e-01
2.52303302e-01 -5.81846297e-01 -1.65296152e-01 -3.97562504e-01
1.35250255e-01 9.00574267e-01 1.38173968e-01 -1.00023508e+00
2.60590225e-01 1.18317187e-01 -1.22801542e+00 3.76627266e-01
9.13403034e-01 3.63262087e-01 4.01718169e-01 1.44668102e+00
-1.22541882e-01 1.37942183e+00 -8.78219485e-01 -3.05970367e-02
-5.34285232e-02 -1.10424960e+00 -1.02122176e+00 3.49882632e-01
-2.96795607e-01 -6.02040231e-01 -3.26000750e-01 8.61495912e-01
6.63966596e-01 6.80003583e-01 6.24835074e-01 -8.49200487e-01
-4.22040880e-01 7.83934355e-01 1.99141696e-01 -8.20055678e-02
6.98481441e-01 4.59690452e-01 8.76288176e-01 -4.41207677e-01
4.12141591e-01 1.58064413e+00 1.84575737e-01 1.21339893e+00
-1.18103135e+00 -7.10688308e-02 1.16992831e-01 -2.07389563e-01
-6.98205292e-01 -7.15673268e-01 7.26937950e-01 -4.15578365e-01
1.07090735e+00 2.96531588e-01 2.52898693e-01 1.27799046e+00
8.84393975e-02 9.38266695e-01 8.02369237e-01 -4.21332002e-01
2.09639877e-01 6.49879277e-01 5.84733665e-01 7.32055247e-01
-9.22935128e-01 -6.89296350e-02 -2.71853149e-01 -4.13919121e-01
8.41257095e-01 -3.09081137e-01 -4.07843024e-01 4.07247432e-03
-7.11507976e-01 1.13078165e+00 -1.67549878e-01 1.55161589e-01
-6.27676368e-01 -2.69945502e-01 6.74228132e-01 6.86771154e-01
2.34538808e-01 5.59498429e-01 -4.67114896e-01 -3.28056425e-01
-4.27178681e-01 4.09679979e-01 1.45770240e+00 9.50645804e-01
2.86118388e-01 5.07296443e-01 -7.54467368e-01 1.30874634e+00
1.32141903e-01 3.85691077e-02 8.29404831e-01 -1.06169903e+00
3.06508869e-01 4.67334092e-01 5.74764907e-01 -5.34570992e-01
-3.72570336e-01 -5.65592162e-02 -6.96793020e-01 -1.59090683e-01
4.62198228e-01 -6.85001373e-01 -5.54005921e-01 1.85722494e+00
9.36989188e-02 -2.69836694e-01 4.68553334e-01 8.72245967e-01
9.68257010e-01 9.94652092e-01 6.30768016e-02 -5.36335409e-01
1.16544700e+00 -1.17605436e+00 -9.94156897e-01 8.28948990e-02
7.75147140e-01 -3.88576984e-01 1.38415980e+00 3.98853511e-01
-1.44964337e+00 -6.51913345e-01 -7.01508582e-01 6.07755445e-02
1.29954016e-04 1.62055343e-01 1.69333238e-02 5.22629201e-01
-1.15864122e+00 3.82767737e-01 -4.29907471e-01 -8.64260867e-02
-6.61701679e-01 4.76619542e-01 2.76290655e-01 5.28777957e-01
-1.69217825e+00 9.64746952e-01 2.65267968e-01 2.99111396e-01
-1.11218655e+00 5.84637187e-02 -1.09355569e+00 2.35932544e-01
3.55278552e-01 -5.05560398e-01 1.91029871e+00 -1.09188485e+00
-2.55162597e+00 4.03231561e-01 -7.61261657e-02 -5.39482713e-01
5.92657387e-01 -1.41064331e-01 -4.49668095e-02 -1.80052355e-01
-4.37188447e-01 3.55801105e-01 6.43811464e-01 -1.17881131e+00
-2.56842762e-01 -1.02047198e-01 3.26086789e-01 7.07185924e-01
-1.21771693e-01 1.52127445e-01 -3.79834473e-01 -1.73185632e-01
-3.51279646e-01 -9.76340055e-01 -5.11995912e-01 -8.55809987e-01
-5.11481583e-01 -4.96339023e-01 3.34434539e-01 -8.33725154e-01
1.30468452e+00 -1.55188787e+00 5.94274819e-01 -9.00396407e-02
2.46492222e-01 4.21147197e-01 -3.66920918e-01 6.47512913e-01
2.23512456e-01 -2.83666134e-01 -6.64228871e-02 -5.55570722e-01
6.63041845e-02 1.81878388e-01 -2.50999004e-01 2.15960350e-02
-1.25427604e-01 8.46435428e-01 -9.03126061e-01 -1.61030889e-01
2.90214628e-01 1.41867712e-01 -6.27668858e-01 1.08318460e+00
-6.90321326e-01 6.86236560e-01 -3.11526358e-01 -1.72795802e-01
1.76308788e-02 -1.60985112e-01 4.82684523e-01 2.67656356e-01
-8.35014582e-02 6.93916559e-01 -7.66466320e-01 1.24233711e+00
-7.09278762e-01 3.52313966e-01 2.76352584e-01 -8.05870593e-01
1.08788848e+00 7.25670397e-01 1.81007388e-04 -6.24706209e-01
3.99946451e-01 -4.28904861e-01 2.74776816e-01 -5.95178604e-01
9.23795760e-01 -1.06868334e-01 -4.73679394e-01 8.52225363e-01
2.47564420e-01 -8.07823837e-02 3.26389372e-02 3.25506985e-01
6.45133734e-01 -7.27035701e-02 2.76727170e-01 5.01798429e-02
9.11418378e-01 -2.46250570e-01 2.76387900e-01 1.02367353e+00
-1.81069642e-01 3.43494058e-01 9.33522701e-01 -2.06106916e-01
-1.03903222e+00 -5.14354467e-01 4.00267035e-01 1.75414312e+00
-2.18127742e-01 -1.61862895e-01 -1.12637687e+00 -6.15687251e-01
-4.13382709e-01 1.23842430e+00 -5.46336114e-01 -2.63294280e-01
-6.71465933e-01 -4.18276429e-01 5.99907219e-01 1.79687604e-01
4.07193750e-01 -1.58784139e+00 -4.63962734e-01 6.14449978e-01
-3.97223979e-01 -8.22085857e-01 -6.37922883e-01 1.72568560e-01
-5.82200289e-01 -6.78931057e-01 -9.15688217e-01 -7.68414378e-01
4.59995180e-01 -1.63589835e-01 9.98430014e-01 -3.49080004e-02
3.84172916e-01 4.72525299e-01 -3.33487988e-01 5.15682139e-02
-1.44690371e+00 1.93499312e-01 2.34469980e-01 -6.54016361e-02
9.16509330e-02 -1.01901755e-01 -2.75818110e-01 4.72754002e-01
-7.25135922e-01 2.85654902e-01 -9.36496351e-03 1.18480146e+00
-3.59111190e-01 -4.83326018e-01 8.11617851e-01 -1.28452337e+00
1.79593015e+00 -5.09398043e-01 -5.41759670e-01 4.66829419e-01
-2.94034243e-01 3.47446352e-01 1.20908558e+00 -6.92624927e-01
-1.35283697e+00 -1.42673224e-01 -4.78024006e-01 -1.48624644e-01
-1.66114062e-01 3.58572245e-01 5.81489615e-02 5.16710520e-01
8.66208434e-01 4.86482859e-01 3.23572814e-01 -4.22267765e-01
5.15951514e-01 1.03471684e+00 3.90629083e-01 -6.66438460e-01
1.62301987e-01 -5.69165528e-01 -8.51355553e-01 -1.09016705e+00
-4.10285562e-01 -3.69082272e-01 -4.95019585e-01 -3.81002605e-01
6.86607182e-01 -5.08411467e-01 -1.17976427e+00 6.09917402e-01
-1.43799436e+00 -7.69228876e-01 2.33467929e-02 1.30201384e-01
-8.49962115e-01 5.55801213e-01 -1.19193602e+00 -1.40543890e+00
-5.56364536e-01 -1.38369417e+00 6.83134854e-01 3.58915240e-01
-6.05221331e-01 -1.11111069e+00 2.33978108e-01 2.06046447e-01
2.99494237e-01 -2.64426351e-01 9.84458029e-01 -1.30967247e+00
8.59537795e-02 -1.80480555e-01 2.05900222e-01 5.32391489e-01
1.08385228e-01 -2.52498686e-01 -8.17384958e-01 -2.70727336e-01
4.46385801e-01 -8.83502960e-01 1.99843615e-01 2.34757587e-01
6.50766492e-01 -8.30618083e-01 2.30066568e-01 3.04020178e-02
6.87103808e-01 4.97170746e-01 4.41073954e-01 -2.34088480e-01
4.65580344e-01 1.10755706e+00 3.81639540e-01 6.96827412e-01
3.92987996e-01 7.33407736e-01 -4.86659817e-02 1.12779811e-01
5.32191515e-01 -3.17550987e-01 7.12794840e-01 1.10745966e+00
2.17584342e-01 -6.69335604e-01 -7.39049196e-01 3.10767472e-01
-1.98026395e+00 -7.72392094e-01 1.07299007e-01 2.19254327e+00
1.07079685e+00 3.24548006e-01 4.95374441e-01 -3.44254971e-01
8.48546803e-01 1.96904868e-01 -6.97894394e-01 -9.10778940e-01
3.32609564e-01 -1.04327731e-01 2.08626851e-04 1.17139590e+00
-5.56512833e-01 1.31852055e+00 5.98981142e+00 3.18025589e-01
-9.39722240e-01 -6.85598925e-02 6.41981721e-01 1.61256701e-01
-1.77211285e-01 -1.92738667e-01 -8.58255863e-01 3.72496367e-01
1.32522678e+00 -2.26360306e-01 5.08088529e-01 6.54114187e-01
6.39006376e-01 1.44010223e-02 -1.04615247e+00 4.38698202e-01
-1.89347118e-02 -9.63608801e-01 5.47319278e-02 -1.57956481e-01
2.78981626e-01 -3.95180821e-01 -1.61894083e-01 8.99173439e-01
9.45602596e-01 -9.24448133e-01 1.61329895e-01 5.90691924e-01
5.99196076e-01 -6.85060501e-01 6.01415753e-01 1.00869489e+00
-5.43190360e-01 2.74595302e-02 -3.37557852e-01 -9.70864296e-02
3.41282278e-01 -4.54911917e-01 -1.43057883e+00 1.50197996e-02
-2.54005462e-01 5.24963625e-02 -7.17128068e-02 2.97105253e-01
-9.66140628e-02 6.73666298e-01 -1.45483091e-01 -5.29694200e-01
4.78383750e-01 -3.15668136e-01 6.09716535e-01 1.22630024e+00
-3.35776627e-01 3.04251909e-01 4.60040182e-01 8.38593066e-01
7.21896365e-02 1.56950951e-01 -6.87657058e-01 -2.07345247e-01
3.32837671e-01 1.07805991e+00 -2.05848292e-01 -3.71915221e-01
-5.39943762e-02 1.25448489e+00 3.57107818e-01 5.10358989e-01
-6.67566419e-01 -3.65266055e-01 2.89344788e-01 -2.01425612e-01
-2.42809519e-01 -1.94996297e-01 1.01267025e-01 -1.10325837e+00
-5.12107909e-01 -1.21484947e+00 1.18931480e-01 -5.60622990e-01
-8.39923978e-01 1.04068279e+00 -1.52711943e-01 -8.31633151e-01
-1.19601178e+00 -1.69489548e-01 -6.46924257e-01 1.35949278e+00
-7.80615091e-01 -5.50266922e-01 6.93976134e-02 5.43343484e-01
1.24574900e+00 -4.58904713e-01 1.19353783e+00 -2.87209749e-01
-8.02421153e-01 8.64647031e-01 1.27998620e-01 2.71612674e-01
5.81579208e-01 -1.28229439e+00 6.83339894e-01 2.10009530e-01
-4.89397079e-01 8.43277574e-01 1.01006317e+00 -6.85593486e-01
-9.93143976e-01 -6.61754549e-01 7.61551321e-01 -3.10311884e-01
3.61158043e-01 -5.84939182e-01 -1.04248571e+00 6.06174588e-01
5.02343178e-01 -9.46530640e-01 6.73317492e-01 9.64576751e-02
3.43305379e-01 6.56326413e-01 -9.35890496e-01 9.72144902e-01
3.37417454e-01 -6.77115858e-01 -7.63868093e-01 4.22020614e-01
9.34549570e-01 -5.36716163e-01 -5.57223082e-01 -3.95336468e-03
4.59626049e-01 -8.21555555e-01 6.31905735e-01 -1.07151949e+00
3.23987424e-01 2.92608917e-01 1.55184075e-01 -1.41396511e+00
-1.51611701e-01 -1.28503048e+00 -1.21971421e-01 9.77236152e-01
3.68450373e-01 -2.98881710e-01 8.40801656e-01 1.18479812e+00
1.96180083e-02 -6.60753846e-01 -3.57413024e-01 -2.23599881e-01
1.12739325e-01 1.27561260e-02 1.13151744e-01 6.85157597e-01
3.63198638e-01 1.13876379e+00 -1.11368930e+00 -2.20832288e-01
1.23184718e-01 -3.46191645e-01 9.44834948e-01 -8.26572478e-01
-5.93156219e-01 -4.28039134e-01 4.88738447e-01 -1.79729640e+00
3.20836991e-01 -4.37602192e-01 4.95843232e-01 -1.18980229e+00
-2.56708503e-01 -5.45515195e-02 5.70788570e-02 1.00752795e-02
-3.10828924e-01 -6.76885605e-01 1.51675180e-01 1.87924251e-01
-6.69932604e-01 9.42256927e-01 1.16865563e+00 1.65778205e-01
-8.91759276e-01 5.97986639e-01 -4.75127161e-01 6.51424170e-01
7.26458967e-01 -2.13621661e-01 -7.53531098e-01 9.32644010e-02
-2.22252980e-01 1.22387433e+00 -5.03072366e-02 -5.06923974e-01
3.70693684e-01 -1.50888801e-01 -8.20000544e-02 -4.89663899e-01
3.50373358e-01 -2.62593836e-01 -4.08676893e-01 4.52828377e-01
-1.25241160e+00 1.37901917e-01 3.32652889e-02 4.53353643e-01
6.10753223e-02 -6.96817100e-01 9.54015076e-01 -5.24561107e-01
-3.28061014e-01 -9.83239710e-02 -1.24530113e+00 1.02477754e-02
5.88694334e-01 -1.66205823e-01 1.11369409e-01 -1.16352296e+00
-1.02559149e+00 6.76297069e-01 8.25969800e-02 5.02586424e-01
6.10680461e-01 -7.59399593e-01 -7.47916222e-01 2.76719928e-01
-1.70388669e-01 -2.19775498e-01 3.57149184e-01 1.13936648e-01
-3.85047257e-01 5.02105653e-01 -8.98241922e-02 -3.82648319e-01
-1.33882129e+00 2.25856751e-01 7.29315996e-01 -7.69294977e-01
-2.78034061e-01 8.64335537e-01 3.76475155e-01 -1.12703300e+00
7.32869446e-01 1.30188465e-01 -8.39682460e-01 1.06653899e-01
4.45264250e-01 2.77352542e-01 -2.91047513e-01 -5.68607748e-01
3.21127415e-01 -3.24040115e-01 -6.96399212e-01 -5.72993279e-01
9.86135244e-01 -3.54449928e-01 2.76339144e-01 7.47442901e-01
9.78108585e-01 -4.33242142e-01 -1.23846936e+00 -5.22831261e-01
-6.37703687e-02 1.07437611e-01 -3.98865461e-01 -9.72385705e-01
-2.13291809e-01 7.79964328e-01 2.30757058e-01 5.85327506e-01
6.69698417e-01 -5.47127426e-01 9.68664289e-01 8.18557680e-01
1.82094052e-01 -1.23991024e+00 2.99190134e-01 1.29306912e+00
9.80095208e-01 -1.04255247e+00 -5.38360119e-01 2.57347226e-01
-1.37569582e+00 1.10847938e+00 8.63675892e-01 -1.25031546e-01
2.82738149e-01 3.42120901e-02 2.98698604e-01 -5.10717556e-02
-1.24662411e+00 -9.05177463e-03 9.27221999e-02 2.48692214e-01
7.16678619e-01 -1.10490754e-01 -2.89296240e-01 7.05607116e-01
-1.76178172e-01 -7.56080747e-02 7.90206969e-01 6.89574182e-01
-5.22571564e-01 -1.20759869e+00 -9.33389217e-02 2.30726719e-01
-2.21648395e-01 -1.51522085e-01 -6.68349683e-01 2.89836705e-01
-8.92255187e-01 1.02359855e+00 -1.87720239e-01 -5.44493735e-01
5.47002137e-01 5.65950692e-01 1.45529345e-01 -9.81548965e-01
-1.35847783e+00 2.33133674e-01 5.25965393e-01 -7.86319599e-02
3.25336426e-01 -4.09452319e-01 -1.03071952e+00 -1.93880141e-01
-5.13389647e-01 6.38415277e-01 3.63904744e-01 9.29899514e-01
6.19653836e-02 5.71151793e-01 1.10055709e+00 -7.36314416e-01
-1.29049122e+00 -1.49577832e+00 -4.05316353e-01 4.11236495e-01
3.78512710e-01 -1.85498491e-01 -1.92383692e-01 -1.21885695e-01]
|
[12.949019432067871, 8.010336875915527]
|
4442c759-a521-4925-9bee-2be8ac753702
|
automatic-image-cropping-a-computational
| null | null |
http://openaccess.thecvf.com/content_cvpr_2016/html/Chen_Automatic_Image_Cropping_CVPR_2016_paper.html
|
http://openaccess.thecvf.com/content_cvpr_2016/papers/Chen_Automatic_Image_Cropping_CVPR_2016_paper.pdf
|
Automatic Image Cropping : A Computational Complexity Study
|
Attention based automatic image cropping aims at preserving the most visually important region in an image. A common task in this kind of method is to search for the smallest rectangle inside which the summed attention is maximized. We demonstrate that under appropriate formulations, this task can be achieved using efficient algorithms with low computational complexity. In a practically useful scenario where the aspect ratio of the cropping rectangle is given, the problem can be solved with a computational complexity linear to the number of image pixels. We also study the possibility of multiple rectangle cropping and a new model facilitating fully automated image cropping.
|
['Gaocheng Bai', 'Jiansheng Chen', 'Zhengqin Li', 'Shaoheng Liang']
|
2016-06-01
| null | null | null |
cvpr-2016-6
|
['image-cropping']
|
['computer-vision']
|
[ 6.83045924e-01 4.13592905e-01 2.55084932e-01 2.64941335e-01
-6.67398930e-01 -5.01964748e-01 1.52274936e-01 2.37504318e-01
-6.64520502e-01 5.18301606e-01 -2.83888876e-01 -1.48114905e-01
-3.98358069e-02 -7.25389123e-01 -8.37476194e-01 -7.88876891e-01
3.42171490e-01 5.75993396e-02 7.53101557e-02 -2.76725501e-01
5.81413269e-01 7.68046260e-01 -1.69211614e+00 -2.61071265e-01
9.39738929e-01 6.07396007e-01 8.66208732e-01 9.69491243e-01
7.24254772e-02 3.57698023e-01 -5.65370023e-01 -1.21287540e-01
4.33578104e-01 -1.14657678e-01 -7.25915551e-01 6.40268087e-01
4.93495882e-01 -3.52304764e-02 1.42260700e-01 1.40385032e+00
3.36205900e-01 2.87417382e-01 6.04569674e-01 -1.06856894e+00
-6.80673778e-01 3.99634272e-01 -1.05110562e+00 4.44445610e-02
4.79934588e-02 -3.17674011e-01 8.58585298e-01 -1.04163766e+00
6.88802481e-01 8.82478237e-01 1.46287754e-01 1.60456702e-01
-1.32847488e+00 -3.28790784e-01 2.33232081e-01 6.38054237e-02
-1.49599493e+00 3.71810608e-02 7.69182801e-01 -2.61339694e-01
7.19284296e-01 2.88597643e-01 4.99315143e-01 1.53128698e-01
2.00680420e-01 4.95164275e-01 8.85132611e-01 -1.05264199e+00
1.94242314e-01 2.26570845e-01 1.22107446e-01 3.86325836e-01
5.33778489e-01 -2.96953827e-01 -6.52630553e-02 2.16984153e-01
7.63719141e-01 -1.57082409e-01 -2.80718774e-01 -5.70281565e-01
-1.20739114e+00 7.82363772e-01 5.21234572e-01 5.25668442e-01
-5.00578046e-01 1.06386498e-01 -7.26506785e-02 1.35296984e-02
4.88574386e-01 9.30130899e-01 -1.46054223e-01 4.00200516e-01
-9.69788611e-01 3.32772553e-01 3.22190225e-01 1.03516388e+00
9.23442543e-01 -4.24415544e-02 9.78248045e-02 7.23348558e-01
-2.78493643e-01 6.50331616e-01 2.78075617e-02 -9.32508230e-01
4.09318119e-01 5.40226996e-01 5.92664480e-01 -1.07107973e+00
-4.16075796e-01 -3.63071769e-01 -7.64608741e-01 7.48282075e-01
6.74655259e-01 -3.67670029e-01 -7.31512427e-01 1.61523700e+00
3.26917142e-01 -4.63871062e-01 5.13774669e-03 7.12013781e-01
3.65340672e-02 6.14472210e-01 1.20337456e-01 -3.19158822e-01
1.66944301e+00 -7.47538686e-01 -7.64903069e-01 -4.68711048e-01
3.22324514e-01 -9.85457063e-01 1.01881313e+00 4.25404578e-01
-1.09874225e+00 -3.19198698e-01 -1.22235906e+00 -3.62899452e-01
-5.00290215e-01 5.88570178e-01 2.96822965e-01 6.41160786e-01
-1.10943329e+00 2.72806436e-01 -1.73297584e-01 -4.29481059e-01
1.20404221e-01 4.02086735e-01 -5.52496672e-01 1.29516954e-02
-6.79724753e-01 1.15417612e+00 6.63625062e-01 2.40087897e-01
-4.17372286e-01 -4.77271587e-01 -9.13560033e-01 4.87749875e-01
6.35428369e-01 -3.94039065e-01 9.13559079e-01 -1.19228888e+00
-1.01434910e+00 9.17567611e-01 -4.25368935e-01 -5.31764567e-01
4.63450879e-01 -3.56038511e-01 3.01637977e-01 3.00348252e-01
1.57888249e-01 9.84771550e-01 1.30009294e+00 -1.27322233e+00
-7.37882555e-01 -3.66405874e-01 3.04951370e-01 5.37146866e-01
-3.46064866e-01 9.92958397e-02 -3.33733380e-01 -8.22151840e-01
-7.03969672e-02 -1.16362739e+00 -6.17398143e-01 1.38679057e-01
-4.21801418e-01 2.82344650e-02 8.07271898e-01 -7.76425302e-01
9.64096308e-01 -2.19328427e+00 3.28073412e-01 1.06606610e-01
1.66108981e-01 7.20481724e-02 -1.56508565e-01 2.62487292e-01
-3.00854862e-01 1.61555842e-01 -5.00869215e-01 -9.78285819e-02
-3.51452112e-01 -2.44056106e-01 -2.16232806e-01 4.85925645e-01
4.88024086e-01 8.76297235e-01 -6.24213934e-01 -4.92047727e-01
5.02276897e-01 4.87486035e-01 -5.03358245e-01 8.69142711e-02
-6.19843490e-02 2.04814255e-01 -1.32545605e-01 2.59678960e-01
9.36717153e-01 -2.23082110e-01 2.36114055e-01 -1.95129644e-02
-3.20477247e-01 -5.34821451e-01 -1.22318172e+00 1.37906063e+00
-7.06942916e-01 1.18647742e+00 1.02347270e-01 -9.25526738e-01
9.24889088e-01 -8.81049410e-02 4.01946247e-01 -4.30741966e-01
1.08011886e-01 -1.13732755e-01 -2.63805747e-01 -2.03843579e-01
1.08493888e+00 4.51511890e-02 -9.75985453e-02 4.00508076e-01
-4.55392987e-01 -1.57170534e-01 2.72409081e-01 -1.57584641e-02
5.76274812e-01 -1.76139802e-01 8.36155176e-01 -5.68850279e-01
3.80186588e-01 1.06245555e-01 3.15893181e-02 6.71303689e-01
3.12401038e-02 7.33917892e-01 6.33917868e-01 -1.38054326e-01
-1.43676996e+00 -5.25969505e-01 6.82390155e-03 9.36178505e-01
4.31156963e-01 1.02590591e-01 -1.39012039e+00 -3.05197954e-01
-2.83166260e-01 8.02172244e-01 -1.11844540e+00 2.70923764e-01
-8.10046434e-01 -7.47715771e-01 -2.53872842e-01 3.47960621e-01
4.29441422e-01 -1.31817031e+00 -1.30852568e+00 3.02309375e-02
-2.90971667e-01 -1.19679475e+00 -6.11091197e-01 2.42650568e-01
-6.64858818e-01 -1.29193401e+00 -1.42555130e+00 -1.07936585e+00
1.23296726e+00 8.44278038e-01 8.33876669e-01 7.75510892e-02
-5.64003944e-01 2.49779999e-01 -2.76648939e-01 -5.00089943e-01
-1.86560616e-01 1.64704904e-01 -4.41488951e-01 -1.19414134e-02
7.56146610e-02 -1.23286471e-01 -4.72735137e-01 1.04612008e-01
-1.08896112e+00 2.47366726e-01 5.37852705e-01 7.62712896e-01
6.08260989e-01 3.54677886e-01 3.77356768e-01 -6.20655537e-01
4.80146140e-01 -1.00381725e-01 -9.78263080e-01 4.62942600e-01
-2.68059015e-01 2.05213696e-01 5.74212790e-01 -4.61419582e-01
-9.19767320e-01 5.40479362e-01 3.47627193e-01 -1.18170850e-01
-3.99261015e-03 7.24239200e-02 -1.80699587e-01 -2.83668280e-01
3.55545938e-01 2.84435302e-01 -1.96328983e-01 -1.86816812e-01
5.67809582e-01 3.72326612e-01 5.11947751e-01 8.19948222e-03
7.17861652e-01 5.25646508e-01 2.91380405e-01 -1.31355417e+00
-4.95444745e-01 -2.32239470e-01 -9.76166308e-01 -2.43396893e-01
9.24607038e-01 -4.67430890e-01 -6.73996210e-01 2.72870570e-01
-1.36992633e+00 -6.56356364e-02 -3.83057624e-01 1.35683805e-01
-6.84998751e-01 5.98246753e-01 2.12576136e-01 -1.13380325e+00
-3.66587371e-01 -1.06528211e+00 1.09618616e+00 3.81618440e-01
-8.26629922e-02 -7.31896639e-01 -4.08147931e-01 4.81007475e-04
1.54393449e-01 3.50666821e-01 1.06434059e+00 -8.47610533e-02
-5.38786471e-01 -1.11182950e-01 -6.79487348e-01 1.52261565e-02
1.73978344e-01 -2.80702561e-02 -7.28592992e-01 -2.90217847e-01
-2.86017179e-01 2.92517185e-01 7.41950452e-01 8.19003880e-01
9.77273583e-01 -3.14794742e-02 -2.25678459e-01 2.99461633e-01
1.56645381e+00 4.48546916e-01 6.23050869e-01 3.88223767e-01
5.60893357e-01 9.27510023e-01 9.52536523e-01 2.52423167e-01
-1.54137105e-01 8.00941706e-01 4.78580683e-01 -3.46697003e-01
-1.25643373e-01 5.71181588e-02 -1.94875225e-01 6.90474808e-02
1.39844716e-01 -3.81907761e-01 -7.41833210e-01 1.03348780e+00
-1.72181988e+00 -7.63545752e-01 2.00086609e-02 2.51596904e+00
3.08295995e-01 -1.85642049e-01 -9.90275517e-02 3.00935835e-01
1.18610632e+00 -7.37841800e-02 -5.36255538e-01 -4.32112426e-01
-3.07168812e-01 -2.10528784e-02 7.43907690e-01 7.46431172e-01
-1.18775415e+00 9.79827344e-01 7.37842512e+00 6.99029326e-01
-9.32733357e-01 -2.04895869e-01 7.58652925e-01 3.16289254e-02
-3.92542705e-02 -5.29657155e-02 -7.41030812e-01 4.16549474e-01
3.90224725e-01 -5.02243459e-01 4.42239523e-01 6.78707600e-01
2.23963216e-01 -7.45734215e-01 -6.03151500e-01 9.59419668e-01
2.34663621e-01 -8.89224827e-01 -3.84998322e-02 1.27357513e-01
6.77498341e-01 -6.13113880e-01 3.27028573e-01 -3.11110348e-01
-5.08883893e-02 -8.55089068e-01 4.97766405e-01 1.10292174e-01
8.11713398e-01 -1.08796477e+00 3.14285129e-01 2.06427395e-01
-1.27917325e+00 -2.79598832e-01 -4.99358237e-01 7.80105442e-02
1.83897376e-01 4.60807711e-01 -7.42262721e-01 2.59780496e-01
5.03823817e-01 2.90843368e-01 -4.78108495e-01 1.13139510e+00
-2.95135051e-01 -1.71804160e-01 -3.82263303e-01 -6.36963695e-02
2.85680979e-01 -2.19741821e-01 6.38825059e-01 1.10297787e+00
6.33656383e-01 -5.28158993e-03 -3.35432202e-01 6.64396882e-01
-1.15187662e-02 4.79050994e-01 -7.64455736e-01 2.35854268e-01
3.11439633e-01 1.20585501e+00 -1.31682062e+00 -2.93756902e-01
-1.09148718e-01 1.11130178e+00 5.49121261e-01 3.41813117e-01
-6.18956745e-01 -8.02291811e-01 3.21191043e-01 1.03661701e-01
7.64239848e-01 -2.46113941e-01 -3.90420198e-01 -8.27207029e-01
1.18404679e-01 -5.29097021e-01 9.75587294e-02 -1.04911256e+00
-4.36478794e-01 7.35539794e-01 1.03265680e-01 -1.05139744e+00
-7.10820332e-02 -6.06225252e-01 -3.73355538e-01 8.27958465e-01
-1.39359140e+00 -9.77391541e-01 -3.64467025e-01 2.23744318e-01
8.41671169e-01 1.87792018e-01 6.40617132e-01 -2.48960070e-02
-5.63469350e-01 3.24692488e-01 1.16122000e-01 -2.02566564e-01
4.33190674e-01 -1.37102568e+00 4.67061579e-01 1.29493380e+00
2.11005867e-01 2.71975279e-01 1.17797911e+00 -4.78592366e-01
-1.00344837e+00 -9.56390440e-01 1.22724330e+00 -6.80854768e-02
2.24761322e-01 -3.60140532e-01 -8.15424204e-01 5.08399189e-01
4.86741185e-01 -2.86423892e-01 6.82601631e-02 -2.92647421e-01
2.33363211e-02 2.37124622e-01 -1.09071958e+00 8.32268476e-01
5.49064517e-01 -1.06370142e-02 -2.99595565e-01 4.23771590e-01
7.04782248e-01 -3.21923018e-01 -5.07797956e-01 -3.77915017e-02
4.04369086e-01 -6.11320972e-01 9.21425402e-01 -1.89200029e-01
4.02962565e-01 -3.88834476e-01 1.02549382e-01 -1.33365762e+00
-2.50478685e-01 -8.35915029e-01 3.18048418e-01 8.02233577e-01
3.42767477e-01 -4.79412585e-01 5.95570028e-01 5.96533298e-01
5.09044111e-01 -4.67873901e-01 -9.37409222e-01 -4.71149772e-01
7.55580291e-02 -1.05998479e-01 4.15885448e-01 4.48460162e-01
-6.28937855e-02 -2.37311958e-03 -6.88876569e-01 3.04175466e-01
5.66858172e-01 2.88851827e-01 6.58491731e-01 -1.12131250e+00
6.37239218e-02 -2.91473240e-01 -3.51064354e-01 -9.07021463e-01
3.17752473e-02 -3.16201359e-01 8.12574923e-02 -1.47583783e+00
2.92341113e-01 -1.93276390e-01 -6.65457547e-02 4.23880935e-01
-3.71047884e-01 3.56633455e-01 5.84435761e-01 -1.74339786e-01
-2.95992374e-01 1.89773157e-01 1.26257682e+00 -1.35089025e-01
-2.82695115e-01 -4.89654839e-02 -1.03496277e+00 5.62539876e-01
8.83848369e-01 -5.57552278e-01 -2.94016421e-01 -3.32670629e-01
2.68803120e-01 -1.12499762e-02 3.08857322e-01 -8.33782673e-01
2.05566272e-01 -1.46638647e-01 2.43743852e-01 -7.75662482e-01
4.46593225e-01 -9.84216988e-01 -8.52981731e-02 3.70132953e-01
-4.85877544e-01 2.75503278e-01 4.45004672e-01 5.06820202e-01
-3.40181813e-02 -7.34892726e-01 1.06770158e+00 -1.50498729e-02
-6.47938848e-01 -1.16427973e-01 -5.19558847e-01 -3.31567258e-01
1.62711012e+00 -2.44030923e-01 -6.70489445e-02 -5.10511339e-01
-8.14934433e-01 -1.45074409e-02 5.81585646e-01 4.38577682e-01
4.87738520e-01 -9.99062359e-01 -6.70959234e-01 -3.98369972e-03
-2.29747295e-02 -1.96471810e-01 4.73312378e-01 6.31193101e-01
-8.04508150e-01 5.90954423e-01 -4.31696415e-01 -3.12561542e-01
-1.79358423e+00 1.03801894e+00 2.36492708e-01 -2.69522607e-01
-8.17787170e-01 4.36993241e-01 7.13093638e-01 3.61670166e-01
-4.13364954e-02 -1.91173017e-01 -5.30654192e-01 1.35655746e-01
7.67492831e-01 4.32674021e-01 -8.54976922e-02 -8.16445947e-01
-1.72942176e-01 1.09234190e+00 -1.70316830e-01 -1.34406462e-01
1.04756546e+00 -5.00038803e-01 -1.52410753e-02 -8.77224877e-02
9.92292762e-01 7.52191544e-02 -1.44169950e+00 4.69210558e-03
-1.31155536e-01 -5.85167587e-01 2.03518823e-01 -3.43068063e-01
-1.03777134e+00 1.00699091e+00 6.22291028e-01 3.58930737e-01
1.50032651e+00 -3.49250376e-01 2.48006433e-01 3.15259486e-01
3.87079984e-01 -1.00653195e+00 -3.21400136e-01 1.47961959e-01
1.17698002e+00 -1.26377738e+00 2.82994330e-01 -7.24106848e-01
-6.90703392e-01 1.22055984e+00 4.43704754e-01 -5.04531085e-01
3.78573000e-01 4.02835965e-01 -1.26403287e-01 1.98517680e-01
-3.90570045e-01 -4.46047753e-01 2.26442337e-01 8.04139197e-01
1.91005215e-01 -4.85579371e-02 -4.05237287e-01 -6.16552979e-02
-9.54794586e-02 -3.74597907e-01 8.21246922e-01 7.49277592e-01
-7.47660339e-01 -8.70241702e-01 -6.06400609e-01 3.76751162e-02
-4.49765682e-01 -2.42178679e-01 -2.99791038e-01 8.90166938e-01
-7.67847002e-02 9.32719469e-01 2.50264913e-01 3.60539079e-01
2.00497389e-01 -2.43652403e-01 5.52213609e-01 -4.18171197e-01
-4.32258606e-01 8.58340785e-02 -4.23661649e-01 -1.96851090e-01
-2.26511225e-01 -4.41719472e-01 -9.12215531e-01 7.49765337e-02
-5.94906569e-01 -1.35510443e-02 7.09886193e-01 6.90300226e-01
3.29187095e-01 4.36252952e-01 5.46028733e-01 -1.03247344e+00
-2.97132265e-02 -6.94510877e-01 -7.52022564e-01 9.49643925e-02
4.27387506e-01 -5.25932550e-01 -1.74221262e-01 2.01305360e-01]
|
[11.002233505249023, -1.0600452423095703]
|
36e59bad-3d78-4fdd-bf52-572cfc735ae7
|
shape-from-projections-via-differentiable
|
2006.16120
| null |
https://arxiv.org/abs/2006.16120v4
|
https://arxiv.org/pdf/2006.16120v4.pdf
|
Shape from Projections via Differentiable Forward Projector for Computed Tomography
|
In computed tomography, the reconstruction is typically obtained on a voxel grid. In this work, however, we propose a mesh-based reconstruction method. For tomographic problems, 3D meshes have mostly been studied to simulate data acquisition, but not for reconstruction, for which a 3D mesh means the inverse process of estimating shapes from projections. In this paper, we propose a differentiable forward model for 3D meshes that bridge the gap between the forward model for 3D surfaces and optimization. We view the forward projection as a rendering process, and make it differentiable by extending recent work in differentiable rendering. We use the proposed forward model to reconstruct 3D shapes directly from projections. Experimental results for single-object problems show that the proposed method outperforms traditional voxel-based methods on noisy simulated data. We also apply the proposed method on electron tomography images of nanoparticles to demonstrate the applicability of the method on real data.
|
['Sara Bals', 'Ja-Keoung Koo', 'J. Andreas Bærentzen', 'Anders B. Dahl', 'Vedrana A. Dahl', 'Qiongyang Chen']
|
2020-06-29
| null | null | null | null |
['electron-tomography']
|
['medical']
|
[ 5.60437083e-01 6.37657717e-02 5.84565461e-01 -2.47079462e-01
-5.90382755e-01 -4.82929274e-02 5.53228498e-01 -2.02937320e-01
-2.82436013e-01 6.65184975e-01 -1.01589836e-01 -4.39029843e-01
4.86517698e-02 -1.19227004e+00 -8.71067286e-01 -6.49353147e-01
2.68335491e-01 8.67978454e-01 2.61272281e-01 1.85168535e-02
3.61959040e-01 7.32914984e-01 -1.20863402e+00 2.64569700e-01
7.80827045e-01 8.68378878e-01 4.48138833e-01 3.01831871e-01
-5.55789828e-01 1.70975149e-01 -1.06136762e-01 -9.92902648e-03
1.31583512e-01 -5.38260937e-01 -7.73461640e-01 4.46868479e-01
3.39416526e-02 -3.15634459e-01 2.07182765e-01 9.89588439e-01
3.94588619e-01 -9.77235138e-02 9.26437616e-01 -7.55603254e-01
-4.34762180e-01 -1.05108432e-01 -6.46180987e-01 -2.65615642e-01
5.58863878e-01 -3.78278881e-01 2.74967790e-01 -1.38946557e+00
8.43578219e-01 1.39844131e+00 9.20170248e-01 4.48570341e-01
-1.57668543e+00 -3.98122132e-01 -9.45142135e-02 -1.68524951e-01
-1.08806324e+00 -1.43889487e-01 9.01522219e-01 -7.88286746e-01
4.73161489e-01 5.38982153e-01 9.74218369e-01 5.93057990e-01
4.50387359e-01 4.66498345e-01 1.82515049e+00 -4.46814358e-01
3.79732937e-01 1.29772097e-01 2.33368173e-01 5.26877463e-01
1.37773633e-01 1.03790209e-01 -5.29958196e-02 -6.73774421e-01
1.39464366e+00 1.31718084e-01 -6.06610060e-01 -3.10467720e-01
-1.02936554e+00 6.31254435e-01 9.27033946e-02 3.65260281e-02
-6.29195392e-01 5.45939878e-02 7.17263371e-02 2.66517073e-01
1.03558040e+00 4.58638780e-02 5.91912046e-02 1.71759933e-01
-8.36386383e-01 3.27948689e-01 9.68939662e-01 8.95541430e-01
6.60507381e-01 -9.79615524e-02 1.15854941e-01 6.26524091e-01
6.46102726e-01 5.15971661e-01 -1.14366658e-01 -8.91679645e-01
7.86070749e-02 3.68597209e-01 3.36164057e-01 -7.94142842e-01
-1.85505867e-01 -1.09461613e-01 -9.19685066e-01 5.62714100e-01
3.29475820e-01 3.23475331e-01 -7.85576582e-01 1.02620780e+00
8.36878479e-01 3.90918106e-01 -2.79039890e-01 9.60974395e-01
9.34159338e-01 8.27432692e-01 -3.17175686e-01 -7.14755893e-01
1.08042812e+00 -4.86411572e-01 -9.02859688e-01 7.20627785e-01
2.34776318e-01 -8.23000729e-01 9.79308903e-01 7.67149270e-01
-1.43299329e+00 -3.64740998e-01 -6.24118507e-01 -6.53516203e-02
1.53630286e-01 -1.06187031e-01 1.22264922e-01 4.36634421e-01
-9.44041073e-01 9.91435528e-01 -8.99925888e-01 -1.97190210e-01
4.22613412e-01 8.97143483e-02 9.38420184e-03 2.43077017e-02
-6.35058641e-01 8.32452416e-01 -1.72065571e-01 8.07141699e-03
-6.85813367e-01 -7.99619794e-01 -2.74080336e-01 -2.73159891e-01
5.83706759e-02 -1.00131214e+00 9.47454631e-01 -4.77834284e-01
-1.99386632e+00 9.09269631e-01 -4.37864959e-01 -9.40361843e-02
9.55574751e-01 6.66632280e-02 2.01296546e-02 2.24168688e-01
-1.11720659e-01 -2.10068151e-02 7.84035802e-01 -1.89153373e+00
9.54082832e-02 -3.99204701e-01 -1.94368646e-01 -2.14083400e-02
6.63604662e-02 -2.05347300e-01 -1.65611312e-01 -3.20513099e-01
7.66412199e-01 -6.80722535e-01 -5.28004885e-01 5.32125354e-01
-3.94703865e-01 -4.98154294e-03 9.77277994e-01 -6.75222635e-01
5.96357822e-01 -1.80481970e+00 1.45029798e-01 4.80507880e-01
3.30741853e-01 -6.69384226e-02 2.06253827e-01 6.78844213e-01
1.85787138e-02 1.08361363e-01 -7.51849115e-01 -6.17452860e-01
-2.31547862e-01 2.67625540e-01 -4.93734330e-01 8.53245378e-01
-6.26076683e-02 6.32700145e-01 -6.78078651e-01 -4.47424442e-01
2.00546280e-01 1.02663040e+00 -6.22231662e-01 1.48262471e-01
-2.44533747e-01 1.19269419e+00 -7.42364109e-01 3.41231585e-01
1.31605661e+00 -3.92349333e-01 1.56235665e-01 -1.78300723e-01
-6.01996481e-01 1.61680683e-01 -1.15424836e+00 1.41325796e+00
-5.87544322e-01 -7.09660649e-02 6.54403925e-01 -9.45036054e-01
1.07138336e+00 4.36602592e-01 6.00802779e-01 -4.45612192e-01
1.26711756e-01 2.92787522e-01 -2.82690525e-01 -4.39768255e-01
2.29016155e-01 -8.12599301e-01 4.95635659e-01 6.84845984e-01
-6.67098582e-01 -7.78521299e-01 -4.73124206e-01 -3.94549668e-02
7.23997772e-01 4.32933360e-01 1.98779285e-01 -7.65864849e-01
5.03204525e-01 2.01553345e-01 2.71899134e-01 3.09519887e-01
5.80928206e-01 8.85638177e-01 2.22125232e-01 -4.35916245e-01
-1.25663483e+00 -1.03814209e+00 -8.47382188e-01 1.37109637e-01
2.82219738e-01 -1.59555718e-01 -8.23474586e-01 -1.19645610e-01
-8.20811465e-03 7.05777287e-01 -5.10496259e-01 5.13888180e-01
-7.12018788e-01 -8.73645604e-01 -3.99112366e-02 1.42725945e-01
3.44868183e-01 -7.06518352e-01 -4.71396506e-01 5.11154354e-01
1.30343651e-02 -8.92697096e-01 -6.86418712e-02 -3.06153744e-01
-1.42275465e+00 -1.03889477e+00 -8.46911371e-01 -5.36940277e-01
1.01445711e+00 2.34428450e-01 1.13088095e+00 4.73406851e-01
1.12283736e-01 6.03697598e-01 -1.33867577e-01 -3.16138893e-01
-7.85222828e-01 -7.69769251e-01 -2.58105894e-04 1.70457870e-01
-5.19865692e-01 -1.10943305e+00 -4.55836922e-01 4.80314255e-01
-8.70645404e-01 6.04506791e-01 7.34366179e-02 8.24271500e-01
1.44441438e+00 -5.68616912e-02 4.57468688e-01 -1.26924467e+00
6.45437419e-01 -4.43474025e-01 -7.94416666e-01 -1.44001693e-01
-5.43381572e-01 -2.25807562e-01 7.05212295e-01 -3.67493421e-01
-1.19913304e+00 -1.65610850e-01 -6.07834935e-01 -4.09386903e-01
4.81158569e-02 3.71853828e-01 -1.52599126e-01 -4.40533340e-01
2.00714052e-01 4.85519081e-01 2.85901040e-01 -1.05099237e+00
-3.87793109e-02 4.26051527e-01 1.74984470e-01 -8.51621032e-01
6.71972096e-01 1.28567231e+00 5.57802260e-01 -1.12329137e+00
-2.91383475e-01 -2.06616104e-01 -4.94007677e-01 -2.74457932e-01
7.63640881e-01 -3.99671406e-01 -8.06180477e-01 2.56985039e-01
-1.51472163e+00 -3.26069742e-01 -5.82388759e-01 5.20266056e-01
-9.83915567e-01 5.79588890e-01 -6.81815982e-01 -1.04383802e+00
-3.63439143e-01 -1.43951452e+00 1.24972177e+00 -4.03667420e-01
-1.04449950e-02 -1.14221001e+00 1.96989402e-01 -8.00023135e-03
4.41916704e-01 4.92986739e-01 9.83556509e-01 4.13350090e-02
-8.08451712e-01 1.12468109e-01 -2.08905682e-01 8.55004862e-02
6.45821467e-02 -1.88902035e-01 -8.91718566e-01 -2.37066194e-01
9.48156714e-01 5.35571165e-02 6.80145502e-01 5.03765285e-01
1.50746417e+00 -9.42558348e-02 -5.44752538e-01 8.72366548e-01
1.66957605e+00 5.64118288e-02 9.80849206e-01 1.52047724e-02
6.95521414e-01 6.02449477e-01 4.77408320e-01 5.69972873e-01
1.41442001e-01 6.15504920e-01 6.15570009e-01 -2.14409396e-01
-2.42605343e-01 -1.51285261e-01 -1.24043673e-01 1.40207613e+00
-7.03531742e-01 -1.78343803e-02 -6.83533430e-01 1.55892745e-02
-1.42129278e+00 -6.12175941e-01 -1.12201488e+00 2.23434544e+00
6.67487860e-01 -2.70117760e-01 -1.34698093e-01 1.50102586e-01
5.89072406e-01 -4.05110866e-01 -3.40286285e-01 -1.78696841e-01
1.69129610e-01 3.67432147e-01 1.07759103e-01 8.00198197e-01
-5.11336982e-01 3.66845071e-01 6.62810087e+00 8.23985577e-01
-1.06656218e+00 4.55874354e-01 3.37371498e-01 2.92854279e-01
-9.68890429e-01 2.35059664e-01 -5.56961656e-01 3.57378632e-01
4.62040007e-01 -1.02995664e-01 2.32168376e-01 4.65871751e-01
6.78302586e-01 -1.08358711e-01 -1.13645947e+00 9.91877019e-01
-4.11768705e-01 -1.50748575e+00 2.31291309e-01 5.29343724e-01
7.85937071e-01 -3.67611080e-01 -6.54242560e-02 -3.79734099e-01
9.40716639e-02 -1.06805611e+00 6.79747224e-01 9.22668278e-01
8.89217675e-01 -4.31028932e-01 2.42512062e-01 8.40266109e-01
-1.01682317e+00 6.92920864e-01 -4.67993975e-01 -8.39166939e-02
7.75372267e-01 1.09088731e+00 -8.46319556e-01 7.70388246e-01
3.82658809e-01 5.91804624e-01 3.47912371e-01 1.09981179e+00
1.55854329e-01 5.83892822e-01 -4.23700750e-01 8.14052373e-02
-1.13054812e-01 -1.04144359e+00 8.93633485e-01 9.49608922e-01
6.49217129e-01 5.36585271e-01 2.58716553e-01 1.39984000e+00
1.30380079e-01 2.29980156e-01 -7.15836823e-01 5.11143804e-01
6.09443448e-02 1.02251995e+00 -7.98130691e-01 -4.48263556e-01
-3.00032705e-01 7.18058109e-01 -4.90412349e-03 2.93456286e-01
-7.50517786e-01 4.09432650e-01 1.43219590e-01 8.58504891e-01
8.58826786e-02 -4.59956735e-01 -4.97430712e-01 -8.73808444e-01
-5.10829277e-02 -4.13297594e-01 -3.11044604e-01 -7.36756563e-01
-1.46922982e+00 6.46885872e-01 1.99426889e-01 -1.45516551e+00
3.26360375e-01 -5.74172199e-01 -6.33099198e-01 1.05811322e+00
-1.56585419e+00 -7.59971976e-01 -1.05404936e-01 3.56318176e-01
3.67452294e-01 5.92788577e-01 8.49258780e-01 2.30221316e-01
1.41723111e-01 -2.99065709e-01 4.39595252e-01 -4.92738217e-01
7.06128627e-02 -1.10228872e+00 5.19911468e-01 2.43716836e-01
-4.14211124e-01 4.93358225e-01 6.74288332e-01 -1.03145134e+00
-1.78547728e+00 -1.00266671e+00 3.52647394e-01 -3.02522391e-01
1.84098676e-01 -2.49313235e-01 -1.36377692e+00 5.42468190e-01
-2.35227984e-03 1.44146889e-01 4.12011087e-01 -5.23313344e-01
4.15976971e-01 4.72115964e-01 -1.61560142e+00 2.23115698e-01
1.31253982e+00 -1.69882402e-01 -3.77215087e-01 5.67117453e-01
4.74745899e-01 -5.57522893e-01 -1.27917182e+00 3.25485706e-01
4.56368476e-01 -9.45852697e-01 1.09254634e+00 -1.30438834e-01
3.35919082e-01 -3.33106309e-01 -2.26950169e-01 -1.36797810e+00
-2.77689189e-01 -4.70379323e-01 1.94178060e-01 7.81534016e-01
2.31516082e-02 -1.01533198e+00 7.03623831e-01 3.61185908e-01
-4.07865167e-01 -9.82825518e-01 -1.06657064e+00 -7.96628594e-01
3.28180194e-01 -4.96531367e-01 6.05547309e-01 8.79652977e-01
-4.68312711e-01 2.43485495e-02 -1.46451250e-01 1.05547972e-01
1.16070783e+00 5.08294046e-01 6.89330816e-01 -1.48704398e+00
-4.58125025e-01 5.92730418e-02 7.66294077e-02 -1.50778008e+00
-3.50889489e-02 -1.04149199e+00 -3.48641723e-02 -1.80057621e+00
3.52925450e-01 -9.58219588e-01 5.65284073e-01 -4.38567072e-01
2.71801293e-01 1.50839150e-01 -5.54540120e-02 5.24943173e-01
9.52317119e-02 8.44758689e-01 2.23927426e+00 1.94103897e-01
-4.30603847e-02 1.25213981e-01 -1.34024560e-01 1.11002254e+00
5.04227221e-01 -6.33338273e-01 -3.24006110e-01 -4.71263230e-01
1.19317584e-01 4.72329408e-01 6.57757044e-01 -7.04835057e-01
6.65643578e-03 -1.66335911e-01 3.41912955e-01 -1.00422633e+00
8.51184964e-01 -1.15610158e+00 8.21228385e-01 4.82880950e-01
1.08518109e-01 -1.90028012e-01 -1.22469172e-01 7.64958501e-01
9.59771723e-02 -4.50037122e-01 8.60192120e-01 -4.94074374e-01
1.84829608e-01 4.89306778e-01 -2.44425356e-01 -1.23247661e-01
7.08447635e-01 -3.97808522e-01 1.80445164e-01 -2.88054824e-01
-1.02618754e+00 -3.14037651e-01 9.55630541e-01 -7.77883291e-01
1.13857687e+00 -1.47515368e+00 -8.46136749e-01 2.50968993e-01
-5.91091096e-01 4.28573310e-01 1.16235398e-01 1.11897576e+00
-6.88318551e-01 -1.25857532e-01 3.95631306e-02 -1.03326547e+00
-1.15318584e+00 3.93042624e-01 3.17427695e-01 -1.78459510e-01
-1.04427946e+00 4.25692469e-01 6.02610350e-01 -6.12293124e-01
-3.85785103e-01 -6.21569633e-01 -1.58179272e-03 -4.08216536e-01
4.69077468e-01 6.42182708e-01 1.52692854e-01 -6.17319047e-01
-9.75202248e-02 1.07158172e+00 3.81906241e-01 -3.62070531e-01
1.70638585e+00 -6.94333091e-02 -4.94414151e-01 5.65474033e-01
1.17692196e+00 1.03883691e-01 -1.16213524e+00 -1.57418817e-01
-5.48928678e-01 -6.55674696e-01 5.70350885e-02 -1.80550635e-01
-8.93844187e-01 1.06603634e+00 2.21145138e-01 4.39322889e-01
8.90464544e-01 8.68614540e-02 9.15694594e-01 -3.43190581e-02
8.31152558e-01 -5.98749936e-01 -1.71856076e-01 2.78464466e-01
1.36490631e+00 -7.05537021e-01 3.05785209e-01 -1.30989897e+00
5.47061823e-02 1.25933194e+00 1.20564200e-01 -4.43610847e-01
1.01660454e+00 5.14985740e-01 -4.86893713e-01 -5.67117870e-01
-5.17513037e-01 2.76761532e-01 -1.26360739e-02 5.39965868e-01
3.92125070e-01 1.68263376e-01 -6.22206092e-01 3.82187635e-01
1.57723222e-02 3.31743985e-01 6.94450140e-01 1.12084901e+00
-3.77982378e-01 -1.33335662e+00 -9.67919111e-01 5.02074897e-01
-3.18181694e-01 -6.44325763e-02 -8.97097960e-02 7.05009758e-01
-1.32028654e-01 5.24976015e-01 1.49111465e-01 1.37096001e-02
4.47860897e-01 -1.74017787e-01 1.00278533e+00 -5.57487488e-01
-2.39173427e-01 3.67301941e-01 1.71031028e-01 -3.30655366e-01
-6.33539796e-01 -7.43088365e-01 -1.43562102e+00 -3.37875456e-01
-5.09319961e-01 1.47013664e-01 8.95598352e-01 6.13373935e-01
8.71293396e-02 4.57295537e-01 6.68480217e-01 -1.34013736e+00
-6.23741090e-01 -7.57604599e-01 -8.67516518e-01 3.85362476e-01
2.10695788e-02 -9.49161589e-01 -3.06742519e-01 5.68808392e-02]
|
[9.402567863464355, -3.136657238006592]
|
9973f84d-5e0f-44d3-8cee-fdf201dd1679
|
querydet-cascaded-sparse-query-for
|
2103.09136
| null |
https://arxiv.org/abs/2103.09136v2
|
https://arxiv.org/pdf/2103.09136v2.pdf
|
QueryDet: Cascaded Sparse Query for Accelerating High-Resolution Small Object Detection
|
While general object detection with deep learning has achieved great success in the past few years, the performance and efficiency of detecting small objects are far from satisfactory. The most common and effective way to promote small object detection is to use high-resolution images or feature maps. However, both approaches induce costly computation since the computational cost grows squarely as the size of images and features increases. To get the best of two worlds, we propose QueryDet that uses a novel query mechanism to accelerate the inference speed of feature-pyramid based object detectors. The pipeline composes two steps: it first predicts the coarse locations of small objects on low-resolution features and then computes the accurate detection results using high-resolution features sparsely guided by those coarse positions. In this way, we can not only harvest the benefit of high-resolution feature maps but also avoid useless computation for the background area. On the popular COCO dataset, the proposed method improves the detection mAP by 1.0 and mAP-small by 2.0, and the high-resolution inference speed is improved to 3.0x on average. On VisDrone dataset, which contains more small objects, we create a new state-of-the-art while gaining a 2.3x high-resolution acceleration on average. Code is available at https://github.com/ChenhongyiYang/QueryDet-PyTorch.
|
['Naiyan Wang', 'Zehao Huang', 'Chenhongyi Yang']
|
2021-03-16
| null |
http://openaccess.thecvf.com//content/CVPR2022/html/Yang_QueryDet_Cascaded_Sparse_Query_for_Accelerating_High-Resolution_Small_Object_Detection_CVPR_2022_paper.html
|
http://openaccess.thecvf.com//content/CVPR2022/papers/Yang_QueryDet_Cascaded_Sparse_Query_for_Accelerating_High-Resolution_Small_Object_Detection_CVPR_2022_paper.pdf
|
cvpr-2022-1
|
['small-object-detection']
|
['computer-vision']
|
[-2.29102582e-01 -5.13175189e-01 -7.39657432e-02 -1.61438867e-01
-8.93705308e-01 -1.54256836e-01 4.84115392e-01 9.40657184e-02
-5.82209349e-01 2.85794258e-01 -1.88742414e-01 1.23540573e-01
1.18013337e-01 -1.21734869e+00 -7.34785914e-01 -7.02867150e-01
3.10533307e-02 2.63441503e-01 1.13537371e+00 -3.60513441e-02
1.96442649e-01 5.68855822e-01 -1.81948650e+00 3.03754479e-01
6.00719213e-01 1.33730495e+00 5.97718298e-01 5.88519633e-01
1.00766197e-01 6.03102505e-01 -2.65868157e-01 -1.39833748e-01
3.89027178e-01 9.47295949e-02 -2.99907774e-01 -9.75617021e-02
5.53925991e-01 -7.74131119e-01 -4.88905609e-01 1.19307220e+00
6.63248479e-01 -2.63157219e-01 2.53859669e-01 -9.81555760e-01
-5.34577847e-01 4.38845754e-01 -1.11426926e+00 6.50470078e-01
-8.78283381e-02 2.21241117e-01 8.89343441e-01 -1.28170264e+00
2.00718626e-01 1.29629135e+00 4.83799040e-01 2.58384105e-02
-9.86876905e-01 -1.03563392e+00 -5.04181720e-02 4.21546429e-01
-1.90103590e+00 -3.10956717e-01 3.06826651e-01 -3.85432065e-01
7.48267055e-01 1.84864372e-01 5.98755300e-01 4.59142476e-01
-7.05742044e-03 7.84983695e-01 7.65006185e-01 -1.44211322e-01
-9.05788466e-02 -6.56791218e-03 6.86195567e-02 8.99736881e-01
5.32907963e-01 1.81997672e-01 -5.85153103e-01 -6.64615631e-03
9.93529856e-01 5.30125201e-01 -1.15198925e-01 -1.72603264e-01
-1.25340724e+00 8.86308372e-01 9.41662133e-01 2.80177146e-01
-6.72331393e-01 1.01449966e-01 3.01117063e-01 -1.85732856e-01
3.59249949e-01 6.14022836e-02 -1.81696504e-01 6.49615526e-02
-9.86503243e-01 3.84220958e-01 2.64261842e-01 8.25744927e-01
9.58747804e-01 -1.94132179e-01 -3.37532252e-01 6.97774291e-01
2.09441528e-01 7.86554813e-01 1.78217366e-01 -8.03451538e-01
3.05714309e-01 7.91482687e-01 2.61642396e-01 -1.31288886e+00
-3.88207197e-01 -6.32949233e-01 -9.84608591e-01 1.95712596e-01
4.59518850e-01 9.57840011e-02 -8.00603986e-01 1.18242800e+00
5.84247291e-01 1.08583502e-01 -3.78202409e-01 1.14684021e+00
8.29518974e-01 9.78758037e-01 -7.85551444e-02 -5.08423410e-02
1.79967809e+00 -1.00665593e+00 -4.00679290e-01 -2.74542332e-01
3.76140177e-01 -8.60237956e-01 9.12738085e-01 3.24508816e-01
-1.02542675e+00 -8.77537072e-01 -1.03553665e+00 -1.47696957e-01
-2.26218119e-01 4.49758381e-01 6.84569120e-01 3.16224694e-01
-7.51505315e-01 3.19140822e-01 -1.02516842e+00 -1.94745392e-01
7.75166571e-01 2.46802822e-01 -3.34389098e-02 -3.83797020e-01
-8.70139718e-01 5.86727619e-01 5.79224467e-01 -6.96264580e-02
-9.11029518e-01 -7.82714605e-01 -5.67923307e-01 3.45112920e-01
7.50770688e-01 -5.20476103e-01 1.14778900e+00 -3.47287238e-01
-8.92750680e-01 6.07458532e-01 -1.82290182e-01 -4.83826220e-01
5.97149968e-01 -5.37119210e-01 -1.10492408e-01 2.22851127e-01
4.53783810e-01 6.93648636e-01 8.25626552e-01 -7.04319060e-01
-1.33974659e+00 -5.64917982e-01 1.07229725e-02 1.60423011e-01
-4.65067506e-01 2.10320160e-01 -9.96376812e-01 -3.39560300e-01
1.56673193e-01 -6.77170157e-01 -3.65438193e-01 3.38196218e-01
-2.46725231e-01 -3.80007774e-01 8.53348613e-01 -4.08918798e-01
1.17674685e+00 -2.31449032e+00 -4.18313235e-01 -1.33338302e-01
5.04869461e-01 4.43997234e-01 1.12372749e-01 -5.06322421e-02
4.20824379e-01 -1.86372757e-01 1.47475839e-01 -3.63477878e-02
-2.30655193e-01 -1.40195832e-01 -3.05333078e-01 5.47587037e-01
2.89381504e-01 9.28079009e-01 -8.42414141e-01 -6.74377322e-01
4.14519221e-01 7.29996681e-01 -6.23929322e-01 1.50555834e-01
9.08069685e-02 6.81544989e-02 -7.32114673e-01 7.42009223e-01
9.38136220e-01 -6.08932376e-01 -2.63454974e-01 -4.92276251e-01
-5.53355753e-01 4.89742756e-02 -1.47153819e+00 1.37821317e+00
-1.35443270e-01 4.76520985e-01 4.31287363e-02 -7.28387713e-01
8.24849188e-01 -1.88239992e-01 4.28268582e-01 -8.06419551e-01
1.17155418e-01 3.10963362e-01 -5.34970872e-02 -2.36588448e-01
5.48706174e-01 2.39591464e-01 1.67186633e-01 4.80949432e-02
-3.08571428e-01 2.47816667e-01 4.37545180e-01 1.74582496e-01
1.01059449e+00 -1.43382668e-01 4.82190579e-01 -8.02019313e-02
4.39483881e-01 1.19198591e-01 5.83489478e-01 8.71616125e-01
2.28609201e-02 4.47766304e-01 2.10190192e-01 -5.66969395e-01
-9.17278171e-01 -1.04018331e+00 -3.46308649e-01 1.27492523e+00
4.27540243e-01 -5.71826339e-01 -5.21751821e-01 -3.36003393e-01
1.51600599e-01 2.30932206e-01 -4.03107613e-01 4.08814065e-02
-6.32735193e-01 -1.01584756e+00 2.79892266e-01 7.09196031e-01
8.97205889e-01 -8.83504391e-01 -1.13737190e+00 1.40949473e-01
-1.46477073e-01 -1.21421385e+00 -2.93517888e-01 -5.34596629e-02
-7.64643848e-01 -9.54768181e-01 -8.20074916e-01 -7.21564531e-01
5.25921583e-01 7.81638980e-01 1.00190902e+00 2.52106786e-01
-7.37902880e-01 -2.35292792e-01 -3.37704748e-01 -5.15290141e-01
1.74517944e-01 7.07749352e-02 4.96132933e-02 -2.12120684e-03
4.46082920e-01 -3.41588855e-01 -9.25350547e-01 3.98155779e-01
-6.47048652e-01 1.45530567e-01 1.01466715e+00 6.71553552e-01
9.31575716e-01 2.10676000e-01 2.97970712e-01 -4.91937846e-01
-6.82072118e-02 -2.37793893e-01 -1.06057143e+00 -7.48119354e-02
-4.61675972e-01 -1.46708161e-01 5.88274658e-01 -4.10414785e-01
-9.05205905e-01 2.67131209e-01 -1.67876750e-01 -5.28140962e-01
-9.82316062e-02 8.65872651e-02 7.95099735e-02 -7.52100199e-02
7.97432601e-01 4.72214550e-01 -3.01955253e-01 -6.35248303e-01
3.21227431e-01 5.99962711e-01 5.65375566e-01 -1.96859062e-01
8.78344774e-01 8.12924981e-01 -7.00465143e-02 -8.80232334e-01
-1.03683341e+00 -6.69258654e-01 -5.08700192e-01 -4.58627790e-02
7.43279040e-01 -1.25549626e+00 -8.33114028e-01 4.50236410e-01
-1.00958598e+00 2.86789946e-02 -1.26545444e-01 4.57537860e-01
-4.35064919e-02 2.55685806e-01 -6.98389411e-01 -6.04371548e-01
-5.17831981e-01 -9.78248060e-01 1.16301167e+00 4.20690268e-01
5.38615108e-01 -7.01993778e-02 -3.74679178e-01 3.96681540e-02
5.87091565e-01 -3.88620384e-02 2.92080551e-01 -2.17680112e-01
-1.17867780e+00 -4.47880477e-01 -9.85482574e-01 2.60384083e-02
-1.17756560e-01 -1.91974521e-01 -8.57253551e-01 -4.04845268e-01
-1.21914029e-01 -2.85865635e-01 1.11841273e+00 5.47732413e-01
1.28953803e+00 -1.14275567e-01 -6.40124619e-01 6.52695954e-01
1.57250273e+00 -1.18294559e-01 4.37756628e-01 2.91714042e-01
7.08874583e-01 1.96420223e-01 1.10725760e+00 7.40292847e-01
3.22745055e-01 8.45698595e-01 5.33278286e-01 -2.62688398e-01
-2.46913567e-01 -1.50624380e-01 1.02164119e-01 4.64192212e-01
-2.20061108e-01 2.91467130e-01 -8.47883999e-01 6.11166179e-01
-1.86135554e+00 -1.00308561e+00 -1.73203439e-01 2.14810181e+00
7.38571882e-01 4.18628871e-01 2.56821066e-01 -1.08756289e-01
8.20134699e-01 1.97094902e-01 -6.16035223e-01 4.81874496e-01
7.74260089e-02 -7.56196827e-02 6.27891958e-01 2.30817765e-01
-1.29244125e+00 9.23351943e-01 4.96757174e+00 1.09546757e+00
-1.17863476e+00 2.87858903e-01 5.59244871e-01 -3.95888716e-01
5.37983954e-01 -2.16266140e-01 -1.50966215e+00 4.84415591e-01
5.49118519e-01 -2.38405362e-01 1.13714293e-01 1.30606019e+00
2.05440730e-01 -2.67898381e-01 -8.24262738e-01 1.26699972e+00
-8.35966095e-02 -1.59213161e+00 -1.36733288e-02 3.01525034e-02
4.66185331e-01 4.88910913e-01 -3.27798948e-02 3.80420387e-01
-2.64573321e-02 -6.82761908e-01 7.05636501e-01 1.91448435e-01
7.35564768e-01 -8.59774888e-01 7.73406148e-01 6.71254456e-01
-1.65864456e+00 -2.94451475e-01 -9.04680252e-01 -2.02903673e-01
1.08292721e-01 8.80981863e-01 -7.01176107e-01 2.65047729e-01
1.21219969e+00 6.43174529e-01 -7.89212525e-01 1.40123391e+00
-1.49947360e-01 4.81451571e-01 -6.52887106e-01 -1.29502341e-01
1.58057466e-01 3.44951116e-02 3.69934678e-01 1.36126769e+00
4.11135674e-01 3.21567059e-01 5.09266436e-01 7.77092338e-01
-6.37307316e-02 6.17319793e-02 -3.69941622e-01 4.20799494e-01
6.92639589e-01 1.62197948e+00 -8.33054066e-01 -6.46740496e-01
-5.31163812e-01 7.22268283e-01 3.95112783e-01 -3.95204611e-02
-1.07983601e+00 -4.57180291e-01 5.03709793e-01 5.07343411e-01
8.30716550e-01 -1.32962078e-01 -9.41383466e-02 -1.02319443e+00
1.77503258e-01 -6.01633251e-01 3.68763864e-01 -5.58071852e-01
-9.61083591e-01 5.50402224e-01 -1.39752388e-01 -1.17178094e+00
2.85150874e-02 -4.94022608e-01 -3.73982847e-01 7.45931447e-01
-1.58408391e+00 -1.09795177e+00 -7.33137906e-01 5.01102924e-01
6.12076581e-01 1.30013406e-01 4.74456131e-01 7.65936852e-01
-5.80252290e-01 4.45563883e-01 9.86021459e-02 3.06625843e-01
5.51491499e-01 -8.35687995e-01 4.38745350e-01 9.97950017e-01
2.29825884e-01 3.67901295e-01 4.65251803e-01 -4.44895059e-01
-1.30417597e+00 -1.31102848e+00 6.07336819e-01 -3.15246731e-01
4.91725922e-01 -3.37467194e-01 -9.43136275e-01 3.64006072e-01
-2.19934657e-01 5.24215221e-01 2.18397789e-02 -8.68665278e-02
-3.69892240e-01 -4.45713878e-01 -8.18307221e-01 4.13693726e-01
9.51289177e-01 -1.83794305e-01 -2.57534236e-01 4.20947701e-01
6.54475331e-01 -4.56449866e-01 -6.93555832e-01 4.86101896e-01
4.32406902e-01 -1.18860793e+00 1.25358450e+00 -1.28852408e-02
2.10351944e-01 -6.34443283e-01 -1.77615970e-01 -6.35786057e-01
-7.93990076e-01 -1.39320523e-01 -2.61280924e-01 9.27844703e-01
1.14696987e-01 -6.11152112e-01 8.04682791e-01 8.66155177e-02
7.74624199e-02 -9.04960692e-01 -7.66709268e-01 -6.83387458e-01
-4.39633578e-01 -3.87670964e-01 3.81897300e-01 5.15363038e-01
-4.95094001e-01 4.83870447e-01 -2.02792510e-01 5.02797008e-01
8.70491147e-01 5.95457911e-01 8.89395535e-01 -1.21791875e+00
-3.25460047e-01 -4.13164347e-01 -5.18581092e-01 -1.50083220e+00
-6.98149920e-01 -6.28465354e-01 2.35430468e-02 -1.46751845e+00
6.71773612e-01 -3.71993154e-01 -3.30418885e-01 5.86203635e-01
-3.54564667e-01 6.69638574e-01 2.87440538e-01 4.11872298e-01
-9.10259366e-01 3.78366470e-01 1.11494231e+00 1.13964349e-01
-1.49874603e-02 5.61275408e-02 -6.56859934e-01 9.61577415e-01
7.05647528e-01 -6.41416848e-01 7.82918632e-02 -4.08486396e-01
7.82187656e-02 -1.44072056e-01 6.88542008e-01 -1.31438112e+00
4.12163466e-01 1.06863327e-01 7.81665325e-01 -1.06815541e+00
4.67589915e-01 -5.36292076e-01 -2.02688858e-01 8.19459081e-01
5.20265587e-02 -1.72563478e-01 2.12493300e-01 6.00073636e-01
-1.65208399e-01 1.01333283e-01 1.15822971e+00 -1.39234677e-01
-1.04949367e+00 5.64610720e-01 -5.24688400e-02 -1.03514619e-01
1.09018016e+00 -9.74659845e-02 -3.77075166e-01 3.43312323e-02
-2.60241836e-01 3.18544507e-01 2.89756328e-01 4.61518943e-01
6.62246346e-01 -1.31929386e+00 -9.87267077e-01 2.28215754e-01
7.55780861e-02 3.65305483e-01 3.39870989e-01 1.05039799e+00
-5.34931898e-01 5.08731782e-01 -1.90787375e-01 -9.80521679e-01
-1.38760173e+00 6.44483030e-01 1.83574468e-01 -2.06131816e-01
-9.52159762e-01 1.06878507e+00 6.42207444e-01 1.30343512e-01
2.17842311e-01 -1.61193654e-01 -1.83992639e-01 4.11993964e-03
1.13357985e+00 4.98039484e-01 -1.36292443e-01 -5.04946411e-01
-5.02353072e-01 7.15812504e-01 -4.04373020e-01 3.62893581e-01
1.34140241e+00 -3.76070943e-03 6.34992076e-03 1.85001772e-02
9.80213940e-01 -8.09362084e-02 -1.46705747e+00 -5.80544651e-01
-2.27843806e-01 -9.95523632e-01 3.06256622e-01 -3.50361437e-01
-1.20425832e+00 8.94008994e-01 9.35117245e-01 2.81512588e-01
1.20401812e+00 4.12070453e-01 8.24186206e-01 3.99852782e-01
3.91292006e-01 -7.40550458e-01 2.67488122e-01 4.50458169e-01
7.50002265e-01 -1.41389573e+00 2.94255763e-01 -6.00902736e-01
-3.59196097e-01 9.00709093e-01 8.10053885e-01 -2.65545636e-01
5.21825373e-01 3.92355502e-01 -1.67588249e-01 -3.09342295e-01
-5.68621576e-01 -4.49459314e-01 2.26238862e-01 3.04453611e-01
1.60853669e-01 5.11431992e-02 -1.51509345e-01 5.64772546e-01
-1.05348406e-02 -3.16664316e-02 1.26817048e-01 6.44534588e-01
-1.05083001e+00 -5.36066949e-01 -6.29774630e-01 7.49717891e-01
-5.43061137e-01 -1.42169759e-01 1.45991772e-01 7.17670619e-01
2.30844960e-01 7.69718647e-01 2.72768855e-01 -2.88491607e-01
4.32655185e-01 -5.94657183e-01 9.24007371e-02 -6.17403865e-01
-1.11466952e-01 3.02654147e-01 -2.04045504e-01 -9.32798803e-01
-1.00450762e-01 -5.18361807e-01 -1.34606600e+00 -3.67101669e-01
-5.74848294e-01 -1.48600280e-01 4.44146216e-01 5.62730134e-01
5.11904061e-01 4.94249433e-01 4.95077938e-01 -1.14324927e+00
-7.26095855e-01 -9.30615962e-01 -5.48310339e-01 8.06855485e-02
2.13327333e-01 -6.93646729e-01 -1.22209229e-01 -4.73562926e-02]
|
[8.743677139282227, -0.5059682130813599]
|
d87570c8-7126-467b-9be9-8a40aa711b47
|
cotype-joint-extraction-of-typed-entities-and
|
1610.08763
| null |
http://arxiv.org/abs/1610.08763v2
|
http://arxiv.org/pdf/1610.08763v2.pdf
|
CoType: Joint Extraction of Typed Entities and Relations with Knowledge Bases
|
Extracting entities and relations for types of interest from text is
important for understanding massive text corpora. Traditionally, systems of
entity relation extraction have relied on human-annotated corpora for training
and adopted an incremental pipeline. Such systems require additional human
expertise to be ported to a new domain, and are vulnerable to errors cascading
down the pipeline. In this paper, we investigate joint extraction of typed
entities and relations with labeled data heuristically obtained from knowledge
bases (i.e., distant supervision). As our algorithm for type labeling via
distant supervision is context-agnostic, noisy training data poses unique
challenges for the task. We propose a novel domain-independent framework,
called CoType, that runs a data-driven text segmentation algorithm to extract
entity mentions, and jointly embeds entity mentions, relation mentions, text
features and type labels into two low-dimensional spaces (for entity and
relation mentions respectively), where, in each space, objects whose types are
close will also have similar representations. CoType, then using these learned
embeddings, estimates the types of test (unlinkable) mentions. We formulate a
joint optimization problem to learn embeddings from text corpora and knowledge
bases, adopting a novel partial-label loss function for noisy labeled data and
introducing an object "translation" function to capture the cross-constraints
of entities and relations on each other. Experiments on three public datasets
demonstrate the effectiveness of CoType across different domains (e.g., news,
biomedical), with an average of 25% improvement in F1 score compared to the
next best method.
|
['Xiang Ren', 'Zeqiu Wu', 'Meng Qu', 'Heng Ji', 'Jiawei Han', 'Clare R. Voss', 'Wenqi He', 'Tarek F. Abdelzaher']
|
2016-10-27
| null | null | null | null |
['joint-entity-and-relation-extraction']
|
['natural-language-processing']
|
[-6.41227588e-02 5.83794057e-01 -4.85802621e-01 -5.80912709e-01
-8.18375766e-01 -8.90206993e-01 4.54334944e-01 7.20285594e-01
-6.37879014e-01 8.73945534e-01 1.35087714e-01 -2.58567184e-01
-1.03905633e-01 -9.34759676e-01 -8.52062225e-01 -3.96091640e-01
6.37274161e-02 1.01794302e+00 1.45477816e-01 4.05002758e-02
-2.23845497e-01 2.36389801e-01 -1.10678887e+00 1.69296563e-01
1.10747278e+00 9.47773218e-01 -1.10384203e-01 4.19199914e-01
-4.63478863e-01 4.71688718e-01 -6.33887649e-01 -9.70699072e-01
6.20705597e-02 1.42598316e-01 -1.45315492e+00 -5.79104498e-02
3.58171090e-02 1.78589653e-02 -1.94516242e-01 8.71487141e-01
3.45496565e-01 6.46157563e-02 1.04118419e+00 -1.23628485e+00
-9.20154512e-01 8.38539958e-01 -3.68145585e-01 4.77351248e-02
3.99701089e-01 -2.63914466e-01 1.59021854e+00 -1.03044987e+00
9.95379746e-01 1.11146486e+00 8.28785360e-01 6.15301013e-01
-1.58558643e+00 -3.81920427e-01 2.80660003e-01 1.37372062e-01
-1.36284053e+00 -1.47773594e-01 4.17461842e-01 -5.44464052e-01
1.30025804e+00 2.28900835e-01 2.47972086e-01 1.13154209e+00
-5.02280235e-01 9.68558252e-01 7.16546714e-01 -4.63599205e-01
2.72744030e-01 6.71891272e-01 6.32116020e-01 5.27049720e-01
6.70278430e-01 -3.94718766e-01 -3.58981371e-01 -3.26766163e-01
1.46125913e-01 -3.21510732e-01 -2.55199969e-01 -2.19082475e-01
-8.89720976e-01 7.38205075e-01 3.30777615e-01 2.95721710e-01
-2.79085159e-01 -4.15953726e-01 3.20030630e-01 5.38649932e-02
8.46899807e-01 6.21104777e-01 -1.32787776e+00 1.43699139e-01
-5.66579282e-01 1.94081917e-01 1.26876664e+00 1.31058133e+00
8.62543643e-01 -7.87172019e-01 -1.06668845e-01 1.15662253e+00
3.95514160e-01 1.94705889e-01 4.58561629e-01 -1.80878222e-01
8.35739255e-01 1.19278574e+00 3.68366390e-02 -6.50932431e-01
-6.30517066e-01 -1.08431868e-01 -4.37340945e-01 -4.52073187e-01
5.34303367e-01 -6.06271088e-01 -9.10065472e-01 1.62094283e+00
1.01585293e+00 1.70258865e-01 4.46346164e-01 7.04789460e-01
1.11874282e+00 4.89281029e-01 2.89567739e-01 5.86091056e-02
1.65065396e+00 -7.35118628e-01 -6.82407320e-01 -3.73306841e-01
1.01757002e+00 -5.04932225e-01 7.69702554e-01 -2.70639695e-02
-5.78270078e-01 -8.65972862e-02 -8.29823315e-01 -4.81048018e-01
-9.45155859e-01 2.13875055e-01 5.07512987e-01 4.55100477e-01
-3.52518320e-01 6.14467144e-01 -8.19550395e-01 -3.54958683e-01
4.91491437e-01 5.05608916e-01 -5.15419841e-01 7.40230363e-03
-1.51752841e+00 9.86367643e-01 7.27474391e-01 8.73640031e-02
-2.87875205e-01 -8.96769702e-01 -1.11370718e+00 2.09605657e-02
6.30686343e-01 -7.15171456e-01 9.79673088e-01 -5.77800274e-01
-1.02956390e+00 1.09646201e+00 -2.46135473e-01 -4.70223963e-01
1.56754166e-01 -5.27279913e-01 -5.11815131e-01 -2.02612147e-01
3.28088105e-01 3.63417268e-01 4.65531439e-01 -1.27090037e+00
-7.85907924e-01 -4.35446829e-01 1.00925110e-01 1.62465125e-01
-5.21485031e-01 6.92672580e-02 -5.18598557e-01 -4.47475165e-01
1.23267218e-01 -8.34377825e-01 -7.31788501e-02 -2.97862947e-01
-8.58750999e-01 -9.34877396e-01 6.94103718e-01 -7.40938544e-01
1.32911980e+00 -1.90998828e+00 2.86039859e-01 1.72915593e-01
4.52106833e-01 3.41028333e-01 1.22450508e-01 1.83818609e-01
-5.25467843e-02 5.22377551e-01 -3.63079637e-01 -3.29647750e-01
2.58597881e-01 5.17819822e-01 -1.03649497e-01 1.70867279e-01
7.92223692e-01 9.41281855e-01 -1.04550374e+00 -6.46003902e-01
-4.00969267e-01 2.77386993e-01 -4.46771979e-01 4.13222402e-01
-5.11851668e-01 -6.96118688e-03 -6.37506545e-01 6.93763316e-01
6.65602386e-01 -4.05423820e-01 5.85297406e-01 -3.14104646e-01
1.66025013e-01 6.92059398e-01 -1.25969303e+00 1.26043439e+00
-4.79262739e-01 3.39471728e-01 -5.91775738e-02 -1.00849247e+00
8.18207443e-01 4.14556533e-01 3.48733127e-01 -1.23641053e-02
1.66112363e-01 3.30294102e-01 -2.92494833e-01 -8.79349709e-01
4.19748068e-01 -2.25476250e-02 -3.43194485e-01 2.16153875e-01
5.61903059e-01 1.11758798e-01 3.53306800e-01 3.98097694e-01
1.31213665e+00 1.40351787e-01 3.64667296e-01 1.12297803e-01
2.65213579e-01 2.95790613e-01 1.00969505e+00 3.93990040e-01
8.56765825e-03 3.38976413e-01 8.21413994e-01 -2.21655041e-01
-8.99061263e-01 -9.34264183e-01 -4.35058415e-01 9.99686599e-01
1.34846106e-01 -4.47229832e-01 -6.41167939e-01 -1.56024337e+00
3.69157672e-01 6.76595151e-01 -6.91003323e-01 2.56442241e-02
-6.88692510e-01 -1.08550656e+00 5.80151856e-01 6.37529969e-01
6.93215430e-02 -9.40786481e-01 1.64105922e-01 3.66760194e-01
-3.04077417e-01 -1.55856431e+00 -2.32785836e-01 7.13953972e-01
-4.79693145e-01 -1.24081945e+00 -4.85712290e-01 -1.05552602e+00
7.80793846e-01 -4.01130289e-01 1.29873502e+00 -7.52850622e-02
-1.53604999e-01 1.84911788e-01 -5.40343881e-01 -3.10079038e-01
-1.31143719e-01 4.54807013e-01 9.87579301e-02 2.16733832e-02
8.73274922e-01 -1.88720152e-01 -1.53307289e-01 1.54400662e-01
-7.72526979e-01 -3.16242725e-01 5.43977857e-01 1.08334970e+00
6.47132277e-01 6.67121867e-03 6.60246372e-01 -1.60884333e+00
4.41454142e-01 -9.60516036e-01 -4.79239315e-01 4.05398190e-01
-6.57901108e-01 2.28758201e-01 6.29394174e-01 -5.00995278e-01
-1.29252326e+00 1.65810317e-01 -6.96137622e-02 7.42447972e-02
-2.92583138e-01 6.73533916e-01 -6.47822559e-01 4.53355700e-01
7.06540883e-01 -2.58056074e-01 -6.50409460e-01 -7.41951346e-01
6.49672091e-01 1.02047515e+00 3.67237568e-01 -8.73635471e-01
8.55289280e-01 1.43899590e-01 -4.90797758e-01 -6.39118850e-01
-1.34775817e+00 -8.05030286e-01 -1.02603042e+00 4.63579625e-01
9.38510358e-01 -8.31598997e-01 -3.41541469e-01 1.43719971e-01
-1.29208350e+00 -1.07734114e-01 -3.91669422e-01 4.93645519e-01
-2.00954899e-02 2.53888458e-01 -6.56175613e-01 -5.53972900e-01
-1.99814707e-01 -7.63220429e-01 1.06772208e+00 3.32702339e-01
-3.93178403e-01 -1.28179955e+00 1.89705089e-01 4.89874691e-01
-3.73049676e-01 1.59346327e-01 1.20529974e+00 -1.55181158e+00
-4.24726278e-01 -3.14505845e-01 -4.46703255e-01 3.46634746e-01
1.28896788e-01 -7.81039223e-02 -1.04472399e+00 1.93202078e-01
-3.53987247e-01 -4.74805057e-01 7.84016967e-01 -5.94873130e-02
7.16929078e-01 -6.44508898e-01 -8.77818048e-01 4.96432334e-01
1.29548001e+00 -1.62287369e-01 2.14377925e-01 4.41335142e-01
9.23827767e-01 7.32580841e-01 7.12441623e-01 2.64713317e-01
6.63549244e-01 5.10878861e-01 1.68032087e-02 -3.60861234e-02
1.06611349e-01 -2.75210708e-01 1.25180423e-01 6.36860669e-01
2.39670187e-01 -5.54646909e-01 -9.63334978e-01 9.39263880e-01
-1.72395277e+00 -4.96976078e-01 -3.93958092e-01 1.83627737e+00
1.61186981e+00 1.38785362e-01 3.81706245e-02 4.22085449e-02
8.03327382e-01 -3.42339069e-01 -6.29243970e-01 -2.91818142e-01
-8.85204375e-02 3.70476633e-01 6.67344868e-01 2.43784681e-01
-1.43490624e+00 1.18696976e+00 4.83638430e+00 7.55678713e-01
-5.22364378e-01 1.62729993e-01 4.79362160e-01 2.86153197e-01
-3.68071884e-01 2.06560478e-01 -1.37999153e+00 4.01451975e-01
8.96608174e-01 -1.79332256e-01 5.64670041e-02 8.78350616e-01
-3.11813176e-01 1.27511337e-01 -1.40463769e+00 3.71974528e-01
-2.36523226e-01 -1.05592620e+00 -3.16800207e-01 -1.95791554e-02
7.65377283e-01 7.84838572e-02 -2.10242316e-01 6.16858721e-01
8.88191938e-01 -7.84723938e-01 3.76582831e-01 9.27712694e-02
7.54889607e-01 -4.98879284e-01 1.03684711e+00 3.12561572e-01
-1.11499512e+00 8.71648937e-02 -3.46461028e-01 3.64769548e-01
7.99347609e-02 1.09343266e+00 -1.25487876e+00 7.65956879e-01
6.25207722e-01 7.03334630e-01 -3.16749245e-01 7.31756270e-01
-6.32735074e-01 7.16630936e-01 -4.97496307e-01 -2.16679439e-01
-2.45667379e-02 5.37275895e-02 4.46714908e-01 1.51304662e+00
-1.22026518e-01 2.39504546e-01 2.32918769e-01 9.59384084e-01
-6.52367353e-01 2.14435577e-01 -4.54075754e-01 -1.93435624e-01
7.26162851e-01 1.62705827e+00 -5.61857879e-01 -6.25079870e-01
-6.28913403e-01 9.30260599e-01 9.25252974e-01 3.98522228e-01
-7.26368427e-01 -6.90665781e-01 8.13991725e-01 -2.57024735e-01
5.13816953e-01 6.58614337e-02 -4.95858490e-01 -1.30044043e+00
2.42971212e-01 -5.67521274e-01 7.89811850e-01 -1.00370057e-01
-1.73948550e+00 5.25600135e-01 -1.67940006e-01 -8.80267859e-01
-9.94816869e-02 -7.66958833e-01 -1.85884759e-01 8.47288430e-01
-1.61738288e+00 -1.07922852e+00 1.17366128e-01 2.15249792e-01
2.63051689e-01 1.09299183e-01 9.51205850e-01 4.49452519e-01
-9.97600138e-01 8.73353601e-01 9.70797837e-02 7.94937909e-01
7.50358164e-01 -1.74388421e+00 4.86315757e-01 4.95999038e-01
3.10567468e-01 6.84740245e-01 4.00472760e-01 -8.91723037e-01
-1.16170859e+00 -1.40527034e+00 1.45587754e+00 -8.45486224e-01
8.79272103e-01 -6.33890212e-01 -1.18727601e+00 1.01307511e+00
-9.79260951e-02 3.34692210e-01 9.69305336e-01 7.73192048e-01
-7.31661737e-01 2.87779033e-01 -1.44021797e+00 2.75875032e-01
1.05963552e+00 -4.13623601e-01 -9.73374963e-01 5.72109222e-01
9.09327388e-01 -4.42597002e-01 -1.15404642e+00 3.39109510e-01
2.77844131e-01 -2.20694005e-01 8.58290374e-01 -1.05877292e+00
3.01951110e-01 -2.26374567e-01 -4.35018465e-02 -1.41614616e+00
-1.11127540e-01 -3.61468196e-01 -4.55445766e-01 1.84179688e+00
1.10820019e+00 -6.17178679e-01 7.80558050e-01 9.56378222e-01
1.31148733e-02 -9.21472490e-01 -7.86400318e-01 -8.81676555e-01
1.56381339e-01 -3.09113741e-01 5.30935228e-01 1.24119544e+00
4.30869877e-01 8.45977426e-01 1.76259473e-01 6.65698349e-01
2.61736691e-01 1.17225848e-01 5.82288921e-01 -1.43485975e+00
-1.93011865e-01 -8.21540952e-02 -3.54454428e-01 -1.04538941e+00
5.46637595e-01 -1.26720011e+00 1.91008791e-01 -1.61983156e+00
1.34735882e-01 -8.93316090e-01 -1.62755564e-01 8.35942507e-01
-5.83964050e-01 -1.36970177e-01 -3.45186979e-01 -1.99073479e-02
-5.93447328e-01 4.12996173e-01 7.34811962e-01 -3.03697318e-01
-2.52067387e-01 -1.11810289e-01 -8.19806576e-01 9.69493747e-01
5.35102069e-01 -8.73182774e-01 8.12541321e-02 -4.62742269e-01
3.33726496e-01 -3.72580856e-01 1.17949508e-02 -4.05224890e-01
2.14478597e-01 9.83160222e-04 2.90819198e-01 -2.56573707e-01
1.99376181e-01 -8.46938729e-01 -3.44759881e-01 -1.59773514e-01
-4.91387129e-01 -4.22915369e-01 1.22535536e-02 7.39587963e-01
-1.57754689e-01 -5.47939301e-01 4.24209684e-01 1.60356402e-01
-5.70123851e-01 2.98801839e-01 2.27647368e-02 7.67266810e-01
9.27438200e-01 1.91294909e-01 -3.20291191e-01 3.99900466e-01
-1.05198371e+00 4.77686375e-01 1.73594221e-01 2.78446853e-01
2.26915032e-01 -1.22499180e+00 -7.00296462e-01 1.89980324e-02
3.39780271e-01 4.89298254e-01 -1.35820746e-01 5.61216295e-01
-4.49745692e-02 1.09963447e-01 6.04571760e-01 -3.01781625e-01
-1.27504742e+00 4.67124939e-01 5.74932173e-02 -7.64087141e-01
-4.08249706e-01 1.31558466e+00 -7.18106776e-02 -1.04714358e+00
1.43898934e-01 -4.56367671e-01 -3.74333590e-01 4.59806144e-01
1.92298070e-01 1.90598965e-01 3.53613645e-01 -4.30128187e-01
-5.57145238e-01 2.43748486e-01 -5.10752201e-01 1.44331276e-01
1.62517011e+00 3.94233651e-02 -9.78750661e-02 3.00576091e-01
1.29296243e+00 1.27029970e-01 -9.49604809e-01 -6.84580505e-01
7.34525263e-01 -1.54084072e-01 -1.74940556e-01 -9.63166893e-01
-8.83064330e-01 5.22963405e-01 -3.20655853e-02 4.36045468e-01
6.31688297e-01 5.35059094e-01 1.01325285e+00 4.08413738e-01
-4.14083898e-03 -1.24221182e+00 -3.26954931e-01 7.25731790e-01
3.53789508e-01 -1.31263447e+00 -9.19745862e-02 -7.84579158e-01
-5.82730412e-01 9.16635931e-01 6.73130333e-01 1.68330774e-01
7.72943079e-01 4.81489748e-01 -7.30962977e-02 -2.26208955e-01
-8.67007196e-01 -5.74524701e-01 4.83894348e-01 7.21337914e-01
6.04447365e-01 1.66634932e-01 -2.76368469e-01 1.11657655e+00
-1.97310410e-02 -2.21578583e-01 1.73921689e-01 8.54408681e-01
-2.71418303e-01 -1.26074862e+00 -3.24648544e-02 7.15678692e-01
-5.99615753e-01 -2.79484749e-01 -5.97015798e-01 6.08024955e-01
5.01113355e-01 9.99705970e-01 -1.23553753e-01 -2.61087358e-01
4.76657033e-01 3.43129903e-01 1.51226133e-01 -1.11558676e+00
-5.95311582e-01 -3.59343320e-01 6.38464928e-01 -2.93666753e-03
-1.99795693e-01 -7.45670080e-01 -1.57331800e+00 8.49173293e-02
-7.72197306e-01 3.65855992e-01 5.75243413e-01 1.24301565e+00
4.73700672e-01 5.63757479e-01 4.93913144e-01 -1.87122479e-01
-5.03371239e-01 -9.97036338e-01 -4.74470168e-01 6.95713937e-01
1.12604536e-01 -7.91796803e-01 -3.96539271e-01 8.94500688e-02]
|
[9.456271171569824, 8.765880584716797]
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.