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
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
de172a39-7111-4c7f-ab2d-0fcc8416f0c5
|
weakly-supervised-data-augmentation-through
|
2210.14169
| null |
https://arxiv.org/abs/2210.14169v3
|
https://arxiv.org/pdf/2210.14169v3.pdf
|
Weakly Supervised Data Augmentation Through Prompting for Dialogue Understanding
|
Dialogue understanding tasks often necessitate abundant annotated data to achieve good performance and that presents challenges in low-resource settings. To alleviate this barrier, we explore few-shot data augmentation for dialogue understanding by prompting large pre-trained language models and present a novel approach that iterates on augmentation quality by applying weakly-supervised filters. We evaluate our methods on the emotion and act classification tasks in DailyDialog and the intent classification task in Facebook Multilingual Task-Oriented Dialogue. Models fine-tuned on our augmented data mixed with few-shot ground truth data are able to approach or surpass existing state-of-the-art performance on both datasets. For DailyDialog specifically, using 10% of the ground truth data we outperform the current state-of-the-art model which uses 100% of the data.
|
['Dilek Hakkani-Tur', 'Zhou Yu', 'Yang Liu', 'Seokhwan Kim', 'Andy Rosenbaum', 'Chenyang Tao', 'Alexandros Papangelis', 'Maximillian Chen']
|
2022-10-25
| null | null | null | null |
['dialogue-understanding', 'intent-classification']
|
['natural-language-processing', 'natural-language-processing']
|
[ 3.74474339e-02 9.67059493e-01 -1.89269349e-01 -7.18342602e-01
-9.25861776e-01 -2.85177946e-01 1.16408229e+00 2.46166140e-01
-8.80182922e-01 1.04412818e+00 1.02533054e+00 -6.47668466e-02
5.22365272e-01 -3.14372420e-01 -9.51137692e-02 1.85871631e-01
1.30540669e-01 9.03658032e-01 -3.21279824e-01 -8.64760458e-01
-1.27315968e-01 -3.83361578e-01 -1.07957387e+00 5.83913147e-01
1.11005318e+00 6.71608269e-01 -2.26859748e-01 7.90803015e-01
-4.09045905e-01 1.13856328e+00 -4.55023706e-01 -5.51514983e-01
-7.83807710e-02 -6.88020825e-01 -1.38831675e+00 2.51937032e-01
3.54105234e-01 -7.73322940e-01 -1.59444898e-01 4.60993558e-01
6.71224594e-01 3.76094967e-01 5.23436487e-01 -9.02862370e-01
-6.28497541e-01 7.26281106e-01 -1.14385642e-01 2.80236393e-01
6.81506574e-01 2.66595572e-01 1.10202610e+00 -9.84987557e-01
8.35914910e-01 1.29453421e+00 6.79892361e-01 1.15017772e+00
-1.56084359e+00 -2.30259314e-01 2.89309293e-01 -7.21439421e-02
-6.16142929e-01 -1.17983592e+00 4.35121477e-01 -3.39880377e-01
1.49967194e+00 4.47006933e-02 4.91791546e-01 1.69719720e+00
-4.67576861e-01 9.76775944e-01 1.27854145e+00 -6.13702118e-01
2.11635500e-01 3.64455998e-01 6.30921125e-01 7.58196533e-01
-6.20671809e-01 -3.82711112e-01 -8.19308281e-01 -3.74683022e-01
1.85409188e-01 -4.78899121e-01 -4.58184123e-01 3.16471606e-02
-1.06694019e+00 1.20328343e+00 3.20783481e-02 2.95693576e-01
-3.07451487e-01 -3.87076885e-01 7.73189783e-01 4.72367674e-01
1.22615588e+00 8.12195718e-01 -7.37543106e-01 -9.73408222e-01
-7.56624222e-01 1.99850366e-01 1.40834892e+00 5.38046002e-01
5.25317430e-01 -1.45344332e-01 -4.84651327e-01 1.39140296e+00
9.34660658e-02 -2.02636346e-01 5.99539161e-01 -9.85015869e-01
7.39425302e-01 7.50065148e-01 1.42869055e-01 -1.88906938e-01
-7.34637737e-01 4.13519423e-03 -6.14733875e-01 -1.47576004e-01
6.67716205e-01 -7.92658269e-01 -7.84625232e-01 1.88809311e+00
4.29258317e-01 1.39831498e-01 4.65422988e-01 7.23627210e-01
1.13900721e+00 5.57954729e-01 2.57325888e-01 -3.43277305e-01
1.42846119e+00 -1.29917216e+00 -1.26453829e+00 -6.57318473e-01
1.26111114e+00 -1.51733890e-01 1.57305062e+00 1.55176789e-01
-8.88188481e-01 -3.00443113e-01 -1.04904294e+00 -3.00531954e-01
-3.52834493e-01 -6.77798837e-02 7.95181811e-01 7.28269398e-01
-9.61349308e-01 5.00310481e-01 -6.56670153e-01 -8.19334388e-01
3.38850737e-01 -9.39033106e-02 -5.83099663e-01 -2.50155516e-02
-1.43062997e+00 1.29246020e+00 2.10995495e-01 -2.29930490e-01
-8.78704190e-01 -9.82793391e-01 -1.21204674e+00 -1.39001369e-01
3.55153590e-01 -4.78054613e-01 1.69972301e+00 -7.00779438e-01
-2.06555796e+00 1.30124986e+00 5.63698560e-02 -8.31874728e-01
6.92760050e-01 -6.02236986e-01 -1.62811294e-01 -5.15220948e-02
-9.70800233e-04 8.15506160e-01 3.01048219e-01 -8.44236612e-01
-3.78916413e-01 -2.40806580e-01 4.22223985e-01 5.10361552e-01
-7.95584798e-01 -5.33840898e-03 -9.43485424e-02 -8.05418640e-02
-4.98658508e-01 -8.83473575e-01 -3.88407230e-01 -2.38369867e-01
-2.78735936e-01 -3.64163697e-01 7.33811975e-01 -7.44357944e-01
1.05533624e+00 -1.74649084e+00 1.58694804e-01 -6.18883550e-01
2.71787852e-01 3.29129159e-01 -2.01509297e-01 5.16061008e-01
9.83743295e-02 1.67593613e-01 -2.20594913e-01 -1.10814929e+00
1.09405078e-01 3.75404388e-01 -9.22045708e-02 2.95549005e-01
4.85246360e-01 7.72379100e-01 -1.07738996e+00 -2.63149500e-01
3.51209998e-01 2.24980086e-01 -7.70251393e-01 6.92488313e-01
-5.02450109e-01 7.01870441e-01 -2.16404304e-01 1.35910600e-01
2.47967109e-01 -3.50109100e-01 2.63440162e-01 -5.52996211e-02
7.34724924e-02 7.22953856e-01 -6.75574601e-01 2.23080254e+00
-8.36812794e-01 6.08860493e-01 3.39494675e-01 -7.60256290e-01
7.37676442e-01 5.76685727e-01 2.42820397e-01 -5.83676994e-01
8.52771997e-02 -3.34067680e-02 -1.88668650e-02 -6.78548634e-01
6.99774325e-01 -3.54229450e-01 -3.61628979e-01 7.98161328e-01
6.72682822e-01 -1.66826323e-01 6.38293475e-02 7.13575542e-01
1.14046109e+00 1.19321629e-01 3.64373446e-01 -2.14706600e-01
4.27722424e-01 -7.03192353e-02 2.32178584e-01 8.06894958e-01
-3.57267261e-01 3.22765112e-01 5.70276380e-01 -3.59561831e-01
-1.01699162e+00 -3.68071496e-01 -9.39267576e-02 1.76532900e+00
-5.39010108e-01 -6.74998462e-01 -8.89546514e-01 -1.04858780e+00
-3.91465992e-01 1.05253220e+00 -9.74578857e-01 4.34591845e-02
-5.13054617e-02 -1.02456057e+00 7.22992063e-01 3.77131581e-01
6.95069373e-01 -9.36573267e-01 -2.84912676e-01 4.54672486e-01
-6.20819807e-01 -1.63763404e+00 -1.00380987e-01 1.47472665e-01
-4.59152222e-01 -7.34436214e-01 -5.88708639e-01 -1.41473889e-01
2.00488955e-01 -2.13555619e-01 1.60645449e+00 6.08834811e-02
-8.73897299e-02 5.21297514e-01 -6.11939430e-01 -3.39249313e-01
-6.74614072e-01 4.25117314e-01 8.56143907e-02 -1.70804989e-02
5.02598345e-01 -4.66885775e-01 -1.07498281e-01 -3.41399983e-02
-3.21697831e-01 4.79080379e-01 -4.18925509e-02 1.20174313e+00
-3.64547312e-01 -1.07081139e+00 9.75179493e-01 -1.23723459e+00
1.15909088e+00 -6.04358256e-01 5.77291511e-02 -2.16413476e-02
-4.48291451e-01 4.82610799e-02 4.09101367e-01 -3.29337716e-01
-1.37054527e+00 -2.14275032e-01 -3.09227377e-01 -8.78785644e-03
-3.05261970e-01 5.53164601e-01 8.45733434e-02 3.66637856e-01
9.95745897e-01 -3.20436686e-01 2.19909459e-01 -5.50177932e-01
9.00832951e-01 9.96628523e-01 4.29963529e-01 -5.83468020e-01
4.90554385e-02 3.81818324e-01 -7.18154669e-01 -1.02356064e+00
-1.50712228e+00 -4.28748339e-01 -7.56291807e-01 -3.23786914e-01
1.12225974e+00 -1.17775869e+00 -4.44340706e-01 1.88629732e-01
-1.25844598e+00 -9.02588248e-01 -3.31089199e-01 3.11835885e-01
-6.54905677e-01 1.11738734e-01 -1.01904964e+00 -1.22491086e+00
-6.27340436e-01 -5.67314446e-01 9.72334921e-01 1.78969681e-01
-7.06691325e-01 -1.26117563e+00 5.32676220e-01 7.23878264e-01
3.44694644e-01 8.23225379e-02 5.66125810e-01 -1.40588450e+00
2.81083435e-01 -8.86802152e-02 -1.54901683e-01 2.28553444e-01
-6.16298430e-02 -4.77785468e-01 -1.48348117e+00 -1.27341777e-01
-1.38704758e-02 -1.47416377e+00 7.18230069e-01 -1.34995997e-01
5.23844123e-01 -4.56564307e-01 -8.35298523e-02 -1.52522931e-02
5.89112282e-01 -4.37546879e-01 3.97385657e-01 2.68843055e-01
4.82286572e-01 9.13360298e-01 7.61794865e-01 9.34028983e-01
8.58221292e-01 6.34461999e-01 1.75343361e-02 -3.47126007e-01
2.14306250e-01 -1.82189688e-01 2.70457566e-01 6.08733714e-01
1.10500850e-01 -2.92325407e-01 -1.03518093e+00 7.16748178e-01
-2.20047283e+00 -7.44338334e-01 7.18502775e-02 1.68346286e+00
1.33759975e+00 1.41118973e-01 2.88080245e-01 -3.21153015e-01
2.52319425e-01 5.79925179e-01 -3.31558257e-01 -6.95288837e-01
5.10505773e-02 6.28395528e-02 -2.08104655e-01 9.01886880e-01
-1.18810248e+00 1.21070457e+00 6.29747629e+00 4.18181956e-01
-6.24523699e-01 5.68535209e-01 9.80913997e-01 -1.95076719e-01
-5.48236072e-02 -2.02828482e-01 -7.03919470e-01 1.57019064e-01
1.44006276e+00 -4.13026996e-02 2.15558082e-01 7.51306653e-01
3.02021712e-01 -2.13879958e-01 -1.40921795e+00 6.86822891e-01
2.88880259e-01 -1.28874683e+00 -3.33281279e-01 6.61992468e-03
6.94142342e-01 3.42972547e-01 -3.79492670e-01 1.03750265e+00
8.22236121e-01 -1.19750118e+00 5.22974320e-03 4.44263965e-01
5.84519625e-01 -3.53834331e-01 8.45291615e-01 7.38946736e-01
-3.72093439e-01 1.47345468e-01 -5.46687543e-02 -5.68670154e-01
5.80817997e-01 2.52539665e-01 -1.23627853e+00 2.22553328e-01
3.94507796e-01 8.53975296e-01 -2.87259489e-01 1.36226624e-01
-2.16683060e-01 5.68566740e-01 -3.84071350e-01 -4.58501577e-02
6.48932934e-01 -5.24108484e-02 3.49825144e-01 1.38329053e+00
-3.73698175e-01 5.65969884e-01 5.80177784e-01 7.39720523e-01
-5.11880159e-01 3.53100121e-01 -6.81969464e-01 -3.23270798e-01
2.58223921e-01 1.56308448e+00 -4.66177315e-02 -7.74506271e-01
-5.98885238e-01 1.11883450e+00 9.56666768e-01 5.98831289e-02
-4.41137761e-01 2.11284354e-01 7.07112610e-01 -1.10328183e-01
-2.52467960e-01 1.97686348e-02 -7.58802593e-02 -1.36664689e+00
-4.70168620e-01 -1.08075774e+00 5.81535757e-01 -6.10848486e-01
-1.49409020e+00 8.56385171e-01 -2.62376428e-01 -7.33208060e-01
-9.51895177e-01 -4.27369624e-01 -7.05277503e-01 7.77741134e-01
-1.22708023e+00 -1.40250456e+00 -5.22337258e-01 4.45909619e-01
1.02812874e+00 -1.85219318e-01 1.57021356e+00 4.23097499e-02
-5.68119586e-01 4.74847943e-01 -2.42230013e-01 3.01473588e-01
1.09008622e+00 -1.30959904e+00 5.02024889e-01 3.33126754e-01
-4.04742695e-02 2.42032826e-01 9.04415429e-01 -6.17432654e-01
-8.18411946e-01 -7.00519264e-01 8.46990585e-01 -8.74232292e-01
1.01906300e+00 -7.76651502e-01 -1.09725547e+00 9.38455999e-01
7.21898377e-01 -1.17166586e-01 1.14486933e+00 9.52172160e-01
-2.84529865e-01 5.53281307e-01 -1.21338332e+00 5.84313571e-01
1.02815890e+00 -7.05455899e-01 -9.67917025e-01 6.71716332e-01
7.32640207e-01 -5.49741983e-01 -1.01997542e+00 2.61136889e-01
3.69972646e-01 -8.15968156e-01 5.14321804e-01 -1.26741827e+00
7.45899916e-01 6.93484008e-01 1.15335602e-02 -1.67409837e+00
2.63553113e-01 -9.58765030e-01 -2.14265823e-01 1.45258379e+00
6.63273573e-01 -3.24373811e-01 7.51782358e-01 1.16549110e+00
-1.68188423e-01 -6.82944953e-01 -9.74573970e-01 -2.53473133e-01
6.42536879e-02 -2.35442400e-01 2.24413816e-02 1.44653428e+00
9.85938191e-01 1.23544931e+00 -8.43542993e-01 -6.35777771e-01
4.17037040e-01 -3.78788173e-01 1.17241049e+00 -1.14269078e+00
-1.25486046e-01 -4.24658209e-02 3.10584437e-02 -1.01997876e+00
8.31296802e-01 -5.53977072e-01 9.10561532e-02 -1.59028459e+00
2.63057619e-01 -1.74049750e-01 2.66320735e-01 7.40443468e-01
-4.65364695e-01 3.25850576e-01 1.39145389e-01 -2.89646178e-01
-9.47298884e-01 1.10155022e+00 6.72619104e-01 7.87387230e-03
-4.95175660e-01 -4.38428402e-01 -6.59617901e-01 8.56227279e-01
5.61546028e-01 6.85804989e-03 -4.20178235e-01 -1.55097112e-01
-1.61549389e-01 1.39582053e-01 4.44116816e-02 -6.95759535e-01
-7.69947544e-02 2.50020415e-01 -3.65162105e-03 -1.74661428e-01
1.00382662e+00 -2.20381916e-01 -6.71444237e-01 7.63575137e-02
-9.19559181e-01 -3.74518931e-01 3.82098883e-01 4.87524569e-01
5.25057130e-02 -3.03494662e-01 7.32675314e-01 -3.33933800e-01
-6.02372348e-01 -2.14020200e-02 -6.23895526e-01 5.48583865e-01
6.80259883e-01 2.86220491e-01 -7.27960706e-01 -1.05038512e+00
-1.17782831e+00 6.38789713e-01 2.34773159e-01 5.98293364e-01
3.14024031e-01 -1.02378917e+00 -1.16313112e+00 -6.39935732e-02
3.62010211e-01 -3.50283623e-01 3.65031213e-01 9.46543157e-01
9.76681113e-02 3.57349247e-01 -1.86758205e-01 -3.89091223e-01
-1.22574008e+00 1.45985037e-01 4.78441805e-01 -7.14597762e-01
-5.72690368e-01 8.65842164e-01 -2.40593310e-02 -9.99016166e-01
1.17670290e-01 5.91443591e-02 -5.37079036e-01 5.54448307e-01
7.45177388e-01 3.76624279e-02 -5.92827704e-03 -3.73134285e-01
-1.47627160e-01 -3.09464008e-01 -5.47219694e-01 -6.34573758e-01
1.55199790e+00 -3.26437145e-01 1.66154951e-01 8.59820247e-01
9.95271802e-01 -2.74636090e-01 -1.30809951e+00 -4.96811330e-01
-3.69611988e-03 -1.85240597e-01 7.92992860e-02 -1.08154905e+00
-2.98311442e-01 9.08024430e-01 4.24364477e-01 2.78804839e-01
3.67278218e-01 2.30211809e-01 4.99827325e-01 6.17387533e-01
1.92645341e-01 -1.45307577e+00 4.86941040e-01 1.01822734e+00
9.10472095e-01 -1.85694718e+00 -1.14880964e-01 -3.11760694e-01
-1.29067647e+00 7.53616095e-01 1.05208647e+00 5.35444245e-02
2.90895313e-01 2.58381754e-01 3.09642047e-01 -2.71217674e-01
-1.29471767e+00 -4.50524420e-01 1.72395930e-02 5.01980126e-01
8.82940829e-01 -1.77618235e-01 -3.85935068e-01 9.36397851e-01
-1.82190716e-01 -5.46665266e-02 7.00795472e-01 7.95116544e-01
-2.12576434e-01 -8.59856308e-01 1.35874897e-01 5.54251730e-01
-5.09722590e-01 -4.98310953e-01 -7.39245415e-01 7.18048573e-01
-5.82522392e-01 1.28708398e+00 2.06609473e-01 -2.63805896e-01
3.86495292e-01 6.31405771e-01 1.62410632e-01 -1.00507677e+00
-8.23447704e-01 -1.70895040e-01 1.35769653e+00 -6.96644008e-01
-5.08798480e-01 -5.07155895e-01 -1.11124599e+00 -2.05872044e-01
-2.81312138e-01 3.29811126e-01 3.76300335e-01 1.38049698e+00
4.22981530e-01 4.93486762e-01 4.01050299e-01 -9.47322071e-01
-5.63964725e-01 -1.81978631e+00 -1.15983732e-01 6.44962013e-01
3.27784240e-01 -5.48942089e-01 -3.53200376e-01 -1.94141552e-01]
|
[12.793781280517578, 7.971930503845215]
|
c890778b-0273-4d4b-bf75-2fa5de4fcd71
|
open-arms-open-source-arms-hands-control
|
2205.12992
| null |
https://arxiv.org/abs/2205.12992v2
|
https://arxiv.org/pdf/2205.12992v2.pdf
|
Open Arms: Open-Source Arms, Hands & Control
|
Open Arms is a novel open-source platform of realistic human-like robotic hands and arms hardware with 28 Degree-of-Freedom (DoF), designed to extend the capabilities and accessibility of humanoid robotic grasping and manipulation. The Open Arms framework includes an open SDK and development environment, simulation tools, and application development tools to build and operate Open Arms. This paper describes these hands controls, sensing, mechanisms, aesthetic design, and manufacturing and their real-world applications with a teleoperated nursing robot. From 2015 to 2022, the authors have designed and established the manufacturing of Open Arms as a low-cost, high functionality robotic arms hardware and software framework to serve both humanoid robot applications and the urgent demand for low-cost prosthetics, as part of the Hanson Robotics Sophia Robot platform. Using the techniques of consumer product manufacturing, we set out to define modular, low-cost techniques for approximating the dexterity and sensitivity of human hands. To demonstrate the dexterity and control of our hands, we present a Generative Grasping Residual CNN (GGR-CNN) model that can generate robust antipodal grasps from input images of various objects in real-time speeds (22ms). We achieved state-of-the-art accuracy of 92.4% using our model architecture on a standard Cornell Grasping Dataset, which contains a diverse set of household objects.
|
['Raviteja Upadrashta', 'Rushali Mohbe', 'Aman Malali', 'Aditya Sagi', 'Vytas Krisciunas', 'Gerardo Morales', 'Alishba Imran', 'David Hanson']
|
2022-05-20
| null | null | null | null |
['robotic-grasping']
|
['robots']
|
[-3.73067021e-01 3.63917470e-01 3.54811817e-01 -6.89676180e-02
1.19348466e-01 -6.65254235e-01 -1.72532141e-01 -1.22044528e+00
8.86619017e-02 3.43634278e-01 -1.08421646e-01 -1.60685167e-01
-4.62121338e-01 -6.38642848e-01 -8.78298283e-01 -6.09557331e-01
-3.89837712e-01 6.88408673e-01 -1.06030650e-01 -5.38689733e-01
-9.73792672e-02 1.13101399e+00 -1.59695864e+00 4.25203204e-01
3.54074866e-01 1.03166080e+00 7.56169915e-01 7.81945467e-01
4.04956490e-01 6.97549105e-01 -4.16354150e-01 5.84401302e-02
7.25564241e-01 2.76456475e-01 -7.48440444e-01 -3.46034467e-01
-1.62616730e-01 -9.85682607e-01 -5.99259317e-01 4.09857839e-01
9.05898571e-01 -2.27236211e-01 2.98377842e-01 -1.42957568e+00
-1.06019008e+00 9.22459304e-01 -3.01490016e-02 -9.03877854e-01
4.93082017e-01 5.34026325e-01 3.85820508e-01 -6.72129691e-01
9.52817678e-01 1.46466887e+00 7.36918747e-01 1.12238097e+00
-7.10348666e-01 -7.14866459e-01 -5.42157888e-01 -3.08208853e-01
-9.25368011e-01 -6.62166849e-02 3.79162818e-01 -3.76488030e-01
1.05366766e+00 -2.22044233e-02 6.49945498e-01 1.80818200e+00
1.07809854e+00 5.42136133e-01 5.78229904e-01 -4.60031360e-01
1.47714689e-01 -4.05417502e-01 -2.12543532e-01 8.55813265e-01
1.98253095e-01 3.10466975e-01 2.67546684e-01 -4.08365168e-02
1.70748186e+00 4.97840732e-01 -6.28891140e-02 -6.10298693e-01
-1.61903870e+00 4.03261542e-01 1.07954335e+00 1.48980081e-01
-7.89972603e-01 7.16633797e-01 1.63950041e-01 1.75828248e-01
-6.12839520e-01 5.14837682e-01 -9.15768206e-01 -1.73074767e-01
4.60745096e-01 4.27443415e-01 1.40716970e+00 2.05278683e+00
-1.60231426e-01 -1.58921778e-01 -2.32872635e-01 7.05019236e-01
4.76546258e-01 6.26963854e-01 3.19798321e-01 -1.28943479e+00
2.36882158e-02 5.61023891e-01 4.75276470e-01 -6.28315210e-01
-7.26128221e-01 3.19649845e-01 -5.24532676e-01 7.23585904e-01
1.60462230e-01 -2.41160050e-01 -1.36466360e+00 1.26072478e+00
2.36755565e-01 -1.12058890e+00 1.50376156e-01 1.21425819e+00
9.31528449e-01 2.79710442e-01 -1.81201056e-01 5.32310843e-01
1.29151189e+00 -1.08193302e+00 -5.17309487e-01 1.75040916e-01
-2.03006878e-03 -7.60369420e-01 1.28142011e+00 7.56624460e-01
-9.70682085e-01 -6.01912022e-01 -9.26138580e-01 -1.59532353e-01
-3.62172514e-01 3.89826208e-01 1.09983945e+00 3.25017542e-01
-8.82203341e-01 1.06145287e+00 -1.21602273e+00 -5.77571213e-01
3.70839119e-01 7.69187152e-01 -3.68016988e-01 -1.52989775e-01
-5.92745960e-01 1.20859432e+00 2.53018916e-01 5.53519242e-02
-1.05235279e+00 -5.53441167e-01 -4.25570339e-01 -9.90198478e-02
-2.11561248e-02 -7.73073137e-01 1.53490853e+00 -2.51304895e-01
-2.03763533e+00 5.55547833e-01 1.09090590e+00 9.08688158e-02
5.51507473e-01 -5.15272200e-01 -1.35607868e-01 1.76556379e-01
6.96078464e-02 9.65480387e-01 7.32400000e-01 -1.19735312e+00
7.19752954e-03 -4.78127033e-01 1.91254050e-01 -3.88041973e-01
-4.69153412e-02 -2.88726725e-02 -6.23434223e-02 -7.42157042e-01
2.06414595e-01 -1.46307588e+00 -7.66279995e-02 8.54680955e-01
-1.64555803e-01 -1.57400176e-01 9.65796828e-01 -6.75451577e-01
5.12968302e-02 -2.09876800e+00 4.69912499e-01 8.81423280e-02
-4.72338311e-02 2.09694672e-02 -1.44034058e-01 5.13560891e-01
2.78694689e-01 -6.18766487e-01 3.75473648e-01 5.44142365e-01
3.44315708e-01 6.24257743e-01 -2.85401613e-01 2.18666524e-01
-1.03255175e-01 9.25016880e-01 -8.29475820e-01 -1.68700784e-01
2.99737334e-01 5.91997325e-01 -6.02887988e-01 5.23978114e-01
-3.83597791e-01 5.37314415e-01 -4.86044347e-01 1.30114114e+00
5.71557283e-01 -1.93209350e-02 2.64689922e-01 -3.46621782e-01
3.13108861e-02 -4.22387451e-01 -9.23809230e-01 2.15435553e+00
-7.72354245e-01 -1.17717251e-01 7.68381596e-01 3.79276425e-02
9.73295271e-01 3.13771516e-01 6.62416637e-01 -2.17932954e-01
7.16339409e-01 6.65539026e-01 3.73118371e-02 -9.56591308e-01
4.43752669e-02 9.04664546e-02 -8.58707204e-02 2.66409487e-01
3.95913184e-01 -5.25730252e-01 -2.71168947e-01 -2.56588578e-01
1.31838560e+00 1.05884373e+00 -1.50406763e-01 -4.05967116e-01
-4.80595708e-01 1.00000180e-01 -1.15125284e-01 2.79641092e-01
6.76283836e-02 5.69505990e-01 -3.20320688e-02 -4.55065548e-01
-1.40529299e+00 -1.39911544e+00 -1.13281913e-01 1.25872350e+00
-1.18228056e-01 3.90715241e-01 -6.07909203e-01 -2.16468330e-02
6.94535196e-01 1.97245508e-01 -4.55215842e-01 6.57299813e-03
-6.47317886e-01 2.86385655e-01 8.14191878e-01 1.20871401e+00
2.94794142e-01 -1.67794716e+00 -1.52099133e+00 4.65289354e-01
4.32363927e-01 -1.02913249e+00 -1.40257761e-01 3.55273485e-01
-5.44587374e-01 -1.17288518e+00 -1.08327436e+00 -1.25002623e+00
7.03359663e-01 4.93739136e-02 6.19376898e-01 -1.83449641e-01
-1.14095712e+00 6.21957779e-01 -6.79176927e-01 -6.80699110e-01
-5.32688618e-01 -1.88121364e-01 4.52961892e-01 -1.08709085e+00
-1.48359001e-01 -5.39228141e-01 -5.92327833e-01 6.13914907e-01
-5.07770479e-01 4.19342816e-02 1.07258618e+00 9.28811193e-01
1.37367845e-01 -8.51954103e-01 6.77531064e-01 2.49179885e-01
6.16956413e-01 -2.04954401e-01 -3.78968596e-01 2.22991973e-01
-2.76605576e-01 -8.76320302e-02 4.88000125e-01 -7.46021152e-01
-8.30223739e-01 5.08923113e-01 -1.40686527e-01 -8.08704197e-01
5.08576855e-02 -2.16707140e-01 -6.81664646e-02 -4.81766552e-01
7.66410530e-01 -3.80272001e-01 7.60434687e-01 -3.63163143e-01
6.45100236e-01 1.45610058e+00 8.44548285e-01 -1.07116592e+00
-8.10115263e-02 3.76373410e-01 3.10890675e-02 -5.16446471e-01
1.43692881e-01 2.61260383e-02 -7.26709485e-01 -3.09231579e-01
6.79388225e-01 -7.66610384e-01 -1.50878108e+00 6.62854970e-01
-1.45071685e+00 -8.25963140e-01 -2.10799426e-01 2.83054650e-01
-1.20177460e+00 -4.67061922e-02 -1.24096644e+00 -5.31679869e-01
-9.41590250e-01 -1.27688098e+00 1.25226545e+00 3.91410552e-02
-4.84121412e-01 6.07468979e-03 -3.74420136e-01 3.38872783e-02
7.66052902e-01 8.83694887e-01 7.33217180e-01 4.74118069e-02
-3.97468269e-01 -3.15970153e-01 -1.91361874e-01 3.89858842e-01
4.07868028e-01 3.51420566e-02 -6.13419950e-01 -6.52861178e-01
-3.11599463e-01 -7.01885641e-01 -3.94289121e-02 2.95057833e-01
1.07618058e+00 -4.70084310e-01 -6.48950934e-01 2.88620800e-01
1.20706975e+00 5.33084393e-01 8.29242945e-01 3.23699266e-02
6.49767578e-01 5.86292744e-01 6.06460273e-01 6.25595927e-01
-2.95596838e-01 3.38974386e-01 8.09309900e-01 4.48020488e-01
-9.52273905e-02 -6.72865734e-02 2.22784370e-01 2.89320707e-01
-6.06737792e-01 1.73560932e-01 -8.44596565e-01 3.42761993e-01
-1.61185730e+00 -4.08350438e-01 -6.98514953e-02 1.75168383e+00
6.40106142e-01 -2.89644927e-01 1.28233746e-01 -1.26013905e-01
4.96521592e-01 -1.03631246e+00 -7.52410293e-01 -4.48168248e-01
6.05676889e-01 5.55617988e-01 4.91270244e-01 -3.17146629e-01
-6.37468219e-01 6.87988341e-01 6.25315666e+00 1.98105313e-02
-1.01020479e+00 -3.79719138e-02 -4.33127105e-01 -2.03810379e-01
4.38340157e-01 -7.16407955e-01 -2.44706452e-01 3.89185339e-01
4.24133480e-01 4.13764030e-01 8.95080149e-01 1.67528760e+00
-4.11580801e-01 4.70604718e-01 -1.48147035e+00 7.76255250e-01
-8.67670253e-02 -1.14315248e+00 -2.10020095e-01 -4.78656441e-02
4.50486690e-01 1.59324050e-01 -3.43897372e-01 5.13610184e-01
4.37953979e-01 -1.09556043e+00 1.09548318e+00 5.67622423e-01
1.11935365e+00 -4.51752782e-01 4.95766193e-01 1.94920719e-01
-7.94252455e-01 -9.19642448e-01 -3.59613478e-01 -1.91570967e-01
6.05033860e-02 -2.03317568e-01 -9.83220637e-01 1.77596405e-01
1.26715040e+00 -3.52602936e-02 2.05440193e-01 4.45592463e-01
3.50086689e-02 -7.51186490e-01 -3.09746772e-01 -5.55466056e-01
-9.53037739e-02 3.28464210e-01 2.76878774e-01 7.72143185e-01
4.34426516e-01 4.17819351e-01 -5.83701357e-02 1.29274464e+00
9.63012278e-02 -5.23445725e-01 -6.90232337e-01 -3.27205509e-01
2.89073616e-01 1.25582945e+00 -5.79493642e-01 1.35833636e-01
7.58432150e-02 1.15818071e+00 -1.18024632e-01 -3.19722146e-01
-8.21735620e-01 -1.26927972e+00 6.27523780e-01 1.04226433e-01
2.64114976e-01 -6.27528727e-01 -3.05628866e-01 -5.72985113e-01
2.50155896e-01 -8.18871260e-01 -3.59630823e-01 -1.34802151e+00
-1.33002043e+00 5.03806949e-01 -7.61094838e-02 -1.32982099e+00
-3.18195462e-01 -1.55635929e+00 -1.53169811e-01 5.80830991e-01
-2.76934415e-01 -1.85506213e+00 -1.02065372e+00 4.75333750e-01
6.24789655e-01 -2.24903330e-01 1.32302558e+00 9.50717628e-02
2.01123610e-01 1.14362590e-01 1.39362574e-01 1.13029450e-01
5.70210099e-01 -7.94385791e-01 2.87475586e-01 -7.67436028e-02
-1.06505549e+00 1.02731872e+00 5.15961766e-01 -7.08004355e-01
-2.46821427e+00 -8.41633856e-01 -2.67494619e-01 -5.74876130e-01
5.20731986e-01 -3.32417727e-01 -2.34408557e-01 1.09718442e+00
-1.13276944e-01 -6.78875763e-03 5.11211902e-02 -5.51135123e-01
-2.81168837e-02 1.33235544e-01 -1.85739863e+00 5.93551517e-01
1.44868898e+00 5.89948893e-02 -8.70761991e-01 5.89911997e-01
8.94133568e-01 -9.94535208e-01 -1.20772743e+00 6.79145992e-01
1.66630936e+00 -3.06580931e-01 1.03297091e+00 -6.26483619e-01
7.82057166e-01 -1.74479801e-02 -4.29002553e-01 -1.05816066e+00
-5.77228367e-01 -6.79947197e-01 -1.82864428e-01 5.78899145e-01
-1.57864258e-01 -6.25662863e-01 3.61014724e-01 7.42891312e-01
-7.27615058e-01 -1.07025909e+00 -7.01063812e-01 -1.11324835e+00
2.87923906e-02 1.18888207e-01 8.04553449e-01 4.18970734e-01
7.55873501e-01 -2.45925248e-01 4.62428220e-02 3.76371183e-02
3.06968600e-01 1.51837543e-01 8.49772453e-01 -1.21566045e+00
-4.27445322e-01 -1.09474070e-01 -5.19517004e-01 -5.04497349e-01
-5.56010641e-02 -6.28136337e-01 5.37269056e-01 -1.89701331e+00
-2.30985031e-01 -5.65324128e-01 3.97560269e-01 1.04684210e+00
9.19666886e-01 -1.91874132e-01 2.32653499e-01 2.21357495e-01
1.60097539e-01 9.40789878e-02 1.52990592e+00 -7.47827590e-02
-2.57695079e-01 -1.27625301e-01 -2.95529574e-01 4.59322989e-01
5.95720589e-01 4.47208099e-02 3.81316394e-02 -5.62995076e-01
-2.87756711e-01 1.98687837e-01 7.48731852e-01 -1.27349436e+00
1.98561829e-02 -1.64121047e-01 7.47291744e-01 -1.74032405e-01
3.46394330e-01 -1.29850447e+00 5.77724993e-01 1.33736002e+00
-2.17653941e-02 -2.01400295e-01 1.43195063e-01 5.08713089e-02
8.48099470e-01 1.70025662e-01 8.05720270e-01 -5.56271374e-01
-4.90270913e-01 1.16364686e-02 -1.50829688e-01 -9.47662354e-01
1.51317501e+00 -1.64822072e-01 -5.73731124e-01 2.34751478e-01
-9.73812640e-01 1.89598784e-01 5.86342037e-01 9.75216031e-01
7.86753595e-01 -1.29227710e+00 -3.38567615e-01 5.76640487e-01
9.05632414e-03 1.58087999e-01 1.33968428e-01 2.91665435e-01
-1.43820369e+00 4.15844262e-01 -1.33421504e+00 -4.51065004e-01
-8.46523166e-01 8.69475961e-01 8.72437134e-02 4.74263251e-01
-1.02621293e+00 5.13770580e-01 -5.76560974e-01 -1.05855441e+00
5.42080104e-01 -9.07357752e-01 4.96976852e-01 -8.88041019e-01
2.26526424e-01 8.64416897e-01 5.59219755e-02 1.08195253e-01
-5.02542555e-01 3.58661890e-01 4.46204215e-01 1.95726603e-01
1.62760544e+00 6.38565063e-01 -1.21593677e-01 4.38145325e-02
8.36276531e-01 -6.69895828e-01 -1.19533265e+00 8.80258262e-01
-6.01379633e-01 -2.16040358e-01 -5.94450831e-01 -1.19993007e+00
-8.31568062e-01 4.31672513e-01 9.84723091e-01 -1.77580535e-01
6.39857471e-01 3.28142494e-01 1.01349401e+00 1.02626419e+00
1.51781583e+00 -9.86697793e-01 4.03079540e-01 5.01869678e-01
2.02742052e+00 -8.00497890e-01 -1.88502625e-01 -4.84361976e-01
-4.88010824e-01 1.46420312e+00 8.76535892e-01 -5.44390500e-01
6.40144348e-01 7.78760254e-01 1.28180264e-02 -3.54501605e-01
-1.05958685e-01 5.08624434e-01 -4.54549432e-01 1.01051581e+00
1.91124976e-01 5.63559949e-01 -9.82259959e-02 1.04891431e+00
-3.94649655e-01 7.63706982e-01 2.64522284e-01 1.79226851e+00
-2.63802230e-01 -4.89474684e-01 -4.40510929e-01 4.49393392e-01
-2.09840760e-01 6.68654561e-01 -8.40517506e-02 8.24757814e-01
1.13989696e-01 6.38201058e-01 -1.87837496e-01 -6.49724007e-01
8.06189299e-01 -2.03940272e-01 1.18895876e+00 -3.51331979e-01
-5.58080733e-01 -4.12958533e-01 -2.04224691e-01 -9.44573998e-01
1.55241638e-01 4.79505323e-02 -1.73630440e+00 -2.54170418e-01
-2.48481348e-01 -6.64324820e-01 1.39201534e+00 3.22993875e-01
9.33335245e-01 7.81121552e-01 3.04965198e-01 -1.98864746e+00
-1.38333130e+00 -1.31839693e+00 -9.07221019e-01 1.68715283e-01
-1.46895066e-01 -9.37220693e-01 1.02387235e-01 -1.02363765e-01]
|
[5.722343444824219, -0.8222242593765259]
|
bd1cdf96-a46d-4dd5-ac18-845d39d0fea1
|
rare-words-degenerate-all-words
|
2109.03127
| null |
https://arxiv.org/abs/2109.03127v3
|
https://arxiv.org/pdf/2109.03127v3.pdf
|
Rare Tokens Degenerate All Tokens: Improving Neural Text Generation via Adaptive Gradient Gating for Rare Token Embeddings
|
Recent studies have determined that the learned token embeddings of large-scale neural language models are degenerated to be anisotropic with a narrow-cone shape. This phenomenon, called the representation degeneration problem, facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models. Although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation, the training dynamics of token embeddings behind the degeneration problem are still not explored. In this study, we analyze the training dynamics of the token embeddings focusing on rare token embedding. We demonstrate that the specific part of the gradient for rare token embeddings is the key cause of the degeneration problem for all tokens during training stage. Based on the analysis, we propose a novel method called, adaptive gradient gating (AGG). AGG addresses the degeneration problem by gating the specific part of the gradient for rare token embeddings. Experimental results from language modeling, word similarity, and machine translation tasks quantitatively and qualitatively verify the effectiveness of AGG.
|
['Sungroh Yoon', 'Woo-Jong Ryu', 'Seong-min Lee', 'Heeseung Kim', 'Jongyoon Song', 'Sangwon Yu']
|
2021-09-07
| null |
https://aclanthology.org/2022.acl-long.3
|
https://aclanthology.org/2022.acl-long.3.pdf
|
acl-2022-5
|
['word-similarity']
|
['natural-language-processing']
|
[-2.14802548e-01 2.00823154e-02 -3.87304932e-01 2.34049372e-02
-2.00892940e-01 -3.21237534e-01 7.17676997e-01 1.29122376e-01
-4.42841172e-01 5.38734853e-01 5.80180585e-01 -2.80313015e-01
1.64486617e-01 -5.86777687e-01 -7.27615356e-01 -8.66492510e-01
2.51932472e-01 2.08608985e-01 9.60838944e-02 -3.08931112e-01
4.32127297e-01 9.56179649e-02 -1.18185341e+00 2.00634569e-01
1.28245711e+00 6.52490377e-01 4.58820283e-01 2.18368605e-01
-6.60210371e-01 2.41931230e-01 -5.67508042e-01 -1.51188344e-01
1.91185623e-01 -3.79991353e-01 -5.66897750e-01 -3.06834131e-01
-7.09603503e-02 -3.27691048e-01 -2.92825699e-01 1.23458707e+00
6.65594339e-01 1.45883203e-01 8.57667327e-01 -1.14996541e+00
-1.31753922e+00 8.75388861e-01 -5.16382396e-01 4.86645103e-01
-1.11019790e-01 -5.53164892e-02 1.22423530e+00 -1.27898943e+00
6.85862422e-01 1.22304094e+00 6.10996604e-01 7.00755775e-01
-8.35273921e-01 -4.55821455e-01 4.36454117e-01 -1.55245718e-02
-1.47859025e+00 -2.68209539e-02 9.30392325e-01 -3.92699599e-01
1.08534265e+00 -2.30939001e-01 7.43132889e-01 1.11680734e+00
4.19968784e-01 7.65615761e-01 7.91958869e-01 -6.35734081e-01
2.47129470e-01 2.87973225e-01 1.91506967e-01 6.75958633e-01
5.56869626e-01 6.99106082e-02 -5.81070364e-01 -2.62402147e-01
8.09570730e-01 -7.12261721e-02 -2.64540780e-02 -1.05840929e-01
-1.03399920e+00 9.36348319e-01 3.68880570e-01 4.39881831e-01
-3.06723654e-01 3.85807842e-01 4.70407903e-01 2.72368878e-01
7.56302238e-01 6.43944860e-01 -4.11435843e-01 -3.99790734e-01
-5.82360625e-01 1.73711956e-01 4.40975904e-01 8.29260588e-01
7.81708777e-01 3.11981499e-01 -3.51429164e-01 9.57793832e-01
4.60780203e-01 2.32057512e-01 1.15036356e+00 -3.48107040e-01
5.55568457e-01 8.03105175e-01 -1.56362578e-02 -6.36399686e-01
-2.64502734e-01 -5.76612055e-01 -6.13313317e-01 -4.33122486e-01
4.18317854e-01 -3.72081161e-01 -8.57355654e-01 2.14302802e+00
2.82905132e-01 1.85264364e-01 6.29310906e-02 9.09899592e-01
5.42244613e-01 7.30940104e-01 3.47243577e-01 5.17913327e-02
1.27514386e+00 -9.07183528e-01 -7.86650419e-01 -1.88050792e-01
1.12009537e+00 -8.54544461e-01 1.35268724e+00 -3.36636484e-01
-8.28871131e-01 -5.28486371e-01 -9.03956890e-01 -1.48659155e-01
-4.39285874e-01 1.57152399e-01 7.05251634e-01 4.06044036e-01
-9.33559120e-01 6.75093651e-01 -7.25534916e-01 -5.18077910e-01
1.91804901e-01 1.02624618e-01 7.81042203e-02 5.53579293e-02
-1.55897367e+00 9.47887242e-01 1.22650005e-01 1.79338202e-01
-4.51855093e-01 -8.64363015e-01 -5.60747623e-01 2.16138199e-01
-3.29521537e-01 -6.60623133e-01 1.04773831e+00 -9.73838627e-01
-1.37481081e+00 4.40336078e-01 -3.70014876e-01 -2.36160770e-01
3.80168021e-01 -4.59924161e-01 9.06663612e-02 -1.94437832e-01
1.25898734e-01 7.10575342e-01 7.19689727e-01 -1.15897381e+00
-1.82524532e-01 -3.16198170e-01 -2.28184730e-01 3.98239285e-01
-9.08232808e-01 -1.70752168e-01 6.63815811e-02 -9.27328348e-01
-1.64666221e-01 -7.20240295e-01 -1.70775667e-01 -3.22394580e-01
-3.43480349e-01 -8.07676733e-01 5.45607150e-01 -1.92879602e-01
1.41618431e+00 -2.33525324e+00 1.46076623e-02 5.45347743e-02
9.77548733e-02 1.75665304e-01 -5.94561577e-01 8.23615909e-01
8.74761865e-02 4.87030715e-01 1.72678292e-01 -2.92557180e-01
7.14868084e-02 2.57203519e-01 -8.76878083e-01 2.46229813e-01
4.98272300e-01 8.86791945e-01 -1.11961079e+00 -3.02590609e-01
-2.81346798e-01 4.11981434e-01 -7.45276690e-01 3.30648363e-01
-1.13320395e-01 -5.46072721e-02 -8.15472126e-01 4.45202410e-01
6.67078793e-01 -5.29573783e-02 3.04135513e-02 -1.35209471e-01
-4.19811457e-01 4.11383450e-01 -6.70367122e-01 1.42969108e+00
-3.41470093e-01 4.89135712e-01 -3.86419117e-01 -8.90952885e-01
9.66865063e-01 4.60712403e-01 2.59073347e-01 -4.97578949e-01
5.23656756e-02 5.65337896e-01 3.47659677e-01 -4.95660305e-01
8.79783869e-01 -4.06554729e-01 2.22698927e-01 7.62109280e-01
9.69856754e-02 5.67746796e-02 7.12780207e-02 2.38199890e-01
8.58233988e-01 -8.80500972e-02 1.28652900e-02 -4.59700584e-01
-9.53559130e-02 -2.27513343e-01 5.28169215e-01 6.92934096e-01
-2.74526030e-01 5.11827230e-01 7.17935324e-01 -1.95492432e-01
-1.39122498e+00 -1.09886277e+00 -7.41802379e-02 1.26024330e+00
2.13299289e-01 -5.06700158e-01 -8.41886520e-01 -4.20108229e-01
1.22006662e-01 5.97608089e-01 -7.32566118e-01 -7.70682812e-01
-5.02972007e-01 -1.11682558e+00 6.18919909e-01 7.85353780e-01
3.31852645e-01 -1.13512456e+00 -2.10783258e-01 2.82052994e-01
-9.16454643e-02 -1.03962314e+00 -7.04385936e-01 1.96642205e-01
-1.12645710e+00 -4.81349558e-01 -6.64204895e-01 -8.57200146e-01
9.48040366e-01 2.88232595e-01 7.41990626e-01 3.18019480e-01
-1.71957836e-01 1.17038712e-01 -4.28888798e-01 -5.08646727e-01
-2.94113636e-01 2.73110092e-01 3.52483571e-01 -2.01537475e-01
4.38666046e-01 -6.05375290e-01 -6.25681102e-01 1.16310239e-01
-9.32648599e-01 -2.43459508e-01 7.51375377e-01 1.02621543e+00
3.04759771e-01 -2.39130303e-01 1.07309830e+00 -5.29282510e-01
1.26090741e+00 -6.27304077e-01 -1.69120505e-01 8.98623094e-02
-9.17370200e-01 5.52304089e-01 8.45340610e-01 -8.10363829e-01
-9.59429443e-01 -5.67562401e-01 -2.16656402e-02 -3.00643474e-01
2.41036132e-01 5.27007163e-01 1.87635392e-01 2.95804262e-01
3.60605478e-01 3.34407121e-01 -2.26583809e-01 -5.03965735e-01
5.74787021e-01 6.00675941e-01 -1.35052711e-01 -8.60932291e-01
8.18789303e-01 4.46536958e-01 -3.85843992e-01 -8.69518161e-01
-7.18671262e-01 -4.00526226e-01 -3.31782758e-01 1.57677144e-01
7.03815818e-01 -8.37523997e-01 7.87427798e-02 4.12839949e-01
-1.52680385e+00 -2.35344082e-01 -4.62992579e-01 6.11708105e-01
-1.83434159e-01 2.90562719e-01 -7.53652871e-01 -8.69924664e-01
-4.72498000e-01 -1.03093910e+00 8.32368135e-01 3.48898411e-01
-2.09358558e-01 -1.33838189e+00 4.03220713e-01 -1.19018443e-01
6.75625622e-01 -3.47439885e-01 1.44449961e+00 -8.44416499e-01
-3.90786439e-01 3.17947231e-02 -3.62181664e-01 3.47380042e-01
9.04054195e-03 1.80705085e-01 -7.92096257e-01 -5.20014316e-02
-1.12735853e-01 -1.28483295e-01 1.04079247e+00 4.00985420e-01
8.51298690e-01 -1.90127790e-01 -1.70035750e-01 3.86815131e-01
1.28254616e+00 -9.59537253e-02 4.53808993e-01 3.50872099e-01
7.06119180e-01 3.45959783e-01 3.81840706e-01 5.60682952e-01
2.27244809e-01 4.42472935e-01 2.16355383e-01 -1.21230990e-01
-1.52512684e-01 -5.94280243e-01 6.93583906e-01 1.48365009e+00
-1.68796390e-01 -5.26181757e-02 -7.08674192e-01 8.36633801e-01
-1.79696417e+00 -6.76175714e-01 1.89496204e-01 2.01805043e+00
1.01949191e+00 1.49721250e-01 -3.27212900e-01 -2.69018292e-01
8.33861589e-01 2.75943547e-01 -5.05249858e-01 -8.90159249e-01
-2.88231652e-02 1.39515296e-01 3.28865379e-01 3.86119604e-01
-3.45204145e-01 1.25109994e+00 6.59382057e+00 9.38520432e-01
-1.15550995e+00 2.75558740e-01 2.07030326e-01 -1.68836664e-03
-7.31708705e-01 2.18471605e-02 -9.42864001e-01 4.82325077e-01
8.20370138e-01 -5.14248908e-01 2.06944302e-01 7.99848974e-01
5.62450409e-01 7.73670003e-02 -1.02754951e+00 5.97387433e-01
-1.52370363e-01 -1.23493540e+00 6.37979925e-01 1.79250494e-01
8.94881845e-01 3.01650703e-01 2.68389523e-01 5.11878431e-01
1.41984135e-01 -8.24156582e-01 8.43240798e-01 2.33562648e-01
5.82873046e-01 -6.42989993e-01 7.23659813e-01 1.61937222e-01
-1.24549448e+00 -1.38247132e-01 -8.09590340e-01 -1.93072259e-01
9.21592042e-02 9.58821774e-01 -9.64352667e-01 1.47764817e-01
1.48052529e-01 4.76897538e-01 -5.53010345e-01 8.08009326e-01
-3.66735339e-01 8.44084799e-01 -5.14410669e-03 -2.95736283e-01
4.62658733e-01 -3.39071602e-01 5.97435892e-01 1.06117964e+00
4.50688422e-01 -4.42993522e-01 -1.24023698e-01 1.32783914e+00
-3.37623596e-01 1.04904942e-01 -6.44612014e-01 -6.48886859e-01
7.49166429e-01 1.09177172e+00 -5.62449574e-01 -9.84734818e-02
-2.81441182e-01 8.25372100e-01 6.33290768e-01 6.20735824e-01
-8.72444510e-01 -3.54629099e-01 8.74270082e-01 1.52079985e-01
3.27407002e-01 -3.59979093e-01 -4.61310357e-01 -1.05146301e+00
2.77729541e-01 -3.74734044e-01 -2.46789098e-01 -4.30345833e-01
-1.48920155e+00 4.18073505e-01 -2.01359361e-01 -8.79050553e-01
-2.21364759e-02 -4.79043186e-01 -1.10347903e+00 9.03981388e-01
-1.81555295e+00 -8.76307964e-01 1.92987576e-01 2.70137668e-01
4.73066419e-01 -9.95666608e-02 6.80849969e-01 1.75138593e-01
-8.36666822e-01 7.17871487e-01 5.49959660e-01 1.20911427e-01
6.81012869e-01 -1.19096303e+00 5.41577637e-01 6.05844915e-01
-4.82836738e-02 1.15647089e+00 4.79843140e-01 -6.57796323e-01
-1.23280728e+00 -1.09434235e+00 1.11780238e+00 -2.09645271e-01
9.30350244e-01 -5.34563959e-01 -9.09169912e-01 3.00627738e-01
1.51016966e-01 -4.41936180e-02 7.82417536e-01 1.74314901e-01
-4.83072609e-01 7.75425509e-02 -6.89521551e-01 6.28217459e-01
1.02603281e+00 -6.37692392e-01 -7.14475036e-01 3.43252867e-01
9.64140236e-01 1.47647873e-01 -5.45492232e-01 4.17391509e-02
4.78620827e-01 -5.73215187e-01 6.84593320e-01 -8.98381650e-01
7.47706413e-01 3.37114222e-02 -1.27324392e-03 -1.67645121e+00
-2.63695657e-01 -4.52173531e-01 -1.13507338e-01 1.19023180e+00
4.43437397e-01 -8.67615879e-01 3.74660164e-01 1.42768249e-01
-3.51345718e-01 -1.14526045e+00 -1.18328345e+00 -9.59422767e-01
6.31176293e-01 -6.95348345e-03 4.35545802e-01 8.45403552e-01
2.12333158e-01 3.13216984e-01 9.69839320e-02 -2.05229014e-01
2.31232420e-01 -3.75919901e-02 1.91112518e-01 -9.06367064e-01
7.02674687e-02 -5.60444832e-01 -1.18621655e-01 -1.20116580e+00
3.02801073e-01 -9.35200572e-01 -7.13891583e-03 -1.59874034e+00
3.12817484e-01 -6.26047254e-01 -4.75384295e-01 1.39200717e-01
-6.14121139e-01 -1.87352702e-01 1.10175647e-01 4.34022874e-01
-3.98493767e-01 1.06068373e+00 1.30588317e+00 2.98813075e-01
-1.55484363e-01 -5.03381729e-01 -8.37293446e-01 5.30842304e-01
8.86264980e-01 -5.18728495e-01 -4.53224808e-01 -8.51221442e-01
5.28368890e-01 -7.40781307e-01 1.09254695e-01 -5.42265892e-01
-4.78818417e-02 -1.31076202e-01 8.62420499e-02 -2.11487740e-01
-1.77636580e-03 -2.67550856e-01 -6.95737302e-01 5.64825118e-01
-4.32602406e-01 3.75054330e-01 2.38935295e-02 5.99701047e-01
-2.17972994e-01 -3.55971962e-01 6.84510410e-01 2.06745956e-02
-3.56288403e-01 2.58189708e-01 -6.60066307e-01 4.77767467e-01
6.65105820e-01 -6.60523474e-02 -2.48994991e-01 2.21320856e-02
-1.41298965e-01 8.56340751e-02 2.42601499e-01 6.48636818e-01
6.02416277e-01 -1.69390786e+00 -5.83132982e-01 2.56284475e-01
-4.71745692e-02 -1.68247566e-01 6.79436475e-02 7.90495038e-01
-8.73681009e-02 5.20258307e-01 3.16120535e-02 -2.13693917e-01
-6.16498590e-01 1.84357703e-01 3.53749007e-01 -6.58867598e-01
-3.86977553e-01 1.01957476e+00 3.40035051e-01 -4.04836088e-01
-6.84365407e-02 -7.25930750e-01 -1.69351771e-01 2.64744222e-01
3.82508367e-01 1.81721136e-01 -1.05377734e-01 -3.18601310e-01
-2.12981805e-01 5.58316946e-01 -4.25865054e-01 -1.24767693e-02
1.24868846e+00 -1.14898555e-01 -1.82225689e-01 5.96021473e-01
1.13225996e+00 -1.55625399e-02 -1.01542377e+00 -1.51620150e-01
-1.29260868e-01 -3.16359580e-01 -8.31417441e-02 -3.67217422e-01
-8.20369840e-01 1.04078603e+00 4.29543138e-01 2.04332024e-01
5.57939410e-01 -4.08629421e-03 1.20550764e+00 1.77741125e-01
-3.58992741e-02 -1.51624465e+00 4.44788545e-01 8.76281500e-01
7.00310290e-01 -9.53177512e-01 -3.44585657e-01 -1.55784205e-01
-6.08809471e-01 1.09346104e+00 7.95588017e-01 -3.09444100e-01
5.96899629e-01 1.59237921e-01 8.76127854e-02 -2.99523994e-02
-9.99884844e-01 1.51092902e-01 -3.86299863e-02 5.65464616e-01
7.00763702e-01 6.43580779e-03 -8.50367725e-01 7.54416525e-01
-2.90659726e-01 -1.95692852e-01 4.96078044e-01 8.60672295e-01
-6.08468890e-01 -1.26507652e+00 -1.99296221e-01 2.06396759e-01
-2.54024059e-01 -5.97917795e-01 -4.10346806e-01 4.79239911e-01
2.36139596e-01 5.71105421e-01 3.08113247e-01 -2.47831225e-01
1.09440193e-01 3.95391762e-01 3.72224718e-01 -7.72585690e-01
-5.93147933e-01 -2.45940432e-01 -2.69290864e-01 -1.82762995e-01
1.19261649e-02 -4.31099534e-01 -1.60137880e+00 -9.15978625e-02
-6.26367092e-01 3.99029136e-01 7.48062789e-01 1.13352096e+00
5.38012743e-01 4.98332620e-01 6.81691110e-01 -4.87889141e-01
-1.15361130e+00 -1.26421201e+00 -6.88210189e-01 5.60168743e-01
2.25434303e-01 -7.99620986e-01 -7.50308871e-01 -3.07570517e-01]
|
[10.841097831726074, 8.640522003173828]
|
7f83561c-6a15-44b4-8b96-54c06d348adf
|
neural-network-kalman-filtering-for-3d-object
|
2111.09631
| null |
https://arxiv.org/abs/2111.09631v3
|
https://arxiv.org/pdf/2111.09631v3.pdf
|
Neural Network Kalman filtering for 3D object tracking from linear array ultrasound data
|
Many interventional surgical procedures rely on medical imaging to visualise and track instruments. Such imaging methods not only need to be real-time capable, but also provide accurate and robust positional information. In ultrasound applications, typically only two-dimensional data from a linear array are available, and as such obtaining accurate positional estimation in three dimensions is non-trivial. In this work, we first train a neural network, using realistic synthetic training data, to estimate the out-of-plane offset of an object with the associated axial aberration in the reconstructed ultrasound image. The obtained estimate is then combined with a Kalman filtering approach that utilises positioning estimates obtained in previous time-frames to improve localisation robustness and reduce the impact of measurement noise. The accuracy of the proposed method is evaluated using simulations, and its practical applicability is demonstrated on experimental data obtained using a novel optical ultrasound imaging setup. Accurate and robust positional information is provided in real-time. Axial and lateral coordinates for out-of-plane objects are estimated with a mean error of 0.1mm for simulated data and a mean error of 0.2mm for experimental data. Three-dimensional localisation is most accurate for elevational distances larger than 1mm, with a maximum distance of 6mm considered for a 25mm aperture.
|
['Andreas Hauptmann', 'Mikko J. Sillanpää', 'Adrien Desjardins', 'Simon Arridge', 'Efthymios Maneas', 'Erwin J. Alles', 'Arttu Arjas']
|
2021-11-18
| null | null | null | null |
['3d-object-tracking']
|
['computer-vision']
|
[ 3.97193164e-01 2.78850406e-01 5.14351904e-01 -1.65388640e-02
-8.35506141e-01 -6.05190575e-01 3.02473009e-01 4.55348700e-01
-8.24880362e-01 6.19520843e-01 -9.92121547e-02 -3.32962960e-01
-8.46498013e-01 -1.63964987e-01 -6.03313327e-01 -9.59014356e-01
-4.98054624e-01 4.13241088e-01 1.15897939e-01 3.87482345e-01
3.46971035e-01 7.46297836e-01 -1.20705569e+00 -2.38829106e-01
6.22303784e-01 9.64742839e-01 5.22516370e-01 9.80903685e-01
4.57863599e-01 1.82358935e-01 -7.09048390e-01 2.19725259e-02
1.68748975e-01 -1.59076303e-01 -3.23679596e-01 2.21695274e-01
1.00899987e-01 -4.69672054e-01 2.49930620e-02 8.55646431e-01
9.86612558e-01 -7.63640180e-02 6.27134264e-01 -7.95392692e-02
2.95217425e-01 1.35693908e-01 -3.31122011e-01 1.95939522e-02
2.39753723e-01 1.60057157e-01 -1.10183232e-01 -7.19921529e-01
7.97365487e-01 3.71499985e-01 9.37129319e-01 9.98548791e-02
-1.03919160e+00 -1.84471637e-01 -7.63301969e-01 -1.26150429e-01
-1.19499969e+00 -3.74052107e-01 6.06370807e-01 -8.77757967e-01
3.08409661e-01 3.34665090e-01 6.36368215e-01 6.02930307e-01
9.12158132e-01 3.74888629e-02 1.06637990e+00 -7.46765614e-01
2.23300502e-01 1.45807818e-01 -4.95674998e-01 5.18285275e-01
4.89380062e-01 3.64519477e-01 1.29859418e-01 3.75991613e-02
1.36479008e+00 -9.48468298e-02 -7.62706578e-01 -8.83758307e-01
-1.47482586e+00 5.03840685e-01 5.26785195e-01 8.01464617e-01
-8.87427926e-01 -7.16553116e-03 3.36789787e-01 -1.61304116e-01
8.19656551e-02 8.09365511e-01 7.04376921e-02 -4.11013246e-01
-8.70018661e-01 -4.32852119e-01 6.31998599e-01 5.30775368e-01
-2.30798833e-02 -7.50574023e-02 -4.21619378e-02 5.13958991e-01
3.40167105e-01 4.06877309e-01 7.53604174e-01 -7.80954599e-01
5.17156757e-02 2.37932615e-02 4.45279360e-01 -1.08619046e+00
-9.53119397e-01 -9.05521512e-01 -1.02460158e+00 4.65562522e-01
6.02485240e-01 -3.46929610e-01 -1.05269182e+00 1.17572665e+00
5.10472834e-01 4.00926441e-01 1.65930212e-01 1.16178620e+00
7.01600730e-01 2.43264511e-01 -5.76599240e-01 -7.10408866e-01
1.15206420e+00 -4.81640875e-01 -7.17970967e-01 -2.15827823e-01
6.78026974e-01 -1.14547575e+00 3.24086279e-01 6.49245918e-01
-1.32967472e+00 -3.89286786e-01 -9.84479070e-01 8.12723339e-01
3.65098774e-01 6.76232755e-01 8.58039111e-02 5.73010683e-01
-7.85919964e-01 6.45582139e-01 -1.11289907e+00 -9.74155068e-02
1.45863835e-02 4.76097405e-01 -5.25310457e-01 1.06207334e-01
-8.21509361e-01 1.02218163e+00 3.06029141e-01 6.44497871e-01
-2.07162350e-01 -7.88000047e-01 -8.07579756e-01 -2.21077472e-01
2.06882358e-01 -5.27590215e-01 1.30536377e+00 -4.10941571e-01
-1.76016247e+00 5.36700904e-01 1.11630335e-01 -3.20902944e-01
6.14822149e-01 -1.67938620e-01 -1.99781373e-01 6.13219857e-01
-8.70480761e-02 -2.68274099e-01 8.29830289e-01 -1.35314333e+00
-2.20057100e-01 -3.73686373e-01 -1.55454442e-01 1.42196134e-01
5.03172614e-02 -3.29159439e-01 -3.93237829e-01 -4.10130829e-01
8.05105865e-01 -1.06754029e+00 -4.56866354e-01 9.19084102e-02
-2.97819912e-01 7.88420439e-01 2.13654429e-01 -7.38297284e-01
1.13533294e+00 -1.84010971e+00 -2.22245958e-02 4.89913642e-01
2.82249991e-02 5.63613296e-01 3.71395200e-01 3.83027405e-01
-2.02972084e-01 -5.72143137e-01 -2.37083763e-01 5.41727729e-02
-5.83284199e-01 -7.10689425e-02 4.69090641e-01 9.70799923e-01
-3.70392382e-01 6.04664803e-01 -9.81287122e-01 -2.34682307e-01
6.31499112e-01 7.17377961e-01 -1.91202238e-01 3.04735512e-01
4.92336184e-01 1.18379605e+00 -3.28051090e-01 2.98292577e-01
6.39601290e-01 -1.34497508e-01 1.46625087e-01 -4.78389055e-01
-5.57515383e-01 -1.90793067e-01 -1.56248748e+00 1.76974666e+00
-1.05631483e+00 6.35905683e-01 3.70808721e-01 -9.04825389e-01
7.14668572e-01 7.15746582e-01 6.59390569e-01 -6.30148828e-01
5.17584205e-01 5.90326905e-01 4.78895336e-01 -8.60419452e-01
9.35209021e-02 -1.11456536e-01 1.57471567e-01 3.07955563e-01
-2.49079466e-01 -3.19419831e-01 3.83818559e-02 -2.71792054e-01
8.44320714e-01 -8.39170516e-02 4.66662347e-01 -1.18013419e-01
7.20917583e-01 -3.10281485e-01 4.07478996e-02 7.51492858e-01
1.60404354e-01 8.33621562e-01 1.48589939e-01 -3.86122733e-01
-8.64585161e-01 -6.08931661e-01 -5.85475445e-01 -2.11537942e-01
3.89823139e-01 3.07980269e-01 -6.95822179e-01 -3.13703835e-01
-2.11702883e-01 5.16824067e-01 -5.11366308e-01 -6.93755150e-02
-8.70817006e-01 -5.98617792e-01 -2.66051404e-02 2.12917790e-01
-9.81587172e-02 -7.81766653e-01 -1.50676680e+00 5.59309304e-01
7.86422864e-02 -9.44412708e-01 1.98164731e-01 1.63301855e-01
-9.44068789e-01 -1.17105663e+00 -1.16675699e+00 -4.44225580e-01
9.98679817e-01 -1.05886534e-01 6.64617538e-01 -1.97292611e-01
-5.08980453e-01 4.37320679e-01 -1.94208056e-01 -3.49907309e-01
-6.57812595e-01 -3.23581725e-01 -1.97284091e-02 -9.53575000e-02
-3.38515222e-01 -3.45995784e-01 -9.54390824e-01 4.24264729e-01
-8.45733881e-01 -6.79424554e-02 9.24290538e-01 1.05041027e+00
2.09550530e-01 -1.21677130e-01 3.07234108e-01 -5.69015563e-01
4.24045473e-01 -5.53582758e-02 -7.55066156e-01 -1.36802867e-01
-1.77060217e-01 -1.77842438e-01 4.03764755e-01 -4.77185607e-01
-1.00463223e+00 8.79143775e-02 -2.77757078e-01 -4.04454023e-01
-4.81895953e-01 9.16292906e-01 6.13970578e-01 -4.09846753e-01
9.59230840e-01 1.50604457e-01 4.27416563e-01 -1.93639815e-01
-8.01348984e-02 6.77546680e-01 7.05325067e-01 -5.98559231e-02
4.91223902e-01 4.46190029e-01 5.39517522e-01 -1.00222039e+00
-3.17691088e-01 -7.03724742e-01 -8.53211701e-01 -4.85768437e-01
5.52981734e-01 -3.74850541e-01 -8.98647368e-01 4.00899202e-01
-1.04897559e+00 -2.88118655e-03 -1.04820050e-01 1.36741757e+00
-5.61951995e-01 4.47171718e-01 -3.31650078e-01 -8.44061792e-01
-3.03284943e-01 -1.49939108e+00 1.05483341e+00 7.75374249e-02
-2.08997041e-01 -1.10097992e+00 -2.07105532e-01 -1.31804377e-01
6.62811577e-01 4.58033860e-01 4.21392053e-01 -1.79351032e-01
-4.74412411e-01 -9.09936905e-01 6.52910918e-02 4.04701009e-02
3.87246609e-01 -2.60330439e-01 -6.59289420e-01 -5.54269373e-01
4.90690470e-01 1.90529794e-01 1.40271366e-01 1.17882228e+00
8.17566395e-01 6.98599545e-03 -6.44834638e-01 3.89572084e-01
1.51749659e+00 4.15802121e-01 3.62435877e-01 2.95487165e-01
3.63822192e-01 4.55756962e-01 9.34088171e-01 4.55061704e-01
-2.98937917e-01 7.87982762e-01 7.67787933e-01 -1.11873940e-01
9.26044583e-02 6.35711670e-01 -5.64792633e-01 7.52260327e-01
-2.00725377e-01 -1.15275577e-01 -8.84616792e-01 5.38670063e-01
-1.44343519e+00 -4.72584605e-01 -2.35516056e-01 2.80941510e+00
6.26267910e-01 1.10623367e-01 -4.43339467e-01 2.79097110e-01
4.24399853e-01 -3.30823839e-01 -2.16008648e-01 -1.99312642e-01
5.14549613e-01 2.20905259e-01 6.64785683e-01 7.03006148e-01
-1.10774016e+00 -1.88637435e-01 4.91793060e+00 3.36738616e-01
-1.69168985e+00 -1.90663308e-01 2.11359367e-01 2.99356896e-02
2.30388954e-01 -4.47859049e-01 -2.56189972e-01 5.84214628e-01
8.63702059e-01 1.56964913e-01 -1.98364407e-01 3.69928807e-01
2.83130735e-01 -7.50500500e-01 -8.20558548e-01 1.07762849e+00
2.70776339e-02 -1.25252306e+00 -8.14259350e-01 4.43246365e-02
5.80087960e-01 -2.85654992e-01 -2.48862430e-03 -4.06127870e-01
-7.10319996e-01 -8.77004802e-01 3.63418698e-01 1.03562272e+00
8.18872035e-01 -6.61833346e-01 1.14282084e+00 8.15959811e-01
-5.84850490e-01 3.66647653e-02 -1.77060649e-01 2.66998231e-01
4.95176256e-01 8.74637604e-01 -1.32169378e+00 8.44113767e-01
3.30615282e-01 2.74342328e-01 1.56632904e-02 1.75613928e+00
-1.61314279e-01 2.50472099e-01 -5.73654294e-01 3.77684385e-02
3.29477936e-01 -2.45172352e-01 8.36180449e-01 1.02184784e+00
9.70163882e-01 4.56496291e-02 -3.00478101e-01 1.15000620e-01
6.09333336e-01 6.58625886e-02 -3.75962436e-01 4.19927925e-01
3.39656979e-01 1.28628600e+00 -7.32099235e-01 -4.21530344e-02
-5.57339340e-02 7.52053738e-01 -4.41882551e-01 2.20325604e-01
-4.82737631e-01 -4.81007546e-01 8.16138461e-02 4.00119245e-01
3.73327643e-01 -3.22714418e-01 6.72625452e-02 -4.71094936e-01
2.73901820e-01 -3.81463140e-01 -1.09047197e-01 -8.19114745e-01
-4.57625687e-01 6.79444849e-01 -2.64249537e-02 -1.67052996e+00
-1.09122562e+00 -7.58701801e-01 -2.80319214e-01 1.08810747e+00
-1.06364298e+00 -7.43297338e-01 -2.81547785e-01 -2.44607449e-01
1.68539345e-01 1.52238235e-01 1.09732330e+00 2.43297204e-01
8.75626728e-02 1.68621361e-01 6.74802721e-01 -1.33479759e-01
6.57286406e-01 -1.13337040e+00 -1.03451289e-01 5.71395993e-01
-2.93834418e-01 8.14402640e-01 9.85902429e-01 -4.11380827e-01
-1.41616213e+00 -6.59289479e-01 4.31079209e-01 -6.43202662e-02
4.16546583e-01 1.17586218e-01 -7.87181258e-01 3.07164460e-01
9.21095163e-02 4.47056770e-01 2.66705751e-01 -4.24000710e-01
6.26781762e-01 2.29875132e-01 -1.30227625e+00 3.22718561e-01
3.95935625e-01 -1.52358040e-02 -2.01852173e-01 2.92510748e-01
-5.45786507e-03 -1.41964781e+00 -1.07153893e+00 7.25616097e-01
9.15979505e-01 -1.18358874e+00 1.02707648e+00 1.84478149e-01
7.42873847e-02 -2.38897502e-01 6.05137765e-01 -1.62025726e+00
-5.81653491e-02 -5.79931259e-01 1.56934217e-01 2.37676069e-01
7.38758668e-02 -7.03759432e-01 9.47310269e-01 3.71874534e-02
-1.70280293e-01 -1.03760600e+00 -1.32022047e+00 -5.62122047e-01
-3.29678386e-01 -4.93034571e-01 -6.11643605e-02 5.98909140e-01
-1.84669103e-02 -1.44741893e-01 -1.59669027e-01 7.14344800e-01
6.77175343e-01 2.19140705e-02 4.75148469e-01 -1.24029326e+00
-3.45059752e-01 -1.32068887e-01 -7.80730307e-01 -1.06092608e+00
-3.78909647e-01 -3.85591149e-01 2.52100497e-01 -1.64079130e+00
-4.62340266e-01 -4.87223595e-01 6.06546253e-02 -3.77135336e-01
1.59096748e-01 4.18918788e-01 -2.20611840e-01 -8.01155418e-02
1.51697651e-01 -2.01652706e-01 1.46264195e+00 3.15694541e-01
-2.51025558e-01 6.73867762e-01 -6.53957501e-02 9.50096965e-01
5.03880024e-01 -2.69694954e-01 -4.73542929e-01 -3.84694606e-01
9.55639035e-03 5.92967868e-01 4.35507208e-01 -1.26192975e+00
4.86755431e-01 3.79713118e-01 6.26286030e-01 -6.75194979e-01
5.91028571e-01 -1.27843881e+00 4.85311002e-01 7.89791703e-01
-4.58683893e-02 -4.18940783e-01 3.56584311e-01 2.46909350e-01
-1.69699013e-01 -9.07004833e-01 8.18138957e-01 -6.79887086e-02
-2.85705984e-01 -1.69792533e-01 -4.12507892e-01 -4.97846067e-01
1.14833152e+00 -6.67087317e-01 2.88551658e-01 -4.83429313e-01
-1.20174921e+00 -3.72602224e-01 2.27104753e-01 -2.25519702e-01
7.02877283e-01 -1.07218611e+00 -5.89434206e-01 6.53898895e-01
2.98781786e-02 1.76958591e-01 6.61631942e-01 1.55777395e+00
-8.78140152e-01 6.69649720e-01 2.07330525e-01 -1.16215539e+00
-1.38598609e+00 2.92971432e-01 6.78332090e-01 -7.09988624e-02
-4.57205981e-01 9.05258238e-01 -1.01275042e-01 -2.42362782e-01
1.16858512e-01 -4.65570956e-01 -3.27906340e-01 -3.03941611e-02
4.78969425e-01 7.80211538e-02 5.66751301e-01 -7.78601944e-01
-1.96250141e-01 1.13486803e+00 1.06720202e-01 -1.18110083e-01
1.16421354e+00 -1.92629844e-01 2.48817950e-01 2.47822404e-01
1.06599021e+00 2.43505567e-01 -1.22051811e+00 -2.11480647e-01
-1.79450944e-01 -8.67986083e-01 2.99591094e-01 -8.61214578e-01
-7.11192429e-01 9.58985507e-01 9.03610229e-01 3.30166548e-01
1.11690903e+00 -1.99688658e-01 2.76536882e-01 -8.29525068e-02
3.56546849e-01 -4.45799053e-01 -2.05796361e-01 9.46138427e-02
1.03026628e+00 -9.18989360e-01 1.70687735e-01 -6.29025340e-01
-1.57232121e-01 1.36333048e+00 2.81846188e-02 1.32767767e-01
5.02173424e-01 4.22047764e-01 4.77398306e-01 -7.84552619e-02
1.05718456e-01 8.49227309e-02 5.69245219e-01 4.22317296e-01
5.60725689e-01 -2.87825227e-01 -5.75076938e-01 -3.74474488e-02
1.18288465e-01 1.02622934e-01 6.41084850e-01 1.33232927e+00
-3.33455950e-01 -7.52396584e-01 -7.34306633e-01 3.29891115e-01
-8.23594630e-01 1.94150358e-01 6.58663452e-01 8.81308913e-01
-4.65812534e-02 5.96783221e-01 1.60665721e-01 2.27418721e-01
7.36477375e-01 -4.70690876e-01 8.29502344e-01 -4.02046949e-01
-4.56644565e-01 6.24314904e-01 -5.14545990e-03 -4.97203141e-01
-3.82424891e-01 -5.83851933e-01 -1.24267721e+00 5.21014154e-01
-4.62324560e-01 2.63823211e-01 1.48485613e+00 7.45997369e-01
3.12301517e-01 8.25679719e-01 5.30956030e-01 -1.31088388e+00
-5.95542669e-01 -9.61329460e-01 -4.00817871e-01 -2.72978637e-02
5.87432623e-01 -6.43261731e-01 -3.34695071e-01 -1.55047268e-01]
|
[13.668266296386719, -2.9699528217315674]
|
ba6ad127-7145-4f79-8b79-0022c82050cd
|
active-self-training-for-weakly-supervised-3d
|
2209.07069
| null |
https://arxiv.org/abs/2209.07069v1
|
https://arxiv.org/pdf/2209.07069v1.pdf
|
Active Self-Training for Weakly Supervised 3D Scene Semantic Segmentation
|
Since the preparation of labeled data for training semantic segmentation networks of point clouds is a time-consuming process, weakly supervised approaches have been introduced to learn from only a small fraction of data. These methods are typically based on learning with contrastive losses while automatically deriving per-point pseudo-labels from a sparse set of user-annotated labels. In this paper, our key observation is that the selection of what samples to annotate is as important as how these samples are used for training. Thus, we introduce a method for weakly supervised segmentation of 3D scenes that combines self-training with active learning. The active learning selects points for annotation that likely result in performance improvements to the trained model, while the self-training makes efficient use of the user-provided labels for learning the model. We demonstrate that our approach leads to an effective method that provides improvements in scene segmentation over previous works and baselines, while requiring only a small number of user annotations.
|
['Ruizhen Hu', 'Hui Huang', 'Oliver van Kaick', 'Gengxin Liu']
|
2022-09-15
| null | null | null | null |
['scene-segmentation']
|
['computer-vision']
|
[ 4.07742888e-01 5.54894567e-01 -4.58113611e-01 -7.83674657e-01
-1.25600386e+00 -7.17149854e-01 4.38880920e-01 6.26288652e-01
-7.30132163e-01 4.41745937e-01 -2.40501240e-01 -1.74519762e-01
2.82735735e-01 -7.05774009e-01 -1.05343497e+00 -6.40500188e-01
-2.37603858e-02 1.06398666e+00 6.55841589e-01 2.19686657e-01
1.32958621e-01 6.70531690e-01 -1.56715834e+00 -9.31413285e-03
9.61016119e-01 1.01545048e+00 4.02870983e-01 4.85712260e-01
-6.12936676e-01 7.04626918e-01 -3.64045322e-01 -1.20188646e-01
5.25088310e-01 -1.38532445e-01 -1.10271406e+00 4.75136369e-01
5.67546487e-01 -3.91938537e-01 4.44619298e-01 9.35783505e-01
2.43620262e-01 3.00919056e-01 7.27628350e-01 -1.00479257e+00
1.80289119e-01 3.79098952e-01 -4.72389430e-01 -1.99882790e-01
3.71908327e-03 7.75570124e-02 1.13148987e+00 -9.45575595e-01
5.26953518e-01 9.32967126e-01 7.11119473e-01 5.93414843e-01
-1.40614057e+00 -4.42293853e-01 3.77432287e-01 -2.73557693e-01
-1.17551732e+00 -5.92686355e-01 9.49583352e-01 -5.77124000e-01
7.33874798e-01 1.51338890e-01 6.97381794e-01 5.62159657e-01
-7.83972621e-01 1.12415481e+00 7.61213362e-01 -6.77509248e-01
6.24525428e-01 4.42748696e-01 3.54983181e-01 7.52850413e-01
5.89481518e-02 -2.23041192e-01 -3.08913320e-01 -3.05964440e-01
8.11261475e-01 -1.23146184e-01 1.27959862e-01 -9.75888371e-01
-7.78928280e-01 8.97710919e-01 4.45321918e-01 -5.69127500e-02
-3.08694094e-01 7.08239675e-02 2.39168152e-01 -5.80346063e-02
1.01039910e+00 4.58312422e-01 -6.97604537e-01 1.75725650e-02
-1.37276125e+00 1.23111822e-01 6.98431790e-01 1.02038944e+00
1.40966570e+00 -4.72886682e-01 1.23700656e-01 9.71124709e-01
3.94918382e-01 3.77749652e-01 -2.51982689e-01 -1.21275473e+00
3.01545173e-01 9.70498860e-01 3.31749737e-01 -5.63695788e-01
-1.58032447e-01 -2.03920141e-01 -7.64010102e-02 4.11566198e-01
5.31017065e-01 -1.16358340e-01 -1.32833636e+00 1.54968226e+00
6.47368729e-01 1.92602769e-01 -2.91149765e-01 7.09332049e-01
4.96921808e-01 5.32392561e-01 2.70022959e-01 -1.48595095e-01
6.79727256e-01 -1.06184375e+00 -2.22942486e-01 -4.01634008e-01
7.58587241e-01 -6.96594536e-01 1.22101545e+00 8.45455527e-02
-1.20456159e+00 -5.35547972e-01 -7.92833030e-01 -1.20104373e-01
-1.56193390e-01 -2.53784452e-02 6.99149907e-01 4.50485796e-01
-1.00519824e+00 7.32668102e-01 -1.25029528e+00 -3.28976184e-01
1.08346570e+00 6.06179893e-01 -4.91970927e-02 1.26116037e-01
-4.42458391e-01 6.36811554e-01 4.18267936e-01 -2.02870399e-01
-9.42305326e-01 -6.87933981e-01 -7.25118995e-01 -1.21596389e-01
6.59806907e-01 -4.28772360e-01 1.45226550e+00 -1.37840986e+00
-1.43853557e+00 1.27711165e+00 -2.49778315e-01 -4.65566576e-01
6.72018826e-01 -3.80067766e-01 5.52810431e-01 3.81362945e-01
2.42326275e-01 1.24291646e+00 6.89102709e-01 -1.73549938e+00
-8.68798375e-01 -3.22604209e-01 2.68157572e-01 4.02231336e-01
-9.34108496e-02 -1.39292389e-01 -8.48999679e-01 -1.89219326e-01
3.62182140e-01 -1.07588017e+00 -6.21629775e-01 4.08235312e-01
-3.82784069e-01 -4.23794925e-01 9.36889946e-01 -1.62298694e-01
4.77248818e-01 -2.04756832e+00 -1.94291964e-01 3.35015297e-01
2.18089134e-01 2.13776201e-01 9.29596350e-02 2.81461719e-02
2.11630717e-01 1.20889306e-01 -7.54514873e-01 -8.50620925e-01
-9.33095142e-02 3.43486607e-01 -2.88008332e-01 3.77182245e-01
3.56533021e-01 8.13047051e-01 -1.04941392e+00 -8.31771076e-01
4.94968355e-01 1.62295654e-01 -6.38413548e-01 5.18007040e-01
-7.18051136e-01 8.13433647e-01 -5.00466585e-01 5.16319454e-01
6.68976843e-01 -4.55919445e-01 -2.54320323e-01 6.74775690e-02
-4.87948321e-02 4.86130089e-01 -1.06245279e+00 1.92272890e+00
-3.72068167e-01 1.84333399e-01 2.05345109e-01 -1.09475553e+00
6.56853139e-01 1.98121235e-01 8.61079037e-01 -1.56122342e-01
1.74921080e-01 2.99984992e-01 -4.58357722e-01 -4.24076825e-01
2.32905045e-01 -4.29620147e-02 1.84356332e-01 5.37255526e-01
1.31243035e-01 -5.72688162e-01 2.80854881e-01 1.74515888e-01
9.70055461e-01 6.14488661e-01 4.56881262e-02 -5.66141866e-02
2.19831213e-01 5.54336071e-01 6.78423166e-01 8.38466644e-01
-3.36679876e-01 6.39210105e-01 2.24306211e-01 -2.53356516e-01
-1.13041139e+00 -9.24498081e-01 -1.61098719e-01 1.09318829e+00
2.58201569e-01 -1.35316551e-01 -9.78343546e-01 -1.21427369e+00
-2.82502621e-01 6.78348541e-01 -3.24528813e-01 1.94855928e-01
-5.65100014e-01 -4.95188743e-01 1.84034139e-01 5.31387866e-01
3.08016747e-01 -1.06847072e+00 -6.60081685e-01 2.49067508e-02
1.31946048e-02 -1.07276177e+00 -1.45191431e-01 6.13272786e-01
-1.31813955e+00 -1.03116047e+00 -6.13529801e-01 -9.05477405e-01
1.22037065e+00 2.39735112e-01 1.35755897e+00 1.60100833e-01
5.88148236e-02 3.44643891e-01 -4.05249953e-01 -5.95803857e-01
-2.43030265e-01 3.07853192e-01 -3.42044622e-01 -1.84813857e-01
4.70595449e-01 -4.47272658e-01 -4.76540416e-01 1.35159373e-01
-6.24114335e-01 2.86461174e-01 4.60231513e-01 4.51391727e-01
9.68279004e-01 -1.69013619e-01 2.09971219e-01 -1.66978478e+00
-6.19291775e-02 -2.41954222e-01 -8.03244412e-01 -5.47793619e-02
-4.38114464e-01 4.20637280e-02 3.82894397e-01 -2.17283994e-01
-1.16338778e+00 8.98168623e-01 -3.18663657e-01 -4.50178415e-01
-6.58194363e-01 2.00062722e-01 -1.59669608e-01 -2.33423635e-01
7.16164291e-01 -4.24598962e-01 -1.63261458e-01 -5.40982246e-01
4.66437697e-01 5.26937962e-01 2.61312962e-01 -7.63827205e-01
8.97772014e-01 6.89345658e-01 -1.65411308e-01 -6.02920830e-01
-1.29726362e+00 -8.97147954e-01 -1.16157830e+00 -2.03441247e-01
7.72993624e-01 -8.46629679e-01 -1.20298058e-01 2.97959149e-01
-1.06637776e+00 -8.24893713e-01 -8.06347668e-01 3.86322021e-01
-7.15462148e-01 2.62030691e-01 -3.84481966e-01 -1.03925979e+00
-7.28214160e-02 -1.16754198e+00 1.28752410e+00 4.30706628e-02
-3.10097426e-01 -1.06366408e+00 1.34127229e-01 5.43806374e-01
-1.32135423e-02 1.74969852e-01 7.63892949e-01 -7.47750938e-01
-9.23887968e-01 -2.63926685e-01 -1.47387296e-01 5.03032506e-01
1.65144429e-01 -7.13209808e-02 -1.40837264e+00 -2.01055124e-01
-5.31585291e-02 -7.40505457e-01 8.21781814e-01 4.25151318e-01
1.35144496e+00 4.83430959e-02 -4.17071044e-01 6.11711204e-01
1.31899142e+00 -1.78798884e-02 3.25498044e-01 7.65257627e-02
8.71589243e-01 8.04596007e-01 8.30290020e-01 -3.61016691e-02
1.87269345e-01 3.63104820e-01 4.64232355e-01 -5.17613649e-01
1.24701656e-01 -2.71188378e-01 -1.47431001e-01 6.01506293e-01
2.19581765e-03 -3.58291492e-02 -1.13925529e+00 6.74677193e-01
-1.94143391e+00 -4.89290774e-01 -5.47310486e-02 2.39305067e+00
1.10793829e+00 5.60157597e-01 1.62937179e-01 1.29465982e-01
6.37337148e-01 -2.98262239e-02 -7.69179642e-01 2.62304336e-01
2.80648559e-01 4.36026156e-01 7.93618858e-01 7.08307087e-01
-1.40878725e+00 1.16536176e+00 6.22737694e+00 6.94437325e-01
-1.08676493e+00 1.87704429e-01 7.23854423e-01 3.84775177e-02
-1.32796362e-01 3.65908831e-01 -6.87354982e-01 2.90005058e-01
6.58860862e-01 4.18108314e-01 -3.79031412e-02 1.16048443e+00
3.52780521e-01 -3.14675719e-01 -1.24324560e+00 8.46755028e-01
-2.03777239e-01 -1.16240060e+00 -1.80628672e-01 -1.29108503e-02
9.45414841e-01 2.63067424e-01 -3.40381950e-01 -4.92358468e-02
6.64070725e-01 -7.04566360e-01 8.39307547e-01 3.10338855e-01
6.31797969e-01 -6.31494462e-01 5.29704869e-01 7.39999115e-01
-8.90451312e-01 2.27634966e-01 -4.09804910e-01 -4.01780196e-02
2.02930227e-01 9.45621014e-01 -1.11035919e+00 1.35507241e-01
6.37123227e-01 6.76391840e-01 -4.87058640e-01 1.18631554e+00
-4.22960997e-01 1.17513847e+00 -6.77056134e-01 2.10124433e-01
2.33214930e-01 -2.12753057e-01 4.28215861e-01 1.03911197e+00
-1.26963526e-01 -4.85737845e-02 7.53496885e-01 9.89963710e-01
-1.71436369e-01 2.12516785e-01 -3.81040722e-01 2.86042802e-02
4.62035179e-01 1.32652771e+00 -1.36074913e+00 -3.16925019e-01
-2.13113591e-01 8.32354546e-01 5.36376297e-01 2.89455503e-01
-4.22072411e-01 -8.64746496e-02 6.70280829e-02 3.57352942e-01
2.72590131e-01 -3.06477845e-01 -6.90569520e-01 -8.51679325e-01
-7.85242319e-02 -2.93510795e-01 8.87830257e-02 -5.91352284e-01
-1.28293967e+00 2.05674261e-01 1.20349757e-01 -1.21025586e+00
-1.10021085e-01 -3.75103742e-01 -5.48507392e-01 8.54933202e-01
-1.50247240e+00 -1.19169581e+00 -4.02221680e-01 7.91260526e-02
5.96458733e-01 1.84601128e-01 7.14866221e-01 2.44680628e-01
-2.86196142e-01 1.42073721e-01 -2.06949323e-01 3.24878991e-01
4.93925005e-01 -1.64315999e+00 4.84461129e-01 7.90373504e-01
3.22098136e-01 5.23226082e-01 5.81611574e-01 -6.10361576e-01
-5.77432454e-01 -1.11544347e+00 7.26396024e-01 -5.55220246e-01
1.31232753e-01 -6.17710292e-01 -9.05734062e-01 7.39623249e-01
-1.85825810e-01 2.82542020e-01 6.85457230e-01 1.15392625e-01
4.38495353e-02 3.24090198e-02 -1.31734836e+00 3.66419941e-01
1.02281797e+00 -3.76634806e-01 -4.79906052e-01 3.56450200e-01
6.70253873e-01 -4.63332564e-01 -3.74986976e-01 4.09815758e-01
-7.86213800e-02 -7.94111490e-01 9.69724536e-01 -2.68939793e-01
2.17153519e-01 -3.66277844e-01 1.64367795e-01 -1.04623294e+00
-7.88393989e-03 -3.53948504e-01 1.73848778e-01 1.31439579e+00
6.27587855e-01 -2.51842171e-01 1.50478470e+00 8.52443695e-01
-3.45439881e-01 -6.48233891e-01 -5.79975963e-01 -4.58738863e-01
-2.96284650e-02 -5.56634903e-01 2.31807694e-01 1.00775468e+00
-5.23705423e-01 4.31809276e-01 1.42837375e-01 8.42615888e-02
1.04705298e+00 -3.45320180e-02 9.67471004e-01 -1.66023993e+00
-3.44125703e-02 -1.63229611e-02 -1.93156883e-01 -1.38672674e+00
3.18252534e-01 -9.03034925e-01 5.73762238e-01 -1.72217536e+00
1.27623558e-01 -1.15507102e+00 -6.80268630e-02 7.63473272e-01
-2.06853136e-01 3.73461097e-01 -1.21532060e-01 4.75788891e-01
-8.12590420e-01 2.70466864e-01 1.10640717e+00 -6.24091960e-02
-5.12712061e-01 3.81183475e-01 -2.71431923e-01 1.10565794e+00
6.57758474e-01 -7.53759921e-01 -7.56941259e-01 -3.86022627e-01
1.20492265e-01 -3.48770291e-01 3.19664121e-01 -8.35642099e-01
2.39398584e-01 -2.58714527e-01 2.13269860e-01 -8.26587617e-01
4.00093049e-01 -9.57343400e-01 -2.47850671e-01 7.50122592e-03
-6.24886811e-01 -5.71783364e-01 5.10978736e-02 4.96781141e-01
-7.44510144e-02 -6.44764066e-01 1.13739181e+00 -6.47854447e-01
-7.10146129e-01 4.53286350e-01 -3.26505862e-02 3.21280956e-01
8.34217787e-01 -4.48343247e-01 3.04830134e-01 -1.40215248e-01
-8.09407890e-01 2.17129380e-01 8.50014269e-01 -1.42587334e-01
2.64323503e-01 -8.63745093e-01 -3.39740604e-01 -2.49442235e-02
7.86820427e-02 1.09470415e+00 -2.29000524e-01 4.57005352e-01
-7.17175245e-01 3.19683999e-02 1.49237573e-01 -1.04925692e+00
-1.15712249e+00 2.06841528e-01 1.88276127e-01 -1.12110257e-01
-6.67631567e-01 1.12969327e+00 1.76307425e-01 -8.15204561e-01
5.89463472e-01 -2.31994763e-01 -2.43010130e-02 2.16044113e-02
-2.00023558e-02 2.09871903e-01 2.17825398e-01 -5.03549218e-01
-1.79003894e-01 5.44590652e-01 -8.96868110e-02 -2.68380374e-01
1.55427849e+00 -1.12197995e-01 -8.16867724e-02 8.82970214e-01
1.13776648e+00 5.54353707e-02 -1.75606370e+00 -3.89042974e-01
3.09016794e-01 -4.90206480e-01 2.09155619e-01 -7.00729191e-01
-9.59769666e-01 8.51726353e-01 5.37393570e-01 1.60444111e-01
8.47869396e-01 2.75894761e-01 6.83869660e-01 5.28038323e-01
5.59818089e-01 -1.33057761e+00 1.12270698e-01 3.92871678e-01
1.89884081e-01 -1.59858012e+00 -3.65044968e-03 -8.79570782e-01
-3.91447604e-01 7.58090019e-01 6.43925309e-01 -2.97293931e-01
6.47485793e-01 3.00913930e-01 4.58155721e-01 -3.02031606e-01
-3.44967425e-01 -4.98689026e-01 2.42742732e-01 5.94301939e-01
3.27644646e-01 -2.09087387e-01 6.66274950e-02 4.11524959e-02
-6.25098199e-02 -2.48337612e-02 1.51249051e-01 1.17643213e+00
-6.97914779e-01 -1.24831355e+00 -2.30995551e-01 5.00814080e-01
-2.48082265e-01 1.20223589e-01 -7.04902530e-01 4.81871277e-01
3.39773208e-01 8.37333083e-01 3.02549571e-01 9.89321992e-02
2.05744222e-01 1.97391450e-01 4.94665533e-01 -1.27185285e+00
-3.78638357e-01 2.18835324e-01 3.76227498e-02 -3.46369326e-01
-9.22524869e-01 -6.14240527e-01 -1.58449888e+00 2.78494298e-01
-6.53457761e-01 3.75645697e-01 6.23898089e-01 1.14100254e+00
1.37521401e-01 1.56022042e-01 6.80014431e-01 -1.15993285e+00
-1.67977259e-01 -6.75666273e-01 -4.16314900e-01 6.69216096e-01
2.87156463e-01 -7.10368872e-01 -5.05757093e-01 4.04541731e-01]
|
[9.469837188720703, 0.5569085478782654]
|
6c5e25c0-324a-4953-8ba4-daf46e5e95a0
|
why-should-i-trust-you-bellman-evaluating-the
| null | null |
https://openreview.net/forum?id=MUpxS9vDbZr
|
https://openreview.net/pdf?id=MUpxS9vDbZr
|
Why Should I Trust You, Bellman? Evaluating the Bellman Objective with Off-Policy Data
|
In this work, we analyze the effectiveness of the Bellman equation as a proxy objective for value prediction accuracy in off-policy evaluation. While the Bellman equation is uniquely solved by the true value function over all state-action pairs, we show that in the finite data regime, the Bellman equation can be satisfied exactly by infinitely many suboptimal solutions. This eliminates any guarantees relating Bellman error to the accuracy of the value function. We find this observation extends to practical settings; when computed over an off-policy dataset, the Bellman error bears little relationship to the accuracy of the value function. Consequently, we show that the Bellman error is a poor metric for comparing value functions, and therefore, an ineffective objective for off-policy evaluation. Finally, we discuss differences between Bellman error and the non-stationary objective used by iterative methods and deep reinforcement learning, and highlight how the effectiveness of this objective relies on generalization during training.
|
['Shixiang Shane Gu', 'Ofir Nachum', 'Doina Precup', 'David Meger', 'Scott Fujimoto']
|
2021-09-29
| null | null | null | null |
['value-prediction']
|
['computer-code']
|
[-1.43046156e-01 1.57232642e-01 -8.39569330e-01 -8.94240141e-02
-9.20815945e-01 -8.36720109e-01 4.40440774e-01 1.07177988e-01
-7.97405601e-01 1.32410586e+00 -8.82480368e-02 -6.65229261e-01
-5.36654353e-01 -4.79385465e-01 -6.10406697e-01 -7.46099234e-01
-1.36044413e-01 3.45402390e-01 -3.13466117e-02 -2.42925629e-01
4.56107885e-01 5.52572966e-01 -1.25225842e+00 -3.10247272e-01
5.68466127e-01 1.37289977e+00 -2.26918370e-01 7.68745482e-01
1.62315980e-01 8.92490149e-01 -9.38831329e-01 -1.53778389e-01
5.01558185e-01 -6.21121228e-01 -8.65210712e-01 -2.77563274e-01
3.95239204e-01 -7.23482668e-01 -3.74568105e-01 1.19502580e+00
4.56263065e-01 4.05359089e-01 4.74825054e-01 -1.47809434e+00
-3.13694268e-01 3.86355877e-01 -1.24250818e-02 2.61715889e-01
2.38012984e-01 3.51605326e-01 1.23901117e+00 -6.13395646e-02
5.38125396e-01 9.87786174e-01 8.06798041e-01 5.31729698e-01
-1.16228366e+00 -2.72787660e-01 1.87356219e-01 6.33003935e-02
-1.05683291e+00 -4.25258160e-01 3.21598947e-01 -2.84801483e-01
9.31395829e-01 1.36226237e-01 8.60027432e-01 8.79993916e-01
4.23342913e-01 4.95172054e-01 1.15245223e+00 -2.35395238e-01
5.90905786e-01 7.33229816e-02 2.16569137e-02 6.46207452e-01
2.88720101e-01 8.41467023e-01 -8.10998753e-02 -1.64654568e-01
1.09212208e+00 -2.99542248e-01 -3.30534190e-01 -4.95620966e-01
-7.74044394e-01 8.08585703e-01 1.29990324e-01 1.44623667e-01
-4.13874984e-01 7.13601708e-01 3.55259538e-01 7.11838961e-01
-1.69708952e-02 9.18121040e-01 -6.27962649e-01 -7.96305358e-01
-7.05960453e-01 6.53230846e-01 9.25658286e-01 5.93962908e-01
3.74305457e-01 2.34643444e-01 -4.55490708e-01 1.89772546e-01
-1.10969581e-01 6.09966755e-01 3.89749974e-01 -1.69841301e+00
4.13745176e-03 8.20442289e-02 9.12585437e-01 -5.08210182e-01
-5.51937699e-01 -6.14842713e-01 1.56532638e-02 5.87768018e-01
1.18781197e+00 -6.58612907e-01 -4.80760187e-01 2.03029823e+00
2.57154882e-01 -2.24922240e-01 2.74672627e-01 8.85896802e-01
2.96048383e-04 4.08136040e-01 -1.77269712e-01 -6.34838700e-01
9.10128176e-01 -6.65618539e-01 -8.64301980e-01 6.65321723e-02
6.48097754e-01 -4.83865559e-01 1.05983949e+00 3.01001817e-01
-1.29248691e+00 -1.85057417e-01 -9.71431613e-01 2.77948141e-01
-7.28270113e-02 -3.91114652e-01 5.87143004e-01 4.99446511e-01
-9.45776880e-01 1.18648958e+00 -6.86165631e-01 -1.89392671e-01
1.54442698e-01 4.74104345e-01 9.13229063e-02 5.49366117e-01
-1.19550514e+00 1.53887224e+00 3.95180225e-01 -2.92917818e-01
-1.07373738e+00 -6.09629571e-01 -3.96711797e-01 3.33427280e-01
6.67983353e-01 -2.09714293e-01 1.90984154e+00 -1.21810973e+00
-1.86100733e+00 3.01894754e-01 1.38645545e-01 -7.50068605e-01
7.26751447e-01 -1.03119485e-01 -1.92383617e-01 1.12657584e-01
-2.85125136e-01 3.25467825e-01 6.96044922e-01 -1.09212613e+00
-7.13899136e-01 -2.20234811e-01 7.48867869e-01 1.25343099e-01
1.24566175e-01 -3.28425407e-01 2.67866731e-01 -2.17392653e-01
-3.01896274e-01 -9.01505828e-01 -8.89616460e-02 -1.33525580e-01
2.53109429e-02 -3.27734917e-01 5.62737763e-01 -5.04739285e-01
1.37728775e+00 -1.88511443e+00 -3.49193931e-01 2.96197802e-01
-2.02411264e-01 2.54627883e-01 -1.88622668e-01 4.49830353e-01
1.89617407e-02 1.96780577e-01 -1.13668434e-01 2.06455365e-01
3.87708724e-01 4.40642297e-01 -3.77773792e-01 6.65099800e-01
5.27804829e-02 1.03584623e+00 -1.18840992e+00 -3.51212204e-01
1.21390268e-01 7.06025884e-02 -6.64145291e-01 1.08682491e-01
-3.27260405e-01 4.60958689e-01 -3.78353029e-01 1.93984941e-01
2.36427665e-01 -1.78121895e-01 4.90249783e-01 2.41258919e-01
-4.03295755e-01 4.63700593e-01 -1.01103926e+00 1.26585102e+00
-3.90690714e-01 5.63838005e-01 -4.75301445e-02 -1.23175263e+00
6.57415390e-01 2.99564928e-01 8.44289243e-01 -9.19774532e-01
3.61687809e-01 3.27829540e-01 2.25670949e-01 -3.21060300e-01
5.05703568e-01 -5.91638446e-01 1.07030526e-01 6.05631649e-01
3.24934609e-02 -1.16823860e-01 1.93848342e-01 -2.96488345e-01
9.51652706e-01 4.76357102e-01 3.89893234e-01 -5.82184494e-01
8.48709792e-02 2.12887511e-01 5.32278895e-01 1.09251201e+00
-5.63194633e-01 1.29184797e-01 1.10896373e+00 -3.65261257e-01
-1.11039126e+00 -9.48079050e-01 -4.37675953e-01 9.78760600e-01
2.36994341e-01 -1.59660086e-01 -7.82833338e-01 -9.13967967e-01
3.34643900e-01 8.23015511e-01 -7.20322371e-01 -4.43152249e-01
-3.76836300e-01 -3.67477626e-01 5.87971747e-01 4.99413788e-01
2.66501278e-01 -8.14316213e-01 -1.07044637e+00 2.70967156e-01
1.56591341e-01 -8.71851683e-01 -4.58549410e-01 2.68045783e-01
-7.89336324e-01 -1.07149422e+00 -4.88907278e-01 -6.72675967e-02
3.14237356e-01 -2.47520521e-01 1.07601142e+00 1.29593790e-01
4.42918152e-01 6.78454041e-01 -8.67474973e-02 -3.13728213e-01
-6.29687130e-01 -1.61992401e-01 1.07899286e-01 -3.64588648e-01
1.42712176e-01 -4.51710790e-01 -6.92414165e-01 3.22685093e-01
-4.31090981e-01 -5.50776482e-01 2.55662054e-01 9.32861090e-01
5.54470420e-01 5.75173795e-02 6.61375225e-01 -3.78327399e-01
7.21827924e-01 -1.46349758e-01 -1.29438686e+00 3.23297054e-01
-1.04436970e+00 6.54965341e-01 8.99627864e-01 -4.83870894e-01
-7.27273464e-01 -2.41238669e-01 4.16101664e-02 -4.21988964e-01
3.92898798e-01 3.73012125e-01 3.68631124e-01 -1.07924506e-01
4.79884714e-01 -6.08389676e-02 4.03107673e-01 -4.36572760e-01
1.41853034e-01 2.93479085e-01 4.61374760e-01 -9.22841489e-01
1.55654088e-01 2.54067898e-01 3.91297877e-01 -4.15337741e-01
-1.17067122e+00 -1.12450413e-01 -1.54393315e-01 -2.11948916e-01
6.26356959e-01 -3.11476141e-01 -1.55436146e+00 -3.74725237e-02
-9.87502158e-01 -7.40755856e-01 -9.48812425e-01 6.39576316e-01
-1.13399327e+00 1.99428380e-01 -4.33690727e-01 -1.35657060e+00
1.42668355e-02 -1.18473804e+00 7.15731263e-01 1.18366048e-01
-1.76214591e-01 -1.19635367e+00 1.67185500e-01 -1.76698551e-01
3.10270816e-01 2.65789270e-01 7.42270410e-01 -4.28774059e-01
-1.78891867e-01 -2.44917572e-01 -3.84720191e-02 5.17503619e-01
-1.53398558e-01 -1.03576981e-01 -7.79914796e-01 -4.62111145e-01
6.35211393e-02 -3.79795909e-01 5.88073015e-01 6.26703739e-01
9.94332135e-01 -6.02734268e-01 2.26257324e-01 3.61854970e-01
1.47078550e+00 5.22722185e-01 4.74337935e-01 5.27410746e-01
1.76666096e-01 3.05801392e-01 6.81913495e-01 8.67128253e-01
8.99308100e-02 6.74001038e-01 4.68733817e-01 1.69655055e-01
2.57767022e-01 -3.68832290e-01 6.82688057e-01 1.37499690e-01
-2.78110981e-01 -4.36518062e-03 -6.35641336e-01 1.31280467e-01
-2.05719519e+00 -1.43776929e+00 3.15574288e-01 2.80867863e+00
9.96873617e-01 2.40774825e-01 5.72458267e-01 -8.95836353e-02
4.41825658e-01 4.16677259e-02 -1.02302885e+00 -8.13523710e-01
1.27345026e-01 3.45548391e-01 1.06737769e+00 8.51416349e-01
-7.86349654e-01 6.99396193e-01 8.20955372e+00 7.10652828e-01
-1.22636986e+00 2.22407743e-01 3.48514736e-01 -1.72600150e-01
-1.97687626e-01 1.89067453e-01 -5.56889474e-01 5.68637490e-01
1.15954030e+00 -5.11162579e-01 8.73813510e-01 8.47229004e-01
4.44728315e-01 -2.62769192e-01 -1.17978883e+00 6.49569452e-01
-6.03700042e-01 -1.09228933e+00 -5.60236037e-01 2.89448321e-01
8.69266748e-01 -1.88613966e-01 1.31272390e-01 5.30051470e-01
5.45042813e-01 -1.07769775e+00 9.39602971e-01 3.52424741e-01
7.23778367e-01 -8.73250067e-01 6.07723415e-01 2.73737162e-01
-6.40238345e-01 -3.90833616e-01 -3.02048892e-01 -4.17946965e-01
-8.10802355e-02 1.67739257e-01 -5.30532956e-01 1.91725031e-01
-2.88959593e-02 2.74869680e-01 1.01336375e-01 8.18430901e-01
-5.91556057e-02 2.91960269e-01 -4.04113770e-01 -2.57981956e-01
7.24127412e-01 -2.63208896e-01 4.14718479e-01 8.07762325e-01
1.63639754e-01 8.19020420e-02 2.49420449e-01 8.06333840e-01
2.08471715e-01 -2.58294761e-01 -3.36614460e-01 -3.85713071e-01
3.52493286e-01 6.46425664e-01 -3.37570071e-01 -3.11867326e-01
-3.12304050e-01 4.46657300e-01 3.98026973e-01 6.65350616e-01
-9.06852424e-01 -1.33539140e-01 1.03050125e+00 -1.89098325e-02
2.85257280e-01 -2.42302403e-01 -2.71942973e-01 -9.83318985e-01
-6.52245283e-02 -8.92995954e-01 3.81975532e-01 -3.39344889e-01
-8.13243866e-01 -8.22546892e-03 1.45179585e-01 -1.25164366e+00
-6.25613391e-01 -8.41954768e-01 -2.77174592e-01 5.27877629e-01
-1.42843807e+00 -1.74292669e-01 3.65176499e-01 2.96788365e-01
5.54410741e-03 2.06146806e-01 6.75380588e-01 -3.00573446e-02
-4.30683404e-01 6.70889437e-01 5.66878438e-01 -3.17013822e-02
2.13589504e-01 -1.35679662e+00 -1.84952915e-01 5.67329288e-01
-2.25464627e-01 5.07547379e-01 1.01893520e+00 -2.62232989e-01
-1.57087314e+00 -5.54603279e-01 2.22080305e-01 -4.85103965e-01
8.24467242e-01 4.01559770e-01 -5.62468767e-01 7.08048284e-01
-9.33820903e-02 3.11614424e-02 6.54389486e-02 1.93692043e-01
-3.57889161e-02 -9.30137634e-02 -1.16658139e+00 5.34018219e-01
8.89275432e-01 -4.95584399e-01 -3.16673517e-01 1.66117653e-01
5.87325990e-01 -4.75673646e-01 -1.06150103e+00 3.97734880e-01
1.03262842e+00 -1.05893934e+00 6.75872862e-01 -1.32577240e+00
3.89484614e-01 8.56874064e-02 -3.20540488e-01 -1.21096039e+00
-1.12884894e-01 -9.82695997e-01 -4.46917653e-01 5.09840846e-01
3.52316707e-01 -8.29025507e-01 6.74496770e-01 9.69060183e-01
1.27017319e-01 -1.09544361e+00 -1.18082130e+00 -1.39219725e+00
5.88326752e-01 -3.85162115e-01 7.36308575e-01 8.95390809e-01
2.89935917e-01 -1.69521924e-02 -3.33466679e-01 -1.59740016e-01
4.77961332e-01 1.85055628e-01 4.44436044e-01 -7.38377750e-01
-6.97615862e-01 -8.95151138e-01 -2.39800543e-01 -1.27217245e+00
2.58503854e-01 -6.21042132e-01 4.81921919e-02 -1.20950556e+00
-1.00019544e-01 -5.71859837e-01 -6.86789870e-01 3.30260843e-01
6.91351220e-02 -3.33695561e-01 4.65378255e-01 -2.45681708e-03
-6.36621594e-01 4.66647238e-01 1.64776504e+00 2.12851062e-01
-3.31079602e-01 3.97778228e-02 -5.02161562e-01 6.39254749e-01
9.78946567e-01 -4.40758198e-01 -4.11408126e-01 -9.09860581e-02
2.17939511e-01 5.00955164e-01 4.31632578e-01 -7.56972790e-01
-2.08572716e-01 -7.11769700e-01 6.17312156e-02 2.85152234e-02
2.91755766e-01 -8.38226080e-01 -1.45596296e-01 7.92213023e-01
-6.49257779e-01 1.79324925e-01 1.17808588e-01 2.46695980e-01
1.91406891e-01 -6.97159231e-01 1.02046680e+00 -8.84158239e-02
-5.64856291e-01 1.99411914e-01 -2.30865315e-01 5.15629709e-01
8.07254374e-01 -9.57960337e-02 -3.74958396e-01 -6.66196823e-01
-5.85165560e-01 3.02016318e-01 4.82053012e-01 -6.16399460e-02
1.87026694e-01 -1.37424874e+00 -3.53180647e-01 -3.93380187e-02
-3.84851247e-01 -5.85224926e-01 -2.33591691e-01 1.17196918e+00
-2.49222770e-01 5.31097710e-01 -4.91491146e-02 -3.14626068e-01
-7.40646064e-01 5.44636309e-01 8.78162384e-01 -5.19739985e-01
-2.75831729e-01 3.32793415e-01 -1.35413215e-01 2.37245150e-02
3.98270428e-01 -3.61964345e-01 2.33758599e-01 -1.32722139e-01
5.19636393e-01 6.15681887e-01 -2.32535452e-01 -2.71952778e-01
-1.48082897e-01 2.57694453e-01 3.13033730e-01 -6.03422284e-01
8.85287642e-01 6.34242669e-02 3.75331163e-01 5.14095843e-01
1.18784034e+00 -2.95430660e-01 -1.69081151e+00 1.45476058e-01
-1.11023471e-01 -3.81084472e-01 2.97306359e-01 -9.62797225e-01
-8.38102460e-01 5.89951456e-01 6.49837077e-01 5.98344684e-01
8.21636081e-01 -4.85039532e-01 7.02184618e-01 6.32008910e-01
4.28368896e-01 -1.48642468e+00 -2.72670150e-01 6.01578176e-01
8.48093808e-01 -1.23217833e+00 1.13654301e-01 3.71722817e-01
-6.40702188e-01 1.03452599e+00 3.80498409e-01 -2.73046672e-01
5.34169614e-01 1.98980778e-01 -8.56326241e-03 1.88380495e-01
-7.65980542e-01 -4.87325490e-01 -9.93244201e-02 4.32954341e-01
1.89657196e-01 2.03978136e-01 -4.97949094e-01 2.46444285e-01
-4.02686387e-01 2.90488988e-01 2.85063595e-01 9.56982672e-01
-5.60566068e-01 -1.02834833e+00 -3.32350522e-01 3.82277071e-01
-6.21504843e-01 1.16274029e-01 -2.74994433e-01 1.04275203e+00
-2.86718488e-01 8.83778989e-01 2.92329818e-01 -1.80818856e-01
2.55393088e-01 3.25819366e-02 1.01097214e+00 -5.59162460e-02
-5.21619439e-01 -2.72443473e-01 2.09923074e-01 -8.00625920e-01
-2.73737252e-01 -5.02522528e-01 -1.40485072e+00 -7.69662142e-01
-2.32050493e-01 3.81685734e-01 5.44394314e-01 1.12537408e+00
1.90716192e-01 3.11747134e-01 7.69012690e-01 -4.60599512e-01
-1.61687553e+00 -6.38375342e-01 -6.25366390e-01 3.83610606e-01
7.48464406e-01 -8.39671493e-01 -4.38465595e-01 -4.92831200e-01]
|
[4.164836406707764, 2.4360311031341553]
|
a20b436e-ec92-4481-bbfe-724e9b7ba3b4
|
interpretable-image-clustering-via
|
2012.09743
| null |
https://arxiv.org/abs/2012.09743v1
|
https://arxiv.org/pdf/2012.09743v1.pdf
|
Interpretable Image Clustering via Diffeomorphism-Aware K-Means
|
We design an interpretable clustering algorithm aware of the nonlinear structure of image manifolds. Our approach leverages the interpretability of $K$-means applied in the image space while addressing its clustering performance issues. Specifically, we develop a measure of similarity between images and centroids that encompasses a general class of deformations: diffeomorphisms, rendering the clustering invariant to them. Our work leverages the Thin-Plate Spline interpolation technique to efficiently learn diffeomorphisms best characterizing the image manifolds. Extensive numerical simulations show that our approach competes with state-of-the-art methods on various datasets.
|
['Behnaam Aazhang', 'Richard Baraniuk', 'Anirvan Sengupta', 'Yanis Bahroun', 'Randall Balestriero', 'Romain Cosentino']
|
2020-12-16
| null | null | null | null |
['image-clustering']
|
['computer-vision']
|
[-4.13262576e-01 1.20635390e-01 -1.42620802e-01 -4.39626604e-01
-5.14903128e-01 -8.41297090e-01 6.12166524e-01 -3.60027909e-01
-5.50827831e-02 -7.36034811e-02 2.20697701e-01 6.25739470e-02
-5.27946413e-01 -3.12516749e-01 -6.83211505e-01 -7.41542518e-01
-5.53249180e-01 3.41626048e-01 -1.63305894e-01 5.73133677e-02
5.62785327e-01 5.80917358e-01 -1.22912335e+00 -2.22661838e-01
8.39061558e-01 6.35917783e-01 -1.05312265e-01 5.76181293e-01
1.43326491e-01 3.68553102e-01 -2.01231334e-03 -1.10590383e-01
6.05551779e-01 -2.75183111e-01 -1.10481822e+00 6.60606503e-01
5.53261399e-01 1.20567672e-01 -4.97654617e-01 1.16521513e+00
-1.36956841e-01 3.00575703e-01 1.09701025e+00 -1.40329027e+00
-1.28220046e+00 1.59453124e-01 -6.29569232e-01 1.98761731e-01
-1.80904970e-01 1.72349349e-01 9.40849543e-01 -9.91400182e-01
7.39089787e-01 1.36323786e+00 7.48693049e-01 3.96141052e-01
-1.39330602e+00 -2.53930509e-01 4.06778865e-02 -1.17946029e-01
-1.60806596e+00 -4.72599924e-01 9.06829834e-01 -7.78737605e-01
3.49660665e-01 3.04799318e-01 2.83273220e-01 1.44027159e-01
1.84043705e-01 3.71544659e-01 1.13385165e+00 -2.94713497e-01
2.51870990e-01 -1.14677347e-01 2.42610604e-01 1.06509709e+00
-2.17250008e-02 -2.70770818e-01 -9.59669277e-02 -1.97522923e-01
1.09043157e+00 1.55111954e-01 -3.11956048e-01 -8.28309059e-01
-1.29691446e+00 8.71613979e-01 5.79613924e-01 1.83518410e-01
-2.33000778e-02 4.95894670e-01 -1.48549587e-01 -1.85864344e-02
5.09693980e-01 6.96749687e-01 -1.30539358e-01 1.25396922e-01
-6.65619910e-01 -6.94063902e-02 5.39187670e-01 1.04188192e+00
1.05999434e+00 -2.65279919e-01 2.12057516e-01 4.42009240e-01
4.81800735e-01 3.47576737e-01 9.02974531e-02 -1.73639512e+00
-3.65613103e-02 8.58834743e-01 -1.40836462e-01 -1.25991011e+00
-3.70797634e-01 9.86303091e-02 -7.37459779e-01 2.20865801e-01
5.02042592e-01 1.40938684e-01 -7.58472264e-01 1.80803061e+00
3.31198871e-01 6.96902275e-01 -1.46652907e-01 8.61112058e-01
1.88178107e-01 3.30650598e-01 -4.79339324e-02 -5.73474960e-03
1.22907472e+00 -7.58942068e-01 -4.92874831e-01 2.49601111e-01
3.74251425e-01 -6.31907701e-01 1.21858454e+00 -1.73177138e-01
-1.10596359e+00 -3.96064639e-01 -8.23498845e-01 3.13382538e-04
-3.11446398e-01 3.57403383e-02 5.76106250e-01 5.71684241e-01
-1.60932672e+00 8.26852858e-01 -1.12751091e+00 -5.10828555e-01
4.83355045e-01 6.29278660e-01 -3.99591982e-01 3.37350190e-01
-3.79759073e-01 5.96553087e-01 3.10540758e-02 -1.10651694e-01
-4.90797698e-01 -9.03317332e-01 -7.78217375e-01 -2.67805129e-01
-1.11903297e-02 -8.11151028e-01 8.12095523e-01 -7.96348691e-01
-1.41179037e+00 1.17333102e+00 -3.66927326e-01 -9.82941613e-02
2.56191164e-01 1.19104259e-01 -7.93040693e-02 6.99723482e-01
-1.30984128e-01 8.52499306e-01 1.12967825e+00 -1.46639037e+00
-1.49676397e-01 -3.95688236e-01 9.32469368e-02 2.40032807e-01
-5.12052357e-01 -1.46438986e-01 -5.66710591e-01 -6.25081360e-01
1.90918043e-01 -1.37182891e+00 -3.54931086e-01 1.72145247e-01
-5.02412856e-01 -1.47149965e-01 1.34678566e+00 -3.12297314e-01
1.14241934e+00 -2.28616548e+00 4.42149460e-01 5.33562541e-01
6.60902560e-01 -1.30615503e-01 -3.59535008e-03 1.37432933e-01
-1.66568175e-01 5.63997149e-01 -3.07919800e-01 -4.69962239e-01
2.43943661e-01 2.20890686e-01 -2.81672359e-01 1.03985918e+00
2.82868475e-01 1.05491877e+00 -7.80837536e-01 -6.16138637e-01
2.40587562e-01 4.87915248e-01 -6.72796607e-01 3.88925038e-02
5.71703501e-02 6.97620153e-01 -5.43742716e-01 3.62321943e-01
6.23166800e-01 -5.80351770e-01 1.27815172e-01 -3.58288676e-01
-7.29164779e-02 -3.08165461e-01 -1.03931248e+00 1.74851990e+00
8.79425630e-02 5.79926074e-01 2.86715448e-01 -1.09833121e+00
7.15058446e-01 5.83546683e-02 8.08605850e-01 2.75811225e-01
5.79930432e-02 6.49698824e-02 -1.87131360e-01 -5.00579774e-01
1.83887273e-01 5.58706634e-02 9.23762098e-02 6.05242133e-01
-9.29509383e-03 -2.21244171e-01 -1.42788053e-01 2.32094109e-01
8.82218957e-01 1.29182547e-01 -3.49175595e-02 -1.25219703e+00
5.48843503e-01 -4.15266352e-03 3.32309961e-01 3.38755280e-01
-4.10550594e-01 5.96634090e-01 1.83162481e-01 -3.78027707e-01
-1.22070622e+00 -1.43116939e+00 -3.94444674e-01 6.91890657e-01
5.01028955e-01 -2.70319849e-01 -1.47961664e+00 -6.29304588e-01
2.94328667e-02 1.53866082e-01 -8.88189018e-01 -7.28586465e-02
-6.92349136e-01 -7.54947245e-01 5.63439310e-01 3.83984685e-01
4.74859923e-01 -5.89384258e-01 -2.96861738e-01 -3.63873571e-01
6.85028313e-03 -1.05609560e+00 -1.25050092e+00 -3.81584257e-01
-9.67270553e-01 -1.54857171e+00 -4.54133451e-01 -1.23995507e+00
1.12154973e+00 4.65036035e-01 1.04515517e+00 2.71718442e-01
-4.88117605e-01 1.09295869e+00 8.19143876e-02 2.15880543e-01
-3.16719562e-01 -1.04706086e-01 3.53404462e-01 1.63668975e-01
3.43809754e-01 -6.71330929e-01 -8.89348924e-01 6.19612098e-01
-1.02261221e+00 -3.69518548e-01 1.68575063e-01 3.92156214e-01
7.56777227e-01 4.22952950e-01 2.00357080e-01 -6.14088953e-01
4.71351504e-01 -5.00516772e-01 -4.09725189e-01 2.54763514e-01
-6.85545325e-01 3.82144034e-01 6.02390230e-01 -4.14936334e-01
-6.93198860e-01 2.19107389e-01 6.21689498e-01 -7.35127985e-01
-8.63293409e-02 5.88472420e-03 -1.12600364e-01 -5.60026765e-01
4.85155970e-01 -9.60592739e-03 3.96329522e-01 -3.40157151e-01
8.54047298e-01 5.57982504e-01 8.82970273e-01 -7.86780119e-01
1.28018498e+00 1.24009955e+00 2.91375369e-01 -8.67632627e-01
-4.12843913e-01 -6.20198905e-01 -1.05986714e+00 -1.38052985e-01
1.21659517e+00 -6.34815276e-01 -9.28153336e-01 3.32878172e-01
-8.92448008e-01 -2.62229264e-01 -3.04082572e-01 3.21414143e-01
-1.00132942e+00 5.44911921e-01 -8.00594449e-01 -3.74139011e-01
-1.41062617e-01 -1.36375785e+00 1.01921415e+00 3.25924680e-02
-1.72819495e-01 -1.61606991e+00 3.35515700e-02 -3.22873332e-02
1.21907644e-01 5.19695759e-01 1.12814033e+00 -2.65235543e-01
-8.68101239e-01 2.64538825e-01 -1.17518589e-01 1.46482900e-01
3.53286684e-01 3.28089893e-01 -7.46988595e-01 -2.77624100e-01
9.02265981e-02 3.13327640e-01 4.86373901e-01 6.26830280e-01
1.37209725e+00 -5.08378029e-01 -5.36624014e-01 1.01476526e+00
1.36165869e+00 -1.25030741e-01 5.66359103e-01 6.16521686e-02
9.11103487e-01 6.46841586e-01 8.83928686e-02 -4.98031359e-03
7.47180760e-01 5.88521898e-01 3.83366287e-01 -1.97368234e-01
-5.43936156e-02 -4.55961712e-02 2.83819169e-01 1.14623785e+00
-2.12686509e-01 3.44846934e-01 -9.46599126e-01 5.67349255e-01
-1.97292769e+00 -9.10150111e-01 -9.51754451e-02 1.85863125e+00
6.27001286e-01 -4.27741021e-01 2.32923344e-01 -2.60958344e-01
1.03257871e+00 2.37992499e-02 -5.57184100e-01 -1.89535379e-01
3.89764011e-02 2.30176579e-02 6.56793118e-01 7.09562659e-01
-1.35138643e+00 8.19873810e-01 7.71570778e+00 5.54174542e-01
-7.32730448e-01 -2.30076276e-02 6.04650021e-01 4.20709401e-01
-4.32035536e-01 -3.69612239e-02 -4.25622195e-01 3.06074589e-01
7.99369752e-01 -3.60674977e-01 8.33119154e-01 7.27756023e-01
2.72779375e-01 5.03462195e-01 -1.38212717e+00 9.45994258e-01
1.72155071e-02 -1.63457239e+00 4.67895865e-02 4.12009656e-01
8.56243670e-01 -7.04170018e-02 6.15479946e-01 -4.46547687e-01
5.42515337e-01 -1.26025498e+00 5.02797961e-01 6.05098248e-01
7.17057943e-01 -6.78959489e-01 -2.25170907e-02 8.14880207e-02
-1.48934340e+00 1.19943812e-01 -1.80661663e-01 2.39560634e-01
-1.95292637e-01 1.08500175e-01 -7.08399773e-01 2.66252726e-01
8.26174438e-01 1.02427781e+00 -7.04543114e-01 6.73776805e-01
1.80316210e-01 2.92137802e-01 -3.15203816e-01 4.54230428e-01
3.09390485e-01 -8.26892912e-01 6.47215068e-01 1.24299335e+00
2.57757246e-01 5.02831221e-01 1.54511273e-01 1.23185432e+00
-2.12311760e-01 -1.36810720e-01 -8.68648052e-01 1.07499130e-01
5.45409977e-01 1.34810483e+00 -9.18146431e-01 -1.08705021e-01
-1.41656443e-01 8.78963709e-01 4.42799747e-01 4.09926325e-01
-5.44493735e-01 -1.42982587e-01 1.23026288e+00 2.30789557e-01
3.50957632e-01 -7.51886547e-01 -4.07655597e-01 -1.23080420e+00
-1.16438143e-01 -4.41838920e-01 2.92821109e-01 -5.65127373e-01
-1.56993437e+00 3.79779369e-01 2.20281109e-01 -1.27656841e+00
-1.86754644e-01 -7.62403429e-01 -8.41797411e-01 4.79244024e-01
-1.28048420e+00 -1.14637756e+00 -2.00311765e-01 9.24417675e-01
2.43633881e-01 1.37677625e-01 8.20791423e-01 -7.32722953e-02
-3.03116292e-01 4.50283974e-01 3.67162943e-01 4.72931951e-01
4.32341069e-01 -1.58025372e+00 4.65968221e-01 8.20365906e-01
8.85046870e-02 1.18391287e+00 5.00914752e-01 -2.21155614e-01
-1.74633873e+00 -1.24419558e+00 1.71956375e-01 -9.43198442e-01
8.74215424e-01 -1.89267382e-01 -8.99180412e-01 1.14572060e+00
3.38860452e-01 2.23970667e-01 8.14361572e-01 -1.47570267e-01
-4.86806482e-01 1.06564865e-01 -1.37204492e+00 8.59316945e-01
1.22989237e+00 -6.59948051e-01 -4.66404796e-01 3.02194834e-01
7.63103545e-01 -2.24484399e-01 -1.33439398e+00 1.69095188e-01
1.91447452e-01 -7.08712339e-01 1.36471200e+00 -7.27052152e-01
2.01587513e-01 -6.23521090e-01 -2.33126074e-01 -1.46745968e+00
-4.95151728e-01 -1.10117745e+00 -7.16554075e-02 1.01034403e+00
4.74723391e-02 -6.20659411e-01 6.93574429e-01 1.07798958e+00
-1.84495851e-01 -5.31195462e-01 -8.72686863e-01 -8.77941370e-01
5.48572004e-01 -4.32442222e-03 5.72772503e-01 1.36457801e+00
4.49059457e-01 -8.94647539e-02 1.06489778e-01 3.66828322e-01
1.17419517e+00 1.37098268e-01 6.95375979e-01 -1.16861773e+00
-7.79526457e-02 -4.88766849e-01 -6.90343022e-01 -9.93714809e-01
6.99248672e-01 -1.20453835e+00 -1.46037564e-01 -8.45690668e-01
1.48442477e-01 -6.49509609e-01 -3.80701525e-03 2.15202689e-01
-1.06571890e-01 4.59083498e-01 4.39604342e-01 7.83911347e-01
-8.17338526e-01 4.37959045e-01 1.39310825e+00 -1.94504987e-02
-2.43919492e-01 -4.78309691e-01 -7.51457870e-01 1.25774515e+00
7.79385686e-01 -2.11297929e-01 -4.29920524e-01 -5.18799365e-01
-3.87578160e-01 -4.42044973e-01 5.86578488e-01 -8.17958236e-01
3.48766387e-01 -2.22067565e-01 6.97584301e-02 -1.18511669e-01
2.19142273e-01 -9.43256676e-01 3.02938759e-01 3.29378664e-01
-2.98206478e-01 2.51332521e-01 -1.67924818e-02 7.79093623e-01
4.31142300e-02 1.66620642e-01 1.01614702e+00 -2.52987407e-02
-6.13739014e-01 5.99266112e-01 -1.63866341e-01 4.27632540e-01
1.25812018e+00 -1.37422532e-01 -4.45634097e-01 -3.26243967e-01
-7.10893452e-01 2.26548046e-01 1.06727433e+00 2.26023272e-01
4.71156538e-01 -1.48787725e+00 -4.21438128e-01 1.12072915e-01
-2.96163354e-02 8.66708234e-02 -2.07955733e-01 1.00037944e+00
-7.26552784e-01 1.32152900e-01 6.04725704e-02 -9.36370969e-01
-9.88012671e-01 7.32067287e-01 6.35013103e-01 1.92927435e-01
-7.28062570e-01 3.90409231e-01 4.39527094e-01 -5.64035892e-01
-2.25485355e-01 -4.26138222e-01 8.56874585e-02 -5.10941803e-01
3.19076687e-01 5.31280696e-01 -4.93546754e-01 -9.95788813e-01
-5.16999602e-01 1.28003037e+00 2.26978153e-01 -1.89978555e-02
1.21057856e+00 -4.95274037e-01 -3.91336858e-01 2.84965783e-01
1.75801098e+00 1.96272228e-03 -1.74153340e+00 -1.50555789e-01
1.67165920e-01 -4.61294144e-01 -1.73302278e-01 2.08696518e-02
-1.18645287e+00 5.21050572e-01 4.01436746e-01 3.53697628e-01
1.02271056e+00 3.96913886e-01 6.46701574e-01 3.78747612e-01
2.38973603e-01 -1.15323603e+00 1.63015246e-01 1.48894712e-01
5.35940766e-01 -1.14775324e+00 1.66153498e-02 -8.06424618e-01
-5.40538132e-01 1.18292677e+00 1.15790047e-01 -8.15316439e-01
1.33589268e+00 1.94768474e-01 2.39417478e-01 -5.93984306e-01
-2.04902753e-01 7.49306083e-02 5.50685585e-01 8.15350533e-01
4.86193262e-02 8.56400356e-02 -3.77747640e-02 1.40817538e-01
-2.06755593e-01 -3.40294302e-01 5.06556809e-01 4.84714478e-01
-3.26233298e-01 -8.34640741e-01 -5.41079402e-01 1.06732674e-01
-3.77390355e-01 6.63422793e-02 -3.58776540e-01 9.85995770e-01
-1.76836371e-01 1.03986883e+00 4.50513482e-01 -3.01615506e-01
-3.53961438e-02 -6.79894239e-02 3.92600387e-01 -3.32343459e-01
-2.93384016e-01 5.53865284e-02 -6.73143268e-01 -7.88026869e-01
-1.01227176e+00 -9.55950141e-01 -1.69221961e+00 -5.83741248e-01
1.20473608e-01 1.89363539e-01 4.95591044e-01 8.22301805e-01
7.90748715e-01 -5.03216833e-02 9.81994629e-01 -1.07892883e+00
-4.91851181e-01 -3.42859566e-01 -8.44567299e-01 9.63401556e-01
3.88762176e-01 -6.75858974e-01 -7.26995826e-01 7.79962838e-01]
|
[8.618050575256348, 3.4356656074523926]
|
bfb262fb-fb54-41d0-9dc0-f01fd80fc833
|
generative-or-contrastive-phrase
|
2204.09358
| null |
https://arxiv.org/abs/2204.09358v2
|
https://arxiv.org/pdf/2204.09358v2.pdf
|
Generative or Contrastive? Phrase Reconstruction for Better Sentence Representation Learning
|
Though offering amazing contextualized token-level representations, current pre-trained language models actually take less attention on acquiring sentence-level representation during its self-supervised pre-training. If self-supervised learning can be distinguished into two subcategories, generative and contrastive, then most existing studies show that sentence representation learning may more benefit from the contrastive methods but not the generative methods. However, contrastive learning cannot be well compatible with the common token-level generative self-supervised learning, and does not guarantee good performance on downstream semantic retrieval tasks. Thus, to alleviate such obvious inconveniences, we instead propose a novel generative self-supervised learning objective based on phrase reconstruction. Empirical studies show that our generative learning may yield powerful enough sentence representation and achieve performance in Sentence Textual Similarity (STS) tasks on par with contrastive learning. Further, in terms of unsupervised setting, our generative method outperforms previous state-of-the-art SimCSE on the benchmark of downstream semantic retrieval tasks.
|
['Hai Zhao', 'Bohong Wu']
|
2022-04-20
| null | null | null | null |
['semantic-retrieval']
|
['natural-language-processing']
|
[ 3.60353589e-01 2.88989127e-01 -3.23310703e-01 -5.25060594e-01
-1.09813809e+00 -3.74909997e-01 1.03664303e+00 4.33911800e-01
-4.04087931e-01 6.00163877e-01 6.63100421e-01 -3.33424181e-01
-1.33641526e-01 -1.00712025e+00 -4.90670085e-01 -5.98869145e-01
4.01951313e-01 8.37475836e-01 1.12314813e-01 -5.59453547e-01
4.25143242e-01 -1.45464733e-01 -1.50693691e+00 4.64268476e-01
1.21283925e+00 6.14361644e-01 4.61834103e-01 1.86124936e-01
-6.59411728e-01 7.33400404e-01 -5.30477226e-01 -4.98017609e-01
-2.72273988e-01 -7.80875862e-01 -1.14635801e+00 -1.83475211e-01
2.46184602e-01 -7.16515929e-02 -1.33016706e-01 1.01676655e+00
5.79769731e-01 2.49032393e-01 1.02719033e+00 -8.19524646e-01
-1.19986999e+00 1.24750245e+00 -1.67307556e-01 3.90292794e-01
4.07232612e-01 -9.28706750e-02 1.55993259e+00 -9.95722711e-01
5.18051505e-01 1.30733871e+00 6.92588329e-01 5.65639913e-01
-1.21556842e+00 -3.81017357e-01 1.59931675e-01 1.96312368e-01
-1.13978624e+00 -4.57409710e-01 7.84044921e-01 -4.39747237e-02
1.26973462e+00 5.54913618e-02 3.79229397e-01 1.41913748e+00
-2.30365247e-02 1.04202795e+00 1.01874411e+00 -5.00714123e-01
1.74363047e-01 2.38329738e-01 2.57795691e-01 3.86725098e-01
1.47150204e-01 -1.58089414e-01 -6.24508262e-01 -8.35274905e-02
3.06523174e-01 1.38379857e-01 -1.57653719e-01 -2.22832561e-01
-1.11177456e+00 1.08373332e+00 4.13127422e-01 7.72186518e-01
-9.91324410e-02 5.35268225e-02 6.26586556e-01 5.09744227e-01
8.88546467e-01 6.43061101e-01 -2.68165410e-01 -1.22992545e-01
-1.05613017e+00 2.58646198e-02 5.56917191e-01 1.04711950e+00
7.89047539e-01 2.04408780e-01 -5.77973843e-01 9.72991586e-01
3.16151202e-01 4.68651116e-01 1.03675234e+00 -3.74052137e-01
5.35101950e-01 5.24764240e-01 -5.59313416e-01 -7.72517443e-01
-1.13653190e-01 -8.54811192e-01 -8.27622652e-01 -7.75891602e-01
-1.51139330e-02 1.72926351e-01 -5.49727142e-01 1.74586082e+00
-3.42905253e-01 5.46325743e-02 4.47366744e-01 6.34791076e-01
1.31443894e+00 6.38552368e-01 4.28929389e-01 -4.55447286e-01
1.07495630e+00 -1.16700399e+00 -6.92956150e-01 -4.80436921e-01
1.11901951e+00 -6.83866203e-01 1.38386095e+00 -1.08259164e-01
-1.02813840e+00 -7.05442131e-01 -9.42044795e-01 -1.03552453e-01
-4.13019240e-01 -1.06388152e-01 9.42370951e-01 6.45858705e-01
-1.08589351e+00 7.08791435e-01 -4.91680443e-01 -6.16127968e-01
4.37433630e-01 -1.44904792e-01 -9.42679048e-02 -1.35479763e-01
-1.44509089e+00 9.44113374e-01 5.58382213e-01 -3.08505446e-01
-8.08867097e-01 -9.09900725e-01 -8.70661795e-01 4.41831052e-01
2.86396980e-01 -1.12790155e+00 1.23347592e+00 -7.93974400e-01
-1.28600979e+00 1.16408646e+00 -2.68823564e-01 -7.15836942e-01
1.07428387e-01 -2.72748947e-01 -1.30059302e-01 3.04905564e-01
3.91685069e-01 6.69549108e-01 9.30543363e-01 -1.30205333e+00
-2.07232952e-01 -2.27459714e-01 1.62836276e-02 5.37023723e-01
-9.35251236e-01 -1.12160675e-01 -7.72403106e-02 -7.41150081e-01
1.67239696e-01 -5.46016574e-01 -1.25739783e-01 -6.06468260e-01
-4.06780779e-01 -8.00012410e-01 5.89639962e-01 -1.43854365e-01
1.21075511e+00 -2.04407263e+00 -6.78678323e-03 -2.81660765e-01
2.06009746e-02 2.96293795e-01 -3.41063350e-01 7.49204278e-01
-2.05021352e-01 1.98587984e-01 -2.83396393e-01 -7.89874375e-01
1.50806457e-01 2.10027695e-01 -7.72984445e-01 3.97993531e-03
2.28924900e-01 1.29963136e+00 -1.37190819e+00 -6.04613483e-01
-3.39900069e-02 2.05033481e-01 -5.78705490e-01 3.56752306e-01
-1.45515084e-01 1.68523282e-01 -6.29118919e-01 3.14299226e-01
3.03888410e-01 -3.44236255e-01 2.22628713e-01 1.71274133e-02
3.69525343e-01 9.43119287e-01 -4.41553771e-01 2.31178331e+00
-6.87794447e-01 3.27247620e-01 -6.17427111e-01 -1.65600026e+00
1.03253245e+00 3.42021227e-01 3.71225983e-01 -9.40526962e-01
-6.75892690e-03 2.65277803e-01 -1.90089345e-01 -4.38889265e-01
7.17658162e-01 -6.12108946e-01 -2.20588267e-01 7.36568511e-01
5.10936677e-01 -3.32086086e-01 1.27402693e-01 6.24090493e-01
1.07785821e+00 1.95678294e-01 2.43461162e-01 -5.05907297e-01
3.33438635e-01 -8.18535462e-02 1.97568908e-01 9.03682649e-01
1.01635754e-01 7.32541382e-01 3.36776972e-01 1.96616933e-01
-7.30832875e-01 -1.25414228e+00 -6.69404864e-02 1.40004873e+00
9.24374908e-02 -7.99733222e-01 -7.47905612e-01 -9.37310457e-01
-1.12715259e-01 8.99522662e-01 -3.96694183e-01 -7.32499957e-01
-3.76164198e-01 -8.52415860e-01 6.19125485e-01 7.35725045e-01
4.69718844e-01 -1.28636575e+00 -1.37951151e-01 1.82720959e-01
-2.70637274e-01 -1.03535843e+00 -4.15043890e-01 3.22533756e-01
-1.11615038e+00 -6.12383842e-01 -6.53668165e-01 -1.02624691e+00
6.64211452e-01 7.53465116e-01 1.56560838e+00 3.88267599e-02
2.99399551e-02 5.53876460e-01 -8.55384469e-01 -2.64131635e-01
-5.87580681e-01 5.17073631e-01 -1.42067119e-01 -3.22447330e-01
5.87075830e-01 -7.88374066e-01 -4.50012624e-01 -2.61917681e-01
-9.16563630e-01 -1.43192872e-01 6.57781541e-01 1.10505581e+00
3.67035717e-01 -1.20857947e-01 1.24097407e+00 -1.09018779e+00
9.45844889e-01 -6.35571003e-01 2.41416037e-01 2.98635095e-01
-9.52469766e-01 2.54472703e-01 7.40716159e-01 -1.75950423e-01
-1.26458132e+00 -4.41381842e-01 -3.38246047e-01 -2.43331462e-01
-9.58897248e-02 7.52786934e-01 -3.37288678e-02 5.04868448e-01
7.01988876e-01 6.50063396e-01 -1.82452220e-02 -5.12816191e-01
5.71295738e-01 6.34589314e-01 3.31077456e-01 -7.27542639e-01
8.22115600e-01 2.90429920e-01 -3.18850607e-01 -6.80847168e-01
-1.46630704e+00 -5.61344385e-01 -5.40923178e-01 1.34319887e-01
7.32072294e-01 -9.10136282e-01 -2.30232432e-01 7.37800449e-03
-1.09918761e+00 2.86597218e-02 -6.02220714e-01 1.45344228e-01
-7.19349563e-01 6.86581314e-01 -4.96381283e-01 -6.11017644e-01
-6.26158714e-01 -7.54812717e-01 1.25439405e+00 -1.08052753e-01
-3.40529919e-01 -1.38017547e+00 2.31570080e-02 5.60213923e-01
6.27203286e-01 -4.57100451e-01 1.13919997e+00 -1.16836166e+00
-2.63046801e-01 -7.14033321e-02 -8.15837532e-02 4.86682177e-01
2.05014125e-01 -5.59200108e-01 -1.04099536e+00 -3.28685313e-01
1.85025468e-01 -7.37243056e-01 1.34229577e+00 1.31584793e-01
1.14112914e+00 -3.61567289e-01 -1.90316603e-01 3.65692466e-01
1.37935567e+00 -2.46903792e-01 6.23493791e-01 3.86555582e-01
5.34665227e-01 6.92111194e-01 5.46992064e-01 2.07095459e-01
4.92623508e-01 4.67577279e-01 1.38110757e-01 2.62227398e-03
-3.09301555e-01 -6.98333383e-01 4.23550814e-01 1.16920388e+00
1.84124291e-01 -2.17448220e-01 -6.93319917e-01 6.22159064e-01
-1.98333383e+00 -1.22801805e+00 7.12491274e-02 1.94822693e+00
1.09358478e+00 1.98520780e-01 -6.17328174e-02 1.54214457e-01
4.75350708e-01 3.70379120e-01 -1.22331582e-01 -3.13523769e-01
-3.99769664e-01 5.05030692e-01 -1.25182584e-01 4.66119610e-02
-1.00675738e+00 1.34823728e+00 6.39531994e+00 1.28937948e+00
-7.97232687e-01 3.33912730e-01 5.11934578e-01 1.54715210e-01
-9.44071829e-01 1.51086226e-01 -8.90089154e-01 3.82358134e-01
1.01095641e+00 -3.48461568e-01 -1.84553996e-01 8.80256593e-01
-2.69008912e-02 1.60236359e-01 -1.26958895e+00 9.46528971e-01
4.55082387e-01 -1.48503554e+00 5.72302222e-01 -2.80917019e-01
7.23499715e-01 -1.33840442e-01 1.24237843e-01 8.79366100e-01
1.23618476e-01 -1.08446527e+00 4.26645070e-01 3.13505232e-01
4.21205342e-01 -6.32535398e-01 9.14977670e-01 5.31994343e-01
-8.87907743e-01 1.31370783e-01 -5.85060716e-01 -1.54712722e-02
1.82536528e-01 5.96448779e-01 -7.02233434e-01 8.80597770e-01
3.20664138e-01 1.15398133e+00 -7.31593192e-01 6.29646480e-01
-4.09620076e-01 7.80981481e-01 2.15182915e-01 -2.57337958e-01
4.51180965e-01 -8.85602161e-02 5.17174423e-01 1.45783615e+00
2.39001125e-01 -2.07180396e-01 1.12844542e-01 9.78079557e-01
-2.37376001e-02 3.98991764e-01 -8.20770621e-01 -2.70349026e-01
5.03033161e-01 1.06373942e+00 -6.03304267e-01 -5.90653777e-01
-3.34954292e-01 1.06235754e+00 5.34630418e-01 2.94744194e-01
-4.30753797e-01 -1.67387933e-01 3.26003879e-01 7.46135861e-02
1.86581761e-01 -5.89813329e-02 -4.33063418e-01 -1.38552129e+00
6.01605140e-03 -6.60419106e-01 4.70835984e-01 -6.17324591e-01
-1.75265110e+00 5.71245730e-01 9.01468098e-02 -1.30156434e+00
-6.46461725e-01 -3.36590230e-01 -7.03439236e-01 5.41889250e-01
-1.80453253e+00 -1.34412169e+00 7.85845518e-02 4.88916278e-01
9.09196138e-01 -4.75707442e-01 1.23752451e+00 9.18207839e-02
-2.03973189e-01 7.90078878e-01 1.87291622e-01 2.86119971e-02
8.59232306e-01 -1.35282063e+00 1.60857961e-01 6.88492119e-01
5.43568730e-01 9.94900107e-01 5.71912169e-01 -2.81066239e-01
-1.33397353e+00 -1.01464200e+00 1.32461059e+00 -4.74139839e-01
7.52204835e-01 -4.08505023e-01 -1.04579830e+00 4.76907194e-01
5.03838122e-01 -2.81418651e-01 8.18961203e-01 4.99454021e-01
-5.79114735e-01 2.75020546e-04 -6.51215196e-01 4.86527771e-01
1.42394078e+00 -7.95895219e-01 -1.33616734e+00 6.79383636e-01
1.03777158e+00 1.60314322e-01 -7.24776924e-01 4.14319783e-01
-6.46544546e-02 -9.59073365e-01 1.12448382e+00 -8.99674535e-01
8.14976811e-01 2.79722989e-01 -6.60258308e-02 -1.48805165e+00
-2.51062006e-01 -4.57417101e-01 3.19999196e-02 1.71413398e+00
4.62312371e-01 -6.85455918e-01 6.35357857e-01 -1.46235198e-01
-6.26253307e-01 -6.62985146e-01 -8.25359225e-01 -1.20858645e+00
5.47516108e-01 -4.49146301e-01 2.78407097e-01 1.18572307e+00
4.21796560e-01 1.08790863e+00 4.88775559e-02 -4.99299377e-01
5.11437953e-01 3.80904675e-01 4.65897202e-01 -1.17276394e+00
-4.45902079e-01 -6.53486669e-01 -2.01947734e-01 -1.25705159e+00
9.53439713e-01 -1.63686967e+00 1.21164858e-01 -1.82151186e+00
6.92845941e-01 -4.60686177e-01 -2.70953387e-01 5.42534649e-01
-5.19625247e-01 1.24971353e-01 -4.89865206e-02 2.89708763e-01
-8.23591828e-01 1.02412629e+00 1.29715240e+00 -3.77037644e-01
1.60599276e-01 -1.71861932e-01 -1.06656194e+00 4.43569511e-01
6.59399509e-01 -4.49948370e-01 -8.69846165e-01 -5.40659070e-01
2.85057008e-01 -2.29924008e-01 2.58240610e-01 -5.97505748e-01
1.42742932e-01 1.30188033e-01 -1.43411785e-01 -4.52085435e-01
1.85670078e-01 -2.93318957e-01 -5.40036142e-01 4.15265441e-01
-7.63708830e-01 -2.55868316e-01 -8.93895775e-02 5.00875533e-01
-6.45220160e-01 -6.62278593e-01 4.31917310e-01 -4.29197252e-01
-6.67941689e-01 1.30812407e-01 -4.49972093e-01 4.22526836e-01
3.06031644e-01 -1.89676836e-01 -5.29576004e-01 -4.93393689e-01
-5.05208254e-01 1.55790504e-02 1.35899872e-01 7.14043856e-01
7.72375882e-01 -1.32766604e+00 -9.07932222e-01 -2.89565213e-02
4.96574312e-01 -1.49558440e-01 2.91600257e-01 6.31164551e-01
2.42508829e-01 8.03616524e-01 1.90736830e-01 -6.50698066e-01
-8.50437760e-01 6.54191971e-01 -1.62871063e-01 -5.42362392e-01
-7.18158543e-01 9.07107890e-01 3.49383414e-01 -5.14475822e-01
5.46218343e-02 4.50078323e-02 -3.91395688e-01 2.23920882e-01
2.59534955e-01 -5.68835586e-02 2.57753670e-01 -4.08647358e-01
-1.75783515e-01 3.63782436e-01 -1.80986837e-01 -7.42602209e-03
1.32733572e+00 -1.24506220e-01 -1.56711787e-01 4.75272417e-01
1.57758355e+00 -3.75825644e-01 -4.45436925e-01 -5.42980552e-01
2.53856421e-01 -1.75255552e-01 1.73505209e-02 -4.16367233e-01
-7.21518040e-01 1.13400877e+00 3.27825025e-02 2.71462530e-01
1.06630898e+00 3.08494985e-01 9.45139289e-01 7.17700779e-01
4.10766155e-01 -1.08728528e+00 6.34508789e-01 7.91830242e-01
7.88109064e-01 -1.33182347e+00 -1.21397644e-01 -5.88662028e-01
-6.72261178e-01 9.02988076e-01 5.01193702e-01 -3.27946573e-01
4.21398342e-01 -8.37636292e-02 -1.74940348e-01 -2.76710063e-01
-8.99479032e-01 -5.35573661e-01 4.23435032e-01 6.30528688e-01
9.08805013e-01 -2.17644796e-01 -6.51445985e-01 7.22799718e-01
-5.07519126e-01 -5.52944124e-01 1.58174917e-01 8.04875135e-01
-6.08924806e-01 -1.15504110e+00 1.82846218e-01 4.11451817e-01
-2.81335354e-01 -6.82161152e-01 -4.34498399e-01 5.43065548e-01
-3.21592867e-01 1.05833316e+00 2.07021996e-01 -1.28012687e-01
6.16491772e-02 3.53624165e-01 5.30646205e-01 -1.17383122e+00
-7.56981552e-01 1.08008727e-01 3.21090847e-01 -3.10482353e-01
-7.00349569e-01 -4.96654332e-01 -1.18420136e+00 2.60416064e-02
-2.63858914e-01 4.84836400e-01 2.80741572e-01 1.39229381e+00
2.44185150e-01 5.85841656e-01 6.70906305e-01 -5.19501626e-01
-9.51621950e-01 -1.29883456e+00 -4.23897177e-01 6.59165204e-01
-2.19017327e-01 -4.84874725e-01 -4.93969202e-01 1.97810959e-02]
|
[10.944986343383789, 8.632025718688965]
|
de92a52e-49c2-4601-8914-b40879f5ee35
|
can-current-nli-systems-handle-german-word
|
2306.04523
| null |
https://arxiv.org/abs/2306.04523v1
|
https://arxiv.org/pdf/2306.04523v1.pdf
|
Can current NLI systems handle German word order? Investigating language model performance on a new German challenge set of minimal pairs
|
Compared to English, German word order is freer and therefore poses additional challenges for natural language inference (NLI). We create WOGLI (Word Order in German Language Inference), the first adversarial NLI dataset for German word order that has the following properties: (i) each premise has an entailed and a non-entailed hypothesis; (ii) premise and hypotheses differ only in word order and necessary morphological changes to mark case and number. In particular, each premise andits two hypotheses contain exactly the same lemmata. Our adversarial examples require the model to use morphological markers in order to recognise or reject entailment. We show that current German autoencoding models fine-tuned on translated NLI data can struggle on this challenge set, reflecting the fact that translated NLI datasets will not mirror all necessary language phenomena in the target language. We also examine performance after data augmentation as well as on related word order phenomena derived from WOGLI. Our datasets are publically available at https://github.com/ireinig/wogli.
|
['Katja Markert', 'Ines Reinig']
|
2023-06-07
| null | null | null | null |
['natural-language-inference']
|
['natural-language-processing']
|
[ 5.37302375e-01 3.49307775e-01 -2.14165032e-01 -3.39619577e-01
-5.29757679e-01 -1.15561104e+00 8.17866862e-01 1.19584687e-01
-5.46429813e-01 1.02445900e+00 3.70955020e-01 -1.03543007e+00
1.11618578e-01 -1.04573441e+00 -1.05347919e+00 -9.09541771e-02
6.18488677e-02 7.06619263e-01 -2.77592957e-01 -5.18790483e-01
-2.29853094e-01 2.58183032e-01 -1.04095376e+00 4.60812181e-01
9.44772005e-01 5.21060586e-01 -2.02282339e-01 7.26583242e-01
1.22628801e-01 6.73398197e-01 -5.66409349e-01 -8.91908288e-01
3.33209723e-01 -4.65934753e-01 -1.01051521e+00 -4.92150277e-01
7.60879576e-01 -3.92814249e-01 -5.97608507e-01 9.87416327e-01
3.54982734e-01 3.10388785e-02 8.51863682e-01 -1.23481512e+00
-8.28238487e-01 1.17752326e+00 1.84953988e-01 3.30825776e-01
6.58583581e-01 3.64481121e-01 1.65451407e+00 -8.66783559e-01
8.55175734e-01 1.28451967e+00 5.88668346e-01 7.43843913e-01
-1.27450728e+00 -5.92538059e-01 1.33966357e-01 2.15243950e-01
-1.23553610e+00 -6.17627203e-01 4.64946359e-01 -7.69010484e-02
1.48754311e+00 4.71754730e-01 4.57074106e-01 1.41499996e+00
3.33816946e-01 8.52439344e-01 1.11304224e+00 -6.81820393e-01
-1.76806748e-01 -2.34990582e-01 2.21384361e-01 7.68104255e-01
3.61592382e-01 4.12314951e-01 -4.33100760e-01 1.48941651e-01
1.99868158e-01 -5.60613573e-01 -1.71848580e-01 5.06780684e-01
-1.51866817e+00 8.10265541e-01 9.16148499e-02 2.36633807e-01
8.58673975e-02 2.25586981e-01 3.69989008e-01 6.37291729e-01
2.63955981e-01 7.32042670e-01 -5.95356166e-01 -5.13895117e-02
-7.40181267e-01 5.13440907e-01 8.78151417e-01 9.82946277e-01
5.98806918e-01 1.12925015e-01 -1.49516240e-01 4.58411366e-01
9.17875171e-02 7.76318789e-01 5.69596946e-01 -7.68667579e-01
7.76873112e-01 2.31161296e-01 -2.18154773e-01 -7.69807220e-01
-3.17627937e-01 -2.67326593e-01 -7.35032797e-01 -1.49905667e-01
7.22777069e-01 -9.37590823e-02 -7.12234557e-01 2.08675313e+00
-1.72338828e-01 -1.08704027e-02 3.70134324e-01 4.14587826e-01
1.13157463e+00 7.32191622e-01 -4.78592375e-03 2.35886365e-01
1.36539614e+00 -3.71875495e-01 -7.36301899e-01 -8.11947763e-01
8.75971138e-01 -7.39061296e-01 1.48783708e+00 3.21115732e-01
-1.37822795e+00 -4.09640074e-01 -1.22057080e+00 -3.46774042e-01
-6.67855978e-01 -2.41832770e-02 7.99966753e-01 4.94332433e-01
-7.89389610e-01 4.21458930e-01 -4.36313838e-01 1.56045005e-01
1.17815301e-01 2.37113148e-01 -6.39317870e-01 -1.53808311e-01
-2.00601459e+00 1.07304037e+00 6.62492454e-01 3.08493435e-01
-6.42869651e-01 -8.55186522e-01 -1.40622973e+00 -2.61141628e-01
1.74652413e-01 -6.83354974e-01 1.15768683e+00 -7.20654070e-01
-1.16105509e+00 1.39740181e+00 -2.98946738e-01 -8.18915963e-01
4.32362735e-01 -1.86810747e-01 -5.59809625e-01 -1.96798787e-01
7.39591196e-02 5.50076008e-01 4.25680757e-01 -9.04760003e-01
-6.24061167e-01 -3.38896632e-01 3.87764215e-01 7.58770183e-02
2.12373082e-02 3.04656010e-02 -4.58638333e-02 -8.40175033e-01
-9.06055570e-02 -9.59798574e-01 2.89863229e-01 -5.00520945e-01
-9.88736629e-01 -3.67557913e-01 1.61299095e-01 -7.76884377e-01
1.40212810e+00 -1.95452595e+00 9.53738242e-02 2.43125051e-01
1.09908685e-01 3.36061984e-01 -2.56614059e-01 3.41420114e-01
-3.02843690e-01 5.84184825e-01 -3.17936093e-01 -1.87945604e-01
4.37927544e-01 7.16474891e-01 -5.66969752e-01 4.54015046e-01
5.61629117e-01 1.30034840e+00 -8.45920742e-01 -3.41715544e-01
1.54162738e-02 -8.04267824e-03 -9.01406169e-01 2.04269066e-01
-3.98218334e-01 8.59015733e-02 4.11167979e-01 4.13494706e-01
5.34919322e-01 1.46580338e-01 2.39147633e-01 -2.23988652e-01
2.34879807e-01 1.03817225e+00 -1.05598545e+00 1.20389724e+00
-8.04798901e-01 5.08741498e-01 -2.57226467e-01 -8.25542092e-01
4.31561708e-01 1.21708937e-01 -1.42328307e-01 -5.70820928e-01
7.85190985e-02 3.88337582e-01 6.32404387e-01 -1.31281957e-01
4.57383633e-01 -4.40791517e-01 -6.26869023e-01 3.81439477e-01
2.73383796e-01 -4.90873247e-01 5.79396009e-01 4.58554626e-01
9.65498209e-01 4.10877392e-02 3.47318798e-01 -2.12654561e-01
4.50500548e-01 -1.92689717e-01 8.30594957e-01 8.06487501e-01
7.26355165e-02 4.12860155e-01 5.48304796e-01 -3.28502089e-01
-8.70993853e-01 -1.56495488e+00 -2.04535484e-01 9.61058676e-01
-1.84458643e-01 -4.57568973e-01 -4.92532313e-01 -7.33394802e-01
1.17938630e-02 1.34889317e+00 -6.22601092e-01 -3.48982275e-01
-1.03454113e+00 -3.60994250e-01 1.18296111e+00 4.72322762e-01
1.85797781e-01 -1.33631563e+00 -1.35871679e-01 1.62066475e-01
-4.74224091e-01 -1.34276271e+00 -6.33164763e-01 4.29767787e-01
-3.35201889e-01 -1.10581958e+00 -1.04338385e-01 -1.01039398e+00
5.73724985e-01 -7.26383328e-01 1.65151405e+00 8.82751793e-02
-1.67750552e-01 4.28072317e-03 -1.46736965e-01 -5.61336875e-01
-9.53591108e-01 3.54174227e-01 1.08241200e-01 -4.61734712e-01
4.56334233e-01 -4.44173217e-01 -1.08219720e-01 -1.25974655e-01
-1.03073478e+00 -4.94178161e-02 2.05579832e-01 9.37927365e-01
5.85095346e-01 -9.81384590e-02 3.73402536e-01 -1.37791502e+00
6.23852074e-01 -2.71010458e-01 -6.73488796e-01 1.98200047e-01
-2.46185958e-01 2.57840723e-01 9.86137569e-01 -2.88158774e-01
-7.37012804e-01 -3.97541553e-01 -4.52188313e-01 1.77423671e-01
-2.80280620e-01 6.27331793e-01 -6.60055101e-01 6.34246051e-01
6.27682030e-01 1.44294381e-01 -1.33619830e-01 1.77059928e-03
6.95164919e-01 3.97225082e-01 7.96230376e-01 -8.23801339e-01
1.01472926e+00 7.35192895e-02 1.48703465e-02 -6.96335852e-01
-9.44626391e-01 1.07672587e-01 -6.92401707e-01 3.73407990e-01
5.89580417e-01 -7.90226460e-01 -6.50422573e-01 4.42259192e-01
-1.38581097e+00 -7.53171206e-01 -5.11838913e-01 2.79368132e-01
-4.25104767e-01 4.53522146e-01 -9.17223752e-01 -3.61978292e-01
-5.05648255e-01 -9.76773500e-01 7.02831209e-01 -3.54014993e-01
-8.76497746e-01 -1.34521365e+00 -9.15480554e-02 1.20609425e-01
1.22219585e-01 1.85698211e-01 1.52695608e+00 -1.01380730e+00
-1.98078021e-01 -1.22351542e-01 2.19513923e-01 6.73320413e-01
1.63125649e-01 7.90057704e-02 -4.98306900e-01 -3.25662903e-02
-2.15273976e-01 -5.07945478e-01 6.96033299e-01 -4.13346477e-02
7.32820272e-01 -8.99145782e-01 1.46683544e-01 8.08053732e-01
1.16594434e+00 -1.79857060e-01 7.57370830e-01 2.85123378e-01
6.52853727e-01 4.53896135e-01 4.15209949e-01 7.97454193e-02
3.83934826e-01 1.91868886e-01 2.69619405e-01 -4.29440513e-02
-3.53166759e-01 -6.04416311e-01 5.41661620e-01 1.01821160e+00
3.85191053e-01 -5.12257934e-01 -1.15262043e+00 8.52777302e-01
-1.24888504e+00 -8.79004598e-01 -5.34987426e-04 2.14575529e+00
1.52797341e+00 4.47066426e-01 -1.74346775e-01 3.71616870e-01
2.94777185e-01 1.49011612e-01 -1.81382433e-01 -8.19114268e-01
-6.01264060e-01 8.95565450e-01 5.05961061e-01 1.30317914e+00
-1.03238106e+00 1.32128298e+00 5.98924160e+00 7.30993509e-01
-7.89456606e-01 -1.23618230e-01 5.02395093e-01 7.92481937e-03
-9.92520690e-01 4.83089276e-02 -1.07000542e+00 5.00852108e-01
8.23641360e-01 -2.45427668e-01 7.35992014e-01 2.33173683e-01
-2.90836900e-01 2.20764190e-01 -1.54337966e+00 6.04867816e-01
-1.15681343e-01 -1.22992122e+00 4.17882323e-01 -1.27318263e-01
4.34456229e-01 -3.76686156e-02 3.71659219e-01 3.79894823e-01
6.20799541e-01 -1.47231889e+00 8.72676969e-01 1.49073273e-01
1.23725557e+00 -1.01666880e+00 7.82624006e-01 2.02533707e-01
-7.42037714e-01 3.35990489e-01 -1.73770100e-01 -3.40538025e-01
3.56062427e-02 4.59848791e-01 -8.41966689e-01 4.53018755e-01
1.01135828e-01 5.66413701e-01 -5.47030389e-01 1.64998665e-01
-1.01646817e+00 1.03751862e+00 -5.62616587e-01 -1.53536037e-01
3.29458952e-01 -2.57649273e-01 6.76770031e-01 1.54391587e+00
-2.66662729e-03 -9.06120911e-02 2.63200030e-02 8.55786920e-01
-2.55069315e-01 -1.61333364e-02 -8.42173755e-01 -6.59558997e-02
5.82226694e-01 6.35053039e-01 -1.65036559e-01 -4.13587660e-01
-4.33043510e-01 1.20883024e+00 3.93675894e-01 3.44737232e-01
-8.83069754e-01 -6.08341575e-01 9.46437895e-01 2.95127928e-01
1.45810351e-01 -2.52843827e-01 -2.62414306e-01 -1.34547758e+00
1.10889142e-02 -1.33836401e+00 5.80618858e-01 -5.03306389e-01
-1.44047546e+00 5.18625200e-01 3.03813443e-03 -6.81535959e-01
-6.70518339e-01 -1.16398811e+00 -5.42074800e-01 9.81625557e-01
-1.57386136e+00 -1.34006214e+00 3.67314368e-01 6.35887682e-01
2.26892814e-01 -7.71396458e-02 1.19000351e+00 2.37214088e-01
-5.14128745e-01 1.25643814e+00 -2.60635287e-01 6.47686541e-01
6.18823171e-01 -1.56092167e+00 8.93696010e-01 1.13952708e+00
5.67241907e-01 1.03472924e+00 7.86487162e-01 -6.28950357e-01
-1.38459873e+00 -1.16196847e+00 1.59579074e+00 -6.59546316e-01
9.72197771e-01 -7.08056927e-01 -6.32424355e-01 1.34969330e+00
3.50000679e-01 1.34096276e-02 7.86011934e-01 3.38499308e-01
-5.88394582e-01 9.92539227e-02 -9.42342222e-01 1.10720456e+00
1.29047465e+00 -7.04006255e-01 -9.54892814e-01 5.39798737e-01
8.33810031e-01 -6.29075050e-01 -7.12007821e-01 4.59241688e-01
3.30582052e-01 -4.20878261e-01 8.42755675e-01 -1.06385922e+00
7.68740296e-01 -1.50539368e-01 -3.26941997e-01 -1.46961391e+00
-2.53039122e-01 -8.08232248e-01 -1.78060532e-01 1.26025701e+00
1.01250041e+00 -9.51869190e-01 4.42281276e-01 4.56136137e-01
-2.46196147e-02 -7.87465751e-01 -1.02616775e+00 -1.05603230e+00
8.68435919e-01 -8.18566620e-01 6.10575616e-01 7.73299217e-01
5.56751080e-02 7.05799699e-01 -5.89417554e-02 8.98051169e-03
2.88030893e-01 9.51692089e-02 6.29008532e-01 -8.01765084e-01
-3.73036474e-01 -4.06242996e-01 -4.13591474e-01 -8.36153924e-01
7.12538838e-01 -1.65771341e+00 -4.40952294e-02 -1.44610679e+00
-2.44438410e-01 -3.95755619e-01 2.46756338e-03 8.12859833e-01
-3.04547340e-01 4.45829570e-01 1.08195253e-01 -3.18665594e-01
-1.46284208e-01 3.19309235e-01 8.51952851e-01 -2.73923635e-01
3.59682560e-01 -1.66992635e-01 -7.42454648e-01 7.97912717e-01
1.04290116e+00 -3.01918805e-01 -2.72808969e-01 -6.83568895e-01
8.11239183e-01 -4.32119220e-01 3.74212116e-01 -7.00180471e-01
-9.53612924e-02 1.35551849e-02 2.04908594e-01 -3.42585206e-01
4.05902117e-02 -4.95865077e-01 -3.33403736e-01 5.08789539e-01
-4.85233247e-01 3.41693252e-01 4.70528841e-01 -4.89797955e-03
-1.42361447e-01 -3.69136393e-01 5.59178591e-01 -2.06135526e-01
-6.52192950e-01 1.97988912e-01 -5.99695921e-01 9.75648642e-01
4.74822938e-01 -9.32790060e-03 -2.62486726e-01 -4.04552281e-01
-5.27356148e-01 -8.76466483e-02 3.92096639e-01 2.94101328e-01
6.12528861e-01 -1.17863953e+00 -1.19165564e+00 3.93807560e-01
1.46447778e-01 1.08776629e-01 -3.04615080e-01 2.87773699e-01
-6.37951851e-01 3.27440172e-01 1.46433368e-01 5.59079908e-02
-1.07077181e+00 5.66187918e-01 2.96186745e-01 -6.37455881e-01
-3.60040426e-01 1.12924850e+00 1.17357805e-01 -1.04360390e+00
3.03589348e-02 -7.91822374e-01 2.53473938e-01 -1.89420596e-01
3.69375885e-01 -6.93430007e-02 1.92639157e-01 -7.20272005e-01
-6.24538302e-01 1.19977415e-01 -2.22681701e-01 -2.73699909e-01
1.06088114e+00 2.02481359e-01 -4.09526110e-01 4.24633712e-01
1.29448140e+00 5.54750383e-01 -5.66357672e-01 -2.78114766e-01
-6.10282719e-02 -5.31697348e-02 -2.93447644e-01 -9.80626285e-01
-5.39826334e-01 7.46342599e-01 -1.81111425e-01 -1.99514955e-01
7.04569459e-01 -1.21914744e-02 1.17159307e+00 4.11439210e-01
5.82232252e-02 -9.78482664e-01 -3.18060577e-01 1.20193541e+00
9.98359978e-01 -9.35831547e-01 -1.59605891e-01 -4.78727102e-01
-4.39202189e-01 7.36266673e-01 4.13806051e-01 -2.53674299e-01
4.77108151e-01 6.57532930e-01 6.30666986e-02 -7.35592935e-03
-6.68191433e-01 -2.27942795e-01 3.47837687e-01 5.80976784e-01
6.59803510e-01 1.62133455e-01 -3.35953861e-01 8.20170879e-01
-1.06782246e+00 -5.18961966e-01 2.80755848e-01 6.44669831e-01
3.22969884e-01 -1.13017702e+00 -1.06275871e-01 5.11211514e-01
-7.62506783e-01 -9.08038139e-01 -5.64136505e-01 1.02450788e+00
1.35468885e-01 9.43608642e-01 2.09696814e-01 -1.19174920e-01
3.54293376e-01 2.11799517e-01 8.61879110e-01 -7.80987561e-01
-5.43722510e-01 -6.82616293e-01 5.32711923e-01 -4.87611383e-01
3.98008049e-01 -4.63349849e-01 -1.35835063e+00 -4.82852519e-01
-1.01088718e-01 -8.15857016e-03 2.57134467e-01 1.00846696e+00
2.37087589e-02 5.89710116e-01 2.05193654e-01 -2.45306507e-01
-5.99155009e-01 -1.00825918e+00 -2.51977503e-01 8.34826529e-01
3.13346297e-01 -2.39608181e-03 -6.71544373e-01 1.52250946e-01]
|
[10.843408584594727, 9.438857078552246]
|
78d53825-16ec-4839-a36e-b9a078d5886d
|
the-generalized-laplacian-distance-and-its
| null | null |
http://openaccess.thecvf.com/content_cvpr_2013/html/Elboer_The_Generalized_Laplacian_2013_CVPR_paper.html
|
http://openaccess.thecvf.com/content_cvpr_2013/papers/Elboer_The_Generalized_Laplacian_2013_CVPR_paper.pdf
|
The Generalized Laplacian Distance and Its Applications for Visual Matching
|
The graph Laplacian operator, which originated in spectral graph theory, is commonly used for learning applications such as spectral clustering and embedding. In this paper we explore the Laplacian distance, a distance function related to the graph Laplacian, and use it for visual search. We show that previous techniques such as Matching by Tone Mapping (MTM) are particular cases of the Laplacian distance. Generalizing the Laplacian distance results in distance measures which are tolerant to various visual distortions. A novel algorithm based on linear decomposition makes it possible to compute these generalized distances efficiently. The proposed approach is demonstrated for tone mapping invariant, outlier robust and multimodal template matching.
|
['Yacov Hel-Or', 'Michael Werman', 'Elhanan Elboer']
|
2013-06-01
| null | null | null |
cvpr-2013-6
|
['tone-mapping']
|
['computer-vision']
|
[ 3.34264606e-01 -2.03002542e-01 -8.09558667e-03 -1.26255065e-01
-5.92047095e-01 -7.03686893e-01 4.17858839e-01 2.76458979e-01
-2.64955670e-01 2.58850157e-01 1.11532211e-02 -1.50545672e-01
-5.71935833e-01 -6.75030291e-01 -3.18279445e-01 -5.77739418e-01
-3.88567120e-01 2.63428718e-01 3.52126718e-01 -1.69528648e-01
5.08067608e-01 7.68232107e-01 -1.47678983e+00 2.31192820e-02
8.54450524e-01 6.60396278e-01 1.28220379e-01 5.63055634e-01
-3.28417331e-01 1.19128607e-01 -4.88388509e-01 -3.40669513e-01
5.27830482e-01 -7.18809962e-01 -7.13549078e-01 3.70969862e-01
7.31352210e-01 5.03088474e-01 -4.12069023e-01 1.40987766e+00
5.80641806e-01 5.12455523e-01 9.10041094e-01 -1.59996951e+00
-9.20323670e-01 3.59255940e-01 -8.17685306e-01 2.12470263e-01
8.07348192e-01 -6.12034261e-01 8.73437524e-01 -9.63172436e-01
6.71436369e-01 1.39029336e+00 8.53314817e-01 1.77485362e-01
-1.57319856e+00 -2.42867693e-01 -4.75005180e-01 6.58593535e-01
-1.84387076e+00 -1.75328851e-01 1.13690531e+00 -4.52851146e-01
5.92877924e-01 6.26534820e-01 4.03521925e-01 6.23129427e-01
2.34370068e-01 5.14783263e-01 1.27204227e+00 -9.21108663e-01
5.81134930e-02 6.81846216e-02 -6.00190684e-02 8.44006956e-01
2.18174428e-01 -1.97108060e-01 -3.86617690e-01 -4.47846979e-01
7.32226074e-01 -1.32043034e-01 -4.81723011e-01 -8.17354798e-01
-1.24301136e+00 8.89230907e-01 3.94276738e-01 6.65140033e-01
-4.15126570e-02 1.82817459e-01 2.93538362e-01 6.67717099e-01
2.83289105e-01 1.97522551e-01 4.78922248e-01 1.80501819e-01
-8.83255720e-01 -4.19344455e-01 7.93999672e-01 9.85522032e-01
1.09356713e+00 7.21104667e-02 1.13445811e-01 9.74513948e-01
3.66012186e-01 4.84740227e-01 4.67555821e-01 -1.00440240e+00
1.35254994e-01 7.78442264e-01 -4.82113600e-01 -1.62935042e+00
-5.24324596e-01 1.60420425e-02 -8.58209908e-01 2.47043312e-01
2.24382639e-01 4.59207237e-01 -6.63375437e-01 1.69288945e+00
1.39046475e-01 5.74181080e-01 -2.40404963e-01 7.62062907e-01
4.93417740e-01 3.83707583e-01 -3.72917831e-01 -3.86243254e-01
7.98726320e-01 -4.21142101e-01 -7.95303583e-01 4.01866108e-01
3.14123631e-01 -1.20859981e+00 1.06015706e+00 3.51690292e-01
-9.16984200e-01 -4.02846456e-01 -1.18555045e+00 7.57484362e-02
-6.21756613e-01 -4.13503498e-01 2.83843577e-01 1.02829766e+00
-1.55634975e+00 6.43846691e-01 -4.70573902e-01 -9.97376204e-01
-2.36867696e-01 4.23873544e-01 -6.08995616e-01 5.50925322e-02
-9.20812190e-01 9.58037555e-01 3.51043582e-01 -7.80418888e-02
6.57574981e-02 -2.17132062e-01 -9.38838303e-01 -5.47528267e-02
5.79485558e-02 -4.25961256e-01 4.57695901e-01 -8.18589091e-01
-1.31269932e+00 1.12975132e+00 -1.12810180e-01 -1.67150036e-01
4.89952862e-01 5.80361724e-01 -5.98101079e-01 3.88244897e-01
-3.60515751e-02 3.14063728e-01 1.12281990e+00 -1.04004514e+00
8.74478072e-02 -2.96761483e-01 -4.02868003e-01 4.36859243e-02
-3.98337573e-01 -7.43735805e-02 -2.97997057e-01 -8.34701240e-01
5.88750541e-01 -1.02235150e+00 1.09119127e-02 -6.68178573e-02
-2.76195943e-01 -1.47594422e-01 1.02245975e+00 -5.39716959e-01
1.30288136e+00 -2.17148614e+00 4.84540075e-01 1.06364131e+00
-1.01581505e-02 -1.08536504e-01 -2.70857543e-01 1.10660231e+00
-4.82237726e-01 2.67733019e-02 -4.43439335e-01 -1.09425969e-01
2.24678114e-01 1.69737011e-01 6.65003359e-02 8.94388139e-01
-2.41533220e-01 7.31191993e-01 -7.23481119e-01 -6.97096586e-01
3.85349810e-01 4.20003355e-01 -2.19520301e-01 -3.66192788e-01
3.66653830e-01 1.13101222e-01 1.45368055e-01 4.80183333e-01
7.59468138e-01 -2.28265822e-02 1.66913167e-01 -6.23107076e-01
1.72512303e-03 -5.58666170e-01 -1.60046792e+00 1.90582061e+00
-3.74189407e-01 1.06369579e+00 -2.20182054e-02 -1.13266110e+00
1.19460940e+00 1.50722861e-01 6.01569772e-01 -3.88393104e-01
1.05540127e-01 2.81521320e-01 -1.80574223e-01 -4.78767872e-01
3.27709973e-01 3.20049711e-02 -1.15088432e-03 3.83998275e-01
-3.42321172e-02 8.67728051e-03 3.51721734e-01 3.44042897e-01
9.81551647e-01 -1.37029737e-01 6.43219948e-01 -6.21238291e-01
9.15608764e-01 -1.79567665e-01 1.73265353e-01 4.39822614e-01
-7.56038949e-02 7.14708090e-01 3.00414443e-01 -1.03125297e-01
-9.71177697e-01 -1.33095646e+00 -1.65143907e-01 7.03724027e-01
3.75606388e-01 -5.21794260e-01 -6.71405315e-01 -3.03544432e-01
1.00243792e-01 2.09153712e-01 -4.21192378e-01 -4.06762749e-01
-4.63075370e-01 -3.70881379e-01 4.88343447e-01 1.57304525e-01
3.10921013e-01 -7.62002170e-01 -1.24783300e-01 3.28581259e-02
-7.38160908e-02 -7.11741507e-01 -9.74570274e-01 -1.23221204e-01
-8.11520517e-01 -1.18805134e+00 -6.44531250e-01 -1.04143631e+00
6.68558180e-01 5.01622319e-01 9.39023733e-01 1.65983468e-01
-7.44358242e-01 1.16788805e+00 -3.73228550e-01 2.55775273e-01
-6.30205274e-01 -2.77756095e-01 2.79307485e-01 5.74637055e-01
3.65531802e-01 -9.41042244e-01 -3.34858596e-01 3.76971096e-01
-1.24314845e+00 -6.04520380e-01 4.22672689e-01 6.86080515e-01
7.01637447e-01 9.43587795e-02 4.11477149e-01 -7.68600225e-01
8.95805597e-01 -2.23521739e-01 -6.11252666e-01 6.88393235e-01
-7.97402143e-01 2.06893548e-01 3.28930616e-01 -5.80840349e-01
-4.04062718e-01 3.47922742e-01 3.96604419e-01 -7.27722168e-01
1.04063332e-01 5.33529043e-01 -1.69315606e-01 -1.06033611e+00
8.30776095e-01 2.71293461e-01 3.23342234e-02 -4.95494723e-01
6.60619676e-01 5.64375281e-01 6.46614790e-01 -3.96515936e-01
1.10446370e+00 4.44577456e-01 5.75171471e-01 -1.27259803e+00
1.43464282e-01 -8.09818029e-01 -7.10566878e-01 -3.90934139e-01
8.40930879e-01 -2.08179504e-01 -8.37449491e-01 9.76470038e-02
-8.53941739e-01 3.11214119e-01 -6.63828850e-02 4.79302436e-01
-6.99546754e-01 1.11117768e+00 -2.53105700e-01 -7.76559591e-01
1.10156415e-02 -6.66877270e-01 7.51374125e-01 1.42536685e-01
-4.45839576e-02 -1.50760126e+00 4.04644638e-01 -9.91029516e-02
2.08367631e-01 5.00678122e-01 9.54074025e-01 -5.10403037e-01
-4.31257337e-01 -2.49388382e-01 -6.29854128e-02 6.73300251e-02
3.73523116e-01 1.48446695e-03 -5.44708371e-01 -5.72859168e-01
-1.14927143e-01 2.81935155e-01 9.17168081e-01 1.57144487e-01
7.36780226e-01 6.72279000e-02 -3.41450334e-01 6.69022560e-01
1.67090929e+00 8.42592046e-02 7.54720628e-01 3.44996125e-01
6.99239552e-01 5.10168970e-01 3.74886751e-01 1.47334874e-01
2.99683632e-03 8.51219952e-01 1.12406962e-01 -1.21295765e-01
-3.15625891e-02 -8.80944505e-02 2.39559293e-01 1.17387235e+00
3.27719897e-02 1.29227951e-01 -7.41643369e-01 3.18269521e-01
-2.03564596e+00 -1.14198434e+00 -2.63447046e-01 2.38421154e+00
4.37636167e-01 -2.56357491e-01 2.55713195e-01 4.25366282e-01
1.27019715e+00 7.56829157e-02 -1.37137920e-01 -7.34653533e-01
-4.13092822e-01 3.65487784e-01 7.08932161e-01 9.22677398e-01
-1.06592238e+00 7.19257951e-01 6.83542633e+00 1.02426672e+00
-9.15558696e-01 2.03117892e-01 -2.48293743e-01 5.22984028e-01
-3.22752565e-01 4.35561612e-02 1.00697145e-01 2.32906908e-01
5.74182689e-01 -6.64034843e-01 6.76706553e-01 5.65629721e-01
2.82415263e-02 9.79599170e-03 -9.85520124e-01 1.59886825e+00
6.35776937e-01 -9.09841478e-01 1.38222039e-01 -1.65911485e-02
7.66214490e-01 -5.21608114e-01 2.06175700e-01 -3.05509627e-01
-4.34061140e-01 -8.20500851e-01 2.22908452e-01 5.67322254e-01
6.87894762e-01 -8.80675316e-01 3.79251510e-01 -7.33647272e-02
-1.78239155e+00 3.56212676e-01 -5.94265163e-01 4.67754543e-01
5.65450713e-02 2.98533857e-01 -6.51973546e-01 8.53288174e-01
2.85196990e-01 6.62928164e-01 -8.42581153e-01 1.51359415e+00
2.76859641e-01 2.68489588e-02 -3.59116703e-01 1.87518045e-01
9.22007635e-02 -8.93441200e-01 7.54502535e-01 1.44174266e+00
6.61430478e-01 -3.00639808e-01 2.75185227e-01 8.25580716e-01
4.46928889e-02 5.95001340e-01 -1.01266134e+00 -1.53387235e-02
5.21000624e-01 1.36487615e+00 -1.31560671e+00 2.35863719e-02
-5.82630634e-01 1.46260726e+00 -6.73502237e-02 3.43905389e-01
-6.00166440e-01 -7.01253772e-01 3.61498713e-01 -1.03099674e-01
1.09358653e-01 -5.18221319e-01 1.60842538e-01 -9.24483716e-01
1.97207872e-02 -6.32090569e-01 5.11809647e-01 -6.00686789e-01
-1.46296751e+00 4.34976786e-01 1.18800029e-01 -1.70246041e+00
-1.54499441e-01 -6.97097659e-01 -5.97528100e-01 7.46061981e-01
-1.04323566e+00 -9.30515468e-01 -3.22506726e-01 1.26078916e+00
4.08618040e-02 -2.14399919e-01 7.79753327e-01 5.30949295e-01
-1.78321183e-01 6.98402584e-01 1.90055966e-01 -1.56592399e-01
9.17744875e-01 -1.63609517e+00 1.04104713e-01 8.10320437e-01
7.00083673e-01 5.58672369e-01 7.72606492e-01 -4.92469847e-01
-1.46059620e+00 -7.39307702e-01 7.99080133e-01 -1.70915022e-01
8.43777955e-01 -1.23071641e-01 -1.05282569e+00 4.67465043e-01
2.98263907e-01 -2.33232826e-01 7.16906726e-01 -4.71513778e-01
-5.24612069e-01 -1.55801520e-01 -1.22733915e+00 4.53581721e-01
1.12522578e+00 -9.47802722e-01 -5.76811314e-01 4.98667747e-01
2.41489261e-01 -8.05189386e-02 -9.56212103e-01 1.22538060e-01
1.95241332e-01 -1.09998691e+00 1.10504127e+00 -2.10430473e-01
-8.37002218e-01 -6.57094419e-01 -2.48408005e-01 -1.29254305e+00
-5.19473553e-01 -1.13190365e+00 2.80531615e-01 1.15643907e+00
1.91002995e-01 -8.11026812e-01 5.73806942e-01 -5.03454022e-02
3.90668288e-02 -4.65362184e-02 -1.32806408e+00 -1.27106297e+00
-3.82371694e-01 -1.31994843e-01 1.49721906e-01 1.31486154e+00
3.48342061e-01 4.85283174e-02 -4.33840871e-01 2.15929821e-01
9.74947035e-01 1.37012482e-01 2.58345544e-01 -1.53968668e+00
-1.62407428e-01 -7.01074898e-01 -1.30582058e+00 -6.17305934e-01
3.19721341e-01 -1.38931203e+00 -3.32523465e-01 -1.32455039e+00
-5.46777360e-02 4.80903462e-02 -3.60119224e-01 2.89074574e-02
1.40785515e-01 7.00161159e-01 2.86213517e-01 1.49936825e-01
-3.43501776e-01 2.32312441e-01 7.81460166e-01 -1.18451029e-01
-2.62107104e-01 -1.67942628e-01 -5.08469567e-02 4.94423568e-01
7.42365539e-01 -3.75885427e-01 -4.81910527e-01 -1.95942391e-02
2.79695243e-01 -1.08274430e-01 1.35903060e-01 -9.30651069e-01
6.07438564e-01 6.05896525e-02 7.30549358e-03 -3.88213277e-01
3.96850348e-01 -1.13480973e+00 5.77246249e-01 5.13138115e-01
-8.79533589e-02 6.24277890e-01 -3.01956218e-02 7.41436183e-01
-4.87150908e-01 -5.13875902e-01 8.06266665e-01 4.91504893e-02
-9.55183625e-01 1.85112178e-01 -3.64533514e-01 -8.98836032e-02
1.11928916e+00 -5.49229443e-01 -9.09674689e-02 -6.36397481e-01
-9.97117221e-01 -2.23848671e-01 6.43620968e-01 3.65807682e-01
6.99868023e-01 -2.00526547e+00 -5.33472478e-01 6.34424612e-02
1.33445293e-01 -1.07299519e+00 -7.94428885e-02 1.37590730e+00
-7.50482142e-01 1.60695329e-01 -3.33890706e-01 -7.31480777e-01
-1.65565681e+00 9.06434655e-01 2.83066779e-01 4.22313094e-01
-4.48822200e-01 3.98133844e-01 -2.07880542e-01 -7.70792142e-02
1.33612543e-01 1.07571736e-01 -2.32537035e-02 1.33105606e-01
1.77342683e-01 7.71506846e-01 4.83937515e-03 -8.96616876e-01
-7.14755714e-01 1.49254560e+00 7.68967867e-01 -3.04717422e-01
7.75773406e-01 -3.73832285e-01 -6.46943569e-01 6.19379222e-01
1.69537151e+00 4.16313738e-01 -2.67761290e-01 -2.78184384e-01
5.71639001e-01 -7.13919342e-01 -1.50775731e-01 -1.22297153e-01
-8.45409691e-01 6.88207328e-01 9.47380185e-01 8.41793716e-01
1.43886089e+00 1.32740224e-02 4.82335299e-01 4.61750150e-01
3.39675248e-01 -9.80365932e-01 4.17532884e-02 1.10273667e-01
1.00531578e+00 -9.70353782e-01 -1.61995515e-01 -4.99440283e-01
-2.49746174e-01 1.50665402e+00 1.65949553e-01 -2.46468142e-01
7.04562008e-01 2.86899228e-02 2.29895458e-01 1.28846215e-02
-5.23879267e-02 -6.78049266e-01 7.89530635e-01 8.17213893e-01
2.50986278e-01 -5.68925366e-02 -6.14235103e-01 -3.63641381e-01
5.27154692e-02 -6.39871478e-01 5.05843580e-01 7.93880403e-01
-6.96070611e-01 -1.41814518e+00 -7.64801681e-01 1.53290872e-02
3.57726254e-02 -3.65787372e-03 -9.35815692e-01 8.09493065e-01
-2.30620775e-04 8.38722110e-01 -1.78280503e-01 -6.98030353e-01
5.12739539e-01 2.95359522e-01 9.13419247e-01 -3.11992526e-01
-2.08011106e-01 1.42706662e-01 -3.01062882e-01 -6.57147527e-01
-8.81446302e-01 -5.26996136e-01 -1.03587329e+00 -4.92593855e-01
-3.09903026e-01 2.64671624e-01 6.10836446e-01 4.26909745e-01
1.19673453e-01 -4.21322435e-02 8.32084537e-01 -7.68926919e-01
-2.46493638e-01 -6.46447361e-01 -8.59517157e-01 5.74334383e-01
1.78006053e-01 -6.91021264e-01 -5.94468653e-01 8.93863663e-02]
|
[7.777602195739746, 4.456758499145508]
|
18cee694-5a2e-4997-ae5a-8d121895f5df
|
the-power-of-motifs-as-inductive-bias-for
|
2306.17246
| null |
https://arxiv.org/abs/2306.17246v1
|
https://arxiv.org/pdf/2306.17246v1.pdf
|
The power of motifs as inductive bias for learning molecular distributions
|
Machine learning for molecules holds great potential for efficiently exploring the vast chemical space and thus streamlining the drug discovery process by facilitating the design of new therapeutic molecules. Deep generative models have shown promising results for molecule generation, but the benefits of specific inductive biases for learning distributions over small graphs are unclear. Our study aims to investigate the impact of subgraph structures and vocabulary design on distribution learning, using small drug molecules as a case study. To this end, we introduce Subcover, a new subgraph-based fragmentation scheme, and evaluate it through a two-step variational auto-encoder. Our results show that Subcover's improved identification of chemically meaningful subgraphs leads to a relative improvement of the FCD score by 30%, outperforming previous methods. Our findings highlight the potential of Subcover to enhance the performance and scalability of existing methods, contributing to the advancement of drug discovery.
|
['Stephan Günnemann', 'Fabian Theis', 'David Lüdke', 'Leon Hetzel', 'Johanna Sommer']
|
2023-04-04
| null | null | null | null |
['drug-discovery']
|
['medical']
|
[ 3.90779108e-01 2.14666709e-01 -5.53013086e-01 1.42950052e-02
-8.42691422e-01 -7.27361262e-01 7.32319295e-01 4.49707806e-01
-9.67496932e-02 1.13027608e+00 3.18970472e-01 -7.44959891e-01
-1.25112291e-02 -9.02852297e-01 -8.34024847e-01 -9.58703458e-01
5.47882216e-03 6.04450643e-01 -1.04831979e-01 -6.03641709e-03
2.31804803e-01 6.84675753e-01 -8.29090476e-01 2.75951982e-01
1.06525612e+00 3.12484056e-01 2.40209475e-01 1.80107489e-01
6.31958395e-02 6.01379693e-01 -5.07964373e-01 -5.15432119e-01
-2.16807976e-01 -7.10438013e-01 -7.50420332e-01 -1.24732777e-01
1.94676220e-01 -6.04181997e-02 -3.36523086e-01 9.03344214e-01
7.75468230e-01 1.80648655e-01 8.83744478e-01 -5.40365279e-01
-7.03774989e-01 7.37939298e-01 -2.99827218e-01 1.43114328e-01
1.37425765e-01 3.29513788e-01 1.20179832e+00 -8.53583515e-01
9.46613908e-01 9.62839305e-01 4.29018706e-01 6.40052676e-01
-1.70413935e+00 -7.45660782e-01 -8.73706788e-02 1.22389823e-01
-1.42429781e+00 -3.31711680e-01 7.51470923e-01 -5.98206103e-01
1.21345174e+00 -2.47799214e-02 6.22256696e-01 1.18855655e+00
3.63918513e-01 4.80435461e-01 7.63211191e-01 -5.45625798e-02
6.54895246e-01 4.69242334e-02 -3.36313248e-01 6.17569923e-01
5.69063067e-01 5.04949642e-03 -4.87907797e-01 -4.25987929e-01
5.51922798e-01 -4.80980650e-02 -3.18164945e-01 -4.14213091e-01
-8.38912427e-01 1.27311373e+00 5.06305516e-01 3.27103406e-01
-4.65427876e-01 1.65812969e-01 1.08507782e-01 -2.32430711e-01
5.82804143e-01 1.19502497e+00 -3.80433679e-01 1.96038950e-02
-8.89844537e-01 4.29945588e-01 7.47981668e-01 5.52218676e-01
4.62636203e-01 1.11639999e-01 -3.13186646e-01 4.44895655e-01
1.47861108e-01 3.50999862e-01 9.25050229e-02 -5.56901991e-01
1.47269726e-01 5.13592660e-01 -7.24405497e-02 -5.00841379e-01
-4.53981668e-01 -8.38735342e-01 -4.72000897e-01 -2.12574944e-01
1.29780486e-01 -1.68974295e-01 -9.04984415e-01 1.75933957e+00
3.15848500e-01 -1.34868205e-01 -6.40416592e-02 4.06938046e-01
7.17861116e-01 5.15715301e-01 6.97916269e-01 -3.51323217e-01
1.01215506e+00 -4.59843934e-01 -4.28640068e-01 2.03925192e-01
8.29775453e-01 -5.23222864e-01 6.73399627e-01 5.04966140e-01
-7.88091123e-01 -1.35762215e-01 -8.55352223e-01 4.06663977e-02
-2.84854293e-01 -1.19713567e-01 1.15470135e+00 8.20745826e-01
-6.36147797e-01 1.00604641e+00 -7.97102153e-01 1.31531870e-02
1.22390115e+00 6.53743386e-01 -3.18936795e-01 -2.28664264e-01
-1.08268201e+00 7.25094616e-01 3.51968884e-01 -3.89156342e-01
-1.17239523e+00 -1.12934637e+00 -7.31534004e-01 2.03521624e-01
3.40246946e-01 -8.92804563e-01 8.50773871e-01 -5.55003822e-01
-1.40711319e+00 2.90693611e-01 -2.53268361e-01 -5.80727041e-01
1.76796421e-01 1.16177052e-01 -2.10142951e-03 3.64792621e-04
8.69816989e-02 6.64011717e-01 3.78464967e-01 -9.09744978e-01
-1.62127748e-01 -4.42314655e-01 -8.73337984e-02 4.58732359e-02
-2.23975435e-01 -5.05010605e-01 3.00084054e-03 -4.72730190e-01
-4.49855447e-01 -8.90825868e-01 -6.85898721e-01 -6.18220150e-01
-5.35816252e-01 -3.76019031e-01 2.41303861e-01 -4.14958864e-01
1.18992913e+00 -1.74349368e+00 5.65087259e-01 3.40311617e-01
4.91436511e-01 2.29223549e-01 -1.74695075e-01 6.66331470e-01
-6.88497573e-02 5.15917063e-01 -2.11138338e-01 4.59919535e-02
-3.92983943e-01 -1.59180135e-01 -1.69649079e-01 4.03200239e-01
3.87230784e-01 1.08156967e+00 -1.15211833e+00 -9.55148637e-02
2.59633064e-02 7.90549695e-01 -9.16035712e-01 -1.21025212e-01
-7.23944008e-01 6.10047400e-01 -6.70576811e-01 5.94992638e-01
3.97386283e-01 -6.29317880e-01 6.96983695e-01 -1.37997821e-01
5.54074273e-02 5.57288945e-01 -4.21267331e-01 1.58539617e+00
-9.76191983e-02 3.60472769e-01 -8.54207575e-01 -6.68887734e-01
6.66787565e-01 1.05092831e-01 5.87186277e-01 -5.72603822e-01
9.84636098e-02 1.39080390e-01 4.13658321e-01 -9.06139389e-02
6.87066466e-02 -6.57512307e-01 2.90900439e-01 1.28200665e-01
2.37164885e-01 -4.49406654e-02 3.26172799e-01 4.16151643e-01
1.06137383e+00 1.11687109e-01 2.37962112e-01 -2.11770549e-01
1.41872734e-01 2.14486942e-01 2.70420849e-01 4.89899129e-01
2.46149540e-01 3.42144370e-01 8.33050013e-01 -6.52112439e-02
-1.00420380e+00 -9.16153967e-01 -2.87379116e-01 7.06717908e-01
-2.83286482e-01 -5.98909914e-01 -6.24605775e-01 -7.77880073e-01
1.50364488e-01 8.93993378e-01 -6.91139281e-01 -4.17244822e-01
-1.95342973e-01 -1.26354778e+00 3.44421923e-01 4.17477697e-01
-1.70158878e-01 -7.54146338e-01 1.82090431e-01 3.96461308e-01
2.18743667e-01 -9.30986762e-01 -4.37631577e-01 3.40348810e-01
-9.48593140e-01 -1.15717745e+00 -8.31518531e-01 -4.88937557e-01
6.33012414e-01 -1.32190613e-02 8.66030931e-01 -3.85198236e-01
-4.68266577e-01 -1.77112117e-01 -1.12347692e-01 -5.28461933e-01
-6.17779016e-01 5.34570277e-01 -7.79778734e-02 -3.93528730e-01
3.57253104e-01 -6.50786638e-01 -8.76501620e-01 -2.50128388e-01
-8.03649724e-01 -1.13346905e-01 6.88441873e-01 7.68387437e-01
7.70903409e-01 -1.34182930e-01 8.34007621e-01 -1.42158353e+00
6.74981534e-01 -7.42765248e-01 -5.93223512e-01 -1.17387123e-01
-1.07059562e+00 5.94931662e-01 6.26237690e-01 -3.98136914e-01
-8.63557816e-01 -3.84797230e-02 -3.34780693e-01 -5.89399002e-02
2.17735901e-01 7.13894844e-01 -3.00304949e-01 -2.39893258e-01
7.90058672e-01 1.37308478e-01 1.40629172e-01 -3.94641429e-01
4.97981757e-01 1.75034344e-01 -2.64671326e-01 -3.93543601e-01
3.43043625e-01 2.92159349e-01 5.12402594e-01 -8.26957345e-01
-5.55555701e-01 -3.08262259e-01 -2.14707181e-01 2.78687239e-01
8.10460508e-01 -9.71018612e-01 -7.71970928e-01 -1.29249349e-01
-8.09720457e-01 -4.44511592e-01 -2.24628106e-01 4.62319016e-01
-2.03682289e-01 3.77635360e-01 -5.44665396e-01 -3.28940004e-01
-4.64544326e-01 -1.31859314e+00 8.99012625e-01 9.38229784e-02
-3.32376599e-01 -1.07408857e+00 5.46569884e-01 3.33628535e-01
2.18261376e-01 3.92692834e-01 1.27925217e+00 -9.18753624e-01
-8.48429263e-01 1.21374846e-01 1.13417454e-01 1.63077459e-01
2.43273258e-01 -2.04991728e-01 -8.92882943e-01 -2.54512906e-01
-5.48696399e-01 -1.95879325e-01 1.13557005e+00 6.75856769e-01
1.03702211e+00 -1.97589487e-01 -7.76485801e-01 6.23787224e-01
1.40799034e+00 4.83619571e-01 7.25368142e-01 4.99046706e-02
9.05280292e-01 2.25605667e-01 1.09898061e-01 5.19368708e-01
2.33797953e-02 4.90396976e-01 3.06166112e-01 -1.98553264e-01
-2.61472017e-01 -5.31283438e-01 1.41045287e-01 2.61441618e-01
-8.14026594e-02 -4.64094609e-01 -6.39794767e-01 4.14153606e-01
-1.21270132e+00 -1.04835427e+00 -2.33497974e-02 2.21342397e+00
1.11933374e+00 -1.84841435e-02 2.36462310e-01 -4.01014835e-01
3.57724011e-01 7.62369931e-02 -7.86931098e-01 -1.47870198e-01
-1.26436412e-01 9.00078833e-01 5.38933218e-01 5.47901094e-01
-7.44511425e-01 1.16234863e+00 6.78325844e+00 9.98503447e-01
-1.21992350e+00 -1.97106823e-01 8.59472513e-01 -1.71301603e-01
-8.07615459e-01 6.88193291e-02 -9.21283901e-01 4.01193917e-01
1.06312597e+00 -2.55680084e-01 3.72177213e-01 6.79086864e-01
3.56519341e-01 2.33130634e-01 -1.20814753e+00 5.25925517e-01
-1.55926552e-02 -2.03450084e+00 5.20936728e-01 5.01574218e-01
9.50346529e-01 3.65796536e-02 3.15791845e-01 1.68118551e-02
2.85585046e-01 -1.40078628e+00 1.27010345e-01 1.75913826e-01
8.74079287e-01 -1.07088029e+00 4.95252490e-01 -1.11775205e-01
-5.75496435e-01 3.04283321e-01 -2.05344200e-01 2.63020009e-01
-8.93499106e-02 5.76853395e-01 -1.50890386e+00 4.54391927e-01
-2.00815111e-01 7.18290567e-01 -4.30411041e-01 9.22706366e-01
-1.70134559e-01 8.56509984e-01 -7.35851303e-02 -4.19948190e-01
2.31386811e-01 -1.85040295e-01 4.68997449e-01 1.03844428e+00
5.99519238e-02 9.43139847e-03 -4.20262925e-02 1.25453126e+00
-5.84979057e-01 2.35602915e-01 -6.63281381e-01 -7.03546643e-01
4.86446470e-01 8.78045678e-01 -6.61108971e-01 -1.07811645e-01
-1.19968675e-01 7.49373019e-01 1.02587007e-01 5.21852434e-01
-8.14675868e-01 -3.53793323e-01 6.01106524e-01 3.42183948e-01
5.15941441e-01 -9.56339911e-02 -2.38217101e-01 -8.33001077e-01
-6.33105278e-01 -1.06000602e+00 2.39669457e-01 -2.83491910e-01
-8.44734013e-01 3.42845649e-01 -1.77359968e-01 -5.61203837e-01
2.94281822e-02 -5.94908714e-01 -2.29717746e-01 8.05095911e-01
-1.25229740e+00 -9.61067557e-01 2.37294897e-01 7.48324096e-02
4.72903818e-01 -1.69149861e-01 7.19570875e-01 1.87192693e-01
-5.60007811e-01 5.02490461e-01 4.77702647e-01 -4.52730179e-01
6.27001584e-01 -1.24380577e+00 4.04682755e-01 4.07658607e-01
4.18238461e-01 8.12398553e-01 8.31400275e-01 -9.15997088e-01
-1.42357051e+00 -1.07397985e+00 5.84833086e-01 -7.48969495e-01
4.87168044e-01 -3.13568771e-01 -6.62220061e-01 3.12736571e-01
-7.20106289e-02 -5.46222150e-01 1.17289340e+00 3.53723854e-01
-3.33419532e-01 1.22133210e-01 -1.05681813e+00 6.48408413e-01
9.33639586e-01 -4.40742284e-01 2.79836822e-02 5.94568372e-01
8.53628337e-01 -9.89819840e-02 -1.13768947e+00 3.08227807e-01
4.42222416e-01 -6.50499701e-01 1.00748909e+00 -9.25175309e-01
6.92745984e-01 -8.02891925e-02 5.92033379e-02 -1.25544465e+00
-7.02530265e-01 -6.91458941e-01 -1.06364250e-01 7.04076231e-01
8.60036671e-01 -5.70518613e-01 1.14825511e+00 3.25808555e-01
-4.30796742e-02 -1.10653102e+00 -6.78086877e-01 -4.26307172e-01
4.06494737e-01 -9.75733325e-02 5.86440325e-01 6.51943207e-01
1.05720229e-01 8.62413585e-01 -1.16211124e-01 -1.32663906e-01
4.40615267e-01 -1.82879996e-02 4.77195472e-01 -1.01103580e+00
-6.83197081e-01 -5.75961232e-01 -2.25858122e-01 -8.13632250e-01
6.46486655e-02 -1.22962880e+00 -5.02644718e-01 -1.56863511e+00
6.84333384e-01 -1.99483976e-01 -1.92340598e-01 2.71789521e-01
-3.16924691e-01 3.78101431e-02 -1.94702670e-01 -4.04763781e-02
-3.70641619e-01 7.65248358e-01 1.38838184e+00 -3.22747350e-01
-2.82611161e-01 -4.93938662e-02 -1.19709098e+00 9.20009837e-02
6.47121191e-01 -3.76008302e-01 -6.82352960e-01 1.62111744e-01
3.23395610e-01 -2.36900151e-01 -1.01603260e-02 -5.47036707e-01
-1.93727970e-01 -2.23600671e-01 5.50021827e-01 -1.11331150e-01
-2.62708822e-03 -2.00826153e-01 3.86683851e-01 7.41655886e-01
-3.90754968e-01 -4.53908831e-01 3.95812303e-01 9.71152902e-01
5.45005240e-02 8.48394353e-03 6.42376959e-01 -7.89133385e-02
-2.19956443e-01 5.49902916e-01 -3.21673304e-01 1.10640809e-01
9.01925445e-01 2.04318296e-02 -1.24371149e-01 -1.85964391e-01
-5.56559443e-01 -2.18648389e-01 4.46531802e-01 9.84767154e-02
4.36328053e-01 -9.10240412e-01 -4.29469109e-01 7.00300559e-02
5.05305864e-02 -8.15741494e-02 1.26373976e-01 6.67020261e-01
-5.54155290e-01 7.14285135e-01 2.52164721e-01 -3.35267425e-01
-1.28219378e+00 5.98913193e-01 3.51534396e-01 -3.83869052e-01
-1.40801713e-01 1.01104176e+00 4.56806093e-01 9.33200344e-02
-3.50433104e-02 -1.15256481e-01 -1.06179960e-01 -5.75210713e-02
2.24902928e-01 2.97693282e-01 1.53867081e-01 -6.63087964e-02
-4.06584918e-01 1.27656758e-01 -4.67841417e-01 3.20651174e-01
1.42882156e+00 5.94678521e-01 -1.15058094e-03 -7.08999634e-02
1.12660301e+00 1.53844163e-01 -1.16267383e+00 2.10574239e-01
-1.73899740e-01 -1.01556420e-01 1.63171127e-01 -8.82387340e-01
-6.99572504e-01 6.12670898e-01 4.74575728e-01 -2.34534159e-01
5.88316023e-01 3.18789274e-01 7.15337574e-01 2.51980573e-01
4.23760772e-01 -5.60251474e-01 2.06586361e-01 1.44456580e-01
4.63071227e-01 -1.05903935e+00 3.94305587e-01 -3.64764571e-01
-6.61863744e-01 8.53250682e-01 1.38146162e-01 1.15281798e-01
3.91354203e-01 -1.80347264e-01 -6.27183676e-01 -6.36156499e-01
-7.02865064e-01 -2.19010457e-01 4.34883803e-01 5.38264811e-01
7.90740728e-01 2.80058354e-01 -3.78643811e-01 3.70557696e-01
1.57910332e-01 2.03631762e-02 2.13183016e-01 5.29471159e-01
-3.83123875e-01 -1.63909805e+00 1.93833634e-01 6.00452483e-01
-7.03091025e-01 -6.36577249e-01 -6.44076526e-01 6.73433721e-01
1.37643233e-01 6.52389288e-01 -3.23359489e-01 -2.18487158e-01
9.88707989e-02 6.29021749e-02 8.72266591e-01 -7.97826231e-01
-3.83594871e-01 2.20158905e-01 2.56356746e-01 -2.36729577e-01
-7.58417696e-02 -5.32505810e-01 -1.01793063e+00 -2.13922799e-01
-5.92760921e-01 4.72504348e-01 4.32003230e-01 7.69743919e-01
9.86072004e-01 7.45611131e-01 5.35032809e-01 -3.95745367e-01
-5.35480738e-01 -6.53914809e-01 -4.04403567e-01 1.96402460e-01
9.37027037e-02 -6.32110655e-01 -7.82706812e-02 -3.46337706e-02]
|
[5.009668350219727, 5.754953861236572]
|
49419a5b-e06f-4e1c-95a1-4781c67333fd
|
a-volumetric-transformer-for-accurate-3d
|
2111.13300
| null |
https://arxiv.org/abs/2111.13300v2
|
https://arxiv.org/pdf/2111.13300v2.pdf
|
A Robust Volumetric Transformer for Accurate 3D Tumor Segmentation
|
We propose a Transformer architecture for volumetric segmentation, a challenging task that requires keeping a complex balance in encoding local and global spatial cues, and preserving information along all axes of the volume. Encoder of the proposed design benefits from self-attention mechanism to simultaneously encode local and global cues, while the decoder employs a parallel self and cross attention formulation to capture fine details for boundary refinement. Empirically, we show that the proposed design choices result in a computationally efficient model, with competitive and promising results on the Medical Segmentation Decathlon (MSD) brain tumor segmentation (BraTS) Task. We further show that the representations learned by our model are robust against data corruptions. \href{https://github.com/himashi92/VT-UNet}{Our code implementation is publicly available}.
|
['Mehrtash Harandi', 'Gary Egan', 'Zhaolin Chen', 'Munawar Hayat', 'Himashi Peiris']
|
2021-11-26
| null | null | null | null |
['volumetric-medical-image-segmentation']
|
['medical']
|
[ 8.34903568e-02 4.54072207e-01 -2.22674802e-01 -4.64176863e-01
-1.03782272e+00 -1.11719064e-01 3.40304375e-01 1.40885890e-01
-2.89601952e-01 6.27167583e-01 5.47441542e-01 -2.32559428e-01
-1.63429882e-02 -6.58125520e-01 -7.34931886e-01 -7.93340504e-01
-5.18816523e-02 4.54123199e-01 2.29265973e-01 -2.86598988e-02
2.01671764e-01 4.97288525e-01 -8.51798713e-01 2.67443061e-01
8.90137911e-01 1.26752377e+00 3.63402754e-01 7.49218762e-01
1.58959314e-01 9.24824536e-01 -1.24453284e-01 -2.16087624e-01
1.80121541e-01 -2.26691738e-01 -1.18078589e+00 -9.38817393e-03
4.63115931e-01 -4.14845884e-01 -6.69566035e-01 1.08484268e+00
5.89582980e-01 -1.98360667e-01 6.82191253e-01 -7.67007232e-01
-7.95273364e-01 4.30825204e-01 -8.08795810e-01 7.78184116e-01
-2.59316146e-01 1.38385862e-01 1.06768501e+00 -8.41066301e-01
6.75341666e-01 8.21415544e-01 5.85813999e-01 6.00330055e-01
-1.21004426e+00 -4.92213517e-01 2.40367562e-01 9.36432928e-02
-1.30240524e+00 -5.33747375e-01 7.31126964e-01 -4.79558796e-01
8.09075236e-01 3.96238118e-01 7.55520046e-01 1.08012021e+00
6.34258747e-01 9.14270401e-01 1.06953394e+00 -7.19101867e-03
7.58763254e-02 -3.45650494e-01 4.76760864e-01 9.86108422e-01
2.19728500e-01 -5.99886924e-02 -3.57512981e-01 9.19285268e-02
1.09977925e+00 -9.17983800e-03 -5.25929391e-01 -3.37345690e-01
-1.07023478e+00 8.46841991e-01 9.09844279e-01 5.41663110e-01
-3.77707154e-01 3.61324549e-01 3.14092785e-01 1.80049483e-02
5.96927345e-01 2.64306724e-01 -2.75869876e-01 2.11198226e-01
-1.10328805e+00 7.80915618e-02 1.43391833e-01 9.73953247e-01
4.29447353e-01 1.01635829e-01 -5.69361031e-01 6.63895130e-01
1.99736670e-01 3.26909035e-01 7.06686378e-01 -8.84446681e-01
3.27668279e-01 3.62996906e-01 -3.29278916e-01 -8.83035123e-01
-7.26070642e-01 -7.35625803e-01 -1.22427428e+00 5.13448007e-02
3.61367136e-01 -2.87315268e-02 -1.28626776e+00 1.76176512e+00
1.39497831e-01 1.38543144e-01 -3.12258929e-01 9.09933448e-01
1.08435166e+00 2.37627417e-01 8.67296830e-02 -1.82832815e-02
1.46440029e+00 -1.09438848e+00 -7.97958553e-01 -4.33355361e-01
5.27241588e-01 -4.15145695e-01 7.06523776e-01 3.69330756e-02
-1.44374287e+00 -3.07557136e-01 -9.89467025e-01 -7.18701005e-01
1.93436369e-02 9.79974568e-02 5.82544327e-01 3.84132236e-01
-1.32608473e+00 5.60689509e-01 -1.25596559e+00 3.45727988e-02
1.11824214e+00 4.33934301e-01 -2.48925030e-01 2.30102483e-02
-8.21137786e-01 6.91272974e-01 9.47012603e-02 1.54118046e-01
-1.01231527e+00 -8.03517997e-01 -8.65591824e-01 1.15409195e-01
7.67737478e-02 -8.32909107e-01 1.19888246e+00 -7.72482634e-01
-1.37599206e+00 1.10449052e+00 -4.41315264e-01 -6.23271465e-01
5.81252456e-01 -1.94549233e-01 6.86686784e-02 4.59242731e-01
1.40221268e-01 8.79230738e-01 7.56659925e-01 -1.01855612e+00
-2.52866536e-01 -6.69487596e-01 -3.69108319e-01 1.12593986e-01
-6.93628490e-02 -1.36857972e-01 -5.12385726e-01 -9.40744996e-01
3.28861535e-01 -6.59068763e-01 -4.98483121e-01 -1.07596582e-02
-7.40640461e-01 3.19016337e-01 5.31464458e-01 -1.05843580e+00
1.11038673e+00 -1.94440854e+00 2.96434104e-01 2.58718491e-01
6.33600354e-01 7.21287588e-03 9.51289982e-02 -2.25269914e-01
-2.39976972e-01 1.42930388e-01 -6.97836876e-01 -5.80723703e-01
-2.35486686e-01 1.25697717e-01 -3.58488932e-02 6.37660205e-01
1.72457129e-01 1.41619885e+00 -5.82164645e-01 -5.56242645e-01
2.67796572e-02 5.76217353e-01 -6.59143865e-01 1.35985568e-01
3.16433306e-03 7.19979942e-01 -6.03591084e-01 7.59414077e-01
6.22000992e-01 -5.81089258e-01 -2.44258195e-02 -1.47790447e-01
1.72453254e-01 3.26800793e-01 -6.77352190e-01 1.98214376e+00
-1.93714485e-01 5.34781098e-01 4.32865292e-01 -9.54185784e-01
6.28236949e-01 2.84651726e-01 7.57433593e-01 -9.46311653e-01
6.14334524e-01 2.90824249e-02 -5.62232248e-02 -2.15507209e-01
2.08799109e-01 7.72520434e-03 8.54571536e-02 2.79770225e-01
1.66183680e-01 -2.04446353e-02 -4.75544743e-02 2.11859122e-01
1.27238119e+00 -1.51626959e-01 2.93312132e-01 -6.30476177e-01
1.92909628e-01 -7.44266137e-02 7.06821501e-01 5.41212738e-01
-4.91159290e-01 9.22293603e-01 6.80188119e-01 -3.90170038e-01
-1.04016745e+00 -1.09073830e+00 -3.98140222e-01 6.11238897e-01
9.42170620e-02 -3.78352165e-01 -9.66100574e-01 -6.08597457e-01
-2.29772821e-01 3.72638017e-01 -9.56343949e-01 -1.01671219e-01
-6.34501457e-01 -8.64134729e-01 4.32253510e-01 7.44822025e-01
4.79311705e-01 -8.73100758e-01 -8.34629357e-01 1.68746516e-01
-1.70369372e-01 -9.82214034e-01 -7.13116586e-01 2.52753049e-01
-1.10838151e+00 -1.06682539e+00 -9.47503686e-01 -7.30573833e-01
8.50121915e-01 1.23098930e-02 1.06440449e+00 2.12079778e-01
-5.95445812e-01 1.47298127e-01 1.56267099e-02 -1.14341021e-01
-1.89948082e-02 2.79169142e-01 -4.66706187e-01 -1.49906501e-01
-1.76354349e-01 -6.18315458e-01 -9.18135405e-01 2.30321996e-02
-8.56249332e-01 4.91270214e-01 4.61779684e-01 9.56024349e-01
1.00562537e+00 -4.78869945e-01 3.39430094e-01 -1.06851792e+00
3.70269626e-01 -5.45524299e-01 -4.57214266e-01 -4.30777222e-02
-3.49624068e-01 7.46439174e-02 3.01335275e-01 1.79286882e-01
-7.10153401e-01 -4.83035343e-03 -6.15406394e-01 -5.03461182e-01
-1.55220211e-01 1.22763194e-01 -1.94143727e-01 -3.59383747e-02
2.25387394e-01 2.10491478e-01 1.35636091e-01 -4.16110724e-01
1.38087496e-01 2.25313142e-01 5.24124682e-01 -4.65770572e-01
3.05322438e-01 7.34488130e-01 -5.80530288e-03 -6.24287963e-01
-8.08558822e-01 -1.57747939e-01 -8.05764735e-01 7.98690785e-03
1.18572438e+00 -8.69718254e-01 -3.79473835e-01 4.18719202e-01
-9.81007934e-01 -6.10830367e-01 -4.58727092e-01 1.99283212e-01
-8.91371548e-01 1.34412229e-01 -9.27786052e-01 -1.47077024e-01
-6.87395632e-01 -1.60973537e+00 1.21468949e+00 2.48512909e-01
-1.83909729e-01 -9.92244601e-01 4.37029935e-02 4.40326184e-01
5.36123455e-01 4.83141005e-01 1.03703475e+00 -5.16876101e-01
-8.16410661e-01 2.95348525e-01 -3.72426093e-01 8.86561871e-02
1.34379223e-01 -3.91590893e-01 -1.04657006e+00 -4.75784898e-01
-2.24322011e-03 -1.80196539e-01 1.20838380e+00 8.83678019e-01
1.75060630e+00 -3.15199852e-01 -4.48234975e-01 1.17943084e+00
1.32774699e+00 -3.45877148e-02 6.18935108e-01 7.06817433e-02
8.62336218e-01 2.23475128e-01 7.75961578e-02 3.40976864e-01
4.89449918e-01 4.09128934e-01 4.79587227e-01 -5.47448397e-01
-5.31247675e-01 1.01814985e-01 -1.58980951e-01 7.91383386e-01
-4.43253964e-02 -7.24014863e-02 -1.04114890e+00 7.21332610e-01
-1.72426665e+00 -8.67422819e-01 -7.80653805e-02 1.84352493e+00
8.82133782e-01 5.87925799e-02 -1.87287033e-01 -7.51532167e-02
4.05665398e-01 3.71696025e-01 -6.44588768e-01 -3.47308189e-01
-5.10675721e-02 4.86858100e-01 7.78489828e-01 7.80366480e-01
-1.24021780e+00 9.09400642e-01 6.55007362e+00 7.68064499e-01
-1.05216205e+00 5.45392513e-01 1.31640756e+00 -2.13442490e-01
-4.63042051e-01 -4.21138912e-01 -5.01213253e-01 3.86012524e-01
7.72995353e-01 6.10920936e-02 7.61461630e-02 3.58055681e-01
1.73608869e-01 -4.25084978e-02 -8.60231698e-01 9.12923634e-01
5.16848974e-02 -1.66413569e+00 -2.08016589e-01 1.34098530e-01
7.26942062e-01 3.44099522e-01 2.68100142e-01 -9.96189117e-02
3.25476825e-01 -1.30130482e+00 6.90373182e-01 6.81297660e-01
9.02599454e-01 -6.10277474e-01 4.71098423e-01 -4.19353396e-02
-1.07326055e+00 2.94823479e-02 -1.27642438e-01 5.05287707e-01
1.66142527e-02 7.31961191e-01 -6.16859078e-01 5.01762986e-01
7.33693302e-01 7.98634648e-01 -7.23430812e-01 1.11733747e+00
-7.01844841e-02 5.77228844e-01 -9.62882116e-02 5.71276724e-01
1.04987547e-01 -8.04719180e-02 6.09194994e-01 1.12800026e+00
1.60748348e-01 1.96399853e-01 -4.52637449e-02 1.07649362e+00
-2.42981121e-01 2.65911724e-02 -4.35540795e-01 3.40425760e-01
9.80615765e-02 1.02875566e+00 -9.59102392e-01 -3.16249847e-01
-1.94282308e-01 9.38011527e-01 6.76336050e-01 4.06413108e-01
-9.66466129e-01 -4.09335680e-02 8.27471197e-01 3.43703419e-01
4.46453691e-01 -2.79574156e-01 -8.74399960e-01 -1.02728248e+00
-2.11478006e-02 -6.57275558e-01 4.97452140e-01 -5.97064853e-01
-1.02208042e+00 7.96657503e-01 -3.51307124e-01 -7.81663656e-01
1.59114495e-01 -5.04333496e-01 -5.99381328e-01 7.85922408e-01
-1.61292994e+00 -1.21436012e+00 -4.05960262e-01 6.97449088e-01
4.56189066e-01 1.15221620e-01 6.15377128e-01 2.83723354e-01
-8.03897738e-01 7.10329473e-01 -1.25331134e-02 3.57972473e-01
3.33214790e-01 -1.20114565e+00 3.54124129e-01 8.01753581e-01
-6.64254650e-03 2.75237739e-01 3.55183780e-01 -7.55837798e-01
-1.01571190e+00 -1.25342190e+00 6.59662068e-01 -1.57388270e-01
5.06451547e-01 -3.59305590e-01 -8.84066939e-01 1.00011289e+00
3.47588748e-01 2.57766813e-01 5.47634602e-01 -2.27822110e-01
-1.44975990e-01 9.40595418e-02 -1.26975405e+00 3.44167560e-01
1.10993123e+00 -2.75193512e-01 -4.60791886e-01 3.86292249e-01
6.90904200e-01 -8.26665938e-01 -9.73533332e-01 4.56191063e-01
1.66871727e-01 -9.61035848e-01 9.80028629e-01 -4.14037645e-01
5.80954194e-01 -1.55880651e-03 -3.72891724e-02 -1.16024804e+00
-6.36479139e-01 -3.96263212e-01 -1.34042203e-01 7.85497725e-01
4.17205453e-01 -6.67785525e-01 8.47677231e-01 5.58522522e-01
-5.84793389e-01 -1.23113859e+00 -1.34081018e+00 -2.50236094e-01
4.45816636e-01 -2.94599414e-01 5.14730990e-01 7.07726419e-01
-1.62646532e-01 7.73889478e-03 -8.80561173e-02 2.17698336e-01
6.33927464e-01 9.67714936e-02 9.38536078e-02 -8.14257860e-01
-2.38113731e-01 -7.09445179e-01 -4.01658028e-01 -1.03211641e+00
2.43488997e-02 -1.23539650e+00 -1.84840128e-01 -1.81549346e+00
3.68369699e-01 -3.87644440e-01 -4.42018807e-01 6.12304568e-01
-2.86357552e-02 3.52519006e-01 1.56933933e-01 1.95876613e-01
-4.64265764e-01 7.28414536e-01 1.66043711e+00 -2.85640657e-01
1.32566959e-01 -2.01724321e-01 -9.15824175e-01 5.88375807e-01
9.39465642e-01 -4.10555184e-01 -2.28596240e-01 -8.25859189e-01
-2.73007989e-01 1.66606694e-01 6.11857235e-01 -9.52370346e-01
1.79035529e-01 9.70970616e-02 6.17225766e-01 -3.79356146e-01
5.07000744e-01 -6.00541890e-01 -2.34648719e-01 5.20634115e-01
-3.22299123e-01 1.73101857e-01 3.39911103e-01 3.37738305e-01
-1.53474435e-01 7.37496242e-02 1.16813588e+00 -1.58500284e-01
-3.22542310e-01 8.35057259e-01 -2.60785431e-01 3.16166848e-01
1.07319915e+00 1.64298981e-03 -2.94189870e-01 -3.60645503e-02
-1.17948973e+00 3.31074417e-01 3.84688973e-01 1.41133174e-01
6.97142243e-01 -1.25128508e+00 -7.68227577e-01 4.35840756e-01
-3.52418631e-01 3.19518685e-01 5.14593482e-01 1.16360343e+00
-7.17013121e-01 5.29341340e-01 -3.05843115e-01 -6.87840939e-01
-9.79295313e-01 1.88665643e-01 7.13132858e-01 -5.56683302e-01
-1.07336497e+00 1.20727849e+00 4.49963629e-01 -2.42936816e-02
3.02730650e-01 -6.57305360e-01 -1.07564509e-01 -1.27255544e-01
5.37126660e-01 3.84133644e-02 3.56597126e-01 -7.32015371e-01
-4.01230186e-01 5.02692699e-01 -2.91190058e-01 1.21606387e-01
1.48998284e+00 -1.52373418e-01 1.95538290e-02 1.89193320e-02
1.17796433e+00 -2.24608064e-01 -1.49953926e+00 -2.12653518e-01
-1.07986219e-01 -3.87797803e-01 5.79099238e-01 -7.08189368e-01
-1.65820980e+00 9.29780185e-01 7.15387821e-01 -1.60024166e-01
1.26862192e+00 7.42873549e-02 1.10576141e+00 -3.46499950e-01
1.93529606e-01 -8.14727485e-01 1.20956767e-02 5.55832982e-01
1.05487037e+00 -1.14114034e+00 -1.48298696e-01 -3.65351558e-01
-6.71881914e-01 7.58377969e-01 6.48542404e-01 -3.90647322e-01
7.89398909e-01 5.66300333e-01 -1.15473635e-01 -5.38683712e-01
-7.27506816e-01 -1.66978776e-01 3.18491548e-01 4.20781523e-01
4.95894939e-01 1.87930927e-01 -7.69244432e-02 7.90578067e-01
-2.18847707e-01 -1.43271402e-01 3.29469323e-01 8.34451795e-01
-3.74856979e-01 -8.62425625e-01 -2.68224210e-01 6.77245796e-01
-4.83771533e-01 -1.91204011e-01 -1.52309939e-01 5.72584569e-01
1.40817598e-01 4.16486531e-01 3.70697141e-01 3.89704071e-02
1.71196878e-01 -7.02965260e-02 6.01754904e-01 -4.60456759e-01
-6.05177522e-01 2.10769325e-01 -1.97661221e-01 -9.02615905e-01
2.84980759e-02 -7.11793184e-01 -1.48440862e+00 -1.88610464e-01
2.63461888e-01 -2.23054752e-01 2.49054387e-01 7.60691464e-01
4.66553420e-01 1.14022291e+00 4.07353371e-01 -7.13382363e-01
-1.74439654e-01 -7.77116358e-01 -5.38578987e-01 2.55221665e-01
6.32088363e-01 -6.82343662e-01 5.77188171e-02 -1.34780839e-01]
|
[14.571266174316406, -2.5321221351623535]
|
28a9141d-1f4d-4aac-944f-d80416516be4
|
towards-more-discriminative-and-robust-iris
| null | null |
https://ieeexplore.ieee.org/abstract/document/9722888
|
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9722888
|
Towards More Discriminative and Robust Iris Recognition by Learning Uncertain Factors
|
The uncontrollable acquisition process limits the performance of iris recognition. In the acquisition process, various inevitable factors, including eyes, devices, and environment, hinder the iris recognition system from learning a discriminative identity representation. This leads to severe performance degradation. In this paper, we explore uncertain acquisition factors and propose uncertainty embedding (UE) and uncertainty-guided curriculum learning (UGCL) to mitigate the influence of acquisition factors. UE represents an iris image using a probabilistic distribution rather than a deterministic point (binary template or feature vector) that is widely adopted in iris recognition methods. Specifically, UE learns identity and uncertainty features from the input image, and encodes them as two independent components of the distribution, mean and variance. Based on this representation, an input image can be regarded as an instantiated feature sampled from the UE, and we can also generate various virtual features through sampling. UGCL is constructed by imitating the progressive learning process of newborns. Particularly, it selects virtual features to train the model in an easy-to-hard order at different training stages according to their uncertainty. In addition, an instance-level enhancement method is developed by utilizing local and global statistics to mitigate the data uncertainty from image noise and acquisition conditions in the pixel-level space. The experimental results on six benchmark iris datasets verify the effectiveness and generalization ability of the proposed method on same-sensor and cross-sensor recognition.
|
['Jianze Wei; Huaibo Huang; Yunlong Wang; Ran He; Zhenan Sun']
|
2022-02-28
| null | null | null |
ieee-transactions-on-information-forensics-9
|
['iris-recognition']
|
['computer-vision']
|
[ 4.12226051e-01 -2.56589085e-01 -3.02755088e-01 -4.78375793e-01
-6.26201093e-01 -1.95295379e-01 1.72837839e-01 -3.69795978e-01
-2.89869696e-01 5.21076500e-01 -2.60971915e-02 3.60581428e-02
-6.88331366e-01 -4.91278619e-01 -6.52236521e-01 -1.11789513e+00
3.39510441e-01 -1.85559839e-02 -3.81087124e-01 4.50548142e-01
3.49216521e-01 2.46601135e-01 -1.89384282e+00 -1.89412057e-01
1.22208214e+00 1.26218462e+00 -9.07815322e-02 2.57773876e-01
-5.25098853e-03 3.10438216e-01 -8.32200825e-01 -4.04009670e-02
2.75670528e-01 -3.88198704e-01 -7.73113668e-02 4.47442085e-01
3.44776005e-01 -4.02583361e-01 -3.74687046e-01 1.40891707e+00
5.88096321e-01 9.73366499e-02 7.34267116e-01 -9.20568168e-01
-8.08347762e-01 3.80592555e-01 -6.74001753e-01 -1.19872775e-03
2.50000715e-01 5.53771317e-01 2.97379524e-01 -6.13493264e-01
1.00840285e-01 1.08216631e+00 1.80079043e-01 5.81225812e-01
-1.06014705e+00 -6.74478292e-01 1.01340160e-01 1.65524006e-01
-1.68611014e+00 -1.52953580e-01 8.16209078e-01 -4.63810772e-01
2.17522696e-01 2.53379762e-01 4.55196589e-01 9.98720229e-01
3.57805282e-01 1.04051399e+00 1.50427818e+00 -4.06280130e-01
2.91751288e-02 3.57565194e-01 5.00466786e-02 4.74138379e-01
1.77964434e-01 8.80006075e-01 -3.85266602e-01 -4.72119190e-02
8.57445359e-01 5.31090098e-03 -6.93090796e-01 -2.29725972e-01
-9.65863466e-01 2.91246682e-01 3.37626249e-01 3.33222747e-02
-2.77409405e-01 -2.44683027e-01 1.34204865e-01 1.46417141e-01
-1.90983638e-01 2.83226460e-01 -3.67754161e-01 -1.80025727e-01
-5.96092522e-01 -2.41080716e-01 3.76848102e-01 8.09011221e-01
6.60131156e-01 1.62999019e-01 -4.99626279e-01 5.50786018e-01
4.39333111e-01 8.08828712e-01 5.83900928e-01 -2.98738420e-01
1.41198516e-01 6.53666556e-01 5.94375394e-02 -1.05357301e+00
-1.18455924e-01 -7.52691567e-01 -9.88851190e-01 2.06426591e-01
2.86208808e-01 -1.91078991e-01 -1.11532784e+00 1.43662179e+00
4.41977024e-01 7.90694475e-01 4.08276208e-02 1.12136555e+00
5.93885005e-01 6.21027231e-01 -1.97223961e-01 -5.34224927e-01
1.00148165e+00 -2.80809939e-01 -7.96680152e-01 2.30739951e-01
-1.44815177e-01 -6.42304242e-01 1.04906154e+00 8.02246511e-01
-5.18114030e-01 -8.69948685e-01 -1.22189391e+00 3.15636575e-01
-6.13186508e-03 6.17238343e-01 3.25739682e-01 7.61968911e-01
-4.54604000e-01 4.71922070e-01 -7.01542616e-01 5.34748971e-01
3.75302881e-01 5.74416995e-01 -6.76948112e-03 -2.66580600e-02
-1.21522915e+00 5.33944607e-01 4.81803894e-01 4.49499249e-01
-6.34589076e-01 -5.53113461e-01 -9.17893112e-01 -6.02434762e-02
4.60206062e-01 -3.91602695e-01 6.64216399e-01 -8.07958961e-01
-1.93200552e+00 3.71023446e-01 -4.53252904e-02 -6.01484627e-03
2.15836480e-01 3.16485669e-03 -7.46251047e-01 6.07932964e-03
-4.58987653e-01 3.99151333e-02 1.62825882e+00 -1.02347279e+00
-6.38718426e-01 -5.70466518e-01 -2.94450670e-01 1.58085734e-01
-3.07470769e-01 -1.18504226e-01 -5.84298670e-01 -5.20087242e-01
1.39436126e-01 -6.13158882e-01 -1.57160033e-02 -2.00626940e-01
-4.05171007e-01 -2.11210653e-01 4.29719388e-01 -5.77210248e-01
1.49077594e+00 -2.52735591e+00 1.08238496e-01 6.63378239e-01
-4.94428016e-02 3.44198704e-01 -3.96394245e-02 -4.60615963e-01
8.53895955e-03 1.07206767e-02 -2.30804518e-01 3.98036055e-02
-5.71865477e-02 1.87870681e-01 -1.86933249e-01 4.45864111e-01
3.36451709e-01 5.58613241e-01 -8.68437290e-01 -5.04282475e-01
6.98294461e-01 6.10760033e-01 -1.26006722e-01 2.81272709e-01
-2.36771643e-01 6.01996899e-01 -7.71662712e-01 8.47784519e-01
9.18843269e-01 -3.99102457e-02 -3.52577716e-01 -3.35569978e-01
9.10571069e-02 -3.08184117e-01 -1.62965393e+00 1.51163530e+00
-4.19823349e-01 1.56951979e-01 -2.63200551e-01 -6.83876872e-01
1.16489685e+00 1.73331767e-01 7.92596042e-02 -4.36100692e-01
4.45158601e-01 1.02509245e-01 1.68040600e-02 -6.58560097e-01
1.17753901e-01 2.29471043e-01 4.50042784e-02 6.51263371e-02
1.75503138e-02 -6.67354614e-02 -4.50242639e-01 -4.53839183e-01
3.15599501e-01 2.49828398e-01 1.83792159e-01 4.18827981e-02
8.75082314e-01 -5.91608346e-01 9.01238263e-01 5.35505176e-01
-2.60324091e-01 4.22729582e-01 3.98638874e-01 -2.53263593e-01
-4.96761203e-01 -1.04078364e+00 -8.43550324e-01 2.43564546e-02
4.84474778e-01 -2.49210671e-02 -5.27095556e-01 -7.26478040e-01
9.64197814e-02 5.47763169e-01 -5.69724381e-01 -5.04634738e-01
-1.19752638e-01 -7.75397062e-01 1.28245831e-01 2.22016901e-01
4.68835890e-01 -7.38042772e-01 -2.61019081e-01 1.12994015e-01
2.99571007e-01 -5.38963079e-01 -6.39763236e-01 -3.52120429e-01
-6.57050014e-01 -1.16419637e+00 -3.37103099e-01 -4.50461477e-01
1.01140416e+00 -3.18225712e-01 4.23096895e-01 -1.20026790e-01
-6.57587647e-01 3.51769686e-01 -4.01447294e-03 -4.94177729e-01
-4.12617438e-02 -5.73192835e-01 4.22089458e-01 6.78035438e-01
7.13030815e-01 -1.63086802e-01 -6.65478706e-01 3.53158891e-01
-9.85730410e-01 -4.14849669e-01 7.94172227e-01 1.49196017e+00
1.04587758e+00 6.80982947e-01 2.76217371e-01 -3.54160428e-01
5.47687232e-01 -1.09770484e-01 -1.14702499e+00 3.40871871e-01
-9.31741118e-01 1.32595196e-01 6.00814283e-01 -7.24794090e-01
-1.18850195e+00 5.95443323e-02 2.68256009e-01 -8.55844021e-01
-1.32782638e-01 5.45683861e-01 -5.39762616e-01 -3.20672840e-01
6.34335637e-01 5.66808760e-01 3.12267184e-01 -2.00341478e-01
2.40902871e-01 1.14691508e+00 5.75179160e-01 -7.96838462e-01
7.68492520e-01 -2.08784267e-02 8.47404543e-03 -6.53831661e-01
-4.05482441e-01 -1.42141923e-01 -1.06513299e-01 -2.34623253e-01
5.15339136e-01 -7.82840848e-01 -9.43063259e-01 9.26419079e-01
-5.90451479e-01 3.46571445e-01 -3.15825611e-01 9.83494580e-01
-3.05433422e-01 3.21587831e-01 -3.73246938e-01 -9.37521279e-01
-1.31256044e-01 -1.48760390e+00 9.72657323e-01 1.15803707e+00
4.20008630e-01 -5.24810553e-01 -4.16459650e-01 1.97141811e-01
1.75255224e-01 3.12440813e-01 6.77744865e-01 -1.55692056e-01
-9.83046174e-01 -4.01412904e-01 -1.01944000e-01 8.53505909e-01
3.53977740e-01 3.65235567e-01 -1.00883067e+00 -4.08037156e-01
4.17935342e-01 -2.61054069e-01 6.44952536e-01 6.30506217e-01
1.71374619e+00 -2.47667477e-01 -2.96652943e-01 9.72349107e-01
1.56461775e+00 4.76521850e-01 7.52708137e-01 8.68910924e-03
3.74950081e-01 3.62531334e-01 7.73793876e-01 5.37422836e-01
-1.23662159e-01 3.84525180e-01 3.40598434e-01 1.57455713e-01
2.42502436e-01 -4.45237786e-01 6.96819723e-02 5.67384422e-01
8.15492496e-02 1.69633199e-02 -4.66586679e-01 3.42213273e-01
-1.56828582e+00 -7.63255775e-01 3.70075375e-01 2.70976162e+00
1.07566893e+00 8.63537788e-02 -5.43556094e-01 7.70896226e-02
7.33338833e-01 -1.01228267e-01 -1.04018450e+00 3.03422790e-02
-1.71806857e-01 3.02702874e-01 2.85280764e-01 4.33529437e-01
-9.92454946e-01 4.05411690e-01 5.08703279e+00 1.20789087e+00
-1.43244040e+00 -5.63553512e-01 8.52593899e-01 -5.56466216e-03
-2.99114645e-01 -2.20176563e-01 -8.97253931e-01 8.53575110e-01
4.98763114e-01 -1.83065996e-01 5.49971640e-01 8.19711983e-01
-7.58766532e-02 -1.21382149e-02 -9.14298236e-01 1.41496253e+00
1.24619320e-01 -9.57791924e-01 -8.93519353e-03 -7.02320337e-02
8.00502658e-01 -4.66860414e-01 8.07087064e-01 2.01610908e-01
-2.84138978e-01 -1.12424254e+00 -1.39809111e-02 1.19884968e+00
1.15867829e+00 -8.63640785e-01 8.51141632e-01 1.97940856e-01
-8.18514884e-01 -2.30809838e-01 -4.71213818e-01 3.19511563e-01
-3.01870733e-01 7.51405180e-01 -5.51935375e-01 8.93225312e-01
4.09148246e-01 6.31199419e-01 -3.90923411e-01 1.17049968e+00
-3.89216334e-01 5.46780169e-01 -4.68173921e-01 -1.07855871e-01
-2.47071132e-01 -5.58669090e-01 6.58480287e-01 4.85914439e-01
4.92748529e-01 3.89037549e-01 7.41975754e-02 9.79348004e-01
1.94317788e-01 5.34464419e-02 -1.87586010e-01 -1.12095863e-01
6.10078812e-01 9.38793480e-01 1.72254011e-01 4.01946669e-03
-2.00783432e-01 6.28693402e-01 -1.27359882e-01 5.17956316e-01
-6.24045491e-01 -6.47679806e-01 5.99096239e-01 -3.55937839e-01
4.05592173e-02 2.46658951e-01 -2.51370162e-01 -1.15441585e+00
3.80034417e-01 -1.06379116e+00 2.75913954e-01 -5.52354991e-01
-1.37025654e+00 5.38358092e-01 -1.64596438e-01 -1.57777679e+00
-1.67531282e-01 -6.11346900e-01 -5.39461970e-01 1.44465423e+00
-1.45688736e+00 -7.80655026e-01 -1.97792932e-01 7.38968670e-01
-3.02043445e-02 -4.85061198e-01 7.21841455e-01 2.07100529e-02
-9.69026625e-01 1.07751250e+00 3.39917362e-01 4.82402295e-02
6.73957586e-01 -1.12184632e+00 -3.10326338e-01 8.84521425e-01
-6.65468723e-02 8.11799288e-01 3.31541508e-01 -4.77269977e-01
-1.47076511e+00 -9.10172343e-01 2.56806672e-01 -2.79498070e-01
4.01727706e-01 1.81373209e-01 -9.29881096e-01 2.96938390e-01
-2.23656252e-01 2.10999414e-01 5.95154107e-01 -8.68258551e-02
-8.80337954e-02 -5.48071444e-01 -1.42022181e+00 4.94943768e-01
4.98886108e-01 -6.44729197e-01 -7.25006819e-01 7.03285113e-02
6.73353791e-01 -9.91547406e-01 -1.12767947e+00 7.76676297e-01
4.48965460e-01 -6.78804457e-01 6.80036485e-01 -3.00837696e-01
1.86464414e-01 -8.64540637e-01 1.57734051e-01 -1.29031503e+00
-8.10485259e-02 -6.72377825e-01 -2.13585988e-01 1.30539155e+00
6.61221668e-02 -8.29187691e-01 6.41064823e-01 6.93503439e-01
7.47262016e-02 -1.08486068e+00 -9.84574795e-01 -7.27778196e-01
-3.43839526e-01 -1.63919851e-01 1.07839787e+00 6.03047729e-01
-1.34837091e-01 -2.15113804e-01 -3.52490515e-01 8.47199976e-01
9.14195061e-01 2.72669643e-01 5.39691985e-01 -9.62244093e-01
-5.23158431e-01 -3.55309904e-01 -7.20011711e-01 -1.13718903e+00
-8.13964009e-02 -4.09360021e-01 3.69129367e-02 -7.62678444e-01
-2.20292985e-01 -6.30350947e-01 -6.92072868e-01 1.06986783e-01
-5.34295619e-01 -4.12087470e-01 -3.08430821e-01 1.29465863e-01
-3.15665849e-03 7.01929390e-01 1.50771177e+00 -4.86576080e-01
-4.19831961e-01 4.47265297e-01 -5.53958476e-01 3.54687810e-01
5.28056920e-01 5.06256288e-03 -7.16622293e-01 -1.75933614e-01
-2.13427186e-01 3.18195641e-01 9.37753916e-02 -9.26810563e-01
3.63534719e-01 -1.14058882e-01 7.63864040e-01 -3.56919914e-01
1.22756220e-01 -9.97542441e-01 7.15889558e-02 2.68029541e-01
-3.16868944e-04 -7.92041242e-01 3.59269559e-01 7.40587831e-01
-4.55143213e-01 -2.26435319e-01 7.04974592e-01 4.18869376e-01
-6.00108922e-01 6.70776010e-01 3.98820609e-01 -2.24964261e-01
9.42887127e-01 -3.59672189e-01 -2.56896913e-01 8.73748884e-02
-7.09207237e-01 4.78988796e-01 2.64275312e-01 3.96344006e-01
1.04526663e+00 -1.30259001e+00 -6.30323768e-01 1.03290451e+00
4.75670189e-01 3.24412197e-01 6.59762681e-01 5.83987117e-01
-9.24171228e-03 1.46892801e-01 4.36688773e-02 -1.05877650e+00
-1.09236872e+00 6.66146278e-01 6.11919701e-01 1.10732473e-01
-1.65772527e-01 9.77479100e-01 -1.13269135e-01 -2.23733738e-01
5.09719133e-01 -4.47472155e-01 -3.12343061e-01 -2.63657928e-01
9.02325571e-01 -1.32182660e-02 -1.24407716e-01 -1.01664558e-01
-2.65781023e-02 8.96626472e-01 -3.30939502e-01 2.02378809e-01
7.05983639e-01 -1.11647718e-01 -1.12408407e-01 1.92986935e-01
9.04108584e-01 -2.77131293e-02 -1.53332543e+00 -5.89169085e-01
-3.66102844e-01 -9.80479658e-01 3.04053515e-01 -1.01850653e+00
-9.79929447e-01 5.75319231e-01 1.14892864e+00 -2.83980191e-01
1.54849792e+00 -4.73128349e-01 4.72581804e-01 -1.71483718e-02
5.36314726e-01 -1.13645744e+00 -2.71835715e-01 -6.67018667e-02
7.05697298e-01 -1.30712020e+00 -4.63368706e-02 -1.47697017e-01
-5.48509717e-01 1.08379257e+00 8.19370925e-01 1.03870265e-01
8.35926592e-01 9.39389914e-02 2.30620161e-01 1.80533081e-01
-2.57046014e-01 1.32667378e-01 6.88910723e-01 7.55493283e-01
-2.59136230e-01 2.47539163e-01 -1.99937388e-01 8.73790920e-01
1.64283484e-01 3.12916785e-01 2.91190624e-01 3.95968676e-01
-1.37861386e-01 -1.07302904e+00 -5.47969818e-01 7.78232813e-01
-1.05416290e-01 2.52822391e-03 3.07907939e-01 5.21354914e-01
5.53465903e-01 8.35504472e-01 7.41332257e-03 -8.49282026e-01
2.70721436e-01 -1.92314193e-01 5.36816418e-01 -3.75227630e-01
-3.58214928e-03 5.70957288e-02 -5.70904553e-01 -5.09441435e-01
-1.85877353e-01 -7.20363498e-01 -1.14826238e+00 1.87754825e-01
-7.19247043e-01 1.55950710e-01 6.27846658e-01 7.83121705e-01
3.08927447e-01 5.42655647e-01 1.20075190e+00 -2.89663762e-01
-1.14433706e+00 -7.30033457e-01 -8.45871925e-01 1.05303444e-01
5.66850185e-01 -6.51192904e-01 -7.97751188e-01 -2.10605115e-01]
|
[13.102278709411621, 0.707775890827179]
|
ee573ea4-42d4-42a5-a383-da91a5ac5d5b
|
rda-reciprocal-distribution-alignment-for
|
2208.04619
| null |
https://arxiv.org/abs/2208.04619v2
|
https://arxiv.org/pdf/2208.04619v2.pdf
|
RDA: Reciprocal Distribution Alignment for Robust Semi-supervised Learning
|
In this work, we propose Reciprocal Distribution Alignment (RDA) to address semi-supervised learning (SSL), which is a hyperparameter-free framework that is independent of confidence threshold and works with both the matched (conventionally) and the mismatched class distributions. Distribution mismatch is an often overlooked but more general SSL scenario where the labeled and the unlabeled data do not fall into the identical class distribution. This may lead to the model not exploiting the labeled data reliably and drastically degrade the performance of SSL methods, which could not be rescued by the traditional distribution alignment. In RDA, we enforce a reciprocal alignment on the distributions of the predictions from two classifiers predicting pseudo-labels and complementary labels on the unlabeled data. These two distributions, carrying complementary information, could be utilized to regularize each other without any prior of class distribution. Moreover, we theoretically show that RDA maximizes the input-output mutual information. Our approach achieves promising performance in SSL under a variety of scenarios of mismatched distributions, as well as the conventional matched SSL setting. Our code is available at: https://github.com/NJUyued/RDA4RobustSSL.
|
['Yinghuan Shi', 'Luping Zhou', 'Lei Wang', 'Lei Qi', 'Yue Duan']
|
2022-08-09
| null | null | null | null |
['semi-supervised-image-classification']
|
['computer-vision']
|
[ 3.30053687e-01 3.11527669e-01 -4.91257280e-01 -5.47383249e-01
-1.03021026e+00 -8.83427799e-01 5.07548988e-01 1.00977175e-01
-1.07784815e-01 9.85616267e-01 -2.50990212e-01 -4.28450167e-01
-1.25423625e-01 -4.36204314e-01 -6.18347287e-01 -1.04004133e+00
5.95134795e-01 8.01375747e-01 1.80295065e-01 2.30003089e-01
6.84223697e-02 2.34230772e-01 -1.67308784e+00 3.20648611e-01
8.89420271e-01 7.92813480e-01 1.37099206e-01 4.32512939e-01
-1.79572240e-01 5.52639663e-01 -4.53587204e-01 -4.41685379e-01
3.18180740e-01 -5.75877964e-01 -6.83215022e-01 3.38410549e-02
3.09229732e-01 6.13058656e-02 -5.49696609e-02 1.32066560e+00
3.39497536e-01 -2.10022524e-01 1.08709216e+00 -1.75638318e+00
-5.48129380e-01 6.58546150e-01 -7.68726587e-01 -2.16150895e-01
2.25254074e-01 -7.82099217e-02 1.02787733e+00 -9.01841164e-01
3.89418036e-01 1.11562288e+00 6.08486056e-01 5.08081973e-01
-1.36634493e+00 -7.83543110e-01 8.36809725e-02 4.61241603e-02
-1.54034007e+00 -2.30752483e-01 5.46721041e-01 -5.39852262e-01
3.36904913e-01 1.77103490e-01 2.27960363e-01 1.13113320e+00
-6.34733588e-02 9.74979699e-01 1.37483799e+00 -7.31019080e-01
2.73208261e-01 5.07463932e-01 3.57965112e-01 4.91848171e-01
2.44048968e-01 2.09484264e-01 -5.66292942e-01 -5.62808990e-01
3.25003266e-01 -1.06493318e-02 -3.26975465e-01 -7.23228514e-01
-1.16060615e+00 6.70676291e-01 -4.79659475e-02 1.37391523e-01
4.70042750e-02 -3.41954470e-01 -2.66726478e-03 2.95812070e-01
3.65731388e-01 1.96160913e-01 -6.89579129e-01 2.61959970e-01
-7.27551639e-01 3.50195840e-02 8.74161839e-01 1.19594920e+00
8.76434624e-01 -3.65792453e-01 -3.00825328e-01 7.36403108e-01
5.29059589e-01 7.83623934e-01 4.31966960e-01 -7.80510843e-01
3.07972223e-01 5.06325483e-01 3.34373504e-01 -6.03466511e-01
-2.96914458e-01 -2.48011962e-01 -7.91754246e-01 2.22658068e-02
1.06042147e+00 2.23956648e-02 -8.34620297e-01 2.02423644e+00
4.85132784e-01 1.80222377e-01 2.89716274e-01 7.35712051e-01
4.93620127e-01 3.78772199e-01 -1.39647767e-01 -4.37034577e-01
1.03107452e+00 -8.05750072e-01 -7.58985221e-01 -1.53839603e-01
7.44324386e-01 -9.75221932e-01 1.20971251e+00 3.60864609e-01
-6.85214460e-01 -3.32649946e-01 -1.13158417e+00 1.89750046e-01
-2.19369650e-01 3.17028284e-01 2.17598990e-01 7.14078903e-01
-8.45389128e-01 5.02104700e-01 -7.30215192e-01 -1.94980100e-01
3.76885772e-01 3.45576108e-01 -4.01860267e-01 -1.92091063e-01
-1.07664621e+00 7.46245742e-01 3.91090453e-01 -2.02124733e-02
-5.36987364e-01 -4.91187394e-01 -6.01894200e-01 -4.74732444e-02
3.97926390e-01 -2.77289748e-01 1.31358171e+00 -8.35224211e-01
-1.25581706e+00 9.92105544e-01 -1.36280894e-01 -2.11404145e-01
6.00699961e-01 -1.03833295e-01 -1.52028248e-01 -4.00937498e-01
1.40595973e-01 4.51582909e-01 6.30995512e-01 -1.37351322e+00
-6.12295210e-01 -4.86961812e-01 -4.85734373e-01 2.19805732e-01
-8.22393820e-02 -4.28712308e-01 -1.72809601e-01 -5.39132416e-01
4.31411684e-01 -1.01530743e+00 -7.14356750e-02 5.85107245e-02
-7.53769934e-01 -3.22174817e-01 5.91497242e-01 -7.88322240e-02
1.04692554e+00 -2.31934166e+00 -7.62520656e-02 4.69127059e-01
1.64158847e-02 3.13966930e-01 -1.10968359e-01 4.40763026e-01
-1.55190647e-01 -1.14372792e-02 -4.75312352e-01 -2.68479794e-01
1.15523063e-01 3.04122180e-01 -4.05720472e-01 6.75084233e-01
8.93162936e-03 4.42596555e-01 -1.01823246e+00 -5.16382575e-01
-7.06730261e-02 1.83953196e-01 -1.25869796e-01 5.72025478e-01
-2.58668542e-01 5.62942922e-01 -2.09537238e-01 5.69681346e-01
8.63224924e-01 -5.02948761e-01 5.38000464e-01 -1.94876418e-01
2.87368298e-01 -6.20571822e-02 -1.31385303e+00 1.35895741e+00
-1.91314593e-02 2.40435511e-01 -2.98761904e-01 -8.88976157e-01
1.19898868e+00 2.75894523e-01 3.61027122e-01 -2.95928955e-01
1.09496377e-01 4.14273143e-01 -6.52929917e-02 -1.95918813e-01
4.96618412e-02 -2.96319157e-01 -4.04413417e-02 7.29885280e-01
2.34246403e-01 -1.09872684e-01 -4.93705161e-02 2.17730030e-02
6.57498300e-01 2.97213703e-01 6.63598657e-01 -3.42137456e-01
3.16541165e-01 -1.83972374e-01 8.15626979e-01 8.77761960e-01
-2.68452346e-01 7.82823622e-01 6.19721293e-01 -2.73582879e-02
-1.00113451e+00 -1.35276628e+00 -3.57978761e-01 7.33327389e-01
4.09061819e-01 -1.27425373e-01 -5.63119769e-01 -1.05201221e+00
9.03456435e-02 8.44660461e-01 -3.93102437e-01 -2.36126646e-01
4.12837416e-02 -8.31394076e-01 5.97573519e-01 4.09732938e-01
1.91571951e-01 -5.37538826e-01 -1.09933717e-02 -1.26841083e-01
-2.40688562e-01 -9.52043831e-01 -4.36926752e-01 6.59375012e-01
-5.19812882e-01 -1.38269925e+00 -5.36710322e-01 -8.20646286e-01
8.59998822e-01 1.11654162e-01 9.29710090e-01 -2.37460002e-01
9.56381485e-02 1.87611476e-01 -2.59204924e-01 -3.28109235e-01
-7.52660275e-01 -6.06135279e-02 1.85351849e-01 1.50082439e-01
7.25705743e-01 -6.04647279e-01 -3.71974140e-01 7.69084632e-01
-6.92450166e-01 1.62054986e-01 4.41978931e-01 1.14279950e+00
8.13685656e-01 -1.65415213e-01 8.29897404e-01 -1.28255832e+00
2.40554854e-01 -7.11495280e-01 -6.93952620e-01 6.90907359e-01
-9.27362204e-01 2.79293805e-01 5.12621641e-01 -6.05876565e-01
-8.95666599e-01 3.28414559e-01 7.01262150e-03 -5.42105198e-01
-4.05557603e-01 2.30075777e-01 -5.13635159e-01 4.31113571e-01
7.16988266e-01 8.76929760e-02 2.88232714e-01 -4.64472741e-01
2.96709418e-01 1.10998368e+00 5.05894423e-01 -6.26615584e-01
6.21647239e-01 3.43398422e-01 -2.56556831e-02 -2.41897672e-01
-1.23443961e+00 -5.39877772e-01 -9.48475838e-01 1.84037704e-02
3.09873730e-01 -6.72964394e-01 -2.91754425e-01 5.62533736e-01
-8.75546515e-01 -4.08054143e-01 -4.30188626e-01 5.19545674e-01
-6.29032612e-01 4.96310592e-01 -2.92599171e-01 -9.02352154e-01
-2.03193575e-02 -1.13015366e+00 9.58254218e-01 3.02339852e-01
-2.44301617e-01 -9.69477654e-01 5.26582748e-02 1.28484935e-01
2.64602602e-02 -2.62144487e-03 1.07622647e+00 -1.39779496e+00
-1.25599697e-01 -2.97033131e-01 -1.62947163e-01 4.24708933e-01
4.29743260e-01 1.00956894e-02 -1.19013822e+00 -3.76286596e-01
-1.19427316e-01 -5.44587076e-01 6.46898687e-01 2.29926914e-01
9.14770365e-01 -1.86361581e-01 -3.68166566e-01 2.92839378e-01
1.07619369e+00 1.53533399e-01 3.21863204e-01 -4.75571491e-02
5.24870634e-01 7.41716743e-01 9.78855610e-01 6.18700564e-01
2.48676986e-01 5.32600045e-01 2.45535478e-01 2.25129016e-02
-7.91064929e-04 -5.94046593e-01 1.62152365e-01 7.09270179e-01
5.29499173e-01 -5.97685397e-01 -9.67072308e-01 3.85310650e-01
-2.05334806e+00 -7.57649660e-01 -1.69302061e-01 2.61710739e+00
1.21352577e+00 -8.10049400e-02 -1.91586055e-02 2.36954391e-01
1.07491517e+00 -2.41915867e-01 -7.34290481e-01 2.68822089e-02
-2.81129688e-01 -9.43734422e-02 4.35066313e-01 5.56105316e-01
-1.16523719e+00 6.10180855e-01 6.21045446e+00 1.10845470e+00
-8.28997731e-01 6.43933713e-02 7.13362336e-01 1.98916867e-01
-5.12474358e-01 1.19192466e-01 -9.20735896e-01 5.78579485e-01
8.81548703e-01 -2.20751777e-01 9.75879952e-02 8.62665534e-01
-1.91046134e-01 -1.73051596e-01 -1.41082764e+00 9.93117392e-01
-8.75391215e-02 -9.21176136e-01 -8.68864134e-02 1.07073840e-02
7.20205665e-01 -1.67170703e-01 1.55161992e-01 2.15458065e-01
5.92732847e-01 -8.04572880e-01 5.43570101e-01 4.03900981e-01
1.07528293e+00 -6.31850064e-01 9.60536242e-01 6.79938078e-01
-7.92258382e-01 5.19084558e-02 -2.68328696e-01 3.13809186e-01
-4.04811949e-02 7.54825830e-01 -9.17700827e-01 5.49215615e-01
4.31528509e-01 6.83764756e-01 -3.54006827e-01 8.11633646e-01
-2.82354444e-01 7.76106298e-01 -4.40455407e-01 2.43918002e-01
-2.60486126e-01 -1.74537599e-01 4.08141911e-01 1.00851882e+00
3.09423745e-01 -1.22021355e-01 4.73428369e-01 6.28633916e-01
5.85395209e-02 3.33417617e-02 -6.00715756e-01 2.76817363e-02
8.73684645e-01 8.28994453e-01 -6.97254121e-01 -2.54790962e-01
-3.57415974e-01 7.67998517e-01 4.26168740e-01 3.77941102e-01
-7.14320600e-01 2.91365106e-03 3.49095464e-01 5.71527742e-02
6.34256303e-02 2.58250624e-01 -2.52102375e-01 -1.26508713e+00
-8.79524946e-02 -7.93801725e-01 7.09152937e-01 -5.55792928e-01
-1.94187403e+00 5.54422140e-01 2.22578749e-01 -1.56178916e+00
-3.02013338e-01 -5.96265674e-01 -1.52297094e-01 7.88954318e-01
-1.38758576e+00 -1.05715621e+00 -3.45055647e-02 4.90104079e-01
2.10117310e-01 -2.87158072e-01 9.40051675e-01 2.40372583e-01
-5.16770840e-01 9.49798584e-01 5.07976949e-01 -8.93690586e-02
1.18528795e+00 -1.43627048e+00 -8.01367164e-02 5.88460565e-01
2.18007013e-01 3.15053105e-01 6.93109930e-01 -5.78687370e-01
-8.48772645e-01 -1.08759570e+00 8.08581829e-01 -3.41207862e-01
6.17153108e-01 -2.06367612e-01 -1.06044972e+00 6.43548608e-01
-1.26794532e-01 2.10318819e-01 1.23440087e+00 2.08990127e-02
-7.13609755e-01 -2.68058795e-02 -1.31865060e+00 3.04145217e-01
7.22732961e-01 -4.72866714e-01 -4.19435620e-01 4.66941357e-01
3.95205051e-01 -3.48718733e-01 -6.74492657e-01 5.75248480e-01
5.79281986e-01 -9.27865565e-01 5.79956949e-01 -5.33273637e-01
7.32536018e-02 -5.49131811e-01 -6.66738570e-01 -1.25007355e+00
-5.18709831e-02 -3.73481959e-01 -1.21582650e-01 1.37915576e+00
6.66779041e-01 -8.56676400e-01 5.90059221e-01 6.53944552e-01
1.01302616e-01 -8.56865585e-01 -9.59505856e-01 -9.14215982e-01
1.78253219e-01 -2.55791903e-01 6.78081572e-01 1.04593790e+00
1.01051822e-01 2.01160520e-01 -4.25170183e-01 3.45315129e-01
7.22049952e-01 4.50069398e-01 6.26541734e-01 -1.23725140e+00
-5.64275920e-01 -2.10514337e-01 -2.52470434e-01 -1.00319445e+00
3.88763994e-01 -1.15811443e+00 2.73480386e-01 -1.08339584e+00
4.85711366e-01 -8.27529669e-01 -4.97287542e-01 7.11246669e-01
-3.37051034e-01 1.88302979e-01 -7.22547024e-02 5.06425083e-01
-4.97286141e-01 5.67475259e-01 1.00040996e+00 9.53117311e-02
-9.15498212e-02 3.83258432e-01 -6.61388934e-01 8.14458668e-01
9.88683820e-01 -6.92970753e-01 -6.91977501e-01 4.66349497e-02
1.02050761e-02 2.89735897e-03 1.01722308e-01 -5.45665145e-01
1.32566154e-01 -3.22402120e-01 1.95288122e-01 -6.02521181e-01
3.83486203e-03 -8.03003788e-01 1.97301000e-01 1.57920524e-01
-6.39135540e-01 -2.70406693e-01 -7.40556642e-02 7.27923632e-01
-1.80516124e-01 -3.37829530e-01 1.02807903e+00 2.62277961e-01
-2.24321708e-01 1.76330343e-01 -1.37345210e-01 1.16076477e-01
1.12296665e+00 -1.17183626e-01 -4.35885459e-01 -2.84039140e-01
-7.69709527e-01 2.58060127e-01 6.10870123e-01 1.02687456e-01
3.33158493e-01 -1.31915224e+00 -6.14292979e-01 3.25123966e-01
3.15738469e-01 1.90584898e-01 1.52269483e-01 7.98860431e-01
-5.52385189e-02 1.09444328e-01 3.73007022e-02 -8.55023026e-01
-1.15766168e+00 6.00456417e-01 3.29370111e-01 -3.30606937e-01
-1.74628258e-01 6.70290351e-01 2.63927400e-01 -9.39535677e-01
3.78288984e-01 1.08334824e-01 -1.33285865e-01 -3.64148244e-02
3.57885629e-01 9.95700434e-02 -1.12081282e-01 -5.06566286e-01
-2.67260373e-01 3.99163961e-01 -2.07380623e-01 3.96332791e-04
7.85361409e-01 -2.72747248e-01 9.10144523e-02 8.53381276e-01
1.00728106e+00 -9.30606667e-03 -1.29727864e+00 -5.90086997e-01
1.36297852e-01 -5.58569729e-01 -3.72806162e-01 -8.94281149e-01
-7.26696491e-01 7.91702747e-01 5.34549773e-01 6.96179867e-02
9.77574706e-01 2.11679235e-01 3.04434747e-01 1.64668635e-01
4.45153832e-01 -8.32776546e-01 -3.91995125e-02 2.65031219e-01
5.12920320e-01 -1.38974559e+00 -3.18933390e-02 -5.55080533e-01
-8.35281074e-01 1.00833571e+00 8.15041423e-01 2.24674970e-01
7.46600986e-01 5.33183157e-01 2.61486590e-01 1.90368593e-01
-8.11033785e-01 -2.02235043e-01 2.25980476e-01 8.01989377e-01
4.29810017e-01 8.73050839e-02 -1.47367850e-01 7.34601915e-01
-4.96834554e-02 -1.43067777e-01 3.85449529e-01 8.95885825e-01
-2.66112566e-01 -1.34178066e+00 -2.61984557e-01 4.74863589e-01
-2.05187142e-01 1.13287121e-01 -3.39196682e-01 6.45809352e-01
1.12315796e-01 9.26739216e-01 4.42260019e-02 -3.73013586e-01
1.18759379e-01 2.67497718e-01 3.41986299e-01 -6.69227540e-01
-2.01277006e-02 2.03166023e-01 -1.46568343e-01 -1.50786981e-01
-4.57173765e-01 -6.75153792e-01 -1.21450281e+00 -7.08721951e-02
-6.17957711e-01 3.55296791e-01 3.51552963e-01 9.53025520e-01
3.55174810e-01 -1.22183874e-01 8.37821841e-01 -4.53155518e-01
-1.01593482e+00 -9.20093596e-01 -8.28296304e-01 3.24004054e-01
2.64287502e-01 -7.62455821e-01 -6.53367341e-01 -3.46220993e-02]
|
[9.356354713439941, 3.916848659515381]
|
2ea39dd1-05ef-48a4-98a8-d6aa3f26dd88
|
distributed-filtered-hyperinterpolation-for
|
1910.02434
| null |
https://arxiv.org/abs/1910.02434v1
|
https://arxiv.org/pdf/1910.02434v1.pdf
|
Distributed filtered hyperinterpolation for noisy data on the sphere
|
Problems in astrophysics, space weather research and geophysics usually need to analyze noisy big data on the sphere. This paper develops distributed filtered hyperinterpolation for noisy data on the sphere, which assigns the data fitting task to multiple servers to find a good approximation of the mapping of input and output data. For each server, the approximation is a filtered hyperinterpolation on the sphere by a small proportion of quadrature nodes. The distributed strategy allows parallel computing for data processing and model selection and thus reduces computational cost for each server while preserves the approximation capability compared to the filtered hyperinterpolation. We prove quantitative relation between the approximation capability of distributed filtered hyperinterpolation and the numbers of input data and servers. Numerical examples show the efficiency and accuracy of the proposed method.
|
['Ding-Xuan Zhou', 'Yu Guang Wang', 'Shao-Bo Lin']
|
2019-10-06
| null | null | null | null |
['geophysics']
|
['miscellaneous']
|
[-1.10321176e+00 -1.27001375e-01 1.01081347e+00 -1.32319510e-01
-6.88652456e-01 -4.82538402e-01 3.80583048e-01 1.80488929e-01
-4.80893672e-01 8.18848431e-01 -1.16388880e-01 -1.40430093e-01
-3.58537018e-01 -1.18336511e+00 -7.87928760e-01 -8.12189460e-01
-2.10758179e-01 9.44127738e-01 5.62380373e-01 -1.19733825e-01
-1.95789456e-01 8.16215754e-01 -1.54567027e+00 1.08384684e-01
1.10814500e+00 1.47796142e+00 2.59547204e-01 7.43243933e-01
-3.65460403e-02 2.07349375e-01 -5.61003089e-01 -4.16379124e-01
6.86440170e-01 1.41397119e-01 -3.93252701e-01 -5.08230329e-01
-1.31697550e-01 2.22671479e-01 -5.38026273e-01 1.37963820e+00
8.87108982e-01 4.89895970e-01 3.73652965e-01 -1.30869448e+00
2.46055052e-02 3.76761258e-01 -3.36507767e-01 6.36186838e-01
-1.28299415e-01 2.04537049e-01 4.60474133e-01 -1.09886158e+00
3.39651823e-01 1.18843424e+00 7.17044711e-01 -2.85286218e-01
-1.01717532e+00 -4.88540053e-01 -4.95370090e-01 1.71093449e-01
-1.92471898e+00 -1.21356465e-01 4.05288897e-02 -1.83705151e-01
6.64813578e-01 7.78801620e-01 9.93249416e-01 2.81845927e-02
2.04016760e-01 8.56436864e-02 1.07951450e+00 7.93770105e-02
7.34209657e-01 5.72770163e-02 2.20147729e-01 2.29688212e-01
6.92008615e-01 7.00021014e-02 -6.36376798e-01 -9.18045998e-01
9.71094608e-01 -5.87651916e-02 -2.35887811e-01 -6.60959631e-02
-7.01604664e-01 5.38702190e-01 2.34335497e-01 -4.80976589e-02
-4.09006596e-01 -1.12650760e-01 4.04944360e-01 5.13032734e-01
8.23720753e-01 4.24314648e-01 -5.35500288e-01 -2.73648828e-01
-7.40941048e-01 5.71454525e-01 1.19411314e+00 1.11915970e+00
6.26460373e-01 -6.91347644e-02 -1.41109005e-01 4.89645123e-01
7.19538182e-02 1.29123425e+00 1.47035167e-01 -1.01907897e+00
2.36085072e-01 2.86322474e-01 4.59353536e-01 -1.00892651e+00
-7.02618778e-01 -3.63930434e-01 -1.24769640e+00 2.57722974e-01
5.27210295e-01 -4.17031854e-01 -1.90621927e-01 9.70121145e-01
1.10618699e+00 4.69044089e-01 -8.79223365e-03 1.36021912e+00
4.62781280e-01 7.80833006e-01 -4.72423375e-01 -4.04003114e-01
1.36456895e+00 -5.00858545e-01 -5.00799358e-01 6.78912342e-01
1.44972414e-01 -1.01405406e+00 8.70679438e-01 7.75898814e-01
-1.25118625e+00 -1.72907531e-01 -3.40589404e-01 7.78803453e-02
-8.50915462e-02 -5.79047441e-01 2.96449214e-01 1.81073785e-01
-1.01439285e+00 7.08325982e-01 -9.87483501e-01 1.03998877e-01
-1.21634088e-01 2.21646085e-01 3.22264060e-02 2.44884029e-01
-1.13237846e+00 7.06143200e-01 5.78304231e-02 1.66921765e-01
-4.61674780e-01 -8.98460329e-01 -2.51229972e-01 3.21807504e-01
1.36792446e-02 -6.23431861e-01 1.11945689e+00 -3.91071767e-01
-1.31119692e+00 1.11131817e-02 1.42649919e-01 -4.26437378e-01
7.87823677e-01 2.77242661e-02 -1.94359779e-01 3.93985771e-02
-1.42473593e-01 -6.57958806e-01 4.89912570e-01 -8.33110034e-01
-4.86135393e-01 -6.34347320e-01 -7.33056784e-01 3.31977546e-01
2.17259303e-02 1.59833491e-01 -3.86008680e-01 -5.92601001e-01
3.50109160e-01 -6.82259023e-01 -5.74423790e-01 2.94965506e-02
-5.88121712e-02 -3.52832615e-01 4.06788439e-01 -3.43920231e-01
8.41923416e-01 -2.04490137e+00 -2.55113870e-01 9.51038718e-01
5.64653337e-01 -3.09571922e-02 3.07387531e-01 4.14747149e-01
1.94712490e-01 -1.82249129e-01 2.50224859e-01 -2.38408402e-01
-5.16282395e-02 2.34582812e-01 -3.27858955e-01 1.15675998e+00
-8.06780696e-01 2.96941638e-01 -6.08117342e-01 -3.95828187e-01
-2.01884404e-01 2.00189516e-01 -6.23709619e-01 3.86202246e-01
-2.02829108e-01 2.37795040e-01 -7.63283134e-01 3.70690748e-02
1.30556297e+00 -3.47440720e-01 -2.33377472e-01 4.13602106e-02
-3.84168983e-01 4.63790409e-02 -1.92673099e+00 1.02335978e+00
-3.51004601e-01 3.31537873e-02 7.94637740e-01 -7.93616056e-01
1.14477873e+00 3.76731962e-01 6.68803036e-01 -3.68679792e-01
2.42494285e-01 5.02271056e-01 -3.19836766e-01 -6.12871766e-01
4.19579268e-01 -1.00698106e-01 1.16346955e-01 7.46940136e-01
-3.38329375e-01 -4.29295689e-01 -1.00232728e-01 2.04260305e-01
9.97942090e-01 -5.77896535e-01 -3.21041830e-02 -8.65690470e-01
1.13876052e-01 -6.60486072e-02 7.52217770e-01 9.19828951e-01
-5.82268946e-02 6.67658389e-01 4.60895956e-01 -6.81416214e-01
-1.42532778e+00 -8.88477087e-01 -5.56996882e-01 8.89970124e-01
3.90097171e-01 -5.35930872e-01 -9.58820701e-01 -5.20374440e-02
4.83664542e-01 2.52505779e-01 -1.85907453e-01 1.64433107e-01
-3.88897836e-01 -8.03309619e-01 3.85965466e-01 8.88052061e-02
4.05669391e-01 -5.45999527e-01 -1.94695279e-01 -1.63937539e-01
-1.01053789e-01 -7.72053301e-01 -5.22162378e-01 -5.91301657e-02
-6.92151368e-01 -1.10196841e+00 -3.57982069e-01 -4.16646302e-01
5.68078458e-01 1.98386684e-01 1.00493801e+00 9.09200534e-02
7.52220349e-03 2.65386123e-02 -2.34166663e-02 -3.92826289e-01
5.41281188e-03 -2.76725829e-01 5.97205579e-01 1.81812719e-01
6.66433992e-03 -6.97497904e-01 -7.19926953e-01 5.60095191e-01
-8.92219484e-01 -3.06195408e-01 -8.07898790e-02 4.89272058e-01
8.30158412e-01 1.21239364e-01 2.14259595e-01 -5.51944494e-01
7.74011433e-01 -9.96962845e-01 -1.46794558e+00 4.92855683e-02
-6.37535751e-01 2.02866390e-01 1.06375980e+00 -2.48552755e-01
-6.85703158e-01 -1.36851639e-01 2.37770781e-01 -5.44160664e-01
3.61827195e-01 2.24170759e-01 -6.81219110e-03 -3.44397634e-01
8.29942346e-01 -5.22164926e-02 2.49419928e-01 -9.81361151e-01
2.88918525e-01 9.57813919e-01 5.12805104e-01 -9.35049832e-01
4.49545979e-01 4.73759651e-01 2.29713485e-01 -6.96647406e-01
-1.70892313e-01 -5.13587534e-01 -1.15032271e-01 -2.25472227e-01
2.37172022e-01 -6.03696048e-01 -1.52926207e+00 3.53355944e-01
-1.27067947e+00 8.02029893e-02 -7.26780295e-01 6.66004062e-01
-6.38974309e-01 3.36061865e-01 -6.36913300e-01 -1.09366083e+00
-4.70364720e-01 -9.62040782e-01 9.06215608e-01 9.99045074e-02
2.24271759e-01 -5.03543556e-01 2.91202009e-01 -8.88269860e-03
5.92051029e-01 5.68209365e-02 2.46192575e-01 -9.38046217e-01
-7.46074140e-01 -5.15345454e-01 -3.10399532e-01 -1.70727223e-01
-6.08655393e-01 -5.09729087e-02 -6.25722468e-01 -2.55257249e-01
5.57905257e-01 -1.51059046e-01 2.29698867e-01 3.41634721e-01
1.22520423e+00 -6.50394738e-01 -1.68875724e-01 8.54840398e-01
1.33460212e+00 -4.85804796e-01 3.74305308e-01 1.67733893e-01
6.85436428e-02 6.99453875e-02 3.24697018e-01 1.14804578e+00
1.81684405e-01 3.00467789e-01 3.99139911e-01 7.86724780e-03
3.13177258e-01 4.66017462e-02 -2.60764360e-01 1.02437782e+00
-7.67155886e-02 2.85391207e-03 -1.10364580e+00 2.30718911e-01
-2.14116740e+00 -5.82614839e-01 -5.13370991e-01 2.61059332e+00
9.38813984e-01 -2.22252935e-01 2.50536591e-01 -6.68554232e-02
8.60246897e-01 -5.89963138e-01 -6.15163803e-01 -1.76333606e-01
-4.10472423e-01 6.87245056e-02 8.95270884e-01 7.20622361e-01
-4.73099917e-01 3.09764802e-01 6.68996143e+00 1.20442367e+00
-6.13636792e-01 3.99360418e-01 6.97158158e-01 -5.44383466e-01
-3.76427978e-01 -2.01683760e-01 -6.24988019e-01 8.43329370e-01
1.07098997e+00 -7.87116528e-01 8.50268245e-01 9.70673382e-01
6.04960859e-01 -3.83078218e-01 -4.86884385e-01 1.04616416e+00
-4.63184237e-01 -1.20880604e+00 -2.66900301e-01 5.66590913e-02
9.38686907e-01 6.77649736e-01 -3.63004446e-01 -3.96028429e-01
5.89503884e-01 -8.89588952e-01 4.91591394e-01 1.35797894e+00
4.21277672e-01 -8.14786315e-01 8.91461790e-01 7.68163085e-01
-1.06044662e+00 1.36595726e-01 -8.70690227e-01 -2.04200909e-01
-2.28413809e-02 1.17399573e+00 -4.91779745e-01 4.23953414e-01
1.21966517e+00 -4.60307807e-01 -1.74713984e-01 1.67354786e+00
5.77084184e-01 4.60658669e-01 -1.36292112e+00 -4.42678273e-01
-2.29681998e-01 -8.72611642e-01 6.75011635e-01 6.07437968e-01
5.46776474e-01 7.11864352e-01 3.75130624e-01 6.41531169e-01
7.01836273e-02 3.54403406e-01 -4.28750098e-01 6.30928755e-01
7.92967796e-01 1.18285549e+00 -5.56831896e-01 -5.20615697e-01
-2.86847264e-01 2.88748920e-01 3.26053917e-01 4.75618780e-01
-4.91977185e-01 -5.57714462e-01 9.09357905e-01 5.87608695e-01
1.23862125e-01 -3.38543534e-01 -1.02749002e+00 -8.76695573e-01
1.50005415e-01 -7.75088251e-01 6.15258098e-01 -8.01797688e-01
-1.44683754e+00 8.24000537e-01 -1.67970628e-01 -9.82771397e-01
1.37757719e-01 -1.90053761e-01 -4.06405210e-01 1.27286148e+00
-7.52617002e-01 -4.24730182e-01 -2.25932777e-01 9.93151844e-01
-2.48803258e-01 -1.23937704e-01 5.73451936e-01 4.03459705e-02
-2.76457518e-01 2.94176430e-01 1.14519393e+00 -1.82641551e-01
3.19087982e-01 -1.21485364e+00 3.06683600e-01 6.34133041e-01
-4.43870395e-01 2.63858020e-01 1.18775022e+00 -8.16722095e-01
-1.59234893e+00 -8.49958062e-01 9.37369943e-01 -2.06993341e-01
8.21256876e-01 -5.64283907e-01 -1.14142513e+00 4.08522561e-02
1.09574303e-01 8.38470936e-01 2.77649879e-01 -9.21065658e-02
1.05930947e-01 -6.91477954e-01 -1.54247391e+00 2.24209592e-01
5.82004786e-01 -1.53386369e-01 -3.31981517e-02 9.18522239e-01
6.09520793e-01 -5.13836026e-01 -1.16692090e+00 6.44289926e-02
-1.01879232e-01 -6.95931911e-01 6.46379173e-01 -6.13794684e-01
-3.63228768e-01 -4.14073914e-01 4.07489426e-02 -1.42058825e+00
-2.78809279e-01 -1.31788456e+00 1.28328493e-02 5.60405314e-01
1.79085210e-01 -8.90574157e-01 5.98784506e-01 1.14214957e+00
9.40575916e-03 -6.44339502e-01 -1.53083730e+00 -7.33886898e-01
2.80069113e-01 -5.32461703e-01 9.43015158e-01 3.55964601e-01
4.65032399e-01 -1.47397891e-01 7.99066946e-02 6.91890001e-01
9.74841356e-01 3.02421898e-01 7.10684836e-01 -1.28820300e+00
-3.21739763e-01 -1.92475781e-01 -2.99798310e-01 -8.65459085e-01
-3.27258885e-01 -7.47674525e-01 -1.15152588e-02 -1.00532568e+00
-7.58873373e-02 -8.76106977e-01 6.48882240e-02 1.16686605e-01
-3.75352576e-02 1.37452766e-01 -6.10220581e-02 6.07593417e-01
-6.90213561e-01 6.49086297e-01 1.49049842e+00 4.70857173e-01
-6.09011911e-02 3.78649682e-01 6.66022077e-02 5.41918516e-01
5.81681490e-01 -5.73101759e-01 -3.19288746e-02 -4.04816061e-01
5.80306351e-01 4.98881340e-01 9.72625837e-02 -8.78775835e-01
8.06854427e-01 -2.22410336e-01 1.90954894e-01 -4.49194074e-01
1.66077882e-01 -8.04679096e-01 2.03376010e-01 1.05774105e-01
-2.04990860e-02 3.67928803e-01 9.05458257e-02 6.55235946e-01
-4.02181707e-02 3.61760333e-02 1.03234887e+00 3.33493575e-02
4.37875003e-01 5.52656114e-01 -4.75332022e-01 3.98300380e-01
1.10119176e+00 4.97758210e-01 -1.77818999e-01 -6.12869024e-01
-1.06216121e+00 4.71576661e-01 3.02404225e-01 -3.38374346e-01
3.14205110e-01 -1.37912428e+00 -8.31371605e-01 7.83875585e-01
-6.77631557e-01 4.32933152e-01 5.74465692e-02 8.20539236e-01
-7.64266372e-01 6.05900101e-02 3.40628713e-01 -5.08716345e-01
-9.02217627e-01 5.45922935e-01 6.82683647e-01 1.74763620e-01
-7.32221603e-01 7.12920487e-01 -3.44027609e-01 -4.00117129e-01
3.38409752e-01 -4.29442734e-01 3.71208191e-01 -2.70678222e-01
5.53877115e-01 1.06225193e+00 6.43692851e-01 -3.00293952e-01
-1.08850710e-01 8.14734846e-02 5.76080143e-01 -3.17134589e-01
1.43607676e+00 -4.00400013e-01 -5.95474243e-01 2.38366589e-01
1.01238298e+00 3.03568635e-02 -1.11121869e+00 -4.56304550e-01
-3.16938072e-01 -5.27200580e-01 1.06739350e-01 -5.08341134e-01
-1.28232753e+00 4.03061599e-01 3.15377146e-01 9.23303485e-01
8.72570932e-01 1.64384425e-01 7.75188625e-01 4.56141472e-01
4.94345993e-01 -1.30334759e+00 -6.53109550e-01 5.65456450e-01
1.06869614e+00 -6.35854661e-01 -1.84335917e-01 -3.88104975e-01
-3.92039895e-01 1.05361366e+00 6.11572683e-01 -5.19611120e-01
1.46132076e+00 1.05956972e+00 -1.73135862e-01 -2.45823994e-01
-9.16507781e-01 7.69586265e-02 -4.28113379e-02 1.25476137e-01
-1.98298126e-01 1.51377350e-01 -5.27423263e-01 1.12933183e+00
-6.75422013e-01 8.20318833e-02 4.09553438e-01 5.24818063e-01
-8.29249561e-01 -3.95892143e-01 -1.05087614e+00 6.40654624e-01
-2.17658177e-01 -1.62152603e-01 2.57964313e-01 5.48324697e-02
-1.57282762e-02 8.73752296e-01 6.73590422e-01 1.19330592e-01
5.68560600e-01 -1.24884889e-01 -4.15011272e-02 -1.17124379e-01
-7.41515934e-01 4.29215074e-01 -1.52165040e-01 -7.62055814e-01
5.83649576e-01 -5.87386966e-01 -1.48887217e+00 -9.43437338e-01
-3.36309522e-01 1.15221918e+00 7.55649447e-01 6.35397136e-01
8.75092745e-01 -3.09003349e-02 1.05675352e+00 -5.82653582e-01
-1.22801709e+00 -1.11614490e+00 -1.28782737e+00 2.83783972e-01
1.65180370e-01 -3.65167350e-01 -7.13235676e-01 -4.97788727e-01]
|
[6.656850814819336, 4.481255054473877]
|
7fb362f7-8cb0-4106-8525-2a9fef04e15b
|
piqi-perceptual-image-quality-index-based-on
|
2305.09214
| null |
https://arxiv.org/abs/2305.09214v1
|
https://arxiv.org/pdf/2305.09214v1.pdf
|
PIQI: Perceptual Image Quality Index based on Ensemble of Gaussian Process Regression
|
Digital images contain a lot of redundancies, therefore, compression techniques are applied to reduce the image size without loss of reasonable image quality. Same become more prominent in the case of videos which contains image sequences and higher compression ratios are achieved in low throughput networks. Assessment of quality of images in such scenarios has become of particular interest. Subjective evaluation in most of the scenarios is infeasible so objective evaluation is preferred. Among the three objective quality measures, full-reference and reduced-reference methods require an original image in some form to calculate the image quality which is unfeasible in scenarios such as broadcasting, acquisition or enhancement. Therefore, a no-reference Perceptual Image Quality Index (PIQI) is proposed in this paper to assess the quality of digital images which calculates luminance and gradient statistics along with mean subtracted contrast normalized products in multiple scales and color spaces. These extracted features are provided to a stacked ensemble of Gaussian Process Regression (GPR) to perform the perceptual quality evaluation. The performance of the PIQI is checked on six benchmark databases and compared with twelve state-of-the-art methods and competitive results are achieved. The comparison is made based on RMSE, Pearson and Spearman correlation coefficients between ground truth and predicted quality scores. The scores of 0.0552, 0.9802 and 0.9776 are achieved respectively for these metrics on CSIQ database. Two cross-dataset evaluation experiments are performed to check the generalization of PIQI.
|
['Hassan Khalid', 'Hafiz Muhammad Shahzad Asif', 'Nisar Ahmed']
|
2023-05-16
| null | null | null | null |
['gpr', 'gpr']
|
['computer-vision', 'miscellaneous']
|
[ 4.10278469e-01 -5.02319396e-01 2.97488034e-01 -3.79298210e-01
-7.37850428e-01 -1.86448544e-01 3.55126500e-01 3.09436500e-01
-5.13941526e-01 8.68849516e-01 6.57254085e-03 6.29255101e-02
-5.92799067e-01 -8.93507421e-01 -3.20396066e-01 -8.80583048e-01
-4.26104695e-01 -2.32693464e-01 3.98943186e-01 -3.51768732e-02
4.89620566e-01 4.14831847e-01 -1.75694871e+00 4.34091948e-02
1.03055108e+00 1.41485548e+00 4.21370149e-01 7.46038675e-01
2.27702886e-01 5.22659123e-01 -7.15057254e-01 -4.70765948e-01
4.59933400e-01 -2.51205564e-01 -3.42383534e-01 2.71263182e-01
-6.78087547e-02 -2.30160296e-01 -2.32091159e-01 1.27942801e+00
5.60581148e-01 2.42122710e-01 6.41605735e-01 -1.18754911e+00
-4.56525058e-01 -2.10275520e-02 -7.98319936e-01 3.60248893e-01
3.67162317e-01 7.58634955e-02 5.78259230e-01 -7.88803935e-01
2.21535921e-01 1.10842240e+00 2.56027371e-01 -2.80289441e-01
-9.93604243e-01 -6.03105783e-01 -4.59247947e-01 5.75684786e-01
-1.47496939e+00 -2.73568094e-01 5.53047717e-01 -2.96321064e-01
6.26507640e-01 3.02930474e-01 2.73003727e-01 2.52746016e-01
5.32690763e-01 -1.92156415e-02 1.45918739e+00 -3.60810071e-01
8.54040682e-02 2.41122276e-01 -1.94096655e-01 4.35619324e-01
3.83367956e-01 2.04492405e-01 -4.28357750e-01 2.59501189e-01
5.74540973e-01 -8.78134146e-02 -4.30350959e-01 -1.26987502e-01
-1.05736387e+00 5.32035172e-01 5.10568917e-01 4.09847140e-01
-7.18316197e-01 -3.27245981e-01 4.04032230e-01 3.64142329e-01
3.76600288e-02 -3.24338824e-02 -1.03626691e-01 -2.57907271e-01
-1.09417689e+00 -1.59816906e-01 4.07490522e-01 7.22943723e-01
4.07557338e-01 1.42509282e-01 -6.37687147e-02 9.15974736e-01
3.19361240e-01 6.72533274e-01 4.88221794e-01 -9.20459449e-01
4.49141622e-01 3.62684995e-01 1.62373185e-01 -1.65781748e+00
-1.57633886e-01 -6.88985407e-01 -1.21941257e+00 4.58047807e-01
6.99257553e-02 2.10169479e-01 -5.89644551e-01 1.16461587e+00
4.67705578e-02 -2.61854902e-02 4.02489185e-01 1.00561821e+00
7.15734363e-01 1.15524471e+00 9.55404863e-02 -5.05185843e-01
1.21364236e+00 -4.54899639e-01 -6.76185608e-01 2.55084872e-01
-1.80848330e-01 -1.35335517e+00 9.50211644e-01 8.78324687e-01
-1.14399636e+00 -1.07924795e+00 -1.32912493e+00 4.98568207e-01
-1.15620956e-01 2.61343509e-01 1.55766666e-01 8.22306871e-01
-7.70003080e-01 6.87571406e-01 -4.60946202e-01 -2.69477785e-01
3.35558265e-01 2.92855412e-01 -4.55837101e-01 -2.47417077e-01
-1.07030904e+00 8.88068199e-01 3.11586797e-01 9.40472558e-02
-6.80888772e-01 -3.02966237e-01 -3.13195974e-01 3.07880137e-02
-5.06552383e-02 -6.20860280e-03 5.96633494e-01 -1.00786531e+00
-1.28993988e+00 6.57048285e-01 1.60023510e-01 -4.15567547e-01
4.58550572e-01 1.34100124e-01 -9.24303472e-01 5.51468313e-01
-1.36041775e-01 3.90909880e-01 6.88654304e-01 -1.24987721e+00
-8.76484752e-01 -3.51404876e-01 -4.59519401e-02 3.46608043e-01
-1.63826376e-01 1.55373439e-01 -6.04560554e-01 -4.74979818e-01
3.38889509e-01 -6.05033755e-01 8.24740231e-02 7.63366446e-02
-1.60014346e-01 2.86694825e-01 7.65706778e-01 -1.02534676e+00
1.24202168e+00 -1.99900222e+00 -2.82865316e-01 4.07946914e-01
-2.18621597e-01 3.58582109e-01 4.94871587e-02 2.89275914e-01
1.32020518e-01 -1.09383225e-01 -2.44076177e-01 2.70235777e-01
-2.58880764e-01 -4.17344160e-02 3.53266358e-01 6.46641970e-01
8.85592997e-02 9.10377875e-02 -5.14855266e-01 -7.40463972e-01
4.52952504e-01 7.61128008e-01 -3.31887990e-01 3.58315855e-01
4.50306743e-01 4.88716483e-01 -2.33068600e-01 6.35212243e-01
1.07937813e+00 -2.41245758e-02 -1.17253810e-01 -8.41925144e-01
-7.93981925e-02 -2.78196990e-01 -1.53329539e+00 1.27281010e+00
-5.75405419e-01 4.62606549e-01 4.66686338e-02 -1.05255127e+00
1.16868818e+00 4.75707173e-01 4.61594045e-01 -1.04620790e+00
2.05650017e-01 1.57559440e-01 1.73561499e-01 -7.71272123e-01
4.48606074e-01 1.68000292e-02 3.07827473e-01 -9.23987105e-02
-1.60493746e-01 -8.45115855e-02 4.35219318e-01 1.67095587e-02
7.39641726e-01 6.83869049e-02 3.59267861e-01 -2.65980363e-01
9.81568635e-01 -3.64711910e-01 4.89496231e-01 2.67809004e-01
-4.13406223e-01 6.79400742e-01 1.28168061e-01 1.32319912e-01
-1.24471068e+00 -1.28405118e+00 -2.70699084e-01 3.36986661e-01
4.62568820e-01 6.33097887e-02 -3.92734587e-01 -9.14057419e-02
-3.36814344e-01 5.08723199e-01 3.86966057e-02 4.75434028e-02
-1.22615457e-01 -8.45350385e-01 5.98228760e-02 1.73246134e-02
1.09884322e+00 -8.62637818e-01 -8.62744629e-01 1.22500554e-01
-2.09476888e-01 -1.26383448e+00 1.61378205e-01 -3.09274226e-01
-8.61083984e-01 -1.00844955e+00 -8.28772485e-01 -4.59987581e-01
4.23226386e-01 3.75376552e-01 9.62607920e-01 -3.73387448e-02
-3.74443650e-01 3.52588110e-02 -5.42456388e-01 -1.28944471e-01
-3.64113390e-01 -5.37895083e-01 -2.27689110e-02 1.58109665e-01
6.75643161e-02 -7.12711871e-01 -9.92484927e-01 4.48360980e-01
-9.86181378e-01 -1.94216326e-01 9.55760956e-01 5.43967962e-01
6.80713713e-01 8.41906786e-01 4.23506141e-01 -3.09964329e-01
6.38810873e-01 -3.48677993e-01 -9.53893006e-01 1.92635745e-01
-6.63717151e-01 -2.32929200e-01 6.92333102e-01 -1.69937313e-02
-1.20357537e+00 -4.90104884e-01 2.07371935e-02 -1.26738295e-01
-1.86868981e-02 5.38819432e-01 -3.83733720e-01 -8.07722956e-02
4.25711870e-01 2.43968189e-01 -5.97497411e-02 -3.18306118e-01
-1.42229766e-01 9.97697592e-01 7.13429272e-01 -1.47745714e-01
8.71917427e-01 1.73973978e-01 3.83474380e-01 -1.00185251e+00
-2.45690033e-01 -5.69916606e-01 -2.71522909e-01 -4.76896822e-01
7.37964153e-01 -7.63707995e-01 -6.94065750e-01 5.14273763e-01
-8.60599399e-01 3.57992679e-01 3.15804362e-01 9.54224288e-01
-4.28836167e-01 5.96387804e-01 -4.59575534e-01 -1.00057852e+00
-4.81939524e-01 -1.44601333e+00 4.09232914e-01 5.24426103e-01
3.03121179e-01 -5.41325629e-01 -5.50743878e-01 3.55858266e-01
6.12967968e-01 3.19029331e-01 6.77928388e-01 -1.21902913e-01
-6.25394046e-01 -3.63076866e-01 -6.72307253e-01 7.48947144e-01
2.32658207e-01 4.35839370e-02 -7.76325762e-01 -1.36614740e-01
1.23129398e-01 3.72524597e-02 8.27768371e-02 5.48435569e-01
1.15908039e+00 -1.77819803e-01 1.43591017e-01 5.12390614e-01
2.08834434e+00 6.35339022e-01 1.07148314e+00 2.86175936e-01
1.31976858e-01 5.19550025e-01 1.06897557e+00 5.99250495e-01
-8.70448649e-02 6.86001539e-01 5.25544524e-01 -6.82275146e-02
-5.59101254e-02 1.75671995e-01 1.25190526e-01 7.94402361e-01
-1.93357408e-01 -5.13402998e-01 -5.39236844e-01 3.34545732e-01
-1.15035868e+00 -9.29429352e-01 -3.15601468e-01 2.63392377e+00
5.62888801e-01 3.59952241e-01 -1.20751277e-01 8.04889560e-01
8.22681665e-01 -7.31711611e-02 -1.49727970e-01 -3.77997756e-01
-2.49303132e-01 2.91100144e-01 7.52778351e-01 2.79894382e-01
-9.96975362e-01 1.86711401e-01 4.96258354e+00 9.44267273e-01
-1.20991933e+00 1.15516782e-01 8.16671133e-01 2.60092854e-01
3.04588914e-01 -7.31739327e-02 -2.28309423e-01 6.26216650e-01
1.05597031e+00 -1.13029212e-01 3.05060536e-01 4.08497334e-01
6.65303230e-01 -7.44150043e-01 -2.65372485e-01 1.38922942e+00
1.73688009e-02 -7.07126796e-01 -1.18317515e-01 2.93235350e-02
7.29573846e-01 -3.45292002e-01 1.17798358e-01 -2.90444255e-01
-1.18023872e-01 -9.82762635e-01 2.99530149e-01 9.39336061e-01
7.31418729e-01 -9.63897169e-01 1.07686555e+00 -7.44711200e-04
-1.04695618e+00 -1.64031386e-02 -6.63663745e-01 2.91417420e-01
2.89519459e-01 8.14492106e-01 -2.72532851e-01 9.59800720e-01
8.98325920e-01 2.97927022e-01 -6.03131473e-01 1.59201300e+00
8.56370330e-02 5.62694609e-01 -2.56000519e-01 1.94654867e-01
8.62035379e-02 -4.50195640e-01 4.38816905e-01 9.58108902e-01
8.85514021e-01 2.11057127e-01 -2.65833616e-01 4.54142153e-01
2.30322152e-01 5.99156260e-01 -1.82104036e-01 1.77292824e-01
3.51107240e-01 1.33440399e+00 -7.07478523e-01 -2.48700380e-01
-5.46385467e-01 9.21097517e-01 -7.42640734e-01 1.94500804e-01
-7.93118179e-01 -5.59068620e-01 3.17087173e-01 1.47683248e-01
2.14508206e-01 -1.84671104e-01 -1.64912269e-02 -6.56069994e-01
-4.14074473e-02 -1.05520737e+00 2.09807426e-01 -1.03428543e+00
-1.02121866e+00 8.05361748e-01 4.52273935e-02 -1.71461248e+00
1.91424102e-01 -3.43983531e-01 -3.79470080e-01 1.14965582e+00
-1.51406062e+00 -6.57023191e-01 -8.06651652e-01 5.49320996e-01
4.35049027e-01 -2.56089896e-01 5.58747709e-01 6.73702896e-01
-3.61458838e-01 4.62788999e-01 2.90805668e-01 -2.20758095e-01
5.31218171e-01 -7.87464738e-01 -4.16079134e-01 1.11434412e+00
-2.68003106e-01 9.93396416e-02 1.19431663e+00 -4.35347497e-01
-1.17350912e+00 -7.58396149e-01 4.75835383e-01 4.27648127e-01
2.28846654e-01 3.96648675e-01 -8.95081818e-01 -2.51550466e-01
2.97396332e-01 1.60109207e-01 5.26369989e-01 -6.10848010e-01
8.88648257e-02 -6.53342664e-01 -1.43026483e+00 1.19025260e-01
4.19873834e-01 -9.47181880e-02 -9.74303335e-02 2.65717972e-02
1.56305969e-01 1.06886737e-01 -1.24621189e+00 5.27126014e-01
5.91575325e-01 -1.49010730e+00 9.27756429e-01 2.43900940e-01
4.58842397e-01 -7.17298985e-01 -6.76952600e-01 -1.06280601e+00
2.47253999e-02 -6.59182891e-02 4.55060005e-01 1.45428360e+00
3.76689285e-01 -3.75991076e-01 5.24901390e-01 1.54888496e-01
1.75271198e-01 -5.42594373e-01 -6.59710407e-01 -8.30406129e-01
-5.91418684e-01 -2.94184566e-01 5.05973876e-01 4.60362971e-01
-2.55076766e-01 6.11038469e-02 -4.65143532e-01 2.81544894e-01
1.02534187e+00 -4.59199585e-02 6.36767805e-01 -1.11135221e+00
-2.99740613e-01 -1.22463189e-01 -9.97212946e-01 -5.43388605e-01
-6.31200552e-01 -4.68161166e-01 -2.51379430e-01 -1.70431709e+00
1.18572608e-01 -4.09965396e-01 -5.32811761e-01 -3.41695100e-01
7.84730166e-02 5.42450666e-01 2.56217629e-01 2.99114347e-01
-3.76872003e-01 5.25529206e-01 1.16974330e+00 -4.93621640e-03
5.89139313e-02 3.48269492e-01 -2.93328613e-01 4.16361928e-01
9.48122859e-01 -2.20549434e-01 -5.52102625e-01 -1.01462521e-01
3.42868343e-02 4.62318212e-01 1.71892464e-01 -1.74621511e+00
2.46682093e-02 3.87185290e-02 5.91985285e-01 -6.89479589e-01
4.24833447e-01 -1.14649069e+00 3.58981252e-01 4.39111441e-01
-1.06418699e-01 2.53718704e-01 -1.82141319e-01 4.37189549e-01
-6.83989584e-01 -3.28694701e-01 1.16667795e+00 -4.46994975e-02
-8.93746674e-01 1.52517691e-01 -5.66017367e-02 -4.87122387e-01
1.05723238e+00 -7.10265160e-01 3.02762184e-02 -6.45426869e-01
-4.95086938e-01 -2.16264859e-01 2.24525034e-01 1.00948878e-01
7.78012335e-01 -1.20914733e+00 -8.12337637e-01 -2.38904264e-03
1.46636531e-01 -5.90639412e-01 5.83950877e-01 1.01905632e+00
-9.30247009e-01 2.88762301e-01 -5.52674055e-01 -5.48594236e-01
-1.55568361e+00 5.04132450e-01 2.36740425e-01 -2.44916752e-01
-5.92021272e-02 4.19414639e-01 -1.31136775e-01 2.76236385e-01
6.29714429e-02 -1.15501538e-01 -5.81622541e-01 -1.35920212e-01
4.64024007e-01 6.57447398e-01 1.06227398e-01 -1.05456078e+00
-1.64733112e-01 7.49061584e-01 4.62224424e-01 -1.44381464e-01
1.30993605e+00 -6.19237721e-01 -6.26219884e-02 1.27408102e-01
1.34252882e+00 9.16984603e-02 -1.07856417e+00 -1.11540332e-01
-7.36353248e-02 -8.63593578e-01 2.74042487e-01 -9.13826823e-01
-1.31164408e+00 9.83045161e-01 1.46282339e+00 2.72657126e-01
1.80375540e+00 -5.89943588e-01 5.77994585e-01 -1.52133912e-01
2.96420515e-01 -1.17035949e+00 2.26203445e-03 -1.14883289e-01
8.25338244e-01 -1.26818538e+00 2.96686977e-01 -4.24114019e-01
-6.51067197e-01 1.11409545e+00 2.46857435e-01 -9.20633525e-02
7.48730183e-01 6.42519891e-02 8.26432835e-03 1.27469063e-01
-3.16932142e-01 -1.27593130e-01 2.59001464e-01 7.10523069e-01
5.75725317e-01 2.84241904e-02 -8.51959229e-01 1.32532761e-01
-1.45138249e-01 1.32255122e-01 5.26681900e-01 7.17505872e-01
-5.99890411e-01 -8.68043780e-01 -7.18281627e-01 5.20423770e-01
-1.02477980e+00 1.21399693e-01 5.88737667e-01 6.17474794e-01
1.65362567e-01 1.58171570e+00 -1.24930114e-01 -3.39164108e-01
3.05023521e-01 -3.97370875e-01 4.02329177e-01 1.22312613e-01
-2.10298926e-01 2.23970518e-01 -5.40599860e-02 -5.48568904e-01
-8.92198265e-01 -3.47978145e-01 -1.02141905e+00 -4.09690171e-01
-2.81298310e-01 2.54847914e-01 1.02131617e+00 5.58677137e-01
-7.83338249e-02 4.87061560e-01 7.48970151e-01 -4.85531479e-01
-2.70480692e-01 -9.40286338e-01 -8.45629930e-01 5.71548045e-01
-6.27302080e-02 -4.98291641e-01 -3.21078151e-01 2.71541655e-01]
|
[11.719757080078125, -1.9838731288909912]
|
a6b125e5-b4e3-4b74-b83e-5ff0e3333efb
|
text-conditional-alt-text-generation-for
|
2305.14779
| null |
https://arxiv.org/abs/2305.14779v1
|
https://arxiv.org/pdf/2305.14779v1.pdf
|
Text Conditional Alt-Text Generation for Twitter Images
|
In this work we present an approach for generating alternative text (or alt-text) descriptions for images shared on social media, specifically Twitter. This task is more than just a special case of image captioning, as alt-text is both more literally descriptive and context-specific. Also critically, images posted to Twitter are often accompanied by user-written text that despite not necessarily describing the image may provide useful context that if properly leveraged can be informative -- e.g. the tweet may name an uncommon object in the image that the model has not previously seen. We address this with a CLIP prefix model that extracts an embedding of the image and passes it to a mapping network that outputs a short sequence in word embedding space, or a ``prefix'', to which we also concatenate the text from the tweet itself. This lets the model condition on both visual and textual information from the post. The combined multimodal prefix is then fed as a prompt to a pretrained language model which autoregressively completes the sequence to generate the alt-text. While prior work has used similar methods for captioning, ours is the first to our knowledge that incorporates textual information from the associated social media post into the prefix as well, and we further demonstrate through ablations that utility of these two information sources stacks. We put forward a new dataset scraped from Twitter and evaluate on it across a variety of automated metrics as well as human evaluation, and show that our approach of conditioning on both tweet text and visual information significantly outperforms prior work.
|
['Taylor Berg-Kirkpatrick', 'Omar Florez', 'Sofia Samaniego', 'Nikita Srivatsan']
|
2023-05-24
| null | null | null | null |
['image-captioning']
|
['computer-vision']
|
[ 6.23431265e-01 4.10796881e-01 -3.83972526e-02 -5.37850976e-01
-9.67459083e-01 -8.69287968e-01 1.25523973e+00 3.49714160e-01
-6.15539551e-01 5.95553041e-01 6.60258770e-01 -3.53870898e-01
5.37849367e-01 -4.72361982e-01 -9.98960972e-01 -4.70996857e-01
1.43775657e-01 4.60222423e-01 -1.99645571e-02 -1.95615634e-01
1.66047662e-01 1.76412016e-01 -1.48095298e+00 5.59815347e-01
1.70450851e-01 7.68595099e-01 2.63408512e-01 6.87883198e-01
-3.19335163e-01 6.53377593e-01 -4.87182617e-01 -7.02518225e-01
1.32540509e-01 -3.08564961e-01 -8.23746562e-01 3.56803179e-01
8.60874712e-01 -4.65020955e-01 -4.71744835e-01 6.37533009e-01
3.22892994e-01 -6.95030252e-03 7.50901759e-01 -1.30234671e+00
-8.12877953e-01 8.29306006e-01 -6.07436717e-01 -3.72223626e-03
6.01271331e-01 4.66305405e-01 1.07061398e+00 -1.10312772e+00
1.11188853e+00 1.12942207e+00 6.00546539e-01 4.20001268e-01
-1.49610114e+00 -4.59195971e-01 1.87145710e-01 -3.69746774e-01
-1.09408319e+00 -3.98627490e-01 5.42798698e-01 -6.08224213e-01
6.92892134e-01 2.24354848e-01 5.53853571e-01 1.66468120e+00
-5.96565902e-02 8.07103217e-01 1.03105628e+00 -3.57904285e-01
-1.83868110e-01 5.73032141e-01 -1.73662946e-01 5.98125875e-01
-5.94846867e-02 -2.04404071e-01 -4.97666121e-01 7.42689706e-03
4.06593919e-01 -1.24661870e-01 -1.19399637e-01 -6.54418468e-02
-1.50085425e+00 8.05102229e-01 6.16588771e-01 2.21421868e-01
-3.87975752e-01 3.71843100e-01 4.23846632e-01 1.97116397e-02
6.40891194e-01 5.19410074e-01 1.03088535e-01 1.43647596e-01
-1.29570448e+00 3.39690953e-01 8.03862929e-01 9.65282261e-01
9.91329312e-01 -3.28835189e-01 -3.84215772e-01 5.53465426e-01
2.15012729e-01 5.84780455e-01 3.54342371e-01 -6.51791215e-01
5.98824918e-01 3.22927177e-01 3.06296557e-01 -1.21094131e+00
-2.07245573e-01 -2.49240965e-01 -3.33316922e-01 -7.32037947e-02
4.11364615e-01 -3.29275519e-01 -9.77726936e-01 1.85251462e+00
-5.10797184e-03 2.18453497e-01 -5.28812110e-02 8.33759785e-01
1.02123535e+00 9.64725256e-01 3.94505829e-01 5.80915436e-02
1.63906133e+00 -8.36997390e-01 -6.87669039e-01 -6.43451333e-01
4.02721018e-01 -1.00521183e+00 1.25345778e+00 -1.51557356e-01
-1.13394058e+00 -2.90809959e-01 -9.66022193e-01 -3.62910837e-01
-7.38606274e-01 6.51836395e-03 2.25708038e-01 1.92244262e-01
-1.29677641e+00 3.09812129e-01 -4.70090151e-01 -8.72904241e-01
3.28783840e-01 -3.57357748e-02 -5.73867381e-01 -4.53182198e-02
-1.08990085e+00 9.88371134e-01 2.62152404e-01 -1.92571461e-01
-7.02531099e-01 -5.27584493e-01 -1.02704632e+00 -3.01629961e-01
3.02023053e-01 -9.70016599e-01 1.37329280e+00 -1.41784573e+00
-1.08344543e+00 1.28575397e+00 -3.39025170e-01 -5.23529708e-01
6.58910096e-01 -1.18941657e-01 -2.97988653e-01 4.99924153e-01
4.59967077e-01 1.38899708e+00 1.12642515e+00 -1.56978297e+00
-4.36386436e-01 6.25391379e-02 3.18295598e-01 2.95971960e-01
-3.28862578e-01 1.73930153e-01 -6.58918619e-01 -7.67685771e-01
-2.86082685e-01 -1.18883169e+00 1.82434414e-02 1.93295121e-01
-7.63276041e-01 2.43638363e-02 9.48436558e-01 -6.53608143e-01
1.07680845e+00 -2.13314533e+00 -9.75904763e-02 2.13021606e-01
6.50784299e-02 -1.11046724e-01 -4.34054673e-01 1.03052092e+00
-2.01652169e-01 6.04028046e-01 -4.05452996e-01 -7.85749376e-01
7.14823082e-02 2.01485023e-01 -6.95131361e-01 4.16485220e-01
5.12835562e-01 1.21988583e+00 -1.02812505e+00 -5.90166211e-01
1.49179056e-01 6.81387544e-01 -3.13018084e-01 1.12703525e-01
-6.13044620e-01 1.94415972e-01 -2.67920643e-01 1.42561600e-01
4.02032137e-01 -5.32892823e-01 -1.38482571e-01 -5.02352655e-01
-2.77998626e-01 1.80922106e-01 -7.44315088e-01 1.56089449e+00
-4.94788080e-01 1.03647172e+00 -1.39306337e-01 -6.00301385e-01
5.18554807e-01 3.38130891e-01 3.41552556e-01 -2.13686138e-01
1.04125783e-01 2.88711544e-02 -5.41691303e-01 -8.23891103e-01
7.94623852e-01 -1.77456066e-01 -2.11794034e-01 7.58304834e-01
-1.03342474e-01 -2.82764375e-01 1.49837524e-01 7.26282239e-01
8.64554346e-01 4.76837248e-01 6.55698031e-02 1.57217309e-01
2.88488239e-01 1.83065042e-01 -4.33533490e-01 9.12575185e-01
8.62086341e-02 1.05672932e+00 5.23039103e-01 -7.77395070e-02
-1.48492122e+00 -9.37815547e-01 -1.93948597e-02 1.13473308e+00
1.01814330e-01 -5.36755443e-01 -6.22976422e-01 -7.14745283e-01
9.72779002e-03 8.85929346e-01 -9.44917202e-01 2.02963278e-01
-4.24700886e-01 -3.42048854e-01 6.20068073e-01 2.17605978e-01
2.41909966e-01 -1.19216132e+00 -4.11716074e-01 2.24261135e-01
-3.70745838e-01 -1.39575016e+00 -7.87224650e-01 -1.54281646e-01
-2.01330438e-01 -7.29374409e-01 -7.82737792e-01 -8.77145946e-01
8.07471216e-01 2.06804767e-01 1.13288188e+00 1.42587990e-01
-1.60085782e-01 8.56075406e-01 -3.79224509e-01 -3.99842948e-01
-5.61858773e-01 8.59632343e-03 -4.32784259e-01 5.03096879e-01
7.11729899e-02 -2.67662406e-01 -5.46018302e-01 -3.66347805e-02
-1.44022155e+00 5.65517485e-01 6.04074419e-01 5.54291964e-01
3.67199033e-01 -5.89773893e-01 3.71732682e-01 -1.12312770e+00
6.05689883e-01 -8.32968116e-01 -1.24307565e-01 -1.10986093e-02
-1.90947130e-01 8.35680962e-02 4.10175055e-01 -5.17915010e-01
-8.40644658e-01 2.63032526e-01 3.59140672e-02 -2.81280816e-01
-1.85694709e-01 7.93531656e-01 3.24868292e-01 3.39006960e-01
6.54995739e-01 2.83416033e-01 2.24759489e-01 -1.82288751e-01
6.90487802e-01 5.78540742e-01 7.29122758e-01 -3.60749245e-01
1.20097077e+00 8.33634257e-01 -6.18016906e-03 -9.18884456e-01
-8.73469591e-01 -4.54257548e-01 -3.79549772e-01 -3.33031446e-01
1.01431096e+00 -9.13223147e-01 -3.19357127e-01 3.69314030e-02
-1.39623344e+00 -1.57168463e-01 -1.84150845e-01 1.79266140e-01
-6.05206370e-01 3.45819861e-01 -4.65177298e-01 -6.45365775e-01
-1.61140691e-02 -1.04344547e+00 1.25388205e+00 -4.71045896e-02
-4.01909709e-01 -1.08660793e+00 -2.57089347e-01 1.10815190e-01
4.50809002e-01 6.56764627e-01 5.81312895e-01 -9.01817799e-01
-5.80375135e-01 -4.41783220e-01 -5.20381987e-01 6.20693937e-02
-2.23985925e-01 2.22086489e-01 -1.10517132e+00 -1.11077152e-01
-3.84124428e-01 -5.23119390e-01 9.24243093e-01 1.56066502e-02
9.00664806e-01 -7.76005685e-01 -4.44208354e-01 2.96452969e-01
1.61877966e+00 -4.91473287e-01 6.35777533e-01 6.78158224e-01
5.84314525e-01 9.50500488e-01 4.03905720e-01 3.26785773e-01
6.41416788e-01 6.27593875e-01 6.84072077e-01 -3.61327916e-01
-1.97593525e-01 -5.86115599e-01 6.01082385e-01 3.15701634e-01
3.58097732e-01 -5.44285476e-01 -8.00940037e-01 7.33893335e-01
-1.67822969e+00 -1.27246523e+00 -1.13183230e-01 1.91329622e+00
9.41379905e-01 -7.44729266e-02 1.30380824e-01 -3.94127041e-01
7.45155990e-01 5.13639629e-01 -2.22494364e-01 -2.75847375e-01
-3.18145961e-01 -1.04942270e-01 5.46466172e-01 7.59529948e-01
-1.20546114e+00 8.87429297e-01 6.01614666e+00 3.91702622e-01
-1.36667562e+00 3.07634324e-02 6.06769562e-01 -1.63235098e-01
-7.10714161e-01 9.27696750e-02 -6.47627950e-01 4.70483780e-01
9.69778657e-01 -1.02205209e-01 2.64136791e-01 4.42620605e-01
3.35179031e-01 -1.33951128e-01 -1.35912406e+00 7.87840247e-01
5.15055120e-01 -1.41601503e+00 4.69827414e-01 5.02329171e-02
6.18864775e-01 1.10049583e-01 4.78148699e-01 4.21474539e-02
-2.39886232e-02 -1.10374653e+00 1.11962688e+00 5.27650952e-01
9.95421767e-01 -3.09095770e-01 5.48217058e-01 7.31996670e-02
-8.56437385e-01 2.44689748e-01 2.00162485e-01 2.75015622e-01
4.18615818e-01 2.51451164e-01 -1.38609052e+00 3.54182363e-01
2.11760789e-01 8.49152088e-01 -9.20563400e-01 8.21964324e-01
-3.19436729e-01 4.92370397e-01 -3.02877337e-01 -1.17336094e-01
7.57795155e-01 1.51929885e-01 7.92484164e-01 1.78432155e+00
2.72243679e-01 -1.42701134e-01 2.11279169e-01 8.26612175e-01
-2.48487145e-01 2.60973245e-01 -9.42327023e-01 -3.90994012e-01
4.29879338e-01 1.42016292e+00 -7.60792375e-01 -5.26413918e-01
-6.08371854e-01 1.10959005e+00 1.94169238e-01 6.26812160e-01
-8.43037367e-01 -2.22514734e-01 1.88794792e-01 5.11735141e-01
2.27927744e-01 -1.09643504e-01 7.94422626e-02 -1.11582589e+00
6.43090382e-02 -6.64449811e-01 2.70691991e-01 -1.51394916e+00
-1.40275395e+00 7.92175591e-01 2.11308792e-01 -1.10317910e+00
-4.41465348e-01 -4.18718308e-01 -4.80994850e-01 9.65170443e-01
-1.71758544e+00 -1.60439456e+00 -2.96431541e-01 3.87638867e-01
5.21484971e-01 2.49448672e-01 6.38785779e-01 7.42137283e-02
-2.57940650e-01 4.74515259e-01 -2.10066289e-01 1.61309212e-01
1.07553840e+00 -1.20973682e+00 3.35255504e-01 7.74516106e-01
2.69655913e-01 6.19611859e-01 1.25593519e+00 -6.79076850e-01
-1.21965694e+00 -1.24211657e+00 1.16524899e+00 -7.59811521e-01
1.01154578e+00 -4.85895157e-01 -7.58839846e-01 1.21997452e+00
8.06914032e-01 -3.23958039e-01 5.94619095e-01 -4.13680404e-01
-5.37287951e-01 3.28533739e-01 -9.59374249e-01 1.00921345e+00
7.30826795e-01 -8.16593349e-01 -6.94043159e-01 5.60980141e-01
8.74780178e-01 -3.17300260e-01 -3.79122734e-01 -1.50229141e-01
5.08128524e-01 -6.98111594e-01 9.14644003e-01 -5.57609558e-01
9.05842245e-01 -3.68478209e-01 -1.06028117e-01 -1.09686565e+00
8.91260505e-02 -7.86152065e-01 4.65212137e-01 1.55174220e+00
6.44681454e-01 -4.40223485e-01 6.12726688e-01 6.37538373e-01
-1.49781346e-01 -4.82629389e-01 -6.24562979e-01 -3.69774073e-01
-1.42541483e-01 -5.54021358e-01 2.98134685e-01 9.98358250e-01
-2.20612362e-02 4.86439914e-01 -4.63466972e-01 -5.10683581e-02
4.18248653e-01 -2.09460065e-01 8.16181898e-01 -6.76159024e-01
1.04024686e-01 -4.96285081e-01 -1.88539535e-01 -9.37557161e-01
2.15546116e-01 -1.21228862e+00 8.01330730e-02 -1.73756623e+00
3.26420724e-01 -2.19326764e-01 2.80733615e-01 5.83497286e-01
-7.23766610e-02 7.34729052e-01 2.98730552e-01 3.57199907e-01
-5.39696753e-01 1.17484510e-01 1.18422294e+00 -1.78659856e-01
7.50121102e-02 -5.44851840e-01 -8.26121509e-01 5.42523801e-01
4.55692828e-01 -3.63226414e-01 -2.85232097e-01 -4.07113105e-01
5.41971087e-01 -1.66682839e-01 8.23571801e-01 -7.08363354e-01
1.66672051e-01 -1.57629047e-02 3.68337065e-01 -4.72686768e-01
6.20624542e-01 -8.81670296e-01 8.65751356e-02 5.94618656e-02
-8.27099621e-01 1.29798159e-01 3.19614023e-01 6.73154652e-01
-1.23622134e-01 -1.79258794e-01 5.12705982e-01 -2.32489064e-01
-6.02570355e-01 3.14004898e-01 -4.91230905e-01 -7.59054795e-02
8.99067223e-01 -4.03681338e-01 -4.67282057e-01 -9.68813539e-01
-8.39957476e-01 2.65647829e-01 7.34401584e-01 5.46041787e-01
5.40688932e-01 -1.32617629e+00 -9.92538393e-01 1.37276913e-03
4.58037019e-01 -3.74280661e-01 6.43866695e-03 8.30750108e-01
-3.72348517e-01 3.14398885e-01 6.77044839e-02 -6.58410549e-01
-1.02961886e+00 5.73154390e-01 2.55405307e-02 7.39771500e-02
-6.83565915e-01 6.20095074e-01 2.99601346e-01 -9.91154835e-02
8.80715102e-02 -1.30192220e-01 -2.98509240e-01 5.12431920e-01
6.03399932e-01 -4.20911521e-01 -3.47381830e-01 -1.18364370e+00
-1.85871422e-01 4.94378090e-01 -1.69797000e-02 -7.76056647e-01
1.27574205e+00 -5.82608283e-01 -1.04496330e-01 6.11869574e-01
1.59583855e+00 1.35856003e-01 -1.13111830e+00 -2.69832402e-01
-3.57089490e-02 -3.99138540e-01 -2.60480136e-01 -9.06461179e-01
-6.49087250e-01 7.65230477e-01 1.52711406e-01 5.22069216e-01
6.77497268e-01 4.35082674e-01 7.09255695e-01 1.93077579e-01
-2.04553291e-01 -7.64470756e-01 4.10856277e-01 4.42919195e-01
1.32135963e+00 -1.29722667e+00 7.00364262e-03 -1.78230673e-01
-9.37764704e-01 1.17327571e+00 3.53225112e-01 -7.53048733e-02
4.06204700e-01 6.02313280e-02 2.18027458e-01 -2.34547839e-01
-7.86036015e-01 -2.06585571e-01 3.63115817e-01 5.54434538e-01
5.15803814e-01 -2.72536635e-01 -5.43338731e-02 2.62055248e-01
-3.98404390e-01 -1.66472316e-01 8.93023014e-01 7.88051665e-01
-3.05506855e-01 -9.12411034e-01 -4.01437908e-01 3.94980669e-01
-5.99975705e-01 -3.65952224e-01 -5.34128904e-01 8.33302617e-01
3.64563949e-02 8.12377512e-01 3.66523385e-01 -2.13675573e-01
8.94192383e-02 2.62684077e-01 2.09496096e-01 -8.88401747e-01
-7.64011741e-01 2.13421434e-02 3.68726879e-01 -3.87295157e-01
-5.83835840e-01 -7.80952215e-01 -1.15407252e+00 -1.42421201e-01
8.84911418e-02 -6.71096295e-02 1.10657489e+00 8.55189145e-01
2.95891404e-01 1.48600847e-01 4.37145323e-01 -1.24576402e+00
-3.37015204e-02 -8.71022522e-01 -6.94293603e-02 9.08552647e-01
7.44080782e-01 -2.60009557e-01 -5.76067090e-01 5.20760298e-01]
|
[11.030924797058105, 0.9951381683349609]
|
88c789f0-4b12-46ff-9cf3-6aa88594f1cf
|
neural-network-based-automatic-liver-tumor
|
1706.00842
| null |
http://arxiv.org/abs/1706.00842v3
|
http://arxiv.org/pdf/1706.00842v3.pdf
|
Neural Network-Based Automatic Liver Tumor Segmentation With Random Forest-Based Candidate Filtering
|
We present a fully automatic method employing convolutional neural networks
based on the 2D U-net architecture and random forest classifier to solve the
automatic liver lesion segmentation problem of the ISBI 2017 Liver Tumor
Segmentation Challenge (LiTS). In order to constrain the ROI in which the
tumors could be located, a liver segmentation is performed first. For the organ
segmentation, an ensemble of convolutional networks is trained to segment a
liver using a set of 179 liver CT datasets from liver surgery planning. Inside
of the liver ROI a neural network, trained using 127 challenge training
datasets, identifies tumor candidates, which are subsequently filtered with a
random forest classifier yielding the final tumor segmentation. The evaluation
on the 70 challenge test cases resulted in a mean Dice coefficient of 0.65,
ranking our method in the second place.
|
['Andrea Schenk', 'Grzegorz Chlebus', 'Jan Hendrik Moltz', 'Hans Meine']
|
2017-06-02
| null | null | null | null |
['liver-segmentation']
|
['medical']
|
[-2.69225100e-03 5.37395000e-01 -4.07152742e-01 -4.13207620e-01
-5.51574409e-01 -5.67880988e-01 5.41779935e-01 1.21121481e-01
-3.36608350e-01 5.20061970e-01 4.29439843e-01 -6.62755072e-01
6.93507046e-02 -5.64738095e-01 -4.73080546e-01 -7.38776743e-01
-4.79251802e-01 8.92999291e-01 4.01768796e-02 5.19700825e-01
-1.38073638e-01 8.45904350e-01 -2.63317525e-01 3.60599667e-01
8.09574723e-01 1.23797476e+00 -3.14235695e-05 5.85772634e-01
7.93748945e-02 8.40908825e-01 -1.38201475e-01 1.45616025e-01
6.30214691e-01 -5.66986978e-01 -1.25814033e+00 1.88245460e-01
1.14456430e-01 -3.39713335e-01 -2.34918162e-01 8.90065789e-01
4.85328019e-01 -3.08553785e-01 8.15896332e-01 -4.51180279e-01
-2.09005177e-02 1.35912359e+00 -5.74755445e-02 3.50553066e-01
9.12618171e-03 3.28927130e-01 6.07103467e-01 -8.80853951e-01
7.40766943e-01 4.61945593e-01 7.90452600e-01 4.55357164e-01
-1.13124871e+00 -3.40441555e-01 -5.34006596e-01 -2.57678509e-01
-1.28746319e+00 -6.91276640e-02 2.89412826e-01 -8.50992382e-01
6.78464472e-01 3.62602949e-01 1.02629066e+00 4.34108168e-01
4.10067081e-01 8.11691105e-01 1.07833230e+00 -3.76534224e-01
1.47173837e-01 -1.29065007e-01 2.25361496e-01 1.05725420e+00
2.03163147e-01 2.25190490e-01 4.50519085e-01 -6.81702942e-02
7.65649915e-01 -1.96016073e-01 -5.24832785e-01 -6.03226066e-01
-1.77735829e+00 9.17766750e-01 1.35276246e+00 4.53869760e-01
-7.92744517e-01 1.86949912e-02 6.59895420e-01 -2.17436403e-01
3.43348533e-01 5.81813931e-01 -2.50456572e-01 7.45023012e-01
-1.12617278e+00 -3.51440966e-01 9.77997303e-01 4.35476929e-01
2.75196463e-01 -3.49922121e-01 -8.51283252e-01 1.57018512e-01
5.55758595e-01 -1.99194074e-01 7.29052246e-01 -3.10599267e-01
-2.80861318e-01 6.36988521e-01 -2.50255287e-01 -9.14065018e-02
-8.78273010e-01 -7.49821901e-01 -1.18917227e+00 2.50475109e-01
8.20890605e-01 -4.66921508e-01 -1.46936202e+00 9.79189038e-01
3.62251669e-01 2.58743107e-01 5.51956296e-02 1.17151380e+00
1.19817948e+00 2.93379072e-02 2.63445556e-01 -2.32833833e-01
1.38664591e+00 -1.33984721e+00 -2.26782039e-01 2.05496788e-01
8.52251410e-01 -7.51901805e-01 2.76284397e-01 1.45850834e-02
-7.88350999e-01 -1.34285122e-01 -8.15362751e-01 4.02152002e-01
-9.79735106e-02 3.35021079e-01 5.97425103e-01 7.00223327e-01
-1.20140481e+00 4.87405002e-01 -1.01413083e+00 -4.04048622e-01
8.12909305e-01 5.87664783e-01 -1.74729764e-01 -9.15427692e-03
-9.58720565e-01 1.27913737e+00 6.74246192e-01 2.40176156e-01
-1.48020327e+00 -9.20473754e-01 -9.21903908e-01 1.40366748e-01
1.29743308e-01 -9.30590034e-01 1.04401469e+00 -9.70256031e-01
-1.53117394e+00 1.09223628e+00 4.24151599e-01 -1.09538972e+00
1.07887042e+00 4.12692457e-01 2.11452603e-01 2.66708642e-01
8.13499913e-02 8.60436618e-01 3.74917895e-01 -9.22043979e-01
-1.93940163e-01 1.99348122e-01 -3.76258194e-01 1.00428201e-01
4.91101414e-01 2.01850329e-02 -3.37328494e-01 -4.51132298e-01
1.16201147e-01 -1.13502431e+00 -7.15435565e-01 -2.55136430e-01
-9.57618892e-01 -1.94470763e-01 4.06982034e-01 -9.86286938e-01
7.21230388e-01 -1.66561258e+00 8.26500356e-02 7.46432781e-01
4.22951639e-01 -9.75397788e-03 1.49189964e-01 -5.33959985e-01
-5.18026114e-01 8.35436359e-02 -1.52067661e-01 2.43831590e-01
-3.81152481e-01 -6.41017109e-02 3.58859122e-01 8.21029961e-01
-1.01155341e-01 1.25040686e+00 -9.03656602e-01 -5.96992314e-01
4.98211861e-01 2.61587799e-01 -3.55997860e-01 3.75895619e-01
-2.13251621e-01 8.85193467e-01 -2.34013706e-01 7.39625335e-01
4.83269989e-01 -3.05480659e-01 1.79185718e-01 -3.37912858e-01
-5.42476662e-02 5.81162348e-02 -4.78102028e-01 1.74261880e+00
-2.56659120e-01 4.34585035e-01 -7.86987226e-03 -7.17577577e-01
6.94582999e-01 6.39927030e-01 1.16207623e+00 2.60465741e-02
4.67111409e-01 2.31033683e-01 6.33539855e-01 -5.64168692e-01
-4.61333454e-01 4.92702946e-02 1.74479306e-01 3.93132746e-01
1.69230178e-01 -1.35803059e-01 3.37877452e-01 -8.17726478e-02
1.15939164e+00 1.09439924e-01 7.03072906e-01 -9.36229229e-01
8.88924718e-01 3.64267886e-01 3.30847412e-01 5.31536162e-01
-7.53728747e-01 7.42510676e-01 5.70396304e-01 -1.24240482e+00
-6.51192069e-01 -8.34613800e-01 -4.07826483e-01 4.21960533e-01
-2.23824739e-01 6.32774010e-02 -9.58706498e-01 -1.41191840e+00
-1.68261677e-01 5.11573911e-01 -1.00729275e+00 2.87711203e-01
-6.50460184e-01 -9.52399731e-01 3.72856200e-01 1.74822733e-01
3.58127534e-01 -1.16872132e+00 -9.96821404e-01 1.95431575e-01
-1.78201273e-01 -8.37990761e-01 -5.98300874e-01 6.00639522e-01
-7.95705438e-01 -1.63012528e+00 -1.07255697e+00 -8.32355857e-01
1.18426251e+00 -3.21058363e-01 1.32942212e+00 2.64335781e-01
-5.21102607e-01 -1.17316591e-02 5.42649701e-02 -2.30692595e-01
-8.29363048e-01 2.25688934e-01 -1.87653750e-01 -1.80168241e-01
2.96795994e-01 1.24497131e-01 -9.04011548e-01 2.72153229e-01
-4.22187388e-01 3.58165890e-01 9.09728646e-01 1.07872748e+00
6.60599232e-01 -3.48532349e-01 1.87274531e-01 -8.03864121e-01
1.73061177e-01 -6.01747870e-01 -7.13526011e-01 3.29381645e-01
-4.18296158e-01 -9.39831957e-02 4.10881430e-01 -3.42666000e-01
-5.50306380e-01 9.83917236e-01 -7.26559535e-02 -3.06594193e-01
-3.14349562e-01 7.95088112e-01 2.87730396e-01 -3.88656795e-01
8.74981523e-01 9.98501759e-03 1.77162319e-01 1.21353075e-01
3.08619916e-01 1.42758623e-01 3.15327346e-01 -4.63831648e-02
4.77443069e-01 8.51404816e-02 2.10162312e-01 -2.12153867e-01
-7.17879534e-01 -3.33809078e-01 -1.03002667e+00 -3.67091119e-01
1.28387713e+00 -7.13062644e-01 -3.12826931e-01 2.71749198e-02
-8.75265479e-01 -6.52090728e-01 -4.53611761e-01 8.83466423e-01
-4.63589370e-01 2.35426072e-02 -8.09172809e-01 -7.96352848e-02
-8.52246761e-01 -1.77067149e+00 3.30465645e-01 2.70369142e-01
-2.92717218e-01 -8.84035468e-01 -5.89321628e-02 2.73522977e-02
6.35039985e-01 5.11898696e-01 7.95512497e-01 -1.36072767e+00
-7.45519519e-01 -4.18229103e-01 -3.62393379e-01 1.61284208e-03
9.86259505e-02 -1.23521641e-01 -7.07446396e-01 -2.95375407e-01
-3.25720161e-01 -1.08916618e-01 8.74392033e-01 1.13984418e+00
1.12544179e+00 -7.88613409e-02 -6.33348465e-01 9.68433082e-01
1.29045045e+00 9.35707390e-02 1.78429246e-01 1.68451235e-01
4.52000797e-01 5.14033325e-02 -1.65635180e-02 1.08686641e-01
9.46809500e-02 1.39501661e-01 8.69151890e-01 -4.74612415e-01
-5.47105193e-01 1.94572657e-01 -2.21988678e-01 3.54177505e-01
-8.75991806e-02 1.72728807e-01 -1.41408932e+00 6.59640312e-01
-1.30928612e+00 -3.84128571e-01 -8.16279650e-02 2.02283311e+00
7.63874948e-01 3.81678343e-02 3.65287140e-02 -4.44916248e-01
6.24232888e-01 -2.67972410e-01 -1.82005495e-01 9.85113531e-02
3.93396765e-01 -1.60932932e-02 8.03534508e-01 5.85976899e-01
-1.52080631e+00 6.84416890e-01 6.42496204e+00 3.78715068e-01
-1.50653374e+00 7.22345849e-03 1.34383523e+00 3.53385150e-01
1.57184049e-01 -8.71489421e-02 -3.12567174e-01 2.93391734e-01
6.52664065e-01 -2.47762464e-02 1.38863936e-01 7.65630603e-01
1.99898303e-01 -2.63697594e-01 -1.21443427e+00 4.65701908e-01
-1.25177592e-01 -1.75320971e+00 -2.15530306e-01 6.26865774e-02
9.31428313e-01 7.62624860e-01 -4.34865147e-01 1.54201791e-01
4.98031288e-01 -1.43596005e+00 1.50321543e-01 6.17264748e-01
8.82201970e-01 -3.84058535e-01 1.20168483e+00 3.44062507e-01
-8.55928421e-01 3.05463731e-01 -8.90197605e-02 4.98774141e-01
-6.59719184e-02 7.21492112e-01 -2.00561833e+00 5.93030095e-01
4.08762991e-01 4.20203507e-01 -5.71969271e-01 1.67656970e+00
-2.74347365e-01 7.83010602e-01 -5.23758054e-01 9.02962685e-02
2.75145233e-01 -7.28568062e-02 5.02635241e-01 1.49547887e+00
3.48341614e-01 -7.00414926e-02 6.56174541e-01 8.53797793e-01
-1.97947621e-01 3.14572155e-01 -4.17260408e-01 4.81301546e-01
-1.63684338e-01 1.84962118e+00 -1.41357994e+00 -5.02175033e-01
-1.24457121e-01 8.32066536e-01 -9.78447050e-02 4.55194041e-02
-6.66764796e-01 2.13406757e-01 -2.22991079e-01 -9.20912698e-02
-4.66054901e-02 2.71425486e-01 -5.49016118e-01 -1.22440434e+00
-7.91429281e-01 -3.91718179e-01 6.32934391e-01 -3.22192490e-01
-9.76531148e-01 1.03482950e+00 -3.26823831e-01 -1.19749129e+00
-3.46804470e-01 -5.53373337e-01 -1.04261661e+00 1.13638020e+00
-1.53721821e+00 -1.38506389e+00 -5.34056187e-01 2.45193690e-01
4.72810656e-01 -2.30031967e-01 1.01087952e+00 -5.21355458e-02
-2.59522080e-01 1.38980314e-01 -4.03135717e-01 8.69915962e-01
4.15200800e-01 -1.61682355e+00 2.42938098e-04 8.66662621e-01
-1.04610085e-01 1.58742636e-01 4.14133370e-01 -7.53252625e-01
-9.44562912e-01 -1.33075058e+00 7.53801227e-01 -2.98964411e-01
5.69390178e-01 2.31151670e-01 -5.00628173e-01 9.25399601e-01
6.30141199e-01 9.16057408e-01 7.47501850e-01 -6.29076302e-01
2.14829594e-01 3.60027373e-01 -1.46706808e+00 2.32200146e-01
1.39154464e-01 1.77126065e-01 -3.73747408e-01 8.19166958e-01
2.66916305e-01 -1.00496209e+00 -1.32242048e+00 7.18479455e-01
3.70345533e-01 -6.30382001e-01 1.05036044e+00 -5.63023925e-01
5.75318456e-01 -2.79229343e-01 4.65390295e-01 -1.67136168e+00
-4.46370989e-01 -4.29943472e-01 1.05138421e-01 2.71103710e-01
6.88606560e-01 -1.94370314e-01 1.05913854e+00 5.18092752e-01
-3.82935286e-01 -1.05370855e+00 -9.43296552e-01 8.15598294e-03
4.46723402e-01 1.22407235e-01 2.90011168e-01 9.65285778e-01
-1.06611475e-01 -3.41457091e-02 2.05519632e-01 1.16936468e-01
7.22641766e-01 6.68592229e-02 2.52336502e-01 -1.14286017e+00
1.32038236e-01 -9.97170269e-01 -1.81240410e-01 -6.82139993e-01
1.60442188e-01 -1.53674006e+00 2.44062096e-01 -1.79435730e+00
3.34297687e-01 -5.92800319e-01 -5.37744284e-01 6.01475358e-01
-1.25100970e-01 3.25432330e-01 3.19215134e-02 3.40776861e-01
-2.43570954e-01 -1.71796247e-01 1.41652358e+00 -4.02513802e-01
-5.95941059e-02 5.79998434e-01 -8.10085535e-02 7.40863919e-01
7.14129269e-01 -2.02020317e-01 2.82800734e-01 -1.56118767e-02
-4.33012575e-01 4.83020395e-01 2.85039157e-01 -1.01782477e+00
3.57315809e-01 3.94088514e-02 1.05565262e+00 -7.94696212e-01
-4.71595585e-01 -1.18720078e+00 6.01421222e-02 1.47552454e+00
-4.89727020e-01 -2.51458228e-01 -7.56081566e-02 -2.63957791e-02
-1.37212753e-01 -3.47927243e-01 1.11369288e+00 -5.92625499e-01
-2.58500636e-01 6.92947686e-01 -5.81890941e-01 -4.94424194e-01
1.47298014e+00 1.90814942e-01 2.24375799e-01 3.84423137e-02
-1.38299334e+00 3.87289584e-01 1.00883637e-02 -1.50741965e-01
2.78122127e-01 -1.19869649e+00 -1.12721229e+00 3.34509939e-01
-1.32291108e-01 4.54502761e-01 -8.67717266e-02 1.55744934e+00
-1.21217990e+00 5.88894069e-01 -1.85819730e-01 -8.58799100e-01
-9.13691700e-01 6.22852743e-01 1.29004920e+00 -9.09380555e-01
-9.31900501e-01 9.56204236e-01 -3.73258106e-02 -5.32425046e-01
7.83270746e-02 -6.00257993e-01 -4.18235600e-01 -1.07354738e-01
2.61719555e-01 -1.36426479e-01 2.63189465e-01 -5.92079520e-01
-4.22675937e-01 -7.02431332e-03 2.05817312e-01 3.05100381e-01
1.14240777e+00 2.81761408e-01 -5.03862500e-01 -4.09406513e-01
9.61703420e-01 -1.12092987e-01 -1.10802495e+00 -2.64840662e-01
3.49383116e-01 3.30652110e-02 2.41737932e-01 -1.39212704e+00
-1.47886240e+00 4.33798879e-01 8.53613257e-01 2.49528289e-01
9.95190799e-01 -1.39237776e-01 3.58318329e-01 -4.77699004e-02
-2.22188085e-01 -2.25633189e-01 -5.22237718e-01 6.38212562e-01
8.79573941e-01 -1.58193719e+00 2.05316637e-02 -5.01976252e-01
-5.13099492e-01 1.80960500e+00 5.55349290e-01 -4.65939373e-01
9.39480782e-01 4.50263411e-01 4.94370461e-01 -2.38405138e-01
-3.37428868e-01 -1.66513473e-01 6.11793160e-01 3.28080863e-01
6.82033181e-01 5.29215693e-01 -3.42754662e-01 4.36402500e-01
-4.95561697e-02 2.75589108e-01 5.47724783e-01 3.63134474e-01
-3.04967225e-01 -5.88055849e-01 -3.05859655e-01 6.52485788e-01
-7.52261043e-01 -2.12439612e-01 -1.63253769e-01 7.64580131e-01
1.47886887e-01 1.18869707e-01 -4.05823104e-02 2.05572873e-01
-2.33693168e-01 3.08061510e-01 3.03407133e-01 -5.55922210e-01
-1.25891256e+00 2.74852663e-01 -8.72284323e-02 -3.64364624e-01
-2.46975183e-01 -4.90728527e-01 -1.28695476e+00 4.20676947e-01
-2.44551659e-01 2.73061574e-01 8.08437705e-01 9.24923241e-01
-2.76554763e-01 7.05863357e-01 5.61441302e-01 -1.01833272e+00
-5.60195446e-01 -1.24414372e+00 -1.46832690e-03 2.59976834e-01
2.69955933e-01 -2.43595079e-01 -1.03756987e-01 3.15751731e-02]
|
[14.498236656188965, -2.641587495803833]
|
1d3d750b-11bb-46dd-a602-f6525f567116
|
distinguish-before-answer-generating
|
2305.08135
| null |
https://arxiv.org/abs/2305.08135v2
|
https://arxiv.org/pdf/2305.08135v2.pdf
|
Distinguish Before Answer: Generating Contrastive Explanation as Knowledge for Commonsense Question Answering
|
Existing knowledge-enhanced methods have achieved remarkable results in certain QA tasks via obtaining diverse knowledge from different knowledge bases. However, limited by the properties of retrieved knowledge, they still have trouble benefiting from both the knowledge relevance and distinguishment simultaneously. To address the challenge, we propose CPACE, a Concept-centric Prompt-bAsed Contrastive Explanation Generation model, which aims to convert obtained symbolic knowledge into a contrastive explanation for better distinguishing the differences among given candidates. Firstly, following previous works, we retrieve different types of symbolic knowledge with a concept-centric knowledge extraction module. After that, we generate corresponding contrastive explanations using acquired symbolic knowledge and explanation prompts as guidance for better modeling the knowledge distinguishment and interpretability. Finally, we regard the generated contrastive explanation as external knowledge for downstream task enhancement. We conduct a series of experiments on three widely-used question-answering datasets: CSQA, QASC, and OBQA. Experimental results demonstrate that with the help of generated contrastive explanation, our CPACE model achieves new SOTA on CSQA (89.8% on the testing set, 0.9% higher than human performance), and gains impressive improvement on QASC and OBQA (4.2% and 3.5%, respectively).
|
['Yin Zhang', 'Luo Si', 'Fei Huang', 'Ji Zhang', 'Ming Yan', 'Guohai Xu', 'Qianglong Chen']
|
2023-05-14
| null | null | null | null |
['explanation-generation']
|
['natural-language-processing']
|
[ 3.04939896e-01 4.95614141e-01 -1.91313043e-01 -4.36344564e-01
-1.04355276e+00 -6.39791727e-01 7.22555995e-01 1.51846185e-01
-1.94219261e-01 8.67927253e-01 2.63164401e-01 -4.04721171e-01
-4.96913999e-01 -8.92243981e-01 -7.29931235e-01 -3.05851609e-01
4.23528254e-01 7.49782741e-01 4.99709785e-01 -6.08435750e-01
2.93655008e-01 7.15045333e-02 -1.68097985e+00 7.86330104e-01
1.79595947e+00 1.02596378e+00 2.96350196e-02 2.74104983e-01
-4.15039629e-01 9.54894364e-01 -7.23156393e-01 -7.95604706e-01
-2.16054082e-01 -6.18201792e-01 -1.36581731e+00 -1.28105551e-01
1.64209619e-01 -1.84621438e-01 5.46266176e-02 8.08885872e-01
1.76865071e-01 2.36056209e-01 5.27246177e-01 -1.22896469e+00
-1.05188596e+00 9.21096385e-01 1.85688157e-02 2.01850474e-01
6.94258749e-01 2.10474566e-01 1.24853587e+00 -9.78155017e-01
4.56799001e-01 1.41856515e+00 -3.51225957e-02 5.97212732e-01
-7.80769229e-01 -5.13757765e-01 1.93632290e-01 9.07283485e-01
-1.07809806e+00 -2.49945626e-01 6.42376602e-01 7.61039834e-03
9.80144978e-01 5.39106369e-01 3.59143585e-01 7.37777710e-01
-3.62879336e-01 1.12005496e+00 1.11927783e+00 -5.55577338e-01
4.34527025e-02 2.90568173e-01 5.26772857e-01 6.77903652e-01
-1.90754216e-02 4.30248566e-02 -6.59845769e-01 -1.09776527e-01
4.62957859e-01 -2.09941223e-01 -6.77059472e-01 8.50786343e-02
-1.13607836e+00 7.63398170e-01 6.42688215e-01 1.16553955e-01
-5.30589759e-01 -2.58671641e-01 -2.23476510e-03 3.58707160e-01
-1.21685416e-01 8.93975198e-01 -7.96537757e-01 -6.82284012e-02
-2.51722902e-01 3.91540587e-01 6.57501817e-01 1.13322508e+00
7.08403170e-01 -2.62524635e-01 -7.37142324e-01 7.16779411e-01
1.70126215e-01 7.90854871e-01 5.63235343e-01 -1.06812584e+00
6.73067749e-01 1.14694858e+00 2.87389517e-01 -9.26990390e-01
-1.52611583e-01 -6.03851795e-01 -5.31348825e-01 -5.14761984e-01
2.64491558e-01 4.57218774e-02 -8.97676349e-01 1.64451063e+00
3.20989043e-01 8.02759640e-03 5.20291984e-01 1.03619456e+00
1.19818282e+00 5.86809695e-01 2.86152691e-01 -1.45605057e-01
1.68424010e+00 -1.34285378e+00 -8.15134883e-01 -3.27184141e-01
5.70406556e-01 -3.84464055e-01 1.37420630e+00 4.72645044e-01
-1.03925633e+00 -4.54869211e-01 -7.36999512e-01 -2.02055931e-01
-2.41513923e-01 2.31158152e-01 7.05392838e-01 1.62508160e-01
-5.39497793e-01 1.78206563e-01 -3.05748522e-01 2.03577816e-01
3.33651811e-01 2.39921987e-01 -1.05917484e-01 -4.25382882e-01
-1.77079403e+00 8.97747815e-01 7.16163993e-01 -7.25781098e-02
-6.38449609e-01 -8.82903218e-01 -8.41788650e-01 5.55580974e-01
9.71520245e-01 -1.03660476e+00 1.47230875e+00 -9.41749036e-01
-1.37795210e+00 3.02171201e-01 -3.55753422e-01 -2.92799890e-01
1.91350952e-01 -4.96118665e-01 -6.51218235e-01 3.33288491e-01
3.77989143e-01 6.53557241e-01 4.54159081e-01 -1.28387594e+00
-8.54698241e-01 -2.77344972e-01 6.11733556e-01 4.25718158e-01
-1.60089016e-01 -3.00332546e-01 -7.39148319e-01 -2.68066108e-01
2.88601756e-01 -5.93757749e-01 -5.63756889e-03 -5.63808084e-01
-3.15971106e-01 -5.94259322e-01 3.62406373e-01 -6.81518853e-01
1.16035485e+00 -1.81718373e+00 2.81308264e-01 1.48077577e-01
2.67748803e-01 4.93367493e-01 -2.01912358e-01 1.50488302e-01
1.01406001e-01 8.32760260e-02 -3.85422349e-01 2.06201673e-01
5.32103740e-02 3.73177499e-01 -6.41761780e-01 -6.80403531e-01
7.04232156e-01 1.29814887e+00 -1.29251838e+00 -5.77419817e-01
-1.47174880e-01 -4.80788723e-02 -6.13661945e-01 3.38736326e-01
-6.21997654e-01 4.46725577e-01 -8.12426746e-01 7.50064194e-01
5.46999335e-01 -3.97681087e-01 1.45049483e-01 -2.45166510e-01
5.09509921e-01 5.13823092e-01 -9.03191805e-01 1.49057627e+00
-3.78493905e-01 1.71949849e-01 -6.60718203e-01 -8.11215460e-01
7.75987327e-01 1.97851658e-01 -2.20371857e-01 -1.17914510e+00
1.57889854e-02 3.19829285e-01 9.55303013e-02 -7.13475764e-01
5.82907796e-01 -2.16957718e-01 2.31340230e-02 2.49832034e-01
2.03274880e-02 -6.86989054e-02 2.04324678e-01 6.20077968e-01
1.12073386e+00 -5.21327779e-02 2.14979813e-01 -1.40655311e-02
8.89183044e-01 5.69343388e-01 4.38603461e-01 7.13758647e-01
3.37808067e-03 3.55287284e-01 4.87299919e-01 -2.20480531e-01
-2.65440315e-01 -9.34093535e-01 1.48216218e-01 1.00495350e+00
4.85551625e-01 -5.15062571e-01 -5.34827650e-01 -1.13115609e+00
-7.32113346e-02 1.15966773e+00 -4.46423441e-01 -5.09919167e-01
-5.72781026e-01 -4.07049119e-01 6.60278082e-01 6.87889636e-01
8.93605292e-01 -1.29220474e+00 -4.10702556e-01 1.52930111e-01
-8.79077852e-01 -1.12177432e+00 -3.46343294e-02 -1.66629210e-01
-8.80576611e-01 -1.45138741e+00 -2.38546982e-01 -6.40376985e-01
6.86573505e-01 5.12020886e-01 1.45690012e+00 5.85845172e-01
3.53090495e-01 3.40070814e-01 -7.55082250e-01 -4.00363714e-01
-2.66707540e-01 -3.38394344e-02 -3.18225533e-01 -2.38102019e-01
4.31149215e-01 -1.95340633e-01 -4.39391136e-01 4.03047740e-01
-1.05416119e+00 2.54051238e-01 7.50091195e-01 1.00770974e+00
5.61206818e-01 -5.96314156e-03 9.24109638e-01 -9.76436496e-01
8.17489326e-01 -6.27072692e-01 -3.74886602e-01 7.31206477e-01
-8.40914130e-01 4.84610885e-01 7.16705680e-01 -2.44661435e-01
-1.61611879e+00 -5.00527203e-01 -1.93857271e-02 -1.22771531e-01
-1.04866199e-01 8.65205348e-01 -4.30568635e-01 4.13690209e-01
7.81949520e-01 4.47661906e-01 -2.34814063e-01 -1.84166580e-01
7.53332317e-01 6.31834507e-01 6.49573624e-01 -9.10622358e-01
7.33420610e-01 2.82250315e-01 -3.94556701e-01 -5.22973165e-02
-1.29459929e+00 -3.50373596e-01 -1.73343375e-01 1.17503993e-01
5.58582902e-01 -7.00350463e-01 -8.64986062e-01 -4.68399376e-02
-1.32702184e+00 6.32183105e-02 -2.43529856e-01 2.94419646e-01
-3.37233365e-01 4.67989713e-01 -3.30931664e-01 -6.02217793e-01
-4.85674948e-01 -1.00428092e+00 9.12035584e-01 4.70306098e-01
4.96545713e-03 -7.09197044e-01 -2.26703510e-01 9.49402928e-01
2.77940482e-01 -1.47641301e-01 1.38715279e+00 -9.63950753e-01
-8.38145196e-01 6.41226992e-02 -3.32164168e-01 2.91446060e-01
9.86397192e-02 -3.44113559e-01 -8.49795878e-01 3.40135127e-01
-6.99260831e-02 -4.60039586e-01 7.84467995e-01 -1.77015156e-01
1.26406372e+00 -4.07710314e-01 -1.77969187e-01 2.37711892e-03
9.70112681e-01 3.39747012e-01 7.10969627e-01 4.76611406e-01
3.21134657e-01 6.85814619e-01 1.20008016e+00 6.46060631e-02
7.34336138e-01 5.85746646e-01 4.81104791e-01 2.38511622e-01
-1.26696646e-01 -2.91126221e-01 1.24242745e-01 7.97595799e-01
-2.06038803e-01 -1.79240316e-01 -9.21151400e-01 8.97164106e-01
-2.05993080e+00 -8.51690590e-01 -2.95600295e-01 1.79221952e+00
1.12602913e+00 3.29848267e-02 -3.17137271e-01 5.79419211e-02
4.46173698e-01 -4.28410113e-01 -5.77206075e-01 -1.25406325e-01
-2.56070167e-01 1.94814578e-01 -2.20057666e-01 4.73668575e-01
-5.03280580e-01 1.18058860e+00 5.11876678e+00 1.03779793e+00
-5.60708642e-01 -1.00180820e-01 2.30926320e-01 2.23885268e-01
-8.39141071e-01 1.68352395e-01 -6.43757701e-01 2.37198412e-01
6.38020933e-01 -3.01947474e-01 2.88322955e-01 7.89585054e-01
-1.92566127e-01 -8.97216797e-02 -9.65503216e-01 6.71537161e-01
3.78822982e-02 -1.32931137e+00 7.82195091e-01 -4.08559382e-01
5.82267940e-01 -6.19305730e-01 -2.39769109e-02 8.78401101e-01
1.96306407e-01 -9.96681809e-01 5.24267673e-01 4.40965384e-01
4.00683045e-01 -7.92141736e-01 1.13028526e+00 5.94642401e-01
-9.48212683e-01 -1.88386306e-01 -2.73654610e-01 -1.09917112e-01
1.64223164e-01 3.52475971e-01 -9.89882171e-01 1.36839414e+00
5.32961369e-01 3.90392363e-01 -6.31169975e-01 6.83058918e-01
-9.65620875e-01 7.53932953e-01 1.20919891e-01 -1.78276040e-02
2.47322753e-01 8.20130482e-02 2.46809259e-01 8.83466065e-01
9.84714106e-02 9.26874220e-01 -2.21582323e-01 1.08029056e+00
-1.18507266e-01 -3.37719880e-02 -1.22112306e-02 1.17726669e-01
7.75638998e-01 1.12808549e+00 -9.33988988e-02 -8.48788738e-01
-1.84280291e-01 8.86102915e-01 5.13293266e-01 3.66586208e-01
-1.07192183e+00 -5.23575723e-01 3.15575570e-01 -3.29706669e-01
1.41526356e-01 3.09401751e-01 -1.13586091e-01 -1.36156833e+00
2.30316803e-01 -1.30029559e+00 7.12874234e-01 -1.15503013e+00
-1.24352837e+00 8.52755845e-01 2.38059610e-01 -9.60662186e-01
-3.26760948e-01 -5.15950263e-01 -4.16630894e-01 9.00317192e-01
-1.97112679e+00 -9.90913332e-01 -6.26585603e-01 6.76640451e-01
6.25417531e-01 -9.33542997e-02 6.76213861e-01 1.02073997e-01
-2.50802934e-01 5.94128907e-01 -5.16135812e-01 -5.74676134e-02
6.73532367e-01 -1.34534919e+00 6.08492903e-02 6.80077255e-01
1.94373757e-01 9.34772551e-01 4.73764777e-01 -5.43763340e-01
-1.41415572e+00 -1.02830756e+00 1.14031696e+00 -8.30532789e-01
2.55290210e-01 4.91319358e-01 -1.35101104e+00 6.63677633e-01
1.81012362e-01 -4.22428668e-01 7.15158701e-01 2.99708307e-01
-4.71102506e-01 1.20668158e-01 -1.07253158e+00 6.55943692e-01
1.13723481e+00 -3.40449393e-01 -1.45286667e+00 2.21450210e-01
1.19870484e+00 -4.87939656e-01 -7.35908389e-01 7.32035100e-01
2.38066211e-01 -8.17251444e-01 1.00794530e+00 -1.12371111e+00
6.92990482e-01 -5.46731174e-01 -3.53584215e-02 -1.28676057e+00
-3.63202423e-01 -1.03109345e-01 -3.04055274e-01 1.17080152e+00
7.39540994e-01 -4.98128057e-01 4.23168689e-01 8.02746058e-01
-4.42008764e-01 -7.05873132e-01 -6.62380993e-01 -7.90419579e-01
-2.57002890e-01 -3.76979649e-01 1.10149324e+00 1.08188403e+00
2.42827073e-01 7.82773674e-01 7.44448602e-02 5.36470056e-01
2.10054353e-01 5.87205470e-01 6.71432912e-01 -9.99787450e-01
-3.04573983e-01 -2.10817605e-01 1.36233672e-01 -1.44065785e+00
1.45113871e-01 -8.82824659e-01 9.08646807e-02 -1.71905756e+00
3.48260254e-01 -3.65183473e-01 -2.85707265e-01 8.17084253e-01
-8.87000620e-01 -3.14705044e-01 2.23886371e-01 1.36994377e-01
-9.83230710e-01 8.08432937e-01 1.59512568e+00 -1.59664527e-01
-1.73790157e-01 -2.99055815e-01 -1.13293970e+00 5.70235252e-01
6.20490432e-01 -2.85895139e-01 -7.89873958e-01 -7.70327330e-01
2.68350542e-01 2.11833090e-01 5.88067353e-01 -5.85118473e-01
3.03186059e-01 -4.52287763e-01 -1.53885245e-01 -4.28957105e-01
2.35632524e-01 -6.24477386e-01 -2.37807304e-01 4.08112824e-01
-4.09270018e-01 -1.87034439e-02 3.20500553e-01 5.63683569e-01
-6.12055719e-01 -6.72333688e-02 1.35727599e-01 -6.68806136e-02
-1.07710230e+00 -1.09571546e-01 4.83466350e-02 3.69234115e-01
6.27581716e-01 1.02240913e-01 -9.43001449e-01 -3.59927803e-01
-5.67526758e-01 7.07138777e-01 -1.70052052e-01 4.37344193e-01
9.55154240e-01 -1.35205221e+00 -7.47695506e-01 -1.31985766e-03
4.53431398e-01 2.43403703e-01 3.69160056e-01 6.98322415e-01
-1.09362230e-02 8.41060519e-01 -1.56264797e-01 -4.59002703e-01
-1.06089568e+00 5.29164076e-01 1.79476693e-01 -4.52751905e-01
-1.89006001e-01 8.13660622e-01 1.76528141e-01 -6.56410635e-01
-8.58779326e-02 -4.63818580e-01 -5.53546429e-01 -2.61442810e-01
5.51653206e-01 2.87262559e-01 8.26948136e-02 -9.80376527e-02
-4.26529050e-01 2.26701409e-01 -8.45484063e-02 -5.53116761e-02
9.51645672e-01 3.25407088e-02 -3.78642827e-02 -2.16690183e-01
3.80609810e-01 1.34092823e-01 -7.81478941e-01 -5.92521548e-01
2.26963669e-01 -4.66516346e-01 -3.64697307e-01 -1.55652630e+00
-8.75131309e-01 7.88197637e-01 -3.78048234e-02 2.16648296e-01
1.42027652e+00 4.11391884e-01 8.73492479e-01 7.22614646e-01
3.23965877e-01 -5.79845250e-01 2.44400874e-01 6.12059593e-01
1.22214162e+00 -1.33351600e+00 -2.64601618e-01 -8.81254673e-01
-9.99512196e-01 8.35959673e-01 1.02926362e+00 3.96828562e-01
-1.07388526e-01 -5.33453166e-01 3.13864425e-02 -5.25209069e-01
-1.00743711e+00 -4.30863768e-01 6.53137445e-01 4.26435113e-01
1.74274728e-01 9.39101130e-02 -4.37936246e-01 1.41777360e+00
-2.33852044e-01 -8.30025673e-02 2.71273047e-01 7.91353524e-01
-6.83648944e-01 -1.10861039e+00 -2.56715953e-01 2.42049173e-01
1.43355140e-02 -4.13969517e-01 -6.02832079e-01 8.54304492e-01
8.54297653e-02 1.29993033e+00 -4.38898146e-01 -3.55477989e-01
7.15345204e-01 1.70541912e-01 2.86211282e-01 -6.46629930e-01
-6.36903107e-01 -4.66153473e-01 4.26622361e-01 -4.87271577e-01
-4.20087129e-01 -5.28404675e-02 -1.78155565e+00 -1.42892018e-01
-6.51489079e-01 7.26808310e-01 1.99401289e-01 1.42889702e+00
7.05191433e-01 7.40137637e-01 2.21759140e-01 2.04587072e-01
-7.46528625e-01 -8.95249844e-01 -6.55470267e-02 7.44854569e-01
1.23525122e-02 -8.16412568e-01 -2.85784513e-01 -9.10675211e-04]
|
[10.7037353515625, 7.969479560852051]
|
aa7ad4f1-d8ec-4c88-a5d5-daf667e725d3
|
are-neural-architecture-search-benchmarks
|
2303.16938
| null |
https://arxiv.org/abs/2303.16938v1
|
https://arxiv.org/pdf/2303.16938v1.pdf
|
Are Neural Architecture Search Benchmarks Well Designed? A Deeper Look Into Operation Importance
|
Neural Architecture Search (NAS) benchmarks significantly improved the capability of developing and comparing NAS methods while at the same time drastically reduced the computational overhead by providing meta-information about thousands of trained neural networks. However, tabular benchmarks have several drawbacks that can hinder fair comparisons and provide unreliable results. These usually focus on providing a small pool of operations in heavily constrained search spaces -- usually cell-based neural networks with pre-defined outer-skeletons. In this work, we conducted an empirical analysis of the widely used NAS-Bench-101, NAS-Bench-201 and TransNAS-Bench-101 benchmarks in terms of their generability and how different operations influence the performance of the generated architectures. We found that only a subset of the operation pool is required to generate architectures close to the upper-bound of the performance range. Also, the performance distribution is negatively skewed, having a higher density of architectures in the upper-bound range. We consistently found convolution layers to have the highest impact on the architecture's performance, and that specific combination of operations favors top-scoring architectures. These findings shed insights on the correct evaluation and comparison of NAS methods using NAS benchmarks, showing that directly searching on NAS-Bench-201, ImageNet16-120 and TransNAS-Bench-101 produces more reliable results than searching only on CIFAR-10. Furthermore, with this work we provide suggestions for future benchmark evaluations and design. The code used to conduct the evaluations is available at https://github.com/VascoLopes/NAS-Benchmark-Evaluation.
|
['Luís A. Alexandre', 'Bruno Degardin', 'Vasco Lopes']
|
2023-03-29
| null | null | null | null |
['architecture-search']
|
['methodology']
|
[-4.01582032e-01 -4.50537473e-01 -1.59688979e-01 -3.63044411e-01
-8.37021530e-01 -7.48362601e-01 4.15220708e-01 -2.42027845e-02
-7.63858378e-01 5.74868083e-01 6.08087070e-02 -6.49849296e-01
-3.42115074e-01 -7.39729881e-01 -7.66752839e-01 -5.23127496e-01
-1.41319573e-01 4.98026580e-01 7.69461840e-02 -2.52989799e-01
2.26454824e-01 6.32365525e-01 -1.60687006e+00 5.23066938e-01
4.97832179e-01 1.12524271e+00 1.67301804e-01 5.60743451e-01
-3.07350550e-02 4.41374719e-01 -8.11635733e-01 -4.49561983e-01
6.29254401e-01 -1.68920651e-01 -5.03222704e-01 -7.50373960e-01
9.02030945e-01 -1.97274029e-01 -2.96022475e-01 8.78063560e-01
6.66436553e-01 1.70619771e-01 4.67602313e-01 -1.15250599e+00
-3.55755121e-01 7.06446826e-01 -2.30783343e-01 7.30330169e-01
-1.40271127e-01 4.67277855e-01 1.24978971e+00 -8.49882960e-01
6.04225993e-01 1.04472542e+00 7.85150409e-01 5.34260929e-01
-1.27907455e+00 -8.91966641e-01 1.69466734e-01 1.99947618e-02
-1.61731100e+00 -6.95243955e-01 2.38412112e-01 -2.53904462e-01
1.41944706e+00 5.32346904e-01 4.69247252e-01 1.03840971e+00
-3.53395678e-02 2.49957308e-01 8.18286657e-01 -2.34423533e-01
4.30978507e-01 2.68723182e-02 3.00268620e-01 6.24085426e-01
6.72511458e-01 1.54939562e-01 -4.26735878e-01 -3.88219267e-01
6.48540556e-01 -2.91634262e-01 -2.85515457e-01 -1.52963951e-01
-1.22661698e+00 6.69186175e-01 6.67104185e-01 5.88903606e-01
-3.95040065e-01 3.59574527e-01 6.48982882e-01 3.83815676e-01
6.75047413e-02 1.14202368e+00 -6.33680344e-01 -1.37385458e-01
-1.04321373e+00 4.26302075e-01 8.33884716e-01 7.16535568e-01
3.92733186e-01 2.45214000e-01 -3.40460956e-01 7.82666624e-01
-2.02711225e-01 1.93461940e-01 5.22673011e-01 -9.82144892e-01
6.82575762e-01 5.08123040e-01 -3.37122269e-02 -8.22864592e-01
-5.09058058e-01 -1.07891345e+00 -5.70167184e-01 2.83034712e-01
7.53960013e-01 -2.91958719e-01 -8.36858988e-01 1.88605177e+00
-1.72446728e-01 -2.34462321e-02 -1.04597457e-01 1.00337195e+00
7.41030991e-01 5.85555911e-01 4.98106927e-02 3.48988652e-01
1.36184895e+00 -1.07293749e+00 3.01576592e-02 -2.48038426e-01
7.30517685e-01 -6.55319035e-01 1.32981563e+00 3.29000592e-01
-1.26246107e+00 -4.85086501e-01 -1.14955950e+00 1.72292709e-01
-5.67344069e-01 2.31319278e-01 6.39574170e-01 6.61571562e-01
-1.33714271e+00 5.12122214e-01 -7.66620636e-01 -9.14807618e-02
4.14603621e-01 4.50716823e-01 -9.86182019e-02 1.74635515e-01
-9.76793408e-01 9.84611988e-01 5.34471214e-01 -7.72084594e-02
-9.04425323e-01 -9.89617467e-01 -4.21807647e-01 5.44349492e-01
1.44488752e-01 -6.60430253e-01 1.27381122e+00 -9.33743238e-01
-9.71510768e-01 6.53843045e-01 1.17332749e-01 -8.86984706e-01
3.17396462e-01 2.27383971e-02 -3.41347665e-01 -1.88179299e-01
-1.95431367e-01 9.12827671e-01 2.16370478e-01 -9.01381016e-01
-5.72094023e-01 -2.22476169e-01 2.78754473e-01 1.27001926e-01
-4.82144028e-01 -4.46737409e-02 -6.53902292e-01 -6.87592983e-01
-9.97496247e-02 -1.00753081e+00 -1.75606996e-01 -3.40226948e-01
-2.83404589e-01 -1.38924662e-02 1.17404722e-01 -2.47090355e-01
1.47270727e+00 -2.17222142e+00 -2.04657972e-01 3.75256807e-01
1.51159063e-01 3.41277838e-01 -5.96400738e-01 3.24245006e-01
-2.14359358e-01 3.85897249e-01 7.78325349e-02 -1.11665182e-01
-3.14922221e-02 -4.12737615e-02 -2.29930520e-01 2.38237292e-01
-8.65788311e-02 6.59609735e-01 -4.96488243e-01 -8.81339237e-02
-1.19921319e-01 3.97080719e-01 -8.06938171e-01 -2.61653602e-01
-2.53224850e-01 -1.66826114e-01 -1.86820745e-01 6.67667806e-01
3.78604740e-01 -4.06279415e-01 -4.52980585e-02 -2.78065592e-01
-1.97850496e-01 6.68154061e-01 -9.95210469e-01 1.50571203e+00
-7.13595450e-01 7.33498752e-01 9.38418414e-03 -6.23204410e-01
6.57720327e-01 8.09600949e-02 7.29781240e-02 -9.51653183e-01
1.29617184e-01 5.72292566e-01 6.15105152e-01 5.40061481e-02
3.67639273e-01 4.54256177e-01 2.24923864e-01 3.85548800e-01
-3.17918807e-02 2.34945580e-01 5.08866251e-01 -2.46540103e-02
1.31967437e+00 -6.23852193e-01 -8.74806568e-02 -5.97578585e-01
2.67907739e-01 1.70606524e-01 4.32929367e-01 9.69324410e-01
1.41098592e-02 6.40421927e-01 5.57172179e-01 -5.44685483e-01
-1.17696440e+00 -1.07967925e+00 -3.32365036e-01 1.28248668e+00
-2.10347146e-01 -4.59985524e-01 -8.67492855e-01 -4.44553852e-01
-6.12152442e-02 1.05679297e+00 -6.52120352e-01 -1.61196768e-01
-5.25911152e-01 -7.66088426e-01 1.11110961e+00 5.50526500e-01
4.90868479e-01 -9.99061108e-01 -8.98248792e-01 -1.83430091e-01
3.48369926e-02 -8.44763339e-01 -5.17759204e-01 3.85417998e-01
-1.01375651e+00 -1.02352870e+00 -6.86917365e-01 -7.34586716e-01
6.96464360e-01 1.31404459e-01 1.47007561e+00 4.56903905e-01
-2.48347372e-01 -4.18183357e-02 2.80358344e-02 -3.51871997e-01
-1.87318385e-01 6.44139469e-01 -1.68746814e-01 -4.30031925e-01
2.30569020e-01 -4.11086082e-01 -7.19772935e-01 4.30433929e-01
-6.69556260e-01 -7.23423138e-02 7.96934783e-01 8.65335584e-01
4.57900673e-01 -4.10561264e-02 4.51713413e-01 -6.52207375e-01
7.65641332e-01 -5.40477157e-01 -8.27320457e-01 2.34154999e-01
-9.41625774e-01 2.95678109e-01 6.56800032e-01 -4.21525925e-01
-6.54591799e-01 -2.20355600e-01 -8.91533494e-02 -5.64121544e-01
-8.12840015e-02 6.43870950e-01 2.41695479e-01 1.89183112e-02
1.20372021e+00 7.93629289e-02 -6.68711811e-02 -5.35937250e-01
-1.89687267e-01 2.24478945e-01 3.84531647e-01 -7.30556011e-01
4.62298870e-01 3.03914756e-01 -2.37170830e-01 -5.09314120e-01
-4.52109993e-01 -3.12986165e-01 1.10730594e-02 2.44601414e-01
5.44878304e-01 -8.68532777e-01 -4.92038786e-01 1.76758125e-01
-9.21486676e-01 -6.43229306e-01 -2.10528802e-02 5.80278695e-01
-9.27829146e-02 -2.87795156e-01 -6.51857436e-01 -4.31939036e-01
-6.14239156e-01 -1.54601669e+00 7.00258076e-01 2.61578560e-01
-3.94493550e-01 -7.92813480e-01 -5.86486086e-02 2.31332749e-01
8.99331570e-01 -1.19194694e-01 1.11053085e+00 -9.77052152e-01
-6.04470968e-01 -1.26029491e-01 -3.41290355e-01 2.61360556e-01
-3.28624815e-01 -7.51736909e-02 -9.66557384e-01 -3.05149138e-01
-4.23965454e-01 -1.75948277e-01 1.04143691e+00 4.85782862e-01
1.39292896e+00 -2.56483197e-01 -2.13913396e-01 8.23322713e-01
1.52591133e+00 3.31314266e-01 6.42521083e-01 6.88626647e-01
1.66073218e-01 3.09948295e-01 1.99227929e-01 6.40197247e-02
-1.55965015e-02 7.44981229e-01 5.66015780e-01 -1.26861818e-02
-8.39444101e-02 1.22956865e-01 3.04267973e-01 4.04446810e-01
-7.12780952e-02 -2.50737727e-01 -1.25454640e+00 6.08625531e-01
-1.36259067e+00 -9.02535617e-01 1.74094379e-01 2.46747375e+00
8.88289452e-01 5.63513160e-01 1.57266110e-01 -1.37608394e-01
5.14452696e-01 7.43927434e-02 -4.81284618e-01 -5.69164157e-01
-6.95902780e-02 3.68983060e-01 7.21791327e-01 3.56307507e-01
-7.77986169e-01 6.56354547e-01 6.40079451e+00 9.44981456e-01
-1.35514617e+00 6.81521520e-02 9.14355576e-01 -7.18520999e-01
-1.88961506e-01 -2.32228771e-01 -9.83552933e-01 5.10348082e-01
1.40624321e+00 -1.05074741e-01 6.32586181e-01 1.02546930e+00
-8.97254944e-02 1.22372009e-01 -1.18627250e+00 9.53693449e-01
-2.21162379e-01 -1.81431115e+00 2.49229688e-02 1.66293666e-01
6.96167171e-01 7.32918084e-01 1.93454176e-01 5.13409376e-01
-8.15203972e-03 -1.33342159e+00 9.30633247e-01 1.23701990e-01
7.43016481e-01 -6.81931794e-01 7.72953272e-01 6.42545372e-02
-9.49090123e-01 -2.37352595e-01 -3.90520692e-01 6.66764081e-02
-3.56929064e-01 5.26399553e-01 -8.35579395e-01 1.83508545e-02
9.01963174e-01 -1.49085643e-02 -8.09644461e-01 1.31891406e+00
3.61043334e-01 7.20076621e-01 -4.18601900e-01 -2.36404344e-01
4.92373794e-01 1.16651341e-01 3.75089437e-01 1.39496636e+00
4.10998076e-01 -2.81943917e-01 -2.09865227e-01 1.00298548e+00
-2.51053512e-01 -6.40049800e-02 -4.74187672e-01 -1.00657798e-01
9.00265872e-01 1.08396149e+00 -7.93317080e-01 -3.20266366e-01
-3.38490516e-01 1.96928114e-01 2.85867721e-01 3.82348061e-01
-1.09616959e+00 -3.80898356e-01 9.13110495e-01 2.79270828e-01
9.51294228e-02 -1.44535691e-01 -7.48802960e-01 -8.36602151e-01
-2.11577630e-03 -1.32944667e+00 4.89571065e-01 -6.59204125e-01
-9.39130366e-01 9.74174798e-01 6.79996312e-02 -7.93175995e-01
-1.15180463e-01 -6.89924300e-01 -6.61594629e-01 9.02497709e-01
-1.06691074e+00 -4.68740731e-01 -4.11991417e-01 3.05415362e-01
6.16452575e-01 -4.92352068e-01 8.72648239e-01 6.18469596e-01
-6.86537743e-01 1.11132336e+00 1.38785690e-01 2.13758200e-01
3.78872752e-01 -9.42853987e-01 6.59332633e-01 5.33258736e-01
4.02083427e-01 1.05992401e+00 6.79248214e-01 -3.11275810e-01
-1.15566945e+00 -8.07549417e-01 5.29860079e-01 -2.19542563e-01
4.91620898e-01 -2.55602956e-01 -8.46748471e-01 5.27478874e-01
1.32553563e-01 -8.55062380e-02 3.36429119e-01 3.84901613e-01
-5.90574503e-01 -3.06684613e-01 -8.35905433e-01 7.92575061e-01
1.13101220e+00 -3.70221436e-01 -1.85364634e-01 2.04566374e-01
5.59206307e-01 -5.42580724e-01 -7.07351804e-01 3.11114818e-01
6.71211302e-01 -1.19403958e+00 1.13668704e+00 -5.19180954e-01
3.88387114e-01 4.84919064e-02 -3.30573559e-01 -1.26934528e+00
-2.58058727e-01 -4.89993319e-02 1.26778543e-01 8.77547085e-01
1.15437376e+00 -9.69717860e-01 9.21024561e-01 4.26314592e-01
-3.16355288e-01 -1.25922024e+00 -7.47486532e-01 -9.36398447e-01
2.04319149e-01 -3.18296492e-01 7.84869432e-01 7.61354268e-01
-5.79768836e-01 -9.91808176e-02 3.87968510e-01 -4.10831384e-02
2.65538990e-01 -1.24350913e-01 5.36405742e-01 -9.39242661e-01
-5.53328633e-01 -1.10462427e+00 -2.04792798e-01 -5.70067286e-01
-5.52511923e-02 -8.66570890e-01 -2.10828692e-01 -1.11412346e+00
-1.52015202e-02 -9.08924520e-01 -6.44832730e-01 6.45138681e-01
2.60005474e-01 2.91424274e-01 2.46287569e-01 3.87728006e-01
-3.04138035e-01 -5.51924389e-03 7.73016334e-01 -3.48750986e-02
-7.12191761e-02 -1.00060672e-01 -7.38876581e-01 5.33321142e-01
1.20862937e+00 -3.95179242e-01 -6.28147364e-01 -8.87213290e-01
3.87598783e-01 -3.05764705e-01 3.96058232e-01 -1.39579546e+00
2.29966268e-01 -2.51498483e-02 2.87743151e-01 -2.99740463e-01
3.83953661e-01 -5.15941501e-01 2.77589798e-01 6.75405502e-01
-7.00542986e-01 7.57389724e-01 5.88885903e-01 5.29254191e-02
-1.22713089e-01 -3.61753136e-01 7.64506161e-01 -1.70275435e-01
-6.41085088e-01 -2.87686214e-02 -7.45306015e-02 3.05490047e-01
7.97349393e-01 -1.12771720e-01 -6.54186010e-01 -1.36291027e-01
-3.51371497e-01 6.56053573e-02 5.00365555e-01 3.60233486e-01
2.73385793e-01 -1.00381911e+00 -5.93412459e-01 2.49525100e-01
8.50228667e-02 -1.51805669e-01 -6.17255941e-02 8.83981824e-01
-1.05462265e+00 8.68716061e-01 -3.66487533e-01 -4.12930310e-01
-1.07020473e+00 1.34074107e-01 7.67094910e-01 -3.03580076e-01
-2.48641774e-01 1.06504714e+00 1.59843385e-01 -3.92613322e-01
6.60432637e-01 -4.93255913e-01 1.68759301e-01 -3.78373601e-02
3.97290707e-01 4.93290573e-01 6.07591331e-01 -2.76268005e-01
-5.47481179e-01 -1.35517996e-02 -1.57836139e-01 -1.96132272e-01
1.21542335e+00 5.28654218e-01 -2.72247791e-02 1.21129021e-01
1.13757634e+00 -4.47939448e-02 -8.77121925e-01 2.01811627e-01
4.14402634e-02 -2.14750007e-01 3.55114102e-01 -1.10697758e+00
-1.35354424e+00 6.54720724e-01 6.76040769e-01 6.01436570e-02
1.06282830e+00 -2.16428727e-01 5.79103291e-01 7.13888466e-01
4.41099435e-01 -9.99441922e-01 -1.77204803e-01 5.49656630e-01
1.00986493e+00 -8.16116810e-01 -1.08385824e-01 1.58947632e-02
-3.46920371e-01 9.43194628e-01 9.85852957e-01 -1.25063956e-01
4.76595044e-01 4.10937577e-01 3.00624091e-02 -3.14530462e-01
-1.20752633e+00 1.48199335e-01 2.88031280e-01 1.25191405e-01
6.06355131e-01 -3.94086465e-02 -2.60330111e-01 4.41823065e-01
-5.13472974e-01 -2.33251542e-01 2.05776691e-01 7.09692180e-01
-2.65203238e-01 -8.15375507e-01 -3.69413435e-01 7.92428434e-01
-6.59530580e-01 -5.79391599e-01 -2.91153044e-01 1.00156176e+00
-4.94401939e-02 5.35166502e-01 3.09290439e-01 -4.10527289e-01
2.96316296e-01 2.25231752e-01 3.05913597e-01 -4.97747988e-01
-1.00997090e+00 -3.12170386e-01 4.28108543e-01 -6.40520871e-01
4.20682698e-01 -4.05136257e-01 -1.24111378e+00 -5.75078189e-01
-3.06253463e-01 3.10077041e-01 9.35838342e-01 4.22516555e-01
7.66743958e-01 5.89165211e-01 -6.50427639e-02 -7.55893290e-01
-1.03621256e+00 -8.39024782e-01 -1.31306082e-01 1.30429998e-01
3.88536900e-02 -6.27315044e-01 -5.06862760e-01 -5.01110137e-01]
|
[8.521553039550781, 3.3130156993865967]
|
88963ac7-0d2d-4b0b-a639-1f7f7fa23d7f
|
beyond-lexical-a-semantic-retrieval-framework
|
2008.03917
| null |
https://arxiv.org/abs/2008.03917v1
|
https://arxiv.org/pdf/2008.03917v1.pdf
|
Beyond Lexical: A Semantic Retrieval Framework for Textual SearchEngine
|
Search engine has become a fundamental component in various web and mobile applications. Retrieving relevant documents from the massive datasets is challenging for a search engine system, especially when faced with verbose or tail queries. In this paper, we explore a vector space search framework for document retrieval. Specifically, we trained a deep semantic matching model so that each query and document can be encoded as a low dimensional embedding. Our model was trained based on BERT architecture. We deployed a fast k-nearest-neighbor index service for online serving. Both offline and online metrics demonstrate that our method improved retrieval performance and search quality considerably, particularly for tail
|
['Kuan Fang', 'RiKang Zhour', 'RuiXing Wang', 'Long Zhao', 'Zhan Shen', 'LiWen Fan']
|
2020-08-10
| null | null | null | null |
['semantic-retrieval']
|
['natural-language-processing']
|
[-2.37929776e-01 -6.61346436e-01 -6.14102066e-01 -2.86210001e-01
-1.11778319e+00 -7.35269129e-01 8.22649479e-01 9.30610020e-03
-4.75440383e-01 9.51443091e-02 4.49773431e-01 -2.34714374e-01
-8.41102183e-01 -8.58957350e-01 -2.97531635e-01 -2.09354296e-01
-1.07920105e-02 8.22038114e-01 3.45283121e-01 -4.31144774e-01
4.70226347e-01 4.64733064e-01 -1.47708058e+00 2.29373842e-01
5.06666005e-01 1.61709130e+00 3.82327825e-01 5.72174072e-01
-4.14240986e-01 3.75154495e-01 -2.44347453e-01 -4.11807090e-01
3.06606233e-01 2.16926977e-01 -1.12096763e+00 -6.16570652e-01
2.70736337e-01 -5.70884168e-01 -1.12123990e+00 9.07022417e-01
8.05911005e-01 3.12999129e-01 6.93331659e-01 -1.10032678e+00
-1.18037415e+00 3.47407669e-01 -6.36784583e-02 5.10028064e-01
4.54531610e-01 -3.46551329e-01 1.41957736e+00 -1.19988692e+00
3.84960234e-01 1.07136655e+00 2.53710896e-01 3.22944880e-01
-7.45863974e-01 -5.00487804e-01 -1.76205531e-01 6.97932363e-01
-1.55937159e+00 -4.88875180e-01 6.46101058e-01 -7.61492550e-02
1.15438175e+00 5.19256592e-01 4.21591818e-01 1.11388493e+00
1.30209476e-01 1.18963063e+00 2.02703357e-01 -1.37465537e-01
2.09872305e-01 3.31345685e-02 3.11999470e-01 3.18943352e-01
1.01075105e-01 1.66165724e-01 -4.41382021e-01 -5.05718946e-01
3.86404663e-01 5.55331290e-01 1.51614724e-02 -2.95951039e-01
-8.77707422e-01 1.11547291e+00 7.48626530e-01 3.93254489e-01
-5.01217782e-01 4.33336973e-01 5.44916689e-01 5.06781936e-01
2.13027194e-01 4.09927934e-01 -1.79141685e-01 -1.04003236e-01
-9.70855057e-01 2.91436911e-01 9.58919942e-01 1.13671803e+00
5.17873228e-01 -6.55168116e-01 -6.96958363e-01 1.08157396e+00
4.30012017e-01 7.12869823e-01 9.69610751e-01 -9.78907943e-01
3.48432958e-01 5.56537926e-01 7.82115534e-02 -1.46789825e+00
-1.06008328e-01 -4.88738954e-01 -7.15196431e-01 -7.72073925e-01
-4.40772146e-01 5.37126005e-01 -7.02002645e-01 1.08225739e+00
5.19707911e-02 -8.61784071e-02 -6.08793646e-02 1.15539706e+00
7.60388076e-01 8.28264058e-01 -5.99727519e-02 1.78730011e-01
1.33322406e+00 -1.31752610e+00 -6.19443417e-01 -2.97701061e-01
6.78012788e-01 -8.76890063e-01 1.13033926e+00 -3.15889344e-02
-1.09372199e+00 -2.97667533e-01 -7.20447898e-01 -4.89827335e-01
-5.79732835e-01 -3.30428034e-01 7.13489413e-01 1.53367475e-01
-1.16662097e+00 4.27369148e-01 -4.94700372e-01 -3.70222867e-01
3.06841224e-01 2.23843142e-01 -1.96272999e-01 -4.60427195e-01
-1.48922062e+00 4.96092021e-01 2.09208682e-01 -9.41713601e-02
-5.47257721e-01 -3.53544563e-01 -4.18967456e-01 6.87364995e-01
2.40355209e-01 -7.84375966e-01 1.49748313e+00 -1.13620728e-01
-1.07251632e+00 7.68332303e-01 -2.57220179e-01 -2.08745465e-01
3.46936099e-02 -2.56914020e-01 -7.47209966e-01 2.52415448e-01
2.14471355e-01 2.34414607e-01 7.24375546e-01 -6.82450056e-01
-6.12312138e-01 -6.39522672e-01 -1.06134199e-01 3.08011055e-01
-8.93756866e-01 4.16290849e-01 -1.28165281e+00 -5.78569472e-01
2.16186091e-01 -6.62957311e-01 3.83762196e-02 1.87230662e-01
1.01442402e-02 -8.54112446e-01 8.74055922e-01 -7.38754869e-01
1.70647025e+00 -2.15314174e+00 -2.21198425e-02 5.41747808e-01
3.14453542e-01 3.34629804e-01 -2.86686003e-01 8.33622277e-01
4.58416671e-01 5.69708198e-02 4.95078325e-01 1.19699948e-02
6.82910085e-01 -1.37796868e-02 -7.27583647e-01 1.51595756e-04
-4.31526750e-01 1.54907918e+00 -8.19556177e-01 -6.93488240e-01
-2.26991594e-01 5.57504654e-01 -5.18596351e-01 4.27048773e-01
-1.59307793e-01 -6.05194569e-01 -1.01397014e+00 9.14822936e-01
3.14712167e-01 -8.29610050e-01 -3.73017222e-01 -1.26646608e-01
5.74647307e-01 2.96161532e-01 -6.29313350e-01 2.14670539e+00
-7.60499775e-01 6.07682765e-01 -4.23501618e-02 -7.86534488e-01
8.65250647e-01 2.07124293e-01 3.75176013e-01 -1.64055097e+00
-1.68496042e-01 6.04574859e-01 -7.82626390e-01 -4.49955165e-01
8.77420723e-01 5.91761708e-01 -1.76441804e-01 6.46069467e-01
-3.40275645e-01 2.19417766e-01 -4.70788740e-02 4.40056533e-01
1.27636683e+00 -4.69873697e-01 -2.82721430e-01 -6.70645460e-02
4.08017933e-01 -7.22012967e-02 -2.62793571e-01 1.00801599e+00
-3.98030356e-02 2.69447595e-01 -1.39875591e-01 -5.40455163e-01
-8.22076857e-01 -9.33123469e-01 -1.53308883e-01 1.64466619e+00
4.70715910e-01 -5.15248597e-01 -3.77868533e-01 -5.30833960e-01
4.38334554e-01 2.75466502e-01 -1.12039477e-01 -7.51223147e-01
-4.37577397e-01 -1.59741580e-01 3.04603547e-01 5.96030593e-01
2.86579490e-01 -1.09217942e+00 -2.24255443e-01 3.33804905e-01
-1.79126590e-01 -9.02682006e-01 -1.21669602e+00 1.66846346e-02
-6.51903272e-01 -7.59753764e-01 -9.72420096e-01 -1.18845987e+00
9.79439542e-02 8.12345564e-01 1.11518621e+00 3.74274641e-01
-4.15803283e-01 6.43736601e-01 -5.63770056e-01 1.45681188e-01
2.37836465e-01 4.03398454e-01 -3.68015245e-02 -3.16695750e-01
1.08350754e+00 -1.96825728e-01 -1.47265887e+00 4.96400952e-01
-1.22877061e+00 -7.53955126e-01 4.95614171e-01 8.18563104e-01
5.63950658e-01 -8.23639557e-02 4.65316415e-01 -1.97191328e-01
1.34098113e+00 -8.18208992e-01 -5.58549345e-01 4.98483509e-01
-1.20513296e+00 2.81597584e-01 4.81263518e-01 -1.33467615e-01
-3.53055865e-01 -5.32406747e-01 -8.64566043e-02 -5.27372360e-01
4.26650047e-01 7.08551407e-01 1.86773330e-01 -6.80335164e-02
6.62742913e-01 6.11084044e-01 -2.33524337e-01 -9.14753616e-01
3.91697794e-01 1.54283535e+00 3.73562068e-01 -4.30039376e-01
6.45560503e-01 2.51792699e-01 -3.14640224e-01 -3.92128855e-01
-7.26623178e-01 -1.17823863e+00 -1.12247527e-01 3.36339206e-01
3.55847865e-01 -8.99309397e-01 -7.78880358e-01 -1.69457626e-02
-1.09464788e+00 1.12986304e-01 -3.99396196e-02 3.24456483e-01
-4.62962717e-01 3.70325327e-01 -5.43618381e-01 -4.01092350e-01
-1.04341567e+00 -1.02774751e+00 1.37208092e+00 9.61354747e-02
2.92385146e-02 -1.05401719e+00 3.45307559e-01 4.98453289e-01
1.01076949e+00 -9.49688315e-01 8.41609836e-01 -1.04378068e+00
-6.35567188e-01 -8.47352982e-01 -9.30015981e-01 -6.52589202e-02
4.11750823e-02 -7.02644944e-01 -7.63603210e-01 -5.94945729e-01
-2.97059834e-01 -5.06506145e-01 9.30438757e-01 -1.35723203e-01
1.78311110e+00 -5.42334139e-01 -6.70793951e-01 8.80993485e-01
1.37487960e+00 1.19605944e-01 3.14694732e-01 6.76142812e-01
4.27027792e-01 2.85567492e-01 5.92206120e-01 4.25682694e-01
4.02268708e-01 1.09348023e+00 2.69473642e-01 3.35975230e-01
-1.18829831e-01 -6.28804386e-01 -1.55866876e-01 7.57042348e-01
6.49426639e-01 -5.44761419e-01 -9.61462319e-01 6.60580397e-01
-1.97857988e+00 -8.84887040e-01 6.38085961e-01 1.95803297e+00
9.05189514e-01 -2.23812625e-01 -1.54198751e-01 -1.53985411e-01
3.92483562e-01 1.90646842e-01 -8.19996357e-01 -1.86809629e-01
3.50167900e-02 3.97305846e-01 3.81776392e-01 3.17888975e-01
-8.45066428e-01 9.80253100e-01 6.64900351e+00 1.33266139e+00
-8.76501083e-01 -9.49616134e-02 1.19756706e-01 -2.93044120e-01
-6.30489469e-01 -1.70805871e-01 -9.02278900e-01 7.52941370e-01
9.81773973e-01 -7.32449055e-01 6.57038510e-01 1.10539591e+00
-1.97652012e-01 5.85820675e-01 -1.13512397e+00 1.46159780e+00
1.22590721e-01 -1.58743191e+00 4.76129562e-01 2.10026398e-01
4.00936544e-01 2.37111464e-01 3.80835772e-01 5.84804773e-01
1.70057118e-01 -1.27152276e+00 1.87291414e-01 6.02451801e-01
7.01002836e-01 -6.27609670e-01 4.97752219e-01 2.40020052e-01
-1.02901542e+00 -3.35970163e-01 -5.76666951e-01 4.67703909e-01
-6.97157159e-02 3.05150360e-01 -4.95867878e-01 2.09846556e-01
9.33965981e-01 5.19742966e-01 -5.40140033e-01 1.24663317e+00
5.35060883e-01 2.02821374e-01 -4.49307263e-01 -3.95619482e-01
4.90898997e-01 -6.52287379e-02 2.15522856e-01 1.17513227e+00
6.26056910e-01 -1.94608971e-01 2.70319302e-02 6.69282854e-01
-4.41835046e-01 3.00665617e-01 -7.14592695e-01 -4.36320901e-01
8.72665048e-01 1.15742350e+00 -1.57361373e-01 -2.30676308e-01
-4.05220896e-01 1.53123403e+00 3.84164989e-01 5.42579889e-01
-4.68108326e-01 -9.49474931e-01 7.90226281e-01 -1.30771399e-02
3.50053668e-01 2.69606914e-02 3.75168741e-01 -1.12385237e+00
5.62290788e-01 -8.16812217e-01 7.31631875e-01 -8.00624967e-01
-1.47487199e+00 6.41314864e-01 -2.95689970e-01 -8.92988384e-01
-6.52513504e-01 -3.95656228e-01 -2.55151510e-01 1.05243278e+00
-1.82858896e+00 -9.11029398e-01 -3.75477940e-01 8.60773861e-01
4.49850321e-01 -5.80254376e-01 1.03352714e+00 6.83180630e-01
-2.03873545e-01 9.54503357e-01 9.10690069e-01 9.04751569e-02
5.96996069e-01 -8.18218768e-01 5.67070127e-01 1.07673243e-01
5.42430282e-01 9.04526353e-01 9.55925211e-02 -2.61493117e-01
-1.92861664e+00 -8.64019513e-01 1.18443525e+00 -5.18988132e-01
9.74758267e-01 -1.48517117e-01 -8.94528627e-01 3.68634641e-01
3.08243446e-02 1.43529922e-01 7.45928943e-01 2.60459986e-02
-6.00058138e-01 -3.42204928e-01 -8.15369368e-01 6.28066897e-01
1.09526992e+00 -1.13901353e+00 -4.80477005e-01 8.38655174e-01
1.29347515e+00 3.04111764e-02 -9.37639117e-01 1.05457723e-01
7.15232015e-01 -4.93374437e-01 1.50992978e+00 -9.00988698e-01
3.95306796e-02 9.41110104e-02 -6.58097923e-01 -9.51918185e-01
-4.98302877e-01 -5.37391722e-01 -9.04517651e-01 6.57945275e-01
2.76762426e-01 -4.25042987e-01 9.29710388e-01 8.52926552e-01
2.23221958e-01 -7.84378767e-01 -7.52482355e-01 -1.01820445e+00
4.71214801e-02 -2.65982300e-01 9.70573068e-01 4.49462742e-01
3.66841583e-03 2.78447241e-01 -1.04080059e-01 -1.39378235e-01
5.17472029e-01 6.51873410e-01 3.42420429e-01 -1.24310410e+00
-2.22222745e-01 -7.40976989e-01 -4.20232803e-01 -1.78410256e+00
2.70470172e-01 -1.40993237e+00 -4.78752881e-01 -1.58959639e+00
5.66197515e-01 -5.24469554e-01 -7.60630131e-01 1.56850502e-01
1.11517319e-02 1.66272298e-01 -1.75266698e-01 7.66821980e-01
-1.13488698e+00 7.45976985e-01 8.89510095e-01 -4.79914635e-01
1.51002750e-01 7.14610964e-02 -8.68332028e-01 -6.00362569e-02
4.62872148e-01 -3.35080385e-01 -5.01469791e-01 -1.12474918e+00
7.45357752e-01 8.62039998e-02 4.19567615e-01 -4.83194649e-01
9.57867026e-01 5.82300453e-03 2.04001799e-01 -5.77598810e-01
4.10553217e-01 -1.11274159e+00 -2.54507512e-01 2.34944060e-01
-7.30749667e-01 7.71641284e-02 -1.60870641e-01 7.76157558e-01
-6.05290473e-01 -2.78164089e-01 2.72088125e-02 2.11368099e-01
-8.82036984e-01 8.35535228e-01 2.43445501e-01 1.72744542e-01
6.38314784e-01 -2.09012795e-02 -2.59435505e-01 -6.42539859e-01
-3.63713026e-01 6.60904229e-01 1.86421186e-01 7.77616680e-01
9.62915063e-01 -1.86333871e+00 -4.18135196e-01 2.54152447e-01
6.74925745e-01 -2.42659062e-01 1.36926640e-02 1.83775067e-01
-4.25856501e-01 1.26994848e+00 3.08432758e-01 -2.52353370e-01
-7.97844768e-01 9.73587692e-01 -7.36209471e-03 -1.40952602e-01
-4.64694917e-01 8.74845922e-01 -1.80478558e-01 -2.72466660e-01
7.56500363e-01 1.83170602e-01 -1.18470579e-01 -3.04460488e-02
8.47082138e-01 3.85912687e-01 1.95409030e-01 -2.27741525e-01
-2.90162057e-01 6.84992671e-01 -4.24231261e-01 5.24420477e-02
1.12436116e+00 -5.30525744e-01 -1.20025851e-01 -8.68512690e-02
2.12091112e+00 -4.62767154e-01 -4.06147420e-01 -8.43523920e-01
3.67883444e-01 -6.83442771e-01 7.15251923e-01 -6.30651414e-01
-9.95373249e-01 6.56685472e-01 7.31058180e-01 3.04458827e-01
9.79438961e-01 3.07246119e-01 1.66281593e+00 1.11711121e+00
4.10507500e-01 -1.33539331e+00 1.30516678e-01 5.31705678e-01
8.44840646e-01 -1.51831067e+00 -4.02856290e-01 3.86438459e-01
-1.34817570e-01 9.66102660e-01 7.32044205e-02 7.16941729e-02
9.29556310e-01 -3.46797585e-01 5.62469885e-02 -6.74609423e-01
-9.65636790e-01 -1.68668449e-01 8.60488415e-01 2.60599345e-01
2.66554385e-01 -2.62340277e-01 -1.49647966e-01 5.47587216e-01
-7.56544769e-02 8.09893459e-02 -5.73788047e-01 9.99947727e-01
-6.46963418e-01 -1.09020066e+00 1.06918640e-01 7.26143122e-01
-3.45994115e-01 -4.70421225e-01 -2.79946208e-01 6.82177767e-02
-9.92680192e-01 9.49330389e-01 2.65239269e-01 -5.49180865e-01
2.77235448e-01 8.38361681e-02 1.08946925e-02 -3.63452673e-01
-2.46882305e-01 -2.39344224e-01 -3.72809559e-01 -1.15546167e+00
3.24126333e-01 -5.19800335e-02 -9.93439436e-01 -3.95771682e-01
-3.09226602e-01 5.44846773e-01 9.28913891e-01 4.71039325e-01
8.67113411e-01 -1.19753137e-01 1.03737724e+00 -2.78431058e-01
-1.26784420e+00 -6.21600032e-01 -7.41384327e-01 5.47832251e-01
2.21941277e-01 -2.99265176e-01 -4.29374039e-01 -5.17986417e-01]
|
[11.350605964660645, 7.439302444458008]
|
a8b874f8-ac18-416b-a8ba-66efb2b9a602
|
is-summary-useful-or-not-an-extrinsic-human
|
2305.15044
| null |
https://arxiv.org/abs/2305.15044v1
|
https://arxiv.org/pdf/2305.15044v1.pdf
|
Is Summary Useful or Not? An Extrinsic Human Evaluation of Text Summaries on Downstream Tasks
|
Research on automated text summarization relies heavily on human and automatic evaluation. While recent work on human evaluation mainly adopted intrinsic evaluation methods, judging the generic quality of text summaries, e.g. informativeness and coherence, our work focuses on evaluating the usefulness of text summaries with extrinsic methods. We carefully design three different downstream tasks for extrinsic human evaluation of summaries, i.e., question answering, text classification and text similarity assessment. We carry out experiments using system rankings and user behavior data to evaluate the performance of different summarization models. We find summaries are particularly useful in tasks that rely on an overall judgment of the text, while being less effective for question answering tasks. The results show that summaries generated by fine-tuned models lead to higher consistency in usefulness across all three tasks, as rankings of fine-tuned summarization systems are close across downstream tasks according to the proposed extrinsic metrics. Summaries generated by models in the zero-shot setting, however, are found to be biased towards the text classification and similarity assessment tasks, due to its general and less detailed summary style. We further evaluate the correlation of 14 intrinsic automatic metrics with human criteria and show that intrinsic automatic metrics perform well in evaluating the usefulness of summaries in the question-answering task, but are less effective in the other two tasks. This highlights the limitations of relying solely on intrinsic automatic metrics in evaluating the performance and usefulness of summaries.
|
['Xiaojun Wan', 'Mingqi Gao', 'Xiao Pu']
|
2023-05-24
| null | null | null | null |
['text-summarization']
|
['natural-language-processing']
|
[ 1.86555088e-01 3.69350821e-01 -7.32987598e-02 -3.19273502e-01
-1.20790768e+00 -6.38655782e-01 1.05337369e+00 1.01343751e+00
-7.23044574e-01 6.80895686e-01 9.22192574e-01 -7.39539936e-02
-2.73494184e-01 -4.43709552e-01 2.06576604e-02 -3.05092633e-01
4.34096575e-01 4.45565701e-01 3.04430485e-01 -2.59657502e-01
9.45770383e-01 -1.95128202e-01 -1.53658020e+00 2.12257907e-01
1.25320160e+00 4.82762516e-01 5.73990047e-02 1.17754030e+00
-3.52915265e-02 7.90597856e-01 -1.07580459e+00 -5.11254013e-01
-4.23942208e-01 -7.71302104e-01 -9.82236385e-01 -8.16489682e-02
6.94914579e-01 -4.46090132e-01 8.38328060e-03 7.92341888e-01
7.23357260e-01 3.19350064e-01 1.05440426e+00 -9.56546247e-01
-5.71253777e-01 8.36139381e-01 -8.78083240e-03 3.30324441e-01
8.60166073e-01 1.13475442e-01 1.52404928e+00 -6.67522848e-01
5.07440507e-01 1.23780823e+00 6.03382826e-01 2.97974229e-01
-1.28229487e+00 -3.79728973e-02 -7.23269656e-02 -2.35957786e-01
-7.58339882e-01 -8.92031431e-01 2.83059448e-01 -6.00581646e-01
9.23226655e-01 6.52168155e-01 7.68706808e-03 9.41923618e-01
1.05664328e-01 6.40477419e-01 7.97284365e-01 -4.96138394e-01
3.92848432e-01 5.21511316e-01 8.71338606e-01 1.31396338e-01
8.48751187e-01 -4.12573218e-01 -5.55520296e-01 -3.77187014e-01
-2.03689113e-02 -6.59961224e-01 -5.03475845e-01 2.62267347e-02
-1.37580276e+00 9.17132854e-01 5.93547383e-03 5.44726670e-01
-5.36847949e-01 -2.47541498e-02 8.05772424e-01 3.02543610e-01
6.81296170e-01 1.25116122e+00 -2.15002030e-01 -3.88191998e-01
-1.24430549e+00 5.38782716e-01 1.20959413e+00 7.88861692e-01
4.08917695e-01 -2.60682821e-01 -1.14553714e+00 9.97216046e-01
1.65943597e-02 4.10005063e-01 7.75963962e-01 -1.20106006e+00
5.51436484e-01 5.57556450e-01 3.78307462e-01 -1.21129417e+00
-4.49406236e-01 -4.49391365e-01 -5.08673966e-01 -1.61038816e-01
3.95064384e-01 -1.58104211e-01 -3.93807530e-01 1.60842502e+00
-1.55723706e-01 -9.47411180e-01 1.26005694e-01 6.27873182e-01
1.35862923e+00 4.03578222e-01 1.69833705e-01 -3.94152224e-01
1.36798894e+00 -1.06507909e+00 -8.90848219e-01 -4.04167846e-02
1.13146901e+00 -8.43458712e-01 1.41775715e+00 1.17429793e-01
-1.26097965e+00 -6.42680287e-01 -1.07476485e+00 -1.50671467e-01
-1.78417832e-01 4.48194116e-01 1.20385438e-01 6.77565813e-01
-1.20943224e+00 7.52437174e-01 -4.60285455e-01 -8.00762057e-01
-1.12733366e-02 -6.69360487e-03 -6.80266395e-02 3.39676708e-01
-9.40076947e-01 1.14221191e+00 2.31522292e-01 -5.23788273e-01
-3.21061671e-01 -3.48710567e-01 -8.15293372e-01 3.00384790e-01
9.48469937e-02 -9.35505927e-01 1.74537671e+00 -6.13742828e-01
-1.13622582e+00 7.30372965e-01 -2.09793881e-01 -4.13655430e-01
4.82820660e-01 -2.67885387e-01 -1.05285132e-02 2.71987140e-01
4.84959513e-01 5.06213188e-01 3.50481242e-01 -1.12100792e+00
-7.37710118e-01 -1.76840872e-01 9.68416855e-02 4.74290341e-01
-6.19658530e-01 -2.23059561e-02 -1.36334538e-01 -5.94785094e-01
-2.61124134e-01 -5.31823635e-01 -1.72132887e-02 -6.93598926e-01
-5.44614911e-01 -6.45480037e-01 5.72725177e-01 -7.60378420e-01
1.93595004e+00 -1.64864016e+00 -1.98190838e-01 -1.51834801e-01
2.98609048e-01 3.25947553e-01 -1.23359732e-01 8.19035649e-01
3.27101707e-01 4.95447338e-01 -1.48328483e-01 -4.29706186e-01
2.91028708e-01 -3.84547263e-01 -1.53231606e-01 1.26858270e-02
5.03190458e-02 9.88654494e-01 -1.20046723e+00 -7.84093916e-01
-8.99928361e-02 4.56349738e-02 -3.66156548e-01 1.62273943e-01
-1.28094256e-01 4.66887048e-03 -5.11317670e-01 2.31753856e-01
-4.16926518e-02 -3.39088529e-01 -2.28421524e-01 -8.68210047e-02
-1.08631417e-01 7.35585690e-01 -5.77690601e-01 1.30106449e+00
-3.58505249e-01 9.37403142e-01 -2.97305286e-01 -5.16399562e-01
7.38078296e-01 3.60447943e-01 -1.78661421e-02 -7.53796101e-01
2.29652375e-02 2.36491486e-01 1.34886205e-01 -6.08390272e-01
1.27549231e+00 1.53991073e-01 -2.39527807e-01 9.78987813e-01
7.55811250e-03 -3.55415702e-01 8.32262874e-01 7.52544224e-01
1.22862852e+00 -1.16304949e-01 5.51696837e-01 -4.81160492e-01
4.40413058e-01 2.48040080e-01 -2.27984339e-01 1.17911041e+00
-1.48073956e-01 7.11005390e-01 7.22985864e-01 2.91250914e-01
-9.79494810e-01 -5.94260454e-01 -8.56100172e-02 1.34890890e+00
6.02489263e-02 -9.08591688e-01 -1.02445924e+00 -7.90172994e-01
-1.27729505e-01 1.40314889e+00 -5.93541682e-01 -2.85787016e-01
-1.96681902e-01 -5.50578833e-01 5.73750913e-01 3.71242672e-01
1.97563365e-01 -1.10280931e+00 -1.01476121e+00 1.97116762e-01
-6.14530325e-01 -8.63691688e-01 -6.41006947e-01 -1.95216447e-01
-1.01462328e+00 -8.85806382e-01 -8.98299992e-01 -3.94950181e-01
4.14959460e-01 4.32384491e-01 1.52435553e+00 2.53170222e-01
3.14143836e-01 8.98080289e-01 -6.79071248e-01 -4.14347291e-01
-7.52880573e-01 5.36551595e-01 -1.39035761e-01 -5.25333345e-01
3.34128767e-01 -2.32129142e-01 -6.76801860e-01 3.38919610e-01
-9.44642425e-01 -1.65385187e-01 6.37780666e-01 7.43629515e-01
-1.27490953e-01 -2.38436610e-01 1.03194761e+00 -1.06917894e+00
1.70380521e+00 -2.51925051e-01 1.82625741e-01 3.23902637e-01
-9.69910204e-01 2.73641676e-01 4.35480833e-01 -8.28908086e-02
-1.03835011e+00 -7.58044124e-01 -6.64760992e-02 3.49755347e-01
-7.53290281e-02 6.45774424e-01 2.42253914e-01 4.65291947e-01
1.32200062e+00 -1.51695937e-01 -6.31581321e-02 -2.11927995e-01
2.40341023e-01 8.52502465e-01 4.06503260e-01 -5.23459136e-01
3.99169475e-01 -9.55414698e-02 -4.82577384e-01 -1.06590641e+00
-1.08418906e+00 -8.73685181e-01 -2.94010103e-01 -2.29631856e-01
7.32896149e-01 -5.42452753e-01 -5.21310091e-01 3.45722549e-02
-1.07996523e+00 -2.41240695e-01 -3.76742303e-01 3.45561445e-01
-6.13112748e-01 8.14271569e-01 -4.78825480e-01 -9.56145048e-01
-8.18720996e-01 -9.85617518e-01 1.35410070e+00 7.33274966e-02
-1.22814989e+00 -1.19784021e+00 2.79310703e-01 4.88271683e-01
5.09476304e-01 1.75924525e-01 9.91410196e-01 -1.18938196e+00
8.10121000e-02 -4.10612345e-01 -1.97585195e-01 3.86974633e-01
2.56856889e-01 5.16810417e-02 -9.87737596e-01 -1.66271761e-01
3.20427045e-02 -4.83306348e-01 1.15038836e+00 4.08266038e-01
6.46822214e-01 -5.85148215e-01 -1.76650196e-01 -3.58988971e-01
1.07131290e+00 -1.49905011e-01 5.48299432e-01 3.55835170e-01
3.51336151e-01 1.15608037e+00 9.01701689e-01 3.31396222e-01
4.59183991e-01 6.42053545e-01 -4.53354903e-02 1.97369725e-01
-2.10441515e-01 -2.66858071e-01 4.29274142e-01 9.36261714e-01
1.60590127e-01 -6.71597719e-01 -8.54731321e-01 6.53451085e-01
-1.95146668e+00 -1.03257263e+00 -4.25045252e-01 2.39527464e+00
8.22164655e-01 3.80321264e-01 3.84328365e-01 1.96658507e-01
6.71108246e-01 1.94631115e-01 -1.62300169e-01 -5.18691480e-01
-2.72420496e-02 -7.83933848e-02 4.78746518e-02 5.10234356e-01
-9.25176322e-01 5.75415969e-01 6.79303789e+00 7.35020876e-01
-5.66321313e-01 -1.25861198e-01 6.69714451e-01 8.86797681e-02
-5.51347613e-01 -3.04992925e-02 -5.97211719e-01 4.77816314e-01
1.06204104e+00 -5.63635647e-01 -2.75323302e-01 5.96102893e-01
4.78580177e-01 -5.95151007e-01 -1.48166454e+00 5.70027053e-01
1.92892522e-01 -1.02051675e+00 2.92434156e-01 -1.10191517e-01
8.03861797e-01 -3.71517360e-01 -2.13360623e-01 5.17321110e-01
7.77626857e-02 -8.15067708e-01 6.78788126e-01 5.82335651e-01
4.64202225e-01 -5.02043128e-01 1.08497977e+00 4.90603417e-01
-4.65598762e-01 2.65547961e-01 -2.29418933e-01 -1.75456658e-01
6.94253743e-02 6.38168037e-01 -8.86855185e-01 2.69942790e-01
2.39243448e-01 4.25173193e-01 -9.49046493e-01 9.34176326e-01
-1.41202539e-01 6.78313434e-01 9.44788083e-02 -5.54418683e-01
2.50006676e-01 -1.27897384e-02 7.40920305e-01 1.64893734e+00
2.11339653e-01 -1.31704807e-01 -8.23045000e-02 6.70709252e-01
-6.70242822e-03 4.73380625e-01 -6.07191682e-01 -2.83904970e-01
4.34570193e-01 1.31153381e+00 -7.10044146e-01 -6.87663257e-01
-5.51502053e-05 5.28152883e-01 1.45609707e-01 1.82478577e-01
-4.14344937e-01 -7.72331774e-01 -5.51596470e-02 1.12791341e-02
-9.28179249e-02 3.39100473e-02 -7.14620769e-01 -9.14704263e-01
1.87721580e-01 -9.41619337e-01 3.56908590e-01 -7.80384123e-01
-1.05763304e+00 5.97153068e-01 1.09420881e-01 -1.06274402e+00
-4.84296054e-01 -1.71674326e-01 -9.37073946e-01 6.87831461e-01
-1.06368685e+00 -5.15929282e-01 -5.74833035e-01 -1.97497964e-01
7.18182445e-01 8.18636715e-02 7.18315542e-01 -1.53100386e-01
-4.09592688e-01 6.42402530e-01 8.58740956e-02 -2.17134252e-01
1.15348458e+00 -1.53053367e+00 4.33089137e-01 7.63707042e-01
-8.00439864e-02 7.18770444e-01 1.21414959e+00 -6.48144782e-01
-6.91988051e-01 -8.58264148e-01 1.20950782e+00 -8.68840635e-01
4.19799477e-01 9.78815705e-02 -8.26912522e-01 1.03235930e-01
4.87244427e-01 -1.12355494e+00 6.87146246e-01 4.00848657e-01
-1.50115281e-01 3.22368681e-01 -8.90792489e-01 6.18952930e-01
7.38661885e-01 -4.36287075e-01 -1.04434979e+00 4.23186332e-01
7.95856237e-01 8.68042409e-02 -8.30684483e-01 2.95051694e-01
4.82527882e-01 -1.08831894e+00 5.76031208e-01 -5.01189590e-01
8.70027781e-01 -9.04155746e-02 4.20112275e-02 -1.52175260e+00
-3.71497124e-01 -4.13267255e-01 4.77330498e-02 1.55614889e+00
7.86894560e-01 -5.20520091e-01 4.47198063e-01 7.15033591e-01
-2.23674104e-01 -4.79246914e-01 -2.34005913e-01 -5.52177012e-01
-6.92389160e-02 -1.66012477e-02 1.77858233e-01 7.05703259e-01
4.04423714e-01 1.26929021e+00 6.11524768e-02 -4.42073464e-01
4.12406921e-01 -5.84521256e-02 9.09037292e-01 -1.35008657e+00
-1.61501542e-01 -1.05361056e+00 -7.46119022e-02 -8.08673441e-01
-6.15143403e-02 -6.69719040e-01 2.98693240e-01 -2.04265237e+00
4.86436337e-01 3.55577692e-02 2.24357173e-02 -3.80737591e-03
-7.30233967e-01 -1.64279453e-02 8.12914595e-02 4.04849887e-01
-1.11862421e+00 5.66630602e-01 1.06969345e+00 -3.69036347e-02
-3.12383592e-01 2.04486907e-01 -1.21155024e+00 5.67234159e-01
7.70999551e-01 -2.04909086e-01 -4.53458965e-01 -3.66102606e-01
3.03417802e-01 8.77285600e-02 6.48322552e-02 -1.10497463e+00
3.54441583e-01 2.23457262e-01 6.50523007e-02 -5.32335699e-01
-5.90537861e-02 -9.97445732e-02 -4.88228291e-01 2.57111400e-01
-1.02681458e+00 1.71285361e-01 -4.33178581e-02 4.22246069e-01
-3.16395491e-01 -7.13106453e-01 4.55748111e-01 -1.25682831e-01
-1.45929217e-01 -2.55441844e-01 -4.26716924e-01 4.34463054e-01
3.76583695e-01 -5.36910892e-01 -5.17765939e-01 -9.72289383e-01
-5.27305305e-01 3.48278344e-01 5.45238256e-01 3.20835203e-01
3.43771905e-01 -9.71333325e-01 -1.15801167e+00 -5.96737683e-01
5.56296110e-01 -4.28041518e-01 -1.39704317e-01 1.01207125e+00
-3.27933758e-01 8.81435454e-01 6.24245070e-02 -6.14322543e-01
-1.20237863e+00 1.08021922e-01 -7.03821629e-02 -6.14526927e-01
-2.10071191e-01 3.73894811e-01 3.58758360e-01 -1.69879377e-01
4.52550858e-01 -3.95267248e-01 -5.96211493e-01 5.76352477e-01
6.14138663e-01 9.36345458e-01 3.63672644e-01 -4.96181279e-01
5.67089505e-02 2.30918825e-01 -1.87373936e-01 -3.54358256e-01
9.11682785e-01 -2.87137657e-01 -4.47115675e-02 5.60040653e-01
1.02476978e+00 8.88324752e-02 -5.13165653e-01 -2.15955894e-03
5.61492026e-01 -1.97371215e-01 3.56855951e-02 -8.57590556e-01
-8.99139047e-02 7.50100493e-01 -9.72026959e-02 8.47469032e-01
7.88704515e-01 -1.47873536e-01 4.98260945e-01 7.34743357e-01
1.32516539e-02 -1.27905250e+00 3.40379655e-01 6.53796375e-01
1.02814150e+00 -1.21478450e+00 2.58292377e-01 -5.10835461e-02
-8.20346594e-01 1.04573143e+00 5.43205857e-01 2.86140352e-01
-1.33336475e-02 -4.50634629e-01 -7.20921978e-02 -3.59943330e-01
-8.90604019e-01 -1.98358610e-01 6.11662865e-01 3.15118939e-01
1.00378358e+00 -1.85068659e-02 -9.28960383e-01 4.68458772e-01
-6.33210957e-01 -3.80921483e-01 7.98971713e-01 6.76532447e-01
-7.66602457e-01 -7.58473039e-01 -1.87437803e-01 1.01562989e+00
-4.47938144e-01 -1.16698872e-02 -8.88583362e-01 6.52410984e-01
-7.36252546e-01 1.55402565e+00 -1.71888843e-01 -1.99232399e-01
5.08538723e-01 2.10792888e-02 1.81211233e-01 -1.01438951e+00
-9.75484490e-01 -1.58966511e-01 8.84790361e-01 -1.69664845e-01
-4.57691669e-01 -7.48398006e-01 -7.60923207e-01 -1.07310988e-01
-6.83889210e-01 7.41846204e-01 5.38059235e-01 8.18661034e-01
2.72777885e-01 4.73420501e-01 4.37315047e-01 -6.67686701e-01
-1.07429373e+00 -1.51025069e+00 -1.32468432e-01 7.05054343e-01
3.29223722e-01 -3.01309496e-01 -6.81196928e-01 -1.01453342e-01]
|
[12.079402923583984, 9.205304145812988]
|
1c37283f-7526-46ad-871e-716b344d5276
|
investigating-monolingual-and-multilingual
|
2103.09519
| null |
https://arxiv.org/abs/2103.09519v1
|
https://arxiv.org/pdf/2103.09519v1.pdf
|
Investigating Monolingual and Multilingual BERTModels for Vietnamese Aspect Category Detection
|
Aspect category detection (ACD) is one of the challenging tasks in the Aspect-based sentiment Analysis problem. The purpose of this task is to identify the aspect categories mentioned in user-generated reviews from a set of pre-defined categories. In this paper, we investigate the performance of various monolingual pre-trained language models compared with multilingual models on the Vietnamese aspect category detection problem. We conduct the experiments on two benchmark datasets for the restaurant and hotel domain. The experimental results demonstrated the effectiveness of the monolingual PhoBERT model than others on two datasets. We also evaluate the performance of the multilingual model based on the combination of whole SemEval-2016 datasets in other languages with the Vietnamese dataset. To the best of our knowledge, our research study is the first attempt at performing various available pre-trained language models on aspect category detection task and utilize the datasets from other languages based on multilingual models.
|
['Ngan Luu-Thuy Nguyen', 'Vu Xuan Hoang', 'Lac Si Le', 'Dang Van Thin']
|
2021-03-17
| null | null | null | null |
['aspect-category-detection']
|
['natural-language-processing']
|
[-2.34267786e-01 -2.83478498e-01 -1.69038489e-01 -3.82842898e-01
-1.06984591e+00 -7.03539729e-01 1.01719475e+00 5.38308024e-01
-8.26706767e-01 3.94685060e-01 3.22545290e-01 -4.22042996e-01
4.96763527e-01 -5.89329958e-01 -1.73079312e-01 -3.46770763e-01
1.95910379e-01 5.89664102e-01 -6.15006164e-02 -5.94022810e-01
6.09864712e-01 -1.50188342e-01 -1.57535899e+00 5.06297946e-01
7.35591292e-01 6.80187225e-01 3.49825397e-02 3.77783716e-01
-3.76270205e-01 4.76955116e-01 -5.35175443e-01 -6.96342051e-01
1.37855262e-01 -2.05310956e-01 -4.18585688e-01 1.88075379e-01
3.29643160e-01 4.47437555e-01 6.20242238e-01 1.01937056e+00
4.10353571e-01 -8.98553953e-02 9.51931357e-01 -9.53900635e-01
-7.03290045e-01 8.54231596e-01 -7.74942577e-01 9.65749025e-02
5.23973346e-01 -3.83467019e-01 1.40718782e+00 -1.22502112e+00
8.41311991e-01 1.31668091e+00 5.62536478e-01 1.48163155e-01
-5.41740358e-01 -5.65898001e-01 5.22784233e-01 5.43792658e-02
-1.36013126e+00 1.85980890e-02 7.62777328e-01 -4.31285679e-01
1.29204464e+00 1.02048904e-01 5.67111731e-01 7.59133637e-01
2.80190110e-01 8.69502962e-01 1.49850917e+00 -6.11781955e-01
1.23139858e-01 8.98339808e-01 6.09884024e-01 4.87797827e-01
4.69968289e-01 -4.61671293e-01 -5.78132331e-01 -1.92207605e-01
-1.39680058e-01 -2.87621439e-01 5.80474176e-02 -1.71186060e-01
-1.17811739e+00 1.13340592e+00 -1.17757112e-01 5.65038741e-01
-2.10818782e-01 -6.57340705e-01 8.42450440e-01 3.35896730e-01
1.06086731e+00 5.97485244e-01 -1.11034572e+00 -1.22450322e-01
-6.60712421e-01 6.50615245e-02 1.19167686e+00 1.10421360e+00
7.11177826e-01 -5.17137982e-02 -1.28762554e-02 1.02098560e+00
5.80352783e-01 7.47002721e-01 5.47647715e-01 -4.07117344e-02
6.38483822e-01 1.01608038e+00 -1.85951516e-01 -7.22107351e-01
-3.61737460e-01 -5.35876870e-01 -3.60985905e-01 -2.95690358e-01
2.14384168e-01 -3.15367252e-01 -1.01470029e+00 1.10031617e+00
4.08248901e-01 -5.32337308e-01 4.20944780e-01 6.90063000e-01
1.31224477e+00 7.56811857e-01 3.40973586e-01 -6.25735968e-02
2.01358056e+00 -1.10513759e+00 -7.45455444e-01 -2.21880630e-01
6.70054197e-01 -1.50585973e+00 1.31924653e+00 4.11187708e-01
-2.90976644e-01 -4.79864895e-01 -9.95581746e-01 -2.63830926e-02
-1.21377170e+00 6.57752156e-01 9.75206792e-01 7.78152227e-01
-7.42995322e-01 -4.79721487e-01 -4.71662909e-01 -8.49855542e-01
-1.71711996e-01 -8.64943936e-02 -3.14099342e-01 -9.24423039e-02
-1.17095733e+00 7.85027742e-01 1.87246904e-01 -9.86788347e-02
-8.46046269e-01 -4.45902944e-01 -1.16698015e+00 -4.56809968e-01
1.91243753e-01 -2.05352500e-01 1.14163148e+00 -1.09139049e+00
-8.28072548e-01 1.31985247e+00 -4.06813413e-01 -2.10561618e-01
-5.00749573e-02 -3.02176028e-01 -6.94098890e-01 -1.87672928e-01
6.81897461e-01 4.60804939e-01 5.82634687e-01 -1.36954045e+00
-1.03769875e+00 -5.12652338e-01 2.12657124e-01 5.16640425e-01
-3.18449259e-01 3.44706148e-01 -5.67391038e-01 -6.10391915e-01
-2.13328212e-01 -1.15607381e+00 -2.47200415e-01 -8.27304006e-01
-3.63845915e-01 -6.07686341e-01 7.60460019e-01 -6.69611216e-01
1.15660512e+00 -2.07958817e+00 -3.34665656e-01 -3.08381487e-02
-1.47618443e-01 -3.03623034e-03 -5.39823659e-02 5.52259922e-01
1.81596085e-01 1.84888124e-01 -1.86307356e-01 -4.58673149e-01
8.49335566e-02 -2.47456897e-02 -2.28921711e-01 2.65872478e-01
7.39431009e-03 6.15259707e-01 -7.06822932e-01 -6.18612289e-01
-1.38585001e-01 5.90958178e-01 -4.53165174e-01 -5.43041946e-03
-4.40003015e-02 9.58300605e-02 -3.68664622e-01 9.52263057e-01
5.39141059e-01 4.64281768e-01 1.28886536e-01 -3.40035290e-01
-4.87352639e-01 6.29104793e-01 -9.40749586e-01 1.45901489e+00
-7.61047184e-01 5.67577362e-01 -2.42592424e-01 -4.92017806e-01
9.97019708e-01 3.59053463e-01 1.58130512e-01 -6.08360350e-01
2.35114202e-01 4.78107244e-01 1.13229759e-01 -2.64647722e-01
1.08725500e+00 -3.38860095e-01 -6.78970814e-01 5.28245807e-01
1.76146954e-01 -4.65723604e-01 8.15324843e-01 2.06288710e-01
4.10137683e-01 -6.19648769e-03 5.95170975e-01 -6.25171363e-01
8.55399489e-01 5.50404489e-01 3.58205497e-01 5.02597272e-01
-3.14767271e-01 5.89267492e-01 5.27449250e-01 -3.77044261e-01
-9.40386236e-01 -8.17457497e-01 -3.77418965e-01 1.23596346e+00
-1.80694401e-01 -8.20792556e-01 -4.96133000e-01 -1.06416345e+00
-2.53345609e-01 9.55255091e-01 -6.60314322e-01 3.15560549e-01
-3.51171672e-01 -1.17597663e+00 2.58452356e-01 2.83277761e-02
6.25467956e-01 -1.02343082e+00 -1.86910719e-01 -3.95841897e-02
-8.20434988e-02 -1.35220349e+00 -6.15276992e-01 2.05654368e-01
-3.76783609e-01 -1.08986819e+00 -4.41369593e-01 -1.35228479e+00
5.63582063e-01 2.36764923e-01 1.48843622e+00 -4.18839902e-01
5.51226288e-02 4.83189166e-01 -7.07467556e-01 -9.42172170e-01
-2.49179959e-01 5.19235194e-01 -1.89192876e-01 -3.19573749e-03
1.28705096e+00 1.87277570e-01 -3.49619210e-01 1.68809041e-01
-7.97489643e-01 -3.41207206e-01 5.00473559e-01 3.32220495e-01
7.97276258e-01 2.02973142e-01 4.55512315e-01 -1.51136315e+00
9.53179836e-01 -5.71457744e-01 -4.07054842e-01 3.60884070e-02
-8.12615991e-01 -1.23983584e-01 4.89628673e-01 -2.40642726e-01
-1.22210062e+00 -6.91016838e-02 -3.24148297e-01 5.63651979e-01
-1.00463077e-01 8.64972472e-01 -2.06117049e-01 3.62335801e-01
1.31646156e-01 2.07583681e-01 -6.13590598e-01 -4.64980632e-01
2.88378000e-01 9.99802828e-01 -6.18829578e-02 -3.47985148e-01
4.29957479e-01 2.99067825e-01 -5.77943027e-01 -1.00494254e+00
-1.15575075e+00 -1.13893104e+00 -7.10685015e-01 -1.27304137e-01
1.04120553e+00 -1.39393699e+00 5.71512692e-02 4.97279942e-01
-9.96267498e-01 3.49605411e-01 -8.31781775e-02 4.76026714e-01
8.85166004e-02 2.49077335e-01 -6.21012568e-01 -7.25580633e-01
-7.39266872e-01 -1.21001089e+00 1.40390384e+00 2.37012352e-03
-3.19978029e-01 -1.23549831e+00 4.28275287e-01 5.18024802e-01
2.75403142e-01 -7.13232309e-02 9.43954468e-01 -1.05951571e+00
-2.73214504e-02 -3.41427118e-01 1.24634668e-01 4.26055193e-01
2.95514166e-01 4.30269912e-02 -8.21291149e-01 -1.53302714e-01
2.67142534e-01 -1.90887004e-01 9.14941132e-01 7.66545767e-03
1.60743207e-01 -7.13073760e-02 9.40144658e-02 6.51593208e-02
1.60152197e+00 2.03127384e-01 1.78001955e-01 7.86251187e-01
8.21509302e-01 6.14284754e-01 1.12344420e+00 2.15252548e-01
1.02758515e+00 2.17760921e-01 5.49896678e-04 -2.10195631e-01
-2.57614627e-02 -2.70833373e-01 5.56299031e-01 1.52399862e+00
2.48546198e-01 -1.20631069e-01 -8.83268952e-01 9.53908682e-01
-1.50626981e+00 -4.17588890e-01 -4.37810987e-01 1.99461401e+00
8.77607048e-01 2.56658882e-01 3.88614573e-02 -6.36028498e-02
4.80062038e-01 4.60013211e-01 -2.54758243e-02 -1.02410698e+00
-2.19276324e-01 1.04493886e-01 2.87330478e-01 3.84179741e-01
-1.74796844e+00 1.18340445e+00 5.48259878e+00 8.94382358e-01
-8.09938133e-01 4.36440766e-01 6.41012132e-01 3.93234819e-01
-3.93039227e-01 1.15026139e-01 -1.39734280e+00 4.83248457e-02
9.46885586e-01 -9.61715654e-02 -1.41409382e-01 1.13493538e+00
1.11981004e-01 -3.35925698e-01 -6.73307955e-01 6.32416487e-01
7.78421402e-01 -4.74292666e-01 2.96859384e-01 -1.01734079e-01
1.15542853e+00 2.73122221e-01 7.69490004e-02 8.44652951e-01
1.50631219e-01 -5.90920210e-01 5.95469654e-01 -8.19242001e-02
5.15381217e-01 -1.03788686e+00 1.12442207e+00 1.30111977e-01
-1.38073540e+00 3.97793651e-01 -5.60216844e-01 8.81969705e-02
1.06911987e-01 6.47169650e-01 -8.19832146e-01 3.37913960e-01
6.46950066e-01 1.08178532e+00 -8.94982696e-01 7.80785918e-01
-4.91962969e-01 8.06656361e-01 -4.52180542e-02 -3.63174915e-01
5.09350419e-01 -4.34100837e-01 6.23161495e-01 1.43981171e+00
1.64419964e-01 -5.89491189e-01 3.90167952e-01 1.96754634e-01
4.75910716e-02 9.49075520e-01 -8.28702152e-01 -9.97890979e-02
-1.71856761e-01 1.72908986e+00 -8.42760861e-01 -5.09521425e-01
-8.43199790e-01 4.51214999e-01 5.28778657e-02 1.57958642e-01
-4.87383008e-01 -4.46487188e-01 5.82966924e-01 -9.19705257e-02
3.70774209e-01 -2.99952924e-01 -4.56242710e-01 -1.51237285e+00
-1.24451056e-01 -1.25177169e+00 6.74612820e-01 -4.31918919e-01
-1.33440626e+00 1.10430717e+00 -6.91229664e-03 -1.39662063e+00
-1.46147415e-01 -6.92612410e-01 -5.90577841e-01 6.51297212e-01
-1.92698860e+00 -1.58513677e+00 1.84919745e-01 4.13283467e-01
1.10449731e+00 -6.17264688e-01 9.01925325e-01 4.18381721e-01
-3.80141437e-01 5.33312857e-01 -9.14777741e-02 1.21144593e-01
9.06688213e-01 -1.35529518e+00 2.85731316e-01 1.00968134e+00
3.80603939e-01 7.55605519e-01 7.23767638e-01 -8.28614712e-01
-1.13254321e+00 -1.09171999e+00 1.66717756e+00 -7.38300443e-01
9.46139514e-01 -5.77913702e-01 -5.69646537e-01 6.52346134e-01
9.14467156e-01 -8.66665959e-01 1.03109372e+00 4.92534280e-01
-4.10009325e-01 4.39563878e-02 -9.42309320e-01 4.25794810e-01
3.55112910e-01 -5.66765666e-01 -7.64175177e-01 3.76576364e-01
6.64412141e-01 5.28762750e-02 -8.78035784e-01 4.58249032e-01
3.37855935e-01 -5.66498041e-01 6.77748501e-01 -3.54094237e-01
5.14273822e-01 -5.24195373e-01 -3.76614153e-01 -1.51193225e+00
2.04448968e-01 1.48367316e-01 7.17888415e-01 1.90880835e+00
1.08147931e+00 -7.28553891e-01 2.10801840e-01 9.03979391e-02
4.07263450e-02 -4.96934950e-01 -5.61312079e-01 -1.84326276e-01
2.17429370e-01 -6.51055634e-01 4.02726442e-01 7.73449421e-01
-1.78124502e-01 1.06331241e+00 1.87418126e-02 -9.66670737e-02
4.84903365e-01 4.96882975e-01 6.07077777e-01 -7.09292114e-01
-9.62018073e-02 -1.18315905e-01 -1.62773013e-01 -3.27262461e-01
3.57075721e-01 -9.69080746e-01 1.17175452e-01 -1.48314118e+00
5.00320911e-01 -2.45658070e-01 -2.25643456e-01 3.30711514e-01
-5.23221195e-01 4.12541538e-01 1.58562541e-01 1.06016122e-01
-9.64976549e-01 5.41353524e-01 1.02464724e+00 -4.36633706e-01
-1.11668766e-01 2.71625876e-01 -1.13467884e+00 8.51127446e-01
1.00011063e+00 -6.74902678e-01 -4.96905655e-01 -3.02567601e-01
4.43777740e-01 -8.39059591e-01 -5.10047495e-01 -5.75534701e-01
-2.07254708e-01 2.53758132e-01 2.93484926e-02 -1.03758693e+00
-1.22024240e-02 -6.07276618e-01 -4.42824394e-01 2.33037457e-01
-1.65226251e-01 6.36893630e-01 2.48081341e-01 2.09310338e-01
-7.22010672e-01 -2.45739028e-01 3.82824302e-01 -3.83361250e-01
-1.13892317e+00 -2.80336291e-02 -9.32046294e-01 4.09596771e-01
8.46273780e-01 3.26697826e-01 -3.56039286e-01 -2.24329919e-01
-2.23877981e-01 1.56083331e-01 4.29731011e-01 8.33241463e-01
4.51974750e-01 -1.05703485e+00 -7.46974051e-01 1.98122799e-01
7.00118601e-01 -4.20473009e-01 -2.26386413e-01 7.23168135e-01
-5.11851966e-01 8.26938868e-01 1.22195713e-01 -4.51585472e-01
-1.43162596e+00 6.20809972e-01 4.82174493e-02 -7.53943086e-01
1.05101533e-01 4.78555948e-01 3.37107003e-01 -1.33647871e+00
-5.59346117e-02 -1.14307612e-01 -1.15843391e+00 4.75379139e-01
2.58573920e-01 -4.30791453e-02 2.47675464e-01 -1.09504616e+00
-6.31657600e-01 7.79978335e-01 -3.00240725e-01 -2.71366954e-01
1.28700197e+00 -2.65250027e-01 -5.60766160e-01 7.83139288e-01
1.39088297e+00 4.29843247e-01 -1.04099050e-01 -6.12169802e-02
2.66632736e-01 -3.96079663e-03 6.83832495e-03 -8.86299253e-01
-9.87564504e-01 5.86386919e-01 5.80751717e-01 1.06562138e-01
1.03017867e+00 -1.32823698e-02 6.16428614e-01 1.69027686e-01
4.10752326e-01 -1.36092865e+00 -3.85375321e-01 9.83792424e-01
6.78668201e-01 -1.63842905e+00 -3.43536586e-02 -4.65012252e-01
-1.21674895e+00 6.88024759e-01 7.47250736e-01 -1.07398994e-01
1.21674490e+00 1.30192161e-01 7.85124660e-01 -4.26755875e-01
-8.14444423e-01 -7.33366311e-01 6.24359190e-01 3.73992920e-01
1.07099140e+00 3.47129405e-01 -1.24991834e+00 6.98628902e-01
-4.64136720e-01 -6.35432720e-01 6.16670370e-01 1.05186343e+00
-2.28135526e-01 -1.39019144e+00 -1.94616452e-01 5.40160477e-01
-9.77990985e-01 -6.00164235e-01 -6.11768782e-01 9.38263595e-01
1.45762518e-01 1.32740676e+00 -3.08585078e-01 -1.04573995e-01
4.50928539e-01 1.75512042e-02 -2.29966789e-01 -9.69489872e-01
-1.07644022e+00 4.69969869e-01 6.08594239e-01 2.06872374e-02
-8.09773982e-01 -9.95737553e-01 -8.78121197e-01 1.35461181e-01
-2.21458822e-02 6.38127685e-01 1.03089345e+00 8.87071550e-01
1.45163253e-01 2.91382909e-01 6.84502959e-01 -4.11402553e-01
-1.41938659e-03 -1.14601111e+00 -7.96486676e-01 4.70644474e-01
-1.08989798e-01 -3.13999683e-01 -5.87718427e-01 -1.07513852e-01]
|
[11.295464515686035, 6.819601535797119]
|
17322890-0052-4951-8231-9cc94de98dd9
|
direct-image-to-point-cloud-descriptors
|
1906.06064
| null |
https://arxiv.org/abs/1906.06064v1
|
https://arxiv.org/pdf/1906.06064v1.pdf
|
Direct Image to Point Cloud Descriptors Matching for 6-DOF Camera Localization in Dense 3D Point Cloud
|
We propose a novel concept to directly match feature descriptors extracted from RGB images, with feature descriptors extracted from 3D point clouds. We use this concept to localize the position and orientation (pose) of the camera of a query image in dense point clouds. We generate a dataset of matching 2D and 3D descriptors, and use it to train a proposed Descriptor-Matcher algorithm. To localize a query image in a point cloud, we extract 2D keypoints and descriptors from the query image. Then the Descriptor-Matcher is used to find the corresponding pairs 2D and 3D keypoints by matching the 2D descriptors with the pre-extracted 3D descriptors of the point cloud. This information is used in a robust pose estimation algorithm to localize the query image in the 3D point cloud. Experiments demonstrate that directly matching 2D and 3D descriptors is not only a viable idea but also achieves competitive accuracy compared to other state-of-the-art approaches for camera pose localization.
|
['Mohammed Bennamoun', 'Mohammad A. A. K. Jalwana', 'Ferdous Sohel', 'Uzair Nadeem', 'Roberto Togneri']
|
2019-06-14
| null | null | null | null |
['camera-localization']
|
['computer-vision']
|
[-8.74336883e-02 -5.01419008e-01 -1.89140335e-01 -1.21184371e-01
-9.24480140e-01 -8.14743817e-01 5.77726960e-01 2.45527819e-01
-4.95251954e-01 -2.91473538e-01 -2.46720597e-01 2.96102256e-01
-1.46535069e-01 -6.09730959e-01 -7.80123830e-01 -4.39920813e-01
1.51774548e-02 7.58643866e-01 6.53456092e-01 4.67645042e-02
7.66191542e-01 1.40175319e+00 -1.71747255e+00 -3.63842815e-01
1.59543440e-01 1.36947489e+00 3.55206370e-01 3.95509690e-01
-1.69293970e-01 -2.19143480e-02 -4.51978803e-01 2.77957908e-04
6.29290998e-01 -1.62175864e-01 -4.36254203e-01 2.02803656e-01
7.60668159e-01 -1.89066499e-01 -3.09482008e-01 9.52594578e-01
2.82038152e-01 5.61962686e-02 4.66065526e-01 -1.31283069e+00
-1.08808249e-01 -5.98211467e-01 -7.38433301e-01 -2.08958298e-01
1.14662945e+00 -2.28164792e-01 6.87747180e-01 -1.21275949e+00
8.93776119e-01 1.10589051e+00 6.00661993e-01 2.53937006e-01
-8.68734181e-01 -6.45553887e-01 -1.84089810e-01 7.02789202e-02
-2.06381369e+00 -1.83613122e-01 1.10102534e+00 -3.09798211e-01
8.42872024e-01 2.39739358e-01 1.10670245e+00 4.87415165e-01
1.95084274e-01 4.09888387e-01 8.63946617e-01 -5.49129546e-01
2.59467930e-01 3.47335860e-02 -2.08484292e-01 9.43722963e-01
5.04019856e-03 2.85450965e-01 -6.46076024e-01 -5.01854181e-01
1.05210006e+00 5.73805928e-01 -1.12112731e-01 -1.07046652e+00
-1.38633609e+00 6.59540236e-01 8.54010165e-01 1.41762197e-01
-6.59349799e-01 3.30964923e-01 -1.65304303e-01 -6.33003041e-02
1.12632215e-01 4.00215298e-01 -1.97365642e-01 -1.61053047e-01
-8.58809710e-01 3.74314189e-01 6.12684488e-01 1.42746770e+00
1.27034497e+00 -7.14011252e-01 2.64076591e-01 3.12271029e-01
6.02894247e-01 1.06697023e+00 3.20863128e-01 -9.22085702e-01
2.35561520e-01 1.10044551e+00 2.53504485e-01 -1.41863978e+00
-2.29064673e-01 1.04051143e-01 -2.48556435e-01 2.15765908e-01
-1.54238030e-01 6.31949842e-01 -6.92378223e-01 8.55657935e-01
7.39573300e-01 1.86835334e-01 -1.77565426e-01 1.11851013e+00
5.87282658e-01 5.18202305e-01 -5.83426178e-01 2.08286375e-01
1.21476531e+00 -5.09621680e-01 -5.30608930e-03 -3.89878362e-01
2.61977375e-01 -1.08183563e+00 4.45187122e-01 -1.73418280e-02
-8.36896062e-01 -5.48433304e-01 -1.04758823e+00 -4.32792678e-02
-2.13645443e-01 2.43187293e-01 1.48111671e-01 1.23076610e-01
-8.05299461e-01 4.61120099e-01 -8.62326026e-01 -5.43306291e-01
2.06213705e-02 6.68997943e-01 -7.45051265e-01 -2.76842684e-01
-4.90678281e-01 9.64104593e-01 2.89715707e-01 -1.56381086e-01
-6.99709833e-01 -3.25149983e-01 -1.02532768e+00 -2.46598706e-01
1.07388169e-01 -6.57837272e-01 9.74690557e-01 -4.98572439e-01
-1.23153150e+00 1.42176378e+00 -5.01860380e-01 1.79063872e-01
7.99031332e-02 -1.55871406e-01 5.78873465e-03 4.53340888e-01
3.65330696e-01 5.97940922e-01 9.07961965e-01 -1.35732758e+00
-7.29723692e-01 -8.29553127e-01 -3.00602913e-01 3.08297724e-01
3.07437927e-01 7.97684267e-02 -8.97627413e-01 6.28089160e-02
1.10691488e+00 -1.10835814e+00 -1.37371212e-01 3.60253990e-01
-2.38955453e-01 -3.19398910e-01 1.00035810e+00 6.13508448e-02
5.76974273e-01 -2.23157859e+00 1.08504429e-01 6.04517519e-01
1.33930907e-01 5.08590117e-02 -1.16064526e-01 5.56763232e-01
6.82611391e-02 -3.13816190e-01 3.90172362e-01 -4.23079133e-01
-1.00568488e-01 2.52115846e-01 -1.16571568e-01 9.27616596e-01
3.35326374e-01 8.15706849e-01 -8.87746811e-01 -5.35360873e-01
5.86394072e-01 6.39314175e-01 -1.52045190e-01 4.79681402e-01
1.69768065e-01 1.79045752e-01 -9.21486974e-01 8.34340513e-01
9.26901042e-01 1.26948982e-01 -4.39611435e-01 -5.10447383e-01
-3.40253145e-01 1.08864143e-01 -1.39927614e+00 1.96791399e+00
-2.17193604e-01 2.97254890e-01 -2.18974248e-01 -5.57686329e-01
1.56076086e+00 9.86584201e-02 7.44667053e-01 -5.61941624e-01
2.49503359e-01 4.73315060e-01 -6.60676122e-01 -2.65284657e-01
4.99788284e-01 1.11910477e-01 -2.59546012e-01 1.07392162e-01
1.16654471e-01 -8.04926157e-01 -2.39087865e-01 -1.70957863e-01
1.09386277e+00 1.69742361e-01 4.78995204e-01 1.29357487e-01
7.45636165e-01 1.96076602e-01 2.20837921e-01 5.32082617e-01
-8.24076030e-03 7.20224917e-01 9.33618098e-02 -6.08309805e-01
-1.26465917e+00 -1.08775675e+00 -7.09539056e-02 2.81832427e-01
7.22279966e-01 -6.48029327e-01 -3.97303879e-01 -5.37716091e-01
3.82362276e-01 -2.86201626e-01 -3.44360054e-01 -1.08334236e-01
-4.38241422e-01 4.28760409e-01 -1.05015248e-01 4.33599174e-01
4.35279459e-01 -2.94024855e-01 -1.17160809e+00 -4.42939959e-02
2.75943369e-01 -1.16509283e+00 -4.45737302e-01 2.71216154e-01
-9.79520261e-01 -1.22037601e+00 -5.94446480e-01 -9.37281370e-01
1.01237154e+00 7.38910973e-01 8.74998331e-01 2.07238019e-01
-6.59038797e-02 8.27490032e-01 -5.79544246e-01 -2.48418555e-01
-7.37236142e-02 -1.49572343e-01 2.40016997e-01 -1.06646597e-01
8.05170774e-01 -2.18883246e-01 -7.35631585e-01 7.41387546e-01
-5.71005166e-01 -3.59862834e-01 6.43374205e-01 4.95804369e-01
1.18531406e+00 -2.51480103e-01 -4.06489015e-01 3.85781340e-02
7.01096430e-02 1.50610939e-01 -1.17378461e+00 7.23311305e-02
-2.69945860e-01 1.66819431e-02 1.12580322e-01 -3.18365693e-01
-2.40279716e-02 1.10668457e+00 4.11519706e-02 -9.59704816e-01
-3.46357644e-01 3.24974447e-01 -5.52268513e-02 -7.02331007e-01
4.25486207e-01 2.64302284e-01 1.08486518e-01 -4.68023181e-01
3.34704131e-01 8.21686685e-01 6.30609751e-01 -4.49679285e-01
1.31970215e+00 6.19453192e-01 4.94131893e-01 -5.89952767e-01
-6.04252994e-01 -1.29684687e+00 -1.43608463e+00 -1.28895432e-01
8.02479625e-01 -1.02647829e+00 -8.65243077e-01 2.01899648e-01
-1.52035081e+00 5.48107028e-01 -2.67095476e-01 7.20100701e-01
-7.47196674e-01 2.99833119e-01 1.40223414e-01 -6.13252461e-01
-2.08378911e-01 -1.36950898e+00 1.96543872e+00 3.31839383e-01
1.14533184e-02 -6.30436003e-01 4.32544172e-01 -1.14078313e-01
-6.22437373e-02 3.77509505e-01 3.88224006e-01 -5.06240308e-01
-8.70401204e-01 -1.14713824e+00 -1.90968856e-01 -5.88293970e-02
-6.27510846e-02 9.02390778e-02 -8.99416804e-01 -2.34814391e-01
9.14174095e-02 7.21911639e-02 2.09724426e-01 4.90170270e-02
5.39045930e-01 2.97872752e-01 -7.36646056e-01 8.95738542e-01
1.71254456e+00 2.02445641e-01 3.70036989e-01 4.91389871e-01
5.11194587e-01 1.97186425e-01 1.03689146e+00 5.00828505e-01
3.77766579e-01 1.05719304e+00 6.24282598e-01 -3.48486342e-02
3.20047200e-01 -6.26633644e-01 1.73284009e-01 6.94682181e-01
6.46263286e-02 5.51506639e-01 -1.09947562e+00 2.38154143e-01
-1.68228459e+00 -4.75531697e-01 -3.37421261e-02 2.56697154e+00
2.85088450e-01 -1.21231616e-01 6.44728765e-02 -2.40957029e-02
6.38702333e-01 -1.41936004e-01 -4.10516471e-01 -4.61698659e-02
1.53248861e-01 3.28478366e-01 7.01810896e-01 2.38068730e-01
-1.13796186e+00 9.40312564e-01 5.93775749e+00 3.62720996e-01
-1.20889997e+00 -2.94671893e-01 -3.32155794e-01 2.11594850e-01
7.71519393e-02 3.67003173e-01 -1.12283802e+00 1.47970483e-01
5.17128170e-01 -2.52400219e-01 5.63412681e-02 1.13844264e+00
-1.01607263e-01 -3.88659120e-01 -1.44904125e+00 1.61428142e+00
4.90697026e-01 -1.10880876e+00 -1.86860502e-01 3.62262815e-01
5.80273151e-01 1.50215194e-01 -1.57764047e-01 -2.33516306e-01
-4.59526509e-01 -6.74446583e-01 8.58409822e-01 5.02474666e-01
6.55247509e-01 -7.36960828e-01 7.94582427e-01 3.90396953e-01
-1.54487133e+00 6.48346543e-02 -8.88426960e-01 8.73373598e-02
-8.60882774e-02 2.55333900e-01 -9.55443144e-01 4.41840559e-01
6.69957280e-01 9.02387261e-01 -6.74831927e-01 1.54329157e+00
-3.12182367e-01 -3.13881516e-01 -6.31379128e-01 -1.04143165e-01
3.13400865e-01 -2.96873748e-01 4.46142167e-01 6.94170237e-01
7.38493204e-01 -5.12266122e-02 3.88079822e-01 8.82646501e-01
-8.74047633e-03 7.83339366e-02 -1.10465205e+00 1.36952147e-01
9.26484764e-01 1.33403063e+00 -7.79718697e-01 -1.61852509e-01
-2.89463580e-01 1.21401000e+00 1.65811423e-02 -1.20433129e-01
-5.35653174e-01 -6.10335648e-01 5.53372324e-01 1.51342973e-01
4.78120118e-01 -7.58812428e-01 5.45657098e-01 -9.21238124e-01
2.15347335e-01 -4.51845974e-01 -4.41993587e-02 -1.18135118e+00
-9.35939848e-01 5.56237698e-01 1.85319126e-01 -1.89717531e+00
-4.15468603e-01 -8.85734022e-01 -4.26296473e-01 1.05654204e+00
-1.42628384e+00 -1.34227169e+00 -7.62643218e-01 8.00320387e-01
-1.00310542e-01 1.23238184e-01 9.59728599e-01 -2.41370443e-02
1.58039793e-01 8.83830190e-02 1.92991540e-01 3.13604712e-01
5.68924248e-01 -1.02281785e+00 5.45841217e-01 3.78961146e-01
4.13184434e-01 6.97845280e-01 2.27684766e-01 -4.38772112e-01
-2.18835163e+00 -6.79502547e-01 6.58699930e-01 -9.61211264e-01
4.63786095e-01 -5.74189067e-01 -4.59766775e-01 5.04554987e-01
-4.05982167e-01 4.44640875e-01 4.00719911e-01 -3.47782373e-01
-3.88280481e-01 -3.37081313e-01 -1.09617293e+00 -1.45909758e-02
8.05233777e-01 -8.49522948e-01 -6.95848525e-01 4.42499816e-01
5.60221493e-01 -9.01464820e-01 -1.08685720e+00 1.16689354e-01
6.56471074e-01 -8.07165146e-01 1.36041439e+00 -3.10423709e-02
-1.94861144e-01 -6.16045773e-01 -4.79154587e-01 -1.13222575e+00
-1.37560353e-01 -3.88779283e-01 2.52355725e-01 8.71093392e-01
-1.66546211e-01 -2.84660310e-01 9.89365697e-01 2.91503131e-01
9.62326527e-02 -4.45845783e-01 -1.35647202e+00 -9.99893367e-01
-3.40946436e-01 -2.41755962e-01 7.59198070e-01 3.98811221e-01
-2.01920211e-01 2.06710532e-01 2.40255207e-01 5.26857734e-01
7.64054120e-01 6.04091406e-01 1.10198617e+00 -1.52244294e+00
4.52339172e-01 -6.90184757e-02 -1.51232815e+00 -1.28600454e+00
3.08347106e-01 -8.51929188e-01 1.24123208e-01 -1.32873905e+00
-1.88877434e-02 -4.75765824e-01 1.14699945e-01 1.40555501e-01
2.67286688e-01 2.32849076e-01 4.10092115e-01 5.34398019e-01
-6.04071856e-01 4.22081590e-01 9.46056545e-01 9.18863267e-02
-1.20899752e-01 1.07283354e-01 -1.99807994e-02 4.51675802e-01
2.31009632e-01 -7.13958204e-01 8.73415768e-02 -4.10633326e-01
4.42077219e-02 -2.07346343e-02 7.04292536e-01 -1.31916869e+00
5.75787127e-01 -8.30106661e-02 6.47874296e-01 -1.18987572e+00
6.48178101e-01 -1.39393342e+00 8.63347426e-02 4.43201244e-01
1.26157701e-01 5.06524265e-01 -1.31130978e-01 3.64233851e-01
-3.24336857e-01 -4.50866789e-01 6.08343959e-01 -3.66327792e-01
-8.50326478e-01 4.98610854e-01 2.54890412e-01 -4.83333200e-01
1.30575287e+00 -6.14158094e-01 8.49445090e-02 -1.91266671e-01
-3.21641237e-01 -1.12377621e-01 1.11621296e+00 3.64494860e-01
1.29045689e+00 -1.71968281e+00 -2.60970473e-01 6.90932930e-01
5.77940524e-01 2.85908461e-01 -3.05500090e-01 7.31836557e-01
-8.51345360e-01 5.58240652e-01 -1.87212557e-01 -1.36016238e+00
-1.34390795e+00 8.76860797e-01 4.65405017e-01 3.98179650e-01
-4.20925528e-01 7.19725788e-01 -3.82169157e-01 -5.77015936e-01
2.17785224e-01 -4.92832005e-01 3.35344613e-01 -1.38719514e-01
4.70616102e-01 1.08464547e-01 2.79441059e-01 -1.20990539e+00
-8.56766522e-01 1.80379474e+00 3.19171697e-01 -5.90659194e-02
1.14190364e+00 -7.56567568e-02 -1.13565318e-01 2.06366405e-01
1.90975225e+00 2.33830735e-02 -9.38285053e-01 -4.95723724e-01
1.55813098e-01 -1.08924985e+00 7.54693449e-02 -1.30168498e-01
-8.57228518e-01 9.19581532e-01 1.02250600e+00 -1.49310604e-01
9.42286968e-01 5.69404662e-01 5.18593848e-01 4.78065759e-01
7.88824916e-01 -6.44456506e-01 1.26724884e-01 5.46572328e-01
7.62688279e-01 -1.11828113e+00 3.14596653e-01 -2.72738367e-01
-2.78918505e-01 1.44794011e+00 2.91538417e-01 -5.08220494e-01
6.77887738e-01 4.84379753e-02 1.24363288e-01 -6.43319905e-01
-1.97658017e-01 -4.17428672e-01 5.42640865e-01 8.85874629e-01
3.22130620e-02 -2.97167748e-01 3.36926013e-01 -2.13074550e-01
-8.74675959e-02 -7.56385922e-02 1.16109438e-01 1.20884526e+00
-6.94119692e-01 -1.29374421e+00 -8.37809622e-01 -2.27211773e-01
3.40077132e-01 3.41069639e-01 -6.97133243e-01 8.92032623e-01
7.55148605e-02 5.54915130e-01 4.48124379e-01 -7.22328365e-01
8.67658377e-01 -1.88470230e-01 7.03613877e-01 -5.89771509e-01
-3.68201673e-01 4.11474615e-01 -7.65124917e-01 -9.88977671e-01
-5.39511263e-01 -6.03075266e-01 -1.12988782e+00 6.62979037e-02
-4.70993876e-01 2.28869244e-01 1.29235256e+00 5.03402054e-01
5.77042520e-01 -4.31887656e-01 1.14593685e+00 -1.40550840e+00
-6.64575040e-01 -3.99237394e-01 -6.23481333e-01 4.38372433e-01
3.96950155e-01 -9.06649947e-01 -3.62665623e-01 -2.21593469e-01]
|
[7.644735813140869, -2.3655097484588623]
|
354e00de-fe4b-45d0-aef7-a56b35508aa7
|
class-specific-variational-auto-encoder-for
|
2304.11734
| null |
https://arxiv.org/abs/2304.11734v1
|
https://arxiv.org/pdf/2304.11734v1.pdf
|
Class-Specific Variational Auto-Encoder for Content-Based Image Retrieval
|
Using a discriminative representation obtained by supervised deep learning methods showed promising results on diverse Content-Based Image Retrieval (CBIR) problems. However, existing methods exploiting labels during training try to discriminate all available classes, which is not ideal in cases where the retrieval problem focuses on a class of interest. In this paper, we propose a regularized loss for Variational Auto-Encoders (VAEs) forcing the model to focus on a given class of interest. As a result, the model learns to discriminate the data belonging to the class of interest from any other possibility, making the learnt latent space of the VAE suitable for class-specific retrieval tasks. The proposed Class-Specific Variational Auto-Encoder (CS-VAE) is evaluated on three public and one custom datasets, and its performance is compared with that of three related VAE-based methods. Experimental results show that the proposed method outperforms its competition in both in-domain and out-of-domain retrieval problems.
|
['Alexandros Iosifidis', 'Mehdi Rafiei']
|
2023-04-23
| null | null | null | null |
['content-based-image-retrieval']
|
['computer-vision']
|
[ 3.45087200e-02 -4.84810024e-01 -4.77340609e-01 -4.27082032e-01
-1.42286551e+00 -3.92748594e-01 9.03777480e-01 1.84932221e-02
-4.70679373e-01 5.96585333e-01 6.86930865e-02 2.45293751e-01
-4.36689556e-01 -6.48348749e-01 -5.53601265e-01 -9.99609947e-01
3.17896336e-01 1.02243507e+00 -3.56936753e-02 -2.92321946e-02
1.76026046e-01 4.33442742e-01 -1.81272900e+00 5.79421699e-01
5.59311271e-01 1.30452955e+00 3.86482745e-01 -5.01170903e-02
-1.28760785e-01 6.82071269e-01 -5.57857752e-01 -1.16506308e-01
2.72879392e-01 -2.84132332e-01 -6.93993032e-01 1.35097250e-01
5.38664818e-01 -2.58013874e-01 -5.15308201e-01 8.27586293e-01
6.07019246e-01 3.98495585e-01 1.26584613e+00 -1.17747986e+00
-7.84707367e-01 1.77371025e-01 -3.39350313e-01 1.44609094e-01
-7.46612325e-02 -4.37642753e-01 1.53895640e+00 -1.20842099e+00
9.41532731e-01 1.10974610e+00 -9.91540328e-02 6.23414993e-01
-1.10106742e+00 -7.34155774e-01 8.02464858e-02 4.53940988e-01
-1.85274136e+00 -2.61947662e-01 1.10536599e+00 -4.52976972e-01
6.64685607e-01 1.27456645e-02 2.80325830e-01 1.24750698e+00
7.40660578e-02 1.30770648e+00 1.03857946e+00 -2.94087321e-01
3.23895425e-01 4.59329635e-01 9.63803679e-02 3.54982585e-01
-1.01218365e-01 -1.51769400e-01 -4.81887847e-01 -2.45712712e-01
5.41016996e-01 1.65257454e-01 -3.03317666e-01 -9.07473028e-01
-7.98670888e-01 1.13143361e+00 5.41149795e-01 5.16809165e-01
-4.15373504e-01 -1.06200181e-01 4.18515831e-01 2.83721149e-01
7.39105761e-01 1.63995072e-01 -3.45638186e-01 5.68735421e-01
-1.33988929e+00 1.56932548e-01 3.09845269e-01 8.02605152e-01
7.56424963e-01 -1.98888615e-01 -6.70504272e-01 1.21317995e+00
6.13576829e-01 4.52681720e-01 7.35655665e-01 -4.72371280e-01
5.40025793e-02 5.22803545e-01 1.63694713e-02 -8.95335317e-01
1.29591897e-01 -6.38791919e-01 -7.23075688e-01 -1.12820350e-01
1.34028466e-02 6.14669323e-01 -1.22086179e+00 1.67798376e+00
7.17136264e-02 1.47953471e-02 1.01240113e-01 1.13162100e+00
1.08431864e+00 9.01733458e-01 4.31729853e-02 -3.04233074e-01
9.09544945e-01 -1.12954938e+00 -7.50808179e-01 5.08410856e-03
3.86132568e-01 -7.03659296e-01 9.31796372e-01 3.18980813e-01
-7.45166004e-01 -5.90277553e-01 -8.30485106e-01 -3.02305073e-01
-4.53955203e-01 4.06609535e-01 3.95434141e-01 2.60199428e-01
-9.73498940e-01 1.07124366e-01 -4.95805919e-01 2.11448818e-02
4.91214335e-01 1.77005053e-01 -3.12001407e-01 -3.98164153e-01
-1.31547713e+00 6.12778068e-01 3.14782232e-01 -3.73178385e-02
-1.35792923e+00 -4.08186436e-01 -6.82057798e-01 4.77610320e-01
2.69667625e-01 -4.87533689e-01 7.78723538e-01 -1.05982935e+00
-1.10972583e+00 1.25777745e+00 -2.47800335e-01 -2.85534590e-01
4.94531065e-01 -8.56519565e-02 -3.51601511e-01 3.05506349e-01
2.34553695e-01 7.88777947e-01 1.29986846e+00 -1.46790075e+00
-2.82569706e-01 -2.60152310e-01 7.40492269e-02 1.10164843e-01
-5.55859506e-01 -1.74157232e-01 -7.79648840e-01 -5.13414860e-01
5.09201661e-02 -8.58323693e-01 3.48878652e-01 2.32185632e-01
-2.90335298e-01 -8.61280501e-01 9.84808385e-01 -3.75907630e-01
1.01240194e+00 -2.22629476e+00 3.44237000e-01 2.49638900e-01
-2.46394370e-02 5.21476269e-01 -3.21925551e-01 4.39311296e-01
-1.00720124e-02 2.22119950e-02 4.87502441e-02 -6.09469056e-01
3.25556334e-06 3.50914508e-01 -5.47984242e-01 4.45136636e-01
1.89546034e-01 7.76549518e-01 -7.36478925e-01 -9.11453187e-01
2.89558798e-01 8.80162895e-01 -5.03263295e-01 2.93378621e-01
-3.41951400e-01 2.80301392e-01 -6.95442080e-01 6.82405889e-01
6.53146863e-01 -6.12547576e-01 1.18139587e-01 -1.99711874e-01
4.33635682e-01 2.15038583e-01 -7.99923122e-01 1.79558492e+00
-6.51766479e-01 7.16958284e-01 -2.85851330e-01 -1.20419228e+00
7.94539869e-01 3.17656338e-01 5.71546912e-01 -8.84302437e-01
3.45608294e-02 3.23941857e-01 -3.30074251e-01 -2.70721525e-01
5.29697180e-01 -7.55290762e-02 6.29990101e-02 2.67365158e-01
5.04834235e-01 1.87818840e-01 1.83312237e-01 4.01873976e-01
5.42840600e-01 2.66649932e-01 -1.49347810e-02 -2.53507286e-01
6.17521465e-01 -1.45135880e-01 4.68106717e-01 8.51279378e-01
-1.37779221e-01 8.10335577e-01 5.70033751e-02 -2.45784819e-01
-8.84972274e-01 -1.11298287e+00 -4.92259234e-01 1.02500880e+00
3.66433769e-01 -2.76966453e-01 -2.16868624e-01 -8.14866483e-01
-1.47669196e-01 6.39999926e-01 -7.17118323e-01 -2.12032035e-01
-1.31380260e-01 -3.57459277e-01 -5.50580695e-02 2.17783876e-04
4.68074739e-01 -1.20574105e+00 -2.27234334e-01 -5.90640567e-02
-4.76132751e-01 -8.63948524e-01 -3.47123086e-01 -7.86805674e-02
-5.19580603e-01 -9.82345045e-01 -1.21595132e+00 -9.73045945e-01
5.72434187e-01 6.09105527e-01 1.13952827e+00 -3.68148163e-02
-2.71811366e-01 7.92544007e-01 -5.71794868e-01 -1.18090719e-01
-5.80792651e-02 6.72325939e-02 -1.45683914e-01 4.01403308e-01
6.29797459e-01 -5.35823964e-02 -7.58067489e-01 2.99181372e-01
-1.19496524e+00 -4.91163701e-01 4.50951964e-01 1.27678442e+00
1.08510482e+00 7.71002099e-02 4.27930504e-01 -6.54618263e-01
3.17292720e-01 -7.58729339e-01 -5.03448129e-01 7.02110648e-01
-6.91160858e-01 1.16033137e-01 2.40548462e-01 -5.10016739e-01
-1.09092963e+00 -2.17876479e-01 2.22376361e-01 -9.25296247e-01
1.01066224e-01 8.17829132e-01 -1.29133791e-01 3.32472205e-01
3.90232712e-01 5.27290404e-01 -2.83996135e-01 -5.86347580e-01
2.05162894e-02 7.69047797e-01 -1.50493771e-01 -5.69733262e-01
4.81922865e-01 3.91326368e-01 -8.64531398e-02 -7.74096727e-01
-1.02070248e+00 -9.23447371e-01 -2.90442020e-01 -1.79286942e-01
7.62113929e-01 -1.31988180e+00 -3.46977301e-02 2.57291883e-01
-9.54748213e-01 8.00356492e-02 -2.26086259e-01 3.78941655e-01
-3.48192871e-01 2.55864382e-01 -2.70437658e-01 -7.72915542e-01
-2.52496302e-01 -1.41229904e+00 1.54566669e+00 1.03257120e-01
9.16477144e-02 -1.03202069e+00 1.07536584e-01 6.01403236e-01
3.63885015e-01 -2.55239159e-01 9.05541956e-01 -1.14806271e+00
-7.51278341e-01 -4.25303996e-01 -2.09907904e-01 4.72279936e-01
1.20029815e-01 -1.35115951e-01 -1.13236499e+00 -6.64175034e-01
-2.11959988e-01 -8.42169106e-01 1.27416885e+00 4.34652269e-01
1.44274032e+00 -7.19000623e-02 -4.59593475e-01 2.20462218e-01
1.56307006e+00 -4.70480211e-02 7.63689637e-01 3.12131822e-01
3.46886426e-01 4.74398941e-01 8.20828795e-01 2.98860431e-01
6.62500411e-02 1.03606248e+00 4.47603613e-01 -3.86452936e-02
-1.76757872e-01 -2.03854367e-01 1.42141059e-01 5.65599263e-01
2.07812279e-01 -7.26491690e-01 -5.60421407e-01 8.79179358e-01
-1.72497594e+00 -9.17425334e-01 3.69470030e-01 2.18437839e+00
7.55773425e-01 -2.70723790e-01 -2.69351333e-01 -1.08229272e-01
6.10153794e-01 4.52464998e-01 -5.19808590e-01 7.44396150e-02
-2.77590007e-01 4.17448163e-01 -2.58969590e-02 3.13971162e-01
-1.24537170e+00 9.44275200e-01 5.22942257e+00 1.54950106e+00
-1.29049492e+00 2.50653177e-01 5.60622692e-01 -2.12852150e-01
-6.08744264e-01 -1.63805157e-01 -7.25229502e-01 3.71907413e-01
5.24807751e-01 1.59427512e-03 7.80074100e-04 8.77197444e-01
-2.06354320e-01 1.31265238e-01 -1.11545086e+00 1.06675601e+00
2.74977773e-01 -1.00086045e+00 6.41740739e-01 1.08052261e-01
7.75052547e-01 1.22230463e-01 5.88530481e-01 5.12377858e-01
-1.00499049e-01 -8.78855467e-01 5.23883045e-01 5.38813531e-01
8.18949580e-01 -6.30207419e-01 8.09869349e-01 3.04012001e-01
-6.56994820e-01 1.01118289e-01 -5.86158097e-01 5.46741843e-01
-1.62887529e-01 4.82971281e-01 -7.19992399e-01 4.83652264e-01
7.11392462e-01 7.77769148e-01 -4.91621196e-01 9.42543745e-01
-1.51384816e-01 6.37265384e-01 -1.01770535e-01 1.94018092e-02
4.15709436e-01 -9.59937871e-02 5.49555659e-01 1.01337922e+00
2.65467137e-01 -1.73476279e-01 4.79057394e-02 8.25353146e-01
-2.96487182e-01 3.86351138e-01 -5.15872777e-01 -2.77146727e-01
3.14221382e-01 1.07036507e+00 -3.71848106e-01 -4.34769154e-01
-1.48724779e-01 1.04390991e+00 3.51591319e-01 5.81104815e-01
-5.23588359e-01 -1.32548064e-01 4.00011778e-01 -8.33847448e-02
6.11181557e-01 1.14116281e-01 3.89068902e-01 -1.47342443e+00
3.79766151e-02 -6.97823942e-01 5.98012984e-01 -9.21878636e-01
-1.47223926e+00 6.39943361e-01 8.97102281e-02 -1.59248900e+00
-3.27859104e-01 -5.71530521e-01 1.11357309e-01 9.03358281e-01
-1.98826778e+00 -1.37874424e+00 2.07191966e-02 7.90647030e-01
7.78417766e-01 -5.90677261e-01 1.03559971e+00 5.72607756e-01
-1.20262004e-01 8.06836486e-01 7.54661620e-01 6.94522113e-02
1.01369488e+00 -7.50370383e-01 -5.98513126e-01 5.18494546e-01
7.14976668e-01 7.31794775e-01 5.01911461e-01 -2.90163606e-01
-1.06909013e+00 -9.37836766e-01 1.11086965e+00 -1.48754239e-01
3.24654102e-01 -3.33335012e-01 -9.71943676e-01 5.19877791e-01
1.93139866e-01 3.38781834e-01 9.37257409e-01 2.32424755e-02
-7.76965857e-01 -7.44327679e-02 -1.17770576e+00 1.04204156e-01
5.33688664e-01 -8.98802876e-01 -4.29889053e-01 6.65451765e-01
3.83881271e-01 -1.46389604e-01 -6.44651771e-01 3.74972612e-01
7.23756254e-01 -7.41930425e-01 1.17079079e+00 -8.13559055e-01
8.19190919e-01 -9.67031345e-02 -5.16019881e-01 -1.25309980e+00
-3.24308395e-01 1.88906372e-01 -7.89336860e-03 1.17388070e+00
1.90316364e-01 -3.51308763e-01 5.99185169e-01 3.24115038e-01
7.50820562e-02 -5.71194112e-01 -1.01023293e+00 -7.86970258e-01
2.52174020e-01 -3.55284773e-02 2.55281955e-01 1.11852443e+00
-5.03239632e-01 2.61791199e-01 -5.32817721e-01 4.99685518e-02
7.09253013e-01 5.28432906e-01 4.98454034e-01 -1.30669916e+00
-3.52353841e-01 -2.34196141e-01 -4.35372591e-01 -1.18518734e+00
5.53458035e-01 -1.13532615e+00 -8.63574818e-02 -1.50794387e+00
6.11923754e-01 -5.87877214e-01 -9.59682822e-01 3.78814161e-01
-9.47756320e-02 3.61426741e-01 1.73124343e-01 6.00330055e-01
-8.48312438e-01 8.93068016e-01 1.16575587e+00 -8.06447148e-01
3.10712494e-02 -8.75133872e-02 -2.80920088e-01 1.87147066e-01
5.26230097e-01 -6.46268845e-01 -6.69341087e-01 -4.91029143e-01
6.50436133e-02 7.27751777e-02 3.92781258e-01 -5.09082556e-01
1.32293463e-01 1.53191360e-02 3.68474573e-01 -7.80107498e-01
6.68535829e-01 -1.00252914e+00 -6.66309074e-02 3.71127091e-02
-7.67220199e-01 -7.31944621e-01 -8.43473300e-02 7.36967862e-01
-8.14573050e-01 -5.07627070e-01 5.97014427e-01 2.84488313e-02
-8.77189517e-01 5.59466898e-01 -2.38346815e-01 8.24736580e-02
6.70580089e-01 7.76501447e-02 -6.94390684e-02 -7.04737484e-01
-7.75982618e-01 2.29599640e-01 2.31938630e-01 6.37011886e-01
8.65744472e-01 -1.43504822e+00 -7.43898392e-01 9.28299874e-02
8.24014246e-01 -1.87578425e-01 4.43740457e-01 2.71645308e-01
-1.32186353e-01 7.08805323e-01 -1.14040144e-01 -8.57311249e-01
-1.35268223e+00 8.28104734e-01 2.15737671e-01 -5.97424090e-01
-3.86710852e-01 1.03760350e+00 5.35381436e-01 -2.36998677e-01
1.84870914e-01 3.71960163e-01 -4.31223631e-01 2.67086744e-01
3.70786935e-01 -8.57217163e-02 1.73036069e-01 -9.46000934e-01
-2.91340292e-01 4.64975387e-01 -5.69787741e-01 -1.89398020e-03
1.27639675e+00 -1.35768458e-01 -4.79045026e-02 4.61912245e-01
1.92253399e+00 -3.31336558e-01 -9.14476097e-01 -8.64294827e-01
-2.64133781e-01 -6.46745443e-01 4.89850432e-01 -6.86329782e-01
-1.29032898e+00 1.17178023e+00 8.75420570e-01 -1.27313346e-01
1.20127416e+00 1.23523399e-01 6.00282371e-01 4.15273339e-01
5.55217624e-01 -1.09232104e+00 2.57427305e-01 2.62404948e-01
1.03490829e+00 -1.52398074e+00 5.73050715e-02 -1.23114951e-01
-5.41506350e-01 8.11968982e-01 2.60354280e-01 -2.93756396e-01
9.52953219e-01 -8.03817511e-01 -3.47576626e-02 -3.99930835e-01
-6.40423059e-01 -1.57593831e-01 8.03232670e-01 3.75542045e-01
3.44746083e-01 -1.34960711e-01 -4.16835546e-01 1.69094875e-01
3.95430744e-01 -1.14746310e-01 6.31094202e-02 8.85517120e-01
-1.29892066e-01 -1.22615647e+00 -8.86971727e-02 4.28606302e-01
-5.47965407e-01 -1.15997799e-01 -4.39053982e-01 7.43170083e-01
-9.54255015e-02 8.66709352e-01 2.01625258e-01 7.41105601e-02
-2.59981543e-01 2.74492264e-01 5.09295285e-01 -5.54285049e-01
-1.46201864e-01 5.34752071e-01 -2.58462518e-01 -3.66586685e-01
-8.53288889e-01 -5.85260093e-01 -6.75742388e-01 4.25483435e-01
-6.08349502e-01 3.58381629e-01 6.22278869e-01 8.85876358e-01
3.63908559e-01 4.10211295e-01 8.26755583e-01 -6.69708490e-01
-5.62786341e-01 -8.52201819e-01 -8.00964057e-01 6.55657947e-01
4.04120505e-01 -1.13105476e+00 -5.22530258e-01 -1.20855570e-01]
|
[11.129494667053223, 1.0602718591690063]
|
aa89f729-4f69-43dc-bd92-92d07a8669f3
|
towards-open-intent-discovery-for
|
1904.08524
| null |
http://arxiv.org/abs/1904.08524v1
|
http://arxiv.org/pdf/1904.08524v1.pdf
|
Towards Open Intent Discovery for Conversational Text
|
Detecting and identifying user intent from text, both written and spoken,
plays an important role in modelling and understand dialogs. Existing research
for intent discovery model it as a classification task with a predefined set of
known categories. To generailze beyond these preexisting classes, we define a
new task of \textit{open intent discovery}. We investigate how intent can be
generalized to those not seen during training. To this end, we propose a
two-stage approach to this task - predicting whether an utterance contains an
intent, and then tagging the intent in the input utterance. Our model consists
of a bidirectional LSTM with a CRF on top to capture contextual semantics,
subject to some constraints. Self-attention is used to learn long distance
dependencies. Further, we adapt an adversarial training approach to improve
robustness and perforamce across domains. We also present a dataset of 25k
real-life utterances that have been labelled via crowd sourcing. Our
experiments across different domains and real-world datasets show the
effectiveness of our approach, with less than 100 annotated examples needed per
unique domain to recognize diverse intents. The approach outperforms
state-of-the-art baselines by 5-15% F1 score points.
|
['Srinivasan Parthasarathy', 'Nikhita Vedula', 'Pranav Maneriker', 'Nedim Lipka']
|
2019-04-17
| null | null | null | null |
['open-intent-discovery', 'intent-discovery']
|
['natural-language-processing', 'natural-language-processing']
|
[ 2.96500951e-01 3.18924248e-01 -9.74889845e-02 -1.01526010e+00
-7.14023113e-01 -8.09109330e-01 8.45397472e-01 -1.16461933e-01
-4.83577460e-01 6.90130234e-01 6.47426486e-01 -3.58280599e-01
5.00945389e-01 -2.53242552e-01 -4.98336762e-01 -2.43487567e-01
1.08395524e-01 7.82570422e-01 1.23851858e-01 -3.70233655e-01
1.64979890e-01 9.17263236e-03 -1.03129625e+00 7.00014353e-01
6.29349113e-01 9.20618057e-01 2.41421759e-01 7.51898468e-01
-4.39472169e-01 1.27871048e+00 -9.54905689e-01 -3.92457664e-01
-1.67370513e-01 -2.22667426e-01 -1.46668470e+00 1.65780097e-01
6.59485906e-02 -4.42086071e-01 -4.00706194e-02 5.93417823e-01
3.53566170e-01 4.29348618e-01 6.49442136e-01 -1.24173522e+00
-6.73756301e-01 5.71188748e-01 5.41824400e-02 3.11522894e-02
7.95321405e-01 2.31080037e-02 1.00818336e+00 -7.80268729e-01
4.59353447e-01 1.35854280e+00 5.26310265e-01 9.87475157e-01
-1.01855826e+00 -5.61706781e-01 4.32515681e-01 -1.05100051e-01
-1.06889164e+00 -7.59587526e-01 7.59196639e-01 -5.02747178e-01
1.26006544e+00 2.89627552e-01 -2.08546028e-01 1.55731297e+00
-9.46927592e-02 1.01434934e+00 1.00101030e+00 -5.21654069e-01
3.65633130e-01 4.91089374e-01 6.44945681e-01 4.11209017e-01
-5.78170717e-01 -2.93266833e-01 -4.99449968e-01 -3.84941220e-01
2.23360196e-01 -6.66408762e-02 -1.40878737e-01 2.86063671e-01
-8.72641385e-01 1.01833820e+00 1.57624170e-01 4.25172716e-01
-1.02065988e-01 -2.04282090e-01 5.67936778e-01 4.01860505e-01
8.16249847e-01 4.18159753e-01 -7.97198355e-01 -1.67192578e-01
-3.83984089e-01 1.79537058e-01 1.36505079e+00 1.13001013e+00
6.92261636e-01 -2.29087397e-01 -2.04199940e-01 1.14153981e+00
4.54608023e-01 3.72162074e-01 8.01081538e-01 -7.03531682e-01
5.63565493e-01 7.14802623e-01 2.53099680e-01 -6.29012525e-01
-5.02239406e-01 8.98990557e-02 -4.49527830e-01 -4.02255237e-01
2.97116339e-01 -5.17056346e-01 -9.20351028e-01 1.78627694e+00
1.69269055e-01 2.09422767e-01 4.24466819e-01 6.22527540e-01
9.97101843e-01 7.59931922e-01 3.88303101e-01 -7.49503449e-02
1.53422904e+00 -1.01847911e+00 -7.80458152e-01 -6.99958384e-01
7.37377346e-01 -6.46918952e-01 1.19536686e+00 1.71431899e-01
-5.65059125e-01 -5.27518630e-01 -5.26388824e-01 -8.72542039e-02
-5.48260391e-01 -1.99034158e-03 3.75528365e-01 6.80861175e-01
-1.04325259e+00 -8.71875361e-02 -5.12963772e-01 -5.24450004e-01
1.37263164e-02 4.02509660e-01 -1.48775652e-01 1.18456753e-02
-1.41210759e+00 7.59573162e-01 5.01804292e-01 -3.74396265e-01
-7.62993455e-01 -4.30220127e-01 -1.10866797e+00 -2.31204256e-01
3.93238246e-01 -4.14678961e-01 1.76328325e+00 -1.15811455e+00
-1.71009076e+00 8.10140073e-01 -7.17050612e-01 -6.51231587e-01
6.36833608e-02 -3.79460245e-01 -5.64879358e-01 -1.23871952e-01
2.59538114e-01 6.54881060e-01 8.26707363e-01 -1.30256975e+00
-7.11467147e-01 -1.75120518e-01 4.23091948e-01 3.53419513e-01
-4.79688078e-01 4.68906939e-01 -3.51134688e-01 -4.49368477e-01
-2.71194905e-01 -9.82404590e-01 -2.00394824e-01 -7.37491012e-01
-4.58565831e-01 -6.60144150e-01 1.04034400e+00 -7.49448180e-01
1.13762283e+00 -1.99323118e+00 -1.74970329e-01 -1.71118394e-01
1.20244838e-01 2.50447392e-01 1.57091662e-03 5.08585274e-01
2.66303301e-01 1.96281537e-01 -2.44343013e-01 -8.55161846e-01
1.79598317e-01 5.22985518e-01 -7.87899613e-01 -1.78606305e-02
2.37848133e-01 8.57496917e-01 -7.23950863e-01 -2.06274152e-01
1.72840729e-01 3.31669033e-01 -4.39494550e-01 7.97281325e-01
-6.69845283e-01 7.30007172e-01 -5.32169044e-01 3.20594370e-01
2.88934618e-01 -3.19673955e-01 2.02036649e-01 2.06211224e-01
1.05376653e-01 8.64397109e-01 -7.77972758e-01 1.65695548e+00
-9.16795015e-01 4.31503057e-01 4.38390076e-02 -9.20148313e-01
1.02719283e+00 6.25478327e-01 1.52243618e-02 -3.87168109e-01
1.27082333e-01 1.21676624e-01 -2.42151871e-01 -5.83214700e-01
5.26809514e-01 -2.42510021e-01 -6.53931916e-01 7.16051340e-01
2.97326922e-01 4.96668778e-02 -1.44572303e-01 1.92998558e-01
1.15286279e+00 -2.83898234e-01 3.66033524e-01 -5.91253862e-02
6.99909389e-01 -4.86269295e-02 4.47060108e-01 8.30513597e-01
-2.28807002e-01 4.94539678e-01 3.81220669e-01 -6.15854919e-01
-4.54231739e-01 -5.22271454e-01 1.30414113e-01 1.86789250e+00
-6.74617440e-02 -5.62771410e-02 -6.36230469e-01 -1.24872649e+00
-2.58079886e-01 1.05606413e+00 -4.82106239e-01 9.57286358e-02
-6.59073174e-01 -4.22656536e-01 8.19436848e-01 4.17162567e-01
6.57944798e-01 -1.35449135e+00 -4.48634803e-01 2.36397177e-01
-5.29415429e-01 -1.48815620e+00 -5.47615945e-01 4.35325325e-01
-2.58817881e-01 -6.73555315e-01 -4.44272101e-01 -9.82565880e-01
4.11738157e-01 -8.53548944e-02 1.31314349e+00 -1.40152827e-01
3.06706965e-01 4.79057640e-01 -7.85326302e-01 -4.36404109e-01
-7.43140101e-01 2.25271881e-01 1.62272170e-01 2.11167693e-01
8.79292369e-01 -1.58759266e-01 -7.47867301e-02 4.00591731e-01
-7.85004795e-01 -8.77534077e-02 2.70163864e-01 8.36005270e-01
-2.60707755e-02 -1.61371276e-01 8.44569385e-01 -9.82579470e-01
9.03832138e-01 -7.73415446e-01 -1.97701156e-02 2.55823612e-01
-8.07163939e-02 7.02555925e-02 8.11778069e-01 -6.08288467e-01
-1.51196170e+00 1.84794411e-01 -4.70507056e-01 -1.12102479e-01
-8.02811503e-01 3.67168725e-01 -3.70758057e-01 2.01167047e-01
4.52561140e-01 2.66039759e-01 -1.93164438e-01 -4.89914507e-01
5.29705882e-01 1.38418245e+00 3.44565451e-01 -5.57415426e-01
3.24663371e-01 2.88621932e-01 -8.82889628e-01 -1.07883322e+00
-1.22086799e+00 -9.34311807e-01 -6.77386701e-01 -3.15633006e-02
1.00001287e+00 -9.49955940e-01 -6.94340050e-01 3.92149955e-01
-1.53744388e+00 -5.91420650e-01 -1.58713199e-02 1.83242962e-01
-3.74229699e-01 1.86760485e-01 -7.67446101e-01 -1.09136868e+00
-4.53217328e-01 -1.11682475e+00 1.18897283e+00 1.54778957e-01
-7.71543860e-01 -1.37999463e+00 6.78047761e-02 6.22792423e-01
4.37772214e-01 -3.23469758e-01 5.46017647e-01 -1.78200972e+00
-3.90987992e-02 -1.20748440e-02 2.12612301e-02 3.92920554e-01
3.19663107e-01 -7.08500981e-01 -1.33932579e+00 -8.77123028e-02
4.98412848e-01 -8.26770544e-01 6.11303091e-01 -3.41351070e-02
7.00401366e-01 -8.04968059e-01 -4.07449186e-01 1.86220258e-01
8.92113626e-01 6.70095384e-01 3.24246168e-01 1.31242827e-01
6.26533210e-01 9.26078022e-01 5.82095802e-01 3.34626824e-01
7.13240743e-01 6.75496280e-01 5.92100322e-02 -4.06819619e-02
1.40431479e-01 -2.44675398e-01 5.35220921e-01 6.63045585e-01
6.31885409e-01 -6.57708168e-01 -1.08729303e+00 5.37930965e-01
-1.73792696e+00 -8.65996182e-01 2.64444828e-01 1.74359941e+00
1.04454887e+00 1.61861420e-01 1.27672926e-01 -3.11589360e-01
5.89359105e-01 2.86907703e-01 -4.18982446e-01 -6.70879602e-01
2.23606586e-01 -4.10373835e-03 1.16140015e-01 8.26176405e-01
-1.38391566e+00 1.25308979e+00 6.23415422e+00 4.55860376e-01
-1.14373481e+00 3.11423123e-01 1.00028062e+00 4.65471417e-01
-2.79196560e-01 -1.30567387e-01 -1.17772961e+00 5.21996021e-01
1.14419544e+00 2.31570557e-01 1.41470209e-01 1.04931366e+00
4.52866144e-02 1.10993423e-01 -1.10151052e+00 4.46965933e-01
5.39146841e-01 -9.37125444e-01 -1.94735348e-01 -2.52849370e-01
4.78841096e-01 8.57519731e-02 -1.39130265e-01 7.67991662e-01
8.24243963e-01 -1.03025901e+00 4.02799338e-01 2.43614480e-01
6.65971637e-01 -3.83881032e-01 8.22490394e-01 7.35765576e-01
-1.05143034e+00 -3.99279296e-02 -1.28912017e-01 -3.73857200e-01
3.00042480e-01 1.80771336e-01 -1.49525845e+00 2.50942677e-01
4.09420401e-01 6.22946978e-01 -2.33550012e-01 2.84639895e-01
-2.52017558e-01 8.27469826e-01 -3.79406959e-01 -3.03564519e-01
2.85708219e-01 1.32116631e-01 4.64609355e-01 1.52068961e+00
-2.24370226e-01 1.80302024e-01 6.32963359e-01 6.73625350e-01
-2.04193398e-01 2.54647974e-02 -7.74219275e-01 1.60136789e-01
4.71709907e-01 9.18318868e-01 -5.09080231e-01 -4.45526063e-01
-6.35933638e-01 1.28961957e+00 3.59470040e-01 4.36981022e-01
-6.83339119e-01 -2.92112768e-01 6.52990818e-01 -2.60657996e-01
1.04408406e-01 -1.33976310e-01 -1.94901839e-01 -1.09764910e+00
-3.87671664e-02 -9.74619567e-01 5.00694573e-01 -4.97014016e-01
-1.54600692e+00 1.00239825e+00 8.25593844e-02 -7.93192863e-01
-6.83242142e-01 -5.00221074e-01 -8.05900753e-01 9.05356228e-01
-1.37423646e+00 -1.26858222e+00 -8.07792842e-02 4.67177451e-01
1.30769300e+00 -3.03154022e-01 1.06651556e+00 2.69754417e-02
-4.55019325e-01 5.18970966e-01 -2.06982940e-01 4.66924608e-01
8.39388728e-01 -1.25216568e+00 8.22415113e-01 6.98420882e-01
1.94248751e-01 8.23951364e-01 7.12687373e-01 -7.71333516e-01
-9.42360163e-01 -1.30254233e+00 1.31513846e+00 -9.07730222e-01
6.58944309e-01 -8.14483523e-01 -1.03891230e+00 1.17423987e+00
5.23733318e-01 -1.23749442e-01 9.04736936e-01 2.70655990e-01
-4.48874354e-01 3.87848169e-01 -1.15902126e+00 4.04459625e-01
8.92801702e-01 -5.69305420e-01 -1.12287688e+00 4.10759836e-01
1.38636625e+00 -3.71613234e-01 -5.72473526e-01 2.75410622e-01
2.62083262e-01 -7.35796571e-01 8.37157071e-01 -7.66118646e-01
9.14611220e-02 5.79310134e-02 -2.85212278e-01 -1.12950826e+00
1.40903309e-01 -7.97220290e-01 2.85448437e-03 1.54599321e+00
5.23694217e-01 -7.08817065e-01 6.48067951e-01 9.86114740e-01
-2.83157110e-01 -4.36762899e-01 -8.17718029e-01 -6.38476729e-01
2.49873791e-02 -6.94302499e-01 4.85513628e-01 9.91403997e-01
3.39354903e-01 1.02903605e+00 -5.26576340e-01 1.61284670e-01
1.26971245e-01 -8.10122415e-02 8.25649977e-01 -1.00051928e+00
-2.80061483e-01 5.22994660e-02 -6.86228797e-02 -1.74734986e+00
8.00432026e-01 -6.14158511e-01 4.02125210e-01 -1.36976945e+00
-6.71821609e-02 -6.30988777e-01 1.46102652e-01 8.02775145e-01
-2.66321123e-01 -6.20387914e-03 6.12535886e-03 3.11130583e-01
-9.38612938e-01 5.70872962e-01 4.84029353e-01 -2.05248177e-01
-5.46904981e-01 3.10327083e-01 -6.28384352e-01 8.78434181e-01
9.24511194e-01 -4.64110196e-01 -5.77134728e-01 -3.14500898e-01
-4.49386686e-01 9.31691155e-02 -4.19856310e-02 -8.60716045e-01
1.85106456e-01 -2.54866719e-01 -8.74857232e-02 -4.18739170e-01
5.81922650e-01 -7.77353704e-01 -3.59321743e-01 2.48958930e-01
-8.39072526e-01 -2.57308900e-01 2.03322425e-01 5.01476765e-01
-2.61788547e-01 -4.80013430e-01 3.72463793e-01 -1.68274581e-01
-9.66031015e-01 -1.42360106e-01 -6.29098177e-01 3.56563956e-01
9.50982988e-01 2.09608465e-01 -3.70999634e-01 -7.63509095e-01
-6.27349734e-01 3.26617181e-01 2.04682663e-01 7.06711233e-01
4.32132840e-01 -8.39031875e-01 -4.77105826e-01 3.17618608e-01
2.69977421e-01 -1.45985633e-01 -4.64636041e-03 2.48993009e-01
7.83136860e-02 6.78297937e-01 4.34885353e-01 -5.86639524e-01
-1.39098310e+00 4.11566585e-01 1.95260346e-01 -4.69065964e-01
-3.59265000e-01 1.10108924e+00 4.43971336e-01 -1.00900567e+00
5.77175438e-01 -2.93651909e-01 -5.33022761e-01 -1.03541300e-01
7.31788576e-01 -1.15568928e-01 -4.14384864e-02 -9.61929023e-01
-5.69787979e-01 5.90158328e-02 -3.79617214e-01 -3.44996244e-01
8.92575145e-01 -4.25309092e-01 1.33163363e-01 7.17974961e-01
1.35241103e+00 -4.95244488e-02 -1.05524385e+00 -4.55893338e-01
2.90207028e-01 -1.37024438e-02 -4.44508642e-01 -9.97472703e-01
-3.42576146e-01 6.31193936e-01 3.51673841e-01 6.09458506e-01
9.25685465e-01 3.63164723e-01 9.58268762e-01 5.80104411e-01
3.31520200e-01 -8.07208180e-01 3.84674758e-01 1.18585801e+00
9.48502541e-01 -1.65710461e+00 -6.51127398e-01 -3.47371012e-01
-1.29475176e+00 7.96652198e-01 6.59901798e-01 1.58987269e-01
5.41209459e-01 3.75081927e-01 5.74084818e-01 -4.12980281e-02
-9.03731883e-01 -2.32171699e-01 2.71265745e-01 6.29049003e-01
6.07045710e-01 -5.95709356e-03 6.63394332e-02 9.30029035e-01
-1.56784341e-01 -2.19139159e-01 3.24965179e-01 9.80465114e-01
-5.45948446e-01 -1.12296283e+00 -2.27809235e-01 3.54880631e-01
-6.19685590e-01 -2.51493931e-01 -6.83587730e-01 4.63975251e-01
-1.57065883e-01 1.50487339e+00 2.86870524e-02 -6.55081213e-01
2.38886014e-01 6.67701304e-01 -4.39711422e-01 -1.06460929e+00
-7.85218060e-01 -1.25101000e-01 5.49656034e-01 -3.74785244e-01
-4.60794061e-01 -5.06644964e-01 -1.21898222e+00 1.37560621e-01
-2.36329168e-01 4.03943270e-01 5.09841084e-01 1.38582611e+00
3.72151524e-01 3.36400360e-01 6.80079043e-01 -5.25157213e-01
-5.13041556e-01 -1.25326622e+00 6.45967107e-03 5.59713960e-01
4.60145712e-01 -3.92954618e-01 -5.08308470e-01 4.12527651e-01]
|
[12.548897743225098, 7.620955944061279]
|
92245db0-7640-4dee-b40b-9d648aea3bb7
|
improving-performance-insensitivity-of-large
|
2304.04071
| null |
https://arxiv.org/abs/2304.04071v2
|
https://arxiv.org/pdf/2304.04071v2.pdf
|
Improving Performance Insensitivity of Large-scale Multiobjective Optimization via Monte Carlo Tree Search
|
The large-scale multiobjective optimization problem (LSMOP) is characterized by simultaneously optimizing multiple conflicting objectives and involving hundreds of decision variables. Many real-world applications in engineering fields can be modeled as LSMOPs; simultaneously, engineering applications require insensitivity in performance. This requirement usually means that the results from the algorithm runs should not only be good for every run in terms of performance but also that the performance of multiple runs should not fluctuate too much, i.e., the algorithm shows good insensitivity. Considering that substantial computational resources are requested for each run, it is essential to improve upon the performance of the large-scale multiobjective optimization algorithm, as well as the insensitivity of the algorithm. However, existing large-scale multiobjective optimization algorithms solely focus on improving the performance of the algorithms, leaving the insensitivity characteristics unattended. In this work, we propose an evolutionary algorithm for solving LSMOPs based on Monte Carlo tree search, the so-called LMMOCTS, which aims to improve the performance and insensitivity for large-scale multiobjective optimization problems. The proposed method samples the decision variables to construct new nodes on the Monte Carlo tree for optimization and evaluation. It selects nodes with good evaluation for further search to reduce the performance sensitivity caused by large-scale decision variables. We compare the proposed algorithm with several state-of-the-art designs on different benchmark functions. We also propose two metrics to measure the sensitivity of the algorithm. The experimental results confirm the effectiveness and performance insensitivity of the proposed design for solving large-scale multiobjective optimization problems.
|
['Gary G. Yen', 'Min Jiang', 'Haokai Hong']
|
2023-04-08
| null | null | null | null |
['multiobjective-optimization']
|
['methodology']
|
[ 3.64026916e-03 -4.74879205e-01 4.66030613e-02 -7.33411387e-02
-3.61964971e-01 -3.69864732e-01 -1.69324040e-01 1.25561401e-01
-6.95229992e-02 1.02109826e+00 -1.80900097e-01 -1.17885232e-01
-8.83891284e-01 -1.01980543e+00 -5.54637432e-01 -1.02290750e+00
-8.61572847e-02 4.43164170e-01 1.95198223e-01 -4.09184605e-01
4.75486606e-01 4.29666340e-01 -1.78644633e+00 -9.85767841e-02
1.45052719e+00 1.04358447e+00 2.52325982e-01 4.14201409e-01
2.57581860e-01 -2.86421296e-03 -9.56846774e-01 1.69069972e-02
1.38636276e-01 -5.44134736e-01 -3.73926431e-01 -3.80468853e-02
-3.29397470e-01 1.45583212e-01 4.55719769e-01 1.11879170e+00
8.40368569e-01 3.02768618e-01 3.61955196e-01 -1.38342059e+00
-4.19180185e-01 5.58165252e-01 -7.60875881e-01 -1.96085405e-02
1.05889635e-02 2.16124073e-01 1.00872505e+00 -4.85487878e-01
2.52377063e-01 1.13636827e+00 2.90317297e-01 -5.32606132e-02
-1.11547005e+00 -6.43246114e-01 1.82730034e-01 2.75228411e-01
-1.63542676e+00 -1.94545507e-01 8.83262455e-01 -2.80893803e-01
7.41940975e-01 7.04654694e-01 7.33427465e-01 2.06584752e-01
6.67266965e-01 3.12576145e-01 1.04477286e+00 -2.56956488e-01
6.91982925e-01 1.62031152e-04 -7.89311826e-02 4.78135586e-01
7.38222837e-01 4.61290628e-01 -2.89338350e-01 -1.41288474e-01
2.58747369e-01 -1.71693295e-01 -1.31021708e-01 -4.79263723e-01
-9.54853952e-01 7.69959033e-01 3.70834827e-01 3.61801326e-01
-4.99193043e-01 8.26509669e-02 2.22897381e-01 9.27805305e-02
4.89274524e-02 7.43168950e-01 -4.95144635e-01 8.03403836e-03
-7.70459116e-01 3.00447911e-01 5.69088578e-01 6.65609777e-01
5.45969963e-01 2.60725051e-01 -2.74678200e-01 8.52525055e-01
3.17090452e-01 4.56375718e-01 4.24454838e-01 -4.78577971e-01
3.45931232e-01 8.73437047e-01 2.05097586e-01 -1.20829225e+00
-6.01263702e-01 -8.68836105e-01 -6.91476107e-01 5.17575383e-01
9.12884809e-03 -4.28420126e-01 -5.72905242e-01 1.57582664e+00
4.89529490e-01 -2.73093522e-01 -7.31819421e-02 9.78351533e-01
3.76309514e-01 9.69734550e-01 -7.85354823e-02 -7.43765175e-01
1.13716102e+00 -7.75036991e-01 -7.39578307e-01 -3.92659903e-02
4.24631238e-01 -8.30871940e-01 8.87100756e-01 3.38755429e-01
-1.00832713e+00 -3.95679742e-01 -1.37894726e+00 9.91966367e-01
-1.44460052e-01 3.22486579e-01 4.75844413e-01 1.00267386e+00
-3.43150437e-01 5.86264014e-01 -6.14169240e-01 1.21755615e-01
9.97865424e-02 5.74691832e-01 3.33504677e-01 6.53871819e-02
-1.13097739e+00 8.56989086e-01 8.03732634e-01 5.52277029e-01
-7.45553732e-01 -7.28550076e-01 -4.23051924e-01 3.96621943e-01
8.55969250e-01 -4.21493381e-01 6.36680007e-01 -6.38434291e-01
-1.67181766e+00 -3.28810140e-02 1.32228294e-02 1.93126664e-01
4.23487514e-01 2.19678238e-01 -6.91526830e-01 -3.72262597e-01
-1.37849808e-01 -1.91278547e-01 3.82128954e-01 -1.24283040e+00
-7.53762305e-01 -3.42269450e-01 7.93182030e-02 6.99937269e-02
-4.26494390e-01 3.55747230e-02 -7.71835297e-02 -6.12137437e-01
-1.01245299e-01 -9.13544893e-01 -4.90456820e-01 -6.84985876e-01
-5.82316279e-01 6.70079561e-03 7.29313135e-01 -4.20021623e-01
1.90056205e+00 -1.74279392e+00 4.70179409e-01 5.43727398e-01
-2.86865801e-01 1.62136599e-01 -1.58483192e-01 3.95202488e-01
-4.32045497e-02 2.49974445e-01 -2.35339761e-01 4.36321318e-01
1.08521506e-01 8.61284807e-02 4.67764616e-01 4.50067997e-01
2.70698569e-03 7.48354733e-01 -6.51244164e-01 -3.56919229e-01
3.17898661e-01 -3.25895175e-02 -6.51223242e-01 9.89042446e-02
-1.93934798e-01 1.49200737e-01 -8.67625535e-01 8.87075305e-01
6.79913819e-01 -1.97346628e-01 1.48672089e-01 -4.69251722e-01
-3.61477613e-01 -4.39007550e-01 -1.75271821e+00 1.07883012e+00
-7.30290830e-01 -4.04339284e-02 7.73967132e-02 -1.25486887e+00
9.12388027e-01 2.59411126e-01 6.96729839e-01 -5.42146921e-01
5.77094615e-01 4.43875194e-01 3.26019436e-01 -3.82475704e-01
3.65611494e-01 -1.95061535e-01 -2.89129108e-01 3.10880542e-01
-2.96706200e-01 -2.24122301e-01 6.43604994e-01 -4.63883132e-01
5.77435732e-01 -1.86523974e-01 5.15546620e-01 -6.15904331e-01
8.25696766e-01 1.24224328e-01 9.88341451e-01 2.35889375e-01
7.82848001e-02 7.09116831e-02 3.57799828e-01 -5.23064807e-02
-9.21494663e-01 -7.92714238e-01 -2.22643450e-01 8.02813709e-01
4.62537587e-01 -7.15497062e-02 -3.60424429e-01 -4.12249625e-01
-3.44877318e-02 7.97010720e-01 -3.47722322e-01 -5.66751420e-01
-5.73845267e-01 -1.50284815e+00 5.03168292e-02 1.74755692e-01
3.61814022e-01 -6.67895436e-01 -7.92206168e-01 4.72553462e-01
1.01519123e-01 -6.57606065e-01 -1.63031980e-01 4.00648825e-02
-7.68282115e-01 -9.96991634e-01 -5.58267057e-01 -6.43111765e-01
7.23468184e-01 -1.90246612e-01 9.33342755e-01 1.64458394e-01
-5.22521019e-01 -5.49849451e-01 -3.54391485e-01 -3.87861341e-01
-2.82669157e-01 -9.70562845e-02 1.00356050e-01 3.05694528e-02
-2.34503075e-01 -5.79708219e-01 -4.42183673e-01 9.49144900e-01
-7.83159077e-01 -4.44896340e-01 8.24608088e-01 9.77770746e-01
7.58396804e-01 1.13523769e+00 9.06695902e-01 -2.80857921e-01
8.27732801e-01 -3.08921576e-01 -1.20986092e+00 7.60831952e-01
-7.58857310e-01 3.67590606e-01 1.04493737e+00 -5.44778168e-01
-9.41635907e-01 -2.01461673e-01 7.37955049e-02 -2.38211423e-01
3.73254269e-01 7.58051991e-01 -5.51456571e-01 -4.34654623e-01
3.71021271e-01 2.34554503e-02 -2.76095867e-01 -3.01262528e-01
6.77549616e-02 4.67599511e-01 8.17331001e-02 -8.30941141e-01
8.67067218e-01 -8.53342786e-02 5.27555525e-01 -5.01793861e-01
-3.12531114e-01 7.75575079e-03 1.28245026e-01 -5.12993693e-01
5.87747753e-01 -2.61070877e-01 -1.23734677e+00 2.87419975e-01
-7.18809724e-01 3.93927217e-01 -1.10770673e-01 4.94549692e-01
-2.58374363e-01 4.95262481e-02 1.36189520e-01 -1.03881693e+00
-1.49106726e-01 -1.65795839e+00 5.29605031e-01 4.26045567e-01
-1.15690306e-01 -7.98058033e-01 -1.35623485e-01 2.61087239e-01
4.79782432e-01 6.93751931e-01 1.07536542e+00 -2.82380164e-01
-5.71928144e-01 -2.64991999e-01 3.09167743e-01 9.59029570e-02
2.20575392e-01 3.53477955e-01 -3.34846973e-01 -4.30433452e-01
1.92491021e-02 -1.75944179e-01 3.33164543e-01 7.61757553e-01
1.07814670e+00 -2.77472407e-01 -3.92434895e-01 4.56979454e-01
1.76285005e+00 7.77905881e-01 5.04930556e-01 4.04012561e-01
3.04595679e-01 7.19958007e-01 1.07762134e+00 8.27362061e-01
-1.46543756e-01 8.36401224e-01 5.35395980e-01 -3.12048122e-02
4.05433267e-01 1.22979194e-01 7.62999654e-02 6.80412591e-01
-1.41849235e-01 -6.48408294e-01 -7.25177884e-01 4.72978592e-01
-1.80400240e+00 -6.81496978e-01 -5.62119903e-03 2.25283027e+00
5.59764922e-01 5.52413315e-02 -4.32378799e-02 3.47395539e-01
1.13867664e+00 1.29122317e-01 -6.79242134e-01 -7.77211845e-01
-1.36089906e-01 1.30564883e-01 5.85092723e-01 1.99614406e-01
-6.66407645e-01 2.80104190e-01 5.95604420e+00 1.28595030e+00
-1.14594066e+00 -1.93123043e-01 7.02996671e-01 -5.03551245e-01
-2.97889352e-01 -1.56736039e-02 -7.31492519e-01 6.53763771e-01
6.67344928e-01 -8.95939112e-01 5.05661547e-01 7.54189909e-01
6.45085275e-01 -3.11077118e-01 -8.06823432e-01 6.49836302e-01
-3.23370755e-01 -1.18502200e+00 -2.28853449e-01 5.47942817e-02
1.23680067e+00 -6.94458723e-01 1.16632894e-01 3.24214026e-02
2.89647937e-01 -1.09306550e+00 6.52718246e-01 7.24819228e-02
3.40692550e-01 -1.29138017e+00 9.85559821e-01 2.54938185e-01
-1.43078685e+00 -4.45684284e-01 -2.87084907e-01 1.95188299e-01
5.74950576e-01 1.00156057e+00 -1.75956339e-01 1.04764211e+00
5.82034886e-01 2.00688273e-01 -2.58237451e-01 1.33564258e+00
-1.28772870e-01 2.54308879e-01 -4.20987219e-01 -6.15968168e-01
6.85640126e-02 -4.54381317e-01 8.33037496e-01 3.12069088e-01
7.14833081e-01 -7.29196751e-03 2.54167497e-01 9.64899421e-01
2.58606583e-01 3.97114903e-01 -2.27258597e-02 -3.10257643e-01
7.19688356e-01 9.40591276e-01 -5.89072347e-01 1.62128583e-02
-2.36807484e-02 3.17645848e-01 -3.81447613e-01 2.97634512e-01
-1.11228037e+00 -6.28239453e-01 5.42520046e-01 -2.80904025e-01
4.21545237e-01 1.01164483e-01 -7.09473491e-01 -6.06536567e-01
2.68264830e-01 -8.23790193e-01 3.96766394e-01 -3.94921154e-01
-9.19976115e-01 3.70430291e-01 -5.87004498e-02 -1.45031130e+00
-7.87285492e-02 -3.74454379e-01 -6.41148686e-01 8.90688241e-01
-1.32240379e+00 -7.15094805e-01 -3.42984982e-02 1.04998648e-01
3.71942312e-01 -1.92437023e-01 3.01621318e-01 3.36926550e-01
-1.06596243e+00 6.39506042e-01 4.76104140e-01 -7.01148748e-01
2.59557605e-01 -5.49871027e-01 -3.88375312e-01 1.00365829e+00
-7.07381248e-01 5.15585840e-01 1.18407369e+00 -6.85656548e-01
-1.55268908e+00 -7.74506032e-01 3.37786973e-01 3.11410517e-01
6.07914031e-01 1.13178212e-02 -4.41896439e-01 -1.49312124e-01
-2.10263416e-01 -2.99418658e-01 5.17878473e-01 3.14802043e-02
6.52994454e-01 -4.86992866e-01 -1.36322057e+00 6.32334173e-01
7.57653058e-01 2.89061695e-01 -2.49629110e-01 1.81816041e-01
5.54377139e-01 -2.85050631e-01 -1.21946108e+00 8.42959881e-01
6.14385307e-01 -6.80023074e-01 1.17235208e+00 -3.68372709e-01
6.57178342e-01 -5.77718198e-01 -1.91714391e-01 -1.66152644e+00
-3.95238847e-01 -4.08564061e-01 6.07327521e-02 1.33802414e+00
5.33642828e-01 -7.25280762e-01 5.64567387e-01 6.09724522e-01
-1.15267977e-01 -1.19681346e+00 -9.39474642e-01 -1.35152423e+00
-1.00363918e-01 1.24277130e-01 1.06528866e+00 6.43186986e-01
-1.67766631e-01 1.81466147e-01 -4.10931379e-01 4.69252467e-01
6.77374482e-01 7.22079992e-01 3.44117224e-01 -1.04939973e+00
-5.21306634e-01 -5.83170772e-01 -3.40038598e-01 -4.16232646e-01
-2.21861675e-01 -4.17772651e-01 9.25081968e-02 -1.44607556e+00
3.51100005e-02 -4.56904918e-01 -4.80036855e-01 6.43228590e-02
-4.66652036e-01 -1.17740296e-01 1.49181504e-02 -1.70885876e-01
-3.63947332e-01 9.41724777e-01 1.37653685e+00 -1.20409586e-01
-5.31981707e-01 1.28039330e-01 -6.70530200e-01 3.79595160e-01
6.74658895e-01 -5.27360559e-01 -4.74121869e-01 -1.18442222e-01
2.89114296e-01 3.36395264e-01 -1.71389431e-01 -1.06117558e+00
-6.78464770e-02 -7.80696809e-01 1.81969464e-01 -5.86024582e-01
-1.36165312e-02 -1.07783008e+00 7.38767266e-01 8.65215242e-01
5.99010959e-02 7.23922327e-02 2.75206089e-01 3.79963458e-01
-3.27741861e-01 -3.61159772e-01 1.01215923e+00 2.16966644e-01
-6.78364038e-01 2.87097488e-02 -2.07481220e-01 -3.56837250e-02
1.58106279e+00 -3.52550298e-01 -2.96173006e-01 2.23036230e-01
-4.57158834e-01 7.87896454e-01 4.38709587e-01 3.17324519e-01
3.39601964e-01 -1.40402365e+00 -6.82728887e-01 -1.99206457e-01
4.59057987e-02 -3.88181746e-01 4.28271443e-01 7.91147530e-01
-4.66150522e-01 4.20234412e-01 -3.19620758e-01 -3.94807100e-01
-1.30358946e+00 6.68272972e-01 3.13891411e-01 -4.16886270e-01
1.87968910e-01 7.84248710e-01 -5.16970046e-02 -2.98831314e-01
-2.67197877e-01 -1.13041095e-01 -1.82972863e-01 1.69239417e-01
1.54648989e-01 8.75733137e-01 2.46538714e-01 -1.99554771e-01
-6.95330262e-01 9.68620896e-01 4.48894083e-01 1.30044773e-01
1.48474073e+00 -7.37224445e-02 -3.04230720e-01 9.01508257e-02
1.10104525e+00 1.88412055e-01 -6.88098907e-01 3.27599525e-01
-4.50387508e-01 -7.26409018e-01 3.47460240e-01 -1.03132427e+00
-1.31719017e+00 3.22714388e-01 5.73453844e-01 -1.77309699e-02
1.66922987e+00 -5.28482974e-01 6.69778168e-01 2.89325476e-01
4.78996575e-01 -1.48428941e+00 -4.06614542e-02 3.07735745e-02
8.59610021e-01 -8.96696448e-01 4.47048426e-01 -4.86074984e-01
-4.99630183e-01 1.04405046e+00 6.01441979e-01 2.43138503e-02
4.58992958e-01 3.41722727e-01 -4.61918414e-01 -1.56471536e-01
-6.26550674e-01 1.56940714e-01 2.85378695e-01 6.74908385e-02
6.29841387e-02 1.47084638e-01 -1.05916214e+00 7.56870806e-01
-1.57427132e-01 -8.94602612e-02 2.40464509e-01 1.10955989e+00
-6.08150363e-01 -1.17641330e+00 -7.95304477e-01 3.78758848e-01
-1.77146047e-01 1.89590335e-01 -1.83251068e-01 6.01083875e-01
1.65255725e-01 1.24491167e+00 -3.73350739e-01 -5.56471825e-01
5.64390242e-01 -3.35260928e-01 1.79906726e-01 -8.53791535e-02
-6.50744736e-01 5.69454990e-02 3.52092236e-01 -5.11879444e-01
-2.14467540e-01 -4.29381967e-01 -1.17870367e+00 -5.31827986e-01
-9.72389102e-01 4.65899944e-01 7.17003345e-01 8.60047102e-01
2.18068138e-01 1.22315264e+00 1.14911747e+00 -3.81161094e-01
-6.36377335e-01 -5.23517668e-01 -6.04484916e-01 -9.20816511e-02
-7.06596226e-02 -1.12398732e+00 -3.37618828e-01 -5.25111139e-01]
|
[5.730721950531006, 3.5296080112457275]
|
883fbc79-4634-496c-a49d-76d41d04b42c
|
quaternion-matrix-completion-using-untrained
|
2305.00416
| null |
https://arxiv.org/abs/2305.00416v1
|
https://arxiv.org/pdf/2305.00416v1.pdf
|
Quaternion Matrix Completion Using Untrained Quaternion Convolutional Neural Network for Color Image Inpainting
|
The use of quaternions as a novel tool for color image representation has yielded impressive results in color image processing. By considering the color image as a unified entity rather than separate color space components, quaternions can effectively exploit the strong correlation among the RGB channels, leading to enhanced performance. Especially, color image inpainting tasks are highly beneficial from the application of quaternion matrix completion techniques, in recent years. However, existing quaternion matrix completion methods suffer from two major drawbacks. First, it can be difficult to choose a regularizer that captures the common characteristics of natural images, and sometimes the regularizer that is chosen based on empirical evidence may not be the optimal or efficient option. Second, the optimization process of quaternion matrix completion models is quite challenging because of the non-commutativity of quaternion multiplication. To address the two drawbacks of the existing quaternion matrix completion approaches mentioned above, this paper tends to use an untrained quaternion convolutional neural network (QCNN) to directly generate the completed quaternion matrix. This approach replaces the explicit regularization term in the quaternion matrix completion model with an implicit prior that is learned by the QCNN. Extensive quantitative and qualitative evaluations demonstrate the superiority of the proposed method for color image inpainting compared with some existing quaternion-based and tensor-based methods.
|
['Juan Han', 'Liqiao Yang', 'Kit Ian Kou', 'Jifei Miao']
|
2023-04-30
| null | null | null | null |
['image-inpainting', 'matrix-completion']
|
['computer-vision', 'methodology']
|
[-9.81498435e-02 -4.39703941e-01 2.13496368e-02 6.79339394e-02
-4.33400154e-01 -2.14561913e-03 2.92377293e-01 -1.23505116e-01
-7.29930639e-01 8.19588423e-01 -1.05341882e-01 -6.74589127e-02
1.67950943e-01 -6.25680685e-01 -5.77602863e-01 -7.98269510e-01
8.95089731e-02 -7.53836408e-02 -1.10737860e-01 -5.55691242e-01
2.57713288e-01 5.29982388e-01 -1.26067972e+00 -2.34428227e-01
1.20661461e+00 9.21161771e-01 -7.04926923e-02 3.90950382e-01
-1.27686068e-01 6.76393628e-01 -6.92709565e-01 -6.80975676e-01
4.06291783e-01 -6.15506411e-01 -3.41977686e-01 1.41595930e-01
1.33409530e-01 -4.38423157e-01 -5.02356231e-01 1.30279791e+00
2.83698380e-01 4.69933003e-02 3.79721314e-01 -1.42822778e+00
-8.40311408e-01 5.71639352e-02 -9.87870574e-01 -3.61695886e-02
2.29544908e-01 -7.50657842e-02 1.01724207e+00 -9.38399971e-01
6.64120913e-01 1.29361069e+00 4.74604607e-01 1.91857181e-02
-1.31286621e+00 -6.29878581e-01 -2.11164162e-01 4.53732133e-01
-1.77340567e+00 2.36382961e-01 1.06371498e+00 -7.77522996e-02
3.44450235e-01 2.93536156e-01 1.01289725e+00 5.24368823e-01
3.38302046e-01 7.94076860e-01 1.42770708e+00 -3.57892185e-01
1.06980927e-01 8.16706046e-02 -4.36072052e-01 7.45807409e-01
2.52544701e-01 1.13994151e-01 -5.22125363e-01 1.14903063e-01
1.13393342e+00 1.09867409e-01 -4.52256501e-01 -4.93210584e-01
-1.52108574e+00 7.38033533e-01 8.31146002e-01 1.58752292e-01
-5.89719057e-01 1.89825758e-01 2.94991702e-01 1.03868663e-01
3.14356297e-01 4.52144474e-01 2.65740335e-01 -2.27161929e-01
-9.15362179e-01 9.78375524e-02 5.40712059e-01 7.78363109e-01
1.19708180e+00 4.47974384e-01 2.47697368e-01 6.89128697e-01
2.36537308e-01 5.90039909e-01 3.74816477e-01 -6.62290335e-01
4.39088702e-01 6.18031383e-01 1.96845606e-01 -1.43458164e+00
-1.72445327e-01 -2.85641760e-01 -1.28790140e+00 4.36171979e-01
2.92265534e-01 -1.36489227e-01 -7.66022623e-01 1.35622513e+00
3.21978480e-01 1.32040173e-01 -9.96234566e-02 1.32605135e+00
4.17254180e-01 7.83256471e-01 4.29189205e-02 -2.48155594e-01
1.14488089e+00 -6.03329599e-01 -9.98801410e-01 8.79171416e-02
1.78602919e-01 -1.06308377e+00 9.41349089e-01 5.97809732e-01
-9.06585574e-01 -4.90725309e-01 -1.62608528e+00 -1.90000087e-01
-3.06282490e-01 3.10457915e-01 6.90122485e-01 5.16852140e-01
-6.79121137e-01 5.25196850e-01 -8.81730855e-01 -2.64070518e-02
-9.98660102e-02 2.34568983e-01 -5.32057941e-01 -9.85164102e-03
-1.26282334e+00 9.74348664e-01 5.25541484e-01 5.94618201e-01
-1.41412035e-01 -3.06740969e-01 -7.10465431e-01 -1.88575432e-01
9.40525755e-02 -3.29735279e-01 8.01413000e-01 -1.30502367e+00
-1.86465788e+00 3.46009910e-01 2.02362001e-01 -2.09472537e-01
5.44910550e-01 -3.53383124e-01 -3.57113004e-01 3.90548527e-01
-1.28880948e-01 7.03358471e-01 1.18180227e+00 -1.28097880e+00
-5.37503362e-01 5.04000597e-02 1.18910477e-01 4.08188939e-01
-4.06849474e-01 -2.00491339e-01 -8.75588238e-01 -1.04251921e+00
5.52683294e-01 -9.65260625e-01 -2.03338385e-01 2.31774300e-01
-4.14100915e-01 1.76677912e-01 4.28738832e-01 -8.81519854e-01
1.27555740e+00 -2.16009092e+00 5.01002371e-01 2.36831695e-01
5.26987463e-02 3.53636712e-01 -3.11822165e-02 6.72142386e-01
-3.05451870e-01 -1.49701342e-01 -2.17280924e-01 -1.74433231e-01
-8.59523937e-02 3.58141840e-01 -2.71240056e-01 8.44298482e-01
5.59929013e-01 4.70793635e-01 -9.24183488e-01 -5.01662910e-01
2.64037579e-01 7.76779592e-01 -3.03854436e-01 2.35116467e-01
1.42857760e-01 3.85411143e-01 -4.21458393e-01 4.81404841e-01
1.09900177e+00 6.79619610e-03 1.61295936e-01 -9.10999894e-01
-3.27663392e-01 -3.00541073e-02 -1.63627958e+00 1.52601957e+00
-4.43103373e-01 4.52930748e-01 9.28451195e-02 -8.66844654e-01
8.34462225e-01 2.59320349e-01 6.40146315e-01 -8.47994566e-01
1.49039343e-01 2.52895743e-01 -6.99338387e-04 -1.97088450e-01
6.89331889e-01 -4.18972045e-01 2.96667099e-01 4.63645309e-01
-1.51243538e-01 -4.24270123e-01 5.08377969e-01 1.26690015e-01
5.32530367e-01 4.67497826e-01 3.86519074e-01 1.33872628e-01
5.86430013e-01 -5.02465926e-02 8.73696029e-01 -1.46004800e-02
-1.28475353e-01 9.24359620e-01 5.90528369e-01 -3.77009630e-01
-1.09599149e+00 -9.75257874e-01 1.35348998e-02 5.17280042e-01
3.53648812e-01 -4.77923959e-01 -4.30618078e-01 -1.56001478e-01
-1.52673706e-01 3.00113648e-01 -4.56855178e-01 -6.91891834e-02
-6.94877863e-01 -7.60824382e-01 3.75539184e-01 4.09356415e-01
9.93327141e-01 -5.62349498e-01 -5.10537624e-01 3.14447403e-01
-2.97978818e-01 -1.01819956e+00 -3.59671831e-01 2.49506847e-04
-8.41175556e-01 -9.94960129e-01 -9.90708590e-01 -4.21196312e-01
9.60782349e-01 6.23555303e-01 8.07973206e-01 3.27742070e-01
-2.15018734e-01 2.91052490e-01 -5.99241912e-01 -3.14725131e-01
-3.34058344e-01 -3.46587330e-01 -6.41752928e-02 3.75297129e-01
1.55040780e-02 -3.20837021e-01 -8.59087646e-01 1.68214738e-01
-1.40187621e+00 3.84723127e-01 8.48634183e-01 1.08386850e+00
5.78364968e-01 9.34680179e-02 3.17758322e-01 -6.20545447e-01
6.74039900e-01 -8.29471797e-02 -7.44905889e-01 2.37879410e-01
-4.84216928e-01 1.61192983e-01 7.65786231e-01 -2.14300305e-01
-1.04000115e+00 8.61522928e-02 -4.49594110e-02 -4.26980585e-01
4.49265152e-01 8.26184452e-01 1.07113376e-01 -3.16416264e-01
4.02915239e-01 2.64099449e-01 4.20250952e-01 -2.64427871e-01
6.38786137e-01 2.96756148e-01 1.92518875e-01 -3.96616071e-01
1.25475740e+00 6.19581580e-01 3.09284210e-01 -8.92367244e-01
-2.63972044e-01 -3.68495494e-01 -6.73469722e-01 -1.26995072e-01
8.82938504e-01 -9.56874669e-01 -7.28368700e-01 4.88145828e-01
-1.09772646e+00 3.33486825e-01 -3.56585234e-02 7.30705976e-01
-3.05038095e-01 9.71742094e-01 -6.91764772e-01 -6.16666138e-01
-1.64858580e-01 -1.37428260e+00 7.84380019e-01 2.83263654e-01
1.02022022e-01 -8.30084026e-01 -1.09511010e-01 -4.04976718e-02
2.93866664e-01 2.64254004e-01 7.31764436e-01 3.31630886e-01
-8.72981489e-01 -4.07381326e-01 -4.52935606e-01 6.64381385e-01
2.84783542e-01 3.24300468e-01 -4.15949792e-01 -3.86109263e-01
-8.96212459e-02 -4.49904472e-01 4.97879773e-01 -2.04842672e-01
6.65808558e-01 -7.33750090e-02 2.95473903e-01 7.48383582e-01
1.72888446e+00 8.13271329e-02 8.38176191e-01 5.82230031e-01
7.51681447e-01 2.25591406e-01 8.23874533e-01 6.73372447e-01
2.96006978e-01 6.48159981e-01 5.06251633e-01 -5.02545178e-01
2.66364366e-02 -1.16893895e-01 2.69663513e-01 1.12864673e+00
-4.79195654e-01 2.50898063e-01 -5.36532760e-01 1.51929945e-01
-1.63392103e+00 -5.95933080e-01 1.33758038e-01 2.38317585e+00
9.88331437e-01 -8.04163441e-02 8.19568615e-03 3.34709644e-01
5.19596040e-01 1.09058276e-01 -4.08289164e-01 -3.23688775e-01
-2.49293521e-01 3.22382390e-01 6.34494007e-01 1.96328565e-01
-1.12596297e+00 6.36611402e-01 5.39342070e+00 7.50207782e-01
-1.58659530e+00 -2.84500599e-01 2.73566604e-01 5.18626332e-01
-2.42142007e-01 1.05811439e-01 -1.14636324e-01 1.44878075e-01
3.26622039e-01 -1.25092849e-01 7.03873336e-01 4.65190887e-01
1.18685335e-01 -3.19817603e-01 -7.83529997e-01 1.29740381e+00
2.90645100e-02 -1.01524103e+00 2.72022337e-01 -1.81713313e-01
5.95080256e-01 -3.55355829e-01 2.90917963e-01 4.22754921e-02
-9.00478289e-02 -8.00480306e-01 7.58903861e-01 5.42748988e-01
7.42975295e-01 -8.69408369e-01 6.26351595e-01 4.01111394e-02
-1.13552129e+00 3.01780224e-01 -5.78849494e-01 2.21042745e-02
4.67237048e-02 4.09732521e-01 -8.11483979e-01 9.08177972e-01
5.29320478e-01 6.77745700e-01 -6.24559402e-01 1.19925690e+00
-5.29630840e-01 4.78801787e-01 -4.18740302e-01 7.56090879e-02
3.44741642e-01 -1.04087079e+00 3.85281116e-01 9.32053626e-01
4.22822952e-01 2.23445895e-04 -1.18676543e-01 7.69326866e-01
1.25450447e-01 4.61250335e-01 -3.32608461e-01 -3.07883024e-01
1.84310377e-02 1.48799515e+00 -8.20580542e-01 -8.96135271e-02
-7.32438564e-01 1.19196963e+00 1.48561686e-01 7.13410258e-01
-8.44390094e-01 -5.99014044e-01 4.15849984e-01 -2.85865873e-01
2.44434282e-01 -7.77643681e-01 -4.99081984e-02 -1.30379367e+00
-4.63731587e-02 -1.11521602e+00 5.62515408e-02 -7.65559494e-01
-1.11693323e+00 5.58430135e-01 -7.45063052e-02 -1.86510587e+00
-1.24788389e-01 -7.87002385e-01 -5.23108304e-01 1.06731439e+00
-1.63699794e+00 -1.19833779e+00 -3.16399008e-01 5.93113363e-01
-1.71342250e-02 1.31553575e-01 6.38708174e-01 5.31403303e-01
-7.09303081e-01 3.62515539e-01 3.53435457e-01 1.37160271e-01
8.34480822e-01 -1.58409703e+00 -3.53652127e-02 1.01401234e+00
9.43704396e-02 9.29151237e-01 8.27629328e-01 -4.05229032e-01
-2.06622505e+00 -7.39613295e-01 2.77323872e-01 3.62832248e-01
6.92006290e-01 8.09224099e-02 -9.63214934e-01 4.74146366e-01
5.17101347e-01 3.03765312e-02 5.56689560e-01 -3.11329335e-01
-2.24783540e-01 -5.26125908e-01 -6.74134374e-01 9.93801236e-01
2.38915935e-01 -5.32211840e-01 -2.06783161e-01 1.94425762e-01
2.25655481e-01 -3.01778942e-01 -9.37278569e-01 2.32208073e-01
5.05306602e-01 -1.13331640e+00 9.86952543e-01 -2.75281131e-01
3.33852232e-01 -7.96715140e-01 9.83455591e-03 -1.50588918e+00
-2.41102576e-01 -6.72527611e-01 3.12322736e-01 1.10876226e+00
-7.45833069e-02 -6.35495782e-01 7.13731408e-01 4.53143924e-01
1.97023183e-01 -6.48171723e-01 -1.08896804e+00 -4.28240389e-01
-1.61222965e-02 -1.52155951e-01 4.28256154e-01 5.79923809e-01
-1.39531076e-01 1.74397096e-01 -6.59069002e-01 1.02113396e-01
6.12425685e-01 1.32596344e-01 9.64736640e-01 -6.45375371e-01
-1.70132667e-01 -3.67357939e-01 -5.01946807e-01 -1.06272638e+00
-1.30967513e-01 -5.05861640e-01 8.77730846e-02 -1.39640427e+00
-3.32070708e-01 -4.04225528e-01 -3.10744673e-01 7.18154088e-02
-4.48290646e-01 5.43569982e-01 5.34243405e-01 3.36125910e-01
-3.02161396e-01 1.02947259e+00 1.87367725e+00 -2.32676864e-01
1.01033468e-02 -1.93767309e-01 -2.89583594e-01 7.38635957e-01
4.77504522e-01 -3.41800973e-02 -5.11228085e-01 -3.07920516e-01
7.48111188e-01 1.89122409e-01 2.95009792e-01 -9.42655027e-01
1.66146755e-01 -2.76596129e-01 4.22314584e-01 -6.15363777e-01
6.42522573e-01 -9.23077404e-01 1.28638327e-01 4.77918923e-01
2.55036801e-01 4.07145739e-01 -2.82384828e-02 5.85620284e-01
-8.05814803e-01 -1.05165131e-01 8.47703695e-01 -1.31659552e-01
-8.01456451e-01 1.94960937e-01 -3.34126353e-01 -3.38247955e-01
8.44735980e-01 -4.00198877e-01 6.54586107e-02 -4.78787035e-01
-2.53628820e-01 -1.82501733e-01 3.76307487e-01 1.36136413e-01
9.47374344e-01 -1.50265336e+00 -4.47788447e-01 2.38401622e-01
-1.06188683e-02 -2.11654693e-01 4.93289642e-02 1.01871920e+00
-1.26815832e+00 1.23554848e-01 -6.92039132e-01 -3.52684677e-01
-8.08634639e-01 5.68868876e-01 1.87584326e-01 -9.93359089e-02
-4.69769537e-01 5.16102314e-01 1.15261070e-01 -5.86814359e-02
1.20671459e-01 -3.66578668e-01 -3.29315245e-01 1.67552590e-01
1.96956769e-01 4.40620124e-01 5.03959842e-02 -9.56061721e-01
-1.69663280e-01 5.30019701e-01 8.47736448e-02 -2.20374912e-01
1.21773696e+00 -1.54008999e-01 -5.21537483e-01 2.85591274e-01
1.20865715e+00 -1.25825163e-02 -1.18087626e+00 -1.03902258e-01
-1.99772611e-01 -6.10878706e-01 1.09275706e-01 -2.74737835e-01
-1.13706768e+00 1.07688522e+00 4.96434867e-01 1.31303534e-01
1.38431263e+00 -1.00143480e+00 8.71200979e-01 3.99863064e-01
3.42053086e-01 -1.31881392e+00 3.08648527e-01 2.36302972e-01
1.20719612e+00 -1.24736881e+00 5.67992568e-01 -5.53793490e-01
-7.05334842e-01 1.59775627e+00 5.59660971e-01 -3.76642495e-01
7.09662914e-01 -3.16178799e-01 2.34720781e-01 2.32187375e-01
-1.08127505e-01 -2.24893644e-01 4.19861257e-01 2.75614768e-01
5.98186970e-01 -2.01986209e-02 -7.85833955e-01 -9.61229950e-02
-1.50429094e-02 -1.07030101e-01 6.41880274e-01 9.98247921e-01
-2.31314916e-02 -1.40771067e+00 -7.50382304e-01 -9.22890306e-02
-3.65193307e-01 4.95303236e-02 2.30852235e-02 1.03775096e+00
2.09065482e-01 6.70407355e-01 -2.84594178e-01 -2.81051189e-01
3.36886525e-01 -3.64885062e-01 5.47332883e-01 -3.16591889e-01
-2.32027858e-01 2.67788887e-01 -2.40718454e-01 -4.54683274e-01
-6.99142873e-01 -2.97783643e-01 -1.34998155e+00 -1.68975696e-01
-2.17021957e-01 1.82054624e-01 8.54230285e-01 7.26838171e-01
-5.40373893e-03 2.49453932e-01 7.60946572e-01 -8.76321733e-01
-8.23589563e-01 -7.21589029e-01 -9.63968158e-01 6.84643984e-01
2.63615608e-01 -8.21517169e-01 -6.72896728e-02 8.80284831e-02]
|
[10.83678913116455, -1.7185115814208984]
|
54e238bd-efd7-4611-9d25-db8f1e5d5e5f
|
robust-point-cloud-registration-framework-1
|
2211.04696
| null |
https://arxiv.org/abs/2211.04696v1
|
https://arxiv.org/pdf/2211.04696v1.pdf
|
Robust Point Cloud Registration Framework Based on Deep Graph Matching(TPAMI Version)
|
3D point cloud registration is a fundamental problem in computer vision and robotics. Recently, learning-based point cloud registration methods have made great progress. However, these methods are sensitive to outliers, which lead to more incorrect correspondences. In this paper, we propose a novel deep graph matching-based framework for point cloud registration. Specifically, we first transform point clouds into graphs and extract deep features for each point. Then, we develop a module based on deep graph matching to calculate a soft correspondence matrix. By using graph matching, not only the local geometry of each point but also its structure and topology in a larger range are considered in establishing correspondences, so that more correct correspondences are found. We train the network with a loss directly defined on the correspondences, and in the test stage the soft correspondences are transformed into hard one-to-one correspondences so that registration can be performed by a correspondence-based solver. Furthermore, we introduce a transformer-based method to generate edges for graph construction, which further improves the quality of the correspondences. Extensive experiments on object-level and scene-level benchmark datasets show that the proposed method achieves state-of-the-art performance. The code is available at: \href{https://github.com/fukexue/RGM}{https://github.com/fukexue/RGM}.
|
['Manning Wang', 'Chenxi Zhang', 'Shaolei Liu', 'Xiaoyuan Luo', 'Jiazheng Luo', 'Kexue Fu']
|
2022-11-09
| null | null | null | null |
['point-cloud-registration', 'graph-matching']
|
['computer-vision', 'graphs']
|
[-2.84982800e-01 -1.75742254e-01 1.90524876e-01 -3.87626499e-01
-6.89569354e-01 -2.31219366e-01 3.23799193e-01 2.45487317e-01
-1.89890459e-01 1.83630347e-01 -2.87673801e-01 -7.52178282e-02
1.25057446e-02 -1.03109276e+00 -1.00412083e+00 -5.56216896e-01
1.86618745e-01 6.48590207e-01 2.34501705e-01 -1.11763343e-01
2.80555189e-01 5.31913221e-01 -1.09347343e+00 -2.57524163e-01
9.93521571e-01 9.76425588e-01 3.28672796e-01 -2.90884431e-02
-7.34848529e-02 1.28009558e-01 -7.68460855e-02 -1.45446584e-01
6.07775033e-01 -2.00819477e-01 -5.81689358e-01 2.04956457e-01
4.08260614e-01 -2.32103512e-01 -4.58324164e-01 1.40160000e+00
4.97913361e-01 1.93517417e-01 3.57464641e-01 -1.36337101e+00
-4.92607594e-01 7.35001191e-02 -6.79551125e-01 -3.20621639e-01
4.64262933e-01 3.32498223e-01 9.39971268e-01 -1.17397797e+00
4.39429283e-01 1.16867435e+00 5.88125885e-01 1.09162815e-01
-9.93576050e-01 -9.72143769e-01 1.21889703e-01 1.57829553e-01
-1.57983553e+00 -1.84425682e-01 1.07872152e+00 -4.56506938e-01
6.45581901e-01 5.20767830e-02 8.34008455e-01 4.67755735e-01
9.64726210e-02 4.44267213e-01 7.01260149e-01 -1.59995943e-01
-2.16548145e-02 -3.63649160e-01 -1.60660505e-01 8.78371298e-01
1.22058921e-01 8.98746625e-02 -9.93153527e-02 -2.50278916e-02
1.12458229e+00 4.93419379e-01 -3.55765104e-01 -6.38067067e-01
-1.35328007e+00 8.54185581e-01 1.06592560e+00 2.48034671e-01
-3.88632357e-01 1.67013437e-01 3.39768827e-02 1.40886113e-01
5.62087238e-01 2.78747439e-01 -6.53310493e-02 2.14883775e-01
-4.96662736e-01 3.52493137e-01 5.28883815e-01 1.25073946e+00
1.30110800e+00 -3.87149930e-01 1.65902004e-01 7.80861318e-01
5.57385206e-01 5.16595483e-01 6.42859787e-02 -8.30556512e-01
6.37374997e-01 9.62774873e-01 -1.57845005e-01 -1.48553300e+00
-4.66339767e-01 -3.90942723e-01 -1.04962373e+00 1.77978754e-01
8.89903232e-02 1.93852976e-01 -9.07127798e-01 1.36398530e+00
5.67332327e-01 7.87353456e-01 -4.05343562e-01 1.21596467e+00
8.03728819e-01 7.08037615e-01 -4.44397718e-01 1.29614636e-01
1.05589950e+00 -7.66474366e-01 -2.34269217e-01 -2.00725302e-01
5.06536543e-01 -9.17858541e-01 9.19785619e-01 -3.67182158e-02
-1.02507865e+00 -4.79150385e-01 -8.91691148e-01 -2.26686731e-01
1.09500997e-02 -6.94989190e-02 3.62479627e-01 -1.34328246e-01
-8.20591271e-01 7.03133166e-01 -1.00747359e+00 -1.73606709e-01
5.50876498e-01 4.99285460e-01 -4.65098083e-01 -3.05262595e-01
-8.67452800e-01 6.23765886e-01 4.76689696e-01 4.76883054e-01
-4.80700970e-01 -6.52447879e-01 -1.07144320e+00 -8.56456757e-02
4.02191937e-01 -8.87775540e-01 8.55824292e-01 -4.24859762e-01
-1.29424596e+00 9.19357181e-01 -3.85116413e-02 1.18573107e-01
5.86729527e-01 -5.66551089e-02 -5.69190197e-02 -5.91806956e-02
2.48986319e-01 4.99152422e-01 5.25990486e-01 -1.31337070e+00
-4.47191298e-01 -4.92813647e-01 4.32504043e-02 1.64857835e-01
1.16812922e-01 -1.23373576e-01 -8.43163431e-01 -3.48746806e-01
6.65578783e-01 -1.09767783e+00 -3.34701687e-01 2.11879209e-01
-5.89592278e-01 -3.15564334e-01 6.47176445e-01 -5.92507720e-01
6.99166894e-01 -2.26891327e+00 2.13944361e-01 5.71581721e-01
4.69262689e-01 3.35913114e-02 -2.03259856e-01 3.22056472e-01
-1.11255147e-01 -3.36371474e-02 -5.38738012e-01 -5.61049223e-01
-4.85338271e-02 6.27435148e-02 1.30163059e-01 6.97619736e-01
2.77946830e-01 9.79028225e-01 -9.51824069e-01 -2.72713870e-01
4.71021712e-01 5.00401080e-01 -5.38395286e-01 3.13471109e-01
-9.41332951e-02 6.84121788e-01 -7.26601183e-01 5.86858988e-01
1.09986949e+00 -3.69704425e-01 -3.47218007e-01 -4.34317797e-01
-1.68128788e-01 2.04405293e-01 -1.38533998e+00 2.04021716e+00
-3.69516850e-01 1.76298276e-01 -9.51089114e-02 -9.37276244e-01
1.18989503e+00 -7.57919326e-02 6.80003047e-01 -5.29471397e-01
4.09460783e-01 3.78060699e-01 -1.41553313e-01 -9.14303362e-02
1.79574117e-01 1.78623557e-01 1.71575382e-01 1.74820013e-02
-2.90406734e-01 -5.20602643e-01 -1.85584396e-01 5.53378016e-02
9.35413897e-01 1.44434735e-01 1.43109902e-03 -3.39279585e-02
5.90885460e-01 5.41291796e-02 8.18027616e-01 6.97177649e-02
3.03894043e-01 9.65423822e-01 1.58032879e-01 -3.78695846e-01
-1.08680582e+00 -1.00135696e+00 -1.21661700e-01 7.49511495e-02
8.14829051e-01 -4.05417711e-01 -5.98028243e-01 -3.80579084e-01
1.07150674e-01 3.00542444e-01 -2.38319725e-01 -4.16974962e-01
-6.90348208e-01 -3.75900149e-01 -1.23874776e-01 3.32897007e-01
6.17356777e-01 -9.49244261e-01 -8.47656950e-02 1.61778837e-01
-2.07004488e-01 -1.09694993e+00 -7.85013497e-01 -3.27958256e-01
-9.57910061e-01 -1.02061152e+00 -5.73823750e-01 -1.00495982e+00
1.18028474e+00 5.04976928e-01 8.65200162e-01 6.89237237e-01
-1.10994957e-01 -2.21724752e-02 -2.95839369e-01 -1.72323361e-01
-1.04204871e-01 1.01244956e-01 -8.89149308e-02 8.96377787e-02
3.43152672e-01 -7.70864367e-01 -7.04238474e-01 5.12391210e-01
-6.26704156e-01 1.93767965e-01 5.96269310e-01 7.47745752e-01
1.05251706e+00 -7.58651048e-02 8.07366669e-02 -6.10283434e-01
3.01140964e-01 -3.43464106e-01 -9.77746129e-01 -1.81769952e-01
-3.66997987e-01 -1.42630085e-01 4.21973377e-01 -1.90307736e-01
-3.69542539e-01 4.02534783e-01 -3.13882023e-01 -1.04452372e+00
-3.45004280e-03 5.54274738e-01 -4.64206576e-01 -4.73806918e-01
2.50757933e-01 1.41481668e-01 1.77829102e-01 -6.28913403e-01
1.35931715e-01 3.98770571e-01 4.83920276e-01 -4.76374924e-01
1.39521158e+00 3.93962473e-01 5.42276576e-02 -4.91837651e-01
-5.10838389e-01 -6.96237504e-01 -7.82833755e-01 -5.87246902e-02
7.68452644e-01 -9.92830634e-01 -7.11603045e-01 7.04390049e-01
-1.27698767e+00 -2.71394670e-01 -6.54198676e-02 6.03569031e-01
-5.18379867e-01 4.25659329e-01 -4.10737216e-01 -2.00360626e-01
-2.92091846e-01 -1.35131717e+00 1.27834749e+00 3.03178996e-01
2.12147236e-01 -8.51876080e-01 6.69165403e-02 2.47043371e-01
-4.08286303e-02 4.31060582e-01 6.48529172e-01 -3.63767087e-01
-1.18741417e+00 -3.86712968e-01 -4.18501496e-01 2.32195556e-01
2.47183233e-01 -4.42369506e-02 -5.51187038e-01 -5.41409731e-01
-9.27789044e-03 1.34968162e-01 4.53319937e-01 1.86198771e-01
1.16408730e+00 -3.59617509e-02 -4.98522013e-01 1.13159811e+00
1.46912587e+00 -4.94921803e-02 6.60295010e-01 4.64505494e-01
1.29229701e+00 3.13127369e-01 8.21477532e-01 3.70262891e-01
6.91592216e-01 7.95782566e-01 7.64138222e-01 -3.10495853e-01
9.28660408e-02 -3.96775395e-01 -1.22587800e-01 1.10961235e+00
-2.04645380e-01 1.36898175e-01 -1.07308936e+00 4.02805060e-01
-2.04267478e+00 -5.36265373e-01 -3.69272560e-01 2.44421959e+00
5.64918935e-01 1.63313463e-01 -6.11936301e-02 -1.98881984e-01
9.25611019e-01 -1.32318679e-02 -6.10837638e-01 2.68777579e-01
2.52921909e-01 1.77121922e-01 4.54035789e-01 5.47473252e-01
-9.49390113e-01 1.02993774e+00 3.77990890e+00 4.92146194e-01
-1.25134861e+00 -5.49537875e-02 2.50097215e-01 1.20562717e-01
-3.32832098e-01 2.85515338e-01 -5.39474249e-01 5.81383944e-01
7.98497573e-02 -2.86288977e-01 4.07217950e-01 6.91419303e-01
1.80911109e-01 2.20518738e-01 -1.11071599e+00 1.28840446e+00
-6.70257360e-02 -1.40904355e+00 -6.50750399e-02 3.14725757e-01
5.73735833e-01 3.14708918e-01 -3.29099685e-01 -5.03958166e-02
6.94536269e-02 -7.01748967e-01 5.29181242e-01 5.78459740e-01
5.94341934e-01 -8.40218067e-01 8.66134048e-01 3.78749847e-01
-1.43289375e+00 4.02966470e-01 -7.11774647e-01 7.06736445e-02
1.53399527e-01 8.44121456e-01 -5.28049886e-01 1.01976836e+00
7.76591241e-01 1.13769078e+00 -4.62350130e-01 1.40150690e+00
-2.95076430e-01 7.23183677e-02 -4.56138730e-01 2.26748660e-01
-3.19351926e-02 -7.51643240e-01 6.87197447e-01 6.61864400e-01
5.42685807e-01 3.08338106e-01 6.23724699e-01 1.15043747e+00
-2.59064943e-01 1.49690032e-01 -6.49254382e-01 4.53659594e-01
6.57200456e-01 1.43806052e+00 -7.07150877e-01 8.92004464e-03
-4.02746409e-01 9.25309002e-01 4.89444733e-01 1.80848449e-01
-6.83862269e-01 -4.40439731e-01 7.02515721e-01 2.44843885e-01
2.58698557e-02 -4.87344891e-01 -2.36560166e-01 -1.35186398e+00
4.44374681e-01 -6.12182558e-01 2.90896967e-02 -7.63140440e-01
-1.34738398e+00 5.31386256e-01 -1.32306144e-01 -1.70221424e+00
9.64967012e-02 -2.69370079e-01 -8.78240287e-01 1.10481369e+00
-1.47958755e+00 -1.07570720e+00 -9.00633872e-01 6.43955886e-01
9.48737040e-02 1.18332647e-01 3.50948989e-01 5.38745940e-01
-5.40376067e-01 4.77599293e-01 -1.83539391e-01 4.31999534e-01
6.20300651e-01 -9.97407973e-01 8.32194328e-01 7.61035979e-01
-1.48248980e-02 6.39369845e-01 2.19379202e-01 -7.42383003e-01
-1.48267972e+00 -1.34899998e+00 6.04246914e-01 -3.35582227e-01
4.93094742e-01 -5.15668988e-01 -1.24833477e+00 7.01296747e-01
-2.74092168e-01 2.85029948e-01 -2.20443355e-03 -2.68550605e-01
-9.35279578e-02 -1.53166071e-01 -1.06684554e+00 4.10292000e-01
1.24805391e+00 -4.65001464e-01 -4.53688592e-01 5.79756081e-01
9.30969119e-01 -9.58581984e-01 -9.81069505e-01 4.89391863e-01
1.66415110e-01 -7.61734843e-01 9.73508477e-01 -1.93254538e-02
3.47373158e-01 -6.78276658e-01 2.06881851e-01 -1.40947878e+00
-4.71440673e-01 -4.75042999e-01 3.76286983e-01 1.20701015e+00
1.97519794e-01 -1.02275240e+00 7.38109887e-01 4.19640839e-01
-5.07327378e-01 -8.13368380e-01 -8.84174883e-01 -8.52789402e-01
2.57395115e-02 -1.37307107e-01 8.74744296e-01 1.17900765e+00
-4.75058079e-01 1.74770758e-01 3.91377555e-03 5.97477913e-01
7.30331182e-01 3.47661853e-01 1.06644309e+00 -1.34959984e+00
8.58435035e-03 -4.48754042e-01 -9.31441188e-01 -1.08087468e+00
2.12818578e-01 -1.10156631e+00 2.34560147e-01 -1.81923676e+00
-1.18890661e-03 -8.61826003e-01 -5.85670061e-02 5.67797065e-01
-2.51673698e-01 1.33628681e-01 3.30322415e-01 4.01334196e-01
-2.10921913e-01 7.43604839e-01 1.35653496e+00 -8.78610238e-02
-2.45162085e-01 8.22511390e-02 -3.80314887e-01 6.63436294e-01
8.02795827e-01 -6.02783918e-01 -4.10207883e-02 -6.12189114e-01
3.60251255e-02 -1.10306621e-01 7.19573081e-01 -1.02906024e+00
4.67295676e-01 -2.18075514e-01 1.85860246e-01 -6.67858124e-01
3.38457227e-01 -9.95270431e-01 3.54371101e-01 3.90474647e-01
1.39123201e-01 2.47813001e-01 7.11200759e-02 3.40409338e-01
-2.78555036e-01 -1.08214475e-01 8.09298515e-01 -5.31792790e-02
-4.21248585e-01 1.10766792e+00 5.79851627e-01 -2.06544735e-02
9.63441491e-01 -2.33734727e-01 -7.62494802e-02 -1.55142441e-01
-3.91949147e-01 6.20570898e-01 9.24906909e-01 5.44417083e-01
8.37199390e-01 -1.57785583e+00 -8.10772359e-01 2.08344519e-01
3.05527240e-01 1.00925112e+00 5.66787422e-02 9.26540911e-01
-6.92966223e-01 -9.06776711e-02 -1.13261677e-01 -9.88402903e-01
-9.83115911e-01 2.99756914e-01 4.39315856e-01 1.76630482e-01
-9.81789529e-01 7.17902064e-01 3.34054351e-01 -8.60198677e-01
-4.56465557e-02 -4.44924533e-01 2.40518898e-01 -4.03474808e-01
1.62683249e-01 1.94218174e-01 2.49175146e-01 -7.72017062e-01
-5.36432862e-01 1.09740937e+00 -2.26565804e-02 3.72621924e-01
1.43808067e+00 8.31392705e-02 -4.28434163e-01 2.37417713e-01
1.47551131e+00 -1.02014311e-01 -1.15804076e+00 -4.08172131e-01
-1.43744439e-01 -8.47844422e-01 1.33207843e-01 -1.64460108e-01
-1.49359250e+00 8.41806233e-01 5.60637116e-01 -8.05858895e-02
1.01193058e+00 9.92928892e-02 1.05600679e+00 2.02294946e-01
5.06848037e-01 -5.78076124e-01 -1.25295877e-01 4.95052785e-01
1.07248616e+00 -1.33266795e+00 1.23506263e-01 -7.75612533e-01
-1.40539095e-01 1.02215767e+00 7.16830790e-01 -6.01332486e-01
6.78550720e-01 -8.59604031e-02 2.98660900e-02 -3.89749676e-01
-8.14949200e-02 -1.47234127e-01 3.54757547e-01 3.77342075e-01
9.17702094e-02 3.42059769e-02 -5.19665442e-02 1.16976574e-01
-3.29105526e-01 -7.11068918e-04 2.11293876e-01 8.81356716e-01
-2.26602823e-01 -1.23557687e+00 -3.37728441e-01 3.94697517e-01
2.40081236e-01 -3.90455537e-02 -3.75620693e-01 5.94785392e-01
2.02104799e-03 6.85161889e-01 2.09885597e-01 -5.41187704e-01
7.85709143e-01 -5.85176408e-01 3.50627482e-01 -8.48341048e-01
-3.89161527e-01 1.51951194e-01 -3.17315042e-01 -7.96390235e-01
-2.75094837e-01 -6.68542564e-01 -1.63497925e+00 -4.13888156e-01
-4.62731838e-01 1.26244113e-01 6.17371738e-01 7.97795773e-01
5.63844800e-01 3.59580934e-01 9.70655620e-01 -1.16347754e+00
-3.23757231e-01 -6.24604404e-01 -3.78583640e-01 5.99941254e-01
2.21072435e-01 -7.66725361e-01 -3.80519152e-01 -2.44187221e-01]
|
[7.718042373657227, -3.111527919769287]
|
dfb8bd41-6845-4375-8fd6-6dc45816e90a
|
attention-based-writer-independent
|
2009.04532
| null |
https://arxiv.org/abs/2009.04532v3
|
https://arxiv.org/pdf/2009.04532v3.pdf
|
Attention based Writer Independent Handwriting Verification
|
The task of writer verification is to provide a likelihood score for whether the queried and known handwritten image samples belong to the same writer or not. Such a task calls for the neural network to make it's outcome interpretable, i.e. provide a view into the network's decision making process. We implement and integrate cross-attention and soft-attention mechanisms to capture the highly correlated and salient points in feature space of 2D inputs. The attention maps serve as an explanation premise for the network's output likelihood score. The attention mechanism also allows the network to focus more on relevant areas of the input, thus improving the classification performance. Our proposed approach achieves a precision of 86\% for detecting intra-writer cases in CEDAR cursive "AND" dataset. Furthermore, we generate meaningful explanations for the provided decision by extracting attention maps from multiple levels of the network.
|
['Mihir Chauhan', 'Mohammad Abuzar Shaikh', 'Tiehang Duan', 'Sargur Srihari']
|
2020-09-07
| null | null | null | null |
['handwriting-verification']
|
['computer-vision']
|
[ 1.85462058e-01 6.49765730e-02 -5.88378161e-02 -8.24939787e-01
-5.91888070e-01 -6.10684335e-01 3.41206104e-01 -1.40423074e-01
-2.63157710e-02 4.01082724e-01 2.48453707e-01 -1.33887053e-01
-1.81145072e-01 -4.90442902e-01 -4.91894662e-01 -6.95689976e-01
3.69201988e-01 2.23982170e-01 -1.48672640e-01 1.13766223e-01
8.96694243e-01 8.20855379e-01 -1.22881317e+00 7.73415148e-01
3.34543407e-01 1.15479147e+00 3.69331837e-01 8.06585848e-01
-9.57377779e-04 8.47095311e-01 -1.02584684e+00 -6.19232535e-01
1.87987965e-02 -2.29574367e-01 -7.07905352e-01 1.19355552e-01
6.00552559e-01 -2.41373524e-01 -1.37816697e-01 1.26441872e+00
2.84888715e-01 -9.97289643e-02 1.09294772e+00 -1.16325676e+00
-1.11839652e+00 5.67527413e-01 -8.58719945e-01 5.25766134e-01
2.70656079e-01 2.25335523e-01 1.12245166e+00 -1.16511989e+00
5.20235717e-01 1.10451484e+00 4.73182350e-01 5.79027832e-01
-9.43055630e-01 -7.99384117e-01 3.47363263e-01 2.40249068e-01
-1.15955985e+00 -3.91568840e-01 9.87257302e-01 -4.21286613e-01
6.94658399e-01 4.04862136e-01 2.21080542e-01 1.08582699e+00
5.13601005e-01 7.92723060e-01 1.00969255e+00 -3.26218158e-01
2.79419590e-02 4.69923049e-01 4.38274890e-01 6.16092622e-01
3.87301631e-02 -1.47154495e-01 -8.73531103e-01 1.57799602e-01
8.94673467e-01 4.28560674e-02 -1.53915718e-01 3.11952084e-01
-9.32208955e-01 5.67384481e-01 7.49332190e-01 2.03503370e-01
-5.09490788e-01 1.08948588e-01 7.29591176e-02 1.10224880e-01
6.42956886e-03 8.56055558e-01 -2.95951158e-01 1.39528990e-01
-6.85968161e-01 8.64287093e-02 5.49557984e-01 7.86430895e-01
2.92550445e-01 4.75109853e-02 -4.68095273e-01 4.52174813e-01
5.81115365e-01 3.96925181e-01 3.69951874e-01 -6.65276766e-01
7.79322624e-01 8.79472077e-01 -7.08458051e-02 -1.28271687e+00
1.24245277e-02 -6.65758669e-01 -6.79980040e-01 6.14360750e-01
4.77099091e-01 7.16111809e-02 -9.16534245e-01 1.31815851e+00
-2.07769603e-01 -3.53827357e-01 7.37215672e-03 1.29154444e+00
6.36430085e-01 3.35498840e-01 -1.54377580e-01 2.93248266e-01
1.77495420e+00 -7.61174083e-01 -8.12654257e-01 -3.78420681e-01
-1.27944261e-01 -7.64841020e-01 1.14566255e+00 4.78218168e-01
-8.91032755e-01 -9.26411450e-01 -1.29635453e+00 -5.14764823e-02
-2.78804898e-01 8.01939547e-01 3.57937187e-01 2.10095599e-01
-6.46098912e-01 5.08724391e-01 -4.17448163e-01 -1.67987905e-02
6.58298135e-01 5.13332546e-01 -2.69233227e-01 3.38908762e-01
-7.76477754e-01 1.00934291e+00 3.97567376e-02 5.51323712e-01
-1.05438614e+00 -1.53374538e-01 -4.29849237e-01 4.16190267e-01
-7.37692639e-02 -2.04140246e-01 9.99585569e-01 -1.12988389e+00
-9.28884447e-01 7.37285554e-01 -5.06782889e-01 -1.45399913e-01
5.38018644e-01 -3.47981930e-01 -4.38720107e-01 1.61496684e-01
2.08109111e-01 5.73705316e-01 1.13553941e+00 -1.40917158e+00
-6.43445015e-01 -8.01704764e-01 -3.53305668e-01 2.09861159e-01
-1.50823236e-01 2.35317439e-01 -3.29748303e-01 -9.11973238e-01
3.64815444e-01 -3.94712418e-01 1.26301304e-01 1.41269937e-01
-8.17578137e-01 -2.85339892e-01 1.32540727e+00 -1.01668954e+00
8.51265967e-01 -2.23697472e+00 -1.84571609e-01 4.97588605e-01
3.35616529e-01 -6.91612214e-02 3.57768498e-02 3.82604375e-02
-2.78525531e-01 1.66624755e-01 7.36873001e-02 -2.99843311e-01
-2.36338153e-01 -1.92929760e-01 -7.51435995e-01 2.89338768e-01
9.15152252e-01 7.47401774e-01 -4.61516649e-01 -2.14734569e-01
-9.29961428e-02 2.57978559e-01 -1.51837647e-01 3.75083089e-01
-6.05649725e-02 2.44620085e-01 -5.59538424e-01 6.61751628e-01
4.56121713e-01 -3.35661352e-01 4.64036427e-02 -2.51579881e-01
1.90137044e-01 1.78395361e-01 -1.19230926e+00 1.07170594e+00
-5.27741760e-02 1.07742786e+00 -1.82706192e-01 -5.55458963e-01
1.51572812e+00 2.09779978e-01 -5.17489851e-01 -4.09698516e-01
4.22041476e-01 -3.19807567e-02 2.30132759e-01 -6.28346741e-01
4.95171100e-01 3.85231823e-02 8.53218213e-02 8.46308470e-01
-2.38254741e-01 5.04240274e-01 -4.88735288e-01 4.86429110e-02
6.35218978e-01 7.35698193e-02 -8.77050906e-02 -3.02634388e-01
5.23905337e-01 -3.55904549e-01 4.33488935e-01 8.57384801e-01
-1.89954892e-01 5.88057220e-01 7.64081836e-01 -5.58995664e-01
-1.01486444e+00 -9.15746331e-01 3.41553092e-02 8.67971063e-01
2.46291012e-01 2.20241711e-01 -6.01351082e-01 -8.97074044e-01
1.92486167e-01 7.78544307e-01 -1.10964632e+00 -1.88977256e-01
-2.46076241e-01 -2.53197134e-01 4.87467289e-01 1.14244020e+00
5.05654573e-01 -1.29242694e+00 -6.13529265e-01 -9.51502696e-02
3.68642807e-02 -6.58006549e-01 -4.46080893e-01 5.46074152e-01
-8.34437132e-01 -1.13950264e+00 -4.38788742e-01 -8.87904942e-01
1.26017809e+00 -1.42526418e-01 6.76839769e-01 6.36347290e-03
-3.75110030e-01 -2.82150120e-01 3.11018918e-02 -4.96614337e-01
-3.16258967e-01 -6.63950518e-02 1.74196228e-01 2.81429619e-01
6.18376732e-01 -5.63537925e-02 -5.16573727e-01 4.72640306e-01
-3.13685387e-01 2.65220478e-02 6.15182817e-01 9.79986668e-01
3.66829932e-01 2.71608699e-02 8.38546753e-01 -6.14887238e-01
1.02411485e+00 -1.08666517e-01 -5.00551701e-01 4.38895583e-01
-5.98013282e-01 4.94137824e-01 6.81824148e-01 -4.10046816e-01
-1.27771747e+00 5.83730787e-02 2.30648935e-01 -4.22065407e-01
-2.25660667e-01 1.27730682e-01 -2.07512543e-01 1.73450008e-01
5.89588225e-01 3.69278133e-01 -3.51442993e-02 -5.35771310e-01
-6.66660536e-03 9.33131993e-01 8.54058385e-01 -4.16958630e-01
4.91538078e-01 2.42732108e-01 -2.86304176e-01 -2.55183637e-01
-6.21086240e-01 -1.57820821e-01 -6.38851225e-01 -2.76570857e-01
9.57903564e-01 -4.68289524e-01 -1.32952273e+00 3.58222544e-01
-1.38166809e+00 2.17541128e-01 4.61222380e-02 2.01339722e-01
-2.84547031e-01 -9.83899236e-02 -5.39316118e-01 -1.28837752e+00
-4.40548062e-01 -1.34846973e+00 1.07410264e+00 5.75993359e-01
-4.87098694e-01 -7.37495780e-01 -5.57179928e-01 3.07667911e-01
-5.19228317e-02 -2.59357877e-02 9.51333284e-01 -1.04840982e+00
-6.50110960e-01 -5.82606494e-01 -6.68782532e-01 3.50739628e-01
8.63400847e-03 2.78458685e-01 -1.75263166e+00 2.07594242e-02
-1.63977668e-02 -6.77188560e-02 8.38769197e-01 2.95088589e-01
1.50857997e+00 -3.54881197e-01 -2.13236198e-01 2.83431828e-01
1.21082199e+00 5.01062453e-01 4.62747961e-01 -7.15080351e-02
4.95806128e-01 7.78960586e-01 4.81781095e-01 3.33019793e-01
-3.32270041e-02 4.97185022e-01 5.91827631e-01 -2.58050174e-01
2.33658366e-02 -2.91385621e-01 2.85268456e-01 1.61865026e-01
-3.82996276e-02 -2.47202724e-01 -8.48305106e-01 4.10090297e-01
-1.74152386e+00 -1.00854206e+00 -1.84647873e-01 1.92387819e+00
5.74144125e-01 3.12569648e-01 -3.63661021e-01 2.89918274e-01
1.04527652e+00 -2.40053862e-01 -7.89922774e-01 -7.19712317e-01
-2.20214635e-01 -1.22459583e-01 1.47067875e-01 5.26452303e-01
-8.39766443e-01 7.00804889e-01 6.43584681e+00 4.67296809e-01
-1.24264812e+00 -3.65857869e-01 1.16749895e+00 1.91411585e-01
-1.15690023e-01 -2.47602075e-01 -1.18331718e+00 3.41374159e-01
2.74615884e-01 6.05232231e-02 1.74990714e-01 8.48449290e-01
1.50514901e-01 -2.01906860e-01 -1.42504191e+00 7.24868834e-01
2.47486711e-01 -1.36908102e+00 5.59568182e-02 1.18599057e-01
5.75545311e-01 -5.40202677e-01 3.47137034e-01 -1.77108020e-01
2.11166248e-01 -1.39325392e+00 8.19592595e-01 9.56732869e-01
7.69796431e-01 -7.33589709e-01 9.73659456e-01 3.34551454e-01
-8.73501897e-01 -3.23817760e-01 -2.72545159e-01 -1.66253686e-01
-3.81842792e-01 6.57477155e-02 -1.42012703e+00 1.13663442e-01
6.47637665e-01 5.05571723e-01 -7.23015010e-01 6.43064737e-01
-6.38866842e-01 4.18176681e-01 1.87861741e-01 -2.92691827e-01
1.17738724e-01 3.81358296e-01 5.07004797e-01 1.00334609e+00
-1.11095905e-02 -1.34581178e-01 -3.22191715e-01 1.49787951e+00
-1.92980826e-01 -3.25490624e-01 -5.22509396e-01 -3.41593698e-02
4.37968701e-01 1.10454774e+00 -8.30228865e-01 -2.64358610e-01
2.05693603e-01 1.19762540e+00 2.94963509e-01 4.32407200e-01
-6.23636603e-01 -8.06745410e-01 5.33562064e-01 -8.26897323e-02
3.15550178e-01 3.16709906e-01 -9.88811255e-01 -8.21316540e-01
2.38292485e-01 -7.12196290e-01 4.20979381e-01 -1.30082834e+00
-1.35869992e+00 8.45031559e-01 -7.99449384e-01 -9.00139093e-01
-1.18288442e-01 -9.99707878e-01 -9.63000000e-01 1.68645227e+00
-1.15381515e+00 -1.02377748e+00 -4.23568130e-01 4.12731081e-01
6.92250669e-01 -5.21941900e-01 8.58223617e-01 -1.57242879e-01
-5.60222566e-01 8.04890752e-01 -2.19098955e-01 7.07079768e-01
7.08158314e-01 -1.26306272e+00 4.10111904e-01 9.49742794e-01
6.73826039e-02 1.00701821e+00 5.69574893e-01 -7.56568909e-01
-1.13698649e+00 -7.41469562e-01 9.55486357e-01 -8.13223541e-01
3.02954346e-01 -3.99560750e-01 -9.08585191e-01 4.43452358e-01
2.34976318e-02 -3.04307491e-01 6.66197777e-01 2.46381357e-01
-5.54648399e-01 -1.77057758e-01 -1.20920789e+00 4.08118695e-01
4.80781347e-01 -8.32354069e-01 -8.93652439e-01 -7.11740479e-02
1.27025589e-01 -1.84206754e-01 -6.19311690e-01 -4.31973400e-04
9.44714189e-01 -6.60766065e-01 7.01630592e-01 -1.03958976e+00
1.03554869e+00 -4.44182247e-01 -2.27267668e-02 -8.13655794e-01
-4.64186549e-01 -6.33813366e-02 -5.04066702e-03 1.21735394e+00
7.59868026e-01 -3.30411077e-01 6.82270706e-01 7.32679248e-01
7.81507045e-02 -7.65455365e-01 -6.63538873e-01 -2.37989068e-01
-4.02622402e-01 -4.22792733e-01 6.82105362e-01 9.41620231e-01
-4.96117435e-02 2.50732213e-01 -2.91702241e-01 5.81358373e-01
4.27646369e-01 3.77471626e-01 4.89515781e-01 -1.09594035e+00
-3.01612258e-01 -6.86453700e-01 -3.94461572e-01 -8.82320821e-01
2.36935094e-02 -8.65606546e-01 8.23123679e-02 -1.27868772e+00
4.22148734e-01 -3.37640226e-01 -3.86832863e-01 6.50585771e-01
-3.19055438e-01 4.23292577e-01 2.64687091e-01 4.62854892e-01
-2.08949879e-01 1.90884531e-01 1.36011791e+00 -1.19089805e-01
1.72772333e-02 4.68825698e-02 -1.06742024e+00 6.22116923e-01
6.96055412e-01 -2.29835510e-01 -3.29918750e-02 -6.39958024e-01
1.85721546e-01 1.02594957e-01 7.44834900e-01 -6.64931417e-01
2.83319056e-01 -6.22755066e-02 1.35590935e+00 -6.06920958e-01
1.72202185e-01 -8.90071929e-01 -3.14806402e-01 5.47257721e-01
-8.99281323e-01 2.26024330e-01 1.69759318e-01 4.77809370e-01
-1.52073056e-01 -2.51007676e-01 5.75318754e-01 1.30685270e-01
-6.78672254e-01 -5.20784222e-02 -2.04618290e-01 -4.05698478e-01
8.04566979e-01 -5.97859800e-01 -5.32520115e-01 -3.30032319e-01
-9.61389184e-01 2.80653805e-01 7.15079084e-02 6.02955401e-01
9.02203023e-01 -1.15965545e+00 -5.49094021e-01 7.97271550e-01
2.40198523e-01 -2.57061213e-01 6.64793130e-04 3.48805308e-01
-3.51611704e-01 2.91904628e-01 -5.98327339e-01 -5.82109094e-01
-1.41595888e+00 1.45262152e-01 4.79091465e-01 6.80202693e-02
-3.62406582e-01 1.18985665e+00 1.95726737e-01 -8.95738304e-02
4.42398190e-01 -2.92673886e-01 -4.02392268e-01 6.79601133e-02
7.75326192e-01 1.40208080e-01 -1.72710255e-01 -4.99796450e-01
-5.97261727e-01 4.49443489e-01 -3.27620655e-01 -1.42566562e-01
1.17328715e+00 2.85453975e-01 1.13229245e-01 4.49849188e-01
7.44384587e-01 -5.57016172e-02 -1.53650451e+00 3.09550902e-03
-6.64191172e-02 -6.19382501e-01 -1.10588349e-01 -1.48304260e+00
-1.06828976e+00 1.27296662e+00 6.02555990e-01 1.03600770e-02
1.06499732e+00 6.97025284e-02 -1.45271989e-02 4.33594257e-01
-1.40505806e-01 -1.12514114e+00 1.86800256e-01 3.01337779e-01
1.34943604e+00 -1.26006472e+00 -1.14527494e-01 -1.52362986e-02
-1.03363287e+00 1.58330894e+00 9.94639397e-01 -1.36655822e-01
1.76968053e-01 4.27893579e-01 3.74349684e-01 -5.34102440e-01
-6.58429146e-01 5.17872036e-01 6.51986837e-01 4.95753616e-01
5.54709315e-01 1.23104965e-02 3.12000036e-01 1.17908287e+00
-2.62784541e-01 -2.11636096e-01 1.53950274e-01 5.30277491e-01
-3.21070641e-01 -4.16413724e-01 -5.44104874e-01 6.28003776e-01
-3.92118931e-01 4.62246947e-02 -8.59063208e-01 3.51638526e-01
2.57508680e-02 6.17307782e-01 3.69029164e-01 -5.53180516e-01
6.38037547e-02 2.31205449e-01 8.04137439e-02 -4.59353894e-01
-8.85255158e-01 -1.00986704e-01 -1.97355688e-01 -1.80554390e-01
2.00443313e-01 -4.70419198e-01 -1.28299510e+00 -3.43886092e-02
-5.96290648e-01 -7.98961613e-03 8.41933846e-01 9.10336912e-01
4.03655171e-01 8.56944263e-01 7.04297662e-01 -3.53856653e-01
-7.65527368e-01 -1.02881062e+00 -4.65416133e-01 2.59073526e-01
3.23127121e-01 -4.06094283e-01 -4.84928973e-02 2.79715210e-01]
|
[11.079718589782715, 2.1311075687408447]
|
a64b16c4-9c29-48b6-8b6e-a1c5a5cce45f
|
on-recoverability-of-graph-neural-network
|
2201.12843
| null |
https://arxiv.org/abs/2201.12843v4
|
https://arxiv.org/pdf/2201.12843v4.pdf
|
Graph Representation Learning via Aggregation Enhancement
|
Graph neural networks (GNNs) have become a powerful tool for processing graph-structured data but still face challenges in effectively aggregating and propagating information between layers, which limits their performance. We tackle this problem with the kernel regression (KR) approach, using KR loss as the primary loss in self-supervised settings or as a regularization term in supervised settings. We show substantial performance improvements compared to state-of-the-art in both scenarios on multiple transductive and inductive node classification datasets, especially for deep networks. As opposed to mutual information (MI), KR loss is convex and easy to estimate in high-dimensional cases, even though it indirectly maximizes the MI between its inputs. Our work highlights the potential of KR to advance the field of graph representation learning and enhance the performance of GNNs. The code to reproduce our experiments is available at https://github.com/Anonymous1252022/KR_for_GNNs
|
['Avi Mendelson', 'Ron Banner', 'Almog David', 'Evgenii Zheltonozhskii', 'Chaim Baskin', 'Maxim Fishman']
|
2022-01-30
| null | null | null | null |
['mutual-information-estimation']
|
['methodology']
|
[ 2.42654979e-01 3.60822201e-01 -3.71534258e-01 -3.29789370e-01
-5.60195923e-01 -6.10762775e-01 4.97427791e-01 6.10400259e-01
-4.92326349e-01 7.12231159e-01 1.27151474e-01 -6.34733975e-01
-3.75083774e-01 -1.16451550e+00 -7.89502561e-01 -6.10826254e-01
-5.25317013e-01 4.01268989e-01 -7.94824064e-02 -1.22919671e-01
-2.67374873e-01 3.39492947e-01 -8.60230207e-01 1.21207774e-01
9.04356062e-01 8.85039330e-01 -1.44910887e-01 5.86879492e-01
-8.87848511e-02 1.08817673e+00 -1.70829073e-01 -6.68849647e-01
1.18877470e-01 -2.00857565e-01 -9.37932312e-01 -4.55954462e-01
3.21314573e-01 9.29276571e-02 -7.56399632e-01 1.03771639e+00
4.04596835e-01 1.79956909e-02 7.87076771e-01 -1.53580821e+00
-7.73083985e-01 1.10257840e+00 -4.33679014e-01 4.10382487e-02
-3.64809148e-02 -1.70116201e-01 1.47043884e+00 -6.84405565e-01
4.07357603e-01 1.04931474e+00 8.88570487e-01 3.61853600e-01
-1.50379372e+00 -7.36848831e-01 1.47314280e-01 1.30920671e-03
-1.27786875e+00 -2.50875950e-01 9.47643459e-01 -3.32877845e-01
9.09979284e-01 3.72082032e-02 5.31809568e-01 9.90527093e-01
-3.91277187e-02 1.10214317e+00 7.52575874e-01 -2.63945431e-01
-1.18292263e-02 -1.06928051e-02 3.57969135e-01 1.05545652e+00
4.11803842e-01 -9.01282653e-02 -3.71777624e-01 -2.05624685e-01
8.91738892e-01 3.46454978e-02 -1.65268734e-01 -4.37199563e-01
-1.08528936e+00 1.03754807e+00 1.01838291e+00 3.64511490e-01
-2.05643401e-01 5.60512662e-01 4.44693953e-01 6.93500757e-01
8.60532463e-01 2.73647010e-01 -4.21095818e-01 1.68251470e-01
-6.24080181e-01 -4.35894728e-02 1.06165171e+00 5.70971489e-01
9.34313297e-01 1.61903314e-02 1.13614053e-02 8.61322165e-01
3.87076378e-01 1.95071682e-01 -4.21292931e-02 -4.80726808e-01
6.57040775e-01 9.73628342e-01 -5.64246774e-01 -1.11392200e+00
-5.56583226e-01 -7.00333118e-01 -1.20935142e+00 -7.98976049e-02
4.78893429e-01 -4.46503490e-01 -8.37078869e-01 1.96053410e+00
8.88772607e-02 1.79981843e-01 -5.19927144e-02 5.45860350e-01
1.29309380e+00 5.06046653e-01 2.10067511e-01 6.27001598e-02
7.53890038e-01 -8.99342835e-01 -4.85115677e-01 -4.32692021e-01
1.18593013e+00 -2.23338351e-01 8.39428961e-01 8.56750086e-02
-9.37999547e-01 4.02357951e-02 -8.97087395e-01 -1.51696041e-01
-5.95292509e-01 4.07325588e-02 1.18102050e+00 3.46029490e-01
-1.51315689e+00 6.93663299e-01 -8.64736855e-01 -3.00404906e-01
9.42378044e-01 6.16784871e-01 -6.48420870e-01 -2.25387141e-01
-1.42825758e+00 6.48801386e-01 4.25269812e-01 1.05940230e-01
-6.33289337e-01 -8.94063413e-01 -1.11523616e+00 7.34196082e-02
3.95145029e-01 -6.21314645e-01 6.64745629e-01 -8.49599779e-01
-1.26106238e+00 9.05091107e-01 2.66038418e-01 -6.36368692e-01
4.38722610e-01 -7.54544362e-02 -8.92440304e-02 -5.27045429e-02
-2.64081925e-01 5.90307534e-01 4.56093162e-01 -1.08288956e+00
-6.91179112e-02 -2.99652070e-01 2.55462736e-01 1.10203065e-01
-5.73502481e-01 -3.04695874e-01 -3.33493501e-01 -5.14387548e-01
-5.66280894e-02 -8.25283229e-01 -3.75161529e-01 5.13814576e-03
-6.46537006e-01 -3.61537308e-01 6.04113996e-01 -5.11294782e-01
1.12525475e+00 -1.85383034e+00 3.12448442e-01 3.05407912e-01
6.57111108e-01 5.02662361e-01 -2.88097829e-01 7.60876000e-01
-1.10216223e-01 2.13489845e-01 -5.16749680e-01 -5.06754339e-01
-1.62248835e-02 9.78569388e-02 2.82043982e-02 5.21332741e-01
3.00225109e-01 1.47104299e+00 -1.04907656e+00 -3.02727491e-01
2.17153475e-01 7.39099920e-01 -3.17978978e-01 1.03719354e-01
-1.74537703e-01 2.23461702e-01 -3.67250800e-01 3.62184048e-01
4.96451616e-01 -8.19373786e-01 3.91821623e-01 -1.62231609e-01
4.97583836e-01 2.77560830e-01 -9.17362869e-01 1.62254143e+00
-5.82786381e-01 7.75764942e-01 2.00579241e-01 -1.41129267e+00
8.77028465e-01 1.10033929e-01 5.17868936e-01 -5.30479789e-01
2.11866900e-01 -1.63111761e-01 -7.84786567e-02 -5.38721941e-02
1.12312444e-01 -3.99962366e-02 -2.63371617e-02 5.66180646e-01
3.15820187e-01 2.54668653e-01 1.69888183e-01 7.30830669e-01
1.54400206e+00 -2.99159795e-01 3.05160373e-01 -2.56620616e-01
2.26837233e-01 -2.27645457e-01 1.83449030e-01 7.49126911e-01
6.55558631e-02 2.39565134e-01 8.56988430e-01 -2.30136469e-01
-6.31081283e-01 -1.26722932e+00 1.20012671e-01 1.08301353e+00
-3.06353997e-02 -5.38558781e-01 -4.31263685e-01 -9.85981464e-01
2.86813706e-01 3.89815122e-01 -5.86867094e-01 -3.59214932e-01
-3.33660990e-01 -9.55002964e-01 7.55715191e-01 5.60871601e-01
4.24967617e-01 -8.53644133e-01 3.90888900e-01 1.50231943e-01
-5.82528152e-02 -1.02506220e+00 -1.49614856e-01 3.74285221e-01
-9.19573069e-01 -1.02902424e+00 -4.46120590e-01 -7.47236192e-01
8.99293900e-01 2.15087906e-01 1.35110343e+00 2.77471393e-01
-2.96532035e-01 4.27307993e-01 -2.30440035e-01 -2.64877766e-01
-3.66562337e-01 4.76159334e-01 -2.32280210e-01 -1.12344332e-01
1.95326731e-01 -7.93000221e-01 -5.37618577e-01 -5.28163314e-02
-1.07057703e+00 1.47011146e-01 6.53219640e-01 7.22802877e-01
1.62406191e-01 -3.70559655e-02 7.74442494e-01 -1.41458094e+00
1.01585305e+00 -8.44686449e-01 -5.23902833e-01 2.26748526e-01
-7.60166466e-01 1.54114023e-01 7.15981185e-01 -1.89891860e-01
-5.93552649e-01 -1.67220473e-01 -4.06710245e-02 -1.38566107e-01
6.21265359e-02 9.71779644e-01 1.91815913e-01 -3.02115679e-01
6.86577797e-01 -1.90802947e-01 1.80409789e-01 -2.89871961e-01
4.95485514e-01 3.27519655e-01 5.04240245e-02 -3.83404255e-01
9.34538126e-01 3.43093783e-01 2.66728997e-01 -6.88955069e-01
-8.94157112e-01 -4.66105610e-01 -6.10645473e-01 -2.40647510e-01
5.10262668e-01 -9.23523903e-01 -6.31192565e-01 6.05303586e-01
-7.75115728e-01 -7.62848377e-01 -9.20125172e-02 4.09855366e-01
-2.54492998e-01 2.77727634e-01 -1.02200246e+00 -6.64870262e-01
-5.41036606e-01 -7.44618595e-01 6.83668852e-01 1.02711141e-01
1.45172521e-01 -1.85498393e+00 -3.99451070e-02 3.87655586e-01
5.39943576e-01 4.49582785e-01 9.96456325e-01 -7.66252339e-01
-4.30874825e-01 -3.77797991e-01 -5.18299997e-01 4.79359180e-01
1.56322256e-01 -2.00712815e-01 -8.00999165e-01 -4.02748734e-01
-7.27333963e-01 -5.98194003e-01 1.44676065e+00 4.41837311e-01
1.02509391e+00 -4.34287071e-01 -3.76737028e-01 6.56895638e-01
1.54175711e+00 -5.14734924e-01 4.96386915e-01 -9.88821313e-02
1.27520669e+00 5.36230266e-01 1.03000380e-01 1.48766205e-01
6.94538772e-01 2.97036260e-01 5.31122029e-01 -5.17667055e-01
1.93598848e-02 -3.40572029e-01 2.91659832e-01 9.16821778e-01
-8.04200675e-03 -6.15103304e-01 -1.05716181e+00 3.76648813e-01
-2.12534523e+00 -7.63288260e-01 -1.37043610e-01 2.18959737e+00
8.72524083e-01 -1.36491784e-03 8.69973898e-02 1.11277610e-01
5.68901777e-01 4.38918203e-01 -4.38803524e-01 -2.35956848e-01
-1.02297388e-01 1.53538361e-01 7.69322872e-01 6.97216928e-01
-1.29400694e+00 1.19545448e+00 5.74906015e+00 7.82960892e-01
-1.07912052e+00 -4.14953642e-02 7.88822830e-01 1.42306864e-01
-3.67009789e-01 -1.03785492e-01 -4.16965812e-01 1.19082421e-01
1.00615609e+00 -1.10399812e-01 5.81179917e-01 4.75618601e-01
-1.04690611e-01 1.68673277e-01 -1.21146858e+00 7.65326500e-01
-1.57435477e-01 -1.48736060e+00 -6.06021769e-02 1.69420376e-01
7.05857456e-01 5.88658035e-01 1.78918943e-01 3.23156506e-01
8.50944102e-01 -1.34318733e+00 4.53339517e-02 3.85158062e-01
6.63406134e-01 -6.88355863e-01 6.97392881e-01 2.61428982e-01
-1.27219379e+00 1.50518164e-01 -3.66180658e-01 -1.23889267e-01
-5.54928593e-02 1.05108547e+00 -1.20877075e+00 6.98957384e-01
4.44002956e-01 1.15055609e+00 -6.80558205e-01 8.45169246e-01
-3.30299288e-01 9.08361256e-01 -4.97261763e-01 -1.19883575e-01
3.00650805e-01 -3.69827390e-01 4.19474870e-01 1.24275005e+00
-1.14884049e-01 -3.79233509e-01 1.76498652e-01 8.88902009e-01
-6.63612604e-01 2.47095302e-01 -9.92326558e-01 -4.31588411e-01
3.83515805e-01 1.54649794e+00 -7.23538220e-01 -5.63433534e-03
-3.67139608e-01 7.29855955e-01 9.57692623e-01 5.71497500e-01
-6.34712934e-01 -4.65368897e-01 4.71480846e-01 1.11595087e-01
8.93587694e-02 -3.81866008e-01 -1.74526021e-01 -1.13180983e+00
6.59988225e-02 -3.90582711e-01 6.80545807e-01 -2.77791530e-01
-1.63317883e+00 5.17643869e-01 -7.67528042e-02 -9.55895901e-01
-1.35225222e-01 -7.44473159e-01 -6.62869751e-01 5.80019772e-01
-1.63003981e+00 -1.46267188e+00 -2.39733040e-01 5.86039782e-01
-2.22111955e-01 6.70648813e-02 6.65685117e-01 3.45682949e-01
-5.02831936e-01 8.20966780e-01 2.69534379e-01 5.55858076e-01
5.45382917e-01 -1.38567007e+00 4.13105339e-01 6.27314866e-01
3.67469877e-01 5.93344092e-01 3.03117454e-01 -5.85910141e-01
-1.53772914e+00 -1.39686286e+00 6.46681786e-01 -3.10326606e-01
1.04414344e+00 -6.76773846e-01 -8.84579897e-01 8.38273287e-01
-3.20499130e-02 3.65775585e-01 9.52703059e-01 5.18109083e-01
-5.51485360e-01 -1.36930361e-01 -9.37839925e-01 5.34153521e-01
1.27369094e+00 -5.70995271e-01 1.19943030e-01 5.79415262e-01
8.20902348e-01 -8.68034363e-02 -1.21679163e+00 4.69581276e-01
2.75529504e-01 -6.80317402e-01 1.02155781e+00 -6.74153268e-01
3.67112249e-01 -7.67671391e-02 1.54572099e-01 -1.61306632e+00
-2.49104455e-01 -5.14054954e-01 -2.07564786e-01 1.03833210e+00
8.66092443e-01 -9.91014183e-01 9.66239512e-01 4.27175134e-01
1.59497440e-01 -1.07671964e+00 -6.80221558e-01 -5.29885232e-01
2.41578653e-01 -5.30004084e-01 1.77080050e-01 1.15739810e+00
1.81022882e-01 5.43726206e-01 -3.18326026e-01 -7.04037920e-02
9.08365965e-01 -1.75462916e-01 6.02266133e-01 -1.43504488e+00
-2.08113432e-01 -4.68121737e-01 -7.18375504e-01 -7.43918598e-01
5.41427851e-01 -1.58297968e+00 -3.20638567e-01 -1.96642816e+00
4.55473475e-02 -7.42155254e-01 -5.41220188e-01 7.96347916e-01
-1.30923957e-01 4.06764984e-01 6.84266835e-02 -4.75365035e-02
-8.14709425e-01 5.79919279e-01 1.16301119e+00 -3.25994849e-01
-2.69507885e-01 1.16221718e-01 -8.28019679e-01 3.90431166e-01
9.45076168e-01 -4.13656622e-01 -6.46086931e-01 -4.97175127e-01
5.84775686e-01 -4.75084186e-02 5.69814026e-01 -7.62258351e-01
4.57470149e-01 1.69756964e-01 2.65100628e-01 -1.86415195e-01
1.98037982e-01 -6.16517723e-01 1.08832316e-02 3.19239110e-01
-6.95172012e-01 -9.77401137e-02 8.66416320e-02 7.72878647e-01
-8.37822556e-02 1.68816209e-01 5.40517390e-01 -1.20332323e-01
-4.11091685e-01 8.00377131e-01 -2.51162034e-02 2.44787693e-01
8.31755638e-01 1.73279807e-01 -7.18178034e-01 -6.69520140e-01
-5.93102694e-01 4.50469136e-01 2.51678109e-01 3.57054919e-01
6.55745566e-01 -1.28941453e+00 -8.48712504e-01 -3.27808559e-02
1.17687382e-01 1.82948992e-01 9.83759984e-02 1.08467984e+00
-4.71111208e-01 2.84026116e-01 9.89529416e-02 -4.26742882e-01
-1.08031726e+00 2.43841767e-01 2.70789415e-01 -6.09910965e-01
-5.59649527e-01 1.04713428e+00 1.52072951e-01 -8.06271553e-01
3.89685720e-01 -1.09748341e-01 -3.09534907e-01 2.04602722e-02
1.73574984e-01 3.10011536e-01 1.49184063e-01 -3.78982693e-01
-3.53986144e-01 6.77653030e-02 -2.81530589e-01 2.20510483e-01
1.63462389e+00 2.07380846e-01 -4.64989752e-01 4.85918403e-01
1.64611399e+00 -3.59629244e-01 -9.59030151e-01 -6.91526711e-01
-1.43312002e-02 -3.06263119e-02 2.71226048e-01 -5.97351968e-01
-1.44546282e+00 8.44844103e-01 1.78569853e-01 4.01574463e-01
8.74553502e-01 3.99710745e-01 6.16190612e-01 5.63022614e-01
1.57270774e-01 -6.92212880e-01 -4.95350361e-02 5.46477735e-01
6.03658140e-01 -1.49482191e+00 -2.66964529e-02 -5.43495715e-01
-5.03817499e-01 9.50651586e-01 4.48645979e-01 -2.91962355e-01
1.03368425e+00 3.92265886e-01 -7.56396428e-02 -3.95053834e-01
-8.64732563e-01 -2.45501801e-01 3.67577642e-01 6.59426451e-01
7.30956852e-01 3.28053087e-01 -9.90903601e-02 1.69632912e-01
-7.51864910e-03 -2.23636881e-01 3.14580709e-01 8.04864347e-01
1.05366018e-03 -1.20100689e+00 4.68913078e-01 1.03461516e+00
-4.04471725e-01 -4.09294218e-01 -5.78722715e-01 5.99404335e-01
-5.81592798e-01 9.20406282e-01 -4.59783375e-02 -6.17059946e-01
-1.03173174e-01 -2.68869728e-01 3.53156447e-01 -5.81830859e-01
-5.81492484e-01 -4.81269568e-01 3.19476724e-01 -5.16014397e-01
-4.63159472e-01 -2.61887759e-01 -1.17899048e+00 -5.49872935e-01
-3.02244633e-01 1.12866215e-01 5.37315667e-01 8.47641170e-01
3.98446470e-01 4.79374498e-01 5.22673666e-01 -6.37371600e-01
-2.71596104e-01 -8.67135048e-01 -4.99923378e-01 3.38913172e-01
3.92551094e-01 -4.71919686e-01 -5.49961984e-01 -4.08776015e-01]
|
[7.074853897094727, 6.234947681427002]
|
d211f2b3-dd99-4322-b59d-a6f8cfae4862
|
co-mining-self-supervised-learning-for
|
2012.01950
| null |
https://arxiv.org/abs/2012.01950v2
|
https://arxiv.org/pdf/2012.01950v2.pdf
|
Co-mining: Self-Supervised Learning for Sparsely Annotated Object Detection
|
Object detectors usually achieve promising results with the supervision of complete instance annotations. However, their performance is far from satisfactory with sparse instance annotations. Most existing methods for sparsely annotated object detection either re-weight the loss of hard negative samples or convert the unlabeled instances into ignored regions to reduce the interference of false negatives. We argue that these strategies are insufficient since they can at most alleviate the negative effect caused by missing annotations. In this paper, we propose a simple but effective mechanism, called Co-mining, for sparsely annotated object detection. In our Co-mining, two branches of a Siamese network predict the pseudo-label sets for each other. To enhance multi-view learning and better mine unlabeled instances, the original image and corresponding augmented image are used as the inputs of two branches of the Siamese network, respectively. Co-mining can serve as a general training mechanism applied to most of modern object detectors. Experiments are performed on MS COCO dataset with three different sparsely annotated settings using two typical frameworks: anchor-based detector RetinaNet and anchor-free detector FCOS. Experimental results show that our Co-mining with RetinaNet achieves 1.4%~2.1% improvements compared with different baselines and surpasses existing methods under the same sparsely annotated setting. Code is available at https://github.com/megvii-research/Co-mining.
|
['Xiangyu Zhang', 'Jiale Cao', 'Tong Yang', 'Tiancai Wang']
|
2020-12-03
| null | null | null | null |
['multi-view-learning']
|
['computer-vision']
|
[-1.86679326e-02 1.51204944e-01 -6.22244120e-01 -4.01211530e-01
-7.61552989e-01 -1.68072134e-01 1.75045729e-01 -1.74737364e-01
-4.10577178e-01 5.79590380e-01 -9.63353813e-02 2.04934001e-01
3.83713573e-01 -4.12267029e-01 -7.80290902e-01 -7.43335605e-01
1.29560336e-01 4.50004429e-01 7.52370417e-01 6.41307756e-02
8.37740228e-02 1.37317747e-01 -1.61719620e+00 5.61267495e-01
7.01755881e-01 1.06277847e+00 2.64017880e-01 2.16248497e-01
-1.77475691e-01 7.75590301e-01 -4.39754397e-01 -4.32470381e-01
7.13472486e-01 -2.97267884e-01 -4.44612205e-01 3.29177737e-01
5.85327923e-01 -2.65978068e-01 -2.62829989e-01 1.37774611e+00
3.63558829e-01 -1.76248133e-01 4.83758658e-01 -1.26069582e+00
-5.46162426e-01 7.59215534e-01 -1.50116956e+00 5.32656908e-01
-1.47158086e-01 1.05035089e-01 1.08684373e+00 -1.45082140e+00
5.07176757e-01 1.19418395e+00 5.57567298e-01 6.14932954e-01
-8.74878645e-01 -1.01578462e+00 5.17256916e-01 2.76154101e-01
-1.59770870e+00 -4.42504168e-01 5.80105603e-01 -1.82543784e-01
5.70118785e-01 6.45611733e-02 6.02209330e-01 7.64738679e-01
-3.47794652e-01 1.29229772e+00 8.20376754e-01 -2.96796709e-01
1.22591469e-03 4.51923221e-01 3.38817686e-01 8.62789512e-01
6.85754538e-01 -2.92934310e-02 -6.15198910e-01 -2.90962279e-01
6.07317567e-01 2.73617119e-01 -1.95084184e-01 -5.62787056e-01
-1.13880301e+00 8.49719942e-01 7.42729902e-01 1.40927076e-01
-2.02787578e-01 -1.56270340e-01 4.84723687e-01 3.76938991e-02
5.35659075e-01 9.19640586e-02 -4.47965711e-01 5.14577210e-01
-9.30694282e-01 -6.51580542e-02 5.70836127e-01 1.14755082e+00
7.74324894e-01 5.11297919e-02 -1.93972588e-01 9.15351331e-01
5.12588441e-01 3.20123047e-01 6.46198511e-01 -6.18027031e-01
6.54160976e-01 9.93476152e-01 3.48627046e-02 -7.97191560e-01
-1.94065213e-01 -7.96032548e-01 -6.78277433e-01 1.04702830e-01
5.15470445e-01 -1.69205904e-01 -1.06324565e+00 1.28522635e+00
6.60410643e-01 4.96604353e-01 2.49326695e-02 1.21563113e+00
9.59285319e-01 4.64297682e-01 6.33914843e-02 -1.40869826e-01
1.31635880e+00 -1.46016109e+00 -4.18135077e-01 -4.21904922e-01
7.53487527e-01 -6.65042162e-01 9.72511768e-01 3.47939402e-01
-1.02532089e+00 -4.69334036e-01 -1.11938906e+00 1.86961442e-01
-7.00830519e-02 6.60038352e-01 6.46107793e-01 3.55041564e-01
-5.11324823e-01 1.43045738e-01 -8.58838379e-01 -1.90899894e-01
9.89757955e-01 3.62286955e-01 -3.52703691e-01 -3.14175725e-01
-6.91315532e-01 5.05099297e-01 5.31193495e-01 1.29239902e-01
-9.49823380e-01 -6.35444701e-01 -6.28179550e-01 9.27603915e-02
7.70517290e-01 -4.39269781e-01 1.30255747e+00 -1.19412017e+00
-7.41824508e-01 1.02909565e+00 -1.45471871e-01 -6.03962362e-01
4.35311317e-01 -2.56807834e-01 -2.94984460e-01 1.15412518e-01
4.64817643e-01 8.81087303e-01 7.87254333e-01 -1.31082880e+00
-1.03499186e+00 -6.63964450e-01 -2.50767827e-01 3.35291415e-01
-4.68970895e-01 -2.62274388e-02 -8.06356132e-01 -5.22559345e-01
4.07023787e-01 -9.06404734e-01 -3.93066913e-01 1.79180190e-01
-5.87878764e-01 -3.43561947e-01 1.08231401e+00 -6.93630502e-02
1.00685227e+00 -2.29210567e+00 -2.71785557e-01 8.35852176e-02
4.59901154e-01 5.60125113e-01 -1.50217891e-01 -1.44601777e-01
-5.34761921e-02 -1.57853678e-01 -1.89283654e-01 -3.93404186e-01
-4.56280291e-01 8.23571011e-02 -1.42223313e-01 6.03520632e-01
4.11848128e-01 7.56913185e-01 -8.35840583e-01 -7.55617797e-01
-1.00825593e-01 1.90106526e-01 -4.79814291e-01 -3.23099941e-02
-1.15073055e-01 1.91724852e-01 -5.89821637e-01 1.07762790e+00
7.11046994e-01 -6.32347107e-01 7.22429827e-02 -2.57418454e-01
1.67110100e-01 -9.06979963e-02 -1.50128520e+00 1.56928682e+00
1.94766551e-01 2.44000331e-01 1.13405168e-01 -1.13172114e+00
9.35322583e-01 5.49613535e-02 4.00600404e-01 -5.32046020e-01
5.56630231e-02 2.98435062e-01 1.11950316e-01 -4.30966586e-01
1.35853603e-01 -3.00577041e-02 2.37974912e-01 1.87110394e-01
1.56879842e-01 5.32080173e-01 3.21055979e-01 5.25849760e-01
9.38349962e-01 -1.07847610e-02 4.44753796e-01 -1.01307660e-01
5.79027891e-01 1.77643687e-01 1.10502839e+00 7.71286905e-01
-5.12132287e-01 6.73970163e-01 3.49724829e-01 -5.27412891e-01
-8.78529668e-01 -8.64785552e-01 -1.40096307e-01 1.23687375e+00
4.27845389e-01 -2.69422293e-01 -5.08902550e-01 -1.21501565e+00
1.60987422e-01 1.75629288e-01 -5.60006738e-01 -5.11076748e-02
-4.15941298e-01 -1.01626647e+00 5.66996872e-01 7.02067256e-01
5.56042790e-01 -8.91293168e-01 -3.56206656e-01 -6.55142292e-02
-5.23473546e-02 -1.08261681e+00 -5.67857921e-01 2.60771215e-01
-9.10897315e-01 -1.20506954e+00 -6.75850809e-01 -9.71582532e-01
8.56283665e-01 9.01191711e-01 9.51436281e-01 3.82720947e-01
-3.58782858e-01 -5.68358228e-02 -2.99807817e-01 -7.00011730e-01
5.65187745e-02 8.85658711e-03 1.40019640e-01 2.84409970e-01
8.46322298e-01 -2.86817282e-01 -7.71181345e-01 5.49793661e-01
-7.73050368e-01 -2.16452390e-01 8.97529185e-01 9.42439318e-01
8.99391174e-01 -3.93158376e-01 8.45674038e-01 -1.24825656e+00
-5.45843281e-02 -7.11226523e-01 -6.16323411e-01 1.73688754e-01
-6.40635252e-01 -1.15249231e-01 3.44546199e-01 -5.52182794e-01
-9.81816590e-01 6.24482512e-01 2.70923346e-01 -8.34238946e-01
-1.19919792e-01 1.95587620e-01 -1.57834351e-01 -9.58435014e-02
7.68576384e-01 9.74899530e-02 -5.06053083e-02 -4.22381401e-01
1.43980309e-01 6.46764576e-01 4.30116206e-01 -1.28844246e-01
8.26564550e-01 8.36233139e-01 -3.20109576e-01 -5.48404992e-01
-1.41771495e+00 -1.15203273e+00 -4.53985900e-01 -3.70188691e-02
5.37581146e-01 -1.38999963e+00 -2.75847673e-01 1.68277457e-01
-7.75809884e-01 1.30015075e-01 -3.32092375e-01 5.94813466e-01
-1.13557070e-01 4.29366618e-01 -4.99373674e-01 -6.83673024e-01
-3.80399346e-01 -1.04118574e+00 1.10496366e+00 4.72581685e-01
2.50376582e-01 -3.92156392e-01 -1.94840655e-01 6.66047990e-01
-6.56927377e-02 -1.80797786e-01 2.57179171e-01 -1.00446403e+00
-6.84588432e-01 -2.95622945e-01 -4.59093183e-01 3.52790236e-01
-1.60831034e-01 -9.16651636e-02 -1.08772826e+00 -3.04168284e-01
-5.00451103e-02 -6.11662805e-01 1.23885989e+00 3.05388123e-01
1.12859440e+00 -1.91213951e-01 -6.86707914e-01 5.39469779e-01
1.39003253e+00 -1.06459059e-01 4.10165519e-01 2.16057897e-01
7.60654867e-01 5.06971121e-01 9.88111079e-01 5.30191541e-01
2.59677261e-01 5.30901194e-01 6.79817855e-01 -2.11984634e-01
-1.64598882e-01 -2.13507488e-01 3.24257582e-01 6.14717126e-01
8.88690576e-02 -4.29352708e-02 -7.83018053e-01 7.68623114e-01
-2.01971889e+00 -9.85440016e-01 -4.32665557e-01 1.99193215e+00
6.91369236e-01 3.61849844e-01 3.88691753e-01 7.15170056e-02
9.27043915e-01 -1.67589337e-01 -8.66350949e-01 3.80599558e-01
-2.03261226e-01 -2.56573111e-01 6.33060336e-01 6.54273992e-03
-1.34531295e+00 9.31663692e-01 5.19838953e+00 8.11115503e-01
-9.18852448e-01 3.87852609e-01 7.55378425e-01 -4.84133244e-01
3.10796827e-01 -5.42251989e-02 -1.40261650e+00 4.62030530e-01
5.53986549e-01 1.48403198e-01 -2.62249142e-01 1.28335953e+00
6.84242547e-02 -2.72576958e-01 -9.36133265e-01 1.04261780e+00
2.51896083e-01 -1.30289865e+00 -5.07467426e-02 -9.39710066e-02
9.18294072e-01 3.93598318e-01 9.24564600e-02 4.11292136e-01
1.27990142e-01 -6.91459119e-01 6.22084022e-01 8.85168836e-02
4.72532660e-01 -5.05023420e-01 8.44394684e-01 5.84532857e-01
-1.22907507e+00 -2.51008809e-01 -5.92695057e-01 1.41302273e-01
-4.20224555e-02 6.98053718e-01 -8.96114051e-01 2.22207606e-01
1.01123893e+00 8.66378665e-01 -6.81502283e-01 1.36487114e+00
-7.42726773e-03 8.35344970e-01 -4.65759158e-01 7.05682114e-02
3.39298218e-01 2.31263395e-02 4.14179027e-01 1.08663487e+00
1.04515254e-02 1.49511710e-01 5.15001953e-01 7.61335075e-01
-1.62538320e-01 1.59261987e-01 -5.87027371e-01 3.37503940e-01
4.86724108e-01 1.42788064e+00 -9.70947266e-01 -5.90489686e-01
-8.48814785e-01 7.01584041e-01 4.80655730e-01 1.69826671e-01
-8.89749706e-01 -1.54189795e-01 3.44414771e-01 2.68626750e-01
6.25876188e-01 2.43691325e-01 -3.65876138e-01 -1.37214899e+00
1.75372109e-01 -8.45197618e-01 6.91708446e-01 -5.08437812e-01
-1.45966649e+00 3.21603000e-01 -2.08893329e-01 -1.44136000e+00
2.30111733e-01 -4.45111603e-01 -6.13919318e-01 4.58244622e-01
-1.69262612e+00 -1.13699305e+00 -4.06781822e-01 6.45923138e-01
7.83644080e-01 -4.10426170e-01 2.48488754e-01 5.83593547e-01
-9.24670517e-01 7.22683609e-01 -6.73817694e-02 3.18653375e-01
7.73178399e-01 -1.09672928e+00 -7.55501464e-02 9.02270496e-01
5.66370010e-01 3.86769056e-01 3.18760812e-01 -6.49337947e-01
-1.15502453e+00 -1.29947674e+00 4.02176738e-01 -3.31065625e-01
5.87944031e-01 -2.46638194e-01 -1.14602780e+00 6.73570812e-01
-1.11096606e-01 6.81388736e-01 7.07124531e-01 -8.25190991e-02
-4.33055609e-01 -2.05583856e-01 -1.08410919e+00 3.06003273e-01
1.10860670e+00 -1.03027776e-01 -4.79078948e-01 5.53323805e-01
5.93028367e-01 -3.37733299e-01 -5.12756288e-01 5.71895003e-01
2.80875236e-01 -1.05381668e+00 9.61085856e-01 -5.88019848e-01
1.82592317e-01 -6.62452698e-01 -9.93245766e-02 -8.41664970e-01
-2.41008803e-01 -1.03837974e-01 -3.03181738e-01 1.04003310e+00
5.88619471e-01 -4.66409683e-01 1.25212467e+00 1.98155999e-01
-1.16810292e-01 -9.37002897e-01 -6.55888975e-01 -7.59880841e-01
-3.00661027e-01 -9.58102494e-02 3.23841363e-01 9.40416873e-01
-2.08945483e-01 5.68059266e-01 -1.86470926e-01 4.72770989e-01
8.31798553e-01 1.68005809e-01 7.60766506e-01 -1.15871787e+00
-3.61564547e-01 -2.46202096e-01 -4.55342978e-01 -1.07031620e+00
2.58551482e-02 -1.00207472e+00 2.41676625e-03 -1.08786714e+00
6.60221994e-01 -6.26006663e-01 -4.47607219e-01 6.34174526e-01
-3.92004550e-01 7.18273640e-01 1.02713011e-01 6.39022231e-01
-1.07110322e+00 5.01008570e-01 1.10998595e+00 -1.76255941e-01
-1.77774087e-01 1.59366533e-01 -7.67771900e-01 1.06529629e+00
8.20301056e-01 -7.11252570e-01 -3.28387290e-01 -3.47125202e-01
-1.00895829e-01 -2.88051963e-01 4.27396268e-01 -1.01335537e+00
4.63296086e-01 5.40962517e-02 4.66055602e-01 -8.10450137e-01
1.80539146e-01 -7.85063922e-01 -2.77458787e-01 5.09871423e-01
-2.10750729e-01 -1.56621143e-01 1.07510146e-02 1.00937581e+00
-4.60481942e-01 -3.57518613e-01 9.52086985e-01 -2.96656966e-01
-9.33499932e-01 4.33229595e-01 8.04062933e-02 1.86188072e-01
1.41503572e+00 -3.60579610e-01 -3.16734433e-01 8.79335105e-02
-7.76865721e-01 5.93291461e-01 2.04003349e-01 2.77046919e-01
6.51126921e-01 -1.23451746e+00 -7.81417549e-01 8.61700922e-02
3.70314300e-01 2.97170103e-01 1.29351780e-01 1.03855503e+00
-2.29813024e-01 2.48028755e-01 9.06089172e-02 -9.06167507e-01
-1.38393486e+00 6.59117222e-01 2.80281961e-01 -2.05331668e-01
-6.30957663e-01 1.12927568e+00 6.38690233e-01 -1.92100286e-01
4.69583303e-01 -9.99279469e-02 -2.79057831e-01 -3.56714963e-03
7.17422783e-01 2.13287741e-01 6.30860925e-02 -5.54063141e-01
-4.32581395e-01 2.60547221e-01 -5.15852511e-01 3.93774390e-01
1.29746759e+00 1.68035217e-02 1.12619966e-01 3.39614749e-01
8.81168664e-01 -2.08779722e-01 -1.38545120e+00 -5.71687460e-01
1.85574472e-01 -5.03319502e-01 -2.10743509e-02 -4.53263044e-01
-1.58022881e+00 6.50282979e-01 6.88805163e-01 -3.16295866e-03
9.97860074e-01 3.74420702e-01 6.17918849e-01 4.20342773e-01
2.08963126e-01 -1.01558506e+00 3.62123013e-01 2.07621649e-01
5.73389888e-01 -1.73419511e+00 3.56203876e-02 -7.06402361e-01
-8.99359524e-01 7.18372464e-01 1.31097519e+00 -3.06654036e-01
5.99305928e-01 3.61439943e-01 2.80979201e-02 -3.48336220e-01
-8.19642842e-01 -3.77772242e-01 2.77505934e-01 4.30144697e-01
2.28723615e-01 -8.81239921e-02 -2.12484226e-01 6.09068036e-01
4.12292212e-01 -7.75370151e-02 4.55517024e-01 8.85205805e-01
-7.64051974e-01 -7.58772433e-01 -5.68380415e-01 8.49595547e-01
-5.66324234e-01 -6.48278296e-02 -3.77943099e-01 7.81635046e-01
3.02896827e-01 6.73677683e-01 -4.21398208e-02 -1.03375696e-01
2.69941479e-01 -4.38757278e-02 2.26810738e-01 -1.01500165e+00
-4.90283877e-01 1.59409463e-01 -1.67064518e-01 -5.64449966e-01
-5.60418069e-01 -7.56722927e-01 -1.39722133e+00 1.88115045e-01
-8.39410305e-01 1.58047788e-02 2.50542670e-01 6.84765577e-01
4.12895054e-01 2.03594387e-01 4.77222711e-01 -6.83905602e-01
-8.53381455e-01 -9.40369308e-01 -6.63948357e-01 3.40312183e-01
1.60047337e-01 -8.27300668e-01 -3.11367303e-01 9.43033919e-02]
|
[9.17454719543457, 1.347322702407837]
|
1f8aaa40-4157-4c3f-9434-f7688158f622
|
mgfn-magnitude-contrastive-glance-and-focus
|
2211.15098
| null |
https://arxiv.org/abs/2211.15098v1
|
https://arxiv.org/pdf/2211.15098v1.pdf
|
MGFN: Magnitude-Contrastive Glance-and-Focus Network for Weakly-Supervised Video Anomaly Detection
|
Weakly supervised detection of anomalies in surveillance videos is a challenging task. Going beyond existing works that have deficient capabilities to localize anomalies in long videos, we propose a novel glance and focus network to effectively integrate spatial-temporal information for accurate anomaly detection. In addition, we empirically found that existing approaches that use feature magnitudes to represent the degree of anomalies typically ignore the effects of scene variations, and hence result in sub-optimal performance due to the inconsistency of feature magnitudes across scenes. To address this issue, we propose the Feature Amplification Mechanism and a Magnitude Contrastive Loss to enhance the discriminativeness of feature magnitudes for detecting anomalies. Experimental results on two large-scale benchmarks UCF-Crime and XD-Violence manifest that our method outperforms state-of-the-art approaches.
|
['Yik-Chung Wu', 'Xiaojuan Qi', 'Wilton Fok', 'Baoheng Zhang', 'Zhengzhe Liu', 'Yingxian Chen']
|
2022-11-28
| null | null | null | null |
['video-anomaly-detection', 'anomaly-detection-in-surveillance-videos', 'anomaly-detection-in-surveillance-videos']
|
['computer-vision', 'computer-vision', 'methodology']
|
[ 1.48010075e-01 -4.77691352e-01 -4.34608199e-02 -6.16089880e-01
-4.67338502e-01 -4.94651437e-01 6.59285724e-01 2.10197404e-01
-2.76614219e-01 1.90395594e-01 3.24318290e-01 -1.73432633e-01
2.03411989e-02 -3.83229792e-01 -7.42060661e-01 -4.98638123e-01
-5.86201727e-01 -5.16714752e-01 4.88879502e-01 -4.45446149e-02
2.94410408e-01 3.70180607e-01 -1.48131180e+00 4.24264133e-01
7.96707511e-01 1.13721240e+00 -4.95303035e-01 5.47374725e-01
2.82952100e-01 9.75863874e-01 -6.15301490e-01 -2.74492919e-01
7.13142514e-01 -3.52798879e-01 -3.30894828e-01 3.54654819e-01
1.00766766e+00 -1.01718700e+00 -6.90045595e-01 1.22990561e+00
1.52130276e-01 5.90343103e-02 4.86541539e-01 -1.50380540e+00
-6.37210786e-01 -2.67602652e-02 -9.32036877e-01 1.07199717e+00
6.71645403e-01 4.66154099e-01 9.69918907e-01 -8.69625151e-01
3.78453076e-01 1.30533040e+00 7.21608818e-01 2.74885535e-01
-8.82391155e-01 -6.04912579e-01 7.19905317e-01 5.50540626e-01
-1.28781104e+00 -4.45267260e-01 7.48824954e-01 -4.60387886e-01
9.56722319e-01 2.19091743e-01 5.10688245e-01 1.22472394e+00
1.40618861e-01 8.44277561e-01 7.95926750e-01 -1.29269481e-01
6.48295954e-02 -4.24366772e-01 1.82824597e-01 8.31437230e-01
5.20467818e-01 9.81263816e-02 -4.84535515e-01 -4.24860895e-01
5.92452407e-01 4.97976303e-01 -2.32556820e-01 -2.71594226e-01
-8.69949520e-01 7.11060584e-01 3.71921122e-01 9.42634344e-02
-4.34993118e-01 1.49217118e-02 6.25001669e-01 4.81960326e-01
6.68933392e-01 3.80350322e-01 -3.79591614e-01 -3.22132915e-01
-7.64066935e-01 3.14436972e-01 2.86795229e-01 6.49914742e-01
4.56449538e-01 1.06501475e-01 -4.79541153e-01 5.74583888e-01
1.63979635e-01 2.48021841e-01 7.10027814e-02 -1.11459839e+00
5.11024058e-01 7.76203811e-01 6.05155937e-02 -1.46083546e+00
-3.07091624e-01 -3.02395284e-01 -5.01427948e-01 9.49866399e-02
5.16143441e-01 -1.61554009e-01 -8.60715687e-01 1.74392521e+00
3.69246095e-01 8.37305307e-01 -2.69105762e-01 9.53678370e-01
6.36278570e-01 4.05519217e-01 1.24301575e-01 -1.41682878e-01
1.09727728e+00 -7.73040771e-01 -7.28796661e-01 -3.71603996e-01
6.35354400e-01 -5.12277722e-01 1.05391335e+00 3.70460212e-01
-7.73660362e-01 -3.63414347e-01 -8.20595860e-01 3.76752317e-01
-2.70496309e-01 -1.77706093e-01 6.25838816e-01 5.61816990e-01
-7.02332795e-01 3.98992032e-01 -9.71222341e-01 -4.33305949e-01
7.72124946e-01 -1.76687256e-01 -4.22050327e-01 -2.08468601e-01
-9.36400354e-01 5.29070139e-01 1.59675166e-01 1.46875590e-01
-9.48648691e-01 -7.77633786e-01 -1.02483773e+00 -8.16967413e-02
5.91342926e-01 -2.04752564e-01 9.83729720e-01 -1.07888031e+00
-7.90465891e-01 5.95316410e-01 -2.28104591e-01 -5.78080893e-01
5.27314961e-01 -6.53763413e-01 -5.28831959e-01 4.65281248e-01
8.50408971e-02 4.26037848e-01 8.93461943e-01 -9.43228245e-01
-9.03135836e-01 -3.84534270e-01 3.02952021e-01 2.74344534e-02
-7.35673189e-01 3.15944850e-01 -4.58192199e-01 -1.02552128e+00
-5.86273195e-03 -6.53665483e-01 -3.39157492e-01 2.54477084e-01
-2.25439608e-01 -7.48429671e-02 1.30684137e+00 -6.94134057e-01
1.50497019e+00 -2.48155236e+00 -2.59679258e-01 1.75300971e-01
2.89953768e-01 3.80820453e-01 -2.43711650e-01 1.31028384e-01
-4.95002605e-02 -9.69745889e-02 -1.66982338e-01 -8.34574774e-02
-2.34931007e-01 2.18604699e-01 -4.01113540e-01 7.53062963e-01
7.06785858e-01 6.14456892e-01 -1.10846102e+00 -4.19664800e-01
2.13442385e-01 2.43781701e-01 -9.50192034e-01 1.74264699e-01
1.03772394e-01 4.00617599e-01 -4.82319504e-01 1.08318329e+00
6.76575601e-01 -1.51012883e-01 -3.21369529e-01 -6.29494935e-02
1.56157613e-01 -1.12099259e-03 -9.74905908e-01 1.32891667e+00
2.25370347e-01 8.16519022e-01 -8.41073468e-02 -1.06963670e+00
4.05383945e-01 1.88663766e-01 6.52166069e-01 -7.57748902e-01
-1.55548289e-01 -3.56743522e-02 8.82223099e-02 -7.78218269e-01
3.71637911e-01 6.07820868e-01 5.00035025e-02 -1.51978284e-01
-1.58785194e-01 5.06540775e-01 3.56995791e-01 3.17548126e-01
1.68564332e+00 2.24529151e-02 1.43800184e-01 -6.85463427e-03
4.27361220e-01 -2.06777856e-01 8.29184353e-01 1.00127947e+00
-7.42579401e-01 6.48682356e-01 8.14284503e-01 -6.41907513e-01
-8.05340409e-01 -1.09770525e+00 -1.75559878e-01 1.20577228e+00
1.21636152e-01 -5.56837499e-01 -6.97425306e-01 -1.08281684e+00
-7.92783592e-03 4.53504831e-01 -7.72220731e-01 -2.37002909e-01
-5.90736270e-01 -8.52185071e-01 6.28860533e-01 6.61241472e-01
5.98154187e-01 -7.80223012e-01 -7.25616276e-01 -3.49383131e-02
-1.99204996e-01 -1.58898723e+00 -5.56429744e-01 -4.58327949e-01
-6.14599407e-01 -1.33420396e+00 -3.43911737e-01 -2.24590346e-01
8.29815865e-01 5.66263974e-01 9.14399385e-01 2.96424478e-01
-6.05689347e-01 7.24908471e-01 -6.18307173e-01 -3.83159071e-01
-2.51417775e-02 -3.31555516e-01 1.41052276e-01 2.09923655e-01
6.94032609e-01 -4.67312813e-01 -8.45738769e-01 2.24282563e-01
-1.04545069e+00 -5.34743965e-01 4.51202452e-01 7.81544864e-01
2.15121359e-01 -3.44291255e-02 4.18132216e-01 -5.97631156e-01
4.62814033e-01 -6.89323008e-01 -5.83349228e-01 7.70830316e-04
-3.20776403e-01 -3.54600281e-01 5.65226078e-01 -4.12660748e-01
-8.53087783e-01 -4.47152443e-02 4.60514016e-02 -7.33135283e-01
-4.06170130e-01 1.34068623e-01 1.72726840e-01 -1.65015638e-01
5.57734013e-01 1.58522010e-01 -2.08450049e-01 -9.31413472e-02
-5.91821857e-02 1.88476399e-01 5.87762773e-01 -2.71720767e-01
6.78373754e-01 7.71345675e-01 5.61228283e-02 -8.58775795e-01
-1.01303375e+00 -6.61321878e-01 -3.87311816e-01 -2.74554133e-01
8.13710570e-01 -1.04958010e+00 -3.20383430e-01 5.61042726e-01
-1.03903770e+00 6.97513483e-03 -1.95665047e-01 4.49575603e-01
-1.74210906e-01 7.63613760e-01 -6.35773897e-01 -7.30841935e-01
1.11068375e-02 -1.08830512e+00 1.16873288e+00 4.02397141e-02
-1.18195035e-01 -7.32615054e-01 5.50696328e-02 6.02116995e-02
4.80036974e-01 6.85641527e-01 4.20100361e-01 -6.08261228e-01
-4.58926529e-01 -4.22951877e-01 -4.56120014e-01 4.16723728e-01
3.55516642e-01 2.94558525e-01 -9.57225800e-01 -3.63887787e-01
-2.93914676e-01 -1.15607746e-01 1.11731243e+00 3.99794281e-01
1.49571550e+00 -5.11977434e-01 -1.89754367e-01 8.09900999e-01
1.02706075e+00 2.88883634e-02 7.38224328e-01 4.25019383e-01
8.62009346e-01 4.95796263e-01 7.80091524e-01 7.85275877e-01
2.43721992e-01 6.77410662e-01 6.29686952e-01 -1.89769998e-01
6.76148534e-02 -4.74493466e-02 6.71606123e-01 1.07585020e-01
-1.58941001e-01 -1.59657091e-01 -8.39205623e-01 6.42674744e-01
-1.99497342e+00 -1.32142591e+00 -1.43659696e-01 1.96074963e+00
2.89523244e-01 2.23446950e-01 3.36014420e-01 6.28771111e-02
4.06841606e-01 5.28636992e-01 -3.91785175e-01 -2.16263115e-01
-1.09906286e-01 -4.01302189e-01 4.37899888e-01 7.74315745e-02
-1.56618810e+00 7.10475683e-01 7.13942242e+00 6.24669433e-01
-9.48887765e-01 -9.84697193e-02 7.10152924e-01 -4.18339849e-01
1.28312156e-01 -3.08200508e-01 -4.63160157e-01 7.45704234e-01
6.44589424e-01 2.50323415e-01 1.49470612e-01 8.93282294e-01
3.70916337e-01 -5.16879484e-02 -1.01933038e+00 8.08299243e-01
1.29930452e-01 -8.25192273e-01 1.71156034e-01 -1.12258442e-01
7.02401638e-01 5.24240062e-02 2.78081596e-01 1.53336525e-01
2.76410412e-02 -8.53990495e-01 5.00459552e-01 3.39719474e-01
3.27241689e-01 -5.89592695e-01 7.80880570e-01 -4.70872372e-02
-1.18061507e+00 -5.09017825e-01 -2.31096104e-01 -3.11365813e-01
2.34491602e-02 5.15124381e-01 -5.62286079e-01 2.86520571e-01
1.22324038e+00 1.05416632e+00 -1.00190806e+00 1.11902285e+00
-5.57293603e-03 8.63109946e-01 -2.42654219e-01 3.42170417e-01
5.87575138e-01 8.66314769e-02 8.34240317e-01 1.52020073e+00
3.19108546e-01 1.07547231e-01 3.30033332e-01 3.69621038e-01
1.86674029e-01 -1.14128061e-01 -9.06826377e-01 2.16653541e-01
2.58505255e-01 9.42152083e-01 -4.74862069e-01 -2.66796440e-01
-9.75589991e-01 8.75321031e-01 1.95740342e-01 5.29286265e-01
-1.00632548e+00 -7.40070865e-02 9.96228397e-01 1.64064020e-01
5.28487027e-01 -1.90425679e-01 6.83634682e-03 -1.33532107e+00
4.89141732e-01 -1.09652162e+00 7.66004086e-01 -3.21154535e-01
-1.47008550e+00 3.53501737e-01 3.12921524e-01 -1.54705715e+00
-3.47185135e-01 -5.22208631e-01 -7.97520518e-01 6.25017807e-02
-1.52539229e+00 -9.19846177e-01 -5.57346940e-01 7.63465762e-01
7.22011864e-01 -2.62565106e-01 4.50288743e-01 5.30012369e-01
-9.04429972e-01 6.87395394e-01 -1.48819000e-01 4.08070296e-01
7.06042528e-01 -1.00788319e+00 4.80469972e-01 1.51682317e+00
6.19917549e-02 3.41990262e-01 7.95834124e-01 -5.80494225e-01
-1.19054663e+00 -1.29750717e+00 1.88375935e-01 -7.21790493e-01
7.59207547e-01 -2.20344990e-01 -1.18611586e+00 7.59169817e-01
1.07907981e-01 6.85605824e-01 4.85821545e-01 -1.11999415e-01
-6.94298029e-01 -8.42617154e-02 -1.23638093e+00 6.33499980e-01
1.36308396e+00 -3.47603738e-01 -4.11379576e-01 1.93732336e-01
4.83265638e-01 -3.19734871e-01 -4.10281062e-01 7.58555770e-01
4.88926262e-01 -1.29814136e+00 1.07367623e+00 -8.26391101e-01
4.67683643e-01 -2.67969728e-01 -1.81598663e-01 -1.09207785e+00
-3.87084574e-01 -4.84248161e-01 -5.13471723e-01 1.10614753e+00
9.10724998e-02 -6.25931203e-01 4.49959993e-01 4.88781154e-01
-7.08915368e-02 -7.42963552e-01 -9.84780788e-01 -8.65879476e-01
-4.55613613e-01 -5.75606287e-01 4.45234448e-01 9.68588173e-01
-1.10559843e-01 -4.08339173e-01 -5.91784060e-01 6.01411164e-01
6.35419011e-01 -3.59413207e-01 6.33649826e-01 -9.52892363e-01
-1.48950443e-01 -4.87233818e-01 -9.58696544e-01 -8.73623729e-01
4.29423247e-03 -1.51444197e-01 4.13073525e-02 -9.40772951e-01
4.06073302e-01 8.01546648e-02 -6.27173901e-01 3.87339890e-01
-6.61226511e-01 4.19274002e-01 -1.00905500e-01 1.57041047e-02
-1.15380681e+00 3.34507674e-01 8.05815935e-01 2.77964994e-02
5.52959228e-03 -1.93827197e-01 -5.28090179e-01 1.06980336e+00
6.94055498e-01 -4.36405897e-01 -3.48407984e-01 -5.20500302e-01
-4.92734350e-02 -4.21123981e-01 8.10847640e-01 -1.06523407e+00
-1.90484766e-02 -2.67092258e-01 5.50581813e-01 -4.40373391e-01
2.13303894e-01 -8.39536011e-01 -6.61661088e-01 3.67824227e-01
-2.84469515e-01 4.68328863e-01 3.16568494e-01 8.93163860e-01
-3.63526642e-01 1.95170894e-01 5.68552434e-01 1.57922119e-01
-1.00573421e+00 4.60764885e-01 -4.23827082e-01 1.95892990e-01
1.20452905e+00 -9.02065039e-02 -4.27869648e-01 -5.05668461e-01
-6.37100190e-02 3.37833524e-01 4.15543139e-01 7.18186736e-01
7.77170420e-01 -1.33210254e+00 -8.03131759e-01 3.19696397e-01
3.58799100e-01 -3.95301729e-01 3.78937691e-01 8.93155575e-01
-4.24512863e-01 1.65412903e-01 -3.03783357e-01 -8.20438206e-01
-1.33293521e+00 6.12384737e-01 2.55121142e-01 -1.67720348e-01
-6.33658111e-01 6.83878541e-01 3.67902905e-01 4.92336415e-02
4.76493001e-01 -4.02799577e-01 -2.37826094e-01 -2.43108347e-01
9.66250420e-01 4.55651551e-01 -1.70057073e-01 -6.93557560e-01
-6.21376991e-01 5.52961528e-01 -1.84948340e-01 3.14544737e-01
1.24785864e+00 -2.74303649e-02 2.17925772e-01 6.24159053e-02
9.71518457e-01 7.83309042e-02 -1.71922779e+00 -2.60474205e-01
6.34537963e-03 -1.02748811e+00 -2.24231817e-02 -3.53357553e-01
-1.28688407e+00 5.76337934e-01 8.18035662e-01 4.33466166e-01
1.50815701e+00 -4.22914363e-02 7.01917708e-01 4.78966117e-01
3.62255275e-02 -9.80636001e-01 4.89957541e-01 4.14117932e-01
7.56512463e-01 -1.72442985e+00 9.11164433e-02 -4.30385172e-01
-6.21118248e-01 9.72249508e-01 1.05469298e+00 -2.62999564e-01
5.36492586e-01 2.06637815e-01 8.41799378e-03 -3.16948444e-01
-7.24195242e-01 -1.46034464e-01 5.15133381e-01 5.74152768e-01
3.34923178e-01 -3.92085403e-01 -1.22936867e-01 2.26192996e-01
3.53492588e-01 -3.19830656e-01 3.60423774e-01 1.04381633e+00
-2.63680845e-01 -5.19184411e-01 -4.52423394e-01 6.49251163e-01
-9.56585109e-01 9.23636407e-02 -2.85300612e-01 6.95817232e-01
1.59940645e-01 1.09948707e+00 3.49429280e-01 -3.62142026e-01
5.86352050e-01 -2.49438569e-01 2.28404045e-01 -2.65114069e-01
-3.93701762e-01 -1.86536893e-01 -4.80575785e-02 -1.20616579e+00
-4.89806443e-01 -7.83499837e-01 -8.72513294e-01 -1.67225257e-01
-7.44153187e-02 -3.16404134e-01 7.16655329e-02 8.98695648e-01
5.23601532e-01 5.24321437e-01 7.44602859e-01 -6.33123577e-01
-5.39308012e-01 -7.89910376e-01 -2.61535734e-01 9.72642839e-01
8.38356853e-01 -8.05955231e-01 -5.74163735e-01 -5.65964356e-02]
|
[7.864446640014648, 1.528595209121704]
|
b457b544-f3c0-4b16-8a73-c24733f1d1c0
|
phonetic-and-visual-priors-for-decipherment
|
2005.02517
| null |
https://arxiv.org/abs/2005.02517v1
|
https://arxiv.org/pdf/2005.02517v1.pdf
|
Phonetic and Visual Priors for Decipherment of Informal Romanization
|
Informal romanization is an idiosyncratic process used by humans in informal digital communication to encode non-Latin script languages into Latin character sets found on common keyboards. Character substitution choices differ between users but have been shown to be governed by the same main principles observed across a variety of languages---namely, character pairs are often associated through phonetic or visual similarity. We propose a noisy-channel WFST cascade model for deciphering the original non-Latin script from observed romanized text in an unsupervised fashion. We train our model directly on romanized data from two languages: Egyptian Arabic and Russian. We demonstrate that adding inductive bias through phonetic and visual priors on character mappings substantially improves the model's performance on both languages, yielding results much closer to the supervised skyline. Finally, we introduce a new dataset of romanized Russian, collected from a Russian social network website and partially annotated for our experiments.
|
['Taylor Berg-Kirkpatrick', 'Maria Ryskina', 'Matthew R. Gormley']
|
2020-05-05
|
phonetic-and-visual-priors-for-decipherment-1
|
https://aclanthology.org/2020.acl-main.737
|
https://aclanthology.org/2020.acl-main.737.pdf
|
acl-2020-6
|
['decipherment']
|
['natural-language-processing']
|
[ 4.14344579e-01 -3.52846347e-02 -2.78894752e-01 -5.29524744e-01
-5.27442753e-01 -1.15956295e+00 7.97115445e-01 -3.12992066e-01
-6.23849511e-01 6.58878446e-01 4.91930604e-01 -5.11209548e-01
2.34465554e-01 -3.65239948e-01 -6.57025933e-01 -7.23923445e-02
5.44726312e-01 8.84612858e-01 -2.21054345e-01 -3.35672677e-01
1.02885030e-01 2.81473041e-01 -1.09491229e+00 8.16120028e-01
7.38924265e-01 -2.18286272e-02 1.24558061e-01 8.77276838e-01
-1.57316551e-01 3.68111104e-01 -3.76790434e-01 -9.79909658e-01
3.47032309e-01 -5.60581148e-01 -9.38161373e-01 -6.83074668e-02
8.00271809e-01 -2.18003452e-01 -4.40613031e-01 1.09059954e+00
2.38170370e-01 -1.47875965e-01 1.08777332e+00 -4.84581232e-01
-1.04920638e+00 1.35669076e+00 -3.24444622e-01 2.37764064e-02
6.08317196e-01 1.94376349e-01 1.43910062e+00 -1.03555882e+00
8.51961493e-01 1.57467306e+00 9.08400834e-01 5.97909153e-01
-1.78953266e+00 -5.60907662e-01 -5.31116948e-02 -1.89589202e-01
-1.59228766e+00 -5.92741191e-01 3.66447866e-01 -6.03040576e-01
8.92102659e-01 5.04240930e-01 5.82081139e-01 1.72726822e+00
-5.67114174e-01 8.99365187e-01 1.24926496e+00 -6.24209344e-01
-4.49960500e-01 4.28211302e-01 1.31450390e-04 8.10742795e-01
7.23037943e-02 -1.59103438e-01 -8.47574592e-01 -1.85019702e-01
8.96781445e-01 -4.95364249e-01 -2.57641941e-01 7.11107776e-02
-1.45247495e+00 4.83085930e-01 -9.15641785e-02 3.17796707e-01
9.09873247e-02 1.61489531e-01 2.69674957e-01 5.48328221e-01
1.68541417e-01 7.53614902e-01 -3.95264179e-01 -6.65035665e-01
-6.61668897e-01 1.53955117e-01 9.14514184e-01 1.20025289e+00
5.31891882e-01 -4.26290035e-02 3.81331034e-02 1.26051235e+00
3.83185387e-01 8.43331933e-01 5.01201928e-01 -7.37662494e-01
7.02140093e-01 1.46646038e-01 1.97988793e-01 -5.86843371e-01
-9.42520052e-02 -5.02657071e-02 -4.57241118e-01 -1.72239438e-01
1.15484953e+00 -9.04692784e-02 -7.84451365e-01 1.84032440e+00
-3.64057064e-01 -2.38183379e-01 -1.08990282e-01 7.86512554e-01
4.88582671e-01 4.41216469e-01 8.34288374e-02 3.20071518e-01
1.41671050e+00 -6.99772000e-01 -3.47201228e-01 -2.26806805e-01
4.82586056e-01 -9.11072612e-01 1.89431548e+00 6.39604926e-01
-9.03889239e-01 -6.28581703e-01 -9.36920643e-01 -3.46204549e-01
-2.50483781e-01 5.09756446e-01 5.49594641e-01 1.11911440e+00
-8.56217325e-01 7.89977849e-01 -5.91756463e-01 -8.42757523e-01
3.58911246e-01 5.49582802e-02 -4.59715009e-01 3.45084846e-01
-1.00263524e+00 7.42805362e-01 1.70514658e-01 -4.09918614e-02
-6.45098805e-01 -3.86400580e-01 -7.20311403e-01 -1.34502143e-01
-1.04001351e-01 -4.56025392e-01 1.45866954e+00 -1.50019217e+00
-1.86874378e+00 1.36169004e+00 -1.66946158e-01 -1.28183469e-01
6.45750344e-01 -4.89583194e-01 -4.30430263e-01 -1.82078153e-01
-7.63713419e-02 5.53882658e-01 9.72799838e-01 -1.12554812e+00
-4.20380890e-01 -3.97410728e-02 -2.22103387e-01 5.01520276e-01
-3.46342921e-01 1.77495614e-01 -8.73325527e-01 -1.02644134e+00
-3.85360979e-02 -1.13512218e+00 9.81987640e-02 -1.48213491e-01
-7.41830945e-01 -5.08609302e-02 8.70578513e-02 -1.07793367e+00
1.14743924e+00 -2.11257219e+00 4.45323884e-01 6.76712513e-01
5.73517755e-02 1.26947075e-01 -2.17683971e-01 5.09314775e-01
2.08464712e-02 5.13472319e-01 -3.17915648e-01 -6.19919837e-01
2.17543259e-01 2.02624872e-01 -7.07533300e-01 5.59821069e-01
1.08585730e-01 9.36321259e-01 -9.26701725e-01 -1.07857436e-01
-1.39133096e-01 2.04391167e-01 -4.72177565e-01 1.67318299e-01
-3.63817364e-01 5.89279175e-01 1.41465753e-01 4.98279095e-01
3.39646310e-01 -1.54029444e-01 8.13038468e-01 2.43517354e-01
1.31266668e-01 7.11669683e-01 -7.98420727e-01 1.72334003e+00
-4.56936806e-01 8.13052654e-01 -2.51341194e-01 -3.02568585e-01
6.64552271e-01 1.41649067e-01 -3.53666455e-01 -4.33084220e-01
-4.10375036e-02 3.88991266e-01 1.27883345e-01 -2.40542471e-01
7.58714616e-01 1.05076104e-01 -4.26062196e-01 8.56571376e-01
8.68502632e-02 -3.14205587e-01 -1.33868987e-02 1.58663422e-01
5.89742541e-01 3.79701585e-01 3.16761315e-01 -3.48999530e-01
3.00350517e-01 -1.21336930e-01 3.64507884e-01 1.31055689e+00
2.48119727e-01 7.20633090e-01 5.35959065e-01 -1.52324900e-01
-1.45258951e+00 -1.47285402e+00 -9.37779769e-02 1.23797989e+00
-4.58427034e-02 -6.18650913e-01 -8.60493898e-01 -6.06569231e-01
1.07843660e-01 6.72824979e-01 -4.48676109e-01 3.17967206e-01
-7.92716444e-01 -1.87002301e-01 1.45892096e+00 4.36041921e-01
1.77277818e-01 -1.04455817e+00 6.38815165e-02 -9.11294017e-03
-9.37729478e-02 -1.07441914e+00 -7.69256771e-01 5.45135774e-02
-4.35445994e-01 -8.92421722e-01 -7.93999612e-01 -8.92062187e-01
6.50562286e-01 -8.05435404e-02 1.22065651e+00 1.30823642e-01
-4.93878387e-02 3.02121609e-01 -2.18555883e-01 -1.87047318e-01
-7.87545979e-01 5.02363801e-01 4.57124382e-01 -1.13091171e-02
6.04269147e-01 -2.96135962e-01 1.00531667e-01 3.02824587e-01
-5.78174472e-01 3.60806227e-01 2.10255176e-01 8.83001626e-01
2.53180206e-01 -6.62831008e-01 -1.16890641e-02 -1.53969514e+00
6.17905021e-01 -4.10160929e-01 -4.81876850e-01 3.03151399e-01
-3.07377100e-01 1.52285725e-01 8.92673075e-01 -5.67308307e-01
-1.17339253e+00 1.21586554e-01 -2.39889547e-01 4.84303758e-02
-2.83051878e-01 3.19068074e-01 -1.71786621e-01 1.99806675e-01
1.00869429e+00 2.65656471e-01 -3.48504871e-01 -7.57404625e-01
7.98690200e-01 1.19705868e+00 9.14920688e-01 -9.11782861e-01
1.05743372e+00 3.62458020e-01 -7.99642920e-01 -1.23048711e+00
-4.17839646e-01 -2.50931799e-01 -7.17706919e-01 1.42607763e-01
8.85464966e-01 -1.10096371e+00 -7.86707163e-01 7.65824258e-01
-1.25250244e+00 -7.14080572e-01 -2.20653966e-01 3.67091775e-01
-5.50557315e-01 6.27709031e-01 -8.85660946e-01 -6.94929004e-01
8.62233415e-02 -8.54561388e-01 8.79900813e-01 2.00874209e-01
-1.07698178e+00 -9.76819575e-01 2.37883955e-01 1.72675937e-01
1.20085858e-01 -4.68015581e-01 1.20243943e+00 -6.48730636e-01
-3.27939004e-01 3.21180075e-01 -2.32368201e-01 2.03576431e-01
2.21997604e-01 1.89109787e-01 -1.12805080e+00 -9.90435928e-02
-6.39030337e-01 -4.43955898e-01 8.81635666e-01 -3.31637233e-01
9.99220967e-01 -1.31108850e-01 -1.05754793e-01 9.70250726e-01
9.32723105e-01 -1.09263755e-01 4.52159286e-01 1.73391942e-02
9.09341633e-01 6.11414909e-01 1.09262288e-01 3.72361034e-01
1.86059505e-01 5.84542334e-01 -3.02243620e-01 7.51388296e-02
-2.38792747e-01 -9.64338481e-01 6.45453393e-01 8.92963827e-01
1.46736372e-02 -4.57147449e-01 -1.04884446e+00 5.32451093e-01
-1.79861939e+00 -9.68414962e-01 -7.69999996e-02 2.28186703e+00
1.32561135e+00 1.55929834e-01 2.36482382e-01 -2.11329013e-01
7.40971982e-01 -4.73084673e-02 -2.25764081e-01 -6.59640312e-01
-6.58305943e-01 4.56064194e-01 6.00373507e-01 9.08828080e-01
-1.01887751e+00 1.60348725e+00 7.44835949e+00 7.48767316e-01
-9.70354855e-01 -1.20584697e-01 5.07785559e-01 8.92421529e-02
-8.40923786e-01 -2.38938350e-02 -8.29340339e-01 4.51500237e-01
8.21969032e-01 6.01787716e-02 1.14179122e+00 4.37159091e-01
-1.12728171e-01 9.56960693e-02 -1.43452024e+00 1.31932831e+00
1.56708479e-01 -1.29637575e+00 8.34301636e-02 -1.59091264e-01
6.69843853e-01 3.01726341e-01 3.37344438e-01 8.88499692e-02
9.37333345e-01 -1.38873553e+00 1.22944653e+00 5.66642880e-01
1.42322147e+00 -3.71250808e-01 -4.56974953e-02 1.36503264e-01
-7.28686154e-01 1.52492940e-01 -1.36336148e-01 -9.95580927e-02
-1.15392260e-01 -1.55678138e-01 -1.19567549e+00 -9.03304443e-02
5.47280610e-01 8.34060848e-01 -7.99197018e-01 5.79305530e-01
-7.94851601e-01 1.04226184e+00 -3.78722638e-01 -2.60767043e-01
3.80369984e-02 -3.73985589e-01 6.24028087e-01 1.63930094e+00
1.19250357e-01 -3.96875978e-01 -2.11214364e-01 1.12133551e+00
-2.60341376e-01 2.05625281e-01 -8.80024910e-01 -5.99191844e-01
5.71812630e-01 9.30483878e-01 -5.80144405e-01 -2.67755359e-01
-4.91239280e-01 1.55071604e+00 5.34166038e-01 6.90195739e-01
-5.91016650e-01 -4.23163652e-01 9.21198785e-01 9.83490050e-02
1.75225928e-01 -6.63512051e-01 -4.83392447e-01 -1.49108088e+00
-2.59674013e-01 -1.34752512e+00 1.38467342e-01 -7.58624494e-01
-1.63705409e+00 5.69694996e-01 -2.66772360e-01 -9.03039038e-01
-4.05303866e-01 -9.82696176e-01 -2.52288401e-01 1.28670025e+00
-7.85959542e-01 -1.19958615e+00 2.37059724e-02 5.55192471e-01
5.21262109e-01 -6.85688555e-01 1.04966426e+00 2.06787691e-01
-3.33414882e-01 9.20215905e-01 3.12797964e-01 6.54081523e-01
1.02444875e+00 -1.64254105e+00 1.34504378e+00 9.36314344e-01
9.02786732e-01 1.13741064e+00 5.87938428e-01 -8.80202651e-01
-1.15045130e+00 -7.37022221e-01 1.10772502e+00 -8.62682879e-01
7.06001818e-01 -1.00893331e+00 -5.98651767e-01 1.01288736e+00
4.08395976e-01 -5.41159153e-01 9.22666609e-01 4.63986099e-01
-9.53945160e-01 5.22619426e-01 -6.93256199e-01 1.35014963e+00
1.45391047e+00 -1.49556875e+00 -6.39459133e-01 2.17670873e-01
4.88664001e-01 -3.19425344e-01 -3.26165587e-01 -3.32782060e-01
8.56636941e-01 -5.19717515e-01 5.85151076e-01 -1.01871979e+00
2.32264623e-01 -3.58069360e-01 -3.34831148e-01 -1.44188190e+00
-4.16015536e-01 -1.23733640e+00 3.86007875e-01 1.12715518e+00
7.55419314e-01 -4.44674313e-01 7.39343286e-01 2.18416065e-01
2.09389940e-01 8.30667764e-02 -7.49734998e-01 -5.93320251e-01
1.25145197e-01 -6.71921313e-01 4.22471642e-01 1.15126300e+00
9.91135240e-02 6.33284330e-01 -5.52306950e-01 -1.12356700e-01
3.17335546e-01 -2.76326984e-01 9.48934734e-01 -1.00190020e+00
-7.95902252e-01 -5.28733611e-01 -8.28720257e-02 -1.49879062e+00
4.65407580e-01 -1.34209549e+00 1.06861509e-01 -6.95069373e-01
8.57238322e-02 -5.54135740e-01 2.36432061e-01 4.22686964e-01
1.31321680e-02 6.44743264e-01 1.82517499e-01 2.90784776e-01
-3.68670523e-01 1.21481955e-01 6.42528892e-01 -1.23433068e-01
-1.62779927e-01 -1.45410314e-01 -8.42177868e-01 9.53985393e-01
4.99419987e-01 -3.50712478e-01 -2.57337153e-01 -8.13268363e-01
8.06179523e-01 -2.69263655e-01 6.08682260e-02 -6.16082549e-01
-4.67351079e-02 -1.39340386e-01 4.65162694e-01 -3.91264930e-02
2.54506439e-01 -5.32539189e-01 3.97187062e-02 1.95949852e-01
-6.42350495e-01 1.79509193e-01 -1.07728755e-02 4.73012537e-01
2.03619644e-01 -2.64076144e-01 6.48920834e-01 -1.22936875e-01
-6.95531130e-01 -3.85089405e-02 -8.97873998e-01 3.22030157e-01
3.71901035e-01 -2.24374160e-01 -2.29955375e-01 -5.82827687e-01
-7.30424702e-01 -2.26520270e-01 8.48054707e-01 6.76622570e-01
3.13686103e-01 -1.34566343e+00 -7.18624115e-01 6.13702238e-01
3.75866950e-01 -6.95382833e-01 -3.76411617e-01 3.48979503e-01
-1.03100657e+00 2.01574892e-01 -1.06395014e-01 -5.37257195e-01
-1.34412873e+00 -4.97993790e-02 3.44951391e-01 9.70286801e-02
-3.44635606e-01 1.06442273e+00 -7.13209435e-02 -1.03329635e+00
2.35779688e-01 -2.23031133e-01 1.04441762e-01 -8.42839256e-02
3.33050787e-01 3.98697965e-02 -3.23000371e-01 -7.86780417e-01
-2.26040468e-01 4.14327323e-01 -2.69973487e-01 -7.46651292e-01
8.94988537e-01 -2.45072514e-01 -7.37417117e-02 7.89905131e-01
7.75852978e-01 7.27597237e-01 -9.48680460e-01 -6.18475735e-01
3.34182501e-01 -5.78912258e-01 -9.20701802e-01 -8.28121006e-01
-1.38779417e-01 7.46725023e-01 2.78822988e-01 -1.18756905e-01
3.44162166e-01 -6.76999539e-02 5.62500179e-01 8.76433671e-01
2.33138263e-01 -1.37275243e+00 3.75621370e-03 1.08332705e+00
6.48547709e-01 -1.07837033e+00 -2.96230882e-01 -3.61191630e-01
-1.02435100e+00 1.13234138e+00 4.48914409e-01 -3.88433039e-02
3.80967408e-01 2.44192868e-01 2.90521473e-01 3.04867506e-01
-3.49320740e-01 -1.34185061e-01 3.28633964e-01 7.50117064e-01
7.79224157e-01 2.72528350e-01 -1.05769180e-01 7.56404877e-01
-5.92246413e-01 -4.74142939e-01 4.81219858e-01 3.56728315e-01
-5.31230159e-02 -1.46355164e+00 -2.93579698e-01 3.04727018e-01
-4.65857208e-01 -6.30589604e-01 -9.78783846e-01 5.78312635e-01
5.28533235e-02 6.16783023e-01 3.63641083e-01 -5.29392600e-01
5.66793568e-02 3.20964307e-01 7.53321886e-01 -1.02410889e+00
-9.51384664e-01 -1.06819071e-01 4.50127602e-01 -2.66355515e-01
1.90667510e-01 -8.49271774e-01 -1.03519142e+00 -4.43526298e-01
2.42566571e-01 -4.36015725e-02 3.47609133e-01 8.95879686e-01
-1.13606844e-02 -1.55827716e-01 3.24234456e-01 -6.18025303e-01
-7.98983216e-01 -9.19822812e-01 -6.95165992e-01 7.81978071e-01
7.86803737e-02 3.95301692e-02 -4.97126803e-02 3.31592858e-01]
|
[11.28786563873291, 9.94622802734375]
|
e2782b31-9402-4382-b6dc-624e989eef93
|
a-study-of-transfer-learning-in-music-source
|
2010.12650
| null |
https://arxiv.org/abs/2010.12650v1
|
https://arxiv.org/pdf/2010.12650v1.pdf
|
A Study of Transfer Learning in Music Source Separation
|
Supervised deep learning methods for performing audio source separation can be very effective in domains where there is a large amount of training data. While some music domains have enough data suitable for training a separation system, such as rock and pop genres, many musical domains do not, such as classical music, choral music, and non-Western music traditions. It is well known that transferring learning from related domains can result in a performance boost for deep learning systems, but it is not always clear how best to do pretraining. In this work we investigate the effectiveness of data augmentation during pretraining, the impact on performance as a result of pretraining and downstream datasets having similar content domains, and also explore how much of a model must be retrained on the final target task, once pretrained.
|
['Prem Seetharaman', 'Bryan Pardo', 'Andreas Bugler']
|
2020-10-23
| null | null | null | null |
['audio-source-separation', 'music-source-separation']
|
['audio', 'music']
|
[ 1.70825496e-01 -7.33867586e-02 3.37479301e-02 -2.92222857e-01
-6.97645783e-01 -6.59986496e-01 3.23379278e-01 1.39887750e-01
-5.30767322e-01 5.62378049e-01 5.32520294e-01 -1.12163544e-01
-2.21261635e-01 -5.91969728e-01 -5.81919789e-01 -6.61310971e-01
-1.42855734e-01 5.77914178e-01 1.97952911e-01 -3.47284138e-01
-4.30177301e-02 3.93380255e-01 -1.36879683e+00 5.33425748e-01
5.84492445e-01 8.36102366e-01 3.55184008e-03 5.55881739e-01
-3.47668156e-02 7.28120327e-01 -1.04723728e+00 -1.93026513e-01
3.15916479e-01 -8.10818970e-01 -9.75499928e-01 -7.07307458e-02
3.10954392e-01 -6.28954545e-02 -7.92226344e-02 7.42142797e-01
7.83015490e-01 2.66250461e-01 7.49890029e-01 -1.07540631e+00
-2.42303550e-01 1.10519731e+00 -4.27856982e-01 2.94820786e-01
3.41969393e-02 -1.61731727e-02 1.19766355e+00 -5.35579681e-01
2.96394020e-01 1.00645936e+00 7.13886559e-01 3.61992419e-01
-1.49209416e+00 -9.83978271e-01 -8.75488594e-02 -5.24249822e-02
-1.00694788e+00 -6.84694231e-01 9.18145120e-01 -3.99882704e-01
5.92062771e-01 8.65715221e-02 5.51741600e-01 9.94390547e-01
-2.46525884e-01 8.04140270e-01 9.52416599e-01 -3.57314348e-01
2.11153612e-01 1.12167187e-01 9.91677959e-03 -2.10002691e-01
-5.60275391e-02 8.18302408e-02 -7.39797533e-01 2.02026498e-02
6.52853370e-01 -3.42558563e-01 -2.89685369e-01 -2.65208244e-01
-1.00551808e+00 6.87533319e-01 4.57425058e-01 6.27788544e-01
-2.70765930e-01 -3.74956727e-02 6.43048525e-01 7.32962728e-01
4.40305203e-01 1.04874671e+00 -5.62544644e-01 -3.07186931e-01
-1.43274641e+00 3.38978559e-01 7.47726500e-01 5.21891832e-01
5.40801942e-01 3.33846956e-01 1.14312537e-01 1.18635297e+00
-2.02772975e-01 5.01560681e-02 7.37337530e-01 -8.35500479e-01
2.21295074e-01 6.09830856e-01 -3.03213298e-01 -5.34676731e-01
-4.60034937e-01 -6.10890329e-01 -6.58063710e-01 3.74556512e-01
8.04821193e-01 -4.28941309e-01 -8.94239306e-01 1.78885365e+00
-5.82025088e-02 2.18270361e-01 2.57651478e-01 1.14988935e+00
8.71858716e-01 4.46695775e-01 -6.21340945e-02 1.17991008e-01
9.72190619e-01 -9.39815700e-01 -2.88884014e-01 -6.74331605e-01
3.60625803e-01 -1.10226071e+00 1.14949358e+00 7.58480787e-01
-1.15717340e+00 -7.21704721e-01 -9.67188776e-01 -8.81209895e-02
-4.32889629e-03 -2.33513936e-02 4.18365330e-01 3.53260964e-01
-5.23381233e-01 9.58214164e-01 -6.32692873e-01 -2.40506873e-01
4.94393587e-01 5.67184448e-01 -4.17623758e-01 1.04832657e-01
-1.07644570e+00 6.90342963e-01 6.75700009e-01 -3.60602409e-01
-1.10318840e+00 -7.08225489e-01 -2.90341645e-01 4.17715192e-01
3.68107706e-01 -4.65909809e-01 1.35418534e+00 -1.56800032e+00
-1.43295121e+00 8.10599446e-01 4.03833389e-01 -6.05219066e-01
2.70891011e-01 -3.86027932e-01 -2.90982455e-01 -1.53964639e-01
-1.45366058e-01 7.01815307e-01 9.47186351e-01 -9.30469930e-01
-5.68355381e-01 -3.46154541e-01 2.10637122e-01 2.65219301e-01
-5.73736966e-01 2.05773547e-01 -2.84499973e-01 -9.10499454e-01
-3.50496545e-02 -1.11313188e+00 -6.32758290e-02 -5.17181575e-01
-3.30974698e-01 -7.98086822e-02 5.87564588e-01 -7.52998233e-01
1.04518998e+00 -2.48977828e+00 3.68773282e-01 2.68771082e-01
-6.17381521e-02 3.68809700e-01 -3.45227182e-01 2.63324499e-01
-3.10273379e-01 -6.16750643e-02 -4.88766804e-02 -7.22763985e-02
-1.57889917e-01 1.32165894e-01 -3.35734278e-01 1.36330217e-01
2.55433470e-01 4.58805382e-01 -7.17149913e-01 4.49583530e-02
-1.97965756e-01 4.60085601e-01 -7.64409840e-01 1.85384184e-01
-2.86500186e-01 6.48710668e-01 -2.61658784e-02 2.06668109e-01
3.05926263e-01 -7.94742182e-02 1.93149656e-01 1.14652142e-01
2.33468682e-01 1.02176118e+00 -1.28550804e+00 1.73207986e+00
-4.14786130e-01 9.46441829e-01 2.76825011e-01 -1.28345883e+00
8.96361470e-01 5.59953392e-01 6.57930791e-01 -5.24856389e-01
2.53033072e-01 2.07094803e-01 8.73943508e-01 -7.51532316e-02
5.18319905e-01 -6.81059897e-01 -9.83167514e-02 6.54930115e-01
1.67655513e-01 -1.35877296e-01 2.05695555e-01 -2.05325484e-01
1.08987784e+00 5.64032458e-02 -1.91910177e-01 -2.20655903e-01
1.22073166e-01 3.45286697e-01 5.75218320e-01 3.17251891e-01
1.15666084e-01 7.80362546e-01 6.38182521e-01 -2.16009900e-01
-1.00242400e+00 -8.21241856e-01 -7.43471161e-02 1.62618232e+00
-2.37233341e-01 -5.21209300e-01 -7.03529894e-01 -4.49332893e-01
-6.43603355e-02 5.51357865e-01 -2.14039594e-01 -5.29035151e-01
-4.84785736e-01 -6.57291830e-01 8.71572793e-01 5.71050048e-01
3.36330831e-01 -1.28334880e+00 -5.41151643e-01 2.26588309e-01
-1.27838124e-02 -9.94133592e-01 -2.27364168e-01 7.62609601e-01
-1.09788465e+00 -8.79159629e-01 -8.23858678e-01 -7.55970597e-01
2.08835289e-01 1.94643915e-01 1.41250288e+00 -1.71573937e-01
1.19357459e-01 -7.64042288e-02 -3.61212879e-01 -7.94802368e-01
-7.28635490e-01 4.80749369e-01 1.64712787e-01 -1.19155765e-01
4.43636179e-01 -7.95012236e-01 -3.97361428e-01 3.16644818e-01
-6.77489400e-01 -2.62436513e-02 5.09990633e-01 7.00210392e-01
3.03666800e-01 3.87898326e-01 7.23888397e-01 -8.67341340e-01
7.17038512e-01 -3.74749631e-01 -1.56789385e-02 -2.21196353e-01
-2.85556108e-01 2.03227643e-02 7.83447206e-01 -7.65758276e-01
-6.81021988e-01 4.89926850e-03 -2.08071768e-01 -5.63854873e-01
-2.93002844e-01 7.12130487e-01 -1.21196486e-01 4.97075021e-01
1.14488590e+00 -2.84528732e-01 1.81487820e-03 -8.42111170e-01
6.63027465e-02 7.22130358e-01 4.38693821e-01 -6.78976655e-01
6.73923731e-01 1.41606912e-01 -4.35164064e-01 -9.36524510e-01
-8.58024538e-01 -5.34637094e-01 -5.37108958e-01 1.20219924e-01
8.67078662e-01 -9.25814629e-01 -2.14150384e-01 2.02998638e-01
-7.53667057e-01 -6.12401664e-01 -6.04758203e-01 6.69225454e-01
-3.87211382e-01 -1.86752960e-01 -3.01367372e-01 -4.94975060e-01
4.46539819e-02 -9.62731302e-01 6.84153318e-01 1.44285813e-01
-7.19913006e-01 -8.96867692e-01 2.08213463e-01 3.15462857e-01
2.60540307e-01 -3.00690308e-02 1.06996310e+00 -1.08601558e+00
-5.13250493e-02 -2.79812515e-02 2.54950792e-01 7.29434431e-01
2.09061101e-01 -3.18310261e-01 -1.31411588e+00 -3.37009490e-01
-1.33677125e-02 -6.24376953e-01 1.09533060e+00 2.69056320e-01
7.89312005e-01 -2.52920091e-02 5.11682853e-02 4.22891766e-01
9.16399896e-01 1.16795376e-01 5.88056147e-01 4.17637527e-01
4.84487951e-01 7.35897839e-01 3.40415657e-01 -1.39602413e-02
-2.39321157e-01 6.92427397e-01 1.20940492e-01 -2.75787592e-01
-2.97180712e-01 -2.36767024e-01 5.59763789e-01 7.81193316e-01
-1.98941469e-01 2.36684252e-02 -1.11733377e+00 5.78041971e-01
-1.45236135e+00 -9.27278340e-01 1.86842650e-01 2.35859632e+00
1.12344158e+00 3.33317459e-01 7.27693617e-01 9.54437435e-01
4.82692510e-01 -8.04960877e-02 -4.00702029e-01 -3.68500561e-01
-7.35539943e-02 5.93226910e-01 1.09190561e-01 2.58941054e-02
-1.17576706e+00 7.76540101e-01 6.19541645e+00 6.32327557e-01
-1.51943123e+00 1.48383766e-01 3.99876654e-01 -4.05192167e-01
-6.84228120e-03 1.61169261e-01 -4.78941202e-01 3.15526456e-01
1.08978009e+00 -6.49892772e-03 3.71685505e-01 7.73319364e-01
3.67245786e-02 1.17604114e-01 -1.40064108e+00 9.35795426e-01
-2.63141662e-01 -9.28949237e-01 -2.69941032e-01 7.59670958e-02
6.25825882e-01 2.28811711e-01 -5.20886108e-02 6.56290412e-01
3.26770723e-01 -1.18567145e+00 6.98924005e-01 -1.67004257e-01
3.61728817e-01 -8.93887699e-01 5.57429671e-01 5.10619283e-01
-6.03517234e-01 -1.52309686e-01 -3.78141671e-01 -2.49298915e-01
-9.07533988e-02 4.90350217e-01 -8.41412544e-01 2.37430304e-01
7.68402457e-01 6.02378011e-01 -6.92191660e-01 1.30901766e+00
-2.05234885e-01 1.13392544e+00 -3.47419500e-01 3.01170707e-01
8.24472234e-02 -2.18638897e-01 5.54129362e-01 1.08609259e+00
2.75812238e-01 -1.15314938e-01 1.54519439e-01 5.08905470e-01
-1.15272209e-01 1.24587603e-01 -4.23523635e-01 -4.79837000e-01
2.11640075e-01 9.52652633e-01 -8.58545184e-01 -2.19406039e-01
-4.16923285e-01 7.88130224e-01 1.16975307e-01 2.92908728e-01
-6.45944715e-01 -1.92788929e-01 8.05564046e-01 6.26299560e-01
2.21580237e-01 -9.74330455e-02 -3.98739696e-01 -9.66236413e-01
-2.43423313e-01 -1.44159746e+00 4.61469650e-01 -6.22406840e-01
-1.21005249e+00 4.64170396e-01 -1.43192858e-01 -1.40570140e+00
-3.85334462e-01 -5.24563253e-01 -7.25240350e-01 8.22998405e-01
-1.14190507e+00 -8.59513164e-01 3.67145687e-02 5.90970576e-01
3.65622669e-01 -5.65478563e-01 8.70291293e-01 5.68381667e-01
-3.08448732e-01 4.88511086e-01 1.76505074e-01 5.40957689e-01
1.12149751e+00 -1.23891902e+00 8.32094029e-02 6.03113770e-01
9.53249395e-01 3.55003059e-01 8.36585641e-01 -2.16121674e-01
-9.67801750e-01 -7.76482046e-01 3.95572484e-01 -3.01112711e-01
5.36068678e-01 -2.82371432e-01 -1.06618321e+00 5.71489036e-01
2.25156531e-01 -2.37881884e-01 8.97477746e-01 7.77945638e-01
-3.50430101e-01 -2.88875401e-01 -5.91380417e-01 4.71247256e-01
7.77768493e-01 -6.77389741e-01 -8.01831841e-01 1.23734493e-02
3.77195567e-01 -4.19290304e-01 -7.35803783e-01 3.72875810e-01
6.07630968e-01 -1.08698964e+00 9.74829495e-01 -7.84145236e-01
7.04933107e-01 -1.04573064e-01 4.84251752e-02 -1.62690687e+00
-2.79536426e-01 -4.52748150e-01 1.23896278e-01 1.44796431e+00
5.72293341e-01 -1.23941697e-01 1.01610434e+00 3.02568585e-01
-2.01297119e-01 -3.24478775e-01 -6.47415996e-01 -9.48485494e-01
4.67762738e-01 -5.29105008e-01 5.16557872e-01 1.11518490e+00
-8.84840339e-02 1.12416625e+00 -2.38474429e-01 -2.07859829e-01
-4.82206196e-02 3.51006478e-01 1.09386170e+00 -1.61024010e+00
-7.95757771e-01 -7.61129797e-01 -4.66297954e-01 -6.28697693e-01
2.03529119e-01 -1.18287230e+00 -8.77619833e-02 -1.45177972e+00
-9.63296145e-02 -6.11900449e-01 -5.62317252e-01 7.19880521e-01
-3.43348011e-02 4.16658998e-01 4.85624582e-01 2.96606362e-01
-2.70577282e-01 1.90487221e-01 1.08192098e+00 -2.76742071e-01
-7.15873837e-01 2.76096433e-01 -9.27418053e-01 8.61384094e-01
9.18910563e-01 -6.44528687e-01 -6.96393967e-01 -6.23399436e-01
2.48320818e-01 -9.79712829e-02 1.61869869e-01 -1.24172568e+00
-1.30902901e-01 9.25769471e-03 4.43411201e-01 -6.42692447e-02
3.76512438e-01 -8.47639084e-01 1.58148572e-01 2.64803797e-01
-5.14177263e-01 -3.18590760e-01 4.95388836e-01 2.02442214e-01
-7.41352797e-01 -3.26598823e-01 1.00674725e+00 -2.00958969e-03
-5.14893591e-01 -1.62605166e-01 -1.99750409e-01 5.58650851e-01
5.92318535e-01 -1.83071136e-01 1.60363838e-01 -5.52293122e-01
-1.00066221e+00 -1.20838672e-01 4.06785846e-01 5.90689719e-01
1.59717470e-01 -1.42482901e+00 -8.80131185e-01 1.90274552e-01
-1.74796119e-01 6.20646700e-02 -4.52496521e-02 8.27607632e-01
-1.60898700e-01 -2.71679834e-02 -4.17638391e-01 -5.16860664e-01
-1.43146956e+00 4.39532816e-01 1.75534919e-01 -1.73595890e-01
-6.71108425e-01 8.86534691e-01 3.69272798e-01 -3.69247526e-01
4.95712101e-01 -3.12823087e-01 -2.81628013e-01 3.92648220e-01
3.91037345e-01 2.07531869e-01 1.89050704e-01 -4.67796087e-01
-2.59876072e-01 2.73779184e-01 7.10439384e-02 -2.91548401e-01
1.63486660e+00 4.19429749e-01 6.66710213e-02 5.96738338e-01
1.01922476e+00 1.89089999e-01 -1.19052148e+00 -1.29861370e-01
1.11015163e-01 -1.77638993e-01 1.70209348e-01 -7.50234663e-01
-1.18290591e+00 1.27384019e+00 5.31036556e-01 3.07942122e-01
1.46737838e+00 -3.33879925e-02 7.41077781e-01 3.05037409e-01
1.26414403e-01 -1.21256614e+00 1.88999444e-01 6.37119234e-01
9.16877687e-01 -1.08492017e+00 1.33266933e-02 -7.43072154e-03
-8.05547535e-01 8.44016016e-01 4.38544214e-01 -5.07859111e-01
7.24771261e-01 5.35242617e-01 1.41711056e-01 1.38632068e-02
-6.58081174e-01 -4.08386946e-01 4.84717071e-01 4.14859086e-01
8.83598268e-01 -1.36489302e-01 9.27966461e-02 1.02339053e+00
-7.37471938e-01 8.58336389e-02 3.80031884e-01 6.78891301e-01
-2.60359168e-01 -1.25829613e+00 -4.90355164e-01 5.55990517e-01
-8.58837008e-01 -4.28958200e-02 -8.84659290e-01 7.24824667e-01
4.74393696e-01 8.50884199e-01 2.66072780e-01 -4.50117975e-01
3.61725181e-01 4.38483715e-01 4.75371450e-01 -1.07594967e+00
-1.06877959e+00 5.43941379e-01 3.31121236e-01 1.07247857e-02
-4.94644344e-01 -4.52106267e-01 -1.31535637e+00 -1.87650725e-01
-3.60906959e-01 4.80266541e-01 2.28376791e-01 8.60530615e-01
5.79445250e-02 7.79185832e-01 2.56293476e-01 -8.93032074e-01
-2.79004931e-01 -1.19139123e+00 -7.61514604e-01 5.39379001e-01
3.49065334e-01 -5.14339328e-01 -2.65062839e-01 2.94531375e-01]
|
[15.775272369384766, 5.294350624084473]
|
5c1d7079-25f1-48f4-b97c-340154e0b8e1
|
improving-gans-for-long-tailed-data-through
|
2208.09932
| null |
https://arxiv.org/abs/2208.09932v1
|
https://arxiv.org/pdf/2208.09932v1.pdf
|
Improving GANs for Long-Tailed Data through Group Spectral Regularization
|
Deep long-tailed learning aims to train useful deep networks on practical, real-world imbalanced distributions, wherein most labels of the tail classes are associated with a few samples. There has been a large body of work to train discriminative models for visual recognition on long-tailed distribution. In contrast, we aim to train conditional Generative Adversarial Networks, a class of image generation models on long-tailed distributions. We find that similar to recognition, state-of-the-art methods for image generation also suffer from performance degradation on tail classes. The performance degradation is mainly due to class-specific mode collapse for tail classes, which we observe to be correlated with the spectral explosion of the conditioning parameter matrix. We propose a novel group Spectral Regularizer (gSR) that prevents the spectral explosion alleviating mode collapse, which results in diverse and plausible image generation even for tail classes. We find that gSR effectively combines with existing augmentation and regularization techniques, leading to state-of-the-art image generation performance on long-tailed data. Extensive experiments demonstrate the efficacy of our regularizer on long-tailed datasets with different degrees of imbalance.
|
['R. Venkatesh Babu', 'Varun Jampani', 'Tejan Karmali', 'Naman Jaswani', 'Harsh Rangwani']
|
2022-08-21
| null | null | null | null |
['conditional-image-generation']
|
['computer-vision']
|
[ 2.50587672e-01 -2.65340388e-01 -1.77921951e-01 -3.96355659e-01
-9.08985019e-01 -3.86570275e-01 6.94381952e-01 -3.12886387e-01
-4.13037129e-02 7.39481926e-01 1.44465908e-01 -1.74744368e-01
2.45497227e-01 -6.56370163e-01 -8.67085636e-01 -9.84296858e-01
2.40272939e-01 6.68561220e-01 -7.49684200e-02 -5.09223789e-02
3.97580303e-03 4.24310833e-01 -1.45495355e+00 4.82641727e-01
9.07436430e-01 1.13337958e+00 -3.69434386e-01 7.00271785e-01
-9.78477001e-02 9.45246220e-01 -1.08462930e+00 -7.03641951e-01
2.13795125e-01 -7.06901252e-01 -2.84242362e-01 3.47128123e-01
8.63546610e-01 -2.79597074e-01 -4.23529387e-01 1.09353995e+00
9.03212249e-01 -9.81288999e-02 1.25140464e+00 -1.71975970e+00
-7.50961542e-01 2.77866453e-01 -1.04418504e+00 1.81860775e-01
-1.18505821e-01 3.00636232e-01 8.29598188e-01 -1.06323791e+00
3.96226794e-01 1.33057427e+00 7.03081310e-01 5.34749210e-01
-1.29568529e+00 -1.07383692e+00 8.40721652e-03 3.22174609e-01
-1.30896115e+00 -2.59988755e-01 8.71383786e-01 -6.45363986e-01
6.02188647e-01 6.84365630e-02 4.35981274e-01 1.36279881e+00
1.42695352e-01 1.03014040e+00 1.02748108e+00 -2.26002231e-01
2.19763085e-01 -1.51029378e-01 -1.24300510e-01 2.72262841e-01
5.29623330e-01 4.82129864e-03 -6.74695313e-01 -8.17661509e-02
5.77565372e-01 -1.51585773e-01 -2.77299225e-01 -5.34227669e-01
-8.13926637e-01 9.29808140e-01 2.72035390e-01 -2.77051657e-01
-2.37734899e-01 1.62754223e-01 4.54716355e-01 2.58150995e-01
8.41289461e-01 1.10041067e-01 -2.33855203e-01 -1.38270214e-01
-1.23112607e+00 3.48020703e-01 7.40016520e-01 8.97376537e-01
5.09768486e-01 4.45204765e-01 -6.90793753e-01 9.31186914e-01
-6.35735691e-02 7.86326468e-01 4.63782549e-01 -5.67989230e-01
5.75885892e-01 4.04192865e-01 6.45362064e-02 -7.13358164e-01
-2.44779959e-01 -1.07178962e+00 -1.34214675e+00 4.71995533e-01
6.21222138e-01 -1.74218133e-01 -1.36821175e+00 1.92875278e+00
8.96102190e-02 1.01540342e-01 -1.81793839e-01 7.61590600e-01
6.69359922e-01 7.31313705e-01 -4.82731536e-02 -1.45317122e-01
1.05590582e+00 -1.04893458e+00 -6.71405852e-01 -3.70971829e-01
1.88939676e-01 -7.64901102e-01 1.22808588e+00 4.60431665e-01
-9.80891585e-01 -4.57172424e-01 -8.99677217e-01 2.06323385e-01
-1.16161667e-02 1.31935021e-02 5.32767117e-01 7.05458164e-01
-6.62208676e-01 3.26865375e-01 -6.57231212e-01 2.32027888e-01
9.07567561e-01 -7.52146989e-02 -1.19907431e-01 -2.11278811e-01
-8.62414002e-01 6.71431720e-01 8.19584802e-02 -4.50007841e-02
-9.68218327e-01 -1.02225626e+00 -7.69311130e-01 1.33202791e-01
-1.86325854e-03 -8.35652649e-01 1.00514519e+00 -1.42413414e+00
-1.17886686e+00 9.78151917e-01 -3.59297395e-02 -3.79104763e-01
8.84950578e-01 -2.00295165e-01 -2.34187678e-01 3.61517891e-02
2.20980018e-01 4.64789450e-01 1.21541154e+00 -1.37102807e+00
-7.64554515e-02 -3.23255748e-01 -7.13696480e-01 -5.72914956e-03
-5.87616742e-01 -1.26273781e-01 -3.77430350e-01 -9.94208336e-01
-3.05581391e-01 -8.34321320e-01 -3.09905224e-03 2.20416263e-02
-5.71810067e-01 -1.15960546e-01 7.77387977e-01 -5.30697405e-01
1.17589629e+00 -2.10447812e+00 -1.55060729e-02 1.64516658e-01
1.99637383e-01 3.95659626e-01 -3.40360403e-01 4.12094921e-01
-4.53630358e-01 -2.42100861e-02 -2.61772633e-01 -6.41585767e-01
2.97226548e-01 1.45985663e-01 -7.94565678e-01 6.30263865e-01
2.29874507e-01 1.05157053e+00 -5.30603290e-01 -2.62934357e-01
-6.35013580e-02 4.16369438e-01 -6.91552699e-01 5.58365345e-01
-3.96272838e-01 4.40563947e-01 2.65920293e-02 6.33185506e-01
1.12394416e+00 -5.99456489e-01 -2.76745439e-01 -2.38010347e-01
4.71441209e-01 -1.35244265e-01 -8.74542236e-01 1.21595931e+00
-1.18352108e-01 6.56467974e-01 -1.40399382e-01 -1.02072167e+00
7.71119595e-01 -3.58292498e-02 -1.59837790e-02 -8.72010946e-01
-6.89064385e-03 3.30879211e-01 -3.73730734e-02 -3.46624464e-01
2.73458660e-01 -6.18465602e-01 9.52539742e-02 3.91833514e-01
1.94436565e-01 -2.50815213e-01 2.38646314e-01 2.25266784e-01
9.10092890e-01 2.13537086e-02 -7.93229267e-02 3.19421589e-02
2.40708254e-02 -5.95535398e-01 5.24213552e-01 9.20309365e-01
-1.57048792e-01 1.13854933e+00 7.94848084e-01 -1.97402954e-01
-1.27306688e+00 -1.20213854e+00 -2.42350593e-01 1.22427785e+00
-2.89550889e-02 -1.21538490e-01 -7.86934614e-01 -5.55484891e-01
-4.60869782e-02 7.05883324e-01 -6.30876124e-01 -4.07865375e-01
-2.77316958e-01 -1.28422976e+00 6.68421924e-01 6.87543571e-01
4.95565623e-01 -1.01491284e+00 -1.54744223e-01 -6.05600625e-02
-1.36806458e-01 -1.11311615e+00 -6.68171167e-01 2.34222591e-01
-5.82488120e-01 -9.06490088e-01 -1.32573867e+00 -6.07494414e-01
6.72286272e-01 -6.43585995e-02 1.60274732e+00 -1.65299147e-01
-4.07004207e-01 1.83130667e-01 -7.67394975e-02 -5.02360642e-01
-3.78408760e-01 -1.69985488e-01 -2.82345772e-01 2.77494639e-01
3.95044014e-02 -7.17288017e-01 -9.45571005e-01 2.85370916e-01
-1.10988259e+00 -1.42223788e-02 4.69951153e-01 1.22377801e+00
5.72860599e-01 3.53056639e-02 8.23802054e-01 -8.34484518e-01
5.55727184e-01 -5.93852103e-01 -4.17353570e-01 -1.29565299e-02
-4.31067169e-01 2.05465518e-02 8.45877647e-01 -6.52103484e-01
-1.01299882e+00 -3.81221414e-01 -3.01327318e-01 -6.38845563e-01
-4.54545915e-02 3.11831027e-01 -3.03648233e-01 1.79216653e-01
8.38886201e-01 3.61487299e-01 -8.71684924e-02 -3.25456411e-01
2.75841326e-01 5.53399086e-01 7.76711762e-01 -6.56898558e-01
7.84344256e-01 5.44511020e-01 1.68927833e-01 -7.35809863e-01
-1.41365409e+00 -2.21848398e-01 2.29125228e-02 -1.80257764e-02
5.88300109e-01 -1.26624763e+00 -3.02418560e-01 1.26382113e+00
-8.88318002e-01 -6.86467469e-01 -4.94658709e-01 9.87635404e-02
-7.06342459e-01 3.36113691e-01 -5.43070495e-01 -8.27553213e-01
-5.01135409e-01 -7.45891750e-01 1.22927904e+00 2.03090087e-01
7.28040338e-02 -9.65105057e-01 1.16601944e-01 6.08935654e-01
5.18889368e-01 4.09761339e-01 1.13803947e+00 -6.01511776e-01
-3.08822960e-01 -2.41343737e-01 -4.89383072e-01 7.87320018e-01
-1.96286604e-01 -2.33509280e-02 -1.03381062e+00 -6.43468797e-01
-1.68241680e-01 -7.46245384e-01 1.34045911e+00 3.79010677e-01
1.62631357e+00 -2.47482762e-01 -3.91297340e-02 9.74667549e-01
1.11030281e+00 -1.51252940e-01 9.30896282e-01 -5.61345136e-03
8.13333511e-01 1.35013998e-01 2.56213099e-01 6.76168859e-01
3.95319521e-01 5.82512081e-01 6.13988042e-01 -5.39137542e-01
-5.67108750e-01 -3.33208323e-01 3.24962646e-01 5.03444850e-01
3.69710952e-01 -6.78177416e-01 -8.89887154e-01 6.89943373e-01
-1.71877718e+00 -8.98654699e-01 -1.01330861e-01 2.27168322e+00
1.20623171e+00 2.28205293e-01 2.66206086e-01 1.80427089e-01
7.08341837e-01 1.45977542e-01 -7.96751976e-01 -8.69301260e-02
-4.64596719e-01 4.17508304e-01 3.02361876e-01 1.29729941e-01
-1.08925533e+00 6.47566378e-01 6.02649927e+00 1.46626365e+00
-1.30381477e+00 -1.44197479e-01 1.13169694e+00 -1.96586505e-01
-5.42388618e-01 -4.91430372e-01 -7.13307679e-01 8.05633545e-01
6.52235448e-01 -9.75345746e-02 1.65470704e-01 6.73264325e-01
-1.75526053e-01 9.53004211e-02 -9.81258631e-01 1.27549231e+00
4.12345499e-01 -1.17307258e+00 2.66786218e-01 -2.40552332e-02
1.22894156e+00 -1.02806419e-01 4.96303886e-01 3.30478758e-01
3.01868200e-01 -1.50207388e+00 8.36444199e-01 4.02872831e-01
1.11377776e+00 -1.02768588e+00 6.92900658e-01 4.23088133e-01
-5.73773921e-01 -3.37242968e-02 -4.00029510e-01 6.12193160e-02
-3.56393382e-02 1.21307755e+00 -5.59083223e-01 1.15587153e-01
4.93967772e-01 5.81061661e-01 -7.45645165e-01 1.19732451e+00
-3.03175598e-01 1.04727912e+00 -1.41214669e-01 3.59763801e-01
1.14721423e-02 -1.19251043e-01 4.30904299e-01 1.27995634e+00
3.72609377e-01 -4.28144723e-01 8.57601222e-03 8.74991834e-01
-3.51830244e-01 -1.17304042e-01 -4.58243936e-01 -1.66462988e-01
7.97059014e-02 1.02937984e+00 -6.14620090e-01 -4.34250712e-01
-2.49768287e-01 1.09996641e+00 4.56775367e-01 6.50758982e-01
-9.03576970e-01 -2.94752210e-01 6.19230151e-01 3.25671166e-01
6.79616451e-01 3.27069350e-02 -5.20718396e-01 -1.32561600e+00
1.70526356e-01 -1.18409014e+00 3.82669836e-01 -7.54438460e-01
-1.87829018e+00 4.30229396e-01 -2.89638191e-01 -1.16487384e+00
-2.50420988e-01 -5.07058442e-01 -6.56745672e-01 7.38090694e-01
-1.63465166e+00 -1.32718062e+00 -4.67155576e-01 7.51986504e-01
2.21675456e-01 -2.13333428e-01 7.12523580e-01 5.19879282e-01
-6.71543658e-01 1.06786811e+00 3.34220558e-01 1.25742510e-01
9.62065935e-01 -1.37945259e+00 3.96716923e-01 8.57098222e-01
5.29016256e-02 2.51747966e-01 8.42834711e-01 -4.06138659e-01
-1.06786537e+00 -1.40874493e+00 4.92973715e-01 -2.20795199e-01
5.59862077e-01 -5.12446582e-01 -9.18518484e-01 5.74601769e-01
1.48534268e-01 4.17232543e-01 8.81122291e-01 -2.50405014e-01
-8.89602900e-01 -1.98817208e-01 -9.70759213e-01 4.68399376e-01
8.78293931e-01 -4.28514987e-01 -2.68935591e-01 3.66485387e-01
2.80904621e-01 -4.34210598e-01 -3.35257024e-01 5.69736063e-01
5.06825984e-01 -1.16463518e+00 9.69290376e-01 -7.06238687e-01
8.35771978e-01 -2.60410815e-01 -3.83146331e-02 -1.71648562e+00
-1.78397372e-01 -6.43610239e-01 -3.74186158e-01 1.24443305e+00
1.41436264e-01 -5.69946468e-01 8.99063110e-01 1.19517222e-01
-6.10187128e-02 -7.35346437e-01 -9.31742370e-01 -9.08098280e-01
5.92547417e-01 -3.71689230e-01 3.47997695e-01 6.48212314e-01
-6.82551086e-01 3.77544999e-01 -7.31439173e-01 -1.94053128e-01
9.69780684e-01 6.00205839e-01 9.88687873e-01 -1.01821375e+00
-4.85837042e-01 -3.64698946e-01 -3.54127139e-01 -1.27934706e+00
3.35281909e-01 -8.12696874e-01 5.19759767e-02 -1.31530738e+00
6.22381330e-01 -3.43370378e-01 -2.81981230e-01 4.51713115e-01
-5.71040750e-01 9.50921476e-01 2.07506493e-01 -8.27947166e-03
-6.57217681e-01 9.30539250e-01 1.67800510e+00 -2.19235227e-01
2.59571880e-01 9.58458111e-02 -8.95429552e-01 7.29474306e-01
5.63230157e-01 -4.59353209e-01 -5.66909134e-01 -1.87196225e-01
6.52895153e-01 -1.09400027e-01 6.16457701e-01 -1.00844884e+00
-4.25249301e-02 -1.62686050e-01 8.03582191e-01 -6.45879686e-01
2.79411763e-01 -4.61917460e-01 -1.91274807e-02 1.49028778e-01
-1.96878552e-01 -3.85548592e-01 2.98121244e-01 5.06861985e-01
-4.12400454e-01 1.06368661e-01 1.18947279e+00 1.69069156e-01
-1.44545108e-01 4.86570388e-01 -2.71913379e-01 8.71134937e-01
7.51033843e-01 -7.72066712e-02 -5.98907471e-01 -9.64665473e-01
-4.14968312e-01 1.09415486e-01 4.49461281e-01 2.24215671e-01
4.02565956e-01 -1.52056766e+00 -1.02667069e+00 5.59111118e-01
1.46631554e-01 2.60674179e-01 3.67567003e-01 7.03524590e-01
-2.14801997e-01 -2.57273495e-01 -2.41633803e-01 -6.42378986e-01
-9.91123080e-01 4.43874151e-01 3.85982424e-01 -4.52350318e-01
-3.93739313e-01 1.18332493e+00 8.28988314e-01 -2.08930656e-01
2.58567661e-01 6.80941939e-02 1.54563740e-01 3.55747014e-01
4.70576346e-01 3.32183838e-01 1.57700434e-01 -4.17638630e-01
-2.51705438e-01 4.99527961e-01 -3.01280599e-02 1.91720456e-01
1.34886944e+00 2.24196136e-01 2.50333436e-02 2.32120663e-01
1.17787600e+00 1.30402490e-01 -1.70949471e+00 -3.37935053e-02
-5.89258313e-01 -3.75016570e-01 -2.23832250e-01 -1.06610930e+00
-1.37639678e+00 1.18027389e+00 5.91789305e-01 1.24806575e-01
1.34050012e+00 -1.69536635e-01 9.38368440e-01 -9.22744647e-02
6.75558951e-03 -9.15697277e-01 6.63114488e-01 5.87718070e-01
1.16451597e+00 -1.10903370e+00 -6.51071519e-02 -2.33136073e-01
-7.36807823e-01 8.91980171e-01 6.39586091e-01 -1.80266857e-01
4.70653564e-01 5.81928372e-01 1.54149547e-01 -1.17394375e-02
-7.30879843e-01 -8.23157281e-03 3.14611346e-01 7.89920390e-01
3.09834003e-01 1.13672972e-01 7.99857527e-02 7.64390290e-01
-4.29096848e-01 -4.56688218e-02 5.50210595e-01 4.94153798e-01
-2.10553110e-01 -9.93056953e-01 -3.92268062e-01 6.41037762e-01
-5.98831058e-01 -3.89140457e-01 -1.26219884e-01 3.75897199e-01
1.31298557e-01 7.21683979e-01 2.47722462e-01 3.12881134e-02
1.38883948e-01 1.76439166e-01 5.21734595e-01 -4.60739344e-01
-2.91776597e-01 1.81807011e-01 -2.65435316e-02 -3.86736482e-01
-2.74544396e-03 -4.06095117e-01 -8.92779469e-01 -1.41440779e-01
-1.80202127e-01 -1.42958522e-01 3.72784942e-01 6.07706547e-01
4.51283902e-01 5.88063300e-01 6.95226192e-01 -9.29553628e-01
-9.19207394e-01 -1.01615894e+00 -8.21636975e-01 1.01955807e+00
6.61041498e-01 -5.93325734e-01 -7.12112427e-01 1.03286341e-01]
|
[9.595014572143555, 2.9111335277557373]
|
c615e08b-f6c6-433a-89a5-a22c97b2f735
|
bag-of-tricks-for-natural-policy-gradient
|
2201.09104
| null |
https://arxiv.org/abs/2201.09104v2
|
https://arxiv.org/pdf/2201.09104v2.pdf
|
Understanding the Effects of Second-Order Approximations in Natural Policy Gradient Reinforcement Learning
|
Natural policy gradient methods are popular reinforcement learning methods that improve the stability of policy gradient methods by utilizing second-order approximations to precondition the gradient with the inverse of the Fisher-information matrix. However, to the best of the authors' knowledge, there has not been a study that has investigated the effects of different second-order approximations in a comprehensive and systematic manner. To address this, five different second-order approximations were studied and compared across multiple key metrics including performance, stability, sample efficiency, and computation time. Furthermore, hyperparameters which aren't typically acknowledged in the literature are studied including the effect of different batch sizes and optimizing the critic network with the natural gradient. Experimental results show that on average, improved second-order approximations achieve the best performance and that using properly tuned hyperparameters can lead to large improvements in performance and sample efficiency ranging up to +181%. We also make the code in this study available at https://github.com/gebob19/natural-policy-gradient-reinforcement-learning.
|
['David A. Clausi', 'Alexander Wong', 'Brennan Gebotys']
|
2022-01-22
| null | null | null | null |
['policy-gradient-methods']
|
['methodology']
|
[-6.25455022e-01 -6.34755343e-02 -5.10973394e-01 -1.12085462e-01
-6.43961608e-01 -4.66959536e-01 7.33735383e-01 -3.16772424e-02
-7.75003374e-01 9.82601523e-01 3.68379980e-01 -6.47797942e-01
-1.29157737e-01 -3.39786381e-01 -6.53955579e-01 -5.69449425e-01
-1.07477650e-01 1.57317415e-01 3.53712738e-01 -2.26019278e-01
6.43831074e-01 3.98741126e-01 -1.39358473e+00 -3.17683697e-01
9.78087544e-01 7.72057950e-01 -6.60090521e-02 5.25546253e-01
2.26116970e-01 6.55475676e-01 -4.27116752e-01 -8.62006247e-02
6.09644592e-01 -5.61577022e-01 -5.75048625e-01 -3.47811997e-01
2.69615322e-01 -8.12627733e-01 -3.91689181e-01 1.13588226e+00
7.07323849e-01 5.42659998e-01 3.69730681e-01 -9.53905761e-01
-4.00159001e-01 6.35837555e-01 -5.31914413e-01 5.06675661e-01
3.48910868e-01 4.95805055e-01 8.96906376e-01 -7.72950053e-01
3.45208973e-01 1.43175042e+00 5.95680296e-01 5.81106663e-01
-1.11519384e+00 -7.40032673e-01 1.96798831e-01 1.98284000e-01
-7.96790123e-01 -3.95333141e-01 6.12901151e-01 -2.30289251e-01
8.82216573e-01 -1.13368027e-01 6.97638333e-01 8.93583059e-01
2.60216951e-01 9.06139016e-01 1.59586072e+00 -5.09796739e-01
4.41779882e-01 4.02197063e-01 -4.62181717e-02 9.80905652e-01
2.58912623e-01 7.74753332e-01 -4.24423009e-01 -8.05850476e-02
9.06987548e-01 -3.15849155e-01 -1.37062252e-01 -4.29144710e-01
-8.79610300e-01 1.11543870e+00 3.47331583e-01 2.89419502e-01
-4.68144864e-01 3.64101827e-01 5.44107378e-01 3.77650470e-01
6.19099379e-01 6.39856756e-01 -2.83237845e-01 -6.69066966e-01
-9.10873294e-01 4.97298092e-01 9.04976308e-01 5.81388712e-01
7.07625389e-01 5.33386827e-01 -2.49550402e-01 7.14479744e-01
3.65340263e-01 5.91772497e-01 7.37728834e-01 -1.26675391e+00
3.55614960e-01 2.39558533e-01 4.92503643e-01 -7.32421756e-01
-5.28050482e-01 -4.57936496e-01 -3.99144322e-01 6.96229875e-01
7.82973528e-01 -6.87936306e-01 -5.65630972e-01 1.56228411e+00
5.54769695e-01 4.83252108e-02 -6.28064051e-02 1.28454041e+00
3.26718241e-01 4.66555655e-01 -1.27130270e-01 -1.20002404e-01
9.52586472e-01 -1.24473834e+00 -5.21784663e-01 -2.65212387e-01
4.51385379e-01 -8.53221476e-01 1.37473428e+00 2.60402679e-01
-1.10228705e+00 -2.68150210e-01 -1.19679987e+00 3.48337919e-01
-1.98505044e-01 2.69736469e-01 7.39462078e-01 7.55913913e-01
-9.75549817e-01 1.07059610e+00 -1.04379714e+00 -1.25810176e-01
3.00779771e-02 9.80014428e-02 1.09330058e-01 2.17727646e-01
-1.28078902e+00 1.17853534e+00 2.04168558e-01 -1.04953974e-01
-9.19563055e-01 -8.26269925e-01 -4.89683837e-01 -3.25947590e-02
6.37753487e-01 -4.41196442e-01 1.78725028e+00 -9.38786328e-01
-2.21892524e+00 -7.18111992e-02 -9.94173437e-02 -5.92373908e-01
8.96748066e-01 -3.83870333e-01 7.77021199e-02 9.34201032e-02
-6.13791049e-02 4.03418690e-01 7.52034903e-01 -8.06149602e-01
-7.12601364e-01 -3.03086877e-01 1.68979213e-01 5.37215769e-01
-3.62363577e-01 -2.77945966e-01 -8.06013048e-02 -5.10498405e-01
-3.18459362e-01 -1.24151564e+00 -2.82300413e-01 -1.87786296e-01
1.05024502e-01 -2.58104414e-01 3.92571777e-01 -6.91134155e-01
1.17963862e+00 -2.03940630e+00 -2.30222672e-01 8.64790827e-02
-1.43949315e-01 5.97711027e-01 -9.61461198e-03 5.81154943e-01
3.24608207e-01 7.35438168e-02 1.38796732e-01 -6.53139949e-02
3.59393567e-01 3.49575803e-02 -3.00343990e-01 6.74998641e-01
-1.90835163e-01 7.13144600e-01 -1.17462897e+00 -9.08219889e-02
4.10471886e-01 4.58333075e-01 -7.60211706e-01 8.18891004e-02
-6.66127503e-02 4.46719199e-01 -5.00347197e-01 1.23073652e-01
1.79781139e-01 -1.33532360e-01 3.21807921e-01 1.46292195e-01
-2.14320093e-01 5.01790941e-01 -1.28268600e+00 1.22440147e+00
-4.38586503e-01 5.72140455e-01 7.74929896e-02 -9.86509740e-01
6.20484889e-01 2.62793481e-01 5.33367634e-01 -8.32624376e-01
2.58220322e-02 4.05498207e-01 2.78464556e-02 -2.18239665e-01
5.40393412e-01 -9.31334570e-02 4.35944110e-01 4.19387013e-01
-1.00102261e-01 -1.69674322e-01 6.16972268e-01 -6.78979307e-02
8.50757539e-01 4.23649162e-01 2.02127963e-01 -4.96891528e-01
3.42132717e-01 -9.38689038e-02 5.12462914e-01 9.13700879e-01
-5.03700495e-01 -5.77175468e-02 6.19851470e-01 -2.93491304e-01
-1.05769742e+00 -7.18976557e-01 1.40950009e-01 1.28400159e+00
-2.12191209e-01 -2.06659257e-01 -6.80557132e-01 -5.43517947e-01
5.34507871e-01 8.01559567e-01 -4.58344430e-01 -1.60155490e-01
-3.56973320e-01 -6.45081878e-01 5.39330304e-01 4.14464474e-01
6.84910595e-01 -9.85587060e-01 -9.01634872e-01 3.49661052e-01
1.93826139e-01 -8.28479707e-01 -5.15897334e-01 -4.11768779e-02
-1.21478879e+00 -9.76364136e-01 -7.96356976e-01 -2.08436191e-01
3.89587194e-01 -3.75427902e-02 8.58641624e-01 -6.22891784e-02
1.72545318e-04 4.45011348e-01 -3.88425082e-01 -4.57132638e-01
-3.77662241e-01 1.18305288e-01 3.47708642e-01 -4.84721035e-01
7.13733025e-04 -4.25337464e-01 -7.73120284e-01 1.51715621e-01
-5.45275807e-01 -3.72398674e-01 6.57156885e-01 9.79129374e-01
2.06704631e-01 -3.90884727e-01 6.59018040e-01 -6.67082310e-01
1.16457081e+00 -2.96688974e-01 -1.16806090e+00 -1.27932161e-01
-1.49388635e+00 6.08864427e-01 6.80629253e-01 -5.70212901e-01
-1.12510395e+00 -2.84060150e-01 -4.09318618e-02 -3.51944834e-01
1.38500467e-01 5.09663403e-01 9.74997401e-01 -3.23238194e-01
8.23069096e-01 1.33419499e-01 4.15440887e-01 -3.64609331e-01
4.55536276e-01 3.08885902e-01 -2.04630699e-02 -7.79890954e-01
3.66443276e-01 2.77062833e-01 -1.70101598e-02 -5.21937728e-01
-7.63185322e-01 -4.12177712e-01 1.16960285e-03 -2.27549642e-01
3.15594524e-01 -6.48535907e-01 -7.80773640e-01 4.19152379e-01
-3.34007263e-01 -9.31959987e-01 -3.06276202e-01 1.00657678e+00
-7.37226009e-01 4.60186839e-01 -8.71587336e-01 -9.66695249e-01
-5.45131862e-01 -1.17698801e+00 4.28918958e-01 5.93337774e-01
5.90437539e-02 -1.06887412e+00 3.26744616e-01 -1.38240054e-01
8.26431334e-01 -2.19277143e-01 5.22847354e-01 -4.31025386e-01
-1.37168124e-01 1.79413214e-01 1.98071990e-02 4.91097718e-01
-2.03473680e-02 3.21417190e-02 -5.72383463e-01 -6.71138287e-01
-1.35157272e-01 -3.92413735e-01 8.12328279e-01 6.28565252e-01
4.47765052e-01 -4.37831521e-01 2.77123810e-03 3.81626934e-01
1.34891987e+00 1.10397950e-01 3.26726526e-01 6.03819072e-01
1.28473908e-01 2.32516170e-01 8.65090668e-01 8.19792569e-01
1.55373469e-01 3.32662970e-01 2.93864697e-01 1.53696492e-01
-8.07601027e-03 -2.66332328e-01 6.32282913e-01 6.07943833e-01
-1.37150362e-01 3.33181262e-01 -6.08024895e-01 3.59376729e-01
-1.88482666e+00 -1.06994307e+00 2.88438886e-01 2.29344130e+00
9.25146520e-01 9.17600393e-02 3.60022277e-01 -1.41517267e-01
4.80481386e-01 1.29074946e-01 -9.18594241e-01 -4.49222863e-01
4.79498535e-01 -2.91186646e-02 7.87529707e-01 8.79218817e-01
-8.79957139e-01 1.05633712e+00 6.59348869e+00 7.63625264e-01
-1.43112874e+00 5.44204265e-02 3.49411696e-01 -4.04617548e-01
8.15408304e-02 2.85301834e-01 -8.75422597e-01 4.62776840e-01
9.50582802e-01 -1.70247957e-01 1.04134917e+00 1.05109453e+00
4.57195789e-01 -3.87513250e-01 -4.94965971e-01 6.74800098e-01
-3.81693304e-01 -9.32387292e-01 -5.11362374e-01 1.06478773e-01
8.89306545e-01 3.03684771e-01 1.79059491e-01 8.05291235e-01
6.17689073e-01 -4.52448606e-01 7.72605062e-01 5.17107666e-01
2.87812799e-01 -9.40769255e-01 4.88051206e-01 2.46502891e-01
-6.90000355e-01 -4.47570950e-01 -5.27421117e-01 -2.35942870e-01
-5.45949712e-02 6.30396128e-01 -8.99215996e-01 3.02498907e-01
7.27676809e-01 4.82907921e-01 -3.15437317e-01 1.22163820e+00
-4.93987620e-01 1.01668596e+00 -3.83237451e-01 -6.49469912e-01
7.37882018e-01 -4.65261430e-01 6.01879179e-01 8.74412537e-01
1.80880189e-01 -1.44282773e-01 2.14374423e-01 6.17398262e-01
2.61925757e-01 2.82081097e-01 -3.92941356e-01 -1.82454094e-01
4.43897814e-01 1.20008230e+00 -5.11629403e-01 -4.29513097e-01
-2.90723115e-01 6.26021624e-01 4.68075871e-01 6.05511427e-01
-7.40191758e-01 -3.34137350e-01 7.62840629e-01 -8.37900937e-02
4.00627255e-01 -3.55612904e-01 -3.53324115e-02 -1.02434111e+00
2.40378287e-02 -1.02627540e+00 2.40795732e-01 -3.73358518e-01
-1.15749753e+00 1.98993713e-01 2.05312625e-01 -7.87080884e-01
-7.29192495e-01 -6.08322740e-01 -2.93459296e-01 6.68711543e-01
-1.42197156e+00 -3.24757844e-01 1.65992379e-01 3.56146604e-01
3.23972583e-01 -1.56735688e-01 6.40843213e-01 1.51934475e-01
-5.56058943e-01 6.16912603e-01 8.67585003e-01 -1.02104247e-01
8.46258759e-01 -1.33563066e+00 4.87555228e-02 6.98517442e-01
-2.28179038e-01 7.22719669e-01 9.71398473e-01 -5.51575959e-01
-1.41048408e+00 -4.89948332e-01 1.54931784e-01 1.82206571e-01
7.78771400e-01 1.01551056e-01 -5.57754219e-01 5.56678653e-01
3.27781975e-01 -1.08692870e-01 2.03400522e-01 2.10443497e-01
-8.92628431e-02 -1.93846941e-01 -9.78878736e-01 7.81010151e-01
4.86421078e-01 -4.06141102e-01 -2.43703991e-01 2.64219493e-01
1.22234136e-01 -6.39572501e-01 -1.00823164e+00 2.92512685e-01
7.98844039e-01 -1.08804727e+00 7.38988280e-01 -4.56913024e-01
1.32061929e-01 -5.07564321e-02 2.46829629e-01 -1.83175528e+00
-3.03063095e-01 -7.19875693e-01 -2.50078708e-01 7.89814770e-01
3.55048835e-01 -1.14521718e+00 7.56520092e-01 5.31753063e-01
-1.46909794e-02 -1.09146786e+00 -7.86076844e-01 -9.79375958e-01
3.60503078e-01 -1.02491722e-01 3.56238127e-01 6.35952115e-01
3.63272339e-01 1.84288859e-01 -3.79063010e-01 -1.80916443e-01
3.69947225e-01 1.17676504e-01 6.72121882e-01 -5.80466509e-01
-4.23321187e-01 -8.79761755e-01 1.21202111e-01 -1.15704751e+00
1.72022969e-01 -5.92070997e-01 3.19229029e-02 -1.60655844e+00
-1.48877501e-01 -4.76934940e-01 -3.30829561e-01 4.37897146e-01
-5.30312657e-01 -1.93891555e-01 2.90677100e-01 1.46490976e-01
-4.48083192e-01 9.03785944e-01 1.30939043e+00 2.24884734e-01
-4.70032305e-01 6.28203079e-02 -5.54807484e-01 5.19734502e-01
1.28725588e+00 -5.13225079e-01 -6.67466044e-01 -1.99130177e-01
-9.83511880e-02 6.71111941e-02 1.10719517e-01 -1.07934690e+00
1.45242468e-01 -2.98467785e-01 3.47892761e-01 -6.87401593e-02
2.02075809e-01 -3.12215835e-01 -3.60904634e-01 7.55680382e-01
-3.79227400e-01 4.64547306e-01 4.06328946e-01 4.57708836e-01
5.13828024e-02 -4.60101277e-01 7.25307465e-01 -2.44954094e-01
-6.33330464e-01 4.28607911e-02 -7.18496382e-01 3.09078097e-01
6.95543528e-01 2.44184613e-01 -2.04567149e-01 -8.18384588e-01
-1.79032475e-01 1.97878957e-01 2.95157731e-01 2.19820529e-01
3.37708861e-01 -1.12363589e+00 -5.95075190e-01 -1.39162600e-01
-6.50306702e-01 -6.00506306e-01 -6.23218864e-02 9.59554434e-01
-4.01150912e-01 5.53648591e-01 -3.66900831e-01 -1.77919969e-01
-8.46576750e-01 5.12916148e-01 5.46386600e-01 -6.02628469e-01
-5.50271034e-01 4.63426888e-01 -4.62202370e-01 -6.99999332e-01
2.92437524e-01 -3.78280163e-01 7.16781020e-02 -7.98073635e-02
4.01734561e-01 7.72162378e-01 -1.24926470e-01 -1.20381273e-01
-1.24707326e-01 2.41451010e-01 -3.19379452e-03 -6.26199484e-01
1.09000897e+00 6.03253543e-02 1.94804460e-01 3.39247376e-01
9.82972503e-01 -1.09982744e-01 -1.73350787e+00 -1.31530687e-01
-1.33956037e-03 -4.47054327e-01 2.54825950e-01 -7.78381050e-01
-1.19823337e+00 6.61456108e-01 9.20322359e-01 5.37309386e-02
7.73756325e-01 -7.40396917e-01 4.99267578e-01 5.03142953e-01
2.61926770e-01 -1.49422657e+00 2.53311574e-01 8.98561776e-01
7.72355735e-01 -1.13938391e+00 3.67899477e-01 2.92322129e-01
-7.31861353e-01 1.00273371e+00 6.67587399e-01 -5.14886141e-01
6.37058020e-01 -1.22423597e-01 2.15088740e-01 1.54595524e-01
-6.11649275e-01 -3.22258353e-01 1.21235199e-01 2.12495461e-01
4.40811366e-01 1.98278669e-02 -9.05794084e-01 -1.43470213e-01
-2.11732224e-01 7.34262988e-02 2.86057770e-01 9.79034007e-01
-4.04360294e-01 -1.07333839e+00 -2.87827194e-01 5.43894887e-01
-8.23817968e-01 -1.29727721e-01 6.86244741e-02 7.69861042e-01
-8.41827273e-01 1.07653737e+00 -3.90198916e-01 -2.16047257e-01
3.34642023e-01 1.74100354e-01 5.96010685e-01 -4.07987013e-02
-7.03510523e-01 -7.37622976e-02 1.47269577e-01 -7.40202069e-01
-2.63335109e-01 -6.25266671e-01 -1.32601416e+00 -6.01194263e-01
-2.63890833e-01 4.17366773e-01 9.82078910e-01 8.72705579e-01
4.37600404e-01 3.04318637e-01 4.85146940e-01 -1.08346879e+00
-1.60639060e+00 -1.26605380e+00 -4.33455318e-01 2.90158659e-01
2.87700415e-01 -9.19776797e-01 -4.52819467e-01 -5.89104176e-01]
|
[4.113807201385498, 2.3905954360961914]
|
2596f4b8-09dc-4476-b92b-9937a3f22adc
|
k-plug-knowledge-injected-pre-trained
| null | null |
https://openreview.net/forum?id=5WcLI0e3cAY
|
https://openreview.net/pdf?id=5WcLI0e3cAY
|
K-PLUG: KNOWLEDGE-INJECTED PRE-TRAINED LANGUAGE MODEL FOR NATURAL LANGUAGE UNDERSTANDING AND GENERATION
|
Existing pre-trained language models (PLMs) have demonstrated the effectiveness of self-supervised learning for a broad range of natural language processing (NLP) tasks. However, most of them are not explicitly aware of domain-specific knowledge, which is essential for downstream tasks in many domains, such as tasks in e-commerce scenarios. In this paper, we propose K-PLUG, a knowledge-injected pre-trained language model based on the encoder-decoder transformer that can be transferred to both natural language understanding and generation tasks. We verify our method in a diverse range of e-commerce scenarios that require domain-specific knowledge. Specifically, we propose five knowledge-aware self-supervised pre-training objectives to formulate the learning of domain-specific knowledge, including e-commerce domain-specific knowledge-bases, aspects of product entities, categories of product entities, and unique selling propositions of product entities. K-PLUG achieves new state-of-the-art results on a suite of domain-specific NLP tasks, including product knowledge base completion, abstractive product summarization, and multi-turn dialogue, significantly outperforms baselines across the board, which demonstrates that the proposed method effectively learns a diverse set of domain-specific knowledge for both language understanding and generation tasks. The code, data, and models will be publicly available.
|
['BoWen Zhou', 'Ying Liu', 'Xiaodong He', 'Youzheng Wu', 'Yujia Wang', 'Peng Yuan', 'Haoran Li', 'Song Xu']
|
2021-01-01
| null | null | null | null |
['knowledge-base-completion', 'knowledge-base-completion']
|
['graphs', 'knowledge-base']
|
[ 1.97346047e-01 5.55848777e-01 -6.52657092e-01 -5.75440109e-01
-1.21357131e+00 -9.78552997e-01 6.60371482e-01 2.48056874e-01
-1.08460568e-01 8.82100463e-01 6.64616168e-01 -2.99866527e-01
6.81061372e-02 -8.52850199e-01 -1.10138857e+00 -1.07290208e-01
1.90486908e-01 8.26408446e-01 -2.90300101e-01 -6.55382454e-01
-7.34328255e-02 -1.08944587e-01 -9.33938980e-01 7.34101236e-01
1.13164544e+00 8.70406985e-01 2.11990133e-01 4.90564704e-01
-3.69761348e-01 9.16156054e-01 -5.19692123e-01 -9.94898081e-01
9.31851938e-02 -3.32024664e-01 -1.24073863e+00 6.51667714e-02
2.71417260e-01 -4.83640343e-01 -9.86948013e-02 8.27158153e-01
2.29325786e-01 5.86170852e-02 8.07877600e-01 -9.77444112e-01
-1.17184448e+00 1.32090425e+00 -2.16991961e-01 -9.17853564e-02
4.48664755e-01 3.89009446e-01 1.54011917e+00 -7.57205665e-01
6.98655307e-01 1.37542105e+00 4.53664988e-01 4.45847422e-01
-1.16826725e+00 -5.66176116e-01 3.53263140e-01 9.95587856e-02
-1.05261290e+00 -3.60957146e-01 6.61889374e-01 -2.13372007e-01
1.57154036e+00 -3.30440193e-01 2.61001587e-01 1.26978946e+00
1.75820570e-02 1.29178512e+00 4.31593478e-01 -3.22835952e-01
3.63564156e-02 4.73973691e-01 3.93502831e-01 4.69505847e-01
4.27340567e-01 -2.35733435e-01 -7.08435297e-01 -1.97825022e-02
5.61771512e-01 -3.19846094e-01 -3.22483899e-03 -7.52154216e-02
-1.28893638e+00 9.97053325e-01 1.05821423e-01 -5.38883209e-02
-6.65715575e-01 -1.24379925e-01 7.04444528e-01 3.65294248e-01
6.45220518e-01 8.74003410e-01 -1.35586917e+00 -1.76926181e-01
-5.95871568e-01 3.65912437e-01 1.45850551e+00 1.45069158e+00
5.75333357e-01 -1.44753214e-02 -2.84683257e-01 8.83157849e-01
2.84464091e-01 6.56399548e-01 5.99168777e-01 -7.87145078e-01
9.71692920e-01 7.38957107e-01 1.59800902e-01 -6.08409822e-01
-2.16137052e-01 -3.49333107e-01 -8.75737369e-01 -8.32447112e-01
4.04382162e-02 -6.04735911e-01 -7.62832165e-01 1.79507577e+00
1.97939470e-01 -1.93902534e-02 9.32277799e-01 5.59843063e-01
1.41780496e+00 1.16059351e+00 3.15971285e-01 -2.41563261e-01
1.54022121e+00 -1.34286153e+00 -8.13601255e-01 -4.98897493e-01
8.19013596e-01 -6.81067944e-01 9.34209824e-01 1.90084487e-01
-1.14861882e+00 -6.13490582e-01 -8.66140187e-01 -4.86067802e-01
-4.91938591e-01 2.33269945e-01 9.81004417e-01 2.17659578e-01
-5.49102783e-01 2.82779127e-01 -4.41318274e-01 -3.02538693e-01
3.27725530e-01 1.15521491e-01 -2.32780188e-01 -1.87980577e-01
-1.66792202e+00 9.51427579e-01 9.19403195e-01 -3.12797688e-02
-8.83974314e-01 -1.09496522e+00 -1.39661014e+00 3.31463128e-01
5.82633495e-01 -1.20508909e+00 1.87842107e+00 -7.80652642e-01
-1.73594522e+00 6.40522301e-01 -1.16867475e-01 -7.84150064e-01
9.48948599e-03 -6.11406147e-01 -6.37456477e-01 -2.12282632e-02
2.34514236e-01 8.20052505e-01 5.66342056e-01 -1.17909110e+00
-8.09174716e-01 -1.34043619e-01 4.38452750e-01 4.24574167e-01
-1.32837579e-01 -2.86324024e-01 -3.06090206e-01 -4.51018095e-01
-5.13827324e-01 -5.98127782e-01 -9.71903577e-02 -8.64175797e-01
-6.57766044e-01 -5.56922436e-01 3.17311257e-01 -8.94881368e-01
1.15720606e+00 -1.96740901e+00 -7.37262657e-03 -2.72534817e-01
-1.88482225e-01 4.75817680e-01 -5.90139091e-01 6.63992047e-01
1.66989252e-01 1.05064869e-01 -9.46651995e-02 -2.31753960e-01
4.02010113e-01 2.84541398e-01 -8.21632028e-01 -5.07247269e-01
6.85633421e-01 1.45569718e+00 -1.05699027e+00 -2.94407696e-01
3.70014124e-02 2.11116567e-01 -5.09426832e-01 2.49167308e-01
-1.01569963e+00 2.06678867e-01 -7.34898269e-01 4.81269389e-01
4.94301647e-01 -4.45582330e-01 2.48467743e-01 -3.82807821e-01
3.50118190e-01 8.57641757e-01 -7.75182724e-01 1.91957116e+00
-1.00775385e+00 1.38546586e-01 -1.00432321e-01 -9.13604617e-01
7.30292022e-01 4.05594468e-01 1.27321184e-01 -5.19589186e-01
-8.14824179e-02 -1.57358143e-02 -2.44074404e-01 -5.56997180e-01
7.88399100e-01 -1.97949514e-01 -4.32518065e-01 5.06188333e-01
6.88296735e-01 -4.04494226e-01 5.65717578e-01 5.72304964e-01
8.94124746e-01 1.11539386e-01 6.55828834e-01 1.01842448e-01
4.32744712e-01 4.01236475e-01 4.46093678e-01 6.89441860e-01
3.08857203e-01 2.23298028e-01 4.99828637e-01 -1.76063985e-01
-8.70119095e-01 -1.00945067e+00 1.29081160e-01 1.34659660e+00
1.03108346e-01 -5.90241849e-01 -5.36226511e-01 -8.50227118e-01
2.71657914e-01 1.39739096e+00 -9.37394649e-02 -4.11935925e-01
-4.49170351e-01 -5.50537109e-01 3.97900581e-01 5.19317687e-01
7.72329509e-01 -1.23379052e+00 1.98304743e-01 4.72675443e-01
-5.10370851e-01 -1.62202132e+00 -5.27785063e-01 -6.13044538e-02
-8.33870769e-01 -9.58562613e-01 -4.40642238e-01 -1.09211981e+00
4.49058980e-01 6.90094829e-02 1.65287507e+00 -8.11839879e-01
2.34480187e-01 6.58581913e-01 -4.95469689e-01 -6.39214277e-01
-8.75721812e-01 5.75856268e-01 -2.32717678e-01 -6.45408928e-02
7.03056455e-01 -3.19061756e-01 -3.00646305e-01 -2.08073575e-02
-7.04220474e-01 2.19850689e-01 1.01344502e+00 7.97069013e-01
5.84451020e-01 8.43121260e-02 1.12085664e+00 -1.21005476e+00
1.08154559e+00 -6.32711172e-01 -3.14156741e-01 5.79263151e-01
-3.18566620e-01 2.62302190e-01 8.92389119e-01 -3.60825241e-01
-1.55171883e+00 4.12001275e-03 -1.47562936e-01 1.97659403e-01
-2.37471968e-01 1.02388251e+00 -5.14725089e-01 7.30552375e-01
5.25543332e-01 5.53597748e-01 -2.42319107e-01 -5.39385200e-01
1.12587786e+00 7.82563329e-01 5.50616860e-01 -6.29702330e-01
4.78587717e-01 -1.32698148e-01 -5.34871459e-01 -6.46274269e-01
-1.48908830e+00 -6.62029982e-01 -5.77965438e-01 5.11358261e-01
6.29568279e-01 -1.42349923e+00 -6.41669452e-01 2.39562824e-01
-1.53515005e+00 -2.73474008e-01 -6.53845012e-01 4.42024291e-01
-5.89767337e-01 2.28374109e-01 -7.46785402e-01 -4.05358434e-01
-9.98675346e-01 -6.43774331e-01 1.24350071e+00 2.31926754e-01
-3.77984196e-01 -1.25687671e+00 -4.91876528e-02 8.39592814e-01
2.66799986e-01 -2.93864399e-01 1.33497524e+00 -1.12172651e+00
-5.81370234e-01 -1.15633503e-01 -5.55499382e-02 8.38627815e-01
2.45823160e-01 -5.17023325e-01 -5.18951058e-01 2.08566915e-02
-1.82724461e-01 -8.09331298e-01 9.96185064e-01 3.03663492e-01
7.01618910e-01 -8.29296827e-01 -2.33897939e-01 2.84321934e-01
9.34449553e-01 -6.55321255e-02 2.62707710e-01 -2.47361120e-02
5.81213593e-01 5.87133944e-01 7.68526614e-01 3.78183842e-01
1.02418792e+00 4.97453839e-01 -9.32524074e-03 4.07025404e-02
-9.57758203e-02 -7.32932091e-01 6.00311697e-01 9.57052529e-01
3.00065458e-01 -4.39353615e-01 -5.73427796e-01 8.77224624e-01
-1.96466768e+00 -8.15712631e-01 2.52899855e-01 1.62166667e+00
1.57545602e+00 8.35575443e-03 -5.42914346e-02 -6.57977998e-01
5.16269147e-01 1.14879154e-01 -1.00350058e+00 -4.50728297e-01
-1.10685945e-01 1.09622985e-01 3.00169498e-01 4.59510028e-01
-1.17055750e+00 1.44894159e+00 6.01462650e+00 9.02851820e-01
-6.32918358e-01 1.11713432e-01 4.10768658e-01 2.72781365e-02
-4.11058098e-01 -1.15029790e-01 -1.29074895e+00 1.75942585e-01
1.01280475e+00 -8.24613690e-01 1.65408239e-01 9.77369308e-01
6.10106327e-02 1.70161873e-01 -1.54307210e+00 8.52124989e-01
1.23324551e-01 -1.38633931e+00 7.18796015e-01 -2.73751408e-01
1.00275588e+00 -1.71453245e-02 -2.94620413e-02 9.57508802e-01
9.41203773e-01 -7.57276595e-01 3.09614897e-01 1.34440348e-01
7.19554901e-01 -6.39152765e-01 8.55806708e-01 5.85630059e-01
-8.43584955e-01 -1.01071022e-01 -3.38768750e-01 1.35503905e-02
7.91450739e-01 8.03149641e-01 -1.16825938e+00 8.75320196e-01
7.39066601e-02 1.09903967e+00 -4.71842177e-02 3.11114848e-01
-6.86541855e-01 7.82957077e-01 -1.72196224e-01 5.07318862e-02
3.56390953e-01 6.18806519e-02 1.32461101e-01 1.39895535e+00
1.33216813e-01 4.04351354e-01 1.54173806e-01 1.05868721e+00
-6.31682873e-01 2.10661054e-01 -5.11392176e-01 -6.30795360e-01
3.35611701e-01 1.17607498e+00 9.54058301e-03 -6.60874426e-01
-5.96551001e-01 9.40658391e-01 1.58875510e-01 3.92836779e-01
-5.47164440e-01 -3.32575530e-01 7.42976606e-01 -1.68608591e-01
5.49672246e-01 -7.93595538e-02 -1.30486965e-01 -1.34434390e+00
1.00667559e-01 -1.01157641e+00 4.70429778e-01 -5.91086864e-01
-1.67501986e+00 2.96650380e-01 1.57078043e-01 -9.12105501e-01
-7.61188447e-01 -5.10172606e-01 -2.99838752e-01 6.66971922e-01
-1.93205464e+00 -1.54794598e+00 2.52757639e-01 4.56230283e-01
9.56312060e-01 -4.36532736e-01 9.21180069e-01 -2.60796584e-02
-2.26436183e-01 5.25563240e-01 1.59192905e-01 3.71940017e-01
8.21514368e-01 -1.10583925e+00 7.97422588e-01 4.78979647e-01
1.52481079e-01 8.15622866e-01 3.85752052e-01 -8.21504474e-01
-1.66338730e+00 -1.47605550e+00 1.31507206e+00 -6.63276911e-01
6.84099913e-01 -4.15751636e-01 -6.95873201e-01 1.14815259e+00
3.87550503e-01 -4.56936628e-01 9.51351821e-01 4.80584890e-01
-4.86954480e-01 -2.24074215e-01 -9.51661527e-01 3.12474996e-01
9.75606561e-01 -4.36820745e-01 -1.06732273e+00 8.60329390e-01
1.36984074e+00 -4.21586722e-01 -1.08537853e+00 4.25139755e-01
3.00234824e-01 -1.72781155e-01 1.05141211e+00 -1.02961361e+00
9.13963139e-01 1.02846265e-01 6.37974218e-02 -1.75848460e+00
-2.10459813e-01 -6.56353295e-01 -4.75207955e-01 1.60945714e+00
1.02017665e+00 -5.20582736e-01 3.94703746e-01 6.22435391e-01
-2.61423081e-01 -5.66237748e-01 -4.85011101e-01 -7.68219173e-01
1.85634971e-01 -2.46804595e-01 6.61731422e-01 9.17069018e-01
4.74780530e-01 1.29470909e+00 -1.99113190e-01 1.32879630e-01
2.39527836e-01 4.32499826e-01 7.93423355e-01 -8.70477319e-01
-3.76078129e-01 -2.97993332e-01 9.76550058e-02 -1.73324513e+00
5.96993327e-01 -1.12870729e+00 -1.47752166e-01 -2.03996801e+00
3.17543209e-01 -8.11069608e-02 1.55885080e-02 8.22957158e-01
-1.10183634e-01 -5.63184738e-01 1.30091280e-01 -8.06627572e-02
-1.07748318e+00 6.81314707e-01 1.44499433e+00 -4.91877317e-01
-4.42676425e-01 6.28975779e-02 -1.38812375e+00 4.47720021e-01
5.73182940e-01 -8.94307569e-02 -7.61492550e-01 -6.82756662e-01
4.07859296e-01 1.19798675e-01 -3.73404175e-02 -2.14643419e-01
2.81744272e-01 -1.45131707e-01 -3.75960320e-02 -4.19668913e-01
2.28532493e-01 -5.48221767e-01 -3.23077530e-01 1.09462790e-01
-8.77693236e-01 -3.10987234e-01 3.77149373e-01 4.63601291e-01
-6.20836973e-01 -1.19155571e-01 1.74530536e-01 -5.00667989e-01
-1.01258802e+00 2.04865098e-01 7.92526379e-02 4.82800841e-01
7.66172111e-01 2.45491967e-01 -5.46139300e-01 -6.98341131e-01
-5.70254326e-01 7.33687699e-01 -2.85145134e-01 7.70790100e-01
4.07498658e-01 -1.17398524e+00 -1.17132676e+00 9.03871953e-02
3.61545563e-01 3.88072729e-01 3.94375384e-01 2.70989656e-01
-1.14778923e-02 1.01631403e+00 6.04393929e-02 -1.98286369e-01
-7.74713397e-01 6.16106927e-01 -2.85533424e-02 -8.34225357e-01
-3.53101611e-01 7.97016323e-01 4.10139292e-01 -7.90715218e-01
9.84756723e-02 -6.31829858e-01 -2.70747066e-01 8.01253319e-03
6.12343729e-01 -3.69510204e-02 2.31583212e-02 -4.30334955e-01
-2.35610381e-02 2.98557043e-01 -6.73563838e-01 2.33889267e-01
1.31454670e+00 -2.70944059e-01 4.35350426e-02 2.23046914e-01
1.09626925e+00 -4.51872677e-01 -1.09292531e+00 -8.12121987e-01
-8.94711912e-02 1.86719418e-01 -1.37885645e-01 -1.53150630e+00
-7.71183193e-01 7.21274972e-01 -3.05472463e-01 -7.30983540e-02
1.02840340e+00 3.66589695e-01 1.37143576e+00 8.54197800e-01
4.43153054e-01 -1.10631168e+00 5.16825635e-03 9.87168968e-01
1.00294209e+00 -1.41983807e+00 -1.82859585e-01 -7.33773410e-01
-1.23882926e+00 8.27185929e-01 6.44525051e-01 3.40900600e-01
4.89131868e-01 4.97137979e-02 4.54471596e-02 -1.06207974e-01
-1.14666319e+00 -2.31520325e-01 2.67510533e-01 6.20717049e-01
4.84745324e-01 2.12595373e-01 -1.13969758e-01 1.31136513e+00
-5.20341098e-01 2.28163257e-01 2.26520121e-01 6.52348101e-01
-1.87486306e-01 -9.43271339e-01 3.78016293e-01 3.64526153e-01
-2.83870459e-01 -4.30520236e-01 -4.71172333e-01 6.18282676e-01
-1.67328671e-01 1.06484091e+00 -1.71809271e-01 -1.23305559e-01
6.75920308e-01 3.32234174e-01 4.78773266e-01 -1.09177101e+00
-6.00594580e-01 -3.72796178e-01 5.98212898e-01 -2.96398610e-01
-4.19174314e-01 -3.38392913e-01 -1.24499130e+00 -1.64910883e-01
-2.96528846e-01 2.25567639e-01 5.04684806e-01 1.05137157e+00
9.19805706e-01 4.11988735e-01 3.11465949e-01 -4.50520754e-01
-9.95300531e-01 -1.18825114e+00 -5.75566113e-01 7.19480157e-01
2.44933173e-01 -2.01047570e-01 -3.66632529e-02 5.09286702e-01]
|
[11.151171684265137, 8.366984367370605]
|
e397da90-e2bb-4869-9e2f-c8a8bec45863
|
joint-community-detection-and-rotational
|
2105.06031
| null |
https://arxiv.org/abs/2105.06031v1
|
https://arxiv.org/pdf/2105.06031v1.pdf
|
Joint Community Detection and Rotational Synchronization via Semidefinite Programming
|
In the presence of heterogeneous data, where randomly rotated objects fall into multiple underlying categories, it is challenging to simultaneously classify them into clusters and synchronize them based on pairwise relations. This gives rise to the joint problem of community detection and synchronization. We propose a series of semidefinite relaxations, and prove their exact recovery when extending the celebrated stochastic block model to this new setting where both rotations and cluster identities are to be determined. Numerical experiments demonstrate the efficacy of our proposed algorithms and confirm our theoretical result which indicates a sharp phase transition for exact recovery.
|
['Zhizhen Zhao', 'Yuehaw Khoo', 'Yifeng Fan']
|
2021-05-13
| null | null | null | null |
['stochastic-block-model']
|
['graphs']
|
[ 1.32316247e-01 -1.34366721e-01 -3.31807107e-01 1.61060810e-01
-6.65554881e-01 -8.81775022e-01 3.15540284e-01 3.02816957e-01
-2.21474677e-01 6.78245127e-01 1.36508167e-01 -1.13838822e-01
-4.89769310e-01 -3.18332583e-01 -5.56535125e-01 -1.15784538e+00
-4.48412925e-01 8.52865577e-01 -3.57110240e-02 2.83448547e-02
1.71548516e-01 3.51754248e-01 -8.24864626e-01 -1.99248999e-01
7.68614650e-01 3.25146496e-01 -1.06132619e-01 7.60976255e-01
5.82161129e-01 5.76623738e-01 -3.41759682e-01 5.54039031e-02
4.73102808e-01 -2.31750473e-01 -8.54716361e-01 7.11995065e-01
2.35113427e-02 6.16060458e-02 -5.51417470e-01 1.30971885e+00
5.10409653e-01 9.23935026e-02 4.37041074e-01 -1.62859476e+00
-5.50512075e-01 8.82799983e-01 -1.32592142e+00 2.67555445e-01
4.07770693e-01 -4.03444082e-01 1.12352526e+00 -6.82519972e-01
8.70425820e-01 9.26300585e-01 6.90496385e-01 5.96358031e-02
-1.61913025e+00 -8.05510879e-01 3.62304837e-01 2.56790847e-01
-1.98015630e+00 -4.37872678e-01 6.88176632e-01 -6.14651918e-01
2.89706409e-01 2.70681053e-01 5.75724423e-01 6.87182009e-01
-4.08592075e-01 6.60291016e-01 1.04999423e+00 -1.33188337e-01
2.51195788e-01 -1.44915670e-01 2.64912218e-01 3.30073535e-01
1.06755054e+00 -4.38846439e-01 -4.77743715e-01 -6.07425809e-01
6.14991724e-01 2.67890364e-01 -6.28757179e-01 -8.25381160e-01
-1.63361287e+00 7.80666471e-01 3.20091039e-01 2.36282438e-01
-3.79090428e-01 6.71126693e-03 1.46467924e-01 1.63970545e-01
3.23594421e-01 1.69744492e-02 1.19256921e-01 2.17350617e-01
-9.41228330e-01 8.89078826e-02 1.04412246e+00 1.18420017e+00
5.67773342e-01 -2.51642048e-01 1.06599316e-01 2.52166599e-01
1.91892475e-01 7.69091845e-01 -1.81520373e-01 -9.33397949e-01
6.57510996e-01 3.54886591e-01 4.90758926e-01 -1.62295723e+00
-3.05987328e-01 -6.10652149e-01 -1.55673385e+00 -6.80230916e-01
4.41326737e-01 -2.29548607e-02 -4.07069743e-01 1.94389451e+00
6.07985079e-01 6.21892929e-01 -1.47266258e-02 1.08019483e+00
3.10240984e-01 4.62355137e-01 -4.34783071e-01 -7.70532668e-01
8.86380613e-01 -6.03156984e-01 -7.58535326e-01 9.93351489e-02
4.17596072e-01 -4.86592799e-01 -4.70362641e-02 2.37258613e-01
-1.06245458e+00 5.81368688e-04 -1.01202667e+00 3.93284678e-01
2.60265440e-01 -2.51561970e-01 5.33676922e-01 3.70584637e-01
-1.23011196e+00 1.64274544e-01 -1.00318980e+00 -3.42106521e-01
1.21900171e-01 7.29067922e-01 -5.66195667e-01 -3.20601493e-01
-8.50448787e-01 2.08321527e-01 2.21013099e-01 5.87130904e-01
-6.01638496e-01 -2.29459271e-01 -5.95194638e-01 -2.20431760e-01
4.62194920e-01 -7.11270630e-01 6.37927175e-01 -6.48679197e-01
-7.15022981e-01 8.36742759e-01 -3.37138414e-01 -3.65328640e-01
6.36750340e-01 2.65135229e-01 -8.25945064e-02 4.03722435e-01
4.82370794e-01 3.24966870e-02 5.30148625e-01 -1.37697554e+00
-5.79384327e-01 -3.37273657e-01 5.74243581e-03 3.30223769e-01
-2.65834153e-01 -1.21586844e-02 -5.79259932e-01 -3.58489692e-01
5.35010457e-01 -1.50412726e+00 -7.12640524e-01 -2.89978176e-01
-8.99773538e-01 1.03158802e-02 6.07987523e-01 -4.12481129e-01
1.16522658e+00 -2.10483718e+00 6.15993977e-01 7.75985777e-01
7.92961597e-01 -2.58941680e-01 9.36270356e-02 5.19402683e-01
-2.64987111e-01 1.04460165e-01 -8.12937915e-02 -3.44755411e-01
-1.07221432e-01 1.92671061e-01 -2.06178397e-01 1.38558853e+00
-3.23582560e-01 2.71406651e-01 -1.06824136e+00 -4.79932725e-01
-1.71667382e-01 2.16751769e-01 -7.00705409e-01 -2.38693401e-01
7.08797097e-01 5.71226358e-01 -5.28149009e-01 3.40313703e-01
1.06316113e+00 -8.82288694e-01 8.24918926e-01 -9.52405334e-02
1.61434054e-01 -2.61090219e-01 -1.93787050e+00 1.09352732e+00
2.03058362e-01 6.11875057e-01 5.58222353e-01 -1.32923388e+00
4.28924292e-01 5.24742782e-01 9.51171041e-01 1.61189169e-01
4.05363590e-02 9.05277282e-02 4.36631739e-02 9.92951356e-03
5.89973569e-01 1.25623085e-02 -3.34741652e-01 6.34753466e-01
-4.64257836e-01 3.33565921e-01 2.67872244e-01 6.28610790e-01
1.21598089e+00 -6.89297020e-01 3.13489377e-01 -5.59164882e-01
3.64532322e-01 -7.14962408e-02 6.83410048e-01 9.45686936e-01
-2.82789201e-01 7.49151051e-01 5.24051428e-01 9.98470411e-02
-1.05378044e+00 -1.11602485e+00 3.67144984e-03 4.35505003e-01
6.80004001e-01 -4.12956774e-01 -4.23948497e-01 -1.01652250e-01
-2.82111578e-02 -1.52404517e-01 -7.49744058e-01 -9.68749151e-02
-3.84049058e-01 -1.11711657e+00 8.64846036e-02 2.50481397e-01
3.25463891e-01 9.89299361e-03 3.35204080e-02 2.06110671e-01
-7.14288771e-01 -1.34295869e+00 -8.30525815e-01 -2.20896918e-02
-7.54713476e-01 -1.34918082e+00 -8.73041928e-01 -8.84747803e-01
9.94396627e-01 9.60389674e-01 8.84165049e-01 2.24569231e-01
-2.04249807e-02 5.38969517e-01 -1.98950231e-01 3.40071678e-01
-1.08458795e-01 1.76156715e-01 4.51554149e-01 3.47846836e-01
-1.43663958e-01 -7.79170215e-01 -7.49795377e-01 4.19304371e-01
-8.73716533e-01 -6.78245798e-02 2.26616368e-01 6.75347686e-01
6.50925875e-01 3.54357898e-01 2.45161921e-01 -7.21494973e-01
4.11222011e-01 -8.15339744e-01 -6.17930233e-01 3.44179124e-01
-2.61705339e-01 -5.82794696e-02 1.65363654e-01 -5.63371599e-01
-5.33859372e-01 4.44498688e-01 8.81471574e-01 -4.19911981e-01
5.00334084e-01 6.42928958e-01 1.81179494e-02 -1.69310004e-01
1.00987121e-01 2.08081335e-01 -1.08143881e-01 -2.62062363e-02
4.07728940e-01 6.23474658e-01 5.62324882e-01 -5.51053822e-01
1.13686323e+00 1.04077780e+00 3.16751860e-02 -9.10159767e-01
-6.83519542e-01 -8.99822950e-01 -6.81690454e-01 -1.65339842e-01
4.39688295e-01 -1.38128555e+00 -9.76724029e-01 2.70773619e-01
-1.07211328e+00 9.34115984e-03 8.35767668e-03 4.24430400e-01
-3.76817226e-01 8.03583920e-01 -5.95518291e-01 -8.14221025e-01
8.66336375e-02 -9.25729990e-01 9.36564207e-01 7.34034330e-02
1.00368308e-02 -1.01610053e+00 3.85864347e-01 2.94479221e-01
-9.38175470e-02 4.30485338e-01 2.73325771e-01 -5.12134075e-01
-7.58597612e-01 -4.06566471e-01 -3.54938924e-01 -3.90824646e-01
2.28071168e-01 2.43351698e-01 -2.32031599e-01 -7.98467577e-01
-1.90171644e-01 1.67639136e-01 6.17487431e-01 5.52593946e-01
5.75180650e-01 -4.86236304e-01 -7.99637735e-01 4.27061439e-01
1.43886697e+00 -3.17268431e-01 1.82827845e-01 8.18314180e-02
8.43058586e-01 2.89315879e-01 4.58671600e-01 8.29569042e-01
6.77139699e-01 6.49596214e-01 2.94921309e-01 5.81661053e-03
4.44396377e-01 2.77621716e-01 1.74449444e-01 1.30048931e+00
-2.75915086e-01 -3.96377712e-01 -9.23246741e-01 8.45863819e-01
-2.24621391e+00 -1.18706143e+00 -5.89839101e-01 2.45222783e+00
1.02307475e+00 -2.99398303e-01 3.24040174e-01 8.27122778e-02
1.50406742e+00 -2.43004765e-02 -2.53470719e-01 2.53026128e-01
-4.34234023e-01 -1.18938617e-01 9.20999169e-01 4.63320017e-01
-1.24428833e+00 4.41192955e-01 7.02675295e+00 5.17417967e-01
-8.76341701e-01 7.22407922e-02 5.85209012e-01 -8.36099386e-02
-8.17940459e-02 2.06776783e-01 -5.61038613e-01 2.02022180e-01
5.94688356e-01 -7.58997023e-01 4.75209981e-01 4.60661888e-01
2.27861911e-01 -8.85397345e-02 -9.37880576e-01 9.72277641e-01
-4.70991954e-02 -1.06931460e+00 -3.74438941e-01 3.65046531e-01
1.33815527e+00 -1.23643540e-02 6.41818419e-02 -4.16799307e-01
9.18514848e-01 -6.80363178e-01 5.27393520e-01 1.46152928e-01
4.80545074e-01 -8.41032207e-01 4.07142639e-01 1.70008227e-01
-1.58118296e+00 -9.39933136e-02 -4.53328192e-01 -9.09019336e-02
2.71121413e-01 6.32521212e-01 -8.02900612e-01 8.64094138e-01
4.30059582e-01 1.06516325e+00 -4.40629870e-01 1.35345876e+00
1.16009288e-01 4.83995646e-01 -7.23174214e-01 3.05787444e-01
-7.24557042e-02 -5.40437162e-01 6.83505893e-01 8.81213009e-01
3.62100631e-01 3.32628459e-01 5.11465371e-01 4.06978160e-01
-5.12138121e-02 -2.31925994e-02 -4.35365379e-01 5.12746125e-02
8.27570558e-01 1.04405355e+00 -1.30162823e+00 -2.55420834e-01
-1.34150088e-01 1.06054521e+00 2.10702643e-01 4.97045606e-01
-8.50757003e-01 -2.20769465e-01 5.18167973e-01 -1.30501866e-01
5.18276036e-01 -5.52765489e-01 -2.39519533e-02 -1.56095517e+00
-7.66409934e-03 -8.63660574e-01 5.44301808e-01 -3.74494314e-01
-1.32932460e+00 2.55284458e-01 -2.71664113e-02 -1.37959433e+00
2.58157104e-02 1.07124560e-01 -3.06973219e-01 3.00615162e-01
-9.04681802e-01 -9.13108408e-01 -1.63755924e-01 8.56551707e-01
-3.64437014e-01 5.41911900e-01 3.97865474e-01 5.56251466e-01
-8.86403024e-01 3.58110279e-01 7.70091295e-01 2.37119257e-01
6.80162668e-01 -1.23519087e+00 -1.61587745e-01 1.16793394e+00
7.21559152e-02 6.23432040e-01 1.06991363e+00 -5.49191415e-01
-1.49540186e+00 -1.12172329e+00 7.94797361e-01 -1.29611343e-01
1.05177760e+00 -4.51597691e-01 -6.03557110e-01 7.96011329e-01
1.94005072e-01 2.67372787e-01 7.54253149e-01 1.30833983e-01
-3.31225246e-01 -9.42862332e-02 -8.05041969e-01 5.08486748e-01
1.20163798e+00 -4.55170482e-01 -2.05702886e-01 8.38537037e-01
3.46294135e-01 -1.85214728e-01 -7.93170989e-01 1.25254929e-01
2.09752142e-01 -5.84017754e-01 1.00803077e+00 -3.40308368e-01
-4.84854504e-02 -5.06431282e-01 -2.32489601e-01 -1.06453919e+00
-3.26163083e-01 -1.05446923e+00 -5.59424097e-03 1.19989920e+00
1.23498239e-01 -6.03609324e-01 8.42962980e-01 4.32167470e-01
6.31393313e-01 -3.04663237e-02 -1.16896594e+00 -8.04498196e-01
-2.45715827e-01 1.09244362e-02 1.54857382e-01 1.44759429e+00
3.37884277e-01 2.26556152e-01 -5.58459699e-01 7.38684058e-01
1.12921417e+00 5.94234228e-01 6.81642830e-01 -1.21341622e+00
-2.39014074e-01 -2.79238492e-01 -5.77248335e-01 -1.05675018e+00
9.20069516e-02 -7.79179990e-01 1.60930306e-01 -1.23778892e+00
9.48955595e-01 -7.09499657e-01 -2.87683547e-01 9.23633128e-02
-2.37387866e-01 5.33928275e-01 1.86542153e-01 6.90025151e-01
-1.34203696e+00 4.02709693e-01 9.32652116e-01 -1.45923957e-01
-2.33933643e-01 2.68739730e-01 -8.72430265e-01 3.95746827e-01
3.41565937e-01 -6.49907589e-01 -2.76630342e-01 -2.25162461e-01
6.20491385e-01 4.46217030e-01 2.74594188e-01 -8.84008169e-01
6.55950606e-01 -2.57886142e-01 -1.92162499e-01 -6.60677195e-01
1.11280680e-01 -7.96654761e-01 6.11996353e-01 6.88639879e-01
-2.08069608e-01 9.59903896e-02 -2.32655928e-01 1.20177436e+00
-1.84263051e-01 4.09922481e-01 6.29314840e-01 4.78473693e-01
-5.37211299e-02 5.94109893e-01 -4.48828220e-01 2.71821022e-01
1.26343167e+00 -2.02898636e-01 -7.90362656e-02 -8.13239694e-01
-1.11087358e+00 6.02575660e-01 6.42324686e-01 -1.75251812e-02
3.09012681e-01 -1.47175932e+00 -9.36159790e-01 -2.46627286e-01
-1.84175447e-02 1.45576775e-01 3.44876677e-01 1.52077055e+00
-4.06509876e-01 3.53704363e-01 2.33888343e-01 -1.02580202e+00
-1.50087643e+00 7.20654786e-01 1.51841015e-01 -3.75024378e-01
-2.08654299e-01 5.26525497e-01 2.93359458e-01 -1.13314554e-01
3.61268744e-02 -4.19661514e-02 -3.30944322e-02 2.32655466e-01
2.79722929e-01 4.63312954e-01 -2.15800777e-01 -9.95418429e-01
-5.56093454e-01 4.74857777e-01 7.38710165e-02 -1.09503753e-02
1.46717954e+00 -6.93879128e-01 -5.38790584e-01 2.06646487e-01
1.12465572e+00 2.83847034e-01 -9.37426686e-01 -5.09161413e-01
-1.17028896e-02 -2.98880994e-01 -3.41920674e-01 2.78120399e-01
-1.33672285e+00 3.65116268e-01 2.72789091e-01 4.28756356e-01
1.05950344e+00 1.84119433e-01 3.44012976e-01 4.78542089e-01
4.76340204e-01 -8.08281004e-01 -1.17411599e-01 3.63893360e-01
3.36395770e-01 -9.24384475e-01 3.51398975e-01 -8.05142164e-01
-2.33308107e-01 7.96392441e-01 5.05460463e-02 -4.94433850e-01
8.66040945e-01 1.64098993e-01 -5.76717257e-01 1.04463473e-02
-7.50364125e-01 -1.09790951e-01 -4.21735011e-02 6.63257003e-01
6.10849932e-02 2.90511340e-01 -2.94011980e-01 3.76612633e-01
-1.87259894e-02 -2.90006489e-01 1.00900924e+00 8.84206176e-01
-9.90575626e-02 -9.85047042e-01 -6.16674721e-01 6.17974475e-02
-3.76732171e-01 8.25157538e-02 -3.71580064e-01 7.08038032e-01
-3.61608118e-01 1.18158233e+00 1.35796428e-01 -1.83595702e-01
-1.22579344e-01 -6.27944410e-01 4.64303076e-01 -5.51666558e-01
-1.52474195e-01 2.48400733e-01 -1.69138983e-01 -2.82796115e-01
-1.01085293e+00 -1.08910048e+00 -7.53560245e-01 -7.74666965e-01
-7.07447946e-01 4.86668825e-01 9.21010971e-02 9.28369701e-01
3.70983958e-01 1.35042191e-01 1.14123964e+00 -8.25578809e-01
-6.05772495e-01 -5.07979751e-01 -8.98382664e-01 2.89855987e-01
4.78405237e-01 -4.95401084e-01 -6.51041269e-01 1.11831196e-01]
|
[6.919951915740967, 5.0616631507873535]
|
5099fe6f-09dd-4ced-ac40-ce31f42364c3
|
acq-improving-generative-data-free
|
2301.07266
| null |
https://arxiv.org/abs/2301.07266v1
|
https://arxiv.org/pdf/2301.07266v1.pdf
|
ACQ: Improving Generative Data-free Quantization Via Attention Correction
|
Data-free quantization aims to achieve model quantization without accessing any authentic sample. It is significant in an application-oriented context involving data privacy. Converting noise vectors into synthetic samples through a generator is a popular data-free quantization method, which is called generative data-free quantization. However, there is a difference in attention between synthetic samples and authentic samples. This is always ignored and restricts the quantization performance. First, since synthetic samples of the same class are prone to have homogenous attention, the quantized network can only learn limited modes of attention. Second, synthetic samples in eval mode and training mode exhibit different attention. Hence, the batch-normalization statistics matching tends to be inaccurate. ACQ is proposed in this paper to fix the attention of synthetic samples. An attention center position-condition generator is established regarding the homogenization of intra-class attention. Restricted by the attention center matching loss, the attention center position is treated as the generator's condition input to guide synthetic samples in obtaining diverse attention. Moreover, we design adversarial loss of paired synthetic samples under the same condition to prevent the generator from paying overmuch attention to the condition, which may result in mode collapse. To improve the attention similarity of synthetic samples in different network modes, we introduce a consistency penalty to guarantee accurate BN statistics matching. The experimental results demonstrate that ACQ effectively improves the attention problems of synthetic samples. Under various training settings, ACQ achieves the best quantization performance. For the 4-bit quantization of Resnet18 and Resnet50, ACQ reaches 67.55% and 72.23% accuracy, respectively.
|
['Huaxiang Lu', 'Wenyu Mao', 'Gang Chen', 'Min Jin', 'Guoliang Gong', 'Benzhe Dai', 'Xiaozhou Guo', 'Jixing Li']
|
2023-01-18
| null | null | null | null |
['data-free-quantization', 'data-free-quantization']
|
['computer-vision', 'methodology']
|
[ 1.21616744e-01 -1.89904839e-01 -2.88132608e-01 -4.70986366e-01
-8.74550760e-01 -4.14681107e-01 3.68366182e-01 -8.38019699e-02
-5.74060500e-01 7.86258280e-01 -5.30477799e-02 -1.01608172e-01
2.31197476e-01 -9.57583964e-01 -6.61510468e-01 -9.13360596e-01
5.64267159e-01 -6.96410006e-03 -1.19372867e-02 -5.29124700e-02
1.57073036e-01 3.37911278e-01 -1.40135145e+00 2.42941961e-01
1.15639293e+00 1.46034312e+00 2.29662254e-01 3.15741986e-01
-2.50217646e-01 6.17785871e-01 -1.21346033e+00 -7.27487206e-01
4.63918895e-01 -5.93840957e-01 -5.32365739e-01 -1.38027966e-01
4.71204579e-01 -6.46862566e-01 -3.45723122e-01 1.64391005e+00
7.90456295e-01 -2.49999296e-02 5.12494326e-01 -1.71264768e+00
-1.07104194e+00 6.88654721e-01 -3.68656129e-01 2.00111911e-01
-3.14375535e-02 2.79820323e-01 9.45115924e-01 -6.78431749e-01
1.28047571e-01 1.37466133e+00 3.10820788e-01 7.95156896e-01
-1.32211697e+00 -1.29372656e+00 1.23952262e-01 3.42384219e-01
-1.81293595e+00 -4.79777694e-01 7.11202621e-01 -2.16171637e-01
3.73240173e-01 4.52735186e-01 3.69517356e-01 1.06731856e+00
8.09909254e-02 8.08242917e-01 7.31989503e-01 1.56482607e-02
4.37067837e-01 3.46025974e-01 -1.97727472e-01 1.01475649e-01
1.86249867e-01 -7.65269026e-02 -2.78074831e-01 -7.94395432e-02
7.87320316e-01 2.17477188e-01 -5.13120234e-01 -6.42497465e-02
-9.80893731e-01 8.27948332e-01 6.62553787e-01 -1.22783789e-02
-2.78272241e-01 1.97528079e-01 5.88866591e-01 3.13129634e-01
1.37987852e-01 2.30832472e-01 -2.66073853e-01 -1.69576891e-02
-8.23216200e-01 2.89592803e-01 2.46562466e-01 1.46451938e+00
9.36921120e-01 4.17612642e-01 -5.86417794e-01 7.13314474e-01
1.58059210e-01 6.58990622e-01 7.72826970e-01 -9.75724101e-01
8.53738725e-01 4.08656716e-01 -5.09901829e-02 -1.24360681e+00
4.01749998e-01 -4.78013873e-01 -1.37951815e+00 -1.80558208e-02
2.73683667e-01 -1.40848935e-01 -8.09498966e-01 1.93474579e+00
9.00215507e-02 1.55105382e-01 5.50881736e-02 8.73609364e-01
5.96705854e-01 6.29551888e-01 1.76523075e-01 -3.07410091e-01
1.20501053e+00 -5.01149893e-01 -1.10029948e+00 1.14193652e-02
4.04744685e-01 -6.17461205e-01 1.36856532e+00 3.89214575e-01
-8.96116853e-01 -7.30326891e-01 -1.19098389e+00 -1.60341449e-02
-4.00847644e-01 1.47017539e-01 5.93473651e-02 7.87268341e-01
-6.26475811e-01 5.11066318e-01 -6.03059113e-01 2.03301370e-01
8.19769740e-01 3.84856045e-01 -2.95109004e-02 -1.97866216e-01
-1.69239378e+00 3.10341030e-01 5.58393180e-01 8.51828307e-02
-7.80680239e-01 -8.16218376e-01 -8.15890312e-01 2.41669595e-01
1.66509137e-01 -2.43359253e-01 1.06409085e+00 -9.91857886e-01
-1.17139995e+00 3.24763268e-01 -9.98664871e-02 -5.29938638e-01
8.27154577e-01 1.00403897e-01 -5.68701625e-01 -8.51501077e-02
3.85332763e-01 7.37311959e-01 1.11148226e+00 -1.11839819e+00
-6.22885823e-01 -1.96512565e-01 -9.68581960e-02 5.89973070e-02
-8.34893644e-01 -3.29788595e-01 -4.95849907e-01 -1.00021207e+00
5.33736981e-02 -4.51986194e-01 -3.12527306e-02 1.19453289e-01
-6.15438402e-01 -8.35501589e-03 8.78212094e-01 -4.49871272e-01
1.51537836e+00 -2.45714998e+00 -3.26650590e-01 3.14412415e-01
2.69261330e-01 3.07287842e-01 7.53834620e-02 -1.32744595e-01
-1.06548674e-01 6.81021035e-01 -1.93150938e-01 -4.39559191e-01
1.55494511e-01 3.15625697e-01 -5.16912162e-01 2.42300928e-01
3.28296244e-01 7.13483453e-01 -9.71365869e-01 -5.24552166e-01
1.14316337e-01 4.48410183e-01 -7.65794337e-01 3.05986881e-01
-6.66994229e-02 1.71104744e-01 -5.01221061e-01 5.58337331e-01
1.16255856e+00 -6.95716217e-03 -2.73604453e-01 -5.03349364e-01
3.02200615e-01 -5.47997095e-03 -1.28131187e+00 1.45008171e+00
-2.91414440e-01 3.69878441e-01 7.56691992e-02 -7.45852172e-01
1.05012953e+00 3.23232859e-01 -6.71461374e-02 -7.69333184e-01
3.00067604e-01 1.22242622e-01 -3.55988480e-02 -2.35575020e-01
6.40580058e-01 -6.16613477e-02 -1.02960460e-01 9.65426192e-02
-1.51225761e-01 -1.97082162e-01 -1.18909873e-01 1.29260257e-01
6.96680307e-01 -4.86551374e-01 -1.20574096e-02 -1.84060976e-01
6.40908659e-01 -5.93817472e-01 1.01861608e+00 7.29162633e-01
-4.85744029e-01 8.98708582e-01 5.10030270e-01 -5.72424009e-02
-1.04098582e+00 -9.36625063e-01 -2.42872179e-01 8.07488084e-01
2.34303564e-01 -3.75888139e-01 -8.86161208e-01 -7.12667704e-01
-9.70893502e-02 8.55154574e-01 -6.00414693e-01 -6.35689676e-01
-2.07620010e-01 -7.08131552e-01 7.06059337e-01 4.88646477e-01
9.38373089e-01 -1.06983769e+00 -4.76516895e-02 1.37145221e-01
-2.62826115e-01 -8.30242574e-01 -9.18136954e-01 8.51118863e-02
-4.63514328e-01 -8.26805174e-01 -6.96436167e-01 -5.26875973e-01
7.75375307e-01 3.07446290e-02 7.67210782e-01 -1.80398393e-02
3.85586545e-02 -2.39455327e-01 -3.29077899e-01 -3.55068237e-01
-5.70639074e-01 2.34908327e-01 1.68386593e-01 3.12804371e-01
6.01050138e-01 -5.82669437e-01 -6.88417315e-01 3.60667169e-01
-1.25224733e+00 -3.88843954e-01 3.49907309e-01 1.01476574e+00
7.06872642e-01 3.85260314e-01 7.32588947e-01 -5.56473315e-01
8.88524055e-01 -5.50685048e-01 -5.59964120e-01 3.42154875e-02
-6.24568105e-01 -1.69198643e-02 1.32882535e+00 -7.40930200e-01
-5.49085021e-01 -2.73480296e-01 -7.75062740e-02 -1.06499946e+00
-7.19841048e-02 -7.00472435e-03 -1.05297339e+00 1.82861716e-01
5.84639549e-01 4.62192118e-01 -2.89601982e-02 -2.23084062e-01
2.64045507e-01 8.69603515e-01 4.70781118e-01 -4.68942434e-01
8.08253288e-01 1.97626606e-01 -4.00101036e-01 -5.38759470e-01
-2.94227302e-01 1.28850371e-01 -1.41770452e-01 1.59780592e-01
9.55210268e-01 -8.43566537e-01 -7.49074876e-01 5.70469141e-01
-1.07182205e+00 -4.23932709e-02 -3.59927595e-01 3.53464782e-01
-2.70763963e-01 3.10689747e-01 -4.43828851e-01 -7.41692305e-01
-3.61079425e-01 -1.62756705e+00 8.89592052e-01 2.69176990e-01
7.25166127e-02 -5.92496037e-01 -6.65832162e-01 7.85431266e-02
3.92758518e-01 7.78169185e-02 7.86214352e-01 -7.45832205e-01
-7.13963270e-01 -1.33357197e-01 -3.36329430e-01 8.54351759e-01
4.78367835e-01 -1.29962355e-01 -1.11009347e+00 -3.94697487e-01
9.12190080e-02 -3.05447459e-01 6.75932109e-01 5.80696091e-02
1.83225489e+00 -5.92916071e-01 4.40033386e-03 8.86993051e-01
1.28841865e+00 4.54295367e-01 8.56517553e-01 1.52695030e-01
8.78614902e-01 1.20815724e-01 6.31974280e-01 5.27464509e-01
1.03897922e-01 3.29464972e-01 7.50040352e-01 8.75788108e-02
1.19977288e-01 -4.42535222e-01 1.40708879e-01 6.29247904e-01
5.05223691e-01 -3.16351116e-01 -5.16746879e-01 5.05582035e-01
-1.41427481e+00 -1.03632557e+00 3.81174326e-01 2.46807790e+00
1.15940905e+00 3.66329700e-01 -3.55649441e-01 4.88025546e-01
1.07594526e+00 1.45466924e-01 -8.33154500e-01 -1.61927879e-01
-1.11573130e-01 5.73612601e-02 7.20786095e-01 3.47234219e-01
-1.02871478e+00 6.77575350e-01 5.13778496e+00 1.73465061e+00
-1.33403480e+00 4.74593341e-02 1.00737083e+00 -9.87055078e-02
-4.98701096e-01 -3.58993620e-01 -1.04794109e+00 1.21045947e+00
5.71876287e-01 -2.56591856e-01 5.86775124e-01 8.67996037e-01
2.41988704e-01 4.04577017e-01 -1.10976017e+00 1.18116260e+00
-1.75266072e-01 -1.11885488e+00 5.08039474e-01 1.51102096e-01
6.70889080e-01 -3.87513429e-01 4.93923604e-01 3.59873205e-01
9.11190808e-02 -1.02655065e+00 8.48910391e-01 3.94109398e-01
1.15017045e+00 -1.23690546e+00 8.35594058e-01 3.39217633e-01
-1.09552121e+00 -9.41166356e-02 -7.35807836e-01 2.51494646e-01
-1.69860959e-01 5.19271255e-01 -7.05920458e-01 5.62623620e-01
7.23580003e-01 5.62306762e-01 -5.24018347e-01 6.63347542e-01
-1.02877170e-01 5.47382355e-01 -1.51312098e-01 2.61904672e-02
1.11674584e-01 -3.18068027e-01 3.25738758e-01 8.06779385e-01
2.73527116e-01 -7.82969818e-02 5.44795915e-02 1.28237951e+00
-4.49926436e-01 5.79592772e-02 -3.42299342e-01 1.47713110e-01
1.13386226e+00 8.71545374e-01 -2.26031810e-01 -4.25988913e-01
-8.49282965e-02 1.04751873e+00 6.34149015e-02 5.41741312e-01
-1.03424168e+00 -9.05182064e-01 9.78529871e-01 -1.35458037e-02
4.57758874e-01 3.82217020e-01 -3.72194767e-01 -1.00047255e+00
1.94074780e-01 -1.01010013e+00 5.95621504e-02 -6.46661580e-01
-1.57528651e+00 6.80221856e-01 -1.46481395e-01 -1.56431198e+00
1.15754595e-03 -5.78695722e-02 -6.29147649e-01 1.28211808e+00
-1.36279452e+00 -7.42045641e-01 -3.89403075e-01 7.01412201e-01
4.02556956e-01 -2.56576002e-01 6.58217430e-01 6.90187573e-01
-7.92839170e-01 1.57712293e+00 3.19668561e-01 5.09759188e-01
8.06155920e-01 -1.11917281e+00 2.96660274e-01 9.31596458e-01
-4.78808105e-01 8.33969772e-01 5.00337958e-01 -5.25796950e-01
-9.48670745e-01 -1.77113175e+00 4.72733974e-01 -9.40425321e-02
3.98650497e-01 -6.41938388e-01 -1.19969666e+00 5.08440495e-01
-4.49452624e-02 2.39811555e-01 6.10369265e-01 -7.30673015e-01
-3.26938123e-01 -4.82783735e-01 -1.57930112e+00 7.92244613e-01
6.34368122e-01 -7.53387332e-01 -2.74082869e-01 5.87522127e-02
1.15181363e+00 -2.66776413e-01 -1.03203917e+00 2.12397054e-01
1.87392175e-01 -8.61730039e-01 9.62871850e-01 -2.88755536e-01
3.13645512e-01 -5.56927145e-01 -4.31283891e-01 -1.06297600e+00
-1.82964638e-01 -6.86743796e-01 -2.64275558e-02 1.80278969e+00
1.50067210e-01 -8.17847610e-01 9.08706367e-01 6.87152147e-01
7.39674792e-02 -6.97853923e-01 -9.35369253e-01 -8.17728996e-01
3.00277859e-01 -3.90327036e-01 1.21899772e+00 1.04016876e+00
-5.67852259e-01 7.51197413e-02 -3.22385103e-01 1.78911552e-01
7.15722740e-01 -3.26702744e-01 6.23114288e-01 -8.03556859e-01
6.77525178e-02 -5.31453371e-01 -4.71879989e-01 -1.18756974e+00
2.22610906e-01 -7.78893888e-01 -8.74209702e-02 -9.27318513e-01
-2.03556031e-01 -7.20925093e-01 -4.08540905e-01 3.62638772e-01
-4.46703434e-01 3.38203996e-01 1.41381532e-01 2.53291398e-01
-2.87152439e-01 9.77723420e-01 1.37071013e+00 -3.88041914e-01
-8.75849128e-02 6.43845797e-02 -7.74949372e-01 2.50970364e-01
1.04498410e+00 -5.11921167e-01 -7.46786416e-01 -2.96048641e-01
-1.79287061e-01 -2.97122598e-01 3.02374840e-01 -9.13452268e-01
2.43322924e-01 -1.33010253e-01 5.90276778e-01 -6.49979413e-01
2.80104369e-01 -1.05306613e+00 6.03505261e-02 3.68763924e-01
-3.76701593e-01 -1.52105018e-01 4.66038547e-02 6.66130900e-01
-2.93658614e-01 -2.54399449e-01 8.35077405e-01 -3.24271061e-02
-3.87081891e-01 7.29890883e-01 -1.03283569e-01 4.00558949e-01
8.33237350e-01 -4.48550582e-01 -1.08012587e-01 -3.30953032e-01
-4.60989296e-01 4.54172254e-01 5.17282069e-01 2.47317553e-01
6.78917050e-01 -1.90398216e+00 -5.76802015e-01 6.19252205e-01
1.62600633e-02 3.95942181e-01 2.54506469e-01 3.27927917e-01
-2.74210900e-01 1.02247819e-01 1.16754603e-02 -5.46702385e-01
-9.15243208e-01 8.32512915e-01 5.05601585e-01 1.22091740e-01
-6.84772283e-02 8.26356292e-01 3.08577776e-01 -3.48688692e-01
4.98841375e-01 -5.48111260e-01 -3.95798236e-02 -8.00659321e-03
7.34400749e-01 4.44249123e-01 2.52818409e-02 -4.69235808e-01
-1.70503199e-01 1.68376073e-01 -3.05821210e-01 9.45650339e-02
6.37872934e-01 -1.58225626e-01 -3.73429023e-02 1.92406997e-01
1.57364929e+00 -3.37473415e-02 -1.31315684e+00 -1.72132954e-01
-6.35225654e-01 -7.28381336e-01 -4.85077910e-02 -3.56744617e-01
-1.39385927e+00 1.09640753e+00 7.03777015e-01 3.43758047e-01
1.28647017e+00 -5.73491454e-01 9.06803668e-01 2.46464089e-02
2.17540339e-01 -9.94694471e-01 -1.19710766e-01 2.60861993e-01
7.02158272e-01 -1.20772457e+00 -2.55772740e-01 -2.10370436e-01
-4.80257869e-01 6.90913975e-01 1.02556610e+00 -3.25173512e-02
5.06007016e-01 1.18673168e-01 -2.84198858e-02 3.48422199e-01
-4.54674006e-01 3.14075738e-01 -3.98309017e-03 7.21899092e-01
-2.18359549e-02 -9.60166380e-02 -7.49431476e-02 9.41907704e-01
-4.16127801e-01 -3.50005150e-01 3.94202143e-01 6.73246980e-01
-1.31851763e-01 -9.26340997e-01 -4.72101241e-01 5.08506536e-01
-6.35365844e-01 -3.59018654e-01 -7.40933865e-02 5.12544692e-01
5.25641024e-01 8.76770973e-01 2.43209362e-01 -6.30123854e-01
4.00336355e-01 -1.10999897e-01 -1.30498305e-01 -3.07040274e-01
-5.76586723e-01 1.88862171e-03 -5.79139411e-01 -4.02006149e-01
5.95871657e-02 -3.79363924e-01 -1.33767581e+00 -5.49611092e-01
-5.34119785e-01 3.45528156e-01 1.96731880e-01 5.09115040e-01
4.06429499e-01 7.53407836e-01 9.29788709e-01 -4.74741340e-01
-9.66307640e-01 -9.23331857e-01 -7.17723727e-01 4.61734027e-01
5.08887410e-01 -3.40994656e-01 -6.79449737e-01 -5.45384549e-02]
|
[8.791245460510254, 3.0163979530334473]
|
264da34e-314c-4930-8b1e-2c3ab0e74371
|
hand-gesture-recognition-through-reflected
|
2301.05955
| null |
https://arxiv.org/abs/2301.05955v2
|
https://arxiv.org/pdf/2301.05955v2.pdf
|
Hand Gesture Recognition through Reflected Infrared Light Wave Signals
|
In this study, we present a wireless (non-contact) gesture recognition method using only incoherent light wave signals reflected from a human subject. In comparison to existing radar, light shadow, sound and camera-based sensing systems, this technology uses a low-cost ubiquitous light source (e.g., infrared LED) to send light towards the subject's hand performing gestures and the reflected light is collected by a light sensor (e.g., photodetector). This light wave sensing system recognizes different gestures from the variations of the received light intensity within a 20-35cm range. The hand gesture recognition results demonstrate up to 96% accuracy on average. The developed system can be utilized in numerous Human-computer Interaction (HCI) applications as a low-cost and non-contact gesture recognition technology.
|
['Li Yu', 'Md Zobaer Islam', 'Sabit Ekin', 'Christopher Crick', "John F. O'Hara", 'Hisham Abuella']
|
2023-01-14
| null | null | null | null |
['hand-gesture-recognition', 'hand-gesture-recognition-1', 'gesture-recognition']
|
['computer-vision', 'computer-vision', 'computer-vision']
|
[ 8.68340969e-01 -6.15928471e-01 2.77281851e-01 -1.46390989e-01
-1.58895981e-02 -5.93235314e-01 3.75315487e-01 -8.49599600e-01
-7.81467080e-01 6.64892972e-01 -1.64907381e-01 -1.13291509e-01
2.43726850e-01 -9.78846908e-01 5.01293465e-02 -1.10347319e+00
5.43826640e-01 -7.51766264e-02 3.03511053e-01 1.29413769e-01
4.46522325e-01 6.15347147e-01 -1.43788016e+00 -1.95580676e-01
1.76218301e-01 9.45811093e-01 2.29042873e-01 1.03926456e+00
-3.65296658e-03 2.62363092e-03 -7.51085401e-01 2.28020430e-01
3.46249849e-01 -3.49575877e-01 7.53159374e-02 -5.28832734e-01
3.40450495e-01 -7.49761462e-01 -1.31560847e-01 7.44167447e-01
7.65267789e-01 -6.14747312e-03 6.16013706e-01 -7.99003005e-01
-3.65475535e-01 -5.33906072e-02 -8.03430796e-01 -6.72311988e-03
1.19315159e+00 4.38693672e-01 2.00078979e-01 -9.25162315e-01
5.14616191e-01 7.76644766e-01 4.65890944e-01 9.84289110e-01
-5.60520113e-01 -1.13763332e+00 -6.84176326e-01 1.76170096e-01
-1.24723232e+00 -3.68965834e-01 9.69553351e-01 -8.20339397e-02
9.14987087e-01 6.92794323e-01 6.28084540e-01 1.12029111e+00
5.82799137e-01 -3.58062945e-02 1.55990767e+00 -8.74136984e-01
2.37759054e-01 -1.16885856e-01 3.96147460e-01 8.11286449e-01
6.05669975e-01 4.93148655e-01 -7.94981062e-01 -1.26212120e-01
8.55936170e-01 3.34660053e-01 -4.44271535e-01 3.58311146e-01
-1.01955450e+00 -1.10199526e-01 3.54820400e-01 6.66484237e-01
-6.21901035e-01 4.06963050e-01 -5.14800131e-01 2.16077194e-01
4.38245125e-02 -5.23506552e-02 1.65469334e-01 -6.05807662e-01
-4.29823369e-01 -5.05779505e-01 1.01983988e+00 5.99149048e-01
4.39720303e-01 3.65210511e-02 -4.51285057e-02 4.11019683e-01
1.03510988e+00 1.43562102e+00 -5.73375486e-02 -2.44948924e-01
4.41489756e-01 2.46116042e-01 4.92403358e-01 -7.53398120e-01
-5.09314001e-01 2.67674804e-01 -4.56656814e-01 8.46779764e-01
4.88538653e-01 -7.15093911e-01 -9.91369903e-01 9.66589928e-01
3.43996346e-01 7.07222939e-01 1.49746194e-01 1.34372067e+00
1.12819719e+00 4.34266061e-01 -6.64388612e-02 -4.22510713e-01
1.52964854e+00 3.68547626e-02 -7.70726204e-01 8.90701786e-02
-2.49017596e-01 -1.18827689e+00 9.68075693e-01 5.02532184e-01
-6.34414673e-01 -1.56574205e-01 -9.49480772e-01 4.30341899e-01
3.38391773e-03 1.10087022e-01 5.29222667e-01 1.38290381e+00
-3.90335828e-01 -4.01498258e-01 -9.12242949e-01 -7.00763106e-01
9.48607773e-02 3.64122093e-01 1.98232248e-01 -2.71949619e-01
-5.69771349e-01 5.24534643e-01 -5.89123726e-01 4.24348712e-01
-1.83018446e-01 -3.68584365e-01 -6.70764074e-02 -3.04467469e-01
-4.71922494e-02 -4.64629680e-01 8.44287634e-01 -4.59041893e-01
-2.43191338e+00 5.74180186e-01 -3.92787129e-01 3.99245679e-01
9.27528217e-02 -4.12244707e-01 -4.78925973e-01 4.02644783e-01
-5.72123826e-01 -3.61600667e-01 8.88694286e-01 -1.29123509e+00
-3.62441778e-01 -6.64096236e-01 -3.25359613e-01 3.32559377e-01
1.57072488e-02 4.71765816e-01 1.44110262e-01 -9.38712135e-02
5.90909958e-01 -9.97577786e-01 3.25749338e-01 -4.96451594e-02
-2.81378686e-01 -3.14747095e-01 1.31299925e+00 -1.02297574e-01
7.09919930e-01 -1.82168329e+00 -6.65024817e-01 6.33833945e-01
-2.22813457e-01 5.29026449e-01 1.04649216e-01 6.16389453e-01
7.35543609e-01 -4.57571894e-01 -1.14346370e-01 2.48082414e-01
-3.15686494e-01 4.03275304e-02 -1.36491701e-01 6.62084281e-01
-5.91774642e-01 6.63229108e-01 -8.99324715e-01 -9.35886726e-02
4.83638108e-01 9.18855250e-01 3.92222971e-01 2.95518190e-01
5.35317659e-01 7.31114388e-01 -6.87847137e-01 1.20776200e+00
8.43689263e-01 6.16808414e-01 -1.33765623e-01 -9.07102749e-02
-3.22068006e-01 -1.08182013e-01 -1.31809747e+00 1.27187932e+00
-6.74825132e-01 8.27600181e-01 2.81038731e-01 -1.22334383e-01
1.42182624e+00 5.89474440e-01 4.35888737e-01 -5.75078607e-01
1.45282671e-01 1.87235162e-01 -3.42161864e-01 -1.20727289e+00
3.53983901e-02 -5.82015693e-01 3.71956944e-01 1.12349510e+00
-7.27819502e-01 1.25107795e-01 -6.27165854e-01 -5.06089807e-01
1.49640024e+00 1.72550082e-01 1.30210057e-01 3.59501600e-01
2.64951110e-01 -1.86450362e-01 6.80917501e-02 7.70719409e-01
-1.16912544e-01 2.05913812e-01 -9.34469819e-01 -3.05540085e-01
9.64009091e-02 -1.24835348e+00 -1.31844297e-01 9.93161619e-01
7.34567702e-01 5.56726813e-01 -3.48124802e-01 -7.02339485e-02
-1.37135357e-01 4.24998671e-01 2.36545563e-01 3.36534977e-01
-6.87905133e-01 -5.75547814e-01 7.06789136e-01 2.49974370e-01
8.25175226e-01 -1.39373958e+00 -1.68130791e+00 2.60706246e-01
1.46696508e-01 -9.83905911e-01 1.71120420e-01 -1.91934258e-01
-6.71373546e-01 -1.07141805e+00 -6.77227974e-01 -7.19914436e-01
5.42035401e-01 7.16641068e-01 6.62654698e-01 3.97473648e-02
-8.59430730e-01 1.22156918e+00 -5.64387918e-01 -9.04693782e-01
3.98030072e-01 -5.98821700e-01 5.67108542e-02 1.79456905e-01
1.32559538e+00 -6.29143715e-01 -8.84451330e-01 1.84201404e-01
-3.88146669e-01 -3.91326278e-01 6.09993100e-01 1.73117429e-01
-3.85475680e-02 -3.94709200e-01 -2.10239366e-02 -5.15817702e-01
9.15007055e-01 7.66451051e-03 -4.62449908e-01 2.43303373e-01
-5.24594784e-01 -5.16333878e-01 -5.79186641e-02 -5.80308557e-01
-1.68426037e+00 2.85226911e-01 1.67007416e-01 3.52337331e-01
-8.83533895e-01 1.19540632e-01 8.91720504e-02 -5.90975821e-01
8.20907474e-01 1.68428421e-01 -2.79810518e-01 -3.93800467e-01
8.89375061e-02 1.31972313e+00 6.06376469e-01 -5.28169274e-02
1.05900753e+00 6.30729973e-01 4.23183054e-01 -1.35725653e+00
3.20414484e-01 -6.59808278e-01 -5.78551233e-01 -8.88056755e-01
8.35526705e-01 -5.30206144e-01 -1.30536532e+00 8.18519652e-01
-1.32077432e+00 -1.14960492e-01 3.57731104e-01 1.13339090e+00
2.33589783e-01 1.05394304e-01 -5.61288655e-01 -1.74679220e+00
-5.82647800e-01 -2.24563971e-01 9.90299702e-01 7.37652600e-01
-9.49653089e-02 -4.57129300e-01 3.35082084e-01 2.86211848e-01
4.81771410e-01 3.72649014e-01 2.39368021e-01 5.46586156e-01
-7.82293618e-01 -6.65353894e-01 -1.62381068e-01 -6.40718102e-01
6.49681032e-01 -1.29517019e-01 -1.30225992e+00 1.33484706e-01
-1.53421015e-02 6.19247817e-02 6.90203011e-01 3.83747488e-01
1.50227785e-01 -3.76264052e-03 -6.79796219e-01 4.56199229e-01
1.85141575e+00 7.20049024e-01 8.99921894e-01 -1.40717655e-01
7.91967511e-01 1.98909342e-01 3.68382126e-01 4.96597022e-01
-2.05679342e-01 7.14636445e-01 1.36953473e-01 -3.44039686e-02
-3.16965312e-01 2.01542720e-01 3.30105394e-01 1.64902821e-01
-1.07468331e+00 -4.60587680e-01 -9.28895056e-01 -3.59028876e-01
-1.10498989e+00 -1.35291946e+00 -8.98216665e-01 2.21904469e+00
6.57851279e-01 -4.62942392e-01 -1.89324886e-01 3.18409622e-01
5.75782120e-01 -1.11899577e-01 -3.96740109e-01 -4.26105797e-01
4.11916494e-01 1.01421869e+00 6.60566390e-01 7.97277927e-01
-3.73793185e-01 6.85115993e-01 5.81373692e+00 -1.39743194e-01
-1.57435906e+00 1.82724133e-01 -7.26686060e-01 -2.32198045e-01
-2.51816124e-01 -4.27658260e-01 -5.59469044e-01 4.43110794e-01
2.72485405e-01 5.14230609e-01 4.54096079e-01 2.52583712e-01
3.68210167e-01 -7.75481284e-01 -6.69536293e-01 1.33586991e+00
1.49083927e-01 -4.76194233e-01 -4.89458829e-01 3.45236138e-02
9.62854251e-02 -1.90478727e-01 -1.03962474e-01 -4.02043790e-01
-2.76362747e-02 -6.88684821e-01 -1.54626608e-01 8.17828357e-01
1.30638850e+00 -2.33636588e-01 5.11772513e-01 2.35359743e-01
-1.16317952e+00 1.28879234e-01 -1.70140818e-01 -7.85987735e-01
3.08451623e-01 6.18242383e-01 -9.18064952e-01 -2.42885947e-02
6.80068970e-01 5.24400081e-03 3.21450442e-01 7.58426726e-01
-5.25873542e-01 1.07665443e+00 -7.03756571e-01 -8.47729981e-01
-1.57317847e-01 -4.99207824e-01 8.05742621e-01 1.34468973e+00
4.96616811e-01 1.13078213e+00 -3.47062647e-01 6.82456017e-01
2.77345568e-01 -2.64402330e-01 -8.65690291e-01 5.87425649e-01
5.17331541e-01 1.20504487e+00 -8.84326875e-01 8.00500624e-03
-4.23694730e-01 1.01631808e+00 -9.30242658e-01 7.79719293e-01
-2.28761360e-01 -1.10260034e+00 3.56393605e-01 -2.51233224e-02
-1.90883189e-01 -6.12282455e-01 -8.10050368e-01 -6.20203912e-01
3.17725480e-01 -7.12798443e-03 -2.39256144e-01 -9.50452447e-01
-9.27829385e-01 2.51491934e-01 -3.23535264e-01 -1.25465071e+00
-1.37451738e-02 -6.21474922e-01 -8.88666570e-01 1.24757397e+00
-1.39572215e+00 -8.78480554e-01 -1.14752841e+00 8.37903023e-01
8.18769559e-02 -9.96025652e-02 9.95762527e-01 -8.56787786e-02
-7.32507110e-02 2.95981824e-01 -7.89292976e-02 1.76751032e-01
6.77481890e-01 -7.47344494e-01 -2.93116033e-01 8.53606343e-01
1.73880547e-01 6.38879657e-01 5.18209517e-01 -8.01941097e-01
-2.11487126e+00 -3.38325232e-01 8.05617571e-01 -4.39437628e-01
-1.18859746e-02 -1.93172708e-01 -2.62600064e-01 1.41604125e-01
2.45087728e-01 1.36291506e-02 8.36407483e-01 -6.13381863e-02
-2.37634152e-01 -5.45238495e-01 -1.68970573e+00 3.53435814e-01
1.22483397e+00 -4.71749485e-01 -6.39200628e-01 -6.94382284e-03
-3.48293513e-01 -2.95995563e-01 -3.30560803e-01 4.32376526e-02
1.54389727e+00 -8.68696213e-01 8.55441689e-01 2.67266244e-01
-2.78772622e-01 -1.22910939e-01 -6.86641708e-02 -8.41748834e-01
-1.38076708e-01 -7.51323760e-01 3.17049682e-01 9.90006864e-01
-4.68241796e-02 -1.16317105e+00 9.45089459e-01 8.58348012e-01
3.40163469e-01 -1.88813463e-01 -1.16436565e+00 -5.81967652e-01
-8.79278123e-01 -5.40918887e-01 4.14339989e-01 3.63995731e-01
4.57580507e-01 3.43747944e-01 -1.92770422e-01 4.59892869e-01
9.75195765e-01 2.20452651e-01 5.55988729e-01 -1.49101007e+00
1.01695806e-01 5.75764030e-02 -3.45282078e-01 -1.01252282e+00
-3.39027047e-01 -4.46758151e-01 5.14298737e-01 -1.74754798e+00
-6.38472512e-02 -5.61277449e-01 -2.51395941e-01 3.97211075e-01
9.04481933e-02 6.42036855e-01 9.80588943e-02 3.60444307e-01
7.82245398e-02 -2.33606949e-01 1.11543941e+00 1.49259239e-01
-6.56633556e-01 4.98966753e-01 -7.11776549e-03 5.55517614e-01
7.07430422e-01 -5.37815213e-01 -6.11610599e-02 -4.75532204e-01
1.26307905e-01 7.58564621e-02 7.09753990e-01 -1.20546877e+00
5.39632440e-01 -5.14194608e-01 4.96468961e-01 -2.31253162e-01
4.28944439e-01 -1.45084703e+00 1.37525529e-01 8.79795671e-01
2.16262594e-01 -4.59150791e-01 -4.48994428e-01 6.12559855e-01
4.98611450e-01 -2.25450620e-01 3.60481709e-01 -2.21140489e-01
-5.67502022e-01 -3.49283874e-01 -7.33582437e-01 -8.19683850e-01
1.01413298e+00 -9.23363805e-01 -7.99029708e-01 -2.90691435e-01
-1.33348564e-02 -6.58368349e-01 9.52745229e-02 2.22232714e-02
1.02568340e+00 -1.14297414e+00 -3.22682321e-01 5.38369536e-01
-9.77881402e-02 -4.92956281e-01 -2.97299266e-01 5.43083310e-01
-5.65438449e-01 4.57350731e-01 -3.26198608e-01 -8.05197358e-01
-1.98859441e+00 -5.37341118e-01 -7.22502731e-03 6.31851614e-01
-7.55535543e-01 8.59874785e-01 -8.08337927e-01 4.08237517e-01
2.18098626e-01 -3.00589055e-01 -1.57423094e-01 -4.45917279e-01
7.03827620e-01 8.76421809e-01 -2.24625662e-01 -3.44217330e-01
-8.32810998e-01 1.50801861e+00 9.37103093e-01 -7.20057249e-01
1.20226049e+00 -6.67404607e-02 2.29280621e-01 5.59980273e-01
6.14936411e-01 3.50432038e-01 -9.91838098e-01 -1.38857529e-01
-3.78122479e-01 -7.67450213e-01 6.16707616e-02 -1.23506320e+00
-7.03987002e-01 7.59452403e-01 1.31540871e+00 3.01585253e-02
1.34399116e+00 -1.97910175e-01 8.11454296e-01 8.74001026e-01
1.07178760e+00 -1.05524325e+00 -6.40942808e-03 6.15064353e-02
5.92534363e-01 -9.03872192e-01 8.41936469e-02 -4.65137511e-01
-1.63624853e-01 1.43296444e+00 3.15360934e-01 -9.87033471e-02
8.56614828e-01 1.01061082e+00 7.91858256e-01 -2.73696333e-01
-9.02281925e-02 -4.12862003e-01 1.12971075e-01 1.18836403e+00
5.95392346e-01 3.94008934e-01 -6.82769418e-01 6.67093545e-02
5.15764654e-02 7.01917350e-01 4.55638319e-01 1.23092306e+00
-8.45153451e-01 -8.29532921e-01 -9.28927839e-01 3.80570769e-01
-3.78118813e-01 2.35614344e-01 -4.98578936e-01 2.96897799e-01
1.75520718e-01 1.60232604e+00 -1.16655819e-01 -5.76632559e-01
4.40247327e-01 2.09514961e-01 9.45481002e-01 -5.64112127e-01
-4.71300036e-01 1.32691577e-01 -2.81737447e-01 -5.54993987e-01
-9.74363685e-01 -4.40890044e-01 -1.55789864e+00 -3.06966286e-02
-2.10868612e-01 -4.75713581e-01 1.26655769e+00 1.01810753e+00
8.67229886e-03 2.21410912e-04 5.56765616e-01 -8.93040895e-01
-1.12411685e-01 -1.20833230e+00 -8.54433239e-01 1.51086688e-01
3.05255145e-01 -6.12442374e-01 -3.33773494e-01 4.15411554e-02]
|
[6.561287879943848, 0.1249130591750145]
|
5f0490f9-1166-4cf5-8e2e-b654db8bfb78
|
a-joint-intensity-and-depth-co-sparse
|
1304.5319
| null |
http://arxiv.org/abs/1304.5319v1
|
http://arxiv.org/pdf/1304.5319v1.pdf
|
A Joint Intensity and Depth Co-Sparse Analysis Model for Depth Map Super-Resolution
|
High-resolution depth maps can be inferred from low-resolution depth
measurements and an additional high-resolution intensity image of the same
scene. To that end, we introduce a bimodal co-sparse analysis model, which is
able to capture the interdependency of registered intensity and depth
information. This model is based on the assumption that the co-supports of
corresponding bimodal image structures are aligned when computed by a suitable
pair of analysis operators. No analytic form of such operators exist and we
propose a method for learning them from a set of registered training signals.
This learning process is done offline and returns a bimodal analysis operator
that is universally applicable to natural scenes. We use this to exploit the
bimodal co-sparse analysis model as a prior for solving inverse problems, which
leads to an efficient algorithm for depth map super-resolution.
|
['Simon Hawe', 'Martin Kiechle', 'Martin Kleinsteuber']
|
2013-04-19
| null | null | null | null |
['depth-map-super-resolution']
|
['computer-vision']
|
[ 7.18359411e-01 1.24187909e-01 1.12751098e-02 -4.49854702e-01
-7.46964395e-01 -1.91940904e-01 6.49132609e-01 -2.79606832e-03
-1.90715045e-01 5.89312315e-01 1.84094459e-01 3.41763020e-01
-4.84729767e-01 -1.06404638e+00 -4.56291109e-01 -7.63411641e-01
4.31276746e-02 6.54003680e-01 3.35486472e-01 -3.01457345e-01
2.03778893e-01 6.92374289e-01 -1.89913714e+00 5.42129695e-01
8.09944034e-01 8.83879900e-01 5.66748381e-01 5.91265023e-01
-1.13197409e-01 7.99179792e-01 1.44261077e-01 1.05574593e-01
2.48120934e-01 -6.47159100e-01 -7.58885622e-01 1.64502054e-01
3.57495666e-01 -4.07128602e-01 -2.50772327e-01 1.09654903e+00
2.39765584e-01 7.42702512e-03 8.19763362e-01 -6.70524180e-01
8.43082666e-02 2.53084242e-01 -6.71445727e-01 2.45733380e-01
5.98724484e-01 -5.37998497e-01 9.09652710e-01 -9.01679039e-01
7.37955928e-01 9.45974946e-01 5.95978081e-01 2.27842614e-01
-1.45617843e+00 -1.58202022e-01 -3.02529067e-01 1.35476455e-01
-1.29047465e+00 -4.61741924e-01 1.01062405e+00 -5.09381473e-01
4.70405221e-01 1.84383824e-01 6.77522719e-01 7.14440405e-01
8.23991820e-02 3.53210926e-01 1.37088943e+00 -7.61954963e-01
3.14183265e-01 -9.34244692e-03 -3.37196551e-02 6.66789889e-01
8.09560716e-03 1.75020576e-01 -8.88690829e-01 -3.28790396e-02
1.16910231e+00 -8.42287019e-02 -3.75647962e-01 -6.39658391e-01
-1.06987715e+00 6.85447216e-01 2.84339786e-01 5.29569387e-01
-6.20318711e-01 -2.06181556e-01 -1.28205474e-02 1.01967931e-01
5.34257591e-01 2.17647091e-01 -7.15624914e-02 2.28704572e-01
-1.01269734e+00 4.70644683e-02 4.71312255e-01 4.28443730e-01
1.44191051e+00 -2.96220869e-01 2.21757755e-01 5.63074827e-01
2.02710450e-01 5.53718388e-01 2.60227323e-01 -1.21278143e+00
6.23466671e-02 3.76779318e-01 2.72558928e-02 -9.30689394e-01
-4.71423268e-01 -2.88413912e-01 -8.61919403e-01 4.03285891e-01
4.77405310e-01 4.14075524e-01 -5.85398316e-01 1.61556375e+00
3.71367693e-01 4.43423003e-01 1.84828967e-01 8.18714619e-01
6.03160381e-01 4.14698213e-01 -5.45569181e-01 -4.41584796e-01
1.11637318e+00 -1.07917845e-01 -8.21325004e-01 -8.21603015e-02
1.23465098e-01 -4.44746822e-01 8.91399264e-01 6.08212352e-01
-1.30578637e+00 -4.77594882e-01 -1.11254716e+00 -2.46067226e-01
1.97946979e-03 -2.00366750e-02 4.24120456e-01 2.64810205e-01
-9.44126964e-01 6.71074927e-01 -8.59971583e-01 -3.98378260e-02
4.01051268e-02 2.45174751e-01 -5.13658166e-01 -2.63448477e-01
-9.10809219e-01 7.62596667e-01 2.34652445e-01 3.87181222e-01
-6.70279264e-01 -6.30289674e-01 -9.60373759e-01 -1.79293171e-01
-3.44263911e-02 -6.60582364e-01 8.14195573e-01 -1.06506050e+00
-1.52779770e+00 1.21947169e+00 -5.00070453e-01 -1.40014336e-01
3.23411196e-01 -3.67378024e-03 -1.02829792e-01 7.94451058e-01
2.67909884e-01 1.80757195e-01 1.15130889e+00 -1.60753989e+00
-4.65884060e-01 -7.52375424e-01 -1.40436932e-01 2.23436639e-01
-6.36266023e-02 -2.49912024e-01 -2.69638151e-01 -2.13856578e-01
8.39855671e-01 -4.21211988e-01 -2.16230258e-01 -2.32758552e-01
-1.36771336e-01 4.43577737e-01 5.30871391e-01 -5.75332999e-01
7.52376258e-01 -2.16454887e+00 7.38357663e-01 6.12585783e-01
2.69104600e-01 -5.49288571e-01 1.82995275e-02 2.35722497e-01
-1.01385608e-01 -4.87737328e-01 -6.60224378e-01 -4.89610583e-01
-4.88093078e-01 3.97169530e-01 -3.49135458e-01 7.80560911e-01
5.66714518e-02 5.32367527e-01 -9.43338335e-01 -4.72281665e-01
6.28396928e-01 7.77708948e-01 -4.54532206e-01 4.57031488e-01
-2.40289196e-02 1.05426717e+00 -3.64659458e-01 3.37661386e-01
8.91894817e-01 -1.72468588e-01 1.94740102e-01 -5.22094548e-01
-4.51237947e-01 1.04637168e-01 -1.35355437e+00 2.02046871e+00
-5.32592058e-01 3.49815637e-01 4.64422464e-01 -1.17130578e+00
9.98446047e-01 2.59974301e-01 8.55799019e-01 -9.03177202e-01
1.46706328e-01 4.60258961e-01 -5.12319505e-01 -5.64714611e-01
2.85344690e-01 -5.72902203e-01 2.31862247e-01 3.06882650e-01
7.49837011e-02 -4.35922682e-01 -5.43794222e-02 4.31512557e-02
9.03990865e-01 9.70351771e-02 2.34376356e-01 -3.45486343e-01
8.39367390e-01 -4.08439994e-01 2.34201565e-01 5.32680273e-01
3.99839401e-01 8.26342523e-01 2.74600655e-01 -4.37535971e-01
-1.01303756e+00 -1.24305642e+00 -5.84814668e-01 7.43671536e-01
1.91485822e-01 -1.17767997e-01 -6.21527672e-01 -7.24950731e-02
-2.63334453e-01 1.73573047e-01 -8.39926362e-01 1.41877174e-01
-4.96699631e-01 -5.50096393e-01 -4.78358120e-02 2.10139200e-01
4.22242969e-01 -7.95241296e-01 -8.63538980e-01 -1.95402671e-02
-4.88806844e-01 -1.36277592e+00 2.00261071e-01 5.46424866e-01
-1.07860339e+00 -1.04307425e+00 -6.95910215e-01 -5.23143888e-01
6.21610820e-01 1.80547297e-01 1.17346931e+00 -1.77395478e-01
-9.29998904e-02 6.61942661e-01 -2.24642053e-01 1.00682385e-01
-3.12856168e-01 -2.01226905e-01 -1.05635785e-01 6.00506961e-01
-9.46693271e-02 -1.11912739e+00 -3.06033522e-01 2.07676843e-01
-1.03291464e+00 1.67564541e-01 5.23374498e-01 7.21479714e-01
1.09102619e+00 1.31851118e-02 2.13490739e-01 -7.67722905e-01
9.40437764e-02 -3.03130001e-01 -8.05987120e-01 6.92965686e-02
-2.54135638e-01 4.17656422e-01 4.79254693e-01 1.23498850e-02
-1.28588867e+00 5.67988038e-01 -2.79963523e-01 -4.71890509e-01
-1.12907290e-01 5.41026413e-01 -3.78581405e-01 -3.34240645e-01
6.71917081e-01 2.72333562e-01 1.14511535e-01 -5.41992903e-01
3.65506381e-01 2.65099227e-01 1.02388275e+00 -6.62984133e-01
6.45628214e-01 1.15047097e+00 5.57382882e-01 -1.25907862e+00
-1.02977979e+00 -7.21365571e-01 -9.75692093e-01 -2.48480126e-01
9.00057554e-01 -8.68774593e-01 -5.49479723e-01 3.64669472e-01
-1.03779101e+00 -3.56241643e-01 -5.10189056e-01 6.23717964e-01
-8.95718098e-01 6.23625875e-01 -4.52562749e-01 -8.84372413e-01
2.35200047e-01 -8.74223769e-01 1.29182005e+00 -4.33869064e-02
2.25097150e-01 -1.01382113e+00 3.94609392e-01 2.24280804e-01
1.48797914e-01 3.27255040e-01 6.02677703e-01 1.56565800e-01
-7.40928113e-01 8.40332508e-02 -2.58438885e-01 3.28452587e-01
1.22459661e-02 -2.32655227e-01 -1.27432609e+00 2.32491689e-03
6.13894105e-01 -1.31860018e-01 9.43141103e-01 7.98166990e-01
9.71220613e-01 1.50368884e-01 -1.86184481e-01 1.07668662e+00
1.67096603e+00 -3.15110624e-01 9.37310755e-01 1.69826195e-01
6.54317379e-01 8.85136604e-01 3.71203363e-01 6.31744504e-01
2.04845801e-01 1.00137687e+00 6.03886127e-01 -1.03830270e-01
3.84164155e-02 -2.91116297e-01 1.32849133e-02 5.36583722e-01
-4.87325072e-01 4.23340678e-01 -7.80100822e-01 4.72082943e-01
-1.64552677e+00 -1.03370464e+00 -4.66168016e-01 2.49890399e+00
6.09035194e-01 -4.00610417e-02 4.14286442e-02 3.92160654e-01
4.94373322e-01 5.74674606e-02 -2.99741626e-01 9.90468115e-02
-5.19080639e-01 6.59384549e-01 4.19790417e-01 1.03078294e+00
-7.78098226e-01 4.48719859e-01 7.15214634e+00 5.37554979e-01
-9.91322756e-01 8.78619626e-02 2.16956258e-01 4.50089991e-01
-7.62189329e-01 7.16154650e-02 -4.99370754e-01 1.59273297e-01
7.33119249e-01 -8.29927549e-02 5.01630068e-01 4.62537497e-01
1.88907772e-01 -4.69046652e-01 -1.10864377e+00 1.21944499e+00
1.56152487e-01 -1.21157265e+00 -1.69025704e-01 1.32499471e-01
6.46316886e-01 -1.11189909e-01 -5.21063171e-02 -4.41427559e-01
-1.11507498e-01 -1.02529180e+00 5.64105988e-01 8.35645378e-01
8.25955629e-01 -5.99620521e-01 5.00386894e-01 3.86651844e-01
-1.33277404e+00 -4.25943658e-02 -2.42232814e-01 -2.48002544e-01
4.82014537e-01 9.90391791e-01 -2.69332737e-01 7.82068014e-01
5.52873135e-01 1.08907461e+00 -2.27887973e-01 5.80345035e-01
-2.60841310e-01 -1.55658415e-02 -4.64898825e-01 8.06266248e-01
-2.11861908e-01 -5.97830772e-01 3.43947470e-01 9.26706612e-01
4.17008519e-01 2.76408821e-01 -9.64693725e-02 9.27113235e-01
4.76697177e-01 -4.34241183e-02 -7.92967081e-01 4.90923434e-01
3.51621099e-02 1.25936890e+00 -8.45141351e-01 -1.27162218e-01
-3.49052161e-01 1.04640996e+00 2.42623538e-01 4.00132775e-01
-4.56120431e-01 1.90232098e-01 3.39234293e-01 3.88512909e-01
2.36970991e-01 -3.35346282e-01 -5.11851251e-01 -1.29485428e+00
4.61823978e-02 -6.16343737e-01 2.94839025e-01 -8.63496900e-01
-1.21744728e+00 5.68123519e-01 1.77931830e-01 -1.19191241e+00
-3.67378503e-01 -4.62775707e-01 -3.21329564e-01 9.68753159e-01
-1.78800666e+00 -9.82650101e-01 -7.52742231e-01 1.04751277e+00
-3.02536916e-02 3.33384424e-01 8.38989079e-01 2.36218125e-01
-1.43257587e-03 -8.12359825e-02 -7.77874142e-02 -2.03641549e-01
2.94862509e-01 -1.21749473e+00 -3.75784248e-01 7.75478661e-01
1.81898564e-01 2.87021965e-01 8.35680723e-01 -3.75379205e-01
-1.31945181e+00 -3.49592060e-01 6.94988251e-01 -4.55634713e-01
3.19809467e-01 -1.52431354e-01 -1.10136223e+00 5.73986173e-01
-1.78332075e-01 3.01527441e-01 4.29981232e-01 -2.91628074e-02
-2.27922514e-01 -2.18961746e-01 -1.06247103e+00 -1.82747096e-01
8.86480212e-01 -1.03322184e+00 -5.52688718e-01 7.52263069e-02
1.33164674e-01 -5.03527284e-01 -8.98696721e-01 5.26166379e-01
4.87850785e-01 -1.61837566e+00 1.31111038e+00 4.19319421e-02
6.20948076e-01 -4.39610630e-01 -5.05443931e-01 -8.75829399e-01
-1.56946793e-01 -3.23046893e-01 -1.40298620e-01 6.83324754e-01
4.25634608e-02 -5.98317206e-01 6.39033437e-01 2.49203712e-01
-1.33690342e-01 -4.19237733e-01 -1.06395316e+00 -5.08732557e-01
-2.31283262e-01 -4.80992347e-01 7.62851015e-02 9.62585151e-01
-7.49257058e-02 3.43607634e-01 -3.92636359e-01 4.67379957e-01
1.14599848e+00 3.23588103e-01 5.14832735e-01 -1.40566361e+00
-5.15013814e-01 -2.68351585e-01 -4.96601284e-01 -1.19588006e+00
3.96192551e-01 -7.40622818e-01 1.13484524e-01 -1.47163260e+00
1.21497154e-01 -5.27304292e-01 -5.91272078e-02 3.23952474e-02
2.91430533e-01 4.68650907e-01 -3.00544888e-01 3.63024861e-01
-2.35798806e-01 5.75596929e-01 1.02953148e+00 2.75177866e-01
-2.82898456e-01 -4.74060960e-02 -2.12138548e-01 8.15234184e-01
2.53500581e-01 -2.15221360e-01 -3.23403478e-01 -2.40690276e-01
6.11113727e-01 3.68546963e-01 4.86965090e-01 -1.04981840e+00
1.64841816e-01 -8.92186910e-02 2.63525695e-01 -4.91282254e-01
6.30032599e-01 -9.58493114e-01 3.73281181e-01 1.00532018e-01
-2.25534856e-01 -3.43104720e-01 -1.35403633e-01 5.01217723e-01
-5.14236867e-01 -3.29035819e-01 9.47173536e-01 -2.36111894e-01
-6.99321032e-01 1.63135722e-01 -1.03459276e-01 -9.84547064e-02
6.00122333e-01 -3.70007873e-01 2.10203543e-01 -4.48271394e-01
-9.36346292e-01 -2.81948000e-01 7.20350206e-01 -2.75165379e-01
7.57312596e-01 -1.28774130e+00 -6.14481986e-01 5.96055388e-01
9.19446051e-02 2.25385621e-01 4.85226691e-01 1.04799545e+00
-5.03652394e-01 -5.66736497e-02 -5.07169425e-01 -1.02035809e+00
-9.20100689e-01 4.71571863e-01 6.01949275e-01 -2.45056331e-01
-8.12143564e-01 5.04810333e-01 4.73872304e-01 -1.12678804e-01
-2.04606131e-01 -2.82692343e-01 -2.95693457e-01 1.11060953e-02
7.22734034e-01 3.23488057e-01 1.26030460e-01 -9.78989303e-01
-3.36725980e-01 1.24334788e+00 4.97844607e-01 -5.76982617e-01
1.43305254e+00 -3.74172807e-01 -4.36620384e-01 7.54243016e-01
1.20681596e+00 3.19280803e-01 -1.26405275e+00 -2.25934610e-01
-1.17221743e-01 -7.65630841e-01 3.23591143e-01 -1.22591719e-01
-9.73273218e-01 8.62970591e-01 4.46280092e-01 3.54444593e-01
1.59159791e+00 3.16182762e-01 2.29444966e-01 1.92337390e-02
4.38700050e-01 -7.63958097e-01 2.23414749e-01 4.99049544e-01
8.88477623e-01 -1.09007704e+00 1.07063398e-01 -7.00420558e-01
-6.55704737e-02 1.30950272e+00 2.35896371e-02 -2.16991246e-01
7.48499036e-01 4.58925098e-01 -2.24966139e-01 -4.18780357e-01
-2.38596141e-01 -5.70550382e-01 1.76870823e-01 7.61660993e-01
3.35070044e-01 -1.07205808e-01 -2.81458050e-01 2.19085693e-01
1.41157899e-02 1.42237574e-01 4.98411775e-01 6.69256508e-01
-4.63727862e-01 -1.07433772e+00 -4.99740839e-01 -9.87830311e-02
-5.13595156e-02 8.67484286e-02 -6.76228404e-02 3.15379173e-01
9.10752416e-02 5.29044986e-01 1.20338045e-01 -2.23170206e-01
2.11489514e-01 -1.28243625e-01 1.04330444e+00 -5.33548772e-01
1.02915786e-01 1.97209656e-01 -2.12866068e-01 -8.25641632e-01
-1.05401754e+00 -6.11182094e-01 -1.26653612e+00 1.45796798e-02
1.43546432e-01 4.46547829e-02 6.36010706e-01 1.06154370e+00
5.18836314e-03 3.39213938e-01 7.69893825e-01 -1.33107686e+00
1.65074188e-02 -5.96558809e-01 -1.07895029e+00 4.51843262e-01
5.66243827e-01 -8.22167397e-01 -6.22991979e-01 1.67675376e-01]
|
[9.528389930725098, -2.630035638809204]
|
420d5bcf-0190-4fba-b89b-f755d3e6a97f
|
ernie-unix2-a-unified-cross-lingual-cross
|
2211.04861
| null |
https://arxiv.org/abs/2211.04861v1
|
https://arxiv.org/pdf/2211.04861v1.pdf
|
ERNIE-UniX2: A Unified Cross-lingual Cross-modal Framework for Understanding and Generation
|
Recent cross-lingual cross-modal works attempt to extend Vision-Language Pre-training (VLP) models to non-English inputs and achieve impressive performance. However, these models focus only on understanding tasks utilizing encoder-only architecture. In this paper, we propose ERNIE-UniX2, a unified cross-lingual cross-modal pre-training framework for both generation and understanding tasks. ERNIE-UniX2 integrates multiple pre-training paradigms (e.g., contrastive learning and language modeling) based on encoder-decoder architecture and attempts to learn a better joint representation across languages and modalities. Furthermore, ERNIE-UniX2 can be seamlessly fine-tuned for varieties of generation and understanding downstream tasks. Pre-trained on both multilingual text-only and image-text datasets, ERNIE-UniX2 achieves SOTA results on various cross-lingual cross-modal generation and understanding tasks such as multimodal machine translation and multilingual visual question answering.
|
['Haifeng Wang', 'Hua Wu', 'Hao Tian', 'Yu Sun', 'Shuohuan Wang', 'Weichong Yin', 'Yaqian Han', 'Bin Shan']
|
2022-11-09
| null | null | null | null |
['multimodal-machine-translation']
|
['natural-language-processing']
|
[ 2.17719346e-01 8.00567269e-02 -3.09182703e-01 -4.69142467e-01
-1.65590680e+00 -7.17924654e-01 1.05004084e+00 -2.34448537e-01
-3.92221242e-01 7.00544477e-01 3.03157121e-01 -5.17314017e-01
5.87322712e-01 -6.39635801e-01 -1.02109873e+00 1.94272362e-02
5.85355222e-01 7.65921116e-01 -5.45182765e-01 -3.81031066e-01
-2.95533389e-01 -2.80959159e-01 -1.11956406e+00 9.28758204e-01
1.10123742e+00 7.09600866e-01 4.15512472e-01 8.22112858e-01
-2.87788093e-01 5.03635228e-01 -7.39186853e-02 -7.87110746e-01
5.28367497e-02 -6.57356918e-01 -1.05316627e+00 -4.95520644e-02
8.84142101e-01 -2.70716101e-01 -3.19771051e-01 6.82133794e-01
5.31033456e-01 -3.17399055e-01 7.58166015e-01 -1.24893999e+00
-1.73078287e+00 7.38195240e-01 -5.80613256e-01 -2.21484572e-01
4.72651869e-01 4.67571229e-01 1.05708671e+00 -1.28283238e+00
8.39718342e-01 1.66660035e+00 3.22391212e-01 8.39187324e-01
-1.43651795e+00 -8.12103033e-01 -4.33057845e-02 2.21531987e-01
-1.25615966e+00 -7.10073173e-01 3.57201368e-01 -3.71185392e-01
1.03037322e+00 -1.27335310e-01 4.07192744e-02 1.62570822e+00
1.49613157e-01 1.33074272e+00 1.51003087e+00 -5.48553288e-01
-4.84069914e-01 1.46242455e-01 -3.39801610e-01 7.48622775e-01
-3.95590216e-01 2.30689436e-01 -9.05290484e-01 4.96325731e-01
6.28055871e-01 -5.95297575e-01 -3.40488285e-01 -5.60967885e-02
-1.83926964e+00 9.94972527e-01 3.66797000e-01 9.69448537e-02
-1.03867399e-02 3.03414077e-01 7.00694025e-01 6.30781889e-01
3.92113417e-01 5.38792968e-01 -6.35875046e-01 -4.61476333e-02
-7.90705323e-01 -1.59534395e-01 2.64701873e-01 1.25085092e+00
9.37274873e-01 1.53347105e-01 -4.92957026e-01 1.04563522e+00
4.81123358e-01 1.00612605e+00 5.36784530e-01 -6.27939761e-01
1.01405048e+00 5.12138665e-01 -4.49408382e-01 -2.32124284e-01
4.15031686e-02 -9.65136569e-03 -1.00897980e+00 -6.63231239e-02
3.03728223e-01 -3.31777960e-01 -1.23096251e+00 2.12217283e+00
-6.52155951e-02 -3.04981083e-01 7.86322713e-01 7.09150136e-01
1.20339775e+00 9.20316041e-01 5.09640098e-01 3.29708755e-01
1.50943410e+00 -1.49295986e+00 -5.71121573e-01 -4.82876390e-01
6.81169271e-01 -1.09612131e+00 1.51896381e+00 4.15050909e-02
-1.19074011e+00 -1.11947191e+00 -6.87661529e-01 -7.64158964e-01
-7.91175544e-01 6.56342745e-01 5.53771198e-01 6.39700219e-02
-1.32075429e+00 -1.75152302e-01 -5.68238080e-01 -5.32944560e-01
7.75955617e-02 -5.27169593e-02 -7.62906134e-01 -5.97790539e-01
-1.44532228e+00 1.04316020e+00 7.07976937e-01 -1.12354457e-02
-1.20729864e+00 -7.51242638e-01 -1.40230060e+00 -3.45880479e-01
2.07924083e-01 -1.10070157e+00 1.15600574e+00 -1.13575840e+00
-1.45672214e+00 1.24414456e+00 -2.23971188e-01 -3.22613329e-01
3.02133024e-01 -3.33032012e-01 -5.83065808e-01 7.26993531e-02
3.79019797e-01 1.65579498e+00 7.33178139e-01 -1.29994977e+00
-3.74081671e-01 -3.14418912e-01 1.59302756e-01 5.58200061e-01
-5.70100099e-02 -2.22706884e-01 -8.24320197e-01 -6.48102105e-01
-4.16929036e-01 -9.18914318e-01 2.72055566e-01 -2.49826565e-01
-4.68010634e-01 -2.90922135e-01 8.24697375e-01 -1.01056373e+00
5.81642926e-01 -1.87299323e+00 4.42805290e-01 -5.36733747e-01
-1.00845218e-01 3.50623094e-02 -1.00051808e+00 7.04857409e-01
-1.40395880e-01 -2.25516707e-02 -3.26133072e-02 -6.64534867e-01
4.02425557e-01 2.81345606e-01 -4.26822990e-01 -7.83897862e-02
5.09171188e-01 1.60792422e+00 -7.10068345e-01 -4.97346014e-01
1.53313160e-01 5.04228592e-01 -4.02557760e-01 5.40880919e-01
-5.57425559e-01 5.99719942e-01 -5.49625643e-02 8.18439424e-01
5.02631068e-01 -3.01461875e-01 1.64825609e-03 -4.69715297e-01
-1.00095011e-01 9.82393324e-02 -3.88007075e-01 2.54965091e+00
-1.11613023e+00 8.62501562e-01 3.38580720e-02 -7.12358236e-01
7.31292129e-01 5.38015783e-01 -7.50552565e-02 -1.22888374e+00
-5.04564233e-02 3.66522074e-01 -3.85041118e-01 -4.69908357e-01
8.60844553e-01 -1.75449669e-01 -4.60370272e-01 5.79106569e-01
8.18199694e-01 -1.00069568e-01 3.58189493e-01 4.22046930e-01
3.44139248e-01 6.65197492e-01 1.67352065e-01 -1.68787658e-01
6.66107178e-01 -6.35222159e-03 -1.74689412e-01 4.63776946e-01
9.01822299e-02 6.56377494e-01 -1.88767798e-02 9.55593362e-02
-9.93907332e-01 -1.37553895e+00 5.65157607e-02 1.50519407e+00
1.23422436e-01 -3.97844434e-01 -6.42074883e-01 -7.26139009e-01
-3.59295197e-02 9.19145286e-01 -7.22767770e-01 -1.65414795e-01
-1.57451689e-01 -4.73199189e-01 9.39305842e-01 6.16023362e-01
6.92675471e-01 -1.14176595e+00 7.01212063e-02 9.76931229e-02
-6.79536939e-01 -1.63293517e+00 -7.02531397e-01 -7.41552263e-02
-4.27290648e-01 -7.40410209e-01 -8.95506203e-01 -1.05196702e+00
3.20148170e-01 -2.07846425e-02 1.75388300e+00 -3.67361754e-01
-1.30729198e-01 9.87291992e-01 -4.15896118e-01 -1.23791978e-01
-7.07610309e-01 3.78458142e-01 -3.41001391e-01 1.84512604e-02
3.78440022e-01 -1.37614012e-01 -3.47272664e-01 9.70101580e-02
-7.59907007e-01 6.78120673e-01 9.20683146e-01 8.65500093e-01
7.75745034e-01 -9.49854255e-01 8.43449056e-01 -4.75776494e-01
7.16256797e-01 -4.59518820e-01 -3.46462667e-01 1.00778592e+00
-3.60218793e-01 2.76257306e-01 5.50064206e-01 -5.09965897e-01
-1.20720935e+00 -2.29546443e-01 -1.18259996e-01 -5.01695096e-01
-3.58403563e-01 6.35098577e-01 -3.82842153e-01 2.78740942e-01
2.62905419e-01 6.20426059e-01 -2.69581914e-01 -4.25940990e-01
1.16840327e+00 4.93153214e-01 9.55391467e-01 -8.42562616e-01
6.97044551e-01 5.10329269e-02 -6.33480847e-01 -5.99591494e-01
-7.87676454e-01 -2.03872100e-01 -7.14874625e-01 -9.58142206e-02
1.57094228e+00 -1.52945065e+00 -4.67539459e-01 4.88189965e-01
-1.38059330e+00 -6.30907714e-01 -1.02018945e-01 4.82874513e-01
-5.89157820e-01 2.64532678e-02 -7.00311542e-01 -4.02067214e-01
-5.21429181e-01 -1.26139522e+00 1.67698443e+00 1.27143651e-01
8.13299343e-02 -1.31109893e+00 2.92145669e-01 9.81938422e-01
3.18666399e-01 -5.15551902e-02 1.10783947e+00 -3.72600287e-01
-7.17979074e-01 2.75700420e-01 -6.69330120e-01 3.66304159e-01
-1.42648771e-01 -3.17426324e-01 -8.76377523e-01 -4.26089704e-01
-7.49906719e-01 -1.31381178e+00 9.04643118e-01 -6.01501130e-02
8.42505634e-01 2.63223499e-02 -2.64550317e-02 7.82939494e-01
1.46241868e+00 -1.68727875e-01 7.30644286e-01 4.23125550e-02
9.28001225e-01 5.60671568e-01 3.27989191e-01 -2.97697037e-01
1.26794314e+00 6.42886460e-01 2.42716640e-01 -6.04824185e-01
-6.19101167e-01 -6.90879762e-01 6.86046481e-01 1.22468436e+00
2.54456669e-01 -4.58588928e-01 -8.72730196e-01 9.24936652e-01
-1.61891687e+00 -7.01701105e-01 -5.61342239e-02 1.66625452e+00
1.14965451e+00 -6.03829741e-01 -3.43369782e-01 -8.85581732e-01
4.77023333e-01 1.84217259e-01 -6.29532218e-01 -4.38791871e-01
-5.09999514e-01 1.65209368e-01 9.90082324e-02 5.23030818e-01
-1.02673829e+00 1.56402409e+00 5.95147038e+00 1.09931314e+00
-1.21353972e+00 5.13475001e-01 6.68281019e-01 1.85117424e-01
-6.37728751e-01 -3.66615579e-02 -8.07940364e-01 9.82550606e-02
9.12901938e-01 -1.18627008e-02 4.21168536e-01 4.22845483e-01
-2.77345568e-01 -3.79861966e-02 -1.34348106e+00 1.26706851e+00
5.27468920e-01 -1.28041303e+00 4.54254746e-01 -6.63663149e-02
1.00375712e+00 4.70903993e-01 3.74158800e-01 1.01654983e+00
5.00535190e-01 -1.42778635e+00 7.10390985e-01 3.97389472e-01
1.48037100e+00 -4.85412449e-01 4.23602790e-01 9.39782783e-02
-1.28309643e+00 4.68873739e-01 2.31583379e-02 3.78034979e-01
6.94558620e-01 -3.66416723e-02 -6.45186186e-01 9.51769412e-01
4.25296605e-01 7.74495542e-01 -8.09900939e-01 6.37721270e-02
-3.38889539e-01 3.77263844e-01 2.52720583e-02 6.05883598e-01
5.74674785e-01 -2.72638768e-01 -2.29318310e-02 1.40536845e+00
4.61648434e-01 -4.42153752e-01 3.55221361e-01 9.90252972e-01
-5.21966219e-01 3.78474176e-01 -6.78759158e-01 -3.99385929e-01
3.28024779e-03 1.17900610e+00 2.09563486e-02 -6.39840484e-01
-8.56005967e-01 1.35203326e+00 6.15814030e-01 6.40431523e-01
-8.98309588e-01 7.46782217e-03 6.32376671e-01 -1.97002962e-01
1.09362155e-01 -4.58032727e-01 -5.36325276e-02 -1.70607638e+00
-4.60007519e-01 -1.28003871e+00 3.64047289e-01 -1.28275681e+00
-1.47584248e+00 9.31295276e-01 -1.13541363e-02 -8.73390794e-01
-6.50500476e-01 -7.83719063e-01 -1.99397355e-01 1.10713887e+00
-1.95372057e+00 -2.33398986e+00 -1.08485788e-01 1.07862616e+00
9.56624687e-01 -5.67356288e-01 1.01891768e+00 3.94823879e-01
-2.59149492e-01 8.43794107e-01 -1.64415985e-02 2.40937456e-01
1.27748847e+00 -1.23176813e+00 4.11913037e-01 9.07835245e-01
4.75153476e-01 4.48829263e-01 3.19108665e-02 -5.11297047e-01
-1.56633294e+00 -1.35434079e+00 9.52022493e-01 -5.51499486e-01
8.42722595e-01 -6.65331900e-01 -6.78056717e-01 1.07265592e+00
1.21106255e+00 -3.97544116e-01 8.12273204e-01 2.54595578e-01
-8.03428590e-01 2.73331165e-01 -5.62361300e-01 7.87330985e-01
8.12472343e-01 -1.36167347e+00 -5.28522134e-01 4.05581027e-01
7.29869127e-01 -3.58811259e-01 -1.05070138e+00 4.62239712e-01
3.61593753e-01 -6.44894123e-01 9.65820491e-01 -7.08345294e-01
9.19811428e-01 -5.09285517e-02 -4.94070739e-01 -1.42552829e+00
7.16956407e-02 -5.31635046e-01 1.63740575e-01 1.35740089e+00
8.49159479e-01 -3.15528005e-01 -6.01781346e-03 -3.82783636e-02
-2.76122957e-01 -4.69179839e-01 -6.62583590e-01 -4.65840489e-01
5.05573750e-01 -5.57619572e-01 2.73737043e-01 1.03710544e+00
-3.30530673e-01 1.08747065e+00 -7.01795578e-01 9.25356895e-02
5.43398798e-01 5.04649222e-01 9.98627484e-01 -5.38871825e-01
-2.78836966e-01 -1.96219146e-01 2.23420680e-01 -1.48656416e+00
6.69110596e-01 -1.45954406e+00 8.14832747e-02 -1.65131915e+00
3.75871778e-01 -1.02065457e-02 9.56360251e-02 8.03904474e-01
-3.01324874e-01 6.49229586e-01 3.22108001e-01 1.01266146e-01
-8.14723849e-01 9.74733889e-01 1.68196177e+00 -3.84876221e-01
2.81905830e-01 -7.97611177e-01 -6.81846976e-01 4.43277657e-01
5.90593159e-01 9.56994891e-02 -6.45380020e-01 -1.39505517e+00
8.04143846e-02 1.93821073e-01 5.03605783e-01 -6.09319925e-01
-7.18254969e-02 2.71160342e-02 3.58971179e-01 -7.48813748e-01
5.29756844e-01 -3.57570410e-01 -1.83838829e-01 -7.50541687e-02
-6.59356952e-01 4.73821521e-01 4.04990554e-01 2.97747999e-01
-6.90823257e-01 1.48735628e-01 5.61864495e-01 -1.67017639e-01
-1.08229530e+00 3.67373735e-01 -5.55350743e-02 4.54611599e-01
7.74554968e-01 2.34913781e-01 -6.68078005e-01 -6.76585674e-01
-8.32199395e-01 6.14141464e-01 3.56244385e-01 9.42977548e-01
5.69718003e-01 -1.64765286e+00 -1.22217953e+00 1.71649575e-01
7.03750610e-01 -4.87850338e-01 5.37083566e-01 7.34605253e-01
-2.10856140e-01 8.83218586e-01 -2.46484235e-01 -7.03154922e-01
-1.10213375e+00 5.52668869e-01 2.22977176e-01 -6.33222699e-01
-4.23034504e-02 8.27465773e-01 7.55066097e-01 -1.05887020e+00
-2.55719602e-01 -6.12498261e-02 1.28065199e-02 3.69286388e-02
3.67750496e-01 -2.89602011e-01 -3.34734201e-01 -1.01080036e+00
-1.34632096e-01 6.10470951e-01 -3.52513678e-02 -5.74948907e-01
7.82406807e-01 -4.80873853e-01 -3.40701975e-02 4.23096955e-01
1.50575888e+00 -2.43787646e-01 -1.08541882e+00 -3.50253105e-01
-2.83298880e-01 8.66063014e-02 -1.58051729e-01 -1.39640486e+00
-1.08029485e+00 1.39930463e+00 4.86834586e-01 -4.96972412e-01
1.13350523e+00 4.13117975e-01 9.69820201e-01 2.72044063e-01
1.88858017e-01 -1.03059864e+00 2.51159370e-01 8.59570146e-01
1.30360675e+00 -1.77496910e+00 -5.21106303e-01 -1.53537681e-02
-1.02255571e+00 9.31777000e-01 1.06920874e+00 3.76108468e-01
1.21042803e-01 -9.90422163e-03 4.90581125e-01 -9.28795431e-03
-1.08113062e+00 -4.86264795e-01 8.40951622e-01 7.13774860e-01
8.16161156e-01 2.83909917e-01 9.68815759e-03 4.40936297e-01
-3.37952882e-01 -1.06807403e-01 -5.43661714e-02 4.51040208e-01
3.03956419e-01 -1.28607690e+00 -2.68752933e-01 -2.29596734e-01
-1.09837994e-01 -7.04585612e-01 -2.94751644e-01 1.09629488e+00
4.13349569e-01 8.55483830e-01 6.40880689e-02 -3.04228336e-01
6.41230196e-02 3.75527948e-01 7.04743087e-01 -5.01854122e-01
-3.88571024e-01 1.44028068e-01 1.86910912e-01 -5.37650168e-01
-5.02914190e-01 -3.03656369e-01 -8.26616645e-01 2.65182313e-02
3.27304029e-03 -9.41613764e-02 6.93031609e-01 9.64021385e-01
5.91154158e-01 5.69962680e-01 5.36169782e-02 -5.92626512e-01
-1.70330107e-01 -1.10295796e+00 -4.40400243e-02 5.01329958e-01
1.30977198e-01 -2.74663210e-01 9.81939062e-02 6.03696167e-01]
|
[11.175018310546875, 1.610640287399292]
|
96968888-1b76-4f97-826c-19152d2701b2
|
learning-from-imperfect-training-data-using-a
|
2208.04941
| null |
https://arxiv.org/abs/2208.04941v1
|
https://arxiv.org/pdf/2208.04941v1.pdf
|
Learning from imperfect training data using a robust loss function: application to brain image segmentation
|
Segmentation is one of the most important tasks in MRI medical image analysis and is often the first and the most critical step in many clinical applications. In brain MRI analysis, head segmentation is commonly used for measuring and visualizing the brain's anatomical structures and is also a necessary step for other applications such as current-source reconstruction in electroencephalography and magnetoencephalography (EEG/MEG). Here we propose a deep learning framework that can segment brain, skull, and extra-cranial tissue using only T1-weighted MRI as input. In addition, we describe a robust method for training the model in the presence of noisy labels.
|
['Richard M. Leahy', 'Anand A Joshi', 'Wenhui Cui', 'Haleh Akrami']
|
2022-08-08
| null | null | null | null |
['brain-image-segmentation']
|
['medical']
|
[ 2.31213838e-01 -5.97912595e-02 2.24809900e-01 -5.67417800e-01
-3.58394712e-01 -1.24167286e-01 2.22511590e-01 2.77120143e-01
-7.46685624e-01 6.45331860e-01 5.22883656e-03 -3.73088419e-01
-1.03723437e-01 -4.37413782e-01 -5.34594595e-01 -7.63763905e-01
-1.17997132e-01 6.76186323e-01 2.22444698e-01 2.90034920e-01
2.29128107e-01 5.72406709e-01 -8.74512672e-01 -1.46273896e-01
7.67866850e-01 1.00300264e+00 3.79891276e-01 1.46561325e-01
-1.60336316e-01 4.32473391e-01 -5.81749499e-01 -1.48216248e-01
-1.30654454e-01 -4.27219003e-01 -7.20030189e-01 2.41994068e-01
-2.07165420e-01 -1.94078818e-01 2.69932970e-02 1.36070895e+00
5.68663895e-01 7.21439123e-02 5.03605247e-01 -6.58431411e-01
-3.05077359e-02 7.06429064e-01 -7.75825381e-01 5.88496089e-01
-1.54465809e-01 -2.69715965e-01 8.35732371e-02 -5.77075899e-01
3.89711142e-01 6.65709555e-01 1.31672114e-01 3.77162129e-01
-7.48056412e-01 -7.45890558e-01 1.64587662e-01 3.62871617e-01
-1.07799757e+00 -3.20989579e-01 6.30402684e-01 -5.96889377e-01
4.19371665e-01 3.36618349e-02 6.60243869e-01 8.06555748e-01
6.88613892e-01 7.41594315e-01 1.26398742e+00 -9.98776779e-02
4.39018786e-01 -3.69308889e-01 4.06747490e-01 3.82423878e-01
-3.90399285e-02 -4.09129471e-01 -3.11142504e-02 -3.17299925e-02
9.93235648e-01 1.64479330e-01 -5.22360384e-01 -1.60289556e-01
-1.33383322e+00 5.62241435e-01 5.11632800e-01 6.96776628e-01
-5.57188749e-01 -1.80261377e-02 3.48867953e-01 -2.06250831e-01
3.67147863e-01 2.05461100e-01 -4.60135676e-02 1.34463480e-03
-1.17344201e+00 -2.97838777e-01 1.22065954e-01 2.18629971e-01
2.62123514e-02 1.18130438e-01 2.83169765e-02 7.67940402e-01
2.02879250e-01 2.71023661e-01 1.10934579e+00 -6.85694814e-01
1.41143482e-02 2.72792578e-01 -2.30688438e-01 -2.31235489e-01
-9.00722444e-01 -6.02856457e-01 -1.00774193e+00 9.32505652e-02
7.14494735e-02 -3.53801623e-02 -1.13689876e+00 1.26980329e+00
2.78869063e-01 2.91243315e-01 -5.51598787e-01 9.65006709e-01
8.73192668e-01 1.32026941e-01 2.27102831e-01 -4.66323435e-01
1.24164855e+00 -6.55869961e-01 -7.25767851e-01 -4.62490708e-01
2.41669744e-01 -1.52433887e-01 6.06402993e-01 3.57905507e-01
-9.68685687e-01 -1.25199795e-01 -7.63002276e-01 6.99434988e-03
-9.27800760e-02 -8.11150745e-02 5.63682616e-01 4.59988445e-01
-8.40400279e-01 3.63425434e-01 -1.17800426e+00 1.69301070e-02
5.79910338e-01 5.88005602e-01 -7.40273058e-01 8.05505738e-02
-8.74590993e-01 1.12608802e+00 1.15117326e-01 3.10119420e-01
-6.40907049e-01 -3.08811843e-01 -6.77753866e-01 8.85337368e-02
1.99555025e-01 -2.44860440e-01 1.07154453e+00 -3.45055819e-01
-1.26304889e+00 1.02900374e+00 -2.11304948e-01 -3.66202474e-01
2.28307322e-01 2.71115392e-01 -4.01682407e-01 2.21615225e-01
5.45068905e-02 6.85342729e-01 8.81664515e-01 -6.83084726e-01
-5.91631792e-02 -1.11166763e+00 -3.55470538e-01 6.65229112e-02
3.49080294e-01 3.73984307e-01 -1.44917145e-01 -4.88437504e-01
6.24524832e-01 -7.41957068e-01 -2.08061263e-01 -5.15473604e-01
-4.45911527e-01 -2.64265984e-02 3.95398170e-01 -1.17727256e+00
6.89954519e-01 -1.84362888e+00 3.02283406e-01 4.11644220e-01
3.93101662e-01 1.13341197e-01 1.65673703e-01 -4.23667431e-01
-2.77868599e-01 -1.73878923e-01 -7.30221212e-01 -2.85224557e-01
-2.55586922e-01 -2.01237649e-01 4.35359292e-02 7.40742862e-01
-3.44164819e-01 8.90760303e-01 -5.43084562e-01 -2.28341714e-01
3.37199837e-01 5.70180237e-01 8.18692371e-02 -7.68684521e-02
2.29076207e-01 1.37019050e+00 -1.26823604e-01 3.82333279e-01
2.62001991e-01 -6.19429350e-02 1.38096094e-01 8.47121875e-04
-6.72366247e-02 1.59134358e-01 -6.79223776e-01 1.69111204e+00
-3.28923941e-01 5.61484098e-01 3.36019814e-01 -1.35327268e+00
5.83382487e-01 4.98052508e-01 6.76232040e-01 -9.53173280e-01
8.57062876e-01 2.43354857e-01 4.88166332e-01 -5.36957204e-01
-2.36353070e-01 -1.88278988e-01 2.13010564e-01 7.92224109e-01
1.59573201e-02 -2.16128126e-01 8.69936123e-02 -1.18071146e-01
7.17385828e-01 -3.39388102e-01 1.77771047e-01 -2.68361628e-01
4.50327367e-01 -6.09527886e-01 5.28398871e-01 1.70854345e-01
-2.43919522e-01 7.48744071e-01 3.63613397e-01 -2.95058012e-01
-5.43824911e-01 -7.88281083e-01 -5.88291585e-01 5.91249347e-01
-1.82766095e-01 2.70080864e-01 -1.19707263e+00 -5.14576614e-01
-4.56907898e-01 5.99682927e-01 -4.87855792e-01 -2.61154361e-02
-5.39926887e-01 -1.05471683e+00 1.52571216e-01 6.69007123e-01
4.81054038e-01 -1.22023380e+00 -8.94779503e-01 2.46129230e-01
-3.17272216e-01 -1.11017847e+00 -2.15456679e-01 4.36193585e-01
-1.02441502e+00 -1.23323262e+00 -1.14293766e+00 -7.60968029e-01
9.29432392e-01 -8.47461373e-02 5.90050697e-01 1.25064060e-01
-4.71693248e-01 6.98805377e-02 1.01673193e-01 -4.83710051e-01
-6.45386204e-02 1.29879266e-01 -2.59824600e-02 2.48836204e-02
1.26809314e-01 -5.99766374e-01 -5.74000895e-01 1.12297148e-01
-9.70844507e-01 1.11203834e-01 3.80666524e-01 4.28313911e-01
7.22020388e-01 5.38902469e-02 5.96422434e-01 -8.44654858e-01
8.80961955e-01 -3.68924409e-01 -6.06428266e-01 3.03057283e-01
2.81965099e-02 -2.00975776e-01 3.55421335e-01 -1.61719650e-01
-5.59985876e-01 -1.52563034e-02 -6.55354798e-01 7.14314077e-03
-3.17700654e-01 5.29033482e-01 -1.90295815e-01 -2.65637308e-01
1.86440066e-01 1.78709120e-01 -1.23272799e-01 -4.71441597e-01
-2.91600049e-01 4.98154491e-01 6.80620313e-01 -7.12634251e-02
-2.98853107e-02 5.25289774e-01 1.38074383e-01 -8.92728627e-01
-4.84074861e-01 -3.22133899e-01 -1.00725210e+00 -1.89293623e-01
1.02319217e+00 -3.27053159e-01 -7.75932372e-01 5.04602253e-01
-9.66747761e-01 -4.78393972e-01 1.53739870e-01 7.72454143e-01
-2.82046825e-01 3.11742425e-01 -5.90978801e-01 -3.80136818e-01
-6.21909797e-01 -1.84067488e+00 8.82924616e-01 1.88359872e-01
-4.28199098e-02 -1.14204109e+00 -2.36429587e-01 3.09020042e-01
3.74612391e-01 9.13969725e-02 1.01297426e+00 -5.71923792e-01
-4.25922386e-02 -2.77898580e-01 -4.52905335e-02 2.72661805e-01
3.60435545e-02 -4.74432409e-01 -8.50135028e-01 -6.69263005e-02
5.60030758e-01 -5.03335223e-02 7.85899043e-01 1.02885926e+00
1.52986264e+00 4.41100895e-01 -1.81584284e-01 6.64688110e-01
8.48960876e-01 5.60033321e-01 5.91726542e-01 1.52369011e-02
5.41743100e-01 6.88495278e-01 -7.43156970e-02 -3.37278936e-03
3.90185326e-01 2.52600968e-01 3.62325966e-01 -9.59065333e-02
2.42436398e-03 3.73787045e-01 -1.15508921e-01 9.25936699e-01
6.66360511e-03 2.37651676e-01 -1.10243714e+00 4.56847131e-01
-1.36712503e+00 -6.05391622e-01 -1.80942327e-01 2.29631329e+00
4.76605505e-01 1.47800064e-02 1.12334769e-02 4.51719582e-01
8.79608750e-01 -2.00431108e-01 -7.77755022e-01 -1.27114847e-01
2.14535624e-01 6.10354424e-01 3.68143559e-01 5.10584891e-01
-8.47623229e-01 4.22239155e-01 6.84414816e+00 2.21978575e-01
-1.59990716e+00 4.05712932e-01 6.33477151e-01 8.75423253e-02
3.50807458e-02 -5.17852783e-01 -3.86489004e-01 7.32633352e-01
6.74658656e-01 1.54414356e-01 4.52047199e-01 3.56676131e-01
3.06183904e-01 -7.63416827e-01 -8.74787033e-01 1.06523943e+00
6.32684976e-02 -9.12618041e-01 -5.56729496e-01 2.66927630e-02
2.74044216e-01 1.93252578e-01 1.00784212e-01 -9.83193144e-02
-3.76793921e-01 -1.14776659e+00 5.48498452e-01 2.32457384e-01
9.09863353e-01 -4.06388998e-01 7.38604248e-01 5.52725911e-01
-5.69346845e-01 1.17274532e-02 -1.97933674e-01 1.45303592e-01
4.78610337e-01 8.87871325e-01 -5.14568448e-01 -6.56006336e-02
5.85761309e-01 1.99015453e-01 -2.52984703e-01 1.49442625e+00
-6.86873972e-01 4.74158168e-01 -1.82019204e-01 3.17090929e-01
-9.44875088e-03 -3.79752457e-01 1.66395262e-01 6.85860395e-01
3.53356689e-01 2.91188329e-01 5.64604923e-02 6.86599016e-01
-3.42201531e-01 3.81405316e-02 -2.27623850e-01 2.69614995e-01
1.10408321e-01 1.25271297e+00 -1.44227695e+00 -1.18550219e-01
-4.68375646e-02 8.23388040e-01 1.29781410e-01 3.65524650e-01
-6.91674411e-01 -2.38243014e-01 -7.47007281e-02 2.89640635e-01
-2.80427992e-01 -3.87787461e-01 -4.08470362e-01 -1.05666542e+00
2.64537204e-02 -3.75135720e-01 1.12603717e-02 -6.59209967e-01
-6.14076316e-01 7.04093635e-01 7.57235140e-02 -4.35325950e-01
-1.63724110e-01 -5.86075902e-01 -8.27097237e-01 8.84248495e-01
-1.56738198e+00 -6.30784631e-01 -3.92882228e-01 7.77466893e-01
3.24553132e-01 1.66426927e-01 8.00702631e-01 4.90218878e-01
-8.00370872e-01 3.26366387e-02 -3.77425961e-02 3.77520591e-01
4.65859473e-01 -9.67569113e-01 3.30062211e-01 7.40227222e-01
9.63905230e-02 6.54570103e-01 2.50350833e-01 -5.37797570e-01
-6.62198007e-01 -6.95663571e-01 6.70624018e-01 3.26307833e-01
4.17365104e-01 -3.36374849e-01 -9.26845253e-01 7.52513587e-01
1.58880889e-01 1.93613216e-01 7.41973221e-01 -2.71639913e-01
3.22902083e-01 -4.56331111e-02 -1.28439212e+00 1.96092665e-01
3.88560742e-01 -4.25462663e-01 -6.08951986e-01 5.25242448e-01
1.97543693e-03 -6.23500466e-01 -7.11603820e-01 3.35728317e-01
3.94276232e-01 -8.33757043e-01 5.31356931e-01 -1.24361366e-01
-1.27934977e-01 1.75709231e-03 5.16733587e-01 -1.55877578e+00
-2.37735569e-01 -2.07796946e-01 1.54630020e-01 5.80396891e-01
-6.69935495e-02 -7.25756764e-01 5.90683103e-01 6.88580453e-01
-2.71466345e-01 -5.38140714e-01 -1.01307380e+00 -3.59679252e-01
6.42515793e-02 -4.31705326e-01 5.70917308e-01 6.85869992e-01
-1.90981384e-02 9.11913142e-02 1.52054951e-01 -1.09933741e-01
7.59165227e-01 -1.66391268e-01 -6.41427785e-02 -1.28375518e+00
2.51031160e-01 -6.50397718e-01 -3.86978626e-01 -6.65721357e-01
4.84828860e-01 -1.02685773e+00 1.63640067e-01 -1.93211997e+00
2.48738989e-01 -2.87944287e-01 -5.22653282e-01 3.04739118e-01
1.08516514e-01 4.06632155e-01 -2.03576893e-01 5.93383461e-02
-1.75515592e-01 1.97775006e-01 1.22229266e+00 -1.07538275e-01
1.25657171e-01 3.51638466e-01 -3.43964517e-01 8.49225283e-01
8.53060901e-01 -4.53182101e-01 -3.03792685e-01 -7.31201053e-01
-2.78134644e-01 1.47401124e-01 2.86471903e-01 -9.97172594e-01
4.39469635e-01 2.62328655e-01 6.42955601e-01 -4.09690887e-01
1.77320167e-01 -7.95790970e-01 -7.65676424e-02 3.27963173e-01
-1.88855901e-01 2.16581985e-01 -4.99578826e-02 -6.37249500e-02
-3.11240852e-01 -4.65690374e-01 1.09644055e+00 -4.42491561e-01
-3.11596662e-01 5.07117212e-01 -6.39841616e-01 -1.29476175e-01
9.83981490e-01 -1.32132706e-03 -1.25844419e-01 -2.67596990e-01
-1.04887545e+00 1.98931411e-01 1.12766266e-01 1.72906980e-01
8.21534395e-01 -7.62341619e-01 -4.11198944e-01 2.93579668e-01
-5.18024981e-01 7.49194324e-02 3.94115537e-01 1.59171963e+00
-5.55777133e-01 4.68877643e-01 -5.14092922e-01 -5.97904146e-01
-1.01301634e+00 2.71975189e-01 5.28168380e-01 -1.10015541e-01
-6.09344363e-01 7.41127491e-01 2.34895378e-01 -4.96323146e-02
5.64063549e-01 -4.77144718e-01 -5.42802453e-01 -6.23982400e-02
6.99890256e-01 1.91427052e-01 6.06371999e-01 -8.64733219e-01
-4.10208732e-01 3.41500849e-01 2.89538782e-02 -2.72392333e-01
1.26062727e+00 -6.22677021e-02 -5.50507426e-01 4.91524190e-01
8.79866362e-01 -3.72856438e-01 -7.49604940e-01 7.40470365e-02
1.48482740e-01 4.00002971e-02 6.22783601e-01 -7.73102760e-01
-1.43304074e+00 1.55524528e+00 5.48451602e-01 -1.77869752e-01
1.07732522e+00 -8.83190185e-02 9.92047191e-01 -7.10563660e-02
6.34604096e-01 -1.04582798e+00 -4.27501559e-01 1.75263777e-01
6.89484715e-01 -1.14432025e+00 -5.54019818e-03 4.84134220e-02
-7.01166451e-01 9.28620040e-01 3.07102829e-01 5.27367555e-02
9.85756099e-01 4.94676232e-01 -1.06840897e-02 -2.87045240e-01
5.69769703e-02 -1.26860797e-01 4.25940961e-01 3.11458141e-01
5.92492819e-01 5.03785647e-02 -3.91062140e-01 7.22205818e-01
1.02337889e-01 -5.40276170e-02 2.22536623e-01 9.31802094e-01
-4.61242855e-01 -9.05630827e-01 -2.92979240e-01 8.33033144e-01
-9.10706997e-01 5.61777828e-03 -9.92262289e-02 3.12371552e-01
5.12811467e-02 6.18141949e-01 -3.10723465e-02 5.97055741e-02
3.41637060e-02 3.10606360e-01 7.94491172e-01 -7.74817288e-01
-6.20321214e-01 2.32939973e-01 -5.62920690e-01 -4.23416317e-01
-4.47951257e-01 -3.87730092e-01 -1.81366158e+00 2.10159034e-01
-1.21016212e-01 1.18815906e-01 1.37484717e+00 1.33005965e+00
7.16440976e-02 6.61226153e-01 4.16993231e-01 -8.02531302e-01
7.26989210e-02 -9.29097652e-01 -8.31845284e-01 2.18000591e-01
1.17343619e-01 -7.09618092e-01 3.02608516e-02 -1.45042837e-01]
|
[14.27866268157959, -2.2535173892974854]
|
69fc128e-01d8-463e-975d-0b9b8cc1bc3f
|
multiresolution-attention-extractor-for-small
|
2006.05941
| null |
https://arxiv.org/abs/2006.05941v1
|
https://arxiv.org/pdf/2006.05941v1.pdf
|
MultiResolution Attention Extractor for Small Object Detection
|
Small objects are difficult to detect because of their low resolution and small size. The existing small object detection methods mainly focus on data preprocessing or narrowing the differences between large and small objects. Inspired by human vision "attention" mechanism, we exploit two feature extraction methods to mine the most useful information of small objects. Both methods are based on multiresolution feature extraction. We initially design and explore the soft attention method, but we find that its convergence speed is slow. Then we present the second method, an attention-based feature interaction method, called a MultiResolution Attention Extractor (MRAE), showing significant improvement as a generic feature extractor in small object detection. After each building block in the vanilla feature extractor, we append a small network to generate attention weights followed by a weighted-sum operation to get the final attention maps. Our attention-based feature extractor is 2.0 times the AP of the "hard" attention counterpart (plain architecture) on the COCO small object detection benchmark, proving that MRAE can capture useful location and contextual information through adaptive learning.
|
['Xu Liu', 'Licheng Jiao', 'Lingling Li', 'Fang Liu', 'Fan Zhang']
|
2020-06-10
| null | null | null | null |
['small-object-detection', 'hard-attention']
|
['computer-vision', 'methodology']
|
[ 2.25766540e-01 7.48661608e-02 2.43759230e-01 -8.47068802e-03
-7.57712662e-01 -3.18867683e-01 5.38882613e-01 8.47767442e-02
-5.99819362e-01 2.36530498e-01 1.06436461e-01 6.23097196e-02
-6.15181699e-02 -7.67124414e-01 -7.32025445e-01 -7.31976926e-01
-7.82524943e-02 1.97372675e-01 8.06362808e-01 -2.84948915e-01
5.14738917e-01 6.35104895e-01 -1.73864162e+00 2.32879430e-01
7.50671804e-01 1.19161880e+00 6.19250357e-01 8.49174559e-01
-1.04533710e-01 6.85386121e-01 -5.59037745e-01 -8.64605829e-02
3.03370625e-01 -1.96558803e-01 -8.75695825e-01 -4.90915217e-02
4.79182035e-01 -1.98007569e-01 -2.05010235e-01 1.02301908e+00
7.13209689e-01 2.95622140e-01 6.06322706e-01 -1.03034365e+00
-1.14581442e+00 6.52585924e-01 -1.00669754e+00 8.50525677e-01
3.39578450e-01 9.30784941e-02 1.14113271e+00 -1.38701522e+00
3.29152882e-01 1.21403444e+00 6.13401055e-01 3.62491786e-01
-1.00314033e+00 -4.64317977e-01 2.58709103e-01 6.09949112e-01
-1.45599186e+00 -2.61070490e-01 6.50768995e-01 -3.34065944e-01
1.21771741e+00 5.47924101e-01 5.65511584e-01 8.97673726e-01
1.60547480e-01 1.01489699e+00 7.55192399e-01 -5.27276397e-01
-1.00606814e-01 2.62911499e-01 3.76395822e-01 5.40951610e-01
2.85346180e-01 -7.16073960e-02 -1.21330880e-01 1.48179844e-01
4.64461505e-01 3.54147613e-01 -3.00037324e-01 -2.08299398e-01
-1.23910248e+00 9.23034668e-01 1.13414168e+00 2.98323959e-01
-6.09850705e-01 4.70641116e-03 2.25966141e-01 3.36984694e-01
2.09525868e-01 5.86072385e-01 -3.45057487e-01 3.44210804e-01
-5.47926486e-01 1.60780042e-01 4.15726632e-01 7.99550414e-01
7.43559957e-01 -1.78914398e-01 -7.01073647e-01 8.02689850e-01
2.18415275e-01 3.95812243e-01 8.31717670e-01 -5.18878043e-01
2.23602980e-01 9.01399016e-01 1.61734655e-01 -1.09830189e+00
-5.42814553e-01 -4.48624015e-01 -7.03819513e-01 3.46293747e-01
1.86102986e-01 -2.62559354e-02 -9.85653222e-01 1.42918670e+00
3.97041529e-01 -8.36425498e-02 -1.88240126e-01 1.12564909e+00
9.89552796e-01 5.89359641e-01 6.43202290e-02 1.54777244e-01
1.74478662e+00 -1.30256796e+00 -3.99581850e-01 -4.63511884e-01
1.95299327e-01 -6.43103600e-01 1.08381808e+00 2.12544724e-01
-1.00471497e+00 -8.26237559e-01 -1.12595904e+00 -2.95534760e-01
-6.83279335e-01 1.41196370e-01 6.87212169e-01 2.96525449e-01
-7.67057478e-01 4.58923072e-01 -4.97813255e-01 -4.16588306e-01
7.12971866e-01 4.99564290e-01 -3.93514872e-01 1.26165822e-01
-8.93313348e-01 1.02487171e+00 3.43771279e-01 1.78365648e-01
-8.22813451e-01 -4.36092854e-01 -7.81893313e-01 5.05598307e-01
6.83264434e-01 -7.09749401e-01 1.00271225e+00 -1.00342941e+00
-1.25258386e+00 5.55854797e-01 5.46361245e-02 -5.61452627e-01
1.95552424e-01 -5.94547331e-01 8.41032937e-02 1.66780040e-01
2.07915634e-01 7.69540966e-01 1.15656102e+00 -8.64629924e-01
-9.39654171e-01 -2.41620734e-01 8.91899168e-02 5.51001169e-02
-4.83722210e-01 4.45692092e-01 -4.72388297e-01 -8.14000845e-01
-5.98227642e-02 -6.43762052e-01 -4.74661976e-01 -1.89330861e-01
-3.95764053e-01 -5.75468719e-01 1.07326782e+00 -5.18735409e-01
1.23810828e+00 -2.02975488e+00 2.74547696e-01 4.35283482e-02
5.81249118e-01 3.53604227e-01 -3.92080873e-01 -9.73448977e-02
-2.24619418e-01 2.92689595e-02 -4.59962301e-02 -9.06250700e-02
-9.95010212e-02 -2.33912215e-01 -1.31869555e-01 2.34316662e-01
7.85997748e-01 1.26197731e+00 -9.00419831e-01 -4.69284683e-01
1.17813073e-01 4.48870748e-01 -6.77446723e-01 2.48342797e-01
1.91129148e-01 -3.08675263e-02 -6.81961358e-01 7.89333582e-01
5.57899117e-01 -2.76244551e-01 -4.77415979e-01 -3.98881853e-01
-2.84348994e-01 -1.28231540e-01 -1.11774778e+00 1.32499051e+00
-3.51753861e-01 8.05616796e-01 -4.63797897e-02 -7.91118741e-01
9.43622530e-01 -1.07499778e-01 2.94951558e-01 -7.01107740e-01
3.96840304e-01 6.73879171e-04 3.80520940e-01 -6.73880219e-01
6.86080098e-01 1.36993289e-01 -1.72829702e-02 3.63529384e-01
8.47133026e-02 1.22780345e-01 2.54527390e-01 1.58369675e-01
1.25809574e+00 -8.74095038e-02 4.21442151e-01 -2.64730006e-01
6.55625105e-01 -4.26142156e-01 3.20684373e-01 1.08092225e+00
-2.97682464e-01 8.22582543e-01 3.88181269e-01 -5.10794818e-01
-8.53760898e-01 -6.22275412e-01 -8.40668287e-03 1.73709357e+00
4.53753695e-02 -3.99306506e-01 -7.65259922e-01 -1.01016486e+00
1.22893550e-01 2.44473979e-01 -1.13807940e+00 -3.45809788e-01
-5.79841971e-01 -8.34504843e-01 3.33642364e-02 9.93648052e-01
3.21291447e-01 -1.70239925e+00 -1.21413708e+00 1.42443284e-01
1.97741091e-02 -8.25012088e-01 -6.13180697e-01 4.08395737e-01
-2.65365779e-01 -1.03847229e+00 -9.22282338e-01 -6.76736116e-01
5.67794800e-01 5.37729204e-01 8.60603809e-01 1.05565891e-01
-6.27911627e-01 2.73734152e-01 -6.52207792e-01 -8.22357953e-01
9.85146761e-02 3.10531437e-01 -6.82886615e-02 3.80494326e-01
4.93337363e-01 -2.71555960e-01 -7.66965270e-01 2.10660771e-01
-6.93765402e-01 -2.56679863e-01 1.24851215e+00 9.54268515e-01
5.59133112e-01 -3.06957424e-01 6.24766290e-01 -5.57190359e-01
5.82005262e-01 -5.23634493e-01 -5.74257195e-01 1.18595339e-01
-2.04927385e-01 2.63443321e-01 6.02945089e-01 -6.48167729e-01
-7.26553261e-01 1.64658174e-01 -6.25274852e-02 -5.69881558e-01
-8.34063441e-02 1.73934504e-01 -5.43978922e-02 -2.98577547e-01
6.61682904e-01 1.66270435e-01 -2.74299592e-01 -6.90471113e-01
2.91661352e-01 7.19858706e-01 3.85179490e-01 -9.76186097e-02
6.83431089e-01 3.03394079e-01 -2.38764346e-01 -7.46486843e-01
-1.06953573e+00 -5.36870182e-01 -7.17620552e-01 1.36970118e-01
8.93325388e-01 -4.18965399e-01 -9.35659051e-01 2.39953011e-01
-1.13306975e+00 -1.49138803e-02 -6.52488470e-01 4.30127621e-01
-3.97997320e-01 2.13885173e-01 -3.57298762e-01 -6.91191912e-01
-7.48390317e-01 -1.09697032e+00 1.28377950e+00 3.55423182e-01
8.82832613e-03 -2.97767162e-01 -1.22432247e-01 -3.34022404e-03
7.10084260e-01 1.01208545e-01 5.77004850e-01 -8.35260451e-01
-5.31852543e-01 -5.32611191e-01 -8.50212514e-01 2.13460237e-01
-1.03993423e-01 -1.14043571e-01 -1.00975776e+00 -3.01711559e-01
7.51912370e-02 -2.29694784e-01 1.36616182e+00 4.80821937e-01
1.34892118e+00 -2.05239072e-01 -4.33214903e-01 5.27267516e-01
1.27791691e+00 -2.79450268e-02 5.77081442e-01 4.14426833e-01
7.69778192e-01 3.75864506e-01 5.69649875e-01 4.63035643e-01
1.96164697e-01 7.39521861e-01 5.91513693e-01 -2.02765197e-01
-2.44717687e-01 2.98008144e-01 2.95516342e-01 4.56067771e-01
-5.28915346e-01 5.49906008e-02 -8.29651415e-01 6.29697621e-01
-1.83463144e+00 -1.06719840e+00 -7.13462141e-05 2.12514734e+00
3.71093899e-01 4.23695773e-01 2.42794633e-01 2.50331722e-02
6.36050463e-01 -5.93112558e-02 -5.91192544e-01 -2.51591057e-01
-5.33680208e-02 3.89519960e-01 3.59043777e-01 1.77645922e-01
-1.32267356e+00 9.51502621e-01 6.23620701e+00 8.67224038e-01
-9.20397818e-01 2.15034425e-01 2.81933099e-01 -1.41933814e-01
2.59254307e-01 -3.56389582e-01 -9.26629186e-01 3.45258385e-01
7.58558929e-01 -1.13093227e-01 1.15634270e-01 1.05629158e+00
-5.18062003e-02 3.79946120e-02 -9.67284679e-01 7.60441482e-01
1.77883253e-01 -1.10267293e+00 2.45161846e-01 -1.26710251e-01
3.99806857e-01 2.41672233e-01 -1.03078848e-02 6.83710933e-01
1.59757901e-02 -1.04719281e+00 7.72305548e-01 6.11448050e-01
3.73644233e-01 -8.32819223e-01 9.29662049e-01 7.30371848e-02
-1.48018658e+00 -7.13656068e-01 -7.18595803e-01 7.36716539e-02
-1.30217120e-01 1.92768380e-01 -5.56984901e-01 2.63395727e-01
1.02435064e+00 5.01592577e-01 -1.04896641e+00 1.45498657e+00
-7.32835159e-02 2.52717078e-01 -3.51324230e-01 -2.15505227e-01
3.38341206e-01 2.62972355e-01 8.37993920e-01 1.50946236e+00
3.03942472e-01 -3.50400209e-02 1.22382678e-01 1.01830053e+00
-1.00899488e-01 9.27693397e-02 -4.66156781e-01 2.03282416e-01
1.60035178e-01 1.74653935e+00 -1.00288260e+00 -4.45500553e-01
-6.51702881e-01 1.10292304e+00 5.30126631e-01 6.34904206e-02
-1.02200890e+00 -1.06873250e+00 3.64389926e-01 5.26853949e-02
1.08576775e+00 1.17632665e-01 1.68491453e-02 -9.24563885e-01
-1.52217463e-01 -7.67709970e-01 4.03580844e-01 -7.82767534e-01
-1.08906555e+00 9.80872393e-01 -1.84785962e-01 -1.02255392e+00
3.96126397e-02 -7.37010598e-01 -7.68825471e-01 7.60394454e-01
-1.50347424e+00 -1.24355459e+00 -5.89541376e-01 5.72309136e-01
9.47629809e-01 -9.50385258e-02 5.91392756e-01 3.39120358e-01
-7.91842818e-01 6.99405372e-01 -2.08948359e-01 1.09713152e-01
4.89826858e-01 -1.39342546e+00 4.29812431e-01 8.07061017e-01
1.55788362e-01 5.24237335e-01 4.34965163e-01 -3.79521847e-01
-1.39391530e+00 -1.10899448e+00 5.39361596e-01 -7.53099382e-01
5.51728189e-01 -4.29838210e-01 -1.03775156e+00 7.64735937e-01
2.86638647e-01 5.05809963e-01 2.61017174e-01 1.14072869e-02
-2.94422120e-01 -1.11384131e-02 -9.91916180e-01 4.27367508e-01
9.36577499e-01 -6.14479445e-02 -8.27266753e-01 6.49547055e-02
9.67081606e-01 3.94204399e-03 -5.79923689e-01 4.63166445e-01
5.56089640e-01 -8.88689399e-01 1.17840660e+00 -7.78270841e-01
2.43342787e-01 -3.75876516e-01 4.52166284e-03 -1.04706168e+00
-9.75472748e-01 -5.93827963e-01 -4.21897501e-01 1.05358505e+00
3.00729394e-01 -4.87671524e-01 4.71022934e-01 1.03246108e-01
-2.86605746e-01 -7.50592589e-01 -6.77769721e-01 -5.67654252e-01
-3.08862805e-01 -2.13320285e-01 5.71214497e-01 6.72048450e-01
-2.28301063e-01 7.85284221e-01 -1.04168663e-02 7.67737627e-02
4.92564023e-01 4.75117743e-01 5.62314987e-01 -1.43659639e+00
-3.65524083e-01 -8.35682869e-01 -4.78166074e-01 -8.96357179e-01
-3.72899055e-01 -7.55275667e-01 1.18793607e-01 -1.35329473e+00
6.46537006e-01 -1.23222262e-01 -7.67642617e-01 6.86291695e-01
-5.53457797e-01 5.65180421e-01 3.81912261e-01 -8.08735937e-02
-1.00218785e+00 5.90678930e-01 1.06708968e+00 -1.54103130e-01
-3.65235925e-01 6.47688110e-04 -1.03766119e+00 8.96830261e-01
5.25071144e-01 -4.13393915e-01 1.48212776e-01 -1.90109029e-01
4.56190156e-03 -4.81920123e-01 4.71597522e-01 -1.17215943e+00
2.50545621e-01 2.52494141e-02 5.72609782e-01 -5.96142411e-01
5.57800792e-02 -8.25554729e-01 -5.12420595e-01 5.02411962e-01
-1.96532533e-01 -5.49530378e-03 3.09169024e-01 5.07119596e-01
-6.22865334e-02 -3.33678782e-01 7.88390815e-01 -1.56419188e-01
-1.04064143e+00 2.56700754e-01 -3.49513173e-01 -1.56242371e-01
1.10264933e+00 -2.78101057e-01 -1.29160628e-01 1.34045139e-01
-8.63535881e-01 1.49715930e-01 -1.57616988e-01 5.47965884e-01
7.69725680e-01 -1.37065959e+00 -9.08087015e-01 4.24038321e-01
1.59457624e-01 6.51370734e-02 9.93065462e-02 1.06267035e+00
-1.59619972e-01 3.74995232e-01 -4.81222600e-01 -4.18349057e-01
-1.28951311e+00 1.09373605e+00 2.65838295e-01 -2.10411280e-01
-7.14468598e-01 1.17641008e+00 4.37895685e-01 1.15911521e-01
1.91736862e-01 -5.10294139e-01 -6.90960884e-01 4.15675133e-01
8.86109531e-01 4.54548478e-01 8.00477266e-02 -5.21901488e-01
-4.26187336e-01 8.54062021e-01 -4.29336697e-01 4.93779004e-01
1.62168252e+00 -1.37315884e-01 5.43427616e-02 2.89352005e-03
9.45939362e-01 1.02044165e-01 -1.25184584e+00 -3.03287178e-01
8.79947096e-02 -5.04802048e-01 1.81694716e-01 -5.67015052e-01
-1.14139545e+00 9.49788392e-01 7.75801778e-01 7.17773795e-01
1.29063082e+00 3.06406349e-01 6.12131476e-01 5.51543117e-01
-9.78379548e-02 -9.53947306e-01 2.91281670e-01 6.16398513e-01
1.50314200e+00 -1.54951394e+00 3.84999812e-02 -1.62000865e-01
-5.93661547e-01 9.55124438e-01 9.31122124e-01 -3.79592896e-01
4.65697289e-01 3.50729436e-01 -3.86403561e-01 -4.23458099e-01
-7.41681695e-01 -7.64596224e-01 6.72859251e-01 6.06275082e-01
2.60808647e-01 -2.41424352e-01 -3.04466691e-02 9.97164130e-01
-5.61865326e-03 -1.85708940e-01 2.74019122e-01 6.56705916e-01
-9.11579251e-01 -4.94423836e-01 -4.20025527e-01 6.16408885e-01
-4.86698031e-01 -1.62447557e-01 -4.16444212e-01 7.20554650e-01
2.09219277e-01 5.78697145e-01 2.87835658e-01 -3.79150867e-01
5.79194665e-01 -2.09014982e-01 4.14372981e-01 -6.06157362e-01
-8.61751854e-01 4.84219380e-02 -4.36543941e-01 -7.45417058e-01
-1.33116543e-01 -3.91993940e-01 -1.07247996e+00 6.75641745e-02
-8.15617979e-01 -6.72684237e-02 2.75048077e-01 6.77942514e-01
5.67489505e-01 9.80499208e-01 6.12365723e-01 -1.42325950e+00
-5.53361297e-01 -1.22339511e+00 -2.40523130e-01 4.14411068e-01
4.53220516e-01 -7.78476238e-01 -1.50015488e-01 -3.00017893e-01]
|
[8.985608100891113, 0.12181495130062103]
|
594baaf7-4f68-480e-b1c6-e3fa341aabef
|
what-the-daam-interpreting-stable-diffusion
|
2210.04885
| null |
https://arxiv.org/abs/2210.04885v5
|
https://arxiv.org/pdf/2210.04885v5.pdf
|
What the DAAM: Interpreting Stable Diffusion Using Cross Attention
|
Large-scale diffusion neural networks represent a substantial milestone in text-to-image generation, but they remain poorly understood, lacking interpretability analyses. In this paper, we perform a text-image attribution analysis on Stable Diffusion, a recently open-sourced model. To produce pixel-level attribution maps, we upscale and aggregate cross-attention word-pixel scores in the denoising subnetwork, naming our method DAAM. We evaluate its correctness by testing its semantic segmentation ability on nouns, as well as its generalized attribution quality on all parts of speech, rated by humans. We then apply DAAM to study the role of syntax in the pixel space, characterizing head--dependent heat map interaction patterns for ten common dependency relations. Finally, we study several semantic phenomena using DAAM, with a focus on feature entanglement, where we find that cohyponyms worsen generation quality and descriptive adjectives attend too broadly. To our knowledge, we are the first to interpret large diffusion models from a visuolinguistic perspective, which enables future lines of research. Our code is at https://github.com/castorini/daam.
|
['Pontus Stenetorp', 'Gefei Yang', 'Zhiying Jiang', 'Linqing Liu', 'Ferhan Ture', 'Jimmy Lin', 'Karun Kumar', 'Akshat Pandey', 'Raphael Tang']
|
2022-10-10
| null | null | null | null |
['unsupervised-semantic-segmentation']
|
['computer-vision']
|
[ 2.11686701e-01 1.17049754e-01 -7.05978647e-02 -3.42708707e-01
-5.42821825e-01 -7.92587936e-01 1.13742220e+00 3.08769971e-01
-5.25863111e-01 3.07139963e-01 9.71859992e-01 -2.06762508e-01
-1.89891130e-01 -7.46375084e-01 -4.92012411e-01 -5.98524809e-01
2.52194732e-01 5.13361990e-01 -1.03198618e-01 -2.70601660e-01
2.51411468e-01 2.25182757e-01 -9.33956861e-01 4.55915391e-01
8.65729630e-01 8.00564766e-01 -4.05132845e-02 3.66637796e-01
1.40545413e-01 6.78971112e-01 -5.94089925e-01 -8.70150030e-01
6.17170427e-03 -7.52964139e-01 -7.49437273e-01 -2.29010303e-02
8.17179918e-01 -1.94883570e-01 -5.67862511e-01 1.26280987e+00
4.57591802e-01 -3.83377522e-02 1.03503931e+00 -1.23646796e+00
-1.58748806e+00 1.06172693e+00 -4.86330509e-01 6.76052570e-01
1.40100539e-01 6.01232231e-01 1.64251661e+00 -7.37522602e-01
1.15779507e+00 1.53834999e+00 4.81329560e-01 4.22802806e-01
-1.50634444e+00 -5.81156194e-01 5.40633835e-02 1.63122207e-01
-1.11315715e+00 -3.81889433e-01 8.96464467e-01 -7.12822199e-01
8.71763527e-01 3.52503248e-02 7.67875433e-01 1.69948947e+00
3.43382895e-01 7.31647789e-01 1.67373300e+00 -4.30371501e-02
1.49568141e-01 -1.68047756e-01 3.38927388e-01 7.68913925e-01
1.41009927e-01 5.89999370e-02 -7.42749751e-01 7.96465650e-02
7.52538562e-01 -3.05153579e-01 -1.26567289e-01 -1.07695073e-01
-1.57099319e+00 1.14814925e+00 9.68276322e-01 5.43114483e-01
-4.82140869e-01 4.21310663e-01 1.61918402e-01 3.13809484e-01
8.25893998e-01 5.78668177e-01 -1.23014137e-01 -2.42427811e-02
-9.50131059e-01 2.40062177e-01 5.21726489e-01 4.41558510e-01
5.05818248e-01 8.31469968e-02 -3.99977535e-01 9.61371303e-01
3.33091497e-01 4.61254746e-01 6.64593458e-01 -1.02287686e+00
2.96540171e-01 3.38438720e-01 -5.09451210e-01 -1.22376215e+00
-6.07461214e-01 -4.92281318e-01 -8.15134108e-01 1.57066867e-01
6.19491696e-01 -1.25917614e-01 -6.78883433e-01 2.06130886e+00
-1.77571714e-01 -4.05401140e-01 -2.65904307e-01 1.14009297e+00
6.44438446e-01 3.66695344e-01 4.04036641e-01 1.07754514e-01
1.58930445e+00 -8.39036822e-01 -5.58517039e-01 -3.60476434e-01
4.97411758e-01 -4.84855235e-01 1.25434077e+00 2.54384100e-01
-1.14027596e+00 -4.49788392e-01 -8.14834774e-01 -4.18736786e-01
-5.30199409e-01 -1.32074937e-01 7.77125478e-01 4.19724137e-01
-1.34192920e+00 7.22626388e-01 -6.52109563e-01 -6.34906292e-01
9.22850788e-01 -7.92218521e-02 -2.03676417e-01 1.27088234e-01
-1.35873330e+00 1.06847441e+00 3.02064698e-02 -2.16905177e-01
-9.16567147e-01 -6.87357247e-01 -7.05167890e-01 -1.45856276e-01
-2.43792813e-02 -9.04228628e-01 1.02522314e+00 -1.40028226e+00
-1.02897644e+00 1.10938847e+00 -1.04109794e-01 -3.47876400e-01
3.75323832e-01 8.06565583e-02 -4.61627811e-01 9.06811953e-02
3.45977783e-01 1.20710409e+00 8.17163050e-01 -1.09926569e+00
-1.23222336e-01 -7.19125807e-01 6.02053069e-02 3.01473409e-01
-5.22262454e-01 1.47839084e-01 -7.82002136e-02 -1.19170511e+00
2.92108059e-02 -8.99371266e-01 4.52191532e-02 3.02264661e-01
-5.02803922e-01 -3.16073477e-01 1.79186970e-01 -8.40577483e-01
9.86491203e-01 -2.14592409e+00 6.02235436e-01 2.08642602e-01
8.12818646e-01 -5.14641821e-01 -3.96122098e-01 3.33342582e-01
-1.85830623e-01 4.33367163e-01 -6.81692243e-01 -5.02000868e-01
4.80789840e-01 -5.11692800e-02 -2.66760260e-01 7.07404912e-01
3.30935031e-01 1.43653309e+00 -9.39557731e-01 -3.23536903e-01
-2.14996979e-01 5.39909542e-01 -5.89764893e-01 -5.18718541e-01
-2.59325266e-01 2.53186107e-01 -5.71402051e-02 5.49833059e-01
4.19858873e-01 -2.17914686e-01 -1.18862271e-01 -2.83090442e-01
-1.27233073e-01 2.56705523e-01 -4.13644016e-01 1.83344948e+00
-2.81264603e-01 1.19234526e+00 1.16065830e-01 -6.98262095e-01
5.66638768e-01 -1.26385927e-01 1.06278285e-01 -8.87619436e-01
2.61452138e-01 2.28763282e-01 6.76502228e-01 -1.59388348e-01
4.82836902e-01 -3.45901817e-01 -2.08787173e-01 7.39718914e-01
2.57409245e-01 -3.26509416e-01 2.10384175e-01 5.36361635e-01
1.07815087e+00 -6.82235062e-02 2.39294823e-02 -5.51651537e-01
-1.71791777e-01 1.11843437e-01 1.68559421e-02 7.16409802e-01
-2.47813061e-01 6.56987667e-01 9.26121533e-01 -3.42690647e-02
-1.18272996e+00 -1.28070831e+00 -2.44235486e-01 1.11368680e+00
-2.18415499e-01 -2.07994223e-01 -9.55810010e-01 -5.07712066e-01
1.63555294e-01 1.09041011e+00 -1.03364825e+00 -2.60407835e-01
-1.15728155e-01 -9.37803745e-01 7.54368424e-01 4.54030693e-01
3.35726291e-01 -1.28371310e+00 -2.14841843e-01 -1.50973246e-01
7.80529380e-02 -8.20014179e-01 -5.54015994e-01 -1.46036129e-02
-6.26662672e-01 -6.39670014e-01 -9.22522843e-01 -5.69989681e-01
5.28526247e-01 -1.58915445e-01 1.14392066e+00 -3.10614388e-02
-2.59393454e-01 3.07195336e-01 -6.98786080e-02 -1.84651479e-01
-5.05335510e-01 2.24401489e-01 -7.13843033e-02 -1.47135913e-01
3.42531741e-01 -5.16010344e-01 -6.67075217e-01 1.08393319e-01
-9.68691468e-01 9.56954584e-02 5.70768893e-01 5.81289113e-01
3.26931655e-01 -2.41851673e-01 4.54207331e-01 -8.06474090e-01
1.37277329e+00 -8.31902206e-01 -2.88400739e-01 -3.06512058e-01
-7.27498114e-01 -7.38756061e-02 3.22126657e-01 -2.83580303e-01
-9.48094070e-01 -4.04773772e-01 -5.80802113e-02 -1.49634793e-01
-2.23748192e-01 5.96683741e-01 1.07951716e-01 3.05869222e-01
8.99057567e-01 -1.40765002e-02 6.42961683e-03 -1.28965676e-01
8.45409214e-01 4.11842197e-01 3.83888185e-01 -4.70961809e-01
6.84650600e-01 6.37546837e-01 -3.05616230e-01 -8.53257477e-01
-9.24461126e-01 -4.99903522e-02 -8.10755610e-01 -1.93621099e-01
1.40693665e+00 -8.49990845e-01 -4.11393970e-01 7.26883948e-01
-1.17501462e+00 -7.02599049e-01 -3.41123074e-01 2.63229668e-01
-3.79485369e-01 1.85539693e-01 -8.03412378e-01 -1.49900243e-01
-3.50110009e-02 -1.08112299e+00 9.66594994e-01 -1.70609683e-01
-7.48106360e-01 -1.26758134e+00 2.30140790e-01 5.53254545e-01
4.04637098e-01 5.87928966e-02 9.76327419e-01 -5.96519411e-01
-3.31824958e-01 2.87090927e-01 -5.77476621e-01 9.23097581e-02
-2.36534581e-01 -4.28581573e-02 -9.90814328e-01 9.71192122e-02
-1.28640458e-01 -3.32825124e-01 1.53246474e+00 4.85886067e-01
1.16193867e+00 -2.83373654e-01 -1.42970353e-01 4.91347104e-01
1.00494492e+00 -3.35441619e-01 4.47946697e-01 1.60985202e-01
8.35376740e-01 7.83764601e-01 -5.91707975e-02 2.70140469e-01
4.98281837e-01 5.10194838e-01 2.15068802e-01 -2.38486305e-01
-5.78735650e-01 -3.48117530e-01 4.07702237e-01 6.59661889e-01
1.98820997e-02 -6.50429666e-01 -1.06024170e+00 6.85821235e-01
-1.50096178e+00 -1.08527052e+00 -4.35355484e-01 1.58724320e+00
7.83493698e-01 8.82298052e-02 5.68166450e-02 -3.03723574e-01
5.23705423e-01 6.78769886e-01 -6.02144182e-01 -3.80887061e-01
-6.23302400e-01 1.25491515e-01 4.01909202e-01 6.69786751e-01
-9.41605687e-01 1.16856432e+00 6.04895020e+00 7.97291338e-01
-8.51190925e-01 5.25476396e-01 9.93188500e-01 -2.66996145e-01
-8.74009192e-01 -1.48712322e-01 -1.95548251e-01 3.30065012e-01
9.68466997e-01 1.11327365e-01 7.36748755e-01 4.62692618e-01
3.00249815e-01 -3.20484824e-02 -1.03807843e+00 7.07293451e-01
4.36057925e-01 -1.32066286e+00 8.72102529e-02 2.05608696e-01
8.40965807e-01 4.86432880e-01 4.16744262e-01 -1.14471406e-01
6.48614407e-01 -1.07774186e+00 1.08572781e+00 5.05282819e-01
6.03142679e-01 -3.57396036e-01 2.25689664e-01 -1.49750888e-01
-5.50978065e-01 -6.34701103e-02 -2.75583297e-01 -4.14913334e-02
4.15332079e-01 6.50759101e-01 -2.48516276e-01 -2.91779399e-01
5.53311229e-01 9.65599656e-01 -8.63944769e-01 4.82411921e-01
-6.28078520e-01 7.07485497e-01 -2.26957165e-02 -1.10688955e-01
3.05188268e-01 -2.27207914e-01 5.62522829e-01 1.26106548e+00
2.52545714e-01 6.98731989e-02 -3.02269220e-01 1.56906986e+00
-5.68118870e-01 1.37429804e-01 -5.81124425e-01 -4.35142666e-01
1.25027582e-01 1.36659753e+00 -1.02203357e+00 -2.76767731e-01
-3.30460370e-01 1.63056588e+00 5.87728441e-01 4.92778718e-01
-9.92971361e-01 -5.16878590e-02 6.96949184e-01 1.45562932e-01
1.89743415e-02 -4.17944759e-01 -6.19479656e-01 -1.14491844e+00
-2.23367631e-01 -6.41261220e-01 2.11261556e-01 -1.22889423e+00
-1.83219743e+00 3.96959215e-01 -7.66359344e-02 -3.86056572e-01
2.05298111e-01 -6.95483983e-01 -5.29271901e-01 8.76305282e-01
-1.08740079e+00 -1.16368306e+00 -1.93664685e-01 5.38241386e-01
4.95314896e-01 -4.64680269e-02 5.49430847e-01 8.36666375e-02
-6.85137630e-01 3.75926644e-01 5.62561005e-02 1.88122094e-01
7.30336726e-01 -1.31133437e+00 7.06727862e-01 7.04450190e-01
5.68777978e-01 7.75137067e-01 5.02295673e-01 -8.94196391e-01
-9.40726697e-01 -9.35833871e-01 8.18102539e-01 -7.84326077e-01
1.24597895e+00 -5.74812233e-01 -6.96564138e-01 7.09069371e-01
6.97641969e-01 -2.28458464e-01 6.85069561e-01 2.54415363e-01
-5.79352021e-01 4.29342628e-01 -9.30476308e-01 8.39768767e-01
1.45637667e+00 -8.03284109e-01 -6.12507284e-01 5.64507306e-01
7.75186241e-01 3.72726955e-02 -1.03833985e+00 -3.78298163e-01
3.73620510e-01 -9.48191583e-01 8.54176819e-01 -4.80443090e-01
1.11529446e+00 5.89957200e-02 -1.32703722e-01 -1.80262649e+00
-8.09319854e-01 -2.48133481e-01 3.68742615e-01 1.25740767e+00
9.62869048e-01 -6.96577668e-01 3.97069126e-01 4.74823594e-01
-2.25270241e-01 -3.70776027e-01 -9.95867312e-01 -4.17463332e-01
5.92451990e-01 -6.03141785e-01 2.30126351e-01 1.26512146e+00
8.67384076e-02 6.79896116e-01 8.72767642e-02 -2.19756916e-01
5.26412129e-01 -2.71393269e-01 9.64543372e-02 -1.07991743e+00
-3.57056260e-01 -1.09486437e+00 -8.16499963e-02 -6.65552735e-01
4.73005772e-01 -1.42330360e+00 -2.52567559e-01 -1.80294335e+00
3.54293883e-01 -1.64899498e-01 -1.87112093e-01 5.29421270e-01
6.25462532e-02 6.78403020e-01 2.42902890e-01 4.34384942e-01
-3.68089348e-01 5.11212230e-01 1.29716289e+00 -3.91921461e-01
2.04705194e-01 -7.12525308e-01 -1.07316685e+00 7.75542200e-01
9.98761475e-01 -4.32085752e-01 -3.88861388e-01 -8.65567327e-01
3.25243026e-01 -4.29917663e-01 7.42310882e-01 -7.10628569e-01
-1.70578748e-01 -2.50403807e-02 6.01091623e-01 1.35947466e-02
3.05513263e-01 -3.96081060e-01 -5.38293235e-02 3.79017681e-01
-7.92103410e-01 2.77881503e-01 1.91455185e-02 3.63361627e-01
3.09104882e-02 3.21017485e-03 6.51057601e-01 -2.03098133e-01
-6.65695727e-01 2.57956386e-01 -6.24081433e-01 3.70422333e-01
7.81019092e-01 -2.40438879e-02 -7.64958680e-01 -4.85520571e-01
-6.31389380e-01 -2.51520455e-01 7.30297387e-01 6.14928603e-01
5.05415380e-01 -1.35470474e+00 -9.41143632e-01 -5.10765649e-02
1.60130233e-01 -7.17927456e-01 3.01959306e-01 1.05737782e+00
-2.80904531e-01 2.44663283e-01 -2.01626167e-01 -2.64423430e-01
-7.66481280e-01 6.37763917e-01 2.33128533e-01 2.19616130e-01
-4.34443772e-01 1.09507918e+00 4.78042096e-01 -3.10005873e-01
-3.31052989e-01 -5.32687485e-01 -3.76258865e-02 6.41875744e-01
1.96701095e-01 3.12581152e-01 -3.77071053e-01 -1.01384473e+00
-2.80076832e-01 4.19417113e-01 1.44438535e-01 -6.45442486e-01
1.25465524e+00 -7.35547766e-02 -4.64846194e-01 5.18395007e-01
1.21184218e+00 1.46222278e-01 -1.17362952e+00 -1.65131390e-02
-3.23351353e-01 -1.91692218e-01 3.34408641e-01 -1.05209315e+00
-1.32080936e+00 9.44458187e-01 5.29390752e-01 2.57170588e-01
7.47261286e-01 4.58778977e-01 5.83507359e-01 -2.28827242e-02
-3.23329031e-01 -1.13429832e+00 2.62349308e-01 3.49841416e-01
1.27723932e+00 -1.21588337e+00 -8.63873214e-02 9.19473097e-02
-1.02693975e+00 6.30535841e-01 4.65221137e-01 -1.60984859e-01
6.07434869e-01 -8.37077871e-02 7.64206890e-03 -6.68799698e-01
-5.74153364e-01 -2.30359316e-01 3.79397839e-01 5.20210743e-01
7.19989002e-01 2.87177145e-01 -4.18004751e-01 5.10068297e-01
-5.65075874e-01 -4.52583373e-01 4.49653387e-01 2.24095896e-01
-2.55400807e-01 -7.65581310e-01 -2.89810508e-01 6.60470426e-01
-3.02180260e-01 -6.57715857e-01 -9.83455658e-01 6.29955411e-01
6.45904168e-02 7.13610351e-01 4.28454518e-01 -2.35191867e-01
9.70929489e-02 1.13524966e-01 3.46374601e-01 -4.31809247e-01
-6.05847895e-01 -7.99031928e-02 2.94125050e-01 -4.20899540e-01
-3.09616238e-01 -1.05713654e+00 -1.13441622e+00 -4.16161239e-01
9.18744430e-02 -4.13664341e-01 6.63610756e-01 6.24522567e-01
3.75068873e-01 6.36022925e-01 1.01786526e-02 -6.41157866e-01
-2.33463407e-01 -1.16667128e+00 -5.61672449e-01 8.06357682e-01
1.08466990e-01 -5.60909748e-01 -3.99039567e-01 5.44349067e-02]
|
[11.24045467376709, 0.2080158293247223]
|
41df5ab0-dc66-4940-9bc8-671fbd352fed
|
a-chinese-math-word-problem-solving-system
| null | null |
https://aclanthology.org/2020.rocling-1.21
|
https://aclanthology.org/2020.rocling-1.21.pdf
|
A Chinese Math Word Problem Solving System Based on Linguistic Theory and Non-statistical Approach
| null |
['Hsin-Hung Lin', 'Chia-Ming Lee', 'Chien-yu Lai', 'Chia-Jung Chen', 'Wen-jet Wang']
| null | null | null | null |
rocling-2020-9
|
['math-word-problem-solving', 'math-word-problem-solving', 'math-word-problem-solving']
|
['knowledge-base', 'reasoning', 'time-series']
|
[-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]
|
[-7.423674583435059, 3.553525686264038]
|
5453d730-b784-4f7e-84ea-9f7dbfb770ae
|
cross-lingual-alignment-of-contextual-word
|
1902.09492
| null |
http://arxiv.org/abs/1902.09492v2
|
http://arxiv.org/pdf/1902.09492v2.pdf
|
Cross-Lingual Alignment of Contextual Word Embeddings, with Applications to Zero-shot Dependency Parsing
|
We introduce a novel method for multilingual transfer that utilizes deep
contextual embeddings, pretrained in an unsupervised fashion. While contextual
embeddings have been shown to yield richer representations of meaning compared
to their static counterparts, aligning them poses a challenge due to their
dynamic nature. To this end, we construct context-independent variants of the
original monolingual spaces and utilize their mapping to derive an alignment
for the context-dependent spaces. This mapping readily supports processing of a
target language, improving transfer by context-aware embeddings. Our
experimental results demonstrate the effectiveness of this approach for
zero-shot and few-shot learning of dependency parsing. Specifically, our method
consistently outperforms the previous state-of-the-art on 6 tested languages,
yielding an improvement of 6.8 LAS points on average.
|
['Ori Ram', 'Regina Barzilay', 'Amir Globerson', 'Tal Schuster']
|
2019-02-25
|
cross-lingual-alignment-of-contextual-word-1
|
https://aclanthology.org/N19-1162
|
https://aclanthology.org/N19-1162.pdf
|
naacl-2019-6
|
['cross-lingual-zero-shot-dependency-parsing']
|
['natural-language-processing']
|
[-1.46698800e-03 -6.94220290e-02 -3.28060597e-01 -5.98462880e-01
-1.28673923e+00 -8.02974045e-01 7.57138431e-01 4.03073698e-01
-7.43523836e-01 6.54087663e-01 5.52944958e-01 -5.21996558e-01
4.25388932e-01 -5.82256854e-01 -7.65715897e-01 -4.92050260e-01
-2.03534812e-02 3.57272387e-01 8.52349401e-02 -3.42810392e-01
-5.95518872e-02 2.07302511e-01 -1.27404094e+00 2.69523561e-01
1.01541245e+00 3.93691778e-01 3.38394016e-01 5.47244310e-01
-4.72607791e-01 3.89012963e-01 -3.90350193e-01 -5.26854396e-01
6.11137599e-02 -3.16458046e-01 -7.14667559e-01 -3.68359745e-01
4.83878165e-01 -1.53794363e-01 -2.23425552e-01 9.58633304e-01
3.78407001e-01 2.56180286e-01 4.70101625e-01 -6.82699144e-01
-1.28346384e+00 7.72010684e-01 -2.27852792e-01 3.76434088e-01
3.37583929e-01 -1.51725277e-01 1.45259368e+00 -1.21302557e+00
9.08299446e-01 1.32704568e+00 4.65790629e-01 6.77056611e-01
-1.55270827e+00 -5.42175949e-01 5.07726789e-01 2.04001993e-01
-9.47106063e-01 -5.93406916e-01 4.13431048e-01 -2.56763130e-01
1.39605916e+00 -4.37626332e-01 4.88308668e-01 1.28834820e+00
2.42556602e-01 7.13743567e-01 1.14167118e+00 -8.93631637e-01
2.46789500e-01 6.18216209e-03 4.06562746e-01 5.73597372e-01
2.70415485e-01 -6.40282258e-02 -5.79116762e-01 1.69822395e-01
5.04267395e-01 -1.63587242e-01 5.38588986e-02 -5.14029801e-01
-1.26424015e+00 1.00537086e+00 4.27197158e-01 5.66239715e-01
-8.14099535e-02 2.02295855e-01 4.83803630e-01 5.41523635e-01
7.60152936e-01 4.04992789e-01 -6.35232449e-01 -2.69834936e-01
-5.45263112e-01 1.81823298e-02 6.52644157e-01 1.05935168e+00
8.30065489e-01 -3.19971703e-02 1.58500369e-03 1.03255761e+00
5.26761264e-02 3.78975421e-01 4.20062214e-01 -3.76083463e-01
7.12302506e-01 3.72561544e-01 -1.47522926e-01 -3.29207689e-01
-2.21176341e-01 -1.67799354e-01 -1.46182716e-01 -8.75613242e-02
5.18955767e-01 -2.66247392e-01 -8.70245278e-01 2.15606737e+00
3.08207870e-01 2.94904172e-01 5.41795194e-01 5.57834983e-01
3.19931686e-01 6.87081337e-01 4.97071028e-01 1.33724332e-01
1.48279727e+00 -1.15074277e+00 -8.14098060e-01 -5.65303028e-01
8.35881889e-01 -8.26731861e-01 1.40663576e+00 -1.08662127e-02
-6.57222807e-01 -4.46906894e-01 -1.22553110e+00 -4.86511439e-01
-4.76330519e-01 -3.06660801e-01 7.59542465e-01 5.84748924e-01
-8.91182601e-01 4.88093287e-01 -1.01777267e+00 -6.40636206e-01
2.60378301e-01 4.16562520e-03 -6.45395637e-01 -3.94297987e-01
-1.19541609e+00 9.78195310e-01 2.45513231e-01 -3.35380197e-01
-6.54267013e-01 -8.31954777e-01 -1.34620106e+00 9.68979076e-02
1.72734201e-01 -3.65966946e-01 1.29130316e+00 -6.56408310e-01
-1.49842632e+00 9.41204846e-01 -2.62714893e-01 -3.21706593e-01
9.82398763e-02 -7.63151407e-01 -4.70502913e-01 4.76413704e-02
2.32886642e-01 5.78749955e-01 4.89640683e-01 -9.05606627e-01
-6.48856163e-01 -2.30939731e-01 4.27664787e-01 1.87555254e-01
-6.09703004e-01 2.77688295e-01 -6.15635276e-01 -7.44287848e-01
-1.00761674e-01 -9.40120161e-01 -3.39538515e-01 -2.38792524e-01
7.91245475e-02 -1.27502128e-01 6.33527577e-01 -7.86241770e-01
1.12017286e+00 -2.18218946e+00 3.43683511e-01 -4.91552114e-01
-2.34265909e-01 1.63066477e-01 -5.54748952e-01 6.66608691e-01
3.66296880e-02 2.37378702e-01 -4.20392573e-01 -6.34626687e-01
-2.88743526e-02 5.53346336e-01 -3.59131753e-01 3.25756162e-01
7.12857723e-01 9.31768715e-01 -1.34629583e+00 -4.11703795e-01
1.60823047e-01 7.02380836e-01 -7.77395964e-01 3.60967577e-01
-2.27062538e-01 5.00426352e-01 -2.57319838e-01 6.35112107e-01
4.70811099e-01 2.07401842e-01 7.86876619e-01 -6.23663440e-02
-9.44431946e-02 5.88933825e-01 -7.11838543e-01 2.35774422e+00
-1.04295552e+00 6.58237219e-01 -1.82395965e-01 -8.25146317e-01
9.11513805e-01 4.36345488e-01 3.26641463e-02 -6.37058020e-01
-8.13266709e-02 3.14751476e-01 -5.71983308e-02 -3.37868363e-01
7.17926681e-01 -3.07088107e-01 -5.94813406e-01 7.15610385e-01
6.66652918e-01 9.93788987e-02 1.68327853e-01 1.96451589e-01
1.08126485e+00 4.99244064e-01 5.42424798e-01 -4.40086901e-01
2.37021238e-01 -1.45402625e-01 7.27410674e-01 4.79048610e-01
-3.52221698e-01 4.23380464e-01 5.65005600e-01 -2.77258486e-01
-1.06588507e+00 -1.39983165e+00 -1.40885755e-01 1.60413444e+00
2.73024924e-02 -4.53159511e-01 -7.80062258e-01 -1.11384606e+00
-1.39814824e-01 9.38783705e-01 -6.21890843e-01 6.63651749e-02
-1.12087667e+00 -5.29902458e-01 3.10334891e-01 8.42057526e-01
-6.66920245e-02 -8.81918967e-01 -3.93862307e-01 4.95798290e-01
-1.03064299e-01 -1.53560936e+00 -3.94846588e-01 4.91572052e-01
-8.67819846e-01 -8.51678431e-01 -4.06069070e-01 -1.14819205e+00
5.80073714e-01 2.28303090e-01 1.21942472e+00 -1.73839182e-01
-9.53774527e-02 3.40200633e-01 -6.71319425e-01 -1.98393151e-01
-4.89513427e-01 3.10242623e-01 1.13766909e-01 -2.43644223e-01
5.91320574e-01 -7.31683671e-01 -4.52677578e-01 -3.27323705e-01
-9.09812868e-01 -2.95602471e-01 6.07985020e-01 1.17382812e+00
5.77521622e-01 -7.95339644e-01 7.28699207e-01 -1.30971587e+00
7.04346001e-01 -5.79551697e-01 -4.47210521e-01 4.06156331e-01
-4.74986374e-01 3.83275062e-01 6.90858662e-01 -4.21306431e-01
-1.24036372e+00 -1.65560935e-02 -1.11186793e-02 -8.10712576e-02
-2.29694471e-01 4.89682436e-01 -2.00230762e-01 1.87531561e-01
4.67437655e-01 -1.87869996e-01 -3.13537806e-01 -7.53688276e-01
1.06553829e+00 5.75424552e-01 6.47943139e-01 -1.04903400e+00
7.11952686e-01 2.38185659e-01 -4.86709297e-01 -5.76317549e-01
-9.97671247e-01 -3.82235557e-01 -1.13585508e+00 2.46666685e-01
1.15795600e+00 -1.06515110e+00 2.67619550e-01 3.70182507e-02
-1.37288070e+00 -2.51481742e-01 -1.92539170e-01 5.33362627e-01
-3.44792455e-01 2.06505388e-01 -7.91347802e-01 -3.26025248e-01
-2.76627868e-01 -1.16341913e+00 9.70270038e-01 9.18561500e-03
-1.91030517e-01 -1.35625350e+00 4.52931106e-01 7.74892792e-02
3.39743644e-01 2.18696013e-01 1.39441037e+00 -9.42833006e-01
-4.18998867e-01 2.19227429e-02 -1.63830101e-01 2.72735059e-01
2.47031048e-01 -2.77113408e-01 -1.22897732e+00 -4.15282816e-01
-2.07667455e-01 -3.86393636e-01 7.41145313e-01 -5.26446216e-02
5.00778854e-01 1.36675119e-01 -2.55774558e-01 5.65970540e-01
1.68654561e+00 -6.06476553e-02 1.95177779e-01 5.32634020e-01
6.21492624e-01 4.51293081e-01 7.48020649e-01 2.01194901e-02
5.66499174e-01 5.49235702e-01 1.47553235e-01 1.23835191e-01
-3.27637881e-01 -3.78267169e-01 5.55633128e-01 1.49374735e+00
1.35339633e-01 9.78098214e-02 -9.84769762e-01 1.00601101e+00
-1.77874792e+00 -5.62971056e-01 4.28524226e-01 2.03029776e+00
1.08611941e+00 1.02285869e-01 -3.49726588e-01 -3.98302644e-01
6.38397634e-01 3.97869140e-01 -2.96252757e-01 -9.45269942e-01
4.88600470e-02 9.12239432e-01 2.98012435e-01 6.19673848e-01
-1.13589013e+00 1.64725280e+00 6.56357431e+00 3.09770197e-01
-1.04813552e+00 6.69733703e-01 1.68420598e-01 -7.46038407e-02
-4.91995364e-01 3.31636220e-01 -9.07953322e-01 1.02260470e-01
1.15474916e+00 -2.90587664e-01 3.41332018e-01 9.17016983e-01
-3.23701411e-01 1.20336860e-01 -1.59835184e+00 5.74063599e-01
1.39373705e-01 -1.13340545e+00 1.04352549e-01 -1.68421894e-01
8.47134650e-01 2.41606161e-01 -3.62290964e-02 8.13983798e-01
5.42352259e-01 -8.06164205e-01 6.59197748e-01 3.46267112e-02
9.95037496e-01 -8.94028187e-01 5.46937108e-01 -2.73390338e-02
-1.34003258e+00 6.18486889e-02 -4.79745895e-01 -1.97260499e-01
3.92254263e-01 2.90517747e-01 -7.01535821e-01 6.49780571e-01
3.94452333e-01 7.81637430e-01 -4.54652667e-01 4.33173269e-01
-7.91281044e-01 6.11310840e-01 9.42117348e-02 4.40592729e-02
4.04052645e-01 -1.51465937e-01 4.07395512e-01 1.72255182e+00
2.66397804e-01 -2.07108393e-01 2.06145704e-01 6.02146804e-01
-3.46784264e-01 3.95327717e-01 -8.09217930e-01 -1.26250908e-01
8.00502717e-01 1.23214042e+00 -4.27281082e-01 -3.38330895e-01
-9.71195459e-01 1.05713999e+00 1.10196459e+00 3.44799519e-01
-7.59191930e-01 -6.23112261e-01 1.21837115e+00 -4.87443089e-01
7.17744648e-01 -6.62211895e-01 -1.76871896e-01 -1.56742990e+00
3.81420329e-02 -8.44687521e-01 3.26529473e-01 -1.56432137e-01
-1.43372750e+00 7.52570689e-01 -2.46090274e-02 -1.06937611e+00
-3.48650783e-01 -8.09714258e-01 -6.55457139e-01 8.27780783e-01
-1.76439106e+00 -1.39057994e+00 1.84704348e-01 2.68801868e-01
8.58055770e-01 -1.95380747e-01 1.43303633e+00 2.25256652e-01
-5.58594048e-01 8.15772176e-01 1.76993281e-01 2.73083985e-01
1.06953669e+00 -1.42584598e+00 9.19203460e-01 1.31500220e+00
6.56880677e-01 9.30914640e-01 4.45182741e-01 -4.30359185e-01
-1.64343965e+00 -1.05205739e+00 1.16847408e+00 -5.76923192e-01
1.12781155e+00 -8.27541292e-01 -9.28642809e-01 9.98359442e-01
5.33876181e-01 2.46963084e-01 1.01172268e+00 7.57166266e-01
-9.91582215e-01 1.76960044e-03 -8.23467255e-01 8.06095600e-01
1.17893565e+00 -9.86496449e-01 -1.16147542e+00 -3.12350318e-03
1.22947192e+00 -1.76917881e-01 -1.05005121e+00 9.13992822e-02
4.70630318e-01 -6.94815338e-01 7.94829726e-01 -7.67376482e-01
5.25286138e-01 1.56398803e-01 -4.97103393e-01 -1.63938510e+00
-1.99108943e-01 -3.60080123e-01 -2.24542860e-02 1.37161016e+00
4.95171219e-01 -6.56215608e-01 3.29125822e-01 3.46979618e-01
-3.84413809e-01 -6.70593143e-01 -1.08072639e+00 -8.78865600e-01
6.52581632e-01 -5.93159497e-01 5.59241116e-01 1.26052248e+00
2.03371868e-01 6.36316657e-01 -1.62222549e-01 1.62456214e-01
5.34382939e-01 1.31118923e-01 5.85441530e-01 -9.24634397e-01
-3.33751738e-01 -1.80146426e-01 -3.71673375e-01 -9.91276264e-01
6.53265715e-01 -1.28003263e+00 9.39792469e-02 -1.51192689e+00
1.12018995e-01 -5.62540829e-01 -8.47177863e-01 4.71668512e-01
-4.79372561e-01 8.53962898e-02 3.82257313e-01 -6.12026155e-02
-5.09716272e-01 3.96840721e-01 8.48603547e-01 3.67585011e-02
-6.47342578e-02 -7.82500684e-01 -7.68078864e-01 4.95253354e-01
8.28331828e-01 -5.33974230e-01 -2.48532042e-01 -1.10001194e+00
-4.97573912e-02 -3.32481593e-01 -1.96615517e-01 -8.77159953e-01
-1.17140755e-01 -2.68141329e-02 1.74608324e-02 -2.84535736e-01
3.61456662e-01 -5.13673365e-01 -5.25457501e-01 1.26126572e-01
-3.19963336e-01 2.07478657e-01 4.46807623e-01 6.72737062e-01
-3.21423769e-01 -2.53212899e-01 5.46769023e-01 3.10477912e-02
-1.01083672e+00 3.40717696e-02 -2.74166524e-01 4.49350595e-01
8.66625249e-01 2.46065393e-01 -2.83271670e-01 1.25397623e-01
-6.22805893e-01 -4.94728498e-02 4.83598202e-01 8.56234014e-01
3.14044446e-01 -1.50094545e+00 -5.84855556e-01 2.42489472e-01
4.58839655e-01 -2.97257990e-01 -2.43740212e-02 4.94978994e-01
-2.86400884e-01 3.95909548e-01 -3.00800711e-01 -4.97143507e-01
-1.08290541e+00 5.89475811e-01 -2.11940810e-01 -4.99540329e-01
-7.77689636e-01 6.98374629e-01 2.08943948e-01 -7.91632593e-01
5.64116202e-02 -5.63723743e-01 -1.30277678e-01 1.33308753e-01
5.44983804e-01 -4.70971316e-02 7.21127689e-02 -4.24851328e-01
-4.00063127e-01 5.87703705e-01 -4.37228829e-01 -3.68415773e-01
1.66465044e+00 -9.65774581e-02 -5.47040328e-02 7.81112075e-01
1.28849256e+00 3.21205497e-01 -1.24449778e+00 -5.53707421e-01
5.56319475e-01 -4.77253616e-01 -2.58646816e-01 -5.43502212e-01
-6.29116595e-01 1.13246572e+00 4.25545216e-01 -2.65828729e-01
7.57049680e-01 1.92576215e-01 9.88593400e-01 5.55983782e-01
5.24472952e-01 -1.01641560e+00 2.91263741e-02 1.02099788e+00
5.10143518e-01 -1.26334119e+00 -2.61921644e-01 -3.19747835e-01
-5.36141098e-01 1.20404053e+00 4.87310797e-01 -2.42578760e-01
5.65309584e-01 4.20843124e-01 3.97479892e-01 2.66887724e-01
-9.37376857e-01 -2.75277317e-01 1.26109142e-02 7.38304496e-01
9.50260043e-01 3.04892153e-01 -4.34914887e-01 6.56479776e-01
-6.66712527e-04 -3.48832428e-01 2.85814703e-01 1.25938404e+00
-3.19361657e-01 -1.78637612e+00 -7.85673708e-02 -1.86536819e-01
-4.95937943e-01 -5.15647888e-01 4.08973172e-02 9.48092401e-01
-1.54199511e-01 8.22054625e-01 1.93208307e-01 -1.79498643e-01
3.83980334e-01 5.65573931e-01 6.56953633e-01 -1.17372048e+00
-5.06046891e-01 -9.45297163e-03 3.81227762e-01 -6.02212131e-01
-2.61662990e-01 -7.27184772e-01 -1.29603744e+00 1.15448609e-01
-2.76300889e-02 -1.30660236e-01 7.13571072e-01 9.36002910e-01
3.12982887e-01 5.06106317e-01 4.43848342e-01 -7.84557343e-01
-6.55876160e-01 -9.12424684e-01 -4.42058071e-02 5.86359739e-01
2.07410723e-01 -6.72555149e-01 -9.68365371e-02 8.62340108e-02]
|
[10.697012901306152, 9.7193021774292]
|
f4f11e78-cf87-4522-8cbf-17bb3b89b646
|
galaxy-morphology-classification-using
|
2008.13611
| null |
https://arxiv.org/abs/2008.13611v2
|
https://arxiv.org/pdf/2008.13611v2.pdf
|
Galaxy Morphology Classification using EfficientNet Architectures
|
We study the usage of EfficientNets and their applications to Galaxy Morphology Classification. We explore the usage of EfficientNets into predicting the vote fractions of the 79,975 testing images from the Galaxy Zoo 2 challenge on Kaggle. We evaluate this model using the standard competition metric i.e. rmse score and rank among the top 3 on the public leaderboard with a public score of 0.07765. We propose a fine-tuned architecture using EfficientNetB5 to classify galaxies into seven classes - completely round smooth, in-between smooth, cigarshaped smooth, lenticular, barred spiral, unbarred spiral and irregular. The network along with other popular convolutional networks are used to classify 29,941 galaxy images. Different metrics such as accuracy, recall, precision, F1 score are used to evaluate the performance of the model along with a comparative study of other state of the art convolutional models to determine which one performs the best. We obtain an accuracy of 93.7% on our classification model with an F1 score of 0.8857. EfficientNets can be applied to large scale galaxy classification in future optical space surveys which will provide a large amount of data such as the Large Synoptic Space Telescope.
|
['Pranav Parwate', 'Hrushikesh Pandit', 'Shreyas Kalvankar']
|
2020-08-31
| null | null | null | null |
['morphology-classification']
|
['computer-vision']
|
[-6.29994750e-01 -1.75703347e-01 1.63594574e-01 -2.31800109e-01
-2.30939016e-01 -9.75923955e-01 7.66736031e-01 -2.02527538e-01
-5.26375830e-01 6.70021594e-01 1.50411308e-01 -4.84197438e-01
-4.58262593e-01 -9.32016730e-01 -4.95065451e-01 -6.70901597e-01
-2.98516333e-01 4.93570805e-01 9.14302886e-01 -1.63992271e-01
4.59887683e-01 6.49305880e-01 -1.37240314e+00 5.46701610e-01
5.11418641e-01 1.55681384e+00 1.33703440e-01 1.15066957e+00
3.63911927e-01 8.80740166e-01 -7.13727534e-01 -5.10327637e-01
3.39256346e-01 1.34221539e-01 -1.02610755e+00 -1.38426587e-01
5.22022605e-01 2.70589311e-02 -1.65911913e-01 8.13303888e-01
3.51629496e-01 2.97839314e-01 6.79160535e-01 -8.90089631e-01
-4.61569041e-01 1.67904049e-01 -2.06998244e-01 5.60042262e-01
-4.85060483e-01 3.29061478e-01 1.21368802e+00 -1.06055975e+00
4.93603945e-01 1.28965592e+00 1.04944837e+00 -6.95893317e-02
-7.58224130e-01 -4.19085622e-01 -3.33145231e-01 3.47044230e-01
-1.36701679e+00 -2.41268858e-01 1.33348688e-01 -6.25260592e-01
1.18930817e+00 5.68053365e-01 6.06134295e-01 5.06933093e-01
4.45934713e-01 8.69517699e-02 1.22641015e+00 5.90997972e-02
1.23762205e-01 6.26868987e-03 7.26412758e-02 9.78843749e-01
3.99497181e-01 1.34859324e-01 -2.54100323e-01 -2.01505899e-01
7.02665925e-01 -5.18002808e-01 1.72748212e-02 5.98166026e-02
-1.07296574e+00 1.06228423e+00 1.13895082e+00 1.90838277e-01
-7.50591280e-03 2.51216441e-01 2.55642772e-01 2.19353110e-01
7.12958634e-01 1.24106431e+00 -6.39366746e-01 1.61692247e-01
-6.84950769e-01 5.36277771e-01 7.79477239e-01 5.20965517e-01
6.82666481e-01 2.57771537e-02 -8.60666111e-02 1.09971952e+00
3.08483601e-01 1.30307943e-01 4.50724572e-01 -1.20867097e+00
-4.29765917e-02 9.06017721e-01 1.78806707e-02 -8.84788990e-01
-6.03450000e-01 -1.11794806e+00 -8.31585348e-01 4.09496963e-01
5.38153648e-01 1.22771293e-01 -6.17978930e-01 1.01091981e+00
1.26632154e-01 2.67930895e-01 -1.55282736e-01 8.88767421e-01
1.40046394e+00 5.42415082e-01 -5.67850173e-01 5.05983233e-01
1.44188869e+00 -1.41135681e+00 3.81059676e-01 -1.65591776e-01
4.43506151e-01 -1.03465128e+00 9.66795564e-01 4.72330362e-01
-7.82016933e-01 -6.67173922e-01 -1.10163414e+00 -2.25574821e-02
-6.17981315e-01 4.16608363e-01 8.13437998e-01 5.26248932e-01
-1.30773878e+00 8.09689105e-01 -7.00621784e-01 -3.98736507e-01
5.04527926e-01 7.21757293e-01 -4.47880596e-01 3.16022605e-01
-5.12217343e-01 5.98093808e-01 3.29462111e-01 -2.90921599e-01
-9.21460092e-01 -6.55467272e-01 -2.39171624e-01 1.80106565e-01
1.41018152e-01 -6.30715013e-01 1.48746002e+00 -7.32204139e-01
-1.17224562e+00 1.02460074e+00 4.06293839e-01 -8.38742375e-01
3.06973398e-01 3.89217079e-01 -2.54033715e-01 -1.26468554e-01
5.22554778e-02 8.22348475e-01 1.18175298e-01 -8.14775348e-01
-1.10926557e+00 1.74297560e-02 4.16280478e-01 -1.75607353e-01
-1.69264019e-01 1.07814923e-01 -4.52877760e-01 -4.07157123e-01
-2.09925487e-03 -1.19669986e+00 -1.35093108e-01 -3.47313523e-01
-2.92253405e-01 -5.59007406e-01 6.25270307e-01 -3.69108140e-01
7.49891698e-01 -1.77286661e+00 -3.24732900e-01 9.69232097e-02
4.31964576e-01 3.08824033e-01 -1.69144914e-01 7.73640424e-02
1.48555666e-01 1.45406082e-01 2.08893344e-01 -1.71343729e-01
-2.16663450e-01 -1.37190461e-01 -1.00549296e-01 5.01404583e-01
6.67754039e-02 8.88348579e-01 -7.20293701e-01 1.09419979e-01
2.04592675e-01 2.73384809e-01 -8.62031102e-01 3.19958776e-01
1.19555546e-02 3.72544914e-01 -1.00189939e-01 7.45762646e-01
6.47327006e-01 -5.98735273e-01 -1.59925491e-01 -1.04652502e-01
-5.65827549e-01 5.12798488e-01 -6.49162114e-01 1.12102842e+00
-4.58592176e-01 9.44589674e-01 1.75583344e-02 -4.36911881e-01
1.05767202e+00 -1.43907547e-01 -2.33093992e-01 -3.73683512e-01
3.09689969e-01 5.70602417e-01 5.26088059e-01 -3.53167159e-03
7.96673954e-01 3.17371070e-01 -3.78875956e-02 1.52583113e-02
3.90754819e-01 -2.12425083e-01 3.52250367e-01 5.94986416e-02
1.36064196e+00 -4.00648147e-01 2.75040448e-01 -8.38095188e-01
5.28757393e-01 1.64588578e-02 1.39219776e-01 1.03602505e+00
-1.04229636e-02 8.53549421e-01 7.32585490e-01 -8.89647603e-01
-1.20618331e+00 -7.33362913e-01 -7.54632056e-02 1.15617299e+00
-1.53127640e-01 -3.41256499e-01 -4.17740732e-01 -6.73495829e-01
-1.73541039e-01 2.28178516e-01 -7.63459444e-01 -3.48575488e-02
-3.05251360e-01 -1.19367039e+00 6.12512946e-01 3.90663862e-01
9.57146585e-01 -1.07358348e+00 -1.68494299e-01 -3.18206511e-02
1.67757452e-01 -7.35942066e-01 -1.04648620e-01 7.55634606e-02
-4.34503108e-01 -1.24852693e+00 -2.75426894e-01 -8.19538832e-01
1.44444749e-01 4.35230851e-01 1.56636703e+00 3.90522242e-01
-7.03076646e-02 -2.43034214e-01 -3.14254582e-01 -1.95018128e-01
-2.63865501e-01 6.18483067e-01 -1.71560138e-01 -6.75619096e-02
-1.00570954e-01 -3.60752374e-01 -1.06100488e+00 7.34534621e-01
-4.72204119e-01 -7.97915980e-02 3.93216819e-01 9.31426406e-01
3.38182658e-01 7.22293258e-02 4.58532572e-01 -8.17466319e-01
3.44394922e-01 -6.29087746e-01 -9.68719959e-01 -3.27886111e-04
-5.15707493e-01 -3.21586072e-01 7.23598897e-01 -2.40219474e-01
-6.95445180e-01 -2.67485261e-01 -5.37288129e-01 -2.48073936e-01
5.82639463e-02 2.31797487e-01 3.15315247e-01 -7.12361813e-01
8.66663516e-01 -1.84340820e-01 -4.03805375e-01 -6.70325875e-01
1.13234945e-01 7.36830711e-01 8.81489098e-01 -2.62089074e-01
5.43429554e-01 4.67658430e-01 1.26710951e-01 -7.37574399e-01
-1.08370984e+00 -8.24371338e-01 -3.73267144e-01 -2.37483129e-01
6.39317155e-01 -8.67133021e-01 -1.03198409e+00 6.62162304e-01
-1.01672232e+00 -4.20302302e-01 1.94676337e-03 4.35659915e-01
-2.10279375e-01 -1.10399403e-01 -6.33246422e-01 -3.90864938e-01
-5.61260462e-01 -1.06720650e+00 9.67204213e-01 3.57859582e-01
-1.07575051e-01 -1.00961137e+00 1.81446642e-01 3.48334521e-01
7.03605771e-01 -5.26798703e-03 7.70823300e-01 -9.29060638e-01
-7.49206722e-01 -2.10642949e-01 -5.82506597e-01 2.98792392e-01
-1.97224781e-01 2.37699784e-02 -1.15354216e+00 -4.94188070e-01
-3.66650105e-01 -5.25624454e-01 1.40787578e+00 3.35036337e-01
1.56345379e+00 -2.67228514e-01 4.12481688e-02 1.08901346e+00
1.36138570e+00 2.34760314e-01 6.87840998e-01 6.70607805e-01
6.29365206e-01 4.12764698e-01 4.43180561e-01 1.62967965e-01
2.67775267e-01 5.92406273e-01 9.37860191e-01 3.94246355e-02
-3.46031249e-01 9.76620242e-03 -1.65880218e-01 8.41769040e-01
-2.82120198e-01 -5.87681532e-01 -1.05256295e+00 4.42751944e-01
-1.55639923e+00 -6.34583592e-01 -4.44729567e-01 2.17301965e+00
1.12092845e-01 3.63660842e-01 1.11589268e-01 -1.22900315e-01
4.65619326e-01 2.11852342e-01 -1.48221478e-01 -3.67264509e-01
-2.52918720e-01 5.17316103e-01 8.76906395e-01 4.10810679e-01
-1.43470943e+00 1.03920794e+00 6.48177195e+00 1.19632518e+00
-1.31474197e+00 1.97839826e-01 9.97721016e-01 -6.37653172e-02
2.57916838e-01 3.92417097e-03 -8.47954988e-01 4.00130451e-01
1.03469408e+00 1.97862089e-01 5.16925991e-01 9.43520069e-01
-5.16869985e-02 -8.78864340e-03 -3.80719662e-01 6.70195937e-01
-1.83454484e-01 -1.87035120e+00 -2.09374890e-01 3.07418346e-01
1.04191864e+00 8.01265836e-01 1.55976817e-01 5.26566267e-01
6.25564814e-01 -1.24592233e+00 6.37710273e-01 3.43362749e-01
8.68420064e-01 -7.21493423e-01 1.01714361e+00 1.63837641e-01
-1.19191158e+00 -1.74420640e-01 -9.02501285e-01 -2.08786353e-01
-6.76863372e-01 2.48379812e-01 -1.34425449e+00 4.24785823e-01
9.80701506e-01 5.42795718e-01 -1.21042383e+00 1.60744202e+00
1.49950385e-01 9.73799944e-01 -1.74335599e-01 -3.15472633e-01
7.07389534e-01 7.71128833e-02 4.27439362e-01 1.19981945e+00
7.05697417e-01 -3.44281822e-01 -1.43028751e-01 5.62565982e-01
-4.15268660e-01 1.43730327e-01 -3.97529572e-01 2.31430382e-01
3.20640445e-01 1.88772559e+00 -1.25014317e+00 -3.98561090e-01
-3.87924165e-01 3.12235892e-01 4.65674132e-01 -2.20003575e-01
-5.92140079e-01 -3.95543784e-01 5.08293450e-01 2.51343161e-01
5.89950204e-01 -1.73596486e-01 -1.92705423e-01 -7.81651080e-01
-7.04293251e-01 -7.22649395e-01 3.69901836e-01 -1.01414275e+00
-1.38773251e+00 1.01754022e+00 -3.50046456e-01 -1.26695454e+00
1.47054717e-01 -1.25027180e+00 -1.09270740e+00 6.51070178e-01
-9.75613773e-01 -1.24444044e+00 -7.53184736e-01 3.55163328e-02
6.04180574e-01 -1.05987227e+00 4.20335531e-01 1.38094217e-01
-2.44642317e-01 3.11413169e-01 5.03746748e-01 1.21049467e-04
5.87680936e-01 -1.63495564e+00 8.24107409e-01 4.47906077e-01
1.63823470e-01 3.01874131e-01 4.38706607e-01 -4.21186447e-01
-7.72856236e-01 -1.61635625e+00 6.52689636e-01 -3.28246981e-01
9.05680954e-01 -3.22172642e-01 -8.22047532e-01 3.88996422e-01
1.67717740e-01 3.57775778e-01 4.14598793e-01 2.51780957e-01
-3.10877860e-01 -2.78422594e-01 -8.70534837e-01 2.38718539e-01
8.30784142e-01 -8.25612918e-02 -1.11226134e-01 4.62229848e-01
5.60056150e-01 -3.73845488e-01 -7.93668211e-01 4.97141033e-01
6.61976039e-01 -1.25250649e+00 9.08682764e-01 -7.03294218e-01
4.84955072e-01 -3.18689138e-01 -1.39734685e-01 -1.58254004e+00
-9.12637591e-01 -1.08006857e-01 4.67611670e-01 7.76219428e-01
4.98257667e-01 -7.96491086e-01 1.01345384e+00 -5.19151092e-01
-5.72307527e-01 -8.89152288e-01 -9.31408823e-01 -7.90880620e-01
2.77085751e-01 -1.19414829e-01 7.68608928e-01 4.80736107e-01
-8.95625651e-01 5.70948660e-01 -2.88027704e-01 -1.97519511e-02
2.17331693e-01 2.81100065e-01 9.70265865e-01 -1.50333881e+00
-4.42316681e-01 -7.16273010e-01 -6.40362859e-01 -8.37205172e-01
-2.57187840e-02 -1.21132123e+00 -1.52985990e-01 -1.19722402e+00
2.90212661e-01 -6.23852432e-01 -4.12342370e-01 5.16430549e-02
1.98109582e-01 1.00233364e+00 -2.10503861e-01 2.34697282e-01
-1.00857031e+00 1.38918132e-01 1.21541274e+00 -8.79187509e-02
2.74894774e-01 3.00537378e-01 -6.44851804e-01 9.24387515e-01
9.99608576e-01 -2.95551121e-01 -1.32996634e-01 -2.81130224e-01
3.48498315e-01 -4.20229614e-01 5.76487601e-01 -1.32006180e+00
3.75299193e-02 2.53682602e-02 5.86289287e-01 -5.88603079e-01
4.17675018e-01 -8.76158774e-02 1.31981254e-01 4.20132875e-01
-6.49506301e-02 1.40975460e-01 -2.85376366e-02 2.74230421e-01
-7.23667070e-02 -4.27636594e-01 1.09536254e+00 -4.55546916e-01
-6.10254586e-01 3.93338025e-01 -3.61226261e-01 -7.00634941e-02
5.44210672e-01 2.62947619e-01 -8.35683167e-01 -1.22063592e-01
-4.53125983e-01 -3.31881493e-02 5.00575423e-01 4.26050454e-01
1.86068997e-01 -1.27266920e+00 -6.90633535e-01 1.30650684e-01
1.68759838e-01 -2.49481738e-01 2.23330278e-02 6.48311675e-01
-1.17128456e+00 7.60875285e-01 -3.32179368e-01 -7.10397482e-01
-1.43800116e+00 -1.48521915e-01 8.52513850e-01 -6.17977560e-01
5.32388650e-02 1.26958168e+00 6.16437733e-01 -8.33165109e-01
-2.42752954e-01 -3.33801925e-01 -4.31512684e-01 -1.90968379e-01
3.07468474e-01 7.00276673e-01 6.52957857e-01 -6.02359831e-01
-3.35632503e-01 2.37044707e-01 1.58583984e-01 2.20871061e-01
1.32428098e+00 4.08387601e-01 -4.14423555e-01 4.37005669e-01
1.11351883e+00 2.05103830e-01 -1.03059828e+00 -5.88996895e-02
3.61023210e-02 -6.66066229e-01 1.36389896e-01 -9.00388360e-01
-1.03100193e+00 6.82458639e-01 6.32791996e-01 6.11135721e-01
6.29121661e-01 3.02503824e-01 3.21360946e-01 5.93715191e-01
1.79752931e-01 -6.63685501e-01 2.78364688e-01 1.03859580e+00
1.10585964e+00 -1.25783646e+00 -2.77137589e-02 -2.23581374e-01
-3.00732106e-01 1.13671744e+00 8.07424486e-01 -5.82995832e-01
5.26483059e-01 -2.41399985e-02 -2.19575211e-01 -3.89950871e-01
-1.33562672e+00 -2.59559184e-01 7.47310340e-01 1.12097844e-01
5.28964400e-01 3.00079376e-01 -2.52603054e-01 5.73135257e-01
-8.86616528e-01 -5.85741818e-01 4.21726823e-01 2.75759995e-01
-7.61721253e-01 -6.86164260e-01 -3.66828799e-01 9.69512463e-01
-7.60032475e-01 7.35390484e-02 -7.53922343e-01 7.74520755e-01
4.87625808e-01 9.13436294e-01 5.19779921e-01 -6.28912866e-01
1.36955023e-01 -3.91431779e-01 2.48589471e-01 -6.51544988e-01
-1.02633476e+00 5.14883064e-02 5.54037690e-01 -6.88486546e-02
-4.09469716e-02 -2.96819001e-01 -5.53760767e-01 -7.27177799e-01
-3.59704494e-01 2.89825350e-01 6.51528716e-01 5.44689119e-01
1.40105560e-01 4.08937067e-01 7.36088812e-01 -9.62100804e-01
-2.05323324e-01 -1.45559084e+00 -5.06704211e-01 -1.69223384e-03
1.99137673e-01 -7.45771646e-01 -5.22280753e-01 -4.83639896e-01]
|
[7.967302322387695, 2.961817741394043]
|
652fa91c-c3ae-48e5-9b2c-074b6178da80
|
adversarial-image-alignment-and-interpolation
|
1707.00067
| null |
http://arxiv.org/abs/1707.00067v1
|
http://arxiv.org/pdf/1707.00067v1.pdf
|
Adversarial Image Alignment and Interpolation
|
Volumetric (3d) images are acquired for many scientific and biomedical
purposes using imaging methods such as serial section microscopy, CT scans, and
MRI. A frequent step in the analysis and reconstruction of such data is the
alignment and registration of images that were acquired in succession along a
spatial or temporal dimension. For example, in serial section electron
microscopy, individual 2d sections are imaged via electron microscopy and then
must be aligned to one another in order to produce a coherent 3d volume. State
of the art approaches find image correspondences derived from patch matching
and invariant feature detectors, and then solve optimization problems that
rigidly or elastically deform series of images into an aligned volume. Here we
show how fully convolutional neural networks trained with an adversarial loss
function can be used for two tasks: (1) synthesis of missing or damaged image
data from adjacent sections, and (2) fine-scale alignment of block-face
electron microscopy data. Finally, we show how these two capabilities can be
combined in order to produce artificial isotropic volumes from anisotropic
image volumes using a super-resolution adversarial alignment and interpolation
approach.
|
['Viren Jain']
|
2017-06-30
| null | null | null | null |
['patch-matching']
|
['computer-vision']
|
[ 6.42936707e-01 3.27605195e-02 5.39096653e-01 -4.83137280e-01
-7.81371772e-01 -4.57664073e-01 4.55515325e-01 7.76304305e-02
-7.91703045e-01 7.05775678e-01 -3.20392966e-01 4.43272926e-02
8.26275535e-03 -6.30360425e-01 -8.35179269e-01 -8.25096071e-01
-6.17566369e-02 1.05183399e+00 3.42727810e-01 7.84829110e-02
4.48935151e-01 1.11453724e+00 -1.09673190e+00 1.30131975e-01
2.70359457e-01 7.50031769e-01 3.59751374e-01 7.18523741e-01
-1.19436510e-01 9.75556299e-02 -1.31820783e-01 -7.49858394e-02
4.24643159e-01 -5.77799261e-01 -1.21785724e+00 2.28144899e-01
4.87086684e-01 -5.02373993e-01 1.08755946e-01 8.72673810e-01
4.22502041e-01 1.85286738e-02 8.60173821e-01 -6.63392246e-01
-6.36497319e-01 2.13411469e-02 -6.35416090e-01 2.55092382e-01
2.07493961e-01 -5.34450114e-02 3.32757264e-01 -9.78715539e-01
1.14927197e+00 1.05814433e+00 6.24075472e-01 7.56339133e-01
-2.08883905e+00 -3.40409994e-01 -4.74790126e-01 -7.97741488e-02
-9.25466478e-01 -5.01825690e-01 9.02373850e-01 -7.82867849e-01
7.45084703e-01 2.38794133e-01 6.31338477e-01 5.38813949e-01
5.76383710e-01 1.08157761e-01 1.44182467e+00 -4.81483608e-01
7.87875578e-02 -2.85409272e-01 -2.76935637e-01 5.70644379e-01
-1.32125050e-01 -6.57208189e-02 -5.47656566e-02 -1.98675990e-01
1.34737766e+00 1.05731390e-01 -2.55771458e-01 -4.25085664e-01
-1.59321940e+00 5.60877562e-01 3.57172847e-01 6.50345981e-01
-3.91404480e-01 -4.13076244e-02 2.46033892e-01 3.91194284e-01
4.49452281e-01 8.10317039e-01 -8.48508999e-02 6.11905158e-01
-1.01822054e+00 3.09301764e-01 3.14814389e-01 3.37859958e-01
9.39177096e-01 -1.93514466e-01 3.11089814e-01 6.35516524e-01
1.30289599e-01 1.89037368e-01 6.50044203e-01 -1.28640127e+00
-6.00810759e-02 4.57381606e-01 1.13265319e-02 -6.97089612e-01
-5.35060406e-01 4.06675577e-01 -9.60687935e-01 6.09811246e-01
6.99094296e-01 3.64209682e-01 -9.46142733e-01 1.51904202e+00
6.31554067e-01 -2.02727672e-02 -2.19730079e-01 8.47503662e-01
5.20047188e-01 2.81227112e-01 -1.55257851e-01 -4.08294529e-01
1.21753764e+00 -2.70254701e-01 -4.28081512e-01 -1.84364598e-02
1.92844078e-01 -9.18337464e-01 7.01289654e-01 -1.29141122e-01
-1.60791457e+00 -2.83805400e-01 -9.70476389e-01 -4.36443478e-01
-1.32828325e-01 -4.05796647e-01 1.69297438e-02 -1.52750090e-01
-1.01300991e+00 9.95402038e-01 -1.16094899e+00 -1.58708423e-01
5.20727515e-01 6.18513167e-01 -9.65023100e-01 1.91328570e-01
-4.67622191e-01 8.85432601e-01 -1.29703775e-01 4.69079092e-02
-7.30491221e-01 -9.86095607e-01 -6.79660857e-01 -3.49762201e-01
-4.07018512e-01 -7.25069225e-01 6.82522595e-01 -7.34353065e-01
-1.10326111e+00 1.71919084e+00 -1.93141490e-01 -3.45393658e-01
4.75008458e-01 3.99584949e-01 7.92199373e-02 3.45747054e-01
2.22612426e-01 6.88863695e-01 8.33232105e-01 -1.33713806e+00
-5.32463659e-03 -1.02540576e+00 -3.74597520e-01 -5.35233766e-02
3.38369995e-01 2.79938519e-01 1.16529606e-01 -4.89978582e-01
6.56673908e-01 -7.43097782e-01 -3.18484575e-01 4.51356232e-01
-3.24178070e-01 3.79669547e-01 1.10603905e+00 -1.09807646e+00
2.66578168e-01 -1.79442251e+00 7.71321893e-01 2.75914490e-01
5.15546501e-01 -1.32790864e-01 -9.15160030e-02 1.81655183e-01
-2.66389489e-01 2.07752921e-03 -6.06268883e-01 -4.66127574e-01
-4.45851833e-01 8.94385800e-02 7.27293193e-02 9.89361286e-01
4.12951559e-02 9.94443417e-01 -6.44455791e-01 -7.27076709e-01
2.10613862e-01 7.74963558e-01 -4.42779034e-01 4.13750619e-01
1.19778529e-01 1.44604766e+00 -1.16605528e-01 3.18647176e-01
6.28195167e-01 -1.90420255e-01 2.51271188e-01 -4.44320738e-01
-1.34142175e-01 -6.20432682e-02 -7.19822466e-01 1.72603178e+00
-3.88038427e-01 6.13215983e-01 5.35835445e-01 -1.31889367e+00
7.00388193e-01 4.19960380e-01 9.11049843e-01 -6.97418511e-01
1.48007736e-01 4.47185785e-01 -9.68482643e-02 -3.28975320e-01
2.95198802e-02 -7.17462718e-01 2.98063785e-01 7.62093067e-01
8.22261348e-02 -7.40319431e-01 -4.64357585e-02 -2.48347089e-01
8.62633467e-01 -1.06201552e-01 9.28844661e-02 -3.31349552e-01
7.24658549e-01 -9.33708549e-02 2.81543761e-01 2.46446282e-02
-4.29038145e-02 1.03932011e+00 3.39629501e-01 -9.33178067e-01
-2.12567568e+00 -1.10608697e+00 -7.44232297e-01 1.07825816e-01
-7.81616196e-02 6.52043700e-01 -8.94444466e-01 -2.75213808e-01
6.34291768e-03 -5.52971996e-02 -7.75332630e-01 1.74212709e-01
-1.06389213e+00 -6.09065711e-01 2.07510397e-01 2.35765219e-01
1.82699665e-01 -1.23475444e+00 -6.82573080e-01 4.63230520e-01
-1.29664555e-01 -1.06593883e+00 -4.81117934e-01 1.72547668e-01
-1.06831157e+00 -1.16516888e+00 -9.58099067e-01 -1.16949368e+00
1.31408072e+00 -7.58558065e-02 1.03389835e+00 3.42592418e-01
-6.37807667e-01 2.71672428e-01 1.83013722e-01 5.16929664e-02
-8.52446973e-01 -3.10514927e-01 3.23127836e-01 3.99734862e-02
-1.86052576e-01 -1.32386947e+00 -7.83891201e-01 4.95680898e-01
-1.18797565e+00 2.03160546e-03 3.09344143e-01 9.92428839e-01
1.38348663e+00 -3.43421012e-01 3.20644528e-01 -8.86067986e-01
3.95251542e-01 4.22831327e-02 -8.14246714e-01 1.05884925e-01
-4.08769101e-01 5.32849431e-02 8.02217662e-01 -3.45551223e-01
-5.61635375e-01 8.56716484e-02 -4.28369075e-01 -5.11122644e-01
-1.67069331e-01 6.07675388e-02 1.13987006e-01 -7.42633343e-01
6.85934603e-01 4.54186797e-01 7.59493470e-01 -2.25990891e-01
-1.08914971e-01 3.06706309e-01 8.54204834e-01 -5.17248869e-01
8.48202288e-01 1.16102040e+00 3.84196043e-01 -7.18302786e-01
-2.41034701e-01 -2.57240355e-01 -1.40587211e+00 -9.18973386e-02
1.16991019e+00 -2.81162322e-01 -5.54636121e-01 3.78880769e-01
-1.14559388e+00 -4.89218265e-01 -4.88204777e-01 3.97844821e-01
-1.01339257e+00 4.72107649e-01 -7.92806983e-01 -2.02058285e-01
-5.40509284e-01 -1.44945514e+00 1.24656439e+00 -9.45530981e-02
-3.81531447e-01 -1.03730488e+00 2.43172035e-01 2.91312814e-01
3.68449181e-01 5.83111286e-01 1.20672083e+00 -1.78256378e-01
-4.25403565e-01 -2.63876796e-01 3.42702158e-02 3.20835054e-01
2.26286054e-01 -7.28495270e-02 -6.37302041e-01 -4.57158595e-01
3.98692816e-01 -3.51932168e-01 3.82414490e-01 6.96083784e-01
1.22142255e+00 -3.56121123e-01 -3.33771527e-01 9.61779952e-01
1.48436117e+00 3.01522195e-01 9.28022444e-01 3.44566345e-01
7.08813906e-01 9.32258606e-01 -3.96527536e-02 -1.57740712e-01
-2.39009894e-02 9.26068127e-01 2.89092243e-01 -2.06829354e-01
-1.55447260e-01 2.05550283e-01 -3.47157329e-01 6.23548448e-01
-4.52866435e-01 6.25686586e-01 -7.11467683e-01 6.43212795e-01
-1.13235652e+00 -1.25004101e+00 -9.26343426e-02 2.36906576e+00
1.00074959e+00 -2.64424473e-01 -2.92255562e-02 1.67342335e-01
7.10449219e-01 -1.88492581e-01 -7.41346180e-01 -4.07168716e-01
6.52897917e-03 3.03344786e-01 3.55793834e-01 6.06757283e-01
-9.42566633e-01 3.29105794e-01 6.76184082e+00 3.14546824e-01
-1.27380800e+00 1.46496579e-01 8.35433960e-01 8.22943971e-02
-4.34070319e-01 -9.34174433e-02 -3.86325061e-01 3.70783210e-01
7.02689409e-01 -7.74688795e-02 6.14932835e-01 1.98030233e-01
1.11134917e-01 3.49141546e-02 -1.32772350e+00 8.56807113e-01
-1.40385166e-01 -1.72804177e+00 -2.97315349e-03 4.77578253e-01
8.12203705e-01 2.25797575e-02 -5.23448735e-02 -6.15718663e-01
1.35566771e-01 -1.13617992e+00 4.48913574e-01 6.07800424e-01
1.03993440e+00 -4.31504697e-01 4.49619651e-01 2.57436305e-01
-9.35295999e-01 4.95017648e-01 -4.71711725e-01 5.39752662e-01
5.66913009e-01 4.07680511e-01 -5.92430234e-01 2.08482459e-01
7.10165203e-01 2.11088374e-01 -4.86389026e-02 6.31522000e-01
3.06291133e-01 -2.12441981e-01 -1.90248072e-01 5.97329855e-01
-1.07636563e-01 -6.76487446e-01 6.58552110e-01 5.78635991e-01
2.49923721e-01 1.03870519e-01 -2.54669100e-01 1.31416357e+00
-1.24257565e-01 -9.78110954e-02 -7.59181917e-01 2.13738322e-01
1.15100734e-01 1.48239028e+00 -1.03503013e+00 -4.14840877e-02
-3.23826522e-01 1.02356589e+00 4.45824474e-01 2.03361705e-01
-4.27371830e-01 -6.92088157e-02 5.06074011e-01 7.29802907e-01
1.27028078e-01 -3.25998306e-01 -1.85250089e-01 -9.39344764e-01
8.54839683e-02 -5.21499217e-01 -7.52037540e-02 -8.62784803e-01
-1.29364681e+00 6.84621930e-01 -2.71619856e-01 -9.72843528e-01
-2.60483295e-01 -4.07296896e-01 -6.65624559e-01 1.09883904e+00
-1.09733427e+00 -9.76009190e-01 -9.35570598e-02 5.15292048e-01
1.82076305e-01 8.34209621e-02 1.01017320e+00 2.31589988e-01
-1.69810113e-02 1.16768003e-01 3.16907823e-01 2.70553291e-01
5.25291324e-01 -1.22883046e+00 4.14756417e-01 6.41793847e-01
-1.14787430e-01 5.25299311e-01 6.74860060e-01 -5.75977683e-01
-1.15371418e+00 -1.00451291e+00 8.30511451e-01 -4.80401278e-01
3.75505179e-01 -1.05818525e-01 -1.29777038e+00 8.92353952e-01
-3.89447138e-02 5.63529015e-01 4.55769926e-01 -7.22735584e-01
-3.23282927e-02 1.77699491e-01 -1.84402990e+00 5.81537306e-01
6.69817924e-01 -6.58809841e-01 -4.50504363e-01 3.73207182e-01
3.12191606e-01 -5.66832304e-01 -1.46895564e+00 1.28569633e-01
8.94315839e-01 -8.64504457e-01 1.29245245e+00 -6.57602131e-01
8.03458631e-01 -2.43335262e-01 4.31131944e-02 -1.29976892e+00
-2.48580396e-01 -4.19329882e-01 5.48440933e-01 7.72020936e-01
8.32580030e-02 -7.08047748e-01 8.07391405e-01 7.77209759e-01
-1.79613218e-01 -7.71085858e-01 -1.41536808e+00 -6.96755290e-01
4.69085217e-01 3.90846550e-01 5.05696654e-01 8.94632638e-01
-2.37041891e-01 5.99088147e-02 1.22923255e-01 -8.58857185e-02
9.41516340e-01 2.88137943e-01 5.74426472e-01 -1.22628927e+00
4.89564911e-02 -3.89778197e-01 -8.10928881e-01 -6.49105012e-01
2.59153634e-01 -1.00681555e+00 7.95966312e-02 -1.41368711e+00
2.11818174e-01 -5.23859203e-01 2.96931535e-01 2.90610287e-02
3.07227105e-01 6.53180182e-01 -3.07540685e-01 7.91450202e-01
6.06096536e-02 2.49783263e-01 1.73244798e+00 -7.21615553e-02
-4.22330610e-02 -2.48825461e-01 -2.62923568e-01 5.65933466e-01
4.84988153e-01 -5.53058147e-01 1.11561865e-01 -3.95781219e-01
-9.57618728e-02 3.29964280e-01 6.39794648e-01 -8.41763318e-01
2.34489366e-01 1.27841115e-01 7.32546568e-01 -3.34605962e-01
3.65525752e-01 -1.04206693e+00 5.92199147e-01 3.94189209e-01
-2.69887090e-01 3.75874758e-01 -2.91346628e-02 2.67055869e-01
-1.48199424e-01 -2.28003889e-01 1.33851957e+00 -5.13551652e-01
-5.73675632e-02 5.09036422e-01 -1.92177027e-01 -1.23510450e-01
1.16119659e+00 -3.47948909e-01 -5.98363094e-02 8.18185101e-04
-1.00907671e+00 -1.58124432e-01 1.02057111e+00 -3.33612144e-01
8.61108661e-01 -1.47327912e+00 -7.56460130e-01 4.66923833e-01
-3.47973108e-01 5.16504169e-01 4.17345971e-01 1.23703957e+00
-1.05416703e+00 1.09075025e-01 -8.32921863e-01 -8.25205147e-01
-1.30218816e+00 4.62556869e-01 7.70206511e-01 -5.16938925e-01
-7.49942064e-01 7.29268134e-01 3.18203211e-01 -7.52849102e-01
-4.82435942e-01 3.10459528e-02 -4.30940092e-02 -4.25991952e-01
5.98443329e-01 3.74597237e-02 3.13010484e-01 -1.01373506e+00
-2.32539311e-01 1.06135595e+00 -7.60322586e-02 -1.59657627e-01
1.78871226e+00 -2.50604093e-01 -6.06928468e-01 2.57123649e-01
1.48455024e+00 -1.42544746e-01 -1.32605755e+00 -1.54047012e-01
-3.97229135e-01 -3.24463367e-01 -9.08727795e-02 -9.29766893e-02
-1.28002703e+00 7.71061540e-01 4.32878405e-01 2.74818152e-01
9.02016640e-01 2.50296205e-01 8.73594224e-01 -1.41418606e-01
4.13478255e-01 -7.48842776e-01 -1.27316877e-01 3.61225791e-02
1.12212527e+00 -1.19597912e+00 1.75692096e-01 -2.16368556e-01
8.84756371e-02 1.24843311e+00 2.21910492e-01 -5.58090985e-01
5.81352949e-01 5.16955674e-01 -2.38301121e-02 -5.45044243e-01
-4.13424760e-01 3.59337211e-01 2.87665427e-01 7.65586019e-01
5.54336190e-01 -3.47744882e-01 -3.47044408e-01 -1.51058156e-02
4.79515176e-03 -2.04674318e-01 3.83989841e-01 8.41149092e-01
-7.94519857e-02 -1.18308032e+00 -5.95687270e-01 5.49005389e-01
-7.59859145e-01 2.22573742e-01 -2.11574867e-01 7.13007987e-01
-2.64649689e-01 6.61243126e-02 5.21243155e-01 -2.71725561e-02
3.48209530e-01 -1.02679348e-02 1.07004166e+00 -4.91771489e-01
-3.65042359e-01 1.01948462e-01 -4.70535219e-01 -5.09871423e-01
-6.83736324e-01 -9.00927961e-01 -1.40041173e+00 -3.60627115e-01
1.04516661e-02 -1.09830111e-01 8.70980859e-01 9.85815287e-01
3.60480770e-02 4.44573492e-01 6.27929807e-01 -1.55015111e+00
-2.08557680e-01 -7.13774383e-01 -7.61065006e-01 7.81102180e-01
3.04190397e-01 -5.46347022e-01 -5.86955726e-01 6.20146453e-01]
|
[13.687034606933594, -2.730955123901367]
|
06bc4c8a-dfa4-4b14-b308-b3cf5bef7477
|
argument-novelty-and-validity-assessment-via
| null | null |
https://aclanthology.org/2022.argmining-1.10
|
https://aclanthology.org/2022.argmining-1.10.pdf
|
Argument Novelty and Validity Assessment via Multitask and Transfer Learning
|
An argument is a constellation of premises reasoning towards a certain conclusion. The automatic generation of conclusions is becoming a very prominent task, raising the need for automatic measures to assess the quality of these generated conclusions. The SharedTask at the 9th Workshop on Argument Mining proposes a new task to assess the novelty and validity of a conclusion given a set of premises. In this paper, we present a multitask learning approach that transfers the knowledge learned from the natural language inference task to the tasks at hand. Evaluation results indicate the importance of both knowledge transfer and joint learning, placing our approach in the fifth place with strong results compared to baselines.
|
['Maja Stahl', 'Milad Alshomary']
| null | null | null | null |
argmining-acl-2022-10
|
['argument-mining']
|
['natural-language-processing']
|
[ 2.09458604e-01 7.40761280e-01 -4.40992713e-01 -3.92771572e-01
-1.19246519e+00 -5.69028080e-01 1.43499768e+00 7.96523809e-01
-3.48474711e-01 1.07201731e+00 5.95998764e-01 -6.82846487e-01
-4.04767662e-01 -7.14591384e-01 -9.50740933e-01 -3.63037288e-01
2.28127271e-01 6.02326632e-01 3.25843871e-01 -2.05905467e-01
7.09959745e-01 -4.45261151e-02 -1.43449426e+00 1.10526490e+00
9.05856013e-01 8.82879972e-01 -2.25593731e-01 5.08950889e-01
-4.59540367e-01 1.23927414e+00 -6.62793159e-01 -1.06355536e+00
-1.65256023e-01 -1.87440127e-01 -1.44999480e+00 -4.00686800e-01
5.34308314e-01 1.08907782e-02 5.92682481e-01 8.35566163e-01
1.75381482e-01 -5.81943393e-02 8.03756058e-01 -1.31750619e+00
-3.53726506e-01 1.30364227e+00 -1.12247162e-01 4.93039191e-01
4.60057020e-01 -2.27328122e-01 1.23858273e+00 -8.09158981e-01
6.69123709e-01 1.46633482e+00 7.35295713e-01 -1.72426794e-02
-1.24163604e+00 -3.97941768e-01 4.25395340e-01 5.35763562e-01
-5.24592400e-01 -6.23713732e-01 9.25617278e-01 -7.00437188e-01
1.10294271e+00 -8.11524987e-02 1.65118396e-01 1.36747003e+00
2.42930472e-01 6.78396702e-01 1.47070217e+00 -8.16961884e-01
4.09028113e-01 3.61156732e-01 2.26760298e-01 4.34774578e-01
5.00271440e-01 -1.77599669e-01 -6.95139229e-01 -5.37694454e-01
-3.42522524e-02 -6.51384294e-01 3.22094411e-01 -2.49998242e-01
-1.28584981e+00 9.81834292e-01 1.58180773e-01 5.78794599e-01
-5.90193033e-01 3.06469556e-02 8.45177531e-01 3.76751751e-01
8.08642745e-01 5.17754257e-01 -6.31724417e-01 -7.93063417e-02
-7.19417453e-01 7.93433428e-01 1.00148392e+00 4.23901469e-01
3.98656726e-01 -4.98567581e-01 -4.63562071e-01 5.05995631e-01
1.79757968e-01 5.96328825e-02 2.29283482e-01 -1.15592456e+00
8.19784880e-01 7.79575169e-01 2.90697604e-01 -8.69614422e-01
-1.53960943e-01 -3.56097102e-01 -2.06440061e-01 4.43802774e-01
6.99150741e-01 -4.14321542e-01 -1.92947969e-01 1.55577064e+00
4.78908718e-01 -1.64084285e-01 5.10818899e-01 3.21220100e-01
8.10093760e-01 3.33423704e-01 2.28368744e-01 -1.22064188e-01
1.07439256e+00 -7.16548085e-01 -8.12587619e-01 -7.82097727e-02
7.11664081e-01 -9.08511162e-01 7.23991632e-01 5.41148841e-01
-1.08759296e+00 -5.12988567e-01 -1.03719735e+00 -1.45324856e-01
-6.00123048e-01 -1.17935658e-01 5.63837886e-01 1.91288173e-01
-5.86089253e-01 6.16921425e-01 -3.69557478e-02 4.48222011e-02
3.40745568e-01 -5.78005016e-02 -1.06863089e-01 1.28884122e-01
-1.43132544e+00 1.38819551e+00 8.65200460e-01 -1.24147363e-01
-2.68244624e-01 -6.94005728e-01 -7.19198942e-01 -1.86740249e-01
6.76603436e-01 -6.84765458e-01 1.37130153e+00 -8.04965258e-01
-1.11571300e+00 1.04681325e+00 -5.82678616e-02 -7.63711452e-01
8.49092662e-01 -3.90328109e-01 -4.42456007e-01 -2.62846440e-01
7.11658478e-01 2.96707481e-01 7.99950361e-01 -1.28418326e+00
-1.14870501e+00 -2.01406717e-01 2.65725762e-01 -1.07307695e-02
1.88734874e-01 2.59377718e-01 5.78323424e-01 -2.47957557e-01
-3.22417915e-01 -4.99696463e-01 1.59440666e-01 -4.09023881e-01
-4.57491517e-01 -1.34179282e+00 7.98721731e-01 -5.30549049e-01
9.31813776e-01 -1.77985013e+00 -4.09816653e-02 6.98226094e-02
7.97513127e-02 4.55254495e-01 3.10149938e-01 4.95907336e-01
-9.45651084e-02 2.25222364e-01 -5.76220490e-02 -2.99380254e-02
1.91274554e-01 1.06220745e-01 -6.36966288e-01 -1.09896816e-01
5.08175671e-01 8.96033049e-01 -1.08421278e+00 -7.52130508e-01
1.02073029e-01 6.44337237e-02 -1.53223902e-01 -2.50870772e-02
-6.06736183e-01 -1.23974830e-02 -6.16503060e-01 3.13103199e-02
2.47455925e-01 -2.41564646e-01 3.35744977e-01 -1.49835676e-01
-2.67766684e-01 1.08107615e+00 -1.02170944e+00 1.45411992e+00
-4.13890511e-01 3.11058134e-01 -2.75128037e-01 -1.45226431e+00
8.86945188e-01 6.91454828e-01 -1.58430666e-01 -6.92183912e-01
1.10421509e-01 5.61119914e-01 2.69004226e-01 -7.89332449e-01
5.96512184e-02 -4.64161754e-01 -2.40131676e-01 6.14954710e-01
8.29879660e-03 -9.29735154e-02 4.08091366e-01 2.72662938e-01
6.14789307e-01 4.17399645e-01 5.86659670e-01 -5.52105248e-01
6.47409201e-01 1.96690112e-01 4.59150404e-01 8.88322949e-01
1.44190922e-01 -4.83201817e-04 8.73450041e-01 -8.61245394e-01
-8.71209621e-01 -8.85240853e-01 -9.69279334e-02 9.69039798e-01
-5.03237605e-01 -2.45402604e-01 -4.09400284e-01 -1.14595675e+00
2.89639831e-01 1.17447400e+00 -8.58328223e-01 1.86197221e-01
-5.63230693e-01 -1.91061080e-01 3.84672821e-01 3.61705363e-01
6.21012211e-01 -1.32510519e+00 -9.27439809e-01 2.99594551e-01
-8.07540834e-01 -1.20305455e+00 3.41219604e-01 3.19516540e-01
-7.58027196e-01 -1.40469909e+00 -6.81236312e-02 -7.76743531e-01
8.21385756e-02 -4.03051406e-01 1.19982338e+00 1.44258756e-02
1.79801479e-01 -1.82631649e-02 -2.88988203e-01 -9.99051690e-01
-9.17464674e-01 7.77681172e-02 -3.07226539e-01 -1.65241763e-01
4.13215160e-01 -3.07771415e-01 1.14846841e-01 -3.23245794e-01
-5.28210759e-01 2.57979423e-01 5.31949401e-01 6.66514516e-01
2.01869339e-01 -8.42731893e-02 1.06572676e+00 -1.17800725e+00
1.24268293e+00 -4.77461874e-01 -2.67680049e-01 6.49272323e-01
-7.89216638e-01 5.87467015e-01 5.58399320e-01 -4.33380231e-02
-1.52663934e+00 -6.02398276e-01 -7.59183802e-03 3.46140772e-01
-2.95709729e-01 7.90351391e-01 2.75928900e-02 6.81220651e-01
9.66469944e-01 -4.57403153e-01 1.81275249e-01 -2.55611837e-01
6.04452491e-01 4.49226201e-01 3.75920177e-01 -1.19467950e+00
5.59191346e-01 3.21917713e-01 5.47693074e-02 -4.27236587e-01
-1.64030862e+00 -1.14313692e-01 -6.03778243e-01 -1.67765945e-01
4.83557165e-01 -6.00933552e-01 -5.57401419e-01 -1.44807696e-01
-1.32744646e+00 -5.04028618e-01 -4.57844555e-01 3.57295156e-01
-5.92165709e-01 3.62518281e-01 -1.34549484e-01 -6.77820146e-01
-4.65775609e-01 -6.89225137e-01 6.24137878e-01 -3.10612857e-01
-7.20918298e-01 -1.36565220e+00 1.39841378e-01 6.29219055e-01
1.53793350e-01 5.21963775e-01 1.41496646e+00 -1.02460682e+00
1.61549449e-01 -3.28476653e-02 -3.45790863e-01 4.64078546e-01
2.25097403e-01 7.99336955e-02 -8.80934894e-01 1.84690118e-01
1.03573844e-01 -6.69518411e-01 8.91489208e-01 2.33605132e-01
6.92562819e-01 -5.18390775e-01 -3.90541255e-01 -4.54687774e-01
9.30837333e-01 -2.09857941e-01 2.47581616e-01 7.58576274e-01
-6.37284294e-03 1.01583922e+00 8.37500572e-01 1.73566677e-02
5.06455183e-01 4.79759246e-01 1.77910224e-01 1.31688699e-01
-2.82529831e-01 -3.66604835e-01 8.03414918e-03 1.90754503e-01
-2.35750169e-01 1.05184525e-01 -1.03232050e+00 7.20497310e-01
-2.17220211e+00 -1.35401356e+00 -4.13998365e-01 1.97139955e+00
9.59124684e-01 8.17697704e-01 9.45867971e-02 4.65495199e-01
7.17027187e-01 -7.93974102e-02 -1.10940166e-01 -1.00914741e+00
-9.42496210e-02 3.06665450e-01 -2.36201450e-01 7.74704576e-01
-9.67099071e-01 7.70270050e-01 6.22731590e+00 4.47906673e-01
-7.36969709e-01 2.62014065e-02 5.15135288e-01 3.00786734e-01
-5.75879276e-01 3.01293790e-01 -7.40399241e-01 1.80382892e-01
7.35352278e-01 -4.56411153e-01 -3.92179370e-01 6.70829892e-01
1.86168760e-01 -1.93369150e-01 -1.25522900e+00 1.88966528e-01
6.14490137e-02 -1.59910989e+00 2.63076603e-01 -1.15726469e-03
6.46392167e-01 -8.22472647e-02 4.02621925e-04 4.17703360e-01
4.93627131e-01 -6.03066683e-01 1.03239083e+00 3.80571783e-01
2.17592061e-01 -6.66992426e-01 9.62919652e-01 5.26883483e-01
-4.33302760e-01 -7.91264027e-02 1.22165643e-02 -7.49434948e-01
6.96452558e-02 7.24685729e-01 -1.22250271e+00 6.19597614e-01
4.88010496e-01 4.89167362e-01 -4.88546968e-01 6.66927636e-01
-9.11553264e-01 7.11260796e-01 -1.81618452e-01 -3.47233713e-01
3.54615152e-01 1.63023382e-01 3.75268042e-01 1.32912481e+00
-2.73622666e-02 -2.08693057e-01 1.35511726e-01 9.22530293e-01
-1.28169343e-01 1.60782367e-01 -7.33345151e-01 3.34041893e-01
2.46243358e-01 1.15707862e+00 -5.14204621e-01 -7.10541725e-01
-2.51359969e-01 3.70739400e-01 6.93086326e-01 -3.12277321e-02
-5.71818113e-01 -5.50988019e-02 2.28925243e-01 -4.88978811e-02
2.93627590e-01 2.35513255e-01 -4.73781943e-01 -7.86623538e-01
3.22502375e-01 -7.58867860e-01 6.53677464e-01 -5.32050312e-01
-1.45840573e+00 1.89302355e-01 3.49978894e-01 -6.28941357e-01
-5.07018030e-01 -4.62444186e-01 -1.03820467e+00 6.90771699e-01
-1.85458982e+00 -1.04527819e+00 1.34585738e-01 2.08680451e-01
5.67189813e-01 1.91171486e-02 8.99221659e-01 -3.98522705e-01
-3.51668000e-02 1.46158919e-01 -3.47061157e-01 1.21060893e-01
8.82293046e-01 -1.37498605e+00 5.26963651e-01 7.58126557e-01
1.23925693e-01 5.94560683e-01 9.15770888e-01 -6.85682893e-01
-6.29639804e-01 -6.80549622e-01 1.83615589e+00 -6.84582710e-01
9.28330421e-01 -8.28281343e-02 -9.25126612e-01 6.29146993e-01
7.14718580e-01 -3.58434558e-01 7.69167066e-01 8.59615982e-01
-7.17452228e-01 3.34340893e-02 -1.07213163e+00 4.10720468e-01
7.11158931e-01 -4.79810834e-01 -1.65959728e+00 3.54386598e-01
3.30633879e-01 -9.68190655e-02 -7.20952511e-01 4.75126892e-01
5.38251877e-01 -9.73458946e-01 6.45240843e-01 -8.34817052e-01
9.20184612e-01 -2.17477411e-01 1.39655292e-01 -1.17506063e+00
-2.63999969e-01 -3.74804854e-01 -7.17235059e-02 1.29299486e+00
1.06240702e+00 -6.00724697e-01 4.14211750e-01 5.93657017e-01
6.22877851e-03 -4.84385550e-01 -9.04918313e-01 -5.63504577e-01
5.59224010e-01 -3.72456610e-01 2.34183773e-01 1.01944304e+00
5.37374258e-01 9.96732771e-01 8.61898512e-02 -2.48175517e-01
7.99430013e-01 5.96146226e-01 7.60439575e-01 -1.71090269e+00
1.95365459e-01 -5.51258683e-01 1.07905939e-01 -2.63872027e-01
4.46384460e-01 -1.13158011e+00 -2.06553668e-01 -2.16516066e+00
1.31023526e-01 -1.74513996e-01 -2.69566387e-01 4.18325067e-01
-4.63768184e-01 -4.52335268e-01 3.68739702e-02 -1.57791257e-01
-7.62177348e-01 1.36508062e-01 9.11728978e-01 6.51369914e-02
1.73516393e-01 1.13850467e-01 -9.04879272e-01 1.01391852e+00
1.06373394e+00 -4.89102364e-01 -2.56241322e-01 -3.58689487e-01
7.42141306e-01 -3.15588415e-01 5.00233710e-01 -7.36598074e-01
4.06045139e-01 -8.30087513e-02 2.29337662e-01 -5.45138955e-01
1.40640326e-02 -4.01978135e-01 -3.60146344e-01 5.32257140e-01
-8.43014121e-01 1.83127254e-01 1.34306431e-01 5.01688838e-01
-2.32356131e-01 -2.49000475e-01 2.37138718e-01 -2.16114298e-01
-5.29467940e-01 -5.67878604e-01 -3.42748046e-01 5.29661059e-01
7.73969412e-01 -7.49120936e-02 -5.49646556e-01 -9.30957496e-03
-6.93308473e-01 2.18739256e-01 -1.60431877e-01 6.29785657e-01
5.79739571e-01 -8.47183108e-01 -1.39361572e+00 -3.66868705e-01
2.53631562e-01 -2.54239738e-02 -3.69235903e-01 7.96237350e-01
1.37496516e-01 6.18516505e-01 -5.13350815e-02 -2.63359040e-01
-1.16960764e+00 6.99846745e-01 -6.44934922e-03 -5.43672562e-01
-7.69068241e-01 5.47448039e-01 -3.78174931e-01 -3.33065420e-01
5.09167127e-02 -5.31944871e-01 -4.66396153e-01 3.65422308e-01
4.10732150e-01 4.04813737e-01 2.00082615e-01 -1.47755399e-01
-7.42804334e-02 9.66491252e-02 -3.37486148e-01 -1.65419921e-01
1.31356668e+00 3.92333508e-01 -2.39293635e-01 8.19718838e-01
7.27115154e-01 -2.81020887e-02 -7.96223402e-01 -5.31044602e-01
8.25353563e-01 -2.41065957e-03 -1.41103208e-01 -1.14744258e+00
1.93734318e-02 6.60294652e-01 -1.73550695e-01 4.58398461e-01
2.02149987e-01 3.40074986e-01 3.10687959e-01 5.79979777e-01
2.08843812e-01 -1.43604612e+00 4.70363759e-02 6.24689341e-01
1.17574835e+00 -1.34720409e+00 7.87340403e-02 -3.79432082e-01
-5.83805799e-01 1.25338411e+00 2.13979125e-01 -1.69169843e-01
4.76349115e-01 2.95965254e-01 1.35474473e-01 -4.37598884e-01
-1.10355592e+00 -4.77143899e-02 2.04454228e-01 4.64489639e-01
6.73642933e-01 -3.03084683e-02 -6.60130084e-01 3.28355610e-01
-2.61049658e-01 4.05044332e-02 2.08806336e-01 9.04446661e-01
-4.80839998e-01 -1.25017118e+00 -1.57217517e-01 5.13159037e-01
-5.87261856e-01 4.16643173e-02 -9.50273871e-01 8.02179813e-01
1.81850612e-01 1.38263440e+00 -2.81210750e-01 2.20776618e-01
4.89475489e-01 5.72293520e-01 4.88714218e-01 -6.75388813e-01
-9.13213253e-01 -4.47822422e-01 8.23896170e-01 -8.13587084e-02
-8.52787614e-01 -9.35834408e-01 -1.16303110e+00 -1.33430466e-01
-3.62653315e-01 6.45549297e-01 6.52820170e-01 1.57995272e+00
1.50484756e-01 4.74764645e-01 1.34021014e-01 -2.21792310e-02
-7.62617826e-01 -1.04283154e+00 1.68620244e-01 7.71132231e-01
1.47328392e-01 -6.69145525e-01 -3.69652033e-01 2.58839428e-02]
|
[9.581153869628906, 9.58338737487793]
|
d7e24f4b-be59-47c2-a8a8-949316b16dcf
|
ptde-personalized-training-with-distillated
|
2210.08872
| null |
https://arxiv.org/abs/2210.08872v1
|
https://arxiv.org/pdf/2210.08872v1.pdf
|
PTDE: Personalized Training with Distillated Execution for Multi-Agent Reinforcement Learning
|
Centralized Training with Decentralized Execution (CTDE) has been a very popular paradigm for multi-agent reinforcement learning. One of its main features is making full use of the global information to learn a better joint $Q$-function or centralized critic. In this paper, we in turn explore how to leverage the global information to directly learn a better individual $Q$-function or individual actor. We find that applying the same global information to all agents indiscriminately is not enough for good performance, and thus propose to specify the global information for each agent to obtain agent-specific global information for better performance. Furthermore, we distill such agent-specific global information into the agent's local information, which is used during decentralized execution without too much performance degradation. We call this new paradigm Personalized Training with Distillated Execution (PTDE). PTDE can be easily combined with many state-of-the-art algorithms to further improve their performance, which is verified in both SMAC and Google Research Football scenarios.
|
['Hongxing Chang', 'Bin Wang', 'Dong Li', 'Jianye Hao', 'Bin Zhang', 'Shiguang Wu', 'Tianle Zhang', 'Hangyu Mao', 'Yiqun Chen']
|
2022-10-17
| null | null | null | null |
['smac-1', 'smac']
|
['playing-games', 'playing-games']
|
[-6.08325839e-01 -1.28288642e-01 -5.16897500e-01 -2.25256667e-01
-8.07328284e-01 -4.16599005e-01 3.96199077e-01 6.91332370e-02
-7.38604844e-01 1.06845427e+00 -5.46576902e-02 -1.77855000e-01
-1.99718848e-01 -8.17068696e-01 -6.38105690e-01 -9.31208491e-01
-2.47262776e-01 7.13205218e-01 3.23464334e-01 -5.35030782e-01
6.24397993e-02 1.46847591e-01 -1.31077397e+00 4.84906957e-02
1.04881644e+00 6.74981356e-01 3.09914410e-01 6.25447810e-01
2.10825518e-01 9.77387130e-01 -9.60851550e-01 -1.15809754e-01
5.45283735e-01 -4.50560719e-01 -6.00445628e-01 -5.14220409e-02
-2.52335846e-01 -7.68144190e-01 -7.70503655e-02 9.83248830e-01
6.41584694e-01 2.51131952e-01 2.64822304e-01 -1.37091625e+00
-1.65585261e-02 1.02221417e+00 -6.79254413e-01 -4.07434534e-03
5.74311465e-02 7.06966996e-01 9.89358366e-01 -8.67438409e-03
4.22496736e-01 1.32146716e+00 1.97577983e-01 4.33957160e-01
-8.99430633e-01 -6.79250479e-01 6.58629775e-01 2.10414782e-01
-1.00107598e+00 -1.21131621e-01 7.67030716e-01 -1.29523024e-01
8.34021330e-01 -3.26503851e-02 7.99612105e-01 8.51613700e-01
2.56569207e-01 1.16821182e+00 1.20673120e+00 -2.63622940e-01
5.79480112e-01 -3.52318101e-02 -3.23199779e-01 7.96038866e-01
1.69641271e-01 3.84637535e-01 -3.08521211e-01 -1.46385819e-01
8.28741670e-01 6.29356429e-02 -5.95740862e-02 -3.60330254e-01
-1.27336860e+00 8.87209952e-01 6.03413343e-01 2.03712076e-01
-5.91996312e-01 5.81779778e-01 6.16211951e-01 7.01198339e-01
2.46105507e-01 6.08466506e-01 -5.48693299e-01 -5.20868719e-01
-5.00095427e-01 6.35246336e-01 7.99192846e-01 5.87654948e-01
1.16171134e+00 4.54867661e-01 -1.53550759e-01 6.85763121e-01
2.29357272e-01 4.25321907e-01 6.69363260e-01 -1.37827766e+00
6.69207335e-01 6.12224698e-01 3.65660071e-01 -6.76633000e-01
-5.08649886e-01 -5.37383139e-01 -5.46867490e-01 6.41723335e-01
4.01805997e-01 -8.51262093e-01 -4.48330730e-01 2.00353742e+00
4.61151272e-01 1.33179426e-01 3.58573705e-01 1.00696468e+00
7.17357472e-02 5.07905245e-01 -6.89316690e-02 -1.97530419e-01
1.01303089e+00 -1.34244800e+00 -3.94662052e-01 -2.74982721e-01
7.57486284e-01 -2.05698863e-01 8.89580965e-01 4.14474696e-01
-9.18427110e-01 -2.74738729e-01 -1.16520405e+00 6.56609535e-01
-6.79696575e-02 -5.09622172e-02 7.44607091e-01 6.07068717e-01
-9.52487826e-01 8.19391251e-01 -1.14494050e+00 1.13066904e-01
1.94939718e-01 5.49027324e-01 8.25760216e-02 2.10777104e-01
-1.15265214e+00 1.03207231e+00 5.87314546e-01 -2.51298696e-01
-1.42827976e+00 -3.17069322e-01 -5.57649970e-01 -4.86394614e-02
1.03139448e+00 -8.44167292e-01 1.70537353e+00 -1.12540483e+00
-2.40743160e+00 -9.65544768e-03 2.79122114e-01 -4.88696396e-01
6.18587911e-01 -1.40417576e-01 1.27372324e-01 6.74465001e-02
4.44033407e-02 4.56813544e-01 9.13749278e-01 -1.23589432e+00
-8.96310210e-01 -3.41968745e-01 6.40879273e-01 6.49044394e-01
-3.03570777e-01 -2.55650014e-01 -2.13528499e-01 -5.79160631e-01
-4.51839954e-01 -1.01355755e+00 -5.82533896e-01 -4.52486455e-01
-1.56343058e-01 -5.71601212e-01 6.75992668e-01 -3.22164893e-01
9.78655457e-01 -1.81503463e+00 3.57166737e-01 2.22624123e-01
2.08856255e-01 2.66765654e-01 -3.77089441e-01 5.99922717e-01
4.81092781e-01 -1.20700493e-01 8.11041966e-02 -3.35145891e-01
3.03749859e-01 4.51317579e-01 6.11500107e-02 3.50607753e-01
-1.45579755e-01 9.82817411e-01 -1.11736298e+00 -2.63621300e-01
1.20331652e-01 2.87433583e-02 -8.89114618e-01 1.70190737e-01
-6.81132257e-01 6.62807047e-01 -1.01238883e+00 3.19071501e-01
3.38234752e-01 -2.33681232e-01 5.78795195e-01 3.06563556e-01
-1.04089744e-01 9.07757282e-02 -1.42726779e+00 1.67114902e+00
-4.88503635e-01 2.17883959e-02 4.46625352e-01 -1.24484086e+00
7.85774827e-01 3.18065286e-01 7.76267052e-01 -1.01857007e+00
7.09600523e-02 2.66934037e-01 1.25521511e-01 -3.74321729e-01
2.90794522e-01 -2.61754915e-02 -1.62523851e-01 8.06307137e-01
-1.11902647e-01 1.07370093e-01 3.97421837e-01 5.86250648e-02
1.20981705e+00 1.51984483e-01 2.62898177e-01 -2.45298460e-01
3.59629035e-01 -1.06737651e-01 1.01218808e+00 1.00089097e+00
-3.04213464e-01 -1.91606045e-01 6.11380935e-01 -4.35290277e-01
-8.24567795e-01 -6.26254141e-01 5.99555612e-01 1.38555551e+00
1.75695270e-01 -4.59877372e-01 -6.63953900e-01 -1.02550495e+00
1.03537083e-01 3.77796382e-01 -3.59674126e-01 -2.42058821e-02
-8.77541780e-01 -7.38409996e-01 3.30374420e-01 3.78374636e-01
8.65557790e-01 -1.14312470e+00 -1.09606230e+00 4.98221189e-01
-1.49233043e-02 -6.07399762e-01 -4.15485710e-01 1.91484556e-01
-6.58166170e-01 -1.00288832e+00 -7.83074677e-01 -2.77341634e-01
2.96981066e-01 1.71777725e-01 8.68206918e-01 1.74181789e-01
3.20769489e-01 3.28713119e-01 -5.36478102e-01 -2.90989548e-01
-4.82741714e-01 2.90751696e-01 1.95714951e-01 -8.99650306e-02
-3.03922921e-01 -6.30191863e-01 -7.17515290e-01 3.70964378e-01
-7.57017493e-01 -5.81729412e-02 7.40065813e-01 9.36932802e-01
1.80047974e-01 2.02506468e-01 9.07456815e-01 -8.38395476e-01
8.80382180e-01 -2.14627877e-01 -8.60765994e-01 2.22796217e-01
-6.48430526e-01 4.65886950e-01 1.03302979e+00 -5.09962440e-01
-1.03948486e+00 -2.68108368e-01 -1.82988614e-01 -3.47606361e-01
2.05695689e-01 7.17068672e-01 -1.36358989e-02 3.96152809e-02
5.21349370e-01 1.59192488e-01 2.91870028e-01 -3.45801532e-01
3.02414447e-01 3.69505554e-01 5.52512221e-02 -1.09405625e+00
4.95907992e-01 1.82801098e-01 -1.33112296e-01 -2.37348050e-01
-5.11813462e-01 -2.28089035e-01 7.49296844e-02 -1.60326660e-01
4.28618789e-01 -1.00271893e+00 -1.42062056e+00 6.29273057e-01
-8.98568869e-01 -9.72476661e-01 -2.67364949e-01 5.46245337e-01
-8.01965654e-01 2.71710426e-01 -6.39591932e-01 -6.93593681e-01
-2.96286613e-01 -1.57380819e+00 6.65013433e-01 4.45125759e-01
4.62465584e-01 -9.54224050e-01 5.15488386e-01 3.61717008e-02
6.20760381e-01 7.95547441e-02 6.08017862e-01 -6.00060225e-01
-6.33640885e-01 1.16975173e-01 1.39831334e-01 3.58616173e-01
-1.19215157e-02 -1.69308022e-01 -4.62788761e-01 -7.85234571e-01
-1.28409073e-01 -5.55592597e-01 5.84886491e-01 1.56496227e-01
8.66300166e-01 -6.27784669e-01 -1.60949230e-01 3.64567608e-01
1.48934436e+00 3.46594870e-01 3.44517440e-01 7.99399018e-01
3.42938393e-01 2.00619072e-01 7.61996031e-01 9.90267456e-01
6.86058044e-01 8.83280754e-01 6.57628417e-01 2.31633335e-01
1.77185610e-01 -2.02144176e-01 8.14030588e-01 7.17739820e-01
-3.67546439e-01 -9.44473818e-02 -7.24421084e-01 2.75983959e-01
-2.40162683e+00 -1.01816607e+00 6.13703310e-01 2.04903460e+00
9.24229622e-01 -9.15137976e-02 5.26610315e-01 -3.42943400e-01
5.42986512e-01 1.77222699e-01 -8.92331183e-01 -2.84208566e-01
1.46031588e-01 4.91899624e-02 5.15305758e-01 4.44911063e-01
-8.08663726e-01 9.04044211e-01 6.07009125e+00 1.02499950e+00
-1.18465412e+00 2.35495567e-01 4.20649379e-01 -5.32624908e-02
-1.20196395e-01 1.37167603e-01 -7.50161231e-01 7.40082204e-01
8.86816621e-01 -3.64359856e-01 8.89262557e-01 1.04780662e+00
3.86900455e-01 -3.51643384e-01 -8.81531298e-01 7.54903376e-01
-4.35972363e-01 -1.26512504e+00 -2.17739582e-01 8.26448128e-02
9.06778932e-01 2.13797018e-01 -1.06752880e-01 8.32341433e-01
9.78988945e-01 -5.97629845e-01 6.56937361e-01 2.18698874e-01
2.18323082e-01 -1.00430512e+00 7.31371880e-01 8.35868716e-01
-1.04972529e+00 -3.90348434e-01 -3.83511961e-01 -1.50484070e-01
3.59196998e-02 3.01877171e-01 -6.90660357e-01 7.75761187e-01
7.02889979e-01 5.86775541e-01 -3.94161522e-01 9.92936373e-01
-3.65175426e-01 3.64756882e-01 -4.62410092e-01 -2.83672154e-01
5.27738035e-01 -2.78974593e-01 4.84496534e-01 6.16039991e-01
1.43844157e-01 -8.91441405e-02 7.66450286e-01 5.49168408e-01
1.99482627e-02 -3.51084322e-02 -2.09189638e-01 6.39132783e-02
4.72772002e-01 1.18592727e+00 -3.33716333e-01 -6.08308077e-01
-2.77268797e-01 6.42303407e-01 6.46391571e-01 4.36639309e-01
-8.93220901e-01 -2.64810830e-01 8.17988515e-01 -3.91424328e-01
3.61607134e-01 -3.35711420e-01 2.41788536e-01 -1.37977517e+00
-9.55513865e-02 -1.27443814e+00 4.05593872e-01 -1.91092566e-01
-1.20178151e+00 4.09482062e-01 -5.80991507e-02 -1.29733253e+00
-7.53895760e-01 -3.04594159e-01 -8.29565465e-01 4.90495980e-01
-1.71893990e+00 -7.95962036e-01 -3.42746004e-02 7.68676996e-01
4.34268206e-01 -3.71234745e-01 7.14158416e-01 2.02055633e-01
-8.49623561e-01 6.70962095e-01 5.02765894e-01 1.31328758e-02
7.20384240e-01 -1.21509254e+00 -1.57217175e-01 5.24699211e-01
-2.14629620e-01 3.80656868e-01 6.52304709e-01 -4.50283259e-01
-1.56909120e+00 -9.39280093e-01 -4.49104384e-02 1.00137085e-01
7.36948431e-01 1.99063927e-01 -4.93586808e-01 7.11488783e-01
4.16847527e-01 -1.07993655e-01 2.00310051e-01 1.60581008e-01
-1.20630264e-01 -5.24606049e-01 -9.06381488e-01 6.63228095e-01
5.50932765e-01 2.02194136e-02 -1.92595378e-01 3.02805036e-01
7.19344020e-01 -4.13154602e-01 -8.48178267e-01 3.96803528e-01
2.72487342e-01 -1.04960430e+00 6.61221445e-01 -5.12921393e-01
2.15020210e-01 -3.98087651e-01 -4.17374307e-03 -1.89622557e+00
-5.92984036e-02 -9.35058236e-01 -1.51298508e-01 8.49921465e-01
2.67494172e-01 -1.06217325e+00 9.38528895e-01 4.46278065e-01
-2.25543156e-01 -8.17241669e-01 -1.05539048e+00 -1.05815697e+00
3.10595036e-01 -1.01574689e-01 7.60382712e-01 6.74796164e-01
3.87024224e-01 2.14776024e-01 -6.24571383e-01 3.11761461e-02
5.43615758e-01 4.09334540e-01 1.17564118e+00 -6.49246633e-01
-8.77504766e-01 -5.42012453e-01 -5.29909208e-02 -1.46576393e+00
2.66433984e-01 -5.71636617e-01 2.02536106e-01 -1.33868420e+00
2.60012120e-01 -8.68214548e-01 -3.76243770e-01 7.41954505e-01
-2.21437871e-01 -2.41192088e-01 6.40738249e-01 1.71046048e-01
-1.21696746e+00 8.57806444e-01 1.59332979e+00 -1.22177243e-01
-4.83572602e-01 1.36024475e-01 -6.19000971e-01 4.65999097e-01
1.13069069e+00 -4.54104394e-01 -4.12305653e-01 -6.05883121e-01
1.43330276e-01 5.05266130e-01 1.46505177e-01 -1.03134537e+00
3.81130904e-01 -6.98308229e-01 -1.90011412e-01 -6.61602393e-02
3.38018417e-01 -6.02994084e-01 -9.27901268e-02 7.46282220e-01
-1.40209496e-01 3.22889477e-01 -3.42039280e-02 5.75532794e-01
-2.56040275e-01 -2.86096871e-01 7.13349104e-01 -4.53817517e-01
-6.91048265e-01 3.42138886e-01 -3.71754318e-01 4.83231917e-02
1.29891229e+00 3.31514060e-01 -6.65753663e-01 -6.09997690e-01
-4.84397441e-01 8.31178784e-01 4.15587932e-01 1.34360284e-01
4.35193568e-01 -1.13783002e+00 -7.45565295e-01 -3.99512798e-02
-2.83633024e-01 -1.18251354e-01 2.81651229e-01 8.51793826e-01
-3.58601421e-01 2.62339532e-01 -3.85303944e-01 -4.92205501e-01
-8.84464800e-01 4.72561687e-01 5.34670472e-01 -7.63566315e-01
-5.50340176e-01 3.68942261e-01 -4.31706160e-02 -4.55869615e-01
4.61584330e-02 -1.56481087e-01 2.47933827e-02 -2.68229276e-01
3.96746337e-01 3.18308473e-01 -2.52910018e-01 -3.27756479e-02
-1.49180785e-01 3.28328639e-01 -2.56067425e-01 -3.45623761e-01
1.42407799e+00 -5.79216843e-03 1.65000692e-01 1.82595432e-01
9.00425375e-01 -1.92897975e-01 -1.76779854e+00 -3.16093951e-01
-2.06801280e-01 -4.23576862e-01 -9.70229283e-02 -7.73640394e-01
-1.46156311e+00 5.43279707e-01 2.95626163e-01 1.91546753e-01
8.92463148e-01 -1.03174247e-01 5.82825422e-01 6.77987635e-01
9.77725565e-01 -1.36194336e+00 5.11781096e-01 6.79105520e-01
6.76431596e-01 -1.35273886e+00 8.05963427e-02 2.48448849e-01
-1.03020561e+00 8.91352534e-01 9.47237968e-01 -4.29615706e-01
3.76926452e-01 1.85278356e-01 -4.00893809e-03 2.75427606e-02
-1.12211490e+00 -3.32971096e-01 -4.16636616e-01 3.94669473e-01
-1.08637132e-01 2.88171232e-01 -2.81438768e-01 5.08958757e-01
3.08634266e-02 -1.75505862e-01 4.50964987e-01 1.23439550e+00
-6.00876212e-01 -1.61969459e+00 -1.15879700e-01 3.54687333e-01
-2.86042213e-01 4.09118295e-01 1.58208385e-01 7.97801375e-01
-4.79365028e-02 8.72432113e-01 -2.77188301e-01 -2.90038526e-01
1.53387204e-01 -1.83575019e-01 4.11793113e-01 -4.97413486e-01
-8.61956656e-01 1.70765579e-01 7.55078765e-03 -7.44248450e-01
-4.87726331e-01 -3.95495474e-01 -1.42249787e+00 -6.29226685e-01
-2.29996711e-01 3.98635268e-01 6.98266268e-01 9.24644709e-01
5.06042182e-01 6.07036173e-01 7.59893417e-01 -9.07491803e-01
-1.05606210e+00 -7.71693945e-01 -4.90747601e-01 1.31268606e-01
3.23935956e-01 -8.15231442e-01 -1.95988581e-01 -5.16619027e-01]
|
[3.7430310249328613, 2.058781385421753]
|
f5655ba5-58b8-4a6a-92e9-1bc82cc58580
|
fast-sparse-classification-for-generalized
|
2202.11389
| null |
https://arxiv.org/abs/2202.11389v2
|
https://arxiv.org/pdf/2202.11389v2.pdf
|
Fast Sparse Classification for Generalized Linear and Additive Models
|
We present fast classification techniques for sparse generalized linear and additive models. These techniques can handle thousands of features and thousands of observations in minutes, even in the presence of many highly correlated features. For fast sparse logistic regression, our computational speed-up over other best-subset search techniques owes to linear and quadratic surrogate cuts for the logistic loss that allow us to efficiently screen features for elimination, as well as use of a priority queue that favors a more uniform exploration of features. As an alternative to the logistic loss, we propose the exponential loss, which permits an analytical solution to the line search at each iteration. Our algorithms are generally 2 to 5 times faster than previous approaches. They produce interpretable models that have accuracy comparable to black box models on challenging datasets.
|
['Cynthia Rudin', 'Margo Seltzer', 'Chudi Zhong', 'Jiachang Liu']
|
2022-02-23
| null | null | null | null |
['additive-models']
|
['methodology']
|
[ 7.91556165e-02 -1.89763576e-01 -3.70522857e-01 -6.82301104e-01
-1.16471910e+00 -3.83383423e-01 3.37058485e-01 3.98824990e-01
-4.61293101e-01 9.57037926e-01 -7.27672055e-02 -3.63500655e-01
-5.12895346e-01 -6.73744977e-01 -5.17604947e-01 -6.08052254e-01
-4.48861003e-01 8.82193148e-01 4.98651601e-02 -1.39312353e-02
1.31832793e-01 4.74404424e-01 -1.47388649e+00 2.01820478e-01
6.98240519e-01 1.11803365e+00 -1.96995199e-01 6.06566012e-01
5.85110746e-02 6.31275773e-01 -2.90526658e-01 -2.12693810e-01
3.81397158e-01 -7.46061131e-02 -3.55854124e-01 -9.91739612e-03
5.30333579e-01 -2.86466002e-01 -6.77738860e-02 5.02713442e-01
2.75265574e-01 2.39027143e-01 8.20036054e-01 -1.77576280e+00
-2.51767427e-01 5.38810968e-01 -9.64598835e-01 3.15482408e-01
4.06011432e-01 -1.67201325e-01 1.13281655e+00 -1.14974391e+00
3.03528875e-01 1.32163954e+00 8.61709595e-01 -7.01244101e-02
-1.62407446e+00 -8.20651472e-01 3.69959265e-01 7.57012963e-02
-1.55224645e+00 -4.84888107e-01 2.25668341e-01 -3.29512596e-01
1.03476238e+00 5.35634756e-01 5.06706059e-01 6.97525799e-01
1.04226664e-01 7.28693366e-01 8.98517668e-01 -4.48839128e-01
2.12620005e-01 5.89449525e-01 5.79872429e-01 7.76308119e-01
6.43445671e-01 1.31251797e-01 -6.67602599e-01 -1.01920378e+00
5.62070012e-01 2.42259309e-01 2.64652725e-02 -5.81077039e-01
-8.90257716e-01 1.19217253e+00 2.70914435e-02 -5.37129402e-01
-3.03232700e-01 2.22942695e-01 3.33872139e-01 4.44524020e-01
5.20738125e-01 4.22525406e-01 -6.40259206e-01 3.60300648e-03
-1.16205287e+00 5.51093817e-01 1.12261546e+00 1.06364107e+00
7.88819432e-01 -7.54171833e-02 9.55317318e-02 8.22185576e-01
1.36908233e-01 4.39400405e-01 4.55787554e-02 -8.91716480e-01
3.78028780e-01 4.35116053e-01 1.75545782e-01 -1.02129436e+00
-5.01748085e-01 -6.22044027e-01 -8.26272547e-01 2.09565237e-01
4.71425384e-01 -1.08103462e-01 -5.60606837e-01 1.34543204e+00
4.54240531e-01 9.35256183e-02 -1.54084116e-01 5.61693728e-01
1.71606153e-01 4.94310826e-01 -3.74590084e-02 -4.33794767e-01
1.21222186e+00 -7.71495163e-01 -3.29926848e-01 -1.34915814e-01
6.42170191e-01 -5.34184396e-01 8.25589478e-01 6.20143652e-01
-1.05641580e+00 -9.01662260e-02 -7.26953268e-01 -5.36961220e-02
-1.32091701e-01 -1.76182345e-01 1.17144811e+00 4.62252647e-01
-8.41264188e-01 4.92791265e-01 -1.02789092e+00 -9.50469598e-02
5.04827499e-01 8.01240981e-01 -4.53078181e-01 -1.58600271e-01
-5.87978959e-01 6.43997788e-01 -1.92532241e-01 -8.45869537e-03
-4.53501076e-01 -1.02796531e+00 -9.24578309e-01 2.50331014e-01
5.26288509e-01 -6.66973948e-01 1.06021917e+00 -7.86009550e-01
-9.30524826e-01 3.63040686e-01 -6.75199151e-01 -5.17776310e-01
3.96374106e-01 -4.72263306e-01 7.50449896e-02 -5.00888862e-02
7.25162625e-02 4.28455442e-01 6.62064791e-01 -8.04360390e-01
-7.31658459e-01 -3.44755441e-01 -4.16512877e-01 1.36522636e-01
-2.79210508e-01 3.23254347e-01 -2.20569938e-01 -5.32905042e-01
1.84800655e-01 -8.39456022e-01 -9.11651194e-01 1.31504655e-01
-2.78433025e-01 -9.72812623e-02 6.03786767e-01 -3.28065783e-01
1.26344192e+00 -1.93757963e+00 -7.75053352e-02 5.86027920e-01
3.52670282e-01 -3.90023679e-01 -5.53721376e-02 3.62778693e-01
-4.84035201e-02 1.45083934e-01 -7.19746649e-02 -4.86386061e-01
-1.65905133e-01 -5.83678484e-02 -3.83965671e-01 7.14604616e-01
2.02014461e-01 5.46608984e-01 -6.57541990e-01 -5.66306233e-01
-1.56868935e-01 2.03475684e-01 -8.89804542e-01 -1.89961895e-01
2.15128601e-01 -1.64120063e-01 -3.68996054e-01 6.69777691e-01
6.18965387e-01 -4.55855936e-01 -2.47966629e-02 1.93521619e-01
-5.10149859e-02 2.32140332e-01 -1.58018649e+00 8.58631790e-01
-5.07449508e-01 4.61040914e-01 1.46597847e-01 -8.16532969e-01
8.76789749e-01 3.46488282e-02 6.58621013e-01 -1.17742747e-01
-3.33261043e-02 2.58909166e-01 -4.24879372e-01 3.64265102e-03
2.58361995e-01 -7.44289383e-02 -1.64946064e-01 5.28244793e-01
-1.43177405e-01 -1.32011902e-03 1.81139007e-01 2.51611054e-01
1.05702400e+00 -1.90357253e-01 5.77656090e-01 -3.86335880e-01
4.35575694e-02 2.53946126e-01 6.38395846e-01 1.02649975e+00
4.54257876e-01 4.48474199e-01 6.95611477e-01 -8.17751110e-01
-8.91143084e-01 -9.31570113e-01 -2.17523217e-01 1.14541161e+00
-2.39160299e-01 -5.82487166e-01 -3.39739323e-02 -5.08824527e-01
5.08186281e-01 4.56659853e-01 -5.20529747e-01 8.48907698e-03
-5.20080507e-01 -1.11745059e+00 1.56890750e-01 5.90724170e-01
-2.97025234e-01 -4.01433855e-01 -3.81088793e-01 2.45875135e-01
2.97624826e-01 -8.14799488e-01 -3.83224159e-01 8.51812124e-01
-1.08792329e+00 -1.01653409e+00 -1.82759747e-01 -4.87960279e-01
8.50461781e-01 2.89414227e-01 1.04962754e+00 -2.48426255e-02
-5.93333006e-01 1.37730762e-02 -1.09045342e-01 -4.47761595e-01
1.06683716e-01 1.13297604e-01 2.85635650e-01 -2.54097432e-01
6.11257613e-01 -5.93369842e-01 -4.02479231e-01 2.85752654e-01
-4.10110623e-01 -1.77659899e-01 5.38220763e-01 1.15066242e+00
8.25789094e-01 -1.01922028e-01 4.52998579e-01 -1.05561626e+00
4.97220993e-01 -9.71544743e-01 -1.00364029e+00 1.56031385e-01
-9.32241380e-01 1.81222618e-01 6.82823539e-01 -6.10431075e-01
-6.24869764e-01 5.34990013e-01 2.58124769e-01 -5.04818916e-01
2.25540936e-01 4.15040255e-01 2.79800028e-01 -2.30175763e-01
7.71257639e-01 -8.26322362e-02 3.91928256e-02 -6.27702773e-01
3.11273754e-01 6.31111264e-01 2.45657027e-01 -5.75225770e-01
7.72287130e-01 3.95676970e-01 2.65660286e-01 -8.04905117e-01
-7.15280414e-01 -6.16480529e-01 -4.07545418e-01 5.46623886e-01
1.63235337e-01 -1.00521851e+00 -7.32661784e-01 -9.17079672e-02
-6.32635176e-01 -1.72803745e-01 -4.05267090e-01 6.21199787e-01
-5.21341681e-01 2.03253210e-01 -4.57337677e-01 -1.12702405e+00
-2.12344855e-01 -7.58918524e-01 1.10725629e+00 -1.15289338e-01
-5.78955054e-01 -7.60228992e-01 -5.19377999e-02 4.47287001e-02
3.25179547e-01 2.03254804e-01 8.48021686e-01 -8.45857680e-01
-5.44657409e-01 -7.22064793e-01 -1.93145007e-01 6.41280040e-02
-1.05990566e-01 1.78882480e-01 -7.18037307e-01 -4.79606390e-01
-2.20627174e-01 -4.90809530e-01 1.03432453e+00 6.04091048e-01
1.28393769e+00 -5.84508181e-01 -6.84219003e-01 9.20000434e-01
1.21200037e+00 -1.03099585e-01 2.68005073e-01 2.45864466e-01
3.15894276e-01 5.36787152e-01 8.27220380e-01 8.99247885e-01
3.31349343e-01 6.11890554e-01 3.89058478e-02 -4.27296996e-01
4.89586473e-01 -4.77308519e-02 1.98044702e-01 3.41570467e-01
1.67348981e-01 -5.88082448e-02 -7.13986337e-01 4.70973939e-01
-1.87653589e+00 -9.46988761e-01 -4.10467803e-01 2.55011296e+00
6.48231983e-01 1.13812841e-01 4.30323869e-01 -1.11723868e-02
2.97752231e-01 -2.27191746e-01 -5.02708793e-01 -5.98117054e-01
-9.50226709e-02 2.29373157e-01 8.26751232e-01 7.12350905e-01
-1.10518467e+00 7.04375148e-01 8.24499798e+00 8.50892186e-01
-7.88317740e-01 -1.54509172e-01 8.61861765e-01 -7.53531516e-01
-1.96770012e-01 1.15380131e-01 -1.26452112e+00 2.82223523e-01
9.03266251e-01 -4.10379767e-01 5.47682643e-01 1.20265365e+00
2.50214159e-01 -3.26431632e-01 -1.19211090e+00 8.95875692e-01
-9.15766135e-02 -1.09493756e+00 -1.98358312e-01 2.39154994e-01
6.62343562e-01 9.57861170e-02 5.36725037e-02 1.67961076e-01
7.01645613e-01 -1.14465833e+00 4.94601667e-01 2.22673431e-01
8.83470595e-01 -8.22942019e-01 4.84753579e-01 3.88837367e-01
-1.07112062e+00 -4.28537846e-01 -6.01309240e-01 -2.54320145e-01
-2.10102387e-02 9.62845445e-01 -9.62302387e-01 9.74262357e-02
7.07144976e-01 3.41474801e-01 -3.97226274e-01 1.24827039e+00
3.02681714e-01 6.39120102e-01 -1.11824286e+00 -1.58550113e-01
9.47762281e-03 -3.96886170e-02 3.72944117e-01 1.29063034e+00
3.50862145e-01 -2.51157898e-02 3.54681790e-01 5.04929185e-01
4.03385401e-01 3.52444530e-01 -5.38051784e-01 2.09436908e-01
6.85868323e-01 1.21660233e+00 -5.77860653e-01 -2.39454478e-01
-5.43369710e-01 4.87112105e-01 5.37709177e-01 3.36208314e-01
-6.20448709e-01 -3.85373771e-01 5.47404647e-01 2.35087574e-01
3.55508000e-01 -3.07550460e-01 -3.92609417e-01 -1.01283312e+00
1.14392720e-01 -9.77420270e-01 7.51501620e-01 -2.31304690e-01
-1.39794707e+00 6.05946362e-01 2.31473982e-01 -9.57346618e-01
-5.35839856e-01 -4.30208534e-01 -3.96318585e-01 8.16505671e-01
-1.29270124e+00 -7.62510896e-01 -4.38807383e-02 5.02756119e-01
3.46921593e-01 -6.36531040e-02 9.26330090e-01 1.46740541e-01
-5.22152543e-01 8.10750008e-01 3.43577117e-01 -4.27219480e-01
7.06590533e-01 -1.18123674e+00 3.61676008e-01 4.16568875e-01
1.09495796e-01 8.01393092e-01 8.32319856e-01 -5.78180492e-01
-1.29581511e+00 -8.35698366e-01 8.64996433e-01 -3.62480909e-01
6.38130546e-01 -4.60352689e-01 -8.64094079e-01 8.45937014e-01
-4.98647451e-01 2.86872417e-01 1.03744602e+00 7.91280031e-01
-4.38257754e-01 -2.94167161e-01 -1.08699000e+00 2.94083148e-01
7.57484257e-01 -3.16952402e-03 -1.16694607e-01 5.68225563e-01
4.10370529e-01 -1.14835054e-01 -6.67839587e-01 2.31684908e-01
8.76602173e-01 -5.92335105e-01 9.10546601e-01 -9.34938490e-01
5.98819219e-02 -1.00786269e-01 -2.96355724e-01 -9.23718512e-01
-5.38828731e-01 -8.99789035e-01 -1.56650871e-01 7.10640252e-01
8.78035605e-01 -6.56138420e-01 9.08931077e-01 8.25170040e-01
3.65619481e-01 -1.21166766e+00 -8.60416770e-01 -7.88765311e-01
-3.03606272e-01 -4.75421369e-01 5.78501225e-01 7.97887266e-01
9.72300768e-02 1.06650285e-01 -4.78935182e-01 1.02057591e-01
8.61309350e-01 2.87994385e-01 7.68256962e-01 -1.35010922e+00
-6.19414806e-01 -2.94109732e-01 -4.53008115e-01 -9.96457458e-01
-8.14893991e-02 -7.08284855e-01 -1.79729104e-01 -1.06776512e+00
4.80922580e-01 -1.03401530e+00 -3.31412852e-01 7.75105655e-01
-2.97919422e-01 3.26350629e-01 -2.02923670e-01 -8.70245602e-03
-7.48062313e-01 2.33996868e-01 4.92894828e-01 1.38339400e-01
-5.47592461e-01 3.12900305e-01 -1.02750218e+00 8.10359180e-01
6.14586353e-01 -9.87711489e-01 -1.94059566e-01 -2.33112007e-01
4.42029804e-01 1.14534624e-01 2.87022173e-01 -5.78762889e-01
2.98764497e-01 -4.30825263e-01 5.53271592e-01 -6.08381748e-01
5.68464577e-01 -7.16202259e-01 1.75870463e-01 3.15647125e-01
-3.96527052e-01 1.58437893e-01 5.57152182e-02 6.23786926e-01
-1.61217302e-01 -6.00776598e-02 7.35311389e-01 -2.09691059e-02
-5.45106977e-02 2.72723645e-01 -4.02845621e-01 -7.84045830e-02
1.10746908e+00 -2.24143878e-01 -1.28036514e-01 -6.02418840e-01
-7.54659414e-01 6.60613537e-01 2.84098506e-01 1.72210276e-01
5.63042164e-01 -1.00729251e+00 -9.58755314e-01 4.02224660e-01
-1.37755424e-01 3.92103940e-02 -5.44744879e-02 1.04861116e+00
-3.03996116e-01 3.86434615e-01 1.47683710e-01 -4.63249058e-01
-1.56261194e+00 7.03945696e-01 1.11316927e-01 -3.57729733e-01
-4.99530107e-01 1.01911080e+00 1.20741367e-01 8.12540855e-03
3.74942243e-01 -1.12111636e-01 1.36708722e-01 1.11407861e-01
8.53497148e-01 7.77556002e-01 -8.35181400e-02 -1.21050879e-01
-5.72233796e-01 3.40136766e-01 -2.63887644e-01 -1.74971223e-02
1.69422865e+00 1.13379713e-02 -1.44789040e-01 4.16200668e-01
1.23718512e+00 3.42808306e-01 -1.24376786e+00 -1.06324017e-01
4.29232046e-02 -7.26984382e-01 3.78424786e-02 -6.28819406e-01
-6.65144861e-01 3.94886762e-01 2.28433803e-01 2.50361264e-01
1.07011580e+00 1.44016758e-01 4.22950208e-01 6.46389723e-01
3.05801988e-01 -7.09893048e-01 -4.09991562e-01 2.31028095e-01
7.48738706e-01 -1.29519856e+00 5.88675261e-01 -8.13435674e-01
-5.71694672e-01 1.03751433e+00 3.50905955e-01 -4.51096177e-01
6.59805000e-01 9.50943470e-01 -2.30737418e-01 3.80849838e-02
-1.35333204e+00 2.46692240e-01 1.61853135e-01 4.30090219e-01
1.61351740e-01 5.21562994e-02 -2.39280298e-01 8.40171874e-01
-3.27831239e-01 -1.00957781e-01 3.42332006e-01 7.91473985e-01
-4.99159515e-01 -1.05236399e+00 -4.70933497e-01 1.16096163e+00
-5.41381419e-01 -2.63380796e-01 -1.60078406e-01 8.81881833e-01
-1.95996970e-01 9.55500960e-01 2.94807971e-01 -1.65565223e-01
2.56716520e-01 7.47162178e-02 3.47802401e-01 -5.92451811e-01
-5.06135821e-01 2.78976798e-01 2.51156986e-01 -7.41709888e-01
2.49511614e-01 -1.02108979e+00 -8.73580456e-01 -4.12644953e-01
-8.33865523e-01 4.04087931e-01 5.46287358e-01 6.48432076e-01
3.25891227e-01 9.69483238e-03 8.69778872e-01 -7.26756155e-01
-9.05070662e-01 -6.25730693e-01 -7.39240468e-01 -7.37289712e-02
5.02378285e-01 -7.32576072e-01 -6.90472543e-01 -1.76873744e-01]
|
[7.655708312988281, 4.410341262817383]
|
8dad28ae-7373-4f4c-be24-455131284713
|
dataset-bias-in-human-activity-recognition
|
2301.10161
| null |
https://arxiv.org/abs/2301.10161v1
|
https://arxiv.org/pdf/2301.10161v1.pdf
|
Dataset Bias in Human Activity Recognition
|
When creating multi-channel time-series datasets for Human Activity Recognition (HAR), researchers are faced with the issue of subject selection criteria. It is unknown what physical characteristics and/or soft-biometrics, such as age, height, and weight, need to be taken into account to train a classifier to achieve robustness towards heterogeneous populations in the training and testing data. This contribution statistically curates the training data to assess to what degree the physical characteristics of humans influence HAR performance. We evaluate the performance of a state-of-the-art convolutional neural network on two HAR datasets that vary in the sensors, activities, and recording for time-series HAR. The training data is intentionally biased with respect to human characteristics to determine the features that impact motion behaviour. The evaluations brought forth the impact of the subjects' characteristics on HAR. Thus, providing insights regarding the robustness of the classifier with respect to heterogeneous populations. The study is a step forward in the direction of fair and trustworthy artificial intelligence by attempting to quantify representation bias in multi-channel time series HAR data.
|
['Christopher Reining', 'Gernot A. Fink', 'Markus Pauly', 'Fernando Moya Rueda', 'Lena Schmid', 'Nilah Ravi Nair']
|
2023-01-19
| null | null | null | null |
['human-activity-recognition', 'human-activity-recognition']
|
['computer-vision', 'time-series']
|
[ 2.43678421e-01 -2.28471816e-01 -8.68904293e-02 -2.63478100e-01
-4.00948852e-01 -1.75216660e-01 4.02044326e-01 2.40395084e-01
-6.85423791e-01 5.38607478e-01 4.92706925e-01 -5.09663559e-02
-3.35132480e-01 -6.56349957e-01 -7.60923803e-01 -8.37650597e-01
-4.68267471e-01 -3.36268283e-02 -1.66976348e-01 -1.43480048e-01
-1.55823417e-02 5.14854670e-01 -1.69610047e+00 -9.22242254e-02
3.18210632e-01 1.16251123e+00 -5.74353576e-01 6.19174123e-01
6.29638910e-01 6.44921720e-01 -9.50703442e-01 -1.66007429e-01
1.47493571e-01 -4.63339239e-01 -3.69565398e-01 8.02620649e-02
7.00726867e-01 -1.34974986e-01 -2.18076453e-01 5.79264820e-01
9.70612586e-01 2.97548939e-02 6.71416640e-01 -1.17088532e+00
-5.09729743e-01 6.19940042e-01 -1.63269848e-01 5.32954097e-01
4.36411411e-01 3.61862242e-01 4.24097389e-01 -3.45201939e-01
4.16622013e-01 8.21174741e-01 1.07997012e+00 1.36291042e-01
-9.34693277e-01 -6.10090792e-01 -2.02058032e-01 4.10705805e-01
-1.48740292e+00 -5.62891364e-01 9.08829093e-01 -6.41737521e-01
5.00942826e-01 3.32693666e-01 8.09434533e-01 1.76538622e+00
3.86123508e-01 3.43474358e-01 9.75454211e-01 -4.38713670e-01
6.07579350e-01 -3.28321606e-02 9.36368182e-02 1.68720275e-01
7.01776981e-01 1.92546159e-01 -6.96479261e-01 -2.57860363e-01
5.69069922e-01 -4.28813994e-01 -1.22097591e-02 -4.66078401e-01
-1.30231750e+00 5.19027650e-01 1.67773277e-01 6.47492051e-01
-6.23428822e-01 1.60571471e-01 6.34283423e-01 1.45735443e-01
2.32525632e-01 6.23990834e-01 -4.20549273e-01 -4.36031312e-01
-8.74166489e-01 3.94904196e-01 6.10507071e-01 3.34122628e-01
1.28542170e-01 3.78749579e-01 -3.54851335e-01 6.72996223e-01
7.95625746e-02 7.69917488e-01 6.78233206e-01 -7.67527282e-01
2.33777776e-01 4.12716776e-01 4.46534008e-02 -1.27113807e+00
-7.91461170e-01 -7.15819180e-01 -7.14516699e-01 -6.23351568e-03
7.58368611e-01 -4.39242452e-01 -6.09591842e-01 1.78175271e+00
1.52439967e-01 1.10492371e-01 -2.00279087e-01 1.04502451e+00
7.07168519e-01 -6.64885491e-02 4.14152622e-01 -9.12775695e-02
1.41375053e+00 -5.60010597e-02 -5.02076566e-01 -2.25428298e-01
5.52133620e-01 -2.85393387e-01 8.04609120e-01 3.34601462e-01
-8.70683551e-01 -8.04760396e-01 -1.13806844e+00 5.50852180e-01
-5.48201799e-01 1.77095070e-01 5.18389165e-01 1.26908624e+00
-2.83220738e-01 6.80341244e-01 -7.28711069e-01 -5.78889072e-01
3.48276347e-01 3.04921329e-01 -3.92722189e-01 3.07843715e-01
-1.47640324e+00 9.46707070e-01 5.49054034e-02 2.59029537e-01
-6.25560164e-01 -6.06707335e-01 -8.26100111e-01 -2.21539363e-01
-6.30044639e-02 -5.36622822e-01 8.42550576e-01 -1.18551290e+00
-1.03121674e+00 6.38489962e-01 3.56227934e-01 -6.16747618e-01
7.32834220e-01 -5.43988384e-02 -8.85827899e-01 -8.48715827e-02
4.50223349e-02 2.90472563e-02 8.05372596e-01 -7.12096274e-01
-1.42244101e-01 -6.36262834e-01 -4.04880077e-01 -1.63286373e-01
-2.57271498e-01 -1.11316726e-01 6.32545799e-02 -7.97274411e-01
-2.95818865e-01 -1.11610079e+00 1.69126540e-01 -2.79012322e-01
-8.47098976e-02 2.52594441e-01 4.64865446e-01 -7.20458388e-01
1.24363863e+00 -2.04239559e+00 -7.23365694e-02 4.11142260e-01
1.23918690e-02 7.84992874e-02 2.81834113e-03 3.65032136e-01
-8.39039981e-02 -1.43056691e-01 -1.16737120e-01 1.51829958e-01
-2.38597021e-02 -9.62511450e-02 3.83106619e-01 9.25646544e-01
7.90960342e-02 8.40297043e-01 -4.94769067e-01 -3.90866399e-01
3.41746479e-01 6.50629103e-01 -1.98579729e-01 -4.17311862e-02
3.24105978e-01 4.64346528e-01 -5.01269698e-01 7.16707587e-01
3.33347291e-01 2.14303091e-01 -5.27844541e-02 -4.09265220e-01
-3.96563634e-02 -2.00479165e-01 -1.13004422e+00 1.22042549e+00
-1.53753325e-01 7.84430623e-01 -4.69967663e-01 -9.84746456e-01
9.46095467e-01 3.30516309e-01 9.66180444e-01 -1.03980768e+00
7.25765586e-01 1.22389413e-01 3.16426486e-01 -8.56093526e-01
3.09823275e-01 1.79519251e-01 -3.58247548e-01 -1.47933066e-02
-1.08291872e-01 5.02720594e-01 1.23375952e-01 -5.48702955e-01
1.17515874e+00 6.46486431e-02 1.98821798e-01 -3.12230021e-01
2.71590978e-01 -1.54850200e-01 5.58858335e-01 7.00758815e-01
-5.96021056e-01 5.96312761e-01 2.04238147e-01 -6.30105078e-01
-1.00180697e+00 -9.60059166e-01 -4.26142454e-01 8.80965710e-01
-2.20111862e-01 9.03666615e-02 -7.41098881e-01 -2.71354377e-01
1.93016946e-01 3.73128861e-01 -1.14620101e+00 -6.51482701e-01
-3.71612161e-01 -1.16868651e+00 1.06895840e+00 7.37260401e-01
5.73369920e-01 -9.74924505e-01 -1.40779817e+00 1.98536873e-01
-9.71224606e-02 -1.13066232e+00 -2.65988354e-02 1.25317395e-01
-5.68133712e-01 -1.23678088e+00 -8.28183055e-01 -1.17592849e-01
2.68776685e-01 -5.05644917e-01 8.96205187e-01 -2.66081482e-01
-3.84928823e-01 6.65949345e-01 -4.36152190e-01 -6.56238496e-01
-1.34973219e-02 2.75590837e-01 2.05300227e-01 2.97571629e-01
7.04353213e-01 -5.30166328e-01 -6.88079536e-01 4.12389964e-01
-6.93533242e-01 -5.76839685e-01 6.08888268e-01 4.72541362e-01
1.87214866e-01 6.03207573e-02 5.40459275e-01 -5.52389860e-01
6.30050719e-01 -5.95690906e-01 3.44926342e-02 2.18847886e-01
-4.01465416e-01 -1.52857170e-01 3.64131838e-01 -8.37928593e-01
-5.71873069e-01 9.50856507e-02 1.46676842e-02 -2.51148373e-01
-4.12372380e-01 5.25085151e-01 -1.54649228e-01 -1.72254771e-01
1.04053640e+00 5.27946651e-02 -5.78382760e-02 -3.29177797e-01
-1.38700694e-01 7.26735115e-01 7.18591809e-01 -5.86340249e-01
4.56260383e-01 3.81661147e-01 1.57643706e-01 -1.17020047e+00
-3.10882509e-01 -2.55473316e-01 -8.10181797e-01 -5.90708017e-01
9.13585603e-01 -8.24078679e-01 -5.84665656e-01 7.54424751e-01
-4.40525472e-01 -2.62792379e-01 -2.30715275e-01 7.27954686e-01
-5.84746301e-01 5.62409982e-02 -1.74346641e-01 -1.03175175e+00
-2.99037486e-01 -7.09681511e-01 8.89869153e-01 1.62820041e-01
-6.85349762e-01 -7.61619687e-01 1.76987410e-01 5.32769084e-01
5.69076836e-01 8.52874517e-01 4.84504670e-01 -6.46489739e-01
1.80238947e-01 -5.37476540e-01 3.71451527e-01 8.78285244e-02
-8.73561203e-02 -7.32445642e-02 -1.09393120e+00 -1.71619862e-01
-2.88996398e-01 -1.14811741e-01 2.72965729e-01 6.00734890e-01
9.42853510e-01 -4.87870760e-02 -2.94276085e-02 4.77765024e-01
1.00223470e+00 1.24208048e-01 8.51291955e-01 5.84170640e-01
4.93872166e-01 5.58612585e-01 3.19109321e-01 6.76165283e-01
2.21097589e-01 7.99701571e-01 2.54908562e-01 -1.35385603e-01
-2.13644914e-02 -3.12208869e-02 3.73765200e-01 1.78535640e-01
-3.78157288e-01 2.35431958e-02 -1.03944850e+00 6.31548405e-01
-1.50806487e+00 -1.10674298e+00 -1.58575669e-01 2.22824621e+00
2.98321664e-01 -3.10481135e-02 7.16381192e-01 7.13289678e-01
3.19480449e-01 1.68112367e-01 -4.47735608e-01 -3.87858003e-01
-2.31042832e-01 -3.58544476e-02 9.31222796e-01 -2.09224179e-01
-1.25097752e+00 3.06164771e-02 6.28856516e+00 2.38999844e-01
-1.43864846e+00 -2.38284618e-01 5.83641648e-01 -2.30198771e-01
2.83842444e-01 -6.22487843e-01 -4.17839259e-01 7.67297387e-01
1.40759933e+00 2.03258455e-01 2.92210519e-01 6.75645173e-01
5.72350919e-01 -1.99443981e-01 -8.12508285e-01 9.08446431e-01
3.89797688e-01 -7.30251014e-01 -4.91579562e-01 9.49917138e-02
3.52744251e-01 -1.00273721e-01 3.36533263e-02 3.29804987e-01
-3.49313647e-01 -9.99474406e-01 9.02047515e-01 8.49220932e-01
4.79918391e-01 -5.24645507e-01 9.71533358e-01 3.83290499e-02
-1.02711391e+00 -3.04342568e-01 1.91838946e-02 -1.64506420e-01
-1.13285169e-01 4.07003582e-01 -4.98203486e-01 3.71324539e-01
9.59248185e-01 2.65056491e-01 -1.01257265e+00 1.01053739e+00
4.87886697e-01 7.68469214e-01 -3.40687275e-01 -8.01083073e-02
-2.41191745e-01 2.61766374e-01 1.78201079e-01 1.13866007e+00
3.57559562e-01 -1.51028723e-01 -6.59034476e-02 6.30678833e-01
2.57238626e-01 1.41549751e-01 -5.20636618e-01 -2.39857063e-01
4.00807858e-01 7.24532664e-01 -4.65271711e-01 1.18355438e-01
-3.99018258e-01 5.49119294e-01 -1.28739148e-01 2.76138604e-01
-7.58612931e-01 -8.33509341e-02 6.88977182e-01 6.15478814e-01
7.65687600e-02 -2.94325352e-01 -4.97111499e-01 -8.55571568e-01
2.05612294e-02 -9.98458922e-01 5.79426408e-01 -6.26262009e-01
-1.21122766e+00 2.03411266e-01 2.10743174e-01 -1.17431664e+00
-2.11232319e-01 -4.59378272e-01 -4.79167491e-01 6.92596436e-01
-9.45536196e-01 -1.31329262e+00 -5.00097930e-01 5.54790735e-01
8.91198441e-02 -2.67346144e-01 7.10893452e-01 4.96769488e-01
-7.44510651e-01 6.83535635e-01 -1.21824116e-01 4.83633041e-01
5.56584954e-01 -8.30415547e-01 1.16644032e-01 7.23288178e-01
-1.43356755e-01 4.65554416e-01 9.29171264e-01 -6.23271585e-01
-1.37638080e+00 -8.01811516e-01 7.39379108e-01 -6.31725311e-01
3.27232331e-01 -1.61357015e-01 -6.50777638e-01 4.12790328e-01
-1.81366652e-01 1.52214542e-01 1.01502442e+00 2.61403918e-01
-1.16682850e-01 -2.53722757e-01 -1.16247165e+00 4.40997094e-01
8.14781010e-01 -3.86306465e-01 -5.87089598e-01 -3.48744869e-01
-2.14046210e-01 -2.52411127e-01 -1.27007937e+00 5.54270983e-01
1.22442770e+00 -8.64256978e-01 1.10629511e+00 -5.31155884e-01
1.52483061e-01 -1.08303927e-01 -1.36098787e-01 -1.09277976e+00
-4.70399290e-01 -2.68490575e-02 -4.49901223e-02 9.71930385e-01
2.97332078e-01 -3.48675668e-01 8.30949724e-01 7.57615626e-01
3.70359540e-01 -5.77139735e-01 -1.00304604e+00 -8.08376491e-01
1.57162398e-01 -6.46875262e-01 7.66530752e-01 8.32264543e-01
-1.37810051e-01 8.22870657e-02 -6.06249690e-01 1.29194349e-01
5.83512962e-01 -3.44415367e-01 8.86057436e-01 -1.35775554e+00
-5.52420281e-02 -2.09665969e-01 -8.51710141e-01 2.34948337e-01
-1.84920609e-01 -2.62554497e-01 -3.17793697e-01 -9.56878006e-01
-2.65249223e-01 -1.15007550e-01 -5.22785425e-01 2.81161845e-01
1.19916022e-01 2.57218748e-01 1.49436519e-01 -2.37729587e-02
-2.95377493e-01 3.73937517e-01 7.27775872e-01 -1.43640712e-01
-2.22701907e-01 1.11711584e-01 -4.75711346e-01 3.99763376e-01
8.69897008e-01 -1.74002185e-01 -3.42173934e-01 -9.59031656e-02
1.02649026e-01 5.87032773e-02 6.44629359e-01 -1.69648850e+00
2.71446165e-02 -9.06493962e-02 9.57518518e-01 -8.78606141e-02
3.70790780e-01 -9.84799147e-01 4.89905506e-01 4.96842206e-01
-4.58353043e-01 4.80096824e-02 2.44384125e-01 3.33499074e-01
1.69121563e-01 2.10567296e-01 6.79584563e-01 -5.17472811e-02
-7.18650639e-01 6.04046211e-02 -4.86974597e-01 7.66622126e-02
9.40793276e-01 -6.39792621e-01 -1.32826820e-01 -4.12407905e-01
-6.73198760e-01 -2.43874133e-01 2.58303732e-01 7.49688387e-01
1.06423080e-01 -1.33521926e+00 -6.39430463e-01 4.04349566e-01
5.07395327e-01 -8.82674694e-01 4.08282429e-01 1.02110052e+00
-3.28850746e-01 3.59557807e-01 -7.78725445e-01 -3.74231726e-01
-1.28904903e+00 4.05780673e-01 7.94840991e-01 2.22555146e-01
-4.41057503e-01 2.37989843e-01 -6.57362878e-01 -1.45452589e-01
3.25239182e-01 -1.84655726e-01 -3.63212913e-01 2.92343646e-01
3.83648396e-01 7.14874566e-01 2.43786454e-01 -7.58931160e-01
-5.88003874e-01 5.94669282e-01 5.59466064e-01 8.46338924e-03
1.15357864e+00 -9.74716023e-02 5.82938254e-01 6.53464615e-01
9.77770507e-01 -8.93850550e-02 -9.29033577e-01 1.07994579e-01
6.63557649e-02 -4.01236683e-01 1.45627990e-01 -9.42289650e-01
-8.90022814e-01 5.63643277e-01 1.35257006e+00 1.55972555e-01
9.91176188e-01 -3.07661533e-01 4.55220282e-01 1.46257177e-01
4.13241506e-01 -1.39406002e+00 -2.23633572e-01 9.42021906e-02
7.93523431e-01 -1.18066764e+00 6.31208643e-02 6.50818869e-02
-6.53749406e-01 9.56738114e-01 4.51726437e-01 4.99624647e-02
5.93750060e-01 1.10694319e-01 2.43897393e-01 -2.25896254e-01
-2.07095444e-01 -2.76837826e-01 4.00663942e-01 9.61261570e-01
6.41762316e-01 2.80120343e-01 -6.03486955e-01 7.22823501e-01
-5.74092925e-01 3.84495020e-01 2.76127547e-01 8.57135057e-01
-1.94479793e-01 -6.51876569e-01 -7.41467118e-01 5.41573405e-01
-5.62761426e-01 5.67263722e-01 -5.17548800e-01 8.86982203e-01
6.15481734e-01 8.33480418e-01 -1.20588169e-02 -5.96946418e-01
6.36090934e-01 2.33492091e-01 3.43426943e-01 -1.38387874e-01
-8.41071844e-01 -4.38045293e-01 4.71876562e-01 -4.32199687e-01
-3.97650301e-01 -1.06864309e+00 -6.57579243e-01 -3.22634608e-01
3.35290954e-02 -1.46361962e-01 7.35958159e-01 9.04045880e-01
2.31137961e-01 6.66481435e-01 1.94628984e-01 -7.40554869e-01
-4.90477949e-01 -9.71251845e-01 -5.59724092e-01 7.30613053e-01
3.16478640e-01 -7.13417113e-01 -2.27649808e-01 9.14156623e-03]
|
[13.725889205932617, 2.1219069957733154]
|
c82b3ea6-c650-4f8d-bbfc-ddf4bcc9d619
|
towards-speech-emotion-recognition-in-the
|
1708.03920
| null |
http://arxiv.org/abs/1708.03920v1
|
http://arxiv.org/pdf/1708.03920v1.pdf
|
Towards Speech Emotion Recognition "in the wild" using Aggregated Corpora and Deep Multi-Task Learning
|
One of the challenges in Speech Emotion Recognition (SER) "in the wild" is
the large mismatch between training and test data (e.g. speakers and tasks). In
order to improve the generalisation capabilities of the emotion models, we
propose to use Multi-Task Learning (MTL) and use gender and naturalness as
auxiliary tasks in deep neural networks. This method was evaluated in
within-corpus and various cross-corpus classification experiments that simulate
conditions "in the wild". In comparison to Single-Task Learning (STL) based
state of the art methods, we found that our MTL method proposed improved
performance significantly. Particularly, models using both gender and
naturalness achieved more gains than those using either gender or naturalness
separately. This benefit was also found in the high-level representations of
the feature space, obtained from our method proposed, where discriminative
emotional clusters could be observed.
|
['Khiet P. Truong', 'Jaebok Kim', 'Gwenn Englebienne', 'Vanessa Evers']
|
2017-08-13
| null | null | null | null |
['cross-corpus']
|
['computer-vision']
|
[-4.38053906e-03 1.06396288e-01 4.27002907e-01 -6.63371146e-01
-7.50886500e-01 -4.08972025e-01 7.70519733e-01 2.12725811e-02
-6.41300619e-01 5.88030517e-01 1.43896058e-01 5.59204109e-02
8.99656340e-02 -4.16330881e-02 -4.48043704e-01 -7.57122338e-01
-1.04053475e-01 3.78101885e-01 -2.56060839e-01 -2.87632555e-01
-1.14740483e-01 4.25390780e-01 -1.97124791e+00 6.04111493e-01
5.35322845e-01 1.24594748e+00 8.57731849e-02 6.14808202e-01
-2.80981749e-01 4.82218325e-01 -1.00920033e+00 -5.13861954e-01
2.02600360e-02 -3.15287709e-01 -6.79437399e-01 -1.29346907e-01
3.79378647e-01 2.84226745e-01 1.42106652e-01 9.02237773e-01
9.56464827e-01 5.05159914e-01 5.84576309e-01 -1.41603839e+00
-3.83349925e-01 4.33190495e-01 -3.15006763e-01 -3.99259776e-02
3.11138272e-01 -3.03000420e-01 8.27146113e-01 -1.20019543e+00
4.53270555e-01 1.54574490e+00 6.35644674e-01 9.13526177e-01
-1.20980024e+00 -9.37483728e-01 2.72853345e-01 2.47622117e-01
-1.10977829e+00 -7.71698177e-01 9.43028271e-01 -2.39706442e-01
1.27725887e+00 3.65417480e-01 3.37555230e-01 1.82681119e+00
2.49458011e-02 9.20825303e-01 1.55931103e+00 -6.55411065e-01
2.47509554e-01 8.03915918e-01 -9.67788100e-02 2.76482970e-01
-4.84079897e-01 2.41093934e-01 -8.42265368e-01 -2.24715352e-01
1.86984599e-01 -5.43905735e-01 -1.17180161e-01 -5.44498786e-02
-9.10304368e-01 9.03074384e-01 -5.37647046e-02 7.16924787e-01
-3.27557504e-01 -1.06414452e-01 9.17071164e-01 5.43310165e-01
9.79707301e-01 5.53929746e-01 -9.00973141e-01 -5.03753781e-01
-9.87990320e-01 -7.46868970e-03 9.39433515e-01 5.58943152e-01
6.01047635e-01 4.21503723e-01 -2.64575362e-01 1.42898428e+00
7.43306652e-02 7.58734569e-02 6.82246268e-01 -6.37559056e-01
1.95757434e-01 2.77190864e-01 -2.68262208e-01 -7.86973119e-01
-5.90735376e-01 -6.47290230e-01 -8.28859627e-01 2.63364822e-01
2.12620497e-01 -4.83135462e-01 -7.74770856e-01 2.11594248e+00
1.28929332e-01 2.37516258e-02 4.31783348e-01 5.35528362e-01
7.85786748e-01 5.77886999e-01 4.34812099e-01 -3.34170163e-01
1.39562821e+00 -9.86448944e-01 -1.00509465e+00 -4.17343974e-01
7.87474573e-01 -8.39705408e-01 1.21585512e+00 6.00429654e-01
-7.32252657e-01 -7.21630454e-01 -8.62495899e-01 2.37139568e-01
-9.74052012e-01 4.45470959e-01 4.82594430e-01 1.08895636e+00
-1.20615983e+00 3.68247807e-01 -3.44499737e-01 -2.65550643e-01
2.34495729e-01 3.28956276e-01 -5.06085098e-01 3.56426835e-01
-1.35412920e+00 1.13293576e+00 3.56348485e-01 5.08869179e-02
-6.58783615e-01 -6.21649206e-01 -8.88319910e-01 1.29352584e-01
1.28087133e-01 -1.59489051e-01 1.20753074e+00 -1.71660972e+00
-1.66214395e+00 1.03006399e+00 -3.37634951e-01 -2.44956285e-01
4.39331383e-01 -2.90081203e-01 -7.45518446e-01 -9.78022963e-02
-3.10512990e-01 8.75097811e-01 8.90922070e-01 -1.44665456e+00
-4.07487273e-01 -3.78694713e-01 -3.68836761e-01 8.25283453e-02
-5.94202578e-01 5.61985731e-01 1.19670011e-01 -6.66673660e-01
-4.01773572e-01 -8.57923627e-01 -4.78288382e-02 -4.80266154e-01
-1.12761401e-01 -5.13776600e-01 1.09278238e+00 -7.58802116e-01
1.07375062e+00 -2.47276402e+00 1.31789550e-01 2.14176387e-01
-1.23201557e-01 4.17605191e-01 -9.81834531e-02 2.89009005e-01
-5.20031452e-01 1.98787391e-01 1.34174243e-01 -9.87278521e-01
3.34236890e-01 4.08175200e-01 8.55936110e-02 8.56885239e-02
3.91651452e-01 4.95223671e-01 -3.53328884e-01 -3.80962789e-01
6.05759993e-02 6.65064991e-01 -2.70947665e-01 2.75532156e-01
5.85691184e-02 4.29089606e-01 1.51934639e-01 3.69929373e-01
6.37317300e-01 4.27287102e-01 -1.37480139e-03 -2.01263204e-01
2.85510346e-03 1.61108747e-01 -1.14377046e+00 1.64266789e+00
-1.01703048e+00 9.45181310e-01 3.71491492e-01 -1.01302981e+00
1.28701770e+00 8.41600299e-01 3.45945030e-01 -7.78238177e-01
2.16054156e-01 2.95245796e-01 2.84032702e-01 -5.42067409e-01
6.11693978e-01 -4.20837164e-01 -3.04944277e-01 9.40009281e-02
5.19329548e-01 4.47112136e-02 -1.76523328e-01 -2.36217380e-01
5.83456576e-01 -3.15326117e-02 1.39613330e-01 -5.45268357e-01
5.72690189e-01 -5.68869352e-01 6.17457569e-01 4.75957990e-01
-5.27826190e-01 2.75754035e-01 6.76391840e-01 -3.18000406e-01
-8.20016801e-01 -6.51557386e-01 -2.54283994e-01 1.62318170e+00
-5.45314074e-01 -3.86336774e-01 -7.67919242e-01 -7.48826146e-01
-3.71004462e-01 8.93350184e-01 -9.57186520e-01 -2.93133497e-01
-3.01238507e-01 -7.94024289e-01 8.79502714e-01 5.02800763e-01
3.99933279e-01 -1.21482968e+00 -7.23256052e-01 1.32421091e-01
-1.40179500e-01 -1.38989079e+00 -2.44427845e-02 8.67920280e-01
-3.78554702e-01 -4.38494831e-01 -5.61087966e-01 -7.44674683e-01
6.37105703e-02 -5.41638076e-01 1.31619334e+00 -2.89627552e-01
-1.56298473e-01 4.34240907e-01 -6.55618429e-01 -8.27896476e-01
-4.75100249e-01 5.43322191e-02 3.85227874e-02 2.75070190e-01
5.53368270e-01 -6.01418436e-01 -2.02535897e-01 1.33628905e-01
-7.76515245e-01 -2.82441944e-01 5.01565933e-01 1.19052315e+00
4.67454381e-02 4.82637063e-02 1.03185570e+00 -7.18954921e-01
9.04753804e-01 -3.86076868e-01 7.88647160e-02 1.72796190e-01
-3.71096760e-01 4.04687785e-03 4.48708683e-01 -6.62263572e-01
-1.24596691e+00 -9.90717560e-02 -5.67307949e-01 -6.06651902e-01
-6.97750747e-01 3.04802835e-01 -3.99266511e-01 1.01088300e-01
4.11166340e-01 -7.86051676e-02 1.20750807e-01 -4.71740633e-01
1.76276788e-02 1.00687623e+00 1.09844185e-01 -8.85705650e-01
-8.39750320e-02 -1.86624050e-01 -2.02059835e-01 -9.56514657e-01
-4.96825755e-01 -3.61668259e-01 -6.25151396e-01 -4.11016613e-01
1.01006699e+00 -9.80370998e-01 -6.17726147e-01 4.80188131e-01
-1.17555404e+00 -3.81622076e-01 -1.86511174e-01 5.71016371e-01
-4.32513237e-01 -4.74239253e-02 -4.13405240e-01 -1.31177306e+00
-2.37500995e-01 -1.24648237e+00 1.22322714e+00 1.24803670e-01
-3.38438690e-01 -1.32198465e+00 1.32360980e-02 4.48503271e-02
7.01238453e-01 3.54008347e-01 1.03382587e+00 -1.34024906e+00
5.69098830e-01 1.35104442e-02 2.02596784e-02 7.87402630e-01
2.30372977e-02 1.89601525e-03 -1.75386834e+00 -1.92857444e-01
2.85477698e-01 -7.32738793e-01 6.96726918e-01 9.28284377e-02
1.15288579e+00 -1.69257030e-01 6.26097545e-02 2.58476943e-01
1.11405396e+00 3.47694844e-01 5.89856565e-01 2.61305749e-01
3.94099385e-01 1.18014050e+00 5.35123825e-01 5.84621966e-01
2.24577397e-01 8.44069064e-01 7.13429078e-02 -4.71675664e-01
-3.29525657e-02 1.76465154e-01 6.30036712e-01 7.43813455e-01
9.81255174e-02 -1.47978693e-01 -9.40306604e-01 5.32687187e-01
-1.69770265e+00 -8.48466337e-01 2.98929140e-02 1.83933330e+00
6.84642196e-01 -5.60555495e-02 2.39749819e-01 2.74595946e-01
5.79233170e-01 1.30127668e-01 -2.81847656e-01 -1.23647630e+00
-4.59561408e-01 5.62046289e-01 -6.85087442e-02 3.11152250e-01
-1.21632195e+00 8.47922385e-01 5.99989223e+00 1.13291621e+00
-1.38473666e+00 3.40259731e-01 9.54720140e-01 -2.20272288e-01
-8.90102684e-02 -4.92853105e-01 -6.07224584e-01 2.57950217e-01
1.42655945e+00 2.23851606e-01 1.68764144e-01 9.03746903e-01
9.83790904e-02 6.44817054e-02 -1.16936731e+00 1.06403482e+00
3.41678977e-01 -5.50298929e-01 -2.78024822e-01 -9.12288129e-02
4.57207412e-01 -1.43013775e-01 1.39532745e-01 8.95944357e-01
-6.09559603e-02 -1.08606732e+00 8.05597305e-01 1.45609975e-01
7.79215693e-01 -9.80540514e-01 9.33379352e-01 2.58625209e-01
-9.78382826e-01 -9.51363072e-02 -1.28638595e-01 -5.48373163e-02
-1.18623145e-01 4.04319257e-01 -7.72578061e-01 6.69909596e-01
1.00878084e+00 3.44290942e-01 -6.99504375e-01 4.53986824e-01
1.64635882e-01 5.01515090e-01 -2.89498121e-01 -2.30086058e-01
3.28369290e-01 7.44203478e-02 4.46858406e-01 1.88199723e+00
3.12435687e-01 -4.96886969e-01 1.59797445e-01 5.58518827e-01
2.38007773e-02 4.23858643e-01 -6.67401135e-01 5.33764847e-02
1.98545054e-01 1.47294784e+00 -5.24719596e-01 -3.55766892e-01
-3.65426362e-01 9.99660969e-01 3.01721066e-01 4.20368701e-01
-7.03262746e-01 -4.43944573e-01 8.46811175e-01 -3.97691488e-01
2.63945878e-01 -5.77746518e-03 -1.91522956e-01 -9.26242709e-01
1.13175809e-02 -1.13938832e+00 1.92146286e-01 -6.04920983e-01
-1.31201661e+00 1.13113880e+00 1.13328367e-01 -8.30626070e-01
-5.11208534e-01 -8.99518967e-01 -6.55414999e-01 9.62153256e-01
-1.24044502e+00 -1.31313753e+00 -5.73975332e-02 7.33108699e-01
7.26451576e-01 -5.13637483e-01 1.27399921e+00 5.68339348e-01
-6.02791309e-01 1.14233756e+00 -7.98911974e-02 -9.31200758e-03
9.06992733e-01 -1.39257717e+00 -1.46857545e-01 5.07579148e-01
2.43266910e-01 2.72105783e-01 8.26110005e-01 -3.85687016e-02
-7.28329480e-01 -7.31300771e-01 9.31590497e-01 -1.07355662e-01
4.90724742e-01 -1.01216614e+00 -9.13605809e-01 4.09754753e-01
6.29009485e-01 -1.84848338e-01 1.07343984e+00 6.54457986e-01
-6.53950155e-01 -4.11207192e-02 -1.24222624e+00 2.96806514e-01
5.62640011e-01 -8.62685323e-01 -4.37586188e-01 7.97620509e-03
5.87593973e-01 -1.39486551e-01 -9.07005489e-01 6.02233946e-01
6.64061904e-01 -1.11820579e+00 6.38949871e-01 -6.44869268e-01
2.23347634e-01 3.54230344e-01 -5.72925687e-01 -1.76698971e+00
9.98566598e-02 -5.08323431e-01 2.20405102e-01 1.71186042e+00
6.64200068e-01 -6.36715651e-01 4.31885511e-01 6.21053576e-01
-3.93127888e-01 -8.05577695e-01 -1.29342401e+00 -7.97154844e-01
4.87045318e-01 -6.47368670e-01 4.61409777e-01 9.59138215e-01
-1.43661909e-02 4.64282036e-01 -3.93264353e-01 -1.35163411e-01
4.92419936e-02 -3.13054442e-01 4.73234922e-01 -1.29036820e+00
-2.60015130e-01 -5.69203079e-01 -3.36916208e-01 -2.63923615e-01
8.78033817e-01 -6.99430346e-01 1.25142068e-01 -8.72920096e-01
-5.74067049e-02 -3.33791345e-01 -4.29523677e-01 5.33457816e-01
-1.30238801e-01 3.12281828e-02 3.35762203e-01 -4.19310927e-01
-4.88129705e-01 8.09376955e-01 6.47084951e-01 2.17681259e-01
-1.54669568e-01 -5.23473620e-02 -5.39385378e-01 6.02638006e-01
9.52731788e-01 -4.08288300e-01 -2.86521614e-01 -1.36004895e-01
8.96668360e-02 -4.25632298e-02 2.88509071e-01 -9.76574600e-01
-1.42801283e-02 3.41496438e-01 3.72953176e-01 -2.04464063e-01
8.89561594e-01 -9.19955909e-01 -5.29383570e-02 8.88830982e-03
-4.91824359e-01 2.36955732e-01 7.59614408e-01 1.39374167e-01
-5.83259463e-01 -1.46498561e-01 7.40878999e-01 1.11288968e-02
-5.67124069e-01 -2.65623361e-01 -4.69634920e-01 1.86086800e-02
7.91856468e-01 -8.32944065e-02 -1.16541676e-01 -3.74459624e-01
-9.55192149e-01 -1.57402143e-01 -8.82593542e-02 6.89535558e-01
3.85688037e-01 -1.25938940e+00 -7.62119055e-01 3.17277223e-01
2.06946075e-01 -5.14797151e-01 4.22602087e-01 9.37317252e-01
3.00352544e-01 3.16547245e-01 -3.35209191e-01 -4.91107196e-01
-1.57362652e+00 2.75199175e-01 6.18610322e-01 -4.15931404e-01
1.35226011e-01 9.63173330e-01 1.55601069e-01 -8.11402202e-01
6.29984915e-01 -5.53638190e-02 -1.52382970e-01 5.14116287e-01
2.14500725e-01 1.49792790e-01 4.14777935e-01 -8.05872321e-01
-5.17552674e-01 2.03479916e-01 -1.24142341e-01 -4.64765936e-01
1.41752803e+00 -8.30604881e-02 4.55447510e-02 1.07540846e+00
1.47970855e+00 2.68771010e-03 -9.94907022e-01 2.20971499e-02
1.81772336e-01 -1.57215878e-01 1.43459275e-01 -1.12715471e+00
-9.74644005e-01 1.23198271e+00 9.26795721e-01 3.16554785e-01
1.39573717e+00 7.71143734e-02 2.45260715e-01 2.67252952e-01
1.84878081e-01 -1.43224800e+00 8.25149342e-02 5.74337065e-01
1.02941680e+00 -1.31445837e+00 -5.47349036e-01 -1.63107499e-01
-1.03555083e+00 1.15599954e+00 6.83683753e-01 1.83253691e-01
7.83632040e-01 3.57519537e-01 4.20537829e-01 -2.23901734e-01
-1.13312972e+00 -1.44686460e-01 2.58670688e-01 7.68422663e-01
8.12479973e-01 7.19686821e-02 -4.82565798e-02 9.62650776e-01
-3.41201305e-01 -5.29758811e-01 1.99539915e-01 5.76187074e-01
-2.53356267e-02 -1.31060266e+00 -2.58846253e-01 2.80516177e-01
-7.62871385e-01 -3.55562977e-02 -6.56388104e-01 9.33736086e-01
4.40516889e-01 1.20059645e+00 2.87466380e-03 -7.12509334e-01
3.96692008e-01 7.62878180e-01 2.10578144e-01 -3.31674427e-01
-1.28238297e+00 1.82341754e-01 5.99295139e-01 -4.97900486e-01
-6.16959035e-01 -7.94284523e-01 -7.80498803e-01 2.56214440e-01
-3.23723316e-01 2.84688443e-01 1.04146731e+00 9.72612500e-01
3.77032638e-01 7.53200293e-01 7.11638987e-01 -9.61274207e-01
-3.86114508e-01 -1.41949880e+00 -6.37394428e-01 6.10502958e-01
3.07221055e-01 -8.25824082e-01 -6.65374160e-01 -7.44781271e-02]
|
[13.580784797668457, 5.819431304931641]
|
a2e38de8-44c4-4d4b-b0d0-6b9b3558df45
|
improving-astrobert-using-semantic-textual
|
2212.00744
| null |
https://arxiv.org/abs/2212.00744v1
|
https://arxiv.org/pdf/2212.00744v1.pdf
|
Improving astroBERT using Semantic Textual Similarity
|
The NASA Astrophysics Data System (ADS) is an essential tool for researchers that allows them to explore the astronomy and astrophysics scientific literature, but it has yet to exploit recent advances in natural language processing. At ADASS 2021, we introduced astroBERT, a machine learning language model tailored to the text used in astronomy papers in ADS. In this work we: - announce the first public release of the astroBERT language model; - show how astroBERT improves over existing public language models on astrophysics specific tasks; - and detail how ADS plans to harness the unique structure of scientific papers, the citation graph and citation context, to further improve astroBERT.
|
['Pavlos Protopapas', 'Taylor Jacovich', 'Jennifer Koch', 'Shinyi Chen', 'Kelly E. Lockhart', 'Matthew R. Templeton', 'Timothy W. Hostetler', 'Donna M. Thompson', 'Carolyn S. Grant', 'Edwin Henneken', 'Golnaz Shapurian', 'Michael J. Kurtz', 'Alberto Accomazzi', 'Sergi Blanco-Cuaresma', 'Thomas Allen', 'Felix Grezes']
|
2022-11-29
| null | null | null | null |
['astronomy']
|
['miscellaneous']
|
[-8.33014607e-01 -2.97344536e-01 -4.83667523e-01 6.92101270e-02
-6.37470424e-01 -1.11830461e+00 1.11521101e+00 7.59531796e-01
-2.36603156e-01 6.33910000e-01 4.99842644e-01 -1.09218645e+00
-4.19896930e-01 -8.09476793e-01 -5.62393963e-01 2.37202328e-02
-2.38655031e-01 4.64990526e-01 -1.39231920e-01 1.36435658e-01
7.74920881e-01 9.88281965e-01 -1.28101611e+00 1.45024061e-01
5.44959903e-01 5.30671895e-01 2.11583287e-01 8.44747126e-01
-9.44506586e-01 9.60588336e-01 -4.40323442e-01 -1.87136203e-01
1.70009241e-01 -1.82030443e-02 -1.05381048e+00 -6.04343474e-01
6.40585065e-01 4.05764490e-01 -9.12675679e-01 8.79007518e-01
8.69074184e-03 1.40484303e-01 6.78393066e-01 -1.25012088e+00
-1.04388404e+00 6.18632734e-01 -4.91658419e-01 9.65380669e-01
2.71167219e-01 1.81178063e-01 1.37875021e+00 -8.02437007e-01
1.08403122e+00 1.29417288e+00 4.13047045e-01 4.99741621e-02
-6.83699489e-01 -4.65620100e-01 3.01390700e-03 1.86674237e-01
-1.16888475e+00 6.09684214e-02 4.85335052e-01 -9.24098372e-01
1.29470122e+00 4.91496027e-01 6.96277499e-01 9.70197856e-01
6.50041282e-01 3.12926501e-01 1.10979128e+00 -4.86835331e-01
2.16437295e-01 -1.56613141e-01 6.27954721e-01 7.83818185e-01
4.72607493e-01 -4.52477597e-02 -7.59280384e-01 -5.41949511e-01
8.61453176e-01 -2.28527308e-01 -8.15676749e-02 1.87388286e-01
-1.43763399e+00 6.69448674e-01 3.95726830e-01 6.76849246e-01
-2.88295895e-01 8.27618986e-02 3.58164281e-01 3.16724211e-01
6.31584227e-01 1.09390366e+00 -5.69907367e-01 -3.00917059e-01
-6.73620522e-01 8.97639215e-01 1.23918605e+00 1.02806401e+00
3.23363155e-01 -9.47478935e-02 -1.41323030e-01 7.81047940e-01
5.18950284e-01 4.12686050e-01 3.47290814e-01 -1.25454307e+00
3.23218375e-01 7.36198843e-01 2.01228861e-04 -9.84132946e-01
-3.69909137e-01 -7.53443182e-01 -3.99206519e-01 -2.78686099e-02
8.65449980e-02 1.46445826e-01 -6.17768943e-01 1.07668447e+00
-1.47917584e-01 2.21133232e-01 6.05768636e-02 7.32870340e-01
1.50709689e+00 7.69030809e-01 4.58087176e-01 1.96705058e-01
1.38919461e+00 -8.41107965e-01 -6.54556513e-01 -2.79671699e-01
7.77810693e-01 -1.11640239e+00 9.56516922e-01 2.47852474e-01
-1.10145056e+00 -1.71421543e-01 -8.59217644e-01 -5.76578259e-01
-1.04944122e+00 1.69901401e-02 1.08017135e+00 2.66970992e-01
-1.03645861e+00 5.02298236e-01 -6.43601418e-01 -5.38864493e-01
3.63120049e-01 -2.77633309e-01 1.87388435e-02 -3.94414887e-02
-1.03179324e+00 9.06500161e-01 3.83247197e-01 -5.83212852e-01
-4.63701248e-01 -1.14405107e+00 -5.40927649e-01 1.56329587e-01
3.26272845e-01 -1.08583438e+00 1.31760180e+00 2.07879484e-01
-9.01386917e-01 8.81402373e-01 -5.99699505e-02 -3.98182690e-01
6.19168095e-02 -1.54826447e-01 -7.22257257e-01 -1.39404833e-02
1.05185166e-01 3.55174989e-01 -1.03553772e-01 -9.06507552e-01
-5.24977744e-01 -3.29635918e-01 1.31427646e-01 -1.28421471e-01
-4.44713205e-01 3.06265950e-01 -7.47421265e-01 -9.81356978e-01
-1.42191693e-01 -8.04133534e-01 -3.75093147e-02 1.86450183e-01
-1.43597186e-01 -9.34184492e-01 7.21936345e-01 -5.55352032e-01
1.11957908e+00 -1.94507957e+00 2.16039792e-01 7.58453012e-02
5.81113517e-01 2.62173153e-02 -2.67376363e-01 7.70501375e-01
-3.28158326e-02 5.91726303e-01 8.59993994e-02 -1.45251393e-01
1.69935878e-02 3.34064871e-01 -5.67783535e-01 3.41086805e-01
-1.53642759e-01 9.74439323e-01 -1.12017000e+00 -2.38711849e-01
-6.24415139e-03 -1.01945773e-02 -4.01381016e-01 1.37827307e-01
-6.62036955e-01 1.55857176e-01 -8.71086419e-01 7.42635727e-01
4.77019608e-01 -5.89281559e-01 -1.46613881e-01 1.64954334e-01
-8.96277428e-01 6.48620129e-01 -5.10366321e-01 1.79295409e+00
-4.32899892e-01 9.00497854e-01 7.36906305e-02 -6.37032807e-01
8.24730992e-01 -4.74434393e-03 6.97354615e-01 -5.00017107e-01
-1.46416575e-01 3.19663316e-01 -2.56624162e-01 -7.10851789e-01
4.58020508e-01 1.61432639e-01 1.53719395e-01 5.80168843e-01
1.91692084e-01 -5.72602570e-01 4.25256491e-01 8.25434566e-01
1.34284604e+00 -4.01421189e-02 8.88366401e-02 -7.41531789e-01
4.17687088e-01 4.31272656e-01 2.78994888e-02 9.69884336e-01
4.34362769e-01 3.06499571e-01 2.76576817e-01 -5.61191022e-01
-1.08178318e+00 -1.06384420e+00 -3.15036505e-01 6.97148263e-01
-3.99817288e-01 -9.79215264e-01 -2.12203264e-01 -5.05627573e-01
4.85600650e-01 1.00825965e+00 -9.14285704e-02 2.84828305e-01
-1.15067184e-01 -4.84254479e-01 5.76292694e-01 1.23842455e-01
2.65586108e-01 -8.60222638e-01 4.99502495e-02 -1.44118831e-01
7.29438663e-02 -1.23652911e+00 -2.47396275e-01 -2.97698855e-01
-7.62693465e-01 -1.16760385e+00 -5.67193687e-01 -5.01465499e-01
6.53171092e-02 2.61910796e-01 1.37465787e+00 1.37153253e-01
-7.85250008e-01 9.63321447e-01 -3.29679161e-01 -9.80761051e-01
-2.56486923e-01 2.94521868e-01 -3.39531079e-02 -9.48997796e-01
5.52540481e-01 -3.50221843e-01 -1.82844877e-01 -5.29771507e-01
-9.37065244e-01 -1.85203061e-01 3.07579279e-01 2.88106382e-01
8.35059136e-02 -5.59925437e-02 6.48712218e-01 -6.14449084e-01
7.11180031e-01 -9.63066161e-01 -9.20697629e-01 2.68292427e-01
-8.31386209e-01 9.74209979e-02 1.59337297e-01 2.16999650e-01
-6.83228910e-01 -5.49115300e-01 -1.47111058e-01 -2.35824674e-01
1.22360617e-01 1.37678015e+00 5.84213495e-01 -4.00504410e-01
5.32240212e-01 -2.32336700e-01 -1.29516855e-01 -1.07932937e+00
3.53115171e-01 7.28417695e-01 6.32197320e-01 -8.93932700e-01
6.96077049e-01 1.29490331e-01 3.81118953e-01 -1.01821446e+00
-9.18055236e-01 -8.28684270e-01 -2.67984957e-01 -8.05821270e-02
6.15221858e-01 -1.04367411e+00 -7.21620917e-01 3.74510847e-02
-1.61458755e+00 3.05742741e-01 -8.05828720e-02 7.51735032e-01
-4.45482284e-02 4.16373342e-01 -7.02960134e-01 -7.14383185e-01
-1.79486275e-01 -6.48069084e-01 9.24079716e-01 1.82716519e-01
-3.29062223e-01 -1.56940544e+00 2.36982539e-01 2.17812419e-01
3.43284190e-01 1.53946783e-02 1.36578476e+00 -7.15833068e-01
-7.19400167e-01 -8.02939981e-02 -4.94338065e-01 6.63099065e-02
6.70171008e-02 1.96073428e-01 -5.88285863e-01 -1.85436890e-01
-1.98206194e-02 3.46111879e-02 8.72909248e-01 1.66901156e-01
1.80768216e+00 -4.07644868e-01 -4.43117827e-01 3.32005143e-01
1.36026609e+00 -9.49874818e-02 3.07023913e-01 6.63291752e-01
7.65188336e-01 4.82683390e-01 2.13314183e-02 3.45124036e-01
3.75467300e-01 2.66030818e-01 4.87430952e-02 1.57666355e-01
-3.49775344e-01 -2.14728251e-01 -1.07123209e-02 1.29478002e+00
-3.60114016e-02 -5.68650305e-01 -1.37036180e+00 5.30712247e-01
-1.66920257e+00 -6.43725812e-01 -6.97943926e-01 1.86103261e+00
7.27818310e-01 8.87046456e-02 -3.66404891e-01 -7.30398893e-01
1.18765414e-01 3.12326998e-01 -2.94450969e-01 -4.81626004e-01
-5.27254462e-01 4.09068614e-01 7.06205666e-01 4.40537214e-01
-6.62190497e-01 7.29707599e-01 7.91747665e+00 7.08308995e-01
-8.90550196e-01 1.14042714e-01 1.82183146e-01 -9.87437516e-02
-8.03047597e-01 1.42633036e-01 -6.50968909e-01 3.28325599e-01
1.10036027e+00 -9.06320930e-01 4.65815604e-01 6.43266320e-01
2.02686206e-01 1.73776865e-01 -9.82100308e-01 9.23946977e-01
7.03089461e-02 -2.20156288e+00 3.93277586e-01 1.44850925e-01
5.58024526e-01 7.18560934e-01 -1.04587168e-01 2.72249103e-01
3.47388923e-01 -1.09222507e+00 2.41299227e-01 8.98333549e-01
3.40711713e-01 -2.39434808e-01 4.16201323e-01 3.87536824e-01
-7.41165340e-01 -1.02905497e-01 -4.35264349e-01 -2.41140291e-01
-2.36917898e-01 8.90680969e-01 -6.79923058e-01 9.94331896e-01
8.01407218e-01 1.02602208e+00 -7.84409046e-01 1.33133459e+00
1.43439785e-01 8.45810592e-01 -1.27344891e-01 -2.67780781e-01
4.52770889e-01 -1.84839666e-01 1.13954854e+00 1.47068524e+00
6.76466644e-01 4.23153520e-01 1.81656808e-01 1.17319286e+00
-5.18415332e-01 9.38934013e-02 -8.92405093e-01 -1.22367561e+00
4.72558051e-01 1.19619310e+00 -2.92819828e-01 -3.51151943e-01
-6.50807202e-01 3.09784174e-01 2.08941862e-01 4.51747596e-01
-1.58280224e-01 -2.45657265e-01 4.44977909e-01 3.16015750e-01
-3.15330416e-01 -1.05331540e+00 -8.88490200e-01 -1.03286052e+00
-4.52820450e-01 -6.18561983e-01 6.33602738e-01 -1.40094209e+00
-1.88660407e+00 3.10759455e-01 8.68233666e-02 -7.50154614e-01
4.83559072e-02 -1.02236950e+00 -5.55582941e-01 1.37912762e+00
-1.24856591e+00 -1.07436323e+00 -1.93724670e-02 2.41559613e-02
6.22817874e-01 -9.13734913e-01 7.75258303e-01 3.24895769e-01
-2.04604402e-01 -2.24938780e-01 2.98049420e-01 -1.20995887e-01
7.36254752e-01 -1.29015565e+00 8.24129164e-01 3.05016071e-01
4.08568144e-01 1.01151657e+00 7.78386414e-01 -1.19131255e+00
-2.02235198e+00 -1.02481127e+00 1.45635080e+00 -8.92400205e-01
1.51544869e+00 -7.02533964e-03 -1.07602453e+00 7.84635484e-01
6.23779893e-01 -2.52504706e-01 5.77548981e-01 4.06435966e-01
-4.41341788e-01 2.68929183e-01 -6.76285028e-01 6.91407442e-01
9.75454330e-01 -6.75284207e-01 -1.09120786e+00 1.02020204e+00
9.31781888e-01 -4.20247704e-01 -1.08125174e+00 2.61874527e-01
2.19180554e-01 1.87258914e-01 1.08556068e+00 -9.57889259e-01
6.62373781e-01 -2.48256505e-01 1.47944599e-01 -1.23205268e+00
-6.44040942e-01 -6.35012448e-01 -2.40854532e-01 1.13449574e+00
3.65127176e-01 -6.84317231e-01 2.85067886e-01 4.25912440e-01
-4.14974272e-01 -1.84726566e-01 -9.05646145e-01 -9.70872104e-01
5.95543623e-01 -6.33095622e-01 5.15472054e-01 1.36754549e+00
9.67637077e-02 2.06785798e-01 3.09010893e-01 1.52150974e-01
6.00148916e-01 7.07302541e-02 4.27738219e-01 -1.58679771e+00
-1.00122929e-01 -7.35755265e-01 -1.38395593e-01 -7.81820357e-01
3.55190992e-01 -1.40991664e+00 -5.68258047e-01 -1.79414535e+00
1.16650328e-01 -5.95222056e-01 -4.28227603e-01 6.26299381e-01
1.02863915e-01 1.54194478e-02 -3.55345942e-02 5.65021813e-01
-2.97045469e-01 2.73230821e-01 1.22896695e+00 -3.44055623e-01
3.29568744e-01 -4.62000608e-01 -8.93028378e-01 4.73981023e-01
5.50455630e-01 -4.80167180e-01 4.49964441e-02 -5.88693321e-01
4.95808661e-01 -1.88416779e-01 7.54162431e-01 -7.88188279e-01
7.28397667e-01 -5.01934648e-01 4.35441315e-01 -7.46262729e-01
-1.69257835e-01 -2.48129994e-01 -3.27349193e-02 1.06599577e-01
-4.91770416e-01 5.86054265e-01 9.20330822e-01 4.41674769e-01
-1.65358573e-01 -4.03928667e-01 3.22131626e-02 -4.17997479e-01
-5.26403606e-01 3.67863655e-01 -5.42238176e-01 -7.13711902e-02
4.68047708e-01 8.32795084e-01 -8.07064891e-01 -8.02865699e-02
-2.57530093e-01 4.13880318e-01 2.49617070e-01 9.84845638e-01
2.25116640e-01 -1.11371696e+00 -8.53701234e-01 -2.94432908e-01
1.58208817e-01 -4.24908400e-01 -3.33346874e-02 3.35013568e-01
-6.90571129e-01 1.05100143e+00 2.27221828e-02 -2.57046461e-01
-9.23508704e-01 5.12489021e-01 -9.47221667e-02 -2.70980075e-02
-7.97722816e-01 6.17885709e-01 -1.64029375e-01 -4.70592439e-01
6.34509474e-02 -1.79918975e-01 -3.10055971e-01 -3.66944164e-01
5.66205621e-01 4.11558330e-01 3.59051406e-01 -1.33415341e-01
-1.72832146e-01 2.86381274e-01 -5.39919473e-02 -2.76063144e-01
1.41642916e+00 -1.51173328e-03 -7.94561505e-01 9.95827675e-01
9.59181726e-01 4.14919317e-01 -1.38991535e-01 -2.20707864e-01
4.17539716e-01 -4.36228573e-01 6.35576308e-01 -1.36351645e+00
-6.66409254e-01 7.39820302e-01 1.62640616e-01 5.23391008e-01
4.01335001e-01 2.79609025e-01 6.10899150e-01 6.12902761e-01
4.47315909e-02 -8.94515753e-01 -3.16319525e-01 9.02017355e-01
1.38128912e+00 -6.75967276e-01 5.24222136e-01 -5.33854902e-01
-1.05132580e-01 1.64876580e+00 2.92043984e-01 1.16758950e-01
8.39607894e-01 4.39270824e-01 -6.25339299e-02 -9.22547102e-01
-7.73583651e-01 2.21966967e-01 7.39802539e-01 1.07549995e-01
7.66046107e-01 -1.80157870e-01 -8.20318580e-01 4.46125150e-01
-2.59426475e-01 3.89344394e-01 5.63426673e-01 1.03490245e+00
-2.82795429e-01 -1.39387202e+00 -5.11381984e-01 5.87163925e-01
-4.98477340e-01 -3.94133449e-01 -9.51660097e-01 5.36542296e-01
-2.31313467e-01 6.38155341e-01 1.87257603e-01 1.60866335e-01
1.37161121e-01 1.77593231e-01 3.09203207e-01 -9.02459264e-01
-6.63109779e-01 -1.72066018e-01 3.35051030e-01 -7.22006857e-02
-2.91649699e-01 -6.42403424e-01 -1.41738176e+00 -6.24768615e-01
3.80000919e-01 6.08281195e-01 1.21749115e+00 1.11940849e+00
8.33320975e-01 8.44118953e-01 8.03691819e-02 -3.44141006e-01
-1.04237579e-01 -7.91517556e-01 -5.88410199e-01 -1.81109399e-01
2.02589259e-01 -6.20495200e-01 -5.52802563e-01 -7.80818164e-02]
|
[9.70857048034668, 8.276421546936035]
|
a0446e61-5369-4a39-866a-5cf2d6af6c30
|
gibbs-duhem-informed-neural-networks-for
|
2306.07937
| null |
https://arxiv.org/abs/2306.07937v1
|
https://arxiv.org/pdf/2306.07937v1.pdf
|
Gibbs-Duhem-Informed Neural Networks for Binary Activity Coefficient Prediction
|
We propose Gibbs-Duhem-informed neural networks for the prediction of binary activity coefficients at varying compositions. That is, we include the Gibbs-Duhem equation explicitly in the loss function for training neural networks, which is straightforward in standard machine learning (ML) frameworks enabling automatic differentiation. In contrast to recent hybrid ML approaches, our approach does not rely on embedding a specific thermodynamic model inside the neural network and corresponding prediction limitations. Rather, Gibbs-Duhem consistency serves as regularization, with the flexibility of ML models being preserved. Our results show increased thermodynamic consistency and generalization capabilities for activity coefficient predictions by Gibbs-Duhem-informed graph neural networks and matrix completion methods. We also find that the model architecture, particularly the activation function, can have a strong influence on the prediction quality. The approach can be easily extended to account for other thermodynamic consistency conditions.
|
['Alexander Mitsos', 'Alexei A. Lapkin', 'Kobi C. Felton', 'Jan G. Rittig']
|
2023-05-31
| null | null | null | null |
['matrix-completion']
|
['methodology']
|
[ 2.14190423e-01 1.65010810e-01 -3.74984384e-01 -4.18212056e-01
-2.18836322e-01 -1.60541832e-01 5.49226165e-01 6.61844090e-02
-3.46867234e-01 1.08689511e+00 -1.35395870e-01 -2.98994333e-01
-3.02709699e-01 -7.98315346e-01 -8.34042907e-01 -1.02630389e+00
-8.78657866e-03 5.08809566e-01 5.08245043e-02 -3.72071028e-01
2.79284596e-01 6.53844535e-01 -1.12402844e+00 -4.77913916e-02
1.00632358e+00 1.12908959e+00 -1.34481583e-02 5.13985336e-01
1.44089907e-01 1.16630852e+00 4.14783657e-02 -3.10831100e-01
-2.55072843e-02 -4.93406177e-01 -5.51215708e-01 -4.35897559e-01
3.20311069e-01 2.92054191e-03 -4.01933789e-01 7.09360063e-01
2.58897305e-01 4.69500303e-01 1.20946753e+00 -8.51535857e-01
-7.08877623e-01 6.14262640e-01 -1.88594908e-01 -2.52777487e-01
-9.59957018e-02 2.04518780e-01 1.39881682e+00 -8.74744654e-01
6.64589942e-01 7.77805686e-01 1.06650472e+00 4.70856428e-01
-1.60956383e+00 -3.86567950e-01 1.32387266e-01 7.42713585e-02
-1.25234568e+00 -4.00275826e-01 1.10007906e+00 -6.32510960e-01
1.16505754e+00 1.08201317e-01 7.74930358e-01 7.65692949e-01
4.53116089e-01 2.74627954e-01 9.32541847e-01 -5.70517123e-01
4.50169146e-01 -3.32535505e-02 3.17966312e-01 8.57850611e-01
5.67182958e-01 5.95452189e-02 -5.36392272e-01 -8.73848423e-02
7.27234721e-01 -2.83499181e-01 -1.49875447e-01 -1.06860828e+00
-5.96978366e-01 1.06239247e+00 4.13816154e-01 1.11521214e-01
-5.66573702e-02 5.11053145e-01 3.56371105e-01 7.13869333e-02
7.12222695e-01 8.68108153e-01 -5.08788347e-01 2.35193029e-01
-1.06577849e+00 3.22994649e-01 1.18300748e+00 4.83766168e-01
1.03371465e+00 4.68479514e-01 7.59387687e-02 7.22857714e-01
5.62493801e-01 -2.48856656e-02 6.46546409e-02 -1.25649250e+00
2.20370442e-01 4.75403279e-01 2.59155594e-02 -5.74550509e-01
-4.12283748e-01 -5.90559185e-01 -8.19808245e-01 3.49986285e-01
5.88725090e-01 -1.93648741e-01 -7.10428774e-01 1.92683733e+00
2.66322613e-01 -3.06689113e-01 -3.99627686e-02 6.69584155e-01
5.61114132e-01 5.70178807e-01 1.73406944e-01 -3.41791272e-01
5.91972291e-01 -9.64876652e-01 -4.56378460e-01 -9.77295861e-02
8.13949764e-01 -2.65173078e-01 8.62246156e-01 4.62811530e-01
-1.34706855e+00 -9.40405577e-02 -1.47497368e+00 -3.11232775e-01
-5.66323161e-01 -1.32291153e-01 1.09044790e+00 6.28567815e-01
-1.01922655e+00 1.52170551e+00 -9.38114583e-01 -6.80896491e-02
4.47778478e-02 8.87540400e-01 -1.74508423e-01 3.68966192e-01
-1.13263512e+00 1.08100843e+00 5.60243905e-01 1.37560561e-01
-4.54634994e-01 -6.99417710e-01 -8.87239695e-01 4.17716056e-02
2.28766337e-01 -7.08526671e-01 1.04151177e+00 -9.67692256e-01
-1.86540997e+00 4.93496925e-01 -2.90480219e-02 -4.44261432e-01
6.60068572e-01 3.11152667e-01 -3.45152467e-02 -2.25206062e-01
-5.45551300e-01 4.73411202e-01 5.10740697e-01 -1.05990469e+00
4.02110845e-01 -4.07567397e-02 -1.04024686e-01 3.84664387e-02
-1.93883717e-01 -5.24578452e-01 -7.35523850e-02 -4.55649525e-01
2.68704921e-01 -8.45901847e-01 -3.63069564e-01 1.01549856e-01
-2.58795947e-01 -3.42527442e-02 3.30002397e-01 -6.37487769e-01
9.97304320e-01 -1.70079541e+00 2.34942809e-01 5.94751835e-01
2.84080327e-01 -1.20247886e-01 1.23932865e-02 5.50413430e-01
-1.23828188e-01 6.36471137e-02 -5.42523921e-01 -3.23417366e-01
2.61653751e-01 2.62795597e-01 3.08784455e-01 4.65934873e-01
1.94329813e-01 9.86531615e-01 -4.85505521e-01 -2.26505652e-01
2.03809753e-01 7.15010345e-01 -9.66210485e-01 -3.93958800e-02
-5.96069753e-01 4.84439224e-01 -3.15110803e-01 2.12825909e-01
4.81629789e-01 -6.32207990e-01 7.14661002e-01 -3.06403577e-01
-1.54883757e-01 8.36361170e-01 -8.43439519e-01 1.57029331e+00
-3.73580217e-01 3.76651645e-01 2.09452957e-01 -1.15039849e+00
8.90725613e-01 4.59605977e-02 7.43221641e-01 -5.95505297e-01
2.72810161e-02 3.80534679e-01 2.32544839e-01 3.92167307e-02
4.69043732e-01 -4.25062060e-01 3.52831632e-01 4.03145730e-01
1.73234299e-01 -1.95746839e-01 4.26097691e-01 1.08441360e-01
4.83744919e-01 7.69702315e-01 2.65944898e-01 -8.12675714e-01
5.22738278e-01 -1.08358286e-01 5.00914574e-01 7.14459956e-01
2.27502629e-01 4.27662164e-01 6.69931650e-01 -3.33310455e-01
-1.39728689e+00 -7.83888519e-01 -4.07244712e-01 1.09679496e+00
-2.64944553e-01 -4.11370337e-01 -6.59635127e-01 -3.31019789e-01
1.34379566e-01 5.85780323e-01 -6.55406177e-01 -3.35277289e-01
-4.87625122e-01 -1.03954577e+00 3.22610646e-01 4.99732405e-01
6.77984431e-02 -6.70774341e-01 3.02044660e-01 3.47326666e-01
4.39019322e-01 -6.37794495e-01 -3.39852899e-01 8.08254242e-01
-9.66011584e-01 -9.22782362e-01 -2.97888219e-01 -4.32667404e-01
5.20275116e-01 -4.67940569e-01 1.04713798e+00 6.89128563e-02
-1.00207580e-02 3.95841859e-02 3.16140443e-01 9.63953733e-02
-7.74676681e-01 4.38448578e-01 -5.16089723e-02 -3.21257234e-01
1.21845268e-01 -1.02533329e+00 -7.41489351e-01 8.02984312e-02
-5.78353882e-01 2.21924275e-01 3.19493115e-01 9.59590673e-01
5.04223466e-01 -3.84906381e-01 5.62785923e-01 -1.14170170e+00
5.02553344e-01 -3.02514523e-01 -6.20838106e-01 1.19030572e-01
-1.37476337e+00 6.81081295e-01 7.60690093e-01 -4.18297648e-01
-9.53249574e-01 1.59471527e-01 -3.48971248e-01 -2.58885380e-02
1.74919307e-01 5.99840581e-01 -1.35732725e-01 -5.95478833e-01
5.82192540e-01 6.64191321e-02 2.18801677e-01 -4.48817372e-01
2.54894912e-01 4.05900553e-02 3.81534807e-02 -9.62923884e-01
4.36966509e-01 -5.16296774e-02 5.55337667e-01 -6.30362391e-01
-5.61400056e-01 1.60311699e-01 -6.71127975e-01 -7.36724585e-02
8.68333340e-01 -7.35915661e-01 -9.61057723e-01 3.83282423e-01
-8.69272470e-01 -5.69811881e-01 -1.15961008e-01 4.80399728e-01
-4.87998515e-01 5.18640757e-01 -1.17580056e+00 -8.66222978e-01
-4.44500327e-01 -1.28583205e+00 4.08632725e-01 -1.05835490e-01
-2.69850671e-01 -1.56529188e+00 1.83562517e-01 3.75328869e-01
5.61162353e-01 3.62162977e-01 1.27770531e+00 -7.69169569e-01
-7.51718044e-01 -5.55552393e-02 -4.87968475e-02 3.75311792e-01
5.33080436e-02 1.96893066e-01 -1.02120781e+00 -1.96399629e-01
-2.65053540e-01 -3.47867489e-01 1.20309877e+00 5.08914053e-01
9.12379444e-01 -6.00042343e-01 -9.57966298e-02 6.23113930e-01
1.45455539e+00 9.69177261e-02 4.34791088e-01 6.40019849e-02
1.03150296e+00 3.72326612e-01 -3.83434743e-01 2.12218240e-01
2.27812365e-01 4.34838325e-01 1.44617751e-01 -5.48463017e-02
8.20904225e-02 -4.17547524e-01 4.00293648e-01 1.03746653e+00
-5.03809750e-01 -1.94356795e-02 -7.72738636e-01 -2.15263724e-01
-1.76848161e+00 -7.32135534e-01 -3.45668495e-01 2.29246354e+00
1.10205233e+00 2.29104146e-01 7.32792243e-02 1.01262301e-01
3.91667753e-01 4.64491904e-01 -8.95729423e-01 -4.99347240e-01
-8.43750164e-02 2.62327939e-01 6.94453239e-01 8.66903901e-01
-6.57396734e-01 6.82155192e-01 7.21899557e+00 7.91305542e-01
-1.19968522e+00 2.62997359e-01 7.15191483e-01 -1.92986384e-01
-4.59345251e-01 2.32071817e-01 -6.61438406e-01 2.60754734e-01
8.61579478e-01 4.21262532e-01 6.86649263e-01 6.92873716e-01
2.53650755e-01 -1.06674857e-01 -1.36723518e+00 3.77200812e-01
-3.30127329e-01 -1.56473315e+00 -1.46510988e-01 2.64150888e-01
7.46716619e-01 8.20925750e-04 -1.32759780e-01 1.31874815e-01
3.83116961e-01 -9.33761060e-01 6.00003004e-01 5.95026255e-01
7.94943511e-01 -7.68225729e-01 2.44749561e-01 1.34793833e-01
-9.01311815e-01 3.41390163e-01 -3.04962337e-01 -3.05947006e-01
-1.94262743e-01 7.91852832e-01 -5.97229898e-01 3.30900908e-01
3.49359028e-02 7.42229402e-01 -3.49385560e-01 4.02742952e-01
-1.47140853e-03 5.96611857e-01 -4.55167562e-01 -3.95020872e-01
1.06149450e-01 -8.99939954e-01 2.27160126e-01 1.09475482e+00
1.32542904e-02 -3.56441408e-01 -2.16431022e-01 1.27778149e+00
-5.81193529e-02 2.42717385e-01 -5.79453230e-01 -3.60609174e-01
9.90246832e-02 9.72737670e-01 -6.23027802e-01 -7.77924806e-02
-2.82087594e-01 5.31123161e-01 7.48607218e-01 6.95249796e-01
-5.75841844e-01 -1.22338630e-01 4.43176299e-01 3.00026238e-01
4.68699872e-01 -4.72167671e-01 -5.29012263e-01 -1.37859488e+00
-2.38454733e-02 -4.56209064e-01 9.94065776e-02 -5.92634439e-01
-1.24531054e+00 7.19979405e-02 -2.31774837e-01 -4.50822800e-01
-3.91304731e-01 -1.10475147e+00 -6.36931956e-01 7.98419058e-01
-1.45386934e+00 -1.02267027e+00 2.82130688e-01 1.27639592e-01
-2.14175463e-01 7.75448978e-02 6.42075360e-01 1.67599201e-01
-8.63724649e-01 3.92778873e-01 5.73795855e-01 -1.53030232e-01
5.17500460e-01 -1.32125318e+00 7.37655535e-02 6.76984549e-01
-2.16714993e-01 8.61540258e-01 9.50594425e-01 -5.92735529e-01
-1.66234338e+00 -7.80981541e-01 7.33950555e-01 -2.09596574e-01
9.15838480e-01 -7.07196951e-01 -1.02982068e+00 5.16914129e-01
2.80479848e-01 -1.97489798e-01 6.67983949e-01 3.81839126e-01
-4.34948146e-01 -2.17039838e-01 -8.98244441e-01 4.69344467e-01
9.64583993e-01 -6.02219939e-01 7.52919912e-02 3.87547463e-01
4.84711051e-01 -1.02792285e-01 -1.29501116e+00 4.51842278e-01
7.48744249e-01 -9.18297708e-01 8.95349264e-01 -6.76845431e-01
6.08328938e-01 1.05269708e-01 -8.55427310e-02 -9.38671708e-01
-4.58717704e-01 -4.63856727e-01 -5.05621970e-01 9.31451321e-01
9.75626051e-01 -9.55492973e-01 1.06846035e+00 1.30517530e+00
-4.85031940e-02 -1.02570260e+00 -8.04615200e-01 -5.81441760e-01
6.65214837e-01 -4.26695079e-01 4.32764947e-01 9.56334114e-01
1.52929321e-01 3.29268515e-01 -4.20119345e-01 -1.67812124e-01
5.37641048e-01 -1.69585243e-01 3.26439381e-01 -1.47822678e+00
-7.28705525e-01 -6.61848307e-01 2.04607278e-01 -8.50656748e-01
3.96580756e-01 -1.16854858e+00 -1.39945403e-01 -1.30810618e+00
1.32569354e-02 -5.74477732e-01 -4.08287138e-01 3.77866209e-01
1.62675172e-01 -1.93138849e-02 -8.59442279e-02 2.61718005e-01
-3.31452459e-01 7.73459256e-01 1.12600160e+00 -2.06258632e-02
-3.53860974e-01 -2.00176001e-01 -4.54834610e-01 3.24230820e-01
8.63810003e-01 -3.02300721e-01 -3.75619501e-01 4.16157767e-02
6.87679350e-01 -1.56542677e-02 1.36511818e-01 -9.58950937e-01
1.92573354e-01 -2.75326043e-01 5.65549254e-01 -9.17724073e-02
5.63610792e-01 -4.38665599e-01 3.80401075e-01 4.32370931e-01
-5.06036043e-01 -3.76830161e-01 -5.18167056e-02 3.92711371e-01
2.21040919e-01 -4.81197655e-01 8.21291566e-01 -2.66176015e-01
-2.40033530e-02 3.04323018e-01 -5.90302944e-01 -1.75872564e-01
3.08596492e-01 -3.07293206e-01 -7.62874112e-02 -1.47277728e-01
-8.20475936e-01 -6.96756765e-02 8.14882696e-01 -3.45685989e-01
1.01519205e-01 -1.21540666e+00 -2.33520195e-01 1.83027685e-01
-2.24493846e-01 -4.78614196e-02 -4.66216765e-02 1.04949105e+00
-6.57927513e-01 4.41634238e-01 1.58446327e-01 -3.46886158e-01
-5.77192903e-01 3.47198099e-01 9.48322058e-01 -5.39960921e-01
-1.38196021e-01 3.54326695e-01 3.30843963e-02 -6.48755074e-01
-1.09085888e-02 -3.40461850e-01 1.70348749e-01 -1.59031168e-01
-2.26739481e-01 3.77339393e-01 2.93197274e-01 -3.03301245e-01
-1.22744493e-01 3.30771863e-01 -6.92523271e-02 -1.20780300e-02
1.57207048e+00 2.55032539e-01 -2.68143833e-01 6.24404371e-01
1.12326086e+00 -5.87590896e-02 -1.54887724e+00 -2.80480832e-02
-7.08067194e-02 5.11899829e-01 2.19739780e-01 -6.73156202e-01
-1.17754519e+00 7.96659946e-01 1.58007950e-01 -1.25597775e-01
5.72587669e-01 -2.64590591e-01 3.17770034e-01 8.09015155e-01
-1.55216288e-02 -1.37284791e+00 -2.09795058e-01 5.55401742e-01
5.99715650e-01 -1.11226666e+00 3.70166987e-01 -3.63655120e-01
-1.96203858e-01 1.26142120e+00 5.95043898e-01 -1.45884603e-01
6.08565927e-01 2.45143980e-01 -3.48153532e-01 -1.17789648e-01
-7.55095899e-01 1.77379936e-01 4.87454534e-01 3.10797423e-01
1.03033495e+00 -8.82901624e-02 -5.22299111e-01 3.11910003e-01
-2.87701078e-02 -1.49784312e-01 3.51801485e-01 8.43829393e-01
-3.55353296e-01 -1.41120946e+00 1.05685368e-01 6.54835165e-01
-2.38564163e-01 -3.54828775e-01 -4.71499860e-01 8.97186816e-01
-1.10198446e-01 5.18539310e-01 -1.84099108e-01 -2.41535231e-01
-2.62283444e-01 7.78247893e-01 6.32683218e-01 -3.02437544e-01
-4.68569160e-01 -8.30445159e-03 3.35012555e-01 -2.31046259e-01
-4.78432596e-01 -5.58805346e-01 -1.30759764e+00 -3.78999591e-01
-5.27011812e-01 3.30044657e-01 8.43163610e-01 1.09795189e+00
3.24420482e-01 2.10420653e-01 2.85067260e-01 -1.01354408e+00
-4.29137200e-01 -9.80313480e-01 -7.47764945e-01 1.93847075e-01
3.41012806e-01 -7.45695055e-01 -5.29827058e-01 -3.93859506e-01]
|
[5.307944297790527, 5.344221591949463]
|
9938e331-c847-4405-81cd-9fb926863a56
|
an-empirical-evaluation-of-zero-resource
|
1702.01360
| null |
http://arxiv.org/abs/1702.01360v1
|
http://arxiv.org/pdf/1702.01360v1.pdf
|
An Empirical Evaluation of Zero Resource Acoustic Unit Discovery
|
Acoustic unit discovery (AUD) is a process of automatically identifying a
categorical acoustic unit inventory from speech and producing corresponding
acoustic unit tokenizations. AUD provides an important avenue for unsupervised
acoustic model training in a zero resource setting where expert-provided
linguistic knowledge and transcribed speech are unavailable. Therefore, to
further facilitate zero-resource AUD process, in this paper, we demonstrate
acoustic feature representations can be significantly improved by (i)
performing linear discriminant analysis (LDA) in an unsupervised self-trained
fashion, and (ii) leveraging resources of other languages through building a
multilingual bottleneck (BN) feature extractor to give effective cross-lingual
generalization. Moreover, we perform comprehensive evaluations of AUD efficacy
on multiple downstream speech applications, and their correlated performance
suggests that AUD evaluations are feasible using different alternative language
resources when only a subset of these evaluation resources can be available in
typical zero resource applications.
|
['Sanjeev Khudanpur', 'Santosh Kesiraju', 'Pegah Ghahremani', 'Jinyi Yang', 'Chunxi Liu', 'Najim Dehak', 'Lucas Ondel', 'Ming Sun', 'Lukas Burget', 'Alena Rott']
|
2017-02-05
| null | null | null | null |
['acoustic-unit-discovery']
|
['speech']
|
[ 2.10270643e-01 8.94732680e-03 -2.86826432e-01 -3.04968357e-01
-1.55204296e+00 -7.20739603e-01 3.64050150e-01 3.78966071e-02
-5.99065304e-01 4.64563310e-01 6.33910120e-01 -6.88422620e-01
3.50771993e-01 -3.97872537e-01 -5.52708626e-01 -2.19138369e-01
-1.06089212e-01 4.15701926e-01 -1.68577164e-01 -1.76170971e-02
-1.44327268e-01 3.21125239e-01 -1.38726616e+00 1.11580715e-01
8.47417295e-01 6.69853091e-01 5.46288192e-01 6.78699493e-01
-1.47577643e-01 3.62947941e-01 -5.00788867e-01 -1.28334910e-01
2.10077405e-01 -3.70378733e-01 -7.45765746e-01 -2.77026705e-02
1.30941406e-01 -6.18950367e-01 1.59971695e-02 7.34976232e-01
7.20861316e-01 4.08869088e-01 6.54382408e-01 -6.97989404e-01
-6.47849262e-01 1.04305327e+00 -3.81311849e-02 3.65104109e-01
2.78853804e-01 4.33421582e-01 1.36681557e+00 -1.21494925e+00
3.43770981e-01 1.14240324e+00 4.51762736e-01 5.18853247e-01
-1.40228093e+00 -8.55660677e-01 6.34865612e-02 -4.04018722e-03
-1.39770806e+00 -1.21320510e+00 7.41330862e-01 -4.31639582e-01
1.39952922e+00 1.28614783e-01 2.47364625e-01 1.09546864e+00
-5.30272901e-01 6.99623525e-01 9.12839890e-01 -6.72353804e-01
2.65354037e-01 2.88066834e-01 2.58801691e-02 4.41848576e-01
-3.35112840e-01 -1.65804606e-02 -9.23318028e-01 -9.47792158e-02
4.29877669e-01 -6.43931270e-01 -7.38661140e-02 1.68956965e-01
-1.18324876e+00 8.27422321e-01 -1.04495645e-01 4.38088983e-01
-3.21135819e-01 -1.67460814e-01 6.75294578e-01 4.49392885e-01
6.56814337e-01 7.38275290e-01 -5.91484606e-01 -6.39489532e-01
-9.12042797e-01 -3.63722146e-01 7.75775135e-01 9.98899877e-01
9.78862286e-01 6.55836046e-01 1.57199368e-01 1.44931531e+00
1.20040312e-01 7.22919345e-01 9.57982481e-01 -8.36668432e-01
5.24116099e-01 1.56095335e-02 -2.92555898e-01 -3.20942938e-01
-2.36853927e-01 -2.26451859e-01 -2.18576595e-01 -3.23884487e-01
2.99627006e-01 -4.28787500e-01 -6.78335965e-01 1.88857329e+00
4.86131161e-02 1.98010057e-01 2.43918344e-01 5.28306365e-01
7.06732213e-01 7.14883447e-01 3.46461505e-01 -3.50096554e-01
1.28747630e+00 -6.99186146e-01 -4.89324361e-01 -3.03560019e-01
9.79674459e-01 -9.51383233e-01 1.77916348e+00 2.83206105e-01
-9.07989502e-01 -6.27374768e-01 -9.40645039e-01 -1.20810062e-01
-3.23804855e-01 2.88999021e-01 6.00083530e-01 9.17522728e-01
-9.94624436e-01 2.86804110e-01 -1.00849652e+00 -3.39574963e-01
1.25479147e-01 3.02796900e-01 -4.77521509e-01 -3.48937400e-02
-1.15039456e+00 6.73556209e-01 1.53794006e-01 4.80493829e-02
-1.14003754e+00 -6.41422689e-01 -1.03180325e+00 8.32242221e-02
2.31253237e-01 -1.09545216e-01 1.29146683e+00 -6.88413560e-01
-1.70560002e+00 5.62870324e-01 -2.33721152e-01 -2.98672706e-01
-6.08826987e-02 -2.41745621e-01 -4.89737421e-01 1.37644811e-02
1.41364023e-01 5.59161544e-01 6.87621057e-01 -7.35932350e-01
-5.68608284e-01 -2.60005563e-01 -2.04202980e-01 4.62429404e-01
-1.04118907e+00 2.66912848e-01 -3.72356087e-01 -5.53483844e-01
-8.96108076e-02 -9.28779066e-01 7.91502446e-02 -6.98220134e-01
-3.11289132e-01 -3.75290126e-01 4.30025697e-01 -1.09831774e+00
1.38526380e+00 -2.27542639e+00 -6.99065775e-02 2.30127230e-01
-2.52233595e-01 1.89298376e-01 -2.18481198e-01 4.49630558e-01
2.61808544e-01 3.89234692e-01 -4.26215418e-02 -5.10491490e-01
5.23988307e-02 1.59569040e-01 -2.42531344e-01 2.37559155e-01
3.76147956e-01 5.42552352e-01 -7.51572609e-01 -4.29412156e-01
1.51524231e-01 2.27891088e-01 -8.50097120e-01 3.48072380e-01
1.64107122e-02 4.96896297e-01 -1.43114597e-01 7.57534504e-01
1.03689216e-01 2.92151630e-01 3.91341984e-01 -1.89759016e-01
-3.07588786e-01 1.02568412e+00 -1.02077794e+00 1.72777700e+00
-1.13599837e+00 5.71326017e-01 8.99210870e-02 -8.92562270e-01
8.90830696e-01 5.80434263e-01 3.97089899e-01 -5.03282189e-01
-1.21813223e-01 3.00246298e-01 3.99529755e-01 -3.90761286e-01
4.84260261e-01 -1.63452715e-01 -2.19560698e-01 7.33570337e-01
3.86711329e-01 -1.96508244e-01 1.02082007e-01 -1.03190817e-01
1.17424226e+00 -1.50463045e-01 1.31496668e-01 -2.56920248e-01
2.56371230e-01 -3.85434483e-03 6.14144087e-01 7.26267815e-01
-3.80274713e-01 3.80384207e-01 3.22441421e-02 1.24366030e-01
-1.18167782e+00 -1.19851828e+00 -4.28456008e-01 1.55086970e+00
-7.01062858e-01 -5.67999601e-01 -6.54460907e-01 -3.49013358e-01
-6.66263700e-02 9.34358418e-01 -3.63790616e-02 -5.32995276e-02
-7.56692767e-01 -3.76958549e-01 9.16504920e-01 7.74362147e-01
-7.78986588e-02 -1.05295193e+00 1.19313173e-01 4.52835113e-01
-2.56947547e-01 -1.28569186e+00 -7.06724882e-01 5.73701620e-01
-4.28480446e-01 -3.65348846e-01 -4.85454530e-01 -8.62856388e-01
2.51125067e-01 1.57880366e-01 8.20419312e-01 -3.27621281e-01
2.65886821e-02 4.44624305e-01 -3.98398161e-01 -3.22995096e-01
-7.84615219e-01 5.23991108e-01 7.07060456e-01 -1.95076033e-01
5.35572112e-01 -7.46983051e-01 -1.99985057e-01 1.82291970e-01
-4.15315866e-01 -2.84918100e-01 5.58454633e-01 8.25818360e-01
4.81696784e-01 -2.98583388e-01 1.21085441e+00 -6.74848676e-01
7.36266732e-01 -4.37155068e-01 -4.79858577e-01 -2.70818714e-02
-5.79042137e-01 -5.63507415e-02 6.54411554e-01 -5.40627241e-01
-1.16163445e+00 -4.16343585e-02 -4.40393001e-01 -3.04188907e-01
-2.33215436e-01 7.53905296e-01 -3.64609927e-01 2.82367349e-01
6.37716889e-01 1.77428022e-01 -4.22717538e-03 -7.90714681e-01
5.21221757e-01 1.33104885e+00 3.68666828e-01 -9.31397259e-01
6.10510170e-01 -1.54067099e-01 -6.81860030e-01 -1.27614844e+00
-3.48587126e-01 -6.35384798e-01 -7.50268817e-01 -1.04861716e-02
7.95214832e-01 -1.24384701e+00 -4.39426750e-01 4.68158303e-03
-8.73319268e-01 -4.86364454e-01 -1.89109206e-01 9.56677794e-01
-3.52122247e-01 2.86316961e-01 -6.52813852e-01 -1.03102863e+00
-3.14270735e-01 -1.40272117e+00 9.70394492e-01 -1.56064659e-01
-5.61265826e-01 -7.11826921e-01 7.51564205e-02 4.74642307e-01
4.51693445e-01 -7.41751969e-01 9.68333602e-01 -1.23901486e+00
-2.08178848e-01 -6.89335167e-02 6.14652634e-02 8.70992601e-01
4.42353427e-01 -2.44233683e-02 -1.32324290e+00 -3.58360797e-01
-3.32814962e-01 -5.72226882e-01 3.61668468e-01 2.06984654e-01
9.47949588e-01 -2.52830714e-01 -1.94773842e-02 4.34806347e-01
8.31016481e-01 3.27930719e-01 -3.74220088e-02 -8.18345621e-02
7.05725312e-01 6.35735810e-01 3.57002854e-01 3.30098808e-01
4.28864121e-01 7.46343553e-01 -3.89673740e-01 4.17695194e-02
-2.76631147e-01 -3.71308506e-01 6.88891053e-01 1.72495115e+00
2.25932747e-01 1.52710795e-01 -1.06508517e+00 7.94195473e-01
-1.17685354e+00 -5.87690473e-01 5.75316787e-01 2.38924026e+00
1.23548400e+00 1.73774466e-01 3.65544885e-01 -2.22000301e-01
5.62935293e-01 -1.13950074e-02 -4.89906728e-01 -4.44588184e-01
-3.25386808e-03 4.28592473e-01 3.39952171e-01 4.57379490e-01
-9.20509815e-01 1.37139893e+00 7.08112431e+00 9.52898800e-01
-1.11977303e+00 3.72161388e-01 5.38958669e-01 -1.72734275e-01
-5.20889342e-01 6.64774701e-02 -8.87861192e-01 3.53682518e-01
1.52181339e+00 -2.07214072e-01 6.89349055e-01 9.30484951e-01
3.75890315e-01 1.69947654e-01 -1.23969078e+00 7.87716806e-01
-1.80241972e-01 -9.94358897e-01 -1.94370285e-01 1.79551974e-01
3.33472788e-01 3.91992778e-01 1.32155772e-02 7.40083396e-01
5.47223508e-01 -9.39253271e-01 3.68117869e-01 -8.92548710e-02
1.27009022e+00 -7.49480188e-01 2.95364767e-01 2.62902856e-01
-1.31781554e+00 9.47303027e-02 -4.63836581e-01 1.53074637e-01
2.26332888e-01 2.47651696e-01 -1.51545513e+00 2.95890987e-01
4.79067534e-01 3.02597702e-01 -3.56014311e-01 5.53083718e-01
5.95509857e-02 1.46602690e+00 -5.96078694e-01 -7.53529966e-02
1.75055072e-01 -3.02049834e-02 5.50293982e-01 1.51153684e+00
3.73590320e-01 -2.31175050e-01 3.31876844e-01 6.61155939e-01
-2.30703518e-01 7.89242506e-01 -7.83075392e-01 -5.04053652e-01
1.02529263e+00 1.09597945e+00 -5.14164031e-01 -8.09335634e-02
-7.08298743e-01 6.10983014e-01 5.49904346e-01 3.06338847e-01
-3.65213096e-01 -2.42107794e-01 9.41785455e-01 -1.84750319e-01
1.47401124e-01 -6.13730431e-01 -2.23871440e-01 -1.12569785e+00
-2.41122320e-01 -9.91173506e-01 1.17634140e-01 -5.30964196e-01
-1.27569127e+00 6.72451735e-01 -9.81552228e-02 -1.09777939e+00
-7.80723989e-01 -4.65037853e-01 -4.09112781e-01 1.18013513e+00
-1.23772216e+00 -1.37490702e+00 3.32854778e-01 4.67500240e-01
9.28801298e-01 -6.18356884e-01 1.29291093e+00 4.89630461e-01
-8.47873151e-01 1.17697334e+00 1.39427990e-01 2.04312384e-01
8.54160964e-01 -1.11793661e+00 3.82881969e-01 8.69384825e-01
5.39305627e-01 8.02901447e-01 4.72204834e-01 -5.52395821e-01
-1.45396328e+00 -8.89176190e-01 7.25814223e-01 -2.40322515e-01
1.06252289e+00 -7.42448092e-01 -9.71737266e-01 7.95141041e-01
5.51317409e-02 -3.00400823e-01 1.22846782e+00 8.10608745e-01
-2.56212234e-01 -2.11141855e-01 -6.07781410e-01 7.36240685e-01
1.09957922e+00 -1.21438730e+00 -3.88704926e-01 2.88471907e-01
1.03953719e+00 2.20925938e-02 -1.16153264e+00 8.69148523e-02
5.49109221e-01 -2.40802571e-01 6.80872262e-01 -6.86742604e-01
5.67975864e-02 -3.08180391e-03 -6.52369797e-01 -1.55414665e+00
3.68488050e-04 -7.04001248e-01 3.10872823e-01 1.92777193e+00
8.66875410e-01 -5.51591277e-01 2.74052382e-01 6.19050205e-01
-6.12620056e-01 -3.63647550e-01 -1.05835032e+00 -8.97841632e-01
2.32469156e-01 -1.02370632e+00 5.68443239e-01 8.82513702e-01
3.94564271e-01 5.04821420e-01 -2.28993475e-01 1.82094470e-01
8.80411938e-02 -4.13594663e-01 7.71644056e-01 -8.35347474e-01
-4.90031868e-01 -3.00525039e-01 -1.97789207e-01 -9.58335578e-01
7.10926890e-01 -1.15080106e+00 2.86306232e-01 -8.90267432e-01
-2.41136387e-01 -8.44803631e-01 -4.99106705e-01 6.37579978e-01
-4.32735309e-02 2.78069880e-02 -8.06556258e-04 4.45205659e-01
-5.20581454e-02 5.23711801e-01 7.07145989e-01 2.49446686e-02
-4.57769006e-01 4.38984670e-02 -8.46372008e-01 5.77554643e-01
6.72167778e-01 -3.62415195e-01 -5.35501659e-01 -4.64328229e-01
-4.83914942e-01 -4.06841226e-02 -3.26815516e-01 -7.69216597e-01
9.02822688e-02 -2.12244820e-02 -5.99743575e-02 -2.51985192e-01
5.03178239e-01 -3.87266129e-01 -3.76002610e-01 -9.86430496e-02
-4.05315757e-01 5.47264405e-02 4.25129265e-01 2.22568661e-01
-2.84108609e-01 -4.29216534e-01 4.12116140e-01 1.78409496e-03
-8.03154469e-01 6.26176819e-02 -7.19948292e-01 1.08607598e-01
4.39218372e-01 1.33993939e-01 -7.86528736e-02 -3.70475799e-01
-6.76072538e-01 -2.23002851e-01 2.96491385e-01 3.80011708e-01
3.18131506e-01 -1.30103183e+00 -8.02522898e-01 4.13694650e-01
2.95406133e-01 -3.53276193e-01 5.45791313e-02 6.13592267e-01
-8.25248212e-02 4.30442303e-01 1.25426248e-01 -4.74660754e-01
-1.02010870e+00 2.57366747e-01 6.13736473e-02 -1.56937391e-02
-5.13639271e-01 8.61775398e-01 2.06387252e-01 -6.39001727e-01
2.39711970e-01 -4.90275072e-03 -1.26765475e-01 6.84115216e-02
4.28982407e-01 2.19503537e-01 3.26459169e-01 -8.42833519e-01
-4.52345371e-01 -8.44629556e-02 -1.23372696e-01 -7.20350862e-01
1.31062603e+00 -3.86626601e-01 2.97411829e-01 7.14290321e-01
1.20331693e+00 4.89370644e-01 -1.01308894e+00 -4.75331068e-01
2.85613444e-02 -1.06280297e-01 3.45585287e-01 -4.99860287e-01
-5.85065842e-01 9.35747385e-01 4.31898117e-01 4.77426238e-02
8.55626702e-01 6.07334971e-02 8.23114276e-01 4.30144578e-01
1.76959112e-01 -1.37246442e+00 -1.55441880e-01 4.46931869e-01
7.34761536e-01 -1.29228854e+00 -3.56852889e-01 -1.56236708e-01
-8.77303839e-01 9.00875390e-01 5.41785777e-01 2.53758848e-01
8.81293595e-01 4.85221982e-01 2.01747313e-01 1.76227897e-01
-7.17951298e-01 -5.07108271e-01 3.64875406e-01 7.47358084e-01
9.60470259e-01 2.91402131e-01 -1.78584248e-01 8.76131415e-01
-6.21354342e-01 -6.21413469e-01 2.76792437e-01 6.63507164e-01
-3.95495385e-01 -1.22243261e+00 -1.53862730e-01 5.84726274e-01
-4.68567073e-01 -7.77911365e-01 8.45421944e-03 5.06851792e-01
-2.95348525e-01 1.25934625e+00 1.46666646e-01 -4.85658884e-01
2.30332837e-01 5.66957772e-01 9.50893760e-02 -1.18800962e+00
-3.07984591e-01 6.23146534e-01 5.96091747e-01 -3.18511844e-01
1.57302424e-01 -9.92646396e-01 -1.09359431e+00 9.45288241e-02
-4.03474897e-01 6.78172112e-02 9.09140229e-01 1.09316778e+00
2.79810578e-01 4.28737044e-01 8.12804639e-01 -8.77729952e-01
-7.29155660e-01 -1.38147831e+00 -6.25489235e-01 1.34239689e-01
4.40344363e-02 -5.75943410e-01 -4.77685660e-01 2.47850388e-01]
|
[14.424230575561523, 6.800588130950928]
|
9529c1b6-f13f-486e-a1d8-09ce9a29700b
|
a-survey-of-video-based-action-quality
|
2204.09271
| null |
https://arxiv.org/abs/2204.09271v1
|
https://arxiv.org/pdf/2204.09271v1.pdf
|
A Survey of Video-based Action Quality Assessment
|
Human action recognition and analysis have great demand and important application significance in video surveillance, video retrieval, and human-computer interaction. The task of human action quality evaluation requires the intelligent system to automatically and objectively evaluate the action completed by the human. The action quality assessment model can reduce the human and material resources spent in action evaluation and reduce subjectivity. In this paper, we provide a comprehensive survey of existing papers on video-based action quality assessment. Different from human action recognition, the application scenario of action quality assessment is relatively narrow. Most of the existing work focuses on sports and medical care. We first introduce the definition and challenges of human action quality assessment. Then we present the existing datasets and evaluation metrics. In addition, we summarized the methods of sports and medical care according to the model categories and publishing institutions according to the characteristics of the two fields. At the end, combined with recent work, the promising development direction in action quality assessment is discussed.
|
['Lihua Zhang', 'Ka Li', 'Zhan Sun', 'Tao Suo', 'Qing Yu', 'Peng Zhai', 'Dingkang Yang', 'Shunli Wang']
|
2022-04-20
| null | null | null | null |
['action-quality-assessment']
|
['computer-vision']
|
[ 3.70305628e-01 -1.84929013e-01 -7.60567009e-01 -1.60286248e-01
-6.11772597e-01 -2.67090321e-01 3.01263154e-01 -6.70125633e-02
-6.56584322e-01 4.75034475e-01 6.96223915e-01 2.09137067e-01
-2.27316618e-01 -7.30511010e-01 -2.65183579e-02 -6.17111862e-01
1.70547605e-01 -1.51409313e-01 3.23709637e-01 1.34238914e-01
4.30882931e-01 3.05038393e-01 -1.82796156e+00 5.00980914e-01
9.03139055e-01 1.02626693e+00 -2.04706654e-01 1.02000153e+00
1.56267866e-01 1.35677469e+00 -9.20115292e-01 -4.70987797e-01
1.39142677e-01 -9.40127850e-01 -1.03019857e+00 5.00622928e-01
4.38520044e-01 -7.36374438e-01 -5.79840362e-01 1.09130275e+00
8.15327942e-01 2.50003099e-01 4.52032477e-01 -1.31526470e+00
-4.09635693e-01 1.03697099e-01 8.14716890e-02 5.99490881e-01
8.47579479e-01 3.98034662e-01 5.42716444e-01 -4.16812509e-01
6.11687779e-01 1.02785456e+00 2.93217450e-01 6.84868038e-01
-2.63195634e-01 -2.34122261e-01 -6.94014356e-02 1.02624238e+00
-9.54931080e-01 -3.27913970e-01 5.31648457e-01 -7.19813645e-01
6.09756649e-01 4.58185107e-01 1.37418604e+00 7.65273273e-01
2.45237693e-01 1.17664754e+00 6.96677923e-01 -3.47794861e-01
2.25654483e-01 -4.55562264e-01 3.59670550e-01 6.68747902e-01
2.42263958e-01 -5.08088470e-02 -6.87991321e-01 9.19073746e-02
6.81580365e-01 1.41777098e-01 -1.02699764e-01 -4.31284755e-02
-1.39860284e+00 2.94743896e-01 -1.80859387e-01 4.79345709e-01
-6.70748651e-01 5.77674098e-02 6.06325686e-01 1.53681353e-01
1.26651302e-01 2.02237770e-01 1.09011091e-01 -8.54653060e-01
-8.48831832e-01 2.41432622e-01 4.38356876e-01 6.20142937e-01
2.46321976e-01 8.35258067e-02 -7.14478195e-01 5.43055832e-01
9.37449038e-02 8.14011097e-01 5.01142919e-01 -1.74986613e+00
5.30722402e-02 9.94481802e-01 7.28149787e-02 -1.22164440e+00
-6.85009882e-02 2.24399015e-01 -6.17794275e-01 3.41391802e-01
3.87468815e-01 1.20727979e-01 -4.46038336e-01 1.00323689e+00
5.23062646e-01 1.05564995e-02 2.95239687e-03 9.09908473e-01
1.22611594e+00 4.79930490e-01 1.88925847e-01 -6.76602364e-01
1.51620674e+00 -1.27749860e+00 -1.29104006e+00 2.31316701e-01
6.19340420e-01 -7.10049152e-01 9.43368614e-01 6.38118088e-01
-1.37992895e+00 -7.03884125e-01 -7.39173591e-01 3.06502972e-02
-5.04748076e-02 2.91431218e-01 5.95804870e-01 7.60337412e-01
-8.12147677e-01 4.21836585e-01 -9.50022221e-01 -4.53139544e-01
5.97546041e-01 -6.37814328e-02 -3.64007235e-01 -2.22501114e-01
-1.11452186e+00 8.21247637e-01 2.17636511e-01 -1.46486148e-01
-9.84621823e-01 -2.38105610e-01 -7.08743095e-01 -4.10878539e-01
5.51049531e-01 -6.38616979e-01 1.58447444e+00 -9.97088194e-01
-1.61547768e+00 1.05290449e+00 -9.34980884e-02 -3.66563410e-01
4.87104803e-01 -3.93315941e-01 -6.12774491e-01 7.98886776e-01
1.19026043e-01 1.78037271e-01 3.09081227e-01 -2.29292870e-01
-1.21938586e+00 -3.77724409e-01 4.36613947e-01 5.56865215e-01
-2.72713929e-01 6.02040946e-01 -6.11538589e-01 -5.08743465e-01
-1.11388661e-01 -5.42322218e-01 -2.33843446e-01 2.37658665e-01
3.86529058e-01 -4.35410947e-01 4.66525644e-01 -5.69589436e-01
1.88684464e+00 -1.93805361e+00 1.36854559e-01 -3.78968388e-01
2.84465551e-01 6.28209531e-01 1.02136722e-02 4.06052142e-01
2.25385979e-01 5.71982153e-02 2.06910193e-01 4.43077803e-01
-1.81299880e-01 3.19415182e-01 3.83632928e-01 5.48161566e-01
-3.18209708e-01 8.35867405e-01 -1.03617656e+00 -1.08497751e+00
5.35101414e-01 1.18278474e-01 -4.53102082e-01 2.90416270e-01
3.62782627e-01 4.57547456e-01 -7.78730929e-01 1.04492474e+00
8.05089101e-02 -8.46656561e-02 1.50818899e-02 -3.86281818e-01
-2.74613172e-01 1.79276487e-03 -1.32805061e+00 1.66801775e+00
1.52591914e-01 4.86188322e-01 -1.72409117e-01 -1.02814984e+00
3.67111474e-01 6.74325466e-01 1.36128044e+00 -9.01235878e-01
9.82398316e-02 6.51757643e-02 -2.34654192e-02 -1.36723328e+00
3.23702544e-01 1.26861766e-01 1.26431845e-02 4.73081380e-01
-3.61629575e-02 1.82503149e-01 7.82764018e-01 7.63474405e-02
1.30264008e+00 1.46333992e-01 1.04623306e+00 2.99616218e-01
7.98926651e-01 3.42573337e-02 8.16158473e-01 5.44567645e-01
-1.07213676e+00 1.87952518e-01 2.60844469e-01 -7.88132191e-01
-5.36188066e-01 -7.94694901e-01 1.96755335e-01 9.85104740e-01
4.37536329e-01 -6.96464062e-01 -1.08313870e+00 -7.22377300e-01
-5.01571417e-01 -1.93956029e-02 -4.79405254e-01 -3.36569518e-01
-7.29017913e-01 -5.82459271e-01 5.77814758e-01 4.39421445e-01
9.48506534e-01 -1.51852810e+00 -8.49474609e-01 2.26980433e-01
-7.99251735e-01 -1.00833476e+00 -2.95919895e-01 -1.09677708e+00
-9.57105815e-01 -1.72990704e+00 -7.71846294e-01 -4.53039259e-01
2.21058890e-01 4.40875202e-01 1.00390136e+00 2.52610356e-01
-3.72842997e-01 9.95788872e-01 -7.42706776e-01 -3.86817306e-01
-2.61772633e-01 -4.43008512e-01 2.25440025e-01 -3.07761878e-02
8.28077137e-01 -7.75584653e-02 -1.03357065e+00 6.69263840e-01
-1.01692641e+00 -1.63581625e-01 5.41192710e-01 2.68190145e-01
5.69136977e-01 2.37075493e-01 5.03135920e-02 -2.48457253e-01
4.19742852e-01 6.56835362e-02 -6.13430478e-02 4.78022665e-01
-6.13576949e-01 -5.20437837e-01 -2.54690886e-01 -2.67312020e-01
-9.77050960e-01 -1.26255706e-01 4.53234017e-02 3.91771235e-02
-3.44449013e-01 1.90696791e-01 -3.43717903e-01 6.56424835e-02
6.65270329e-01 8.37082490e-02 2.37651039e-02 -4.07088846e-01
-9.25037190e-02 5.87992907e-01 4.74478096e-01 -1.43628120e-01
3.98612954e-02 5.07939160e-01 2.78165843e-02 -8.64084065e-01
-9.78627503e-01 -7.09258258e-01 -5.40415943e-01 -1.24045253e+00
1.45452988e+00 -5.88748038e-01 -1.04019189e+00 8.83282483e-01
-1.04122078e+00 -3.63779180e-02 -6.08247817e-01 1.07272875e+00
-8.05412352e-01 8.62291813e-01 -6.96042478e-01 -9.14195061e-01
-3.77370745e-01 -1.25528681e+00 1.02576780e+00 3.42749834e-01
-1.33520827e-01 -5.76446354e-01 2.97290474e-01 1.05602992e+00
7.05888355e-03 4.11570549e-01 8.53079837e-03 9.96056870e-02
-6.45725191e-01 -4.94215399e-01 2.22012460e-01 6.20626509e-01
2.06683233e-01 2.63496842e-02 -4.04595345e-01 2.79653758e-01
2.00172767e-01 -8.86502638e-02 5.47980845e-01 8.92808616e-01
9.18375015e-01 -3.59215021e-01 -3.03936694e-02 2.06774652e-01
9.26104367e-01 6.80760324e-01 1.18523729e+00 3.21820468e-01
4.57591623e-01 6.47774994e-01 1.34650671e+00 4.82557625e-01
1.23894729e-01 7.92843521e-01 1.88833877e-01 1.02529787e-01
-1.30315319e-01 -8.22189357e-03 6.98529184e-01 9.35294449e-01
-1.12914836e+00 -2.82077432e-01 -7.46499360e-01 1.91450790e-01
-2.07757831e+00 -1.73936045e+00 -2.85034806e-01 2.00371027e+00
5.42804539e-01 -9.15972590e-02 5.89795291e-01 5.03858984e-01
7.18916178e-01 1.80039570e-01 -2.40160272e-01 2.48190854e-02
2.23619845e-02 -1.78479135e-01 8.53606388e-02 2.65406340e-01
-1.22343540e+00 7.77587354e-01 7.48412752e+00 8.57453346e-01
-4.59838957e-01 2.93895066e-01 1.88406065e-01 -3.11079592e-01
5.04158795e-01 -3.48865777e-01 -4.59638834e-01 3.77521724e-01
7.32553661e-01 -3.70496511e-01 1.00662112e-01 7.69192636e-01
7.09904909e-01 -6.02926731e-01 -7.58628190e-01 1.34192216e+00
3.29591125e-01 -1.09254849e+00 5.36263697e-02 1.37073025e-01
4.12611246e-01 -5.53641140e-01 -3.66448730e-01 6.90524578e-02
-2.25129023e-01 -4.93464708e-01 5.25600910e-01 1.23755634e+00
6.82463229e-01 -4.39397216e-01 7.64029682e-01 2.12778136e-01
-1.22851336e+00 -9.68822166e-02 -2.31020376e-01 -5.06102502e-01
5.32464027e-01 4.27795053e-01 1.66727856e-01 4.72866386e-01
8.91193926e-01 1.17417204e+00 -4.94091600e-01 1.20976067e+00
-3.47849846e-01 5.74730217e-01 3.03514838e-01 -2.41461489e-02
-6.35069236e-02 -3.46169859e-01 6.28417313e-01 9.47522163e-01
2.56601959e-01 7.44348586e-01 3.89908463e-01 -7.06037059e-02
4.23978180e-01 4.10862327e-01 -5.96612036e-01 -4.03361440e-01
-1.21900588e-01 9.15737152e-01 -3.80941480e-01 -8.11955929e-01
-5.96760452e-01 8.36029351e-01 -3.49082828e-01 9.63448882e-02
-9.44997609e-01 -2.13530049e-01 7.63393104e-01 1.73172593e-01
-4.75476980e-01 -1.24247335e-01 -1.96723863e-02 -1.30319798e+00
1.82714984e-01 -1.18165076e+00 7.72428453e-01 -6.70837700e-01
-6.30061150e-01 1.98039696e-01 2.16108635e-01 -1.84196103e+00
-1.76596448e-01 -4.74654108e-01 -3.18841010e-01 1.29372254e-01
-6.44437373e-01 -8.25218022e-01 -5.06450951e-01 4.63741064e-01
8.34064245e-01 -2.95448333e-01 5.40723383e-01 6.15498602e-01
-6.27514660e-01 1.15729488e-01 -5.47245026e-01 2.14503601e-01
5.16855240e-01 -7.56359100e-01 -3.16333562e-01 9.02524471e-01
-2.94061184e-01 8.99105966e-02 5.24607718e-01 -8.17555726e-01
-1.12294984e+00 -6.48240089e-01 9.83169377e-01 -6.47263288e-01
2.38620415e-01 5.03997624e-01 -3.30297589e-01 2.20198676e-01
1.86121747e-01 -2.90622622e-01 1.01749361e+00 -2.32608572e-01
2.61407584e-01 -2.71397740e-01 -1.08865774e+00 7.25153029e-01
1.50277746e+00 -2.33086750e-01 -6.67646945e-01 4.98130679e-01
3.86593670e-01 -7.12399408e-02 -1.08804214e+00 4.39233541e-01
9.90445375e-01 -1.10878241e+00 9.78513241e-01 -8.64820480e-01
4.82792944e-01 -4.37781572e-01 -1.74266044e-02 -4.58231449e-01
-6.57053769e-01 -3.58765453e-01 -3.46700221e-01 7.85802782e-01
-2.61268348e-01 2.66771261e-02 8.04081917e-01 7.31097639e-01
-2.59961277e-01 -5.56573629e-01 -7.13344097e-01 -7.20295250e-01
-6.72232151e-01 -5.18272400e-01 3.38014752e-01 5.62744796e-01
3.43298674e-01 1.47893205e-01 -7.04687297e-01 -3.04601431e-01
5.92612565e-01 -3.72190885e-02 7.70396709e-01 -9.95784044e-01
-1.68944418e-01 -5.30416548e-01 -1.13465822e+00 -7.90026486e-01
-3.96999508e-01 -2.61934042e-01 -1.89808682e-01 -2.01919365e+00
4.50676590e-01 7.34307587e-01 -2.38550112e-01 2.08337083e-01
-2.87763685e-01 4.07691389e-01 7.40927383e-02 3.08966190e-01
-1.35797465e+00 3.13210487e-01 1.63456535e+00 -1.33831784e-01
3.30881361e-04 2.79248416e-01 -2.22491324e-01 1.10147011e+00
8.53051662e-01 -3.05786729e-01 -3.08829039e-01 -2.68905282e-01
3.01500916e-01 2.92618901e-01 2.74782330e-01 -1.33995855e+00
1.38398349e-01 -7.71908104e-01 -9.83379260e-02 -6.15967691e-01
8.17001835e-02 -8.20930064e-01 1.77256897e-01 7.96282470e-01
-3.01319331e-01 3.50155979e-02 -5.39851904e-01 5.66737294e-01
-5.77818036e-01 -3.20458263e-01 8.43445361e-01 -4.19472218e-01
-1.03279674e+00 4.48657125e-01 -1.02239323e+00 1.26816392e-01
1.40260696e+00 -5.60316443e-01 -2.10619882e-01 -5.16345561e-01
-1.05794621e+00 9.04512405e-02 2.85223693e-01 3.22686672e-01
6.61701858e-01 -1.76762855e+00 -6.31867349e-01 -2.61359245e-01
3.38095039e-01 -6.93238735e-01 8.04557025e-01 1.15311575e+00
-7.30320215e-01 3.24566096e-01 -5.12250006e-01 -3.61255527e-01
-1.80158269e+00 5.83933532e-01 4.38699841e-01 -2.57591665e-01
-3.40392858e-01 1.60431921e-01 3.26916277e-02 2.35215902e-01
5.66099584e-01 -2.18782261e-01 -8.58065426e-01 5.38178198e-02
1.24711514e+00 1.28816974e+00 -2.84541965e-01 -8.27355087e-01
-3.97734642e-01 6.76664352e-01 5.38123131e-01 -2.23292947e-01
8.62780809e-01 -2.92362124e-01 -1.11330211e-01 3.54578644e-01
7.63672888e-01 -3.67456764e-01 -7.32011020e-01 -5.16169518e-03
-6.18997701e-02 -7.18556166e-01 -1.37800694e-01 -6.16308868e-01
-1.14204299e+00 7.59633303e-01 1.01919830e+00 1.31453201e-01
1.49396896e+00 -1.01460941e-01 7.09285080e-01 3.89013499e-01
4.43874359e-01 -1.78727186e+00 4.63700533e-01 1.18792914e-01
7.79886425e-01 -1.17338634e+00 3.28268856e-01 -4.90229398e-01
-7.56010950e-01 7.94003308e-01 6.43087387e-01 1.28962338e-01
6.85818195e-01 -2.41535351e-01 1.87342808e-01 -4.45303798e-01
-4.01514143e-01 -4.73721981e-01 5.89563608e-01 6.57406092e-01
5.27205408e-01 8.20050687e-02 -1.22857225e+00 5.10225177e-01
3.44189405e-01 7.75314987e-01 3.50555271e-01 1.19838464e+00
-8.51012945e-01 -1.12436306e+00 -2.97420204e-01 3.29006106e-01
-7.70101368e-01 3.70119244e-01 -4.97684091e-01 3.42501730e-01
3.47566932e-01 1.30720878e+00 -3.00377339e-01 -4.21837419e-01
8.74089658e-01 -2.20541563e-02 4.95145082e-01 -4.29570585e-01
-5.01944125e-01 -9.87611189e-02 2.66805798e-01 -1.21948802e+00
-1.29761779e+00 -7.90159047e-01 -9.30556059e-01 -3.56522083e-01
1.82044581e-01 3.38539332e-02 2.12132379e-01 7.99000680e-01
2.09738493e-01 4.30264384e-01 2.95629352e-01 -3.67240399e-01
-3.02960370e-02 -8.97457063e-01 -7.31023312e-01 5.05551040e-01
-1.41177550e-01 -6.06436670e-01 -1.11355297e-01 6.11747205e-01]
|
[7.965075969696045, 0.4201321005821228]
|
3a365f30-9419-4a65-a0d0-1b1265de74f4
|
norwegian-native-language-identification
| null | null |
https://aclanthology.org/R15-1053
|
https://aclanthology.org/R15-1053.pdf
|
Norwegian Native Language Identification
| null |
['Mark Dras', 'Irina Temnikova', 'Shervin Malmasi']
|
2015-09-01
|
norwegian-native-language-identification-1
|
https://aclanthology.org/R15-1053
|
https://aclanthology.org/R15-1053.pdf
|
ranlp-2015-9
|
['native-language-identification']
|
['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]
|
[-7.423096179962158, 3.8073649406433105]
|
f8f9cc21-21c9-4347-be19-922538affff8
|
r3-refined-retriever-reader-pipeline-for
| null | null |
https://aclanthology.org/2022.dialdoc-1.17
|
https://aclanthology.org/2022.dialdoc-1.17.pdf
|
R3 : Refined Retriever-Reader pipeline for Multidoc2dial
|
In this paper, we present our submission to the DialDoc shared task based on the MultiDoc2Dial dataset. MultiDoc2Dial is a conversational question answering dataset that grounds dialogues in multiple documents. The task involves grounding a user’s query in a document followed by generating an appropriate response. We propose several improvements over the baseline’s retriever-reader architecture to aid in modeling goal-oriented dialogues grounded in multiple documents. Our proposed approach employs sparse representations for passage retrieval, a passage re-ranker, the fusion-in-decoder architecture for generation, and a curriculum learning training paradigm. Our approach shows a 12 point improvement in BLEU score compared to the baseline RAG model.
|
['Eric Nyberg', 'Teruko Mitamura', 'Ritam Dutt', 'Aditya Srikanth Veerubhotla', 'Sireesh Gururaja', 'Sumit Agarwal', 'Suraj Tripathi', 'Srijan Bansal']
| null | null | null | null |
dialdoc-acl-2022-5
|
['passage-retrieval']
|
['natural-language-processing']
|
[ 1.14046149e-01 7.69432068e-01 2.59910643e-01 -4.18682784e-01
-1.93710327e+00 -5.64168215e-01 1.24493575e+00 4.12753999e-01
-2.83512533e-01 1.08440042e+00 1.26367223e+00 -3.45435232e-01
3.97274680e-02 -6.55161858e-01 -6.24671280e-01 -1.49036214e-01
2.16319263e-01 1.14275861e+00 3.25855613e-01 -1.05072761e+00
4.56507325e-01 -5.37595868e-01 -1.01235723e+00 1.11412930e+00
8.17975163e-01 4.04934078e-01 3.93307477e-01 1.47393179e+00
-3.84702951e-01 1.42027938e+00 -9.40925837e-01 -5.02049804e-01
-3.18660468e-01 -1.00351715e+00 -1.75248766e+00 -2.03405380e-01
7.16546357e-01 -6.93284333e-01 -4.80015635e-01 3.98560464e-01
7.96208620e-01 5.71848154e-01 7.75990605e-01 -8.08465481e-01
-8.69806707e-01 1.08504701e+00 5.00732102e-02 1.45500541e-01
9.35404480e-01 -1.79343387e-01 1.49460173e+00 -6.81174457e-01
7.95239210e-01 1.60645843e+00 1.36766702e-01 7.97944069e-01
-1.14635634e+00 7.34895319e-02 -2.36387357e-01 1.68457162e-02
-7.96788871e-01 -6.93158388e-01 4.31698471e-01 -2.72478521e-01
1.33203244e+00 2.81749398e-01 1.04401745e-01 1.34213793e+00
1.89113349e-01 1.19273984e+00 6.32204413e-01 -7.72845507e-01
-2.69466583e-02 7.72040337e-02 5.54718256e-01 7.16413260e-01
-4.94892538e-01 -1.21010251e-01 -7.08721995e-01 -5.77943265e-01
2.89310038e-01 -6.97357893e-01 -3.70458812e-01 2.67719150e-01
-1.09462929e+00 1.23790526e+00 1.25699818e-01 -1.37410715e-01
-2.74293780e-01 6.58945292e-02 3.67095023e-01 5.98271489e-01
4.67480063e-01 8.79151642e-01 -3.03244174e-01 -2.30839357e-01
-8.35827291e-01 1.09367990e+00 1.41415334e+00 1.17140853e+00
4.69655871e-01 -6.54867411e-01 -1.17568707e+00 1.10335076e+00
6.54743969e-01 2.35379487e-01 4.65970993e-01 -1.00054502e+00
8.78921390e-01 3.23447585e-01 3.60279739e-01 -5.98305464e-01
-3.36949170e-01 -1.51459098e-01 -3.76315385e-01 -5.12170374e-01
3.07158470e-01 -6.13812864e-01 -4.07009065e-01 1.51877153e+00
1.00590840e-01 -1.76608637e-01 9.35204208e-01 8.21313143e-01
1.49056435e+00 1.24372649e+00 4.41609835e-03 1.07394166e-01
1.58347380e+00 -1.62717295e+00 -7.45167613e-01 -7.01021254e-02
8.06347847e-01 -1.13030529e+00 9.09619451e-01 8.10903832e-02
-1.40977120e+00 -5.03579021e-01 -8.46987307e-01 -7.72572696e-01
-4.38263826e-02 1.11749515e-01 1.92186192e-01 1.51294723e-01
-1.30385864e+00 1.49939626e-01 -1.85383856e-01 -6.00279570e-01
-5.13849854e-01 -3.86579841e-01 1.15886584e-01 1.98811130e-03
-1.55253983e+00 8.07760179e-01 2.24182397e-01 -1.96641102e-01
-1.13079393e+00 -5.44003904e-01 -7.95417368e-01 1.26149863e-01
5.00600673e-02 -1.11942184e+00 2.19352055e+00 1.62099767e-02
-1.86772537e+00 7.47139692e-01 -5.34136146e-02 -7.89483428e-01
2.13638246e-01 -4.99459207e-01 -2.57259518e-01 3.14351231e-01
2.08516449e-01 8.03468704e-01 3.42522502e-01 -9.61528540e-01
-7.11199224e-01 5.58358093e-04 6.28392339e-01 8.23119998e-01
2.69259095e-01 3.10593378e-02 -5.49920619e-01 -2.00983793e-01
-2.20533475e-01 -8.28567982e-01 -1.93170667e-01 -8.88591945e-01
-7.05537558e-01 -6.57080770e-01 2.45914385e-01 -9.35856581e-01
1.15724647e+00 -1.45833957e+00 4.33063775e-01 -3.90176564e-01
9.80000664e-03 -1.71668112e-01 -4.64697152e-01 1.34642208e+00
6.44802690e-01 -1.42563879e-01 1.59544796e-01 -4.80763108e-01
2.85804480e-01 -2.93999821e-01 -7.22192466e-01 -2.76282430e-01
1.18803367e-01 7.46828079e-01 -1.11833906e+00 -4.73077029e-01
-4.44650292e-01 2.41686761e-01 -6.74781382e-01 9.18898046e-01
-9.07959640e-01 2.15859190e-01 -5.65144360e-01 7.49490112e-02
9.31872874e-02 -3.95118862e-01 1.84292406e-01 1.72657192e-01
4.03615683e-02 1.18229806e+00 -6.05540156e-01 2.32061481e+00
-4.88316953e-01 6.96191072e-01 6.03583902e-02 -2.98992962e-01
1.09524846e+00 8.59565854e-01 -1.53959349e-01 -9.81716156e-01
-2.07756698e-01 5.36430962e-02 -3.88967335e-01 -4.01411265e-01
1.45323968e+00 2.47954160e-01 -3.91756266e-01 1.01093006e+00
3.86885971e-01 -4.12786186e-01 4.36330855e-01 1.09252942e+00
1.00606275e+00 1.25999272e-01 -1.64387256e-01 -3.55355650e-01
5.74450791e-01 2.97883868e-01 -2.47910708e-01 1.11473954e+00
3.33060473e-01 5.94140291e-01 6.13911927e-01 6.79906532e-02
-8.52613509e-01 -7.71793127e-01 1.66610405e-01 1.56443322e+00
-9.21093374e-02 -6.33906364e-01 -7.87836492e-01 -7.66862631e-01
-9.85835642e-02 1.12945271e+00 -4.02735382e-01 -3.14567164e-02
-6.12754583e-01 -4.27171290e-01 8.99932146e-01 1.51760817e-01
4.61812615e-01 -9.20844793e-01 -3.48568968e-02 5.60093760e-01
-8.91689062e-01 -9.05245125e-01 -7.13375509e-01 -1.16880327e-01
-6.78065062e-01 -9.43207443e-01 -9.42887008e-01 -9.52046394e-01
1.16399080e-01 3.54387969e-01 1.69874132e+00 4.04092073e-02
1.97333261e-01 5.09460866e-01 -6.19914770e-01 -2.59208739e-01
-9.91491079e-01 5.41040123e-01 -6.02827549e-01 -5.27454913e-01
2.40909740e-01 1.11249417e-01 -6.39637649e-01 -2.45677531e-02
-6.11249506e-01 3.29691648e-01 3.96252155e-01 1.04161096e+00
4.53767143e-02 -9.73786473e-01 8.80341351e-01 -1.00131583e+00
1.60266733e+00 -7.11709321e-01 -3.87357354e-01 4.61703449e-01
-4.27455425e-01 4.10549343e-01 1.41785502e-01 3.54600549e-01
-1.36575401e+00 -6.22442067e-01 -3.69332939e-01 6.08410180e-01
9.42025557e-02 7.73777366e-01 2.77905047e-01 7.35520780e-01
1.00885332e+00 3.27877253e-01 1.87495927e-04 -6.42915428e-01
7.81337619e-01 9.38837290e-01 4.36306059e-01 -9.15232778e-01
8.90803188e-02 -4.63244438e-01 -6.80229664e-01 -7.34475434e-01
-1.05998147e+00 -8.32907557e-01 -3.23027015e-01 -2.53480405e-01
8.85327816e-01 -1.35071933e+00 -4.00761455e-01 9.26291347e-02
-1.50978804e+00 -4.50227469e-01 -2.84572449e-02 3.01782519e-01
-7.46444285e-01 2.58009762e-01 -9.61727798e-01 -5.82848310e-01
-8.69465292e-01 -9.00604546e-01 1.16362715e+00 4.81859148e-01
-3.80456805e-01 -1.08228290e+00 7.63504326e-01 9.61651504e-01
4.05743152e-01 -2.32410192e-01 9.58642721e-01 -1.09808481e+00
-6.15866780e-01 6.00758530e-02 -1.33984745e-01 -4.43413816e-02
-2.86360294e-01 -3.99571836e-01 -9.74036753e-01 -3.86744380e-01
-4.93177950e-01 -1.04525995e+00 9.47245717e-01 -1.04992753e-02
2.86314875e-01 -6.74382508e-01 -1.59981754e-02 -1.51077032e-01
1.04865420e+00 1.44652892e-02 6.39176607e-01 3.29603016e-01
1.16248131e-01 8.26117337e-01 5.02215624e-01 3.83024395e-01
9.94062424e-01 7.37856865e-01 1.66209728e-01 1.76228151e-01
-3.43743443e-01 -5.99378884e-01 5.01972556e-01 1.02783084e+00
2.90395707e-01 -9.42132950e-01 -6.58954859e-01 7.53204823e-01
-1.95597708e+00 -1.05854583e+00 -2.78776914e-01 1.77996576e+00
1.28405344e+00 -3.57560039e-01 6.83687851e-02 -7.72673368e-01
3.39436173e-01 3.46286356e-01 1.12026721e-01 -8.35436404e-01
1.97809711e-02 1.42553942e-02 -9.06192362e-02 1.20759082e+00
-7.43321836e-01 9.09770131e-01 6.66530609e+00 2.25338772e-01
-3.47905308e-01 1.26492023e-01 4.02189255e-01 7.39995912e-02
-5.21268547e-01 8.93096253e-02 -1.37792587e+00 2.00823154e-02
1.49822831e+00 -4.84258771e-01 1.24333344e-01 5.95491290e-01
7.37510696e-02 -5.07092290e-02 -1.14637375e+00 2.25546807e-01
5.51350296e-01 -1.64652932e+00 3.82802367e-01 -2.43086711e-01
5.80430984e-01 1.09918743e-01 -2.88852811e-01 9.75573421e-01
1.04511237e+00 -8.38706017e-01 3.59494120e-01 7.29644716e-01
1.08405419e-01 -6.91213727e-01 6.85271382e-01 3.53024811e-01
-4.16801542e-01 3.89900476e-01 -5.58468699e-01 2.21089348e-01
2.39592016e-01 1.20573267e-02 -1.71288645e+00 6.06033087e-01
2.51415253e-01 1.09430842e-01 -3.35993141e-01 8.90503228e-01
-2.41309032e-01 7.93868363e-01 7.76931867e-02 -4.65209752e-01
8.44883680e-01 -6.91409409e-02 5.94323397e-01 1.52605081e+00
1.44230157e-01 1.15747847e-01 3.25299531e-01 7.42084563e-01
-3.33142966e-01 4.61779326e-01 -2.70689547e-01 -1.24701500e-01
3.79804194e-01 1.15237558e+00 8.44970793e-02 -5.42645514e-01
-3.42503607e-01 1.26797819e+00 3.34037900e-01 3.46922845e-01
-3.04563671e-01 -6.37169838e-01 3.45530272e-01 -3.52123797e-01
3.81417423e-02 -8.80319104e-02 4.08946067e-01 -1.01707256e+00
-3.40120375e-01 -1.22829199e+00 6.87920988e-01 -9.45493579e-01
-9.61324275e-01 8.72758687e-01 -1.40603587e-01 -7.93515205e-01
-1.06097627e+00 -4.25724424e-02 -6.11734569e-01 1.27798021e+00
-1.69615602e+00 -1.14898884e+00 -1.42598944e-02 6.52035296e-01
1.09450912e+00 -3.76636446e-01 1.35841942e+00 2.85024079e-03
-2.16957659e-01 4.85531807e-01 4.31717604e-01 3.87875229e-01
1.21312606e+00 -1.60277438e+00 5.37080348e-01 4.49506074e-01
2.10045204e-01 8.15354168e-01 9.46227074e-01 -5.67713380e-01
-1.39459276e+00 -9.05829847e-01 1.56267583e+00 -6.35547340e-01
6.06783450e-01 -3.10722142e-01 -7.74519622e-01 7.73400962e-01
1.31036472e+00 -1.04286146e+00 1.04451323e+00 5.42469323e-01
-1.60408199e-01 4.31898445e-01 -7.26813078e-01 5.87978840e-01
9.56098512e-02 -7.02928662e-01 -1.21748483e+00 9.01872635e-01
1.05732012e+00 -8.00964773e-01 -8.99936140e-01 -4.30570692e-01
2.46147737e-01 -5.47177970e-01 7.64677286e-01 -1.03775811e+00
8.86750579e-01 4.52888384e-02 -2.71608502e-01 -1.46782637e+00
-2.20808193e-01 -1.06210184e+00 -2.73659229e-01 1.23249674e+00
8.71857882e-01 1.58343345e-01 5.13315916e-01 4.36133385e-01
-6.37239873e-01 -2.21953109e-01 -6.65689468e-01 -2.64183253e-01
4.65196073e-01 2.51526713e-01 5.46230555e-01 5.09186268e-01
5.27882755e-01 1.31622243e+00 -5.07660151e-01 8.42168629e-02
4.58180100e-01 2.77921319e-01 1.05163467e+00 -1.08326387e+00
-5.00776052e-01 -9.41561535e-02 4.68441129e-01 -1.79250193e+00
-5.43528120e-04 -1.09102857e+00 5.29929698e-01 -2.08553219e+00
2.40395889e-01 -6.63415492e-02 1.80365697e-01 1.65452302e-01
-3.31221938e-01 -2.45279610e-01 5.12365997e-02 1.63012102e-01
-1.09638083e+00 7.46093750e-01 1.27226830e+00 -3.55735987e-01
-2.92719573e-01 4.13365178e-02 -1.10671854e+00 6.68191612e-02
5.30553281e-01 -3.01926196e-01 -6.50743604e-01 -7.30488479e-01
2.97547758e-01 9.79420424e-01 7.83923492e-02 -6.38740957e-01
5.89879692e-01 1.71799764e-01 -6.35422096e-02 -8.74677122e-01
4.88486826e-01 1.07368223e-01 -7.61996627e-01 5.48701920e-02
-1.36380005e+00 1.31921604e-01 1.98756158e-01 5.76673508e-01
-3.75444263e-01 -4.76525903e-01 2.55870849e-01 -4.06487137e-01
-5.51724017e-01 -1.92690387e-01 -9.27338839e-01 5.57677805e-01
2.44984597e-01 6.22452438e-01 -1.05022514e+00 -9.61854517e-01
-6.18676364e-01 8.03177297e-01 -1.79991677e-01 7.39800572e-01
6.51365280e-01 -1.06232512e+00 -1.58440328e+00 -3.10770243e-01
3.09293330e-01 -2.38209307e-01 2.22072944e-01 4.10747647e-01
-4.51110989e-01 1.16108620e+00 8.76680687e-02 -3.33000034e-01
-1.13754165e+00 -2.42406309e-01 1.15871668e-01 -8.65955472e-01
-6.80216253e-01 1.09019041e+00 -2.26818457e-01 -8.80378842e-01
2.58993238e-01 5.48898131e-02 -6.95138216e-01 4.01192248e-01
1.07147050e+00 2.62276798e-01 2.58571774e-01 -2.63072014e-01
2.56946623e-01 -2.75281310e-01 -7.85379171e-01 -8.48914146e-01
9.87389982e-01 -4.92428571e-01 -2.92968675e-02 4.00076807e-01
1.15963423e+00 -1.46518331e-02 -7.62871861e-01 -5.66975474e-01
1.04495563e-01 1.72064811e-01 2.47409478e-01 -1.37286806e+00
-1.64600685e-01 6.51202083e-01 1.73590243e-01 2.33394042e-01
3.90606344e-01 6.77991956e-02 1.02278733e+00 1.22473931e+00
1.32599831e-01 -1.19912994e+00 2.91430116e-01 1.22172976e+00
1.26825547e+00 -1.19135046e+00 -1.90781966e-01 2.31326848e-01
-9.94222105e-01 1.20643246e+00 6.93099320e-01 1.99982569e-01
1.19973287e-01 -3.23408872e-01 4.26336884e-01 -3.83994788e-01
-1.43475258e+00 -2.17869416e-01 3.75431418e-01 3.46837312e-01
1.01686156e+00 -2.53796071e-01 -4.61853862e-01 4.95817721e-01
-4.80620861e-01 -3.27844024e-01 9.77460146e-01 8.82137835e-01
-8.33599329e-01 -9.94094193e-01 -1.08503222e-01 1.32835403e-01
-4.50309694e-01 -6.27426386e-01 -7.30668187e-01 3.65602374e-01
-9.76449668e-01 1.55070388e+00 1.89145327e-01 -1.97639644e-01
3.43153358e-01 6.06584668e-01 2.61268258e-01 -1.07923710e+00
-1.03423011e+00 6.03988506e-02 1.01298416e+00 -3.91474724e-01
-2.30090797e-01 -6.16060197e-01 -1.06875563e+00 -6.73114285e-02
-2.30984926e-01 1.15284240e+00 6.87608302e-01 7.42591977e-01
6.48742557e-01 6.07767403e-01 5.95247984e-01 -5.87664843e-02
-7.77447701e-01 -1.50148737e+00 -2.10809484e-01 -1.78370122e-02
3.96660805e-01 2.12755967e-02 -1.11725241e-01 6.22048900e-02]
|
[12.400299072265625, 8.090709686279297]
|
8ad921a6-335e-4fbb-b596-6f44912e4b59
|
shot-in-the-dark-few-shot-learning-with-no-1
|
2010.02430
| null |
https://arxiv.org/abs/2010.02430v2
|
https://arxiv.org/pdf/2010.02430v2.pdf
|
Shot in the Dark: Few-Shot Learning with No Base-Class Labels
|
Few-shot learning aims to build classifiers for new classes from a small number of labeled examples and is commonly facilitated by access to examples from a distinct set of 'base classes'. The difference in data distribution between the test set (novel classes) and the base classes used to learn an inductive bias often results in poor generalization on the novel classes. To alleviate problems caused by the distribution shift, previous research has explored the use of unlabeled examples from the novel classes, in addition to labeled examples of the base classes, which is known as the transductive setting. In this work, we show that, surprisingly, off-the-shelf self-supervised learning outperforms transductive few-shot methods by 3.9% for 5-shot accuracy on miniImageNet without using any base class labels. This motivates us to examine more carefully the role of features learned through self-supervision in few-shot learning. Comprehensive experiments are conducted to compare the transferability, robustness, efficiency, and the complementarity of supervised and self-supervised features.
|
['Erik Learned-Miller', 'Subhransu Maji', 'Zitian Chen']
|
2020-10-06
|
shot-in-the-dark-few-shot-learning-with-no
| null | null | null |
['unsupervised-few-shot-learning', 'unsupervised-few-shot-image-classification']
|
['computer-vision', 'computer-vision']
|
[ 3.93922031e-01 2.98857093e-01 -3.99555355e-01 -7.35311806e-01
-6.13422990e-01 -2.80809075e-01 6.54783845e-01 3.02291363e-01
-5.05457520e-01 9.13099229e-01 7.23101720e-02 2.63839245e-01
3.37421261e-02 -8.98877382e-01 -6.24504507e-01 -8.54731202e-01
1.52833670e-01 4.89379674e-01 4.93651062e-01 -2.66951144e-01
-1.14835184e-02 2.06917614e-01 -1.98772168e+00 4.05907810e-01
6.97417200e-01 8.07423532e-01 8.17484856e-02 2.26003081e-01
-1.71052366e-01 7.33231544e-01 -5.24463594e-01 -1.60554230e-01
1.99066713e-01 -8.69151175e-01 -7.18728304e-01 4.40355510e-01
3.09900314e-01 -2.80312240e-01 -1.68517604e-01 1.01942337e+00
4.93188798e-01 4.68701363e-01 6.45839810e-01 -1.36150098e+00
-7.06084192e-01 5.86581886e-01 -3.77349883e-01 3.62899095e-01
1.30942196e-01 1.35861129e-01 8.90476227e-01 -1.00953341e+00
1.01122999e+00 8.41285586e-01 6.91852450e-01 8.47670138e-01
-1.31130528e+00 -7.12890625e-01 -1.05434366e-01 2.99502820e-01
-1.22859335e+00 -6.62038505e-01 7.40183175e-01 -3.80913407e-01
7.11996973e-01 -1.19961485e-01 5.67960680e-01 1.02511680e+00
-1.79912299e-01 6.14570379e-01 1.29246569e+00 -7.74230838e-01
7.77013481e-01 7.75499046e-01 5.99793136e-01 5.22924900e-01
2.95757174e-01 3.53376091e-01 -4.77507055e-01 -1.92962497e-01
1.60622925e-01 3.02782238e-01 -1.47167578e-01 -8.85344505e-01
-7.12339997e-01 1.16460013e+00 6.87589049e-01 4.97715354e-01
-1.67087451e-01 -2.21250564e-01 5.25359035e-01 3.41849118e-01
6.57629609e-01 5.21954596e-01 -4.46963519e-01 9.03711841e-02
-6.34351313e-01 -1.96999729e-01 7.35822558e-01 1.08853936e+00
1.35678101e+00 7.86557570e-02 -7.74443895e-02 1.01105464e+00
-1.87798478e-02 2.49288142e-01 9.26628768e-01 -5.23070335e-01
-5.89095987e-02 6.47874534e-01 -3.06612521e-01 -4.07514364e-01
-1.49338037e-01 -3.72589469e-01 -3.30640048e-01 9.70096737e-02
3.40770692e-01 -3.06855530e-01 -1.21046662e+00 1.69171894e+00
4.47806656e-01 1.81929052e-01 2.21939191e-01 6.43042028e-01
9.32405770e-01 4.88767415e-01 1.34933248e-01 -4.59869832e-01
9.01532531e-01 -8.78257036e-01 -4.37791169e-01 -2.92160809e-01
1.05935788e+00 -3.85153115e-01 1.09860742e+00 -8.56658891e-02
-4.56485152e-01 -5.04741073e-01 -1.12011123e+00 3.71132374e-01
-7.80466199e-01 -3.47016335e-01 6.06480539e-01 6.26653910e-01
-5.79537332e-01 8.14995706e-01 -4.00746346e-01 -7.40153372e-01
8.03012788e-01 1.54343605e-01 -4.37126517e-01 -4.24937993e-01
-1.11820936e+00 1.00404358e+00 4.71973240e-01 -5.86681962e-01
-8.34346175e-01 -7.53800035e-01 -9.10263479e-01 3.80914845e-02
5.00909507e-01 -1.39001593e-01 1.21587706e+00 -1.24649870e+00
-1.22200882e+00 1.08842206e+00 5.45904189e-02 -4.15240556e-01
1.69090852e-01 6.22427789e-04 -2.85484284e-01 2.25805745e-01
2.10563049e-01 6.58949018e-01 9.30194139e-01 -1.24772179e+00
-6.57564282e-01 -4.15893555e-01 -1.21533819e-01 1.12419672e-01
-4.60184783e-01 -2.53951848e-01 1.03558294e-01 -3.49033952e-01
-7.39555284e-02 -1.02080977e+00 -8.40920657e-02 -2.48431504e-01
-1.67490929e-01 -2.76675195e-01 8.78521800e-01 1.68384105e-01
8.30795884e-01 -2.31689358e+00 -2.89923668e-01 1.28640488e-01
1.94659740e-01 4.64035064e-01 -7.30320737e-02 2.87573159e-01
-4.41026270e-01 -3.33348721e-01 -1.72071233e-01 1.92151606e-01
-3.88791412e-01 3.64795327e-01 -3.45228165e-01 4.90668803e-01
1.93736017e-01 8.03842485e-01 -1.25103939e+00 -3.81118476e-01
2.15119675e-01 1.05365351e-01 -3.66947174e-01 9.46709588e-02
8.70056301e-02 1.08387098e-01 -1.92223251e-01 6.47513151e-01
3.23688209e-01 -2.08947361e-01 2.35233102e-02 -3.65464315e-02
2.32064396e-01 5.88287972e-03 -8.51850629e-01 1.49822414e+00
-1.40215725e-01 6.39742732e-01 -5.91470301e-01 -1.15774906e+00
1.05557644e+00 2.74474084e-01 2.51970798e-01 -6.01015508e-01
3.57597649e-01 1.48515791e-01 2.26193607e-01 -4.13197517e-01
-1.47943730e-02 -9.37984526e-01 7.52450153e-02 6.60815299e-01
7.40326524e-01 -2.04058751e-01 2.04957008e-01 1.78378329e-01
9.90933955e-01 -5.92860207e-02 7.02093005e-01 -1.44733027e-01
-9.04051065e-02 3.10704380e-01 6.72162294e-01 9.31735098e-01
-5.17381012e-01 6.01014435e-01 1.89930812e-01 -4.25253659e-01
-9.17315781e-01 -1.08768129e+00 -3.98507655e-01 1.54133654e+00
5.03619276e-02 -3.65619183e-01 -6.01845682e-01 -1.12335396e+00
2.24774480e-02 1.02989054e+00 -9.84761000e-01 -8.02283883e-01
-7.49195144e-02 -7.16074049e-01 1.68890536e-01 6.52116895e-01
2.77873695e-01 -1.06568575e+00 -7.74818838e-01 2.05343738e-01
2.62134284e-01 -6.98331535e-01 -2.47672379e-01 7.92414367e-01
-9.55391943e-01 -1.27937841e+00 -5.77207029e-01 -7.24266469e-01
9.06937361e-01 5.80006480e-01 9.09032285e-01 -1.38145864e-01
-4.75114018e-01 5.22836685e-01 -7.92489350e-01 -5.93390167e-01
-3.49098176e-01 -6.84817061e-02 1.64568424e-01 9.10471752e-02
9.30330396e-01 -5.53288341e-01 -1.80501640e-01 3.40306908e-01
-9.34232175e-01 -2.24495113e-01 5.14303744e-01 1.18461144e+00
3.96113694e-01 1.37102557e-02 8.23911965e-01 -1.37291312e+00
2.92556614e-01 -7.76928365e-01 -9.08519849e-02 3.36965621e-01
-7.57843316e-01 1.19856432e-01 5.15598536e-01 -8.33423674e-01
-1.14986849e+00 2.62327701e-01 4.34846133e-01 -4.12011474e-01
-3.58474225e-01 4.15486693e-01 1.25217792e-02 -1.23712607e-01
1.41790318e+00 6.93512261e-02 2.34423935e-01 -3.13501298e-01
5.14995039e-01 7.92821765e-01 1.19953275e-01 -3.47764015e-01
6.86718583e-01 5.06854892e-01 -2.70200193e-01 -9.17000651e-01
-1.31053185e+00 -6.29411578e-01 -9.27959800e-01 -1.99353188e-01
5.07369339e-01 -6.88350439e-01 2.46026337e-01 2.18873113e-01
-5.69815099e-01 -2.73009986e-01 -1.05358398e+00 5.62741995e-01
-6.39754236e-01 -1.13793109e-02 -3.40482354e-01 -5.68342268e-01
-1.00643151e-01 -8.32114279e-01 5.56898952e-01 4.00217950e-01
-4.41448390e-01 -9.91707265e-01 2.45824814e-01 1.20584324e-01
1.61191761e-01 5.96995093e-02 7.85972297e-01 -1.35825932e+00
-5.66633604e-02 -6.64427400e-01 2.16042530e-02 4.81378585e-01
4.92271543e-01 -3.93947542e-01 -1.35318553e+00 -3.60796422e-01
1.46163300e-01 -8.99607778e-01 1.04653478e+00 1.33989707e-01
4.66746986e-01 9.84933749e-02 -4.27067310e-01 3.05694103e-01
1.31262183e+00 3.33229423e-01 4.95005190e-01 1.75385445e-01
3.82027268e-01 7.31083870e-01 8.08166921e-01 3.79809231e-01
-1.50968447e-01 3.10240895e-01 -9.45855305e-02 1.13663711e-01
-7.67149553e-02 -2.20510066e-01 -5.54153845e-02 4.60659683e-01
1.74822658e-01 2.81511635e-01 -8.96061242e-01 4.38635319e-01
-1.70865190e+00 -1.03994966e+00 5.53017497e-01 2.32411623e+00
1.10843897e+00 2.24521309e-01 1.11975498e-01 1.66444808e-01
1.03765488e+00 -1.23476654e-01 -7.41237581e-01 -8.75012949e-02
6.30134996e-03 2.73272783e-01 2.77086139e-01 1.12577230e-01
-1.16273606e+00 9.57696915e-01 6.62881708e+00 7.76424408e-01
-1.09289503e+00 1.64381355e-01 5.78518510e-01 -3.82484615e-01
6.03936538e-02 1.94933519e-01 -9.53970969e-01 3.54113221e-01
9.18897629e-01 -3.91268313e-01 1.98740035e-01 1.22371304e+00
-3.17710131e-01 -2.25367755e-01 -1.40993476e+00 8.66892457e-01
4.85839963e-01 -1.24438453e+00 -1.14270553e-01 -7.12060854e-02
8.46319795e-01 9.78131667e-02 -1.24399237e-01 6.59182847e-01
3.36669117e-01 -6.27118111e-01 2.69521356e-01 2.57698804e-01
8.29114854e-01 -6.77770913e-01 7.04135895e-01 3.84360373e-01
-6.18398964e-01 -3.01361889e-01 -5.58638215e-01 -1.80781975e-01
-2.66526639e-01 5.25061965e-01 -1.14606869e+00 -8.38469192e-02
6.32900596e-01 7.85312951e-01 -6.47805512e-01 1.04560161e+00
-9.62033421e-02 7.12899506e-01 -1.45493269e-01 -8.80992562e-02
3.01426291e-01 9.25169438e-02 2.92277426e-01 9.24102247e-01
-2.77390443e-02 2.84076273e-01 1.35966018e-01 5.32156587e-01
-1.63109556e-01 9.97072533e-02 -9.56597388e-01 -1.10267878e-01
3.71639550e-01 1.13043094e+00 -9.34756637e-01 -7.35457778e-01
-5.18926322e-01 6.66976035e-01 5.03354967e-01 3.02942783e-01
-4.23853546e-01 -5.64186692e-01 2.28331894e-01 2.81043112e-01
3.70717406e-01 4.29500580e-01 -6.39581904e-02 -1.14299059e+00
-3.41310292e-01 -5.71431041e-01 6.00366354e-01 -7.43306816e-01
-1.62780142e+00 4.26062793e-01 3.07688832e-01 -1.39093256e+00
-3.18071723e-01 -3.55680406e-01 -6.79756582e-01 4.70056117e-01
-1.12143910e+00 -7.96942294e-01 -2.99965113e-01 5.16635418e-01
5.98170519e-01 -5.18583775e-01 9.60412860e-01 -5.07209487e-02
-2.88839251e-01 5.68438470e-01 3.71664643e-01 1.64429918e-01
1.02594483e+00 -1.09597421e+00 -2.61785239e-01 4.49409783e-01
2.20235661e-01 4.90686685e-01 7.85558939e-01 -6.11761510e-01
-8.17137599e-01 -1.08915055e+00 7.10098863e-01 -3.14548939e-01
5.51891685e-01 -2.02671349e-01 -1.19681287e+00 7.12639868e-01
4.53737145e-03 4.12463695e-01 1.34029794e+00 1.86986268e-01
-5.88719308e-01 -2.61332184e-01 -1.40440476e+00 3.53419751e-01
9.43067789e-01 -4.66810763e-01 -1.09245193e+00 3.71101022e-01
5.59538901e-01 1.29167542e-01 -5.07695973e-01 2.50601709e-01
3.11983913e-01 -9.53888774e-01 7.46311903e-01 -9.43348050e-01
3.65610868e-01 4.03680056e-02 -2.08082914e-01 -1.73055053e+00
-3.74362588e-01 -6.76777214e-02 -2.40337048e-02 1.13606572e+00
2.57102609e-01 -6.90086126e-01 8.66757393e-01 6.82481349e-01
-3.68984938e-02 -6.16605163e-01 -7.25203097e-01 -9.66294229e-01
8.12263638e-02 -1.47488028e-01 3.49307209e-01 1.33914649e+00
4.36826199e-01 7.75410652e-01 -2.45428115e-01 -4.88996178e-01
6.14066958e-01 2.41733953e-01 5.71047723e-01 -1.38320863e+00
-1.90623268e-01 -6.17957376e-02 -7.68645704e-01 -3.07339072e-01
2.70501047e-01 -1.03399587e+00 3.41377109e-01 -1.04834855e+00
5.40479124e-01 -4.01360154e-01 -4.54769939e-01 7.32520163e-01
-2.41484195e-01 4.53939497e-01 2.24884167e-01 2.08919317e-01
-7.31001914e-01 6.56235039e-01 1.03415740e+00 -1.39230594e-01
-5.20768702e-01 -3.11903842e-02 -8.16360235e-01 7.70262003e-01
9.22550797e-01 -6.78313613e-01 -6.82313621e-01 1.78375348e-01
-3.26790065e-01 -3.34178448e-01 -5.17361984e-02 -1.01071119e+00
2.05399618e-01 -1.35856569e-01 5.18513858e-01 -2.29753196e-01
3.75367105e-01 -6.04772627e-01 -2.65016377e-01 3.87620002e-01
-5.47016978e-01 -7.45329797e-01 2.38184463e-02 7.94564903e-01
-1.50007173e-01 -7.58002937e-01 1.19940257e+00 -3.18474919e-01
-9.98266220e-01 1.15990534e-01 -3.58705461e-01 3.47140998e-01
1.44958937e+00 -4.83741105e-01 -3.51507187e-01 -1.75576538e-01
-1.07725310e+00 -1.98970884e-01 4.89989340e-01 3.47822607e-01
6.91313326e-01 -1.45748281e+00 -3.82623315e-01 2.99756467e-01
8.44379783e-01 -3.18999112e-01 5.62740304e-02 6.74021602e-01
1.30551802e-02 1.34642497e-01 -3.53446901e-01 -4.79736775e-01
-1.16661036e+00 8.27093542e-01 1.76946610e-01 2.40915537e-01
-6.24451280e-01 7.45013237e-01 1.75931558e-01 -4.48446929e-01
9.12437290e-02 1.80279493e-01 -9.02707651e-02 4.88594502e-01
6.27083242e-01 4.57205802e-01 -2.19881348e-02 -5.12064397e-01
-1.77367657e-01 4.43735495e-02 -4.08322036e-01 4.03564833e-02
1.40910006e+00 9.47767049e-02 3.06628764e-01 1.04872596e+00
1.35512733e+00 -4.90713924e-01 -1.16233504e+00 -7.83784449e-01
9.22214985e-02 -5.53376317e-01 3.37897539e-02 -7.67051280e-01
-9.31908011e-01 7.63566077e-01 8.18373740e-01 4.38566655e-02
7.61889160e-01 2.93590128e-01 4.74471450e-01 6.14687502e-01
5.04709482e-01 -1.37913394e+00 2.60404706e-01 3.70244771e-01
4.15574431e-01 -1.60120189e+00 -3.57430391e-02 -3.28517795e-01
-7.79705584e-01 9.94628012e-01 7.41731763e-01 -1.88836485e-01
9.31366563e-01 -1.68729454e-01 3.15633998e-03 -1.90029755e-01
-7.50744522e-01 -5.17282069e-01 1.16999261e-01 8.98884237e-01
2.45952979e-01 -1.77157298e-02 4.06698920e-02 4.85193044e-01
4.05640937e-02 1.56039938e-01 5.66977739e-01 1.42686141e+00
-8.21869612e-01 -8.14175010e-01 -2.31339484e-01 9.44489896e-01
1.14302047e-01 -1.18535109e-01 -4.97944146e-01 7.34712601e-01
2.05903754e-01 9.01403546e-01 1.80122450e-01 -3.66431594e-01
2.30392128e-01 5.89325666e-01 4.86392498e-01 -1.30373526e+00
-2.79233754e-01 -2.19741613e-01 -1.72223113e-02 -1.94053143e-01
-5.43426633e-01 -5.93263924e-01 -1.14836931e+00 1.46871740e-02
-6.94161057e-01 2.82122374e-01 2.88394600e-01 1.12615752e+00
2.35889614e-01 1.75354004e-01 7.80712366e-01 -6.95489049e-01
-8.07076097e-01 -1.12013555e+00 -8.97431612e-01 7.92422235e-01
1.95939615e-01 -1.02907491e+00 -7.68647909e-01 1.32327527e-01]
|
[9.978893280029297, 2.990468740463257]
|
c0718dea-69c5-4368-83e4-2c4c01eb0609
|
multiple-sequence-alignment-is-not-a-solved
|
1808.07717
| null |
http://arxiv.org/abs/1808.07717v1
|
http://arxiv.org/pdf/1808.07717v1.pdf
|
Multiple Sequence Alignment is not a Solved Problem
|
Multiple sequence alignment is a basic procedure in molecular biology, and it
is often treated as being essentially a solved computational problem. However,
this is not so, and here I review the evidence for this claim, and outline the
requirements for a solution. The goal of alignment is often stated to be to
juxtapose nucleotides (or their derivatives, such as amino acids) that have
been inherited from a common ancestral nucleotide (although other goals are
also possible). Unfortunately, this is not an operational definition, because
homology (in this sense) refers to unique and unobservable historical events,
and so there can be no objective mathematical function to optimize.
Consequently, almost all algorithms developed for multiple sequence alignment
are based on optimizing some sort of compositional similarity (similarity =
homology + analogy). As a result, many, if not most, practitioners either
manually modify computer-produced alignments or they perform de novo manual
alignment, especially in the field of phylogenetics. So, if homology is the
goal, then multiple sequence alignment is not yet a solved computational
problem. Several criteria have been developed by biologists to help them
identify potential homologies (compositional, ontogenetic, topographical and
functional similarity, plus conjunction and congruence), and these criteria can
be applied to molecular data, in principle. Current computer programs do
implement one (or occasionally two) of these criteria, but no program
implements them all. What is needed is a program that evaluates all of the
evidence for the sequence homologies, optimizes their combination, and thus
produces the best hypotheses of homology. This is basically an inference
problem not an optimization problem.
|
[]
|
2018-08-23
| null | null | null | null |
['multiple-sequence-alignment']
|
['medical']
|
[ 5.16969621e-01 -2.89172024e-01 -9.95266214e-02 -2.74601668e-01
-2.43715525e-01 -8.22349429e-01 3.67984176e-01 3.72089356e-01
-4.07481283e-01 9.73270774e-01 -8.39188695e-02 -7.15796053e-01
-2.30900109e-01 -4.51120943e-01 -4.40498143e-01 -1.02930558e+00
7.70471320e-02 6.72281861e-01 4.52155232e-01 -4.04775232e-01
7.70944059e-01 7.31407404e-01 -1.56709278e+00 -1.89238295e-01
9.39742684e-01 7.52729177e-02 5.72844803e-01 8.18854928e-01
-4.45872217e-01 2.40342319e-01 -6.47490442e-01 -5.47283471e-01
5.65434806e-02 -1.05339575e+00 -1.11039507e+00 -4.20476943e-02
8.98796692e-02 1.27750605e-01 4.87899929e-01 8.17330480e-01
3.43407840e-01 -1.61570400e-01 7.54419386e-01 -9.99110639e-01
-4.51447338e-01 1.65976882e-01 -2.98323363e-01 2.11100012e-01
5.62267184e-01 1.46563143e-01 1.04910505e+00 -5.04176438e-01
7.26788819e-01 8.51489365e-01 7.06552744e-01 2.24623308e-01
-1.34350359e+00 -1.85535878e-01 -2.47592747e-01 1.65206432e-01
-1.42657971e+00 -1.35489970e-01 1.83978185e-01 -6.11952722e-01
9.72958982e-01 7.09940255e-01 9.91911829e-01 6.70652807e-01
5.26039720e-01 2.81077236e-01 1.08337605e+00 -6.09532654e-01
1.12319022e-01 3.97783630e-02 -8.41860697e-02 3.73077691e-01
2.69189149e-01 -1.49154589e-01 -2.84272611e-01 -4.48909789e-01
6.00348711e-01 -2.89281636e-01 -4.57444787e-01 -2.28279561e-01
-1.14694786e+00 7.85769045e-01 -6.79566339e-02 8.05957496e-01
-3.12165171e-01 -2.48468965e-01 2.99932659e-01 3.61422002e-01
-8.01503584e-02 8.35374951e-01 -5.40459216e-01 -2.25616083e-01
-9.37087476e-01 4.23911542e-01 9.61862922e-01 4.61820513e-01
7.64852405e-01 -2.34642431e-01 8.11904192e-01 6.63754284e-01
1.46111906e-01 -2.75541190e-02 6.62061930e-01 -7.46034443e-01
-2.85397112e-01 5.44606030e-01 2.16956586e-02 -1.02251482e+00
-4.88682717e-01 -8.77525806e-02 -5.44206083e-01 3.49747092e-01
7.12683499e-01 3.10537755e-01 -5.47399819e-01 1.83554447e+00
4.37471837e-01 -1.14410380e-02 1.24349184e-02 7.70951569e-01
3.50551397e-01 6.86567128e-01 7.56046399e-02 -7.57450998e-01
1.23188651e+00 -4.53133315e-01 -5.97071946e-01 -4.73536514e-02
5.63951731e-01 -1.16735733e+00 8.26707006e-01 3.30983758e-01
-1.00287473e+00 -2.38987237e-01 -1.21680224e+00 -7.10593536e-02
-3.52752507e-01 -6.44024372e-01 6.39327765e-01 5.90641201e-01
-9.73820448e-01 9.93293107e-01 -6.12255991e-01 -7.38280237e-01
-5.52690864e-01 3.95559579e-01 -4.46754754e-01 4.49307650e-01
-1.11852467e+00 1.45806956e+00 6.28095090e-01 4.74007130e-02
-1.01699591e-01 -1.79276407e-01 -4.60612237e-01 -2.05282718e-01
1.22326724e-01 -9.33419526e-01 8.34171176e-01 -1.17448783e+00
-1.23324132e+00 1.29238188e+00 -3.21662962e-01 -9.09564942e-02
2.88119912e-01 3.49654853e-01 -2.44050607e-01 -2.10742317e-02
3.85347232e-02 3.97622973e-01 2.17535764e-01 -1.18805516e+00
-5.82041144e-01 -4.21374261e-01 -1.80657938e-01 3.11119378e-01
2.75419682e-01 4.31487471e-01 -9.12918597e-02 -5.91784835e-01
4.87951040e-01 -8.93215001e-01 -3.68491977e-01 -1.80935264e-02
4.58433442e-02 -2.78161019e-01 1.82864323e-01 -9.59435403e-01
1.13855278e+00 -2.08255029e+00 5.48149765e-01 3.01040530e-01
-3.79455984e-02 1.96636781e-01 7.83888400e-02 7.98440397e-01
-3.83108050e-01 3.93526584e-01 -6.05287731e-01 3.74845058e-01
-2.86555469e-01 4.34885681e-01 -1.61393389e-01 6.70842946e-01
-5.49995266e-02 4.77546751e-01 -1.02158082e+00 -5.10917962e-01
1.28661633e-01 2.30641082e-01 -1.51852146e-01 -1.15236782e-01
-9.60517302e-02 4.01689142e-01 -2.92375475e-01 6.06281102e-01
4.17862117e-01 -9.98005792e-02 7.75518656e-01 1.09388404e-01
-5.55215538e-01 3.07470471e-01 -9.87279177e-01 1.36555934e+00
9.63977277e-02 4.76630241e-01 -1.48388013e-01 -1.03266776e+00
1.08011913e+00 5.41050375e-01 7.11836517e-01 -2.00784683e-01
-5.18709421e-02 8.18594635e-01 5.71671844e-01 -5.22947729e-01
3.54887962e-01 -4.29097831e-01 3.20103645e-01 2.98716545e-01
-1.88904986e-01 -3.64180535e-01 3.43008250e-01 -2.20280558e-01
1.02978539e+00 3.04880261e-01 9.66789782e-01 -3.20829272e-01
7.46823132e-01 4.60266292e-01 9.23837602e-01 4.49214816e-01
-1.05794951e-01 4.94572937e-01 4.37255442e-01 -5.34864664e-01
-1.66357446e+00 -9.18367863e-01 -4.04558599e-01 7.21702158e-01
2.44822681e-01 -2.44294584e-01 -7.36174881e-01 -1.50606930e-02
-2.10182220e-01 5.34399092e-01 -2.67449141e-01 1.14216663e-01
-7.64613569e-01 -1.03386438e+00 4.62623060e-01 -1.75480470e-02
-2.19918177e-01 -1.03994548e+00 -9.19751167e-01 5.91845036e-01
-4.68891859e-02 -4.33817863e-01 -6.86056614e-02 3.82710993e-01
-8.66545975e-01 -1.18864989e+00 -5.98383129e-01 -7.85328507e-01
3.91023278e-01 3.15887600e-01 8.36564779e-01 6.02817178e-01
-4.43982989e-01 -1.28599033e-01 -4.43568259e-01 -2.48263687e-01
-7.34993458e-01 -1.60066947e-01 1.34254068e-01 -5.15939832e-01
3.88321847e-01 -1.02425647e+00 -2.14454576e-01 5.79232275e-01
-1.06897902e+00 -8.78620073e-02 5.24831593e-01 9.47974980e-01
5.38132489e-01 -2.56413165e-02 5.45288444e-01 -5.63780248e-01
5.75952888e-01 -2.34528929e-01 -4.66133982e-01 6.81579590e-01
-3.74084145e-01 -1.28437519e-01 4.71402019e-01 -2.84544826e-01
-5.68889558e-01 1.66584268e-01 -5.65340996e-01 1.02175876e-01
-3.50918055e-01 7.09742665e-01 -2.81734079e-01 -1.44643620e-01
6.61487937e-01 5.90980709e-01 2.79097617e-01 -3.59155983e-01
5.18484749e-02 6.22070909e-01 5.09637952e-01 -6.22006893e-01
5.73474050e-01 2.70592719e-01 1.01525135e-01 -1.11813045e+00
-9.78894457e-02 -5.36491871e-01 -9.14910853e-01 -1.60243109e-01
9.51717794e-01 -2.31357023e-01 -7.76585996e-01 2.13246211e-01
-1.18796921e+00 -2.24492606e-02 5.07091321e-02 5.77785552e-01
-7.15187550e-01 8.14031363e-01 -3.55613470e-01 -7.10685074e-01
-1.18004411e-01 -1.34116316e+00 5.33025324e-01 1.54670149e-01
-6.88593626e-01 -8.56543183e-01 2.82691330e-01 1.83422506e-01
1.63160250e-01 4.53618109e-01 1.33068895e+00 -8.02002907e-01
-2.67828882e-01 -1.09954573e-01 3.59398991e-01 4.86966176e-03
4.10927415e-01 7.07962751e-01 -3.78507107e-01 -1.07806757e-01
3.09246749e-01 -8.15110467e-03 3.05899084e-01 2.46694893e-01
5.40806890e-01 -2.81446606e-01 -4.12265122e-01 4.46132839e-01
1.37954545e+00 8.04926157e-01 7.40241289e-01 5.71274579e-01
1.81232989e-01 1.07358265e+00 6.50436819e-01 1.18291184e-01
-2.16033217e-03 9.54777002e-01 2.46655375e-01 2.04488523e-02
4.35316294e-01 2.00031057e-01 2.64591664e-01 8.87748718e-01
-4.25739408e-01 -2.16474026e-01 -1.01704562e+00 4.29040968e-01
-1.72927225e+00 -1.33537591e+00 -6.19920969e-01 2.31421661e+00
1.03293896e+00 -4.43866998e-02 3.77722204e-01 1.89233407e-01
8.55499864e-01 -2.43466541e-01 -5.45389175e-01 -8.03558469e-01
-3.68773013e-01 1.21212609e-01 3.96709889e-01 6.17857933e-01
-6.70440316e-01 7.54822135e-01 7.58228970e+00 6.43149674e-01
-1.18965650e+00 -4.28403497e-01 2.73259610e-01 3.15869361e-01
-4.86211926e-01 4.86090064e-01 -4.97422904e-01 4.52102005e-01
7.40725875e-01 -4.65852380e-01 3.08830947e-01 7.05304503e-01
1.74042478e-01 -1.69634670e-01 -8.26758981e-01 5.61773181e-01
-1.74120307e-01 -1.21675789e+00 -3.54517810e-02 3.36388081e-01
1.88505262e-01 -3.27153921e-01 -3.31224620e-01 -4.56133872e-01
2.36624584e-01 -9.66806054e-01 5.87163508e-01 3.58827770e-01
1.86525732e-01 -7.51465023e-01 7.47065783e-01 4.80723888e-01
-8.47152948e-01 4.11600769e-01 -4.17466730e-01 -6.91928715e-02
6.72144592e-01 4.17647749e-01 -7.98149943e-01 7.07935989e-01
3.92775536e-01 1.91921979e-01 -3.77519339e-01 1.30535126e+00
-3.28769654e-01 2.75945157e-01 -3.68166298e-01 -1.96565226e-01
2.52761275e-01 -6.02528036e-01 7.43691981e-01 1.07673991e+00
3.26176494e-01 2.74944812e-01 3.32500935e-02 5.05486012e-01
7.96047509e-01 6.04106009e-01 -4.29908931e-01 -1.71602443e-01
5.17699540e-01 9.50002968e-01 -9.85420346e-01 -2.20069960e-01
-4.52862054e-01 1.09203947e+00 -4.51777875e-02 -1.72764286e-02
-7.69432962e-01 -2.34477252e-01 1.00438738e+00 -1.95816606e-02
-5.21483570e-02 -2.48290077e-01 -1.85001805e-01 -7.74210751e-01
-6.01506606e-02 -1.23650146e+00 3.53141606e-01 -6.11667991e-01
-1.17081118e+00 3.98897827e-01 -5.39766923e-02 -8.36381197e-01
-5.74691713e-01 -7.15015590e-01 -4.50342983e-01 1.05095351e+00
-8.05321753e-01 -8.56197774e-01 2.48703524e-01 1.25574380e-01
2.35805541e-01 2.19739795e-01 8.90129745e-01 8.11274797e-02
-3.21072221e-01 1.85188219e-01 2.78159648e-01 -3.62565875e-01
6.39734507e-01 -1.14747989e+00 3.00241679e-01 6.79722726e-01
-9.23141651e-03 1.08290195e+00 1.17928159e+00 -8.34375143e-01
-1.09080565e+00 -3.48053932e-01 1.24910200e+00 -1.57241657e-01
7.53323138e-01 7.84329772e-02 -1.29516077e+00 5.48039794e-01
1.97362766e-01 -9.31028545e-01 1.09657156e+00 -5.65783493e-02
-4.16318588e-02 4.60498959e-01 -9.89718497e-01 9.06567454e-01
9.67644870e-01 -2.75094330e-01 -8.65904272e-01 4.35225397e-01
6.00915670e-01 -1.06592350e-01 -1.11698675e+00 3.98573279e-01
8.53052497e-01 -1.19026494e+00 9.95243430e-01 -7.56087840e-01
9.91830975e-02 -8.36961687e-01 -1.36048526e-01 -8.94090593e-01
-4.96043146e-01 -5.18545926e-01 7.79680252e-01 1.11501992e+00
5.05554616e-01 -7.01132298e-01 6.35343850e-01 3.68103087e-01
-4.72713083e-01 -5.85654974e-01 -9.26064789e-01 -9.98862803e-01
8.72508511e-02 1.27757266e-01 6.47841811e-01 1.44337153e+00
2.95492977e-01 3.90235335e-01 -2.68607348e-01 -2.87860129e-02
4.03893381e-01 1.58830225e-01 7.97674716e-01 -1.13185918e+00
-4.18244541e-01 -9.52031732e-01 -7.86511421e-01 -5.54607213e-01
1.32680424e-02 -6.76621556e-01 -3.79508398e-02 -1.43850040e+00
-4.25554700e-02 -2.67806023e-01 2.72937924e-01 2.88850695e-01
-3.15681964e-01 -1.92157730e-01 -1.58352345e-01 4.83145505e-01
3.87530357e-01 -3.87558937e-02 8.35535765e-01 3.87628645e-01
-1.45124406e-01 -1.43979967e-01 -5.29347479e-01 8.71608853e-01
9.38755751e-01 -3.86084795e-01 -2.52079573e-02 9.96787697e-02
4.46964800e-01 8.28764215e-02 1.28974080e-01 -7.44656563e-01
1.31056249e-01 -8.12894344e-01 9.95549858e-02 -4.91481334e-01
2.25321665e-01 -7.02662766e-01 1.09848320e+00 8.89945567e-01
6.91262931e-02 4.29546237e-01 -1.02973074e-01 1.49616733e-01
-1.76645309e-01 -8.28394949e-01 8.77344251e-01 -4.11945730e-01
-8.67557883e-01 -2.37737879e-01 -7.05177784e-01 -5.49844205e-01
1.22121179e+00 -7.43543863e-01 -2.02311113e-01 -1.22768961e-01
-6.34407341e-01 3.92700806e-02 1.19247639e+00 -2.29916889e-02
4.03724402e-01 -8.38751912e-01 -6.96629524e-01 -2.87710667e-01
1.22420765e-01 -3.62403154e-01 3.85039821e-02 9.28080201e-01
-1.24873340e+00 2.46663094e-01 -5.13279676e-01 -5.43785095e-01
-1.61203730e+00 8.96638870e-01 3.23813379e-01 4.81556319e-02
-3.57835293e-01 7.49912202e-01 6.49920255e-02 -3.91187549e-01
-2.68485636e-01 2.13261321e-01 -2.23337844e-01 6.29100874e-02
4.67289984e-01 1.05617851e-01 -1.51880175e-01 -8.79489005e-01
-4.34962511e-01 7.02799380e-01 3.76839750e-02 -2.09455177e-01
1.32568800e+00 -1.38261706e-01 -6.82398617e-01 4.80838656e-01
9.21053946e-01 4.09279913e-02 -4.12034959e-01 1.28406420e-01
3.18181008e-01 -5.55002093e-01 -8.52176845e-01 -4.25369114e-01
-3.15936863e-01 5.61519682e-01 1.30867481e-01 2.78623343e-01
1.08701873e+00 -8.49985555e-02 3.56307328e-01 2.77172744e-01
2.74960071e-01 -8.61187458e-01 -4.20855880e-01 3.93245965e-01
8.09955239e-01 -6.25360429e-01 1.33377612e-01 -5.51261067e-01
-3.42024773e-01 1.18038416e+00 3.57414752e-01 1.60650700e-01
5.56554645e-02 1.29466981e-01 -4.37075570e-02 -9.22543555e-02
-6.81626856e-01 -2.64042825e-01 -1.20721139e-01 3.89147609e-01
8.40055943e-01 -2.45732337e-01 -1.40512621e+00 -2.56387621e-01
-5.79338551e-01 -2.76629269e-01 3.69279951e-01 1.09977019e+00
-9.03696835e-01 -1.59534299e+00 -6.55542970e-01 6.57968596e-02
-4.45120126e-01 -4.12954539e-02 -9.00491714e-01 9.22706902e-01
3.24963629e-01 6.81596875e-01 -7.02503370e-03 -2.93373495e-01
1.90605223e-02 2.69577593e-01 5.49813688e-01 -3.24778229e-01
-7.79938459e-01 2.80129433e-01 1.33875072e-01 -1.07328460e-01
-5.15734673e-01 -8.12552810e-01 -1.23386872e+00 -7.05798030e-01
-5.15872478e-01 6.02471709e-01 9.50706005e-01 9.91393685e-01
-1.83114752e-01 1.16636746e-01 3.36844951e-01 -4.88451511e-01
-2.57125467e-01 -5.69144130e-01 -5.44451237e-01 2.73953617e-01
-5.83450459e-02 -3.59479815e-01 -4.20080930e-01 1.66567236e-01]
|
[4.8786492347717285, 5.230973720550537]
|
5d895066-ca2c-428c-bef9-de941eac07e9
|
vint-a-foundation-model-for-visual-navigation
|
2306.14846
| null |
https://arxiv.org/abs/2306.14846v1
|
https://arxiv.org/pdf/2306.14846v1.pdf
|
ViNT: A Foundation Model for Visual Navigation
|
General-purpose pre-trained models ("foundation models") have enabled practitioners to produce generalizable solutions for individual machine learning problems with datasets that are significantly smaller than those required for learning from scratch. Such models are typically trained on large and diverse datasets with weak supervision, consuming much more training data than is available for any individual downstream application. In this paper, we describe the Visual Navigation Transformer (ViNT), a foundation model that aims to bring the success of general-purpose pre-trained models to vision-based robotic navigation. ViNT is trained with a general goal-reaching objective that can be used with any navigation dataset, and employs a flexible Transformer-based architecture to learn navigational affordances and enable efficient adaptation to a variety of downstream navigational tasks. ViNT is trained on a number of existing navigation datasets, comprising hundreds of hours of robotic navigation from a variety of different robotic platforms, and exhibits positive transfer, outperforming specialist models trained on singular datasets. ViNT can be augmented with diffusion-based subgoal proposals to explore novel environments, and can solve kilometer-scale navigation problems when equipped with long-range heuristics. ViNT can also be adapted to novel task specifications with a technique inspired by prompt-tuning, where the goal encoder is replaced by an encoding of another task modality (e.g., GPS waypoints or routing commands) embedded into the same space of goal tokens. This flexibility and ability to accommodate a variety of downstream problem domains establishes ViNT as an effective foundation model for mobile robotics. For videos, code, and model checkpoints, see our project page at https://visualnav-transformer.github.io.
|
['Sergey Levine', 'Noriaki Hirose', 'Kevin Black', 'Kyle Stachowicz', 'Nitish Dashora', 'Ajay Sridhar', 'Dhruv Shah']
|
2023-06-26
| null | null | null | null |
['visual-navigation']
|
['robots']
|
[-1.67778596e-01 2.82182634e-01 -2.48837277e-01 -2.24595010e-01
-6.27635181e-01 -9.73650575e-01 5.06780386e-01 -3.43818069e-01
-4.63983506e-01 4.38961565e-01 1.97340116e-01 -6.95013165e-01
-3.35066408e-01 -4.79578793e-01 -1.01449573e+00 -3.26597244e-01
-2.96674103e-01 5.95880151e-01 2.71474510e-01 -9.24238741e-01
2.74587095e-01 3.35758954e-01 -1.67310178e+00 3.09761036e-02
7.75941491e-01 7.87258744e-01 1.03644753e+00 6.44038856e-01
1.79702669e-01 4.94353890e-01 -1.38453990e-01 1.47427082e-01
6.85608685e-01 1.58837423e-01 -9.36071873e-01 -2.39037260e-01
3.04591805e-01 -2.28961393e-01 -6.42926574e-01 4.97214705e-01
4.28190351e-01 6.29778206e-01 5.67526579e-01 -1.39953089e+00
-1.00910437e+00 3.44293028e-01 1.36058837e-01 8.04878548e-02
4.89019603e-01 6.63903296e-01 9.26376104e-01 -7.45382130e-01
8.80532980e-01 1.16470456e+00 9.98695195e-01 9.01301444e-01
-1.08066952e+00 -1.06687516e-01 3.90067190e-01 1.37241751e-01
-8.69966567e-01 -3.71156991e-01 3.31193149e-01 -4.05283421e-01
1.43342841e+00 -9.80084836e-02 7.17112601e-01 1.42553449e+00
3.63994539e-01 8.22345018e-01 4.29870874e-01 3.43928821e-02
3.20794404e-01 -2.30802909e-01 -3.13942403e-01 8.83140743e-01
5.04291169e-02 2.27247790e-01 -3.58660936e-01 1.34772018e-01
7.25293517e-01 1.09695539e-01 -4.02932733e-01 -1.12388194e+00
-1.41138041e+00 7.55837858e-01 1.07652771e+00 -6.64159330e-03
-2.35625640e-01 5.25268018e-01 4.35162783e-01 5.44710338e-01
-2.95775533e-01 1.02709568e+00 -1.01396811e+00 -3.20365191e-01
-1.32012814e-01 4.50537473e-01 8.17510009e-01 1.61782920e+00
9.01535511e-01 1.12894043e-01 3.24130058e-01 6.62845254e-01
3.07634205e-01 4.64243472e-01 9.19374704e-01 -1.47618580e+00
6.58834577e-01 4.53979164e-01 1.56688169e-01 -6.99381709e-01
-9.02623653e-01 -3.14050317e-01 -2.21749619e-01 4.28107530e-01
4.45940763e-01 -3.42279047e-01 -1.30545509e+00 1.72970319e+00
4.01019782e-01 -5.16237557e-01 4.49624717e-01 9.40472364e-01
7.66538441e-01 6.57807350e-01 -1.63531095e-01 7.80409396e-01
9.56245720e-01 -1.67382812e+00 4.74095121e-02 -9.70581234e-01
1.27204537e+00 -1.80830777e-01 1.57504010e+00 4.30849433e-01
-6.50071323e-01 -4.91206795e-01 -1.00239730e+00 -4.94291544e-01
-7.46697724e-01 -1.50358960e-01 1.02380753e+00 1.48140892e-01
-1.51459944e+00 3.87612998e-01 -8.38198125e-01 -8.98483932e-01
3.70368659e-01 4.05227810e-01 -7.81432807e-01 -4.87915307e-01
-8.61317337e-01 1.27549076e+00 4.06183779e-01 1.65813193e-01
-1.37685251e+00 -4.15101260e-01 -1.25920725e+00 -2.33620748e-01
5.02792597e-01 -1.01304173e+00 1.54956651e+00 -6.83600008e-01
-1.52748656e+00 3.90329987e-01 1.95044681e-01 -3.63357067e-01
2.39809915e-01 -1.80396780e-01 4.95786034e-02 -1.31951600e-01
6.02989733e-01 1.05518413e+00 7.28317559e-01 -1.00405240e+00
-6.61383867e-01 -2.23512098e-01 5.40876448e-01 6.34942055e-01
-7.04718083e-02 -6.34211183e-01 -6.31328106e-01 -1.79980710e-01
2.63820767e-01 -1.37458014e+00 -7.51204073e-01 1.57732889e-01
-8.72909278e-02 -1.32419556e-01 1.00666487e+00 -5.15034378e-01
4.18372154e-01 -2.13173509e+00 6.77488923e-01 -1.86649352e-01
-1.53616285e-02 -1.65290162e-01 -6.46700323e-01 6.13077641e-01
2.42321879e-01 -1.67362113e-02 -1.11142114e-01 -1.89758256e-01
1.64617956e-01 5.77287316e-01 -2.04116702e-01 1.20693117e-01
-1.52322948e-01 1.11203277e+00 -1.22410691e+00 1.05042540e-01
2.41047904e-01 4.37716134e-02 -8.99710953e-01 -5.64532205e-02
-6.31238341e-01 6.95364475e-01 -5.72241843e-01 7.79146016e-01
1.71218980e-02 -1.86516657e-01 2.11910345e-03 2.11349264e-01
-4.25439924e-02 3.56291056e-01 -6.49970233e-01 2.67367482e+00
-9.43771482e-01 7.30529845e-01 9.05391052e-02 -9.24075663e-01
8.86424184e-01 -1.39179766e-01 1.76625118e-01 -7.93091834e-01
-8.93309936e-02 4.42281008e-01 1.48785645e-02 -8.11529040e-01
8.22262585e-01 3.44516784e-01 -5.32175183e-01 2.17720702e-01
3.53257030e-01 -6.53195679e-01 1.62577510e-01 1.14801049e-01
1.44744825e+00 7.38149285e-01 4.50256504e-02 -2.84228832e-01
-8.54444355e-02 7.13022232e-01 2.83827931e-01 1.02574861e+00
-2.16234952e-01 4.65595692e-01 1.19762436e-01 -7.38448918e-01
-1.11725700e+00 -9.57306743e-01 9.44585353e-02 1.55103183e+00
4.77237701e-01 -4.77418661e-01 -3.44338208e-01 -7.30758727e-01
3.61642659e-01 5.45076966e-01 -6.73569739e-01 -2.44582117e-01
-5.23390532e-01 -2.72291929e-01 4.42946881e-01 6.25041366e-01
4.70225632e-01 -1.21564925e+00 -9.41228211e-01 2.21602812e-01
-1.06306612e-01 -9.23920512e-01 -2.15045169e-01 7.62079179e-01
-9.53570783e-01 -1.02897573e+00 -5.01683354e-01 -1.18780458e+00
5.40428758e-01 6.65615678e-01 8.88153613e-01 1.53844915e-02
-1.49855286e-01 7.81376600e-01 -3.98581862e-01 -1.46165892e-01
-1.72396258e-01 6.08595669e-01 1.95281819e-01 -8.95397246e-01
-3.51790525e-02 -6.39178276e-01 -4.43508685e-01 4.47321802e-01
-6.74928248e-01 6.32718801e-02 6.64437950e-01 1.12038028e+00
3.12725127e-01 -4.20706153e-01 5.07826447e-01 -3.73947233e-01
6.79580271e-01 -9.23672557e-01 -4.64495450e-01 -4.58415784e-02
-6.21524036e-01 2.51204371e-01 7.78993785e-01 -5.08412242e-01
-7.67191470e-01 1.63358316e-01 3.09711397e-02 -2.76659310e-01
-1.65550709e-01 8.23502064e-01 4.28452669e-03 -2.49827266e-01
1.19908297e+00 2.97357917e-01 1.43586785e-01 -3.94656569e-01
8.67714226e-01 4.69480634e-01 6.40908003e-01 -8.81921649e-01
6.03609085e-01 3.13693732e-01 -1.54018059e-01 -5.75558603e-01
-5.61502218e-01 -2.60553449e-01 -5.53591788e-01 1.58579424e-01
5.92637300e-01 -9.97950017e-01 -5.34507990e-01 3.55865657e-01
-6.60088480e-01 -1.41887677e+00 -3.07389081e-01 3.26208472e-01
-1.02252197e+00 2.91562974e-02 -3.42709780e-01 -1.89613223e-01
7.96092376e-02 -1.32906890e+00 9.92685258e-01 4.18963224e-01
-1.70555547e-01 -1.16586423e+00 1.19577814e-03 2.12066963e-01
6.92591906e-01 -8.02242309e-02 6.69119656e-01 -5.27661860e-01
-7.47736990e-01 -2.11562261e-01 -2.20641512e-02 6.12681098e-02
5.36820516e-02 -5.12271225e-01 -4.70356733e-01 -4.44354028e-01
-4.00317848e-01 -8.37564945e-01 8.24513733e-01 1.50972068e-01
9.23838735e-01 -1.92625135e-01 -6.99215531e-01 1.00389540e+00
1.20406103e+00 1.93978667e-01 4.66086328e-01 1.28585219e+00
7.64765203e-01 4.31654096e-01 8.27204943e-01 -5.24218054e-03
9.85574305e-01 6.50485218e-01 1.07412887e+00 2.70931184e-01
8.51021856e-02 -4.15547609e-01 3.56484532e-01 2.20610648e-01
2.16758368e-03 -2.49026567e-01 -1.26167309e+00 9.34789956e-01
-2.15698528e+00 -6.51909769e-01 1.70372933e-01 1.77135062e+00
3.19402367e-01 4.20098417e-02 -5.51644601e-02 -6.33245289e-01
1.29245922e-01 5.65860048e-02 -1.11718082e+00 -5.37478566e-01
8.35780725e-02 -2.87873268e-01 7.39402056e-01 3.90145361e-01
-9.81840253e-01 1.36768174e+00 6.56209946e+00 4.07672048e-01
-1.06464040e+00 -2.71535915e-04 -7.44689032e-02 -7.95574784e-02
-3.89653981e-01 3.33748788e-01 -5.74482501e-01 1.26659870e-01
7.27725208e-01 7.39431754e-02 8.26895773e-01 1.42098212e+00
-3.98056507e-02 -1.08402714e-01 -1.22726023e+00 8.16319406e-01
6.03508856e-03 -1.28324962e+00 -9.56698731e-02 1.39964551e-01
6.06091738e-01 8.19586396e-01 4.06961977e-01 9.35968876e-01
8.68722022e-01 -1.18364453e+00 8.64480197e-01 3.57040048e-01
6.47325516e-01 -2.24550843e-01 4.52316880e-01 4.85850781e-01
-9.72571075e-01 -8.08090866e-01 -5.26040614e-01 -4.65806901e-01
1.72894463e-01 -3.70675296e-01 -1.03238916e+00 3.75362307e-01
1.08781779e+00 1.05863917e+00 -5.38003325e-01 1.04112649e+00
-3.96867543e-01 -1.05239652e-01 -6.07392967e-01 -2.51347691e-01
7.16047287e-01 -1.05519548e-01 5.95380843e-01 8.52296710e-01
6.15855932e-01 -3.59053582e-01 2.19481438e-01 5.54219663e-01
2.11049214e-01 -2.65755475e-01 -1.29172504e+00 2.53424823e-01
6.07568741e-01 1.18055689e+00 -2.72616982e-01 9.50538591e-02
-2.88479030e-01 9.85919893e-01 7.67775714e-01 6.07689977e-01
-6.92016184e-01 -3.76487613e-01 9.14763927e-01 -2.00671032e-01
5.53505123e-01 -6.93713069e-01 -1.57663763e-01 -1.11000323e+00
-4.90226559e-02 -8.74949932e-01 2.03016251e-01 -1.33084273e+00
-8.91948521e-01 6.12621367e-01 -1.64286360e-01 -1.29852974e+00
-5.24997592e-01 -9.91839051e-01 -4.23272282e-01 7.38062203e-01
-1.42482269e+00 -1.38696563e+00 -5.50953448e-01 7.17553616e-01
8.04948092e-01 -3.97562921e-01 9.15113628e-01 -2.28205681e-01
-2.97129631e-01 3.25682580e-01 1.81103259e-01 -1.65168911e-01
5.95730782e-01 -1.26142943e+00 9.20435727e-01 6.58107400e-01
-1.23765565e-01 7.66968787e-01 7.29572654e-01 -4.04189795e-01
-1.89094722e+00 -1.11681914e+00 2.45402783e-01 -1.09749043e+00
7.43184507e-01 -3.52714002e-01 -4.42501992e-01 1.40444684e+00
-8.40410292e-02 1.25993742e-02 2.15902507e-01 3.92986864e-01
-4.60216641e-01 1.35306925e-01 -9.74919260e-01 1.13524270e+00
1.80118525e+00 -3.45993280e-01 -6.83844268e-01 4.05641258e-01
9.80315626e-01 -9.83641863e-01 -6.60862684e-01 2.70790607e-01
6.32925093e-01 -7.02455401e-01 1.12794352e+00 -7.76268363e-01
3.91106069e-01 -3.81329387e-01 -4.09478694e-01 -1.71679449e+00
-5.86879313e-01 -6.47811472e-01 -5.23533225e-02 4.17023301e-01
9.54629123e-01 -8.19393575e-01 5.80771446e-01 6.36796176e-01
-9.66959655e-01 -6.60698056e-01 -8.35822284e-01 -8.85690987e-01
7.20791221e-02 -6.57828510e-01 6.16917968e-01 7.99351692e-01
4.30031598e-01 4.47173089e-01 -2.85367161e-01 1.71642780e-01
1.41901687e-01 -1.22581705e-01 1.33680987e+00 -9.90843832e-01
-2.62506902e-01 -3.85351121e-01 -4.97647524e-01 -1.69998372e+00
6.74686804e-02 -1.10379684e+00 5.12746751e-01 -2.07910776e+00
-4.95455235e-01 -9.95646358e-01 5.35346046e-02 1.03282142e+00
3.90934408e-01 -1.51605234e-01 2.01390892e-01 3.81874472e-01
-7.18578219e-01 6.65208399e-01 1.44464362e+00 -2.46961385e-01
-6.17621303e-01 -1.28256261e-01 -1.14455092e+00 7.01259375e-01
9.19030726e-01 -2.35834166e-01 -7.52545357e-01 -1.13263798e+00
5.37893534e-01 -1.65687084e-01 4.00684536e-01 -1.24730432e+00
5.24807155e-01 -2.91916102e-01 1.31707519e-01 -5.10381944e-02
4.34364080e-01 -8.20286512e-01 -7.87597150e-02 3.13511044e-01
-1.17087938e-01 3.32758486e-01 5.20230532e-01 8.70443165e-01
1.84899047e-01 -9.74652991e-02 2.14116916e-01 -4.64402944e-01
-1.55240142e+00 1.94051236e-01 -5.90898871e-01 -1.59397442e-02
9.74559665e-01 -5.30196130e-01 -7.08439529e-01 -4.62795913e-01
-9.12243843e-01 7.73364067e-01 9.58257198e-01 9.32481766e-01
6.18260682e-01 -1.12624025e+00 -1.54413626e-01 1.31042957e-01
5.53835571e-01 5.27407408e-01 2.50373244e-01 7.22236276e-01
-7.53601670e-01 4.80544090e-01 -4.00305003e-01 -8.02487552e-01
-3.87390733e-01 7.18030274e-01 4.61825877e-01 1.12897702e-01
-9.32877004e-01 9.69505310e-01 1.94954008e-01 -1.36246896e+00
-1.95267946e-02 -6.00140870e-01 -3.63129340e-02 -5.86031377e-01
6.94165155e-02 -6.10234179e-02 -1.95431396e-01 -3.03288698e-01
-3.11302513e-01 3.88386965e-01 1.78726867e-01 -6.45681322e-02
1.43379295e+00 -4.15542185e-01 2.27347001e-01 3.58582497e-01
8.68872881e-01 -4.61167812e-01 -1.87218916e+00 -2.18313895e-02
-3.97590511e-02 -1.06763542e-01 -2.13119909e-01 -9.77980554e-01
-6.39885128e-01 6.17062986e-01 2.37265438e-01 7.06732320e-03
6.46663427e-01 1.25996158e-01 6.98363900e-01 1.27885163e+00
9.71333086e-01 -1.10877550e+00 3.05243015e-01 1.13839960e+00
1.11154234e+00 -1.39533639e+00 -5.94338179e-01 1.38640419e-01
-6.56966329e-01 1.10986865e+00 9.42290187e-01 -1.33251086e-01
3.49969834e-01 -5.04474714e-02 1.53016910e-01 -2.25892708e-01
-9.21405196e-01 -2.46388972e-01 2.51795780e-02 1.44268060e+00
-2.05060169e-01 -2.42625028e-01 3.02241921e-01 5.08841753e-01
-5.24731040e-01 -1.00385979e-01 6.84361875e-01 1.19990647e+00
-6.27492011e-01 -7.84875751e-01 -2.25879662e-02 3.43994677e-01
4.67062294e-01 -1.08598918e-01 -3.22544463e-02 1.05330563e+00
-2.85824686e-02 8.95085871e-01 -2.25005805e-01 -6.44520819e-01
4.13154662e-01 -6.55811578e-02 4.54898536e-01 -9.75617290e-01
-2.41573617e-01 -5.37766397e-01 3.24701637e-01 -9.56171691e-01
4.74854931e-02 -6.41240776e-01 -1.32810521e+00 -1.02970518e-01
1.04409412e-01 -2.12282073e-02 9.62746382e-01 8.49101126e-01
7.82114029e-01 4.10254657e-01 5.55117615e-02 -1.54886913e+00
-5.38857639e-01 -8.68765235e-01 -2.63346404e-01 1.07827097e-01
5.05444467e-01 -8.83433282e-01 -3.81317735e-01 -1.17433093e-01]
|
[4.504004955291748, 0.6575742959976196]
|
3969186a-eeb7-4900-8ed9-80b15437c6c7
|
a-simple-and-general-graph-neural-network
|
2009.02562
| null |
https://arxiv.org/abs/2009.02562v2
|
https://arxiv.org/pdf/2009.02562v2.pdf
|
Permutation-equivariant and Proximity-aware Graph Neural Networks with Stochastic Message Passing
|
Graph neural networks (GNNs) are emerging machine learning models on graphs. Permutation-equivariance and proximity-awareness are two important properties highly desirable for GNNs. Both properties are needed to tackle some challenging graph problems, such as finding communities and leaders. In this paper, we first analytically show that the existing GNNs, mostly based on the message-passing mechanism, cannot simultaneously preserve the two properties. Then, we propose Stochastic Message Passing (SMP) model, a general and simple GNN to maintain both proximity-awareness and permutation-equivariance. In order to preserve node proximities, we augment the existing GNNs with stochastic node representations. We theoretically prove that the mechanism can enable GNNs to preserve node proximities, and at the same time, maintain permutation-equivariance with certain parametrization. We report extensive experimental results on ten datasets and demonstrate the effectiveness and efficiency of SMP for various typical graph mining tasks, including graph reconstruction, node classification, and link prediction.
|
['Jian Pei', 'Peng Cui', 'Ziwei Zhang', 'Bo Zhang', 'Wenwu Zhu', 'Chenhao Niu']
|
2020-09-05
| null |
https://openreview.net/forum?id=fhcMwjavKEZ
|
https://openreview.net/pdf?id=fhcMwjavKEZ
| null |
['graph-reconstruction']
|
['graphs']
|
[ 1.02290206e-01 2.02106059e-01 -4.28309470e-01 -1.54364526e-01
-1.12018555e-01 -3.31799954e-01 4.52752173e-01 4.67199773e-01
8.85626599e-02 6.93180025e-01 -1.37449317e-02 -5.33891737e-01
-7.06601501e-01 -1.20891285e+00 -8.86715651e-01 -8.47407460e-01
-8.60792398e-01 4.48959261e-01 2.10274190e-01 -4.99844849e-02
1.13433972e-01 4.66177493e-01 -1.18978572e+00 -5.26297055e-02
9.15069878e-01 3.45937312e-01 1.82485983e-01 7.47959614e-01
-1.34479865e-01 8.25235128e-01 1.26867861e-01 -3.06958646e-01
3.69850755e-01 -1.99685931e-01 -7.09279001e-01 -9.09083709e-02
1.58322975e-01 -3.65798548e-02 -9.30597663e-01 1.30821788e+00
2.25631595e-01 -9.76701304e-02 4.75534737e-01 -1.69819713e+00
-8.09352338e-01 1.31965649e+00 -8.17327321e-01 1.69384986e-01
-8.00168701e-03 -1.86785623e-01 1.39107764e+00 -3.11957419e-01
5.27893007e-01 1.28395927e+00 8.19580019e-01 2.85925627e-01
-1.19076812e+00 -7.39172339e-01 3.73554975e-01 2.29869112e-01
-1.47406650e+00 -7.48159513e-02 1.00719106e+00 -1.15713917e-01
5.85016251e-01 2.76873559e-01 4.47577119e-01 7.85591841e-01
2.23233640e-01 6.23240948e-01 6.31211281e-01 -2.44366333e-01
-2.64540557e-02 -1.87405422e-01 2.30232328e-01 1.06170249e+00
6.26352489e-01 -2.06908658e-01 -4.27755803e-01 -2.49607578e-01
8.73352587e-01 3.45361620e-01 -3.40733707e-01 -7.28001952e-01
-1.18956375e+00 8.91034782e-01 8.73769164e-01 3.30245435e-01
-3.28338474e-01 3.36870551e-01 3.91253680e-01 4.24051553e-01
2.72282869e-01 2.93788370e-02 -1.55669600e-01 3.76010537e-01
-5.16931593e-01 -1.85351804e-01 1.00375032e+00 1.03324580e+00
8.80050838e-01 1.02584166e-02 -6.82726502e-02 5.32171071e-01
4.65652943e-01 4.11993086e-01 2.22610697e-01 -4.75924939e-01
4.42026049e-01 8.42710018e-01 -5.68794489e-01 -1.88224423e+00
-4.72671330e-01 -8.57476473e-01 -1.69826019e+00 -6.03357613e-01
-4.31955084e-02 1.02406882e-01 -5.51181257e-01 2.07910013e+00
4.06639695e-01 6.67232573e-01 8.02505668e-03 4.25245792e-01
9.79809165e-01 7.99835384e-01 -6.10497184e-02 -1.81446731e-01
8.42878461e-01 -7.60699272e-01 -3.78310710e-01 -4.78921831e-03
7.95989454e-01 -2.16670364e-01 7.24530518e-01 -1.58232182e-01
-9.20077026e-01 -2.02933028e-01 -9.83264089e-01 4.86976281e-02
-2.19841097e-02 -2.05847770e-01 1.05080533e+00 5.91821671e-01
-1.28661656e+00 8.70696545e-01 -8.53613436e-01 -4.59563076e-01
3.83483499e-01 4.32682365e-01 -5.84591866e-01 -1.72329083e-01
-1.07899153e+00 5.73712625e-02 3.85683507e-01 1.96429551e-01
-9.69240546e-01 -5.65330982e-01 -9.42688525e-01 4.29257184e-01
3.54533523e-01 -6.26318693e-01 6.49621308e-01 -6.63044155e-01
-9.70254540e-01 5.21502912e-01 -1.11980498e-01 -6.82423532e-01
1.97500527e-01 4.42046374e-01 -2.73611575e-01 1.55037388e-01
2.46653538e-02 3.09131384e-01 2.95029581e-01 -1.17547035e+00
-5.35047174e-01 -3.10580015e-01 1.73531055e-01 2.93103401e-02
-6.96743965e-01 -3.20572346e-01 -4.28424865e-01 -4.97818232e-01
6.13435864e-01 -7.95416415e-01 -4.32062536e-01 -2.21046507e-02
-9.14552271e-01 -3.85892600e-01 6.35641754e-01 -4.97883022e-01
1.22601640e+00 -1.85746968e+00 1.02723457e-01 8.98841739e-01
7.58444965e-01 -2.90347077e-03 -4.26082492e-01 5.73352277e-01
-2.29682773e-02 1.78846925e-01 -1.81223944e-01 -1.86624214e-01
-9.77923498e-02 3.74559134e-01 -1.27617896e-01 7.05758095e-01
3.41362357e-02 9.97120738e-01 -9.24817145e-01 -5.38062155e-01
-9.44396555e-02 2.99781829e-01 -6.38030648e-01 -1.55992746e-01
-1.15663916e-01 6.30251318e-02 -4.74570453e-01 4.51654971e-01
9.10059154e-01 -6.49163008e-01 8.77367198e-01 -8.18536729e-02
3.49577665e-01 2.25155160e-01 -1.36302865e+00 1.15006959e+00
-1.78075597e-01 3.29765767e-01 2.00652599e-01 -1.49107838e+00
9.84596312e-01 -8.67179334e-02 4.87762481e-01 -4.75070000e-01
-5.38904667e-02 8.54316428e-02 1.36486724e-01 -1.45867825e-01
2.57253945e-01 2.91573912e-01 2.38565549e-01 7.06484318e-01
-1.79602072e-01 8.10518861e-01 2.14596942e-01 7.94357657e-01
1.29026306e+00 -5.55625021e-01 1.49973541e-01 -6.14748180e-01
4.61746246e-01 -4.63234276e-01 7.13023841e-01 9.32697475e-01
1.56303793e-01 3.50343674e-01 9.25726354e-01 -4.70482826e-01
-8.85807157e-01 -1.05040860e+00 4.74006474e-01 1.03801394e+00
2.96548545e-01 -5.18566489e-01 -4.60906386e-01 -6.85039937e-01
-2.38135993e-03 2.36959636e-01 -4.94393677e-01 -4.49851871e-01
-5.16122878e-01 -8.90102446e-01 5.37414193e-01 3.57717812e-01
4.62752759e-01 -6.75710380e-01 3.98427606e-01 2.69871652e-01
-2.31773153e-01 -9.71620917e-01 -5.76394022e-01 -9.41231102e-02
-1.06151438e+00 -1.24138486e+00 -3.14568400e-01 -1.29605567e+00
9.69039679e-01 7.50042558e-01 9.38018858e-01 6.98633373e-01
1.92809865e-01 -1.05663305e-02 -2.09110335e-01 1.21751063e-01
-4.83232647e-01 6.16800427e-01 1.61039010e-01 1.87240481e-01
-3.14486101e-02 -1.41814792e+00 -4.25568223e-01 1.83775976e-01
-8.44053388e-01 3.17354918e-01 7.50620246e-01 5.46497464e-01
5.45898616e-01 3.62933666e-01 5.96177399e-01 -1.14229548e+00
6.04033470e-01 -7.27762997e-01 -5.53214848e-01 4.90286469e-01
-6.29071951e-01 2.87178159e-01 7.18642652e-01 -4.05759841e-01
-4.54240471e-01 -4.53545749e-02 -4.73938920e-02 -2.99160570e-01
4.24743831e-01 8.76225650e-01 -3.71577680e-01 -3.00822943e-01
4.17525381e-01 5.76569140e-01 1.75541222e-01 -3.46945465e-01
2.67724603e-01 3.38589221e-01 5.07602394e-01 -5.24415612e-01
1.08105254e+00 7.31794894e-01 4.40651566e-01 -6.81963980e-01
-3.71576399e-01 -2.63448656e-01 -3.67541760e-01 7.08029047e-02
6.25961199e-02 -7.50163913e-01 -1.02958202e+00 3.86188954e-01
-9.59131837e-01 -5.87382391e-02 2.40156293e-01 3.09790075e-01
-9.07751098e-02 7.74554551e-01 -7.54926622e-01 -6.57486320e-01
-6.39508903e-01 -6.89732134e-01 4.19892669e-01 2.24545404e-01
4.22557354e-01 -1.18248773e+00 6.40808716e-02 -1.97666749e-01
3.59996915e-01 2.55020589e-01 1.29037976e+00 -8.64710629e-01
-7.79283822e-01 -9.34490114e-02 -5.80616474e-01 -8.95047858e-02
2.78314859e-01 8.17135200e-02 -3.61424267e-01 -6.20290816e-01
-5.59126556e-01 9.33966264e-02 1.04525411e+00 3.62569958e-01
1.25529695e+00 -8.04748476e-01 -5.73024750e-01 8.13071668e-01
1.32269382e+00 -3.89331847e-01 6.28152966e-01 3.23813781e-02
9.68498170e-01 5.62217295e-01 -1.69502065e-01 3.50983202e-01
8.27970386e-01 5.33301830e-02 7.25627184e-01 1.21428529e-02
4.49586380e-03 -8.84626627e-01 2.25365832e-01 1.36929333e+00
4.76185195e-02 -3.42145950e-01 -7.11709023e-01 7.21885622e-01
-1.91627872e+00 -8.06904674e-01 -3.68782133e-01 2.00151038e+00
5.88197887e-01 7.30148852e-02 4.91291285e-02 1.04035906e-01
1.23221314e+00 3.20501029e-01 -4.50899571e-01 -7.85214677e-02
-3.64890724e-01 -7.89860338e-02 7.34063387e-01 4.31263924e-01
-9.31676507e-01 7.32490182e-01 5.51550817e+00 8.50088477e-01
-7.23921895e-01 -6.25014156e-02 5.47406137e-01 3.75496030e-01
-7.41323292e-01 2.56064296e-01 -5.05213082e-01 3.69466215e-01
7.22510636e-01 -3.95082206e-01 5.51550806e-01 8.62744212e-01
1.38683273e-02 5.38043618e-01 -9.69334960e-01 8.72612417e-01
-1.63416028e-01 -1.68173897e+00 4.82001692e-01 1.49813190e-01
8.45540702e-01 1.32359207e-01 -1.65114000e-01 2.39654467e-01
6.70450389e-01 -1.03435695e+00 1.99341774e-01 3.81786525e-01
5.88677227e-01 -9.95108485e-01 6.58685267e-01 4.49697107e-01
-1.58278656e+00 -6.65465593e-02 -6.38850629e-01 -2.35716230e-03
9.90064070e-02 9.18613732e-01 -6.81030810e-01 1.08001149e+00
4.95436430e-01 8.78823876e-01 -5.12127459e-01 1.16344905e+00
-1.10712387e-01 6.69675350e-01 -5.72672009e-01 -3.12816739e-01
3.21079940e-01 -3.00002366e-01 6.48520052e-01 8.90917122e-01
3.10065806e-01 -7.41001070e-02 4.23288733e-01 7.77731180e-01
-5.45793295e-01 1.35828510e-01 -6.18122816e-01 -1.96896300e-01
8.32824945e-01 1.23036408e+00 -8.59627724e-01 -7.10895797e-03
-1.07541032e-01 6.62753522e-01 6.15218222e-01 2.83017427e-01
-7.00384259e-01 -6.01659596e-01 6.40230954e-01 1.28539726e-01
1.86212942e-01 -4.07734096e-01 7.43337199e-02 -1.05475736e+00
1.22363515e-01 -8.08044136e-01 5.22472441e-01 -2.57910848e-01
-1.37822044e+00 5.37276328e-01 -3.22106034e-01 -8.95695210e-01
3.30158293e-01 -3.42221856e-01 -1.02031267e+00 3.00649792e-01
-1.58888745e+00 -1.42831802e+00 -2.36559525e-01 7.33102143e-01
-2.03785729e-02 -1.68983005e-02 4.73728836e-01 2.96745867e-01
-6.77125812e-01 7.27159142e-01 3.34231734e-01 3.43762487e-01
2.16582239e-01 -1.00628066e+00 5.58079541e-01 8.93450201e-01
4.24118936e-01 8.47525895e-01 5.07632315e-01 -8.89700472e-01
-1.72683585e+00 -1.42428625e+00 8.51833761e-01 2.67938465e-01
6.64881349e-01 -4.45371807e-01 -1.03769195e+00 8.21082413e-01
-2.59173125e-01 6.91939294e-02 1.67868987e-01 3.08115929e-01
-4.84254897e-01 -3.93112928e-01 -9.06510651e-01 6.30316019e-01
1.51470387e+00 -4.40258801e-01 -5.56716025e-02 6.55311167e-01
9.97957289e-01 -5.85385859e-02 -6.13874972e-01 5.60958564e-01
4.30937737e-01 -7.75040329e-01 1.00480020e+00 -8.26369405e-01
1.07465886e-01 -2.68341154e-01 -7.66329914e-02 -1.14787424e+00
-5.76830208e-01 -7.82938242e-01 -1.70067027e-01 1.27314341e+00
4.56130832e-01 -9.34831083e-01 1.04772103e+00 -9.67483521e-02
2.66141206e-01 -6.48611963e-01 -8.91172111e-01 -9.00512099e-01
-6.93142414e-02 -1.90279886e-01 8.48887861e-01 1.07701063e+00
-1.98319480e-01 3.50553781e-01 -6.38854444e-01 5.99424422e-01
8.14761221e-01 2.93639839e-01 7.61830091e-01 -1.57759130e+00
-1.29983872e-01 -4.97856468e-01 -4.70943063e-01 -1.13751519e+00
2.96987444e-01 -1.37816155e+00 -1.88512489e-01 -1.79208589e+00
6.34391963e-01 -6.72724545e-01 -4.62637812e-01 3.03908676e-01
5.31523861e-02 -1.85075864e-01 -1.71265513e-01 2.86715508e-01
-8.49315703e-01 6.26528859e-01 9.64930773e-01 -1.72626764e-01
-3.09630603e-01 1.50417924e-01 -9.68307257e-01 5.89956999e-01
1.00026655e+00 -6.03413165e-01 -6.50916159e-01 -5.16412437e-01
4.72029656e-01 -2.73385674e-01 5.12715816e-01 -7.75562167e-01
7.12100923e-01 -1.08241566e-01 -2.15312932e-02 -4.98656482e-01
-2.10979521e-01 -5.06430805e-01 3.61978471e-01 8.82059336e-01
-3.48808914e-01 2.44992360e-01 -3.28916490e-01 1.03233314e+00
1.54647455e-01 1.87569261e-01 6.36128187e-01 1.75035432e-01
-4.35357988e-01 8.42315674e-01 5.18665873e-02 -9.26128253e-02
8.65289569e-01 2.37682872e-02 -4.11686301e-01 -6.92902803e-01
-3.05499285e-01 6.49348378e-01 2.34944895e-01 2.62413800e-01
7.79850543e-01 -1.34281945e+00 -9.91488278e-01 3.12298834e-01
-1.26339570e-02 -3.94682847e-02 3.44440550e-01 9.98206556e-01
-6.42193735e-01 1.27200603e-01 -3.55956107e-02 -5.14584720e-01
-1.33657777e+00 5.71461856e-01 1.70026764e-01 -5.37027717e-01
-7.83877611e-01 7.47370899e-01 1.38725847e-01 -8.96491826e-01
3.00744742e-01 -9.72248167e-02 -1.33102298e-01 -4.92015511e-01
3.14835906e-01 3.27700108e-01 -1.25313044e-01 -3.40855300e-01
-3.61069053e-01 3.31494600e-01 -1.93476528e-01 4.06939268e-01
1.47744679e+00 -2.38665983e-01 -6.84541523e-01 -6.06670752e-02
1.14873421e+00 -1.94220636e-02 -7.65587866e-01 -7.06281662e-01
1.68194637e-01 -3.12624723e-01 -2.15720966e-01 8.16760883e-02
-1.36562121e+00 6.75633192e-01 6.34267107e-02 4.63354528e-01
9.08629894e-01 1.71423703e-01 8.13103616e-01 7.82547534e-01
4.20345068e-01 -4.98028904e-01 -2.01048300e-01 3.80798876e-01
4.89096433e-01 -7.60957181e-01 -6.77409917e-02 -7.53387511e-01
-1.30152494e-01 9.57664907e-01 6.13049686e-01 -2.66641766e-01
9.24834430e-01 -1.31638557e-01 -7.64831603e-01 -1.22522399e-01
-8.80487442e-01 -4.08829376e-02 -8.60783607e-02 7.09551275e-01
-9.22987610e-02 1.72244206e-01 -1.90923557e-01 6.92281008e-01
-6.47714511e-02 -4.55854297e-01 6.12595141e-01 5.50582230e-01
-6.22610986e-01 -9.17721391e-01 1.31552398e-01 6.70986235e-01
-1.69637665e-01 -3.07574093e-01 -3.66613686e-01 6.50166929e-01
-3.60391557e-01 8.16090643e-01 5.54623839e-04 -6.56467140e-01
-1.47587448e-01 -5.84618807e-01 1.01819493e-01 -4.37067270e-01
-9.23485234e-02 -3.66067141e-01 5.01936227e-02 -2.93503433e-01
-2.47201368e-01 -2.23858580e-01 -1.09055352e+00 -1.05668390e+00
-4.69650000e-01 3.04620951e-01 3.83523673e-01 6.73672140e-01
7.02265561e-01 3.93797338e-01 9.53589201e-01 -1.68652162e-01
-5.70988178e-01 -8.22709799e-01 -8.76605093e-01 7.44036138e-02
2.42119402e-01 -3.79899949e-01 -5.59386611e-01 -5.12533426e-01]
|
[7.064744472503662, 6.124591827392578]
|
dc42423c-e2f5-4f6c-9317-6ee70b85be22
|
bias-reducing-multitask-learning-on-mental
|
2208.03621
| null |
https://arxiv.org/abs/2208.03621v1
|
https://arxiv.org/pdf/2208.03621v1.pdf
|
Bias Reducing Multitask Learning on Mental Health Prediction
|
There has been an increase in research in developing machine learning models for mental health detection or prediction in recent years due to increased mental health issues in society. Effective use of mental health prediction or detection models can help mental health practitioners re-define mental illnesses more objectively than currently done, and identify illnesses at an earlier stage when interventions may be more effective. However, there is still a lack of standard in evaluating bias in such machine learning models in the field, which leads to challenges in providing reliable predictions and in addressing disparities. This lack of standards persists due to factors such as technical difficulties, complexities of high dimensional clinical health data, etc., which are especially true for physiological signals. This along with prior evidence of relations between some physiological signals with certain demographic identities restates the importance of exploring bias in mental health prediction models that utilize physiological signals. In this work, we aim to perform a fairness analysis and implement a multi-task learning based bias mitigation method on anxiety prediction models using ECG data. Our method is based on the idea of epistemic uncertainty and its relationship with model weights and feature space representation. Our analysis showed that our anxiety prediction base model introduced some bias with regards to age, income, ethnicity, and whether a participant is born in the U.S. or not, and our bias mitigation method performed better at reducing the bias in the model, when compared to the reweighting mitigation technique. Our analysis on feature importance also helped identify relationships between heart rate variability and multiple demographic groupings.
|
['Akane Sano', 'Han Yu', 'Kusha Sridhar', 'Khadija Zanna']
|
2022-08-07
| null | null | null | null |
['heart-rate-variability']
|
['medical']
|
[ 3.23814243e-01 2.12495834e-01 -4.36354935e-01 -5.14070690e-01
-6.02211475e-01 6.47544712e-02 1.22688547e-01 6.33823335e-01
-4.94118035e-01 7.55290449e-01 6.52926922e-01 -2.39606321e-01
-5.99075794e-01 -6.66240513e-01 -2.22658813e-01 -4.52737629e-01
-2.15285689e-01 1.83959708e-01 -4.27111208e-01 -1.39998659e-01
3.62371206e-01 1.53636694e-01 -1.09092915e+00 2.68336028e-01
1.03840029e+00 8.53866577e-01 -3.36501151e-01 3.26817662e-01
1.66235790e-01 5.40432036e-01 -4.12380606e-01 -5.32686651e-01
-2.05863237e-01 -5.55660307e-01 -6.09455884e-01 -3.86360198e-01
1.59340929e-02 -1.53638870e-01 1.82400391e-01 8.29285681e-01
1.07259548e+00 -7.77477100e-02 7.31662214e-01 -1.28075171e+00
-5.77665806e-01 7.64300108e-01 -4.52460110e-01 3.19996983e-01
3.28588098e-01 -8.97473190e-03 6.05902493e-01 -6.11682415e-01
2.87374586e-01 1.21980572e+00 1.17056513e+00 7.00055897e-01
-1.44436467e+00 -1.00683594e+00 1.45709172e-01 2.31446430e-01
-1.25880253e+00 -6.01647973e-01 7.37164438e-01 -7.14339972e-01
6.68559551e-01 2.35211179e-01 1.00525939e+00 1.08883357e+00
5.55225551e-01 -7.84736797e-02 1.22346449e+00 -3.31530511e-01
1.99399114e-01 3.94005090e-01 1.86120853e-01 3.74513775e-01
4.78989661e-01 -2.86997072e-02 -4.60465550e-01 -6.34500980e-01
4.93594974e-01 1.35175675e-01 -8.95993337e-02 -2.31123611e-01
-8.93150449e-01 9.79856968e-01 1.23645313e-01 5.05906284e-01
-6.28473878e-01 3.63346972e-02 5.48841178e-01 5.75893223e-02
8.50767493e-01 5.24476945e-01 -5.09908617e-01 -2.41720334e-01
-1.10892737e+00 1.83791280e-01 5.23814499e-01 -3.00150275e-01
3.06871206e-01 4.03571688e-02 -1.24309182e-01 8.96695733e-01
4.84849155e-01 5.13753772e-01 7.16225803e-01 -9.82226551e-01
2.38305181e-01 5.98552525e-01 -1.41926348e-01 -1.42295372e+00
-9.37275827e-01 -4.76579189e-01 -9.70156848e-01 6.95114136e-02
4.46039528e-01 -4.74114060e-01 -4.88823950e-01 2.04613972e+00
7.22561032e-02 9.66914594e-02 8.09557661e-02 7.48506784e-01
4.93568927e-01 1.06608775e-02 4.29414779e-01 -5.22359312e-01
1.50293922e+00 9.16845575e-02 -8.11178505e-01 -4.26561713e-01
6.98915243e-01 -3.27157378e-01 7.54483104e-01 3.81507427e-01
-1.00358057e+00 -3.53447884e-01 -6.00529969e-01 2.88953632e-01
-2.41153717e-01 -9.58694443e-02 8.34994733e-01 1.45989704e+00
-8.03468406e-01 6.62016511e-01 -8.22459042e-01 -6.37153208e-01
7.28675604e-01 4.44388121e-01 -8.14257637e-02 8.21499005e-02
-1.50879085e+00 1.22246885e+00 3.46428305e-02 4.01075222e-02
-3.04876417e-01 -9.83394861e-01 -8.31331313e-01 -1.67064779e-02
9.73038450e-02 -7.26762295e-01 5.99311531e-01 -1.22912741e+00
-1.03172815e+00 7.45466650e-01 -2.19528735e-01 -5.02345204e-01
3.94678295e-01 -2.90667918e-02 -8.15126002e-01 8.77445564e-03
4.65888418e-02 5.65523148e-01 8.01570773e-01 -7.11662292e-01
-2.33249813e-01 -8.72262597e-01 -3.11627954e-01 1.28583536e-01
-5.07804632e-01 1.50303468e-01 5.33816397e-01 -2.95471162e-01
2.64224857e-01 -8.04035008e-01 -3.93151790e-01 -3.38249989e-02
-1.29798591e-01 5.51524982e-02 2.88134009e-01 -7.80706465e-01
1.31754947e+00 -2.04616785e+00 -1.08455583e-01 4.75535572e-01
4.51506555e-01 1.17332608e-01 2.60423034e-01 1.32561028e-01
-3.06464970e-01 5.40533304e-01 4.01801802e-02 9.03643742e-02
-2.85326898e-01 -8.39409456e-02 7.50632305e-03 5.22712886e-01
4.66426581e-01 6.37168765e-01 -8.29204798e-01 -4.81449187e-01
1.99205875e-01 9.27853286e-01 -8.38793218e-01 -3.63763839e-01
5.34615815e-01 5.19270658e-01 -2.46329650e-01 5.55626392e-01
4.98398840e-01 -6.44285455e-02 2.37511143e-01 -1.16303295e-01
-2.89953742e-02 4.07217622e-01 -1.18456388e+00 1.32543325e+00
-1.50203243e-01 3.50089431e-01 -1.61821753e-01 -1.19251299e+00
9.44672227e-01 5.23891091e-01 7.99566329e-01 -5.06643653e-01
1.43074945e-01 2.62918085e-01 6.51086271e-01 -6.01895273e-01
1.82834659e-02 -4.75341737e-01 1.43187150e-01 2.25474343e-01
-3.39562416e-01 -5.00568375e-02 -2.89083153e-01 -5.18916249e-02
6.95650101e-01 -1.92043006e-01 4.40397859e-01 -3.57783228e-01
2.50066370e-01 -3.62750739e-01 8.33875239e-01 5.90754569e-01
-6.40193701e-01 4.05618429e-01 6.78288043e-01 -3.33441854e-01
-4.23845619e-01 -6.76361680e-01 -7.58627057e-01 7.41262078e-01
-4.88676667e-01 -2.92514354e-01 -3.68060201e-01 -1.92396820e-01
1.83423772e-01 7.22829580e-01 -8.92751157e-01 -8.32060218e-01
4.93219607e-02 -1.43695402e+00 6.58761501e-01 4.13590938e-01
3.40124577e-01 -6.55854106e-01 -1.01569188e+00 3.11517894e-01
-2.04745561e-01 -4.34324861e-01 1.28283218e-01 1.67837650e-01
-1.18816912e+00 -9.69342947e-01 -7.74731159e-01 -1.43932298e-01
3.09251666e-01 -3.04211050e-01 8.98903430e-01 -8.31921324e-02
-2.97610641e-01 4.56802309e-01 -1.39805838e-03 -1.01769030e+00
-2.62202919e-01 -2.25784648e-02 1.82232991e-01 9.24025662e-03
6.61813140e-01 -5.78780293e-01 -7.86065996e-01 -1.41631797e-01
-6.10367477e-01 -2.46968225e-01 3.62351656e-01 6.65711880e-01
6.58088699e-02 -2.14891564e-02 1.13093972e+00 -9.32984591e-01
8.77161682e-01 -9.11110640e-01 1.61044747e-01 -5.18620089e-02
-1.31111002e+00 -2.28763789e-01 -2.06810504e-01 -4.99346226e-01
-9.00507390e-01 -2.38761798e-01 -9.06466320e-02 2.47391224e-01
-1.58240452e-01 8.02683234e-01 1.44068658e-01 1.18364699e-01
9.23420429e-01 -3.66932511e-01 3.95918965e-01 -1.14443921e-01
-2.20124304e-01 6.88181818e-01 -1.77060619e-01 -3.30075473e-01
9.68271419e-02 4.03665215e-01 1.07045874e-01 -7.99742162e-01
-6.18172705e-01 -3.52066271e-02 -4.50167030e-01 -2.65730858e-01
8.30873191e-01 -9.02340412e-01 -7.95331061e-01 3.32767695e-01
-6.25135660e-01 -9.30477306e-02 5.80254570e-02 9.44651186e-01
-1.89858377e-01 7.22470433e-02 -3.16183567e-01 -1.24045944e+00
-4.70612705e-01 -1.01161921e+00 5.19185543e-01 6.83290362e-02
-8.18928182e-01 -1.16626978e+00 2.24559501e-01 4.65466201e-01
8.83950472e-01 3.69417369e-01 1.14693522e+00 -6.18774235e-01
2.60330796e-01 -1.13936193e-01 -1.25503927e-01 2.45972723e-01
3.05142820e-01 -5.06384932e-02 -1.02678776e+00 -3.40928659e-02
1.34198248e-01 -3.80512327e-01 7.38112569e-01 9.94031787e-01
9.97865796e-01 -3.84016591e-03 -3.38801980e-01 1.96051508e-01
1.13322163e+00 3.04276288e-01 6.39003277e-01 2.31829174e-02
4.71969694e-01 9.47756231e-01 2.49993443e-01 6.36665046e-01
6.01200759e-01 4.41292942e-01 2.73465574e-01 -2.51055956e-01
3.49402070e-01 1.75676793e-01 1.45904317e-01 2.13406578e-01
-4.01075333e-01 3.65357280e-01 -1.16447437e+00 4.02316540e-01
-1.57981026e+00 -9.92515564e-01 -1.51674926e-01 2.38011026e+00
8.00456762e-01 1.93951219e-01 4.66286421e-01 4.12371755e-01
5.85662425e-01 -1.86280027e-01 -6.93996906e-01 -6.32949114e-01
4.08416912e-02 1.01293296e-01 2.26056769e-01 2.82490402e-01
-8.57733667e-01 2.17455775e-01 6.70997334e+00 2.01990129e-03
-1.37593448e+00 1.05117418e-01 1.15857387e+00 -2.70292789e-01
-1.80329666e-01 -1.65631965e-01 -4.19735134e-01 4.93275195e-01
1.18939471e+00 -1.82612255e-01 1.86268374e-01 5.82639992e-01
6.27594531e-01 -2.09686249e-01 -8.42696309e-01 9.38097119e-01
2.45358020e-01 -8.02886426e-01 -4.61553186e-01 8.94729644e-02
3.12677115e-01 -2.33402967e-01 2.76293635e-01 3.14506471e-01
-2.85495639e-01 -1.06673932e+00 2.84959733e-01 7.26295412e-01
5.81188858e-01 -6.98629797e-01 8.28725100e-01 -6.40876442e-02
-5.00789821e-01 -2.38611698e-01 -1.78685233e-01 -5.71354389e-01
-1.78680390e-01 1.05252051e+00 -9.08426702e-01 4.55838107e-02
6.75306916e-01 5.51931202e-01 -3.76264930e-01 8.94488752e-01
3.05660278e-01 8.30468833e-01 -5.16199321e-02 6.83433935e-02
-2.60908544e-01 -6.57408312e-02 1.13390259e-01 9.32596147e-01
5.07188797e-01 2.49392278e-02 -1.82935491e-01 8.62159193e-01
3.46902311e-01 5.14208853e-01 -6.60258710e-01 -1.82690918e-01
2.51542151e-01 1.15485656e+00 -7.01351166e-01 -2.37877801e-01
-6.20989799e-01 4.84445602e-01 -1.06456846e-01 1.33275852e-01
-6.77035928e-01 -1.18371584e-01 8.01110148e-01 3.40012550e-01
-4.85610396e-01 3.05613697e-01 -7.52272367e-01 -9.85604644e-01
-4.17955995e-01 -1.03933060e+00 5.10708392e-01 -4.40199226e-01
-1.08170009e+00 1.21392772e-01 1.01223756e-02 -6.68415606e-01
-2.63108522e-01 -1.35493219e-01 -3.74637991e-01 1.00604486e+00
-1.26595616e+00 -8.03040862e-01 -6.28304780e-02 3.94783467e-01
-2.95630042e-02 6.20982237e-02 1.26545751e+00 4.18762088e-01
-7.46362448e-01 5.08858204e-01 -2.18741640e-01 -1.12944439e-01
1.13225436e+00 -9.91244793e-01 -3.78456324e-01 3.97877336e-01
-3.51551205e-01 6.57748342e-01 6.32348061e-01 -7.81660616e-01
-7.88237393e-01 -7.56631613e-01 1.08803153e+00 -5.11857629e-01
4.35490876e-01 -1.11536615e-01 -9.85156953e-01 3.91514242e-01
-1.81420058e-01 -3.11596721e-01 1.30670965e+00 7.82655001e-01
-1.38760954e-01 -2.71567523e-01 -1.51984549e+00 5.56997180e-01
6.31151080e-01 -4.38540131e-01 -5.79139531e-01 1.62667856e-01
1.74744636e-01 -1.65098812e-02 -1.20680296e+00 4.68258709e-01
8.96133304e-01 -1.04485083e+00 1.08083045e+00 -7.21476555e-01
3.78947616e-01 4.63950753e-01 1.25020325e-01 -1.49781573e+00
-4.56453949e-01 -2.08836034e-01 2.87157148e-01 1.24632204e+00
6.73478961e-01 -1.01404452e+00 6.61427915e-01 1.43684793e+00
2.45691150e-01 -7.00771987e-01 -8.30881238e-01 -1.69810832e-01
2.16431573e-01 -6.83294296e-01 4.39903677e-01 1.28172183e+00
4.53731596e-01 2.00830162e-01 -4.57118422e-01 1.69383548e-02
3.31179440e-01 -2.09922925e-01 2.06140906e-01 -1.75128639e+00
-1.60700724e-01 -3.92453879e-01 -5.77934086e-01 3.23093832e-01
-2.47067902e-02 -6.62300825e-01 -5.76684535e-01 -1.55467272e+00
3.95972341e-01 -4.88264859e-01 -7.23302364e-01 6.76471710e-01
-4.06777471e-01 2.39433512e-01 -6.93857297e-02 -8.28088000e-02
4.98879664e-02 1.96648091e-01 8.47015917e-01 3.90526168e-02
-6.84325576e-01 2.00070679e-01 -1.31705117e+00 8.21135461e-01
1.15386140e+00 -6.30425155e-01 -6.00059688e-01 -9.54257417e-03
5.65202415e-01 2.18852967e-01 3.47084314e-01 -1.11243129e+00
-2.32030898e-01 -7.05617890e-02 7.84597516e-01 2.81956773e-02
5.61741233e-01 -7.25580573e-01 1.30083591e-01 9.91074026e-01
-5.04088819e-01 -6.25535473e-02 2.82639682e-01 1.49458274e-01
2.31513590e-01 -6.19981773e-02 5.50824404e-01 6.98150555e-03
-1.70752659e-01 -2.07116753e-01 -5.97095490e-01 1.65156230e-01
7.06969678e-01 -1.19104534e-01 -7.95382112e-02 -5.98809183e-01
-1.13911760e+00 -9.80968680e-03 2.51917094e-02 3.15798789e-01
4.36382025e-01 -1.07439947e+00 -7.57313013e-01 2.79235482e-01
8.02754685e-02 -7.71011353e-01 4.10774320e-01 1.51155698e+00
9.90791097e-02 3.17251354e-01 -4.51859981e-01 -3.11916918e-01
-1.32014155e+00 3.41071129e-01 4.81766880e-01 -3.62401456e-03
-2.13648736e-01 5.65693855e-01 -2.92442054e-01 -1.27479434e-01
2.10460976e-01 -5.55229783e-01 -4.10661370e-01 5.23504019e-01
6.05292439e-01 7.58079112e-01 3.00707873e-02 -4.30531323e-01
-6.49351180e-01 3.43154460e-01 9.62022394e-02 -1.35664716e-01
1.41098869e+00 -1.67477980e-01 -1.82884350e-01 7.59490430e-01
6.58923924e-01 2.30941959e-02 -6.33389175e-01 1.98592484e-01
1.58888754e-02 -2.81640738e-01 3.43823820e-01 -9.89994407e-01
-9.05664682e-01 9.90501404e-01 1.25741124e+00 1.94132954e-01
1.12266767e+00 -2.84765899e-01 1.68546751e-01 1.13654785e-01
2.00387925e-01 -1.14079726e+00 -6.26796931e-02 -3.75948474e-02
6.19392216e-01 -1.45643890e+00 1.63049418e-02 -1.18582711e-01
-8.03965330e-01 8.29345882e-01 3.28301966e-01 9.30532068e-02
9.02253330e-01 -1.23559974e-01 2.87874252e-01 -2.50384718e-01
-4.73181248e-01 -4.64840569e-02 9.99548584e-02 8.42977941e-01
9.01500642e-01 2.26219997e-01 -8.21304619e-01 8.41126680e-01
2.65487470e-03 2.77718604e-01 3.75054359e-01 5.98154604e-01
-3.33704263e-01 -1.15089905e+00 -6.23383462e-01 1.08616400e+00
-9.19955432e-01 -1.72830999e-01 -3.01723868e-01 5.19540071e-01
4.37079787e-01 1.12016714e+00 1.28086567e-01 -2.20917538e-01
1.25257760e-01 7.74217308e-01 2.19136968e-01 -7.24934697e-01
-6.10932410e-01 9.75784287e-02 2.82826215e-01 -4.59123641e-01
-6.11111462e-01 -1.02311695e+00 -1.08511961e+00 1.08176759e-02
-2.75120199e-01 9.08071622e-02 8.82954121e-01 9.10019636e-01
5.91331899e-01 7.12201416e-01 2.50215799e-01 -4.67095137e-01
-3.96652132e-01 -1.06017923e+00 -6.15380228e-01 2.96028584e-01
1.48199469e-01 -6.69291556e-01 -2.76926994e-01 -3.40157777e-01]
|
[13.556145668029785, 3.3938655853271484]
|
60dc86e9-92c4-4f30-98ca-d502c19fccd1
|
an-algorithm-with-optimal-dimension
|
2307.04504
| null |
https://arxiv.org/abs/2307.04504v1
|
https://arxiv.org/pdf/2307.04504v1.pdf
|
An Algorithm with Optimal Dimension-Dependence for Zero-Order Nonsmooth Nonconvex Stochastic Optimization
|
We study the complexity of producing $(\delta,\epsilon)$-stationary points of Lipschitz objectives which are possibly neither smooth nor convex, using only noisy function evaluations. Recent works proposed several stochastic zero-order algorithms that solve this task, all of which suffer from a dimension-dependence of $\Omega(d^{3/2})$ where $d$ is the dimension of the problem, which was conjectured to be optimal. We refute this conjecture by providing a faster algorithm that has complexity $O(d\delta^{-1}\epsilon^{-3})$, which is optimal (up to numerical constants) with respect to $d$ and also optimal with respect to the accuracy parameters $\delta,\epsilon$, thus solving an open question due to Lin et al. (NeurIPS'22). Moreover, the convergence rate achieved by our algorithm is also optimal for smooth objectives, proving that in the nonconvex stochastic zero-order setting, nonsmooth optimization is as easy as smooth optimization. We provide algorithms that achieve the aforementioned convergence rate in expectation as well as with high probability. Our analysis is based on a simple yet powerful geometric lemma regarding the Goldstein-subdifferential set, which allows utilizing recent advancements in first-order nonsmooth nonconvex optimization.
|
['Ohad Shamir', 'Guy Kornowski']
|
2023-07-10
| null | null | null | null |
['stochastic-optimization']
|
['methodology']
|
[-1.68956071e-01 3.65072370e-01 1.44058257e-01 8.41630176e-02
-1.21593356e+00 -5.35494685e-01 -3.13883483e-01 7.18452707e-02
-5.86638570e-01 1.15353596e+00 -1.36537999e-01 -2.06127942e-01
-6.30044401e-01 -6.01042092e-01 -1.06246483e+00 -1.17545605e+00
-5.32986760e-01 4.46529120e-01 3.25897895e-02 -4.95618373e-01
3.69090855e-01 2.58768767e-01 -1.22426200e+00 -5.59011638e-01
1.20868361e+00 1.31036556e+00 -4.60297279e-02 9.33852494e-01
2.29328230e-01 2.12240815e-01 -3.92623395e-01 -4.32059795e-01
4.94104385e-01 -3.99276257e-01 -7.82674551e-01 1.29383057e-01
2.81293035e-01 4.55336869e-02 1.86979130e-01 1.72770262e+00
5.60269237e-01 4.88663316e-01 5.96324682e-01 -9.75199759e-01
-5.75733662e-01 3.47653329e-01 -6.67111099e-01 1.22551210e-01
1.36766850e-03 -5.86835667e-02 9.02324975e-01 -8.21124971e-01
2.10611060e-01 9.19463754e-01 9.01113272e-01 4.78930742e-01
-9.70248520e-01 -4.41149831e-01 -1.93914175e-02 -3.12373936e-01
-1.58358979e+00 -2.53197432e-01 6.00906610e-01 -4.81246501e-01
3.49882454e-01 4.69788700e-01 3.77807587e-01 2.65012175e-01
2.70260364e-01 4.19550389e-01 1.11418080e+00 -4.38507020e-01
3.24081719e-01 1.46598652e-01 2.38963496e-02 1.14616001e+00
5.43872297e-01 -9.43113342e-02 -7.78042944e-03 -1.47746310e-01
8.13440979e-01 -4.54535574e-01 -5.82757831e-01 8.48065037e-03
-8.78104448e-01 1.06750154e+00 1.15636192e-01 5.09950399e-01
-1.59371004e-01 2.88060397e-01 -6.09866418e-02 3.26135218e-01
8.15179646e-01 5.69125831e-01 -5.14936388e-01 -2.78865576e-01
-6.50844812e-01 2.70889401e-01 9.85701323e-01 1.18328166e+00
5.05020380e-01 1.22292951e-01 8.27637240e-02 5.37501276e-01
2.43551269e-01 9.04804051e-01 -6.72050789e-02 -1.27730858e+00
6.37707055e-01 1.12402350e-01 5.27734637e-01 -1.17817330e+00
-5.28906167e-01 -5.93536258e-01 -9.08928871e-01 2.96255469e-01
9.10183489e-01 -4.22675818e-01 -2.51955181e-01 1.92422140e+00
2.67382711e-01 -3.01161353e-02 -1.42812639e-01 9.64720607e-01
-8.08189288e-02 5.05873322e-01 -5.03032982e-01 -6.92939460e-01
1.12803388e+00 -5.00135839e-01 -6.72118902e-01 4.05766726e-01
6.98412478e-01 -7.79492795e-01 1.34771574e+00 5.91692805e-01
-1.59247172e+00 -1.72342174e-02 -9.68412459e-01 1.27725571e-01
6.45480528e-02 1.32572707e-02 3.57723802e-01 8.10523510e-01
-1.15811372e+00 6.97726429e-01 -6.45090461e-01 2.22723365e-01
3.83327574e-01 5.11929274e-01 -1.73170418e-01 1.99514702e-01
-8.40412676e-01 5.75348556e-01 -1.50647670e-01 2.37460151e-01
-4.21664596e-01 -8.73983085e-01 -6.97576463e-01 -6.83709756e-02
5.11226654e-01 -3.93047690e-01 7.51088202e-01 -5.36735058e-01
-1.60782063e+00 5.66775143e-01 -4.75540936e-01 -3.71357769e-01
8.64389122e-01 -8.49485546e-02 2.52272766e-02 3.31083238e-01
1.95561394e-01 -2.48891577e-01 6.66184962e-01 -1.14298391e+00
-5.98759949e-01 -8.34172666e-01 1.88167796e-01 1.25208959e-01
-2.03773081e-01 -1.32012159e-01 -2.76228525e-02 -5.37911475e-01
1.69993058e-01 -9.88721907e-01 -6.08336091e-01 4.96672280e-02
-2.79931873e-01 -1.22688405e-01 2.39439413e-01 -6.14980817e-01
1.11492848e+00 -2.13286829e+00 6.35929853e-02 3.94364893e-01
2.57545799e-01 -7.90182594e-03 2.63273239e-01 -5.02820797e-02
2.15747237e-01 5.91823995e-01 -6.91577196e-01 -4.41536695e-01
4.74689007e-02 -5.50508173e-03 1.01590157e-02 1.20275927e+00
-1.16748244e-01 5.16409993e-01 -9.08741951e-01 -1.84495315e-01
-8.68736804e-02 3.82664919e-01 -7.16650248e-01 -4.34677184e-01
5.14114648e-02 4.34762925e-01 -5.73835075e-01 4.19081658e-01
9.70382929e-01 -3.13813359e-01 -2.19255269e-01 2.00749576e-01
-2.72007942e-01 -3.20272535e-01 -1.70421267e+00 1.29654765e+00
-6.67228103e-01 5.01409650e-01 8.01788449e-01 -1.40309155e+00
6.00606203e-01 2.10124776e-01 7.71589756e-01 -1.82261601e-01
3.59756619e-01 6.83186710e-01 -5.01140475e-01 -5.95721483e-01
2.53325641e-01 -9.02492523e-01 3.72886844e-02 4.36240695e-02
-3.21892381e-01 -1.33170575e-01 2.01473713e-01 -2.33688086e-01
1.04605961e+00 -5.74278772e-01 -5.48266470e-02 -1.06455326e+00
5.69101512e-01 -9.07884911e-02 5.34886420e-01 7.04190910e-01
-2.03308970e-01 6.51128113e-01 8.90128076e-01 4.52381596e-02
-8.00845206e-01 -8.64218056e-01 -6.07536495e-01 8.01821947e-01
5.25507212e-01 4.79983613e-02 -8.34528327e-01 -5.20969510e-01
-4.19297889e-02 6.17355883e-01 -9.30623889e-01 1.35305732e-01
-6.36034727e-01 -9.84667242e-01 2.92479247e-01 4.11373287e-01
4.37524110e-01 -3.37963283e-01 -2.16047749e-01 2.11596832e-01
-1.36500141e-02 -9.31061447e-01 -8.95115614e-01 2.51552224e-01
-8.67149770e-01 -1.14094317e+00 -1.13859642e+00 -7.85333335e-01
1.08415210e+00 1.01068385e-01 8.04589331e-01 -6.47338340e-03
1.23828419e-01 3.72531593e-01 -2.48429235e-02 -5.06225944e-01
-3.00270393e-02 -1.75646082e-01 1.95219740e-01 1.42543972e-01
-3.50731730e-01 -3.85642290e-01 -5.81001222e-01 4.80353773e-01
-8.51584196e-01 -6.85741425e-01 1.49902865e-01 9.05043364e-01
9.78725493e-01 5.49938202e-01 4.73964334e-01 -7.34942138e-01
6.83986187e-01 -4.81745213e-01 -1.28178656e+00 -2.78346706e-03
-5.98360896e-01 2.95544893e-01 1.09678042e+00 -3.26602370e-01
-5.01701832e-01 -8.77595395e-02 -2.37585992e-01 -3.63676220e-01
4.17623937e-01 4.10769194e-01 9.34438258e-02 -5.54857552e-01
6.09444678e-01 1.43379435e-01 8.30791369e-02 -4.17187542e-01
1.27310470e-01 2.52018839e-01 4.53360260e-01 -7.77462065e-01
8.08510184e-01 9.35242772e-01 5.77837288e-01 -1.11928463e+00
-1.04105127e+00 -4.69037563e-01 -1.48196429e-01 -3.00244195e-03
7.33416438e-01 -3.20229411e-01 -1.39874196e+00 -3.53476927e-02
-9.30781305e-01 -3.50964218e-01 -7.36786366e-01 6.85099363e-01
-8.26596618e-01 4.38085079e-01 -3.45043361e-01 -1.28950965e+00
2.06385050e-02 -1.14623821e+00 8.49595726e-01 5.67724258e-02
2.75405854e-01 -1.39750457e+00 -2.04343587e-01 2.54614830e-01
5.78913093e-01 4.21412557e-01 6.91053450e-01 -8.97408873e-02
-2.57177472e-01 -2.51101077e-01 -9.32712927e-02 6.26805067e-01
-1.62932649e-01 -2.27039695e-01 -1.72714010e-01 -2.84417421e-01
7.53261864e-01 9.69462022e-02 4.87175524e-01 9.69099283e-01
1.19433701e+00 -9.99745965e-01 1.48236533e-04 7.49920726e-01
1.73011434e+00 -1.28418431e-01 4.09597933e-01 -2.27802945e-03
3.36412996e-01 4.16396797e-01 4.57266331e-01 6.47347331e-01
2.82017171e-01 4.69704717e-01 6.48396671e-01 -4.71433513e-02
3.45088959e-01 2.49351338e-01 2.28372589e-01 6.94678009e-01
-3.09760034e-01 -8.05063173e-02 -4.72099483e-01 7.83861220e-01
-1.58498287e+00 -8.04499686e-01 -6.73294425e-01 2.57881689e+00
8.84098172e-01 1.00945905e-01 2.58068562e-01 2.59796202e-01
6.85594618e-01 -2.25657374e-01 -2.46056482e-01 -4.93551850e-01
-4.55335528e-01 6.38581872e-01 1.10925758e+00 1.02981591e+00
-7.99171746e-01 3.82616431e-01 5.70112705e+00 9.84799266e-01
-9.29400861e-01 1.55543953e-01 6.58705533e-01 -1.39102206e-01
-4.45170552e-01 -4.74294089e-02 -6.95656657e-01 8.53018880e-01
6.92674458e-01 -3.71947318e-01 5.51422358e-01 9.69818652e-01
4.70502168e-01 -3.17197025e-01 -4.84765917e-01 8.52496922e-01
3.13674547e-02 -1.29304993e+00 -6.97943151e-01 4.49874580e-01
1.20629394e+00 -3.70836496e-01 3.20096731e-01 -1.72481894e-01
3.11897457e-01 -1.18223059e+00 4.09966260e-01 2.95022488e-01
6.69443548e-01 -9.70833957e-01 8.97403717e-01 4.15440470e-01
-1.28396940e+00 -1.65632188e-01 -3.60744506e-01 4.67128754e-02
3.95027101e-01 1.02652740e+00 -2.84028083e-01 5.87411344e-01
7.21652806e-01 1.75692216e-01 2.02494591e-01 1.02003419e+00
6.02801815e-02 5.02515852e-01 -9.12960827e-01 -2.12995097e-01
3.26933235e-01 -5.36648333e-01 8.45951140e-01 9.25829351e-01
6.68250442e-01 6.17770612e-01 1.17065936e-01 5.72464943e-01
-1.67341799e-01 3.52241904e-01 -2.44055748e-01 4.10571605e-01
7.72434995e-02 7.81178892e-01 -6.17001951e-01 -5.14443219e-02
-1.34131208e-01 6.52400494e-01 -9.84478742e-02 3.78170997e-01
-8.26763749e-01 -6.01012170e-01 6.66858375e-01 4.58200097e-01
4.08972591e-01 -6.16123676e-01 -8.54833901e-01 -1.02162993e+00
6.67979479e-01 -4.20245200e-01 3.41423154e-01 6.59011081e-02
-1.22053266e+00 4.28056508e-01 -9.44277421e-02 -1.00148618e+00
2.42126375e-01 -7.17464507e-01 -2.30110034e-01 8.78605068e-01
-1.37408948e+00 -3.57199520e-01 2.03173116e-01 6.22425139e-01
2.38231659e-01 7.74793178e-02 4.91062313e-01 4.06384975e-01
-3.87913346e-01 6.89836502e-01 5.66578150e-01 -2.20015526e-01
1.24114551e-01 -1.39187551e+00 -5.21915793e-01 8.35981607e-01
-3.95618021e-01 3.66297513e-01 1.00018966e+00 -2.74948776e-01
-1.46067381e+00 -7.91886389e-01 1.04163921e+00 -2.60071725e-01
7.22320557e-01 -1.22610986e-01 -8.23861063e-01 2.29137748e-01
-2.97964364e-01 3.50514948e-01 3.86330366e-01 -2.06253171e-01
4.82102811e-01 -1.59677133e-01 -1.48872983e+00 5.64870715e-01
1.05705845e+00 -3.39659154e-02 -2.23500188e-02 5.14191210e-01
4.89498556e-01 -4.74118471e-01 -1.12308025e+00 4.53138411e-01
2.24757627e-01 -7.93980300e-01 7.62558699e-01 -5.23983300e-01
2.31235966e-01 -2.51122057e-01 -4.79658663e-01 -1.02772021e+00
9.43514630e-02 -1.24489474e+00 -9.22844708e-02 6.42326415e-01
6.37861550e-01 -8.50566924e-01 8.67668211e-01 4.53735679e-01
-2.53578007e-01 -1.32250118e+00 -1.50923645e+00 -1.17733407e+00
5.50974131e-01 -5.51797211e-01 1.19895451e-02 7.62564898e-01
-3.82932350e-02 -1.24273852e-01 -2.77846843e-01 4.06519502e-01
6.45068586e-01 -3.63639176e-01 4.21960860e-01 -1.11780584e+00
-4.77672219e-01 -6.02615714e-01 -2.50421107e-01 -1.39588916e+00
2.38454901e-02 -7.42952287e-01 3.32508326e-01 -1.16837490e+00
-3.32351208e-01 -8.84864211e-01 -1.48359552e-01 2.63536088e-02
-1.50389031e-01 8.15911219e-02 -1.01371668e-01 -2.52069980e-02
-3.55318755e-01 6.20957136e-01 1.53600574e+00 3.76146846e-02
-4.16761041e-01 3.73167753e-01 -8.55168462e-01 9.78915453e-01
5.94497025e-01 -4.46612120e-01 -3.01795363e-01 -5.09338558e-01
3.92670214e-01 4.54161704e-01 1.58733070e-01 -9.45670247e-01
7.47869983e-02 -2.89530694e-01 -2.70416915e-01 -1.22558460e-01
3.17325503e-01 -6.82772934e-01 -3.07223976e-01 3.37601572e-01
-3.50087613e-01 7.97458217e-02 -8.72908533e-02 6.01718307e-01
1.00844063e-01 -5.82533121e-01 1.32484448e+00 -9.15891584e-03
1.78337678e-01 3.27354580e-01 -1.06566414e-01 6.06032908e-01
1.04159689e+00 -5.52480072e-02 -7.29559436e-02 -6.16134286e-01
-8.00410509e-01 9.44220573e-02 2.79879630e-01 -4.22890842e-01
3.65582675e-01 -1.05620325e+00 -8.69365215e-01 -4.19743508e-02
-5.65704107e-01 3.25763643e-01 2.58215219e-01 1.50264168e+00
-6.34731352e-01 1.85686246e-01 5.45186639e-01 -5.81369221e-01
-6.05207145e-01 3.15986991e-01 3.41716021e-01 -3.43239784e-01
-2.99232364e-01 1.26822054e+00 -1.11381218e-01 1.38824031e-01
1.94353446e-01 -5.06058693e-01 3.88843298e-01 -3.07609648e-01
2.50068337e-01 8.89779866e-01 1.21319830e-01 -3.35297018e-01
-3.67108196e-01 8.50457847e-01 4.32725608e-01 -4.77700770e-01
1.18593061e+00 -2.97980636e-01 -1.49220720e-01 1.48527712e-01
1.70792484e+00 5.10188043e-01 -1.40173697e+00 8.50071386e-02
-1.75495684e-01 -3.63642782e-01 -8.10831636e-02 -3.00769776e-01
-1.18892670e+00 6.11461759e-01 3.76180351e-01 5.21323740e-01
1.12605250e+00 2.99886614e-03 8.91106188e-01 1.26364455e-01
4.04681534e-01 -1.37161934e+00 -8.40877835e-03 6.54562294e-01
9.58224177e-01 -1.25651908e+00 -1.36001140e-01 -6.58378720e-01
-3.11923862e-01 1.04021275e+00 8.69362503e-02 -5.06657898e-01
1.08809376e+00 2.45403126e-01 -3.78679842e-01 6.53855726e-02
-1.66142315e-01 -2.32243270e-01 2.93026060e-01 1.84639201e-01
2.84735203e-01 5.83366975e-02 -9.77388561e-01 6.36921823e-01
-3.43257099e-01 -9.72959250e-02 4.27561969e-01 7.57483840e-01
-5.50129950e-01 -8.27378750e-01 -4.41374630e-01 4.33832735e-01
-7.58078396e-01 2.45510545e-02 1.97768748e-01 7.81046033e-01
1.11522980e-01 1.07786727e+00 -3.24349612e-01 1.79414274e-04
2.64850527e-01 -3.90306383e-01 5.04778564e-01 -2.76672363e-01
-8.10925737e-02 9.87343565e-02 -9.15514752e-02 -3.84801418e-01
-1.27137065e-01 -5.82955837e-01 -1.42703247e+00 -4.79494601e-01
-3.80004644e-01 4.91802722e-01 6.64280176e-01 1.11294568e+00
8.46109167e-02 1.92203909e-01 1.04417825e+00 -4.53790784e-01
-9.08451796e-01 -6.35869920e-01 -8.70688438e-01 4.18306291e-02
4.97189850e-01 -5.80972254e-01 -9.62465227e-01 -1.06896207e-01]
|
[6.536753177642822, 4.4715752601623535]
|
2ee0d30e-11e8-4953-a2ad-fe01dc31fd4f
|
robustswap-a-simple-yet-robust-face-swapping
|
2303.15768
| null |
https://arxiv.org/abs/2303.15768v1
|
https://arxiv.org/pdf/2303.15768v1.pdf
|
RobustSwap: A Simple yet Robust Face Swapping Model against Attribute Leakage
|
Face swapping aims at injecting a source image's identity (i.e., facial features) into a target image, while strictly preserving the target's attributes, which are irrelevant to identity. However, we observed that previous approaches still suffer from source attribute leakage, where the source image's attributes interfere with the target image's. In this paper, we analyze the latent space of StyleGAN and find the adequate combination of the latents geared for face swapping task. Based on the findings, we develop a simple yet robust face swapping model, RobustSwap, which is resistant to the potential source attribute leakage. Moreover, we exploit the coordination of 3DMM's implicit and explicit information as a guidance to incorporate the structure of the source image and the precise pose of the target image. Despite our method solely utilizing an image dataset without identity labels for training, our model has the capability to generate high-fidelity and temporally consistent videos. Through extensive qualitative and quantitative evaluations, we demonstrate that our method shows significant improvements compared with the previous face swapping models in synthesizing both images and videos. Project page is available at https://robustswap.github.io/
|
['Jaegul Choo', 'Younggun Lee', 'Sunghyun Park', 'Taewoo Kim', 'Jaeseong Lee']
|
2023-03-28
| null | null | null | null |
['face-swapping']
|
['computer-vision']
|
[ 3.89406025e-01 2.98407584e-01 -2.39362568e-01 -2.66860515e-01
-6.63596809e-01 -9.24324989e-01 6.12690687e-01 -8.09284866e-01
1.06037989e-01 5.96291780e-01 3.79487842e-01 2.87956707e-02
2.36648381e-01 -5.27313232e-01 -8.52879226e-01 -8.32111001e-01
2.56414294e-01 -1.36066020e-01 -3.89354914e-01 -1.19971670e-01
2.89599109e-03 3.73958796e-01 -1.56719625e+00 3.28885168e-01
5.24044216e-01 1.02326322e+00 -1.27498820e-01 2.79834718e-01
2.79272258e-01 7.75981486e-01 -4.79214936e-01 -6.11375570e-01
7.48744249e-01 -6.89042270e-01 -6.49530053e-01 2.60086745e-01
9.47975159e-01 -6.53472841e-01 -5.16672909e-01 1.14354181e+00
3.55278224e-01 -2.44264022e-01 2.24352255e-01 -1.75682878e+00
-9.36199129e-01 1.67854175e-01 -5.20221889e-01 -2.44626611e-01
4.68320578e-01 4.64475572e-01 7.28287876e-01 -9.97735381e-01
9.35461402e-01 1.29271770e+00 5.66957057e-01 9.37672257e-01
-1.23573899e+00 -1.30592465e+00 6.51354268e-02 -2.12207269e-02
-1.53591168e+00 -1.21876311e+00 9.87828910e-01 -3.56611192e-01
1.49886206e-01 3.61291736e-01 5.09247303e-01 1.43865669e+00
-4.27842140e-03 4.34375554e-01 1.27635002e+00 -2.76251376e-01
-6.47247806e-02 3.70270610e-01 -6.05526805e-01 8.15132558e-01
-1.03557602e-01 4.51304972e-01 -8.55383873e-01 -2.82160640e-01
8.39663327e-01 -2.30051786e-01 -4.87125695e-01 -5.96937954e-01
-1.16067779e+00 6.82572365e-01 1.83945879e-01 2.85871048e-03
-1.87839270e-01 2.22466707e-01 -1.51838720e-01 3.32518846e-01
3.09340984e-01 1.51624098e-01 -2.18656808e-01 1.08792573e-01
-7.99950659e-01 3.22464071e-02 4.86964941e-01 1.15241945e+00
8.38169754e-01 1.67378187e-01 -2.09016800e-01 5.10841310e-01
4.07258570e-01 7.44829416e-01 7.76752532e-02 -1.51115811e+00
3.21673274e-01 2.89400727e-01 2.10051477e-01 -1.30461526e+00
3.21231902e-01 -1.46455988e-01 -5.80288768e-01 3.88506144e-01
3.44593078e-01 -1.76806390e-01 -8.98311675e-01 2.38079906e+00
3.72525722e-01 3.60525489e-01 -3.36829829e-03 8.33834231e-01
7.29728997e-01 3.86568338e-01 -1.35361239e-01 -2.35666946e-01
1.22942090e+00 -7.90244520e-01 -8.99420738e-01 -1.88742906e-01
2.11243764e-01 -9.37091649e-01 9.59269822e-01 -1.04591638e-01
-1.17581272e+00 -3.87882710e-01 -9.98225272e-01 -1.76468957e-03
-2.46595256e-02 1.61432743e-01 4.70018506e-01 9.40805197e-01
-1.48477173e+00 3.91448617e-01 -4.55176592e-01 -2.50223190e-01
7.01560318e-01 5.44152498e-01 -8.82298231e-01 -9.24291387e-02
-1.21259534e+00 5.85261047e-01 -5.37946038e-02 9.46708620e-02
-1.02089286e+00 -7.50769198e-01 -9.59211648e-01 -2.48946771e-01
3.44640434e-01 -7.32882679e-01 9.99954224e-01 -1.52577400e+00
-1.47268164e+00 1.04833436e+00 -3.24424446e-01 -1.92240290e-02
6.63306534e-01 8.11068714e-02 -2.97875226e-01 4.36634332e-01
2.95185059e-01 1.13060415e+00 1.35565853e+00 -1.56565106e+00
-5.11604130e-01 -3.03307444e-01 1.70569211e-01 1.90019265e-01
-5.18927693e-01 1.23135932e-01 -6.98265314e-01 -6.97472692e-01
-1.16731465e-01 -1.13488495e+00 2.87182122e-01 5.36290646e-01
-4.76483762e-01 4.87497002e-01 1.16858399e+00 -7.81647861e-01
7.64456749e-01 -2.49229074e+00 1.91421196e-01 1.84457302e-01
2.42840961e-01 7.51104355e-02 -4.07304347e-01 2.33554319e-01
-3.24588865e-01 2.60929376e-01 -1.52238742e-01 -6.57605946e-01
-1.39366895e-01 1.22383490e-01 -4.44507062e-01 7.51152337e-01
1.75287575e-01 9.50012624e-01 -8.51748824e-01 -5.21425664e-01
8.44830200e-02 8.09207559e-01 -5.58565855e-01 2.24822909e-01
1.15813024e-01 9.95153666e-01 -1.74064413e-01 8.59116852e-01
8.86747122e-01 -1.16490938e-01 2.89209276e-01 -5.20727634e-01
5.01732081e-02 -7.84119368e-02 -1.03517115e+00 1.59227598e+00
-1.43800125e-01 6.63879395e-01 4.14073884e-01 -2.77567118e-01
6.81902885e-01 5.09209335e-01 6.58464849e-01 -6.85512483e-01
1.09532654e-01 2.59346753e-01 -2.08359510e-01 -2.17935666e-01
2.48399287e-01 -1.28062516e-01 1.05711341e-01 5.75669527e-01
1.25439828e-02 2.17872769e-01 -2.35461980e-01 3.28793287e-01
8.50036323e-01 3.72350484e-01 -1.21477835e-01 -1.68529227e-01
5.08588850e-01 -3.88113081e-01 7.20527053e-01 3.64453763e-01
-3.49963605e-01 8.22959721e-01 4.46314484e-01 -1.74300283e-01
-1.03308296e+00 -1.08422649e+00 8.42261780e-03 7.25531936e-01
2.16047391e-01 -3.91947180e-01 -9.91609275e-01 -7.15134561e-01
1.02591421e-02 4.44724411e-01 -8.53745162e-01 -3.16286474e-01
-4.48632538e-01 -2.67562538e-01 7.48154700e-01 2.28667095e-01
7.01550663e-01 -8.27456832e-01 -2.88471401e-01 -2.97819197e-01
-5.99589527e-01 -1.17584205e+00 -1.08814466e+00 -6.43785655e-01
-4.89235580e-01 -1.17134976e+00 -4.60114121e-01 -7.41947114e-01
1.09435189e+00 4.66609001e-01 7.59045005e-01 2.37301871e-01
-1.78432856e-02 7.19825089e-01 -1.68278053e-01 -7.20186308e-02
-5.00508964e-01 -3.20599675e-01 1.59627244e-01 5.53209782e-01
-1.31805703e-01 -6.46520317e-01 -7.39747465e-01 4.92028892e-01
-9.89579916e-01 3.12772095e-01 2.86924750e-01 7.78900862e-01
3.89832109e-01 -2.89192349e-02 3.00916135e-01 -9.03605163e-01
1.71905145e-01 -3.85969698e-01 -3.75901759e-01 2.24105090e-01
-5.45595109e-01 -2.06237108e-01 3.45385671e-01 -4.48913068e-01
-1.14936483e+00 1.99423283e-01 1.03864796e-01 -7.83434987e-01
1.17247716e-01 -8.76642317e-02 -6.47548914e-01 -5.13337433e-01
2.46364281e-01 2.54031628e-01 4.49734926e-01 -2.89192319e-01
3.75882894e-01 4.64895159e-01 6.44198298e-01 -5.58643878e-01
1.17428291e+00 9.67056096e-01 -1.67415917e-01 -4.36579138e-01
-6.62048340e-01 3.24460231e-02 -5.76233208e-01 -3.60898644e-01
5.19433498e-01 -1.16181457e+00 -6.90460086e-01 8.39051664e-01
-1.06514442e+00 -1.54027835e-01 -3.60478669e-01 1.24661908e-01
-6.03429556e-01 2.39278406e-01 -4.61537063e-01 -4.74623024e-01
-7.09998086e-02 -1.20884633e+00 1.11012781e+00 1.04929715e-01
-2.31364712e-01 -8.65199387e-01 -2.01850608e-01 5.99732399e-01
4.06947970e-01 4.19608265e-01 6.23415411e-01 -4.33685705e-02
-9.67413604e-01 7.92396963e-02 -2.81329006e-01 4.09183353e-01
6.92501187e-01 6.32237568e-02 -1.17265713e+00 -6.78742111e-01
2.87370354e-01 -2.02546105e-01 4.73926455e-01 1.02975354e-01
8.67809772e-01 -7.58883715e-01 -2.45094717e-01 1.11196756e+00
1.29288304e+00 1.12257339e-01 9.24238622e-01 1.51960984e-01
8.68301332e-01 8.32959831e-01 5.21274090e-01 3.31425637e-01
3.51536065e-01 7.49473631e-01 5.09405255e-01 -2.68758565e-01
-4.50502783e-01 -6.72175169e-01 7.41309166e-01 4.68127489e-01
7.52754956e-02 -1.69800296e-01 -3.99995208e-01 5.01145303e-01
-1.57299435e+00 -1.00676215e+00 3.56968641e-01 2.13072038e+00
1.01121914e+00 -4.33243752e-01 -1.45359278e-01 -1.73297912e-01
8.39835584e-01 3.29030871e-01 -5.73079288e-01 1.88256893e-02
-3.23599070e-01 7.24755973e-02 5.87056518e-01 5.63676715e-01
-1.00131369e+00 9.40336823e-01 6.21211958e+00 6.88945889e-01
-1.15039527e+00 3.41857344e-01 6.84311807e-01 -3.20500344e-01
-7.05237329e-01 1.64939448e-01 -6.19823873e-01 7.89753139e-01
6.20628834e-01 -1.37457430e-01 6.67402208e-01 3.08932334e-01
1.59162283e-01 1.69542372e-01 -1.15233433e+00 9.29996312e-01
3.22856009e-01 -1.44083667e+00 1.80090621e-01 3.58887464e-01
8.11424851e-01 -5.75653970e-01 5.52770317e-01 -2.07903400e-01
1.51086539e-01 -1.14923298e+00 1.13440359e+00 4.13308114e-01
1.40606689e+00 -6.14557564e-01 4.24494594e-01 -1.12182796e-01
-1.01681018e+00 -5.20442473e-03 2.07132071e-01 2.11267576e-01
-3.94858606e-02 1.14557698e-01 -4.43320483e-01 3.47485304e-01
7.15473354e-01 7.38961577e-01 -5.63208222e-01 2.56743848e-01
-4.37922478e-01 3.48876894e-01 -1.49508521e-01 9.02658522e-01
-1.37363493e-01 -2.81929821e-01 6.32230163e-01 6.15429282e-01
5.17597854e-01 1.84422433e-01 -4.09093015e-02 1.05492914e+00
-3.95640314e-01 -6.94498718e-02 -9.09882843e-01 -7.37828985e-02
6.86858058e-01 1.09324956e+00 -3.29856634e-01 3.73754464e-02
-3.37990373e-01 1.17393458e+00 -2.47737877e-02 4.48270291e-01
-1.03863180e+00 1.11839250e-01 1.07812846e+00 2.04981640e-01
2.41245642e-01 -5.05538248e-02 -2.00026572e-01 -1.11961901e+00
1.93584278e-01 -1.22103155e+00 1.66838050e-01 -8.55095565e-01
-9.99553502e-01 4.71253455e-01 -4.60983887e-02 -1.27697313e+00
-7.86328837e-02 -2.20453873e-01 -4.03488785e-01 9.51855302e-01
-1.48536134e+00 -1.84377015e+00 -2.65853137e-01 9.59755242e-01
1.64419003e-02 -2.57205665e-01 7.83468366e-01 3.86740774e-01
-4.90991801e-01 1.03076005e+00 -1.06558710e-01 2.88308144e-01
1.03863811e+00 -6.29726768e-01 2.15070948e-01 9.40666020e-01
4.29102741e-02 8.29282880e-01 7.19089866e-01 -7.38267839e-01
-1.60709035e+00 -1.01973093e+00 7.41178691e-01 -5.17877996e-01
4.76638436e-01 -4.07729447e-01 -6.72180057e-01 9.52030241e-01
3.74703646e-01 8.80258754e-02 6.98196709e-01 -5.08490384e-01
-6.44467711e-01 -1.80087730e-01 -1.50734925e+00 6.40321910e-01
1.36286449e+00 -8.02867770e-01 -3.48030031e-02 1.09113447e-01
7.09047914e-01 -4.75187212e-01 -6.73831582e-01 3.66075277e-01
7.27545142e-01 -1.05794179e+00 1.04668128e+00 -2.60346621e-01
4.43782270e-01 -4.45264667e-01 -2.58960754e-01 -1.04747570e+00
-2.60125220e-01 -9.40530419e-01 -7.24107549e-02 1.71368515e+00
5.82612269e-02 -8.52806211e-01 9.01961148e-01 7.99736679e-01
2.92961359e-01 -3.66030753e-01 -1.04276085e+00 -6.67626858e-01
-9.48498324e-02 1.38283027e-02 8.70724320e-01 1.07747626e+00
-4.98197943e-01 -1.05755225e-01 -8.80962849e-01 3.34901899e-01
9.15789366e-01 1.73616409e-01 7.93614626e-01 -6.79336667e-01
-5.29029332e-02 -1.24701262e-01 -2.95670778e-01 -6.15505397e-01
4.34232444e-01 -7.11896896e-01 -1.12895221e-01 -8.40350330e-01
3.06843013e-01 -4.59991157e-01 -2.02358708e-01 8.46506119e-01
8.12035203e-02 8.04342330e-01 3.82329822e-01 4.36662495e-01
-9.48444009e-02 5.82262695e-01 1.44213164e+00 -1.82078570e-01
1.04018569e-01 -2.84224093e-01 -1.05751777e+00 4.67368722e-01
8.70821536e-01 -5.43207169e-01 -6.62409008e-01 -4.92536694e-01
-3.96371931e-02 -1.75302904e-02 7.16066539e-01 -6.26859188e-01
2.19309166e-01 -4.08344120e-01 3.43030989e-01 -5.35585098e-02
5.88946760e-01 -9.17919099e-01 7.63690472e-01 3.20849538e-01
-2.07788914e-01 -1.30507916e-01 2.61074543e-01 3.97466809e-01
-1.88781455e-01 3.64248194e-02 8.96886945e-01 1.16236424e-02
-6.00049973e-01 6.60844982e-01 5.75803593e-02 -8.18275362e-02
1.15716004e+00 -5.07087588e-01 -4.25309181e-01 -6.64922655e-01
-4.78336185e-01 6.52211979e-02 1.19916570e+00 6.51153445e-01
7.27216899e-01 -1.66574395e+00 -6.91051424e-01 6.43608391e-01
8.03617164e-02 -3.49907577e-01 3.59976470e-01 7.67727673e-01
-2.45488316e-01 8.30002353e-02 -5.68080008e-01 -3.74048263e-01
-1.56651866e+00 4.82065916e-01 4.02872771e-01 2.37307131e-01
-4.13969666e-01 7.82340944e-01 6.57842755e-01 -3.37758064e-01
5.37286103e-02 4.16082859e-01 1.85718611e-01 1.10155173e-01
4.45056021e-01 5.56176789e-02 -3.45397860e-01 -1.21842349e+00
-4.53289807e-01 6.15768552e-01 -2.21931487e-02 -2.95514971e-01
1.08620393e+00 -5.36623299e-01 -2.12741166e-01 1.69263314e-02
1.37446558e+00 4.58750516e-01 -1.77884150e+00 -1.73228130e-01
-5.11090636e-01 -1.16574299e+00 -1.07948713e-01 -7.04123974e-01
-1.54736316e+00 5.17630637e-01 6.29418910e-01 -4.50841635e-01
1.31420922e+00 -8.99118260e-02 8.63904357e-01 -3.87034535e-01
6.43041074e-01 -8.70716751e-01 1.25984818e-01 -1.08275600e-02
9.51881707e-01 -1.22020018e+00 1.16077717e-03 -6.37983382e-01
-6.10449433e-01 7.55615175e-01 5.97326875e-01 3.59541684e-01
5.43432117e-01 2.68602639e-01 2.65693724e-01 -1.38087571e-01
-5.92158020e-01 1.86058730e-01 1.30568191e-01 7.35877752e-01
-3.00944019e-02 -6.26673326e-02 3.49759549e-01 1.30042821e-01
-2.78285772e-01 3.18492688e-02 4.45226431e-01 8.78008068e-01
2.50045240e-01 -1.41765368e+00 -4.71266806e-01 -5.23168268e-03
-4.26091433e-01 -7.07237124e-02 -5.22828221e-01 6.38552547e-01
4.77994055e-01 8.45086336e-01 -3.59607232e-03 -5.07186830e-01
2.86310948e-02 -3.91879417e-02 6.29511774e-01 -5.12208462e-01
-3.74941081e-01 -1.53795779e-01 -9.27346051e-02 -8.18446219e-01
-7.11561263e-01 -7.59404361e-01 -8.98005188e-01 -7.52815187e-01
-1.37545303e-01 -1.51546761e-01 5.51195025e-01 6.88208103e-01
6.53929293e-01 1.97356064e-02 1.02837861e+00 -7.08105505e-01
-2.60661006e-01 -3.96443039e-01 -4.74920034e-01 6.87006235e-01
5.65576971e-01 -6.96663857e-01 -4.92355913e-01 5.04519343e-01]
|
[12.752466201782227, 0.030547354370355606]
|
2035c1cc-ba85-4a7a-a4ce-338ca38bdfc9
|
accented-speech-recognition-benchmarking-pre
|
2205.08014
| null |
https://arxiv.org/abs/2205.08014v1
|
https://arxiv.org/pdf/2205.08014v1.pdf
|
Accented Speech Recognition: Benchmarking, Pre-training, and Diverse Data
|
Building inclusive speech recognition systems is a crucial step towards developing technologies that speakers of all language varieties can use. Therefore, ASR systems must work for everybody independently of the way they speak. To accomplish this goal, there should be available data sets representing language varieties, and also an understanding of model configuration that is the most helpful in achieving robust understanding of all types of speech. However, there are not enough data sets for accented speech, and for the ones that are already available, more training approaches need to be explored to improve the quality of accented speech recognition. In this paper, we discuss recent progress towards developing more inclusive ASR systems, namely, the importance of building new data sets representing linguistic diversity, and exploring novel training approaches to improve performance for all users. We address recent directions within benchmarking ASR systems for accented speech, measure the effects of wav2vec 2.0 pre-training on accented speech recognition, and highlight corpora relevant for diverse ASR evaluations.
|
['Gary Wang', 'Suzan Schwartz', 'Andrew Rosenberg', 'Bhuvana Ramabhadran', 'Levi King', 'Wei Han', 'Pavel Golik', 'Daan van Esch', 'Chung-Cheng Chiu', 'Zhehuai Chen', 'Alëna Aksënova']
|
2022-05-16
| null | null | null | null |
['accented-speech-recognition']
|
['speech']
|
[-6.58181161e-02 -7.67123029e-02 5.24374135e-02 -7.88705826e-01
-9.08666432e-01 -7.76017606e-01 3.58582616e-01 -1.22214794e-01
-4.78563935e-01 2.88851440e-01 6.10393286e-01 -6.78481698e-01
2.63508558e-01 -4.93703365e-01 -5.23322225e-02 -5.81687033e-01
3.76518279e-01 6.44368887e-01 -6.19130395e-02 -9.24309850e-01
-1.29181609e-01 7.03835964e-01 -1.45540214e+00 1.30513832e-01
7.82520652e-01 5.28394401e-01 3.73124689e-01 8.51602972e-01
-4.57507759e-01 5.70543766e-01 -1.01350892e+00 -4.99745995e-01
1.83219790e-01 -3.84823322e-01 -8.36118639e-01 3.43144745e-01
5.57566762e-01 5.75908013e-02 -1.84502348e-01 9.58559453e-01
8.04477394e-01 4.98134255e-01 3.77073765e-01 -5.47935426e-01
-9.02192354e-01 1.06224155e+00 9.47937146e-02 4.81333196e-01
1.94522038e-01 3.56945276e-01 9.03023958e-01 -7.39700556e-01
3.70366007e-01 1.31181622e+00 6.34647235e-02 9.68376100e-01
-1.02145445e+00 -5.12609661e-01 3.30305904e-01 6.29784986e-02
-1.29448700e+00 -1.12734354e+00 6.23820961e-01 -1.91223025e-01
1.20083165e+00 6.13275528e-01 4.52916384e-01 1.12012398e+00
-6.90678537e-01 6.64052308e-01 9.08686996e-01 -7.09790051e-01
8.07554349e-02 4.78110254e-01 4.75290656e-01 2.28952557e-01
-1.43678576e-01 -1.28662437e-01 -5.17833471e-01 3.30195189e-01
4.06163126e-01 -6.54357016e-01 -4.01628017e-01 9.52961966e-02
-1.10632646e+00 7.49668360e-01 8.11559558e-02 6.65818691e-01
-1.51101336e-01 -4.17668521e-01 5.87437153e-01 5.38126647e-01
5.40644884e-01 6.83749616e-01 -6.42171502e-01 -4.95557338e-01
-9.06780362e-01 3.08789350e-02 8.11574757e-01 8.78923714e-01
3.41831982e-01 7.29506671e-01 1.00629993e-01 1.53910506e+00
2.51125276e-01 7.49153376e-01 6.43496335e-01 -7.38220274e-01
4.91862506e-01 4.87209857e-01 -2.53728658e-01 -1.35214716e-01
-1.73378885e-01 -4.08651710e-01 -4.81564641e-01 1.41109139e-01
4.18407232e-01 -3.07569444e-01 -1.19931328e+00 1.50992811e+00
2.15418842e-02 -3.14779967e-01 5.59150279e-01 8.33829820e-01
9.36873257e-01 8.74884903e-01 7.26121664e-02 -3.37624587e-02
1.36499298e+00 -7.50846744e-01 -6.93286717e-01 -4.91675913e-01
7.94103622e-01 -1.25008547e+00 1.31737244e+00 1.97828591e-01
-1.09674621e+00 -6.46421969e-01 -1.05683613e+00 -7.80911073e-02
-7.48118401e-01 2.76312064e-02 2.57432431e-01 1.40985942e+00
-1.25104058e+00 -1.58456057e-01 -6.54675901e-01 -3.60867888e-01
4.90319729e-02 3.06247920e-01 -4.73477334e-01 -1.74647987e-01
-1.34199667e+00 1.14416909e+00 2.24216267e-01 1.39082089e-01
-4.14985895e-01 -6.52687967e-01 -1.05366528e+00 1.38683468e-01
-8.18166137e-02 -1.77196711e-01 1.45822525e+00 -1.05846107e+00
-1.49331629e+00 8.56544912e-01 -2.19097227e-01 -4.50624198e-01
1.17218092e-01 -2.53155408e-03 -9.02701974e-01 -5.34162223e-01
-4.04886127e-01 5.20896733e-01 3.55349272e-01 -1.08697641e+00
-6.76706970e-01 -5.65673172e-01 -1.25457466e-01 5.01566291e-01
-5.17245114e-01 4.74907070e-01 -2.40224510e-01 -4.48034644e-01
-1.28176898e-01 -9.65665877e-01 -2.64965713e-01 -7.90893495e-01
-2.69892514e-01 -2.22802728e-01 7.71767378e-01 -1.09340775e+00
1.17517662e+00 -2.45573139e+00 -3.54958400e-02 2.86726981e-01
-1.19396776e-01 1.00251639e+00 -1.97616503e-01 1.86002150e-01
5.96063286e-02 2.74297148e-01 7.78486729e-02 -4.22859043e-01
9.39011052e-02 2.93993264e-01 -3.87969375e-01 6.75911009e-02
2.07733244e-01 5.20117044e-01 -6.24918103e-01 -3.74499359e-03
5.29398203e-01 8.53972316e-01 -2.21145973e-01 2.74245262e-01
-1.46175027e-01 2.55110621e-01 -6.29929677e-02 4.05799419e-01
4.74563628e-01 4.46009994e-01 5.92585616e-02 2.47378796e-01
-3.84829611e-01 8.15859377e-01 -1.18912578e+00 1.07677042e+00
-9.43934798e-01 7.78965235e-01 3.62417728e-01 -8.40365529e-01
1.31387985e+00 6.65638924e-01 1.21793188e-01 -5.61866701e-01
-7.45676178e-03 5.12198448e-01 4.57775205e-01 -1.63553149e-01
7.84373879e-01 -6.93801790e-02 7.06798062e-02 -1.74214039e-02
1.17020376e-01 -3.59826922e-01 3.06430697e-01 -2.11186290e-01
7.39965320e-01 -6.17961049e-01 3.24034393e-02 -2.32737035e-01
6.38288677e-01 7.45450184e-02 4.70329314e-01 6.00398004e-01
-3.12541068e-01 6.76139891e-01 -3.09677068e-02 -3.16864163e-01
-1.08952343e+00 -9.94936764e-01 -1.05247125e-01 1.35374272e+00
-3.66631508e-01 -1.49306133e-01 -7.90259004e-01 -2.64144182e-01
-3.26385230e-01 1.09177792e+00 -1.01821005e-01 9.80772674e-02
-7.37878919e-01 -5.82470417e-01 7.71315336e-01 4.87390012e-01
2.81119466e-01 -1.08042574e+00 9.41888541e-02 1.94154054e-01
-1.23498430e-02 -1.19221866e+00 -6.24219954e-01 3.38505030e-01
-3.64697039e-01 -4.47187752e-01 -8.81656110e-01 -1.14754951e+00
2.18648046e-01 2.84738481e-01 1.10799849e+00 -1.09554976e-01
5.82901873e-02 3.78584415e-01 -4.97875601e-01 -5.55631876e-01
-1.10599911e+00 5.24010718e-01 1.10659227e-01 -9.37728509e-02
6.92430019e-01 -1.83667839e-01 -3.89566980e-02 2.50611216e-01
-5.47153175e-01 -3.04496437e-01 4.57044989e-01 6.22583330e-01
3.39652419e-01 -5.39961345e-02 6.50115490e-01 -7.33042538e-01
7.00792849e-01 -2.62160331e-01 -4.30114150e-01 3.79269421e-01
-2.55954713e-01 -5.41524664e-02 7.84510851e-01 -3.43970984e-01
-1.25821292e+00 1.00044951e-01 -9.98909056e-01 -1.34039015e-01
-5.87613821e-01 4.07124907e-01 -5.69047213e-01 1.01055779e-01
7.74128199e-01 2.75009871e-01 9.50818975e-03 -5.22017896e-01
6.91356897e-01 1.41097069e+00 3.54369015e-01 -4.12386000e-01
4.28482652e-01 -1.69572666e-01 -6.66754246e-01 -1.85360205e+00
-4.33014065e-01 -8.19701552e-01 -5.73943913e-01 3.16227484e-03
7.66766071e-01 -8.55556428e-01 -1.78660721e-01 4.46449786e-01
-1.07821941e+00 -2.26375252e-01 -3.17124426e-01 4.91672456e-01
-3.18745613e-01 4.45473613e-03 -3.63947541e-01 -9.40339327e-01
-4.01614368e-01 -1.72619188e+00 8.58403802e-01 2.65777260e-01
-4.01337177e-01 -1.16168141e+00 1.45940632e-01 7.58770704e-01
8.20523679e-01 -6.87450349e-01 7.55532622e-01 -1.15422142e+00
-2.05340490e-01 -1.09434366e-01 3.43267500e-01 8.56866479e-01
4.67654526e-01 2.24114090e-01 -1.26045024e+00 -1.53358445e-01
-1.58681989e-01 -1.78447783e-01 5.38787305e-01 4.54158396e-01
6.47071064e-01 -2.06707045e-01 1.14273630e-01 4.30161059e-01
7.74720430e-01 5.12827277e-01 6.53057396e-01 3.18063274e-02
6.42598987e-01 7.65699387e-01 1.96701050e-01 -2.68694788e-01
4.56328332e-01 7.69084394e-01 -2.69010961e-01 -2.61143297e-01
-3.92071486e-01 2.10412964e-01 5.38226604e-01 1.34423566e+00
6.35810420e-02 -3.50904554e-01 -1.04640603e+00 7.98190534e-01
-1.11571038e+00 -1.11711025e+00 -3.73309813e-02 2.30395842e+00
8.05807114e-01 1.24671487e-02 3.90144318e-01 2.78715461e-01
8.12963367e-01 3.47558349e-01 -3.14371169e-01 -9.62992668e-01
-5.83702028e-01 3.19648057e-01 2.79760301e-01 8.74153852e-01
-1.03222859e+00 1.28606915e+00 5.93003225e+00 5.98987818e-01
-1.55248988e+00 -1.17104143e-01 7.96255529e-01 6.50532171e-02
-3.60432625e-01 -3.39999616e-01 -1.10712934e+00 2.84105182e-01
1.41294980e+00 -2.84315437e-01 4.98698443e-01 1.04256713e+00
2.73072273e-01 4.32948858e-01 -6.76939189e-01 9.04666603e-01
5.51778972e-02 -1.08493209e+00 -1.38574064e-01 -1.25991404e-01
4.14390445e-01 5.18425584e-01 1.30854681e-01 3.92162293e-01
6.28490210e-01 -1.08362627e+00 4.54570115e-01 -1.62119180e-01
5.16704202e-01 -8.48516047e-01 7.78579414e-01 -2.59420574e-02
-1.03723681e+00 2.64088303e-01 -2.46132404e-01 3.24127942e-01
1.39181077e-01 2.98447788e-01 -1.14391851e+00 1.89745799e-01
4.39150900e-01 2.49219298e-01 -5.86592972e-01 7.95852423e-01
3.76346558e-02 9.99659896e-01 -3.48420709e-01 -3.61952603e-01
1.63192973e-01 -6.05013371e-02 5.99115610e-01 1.59299338e+00
3.13705772e-01 -8.88730213e-02 2.21875668e-01 1.61699757e-01
1.29832655e-01 4.82048064e-01 -6.83875442e-01 -3.00328314e-01
9.13947701e-01 1.02342451e+00 -3.66969734e-01 -1.25872478e-01
-6.99464381e-01 7.91136622e-01 4.07365263e-01 3.22872132e-01
-1.45624146e-01 -3.56472731e-01 1.44044137e+00 5.92939034e-02
5.27745374e-02 -5.70189297e-01 -3.43014240e-01 -9.65403318e-01
-5.81119210e-02 -1.38059473e+00 3.41303647e-01 -4.48466808e-01
-1.19881451e+00 9.90871549e-01 -4.42602664e-01 -5.77046156e-01
-5.66674590e-01 -8.02297652e-01 -5.55256724e-01 1.29523039e+00
-1.19658768e+00 -1.16566074e+00 -3.51708904e-02 2.60444075e-01
1.10895813e+00 -5.61591148e-01 1.03829896e+00 3.25763941e-01
-5.42187691e-01 7.22089648e-01 8.94433931e-02 2.97363818e-01
7.23212421e-01 -1.32574952e+00 9.19533193e-01 9.24954712e-01
6.60068333e-01 6.32297456e-01 6.20867193e-01 -2.88506359e-01
-1.31744421e+00 -9.36393261e-01 8.86569679e-01 -5.04680574e-01
6.58308923e-01 -4.70135063e-01 -1.13658786e+00 8.17890227e-01
3.07243913e-01 -4.24557924e-01 9.40445840e-01 7.10839331e-01
-2.55893975e-01 -2.12108523e-01 -7.35386014e-01 8.47559392e-01
8.15305471e-01 -6.41891420e-01 -5.86760402e-01 9.23949033e-02
7.77003527e-01 -4.35520232e-01 -9.27165866e-01 9.58705172e-02
2.21350700e-01 -7.53996611e-01 8.67314458e-01 -6.07634366e-01
-2.59381980e-01 -2.50205427e-01 -4.90750968e-01 -1.97140515e+00
-1.99268475e-01 -5.97974837e-01 5.13373435e-01 1.70823979e+00
9.00584936e-01 -6.14609897e-01 5.74291050e-01 6.84429407e-01
-5.40916622e-01 -4.16234434e-01 -8.78851533e-01 -8.98242831e-01
3.56487393e-01 -6.46267593e-01 8.63111198e-01 1.06937790e+00
-1.18749931e-01 5.67133307e-01 4.25599366e-02 3.55687797e-01
8.65639299e-02 -4.69556451e-01 6.25870168e-01 -8.41283441e-01
-1.84194315e-02 -6.54710770e-01 -7.90196061e-01 -8.23575318e-01
2.83536345e-01 -8.29230785e-01 5.87608553e-02 -1.39751518e+00
-5.69725633e-01 -6.57333612e-01 -9.86240059e-02 4.19641435e-01
-3.66248578e-01 -6.12974465e-02 2.42647350e-01 -2.47370541e-01
-2.00224873e-02 3.79622489e-01 8.52628767e-01 -2.64859676e-01
-4.78061736e-01 1.63315654e-01 -8.76338959e-01 3.89367104e-01
1.18320894e+00 1.66326985e-01 -5.80035448e-01 -6.88800335e-01
-4.71758157e-01 -2.31484368e-01 -2.72901416e-01 -1.00641394e+00
2.36893043e-01 -1.08014829e-01 1.33148044e-01 -2.55512208e-01
5.69809914e-01 -4.27478313e-01 -1.25724033e-01 -1.69576064e-01
-3.48791927e-01 2.08535224e-01 2.59995103e-01 -3.29769373e-01
-4.38232511e-01 -2.28542045e-01 1.10699022e+00 -9.24339890e-02
-9.59376097e-01 9.82775912e-02 -5.02695620e-01 3.99534374e-01
5.62694967e-01 -1.02128638e-02 -3.97503227e-01 -4.99825776e-01
-7.76677608e-01 1.06360376e-01 5.19032359e-01 8.08172882e-01
3.59297901e-01 -9.00155842e-01 -1.00574827e+00 5.32732010e-01
2.68841565e-01 -2.75464445e-01 1.78598017e-01 1.65835291e-01
-4.16622043e-01 4.79426473e-01 -1.27312839e-02 -1.69589326e-01
-1.66361785e+00 3.74758214e-01 5.65107107e-01 2.04919800e-01
-2.84382164e-01 1.02975106e+00 -2.23381594e-02 -8.55607271e-01
3.12686831e-01 5.01707532e-02 -3.69377226e-01 -7.80907124e-02
7.44939148e-01 2.10629135e-01 5.90503335e-01 -1.16719806e+00
-3.27920765e-01 1.44672379e-01 -3.93051684e-01 -4.50130105e-01
1.17375374e+00 -2.05462486e-01 1.47415906e-01 5.44246972e-01
9.52214181e-01 3.91165823e-01 -6.64112210e-01 -4.58720773e-02
5.57798147e-02 -2.15430915e-01 2.74719566e-01 -8.03960264e-01
-9.88313437e-01 9.60901976e-01 6.33173823e-01 5.93808174e-01
9.28232551e-01 3.46053578e-02 5.71964383e-01 3.71365666e-01
9.82997939e-02 -1.30588603e+00 -6.06599867e-01 8.35769296e-01
7.83159852e-01 -1.27904296e+00 -5.90639353e-01 -3.91037583e-01
-8.81538570e-01 9.62643623e-01 5.82638621e-01 3.89610976e-01
6.63690984e-01 4.44752365e-01 9.21553195e-01 7.45278671e-02
-5.20604312e-01 -5.81801057e-01 2.61355013e-01 9.25391495e-01
1.04842293e+00 4.32892919e-01 2.38417350e-02 2.09835380e-01
-8.97511482e-01 -6.95989430e-01 4.30631012e-01 5.08470774e-01
-5.60360014e-01 -1.49206364e+00 -4.88587677e-01 4.18784916e-01
-3.98418665e-01 -2.56093621e-01 -6.72346413e-01 6.09381080e-01
-2.04854324e-01 1.29339194e+00 2.28097484e-01 -3.89810592e-01
6.67576969e-01 4.95506316e-01 2.56909132e-01 -8.99216890e-01
-4.48026121e-01 -1.30430251e-01 7.49213398e-01 4.65334915e-02
1.11888394e-01 -7.66971886e-01 -8.93393517e-01 -3.73749346e-01
-3.30341041e-01 3.58880490e-01 1.10203886e+00 7.95534730e-01
1.55509517e-01 3.87021959e-01 6.85405493e-01 -5.51479995e-01
-5.34808338e-01 -9.86416042e-01 -5.66146314e-01 3.06613952e-01
3.13396960e-01 -1.65054798e-01 -3.74923050e-01 3.13360803e-02]
|
[14.257556915283203, 6.75278377532959]
|
2e3cc3af-62ab-431d-ae5e-578603c02309
|
links-a-high-dimensional-online-clustering
|
1801.10123
| null |
http://arxiv.org/abs/1801.10123v1
|
http://arxiv.org/pdf/1801.10123v1.pdf
|
Links: A High-Dimensional Online Clustering Method
|
We present a novel algorithm, called Links, designed to perform online
clustering on unit vectors in a high-dimensional Euclidean space. The algorithm
is appropriate when it is necessary to cluster data efficiently as it streams
in, and is to be contrasted with traditional batch clustering algorithms that
have access to all data at once. For example, Links has been successfully
applied to embedding vectors generated from face images or voice recordings for
the purpose of recognizing people, thereby providing real-time identification
during video or audio capture.
|
['Carlton Downey', 'Philip Andrew Mansfield', 'Li Wan', 'Ignacio Lopez Moreno', 'Quan Wang']
|
2018-01-30
| null | null | null | null |
['online-clustering']
|
['computer-vision']
|
[ 9.93392542e-02 -2.34225124e-01 -2.10306030e-02 -4.88117248e-01
-3.24667931e-01 -4.57437128e-01 5.20650089e-01 5.38611591e-01
-5.26932240e-01 -3.21869925e-02 1.40523076e-01 -2.90742844e-01
-3.49920005e-01 -6.21470392e-01 -7.34407008e-02 -7.30862260e-01
-6.92445457e-01 7.06748664e-01 -1.58040568e-01 2.21877903e-01
4.34533972e-03 7.63311446e-01 -2.11635470e+00 3.11867267e-01
1.38159141e-01 8.98049772e-01 -1.08727463e-01 8.54317725e-01
-4.05090511e-01 6.06468201e-01 -5.13437152e-01 -2.35697791e-01
8.84064585e-02 -4.11351174e-01 -5.30954480e-01 4.42286372e-01
-3.19657438e-02 -1.12011544e-01 -4.51821446e-01 6.15645587e-01
6.64515972e-01 2.64421821e-01 7.90760636e-01 -1.48308825e+00
-1.32250890e-01 4.41443950e-01 -1.41942620e-01 1.86275497e-01
9.09661055e-01 -5.07855952e-01 8.58882844e-01 -1.11505687e+00
6.68399632e-01 9.63717639e-01 3.94182563e-01 5.40088475e-01
-1.17503703e+00 -4.38404232e-01 -1.84586182e-01 4.67633814e-01
-1.80719268e+00 -7.33555019e-01 7.30188310e-01 -6.31394029e-01
7.61996746e-01 3.75705332e-01 6.84503198e-01 9.75205898e-01
-3.06400329e-01 1.07042706e+00 3.47098470e-01 -5.15608132e-01
6.16534531e-01 1.25942513e-01 3.94745171e-02 9.90728959e-02
-1.36654049e-01 -2.48822972e-01 -6.78815782e-01 -5.94486833e-01
2.30643019e-01 1.99413568e-01 -1.70434132e-01 -3.83813024e-01
-1.08324254e+00 8.74224246e-01 -1.57506242e-01 5.33533931e-01
-4.89589989e-01 -2.26191789e-01 7.04868138e-01 6.09753549e-01
4.68322188e-01 4.65273634e-02 -5.66646345e-02 -6.27101839e-01
-1.13955486e+00 -1.16140276e-01 1.11770427e+00 9.96537089e-01
5.36986113e-01 2.81479508e-02 2.89501607e-01 7.76310444e-01
2.30575606e-01 -3.01283784e-03 6.82506323e-01 -7.32594132e-01
1.91300243e-01 5.67591488e-01 -1.27085984e-01 -9.74662960e-01
-2.98704952e-01 2.17413008e-01 -9.76502776e-01 -5.00460528e-02
-1.50142070e-02 -1.71518981e-01 -2.74144650e-01 1.15532649e+00
5.99235177e-01 6.14843547e-01 1.55156434e-01 6.71868384e-01
6.94758177e-01 7.18739092e-01 -2.80029029e-01 -6.81157589e-01
9.19967353e-01 -2.22289816e-01 -8.27240229e-01 3.63029242e-01
4.58995998e-01 -7.35372007e-01 5.97172141e-01 6.27215624e-01
-7.48265505e-01 -4.81926143e-01 -7.44895220e-01 5.97529650e-01
-4.97046083e-01 -2.27199703e-01 5.48758149e-01 1.06357539e+00
-1.16283178e+00 5.56295514e-01 -6.32212400e-01 -4.60445106e-01
2.67025769e-01 6.81752026e-01 -4.93148804e-01 -1.37181040e-02
-9.46401536e-01 3.66726629e-02 2.97564059e-01 1.44675612e-01
-5.15706718e-01 -3.38242888e-01 -8.35848093e-01 -5.88088669e-02
-4.78321165e-02 2.32703779e-02 8.48170102e-01 -9.27116394e-01
-1.57149506e+00 7.50177443e-01 -2.80164301e-01 -2.47545898e-01
3.59643787e-01 1.82045829e-02 -7.65214205e-01 4.16681141e-01
-4.12159830e-01 3.01538706e-01 1.25201154e+00 -8.15222621e-01
-5.41206598e-01 -4.16013360e-01 -5.39488494e-01 1.52211607e-01
-9.50096607e-01 4.23082501e-01 -6.05496466e-01 -5.82871735e-01
1.14085652e-01 -7.08486915e-01 -1.22268088e-01 -3.65383178e-01
-6.73749372e-02 -6.06521606e-01 1.34088469e+00 -2.52366722e-01
1.41415465e+00 -2.35455298e+00 3.35984051e-01 6.48524880e-01
1.62692055e-01 2.41726547e-01 1.46156892e-01 7.97336876e-01
-2.15986311e-01 -9.70302969e-02 3.08746863e-02 -4.99506384e-01
3.98511551e-02 2.55052179e-01 -1.34339720e-01 7.74956226e-01
6.84604887e-03 3.44447374e-01 -8.77224565e-01 -6.24654293e-01
4.66343045e-01 5.28054416e-01 -5.11814892e-01 3.57682586e-01
2.85810977e-01 4.48527634e-02 -3.18618864e-02 5.26669323e-01
4.03776854e-01 3.63120139e-02 3.22199643e-01 1.52621120e-01
-2.06460692e-02 -1.20684654e-01 -1.62449276e+00 1.31047857e+00
9.01281368e-03 1.10626543e+00 3.75128329e-01 -1.27768755e+00
7.62196183e-01 8.58974934e-01 1.43365967e+00 -8.75212774e-02
3.41617316e-01 -2.18486607e-01 -3.78813565e-01 -3.76008451e-01
4.64605093e-01 1.39980376e-01 -5.46627007e-02 7.94599533e-01
2.19978392e-01 3.34084094e-01 2.41124228e-01 4.41417098e-01
9.81097281e-01 -6.26040578e-01 8.43126401e-02 4.54573333e-03
5.86852908e-01 -1.87122539e-01 1.09231144e-01 3.45174134e-01
-1.14069253e-01 5.70521474e-01 1.47290245e-01 -1.92163318e-01
-1.02593756e+00 -1.06317139e+00 -3.11520755e-01 9.68110800e-01
-1.17617197e-01 -9.12930906e-01 -6.32899225e-01 -4.17281419e-01
6.56930730e-03 9.81694683e-02 -4.45104867e-01 7.73736835e-02
-2.00652778e-01 -5.21206677e-01 3.71541977e-01 3.22493285e-01
-2.63845444e-01 -9.61090684e-01 -4.88455594e-01 4.80298072e-01
2.06651464e-01 -1.05081832e+00 -4.67917055e-01 1.81151137e-01
-6.79517925e-01 -1.20147192e+00 -4.32344884e-01 -8.69633496e-01
6.72957301e-01 2.63713747e-01 6.88078463e-01 -5.51949739e-02
-7.09746182e-01 1.21014035e+00 -7.45601356e-01 -2.54410386e-01
-2.34835427e-02 -1.05974957e-01 6.39798880e-01 7.47387230e-01
9.51974928e-01 -3.89579237e-01 -3.23464423e-01 4.32045370e-01
-1.17376316e+00 -6.79921389e-01 -1.47679716e-01 6.17630482e-01
1.74355924e-01 5.15679121e-01 4.51369792e-01 -7.54437864e-01
9.74927425e-01 -7.99596727e-01 -2.78873026e-01 -2.80059241e-02
-4.15947497e-01 -5.79071343e-01 6.65227950e-01 -5.52343547e-01
-2.93244809e-01 4.47580516e-01 -1.43064827e-01 -8.44799936e-01
-4.88639146e-01 2.82900870e-01 -1.48426414e-01 -4.26570810e-02
3.78702939e-01 2.60389626e-01 2.72571206e-01 -4.49166894e-01
4.04444277e-01 1.43922734e+00 2.61779666e-01 -3.03636432e-01
5.89472413e-01 5.38425744e-01 -3.99708748e-01 -1.53308392e+00
1.96436629e-01 -1.31996334e+00 -1.17739952e+00 -7.24442303e-01
6.73353672e-01 -7.08256841e-01 -1.07215881e+00 2.81659663e-01
-7.16936469e-01 -1.89234857e-02 -4.63678896e-01 7.22712994e-01
-5.66048324e-01 5.65468729e-01 -4.81544405e-01 -1.04006481e+00
-8.80661383e-02 -5.92565596e-01 6.99916065e-01 -3.38565782e-02
-3.83164316e-01 -1.11310732e+00 4.28699739e-02 -2.90575594e-01
-4.72198008e-03 1.36775911e-01 3.64361376e-01 -9.68344867e-01
1.18565328e-01 -7.11177826e-01 2.79231519e-01 4.35005963e-01
4.33373958e-01 4.28176612e-01 -9.29915011e-01 -6.12219453e-01
-4.94631520e-03 -1.87864155e-01 3.00238639e-01 1.67761892e-01
1.25538945e+00 -2.74878979e-01 -3.63301158e-01 2.66874671e-01
1.15009809e+00 4.07134324e-01 3.91804695e-01 -1.48303822e-01
4.11906540e-01 9.48843539e-01 3.48912895e-01 1.04936862e+00
7.44012296e-02 5.37432849e-01 -1.72876064e-02 5.54000996e-02
5.12593508e-01 5.71099967e-02 3.90321285e-01 1.32811415e+00
3.04766536e-01 -2.33189851e-01 -8.10143113e-01 6.85699463e-01
-1.89618528e+00 -1.38692689e+00 -1.02799870e-01 2.18394899e+00
4.14829344e-01 -2.29183480e-01 7.26959050e-01 7.26853430e-01
8.11957955e-01 -3.23236175e-02 -3.74813050e-01 -4.62034196e-01
2.20194548e-01 2.24336058e-01 2.32944563e-01 2.26614520e-01
-1.11684680e+00 6.29495442e-01 7.83972836e+00 6.88338637e-01
-8.94284904e-01 1.61973312e-02 3.29065561e-01 -3.89382243e-01
-7.60339200e-02 -2.74706125e-01 -5.81220031e-01 6.35516882e-01
1.39023256e+00 -3.10092419e-01 3.85344148e-01 8.70007932e-01
1.66993275e-01 6.79966658e-02 -1.34170091e+00 1.55686343e+00
4.15254980e-01 -1.07507944e+00 -7.21468702e-02 1.28240943e-01
3.46023023e-01 -2.56916016e-01 9.66199934e-02 -1.53524116e-01
1.88451290e-01 -9.31746423e-01 2.27063745e-01 1.83886185e-01
7.32030809e-01 -1.18774283e+00 4.10429269e-01 3.42023462e-01
-1.37989986e+00 -1.76953271e-01 -3.48712474e-01 4.84693609e-02
2.70836115e-01 5.80747068e-01 -1.12449276e+00 3.17853719e-01
8.56344163e-01 6.96725309e-01 -2.13030487e-01 1.19515538e+00
4.37027514e-01 7.26641417e-01 -6.20888770e-01 -7.01709911e-02
3.15152764e-01 -4.87852186e-01 3.95997345e-01 1.44626737e+00
3.51287931e-01 2.19454229e-01 2.12854102e-01 -3.19868959e-02
5.24463169e-02 5.86055040e-01 -9.80280638e-01 -1.93507016e-01
7.40006387e-01 1.17875111e+00 -8.38904321e-01 -5.07890224e-01
-3.66617352e-01 1.14570129e+00 -1.80352435e-01 1.72452271e-01
-4.38631862e-01 -8.19880545e-01 9.15062249e-01 1.19759597e-01
4.71881956e-01 -6.16791368e-01 3.37874234e-01 -9.96121049e-01
-2.73292661e-01 -5.31920016e-01 7.63759434e-01 -1.81484878e-01
-1.20726573e+00 6.04933560e-01 -4.48440202e-02 -1.51656246e+00
-7.12350249e-01 -5.80298245e-01 -4.95344669e-01 2.60377109e-01
-8.61686528e-01 -4.34548020e-01 -7.51273111e-02 1.35900450e+00
4.19738621e-01 -5.50304651e-01 9.14076984e-01 6.34903967e-01
-6.39592528e-01 6.29830837e-01 7.03045189e-01 3.73006940e-01
6.28910661e-01 -1.17873478e+00 1.02084063e-01 5.51423788e-01
7.34503031e-01 5.89431167e-01 5.67180872e-01 -2.50087202e-01
-1.73063207e+00 -8.96853566e-01 7.84022689e-01 -3.67968380e-01
6.31474376e-01 -7.55155325e-01 -7.38658905e-01 3.22137535e-01
1.31768286e-01 -5.32648563e-02 1.47084737e+00 1.19765304e-01
-1.62498593e-01 -1.70294330e-01 -1.09564519e+00 2.92751074e-01
8.02712262e-01 -7.58912802e-01 -3.12791497e-01 5.09348631e-01
3.03645581e-01 1.13541320e-01 -1.12847865e+00 -1.15368888e-01
4.10385907e-01 -7.67037809e-01 9.48542774e-01 -6.58706486e-01
-3.95640641e-01 -3.46733540e-01 -2.21861079e-01 -9.63949025e-01
-3.10394973e-01 -1.19341886e+00 -2.05615729e-01 1.35203612e+00
2.51575708e-02 -3.63720387e-01 1.06554353e+00 3.92652601e-01
4.56423491e-01 -2.05457643e-01 -1.29207087e+00 -7.53376305e-01
-5.63284278e-01 -7.92217731e-01 6.02190435e-01 1.19030178e+00
5.54591656e-01 1.31638616e-01 -3.96124959e-01 1.07797049e-01
8.96721661e-01 -1.55445099e-01 8.38330209e-01 -1.47462225e+00
7.82443359e-02 -3.58233184e-01 -9.10606444e-01 -8.07945788e-01
3.19224447e-01 -7.75531292e-01 -2.02281654e-01 -1.03417957e+00
-3.67967218e-01 -4.36880827e-01 -1.88402578e-01 8.39183405e-02
3.57595742e-01 3.91586810e-01 6.82171881e-02 4.77799445e-01
-7.28445649e-01 5.14795363e-01 1.10411204e-01 -1.92090981e-02
-4.89875972e-01 3.01181108e-01 -3.30160916e-01 2.95374006e-01
6.72525465e-01 -2.91858763e-01 -5.52411497e-01 2.53513195e-02
-4.85199466e-02 -1.93176027e-02 -1.87955514e-01 -1.08664358e+00
8.20537090e-01 2.02878863e-01 3.97625774e-01 -6.89855337e-01
4.26881641e-01 -1.35203671e+00 3.20677519e-01 1.64717004e-01
-8.19347128e-02 1.46946877e-01 -1.33682653e-01 7.55350173e-01
-5.66669703e-01 -1.85347840e-01 5.62783301e-01 1.79918110e-01
-8.33910763e-01 4.85063553e-01 -9.44838583e-01 -3.69334221e-01
1.52974427e+00 -5.98809183e-01 4.46879894e-01 -6.29536390e-01
-9.15051401e-01 1.81083262e-01 4.34891731e-01 6.10156178e-01
9.52908278e-01 -1.61447597e+00 -4.94630724e-01 6.59524083e-01
1.37643397e-01 -3.28930020e-01 5.56863919e-02 4.97795790e-01
-4.78177875e-01 3.23277175e-01 -3.91991623e-02 -8.52441251e-01
-1.70064497e+00 8.25685203e-01 -3.21756274e-01 5.24240315e-01
-6.67558908e-01 7.82230079e-01 -6.89650297e-01 -1.81364521e-01
6.08973861e-01 3.47215563e-01 -4.28734511e-01 7.64737546e-01
9.48076725e-01 5.05881190e-01 2.82523185e-01 -8.24486673e-01
-4.26245928e-01 4.53471690e-01 6.66992888e-02 -3.48714679e-01
1.37718999e+00 -2.04291150e-01 -2.39353523e-01 8.09011698e-01
1.57893097e+00 -2.62290686e-01 -8.71452510e-01 -2.16501653e-01
1.70687884e-01 -6.65135503e-01 -1.02699734e-01 4.72997487e-01
-8.01102579e-01 7.80252635e-01 7.08672762e-01 7.52212703e-01
1.02858961e+00 2.41076332e-02 5.53814411e-01 5.66716433e-01
3.71345818e-01 -1.51819062e+00 1.88421905e-01 2.83624262e-01
2.11438596e-01 -9.32873547e-01 -9.39194486e-02 3.44119258e-02
-6.12592638e-01 1.28876650e+00 3.76941115e-02 -1.06088229e-01
1.19913387e+00 4.58960354e-01 1.07678177e-03 -1.29209921e-01
-8.72526109e-01 -1.89000174e-01 8.47429857e-02 9.54612136e-01
2.87493825e-01 1.86944241e-03 6.49785772e-02 -7.24560916e-02
-1.44048005e-01 -3.89547378e-01 3.72917980e-01 1.12069309e+00
-5.72740912e-01 -1.24963987e+00 -6.78972304e-01 6.73203707e-01
-3.13187122e-01 2.56092161e-01 -6.08185530e-01 2.78593421e-01
-1.60973832e-01 1.38497615e+00 4.69576955e-01 -6.85904682e-01
9.32208374e-02 3.48601460e-01 1.37454554e-01 -5.01765609e-01
-5.00014961e-01 1.35734454e-01 -1.66949570e-01 -7.02757239e-01
-5.89331865e-01 -9.78048563e-01 -8.75358224e-01 -4.63210285e-01
-1.20711021e-01 7.53544092e-01 7.81492233e-01 4.96056378e-01
3.02996635e-01 5.63630499e-02 1.37123775e+00 -9.89203215e-01
-7.29683042e-02 -6.08399212e-01 -9.30031598e-01 5.14517248e-01
2.77060360e-01 -3.31058264e-01 -4.19200510e-01 2.37213522e-01]
|
[13.432961463928223, 1.1367790699005127]
|
3cd78eb1-b36f-46cf-af1e-575778559832
|
heterogeneous-graph-transformer
|
2003.01332
| null |
https://arxiv.org/abs/2003.01332v1
|
https://arxiv.org/pdf/2003.01332v1.pdf
|
Heterogeneous Graph Transformer
|
Recent years have witnessed the emerging success of graph neural networks (GNNs) for modeling structured data. However, most GNNs are designed for homogeneous graphs, in which all nodes and edges belong to the same types, making them infeasible to represent heterogeneous structures. In this paper, we present the Heterogeneous Graph Transformer (HGT) architecture for modeling Web-scale heterogeneous graphs. To model heterogeneity, we design node- and edge-type dependent parameters to characterize the heterogeneous attention over each edge, empowering HGT to maintain dedicated representations for different types of nodes and edges. To handle dynamic heterogeneous graphs, we introduce the relative temporal encoding technique into HGT, which is able to capture the dynamic structural dependency with arbitrary durations. To handle Web-scale graph data, we design the heterogeneous mini-batch graph sampling algorithm---HGSampling---for efficient and scalable training. Extensive experiments on the Open Academic Graph of 179 million nodes and 2 billion edges show that the proposed HGT model consistently outperforms all the state-of-the-art GNN baselines by 9%--21% on various downstream tasks.
|
['Ziniu Hu', 'Kuansan Wang', 'Yuxiao Dong', 'Yizhou Sun']
|
2020-03-03
| null | null | null | null |
['graph-sampling']
|
['graphs']
|
[-5.59644103e-02 4.96994495e-01 -4.51667517e-01 -1.81123689e-01
-2.94801027e-01 -5.96461177e-01 3.63867372e-01 2.04017296e-01
9.64530036e-02 4.49410468e-01 2.29200765e-01 -5.86918950e-01
2.10876793e-01 -1.25465608e+00 -8.35702360e-01 -3.74602675e-01
-6.04568362e-01 5.37304640e-01 4.01228100e-01 -1.95329249e-01
-3.78518343e-01 1.57773823e-01 -1.01119494e+00 3.23677152e-01
7.29087353e-01 8.21765065e-01 4.07372117e-01 6.43106401e-01
-3.54110479e-01 1.01289177e+00 -4.36214119e-01 -6.73956990e-01
2.79748052e-01 -2.66156346e-01 -6.72889113e-01 5.72007596e-02
4.98335779e-01 -2.34953776e-01 -9.78216112e-01 8.37176800e-01
5.93320847e-01 -6.60721073e-03 4.70712572e-01 -1.29156625e+00
-1.02919674e+00 1.43957710e+00 -5.91556787e-01 2.77879775e-01
1.31587625e-01 6.73396438e-02 1.53006518e+00 -3.22221071e-01
1.00871575e+00 1.34082019e+00 8.36557686e-01 5.69204688e-01
-1.14223933e+00 -4.25702125e-01 1.00387454e+00 1.85006604e-01
-1.30727828e+00 -1.34120345e-01 9.24073696e-01 -6.43373728e-02
1.47742343e+00 3.96369323e-02 7.87047505e-01 1.57864082e+00
2.70385772e-01 8.55758131e-01 4.04881388e-01 -1.70365218e-02
-3.14807706e-02 -4.91344601e-01 3.35019737e-01 9.40931022e-01
5.65284908e-01 -2.76941448e-01 -2.26525411e-01 -3.74524519e-02
7.07559347e-01 -8.91999155e-02 -7.64974505e-02 -3.78557742e-01
-7.94394732e-01 7.48265743e-01 9.29039955e-01 1.76831588e-01
-5.45757413e-01 5.88608205e-01 7.71504164e-01 4.32064205e-01
6.17459178e-01 9.96481776e-02 -5.17632425e-01 2.13369191e-01
-5.31445861e-01 4.03690077e-02 9.78551149e-01 1.53583205e+00
4.82471973e-01 4.30497646e-01 -4.31553900e-01 7.19778776e-01
1.60116091e-01 7.65530393e-02 2.95601219e-01 -4.09124583e-01
9.32252288e-01 9.24669921e-01 -5.40001512e-01 -1.02880335e+00
-4.78947043e-01 -8.47864449e-01 -1.24447620e+00 -7.00158477e-01
6.54682294e-02 -2.28363127e-02 -1.34815907e+00 1.90514815e+00
2.10870922e-01 3.42096925e-01 -1.74521819e-01 4.03954923e-01
1.17415249e+00 7.20629275e-01 2.61272013e-01 1.57796949e-01
1.20619237e+00 -1.25560212e+00 -5.11329114e-01 -4.75378394e-01
8.21713448e-01 -1.64723992e-02 1.15041089e+00 -1.83831081e-02
-1.10966599e+00 -3.13461632e-01 -8.44818234e-01 -8.98075476e-02
-4.47630614e-01 -4.88114089e-01 7.67338634e-01 4.32892025e-01
-1.44401348e+00 6.49624825e-01 -8.58108819e-01 -4.44554359e-01
3.54549617e-01 2.49227375e-01 -3.58754128e-01 -2.71049440e-01
-1.32518482e+00 4.35689390e-01 4.43194300e-01 7.95584172e-02
-1.16910923e+00 -8.29911411e-01 -1.17800796e+00 5.77412903e-01
5.51155031e-01 -1.01568711e+00 1.07901287e+00 -7.30508924e-01
-1.00608850e+00 6.69904470e-01 -1.42296255e-01 -6.04486406e-01
1.95430383e-01 3.08346391e-01 -6.22322202e-01 7.50403255e-02
2.85907239e-02 3.53656858e-01 6.53075159e-01 -1.24841070e+00
-2.53016561e-01 -3.57424021e-01 2.48250946e-01 1.34442374e-01
-5.31873524e-01 -1.71159893e-01 -1.02308774e+00 -7.89606273e-01
-8.34591314e-02 -9.96725976e-01 -2.12284803e-01 -5.51516831e-01
-8.12376320e-01 -2.92841613e-01 7.80743837e-01 -6.76856279e-01
1.73739481e+00 -1.95076871e+00 3.19846511e-01 9.92082730e-02
8.52207839e-01 1.25980109e-01 -6.01475954e-01 6.43045604e-01
-9.44660380e-02 2.96476722e-01 1.24454036e-01 -5.82156837e-01
1.98358208e-01 3.93037319e-01 3.43226679e-02 6.58911839e-02
3.17453444e-02 1.47137582e+00 -9.17089820e-01 -3.95287544e-01
-2.30533645e-01 4.22122031e-01 -5.93778610e-01 2.19205946e-01
-5.04849732e-01 -8.50797817e-02 -5.24608076e-01 7.20992327e-01
5.62585890e-01 -8.99611294e-01 7.47229815e-01 -2.05671668e-01
4.98386830e-01 4.73828435e-01 -8.25893998e-01 1.61212730e+00
-5.06992996e-01 2.94010520e-01 2.37517565e-01 -7.71689892e-01
7.42931783e-01 6.97369426e-02 2.91234583e-01 -8.07966948e-01
7.43205845e-02 -3.03390814e-04 2.38059655e-01 -1.44431561e-01
5.13097227e-01 4.00739491e-01 -1.45123109e-01 2.68176675e-01
2.38418147e-01 2.01126501e-01 6.41378760e-01 7.34852493e-01
1.79054296e+00 -6.00554422e-02 -5.17760031e-02 -2.48095930e-01
-9.22459643e-03 -4.93984789e-01 5.88278651e-01 8.32284212e-01
-1.28978193e-01 3.90924424e-01 8.48842144e-01 -6.70024514e-01
-1.06670141e+00 -9.30065811e-01 5.66058040e-01 1.26462841e+00
-5.63849835e-03 -9.04913962e-01 -6.02820635e-01 -9.74316895e-01
2.26735443e-01 5.25876164e-01 -7.24450588e-01 -3.75121742e-01
-8.11839640e-01 -7.60210335e-01 3.47598106e-01 8.05433810e-01
3.89819384e-01 -1.14422929e+00 1.43041223e-01 4.75892454e-01
-3.17044277e-03 -1.31600749e+00 -8.12517345e-01 2.34150171e-01
-7.44393885e-01 -1.00575387e+00 -4.98279750e-01 -1.11346471e+00
6.05879247e-01 3.87832850e-01 1.79270506e+00 3.12520325e-01
2.25133318e-02 2.42931336e-01 -4.88147557e-01 1.33761510e-01
-2.91961461e-01 8.53410959e-01 -5.56471407e-01 -2.09105566e-01
-1.78437661e-02 -1.05207217e+00 -4.48996335e-01 -1.03120252e-01
-8.31661642e-01 1.72344759e-01 5.73523343e-01 7.18823552e-01
4.43958640e-01 1.55878365e-01 3.75817448e-01 -1.62059116e+00
7.81579614e-01 -7.76033521e-01 -3.08172196e-01 4.11235154e-01
-8.25995326e-01 1.99887037e-01 1.06710565e+00 -5.77656090e-01
-6.61029756e-01 -4.90419626e-01 9.85504128e-03 -5.92365086e-01
3.56879652e-01 1.01996028e+00 -4.02346849e-01 1.74209066e-02
4.02280420e-01 2.40324646e-01 -3.67665768e-01 -2.87725747e-01
4.38933134e-01 1.08365774e-01 1.46837413e-01 -7.22139955e-01
6.80542111e-01 4.73606288e-02 8.40968862e-02 -4.67857957e-01
-6.60575688e-01 -1.30002305e-01 -5.03101945e-02 2.06947960e-02
5.47203362e-01 -1.10673356e+00 -3.74688894e-01 7.01515794e-01
-9.81964409e-01 -1.00919735e+00 -1.00277945e-01 -1.82790846e-01
-8.78436863e-02 3.52283388e-01 -1.37155616e+00 -3.32384169e-01
-6.25968754e-01 -1.20623505e+00 1.26979482e+00 -1.09554175e-03
1.99195653e-01 -1.29029763e+00 -2.09499493e-01 2.67887325e-03
7.30958045e-01 2.36373618e-01 1.39206743e+00 -7.33415067e-01
-8.08099270e-01 -1.21448383e-01 -4.06479836e-01 8.74536037e-02
2.85650138e-03 2.04801355e-02 -4.27077085e-01 -6.80373073e-01
-7.98279881e-01 -1.48182526e-01 8.74949217e-01 3.78358364e-01
1.45894873e+00 -5.04502237e-01 -5.93773484e-01 9.62646365e-01
1.44511223e+00 -9.01302546e-02 5.12308061e-01 1.42480209e-01
1.31898308e+00 2.07948610e-01 -2.16316521e-01 2.26136446e-01
9.63667274e-01 4.28772748e-01 8.18071961e-01 -5.05501330e-02
-4.53265637e-01 -6.38668239e-01 2.03203201e-01 1.11860979e+00
1.28598586e-01 -1.10269296e+00 -8.33557069e-01 6.89564526e-01
-1.82452154e+00 -7.64269650e-01 -2.13170111e-01 1.75857532e+00
4.50902194e-01 4.47961956e-01 1.67211100e-01 -4.36219335e-01
1.03353870e+00 6.85706735e-01 -7.14135110e-01 -1.85523987e-01
-1.40241012e-01 1.51861280e-01 6.94840908e-01 4.08922791e-01
-8.04113626e-01 1.11940861e+00 5.90102100e+00 6.68237627e-01
-8.70176494e-01 2.94441264e-02 6.46703005e-01 8.31765607e-02
-8.07680249e-01 9.21890959e-02 -6.86965227e-01 6.38814807e-01
1.22594881e+00 -4.57142591e-01 6.82764769e-01 8.57956886e-01
-1.41492322e-01 6.89918697e-01 -8.62627685e-01 6.18457079e-01
-1.48077846e-01 -1.35725331e+00 5.16306341e-01 1.05582640e-01
7.69828200e-01 4.21189696e-01 -7.44175836e-02 7.93245852e-01
8.84546518e-01 -7.61314273e-01 6.05215490e-01 3.02214418e-02
6.92449808e-01 -5.57591915e-01 4.36316520e-01 7.99244046e-02
-1.92495155e+00 -1.39936537e-01 -3.46539795e-01 3.55413854e-01
1.71006426e-01 5.29090941e-01 -6.25444233e-01 9.77367103e-01
6.39035523e-01 9.91631448e-01 -7.17150450e-01 6.67524397e-01
-1.89859375e-01 7.06946194e-01 -2.39719361e-01 1.25615954e-01
3.46570998e-01 -2.26180315e-01 3.48483533e-01 1.07409167e+00
4.36878234e-01 -3.08644056e-01 3.32610995e-01 7.72639036e-01
-7.00612605e-01 -4.22088020e-02 -7.26165116e-01 -4.69085217e-01
8.51595700e-01 1.25729072e+00 -6.96590602e-01 -3.26515049e-01
-6.93261743e-01 9.98368263e-01 9.72190917e-01 5.72241724e-01
-9.48018968e-01 -4.02947217e-01 5.33890247e-01 3.44701439e-01
5.46719313e-01 -2.13732898e-01 3.10500950e-01 -1.37769127e+00
1.30014256e-01 -9.86157537e-01 8.05736423e-01 -6.96210086e-01
-1.49122596e+00 1.08922505e+00 -1.06475823e-01 -5.03118873e-01
-7.87894949e-02 -4.91281629e-01 -7.13300467e-01 7.34607875e-01
-1.37307358e+00 -1.69605601e+00 -5.01957417e-01 6.42496407e-01
2.64353424e-01 -9.59902711e-04 5.16772330e-01 4.84901220e-01
-8.14924479e-01 7.84994125e-01 -2.60004610e-01 4.22670335e-01
3.41422558e-01 -1.44835138e+00 1.33660150e+00 1.02866733e+00
7.81508833e-02 6.00509346e-01 5.98491728e-01 -9.64917660e-01
-1.71710730e+00 -1.57309663e+00 8.42010677e-01 -2.56227590e-02
9.58857417e-01 -7.64036417e-01 -1.02568853e+00 1.26660037e+00
2.16694623e-01 5.85710704e-01 2.90651441e-01 3.98191005e-01
-7.99317241e-01 -2.48824835e-01 -9.16926444e-01 8.40323925e-01
1.85247970e+00 -5.43386757e-01 1.07194632e-01 4.42682445e-01
1.39428520e+00 -4.88355666e-01 -1.12720954e+00 4.94342625e-01
1.67789087e-01 -6.71175957e-01 7.97734976e-01 -8.76120806e-01
2.33195901e-01 6.53856471e-02 -9.13866758e-02 -1.58619690e+00
-8.63989055e-01 -7.77950883e-01 -7.54695773e-01 1.19611979e+00
6.21793628e-01 -7.91415989e-01 1.15859342e+00 3.54330719e-01
-5.65022230e-01 -8.11741650e-01 -7.22889304e-01 -7.63827920e-01
-4.85104918e-02 -1.53389378e-02 9.83209908e-01 8.81765723e-01
-2.40111426e-01 6.53686404e-01 -2.93869883e-01 2.39907220e-01
5.64174950e-01 2.20149457e-01 5.98516703e-01 -1.18586612e+00
-5.03445625e-01 -5.59807956e-01 -3.94851536e-01 -1.11525142e+00
5.78954518e-01 -1.24684632e+00 -4.19694453e-01 -1.99261761e+00
2.63329118e-01 -2.62162179e-01 -2.74629712e-01 6.29049361e-01
-3.40415627e-01 -2.38184765e-01 5.18129915e-02 -2.27320135e-01
-7.60149419e-01 6.00006461e-01 1.41676021e+00 -4.80783910e-01
-5.29110320e-02 -2.09366918e-01 -9.01754558e-01 1.49376420e-02
6.36634827e-01 -2.04856917e-01 -9.55683768e-01 -8.21199119e-01
4.30173665e-01 4.85869721e-02 3.12346946e-02 -6.68240190e-01
3.68530080e-02 4.20015268e-02 -1.52676105e-01 -2.46654615e-01
-1.05075009e-01 -7.22009301e-01 4.73347157e-01 3.05329174e-01
-3.06483954e-01 5.59037685e-01 3.09675708e-02 8.28981638e-01
-1.12281866e-01 2.69636810e-01 5.00112414e-01 -3.95473033e-01
-7.18303263e-01 1.11812246e+00 5.24881408e-02 3.94424170e-01
9.66623247e-01 1.62847251e-01 -8.81858110e-01 -5.83776534e-01
-7.24089384e-01 4.32908326e-01 5.07289767e-01 5.68546236e-01
3.61659497e-01 -1.25422561e+00 -7.27948606e-01 1.76592991e-01
2.76070565e-01 2.68445462e-01 4.79250938e-01 5.51069140e-01
-5.13697147e-01 8.53866339e-02 1.82329163e-01 -1.32204086e-01
-8.87550533e-01 1.01814818e+00 4.66461718e-01 -9.06534493e-01
-9.55900729e-01 8.22988033e-01 1.93255574e-01 -3.86585474e-01
2.40261167e-01 -4.53019470e-01 1.33355975e-01 -2.56910414e-01
-2.80545801e-02 9.06560421e-02 3.26287329e-01 -3.75033021e-01
-8.94089565e-02 -1.06181055e-01 -3.99186134e-01 6.55818880e-01
1.36654127e+00 -3.02290916e-02 -2.04617292e-01 2.00273156e-01
1.16472256e+00 -2.81334579e-01 -1.08008730e+00 -3.20598125e-01
-5.81005998e-02 -2.48792972e-02 -6.10278323e-02 -6.05981231e-01
-1.55721807e+00 4.63417917e-01 -2.22800210e-01 7.77774334e-01
1.01990891e+00 9.21136960e-02 1.01247752e+00 1.60499841e-01
7.03696728e-01 -6.33538008e-01 -1.86129406e-01 6.10744178e-01
6.53604031e-01 -8.60548496e-01 -3.09761256e-01 -6.44232333e-01
-4.26311135e-01 7.11271048e-01 9.80898976e-01 -2.02286199e-01
6.61592722e-01 3.37087512e-01 -5.12290418e-01 -3.30074400e-01
-1.28069675e+00 -1.11557379e-01 1.77748784e-01 6.22408569e-01
4.51727152e-01 3.14788610e-01 1.78603120e-02 4.98727500e-01
9.66945067e-02 -4.17812437e-01 4.31042552e-01 6.77141190e-01
-5.46714291e-03 -1.43618608e+00 5.26297152e-01 8.43946338e-01
-3.09797406e-01 -4.49451894e-01 -4.00212198e-01 9.95549858e-01
-4.73217160e-01 6.70977056e-01 -1.38383526e-02 -5.87284803e-01
3.60632300e-01 -1.35223165e-01 5.84337413e-01 -7.57615507e-01
-7.49226511e-01 -1.10078931e-01 4.94909316e-01 -7.67940879e-01
1.06345005e-01 -2.82122821e-01 -1.05970216e+00 -7.05502510e-01
-1.39048845e-01 1.22974277e-01 1.86407492e-01 3.56225222e-01
8.04178953e-01 1.03139496e+00 3.73152256e-01 -7.66291797e-01
-5.37451625e-01 -1.16436756e+00 -9.90586281e-01 5.80081820e-01
1.26507998e-01 -3.95046294e-01 -4.37152952e-01 -5.23414314e-01]
|
[6.989778995513916, 6.274105072021484]
|
f1e45a7e-a061-452b-ae14-c124c270d174
|
reproducible-evaluation-of-classification
|
1808.06452
| null |
http://arxiv.org/abs/1808.06452v1
|
http://arxiv.org/pdf/1808.06452v1.pdf
|
Reproducible evaluation of classification methods in Alzheimer's disease: framework and application to MRI and PET data
|
A large number of papers have introduced novel machine learning and feature
extraction methods for automatic classification of AD. However, they are
difficult to reproduce because key components of the validation are often not
readily available. These components include selected participants and input
data, image preprocessing and cross-validation procedures. The performance of
the different approaches is also difficult to compare objectively. In
particular, it is often difficult to assess which part of the method provides a
real improvement, if any. We propose a framework for reproducible and objective
classification experiments in AD using three publicly available datasets (ADNI,
AIBL and OASIS). The framework comprises: i) automatic conversion of the three
datasets into BIDS format, ii) a modular set of preprocessing pipelines,
feature extraction and classification methods, together with an evaluation
framework, that provide a baseline for benchmarking the different components.
We demonstrate the use of the framework for a large-scale evaluation on 1960
participants using T1 MRI and FDG PET data. In this evaluation, we assess the
influence of different modalities, preprocessing, feature types, classifiers,
training set sizes and datasets. Performances were in line with the
state-of-the-art. FDG PET outperformed T1 MRI for all classification tasks. No
difference in performance was found for the use of different atlases, image
smoothing, partial volume correction of FDG PET images, or feature type. Linear
SVM and L2-logistic regression resulted in similar performance and both
outperformed random forests. The classification performance increased along
with the number of subjects used for training. Classifiers trained on ADNI
generalized well to AIBL and OASIS. All the code of the framework and the
experiments is publicly available at:
https://gitlab.icm-institute.org/aramislab/AD-ML.
|
['the Australian Imaging Biomarkers', "for the Alzheimer's Disease Neuroimaging Initiative", 'Marie-Odile Habert', 'Junhao Wen', 'Jérémy Guillon', 'Alexandre Routier', 'Simona Bottani', 'Jorge Samper-González', 'Pascal Lu', 'Lifestyle flagship study of ageing', 'Sabrina Fontanella', 'Stanley Durrleman', 'Olivier Colliot', 'Ninon Burgos', 'Michael Bacci', 'Arnaud Marcoux', 'Anne Bertrand', 'Theodoros Evgeniou', 'Hugo Bertin']
|
2018-08-20
| null | null | null | null |
['image-smoothing']
|
['computer-vision']
|
[ 6.29805177e-02 -2.64476240e-01 -2.25636140e-01 -6.47441506e-01
-1.04140747e+00 -5.47150195e-01 7.59861469e-01 3.72343600e-01
-8.52334976e-01 9.31113839e-01 -1.86270792e-02 -9.93866101e-02
-1.96988285e-01 -4.98461813e-01 -3.89910638e-01 -6.15708351e-01
-2.25288898e-01 8.60534728e-01 4.79669958e-01 2.11852744e-01
-7.52389729e-02 5.42697072e-01 -1.38881981e+00 6.15716577e-01
5.86897254e-01 1.12945557e+00 2.54186392e-01 4.04582143e-01
4.14822847e-02 3.42107296e-01 -3.45721632e-01 -1.79571927e-01
1.85379088e-01 -2.94999927e-01 -9.11222160e-01 -5.46749793e-02
5.22110522e-01 -3.84911805e-01 -3.25877778e-02 5.82054675e-01
7.98827529e-01 -2.37293780e-01 7.65884936e-01 -1.20298612e+00
-1.56880021e-01 3.88280272e-01 -2.74723887e-01 5.33976614e-01
1.50615275e-01 3.35163802e-01 5.74990034e-01 -7.78880894e-01
7.30046034e-01 1.00524414e+00 9.64812219e-01 4.28658426e-01
-1.46487582e+00 -4.88681138e-01 -1.98451906e-01 4.90187883e-01
-1.20078313e+00 -5.21886706e-01 6.37986735e-02 -7.00528145e-01
9.19103205e-01 4.25580680e-01 9.42536414e-01 1.25256991e+00
4.55990173e-02 3.63832355e-01 1.86125755e+00 -4.61637706e-01
4.00877029e-01 5.55778623e-01 6.28983259e-01 4.70317662e-01
3.69049758e-01 1.31837621e-01 1.03356987e-01 -5.67914128e-01
5.47767520e-01 -1.40847355e-01 -1.22173637e-01 -3.80153298e-01
-1.26011634e+00 8.82313967e-01 5.60047925e-01 5.21274865e-01
-4.03609216e-01 -1.20853722e-01 6.62343264e-01 1.66204810e-01
4.17597473e-01 6.60484582e-02 -5.13136804e-01 1.86044306e-01
-1.16524243e+00 3.77742112e-01 4.84837621e-01 3.11918110e-01
1.38590768e-01 -4.02024508e-01 -3.17751437e-01 1.15889478e+00
3.68314177e-01 3.96807790e-01 9.18895960e-01 -7.85651684e-01
2.15009257e-01 6.21245325e-01 -2.05389097e-01 -4.86931115e-01
-1.01395333e+00 -4.26017463e-01 -6.26335084e-01 4.43308562e-01
7.46192455e-01 -4.37015630e-02 -1.11712635e+00 1.48366284e+00
1.66454375e-01 -3.74308586e-01 -4.01159286e-01 1.01019812e+00
7.17557013e-01 -3.91298644e-02 4.65049744e-01 -9.19420496e-02
1.63667202e+00 -7.62568653e-01 -4.23998564e-01 -1.90759584e-01
7.66138852e-01 -4.22808766e-01 1.06908548e+00 3.88777882e-01
-1.08251846e+00 -4.67099786e-01 -9.17526186e-01 7.33034983e-02
-5.62772036e-01 4.89605904e-01 6.78633690e-01 9.80158806e-01
-1.04927802e+00 5.38444340e-01 -1.00877893e+00 -6.83068395e-01
7.47584999e-01 5.66358387e-01 -6.97175443e-01 -1.53238639e-01
-1.00674748e+00 1.30756593e+00 3.89521211e-01 -6.38905540e-02
-7.36926258e-01 -7.84685373e-01 -4.52871948e-01 -4.13813680e-01
-1.34011999e-01 -7.76723146e-01 1.27023518e+00 -1.22401655e+00
-9.93064344e-01 1.19947445e+00 1.25695959e-01 -6.13261402e-01
1.04092610e+00 4.64994274e-02 -3.65939260e-01 2.02877536e-01
3.25526536e-01 9.19801354e-01 5.34927309e-01 -7.63737857e-01
-4.33329672e-01 -8.58886063e-01 -4.02370006e-01 -1.58466130e-01
1.51199058e-01 2.99549937e-01 3.44981290e-02 -4.58345026e-01
-4.75533232e-02 -1.10776687e+00 -2.13642582e-01 1.07217498e-01
-1.51888207e-01 3.25242169e-02 3.82971406e-01 -8.50837469e-01
7.44655848e-01 -1.88892674e+00 -2.50440001e-01 1.70443401e-01
3.67977589e-01 1.32429138e-01 2.27279231e-01 -6.95453361e-02
-4.83241320e-01 2.39918411e-01 -2.75138855e-01 -3.64311896e-02
-4.91472483e-02 -8.36698562e-02 2.74114698e-01 7.71362484e-01
1.38170704e-01 8.04022908e-01 -4.41247314e-01 -3.96381140e-01
4.12060529e-01 5.84839404e-01 -4.10764724e-01 -2.43475333e-01
1.63063258e-01 5.42215109e-01 -3.10737222e-01 5.19119442e-01
4.85639870e-01 -2.23634154e-01 2.16433898e-01 -3.20511937e-01
-2.17828667e-03 2.07077667e-01 -9.34237182e-01 1.40930891e+00
-6.87705502e-02 3.46449435e-01 -1.04130417e-01 -9.27155912e-01
5.52291512e-01 4.50283885e-01 5.66903591e-01 -7.94848740e-01
3.84821296e-01 6.05591714e-01 3.37371409e-01 -3.67006093e-01
-3.00796181e-01 -1.62895799e-01 1.84101462e-01 2.90182114e-01
2.13089868e-01 2.64495820e-01 4.60963041e-01 -1.48534209e-01
1.10390985e+00 1.75106023e-02 5.23102462e-01 -3.97005260e-01
5.68196535e-01 2.81572074e-01 1.83633730e-01 5.05272567e-01
-4.64471072e-01 7.75873423e-01 5.48431933e-01 -3.78252506e-01
-1.04564416e+00 -1.04967618e+00 -7.71225393e-01 7.83420324e-01
-5.43637991e-01 -3.98935974e-01 -1.00062633e+00 -8.49709034e-01
-1.75787397e-02 6.32884681e-01 -8.78229320e-01 1.33453846e-01
-3.35982591e-01 -1.31676483e+00 5.96412420e-01 5.47493219e-01
5.23897290e-01 -8.65585089e-01 -6.16679788e-01 1.25029638e-01
-2.51634531e-02 -1.03766811e+00 1.38068041e-02 2.26302594e-01
-1.24925923e+00 -1.32850897e+00 -9.82867181e-01 -4.61376667e-01
5.61131716e-01 -1.25103310e-01 9.83895421e-01 -2.57577207e-02
-5.96986890e-01 3.79200995e-01 -2.82779276e-01 -4.15768087e-01
-3.25025260e-01 1.91512182e-01 -6.87612081e-03 -3.35344791e-01
2.33832762e-01 -4.90435123e-01 -7.21329331e-01 4.54490960e-01
-6.24458075e-01 -1.68859094e-01 8.81370425e-01 6.68883502e-01
8.64549339e-01 -3.21270317e-01 4.82302606e-01 -7.78434515e-01
4.49583888e-01 -5.08771181e-01 -3.29244554e-01 1.58505902e-01
-8.47100258e-01 -1.92461520e-01 1.86565116e-01 -2.98688889e-01
-5.59067309e-01 1.52780727e-01 -3.03119451e-01 1.48375630e-01
-6.26286805e-01 3.16529959e-01 -1.72975034e-01 -1.78273432e-02
9.48745787e-01 -1.76499844e-01 1.86542705e-01 -7.16705322e-01
6.91043064e-02 6.75069511e-01 1.18541248e-01 -3.60455513e-01
2.78824061e-01 4.08334047e-01 -1.14387609e-01 -6.59493148e-01
-4.30433661e-01 -2.92974323e-01 -1.11741555e+00 -2.66045462e-02
9.79959726e-01 -6.77247047e-01 1.52760698e-02 5.50203145e-01
-7.20988512e-01 -5.67751348e-01 -3.98066908e-01 7.56843805e-01
-4.02076632e-01 4.73339520e-02 -2.38943174e-01 -2.90532559e-01
-6.08552933e-01 -1.53828609e+00 7.47363985e-01 -8.37584659e-02
-4.20566142e-01 -8.39683771e-01 1.84270978e-01 5.22670150e-01
5.07706106e-01 3.26961040e-01 9.04776454e-01 -1.02268040e+00
-5.48963919e-02 -3.36145490e-01 -3.67071152e-01 3.13556969e-01
5.55793457e-02 -1.78724751e-01 -1.02864647e+00 -1.39864922e-01
-2.06512108e-01 -3.68693888e-01 8.03800106e-01 5.78530312e-01
1.04539132e+00 -5.62343262e-02 -4.52747792e-01 2.58495897e-01
1.12129104e+00 1.41956702e-01 7.13784933e-01 8.23668599e-01
1.84371218e-01 5.48368573e-01 2.93355018e-01 -9.57751647e-03
2.85080045e-01 1.02964246e+00 1.38141125e-01 -1.60105437e-01
-3.74756277e-01 4.17052507e-01 4.50359523e-01 2.11862743e-01
-3.46212685e-01 3.09438139e-01 -1.14025962e+00 2.93863297e-01
-1.34587598e+00 -7.35404611e-01 -5.43973744e-01 2.20719814e+00
5.43601334e-01 1.36729047e-01 7.79358923e-01 3.73584419e-01
4.80109066e-01 -3.24560255e-01 -2.82023638e-01 -1.72605872e-01
-1.39586655e-02 4.07482028e-01 5.46336532e-01 1.62957460e-01
-1.10500896e+00 3.44218880e-01 6.46966267e+00 4.85491544e-01
-1.29365492e+00 7.87623286e-01 8.54843318e-01 -2.49723166e-01
3.06107968e-01 -2.69408315e-01 -6.65339947e-01 5.41918874e-01
1.36756063e+00 1.19481429e-01 2.33768106e-01 8.02927196e-01
3.23271036e-01 -3.26096654e-01 -1.00281560e+00 6.91924870e-01
-9.44086611e-02 -1.01929438e+00 -3.36641699e-01 1.51361242e-01
1.90634742e-01 6.43514216e-01 -2.74149209e-01 2.36084029e-01
-1.91869363e-01 -1.05658996e+00 8.37535679e-01 4.05046254e-01
9.12983716e-01 -1.45455241e-01 8.37955654e-01 -1.06781423e-01
-7.09725499e-01 -1.26049398e-02 -4.24721986e-02 1.93152621e-01
-3.98829430e-02 5.86360633e-01 -8.73963237e-01 5.17038763e-01
8.96606445e-01 2.62989104e-01 -1.19292355e+00 1.36184859e+00
-1.11017473e-01 6.76005960e-01 -4.07735646e-01 2.65477747e-01
8.24518427e-02 1.39035583e-02 2.86008984e-01 1.35682321e+00
2.24754497e-01 -8.77694115e-02 1.03500135e-01 7.70507276e-01
3.00024837e-01 4.53044862e-01 -4.83973473e-02 -5.31324148e-02
-5.61868586e-02 1.58973670e+00 -1.11859596e+00 -2.28467971e-01
-5.85256159e-01 6.84545696e-01 4.66402061e-02 -3.33307199e-02
-9.51868653e-01 3.63506824e-02 2.80225158e-01 8.01400065e-01
1.05206687e-02 -6.12588637e-02 -3.02286834e-01 -8.77566397e-01
-1.17186410e-02 -8.45137537e-01 7.56026030e-01 -8.57294917e-01
-1.29512823e+00 7.55344570e-01 3.32534909e-01 -8.62599194e-01
-9.98696834e-02 -8.65348697e-01 -3.23838979e-01 8.34753454e-01
-1.17316473e+00 -1.07997870e+00 -6.18665040e-01 5.85786104e-01
9.03427824e-02 -1.90531448e-01 1.02220643e+00 6.73279941e-01
-6.04752600e-01 3.69332552e-01 -5.81440963e-02 9.96152535e-02
9.22550678e-01 -1.19515800e+00 -3.20533814e-04 2.64170974e-01
-3.84001136e-02 3.84404570e-01 1.60364866e-01 -6.76612020e-01
-7.35263765e-01 -1.06997561e+00 5.42251825e-01 -4.56998914e-01
7.44669676e-01 -2.87493110e-01 -7.23922491e-01 7.90790260e-01
1.36871974e-03 1.92444786e-01 7.87167847e-01 -5.90355843e-02
-1.28235638e-01 -1.79699004e-01 -1.50389802e+00 1.18022151e-01
8.22679579e-01 -1.19990043e-01 -6.44738436e-01 6.04213357e-01
-5.43510308e-03 -2.85693914e-01 -1.08023214e+00 3.48168731e-01
8.32705379e-01 -1.05593860e+00 1.04237258e+00 -5.57150304e-01
1.44250125e-01 -5.61621152e-02 -1.81780577e-01 -1.09024501e+00
-3.56990457e-01 2.04901412e-01 3.39971244e-01 1.11155975e+00
6.32694006e-01 -9.87646520e-01 4.71673459e-01 7.07028508e-01
-1.39842648e-02 -6.80903256e-01 -1.03009319e+00 -6.32884324e-01
1.86435506e-01 -6.31956577e-01 2.32386857e-01 7.68997014e-01
-3.56529027e-01 1.89558789e-01 3.26197624e-01 -1.54685810e-01
4.88627762e-01 -3.92121375e-01 3.83737922e-01 -1.42764866e+00
-1.85056150e-01 -5.38532555e-01 -7.86367893e-01 2.90872812e-01
-1.03915378e-01 -1.35365748e+00 -5.40903687e-01 -1.74867797e+00
4.30065989e-01 -7.32438326e-01 -3.49396855e-01 8.35603356e-01
6.34082779e-02 5.12384057e-01 1.17527954e-01 5.37827432e-01
-7.98977986e-02 1.61750037e-02 8.45553458e-01 -1.02381438e-01
-3.13778698e-01 6.54517785e-02 -5.25017023e-01 7.69475102e-01
1.11467028e+00 -6.36188865e-01 -4.21546042e-01 -1.46106854e-01
-3.57440621e-01 -4.86531854e-01 1.05800927e+00 -1.36182475e+00
-3.07312340e-01 2.84115613e-01 1.03297007e+00 -3.00675809e-01
2.37334400e-01 -8.38013113e-01 5.58821797e-01 6.92507029e-01
-1.70973361e-01 1.57996505e-01 3.30673695e-01 2.29987986e-02
2.22098842e-01 -3.62945139e-01 1.04463065e+00 -3.18652868e-01
-4.58142072e-01 1.44158915e-01 -3.64291966e-01 -6.38218969e-03
1.06321788e+00 -2.12419868e-01 -3.65170747e-01 8.16478431e-02
-1.23733687e+00 -2.06056729e-01 2.90035427e-01 1.97241351e-01
6.90492615e-02 -1.27363920e+00 -8.31882238e-01 1.63830519e-02
3.77465822e-02 -4.83515322e-01 1.79216847e-01 1.69614255e+00
-4.58855003e-01 4.85607922e-01 -7.14689851e-01 -7.07317591e-01
-1.42532206e+00 2.86441624e-01 7.25812495e-01 -5.01509309e-01
-5.95809400e-01 2.91190714e-01 -1.89077184e-02 -3.86044890e-01
2.36646980e-02 -5.05317330e-01 -1.24951966e-01 4.92222309e-01
7.57600725e-01 5.60985446e-01 6.49342179e-01 -6.66925669e-01
-7.12832272e-01 2.11409479e-01 -2.13216007e-01 -3.64076763e-01
1.63886678e+00 5.56394048e-02 -9.04284120e-02 4.71566558e-01
1.02087867e+00 -3.91477346e-01 -6.47049487e-01 2.34481901e-01
1.75524592e-01 -1.63859889e-01 4.06296283e-01 -1.32203531e+00
-1.10085416e+00 7.40988553e-01 1.55584919e+00 6.34856820e-02
9.01070535e-01 1.12353534e-01 3.92438501e-01 -1.13135375e-01
3.84302497e-01 -8.44621956e-01 -3.98453921e-01 2.19192961e-03
1.18120265e+00 -1.12319136e+00 2.88780779e-01 -2.60216624e-01
-6.49044216e-01 1.10813165e+00 2.27727175e-01 -8.26240405e-02
6.47503376e-01 2.32475683e-01 7.96471313e-02 -2.58193702e-01
-2.63655275e-01 -2.00842574e-01 5.04892349e-01 7.88809657e-01
5.91316760e-01 6.16467670e-02 -7.93094218e-01 1.09265137e+00
-2.83696175e-01 4.40772504e-01 2.08547071e-01 7.57229507e-01
-1.44710705e-01 -1.27820134e+00 -4.59974051e-01 9.25133884e-01
-6.56533241e-01 1.45989060e-01 -6.21933818e-01 1.27682006e+00
3.97184819e-01 5.09904146e-01 -1.09504774e-01 -3.34768184e-02
4.89376426e-01 4.29844618e-01 7.68930435e-01 -5.01839936e-01
-8.38240445e-01 7.18044713e-02 4.31509823e-01 -5.56033313e-01
-5.61143935e-01 -1.21773827e+00 -1.18794429e+00 -1.25035316e-01
-1.19226336e-01 -1.35618895e-01 8.57215166e-01 9.14246261e-01
2.44674474e-01 5.47851503e-01 3.60968299e-02 -9.04494107e-01
-3.16782951e-01 -1.22435129e+00 -5.91084421e-01 1.89796284e-01
-1.16362564e-01 -9.93863225e-01 -2.00091735e-01 5.43662347e-02]
|
[14.22966194152832, -1.8495522737503052]
|
816a4200-8694-4d94-a312-74f9def6fa65
|
galaxy-classification-using-transfer-learning
|
2305.00002
| null |
https://arxiv.org/abs/2305.00002v1
|
https://arxiv.org/pdf/2305.00002v1.pdf
|
Galaxy Classification Using Transfer Learning and Ensemble of CNNs With Multiple Colour Spaces
|
Big data has become the norm in astronomy, making it an ideal domain for computer science research. Astronomers typically classify galaxies based on their morphologies, a practice that dates back to Hubble (1936). With small datasets, classification could be performed by individuals or small teams, but the exponential growth of data from modern telescopes necessitates automated classification methods. In December 2013, Winton Capital, Galaxy Zoo, and the Kaggle team created the Galaxy Challenge, which tasked participants with developing models to classify galaxies. The Kaggle Galaxy Zoo dataset has since been widely used by researchers. This study investigates the impact of colour space transformation on classification accuracy and explores the effect of CNN architecture on this relationship. Multiple colour spaces (RGB, XYZ, LAB, etc.) and CNN architectures (VGG, ResNet, DenseNet, Xception, etc.) are considered, utilizing pre-trained models and weights. However, as most pre-trained models are designed for natural RGB images, we examine their performance with transformed, non-natural astronomical images. We test our hypothesis by evaluating individual networks with RGB and transformed colour spaces and examining various ensemble configurations. A minimal hyperparameter search ensures optimal results. Our findings indicate that using transformed colour spaces in individual networks yields higher validation accuracy, and ensembles of networks and colour spaces further improve accuracy. This research aims to validate the utility of colour space transformation for astronomical image classification and serve as a benchmark for future studies.
|
['Yevonnael Andrew']
|
2023-03-26
| null | null | null | null |
['astronomy']
|
['miscellaneous']
|
[-2.26788253e-01 -3.93306971e-01 4.90578771e-01 -5.35363816e-02
2.36714259e-01 -8.15220177e-01 1.13572991e+00 -3.06376278e-01
-8.00389469e-01 4.13427234e-01 -5.01541942e-02 -7.39776552e-01
-8.44605193e-02 -1.06028640e+00 -3.46758395e-01 -7.14667797e-01
1.83960181e-02 2.75435746e-01 2.39072993e-01 -1.65796936e-01
3.66192490e-01 7.18451142e-01 -1.64298749e+00 1.42262116e-01
5.65159798e-01 1.03889763e+00 1.01030923e-01 1.04556453e+00
-7.64625221e-02 7.50628710e-01 -8.02546978e-01 -4.10400361e-01
7.35599101e-01 -5.69079101e-01 -8.85889232e-01 -9.45766717e-02
4.71572220e-01 3.34961742e-01 -9.22792628e-02 8.01131487e-01
4.93089676e-01 4.89314109e-01 5.20201981e-01 -1.14797080e+00
-7.46965468e-01 2.05461644e-02 3.67727242e-02 3.26416373e-01
-1.86729088e-01 4.47194308e-01 9.45437014e-01 -5.06484985e-01
4.48310673e-01 1.11379254e+00 7.89272010e-01 2.75246799e-01
-1.14736354e+00 -5.93916714e-01 -3.52204919e-01 5.54054797e-01
-1.27213681e+00 -7.27357492e-02 3.96215737e-01 -5.17275751e-01
1.22307277e+00 4.38834608e-01 1.06143081e+00 7.63607919e-01
2.41724849e-01 -1.65095642e-01 1.59389639e+00 -6.03694797e-01
2.95270324e-01 2.00525567e-01 -2.33658522e-01 3.48303944e-01
3.56639087e-01 3.44182700e-01 -1.16208844e-01 -5.86116016e-02
1.06755006e+00 -4.33648564e-02 -1.69412211e-01 -2.15550691e-01
-1.30395281e+00 8.95784795e-01 8.23655665e-01 5.70180237e-01
-1.02598600e-01 6.52914047e-02 1.06632248e-01 7.54730642e-01
5.75502932e-01 1.22814190e+00 -3.88771176e-01 -1.77078173e-01
-6.08171821e-01 1.15086548e-01 8.70658040e-01 4.20637071e-01
6.75141573e-01 9.47355479e-02 3.42768699e-01 1.03301001e+00
1.62361458e-01 7.66623691e-02 7.31421113e-01 -1.02924395e+00
-6.06646761e-02 9.06289756e-01 -9.12971273e-02 -9.51439559e-01
-6.65074050e-01 -6.94315970e-01 -1.04672718e+00 5.77008903e-01
8.91420066e-01 1.86033756e-01 -9.78776276e-01 1.13245046e+00
2.46046558e-01 2.20784582e-02 -1.26273762e-02 1.17825341e+00
6.87339544e-01 3.01392645e-01 -2.22567648e-01 4.26545084e-01
1.32136798e+00 -8.71853352e-01 1.98546514e-01 -3.00656348e-01
6.22652471e-01 -8.69010925e-01 1.13149452e+00 6.77936256e-01
-6.01625025e-01 -7.72904932e-01 -8.60945582e-01 -7.86040910e-03
-7.68182039e-01 -2.16301948e-01 8.64099205e-01 9.93323624e-01
-1.33796775e+00 9.00624216e-01 -6.83611274e-01 -5.82089543e-01
3.45261365e-01 3.68151456e-01 -5.79287469e-01 4.39860448e-02
-1.01787221e+00 1.16518366e+00 5.30561686e-01 1.82623014e-01
-4.98246491e-01 -3.21427852e-01 -2.84484476e-01 -4.01430242e-02
-1.05548941e-01 -6.26103878e-01 1.10774601e+00 -1.36746001e+00
-1.53015816e+00 9.31595385e-01 4.42336321e-01 -5.83249867e-01
1.91809073e-01 3.00487787e-01 -5.14764845e-01 3.88924181e-02
-4.77892190e-01 6.65810108e-01 7.18388796e-01 -9.35216010e-01
-7.28755951e-01 -1.85309634e-01 3.04213762e-01 -6.45985007e-02
-3.37456197e-01 1.03542514e-01 -2.18220323e-01 -4.02080774e-01
9.29488838e-02 -1.28470397e+00 -2.22537205e-01 -3.12352359e-01
1.21635638e-01 -3.61984432e-01 3.96772474e-01 -4.09081727e-01
6.80354178e-01 -2.11801100e+00 4.67752246e-03 4.27278221e-01
5.41747868e-01 2.23377094e-01 -2.47217286e-02 4.79893908e-02
-3.49429756e-01 3.35967630e-01 1.42703494e-02 6.43362924e-02
-9.81971920e-02 1.60044327e-01 2.95728207e-01 5.45885563e-01
6.70385826e-03 8.27552497e-01 -6.59860849e-01 -1.78075343e-01
4.56717104e-01 5.33755779e-01 -5.22728026e-01 4.84455936e-03
1.48903519e-01 5.12809515e-01 1.54154584e-01 3.78553629e-01
3.92285943e-01 -5.35866857e-01 -1.82819292e-01 -1.26154507e-02
-3.85341018e-01 1.50049523e-01 -8.27979088e-01 1.11472392e+00
-6.64240599e-01 1.09177673e+00 -1.18018679e-01 -9.66123700e-01
1.13656366e+00 1.94371894e-01 1.95656300e-01 -6.83104038e-01
2.60631144e-01 4.09425586e-01 7.92584896e-01 -5.22406064e-02
5.90183854e-01 -3.62196863e-01 2.64999956e-01 3.16768169e-01
1.73795238e-01 -5.69681227e-01 1.32267416e-01 -1.19112059e-01
1.20081341e+00 -7.89413452e-02 1.93502381e-02 -1.41436860e-01
2.35744238e-01 2.24472195e-01 2.18224913e-01 7.69417703e-01
-1.04555741e-01 8.66803944e-01 4.40677345e-01 -8.28775644e-01
-1.46396601e+00 -6.31197393e-01 -2.09120169e-01 1.08427489e+00
-2.76168406e-01 -1.78112268e-01 -3.80658329e-01 -3.78525078e-01
-2.15987936e-01 5.80427825e-01 -6.71758950e-01 -3.67617965e-01
-3.32720101e-01 -9.36590731e-01 5.05748987e-01 1.31892726e-01
7.77983189e-01 -1.38761210e+00 -8.17711174e-01 -1.48754671e-01
2.37140015e-01 -5.08509696e-01 2.91409016e-01 2.77839661e-01
-8.05855811e-01 -1.20814300e+00 -7.13428557e-01 -5.55539250e-01
5.89703977e-01 3.99250418e-01 1.29378295e+00 4.53467935e-01
-3.91798377e-01 2.72063315e-01 -6.68662369e-01 -3.97728711e-01
-3.09510738e-01 2.28040233e-01 -1.68061718e-01 -1.13508999e-01
3.31156582e-01 -4.12324816e-01 -7.00565100e-01 4.51860279e-01
-9.16964173e-01 1.67208210e-01 6.41446173e-01 7.91631579e-01
7.70252571e-02 3.46389264e-01 7.78164640e-02 -7.12949216e-01
4.95902181e-01 -3.32663566e-01 -6.54865444e-01 2.08632159e-03
-6.41167641e-01 -8.30120146e-02 8.82553041e-01 -3.21165949e-01
-8.86605859e-01 -2.69549936e-01 -1.60740495e-01 -3.44032615e-01
-3.69988531e-01 3.17255288e-01 3.48434359e-01 -8.07766318e-01
1.06185687e+00 -1.97168916e-01 8.21154118e-02 -4.33574319e-01
1.66899592e-01 8.03615272e-01 5.28948665e-01 -3.35482478e-01
9.76275861e-01 3.76727462e-01 5.91731444e-02 -8.12955976e-01
-5.27340949e-01 -5.66638529e-01 -8.16093266e-01 -4.10406172e-01
7.32348323e-01 -6.50444210e-01 -7.23266244e-01 5.45033753e-01
-7.74761140e-01 -4.21681374e-01 -2.35164225e-01 7.06976414e-01
-1.68206677e-01 1.71605672e-03 -2.55999327e-01 -5.17798722e-01
2.74761338e-02 -9.60856318e-01 4.85408962e-01 4.42417204e-01
-1.82462260e-01 -1.21754706e+00 1.26009360e-01 2.56299317e-01
8.18938076e-01 2.54251331e-01 6.89136684e-01 -6.37604356e-01
-5.02482891e-01 -4.44699645e-01 -4.47630823e-01 5.10766029e-01
-9.79822478e-04 5.97799681e-02 -1.10349071e+00 -3.02105427e-01
9.06976387e-02 -3.01743507e-01 9.91779685e-01 9.17740241e-02
1.35630405e+00 8.70607980e-03 1.75727963e-01 7.39372909e-01
1.23968995e+00 1.43436342e-01 8.98005009e-01 1.05126941e+00
9.04245734e-01 6.43022239e-01 4.29670736e-02 8.05651695e-02
1.84884205e-01 2.26588994e-01 5.11689425e-01 -1.29768893e-01
-8.70996490e-02 4.43074197e-01 -1.20167218e-01 7.03959644e-01
-8.29048991e-01 -8.99481475e-02 -1.19921470e+00 3.20124179e-01
-1.24480975e+00 -9.39678311e-01 -2.93516427e-01 2.40297246e+00
6.19463027e-01 3.38008225e-01 9.74869654e-02 2.47118980e-01
4.23332244e-01 -1.10390056e-02 -1.15291484e-01 -4.18691188e-01
-3.43516916e-01 5.05374312e-01 7.25349069e-01 2.22039502e-02
-9.15645778e-01 6.74070776e-01 6.16219139e+00 6.54420376e-01
-1.43999958e+00 1.28944740e-01 7.63807118e-01 -2.09140450e-01
-2.83553377e-02 2.09674224e-01 -2.97326952e-01 3.30533117e-01
1.26917684e+00 1.66326672e-01 7.96525359e-01 6.02154016e-01
-6.09563254e-02 -2.43853226e-01 -6.62018001e-01 8.33492041e-01
-1.51554212e-01 -1.20744908e+00 -4.40668941e-01 2.81371325e-01
7.66517878e-01 4.81567383e-01 -1.32884279e-01 3.91925931e-01
4.09474283e-01 -1.30220675e+00 4.96305048e-01 6.09610498e-01
6.13436460e-01 -6.66367888e-01 8.54699075e-01 9.64690968e-02
-9.24168348e-01 -2.67012000e-01 -6.54520392e-01 -4.10725802e-01
-5.40353656e-01 2.94748843e-01 -9.21262026e-01 2.97016650e-01
9.48722601e-01 4.59916413e-01 -1.20175564e+00 1.40574586e+00
-6.29088879e-02 9.21839476e-01 -4.44303662e-01 -3.34205002e-01
4.29691821e-01 -3.44742715e-01 5.88402478e-03 9.77420628e-01
4.49358404e-01 -8.34005922e-02 -2.27187052e-01 6.04753554e-01
9.11237076e-02 -3.02205980e-02 -3.84561211e-01 -3.27763259e-01
4.18078244e-01 1.65763402e+00 -1.43628597e+00 -2.59222865e-01
-6.71415448e-01 4.66705918e-01 1.45709619e-01 2.81589210e-01
-4.38673228e-01 -4.82565552e-01 5.00747502e-01 1.04680829e-01
2.72653289e-02 -3.07103217e-01 -5.85393846e-01 -9.04991210e-01
-5.30911982e-01 -1.01093280e+00 2.01595813e-01 -9.61355686e-01
-1.13959742e+00 6.09512150e-01 -2.19063029e-01 -1.36009133e+00
-6.42730296e-02 -1.09404922e+00 -8.00422966e-01 9.28515553e-01
-9.43669200e-01 -8.96985948e-01 -6.61758363e-01 3.24010730e-01
1.23267144e-01 -5.65531433e-01 7.48693407e-01 1.31548956e-01
-5.00194073e-01 1.27372354e-01 4.18969929e-01 2.28444010e-01
5.97455442e-01 -1.46858728e+00 7.14817166e-01 5.55433393e-01
3.40538561e-01 6.15633011e-01 6.57358527e-01 -3.96891415e-01
-1.09790528e+00 -9.00565863e-01 6.17386580e-01 -4.62989420e-01
6.21875644e-01 -3.48121524e-01 -9.03021991e-01 2.87372202e-01
2.15387225e-01 1.16964802e-01 4.78438377e-01 2.10714966e-01
-3.11236024e-01 2.09686570e-02 -9.78493869e-01 5.46460330e-01
7.67735064e-01 -6.33914530e-01 -2.11677253e-01 3.90185177e-01
3.60325426e-01 -2.42761344e-01 -1.11313653e+00 3.33073433e-03
5.32001793e-01 -1.41570568e+00 1.05463922e+00 -3.35204840e-01
3.01431447e-01 -2.79970944e-01 2.73772120e-01 -1.78086531e+00
-6.90293849e-01 -2.08524138e-01 4.37566668e-01 7.08081305e-01
3.32698852e-01 -1.00494063e+00 8.08803976e-01 6.63414836e-01
-2.11510390e-01 -3.63978356e-01 -9.31438506e-01 -8.64380062e-01
2.46137366e-01 -5.11831164e-01 7.12235808e-01 1.01298702e+00
-4.13662285e-01 9.95069668e-02 -4.62106019e-02 -2.08679959e-01
2.47622654e-01 2.07800027e-02 9.32661474e-01 -1.62010753e+00
-4.14975822e-01 -1.00197875e+00 -8.03539157e-01 -3.96630801e-02
-1.94138661e-01 -9.85623181e-01 -4.40824740e-02 -1.39135325e+00
-3.28376234e-01 -6.04555309e-01 -2.63175040e-01 5.33159733e-01
7.90918246e-03 8.25306833e-01 3.54344100e-01 4.15815294e-01
-3.91827017e-01 5.67337126e-02 1.29122388e+00 1.60364270e-01
-1.50204180e-02 -4.54940647e-03 -5.57938933e-01 5.97493589e-01
1.31996417e+00 -1.69251934e-01 -5.21783866e-02 -1.77085191e-01
5.11281967e-01 -6.36203051e-01 8.46059680e-01 -1.42639005e+00
1.69347018e-01 -6.80024698e-02 9.03042436e-01 -2.65064202e-02
3.71617943e-01 -6.21496439e-01 4.63542670e-01 2.87201077e-01
2.70566028e-02 -4.37275022e-02 1.53438225e-01 4.36346866e-02
-2.42763564e-01 -3.23482186e-01 9.43971157e-01 -5.24493694e-01
-7.32137561e-01 -1.02669515e-01 -3.45288426e-01 -3.11213676e-02
7.64347672e-01 -5.09936452e-01 -4.97224122e-01 -1.30896166e-01
-4.31802183e-01 -2.05848113e-01 7.43450522e-01 4.41238552e-01
3.14639390e-01 -1.21716189e+00 -4.30486500e-01 4.13937151e-01
-1.79011747e-02 1.75328672e-01 -1.87968984e-02 8.94982874e-01
-1.08662117e+00 3.88952047e-01 -6.12787843e-01 -4.71990108e-01
-1.20585299e+00 2.04211965e-01 5.62280178e-01 -1.53704034e-02
-4.30844665e-01 1.25132132e+00 3.84829752e-02 -7.18069851e-01
-1.65268809e-01 -1.67998627e-01 -2.76055038e-01 -1.78435128e-02
2.69353718e-01 4.81531829e-01 3.81949842e-01 -5.00227392e-01
-4.95912731e-02 2.83807039e-01 1.26371950e-01 -1.03252776e-01
1.46331871e+00 1.81646749e-01 -3.17300826e-01 6.23871148e-01
1.05269217e+00 -4.55408692e-01 -1.02653289e+00 4.53578755e-02
1.16230682e-01 -6.70974910e-01 2.28668109e-01 -7.46474862e-01
-1.03571689e+00 9.51974571e-01 6.86623573e-01 8.35157931e-01
1.10661483e+00 -1.84960738e-02 1.85596749e-01 6.31249845e-01
2.96585649e-01 -1.09945309e+00 -3.40651385e-02 7.77412713e-01
8.29048872e-01 -1.15307784e+00 -1.28300309e-01 3.50154877e-01
-4.65750724e-01 1.25420177e+00 7.57348001e-01 -2.62672842e-01
4.70675945e-01 -2.20908910e-01 1.14164695e-01 -3.37898016e-01
-6.45707667e-01 -2.49074906e-01 4.73865777e-01 4.35899079e-01
7.79248536e-01 9.87048671e-02 -4.55884868e-03 2.15950739e-02
-9.30310845e-01 -1.73088223e-01 5.97325683e-01 8.63086760e-01
-7.40751386e-01 -9.68352377e-01 -1.06863797e+00 1.07153261e+00
-2.78448790e-01 -9.45122689e-02 -8.45606744e-01 8.02052617e-01
5.41773856e-01 8.63280237e-01 4.03249681e-01 -5.41561186e-01
1.80777371e-01 1.75996035e-01 4.99519259e-01 -5.40995359e-01
-1.04216933e+00 -1.40993729e-01 5.08635081e-02 -1.81633934e-01
-3.86250973e-01 -5.50361156e-01 -9.51566279e-01 -6.83335185e-01
-4.72779125e-01 3.57993931e-01 9.03007090e-01 8.27061176e-01
1.04525045e-01 5.91889322e-01 6.34352386e-01 -1.22169697e+00
-8.03446621e-02 -1.40207589e+00 -6.09474719e-01 4.06469792e-01
5.73682562e-02 -6.20334983e-01 -7.08099842e-01 -1.25454724e-01]
|
[7.918581008911133, 2.969149589538574]
|
6cee4325-d403-4ec9-ba9a-2d33b9b5adfd
|
class-incremental-learning-for-video-action
| null | null |
https://ieeexplore.ieee.org/document/9506788
|
https://ieeexplore.ieee.org/document/9506788
|
Class incremental learning for video action classification
|
Class Incremental Learning (CIL) is a hot topic in machine learning for CNN models to learn new classes incrementally. However, most of the CIL studies are for image classification and object recognition tasks and few CIL studies are available for video action classification. To mitigate this problem, in this paper, we present a new Grow When Required network (GWR) based video CIL framework for action classification. GWR learns knowledge incrementally by modeling the manifold of video frames for each encountered action class in feature space. We also introduce a Knowledge Consolidation (KC) method to separate the feature manifolds of old class and new class and introduce an associative matrix for label prediction. Experimental results on KTH and Weizmann demonstrate the effectiveness of the framework.
|
['Yihong Gong', 'Xiaopeng Hong', 'Jianxing Ma', 'Xiaoyu Tao', 'Jiawei Ma']
|
2021-09-19
| null | null | null |
ieee-international-conference-on-image-9
|
['class-incremental-learning', 'action-classification', 'action-recognition-in-videos-2']
|
['computer-vision', 'computer-vision', 'computer-vision']
|
[ 3.31249952e-01 -2.75333613e-01 -8.01228046e-01 -2.60536104e-01
-6.29859269e-02 -3.23485315e-01 4.91489857e-01 -8.42896476e-02
-3.53281736e-01 7.37849772e-01 1.22977823e-01 -5.01472466e-02
-2.46643826e-01 -6.17250323e-01 -7.26033330e-01 -5.31403184e-01
-3.57832670e-01 -2.67947257e-01 5.51062942e-01 3.62340868e-01
3.24682802e-01 4.49416280e-01 -1.51693559e+00 8.00410688e-01
4.98484701e-01 1.07622564e+00 -5.39001217e-03 6.90874934e-01
5.25829047e-02 1.87357962e+00 -1.84252843e-01 -9.27516595e-02
3.38927448e-01 -5.73164880e-01 -1.21031988e+00 4.10158217e-01
5.67781329e-01 -3.98725301e-01 -7.33853638e-01 8.91887009e-01
9.73524433e-03 5.79209507e-01 8.03509772e-01 -1.63431823e+00
-9.03931618e-01 5.73572755e-01 -3.77668738e-01 6.21046603e-01
1.57736823e-01 -1.91968996e-02 6.36485994e-01 -1.22719276e+00
9.31459904e-01 7.91385055e-01 6.77410007e-01 7.96504200e-01
-5.58682084e-01 -6.66316748e-01 6.18266523e-01 1.17703283e+00
-1.28292477e+00 -3.31434757e-01 8.39465678e-01 -4.90480393e-01
1.03382730e+00 -4.15562876e-02 1.16296566e+00 8.45359981e-01
9.88004878e-02 1.28803229e+00 9.08876598e-01 -3.14524502e-01
4.05500561e-01 -2.38314360e-01 3.93845379e-01 8.06092858e-01
-7.46189151e-03 -1.92822918e-01 -9.89543915e-01 4.39057916e-01
8.09795380e-01 4.32528794e-01 -1.10792495e-01 -6.54981732e-01
-1.08806229e+00 5.54507077e-01 3.75252932e-01 3.95770192e-01
-2.21004337e-01 4.40015733e-01 5.32632828e-01 5.07032812e-01
5.16014576e-01 2.86478698e-02 -4.69633192e-01 -4.18712646e-01
-5.80608845e-01 -1.15676358e-01 3.04045558e-01 9.13929939e-01
7.26720870e-01 3.27395290e-01 -1.65168718e-01 7.46967018e-01
8.49921182e-02 3.84644233e-02 1.04545701e+00 -1.19301724e+00
1.35160506e-01 1.00828242e+00 -2.80233413e-01 -9.66755569e-01
-1.07866302e-01 -5.01535609e-02 -7.10502744e-01 1.39737800e-01
6.46959394e-02 1.66625619e-01 -9.87026095e-01 1.25466645e+00
3.22769552e-01 1.16967607e+00 3.40128481e-01 4.61788476e-01
7.48300433e-01 5.63839316e-01 2.67448008e-01 -4.53490078e-01
6.60216689e-01 -1.27271175e+00 -7.23486423e-01 8.37571323e-02
8.84586692e-01 -7.58143291e-02 5.65383971e-01 2.73758084e-01
-6.67132735e-01 -8.23284388e-01 -9.12128568e-01 9.25144404e-02
-5.18748581e-01 1.93121165e-01 8.45412374e-01 3.14808786e-01
-8.64875972e-01 7.14022577e-01 -1.01563203e+00 -3.84443134e-01
1.05491972e+00 3.04527909e-01 -6.52936339e-01 -2.35487700e-01
-7.44004726e-01 6.97431505e-01 7.97596633e-01 6.06034324e-02
-1.02710879e+00 -6.71248674e-01 -6.80431008e-01 -1.96501806e-01
3.11478704e-01 -2.72796333e-01 1.18590796e+00 -1.80033612e+00
-1.72911894e+00 5.94646454e-01 5.01721948e-02 -9.06576157e-01
1.35807529e-01 -4.07598436e-01 -4.20569152e-01 4.53126132e-01
-2.04121873e-01 8.37763786e-01 1.03142309e+00 -8.46166551e-01
-8.40671062e-01 -3.87518853e-01 2.07098126e-01 3.46612334e-01
-7.29993582e-01 -3.75121564e-01 2.99462164e-03 -6.71970606e-01
1.43110812e-01 -8.82409215e-01 8.98943692e-02 -1.68015122e-01
3.70626450e-01 -6.23901486e-01 1.28716397e+00 -5.55771768e-01
1.31530070e+00 -2.34160256e+00 2.32417718e-01 -1.89265758e-01
1.43426731e-01 6.57309055e-01 -1.20927252e-01 6.55997768e-02
-5.41522801e-01 -2.78043002e-01 1.82847396e-01 1.01905659e-01
-6.05024457e-01 3.79289418e-01 -4.12810355e-01 3.48327577e-01
2.27977470e-01 1.18530381e+00 -1.17062962e+00 -4.06720936e-01
2.91141808e-01 2.28495851e-01 -4.85466719e-01 1.68191433e-01
-1.09930769e-01 2.94586211e-01 -1.89209014e-01 8.92836273e-01
2.83330321e-01 -1.94126964e-01 8.81761685e-02 -4.19237345e-01
-1.00652061e-01 -3.76600206e-01 -1.05500531e+00 1.74376988e+00
3.69739048e-02 8.65534663e-01 -8.02396178e-01 -1.41418648e+00
5.61064839e-01 3.63984734e-01 6.61468923e-01 -3.96267265e-01
1.06588535e-01 8.28330033e-03 -4.73102331e-02 -7.83321261e-01
3.76988560e-01 -2.83712521e-02 4.42962140e-01 1.18319087e-01
5.83705068e-01 5.87619781e-01 1.04314312e-01 1.84076428e-01
1.27460361e+00 3.63651663e-01 3.33471656e-01 6.18747212e-02
5.72973073e-01 1.35329500e-01 7.38301158e-01 5.97422302e-01
-7.37703502e-01 1.66311845e-01 -1.13197304e-01 -9.42206323e-01
-5.80944121e-01 -1.02991676e+00 1.96773529e-01 1.23349917e+00
1.22615397e-01 -3.26194078e-01 -6.45935655e-01 -1.06230676e+00
-8.88590217e-02 4.50852364e-01 -8.18070769e-01 -6.54126525e-01
-6.84973598e-01 -4.25912261e-01 2.02385798e-01 1.04518032e+00
9.35976744e-01 -1.24531865e+00 -5.51074505e-01 1.52560890e-01
-1.22113608e-01 -9.87005413e-01 -2.64051586e-01 2.41764653e-02
-1.28991055e+00 -1.51551545e+00 -5.10580420e-01 -1.09123516e+00
8.19823384e-01 6.93183064e-01 4.37531143e-01 -5.16346768e-02
-5.84859848e-01 1.35650706e+00 -8.73467267e-01 -1.56327829e-01
-5.68482652e-02 -1.65466130e-01 4.03932661e-01 5.44817388e-01
5.69189548e-01 -3.90791297e-01 -6.53750956e-01 1.81588978e-01
-9.55278039e-01 1.13733344e-01 4.19274092e-01 7.79453218e-01
6.56082392e-01 2.08715782e-01 8.12380850e-01 -7.37479270e-01
1.94067717e-01 -5.52711189e-01 -1.73681617e-01 4.15369421e-01
-4.14685518e-01 -2.71920294e-01 3.37646782e-01 -1.03822768e+00
-1.04665124e+00 3.57132673e-01 5.04729211e-01 -8.29452693e-01
-7.77095780e-02 5.59934795e-01 2.12831765e-01 -4.01443571e-01
3.46911669e-01 4.95098919e-01 -2.73330007e-02 -2.03947395e-01
5.63627601e-01 6.92596436e-01 4.89994079e-01 -3.03887039e-01
4.27353323e-01 6.00273788e-01 -6.64263591e-02 -8.60039115e-01
-1.05288339e+00 -7.31682777e-01 -1.23072362e+00 -8.78424406e-01
1.15447628e+00 -1.07463062e+00 -4.27555174e-01 9.59347069e-01
-7.01041996e-01 -6.42693877e-01 -7.60212958e-01 7.34460235e-01
-8.11225533e-01 3.65851671e-01 -4.60020840e-01 -6.62868798e-01
-2.38156263e-02 -6.89971626e-01 3.69808316e-01 3.16627681e-01
1.28579170e-01 -9.13161099e-01 2.67022014e-01 3.30714464e-01
2.94002205e-01 -9.96856019e-03 6.90416515e-01 -4.56040323e-01
-7.82656670e-01 -2.52123713e-01 -1.32046521e-01 6.96154356e-01
4.33962703e-01 -1.28370419e-01 -7.14177310e-01 -3.17980617e-01
-2.29333684e-01 -5.62085927e-01 1.05331075e+00 1.28261298e-01
1.34131670e+00 -2.30709463e-01 -3.75111073e-01 2.34626308e-01
1.34942448e+00 6.67324662e-01 8.34570110e-01 3.37077379e-01
9.70016778e-01 1.15242690e-01 5.85461438e-01 2.43861303e-01
2.57633090e-01 4.22770858e-01 1.40539289e-01 5.57363212e-01
-5.62303722e-01 -1.81768194e-01 7.71784961e-01 1.17535853e+00
-5.32640576e-01 2.75166035e-01 -7.85281718e-01 4.92700338e-01
-2.25654078e+00 -1.34542561e+00 3.95300776e-01 1.85493124e+00
6.80275857e-01 -8.24360475e-02 -6.63832873e-02 1.23979658e-01
6.07998133e-01 -1.87356502e-01 -7.29311109e-01 4.93190326e-02
9.55993459e-02 1.97009996e-01 1.82426035e-01 1.89315826e-01
-1.44657075e+00 1.30080760e+00 6.45286751e+00 9.71168578e-01
-9.17667806e-01 1.35459229e-01 6.02156222e-01 1.41700236e-02
6.18251324e-01 1.02340981e-01 -7.54185677e-01 7.66034946e-02
9.05786633e-01 -1.26837313e-01 2.15516463e-01 1.11761081e+00
-2.40260974e-01 -2.52541214e-01 -1.12377787e+00 1.44674528e+00
4.79132295e-01 -1.44116390e+00 5.55534840e-01 -4.19720203e-01
9.44235265e-01 -5.62777482e-02 2.04722151e-01 6.39606893e-01
1.92216650e-01 -6.48815930e-01 3.35411012e-01 9.35197055e-01
4.99947578e-01 -4.94063437e-01 3.51527929e-01 3.02551240e-02
-1.44677985e+00 -6.96526408e-01 -4.85653251e-01 -2.83327907e-01
-3.08642685e-01 -7.40521327e-02 -6.59100711e-01 3.63937646e-01
8.01617384e-01 1.59657323e+00 -1.09319699e+00 1.19272685e+00
7.28260502e-02 8.11132371e-01 2.25927711e-01 1.69563919e-01
1.69278562e-01 5.09045273e-02 2.73706257e-01 8.72680068e-01
1.04743809e-01 3.73200953e-01 1.79923132e-01 -1.37869678e-02
-2.38688260e-01 1.50819242e-01 -8.50998223e-01 -1.75346866e-01
1.05501801e-01 9.97431040e-01 -9.87182081e-01 -4.82678682e-01
-5.75307131e-01 1.45294106e+00 6.37880802e-01 4.14003849e-01
-6.82458103e-01 -4.43237834e-02 4.36392277e-01 -7.47670606e-02
5.08546293e-01 -5.19129097e-01 3.35337102e-01 -1.43736041e+00
-9.35459211e-02 -4.65026289e-01 6.48094714e-01 -8.50614071e-01
-1.09101963e+00 2.03155443e-01 1.29984021e-01 -1.65815580e+00
8.34335908e-02 -8.09440136e-01 -3.78920466e-01 -4.40407723e-01
-1.48945272e+00 -1.18781924e+00 -3.60036254e-01 9.49359894e-01
8.87027025e-01 -7.97917366e-01 7.63281822e-01 3.01413357e-01
-6.30242288e-01 5.28189182e-01 2.09049776e-01 4.37841505e-01
4.34462994e-01 -9.50707674e-01 -2.96226144e-01 6.97816312e-01
5.73703587e-01 4.38455224e-01 -1.69305444e-01 -7.12437332e-01
-1.44360852e+00 -1.47796750e+00 4.04721886e-01 -5.92067719e-01
7.28796661e-01 6.39587045e-02 -8.51824462e-01 1.05332911e+00
1.12644054e-01 3.26847464e-01 9.64253664e-01 -1.09194778e-01
-5.85882902e-01 -3.14838171e-01 -7.35347092e-01 5.61561286e-01
1.34480298e+00 -4.53534752e-01 -6.06851101e-01 4.36716408e-01
6.46944404e-01 -1.90526526e-03 -7.77796566e-01 5.84394634e-01
6.91864729e-01 -6.55495644e-01 8.23908985e-01 -1.12033188e+00
1.85627222e-01 -4.73979712e-01 -2.99055457e-01 -1.03238332e+00
-4.71668124e-01 -3.29231285e-02 -6.41138911e-01 1.02456665e+00
7.60909319e-02 -2.88641691e-01 9.23444867e-01 3.95007372e-01
-2.21864998e-01 -7.70824671e-01 -1.19003260e+00 -1.07652521e+00
-9.74199176e-02 -6.09845936e-01 -7.01130033e-02 1.22646749e+00
2.58295715e-01 2.13773370e-01 -4.32220221e-01 -2.32804328e-01
3.51110101e-01 7.85484686e-02 5.41682661e-01 -9.55569625e-01
-9.73819867e-02 -2.35588625e-01 -1.15638781e+00 -6.91222429e-01
2.47550026e-01 -1.22028041e+00 -3.74050230e-01 -1.39793861e+00
5.51283121e-01 -1.35536820e-01 -7.59869039e-01 7.98448443e-01
-3.59817110e-02 2.97414094e-01 2.15366289e-01 3.47984910e-01
-1.54586422e+00 6.73584640e-01 8.78308058e-01 -2.90152818e-01
-2.76543558e-01 -1.87458009e-01 -1.60818025e-01 8.27347517e-01
8.96830976e-01 -2.26624936e-01 -7.07454264e-01 -4.21694368e-01
1.44428879e-01 -5.77527285e-01 4.45616424e-01 -1.50221181e+00
4.14259583e-01 -1.69345215e-01 6.37318492e-01 -4.56964403e-01
6.80002049e-02 -9.74965394e-01 4.37125564e-03 5.18246710e-01
-3.09045643e-01 -8.65241587e-02 1.35537341e-01 1.03174138e+00
-3.36844087e-01 -1.91315517e-01 7.59080470e-01 -1.98706210e-01
-1.51700425e+00 7.13217080e-01 -6.46273792e-01 -1.11386172e-01
1.54917061e+00 -5.72536469e-01 -1.77313015e-01 -4.75914404e-02
-1.08331835e+00 4.57584560e-02 -1.34911507e-01 7.02589691e-01
1.28105831e+00 -1.84323227e+00 -3.43433797e-01 8.77834298e-03
3.44296038e-01 -3.15136343e-01 4.98266518e-01 6.73204303e-01
-4.32008147e-01 2.17647552e-01 -5.06248832e-01 -3.54009092e-01
-1.39322937e+00 6.76022828e-01 2.56031305e-01 -6.42558560e-03
-6.91575348e-01 1.02697921e+00 4.71250201e-03 -1.22561096e-03
3.72081250e-01 -1.05352975e-01 -6.46050513e-01 2.61344939e-01
8.54924023e-01 6.97776258e-01 -2.33157545e-01 -6.41198874e-01
-1.20938063e-01 5.34057617e-01 -3.82207513e-01 3.88723254e-01
1.31663132e+00 -1.01705328e-01 -1.03693768e-01 1.13115835e+00
1.29337299e+00 -9.67870176e-01 -1.36724651e+00 -3.31541449e-01
1.37006372e-01 -3.52857381e-01 -1.21992655e-01 -5.93426168e-01
-1.06285286e+00 7.38952041e-01 1.21345270e+00 -3.07814449e-01
1.10400188e+00 -8.91077593e-02 6.04806244e-01 7.48938203e-01
5.19313812e-01 -1.60765755e+00 9.99602735e-01 6.93910182e-01
6.22923493e-01 -1.18357873e+00 -6.75676018e-02 -2.20529020e-01
-6.89219356e-01 1.43016732e+00 8.70743930e-01 -2.97246873e-01
1.19621289e+00 -3.51189286e-01 -1.00538671e-01 -1.66408673e-01
-7.37602532e-01 -1.82420388e-01 1.70765311e-01 7.33641863e-01
1.79468393e-01 -1.07657611e-01 -1.46853432e-01 5.28971374e-01
5.09090424e-01 5.67673326e-01 5.81361711e-01 1.36787724e+00
-6.32140040e-01 -9.49104369e-01 2.63685048e-01 5.51892936e-01
-2.38478735e-01 7.70134199e-03 -2.17238665e-01 4.37852561e-01
3.71534318e-01 5.70287585e-01 5.79563826e-02 -7.71176159e-01
1.45603623e-02 4.83321846e-01 9.02162492e-01 -7.28441119e-01
-4.39206092e-03 -5.39904594e-01 -4.16834921e-01 -5.87038517e-01
-1.11936688e+00 -7.86896110e-01 -1.23577821e+00 3.90913963e-01
-4.34889346e-01 -9.83996913e-02 2.13431299e-01 9.37255263e-01
3.73540312e-01 4.72612351e-01 6.61917329e-01 -5.58354795e-01
-6.34289086e-02 -9.06478763e-01 -6.57514274e-01 4.59837973e-01
-1.03947008e-02 -1.10326636e+00 -1.81056768e-01 8.33169520e-01]
|
[8.681632041931152, 0.8974113464355469]
|
e7bae880-81a5-4fa6-9d1b-377e89b8ddd4
|
bolt-an-automated-deep-learning-framework-for
|
2303.17727
| null |
https://arxiv.org/abs/2303.17727v3
|
https://arxiv.org/pdf/2303.17727v3.pdf
|
BOLT: An Automated Deep Learning Framework for Training and Deploying Large-Scale Search and Recommendation Models on Commodity CPU Hardware
|
Efficient large-scale neural network training and inference on commodity CPU hardware is of immense practical significance in democratizing deep learning (DL) capabilities. Presently, the process of training massive models consisting of hundreds of millions to billions of parameters requires the extensive use of specialized hardware accelerators, such as GPUs, which are only accessible to a limited number of institutions with considerable financial resources. Moreover, there is often an alarming carbon footprint associated with training and deploying these models. In this paper, we take a step towards addressing these challenges by introducing BOLT, a sparse deep learning library for training large-scale search and recommendation models on standard CPU hardware. BOLT provides a flexible, high-level API for constructing models that will be familiar to users of existing popular DL frameworks. By automatically tuning specialized hyperparameters, BOLT also abstracts away the algorithmic details of sparse network training. We evaluate BOLT on a number of information retrieval tasks including product recommendations, text classification, graph neural networks, and personalization. We find that our proposed system achieves competitive performance with state-of-the-art techniques at a fraction of the cost and energy consumption and an order-of-magnitude faster inference time. BOLT has also been successfully deployed by multiple businesses to address critical problems, and we highlight one customer deployment case study in the field of e-commerce.
|
['Anshumali Shrivastava', 'Tharun Medini', 'Yashwanth Adunukota', 'Shubh Gupta', 'Benjamin Meisburger', 'Benjamin Coleman', 'Pratik Pranav', 'David Torres Ramos', 'Joshua Engels', 'Benito Geordie', 'Vihan Lakshman', 'Nicholas Meisburger']
|
2023-03-30
| null | null | null | null |
['fraud-detection']
|
['miscellaneous']
|
[-2.22557515e-01 -9.42469090e-02 -5.85839748e-01 -3.57130349e-01
-3.08396667e-01 -3.50776970e-01 3.54046196e-01 2.65214026e-01
-3.41257006e-01 2.36233249e-01 -7.81009495e-02 -8.86910737e-01
-1.71173006e-01 -1.06499195e+00 -7.67069995e-01 -5.29470742e-01
-1.49236217e-01 7.72821784e-01 -2.07946658e-01 -1.86342224e-01
1.35492012e-02 5.71531653e-01 -1.53519368e+00 3.46137881e-01
4.23037291e-01 1.31115317e+00 9.74178836e-02 4.56036657e-01
-1.21908322e-01 6.45909429e-01 -2.95188278e-01 -5.64788938e-01
4.18027759e-01 2.77533591e-01 -5.82145751e-01 -1.62212312e-01
6.40382886e-01 -5.33723474e-01 -5.58088183e-01 8.25218201e-01
5.37672698e-01 2.70078421e-01 1.97783560e-01 -1.04381406e+00
-4.07024533e-01 7.67857313e-01 -5.47907352e-01 3.18003297e-01
-1.94184601e-01 9.13726613e-02 1.21003795e+00 -8.04536104e-01
2.55301476e-01 9.95571613e-01 8.85876060e-01 9.53430831e-02
-1.06131053e+00 -7.60881901e-01 2.22196445e-01 1.21322468e-01
-1.33795607e+00 -4.67287779e-01 5.00231743e-01 -7.17783794e-02
1.56867957e+00 1.58938915e-01 7.49667525e-01 1.13308489e+00
1.54439926e-01 7.23667920e-01 3.59473944e-01 -1.79359525e-01
3.76125425e-01 1.22023344e-01 5.48086882e-01 1.06550658e+00
6.50281370e-01 -8.17088634e-02 -5.89576840e-01 -4.70347077e-01
8.24271679e-01 1.31731346e-01 3.10219228e-01 -1.15382679e-01
-8.84314120e-01 1.19227540e+00 5.13681531e-01 2.19201416e-01
-4.76504475e-01 4.88032460e-01 7.47024477e-01 1.46647662e-01
5.88180065e-01 2.86057502e-01 -6.86717033e-01 -7.26127923e-02
-7.66674936e-01 1.66317686e-01 1.22846246e+00 9.21347797e-01
5.90463340e-01 4.52562213e-01 8.41950625e-02 8.78830254e-01
1.70767665e-01 2.70390779e-01 5.05997658e-01 -8.39072049e-01
4.18667436e-01 4.98800278e-01 -3.56683016e-01 -1.28157938e+00
-5.05515516e-01 -9.35609698e-01 -1.33142936e+00 -3.27658623e-01
1.93236858e-01 -2.89404124e-01 -5.76450408e-01 1.13527882e+00
5.06182849e-01 1.89562365e-01 -3.09862852e-01 7.17726171e-01
7.82758355e-01 8.69786143e-01 1.01920798e-01 2.19349056e-01
1.47000349e+00 -1.28653336e+00 -2.48198211e-01 -5.04240215e-01
1.00496614e+00 -5.70610523e-01 1.30007875e+00 6.61441922e-01
-9.29608703e-01 -4.07113463e-01 -1.00955498e+00 -3.25881243e-01
-3.96238148e-01 2.24054173e-01 1.75171065e+00 7.18502641e-01
-9.44857061e-01 7.27208912e-01 -1.05450976e+00 -4.20708090e-01
7.01660275e-01 5.92735589e-01 -3.61841917e-02 -4.28790957e-01
-8.56434464e-01 5.38767874e-01 2.14157209e-01 1.61145240e-01
-4.78889763e-01 -8.96587610e-01 -6.94782495e-01 5.16115844e-01
3.63795966e-01 -9.33055103e-01 1.12651753e+00 -5.08253038e-01
-1.46304560e+00 4.66130763e-01 1.33808494e-01 -6.46135330e-01
-5.28027527e-02 -1.37525827e-01 -4.05451268e-01 -2.79964954e-01
-4.12207931e-01 2.79397637e-01 8.14579964e-01 -5.15454650e-01
-5.14519274e-01 -4.58549470e-01 1.76459029e-01 1.74167454e-02
-1.08818448e+00 -5.24748638e-02 -8.09470892e-01 -4.51606810e-01
-3.02719384e-01 -1.07510400e+00 -4.51390833e-01 -1.77714840e-01
-2.99960911e-01 -4.24112052e-01 8.22657168e-01 -4.71516728e-01
1.22099900e+00 -1.89563990e+00 -1.61762580e-01 4.61067408e-01
5.11206090e-01 4.23738062e-01 -1.72840551e-01 7.18595609e-02
3.16367686e-01 -1.48846880e-01 4.81855184e-01 -3.29249918e-01
2.21311942e-01 3.84919167e-01 -1.97404727e-01 3.39293540e-01
-3.42946798e-01 9.00080919e-01 -6.32137775e-01 -2.12184772e-01
1.12919576e-01 6.69773638e-01 -9.52666879e-01 -3.50070857e-02
-2.92836219e-01 -1.50278404e-01 -6.27556264e-01 7.51341581e-01
3.24709654e-01 -1.01851785e+00 5.74893653e-01 -4.92596865e-01
2.82010615e-01 5.55364251e-01 -9.97326672e-01 1.81321728e+00
-9.46177661e-01 4.56429005e-01 8.48321691e-02 -1.16247308e+00
5.64971268e-01 -1.30822420e-01 3.83469820e-01 -7.33675003e-01
3.86371404e-01 7.74650127e-02 -9.51819494e-02 -1.03170574e-01
6.25561833e-01 4.02468443e-01 -2.72806492e-02 6.72093272e-01
1.80664346e-01 2.75740266e-01 2.07287133e-01 2.79947847e-01
1.16055346e+00 -3.72364998e-01 -9.20980796e-03 -3.86190057e-01
-7.72316232e-02 -3.51360999e-02 2.86357045e-01 8.35216045e-01
4.90681380e-01 -2.18932092e-01 4.01534468e-01 -1.00687480e+00
-1.11635721e+00 -4.46624428e-01 -1.69118077e-01 1.67446637e+00
-4.85538781e-01 -8.14926803e-01 -6.06684804e-01 -4.63171542e-01
3.94620568e-01 6.33973062e-01 -3.14861715e-01 -8.18297341e-02
-6.16749048e-01 -1.10122216e+00 3.72513473e-01 6.42470896e-01
5.09903252e-01 -6.73160017e-01 -3.87317449e-01 5.64606003e-02
4.49572235e-01 -1.27982795e+00 -4.57303852e-01 3.49745870e-01
-1.12130678e+00 -6.23776078e-01 -2.60331128e-02 -7.65680790e-01
7.46863782e-01 5.01737475e-01 1.43648374e+00 5.18934846e-01
-4.66516733e-01 5.37428595e-02 -4.87952568e-02 -3.01401973e-01
-6.51812032e-02 8.19547117e-01 1.99873269e-01 -3.18955481e-01
4.50770974e-01 -7.40028501e-01 -6.74385011e-01 6.41414821e-02
-6.67526364e-01 1.14005774e-01 7.45337009e-01 7.91006207e-01
5.34959435e-01 2.72425801e-01 2.45293021e-01 -1.24366665e+00
7.17202365e-01 -6.96596324e-01 -1.01423383e+00 3.99244055e-02
-1.09627402e+00 4.17286977e-02 7.87799239e-01 -5.19849658e-01
-6.12563491e-01 -1.44249931e-01 -4.05064940e-01 -4.24011797e-01
2.01931134e-01 8.70175898e-01 2.51273632e-01 -3.71131748e-01
7.28371561e-01 -1.97585613e-01 -1.56782925e-01 -5.85019171e-01
4.58990186e-01 5.13548911e-01 3.05467039e-01 -7.93506801e-01
5.65846324e-01 1.43035501e-01 1.35193229e-01 -7.22451031e-01
-1.15824819e+00 -2.33261496e-01 -1.49694249e-01 2.03572363e-01
3.05604160e-01 -9.76640642e-01 -1.00058746e+00 2.21111700e-01
-7.80473948e-01 -5.96170902e-01 -1.62226528e-01 2.83932298e-01
3.08644772e-01 1.37064233e-01 -1.01634777e+00 -1.57368749e-01
-1.03031766e+00 -1.05112505e+00 9.72626269e-01 9.33151096e-02
-1.05684832e-01 -1.18751097e+00 -4.05262113e-01 5.61463296e-01
8.55485916e-01 -2.15979815e-01 1.12833524e+00 -7.13892519e-01
-6.97750032e-01 -5.39071381e-01 -2.73692995e-01 3.96197110e-01
-3.29865575e-01 -2.16808423e-01 -7.80959904e-01 -7.55042493e-01
-1.46842092e-01 -4.62761879e-01 6.39313400e-01 3.14271212e-01
1.87471545e+00 -5.80892563e-01 -4.32689071e-01 1.12307358e+00
1.36396730e+00 -1.54194027e-01 2.86621630e-01 2.28413090e-01
1.12902629e+00 -4.38580429e-03 1.71320915e-01 5.06185472e-01
3.35360169e-01 6.24379694e-01 2.31554151e-01 -1.91987962e-01
1.56861261e-01 -2.09743798e-01 -2.76590306e-02 1.19590712e+00
-9.24431160e-02 -2.31953353e-01 -8.55294228e-01 1.66588277e-01
-1.76198018e+00 -6.59936965e-01 1.20442480e-01 2.00848866e+00
6.41810536e-01 3.78791094e-01 -3.82025680e-03 -1.02668002e-01
1.33187756e-01 1.55109912e-01 -8.90362322e-01 -5.53841293e-01
2.96668828e-01 5.27879775e-01 9.14800942e-01 2.41623178e-01
-9.38052893e-01 9.76145446e-01 6.24413967e+00 1.15059876e+00
-1.22922599e+00 3.16888213e-01 9.24031734e-01 -4.03090686e-01
-1.99929193e-01 -2.40978450e-01 -1.12020540e+00 3.69570673e-01
1.47786963e+00 -7.98558742e-02 7.39070475e-01 1.45071971e+00
-1.86613053e-01 1.23881847e-01 -1.18276215e+00 1.20794833e+00
-8.62233639e-02 -1.94585741e+00 1.80069372e-01 4.14561838e-01
7.42545784e-01 4.85914201e-01 2.55953938e-01 5.95526457e-01
6.83489382e-01 -1.06308424e+00 2.56413907e-01 -1.11925729e-01
6.87731326e-01 -9.02232707e-01 6.98900282e-01 3.09459299e-01
-1.00240231e+00 -2.58627951e-01 -7.37045467e-01 -1.56693757e-01
-1.54090300e-01 9.96028960e-01 -7.72075176e-01 1.39507234e-01
8.71488452e-01 6.04381323e-01 -4.86548185e-01 7.02524662e-01
7.43421093e-02 8.19190919e-01 -5.59079766e-01 -2.57609129e-01
2.03197718e-01 -3.12767714e-01 -1.77428588e-01 1.29678881e+00
4.12084043e-01 6.55726790e-02 1.68980271e-01 4.31156486e-01
-5.46734929e-01 3.38674970e-02 -3.98146212e-01 -1.68227270e-01
5.58752358e-01 1.71125817e+00 -6.93657696e-01 -4.59495217e-01
-5.99480748e-01 6.53316617e-01 4.87361610e-01 2.36837361e-02
-8.18989515e-01 -1.42926767e-01 7.77076066e-01 7.15562180e-02
2.31260315e-01 -3.77356350e-01 -4.63669479e-01 -1.15396273e+00
-1.10693663e-01 -1.18289912e+00 4.02775258e-01 -4.47815388e-01
-1.25232840e+00 5.69562793e-01 -2.98775524e-01 -3.80798012e-01
-2.46756732e-01 -7.48445153e-01 -3.42970848e-01 7.01853335e-01
-1.31571567e+00 -1.06993568e+00 -5.47663867e-01 6.66459799e-01
4.01341558e-01 -4.97342855e-01 9.73344743e-01 7.84660041e-01
-9.34934735e-01 8.61016870e-01 2.91500956e-01 -3.31338705e-03
1.28213644e-01 -7.43570685e-01 1.00112522e+00 4.63046879e-01
2.87523210e-01 8.64267349e-01 5.17130554e-01 -3.56677145e-01
-2.27862239e+00 -1.18701100e+00 6.30675972e-01 -2.32877329e-01
1.05145442e+00 -7.56693125e-01 -7.75815308e-01 9.17691529e-01
-1.18097700e-01 2.59694338e-01 7.42036402e-01 7.78708696e-01
-3.57825458e-01 -4.49158132e-01 -8.18159819e-01 4.93466467e-01
1.13367891e+00 -4.92449462e-01 1.41029328e-01 1.05261636e+00
7.92286634e-01 -6.23189449e-01 -1.04336631e+00 1.39151677e-01
5.86281180e-01 -6.16378188e-01 1.29143250e+00 -5.95133007e-01
1.38513952e-01 3.76076609e-01 -1.07952468e-01 -1.00220609e+00
-5.69921374e-01 -5.55373490e-01 -5.93638718e-01 7.30684936e-01
2.52968967e-01 -7.27554321e-01 1.25050998e+00 6.57716095e-01
-1.19731501e-01 -1.08625352e+00 -4.96638894e-01 -5.17945647e-01
-9.49774757e-02 -5.29415846e-01 7.74990439e-01 1.03035498e+00
-2.66470820e-01 6.19297206e-01 -3.04529786e-01 1.46203890e-01
7.11264074e-01 1.36681393e-01 8.19034159e-01 -1.38852096e+00
-7.53298938e-01 -3.24468553e-01 -1.32309109e-01 -1.15256941e+00
2.45555416e-01 -1.06565940e+00 -5.60084343e-01 -1.41375220e+00
-8.17892794e-03 -8.14007819e-01 -2.80774653e-01 7.14577675e-01
4.09925371e-01 1.78412020e-01 -1.76210672e-01 1.87404320e-01
-5.90118289e-01 1.89008012e-01 9.45070922e-01 -2.24361494e-01
3.16990726e-02 6.47160187e-02 -1.01651144e+00 7.00389922e-01
7.82299340e-01 -4.62480873e-01 -5.59563637e-01 -9.85752404e-01
6.82550251e-01 -1.87721893e-01 2.06877947e-01 -8.55010033e-01
4.26993459e-01 6.73426837e-02 3.67237717e-01 -3.41906965e-01
3.23922902e-01 -7.27018416e-01 1.20776355e-01 3.91122520e-01
-2.17837468e-01 2.75536120e-01 3.07645828e-01 4.17837709e-01
2.21942395e-01 -3.27178603e-03 6.51473939e-01 -1.11683905e-01
-5.37679672e-01 6.66865885e-01 -8.19947049e-02 -3.43929172e-01
5.67310452e-01 2.93570936e-01 -6.01917326e-01 -1.54643744e-01
-2.40503430e-01 1.24504253e-01 1.04585171e-01 3.25190693e-01
2.42584676e-01 -1.17661870e+00 -2.75496662e-01 3.68925720e-01
-2.73073912e-01 6.54502586e-02 1.86576709e-01 6.67172015e-01
-6.11492157e-01 7.67935336e-01 6.08945377e-02 -4.16130275e-01
-1.11582971e+00 5.90796351e-01 7.92926326e-02 -5.77836812e-01
-8.96981657e-01 9.61105525e-01 -2.07046911e-01 -3.31170529e-01
4.76563841e-01 -4.81036723e-01 2.33517319e-01 -1.32075667e-01
5.15655696e-01 3.51233840e-01 6.48383141e-01 -8.14931281e-03
-2.08247453e-02 1.25127017e-01 -3.99002284e-01 6.92751467e-01
1.58609974e+00 3.41980338e-01 -2.55574524e-01 -8.47263932e-02
1.37792218e+00 -2.57939994e-01 -8.98715079e-01 -3.75350356e-01
-2.61321127e-01 -3.41848135e-01 8.15126777e-01 -5.72858810e-01
-1.67366886e+00 7.03929186e-01 4.85809267e-01 2.12959290e-01
8.55369508e-01 -6.09228062e-03 1.27213490e+00 1.07216060e+00
7.27835059e-01 -1.13355207e+00 -4.15051281e-02 5.51918030e-01
3.51606786e-01 -1.14327061e+00 3.32873821e-01 -2.02721432e-01
-1.45650029e-01 9.46536660e-01 4.19635952e-01 -1.25769988e-01
9.12592053e-01 5.79534709e-01 -5.09931862e-01 -4.32010651e-01
-1.02410507e+00 4.11349028e-01 2.25639865e-01 2.16733515e-01
3.85538787e-01 1.86250024e-02 -1.20219365e-02 5.47194600e-01
-3.21881115e-01 6.20165747e-03 6.11240044e-02 8.10024440e-01
-3.69096011e-01 -1.04960740e+00 3.99488434e-02 1.08264494e+00
-4.17336196e-01 -4.87488449e-01 3.62406671e-01 5.90766132e-01
-2.96687365e-01 6.59765542e-01 2.90695369e-01 -5.57299972e-01
1.83024332e-01 -1.13819361e-01 2.28052586e-01 -5.11965632e-01
-7.57614315e-01 -8.90445933e-02 4.50479776e-01 -7.58323431e-01
1.79725349e-01 -3.43546689e-01 -9.87312913e-01 -1.09191585e+00
-1.20472424e-01 5.03435433e-02 1.23733246e+00 7.44853199e-01
7.20389605e-01 5.87752342e-01 3.17315876e-01 -1.19221210e+00
-7.66061425e-01 -7.50937939e-01 -6.02999210e-01 2.14692205e-02
-2.17080280e-01 -4.85967904e-01 -1.97742179e-01 -3.63020003e-01]
|
[7.13943338394165, 5.4696879386901855]
|
fdd8556c-245a-4193-912f-ab93317b4566
|
graph-neural-networks-provably-benefit-from
|
2306.13926
| null |
https://arxiv.org/abs/2306.13926v1
|
https://arxiv.org/pdf/2306.13926v1.pdf
|
Graph Neural Networks Provably Benefit from Structural Information: A Feature Learning Perspective
|
Graph neural networks (GNNs) have pioneered advancements in graph representation learning, exhibiting superior feature learning and performance over multilayer perceptrons (MLPs) when handling graph inputs. However, understanding the feature learning aspect of GNNs is still in its initial stage. This study aims to bridge this gap by investigating the role of graph convolution within the context of feature learning theory in neural networks using gradient descent training. We provide a distinct characterization of signal learning and noise memorization in two-layer graph convolutional networks (GCNs), contrasting them with two-layer convolutional neural networks (CNNs). Our findings reveal that graph convolution significantly augments the benign overfitting regime over the counterpart CNNs, where signal learning surpasses noise memorization, by approximately factor $\sqrt{D}^{q-2}$, with $D$ denoting a node's expected degree and $q$ being the power of the ReLU activation function where $q > 2$. These findings highlight a substantial discrepancy between GNNs and MLPs in terms of feature learning and generalization capacity after gradient descent training, a conclusion further substantiated by our empirical simulations.
|
['Taiji Suzuki', 'Xin Cao', 'Haonan Wang', 'Yuan Cao', 'Wei Huang']
|
2023-06-24
| null | null | null | null |
['graph-representation-learning', 'memorization']
|
['methodology', 'natural-language-processing']
|
[ 3.81011248e-01 4.32739675e-01 4.09007929e-02 -2.64357060e-01
7.04653934e-02 -3.15024287e-01 4.86996293e-01 5.48783720e-01
-5.87313235e-01 4.82742310e-01 -2.03762025e-01 -7.68891156e-01
-3.92426193e-01 -1.15478873e+00 -8.31632435e-01 -6.81891739e-01
-7.97148049e-01 -1.33582175e-01 -7.41109475e-02 -4.66082424e-01
1.21515706e-01 6.47227943e-01 -1.28927195e+00 1.89096909e-02
5.02908349e-01 1.32527363e+00 -2.65869588e-01 9.10255671e-01
-2.26829782e-01 8.87889981e-01 -5.32481313e-01 -5.19003868e-01
2.55730331e-01 -4.15316433e-01 -7.07901061e-01 -1.82883903e-01
4.35567528e-01 2.50896543e-01 -8.57690215e-01 1.09109223e+00
4.00332093e-01 2.65114337e-01 6.03984773e-01 -1.14812589e+00
-1.04801965e+00 6.49693370e-01 -2.09669024e-01 7.83809483e-01
8.16617087e-02 2.60344177e-01 1.35220706e+00 -5.67235649e-01
3.55373472e-01 1.04413319e+00 1.05202997e+00 2.89620191e-01
-1.38537908e+00 -4.74329442e-01 3.43741745e-01 -1.72659606e-01
-1.24921453e+00 -8.79889876e-02 7.81110227e-01 -3.78012061e-01
1.37319601e+00 -9.42778308e-03 7.78312087e-01 8.22545409e-01
3.42408925e-01 4.94043171e-01 8.18220496e-01 -4.78018135e-01
1.18475839e-01 -1.35553226e-01 2.28465214e-01 1.04805446e+00
5.83165705e-01 2.48195827e-01 -3.78396809e-01 1.31167158e-01
1.03853476e+00 4.54360396e-02 -1.66178837e-01 -2.71025058e-02
-5.89257360e-01 9.33518171e-01 1.10100043e+00 4.13409591e-01
-3.48679870e-01 8.26428175e-01 3.42997611e-01 8.55856359e-01
3.55485231e-01 5.60959935e-01 -4.26608264e-01 1.00471139e-01
-5.75566471e-01 -2.34344512e-01 6.66005075e-01 6.83320463e-01
1.03233254e+00 5.77086806e-01 -7.81425610e-02 4.60556984e-01
6.56289682e-02 3.10876787e-01 3.14274281e-01 -1.33718804e-01
2.94505060e-01 1.08230567e+00 -6.97545826e-01 -1.17305839e+00
-8.35344911e-01 -1.09459758e+00 -1.25034845e+00 1.72060996e-01
4.43858892e-01 -3.64344895e-01 -8.87173176e-01 1.65283251e+00
-4.64905053e-01 -8.41882750e-02 -1.53247071e-02 5.18071949e-01
8.99937510e-01 3.33642423e-01 3.51243198e-01 2.09049270e-01
1.16729105e+00 -4.56337661e-01 -1.34361953e-01 -4.85961407e-01
6.82642341e-01 -3.21140811e-02 1.04977930e+00 1.02379359e-01
-8.46083701e-01 -6.06959641e-01 -1.27600813e+00 2.03908846e-01
-7.19234407e-01 -1.51041538e-01 1.19019127e+00 1.01202619e+00
-1.36921275e+00 1.05869615e+00 -6.17035031e-01 -3.34348261e-01
8.07655931e-01 8.09405744e-01 -3.46017450e-01 1.06689677e-01
-1.25307155e+00 6.30153120e-01 3.41019183e-01 2.47452110e-01
-5.39403200e-01 -6.22050345e-01 -9.44283247e-01 4.59918231e-01
7.77604356e-02 -5.21816254e-01 7.69641101e-01 -1.07872856e+00
-1.17300224e+00 7.30066895e-01 3.95622104e-01 -7.88738191e-01
-5.20561747e-02 3.90802264e-01 -6.64386213e-01 -4.58873473e-02
-4.17161345e-01 3.61212999e-01 6.96409643e-01 -7.24703908e-01
-1.11874416e-01 -3.51755589e-01 1.26745701e-01 -1.40984938e-01
-3.72947037e-01 -2.80748129e-01 1.26111373e-01 -5.82377613e-01
3.03207010e-01 -6.97386205e-01 -5.59498012e-01 -1.69614285e-01
-2.88243920e-01 4.56963964e-02 2.77372897e-01 -2.13257894e-01
1.28211701e+00 -2.07283163e+00 -2.20862374e-01 7.37061441e-01
7.12199926e-01 3.25462461e-01 -2.35280842e-01 7.89877176e-01
-2.85007715e-01 1.74440295e-01 -1.67590529e-01 1.26641601e-01
3.80067937e-02 1.86476633e-01 5.79564832e-02 5.52053988e-01
6.89824343e-01 1.33954501e+00 -9.00687456e-01 1.48039311e-01
-3.92300747e-02 5.78518808e-01 -4.54824507e-01 -2.71632701e-01
-2.20443204e-01 -8.66624620e-03 -3.55737001e-01 6.05069280e-01
4.32623118e-01 -7.99563766e-01 4.95601147e-01 6.88228682e-02
1.43841773e-01 1.70913294e-01 -1.03425169e+00 1.18889356e+00
-2.55533248e-01 9.79781330e-01 -1.46640509e-01 -1.36542606e+00
1.16381860e+00 6.88156039e-02 2.17793986e-01 -1.05299890e+00
3.58053267e-01 8.63801837e-02 4.81819630e-01 -1.25199333e-01
3.74870986e-01 -2.81944364e-01 -3.27379778e-02 3.87373954e-01
4.56270188e-01 3.26170772e-01 1.74989119e-01 3.21635194e-02
1.52505696e+00 -5.14631987e-01 1.51750848e-01 -4.07842904e-01
2.77347714e-01 -3.23997855e-01 5.51242977e-02 1.05625927e+00
-2.10724145e-01 2.06250444e-01 1.22844195e+00 -6.60237908e-01
-6.65363014e-01 -9.68198061e-01 2.94655887e-03 1.37664592e+00
-1.96440443e-01 -3.30182493e-01 -4.32305813e-01 -6.41641438e-01
2.01168656e-01 1.89876139e-01 -9.84095991e-01 -5.31281412e-01
-5.33688247e-01 -1.03718793e+00 7.79848397e-01 7.34321117e-01
4.76371408e-01 -1.38456750e+00 -5.69214761e-01 2.01539502e-01
8.32338512e-01 -8.16341460e-01 -9.35518444e-02 8.27452838e-01
-9.45992231e-01 -1.21156538e+00 -4.47350174e-01 -9.81475472e-01
1.03855240e+00 -5.26270568e-02 1.29023373e+00 5.60602903e-01
-3.93245012e-01 3.72891068e-01 -1.37910187e-01 -2.72208184e-01
-3.94469053e-01 3.84398460e-01 -8.58420283e-02 -2.57696837e-01
3.65890324e-01 -1.05040371e+00 -6.56905651e-01 -1.46846682e-01
-1.03096294e+00 -3.15740794e-01 1.03865457e+00 8.56562734e-01
2.00729862e-01 1.51810214e-01 5.95359564e-01 -1.09125268e+00
1.08081806e+00 -5.03802538e-01 -6.46995842e-01 -6.85947482e-03
-9.65590835e-01 4.60767686e-01 8.71089637e-01 -2.60736763e-01
-2.77579963e-01 -2.72461653e-01 -1.51134059e-01 -1.82033889e-02
5.05223647e-02 9.03584063e-01 1.95066914e-01 -7.46419013e-01
9.06510830e-01 2.42396295e-01 4.84588556e-02 -2.10560724e-01
3.64371389e-01 -6.98800310e-02 3.60928357e-01 -4.00965273e-01
7.25452840e-01 6.53488487e-02 4.29573298e-01 -7.61747360e-01
-3.43562424e-01 -4.99669462e-02 -4.89267468e-01 -1.23819329e-01
3.79385144e-01 -5.81834197e-01 -1.11696601e+00 3.33844006e-01
-6.67153895e-01 -5.05860865e-01 -4.44159836e-01 3.38601351e-01
-2.04425827e-01 1.91555284e-02 -7.64667928e-01 -7.53387749e-01
-4.00748968e-01 -7.41476476e-01 3.71109694e-01 3.88699681e-01
2.07414880e-01 -1.36630869e+00 -2.74859607e-01 -4.42300767e-01
8.34477007e-01 5.50765514e-01 1.38435304e+00 -8.91593397e-01
-2.92096138e-01 -4.90778923e-01 -5.66626072e-01 4.11798358e-01
-1.41918421e-01 -2.48622015e-01 -8.99079025e-01 -3.47964466e-01
-4.72536564e-01 -8.96202475e-02 1.04479241e+00 5.01722336e-01
8.18695962e-01 -2.80405253e-01 -1.04777180e-01 5.84173143e-01
1.70605242e+00 -7.77327195e-02 5.24164736e-01 1.58654988e-01
5.29664457e-01 1.06627889e-01 -5.65931737e-01 2.67032951e-01
1.33982345e-01 -2.21361574e-02 7.09759116e-01 -3.77939552e-01
-2.30581552e-01 -3.74071091e-01 7.89111555e-02 4.80037481e-01
-1.23031050e-01 -2.69029468e-01 -1.07356572e+00 2.81471103e-01
-1.36197531e+00 -5.82140028e-01 -4.75009568e-02 2.03625870e+00
3.45779419e-01 6.15026593e-01 1.34752542e-01 2.35934958e-01
4.64461863e-01 2.95733482e-01 -5.13523281e-01 -6.63608074e-01
-3.56261641e-01 7.74602592e-01 7.78393447e-01 3.08105528e-01
-9.08372700e-01 8.17919374e-01 6.54483318e+00 3.91018629e-01
-1.26029730e+00 -3.79698217e-01 7.38297224e-01 2.90729314e-01
-3.26452971e-01 -1.41807199e-01 -2.35448077e-01 1.05699442e-01
1.20213342e+00 1.38613105e-01 6.59021020e-01 6.41372502e-01
-3.18321228e-01 1.49091020e-01 -1.13241756e+00 7.21161246e-01
-2.16215447e-01 -1.41200876e+00 5.36122406e-03 1.82551906e-01
5.43387949e-01 3.52453262e-01 4.23313349e-01 7.51599669e-01
3.71755123e-01 -1.68112791e+00 2.88022131e-01 3.84507447e-01
7.50427485e-01 -6.11799181e-01 7.01773465e-01 6.11647666e-02
-1.30575037e+00 -3.81216705e-01 -4.40136552e-01 -5.97263396e-01
-4.02025700e-01 5.58940113e-01 -9.40517306e-01 4.96284336e-01
4.26718593e-01 3.81594151e-01 -8.01222920e-01 8.25129628e-01
-1.49110049e-01 6.74164772e-01 -1.77265376e-01 -3.62162799e-01
5.94961107e-01 -1.62262321e-01 1.69696137e-01 1.39530671e+00
4.99377809e-02 4.46236990e-02 2.31105066e-03 1.00246024e+00
-4.46570367e-01 8.27069581e-02 -4.81304407e-01 -5.90156674e-01
1.55373380e-01 1.16733372e+00 -1.12399507e+00 3.74903008e-02
-4.43211287e-01 7.03127384e-01 6.48085475e-01 4.39114720e-01
-4.01479095e-01 -8.68491411e-01 4.93354797e-01 2.48357862e-01
4.95811045e-01 -3.41642439e-01 -3.31563175e-01 -6.49127066e-01
-4.16878425e-03 -4.20052826e-01 4.37749803e-01 -2.55833179e-01
-1.20708477e+00 7.35604465e-01 -3.71686727e-01 -6.91440105e-01
-1.07835583e-01 -1.16024303e+00 -7.54002690e-01 8.37150455e-01
-1.55727053e+00 -1.10269952e+00 -2.11174071e-01 4.34153259e-01
-3.83116901e-01 -1.25050366e-01 8.75110388e-01 3.26552480e-01
-6.27907693e-01 9.01097476e-01 -2.30570752e-02 6.16309464e-01
-2.46384040e-01 -1.27206099e+00 8.25587690e-01 6.63339138e-01
2.21811667e-01 6.97774768e-01 4.53520864e-01 -3.35342914e-01
-1.62727940e+00 -8.20880234e-01 6.90502286e-01 -6.87276898e-03
8.36424470e-01 -6.20108962e-01 -8.07603359e-01 6.24360025e-01
-6.38426691e-02 6.01393223e-01 7.01896846e-01 3.31967324e-01
-5.57066560e-01 1.87147148e-02 -1.13994646e+00 4.88291562e-01
1.28471947e+00 -7.20001578e-01 -3.43201458e-02 3.13061066e-02
4.84607846e-01 -1.21697001e-01 -9.28258300e-01 2.44116694e-01
4.24852431e-01 -1.00151825e+00 8.10838461e-01 -9.26260114e-01
1.01105899e-01 1.91679895e-01 -2.43910626e-01 -1.28308177e+00
-4.33946311e-01 -6.64243400e-01 -1.40424982e-01 5.92920303e-01
7.33946264e-01 -8.49153578e-01 1.06816852e+00 3.91236693e-01
-2.15645671e-01 -9.73193228e-01 -8.94249201e-01 -6.65129483e-01
3.01295310e-01 -4.58144248e-01 4.06637251e-01 7.58613050e-01
5.45273758e-02 2.93197513e-01 1.77880526e-01 1.21719591e-01
9.65743586e-02 -1.76586524e-01 2.36368403e-01 -1.48003280e+00
-4.05883461e-01 -9.27918494e-01 -9.46951032e-01 -7.66427398e-01
1.14979006e-01 -1.32638693e+00 -3.83564681e-01 -1.39788878e+00
-2.47540578e-01 -3.63492548e-01 -7.54408538e-01 6.26870096e-01
8.55500102e-02 2.77742654e-01 6.32210895e-02 -3.68265718e-01
-5.18637955e-01 2.86557168e-01 1.01176178e+00 -1.15580566e-01
-2.44907171e-01 1.74488589e-01 -1.08343983e+00 4.51034218e-01
7.11503625e-01 -2.64349163e-01 -3.67339820e-01 -2.63008744e-01
7.59160697e-01 -8.81849155e-02 5.57065189e-01 -1.22950304e+00
2.56868094e-01 2.55529016e-01 6.20760620e-01 1.38497189e-01
-2.65062507e-02 -5.46826720e-01 -7.51340389e-02 7.01235771e-01
-3.58106285e-01 3.33420962e-01 5.27206302e-01 7.21012294e-01
3.36098224e-02 -6.82363734e-02 4.89541799e-01 -3.15314919e-01
-6.26022220e-01 4.98509526e-01 -4.59667146e-01 3.48549075e-02
3.89107466e-01 -4.54499096e-01 -4.18432057e-01 -4.67260838e-01
-7.41242886e-01 -7.73825571e-02 2.24892218e-02 1.76973585e-02
6.10106647e-01 -1.09091949e+00 -5.15159726e-01 6.77236319e-01
1.57602830e-03 -3.09781641e-01 1.56358376e-01 8.47098351e-01
-4.83052135e-01 3.95039707e-01 -2.73437381e-01 -1.99536249e-01
-5.22908211e-01 3.92403692e-01 5.68030655e-01 -3.88684034e-01
-5.88826835e-01 1.24078727e+00 -1.97972983e-01 -3.45955014e-01
3.34913820e-01 -5.77342868e-01 2.16987096e-02 -2.05603480e-01
8.35065469e-02 2.81505287e-01 4.01676536e-01 -2.95692176e-01
-2.71262020e-01 2.18491375e-01 -5.41449711e-02 4.30097878e-01
1.31706953e+00 3.18127990e-01 -9.66030825e-03 2.66602010e-01
1.35070205e+00 -3.37932408e-01 -1.22764027e+00 -3.91136885e-01
3.45167279e-01 2.23677203e-01 9.07958834e-04 -7.60975242e-01
-1.26589286e+00 8.39421034e-01 5.90117037e-01 7.92439878e-01
9.29061711e-01 3.46803851e-02 4.72230613e-01 6.20087683e-01
2.59646717e-02 -8.33181322e-01 2.19981343e-01 6.60757303e-01
5.62068462e-01 -1.07707047e+00 -9.69055966e-02 -8.59409273e-02
-1.64423630e-01 1.29555893e+00 4.59247887e-01 -6.12891257e-01
1.02171087e+00 9.80913043e-02 -3.04957002e-01 -7.67220378e-01
-4.85738814e-01 -3.75277519e-01 4.18297261e-01 6.60186529e-01
5.97444534e-01 2.04792783e-01 5.00916094e-02 7.63560712e-01
-3.82755309e-01 -3.36402088e-01 3.27137947e-01 9.05278385e-01
-4.92304862e-01 -7.79771745e-01 3.22714508e-01 8.33992481e-01
-4.13916022e-01 -4.83044744e-01 -3.28371137e-01 1.03188896e+00
-9.79947150e-02 7.27018714e-01 2.46871755e-01 -5.57286024e-01
4.18058157e-01 1.10301547e-01 4.98669803e-01 -5.65123141e-01
-9.84647453e-01 -5.49472868e-01 -6.76916912e-02 -2.51241535e-01
1.49529194e-02 -7.70658627e-02 -1.40150690e+00 -4.55145419e-01
-3.18116099e-01 3.21721211e-02 6.46694660e-01 7.30236650e-01
3.31699610e-01 8.79246831e-01 4.27269757e-01 -5.77350318e-01
-4.52156037e-01 -8.48140657e-01 -7.84967363e-01 1.44646004e-01
4.80907261e-01 -3.76837105e-01 -3.75105530e-01 -5.71617782e-01]
|
[6.860459804534912, 6.098934173583984]
|
2e309c82-e39a-43de-aae2-e9ef7f73cb7f
|
molweni-a-challenge-multiparty-dialogues
|
2004.05080
| null |
https://arxiv.org/abs/2004.05080v3
|
https://arxiv.org/pdf/2004.05080v3.pdf
|
Molweni: A Challenge Multiparty Dialogues-based Machine Reading Comprehension Dataset with Discourse Structure
|
Research into the area of multiparty dialog has grown considerably over recent years. We present the Molweni dataset, a machine reading comprehension (MRC) dataset with discourse structure built over multiparty dialog. Molweni's source samples from the Ubuntu Chat Corpus, including 10,000 dialogs comprising 88,303 utterances. We annotate 30,066 questions on this corpus, including both answerable and unanswerable questions. Molweni also uniquely contributes discourse dependency annotations in a modified Segmented Discourse Representation Theory (SDRT; Asher et al., 2016) style for all of its multiparty dialogs, contributing large-scale (78,245 annotated discourse relations) data to bear on the task of multiparty dialog discourse parsing. Our experiments show that Molweni is a challenging dataset for current MRC models: BERT-wwm, a current, strong SQuAD 2.0 performer, achieves only 67.7% F1 on Molweni's questions, a 20+% significant drop as compared against its SQuAD 2.0 performance.
|
['Ting Liu', 'Zekun Wang', 'Zihao Zheng', 'Min-Yen Kan', 'Jiaqi Li', 'Bing Qin', 'Wenqiang Lei', 'Ming Liu']
|
2020-04-10
| null |
https://aclanthology.org/2020.coling-main.238
|
https://aclanthology.org/2020.coling-main.238.pdf
|
coling-2020-8
|
['dialogue-understanding']
|
['natural-language-processing']
|
[ 1.03413045e-01 1.15257311e+00 1.04655838e-02 -5.73141932e-01
-1.26222610e+00 -1.09310687e+00 7.67784655e-01 2.42492765e-01
-2.17193022e-01 1.04549444e+00 1.07012141e+00 -6.65505290e-01
1.76543176e-01 -3.87188584e-01 -3.51537913e-01 -2.03950480e-02
1.79758787e-01 1.22843552e+00 4.34813499e-01 -8.16800892e-01
6.32321835e-02 -5.23525178e-01 -7.80929923e-01 9.31870997e-01
7.42604136e-01 6.30517185e-01 1.02543242e-01 1.35208821e+00
-1.10372394e-01 1.88121283e+00 -1.23768449e+00 -8.75268579e-01
-5.47801852e-01 -4.63441044e-01 -2.02786326e+00 -3.35392684e-01
5.66554427e-01 -6.11310720e-01 -2.77028203e-01 2.71383107e-01
3.36127967e-01 2.47750744e-01 5.64158797e-01 -9.47166145e-01
-6.95363164e-01 1.24280715e+00 7.64536709e-02 3.69916677e-01
7.68708110e-01 1.24337398e-01 1.60647345e+00 -5.70882261e-01
1.05218208e+00 1.76223230e+00 3.33835572e-01 8.19932878e-01
-1.01345670e+00 -1.28529668e-01 -1.51519358e-01 2.58112848e-01
-4.49569702e-01 -7.33939826e-01 4.37966317e-01 -3.99503350e-01
1.39316285e+00 6.43985033e-01 -9.48661342e-02 1.32405388e+00
-2.72655398e-01 1.26648962e+00 1.00855803e+00 -3.32674593e-01
-3.19784172e-02 8.59499257e-03 8.04766476e-01 4.09295499e-01
-6.55387104e-01 -5.64502478e-01 -6.40078902e-01 -3.38225424e-01
1.04375072e-01 -9.48624253e-01 -3.85190547e-01 4.04176772e-01
-1.06776702e+00 1.18503094e+00 3.42723668e-01 2.28481039e-01
1.52859032e-01 -3.72263134e-01 6.72672749e-01 6.82142258e-01
3.75760466e-01 8.48199785e-01 -5.47234416e-01 -8.26025486e-01
-3.53322662e-02 6.44043982e-01 1.65275848e+00 1.16512084e+00
1.98533222e-01 -5.68775237e-01 -5.30242503e-01 1.16897213e+00
2.58510202e-01 2.44651452e-01 3.29524398e-01 -1.74217331e+00
1.23977506e+00 6.60573840e-01 1.14747897e-01 -5.40326297e-01
-5.37630916e-01 5.77179492e-01 -3.79873484e-01 -3.52187932e-01
1.12891555e+00 -6.29811227e-01 4.03191224e-02 1.64350331e+00
3.41478169e-01 -8.62444639e-01 6.05506957e-01 6.82392240e-01
1.70857382e+00 9.47344184e-01 1.89142764e-01 -1.14744991e-01
1.67330611e+00 -1.39429057e+00 -9.16224182e-01 -3.99745375e-01
7.26134300e-01 -1.01110637e+00 1.04441547e+00 2.36948654e-01
-1.51378810e+00 -1.51100054e-01 -6.27862275e-01 -9.02063966e-01
-8.03959072e-02 -1.30032867e-01 4.71583933e-01 4.17990148e-01
-9.75306571e-01 6.85133738e-03 -3.00403178e-01 -1.45961389e-01
1.11548223e-01 2.66384352e-02 5.27913868e-02 9.36785936e-02
-1.35087311e+00 1.46961999e+00 1.57328799e-01 -2.77766943e-01
-8.41644704e-01 -6.60025120e-01 -8.65960419e-01 6.42363280e-02
5.85864961e-01 -2.22079307e-02 2.24714589e+00 -3.28342348e-01
-1.89487588e+00 1.02699244e+00 -2.83364117e-01 -7.03894436e-01
5.25793970e-01 -3.25631320e-01 9.97457281e-02 2.44765669e-01
8.99342895e-02 8.09377134e-01 1.96746379e-01 -8.98141205e-01
-5.54934621e-01 -2.36258283e-01 6.63627803e-01 4.50189501e-01
9.55148488e-02 4.40196395e-01 7.25015327e-02 -1.27703100e-01
-4.04767215e-01 -8.67021799e-01 2.93430567e-01 -6.31511688e-01
-6.23309374e-01 -1.01306522e+00 9.57560658e-01 -1.12108135e+00
1.11760652e+00 -1.66894698e+00 5.26730597e-01 -7.24914968e-01
5.68917274e-01 3.30547512e-01 -8.81017838e-03 8.05923522e-01
4.08086509e-01 -5.35495691e-02 -4.63788629e-01 -6.70978129e-01
2.42156774e-01 4.33728546e-01 -5.59656143e-01 -1.29897594e-01
2.34929293e-01 1.10568511e+00 -7.60834277e-01 -5.53403497e-01
1.48114577e-01 -6.41685724e-02 -5.03265977e-01 9.03722703e-01
-9.24869239e-01 7.07301140e-01 -2.94651359e-01 3.62150937e-01
1.12906277e-01 -2.99720466e-01 5.18561065e-01 2.56008238e-01
-2.06581339e-01 1.05415559e+00 -2.38062754e-01 1.65258765e+00
-5.30246139e-01 1.19150794e+00 7.23202944e-01 -4.47898686e-01
8.34277511e-01 6.98851228e-01 -6.86458647e-02 -3.59812438e-01
5.65134048e-01 1.75752178e-01 2.94755697e-01 -5.39472640e-01
9.24436748e-01 2.41892025e-01 -5.42274594e-01 6.56544864e-01
2.47546300e-01 -3.24044436e-01 3.74710232e-01 7.06430554e-01
1.30584228e+00 -3.62169921e-01 2.93267459e-01 -3.45152766e-01
6.88932538e-01 5.43176770e-01 9.91690084e-02 6.09246135e-01
-4.05595660e-01 3.95490289e-01 1.08275044e+00 -1.16047300e-01
-7.41390288e-01 -9.31247771e-01 -2.69783258e-01 1.75550985e+00
-1.61819801e-01 -3.21557373e-01 -8.90692830e-01 -7.15097964e-01
-1.86102435e-01 9.51287925e-01 -4.67223644e-01 5.67398965e-01
-1.18111336e+00 -4.84976560e-01 9.14767802e-01 3.69561732e-01
8.23272765e-01 -1.44811261e+00 -6.40295327e-01 3.92577499e-01
-8.64098191e-01 -1.23440027e+00 -3.18502188e-01 2.03698482e-02
-3.82799238e-01 -1.46100152e+00 -5.74179888e-01 -8.75526130e-01
-1.68498963e-01 -1.27287777e-02 1.76150548e+00 9.07639042e-02
-1.84150618e-02 3.95034790e-01 -6.97666705e-01 -2.83822119e-01
-1.33904755e+00 4.83066380e-01 -7.48082280e-01 -7.43537366e-01
3.72073442e-01 -2.20164508e-01 -2.10674196e-01 4.93789315e-01
-2.33384624e-01 9.80182439e-02 2.45914925e-02 9.62442935e-01
-3.43479872e-01 -9.84414935e-01 9.95544434e-01 -1.28500700e+00
1.10167265e+00 -7.18512654e-01 -3.57150078e-01 4.20472443e-01
7.20914975e-02 -3.11324567e-01 4.83410925e-01 -1.05031319e-01
-1.64404714e+00 -5.57327867e-01 -5.65097868e-01 4.40796256e-01
-1.89265788e-01 2.04216585e-01 -9.67134610e-02 5.99620283e-01
7.92159915e-01 -4.00531828e-01 1.43810838e-01 -6.01748824e-01
8.20366383e-01 1.16841948e+00 8.05813313e-01 -9.70542669e-01
7.89435059e-02 -1.70297444e-01 -6.92346096e-01 -1.12828183e+00
-1.21497190e+00 -5.96848428e-01 -5.79115987e-01 -1.91966817e-01
1.28880703e+00 -8.38272691e-01 -1.46512771e+00 2.58711040e-01
-1.74667537e+00 -1.06662536e+00 -5.26711382e-02 -1.64985523e-01
-5.66587448e-01 2.58840144e-01 -1.35586464e+00 -1.04527915e+00
-5.86343884e-01 -1.05198264e+00 9.07533467e-01 1.71759591e-01
-1.05254841e+00 -1.17188704e+00 3.28852117e-01 1.30456126e+00
3.96678984e-01 1.33664280e-01 1.25785899e+00 -1.18181396e+00
-3.59478682e-01 3.65810633e-01 -3.51519078e-01 1.60719380e-01
-1.59938619e-01 -2.82943219e-01 -1.15813446e+00 -1.20936416e-01
2.44318023e-01 -1.35547948e+00 4.63196695e-01 -8.40401798e-02
4.54958677e-01 -6.86338007e-01 2.93897778e-01 -4.26194638e-01
5.58614492e-01 3.01982552e-01 4.14463788e-01 1.91741437e-01
5.49795389e-01 1.15292680e+00 5.11981905e-01 2.13670209e-01
1.09605956e+00 5.12061477e-01 3.28756124e-01 5.73746443e-01
-3.03434491e-01 -9.84210894e-02 3.17933708e-01 1.28212488e+00
2.27392793e-01 -6.20740771e-01 -1.23551142e+00 7.18374789e-01
-2.00504899e+00 -7.46514618e-01 -4.80841577e-01 1.35449743e+00
1.44263256e+00 -8.55921730e-02 2.86339343e-01 -2.54150867e-01
5.41911483e-01 4.85985458e-01 -5.73640049e-01 -7.92857409e-01
-2.37695903e-01 6.05337322e-02 -1.44114599e-01 1.20774305e+00
-9.99848306e-01 9.60778415e-01 5.79924917e+00 3.89502466e-01
-2.00410590e-01 3.82427335e-01 8.12859058e-01 2.17293248e-01
-1.16288789e-01 -3.00939698e-02 -8.50869179e-01 2.03061894e-01
1.36173606e+00 -1.75525710e-01 3.51944804e-01 8.27926815e-01
-4.09264743e-01 -3.58461678e-01 -1.33477807e+00 3.45832527e-01
1.58547655e-01 -1.47772992e+00 -3.35295051e-01 -2.49910370e-01
5.11629879e-01 5.58754243e-02 -2.69649357e-01 9.26121473e-01
9.88891602e-01 -1.23965037e+00 4.13834870e-01 -1.40467614e-01
3.23067009e-01 -3.58312249e-01 7.49849439e-01 7.23515809e-01
-5.59475601e-01 -3.54490727e-02 -1.48191974e-01 -4.00744587e-01
4.40298438e-01 -3.12610656e-01 -1.33538282e+00 1.85582817e-01
4.50792313e-01 3.63662720e-01 -4.31934208e-01 -1.00467745e-02
-4.10976499e-01 1.06172502e+00 -1.80029437e-01 -5.00568509e-01
3.30766767e-01 5.67917116e-02 6.91811979e-01 1.20489776e+00
-6.62810445e-01 7.02538252e-01 1.12436138e-01 6.20531499e-01
-5.09165943e-01 -1.18006796e-01 -2.79796243e-01 1.35418728e-01
7.90378571e-01 1.06378889e+00 -1.42514393e-01 -5.45836091e-01
-3.01467836e-01 6.73607588e-01 6.82874382e-01 -4.78867218e-02
-4.75778788e-01 -1.88458756e-01 4.46873456e-01 -4.31943148e-01
-1.46417901e-01 -1.48456454e-01 -3.26463103e-01 -8.86329353e-01
-2.00350419e-01 -1.30058753e+00 7.21564531e-01 -6.68709934e-01
-1.48432708e+00 7.53740966e-01 1.13952421e-01 -3.53795260e-01
-7.48999178e-01 -6.84764206e-01 -8.06042314e-01 7.42049813e-01
-1.23836243e+00 -1.16343594e+00 -2.98249215e-01 4.79158014e-01
1.32042873e+00 -8.45065415e-02 1.13493276e+00 7.84224942e-02
-4.89564240e-01 3.90908003e-01 -8.04262981e-02 5.91769814e-01
9.22216833e-01 -1.69492888e+00 5.33284068e-01 2.02085767e-02
-2.75775462e-01 4.71352488e-01 6.83109343e-01 -4.33385849e-01
-1.29089153e+00 -6.77503884e-01 1.26871574e+00 -1.37266052e+00
1.03351164e+00 -3.23241055e-01 -1.24684989e+00 1.14518356e+00
1.13678586e+00 -1.08683836e+00 7.80793607e-01 3.88168961e-01
-2.63817310e-01 6.55138254e-01 -1.11370385e+00 6.42756522e-01
7.40962863e-01 -5.59212089e-01 -1.47439063e+00 7.64636159e-01
1.29918551e+00 -1.02261424e+00 -1.40632153e+00 1.74688905e-01
1.83606565e-01 -9.75710034e-01 7.11743414e-01 -8.13646555e-01
1.00066674e+00 4.99884516e-01 -3.42730671e-01 -9.54843044e-01
4.44478631e-01 -1.00867867e+00 -3.40431005e-01 1.58469391e+00
7.10672438e-01 -4.02917534e-01 5.87964892e-01 1.26571310e+00
-4.74721342e-01 -6.11845911e-01 -1.22588837e+00 -2.12482885e-01
8.48364949e-01 -9.97866988e-02 3.79977584e-01 8.53397906e-01
7.52696097e-01 1.24509954e+00 -2.12841332e-01 -4.45209563e-01
2.75225371e-01 2.62042254e-01 1.07233393e+00 -1.08212614e+00
-4.17821348e-01 -3.57549548e-01 4.62728411e-01 -1.64003539e+00
3.66248876e-01 -7.33045042e-01 2.02646896e-01 -1.52928090e+00
-8.23896527e-02 -2.63913989e-01 8.92822385e-01 4.08480823e-01
-1.73515186e-01 -1.83692247e-01 4.57790613e-01 4.19823408e-01
-8.96933079e-01 7.15986013e-01 1.30157757e+00 -3.36234331e-01
-2.32072204e-01 5.19341640e-02 -6.92902863e-01 8.61349404e-01
6.86364353e-01 3.22355032e-02 -3.10785145e-01 -7.40094841e-01
-2.74726957e-01 9.64042544e-01 7.55416462e-03 -3.83428663e-01
3.32855701e-01 1.45813420e-01 -3.51159275e-01 -7.04192579e-01
7.74671435e-01 -1.39316812e-01 -7.11553812e-01 2.49877498e-01
-1.12789416e+00 7.29653537e-02 1.75664499e-01 2.40068525e-01
-4.12863284e-01 -5.03238499e-01 5.87488055e-01 -2.46408761e-01
-3.89603972e-01 -5.73566496e-01 -7.22894788e-01 1.11765039e+00
7.28250444e-01 5.63667655e-01 -1.37440324e+00 -8.54201317e-01
-7.73463607e-01 7.93152153e-01 -8.73464420e-02 5.14881134e-01
1.53378502e-01 -5.70177495e-01 -9.84554768e-01 -5.93726516e-01
-1.72693387e-01 4.34823304e-01 2.27065459e-01 3.79368007e-01
-4.08235162e-01 7.56800592e-01 2.82793045e-02 -4.11674440e-01
-1.51600528e+00 -1.46160036e-01 -1.17366724e-01 -6.86389744e-01
-5.77034593e-01 1.06784201e+00 -8.29252377e-02 -7.86191106e-01
3.31577927e-01 -2.10443333e-01 -5.55134356e-01 3.96143705e-01
6.41481340e-01 6.46988869e-01 -1.80924773e-01 -5.78980505e-01
-1.32798970e-01 -2.02204809e-01 -2.69266546e-01 -5.15004098e-01
1.06576455e+00 -4.43129897e-01 -3.34280998e-01 6.82577848e-01
1.24465048e+00 -1.28649138e-02 -9.49971735e-01 -3.00675660e-01
3.06550980e-01 2.07153689e-02 -7.30119824e-01 -1.28173816e+00
-8.83992389e-02 8.93851995e-01 -2.84961522e-01 7.33030021e-01
3.34804416e-01 5.21438479e-01 1.17131078e+00 9.81032133e-01
2.19166607e-01 -8.81402075e-01 1.91879511e-01 1.37094140e+00
1.35235357e+00 -1.49658537e+00 -3.01264107e-01 -4.53666985e-01
-1.12696207e+00 9.60956991e-01 9.80295181e-01 6.88587874e-02
1.35849833e-01 -7.09690303e-02 4.04543549e-01 -4.26154643e-01
-1.19322741e+00 2.58730412e-01 -1.05703995e-01 4.36771870e-01
7.40220428e-01 1.04286611e-01 -1.50031358e-01 7.47461319e-01
-8.97857845e-01 -8.05939078e-01 7.85259008e-01 7.87506104e-01
-5.74608862e-01 -9.08180118e-01 -2.37047732e-01 1.20568790e-01
-2.56637543e-01 -2.25949064e-02 -1.00770962e+00 9.68048930e-01
-8.21161687e-01 1.77127397e+00 1.55945897e-01 5.32828383e-02
4.29182172e-01 2.89883643e-01 1.06199861e-01 -8.12125087e-01
-1.29883838e+00 -4.71823633e-01 1.22684085e+00 -2.50423521e-01
-3.74728233e-01 -6.50451958e-01 -1.26782310e+00 -5.88477254e-01
-5.72792254e-02 6.23915672e-01 3.50006402e-01 9.73983169e-01
1.00926898e-01 4.95561153e-01 3.46578956e-01 -4.85194206e-01
-6.82751060e-01 -1.54251909e+00 2.35366926e-01 3.21752220e-01
3.05416673e-01 -2.93397665e-01 -3.49503130e-01 5.10459952e-03]
|
[12.28313159942627, 8.05882453918457]
|
f3ae4eb1-8584-4a83-87b0-476b7347022a
|
energy-efficient-wearable-to-mobile-offload
|
2306.06129
| null |
https://arxiv.org/abs/2306.06129v1
|
https://arxiv.org/pdf/2306.06129v1.pdf
|
Energy-efficient Wearable-to-Mobile Offload of ML Inference for PPG-based Heart-Rate Estimation
|
Modern smartwatches often include photoplethysmographic (PPG) sensors to measure heartbeats or blood pressure through complex algorithms that fuse PPG data with other signals. In this work, we propose a collaborative inference approach that uses both a smartwatch and a connected smartphone to maximize the performance of heart rate (HR) tracking while also maximizing the smartwatch's battery life. In particular, we first analyze the trade-offs between running on-device HR tracking or offloading the work to the mobile. Then, thanks to an additional step to evaluate the difficulty of the upcoming HR prediction, we demonstrate that we can smartly manage the workload between smartwatch and smartphone, maintaining a low mean absolute error (MAE) while reducing energy consumption. We benchmark our approach on a custom smartwatch prototype, including the STM32WB55 MCU and Bluetooth Low-Energy (BLE) communication, and a Raspberry Pi3 as a proxy for the smartphone. With our Collaborative Heart Rate Inference System (CHRIS), we obtain a set of Pareto-optimal configurations demonstrating the same MAE as State-of-Art (SoA) algorithms while consuming less energy. For instance, we can achieve approximately the same MAE of TimePPG-Small (5.54 BPM MAE vs. 5.60 BPM MAE) while reducing the energy by 2.03x, with a configuration that offloads 80\% of the predictions to the phone. Furthermore, accepting a performance degradation to 7.16 BPM of MAE, we can achieve an energy consumption of 179 uJ per prediction, 3.03x less than running TimePPG-Small on the smartwatch, and 1.82x less than streaming all the input data to the phone.
|
['Daniele Jahier Pagliari', 'Massimo Poncino', 'Enrico Macii', 'Luca Benini', 'Yukai Chen', 'Noemi Tomasello', 'Matteo Risso', 'Alessio Burrello']
|
2023-06-08
| null | null | null | null |
['heart-rate-estimation']
|
['medical']
|
[ 1.42632112e-01 3.54431391e-01 -4.02996317e-02 -2.36966178e-01
-5.56595683e-01 -4.41974014e-01 -2.48743519e-01 -3.02546378e-02
-2.55414635e-01 8.17967892e-01 -1.90265104e-01 -5.15324652e-01
-6.35605380e-02 -9.21695411e-01 -3.84150684e-01 -6.65467620e-01
1.01261608e-01 1.62321374e-01 -1.30361840e-01 3.91289353e-01
-8.72237086e-02 2.53860299e-02 -1.59736717e+00 -2.98463702e-01
1.01913691e+00 1.52885044e+00 -2.18024030e-01 1.25767124e+00
5.39898574e-01 4.92535323e-01 -7.67327130e-01 -7.57051483e-02
8.60935450e-02 -7.65576839e-01 -1.34595633e-01 -4.96968657e-01
2.26385728e-01 -3.80347967e-01 -1.51704764e-02 3.56809855e-01
9.52317655e-01 -1.30509548e-02 1.72811951e-02 -1.38892019e+00
5.29829144e-01 6.41253293e-01 -2.47726113e-01 6.97448403e-02
3.47494513e-01 3.71660680e-01 5.34986734e-01 3.88790015e-03
-6.47732243e-02 4.20444846e-01 1.09891546e+00 4.53561187e-01
-1.45633900e+00 -6.59157872e-01 -4.74773496e-01 2.04557657e-01
-1.77822304e+00 -7.14325726e-01 7.04604864e-01 7.55184144e-02
9.26448464e-01 1.05279028e+00 1.17594540e+00 6.24902248e-01
5.54863691e-01 -2.06976179e-02 1.17056739e+00 -3.03722471e-01
5.40390193e-01 2.28086457e-01 4.75304164e-02 4.96095181e-01
3.22630793e-01 -2.70721391e-02 -5.46217918e-01 -5.66646993e-01
3.56408983e-01 1.13369584e-01 -2.77469605e-01 3.09862107e-01
-1.02223289e+00 6.46145865e-02 -3.27655613e-01 7.88813457e-02
-4.37146902e-01 5.18195331e-01 -5.87220490e-02 -1.46916017e-01
1.23490177e-01 3.19185436e-01 -6.20433509e-01 -8.07990909e-01
-1.13222718e+00 -2.39244804e-01 1.59422529e+00 7.14399278e-01
5.16004682e-01 -4.41111103e-02 -4.84445900e-01 3.42104852e-01
3.87675107e-01 1.05119824e+00 2.98160255e-01 -1.43974578e+00
1.70532137e-01 4.65524763e-01 4.20868367e-01 -7.05285132e-01
-6.52994096e-01 -1.33993387e-01 -8.85560572e-01 -2.08914682e-01
4.83431548e-01 -7.46183574e-01 -1.34917021e-01 1.56389678e+00
5.22748172e-01 6.31444097e-01 -2.44727984e-01 6.02807045e-01
3.50340277e-01 5.90387940e-01 3.01810324e-01 -8.53931904e-01
1.60959578e+00 -4.69275981e-01 -6.85665369e-01 -4.01032269e-02
2.93547958e-01 -4.48593527e-01 1.12568855e+00 2.48308048e-01
-1.24516165e+00 -6.72210753e-01 -1.05828297e+00 4.66751307e-02
1.01953231e-01 2.04937365e-02 1.72698095e-01 1.43284714e+00
-8.77715409e-01 8.88163865e-01 -1.04754305e+00 -2.04994664e-01
-1.11066200e-01 4.16012853e-01 6.59395635e-01 6.59998417e-01
-1.07712924e+00 7.97018468e-01 -2.11060286e-01 -1.52036235e-01
-3.27306509e-01 -1.30654192e+00 -3.43475729e-01 5.84617496e-01
2.31316596e-01 -1.02422380e+00 8.98159862e-01 -3.89527947e-01
-2.33001423e+00 4.24496472e-01 -1.20582268e-01 -6.99545741e-01
5.94548583e-01 -2.97336459e-01 -6.00118876e-01 1.25234261e-01
-6.97586060e-01 9.46758091e-02 6.04825556e-01 -6.26033187e-01
-2.66382098e-01 -2.85629660e-01 -4.24620330e-01 9.68947634e-02
-4.62147921e-01 -6.18682146e-01 -2.01642871e-01 4.85288538e-02
-2.69852906e-01 -1.20400321e+00 -8.51325020e-02 -3.58405948e-01
-2.73542881e-01 2.73768663e-01 6.01054132e-01 -6.78227127e-01
1.79189408e+00 -1.78615725e+00 -4.08218086e-01 4.11080986e-01
1.10205248e-01 1.23513207e-01 6.74225211e-01 1.01899743e-01
4.85338598e-01 3.93202230e-02 -1.61761120e-01 -4.17155385e-01
6.97704926e-02 2.32462138e-01 6.04296550e-02 4.83660042e-01
-7.75011301e-01 1.09893978e+00 -4.69916850e-01 -5.32651961e-01
6.12644494e-01 5.14971435e-01 -4.00201172e-01 4.08381552e-01
2.05892652e-01 3.87228817e-01 3.81632410e-02 5.95762968e-01
3.23771805e-01 -2.90572107e-01 8.73552799e-01 -4.23190713e-01
-1.61810890e-01 2.83772379e-01 -1.25895941e+00 1.33621752e+00
-9.08612549e-01 3.04941952e-01 2.50018556e-02 -5.55097163e-01
9.89875913e-01 3.14128757e-01 9.58414555e-01 -6.89540684e-01
3.10154855e-01 3.29395048e-02 -2.01403975e-01 -6.19947612e-01
1.83791190e-01 -9.33317393e-02 -5.96722364e-02 7.45758832e-01
-3.58004957e-01 1.28847659e-01 -3.46359372e-01 -2.46810704e-01
1.29015231e+00 7.53207579e-02 5.56876838e-01 -4.90688384e-01
3.93366307e-01 -4.86264378e-01 5.89472175e-01 6.78612649e-01
-3.82942080e-01 6.20226599e-02 -9.32117924e-03 -1.52428597e-01
-6.62730932e-01 -9.38787580e-01 -3.07600778e-02 8.32617998e-01
1.58977434e-01 -4.93716687e-01 -8.29740822e-01 -8.48075449e-02
1.00211032e-01 9.50149000e-01 -2.18448162e-01 -2.14319557e-01
-5.75890660e-01 -8.06703687e-01 7.33736575e-01 4.01147872e-01
6.27259970e-01 -4.42389250e-01 -1.57503247e+00 3.43543738e-01
-3.89088213e-01 -9.66597319e-01 -5.58237493e-01 -6.04796335e-02
-1.13802326e+00 -8.01126540e-01 -2.36658275e-01 1.21051915e-01
2.04529598e-01 -1.99257061e-01 1.27029347e+00 -1.69822481e-02
-3.16214025e-01 8.65440249e-01 1.74631044e-01 -4.90915209e-01
-9.75717902e-02 1.56299248e-01 1.64171889e-01 -5.72925732e-02
3.43711227e-01 -9.93460059e-01 -1.32734764e+00 3.90460879e-01
8.06396548e-03 8.46343264e-02 1.00528039e-02 1.11918055e-01
7.56765425e-01 -2.60593206e-01 5.88740408e-01 -4.14091021e-01
1.97583884e-01 -4.71635580e-01 -7.61493146e-01 1.79628640e-01
-1.42029798e+00 -2.01189101e-01 7.89967000e-01 -5.63397527e-01
-7.83288896e-01 4.14969742e-01 3.14266458e-02 -2.29155242e-01
8.26160461e-02 -2.07189694e-01 -2.22636059e-01 7.77721331e-02
4.59550500e-01 3.71553272e-01 2.66348422e-01 -3.23119074e-01
2.02089280e-01 8.53482127e-01 6.63713336e-01 -4.20298368e-01
3.19003582e-01 1.76556543e-01 3.89544040e-01 -9.23695862e-01
-3.23269308e-01 -1.30364001e-01 -4.09680158e-02 -6.67402983e-01
6.70963883e-01 -8.65127206e-01 -2.00644374e+00 1.60562977e-01
-4.38712925e-01 -5.97276092e-01 -4.66407895e-01 4.87560630e-01
-6.65131330e-01 3.75798225e-01 -3.17826182e-01 -1.37497997e+00
-1.14901602e+00 -4.13879603e-01 7.86870599e-01 6.25653565e-01
-6.28305435e-01 -8.77848983e-01 9.12432745e-02 6.01712406e-01
7.24887311e-01 5.10104358e-01 3.38997006e-01 -2.55629737e-02
-3.59203190e-01 -1.64913759e-01 4.08352882e-01 1.65248007e-01
1.05699543e-02 -3.24030727e-01 -1.23486149e+00 -2.67245561e-01
2.89556652e-01 2.67904192e-01 9.60407183e-02 6.44984305e-01
1.38676739e+00 -5.68673968e-01 -4.91224229e-01 7.50802994e-01
1.65766990e+00 2.88328230e-01 9.38311756e-01 -1.62644073e-01
4.08693790e-01 5.46106733e-02 3.89500141e-01 8.78495991e-01
6.96748316e-01 7.26765990e-01 2.46849358e-01 2.58401968e-02
1.89532235e-01 -1.39922023e-01 5.76262474e-01 8.38094115e-01
-3.46962839e-01 -8.59662071e-02 -6.86113656e-01 3.94449979e-02
-1.74760795e+00 -7.64859498e-01 -2.34831467e-01 2.89232993e+00
1.06133389e+00 -1.81505218e-01 4.74609166e-01 3.87365639e-01
4.92063254e-01 -1.06826052e-01 -8.09973896e-01 -5.92863798e-01
5.77404141e-01 5.05766213e-01 9.15685773e-01 5.19129515e-01
-5.21625578e-01 9.18481275e-02 5.99748755e+00 2.48308867e-01
-1.03303421e+00 2.56955385e-01 9.05888081e-01 -7.12491333e-01
9.98724923e-02 -1.01063222e-01 -7.18307376e-01 1.17025673e+00
2.15033650e+00 -6.77510619e-01 7.43627489e-01 9.15865600e-01
6.63300335e-01 -5.18393934e-01 -1.25955856e+00 1.26847041e+00
-1.40965685e-01 -1.08816540e+00 -1.04571104e+00 1.20172769e-01
2.24146307e-01 -2.58291930e-01 -5.21814167e-01 2.28754476e-01
-2.36155555e-01 -6.42565906e-01 1.26204506e-01 1.11330080e+00
1.09622490e+00 -6.96624100e-01 5.05178273e-01 3.87449741e-01
-1.33581805e+00 -1.36831626e-02 4.67191003e-02 -2.00049818e-01
2.03320667e-01 1.11821425e+00 -6.64009571e-01 2.49495834e-01
7.22085774e-01 -9.08457488e-02 -2.28755876e-01 8.03804159e-01
6.96903467e-02 8.33723187e-01 -9.51822698e-01 -3.75534803e-01
-8.40456843e-01 -2.17143953e-01 2.58071691e-01 9.14608717e-01
5.81792474e-01 6.19600534e-01 1.98905859e-02 8.08697402e-01
-6.52489364e-02 -1.34641945e-01 -6.42165840e-02 5.04969954e-01
1.02305889e+00 1.43660188e+00 -3.51984382e-01 -6.23048544e-01
-2.45893579e-02 8.07788253e-01 -5.48379123e-01 1.38108237e-02
-1.41852665e+00 -5.74716747e-01 8.41410756e-01 3.79243523e-01
-1.57128304e-01 -5.85619640e-03 -8.96767974e-01 -8.59582663e-01
8.87474790e-02 -2.94721097e-01 2.96811372e-01 -6.04003906e-01
-7.22480774e-01 -2.72724107e-02 -2.41100088e-01 -1.13871741e+00
-2.99838960e-01 1.96100980e-01 -7.28451431e-01 8.46198022e-01
-1.18521500e+00 -3.85803610e-01 -8.13376248e-01 5.66530824e-01
6.40904605e-02 6.88066065e-01 1.02283716e+00 4.76858050e-01
-5.66765428e-01 7.32682407e-01 -7.91201741e-02 -5.39654195e-01
4.86217380e-01 -1.19693100e+00 1.26000708e-02 5.46413302e-01
-5.50140321e-01 3.99812371e-01 6.65168881e-01 -4.09427136e-01
-2.22975898e+00 -9.58083391e-01 8.77882123e-01 -5.33058167e-01
3.01709652e-01 -2.33666599e-01 -5.16836643e-01 2.02613771e-01
1.69087529e-01 1.17817789e-01 9.19025064e-01 5.91317676e-02
3.61817002e-01 -8.56282294e-01 -1.56025124e+00 4.92016882e-01
8.84592414e-01 -2.50446886e-01 -1.18597388e-01 4.72994968e-02
2.97937810e-01 -5.09276867e-01 -1.71770358e+00 1.22232795e-01
1.21036065e+00 -7.97810853e-01 8.51529717e-01 1.46987021e-01
-2.54227698e-01 -2.81019539e-01 7.69847864e-03 -1.00807393e+00
-2.49479972e-02 -1.27765584e+00 -1.04942429e+00 1.26414037e+00
8.03129822e-02 -8.98826897e-01 8.72805715e-01 1.27028906e+00
2.34281674e-01 -8.06739390e-01 -1.17834604e+00 -6.88967705e-01
-5.80516100e-01 -4.88706917e-01 6.15465641e-01 5.66645682e-01
4.51668143e-01 3.89204174e-01 -5.97415447e-01 1.48450628e-01
9.27175164e-01 3.73481393e-01 7.99284160e-01 -1.13998783e+00
-5.41940272e-01 1.10442489e-01 2.76780920e-03 -8.10870469e-01
-6.63722157e-01 -3.81806880e-01 4.85985950e-02 -1.20700765e+00
4.77781892e-02 -4.05783623e-01 -5.02501070e-01 6.77083373e-01
-1.68095842e-01 3.01843733e-01 2.88644940e-01 -1.22901544e-01
-5.43005109e-01 6.07881323e-02 5.58579624e-01 2.13619500e-01
-8.21857750e-01 2.00930744e-01 -5.29498100e-01 4.36502665e-01
1.04277039e+00 -4.21553969e-01 -4.78740782e-01 2.47394726e-01
2.91528732e-01 5.93423486e-01 2.68347889e-01 -1.39999342e+00
2.75917530e-01 -5.30534387e-02 4.65481430e-01 -2.95839041e-01
3.71787161e-01 -1.09136450e+00 8.59102190e-01 9.18997228e-01
4.94202189e-02 -3.45975250e-01 1.41324788e-01 3.81666213e-01
6.74416900e-01 3.59913737e-01 8.49961877e-01 7.20357522e-02
1.70037612e-01 -3.24631669e-02 -5.39642334e-01 -2.00027972e-01
1.17653930e+00 -2.98934579e-01 -5.08662224e-01 -3.28095108e-01
-7.65284061e-01 4.05349731e-01 2.62512565e-01 8.10243562e-02
1.83672652e-01 -1.01264679e+00 -1.58674553e-01 1.12346284e-01
-3.96861553e-01 -4.99965727e-01 6.65023744e-01 1.35930371e+00
-2.11813018e-01 2.31442720e-01 -4.55974638e-02 -6.78088546e-01
-1.59747076e+00 1.52415350e-01 6.62756503e-01 -1.52486727e-01
-6.19969368e-01 -1.08785024e-02 -8.52031708e-01 2.78829098e-01
1.75441980e-01 -5.05256534e-01 1.90284848e-01 1.36903871e-03
5.15660465e-01 1.29736197e+00 1.55375481e-01 3.30136210e-01
-5.99635303e-01 6.07811809e-01 1.02664864e+00 1.49676099e-01
1.00139892e+00 -6.13916874e-01 1.37472361e-01 4.74302858e-01
8.74934435e-01 2.59054571e-01 -1.14352334e+00 3.30590427e-01
-2.73700237e-01 -2.05730259e-01 1.68915227e-01 -1.14511919e+00
-1.02056944e+00 2.16436774e-01 1.27635896e+00 5.53365588e-01
1.65765703e+00 -3.99518818e-01 1.16839683e+00 2.20738664e-01
4.90041018e-01 -1.40135550e+00 -5.62318563e-01 -1.98792234e-01
5.55272289e-02 -6.73901856e-01 3.49721789e-01 -1.53457314e-01
-5.49689293e-01 8.20253789e-01 3.49327683e-01 1.90706477e-01
8.35859239e-01 5.67634046e-01 -1.20248109e-01 2.17192575e-01
-8.51576209e-01 9.19182748e-02 -1.67198583e-01 3.23255450e-01
2.38410279e-01 6.15212619e-01 -6.80932283e-01 8.55646670e-01
-2.04144537e-01 6.07083976e-01 5.45361817e-01 7.38452137e-01
-4.67752367e-01 -6.91726923e-01 -3.93684268e-01 7.49696851e-01
-6.11896396e-01 1.25326723e-01 1.48665801e-01 2.07177132e-01
1.76403135e-01 1.31159663e+00 2.68136531e-01 -5.35698056e-01
5.25649190e-01 4.97803807e-01 2.24717110e-01 -1.03033155e-01
-9.42019463e-01 1.21366844e-01 2.17596397e-01 -9.83647346e-01
-3.71459454e-01 -3.86840820e-01 -1.16508090e+00 -8.14495444e-01
-3.94491181e-02 6.79966733e-02 9.23027039e-01 6.71691298e-01
1.04179609e+00 6.70228899e-01 9.54892695e-01 -5.39804220e-01
-3.36260021e-01 -8.38674366e-01 -7.48749256e-01 -2.61890560e-01
-1.20036034e-02 -1.06715225e-01 -5.28310835e-01 9.91884172e-02]
|
[13.92483901977539, 3.038970947265625]
|
8033067f-24b0-4f21-bdb6-8896b0a8ba4b
|
modeling-variable-space-with-residual-tensor
| null | null |
https://openreview.net/forum?id=Qx0EswNY_bW
|
https://openreview.net/pdf?id=Qx0EswNY_bW
|
Modeling Variable Space with Residual Tensor Networks for Multivariate Time Series
|
Multivariate time series involve a series of valuable applications in the real world, and the basic premise of which is that multiple variables are interdependent. However, the relationship between variables in the latent space is dynamic and complex, and as the time window increases, the size of the space also increases exponentially. For fully exploiting the dependencies in the variable space, we propose Modeling Variable Space with Residual Tensor Networks (MVSRTN) for multivariate time series. In this framework, we derive the mathematical representation of the variable space, and then use a tensor network based on the idea of low-rank approximation to model the variable space. The tensor components are shared to ensure the translation invariance of the network. In order to improve the ability to model long-term sequences, we propose an N-order residual connection approach and couple it to the space-approximated tensor network. Moreover, the series-variable encoder is designed to improve the quality of the variable space, and we use the skip-connection layer to achieve the dissemination of information such as scale. Experimental results verify the effectiveness of our proposed method on four multivariate time series forecasting benchmark datasets.
|
['Guangjian Tian', 'Jun Wang', 'Siwei Rao', 'Yupeng He', 'Peng Zhang', 'Jing Zhang']
|
2021-09-29
| null | null | null | null |
['tensor-networks']
|
['methodology']
|
[-2.17058033e-01 -5.38605452e-01 -2.15590999e-01 -2.00825214e-01
1.58187628e-01 -5.31537294e-01 3.04846376e-01 -3.14751983e-01
1.70317199e-02 2.67491043e-01 3.06827337e-01 -2.59186536e-01
-4.00096059e-01 -6.91776335e-01 -5.35850644e-01 -7.80466557e-01
-4.13088143e-01 -6.07365035e-02 -9.25525427e-02 -3.53971392e-01
1.05326138e-01 3.22665900e-01 -9.46891189e-01 1.79807954e-02
5.90901136e-01 1.37817276e+00 2.34195858e-01 4.10524398e-01
-4.22181606e-01 1.00514877e+00 -4.49707240e-01 -2.32044831e-01
1.94862098e-01 -3.35799426e-01 -6.27138555e-01 1.83267176e-01
-3.07848364e-01 -1.74151093e-01 -5.93733728e-01 9.40332472e-01
4.23982069e-02 3.03435951e-01 2.50077456e-01 -1.45208812e+00
-9.58629012e-01 8.25264335e-01 -4.25482452e-01 2.61398762e-01
9.74552613e-03 -1.02050276e-03 1.28125834e+00 -7.91678011e-01
4.73387659e-01 1.43996787e+00 5.43224573e-01 -1.38639873e-02
-1.15136290e+00 -6.83544815e-01 6.72398925e-01 5.34549952e-01
-1.33438480e+00 -1.23423889e-01 1.39002919e+00 -6.53034151e-01
6.23596609e-01 2.97048002e-01 7.39072978e-01 1.06111395e+00
2.78114140e-01 6.42855287e-01 6.11560643e-01 1.24765687e-01
-7.72999553e-03 -2.36337021e-01 1.26637310e-01 3.87428761e-01
-2.78275222e-01 -5.45902029e-02 -4.98259217e-01 -1.58629298e-01
1.00068069e+00 9.28765297e-01 -4.09890473e-01 -3.85302722e-01
-1.72248042e+00 8.33220541e-01 6.60311937e-01 6.78076267e-01
-6.83236599e-01 2.22411960e-01 6.13009214e-01 5.31785548e-01
6.86113775e-01 1.54897600e-01 -6.57831430e-01 -3.08531880e-01
-7.63086081e-01 -7.50605986e-02 5.29101133e-01 1.00949168e+00
5.67598820e-01 3.24039608e-01 -4.92523164e-02 6.54044211e-01
3.61865342e-01 3.89966846e-01 9.02199447e-01 -7.73452103e-01
9.00952220e-01 7.06827760e-01 -7.55152851e-02 -1.60096300e+00
-3.89226943e-01 -7.97214389e-01 -1.31021023e+00 -5.42170286e-01
1.77106336e-01 3.97576056e-02 -5.00372767e-01 1.74204528e+00
2.52246380e-01 4.84559864e-01 -1.44971877e-01 9.50552046e-01
2.24706322e-01 8.43327761e-01 -3.26034755e-01 -5.62833369e-01
1.07754731e+00 -9.45350528e-01 -1.12356925e+00 3.22937250e-01
5.19073844e-01 -5.41877389e-01 9.02794600e-01 1.42519549e-01
-7.47638524e-01 -7.31769860e-01 -8.90826881e-01 -2.01715931e-01
-1.84746981e-01 -9.83028412e-02 7.17828691e-01 -3.02233957e-02
-6.65496767e-01 8.39660585e-01 -1.07148159e+00 1.62407547e-01
2.09603123e-02 1.02929294e-01 -2.49936834e-01 7.55568817e-02
-1.40653002e+00 3.79072249e-01 2.04039916e-01 7.68231869e-01
-6.17666841e-01 -5.81224203e-01 -6.40000224e-01 1.49975717e-01
1.57196522e-01 -5.61592758e-01 7.76225686e-01 -9.41060424e-01
-1.36713791e+00 -2.80207425e-01 -3.47460836e-01 -1.20373718e-01
3.90522838e-01 -4.70767692e-02 -7.33310759e-01 1.18724689e-01
-1.07381925e-01 -1.65346175e-01 1.08482230e+00 -8.11830103e-01
-5.05967081e-01 -4.24098760e-01 3.41322094e-01 -1.06793545e-01
-8.50643635e-01 -9.58349705e-02 -3.30829442e-01 -9.21205580e-01
6.35803640e-01 -9.47209120e-01 -3.54151577e-01 -1.38545200e-01
-1.25437334e-01 -2.63919175e-01 9.61939454e-01 -1.07808518e+00
1.74156308e+00 -2.42046714e+00 8.16583395e-01 4.15968686e-01
5.90269387e-01 -3.70258689e-01 -2.72781610e-01 5.28428078e-01
-4.89923835e-01 6.75476342e-02 -2.42987737e-01 -4.18601573e-01
-1.29923448e-01 5.20786941e-01 -7.64040470e-01 5.37541807e-01
1.33901671e-01 8.28390062e-01 -9.48697865e-01 -2.28321984e-01
-4.80384305e-02 6.72802806e-01 -4.32129622e-01 2.96680450e-01
-6.79044873e-02 7.73462474e-01 -8.51731420e-01 3.74444276e-01
6.47764087e-01 -5.41726887e-01 3.81719880e-02 -6.19464993e-01
-3.41767043e-01 2.34441534e-01 -1.18522036e+00 1.99321270e+00
-5.02091050e-01 3.79779249e-01 -2.21925557e-01 -1.13416123e+00
9.47170019e-01 4.49203759e-01 9.06704783e-01 -5.61434507e-01
-2.62717856e-03 1.59282848e-01 3.27048227e-02 -7.23503828e-01
3.81400049e-01 -2.92949192e-02 2.42044434e-01 1.73006147e-01
-2.09866002e-01 3.11163306e-01 1.11297771e-01 1.68128669e-01
7.94371068e-01 1.60947651e-01 -2.30932564e-01 4.99477759e-02
8.31062138e-01 -5.15912056e-01 8.21097195e-01 -1.58874482e-01
1.36536837e-01 1.29792541e-01 8.65640104e-01 -7.49639869e-01
-1.00118029e+00 -5.18434465e-01 -1.98832843e-02 8.78513157e-01
2.53985897e-02 -6.26223743e-01 -2.02845916e-01 -4.40170348e-01
-3.37723494e-02 3.68729949e-01 -7.89148271e-01 -1.58092663e-01
-7.36756206e-01 -4.54615176e-01 -1.24359932e-02 5.15421569e-01
1.52923122e-01 -6.68807030e-01 -9.95888859e-02 4.89838421e-01
-6.59478366e-01 -8.62178087e-01 -6.87822938e-01 -4.54864837e-02
-1.10807419e+00 -6.66132689e-01 -7.77727544e-01 -5.51735282e-01
5.30240476e-01 4.53764111e-01 7.55017996e-01 2.24434003e-01
3.87068003e-01 1.76511481e-01 -6.00019455e-01 2.49006435e-01
1.92703065e-02 5.87026998e-02 1.44176573e-01 6.38812542e-01
9.28213373e-02 -1.16865242e+00 -7.37429976e-01 2.36626074e-01
-1.23408592e+00 5.95502183e-02 3.50325674e-01 1.01365685e+00
5.02432883e-01 2.18669683e-01 2.98834980e-01 -4.14248168e-01
4.93245155e-01 -7.97420144e-01 -5.38492143e-01 7.58381933e-02
-6.14907682e-01 1.33848116e-01 1.04607511e+00 -9.99488294e-01
-4.50192302e-01 -2.97437042e-01 2.81664878e-01 -1.01996756e+00
7.24197090e-01 1.14770210e+00 1.90623675e-03 1.50028571e-01
2.05233209e-02 3.35898042e-01 4.92242873e-02 -8.74117374e-01
4.84327257e-01 3.30929726e-01 1.40063137e-01 -3.60869139e-01
1.14124179e+00 5.74683666e-01 3.02918702e-01 -4.29025233e-01
-5.86422145e-01 -3.46550792e-01 -7.42263496e-01 -1.09970875e-01
6.40352428e-01 -8.47062230e-01 -8.73279512e-01 3.50968093e-01
-1.28501225e+00 1.08177185e-01 -1.54891551e-01 8.38010609e-01
-2.28211179e-01 6.15884185e-01 -7.88870215e-01 -5.57223976e-01
-1.12248458e-01 -1.17446780e+00 8.12782645e-01 -1.68172181e-01
1.18974857e-01 -1.22127140e+00 -2.50871480e-02 -1.09730557e-01
6.87123716e-01 2.36921921e-01 1.00065875e+00 -2.29395449e-01
-6.76136315e-01 -2.09090114e-01 -1.14452414e-01 4.27865237e-01
1.60037532e-01 6.94458783e-02 -4.99356508e-01 -4.39583778e-01
4.32578564e-01 2.89779067e-01 5.39390385e-01 1.22214109e-01
1.46365297e+00 -5.50056517e-01 -1.35435564e-02 8.60562444e-01
1.20687246e+00 2.11302936e-02 2.95748949e-01 2.29669794e-01
1.29364800e+00 6.72655702e-01 3.73582095e-01 6.62602425e-01
8.12730551e-01 6.19479835e-01 3.63497108e-01 -1.69767663e-02
2.70439953e-01 -2.67418325e-01 4.22885031e-01 2.03246284e+00
-3.52069259e-01 1.51393920e-01 -6.03941381e-01 3.86738867e-01
-1.95327759e+00 -9.68620479e-01 -3.31556737e-01 1.99815059e+00
6.27254546e-01 -4.03099880e-02 -5.78324609e-02 1.60636976e-01
5.60487509e-01 5.06536841e-01 -7.09428251e-01 4.86675277e-02
-1.04914486e-01 -3.99644256e-01 4.05390024e-01 2.94037282e-01
-7.41304219e-01 4.32907581e-01 5.43160534e+00 5.87188184e-01
-1.64459968e+00 1.72735065e-01 2.77727216e-01 -4.13860194e-02
-6.95026457e-01 2.92142957e-01 -1.89846963e-01 7.46138334e-01
1.02079022e+00 -3.80763710e-01 9.10499573e-01 5.77300966e-01
3.86660337e-01 9.63004053e-01 -1.04308045e+00 1.00099385e+00
-2.54582316e-01 -1.07172298e+00 4.31049913e-02 3.40004675e-02
4.83061522e-01 1.39533950e-03 1.83546230e-01 4.73063767e-01
-1.47201836e-01 -6.41928732e-01 8.78536940e-01 9.02053356e-01
6.74189210e-01 -6.10469878e-01 5.06473720e-01 2.52473086e-01
-1.78562665e+00 -2.93564647e-01 -5.63278258e-01 -1.88799590e-01
2.26188228e-01 7.62691259e-01 -1.13365710e-01 8.08515728e-01
6.93455100e-01 1.45622802e+00 -4.81056988e-01 6.17503166e-01
-9.94460881e-02 5.50344408e-01 -1.77145660e-01 1.66317135e-01
2.21436948e-01 -7.34101176e-01 5.65415144e-01 6.95711613e-01
3.78739953e-01 -4.64397296e-02 3.11863005e-01 8.00339818e-01
1.52473181e-01 1.55796558e-01 -2.55426437e-01 -4.03107733e-01
4.03193772e-01 1.22352254e+00 -3.60408843e-01 -2.92238414e-01
-5.99934280e-01 9.88488853e-01 2.56347716e-01 8.58888328e-01
-9.05379832e-01 -2.53745556e-01 5.89629173e-01 -4.40058708e-01
5.99653482e-01 -7.52679884e-01 -5.87647706e-02 -1.61766875e+00
6.75367773e-01 -9.20358181e-01 2.51754612e-01 -5.15841365e-01
-1.37096632e+00 8.21196556e-01 -3.00964534e-01 -1.77439547e+00
-3.34398329e-01 -2.55459309e-01 -2.58359343e-01 1.15239346e+00
-1.62191916e+00 -1.41104829e+00 -1.51184797e-01 1.01765168e+00
3.74599367e-01 8.32167640e-03 6.39914155e-01 6.61001027e-01
-7.25898027e-01 4.29515570e-01 4.14605141e-01 8.05552527e-02
2.97925293e-01 -8.97583127e-01 3.65874559e-01 8.16561699e-01
-2.06464026e-02 1.08794427e+00 5.20382047e-01 -6.09837413e-01
-2.02772760e+00 -1.12814724e+00 8.40537548e-01 -2.22976103e-01
1.39878726e+00 -4.51346308e-01 -1.17023134e+00 8.19130301e-01
-3.10035378e-01 2.11013019e-01 6.09051824e-01 3.50240737e-01
-6.96538866e-01 -4.00211096e-01 -4.68830884e-01 4.99952674e-01
9.07698035e-01 -1.04505253e+00 -3.34794551e-01 5.03001690e-01
1.49957991e+00 -1.57496154e-01 -1.42203569e+00 2.45105445e-01
6.55765474e-01 -3.82643968e-01 8.77661347e-01 -7.74002612e-01
5.48732400e-01 -6.18189037e-01 -2.91074365e-01 -1.37703681e+00
-8.64873409e-01 -6.79349244e-01 -4.96817261e-01 1.38820362e+00
2.51481831e-01 -8.00837576e-01 2.50622600e-01 5.74942112e-01
-4.96614212e-03 -8.11407149e-01 -1.03860080e+00 -8.11655700e-01
-1.14341617e-01 -4.07087892e-01 1.00386775e+00 1.36251926e+00
-4.15884331e-02 4.31850255e-01 -6.99061155e-01 3.35533917e-01
3.53658438e-01 4.63876843e-01 5.38512826e-01 -1.05800879e+00
-4.74334478e-01 -4.01265621e-01 -5.61637938e-01 -1.41715157e+00
3.39570045e-02 -7.21037984e-01 -4.17652220e-01 -1.15030169e+00
-7.50784650e-02 -4.04401660e-01 -8.87040079e-01 8.12433660e-02
-1.78595886e-01 -4.38978791e-01 2.29086176e-01 8.47578347e-01
-3.03271890e-01 1.06850779e+00 1.54036534e+00 -1.06443100e-01
-1.36515960e-01 -6.49105012e-02 -3.13262463e-01 2.87353247e-01
3.63043249e-01 -2.77621746e-01 -5.57923794e-01 -7.60036826e-01
3.10165346e-01 4.47810709e-01 1.20269671e-01 -5.43843150e-01
2.72253454e-01 -4.14201051e-01 1.36072114e-01 -5.70431709e-01
4.21574563e-01 -1.31698215e+00 3.79527807e-01 2.67427772e-01
-3.08267117e-01 6.34612739e-01 -2.95417041e-01 6.00252450e-01
-5.20141542e-01 3.84605408e-01 5.35555407e-02 2.96902955e-01
-4.03998554e-01 8.70323658e-01 2.08390802e-01 -4.51571643e-01
5.76043487e-01 2.60878533e-01 -2.18756720e-01 -3.26612949e-01
-5.49312353e-01 3.55916291e-01 2.03980491e-01 7.98771977e-01
6.06594682e-01 -1.83073342e+00 -5.18525362e-01 3.16503495e-01
5.88260293e-02 -4.76408005e-02 4.67024028e-01 1.30182421e+00
-2.16576740e-01 2.88632751e-01 -4.86301333e-02 -6.44481301e-01
-6.63439512e-01 1.13597858e+00 1.43337756e-01 -2.90663749e-01
-8.40585709e-01 5.97694337e-01 2.44214550e-01 -2.20048994e-01
2.17060223e-01 -6.38026834e-01 -5.77668548e-01 2.14691833e-01
6.84370577e-01 2.15635940e-01 -3.31415236e-01 -7.79388845e-01
-1.68876827e-01 5.45357883e-01 1.09045133e-01 -8.34585279e-02
1.65079355e+00 -5.58097124e-01 -6.82794094e-01 1.20850396e+00
1.67418325e+00 -5.25155254e-02 -1.15621674e+00 -6.93918765e-01
-7.75160864e-02 -5.36724150e-01 1.84464648e-01 1.13142729e-02
-1.39585912e+00 8.83172393e-01 4.06904101e-01 6.53647184e-01
1.27809238e+00 -4.58265394e-01 9.60615575e-01 2.76578277e-01
2.55965441e-01 -7.52160251e-01 -8.12643096e-02 6.52408838e-01
1.06992400e+00 -7.75542378e-01 -7.68770203e-02 -2.69373924e-01
-4.05296057e-01 1.39549243e+00 1.10020511e-01 -1.41601920e-01
9.23876762e-01 -2.59031504e-01 -4.01879698e-02 -2.37867743e-01
-8.03536117e-01 3.25484455e-01 3.66984785e-01 1.14713944e-01
5.62164962e-01 1.87316269e-01 -4.39607680e-01 6.02954090e-01
-2.47359812e-01 -1.33077085e-01 1.98971570e-01 5.43381214e-01
1.66770637e-01 -1.14312363e+00 -3.84506702e-01 2.07820565e-01
-3.25328320e-01 -6.23434521e-02 1.07713372e-01 2.52808154e-01
4.20722691e-03 8.91415298e-01 -6.14851303e-02 -7.68889189e-01
3.45390469e-01 -1.08438171e-01 -1.74953490e-01 -6.69951970e-03
-2.84887254e-01 3.53107750e-01 -3.54665935e-01 -7.78461933e-01
-5.25766909e-01 -6.01602674e-01 -9.97617483e-01 -3.51420075e-01
-2.77190149e-01 2.30756804e-01 9.23312545e-01 8.86213720e-01
3.88988197e-01 9.18667495e-01 1.41655767e+00 -5.28211117e-01
-6.81295335e-01 -1.10028875e+00 -7.75684953e-01 7.46065080e-01
7.30570614e-01 -6.83874071e-01 -4.65960145e-01 1.34592071e-01]
|
[6.900774002075195, 2.9222614765167236]
|
1b67de87-724a-4211-b7c0-7b5b201e1fed
|
webformer-the-web-page-transformer-for
|
2202.00217
| null |
https://arxiv.org/abs/2202.00217v1
|
https://arxiv.org/pdf/2202.00217v1.pdf
|
WebFormer: The Web-page Transformer for Structure Information Extraction
|
Structure information extraction refers to the task of extracting structured text fields from web pages, such as extracting a product offer from a shopping page including product title, description, brand and price. It is an important research topic which has been widely studied in document understanding and web search. Recent natural language models with sequence modeling have demonstrated state-of-the-art performance on web information extraction. However, effectively serializing tokens from unstructured web pages is challenging in practice due to a variety of web layout patterns. Limited work has focused on modeling the web layout for extracting the text fields. In this paper, we introduce WebFormer, a Web-page transFormer model for structure information extraction from web documents. First, we design HTML tokens for each DOM node in the HTML by embedding representations from their neighboring tokens through graph attention. Second, we construct rich attention patterns between HTML tokens and text tokens, which leverages the web layout for effective attention weight computation. We conduct an extensive set of experiments on SWDE and Common Crawl benchmarks. Experimental results demonstrate the superior performance of the proposed approach over several state-of-the-art methods.
|
['Dongfang Liu', 'Xiaojun Quan', 'Fuli Feng', 'Anirudh Ravula', 'Yi Fang', 'Qifan Wang']
|
2022-02-01
| null | null | null | null |
['deep-attention', 'deep-attention']
|
['computer-vision', 'natural-language-processing']
|
[ 4.27552611e-01 1.11795664e-01 -6.39777601e-01 -1.68000117e-01
-8.91543567e-01 -8.82044911e-01 4.67996478e-01 4.58158910e-01
-2.09053472e-01 2.82492250e-01 5.02914369e-01 -5.98471284e-01
-1.84333801e-01 -7.52477050e-01 -7.92150915e-01 -4.14648205e-01
-1.36064351e-01 2.08995864e-01 1.92125708e-01 4.82289866e-02
2.85696954e-01 4.45450516e-03 -1.37316954e+00 5.88601530e-01
9.87375021e-01 1.16816032e+00 4.00848716e-01 3.99129152e-01
-1.15241265e+00 9.01672184e-01 -2.56074071e-01 -7.08476067e-01
8.15585181e-02 -1.89455792e-01 -1.07276416e+00 3.49735171e-01
4.51013654e-01 -3.93172413e-01 -5.00527799e-01 1.32677901e+00
2.44727880e-01 -2.89685547e-01 4.63722885e-01 -1.12426865e+00
-8.50412071e-01 1.29633760e+00 -9.76296604e-01 1.79538354e-01
4.89017308e-01 -1.73970595e-01 1.83156955e+00 -8.23274016e-01
8.64648342e-01 9.33039546e-01 3.52934510e-01 6.89400434e-02
-9.76455808e-01 -3.20716143e-01 5.33780932e-01 3.74996543e-01
-1.00408840e+00 1.40231982e-01 1.07353294e+00 -1.81277096e-01
1.14428186e+00 5.35281003e-02 5.40254951e-01 1.08651018e+00
2.62046397e-01 1.50700378e+00 6.34558499e-01 -5.17175734e-01
-2.06541777e-01 1.14459164e-01 8.04886341e-01 8.30203056e-01
5.93913674e-01 -6.14336073e-01 -4.04489040e-01 -2.22192938e-03
3.62970918e-01 1.32230893e-02 -2.42601875e-02 -5.07762730e-01
-9.28222358e-01 8.27563405e-01 3.81880105e-01 6.60603121e-02
-5.19972563e-01 6.43937886e-02 6.85094535e-01 -6.88872561e-02
1.59096509e-01 9.63398516e-02 -5.35422623e-01 -2.22349912e-02
-4.63485926e-01 2.54104584e-01 9.73637104e-01 1.48211193e+00
5.86157739e-01 -2.40704492e-01 -7.52676204e-02 9.05473530e-01
4.47752535e-01 4.32862073e-01 5.47875762e-01 -2.39075586e-01
1.13712907e+00 9.10584927e-01 -1.42252386e-01 -9.38504696e-01
-2.21125945e-01 -2.79456019e-01 -5.04914641e-01 -3.77261341e-01
1.68308169e-01 -3.49299535e-02 -9.17070389e-01 9.90594506e-01
3.69991124e-01 -4.60010678e-01 -1.90180376e-01 6.70761645e-01
8.87769580e-01 6.82486296e-01 1.99175343e-01 1.69981971e-01
1.87507665e+00 -1.09114659e+00 -9.23220992e-01 -5.05559683e-01
4.71727401e-01 -7.83851266e-01 1.07348335e+00 2.46484607e-01
-8.80136132e-01 -3.17799985e-01 -1.04646671e+00 -4.45565015e-01
-7.97889769e-01 2.81618498e-02 7.77463377e-01 2.83790171e-01
-3.40015739e-01 3.50312799e-01 -6.20302379e-01 -4.03838903e-01
4.16756451e-01 1.01442441e-01 -8.62676650e-02 -1.01888098e-01
-1.30745244e+00 1.40556827e-01 5.29063404e-01 5.49968518e-02
-3.81079823e-01 -5.31464159e-01 -1.21133828e+00 6.59151196e-01
1.04890561e+00 -3.19609642e-01 1.37637198e+00 -7.20683157e-01
-9.30657268e-01 7.50489295e-01 -1.06844336e-01 -5.40189207e-01
7.20513612e-02 -5.29073775e-01 -6.20640278e-01 5.71357831e-02
2.32970685e-01 1.79866523e-01 6.98974967e-01 -1.19572544e+00
-8.86458516e-01 -5.44617176e-01 -2.12366898e-02 -3.95957865e-02
-6.67979538e-01 -4.94747348e-02 -1.07980919e+00 -8.01935673e-01
3.23883146e-02 -6.19289994e-01 -5.01792058e-02 -4.36617762e-01
-9.78943825e-01 -4.91040468e-01 7.20417678e-01 -1.11459935e+00
1.62156343e+00 -1.85589361e+00 -3.26276898e-01 4.66077149e-01
4.95045394e-01 8.25614706e-02 -3.37654710e-01 7.46123552e-01
-4.75813188e-02 6.89762771e-01 2.71207958e-01 2.28730470e-01
6.40783310e-01 -1.31869614e-01 -5.75457931e-01 -1.92242146e-01
2.24079356e-01 1.19908881e+00 -7.99102426e-01 -9.03876781e-01
-1.58461511e-01 2.49963269e-01 -4.96011794e-01 9.85984132e-02
-5.89779496e-01 -7.09013164e-01 -9.44864213e-01 7.55659699e-01
5.87954223e-01 -7.20949352e-01 5.49450099e-01 -4.12894875e-01
8.85786712e-02 8.27700436e-01 -1.08529270e+00 1.47578311e+00
-3.64117026e-01 5.25310159e-01 -7.95563981e-02 -5.47253430e-01
5.27987242e-01 -1.89095049e-03 6.28121316e-01 -9.72783983e-01
1.60550207e-01 -1.78470194e-01 -6.83414638e-02 -7.39234209e-01
5.80651700e-01 7.22519577e-01 -2.68265992e-01 5.64616024e-01
1.22037500e-01 4.66577202e-01 7.62336791e-01 5.45455337e-01
1.23172510e+00 2.77295202e-01 3.64771605e-01 -1.80387259e-01
4.00229722e-01 -7.57986680e-02 3.78735006e-01 6.73504710e-01
1.66945815e-01 3.46087888e-02 7.59065211e-01 -3.79586190e-01
-1.22279048e+00 -1.08819616e+00 2.10991457e-01 1.16531050e+00
1.36021256e-01 -1.02076793e+00 -1.07822299e+00 -1.18265283e+00
-1.69728766e-03 5.14327109e-01 -7.16657460e-01 2.88135290e-01
-8.05976570e-01 -4.78425354e-01 2.23326325e-01 7.58309424e-01
5.15084267e-01 -1.27347386e+00 -1.43303454e-01 4.89728242e-01
-4.31555688e-01 -1.42591441e+00 -9.11036789e-01 2.51729488e-01
-5.72387278e-01 -1.24163556e+00 -5.09542167e-01 -1.30526805e+00
6.52121902e-01 4.23015237e-01 1.33654618e+00 2.21235529e-01
-5.08269012e-01 3.13086689e-01 -5.60069740e-01 -3.35112065e-01
2.18062364e-02 7.12972164e-01 -5.46327412e-01 -4.13581952e-02
7.97484219e-01 -1.60337433e-01 -4.41822797e-01 4.46810424e-02
-1.28175390e+00 6.37277812e-02 1.00043452e+00 7.94232547e-01
5.80991626e-01 4.51961488e-01 1.85487360e-01 -1.45113325e+00
9.48288083e-01 -5.52138388e-01 -7.41344333e-01 4.85531658e-01
-7.06503749e-01 6.78299665e-01 9.13208306e-01 -2.71735907e-01
-1.20363390e+00 -1.33734038e-02 -1.28250018e-01 2.01564953e-01
-6.82403371e-02 1.06719673e+00 -5.81303835e-01 5.21652937e-01
2.80506257e-03 5.38344443e-01 -4.79823500e-01 -8.25634837e-01
3.99414927e-01 7.55735338e-01 2.81171352e-01 -5.82902670e-01
9.31068480e-01 1.49135098e-01 -4.56347734e-01 -8.40721548e-01
-8.49288285e-01 -6.76371455e-01 -6.69183373e-01 1.32998198e-01
6.95651114e-01 -4.03721869e-01 -7.50605345e-01 2.05848798e-01
-1.07171500e+00 -6.66379631e-02 6.65934011e-02 -3.24915871e-02
-1.96065545e-01 8.69886935e-01 -7.89920390e-01 -6.01109862e-01
-6.17967904e-01 -8.46162319e-01 1.14543355e+00 1.79013863e-01
9.22446623e-02 -9.47854519e-01 -1.21608026e-01 4.07247126e-01
-1.42005637e-01 -9.46677923e-02 1.49598551e+00 -9.40860689e-01
-8.51451635e-01 -3.41395676e-01 -4.72455651e-01 -2.21874952e-01
2.61484236e-01 -8.66564587e-02 -5.02540588e-01 8.60820562e-02
-5.21635532e-01 -1.39140308e-01 9.29171920e-01 -3.34303491e-02
1.38799942e+00 -7.92487979e-01 -5.68050742e-01 5.14623046e-01
1.64600909e+00 3.00781041e-01 6.04865015e-01 7.66001284e-01
8.98550808e-01 6.53808653e-01 4.86529857e-01 4.82082546e-01
4.18285072e-01 3.68991077e-01 4.15407807e-01 -5.29079251e-02
-4.35581021e-02 -7.72511721e-01 2.30937093e-01 9.02014017e-01
5.93919396e-01 -5.50347745e-01 -7.16992259e-01 5.57249069e-01
-1.64278388e+00 -9.55171287e-01 3.20343953e-03 1.77540696e+00
9.35522377e-01 6.02692306e-01 2.23555285e-02 1.28787130e-01
7.49626637e-01 3.45192760e-01 -5.93439281e-01 -7.88030401e-02
1.37023598e-01 -4.85896431e-02 7.93008566e-01 1.88032642e-01
-1.16856694e+00 1.00514948e+00 5.32828045e+00 8.77860844e-01
-4.50223207e-01 -4.66968119e-01 3.36351365e-01 3.02829295e-01
-6.42308652e-01 -9.91619900e-02 -1.32819605e+00 6.76639378e-01
4.84887332e-01 -2.42449045e-01 5.47018647e-01 1.20026040e+00
-1.21083669e-02 2.72898704e-01 -9.79960501e-01 6.85082138e-01
2.16820836e-01 -1.21736646e+00 4.16353703e-01 2.57605016e-01
5.35450637e-01 -1.87763140e-01 -5.70885725e-02 2.75046140e-01
6.23599410e-01 -6.40695095e-01 5.80803275e-01 -4.57323045e-02
5.47253788e-01 -7.18259513e-01 8.02071750e-01 3.64313126e-02
-1.65787888e+00 -1.49065834e-02 -2.30517566e-01 6.49940908e-01
8.34203213e-02 5.53074837e-01 -8.82168412e-01 4.25296009e-01
8.77492070e-01 6.23425663e-01 -7.46863723e-01 9.47696447e-01
-4.01681006e-01 7.05868065e-01 -1.34247571e-01 -7.42272019e-01
5.67791879e-01 -2.05538720e-01 1.87528968e-01 1.43930519e+00
-8.20694640e-02 -2.21520007e-01 1.85103044e-01 7.63962626e-01
-6.05336428e-01 4.17168111e-01 -4.84337538e-01 -6.72104537e-01
2.16971099e-01 1.49951494e+00 -9.55735445e-01 -3.28121364e-01
-8.27954054e-01 8.13875377e-01 4.40604478e-01 5.01205564e-01
-8.85211051e-01 -1.25544441e+00 5.52757680e-01 2.94123381e-01
9.55938160e-01 -3.34744722e-01 -1.59430683e-01 -1.17925882e+00
4.89248842e-01 -1.00058389e+00 5.58643997e-01 -7.99746871e-01
-1.25579107e+00 6.23564482e-01 -2.04938337e-01 -1.01533163e+00
-2.42816716e-01 -8.45624447e-01 -3.38843435e-01 5.10276437e-01
-1.71201098e+00 -1.19745731e+00 -2.85117418e-01 3.01347852e-01
9.57553923e-01 9.76479724e-02 3.00365806e-01 1.69860944e-01
-5.48290193e-01 5.88170707e-01 2.88554341e-01 1.00098777e+00
2.95881212e-01 -1.49887753e+00 1.03062141e+00 8.51037264e-01
5.32680929e-01 9.21186864e-01 3.00647169e-01 -9.73878145e-01
-1.95644283e+00 -9.84801352e-01 1.11111319e+00 -1.38158187e-01
1.05489028e+00 -7.17447400e-01 -8.80312026e-01 9.33752000e-01
6.20171547e-01 -5.40314674e-01 8.56971562e-01 2.22415939e-01
-6.09516323e-01 -1.10651217e-01 -5.18399179e-01 9.88385081e-01
1.09724283e+00 -3.54809105e-01 -6.23984277e-01 3.61658037e-01
9.63230729e-01 -3.79092395e-02 -7.56629527e-01 -1.20657302e-01
6.41699731e-01 -3.71084958e-01 9.80876327e-01 -7.93855071e-01
5.95898092e-01 5.96582182e-02 2.76376784e-01 -1.03195667e+00
-5.44631839e-01 -6.39721870e-01 -2.99978465e-01 1.68234146e+00
6.92350984e-01 -2.87543446e-01 1.08085465e+00 3.81460816e-01
2.73891509e-01 -6.81599855e-01 -1.70238927e-01 -5.97345889e-01
-3.72689992e-01 -3.29724193e-01 7.78290987e-01 5.76069534e-01
3.39270115e-01 7.95151234e-01 -1.32594481e-01 -3.17204446e-02
8.10335338e-01 4.80255127e-01 6.50753260e-01 -1.11003506e+00
-1.40911832e-01 -5.66152096e-01 7.48751834e-02 -1.43372369e+00
1.37385279e-01 -1.10820174e+00 1.25884756e-01 -2.01460648e+00
4.66213644e-01 9.07879397e-02 -3.14038515e-01 4.48832363e-01
-2.22625598e-01 -3.76646399e-01 1.31872028e-01 -2.88574472e-02
-9.37596798e-01 3.09254467e-01 1.09798753e+00 -6.26603127e-01
8.40699896e-02 -2.41038278e-01 -9.35935915e-01 5.83188474e-01
6.87022328e-01 -5.27050495e-01 -3.56876314e-01 -5.04130840e-01
5.71404636e-01 -1.53125644e-01 -1.78279191e-01 -4.85247999e-01
2.70159006e-01 -1.10014386e-01 4.91346389e-01 -8.83650541e-01
-2.60500371e-01 -1.06166744e+00 -6.36391401e-01 2.33763218e-01
-6.49331212e-01 2.21154302e-01 1.17524743e-01 7.72528708e-01
-1.51553780e-01 -5.14229178e-01 1.40748188e-01 -2.51386821e-01
-1.00806069e+00 5.35965383e-01 -5.75916350e-01 4.00165379e-01
3.89663965e-01 3.16333026e-02 -2.83401161e-01 -2.41199926e-01
-2.48014063e-01 4.09006745e-01 5.12172803e-02 8.37869048e-01
6.46010697e-01 -1.31574869e+00 -4.06892955e-01 2.72135854e-01
3.54611754e-01 -2.42202222e-01 -7.44979084e-02 1.60268068e-01
-5.49495280e-01 7.88143218e-01 2.07775217e-02 -1.14555217e-01
-1.33836865e+00 1.06172872e+00 -4.30362135e-01 -8.64703000e-01
-6.53450370e-01 2.74994105e-01 2.18306720e-01 -9.87003893e-02
4.30391133e-01 -6.99583054e-01 -4.15431798e-01 -2.22242456e-02
6.80913150e-01 8.49601924e-02 1.32949680e-01 -1.30883187e-01
-1.49424002e-01 6.95312500e-01 -7.15538621e-01 1.57177374e-01
1.20603752e+00 -1.59683630e-01 2.64859833e-02 -1.22447781e-01
1.38195634e+00 1.60737276e-01 -1.08455622e+00 -4.94964004e-01
7.02056825e-01 -2.97341675e-01 1.05747124e-02 -7.98543215e-01
-1.07129836e+00 7.53944218e-01 1.94183625e-02 5.84379733e-01
9.09061551e-01 -1.67306475e-02 1.35685849e+00 6.96115792e-01
2.14429095e-01 -1.48956168e+00 1.21834673e-01 5.26939929e-01
6.28690898e-01 -1.28222167e+00 -4.89226431e-02 -7.76884377e-01
-5.47740579e-01 1.16930485e+00 6.54536545e-01 1.57105234e-02
7.01248467e-01 6.49670482e-01 -7.97356144e-02 -2.39386231e-01
-6.69531882e-01 -5.14270842e-01 4.82091069e-01 4.50603634e-01
6.33668602e-01 -3.70426208e-01 -3.12887669e-01 1.02251959e+00
4.72372361e-02 -2.78976083e-01 1.50009871e-01 1.22989869e+00
-4.95273173e-01 -1.25085652e+00 3.57565135e-02 7.52446830e-01
-8.46214235e-01 -6.59993529e-01 -6.33403003e-01 9.52256262e-01
-4.28711891e-01 7.30439603e-01 1.14445426e-01 -2.23091766e-01
2.12777883e-01 2.96428323e-01 1.99145690e-01 -4.65184569e-01
-5.71326613e-01 4.20386583e-01 1.06023066e-01 -3.63122493e-01
1.95645660e-01 -6.43347025e-01 -1.44126153e+00 4.92192572e-03
-3.69964272e-01 3.87115747e-01 8.20833027e-01 5.28610110e-01
4.92028087e-01 7.21093357e-01 5.10203242e-01 -2.87604064e-01
-5.36648154e-01 -8.45468044e-01 -7.07038522e-01 6.88574135e-01
1.60093885e-02 -3.42846572e-01 -4.64564897e-02 3.56991887e-01]
|
[9.8545560836792, 7.919076919555664]
|
255419cb-b2f2-4eff-a154-4c91429747c5
|
prompt-based-multi-modal-image-segmentation
|
2112.10003
| null |
https://arxiv.org/abs/2112.10003v2
|
https://arxiv.org/pdf/2112.10003v2.pdf
|
Image Segmentation Using Text and Image Prompts
|
Image segmentation is usually addressed by training a model for a fixed set of object classes. Incorporating additional classes or more complex queries later is expensive as it requires re-training the model on a dataset that encompasses these expressions. Here we propose a system that can generate image segmentations based on arbitrary prompts at test time. A prompt can be either a text or an image. This approach enables us to create a unified model (trained once) for three common segmentation tasks, which come with distinct challenges: referring expression segmentation, zero-shot segmentation and one-shot segmentation. We build upon the CLIP model as a backbone which we extend with a transformer-based decoder that enables dense prediction. After training on an extended version of the PhraseCut dataset, our system generates a binary segmentation map for an image based on a free-text prompt or on an additional image expressing the query. We analyze different variants of the latter image-based prompts in detail. This novel hybrid input allows for dynamic adaptation not only to the three segmentation tasks mentioned above, but to any binary segmentation task where a text or image query can be formulated. Finally, we find our system to adapt well to generalized queries involving affordances or properties. Code is available at https://eckerlab.org/code/clipseg.
|
['Alexander S. Ecker', 'Timo Lüddecke']
|
2021-12-18
| null |
http://openaccess.thecvf.com//content/CVPR2022/html/Luddecke_Image_Segmentation_Using_Text_and_Image_Prompts_CVPR_2022_paper.html
|
http://openaccess.thecvf.com//content/CVPR2022/papers/Luddecke_Image_Segmentation_Using_Text_and_Image_Prompts_CVPR_2022_paper.pdf
|
cvpr-2022-1
|
['referring-image-matting-refmatte-rw100', 'zero-shot-segmentation', 'referring-image-matting-keyword-based', 'one-shot-segmentation', 'referring-expression-segmentation', 'referring-image-matting-expression-based', 'multi-modal-image-segmentation']
|
['computer-vision', 'computer-vision', 'computer-vision', 'computer-vision', 'computer-vision', 'computer-vision', 'computer-vision']
|
[ 7.54994094e-01 4.07535642e-01 -3.46878283e-02 -6.27208292e-01
-1.16423500e+00 -9.83787000e-01 6.01904809e-01 1.67668402e-01
-4.47768033e-01 6.88975677e-02 -3.29280108e-01 -3.40813965e-01
1.59635738e-01 -7.92633474e-01 -9.31962430e-01 -3.84107053e-01
6.16319478e-01 8.73087049e-01 8.37315440e-01 -1.49677724e-01
1.62125543e-01 2.99789160e-01 -1.76555276e+00 3.66642058e-01
7.43816853e-01 1.07779884e+00 5.95850945e-01 9.39378619e-01
-3.67091864e-01 4.62722778e-01 -5.15141845e-01 -6.00618958e-01
3.07191789e-01 -3.30746204e-01 -1.23131013e+00 3.98655206e-01
5.69975019e-01 -2.58640379e-01 3.44266742e-01 8.49098861e-01
4.71755236e-01 1.68219611e-01 5.84664702e-01 -1.22256351e+00
-4.10689771e-01 7.83102810e-01 -3.41391452e-02 -1.46282315e-01
6.73307717e-01 3.15534890e-01 1.11285746e+00 -7.58981884e-01
8.64462674e-01 9.80873644e-01 4.11430150e-01 6.76014006e-01
-1.47365773e+00 -8.56207460e-02 2.43425876e-01 1.32844821e-01
-1.09180558e+00 -3.95305246e-01 5.79533935e-01 -6.09593809e-01
9.95143056e-01 4.52817500e-01 7.25886345e-01 1.02522433e+00
-2.57070005e-01 1.04044497e+00 8.85333598e-01 -5.19809842e-01
3.06990921e-01 3.16699833e-01 3.28043789e-01 5.58458388e-01
-4.55792755e-01 -1.44694567e-01 -1.02942310e-01 1.63392365e-01
5.31445622e-01 -3.07821721e-01 -2.70352542e-01 -3.26873839e-01
-9.86765027e-01 6.22915924e-01 2.01812491e-01 2.53083020e-01
-6.16053082e-02 4.93058898e-02 3.44769090e-01 3.36143970e-01
1.64430037e-01 4.93549854e-01 -6.40256584e-01 -1.83828220e-01
-1.21560431e+00 3.02733749e-01 1.15572655e+00 1.25094092e+00
1.14982784e+00 -4.60088313e-01 -4.89711851e-01 8.65503430e-01
3.14831175e-02 1.82474196e-01 2.28540018e-01 -1.05626225e+00
1.98630735e-01 5.31574070e-01 -1.65546332e-02 -5.95232248e-01
-4.09460276e-01 -1.70486227e-01 -1.44090116e-01 -1.33838747e-02
4.96680915e-01 -7.66509250e-02 -1.18305755e+00 1.79686737e+00
3.83366406e-01 -4.48299153e-03 -1.48840368e-01 8.82907689e-01
9.50603366e-01 5.11132598e-01 2.01497987e-01 9.91978645e-02
1.51109946e+00 -1.11336601e+00 -4.67683762e-01 -3.15490633e-01
7.79026330e-01 -6.80265367e-01 1.48785651e+00 4.56986517e-01
-1.18999696e+00 -6.50517285e-01 -7.91938722e-01 -5.46378076e-01
-7.02889919e-01 4.84156050e-03 3.84699047e-01 5.86867332e-01
-1.38855338e+00 4.75024164e-01 -7.53265321e-01 -6.40706182e-01
2.05511630e-01 5.92303574e-01 -4.65126634e-02 8.71563405e-02
-1.00721586e+00 7.11383045e-01 5.53629160e-01 -2.12129399e-01
-6.23986483e-01 -6.23588145e-01 -1.11807883e+00 2.51915045e-02
7.26345122e-01 -7.87175477e-01 1.66913450e+00 -1.37526917e+00
-1.59013593e+00 1.36455369e+00 -6.25923127e-02 -3.66975963e-01
5.69810212e-01 3.93593963e-03 4.47274968e-02 3.30735117e-01
2.99137026e-01 1.37523735e+00 8.03867817e-01 -1.15664434e+00
-4.07493681e-01 -1.65371343e-01 5.94231009e-01 1.80127740e-01
1.39449060e-01 2.88655341e-01 -1.10832787e+00 -5.47695339e-01
-1.48182422e-01 -1.02888739e+00 -2.38718688e-01 -2.61397753e-02
-8.30479860e-01 -1.97384432e-01 8.18051755e-01 -4.58427310e-01
1.00157619e+00 -2.16690445e+00 3.19525450e-01 1.11164078e-01
-1.99314505e-01 -5.21857943e-03 -3.74013990e-01 3.47371310e-01
-1.13803819e-01 2.59408385e-01 -7.50747383e-01 -5.75872779e-01
2.47773498e-01 3.72680128e-01 -1.80257723e-01 -7.18420297e-02
5.48447371e-01 1.14152372e+00 -7.67067015e-01 -7.80989707e-01
2.71360666e-01 2.08694056e-01 -6.94371998e-01 3.32354546e-01
-9.45476174e-01 5.48621774e-01 -2.91488886e-01 6.22448504e-01
4.82049316e-01 -3.59289557e-01 -1.24690443e-01 -1.97553396e-01
-1.30994648e-01 -1.39038144e-02 -1.11959708e+00 2.17087197e+00
-4.66054708e-01 3.99428725e-01 -8.69652256e-02 -1.03659320e+00
7.45986938e-01 2.24085048e-01 4.50014114e-01 -5.53066671e-01
2.38930449e-01 1.80411622e-01 -3.09721559e-01 -7.15210557e-01
5.42703509e-01 -1.34402746e-02 -5.32761693e-01 5.74882686e-01
5.01660645e-01 -6.00407243e-01 6.43822432e-01 3.15003633e-01
1.00022197e+00 7.37255156e-01 1.06930800e-01 -1.60378486e-01
5.11563957e-01 1.70505926e-01 2.45499477e-01 9.43912745e-01
4.60093915e-02 9.21545923e-01 7.78858721e-01 -1.18809313e-01
-9.21970308e-01 -8.21551681e-01 -2.99686909e-01 1.49136889e+00
2.96821624e-01 -5.10240734e-01 -1.01635623e+00 -6.73265278e-01
-3.99526775e-01 8.36815655e-01 -5.41065037e-01 2.07987264e-01
-3.97924423e-01 -4.12085146e-01 3.03682446e-01 4.14214760e-01
3.45936239e-01 -1.24757302e+00 -9.91971672e-01 1.08131193e-01
-9.46723148e-02 -1.38971221e+00 -4.87685382e-01 4.95803148e-01
-5.36208510e-01 -1.07919919e+00 -7.53317058e-01 -1.06845164e+00
7.22643793e-01 -3.30119163e-01 1.25615013e+00 1.69498622e-01
-3.47798347e-01 9.11576331e-01 -5.52330554e-01 -1.83119029e-01
-5.13779581e-01 4.40946192e-01 -7.05537736e-01 8.49497840e-02
1.41307712e-01 -4.37747002e-01 -4.60031182e-01 3.92791331e-01
-1.32295775e+00 4.70133603e-01 3.25662941e-01 4.22357053e-01
8.96172464e-01 -5.54777265e-01 2.03941956e-01 -1.24550796e+00
4.05295432e-01 -3.21344852e-01 -6.85596406e-01 4.02527660e-01
-1.26931027e-01 1.32520974e-01 5.48437774e-01 -4.41675276e-01
-9.94171858e-01 7.44769514e-01 -4.41302627e-01 -2.74589896e-01
-6.11276031e-01 5.23957133e-01 -3.36254776e-01 1.37167379e-01
5.59936106e-01 1.21388026e-01 -2.23991603e-01 -4.79909539e-01
7.71862805e-01 5.72731316e-01 7.47933805e-01 -8.38536739e-01
4.53378528e-01 1.16849378e-01 -3.64026368e-01 -6.17863774e-01
-9.05073404e-01 -6.02889895e-01 -1.03704131e+00 -1.97684631e-01
1.22245073e+00 -5.75238764e-01 -4.07492995e-01 3.66648585e-01
-1.36003494e+00 -1.00778735e+00 -5.69871902e-01 -1.84082374e-01
-1.04211605e+00 2.48976901e-01 -5.19649863e-01 -3.41241866e-01
-1.35694444e-01 -1.55980909e+00 1.58421206e+00 2.20684543e-01
-2.19286114e-01 -9.25026715e-01 -8.60703886e-02 1.79229245e-01
1.74877793e-01 1.93556994e-01 7.97959566e-01 -9.35981214e-01
-7.97796130e-01 -1.24514021e-01 -2.25577027e-01 2.22395778e-01
-2.33090520e-01 2.57198066e-01 -9.55593526e-01 8.34884346e-02
-7.68817589e-02 -5.32076538e-01 8.13951194e-01 1.30139738e-01
1.28028715e+00 -1.20858014e-01 -3.25315863e-01 8.03477585e-01
1.56173778e+00 1.09342650e-01 6.24595404e-01 2.49149323e-01
5.62445462e-01 5.91164589e-01 6.14639461e-01 1.44996151e-01
5.21878123e-01 1.01098740e+00 3.56841296e-01 -7.88052976e-02
2.65888032e-02 -4.82499078e-02 2.34019801e-01 4.21494603e-01
2.34955907e-01 -3.49314570e-01 -1.05169117e+00 4.70616460e-01
-1.82446611e+00 -7.49135494e-01 -1.62758529e-01 1.97714961e+00
1.05509663e+00 6.78355396e-02 2.93953568e-01 -3.25544626e-02
6.27020538e-01 4.09638882e-02 -6.08738303e-01 -5.99758446e-01
4.74489201e-03 4.38426405e-01 2.63538927e-01 6.51227236e-01
-1.29311574e+00 1.32801497e+00 6.18967772e+00 8.10250461e-01
-1.31680048e+00 1.85306400e-01 6.31906271e-01 1.94364637e-01
-4.03297365e-01 2.93539166e-01 -8.82236242e-01 3.35204333e-01
8.67543817e-01 -1.05128862e-01 3.52325946e-01 8.00343215e-01
-3.35369892e-02 -4.62689936e-01 -1.53397560e+00 7.76346326e-01
2.15375796e-01 -1.02472878e+00 7.34085366e-02 -3.05405885e-01
3.74061614e-01 1.70762446e-02 -7.66247362e-02 4.38930959e-01
1.69937536e-01 -7.29063511e-01 9.52796102e-01 5.45902073e-01
1.06762409e+00 -3.57314199e-02 2.12497592e-01 5.11492074e-01
-9.90102768e-01 1.50459945e-01 4.17141709e-03 1.24123380e-01
3.52475584e-01 2.18363136e-01 -6.70370221e-01 3.92743111e-01
3.99616212e-01 6.55060828e-01 -8.87489974e-01 1.13803470e+00
-4.16728079e-01 4.28368360e-01 -5.51599383e-01 1.93936765e-01
1.73673585e-01 -1.10921904e-01 4.28755909e-01 1.45793748e+00
2.08845422e-01 3.65220495e-02 4.87822533e-01 1.19537449e+00
1.33730859e-01 3.10951173e-01 -4.41424221e-01 1.70684353e-01
1.39984474e-01 1.48956978e+00 -1.13897407e+00 -5.92116714e-01
-5.21632552e-01 1.35311306e+00 1.97955534e-01 4.97596920e-01
-1.01740956e+00 -4.90029961e-01 -6.55800477e-02 6.73326030e-02
5.44604540e-01 -2.96904761e-02 -9.62408409e-02 -1.14117193e+00
-1.06217273e-01 -7.46110797e-01 3.04684192e-01 -1.14205694e+00
-7.75882185e-01 6.72500968e-01 3.25851858e-01 -9.36805665e-01
-5.04225791e-01 -6.43444896e-01 -5.90925217e-01 6.25181258e-01
-1.27286792e+00 -1.27556717e+00 -4.55312371e-01 6.26401365e-01
7.90076733e-01 3.82679820e-01 9.09159422e-01 3.57943654e-01
-5.82553208e-01 3.75359207e-01 -7.19687760e-01 1.92439422e-01
6.12859249e-01 -1.60239959e+00 3.64837855e-01 7.63807237e-01
3.59578669e-01 3.61644834e-01 7.75758088e-01 -4.40842569e-01
-9.46777046e-01 -9.64774072e-01 6.89377248e-01 -5.62222004e-01
4.79162484e-01 -6.42946184e-01 -9.09832716e-01 9.66114283e-01
3.73382390e-01 -1.47260167e-02 6.86232328e-01 -2.04557970e-01
-2.03847378e-01 2.20137417e-01 -8.93962562e-01 5.24998188e-01
9.38280582e-01 -5.03206193e-01 -4.44440663e-01 4.76388007e-01
1.03965497e+00 -8.56633425e-01 -7.86867797e-01 2.83761710e-01
2.84825027e-01 -1.07780516e+00 6.59693956e-01 -4.11277831e-01
4.68547612e-01 -2.31294468e-01 -4.21236046e-02 -8.94055605e-01
1.68094546e-01 -7.07206845e-01 2.04138964e-01 1.19536400e+00
7.07524419e-01 -3.46160889e-01 5.81364751e-01 9.46576893e-01
-3.50135177e-01 -8.51334453e-01 -7.00237095e-01 -5.58005035e-01
-8.09713230e-02 -9.17851686e-01 5.13580084e-01 5.60717762e-01
-4.74426299e-02 5.19215822e-01 1.50338501e-01 -9.95304286e-02
2.02152818e-01 3.69079679e-01 8.73193502e-01 -1.08850098e+00
-6.25843942e-01 -4.93893385e-01 -3.09100419e-01 -1.47184134e+00
1.91623330e-01 -1.18444407e+00 3.04306865e-01 -1.56405890e+00
1.12357564e-01 -4.90200698e-01 9.23462734e-02 8.57014775e-01
-1.12449899e-02 3.19364339e-01 4.34057236e-01 1.48989439e-01
-9.25186038e-01 6.08298033e-02 1.15101326e+00 -1.32979020e-01
-3.75920117e-01 1.01194061e-01 -5.81911206e-01 6.87980473e-01
7.30495095e-01 -3.37601036e-01 -4.31980640e-01 -4.79793698e-01
4.51098263e-01 3.07437181e-01 5.89279056e-01 -1.05360591e+00
3.09477508e-01 -9.10647120e-03 -1.48187429e-01 -5.28610945e-01
3.62039655e-01 -6.97064281e-01 1.47924140e-01 -6.90054074e-02
-5.84264100e-01 -2.56036729e-01 2.38458246e-01 2.57670075e-01
-1.43623665e-01 -7.72378981e-01 6.46294892e-01 -4.28812742e-01
-9.66309845e-01 2.25578576e-01 -1.81171402e-01 1.01063669e-01
1.09989083e+00 -4.87944454e-01 -8.97281095e-02 -2.90746003e-01
-1.38098645e+00 3.40763032e-01 8.01576257e-01 3.51657659e-01
3.26670170e-01 -8.81658614e-01 -2.68244684e-01 1.26478374e-01
2.52054572e-01 3.58355999e-01 6.20061569e-02 1.00279832e+00
-4.76623803e-01 1.86854705e-01 -4.25708219e-02 -9.06655133e-01
-1.02913010e+00 6.64227903e-01 4.91669387e-01 -1.90681145e-01
-4.51645017e-01 7.81287134e-01 3.12888473e-01 -4.99881774e-01
1.15416825e-01 -5.55340707e-01 -1.36687219e-01 2.01950833e-01
1.54178709e-01 -2.39194244e-01 2.35931738e-03 -7.06807613e-01
-8.51453096e-02 8.61459434e-01 1.66077301e-01 -4.17744309e-01
1.10413504e+00 -1.17347449e-01 -9.10428166e-02 6.32233560e-01
1.14968872e+00 -2.52542406e-01 -1.29547167e+00 -7.19072372e-02
2.21809909e-01 -1.15959980e-01 -2.97438353e-01 -9.13963497e-01
-9.15153682e-01 8.78254592e-01 2.34052569e-01 5.35745680e-01
1.32136035e+00 4.29156333e-01 6.76613271e-01 2.96673775e-01
3.33353490e-01 -1.03269267e+00 1.36649847e-01 5.72029650e-01
9.34544384e-01 -1.14445806e+00 -4.54740852e-01 -6.27412617e-01
-6.46134079e-01 1.26016927e+00 5.62022805e-01 -4.30984162e-02
4.93997574e-01 4.47528243e-01 1.06772490e-01 -2.08739772e-01
-6.47140205e-01 -6.97841823e-01 2.76257396e-01 5.99592805e-01
3.43845993e-01 -2.14976788e-01 -2.63745487e-01 6.83473408e-01
-3.12464714e-01 1.91613898e-01 4.55392748e-01 8.13999712e-01
-3.99128467e-01 -1.46717453e+00 -1.22353651e-01 4.32506651e-01
-3.31163138e-01 -1.61157593e-01 -4.57275599e-01 6.28555179e-01
3.90267700e-01 7.33373165e-01 2.16436490e-01 -2.16323927e-01
4.13957626e-01 3.60998780e-01 5.38695335e-01 -1.23059714e+00
-6.60052180e-01 1.96579888e-01 2.45626774e-02 -6.75421357e-01
-6.59782231e-01 -7.71420181e-01 -1.35014188e+00 5.18249512e-01
-2.65092760e-01 -1.10720791e-01 5.79271853e-01 9.66471076e-01
2.37334773e-01 3.23868722e-01 1.42258629e-01 -1.03445733e+00
-7.88604394e-02 -6.85726702e-01 -2.30455592e-01 5.54384530e-01
3.87323685e-02 -3.64722341e-01 -1.14937089e-01 7.13309348e-01]
|
[10.201324462890625, 1.1208386421203613]
|
338ef2a1-081d-43fe-b4d5-f1d7926ee566
|
escnet-gaze-target-detection-with-the
| null | null |
http://openaccess.thecvf.com//content/CVPR2022/html/Bao_ESCNet_Gaze_Target_Detection_With_the_Understanding_of_3D_Scenes_CVPR_2022_paper.html
|
http://openaccess.thecvf.com//content/CVPR2022/papers/Bao_ESCNet_Gaze_Target_Detection_With_the_Understanding_of_3D_Scenes_CVPR_2022_paper.pdf
|
ESCNet: Gaze Target Detection With the Understanding of 3D Scenes
|
This paper aims to address the single image gaze target detection problem. Conventional methods either focus on 2D visual cues or exploit additional depth information in a very coarse manner. In this work, we propose to explicitly and effectively model 3D geometry under challenging scenario where only 2D annotations are available. We first obtain 3D point clouds of given scene with estimated depth and reference objects. Then we figure out the front-most points in all possible 3D directions of given person. These points are later leveraged in our ESCNet model. Specifically, ESCNet consists of geometry and scene parsing modules. The former produces an initial heatmap inferring the probability that each front-most point has been looking at according to estimated 3D gaze direction. And the latter further explores scene contextual cues to regulate detection results. We validate our idea on two publicly available dataset, GazeFollow and VideoAttentionTarget, and demonstrate the state-of-the-art performance. Our method also beats the human in terms of AUC on GazeFollow.
|
['Jun Yu', 'Buyu Liu', 'Jun Bao']
|
2022-01-01
| null | null | null |
cvpr-2022-1
|
['scene-parsing']
|
['computer-vision']
|
[ 1.63787171e-01 1.46018550e-01 -9.06143114e-02 -5.35639942e-01
-5.34286201e-01 -6.27253771e-01 5.65948546e-01 6.72484338e-02
-4.24129725e-01 2.59372741e-01 2.72783488e-01 -1.51148677e-01
2.04966769e-01 -4.52783585e-01 -8.20350170e-01 -4.46589351e-01
1.84780642e-01 2.75590986e-01 4.39386725e-01 2.14327797e-02
6.54051840e-01 4.46797669e-01 -1.93278217e+00 1.75598171e-02
7.22805738e-01 1.05410743e+00 1.66581556e-01 8.23858917e-01
1.54250547e-01 4.54219908e-01 -3.16285968e-01 -3.38055015e-01
2.85045415e-01 -1.45949289e-01 -5.43223739e-01 2.71275729e-01
1.09865534e+00 -6.04144275e-01 2.31958199e-02 1.07579744e+00
4.67241973e-01 -7.40724951e-02 7.79407680e-01 -1.23392069e+00
-3.16809148e-01 -1.32920459e-01 -1.23222280e+00 5.44971168e-01
9.61811841e-01 5.31549096e-01 7.45416701e-01 -9.45323646e-01
5.41318595e-01 1.30352902e+00 5.72026253e-01 5.17480791e-01
-1.02804184e+00 -6.65559173e-01 5.07256150e-01 1.28385901e-01
-1.34873044e+00 -6.24608159e-01 8.40329349e-01 -5.62334836e-01
9.48745310e-01 2.93965321e-02 6.28943622e-01 9.25875902e-01
-1.09328590e-01 8.03525746e-01 1.17224240e+00 -5.35435259e-01
4.35050577e-02 3.13595608e-02 3.08688939e-01 8.38085175e-01
1.79266065e-01 1.58456072e-01 -8.70004594e-01 1.95558533e-01
6.33430004e-01 9.99716446e-02 -3.30796242e-01 -4.64055508e-01
-1.05576098e+00 4.80523169e-01 8.05951536e-01 -2.79859990e-01
-4.66228396e-01 1.59278419e-02 -2.91823477e-01 -2.91401982e-01
4.60155755e-01 2.52081193e-02 -2.49386236e-01 -1.44891724e-01
-9.18763995e-01 4.34704393e-01 3.16101581e-01 1.04979753e+00
8.46376300e-01 -5.92728019e-01 -1.65805459e-01 4.37629819e-01
6.89053416e-01 7.37776995e-01 -1.16267599e-01 -9.70556796e-01
6.60604656e-01 8.34584534e-01 3.52413863e-01 -1.08352697e+00
-5.43985248e-01 -1.34084567e-01 -2.68782049e-01 5.09068608e-01
7.29663908e-01 -1.27930671e-01 -1.06885266e+00 1.63893521e+00
8.29413891e-01 1.58923432e-01 -3.59754235e-01 1.26713169e+00
9.21000123e-01 2.68236816e-01 1.24212705e-01 9.53611434e-02
1.65322828e+00 -8.67133439e-01 -3.71744484e-01 -4.99649495e-01
4.06544268e-01 -6.04889810e-01 1.07468712e+00 3.57834816e-01
-1.21633804e+00 -5.30535579e-01 -7.44440675e-01 -3.79024446e-01
-1.42862454e-01 1.77150339e-01 4.64744925e-01 7.38128543e-01
-1.31066179e+00 -6.32211268e-02 -7.19882727e-01 -5.69497585e-01
7.12019920e-01 2.89421797e-01 -3.32544625e-01 -2.17142314e-01
-6.09912515e-01 6.35218620e-01 3.97984311e-02 5.19669540e-02
-9.00147200e-01 -6.01354539e-01 -1.01206291e+00 -5.49244024e-02
4.96127158e-01 -1.01145542e+00 1.36626661e+00 -5.87135732e-01
-1.20298290e+00 1.22029018e+00 -8.25703382e-01 -2.94740140e-01
5.59879184e-01 -3.77085418e-01 9.80369672e-02 2.84575343e-01
2.58436799e-01 1.14055800e+00 9.53815222e-01 -1.29424345e+00
-9.46904242e-01 -8.61629426e-01 3.96644920e-01 7.32970476e-01
-3.38902697e-03 1.17740057e-01 -9.24568415e-01 1.90039258e-02
2.62455612e-01 -7.70829082e-01 1.61944374e-01 2.52050400e-01
-9.73697841e-01 -4.96660471e-01 4.16565001e-01 -3.09000254e-01
1.03755617e+00 -1.89105284e+00 6.09688871e-02 -5.69026954e-02
5.71469188e-01 -9.96325165e-02 2.92745531e-01 -3.95350493e-02
1.33552328e-01 2.84578353e-02 1.30133599e-01 -8.12879920e-01
-6.60167411e-02 -4.91241753e-01 -4.82890159e-02 6.69299006e-01
2.44122133e-01 9.50303733e-01 -8.78864706e-01 -8.03364515e-01
4.68812078e-01 5.37895262e-01 -6.27788961e-01 1.83445707e-01
-1.47127181e-01 4.68896419e-01 -6.20617509e-01 8.31407666e-01
8.51033449e-01 -5.36906838e-01 -3.47038895e-01 -2.30058352e-03
-4.40649390e-01 9.95324850e-02 -8.52427185e-01 1.82909596e+00
-2.62898952e-02 7.11112499e-01 -3.19498569e-01 -1.37041956e-01
6.12120569e-01 -2.54841566e-01 1.06775738e-01 -7.03678012e-01
2.51895875e-01 -2.26944715e-01 -3.43615949e-01 -6.92061663e-01
4.92766440e-01 2.61159629e-01 4.37598266e-02 1.77465454e-01
-2.47508764e-01 9.32929963e-02 -2.55298048e-01 7.34013990e-02
7.85200715e-01 4.37173277e-01 3.75079900e-01 6.96402974e-03
5.21090448e-01 1.71478800e-02 1.10054910e-01 7.19619870e-01
-5.38184822e-01 9.78840828e-01 6.66289866e-01 -1.31915674e-01
-7.96672106e-01 -9.68201637e-01 -2.79782638e-02 1.01403892e+00
6.08483493e-01 -3.49346519e-01 -9.51421142e-01 -7.67944276e-01
-1.57255217e-01 7.39465535e-01 -1.11543834e+00 3.29480797e-01
-2.80676842e-01 -2.39642650e-01 2.55971283e-01 4.23679680e-01
6.98688388e-01 -7.47073293e-01 -1.20227110e+00 -5.80096304e-01
-1.58491299e-01 -1.12809062e+00 -5.84039211e-01 -4.10089254e-01
-6.20533168e-01 -1.46920407e+00 -7.54094124e-01 -4.31606531e-01
7.67104864e-01 7.50943363e-01 1.03827548e+00 -2.14205757e-02
-1.33794010e-01 4.03131962e-01 -1.27301261e-01 -6.06325984e-01
3.60706240e-01 -3.75368670e-02 -5.52312508e-02 7.92433023e-02
8.52842033e-01 -1.93866745e-01 -1.10739160e+00 2.08481982e-01
-6.39813021e-02 2.08966821e-01 4.37505782e-01 1.86039075e-01
6.49204910e-01 -3.97506446e-01 -1.20716244e-01 -6.67966425e-01
2.50335366e-01 -3.95809978e-01 -7.13414311e-01 1.72074422e-01
-3.07339877e-01 -2.17665792e-01 -1.87694490e-01 -8.66983458e-02
-1.02110934e+00 3.01036954e-01 1.42021835e-01 -7.59433031e-01
-7.49907553e-01 -8.57810229e-02 -1.66406572e-01 4.26473953e-02
8.42164397e-01 9.73118003e-03 -1.65664360e-01 -4.14075047e-01
3.25337827e-01 6.20674372e-01 4.74023223e-01 -2.33693719e-01
4.92244393e-01 8.05118740e-01 2.13687912e-01 -7.65789807e-01
-1.26592457e+00 -6.93465233e-01 -9.91840661e-01 -5.80348730e-01
1.28847587e+00 -9.50550973e-01 -1.29905903e+00 4.50804561e-01
-1.23665237e+00 -2.26591572e-01 2.12089226e-01 2.82164186e-01
-5.43082952e-01 2.70231098e-01 8.65910053e-02 -1.17497623e+00
-2.12809905e-01 -1.11177933e+00 1.87877321e+00 4.88785863e-01
-1.34271264e-01 -7.43030727e-01 3.67389135e-02 2.61437535e-01
-1.67472586e-01 5.12848735e-01 3.28784168e-01 -1.11992292e-01
-8.35395038e-01 -9.25260782e-02 -5.83680153e-01 -3.24065119e-01
-2.28824615e-01 -1.59458399e-01 -1.42979240e+00 1.07154310e-01
-2.44602069e-01 -1.52702257e-01 7.84038246e-01 9.09575224e-01
1.07033587e+00 1.27234384e-01 -7.04841852e-01 8.41551185e-01
1.39231384e+00 -2.09733620e-01 4.26852077e-01 4.21795458e-01
9.27609026e-01 9.74228859e-01 8.74498904e-01 2.45930746e-01
8.01783204e-01 8.12919974e-01 7.36680925e-01 5.03740571e-02
-2.55765676e-01 -3.93130332e-01 -2.81831361e-02 -2.02667102e-01
-2.67226070e-01 -3.76663238e-01 -1.28146231e+00 4.29969490e-01
-1.67996883e+00 -8.93917084e-01 -4.46457267e-01 2.18409920e+00
3.54145467e-01 2.37537846e-01 3.40765059e-01 -9.06279013e-02
6.77411735e-01 -2.27151215e-02 -8.33395898e-01 1.65093943e-01
1.06902517e-01 -3.10452163e-01 4.65792596e-01 5.61833501e-01
-1.12268376e+00 1.01823938e+00 5.91919851e+00 6.61820844e-02
-8.63144875e-01 -1.14329405e-01 6.36017263e-01 -4.94790792e-01
9.42934901e-02 -1.40923500e-01 -1.21475518e+00 4.79563057e-01
6.46383643e-01 2.98457503e-01 6.79684952e-02 7.84446061e-01
3.89008820e-01 -6.42159522e-01 -1.23759210e+00 1.44394290e+00
4.19485629e-01 -1.00679088e+00 -3.58323365e-01 3.91189069e-01
4.29011494e-01 2.05080658e-01 2.86127299e-01 -4.13340889e-02
-9.04531553e-02 -1.06097782e+00 9.67043400e-01 9.42927539e-01
1.00260246e+00 -6.50190830e-01 2.82087296e-01 5.37866831e-01
-9.17675793e-01 3.13013531e-02 -8.03873464e-02 -1.27526656e-01
1.81072399e-01 6.24371096e-02 -1.11723280e+00 2.11025640e-01
1.06719565e+00 9.43898857e-01 -9.99336362e-01 1.31303954e+00
-4.05900300e-01 2.73752004e-01 -4.63998646e-01 -1.15213089e-01
5.53135462e-02 1.77367017e-01 6.74279094e-01 8.64604533e-01
1.53404102e-01 2.33161628e-01 -1.23660281e-01 8.18456292e-01
9.76461843e-02 -2.35437080e-01 -6.09715581e-01 6.55876100e-01
5.78977764e-01 1.12302911e+00 -6.90841317e-01 -1.05632827e-01
-3.45883578e-01 9.38320696e-01 5.23611844e-01 5.44654787e-01
-7.38472998e-01 -8.01073834e-02 7.03453124e-01 5.05674422e-01
2.75220335e-01 -1.62784666e-01 -4.33447123e-01 -1.12963486e+00
1.04864083e-01 -3.77582222e-01 3.42832536e-01 -1.38590300e+00
-1.00445747e+00 6.13336504e-01 2.16428220e-01 -1.27764034e+00
-3.68161082e-01 -6.52809858e-01 -4.67343152e-01 1.17846942e+00
-1.92524791e+00 -1.26827991e+00 -9.96782124e-01 8.31867278e-01
7.28896856e-01 2.25497827e-01 5.30741334e-01 -2.17458069e-01
-5.43098271e-01 5.88857412e-01 -4.48109388e-01 -3.78281511e-02
6.65477097e-01 -1.32874727e+00 4.86032546e-01 9.53632176e-01
-2.67366990e-02 5.84945083e-01 7.30212748e-01 -6.65368855e-01
-1.14887202e+00 -7.73112476e-01 9.85502839e-01 -1.24062312e+00
1.93333805e-01 -5.65621138e-01 -4.94074702e-01 6.00025058e-01
1.14793584e-01 -5.47660626e-02 3.29547346e-01 3.39552343e-01
-3.52560788e-01 2.45107323e-01 -1.08161283e+00 6.57772541e-01
1.39286411e+00 -3.04730028e-01 -6.44850552e-01 1.98530689e-01
5.52871943e-01 -7.15236604e-01 -2.57965893e-01 1.20183468e-01
5.50360739e-01 -1.43101335e+00 1.09273231e+00 -4.71646190e-01
3.13034266e-01 -3.45637292e-01 3.30274589e-02 -7.88820744e-01
-1.84211284e-02 -4.36003566e-01 -3.62054139e-01 9.13539171e-01
1.85637355e-01 -2.95069277e-01 9.64344919e-01 6.99057162e-01
2.27157269e-02 -6.73855126e-01 -6.25652134e-01 -8.29953179e-02
-3.40898097e-01 -6.40291989e-01 7.07209885e-01 5.01174569e-01
-2.64977753e-01 4.12357241e-01 -7.29714781e-02 4.67985868e-01
1.02962220e+00 8.26983973e-02 1.21333241e+00 -1.41460872e+00
9.93704051e-02 -3.73192251e-01 -5.59042335e-01 -1.53416598e+00
-4.36983891e-02 -3.55138928e-01 8.44508633e-02 -1.51085770e+00
2.93012917e-01 -2.72590220e-01 -4.01935540e-02 3.47034872e-01
-4.83603805e-01 4.03630853e-01 1.78531572e-01 4.73680675e-01
-8.78441691e-01 1.82154208e-01 1.16738462e+00 2.80731678e-01
-3.82219046e-01 3.42753194e-02 -9.17291164e-01 9.04997289e-01
6.18360162e-01 -2.15438440e-01 -4.37292188e-01 -6.56593144e-01
2.18355805e-01 -1.04336413e-02 8.84230614e-01 -9.75202560e-01
4.12844777e-01 -1.21040970e-01 7.94468522e-01 -1.17062700e+00
6.82324231e-01 -5.95525503e-01 -5.40287137e-01 -8.94419104e-02
-2.02569470e-01 -3.47928777e-02 1.55438051e-01 9.61022079e-01
2.71386713e-01 -3.45018022e-02 6.08017623e-01 6.15278445e-02
-6.70778453e-01 5.18278420e-01 3.14784884e-01 -2.10650116e-02
1.23343992e+00 -6.64126277e-01 -2.68620312e-01 -3.85880262e-01
-6.09323263e-01 3.92758131e-01 8.72969151e-01 5.28425992e-01
7.59197950e-01 -9.28286076e-01 -4.56144840e-01 3.52374405e-01
3.48956048e-01 3.02743614e-01 3.77394527e-01 9.68282282e-01
-4.54574078e-01 6.52530849e-01 -2.26130746e-02 -1.28151858e+00
-1.53830147e+00 7.45425642e-01 4.96382415e-01 4.24817383e-01
-5.76830566e-01 1.11549509e+00 7.47258008e-01 5.38678057e-02
3.83132577e-01 -4.04474288e-01 -5.95629990e-01 -1.08354561e-01
7.69200921e-01 1.56646326e-01 -2.14124128e-01 -1.09021366e+00
-5.63992262e-01 1.10387635e+00 -9.09831896e-02 -2.26540267e-01
1.03364444e+00 -6.52305961e-01 2.73123294e-01 3.80534261e-01
9.77953494e-01 6.45855218e-02 -1.77554870e+00 -1.90462619e-01
-2.80169070e-01 -9.91007447e-01 8.19040462e-02 -6.96673870e-01
-5.41452408e-01 1.15106630e+00 6.97718561e-01 1.23073056e-01
1.22348142e+00 3.81782204e-01 2.01019228e-01 2.34367233e-02
3.16782951e-01 -5.73665202e-01 1.11575782e-01 2.52019972e-01
5.77211678e-01 -1.62834084e+00 -3.40867229e-02 -6.43941224e-01
-6.73395395e-01 9.27106917e-01 9.20467079e-01 -4.33653332e-02
6.02752686e-01 -2.75448561e-01 -5.17595485e-02 -6.21379435e-01
-5.53150356e-01 -6.66446447e-01 7.53742814e-01 8.92137170e-01
1.67578012e-01 -3.18212450e-01 3.98614049e-01 4.48620319e-02
-4.76587951e-01 -1.98019266e-01 2.15626299e-01 6.59312248e-01
-5.18516600e-01 -3.88529181e-01 -5.72375774e-01 2.26894662e-01
-3.78607929e-01 -3.97644117e-02 -5.40646672e-01 8.46708119e-01
9.15565565e-02 1.06771207e+00 3.72669786e-01 -2.14502662e-01
3.75587463e-01 -1.36387303e-01 7.08658338e-01 -7.18682706e-01
-1.94704100e-01 1.71847180e-01 -1.09615341e-01 -8.66999626e-01
-6.36886179e-01 -1.00132453e+00 -1.01566565e+00 -1.94469556e-01
-2.07633078e-01 -4.65322584e-01 7.34521270e-01 6.73749685e-01
5.66509724e-01 2.78771698e-01 4.06820536e-01 -1.20852494e+00
-4.92334180e-02 -9.91437078e-01 -1.67531237e-01 4.06064302e-01
8.11595798e-01 -1.00698686e+00 -2.45334089e-01 1.03506751e-01]
|
[14.098119735717773, 0.05728166177868843]
|
f9651b31-486d-4e4d-bf8e-0b9da1b728c7
|
labeling-cutting-grouping-an-efficient-text
|
1906.11894
| null |
https://arxiv.org/abs/1906.11894v2
|
https://arxiv.org/pdf/1906.11894v2.pdf
|
Labeling, Cutting, Grouping: an Efficient Text Line Segmentation Method for Medieval Manuscripts
|
This paper introduces a new way for text-line extraction by integrating deep-learning based pre-classification and state-of-the-art segmentation methods. Text-line extraction in complex handwritten documents poses a significant challenge, even to the most modern computer vision algorithms. Historical manuscripts are a particularly hard class of documents as they present several forms of noise, such as degradation, bleed-through, interlinear glosses, and elaborated scripts. In this work, we propose a novel method which uses semantic segmentation at pixel level as intermediate task, followed by a text-line extraction step. We measured the performance of our method on a recent dataset of challenging medieval manuscripts and surpassed state-of-the-art results by reducing the error by 80.7%. Furthermore, we demonstrate the effectiveness of our approach on various other datasets written in different scripts. Hence, our contribution is two-fold. First, we demonstrate that semantic pixel segmentation can be used as strong denoising pre-processing step before performing text line extraction. Second, we introduce a novel, simple and robust algorithm that leverages the high-quality semantic segmentation to achieve a text-line extraction performance of 99.42% line IU on a challenging dataset.
|
['Rolf Ingold', 'Lars Vögtlin', 'Vinaychandran Pondenkandath', 'Michele Alberti', 'Mathias Seuret', 'Marcus Liwicki']
|
2019-06-11
| null | null | null | null |
['text-line-extraction']
|
['computer-vision']
|
[ 4.44983274e-01 -4.25775260e-01 2.40526661e-01 -2.86338806e-01
-1.15206957e+00 -6.37750268e-01 7.47266829e-01 2.28652596e-01
-5.88783026e-01 6.89019442e-01 -2.39562839e-01 -1.23020247e-01
2.21419483e-01 -8.53435636e-01 -8.00424695e-01 -5.75499296e-01
4.03434008e-01 4.22054112e-01 5.13396263e-01 -8.43399316e-02
7.25222647e-01 8.02705824e-01 -1.13224435e+00 4.02066827e-01
1.09411693e+00 9.87405300e-01 -7.12666661e-02 7.33739734e-01
-4.57692444e-01 4.78794783e-01 -9.57924783e-01 -6.78232133e-01
1.97252214e-01 -4.18579787e-01 -4.82746631e-01 4.51353639e-01
9.00348425e-01 -3.03717643e-01 -2.43998230e-01 9.23587680e-01
5.22734523e-01 -1.49665192e-01 8.42467070e-01 -7.25925207e-01
-2.42270306e-01 5.59419215e-01 -9.59296763e-01 -2.60187566e-01
-2.08378136e-02 1.86079238e-02 7.53780186e-01 -9.31368768e-01
8.13656628e-01 8.59462500e-01 9.42737520e-01 2.56935269e-01
-1.09039068e+00 -3.47429752e-01 -3.12593043e-01 1.86291113e-01
-1.01371431e+00 -4.37820911e-01 9.72796977e-01 -5.35239995e-01
5.53707778e-01 1.92791864e-01 4.76391226e-01 8.98298979e-01
2.01439112e-01 1.09255505e+00 1.48264897e+00 -7.63256490e-01
2.33029112e-01 -1.38790190e-01 4.81063157e-01 8.19463611e-01
2.88857579e-01 -3.78564090e-01 -6.36420727e-01 3.95451307e-01
5.77869773e-01 -3.83546054e-01 -2.84954846e-01 7.76443779e-02
-1.03055048e+00 5.40586889e-01 5.30896634e-02 3.29689860e-01
-1.69368744e-01 8.53630081e-02 3.68846059e-01 -6.82672411e-02
5.26893497e-01 4.37186241e-01 -1.39864817e-01 -3.65067273e-01
-1.92208815e+00 1.79929078e-01 9.33234274e-01 8.47129405e-01
6.15486264e-01 -2.89292671e-02 -3.86225551e-01 1.04187357e+00
1.74165279e-01 7.17321932e-01 1.91113487e-01 -5.69963694e-01
6.05876327e-01 3.90709609e-01 -9.07398760e-02 -1.04998946e+00
-4.68614787e-01 -5.15211642e-01 -8.15287054e-01 3.73201221e-01
6.99182034e-01 -7.93013200e-02 -1.37950838e+00 8.25373411e-01
-3.09839547e-02 -3.77623141e-01 -1.53611809e-01 7.08925188e-01
6.33335412e-01 6.15191877e-01 -3.36596906e-01 -1.03086181e-01
1.56624389e+00 -1.39546680e+00 -9.22939241e-01 -2.13267729e-01
1.62744910e-01 -1.31367683e+00 1.02315509e+00 1.00273454e+00
-1.09559190e+00 -3.12193424e-01 -1.54973853e+00 -2.24511683e-01
-1.65502965e-01 5.70834577e-01 4.87844914e-01 8.16417217e-01
-5.91079533e-01 8.28175485e-01 -8.57313216e-01 -4.14717913e-01
7.58257210e-01 7.84342363e-03 -1.75312385e-01 -9.05660838e-02
-5.16158581e-01 6.15887702e-01 3.19736749e-01 1.56646699e-01
-4.63772207e-01 -6.56528354e-01 -3.97505164e-01 -6.80084080e-02
6.78540587e-01 -4.50401485e-01 1.04679680e+00 -5.82153738e-01
-1.74721277e+00 9.16026652e-01 -1.57704681e-01 -4.60634530e-01
1.25614250e+00 -8.61665547e-01 -3.08995724e-01 5.19180179e-01
-1.15405500e-01 2.37484336e-01 1.13785219e+00 -1.48611271e+00
-4.83151853e-01 -2.52056748e-01 -7.13208795e-01 -1.50088444e-01
-3.17797959e-01 -9.78888124e-02 -1.06749570e+00 -1.08548927e+00
3.80110770e-01 -6.30334020e-01 1.60636246e-01 1.61596149e-01
-9.16811168e-01 1.88206732e-01 9.18096185e-01 -9.95101750e-01
1.15220797e+00 -1.79821682e+00 -3.83211374e-01 2.81040311e-01
6.36525750e-02 2.80320942e-01 1.65786013e-01 4.43557352e-01
4.53937948e-01 2.82513499e-01 -8.48323941e-01 -6.26413643e-01
-1.36202499e-02 -2.19466940e-01 -3.48817497e-01 4.83589917e-01
3.29617292e-01 8.91395748e-01 -3.53289902e-01 -6.32166624e-01
1.78791687e-01 3.95811915e-01 -1.46996260e-01 -1.11687355e-01
-2.84329295e-01 8.86866357e-03 -1.20914653e-02 1.01186287e+00
8.62062633e-01 8.92490670e-02 -1.72635004e-01 -5.57593882e-01
-3.61014634e-01 -2.73141295e-01 -1.15338564e+00 1.69444335e+00
-1.05237784e-02 1.02090073e+00 -9.06152353e-02 -6.63730085e-01
1.21342134e+00 -2.49419227e-01 3.28090817e-01 -8.79347384e-01
2.68363267e-01 5.06818593e-01 -4.43664670e-01 -3.49324822e-01
9.32891786e-01 7.96048269e-02 -4.77455556e-02 3.28187853e-01
2.36636661e-02 -7.61896610e-01 5.30990958e-01 1.77726611e-01
8.32545877e-01 4.06278461e-01 -1.72152981e-01 -3.42238516e-01
4.48012620e-01 1.83881924e-01 4.15506750e-01 1.12178862e+00
-3.68012674e-02 1.06512296e+00 6.24791741e-01 -1.47882625e-01
-1.39114904e+00 -9.36839163e-01 -1.69438854e-01 5.03238440e-01
-1.05226189e-01 -3.93737257e-01 -1.25616288e+00 -5.72566867e-01
-1.22305132e-01 4.12512690e-01 -3.35278034e-01 2.51836568e-01
-7.89573193e-01 -1.03807867e+00 9.37142968e-01 4.51195866e-01
1.10199785e+00 -7.54477441e-01 -2.73715407e-01 1.83543667e-01
1.03019796e-01 -1.54750860e+00 -4.44913417e-01 1.77045867e-01
-9.81245995e-01 -8.58901322e-01 -1.02878463e+00 -7.35521019e-01
6.74350023e-01 -8.37634578e-02 1.01577449e+00 -2.57145055e-02
-6.19369149e-01 1.33640736e-01 -2.22933322e-01 -3.79316419e-01
-3.27274054e-01 1.77672118e-01 -4.36888844e-01 7.34070763e-02
-4.26833481e-02 1.70167331e-02 -6.36630356e-01 1.39977969e-02
-9.51683939e-01 1.59941837e-01 7.49965429e-01 8.37576747e-01
6.57156169e-01 1.38649061e-01 1.07617810e-01 -1.13380814e+00
6.94338262e-01 3.00225049e-01 -6.81268692e-01 1.90850556e-01
-6.95717692e-01 -7.55085517e-03 6.36191428e-01 5.42746158e-03
-1.20595539e+00 1.24762274e-01 -2.09240019e-01 2.83956468e-01
-2.28569299e-01 4.21186537e-01 -1.91144556e-01 -1.44613981e-01
4.89348203e-01 4.07564968e-01 -2.10658938e-01 -5.85157931e-01
4.16812748e-01 7.62021661e-01 1.12830102e+00 -6.06415868e-01
1.16284788e+00 8.93781364e-01 1.49387419e-01 -1.42246878e+00
-8.87002826e-01 -3.93556714e-01 -9.39401925e-01 -2.49037430e-01
8.98789167e-01 -5.69106221e-01 -4.99174058e-01 1.54058492e+00
-1.05661869e+00 -5.79106569e-01 4.50186469e-02 -1.58312935e-02
-3.94883692e-01 9.85120952e-01 -9.08505976e-01 -6.12517774e-01
-6.69315577e-01 -9.51498628e-01 1.03837538e+00 5.48348784e-01
-6.63793227e-03 -7.51923203e-01 -2.05745012e-01 8.60457659e-01
1.83295742e-01 3.29482228e-01 8.05224359e-01 -4.05944407e-01
-6.11169040e-01 -2.46989623e-01 -4.72555369e-01 4.74914610e-01
-5.71068153e-02 5.77512085e-01 -1.14369893e+00 -3.09230648e-02
-2.94618636e-01 -1.80470914e-01 1.30573678e+00 1.46968991e-01
1.15733814e+00 7.91566595e-02 -3.76600549e-02 8.20273042e-01
1.61817241e+00 5.53120412e-02 9.69308257e-01 8.15658808e-01
9.04913425e-01 3.10615122e-01 5.42095780e-01 5.12114525e-01
3.61894704e-02 5.36702991e-01 -1.53444767e-01 -4.21912223e-01
-4.69617397e-01 -1.51718417e-02 9.67423394e-02 8.57955158e-01
2.02616341e-02 -4.54931408e-01 -1.06258941e+00 3.67017388e-01
-1.57237601e+00 -8.08546901e-01 -7.48809218e-01 2.00509143e+00
1.07625997e+00 5.89784086e-01 -4.89404537e-02 6.49506629e-01
4.31335986e-01 9.89101827e-03 -3.16236317e-01 -3.51656258e-01
-6.16416574e-01 5.51009655e-01 7.89902925e-01 4.43705708e-01
-1.49082839e+00 1.34479439e+00 5.70665836e+00 1.00120890e+00
-1.29106247e+00 -2.71465778e-01 7.35894263e-01 4.20763850e-01
-6.34767711e-02 -2.60428786e-01 -8.73566091e-01 4.66568530e-01
4.45980459e-01 4.64617431e-01 -1.73100576e-01 3.00955743e-01
2.61831433e-01 -5.81417441e-01 -6.50440097e-01 1.03843689e+00
4.96405065e-01 -1.39735341e+00 -1.17174052e-02 -3.51896435e-01
9.22833502e-01 -3.44429463e-01 -7.75198340e-02 -1.57247558e-01
-1.17810719e-01 -9.53403831e-01 9.23376441e-01 7.87306666e-01
6.27180636e-01 -7.43544638e-01 5.64272583e-01 -5.71098886e-02
-7.50735819e-01 4.03828055e-01 3.59777967e-03 2.65549570e-01
3.03902984e-01 1.25332940e+00 -7.01793373e-01 6.53717637e-01
5.12637913e-01 6.49692297e-01 -9.05595720e-01 1.46816206e+00
-4.94994581e-01 8.62573504e-01 -4.19260591e-01 1.08376950e-01
3.25010210e-01 -3.48083854e-01 3.16766709e-01 1.75631809e+00
1.90821692e-01 -3.58024359e-01 6.02597278e-03 8.28315377e-01
-3.22847039e-01 2.48190194e-01 8.16453919e-02 -2.66190320e-01
2.43229374e-01 1.40004134e+00 -1.47430789e+00 -3.22301567e-01
-2.52389222e-01 1.51401317e+00 1.06818392e-03 4.33960527e-01
-6.47197425e-01 -9.70600307e-01 -3.93491685e-02 -8.05317983e-02
4.83868033e-01 -6.63349926e-01 -1.04769015e+00 -1.06065512e+00
3.88344318e-01 -7.91504145e-01 -8.00039172e-02 -4.68003154e-01
-1.08329928e+00 2.40648523e-01 -6.69904947e-01 -9.40371156e-01
4.17085975e-01 -8.67411196e-01 -5.27087927e-01 5.57424426e-01
-1.67918670e+00 -1.39331532e+00 -5.62457621e-01 1.66030392e-01
7.33735621e-01 -1.36591583e-01 4.01685119e-01 2.69635797e-01
-7.26489484e-01 6.16715312e-01 6.93933666e-01 5.63817382e-01
9.45856929e-01 -1.43076003e+00 5.46950519e-01 1.23988903e+00
2.42936954e-01 1.31733835e-01 5.70974410e-01 -7.25794733e-01
-1.42860186e+00 -8.75854254e-01 6.16575956e-01 -1.51613072e-01
6.12520456e-01 -6.53415620e-01 -8.93458724e-01 2.42662713e-01
1.62561923e-01 -3.66561145e-01 4.19163704e-01 -3.43481213e-01
-1.45618841e-01 -1.74588293e-01 -1.00691009e+00 7.36595929e-01
4.55914021e-01 -3.32664192e-01 -5.45146286e-01 1.30093321e-01
1.51573405e-01 -4.30509657e-01 -8.16799641e-01 3.31739902e-01
6.66196823e-01 -1.08741271e+00 8.96160603e-01 3.55988406e-02
8.16576123e-01 -2.61060476e-01 2.03359678e-01 -9.55792665e-01
1.17021739e-01 -7.85950482e-01 -7.57476538e-02 1.64191794e+00
4.46487188e-01 -1.70437872e-01 1.01532483e+00 2.75725931e-01
-2.77635843e-01 -5.32859147e-01 -5.68306625e-01 -7.03249991e-01
1.94848090e-01 -4.97143656e-01 3.56045067e-02 7.68536091e-01
-4.82076943e-01 9.20558199e-02 -4.00323987e-01 -1.80573121e-01
1.08346665e+00 2.49456868e-01 7.68620253e-01 -1.09648001e+00
-1.00280777e-01 -7.83398449e-01 -2.94420689e-01 -1.13367867e+00
-1.40128911e-01 -6.27012610e-01 3.57057512e-01 -1.71574295e+00
1.33031577e-01 -1.79418355e-01 2.16836855e-01 2.55937099e-01
-3.03911656e-01 6.53808892e-01 4.76153672e-01 2.57452041e-01
-3.49297762e-01 3.51180881e-01 1.24399507e+00 -5.61054111e-01
-1.42944247e-01 -2.06121981e-01 -3.33530694e-01 9.19909179e-01
7.00330436e-01 -2.36981690e-01 2.32898042e-01 -4.96514946e-01
1.03209682e-01 -4.01872456e-01 1.97030827e-01 -1.33143437e+00
3.61487776e-01 2.31490418e-01 7.82701075e-01 -1.04469502e+00
1.74671426e-01 -5.13094842e-01 -6.27452135e-01 4.93846327e-01
6.07602783e-02 -5.34674466e-01 4.17806685e-01 3.45776558e-01
-1.28859714e-01 -4.60271776e-01 8.64421368e-01 6.83218911e-02
-9.63194609e-01 -4.74104024e-02 -4.49208468e-01 1.30267933e-01
9.22642708e-01 -4.76186007e-01 -5.62044442e-01 -8.59144330e-02
-4.27501827e-01 -2.15187408e-02 5.98025858e-01 1.25985056e-01
6.89793348e-01 -7.60760903e-01 -8.41374516e-01 2.39920542e-02
-1.03257529e-01 6.91284910e-02 -2.22870372e-02 7.51367509e-01
-1.41377223e+00 2.28326842e-01 -1.36022165e-01 -7.43731797e-01
-1.30533934e+00 -7.19103962e-02 1.34634823e-01 -1.40090823e-01
-8.28988254e-01 8.22469473e-01 -4.76383179e-01 -3.54313478e-02
1.91454142e-01 -3.13015103e-01 7.20374659e-02 3.90284181e-01
2.91078746e-01 6.33918822e-01 4.25784051e-01 -2.35219598e-01
-1.91597253e-01 1.03546214e+00 -2.97895193e-01 -3.65986705e-01
1.42132437e+00 1.73871711e-01 -9.01197493e-02 4.67763186e-01
9.22697663e-01 3.40885311e-01 -1.34303057e+00 2.09724549e-02
3.79687399e-01 -3.88576388e-01 9.75699276e-02 -1.04661393e+00
-1.04073191e+00 9.62463260e-01 4.51617539e-01 -1.87974438e-01
1.05553126e+00 -3.68674099e-01 1.09877539e+00 4.45609242e-01
4.35649231e-02 -1.74910796e+00 2.46385276e-01 4.84551221e-01
6.94531977e-01 -1.41462886e+00 4.62894112e-01 -5.32560468e-01
-3.77925992e-01 1.69547045e+00 2.77278453e-01 -1.47351116e-01
3.15538138e-01 6.51204705e-01 3.59695852e-01 -1.17781155e-01
-1.05477571e-02 -2.38510855e-02 3.95736217e-01 1.99695736e-01
5.78719497e-01 -1.17339022e-01 -5.00470817e-01 4.64832544e-01
-3.54064882e-01 4.13340293e-02 8.00369740e-01 1.21694732e+00
-5.03996193e-01 -9.91288304e-01 -6.79886758e-01 4.40350950e-01
-6.94833875e-01 -2.96112537e-01 -6.65615618e-01 7.45877504e-01
-2.90153533e-01 8.54225039e-01 2.53778941e-04 -2.95724673e-03
2.97445983e-01 8.51134211e-02 6.56266809e-01 -1.47865310e-01
-6.59512281e-01 3.30026358e-01 7.61186853e-02 -3.43712032e-01
-2.34261721e-01 -6.04532599e-01 -1.30994284e+00 -1.35673955e-01
-3.43023926e-01 -3.46045017e-01 9.67842937e-01 9.95592177e-01
-3.46158929e-02 7.09857464e-01 3.63180101e-01 -6.68306053e-01
-3.22123438e-01 -8.39768350e-01 -5.93302488e-01 6.40648007e-01
2.14715615e-01 -1.24386005e-01 -1.32715717e-01 5.73350728e-01]
|
[11.808948516845703, 2.6025524139404297]
|
a405aab6-bd99-4ac7-905e-2e41785177c8
|
infrared-and-visible-image-fusion-with-resnet
|
1806.07119
| null |
https://arxiv.org/abs/1806.07119v7
|
https://arxiv.org/pdf/1806.07119v7.pdf
|
Infrared and Visible Image Fusion with ResNet and zero-phase component analysis
|
Feature extraction and processing tasks play a key role in Image Fusion, and the fusion performance is directly affected by the different features and processing methods undertaken. By contrast, most of deep learning-based methods use deep features directly without feature extraction or processing. This leads to the fusion performance degradation in some cases. To solve these drawbacks, we propose a deep features and zero-phase component analysis (ZCA) based novel fusion framework is this paper. Firstly, the residual network (ResNet) is used to extract deep features from source images. Then ZCA is utilized to normalize the deep features and obtain initial weight maps. The final weight maps are obtained by employing a soft-max operation in association with the initial weight maps. Finally, the fused image is reconstructed using a weighted-averaging strategy. Compared with the existing fusion methods, experimental results demonstrate that the proposed framework achieves better performance in both objective assessment and visual quality. The code of our fusion algorithm is available at https://github.com/hli1221/imagefusion_resnet50
|
['Xiao-Jun Wu', 'Hui Li', 'Tariq S. Durrani']
|
2018-06-19
| null | null | null | null |
['infrared-and-visible-image-fusion']
|
['computer-vision']
|
[ 1.38651207e-01 -5.24155617e-01 1.82670295e-01 -2.72267163e-01
-5.51049531e-01 8.08094963e-02 5.79318762e-01 1.76087081e-01
-5.04809737e-01 5.09939313e-01 2.31367752e-01 2.51767278e-01
-1.41859069e-01 -8.56137395e-01 -2.51765043e-01 -1.06067133e+00
3.03864866e-01 -4.11821842e-01 1.71809897e-01 -1.84548661e-01
1.16483934e-01 3.35885823e-01 -1.56782198e+00 -5.65729514e-02
1.21152306e+00 1.52792454e+00 3.88169378e-01 2.99819797e-01
-1.27735943e-01 7.00264275e-01 -4.56603080e-01 -2.87885457e-01
3.69593501e-01 -4.35753196e-01 -3.01093847e-01 2.14482009e-01
-3.80922817e-02 -3.80804807e-01 -5.01962423e-01 1.55370665e+00
6.75342619e-01 3.63250732e-01 3.53913844e-01 -1.21299732e+00
-5.52495778e-01 3.18608135e-01 -8.71013880e-01 4.50812429e-01
1.63402781e-01 1.00164987e-01 4.81998056e-01 -1.16121686e+00
-3.70812640e-02 1.08354557e+00 4.43872422e-01 6.55963421e-02
-8.20487559e-01 -9.37942743e-01 -9.27011147e-02 4.89994437e-01
-1.53344786e+00 -6.94307029e-01 9.16668296e-01 -2.05353707e-01
4.69747722e-01 6.07581548e-02 6.42039180e-01 4.91348684e-01
5.19000173e-01 6.74308836e-01 1.23598504e+00 -2.65141666e-01
-8.07186589e-03 -1.36699021e-01 1.24647096e-02 7.75631905e-01
2.58108944e-01 1.85986366e-02 -4.01094645e-01 1.60898075e-01
7.16690421e-01 5.50749958e-01 -6.34338081e-01 3.12034450e-02
-1.39900696e+00 5.86297989e-01 9.26133811e-01 4.87184316e-01
-8.93432736e-01 -9.20982212e-02 2.49947131e-01 2.89214075e-01
4.04012054e-01 -1.56755105e-01 -1.40228704e-01 2.42759317e-01
-1.06583953e+00 2.07639366e-01 1.47666126e-01 3.85761201e-01
1.07354820e+00 1.67768344e-01 -2.52647221e-01 9.11906183e-01
4.32669640e-01 5.30800700e-01 8.76453578e-01 -9.23204601e-01
2.64500648e-01 7.61783063e-01 -1.10218614e-01 -1.34795129e+00
-4.04432505e-01 -5.51461935e-01 -1.21502292e+00 5.65033853e-01
7.71050751e-02 -2.81845599e-01 -1.05330586e+00 1.50083613e+00
3.08836102e-01 3.78839463e-01 2.63349921e-01 1.06721413e+00
1.18163300e+00 6.85510397e-01 2.10032552e-01 -3.07459742e-01
1.46398842e+00 -8.63434255e-01 -1.08869696e+00 1.25070781e-01
6.97379708e-02 -1.19786990e+00 5.54295421e-01 2.55240530e-01
-1.15116620e+00 -9.37715471e-01 -1.34811795e+00 -3.58304344e-02
-4.60832238e-01 2.32268825e-01 5.08078158e-01 3.11049640e-01
-1.02355540e+00 5.06182551e-01 -9.26837862e-01 -7.69744068e-02
5.21755457e-01 4.82373536e-01 -5.58828175e-01 -1.71268478e-01
-1.14191794e+00 1.00648785e+00 5.64925134e-01 4.01726156e-01
-5.05418777e-01 -4.82815653e-01 -1.00707626e+00 1.79227680e-01
2.10173219e-01 -7.67617464e-01 1.03867292e+00 -9.10577834e-01
-1.37059581e+00 2.81075835e-01 -2.13604942e-01 -2.85196573e-01
2.86117315e-01 -2.11453393e-01 -4.67694163e-01 2.24131674e-01
-2.99775451e-02 6.23803079e-01 9.94856179e-01 -1.01493227e+00
-8.89642000e-01 -2.68863738e-01 -6.51316792e-02 5.54930806e-01
-3.19098890e-01 7.83165544e-02 -4.53113943e-01 -7.47263908e-01
3.78797799e-01 -4.89628643e-01 -2.95474619e-01 -1.61201164e-01
-1.24878593e-01 -1.11373715e-01 8.21500123e-01 -8.38032126e-01
1.14714777e+00 -2.24761415e+00 9.33539867e-02 2.00266853e-01
5.31391919e-01 4.11534429e-01 -6.32113069e-02 3.21537144e-02
-2.06538424e-01 -2.67619997e-01 -2.86924541e-01 -2.77547628e-01
-2.86860287e-01 -1.95262358e-01 5.24379611e-02 5.38974643e-01
2.65393943e-01 8.06317985e-01 -5.86907804e-01 -7.37292647e-01
8.48176599e-01 8.76573980e-01 -1.94651261e-01 3.19370121e-01
4.78398651e-01 4.42395061e-01 -4.08866078e-01 6.37734890e-01
1.17907536e+00 -1.20137282e-01 -3.34955871e-01 -7.26032794e-01
-1.63805068e-01 -2.39070088e-01 -1.23978686e+00 1.82557392e+00
-3.55497420e-01 3.40715408e-01 1.90232798e-01 -9.52431619e-01
9.17998374e-01 3.88978690e-01 6.34041011e-01 -6.88797772e-01
7.34737098e-01 6.18881211e-02 1.16738252e-01 -4.01066691e-01
4.72376764e-01 -5.09553701e-02 1.44102216e-01 3.55557017e-02
3.29241812e-01 -3.89537551e-02 1.35743320e-01 1.48537576e-01
8.46718311e-01 -1.85387637e-02 5.51613748e-01 -3.57414037e-02
9.39579725e-01 -3.50740761e-01 8.82981420e-01 2.95497835e-01
-3.69330317e-01 6.74944103e-01 -6.26113638e-02 -3.48730326e-01
-7.14774132e-01 -9.99411464e-01 -8.98193792e-02 5.47131836e-01
4.76180613e-01 -3.72288376e-01 -6.49031937e-01 -2.42191121e-01
-2.02551931e-01 2.46745944e-01 -5.44075608e-01 -4.47344869e-01
-2.92849511e-01 -8.37713957e-01 -1.02489046e-03 3.29125345e-01
1.05902052e+00 -1.03065467e+00 -5.74585557e-01 3.03001940e-01
-3.48905623e-01 -9.76159751e-01 -2.77575582e-01 -1.00638933e-01
-7.25448370e-01 -1.00079918e+00 -1.00111365e+00 -6.28250062e-01
6.68784797e-01 6.34857833e-01 4.92198437e-01 2.78399020e-01
-1.59406275e-01 -1.60837490e-02 -4.11255419e-01 -3.42406094e-01
-1.91503186e-02 -1.62319675e-01 -3.29038166e-02 3.05556267e-01
3.92270088e-01 -6.32447720e-01 -1.02260399e+00 -7.51496926e-02
-1.01472354e+00 2.63028622e-01 9.02940214e-01 7.29225457e-01
4.82449651e-01 4.32159573e-01 5.60483456e-01 -1.47667587e-01
7.90156603e-01 -3.71103019e-01 -6.19995236e-01 9.81518030e-02
-4.33088928e-01 -1.64152279e-01 4.43218350e-01 -1.45851552e-01
-1.32322848e+00 2.50867978e-02 -1.86272070e-01 -5.71841657e-01
-3.68853398e-02 6.15187943e-01 -2.50377744e-01 -2.57249534e-01
1.25590011e-01 3.68849933e-01 2.47074753e-01 -4.00618732e-01
2.04919249e-01 5.47428310e-01 6.80652618e-01 -5.27143739e-02
8.48157644e-01 3.41067582e-01 -1.38429582e-01 -4.63409603e-01
-5.67736149e-01 -1.95419386e-01 -4.06369776e-01 -3.59097213e-01
1.07532704e+00 -1.20599842e+00 -7.31728911e-01 8.67811501e-01
-8.86047959e-01 2.24515945e-01 -9.54275355e-02 9.59586442e-01
-2.19816074e-01 4.13847029e-01 -5.12792706e-01 -5.20262420e-01
-7.43329704e-01 -1.38201702e+00 6.59712553e-01 9.14551973e-01
3.37764859e-01 -6.18464112e-01 -1.92182064e-01 1.00658402e-01
6.55592740e-01 2.81767368e-01 3.48392218e-01 -2.65881777e-01
-3.12915474e-01 -2.90439397e-01 -4.98795092e-01 6.16692007e-01
4.29634094e-01 -7.10838959e-02 -8.35686386e-01 -1.67532772e-01
2.82120079e-01 -2.92820204e-03 1.12213743e+00 6.35145187e-01
1.07423055e+00 1.19510572e-02 -1.32812187e-01 7.62900531e-01
1.55193293e+00 2.53431141e-01 5.81396043e-01 1.76414102e-01
6.56403661e-01 2.53105938e-01 6.22452199e-01 5.01398146e-01
4.02819127e-01 3.58240426e-01 4.32604492e-01 -4.82316136e-01
-2.84522623e-01 1.77167490e-01 2.28378132e-01 1.01477623e+00
-2.05768958e-01 9.29713398e-02 -7.06158876e-01 4.23827559e-01
-1.86837351e+00 -7.68061519e-01 1.11622095e-01 1.96399999e+00
6.46303773e-01 -1.76267661e-02 -3.65653306e-01 4.15291011e-01
9.02948201e-01 2.60954440e-01 -3.71024311e-01 1.45540014e-01
-2.21914440e-01 4.79657024e-01 4.57713604e-01 3.56772840e-01
-1.28909957e+00 5.96584260e-01 5.24349117e+00 9.47489560e-01
-1.25941622e+00 2.93798417e-01 4.21932697e-01 -6.05497509e-02
9.18458477e-02 -1.58075377e-01 -2.44406909e-01 6.21774912e-01
5.70232749e-01 -2.82732725e-01 3.38694304e-01 3.27463984e-01
2.24378243e-01 -3.45136911e-01 -3.45808804e-01 1.35631752e+00
5.02068847e-02 -1.10206759e+00 -1.63437098e-01 -1.11050755e-01
5.56389332e-01 8.97652507e-02 6.55517727e-02 1.10982150e-01
2.69984961e-01 -6.93157971e-01 4.81236935e-01 8.57579231e-01
4.83628660e-01 -1.05361462e+00 1.07142210e+00 -1.80621017e-02
-1.39448702e+00 -1.72458842e-01 -4.37967330e-01 1.13508403e-02
2.14810967e-01 9.85565186e-01 -2.23742068e-01 9.73123252e-01
8.02010417e-01 8.11874092e-01 -6.58327579e-01 1.21789110e+00
-2.05044270e-01 2.64894038e-01 -2.53720134e-01 5.77168286e-01
-6.02900758e-02 -2.47457638e-01 4.23171639e-01 9.42395568e-01
6.79528296e-01 3.06587815e-01 1.30858555e-01 4.64701265e-01
-1.48346066e-01 1.38165802e-01 -3.22421759e-01 3.12542379e-01
3.91682476e-01 1.81717682e+00 -7.79877603e-01 -5.94801247e-01
-5.85557342e-01 8.60210478e-01 -8.93713310e-02 3.73084664e-01
-7.54363716e-01 -7.30582714e-01 6.99033082e-01 -2.95308053e-01
3.53067219e-01 -1.29936025e-01 -8.01985785e-02 -1.48841131e+00
5.35327569e-03 -6.90705061e-01 1.94431901e-01 -8.40837538e-01
-1.06033444e+00 7.16574013e-01 -8.89345556e-02 -1.47634017e+00
3.91572639e-02 -1.98306486e-01 -6.93940759e-01 1.11390746e+00
-1.81551802e+00 -9.69531238e-01 -8.12402189e-01 8.81178677e-01
3.68680388e-01 -3.13123792e-01 5.02035499e-01 5.17970622e-01
-6.76330566e-01 3.61692727e-01 -3.97951864e-02 8.69891793e-02
6.73238933e-01 -8.57773066e-01 -5.16818240e-02 1.17887282e+00
-3.72142762e-01 3.85423094e-01 4.93625879e-01 -4.65536505e-01
-1.12643349e+00 -9.90403831e-01 4.85996187e-01 4.15348321e-01
3.95707607e-01 -4.28635348e-03 -9.79956806e-01 2.92099148e-01
7.54482985e-01 4.25775141e-01 6.10848904e-01 -5.28049529e-01
5.36755584e-02 -4.37763602e-01 -1.32481349e+00 3.77020866e-01
5.22015393e-01 -2.18743473e-01 -5.23869097e-01 -9.99520048e-02
6.23197317e-01 -3.46846908e-01 -1.15061319e+00 7.13829994e-01
5.54154813e-01 -1.04623985e+00 8.18920970e-01 2.26391137e-01
3.45967233e-01 -8.52667689e-01 -1.78614601e-01 -1.33033705e+00
-5.80495477e-01 -1.83340177e-01 2.93321479e-02 1.37926924e+00
-1.15328491e-01 -8.71522486e-01 2.44715124e-01 3.47121149e-01
-8.43043774e-02 -6.47112608e-01 -8.09595585e-01 -2.49633953e-01
-4.09198105e-01 -6.87721819e-02 8.09435248e-01 8.70492101e-01
-2.23370567e-01 4.63267304e-02 -1.80256695e-01 1.46265164e-01
8.17836165e-01 -5.80891930e-02 4.42946672e-01 -1.01851869e+00
1.44237190e-01 -4.78718817e-01 -6.79560661e-01 -5.33942342e-01
6.30719261e-03 -7.89008021e-01 -5.87049015e-02 -1.61307907e+00
1.97000727e-01 -2.81224281e-01 -9.33763504e-01 4.85417277e-01
-5.00894010e-01 5.02309680e-01 4.65306610e-01 2.99684405e-01
-4.76502925e-01 8.75687897e-01 1.23381579e+00 -1.32704839e-01
-6.02375753e-02 -2.24485055e-01 -8.64842057e-01 7.73673475e-01
9.56786215e-01 -2.74211437e-01 -3.18839610e-01 -4.28014278e-01
-3.09772789e-01 7.55191371e-02 3.57354939e-01 -1.51754344e+00
4.62197274e-01 4.11015898e-02 1.02487230e+00 -6.82434320e-01
4.48706448e-01 -1.01838541e+00 1.99565947e-01 5.38668394e-01
5.81992120e-02 1.61376864e-01 1.19649246e-01 1.21446565e-01
-6.19043410e-01 -7.08194450e-02 7.27455735e-01 -8.13413113e-02
-7.19314933e-01 3.49190623e-01 -2.99931914e-01 -6.47477090e-01
1.08682132e+00 -2.15925604e-01 -1.97966054e-01 -2.41422981e-01
-6.37304664e-01 2.16399312e-01 3.35317641e-01 3.86905402e-01
9.69488084e-01 -1.63095534e+00 -8.67378712e-01 3.25278819e-01
-4.63566743e-02 5.01864683e-03 6.15675151e-01 1.18032813e+00
-3.49136144e-01 -4.56764456e-03 -5.34083486e-01 -4.18609202e-01
-1.23531270e+00 5.05178392e-01 2.58565426e-01 -1.81521162e-01
-5.11135876e-01 5.79631984e-01 2.89271086e-01 1.32690132e-01
-4.61561605e-02 -2.19467744e-01 -4.75394487e-01 1.66895241e-01
7.81744182e-01 2.82701194e-01 1.75188914e-01 -9.87008154e-01
-3.39272469e-01 6.61307454e-01 -2.31386051e-01 1.15318429e-02
1.31818247e+00 -3.46165717e-01 -2.66887218e-01 -8.71215612e-02
1.24261940e+00 -2.46669590e-01 -1.14482820e+00 -5.86705387e-01
-5.67537487e-01 -5.70486546e-01 4.99093950e-01 -5.52214026e-01
-1.72925353e+00 8.10392857e-01 1.10693967e+00 -8.69033113e-02
1.86300206e+00 -3.18051845e-01 9.70063686e-01 -9.11052898e-02
-2.27680430e-03 -7.01008081e-01 -2.73323834e-01 6.88242167e-02
8.56225848e-01 -1.37965655e+00 2.83113062e-01 -1.55423045e-01
-4.06274915e-01 9.83850658e-01 4.93383139e-01 -4.50933486e-01
9.82952058e-01 2.80568779e-01 1.91521361e-01 -9.98580828e-02
-4.57402110e-01 -3.54746789e-01 2.81850547e-01 3.39572608e-01
3.34429145e-01 -9.45003238e-03 -5.01506984e-01 6.10824525e-01
-1.50196657e-01 2.75591344e-01 2.71559626e-01 9.32185769e-01
-5.06683886e-01 -9.36057746e-01 -5.32399893e-01 6.34800255e-01
-6.93746030e-01 -8.74660388e-02 2.66073316e-01 3.04913759e-01
4.34887558e-01 1.21522808e+00 8.95832181e-02 -7.08630800e-01
2.11354420e-01 -2.42412671e-01 4.31883037e-01 -1.08283550e-01
-6.12422764e-01 2.78611034e-01 -4.78393346e-01 -5.53201973e-01
-8.63744318e-01 -5.12766719e-01 -1.28109539e+00 -3.78165632e-01
-4.81354922e-01 7.47594908e-02 7.16506124e-01 7.93720543e-01
2.52615064e-01 7.59690404e-01 7.57821321e-01 -1.18267620e+00
-1.68704167e-02 -1.13996494e+00 -4.23261851e-01 4.34529036e-01
4.98061001e-01 -9.58153069e-01 -2.60500520e-01 1.42406687e-01]
|
[10.548544883728027, -1.8562434911727905]
|
b3852995-6b96-4a84-9c5e-bbf4f0d5818c
|
nslf-ol-online-learning-of-neural-surface
|
2305.00282
| null |
https://arxiv.org/abs/2305.00282v1
|
https://arxiv.org/pdf/2305.00282v1.pdf
|
NSLF-OL: Online Learning of Neural Surface Light Fields alongside Real-time Incremental 3D Reconstruction
|
Immersive novel view generation is an important technology in the field of graphics and has recently also received attention for operator-based human-robot interaction. However, the involved training is time-consuming, and thus the current test scope is majorly on object capturing. This limits the usage of related models in the robotics community for 3D reconstruction since robots (1) usually only capture a very small range of view directions to surfaces that cause arbitrary predictions on unseen, novel direction, (2) requires real-time algorithms, and (3) work with growing scenes, e.g., in robotic exploration. The paper proposes a novel Neural Surface Light Fields model that copes with the small range of view directions while producing a good result in unseen directions. Exploiting recent encoding techniques, the training of our model is highly efficient. In addition, we design Multiple Asynchronous Neural Agents (MANA), a universal framework to learn each small region in parallel for large-scale growing scenes. Our model learns online the Neural Surface Light Fields (NSLF) aside from real-time 3D reconstruction with a sequential data stream as the shared input. In addition to online training, our model also provides real-time rendering after completing the data stream for visualization. We implement experiments using well-known RGBD indoor datasets, showing the high flexibility to embed our model into real-time 3D reconstruction and demonstrating high-fidelity view synthesis for these scenes. The code is available on github.
|
['Andreas Nuchter', 'Yijun Yuan']
|
2023-04-29
| null | null | null | null |
['3d-reconstruction']
|
['computer-vision']
|
[ 3.35021198e-01 3.11113417e-01 4.09683347e-01 -2.34573722e-01
-2.03960449e-01 -4.02058065e-01 3.99038881e-01 -2.73854703e-01
-3.33417267e-01 4.47895378e-01 -2.90379167e-01 -3.49910408e-01
1.05156638e-01 -1.22466087e+00 -1.11571622e+00 -5.52002788e-01
-2.24368498e-01 7.27804661e-01 3.76300365e-01 -4.00649399e-01
2.18525514e-01 9.09633458e-01 -2.22898316e+00 2.62966543e-01
5.85157216e-01 1.01839960e+00 8.94177496e-01 9.22417998e-01
-7.23540410e-02 5.96489131e-01 -2.06241772e-01 1.83525950e-01
6.62555695e-01 -1.39318854e-01 -5.57058513e-01 2.59554595e-01
4.28960949e-01 -5.73704362e-01 -2.56013393e-01 6.64703012e-01
5.48165619e-01 8.97058174e-02 1.89756483e-01 -1.18910468e+00
-2.80257136e-01 -8.53187814e-02 -3.75224531e-01 -3.84615123e-01
5.93296647e-01 2.83972949e-01 5.03840744e-01 -1.13551116e+00
1.16135156e+00 1.08309281e+00 3.09207052e-01 8.63088965e-01
-9.09413815e-01 -4.09157574e-01 2.55848110e-01 1.51582986e-01
-9.88081396e-01 -3.32569897e-01 8.97389412e-01 -1.79191187e-01
1.11169922e+00 4.94595587e-01 1.23071742e+00 9.67123687e-01
3.78314823e-01 7.53610075e-01 1.10551703e+00 -2.36072764e-01
6.59260452e-01 -2.19106339e-02 -4.64398742e-01 8.18565667e-01
-3.25201713e-02 2.82235503e-01 -7.46020257e-01 1.72329888e-01
1.41834760e+00 2.79534101e-01 -5.35679460e-01 -9.85720217e-01
-1.34150422e+00 4.43882167e-01 7.78072000e-01 -1.73378631e-01
-3.53188336e-01 3.85560215e-01 -6.31412491e-02 3.36754113e-01
4.99797612e-01 2.56531447e-01 -5.44124663e-01 -1.66805372e-01
-2.04677165e-01 3.90637457e-01 9.67931032e-01 1.31697154e+00
1.09443605e+00 -4.08040956e-02 4.87919778e-01 5.70747674e-01
1.83883876e-01 7.11621642e-01 5.29050902e-02 -1.39045787e+00
3.93035322e-01 5.95373750e-01 3.55530232e-01 -9.01785851e-01
-5.60615063e-01 -3.17612469e-01 -9.71915066e-01 1.01584768e+00
2.01652497e-01 1.00460254e-01 -8.17318738e-01 1.43023705e+00
7.17539012e-01 -1.69814024e-02 1.53650250e-02 1.03590810e+00
5.82386672e-01 7.48401582e-01 -7.15767264e-01 -1.07764952e-01
9.95031536e-01 -1.01630950e+00 -3.47450078e-01 -4.15953189e-01
4.50804442e-01 -4.43550020e-01 1.29298997e+00 7.29720712e-01
-1.21215010e+00 -2.75081784e-01 -9.74583566e-01 -3.68281037e-01
-3.41799408e-01 -3.14181149e-01 9.18385565e-01 2.39543691e-01
-1.37052155e+00 5.28494239e-01 -1.11426747e+00 -5.12138247e-01
2.85328925e-01 3.67276937e-01 -4.25785631e-01 -4.63029534e-01
-5.10316193e-01 6.16512835e-01 -2.03724112e-02 1.99280292e-01
-8.09564829e-01 -5.44547677e-01 -6.71499789e-01 -1.86697617e-01
4.31858033e-01 -1.14176834e+00 1.17862642e+00 -5.51355600e-01
-2.01961970e+00 8.57314408e-01 3.10952272e-02 -1.47289678e-01
7.97723114e-01 -2.13909060e-01 2.94743866e-01 6.99381065e-03
-1.20264508e-01 1.03509331e+00 7.00210512e-01 -1.64607799e+00
-6.00853503e-01 -5.79811215e-01 3.93525183e-01 6.52875721e-01
-7.09732547e-02 -5.47463715e-01 -4.14551824e-01 -1.60083234e-01
6.01569116e-01 -1.02017808e+00 -5.21650255e-01 7.43723512e-01
-7.64282346e-02 -1.44754304e-02 8.93373549e-01 -6.39713258e-02
3.75272095e-01 -2.05095458e+00 3.41503680e-01 -2.46932972e-02
1.55033022e-01 -3.68028611e-01 1.26630411e-01 4.73151326e-01
3.45957577e-01 -2.72245735e-01 -3.91125292e-01 -6.58584356e-01
-9.96665582e-02 6.88127100e-01 -4.02485877e-01 3.68435621e-01
-1.82899103e-01 5.88707447e-01 -1.04539704e+00 -2.10274071e-01
4.95027781e-01 5.33979356e-01 -1.04589200e+00 4.49989796e-01
-3.54239315e-01 7.55433381e-01 -3.76763672e-01 5.36344886e-01
9.82952833e-01 -3.53708327e-01 -3.16679403e-02 1.48264155e-01
-4.24933374e-01 6.66834116e-02 -1.34377348e+00 2.31201315e+00
-9.19304609e-01 2.97709703e-01 4.18973446e-01 -5.06992996e-01
8.59118402e-01 -1.30403014e-02 4.03360903e-01 -6.73469782e-01
-9.80028361e-02 3.41162562e-01 -5.09686053e-01 -4.83273447e-01
6.89555943e-01 5.17734922e-02 6.78995922e-02 5.15367687e-01
-3.47725034e-01 -8.02761734e-01 -1.47117451e-01 1.30340785e-01
1.29835844e+00 6.21063232e-01 8.20859745e-02 -2.57100873e-02
8.91734904e-04 2.70677134e-02 4.40317512e-01 6.06670022e-01
3.22839707e-01 8.98901582e-01 -2.46867910e-02 -7.30453491e-01
-1.17503822e+00 -1.12896192e+00 -1.44231571e-02 8.21079314e-01
7.11727917e-01 -1.84850916e-01 -3.55051786e-01 -3.39284390e-01
-8.62436146e-02 6.26263380e-01 -5.35299420e-01 2.52530456e-01
-8.42563093e-01 -3.28622878e-01 -2.42068142e-01 3.37056071e-01
4.09918845e-01 -1.45649028e+00 -1.68252432e+00 1.95320189e-01
2.10698717e-03 -9.35665905e-01 4.24170047e-02 5.36478162e-01
-1.18038785e+00 -9.46060359e-01 -6.70209169e-01 -7.54054368e-01
7.60586500e-01 7.62807190e-01 1.05467963e+00 5.70450984e-02
-3.15643460e-01 6.24857426e-01 -2.76277721e-01 -4.54391181e-01
-3.02980393e-01 -2.87164897e-01 -1.72541872e-01 -3.35934073e-01
-3.71953756e-01 -9.69623685e-01 -8.05724859e-01 3.16829860e-01
-9.41242337e-01 8.55376124e-01 3.79227668e-01 7.56872058e-01
8.47855806e-01 -3.20828199e-01 5.26408665e-02 -7.56944001e-01
1.70006365e-01 -3.44781846e-01 -8.32659483e-01 -1.08366214e-01
-3.48452061e-01 -1.51988536e-01 6.97807133e-01 -2.29230896e-01
-1.23560369e+00 2.84413874e-01 -1.83697015e-01 -5.13303936e-01
-3.29056621e-01 1.07358873e-01 2.60035209e-02 -1.33534312e-01
6.40788138e-01 1.07660927e-01 -8.15843940e-02 -3.61086011e-01
3.10291946e-01 3.66801262e-01 4.28182453e-01 -3.91695857e-01
5.56573033e-01 1.08081174e+00 1.70227766e-01 -7.73045242e-01
-3.25859398e-01 -2.28561014e-01 -5.61412871e-01 -4.76563483e-01
5.77911735e-01 -7.50870287e-01 -9.37189639e-01 5.95904946e-01
-1.44567335e+00 -8.78146350e-01 -7.20571876e-01 3.72970313e-01
-1.13928533e+00 8.16585347e-02 -4.79109645e-01 -9.94704843e-01
-1.33622900e-01 -1.12446916e+00 1.48105252e+00 9.66581330e-02
1.06986225e-01 -6.83653235e-01 -4.52015027e-02 -1.43820215e-02
3.43453884e-01 3.77603143e-01 7.11453676e-01 3.53073567e-01
-1.42368186e+00 9.33138207e-02 -1.43565744e-01 -2.38210395e-01
-7.86638707e-02 -3.92948151e-01 -1.14345145e+00 -2.44489685e-01
1.92192197e-01 -5.49058199e-01 5.67570567e-01 1.76028907e-01
1.26370656e+00 -4.14296612e-03 -4.14540976e-01 1.01230335e+00
1.48263025e+00 2.25541130e-01 6.18876934e-01 3.90198052e-01
7.13622630e-01 7.55144894e-01 6.62587225e-01 7.50058055e-01
5.27200580e-01 6.40011013e-01 1.18261254e+00 -1.13320112e-01
-3.37839918e-03 -2.40986243e-01 2.31242001e-01 9.03189898e-01
-3.62993479e-01 -4.85747218e-01 -6.76050305e-01 2.70161361e-01
-1.88266718e+00 -5.94668984e-01 -9.27952379e-02 2.29718089e+00
4.33164775e-01 -5.98660931e-02 -3.99473250e-01 5.06731011e-02
1.60244867e-01 6.39666170e-02 -8.53844881e-01 -3.77803624e-01
2.19653901e-02 1.68472916e-01 5.41979849e-01 3.91844600e-01
-4.81245369e-01 7.94963837e-01 5.03590107e+00 2.92925358e-01
-1.24304342e+00 6.33180663e-02 3.32267165e-01 -2.35299706e-01
-6.46287084e-01 2.96265073e-02 -3.57316613e-01 -3.40474918e-02
2.94704735e-01 1.87019214e-01 8.51160467e-01 9.69927847e-01
7.72720426e-02 -4.20307040e-01 -1.25584340e+00 1.27739894e+00
-4.66954000e-02 -1.33577335e+00 -1.04657784e-01 3.05374295e-01
6.73527122e-01 3.27031970e-01 -5.01946472e-02 -1.38298929e-01
4.64758724e-01 -6.39107585e-01 1.01023901e+00 5.43660998e-01
1.00465274e+00 -4.95308101e-01 2.79502749e-01 9.39068258e-01
-8.77308130e-01 -6.61691874e-02 -6.58194065e-01 -2.26315811e-01
5.10249734e-01 5.34060776e-01 -8.01974893e-01 4.41101015e-01
1.13298905e+00 7.19784200e-01 -8.59137103e-02 9.31732595e-01
-1.09801404e-01 -1.33025616e-01 -7.21609771e-01 -1.58838242e-01
-5.36242574e-02 -2.47969285e-01 6.30668819e-01 5.57393789e-01
7.26763189e-01 1.80895790e-01 8.59236717e-02 8.25587273e-01
5.31354696e-02 -8.46687630e-02 -9.93743896e-01 5.53334832e-01
2.05469951e-01 1.08123720e+00 -9.91633415e-01 -8.67289379e-02
-2.52072543e-01 1.33481896e+00 5.22234380e-01 2.92162508e-01
-4.42926615e-01 -2.08642274e-01 3.62957716e-01 2.40762919e-01
1.64543316e-01 -5.90133429e-01 -4.67149705e-01 -1.21547997e+00
2.05861345e-01 -3.16272616e-01 -1.68548450e-01 -1.15782726e+00
-9.54277813e-01 6.87908769e-01 -1.70259669e-01 -1.48925102e+00
-2.65890867e-01 -7.19332874e-01 -1.17713518e-01 5.94139993e-01
-1.61408365e+00 -1.05218685e+00 -9.16379809e-01 6.21328592e-01
8.42084110e-01 3.17716867e-01 1.05460620e+00 2.55419500e-02
1.22389480e-01 -1.03586830e-01 -3.76064740e-02 -6.58994079e-01
4.91689622e-01 -1.30386114e+00 6.48765564e-01 5.18686771e-01
2.51551531e-02 2.42376149e-01 5.87024271e-01 -5.57891309e-01
-1.98689353e+00 -7.86309719e-01 4.84281212e-01 -3.88845891e-01
8.83657560e-02 -8.62420201e-01 -7.39593208e-01 6.10906184e-01
2.16079969e-02 2.66407430e-01 1.08551487e-01 -2.30102822e-01
7.68766478e-02 -5.52527010e-02 -1.12892580e+00 6.04614079e-01
1.88868213e+00 -1.84408262e-01 -9.55055375e-03 3.76485080e-01
7.83098102e-01 -1.09510887e+00 -4.33016658e-01 3.16329539e-01
6.77779913e-01 -1.55823755e+00 8.98975909e-01 7.85548706e-03
5.35301983e-01 -3.27532262e-01 -2.54662722e-01 -1.34678078e+00
-1.21346645e-01 -5.82605243e-01 -1.68467984e-01 3.25400591e-01
1.15169957e-01 -8.72127473e-01 9.61214900e-01 5.40302217e-01
-5.40860355e-01 -8.93033624e-01 -9.34238493e-01 -4.31861132e-01
-4.02441055e-01 -6.79916620e-01 6.35419607e-01 6.80028319e-01
-2.96764404e-01 8.26426893e-02 -3.07731152e-01 2.77059346e-01
6.30940437e-01 5.94636738e-01 1.23225009e+00 -1.24716282e+00
-4.25241590e-01 -1.83831751e-02 -2.85513937e-01 -1.59135067e+00
-2.83244312e-01 -7.73187637e-01 3.81071419e-01 -1.76728582e+00
-2.14999199e-01 -9.07363832e-01 3.04360837e-01 2.49345422e-01
4.45590675e-01 1.77315980e-01 2.82473683e-01 2.27334067e-01
-4.43133295e-01 7.26281822e-01 1.76231301e+00 2.10389763e-01
-3.43156427e-01 8.22391585e-02 -1.17949665e-01 8.41262817e-01
4.44182336e-01 -2.53973454e-01 -6.24300241e-01 -8.76431584e-01
8.43833268e-01 4.33543026e-01 5.52827179e-01 -1.14245999e+00
4.04259890e-01 -3.19477320e-01 3.56576532e-01 -7.13953733e-01
7.93492675e-01 -1.17060483e+00 1.93324104e-01 4.21421587e-01
2.06486061e-02 1.36773333e-01 -1.44362360e-01 7.69614518e-01
2.10238963e-01 6.76226914e-02 4.76042479e-01 -5.96701443e-01
-7.69167006e-01 5.85896134e-01 -1.89176917e-01 -1.85355246e-01
9.64947939e-01 -6.36422694e-01 -7.20562413e-02 -4.49651808e-01
-6.77269995e-01 1.08177274e-01 1.08092952e+00 3.70370209e-01
1.01184142e+00 -1.01305103e+00 -3.81487697e-01 6.29162729e-01
2.07250595e-01 1.11260056e+00 4.52266991e-01 3.60317618e-01
-9.90150273e-01 -1.01418309e-01 -2.68167108e-01 -9.98408616e-01
-7.52186537e-01 4.47198153e-01 8.47385600e-02 9.80895478e-04
-1.12146711e+00 9.34300125e-01 6.78299904e-01 -7.73764014e-01
1.30251765e-01 -5.62349260e-01 1.66133031e-01 -5.54154813e-01
2.43964210e-01 3.96826804e-01 1.56407177e-01 -1.38386965e-01
-9.55316499e-02 7.21929550e-01 4.19148743e-01 -2.44560570e-01
1.73294854e+00 -2.39744261e-01 -4.74265106e-02 6.50925100e-01
1.00154114e+00 -3.13098170e-02 -1.67310166e+00 -4.00021039e-02
-5.20575047e-01 -7.31844664e-01 -1.12070650e-01 -4.70091432e-01
-8.63777161e-01 1.10710752e+00 6.76037967e-01 4.47970033e-02
1.16789627e+00 4.79092598e-02 8.08809698e-01 6.64635420e-01
1.38167644e+00 -9.39469457e-01 3.78018208e-02 5.47296047e-01
1.29696167e+00 -1.05965745e+00 -7.78989494e-02 -6.56997025e-01
-2.29765058e-01 1.26068366e+00 7.61522055e-01 -2.94856250e-01
6.10529900e-01 6.36187434e-01 1.31275738e-02 -3.35957497e-01
-9.36470151e-01 1.87111288e-01 -4.58080530e-01 7.53721714e-01
-1.26068547e-01 2.59716921e-02 2.84026027e-01 -1.10351846e-01
-4.32428926e-01 1.52673900e-01 7.25290835e-01 1.28685439e+00
-3.75102013e-01 -9.25649405e-01 -1.85898975e-01 2.09100172e-01
3.23491961e-01 2.29797676e-01 -2.11452935e-02 6.65337503e-01
1.43179193e-01 4.73670483e-01 3.68966609e-01 -2.01088190e-01
4.08667952e-01 -4.34203804e-01 8.05614889e-01 -6.96964324e-01
-3.11965883e-01 -4.27890196e-02 -1.20837256e-01 -1.03738642e+00
-4.56111491e-01 -5.67960799e-01 -1.43227243e+00 -1.48568809e-01
-1.66261986e-01 -3.68063539e-01 9.21607673e-01 5.43458283e-01
6.52500093e-01 4.47988659e-01 7.22733498e-01 -1.61820018e+00
-1.64214075e-01 -6.71832383e-01 -5.49537063e-01 2.96801925e-01
2.83716887e-01 -7.27691531e-01 -2.31911391e-01 1.15380213e-02]
|
[8.763050079345703, -2.8462026119232178]
|
ef59883c-10d8-4ee7-ab3b-c0d138023899
|
tracing-semantic-variation-in-slang
|
2210.08635
| null |
https://arxiv.org/abs/2210.08635v2
|
https://arxiv.org/pdf/2210.08635v2.pdf
|
Tracing Semantic Variation in Slang
|
The meaning of a slang term can vary in different communities. However, slang semantic variation is not well understood and under-explored in the natural language processing of slang. One existing view argues that slang semantic variation is driven by culture-dependent communicative needs. An alternative view focuses on slang's social functions suggesting that the desire to foster semantic distinction may have led to the historical emergence of community-specific slang senses. We explore these theories using computational models and test them against historical slang dictionary entries, with a focus on characterizing regularity in the geographical variation of slang usages attested in the US and the UK over the past two centuries. We show that our models are able to predict the regional identity of emerging slang word meanings from historical slang records. We offer empirical evidence that both communicative need and semantic distinction play a role in the variation of slang meaning yet their relative importance fluctuates over the course of history. Our work offers an opportunity for incorporating historical cultural elements into the natural language processing of slang.
|
['Yang Xu', 'Zhewei Sun']
|
2022-10-16
| null | null | null | null |
['culture']
|
['speech']
|
[-5.77000016e-03 -4.34240371e-01 -2.83098817e-01 -5.45722246e-01
-5.62464558e-02 -5.62851250e-01 9.61687744e-01 4.24439490e-01
-5.88050365e-01 2.09073246e-01 1.25860476e+00 -3.36755872e-01
-4.67465490e-01 -7.06276357e-01 -1.89749613e-01 -3.19291174e-01
2.81417221e-01 3.72144550e-01 1.49692353e-02 -8.13998699e-01
8.68920684e-01 1.64923921e-01 -1.18362331e+00 1.05197184e-01
8.12763929e-01 6.79923147e-02 5.72475135e-01 2.87211895e-01
-4.06348795e-01 4.52494234e-01 -4.48045492e-01 -1.17558874e-01
1.45994157e-01 -7.66497254e-01 -9.34675515e-01 1.51528046e-01
4.25606430e-01 3.65043074e-01 -4.81266789e-02 8.95791709e-01
1.77132204e-01 1.10081695e-02 4.61786449e-01 -7.50946164e-01
-8.72284949e-01 9.31569874e-01 -2.70750187e-02 4.76348884e-02
3.54271501e-01 -2.73707032e-01 1.48326969e+00 -7.17795610e-01
1.27639854e+00 1.37631285e+00 7.27799416e-01 2.69445837e-01
-1.24407876e+00 -4.00158674e-01 1.52069792e-01 3.03711355e-01
-1.36960351e+00 -5.24187028e-01 9.99615490e-01 -9.97195959e-01
4.97110307e-01 -9.24952701e-02 1.20679450e+00 7.05676019e-01
3.90060216e-01 3.61252666e-01 1.14976311e+00 -7.20261991e-01
-8.25520456e-02 -4.21350896e-02 -1.48862312e-02 5.22385895e-01
1.29210711e-01 -1.84581861e-01 -1.24762428e+00 -3.77820134e-01
4.51939374e-01 -3.86358760e-02 -1.68534219e-01 -1.97867453e-01
-1.27667212e+00 1.21828556e+00 3.58106494e-01 8.59178126e-01
-1.15069292e-01 2.02467546e-01 5.93829453e-01 5.13617039e-01
6.91429555e-01 6.35228872e-01 -5.73574185e-01 -4.48990226e-01
-8.04904699e-01 4.02574658e-01 5.83453417e-01 3.29958588e-01
7.67619431e-01 -4.69206423e-01 7.75132954e-01 1.26146865e+00
4.38990533e-01 3.20034683e-01 5.84433496e-01 -5.44728458e-01
-5.99425193e-03 6.03580058e-01 -1.41203657e-01 -1.20247996e+00
-2.58650005e-01 1.36663705e-01 1.38675958e-01 -2.71259159e-01
5.82897842e-01 6.77945465e-02 -3.61926109e-01 1.95983732e+00
6.67440668e-02 -3.60615194e-01 -5.18897511e-02 5.91111958e-01
3.11217725e-01 2.56519079e-01 2.03438550e-01 -3.59740620e-03
1.25518298e+00 -2.27768973e-01 -2.22701520e-01 -4.83913779e-01
8.32534671e-01 -9.06340778e-01 1.42678320e+00 -3.11859958e-02
-5.17365813e-01 -3.91860008e-02 -8.82978916e-01 -1.19996838e-01
-2.90741354e-01 -3.33844155e-01 1.22216880e+00 8.95417988e-01
-7.28048623e-01 5.62771380e-01 -6.36652410e-01 -1.10591102e+00
-1.95913110e-02 -2.75315732e-01 -3.89996804e-02 1.23572059e-01
-1.21302116e+00 9.88345385e-01 2.92253971e-01 7.12949131e-03
-1.93217061e-02 -4.24221873e-01 -8.91579270e-01 -6.39579833e-01
1.84666723e-01 -2.89905071e-01 8.51898432e-01 -1.38859415e+00
-1.20354068e+00 1.39729786e+00 -1.37291923e-01 -1.65303022e-01
2.90373474e-01 -1.77940115e-01 -6.79382443e-01 -2.08490372e-01
4.83783513e-01 2.25511223e-01 4.63624209e-01 -9.18897212e-01
-5.44336438e-01 -4.34120059e-01 -3.19020785e-02 3.48506898e-01
4.25493419e-02 4.68398839e-01 3.13574463e-01 -6.24528587e-01
7.48246670e-01 -1.05176592e+00 2.09841281e-01 -2.16911644e-01
3.20716321e-01 -2.23875135e-01 2.77101189e-01 -7.28747666e-01
1.28010702e+00 -2.30417824e+00 5.77114448e-02 2.54082918e-01
-8.79357532e-02 -6.21024370e-01 1.38551176e-01 1.40516400e+00
4.68155712e-01 2.40096509e-01 -1.41938657e-01 2.52340347e-01
2.32760087e-01 5.92655599e-01 -2.64275610e-01 9.38238978e-01
-3.92441839e-01 8.61141443e-01 -1.10944116e+00 -2.76203990e-01
1.03074491e-01 3.23857337e-01 -7.58257031e-01 -6.10802710e-01
-1.53354844e-02 1.51681885e-01 -1.48704380e-01 5.75097144e-01
1.45698085e-01 1.07201375e-01 6.63123608e-01 4.50503886e-01
-6.94667399e-01 8.72345030e-01 -7.14318156e-01 2.12362075e+00
-5.84997714e-01 1.12548792e+00 1.43249720e-01 -3.51890355e-01
7.80002654e-01 -4.20686603e-02 9.95321274e-02 -4.05109763e-01
1.61898419e-01 6.56519592e-01 6.76513612e-01 -4.31799918e-01
1.02366245e+00 -7.47424841e-01 -5.75953007e-01 5.20425320e-01
-2.49385506e-01 -5.84866047e-01 -9.27681401e-02 2.97965050e-01
5.92683852e-01 1.41507491e-01 6.44895256e-01 -1.09074259e+00
3.11789244e-01 1.99185893e-01 8.11706245e-01 3.67685348e-01
-1.86508700e-01 1.03191860e-01 2.70267904e-01 -4.88971651e-01
-1.09524047e+00 -1.07988894e+00 -6.10573292e-01 1.35028553e+00
4.46369827e-01 -6.83584809e-01 -2.71611601e-01 2.04012040e-02
5.85044809e-02 9.64656889e-01 -6.70922637e-01 -4.02006283e-02
-4.60326284e-01 -6.13408983e-01 6.08296216e-01 -8.96740854e-02
2.47351810e-01 -1.09774804e+00 -1.01522994e+00 3.45694005e-01
-1.07622251e-01 -7.34816730e-01 -2.74191469e-01 -4.43558574e-01
-3.60772103e-01 -1.01899672e+00 -1.85274482e-01 -6.99987710e-01
3.22468013e-01 9.76679176e-02 7.88713932e-01 4.23592776e-01
-8.43796283e-02 6.11384571e-01 -4.64550763e-01 -5.99081814e-01
-8.58741701e-01 3.71321887e-01 1.52310789e-01 -3.10854942e-01
5.13364851e-01 -4.78929192e-01 -1.70007035e-01 1.01334326e-01
-8.02083433e-01 1.20194294e-02 -1.07962944e-01 6.25595510e-01
-3.25591415e-01 -4.39563185e-01 6.97451949e-01 -1.12372494e+00
8.12185466e-01 -6.51182950e-01 -2.20224231e-01 2.97728665e-02
-6.52897120e-01 -3.57766673e-02 4.79677133e-02 -1.57108128e-01
-1.03786945e+00 -4.12035882e-01 -2.49683633e-01 8.09979498e-01
-6.19681291e-02 7.81590760e-01 2.64811546e-01 -6.06322773e-02
6.16999269e-01 1.00662604e-01 4.55835834e-03 -2.43009984e-01
5.37114143e-01 6.01086080e-01 3.40561569e-01 -1.07398045e+00
7.03592837e-01 5.84509075e-01 -4.12329614e-01 -1.64570010e+00
-5.74206591e-01 -5.80191553e-01 -8.32425654e-01 -3.55677396e-01
7.61173606e-01 -8.84083271e-01 -2.80056208e-01 3.99141371e-01
-8.95101428e-01 -5.46495497e-01 -1.39560714e-01 4.19536203e-01
-4.46760416e-01 4.21437889e-01 -7.40459323e-01 -7.10500777e-01
7.65714496e-02 -6.54340982e-01 7.72148728e-01 -3.16689402e-01
-8.90872717e-01 -1.36821389e+00 4.59616035e-01 2.25566357e-01
2.61027098e-01 3.51786494e-01 1.24489653e+00 -4.75472718e-01
-2.89544743e-02 2.69594699e-01 1.90998256e-01 -1.84084460e-01
5.69061220e-01 -8.05677548e-02 -4.18030828e-01 5.09733856e-02
3.78111415e-02 -1.95615128e-01 5.40229201e-01 5.65069765e-02
-1.65455565e-01 -1.56043306e-01 -2.13609058e-02 4.18024063e-01
1.76836550e+00 1.11972019e-01 2.81602710e-01 5.79079449e-01
4.13816273e-01 8.48352134e-01 2.25651845e-01 1.49556801e-01
4.40329939e-01 4.56514537e-01 -3.34151655e-01 4.30372864e-01
1.33573264e-01 -3.43497306e-01 5.05656481e-01 1.07930005e+00
3.89078036e-02 2.45664150e-01 -1.53487790e+00 1.05709291e+00
-1.82249880e+00 -9.84502852e-01 -3.33894104e-01 1.97265065e+00
6.12717688e-01 -2.35512182e-02 8.15131813e-02 -1.81857914e-01
4.12157744e-01 7.07946002e-01 -4.03723167e-03 -6.52277052e-01
-1.94317549e-01 1.69122536e-02 5.11530697e-01 1.05069900e+00
-3.01879972e-01 1.58071041e+00 6.82489491e+00 3.02429199e-01
-1.02219832e+00 3.95916812e-02 -2.48315021e-01 7.75598586e-02
-8.43722641e-01 2.53592670e-01 -2.95247465e-01 3.92434150e-01
5.64002156e-01 -3.57426018e-01 4.61061239e-01 3.98897052e-01
3.24924171e-01 -3.37286055e-01 -8.00327063e-01 6.15349948e-01
4.06258017e-01 -1.19653463e+00 -3.17179471e-01 1.52565673e-01
6.49552107e-01 3.91693056e-01 -4.50063914e-01 -2.01594964e-01
5.59226632e-01 -6.81364000e-01 1.27664304e+00 1.19468905e-01
5.79237759e-01 -5.12506545e-01 2.22651392e-01 1.74309134e-01
-1.14500666e+00 1.02313891e-01 -1.67210460e-01 -8.23032320e-01
6.60968423e-01 1.38322145e-01 -9.02040303e-01 8.02970827e-02
1.18485585e-01 7.81363130e-01 -6.27968550e-01 1.86915666e-01
-6.57283425e-01 7.16540277e-01 -1.42904624e-01 -1.46064967e-01
5.78313887e-01 -5.31218708e-01 7.50965357e-01 1.21578860e+00
1.67378653e-02 -1.98070094e-01 2.09463298e-01 7.38468409e-01
5.77570736e-01 6.35134995e-01 -7.06438303e-01 -2.94931799e-01
5.72163880e-01 5.99673152e-01 -1.08918715e+00 8.31607059e-02
-4.12454754e-01 1.32219994e+00 2.19736889e-01 2.26295933e-01
-4.17844951e-01 -3.58309550e-03 1.20132601e+00 5.37089705e-01
-2.53006548e-01 -8.90971184e-01 -2.82716423e-01 -1.05525780e+00
-4.11366910e-01 -3.81217062e-01 1.88449666e-01 -4.55356807e-01
-1.35119283e+00 2.22118031e-02 -2.67830975e-02 -1.40610799e-01
-1.72208950e-01 -4.06872600e-01 -1.59087241e-01 7.37330019e-01
-8.84526253e-01 -1.38043928e+00 3.52802910e-02 3.93454611e-01
7.10399687e-01 -2.78460309e-02 6.58445060e-01 -1.33422479e-01
-7.41576403e-02 -1.01789370e-01 3.58111635e-02 9.18843001e-02
7.81360984e-01 -1.02972281e+00 6.87330186e-01 8.37101996e-01
5.41523457e-01 9.52654541e-01 7.51813352e-01 -9.29602981e-01
-7.71707118e-01 -3.05921972e-01 1.48167884e+00 -6.37452602e-01
1.36475849e+00 -5.13025820e-01 -6.36233449e-01 9.03899789e-01
3.45356047e-01 -9.07372534e-01 1.07115805e+00 7.44196773e-01
-4.56047565e-01 4.26898897e-01 -9.27599251e-01 9.80025232e-01
1.61608863e+00 -1.16930306e+00 -1.08171332e+00 1.14623360e-01
5.32449245e-01 2.05076471e-01 -5.17085969e-01 -1.97961539e-01
9.11964834e-01 -8.88829529e-01 4.45812106e-01 -5.16030610e-01
3.51725161e-01 -1.81449711e-01 -5.04739344e-01 -1.48533154e+00
-5.47042787e-01 -5.38019478e-01 1.27258587e+00 1.12695491e+00
3.11991990e-01 -9.98344481e-01 4.84134465e-01 5.83377659e-01
-1.90434337e-01 -5.99052757e-02 -1.01175010e+00 -4.93532002e-01
3.53499830e-01 -6.48993313e-01 5.54251313e-01 1.56477499e+00
4.23105359e-01 5.79765141e-01 -2.42084339e-02 -1.81983829e-01
2.24464208e-01 2.07622210e-03 5.86462855e-01 -1.58027208e+00
8.08482990e-02 -5.75518847e-01 -7.30721951e-01 -5.48687398e-01
6.23399079e-01 -1.18735254e+00 7.20577091e-02 -1.43235517e+00
-1.88235462e-01 -5.82053900e-01 3.14598083e-01 7.09173158e-02
4.44681644e-01 7.03800842e-02 5.07153213e-01 3.56087387e-01
-1.12040162e-01 3.05743814e-01 6.46884143e-01 3.14038038e-01
-6.34732306e-01 -6.20211184e-01 -7.52187908e-01 1.06706142e+00
9.89676178e-01 -4.20294851e-01 -4.04112697e-01 -4.70848143e-01
1.13964725e+00 -4.96934682e-01 1.92311645e-01 -6.93975866e-01
3.98847535e-02 -8.22692752e-01 -4.08775032e-01 -3.89578199e-05
-9.01502557e-03 -7.29904592e-01 2.55928069e-01 6.58223033e-01
-2.15145648e-01 2.19006807e-01 -1.32108614e-01 3.27951819e-01
1.61982030e-02 -1.82758331e-01 5.55552781e-01 -3.65770727e-01
-1.22431612e+00 -3.17862391e-01 -9.00102139e-01 2.36304954e-01
6.48226619e-01 -4.92628217e-01 5.14141582e-02 -2.50803977e-01
-4.37941700e-01 -1.33856788e-01 1.37515974e+00 7.77241409e-01
2.16421068e-01 -1.28583372e+00 -8.40748072e-01 2.24711522e-01
3.83441955e-01 -7.86786079e-01 1.36793569e-01 4.87844586e-01
-9.42224622e-01 -2.10127700e-02 -1.92825288e-01 -2.93921024e-01
-1.10447574e+00 1.13494188e-01 2.99883991e-01 4.07061368e-01
-7.20045686e-01 7.37396836e-01 9.10417885e-02 -5.63375413e-01
-6.70395851e-01 -1.25925913e-01 9.86815393e-02 3.34951639e-01
1.16237320e-01 3.51172149e-01 -5.91595531e-01 -1.30634058e+00
-5.63045979e-01 7.09228933e-01 3.22803497e-01 -7.22163200e-01
1.20806074e+00 -7.05608606e-01 -4.37627614e-01 1.34056032e+00
8.82709146e-01 7.68677175e-01 -5.17052174e-01 -8.03859010e-02
3.54179025e-01 -7.18739629e-01 -5.15434325e-01 -6.15802526e-01
-3.00145626e-01 1.90577149e-01 6.01601861e-02 3.97487953e-02
4.13378865e-01 3.02282333e-01 6.94662571e-01 4.09012474e-02
6.18520319e-01 -1.44523358e+00 -4.17990357e-01 7.20006526e-01
9.56070781e-01 -7.43782043e-01 3.67074572e-02 -4.00903225e-01
-7.38464952e-01 8.35681200e-01 1.22341752e-01 -3.51989329e-01
7.09946930e-01 -1.44233659e-01 3.51275027e-01 -5.77319443e-01
-4.06083941e-01 -1.45156533e-01 8.67593512e-02 3.75343025e-01
8.96301687e-01 5.76562583e-01 -1.40284121e+00 -4.04699743e-02
-1.04911637e+00 -4.29253876e-01 7.61519253e-01 6.08164072e-01
-5.15399516e-01 -1.35254264e+00 -2.35207930e-01 1.17953196e-01
-3.83695781e-01 -5.16231179e-01 -9.46990013e-01 1.01826477e+00
5.39925098e-01 7.79901981e-01 1.68789312e-01 -2.95570552e-01
-1.64637819e-01 6.11334741e-01 6.35209084e-01 -9.87389624e-01
-8.96132827e-01 -1.19356588e-02 5.30683339e-01 -1.59591302e-01
-5.41673541e-01 -1.09190142e+00 -1.29625082e+00 -5.28186321e-01
-4.98045944e-02 3.35978031e-01 6.66373610e-01 1.05250120e+00
2.64787674e-02 -4.13774177e-02 -2.23018825e-01 -4.65296447e-01
1.01127021e-01 -7.26404428e-01 -1.20722628e+00 5.89547455e-01
-7.68766105e-02 -5.36293447e-01 -3.04081112e-01 -1.40374610e-02]
|
[10.348642349243164, 9.698309898376465]
|
4830e219-51a7-48cf-887c-0cfa33aee62e
|
apr-online-distant-point-cloud-registration
|
2305.02893
| null |
https://arxiv.org/abs/2305.02893v2
|
https://arxiv.org/pdf/2305.02893v2.pdf
|
APR: Online Distant Point Cloud Registration Through Aggregated Point Cloud Reconstruction
|
For many driving safety applications, it is of great importance to accurately register LiDAR point clouds generated on distant moving vehicles. However, such point clouds have extremely different point density and sensor perspective on the same object, making registration on such point clouds very hard. In this paper, we propose a novel feature extraction framework, called APR, for online distant point cloud registration. Specifically, APR leverages an autoencoder design, where the autoencoder reconstructs a denser aggregated point cloud with several frames instead of the original single input point cloud. Our design forces the encoder to extract features with rich local geometry information based on one single input point cloud. Such features are then used for online distant point cloud registration. We conduct extensive experiments against state-of-the-art (SOTA) feature extractors on KITTI and nuScenes datasets. Results show that APR outperforms all other extractors by a large margin, increasing average registration recall of SOTA extractors by 7.1% on LoKITTI and 4.6% on LoNuScenes. Code is available at https://github.com/liuQuan98/APR.
|
['Minyi Guo', 'Shan Chang', 'Hongzi Zhu', 'Yunsong Zhou', 'Quan Liu']
|
2023-05-04
| null | null | null | null |
['point-cloud-reconstruction', 'point-cloud-registration']
|
['computer-vision', 'computer-vision']
|
[-4.38111931e-01 -2.70181507e-01 -1.53578982e-01 -6.02115989e-01
-9.88992929e-01 -5.25535583e-01 6.72859490e-01 7.59146968e-03
-4.24753040e-01 1.67759657e-01 -5.83671965e-02 -1.19960427e-01
-7.30681941e-02 -1.05757952e+00 -1.14342713e+00 -5.14880061e-01
3.72018069e-02 6.66378260e-01 1.89132422e-01 -3.53340745e-01
3.00186463e-02 9.68476713e-01 -1.76885843e+00 -1.26110971e-01
6.18699372e-01 1.07555807e+00 2.30478600e-01 5.06544471e-01
3.06431041e-03 3.27027887e-01 -4.24024284e-01 -5.86507320e-01
7.49773860e-01 4.80077893e-01 -2.00265288e-01 -3.71999025e-01
9.95976567e-01 -7.02601135e-01 -8.12091529e-01 9.11175668e-01
3.94966513e-01 2.02410862e-01 4.81488615e-01 -1.45037210e+00
-6.95097625e-01 2.91392297e-01 -5.60434103e-01 3.36928889e-02
9.30962898e-03 2.18676507e-01 9.27529454e-01 -1.13051975e+00
4.57407981e-01 8.16817343e-01 9.11424279e-01 2.99981087e-01
-6.67663634e-01 -1.08868992e+00 -2.01999113e-01 3.53690833e-01
-1.78101718e+00 -4.75785583e-01 8.20676684e-01 -4.37296033e-01
1.07357311e+00 8.84992108e-02 8.05926442e-01 7.91404784e-01
2.49889866e-01 5.29629290e-01 6.48843348e-01 1.81218207e-01
9.82864946e-02 -2.11519808e-01 6.25750422e-02 5.87673664e-01
4.90940303e-01 5.89568377e-01 -6.36817694e-01 -5.15422001e-02
6.55548096e-01 5.19170523e-01 -2.81620119e-02 -2.41897628e-01
-1.17880690e+00 9.72576559e-01 9.93325889e-01 -2.73390803e-02
-4.18232292e-01 4.13294077e-01 1.39763029e-02 2.44285688e-01
4.64111984e-01 -8.94649606e-03 -3.18422765e-01 -1.98828638e-01
-7.82454550e-01 3.38316053e-01 4.77109760e-01 1.38864768e+00
1.20767868e+00 -8.59100930e-03 4.62822586e-01 3.83293539e-01
3.92801225e-01 1.15256727e+00 2.14775592e-01 -1.08190262e+00
7.22263396e-01 4.38889593e-01 -5.39036803e-02 -1.04726315e+00
-1.20236501e-01 -1.66060522e-01 -7.62475193e-01 5.87050974e-01
-4.87374030e-02 -7.63661936e-02 -8.40396702e-01 1.26109493e+00
4.86557066e-01 5.62990248e-01 8.12780783e-02 1.06060076e+00
1.04444826e+00 8.28198791e-01 -2.51523018e-01 5.45088649e-01
1.19368410e+00 -5.08349121e-01 -4.04750407e-01 -3.61799568e-01
3.19188654e-01 -6.23351514e-01 4.26218271e-01 -3.18649746e-02
-8.83587837e-01 -7.71876514e-01 -1.30208826e+00 -4.63773102e-01
-3.09581101e-01 3.09997555e-02 6.00234449e-01 3.97620648e-02
-9.41039383e-01 6.94175661e-01 -1.15857267e+00 9.16763954e-03
5.87890387e-01 6.18455887e-01 -6.35260820e-01 -1.81563258e-01
-8.66066098e-01 9.22865927e-01 1.57914266e-01 2.08407640e-01
-3.97480309e-01 -9.69725013e-01 -1.13814187e+00 -7.28403926e-02
-5.11103235e-02 -8.76247168e-01 1.14011252e+00 -2.30135858e-01
-1.15286934e+00 6.17239296e-01 -3.15848351e-01 -6.51538014e-01
4.08253670e-01 -5.61012328e-01 -4.95584935e-01 1.89843047e-02
2.80810177e-01 1.01879144e+00 8.71928930e-01 -1.33019435e+00
-8.57004106e-01 -5.97754478e-01 -2.33254075e-01 1.10008553e-01
1.82943851e-01 -3.75537246e-01 -4.16104972e-01 -1.47720814e-01
2.71070987e-01 -1.16407084e+00 -6.39431775e-02 1.74279407e-01
-2.10873365e-01 -4.60781485e-01 1.13885880e+00 -4.26120609e-01
4.07774895e-01 -2.37671900e+00 -2.60230273e-01 1.91023409e-01
5.54878175e-01 -6.24951022e-03 -1.29536957e-01 2.37756461e-01
9.41075161e-02 -2.87119240e-01 -1.30584359e-01 -4.61154848e-01
1.27864823e-01 3.11433971e-01 -4.72177535e-01 7.08510041e-01
3.54472488e-01 1.18023729e+00 -8.01810920e-01 -1.70750812e-01
6.18504226e-01 9.46855783e-01 -3.46635729e-01 4.75775218e-03
2.19647393e-01 1.20496340e-01 -5.45842052e-01 6.79256976e-01
1.05439496e+00 -6.89251497e-02 -5.90803206e-01 -3.58027339e-01
-2.37952515e-01 1.25922367e-01 -8.63862157e-01 1.73561394e+00
-4.28699374e-01 9.71601963e-01 -3.46071869e-01 -3.80182117e-01
1.24132562e+00 -3.76355345e-03 7.71732211e-01 -7.12738454e-01
2.72845984e-01 2.37917617e-01 -2.16692075e-01 -1.43957213e-01
8.15064013e-01 3.12315784e-02 -1.53972685e-01 -8.20981339e-03
-2.30049249e-02 -4.64583188e-01 -3.85255784e-01 2.10455549e-03
1.13654220e+00 -2.84734629e-02 8.90480429e-02 1.87047586e-01
1.87215015e-01 1.64074674e-01 5.60411215e-01 4.11570102e-01
-2.33242944e-01 9.30582106e-01 -4.43765432e-01 -7.18505740e-01
-1.07942367e+00 -1.32398486e+00 -3.15159589e-01 3.53211015e-01
4.94570404e-01 -3.89396459e-01 -3.14409435e-01 -4.31468517e-01
6.14113033e-01 4.88126367e-01 -3.32608223e-01 -2.15455279e-01
-8.08166921e-01 -3.89963873e-02 5.13111711e-01 8.51968646e-01
4.75080639e-01 -7.47636080e-01 -7.90587962e-01 -1.68430936e-02
-1.05592594e-01 -1.48425055e+00 -3.58889520e-01 -2.39199251e-02
-9.63055909e-01 -8.62786949e-01 -3.13245535e-01 -6.81107640e-01
4.45196748e-01 7.60703504e-01 1.12082791e+00 6.74441317e-03
1.37768447e-01 1.92301586e-01 -2.58762598e-01 -8.52092922e-01
1.75685380e-02 9.84898731e-02 2.35022485e-01 -7.06437379e-02
1.08376276e+00 -6.79089427e-01 -5.13778627e-01 1.65989488e-01
-5.82666755e-01 -1.92280203e-01 7.10177779e-01 4.52162355e-01
9.13201213e-01 -1.86141968e-01 4.60761636e-02 -2.35045195e-01
6.65738434e-02 -5.73693991e-01 -9.61114228e-01 -5.31977952e-01
-2.12326527e-01 -9.39827934e-02 2.91998059e-01 -5.71357943e-02
-4.92191166e-01 3.84521365e-01 -4.44746137e-01 -1.21602225e+00
-3.56425345e-01 2.86111295e-01 -3.53568792e-02 -2.77946442e-01
6.63842797e-01 1.26251772e-01 5.27471378e-02 -3.87068123e-01
2.87938833e-01 7.58778930e-01 8.80894840e-01 -1.75474361e-01
1.52842939e+00 8.25318277e-01 -9.42470431e-02 -7.66865671e-01
-5.46479225e-01 -7.36789048e-01 -8.99709821e-01 -4.48506847e-02
8.96208048e-01 -1.36626744e+00 -9.06037986e-01 2.82912821e-01
-1.27198291e+00 -2.26438686e-01 -4.81072187e-01 7.19411850e-01
-5.23356557e-01 1.71117447e-02 -3.09326947e-01 -3.03942353e-01
-4.78204459e-01 -1.04734004e+00 1.68843675e+00 2.08135426e-01
3.23236501e-03 -6.17867768e-01 2.39288792e-01 4.41866368e-01
1.54190391e-01 4.26264048e-01 1.04796186e-01 -3.24149966e-01
-1.15873623e+00 -5.93210995e-01 -1.58323407e-01 1.59355581e-01
1.37535244e-01 -1.68647722e-03 -1.14992261e+00 -2.42485791e-01
6.78751022e-02 1.03940420e-01 8.85450840e-01 2.71454841e-01
1.00504184e+00 -7.15424642e-02 -4.62644458e-01 9.77334321e-01
1.51296318e+00 -2.58986298e-02 5.70756316e-01 3.02549660e-01
1.05359054e+00 2.60810554e-01 7.36427128e-01 2.29951933e-01
7.42114723e-01 7.32701838e-01 6.73372328e-01 1.69172958e-01
-1.43024653e-01 -3.29385877e-01 3.17480177e-01 1.14092386e+00
-2.94084519e-01 2.21783310e-01 -1.16485035e+00 7.60925770e-01
-1.69171238e+00 -9.63771582e-01 -1.90092817e-01 2.04816866e+00
3.05945277e-01 -1.19304610e-03 -1.51399136e-01 3.06375604e-02
5.51349461e-01 1.83054700e-01 -6.09556258e-01 8.60692412e-02
6.64154291e-02 3.90412033e-01 8.83891523e-01 4.94758010e-01
-9.94963825e-01 9.51958179e-01 4.79887295e+00 4.36969608e-01
-1.35115564e+00 2.54633516e-01 5.39276488e-02 -2.19307184e-01
-1.59061849e-01 6.41673384e-03 -9.74799514e-01 5.32167912e-01
1.11106241e+00 -2.70251274e-01 2.08312958e-01 1.08655310e+00
-1.74988195e-01 2.94224143e-01 -1.03476405e+00 1.32331443e+00
-3.48714320e-03 -1.52820933e+00 -1.49145141e-01 4.19341445e-01
5.71311474e-01 1.15864480e+00 8.99442881e-02 2.64411628e-01
4.01497483e-01 -9.16874290e-01 8.21021795e-01 4.62097108e-01
8.25887561e-01 -1.02080643e+00 8.78921688e-01 4.52082276e-01
-1.43156838e+00 7.35320076e-02 -6.81854546e-01 1.14398506e-02
2.51821280e-01 7.66041756e-01 -7.85784006e-01 6.87290132e-01
1.01749933e+00 1.04330170e+00 -5.28247535e-01 9.74482894e-01
-3.65034752e-02 2.08148986e-01 -7.79818416e-01 2.33958095e-01
6.16623014e-02 -2.39159018e-01 5.80157280e-01 8.10712337e-01
6.38444960e-01 2.43115261e-01 6.58768192e-02 8.16189349e-01
-2.83431798e-01 -4.18107241e-01 -1.26163077e+00 5.28826952e-01
9.01677430e-01 1.36778927e+00 -9.59922075e-02 -2.51511991e-01
-5.93164623e-01 6.03551805e-01 3.01201433e-01 3.38464715e-02
-9.80162263e-01 -3.79053563e-01 1.24592018e+00 1.14156611e-01
5.01325667e-01 -6.79472566e-01 -3.36431921e-01 -1.12646437e+00
3.51847976e-01 -4.69846398e-01 -5.29617332e-02 -8.54771018e-01
-1.37973893e+00 7.35226452e-01 -1.49460428e-03 -1.86630797e+00
-2.00728521e-01 -4.01277483e-01 -5.34432650e-01 7.40376115e-01
-1.77867997e+00 -1.35519779e+00 -9.29971993e-01 7.13479042e-01
3.96263510e-01 -1.26206383e-01 4.76593554e-01 4.62648869e-01
-1.18597522e-01 4.10840899e-01 1.35336248e-02 3.97409499e-01
3.50161880e-01 -1.01382816e+00 1.15268135e+00 6.99853480e-01
5.77788889e-01 5.91730595e-01 2.56423324e-01 -7.21931756e-01
-1.94336343e+00 -1.59076869e+00 7.92561650e-01 -8.95119309e-01
4.62723464e-01 -3.52846652e-01 -9.77458477e-01 1.04641557e+00
1.40280277e-02 4.94411826e-01 4.28854436e-01 -5.01619652e-02
-3.03654850e-01 -4.46575940e-01 -1.06806695e+00 4.21599090e-01
1.23633206e+00 -6.84652627e-01 -5.99463165e-01 2.66792893e-01
8.28005314e-01 -9.77729738e-01 -1.20029891e+00 5.53871810e-01
4.43052053e-01 -7.94848084e-01 1.25230110e+00 -7.38675743e-02
3.65746528e-01 -6.02532923e-01 -3.99351150e-01 -1.22822893e+00
-5.00705719e-01 -2.88518876e-01 -5.03277004e-01 7.67813444e-01
8.10493454e-02 -8.49997640e-01 9.21374619e-01 3.52044076e-01
-4.03835297e-01 -5.88867009e-01 -1.23314416e+00 -9.78458107e-01
1.10101581e-01 -7.07867622e-01 1.18974435e+00 9.08695817e-01
-5.45001090e-01 2.97466427e-01 4.21276391e-02 6.73909783e-01
7.67200828e-01 2.61339158e-01 1.27841926e+00 -1.44671762e+00
1.70805454e-02 -1.67142794e-01 -1.08866048e+00 -1.17434669e+00
3.40217441e-01 -1.04301536e+00 1.63402900e-01 -1.24240947e+00
-3.77926290e-01 -6.65802419e-01 -2.59491708e-02 4.97245431e-01
5.25912754e-02 4.58994746e-01 4.19708520e-01 4.43495274e-01
-1.84393242e-01 8.30315709e-01 8.74915600e-01 -2.61459053e-01
-8.01826641e-02 2.90014092e-02 -5.68235219e-01 6.50080204e-01
9.69053209e-01 -5.76956093e-01 -1.49585679e-01 -1.14062548e+00
3.83029170e-02 -1.54128328e-01 7.84994006e-01 -1.51478994e+00
6.46215498e-01 -1.56698022e-02 5.76881289e-01 -1.18486536e+00
7.76210666e-01 -1.13925064e+00 4.10024881e-01 1.31459460e-01
3.90594423e-01 4.74364519e-01 3.61637920e-01 5.60576499e-01
-3.34621221e-01 2.30129644e-01 4.03241932e-01 2.47597009e-01
-8.56881022e-01 8.93121839e-01 2.70958424e-01 -4.08105731e-01
1.02416611e+00 -3.80582064e-01 -2.65135199e-01 -3.22064519e-01
-1.60153851e-01 1.77225351e-01 6.15125060e-01 7.57375598e-01
9.47340846e-01 -1.77817082e+00 -8.15927625e-01 4.45372403e-01
3.60166311e-01 8.59644949e-01 9.49210078e-02 6.14432395e-01
-6.31748974e-01 3.72893512e-01 -2.56472260e-01 -1.12940598e+00
-1.16263294e+00 3.01235229e-01 2.14945212e-01 4.33845758e-01
-1.19510674e+00 6.64113641e-01 -7.59225190e-02 -6.57288015e-01
-3.21550608e-01 -5.26772499e-01 1.15315527e-01 -2.54394919e-01
3.99756253e-01 2.72190839e-01 4.80008960e-01 -1.12770438e+00
-3.90431613e-01 1.09965110e+00 -2.40880065e-02 4.77478094e-02
1.49873757e+00 1.55665308e-01 3.63304734e-01 2.41965503e-01
1.51136172e+00 1.75538495e-01 -1.42262411e+00 -3.15070063e-01
-2.38046825e-01 -8.28996956e-01 3.30818236e-01 -9.39034224e-02
-1.48032284e+00 7.34911680e-01 6.77938104e-01 -5.20631708e-02
7.74878383e-01 1.03682332e-01 1.31350207e+00 6.87978566e-01
7.29716182e-01 -5.53824782e-01 -4.74008620e-01 6.62914395e-01
8.71854603e-01 -1.48211813e+00 1.05197646e-01 -3.42817873e-01
-4.95017588e-01 9.46386874e-01 4.45219755e-01 -6.32128358e-01
7.44584799e-01 4.27088171e-01 2.57516384e-01 -5.28995693e-01
-5.57820857e-01 -2.16555208e-01 3.84609699e-01 7.13574111e-01
-1.51244044e-01 2.92352885e-01 6.14075243e-01 2.22569555e-01
-9.22608852e-01 3.18242125e-02 2.08667383e-01 1.07168901e+00
-3.04965079e-01 -8.40383291e-01 -4.00625944e-01 4.85855937e-01
1.25126421e-01 6.49645478e-02 -2.36688092e-01 9.61735785e-01
1.81288332e-01 7.16571152e-01 6.17907405e-01 -7.42545187e-01
4.52603370e-01 -3.06880325e-01 2.62915432e-01 -4.05686766e-01
-3.68503600e-01 -3.59809309e-01 -2.73907065e-01 -8.14169645e-01
-2.16497704e-01 -9.20656562e-01 -1.40093136e+00 -8.19690406e-01
-2.47208104e-01 -5.05462438e-02 1.02985430e+00 7.73392379e-01
8.15057933e-01 1.44886315e-01 7.60998189e-01 -1.35093880e+00
-3.60827565e-01 -6.39962196e-01 -2.10608259e-01 3.75410289e-01
8.32863152e-01 -7.07057714e-01 -7.55476728e-02 -4.71534468e-02]
|
[7.672710418701172, -2.942967414855957]
|
3faa38f6-5535-4dd8-a41b-5096ee9a76d6
|
the-challenge-of-imputation-in-explainable
|
1907.12669
| null |
https://arxiv.org/abs/1907.12669v1
|
https://arxiv.org/pdf/1907.12669v1.pdf
|
The Challenge of Imputation in Explainable Artificial Intelligence Models
|
Explainable models in Artificial Intelligence are often employed to ensure transparency and accountability of AI systems. The fidelity of the explanations are dependent upon the algorithms used as well as on the fidelity of the data. Many real world datasets have missing values that can greatly influence explanation fidelity. The standard way to deal with such scenarios is imputation. This can, however, lead to situations where the imputed values may correspond to a setting which refer to counterfactuals. Acting on explanations from AI models with imputed values may lead to unsafe outcomes. In this paper, we explore different settings where AI models with imputation can be problematic and describe ways to address such scenarios.
|
['Ankur Teredesai', 'Muhammad Aurangzeb Ahmad', 'Carly Eckert']
|
2019-07-29
| null | null | null | null |
['explainable-models']
|
['computer-vision']
|
[ 5.54552674e-01 6.43718302e-01 -1.89865515e-01 -6.54569566e-01
-3.65574658e-01 -4.61754471e-01 5.27242780e-01 -6.36186497e-03
-1.24623232e-01 1.44351840e+00 5.84501684e-01 -5.54532409e-01
-4.47831184e-01 -8.33915830e-01 -9.30830896e-01 -4.82818842e-01
4.70166028e-01 5.70245862e-01 -7.19343901e-01 2.62425154e-01
5.12732208e-01 9.69739333e-02 -1.28628492e+00 4.51038748e-01
1.31086659e+00 2.09676296e-01 -5.37214816e-01 6.50499165e-01
-2.64326692e-01 1.03979993e+00 -8.69266152e-01 -8.49766314e-01
4.33877707e-01 -4.58326191e-01 -7.26052523e-01 -8.36114287e-02
1.18912630e-01 -5.83213151e-01 8.94554630e-02 1.01316154e+00
1.30057121e-02 -2.66612113e-01 9.55811679e-01 -1.77861381e+00
-8.45301211e-01 1.09662890e+00 -3.95547152e-01 -3.33782285e-01
5.98924637e-01 4.89591271e-01 5.05184889e-01 -1.84152871e-01
5.67617476e-01 1.51043773e+00 7.20198870e-01 4.83397394e-01
-1.70214844e+00 -8.39342356e-01 -1.42891943e-01 1.67620003e-01
-8.04914296e-01 -6.67474151e-01 4.44904119e-01 -5.22098422e-01
4.57560122e-01 5.68026125e-01 5.42056322e-01 1.42742896e+00
4.32392836e-01 1.20998368e-01 1.45345926e+00 -3.17690343e-01
4.73202765e-01 1.75120756e-01 1.76674053e-01 3.63291837e-02
1.00739074e+00 5.10789514e-01 -4.36815292e-01 -6.38632774e-01
4.73351091e-01 1.84403911e-01 -2.42773592e-01 -1.68797508e-01
-1.25052333e+00 1.00616324e+00 3.85098487e-01 -1.53658822e-01
-8.24831188e-01 3.08869153e-01 6.47875145e-02 5.04804671e-01
2.04321407e-02 9.24610496e-01 -5.57577014e-01 -2.91488141e-01
-6.34820819e-01 4.86225396e-01 8.67099762e-01 6.30957067e-01
5.24824202e-01 -5.83147034e-02 -2.49790534e-01 3.16453129e-01
3.44128191e-01 3.29036713e-01 8.93394500e-02 -1.40378523e+00
4.83090162e-01 6.91063702e-01 7.43118405e-01 -9.13076758e-01
-1.90131620e-01 9.68464557e-03 -9.39063430e-01 5.29928684e-01
9.86965537e-01 -2.34475851e-01 -1.16609347e+00 1.80306923e+00
1.48237467e-01 -1.86173320e-01 3.69277358e-01 8.74439240e-01
2.92107552e-01 3.67555261e-01 3.96746129e-01 -2.76862055e-01
7.83512712e-01 -2.29378179e-01 -1.11954284e+00 -3.32629830e-01
6.99339807e-01 -4.84885186e-01 1.03639674e+00 2.83129811e-01
-1.30199456e+00 7.26373196e-02 -6.01664662e-01 1.16515011e-01
-7.96671882e-02 -6.07450128e-01 8.44260991e-01 7.60127842e-01
-7.00198591e-01 5.71680844e-01 -5.78873694e-01 -1.34206846e-01
5.13039708e-01 5.15126467e-01 -5.41835904e-01 -2.13210836e-01
-1.18737411e+00 1.28886974e+00 2.96469212e-01 2.29044601e-01
-4.21508729e-01 -7.37802923e-01 -6.19001567e-01 3.02086651e-01
3.93090487e-01 -9.93207216e-01 8.96580815e-01 -1.39250779e+00
-8.75796556e-01 5.58283091e-01 -2.05943778e-01 -6.20480537e-01
1.01515579e+00 1.78463403e-02 -1.16172925e-01 -5.78554094e-01
2.42572725e-01 4.10821855e-01 3.86029571e-01 -1.24327052e+00
-2.79339224e-01 -4.80347216e-01 1.48390979e-01 8.68680421e-03
3.82603973e-01 -1.38706014e-01 6.41737401e-01 -3.15660089e-01
1.07167013e-01 -8.79101634e-01 -5.05593657e-01 -1.17610684e-02
-6.62214339e-01 1.94615155e-01 1.98600322e-01 -3.56059074e-01
9.69694555e-01 -1.78437650e+00 -1.77685946e-01 3.29823434e-01
1.79721609e-01 -1.71796218e-01 2.21017376e-01 2.18737319e-01
-1.07672408e-01 6.75499380e-01 -4.90645111e-01 7.60076568e-02
1.55870274e-01 5.17283618e-01 -3.80566746e-01 3.82551968e-01
2.31944844e-01 7.60899603e-01 -3.56870681e-01 -3.49543363e-01
2.36808375e-01 3.29053283e-01 -6.94016099e-01 2.94452697e-01
-6.28765952e-03 6.54429138e-01 -3.87374818e-01 4.09969896e-01
7.89745569e-01 -2.58468427e-02 2.15786263e-01 3.86621982e-01
-6.99954405e-02 4.04237509e-01 -1.11392009e+00 8.11863422e-01
-1.03050686e-01 3.31103325e-01 -1.14135914e-01 -8.05313230e-01
6.87530160e-01 3.28162313e-01 4.18334194e-02 -2.48860002e-01
-4.01366828e-03 1.16135612e-01 4.64544415e-01 -5.75605333e-01
1.97677553e-01 -4.62086946e-01 -5.64315282e-02 7.16272712e-01
-6.54607892e-01 -2.62838572e-01 -2.60594994e-01 -2.36795396e-02
1.06149077e+00 9.39578842e-03 6.46522224e-01 -1.55937687e-01
6.44767890e-03 4.15826619e-01 1.11919105e+00 1.23348117e+00
-2.39622578e-01 6.09948754e-01 8.71658742e-01 -8.77092123e-01
-1.19423068e+00 -9.69183087e-01 -4.42762852e-01 2.32597515e-01
-2.35961065e-01 1.84090808e-01 -6.69467568e-01 -5.69541991e-01
2.91386992e-01 1.48048806e+00 -7.71029890e-01 -4.25362021e-01
-9.56958160e-02 -7.91960180e-01 3.00344884e-01 3.67612690e-01
1.65042579e-01 -1.12069321e+00 -8.64372313e-01 2.89454728e-01
-2.08599895e-01 -6.35510683e-01 9.45090130e-02 4.20890450e-02
-9.12960649e-01 -1.07791579e+00 -2.29789000e-02 3.75127017e-01
1.10055768e+00 -1.03249028e-01 1.24308968e+00 4.85904276e-01
3.61632794e-01 -7.66736344e-02 -1.12883933e-01 -7.83764243e-01
-9.01904821e-01 -4.09064800e-01 -4.75488342e-02 -2.01977253e-01
5.68254590e-01 -5.13370395e-01 -2.91556180e-01 1.17346108e-01
-1.11106634e+00 4.49221730e-01 4.35662985e-01 9.39333141e-01
3.14713776e-01 -1.59757778e-01 6.42555952e-01 -1.58706379e+00
5.97857893e-01 -8.97447884e-01 -6.66667521e-01 4.13666576e-01
-9.79249775e-01 3.61067116e-01 7.05197334e-01 -5.09273767e-01
-9.80092645e-01 4.15914766e-02 4.68655825e-01 -2.83492021e-02
-5.94244003e-01 6.84619904e-01 -3.52066875e-01 2.85046607e-01
7.27260530e-01 -5.00811100e-01 -8.41425732e-02 -8.89536664e-02
2.29655579e-01 8.10453892e-01 3.52468163e-01 -5.42336941e-01
6.15792811e-01 3.16687852e-01 1.12532295e-01 -7.31060654e-02
-7.31840074e-01 6.11763775e-01 -4.09836739e-01 -3.11958846e-02
5.76406598e-01 -4.64191526e-01 -6.67178392e-01 -6.78061470e-02
-1.07465231e+00 -2.68274546e-01 -4.10220146e-01 7.34828472e-01
-6.47296906e-01 -1.75222844e-01 6.48755506e-02 -1.14041078e+00
1.72181174e-01 -1.42113292e+00 3.26126724e-01 2.69872665e-01
-8.60190153e-01 -8.87433410e-01 -2.40300059e-01 5.32347679e-01
5.81786513e-01 7.82110274e-01 1.04030252e+00 -6.47351384e-01
-6.39487445e-01 -2.19209105e-01 3.25676538e-02 -4.32419360e-01
1.68348238e-01 2.03954935e-01 -9.15600181e-01 1.98833004e-01
2.31824312e-02 -1.89183399e-01 4.49797809e-01 7.59504378e-01
1.00751626e+00 -1.09035373e+00 -1.05232581e-01 4.28444147e-01
1.12094474e+00 2.52944708e-01 7.70955622e-01 4.37071025e-01
4.23271924e-01 9.24930871e-01 4.93859321e-01 4.06028509e-01
5.07964551e-01 4.94841278e-01 4.80730563e-01 -6.65466562e-02
4.90697980e-01 -3.04306030e-01 9.15294588e-02 -1.07254259e-01
7.16676638e-02 -2.43805125e-01 -1.14767170e+00 5.87546885e-01
-2.09769654e+00 -1.16946161e+00 -6.16477549e-01 2.51574206e+00
8.65123451e-01 8.44829902e-02 -1.04941763e-01 2.37561285e-01
6.88787043e-01 -4.63648379e-01 -7.06575334e-01 -1.08397079e+00
-1.32603332e-01 -4.61382866e-01 4.89791751e-01 5.78940392e-01
-4.18864429e-01 3.29340488e-01 7.35944033e+00 -1.50942564e-01
-5.09088933e-01 -3.33140679e-02 8.70367885e-01 -1.31732821e-01
-1.04741776e+00 5.81853211e-01 -5.90311773e-02 8.75755131e-01
1.14693236e+00 -6.91895783e-01 4.72656757e-01 4.32698101e-01
4.63358879e-01 -3.18038881e-01 -1.37132490e+00 4.64982212e-01
-4.07916754e-01 -1.11639977e+00 9.17315781e-02 3.37188184e-01
8.96737516e-01 -5.61191857e-01 -2.35810969e-02 3.01391166e-02
9.79680657e-01 -1.56994057e+00 6.74797833e-01 7.91500926e-01
5.93824804e-01 -8.60198081e-01 1.04314530e+00 5.73630333e-01
-2.71181148e-02 -1.71082720e-01 -6.18878961e-01 -7.06950605e-01
4.27430086e-02 6.80992782e-01 -8.25513661e-01 9.15429965e-02
4.85163808e-01 2.01827452e-01 -1.39533922e-01 9.91224349e-01
-4.86987680e-01 6.93852365e-01 -1.37831852e-01 2.34277397e-01
-9.25830752e-02 -3.92072201e-01 4.17463481e-01 5.27955830e-01
3.97751868e-01 4.28159446e-01 -4.26948547e-01 1.44435632e+00
9.48804151e-03 -3.48333448e-01 -1.19615495e+00 -7.33734369e-02
6.25788927e-01 4.98659164e-01 -2.07110599e-01 -3.22866440e-01
-4.73042190e-01 4.37914342e-01 1.96981385e-01 4.24874544e-01
-6.96260095e-01 2.90370673e-01 8.25822175e-01 2.50422418e-01
-5.26947021e-01 2.05456004e-01 -9.04415250e-01 -1.25907636e+00
-1.09730028e-01 -1.14395022e+00 4.78495866e-01 -1.06734467e+00
-1.36213601e+00 1.52143687e-01 1.07640475e-01 -9.93930221e-01
-4.60946381e-01 3.79893817e-02 -6.27357185e-01 9.55837190e-01
-1.19562840e+00 -8.83743763e-01 -3.73429842e-02 2.83237457e-01
1.02055945e-01 1.58446699e-01 9.07465994e-01 -3.61942500e-01
-3.38890105e-01 4.38539654e-01 -4.17712554e-02 -3.05176169e-01
6.92053914e-01 -1.30930388e+00 3.81594092e-01 8.78154695e-01
3.22046988e-02 8.17942917e-01 1.16429329e+00 -8.61306429e-01
-1.26756191e+00 -8.42995644e-01 1.14918065e+00 -6.78407967e-01
2.35394701e-01 1.16978191e-01 -1.20242143e+00 1.17087209e+00
1.23937145e-01 -1.95413291e-01 7.85497308e-01 2.43887186e-01
-2.67821044e-01 2.30645314e-01 -1.74761593e+00 8.17350328e-01
8.83856177e-01 -1.75274521e-01 -7.71209478e-01 1.96943164e-01
5.35100698e-01 -3.81448478e-01 -7.64997840e-01 2.18604892e-01
6.07409954e-01 -1.15182555e+00 6.31435812e-01 -1.27627385e+00
7.23711073e-01 -3.17501068e-01 6.77029695e-03 -1.47152114e+00
-3.26380998e-01 -4.89609897e-01 1.69396177e-01 1.30069911e+00
7.30778873e-01 -9.17730927e-01 4.76507604e-01 2.11296058e+00
3.90882462e-01 -1.14820778e-01 -8.91538262e-01 -2.97853470e-01
1.50407553e-01 -5.39979219e-01 1.42387378e+00 1.32829320e+00
4.00579581e-03 -2.37715900e-01 -5.93982458e-01 3.08564365e-01
9.98403430e-01 2.67310679e-01 8.68484616e-01 -1.49482167e+00
4.38888222e-02 -1.20719276e-01 -4.41583574e-01 2.21512839e-02
3.03763837e-01 -4.99766529e-01 -1.64850101e-01 -1.46526825e+00
5.49975991e-01 -5.48171639e-01 -8.41151327e-02 6.94232821e-01
-4.97181356e-01 -2.17581298e-02 3.71100575e-01 2.94882238e-01
2.57335976e-02 1.26085073e-01 8.75062406e-01 3.14830914e-02
-1.66277006e-01 1.88801795e-01 -1.24920678e+00 8.80379081e-01
9.54991400e-01 -1.08945203e+00 -2.19711319e-01 -6.52494133e-01
5.00765324e-01 5.15568018e-01 6.44098938e-01 -4.16211367e-01
8.79577696e-02 -9.61846471e-01 5.99280894e-01 -2.40113027e-02
-9.04267803e-02 -1.22698057e+00 1.24314046e+00 6.27550602e-01
-8.66493285e-01 4.72953133e-02 -5.08087203e-02 3.45603466e-01
4.05571610e-02 -3.44749182e-01 5.57385147e-01 -2.08211452e-01
1.55471951e-01 -1.01034611e-01 -4.28685755e-01 -2.04730153e-01
1.05611444e+00 -5.16194522e-01 -6.24491274e-01 -6.69365466e-01
-7.02449501e-01 4.10750151e-01 9.90667820e-01 1.39851555e-01
3.99317414e-01 -1.33087814e+00 -9.33763146e-01 2.15191916e-01
-7.24368095e-02 -1.46338791e-01 4.01670896e-02 6.64706171e-01
-1.06974721e-01 2.14293391e-01 -4.87546086e-01 -9.05743390e-02
-8.56693745e-01 5.25674820e-01 4.18154299e-01 3.11289262e-02
-4.04204279e-01 1.65017098e-01 3.61047119e-01 -4.74502742e-01
-1.09900482e-01 -2.19189793e-01 -1.59887820e-01 -3.83613974e-01
3.88784200e-01 3.57358187e-01 -3.67255121e-01 -4.51770544e-01
-3.27479094e-01 1.15459442e-01 1.17101192e-01 -1.78045034e-01
1.44274282e+00 -3.18760008e-01 -7.10558891e-02 5.60467362e-01
2.79596061e-01 -6.39481694e-02 -1.30869102e+00 1.82425961e-01
-2.50798315e-02 -1.04434264e+00 -8.42545107e-02 -1.21589077e+00
-7.76445627e-01 7.16402352e-01 1.47632658e-01 4.20676827e-01
9.92707908e-01 -5.00001132e-01 1.14767134e-01 1.00290522e-01
4.90794748e-01 -6.92038953e-01 -9.61862564e-01 -3.48631561e-01
1.12993300e+00 -1.54714847e+00 1.63349867e-01 4.23706323e-02
-1.01923573e+00 9.15944695e-01 5.59323192e-01 2.86740512e-01
7.42104873e-02 2.72908628e-01 3.51803333e-01 4.37557995e-02
-1.07846689e+00 2.12572962e-01 -1.14773698e-02 7.95489371e-01
5.82795680e-01 4.86939728e-01 -6.91934705e-01 7.17252254e-01
-4.79249239e-01 3.53334397e-01 1.24648285e+00 5.95870256e-01
1.77913412e-01 -1.05498350e+00 -9.90174055e-01 8.55779231e-01
-7.01027334e-01 9.70136002e-02 -7.74362326e-01 6.58866823e-01
2.42580529e-02 1.30725431e+00 -1.89034846e-02 9.12679918e-03
3.82809460e-01 9.25260857e-02 1.57062951e-02 -3.11610788e-01
-7.24892080e-01 -5.01535177e-01 1.91735834e-01 -6.54868782e-01
-3.94764185e-01 -8.91056716e-01 -1.15937173e+00 -9.26749110e-01
-2.89678305e-01 1.82890087e-01 3.58539045e-01 1.04900956e+00
4.65418845e-01 1.29409969e-01 3.55512172e-01 -2.49526352e-01
-7.26401687e-01 -7.43680954e-01 -3.38106394e-01 8.27240586e-01
5.59865057e-01 -6.11260772e-01 -5.91366053e-01 -8.01781863e-02]
|
[8.714235305786133, 5.6551513671875]
|
66fba441-fbe8-410f-8ee3-16f202c9aa33
|
spectral-feature-mapping-with-mimic-loss-for
|
1803.09816
| null |
http://arxiv.org/abs/1803.09816v1
|
http://arxiv.org/pdf/1803.09816v1.pdf
|
Spectral feature mapping with mimic loss for robust speech recognition
|
For the task of speech enhancement, local learning objectives are agnostic to
phonetic structures helpful for speech recognition. We propose to add a global
criterion to ensure de-noised speech is useful for downstream tasks like ASR.
We first train a spectral classifier on clean speech to predict senone labels.
Then, the spectral classifier is joined with our speech enhancer as a noisy
speech recognizer. This model is taught to imitate the output of the spectral
classifier alone on clean speech. This \textit{mimic loss} is combined with the
traditional local criterion to train the speech enhancer to produce de-noised
speech. Feeding the de-noised speech to an off-the-shelf Kaldi training recipe
for the CHiME-2 corpus shows significant improvements in WER.
|
['Eric Fosler-Lussier', 'Deblin Bagchi', 'Peter Plantinga', 'Adam Stiff']
|
2018-03-26
| null | null | null | null |
['robust-speech-recognition']
|
['speech']
|
[ 5.20704269e-01 5.78796744e-01 2.20504925e-01 -5.65313637e-01
-1.43150830e+00 -5.04573703e-01 2.78650403e-01 -2.92776763e-01
-4.37806934e-01 3.43051195e-01 4.18457240e-01 -7.77427554e-01
1.85066059e-01 -3.50252151e-01 -5.65081835e-01 -8.69065762e-01
3.73754948e-01 7.12275803e-02 1.29776821e-01 -3.79288763e-01
-4.15258676e-01 3.03117007e-01 -1.25843263e+00 5.60101271e-01
7.63096929e-01 7.01084733e-01 7.66891599e-01 1.03821290e+00
1.30013674e-01 7.14285612e-01 -7.50527084e-01 -2.72936523e-01
3.54822040e-01 -7.15284109e-01 -5.07611096e-01 -1.17465770e-02
2.93799460e-01 -2.79353112e-01 -4.92160618e-01 1.23371994e+00
7.83402085e-01 4.33986872e-01 3.90042126e-01 -3.65255862e-01
-2.85226673e-01 1.23941934e+00 -1.86671644e-01 4.02799010e-01
-1.54172778e-01 1.87197566e-01 1.12678051e+00 -9.08484578e-01
2.04092622e-01 1.31802440e+00 4.32198107e-01 7.29896069e-01
-1.33174086e+00 -6.24564290e-01 1.01993740e-01 -1.56126648e-01
-8.20725679e-01 -1.25565863e+00 7.98189044e-01 -5.45256436e-02
1.11959147e+00 4.44616228e-01 1.90447137e-01 1.11097658e+00
-4.54366028e-01 7.69428253e-01 7.91794002e-01 -6.06696248e-01
2.23937064e-01 1.03003189e-01 1.01301298e-01 3.01565677e-01
-6.22779012e-01 4.44888741e-01 -6.16441488e-01 1.65107489e-01
2.62505054e-01 -9.34704244e-01 -3.31451416e-01 3.67851377e-01
-6.03082716e-01 6.71994865e-01 2.59296298e-01 3.45788240e-01
-3.96240354e-01 2.41370261e-01 4.55345869e-01 5.84139824e-01
9.26087022e-01 6.61286056e-01 -7.09588587e-01 -3.94093633e-01
-1.01676035e+00 -7.52165616e-02 4.11396444e-01 4.99367356e-01
5.40244401e-01 7.08614707e-01 -1.76554605e-01 1.62207663e+00
3.22163135e-01 5.31959951e-01 3.45695406e-01 -1.14765477e+00
5.69084167e-01 -3.52429420e-01 -3.60764533e-01 -4.91902269e-02
-5.78250140e-02 -8.27889800e-01 -4.41049367e-01 2.43850410e-01
2.57298261e-01 -4.92311209e-01 -1.25170648e+00 1.80542731e+00
1.16830960e-01 1.44884244e-01 3.01947236e-01 8.12392592e-01
5.23503244e-01 1.23135984e+00 -1.43898306e-02 -2.95309752e-01
8.70816588e-01 -1.22591102e+00 -9.04752791e-01 -5.38339436e-01
8.15428615e-01 -1.00307083e+00 1.08432007e+00 4.16472256e-01
-1.41117918e+00 -7.02396750e-01 -9.38188016e-01 -1.46404862e-01
-1.07244454e-01 3.84329766e-01 -4.43675108e-02 8.62165034e-01
-1.11407697e+00 5.40709019e-01 -7.55838692e-01 3.42169702e-02
-1.32303557e-03 1.85729429e-01 6.30985349e-02 1.14029244e-01
-1.24635088e+00 9.53073502e-01 4.87874329e-01 1.65189542e-02
-1.10115814e+00 -7.93546736e-01 -9.64995742e-01 2.02248842e-01
1.83881596e-01 -2.80051410e-01 1.81967783e+00 -1.21537769e+00
-2.09128952e+00 6.19193375e-01 -2.53911048e-01 -3.95591795e-01
3.01608771e-01 -1.73986331e-01 -6.11665845e-01 3.26359160e-02
-1.55579701e-01 3.74030530e-01 1.05761755e+00 -1.15107441e+00
-6.88657701e-01 2.81559676e-02 -3.76273423e-01 5.39667487e-01
-9.36514288e-02 3.69756430e-01 -3.37269127e-01 -9.25938904e-01
1.65074095e-01 -6.72744274e-01 -1.61260828e-01 -6.77457809e-01
-5.63084483e-01 -1.50626585e-01 9.92762446e-01 -1.29496479e+00
1.12408841e+00 -2.50276709e+00 3.26411203e-02 4.36856180e-01
-4.38987792e-01 6.32972598e-01 -4.38322037e-01 -3.18527445e-02
-5.15774906e-01 -5.25688045e-02 -2.92973012e-01 -7.41660893e-01
-1.23747930e-01 6.39646947e-02 -3.23380738e-01 1.87408373e-01
5.21960974e-01 5.84491134e-01 -9.21137214e-01 2.68840417e-02
4.30814862e-01 5.04461646e-01 -6.07110202e-01 3.51211160e-01
-2.60464162e-01 3.73286337e-01 1.47290513e-01 9.02677551e-02
4.34237629e-01 5.51778018e-01 1.16963796e-01 6.47083968e-02
-1.73995197e-01 1.31039548e+00 -1.02599943e+00 1.49672973e+00
-6.03630543e-01 5.32567918e-01 6.42057598e-01 -9.68453348e-01
7.87892938e-01 6.95246100e-01 1.25977444e-02 -6.10080481e-01
-2.68013701e-02 4.31029260e-01 3.84182334e-01 -1.95249990e-01
2.50639260e-01 -2.98421562e-01 3.20201278e-01 1.35158852e-01
3.82945240e-01 -4.77379978e-01 -2.36747503e-01 7.56183341e-02
1.30087554e+00 3.69961262e-02 -1.94183618e-01 -2.43043453e-01
3.34266037e-01 -3.05666000e-01 5.18258154e-01 6.22824907e-01
-1.26823008e-01 7.37471104e-01 3.02128285e-01 4.62977231e-01
-1.36615229e+00 -1.33896244e+00 -9.97332707e-02 1.59619105e+00
-5.97213089e-01 -3.17677677e-01 -1.09546173e+00 -5.89932561e-01
-3.27514023e-01 1.23194814e+00 -6.97323829e-02 -2.11012051e-01
-8.31748068e-01 -4.49678272e-01 8.08555603e-01 3.84137839e-01
-7.66527429e-02 -9.83429611e-01 4.65798587e-01 4.82796848e-01
-6.53824434e-02 -8.89487028e-01 -7.04538643e-01 9.53339994e-01
-4.24632460e-01 -2.61226058e-01 -7.16745198e-01 -1.05945635e+00
3.93575281e-01 -3.23323496e-02 6.37719214e-01 -9.81796384e-02
2.54573107e-01 -2.49185160e-01 -3.68104666e-01 -2.35137060e-01
-1.27469909e+00 1.58679724e-01 1.31990358e-01 -2.11056974e-02
4.42144498e-02 -4.85482484e-01 -8.83590356e-02 1.14207312e-01
-5.49665213e-01 -7.42540881e-02 4.06073868e-01 9.62466180e-01
4.33146834e-01 1.85730979e-01 9.81562912e-01 -7.29419112e-01
5.62248170e-01 -7.54507035e-02 -6.46578133e-01 -1.16612792e-01
-1.55689955e-01 1.98215723e-01 7.90450573e-01 -4.50253636e-01
-1.66525817e+00 3.79384220e-01 -1.05400527e+00 -4.34354842e-01
-1.41548902e-01 5.36584377e-01 -6.97368562e-01 4.61089462e-01
7.88895249e-01 -7.35194189e-03 -1.83210701e-01 -8.26111019e-01
5.37237942e-01 1.22656643e+00 7.89522171e-01 -3.89019251e-01
8.56826007e-01 -4.10489231e-01 -6.27047956e-01 -1.30760050e+00
-8.42422485e-01 -5.88675737e-01 -2.34893799e-01 1.24158762e-01
6.50276065e-01 -8.88838053e-01 -1.77932143e-01 5.96664250e-01
-1.23676348e+00 -8.05930018e-01 -4.99820292e-01 5.99424422e-01
-6.37736797e-01 1.40248582e-01 -9.32224631e-01 -1.26251614e+00
-2.74448603e-01 -1.24786067e+00 8.70231748e-01 -1.29553467e-01
-2.12452170e-02 -6.04141891e-01 -5.52860685e-02 4.03313816e-01
5.45463860e-01 -8.79852772e-01 8.88297677e-01 -8.19796741e-01
-1.42716959e-01 1.48730516e-01 1.36657432e-01 1.16620612e+00
2.10663274e-01 -1.89325623e-02 -1.77280903e+00 -2.75986400e-02
1.61482960e-01 -1.82705030e-01 1.05560660e+00 6.55537784e-01
1.09351754e+00 -1.81574225e-01 5.11936583e-02 6.61744833e-01
7.32304454e-01 5.51804125e-01 6.68284416e-01 -1.36000693e-01
5.03448665e-01 7.24514902e-01 3.65969896e-01 -1.60929441e-01
-2.92770177e-01 5.93059301e-01 -1.02035403e-01 -3.47304910e-01
-6.46311641e-01 -3.92377585e-01 8.80951345e-01 1.43806767e+00
3.96596044e-01 -2.27482691e-01 -7.72472024e-01 6.13611698e-01
-1.52826011e+00 -9.61224377e-01 -2.16276627e-02 1.92416322e+00
1.33082128e+00 2.95231760e-01 -1.00647891e-02 2.13225231e-01
9.08892214e-01 2.84341127e-01 -2.57526606e-01 -6.74325407e-01
-2.30008900e-01 5.35571754e-01 4.35193390e-01 1.08009446e+00
-1.12185454e+00 1.45869529e+00 6.26274586e+00 1.33163202e+00
-1.02002108e+00 2.84375340e-01 7.39684463e-01 -1.06889546e-01
-3.99680972e-01 -1.18369339e-02 -7.57260263e-01 3.40290636e-01
1.50130510e+00 2.13598117e-01 9.02709782e-01 7.39281058e-01
7.95384109e-01 1.14280313e-01 -1.02172577e+00 6.88043654e-01
-3.17023546e-01 -9.66044426e-01 -4.35949326e-01 -2.05461934e-01
5.87569892e-01 3.10487926e-01 1.86627060e-01 5.28025329e-01
6.05148971e-01 -9.69255567e-01 8.13976884e-01 1.88525002e-02
8.35773468e-01 -9.70504284e-01 5.59712350e-01 4.34233159e-01
-8.94455850e-01 1.30785704e-01 -2.59029299e-01 2.52438843e-01
3.47589970e-01 8.17666829e-01 -1.19734800e+00 1.85350433e-01
4.00350511e-01 1.74276993e-01 -8.70035514e-02 8.97022605e-01
-5.33408701e-01 1.38121712e+00 -5.26025057e-01 3.66900295e-01
2.23610684e-01 -6.03352934e-02 7.34687567e-01 1.47629762e+00
2.60480762e-01 5.08919172e-03 -1.52858734e-01 6.93466842e-01
-1.88031256e-01 1.68289408e-01 -4.09759909e-01 -2.98996896e-01
5.43308556e-01 9.63722765e-01 -1.22958362e-01 -3.54599983e-01
-2.48405114e-01 1.01351476e+00 2.97418326e-01 5.86089253e-01
-4.31988269e-01 -5.32366276e-01 9.35094953e-01 -1.24847755e-01
3.87587458e-01 -1.12333611e-01 -3.29326421e-01 -7.63809621e-01
-1.73262358e-01 -1.17277706e+00 -1.15015969e-01 -9.40164089e-01
-1.19638801e+00 6.67632937e-01 -4.13842320e-01 -5.36439419e-01
-5.16650915e-01 -6.88640416e-01 -7.26044357e-01 1.63563406e+00
-1.48150814e+00 -9.01237726e-01 4.63740855e-01 3.09533149e-01
9.28227544e-01 -1.46174297e-01 7.11455107e-01 4.36975360e-01
-5.32077730e-01 6.44991755e-01 1.71006560e-01 2.65662938e-01
7.72498190e-01 -1.44265652e+00 8.54491591e-01 1.20966315e+00
1.63352024e-02 3.88632745e-01 8.19483876e-01 -7.56519377e-01
-7.19961107e-01 -1.27317095e+00 9.00569081e-01 -7.08584636e-02
7.01401949e-01 -5.04578710e-01 -1.04479027e+00 6.70611739e-01
2.28792280e-01 -2.76082516e-01 3.95487338e-01 2.35540450e-01
-1.52144924e-01 -2.60559767e-01 -8.25340927e-01 5.77064097e-01
9.64389384e-01 -9.29036498e-01 -8.33011329e-01 1.91174820e-01
1.32404113e+00 -3.54068756e-01 -3.67240101e-01 1.06209725e-01
-8.41434151e-02 -3.20570111e-01 6.71994805e-01 -5.70592046e-01
1.10681936e-01 -2.79394299e-01 -3.44930023e-01 -2.10337758e+00
-7.47130439e-02 -1.05262041e+00 2.12724671e-01 1.57415748e+00
1.00223422e+00 -4.51410949e-01 4.90752786e-01 2.12162286e-01
-1.00842834e+00 -8.19084719e-02 -9.79287207e-01 -8.70159566e-01
9.12149400e-02 -8.57517123e-01 1.95529670e-01 6.90587759e-01
4.34249453e-02 5.19555867e-01 -2.83917546e-01 4.23709214e-01
4.28387016e-01 -8.16746593e-01 3.70043367e-01 -7.11868465e-01
-7.34986007e-01 -3.85039061e-01 1.31875068e-01 -1.22534192e+00
3.22190374e-01 -1.05599701e+00 7.94220328e-01 -1.13331902e+00
-3.56864393e-01 -4.20128107e-01 -2.98412830e-01 5.27209163e-01
-4.61779952e-01 -1.36384815e-01 2.17653498e-01 -3.12382638e-01
1.46676794e-01 5.89578688e-01 9.46968496e-01 -7.08331764e-02
-5.39082110e-01 3.69805306e-01 -4.95120108e-01 6.52309358e-01
8.91781628e-01 -5.58566511e-01 -2.69909263e-01 -4.52195466e-01
-2.69353211e-01 1.01634555e-01 -6.39660358e-02 -8.82862449e-01
1.55661330e-01 6.86148182e-02 -6.21884651e-02 -5.38752735e-01
7.18364179e-01 -4.49374199e-01 -2.56603271e-01 6.03644252e-02
-6.66284084e-01 -5.54730654e-01 2.61478394e-01 4.63572070e-02
-2.95248538e-01 -6.36041164e-01 1.21123290e+00 1.61424890e-01
-3.71239543e-01 -1.87543020e-01 -6.57523215e-01 7.14376569e-02
3.92166674e-01 1.93330973e-01 -2.49569744e-01 -5.30997038e-01
-9.83672500e-01 -1.61488578e-01 3.30819525e-02 1.61602497e-01
3.15053523e-01 -9.73912358e-01 -8.70672524e-01 5.29086292e-01
-3.50968212e-01 -3.91539559e-02 7.19415247e-02 5.20509541e-01
-1.83444202e-01 2.58122087e-01 5.58307171e-01 -3.04351956e-01
-1.28061306e+00 3.63146126e-01 8.08408380e-01 6.52450547e-02
-4.13044840e-01 1.26787102e+00 3.49209875e-01 -7.42911637e-01
5.62088907e-01 -4.07609463e-01 3.47340196e-01 -1.72948852e-01
4.21440601e-01 2.32990384e-01 5.57099104e-01 -4.79475081e-01
-1.47699833e-01 -1.16902076e-01 9.18051451e-02 -8.22227120e-01
1.40491951e+00 -2.22587124e-01 2.51398832e-01 4.74440128e-01
1.30253518e+00 4.47956741e-01 -1.30306947e+00 -2.52108663e-01
1.96668789e-01 -5.08414917e-02 6.45475864e-01 -1.16804063e+00
-8.31166625e-01 1.03891802e+00 4.35121715e-01 3.22207540e-01
1.07634842e+00 1.42302886e-01 7.01730788e-01 4.01147157e-01
-2.37786889e-01 -1.60448921e+00 -1.04323305e-01 1.06043684e+00
9.32934582e-01 -9.92381692e-01 -7.51645923e-01 -4.95006382e-01
-7.16030359e-01 9.49167073e-01 3.15518945e-01 1.30023047e-01
5.81014037e-01 7.73356318e-01 3.94055784e-01 2.75624603e-01
-6.38855398e-01 -4.73319292e-01 3.17953378e-01 6.40020847e-01
4.87419516e-01 1.03662625e-01 1.78589866e-01 5.64513505e-01
-7.12907732e-01 -5.22193491e-01 2.42865473e-01 4.19076085e-01
-9.00072217e-01 -1.18272853e+00 -5.01386046e-01 2.98965603e-01
-5.81431270e-01 -6.46628141e-01 -3.90555292e-01 1.31877847e-02
-7.33860284e-02 1.33061910e+00 1.41230687e-01 -2.61480778e-01
4.61119533e-01 6.11323118e-01 -5.24020800e-03 -1.04155314e+00
-7.93552041e-01 8.77085686e-01 6.74691260e-01 -2.24408343e-01
1.72337860e-01 -5.67859888e-01 -1.31678116e+00 2.70972047e-02
-5.96876979e-01 1.38048798e-01 1.04036510e+00 9.39504445e-01
-1.61709726e-01 9.70756590e-01 7.90486455e-01 -6.94940448e-01
-8.69566917e-01 -1.25058103e+00 -6.70755744e-01 1.42081708e-01
6.73848510e-01 4.65902612e-02 -7.29389191e-01 1.15208939e-01]
|
[14.682263374328613, 6.289509296417236]
|
a85c2c50-839c-4fd2-894f-06b5b453bd27
|
videomae-masked-autoencoders-are-data-1
|
2203.12602
| null |
https://arxiv.org/abs/2203.12602v3
|
https://arxiv.org/pdf/2203.12602v3.pdf
|
VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training
|
Pre-training video transformers on extra large-scale datasets is generally required to achieve premier performance on relatively small datasets. In this paper, we show that video masked autoencoders (VideoMAE) are data-efficient learners for self-supervised video pre-training (SSVP). We are inspired by the recent ImageMAE and propose customized video tube masking with an extremely high ratio. This simple design makes video reconstruction a more challenging self-supervision task, thus encouraging extracting more effective video representations during this pre-training process. We obtain three important findings on SSVP: (1) An extremely high proportion of masking ratio (i.e., 90% to 95%) still yields favorable performance of VideoMAE. The temporally redundant video content enables a higher masking ratio than that of images. (2) VideoMAE achieves impressive results on very small datasets (i.e., around 3k-4k videos) without using any extra data. (3) VideoMAE shows that data quality is more important than data quantity for SSVP. Domain shift between pre-training and target datasets is an important issue. Notably, our VideoMAE with the vanilla ViT can achieve 87.4% on Kinetics-400, 75.4% on Something-Something V2, 91.3% on UCF101, and 62.6% on HMDB51, without using any extra data. Code is available at https://github.com/MCG-NJU/VideoMAE.
|
['LiMin Wang', 'Jue Wang', 'Yibing Song', 'Zhan Tong']
|
2022-03-23
|
videomae-masked-autoencoders-are-data
|
https://arxiv.org/abs/2203.12602
|
https://arxiv.org/pdf/2203.12602
| null |
['video-reconstruction', 'self-supervised-action-recognition']
|
['computer-vision', 'computer-vision']
|
[ 1.29563600e-01 -2.69148022e-01 -4.67317283e-01 -1.88978299e-01
-7.13064492e-01 -2.64460385e-01 4.00028378e-01 -6.53827608e-01
-7.20828354e-01 7.81979442e-01 3.30604881e-01 -3.49041373e-01
3.05547118e-01 -4.75717276e-01 -1.22439170e+00 -7.67003655e-01
-2.18998328e-01 -2.35616669e-01 3.34196061e-01 -2.53590912e-01
-1.52102634e-01 -2.40118597e-02 -1.80533493e+00 6.87468708e-01
7.55957544e-01 9.65504408e-01 6.17563307e-01 8.08719695e-01
2.33686700e-01 1.05900013e+00 -6.68368101e-01 -3.29511195e-01
4.85543251e-01 -4.51473147e-01 -6.50159955e-01 3.24796855e-01
5.30430913e-01 -6.28976703e-01 -6.67835832e-01 9.53994632e-01
4.09322828e-01 1.69389009e-01 4.44375277e-01 -1.32724714e+00
-6.37993634e-01 7.04311788e-01 -5.96789658e-01 5.49509287e-01
8.19121078e-02 5.12496054e-01 7.17774808e-01 -1.07479346e+00
5.42568624e-01 8.61960113e-01 3.83144975e-01 7.07016706e-01
-9.45663869e-01 -1.02055311e+00 7.69798383e-02 5.50771475e-01
-1.55839384e+00 -6.90034866e-01 4.25811172e-01 -4.56701398e-01
8.93045068e-01 2.92224199e-01 4.67143446e-01 1.48568320e+00
-5.65348119e-02 7.48802960e-01 1.18576837e+00 -2.73749530e-01
1.68026552e-01 3.68150175e-01 -1.41686901e-01 5.14931262e-01
1.62410989e-01 2.03394040e-01 -7.81163752e-01 3.70896250e-01
8.29870760e-01 8.00918490e-02 -6.67841613e-01 1.19987294e-01
-1.23432970e+00 8.77007544e-01 3.21668237e-01 4.04728800e-01
-3.09679508e-01 1.28157347e-01 3.75471503e-01 7.05269635e-01
2.04824701e-01 8.22656974e-02 -4.60356206e-01 -2.91382581e-01
-9.65969086e-01 -2.16816083e-01 3.87896895e-01 1.16301894e+00
6.84504330e-01 5.85698068e-01 -1.39022142e-01 8.33600104e-01
2.96438914e-02 5.93615115e-01 7.08446383e-01 -1.05217993e+00
7.29898751e-01 -8.98109190e-03 -1.61724120e-01 -7.85712063e-01
7.65724620e-03 -3.11566740e-01 -1.16898346e+00 1.83447272e-01
3.84984910e-01 -3.27770293e-01 -9.81524467e-01 1.88110161e+00
-2.30782684e-02 4.62146461e-01 4.08358395e-01 9.57158387e-01
1.11254954e+00 1.13285410e+00 4.07152660e-02 -4.84941959e-01
1.40629876e+00 -1.10203290e+00 -7.95380831e-01 -2.43795067e-01
4.56825674e-01 -5.12766600e-01 1.28044236e+00 5.76538086e-01
-1.01413524e+00 -9.05482292e-01 -1.18909454e+00 1.53871924e-01
-1.39844269e-01 1.69845432e-01 5.59938073e-01 6.24524176e-01
-1.23889983e+00 5.18490314e-01 -6.50324106e-01 -1.27279386e-01
6.21744275e-01 4.61975574e-01 -5.69228172e-01 -3.26600075e-01
-1.27351010e+00 5.23442566e-01 5.30652344e-01 -2.38183334e-01
-1.07880843e+00 -7.40290344e-01 -7.51187623e-01 3.04344706e-02
5.93511224e-01 -2.60137618e-01 1.04629922e+00 -1.37768793e+00
-1.44480336e+00 6.88702106e-01 -9.11692008e-02 -7.36208260e-01
4.65484291e-01 -3.51862252e-01 -7.45574832e-01 5.84085643e-01
-1.12205319e-01 1.05631983e+00 1.24894583e+00 -1.16371632e+00
-6.48644805e-01 7.62547627e-02 -8.52598399e-02 2.40262926e-01
-6.90700650e-01 3.48022650e-03 -8.64545465e-01 -8.99330676e-01
-3.40847790e-01 -8.99057329e-01 6.14481419e-02 -1.96919769e-01
-5.80670424e-02 1.22093983e-01 9.57727373e-01 -9.01432812e-01
1.30591071e+00 -2.42746329e+00 8.54546875e-02 -1.15329169e-01
3.25531989e-01 6.04624152e-01 -2.29019165e-01 3.56580541e-02
-3.54379177e-01 9.04896259e-02 -1.84110209e-01 -1.93829522e-01
-3.58499825e-01 1.67908564e-01 -1.77453130e-01 3.21591765e-01
2.38576233e-01 8.15570235e-01 -6.33742869e-01 -5.07556975e-01
2.82118350e-01 5.46919644e-01 -7.97332346e-01 2.41471455e-01
7.19844922e-02 2.64319360e-01 1.32019415e-01 6.33908570e-01
7.18865037e-01 -4.23513114e-01 4.14073579e-02 -3.09305936e-01
-2.16984004e-02 -2.06273794e-01 -1.15830410e+00 1.51700270e+00
-2.22737566e-01 1.06762397e+00 1.28457159e-01 -9.30789471e-01
4.91464257e-01 5.19373894e-01 4.92383868e-01 -9.11835790e-01
2.41189286e-01 3.11728064e-02 -5.89581802e-02 -5.23661554e-01
5.36542773e-01 5.41766360e-02 2.27707371e-01 1.88501224e-01
4.00436550e-01 2.98725605e-01 4.66566622e-01 4.42370921e-01
1.08049595e+00 -2.12430954e-01 2.23412141e-01 -2.66117334e-01
3.14300597e-01 -1.30129784e-01 6.64532065e-01 5.73493600e-01
-2.35850230e-01 6.54825151e-01 1.11965962e-01 1.82053074e-03
-1.12224567e+00 -9.04785573e-01 -9.70140919e-02 1.00988388e+00
1.47772327e-01 -6.25155210e-01 -7.59839416e-01 -4.83578771e-01
-1.77539930e-01 4.67998266e-01 -4.47556078e-01 -3.27502817e-01
-5.90261817e-01 -6.02991879e-01 3.47292513e-01 6.06298387e-01
9.36305821e-01 -1.01201081e+00 -3.32857698e-01 -4.48946692e-02
-3.24214429e-01 -1.54868376e+00 -5.62010229e-01 1.07415803e-01
-8.60942781e-01 -7.83228517e-01 -9.63032961e-01 -9.04513240e-01
5.64817727e-01 7.42309153e-01 1.01258647e+00 1.04258649e-01
1.16346292e-01 3.19919914e-01 -7.28388608e-01 -1.37299150e-01
-6.47546709e-01 -1.66821420e-01 3.55456173e-01 3.55735384e-02
2.70444572e-01 -6.20082736e-01 -6.13625884e-01 6.33586049e-01
-1.03947854e+00 2.35059887e-01 7.98035622e-01 8.26281488e-01
6.06392682e-01 2.83025742e-01 4.51915801e-01 -6.54963017e-01
8.14385414e-02 -6.97087765e-01 -5.72011173e-01 6.27092971e-03
-3.70259732e-01 -2.07159653e-01 8.25183034e-01 -8.91784668e-01
-8.52039695e-01 -1.27570093e-01 -1.59240559e-01 -1.01602399e+00
-8.92897323e-02 2.04994246e-01 -1.18123189e-01 7.14276731e-02
7.05182672e-01 3.86641055e-01 1.50023788e-01 -3.31368357e-01
7.05655143e-02 8.01384687e-01 7.20476031e-01 -2.06389800e-01
8.85878801e-01 4.37865883e-01 -4.77595121e-01 -1.11088455e+00
-3.19875658e-01 -2.92386293e-01 -2.10660741e-01 -1.95857108e-01
9.60449815e-01 -1.56180429e+00 -5.77720702e-01 4.88278717e-01
-5.40403306e-01 -7.41058707e-01 -2.16612265e-01 7.85702825e-01
-3.76376510e-01 3.77205074e-01 -7.27200806e-01 -4.14161474e-01
-2.61925519e-01 -1.27745330e+00 5.14917731e-01 1.76899478e-01
5.69809489e-02 -5.27084947e-01 -4.09364015e-01 6.10985041e-01
4.43756402e-01 -9.59504247e-02 3.44858527e-01 -2.63979405e-01
-6.84815824e-01 2.51808822e-01 -2.55434304e-01 8.17433119e-01
1.62946492e-01 3.81152965e-02 -1.09746587e+00 -6.18420780e-01
1.27107248e-01 -4.30029035e-01 1.03640747e+00 4.19944197e-01
1.42322731e+00 -3.73234808e-01 4.08564806e-02 9.11321759e-01
1.37688744e+00 3.76299739e-01 9.75596786e-01 2.40939185e-01
7.48498976e-01 1.73703253e-01 7.02569723e-01 4.58950400e-01
1.60018608e-01 6.70411348e-01 2.94381469e-01 -1.07144609e-01
-3.90126765e-01 -1.40023455e-01 1.01222920e+00 1.19595098e+00
-1.89225584e-01 -4.15009320e-01 -6.48790061e-01 3.80549878e-01
-1.51944387e+00 -1.10051537e+00 -2.65639853e-02 2.07359481e+00
9.71545339e-01 1.90924972e-01 3.00235629e-01 3.28505129e-01
7.39573121e-01 2.86091566e-01 -3.80301744e-01 6.74706846e-02
-4.17777061e-01 1.78133518e-01 7.21014142e-01 2.98328936e-01
-1.32026994e+00 9.61863577e-01 5.54393053e+00 1.20113301e+00
-1.13176203e+00 3.18326503e-01 6.76246345e-01 -3.76260757e-01
9.28943418e-03 -3.44440520e-01 -8.08320403e-01 1.02686810e+00
1.28475821e+00 -7.17291087e-02 5.46635509e-01 7.93821037e-01
6.12887964e-02 -1.34138197e-01 -9.14807796e-01 1.46918118e+00
3.84937860e-02 -1.52866697e+00 5.55068813e-02 3.84179018e-02
7.93225586e-01 2.13459685e-01 2.28733569e-01 5.39318860e-01
4.86562029e-02 -1.07942224e+00 7.56163478e-01 -7.24695921e-02
1.33785212e+00 -5.58443308e-01 6.29817247e-01 1.78870752e-01
-1.10709584e+00 -2.54808962e-01 -5.53707898e-01 2.22469494e-01
2.54150741e-02 4.27294940e-01 -5.26513755e-01 3.63585442e-01
1.07162607e+00 7.59732544e-01 -5.25926888e-01 8.05243850e-01
-6.47407025e-02 1.04853988e+00 -3.24202746e-01 2.46078029e-01
9.96682644e-02 6.08264061e-04 4.44710046e-01 1.36793029e+00
2.73817390e-01 2.66743720e-01 -1.23617589e-01 2.69731492e-01
-4.38913852e-01 4.88551194e-03 -5.21115243e-01 2.07555946e-04
3.84677440e-01 9.40825999e-01 -4.64731097e-01 -5.43148160e-01
-6.98431253e-01 1.09015882e+00 7.07069859e-02 5.42456686e-01
-1.17142284e+00 -5.82139753e-02 7.54655957e-01 -5.83493267e-04
7.30730534e-01 -1.09559886e-01 2.68423259e-01 -1.47578585e+00
1.09883524e-01 -1.39057446e+00 5.26070356e-01 -8.12758327e-01
-8.68658364e-01 7.46386349e-01 -1.23576615e-02 -1.59261966e+00
-8.47467631e-02 -6.32556260e-01 -3.40067804e-01 3.10628563e-01
-1.47030294e+00 -4.39565390e-01 -6.12382293e-01 9.92163002e-01
8.36221337e-01 -4.31286395e-01 5.47966540e-01 6.93632424e-01
-7.58965492e-01 9.38858449e-01 2.47422561e-01 2.68917888e-01
6.92192852e-01 -8.96106005e-01 1.57859936e-01 1.10229599e+00
4.17538971e-01 2.55178720e-01 5.88661849e-01 -4.23933327e-01
-1.60091150e+00 -1.16102004e+00 1.99164867e-01 -3.53797793e-01
3.92580122e-01 -3.92149746e-01 -1.07163608e+00 6.12317562e-01
4.19218570e-01 1.03856646e-01 6.13012075e-01 -4.07140851e-01
-4.00187761e-01 -2.70894825e-01 -1.00527298e+00 6.86524391e-01
1.21978807e+00 -4.61516500e-01 -3.22473586e-01 2.28078827e-01
1.09837174e+00 -4.40846503e-01 -9.56184149e-01 4.67660040e-01
2.12552324e-01 -1.08849370e+00 9.70845878e-01 -3.63543838e-01
7.46811032e-01 -2.91827112e-01 -4.76904720e-01 -1.19065785e+00
-3.06414723e-01 -7.34800398e-01 -5.08972466e-01 1.07613206e+00
3.64897400e-01 -4.35725927e-01 9.56248522e-01 2.75154084e-01
-8.86892825e-02 -6.29967213e-01 -7.07798898e-01 -1.02588964e+00
-1.53596491e-01 -5.58857262e-01 3.28357697e-01 1.13416898e+00
-2.81674325e-01 1.59858853e-01 -9.22725081e-01 1.35392085e-01
4.59706962e-01 -2.64635086e-01 8.73653352e-01 -5.08134186e-01
-5.83513141e-01 -1.40561432e-01 -3.53596270e-01 -1.22680724e+00
-3.85329947e-02 -7.10415781e-01 -2.34348014e-01 -8.78109455e-01
2.12706864e-01 -1.12697482e-01 -3.81489456e-01 4.56269711e-01
-3.16145957e-01 4.41557974e-01 4.04371411e-01 2.79622197e-01
-5.98052919e-01 6.16083503e-01 1.20529437e+00 -2.58816630e-02
-1.29943430e-01 -1.69223532e-01 -7.45835543e-01 5.07148147e-01
1.02502024e+00 -4.07468438e-01 -6.10087931e-01 -5.30696094e-01
-3.06817383e-01 9.25274715e-02 1.19772233e-01 -1.16902208e+00
1.22727856e-01 1.81756448e-02 4.77039367e-01 -3.84041101e-01
4.74366158e-01 -8.09498668e-01 2.32969701e-01 5.18219113e-01
-3.71071398e-02 1.35693163e-01 3.56317282e-01 5.34986913e-01
-4.60938275e-01 -2.51909439e-02 8.76152456e-01 -4.11872156e-02
-1.33002114e+00 3.05756539e-01 -4.76192206e-01 2.10326925e-01
9.61010993e-01 -3.45760405e-01 -3.94701004e-01 -6.95330918e-01
-4.79330570e-01 2.44893566e-01 3.71720344e-01 4.57937121e-01
8.05335879e-01 -1.31294119e+00 -7.47856140e-01 2.95073390e-01
-4.35094684e-02 -8.90710726e-02 6.47677481e-01 7.21940041e-01
-4.69096422e-01 1.53543711e-01 -3.61227602e-01 -7.69060791e-01
-1.63068092e+00 6.90222979e-01 -6.22821860e-02 1.06220499e-01
-7.20300078e-01 1.05484200e+00 1.63847432e-01 1.61073983e-01
4.63377744e-01 -3.07750672e-01 -1.23304985e-01 3.46823335e-02
8.88079643e-01 2.85001755e-01 -6.97525963e-02 -7.27788627e-01
-2.82954514e-01 3.34750265e-01 -1.70699880e-01 6.36530370e-02
1.41252923e+00 7.07344562e-02 4.88992810e-01 2.61265226e-02
1.49687457e+00 5.85811734e-02 -1.64666450e+00 -3.84209633e-01
-4.19350415e-01 -7.13702977e-01 1.69711351e-01 -3.88260692e-01
-1.53089631e+00 6.94718540e-01 6.76810741e-01 -5.69824241e-02
1.42413545e+00 -1.67000052e-02 8.06754649e-01 3.32114607e-01
3.40012610e-01 -1.13108575e+00 3.97231102e-01 2.80288935e-01
8.69455040e-01 -1.57055962e+00 1.80820242e-01 -3.52601498e-01
-1.14826727e+00 6.15752041e-01 7.87933469e-01 1.00678213e-01
4.92378891e-01 3.06186795e-01 3.22479941e-02 3.07330117e-02
-9.90248680e-01 -1.90595776e-01 1.91938370e-01 4.74921137e-01
7.25950394e-03 -1.23100281e-01 1.52266756e-01 5.87473392e-01
-2.49823228e-01 -1.76234804e-02 5.91811419e-01 9.03959155e-01
-3.24419796e-01 -6.52651370e-01 -3.66955251e-01 3.99343312e-01
-6.24720514e-01 -2.79139221e-01 1.34935215e-01 1.00755823e+00
-4.10702787e-02 1.05107355e+00 1.00249365e-01 -8.95153105e-01
9.97143388e-02 -2.25938305e-01 3.14662635e-01 -3.76946449e-01
-4.27628398e-01 1.47482336e-01 5.17684892e-02 -6.98387384e-01
-5.87646961e-01 -4.08526987e-01 -9.83497322e-01 -6.69586480e-01
-2.88617224e-01 1.50224283e-01 4.42210138e-01 5.27717471e-01
4.31207240e-01 4.61353242e-01 7.77879536e-01 -8.06481063e-01
-2.97860444e-01 -8.10142577e-01 -4.60023671e-01 5.50237238e-01
3.46937567e-01 -6.05468869e-01 -6.78873837e-01 2.89850116e-01]
|
[9.43221664428711, 0.7815083861351013]
|
46c53f40-94fd-41da-8803-532a3362e4fe
|
sug-single-dataset-unified-generalization-for
|
2305.09160
| null |
https://arxiv.org/abs/2305.09160v1
|
https://arxiv.org/pdf/2305.09160v1.pdf
|
SUG: Single-dataset Unified Generalization for 3D Point Cloud Classification
|
Although Domain Generalization (DG) problem has been fast-growing in the 2D image tasks, its exploration on 3D point cloud data is still insufficient and challenged by more complex and uncertain cross-domain variances with uneven inter-class modality distribution. In this paper, different from previous 2D DG works, we focus on the 3D DG problem and propose a Single-dataset Unified Generalization (SUG) framework that only leverages a single source dataset to alleviate the unforeseen domain differences faced by a well-trained source model. Specifically, we first design a Multi-grained Sub-domain Alignment (MSA) method, which can constrain the learned representations to be domain-agnostic and discriminative, by performing a multi-grained feature alignment process between the splitted sub-domains from the single source dataset. Then, a Sample-level Domain-aware Attention (SDA) strategy is presented, which can selectively enhance easy-to-adapt samples from different sub-domains according to the sample-level inter-domain distance to avoid the negative transfer. Experiments demonstrate that our SUG can boost the generalization ability for unseen target domains, even outperforming the existing unsupervised domain adaptation methods that have to access extensive target domain data. Our code is available at https://github.com/SiyuanHuang95/SUG.
|
['Hongsheng Li', 'Yikang Li', 'Peng Gao', 'Botian Shi', 'Bo Zhang', 'Siyuan Huang']
|
2023-05-16
| null | null | null | null |
['3d-point-cloud-classification', 'point-cloud-classification', 'unsupervised-domain-adaptation']
|
['computer-vision', 'computer-vision', 'methodology']
|
[ 3.09424758e-01 -1.50153860e-01 -2.80733913e-01 -5.25123298e-01
-8.42243195e-01 -7.27069497e-01 4.74437982e-01 -1.86943665e-01
-1.63003698e-01 6.67590141e-01 7.87110701e-02 -6.96693063e-02
-2.02003032e-01 -6.08729541e-01 -5.87294221e-01 -8.28671813e-01
2.84608543e-01 5.93051910e-01 2.40764827e-01 -1.24550432e-01
1.04134299e-01 5.31264305e-01 -1.47966969e+00 1.17857456e-01
1.29038632e+00 1.14582956e+00 5.07026136e-01 -3.14920656e-02
-2.70620763e-01 7.03883693e-02 -4.80742723e-01 1.00760229e-01
3.89920563e-01 -3.24838728e-01 -5.34840584e-01 2.18713596e-01
4.32133198e-01 -1.65337503e-01 -1.95507258e-01 1.20342350e+00
6.99815810e-01 1.11962348e-01 8.88833225e-01 -1.35244191e+00
-1.08468306e+00 1.89272724e-02 -7.30386376e-01 2.73607552e-01
7.95904920e-02 1.80754155e-01 6.03929341e-01 -1.01963663e+00
6.22333050e-01 1.13783264e+00 3.08175445e-01 7.36419439e-01
-1.30386257e+00 -1.10337663e+00 6.45560086e-01 1.02536581e-01
-1.37150574e+00 -8.23015869e-02 1.24863827e+00 -5.91222346e-01
6.85097456e-01 -2.57655472e-01 3.12833905e-01 1.47928631e+00
-1.72967926e-01 8.65720093e-01 1.14338756e+00 -1.88102439e-01
3.48903924e-01 2.83902228e-01 -7.66634718e-02 5.37994057e-02
1.89554051e-01 1.03118636e-01 -4.08197105e-01 1.52880577e-02
8.45436096e-01 4.05532643e-02 -4.02175158e-01 -1.00346327e+00
-1.28153062e+00 7.66945243e-01 5.37098348e-01 2.95404911e-01
-3.71542722e-01 -7.30638802e-01 3.86647195e-01 4.20616865e-01
7.44069874e-01 4.10692036e-01 -8.31809938e-01 1.59744322e-01
-6.64124846e-01 2.68160611e-01 3.02412361e-01 1.33429980e+00
9.14108992e-01 -5.79874068e-02 -7.79849961e-02 1.00331140e+00
5.93377016e-02 6.65493786e-01 6.94642246e-01 -4.93062139e-01
6.81982160e-01 8.06771696e-01 -1.40220886e-02 -8.12277853e-01
-2.64624894e-01 -6.79318786e-01 -8.90091598e-01 2.70310819e-01
3.16569358e-01 -6.33349940e-02 -1.04916418e+00 2.02500105e+00
6.76997006e-01 1.55767187e-01 8.66827890e-02 1.09349871e+00
7.11827099e-01 4.04646367e-01 1.55914605e-01 6.81240484e-02
1.02724290e+00 -8.03054273e-01 -1.50922775e-01 -4.64627415e-01
4.40484047e-01 -4.33966756e-01 1.33018303e+00 2.36457735e-01
-6.07627690e-01 -8.28172863e-01 -1.10747254e+00 -2.54743285e-02
-4.96863216e-01 -1.19978925e-02 1.75165519e-01 3.80217195e-01
-5.87968588e-01 1.91085353e-01 -5.38441658e-01 -5.62616169e-01
8.43758225e-01 2.67366648e-01 -6.18722498e-01 -2.95537353e-01
-1.13832557e+00 8.33311617e-01 6.10510707e-01 -2.56021440e-01
-8.32917511e-01 -9.26501274e-01 -8.30001056e-01 -2.10255951e-01
3.59904766e-01 -6.99707508e-01 9.88782823e-01 -1.32750762e+00
-1.49053526e+00 1.14925194e+00 5.25887264e-03 1.57613829e-02
3.99297446e-01 -1.94898933e-01 -4.83775735e-01 6.15509488e-02
3.94863814e-01 7.79322624e-01 1.01967561e+00 -1.48580194e+00
-5.45877159e-01 -7.14280128e-01 -1.54318243e-01 5.26515901e-01
-5.21131039e-01 -3.87990326e-01 -3.39439601e-01 -9.60552156e-01
2.67719030e-01 -8.46477747e-01 -4.80543114e-02 1.06212921e-01
-1.12926990e-01 -1.82089359e-01 9.05550778e-01 -4.44309294e-01
9.33087528e-01 -2.40867710e+00 3.97980809e-01 8.35404098e-02
1.00545706e-02 3.74553263e-01 -3.64975750e-01 1.52280852e-01
-3.19010168e-01 -2.68155307e-01 -6.68682694e-01 -1.39991537e-01
1.75011232e-02 2.58961260e-01 -3.86651933e-01 3.90551388e-01
5.92527926e-01 6.17807209e-01 -9.26511109e-01 -3.15644145e-01
1.90722987e-01 3.06206882e-01 -5.05449355e-01 3.20545435e-01
-2.42604777e-01 1.01436353e+00 -7.09286988e-01 8.19132984e-01
1.22298503e+00 -3.26837301e-01 -4.32811193e-02 -7.49030039e-02
1.79473653e-01 9.55226496e-02 -1.10388625e+00 2.14798617e+00
-4.01005358e-01 1.18547335e-01 7.91953318e-03 -1.42939651e+00
1.35635257e+00 5.83899440e-03 4.03258532e-01 -8.26871753e-01
-3.15254554e-02 4.93089855e-01 -2.18435541e-01 -3.13660115e-01
1.55599564e-01 -4.24960256e-01 -2.93071151e-01 9.36816931e-02
3.03518623e-01 -1.57710776e-01 -3.22825938e-01 -6.56419992e-02
6.87673450e-01 3.68442684e-01 3.08576494e-01 -3.53380591e-01
7.01760352e-01 2.80285906e-02 7.16601670e-01 4.41766530e-01
-4.52456921e-01 8.84797394e-01 2.80584991e-01 -1.10816419e-01
-9.42026496e-01 -1.38277102e+00 -2.57182688e-01 8.93278122e-01
7.16328204e-01 2.39648461e-01 -4.32101935e-01 -1.14457917e+00
3.34736615e-01 7.10048318e-01 -5.82402170e-01 -4.63840812e-01
-3.78451288e-01 -4.77633595e-01 1.28253713e-01 6.80205822e-01
6.41414881e-01 -8.33494246e-01 -2.98763752e-01 3.66485864e-02
7.30665848e-02 -1.07650435e+00 -5.00891864e-01 3.16381574e-01
-9.93139625e-01 -8.07420433e-01 -1.07959461e+00 -9.49292183e-01
7.41744995e-01 3.55789155e-01 1.09632516e+00 -5.09439945e-01
9.24346223e-02 4.25982535e-01 -5.54993689e-01 -4.06123549e-01
-7.91075826e-02 2.81419486e-01 2.87733495e-01 3.16166244e-02
8.83896053e-01 -8.91359329e-01 -5.21527469e-01 4.74099934e-01
-9.09162998e-01 -4.32990827e-02 7.46410906e-01 1.05834651e+00
8.28318357e-01 -6.74674809e-02 1.02615774e+00 -6.74950182e-01
4.13970113e-01 -8.91867042e-01 -4.40086067e-01 7.11644888e-02
-5.73334754e-01 -7.40417913e-02 7.70197392e-01 -8.30869794e-01
-1.17254031e+00 7.01578110e-02 1.36658832e-01 -8.11185598e-01
-6.88136697e-01 2.30747357e-01 -8.23880076e-01 3.97448502e-02
6.95609570e-01 4.15726662e-01 1.97708800e-01 -6.47836924e-01
2.99420416e-01 7.57034838e-01 3.79296839e-01 -9.13222313e-01
1.05440819e+00 4.06282425e-01 -3.20765883e-01 -5.46449721e-01
-7.65502274e-01 -4.96648639e-01 -9.42887783e-01 1.09754233e-02
6.88385546e-01 -1.13594675e+00 6.92246482e-03 5.99225104e-01
-8.89214993e-01 -3.18107098e-01 -3.72444421e-01 5.44426024e-01
-4.53854293e-01 4.49898064e-01 -4.37061340e-02 -4.14022624e-01
-6.12439355e-03 -1.06556153e+00 1.20355320e+00 2.82205611e-01
2.54838681e-03 -8.84319365e-01 2.13630684e-02 1.64238572e-01
2.65402406e-01 1.93411380e-01 9.32941437e-01 -1.06351686e+00
-4.13982391e-01 -4.00469154e-02 -4.70193326e-01 7.39257455e-01
4.14159715e-01 -6.59044862e-01 -9.30221856e-01 -4.70129699e-01
9.65010747e-02 -3.98999572e-01 6.18260622e-01 2.33998194e-01
1.34874249e+00 3.53448913e-02 -4.58153188e-01 6.93396211e-01
1.30981541e+00 9.38271359e-02 2.26242572e-01 5.25359631e-01
6.13713622e-01 6.44265950e-01 8.87283862e-01 4.06358153e-01
3.38182509e-01 6.38248205e-01 4.45578814e-01 -1.09897189e-01
-7.78284073e-02 -4.64343876e-01 1.73612192e-01 4.40405965e-01
2.75796145e-01 -9.82465744e-02 -8.47822547e-01 7.98062444e-01
-1.65778375e+00 -5.95669329e-01 3.72350425e-01 2.20179892e+00
7.11678565e-01 1.06115945e-01 2.54474252e-01 -2.42711648e-01
8.67365479e-01 5.43323420e-02 -1.12345612e+00 6.33305758e-02
-2.76086688e-01 1.94741592e-01 4.31063086e-01 1.78671807e-01
-1.30047774e+00 8.45331967e-01 4.76554918e+00 1.08430016e+00
-1.29052603e+00 4.22904342e-02 3.62182379e-01 -1.26652926e-01
-4.06715155e-01 -2.61182517e-01 -6.44316137e-01 7.65230536e-01
3.92342955e-01 -2.04689980e-01 1.22221649e-01 1.18363523e+00
-1.58241421e-01 1.62549093e-01 -1.08813155e+00 1.00160921e+00
-4.31621037e-02 -9.04802740e-01 2.33021751e-01 1.40586346e-01
7.76393175e-01 1.06220096e-01 2.70417154e-01 5.21579623e-01
3.03022303e-02 -6.63585663e-01 5.48303664e-01 1.67825222e-01
9.12852466e-01 -7.39133537e-01 5.31301439e-01 6.12772465e-01
-1.01771796e+00 -2.36937404e-01 -4.48353738e-01 1.36509299e-01
1.36251850e-02 6.41723216e-01 -6.29779577e-01 8.87256324e-01
9.21354234e-01 9.00680780e-01 -4.23503816e-01 7.96974421e-01
-1.30906299e-01 2.75854468e-01 -2.99581587e-01 2.36454979e-01
1.31530374e-01 -2.59142101e-01 7.78037846e-01 9.07440543e-01
5.40734649e-01 1.02969222e-01 2.48497143e-01 7.54452288e-01
1.53094362e-02 2.51680966e-02 -7.98446596e-01 2.78884232e-01
7.14626431e-01 9.12237227e-01 -3.42264086e-01 -2.89744586e-01
-5.51066995e-01 1.25415754e+00 3.74768108e-01 5.53316712e-01
-7.12115288e-01 -3.66190672e-01 1.01777816e+00 9.36996341e-02
6.70038521e-01 -2.60235757e-01 -5.71426809e-01 -1.36701894e+00
3.24195057e-01 -8.67274106e-01 6.47439182e-01 -6.44890249e-01
-1.96929300e+00 5.73307514e-01 2.57785112e-01 -1.83063054e+00
5.41562699e-02 -7.81237960e-01 -3.49521488e-01 1.10308528e+00
-1.94508410e+00 -1.30910385e+00 -3.66963208e-01 8.86381924e-01
5.20826697e-01 -3.74177843e-01 6.31277502e-01 5.12363791e-01
-3.72471988e-01 7.94163287e-01 2.84456402e-01 -1.03748515e-01
9.80228186e-01 -1.04958713e+00 1.41801983e-01 7.28240132e-01
-3.38647544e-01 5.82391858e-01 4.32461768e-01 -5.61357796e-01
-1.02111912e+00 -1.14978218e+00 4.96988207e-01 -5.23025692e-01
5.01488745e-01 -4.76438642e-01 -1.34351945e+00 6.01530075e-01
-8.04291889e-02 1.46941394e-01 6.89924598e-01 -3.25713977e-02
-6.06017768e-01 -3.28829378e-01 -1.37850809e+00 3.57827485e-01
1.35605037e+00 -4.86747950e-01 -8.00739944e-01 1.11275777e-01
7.07516730e-01 -4.51987565e-01 -8.34847629e-01 7.14460671e-01
2.16362357e-01 -7.73601592e-01 1.07591665e+00 -7.05708027e-01
3.48700553e-01 -4.81801331e-01 -3.39541912e-01 -1.42033422e+00
-3.97838086e-01 2.69662719e-02 1.37803361e-01 1.25792015e+00
2.06625625e-01 -8.36158097e-01 8.33707452e-01 3.57307792e-01
-3.38772923e-01 -5.94275355e-01 -1.08492935e+00 -1.16773081e+00
5.91570020e-01 -2.54832774e-01 9.40073073e-01 1.21210277e+00
-1.84083790e-01 3.26527774e-01 -1.44074365e-01 5.27002990e-01
6.07538342e-01 4.77057874e-01 8.70908082e-01 -1.39907670e+00
-1.43385544e-01 -4.39418525e-01 -3.83020788e-01 -1.43893576e+00
3.00651312e-01 -1.05048418e+00 -3.13773809e-04 -1.18523014e+00
-5.16179986e-02 -5.73570251e-01 -5.66887856e-01 3.21876317e-01
-1.42916992e-01 1.18375383e-02 -1.50169814e-02 2.76716739e-01
-3.96166742e-01 1.04533947e+00 1.46353507e+00 -1.14389822e-01
-2.34387442e-01 -1.75889023e-02 -8.93029690e-01 3.90959471e-01
7.37376809e-01 -4.31913555e-01 -7.32239306e-01 -4.94394124e-01
-4.42214966e-01 -2.89243102e-01 4.61831361e-01 -9.84009922e-01
7.68528059e-02 -3.30328226e-01 5.81356287e-01 -6.38148010e-01
2.13297904e-01 -1.05806446e+00 -9.09748301e-02 3.91632803e-02
-1.23762377e-01 -2.74569571e-01 3.94655168e-01 7.74683893e-01
-3.87184054e-01 1.10244431e-01 9.26364005e-01 -8.98626894e-02
-1.04848993e+00 6.34618759e-01 1.82787910e-01 1.59614921e-01
1.20028174e+00 -4.69375849e-01 -1.90887287e-01 7.73342773e-02
-6.30718529e-01 3.92190188e-01 7.79423892e-01 6.61197245e-01
4.52749759e-01 -1.51263630e+00 -6.11384988e-01 5.19778252e-01
6.67410851e-01 4.85768110e-01 6.70521617e-01 5.27845740e-01
1.82959810e-01 1.33764014e-01 -5.77778816e-01 -9.36983109e-01
-7.81421065e-01 8.49160194e-01 2.37580732e-01 -3.03803757e-02
-6.45655811e-01 9.79797244e-01 7.14434087e-01 -9.62664723e-01
-1.01177050e-02 -9.76706296e-02 -4.01593670e-02 4.85749915e-02
2.71921545e-01 2.65222099e-02 7.16696307e-02 -5.34675837e-01
-6.32263899e-01 8.45733821e-01 -1.56956747e-01 2.64009863e-01
1.34564686e+00 -2.98834205e-01 3.16231042e-01 3.31421554e-01
1.29077458e+00 -2.79707611e-01 -1.69006217e+00 -5.72871387e-01
-6.58126101e-02 -6.93915784e-01 -2.47961715e-01 -9.16790724e-01
-9.23267186e-01 9.70203340e-01 7.44461834e-01 -6.29090965e-02
1.57720900e+00 3.24116379e-01 6.71321332e-01 -7.93110579e-03
3.19828212e-01 -1.12549341e+00 1.87469378e-01 4.38524097e-01
9.42574382e-01 -1.49130976e+00 -2.39540160e-01 -2.98783749e-01
-9.81509924e-01 7.94409871e-01 1.05492866e+00 -2.14480862e-01
6.60684645e-01 -1.70429677e-01 1.12176090e-01 -3.18365917e-03
-3.98556292e-01 -1.67329371e-01 3.42876226e-01 1.20592046e+00
3.47512774e-02 -2.47513670e-02 1.00403704e-01 9.59569812e-01
1.10497370e-01 5.03127128e-02 -2.93342844e-02 8.14948142e-01
-3.03984195e-01 -1.22580063e+00 -3.26354414e-01 2.53723353e-01
2.11908340e-01 1.73664644e-01 -1.95103884e-01 8.63933265e-01
2.68132448e-01 4.50574309e-01 7.87434578e-02 -3.87785912e-01
6.66900873e-01 1.61379620e-01 4.47476894e-01 -7.55928934e-01
-3.03407572e-02 -1.37924671e-01 -3.37898880e-01 -3.71220142e-01
-4.48373944e-01 -8.26664209e-01 -1.01952243e+00 1.09656975e-01
-5.51849082e-02 -1.22882150e-01 3.17448884e-01 7.84741759e-01
8.56577337e-01 2.58794963e-01 7.60228693e-01 -1.01253235e+00
-7.55384982e-01 -8.87125254e-01 -6.96680963e-01 6.20246053e-01
4.79939312e-01 -1.25952566e+00 -4.44411963e-01 -1.22283980e-01]
|
[10.343320846557617, 2.984407663345337]
|
141a765c-0211-4946-b6fe-9a23b0c199fe
|
directional-mean-curvature-for-textured-image
|
1611.08625
| null |
http://arxiv.org/abs/1611.08625v1
|
http://arxiv.org/pdf/1611.08625v1.pdf
|
Directional Mean Curvature for Textured Image Demixing
|
Approximation theory plays an important role in image processing, especially
image deconvolution and decomposition. For piecewise smooth images, there are
many methods that have been developed over the past thirty years. The goal of
this study is to devise similar and practical methodology for handling textured
images. This problem is motivated by forensic imaging, since fingerprints,
shoeprints and bullet ballistic evidence are textured images. In particular, it
is known that texture information is almost destroyed by a blur operator, such
as a blurred ballistic image captured from a low-cost microscope. The
contribution of this work is twofold: first, we propose a mathematical model
for textured image deconvolution and decomposition into four meaningful
components, using a high-order partial differential equation approach based on
the directional mean curvature. Second, we uncover a link between functional
analysis and multiscale sampling theory, e.g., harmonic analysis and filter
banks. Both theoretical results and examples with natural images are provided
to illustrate the performance of the proposed model.
|
['David Banks', 'Duy Hoang Thai']
|
2016-11-25
| null | null | null | null |
['image-deconvolution']
|
['computer-vision']
|
[ 4.06590998e-01 -2.48983502e-01 4.35578972e-01 -5.23063801e-02
-5.91812171e-02 -2.25855589e-01 1.78627729e-01 -2.81206012e-01
-2.22208813e-01 7.40436018e-01 -6.29991069e-02 7.24825487e-02
-2.43321285e-01 -4.44423378e-01 -4.15972739e-01 -9.51943099e-01
2.12309197e-01 -1.53003022e-01 2.62942553e-01 -7.56157190e-02
5.99647403e-01 6.58529878e-01 -1.49157703e+00 8.49286243e-02
8.58196914e-01 1.05475783e+00 4.08285916e-01 3.78527880e-01
2.78024506e-02 6.04618609e-01 -3.26132625e-01 -5.83482862e-01
1.15418419e-01 -5.67228615e-01 -7.88618684e-01 5.38558900e-01
-9.04024243e-02 -4.20233041e-01 -2.70322382e-01 1.29035819e+00
3.34517986e-01 1.68999404e-01 8.37798715e-01 -5.96682012e-01
-9.76765931e-01 1.16350101e-02 -9.30969834e-01 2.46810034e-01
1.86136470e-03 1.04667842e-01 2.35654235e-01 -1.05262184e+00
4.40040916e-01 1.20334125e+00 6.64263725e-01 3.43967497e-01
-1.41953897e+00 -3.98568772e-02 -6.10648215e-01 3.55345577e-01
-1.28708565e+00 -2.37258971e-01 1.14664519e+00 -5.72505713e-01
8.42572376e-02 2.39979401e-01 5.44737339e-01 6.93935871e-01
7.12051451e-01 5.68044484e-01 1.68542755e+00 -5.64637959e-01
-6.98412163e-03 3.97505641e-01 4.05573100e-01 6.90601707e-01
5.29869676e-01 -1.00626349e-01 -2.64126688e-01 -2.34066337e-01
1.04020071e+00 5.50198443e-02 -6.69697106e-01 -1.55533895e-01
-8.65794063e-01 4.61726367e-01 -1.74466565e-01 6.11515880e-01
-4.01943594e-01 -1.36232898e-01 1.14910640e-01 8.56428966e-02
7.01806962e-01 1.98161572e-01 4.65471685e-01 -6.04606085e-02
-8.50808382e-01 1.66654482e-01 6.02778733e-01 4.18765157e-01
6.40823662e-01 -8.58232677e-02 9.29197744e-02 8.99426520e-01
1.57939419e-01 5.50204277e-01 4.59121794e-01 -1.05232477e+00
-2.00105235e-01 8.74015838e-02 3.21510285e-01 -1.27316940e+00
-4.21921834e-02 -1.43916547e-01 -8.67333055e-01 3.76898289e-01
6.48977637e-01 3.02858084e-01 -3.01518410e-01 1.25533772e+00
2.99238265e-01 2.30748489e-01 -1.66467130e-01 1.04120409e+00
4.31037337e-01 4.02152151e-01 -4.91795480e-01 -5.42698145e-01
1.58756769e+00 -3.66557986e-01 -1.13790667e+00 4.86528873e-01
-2.28471920e-01 -1.09177589e+00 9.72926497e-01 7.32369661e-01
-1.23009181e+00 -4.74541664e-01 -7.51720548e-01 -4.78121154e-02
2.31425911e-02 1.52462006e-01 3.53513420e-01 6.74987733e-01
-7.47280896e-01 8.49966407e-01 -7.10936725e-01 -5.45888245e-01
1.81061715e-01 -7.53678977e-02 -2.51664847e-01 -8.99410322e-02
-8.16503584e-01 9.49854851e-01 -1.25572115e-01 4.69939202e-01
-4.46352065e-01 -3.52745891e-01 -4.20716047e-01 -7.05599934e-02
-3.95151637e-02 -4.56014603e-01 8.57870162e-01 -8.35784316e-01
-1.59035540e+00 9.87656653e-01 -4.23011065e-01 -2.88041592e-01
4.92860585e-01 -4.77313250e-02 -2.86357433e-01 6.46943033e-01
-3.03621665e-02 -3.62951547e-01 1.15863419e+00 -1.35272813e+00
1.23395957e-02 -6.31243706e-01 -3.12490612e-01 4.77214053e-04
-3.42714310e-01 8.11458230e-02 -7.73593858e-02 -7.07166910e-01
2.07982406e-01 -5.71114302e-01 6.40657842e-02 5.94861060e-02
-4.22502667e-01 9.49852020e-02 7.01433420e-01 -8.68390739e-01
1.14179730e+00 -2.31352425e+00 1.38539657e-01 4.20139991e-02
3.29908788e-01 8.42017382e-02 3.08607340e-01 5.61046779e-01
4.96331863e-02 -1.58674225e-01 -5.32766342e-01 -2.16744229e-01
-3.40320200e-01 4.89264168e-03 -4.63142097e-01 9.12662208e-01
3.94389369e-02 5.46729743e-01 -4.41100359e-01 -5.67341089e-01
1.34143457e-01 6.05393291e-01 -1.13024831e-01 7.01125190e-02
4.40276504e-01 7.10056424e-01 -4.79245603e-01 6.74015701e-01
1.10222220e+00 -1.07879318e-01 -8.67912546e-02 -6.44353688e-01
-6.16702557e-01 -5.91441214e-01 -9.95983541e-01 1.09097993e+00
-1.01262689e-01 6.64640367e-01 5.25791824e-01 -1.28645682e+00
7.66695917e-01 4.45203185e-01 4.61020023e-01 -3.93682688e-01
2.54221171e-01 5.67766786e-01 -3.04698855e-01 -1.06157124e+00
5.75157642e-01 -6.61292553e-01 4.28901643e-01 3.42791647e-01
-3.61031532e-01 -7.20385239e-02 4.21816632e-02 -2.03991294e-01
6.85441375e-01 1.01642914e-01 1.69091150e-01 -5.77113569e-01
8.07414412e-01 -8.37456882e-02 2.43506774e-01 4.87764120e-01
-2.23762348e-01 7.26630390e-01 5.17202675e-01 -5.73249221e-01
-1.11230397e+00 -7.04073906e-01 -8.11436892e-01 4.45522368e-02
6.54441595e-01 3.42535973e-01 -1.21605837e+00 1.95124149e-01
5.63149899e-02 4.01999623e-01 -6.30239546e-01 -5.04889451e-02
-4.44583356e-01 -9.03536260e-01 4.65873599e-01 -5.35500944e-02
7.88968980e-01 -8.83235395e-01 -6.71885908e-01 3.42045724e-01
-4.12506908e-01 -1.01057565e+00 -3.09134126e-01 -4.15779769e-01
-1.04886842e+00 -1.27902019e+00 -1.24013543e+00 -5.69140851e-01
5.97294927e-01 6.67295516e-01 6.25642598e-01 -6.90804273e-02
-5.43624580e-01 7.60281980e-01 -1.41559437e-01 -5.26979938e-02
-3.05466235e-01 -6.92881584e-01 1.70264423e-01 8.77713203e-01
2.71983922e-01 -6.06865406e-01 -7.43918359e-01 5.32966018e-01
-1.17307591e+00 -3.67660001e-02 6.43021762e-01 9.51125324e-01
6.67849660e-01 3.83066654e-01 3.38829845e-01 -4.88565713e-01
8.77060711e-01 -2.31463596e-01 -5.15820324e-01 8.03774744e-02
-4.02735084e-01 -4.46972363e-02 6.35746300e-01 -3.70516658e-01
-1.43878138e+00 -3.81757051e-01 1.43427983e-01 -3.62453133e-01
-1.97479546e-01 3.39011848e-01 1.04070731e-01 -4.21972334e-01
5.68908572e-01 8.39001656e-01 5.30709267e-01 -9.26812649e-01
-1.76908746e-01 8.36217642e-01 8.16633403e-01 -6.66714251e-01
6.40643537e-01 1.12318301e+00 2.35767081e-01 -1.60836744e+00
-5.00581920e-01 -5.04782796e-01 -3.11259180e-01 -4.79236811e-01
9.18352723e-01 -1.92967132e-01 -1.16485775e+00 9.05860126e-01
-1.24220693e+00 -1.76936202e-02 -1.19511552e-01 6.80505395e-01
-6.45780563e-01 1.05952537e+00 -7.76010871e-01 -1.21815622e+00
-2.20520735e-01 -1.08290362e+00 7.53387511e-01 3.38889986e-01
2.57727683e-01 -9.40714896e-01 -1.90393962e-02 5.52617669e-01
4.53260571e-01 4.41375643e-01 7.76651263e-01 1.80652872e-01
-5.25390685e-01 -9.07195434e-02 -2.74960846e-01 5.91476262e-01
1.54554695e-01 -1.71684042e-01 -9.91633534e-01 -1.72139898e-01
1.10203421e+00 1.15079246e-01 6.69981778e-01 7.14209557e-01
1.16264141e+00 -7.73150846e-02 -1.61242932e-01 5.27911186e-01
1.53887677e+00 1.47205561e-01 9.31324661e-01 1.83284819e-01
3.70318741e-01 9.78751481e-01 4.46195841e-01 2.68312305e-01
-1.06591180e-01 7.11803854e-01 3.04133177e-01 -7.63983577e-02
-1.94771111e-01 2.59611100e-01 9.79770645e-02 6.64791822e-01
-6.53942347e-01 6.77491650e-02 -4.47969168e-01 4.50204045e-01
-1.60622215e+00 -1.14704442e+00 -6.32224083e-01 2.42505860e+00
5.33312619e-01 -2.25496545e-01 -7.93541372e-02 2.98123330e-01
9.30873632e-01 -3.20254356e-01 -3.29223871e-01 -1.37497574e-01
-2.60596693e-01 5.46887778e-02 3.45980376e-01 5.85973501e-01
-8.00819039e-01 3.53167027e-01 5.98874283e+00 9.69591498e-01
-1.25175285e+00 1.41804859e-01 5.89081466e-01 3.71016085e-01
-3.25094968e-01 -2.35433336e-02 -2.82223880e-01 8.57263803e-01
4.24274087e-01 -1.08781466e-02 3.26170743e-01 3.98462415e-01
7.44583786e-01 -5.24787605e-01 -4.40965861e-01 1.16745281e+00
9.66344625e-02 -9.67969596e-01 -1.57051206e-01 4.26355779e-01
3.29486042e-01 -7.24671721e-01 2.80862033e-01 -5.93650937e-01
-6.35310411e-01 -8.32527757e-01 5.14688253e-01 9.82601821e-01
7.18707323e-01 -6.66143000e-01 8.30896795e-01 2.21901909e-01
-8.61830771e-01 1.02937005e-01 -6.28236651e-01 -1.09717764e-01
2.70847976e-01 1.03903794e+00 -2.05692619e-01 6.19136572e-01
5.50652623e-01 6.16429508e-01 -1.36730835e-01 1.25619471e+00
1.94195777e-01 6.17166936e-01 -1.73575375e-02 1.12783752e-01
-5.48471287e-02 -9.05871630e-01 7.95426250e-01 9.69798982e-01
5.01919806e-01 4.52477157e-01 -5.40501177e-01 1.17580307e+00
4.68895882e-01 1.04157045e-01 -6.01487398e-01 8.46966952e-02
-5.57328612e-02 1.35874403e+00 -9.15583849e-01 -1.82763055e-01
-4.80282456e-01 1.12594604e+00 -3.94568831e-01 6.91051900e-01
-7.00651109e-01 -5.93393385e-01 4.07291114e-01 4.57085758e-01
1.05075479e-01 -1.40388861e-01 -4.63854045e-01 -1.37915218e+00
-1.99570460e-03 -5.92403531e-01 -2.11556956e-01 -7.29538977e-01
-1.17702401e+00 3.61779362e-01 1.49900299e-02 -1.28902245e+00
4.46598262e-01 -6.80418432e-01 -5.65873921e-01 1.18014050e+00
-1.44371009e+00 -7.87539184e-01 -3.04811984e-01 6.55161023e-01
3.56246442e-01 1.01190746e-01 4.09961641e-01 3.90220106e-01
-4.99630362e-01 1.85988266e-02 5.46927810e-01 -1.55699894e-01
5.01206696e-01 -1.05523312e+00 -2.18141854e-01 8.71354163e-01
-6.12688720e-01 8.26301336e-01 1.09949720e+00 -4.59955513e-01
-1.50748146e+00 -3.46304595e-01 7.49138534e-01 -3.82278487e-02
5.45530260e-01 7.03616887e-02 -1.19665575e+00 9.99191627e-02
1.37498826e-01 -2.74468046e-02 4.19110835e-01 -7.28702128e-01
2.63570577e-01 -1.44390419e-01 -1.44577134e+00 4.12921876e-01
4.91093963e-01 -4.32366371e-01 -6.30199254e-01 9.90489200e-02
-1.13312006e-01 -3.21270563e-02 -8.53148043e-01 5.87511808e-02
8.76026750e-01 -1.23044038e+00 8.44435751e-01 1.54787702e-02
3.84303153e-01 -4.23636973e-01 5.39589971e-02 -9.91163552e-01
-2.55458951e-01 -8.15973997e-01 9.93091911e-02 9.44529951e-01
-3.06441754e-01 -8.92385244e-01 5.22994220e-01 3.70966077e-01
-1.14538953e-01 -6.78170860e-01 -8.70074749e-01 -8.64352703e-01
-2.12463468e-01 8.03003833e-02 -5.84092513e-02 7.45216787e-01
1.84955239e-01 -7.50901029e-02 -5.97255647e-01 -1.99921839e-02
1.35249949e+00 1.51292816e-01 3.75639588e-01 -1.29698527e+00
-4.18441325e-01 -3.37103158e-01 -4.46663648e-01 -1.00969744e+00
-2.33855635e-01 -4.48579282e-01 -4.98986654e-02 -1.17997992e+00
3.73071551e-01 -4.13588919e-02 -3.92683148e-02 -2.26318285e-01
-3.10403984e-02 4.01788145e-01 -1.74202815e-01 7.09213614e-01
1.45091131e-01 4.44990546e-01 1.62148786e+00 1.36947542e-01
1.14969850e-01 1.23306029e-01 -5.71387649e-01 8.06087673e-01
5.28061211e-01 -2.33990639e-01 -3.37928861e-01 -1.35401934e-01
-7.25089386e-02 2.45357499e-01 6.34727836e-01 -8.84715974e-01
7.60717466e-02 -6.15256913e-02 1.44962743e-01 -2.01206148e-01
4.25169617e-01 -7.39786029e-01 3.59707743e-01 5.15398562e-01
4.84601781e-02 -2.51449347e-01 -1.21289551e-01 7.38944352e-01
-4.29896146e-01 -6.20007277e-01 1.17782998e+00 -2.94629991e-01
-2.60856092e-01 -1.81293890e-01 -6.54500186e-01 -2.96089470e-01
8.33961427e-01 -6.19222581e-01 -2.85741091e-01 -3.28086048e-01
-6.21565878e-01 -5.27048767e-01 5.53942382e-01 -4.36590314e-01
8.57525706e-01 -1.18205750e+00 -5.68677545e-01 1.20304696e-01
-3.73098046e-01 -4.13067073e-01 6.38132572e-01 1.58913791e+00
-9.07823861e-01 2.25252032e-01 -3.13451320e-01 -5.34295082e-01
-1.19737017e+00 4.72596288e-01 3.88754159e-01 6.96505755e-02
-7.74645090e-01 4.56527710e-01 4.76998925e-01 2.40342215e-01
-2.09023103e-01 -2.11897761e-01 -3.15653741e-01 -1.71331793e-01
7.80650258e-01 7.66133249e-01 -1.84306398e-01 -8.09956491e-01
-6.32437393e-02 1.19364810e+00 1.69705659e-01 -1.93678349e-01
1.26165509e+00 -5.07064342e-01 -6.61904693e-01 4.63001221e-01
1.14205050e+00 1.53445661e-01 -1.21452570e+00 -1.27494335e-01
-9.81568992e-02 -5.87378204e-01 4.07983586e-02 -1.24161959e-01
-7.34966040e-01 9.91577446e-01 5.82005322e-01 7.38022864e-01
1.18103576e+00 -2.56791502e-01 8.82197976e-01 8.01231340e-02
3.01406771e-01 -1.12698472e+00 1.39773056e-01 6.89647673e-03
9.47978139e-01 -8.05583239e-01 -1.33136407e-01 -5.46874166e-01
-3.39017719e-01 1.44651473e+00 -9.53347608e-02 -1.91873729e-01
5.14222383e-01 -4.18213084e-02 -1.69919848e-01 -2.60555774e-01
-1.13023661e-01 -9.30465609e-02 1.39618710e-01 4.28724676e-01
4.89229947e-01 -1.69542491e-01 -9.84732985e-01 4.81128454e-01
3.27151626e-01 3.16242844e-01 9.37420011e-01 7.32160211e-01
-6.86251640e-01 -6.90428495e-01 -9.93415594e-01 7.99741000e-02
-9.25025225e-01 1.18729360e-01 3.50393099e-03 4.96105224e-01
-5.10731302e-02 8.93392146e-01 -2.23229975e-01 -1.21358025e-04
2.64461309e-01 1.28030181e-02 7.33960986e-01 -8.19676816e-02
5.28271357e-03 3.12633395e-01 -3.74240994e-01 -3.91530663e-01
-5.55198610e-01 -6.60072088e-01 -8.88393402e-01 -3.20380181e-01
-2.89517313e-01 4.09534663e-01 7.74302185e-01 7.50812531e-01
-3.85645591e-02 2.92007357e-01 6.13549590e-01 -9.60770547e-01
-4.95694339e-01 -9.75441754e-01 -1.38135493e+00 4.59495693e-01
3.78981709e-01 -8.43964398e-01 -6.91996932e-01 2.69690543e-01]
|
[11.6507568359375, -2.6204919815063477]
|
efeb6739-3365-4d9c-9ac9-ab3c0965a518
|
a-dynamic-window-neural-network-for-ccg
|
1610.02749
| null |
http://arxiv.org/abs/1610.02749v1
|
http://arxiv.org/pdf/1610.02749v1.pdf
|
A Dynamic Window Neural Network for CCG Supertagging
|
Combinatory Category Grammar (CCG) supertagging is a task to assign lexical
categories to each word in a sentence. Almost all previous methods use fixed
context window sizes as input features. However, it is obvious that different
tags usually rely on different context window sizes. These motivate us to build
a supertagger with a dynamic window approach, which can be treated as an
attention mechanism on the local contexts. Applying dropout on the dynamic
filters can be seen as drop on words directly, which is superior to the regular
dropout on word embeddings. We use this approach to demonstrate the
state-of-the-art CCG supertagging performance on the standard test set.
|
['Cheng-qing Zong', 'Jiajun Zhang', 'Huijia Wu']
|
2016-10-10
| null | null | null | null |
['ccg-supertagging']
|
['natural-language-processing']
|
[-8.74482933e-03 2.61727095e-01 -1.19560063e-01 -6.86978042e-01
-7.09051609e-01 -5.87376654e-01 7.57596850e-01 1.31923109e-01
-7.80115306e-01 4.22849149e-01 4.56086457e-01 -6.17297292e-01
4.11281466e-01 -8.11702967e-01 -3.79346907e-01 -7.62420118e-01
-8.34765732e-02 3.16395372e-01 7.00981617e-01 -3.52178335e-01
1.99037999e-01 -1.87901467e-01 -1.25935650e+00 3.62213939e-01
8.01822782e-01 6.97194457e-01 5.58784127e-01 4.58201259e-01
-4.60049659e-01 2.77759671e-01 -6.42213702e-01 -5.89000463e-01
-3.82332131e-02 -2.30729923e-01 -6.46601856e-01 -2.65358686e-01
6.66333020e-01 1.46172583e-01 -2.12766588e-01 1.22078490e+00
3.91805828e-01 3.00473034e-01 3.36097777e-01 -6.64113879e-01
-1.08879280e+00 1.35429192e+00 -7.85397440e-02 4.06182289e-01
5.69278337e-02 -1.63682789e-01 1.68388951e+00 -1.12045228e+00
4.33291137e-01 1.52024150e+00 4.93136913e-01 7.07770050e-01
-1.15402758e+00 -6.08490109e-01 9.47869241e-01 4.09026295e-02
-1.18911123e+00 1.13678120e-01 3.50502640e-01 -1.73250839e-01
1.20502436e+00 2.21805736e-01 6.42173231e-01 1.18234456e+00
4.93713051e-01 8.62878382e-01 9.19314623e-01 -5.95725000e-01
2.19679102e-01 -1.16145037e-01 9.86960709e-01 6.24811828e-01
2.14187339e-01 -7.50725269e-02 -5.31871855e-01 -1.20865010e-01
1.56677857e-01 4.52373996e-02 -1.38407558e-01 7.38606155e-02
-9.67559814e-01 1.35381746e+00 4.39676195e-01 3.33685935e-01
1.62156865e-01 6.07489169e-01 4.53929573e-01 4.65884894e-01
9.85558450e-01 5.05686343e-01 -7.27756023e-01 -1.18293017e-01
-3.60156953e-01 1.46763653e-01 6.79550707e-01 9.66409147e-01
6.85815036e-01 5.70987090e-02 -5.07390261e-01 8.37433517e-01
4.97127086e-01 3.88757974e-01 7.03754246e-01 -2.46781856e-01
4.39678431e-01 4.33899403e-01 -4.09785956e-01 -4.16114181e-01
-3.70290101e-01 -5.60281873e-01 -2.24627122e-01 -6.50276542e-02
4.77314502e-01 -2.66606152e-01 -1.48307168e+00 1.92799199e+00
1.49501234e-01 4.58136439e-01 -3.50375533e-01 7.61331439e-01
9.48524952e-01 5.32745004e-01 4.78641510e-01 1.47765219e-01
1.71839273e+00 -7.36538172e-01 -8.43898356e-01 -6.75437152e-01
1.10731030e+00 -6.15829349e-01 1.75964499e+00 2.57142872e-01
-5.80991030e-01 -3.56954515e-01 -1.03199434e+00 -2.37047672e-01
-8.46216679e-01 -4.31426525e-01 1.00608909e+00 7.84655511e-01
-1.00027990e+00 6.73984766e-01 -1.12339735e+00 -4.18392271e-01
1.49655342e-01 5.59976935e-01 -1.33625969e-01 -4.60217670e-02
-1.63711143e+00 9.24780607e-01 5.25399148e-01 5.21453060e-02
-6.31796956e-01 -7.18902886e-01 -9.57847118e-01 2.51817685e-02
3.49950999e-01 -5.23456991e-01 1.15721393e+00 -5.61879337e-01
-1.30733943e+00 1.00839734e+00 -1.54169917e-01 -4.77698654e-01
-1.60974160e-01 -4.65732306e-01 -3.62954676e-01 -4.11527693e-01
-6.32135719e-02 3.33802938e-01 6.29667401e-01 -5.79528272e-01
-1.00754964e+00 -2.02883184e-01 1.71573699e-01 1.12216927e-01
-4.92067099e-01 3.85800302e-01 -5.23890436e-01 -8.85115027e-01
2.16145009e-01 -8.96471143e-01 -2.03536853e-01 -7.74502456e-01
-4.87460405e-01 -9.08469260e-01 6.96411908e-01 -4.24191982e-01
1.69266760e+00 -2.24654245e+00 1.28937408e-01 6.70651421e-02
1.03919208e-01 6.51459694e-02 -3.05149287e-01 2.23116234e-01
2.06264779e-02 4.16268051e-01 8.35537165e-02 -5.25569141e-01
2.16411978e-01 3.84649843e-01 -4.28943545e-01 2.94349551e-01
1.18110210e-01 9.83535290e-01 -1.19743681e+00 -1.71393171e-01
-1.41134098e-01 2.21885994e-01 -7.68729329e-01 -7.59680420e-02
-3.16452146e-01 -3.01300794e-01 -4.51656312e-01 2.40765348e-01
4.87818480e-01 1.30131453e-01 2.53657103e-01 2.23093748e-01
-7.62469172e-02 9.85822797e-01 -7.05517590e-01 1.65660048e+00
-5.44761062e-01 3.74657512e-01 -4.49741602e-01 -7.40578592e-01
7.64585257e-01 2.15331718e-01 -2.66795486e-01 -2.36021549e-01
2.71784455e-01 3.07898790e-01 2.83241391e-01 -4.62691933e-02
6.45919681e-01 -3.15800488e-01 -6.35682046e-01 3.80343467e-01
4.07899410e-01 3.11830752e-02 1.92675591e-01 2.92312354e-01
1.23780572e+00 -1.62739754e-01 -1.44908682e-01 -6.71118498e-01
-7.44799301e-02 -4.70411181e-01 7.53446460e-01 9.75860059e-01
1.89270452e-01 6.70335948e-01 6.74216747e-01 -5.24262607e-01
-6.43749356e-01 -1.00075328e+00 -5.00924923e-02 1.83644009e+00
-8.26438442e-02 -7.56066918e-01 -7.15234697e-01 -1.11166584e+00
3.24700437e-02 8.67819309e-01 -9.47336972e-01 -4.16787863e-01
-7.75396466e-01 -1.04388356e+00 1.72420055e-01 7.40719497e-01
7.27809817e-02 -1.19422960e+00 -1.12576388e-01 4.43624914e-01
1.53599650e-01 -1.04595459e+00 -1.06689656e+00 8.44862878e-01
-7.08657384e-01 -8.05247605e-01 -1.44847780e-01 -1.07736230e+00
3.37192029e-01 2.03010589e-01 1.38577068e+00 3.74369234e-01
1.73083901e-01 -1.54327706e-01 -8.06154251e-01 -3.16655815e-01
-2.37438709e-01 7.63901055e-01 2.26699258e-03 3.51106413e-02
8.42593849e-01 -4.05165166e-01 -4.37523782e-01 -4.22830656e-02
-7.74271667e-01 -2.06923887e-01 2.95115709e-01 1.08370376e+00
4.46128845e-01 -5.13758957e-01 4.27455723e-01 -1.55901730e+00
7.50362277e-01 -2.85153687e-01 -6.10802472e-01 2.09302455e-01
-4.71482903e-01 1.43788040e-01 5.42372406e-01 -8.00386667e-01
-6.90242767e-01 -2.63586253e-01 -2.69869089e-01 5.20134158e-02
1.58322945e-01 7.95919776e-01 -4.78115976e-01 1.98057994e-01
2.30846271e-01 -7.52535313e-02 -4.29476678e-01 -6.31683886e-01
5.84501743e-01 6.34921134e-01 1.45782113e-01 -4.07997578e-01
6.24212444e-01 6.30973354e-02 -4.88996446e-01 -6.68400645e-01
-1.39207590e+00 -4.64436918e-01 -4.99352902e-01 2.75826126e-01
9.82609272e-01 -7.05120206e-01 -2.58177489e-01 3.48451227e-01
-1.01952195e+00 -8.04445267e-01 -1.41211033e-01 4.73762840e-01
1.27210125e-01 8.35992694e-02 -7.15514064e-01 -4.24516290e-01
-2.06773221e-01 -6.78568304e-01 1.07895279e+00 4.43811230e-02
-2.32947186e-01 -1.49882066e+00 -4.59425896e-02 -2.90739357e-01
4.13933873e-01 -1.58951983e-01 9.47159171e-01 -1.01256120e+00
-3.11424643e-01 -2.95368075e-01 8.90158191e-02 4.21322048e-01
1.51187584e-01 -1.46534130e-01 -9.98094082e-01 -5.36496043e-01
-1.81589723e-01 -3.84059846e-02 1.36783314e+00 3.29781681e-01
9.86311734e-01 -1.33059502e-01 -4.38533634e-01 7.37746298e-01
1.30007899e+00 -2.91311052e-02 6.30534530e-01 2.21915945e-01
9.97215629e-01 1.35670975e-01 7.08564937e-01 4.35107611e-02
2.61675119e-01 7.82450497e-01 2.74002403e-01 1.14692368e-01
-2.05321878e-01 -3.63148630e-01 6.33532822e-01 1.12519658e+00
3.61159086e-01 -5.75683475e-01 -8.39404404e-01 5.60963511e-01
-1.68625009e+00 -7.00966358e-01 -2.57437646e-01 2.11417937e+00
9.21154439e-01 6.17668271e-01 -1.71155766e-01 -2.25673318e-01
8.36220801e-01 3.27645481e-01 -2.03498248e-02 -6.26142383e-01
-1.11098990e-01 5.50598562e-01 6.48508549e-01 8.36208463e-01
-1.42160320e+00 1.67247951e+00 6.52016068e+00 1.20175660e+00
-1.11626208e+00 5.48419476e-01 3.47699881e-01 -1.53210446e-01
-6.01588666e-01 2.14114085e-01 -1.43983698e+00 7.67227829e-01
9.70587790e-01 3.13041806e-02 7.94602558e-02 6.84462965e-01
5.96570820e-02 4.13083546e-02 -1.02832770e+00 5.22741437e-01
-6.57628700e-02 -9.95758295e-01 8.11449364e-02 5.62923104e-02
4.36055005e-01 4.81829375e-01 -1.21336803e-02 6.41842186e-01
8.69324803e-01 -6.67204261e-01 7.54049361e-01 4.61613573e-02
6.79185688e-01 -6.16228640e-01 9.27109301e-01 -1.16017669e-01
-1.24524283e+00 7.09034875e-02 -6.64737165e-01 -1.61231712e-01
1.57195851e-01 6.09171808e-01 -6.64597332e-01 1.13482304e-01
4.66077745e-01 4.24985111e-01 -8.45434785e-01 9.39760268e-01
-7.19277084e-01 1.32451046e+00 -4.26351309e-01 -5.82585096e-01
4.97841895e-01 -6.68727979e-02 3.37651849e-01 1.47557235e+00
1.45255506e-01 -9.28170308e-02 4.98999208e-01 3.85308355e-01
-1.83370680e-01 8.66436865e-03 -4.45280373e-01 1.56555906e-01
5.61249673e-01 1.17802262e+00 -8.92829418e-01 -3.80376399e-01
-8.09271336e-01 7.06152380e-01 5.62082827e-01 2.72933245e-01
-7.74879992e-01 -7.03563929e-01 9.23229337e-01 1.21759780e-01
6.00351810e-01 -2.50978410e-01 -3.29911977e-01 -1.17667019e+00
-2.64292389e-01 -3.79422635e-01 7.78621078e-01 -2.71797121e-01
-1.32974291e+00 7.19982684e-01 -6.64681103e-03 -8.97315264e-01
1.05729580e-01 -8.22717786e-01 -7.80452490e-01 8.58882248e-01
-1.44256139e+00 -1.19109857e+00 4.16087024e-02 2.59215713e-01
5.88930428e-01 -8.20564777e-02 8.88804793e-01 1.17205396e-01
-4.58232254e-01 1.00823081e+00 -3.13640945e-02 3.31785530e-01
5.95297098e-01 -1.77387726e+00 9.01675344e-01 1.10886669e+00
3.60605747e-01 8.32353532e-01 7.03475475e-01 -7.05934644e-01
-1.01641786e+00 -1.40968072e+00 1.48110735e+00 -7.29959428e-01
1.05167413e+00 -1.21161044e+00 -1.20880413e+00 8.49975228e-01
3.10288072e-01 2.20879540e-01 6.04994357e-01 7.50503123e-01
-3.68194073e-01 -2.68677026e-02 -5.47760189e-01 7.05269158e-01
1.28266370e+00 -4.92345691e-01 -1.03341746e+00 3.71264756e-01
1.35189235e+00 -2.31187329e-01 -3.90103966e-01 -8.97012055e-02
1.35850593e-01 -2.56170958e-01 4.77936804e-01 -8.87217045e-01
-8.87667835e-02 -2.04163730e-01 -7.67032430e-02 -1.72090697e+00
-6.06158018e-01 -8.41593802e-01 2.16452792e-01 1.38544941e+00
7.45513141e-01 -1.01222384e+00 7.39518464e-01 5.46135902e-01
-7.06388950e-01 -5.99509835e-01 -1.03295004e+00 -1.05003202e+00
5.88472903e-01 -6.57008111e-01 6.41173363e-01 7.82186210e-01
2.88694978e-01 6.47683024e-01 -8.36810246e-02 5.33329882e-02
2.49345943e-01 -1.01646282e-01 1.48364782e-01 -1.23062420e+00
-1.62360325e-01 -3.04286629e-01 -3.87933284e-01 -1.05908322e+00
3.61110479e-01 -1.23904252e+00 3.21477503e-01 -1.21790314e+00
6.58748522e-02 -4.72877383e-01 -6.51151896e-01 7.54885375e-01
-8.14113855e-01 4.41729277e-01 1.11313909e-01 -3.91337007e-01
-7.15704858e-01 5.04610300e-01 9.11071002e-01 -1.04487464e-01
1.11739095e-02 -1.55232295e-01 -7.93366969e-01 6.37784541e-01
8.38263750e-01 -7.81025827e-01 -2.59184599e-01 -6.96523666e-01
5.14119625e-01 -7.07534790e-01 -7.60810673e-02 -5.45678854e-01
9.02772397e-02 -4.76686703e-03 -4.11742985e-01 -3.60669523e-01
6.01495802e-02 -2.37805709e-01 -3.62721711e-01 3.63627821e-01
-3.02442729e-01 9.55929160e-02 1.93219155e-03 3.91804427e-01
-2.04176933e-01 -3.91154289e-01 5.59499860e-01 -2.15638667e-01
-8.93737853e-01 3.57773334e-01 -4.64477152e-01 4.75034475e-01
4.60834444e-01 -1.46728903e-01 -3.04064125e-01 1.41455531e-01
-8.60856116e-01 3.79657030e-01 2.18484297e-01 6.75849140e-01
6.16804540e-01 -1.50290430e+00 -7.15756357e-01 3.28695625e-01
2.38902912e-01 -1.79293975e-01 -2.18917459e-01 5.42336166e-01
-1.17769785e-01 3.32198381e-01 4.66762394e-01 -3.07326257e-01
-1.26913416e+00 4.30392891e-01 2.40693212e-01 -3.39412749e-01
-7.65342414e-01 1.45378244e+00 5.22898257e-01 -3.37907284e-01
2.74097651e-01 -8.57497096e-01 -2.29123905e-01 2.18770504e-01
5.20836890e-01 -3.54977608e-01 2.35981941e-01 -8.71380791e-02
-4.66939479e-01 5.42385757e-01 -3.16823483e-01 -4.69818823e-02
1.32637119e+00 -6.98339194e-02 -6.06816709e-02 6.92451954e-01
1.16761756e+00 -2.60154568e-02 -9.90673959e-01 -3.14430624e-01
5.22701025e-01 -3.93325984e-01 3.19753081e-01 -5.19058526e-01
-1.07882106e+00 8.82295132e-01 4.26953822e-01 6.78709269e-01
6.41426682e-01 3.15063387e-01 8.13147306e-01 1.65388420e-01
5.31459749e-01 -1.24761093e+00 -1.43833160e-01 1.14123535e+00
7.05062747e-01 -9.76914108e-01 -4.29010034e-01 -6.70060098e-01
-4.73733276e-01 9.31699038e-01 5.04124701e-01 -5.33146262e-01
1.03615010e+00 3.81229132e-01 2.61459798e-01 -1.48688287e-01
-1.10018897e+00 -5.97965539e-01 2.31502756e-01 3.20096791e-01
8.13932598e-01 1.44862771e-01 -7.20397830e-01 7.01607049e-01
-3.68188292e-01 -5.04046500e-01 5.97829342e-01 7.25874841e-01
-7.34319508e-01 -1.46298027e+00 1.80481523e-02 6.38280988e-01
-8.42440486e-01 -7.30632067e-01 -2.05442175e-01 8.11026096e-01
1.75430074e-01 9.32032764e-01 2.25189820e-01 -4.85772461e-01
4.38337147e-01 4.15554225e-01 2.76395112e-01 -1.52318788e+00
-9.12749469e-01 1.61967427e-01 3.18315685e-01 -4.33687896e-01
1.25165328e-01 -7.19157755e-01 -1.16948521e+00 -1.38023496e-01
-7.56966531e-01 4.63340342e-01 1.77267686e-01 9.25964713e-01
1.05567798e-02 6.28654599e-01 3.44004869e-01 -4.61838841e-01
-5.95041513e-01 -1.40735805e+00 -9.27435100e-01 6.05176866e-01
6.80652112e-02 -7.34554768e-01 -5.96016943e-01 -2.35131085e-01]
|
[10.539226531982422, 9.240255355834961]
|
f8873b85-d78d-44bd-a84a-9506a1002c94
|
beyond-rule-based-named-entity-recognition
|
2305.03960
| null |
https://arxiv.org/abs/2305.03960v1
|
https://arxiv.org/pdf/2305.03960v1.pdf
|
Beyond Rule-based Named Entity Recognition and Relation Extraction for Process Model Generation from Natural Language Text
|
Automated generation of business process models from natural language text is an emerging methodology for avoiding the manual creation of formal business process models. For this purpose, process entities like actors, activities, objects etc., and relations among them are extracted from textual process descriptions. A high-quality annotated corpus of textual process descriptions (PET) has been published accompanied with a basic process extraction approach. In its current state, however, PET lacks information about whether two mentions refer to the same or different process entities, which corresponds to the crucial decision of whether to create one or two modeling elements in the target model. Consequently, it is ambiguous whether, for instance, two mentions of data processing mean processing of different, or the same data. In this paper, we extend the PET dataset by clustering mentions of process entities and by proposing a new baseline technique for process extraction equipped with an additional entity resolution component. In a second step, we replace the rule-based relation extraction component with a machine learning-based alternative, enabling rapid adaption to different datasets and domains. In addition, we evaluate a deep learning-approach built for solving entity and relation extraction as well as entity resolution in a holistic manner. Finally, our extensive evaluation of the original PET baseline against our own implementation shows that a pure machine learning-based process extraction technique is competitive, while avoiding the massive overhead arising from feature engineering and rule definition needed to adapt to other datasets, different entity and relation types, or new domains.
|
['Stefan Jablonski', 'Lars Ackermann', 'Julian Neuberger']
|
2023-05-06
| null | null | null | null |
['feature-engineering', 'relation-extraction', 'entity-resolution']
|
['methodology', 'natural-language-processing', 'natural-language-processing']
|
[ 3.99069101e-01 6.77597463e-01 7.71114305e-02 -1.15970396e-01
-4.59612101e-01 -7.17003107e-01 1.29925466e+00 7.90408134e-01
-4.10245329e-01 6.97045624e-01 1.87639058e-01 -3.80557150e-01
-3.55646372e-01 -1.12185931e+00 -4.75552738e-01 -2.50860751e-01
1.88560858e-01 1.03843224e+00 3.62808228e-01 2.58948445e-01
6.47256449e-02 6.47587776e-01 -1.49643850e+00 2.31281474e-01
5.77337801e-01 1.07190800e+00 -7.92217702e-02 2.71811545e-01
-8.75837326e-01 1.14053380e+00 -5.10588586e-01 -7.06961870e-01
3.27687055e-01 -3.47616166e-01 -1.04003108e+00 2.74470687e-01
-6.37421384e-02 2.56514307e-02 4.85014953e-02 6.01615310e-01
5.03254905e-02 -1.08065508e-01 8.37870002e-01 -1.34535646e+00
-1.62969276e-01 9.65469897e-01 -3.27784419e-01 -1.01060256e-01
3.67665201e-01 1.81438178e-01 1.03364313e+00 -6.38383448e-01
9.74716544e-01 9.81498659e-01 5.66510618e-01 5.42637587e-01
-1.39556992e+00 -2.88415402e-01 7.62556195e-02 3.67788933e-02
-1.23924637e+00 -4.61075187e-01 5.51024258e-01 -6.16585553e-01
1.02930665e+00 1.07059330e-01 6.99048996e-01 1.13822925e+00
-4.54388186e-02 6.55237794e-01 9.56360817e-01 -5.24436057e-01
5.35285413e-01 3.64922315e-01 2.06290066e-01 4.63948041e-01
7.77762175e-01 -4.50604677e-01 -3.50429446e-01 -2.19477132e-01
6.28617525e-01 -1.62392870e-01 3.64617482e-02 -5.32546937e-01
-1.12434983e+00 5.35126448e-01 -2.65125185e-01 7.48960435e-01
-6.80219650e-01 -2.59828963e-03 5.17324090e-01 1.73205167e-01
1.55322492e-01 8.01135421e-01 -7.67022192e-01 -3.56364399e-01
-9.63612974e-01 3.67201835e-01 1.60548067e+00 1.15368533e+00
8.10704887e-01 -2.67875195e-01 -2.86846846e-01 1.65805548e-01
2.67375648e-01 -9.68969613e-02 4.63737249e-01 -7.96828926e-01
5.18401980e-01 1.24267614e+00 3.25212777e-01 -8.73350203e-01
-3.41680586e-01 -2.60385901e-01 -6.66319966e-01 -1.55219421e-01
4.86468703e-01 -1.91281304e-01 -7.11563945e-01 1.34135926e+00
3.09055477e-01 8.82059429e-03 2.30951682e-01 2.99239844e-01
8.45281005e-01 3.39932352e-01 4.71509635e-01 -3.98163289e-01
1.70638692e+00 -7.45961607e-01 -9.24265027e-01 -2.60876685e-01
4.01291668e-01 -3.08062643e-01 4.37236845e-01 4.45493698e-01
-1.03470552e+00 -2.46419445e-01 -7.34233260e-01 1.05749285e-02
-7.19785631e-01 3.19631817e-03 9.15023863e-01 5.09042740e-01
-6.93867624e-01 8.07116985e-01 -6.73286498e-01 -4.93185133e-01
4.41502780e-01 3.75437468e-01 -5.57440341e-01 3.12903047e-01
-1.09832823e+00 9.57812607e-01 7.82720029e-01 -1.56727955e-01
-5.93358219e-01 -5.96660256e-01 -9.55924988e-01 4.71611559e-01
1.04111624e+00 -7.24761665e-01 1.37318051e+00 -9.52760220e-01
-1.41029704e+00 7.10414886e-01 -1.78631023e-01 -8.96032274e-01
7.45164514e-01 -8.43399763e-02 -5.66476524e-01 -5.07631674e-02
8.74491110e-02 2.35109761e-01 7.82909691e-01 -1.35154688e+00
-1.04554892e+00 -2.81245828e-01 1.55458197e-01 -5.93394004e-02
-4.13992777e-02 1.46997347e-01 -4.84128863e-01 -1.56688556e-01
6.40581846e-02 -4.59706634e-01 -2.30287313e-01 -4.49606270e-01
-5.32023013e-01 -3.91331941e-01 5.60847223e-01 -6.08535707e-01
1.25638604e+00 -1.78926969e+00 2.17393890e-01 4.21010286e-01
6.52934551e-01 1.55384466e-01 1.96306571e-01 6.78392410e-01
-6.19098023e-02 5.98723710e-01 -2.63887137e-01 -3.93156052e-01
4.56342638e-01 2.16185927e-01 -1.40061766e-01 -1.04985133e-01
6.41515791e-01 6.79539204e-01 -7.43792117e-01 -7.51416624e-01
1.63250998e-01 2.67197907e-01 -3.71561162e-02 5.43605201e-02
-4.64063644e-01 4.52603996e-01 -3.95167738e-01 5.04999816e-01
3.59377205e-01 -2.13235945e-01 6.48346782e-01 -1.92347527e-01
-2.20940813e-01 4.63152945e-01 -1.70156908e+00 1.34953237e+00
-5.92005372e-01 2.76177198e-01 -6.60729036e-02 -7.21213341e-01
9.87865210e-01 6.97181880e-01 8.84781241e-01 -4.47436392e-01
1.73888370e-01 2.43662640e-01 2.53978502e-02 -3.68042797e-01
4.86328214e-01 -7.54281878e-02 -1.94308385e-01 5.32482803e-01
3.46855402e-01 -1.29581735e-01 9.13177192e-01 -4.95788790e-02
1.47161508e+00 4.54652309e-01 9.14134860e-01 -4.54910956e-02
9.02595699e-01 1.66709460e-02 7.47583985e-01 5.01342058e-01
3.65839107e-03 2.44854853e-01 1.05618954e+00 -5.27977347e-01
-9.59372759e-01 -8.40004086e-01 1.96795180e-01 5.31781435e-01
-2.54968613e-01 -9.60567772e-01 -6.34826779e-01 -1.11389530e+00
-1.00747660e-01 7.75611937e-01 -4.67491835e-01 2.25867733e-01
-6.07233047e-01 -4.50265288e-01 5.46768844e-01 3.98374915e-01
4.24147993e-01 -9.95081007e-01 -7.87824392e-01 5.41568100e-01
-2.92653944e-02 -1.48355162e+00 3.12275529e-01 4.51677322e-01
-5.42485058e-01 -1.31642187e+00 2.16499972e-03 -3.55905473e-01
2.62784541e-01 -7.13635862e-01 1.29544735e+00 -3.94505203e-01
2.19537273e-01 1.19839139e-01 -3.13380688e-01 -7.05743849e-01
-7.72070229e-01 4.77168173e-01 -3.27566862e-01 1.72446027e-01
6.96189523e-01 -4.52257752e-01 2.95588630e-03 -6.08501881e-02
-9.10185039e-01 1.14380218e-01 8.08160722e-01 2.39222497e-01
3.74300510e-01 4.03123081e-01 4.27970916e-01 -1.02795815e+00
8.61976504e-01 -3.18016559e-01 -5.41307569e-01 3.60643238e-01
-8.23342979e-01 4.15855765e-01 6.51216269e-01 -3.09852481e-01
-1.28463638e+00 2.41399050e-01 1.04532778e-01 1.92558486e-03
-7.23978579e-01 6.98727310e-01 -7.09969163e-01 7.14275241e-01
4.62562948e-01 9.37120318e-02 -2.50692517e-01 -4.67547059e-01
3.11190724e-01 2.83772469e-01 1.08056501e-01 -6.86309695e-01
9.80217695e-01 3.88409317e-01 3.29593986e-01 -4.06653613e-01
-3.19135487e-01 -4.97225225e-01 -9.27337527e-01 9.33431759e-02
8.83900225e-01 -5.73900163e-01 -6.54423118e-01 2.06601113e-01
-1.43481469e+00 -2.13738322e-01 -8.81983519e-01 2.13553861e-01
-5.97326994e-01 2.72957951e-01 -7.02107012e-01 -7.34456182e-01
-4.48298901e-01 -9.68638897e-01 9.54254031e-01 1.25520319e-01
-6.12176418e-01 -8.59393418e-01 3.59579585e-02 1.08507462e-01
2.05510482e-01 2.45597452e-01 9.37848926e-01 -1.53676331e+00
-5.75742185e-01 -3.62450302e-01 -2.63360739e-01 1.54038727e-01
4.73115981e-01 9.03478116e-02 -6.20752931e-01 3.44133347e-01
-8.59236624e-03 4.09617513e-01 1.91152662e-01 -2.92783022e-01
6.12518311e-01 -4.44907099e-01 -4.30684268e-01 1.32326871e-01
1.44669163e+00 3.77501458e-01 7.26716399e-01 6.92382276e-01
7.64874339e-01 9.19514656e-01 4.01405931e-01 4.22634095e-01
2.37244710e-01 6.00573838e-01 -7.65398815e-02 2.02085212e-01
2.58107297e-02 -1.08851552e-01 1.00960098e-01 2.80506164e-01
-3.97720575e-01 -1.11128040e-01 -1.28208995e+00 4.35655236e-01
-1.88331842e+00 -1.19088161e+00 -4.14240569e-01 1.94796944e+00
9.44687665e-01 2.83179462e-01 7.00046942e-02 4.20552552e-01
5.09129822e-01 -1.19735137e-01 -8.76445174e-02 -3.58539641e-01
1.31775960e-01 4.46839869e-01 4.41714555e-01 1.91676289e-01
-1.30896628e+00 8.65479469e-01 4.74822712e+00 6.28236055e-01
-8.10897052e-01 1.36773929e-01 1.60989881e-01 2.73573101e-01
-1.46396115e-01 2.66469359e-01 -1.03348672e+00 2.70314723e-01
1.06927359e+00 -3.13649148e-01 2.36896187e-01 7.25123227e-01
4.78182852e-01 6.34789094e-02 -1.61663115e+00 7.56805539e-01
-8.81694183e-02 -1.27592564e+00 2.43453726e-01 3.25218111e-01
4.35015023e-01 -3.86527002e-01 -8.19803536e-01 4.34611052e-01
4.52882826e-01 -8.97913694e-01 1.01212502e+00 7.32280970e-01
1.84871703e-01 -4.93160993e-01 1.00065637e+00 3.07573646e-01
-1.45430958e+00 -2.67555088e-01 3.12500507e-01 -3.09350956e-02
1.77306280e-01 6.05977833e-01 -1.15902627e+00 1.17131269e+00
4.11187887e-01 4.71357256e-01 -6.63052857e-01 6.92544401e-01
-4.11650211e-01 3.30379456e-01 -1.95889428e-01 2.09452078e-01
1.40886247e-01 -2.68153906e-01 4.95803624e-01 1.41272557e+00
1.90337896e-01 -4.21203732e-01 9.10193548e-02 1.08081686e+00
-7.57845193e-02 3.34853858e-01 -6.00071788e-01 -2.38916263e-01
2.83575982e-01 1.46488607e+00 -9.59708989e-01 -5.33793032e-01
-4.62125629e-01 6.76781058e-01 1.68547139e-01 3.40808451e-01
-6.83988273e-01 -1.69380113e-01 3.70601654e-01 4.25974309e-01
3.85722548e-01 -1.63199335e-01 -4.00498271e-01 -1.10321414e+00
1.71674445e-01 -9.49091852e-01 2.95095772e-01 -4.43022966e-01
-1.20530105e+00 5.94651282e-01 1.73191234e-01 -9.84696805e-01
-6.35259986e-01 -3.85939330e-01 -3.24113339e-01 8.46315503e-01
-1.26795650e+00 -1.36764967e+00 -1.53912336e-01 2.66492546e-01
4.50703621e-01 -1.32394983e-02 7.54710317e-01 2.40452543e-01
-4.94718611e-01 -1.67662323e-01 -4.80589956e-01 3.23273838e-01
5.62910855e-01 -1.58720934e+00 3.67898047e-01 9.78080988e-01
2.10445061e-01 6.88684642e-01 6.66838229e-01 -8.68523896e-01
-1.22509313e+00 -1.18304658e+00 1.39560878e+00 -5.76532304e-01
9.39351320e-01 -4.17909473e-01 -9.06215250e-01 9.83942628e-01
3.86654675e-01 -3.38210911e-01 5.12770295e-01 4.85553928e-02
-1.43091708e-01 -2.88298994e-01 -1.11668599e+00 6.47669613e-01
1.00923014e+00 -3.49919885e-01 -9.97936130e-01 -1.50459260e-01
4.82305974e-01 -1.10576767e-03 -1.22924578e+00 4.24817234e-01
3.17619592e-01 -6.72989964e-01 6.44981802e-01 -5.68166018e-01
4.88282084e-01 -4.92129952e-01 6.73307478e-02 -7.79143095e-01
-1.07666448e-01 -6.58765554e-01 -7.21277356e-01 1.82723069e+00
5.69948137e-01 -5.03664851e-01 6.08864188e-01 9.39595044e-01
1.06886737e-01 -5.51496327e-01 -7.25109637e-01 -6.66858912e-01
-2.65927047e-01 -4.64458227e-01 9.77724016e-01 1.03930521e+00
-2.14172457e-03 7.35584795e-01 1.50825813e-01 -8.24875943e-03
2.97598183e-01 2.63088904e-02 9.67066109e-01 -1.90225291e+00
-3.52610379e-01 -7.60935128e-01 -3.00941348e-01 -2.68829346e-01
1.04913317e-01 -6.99223518e-01 -1.02217346e-01 -2.30698967e+00
-5.07705435e-02 -3.89640212e-01 -2.84305308e-02 5.59030950e-01
2.51798272e-01 -4.55042869e-01 3.16411167e-01 3.04594129e-01
-4.24973816e-01 2.11973801e-01 9.35074091e-01 -2.51003832e-01
-4.53820735e-01 -9.67990665e-04 -6.02592170e-01 9.28134501e-01
8.68722856e-01 -5.10014117e-01 -3.51503044e-01 4.11865860e-02
4.02605355e-01 -1.71049505e-01 2.34900177e-01 -1.04060626e+00
3.18476140e-01 -1.10300355e-01 6.87184334e-02 -1.48931623e-01
1.32292628e-01 -1.29129159e+00 6.24524236e-01 2.86996216e-01
-2.65138149e-01 -1.02278829e-01 3.40036191e-02 5.10772705e-01
-2.33523026e-01 -5.53370297e-01 2.37673774e-01 -4.51881886e-01
-8.00917208e-01 6.18245676e-02 -7.26935923e-01 -2.57220030e-01
1.14771140e+00 -2.64807135e-01 -9.43666548e-02 -8.92710313e-03
-9.76032615e-01 -2.45123893e-01 3.82805496e-01 2.99303263e-01
3.27503346e-02 -7.03184664e-01 -5.36281466e-01 -7.82314911e-02
1.57777220e-01 3.59622449e-01 -3.25004160e-01 9.08122778e-01
-3.62596065e-01 2.11052835e-01 -4.21188548e-02 -1.32456645e-01
-1.03939915e+00 7.76747227e-01 2.66204774e-01 -1.06762469e+00
-4.53680336e-01 -1.21211685e-01 -2.28969276e-01 -3.27420741e-01
-4.93194424e-02 -7.31490493e-01 -6.17378592e-01 4.59806293e-01
3.15739483e-01 1.83475509e-01 3.75504047e-01 -4.54944551e-01
-3.18734676e-01 5.94242886e-02 1.84153304e-01 -3.70607257e-01
1.24785161e+00 1.60350911e-02 -3.60375404e-01 6.09832346e-01
3.23050886e-01 1.69799566e-01 -9.06308770e-01 -1.32373378e-01
9.43297327e-01 3.32703665e-02 -3.92076343e-01 -8.89085948e-01
-7.61669934e-01 5.00168502e-01 2.76666246e-02 6.73204482e-01
8.26657057e-01 2.57082850e-01 3.16502690e-01 5.36003709e-01
3.68471205e-01 -1.07188118e+00 -4.00222987e-01 5.41961670e-01
6.92356229e-01 -8.18373203e-01 2.75507383e-02 -7.58082151e-01
-5.00286222e-01 1.06884563e+00 5.63766718e-01 2.90074885e-01
4.34818447e-01 7.05771685e-01 -3.85528475e-01 -3.84337962e-01
-6.77363873e-01 -4.85922307e-01 5.60773797e-02 5.91862917e-01
3.34107935e-01 -2.52376869e-03 -6.55275524e-01 8.49841297e-01
4.46008667e-02 3.18419009e-01 4.93250847e-01 1.22439754e+00
-5.87260686e-02 -1.46502912e+00 -1.21958107e-01 5.76562941e-01
-6.54055119e-01 -1.23867944e-01 -6.81632698e-01 1.14533234e+00
5.48465490e-01 1.03602457e+00 2.56963462e-01 -1.22056477e-01
6.17896497e-01 6.07608795e-01 3.84208202e-01 -8.71824145e-01
-9.86848652e-01 1.40955229e-03 6.76727474e-01 -2.74120182e-01
-6.27767146e-01 -1.01175487e+00 -1.27919722e+00 -8.35906491e-02
-1.73383310e-01 1.98932022e-01 6.34676695e-01 1.34138465e+00
3.72971773e-01 7.88339555e-01 -9.87182185e-02 -4.54085648e-01
-2.01068476e-01 -1.03424788e+00 -3.09271127e-01 5.95291615e-01
-3.00154477e-01 -4.96671796e-01 4.60988347e-04 5.18749893e-01]
|
[9.249055862426758, 8.677992820739746]
|
d6479e11-de20-46c4-a473-11f088232531
|
190600910
|
1906.00910
| null |
https://arxiv.org/abs/1906.00910v2
|
https://arxiv.org/pdf/1906.00910v2.pdf
|
Learning Representations by Maximizing Mutual Information Across Views
|
We propose an approach to self-supervised representation learning based on maximizing mutual information between features extracted from multiple views of a shared context. For example, one could produce multiple views of a local spatio-temporal context by observing it from different locations (e.g., camera positions within a scene), and via different modalities (e.g., tactile, auditory, or visual). Or, an ImageNet image could provide a context from which one produces multiple views by repeatedly applying data augmentation. Maximizing mutual information between features extracted from these views requires capturing information about high-level factors whose influence spans multiple views -- e.g., presence of certain objects or occurrence of certain events. Following our proposed approach, we develop a model which learns image representations that significantly outperform prior methods on the tasks we consider. Most notably, using self-supervised learning, our model learns representations which achieve 68.1% accuracy on ImageNet using standard linear evaluation. This beats prior results by over 12% and concurrent results by 7%. When we extend our model to use mixture-based representations, segmentation behaviour emerges as a natural side-effect. Our code is available online: https://github.com/Philip-Bachman/amdim-public.
|
['R. Devon Hjelm', 'William Buchwalter', 'Philip Bachman']
|
2019-06-03
|
learning-representations-by-maximizing-mutual
|
http://papers.nips.cc/paper/9686-learning-representations-by-maximizing-mutual-information-across-views
|
http://papers.nips.cc/paper/9686-learning-representations-by-maximizing-mutual-information-across-views.pdf
|
neurips-2019-12
|
['self-supervised-image-classification']
|
['computer-vision']
|
[ 5.40708959e-01 1.60975218e-01 -2.01739937e-01 -4.95528281e-01
-9.55488801e-01 -7.15469480e-01 8.56508732e-01 5.58392704e-02
-3.82342726e-01 5.44549704e-01 4.21194613e-01 5.74341081e-02
1.53125077e-01 -7.23723173e-01 -1.13228202e+00 -5.16022980e-01
1.00494316e-03 1.68507732e-02 -5.52306622e-02 4.01759088e-01
1.84464350e-01 2.83933759e-01 -1.73233247e+00 5.92946291e-01
2.77878106e-01 7.83521533e-01 6.29181802e-01 8.58322561e-01
1.65964216e-01 8.34041417e-01 -3.61190379e-01 9.64915305e-02
1.23828866e-01 -3.36177528e-01 -7.34737635e-01 6.39778018e-01
7.36122489e-01 -3.42549562e-01 -4.14597362e-01 8.13171029e-01
1.23859107e-01 1.15396664e-01 7.10339606e-01 -1.13456333e+00
-6.11740887e-01 4.07092154e-01 -6.71458721e-01 2.69908816e-01
4.99972045e-01 1.76480144e-01 1.03144288e+00 -8.51212382e-01
7.67561913e-01 1.06408286e+00 1.04544774e-01 4.83749479e-01
-1.56836784e+00 -4.64809358e-01 4.40623730e-01 1.98371083e-01
-1.09076416e+00 -5.41237712e-01 6.87388301e-01 -4.54758108e-01
7.42441416e-01 2.32765764e-01 3.58180732e-01 1.37931645e+00
1.87404573e-01 1.09219193e+00 1.25596929e+00 -3.25318873e-01
2.59331405e-01 2.22176746e-01 -1.96039557e-01 5.46187878e-01
-2.67883465e-02 1.88832462e-01 -7.71199703e-01 -2.48555452e-01
1.04649639e+00 3.59451294e-01 -1.04231136e-02 -4.56923723e-01
-1.41185582e+00 5.80521047e-01 3.49470168e-01 3.38845730e-01
-4.10490215e-01 3.77346933e-01 7.14421570e-02 1.73005462e-01
3.64014506e-01 3.03756237e-01 -5.17492533e-01 2.98917666e-02
-8.10534954e-01 6.61744922e-02 4.80424464e-01 9.10459161e-01
1.11001921e+00 -1.35008484e-01 1.66038901e-01 6.68813109e-01
3.98197114e-01 4.78458613e-01 3.90782565e-01 -1.34802663e+00
3.44442248e-01 4.82621461e-01 6.85949251e-02 -8.34521234e-01
-1.88216060e-01 -1.66098341e-01 -7.14737773e-01 2.48686403e-01
3.50515127e-01 -2.33379111e-01 -1.02894163e+00 2.08566260e+00
1.87467203e-01 5.28058529e-01 5.56047671e-02 8.15309048e-01
7.76694715e-01 6.64907336e-01 1.83854729e-01 -2.43267119e-01
1.32788336e+00 -5.96758425e-01 -5.12089968e-01 -4.74714726e-01
2.86786079e-01 -7.22184598e-01 7.18096316e-01 4.25656497e-01
-1.04448724e+00 -6.79748714e-01 -8.64544988e-01 2.08520800e-01
-3.01867664e-01 1.81973323e-01 5.44196367e-01 1.65423930e-01
-1.04431283e+00 3.87660980e-01 -9.40233588e-01 -5.41848242e-01
3.03366482e-01 2.44667977e-01 -6.99414313e-01 -2.28248224e-01
-7.68576860e-01 7.14296699e-01 7.14911222e-02 -4.79629606e-01
-1.20426691e+00 -4.70552862e-01 -9.70371008e-01 -1.79590881e-01
4.56088871e-01 -6.59135938e-01 1.13636303e+00 -1.11411858e+00
-1.10577512e+00 9.91166294e-01 -5.01659930e-01 -3.24741244e-01
1.96495622e-01 -5.31832993e-01 -2.94692874e-01 3.33009273e-01
3.11527729e-01 9.18738306e-01 8.74901295e-01 -1.71713281e+00
-4.97363836e-01 -4.94620919e-01 3.03917021e-01 5.04110098e-01
1.18053518e-02 2.42999289e-03 -4.60242569e-01 -4.40148562e-01
2.49561787e-01 -1.02779150e+00 -4.88574177e-01 7.01508820e-02
-3.87047589e-01 5.99862337e-02 8.15569699e-01 -5.46451926e-01
6.91365182e-01 -2.22985029e+00 2.62041092e-01 1.11155912e-01
1.97512388e-01 -1.66717827e-01 -9.04005468e-02 5.60560346e-01
-1.82532012e-01 1.52085409e-01 -4.09845203e-01 -4.95544195e-01
-3.91687274e-01 3.67581725e-01 -2.33702093e-01 4.37310249e-01
3.08075964e-01 6.51466370e-01 -9.94282603e-01 -4.18603659e-01
6.18170977e-01 5.87026477e-01 -4.59514052e-01 2.14207947e-01
-2.65676081e-01 7.47718334e-01 -2.20737636e-01 4.00904506e-01
3.51813585e-01 -6.10205650e-01 3.08800161e-01 -2.04088122e-01
1.34869246e-02 5.94225563e-02 -1.37646556e+00 2.08369374e+00
-7.02722311e-01 6.50556862e-01 -1.20078191e-01 -9.59861100e-01
5.41181564e-01 5.10035455e-01 5.82034826e-01 -3.71929407e-01
-1.14130117e-01 -2.90763348e-01 -2.94556022e-01 -5.84769726e-01
2.18966573e-01 8.45299512e-02 -1.00261681e-02 7.37113714e-01
3.58714819e-01 -1.09830774e-01 -7.71856457e-02 5.75332165e-01
1.18589425e+00 3.21621269e-01 5.95764279e-01 1.98011011e-01
7.68529624e-02 -2.44136527e-01 3.98544639e-01 7.63440609e-01
-1.14027619e-01 8.60203624e-01 3.16631168e-01 -2.11513862e-01
-8.98437142e-01 -1.29800987e+00 -9.88664627e-02 1.10967541e+00
2.06437886e-01 -4.10135716e-01 -3.10514659e-01 -7.15489388e-01
-1.18749879e-01 7.96701610e-01 -9.30121601e-01 -7.40687847e-02
-1.67614788e-01 -2.81155348e-01 1.47672951e-01 5.60479760e-01
3.36696893e-01 -1.11003518e+00 -9.13912892e-01 9.77054834e-02
-3.10673624e-01 -1.17682695e+00 -2.14881539e-01 2.94387907e-01
-9.38635707e-01 -1.03953123e+00 -5.24653435e-01 -4.23060358e-01
7.44586945e-01 4.46762443e-01 1.19309568e+00 -2.72152543e-01
-3.35048676e-01 1.15955472e+00 -3.36033523e-01 -1.98502243e-01
-1.93381011e-01 -3.16392660e-01 5.84775768e-02 3.19501370e-01
2.38497093e-01 -8.11773121e-01 -7.91090488e-01 2.89640099e-01
-1.12361288e+00 1.77210644e-01 7.72869527e-01 6.62699997e-01
7.56445944e-01 -3.69449586e-01 5.20151854e-01 -9.26410079e-01
8.81557167e-02 -1.01343751e+00 -1.78487092e-01 2.03022361e-01
-1.08274445e-01 -3.76811624e-02 2.34848499e-01 -5.83121300e-01
-1.14945745e+00 5.06958723e-01 3.49879652e-01 -6.99354053e-01
-8.88422489e-01 3.52327764e-01 -2.30681486e-02 3.51768434e-01
8.55958700e-01 3.08372140e-01 -1.76603384e-02 -1.66450649e-01
6.29277170e-01 4.23275143e-01 4.06750143e-01 -5.04888177e-01
4.81137812e-01 7.60571599e-01 -3.55921984e-01 -8.80106807e-01
-8.65196586e-01 -6.58393562e-01 -6.33357763e-01 -3.06657314e-01
9.40472007e-01 -1.13811612e+00 -3.49230349e-01 1.36587948e-01
-1.11116648e+00 -2.60118306e-01 -2.24590749e-01 5.80998063e-01
-8.36396635e-01 1.80251107e-01 -3.60175222e-01 -8.06263328e-01
2.89789915e-01 -9.30673003e-01 1.30715156e+00 2.26044104e-01
-5.72793484e-01 -9.51580524e-01 9.44047347e-02 5.00137806e-01
1.34557322e-01 3.97932649e-01 3.59535396e-01 -5.45427382e-01
-7.77154148e-01 7.03288540e-02 -1.36693433e-01 3.66710812e-01
5.68047762e-01 1.02460943e-03 -1.30034566e+00 -1.47895351e-01
-2.59117186e-02 -4.81819987e-01 8.76401484e-01 4.99571741e-01
1.28509152e+00 -3.48149329e-01 -4.20865595e-01 2.66678214e-01
1.48403811e+00 2.32541502e-01 4.27356958e-01 1.77814942e-02
6.00714445e-01 7.15195298e-01 4.51639712e-01 6.34881735e-01
4.77008879e-01 5.03690600e-01 6.54561102e-01 -8.81526545e-02
-9.56975389e-03 -2.17887580e-01 2.81702280e-01 2.99890876e-01
-1.01783141e-01 -8.38127434e-02 -7.35772431e-01 8.09352040e-01
-1.86577630e+00 -1.05136013e+00 3.24518919e-01 2.15380716e+00
5.62823296e-01 -1.34418219e-01 2.13607270e-02 -1.25620037e-01
5.34131050e-01 3.18661034e-01 -9.92877305e-01 -8.19929242e-02
1.50539465e-02 -5.38597703e-02 3.43882084e-01 5.00072420e-01
-1.16605914e+00 7.78495908e-01 6.02832890e+00 2.97139168e-01
-9.23619866e-01 1.13435149e-01 6.73609972e-01 -4.37882841e-01
-4.92184281e-01 -2.68761311e-02 -4.17250246e-01 2.79197812e-01
7.64559805e-01 -3.45503688e-02 4.34464812e-01 6.57914519e-01
1.71741366e-01 -4.95101005e-01 -1.33521366e+00 1.01465964e+00
3.17990303e-01 -1.25679421e+00 -1.29828164e-02 3.00985515e-01
9.00409222e-01 1.82374299e-01 2.60151267e-01 -6.77385479e-02
6.65249825e-01 -9.20830905e-01 5.00807703e-01 6.82650387e-01
6.91474319e-01 -4.82954472e-01 1.65781915e-01 4.96222109e-01
-9.29923058e-01 1.07081518e-01 -3.23827751e-03 -9.13808495e-02
1.23043880e-01 6.05823278e-01 -7.74712086e-01 4.22009706e-01
6.51377678e-01 7.82665253e-01 -4.82805073e-01 6.34693503e-01
-2.13860959e-01 7.32927501e-01 -3.90254855e-01 3.12506616e-01
6.26700968e-02 5.91281392e-02 5.51847517e-01 1.07490623e+00
2.52866179e-01 2.33392507e-01 1.84148088e-01 8.86689663e-01
6.78197062e-03 -1.13871679e-01 -1.28017569e+00 2.43362918e-01
4.18329567e-01 1.34244990e+00 -5.33129454e-01 -5.11963189e-01
-6.93259954e-01 1.11595213e+00 4.12285447e-01 7.66459584e-01
-6.59306526e-01 2.91610181e-01 6.49721086e-01 6.26806021e-02
2.53608286e-01 -3.01575661e-01 -2.78706253e-01 -1.27153325e+00
9.10473801e-03 -7.00541019e-01 2.68391341e-01 -1.00969672e+00
-1.20910430e+00 4.15185899e-01 2.01144278e-01 -1.37775433e+00
-5.17173350e-01 -3.14767838e-01 -4.91236031e-01 6.12384260e-01
-1.06834173e+00 -1.12863910e+00 -1.99400455e-01 7.08489239e-01
7.70339370e-01 -1.37947842e-01 8.99486780e-01 -8.77624527e-02
-1.76570475e-01 9.65593159e-02 -2.61085689e-01 -5.80979995e-02
7.22991228e-01 -1.30159450e+00 2.44847387e-01 7.16201901e-01
8.30738187e-01 6.06630683e-01 5.48034787e-01 -5.13494372e-01
-1.24244165e+00 -9.26251531e-01 6.66427851e-01 -6.97238326e-01
4.56496298e-01 -2.73999155e-01 -9.10704076e-01 1.03500235e+00
4.62684572e-01 1.83270782e-01 9.62431788e-01 1.62516922e-01
-5.10963798e-01 1.43880516e-01 -1.13442290e+00 6.53956532e-01
1.04451799e+00 -7.83001661e-01 -5.74995875e-01 2.87217766e-01
5.24239719e-01 -3.61336082e-01 -7.86372602e-01 3.19162548e-01
5.13723969e-01 -1.16156268e+00 8.63644421e-01 -6.45479679e-01
8.14412296e-01 -2.04583168e-01 -5.48747897e-01 -1.37907481e+00
-2.79721200e-01 -2.37632900e-01 -2.51549304e-01 1.06828880e+00
4.07080173e-01 -5.35340428e-01 6.21914506e-01 6.76516771e-01
2.75791526e-01 -6.66993856e-01 -8.22658300e-01 -4.54619706e-01
-1.87087387e-01 -7.12047815e-01 1.35446906e-01 1.03309536e+00
4.26502712e-02 2.92910039e-01 -4.53048646e-01 5.44764400e-01
7.18063474e-01 1.10045850e-01 7.04063296e-01 -7.49476016e-01
-3.74615580e-01 -9.54793692e-02 -4.36923683e-01 -1.06007099e+00
1.33379251e-01 -7.36678123e-01 -9.04316269e-03 -1.61212230e+00
5.38924336e-01 -2.22560570e-01 -4.99028742e-01 4.66269016e-01
-1.46075696e-01 1.37934446e-01 3.60609829e-01 2.12481380e-01
-7.01994002e-01 2.42916137e-01 1.17278624e+00 7.48364776e-02
-4.19503078e-04 -2.35073976e-02 -8.22152734e-01 8.81525517e-01
9.42541242e-01 -3.36426318e-01 -5.01008928e-01 -4.94244456e-01
1.35913625e-01 4.79308069e-01 6.65543079e-01 -9.14344132e-01
2.56993353e-01 -2.86502779e-01 6.43406749e-01 -3.94080460e-01
5.57324946e-01 -9.22618389e-01 1.98499292e-01 2.43566185e-02
-6.14779711e-01 -4.31812257e-02 1.94874614e-01 9.49885130e-01
-3.26943666e-01 7.16184732e-04 4.93563533e-01 -5.53713620e-01
-8.02536547e-01 2.24466801e-01 -6.05222762e-01 -1.30298913e-01
1.08914638e+00 -1.97971851e-01 -1.58230647e-01 -7.47249246e-01
-1.18203807e+00 6.01660013e-02 5.49511135e-01 5.75438380e-01
8.70033324e-01 -1.35128736e+00 -7.00090826e-01 2.36658782e-01
3.24745297e-01 3.86641324e-02 4.13008928e-01 4.51862603e-01
1.79743677e-01 1.04957275e-01 -1.52322024e-01 -9.18898880e-01
-1.31001854e+00 2.00228184e-01 -1.25259981e-02 -1.21024445e-01
-4.05140251e-01 6.65793538e-01 6.31375670e-01 -4.54057068e-01
-1.54044610e-02 -1.82754278e-01 -3.12483311e-01 -2.14951634e-02
5.12497008e-01 2.18995363e-01 -3.33875358e-01 -7.30032265e-01
-3.21295023e-01 5.38194895e-01 -5.79779334e-02 -6.36122108e-01
1.15777445e+00 -2.68342614e-01 2.94857085e-01 9.55370069e-01
1.22758269e+00 -1.41311437e-01 -1.64687371e+00 -5.44969022e-01
-3.22558284e-01 -4.33795899e-01 -5.18269837e-02 -8.82302225e-01
-1.03808641e+00 7.24115014e-01 5.74530602e-01 1.15625322e-01
1.04827857e+00 5.48018575e-01 2.37409458e-01 1.61798954e-01
5.24160147e-01 -8.11647534e-01 4.17623430e-01 2.60765433e-01
1.03093553e+00 -1.55171585e+00 -2.26895809e-02 -1.55045331e-01
-1.03565764e+00 7.69825220e-01 6.46489799e-01 -3.16052258e-01
6.76574945e-01 2.82320738e-01 1.52773246e-01 -2.33802825e-01
-1.11385000e+00 -3.21120292e-01 3.84609759e-01 5.00079691e-01
4.86564875e-01 3.40472817e-01 2.71648556e-01 1.71112940e-01
1.11372024e-01 -6.90799803e-02 4.85908926e-01 9.68839943e-01
-1.83123022e-01 -8.00616443e-01 -3.50527763e-01 6.66697502e-01
-3.58006448e-01 1.11264572e-01 -3.52651715e-01 5.20675361e-01
1.68468744e-01 1.06861019e+00 3.10660660e-01 -3.77058238e-01
7.27869645e-02 6.69473410e-02 6.58922136e-01 -1.02911353e+00
-1.17934562e-01 2.02319115e-01 -3.64136472e-02 -7.77166545e-01
-1.02048421e+00 -1.13976967e+00 -1.07213187e+00 2.20338374e-01
-3.36455479e-02 -4.60008889e-01 5.96431732e-01 1.01447356e+00
3.21909934e-01 4.34640944e-01 6.69644058e-01 -1.05941784e+00
-3.45260277e-02 -9.26697016e-01 -4.40235257e-01 6.35628700e-01
4.45600033e-01 -5.86350143e-01 -4.63628471e-01 5.98730147e-01]
|
[9.94236946105957, 0.32377997040748596]
|
92869db6-de7a-46ba-a58e-8757b61d0854
|
towards-robustness-of-text-to-sql-models-3
|
2212.09994
| null |
https://arxiv.org/abs/2212.09994v1
|
https://arxiv.org/pdf/2212.09994v1.pdf
|
Towards Robustness of Text-to-SQL Models Against Natural and Realistic Adversarial Table Perturbation
|
The robustness of Text-to-SQL parsers against adversarial perturbations plays a crucial role in delivering highly reliable applications. Previous studies along this line primarily focused on perturbations in the natural language question side, neglecting the variability of tables. Motivated by this, we propose the Adversarial Table Perturbation (ATP) as a new attacking paradigm to measure the robustness of Text-to-SQL models. Following this proposition, we curate ADVETA, the first robustness evaluation benchmark featuring natural and realistic ATPs. All tested state-of-the-art models experience dramatic performance drops on ADVETA, revealing models' vulnerability in real-world practices. To defend against ATP, we build a systematic adversarial training example generation framework tailored for better contextualization of tabular data. Experiments show that our approach not only brings the best robustness improvement against table-side perturbations but also substantially empowers models against NL-side perturbations. We release our benchmark and code at: https://github.com/microsoft/ContextualSP.
|
['Jian-Guang Lou', 'Zhoujun Li', 'Jiaqi Guo', 'Yan Gao', 'Bing Wang', 'Xinyu Pi']
|
2022-12-20
|
towards-robustness-of-text-to-sql-models-2
|
https://aclanthology.org/2022.acl-long.142
|
https://aclanthology.org/2022.acl-long.142.pdf
|
acl-2022-5
|
['text-to-sql']
|
['computer-code']
|
[ 4.28202078e-02 3.27492833e-01 -2.68149609e-03 -2.29088202e-01
-1.38521254e+00 -1.14549530e+00 5.57702899e-01 1.23995863e-01
1.01762772e-01 5.47508836e-01 1.96442783e-01 -8.74904752e-01
2.85256356e-01 -9.39905047e-01 -1.35319114e+00 -3.85085285e-01
6.34156987e-02 2.60433316e-01 4.55548257e-01 -6.00145400e-01
1.70037121e-01 4.30015624e-01 -9.94407058e-01 7.05652833e-01
8.26508880e-01 5.66884458e-01 -5.22250295e-01 8.87400866e-01
-9.59654525e-02 1.06666887e+00 -9.77545977e-01 -1.22320902e+00
3.95253301e-01 7.12151080e-02 -8.26985717e-01 -6.52830541e-01
4.66013223e-01 -1.81683093e-01 -5.40981889e-01 1.10914373e+00
7.45359838e-01 -4.09772515e-01 9.92089063e-02 -1.47049212e+00
-6.88617289e-01 1.46298277e+00 -2.43701115e-01 1.18796110e-01
5.47900677e-01 7.08820939e-01 8.82025361e-01 -2.72467673e-01
6.34113193e-01 1.41981745e+00 7.25540161e-01 7.36344397e-01
-1.32523084e+00 -5.65353990e-01 6.44301400e-02 -6.61660358e-02
-9.52469587e-01 -5.74314535e-01 8.00006986e-01 -1.31493002e-01
8.70403707e-01 8.16771567e-01 -4.25411880e-01 1.92237008e+00
4.19454038e-01 7.31382251e-01 9.93066907e-01 -3.59682262e-01
2.26179734e-01 4.95234460e-01 4.93783876e-02 2.96624064e-01
3.18830073e-01 1.68987527e-01 -4.55882251e-01 -5.70309997e-01
-2.57141758e-02 -5.18927991e-01 -1.89957559e-01 -2.66048729e-01
-1.01348054e+00 6.99832618e-01 1.85455337e-01 6.88817948e-02
-7.72803500e-02 2.63910025e-01 9.99050081e-01 5.00870645e-01
1.50444984e-01 8.47999454e-01 -8.64800513e-01 -4.34691459e-01
-2.69862711e-01 5.98829806e-01 1.15050268e+00 1.14269567e+00
1.33268207e-01 -9.79193137e-04 -4.52177793e-01 4.09942269e-01
-3.06990426e-02 7.87284255e-01 2.02921599e-01 -7.54460990e-01
1.08791947e+00 4.22633767e-01 1.71229206e-02 -8.75882685e-01
2.89718527e-02 -1.76014975e-01 -4.67799366e-01 -4.48320934e-04
5.55033684e-01 -2.95909584e-01 -6.21463835e-01 1.71505606e+00
4.02334660e-01 -2.79290020e-01 2.90780991e-01 3.40551585e-01
5.81757426e-01 3.88706535e-01 1.18166789e-01 1.48976296e-01
1.38241148e+00 -6.22725606e-01 -5.68259954e-01 -1.97244793e-01
7.32369542e-01 -7.30646193e-01 1.80407393e+00 3.64290446e-01
-1.01782072e+00 -2.48450264e-01 -1.00205398e+00 9.97022316e-02
-8.29801857e-01 -6.27867639e-01 4.22920734e-01 1.26807582e+00
-7.69193411e-01 5.87515473e-01 -9.69299912e-01 -6.93665892e-02
4.08531666e-01 -5.19098341e-03 -4.65243697e-01 1.46295622e-01
-1.39749420e+00 8.78926635e-01 1.57532692e-01 -2.54341066e-01
-8.20289671e-01 -1.20976901e+00 -8.34137976e-01 -6.42124489e-02
6.52970195e-01 -4.67523605e-01 1.44356728e+00 -2.18105704e-01
-1.48078048e+00 6.76304460e-01 5.61338477e-02 -7.42449403e-01
1.01467025e+00 -5.18165648e-01 -4.25098717e-01 -1.70142278e-01
-9.74372998e-02 5.58855664e-03 7.06805110e-01 -1.48611069e+00
-2.60957517e-02 -1.80956706e-01 4.00119722e-01 -4.77987945e-01
-3.80854309e-01 3.04028839e-01 -3.01722497e-01 -6.97237313e-01
-5.03964484e-01 -7.82192647e-01 -1.74690425e-01 -5.05122364e-01
-1.14038587e+00 2.00680882e-01 7.71115243e-01 -7.11096346e-01
1.36656082e+00 -2.09994340e+00 -7.65400603e-02 1.03865594e-01
9.69037879e-03 4.62303013e-01 -2.42920622e-01 7.79165626e-01
-2.68481165e-01 8.00922692e-01 -3.39573652e-01 -1.41770214e-01
4.63199317e-01 -2.23852713e-02 -9.91259694e-01 1.88982695e-01
4.40107167e-01 1.22800076e+00 -7.10982919e-01 -4.05360222e-01
5.56694455e-02 1.88077539e-01 -7.98069417e-01 3.78152490e-01
-8.20660472e-01 2.89606571e-01 -4.78041291e-01 9.22593951e-01
8.49659801e-01 2.40443096e-01 1.17218032e-01 1.48976082e-03
2.87508368e-01 4.63697076e-01 -9.26518679e-01 1.35329700e+00
-4.87177312e-01 2.49047101e-01 -1.79301500e-01 -4.55401570e-01
7.41675735e-01 7.97520671e-03 -9.63232573e-03 -7.40596056e-01
-3.94220613e-02 8.99512619e-02 -1.12635776e-01 -6.31136894e-01
4.68344659e-01 2.84419924e-01 -6.52334511e-01 2.48317137e-01
-2.97856152e-01 -3.91517282e-01 1.32158905e-01 4.75651801e-01
1.62111795e+00 1.84137672e-01 1.37238085e-01 -6.55388460e-02
6.65668190e-01 -6.73516393e-02 3.62206161e-01 1.08496678e+00
-2.36412019e-01 6.07457876e-01 1.21292055e+00 -4.06463176e-01
-1.06869197e+00 -1.28889799e+00 -1.50951445e-01 9.82573748e-01
-3.35336715e-01 -6.03202701e-01 -1.17177415e+00 -1.37502110e+00
2.69404799e-01 1.22501719e+00 -9.22305942e-01 -4.37988639e-01
-7.59433210e-01 -7.41571009e-01 1.31376970e+00 4.56826538e-01
3.30356240e-01 -1.25860155e+00 -2.16070816e-01 1.83089338e-02
-1.64092615e-01 -1.32812119e+00 -4.02939051e-01 3.46608162e-01
-4.77514774e-01 -1.15767741e+00 1.50330560e-02 -1.91105217e-01
1.57602951e-01 -3.48877281e-01 1.59482074e+00 -3.89024569e-03
-1.68076769e-01 2.00869888e-01 -5.42584717e-01 -6.15255952e-01
-1.30741775e+00 4.29165483e-01 -1.60256609e-01 -4.50675964e-01
2.57953942e-01 -6.92264855e-01 -1.83337152e-01 3.56459916e-01
-1.20485902e+00 -6.07035637e-01 3.27350467e-01 5.76540589e-01
3.57122362e-01 -1.34339079e-01 5.91193914e-01 -1.54203331e+00
6.80538177e-01 -6.22588098e-01 -7.91621745e-01 4.61735934e-01
-4.65893775e-01 2.94256359e-01 1.23104215e+00 -4.53535587e-01
-8.31692040e-01 -3.46488059e-01 -4.98698920e-01 -1.95448503e-01
-3.85962188e-01 2.71934628e-01 -8.39531243e-01 -1.03962332e-01
1.24417853e+00 1.12615265e-01 -2.49940246e-01 -3.88956726e-01
6.54457450e-01 4.83501107e-01 6.69769287e-01 -9.76870894e-01
1.40720570e+00 1.50857925e-01 -1.55188948e-01 -2.04301566e-01
-3.95535499e-01 2.28174478e-01 -3.83197933e-01 4.24735487e-01
3.70787621e-01 -4.70187187e-01 -9.87619460e-01 5.39028168e-01
-1.10312986e+00 -6.49594665e-01 -2.64538020e-01 -3.47698063e-01
-5.64766467e-01 6.70830309e-01 -5.99911690e-01 -5.71463346e-01
-4.73892838e-01 -1.31965375e+00 9.85910952e-01 -2.89152414e-01
-1.46142632e-01 -8.39237869e-01 2.38940820e-01 3.89781743e-01
6.04761541e-01 7.47870684e-01 1.16103768e+00 -1.12612236e+00
-4.76479083e-01 -4.17833328e-01 1.46125525e-01 5.32881320e-01
-1.27951890e-01 3.68252069e-01 -1.20083916e+00 -7.92313665e-02
8.86896029e-02 -2.95329809e-01 3.82345825e-01 -2.95690745e-01
1.28027356e+00 -6.22829854e-01 -1.86423641e-02 8.64885926e-01
1.33499420e+00 -8.24132636e-02 1.04867446e+00 7.11634576e-01
7.57475615e-01 4.31694031e-01 6.41473353e-01 3.62005293e-01
2.78041929e-01 5.98523498e-01 7.63066232e-01 2.41291448e-01
1.58364698e-01 -4.78311747e-01 7.34416604e-01 3.65230918e-01
5.72966576e-01 -5.33751726e-01 -1.15439630e+00 3.99001390e-01
-1.42352879e+00 -7.14227974e-01 -1.07691638e-01 2.10912228e+00
1.17440009e+00 6.17949188e-01 6.47223294e-02 2.88062871e-01
4.34773594e-01 2.76843846e-01 -5.55548608e-01 -8.33014727e-01
-1.93499193e-01 2.79754907e-01 6.37669027e-01 4.23965573e-01
-1.14143956e+00 1.25984931e+00 5.82478333e+00 8.50209177e-01
-9.50379431e-01 -2.68618651e-02 7.68029273e-01 2.43387260e-02
-5.78348756e-01 -1.82601422e-01 -7.45777488e-01 6.06777906e-01
1.47864389e+00 -3.65876079e-01 4.37701523e-01 1.14462066e+00
5.10041928e-03 3.92141014e-01 -1.21813214e+00 1.95856705e-01
-2.87429333e-01 -1.12968230e+00 2.20003977e-01 -1.42481312e-01
3.27717423e-01 -6.81010038e-02 5.32380104e-01 6.43739700e-01
6.35523260e-01 -1.11642015e+00 8.36499572e-01 2.04986975e-01
7.64078379e-01 -8.64975512e-01 9.33556497e-01 1.46896392e-01
-6.33485675e-01 9.12356898e-02 -3.70427758e-01 3.49417865e-01
-1.00798115e-01 3.38181198e-01 -9.85952497e-01 6.99196160e-01
8.61140490e-01 8.52746964e-02 -1.10641360e+00 3.84325832e-01
-2.87170887e-01 9.49923456e-01 -2.57445842e-01 1.12792984e-01
-3.60885039e-02 4.32215542e-01 6.31947815e-01 1.38003218e+00
-8.12752992e-02 -2.73642033e-01 -3.66121799e-01 9.52188849e-01
-3.04573029e-01 -2.98570599e-02 -9.48623717e-01 -1.51096463e-01
7.81119645e-01 9.84294176e-01 -2.61338025e-01 -1.11716986e-03
-2.08229572e-01 8.04021120e-01 4.21436608e-01 1.78821430e-01
-1.19034219e+00 -4.82869148e-01 8.61586690e-01 1.25662133e-01
2.71087319e-01 1.25764012e-01 -6.23784959e-01 -1.03524101e+00
5.67515731e-01 -1.89543593e+00 3.47567409e-01 -3.00825626e-01
-1.42048073e+00 8.63700509e-01 2.74785571e-02 -8.92822981e-01
-2.78304309e-01 -5.52988946e-01 -7.27462530e-01 8.43675733e-01
-1.19875944e+00 -1.25123215e+00 -5.05992435e-02 5.50291777e-01
2.25621000e-01 -2.76683956e-01 9.85381484e-01 -2.22185254e-03
-7.22842157e-01 1.61133969e+00 4.01970521e-02 2.15900242e-01
8.20863485e-01 -1.62971866e+00 1.54446089e+00 1.28223562e+00
-1.64701834e-01 9.25320208e-01 1.15647292e+00 -5.47547936e-01
-1.85254347e+00 -1.30390155e+00 5.80571711e-01 -1.38542438e+00
1.10835314e+00 -8.04737687e-01 -1.22854304e+00 8.24980974e-01
1.19449191e-01 9.69660804e-02 4.91014183e-01 -9.93789509e-02
-1.01734746e+00 -1.85546324e-01 -1.56087470e+00 1.00302684e+00
8.17076385e-01 -6.61422074e-01 -3.30881774e-01 5.09994030e-01
1.50509274e+00 -7.43398964e-01 -1.23261070e+00 5.13350427e-01
3.57124180e-01 -1.17205036e+00 1.07386672e+00 -8.94725919e-01
6.95792854e-01 -5.47962189e-02 -4.55491364e-01 -1.08308291e+00
1.75058141e-01 -1.12854064e+00 -3.08200479e-01 1.65771163e+00
7.50088036e-01 -9.82113302e-01 8.23293567e-01 8.03310513e-01
-7.98667222e-02 -6.16941571e-01 -7.96001017e-01 -8.34453702e-01
7.28164911e-01 -6.56351388e-01 1.07596326e+00 8.56044173e-01
-3.47555466e-02 -2.91704357e-01 -1.83865398e-01 5.89602828e-01
5.30919194e-01 -4.45337892e-01 1.24429524e+00 -3.90809655e-01
-5.63253045e-01 -2.65474439e-01 -2.00806618e-01 -2.70120889e-01
2.09653199e-01 -6.46540046e-01 4.12763357e-02 -7.14124799e-01
-1.19850144e-01 -3.14469934e-01 -9.25418809e-02 4.91987646e-01
-6.27461314e-01 1.35005778e-02 3.92028660e-01 -2.50566453e-01
-4.23226655e-01 1.79083690e-01 7.79295862e-01 -8.73336270e-02
-4.33144066e-03 1.42265707e-01 -1.04243577e+00 3.38452697e-01
1.07978332e+00 -7.81459093e-01 -2.66436428e-01 -2.81219035e-01
3.78205001e-01 -1.94829935e-03 2.59243309e-01 -1.05340052e+00
-1.74341351e-01 -1.89913720e-01 -2.70734608e-01 -2.41864815e-01
-2.15489134e-01 -5.99466622e-01 -9.71470997e-02 4.61870372e-01
-5.04480422e-01 4.24892783e-01 7.50455916e-01 3.35754067e-01
-3.13104764e-02 -6.28547817e-02 6.34903371e-01 2.82159839e-02
-8.78036395e-02 1.61332831e-01 -3.91441770e-02 5.45411348e-01
9.08972085e-01 3.31781834e-01 -9.28046763e-01 -1.96639940e-01
-3.19996566e-01 1.06134705e-01 6.70682788e-01 6.54396176e-01
2.59074837e-01 -8.64984930e-01 -8.94800305e-01 1.59431636e-01
3.30413610e-01 -1.40937433e-01 1.25090539e-01 3.65177184e-01
-8.09797287e-01 4.66069341e-01 6.52098358e-02 -2.56903410e-01
-1.19845128e+00 1.17245507e+00 3.33969116e-01 -6.86445832e-01
-4.32588041e-01 8.07278752e-01 -1.55947983e-01 -9.17627454e-01
3.51422578e-01 -4.01343077e-01 2.99268544e-01 -5.76373458e-01
3.68193746e-01 2.90480882e-01 4.80591118e-01 -9.28280354e-02
-5.19127607e-01 1.35693897e-03 -4.58882302e-01 4.11373330e-03
1.00293612e+00 4.28369045e-01 -1.05307989e-01 1.59896299e-01
1.12342763e+00 5.79890072e-01 -9.02813256e-01 -7.08812401e-02
2.22579688e-01 -4.71324265e-01 -5.77552140e-01 -1.23145270e+00
-8.22647393e-01 7.89278150e-01 2.07996070e-01 5.50966263e-01
8.66275787e-01 -2.91908950e-01 8.23462069e-01 4.55548733e-01
3.38515610e-01 -3.71447712e-01 -4.44748029e-02 4.53917950e-01
1.11606205e+00 -1.28020394e+00 -2.09040299e-01 -5.81531644e-01
-8.50366473e-01 8.77080083e-01 7.42926061e-01 1.06052801e-01
3.10700893e-01 8.43023300e-01 4.07603383e-01 1.41580895e-01
-1.00964141e+00 4.75363314e-01 -9.60980803e-02 9.21172023e-01
3.63097548e-01 -2.31494904e-02 1.55127972e-01 7.91307688e-01
-6.13333642e-01 -4.10465032e-01 7.29915619e-01 9.99668837e-01
2.38773599e-01 -1.56347799e+00 -5.69632590e-01 8.54511708e-02
-1.02351403e+00 -4.39474463e-01 -5.86197019e-01 1.06544566e+00
-3.55349749e-01 1.00954092e+00 -6.11364007e-01 -8.13610733e-01
6.99733615e-01 2.37200394e-01 5.76510355e-02 -3.68587226e-01
-1.36906230e+00 -6.90249503e-01 3.29441816e-01 -9.86445785e-01
6.86132848e-01 -5.92187643e-01 -7.33945549e-01 -8.17079544e-01
7.03112856e-02 1.60355419e-01 6.81097388e-01 4.07403201e-01
5.78697383e-01 7.27034211e-01 7.76908278e-01 -2.88781762e-01
-1.40746307e+00 -9.83246505e-01 -6.92523494e-02 6.35303080e-01
2.68360674e-01 -8.77248347e-02 -6.91081882e-01 -1.07771397e-01]
|
[6.085960388183594, 8.104360580444336]
|
d9c9ca0c-69cc-4b4f-a90f-f9a999e930a3
|
semantic-unfolding-of-stylegan-latent-space
|
2206.14892
| null |
https://arxiv.org/abs/2206.14892v1
|
https://arxiv.org/pdf/2206.14892v1.pdf
|
Semantic Unfolding of StyleGAN Latent Space
|
Generative adversarial networks (GANs) have proven to be surprisingly efficient for image editing by inverting and manipulating the latent code corresponding to an input real image. This editing property emerges from the disentangled nature of the latent space. In this paper, we identify that the facial attribute disentanglement is not optimal, thus facial editing relying on linear attribute separation is flawed. We thus propose to improve semantic disentanglement with supervision. Our method consists in learning a proxy latent representation using normalizing flows, and we show that this leads to a more efficient space for face image editing.
|
['Pierre Hellier', 'Bharath Bushan Damodaran', 'Xu Yao', 'Mustafa Shukor']
|
2022-06-29
| null | null | null | null |
['facial-editing']
|
['computer-vision']
|
[ 7.59984791e-01 5.55845857e-01 -3.82871218e-02 -2.23890364e-01
-4.24660951e-01 -9.10281658e-01 8.79204154e-01 -6.59074664e-01
-1.55030310e-01 9.01027560e-01 2.52681524e-01 -9.62744504e-02
-1.16010882e-01 -7.54236758e-01 -7.02524841e-01 -8.15537691e-01
3.36356789e-01 3.59622419e-01 -7.73571074e-01 -6.63220510e-02
-1.29138753e-02 7.25172937e-01 -1.31854260e+00 8.68589990e-03
8.90982151e-01 4.84275997e-01 -4.39926833e-01 8.81657302e-01
-8.42692424e-03 7.63725340e-01 -5.50380647e-01 -8.09798181e-01
5.96749306e-01 -8.37891638e-01 -8.09511185e-01 1.92083433e-01
7.08036959e-01 -3.87833625e-01 -2.88349867e-01 1.16700840e+00
1.53102100e-01 -1.75558671e-01 9.27041650e-01 -1.81319559e+00
-1.07449317e+00 2.72175282e-01 -3.17948520e-01 -3.09333175e-01
2.76768744e-01 -1.91261709e-01 1.15930951e+00 -7.76979744e-01
7.20369220e-01 1.22509289e+00 4.68634397e-01 9.65623081e-01
-1.79147744e+00 -7.25538909e-01 -3.17415267e-01 -1.05178818e-01
-1.28963935e+00 -6.77737057e-01 1.13404727e+00 -4.82488364e-01
3.12188238e-01 5.63003063e-01 6.46504998e-01 1.51888108e+00
-7.26491120e-03 4.89627957e-01 1.31961238e+00 -6.28960252e-01
9.48088244e-02 1.32403344e-01 -5.98375857e-01 8.43416214e-01
3.40437949e-01 2.29606718e-01 -4.61988479e-01 -2.47553170e-01
1.03442156e+00 -5.33795618e-02 -2.00247332e-01 -9.14931595e-01
-1.13099825e+00 1.26194572e+00 1.42740488e-01 -3.74766961e-02
-2.21459582e-01 4.05687749e-01 3.17054331e-01 6.84562087e-01
5.78493357e-01 7.97491610e-01 -3.64908911e-02 -1.13855176e-01
-7.33477354e-01 1.54348731e-01 8.14300537e-01 7.81705201e-01
6.76454663e-01 3.50048155e-01 -8.30189735e-02 6.15974069e-01
1.07886836e-01 4.63791281e-01 1.26351744e-01 -1.61435688e+00
1.83823079e-01 4.24020827e-01 -6.47169724e-02 -9.69536960e-01
1.76232740e-01 -1.20891497e-01 -8.53720427e-01 7.86150277e-01
4.76535738e-01 -9.21978578e-02 -9.04753506e-01 2.25597334e+00
-3.24733146e-02 7.80437142e-02 2.12156892e-01 5.50891340e-01
-2.65256241e-02 2.56634384e-01 1.30561233e-01 -1.16895720e-01
1.14757180e+00 -7.29870379e-01 -1.07612360e+00 -9.52685997e-02
2.15068057e-01 -6.34246171e-01 1.14370477e+00 1.49622396e-01
-1.05233037e+00 -1.78496122e-01 -1.25952721e+00 -2.93364555e-01
-2.76292413e-01 8.78124237e-02 9.13252473e-01 1.00637615e+00
-1.07022679e+00 5.53974926e-01 -7.57099748e-01 -1.99193045e-01
5.74526429e-01 3.45601678e-01 -9.89376307e-01 2.57400990e-01
-1.15959466e+00 8.20667863e-01 6.36890158e-02 2.16818172e-02
-8.86725545e-01 -6.35016561e-01 -9.31552112e-01 5.48303910e-02
2.97340393e-01 -1.11334431e+00 8.14114988e-01 -1.71459913e+00
-1.88019049e+00 1.20559263e+00 -1.07634842e-01 -2.68838108e-01
9.17274773e-01 -1.72013879e-01 -7.14662299e-02 2.66655624e-01
8.57642218e-02 5.72965801e-01 1.55706716e+00 -1.34129381e+00
1.81449592e-01 -3.92431289e-01 2.67412961e-01 -1.75572671e-02
-4.60183054e-01 5.37040457e-02 1.18580714e-01 -9.13257897e-01
-5.88039085e-02 -1.16135728e+00 2.03950740e-02 3.46145749e-01
-4.79998320e-01 3.59182060e-01 7.16294706e-01 -6.34794474e-01
7.31612504e-01 -2.13703966e+00 5.08717477e-01 1.99300319e-01
6.75382257e-01 2.27617901e-02 -1.65042683e-01 3.23446870e-01
-5.79450071e-01 2.68511713e-01 -4.54364747e-01 -6.22401953e-01
1.49451792e-01 5.12344599e-01 -5.71192265e-01 6.34066284e-01
3.16714883e-01 1.04857063e+00 -7.50753880e-01 -4.10962552e-01
-5.11518978e-02 7.33660877e-01 -9.14472997e-01 3.28773737e-01
1.40353674e-02 7.15969026e-01 -2.55438536e-01 3.72040242e-01
6.40440643e-01 -2.53217877e-03 1.75183415e-01 -2.56425887e-01
1.50692344e-01 -1.00319833e-01 -8.32829833e-01 1.79155743e+00
-5.64371645e-01 8.61479878e-01 1.17452398e-01 -8.93669844e-01
7.03304827e-01 4.96594012e-01 5.11039555e-01 -2.05676660e-01
5.90878241e-02 1.23575948e-01 -1.56395301e-01 -3.22205514e-01
2.85042882e-01 -5.19382536e-01 6.18733540e-02 5.68698764e-01
2.38549098e-01 -1.29673317e-01 -2.28654325e-01 3.44478130e-01
9.41253304e-01 2.90626496e-01 3.02053928e-01 -3.04019153e-01
4.67425287e-01 -4.48054641e-01 4.57456529e-01 5.55426240e-01
-1.32345527e-01 7.08857894e-01 1.05270767e+00 -3.60958993e-01
-1.53783941e+00 -1.42134368e+00 9.38415453e-02 7.17732728e-01
-4.01443899e-01 -3.93671095e-01 -1.05158353e+00 -7.28086889e-01
-7.04789236e-02 7.45924234e-01 -1.03576016e+00 -5.59682965e-01
-6.36827350e-01 -4.43014324e-01 9.34540689e-01 5.52698016e-01
2.92813689e-01 -7.10087419e-01 -4.08550113e-01 -2.25182936e-01
-3.25046405e-02 -1.08959377e+00 -5.10080457e-01 9.88874808e-02
-6.81498766e-01 -9.51804221e-01 -7.07173049e-01 -2.45848492e-01
1.07419860e+00 -2.02697843e-01 9.55296934e-01 -1.95294872e-01
-2.15844274e-01 5.23958385e-01 5.45401243e-04 -1.88303292e-01
-8.16310942e-01 -1.23020436e-03 2.05068275e-01 3.93629432e-01
-4.33420911e-02 -1.04144871e+00 -5.15953004e-01 1.15884773e-01
-1.09837830e+00 2.60745883e-01 3.40372890e-01 1.02197385e+00
1.66844577e-01 -4.13471431e-01 3.65363151e-01 -1.34374774e+00
6.78081810e-01 -1.22281820e-01 -5.22271395e-01 2.82223344e-01
-8.61450970e-01 4.68765497e-01 8.59231353e-01 -3.46276283e-01
-1.15295112e+00 1.31157115e-01 -6.36885390e-02 -6.60159528e-01
-4.01021689e-02 -1.50053367e-01 -5.07286191e-01 -3.22664440e-01
6.12703860e-01 -3.56806032e-02 4.67747152e-01 -3.00258279e-01
8.80870461e-01 2.52578527e-01 5.61090350e-01 -6.87663972e-01
1.12002921e+00 8.92613590e-01 3.21815550e-01 -3.68215173e-01
-6.51512742e-01 2.92856485e-01 -8.28639269e-01 4.26268484e-03
1.08580399e+00 -7.26507246e-01 -8.42991352e-01 2.11515918e-01
-1.20628095e+00 3.61317024e-03 -8.19392622e-01 3.29755813e-01
-1.06817889e+00 2.51672179e-01 -5.30246019e-01 -5.31619906e-01
1.78149436e-02 -1.03538251e+00 9.58721280e-01 -2.86882669e-01
-3.88077438e-01 -8.60722482e-01 2.99707919e-01 2.58872539e-01
4.66869563e-01 5.89558661e-01 9.50832248e-01 -4.42361325e-01
-5.41496456e-01 -1.19486481e-01 -2.09470704e-01 4.54154849e-01
4.77514744e-01 1.76894501e-01 -1.21524715e+00 -2.40302682e-01
2.40118742e-01 -1.57456473e-01 7.71620333e-01 -9.50459689e-02
1.17666805e+00 -6.98695898e-01 1.02781035e-01 1.11733580e+00
1.31439459e+00 -1.02925815e-01 8.01432192e-01 1.21505104e-01
1.14793336e+00 6.35090351e-01 -1.25701010e-01 1.19520776e-01
-9.05686691e-02 7.54978418e-01 4.76876259e-01 -1.10435173e-01
-1.19071476e-01 -4.99112040e-01 3.92301142e-01 5.99699140e-01
-3.56315643e-01 6.08858801e-02 -5.82480073e-01 2.31215730e-01
-1.51767695e+00 -9.93258774e-01 3.22476834e-01 2.00334048e+00
9.36944008e-01 -2.12957263e-01 -6.82449341e-02 1.56580463e-01
5.38580835e-01 2.85248190e-01 -4.60443467e-01 -6.37126505e-01
-2.39453837e-01 4.29471850e-01 6.93897009e-01 7.21956968e-01
-8.15218627e-01 8.29373181e-01 7.13398266e+00 6.32034302e-01
-8.60995412e-01 4.07782584e-01 2.90562898e-01 -3.23009193e-02
-8.31063509e-01 2.16189191e-01 -1.65997699e-01 2.53983766e-01
6.81321084e-01 -2.87958831e-01 6.92771256e-01 7.03618348e-01
-1.52374238e-01 5.03585935e-01 -1.31199908e+00 9.38911140e-01
3.23012292e-01 -1.19949460e+00 5.53026915e-01 2.73670614e-01
7.58425415e-01 -7.36923397e-01 4.80593115e-01 -1.18581645e-01
2.81581461e-01 -1.37577355e+00 6.36660993e-01 4.96123314e-01
1.22093737e+00 -7.78035820e-01 2.19359621e-01 -1.03236534e-01
-6.67725265e-01 1.30915627e-01 -1.03961989e-01 7.28740394e-02
-4.50355047e-03 2.55029440e-01 -4.75660622e-01 4.72448945e-01
5.21382317e-03 4.25103635e-01 -4.18600023e-01 1.70126736e-01
-5.57428539e-01 3.28191400e-01 -1.48023572e-02 5.86348593e-01
-7.64495879e-02 -6.21129334e-01 8.19869637e-01 7.32260108e-01
3.05046171e-01 -8.20540413e-02 -5.28284669e-01 1.38912022e+00
-4.00852978e-01 -1.65233165e-01 -1.07596874e+00 -3.53135198e-01
2.13472787e-02 1.13601613e+00 -6.12594724e-01 -1.05425015e-01
-2.74475992e-01 1.65480876e+00 2.63799429e-01 4.71931368e-01
-8.62003922e-01 -2.81233519e-01 1.06407332e+00 -1.80924535e-01
-3.58766802e-02 -2.94894516e-01 -3.96192402e-01 -1.48018527e+00
-8.70262682e-02 -8.60752940e-01 9.53688994e-02 -6.22795522e-01
-1.23208869e+00 6.03774250e-01 -2.71480531e-02 -1.05483711e+00
-3.91890764e-01 -5.93669116e-01 -4.61675584e-01 9.56235468e-01
-1.31417179e+00 -1.41904044e+00 -1.48288473e-01 7.83208549e-01
1.54143512e-01 -1.90019056e-01 1.22719300e+00 2.76204497e-01
-3.41463208e-01 9.27179873e-01 -1.04709174e-02 8.67673680e-02
6.34048760e-01 -1.55669785e+00 3.78060967e-01 7.51909554e-01
3.21437746e-01 8.29484463e-01 9.00713325e-01 -2.98984200e-01
-1.44923663e+00 -8.90537977e-01 7.56105185e-01 -7.26110220e-01
6.11412942e-01 -7.09837079e-01 -6.48006618e-01 1.05803812e+00
4.20935899e-01 2.46425550e-02 8.23654830e-01 -1.57963172e-01
-8.44826221e-01 6.80034421e-03 -1.23378360e+00 8.56104314e-01
1.27384686e+00 -1.27999151e+00 -3.96831751e-01 1.58627167e-01
9.27247882e-01 1.43885124e-03 -6.28046811e-01 2.90179729e-01
6.92546070e-01 -7.95723915e-01 1.12736976e+00 -9.01749134e-01
4.89575386e-01 -4.99077961e-02 -7.69476593e-02 -1.28678155e+00
-6.75080642e-02 -1.01674664e+00 -4.15999405e-02 1.25986242e+00
1.23986520e-01 -8.97175610e-01 9.76447403e-01 7.29797661e-01
3.84209424e-01 -2.81672210e-01 -9.43109870e-01 -7.90054679e-01
2.38277957e-01 -3.26651335e-02 5.94211459e-01 1.34376395e+00
-2.86210984e-01 2.43862927e-01 -6.36269450e-01 -2.13441290e-02
8.18896890e-01 -1.29653944e-03 7.10824728e-01 -1.05825508e+00
-3.31621975e-01 -4.00039554e-01 -5.10173678e-01 -6.26239002e-01
7.19883978e-01 -9.93501902e-01 -4.35733497e-01 -7.42964447e-01
1.62459314e-01 -2.18575120e-01 -6.37753904e-02 4.69762504e-01
1.13202140e-01 5.61295569e-01 2.68545181e-01 2.48342305e-01
3.14598009e-02 7.77290523e-01 1.35816920e+00 -1.27762601e-01
1.97068900e-01 -1.82512730e-01 -7.50140488e-01 7.38235354e-01
7.25245297e-01 -5.60629964e-01 -4.28500295e-01 -3.52405071e-01
6.35831594e-01 -4.45926934e-02 7.34817266e-01 -5.03855109e-01
-4.66198958e-02 -5.79653457e-02 2.93326348e-01 3.00967187e-01
5.08390784e-01 -1.03189266e+00 4.88489211e-01 3.50679696e-01
-6.66690171e-01 8.61950070e-02 -1.14493757e-01 4.77291197e-01
-2.55412936e-01 -2.50808239e-01 7.20895708e-01 -2.43784636e-02
-1.70192018e-01 3.16943914e-01 -1.94316328e-01 -7.59091154e-02
9.28302586e-01 -3.10877532e-01 -1.60278961e-01 -5.01014411e-01
-1.01310170e+00 -4.90985155e-01 7.13008940e-01 2.37350255e-01
4.61024761e-01 -1.64663875e+00 -6.47404969e-01 6.49340570e-01
-2.85477433e-02 -5.80689132e-01 4.63850610e-02 6.77178383e-01
-6.73579156e-01 1.61995158e-01 -7.34026849e-01 -1.99867174e-01
-1.30328131e+00 5.05108595e-01 3.41947317e-01 -3.74073610e-02
-4.61525828e-01 7.60682821e-01 3.78188968e-01 -3.22096258e-01
-2.20128477e-01 2.47897446e-01 1.57383643e-02 1.28789797e-01
7.08246455e-02 3.85635376e-01 -2.12467894e-01 -7.10042894e-01
-8.01889375e-02 6.23817027e-01 9.79870632e-02 -3.96207333e-01
1.10507882e+00 -1.80961519e-01 -4.28893596e-01 2.71811396e-01
1.60740972e+00 5.68666518e-01 -1.42366767e+00 1.37220174e-01
-3.56461555e-01 -7.82344222e-01 -3.29321474e-01 -3.51933450e-01
-1.25617743e+00 8.78160536e-01 4.79614645e-01 3.55832815e-01
1.19097745e+00 -2.10606888e-01 4.50189918e-01 1.57141909e-01
1.01868846e-01 -7.26827264e-01 4.09966484e-02 1.19502455e-01
1.05411029e+00 -9.50117469e-01 -7.37840608e-02 -4.83007103e-01
-4.76276368e-01 1.00848639e+00 3.20068777e-01 -3.98507863e-01
5.23423612e-01 2.84914762e-01 -2.93020215e-02 -2.66014606e-01
-4.18895900e-01 9.32159275e-02 1.65428475e-01 5.57743311e-01
1.79607049e-01 2.40727827e-01 -2.32499808e-01 5.00740595e-02
-4.54064280e-01 -1.90799177e-01 5.62411249e-01 6.13128841e-01
4.06829953e-01 -1.57385385e+00 -3.26379955e-01 2.69817133e-02
-6.39142513e-01 -3.69143113e-02 -5.66877484e-01 6.63678408e-01
1.01044901e-01 4.30330336e-01 -8.69306773e-02 -7.47497156e-02
1.86850913e-02 3.04954916e-01 8.89551044e-01 -3.92706931e-01
-1.22890569e-01 -3.77652556e-01 -1.09190352e-01 -8.08093846e-01
-6.03231490e-01 -4.42001671e-01 -7.03302801e-01 -4.45929080e-01
-1.43843278e-01 1.91200674e-01 7.05549002e-01 6.77618623e-01
3.77895504e-01 4.72711086e-01 8.20254743e-01 -6.09631240e-01
-6.70835674e-01 -4.74919766e-01 -5.77785552e-01 6.85325563e-01
5.98500550e-01 -7.12511599e-01 -6.77725554e-01 3.89038384e-01]
|
[11.8656005859375, -0.2794452905654907]
|
370eb6f2-9926-43aa-a10e-99bd33bc794f
|
transformation-consistent-self-ensembling
|
1903.00348
| null |
https://arxiv.org/abs/1903.00348v3
|
https://arxiv.org/pdf/1903.00348v3.pdf
|
Transformation Consistent Self-ensembling Model for Semi-supervised Medical Image Segmentation
|
Deep convolutional neural networks have achieved remarkable progress on a variety of medical image computing tasks. A common problem when applying supervised deep learning methods to medical images is the lack of labeled data, which is very expensive and time-consuming to be collected. In this paper, we present a novel semi-supervised method for medical image segmentation, where the network is optimized by the weighted combination of a common supervised loss for labeled inputs only and a regularization loss for both labeled and unlabeled data. To utilize the unlabeled data, our method encourages the consistent predictions of the network-in-training for the same input under different regularizations. Aiming for the semi-supervised segmentation problem, we enhance the effect of regularization for pixel-level predictions by introducing a transformation, including rotation and flipping, consistent scheme in our self-ensembling model. With the aim of semi-supervised segmentation tasks, we introduce a transformation consistent strategy in our self-ensembling model to enhance the regularization effect for pixel-level predictions. We have extensively validated the proposed semi-supervised method on three typical yet challenging medical image segmentation tasks: (i) skin lesion segmentation from dermoscopy images on International Skin Imaging Collaboration (ISIC) 2017 dataset, (ii) optic disc segmentation from fundus images on Retinal Fundus Glaucoma Challenge (REFUGE) dataset, and (iii) liver segmentation from volumetric CT scans on Liver Tumor Segmentation Challenge (LiTS) dataset. Compared to the state-of-the-arts, our proposed method shows superior segmentation performance on challenging 2D/3D medical images, demonstrating the effectiveness of our semi-supervised method for medical image segmentation.
|
['Pheng-Ann Heng', 'Chi-Wing Fu', 'Lequan Yu', 'Lei Xing', 'Hao Chen', 'Xiaomeng Li']
|
2019-02-28
| null | null | null | null |
['semi-supervised-medical-image-segmentation', 'skin-lesion-segmentation', 'liver-segmentation']
|
['computer-vision', 'medical', 'medical']
|
[ 5.97620547e-01 4.02317166e-01 -4.34891164e-01 -7.23151267e-01
-8.03207517e-01 -3.58554095e-01 1.82798207e-01 3.66008617e-02
-7.11159527e-01 5.21995723e-01 -1.28047958e-01 -4.69752222e-01
1.36743551e-02 -4.49405909e-01 -6.26681745e-01 -6.93660796e-01
1.14264593e-01 4.37375188e-01 1.18251435e-01 2.64690071e-01
-1.33339584e-01 4.63572085e-01 -1.17037523e+00 2.82858074e-01
1.27261448e+00 1.13972425e+00 1.00160204e-01 4.19285357e-01
-2.97369093e-01 6.68497562e-01 -9.60832536e-02 -8.16792846e-02
5.68686485e-01 -4.50982690e-01 -1.29701090e+00 8.10465455e-01
5.73421419e-01 -1.76400363e-01 9.22364220e-02 1.21917295e+00
4.59370136e-01 -2.90751964e-01 6.80458188e-01 -8.01417947e-01
-3.14671338e-01 4.69623387e-01 -8.38850379e-01 -4.82138917e-02
-3.39900523e-01 2.80024558e-01 6.17586970e-01 -3.02985042e-01
7.61184096e-01 6.66953146e-01 6.81667447e-01 8.10701907e-01
-1.16649520e+00 -3.05773228e-01 -1.21201433e-01 -3.38248044e-01
-1.25076318e+00 7.86843803e-03 5.61502635e-01 -6.37602746e-01
5.59607089e-01 2.14991286e-01 6.65392995e-01 5.04552186e-01
2.66319700e-02 8.38982344e-01 1.48005235e+00 -4.38210040e-01
2.95097411e-01 1.88434452e-01 2.74408191e-01 9.23738301e-01
6.49281144e-02 7.08803162e-02 1.60465255e-01 8.46313164e-02
7.65961885e-01 5.75741269e-02 -2.49950439e-01 -3.56291980e-01
-8.67299199e-01 6.48242474e-01 5.94807923e-01 2.18436956e-01
-3.97009313e-01 -5.97857907e-02 3.35261166e-01 -3.68688256e-02
7.08032548e-01 2.38117352e-01 -5.90081096e-01 2.71491736e-01
-1.29416955e+00 -3.38645965e-01 6.16812944e-01 5.07530749e-01
7.54611671e-01 -1.89949751e-01 -3.49382818e-01 8.97402525e-01
4.73066777e-01 8.78136307e-02 7.37619340e-01 -6.55219138e-01
6.87944368e-02 9.85355139e-01 -1.95762128e-01 -2.38400444e-01
-6.86221004e-01 -6.78667068e-01 -1.04330790e+00 4.25021052e-01
7.37508655e-01 -2.55735874e-01 -1.61224985e+00 1.46634114e+00
5.36170602e-01 1.12800360e-01 -1.05183840e-01 1.12682581e+00
1.01930368e+00 2.25240752e-01 2.01983467e-01 -3.34295750e-01
1.17582083e+00 -1.18085802e+00 -3.56383830e-01 -5.92416339e-02
9.48417366e-01 -5.89592397e-01 9.15864348e-01 4.22897965e-01
-1.01871181e+00 -4.29149240e-01 -9.04007375e-01 2.50744950e-02
-9.64877978e-02 5.06427646e-01 7.22900569e-01 6.96777821e-01
-1.00622904e+00 6.14436805e-01 -9.72972751e-01 -3.74216169e-01
9.09024179e-01 5.96330583e-01 -3.30708832e-01 1.99659113e-02
-7.33999312e-01 5.61279655e-01 2.70234853e-01 1.50504664e-01
-6.54681742e-01 -8.21280956e-01 -7.24467814e-01 -2.28666797e-01
4.44204032e-01 -4.99262154e-01 1.02010775e+00 -1.47611034e+00
-1.55688632e+00 1.48157871e+00 6.85022324e-02 -8.50815594e-01
7.32495189e-01 9.56990048e-02 -4.17136662e-02 2.90403932e-01
1.33142872e-02 1.07272267e+00 8.36139619e-01 -1.26087630e+00
-4.95120585e-01 -5.13952732e-01 -3.31709266e-01 1.48178309e-01
-8.91734287e-02 -1.56849161e-01 -6.76756978e-01 -4.74472374e-01
1.15353473e-01 -1.01169920e+00 -5.94963074e-01 2.69546270e-01
-7.77470708e-01 -6.78342208e-02 6.79740608e-01 -7.21309543e-01
9.44295108e-01 -2.00916004e+00 -4.44882065e-02 3.63913625e-01
3.56373399e-01 6.90478563e-01 -1.48822531e-01 -4.12560046e-01
-3.39323044e-01 2.80988038e-01 -7.60646880e-01 -6.30105674e-01
-4.62433100e-01 2.05454707e-01 1.27316475e-01 6.18633807e-01
3.24100256e-01 9.91271794e-01 -6.68860078e-01 -7.13514924e-01
3.50789249e-01 4.74969029e-01 -3.77824247e-01 1.30352303e-01
-3.94649565e-01 9.37128305e-01 -2.64946043e-01 6.73223078e-01
6.53143644e-01 -6.44872844e-01 1.80473775e-01 -2.68026948e-01
1.10861763e-01 -8.57276097e-02 -8.37891221e-01 1.90521669e+00
-3.53666693e-01 3.47470522e-01 1.42794982e-01 -1.06945312e+00
7.12680519e-01 2.50536114e-01 8.98302019e-01 -6.20043099e-01
3.72787297e-01 2.21127748e-01 1.16608620e-01 -5.69836259e-01
-1.42582536e-01 -1.81929141e-01 4.26442951e-01 4.62411374e-01
2.83616185e-01 -1.07607953e-01 3.29712301e-01 -3.02423933e-03
7.12182224e-01 2.80801028e-01 1.32373273e-01 -2.35741019e-01
5.80234706e-01 2.79664308e-01 5.80767870e-01 4.71762657e-01
-3.91344726e-01 8.82783771e-01 5.19006789e-01 -5.58180749e-01
-8.92022073e-01 -6.98443234e-01 -5.25561452e-01 4.20170933e-01
2.64102649e-02 -1.60190225e-01 -1.21241701e+00 -1.18005991e+00
-2.10108116e-01 1.27411082e-01 -7.50564575e-01 1.92100376e-01
-1.66872978e-01 -1.13482022e+00 4.95951742e-01 3.69002253e-01
6.68176591e-01 -9.10570264e-01 -4.91561562e-01 6.82104975e-02
1.11468077e-01 -1.10307372e+00 -4.26643670e-01 1.92649186e-01
-1.04529965e+00 -1.33122230e+00 -1.04014087e+00 -1.00274920e+00
1.25282145e+00 -9.52021033e-02 8.89065087e-01 2.82900542e-01
-8.50544095e-01 1.95004046e-01 -1.22837774e-01 -2.41778776e-01
-4.09180731e-01 1.65054779e-02 -3.48129511e-01 2.91363329e-01
1.15232229e-01 -1.98650956e-01 -8.23578238e-01 3.71819556e-01
-1.02389169e+00 3.67230713e-01 7.14552462e-01 1.10417676e+00
1.09746027e+00 1.14900630e-03 3.45985860e-01 -1.40558815e+00
2.60396034e-01 -1.27884120e-01 -6.36302114e-01 3.70005757e-01
-7.14484692e-01 -1.00648507e-01 4.49512303e-01 -3.07845443e-01
-1.01973832e+00 6.47658408e-01 -2.28340596e-01 -3.21635783e-01
-3.33889693e-01 5.46193123e-01 2.73068637e-01 -4.27045971e-01
7.38459647e-01 3.68388519e-02 5.33699691e-01 -2.90312409e-01
3.49950671e-01 7.72824526e-01 3.74520332e-01 -3.08150798e-01
4.33468699e-01 7.37125218e-01 9.63388234e-02 -6.36230826e-01
-9.96484697e-01 -6.05675578e-01 -7.65235186e-01 -7.17916042e-02
1.13070703e+00 -7.11039364e-01 -4.35031563e-01 8.27616692e-01
-8.62286270e-01 -6.98409617e-01 -5.77568412e-01 4.25213784e-01
-3.54594469e-01 5.91710508e-01 -5.97141922e-01 -5.80204904e-01
-6.24296725e-01 -1.55080521e+00 9.41724300e-01 5.13993204e-01
2.73008440e-02 -1.24096584e+00 -2.43361201e-02 7.32434809e-01
3.12890619e-01 4.37211156e-01 8.32670033e-01 -7.79207647e-01
-3.70827466e-01 -1.65391386e-01 -4.45433140e-01 9.75140095e-01
2.89647967e-01 4.71886732e-02 -7.64989614e-01 -1.99199140e-01
-2.00013548e-01 -5.74477911e-01 1.19177330e+00 7.95253098e-01
1.56377947e+00 3.59830447e-02 -1.32071793e-01 1.06559145e+00
1.52682745e+00 2.28888299e-02 6.25164092e-01 -8.76383856e-02
8.23719323e-01 7.41975188e-01 3.36869985e-01 1.08697332e-01
2.79175907e-01 2.85507888e-01 4.68107581e-01 -1.06076968e+00
-3.58643621e-01 8.04965869e-02 -2.26579487e-01 3.73482317e-01
-4.05097753e-02 3.18794176e-02 -1.00686395e+00 5.97145557e-01
-1.77439594e+00 -2.15293765e-01 -1.49334714e-01 2.19334769e+00
1.15760028e+00 -1.05848340e-02 1.39798194e-01 -1.15311779e-01
6.78027093e-01 -2.17275843e-01 -7.62691259e-01 -1.83772326e-01
5.04628867e-02 5.45392036e-01 7.18586087e-01 3.82346153e-01
-1.43137181e+00 9.11587834e-01 5.37267256e+00 9.68159139e-01
-1.57697248e+00 1.00412443e-01 1.22231591e+00 1.43331304e-01
-5.36398701e-02 -4.77072895e-02 -5.22242427e-01 4.57574338e-01
5.29383719e-01 5.13616800e-01 2.12417036e-01 6.41646624e-01
1.51763141e-01 -4.04818326e-01 -8.99185956e-01 6.62379980e-01
-3.35524119e-02 -1.51920474e+00 2.10510436e-02 5.47527634e-02
1.00831330e+00 1.67655617e-01 1.47805035e-01 -1.11918658e-01
5.61766364e-02 -1.27582788e+00 4.02372144e-02 3.88873041e-01
1.11631620e+00 -3.15067172e-01 8.66931081e-01 2.41356537e-01
-6.78187490e-01 2.62460947e-01 -7.34223723e-02 3.89359444e-01
8.34392831e-02 7.61486351e-01 -1.02084696e+00 4.40008581e-01
4.42764521e-01 7.96037138e-01 -6.06545329e-01 1.18374944e+00
-1.80462450e-01 9.15405452e-01 -1.61006913e-01 3.26373309e-01
4.57558841e-01 -5.17596960e-01 2.60067195e-01 9.88474011e-01
-1.52866215e-01 -1.16718136e-01 1.19590618e-01 9.81683195e-01
-1.86932147e-01 3.32870334e-01 -8.68847817e-02 -1.14884980e-01
-1.89210102e-01 1.60428095e+00 -1.00015938e+00 -1.73180312e-01
-2.24861383e-01 8.07554364e-01 9.66929942e-02 3.69369686e-01
-7.38499105e-01 -1.19555682e-01 -8.66442546e-02 2.29534522e-01
-9.24598053e-02 1.64763898e-01 -6.48785174e-01 -8.61229122e-01
-2.16994882e-01 -5.96080363e-01 3.73767585e-01 -4.48529691e-01
-1.18271661e+00 6.86314702e-01 -2.69185066e-01 -1.26956165e+00
9.16673914e-02 -6.60632432e-01 -7.05205679e-01 8.77638102e-01
-1.97995842e+00 -1.42149329e+00 -4.04719204e-01 5.61485529e-01
1.80293202e-01 -2.77617276e-01 5.87673187e-01 2.84304082e-01
-7.59665966e-01 5.09675324e-01 -9.49807540e-02 3.19697678e-01
7.62356043e-01 -1.39921570e+00 8.38210732e-02 6.83964849e-01
3.53602655e-02 4.98570770e-01 3.49653512e-02 -5.76684773e-01
-9.28007066e-01 -1.39226544e+00 3.81256610e-01 1.41839981e-01
3.95998389e-01 1.59433991e-01 -8.33235800e-01 4.53079164e-01
1.51289955e-01 5.27385950e-01 8.80600512e-01 -1.95381612e-01
-1.43689457e-02 -1.10991538e-01 -1.56748092e+00 3.31356853e-01
4.90659833e-01 -1.64831907e-01 -1.72109321e-01 8.22420955e-01
6.04974210e-01 -7.11398125e-01 -9.11148131e-01 6.63477063e-01
3.70928556e-01 -7.79956639e-01 7.90120244e-01 -6.36875212e-01
5.74328721e-01 -3.03727835e-01 3.43528628e-01 -1.03920817e+00
1.72244340e-01 -7.61671484e-01 3.00043106e-01 9.40758109e-01
5.24593174e-01 -6.31829679e-01 1.16772366e+00 6.43979311e-01
-1.66245803e-01 -1.40001595e+00 -8.04379702e-01 -3.31975013e-01
2.48251662e-01 -8.49481747e-02 5.14438376e-02 7.49501228e-01
-4.01730508e-01 -4.49531525e-02 -1.58757702e-01 -1.27857924e-01
7.53960907e-01 -6.01429380e-02 5.37730455e-01 -1.08772647e+00
-3.60305041e-01 -2.49054387e-01 -2.51387805e-01 -7.74047732e-01
1.05288811e-01 -1.04834819e+00 -6.36252835e-02 -1.68148720e+00
3.89696449e-01 -6.40802145e-01 -1.71729729e-01 7.47862041e-01
-1.36278093e-01 5.34694731e-01 -2.63654683e-02 2.83490747e-01
-5.24022520e-01 7.75109082e-02 1.62586224e+00 -2.80518055e-01
-3.38404983e-01 1.97615519e-01 -5.86070180e-01 8.61361861e-01
8.33845496e-01 -2.93217897e-01 -3.14395458e-01 -2.86329985e-01
-1.23014778e-01 4.85046245e-02 4.89857048e-01 -8.26306403e-01
2.74720550e-01 4.44953367e-02 3.31564903e-01 -4.03133512e-01
-6.84987381e-02 -7.23524213e-01 -2.14930549e-01 6.50765896e-01
-4.06289786e-01 -7.41382420e-01 1.97496414e-01 3.29210371e-01
-3.40642005e-01 -2.48721942e-01 1.27120590e+00 -2.21461594e-01
-6.05080962e-01 6.19661987e-01 -3.34265642e-02 1.20161951e-01
1.17518461e+00 -3.38306248e-01 -3.64469975e-01 -4.67950739e-02
-9.69721854e-01 4.98772770e-01 3.31911087e-01 -7.10638613e-02
4.77291524e-01 -8.17307413e-01 -7.06044197e-01 2.71181852e-01
9.00121685e-03 5.61306119e-01 3.90591443e-01 1.34895086e+00
-7.54067421e-01 2.06504017e-01 -2.33036429e-01 -9.41203952e-01
-1.24293637e+00 1.15675628e-01 6.91990614e-01 -6.43675029e-01
-3.48451555e-01 9.84093368e-01 1.12997383e-01 -6.99762523e-01
3.22305322e-01 -5.27410924e-01 -9.40710008e-02 -2.40378410e-01
1.53198555e-01 1.81409006e-03 2.40619034e-01 -5.60188890e-01
-1.94907129e-01 7.04172790e-01 -3.45199734e-01 2.58511007e-01
1.14528859e+00 3.47441137e-02 -3.32698733e-01 -7.99781308e-02
1.28613794e+00 -3.75199318e-01 -1.47040200e+00 -3.97839606e-01
-2.08808869e-01 1.01450859e-02 4.04477715e-01 -1.26265860e+00
-1.44958425e+00 9.41697419e-01 9.17139888e-01 -1.54955685e-01
1.29783642e+00 -1.42202556e-01 8.98296177e-01 -2.63996590e-02
2.03782752e-01 -1.11947405e+00 -1.34417668e-01 1.98114380e-01
4.58223134e-01 -1.56805325e+00 1.17201835e-01 -7.16590047e-01
-8.86717796e-01 1.04547536e+00 5.92971683e-01 -1.30346581e-01
6.47862554e-01 2.71431834e-01 3.67616206e-01 -1.31063759e-01
-2.58629352e-01 -4.88232791e-01 6.97468221e-01 4.68196899e-01
5.11908770e-01 1.51098333e-03 -5.24814010e-01 5.55903554e-01
4.01131034e-01 2.38602966e-01 2.89462775e-01 6.75425053e-01
-1.75564885e-01 -1.24538612e+00 9.01580229e-02 7.26494670e-01
-7.38659382e-01 -7.78782964e-02 -4.25920576e-01 7.58841872e-01
3.43011439e-01 6.83035016e-01 5.79585023e-02 -7.86482170e-02
-1.74117982e-01 -1.54365018e-01 3.97686571e-01 -8.55827212e-01
-8.48548532e-01 3.13723952e-01 -2.00253844e-01 -4.91112024e-01
-7.66303778e-01 -3.17532986e-01 -1.58379972e+00 4.93485689e-01
-3.02427173e-01 -1.56905502e-02 8.75251114e-01 9.85596180e-01
2.53859907e-01 5.92886806e-01 6.89867973e-01 -5.56881011e-01
-6.16066456e-01 -8.19172323e-01 -6.89682186e-01 6.47060871e-01
1.41063675e-01 -3.13586533e-01 -3.02029759e-01 3.00819010e-01]
|
[14.647855758666992, -2.3577706813812256]
|
9eae4087-327f-40f8-966b-a506e83e4963
|
explicit-shape-encoding-for-real-time
|
1908.04067
| null |
https://arxiv.org/abs/1908.04067v1
|
https://arxiv.org/pdf/1908.04067v1.pdf
|
Explicit Shape Encoding for Real-Time Instance Segmentation
|
In this paper, we propose a novel top-down instance segmentation framework based on explicit shape encoding, named \textbf{ESE-Seg}. It largely reduces the computational consumption of the instance segmentation by explicitly decoding the multiple object shapes with tensor operations, thus performs the instance segmentation at almost the same speed as the object detection. ESE-Seg is based on a novel shape signature Inner-center Radius (IR), Chebyshev polynomial fitting and the strong modern object detectors. ESE-Seg with YOLOv3 outperforms the Mask R-CNN on Pascal VOC 2012 at mAP$^r$@0.5 while 7 times faster.
|
['Cewu Lu', 'Haiyang Wang', 'Wenqiang Xu', 'Fubo Qi']
|
2019-08-12
|
explicit-shape-encoding-for-real-time-1
|
http://openaccess.thecvf.com/content_ICCV_2019/html/Xu_Explicit_Shape_Encoding_for_Real-Time_Instance_Segmentation_ICCV_2019_paper.html
|
http://openaccess.thecvf.com/content_ICCV_2019/papers/Xu_Explicit_Shape_Encoding_for_Real-Time_Instance_Segmentation_ICCV_2019_paper.pdf
|
iccv-2019-10
|
['real-time-instance-segmentation']
|
['computer-vision']
|
[-1.13739066e-01 2.28617311e-01 1.68975338e-01 -3.70632291e-01
-8.31376851e-01 -7.70277798e-01 2.35448614e-01 1.15440693e-02
-4.91527855e-01 5.53756133e-02 -6.48299158e-01 -2.81782627e-01
7.70151392e-02 -7.06960857e-01 -9.92111862e-01 -5.05938530e-01
-1.04932729e-02 5.39271355e-01 7.82526255e-01 -1.33416191e-01
4.30290908e-01 7.13418305e-01 -1.29560649e+00 3.09674114e-01
7.89158583e-01 1.71741974e+00 7.24897087e-02 8.72158289e-01
-2.48640567e-01 1.47681534e-01 -5.36572039e-01 -9.84439790e-01
6.10686660e-01 3.78275871e-01 -7.40183234e-01 1.83942810e-01
9.59055066e-01 -2.70949543e-01 -1.24718271e-01 1.10572982e+00
3.17407906e-01 1.53601155e-01 6.66637421e-01 -1.01235032e+00
-8.44414532e-01 4.79803622e-01 -8.20168555e-01 1.95941523e-01
-2.85306096e-01 2.77846098e-01 9.31715906e-01 -1.38314867e+00
4.54601169e-01 1.21471024e+00 9.35415208e-01 3.12033743e-01
-1.12772334e+00 -3.84820521e-01 2.81545669e-01 1.79893866e-01
-1.51228178e+00 1.15852796e-01 4.48288292e-01 -4.02381390e-01
1.09276521e+00 4.97707307e-01 4.15456533e-01 4.00570601e-01
-1.71774626e-01 1.08857763e+00 9.52769041e-01 2.58591861e-01
3.48949842e-02 -1.09290682e-01 5.01268208e-01 1.01513445e+00
3.84180874e-01 -5.33377714e-02 1.91854965e-02 2.76175618e-01
7.76249111e-01 -2.33655453e-01 2.24765111e-02 -1.81359097e-01
-1.13537717e+00 4.37837720e-01 8.99363518e-01 -3.54909748e-02
-1.39190137e-01 4.88228798e-01 4.65026170e-01 -1.78628579e-01
3.89194667e-01 4.53128129e-01 -6.67971075e-01 2.06195876e-01
-1.07158220e+00 2.30533302e-01 7.02509880e-01 1.46235561e+00
7.32497692e-01 2.66342461e-01 -4.05133456e-01 7.23772049e-01
2.40794733e-01 7.26431608e-01 -1.49666235e-01 -1.05254376e+00
4.62744236e-01 8.19902778e-01 -6.41051866e-03 -9.77571607e-01
-3.99534315e-01 -7.46790171e-01 -6.20228350e-01 5.74769415e-02
4.74213302e-01 9.43863317e-02 -1.42782068e+00 8.46718431e-01
6.09280407e-01 2.22265482e-01 -2.64094025e-01 9.11560297e-01
1.34433961e+00 6.28285468e-01 -5.37560731e-02 3.92149568e-01
1.57344556e+00 -1.35755670e+00 -3.09230447e-01 -1.29989579e-01
4.67848092e-01 -7.19817519e-01 7.31965423e-01 5.21209359e-01
-1.00791192e+00 -7.54264176e-01 -1.00616431e+00 -2.84638882e-01
-5.62663496e-01 6.70119524e-01 6.70105696e-01 6.97609901e-01
-9.59791839e-01 6.68188155e-01 -7.48690605e-01 2.56968915e-01
7.88487017e-01 6.48651123e-01 -8.88194591e-02 8.20503756e-02
-4.06677753e-01 3.99946362e-01 6.57952726e-01 5.34787059e-01
-7.64307737e-01 -8.24993193e-01 -9.71586108e-01 -1.25932738e-01
6.72954023e-01 -3.96917313e-01 9.50329483e-01 -4.76344883e-01
-1.46324289e+00 1.08829081e+00 -7.75209116e-03 -7.81676352e-01
6.35149121e-01 -4.06808496e-01 -4.31314930e-02 2.96467543e-01
4.50548157e-02 1.04280412e+00 1.01370549e+00 -1.27118945e+00
-5.34340799e-01 -4.96146023e-01 -2.09718924e-02 -1.77679166e-01
2.45171279e-01 -5.12912571e-02 -9.58224595e-01 -6.36681080e-01
3.96419168e-01 -7.98637450e-01 -3.41611147e-01 -1.12450078e-01
-6.89339817e-01 -4.92145777e-01 1.10032952e+00 -6.49238586e-01
9.96821642e-01 -2.09036350e+00 -6.21925890e-02 1.48722827e-01
3.48200321e-01 6.35116220e-01 -1.22410946e-01 -1.84290126e-01
2.27699533e-01 3.04488420e-01 -7.84758806e-01 -4.33030784e-01
1.02441020e-01 1.76148742e-01 -2.61719704e-01 5.70417225e-01
5.88614166e-01 1.32662582e+00 -6.16332293e-01 -6.85053229e-01
2.55668938e-01 3.14313650e-01 -5.35494328e-01 2.45903004e-02
-4.68772829e-01 4.82571870e-02 -4.29546177e-01 9.85798419e-01
1.26566505e+00 -2.35984996e-01 -4.17820066e-01 -4.83368546e-01
-3.97721350e-01 -8.62937346e-02 -1.31431675e+00 1.43312430e+00
1.35046303e-01 4.48419243e-01 4.73239779e-01 -1.03910327e+00
1.09147263e+00 -1.20364234e-01 4.85654175e-01 -5.69520056e-01
2.59105474e-01 4.00860459e-01 -1.21693738e-01 -3.07525933e-01
5.37940323e-01 4.41558599e-01 9.12669674e-02 -1.93285108e-01
7.92586505e-02 -6.39228284e-01 4.41715002e-01 1.53679118e-01
6.56046152e-01 5.55589736e-01 -1.28685594e-01 -4.58874255e-01
5.85622311e-01 2.12031946e-01 6.85301244e-01 7.88016319e-01
-4.04463232e-01 8.40497792e-01 3.10062468e-01 -4.55341667e-01
-9.35937464e-01 -9.22363877e-01 -6.98412538e-01 9.72676277e-01
4.71923888e-01 -4.08354491e-01 -1.15739262e+00 -7.89711475e-01
2.42958158e-01 5.49263239e-01 -6.09514713e-01 4.30804461e-01
-9.15497780e-01 -6.84694290e-01 7.83389211e-01 8.59968185e-01
8.57948184e-01 -9.48157787e-01 -3.97424340e-01 1.33914918e-01
2.14681968e-01 -1.54829490e+00 -6.61955893e-01 1.14267193e-01
-1.10801935e+00 -1.12454379e+00 -7.92905688e-01 -7.18402267e-01
7.24231482e-01 1.07528074e-02 1.02724075e+00 1.29585788e-01
-6.96789682e-01 3.14072490e-01 -2.47912958e-01 -4.72024530e-01
1.78538933e-01 -7.81389549e-02 -4.62098092e-01 1.60661757e-01
1.79322064e-01 1.07968725e-01 -7.81445503e-01 4.83507127e-01
-8.56924713e-01 -1.17693007e-01 5.45529068e-01 5.74854136e-01
1.15517557e+00 -2.32624516e-01 1.39663085e-01 -8.16627681e-01
-1.12615786e-01 2.86775175e-02 -1.02051651e+00 1.24914981e-01
-4.18584526e-01 -1.37056798e-01 4.98438478e-01 -1.39267653e-01
-7.47489154e-01 3.04508269e-01 -3.53545189e-01 -6.51883602e-01
-1.64426327e-01 -1.62172720e-01 -1.29675772e-02 -3.12006325e-01
3.66070092e-01 3.41520846e-01 -4.60192800e-01 -6.59915388e-01
6.04201376e-01 4.33457732e-01 8.75710487e-01 -6.82425618e-01
7.78369844e-01 6.86206162e-01 1.14534453e-01 -9.74660397e-01
-1.01116359e+00 -7.63871729e-01 -9.20671940e-01 -5.52585796e-02
1.25012839e+00 -6.33303881e-01 -9.36052263e-01 7.85148144e-01
-1.34658396e+00 -2.59636253e-01 -2.53221720e-01 4.22619656e-02
-5.04863083e-01 4.19559419e-01 -7.58880973e-01 -8.87116849e-01
-5.27558565e-01 -1.40284371e+00 1.49585009e+00 2.39266008e-01
6.83412910e-01 -5.19575655e-01 -6.86520934e-01 7.75873125e-01
8.56801346e-02 3.89254451e-01 4.88487810e-01 -6.24466240e-01
-1.17803025e+00 -8.69814754e-02 -9.29967642e-01 5.92447281e-01
-4.71472591e-01 2.27654040e-01 -9.63686824e-01 -1.02222569e-01
-1.13815822e-01 6.98818415e-02 1.19097948e+00 5.54639816e-01
1.71183527e+00 -1.53193444e-01 -1.43468827e-01 1.12981033e+00
1.44764984e+00 1.60466269e-01 6.31696105e-01 1.28169984e-01
1.16940737e+00 4.18060064e-01 7.23776639e-01 1.11757636e-01
5.28830826e-01 5.10642111e-01 7.95989990e-01 -1.42443046e-01
-3.44388306e-01 1.08904108e-01 1.48019955e-01 6.60544693e-01
-1.79858372e-01 1.46322936e-01 -9.67234373e-01 6.38928413e-01
-1.64610767e+00 -4.64888364e-01 -7.76022375e-01 1.87821734e+00
5.43904424e-01 3.25600028e-01 1.10414430e-01 -1.38219684e-01
7.43277133e-01 -3.12544964e-02 -6.25291526e-01 -4.43319857e-01
-1.34828374e-01 4.06967402e-01 1.15158570e+00 2.76951134e-01
-1.46208894e+00 1.37918365e+00 6.08510780e+00 1.00021839e+00
-9.42578852e-01 7.29386136e-02 8.58478010e-01 4.32676494e-01
1.78713590e-01 -1.08381726e-01 -1.27080917e+00 2.55553037e-01
4.60755855e-01 4.13493425e-01 2.40803480e-01 1.13908696e+00
-2.28946209e-01 -1.97024131e-03 -9.40035224e-01 1.09229589e+00
7.32373595e-02 -1.37147403e+00 -9.74693000e-02 -1.64535761e-01
7.19972849e-01 3.52104753e-01 1.10383488e-01 4.47689384e-01
-6.54947981e-02 -1.16333640e+00 1.02703583e+00 3.03221017e-01
8.87405634e-01 -5.35539627e-01 6.21726811e-01 3.10296535e-01
-1.73664474e+00 2.14873329e-02 -5.62447071e-01 6.94724262e-01
-3.92354280e-02 5.39099872e-01 -9.01585281e-01 5.69460094e-01
8.18615317e-01 5.38475156e-01 -7.77972102e-01 1.29057670e+00
-1.19495220e-01 6.78840637e-01 -6.12654984e-01 9.84283909e-02
5.81478536e-01 -4.20555115e-01 8.35694313e-01 1.83241498e+00
-5.33999084e-03 3.66919160e-01 3.66855919e-01 1.33522618e+00
-1.87503904e-01 1.23720011e-02 -1.15311019e-01 -3.06856986e-02
1.20140620e-01 1.46759164e+00 -1.24338818e+00 -5.16827047e-01
-3.30503099e-02 8.88476431e-01 2.85209324e-02 2.57715553e-01
-1.04296732e+00 -4.76999015e-01 3.37183535e-01 8.52641612e-02
9.65718389e-01 -5.47452509e-01 -7.25066006e-01 -7.99978912e-01
3.95809337e-02 -5.71102262e-01 1.92835674e-01 -5.45016646e-01
-1.05338645e+00 4.92679417e-01 -1.59811392e-01 -8.27080131e-01
5.40422022e-01 -1.34153867e+00 -5.23291409e-01 4.63633120e-01
-1.52962351e+00 -1.21500862e+00 -1.26035243e-01 3.55295897e-01
6.61051989e-01 2.18343228e-01 2.83846140e-01 2.88507491e-01
-9.26850438e-01 6.51935399e-01 -2.93979257e-01 5.59691012e-01
-3.07009965e-02 -1.77930391e+00 8.39154601e-01 7.07565904e-01
1.77320257e-01 5.56110919e-01 3.71015608e-01 -6.94197714e-01
-1.51082289e+00 -1.35156286e+00 1.65257782e-01 -4.49368775e-01
5.22650063e-01 -2.74600774e-01 -9.14245129e-01 4.84965593e-01
-3.85715216e-02 6.28379941e-01 6.75725415e-02 -4.10025179e-01
-6.12874627e-01 -1.78960800e-01 -1.36905777e+00 3.41395020e-01
1.16296566e+00 -2.47590080e-01 -4.73503619e-01 3.67752463e-01
1.09773731e+00 -7.65644550e-01 -1.02983987e+00 8.22581291e-01
1.24369279e-01 -8.52126241e-01 1.27123582e+00 -4.08793390e-01
9.38209668e-02 -6.23665154e-01 -1.52543455e-01 -6.22004271e-01
-1.40768275e-01 -6.76762342e-01 -2.50622302e-01 9.42668080e-01
4.81405467e-01 -5.20913541e-01 8.24325562e-01 4.16888237e-01
-6.66479528e-01 -1.17740667e+00 -1.04496336e+00 -9.56692398e-01
-2.42868531e-02 -9.60544884e-01 5.91960192e-01 3.40301156e-01
-1.04255700e+00 -2.19844386e-01 3.49109858e-01 3.71268392e-01
9.46021676e-01 3.64628464e-01 6.84858978e-01 -1.07702386e+00
-3.76763284e-01 -7.10955501e-01 -4.46090549e-01 -1.57466221e+00
-2.40526974e-01 -1.08911872e+00 1.12787582e-01 -1.56376195e+00
-1.26205817e-01 -4.61635143e-01 -2.70303249e-01 4.95210856e-01
-9.89495963e-02 7.22944140e-01 3.52376610e-01 -5.39869331e-02
-9.33858931e-01 3.30672890e-01 1.67063904e+00 -2.56418377e-01
3.97593640e-02 1.40014648e-01 -3.18285346e-01 1.12836587e+00
4.30123299e-01 -4.12961423e-01 2.57892728e-01 -3.61816108e-01
1.12737954e-01 -2.72654265e-01 5.35619259e-01 -9.61131513e-01
7.58806467e-02 5.06428406e-02 3.25040072e-01 -1.31795883e+00
4.50829118e-01 -6.26413167e-01 -4.82813150e-01 3.21994603e-01
1.46045923e-01 -2.87696421e-02 5.05770922e-01 5.24517655e-01
-6.11135997e-02 -4.08673495e-01 9.86027479e-01 -2.29774594e-01
-8.78182173e-01 5.73653936e-01 2.62908161e-01 3.09508979e-01
1.06870985e+00 -6.25521362e-01 -2.80298471e-01 5.53850114e-01
-9.09462452e-01 3.80396992e-01 5.74431345e-02 2.37439319e-01
8.46170664e-01 -1.07553482e+00 -6.10354960e-01 1.10453129e-01
-1.79556832e-02 6.60163105e-01 1.17397889e-01 1.03385353e+00
-9.61340010e-01 3.89266342e-01 3.57997894e-01 -1.02573013e+00
-1.20288968e+00 3.60057086e-01 5.43310940e-01 -5.90571165e-02
-9.22522843e-01 1.47730124e+00 4.33527023e-01 -5.68416238e-01
3.01155835e-01 -6.27542019e-01 -5.55639043e-02 -3.42609361e-02
3.06752086e-01 7.04721034e-01 2.26436228e-01 -7.24138260e-01
-3.94575298e-01 7.71356285e-01 1.07750103e-01 1.85817286e-01
1.32912636e+00 2.73113281e-01 -3.94672960e-01 5.64727373e-02
1.05148995e+00 -5.43045625e-03 -1.55014837e+00 -3.61812152e-02
2.79106557e-01 -3.67263973e-01 5.22662401e-02 -7.99248993e-01
-1.31804192e+00 7.95122325e-01 5.43096542e-01 2.33906627e-01
7.58930385e-01 1.20118130e-02 1.09396708e+00 5.92868507e-01
3.34461719e-01 -1.33358538e+00 8.50219354e-02 7.57983088e-01
1.04746664e+00 -1.05980265e+00 4.92052287e-02 -8.56225848e-01
-4.58026290e-01 1.18401706e+00 7.40981400e-01 -4.81265515e-01
7.41316736e-01 4.18973207e-01 -1.86708614e-01 -4.72399712e-01
-9.18458700e-02 -5.06839335e-01 1.00332689e+00 2.59720087e-01
2.14527100e-01 3.85936618e-01 -2.44433954e-01 5.83499134e-01
-2.36201212e-01 -4.73522574e-01 1.58278197e-01 3.96120429e-01
-6.10339463e-01 -7.84320474e-01 -5.95284522e-01 5.29735923e-01
-7.03551888e-01 -9.28666592e-02 -4.25595760e-01 7.59852707e-01
5.76064169e-01 6.19634688e-01 1.93007633e-01 -2.22108677e-01
4.77426887e-01 -1.42093942e-01 4.22963649e-01 -7.59670436e-01
-8.94787490e-01 1.38733447e-01 -1.40826926e-01 -7.89779067e-01
-9.28979665e-02 -5.53749561e-01 -1.96637261e+00 7.28340214e-03
-4.91200924e-01 -2.91611522e-01 9.47991550e-01 8.29784334e-01
4.10522968e-01 4.54243153e-01 2.94677675e-01 -8.83978665e-01
-5.81117332e-01 -7.54405022e-01 -3.51346731e-01 3.43852580e-01
9.92579386e-02 -5.21504402e-01 -2.85384208e-01 -6.07542060e-02]
|
[9.49328899383545, 0.0800282210111618]
|
53d43ba4-3588-403a-ad6d-de7781ab99d4
|
enhancing-keyphrase-extraction-from-academic
|
2111.14106
| null |
https://arxiv.org/abs/2111.14106v2
|
https://arxiv.org/pdf/2111.14106v2.pdf
|
Enhancing Keyphrase Extraction from Academic Articles with their Reference Information
|
With the development of Internet technology, the phenomenon of information overload is becoming more and more obvious. It takes a lot of time for users to obtain the information they need. However, keyphrases that summarize document information highly are helpful for users to quickly obtain and understand documents. For academic resources, most existing studies extract keyphrases through the title and abstract of papers. We find that title information in references also contains author-assigned keyphrases. Therefore, this article uses reference information and applies two typical methods of unsupervised extraction methods (TF*IDF and TextRank), two representative traditional supervised learning algorithms (Na\"ive Bayes and Conditional Random Field) and a supervised deep learning model (BiLSTM-CRF), to analyze the specific performance of reference information on keyphrase extraction. It is expected to improve the quality of keyphrase recognition from the perspective of expanding the source text. The experimental results show that reference information can increase precision, recall, and F1 of automatic keyphrase extraction to a certain extent. This indicates the usefulness of reference information on keyphrase extraction of academic papers and provides a new idea for the following research on automatic keyphrase extraction.
|
['Yingyi Zhang', 'Mengyuan Zhao', 'Lei Zhao', 'Chengzhi Zhang']
|
2021-11-28
| null | null | null | null |
['keyphrase-extraction']
|
['natural-language-processing']
|
[-1.16819784e-01 -2.88882345e-01 -7.09668994e-01 1.16529949e-01
-5.98837435e-01 -5.54561138e-01 9.10092592e-01 8.03606570e-01
-6.31944537e-01 1.00540924e+00 4.52980429e-01 -4.71643269e-01
-3.30559134e-01 -9.79739666e-01 -3.74038815e-01 -4.63270158e-01
1.98874816e-01 7.71398023e-02 1.57385692e-01 6.80639446e-02
8.79090250e-01 7.56826520e-01 -1.41709590e+00 1.96172863e-01
9.40786183e-01 8.99139464e-01 4.71686125e-01 4.11878914e-01
-8.59611332e-01 9.43975389e-01 -9.85657096e-01 -2.38935232e-01
-1.76265791e-01 -6.41725957e-02 -9.56506908e-01 -1.27172500e-01
3.21524262e-01 -4.63179618e-01 -6.54459655e-01 1.10264468e+00
2.86919236e-01 3.20362952e-03 7.24836886e-01 -1.08923268e+00
-7.57205307e-01 1.02492726e+00 -7.56056428e-01 6.40542030e-01
3.26490253e-01 -5.93138516e-01 1.17921638e+00 -7.59854674e-01
6.40710711e-01 9.92238522e-01 2.03625500e-01 -1.85232788e-01
-3.90069544e-01 -8.78857851e-01 3.64045836e-02 2.19965160e-01
-1.45527673e+00 -1.07659828e-02 9.04534042e-01 -4.00891602e-01
8.28105748e-01 2.02459842e-01 5.38740516e-01 9.19407845e-01
4.97859120e-01 1.20239270e+00 9.62324619e-01 -7.57777691e-01
-1.46254703e-01 3.63473177e-01 7.00412691e-01 3.41808498e-01
6.12851739e-01 -1.31943420e-01 -6.99888051e-01 -1.80192083e-01
4.26112890e-01 3.60722870e-01 -2.30851904e-01 3.79541963e-01
-1.31039906e+00 6.46429837e-01 -9.70001891e-03 7.24258482e-01
-4.90332007e-01 -3.91219765e-01 4.93030041e-01 1.43052652e-01
2.76657909e-01 7.67401040e-01 -6.72870815e-01 -1.37519836e-01
-1.29584885e+00 5.50466955e-01 8.76208067e-01 1.10785508e+00
4.69495833e-01 -1.65542513e-01 -5.74918807e-01 5.18310905e-01
2.32897162e-01 6.58716798e-01 6.62442744e-01 -6.63721085e-01
4.62625355e-01 9.53571677e-01 1.44499943e-01 -1.47234678e+00
-2.85450339e-01 -7.62919426e-01 -9.01584268e-01 -7.46258378e-01
3.58936042e-02 -1.14768185e-01 -6.31083190e-01 1.04314053e+00
-9.94190723e-02 -3.65833670e-01 -1.79678947e-01 3.42766017e-01
1.40515435e+00 9.42203760e-01 2.31955349e-01 -5.69635093e-01
1.57070386e+00 -6.57792270e-01 -1.32042134e+00 2.29527295e-01
4.14569676e-01 -1.30637586e+00 6.01752579e-01 5.49724817e-01
-6.13339603e-01 -5.40660441e-01 -9.82008576e-01 -1.42522275e-01
-9.51786995e-01 6.43480539e-01 7.15924501e-01 3.03212494e-01
-4.17935550e-01 6.50491178e-01 -3.51616114e-01 -3.57888453e-02
2.56853282e-01 5.96820414e-02 -5.90745173e-02 3.48087661e-02
-1.53527510e+00 6.88656330e-01 6.13460600e-01 -2.19786420e-01
-1.58632234e-01 -8.17948699e-01 -4.87239122e-01 2.70608604e-01
6.24144912e-01 -3.47550750e-01 1.25456607e+00 -3.70262295e-01
-8.78753483e-01 5.19195437e-01 -1.65852636e-01 -2.29640335e-01
-5.52283898e-02 -7.33939111e-01 -6.57746077e-01 2.84420997e-01
2.93418109e-01 1.94390178e-01 7.79590309e-01 -7.90192544e-01
-1.02372146e+00 -2.64275312e-01 -3.01651239e-01 -6.03065342e-02
-9.06221986e-01 3.38904023e-01 -4.87477243e-01 -1.03418827e+00
-1.23262800e-01 -4.28467482e-01 1.80434838e-01 -7.34286368e-01
-6.87141240e-01 -8.76024485e-01 9.48961794e-01 -1.03543603e+00
1.98516929e+00 -1.67619407e+00 -2.56625921e-01 2.56178856e-01
5.05113065e-01 3.48732799e-01 3.08836490e-01 7.80672967e-01
7.47253075e-02 5.88528514e-01 2.70152628e-01 5.18262267e-01
-8.76310244e-02 -1.22914329e-01 -7.83996820e-01 -1.63073093e-01
-3.54867160e-01 8.39295805e-01 -8.50268304e-01 -9.26446855e-01
1.15012527e-02 1.70124277e-01 1.96010202e-01 2.67462611e-01
7.55044119e-03 -1.16406433e-01 -9.92756784e-01 6.18952751e-01
3.70027661e-01 -3.33017677e-01 -3.11870426e-01 -4.69553620e-01
-4.84005421e-01 5.36919475e-01 -1.14577937e+00 1.03502464e+00
-2.15533242e-01 8.44393551e-01 -4.88429159e-01 -8.27409744e-01
9.19588447e-01 3.97466511e-01 7.59394765e-01 -4.65459794e-01
2.87095964e-01 2.13822052e-01 -3.71500611e-01 -4.81030017e-01
9.18303847e-01 5.72439492e-01 -1.33293852e-01 4.02737349e-01
1.00364469e-01 1.47946030e-01 6.94484591e-01 7.51808882e-01
7.21553564e-01 -9.29203629e-02 5.03533721e-01 -4.06185091e-01
6.36881113e-01 -3.98423746e-02 2.80536354e-01 8.73521686e-01
3.26269537e-01 -4.05545644e-02 3.36378336e-01 -3.28485191e-01
-8.96641016e-01 -4.94770944e-01 -1.36868760e-01 8.34038198e-01
-1.98865011e-01 -9.11881745e-01 -6.07558191e-01 -5.80466747e-01
1.77641630e-01 6.41598344e-01 -1.80140600e-01 -1.58091728e-02
-4.40143436e-01 -6.08202696e-01 4.09874380e-01 3.55198205e-01
6.60976887e-01 -1.03520548e+00 -2.96821564e-01 2.01758549e-01
-3.24300677e-01 -1.16983092e+00 -4.03851539e-01 1.18771307e-01
-6.97029829e-01 -9.72931564e-01 -1.11290014e+00 -6.76149011e-01
6.06241047e-01 4.77582157e-01 1.17228067e+00 4.07845601e-02
-7.69557729e-02 3.29455405e-01 -6.45839095e-01 -5.99564910e-01
-6.51911870e-02 6.77789450e-01 3.46188210e-02 -4.73853409e-01
9.17744577e-01 -3.09123635e-01 -2.79607385e-01 -2.65097558e-01
-8.04267466e-01 -8.06596354e-02 8.52970123e-01 5.50821245e-01
3.48516494e-01 7.00361490e-01 4.81793284e-01 -6.77223682e-01
1.22615814e+00 -2.49753192e-01 -6.18617058e-01 3.98769945e-01
-1.10723448e+00 2.92274266e-01 6.62031174e-01 -3.51366132e-01
-1.03587949e+00 -5.18401504e-01 2.25742757e-01 -1.44538194e-01
-1.55889839e-01 9.40042496e-01 -1.09414481e-01 2.88871169e-01
2.78121889e-01 6.26596153e-01 -6.98891938e-01 -8.81448984e-01
2.36637592e-01 1.25440323e+00 3.31101328e-01 -6.16773248e-01
8.25422645e-01 -1.96374983e-01 7.34860525e-02 -1.09431040e+00
-1.05091619e+00 -8.32930207e-01 -6.86619520e-01 -2.16507062e-01
5.72001755e-01 -8.20321441e-01 -7.46233463e-01 3.58758897e-01
-1.28918898e+00 6.56763673e-01 8.63012895e-02 7.06584156e-01
3.66731882e-01 7.12757885e-01 -6.30410790e-01 -5.47693968e-01
-6.97023034e-01 -7.33180761e-01 8.93250108e-01 6.57182515e-01
-3.95778954e-01 -6.53486431e-01 -1.23764284e-01 2.05465853e-01
6.00193925e-02 -1.52583376e-01 1.05849910e+00 -1.05161321e+00
-5.83136737e-01 -5.91053724e-01 -4.42552924e-01 2.09560260e-01
4.85384881e-01 2.16480076e-01 -5.67486763e-01 6.37998208e-02
-1.69506699e-01 1.77402794e-01 1.09538209e+00 2.50119179e-01
1.43467677e+00 -7.45270610e-01 -6.12010181e-01 6.36806190e-02
1.11520720e+00 3.51003855e-01 2.29367629e-01 5.59944630e-01
9.03880537e-01 5.09120762e-01 5.71983218e-01 4.38129723e-01
3.66604418e-01 3.05941463e-01 -3.71664733e-01 1.63543150e-01
1.30587563e-01 -4.60892290e-01 -1.56950783e-02 1.35755634e+00
-1.48395121e-01 -1.64732307e-01 -8.96087766e-01 5.59342623e-01
-1.45281756e+00 -9.54302073e-01 -2.57300168e-01 1.87698865e+00
1.34817219e+00 3.22704077e-01 -7.73232728e-02 5.39841473e-01
5.41279435e-01 1.79903939e-01 5.51825315e-02 -1.06397986e-01
7.92525411e-02 1.84684962e-01 7.46264100e-01 1.10316589e-01
-1.14496946e+00 1.03163683e+00 5.60476351e+00 1.18381202e+00
-8.63292396e-01 -4.82619792e-01 5.37046254e-01 5.57921112e-01
-2.50845134e-01 2.65964326e-02 -1.52842069e+00 5.56526005e-01
8.61538112e-01 -5.66377342e-01 3.47533673e-02 1.04503167e+00
1.45275980e-01 -4.27980423e-01 -8.39973569e-01 9.84461665e-01
-6.99479729e-02 -1.33052814e+00 4.38195616e-01 5.14079183e-02
6.68816924e-01 -5.17157495e-01 -2.16540411e-01 2.42080107e-01
2.82098323e-01 -5.99743962e-01 1.77216530e-01 6.83713734e-01
1.85510233e-01 -1.01411057e+00 8.59697580e-01 4.95510608e-01
-1.11873472e+00 1.26977459e-01 -3.54645699e-01 -1.19626194e-01
-3.34030211e-01 1.16631246e+00 -6.11313760e-01 6.39642894e-01
7.25681901e-01 7.90681779e-01 -7.57184267e-01 1.00822818e+00
-4.65210289e-01 6.40193105e-01 -1.27008691e-01 -5.57075202e-01
1.75042361e-01 1.39614595e-02 3.77841115e-01 1.45849895e+00
2.77943999e-01 1.12381987e-01 3.59578192e-01 6.57881260e-01
-3.10664892e-01 5.36719859e-01 -4.20380324e-01 -6.82484448e-01
7.65745938e-01 1.40944266e+00 -1.21113467e+00 -7.15746880e-01
-3.84690434e-01 2.66500264e-01 -2.33679742e-01 3.39929849e-01
-2.62316436e-01 -1.07160032e+00 -9.14177150e-02 3.13448682e-02
1.54745564e-01 -5.51776767e-01 -3.22749197e-01 -1.12368774e+00
2.14515980e-02 -9.71259892e-01 3.52191836e-01 -6.43477380e-01
-1.14703071e+00 2.76031584e-01 2.69806892e-01 -1.01276731e+00
-1.80506289e-01 -6.23920858e-01 -2.63048768e-01 8.38450372e-01
-1.46166635e+00 -8.34746301e-01 -1.73498914e-01 2.18840241e-01
5.81213415e-01 -4.55692202e-01 6.58901155e-01 2.02820554e-01
-5.60716987e-01 2.54096538e-01 2.49721333e-01 6.90389633e-01
6.92692399e-01 -1.20033598e+00 2.12899521e-01 8.91986609e-01
4.75707799e-01 1.12147319e+00 5.92166603e-01 -1.08118033e+00
-1.44402921e+00 -5.86251616e-01 1.53622985e+00 -4.06191975e-01
6.33861184e-01 1.47775352e-01 -8.61650288e-01 2.05469891e-01
5.72808087e-01 -6.31814003e-01 7.87670493e-01 1.65347025e-01
-2.16359854e-01 -4.25436109e-01 -5.46456933e-01 6.72103822e-01
2.83268332e-01 -5.67140400e-01 -1.07257664e+00 3.02316040e-01
9.21660483e-01 4.52692360e-02 -9.80134428e-01 4.37432051e-01
3.97637814e-01 -1.56227872e-01 9.40847695e-01 -4.27615970e-01
5.12221158e-01 -1.71050981e-01 1.48037538e-01 -9.74830270e-01
-2.37337410e-01 -6.73896909e-01 -7.18311965e-01 1.58621764e+00
2.25790992e-01 -1.17222965e-01 3.01310718e-01 5.14978349e-01
4.47258502e-01 -5.25833726e-01 -3.35244417e-01 -5.77620685e-01
-7.14929216e-03 -1.05047353e-01 6.41744614e-01 1.11809063e+00
2.41697162e-01 7.47106850e-01 -2.30476499e-01 -3.05000216e-01
5.55123568e-01 4.30524886e-01 5.57841957e-01 -1.64605141e+00
2.68821418e-01 -6.37576818e-01 2.92356819e-01 -1.15111768e+00
1.44376261e-02 -6.93215609e-01 -4.85003740e-01 -1.72381783e+00
3.62357080e-01 -2.02765632e-02 -5.36654949e-01 4.39352214e-01
-3.83102328e-01 -5.98315835e-01 -1.53154895e-01 6.39942467e-01
-5.52050412e-01 3.12619328e-01 1.44139051e+00 -3.30799520e-01
-3.10038120e-01 2.33000919e-01 -9.17866826e-01 6.66975796e-01
6.08309209e-01 -5.94184995e-01 -1.63643226e-01 1.00721709e-01
5.36275446e-01 -1.13124423e-01 -2.42492314e-02 -6.25861108e-01
5.32591164e-01 -4.16367888e-01 7.80943871e-01 -1.05682695e+00
-4.37932938e-01 -7.81403780e-01 -6.28696740e-01 2.36265719e-01
-5.55371642e-01 1.55729815e-01 1.71930701e-01 1.93195447e-01
-3.29893440e-01 -5.07160127e-01 7.50857294e-02 -3.76203418e-01
-6.74338341e-01 1.97577327e-01 -7.75344789e-01 1.00908373e-02
1.78583875e-01 3.08260441e-01 -1.16863832e-01 -1.30391225e-01
-4.74104360e-02 3.87972035e-02 -3.33995104e-01 6.89230561e-01
6.51600897e-01 -1.17984521e+00 -6.18420362e-01 -1.25325009e-01
-1.16347395e-01 -2.00085044e-01 -2.73805380e-01 5.31233966e-01
-2.65361518e-01 1.04938996e+00 -4.20561619e-02 -1.06690854e-01
-1.22375262e+00 6.73153579e-01 -5.23878396e-01 -6.51549578e-01
-4.76303607e-01 5.45842350e-01 -3.25283170e-01 8.75129104e-02
5.98549962e-01 -5.51350296e-01 -1.01657367e+00 6.24838114e-01
8.29185367e-01 4.22408998e-01 1.17348328e-01 -3.78148675e-01
-3.61381806e-02 3.96895319e-01 -7.46115923e-01 3.96130420e-02
1.11654508e+00 2.42743175e-02 -5.91999054e-01 3.26153129e-01
1.15109181e+00 3.62148851e-01 -2.91911304e-01 -5.91894209e-01
6.77154779e-01 -2.00993076e-01 7.04247892e-01 -7.39577711e-01
-8.03853631e-01 7.55330026e-01 1.51937127e-01 3.46240431e-01
1.05377829e+00 -1.57372534e-01 9.00012136e-01 9.00510073e-01
1.33248299e-01 -1.39574862e+00 -1.21708542e-01 5.99361897e-01
4.92092788e-01 -9.42574203e-01 8.05669904e-01 -3.11602920e-01
1.03086960e-02 1.44041085e+00 4.29247975e-01 3.18552732e-01
9.44487691e-01 3.23421121e-01 -1.88398212e-01 -1.23511225e-01
-2.89482951e-01 -8.83490443e-02 8.18159878e-01 1.28387928e-03
7.28899717e-01 -1.88589901e-01 -7.14302599e-01 1.06539726e+00
-3.43924761e-01 6.50257319e-02 4.08903003e-01 9.28361356e-01
-8.36238980e-01 -1.23135984e+00 -3.74351233e-01 7.85396457e-01
-1.15695643e+00 -3.90492767e-01 -6.13247037e-01 6.20877922e-01
-5.81680275e-02 9.10805821e-01 -3.04800630e-01 -2.95585304e-01
-9.31240693e-02 1.59797728e-01 2.18712807e-01 -4.67661113e-01
-3.79090816e-01 2.42675915e-01 -1.82905182e-01 1.12419911e-01
-5.55193245e-01 -3.26272935e-01 -1.05805004e+00 -2.80335158e-01
-6.06413603e-01 7.11903989e-01 6.19754434e-01 1.18354666e+00
2.10697368e-01 8.66606534e-01 7.39795566e-01 -2.50117123e-01
-2.10686281e-01 -1.31776738e+00 -3.41060162e-01 -1.06245726e-01
1.47751063e-01 -4.77000594e-01 -1.15720503e-01 3.06949943e-01]
|
[12.209250450134277, 8.923418045043945]
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.