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
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
5919a1cb-c1e7-4c04-be08-bc937cb533e7
|
lipschitz-continuity-of-signal-temporal-logic
|
2304.03849
| null |
https://arxiv.org/abs/2304.03849v1
|
https://arxiv.org/pdf/2304.03849v1.pdf
|
Lipschitz Continuity of Signal Temporal Logic Robustness Measures: Synthesizing Control Barrier Functions from One Expert Demonstration
|
Control Barrier Functions (CBFs) allow for efficient synthesis of controllers to maintain desired invariant properties of safety-critical systems. However, the problem of identifying a CBF remains an open question. As such, this paper provides a constructive method for control barrier function synthesis around one expert demonstration that realizes a desired system specification formalized in Signal Temporal Logic (STL). First, we prove that all STL specifications have Lipschitz-continuous robustness measures. Second, we leverage this Lipschitz continuity to synthesize a time-varying control barrier function. By filtering control inputs to maintain the positivity of this function, we ensure that the system trajectory satisfies the desired STL specification. Finally, we demonstrate the effectiveness of our approach on the Robotarium.
|
['Aaron D. Ames', 'Richard M. Murray', 'Apurva Badithela', 'Prithvi Akella']
|
2023-04-07
| null | null | null | null |
['open-question']
|
['natural-language-processing']
|
[ 5.29326618e-01 7.22368002e-01 -3.48550320e-01 3.21221828e-01
-6.85563862e-01 -1.08446491e+00 3.84309262e-01 1.94249168e-01
2.21797779e-01 7.78051615e-01 -4.37268913e-01 -6.00710213e-01
-4.89891380e-01 -5.67860544e-01 -1.11688793e+00 -6.01920247e-01
-2.09089473e-01 -5.13275564e-01 4.53765661e-01 -2.35169813e-01
6.31565452e-02 7.24450231e-01 -9.91605997e-01 -1.05504781e-01
5.59977233e-01 9.55796361e-01 -2.76168764e-01 7.11524546e-01
8.28168154e-01 8.70319486e-01 -4.82394814e-01 3.63009006e-01
3.58575821e-01 -5.24083018e-01 -8.42720807e-01 -9.59042460e-02
6.80872351e-02 -1.77365467e-01 -1.00252785e-01 1.38233852e+00
-1.88710559e-02 -6.29754290e-02 6.81282938e-01 -2.10895276e+00
-2.27594659e-01 6.11136138e-01 7.26675391e-02 -5.44405818e-01
2.91030377e-01 4.34462249e-01 7.92011082e-01 -1.61603764e-01
6.25979304e-01 1.01135230e+00 6.04662716e-01 7.40593612e-01
-1.65369761e+00 -4.53742981e-01 2.42974088e-01 -5.61445475e-01
-1.46427786e+00 -1.82006195e-01 4.42016542e-01 -7.16635048e-01
6.94521666e-01 4.24664855e-01 7.32018948e-01 8.11016142e-01
8.18520010e-01 4.36146796e-01 8.95713091e-01 -2.07729429e-01
4.14286941e-01 -7.82658625e-03 1.55917645e-01 7.42372811e-01
3.23888689e-01 6.07430041e-01 9.40957516e-02 -2.03969225e-01
1.05148423e+00 -3.66280198e-01 -4.89627361e-01 -4.38959509e-01
-1.31542313e+00 5.96075058e-01 7.61445835e-02 1.93180248e-01
-1.31102830e-01 7.80449331e-01 3.27066034e-01 8.45815897e-01
-5.60534559e-02 7.91641176e-01 -4.99654680e-01 1.67051144e-02
-7.66188353e-02 4.25068259e-01 8.52611184e-01 1.30089653e+00
1.99730814e-01 3.42127711e-01 -2.14522108e-01 -3.20130348e-01
1.51535764e-01 8.15693796e-01 -4.09082353e-01 -1.30999708e+00
-1.05870314e-01 5.85188866e-01 6.78244412e-01 -6.52426779e-01
-1.60047755e-01 -4.32436094e-02 -4.61263567e-01 6.50720179e-01
4.42477852e-01 -3.24755132e-01 -4.24288034e-01 2.07936692e+00
2.76833236e-01 2.81080995e-02 1.06604144e-01 6.53867304e-01
-4.64803725e-01 8.04677546e-01 -3.14572334e-01 -5.65764666e-01
7.04077661e-01 -2.49357015e-01 -7.73957014e-01 2.40007788e-01
4.46401775e-01 -7.81928003e-02 1.12198317e+00 5.44009507e-01
-1.23492169e+00 1.10891812e-01 -1.34101462e+00 5.53177536e-01
1.25556707e-01 -6.57913685e-02 -1.04557827e-01 3.33811760e-01
-1.02200890e+00 5.79806805e-01 -1.02536988e+00 -4.47728075e-02
-2.19862372e-01 5.36643863e-01 -3.08678359e-01 7.24526584e-01
-1.11714399e+00 8.42490017e-01 3.09091955e-01 1.34334400e-01
-1.37352574e+00 -1.05912840e+00 -7.43048191e-01 -5.18259592e-02
6.34924173e-01 -2.86321402e-01 1.55062938e+00 -7.39480257e-01
-1.95479846e+00 2.59922981e-01 3.06652486e-01 -6.71435535e-01
7.22341597e-01 1.25445858e-01 -1.18773893e-01 2.59854347e-01
-1.48488477e-01 2.57846296e-01 1.03576875e+00 -1.07259047e+00
-6.08500957e-01 1.72165021e-01 5.58209479e-01 -6.37555182e-01
-9.96026173e-02 -1.25789478e-01 4.32178169e-01 -1.92585036e-01
-2.72744030e-01 -9.51663494e-01 -5.70017159e-01 3.20021480e-01
-2.73488820e-01 4.06347550e-02 5.18521190e-01 -1.02587059e-01
1.31480968e+00 -2.40694427e+00 3.20294470e-01 5.96941352e-01
-4.89582494e-02 -5.02547808e-02 -2.27228831e-02 5.70326686e-01
5.89757375e-02 2.43159100e-01 -3.75819653e-01 5.48827291e-01
2.96111226e-01 -1.12190381e-01 -9.62049007e-01 1.14778805e+00
6.98376477e-01 6.62851334e-01 -8.59942317e-01 -2.11679593e-01
2.53121704e-01 2.45616838e-01 -7.23352432e-01 5.57326563e-02
-6.48810685e-01 6.48752511e-01 -7.34357238e-01 1.55667946e-01
1.49231106e-01 2.51910508e-01 1.99561179e-01 1.82062671e-01
-6.45639002e-01 -2.95269657e-02 -1.20608234e+00 6.35758936e-01
-5.36014318e-01 3.76927316e-01 6.88028395e-01 -7.43668675e-01
7.97563612e-01 5.58250844e-01 5.16205966e-01 -1.71100378e-01
6.80405080e-01 3.82571459e-01 -2.08609849e-01 -1.43969461e-01
-2.02656105e-01 -3.67654204e-01 -3.64613354e-01 1.02879189e-01
-5.23660064e-01 -7.62841284e-01 -5.93532920e-02 -1.90880165e-01
1.37327027e+00 -2.72107534e-02 3.89098018e-01 -7.18355358e-01
8.77657771e-01 2.53814816e-01 6.56200469e-01 2.46847123e-01
-4.18695718e-01 -5.02094403e-02 1.16954839e+00 2.01841801e-01
-1.16019404e+00 -9.10030007e-01 1.89327642e-01 3.28715086e-01
2.77092129e-01 -2.67419308e-01 -8.64303827e-01 -2.86686033e-01
1.26026213e-01 7.29260564e-01 -6.39591038e-01 -7.21143246e-01
-6.60362542e-01 5.86057246e-01 8.16955507e-01 5.68529189e-01
6.14394881e-02 -5.63679755e-01 -1.15456462e+00 2.84619331e-01
2.96519846e-01 -8.34415615e-01 -8.06293249e-01 2.69590914e-01
-6.50476754e-01 -1.29517961e+00 -5.65853454e-02 -9.21722412e-01
9.84423101e-01 -1.66981608e-01 2.13923678e-01 -1.48473009e-01
-1.81458041e-01 7.51372457e-01 1.88697785e-01 -3.48121434e-01
-7.01792002e-01 -4.02597785e-01 1.23488508e-01 -1.73508286e-01
-6.72441840e-01 -1.88339025e-01 -1.41451463e-01 5.49001634e-01
-8.19942117e-01 -6.47717938e-02 -2.56779194e-01 7.60858178e-01
8.10333908e-01 4.12849754e-01 7.16239572e-01 -7.27629215e-02
1.00145805e+00 7.55900145e-02 -1.80896294e+00 3.42351615e-01
-4.86081898e-01 1.42269075e-01 1.04392540e+00 -8.49934816e-01
-5.07717550e-01 6.20436728e-01 4.69422638e-01 -4.14556533e-01
1.86702415e-01 3.11785519e-01 -2.08615929e-01 -1.35824755e-01
6.57656670e-01 -5.29810041e-02 3.87714177e-01 3.01387787e-01
3.92363608e-01 2.53614873e-01 5.82003057e-01 -9.30677533e-01
8.94666791e-01 4.24925238e-01 4.15871918e-01 -5.18758297e-01
-2.34226257e-01 -6.92260787e-02 -2.21393317e-01 -5.44904709e-01
6.50909960e-01 -4.42800909e-01 -1.70310676e+00 1.37132168e-01
-1.16488838e+00 -8.03467691e-01 -6.79379761e-01 1.65371850e-01
-1.29825211e+00 -2.71247447e-01 -4.32191104e-01 -1.75158989e+00
-3.18880044e-02 -1.04335463e+00 1.10317242e+00 -2.96054244e-01
-5.68609476e-01 -6.30065560e-01 1.58308908e-01 -7.42527783e-01
2.97078073e-01 8.74504387e-01 8.01777482e-01 1.06473215e-01
-7.28651464e-01 -4.73570824e-01 2.47326598e-01 2.52802223e-01
-3.39680258e-03 3.82358789e-01 -4.30186749e-01 -6.12527013e-01
3.86153579e-01 -6.67266175e-02 6.63544610e-02 4.20326978e-01
4.23687518e-01 -7.80037344e-01 -3.71823162e-01 2.35379443e-01
1.47090888e+00 5.43629289e-01 2.64566600e-01 2.22372741e-01
2.97595840e-02 8.52025092e-01 9.03206825e-01 4.08431470e-01
-1.91184834e-01 2.80506253e-01 4.77922946e-01 4.78687972e-01
4.51404512e-01 -2.76910305e-01 9.30618942e-01 7.62609988e-02
3.13878506e-01 2.67428577e-01 -8.35427284e-01 8.74190032e-01
-1.94927955e+00 -6.93163753e-01 -1.71544962e-02 2.39736485e+00
9.83970881e-01 1.45791665e-01 3.94577235e-01 5.84770560e-01
9.46325958e-01 -7.15340137e-01 -5.68844855e-01 -5.65865636e-01
2.56940722e-01 4.14353907e-02 9.08378303e-01 1.03365374e+00
-8.37681890e-01 5.95638096e-01 6.73248196e+00 3.25173169e-01
-1.34026718e+00 -4.39186841e-01 2.15701852e-02 2.39904225e-01
-2.85271317e-01 1.18424833e-01 -4.29803908e-01 1.35324001e-01
1.08380604e+00 -9.46731150e-01 6.05697632e-01 1.04542744e+00
7.04358578e-01 2.04921708e-01 -1.44559109e+00 2.31500179e-01
-7.86867976e-01 -1.03430736e+00 -7.02500522e-01 -1.61359578e-01
7.33166993e-01 -8.27594280e-01 2.75788993e-01 2.87249833e-02
5.30218482e-01 -8.51974726e-01 1.20702660e+00 2.39562050e-01
7.66068161e-01 -8.85398030e-01 2.58202970e-01 4.48067486e-01
-1.30951071e+00 -3.17159623e-01 1.50650725e-01 -1.72142342e-01
3.29320788e-01 2.79320806e-01 -8.28921139e-01 2.05546528e-01
-9.82228145e-02 3.59494448e-01 -3.06117237e-02 9.00320828e-01
-5.18804133e-01 4.25453901e-01 -4.71533626e-01 -2.71747231e-01
2.68132210e-01 -8.64738449e-02 7.40905225e-01 8.38551581e-01
1.45730078e-01 1.53818160e-01 1.81576505e-01 1.23243046e+00
5.20203173e-01 -3.33479583e-01 -1.05210042e+00 -5.31560779e-01
2.52166539e-01 6.69818878e-01 -5.09355009e-01 8.98444131e-02
1.44173607e-01 4.00827497e-01 -1.16272315e-01 3.23174894e-01
-1.22775733e+00 -7.00490475e-01 7.83744097e-01 7.28183463e-02
4.29794155e-02 -6.36480510e-01 -5.35480261e-01 -6.22836769e-01
1.49639621e-01 -7.88428426e-01 8.46086442e-02 -3.96233439e-01
-9.20434833e-01 2.46097878e-01 1.30688578e-01 -1.49186575e+00
-4.72788930e-01 -4.34539169e-01 -4.40251708e-01 4.41304982e-01
-9.24869299e-01 -7.79957056e-01 3.18230420e-01 6.97098434e-01
-1.21689755e-02 4.38542604e-01 5.83802223e-01 -5.01243249e-02
-4.40845639e-01 9.42876115e-02 -3.37963164e-01 -3.13769817e-01
4.36471283e-01 -1.22295249e+00 1.60535440e-01 1.07226169e+00
-9.47275221e-01 9.47694421e-01 1.10732865e+00 -7.75141776e-01
-1.99722731e+00 -1.44367611e+00 5.21184325e-01 -3.40402305e-01
1.43233085e+00 -1.87174022e-01 -5.48472762e-01 9.44349289e-01
-4.91029061e-02 -3.76862288e-02 -3.76025200e-01 -9.00344849e-01
-3.27945769e-01 -1.51631176e-01 -1.42346859e+00 1.02122808e+00
7.32193649e-01 -4.81183082e-01 -4.02890384e-01 -1.69381779e-02
1.25999665e+00 -3.05452049e-01 -9.72866535e-01 5.66638947e-01
5.39636910e-01 -7.41034746e-02 6.14002824e-01 -6.33751333e-01
7.53170550e-02 -9.62292194e-01 -1.80088907e-01 -1.14067638e+00
-5.92923164e-03 -1.37908554e+00 -1.11219302e-01 8.80137086e-01
4.38489258e-01 -8.34731758e-01 1.50476754e-01 8.04629743e-01
-1.82426512e-01 -5.98502457e-01 -9.11528826e-01 -1.65595758e+00
3.68294030e-01 -4.40917194e-01 2.32712761e-01 6.39663577e-01
9.71513391e-01 1.31942615e-01 -1.28980458e-01 5.20867646e-01
4.59188759e-01 -6.81319237e-02 4.30441439e-01 -7.67997980e-01
-6.58961684e-02 -6.40139997e-01 -1.72789842e-01 -5.53104877e-01
3.67981732e-01 -5.14014959e-01 6.70192480e-01 -1.17960882e+00
-6.13797843e-01 -2.11012259e-01 -1.21494375e-01 5.23078799e-01
5.11436760e-01 -3.75400752e-01 3.94705795e-02 -4.33867455e-01
-2.64204144e-01 3.75739664e-01 1.25906622e+00 -3.13077688e-01
-6.31350458e-01 2.76321061e-02 -2.78527200e-01 5.14236391e-01
9.09349859e-01 -1.94257081e-01 -8.41775358e-01 7.08443597e-02
5.19752502e-01 5.91519892e-01 5.20212352e-01 -8.81823182e-01
2.15831444e-01 -8.98647130e-01 -7.48591065e-01 -3.26441824e-01
-1.17719732e-01 -1.29787743e+00 5.33317447e-01 1.19445562e+00
-6.54052615e-01 -6.70140311e-02 2.82987088e-01 5.53973556e-01
-1.31140202e-01 7.97080919e-02 1.11347449e+00 6.55820012e-01
-1.19595185e-01 -7.90535659e-02 -8.72248709e-01 -2.28876606e-01
1.49668622e+00 1.22713909e-01 -2.42619053e-01 -2.79322177e-01
-4.49243516e-01 5.95762670e-01 4.72539514e-01 3.66547517e-02
5.55085897e-01 -1.08192730e+00 -2.60266155e-01 1.97587520e-01
7.96066150e-02 -2.67676443e-01 -2.26574779e-01 1.11560893e+00
-3.34547341e-01 6.00982010e-01 -1.56567305e-01 -5.16018033e-01
-1.03295481e+00 8.91542017e-01 5.45725703e-01 2.81598926e-01
-5.85101366e-01 3.72766107e-01 2.30262369e-01 -2.67813891e-01
3.75952572e-01 -1.13513911e+00 4.63210434e-01 -4.34827834e-01
4.67089921e-01 3.37408721e-01 -2.75180787e-01 2.08098933e-01
-6.34210348e-01 5.38437963e-01 7.39147723e-01 -6.84596539e-01
1.13761353e+00 -2.63867415e-02 -2.79803276e-01 4.43187237e-01
9.11145031e-01 -1.85816109e-01 -1.58090031e+00 6.33632004e-01
1.80657417e-01 -4.08596583e-02 -3.54179233e-01 -3.93897086e-01
-6.27047420e-01 4.20535147e-01 9.18534100e-02 5.37100255e-01
1.10442615e+00 -3.72934431e-01 4.20360714e-01 6.02495134e-01
7.82687008e-01 -9.13100541e-01 -6.06725179e-02 7.49677896e-01
1.18512499e+00 -2.58690476e-01 -3.97244483e-01 -8.39320362e-01
-3.36027235e-01 1.12691402e+00 5.74598014e-01 -6.93279743e-01
5.58773279e-01 1.06497788e+00 -5.29802442e-01 2.98565745e-01
-9.73818004e-01 2.56359633e-02 -1.12612866e-01 5.85010350e-01
1.02946244e-01 2.35443725e-03 -5.55329084e-01 5.72510421e-01
1.27560750e-01 2.90361792e-01 7.78112113e-01 1.21337605e+00
-5.83383203e-01 -8.19336355e-01 -4.84251350e-01 -3.70487422e-01
-1.95416927e-01 4.43187147e-01 -5.99627793e-01 9.93225753e-01
-4.96470511e-01 1.05757499e+00 -4.11075532e-01 -2.60986865e-01
7.92553723e-01 -1.59277856e-01 8.13714743e-01 -4.66876298e-01
-5.71502566e-01 1.15049861e-01 2.00175464e-01 -7.60888875e-01
-2.71629561e-02 -2.31561109e-01 -1.57878149e+00 -1.42484501e-01
-4.98153806e-01 1.53306499e-01 4.06602979e-01 6.39076889e-01
7.46841729e-02 6.87706947e-01 7.87843883e-01 -2.19830662e-01
-1.01902139e+00 -2.61608928e-01 -5.38598835e-01 -2.98480004e-01
9.60594296e-01 -5.71151376e-01 -5.81329405e-01 3.84670407e-01]
|
[4.809057712554932, 2.2785587310791016]
|
0df338bd-0b2b-435a-9600-b88909514daa
|
on-the-optimization-landscape-of-dynamic
|
2209.05042
| null |
https://arxiv.org/abs/2209.05042v2
|
https://arxiv.org/pdf/2209.05042v2.pdf
|
On the Optimization Landscape of Dynamic Output Feedback: A Case Study for Linear Quadratic Regulator
|
The convergence of policy gradient algorithms hinges on the optimization landscape of the underlying optimal control problem. Theoretical insights into these algorithms can often be acquired from analyzing those of linear quadratic control. However, most of the existing literature only considers the optimization landscape for static full-state or output feedback policies (controllers). We investigate the more challenging case of dynamic output-feedback policies for linear quadratic regulation (abbreviated as dLQR), which is prevalent in practice but has a rather complicated optimization landscape. We first show how the dLQR cost varies with the coordinate transformation of the dynamic controller and then derive the optimal transformation for a given observable stabilizing controller. One of our core results is the uniqueness of the stationary point of dLQR when it is observable, which provides an optimality certificate for solving dynamic controllers using policy gradient methods. Moreover, we establish conditions under which dLQR and linear quadratic Gaussian control are equivalent, thus providing a unified viewpoint of optimal control of both deterministic and stochastic linear systems. These results further shed light on designing policy gradient algorithms for more general decision-making problems with partially observed information.
|
['Lin Zhao', 'Yang Zheng', 'Wenhan Cao', 'Jingliang Duan']
|
2022-09-12
| null | null | null | null |
['policy-gradient-methods']
|
['methodology']
|
[ 3.41478512e-02 2.98528552e-01 -7.26198792e-01 4.23620969e-01
-5.99729300e-01 -8.66881490e-01 2.71083593e-01 3.38784531e-02
-3.30006987e-01 8.63327086e-01 2.49227546e-02 -7.90901065e-01
-4.37608868e-01 -3.31306845e-01 -8.21990967e-01 -1.10558784e+00
1.69641316e-01 -2.68157832e-02 -1.72517642e-01 -4.05801117e-01
-5.25368080e-02 4.24763620e-01 -9.80287135e-01 -5.63044548e-01
1.01338494e+00 9.39668119e-01 1.74826637e-01 7.15326726e-01
5.57037055e-01 4.22574818e-01 -2.17287868e-01 2.37373665e-01
6.35758936e-01 -4.08881485e-01 -2.80334085e-01 3.40239912e-01
2.09822819e-01 -2.17396155e-01 -3.97610039e-01 1.38558722e+00
6.33602977e-01 3.24836761e-01 5.60857356e-01 -1.20924771e+00
-6.09503925e-01 3.11321497e-01 -2.86535650e-01 9.89909321e-02
1.01484880e-02 3.07245344e-01 9.94927824e-01 -3.21190625e-01
5.59539795e-01 1.38443625e+00 4.15182084e-01 6.52391493e-01
-1.51764524e+00 -1.90530017e-01 6.54384613e-01 7.16626346e-02
-1.00894475e+00 -4.09114718e-01 2.88096577e-01 -8.33545327e-01
2.66808093e-01 2.44303912e-01 4.53916788e-01 8.22900355e-01
2.93448389e-01 7.06501603e-01 1.24703348e+00 -1.98026717e-01
3.30212027e-01 4.52939957e-01 1.80881321e-01 8.23507011e-01
5.98666072e-01 4.80317146e-01 2.01652914e-01 -9.00254026e-02
8.03626299e-01 -1.75105006e-01 -6.34859800e-01 -9.14061725e-01
-1.14607012e+00 9.66791987e-01 2.10100919e-01 2.83904281e-03
-5.31484365e-01 1.30273372e-01 2.01067269e-01 9.11699533e-01
1.64948642e-01 6.15514815e-01 -4.19980109e-01 5.03864996e-02
-1.06003463e-01 5.40515482e-01 1.00919831e+00 9.60932791e-01
4.62667584e-01 6.20911598e-01 -5.07084131e-01 3.32647741e-01
2.45138600e-01 1.18182600e+00 -1.04053296e-01 -1.26640725e+00
5.42680025e-01 2.47157887e-01 8.84641469e-01 -1.02271354e+00
-2.79143810e-01 -6.92974329e-01 -6.16099715e-01 3.78876090e-01
8.27589452e-01 -8.87568951e-01 -2.81957746e-01 2.04132104e+00
3.89001429e-01 -2.44481251e-01 1.35243833e-01 1.12959158e+00
-4.12724078e-01 6.46603286e-01 -5.51300645e-01 -8.40116024e-01
8.17968369e-01 -4.82861519e-01 -1.05544436e+00 -1.23607934e-01
4.77513969e-01 -1.89297929e-01 1.19995928e+00 9.36862156e-02
-8.56517911e-01 2.92026401e-02 -1.01257145e+00 3.18009853e-01
1.03499391e-03 4.31328148e-01 -5.27712107e-02 4.24641341e-01
-1.12788129e+00 5.64763427e-01 -8.13111722e-01 -4.34241712e-01
-3.94661158e-01 3.96467716e-01 1.28737882e-01 3.68416011e-01
-1.10327590e+00 1.08731234e+00 6.16958216e-02 2.05402747e-01
-1.08984077e+00 -7.87043631e-01 -5.98117352e-01 -2.33624205e-02
1.28202498e+00 -7.52628803e-01 1.58474183e+00 -9.04113472e-01
-1.98454916e+00 1.81422710e-01 -2.40197465e-01 -3.16375852e-01
8.02977681e-01 -1.43844128e-01 1.47063687e-01 -4.71085049e-02
6.06252663e-02 -3.19217920e-01 1.14455807e+00 -1.17575026e+00
-6.85580730e-01 -4.99520153e-01 3.45928997e-01 3.63045543e-01
-2.31210887e-01 -1.66581154e-01 5.84743805e-02 -3.10372800e-01
-5.06312072e-01 -1.21621931e+00 -6.23188615e-01 9.95042250e-02
-4.68952090e-01 3.16732447e-03 8.52268934e-01 -4.92773443e-01
1.45566928e+00 -1.76271117e+00 6.29465997e-01 1.74401805e-01
-2.18224735e-03 9.33957398e-02 8.98084491e-02 5.91257930e-01
1.16062639e-02 -1.66697036e-02 -9.88386348e-02 3.70267220e-02
3.61831844e-01 4.91404086e-02 -9.52778041e-01 9.68662560e-01
1.15707502e-01 6.85035765e-01 -9.80133057e-01 6.55841455e-02
1.68004602e-01 -4.20373958e-03 -6.04600608e-01 -2.87393481e-02
-3.65351915e-01 7.67413020e-01 -1.17847884e+00 1.19192034e-01
1.02667876e-01 -1.72892213e-01 1.87935680e-01 2.61798620e-01
-7.20242977e-01 -2.43226916e-01 -1.37454522e+00 8.19769204e-01
-5.31938970e-01 5.74678063e-01 1.00517094e+00 -1.40597963e+00
4.12537068e-01 3.06985021e-01 5.88418782e-01 -1.61925852e-01
4.01019990e-01 1.03759803e-01 -1.99613824e-01 -4.50129688e-01
1.53175130e-01 -2.32189119e-01 -5.08133806e-02 2.86752909e-01
-4.25593048e-01 -1.43974394e-01 1.26005962e-01 -1.29356012e-01
8.48309636e-01 -2.04850391e-01 4.66232777e-01 -7.66764879e-01
6.44008040e-01 -5.19716702e-02 5.92377067e-01 8.83106530e-01
-3.95352006e-01 -1.14255354e-01 1.05114591e+00 3.02416444e-01
-1.04042923e+00 -6.47284150e-01 7.40150269e-03 9.81025577e-01
1.00447223e-01 6.59595802e-02 -6.17893398e-01 -2.66208768e-01
6.05045736e-01 4.76670980e-01 -7.44674027e-01 -3.37330043e-01
-4.08660114e-01 -5.07499993e-01 -2.50884056e-01 1.94102444e-03
1.64175645e-01 -1.83091626e-01 -4.58320022e-01 4.63821560e-01
2.19713464e-01 -9.85421240e-01 -7.81220615e-01 -8.50804299e-02
-7.35556126e-01 -1.28438282e+00 -7.85564125e-01 -2.40516603e-01
7.34573364e-01 1.46623164e-01 3.20831239e-01 -4.28523570e-01
2.93323874e-01 1.03430808e+00 2.53376156e-01 -5.94160795e-01
-4.61596578e-01 1.68650612e-01 6.17956042e-01 5.45414984e-01
-4.97270614e-01 -1.51492670e-01 -3.81737143e-01 6.56253457e-01
-5.88803351e-01 -2.53865033e-01 2.72175908e-01 8.48225713e-01
7.21234620e-01 -2.04692543e-01 6.01462424e-01 -5.85178316e-01
1.21161199e+00 -4.77786213e-01 -1.57335985e+00 3.49188447e-01
-8.36108148e-01 4.99850124e-01 1.12161922e+00 -5.69936633e-01
-9.44437087e-01 1.05625987e-01 5.41687787e-01 -5.12452900e-01
4.54260558e-01 3.44821364e-01 -4.19027731e-02 -5.32808788e-02
5.80030322e-01 2.58299708e-01 4.72090095e-01 -3.55320424e-01
5.30995786e-01 5.77537179e-01 3.45220983e-01 -9.31628108e-01
8.37724030e-01 6.25196338e-01 2.83418775e-01 -9.10480797e-01
-9.73772168e-01 -4.03867394e-01 -3.48837733e-01 -3.47210228e-01
6.26728117e-01 -6.23340428e-01 -1.34027517e+00 1.54935032e-01
-8.05278361e-01 -7.48217881e-01 -5.53351820e-01 3.87101412e-01
-1.18332732e+00 8.23914260e-02 -3.50534678e-01 -1.48659897e+00
9.99231189e-02 -1.21551013e+00 9.08660412e-01 1.47641703e-01
3.13192070e-01 -1.21209276e+00 2.54204959e-01 -3.34260434e-01
5.20937204e-01 4.20951515e-01 6.83692992e-01 9.68654603e-02
-5.27359843e-01 -1.85995609e-01 1.12233080e-01 3.23959410e-01
-8.24775323e-02 -2.26135194e-01 -3.60054165e-01 -7.63901234e-01
3.79339397e-01 -7.34755397e-02 5.51842332e-01 8.04572105e-01
5.56150556e-01 -7.29039729e-01 -4.02487636e-01 3.68811667e-01
1.50452054e+00 3.63863021e-01 -2.85655439e-01 2.71703571e-01
6.01415396e-01 7.00525165e-01 7.16872394e-01 6.45161510e-01
2.45481670e-01 7.74464309e-01 3.72847319e-01 1.92297637e-01
6.65681660e-01 -3.23815286e-01 8.15141976e-01 5.50845265e-01
1.44710168e-01 1.41858935e-01 -5.92677057e-01 4.72281814e-01
-2.30856180e+00 -7.98467636e-01 1.21235758e-01 2.65070057e+00
8.99956405e-01 -4.15733665e-01 2.83672035e-01 -2.58815616e-01
1.11430228e+00 -6.80621192e-02 -1.23015463e+00 -4.63635445e-01
-8.03463012e-02 -5.14351904e-01 1.15031898e+00 8.31148028e-01
-1.09289992e+00 6.42820537e-01 6.66747332e+00 6.25053704e-01
-1.35231137e+00 1.53271947e-02 4.45343554e-01 -8.84442255e-02
1.26966491e-01 -4.95636165e-02 -1.06844020e+00 4.45524454e-01
1.00809109e+00 -9.22046244e-01 8.80350769e-01 1.12380302e+00
1.31107605e+00 2.16081291e-02 -8.57732892e-01 6.33043408e-01
-6.01163387e-01 -8.42184424e-01 -5.27392507e-01 4.68383998e-01
1.12773645e+00 -2.44769648e-01 4.60799068e-01 4.39617634e-01
5.53632915e-01 -4.94201243e-01 6.09568954e-01 5.52890182e-01
8.00250769e-01 -6.57455206e-01 2.99221426e-01 5.68517268e-01
-9.02017832e-01 -7.92786419e-01 -4.49196935e-01 -2.31489629e-01
2.74204940e-01 3.36208731e-01 -3.00107628e-01 4.90922540e-01
-9.40737128e-02 8.17206860e-01 3.24061662e-02 1.01549077e+00
-2.31118500e-01 4.59553629e-01 -5.08841157e-01 -3.56144071e-01
3.99638683e-01 -7.20002055e-01 1.10731184e+00 7.60277331e-01
2.01266944e-01 -2.49500815e-02 5.15795767e-01 7.92832971e-01
2.85022199e-01 2.53615379e-01 -9.66925204e-01 -3.14319909e-01
1.09741889e-01 8.18691075e-01 -2.35925242e-01 -1.64641321e-01
-2.21875340e-01 4.59527969e-01 3.66336346e-01 6.33842289e-01
-6.25050843e-01 -3.39343309e-01 1.27818775e+00 -1.35286033e-01
3.39282840e-01 -5.16156137e-01 -1.21266715e-01 -1.20696914e+00
2.22439319e-01 -9.26931560e-01 1.60926595e-01 -5.45890257e-03
-9.42090869e-01 -4.90106121e-02 1.70151800e-01 -1.19088185e+00
-6.23807728e-01 -6.58665776e-01 -3.53853822e-01 5.78875542e-01
-1.23084831e+00 -2.75413752e-01 4.07084823e-01 5.70614636e-01
2.33174697e-01 1.05589949e-01 3.73506695e-01 -1.89930853e-02
-1.06661057e+00 2.47401297e-01 1.17438579e+00 -3.89856637e-01
4.81310278e-01 -1.65128696e+00 -3.11890960e-01 9.51496959e-01
-6.32746041e-01 6.54867113e-01 1.17873132e+00 -3.81890297e-01
-2.15886211e+00 -1.24469078e+00 1.65062547e-01 -2.73562402e-01
1.41229308e+00 -2.00061217e-01 -5.32177269e-01 6.47945583e-01
-1.55972272e-01 -1.98367804e-01 -2.36617252e-01 -3.57789487e-01
2.50794113e-01 -1.91233039e-01 -7.09510744e-01 1.02821290e+00
8.55135858e-01 -3.78234595e-01 2.95509007e-02 4.88712043e-01
9.26273525e-01 -3.91935021e-01 -6.96008742e-01 1.34220779e-01
3.69362235e-01 -2.10928157e-01 5.77690899e-01 -1.14095008e+00
-1.56656191e-01 -4.03983295e-01 -4.46296744e-02 -1.75950420e+00
-2.99440194e-02 -1.42436242e+00 -4.10817683e-01 5.10338843e-01
3.60354036e-01 -9.83110845e-01 4.30968761e-01 4.67863798e-01
1.36999590e-02 -1.00324225e+00 -7.98176110e-01 -1.26484394e+00
3.88496190e-01 -3.33145559e-01 -2.06057966e-01 7.26486981e-01
1.74974829e-01 2.82483846e-01 -5.84684789e-01 4.34070408e-01
6.03200436e-01 1.14326805e-01 6.66987002e-01 -7.04548061e-01
-4.57645297e-01 -7.01278806e-01 3.31215151e-02 -1.37267184e+00
5.00574946e-01 -5.79679608e-01 2.85299897e-01 -1.28470314e+00
-2.77873546e-01 -7.85991922e-02 -8.29686970e-02 -1.19405948e-02
-2.64006943e-01 -5.79639196e-01 2.92688727e-01 5.80676645e-02
-4.76444960e-01 8.85399401e-01 1.48658442e+00 -4.44897786e-02
-5.94309151e-01 6.42368078e-01 -8.29892278e-01 3.66622657e-01
7.05413282e-01 -1.94490969e-01 -7.03151226e-01 -3.33096921e-01
-1.28075760e-02 4.29437578e-01 5.04499637e-02 -5.29296637e-01
1.26278684e-01 -6.77552402e-01 -6.37030363e-01 2.92429272e-02
-3.52755412e-02 -7.44566739e-01 -1.02250502e-01 8.33994269e-01
-6.03339672e-01 7.56950006e-02 -8.27819631e-02 1.18433952e+00
8.08290318e-02 -8.72386098e-02 1.05787480e+00 2.34640598e-01
-2.68272310e-01 5.49605846e-01 -7.83333957e-01 3.50032359e-01
1.02705169e+00 2.66486168e-01 -1.42036274e-01 -9.46143985e-01
-9.93703723e-01 7.36113131e-01 3.10266912e-01 2.64016151e-01
6.13020211e-02 -1.14576101e+00 -4.08746094e-01 -1.76790297e-01
-3.44517678e-01 -5.22872567e-01 -9.65989679e-02 1.26656389e+00
1.38037473e-01 9.10261750e-01 2.97823191e-01 -4.48595226e-01
-8.90054643e-01 8.32478702e-01 8.50492477e-01 -3.12005281e-01
-2.67949730e-01 5.58068603e-02 2.32932970e-01 -2.01927066e-01
3.08535308e-01 -7.19343781e-01 -3.35582090e-03 1.08550694e-02
4.21634942e-01 4.95676279e-01 -2.14785218e-01 -3.44193816e-01
4.89693098e-02 5.82705021e-01 3.36001694e-01 -4.72047299e-01
9.02051032e-01 -4.80675220e-01 1.67706847e-01 7.36527979e-01
1.09607911e+00 -2.19165534e-01 -1.67086565e+00 -4.31079596e-01
1.68161795e-01 -2.44765013e-01 7.91807026e-02 -2.52321184e-01
-1.02021492e+00 5.56279004e-01 4.68522519e-01 4.04922515e-01
7.62415171e-01 -3.81926805e-01 1.50234491e-01 8.51053178e-01
5.20619631e-01 -1.14775813e+00 -2.60689527e-01 8.50598037e-01
1.08267879e+00 -1.01074886e+00 -2.57720828e-01 -2.26003528e-01
-5.87837696e-01 1.00116551e+00 4.36835498e-01 -4.31810290e-01
6.48603857e-01 1.44193411e-01 -1.59832582e-01 3.56460899e-01
-9.20861363e-01 -5.01014829e-01 1.32592320e-01 4.53552395e-01
-4.17940058e-02 1.21205263e-01 -6.88987195e-01 2.59020358e-01
1.31441340e-01 -1.38088673e-01 6.34383798e-01 9.07031357e-01
-6.59640193e-01 -8.23034167e-01 -4.82700109e-01 2.50200033e-01
-4.57642585e-01 3.81207585e-01 -3.87657911e-01 8.29414248e-01
-6.35245681e-01 9.16896582e-01 -1.96523547e-01 1.27829701e-01
7.81042814e-01 -1.60300940e-01 4.21552837e-01 -5.73712289e-01
-1.01078615e-01 1.17171342e-02 -1.66919947e-01 -7.75497675e-01
5.08339293e-02 -7.39674866e-01 -7.53087461e-01 -1.89126641e-01
-3.47856700e-01 2.29565889e-01 5.96391499e-01 8.80173087e-01
3.97252947e-01 3.50294948e-01 9.60661590e-01 -7.29325235e-01
-1.74148405e+00 -6.93099201e-01 -7.26284504e-01 -1.09032445e-01
9.16565478e-01 -8.78367543e-01 -5.69838762e-01 -2.08671287e-01]
|
[4.761800765991211, 2.560778856277466]
|
048ff087-806b-4e79-8f76-c89d5b1bcf1e
|
semi-mae-masked-autoencoders-for-semi
|
2301.01431
| null |
https://arxiv.org/abs/2301.01431v1
|
https://arxiv.org/pdf/2301.01431v1.pdf
|
Semi-MAE: Masked Autoencoders for Semi-supervised Vision Transformers
|
Vision Transformer (ViT) suffers from data scarcity in semi-supervised learning (SSL). To alleviate this issue, inspired by masked autoencoder (MAE), which is a data-efficient self-supervised learner, we propose Semi-MAE, a pure ViT-based SSL framework consisting of a parallel MAE branch to assist the visual representation learning and make the pseudo labels more accurate. The MAE branch is designed as an asymmetric architecture consisting of a lightweight decoder and a shared-weights encoder. We feed the weakly-augmented unlabeled data with a high masking ratio to the MAE branch and reconstruct the missing pixels. Semi-MAE achieves 75.9% top-1 accuracy on ImageNet with 10% labels, surpassing prior state-of-the-art in semi-supervised image classification. In addition, extensive experiments demonstrate that Semi-MAE can be readily used for other ViT models and masked image modeling methods.
|
['Xiaoming Xu', 'Kang Zhao', 'Haojie Yu']
|
2023-01-04
| null | null | null | null |
['semi-supervised-image-classification']
|
['computer-vision']
|
[ 4.19971198e-01 4.82687086e-01 -3.71789664e-01 -5.41333973e-01
-6.65899038e-01 -9.65113491e-02 6.62602663e-01 -5.27699709e-01
-2.24001691e-01 6.20158613e-01 1.33898467e-01 -3.24220121e-01
7.10638642e-01 -4.74527806e-01 -1.05481219e+00 -7.75869787e-01
5.57039082e-01 4.13819462e-01 1.63720116e-01 9.62594599e-02
-3.19938660e-01 -9.48486403e-02 -1.78113985e+00 6.91824615e-01
7.46648073e-01 1.15309596e+00 5.56997478e-01 2.83832073e-01
-2.56736666e-01 1.71850801e+00 -3.14545870e-01 -3.76684964e-01
1.65882349e-01 -6.49583638e-01 -7.65929282e-01 5.56528628e-01
5.19840598e-01 -3.72654408e-01 -5.66827178e-01 1.02786207e+00
2.68473506e-01 -5.01360893e-01 6.79450572e-01 -1.50785196e+00
-9.65392888e-01 1.03143311e+00 -6.43347502e-01 -2.37540007e-01
-2.60995954e-01 1.06299974e-01 8.24888408e-01 -1.31950665e+00
4.73398745e-01 1.02705193e+00 6.44903064e-01 9.05137539e-01
-1.34507298e+00 -9.03629124e-01 1.90672889e-01 2.19941184e-01
-1.32642150e+00 -6.58654809e-01 7.21872568e-01 -4.84250009e-01
7.25825071e-01 -3.34632665e-01 4.60792214e-01 1.30894661e+00
-5.89944012e-02 1.15868318e+00 1.51868844e+00 -5.69333255e-01
1.82203725e-01 5.22644937e-01 1.89179350e-02 1.06333172e+00
-1.10899404e-01 2.33689755e-01 -7.95307577e-01 2.66161025e-01
7.12127030e-01 1.79175735e-01 -1.44409463e-01 -5.57224691e-01
-1.25450861e+00 8.08052361e-01 6.64952695e-01 -8.12823549e-02
-2.89283395e-01 1.49065107e-01 1.31492287e-01 4.90361243e-01
4.58911777e-01 -1.38856232e-01 -2.83057779e-01 4.36289281e-01
-1.10686803e+00 -4.74963546e-01 4.79740173e-01 1.09446847e+00
9.67331409e-01 6.02219641e-01 -1.19434521e-01 9.69543755e-01
6.18922055e-01 4.26242322e-01 6.66917443e-01 -7.44457364e-01
2.42017493e-01 7.05268323e-01 -3.22699130e-01 -3.98148447e-02
8.96410857e-05 -7.12629318e-01 -1.09217358e+00 5.60604274e-01
4.11503203e-02 1.72452733e-01 -1.45188749e+00 1.74861097e+00
1.18223868e-01 3.25384736e-01 2.47417331e-01 7.73876131e-01
1.04044580e+00 7.20774710e-01 2.79917181e-01 -2.28330985e-01
1.27470827e+00 -1.66829324e+00 -6.83717787e-01 -6.08689487e-01
5.02779841e-01 -6.02986217e-01 1.11438119e+00 2.45198399e-01
-1.06543052e+00 -8.53859484e-01 -1.27695262e+00 -1.99684739e-01
-2.19536930e-01 5.43473184e-01 3.43699545e-01 5.60363352e-01
-1.25713301e+00 2.65619010e-01 -8.18466842e-01 2.87597347e-02
8.91003788e-01 2.31187835e-01 -3.56055439e-01 -1.93507671e-01
-8.60455871e-01 9.71368313e-01 4.01258886e-01 2.84583773e-02
-1.68783176e+00 -6.49249375e-01 -1.04498017e+00 -8.37569088e-02
2.80169845e-01 -4.32918996e-01 1.27192664e+00 -1.42748380e+00
-1.43643153e+00 1.35294020e+00 -3.81751984e-01 -7.90737808e-01
4.53207493e-01 1.62752628e-01 -6.12781167e-01 3.41934264e-02
1.52542904e-01 1.14821708e+00 1.39158237e+00 -1.52234757e+00
-5.28136194e-01 -2.61375159e-01 -3.56278479e-01 8.61273929e-02
-4.54611629e-01 -2.26202458e-01 -2.55171448e-01 -8.20149362e-01
7.97204003e-02 -8.68660510e-01 -9.26345587e-02 2.05937415e-01
-2.46476889e-01 -5.16910069e-02 9.49563563e-01 -5.79532206e-01
8.23668361e-01 -2.13982749e+00 -3.38262320e-02 -1.90629825e-01
5.51303804e-01 6.07589483e-01 -2.34850809e-01 2.05157965e-01
-3.13751638e-01 -4.93667126e-01 -6.14113808e-01 -9.19194162e-01
-1.25124544e-01 4.62389320e-01 -4.69477296e-01 3.30177784e-01
3.65054250e-01 1.22095573e+00 -7.77153254e-01 -7.31677353e-01
3.75532806e-01 4.87709492e-01 -1.52432427e-01 3.43401164e-01
-1.97356492e-01 3.07390809e-01 7.01849833e-02 9.56219375e-01
7.08331704e-01 -7.29405642e-01 1.61500871e-01 -2.52283871e-01
2.05029622e-02 2.44167328e-01 -9.25749958e-01 1.72637558e+00
-5.15938103e-01 7.11994767e-01 1.34700343e-01 -1.04382300e+00
9.62940991e-01 3.03286463e-01 1.11414842e-01 -7.76380420e-01
1.56565011e-01 1.50146380e-01 -4.13577139e-01 -1.27251893e-01
2.33782500e-01 -1.26692623e-01 4.87660438e-01 5.96786737e-01
6.65276289e-01 3.05503190e-01 -1.17045894e-01 3.56579721e-01
7.16475368e-01 4.08039123e-01 3.00282121e-01 -1.91606656e-01
5.17322421e-01 4.90511134e-02 6.07082188e-01 6.21185362e-01
-2.25859001e-01 7.12199092e-01 -1.01149395e-01 -3.82768810e-01
-1.13958907e+00 -1.15747595e+00 -1.67077526e-01 1.10770988e+00
1.08296238e-01 -3.60073835e-01 -7.77538896e-01 -9.20614660e-01
-7.33984187e-02 5.76126635e-01 -6.82265341e-01 -3.70217234e-01
-1.40038803e-01 -5.98813415e-01 4.44150478e-01 8.14480662e-01
9.53234196e-01 -1.09285724e+00 -1.60375297e-01 2.38929078e-01
-2.66250491e-01 -1.28452444e+00 -4.68751013e-01 6.02129400e-01
-8.44263136e-01 -8.92720163e-01 -7.00259984e-01 -1.37826633e+00
9.72815037e-01 5.98190725e-01 1.14646733e+00 -8.92149732e-02
1.66581348e-02 1.50192574e-01 -3.45608354e-01 -3.64487916e-01
-9.63007689e-01 -2.16044784e-01 -6.90613091e-02 4.24079180e-01
4.92843181e-01 -5.15508115e-01 -5.37798226e-01 6.36680365e-01
-9.30495322e-01 6.71976447e-01 7.88327575e-01 1.24552488e+00
9.27930295e-01 -2.44041443e-01 6.12813771e-01 -1.02308106e+00
-2.72132903e-01 -4.59062248e-01 -5.29510915e-01 4.83261347e-01
-1.14486456e+00 1.12716019e-01 6.27435565e-01 -6.13609731e-01
-1.07451117e+00 5.77610791e-01 -5.59727140e-02 -7.61840284e-01
-2.58152038e-02 7.97215626e-02 -1.88057333e-01 -3.27837199e-01
6.34321570e-01 7.56123960e-01 2.30908245e-01 -5.35391569e-01
4.27271158e-01 1.19031537e+00 8.24420512e-01 -7.90972263e-02
6.98087573e-01 5.75475574e-01 -3.99901092e-01 -5.23605824e-01
-1.18606114e+00 -2.68904001e-01 -4.83629793e-01 -2.80971199e-01
7.85368383e-01 -1.70708835e+00 -2.16103271e-01 7.46078789e-01
-6.79310501e-01 -6.94189787e-01 -5.68304539e-01 2.87346721e-01
-5.75365961e-01 2.25039467e-01 -6.83659494e-01 -6.05525970e-01
-3.90793175e-01 -1.21590292e+00 9.72379208e-01 6.40234426e-02
3.13337982e-01 -7.63174951e-01 -2.82270581e-01 9.14567947e-01
2.98714370e-01 -2.77299941e-01 5.85245252e-01 -4.26160902e-01
-5.63508511e-01 1.28170863e-01 -3.99534404e-01 9.25876975e-01
-6.95117936e-02 -4.84306931e-01 -1.64258981e+00 -4.08820957e-01
4.52327877e-02 -1.00947833e+00 1.38812101e+00 1.41786441e-01
1.11918998e+00 -2.34128416e-01 -1.66282535e-01 8.25287163e-01
1.41925657e+00 -1.93719685e-01 6.57485664e-01 1.76178247e-01
1.00747740e+00 2.61687130e-01 2.47183636e-01 2.58049250e-01
7.88579166e-01 2.30415180e-01 5.66164494e-01 -3.40398580e-01
-8.49853814e-01 -7.30983198e-01 7.63758183e-01 1.12053299e+00
2.87772208e-01 4.99100648e-02 -6.63350224e-01 6.04949653e-01
-1.76118386e+00 -6.63022995e-01 -7.39935040e-02 1.92877162e+00
1.14939773e+00 1.54212013e-01 -1.89433768e-02 1.68884784e-01
7.60533690e-01 3.80150169e-01 -7.34923601e-01 6.76549673e-02
-4.84320492e-01 1.72778517e-01 6.96185648e-01 1.82699308e-01
-1.21747780e+00 1.19142091e+00 6.39662790e+00 9.56910908e-01
-1.02701759e+00 6.13157809e-01 5.98876357e-01 2.98187941e-01
-1.16481200e-01 5.09455130e-02 -9.23490524e-01 5.16094983e-01
1.02841187e+00 4.45904046e-01 4.51432735e-01 1.16234469e+00
-3.51405442e-01 2.23705769e-02 -1.10786211e+00 1.12243462e+00
3.71094882e-01 -1.46154773e+00 1.66222006e-01 -7.63451606e-02
1.09736466e+00 4.07085150e-01 1.67310223e-01 4.28464711e-01
4.75159168e-01 -9.52092886e-01 1.00771177e+00 1.43299997e-01
1.22061896e+00 -2.42705554e-01 4.84282583e-01 4.56329435e-01
-1.19359899e+00 -2.03296229e-01 -5.23771226e-01 8.07608292e-02
6.77623728e-04 4.40325707e-01 -6.67847931e-01 1.73242405e-01
6.44653857e-01 1.05549657e+00 -6.11215711e-01 5.97706676e-01
-5.12797534e-01 1.16684747e+00 4.41856980e-02 4.25732404e-01
2.81905681e-01 2.68726982e-02 6.11882173e-02 1.02832484e+00
-2.56319195e-01 -2.96603143e-01 1.53097942e-01 9.41424668e-01
-4.50936019e-01 -3.05031776e-01 -4.21795368e-01 -5.14263287e-02
5.29596031e-01 1.10429931e+00 -5.16560912e-01 -5.81040978e-01
-8.07405472e-01 1.27060986e+00 5.18645704e-01 3.48878384e-01
-6.81608677e-01 6.06160201e-02 2.23691940e-01 -5.30407615e-02
6.24546051e-01 1.07301891e-01 -2.52879888e-01 -1.38865519e+00
-9.84642357e-02 -1.10112941e+00 2.85993963e-01 -1.07236803e+00
-1.24562120e+00 8.22234452e-01 -4.84909296e-01 -1.59420884e+00
-1.39085844e-01 -6.15390420e-01 -2.71563947e-01 4.92562205e-01
-2.11992145e+00 -1.74147868e+00 -4.85262692e-01 9.09929395e-01
7.97856152e-01 -6.21820688e-01 9.23820078e-01 2.55768448e-01
-6.75724208e-01 8.05894971e-01 1.96022719e-01 2.72194624e-01
7.17900693e-01 -1.21682119e+00 3.72796446e-01 9.57852244e-01
5.87924719e-01 1.24217957e-01 2.46517137e-01 -4.65561897e-01
-1.18459892e+00 -1.48930609e+00 8.19011867e-01 -2.64937311e-01
5.74875236e-01 -5.55984557e-01 -8.44189286e-01 1.07479084e+00
3.76871824e-01 4.09249693e-01 5.73880374e-01 -5.18300235e-01
-8.58240247e-01 -2.51267552e-01 -1.03686571e+00 4.25541192e-01
1.07621193e+00 -9.21023011e-01 -7.52598882e-01 4.18800145e-01
7.62831986e-01 -2.76487321e-01 -6.00011230e-01 2.11675078e-01
4.06765014e-01 -9.28759634e-01 9.13922429e-01 -3.05439234e-01
6.17199898e-01 -3.99179727e-01 -3.17688763e-01 -1.24381149e+00
-3.34783584e-01 -4.23476160e-01 -4.69791293e-01 1.21466672e+00
3.69308025e-01 -5.05098820e-01 9.95779514e-01 6.55233562e-02
-2.39838004e-01 -5.92031121e-01 -6.69899881e-01 -7.99391627e-01
-2.24872813e-01 -3.06068808e-01 4.17047232e-01 1.12223661e+00
-4.18213665e-01 4.42678064e-01 -8.64655256e-01 1.49719656e-01
1.18548441e+00 1.44337326e-01 4.89594400e-01 -1.04876471e+00
-3.59378606e-01 -5.34320287e-02 -1.69035211e-01 -1.28264380e+00
4.60591018e-01 -1.25041163e+00 1.74555153e-01 -1.39269018e+00
4.27841157e-01 -3.31695914e-01 -5.58060527e-01 9.18059409e-01
-8.47311914e-02 7.69558907e-01 5.93003184e-02 5.41185737e-01
-6.42090142e-01 8.54577899e-01 1.20648861e+00 -5.01277566e-01
1.89315557e-01 -1.90156311e-01 -5.44239759e-01 7.35503495e-01
5.72423875e-01 -6.51071668e-01 -5.24601877e-01 -4.67531681e-01
-2.49114722e-01 -1.83513865e-01 4.32256997e-01 -1.06624866e+00
4.62666661e-01 1.30135864e-01 4.86462861e-01 -6.99622869e-01
3.83002758e-01 -9.38240945e-01 -2.22555567e-02 4.33785975e-01
-5.96542358e-01 -3.71372491e-01 -2.48610422e-01 6.12323701e-01
-5.03286481e-01 -1.71606362e-01 9.59672511e-01 -1.54397726e-01
-1.12171161e+00 3.16964090e-01 -3.02914113e-01 -5.18375263e-02
9.20682788e-01 -3.02053541e-01 -4.14925814e-01 -2.07205236e-01
-6.37114406e-01 3.87695432e-01 3.83016378e-01 2.82270610e-01
9.60489810e-01 -1.38693643e+00 -6.67429030e-01 5.57643235e-01
5.91377437e-01 -6.07599923e-03 2.76060134e-01 7.22934663e-01
-2.41340384e-01 5.72257861e-02 -3.26844543e-01 -5.90154052e-01
-1.10644054e+00 6.51241243e-01 3.47568542e-01 -1.58373509e-02
-7.62855828e-01 1.07413447e+00 3.22645724e-01 -4.50221270e-01
6.34957671e-01 3.83066870e-02 -2.57236827e-02 -1.39520634e-02
7.45974541e-01 1.00636698e-01 3.49195749e-02 -7.44983017e-01
-2.19437242e-01 2.00674117e-01 -2.00316951e-01 3.48505154e-02
1.34794188e+00 -3.90708476e-01 -7.31662428e-03 5.63438714e-01
1.10923374e+00 -4.82973188e-01 -1.76084816e+00 -9.12298799e-01
-3.21193278e-01 -6.24016859e-03 5.73262572e-01 -8.32392573e-01
-1.31132495e+00 9.54829991e-01 7.99940586e-01 -3.57164055e-01
1.13418114e+00 1.38749138e-01 8.15264285e-01 3.37125629e-01
4.06510830e-01 -1.07057309e+00 1.17132775e-01 2.34065846e-01
6.29082918e-01 -1.68843138e+00 -1.71652213e-01 -2.29710251e-01
-1.19174373e+00 7.30914295e-01 7.26372957e-01 -2.46194806e-02
7.81585872e-01 3.35947692e-01 4.75392908e-01 -3.18287387e-02
-8.62075329e-01 -4.33637917e-01 1.74634606e-01 7.20648050e-01
9.96373668e-02 2.58377977e-02 2.83002317e-01 5.09875834e-01
8.43906030e-02 2.18058944e-01 4.43700105e-01 9.82346356e-01
-3.85063410e-01 -9.55246747e-01 -1.23253971e-01 4.19645727e-01
-9.91584957e-02 -3.34672004e-01 -2.06072658e-01 4.41644847e-01
1.24116383e-01 9.69157100e-01 2.33915374e-02 -6.90401614e-01
-1.03045866e-01 3.63217473e-01 3.19116265e-01 -5.77538490e-01
-5.25221527e-01 2.20303936e-03 -1.56938657e-01 -5.04448116e-01
-7.22602606e-01 -1.76942766e-01 -1.02373099e+00 -3.70006226e-02
-3.83370697e-01 -1.03710443e-01 6.82504773e-01 9.33598757e-01
2.67057478e-01 4.77857202e-01 9.20275450e-01 -6.05698109e-01
-7.66107142e-01 -9.77888763e-01 -8.07795584e-01 2.95780391e-01
5.45923471e-01 -5.47429800e-01 -2.78230190e-01 4.06722426e-01]
|
[9.570018768310547, 1.5485973358154297]
|
63f86b7d-dda4-43bd-a375-e6e7e8d4a29a
|
original-loop-closure-detection-algorithm-for
|
1707.04771
| null |
http://arxiv.org/abs/1707.04771v1
|
http://arxiv.org/pdf/1707.04771v1.pdf
|
Original Loop-closure Detection Algorithm for Monocular vSLAM
|
Vision-based simultaneous localization and mapping (vSLAM) is a
well-established problem in mobile robotics and monocular vSLAM is one of the
most challenging variations of that problem nowadays. In this work we study one
of the core post-processing optimization mechanisms in vSLAM, e.g. loop-closure
detection. We analyze the existing methods and propose original algorithm for
loop-closure detection, which is suitable for dense, semi-dense and
feature-based vSLAM methods. We evaluate the algorithm experimentally and show
that it contribute to more accurate mapping while speeding up the monocular
vSLAM pipeline to the extent the latter can be used in real-time for
controlling small multi-rotor vehicle (drone).
|
['Konstantin Yakovlev', 'Andrey Bokovoy']
|
2017-07-15
| null | null | null | null |
['loop-closure-detection']
|
['computer-vision']
|
[-1.78801179e-01 -3.49648565e-01 1.52221814e-01 -7.57981464e-02
-1.60266295e-01 -7.39045382e-01 7.99697220e-01 8.13623592e-02
-7.44722545e-01 6.69562995e-01 -3.55905890e-01 -3.74444604e-01
-9.37561318e-02 -5.36506116e-01 -7.73201942e-01 -1.88211173e-01
6.71339333e-02 7.82943666e-01 6.52944982e-01 -4.53359216e-01
3.63261402e-01 1.14639926e+00 -2.03781605e+00 -3.06849688e-01
5.82123458e-01 7.14644253e-01 8.45589340e-01 6.62215650e-01
3.85905743e-01 2.18035147e-01 -1.09903112e-01 3.77425373e-01
3.68059278e-01 -5.97822256e-02 -5.07912099e-01 -3.85612657e-04
7.92897880e-01 -6.67097345e-02 -1.68945923e-01 1.08235550e+00
4.07861710e-01 1.25813112e-01 4.24261063e-01 -1.28993630e+00
2.98182338e-01 -2.11257696e-01 -6.20931685e-01 1.61082983e-01
5.29540598e-01 9.32478718e-03 4.60450828e-01 -1.33387423e+00
1.04168391e+00 1.25626469e+00 8.33501816e-01 -3.40661108e-02
-1.03813446e+00 -4.09578443e-01 -2.26451188e-01 3.62123638e-01
-1.42889929e+00 -5.90302229e-01 3.87246460e-01 -6.41151309e-01
9.06769037e-01 6.73839301e-02 6.82066798e-01 3.29393983e-01
3.97735447e-01 4.11268681e-01 1.20022368e+00 -4.91552651e-01
7.80180767e-02 1.12201929e-01 -5.94057590e-02 1.06855416e+00
6.39847219e-01 4.92363185e-01 -4.64816570e-01 1.51147783e-01
8.62489998e-01 -1.26305208e-01 -3.06205183e-01 -1.29504168e+00
-1.58456564e+00 7.59384871e-01 5.17400682e-01 1.59974113e-01
-1.90720841e-01 2.73989022e-01 1.45979375e-01 2.90654182e-01
1.84832215e-01 3.71835530e-01 -4.23334241e-01 -8.10633898e-02
-1.04739046e+00 4.97900754e-01 4.48117971e-01 1.19910729e+00
1.30308771e+00 -1.48489520e-01 2.48340130e-01 4.11440015e-01
3.69685322e-01 7.54817665e-01 2.65659600e-01 -9.28579330e-01
2.46505141e-01 6.07951760e-01 4.15774345e-01 -1.01624060e+00
-8.40833604e-01 -3.50764334e-01 -3.16290021e-01 7.37949133e-01
3.25972408e-01 -1.47199214e-01 -8.19541276e-01 1.09341812e+00
5.44488072e-01 1.03470802e-01 -2.00542286e-01 9.59022105e-01
6.70147479e-01 3.80209416e-01 -5.63652337e-01 -1.19912580e-01
1.19189310e+00 -1.03471434e+00 -4.79843616e-01 -7.06783354e-01
6.33808672e-01 -1.02840972e+00 5.70696771e-01 2.54479915e-01
-5.98442256e-01 -6.80404484e-01 -1.18015695e+00 -2.61231899e-01
-5.12754142e-01 6.83022261e-01 4.40071851e-01 3.38487625e-01
-1.27326584e+00 2.49071389e-01 -8.82535756e-01 -9.46774840e-01
2.61000567e-03 4.71595883e-01 -9.45693970e-01 -1.20462783e-01
-7.38286972e-01 1.38557935e+00 4.05280888e-01 1.48454532e-01
-6.88792527e-01 -3.15387398e-01 -1.17425144e+00 -5.56687534e-01
4.83860552e-01 -6.77509010e-01 9.55253601e-01 -3.20129126e-01
-1.27608144e+00 1.18267512e+00 -4.68266368e-01 -7.65942812e-01
7.84633994e-01 -4.55915272e-01 2.41958231e-01 1.16053618e-01
2.89409608e-01 1.01772046e+00 7.47881234e-01 -1.26776755e+00
-9.29737270e-01 -4.90686804e-01 -6.89204261e-02 3.60745162e-01
4.14037764e-01 -1.65219128e-01 -4.06168848e-01 9.90211312e-03
4.30549473e-01 -1.33597577e+00 -2.50861913e-01 2.41156802e-01
-1.29744470e-01 3.34269442e-02 1.22129476e+00 -3.79787177e-01
6.80455446e-01 -1.81382418e+00 2.05151498e-01 -7.76013033e-03
5.27159348e-02 3.41770679e-01 1.75410315e-01 5.46599627e-01
3.26788396e-01 -6.70690179e-01 -1.11793987e-01 -4.93609428e-01
-1.70719877e-01 3.76829863e-01 -1.39691666e-01 1.12608790e+00
-1.04506433e-01 9.49251354e-01 -9.09783661e-01 -2.55196571e-01
9.92864311e-01 3.83683771e-01 -3.49663943e-01 9.74302888e-02
-7.20191598e-02 6.60747528e-01 4.14418094e-02 5.17671764e-01
9.19899285e-01 5.34057856e-01 -2.52539545e-01 -1.55499056e-01
-9.71276999e-01 -1.10963210e-01 -1.32736194e+00 1.99937296e+00
-5.23771405e-01 1.20152140e+00 5.29400527e-01 -7.08337307e-01
1.03611422e+00 -3.07541877e-01 2.00207278e-01 -5.30226827e-01
2.69864857e-01 4.61255282e-01 -3.38037699e-01 -2.90352613e-01
1.00987792e+00 1.90295860e-01 6.09004796e-02 -5.66381179e-02
-3.41545008e-02 -3.77537310e-01 3.07541847e-01 -6.81730434e-02
9.08573508e-01 5.24899721e-01 7.61492789e-01 -5.38151681e-01
6.96865380e-01 7.20761955e-01 2.25086734e-01 6.49441838e-01
-1.66570634e-01 6.87669516e-01 -1.38232317e-02 -2.62807548e-01
-9.22350526e-01 -5.23412168e-01 -1.66855648e-01 6.16311431e-01
7.06533670e-01 -2.07784355e-01 -4.41511601e-01 -2.70277649e-01
3.74395669e-01 2.20232397e-01 -4.32665199e-01 2.81994820e-01
-6.27927363e-01 -1.90439910e-01 1.95520490e-01 1.85143173e-01
5.01653433e-01 -6.87435627e-01 -1.47246730e+00 1.61047608e-01
1.15485698e-01 -1.30419731e+00 9.65596586e-02 3.43373448e-01
-7.01995373e-01 -1.29728568e+00 -4.99291837e-01 -8.07668507e-01
5.27807236e-01 9.72966433e-01 7.59318769e-01 -1.60514802e-01
-4.15358126e-01 3.97136688e-01 -2.56534427e-01 -4.52914029e-01
-2.22654685e-01 1.26355384e-02 3.38429987e-01 -4.12724167e-01
1.83640063e-01 -2.38284901e-01 -3.21519881e-01 4.60996866e-01
-3.66284579e-01 5.73226139e-02 6.45675123e-01 5.20157278e-01
7.52391219e-01 -2.75635570e-01 -9.50936750e-02 -6.88076794e-01
2.02939197e-01 -2.23925754e-01 -1.41461134e+00 -1.83525681e-01
-4.60175604e-01 -7.46792704e-02 3.46834004e-01 8.00020024e-02
-4.42150861e-01 8.07170331e-01 -6.33769333e-02 -5.90044260e-01
-3.81944776e-01 3.96157473e-01 4.73233052e-02 -8.26938808e-01
6.36779428e-01 1.44264579e-01 2.42448121e-01 -5.27339876e-01
4.92509723e-01 5.22942245e-01 7.10314214e-01 2.03406841e-01
1.05797362e+00 9.05128598e-01 4.41551626e-01 -1.08305168e+00
-3.46381634e-01 -1.02991021e+00 -1.17932189e+00 -2.92730212e-01
5.89355588e-01 -1.15527725e+00 -3.41120213e-01 2.37902731e-01
-1.27367473e+00 -2.50262469e-01 4.02340889e-02 5.52480400e-01
-6.82219863e-01 3.10613394e-01 3.71397473e-02 -7.20779955e-01
-3.31658535e-02 -1.22616041e+00 1.35672581e+00 1.13883123e-01
7.61957541e-02 -8.88379633e-01 4.18025285e-01 -2.81140842e-02
2.72221237e-01 4.09621447e-01 1.97729096e-02 2.16632243e-02
-6.61004484e-01 -3.69754374e-01 -2.29716837e-01 -1.48011819e-01
-2.10716143e-01 -1.18035614e-01 -8.02211523e-01 -5.99888146e-01
-3.45732808e-01 -1.20297289e-02 8.52450073e-01 3.00290316e-01
2.84011960e-01 4.07653004e-01 -7.38535345e-01 6.98331594e-01
1.71404898e+00 -1.08776502e-01 4.83423948e-01 7.76144922e-01
8.28034997e-01 5.22669852e-01 1.43306434e+00 1.65165290e-01
4.68671560e-01 1.15644109e+00 8.66647601e-01 -8.95608589e-02
-1.37677774e-01 -2.79228330e-01 3.97684753e-01 4.38391268e-01
2.86141485e-01 3.09705585e-01 -1.02486539e+00 6.77807510e-01
-2.15728021e+00 -4.66653317e-01 -7.74953961e-01 2.17398834e+00
5.56256473e-02 -1.98812962e-01 -3.01834922e-02 1.31521374e-01
6.50121927e-01 -9.55298170e-02 -2.48702735e-01 -3.75773251e-01
-7.47244060e-02 -5.69689311e-02 1.07407582e+00 8.46125185e-01
-1.41961050e+00 1.33952904e+00 6.07441568e+00 4.00774658e-01
-1.24181318e+00 1.62829250e-01 -3.77407134e-01 3.66490394e-01
3.51342767e-01 2.18486339e-01 -1.12966549e+00 4.22725454e-02
4.89122778e-01 1.20855488e-01 2.58787781e-01 1.14714766e+00
4.30315316e-01 -1.02850091e+00 -8.70373845e-01 1.31437266e+00
2.10938290e-01 -1.35979581e+00 -4.28716749e-01 1.78364918e-01
7.42069483e-01 5.63342869e-01 -4.29448664e-01 -7.63361454e-02
-1.01193666e-01 -7.08902597e-01 1.01169181e+00 2.15218350e-01
8.16783071e-01 -7.02079177e-01 8.28318357e-01 7.69857407e-01
-1.50015557e+00 -6.20811135e-02 -6.89265251e-01 -3.48975182e-01
2.60912418e-01 6.96030319e-01 -9.68983769e-01 5.40758729e-01
5.30479431e-01 9.19674397e-01 -8.54001522e-01 1.43028235e+00
-2.80071259e-01 -1.32058576e-01 -6.21742368e-01 6.92099705e-02
3.06927562e-01 -3.09606522e-01 8.90476286e-01 1.29017186e+00
4.62566793e-01 -5.93433857e-01 3.07551086e-01 6.56982660e-01
4.97800022e-01 -1.17680967e-01 -1.13387871e+00 4.37838674e-01
4.16115969e-01 1.54796672e+00 -8.59552860e-01 4.08586208e-03
-2.27223456e-01 9.49175298e-01 3.14810753e-01 -1.76250488e-01
-5.57489932e-01 -4.75077540e-01 6.47316694e-01 3.09219509e-01
4.95750755e-01 -8.65915596e-01 -1.68853700e-02 -9.83739138e-01
-3.97727117e-02 -2.68090963e-01 -2.32502997e-01 -9.33001459e-01
-3.97224218e-01 4.62782234e-01 -4.81142960e-02 -1.55436921e+00
-3.66082907e-01 -8.49901915e-01 -2.03777194e-01 6.98008180e-01
-2.04941845e+00 -1.38363397e+00 -8.19268703e-01 5.85183144e-01
5.67669094e-01 -3.51202488e-02 6.41261876e-01 2.22815529e-01
-6.23875447e-02 -1.25634253e-01 8.76922756e-02 -2.67870903e-01
7.14749634e-01 -1.04800558e+00 3.75785887e-01 1.37873602e+00
3.67224395e-01 5.87736547e-01 9.47358429e-01 -7.15552211e-01
-1.79163563e+00 -1.27380073e+00 7.63266683e-01 -5.82221746e-01
5.63162446e-01 -4.25747067e-01 -3.55480134e-01 7.14292407e-01
-4.65213098e-02 2.13385746e-01 -3.38459551e-01 -4.58171368e-01
2.92881221e-01 8.24693125e-04 -1.01056600e+00 2.18764350e-01
9.90443528e-01 -3.68615568e-01 -6.08757675e-01 2.65320837e-01
1.17728271e-01 -9.45018411e-01 -1.98069096e-01 4.23409164e-01
4.58875000e-01 -1.31927991e+00 8.99922073e-01 3.50278109e-01
-1.38770431e-01 -1.05612743e+00 2.56871320e-02 -1.37280142e+00
-9.15125310e-02 -6.09855652e-01 5.57513945e-02 8.24490428e-01
-2.15737391e-02 -7.12957084e-01 6.48158669e-01 -4.96154666e-01
-4.06442255e-01 -3.00196201e-01 -1.12940156e+00 -7.91209757e-01
-4.55489963e-01 -2.84470171e-01 4.26291302e-02 5.17389357e-01
-2.02376977e-01 2.25005448e-01 -3.09781700e-01 4.67686176e-01
6.21764481e-01 2.79859424e-01 1.35265207e+00 -1.37407708e+00
1.77569419e-01 -2.18417048e-02 -9.71066833e-01 -1.32351387e+00
1.07912101e-01 -7.64335871e-01 4.61079627e-01 -1.94584382e+00
-3.12565416e-01 -5.32689512e-01 5.04039586e-01 1.20586582e-01
2.79055446e-01 6.38697922e-01 5.48563674e-02 2.38328442e-01
-6.37958050e-01 2.07811385e-01 8.17126155e-01 4.21520412e-01
-8.34719837e-02 -1.17202653e-02 2.50601862e-02 7.52499640e-01
5.30997515e-01 -2.28849754e-01 -2.44287223e-01 -3.03721339e-01
1.60336539e-01 -2.47701719e-01 6.53809547e-01 -1.38815141e+00
6.54453218e-01 7.73954391e-02 9.60330218e-02 -1.18834984e+00
4.07331407e-01 -1.05778885e+00 6.25827163e-02 6.87618971e-01
4.01615858e-01 3.79896581e-01 3.76660585e-01 6.02539659e-01
-2.97530502e-01 -1.05041690e-01 8.67476702e-01 -2.50372112e-01
-1.46542549e+00 5.14667183e-02 -4.97355103e-01 -2.91562498e-01
1.35686111e+00 -4.43532646e-01 -2.02601522e-01 -1.64569005e-01
-3.63394469e-01 2.39666775e-01 9.26498175e-01 4.79791313e-01
7.80131519e-01 -1.01007795e+00 -5.50719738e-01 4.18157816e-01
3.87842774e-01 2.50348955e-01 -4.18450594e-01 1.33547473e+00
-1.32629097e+00 8.29709232e-01 -3.94601643e-01 -1.12437081e+00
-1.62922359e+00 4.80658859e-01 2.89109319e-01 1.82298690e-01
-6.98167801e-01 7.19765007e-01 -4.77316193e-02 -6.47967637e-01
5.21043874e-02 -3.66590589e-01 -3.87919426e-01 2.47323029e-02
2.81739563e-01 6.47760868e-01 2.36232281e-01 -1.23325610e+00
-6.97328985e-01 1.11012161e+00 4.82202828e-01 -1.28664479e-01
1.19892168e+00 -6.12413824e-01 -4.73761201e-01 4.78368104e-01
9.44311678e-01 2.26231664e-01 -1.19438815e+00 1.37311205e-01
1.19810238e-01 -4.78317946e-01 3.50627571e-01 -2.19209284e-01
-5.94285131e-01 8.82602870e-01 8.71015668e-01 -2.06771031e-01
5.29259443e-01 -1.57912955e-01 3.96717280e-01 5.35927474e-01
1.16596961e+00 -1.04299045e+00 -5.83143234e-01 7.93988883e-01
8.54376674e-01 -1.32630181e+00 4.50081408e-01 -6.96224868e-01
-2.00213209e-01 1.24791002e+00 3.76964390e-01 -4.22622263e-01
5.37930191e-01 4.13868278e-01 1.11604847e-01 -2.09780961e-01
-2.53667653e-01 -6.71741903e-01 1.34419918e-01 8.17139208e-01
-9.12291855e-02 -5.42445667e-02 -2.21986085e-01 -3.70616287e-01
-7.15024397e-02 -2.89246682e-02 5.76640844e-01 1.23226142e+00
-9.06829119e-01 -8.74293268e-01 -7.44129300e-01 -7.69526279e-03
2.40358487e-01 1.36868253e-01 -4.17854667e-01 1.01863635e+00
4.35146481e-01 9.05789912e-01 1.26550812e-02 -3.91791880e-01
5.50763249e-01 -3.39677721e-01 6.29995942e-01 -6.98523104e-01
-3.47006559e-01 7.92054832e-02 4.21350636e-02 -9.45105374e-01
-5.38858294e-01 -8.43406916e-01 -1.23624015e+00 -6.42658547e-02
-5.72923720e-01 -8.19902420e-02 1.16633058e+00 1.01602912e+00
4.86831784e-01 7.20591322e-02 3.83418888e-01 -1.55155647e+00
-3.06953993e-02 -7.69338191e-01 -4.56578821e-01 -9.68355164e-02
5.32270312e-01 -1.23433530e+00 -2.77237654e-01 -1.70435697e-01]
|
[7.398834705352783, -2.0896337032318115]
|
bf22dafb-9603-4063-843a-cc6b45430d63
|
pure-transformer-with-integrated-experts-for
|
2211.04963
| null |
https://arxiv.org/abs/2211.04963v1
|
https://arxiv.org/pdf/2211.04963v1.pdf
|
Pure Transformer with Integrated Experts for Scene Text Recognition
|
Scene text recognition (STR) involves the task of reading text in cropped images of natural scenes. Conventional models in STR employ convolutional neural network (CNN) followed by recurrent neural network in an encoder-decoder framework. In recent times, the transformer architecture is being widely adopted in STR as it shows strong capability in capturing long-term dependency which appears to be prominent in scene text images. Many researchers utilized transformer as part of a hybrid CNN-transformer encoder, often followed by a transformer decoder. However, such methods only make use of the long-term dependency mid-way through the encoding process. Although the vision transformer (ViT) is able to capture such dependency at an early stage, its utilization remains largely unexploited in STR. This work proposes the use of a transformer-only model as a simple baseline which outperforms hybrid CNN-transformer models. Furthermore, two key areas for improvement were identified. Firstly, the first decoded character has the lowest prediction accuracy. Secondly, images of different original aspect ratios react differently to the patch resolutions while ViT only employ one fixed patch resolution. To explore these areas, Pure Transformer with Integrated Experts (PTIE) is proposed. PTIE is a transformer model that can process multiple patch resolutions and decode in both the original and reverse character orders. It is examined on 7 commonly used benchmarks and compared with over 20 state-of-the-art methods. The experimental results show that the proposed method outperforms them and obtains state-of-the-art results in most benchmarks.
|
['Jung-jae Kim', 'Adams Wai-Kin Kong', 'Yew Lee Tan']
|
2022-11-09
| null | null | null | null |
['scene-text-recognition']
|
['computer-vision']
|
[ 4.95512962e-01 -2.43319318e-01 2.30908599e-02 -6.66243583e-02
-4.56564218e-01 -2.79796898e-01 7.39543617e-01 -3.10057282e-01
-2.47851178e-01 3.20324421e-01 2.02332243e-01 -2.88526624e-01
1.75197005e-01 -7.66004205e-01 -7.88018584e-01 -6.64422452e-01
5.85341930e-01 2.01844841e-01 6.39613986e-01 -4.62691456e-01
5.41113496e-01 2.69986600e-01 -1.43387353e+00 9.65822220e-01
7.99920321e-01 9.93818939e-01 5.79493463e-01 8.39766324e-01
-3.42091680e-01 1.32684350e+00 -4.49616283e-01 -5.93001842e-01
1.73120081e-01 -5.13938904e-01 -5.72409630e-01 1.53274864e-01
4.45552468e-01 -4.10518140e-01 -5.45061946e-01 7.62342095e-01
5.39401591e-01 -2.64060766e-01 6.27847731e-01 -6.89053774e-01
-7.08226442e-01 6.85852408e-01 -7.15365767e-01 3.33521992e-01
3.44576031e-01 -2.00828705e-02 8.75006676e-01 -1.11017501e+00
3.08372915e-01 1.12683618e+00 8.19454253e-01 1.39661729e-01
-9.05605197e-01 -4.30602610e-01 1.50340274e-01 5.36352515e-01
-1.23403037e+00 -3.15796137e-01 7.75598466e-01 -2.59629488e-01
1.60656333e+00 7.92797729e-02 4.77025688e-01 1.12187612e+00
6.80293202e-01 1.10799015e+00 1.16330314e+00 -5.41369975e-01
-2.28263050e-01 7.01266155e-02 5.87496981e-02 6.81640029e-01
-1.96052119e-01 -8.47012028e-02 -7.90826201e-01 6.10470414e-01
8.06080341e-01 5.97106367e-02 -3.15943450e-01 4.74331565e-02
-1.13209796e+00 5.75965822e-01 5.14604151e-01 3.99352431e-01
-3.74430120e-01 -1.13187492e-01 5.58150709e-01 4.00076121e-01
2.69321799e-01 -7.42726326e-02 -2.49477401e-01 -1.88482076e-01
-1.32224131e+00 -7.10133165e-02 5.93661785e-01 9.39781249e-01
4.34366405e-01 4.02141839e-01 -3.76083016e-01 1.10076439e+00
2.12518752e-01 3.54523599e-01 6.77779138e-01 -1.06396995e-01
9.45048630e-01 8.81508887e-01 -3.69168222e-01 -1.01123703e+00
-3.27640884e-02 -4.92958814e-01 -1.26922250e+00 1.05056040e-01
1.60256371e-01 3.03672463e-01 -1.25388777e+00 1.00623536e+00
-2.41881669e-01 5.31901047e-02 2.67670959e-01 7.33329654e-01
9.12447393e-01 1.06849945e+00 -3.08595210e-01 9.01295394e-02
1.43061686e+00 -1.38243163e+00 -7.32059956e-01 -3.27061713e-01
1.86731726e-01 -1.13741326e+00 9.10283148e-01 7.25546241e-01
-1.16612101e+00 -8.88118148e-01 -1.29579866e+00 -3.38892311e-01
-5.53141057e-01 5.40860116e-01 -8.32399074e-03 5.11645734e-01
-1.20619178e+00 4.82136846e-01 -7.13195622e-01 -5.14681756e-01
2.32374892e-01 2.06222579e-01 -1.74933329e-01 -2.32216284e-01
-1.19462240e+00 9.37634349e-01 2.98405498e-01 4.27922755e-01
-8.40511084e-01 -2.69276857e-01 -7.72651672e-01 1.98389411e-01
2.40578130e-01 -3.90857369e-01 1.24351704e+00 -1.16512430e+00
-1.77269137e+00 4.64523077e-01 -1.57403171e-01 -6.07639611e-01
7.48901963e-01 -3.20176750e-01 -3.41755629e-01 2.73992747e-01
-1.49863705e-01 5.36410034e-01 1.28057957e+00 -9.22090828e-01
-7.50708222e-01 -6.26507327e-02 -4.05630544e-02 4.66148853e-01
-3.04209828e-01 1.11057557e-01 -7.10425138e-01 -8.29970539e-01
1.10758781e-01 -6.59739316e-01 2.59579450e-01 -1.59487933e-01
-5.44246495e-01 -9.45361331e-02 1.25458133e+00 -8.84537578e-01
1.25603342e+00 -1.98992419e+00 1.02438360e-01 -2.54622668e-01
8.39816332e-02 5.39696753e-01 8.84498295e-04 6.88402653e-01
-1.01467401e-01 -3.10561396e-02 -1.70953479e-02 -4.13734853e-01
-1.18973285e-01 1.57885283e-01 -4.01323676e-01 1.12447456e-01
3.58001560e-01 9.25657451e-01 -3.49167556e-01 -5.24161041e-01
4.81233716e-01 7.45915055e-01 -3.02183300e-01 2.09484473e-01
-6.74498528e-02 1.90088246e-02 -2.96441048e-01 6.50306940e-01
7.99924731e-01 -3.27237099e-01 -3.44723091e-02 -4.67057437e-01
-4.57330585e-01 2.16149062e-01 -9.28720117e-01 1.24942505e+00
-3.14545035e-01 9.62487042e-01 -2.57470161e-01 -1.10715234e+00
1.06620884e+00 4.44063753e-01 6.68117628e-02 -1.05216229e+00
2.18052879e-01 1.82026342e-01 -1.75919816e-01 -4.80036378e-01
8.55136156e-01 1.31124064e-01 2.60751337e-01 1.10482574e-01
-7.87137076e-02 8.74292552e-02 1.46289691e-01 3.89637835e-02
1.00261319e+00 2.52289116e-01 1.55549482e-01 -4.10792455e-02
7.23932445e-01 -6.21973313e-02 3.39452207e-01 6.89282715e-01
6.77580386e-02 1.01374781e+00 4.95631248e-01 -4.28002089e-01
-1.45516479e+00 -7.62152851e-01 1.54920656e-03 8.87437403e-01
1.48260668e-01 -2.89436281e-01 -6.30120397e-01 -4.39022392e-01
-5.34735680e-01 6.04768455e-01 -6.92131042e-01 5.73086813e-02
-6.49523914e-01 -6.41716897e-01 8.41381013e-01 7.47912765e-01
1.44348717e+00 -1.14933741e+00 -8.03322017e-01 2.26532385e-01
-3.16713989e-01 -1.36767137e+00 -3.72278184e-01 4.21536148e-01
-8.12795699e-01 -8.91603589e-01 -1.08396387e+00 -9.39813316e-01
4.70742613e-01 4.07650143e-01 8.92788410e-01 -1.34369805e-01
-6.83467314e-02 6.29744828e-02 -8.15079510e-01 -2.14014858e-01
-3.39690447e-01 1.33844763e-01 -5.87483168e-01 1.70948699e-01
2.90602773e-01 -2.00471997e-01 -5.57072282e-01 4.32675660e-01
-9.19484377e-01 6.47027493e-01 9.21235621e-01 1.03117907e+00
3.73833150e-01 2.31055528e-01 4.67335172e-02 -8.38249683e-01
6.25794232e-01 -1.25499651e-01 -3.88785303e-01 5.16572475e-01
-5.47534108e-01 -5.42990305e-02 1.12425411e+00 -4.33805496e-01
-1.30186725e+00 -1.04364902e-01 -2.18786508e-01 -4.72781330e-01
8.18166602e-03 6.11699581e-01 -2.84838006e-02 5.56586236e-02
4.29734588e-01 7.98830092e-01 -2.35146999e-01 -4.58453298e-01
-2.37052694e-01 7.88726687e-01 4.45534140e-01 -2.40606278e-01
5.48902452e-01 8.65781158e-02 -2.60741949e-01 -9.49359179e-01
-5.32454610e-01 -3.03047210e-01 -6.46266043e-01 -2.21893534e-01
8.27993691e-01 -9.88705099e-01 -2.76714534e-01 9.96863425e-01
-1.20946848e+00 -3.84829849e-01 1.43508866e-01 1.89575776e-01
-3.56171817e-01 4.85940456e-01 -8.68236363e-01 -5.52875876e-01
-5.51055014e-01 -1.53681588e+00 1.13038778e+00 2.83375382e-01
2.80418396e-01 -8.07826221e-01 -1.95922554e-01 3.50878477e-01
6.69883430e-01 -1.47993460e-01 9.32830930e-01 -3.62080604e-01
-5.88444352e-01 -1.25683457e-01 -5.67477584e-01 4.28764969e-01
4.80853654e-02 2.25311235e-01 -9.61472392e-01 -1.86748773e-01
-4.26831841e-02 -1.80297270e-01 1.01586366e+00 2.47468710e-01
1.02093613e+00 -1.50289372e-01 -5.20884246e-02 6.23574853e-01
1.72623897e+00 3.51409733e-01 1.35405385e+00 5.08162320e-01
7.58449018e-01 1.98617265e-01 3.39486867e-01 3.14823151e-01
3.83847028e-01 6.27994776e-01 2.86467284e-01 -1.93768486e-01
-2.62373477e-01 -4.05188620e-01 7.04721868e-01 1.12622643e+00
-3.94262038e-02 -7.66626656e-01 -8.97926569e-01 3.63272160e-01
-1.90114617e+00 -9.30357695e-01 -2.72506982e-01 1.94647920e+00
5.88783205e-01 3.67473692e-01 -2.45581746e-01 6.25626326e-01
7.41448522e-01 2.22169876e-01 -4.21816438e-01 -9.23461080e-01
-4.37119395e-01 3.24338108e-01 6.68424249e-01 2.18214780e-01
-9.37124550e-01 1.07458067e+00 6.05144787e+00 9.85812366e-01
-1.53429830e+00 -2.09904328e-01 6.81381047e-01 3.86970341e-01
1.81230679e-01 -1.33624762e-01 -9.52395022e-01 4.36699808e-01
8.56452942e-01 2.42184877e-01 3.18532854e-01 5.32251716e-01
1.97773099e-01 -1.53138265e-01 -8.42284858e-01 9.03860748e-01
2.64054716e-01 -1.24095678e+00 4.11873251e-01 -1.87926859e-01
6.13266468e-01 -2.16212124e-02 2.19128728e-01 3.26915383e-01
-1.32852405e-01 -1.22896421e+00 8.89635742e-01 5.79436600e-01
1.08005095e+00 -5.99540710e-01 9.10876572e-01 4.66566950e-01
-1.54348183e+00 -1.02171682e-01 -5.44485271e-01 4.71327230e-02
-1.04576014e-01 2.81946480e-01 -7.38326073e-01 7.41564751e-01
8.65538120e-01 1.06347358e+00 -8.00742686e-01 9.00875509e-01
-1.11446962e-01 7.09069073e-01 -2.14676127e-01 -3.36047038e-02
4.93524671e-01 1.76731832e-02 3.23688328e-01 1.54542208e+00
4.60681170e-01 -4.41853181e-02 -1.24948032e-01 6.01399243e-01
1.01209059e-01 1.26883700e-01 -5.86208045e-01 -1.50870867e-02
7.13304952e-02 9.84071553e-01 -8.73789966e-01 -3.75378370e-01
-6.32240415e-01 1.26690555e+00 1.05177537e-01 2.94707030e-01
-8.26992273e-01 -3.29754025e-01 -6.37319237e-02 1.03014708e-01
9.30303574e-01 1.66802050e-03 -3.06060702e-01 -1.11621714e+00
1.19421490e-01 -1.22826564e+00 2.13937595e-01 -1.18322837e+00
-9.34104800e-01 9.42047954e-01 -1.09331854e-01 -1.50495720e+00
-1.44154444e-01 -9.14164841e-01 -6.26713753e-01 8.25310349e-01
-1.79515851e+00 -1.48689008e+00 -4.50562924e-01 8.20876598e-01
1.34453726e+00 -3.33364248e-01 5.19336641e-01 2.35560104e-01
-7.99828887e-01 8.44928920e-01 2.37008303e-01 1.85819298e-01
5.96957564e-01 -1.03887522e+00 3.17169935e-01 1.09602821e+00
-1.67775732e-02 3.11838537e-01 5.38335085e-01 -5.83809018e-01
-1.49237871e+00 -1.02942443e+00 8.90898943e-01 4.07697298e-02
3.12691629e-01 -3.27013224e-01 -9.71344471e-01 6.33756757e-01
7.76184499e-01 -2.39686519e-01 3.73419374e-02 -5.34439564e-01
-4.75940108e-01 -2.35731214e-01 -8.99497449e-01 5.47546446e-01
5.01879811e-01 -4.97584492e-01 -5.03070176e-01 -1.55344173e-01
2.91753083e-01 -5.41578829e-01 -5.46896756e-01 3.28733146e-01
5.84618330e-01 -1.41994500e+00 6.89994216e-01 2.13515386e-01
1.00163913e+00 -3.25573772e-01 -2.07144842e-01 -9.32880461e-01
-3.36840659e-01 -3.40399325e-01 -1.53112095e-02 1.19211328e+00
4.47581232e-01 -5.90456843e-01 6.26405597e-01 4.24550250e-02
-1.56758577e-01 -9.02040303e-01 -7.30086923e-01 -4.24552888e-01
6.26190454e-02 -2.41180316e-01 2.59888351e-01 6.22130692e-01
-3.21871817e-01 6.06709599e-01 -6.63397551e-01 -9.46749747e-03
3.09286594e-01 -6.36279061e-02 5.98644972e-01 -8.28009486e-01
-1.28389835e-01 -5.35568178e-01 -4.03297335e-01 -1.47508717e+00
-3.01215082e-01 -5.02745688e-01 1.02405705e-01 -1.65760565e+00
3.27948898e-01 2.16772873e-02 -1.30006194e-01 4.09690678e-01
-2.72746291e-03 3.35121065e-01 3.43777329e-01 2.38782525e-01
-4.21825886e-01 5.56455314e-01 1.24998212e+00 -3.95813972e-01
1.29103720e-01 -1.47214115e-01 -4.07775849e-01 4.95369524e-01
8.49706888e-01 -1.62722170e-01 -5.34371495e-01 -6.98991537e-01
3.06245834e-01 1.23754419e-01 3.28345418e-01 -1.25636709e+00
6.46065295e-01 2.16293380e-01 7.06944525e-01 -1.19173443e+00
2.68932968e-01 -8.10936391e-01 1.50476798e-01 2.79832989e-01
-2.89569199e-01 5.24301767e-01 2.89661080e-01 3.86496603e-01
-5.02332032e-01 -2.94043928e-01 8.89235318e-01 -1.53502971e-01
-8.34296763e-01 1.43905915e-03 -6.85110450e-01 -2.59016782e-01
6.64755344e-01 -8.47647786e-01 -3.25572491e-01 -3.15365911e-01
-2.08353117e-01 1.84757803e-02 3.33756864e-01 4.19260085e-01
1.07543278e+00 -9.32294667e-01 -8.55425239e-01 2.37658784e-01
-4.79484648e-02 5.00206240e-02 2.46466056e-01 8.89477491e-01
-8.78283918e-01 6.00690842e-01 -2.69617260e-01 -8.12513590e-01
-1.29522038e+00 3.09366703e-01 4.65555638e-01 -6.96893752e-01
-9.39837754e-01 6.50734305e-01 2.69483805e-01 -1.38842538e-01
2.89773762e-01 -3.35001141e-01 -5.31112850e-01 -1.84369892e-01
4.79051232e-01 1.00067995e-01 1.96907535e-01 -7.85572827e-01
-2.55672522e-02 9.45458531e-01 -4.89486605e-01 6.31783679e-02
1.18741429e+00 -1.39096305e-01 1.76835731e-02 4.95562851e-01
1.06429946e+00 -3.19284737e-01 -1.26049519e+00 -3.63691360e-01
-2.90974528e-01 -3.15837234e-01 2.06760988e-01 -9.33434010e-01
-1.17738223e+00 1.25588071e+00 5.33600986e-01 1.36214510e-01
1.53802586e+00 -7.35768139e-01 7.95772135e-01 2.98703074e-01
7.97842443e-02 -1.19946373e+00 3.60103369e-01 9.73590016e-01
9.81664598e-01 -1.04276752e+00 -9.74711124e-03 -3.41119885e-01
-8.87521863e-01 1.58204961e+00 7.17238426e-01 -2.14019522e-01
4.14091974e-01 5.63972473e-01 1.02547091e-02 -4.49524671e-02
-1.04364717e+00 -9.71166417e-02 2.11793706e-01 2.82217592e-01
6.62125707e-01 -2.73360163e-01 -3.80503619e-03 1.92761660e-01
-1.81336790e-01 1.56763326e-02 5.10946751e-01 1.02338171e+00
-2.22944260e-01 -9.89211202e-01 -4.36284035e-01 5.07405996e-01
-5.59142232e-01 -4.05030966e-01 -3.97587121e-01 7.21968412e-01
-4.65922579e-02 8.67720068e-01 -7.36043677e-02 -5.34717560e-01
3.51380676e-01 -3.48735228e-02 3.82341534e-01 -1.32797614e-01
-9.27881002e-01 3.27600449e-01 -1.83670089e-01 -3.54548514e-01
-2.98352033e-01 -4.86073703e-01 -8.70685041e-01 -3.92843455e-01
-4.11593616e-01 -1.69461235e-01 5.86410761e-01 9.61580455e-01
6.41783625e-02 8.99818361e-01 6.57984376e-01 -7.71830618e-01
-2.98110783e-01 -1.11301541e+00 -3.20236653e-01 -3.36554460e-02
3.63517493e-01 -3.39590728e-01 -9.84270722e-02 3.12801778e-01]
|
[11.857671737670898, 2.1943657398223877]
|
5c0afaae-4bdb-49ab-9ea0-320f992c9672
|
odn-opening-the-deep-network-for-open-set
|
1901.07757
| null |
http://arxiv.org/abs/1901.07757v1
|
http://arxiv.org/pdf/1901.07757v1.pdf
|
ODN: Opening the Deep Network for Open-set Action Recognition
|
In recent years, the performance of action recognition has been significantly
improved with the help of deep neural networks. Most of the existing action
recognition works hold the \textit{closed-set} assumption that all action
categories are known beforehand while deep networks can be well trained for
these categories. However, action recognition in the real world is essentially
an \textit{open-set} problem, namely, it is impossible to know all action
categories beforehand and consequently infeasible to prepare sufficient
training samples for those emerging categories. In this case, applying
closed-set recognition methods will definitely lead to unseen-category errors.
To address this challenge, we propose the Open Deep Network (ODN) for the
open-set action recognition task. Technologically, ODN detects new categories
by applying a multi-class triplet thresholding method, and then dynamically
reconstructs the classification layer and "opens" the deep network by adding
predictors for new categories continually. In order to transfer the learned
knowledge to the new category, two novel methods, Emphasis Initialization and
Allometry Training, are adopted to initialize and incrementally train the new
predictor so that only few samples are needed to fine-tune the model. Extensive
experiments show that ODN can effectively detect and recognize new categories
with little human intervention, thus applicable to the open-set action
recognition tasks in the real world. Moreover, ODN can even achieve comparable
performance to some closed-set methods.
|
['Yao-Wei Wang', 'Yu Shu', 'Yixiong Zou', 'Yemin Shi', 'Qingsheng Yuan', 'Yonghong Tian']
|
2019-01-23
| null | null | null | null |
['open-set-action-recognition']
|
['computer-vision']
|
[ 7.58340240e-01 -9.46593750e-03 -2.81049848e-01 -3.27596426e-01
-1.70116603e-01 -4.77461129e-01 3.85043412e-01 -2.39654854e-01
-3.96474898e-01 7.38996267e-01 -1.95873603e-01 -2.20640544e-02
-1.51327968e-01 -9.50873494e-01 -5.26459157e-01 -9.53798175e-01
3.28574449e-01 6.23846412e-01 4.18348342e-01 -1.64744759e-03
1.68949917e-01 5.81975818e-01 -2.06165314e+00 4.33792681e-01
8.63725483e-01 1.28520954e+00 8.68582353e-02 2.11475223e-01
-1.98360041e-01 5.74501812e-01 -6.47649765e-01 -7.11569712e-02
4.89075124e-01 -4.00707960e-01 -7.88223386e-01 3.25452447e-01
4.83200073e-01 -5.07779956e-01 -3.26903552e-01 1.00857401e+00
3.90366137e-01 4.42090869e-01 5.49395323e-01 -1.15968788e+00
-5.57879210e-01 3.72442365e-01 -3.80996227e-01 1.26655743e-01
2.08741128e-01 3.75415921e-01 6.45959139e-01 -6.64325714e-01
3.28783840e-01 1.01462507e+00 5.60954571e-01 9.24150705e-01
-1.03330433e+00 -7.25025296e-01 3.81802529e-01 5.74193537e-01
-1.10643172e+00 -4.61802542e-01 7.21504807e-01 -4.44959641e-01
8.07444990e-01 3.51281792e-01 9.27874267e-01 1.38269341e+00
-5.89743070e-02 8.38217020e-01 9.52678800e-01 -4.21241194e-01
4.13543940e-01 -7.29128718e-02 1.96581468e-01 4.52816755e-01
2.83892870e-01 3.09640020e-01 -3.84953260e-01 3.74877065e-01
8.91138732e-01 4.43988681e-01 -2.45425120e-01 -4.55253750e-01
-1.30917645e+00 4.49414313e-01 4.12036091e-01 6.27308309e-01
-4.28059250e-01 -1.38559848e-01 5.35473526e-01 2.42547512e-01
1.60735965e-01 4.23629701e-01 -4.46991980e-01 -2.15279222e-01
-7.07741380e-01 -2.16825783e-01 5.81232131e-01 6.28410101e-01
7.23572195e-01 1.17190108e-01 -1.24866568e-01 9.62144315e-01
-7.08230585e-02 4.87206459e-01 8.63380134e-01 -9.58803058e-01
2.32245550e-01 1.06289876e+00 -1.70600023e-02 -9.15539742e-01
-2.17286453e-01 -3.27332824e-01 -1.06691504e+00 2.67449260e-01
6.35326743e-01 9.34088454e-02 -1.13545883e+00 1.56895018e+00
4.16756272e-01 2.35650226e-01 6.36960119e-02 7.52736509e-01
6.85651243e-01 3.85342956e-01 -3.30558382e-02 -2.52830118e-01
1.00675499e+00 -7.26608813e-01 -5.47683537e-01 -4.19581532e-01
5.03864825e-01 -1.45741791e-01 1.01994109e+00 6.09096289e-01
-2.94690907e-01 -9.93793249e-01 -1.06967115e+00 3.43472600e-01
-4.85906929e-01 3.03132176e-01 6.88701928e-01 5.38002133e-01
-6.78561628e-01 6.36682212e-01 -8.18430722e-01 -4.41228300e-01
8.10263932e-01 5.80125272e-01 -4.53345656e-01 -2.77388185e-01
-1.13688815e+00 6.42627001e-01 8.68261099e-01 5.29251099e-01
-1.10086417e+00 -1.60182327e-01 -7.52173722e-01 4.84506823e-02
7.59786069e-01 -4.05274510e-01 1.01408005e+00 -1.56514502e+00
-1.57613504e+00 7.44772553e-01 9.56037715e-02 -2.79871166e-01
3.22499365e-01 -5.41626886e-02 -4.94818270e-01 1.62588179e-01
1.72859170e-02 7.55648077e-01 9.81810451e-01 -1.03463197e+00
-8.39556813e-01 -4.62552011e-01 2.52955347e-01 4.84936908e-02
-6.61389649e-01 -2.34826148e-01 2.07880270e-02 -4.19930279e-01
4.81735021e-01 -7.89549351e-01 -1.28470361e-01 2.01472610e-01
-2.11334407e-01 -5.17774582e-01 8.99593592e-01 -3.49421859e-01
1.04674447e+00 -2.12599993e+00 -3.78590077e-02 1.12473965e-01
2.36905411e-01 8.16241980e-01 -2.39224702e-01 -1.19987436e-01
-2.95745134e-01 -3.29558820e-01 -2.04362541e-01 2.72756398e-01
-6.46873340e-02 7.76643991e-01 -2.70352811e-01 2.34114960e-01
2.16655638e-02 6.80249095e-01 -9.28441465e-01 -3.45757604e-01
4.41701144e-01 9.13472474e-02 -3.80525768e-01 4.84351143e-02
-3.70075434e-01 5.97421348e-01 -3.59040797e-01 7.10336030e-01
4.89243180e-01 -2.54701495e-01 3.04448515e-01 -9.33567435e-02
-6.58571869e-02 -1.02207087e-01 -1.39004838e+00 1.41505480e+00
-1.31151497e-01 3.62945408e-01 -4.17701840e-01 -1.61810362e+00
1.14755130e+00 2.14459419e-01 6.83351278e-01 -6.75382555e-01
3.85366291e-01 2.17672780e-01 1.81616858e-01 -7.15538859e-01
4.95968126e-02 -2.09370241e-01 6.17221780e-02 3.58296633e-01
3.03844139e-02 3.44036102e-01 3.32323432e-01 -4.25780624e-01
1.09313822e+00 2.61025488e-01 2.36512870e-01 2.52785623e-01
6.75569236e-01 5.31002469e-02 9.16477323e-01 7.34107852e-01
-4.70382065e-01 3.25265199e-01 2.84415513e-01 -9.87227142e-01
-6.59673989e-01 -1.02964509e+00 -1.35259897e-01 1.00404465e+00
1.44567132e-01 1.75291598e-01 -7.91666448e-01 -1.02160859e+00
-1.35408312e-01 4.67025489e-01 -6.15480721e-01 -5.82704246e-01
-5.43182373e-01 -5.65625846e-01 3.31210047e-01 7.89431989e-01
9.12966967e-01 -1.41255176e+00 -7.52846479e-01 2.48477519e-01
-3.58809352e-01 -7.72332311e-01 -2.61589810e-02 1.33572206e-01
-9.65675056e-01 -1.38179159e+00 -5.37774205e-01 -8.27559352e-01
8.71503413e-01 1.46017328e-01 6.87451541e-01 9.39797014e-02
-2.75518000e-01 1.90132111e-01 -5.37941694e-01 -2.28512928e-01
-4.30325449e-01 -1.85312212e-01 2.71981210e-01 6.12705052e-01
8.70933771e-01 -6.46507800e-01 -5.00085711e-01 6.03056252e-01
-8.37800741e-01 -6.64315820e-02 8.40304613e-01 8.51832926e-01
7.20413804e-01 3.44815880e-01 7.14146614e-01 -3.99963707e-01
1.09844238e-01 -1.66706219e-01 -4.63519841e-01 4.21650976e-01
-3.67243439e-01 -6.04150556e-02 8.09427977e-01 -9.13901448e-01
-9.19501305e-01 3.00864607e-01 -8.56550336e-02 -6.30408049e-01
-6.51191711e-01 2.64134258e-01 -4.72397268e-01 -1.40564973e-02
7.10531771e-01 5.29941678e-01 9.68143567e-02 -4.71570700e-01
1.32506385e-01 9.47930038e-01 5.30650258e-01 -3.41234148e-01
6.18492305e-01 5.86293459e-01 -1.04939356e-01 -6.46892488e-01
-1.09507608e+00 -3.10750961e-01 -1.22575700e+00 -5.94393134e-01
8.24011087e-01 -6.16037428e-01 -6.91113472e-01 9.05076563e-01
-8.05227697e-01 -6.17422342e-01 -7.77910113e-01 4.90861535e-01
-4.96415466e-01 5.30643582e-01 -1.57115221e-01 -5.73882222e-01
-8.72067064e-02 -1.03621650e+00 8.62324178e-01 2.69216865e-01
-1.11745968e-01 -7.08304286e-01 -1.88887551e-01 4.87371117e-01
1.35287255e-01 2.32060015e-01 6.55154228e-01 -8.83172095e-01
-5.34324050e-01 -4.27840889e-01 -6.50577396e-02 9.08458591e-01
4.08623576e-01 -2.07050636e-01 -8.62615347e-01 -1.56919882e-01
2.65664589e-02 -4.94403720e-01 8.86056781e-01 1.19800650e-01
1.43630898e+00 -1.44369766e-01 -3.41640681e-01 3.09706420e-01
9.65721548e-01 6.53312027e-01 8.39889884e-01 3.41672897e-01
6.89238667e-01 2.69108742e-01 7.30414510e-01 5.73923528e-01
6.52708113e-02 4.85851496e-01 2.49622896e-01 6.00488856e-02
-1.32712126e-01 -3.55678946e-02 3.50242555e-01 5.58664083e-01
-1.62035465e-01 -8.94060582e-02 -7.54020691e-01 2.88098127e-01
-1.70665359e+00 -1.17517555e+00 2.24180631e-02 2.36202574e+00
8.79885972e-01 2.32242659e-01 -7.70333037e-02 5.90405762e-01
8.30574274e-01 -7.67542571e-02 -9.64341283e-01 -9.95622873e-02
1.03380442e-01 2.30898678e-01 -4.94041061e-03 -7.38244653e-02
-1.25923729e+00 8.09461713e-01 5.89605141e+00 8.42328072e-01
-1.24091482e+00 -1.70724764e-02 5.14026523e-01 4.84551042e-02
2.95142114e-01 2.70451233e-02 -8.79819036e-01 4.70276952e-01
6.26008630e-01 2.96723247e-01 2.65123367e-01 1.00135422e+00
7.84970596e-02 -3.41475129e-01 -1.36711633e+00 1.09282005e+00
2.74153352e-01 -9.35272217e-01 2.58576363e-01 -2.30836938e-03
5.51536024e-01 -3.74417216e-01 -1.35833770e-01 6.50745392e-01
1.12039410e-01 -5.13894439e-01 4.52853948e-01 7.40784109e-01
9.59621668e-01 -3.92288089e-01 5.82848847e-01 5.95750093e-01
-1.02811241e+00 -6.65720761e-01 -5.73494494e-01 -1.96190059e-01
-1.59453779e-01 5.34482121e-01 -5.86566687e-01 3.23730141e-01
6.41722739e-01 9.26044703e-01 -6.20849848e-01 1.03561234e+00
-3.42514604e-01 5.00213146e-01 -3.49306911e-01 4.32475358e-02
1.04551993e-01 -2.20919222e-01 2.79110759e-01 3.94336075e-01
2.55652785e-01 2.38704622e-01 4.81246829e-01 5.71620941e-01
1.14211492e-01 -3.07245672e-01 -4.90747333e-01 -1.48182347e-01
1.12318806e-01 8.51504862e-01 -7.88027644e-01 -6.05286062e-01
-2.78973103e-01 9.46840525e-01 1.51292205e-01 2.46141464e-01
-7.91356027e-01 -4.02656436e-01 4.37318653e-01 -2.72853523e-02
3.74241084e-01 -2.93215108e-03 -7.70671517e-02 -1.25369787e+00
1.34668291e-01 -1.00605512e+00 6.91388726e-01 -6.68278396e-01
-1.13074160e+00 4.36116338e-01 -8.48443136e-02 -1.52042854e+00
-1.75650612e-01 -7.90597618e-01 -3.41298997e-01 2.28728801e-01
-1.17668641e+00 -7.71535397e-01 -4.79817271e-01 7.56274700e-01
6.48776829e-01 -2.79651999e-01 8.22284222e-01 4.10774291e-01
-9.37611699e-01 6.19542837e-01 2.51629651e-01 3.47545683e-01
4.95845526e-01 -8.41771185e-01 -1.24318227e-01 8.79105806e-01
1.32182151e-01 3.42417598e-01 2.56010890e-01 -6.59121275e-01
-1.12056494e+00 -1.15507543e+00 5.67607999e-01 -4.23418641e-01
5.23637891e-01 -4.05712217e-01 -9.91908610e-01 6.44956708e-01
-3.26793522e-01 2.66513407e-01 7.52765417e-01 1.98124759e-02
-2.39731818e-01 -6.99125409e-01 -1.02846968e+00 3.41483474e-01
1.36902487e+00 -1.47856683e-01 -8.91985595e-01 5.24628341e-01
3.82441223e-01 -1.36263877e-01 -7.60392070e-01 5.89055598e-01
5.63067555e-01 -9.58344460e-01 7.40847409e-01 -6.74249113e-01
1.52942017e-01 -3.04052949e-01 -6.54970631e-02 -1.06664050e+00
-3.74443263e-01 -2.69597024e-01 -1.61697313e-01 9.58431244e-01
8.04990754e-02 -8.20807457e-01 8.72333765e-01 5.23528099e-01
-3.76762390e-01 -7.44097114e-01 -1.19208062e+00 -1.00700092e+00
-2.16447890e-01 -2.23671734e-01 6.22931838e-01 8.32911730e-01
-1.60487905e-01 8.01165849e-02 -1.74769610e-01 -5.20843714e-02
4.16456819e-01 3.89519125e-01 7.13300943e-01 -1.53244591e+00
-3.18969697e-01 -3.54358226e-01 -7.36086965e-01 -1.18643510e+00
2.25407913e-01 -6.52076066e-01 2.50846237e-01 -1.50337851e+00
1.80846050e-01 -6.72317386e-01 -6.02325678e-01 1.09851611e+00
-4.93839383e-02 2.87390560e-01 -4.61950041e-02 2.44628936e-01
-8.75554383e-01 6.88700557e-01 1.40630555e+00 -3.30782086e-01
-2.14215994e-01 2.62206465e-01 -6.11769378e-01 7.71601975e-01
8.09910059e-01 -3.26401442e-01 -4.87093091e-01 -2.76760370e-01
-1.27380565e-01 -2.99608111e-01 4.46738243e-01 -1.61127400e+00
8.85258149e-03 -3.97425115e-01 6.20419800e-01 -5.13047338e-01
2.84104496e-01 -9.14105117e-01 8.40597153e-02 6.25228226e-01
-2.08689630e-01 -6.77825093e-01 6.51830733e-02 6.55231476e-01
-1.02055259e-01 -3.86974007e-01 8.31642687e-01 -2.67011434e-01
-1.21631539e+00 4.50769544e-01 -4.26145107e-01 -1.50433600e-01
1.32698071e+00 -9.23291743e-01 -1.01735964e-01 1.96671277e-01
-1.08287013e+00 9.00816396e-02 2.14776516e-01 4.82775211e-01
7.26221561e-01 -1.28233993e+00 -2.30269119e-01 5.60421109e-01
5.82554489e-02 1.49178877e-01 4.06207472e-01 8.36368263e-01
-1.30294502e-01 1.91739619e-01 -4.42093194e-01 -7.18676984e-01
-1.05179811e+00 8.89172971e-01 5.05287170e-01 -1.48960978e-01
-6.69394493e-01 6.83793187e-01 1.43476665e-01 -4.94108081e-01
5.00156045e-01 -2.61096627e-01 -5.44933438e-01 1.86016470e-01
6.48042500e-01 1.98767632e-01 -8.67770240e-02 -4.96303350e-01
-2.15456456e-01 4.46243078e-01 -1.73118368e-01 5.40673912e-01
1.37677550e+00 8.75190869e-02 -1.60419375e-01 5.61389267e-01
8.93524468e-01 -8.14754009e-01 -1.40168321e+00 -2.82543331e-01
-2.36742869e-01 -5.14805675e-01 -2.55222768e-01 -8.30944717e-01
-1.16003716e+00 9.72623944e-01 9.58473444e-01 3.66017632e-02
1.46261799e+00 -1.64867342e-01 8.29635084e-01 8.58541846e-01
4.51743394e-01 -1.46148765e+00 4.50086385e-01 5.31180978e-01
7.27144480e-01 -1.18111610e+00 -1.54674664e-01 -1.74897835e-01
-4.25335050e-01 1.35049021e+00 1.07187438e+00 6.14812113e-02
5.27053535e-01 -2.10836112e-01 -4.50442582e-02 -1.20287865e-01
-3.86570752e-01 -1.70640260e-01 -2.75727769e-04 7.24042356e-01
-2.72433758e-01 5.82156591e-02 -7.63932019e-02 4.75216597e-01
1.06794067e-01 3.65613848e-01 3.55565101e-01 9.21220660e-01
-7.14526832e-01 -1.07650959e+00 -4.28692609e-01 8.41949582e-01
1.56023815e-01 3.30228597e-01 -3.02312255e-01 5.61382532e-01
6.23582959e-01 9.27081108e-01 2.27207839e-01 -6.39832497e-01
4.89919543e-01 1.65282071e-01 4.72812235e-01 -6.67867959e-01
-2.32235596e-01 -3.16239893e-01 -1.28069267e-01 -6.31633639e-01
-6.26086473e-01 -8.24876249e-01 -1.28928137e+00 -1.20149553e-03
-4.57012802e-01 -1.36356890e-01 1.22954182e-01 1.23833776e+00
3.45934361e-01 5.64007461e-01 6.42185271e-01 -7.00309455e-01
-8.50822687e-01 -9.78621364e-01 -5.26523769e-01 4.55008239e-01
1.37242954e-02 -1.16378558e+00 -3.39241207e-01 1.82808131e-01]
|
[8.438355445861816, 0.8800064325332642]
|
446c3ed4-17d0-444a-a60c-d186b086c474
|
optimizing-protein-fitness-using-gibbs
|
2307.00494
| null |
https://arxiv.org/abs/2307.00494v1
|
https://arxiv.org/pdf/2307.00494v1.pdf
|
Optimizing protein fitness using Gibbs sampling with Graph-based Smoothing
|
The ability to design novel proteins with higher fitness on a given task would be revolutionary for many fields of medicine. However, brute-force search through the combinatorially large space of sequences is infeasible. Prior methods constrain search to a small mutational radius from a reference sequence, but such heuristics drastically limit the design space. Our work seeks to remove the restriction on mutational distance while enabling efficient exploration. We propose Gibbs sampling with Graph-based Smoothing (GGS) which iteratively applies Gibbs with gradients to propose advantageous mutations using graph-based smoothing to remove noisy gradients that lead to false positives. Our method is state-of-the-art in discovering high-fitness proteins with up to 8 mutations from the training set. We study the GFP and AAV design problems, ablations, and baselines to elucidate the results. Code: https://github.com/kirjner/GGS
|
['Ila Fiete', 'Regina Barzilay', 'Tommi Jaakkola', 'Raman Samusevich', 'Jason Yim', 'Andrew Kirjner']
|
2023-07-02
| null | null | null | null |
['efficient-exploration']
|
['methodology']
|
[ 7.57111013e-01 2.37997901e-02 2.30483748e-02 -3.14913355e-02
-9.58811104e-01 -8.50240052e-01 1.62081599e-01 -3.16760577e-02
-3.55266422e-01 1.35448503e+00 -1.56721964e-01 -4.91526544e-01
-1.90765366e-01 -5.12089968e-01 -1.04606175e+00 -9.32994723e-01
-1.41338989e-01 3.41166317e-01 8.75101015e-02 -9.16048512e-02
5.31303704e-01 3.21196049e-01 -1.45603251e+00 6.40006587e-02
1.37370384e+00 1.90993071e-01 6.99077606e-01 7.28757918e-01
1.52837783e-01 -1.64979368e-01 -4.68251705e-01 -1.11714482e-01
2.26923123e-01 -8.67204487e-01 -6.18030190e-01 -3.37400019e-01
1.37784034e-01 1.22843437e-01 2.54231155e-01 9.69334483e-01
8.57708931e-01 2.59902030e-01 5.55633605e-01 -7.50721753e-01
-7.16897070e-01 2.07867399e-01 -5.76859415e-01 7.94519633e-02
3.29904318e-01 6.52751446e-01 1.00393975e+00 -7.09999144e-01
9.80219245e-01 9.56262529e-01 5.63811898e-01 7.13545144e-01
-1.77470672e+00 -4.65527654e-01 -1.35463104e-02 -1.77532941e-01
-1.26297534e+00 -3.19688022e-01 5.46958804e-01 -3.33473295e-01
1.30176604e+00 5.17790437e-01 9.87832844e-01 1.03313553e+00
2.58234590e-01 6.50573015e-01 8.33244860e-01 -3.86685908e-01
6.71677649e-01 -6.04644299e-01 -3.23062003e-01 9.77213860e-01
3.83736610e-01 2.13758215e-01 -7.31199682e-01 -7.29221284e-01
4.28909063e-01 -1.53974937e-02 -5.04260540e-01 -6.99507475e-01
-9.42532301e-01 1.02777946e+00 9.71483439e-02 3.77021320e-02
-4.40163791e-01 2.28725791e-01 -1.19045049e-01 -6.99302778e-02
9.90760922e-02 1.29635012e+00 -7.86343038e-01 -4.26469803e-01
-8.50850105e-01 4.63286519e-01 5.58790803e-01 8.23660076e-01
8.43443394e-01 -1.77787781e-01 -7.21798688e-02 8.14475834e-01
1.69242099e-01 1.69317737e-01 3.09648812e-01 -9.83246505e-01
-1.96248338e-01 3.52483571e-01 3.29691023e-01 -3.96658510e-01
-4.13558245e-01 -4.36456263e-01 -1.39488041e-01 2.72938639e-01
4.13070083e-01 -4.73762304e-01 -1.23455358e+00 1.98886645e+00
5.94748020e-01 -1.21271104e-01 -2.33356208e-01 6.77576125e-01
1.55061319e-01 5.70065796e-01 9.42392573e-02 -4.73958671e-01
1.02697551e+00 -6.49013519e-01 -3.63702297e-01 -2.20986620e-01
9.11145985e-01 -5.67169189e-01 1.37031984e+00 4.73432541e-01
-1.04535568e+00 4.68675941e-02 -9.51061249e-01 1.74202949e-01
-1.53102994e-01 -1.11892849e-01 8.98536146e-01 7.25996017e-01
-9.29682255e-01 9.68601406e-01 -9.09779549e-01 -4.91626233e-01
8.61579418e-01 5.16619742e-01 7.66607374e-02 8.40111896e-02
-8.90035510e-01 7.86892891e-01 3.88085216e-01 -4.72005317e-03
-9.96639729e-01 -1.02586460e+00 -4.26982522e-01 -1.13779217e-01
5.68595827e-01 -9.14299786e-01 8.75279069e-01 -3.56291145e-01
-1.46294498e+00 5.08423030e-01 -3.34126472e-01 -3.51955891e-01
3.42168927e-01 -1.80629790e-01 3.01119924e-01 -1.88039392e-01
-3.03894673e-02 8.85672808e-01 5.44265926e-01 -9.11970317e-01
-4.23917681e-01 -2.37960398e-01 -3.60691428e-01 1.44665495e-01
6.30300045e-02 -2.88884044e-01 -8.37107971e-02 -5.80856383e-01
7.66072189e-03 -1.26458395e+00 -6.22469127e-01 -2.34022051e-01
-4.76299733e-01 -1.07243255e-01 3.07181656e-01 -4.81039584e-01
1.11739159e+00 -1.83181298e+00 3.26609105e-01 1.58839092e-01
1.46898078e-02 1.65360793e-01 -4.02475148e-01 6.12966299e-01
4.02659364e-02 3.21366578e-01 -5.97438812e-01 5.02280712e-01
-2.33831078e-01 -9.08791274e-02 1.31286141e-02 4.69874233e-01
2.46418402e-01 9.97333050e-01 -1.11062467e+00 -1.47184357e-01
-2.09375843e-01 5.34065127e-01 -1.04495382e+00 -9.77604166e-02
-7.57373750e-01 5.45083582e-01 -7.73227036e-01 7.63626099e-01
4.52560335e-01 -5.20837665e-01 5.77082872e-01 1.30997971e-01
-1.08073935e-01 1.37969360e-01 -6.52295709e-01 2.05960393e+00
6.93654567e-02 1.40584782e-01 -2.42558405e-01 -6.86353683e-01
8.02175283e-01 -2.34650657e-01 4.90498245e-01 -1.21623099e-01
-1.06327653e-01 1.87502369e-01 2.58395731e-01 -2.45032340e-01
8.42085630e-02 -1.77764431e-01 2.51352727e-01 2.51794666e-01
-7.87259117e-02 -2.25314587e-01 4.46147472e-01 4.89585139e-02
1.57255435e+00 7.10848808e-01 2.53345460e-01 -3.74961555e-01
-1.52652666e-01 3.26825827e-01 1.00824177e+00 9.11750674e-01
-1.00217849e-01 4.63204652e-01 5.25260448e-01 -2.25155696e-01
-1.12722874e+00 -1.02416790e+00 -9.70303193e-02 1.04075849e+00
-1.66149199e-01 -4.56533015e-01 -6.16346836e-01 -6.87059939e-01
2.18243584e-01 9.37458754e-01 -5.12064099e-01 -3.26038748e-01
-3.51237655e-01 -1.17097449e+00 5.38299024e-01 2.89229061e-02
-2.26052463e-01 -9.18500364e-01 -9.45171416e-01 2.09453940e-01
3.29514109e-02 -2.94242918e-01 -5.61945260e-01 6.62589490e-01
-8.93674493e-01 -1.01883984e+00 -8.36009562e-01 -7.53436327e-01
8.81196678e-01 -1.79538950e-01 8.51430118e-01 4.22872342e-02
-8.16860855e-01 -2.10248977e-01 -1.90765545e-01 -4.14631367e-01
-3.18177104e-01 -8.72114003e-02 -7.91227743e-02 -8.06869864e-01
4.52538133e-01 -7.51368105e-01 -7.84122765e-01 1.86503842e-01
-6.65015638e-01 1.31062403e-01 7.42522001e-01 1.33385754e+00
7.84548879e-01 -1.27826676e-01 7.36828387e-01 -9.02685881e-01
8.02095652e-01 -1.93255484e-01 -9.71685529e-01 3.85560840e-01
-6.29383266e-01 5.43000698e-01 4.39664751e-01 -6.16436541e-01
-7.94588208e-01 4.93115455e-01 -1.05968621e-02 1.33570461e-02
2.99832877e-02 4.15993273e-01 1.91507749e-02 -2.43025988e-01
1.00080502e+00 2.40336105e-01 5.71979359e-02 -4.47449237e-01
6.27140582e-01 1.18315704e-01 6.42013028e-02 -7.34150529e-01
3.23608100e-01 2.28319183e-01 2.04581857e-01 -8.74068320e-01
-3.40564072e-01 -6.58405945e-02 -2.47309476e-01 1.67364717e-01
5.18367648e-01 -4.61536855e-01 -9.82238948e-01 -5.02703041e-02
-7.57424593e-01 -6.33516431e-01 -3.66988629e-01 4.11525548e-01
-7.96840787e-01 4.63001609e-01 -4.09141988e-01 -9.19874251e-01
-4.26447123e-01 -1.25390828e+00 9.13550556e-01 2.21520603e-01
-6.19145274e-01 -5.01162529e-01 4.83940750e-01 1.63979486e-01
1.90473735e-01 3.60497564e-01 1.10132456e+00 -4.89333659e-01
-8.25435281e-01 1.71533823e-01 4.17056978e-01 -1.72805801e-01
2.57172763e-01 2.79939845e-02 -5.76622069e-01 -4.03189808e-01
-2.72834957e-01 -3.57066005e-01 1.05933380e+00 7.62431681e-01
9.64785159e-01 -2.71320045e-01 -6.87255859e-01 7.12887466e-01
1.24970782e+00 5.57541311e-01 6.31915987e-01 3.53924036e-01
3.17793518e-01 2.21879974e-01 7.71351218e-01 6.04961932e-01
-4.00292516e-01 5.47306240e-01 1.36867464e-01 1.18527479e-01
2.66016960e-01 -3.85583431e-01 5.70975132e-02 -1.87638286e-03
1.44524639e-02 -3.64143163e-01 -9.42677975e-01 5.95147908e-01
-1.78228176e+00 -7.71263659e-01 3.90477449e-01 2.27004099e+00
1.39372218e+00 5.80851436e-02 1.55050457e-01 -4.84686166e-01
7.82627285e-01 -2.19383150e-01 -1.09319079e+00 -2.83790171e-01
-1.52422801e-01 3.46618444e-01 6.70046568e-01 4.87933934e-01
-7.11773157e-01 1.06970036e+00 6.85470581e+00 1.02665699e+00
-7.18586266e-01 -4.03295010e-01 6.90634072e-01 -5.57704747e-01
-5.79894960e-01 4.01592642e-01 -9.07450438e-01 3.78813893e-01
7.71352112e-01 -1.43872291e-01 8.55595767e-01 6.83284760e-01
3.47893357e-01 -3.06340724e-01 -8.47311318e-01 6.08182669e-01
-3.26158673e-01 -1.73969579e+00 -1.47852451e-01 1.98038071e-01
9.06215370e-01 1.00079186e-01 1.22217171e-01 -1.71751007e-01
7.46186793e-01 -1.08287287e+00 1.46077082e-01 4.52311963e-01
6.02166355e-01 -8.38153601e-01 1.19702160e-01 1.86700538e-01
-4.72857028e-01 1.62900537e-01 -4.17213410e-01 2.33385131e-01
2.07194462e-01 8.08836043e-01 -1.21720648e+00 1.49602685e-02
5.21255553e-01 3.62409532e-01 -3.71172011e-01 1.00014114e+00
-2.36699045e-01 6.06129646e-01 -6.48494184e-01 -6.42748892e-01
-2.23163627e-02 -4.46477264e-01 7.75164247e-01 8.87338281e-01
5.22390425e-01 7.17200786e-02 -2.06569117e-03 1.35771096e+00
1.46226333e-02 -7.18753831e-03 -5.46128631e-01 -5.60587406e-01
5.48205495e-01 9.09967959e-01 -9.04400289e-01 2.82009772e-04
1.12615250e-01 9.47375715e-01 4.73319203e-01 6.50439918e-01
-7.43394613e-01 -6.48403883e-01 8.27894926e-01 -6.98727230e-03
5.97273171e-01 -1.15836270e-01 -1.88477561e-01 -6.64730489e-01
-1.23223476e-01 -1.04407573e+00 3.55756760e-01 -6.52168930e-01
-1.06344366e+00 1.21440170e-02 -4.14105952e-01 -6.52056932e-01
-1.08847842e-01 -3.37557673e-01 -2.54793763e-01 9.97621775e-01
-1.03254414e+00 -5.29790103e-01 3.57378364e-01 -1.77585751e-01
5.81960440e-01 8.96991268e-02 6.93638325e-01 -1.76458016e-01
-4.08403665e-01 4.01955903e-01 3.77099216e-01 -6.78079009e-01
5.80666482e-01 -1.01971650e+00 6.89466357e-01 6.57718539e-01
-2.08541960e-01 1.24736905e+00 1.05866492e+00 -1.24893475e+00
-1.69333875e+00 -7.92318523e-01 4.91685301e-01 -4.55536216e-01
3.96150112e-01 -3.56439233e-01 -9.07395720e-01 4.44746584e-01
-1.09771371e-01 -3.50344956e-01 7.39140570e-01 2.91551858e-01
-3.56402770e-02 5.27876556e-01 -1.29193127e+00 1.10150516e+00
1.40444279e+00 -2.08956271e-01 -1.20352410e-01 5.75401127e-01
7.60307133e-01 -2.74086624e-01 -7.31575072e-01 4.10927773e-01
5.64483881e-01 -6.02501929e-01 1.01329792e+00 -8.70751023e-01
1.24526039e-01 -6.82982206e-01 7.83609524e-02 -1.44144666e+00
-5.15313864e-01 -1.39681065e+00 8.89229104e-02 8.20055127e-01
9.36746061e-01 -6.22688949e-01 1.11427569e+00 4.82752740e-01
-1.00206278e-01 -1.24095309e+00 -7.57747769e-01 -1.02490187e+00
6.30379021e-02 1.07953295e-01 5.29335797e-01 6.76155150e-01
4.11912501e-01 3.57753307e-01 -3.07191044e-01 -1.06980436e-01
5.41574955e-01 1.35969907e-01 5.54089844e-01 -7.90434897e-01
-5.76354802e-01 -4.64437544e-01 -1.17230304e-01 -1.04230058e+00
-2.65916944e-01 -7.51273215e-01 4.22632933e-01 -1.48948276e+00
3.66780192e-01 -3.19559574e-01 -9.86205563e-02 4.04467762e-01
-4.46841806e-01 -1.93944678e-01 -1.77219898e-01 -8.82334858e-02
-6.00081563e-01 7.41343498e-01 1.12997437e+00 1.00054830e-01
-5.53792715e-01 -2.98709333e-01 -8.88612866e-01 4.43840474e-01
1.03948414e+00 -6.40946448e-01 -4.38498318e-01 2.47340575e-02
5.70986390e-01 -1.49718255e-01 -3.88501324e-02 -5.79064846e-01
-1.64479122e-01 -5.39256334e-01 4.22924697e-01 -5.07827103e-01
2.42655203e-01 -2.34883934e-01 5.00313878e-01 7.66111970e-01
-6.66943312e-01 -1.53436780e-01 2.85311520e-01 8.22665215e-01
5.58598399e-01 -3.22078228e-01 8.54479611e-01 -3.14589381e-01
-2.78490663e-01 6.82731494e-02 -4.90614712e-01 3.20684761e-01
9.15244460e-01 -2.39116266e-01 -3.86088878e-01 -1.09865993e-01
-7.35035062e-01 1.95847720e-01 9.73003387e-01 -6.73470646e-02
6.92786992e-01 -8.65956068e-01 -5.28000057e-01 4.34431843e-02
3.54628116e-02 -1.92801431e-01 7.15419576e-02 6.33361459e-01
-6.07442737e-01 3.81940067e-01 -9.79588740e-03 -5.77228010e-01
-1.31313336e+00 7.29538202e-01 2.02177092e-01 1.15931937e-02
-3.72976214e-01 1.23436081e+00 1.68842927e-01 -4.47593033e-01
-1.32967485e-02 -1.19052321e-01 3.57973099e-01 -4.96502221e-01
1.40438929e-01 3.37540925e-01 -1.86683714e-01 1.40966460e-01
-4.88360077e-01 4.67220962e-01 -1.39556587e-01 -1.04293404e-02
1.48707294e+00 9.58266780e-02 2.23797336e-02 -2.20662616e-02
1.00913572e+00 -6.39140680e-02 -1.50857663e+00 3.07619423e-01
2.19004065e-01 -6.95715666e-01 -1.52329639e-01 -1.11004627e+00
-3.79326195e-01 3.31677616e-01 7.18446314e-01 -3.32310289e-01
9.50488627e-01 4.66409698e-02 7.13006675e-01 6.35435164e-01
3.41981441e-01 -1.19269192e+00 -4.91980501e-02 2.04103380e-01
5.94554722e-01 -9.28871870e-01 2.37972841e-01 -3.24973613e-01
-4.08829749e-01 7.19570994e-01 3.60598862e-01 1.41350150e-01
1.77088007e-01 1.81199491e-01 -3.84660184e-01 -2.36790940e-01
-9.50497389e-01 -2.62230784e-01 -7.23167583e-02 8.18850160e-01
6.04027510e-01 -5.75508252e-02 -7.54827201e-01 2.19204351e-01
8.76413379e-03 2.12574825e-01 2.83391476e-01 1.34588635e+00
-7.35577166e-01 -1.23025703e+00 -1.60914540e-01 6.51859403e-01
-4.41181034e-01 -5.42934299e-01 -5.95975161e-01 2.49423563e-01
-1.34702727e-01 8.57650161e-01 -4.56792206e-01 -7.36373588e-02
1.04772098e-01 2.21646354e-01 7.11552262e-01 -4.59025890e-01
-2.32094690e-01 4.43621457e-01 3.17211062e-01 -4.99744147e-01
5.97808659e-02 -9.58145320e-01 -1.46841502e+00 -5.76576181e-02
-6.75814748e-01 4.58308190e-01 5.88693798e-01 5.28921306e-01
1.01825404e+00 4.84376758e-01 2.54258692e-01 -3.57900977e-01
-4.37376261e-01 -5.46123147e-01 -4.31122214e-01 8.92366916e-02
2.04168186e-01 -5.37735224e-01 -2.89615244e-01 1.29479915e-01]
|
[4.754087448120117, 5.545462608337402]
|
ee204be6-d12b-4f7c-b7c2-eda0cecdabc4
|
paravs-a-simple-fast-efficient-and-flexible
|
2102.06086
| null |
https://arxiv.org/abs/2102.06086v1
|
https://arxiv.org/pdf/2102.06086v1.pdf
|
ParaVS: A Simple, Fast, Efficient and Flexible Graph Neural Network Framework for Structure-Based Virtual Screening
|
Structure-based virtual screening (SBVS) is a promising in silico technique that integrates computational methods into drug design. An extensively used method in SBVS is molecular docking. However, the docking process can hardly be computationally efficient and accurate simultaneously because classic mechanics scoring function is used to approximate, but hardly reach, the quantum mechanics precision in this method. In order to reduce the computational cost of the protein-ligand scoring process and use data driven approach to boost the scoring function accuracy, we introduce a docking-based SBVS method and, furthermore, a deep learning non-docking-based method that is able to avoid the computational cost of the docking process. Then, we try to integrate these two methods into an easy-to-use framework, ParaVS, that provides both choices for researchers. Graph neural network (GNN) is employed in ParaVS, and we explained how our in-house GNN works and how to model ligands and molecular targets. To verify our approaches, cross validation experiments are done on two datasets, an open dataset Directory of Useful Decoys: Enhanced (DUD.E) and an in-house proprietary dataset without computational generated artificial decoys (NoDecoy). On DUD.E we achieved a state-of-the-art AUC of 0.981 and a state-of-the-art enrichment factor at 2% of 36.2; on NoDecoy we achieved an AUC of 0.974. We further finish inference of an open database, Enamine REAL Database (RDB), that comprises over 1.36 billion molecules in 4050 core-hours using our ParaVS non-docking method (ParaVS-ND). The inference speed of ParaVS-ND is about 3.6e5 molecule / core-hour, while this number of a conventional docking-based method is around 20, which is about 16000 times faster. The experiments indicate that ParaVS is accurate, computationally efficient and can be generalized to different molecular.
|
['Lurong Pan', 'Dawei Leng', 'Junfeng Wu']
|
2021-02-08
| null | null | null | null |
['molecular-docking']
|
['medical']
|
[-4.01650637e-01 -3.09576690e-01 -1.26891077e-01 -5.44265434e-02
-7.42085993e-01 -5.73689580e-01 2.08764419e-01 2.68739074e-01
-5.75816274e-01 1.44678211e+00 -5.04257023e-01 -6.45328224e-01
3.51975411e-02 -9.43446994e-01 -1.11324251e+00 -7.99474359e-01
-9.78330076e-02 7.84918845e-01 3.71116668e-01 -4.76592213e-01
1.00165263e-01 8.58945787e-01 -1.18072820e+00 4.50680666e-02
1.08595860e+00 5.24118900e-01 1.05703652e-01 2.65848428e-01
4.19342890e-02 2.84144551e-01 -5.13986826e-01 -6.66476727e-01
1.74008474e-01 -4.37876850e-01 -6.21285379e-01 -8.85624766e-01
1.98590487e-01 2.40931045e-02 -1.32726744e-01 9.86930311e-01
1.07659507e+00 6.57620281e-02 6.94049656e-01 -7.02521563e-01
-2.42241144e-01 1.77554056e-01 -9.20948833e-02 -1.55585051e-01
5.77647805e-01 5.12603045e-01 6.83561325e-01 -1.00065792e+00
9.97137904e-01 8.84249866e-01 7.80995369e-01 5.50971687e-01
-1.40344393e+00 -9.84197617e-01 -4.39554483e-01 1.37998983e-01
-1.70071876e+00 -1.70518290e-02 2.33562082e-01 -5.56195736e-01
1.27522182e+00 3.56243581e-01 7.45436847e-01 1.27025580e+00
6.63640976e-01 -4.12549227e-02 1.00026798e+00 -7.54961148e-02
6.88697934e-01 1.33232411e-03 -2.56571658e-02 7.71637321e-01
4.26416308e-01 3.63395721e-01 -4.75377828e-01 -5.75943410e-01
6.06041908e-01 1.21550456e-01 -2.85791993e-01 -3.15446824e-01
-7.98691332e-01 1.14191318e+00 6.07371688e-01 8.25181156e-02
-4.42346573e-01 3.32524278e-03 2.93602496e-01 2.25504324e-01
7.59948939e-02 7.40922272e-01 -6.38215721e-01 1.25671148e-01
-7.49937773e-01 5.65373063e-01 1.11963534e+00 4.51159865e-01
6.05496407e-01 -1.21403858e-01 5.67759164e-02 3.49673003e-01
2.70551115e-01 3.07324141e-01 2.81004608e-01 -3.17038685e-01
7.52319722e-03 5.22329330e-01 2.33654276e-01 -7.07217097e-01
-7.36655593e-01 -2.96430886e-01 -8.59899402e-01 3.85654926e-01
4.79300201e-01 -1.59461349e-01 -8.51100743e-01 1.64122176e+00
6.00574732e-01 -1.19481355e-01 -2.97344178e-02 8.92321467e-01
1.09360158e+00 5.75477600e-01 3.32121283e-01 -4.94583964e-01
1.22147667e+00 -7.08359480e-01 -3.44748110e-01 6.43257439e-01
8.12103629e-01 -6.37730718e-01 8.84404063e-01 8.06863368e-01
-8.20721269e-01 -4.26313013e-01 -1.23046517e+00 2.14062318e-01
-6.63223982e-01 -1.66332185e-01 8.65075469e-01 7.29192138e-01
-8.30377877e-01 1.21761942e+00 -8.19172144e-01 -8.11413899e-02
4.19367880e-01 1.01154053e+00 -7.84710169e-01 1.83836192e-01
-1.44984913e+00 1.00435102e+00 5.60750306e-01 -3.21918219e-01
-1.27909052e+00 -8.10745478e-01 -4.26407009e-01 -4.23899330e-02
3.17535520e-01 -8.97053897e-01 6.61339104e-01 -4.66189533e-01
-1.66263020e+00 4.75039780e-01 3.22853075e-03 -4.17203844e-01
4.09047604e-01 1.05753213e-01 -2.79549628e-01 -1.00300610e-01
-6.69558570e-02 4.23765659e-01 5.36374487e-02 -7.45387793e-01
1.89445898e-01 -4.84172344e-01 -9.24464911e-02 -1.88946109e-02
1.14919499e-01 -9.75012034e-02 -3.71135443e-01 -2.71792620e-01
-2.81770349e-01 -9.82857168e-01 -5.09862542e-01 -2.16090545e-01
-6.90584600e-01 -2.19677135e-01 7.72598162e-02 -4.36895162e-01
1.20109284e+00 -1.48298967e+00 2.23896116e-01 4.98882234e-01
4.35708404e-01 5.48718214e-01 -1.74078852e-01 8.31004500e-01
-3.72872889e-01 2.04643413e-01 -9.68848914e-03 1.48762092e-01
-2.89981037e-01 -5.95897473e-02 1.31779626e-01 6.11600578e-01
-8.90365392e-02 8.56387436e-01 -7.78636992e-01 -8.89958888e-02
2.95881201e-02 7.44845152e-01 -8.31390321e-01 6.72908649e-02
-5.06400228e-01 6.26125515e-01 -5.57020724e-01 9.04065490e-01
9.17078793e-01 -4.16837305e-01 5.53201318e-01 -3.23016047e-01
-2.89552540e-01 2.17082232e-01 -9.58377838e-01 1.82348382e+00
2.35873945e-02 -1.60086066e-01 -5.05138755e-01 -6.45984709e-01
9.35811818e-01 7.72090033e-02 4.99090880e-01 -7.49029160e-01
2.36768425e-01 4.49292749e-01 3.05532038e-01 -1.40207887e-01
-6.63412139e-02 -1.43074125e-01 3.78013015e-01 -1.81585774e-01
2.65274793e-01 3.57695252e-01 2.43534058e-01 4.87807952e-02
1.24856663e+00 3.75819445e-01 4.98536676e-01 -3.63404304e-01
6.52176619e-01 2.65169829e-01 4.74247336e-01 4.34142768e-01
1.55378342e-01 9.69232991e-02 5.97807169e-01 -8.71330976e-01
-9.75306988e-01 -7.02293336e-01 -4.82767820e-01 8.16889942e-01
-2.35299096e-02 -8.57238591e-01 -8.47412050e-01 -6.43437266e-01
1.49757266e-01 3.70221317e-01 -4.80581284e-01 -3.17496099e-02
-2.11261362e-01 -1.28314078e+00 6.84272349e-01 -8.41574669e-02
6.78901449e-02 -8.05980861e-01 7.78656304e-02 5.91930151e-01
3.17141533e-01 -4.47923124e-01 -7.54563361e-02 4.68532652e-01
-6.43551767e-01 -1.45397902e+00 -5.60348153e-01 -2.88403273e-01
2.58065045e-01 -2.43926778e-01 1.06324768e+00 1.21440008e-01
-3.38571668e-01 -7.27262437e-01 -2.17002586e-01 -3.24235916e-01
-4.54616189e-01 3.99521589e-02 4.70205337e-01 -3.76483828e-01
5.88952184e-01 -8.68104696e-01 -8.16135049e-01 4.47818190e-01
-5.13963103e-01 -1.36821657e-01 5.79118669e-01 1.08592796e+00
1.07975495e+00 -4.26965475e-01 3.84370923e-01 -8.35968852e-01
3.78275067e-01 -4.15477246e-01 -1.09565449e+00 -2.75882818e-02
-8.91148686e-01 2.74363548e-01 8.74274194e-01 -4.46708620e-01
-2.17336282e-01 4.46555972e-01 -8.45725358e-01 -4.50031430e-01
3.48155424e-02 7.50249267e-01 -3.37850690e-01 -6.96810722e-01
9.46047723e-01 1.86396167e-01 2.40388200e-01 -7.21550584e-01
-5.10845743e-02 3.71742606e-01 -3.04390565e-02 -4.33562785e-01
4.39090014e-01 1.04196966e-01 6.01899683e-01 -5.32238066e-01
-3.26263815e-01 -2.19908893e-01 -3.11518073e-01 3.64954680e-01
8.47996294e-01 -9.42520320e-01 -1.69757211e+00 2.01014727e-01
-1.12972724e+00 -3.77072215e-01 3.79619181e-01 5.61677098e-01
-2.87183464e-01 5.29136896e-01 -5.03588915e-01 -4.23603117e-01
-6.84572697e-01 -1.65870619e+00 7.28191674e-01 -1.02386646e-01
3.51960212e-03 -8.28718781e-01 5.97380936e-01 1.95051819e-01
2.44074836e-01 8.48990560e-01 7.81202853e-01 -1.21877635e+00
-5.17204404e-01 -1.30353749e-01 9.96220764e-03 7.55153149e-02
-1.47023514e-01 2.69776769e-03 -7.88832903e-01 -4.05800670e-01
-3.56722295e-01 -3.50088358e-01 7.31865942e-01 3.93893808e-01
1.09331632e+00 -2.58425504e-01 -6.01153314e-01 8.65059257e-01
1.81394362e+00 5.01303911e-01 1.06932056e+00 3.90207827e-01
6.79854631e-01 -2.14542332e-03 6.91157281e-01 4.63344574e-01
3.44976299e-02 1.07010138e+00 7.42054820e-01 -3.55778873e-01
3.00742626e-01 -2.25948140e-01 3.51713061e-01 3.84365886e-01
-7.20097423e-01 -3.57080996e-01 -8.63162935e-01 -4.43906218e-01
-1.58841419e+00 -1.01899183e+00 -6.03835404e-01 2.52793264e+00
1.30300307e+00 -2.15787869e-02 2.91432709e-01 -2.68179804e-01
3.05029780e-01 -5.82027793e-01 -7.32011795e-01 -4.27479148e-01
-8.46392810e-02 8.59146833e-01 4.89201903e-01 4.92450088e-01
-9.88197148e-01 1.00900340e+00 5.80138779e+00 1.36982846e+00
-1.31554341e+00 6.57830015e-02 5.91728926e-01 3.81915718e-02
3.68298814e-02 6.83448091e-02 -9.89108086e-01 7.47470140e-01
1.50951302e+00 1.16352923e-01 4.25569087e-01 9.85303044e-01
3.00174505e-01 -2.02071834e-02 -1.24304044e+00 1.01859879e+00
-4.10760343e-01 -1.92603731e+00 2.76378900e-01 6.20205343e-01
3.40904742e-01 3.50728512e-01 -3.55078459e-01 3.54978800e-01
6.41354397e-02 -1.43004322e+00 9.88015383e-02 6.61121607e-01
1.19854391e+00 -1.00130677e+00 9.69527483e-01 3.14304858e-01
-8.67242873e-01 5.06119728e-01 -6.25742078e-01 8.96873549e-02
-2.05342174e-01 5.05597532e-01 -8.28870237e-01 8.53136182e-01
4.90536541e-01 4.31160241e-01 -4.33376133e-01 1.08476424e+00
9.84816998e-02 2.76252508e-01 -2.89106667e-01 -4.68792349e-01
1.99305296e-01 -5.20920992e-01 2.10687503e-01 9.85211551e-01
1.00294963e-01 8.93612504e-02 3.21126819e-01 9.18625653e-01
-1.62458524e-01 3.91288936e-01 -4.65449363e-01 5.72494231e-02
4.22414660e-01 1.13757348e+00 -3.31857592e-01 -2.26962239e-01
-8.40575397e-02 9.89027262e-01 2.06836760e-01 5.11622317e-02
-1.26837957e+00 -5.90449452e-01 8.24237108e-01 1.52191743e-01
1.39941543e-01 1.08238548e-01 3.94354612e-01 -1.06428719e+00
-4.30682302e-01 -1.18697858e+00 1.55194297e-01 -5.47149062e-01
-1.21288943e+00 6.83449805e-01 -3.31864864e-01 -1.21099889e+00
1.63236171e-01 -1.12849271e+00 -3.34434062e-01 1.20956862e+00
-1.24031043e+00 -7.37617552e-01 -6.97614402e-02 6.51269257e-01
7.59103373e-02 -1.93213537e-01 1.34663045e+00 6.53770268e-01
-7.42792130e-01 7.69713879e-01 5.06478250e-01 -2.70449758e-01
7.86253691e-01 -1.07753444e+00 2.73738056e-01 1.55292168e-01
-2.64256418e-01 1.20905232e+00 6.27289772e-01 -7.85534203e-01
-1.84092736e+00 -9.89567161e-01 5.37930727e-01 -5.89772999e-01
6.45846605e-01 -4.07893509e-01 -9.44183052e-01 -2.48075537e-02
-2.52085984e-01 -5.92989139e-02 1.19965136e+00 -8.74818489e-02
-2.66408414e-01 1.98185235e-01 -1.30074179e+00 4.01812524e-01
8.62868369e-01 -1.10652573e-01 -1.13136142e-01 8.88418078e-01
6.69146895e-01 -7.24377751e-01 -1.18426979e+00 4.51666415e-01
6.52335286e-01 -1.13443208e+00 1.13909304e+00 -9.27203119e-01
-2.24944223e-02 -5.41922271e-01 -1.21082202e-01 -9.52437699e-01
-2.62880266e-01 -7.22547412e-01 1.65290851e-02 5.04651845e-01
6.13857746e-01 -8.71750414e-01 7.80445814e-01 2.89255559e-01
-2.60195553e-01 -1.30154920e+00 -1.13190150e+00 -9.83696878e-01
2.73730874e-01 -3.96738388e-02 6.14684522e-01 9.08393681e-01
-1.15495585e-01 5.51992178e-01 -3.04540455e-01 6.03840780e-03
3.90265256e-01 -2.20171779e-01 9.58451390e-01 -1.49138427e+00
-5.77179134e-01 -1.79814905e-01 -5.61212778e-01 -6.03384078e-01
-2.18024582e-01 -9.91966724e-01 -6.22411430e-01 -1.19586289e+00
4.50162083e-01 -2.02828154e-01 -2.64569700e-01 6.54122055e-01
1.65631533e-01 1.99223861e-01 -2.95866728e-01 2.72384465e-01
-4.40957755e-01 5.12996674e-01 1.34850025e+00 -6.69098869e-02
-4.78896350e-01 -2.18055710e-01 -4.32614893e-01 4.90034342e-01
5.85037649e-01 -6.56736791e-01 6.37372062e-02 5.64069629e-01
5.06730378e-01 -6.05165586e-03 3.27790886e-01 -8.85259509e-01
-5.59975766e-02 -1.61279336e-01 5.64766824e-01 -6.97143316e-01
3.17164302e-01 -4.44576949e-01 8.13885272e-01 1.02362919e+00
4.90977734e-01 -2.08208025e-01 4.04360533e-01 4.73084927e-01
5.30158803e-02 -3.00711785e-02 7.30973303e-01 -1.27651855e-01
-1.37594089e-01 5.77533841e-01 -1.49714112e-01 -4.53485817e-01
9.06045973e-01 -3.40323329e-01 -2.82333970e-01 1.79791182e-01
-9.02032256e-01 -1.65219717e-02 5.36514640e-01 -4.11504388e-01
4.99243379e-01 -1.03737414e+00 -4.40531701e-01 -2.08764412e-02
9.76009220e-02 -3.90343815e-01 8.14685673e-02 1.13096881e+00
-1.17457688e+00 6.21713340e-01 -1.70567036e-01 -4.89856064e-01
-1.30696225e+00 7.93179214e-01 6.89440966e-01 -3.65570873e-01
-1.38865009e-01 6.55068755e-01 2.31789537e-02 -4.21082258e-01
6.57543689e-02 -1.55658394e-01 -1.51043877e-01 -1.30132318e-01
5.61165988e-01 2.50882536e-01 4.77631897e-01 -3.54699194e-01
-6.81421757e-01 4.81520176e-01 -7.24605620e-02 5.99931598e-01
1.49075472e+00 7.56429493e-01 -3.49100947e-01 -1.92648247e-01
1.04026949e+00 9.57677215e-02 -7.48801231e-01 4.63494301e-01
-3.14091206e-01 -4.62819152e-02 -6.65831938e-03 -1.11895990e+00
-5.01623690e-01 6.04228735e-01 9.78289366e-01 -2.18490556e-01
7.28347003e-01 -2.22526416e-01 5.95123470e-01 8.75620425e-01
7.21157491e-01 -4.85572338e-01 -1.79949179e-01 4.85407919e-01
8.72469127e-01 -1.19434166e+00 4.08392936e-01 -3.64472061e-01
-2.83931464e-01 1.23727655e+00 4.51089889e-01 -9.21430252e-03
3.74053359e-01 -1.60977364e-01 -4.80601788e-01 -6.21219218e-01
-5.63193142e-01 -6.14015944e-02 2.84500569e-01 3.95187855e-01
5.42893946e-01 -4.20015268e-02 -7.47065783e-01 8.33206415e-01
-4.06606309e-02 1.88264415e-01 2.16572478e-01 6.57785296e-01
-3.62535268e-01 -1.65804744e+00 -2.19770461e-01 6.29515648e-02
-6.40125334e-01 -4.85539705e-01 -7.42470980e-01 1.08725393e+00
2.62545884e-01 6.04666889e-01 -6.00115538e-01 -6.34906471e-01
4.18162674e-01 1.77764595e-02 5.14652133e-01 -4.01737094e-01
-7.93856680e-01 1.77921116e-01 6.79174587e-02 -9.70241368e-01
-2.26715654e-01 1.30259335e-01 -1.43905091e+00 -6.78072274e-01
-7.82962441e-01 6.59663618e-01 8.44497919e-01 5.25682330e-01
7.78245568e-01 2.54137248e-01 4.79904801e-01 -8.96082580e-01
-4.03975874e-01 -6.98148131e-01 -5.38909078e-01 -3.57055478e-02
-7.43275285e-02 -8.42547834e-01 -2.13269725e-01 -4.12935376e-01]
|
[4.8986077308654785, 5.592639923095703]
|
fd0f7542-0a66-4c4f-b963-ba24060e15d8
|
is-model-attention-aligned-with-human
|
2306.01220
| null |
https://arxiv.org/abs/2306.01220v1
|
https://arxiv.org/pdf/2306.01220v1.pdf
|
Is Model Attention Aligned with Human Attention? An Empirical Study on Large Language Models for Code Generation
|
Large Language Models (LLMs) have been demonstrated effective for code generation. Due to the complexity and opacity of LLMs, little is known about how these models generate code. To deepen our understanding, we investigate whether LLMs attend to the same parts of a natural language description as human programmers during code generation. An analysis of five LLMs on a popular benchmark, HumanEval, revealed a consistent misalignment between LLMs' and programmers' attention. Furthermore, we found that there is no correlation between the code generation accuracy of LLMs and their alignment with human programmers. Through a quantitative experiment and a user study, we confirmed that, among twelve different attention computation methods, attention computed by the perturbation-based method is most aligned with human attention and is constantly favored by human programmers. Our findings highlight the need for human-aligned LLMs for better interpretability and programmer trust.
|
['Tianyi Zhang', 'Lei Ma', 'Zhijie Wang', 'Shengmai Chen', 'Bonan Kou']
|
2023-06-02
| null | null | null | null |
['code-generation']
|
['computer-code']
|
[-8.25405568e-02 3.83882582e-01 2.42377687e-02 -1.20027393e-01
-3.73247325e-01 -5.84387600e-01 5.71553051e-01 4.98533666e-01
-1.45264611e-01 2.16745585e-01 4.05834079e-01 -4.17534590e-01
4.11446184e-01 -4.34140027e-01 -7.16696262e-01 -3.68464813e-02
2.60748714e-01 1.55791566e-01 -6.50295392e-02 -2.99973696e-01
9.43930447e-01 -1.34224454e-02 -1.37974286e+00 5.16029239e-01
1.28566504e+00 2.03285083e-01 5.57265460e-01 8.20536375e-01
-2.74427861e-01 1.22552204e+00 -7.79387772e-01 -6.69154048e-01
-4.94615100e-02 -4.05746192e-01 -1.03448176e+00 -3.37543845e-01
5.70759714e-01 -5.95122650e-02 8.49715471e-02 1.07311666e+00
4.16430205e-01 1.75567158e-02 3.78602833e-01 -1.24624884e+00
-1.29903102e+00 8.60869527e-01 -5.39681017e-01 3.67483377e-01
7.00403631e-01 4.82432008e-01 9.46304381e-01 -8.29764426e-01
6.20010316e-01 1.24320328e+00 8.30452621e-01 6.70616627e-01
-1.42706406e+00 -5.11685193e-01 4.75623794e-02 -2.45118633e-01
-1.28831542e+00 -4.53802139e-01 4.16057169e-01 -9.32434082e-01
1.23345363e+00 2.56723583e-01 4.42681551e-01 1.15103686e+00
6.52993560e-01 4.18860018e-01 9.90452707e-01 -5.54548323e-01
1.31670609e-01 6.20168746e-01 3.51008952e-01 7.19686866e-01
4.68069702e-01 1.46797106e-01 -6.27690017e-01 -5.68811357e-01
7.11751699e-01 -3.11886162e-01 -4.02729511e-01 -1.52032763e-01
-1.33272111e+00 8.32288921e-01 3.44740599e-01 4.74937826e-01
-2.05214813e-01 4.17933553e-01 3.17394614e-01 1.11251242e-01
1.26927614e-01 1.42620552e+00 -3.40386897e-01 -5.73934376e-01
-5.22495270e-01 2.87055403e-01 9.56511438e-01 1.14441824e+00
5.69415390e-01 -1.80849396e-02 -4.05323476e-01 4.77898568e-01
3.41858536e-01 3.08342695e-01 7.33299494e-01 -1.06098926e+00
3.75064909e-01 8.20653200e-01 3.17898571e-01 -1.45475912e+00
-1.10447273e-01 -5.69642484e-01 -3.34892273e-01 3.44737440e-01
2.79201865e-01 7.04619661e-02 -6.56232908e-02 1.78148389e+00
-3.16775769e-01 -3.82297665e-01 -1.28822312e-01 7.12196410e-01
4.76472020e-01 4.00653183e-01 1.63076088e-01 1.27824351e-01
1.20205140e+00 -9.88586605e-01 -2.76585400e-01 -5.98926544e-01
7.23967671e-01 -8.19922745e-01 1.72133052e+00 1.46982223e-01
-1.19729853e+00 -7.76177406e-01 -8.08009803e-01 -2.05907986e-01
4.02757572e-03 2.44097766e-02 6.24821484e-01 5.16067863e-01
-1.20499778e+00 6.28392458e-01 -7.64957190e-01 -4.03280795e-01
3.04841697e-01 5.71215972e-02 -3.10410112e-02 2.72855610e-01
-7.59130359e-01 1.07237530e+00 -2.89930478e-02 -2.38700479e-01
-9.18287218e-01 -6.50873840e-01 -8.20399106e-01 1.84783578e-01
-2.13736296e-01 -7.42746532e-01 1.70463502e+00 -1.34118903e+00
-1.06833994e+00 9.61999118e-01 -2.99596518e-01 -3.56901020e-01
2.33439222e-01 -3.75888854e-01 8.81009474e-02 -3.77695322e-01
2.24875957e-01 6.07411444e-01 8.38959992e-01 -1.51192296e+00
-1.57217890e-01 2.15760350e-01 3.02820921e-01 -8.31265301e-02
-2.31358796e-01 2.32576549e-01 7.01665059e-02 -7.56967723e-01
-4.65752631e-01 -9.04822826e-01 -2.61402726e-01 -1.71408698e-01
-3.06742907e-01 -2.12829679e-01 2.03666046e-01 -6.24080837e-01
1.70606053e+00 -2.16052985e+00 3.33818138e-01 8.05638060e-02
7.74615586e-01 1.54679179e-01 -3.23970437e-01 6.45907283e-01
-9.23446491e-02 5.71541667e-01 -1.07413255e-01 -2.56210715e-01
1.90175131e-01 -2.41202414e-01 -4.91616100e-01 1.56165853e-01
6.75985366e-02 1.08626997e+00 -1.27683282e+00 -3.02699894e-01
-3.40028107e-01 3.14415365e-01 -1.02750969e+00 5.88195384e-01
-2.29224399e-01 3.46620947e-01 -2.00077206e-01 3.98375362e-01
3.02622825e-01 -6.60919428e-01 -1.47598788e-01 2.49842897e-01
-3.33901346e-01 3.19508225e-01 -2.53072888e-01 1.50326669e+00
-6.98656678e-01 1.05024815e+00 -1.57664672e-01 -9.27352235e-02
7.37221003e-01 1.62994921e-01 -4.28821206e-01 -7.14805484e-01
-1.90045554e-02 8.63270834e-02 5.63876212e-01 -6.75319374e-01
7.57142663e-01 1.04776688e-01 -1.29795879e-01 1.05174839e+00
-4.25257862e-01 -3.73850018e-01 6.98331147e-02 3.86995882e-01
1.07549703e+00 -1.51246026e-01 4.82273340e-01 -5.87802649e-01
1.81596175e-01 -3.65033187e-02 2.68407375e-01 1.17751670e+00
-1.64968774e-01 4.29615140e-01 6.74629569e-01 -4.04082149e-01
-1.00523877e+00 -7.40049422e-01 2.89119959e-01 1.32007587e+00
-9.96708795e-02 -8.42462182e-01 -1.19046938e+00 -4.07214463e-01
-2.14201599e-01 1.27438962e+00 -7.28188813e-01 -4.29992408e-01
-3.60900700e-01 -4.05885100e-01 6.10452771e-01 5.12084544e-01
6.31395355e-02 -1.34437490e+00 -1.29031456e+00 -9.59388074e-03
-1.82987705e-01 -6.29694760e-01 -9.55786347e-01 -3.33590031e-01
-6.41209841e-01 -8.91459227e-01 -2.68595815e-01 -6.16514385e-01
1.14310622e+00 1.92377269e-01 1.67380488e+00 8.14125896e-01
-2.49771670e-01 3.60762358e-01 -3.04023832e-01 -4.87523496e-01
-1.02456760e+00 2.00047135e-01 -1.20534964e-01 -6.40453458e-01
2.49352187e-01 -3.82500261e-01 -2.99143910e-01 2.12181538e-01
-5.78174591e-01 4.46697116e-01 6.44431233e-01 7.08401382e-01
-2.36467838e-01 -3.56311619e-01 1.80957854e-01 -9.54575717e-01
1.29687524e+00 -4.97574121e-01 -5.12486875e-01 1.26899168e-01
-7.60790706e-01 2.65313268e-01 5.47535241e-01 -5.01817346e-01
-9.71404374e-01 -3.66993904e-01 1.73106521e-01 1.89224690e-01
4.36857603e-02 5.49248755e-01 4.13689554e-01 -2.25886390e-01
1.14112270e+00 9.27745849e-02 -1.21102512e-01 -1.07706398e-01
1.48280725e-01 4.74632829e-01 2.81934649e-01 -1.00832939e+00
5.94725609e-01 -1.52575061e-01 -7.74458408e-01 -5.79619110e-01
-6.82513177e-01 2.20956475e-01 -2.42778629e-01 1.48649430e-02
8.35400820e-01 -6.98298037e-01 -7.94179380e-01 3.76230896e-01
-1.68710208e+00 -6.41805530e-01 5.08626131e-03 9.96527448e-02
-3.99035156e-01 3.20557833e-01 -7.29186773e-01 -6.72077358e-01
-3.68085206e-01 -1.54570925e+00 9.56081033e-01 2.57847011e-01
-1.29994011e+00 -9.54211175e-01 2.73280084e-01 3.71862561e-01
9.57276165e-01 -8.04676861e-02 1.39270937e+00 -5.29924035e-01
-6.33131564e-01 -7.11129680e-02 -3.01434338e-01 -1.42685126e-03
1.04954451e-01 2.70931989e-01 -9.21025038e-01 -3.21599811e-01
3.99063639e-02 -2.37362057e-01 1.63225025e-01 7.06736594e-02
1.17399395e+00 -6.40699744e-01 -2.72682458e-01 3.63085926e-01
1.14747500e+00 1.20360561e-01 5.25506437e-01 2.23485813e-01
7.97749758e-01 5.55048704e-01 2.88972825e-01 3.63920927e-01
4.51653928e-01 4.56807137e-01 3.40639710e-01 1.01574957e-01
1.81073129e-01 -3.55299145e-01 4.77711052e-01 9.23318207e-01
-8.83707479e-02 -2.22663879e-01 -1.40291047e+00 4.34201896e-01
-1.72387624e+00 -9.01406407e-01 -2.67136872e-01 2.16248417e+00
1.06080198e+00 3.15463662e-01 -2.79861718e-01 -4.07551169e-01
5.53863406e-01 -3.28662694e-02 -3.34375679e-01 -7.91528702e-01
4.47782069e-01 -2.70161089e-02 4.41560261e-02 6.27706289e-01
-4.28015560e-01 7.16845810e-01 7.61576509e+00 3.43215466e-01
-9.31534648e-01 1.08915210e-01 6.53666854e-01 1.27271786e-01
-6.50471449e-01 1.68714240e-01 -5.39509714e-01 6.11360073e-01
1.02582181e+00 -7.78294504e-01 6.86765730e-01 1.01361609e+00
2.09662661e-01 -9.33795497e-02 -1.62820721e+00 7.11157382e-01
2.11804420e-01 -1.23902547e+00 -3.80019434e-02 -5.71533516e-02
9.18161809e-01 -1.86413690e-01 5.08171581e-02 5.18972218e-01
3.94955605e-01 -1.37285388e+00 1.00939023e+00 6.65839612e-01
3.38650405e-01 -5.31761467e-01 7.20397115e-01 7.01002240e-01
-8.27275693e-01 -1.26859531e-01 -1.83465034e-01 -6.26317203e-01
-1.16702199e-01 1.38473019e-01 -9.49588597e-01 -3.95816743e-01
5.45112848e-01 3.26517910e-01 -1.12925339e+00 7.49582827e-01
-2.11145908e-01 3.77560914e-01 5.28456450e-01 -2.21416041e-01
-1.46735549e-01 2.83930600e-01 3.97533298e-01 1.35501266e+00
1.47292092e-01 -1.59518436e-01 6.72921315e-02 1.57736790e+00
-2.90827040e-04 -1.44146476e-02 -5.92085600e-01 -7.57838607e-01
5.57865143e-01 1.01005340e+00 -5.33539593e-01 -3.14008176e-01
-3.84121895e-01 9.40045059e-01 5.33026755e-01 2.10022658e-01
-9.44902658e-01 -3.31245154e-01 7.37581074e-01 4.02208000e-01
-3.23687404e-01 -2.52421260e-01 -6.62107348e-01 -9.30790067e-01
-1.11662753e-01 -1.31780279e+00 -3.23840320e-01 -1.24675024e+00
-1.26140690e+00 9.03660357e-01 -7.04985783e-02 -7.84212232e-01
-2.01858088e-01 -3.38783145e-01 -9.06426549e-01 1.03490853e+00
-8.28735352e-01 -6.03304923e-01 -5.84336936e-01 -1.26089975e-01
6.44914746e-01 -1.76534861e-01 8.91309679e-01 6.45029685e-03
-3.97119701e-01 6.75700784e-01 -6.22213840e-01 -4.27349210e-02
5.69313705e-01 -1.21271515e+00 1.16257095e+00 7.86282003e-01
-1.05699360e-01 1.56565320e+00 9.92424965e-01 -7.28229105e-01
-1.25475907e+00 -8.74136329e-01 1.17556798e+00 -1.19672751e+00
6.60006821e-01 -3.02659780e-01 -1.12344825e+00 7.85815537e-01
6.19970977e-01 -4.75436389e-01 7.52143919e-01 -1.91585682e-02
-4.36683238e-01 5.77098608e-01 -8.35508049e-01 7.16116786e-01
1.01254535e+00 -8.92584205e-01 -8.18684936e-01 2.18516946e-01
1.02942371e+00 -5.03558934e-01 -5.35066962e-01 -1.40891269e-01
4.75317836e-01 -1.13776839e+00 6.58030212e-01 -4.94745612e-01
1.06476116e+00 -2.41174757e-01 4.81519699e-02 -1.42274320e+00
-5.76749444e-01 -5.97156525e-01 -1.10584106e-02 1.31275272e+00
4.43205118e-01 -6.23592436e-01 1.39562398e-01 1.07610059e+00
-2.44894907e-01 -5.62964261e-01 -1.23322576e-01 -3.31051856e-01
-2.05198489e-02 -2.17067018e-01 4.28454995e-01 1.05739284e+00
3.24635208e-01 2.56426960e-01 -1.46632388e-01 -5.13182487e-03
2.70412296e-01 6.80368021e-02 8.10764253e-01 -1.18940878e+00
-7.60852993e-01 -7.04014122e-01 -3.81132327e-02 -7.06971586e-01
3.62534583e-01 -1.08578336e+00 2.49785945e-01 -1.32036293e+00
6.45127535e-01 -1.62944332e-01 1.88738137e-01 3.93756986e-01
-4.56697315e-01 -1.77029699e-01 1.98424593e-01 1.68901160e-01
-5.02843082e-01 2.88036972e-01 9.52824116e-01 1.71661139e-01
-1.47020310e-01 -1.74471900e-01 -1.32956874e+00 8.46741378e-01
6.51511610e-01 -5.88844240e-01 -4.78257924e-01 -8.37557375e-01
8.76201928e-01 -1.24431193e-01 3.55097771e-01 -9.26881850e-01
2.66059995e-01 -3.71362537e-01 5.87694496e-02 2.24771760e-02
-3.20001096e-01 -3.73472899e-01 1.61866605e-01 6.97836518e-01
-6.76416636e-01 7.62902260e-01 4.26161647e-01 1.38274044e-01
3.45197879e-02 -4.93581176e-01 6.74981296e-01 -4.37191248e-01
-5.92348218e-01 -2.40260080e-01 -6.87636077e-01 3.81676435e-01
8.40402603e-01 -9.53846276e-02 -5.22987783e-01 -2.60669768e-01
-1.26723900e-01 1.44186839e-01 1.08632600e+00 5.87797344e-01
5.14068246e-01 -1.12609303e+00 -5.71055889e-01 2.73213953e-01
3.94484550e-01 -2.73446172e-01 -1.17242523e-01 6.90583110e-01
-6.95603371e-01 2.30252862e-01 -1.56344831e-01 -4.43291038e-01
-1.06793058e+00 5.72321177e-01 3.43684226e-01 3.12482640e-02
-1.98228925e-01 1.08810747e+00 4.45880055e-01 -4.11932915e-01
5.18261418e-02 -6.05675817e-01 2.76094586e-01 -3.84858459e-01
7.20638156e-01 2.44967327e-01 -2.66847312e-01 -4.05495763e-01
-3.25991333e-01 3.55634242e-01 -1.97707370e-01 1.30060330e-01
1.07505834e+00 9.61889252e-02 -5.58634222e-01 6.49432421e-01
6.64124131e-01 4.37819868e-01 -1.12831128e+00 2.58926183e-01
1.60161436e-01 -8.09338987e-01 -4.07302350e-01 -6.88185632e-01
-6.20091558e-01 8.72710526e-01 1.24746054e-01 3.01965952e-01
6.09194338e-01 -1.72547977e-02 4.26360399e-01 3.25235516e-01
6.16901696e-01 -7.41334558e-01 3.88613492e-01 6.68348253e-01
1.22286201e+00 -1.32573986e+00 -2.97777444e-01 -4.92935181e-02
-8.57801139e-01 8.33025455e-01 1.24637675e+00 4.73308228e-02
1.05291039e-01 3.66293639e-01 -1.24037907e-01 -2.32330412e-01
-1.21903777e+00 6.33628607e-01 2.45348752e-01 5.28596282e-01
1.20398402e+00 -6.73704445e-02 -1.13953449e-01 7.05452323e-01
-5.88379264e-01 1.70789715e-02 9.66732264e-01 8.69060338e-01
-3.82467955e-01 -8.17099810e-01 -3.01501751e-01 3.00958008e-01
-2.63419062e-01 -5.69794893e-01 -5.60179651e-01 4.69220251e-01
9.20879394e-02 8.71223986e-01 4.16961461e-02 -3.59733880e-01
2.33983457e-01 4.90544774e-02 4.39245224e-01 -1.26132989e+00
-1.08759248e+00 -6.97025180e-01 1.04438802e-02 -6.81086302e-01
2.38392770e-01 -4.42078650e-01 -1.29635704e+00 -4.76642162e-01
-2.61994004e-01 2.31293723e-01 3.57534468e-01 7.14844584e-01
6.39423192e-01 5.38504779e-01 1.98686182e-01 -8.72571230e-01
-5.45482337e-01 -8.20567191e-01 -7.57170515e-03 5.59022427e-01
3.33268136e-01 -1.89534143e-01 -3.91818285e-01 2.04469815e-01]
|
[7.957571029663086, 7.737530708312988]
|
53b88b21-3f76-4226-9e77-06e89bb294f8
|
simultaneous-job-interview-system-using
| null | null |
https://aclanthology.org/2022.sigdial-1.12
|
https://aclanthology.org/2022.sigdial-1.12.pdf
|
Simultaneous Job Interview System Using Multiple Semi-autonomous Agents
|
In recent years, spoken dialogue systems have been applied to job interviews where an applicant talks to a system that asks pre-defined questions, called on-demand and self-paced job interviews. We propose a simultaneous job interview system, where one interviewer can conduct one-on-one interviews with multiple applicants simultaneously by cooperating with the multiple autonomous job interview dialogue systems. However, it is challenging for interviewers to monitor and understand all the parallel interviews done by the autonomous system at the same time. As a solution to this issue, we implemented two automatic dialogue understanding functions: (1) response evaluation of each applicant’s responses and (2) keyword extraction as a summary of the responses. It is expected that interviewers, as needed, can intervene in one dialogue and smoothly ask a proper question that elaborates the interview. We report a pilot experiment where an interviewer conducted simultaneous job interviews with three candidates.
|
['Tatsuya Kawahara', 'Koji Inoue', 'Divesh Lala', 'Kenta Yamamoto', 'Yusuke Muraki', 'Haruki Kawai']
| null | null | null | null |
sigdial-acl-2022-9
|
['keyword-extraction', 'dialogue-understanding', 'spoken-dialogue-systems']
|
['natural-language-processing', 'natural-language-processing', 'speech']
|
[ 3.52907568e-01 7.67648458e-01 2.55555157e-02 -8.12578440e-01
-9.52911556e-01 -1.09551775e+00 6.64658368e-01 2.68593311e-01
-4.84733105e-01 7.98775613e-01 1.92995936e-01 -6.66042507e-01
1.61973849e-01 -5.67209065e-01 4.13267195e-01 -1.53460249e-01
4.63273793e-01 1.37988341e+00 1.68820634e-01 -6.15816116e-01
2.79103637e-01 3.40258479e-01 -1.01681519e+00 1.60098210e-01
8.64765346e-01 2.63544261e-01 4.94753331e-01 1.18583310e+00
-5.77071846e-01 9.92638528e-01 -1.02830040e+00 -2.05151826e-01
-3.67668793e-02 -6.97427511e-01 -1.54515123e+00 3.83109480e-01
-3.12467098e-01 -5.15590608e-01 -4.69873194e-03 6.59680545e-01
5.69083929e-01 2.45746613e-01 3.27092111e-01 -1.14321029e+00
1.30889798e-03 6.78349435e-01 -1.15021177e-01 -1.64540708e-01
1.13091350e+00 2.63576686e-01 8.19254756e-01 -5.86933315e-01
5.00404060e-01 1.30763233e+00 2.63307124e-01 5.65958798e-01
-1.21850646e+00 -2.60978311e-01 2.36600358e-02 -2.51513660e-01
-9.27839577e-01 -4.81072277e-01 6.33125603e-01 -3.49957168e-01
7.31302023e-01 5.32645822e-01 1.60146356e-01 7.93282747e-01
1.45898372e-01 5.13171554e-01 1.05551505e+00 -5.00537097e-01
1.90666780e-01 9.76991415e-01 4.57026780e-01 3.37389588e-01
-6.73393786e-01 -5.50088286e-01 -3.86655658e-01 -5.82775176e-01
5.58152020e-01 -4.65857983e-01 4.02829722e-02 2.74450570e-01
-1.11492419e+00 8.12119663e-01 -4.39092249e-01 4.92573470e-01
-6.83811367e-01 -5.19540608e-01 6.08408749e-01 6.92713857e-01
2.66579479e-01 7.33313322e-01 -2.45274276e-01 -6.85113668e-01
-5.79689562e-01 4.10673112e-01 1.78780866e+00 1.00963557e+00
7.77067065e-01 -4.62658942e-01 -3.04923981e-01 7.47711718e-01
-2.64861863e-02 2.18558624e-01 1.78910047e-01 -1.13277352e+00
5.08579314e-01 6.77248180e-01 7.62729347e-01 -8.32042158e-01
-3.91010851e-01 1.52107254e-01 -5.53637564e-01 -1.61565661e-01
5.07017136e-01 -6.53773725e-01 -2.73847252e-01 1.22925329e+00
5.34462094e-01 -8.42543423e-01 2.35429257e-01 9.69497263e-01
1.01719654e+00 6.97598159e-01 9.85884592e-02 -5.15656650e-01
1.80730140e+00 -9.39758956e-01 -1.23454607e+00 -4.62959737e-01
5.11941195e-01 -9.47099209e-01 1.10225224e+00 2.59903640e-01
-1.28409123e+00 -7.06103981e-01 -4.85560894e-01 2.72963289e-02
9.54349488e-02 2.74273336e-01 4.56782401e-01 6.01167381e-01
-9.42258894e-01 2.30870694e-02 -4.24615145e-01 -5.93846500e-01
-9.16987717e-01 8.07190895e-01 -3.06564361e-01 2.41818562e-01
-1.50416195e+00 8.90429556e-01 -1.65196776e-01 3.11120540e-01
-4.82507765e-01 5.31819500e-02 -7.88321137e-01 2.12496161e-01
5.47158718e-01 -4.53754216e-01 2.17498803e+00 -8.47847879e-01
-2.00093794e+00 8.88874829e-01 -4.81776178e-01 9.68271792e-02
5.09038806e-01 -3.92085537e-02 -1.06596716e-01 2.19291791e-01
4.70731467e-01 2.51203328e-01 2.90235411e-02 -1.21831131e+00
-6.19854569e-01 -5.21629989e-01 4.81975228e-01 6.02859437e-01
1.00145027e-01 5.33495367e-01 -4.30364996e-01 3.37912619e-01
-1.05179846e-02 -1.00586402e+00 -4.27876443e-01 -8.17077935e-01
-4.26959455e-01 -3.90072942e-01 7.12275624e-01 -5.58474541e-01
1.15715814e+00 -1.83901656e+00 -1.20904498e-01 1.17204733e-01
6.42240494e-02 2.02302769e-01 -4.74580154e-02 1.02399790e+00
9.62215587e-02 -1.22877821e-01 -2.95816711e-03 -5.43370306e-01
2.83043683e-01 1.80081695e-01 -1.58402637e-01 5.47456630e-02
6.98281899e-02 6.02991819e-01 -9.77777421e-01 -6.19254768e-01
-4.41594869e-02 -3.32232863e-01 2.52590813e-02 9.32908237e-01
-5.70162013e-02 8.61714840e-01 -7.28355110e-01 3.23575854e-01
4.21129733e-01 -9.70707908e-02 6.73679769e-01 5.08500278e-01
-2.90723622e-01 6.96398377e-01 -1.07249570e+00 1.25631630e+00
-6.08573854e-01 5.71814597e-01 8.90881002e-01 -8.44661593e-01
1.25661302e+00 9.51696694e-01 3.51537675e-01 -5.20244181e-01
-3.30956690e-02 1.21113189e-01 -1.48062155e-01 -5.98540425e-01
9.72040176e-01 -1.09579206e-01 -6.83114648e-01 1.04304254e+00
-2.02134892e-01 -3.01881820e-01 1.05607606e-01 4.18275326e-01
1.18106949e+00 -3.23254198e-01 5.97434580e-01 6.74576536e-02
7.68779159e-01 4.23738331e-01 4.52593565e-01 1.03569698e+00
-4.10636306e-01 2.32607454e-01 7.64916182e-01 -5.31960189e-01
-6.63300216e-01 -5.89183509e-01 4.54637021e-01 1.29446793e+00
2.37114549e-01 -3.79188389e-01 -7.17547596e-01 -6.64265275e-01
-4.25495058e-01 6.90945983e-01 5.61565347e-03 3.29648107e-01
-5.59882939e-01 1.55925483e-01 4.92495924e-01 1.01456448e-01
7.77568936e-01 -1.29486358e+00 -7.29711890e-01 6.57329679e-01
-7.90605545e-01 -1.11659670e+00 -7.77154863e-01 3.80589455e-01
-1.33010745e-01 -5.44880986e-01 -6.20138288e-01 -8.43125761e-01
6.02645695e-01 3.88792366e-01 1.06190014e+00 6.30811900e-02
1.81971654e-01 5.72970808e-01 -3.07000041e-01 -3.29877138e-01
-8.85277033e-01 6.27732456e-01 1.38771785e-02 -6.91940933e-02
3.40612948e-01 -3.13582361e-01 -2.54616022e-01 8.42876256e-01
-7.05023050e-01 1.35229617e-01 3.50483656e-01 7.23815143e-01
-1.71215862e-01 -1.93183094e-01 9.50183928e-01 -1.35035324e+00
1.54360652e+00 -3.15412074e-01 -4.50513512e-01 3.92487764e-01
-3.78313661e-01 -1.21715568e-01 6.11959577e-01 -3.13974917e-01
-1.39140368e+00 4.32976663e-01 -2.07847074e-01 4.75711495e-01
-5.39214194e-01 6.50460482e-01 -1.67745262e-01 1.50223002e-01
6.87696755e-01 2.78294384e-01 1.40886828e-01 8.53554532e-03
3.93314287e-02 1.45258951e+00 6.77920520e-01 -5.94850719e-01
6.30903661e-01 -3.18189979e-01 -6.82020009e-01 -7.65879273e-01
-2.55792946e-01 -9.86341357e-01 -6.11172616e-01 -5.20708084e-01
8.14280212e-01 -7.70536542e-01 -1.42037070e+00 6.04946241e-02
-1.69421303e+00 -4.56462592e-01 2.18104392e-01 3.01187336e-01
-4.20622319e-01 2.42372319e-01 -6.67498469e-01 -1.42091382e+00
-3.59858900e-01 -1.20702899e+00 1.00403488e+00 6.58728302e-01
-1.18932509e+00 -6.92219794e-01 7.01339319e-02 7.16009974e-01
4.31529164e-01 -3.21349382e-01 6.75618052e-01 -1.04939973e+00
-1.39859036e-01 -4.81151432e-01 4.62982766e-02 -2.03717917e-01
3.27975273e-01 -3.03530186e-01 -9.83520687e-01 -5.51973656e-02
4.27841425e-01 -5.86284459e-01 -1.47382721e-01 -8.24467316e-02
3.31950217e-01 -5.65852702e-01 -2.59110570e-01 -4.45041120e-01
4.54367816e-01 7.32961953e-01 3.44941556e-01 4.10661660e-02
7.03225583e-02 1.17798555e+00 1.03734791e+00 5.01736283e-01
7.74189353e-01 7.43455589e-01 -2.71971166e-01 -2.55611867e-01
8.44756365e-01 6.77296752e-03 3.22812945e-01 7.06515372e-01
2.52602071e-01 -2.83150047e-01 -1.11928272e+00 6.36210799e-01
-2.07355475e+00 -7.41507947e-01 -1.44656628e-01 1.89539027e+00
1.08742976e+00 4.07367013e-02 3.06002378e-01 -2.28177086e-01
9.12846804e-01 -1.21820047e-02 -2.76835740e-01 -9.35272455e-01
6.30702913e-01 1.50965396e-04 8.28050300e-02 9.08652842e-01
-6.90804183e-01 9.09088135e-01 6.27554798e+00 -3.40230279e-02
-8.65563095e-01 -3.06688219e-01 7.66558528e-01 4.27313238e-01
-3.45198214e-01 5.70336580e-01 -8.76001954e-01 1.17394045e-01
1.09708095e+00 -4.80827719e-01 2.43852615e-01 8.62188339e-01
4.70328838e-01 -7.25106835e-01 -1.13009107e+00 5.16701579e-01
-2.90137351e-01 -9.25730586e-01 -5.59715152e-01 4.86143678e-02
9.59743783e-02 -7.63216376e-01 -5.50186217e-01 6.61740005e-01
5.39676964e-01 -8.40606034e-01 2.05995098e-01 2.17786327e-01
5.56662619e-01 -8.00263584e-01 9.40753400e-01 1.10676718e+00
-1.15040827e+00 -3.85506032e-03 8.20700973e-02 -5.19309282e-01
4.33871210e-01 1.20532073e-01 -1.76065052e+00 6.41799152e-01
4.62065227e-02 -5.27831733e-01 6.71543460e-03 3.31657916e-01
-1.26692787e-01 3.78871381e-01 -3.45535023e-04 -2.88276106e-01
1.70430958e-01 -4.14964527e-01 3.98502946e-01 1.13932061e+00
-1.04108661e-01 6.54796600e-01 7.65759766e-01 7.58278787e-01
2.31427282e-01 5.00748530e-02 -6.55441940e-01 -2.23016739e-01
6.73207164e-01 1.56322408e+00 -7.53215492e-01 -4.44405407e-01
-1.10952742e-01 1.14122355e+00 -1.11553356e-01 4.46491122e-01
-4.67607468e-01 -9.49219704e-01 1.82255819e-01 8.14398900e-02
-3.91693503e-01 -4.11204278e-01 -7.14405775e-02 -5.52920818e-01
1.54072031e-01 -1.31086981e+00 7.62021840e-02 -6.38818741e-01
-8.10549617e-01 8.73253345e-01 -1.59836575e-01 -6.42329514e-01
-9.66321290e-01 1.76959425e-01 -8.11955512e-01 1.37914777e+00
-8.76428545e-01 -5.81791043e-01 -3.13642919e-01 1.51979625e-01
9.21277523e-01 2.65420862e-02 1.29624283e+00 1.46203697e-01
-3.89762163e-01 2.58421779e-01 -8.73783469e-01 2.51066059e-01
9.89031315e-01 -1.35677648e+00 6.03032351e-01 1.08568422e-01
-4.12013322e-01 8.96644056e-01 8.08978438e-01 -8.10753763e-01
-1.51184905e+00 -3.99914324e-01 1.59285426e+00 -9.84370783e-02
1.60352737e-01 -5.69901586e-01 -9.65735972e-01 4.85501111e-01
6.73322201e-01 -6.47070169e-01 6.63256407e-01 2.68107980e-01
6.15193129e-01 5.13656698e-02 -1.00131094e+00 5.20160019e-01
2.37101987e-01 -8.71562183e-01 -5.84796369e-01 4.36827153e-01
6.95328653e-01 -7.62353659e-01 -5.72894394e-01 -1.35637522e-01
4.09957051e-01 -7.19271064e-01 1.85283542e-01 -3.55409950e-01
4.12151311e-03 -1.90067291e-01 3.80915940e-01 -1.18024015e+00
-3.62139605e-02 -1.34731400e+00 7.86660373e-01 1.49975371e+00
5.07780194e-01 -7.53296375e-01 6.34388864e-01 1.61844623e+00
1.98818475e-01 -3.56565416e-01 -6.34876072e-01 -2.55382061e-01
-6.39229476e-01 3.04111224e-02 4.59335297e-01 7.07894146e-01
8.55109751e-01 1.09339869e+00 -4.00905639e-01 2.37351462e-01
-1.47983134e-01 7.29623660e-02 1.30766749e+00 -1.17538130e+00
-3.76008093e-01 1.63827911e-01 2.27369830e-01 -1.35099792e+00
1.48092851e-01 -2.79303670e-01 7.30234981e-01 -1.61988819e+00
8.40223730e-02 -2.11287305e-01 6.52012050e-01 2.73557752e-01
-3.34259868e-01 -7.62253523e-01 7.09399804e-02 2.04748020e-01
-6.51477993e-01 1.80603996e-01 1.09751832e+00 -2.63496544e-02
-7.69314408e-01 7.17699170e-01 -8.21982026e-01 3.26973528e-01
6.83985353e-01 -3.98613125e-01 -4.62973654e-01 -8.10397416e-02
2.52770633e-02 1.29588270e+00 -1.29327208e-01 -7.62370527e-01
1.02360713e+00 -3.49865168e-01 -1.06514893e-01 -5.47705412e-01
1.67907417e-01 -5.53613424e-01 1.05597302e-01 1.97427094e-01
-8.34349453e-01 3.46962631e-01 5.03819473e-02 1.67827412e-01
-4.70406890e-01 -3.89127314e-01 3.59853089e-01 -4.59150940e-01
-2.10263416e-01 -1.93860278e-01 -1.38970852e+00 -2.69952923e-01
9.94268239e-01 -1.56331107e-01 -1.28977463e-01 -1.20688498e+00
-7.10643172e-01 7.60552347e-01 1.63350508e-01 1.00563429e-01
3.53815496e-01 -7.70423412e-01 -4.89381850e-01 1.24254469e-02
-1.00837126e-01 9.18382555e-02 -1.38344113e-02 5.09082735e-01
-2.70578980e-01 6.99878395e-01 3.42544504e-02 -5.26781440e-01
-1.68054843e+00 6.38500368e-03 1.49772257e-01 -6.52534723e-01
-1.55056939e-02 4.50966567e-01 1.01690114e-01 -9.86641467e-01
4.92119879e-01 1.05843641e-01 -3.32943857e-01 2.54204243e-01
5.72766364e-01 -8.69883001e-02 8.27020733e-04 -2.73208290e-01
-3.58618408e-01 4.52086031e-02 -3.05758506e-01 -8.81336093e-01
8.67188632e-01 -5.60817182e-01 -1.91508278e-01 8.17111075e-01
5.87118745e-01 1.03796154e-01 -6.74351156e-01 -4.49783355e-01
2.69482970e-01 -3.16706717e-01 -5.77061415e-01 -7.54830956e-01
-1.02148034e-01 4.75308478e-01 -6.70629516e-02 7.88990855e-01
8.66999388e-01 -1.21014968e-01 7.79790521e-01 8.67947519e-01
3.76918525e-01 -1.36496818e+00 1.64800599e-01 8.28924954e-01
8.34105790e-01 -1.28755021e+00 -2.79033214e-01 -4.74062115e-01
-1.32343626e+00 1.29524267e+00 8.15579891e-01 5.71949899e-01
-8.73553008e-02 2.58020222e-01 6.16267800e-01 -3.57068270e-01
-1.12492609e+00 7.00903907e-02 -1.37544066e-01 3.09314162e-01
7.48649180e-01 3.97814065e-02 -4.98490334e-01 7.24802732e-01
-2.03420505e-01 6.85316995e-02 7.98452735e-01 1.13440716e+00
-5.24396300e-01 -1.48805332e+00 -5.76966286e-01 -3.77244353e-02
-1.78377494e-01 2.94073522e-01 -1.34305346e+00 5.49413204e-01
-5.52152038e-01 1.59051085e+00 -1.75480172e-01 -3.53025407e-01
7.09734678e-01 3.47337812e-01 -2.76868999e-01 -1.15176487e+00
-1.32683825e+00 -1.16672799e-01 8.15836847e-01 -2.35514730e-01
2.94251293e-02 -3.66180092e-01 -1.46647513e+00 -1.89080447e-01
-2.71715283e-01 9.35056686e-01 6.24304056e-01 1.06907964e+00
2.57371038e-01 2.93760121e-01 1.22959256e+00 -6.07361019e-01
-8.71258557e-01 -1.18665099e+00 -4.87869650e-01 2.06772907e-04
2.87852556e-01 1.03713490e-01 -5.14754355e-02 -8.71379450e-02]
|
[12.932321548461914, 7.986958026885986]
|
9f935eca-ea70-45d5-9121-60103a47477b
|
deep-machine-learning-based-egyptian-vehicle
|
2107.11640
| null |
https://arxiv.org/abs/2107.11640v1
|
https://arxiv.org/pdf/2107.11640v1.pdf
|
Deep Machine Learning Based Egyptian Vehicle License Plate Recognition Systems
|
Automated Vehicle License Plate (VLP) detection and recognition have ended up being a significant research issue as of late. VLP localization and recognition are some of the most essential techniques for managing traffic using digital techniques. In this paper, four smart systems are developed to recognize Egyptian vehicles license plates. Two systems are based on character recognition, which are (System1, Characters Recognition with Classical Machine Learning) and (System2, Characters Recognition with Deep Machine Learning). The other two systems are based on the whole plate recognition which are (System3, Whole License Plate Recognition with Classical Machine Learning) and (System4, Whole License Plate Recognition with Deep Machine Learning). We use object detection algorithms, and machine learning based object recognition algorithms. The performance of the developed systems has been tested on real images, and the experimental results demonstrate that the best detection accuracy rate for VLP is provided by using the deep learning method. Where the VLP detection accuracy rate is better than the classical system by 32%. However, the best detection accuracy rate for Vehicle License Plate Arabic Character (VLPAC) is provided by using the classical method. Where VLPAC detection accuracy rate is better than the deep learning-based system by 6%. Also, the results show that deep learning is better than the classical technique used in VLP recognition processes. Where the recognition accuracy rate is better than the classical system by 8%. Finally, the paper output recommends a robust VLP recognition system based on both statistical and deep machine learning.
|
['Hany Elnashar', 'Mohamed Taha Abou-Kreisha', 'Mohamed Shehata']
|
2021-07-24
| null | null | null | null |
['license-plate-recognition']
|
['computer-vision']
|
[-4.09330904e-01 -7.43523419e-01 3.30449119e-02 8.87031779e-02
-5.14709711e-01 -5.36125660e-01 6.63618267e-01 -4.78356928e-01
-4.49175090e-01 4.50390875e-01 -4.62939709e-01 -2.61993468e-01
3.55265588e-01 -9.43119228e-01 -5.34833014e-01 -8.73193324e-01
2.78312922e-01 6.74404025e-01 8.60577345e-01 -2.96507478e-01
9.43077266e-01 8.27876627e-01 -1.36866093e+00 4.49207902e-01
5.77120960e-01 9.96007442e-01 1.52662965e-02 8.62704337e-01
-5.43716908e-01 9.32483673e-01 -5.68168819e-01 -2.83805221e-01
4.63209748e-01 -1.09904341e-01 -2.30097517e-01 4.72375713e-02
1.39517695e-01 -4.72196668e-01 -5.58388770e-01 1.09744108e+00
3.50316137e-01 9.07411799e-02 1.11697686e+00 -1.31391180e+00
-8.17199647e-01 -2.63777256e-01 -5.95792472e-01 2.92751074e-01
-1.37362601e-02 1.98779777e-02 1.58413604e-01 -1.30437875e+00
1.74681470e-01 1.05174756e+00 1.10380018e+00 4.46308374e-01
-3.93201530e-01 -8.48735034e-01 -7.32639372e-01 8.71698976e-01
-1.83888721e+00 -1.94094449e-01 4.65894789e-01 -7.61499286e-01
1.06016016e+00 1.95209876e-01 2.97971427e-01 3.10017854e-01
5.77999413e-01 9.04195309e-01 1.15583456e+00 -6.78368092e-01
1.83246225e-01 5.54331064e-01 6.55010223e-01 8.35330129e-01
1.82298303e-01 -5.98867284e-03 3.29584628e-01 3.59780371e-01
8.09366524e-01 1.85586423e-01 4.70362753e-01 3.67695004e-01
-4.95091379e-01 9.05264378e-01 -8.22306424e-02 6.63490117e-01
-2.05240503e-01 -2.21889675e-01 2.51048446e-01 1.24457836e-01
-8.70912820e-02 1.82129238e-02 -1.34786099e-01 -2.85832494e-01
-1.13665712e+00 1.25038981e-01 7.98554897e-01 9.10574853e-01
7.91834474e-01 6.59658253e-01 -4.75739092e-02 1.16939008e+00
6.53625011e-01 9.30162728e-01 8.98262918e-01 -4.63394463e-01
3.18249375e-01 6.61063969e-01 1.99316651e-01 -1.42344749e+00
-3.29915792e-01 2.82861650e-01 -6.07629776e-01 7.72363782e-01
4.28287148e-01 -3.95912439e-01 -1.18787014e+00 4.49432462e-01
-4.30782318e-01 2.38350838e-01 3.45241994e-01 7.14442074e-01
1.00256658e+00 1.59229779e+00 -1.96775153e-01 2.34819055e-01
1.38468647e+00 -1.22615409e+00 -7.64603257e-01 -3.52926910e-01
5.08793950e-01 -1.07599962e+00 5.64769328e-01 4.22392249e-01
-6.43265963e-01 -7.37776577e-01 -1.27856886e+00 3.06205601e-01
-8.54057789e-01 6.79687679e-01 1.44086689e-01 1.28913999e+00
-8.71648669e-01 -8.68579969e-02 -5.58452129e-01 -5.64863086e-01
2.16620490e-01 5.57147741e-01 -4.40701336e-01 -1.21028334e-01
-8.86131406e-01 1.20239818e+00 4.44548547e-01 1.27394021e-01
-5.19449711e-01 2.18288869e-01 -4.12417948e-01 -5.47463410e-02
-1.84084829e-02 5.38270473e-01 7.81335056e-01 -1.12386286e+00
-1.83990598e+00 7.47684002e-01 3.87659185e-02 -2.07643405e-01
4.32674050e-01 4.82988618e-02 -1.04412985e+00 2.47854784e-01
-1.06524646e-01 2.10164577e-01 5.88505805e-01 -1.04106104e+00
-9.66245472e-01 -1.29567146e-01 -4.40409213e-01 -5.18158227e-02
-8.64562765e-02 3.72609794e-01 -7.13511109e-01 -3.48227359e-02
5.67268319e-02 -9.28592265e-01 3.78814846e-01 -3.53445113e-01
-1.69610500e-01 -4.34477866e-01 1.21750534e+00 -1.07455909e+00
7.96760857e-01 -2.17123866e+00 -9.60525036e-01 5.87455750e-01
-2.28051186e-01 1.11536109e+00 -7.88594538e-04 4.74931628e-01
2.24004298e-01 9.12513360e-02 7.00327605e-02 1.52849168e-01
7.87675008e-02 2.54112542e-01 3.05282939e-02 4.79331374e-01
1.71638802e-01 6.51108980e-01 -2.47815959e-02 -5.72290719e-01
5.93641222e-01 4.30162936e-01 -5.99416420e-02 -2.10277885e-01
5.08131206e-01 -3.08515280e-01 -3.46772701e-01 9.29033875e-01
1.23163569e+00 2.80819923e-01 -3.02786767e-01 -1.83401540e-01
-5.29929221e-01 -7.03705192e-01 -1.43474519e+00 2.24280074e-01
-1.69487402e-01 1.42395115e+00 -1.38434634e-01 -1.23258173e+00
1.44849527e+00 2.45266289e-01 -2.04946846e-02 -8.29045653e-01
2.79406607e-01 5.66673219e-01 -1.90596990e-02 -1.05379570e+00
4.24511045e-01 -6.70044646e-02 4.12866205e-01 -1.44119844e-01
-7.91969895e-02 1.93515271e-01 1.82251602e-01 -3.17371905e-01
5.70509911e-01 -2.56737381e-01 -1.34360269e-01 -7.99397826e-02
1.23832369e+00 3.05911064e-01 4.32139158e-01 8.45259190e-01
-4.73442942e-01 5.56761265e-01 4.02938038e-01 -5.00343263e-01
-1.21697497e+00 -7.23013997e-01 -3.13612700e-01 6.71882153e-01
2.71448731e-01 5.55987716e-01 -6.58344448e-01 -3.11901867e-01
-1.02954097e-02 5.12232006e-01 -1.81702539e-01 1.47247657e-01
-7.59702027e-01 -8.36491644e-01 1.13217926e+00 7.00907052e-01
1.34753644e+00 -1.20226395e+00 1.02111101e-01 -6.99317316e-03
1.19490474e-01 -1.10685802e+00 -7.39281699e-02 -4.91617203e-01
-3.95989954e-01 -1.20068014e+00 -8.96019399e-01 -1.52920592e+00
4.68156427e-01 1.31970152e-01 3.47636789e-01 1.92227185e-01
-8.67461190e-02 2.26848349e-01 -3.63122821e-01 -3.69319201e-01
-7.58919477e-01 -3.74809116e-01 1.49601087e-01 3.68614584e-01
1.00781393e+00 2.24791721e-01 -1.65516675e-01 3.88092130e-01
-6.49218619e-01 -3.28571171e-01 1.16922069e+00 4.88427669e-01
1.76294878e-01 3.09883416e-01 4.85576391e-01 -3.48867178e-01
5.76309621e-01 -2.90498137e-01 -1.02286506e+00 2.47146010e-01
-5.34616947e-01 -5.58712482e-01 5.87580144e-01 -1.49007276e-01
-1.00507689e+00 -1.32289380e-01 -7.29689121e-01 -2.37068728e-01
-4.91681337e-01 2.42634073e-01 -3.51334102e-02 -6.03505313e-01
4.31241691e-01 9.94091451e-01 3.24259609e-01 -2.73563862e-01
-2.34518170e-01 1.37624443e+00 4.18912649e-01 2.68359054e-02
7.13276565e-01 1.43897966e-01 -1.80360720e-01 -1.40993118e+00
3.41886133e-01 -6.39433026e-01 -5.21229684e-01 -5.65073252e-01
1.10214615e+00 -5.66885889e-01 -1.05135608e+00 1.31878054e+00
-9.63679254e-01 9.27184597e-02 3.63080829e-01 7.57304013e-01
-2.23500848e-01 5.33148348e-01 -8.03994298e-01 -1.12130630e+00
-3.83866765e-02 -1.28950977e+00 5.45229673e-01 4.78381068e-01
4.47307736e-01 -1.02629232e+00 1.71908483e-01 5.61242998e-01
5.11837482e-01 2.53682993e-02 6.28390193e-01 -1.29640925e+00
-4.64933842e-01 -1.13166952e+00 -6.44863427e-01 8.43906224e-01
-9.31258425e-02 4.60895032e-01 -6.42171979e-01 3.05542499e-01
-1.74575999e-01 8.25734902e-03 6.21626616e-01 3.62367839e-01
6.44653022e-01 -2.80791581e-01 -2.81891823e-01 3.39048624e-01
1.72733629e+00 1.02743912e+00 1.32918286e+00 8.66814792e-01
5.41251123e-01 2.96024770e-01 4.62711394e-01 -1.07223973e-01
1.88403904e-01 4.37771559e-01 -8.49014055e-03 3.32063623e-02
-9.24035087e-02 1.43943429e-01 6.79046333e-01 8.68823528e-01
-2.81425357e-01 -1.81192428e-01 -1.51227963e+00 1.75590575e-01
-1.53420138e+00 -1.47487223e+00 -8.31294894e-01 1.87104511e+00
2.91974813e-01 3.92381251e-02 1.76909104e-01 3.77339423e-01
1.00075305e+00 -2.33897403e-01 -2.77713574e-02 -8.97484720e-01
-2.37613782e-01 -7.26014003e-02 7.83404350e-01 5.54747403e-01
-1.36954486e+00 1.09563148e+00 6.07733202e+00 9.94236171e-01
-1.45471931e+00 4.26358730e-02 3.48312378e-01 6.50022686e-01
6.84612870e-01 -5.07121623e-01 -1.18701577e+00 8.42331529e-01
9.62281883e-01 5.48666157e-02 1.21774472e-01 8.68281305e-01
5.66741340e-02 -2.20316738e-01 -4.53417212e-01 1.30746555e+00
5.41949630e-01 -1.49119151e+00 1.04864106e-01 -4.67063338e-02
5.39349794e-01 1.44059852e-01 1.02126533e-02 8.33594143e-01
-1.76672816e-01 -9.39795256e-01 5.36781013e-01 7.30894923e-01
4.74052191e-01 -9.22970295e-01 1.45403636e+00 5.59906006e-01
-8.64404202e-01 -2.16287717e-01 -6.71058774e-01 -4.90123704e-02
-3.87686521e-01 -5.50386645e-02 -9.91571248e-01 1.26872128e-02
4.08839077e-01 5.15272677e-01 -4.90113407e-01 1.39212978e+00
1.38263568e-01 9.36141372e-01 -2.28455424e-01 -7.19839811e-01
6.34947240e-01 -5.51283717e-01 4.34067369e-01 1.83714771e+00
4.74581480e-01 7.26246759e-02 3.23937982e-02 6.00375891e-01
1.56860858e-01 3.77606958e-01 -5.46507299e-01 -1.59115680e-02
3.82164717e-01 1.15286434e+00 -8.27364981e-01 -5.67934036e-01
-6.76729918e-01 8.02243352e-01 -6.65600657e-01 3.59640032e-01
-9.85415220e-01 -8.52257788e-01 4.23881263e-02 1.90401509e-01
4.16527539e-01 -2.43774414e-01 -3.67161602e-01 -6.80107594e-01
-2.29783759e-01 -5.73671639e-01 -5.93226701e-02 -7.91993439e-01
-1.06218100e+00 3.33367288e-01 -2.93932974e-01 -1.53992403e+00
-4.12856713e-02 -1.48201764e+00 -8.52747083e-01 9.42730248e-01
-1.26008630e+00 -1.11314619e+00 -1.19595304e-01 7.77881563e-01
8.12893093e-01 -1.16833973e+00 5.67141891e-01 5.56662619e-01
-8.00867021e-01 6.03195786e-01 1.06868160e+00 8.38287532e-01
4.07867372e-01 -7.14548945e-01 -2.51268834e-01 9.39642847e-01
-2.19407186e-01 1.59617677e-01 2.28618741e-01 -6.05940759e-01
-1.39746833e+00 -9.85227466e-01 7.36072361e-01 -1.90223485e-01
2.03688979e-01 -8.23575817e-03 -8.15870404e-01 6.00083351e-01
1.32081702e-01 -2.52318770e-01 7.27868736e-01 -5.83596170e-01
-1.37993125e-02 -3.51348042e-01 -1.62324464e+00 3.55493754e-01
-4.50214684e-01 -2.95544833e-01 -7.72912979e-01 3.49600792e-01
-3.81134599e-01 -9.27099437e-02 -2.57439196e-01 -3.43956202e-02
8.49518955e-01 -7.31755197e-01 7.97005177e-01 -8.92157108e-03
1.68360978e-01 -8.11400414e-01 -4.17794585e-01 -5.37203312e-01
-5.77237904e-01 3.14768046e-01 2.91497797e-01 1.32342350e+00
4.36258674e-01 -9.82830167e-01 8.11234057e-01 7.13645160e-01
-2.48382285e-01 -1.61749676e-01 -9.57842171e-01 -1.04060638e+00
2.57795393e-01 -3.28428715e-01 1.13111734e-03 8.68909717e-01
-4.09014344e-01 -2.08941951e-01 -3.75241369e-01 3.99960816e-01
2.36083627e-01 -4.37794536e-01 6.18724823e-01 -9.06744421e-01
1.57370448e-01 -6.74005628e-01 -1.17575419e+00 -6.85441911e-01
-6.35270551e-02 -5.62932253e-01 1.88970536e-01 -1.68314064e+00
6.65522888e-02 -1.09437788e-02 -2.71964818e-01 4.90295619e-01
4.11634803e-01 7.18225002e-01 2.53998578e-01 5.78530908e-01
-3.74246657e-01 -1.27896238e-02 7.11020768e-01 -4.78353500e-01
-2.52518207e-01 1.79571688e-01 -6.56735376e-02 1.01905465e+00
9.81010556e-01 -1.05570219e-01 1.60813853e-01 -3.74244779e-01
-2.25750774e-01 -2.24211141e-01 1.54767543e-01 -1.48031950e+00
6.62475049e-01 -9.61314514e-03 8.38918090e-01 -8.81332994e-01
3.45167249e-01 -7.58516133e-01 -3.28834981e-01 6.97772205e-01
4.20167685e-01 -1.20952256e-01 6.29169345e-01 4.93287183e-02
-4.34928805e-01 -6.62319601e-01 1.02227581e+00 -5.00079282e-02
-1.67101204e+00 -6.43667057e-02 -1.27806711e+00 -3.89287323e-01
1.35643589e+00 -1.05917490e+00 -2.77780920e-01 -1.56494364e-01
-5.29901028e-01 -3.00319418e-02 1.14836179e-01 4.48847264e-01
9.29018736e-01 -1.42367756e+00 -7.99323738e-01 4.02903467e-01
-1.31011710e-01 -8.35836291e-01 2.72062384e-02 7.31655002e-01
-1.81900752e+00 7.80888438e-01 -7.41152287e-01 -5.01696765e-01
-1.32882750e+00 1.45816550e-01 5.98638058e-01 2.87358791e-01
-4.08717752e-01 6.12685978e-01 -2.71460980e-01 -3.80910367e-01
-9.36259851e-02 3.19152951e-01 -8.82581830e-01 -7.43468776e-02
8.94303083e-01 9.81517494e-01 2.08340406e-01 -1.11558211e+00
-6.67341948e-01 1.13728869e+00 -1.22999921e-01 -1.09423034e-01
1.19725704e+00 2.50526786e-01 -1.60036713e-01 3.30924779e-01
1.14087391e+00 2.05828175e-01 -7.77149677e-01 2.49161199e-01
7.39876833e-03 -5.75679779e-01 -5.22073060e-02 -1.09839618e+00
-9.24663961e-01 9.99958038e-01 1.18414080e+00 1.78409472e-01
7.42369950e-01 -6.78309500e-01 6.07365966e-01 7.58387089e-01
2.31347203e-01 -1.59929717e+00 -2.30447471e-01 9.66068387e-01
8.12522709e-01 -1.35297370e+00 -3.91412795e-01 -5.37769310e-02
-7.23084152e-01 1.78211534e+00 7.09723353e-01 -6.62420094e-01
8.00062180e-01 3.57833475e-01 2.26774812e-01 2.09968910e-01
1.40709812e-02 1.08474158e-01 2.64296293e-01 7.52138495e-01
1.84342712e-01 3.81822400e-02 -4.01158273e-01 7.82327056e-01
2.24742085e-01 8.76559094e-02 7.46432304e-01 7.51324177e-01
-1.04632342e+00 -8.09814870e-01 -1.01712239e+00 3.76464933e-01
-4.32108670e-01 -3.85125577e-02 -3.88615936e-01 1.11042440e+00
4.24998224e-01 1.07926905e+00 4.54467565e-01 -4.55722809e-01
2.73295194e-01 3.65363181e-01 1.01630144e-01 -9.91068110e-02
-9.50924382e-02 -3.13561589e-01 -1.54179931e-01 3.55916545e-02
-2.21796688e-02 -5.03148079e-01 -1.51809275e+00 -7.95445085e-01
-2.25172088e-01 5.46185225e-02 9.90617037e-01 1.16110599e+00
-4.44264747e-02 3.45839560e-02 5.29693663e-01 -7.19424427e-01
-2.30828911e-01 -9.35343206e-01 -8.54191124e-01 -1.77697837e-02
1.54165998e-01 -3.47054034e-01 -1.51584551e-01 1.38084874e-01]
|
[9.823433876037598, -4.968441963195801]
|
26bf426c-731b-4462-bac3-3b1d824f7a43
|
demystifying-misconceptions-in-social-bots
|
2303.17251
| null |
https://arxiv.org/abs/2303.17251v1
|
https://arxiv.org/pdf/2303.17251v1.pdf
|
Demystifying Misconceptions in Social Bots Research
|
The science of social bots seeks knowledge and solutions to one of the most debated forms of online misinformation. Yet, social bots research is plagued by widespread biases, hyped results, and misconceptions that set the stage for ambiguities, unrealistic expectations, and seemingly irreconcilable findings. Overcoming such issues is instrumental towards ensuring reliable solutions and reaffirming the validity of the scientific method. In this contribution we revise some recent results in social bots research, highlighting and correcting factual errors as well as methodological and conceptual issues. More importantly, we demystify common misconceptions, addressing fundamental points on how social bots research is discussed. Our analysis surfaces the need to discuss misinformation research in a rigorous, unbiased, and responsible way. This article bolsters such effort by identifying and refuting common fallacious arguments used by both proponents and opponents of social bots research as well as providing indications on the correct methodologies and sound directions for future research in the field.
|
['Marinella Petrocchi', 'Maurizio Tesconi', 'Angelo Spognardi', 'Roberto Di Pietro', 'Stefano Cresci']
|
2023-03-30
| null | null | null | null |
['misconceptions', 'misinformation']
|
['miscellaneous', 'miscellaneous']
|
[ 7.50797018e-02 2.30189309e-01 -4.10068184e-01 1.18105732e-01
3.17901820e-02 -9.37090397e-01 7.82291353e-01 5.29820085e-01
-4.54296917e-01 6.15528584e-01 4.27864939e-01 -1.23090672e+00
-7.26477951e-02 -4.75067586e-01 -4.51123238e-01 -4.38505709e-01
3.93420637e-01 -7.09097972e-03 2.55152285e-01 -3.65681052e-01
1.15587318e+00 2.46947765e-01 -9.96409535e-01 1.40178194e-02
8.78993750e-01 2.08314806e-01 -1.43455327e-01 3.51431787e-01
-3.40734750e-01 1.34687853e+00 -1.20233750e+00 -9.81526673e-01
-1.08606331e-01 -3.34658265e-01 -7.65495718e-01 -1.63666874e-01
3.46809000e-01 -3.41611803e-01 -3.60046253e-02 1.52217162e+00
2.12634355e-01 -4.12037015e-01 2.03569770e-01 -1.31977487e+00
-1.01301181e+00 9.34988379e-01 -6.84490621e-01 6.97130203e-01
1.35319844e-01 6.05046928e-01 6.54155493e-01 -2.85948485e-01
7.13754117e-01 1.50976300e+00 7.39748180e-01 5.13150871e-01
-8.86056781e-01 -8.71862113e-01 1.42396539e-01 8.38037059e-02
-7.72221506e-01 -5.42801857e-01 2.77489036e-01 -9.95133162e-01
4.17957932e-01 3.21232557e-01 6.90422773e-01 1.42759252e+00
4.07381833e-01 5.01706228e-02 1.38858092e+00 -1.77173257e-01
1.40259877e-01 4.82049257e-01 2.53108650e-01 2.73004562e-01
1.30710077e+00 2.38688067e-02 -2.18271852e-01 -6.97657049e-01
5.14682710e-01 -1.23595195e-02 -1.55244768e-01 3.41596752e-01
-9.82675433e-01 1.03150821e+00 3.56538564e-01 9.64284241e-01
-5.02025664e-01 1.24814004e-01 6.11275673e-01 5.81691824e-02
5.95831037e-01 5.83258867e-01 -9.06175897e-02 -5.57554364e-01
-3.67241055e-01 8.33332315e-02 7.49781430e-01 1.49665594e-01
3.30788374e-01 1.07845545e-01 1.17727779e-01 3.18422586e-01
5.28173685e-01 5.79117894e-01 2.54134357e-01 -1.11470509e+00
2.93234766e-01 4.15000200e-01 4.77877289e-01 -1.92561328e+00
2.20201467e-03 -4.58961755e-01 -1.54254809e-01 1.83024555e-01
5.88733912e-01 -3.52410316e-01 -1.93161190e-01 1.41366172e+00
8.47078040e-02 -2.34859914e-01 -3.83857369e-01 7.30384588e-01
5.08949578e-01 2.13417634e-01 6.11260772e-01 -1.70360625e-01
1.23029208e+00 -3.56686682e-01 -9.41517770e-01 -4.45685655e-01
6.04421020e-01 -9.39851761e-01 9.61981356e-01 1.75132994e-02
-8.62872481e-01 3.43882203e-01 -7.18280196e-01 1.33488640e-01
-3.19649607e-01 -4.88546401e-01 3.71184796e-01 1.04303753e+00
-7.44107068e-01 6.87797666e-01 -4.93429869e-01 -6.75327420e-01
5.83465338e-01 -2.56860703e-01 2.60158807e-01 1.47332639e-01
-7.47711301e-01 1.12781954e+00 1.78351745e-01 1.16096079e-01
-3.85468632e-01 -4.40881699e-01 -2.33265415e-01 -3.38125616e-01
8.60873997e-01 -2.31510893e-01 1.14155459e+00 -7.34976828e-01
-8.76045763e-01 8.76147032e-01 1.08861953e-01 -6.72902107e-01
4.69617993e-01 1.15711041e-01 -4.80986595e-01 6.59532472e-02
7.10951507e-01 -1.40185758e-01 8.09969485e-01 -1.28870344e+00
-3.65622371e-01 -5.18280327e-01 7.09299296e-02 -3.84934634e-01
-3.99956763e-01 6.29032850e-01 8.09664190e-01 -4.59818006e-01
1.00440137e-01 -7.28009522e-01 -1.72946841e-01 -3.43291461e-01
-3.74338597e-01 -1.75333783e-01 9.83262002e-01 -7.20130563e-01
1.30345201e+00 -2.00137830e+00 -5.74267328e-01 -6.73182905e-02
1.00438929e+00 6.69130325e-01 1.69538647e-01 9.94877458e-01
2.56521821e-01 1.06914294e+00 3.05018485e-01 2.92933315e-01
-4.56761904e-02 -1.04177527e-01 -4.91603673e-01 7.25829124e-01
-4.78408523e-02 9.49195325e-01 -1.40867734e+00 -1.50629595e-01
1.89798214e-02 2.90038377e-01 -1.03101715e-01 -2.35500678e-01
7.62546510e-02 5.53505838e-01 -6.26043856e-01 6.08926654e-01
6.11192167e-01 -5.52563310e-01 4.42060202e-01 4.38026413e-02
-8.47475052e-01 9.42694902e-01 -2.93468714e-01 3.87723655e-01
-3.30900177e-02 1.02278805e+00 3.47726256e-01 -6.09714210e-01
5.67967832e-01 2.12448332e-02 9.83228013e-02 -3.98071736e-01
7.03251123e-01 5.16260803e-01 2.53274471e-01 -8.00731182e-01
5.75445473e-01 -1.51741460e-01 3.23227257e-01 9.53246295e-01
-5.65277994e-01 -3.42038386e-02 -1.81353256e-01 4.76717353e-01
9.89919126e-01 -3.68005842e-01 2.23094583e-01 -3.12197953e-01
6.12621456e-02 2.03663751e-01 3.35880280e-01 9.36916828e-01
-6.77900434e-01 -3.16996336e-01 7.93033838e-01 -4.57031697e-01
-8.86983097e-01 -4.05982971e-01 1.97922349e-01 8.21946204e-01
1.83753416e-01 -6.24621868e-01 -7.71376550e-01 -9.30803359e-01
1.07695526e-02 1.00203013e+00 -6.30078912e-01 -5.59987463e-02
-2.73421854e-01 -7.46881783e-01 5.95031798e-01 -1.34646311e-01
6.81932986e-01 -7.31393814e-01 -9.19295251e-01 3.13831791e-02
-4.69279259e-01 -1.15486431e+00 8.18613991e-02 -6.19257450e-01
-5.56904614e-01 -1.78958571e+00 -2.62681305e-01 -5.94271719e-02
6.37426019e-01 1.17521155e+00 5.70500016e-01 7.62780011e-01
1.81342915e-01 2.10105658e-01 -3.58845621e-01 -6.34763181e-01
-1.15928924e+00 -2.87552893e-01 -1.73498541e-01 -4.01634008e-01
6.77774608e-01 -4.54226404e-01 -2.44737312e-01 2.92786181e-01
-8.30708921e-01 -3.51589322e-01 4.18214560e-01 2.66626626e-01
-6.06145382e-01 -5.16684949e-02 7.44051695e-01 -9.82680559e-01
1.10504019e+00 -9.79665577e-01 -4.33352888e-01 -9.22523066e-02
-9.46302116e-01 -6.05560660e-01 7.45533928e-02 -3.61185551e-01
-8.44261527e-01 -1.39268541e+00 1.51488081e-01 3.07558328e-01
-1.36186033e-01 4.05904651e-01 6.31692946e-01 -4.72649485e-01
1.30587041e+00 -3.82499546e-01 3.99914831e-01 -3.82701576e-01
3.00236978e-02 8.98575544e-01 -3.75272147e-02 -2.90160865e-01
9.38158154e-01 7.37776518e-01 -5.03821373e-01 -1.04331172e+00
-8.61098945e-01 -2.22728699e-01 1.26118109e-01 -4.37289387e-01
5.28204560e-01 -4.25271541e-01 -1.00160933e+00 4.35486645e-01
-1.41048729e+00 -2.82600880e-01 3.28952703e-03 2.01856881e-01
3.05965766e-02 8.32725465e-01 -6.85291409e-01 -1.10286534e+00
-5.86616509e-02 -9.15874600e-01 2.38089934e-01 1.72018319e-01
-6.63381517e-01 -1.11050987e+00 -4.37566154e-02 9.23894405e-01
1.02255428e+00 4.13236082e-01 5.17308712e-01 -6.49359405e-01
-2.80418754e-01 -1.84231311e-01 -5.43770790e-01 6.37595505e-02
2.75532603e-01 3.43204647e-01 -6.54394865e-01 -7.07406085e-03
4.71513987e-01 -3.64697695e-01 3.87071848e-01 -1.74841750e-02
3.71564806e-01 -1.01770532e+00 -2.43109420e-01 -5.34803629e-01
1.18361986e+00 1.96547329e-01 3.73536348e-01 8.10192287e-01
3.58165562e-01 1.04207659e+00 3.88755649e-01 4.10145998e-01
4.04420644e-01 1.59090132e-01 6.78482652e-01 5.59171379e-01
1.67585805e-01 -4.45375413e-01 2.87268847e-01 3.98846716e-01
6.49738833e-02 -2.08942473e-01 -1.01166105e+00 5.85876822e-01
-1.72444665e+00 -1.17779469e+00 -6.71883047e-01 2.04912758e+00
2.85420895e-01 3.02897453e-01 4.39316988e-01 1.22758426e-01
1.19305706e+00 5.43697953e-01 -1.49546089e-02 -3.84911746e-01
7.92928934e-02 -5.85535765e-01 5.07862329e-01 5.66108644e-01
-5.90301216e-01 1.14203012e+00 6.66752005e+00 5.22898018e-01
-1.17143404e+00 3.28149378e-01 2.82761246e-01 4.90785062e-01
-4.92595017e-01 4.49600399e-01 -3.79587650e-01 5.64144552e-01
8.74216139e-01 -4.25136745e-01 5.38802624e-01 6.38107598e-01
7.57518530e-01 -3.97904962e-01 -5.33561967e-02 2.94832349e-01
4.29529548e-02 -1.53005397e+00 -3.69792402e-01 3.82754505e-01
4.88990068e-01 2.62087017e-01 1.97881311e-01 -2.22368836e-01
5.69315493e-01 -7.48580039e-01 8.58042240e-01 -1.96031556e-01
1.65392160e-01 1.59737691e-01 6.97357059e-01 3.98861468e-01
-2.90226527e-02 -3.29342395e-01 -2.92932540e-01 -8.01014423e-01
1.39966100e-01 6.03150606e-01 -9.86530006e-01 -3.00242528e-02
5.33341825e-01 7.62152314e-01 -5.17330050e-01 7.42222071e-01
-5.04010081e-01 9.90645051e-01 4.17495184e-02 -4.35490966e-01
5.11967957e-01 7.37977354e-03 8.99944901e-01 8.61024737e-01
-1.62442088e-01 5.59884822e-03 -4.86778200e-01 1.08101141e+00
3.14351916e-01 -2.60913353e-02 -9.07771826e-01 -1.10630929e+00
9.25178647e-01 1.02637041e+00 -1.08106446e+00 -9.26382691e-02
-1.63443968e-01 2.07248986e-01 2.10042536e-01 3.25929672e-01
-5.08388460e-01 1.76113740e-01 4.32466745e-01 5.77652216e-01
-1.61538109e-01 -4.84142154e-01 -6.86497509e-01 -1.13036633e+00
-2.93629438e-01 -1.01213181e+00 1.16387539e-01 -3.05461764e-01
-1.15405190e+00 1.85976118e-01 -3.01729348e-02 -6.18819177e-01
8.56238380e-02 -3.86569530e-01 -8.30028832e-01 4.91235256e-01
-1.25434291e+00 -7.95782030e-01 -7.00629456e-03 -2.34753713e-01
3.03744096e-02 8.10307935e-02 2.64101565e-01 -1.62573624e-02
-5.10832667e-01 -3.46853165e-03 -1.36523888e-01 -1.25568286e-01
5.88143945e-01 -4.69510585e-01 5.55115461e-01 5.49399555e-01
-4.27313268e-01 9.76541996e-01 1.19376755e+00 -1.09088731e+00
-1.08165777e+00 -5.73472738e-01 1.27784050e+00 -5.74420869e-01
1.50466013e+00 8.41725022e-02 -7.99469411e-01 7.84707487e-01
2.99124569e-01 -7.73817062e-01 7.07004070e-01 7.64976963e-02
-5.20358205e-01 6.31379247e-01 -1.44407988e+00 1.04660285e+00
8.51922393e-01 -3.42964470e-01 -5.44216692e-01 6.55691683e-01
5.77668428e-01 2.85459012e-02 -1.98693335e-01 -2.85813540e-01
4.13606435e-01 -1.26487935e+00 4.57571268e-01 -7.21550882e-01
3.98994952e-01 1.41331017e-01 4.56235893e-02 -8.41649830e-01
-1.70618728e-01 -8.35879028e-01 4.27534163e-01 1.09274971e+00
1.00988857e-01 -1.17344069e+00 5.31665325e-01 6.87906981e-01
8.20467696e-02 -4.89851773e-01 -5.41430175e-01 -5.99930763e-01
1.13410123e-01 -3.09382528e-01 2.30088338e-01 1.46976781e+00
4.30186242e-01 1.85189366e-01 -2.07494393e-01 3.12718488e-02
6.82095826e-01 -3.63736242e-01 9.02965307e-01 -1.10328412e+00
2.13606521e-01 -5.58840215e-01 -3.19199592e-01 -3.74708861e-01
-3.16320695e-02 -4.41121012e-01 -4.93226409e-01 -1.31950581e+00
1.04299508e-01 -3.05396587e-01 4.65070128e-01 2.19434768e-01
1.02596022e-01 2.99744695e-01 4.75470632e-01 4.09506530e-01
-3.89101595e-01 -3.08524538e-03 1.05315018e+00 5.82753681e-02
-1.32944986e-01 -2.54752487e-01 -1.64928472e+00 1.11964774e+00
1.20647991e+00 -6.90312862e-01 -8.79403651e-02 -4.06044215e-01
8.57972801e-01 -4.11496401e-01 9.16585267e-01 -5.18229723e-01
2.50991940e-01 -7.88953006e-01 -5.15002668e-01 -1.25397459e-01
-7.08833113e-02 -5.42686999e-01 6.28705695e-02 7.53586113e-01
-1.99622557e-01 1.05587825e-01 1.68872148e-01 5.19318342e-01
3.45812410e-01 -6.09335899e-01 6.38513029e-01 -2.66658306e-01
-9.75604653e-02 -3.11565071e-01 -7.46254921e-01 3.54443640e-01
8.63702834e-01 -1.29427567e-01 -1.16049945e+00 -7.24609375e-01
-3.40380490e-01 -1.44516472e-02 8.12179804e-01 3.68456781e-01
2.70185053e-01 -7.50161231e-01 -6.63738728e-01 -2.91792303e-01
-2.47512102e-01 -8.57644498e-01 2.26327956e-01 1.11937022e+00
-5.62914729e-01 2.60857582e-01 -6.86251149e-02 5.28047606e-02
-9.46243286e-01 5.24333894e-01 7.39609376e-02 3.65700275e-01
-5.09618223e-01 3.44528705e-01 -3.69642526e-01 9.59423259e-02
-5.01598716e-02 2.79512137e-01 -2.39692762e-01 -7.23174633e-03
7.14867711e-01 9.89124238e-01 -3.50610763e-01 -8.55927408e-01
-3.02741349e-01 6.10113777e-02 -2.99979419e-01 -2.65484154e-01
9.89431739e-01 -4.73330408e-01 -7.32310176e-01 4.59863365e-01
6.33016407e-01 6.92396760e-01 -4.47845310e-01 2.84758676e-02
2.17307210e-01 -1.13288164e+00 -2.09771529e-01 -9.57711637e-01
-4.08814281e-01 6.71146870e-01 -2.64475256e-01 1.04834282e+00
1.24942474e-01 -3.91791947e-02 6.54588223e-01 1.10816620e-01
4.36646104e-01 -9.63714659e-01 2.52112865e-01 2.85917968e-01
8.32610428e-01 -1.10119855e+00 9.22517292e-03 -7.22834289e-01
-4.90929008e-01 9.47006106e-01 3.29211861e-01 -1.92545261e-02
4.80509371e-01 -2.16787636e-01 2.75401950e-01 -6.56084776e-01
-4.95937675e-01 2.43280709e-01 -5.89417636e-01 7.04145610e-01
5.40642083e-01 4.51043583e-02 -1.52551401e+00 4.27646279e-01
-9.63317305e-02 -2.11582500e-02 1.09411073e+00 1.02603483e+00
-7.79299498e-01 -8.79387259e-01 -9.19671059e-01 2.99561322e-01
-9.62209225e-01 9.21258479e-02 -1.12798655e+00 7.67957866e-01
5.73800458e-03 1.60669005e+00 -4.16482627e-01 -6.21612251e-01
-1.70429662e-01 -3.57788444e-01 -1.68145560e-02 -4.85244870e-01
-8.18475366e-01 -2.11291179e-01 4.38629776e-01 -2.45223403e-01
-2.68414557e-01 -5.48903763e-01 -6.94517493e-01 -1.13935542e+00
-6.25327289e-01 5.14762819e-01 9.73665059e-01 1.01520991e+00
6.19644046e-01 -4.35321592e-02 3.76286209e-01 -3.69913548e-01
-9.64184999e-01 -9.04768586e-01 -1.41861722e-01 6.81445599e-02
3.13115954e-01 -5.00401616e-01 -8.41804743e-01 -6.69527769e-01]
|
[8.455018997192383, 9.975789070129395]
|
69adad40-269d-4a78-8e38-717c517230a4
|
contranet-a-single-end-to-end-hybrid-network
|
2206.10677
| null |
https://arxiv.org/abs/2206.10677v1
|
https://arxiv.org/pdf/2206.10677v1.pdf
|
ConTraNet: A single end-to-end hybrid network for EEG-based and EMG-based human machine interfaces
|
Objective: Electroencephalography (EEG) and electromyography (EMG) are two non-invasive bio-signals, which are widely used in human machine interface (HMI) technologies (EEG-HMI and EMG-HMI paradigm) for the rehabilitation of physically disabled people. Successful decoding of EEG and EMG signals into respective control command is a pivotal step in the rehabilitation process. Recently, several Convolutional neural networks (CNNs) based architectures are proposed that directly map the raw time-series signal into decision space and the process of meaningful features extraction and classification are performed simultaneously. However, these networks are tailored to the learn the expected characteristics of the given bio-signal and are limited to single paradigm. In this work, we addressed the question that can we build a single architecture which is able to learn distinct features from different HMI paradigms and still successfully classify them. Approach: In this work, we introduce a single hybrid model called ConTraNet, which is based on CNN and Transformer architectures that is equally useful for EEG-HMI and EMG-HMI paradigms. ConTraNet uses CNN block to introduce inductive bias in the model and learn local dependencies, whereas the Transformer block uses the self-attention mechanism to learn the long-range dependencies in the signal, which are crucial for the classification of EEG and EMG signals. Main results: We evaluated and compared the ConTraNet with state-of-the-art methods on three publicly available datasets which belong to EEG-HMI and EMG-HMI paradigms. ConTraNet outperformed its counterparts in all the different category tasks (2-class, 3-class, 4-class, and 10-class decoding tasks). Significance: The results suggest that ConTraNet is robust to learn distinct features from different HMI paradigms and generalizes well as compared to the current state of the art algorithms.
|
['Christian Klaes', 'Ioannis Iossifidis', 'Tobias Glasmachers', 'Muhammad Saif-ur-Rehman', 'Omair Ali']
|
2022-06-21
| null | null | null | null |
['electromyography-emg']
|
['medical']
|
[ 4.59633768e-01 -7.26591125e-02 1.42699987e-01 -2.36169219e-01
-4.23581630e-01 -4.44105305e-02 3.64903688e-01 -4.48368907e-01
-5.67438126e-01 1.07501471e+00 1.85003489e-01 -7.36520961e-02
-7.70516455e-01 -4.32429463e-01 -7.75454760e-01 -6.63354993e-01
-5.42453170e-01 3.40076208e-01 -7.84434006e-02 -4.24008876e-01
1.67051896e-01 4.21183497e-01 -1.83957660e+00 6.99972808e-01
8.84877443e-01 1.29685879e+00 3.68744880e-01 5.67397118e-01
3.95399034e-01 4.65578616e-01 -8.38458478e-01 4.41785604e-01
-1.39294043e-01 -4.90246117e-01 -8.84788811e-01 -5.52918077e-01
-6.17160238e-02 1.71693951e-01 -3.60912800e-01 7.06892848e-01
9.39933240e-01 -1.78772267e-02 9.13053632e-01 -1.25349355e+00
-4.43431169e-01 4.35400218e-01 -1.29527897e-02 4.72793579e-01
5.46960115e-01 2.38091528e-01 4.51797813e-01 -6.02048278e-01
4.27474976e-01 6.16021276e-01 7.80656934e-01 8.12191248e-01
-1.08483016e+00 -9.78328526e-01 -1.29673854e-02 8.56891811e-01
-1.37353814e+00 -9.74196345e-02 7.66137779e-01 -4.76990402e-01
1.40419304e+00 4.97990072e-01 1.11596143e+00 1.80545223e+00
6.94829762e-01 8.68937850e-01 1.48359847e+00 -1.66834056e-01
1.93479940e-01 -6.92795217e-02 3.72705668e-01 -1.45320907e-01
-2.57128358e-01 4.53812093e-01 -8.96391034e-01 2.45927215e-01
5.80417514e-01 -1.14816286e-01 -9.36417580e-01 4.01658118e-01
-1.26667023e+00 2.87239522e-01 4.36511129e-01 8.92792046e-01
-8.80829155e-01 3.88921425e-02 4.42057759e-01 6.41663909e-01
3.36018026e-01 5.38702071e-01 -5.59307277e-01 -6.39314771e-01
-8.88005555e-01 2.03400478e-01 8.45965683e-01 8.85422349e-01
2.98646331e-01 1.01007633e-01 -4.61730659e-01 6.73800766e-01
-2.39959925e-01 3.17982644e-01 1.00883865e+00 -3.35196316e-01
3.85551155e-01 5.01418412e-01 -3.12608063e-01 -6.27578318e-01
-9.30097759e-01 -6.65919185e-01 -1.08686066e+00 2.78723210e-01
7.79218897e-02 -2.54017740e-01 -9.50994313e-01 1.68182647e+00
-4.42784280e-01 4.18670982e-01 -1.11456245e-01 9.93332148e-01
9.21725690e-01 3.64061534e-01 -7.13468269e-02 5.50160483e-02
1.15247250e+00 -4.25326347e-01 -7.31412113e-01 -4.02518243e-01
2.90092200e-01 -1.05978809e-01 9.93410468e-01 8.27097654e-01
-9.38656092e-01 -7.41401196e-01 -1.19270182e+00 4.14067924e-01
-5.74587166e-01 2.59798348e-01 3.67226839e-01 2.66494185e-01
-1.02444601e+00 1.01168561e+00 -7.32891738e-01 -3.64802063e-01
2.26944417e-01 1.14216149e+00 -7.18091905e-01 4.71210748e-01
-1.60995924e+00 1.27593696e+00 5.86961687e-01 4.87180114e-01
-9.26430762e-01 -5.16810715e-01 -3.94369513e-01 8.96309465e-02
-1.14214957e-01 -5.25443077e-01 6.60025179e-01 -1.09625554e+00
-1.78698635e+00 7.18809724e-01 2.03840777e-01 -3.91452610e-01
2.50576645e-01 -3.76510203e-01 -3.61107528e-01 -1.73383072e-01
-2.79661894e-01 4.49541301e-01 8.93062890e-01 -7.80753732e-01
-5.21739602e-01 -5.67436397e-01 -2.34271824e-01 2.74621379e-02
-4.45282519e-01 -2.58697495e-02 8.74586478e-02 -6.31115615e-01
-1.74097076e-01 -9.42134321e-01 5.06235301e-01 -7.06150174e-01
-4.18614894e-01 -3.35795432e-01 5.72200716e-01 -8.19153130e-01
1.18852663e+00 -2.12367415e+00 7.61527479e-01 2.77017593e-01
1.87649041e-01 2.32802615e-01 -1.77391335e-01 4.21496034e-01
-6.23142123e-01 -2.09899664e-01 -1.84014872e-01 1.20833926e-01
-7.18246624e-02 3.62930089e-01 1.89878345e-01 4.25571680e-01
1.98668122e-01 1.00081766e+00 -5.06917536e-01 1.76338434e-01
2.26343349e-01 5.47510386e-01 -2.60955781e-01 6.22697771e-01
3.81929278e-01 1.04000306e+00 3.48247662e-02 3.30301791e-01
3.24582189e-01 2.69550920e-01 -9.74908322e-02 -5.65045595e-01
1.03540318e-02 3.45094651e-01 -9.27312493e-01 1.81587672e+00
-6.32789016e-01 8.45906734e-01 -6.57032877e-02 -1.66801441e+00
7.20342159e-01 7.37667978e-01 5.68127930e-01 -7.10975111e-01
7.31214583e-01 3.92827481e-01 4.27186280e-01 -1.02355278e+00
-4.02604610e-01 -1.69561893e-01 1.99061334e-01 3.96034978e-02
6.46787107e-01 2.21784800e-01 -7.33318105e-02 -6.91221833e-01
1.13731718e+00 3.28147471e-01 1.12522922e-01 -5.31893790e-01
5.95633447e-01 -4.23234165e-01 5.23118138e-01 6.81886911e-01
1.77903816e-01 6.37159646e-01 2.19207868e-01 -3.27667445e-01
-4.47934121e-01 -7.78875589e-01 -2.14293599e-01 7.23301172e-01
-4.95442934e-02 -9.49803144e-02 -7.97825754e-01 -2.65171379e-01
-1.11660495e-01 6.43588841e-01 -9.55581903e-01 -7.29572952e-01
-6.47806406e-01 -8.80028486e-01 4.37529713e-01 9.63669538e-01
3.79998565e-01 -1.61202502e+00 -8.82591844e-01 4.48429704e-01
-3.86005580e-01 -8.86749983e-01 2.73573976e-02 1.04552507e+00
-8.21739674e-01 -1.07499731e+00 -9.02898550e-01 -8.07844102e-01
1.56163245e-01 -5.58105946e-01 7.84908593e-01 -2.62825936e-01
-3.31165940e-01 3.31145376e-01 -5.31331897e-01 -6.84537590e-01
7.38145504e-03 3.68106127e-01 2.22189873e-01 6.31605163e-02
5.73720694e-01 -1.30204844e+00 -6.55169666e-01 1.68274865e-01
-7.28755951e-01 1.08513713e-01 7.97489405e-01 1.14684963e+00
1.36978328e-01 -1.96079031e-01 8.25367153e-01 -3.43863696e-01
9.57304120e-01 -6.01594150e-01 3.11955184e-01 1.97935253e-01
-4.46031123e-01 -1.90093786e-01 6.97688460e-01 -8.26877117e-01
-5.14867544e-01 -3.08088303e-01 -4.80376303e-01 -4.43137974e-01
-4.09905583e-01 5.87241054e-01 -1.50980949e-01 -1.89578503e-01
7.35099673e-01 5.84146976e-01 -1.30086660e-01 -5.36402643e-01
-2.55039126e-01 1.17762256e+00 8.07962000e-01 -5.37519097e-01
1.52731374e-01 5.18839993e-03 -2.37339675e-01 -8.03874314e-01
-2.35946834e-01 -3.87229145e-01 -7.51241744e-01 -5.41284263e-01
1.04437768e+00 -5.61367691e-01 -8.03869724e-01 8.74115646e-01
-1.30893171e+00 -4.87471193e-01 -1.29342005e-01 9.32187796e-01
-9.23367083e-01 -1.19279228e-01 -4.91820306e-01 -7.76802897e-01
-7.37896621e-01 -1.18959010e+00 1.00133038e+00 8.97878781e-02
-4.89227444e-01 -6.20017827e-01 -1.19694173e-01 -4.09218259e-02
6.36956096e-01 4.19053912e-01 1.02215016e+00 -8.16989183e-01
2.58619696e-01 -3.15987229e-01 1.02060907e-01 7.21434176e-01
4.13570292e-02 -6.67303324e-01 -1.22064221e+00 -3.74909192e-01
4.41791683e-01 -2.01429516e-01 4.92611796e-01 4.40561086e-01
1.35925972e+00 -3.49640697e-02 -2.68200904e-01 7.40709901e-01
1.23498607e+00 6.85566604e-01 1.11015379e+00 5.07296503e-01
3.21891814e-01 5.76840341e-01 2.09637564e-02 1.30811006e-01
1.92297827e-02 7.12669969e-01 3.50920320e-01 -1.00295268e-01
1.08646555e-02 2.80654997e-01 3.44858468e-01 1.14988184e+00
-8.12846124e-01 -1.11428186e-01 -8.51555347e-01 4.45726782e-01
-1.86519027e+00 -6.50533199e-01 1.17136957e-03 2.14034295e+00
6.85834825e-01 1.55854881e-01 4.91155265e-03 7.32596993e-01
4.02475148e-01 -5.30481458e-01 -6.19577646e-01 -2.90800631e-01
-5.90814054e-02 1.08928120e+00 1.80581808e-01 -1.01170741e-01
-8.10272157e-01 3.19800109e-01 5.91194677e+00 7.04050422e-01
-1.50716233e+00 3.55310887e-01 3.17746848e-02 -8.38529393e-02
3.89215231e-01 -5.69318354e-01 -3.68040115e-01 7.03459740e-01
1.37633491e+00 9.52724814e-02 7.31997013e-01 3.82394642e-01
1.57548040e-01 -2.94732228e-02 -1.57271576e+00 1.38902271e+00
-7.20457081e-03 -1.07567990e+00 -3.28854442e-01 -7.48218372e-02
2.40124837e-01 2.35990182e-01 -2.39561111e-01 5.64367712e-01
-6.94615722e-01 -1.32595730e+00 7.54329443e-01 1.00361216e+00
9.54548895e-01 -5.46661556e-01 1.26574361e+00 3.98655772e-01
-1.07489705e+00 -4.64153707e-01 5.57760559e-02 -1.78719580e-01
-5.15047386e-02 4.75708768e-02 -1.19221680e-01 9.06684399e-01
1.20676601e+00 1.03410065e+00 -3.69363308e-01 9.03655708e-01
-1.86944723e-01 7.85431862e-01 -3.06911498e-01 -2.00517997e-01
1.02949142e-01 7.25599146e-03 6.22405648e-01 1.24862015e+00
5.13849080e-01 1.47265950e-02 -2.21098244e-01 9.31409061e-01
3.43102753e-01 -1.47155687e-01 -4.95429128e-01 1.30268484e-01
8.47324729e-02 8.04553211e-01 -2.12407559e-01 -1.74287155e-01
-2.75466710e-01 1.19808829e+00 1.72280625e-01 5.37464082e-01
-8.25750768e-01 -8.04922998e-01 4.60105985e-01 6.37919381e-02
-1.02910228e-01 4.64257076e-02 -3.70144129e-01 -1.10801840e+00
2.75199473e-01 -1.05540836e+00 8.27407688e-02 -1.06477940e+00
-1.35613251e+00 1.05612504e+00 2.84156114e-01 -1.38580751e+00
-4.45208311e-01 -1.04129410e+00 -7.38737702e-01 1.02998459e+00
-1.35729873e+00 -8.76471519e-01 -7.02898085e-01 1.02259481e+00
3.59781563e-01 -1.24674194e-01 1.06304634e+00 6.02417350e-01
-4.88356799e-01 4.09247726e-01 1.58895612e-01 -2.30584722e-02
5.12543440e-01 -1.09697807e+00 -1.34074703e-01 3.68697584e-01
-1.97906777e-01 6.96304083e-01 4.21139359e-01 -2.23968133e-01
-1.38619637e+00 -6.64657176e-01 5.34177065e-01 -4.80296537e-02
4.69356924e-01 -4.66433883e-01 -9.10331905e-01 4.97074068e-01
4.22252566e-01 -3.58419091e-01 6.50898159e-01 -3.56228203e-02
2.91069478e-01 -3.25763226e-01 -8.37709785e-01 2.56531686e-01
1.27502632e+00 -5.32738805e-01 -1.00389194e+00 1.65652394e-01
-3.49773690e-02 -2.98100412e-01 -1.22478795e+00 7.55320489e-01
9.17777658e-01 -8.78573954e-01 6.68623269e-01 -7.17159033e-01
1.75108969e-01 3.02279387e-02 -9.51747820e-02 -1.73948741e+00
-3.28655154e-01 -5.72258472e-01 -1.20977871e-01 6.17029428e-01
2.75492311e-01 -9.16925251e-01 2.01126546e-01 3.18804771e-01
-4.56132293e-01 -1.16411591e+00 -1.23125792e+00 -1.06365752e+00
5.91772646e-02 -7.60821342e-01 4.78426397e-01 4.46393311e-01
5.34241557e-01 2.66025811e-01 -5.00010431e-01 -2.46683329e-01
8.58367980e-02 -2.80513197e-01 3.49475890e-01 -1.47369576e+00
-2.01641262e-01 -5.44951439e-01 -1.02014637e+00 -7.05014467e-01
2.38586828e-01 -1.06462598e+00 2.05013484e-01 -1.60846627e+00
-1.69169791e-02 6.82049841e-02 -8.30086589e-01 7.05188453e-01
6.87624812e-02 1.13163345e-01 4.15165909e-03 1.48373142e-01
1.93968952e-01 6.02421463e-01 8.13632727e-01 -3.38787228e-01
-2.64114738e-01 1.03968024e-01 -2.74736911e-01 3.47514391e-01
8.39868665e-01 -4.39011633e-01 -3.87046665e-01 -2.51216114e-01
1.57426388e-04 4.14759889e-02 4.99337643e-01 -1.74817777e+00
2.46169373e-01 3.73116791e-01 5.46815932e-01 -4.40245897e-01
3.28627020e-01 -8.28578591e-01 2.84798741e-01 4.93665725e-01
-2.63088137e-01 -4.70408536e-02 3.09492350e-01 3.39580685e-01
-3.63112509e-01 -5.37809134e-02 5.88726580e-01 1.44012915e-02
-5.30140281e-01 1.54700026e-01 -8.09031010e-01 -1.81697607e-02
9.82304871e-01 -6.07384920e-01 3.70559692e-02 -2.72318959e-01
-1.17192185e+00 -3.25074941e-02 -4.54511672e-01 7.09207475e-01
7.20733345e-01 -1.27379000e+00 -7.81965911e-01 5.93845248e-01
1.53746173e-01 -3.64197671e-01 2.00216815e-01 1.49448991e+00
-2.29797643e-02 5.78931808e-01 -8.82090509e-01 -7.62826204e-01
-1.03579152e+00 2.37002239e-01 6.48749113e-01 4.64310013e-02
-9.42157626e-01 6.45313621e-01 2.15484295e-02 -3.26128215e-01
6.09948456e-01 -6.40235007e-01 -6.28735065e-01 -9.08056796e-02
4.11705345e-01 2.82697558e-01 5.81756055e-01 -4.74558771e-01
-5.00724912e-01 5.28254628e-01 2.95911610e-01 1.13194482e-02
1.76888824e+00 2.62503773e-01 -9.97610316e-02 6.99693501e-01
1.26896250e+00 -9.31339800e-01 -8.32088590e-01 2.96867669e-01
5.96939214e-02 1.92514151e-01 9.27234590e-02 -1.18758297e+00
-1.11993206e+00 1.15300286e+00 1.23612118e+00 -6.04080558e-02
1.55768979e+00 -4.17967349e-01 8.34562719e-01 2.10609764e-01
6.54029787e-01 -1.15204751e+00 -2.32020617e-01 3.44118655e-01
1.39486778e+00 -6.39370978e-01 -4.15390164e-01 1.25545263e-01
-3.97772878e-01 1.43089151e+00 6.55588865e-01 -2.89113402e-01
8.03637683e-01 3.81868124e-01 -2.81800061e-01 -2.95781106e-01
-4.67667758e-01 -1.73856229e-01 7.78525591e-01 8.49124670e-01
4.95198905e-01 5.81308976e-02 -6.56857908e-01 1.30989754e+00
-2.04142570e-01 6.05811536e-01 -8.59411620e-03 1.01442194e+00
-2.66513154e-02 -7.11039007e-01 -2.28771389e-01 7.96214879e-01
-2.64537841e-01 -6.77861199e-02 -2.60260284e-01 1.04537153e+00
4.52634543e-01 9.43243325e-01 -2.37791270e-01 -1.16426778e+00
7.79214621e-01 4.32329178e-01 7.37140894e-01 -4.01503175e-01
-1.21688080e+00 -2.06715629e-01 4.40992415e-02 -7.68045127e-01
-5.08832872e-01 -2.46211499e-01 -1.12802291e+00 2.72038996e-01
-3.68980050e-01 3.50557938e-02 7.40224421e-01 1.25386858e+00
4.10467625e-01 1.14908195e+00 3.85595918e-01 -1.33824372e+00
-3.78225654e-01 -1.64575493e+00 -6.82888150e-01 4.80090857e-01
1.78382471e-01 -9.14785147e-01 -4.73290026e-01 -1.76127777e-01]
|
[13.05759334564209, 3.4565787315368652]
|
4997e621-b132-4a51-9cf6-ecfe7bff6f93
|
t2ranking-a-large-scale-chinese-benchmark-for
|
2304.03679
| null |
https://arxiv.org/abs/2304.03679v1
|
https://arxiv.org/pdf/2304.03679v1.pdf
|
T2Ranking: A large-scale Chinese Benchmark for Passage Ranking
|
Passage ranking involves two stages: passage retrieval and passage re-ranking, which are important and challenging topics for both academics and industries in the area of Information Retrieval (IR). However, the commonly-used datasets for passage ranking usually focus on the English language. For non-English scenarios, such as Chinese, the existing datasets are limited in terms of data scale, fine-grained relevance annotation and false negative issues. To address this problem, we introduce T2Ranking, a large-scale Chinese benchmark for passage ranking. T2Ranking comprises more than 300K queries and over 2M unique passages from real-world search engines. Expert annotators are recruited to provide 4-level graded relevance scores (fine-grained) for query-passage pairs instead of binary relevance judgments (coarse-grained). To ease the false negative issues, more passages with higher diversities are considered when performing relevance annotations, especially in the test set, to ensure a more accurate evaluation. Apart from the textual query and passage data, other auxiliary resources are also provided, such as query types and XML files of documents which passages are generated from, to facilitate further studies. To evaluate the dataset, commonly used ranking models are implemented and tested on T2Ranking as baselines. The experimental results show that T2Ranking is challenging and there is still scope for improvement. The full data and all codes are available at https://github.com/THUIR/T2Ranking/
|
['Jin Ma', 'Yiqun Liu', 'Haitao Li', 'Xiangsheng Li', 'Zhijing Wu', 'Weinan Gan', 'Ting Yao', 'Feiyang Lv', 'Bingning Wang', 'Qian Dong', 'Xiaohui Xie']
|
2023-04-07
| null | null | null | null |
['passage-ranking', 'passage-re-ranking', 'passage-retrieval']
|
['natural-language-processing', 'natural-language-processing', 'natural-language-processing']
|
[-2.08798632e-01 -6.37131512e-01 -3.54348332e-01 -6.87999725e-02
-1.50782847e+00 -8.75572264e-01 6.11218452e-01 5.49403548e-01
-7.67015815e-01 9.04300570e-01 6.90798163e-01 -1.46653101e-01
-2.99718618e-01 -6.31524682e-01 -3.00125211e-01 -2.27692142e-01
1.27387214e-02 4.11345929e-01 6.59061491e-01 -6.64499938e-01
7.73184896e-01 -1.89739764e-01 -1.50024247e+00 5.90750098e-01
1.17860079e+00 6.85091376e-01 3.10140878e-01 6.51113749e-01
-2.94734389e-01 4.98958021e-01 -8.59646261e-01 -3.83744419e-01
1.50274439e-02 -4.19318557e-01 -1.22655523e+00 -7.53586590e-01
2.26808831e-01 -3.22702169e-01 -3.38472366e-01 9.53909397e-01
7.69317091e-01 5.13477445e-01 5.85349321e-01 -8.74349177e-01
-7.60609031e-01 5.37350535e-01 -1.92541912e-01 5.91185153e-01
9.54863727e-01 -2.87992477e-01 1.45713687e+00 -9.99798715e-01
4.75920945e-01 9.15751517e-01 2.42798254e-01 2.79811352e-01
-5.05635679e-01 -4.89964247e-01 1.15856491e-02 4.71834660e-01
-1.55707300e+00 -1.36217445e-01 4.16032851e-01 -9.08425078e-02
7.53217399e-01 9.99666572e-01 4.05580252e-01 9.64236796e-01
-2.59347767e-01 9.07842875e-01 7.11043239e-01 -5.72829425e-01
-1.77225936e-02 1.24869853e-01 3.70366007e-01 -1.08365737e-01
1.42376870e-01 -3.09957027e-01 -3.65430415e-01 -4.15437102e-01
5.41653454e-01 1.19124793e-01 -4.95059043e-01 3.75718296e-01
-1.34365320e+00 5.90485752e-01 4.71474588e-01 6.44408226e-01
-2.15474918e-01 -5.18525362e-01 7.76219070e-01 5.55392206e-01
3.38929564e-01 7.45314598e-01 -6.17207944e-01 -3.68899882e-01
-8.12281609e-01 4.95511532e-01 6.89448416e-01 1.12636030e+00
5.73111057e-01 -8.90864134e-01 -1.14373553e+00 1.24911857e+00
2.60324508e-01 3.73405427e-01 8.90038788e-01 -5.74668765e-01
9.35540080e-01 7.74869502e-01 4.79245603e-01 -1.06772184e+00
-1.28741115e-01 -3.99034947e-01 -8.14171314e-01 -8.36188614e-01
-9.54973605e-03 2.13419974e-01 -5.59454024e-01 1.19590712e+00
1.20624989e-01 -1.91618457e-01 4.65500243e-02 1.17365670e+00
1.32769084e+00 8.40518951e-01 7.51085207e-02 -2.25279123e-01
1.57036459e+00 -1.25234723e+00 -4.56438512e-01 9.72022787e-02
7.88895845e-01 -1.29981792e+00 1.81501639e+00 -1.54633820e-01
-9.08356190e-01 -5.14213383e-01 -7.17441499e-01 -3.53510290e-01
-3.37031245e-01 1.72271281e-01 1.44244924e-01 1.75599247e-01
-9.64435339e-01 2.30342567e-01 -2.86716014e-01 -4.84602273e-01
-2.48497412e-01 -1.46949112e-01 -8.72688293e-02 -2.66046673e-01
-2.09923124e+00 6.14293098e-01 4.12594825e-01 -4.53277677e-03
-4.42131698e-01 -6.24828577e-01 -4.30305272e-01 1.08802393e-01
1.69461608e-01 -5.61080217e-01 1.46629214e+00 -2.74214387e-01
-1.02202439e+00 7.54204750e-01 -1.39841512e-01 1.33826941e-01
6.24265432e-01 -4.40468878e-01 -6.81406617e-01 2.65881568e-01
5.15337229e-01 1.99431390e-01 1.37214705e-01 -8.28456879e-01
-9.19301152e-01 8.63076448e-02 4.54092979e-01 7.07056403e-01
-6.33088648e-01 3.16657901e-01 -1.33368564e+00 -7.89975345e-01
-1.48279041e-01 -6.98738754e-01 -1.56293914e-01 -5.58142960e-01
-4.51745570e-01 -6.49090230e-01 3.70180398e-01 -6.88111782e-01
1.92062914e+00 -1.88376915e+00 -4.60900962e-01 1.36515722e-01
-4.25239466e-02 3.05046290e-01 -4.45921600e-01 9.84113693e-01
4.00014341e-01 5.47947109e-01 2.33776197e-01 4.24697772e-02
-5.01627009e-03 -2.24658147e-01 -2.83839792e-01 -2.55602479e-01
-2.02256992e-01 9.87000525e-01 -1.25593174e+00 -8.21336329e-01
-2.16453880e-01 9.64173302e-02 -2.74764329e-01 3.62867296e-01
-3.21856998e-02 5.47934353e-01 -1.08465958e+00 5.91020763e-01
2.17959464e-01 -3.86351347e-01 -3.05216759e-01 -9.37960073e-02
-3.76069956e-02 7.77770221e-01 -8.77113163e-01 1.69159973e+00
-4.44776356e-01 3.33168864e-01 -6.19640946e-01 -3.85873288e-01
6.30630672e-01 5.49123943e-01 4.10738826e-01 -1.26393223e+00
-1.44999042e-01 4.23955083e-01 -4.42307919e-01 -7.56166756e-01
1.20970488e+00 4.09570396e-01 -4.68600512e-01 3.48027736e-01
-5.49174190e-01 -1.07492432e-02 9.24552143e-01 5.36467135e-01
1.12984765e+00 -1.46942094e-01 1.29985809e-01 -1.70900628e-01
7.82213688e-01 4.68138605e-01 4.74152803e-01 1.04331005e+00
3.24405357e-02 8.91531348e-01 -8.65123933e-04 -7.19352663e-02
-9.76779521e-01 -7.02038527e-01 -1.82991207e-01 1.36378324e+00
5.83611131e-01 -7.65418828e-01 -4.49078619e-01 -6.21575236e-01
-3.36806715e-01 4.12769437e-01 -3.45343709e-01 -1.17926732e-01
-4.64876562e-01 -6.83373928e-01 6.47776723e-01 3.19649041e-01
5.25515616e-01 -1.31572890e+00 -1.30312800e-01 2.54075944e-01
-1.08061671e+00 -8.80998194e-01 -1.02129090e+00 -5.64600348e-01
-6.92331493e-01 -1.18248904e+00 -1.16916764e+00 -9.01668191e-01
5.49069047e-01 6.83058500e-01 1.49416959e+00 6.74060822e-01
-8.30209181e-02 4.34541702e-01 -1.22838163e+00 -9.55494121e-02
-1.61491469e-01 5.18443167e-01 -1.92836598e-01 -7.09903598e-01
4.22211677e-01 -5.53124696e-02 -1.07467091e+00 7.79277742e-01
-1.27234960e+00 -2.74809599e-01 5.53770006e-01 8.28070819e-01
5.90483725e-01 8.67933854e-02 7.12957025e-01 -6.66460633e-01
1.39967692e+00 -5.42563915e-01 -3.91321957e-01 7.84819543e-01
-6.08811140e-01 -2.28500023e-01 6.70753539e-01 -3.61921638e-01
-8.78699303e-01 -9.91219938e-01 -3.04528505e-01 2.21283630e-01
1.30171463e-01 1.06657040e+00 1.77995637e-01 2.66579211e-01
7.09971011e-01 2.94766635e-01 -6.72932386e-01 -7.26922512e-01
1.13228217e-01 9.87339973e-01 1.73675537e-01 -7.68634856e-01
6.34179235e-01 -2.55683064e-01 -6.21647298e-01 -4.73483920e-01
-8.62200975e-01 -1.15637136e+00 -4.04738337e-01 -2.17663065e-01
3.31405550e-01 -9.91690755e-01 -3.16506386e-01 2.06549183e-01
-8.93031716e-01 1.15511790e-02 -1.33198006e-02 6.87287629e-01
-1.20372362e-02 6.29262149e-01 -9.57544088e-01 -3.66594762e-01
-8.33607316e-01 -9.34999228e-01 9.93258119e-01 5.58067620e-01
-1.32850930e-01 -8.07352304e-01 3.04024875e-01 6.30207658e-01
5.50239325e-01 -4.45106447e-01 9.55899835e-01 -9.34252143e-01
-3.26101214e-01 -6.44621491e-01 -1.19256310e-01 1.70354564e-02
1.93663076e-01 7.65905306e-02 -5.38234353e-01 -4.50604349e-01
-3.96827489e-01 -5.56333065e-01 7.11450040e-01 -1.56233907e-01
1.20322239e+00 -5.17027617e-01 -2.45122820e-01 -2.02967376e-01
1.13042271e+00 1.56262428e-01 7.79448390e-01 8.74173045e-01
3.74028563e-01 4.51376319e-01 1.22683990e+00 4.20549691e-01
6.87582135e-01 7.17832983e-01 -6.71770945e-02 1.53930262e-02
5.98519333e-02 -4.12607968e-01 9.95695740e-02 1.47978175e+00
-1.08435266e-01 -4.87260461e-01 -8.13360572e-01 6.30713582e-01
-1.68598485e+00 -1.05415738e+00 -1.06805615e-01 2.67069268e+00
1.19602001e+00 9.37401950e-02 8.21839347e-02 1.63573742e-01
6.94327474e-01 -5.08118980e-02 -2.47619212e-01 9.61248055e-02
-6.05963245e-02 -1.79901019e-01 -8.06912035e-02 3.95321876e-01
-9.69478846e-01 7.35940099e-01 5.65096569e+00 1.23173070e+00
-8.71336102e-01 -1.53107688e-01 5.47530591e-01 7.00685829e-02
-6.01405263e-01 1.93250664e-02 -9.36576307e-01 7.04266548e-01
5.05146623e-01 -5.75031638e-01 1.41651928e-01 5.43783247e-01
2.08281070e-01 3.92163033e-03 -6.78528011e-01 7.91316807e-01
-1.11778967e-01 -9.96214807e-01 1.89602017e-01 -3.91537338e-01
7.59414852e-01 2.40271479e-01 -1.09066635e-01 8.59285653e-01
1.47633865e-01 -5.71172357e-01 3.50073129e-01 3.63202780e-01
7.99674749e-01 -4.80031788e-01 9.61110830e-01 4.41456914e-01
-1.40243399e+00 1.49822056e-01 -5.65234840e-01 1.09056272e-01
1.05884716e-01 6.23662114e-01 -5.35716474e-01 9.64207888e-01
1.11541557e+00 5.74598968e-01 -8.60920250e-01 1.61142039e+00
-2.90905237e-01 6.10520780e-01 -3.17385763e-01 -4.50965196e-01
1.65016249e-01 -2.96625905e-02 4.73467112e-01 1.26642728e+00
5.23879588e-01 1.77651048e-01 2.90575296e-01 1.39993131e-01
-3.11633110e-01 7.12697268e-01 -4.29629534e-02 1.34296745e-01
9.40136075e-01 1.34051692e+00 -5.92180669e-01 -4.79628474e-01
-4.42708075e-01 8.21095109e-01 2.44995922e-01 5.09238660e-01
-5.27010679e-01 -7.07105339e-01 2.01496542e-01 -6.86802566e-02
-1.86278045e-01 8.31092894e-02 3.22114080e-01 -1.44503593e+00
4.49542910e-01 -8.47230077e-01 8.72651517e-01 -6.06737196e-01
-1.60255337e+00 7.25811362e-01 1.68413743e-01 -1.97613227e+00
-3.17182630e-01 -4.95743081e-02 -6.08844638e-01 9.03119445e-01
-1.81354594e+00 -6.89653754e-01 -4.35339242e-01 5.36439657e-01
7.68372953e-01 1.82538450e-01 7.97259748e-01 8.42075527e-01
-4.41726834e-01 9.64102566e-01 4.91264105e-01 3.42101455e-01
1.08238542e+00 -9.84543562e-01 3.35763574e-01 6.85511589e-01
2.16180891e-01 1.02283299e+00 4.71406221e-01 -6.17563665e-01
-9.65188205e-01 -9.74570572e-01 1.26752794e+00 -5.06502211e-01
4.66408581e-01 1.30505368e-01 -1.05237532e+00 7.96392560e-02
1.34349823e-01 -1.00256816e-01 9.08031762e-01 2.77124643e-01
-8.00906569e-02 -3.04828994e-02 -7.62736142e-01 7.43120730e-01
8.73599112e-01 -6.49846971e-01 -6.56573951e-01 4.33594435e-01
8.55459273e-01 -4.85389352e-01 -1.09775722e+00 5.73763609e-01
5.78260541e-01 -5.61663687e-01 1.05130565e+00 -4.63911504e-01
3.39154869e-01 -5.50371528e-01 -4.41156402e-02 -1.14439356e+00
-3.70325834e-01 -1.77868590e-01 3.02796494e-02 1.50877047e+00
6.85835302e-01 -4.38920259e-01 4.35627967e-01 5.59110820e-01
-6.33563399e-02 -8.52607906e-01 -6.75739408e-01 -6.66704893e-01
1.17717378e-01 2.69065611e-02 8.13589871e-01 9.20019805e-01
1.62921414e-01 4.99483734e-01 -1.34204298e-01 -6.34798780e-02
1.04400944e-02 3.76675278e-01 5.24532139e-01 -1.10517919e+00
6.94938079e-02 -6.01949573e-01 2.86533255e-02 -1.40955341e+00
-3.62679869e-01 -8.06183159e-01 1.59271345e-01 -1.88654816e+00
5.47463894e-01 -6.62259042e-01 -7.19082057e-01 2.68243104e-01
-8.97008777e-01 3.26575130e-01 2.19919607e-02 8.87037277e-01
-1.20777380e+00 6.94673300e-01 1.44053459e+00 -1.76986650e-01
-2.30022907e-01 2.07739979e-01 -7.42694557e-01 2.01316714e-01
7.02769637e-01 -5.19166410e-01 -5.65687478e-01 -7.15353847e-01
5.28576255e-01 1.74641639e-01 1.58173591e-02 -6.93875015e-01
5.23524463e-01 -2.40530878e-01 3.10419142e-01 -7.61460721e-01
-1.39569387e-01 -4.57442850e-01 -2.89449453e-01 2.67501292e-03
-8.04679632e-01 5.42233825e-01 1.74493380e-02 2.33872011e-01
-7.45711386e-01 -4.46145952e-01 -8.67155474e-03 -1.97275013e-01
-7.91594982e-01 4.69483018e-01 -1.67957991e-01 5.68053961e-01
4.57830846e-01 1.04519069e-01 -5.68845987e-01 -5.51001370e-01
-1.66116327e-01 7.73402989e-01 4.81023461e-01 8.84642959e-01
5.50919771e-01 -1.58584070e+00 -1.06144428e+00 -4.30579901e-01
7.74477065e-01 1.26634106e-01 4.98761952e-01 6.46431684e-01
-3.64818931e-01 8.30486834e-01 2.23078713e-01 -2.88927108e-01
-1.18049586e+00 2.42581978e-01 -1.67923167e-01 -7.94591665e-01
-2.99856693e-01 7.17723370e-01 -5.10648899e-02 -6.06342852e-01
2.19959036e-01 -1.58364683e-01 -9.39684689e-01 9.36090052e-02
9.75337982e-01 3.00325930e-01 3.70467246e-01 -5.14986396e-01
-1.96285069e-01 4.74309474e-01 -4.38177377e-01 -9.60010663e-02
7.51994073e-01 -5.75968623e-01 -3.14025015e-01 2.38532990e-01
1.24892890e+00 1.90187395e-01 -5.15603185e-01 -4.69740212e-01
3.07896674e-01 -5.44288576e-01 -2.96687126e-01 -9.24513280e-01
-5.56533456e-01 4.80243951e-01 2.54000813e-01 3.67623985e-01
1.24008107e+00 -1.01239339e-01 9.21190977e-01 7.07720876e-01
5.61244369e-01 -1.15854275e+00 1.49425015e-01 7.93419480e-01
1.02465200e+00 -1.18889070e+00 -5.45849316e-02 -1.26094401e-01
-5.57493329e-01 7.04085529e-01 7.03911006e-01 3.52663547e-01
3.09669435e-01 -5.01550019e-01 4.44982499e-01 6.72467873e-02
-5.94618380e-01 -1.19192794e-01 8.16563010e-01 4.86680865e-02
9.11293209e-01 -2.31626317e-01 -9.45624530e-01 7.32932806e-01
-2.85380065e-01 -7.53008947e-02 1.33544356e-01 9.43542778e-01
-2.76827186e-01 -1.34451962e+00 -3.15323412e-01 7.99916387e-01
-7.48395562e-01 -4.70759481e-01 -1.34302974e-01 5.05959213e-01
-5.47384202e-01 1.29553437e+00 -3.37235332e-01 -3.35392296e-01
4.68384922e-01 -4.46982086e-01 -3.24053578e-02 -5.86983442e-01
-7.25097835e-01 -1.22481018e-01 1.31102547e-01 -2.20168099e-01
-3.48951489e-01 -3.69735807e-01 -1.05221057e+00 -4.99860682e-02
-6.53325796e-01 1.09342766e+00 2.18886927e-01 7.68393993e-01
3.88792485e-01 2.42072567e-01 9.60105836e-01 -4.38615948e-01
-5.38677037e-01 -1.31081617e+00 -2.94035375e-01 6.63638473e-01
1.40662983e-01 -2.68917471e-01 -3.78201276e-01 -2.77984023e-01]
|
[11.47692584991455, 7.733522415161133]
|
31bff8f5-836f-4b05-a99b-76e0b7ae6461
|
2-bit-conformer-quantization-for-automatic
|
2305.16619
| null |
https://arxiv.org/abs/2305.16619v1
|
https://arxiv.org/pdf/2305.16619v1.pdf
|
2-bit Conformer quantization for automatic speech recognition
|
Large speech models are rapidly gaining traction in research community. As a result, model compression has become an important topic, so that these models can fit in memory and be served with reduced cost. Practical approaches for compressing automatic speech recognition (ASR) model use int8 or int4 weight quantization. In this study, we propose to develop 2-bit ASR models. We explore the impact of symmetric and asymmetric quantization combined with sub-channel quantization and clipping on both LibriSpeech dataset and large-scale training data. We obtain a lossless 2-bit Conformer model with 32% model size reduction when compared to state of the art 4-bit Conformer model for LibriSpeech. With the large-scale training data, we obtain a 2-bit Conformer model with over 40% model size reduction against the 4-bit version at the cost of 17% relative word error rate degradation
|
['Yanzhang He', 'David Rim', 'Jian Li', 'David Qiu', 'Shaojin Ding', 'Phoenix Meadowlark', 'Oleg Rybakov']
|
2023-05-26
| null | null | null | null |
['model-compression', 'automatic-speech-recognition']
|
['methodology', 'speech']
|
[ 4.40718114e-01 -1.13442661e-02 -3.83188307e-01 -3.62952352e-01
-1.13143384e+00 -7.36052841e-02 2.39392698e-01 2.31742859e-01
-6.50887907e-01 4.26591367e-01 3.29903096e-01 -5.11807621e-01
2.62728661e-01 -5.04737079e-01 -5.05535901e-01 -4.19361681e-01
4.30982420e-03 3.54310304e-01 2.85909384e-01 -2.00347289e-01
2.07518756e-01 4.07245755e-01 -1.39500391e+00 4.53861862e-01
6.18913412e-01 1.33738649e+00 5.16679764e-01 9.02397811e-01
-1.02788799e-01 5.74448347e-01 -8.58337402e-01 -3.95881832e-01
2.50760943e-01 -1.06722616e-01 -6.00855827e-01 -1.89104408e-01
3.55615258e-01 -5.69466174e-01 -1.15717125e+00 1.04796267e+00
7.87389636e-01 3.74582112e-02 3.22835088e-01 -7.99065530e-01
-3.96011859e-01 8.50827694e-01 -3.17420989e-01 1.76152974e-01
-7.45241195e-02 -3.59520465e-02 8.83642375e-01 -6.36215448e-01
2.85328031e-02 1.52497482e+00 4.41555649e-01 7.59962320e-01
-9.62135553e-01 -1.03190041e+00 -4.20625091e-01 6.10870540e-01
-1.87699640e+00 -1.26816761e+00 3.79740328e-01 3.50889474e-01
1.67787313e+00 7.48210728e-01 5.53239048e-01 6.82747543e-01
-8.26655477e-02 7.89991975e-01 6.66696370e-01 -5.00507414e-01
5.02883971e-01 -1.58003107e-01 1.50387675e-01 3.91007781e-01
4.21857208e-01 7.06740171e-02 -8.32483590e-01 -2.38592297e-01
4.15089279e-01 -1.54197678e-01 -3.00939411e-01 5.82396746e-01
-6.00320876e-01 5.40003479e-01 8.39959383e-02 2.99139135e-02
-3.80674630e-01 5.96296191e-01 4.61784869e-01 1.79393113e-01
3.33201885e-01 -3.64660136e-02 -3.12002361e-01 -8.25613201e-01
-1.43490076e+00 3.08361918e-01 5.49432397e-01 1.30114532e+00
3.11199903e-01 5.61330080e-01 -1.02669932e-01 1.29108620e+00
7.40208805e-01 8.37978899e-01 1.08281481e+00 -8.89776886e-01
8.45037699e-01 1.04398072e-01 -1.23303629e-01 -4.78105813e-01
1.88834146e-01 -3.82951558e-01 -1.08036292e+00 -3.14399064e-01
-4.06425148e-01 1.89129859e-01 -1.27532446e+00 1.37219584e+00
-8.72443840e-02 8.79133269e-02 1.68870032e-01 5.96987963e-01
4.11169976e-01 1.31627750e+00 -2.49926329e-01 -5.13563573e-01
1.31778646e+00 -9.22268271e-01 -1.17134297e+00 -3.85653883e-01
8.71748865e-01 -9.72655892e-01 7.35459745e-01 4.59396183e-01
-1.51977062e+00 -4.81577873e-01 -1.38491488e+00 -2.40513921e-01
7.90599436e-02 1.30748078e-02 -6.77588806e-02 1.08365798e+00
-1.21970010e+00 4.44821060e-01 -1.03353524e+00 2.05699936e-01
3.68486047e-01 3.60140979e-01 -6.60365522e-02 -2.37278685e-01
-1.20665765e+00 8.06886792e-01 4.08579588e-01 -3.09551328e-01
-6.66215897e-01 -6.76241279e-01 -6.47653997e-01 5.31360626e-01
-1.14005916e-01 2.83393860e-02 1.56987214e+00 -2.50641435e-01
-1.56231594e+00 3.05665880e-01 -7.10364461e-01 -1.15684545e+00
2.23527420e-02 -1.22463703e-02 -7.90797651e-01 1.44711792e-01
-8.81058812e-01 6.47678971e-01 7.83749580e-01 -7.53576458e-01
-5.56336164e-01 -3.84791970e-01 -5.44244230e-01 9.64143500e-02
-8.39271307e-01 2.18895942e-01 -3.94207954e-01 -1.09704781e+00
3.14315289e-01 -8.49311829e-01 -1.00198746e-01 6.24404941e-03
-1.40455782e-01 -4.76661101e-02 1.19083834e+00 -1.07000923e+00
2.13985801e+00 -2.26574039e+00 -2.29060590e-01 6.97300583e-02
1.00466177e-01 1.04047275e+00 -1.32826529e-02 3.83179218e-01
2.34839618e-01 4.48159695e-01 -2.89855629e-01 -7.38551319e-01
1.30865887e-01 2.69199282e-01 -3.98253411e-01 1.72342911e-01
-2.77135253e-01 6.37265384e-01 -2.61079609e-01 -4.34139758e-01
8.03595185e-02 6.80518866e-01 -6.52519524e-01 2.55705625e-01
3.10933776e-02 -6.38788640e-01 3.60669866e-02 5.50383151e-01
9.22643006e-01 6.91922829e-02 5.40071391e-02 -1.69763044e-01
3.14581841e-01 9.48713005e-01 -9.20323670e-01 1.44662654e+00
-4.74974960e-01 8.80088866e-01 4.06320393e-01 -5.75619340e-01
1.17910600e+00 7.41039753e-01 -9.46558639e-02 -1.12316406e+00
-6.66586682e-02 5.31033278e-01 1.18861005e-01 6.89375475e-02
1.01972914e+00 -3.20808649e-01 2.56336987e-01 2.02305451e-01
-2.25539505e-01 -4.40427154e-01 -6.26208261e-02 5.46093099e-02
1.13304210e+00 -8.35731685e-01 -1.68364067e-02 6.63057119e-02
2.84385175e-01 -5.45412362e-01 4.24566537e-01 2.33384863e-01
-4.03269440e-01 5.78322709e-01 -1.82879761e-01 -8.46974701e-02
-1.58476734e+00 -5.71823001e-01 -2.94408977e-01 5.32600045e-01
-2.44150922e-01 -8.75739038e-01 -1.03739309e+00 1.42888606e-01
-2.62177527e-01 9.00614440e-01 3.73293340e-01 -4.98000026e-01
-9.05480266e-01 -6.50293410e-01 1.04069984e+00 3.65966886e-01
6.29339159e-01 -6.61805630e-01 -5.28367996e-01 3.55067372e-01
-8.62268135e-02 -1.29953969e+00 -7.69115388e-01 2.43924126e-01
-1.11690748e+00 -1.17317632e-01 -8.09089661e-01 -7.40605175e-01
1.34282991e-01 3.66397828e-01 6.38239563e-01 1.78389307e-02
-1.63501911e-02 -2.91743934e-01 -5.40493190e-01 -1.31564379e-01
-9.82983172e-01 1.38679415e-01 2.59070396e-01 -3.24407697e-01
2.97524631e-01 -4.49263632e-01 -5.04310966e-01 1.37542218e-01
-1.06081116e+00 7.83179402e-02 5.22288263e-01 6.60509825e-01
6.89099312e-01 9.80585963e-02 7.46616304e-01 -1.91324443e-01
7.16471434e-01 -9.99718308e-02 -5.97049475e-01 2.81150728e-01
-1.03452778e+00 2.83835232e-01 7.12204158e-01 -2.99413741e-01
-7.21664608e-01 -2.81255156e-01 -6.33671403e-01 -7.41469979e-01
2.54245460e-01 3.07252377e-01 -2.35384911e-01 3.56346332e-02
4.15221870e-01 5.70897877e-01 4.40296270e-02 -7.18403637e-01
2.85647988e-01 1.78942013e+00 1.19681850e-01 -2.06318885e-01
3.51939976e-01 -9.90887806e-02 -2.09043950e-01 -1.19422448e+00
2.07740530e-01 -3.59299153e-01 9.53382533e-03 3.40148181e-01
3.69231284e-01 -9.32930410e-01 -2.82001942e-01 5.31526864e-01
-1.25417686e+00 -3.60281080e-01 -2.66243190e-01 4.87374842e-01
-5.48924327e-01 6.55871212e-01 -9.15298522e-01 -1.11176276e+00
-1.01341403e+00 -1.28933203e+00 1.05754185e+00 -2.80029953e-01
-1.41964123e-01 -1.95108563e-01 -2.54972488e-01 4.86184061e-01
6.82216585e-01 -8.90078306e-01 9.33834136e-01 -5.68890631e-01
-4.72717404e-01 -5.52578032e-01 -1.38842911e-01 7.14432895e-01
8.36509243e-02 -4.11174238e-01 -1.02492774e+00 -7.28120387e-01
1.13209091e-01 -3.64978105e-01 6.88864589e-01 1.16853178e-01
1.56963944e+00 -8.24912965e-01 -4.34749350e-02 6.76124215e-01
1.09341633e+00 6.63401902e-01 9.58641112e-01 -3.37323427e-01
3.99109662e-01 1.13225959e-01 3.35964233e-01 6.84923649e-01
2.23981366e-02 1.01842940e+00 1.35629937e-01 5.48267305e-01
-7.28460431e-01 -3.54559541e-01 5.03504574e-01 1.94968033e+00
4.45503801e-01 -6.19540036e-01 -8.42150867e-01 3.43987674e-01
-1.15252709e+00 -8.71861696e-01 2.16862693e-01 2.52978706e+00
1.25457263e+00 2.13013604e-01 -1.65467083e-01 7.57448614e-01
7.19889462e-01 3.47748458e-01 -4.30446059e-01 -6.95211411e-01
-1.89732403e-01 3.88831347e-01 8.93872976e-01 6.51283324e-01
-4.50731426e-01 8.71838868e-01 6.50661135e+00 1.88241220e+00
-1.23517585e+00 2.54139632e-01 1.05563354e+00 -4.77875501e-01
-3.14775199e-01 -2.75682032e-01 -1.18037939e+00 7.92055011e-01
2.15337515e+00 -3.66184056e-01 7.81612933e-01 6.19863927e-01
4.17375386e-01 5.74750662e-01 -7.79469252e-01 1.55189764e+00
-6.31474331e-02 -1.40023077e+00 4.22478348e-01 2.91571707e-01
4.01321948e-01 1.30807430e-01 1.38949186e-01 1.85255095e-01
-1.84381142e-01 -1.29849589e+00 9.85490859e-01 -4.61181328e-02
1.44336557e+00 -8.78532231e-01 4.09301668e-01 4.72901464e-01
-1.36358976e+00 -9.62728485e-02 -6.96014404e-01 2.93706834e-01
4.56327528e-01 2.64430374e-01 -7.48396039e-01 1.22930948e-02
4.48249280e-01 1.36052892e-01 -1.33014113e-01 1.03430796e+00
5.04006088e-01 1.15585053e+00 -6.12989068e-01 -2.90134460e-01
1.55788749e-01 2.12504342e-01 3.98914218e-01 1.26046085e+00
7.14173257e-01 2.48995170e-01 -4.62628752e-01 3.88573349e-01
-3.61773342e-01 -3.09270676e-02 -1.21226713e-01 -2.23462388e-01
1.32517755e+00 5.38881958e-01 1.76935047e-02 -6.80806160e-01
-9.28836912e-02 9.99710917e-01 7.07963407e-02 1.24067418e-01
-5.43114066e-01 -6.56414151e-01 7.21861422e-01 2.38365889e-01
3.98523062e-01 -4.63374138e-01 -3.34493756e-01 -8.87223423e-01
2.14362293e-01 -1.26257873e+00 -1.73154354e-01 -6.45369768e-01
-4.63806987e-01 7.09525645e-01 -7.16900826e-02 -1.00356138e+00
-3.21240962e-01 -3.02730203e-01 -9.54320133e-02 9.60287094e-01
-1.51406336e+00 -6.39770150e-01 8.17541480e-02 2.23405510e-01
9.78381515e-01 -2.92938471e-01 9.76082683e-01 7.49146402e-01
-3.89111727e-01 1.32218635e+00 5.14840007e-01 -3.14849555e-01
2.22157016e-01 -5.70794940e-01 9.00125146e-01 7.09414601e-01
1.41219050e-01 5.05523026e-01 4.59104776e-01 -4.73110408e-01
-1.80943346e+00 -1.15619802e+00 1.14804983e+00 3.33763778e-01
2.64089972e-01 -4.60835457e-01 -1.12473857e+00 2.88098603e-01
1.00521788e-01 1.24938693e-02 4.62167561e-01 -6.43052459e-01
-4.19377327e-01 -3.43857229e-01 -1.42228258e+00 6.84664369e-01
8.63457382e-01 -7.90043235e-01 -1.97501436e-01 4.87131104e-02
1.44834590e+00 -3.20965320e-01 -9.61111605e-01 2.78920770e-01
3.07657987e-01 -4.56547648e-01 8.86538446e-01 -1.32737309e-01
-8.58643875e-02 2.89022289e-02 -9.57184315e-01 -9.84885514e-01
-1.87408738e-02 -1.00494051e+00 -7.56990731e-01 9.78786647e-01
3.88494134e-01 -4.50912893e-01 9.90609765e-01 7.77699769e-01
-5.30382454e-01 -8.45912814e-01 -1.66179299e+00 -9.90781903e-01
2.12062076e-01 -6.70655668e-01 8.77104700e-01 2.26917922e-01
1.57226905e-01 2.57103164e-02 -4.86620426e-01 -4.94201779e-02
5.30777335e-01 -7.20216215e-01 2.19476640e-01 -4.01121825e-01
-3.10338408e-01 -4.77631599e-01 -5.42609751e-01 -1.80900717e+00
1.43622169e-02 -9.12661672e-01 2.34006401e-02 -1.05234766e+00
-3.93087342e-02 -6.14315510e-01 -2.09441766e-01 2.60728896e-01
1.97409436e-01 6.11806288e-02 3.92063558e-01 2.31966332e-01
-2.66326934e-01 9.69942808e-01 7.94322491e-01 -4.11856532e-01
9.97716635e-02 -2.06652984e-01 -1.69698671e-01 1.11081205e-01
7.95251846e-01 -3.48686635e-01 -5.35359621e-01 -6.16649747e-01
-2.91578352e-01 3.53746057e-01 -1.80057481e-01 -1.32410419e+00
3.93544674e-01 1.53589457e-01 -1.56776816e-01 -5.97883463e-01
9.53139603e-01 -9.04638410e-01 1.18272491e-01 7.35508263e-01
-5.49088657e-01 1.57684311e-01 3.89487475e-01 3.19693029e-01
-4.16349262e-01 -5.10410666e-01 9.97435510e-01 2.42314190e-01
-3.20816398e-01 3.61493438e-01 -4.07543927e-01 -1.05920412e-01
5.24938166e-01 -2.85575897e-01 -3.11361581e-01 -5.46039760e-01
-2.94456363e-01 -1.95360363e-01 2.65363276e-01 3.11671644e-01
1.20282316e+00 -1.28630793e+00 -7.95583427e-01 5.71220279e-01
-1.42450482e-01 -5.25661446e-02 2.52225995e-01 7.00451806e-02
-8.07577670e-01 1.03099155e+00 3.43495488e-01 -3.03979903e-01
-1.55707479e+00 2.05296963e-01 1.77733138e-01 -2.97601167e-02
-4.02957082e-01 8.09508801e-01 -4.36663955e-01 9.73875374e-02
5.66032052e-01 -5.39077818e-01 4.42880720e-01 -5.14855504e-01
9.20277774e-01 7.95811832e-01 5.61732531e-01 -7.28814483e-01
-4.42998782e-02 1.59737453e-01 -2.66391933e-01 -4.80297953e-01
1.07686961e+00 -2.53891021e-01 1.14772178e-01 5.08574620e-02
1.62785912e+00 -3.02938968e-01 -9.62208927e-01 -2.30733395e-01
-1.48892492e-01 -5.83461165e-01 5.97539842e-01 -5.58024526e-01
-9.04305518e-01 1.26694715e+00 1.06814110e+00 1.47400245e-01
1.07142019e+00 -3.09402883e-01 1.69626248e+00 4.17415500e-01
5.51642358e-01 -1.12131464e+00 -1.69650212e-01 5.42396486e-01
8.06670666e-01 -6.84122980e-01 -2.07643449e-01 -1.10217229e-01
-2.61108756e-01 8.27773213e-01 1.38813868e-01 2.89690763e-01
6.60263538e-01 5.73194981e-01 -1.56477839e-01 3.65844488e-01
-1.11766720e+00 4.16291356e-01 1.09015852e-01 4.00543392e-01
4.02947038e-01 3.21099669e-01 -2.46593744e-01 4.75895017e-01
-5.88591814e-01 -1.05421074e-01 2.67881334e-01 8.85414720e-01
-8.87680471e-01 -1.33269691e+00 -3.28018963e-01 6.26349151e-01
-5.24874032e-01 -6.47648990e-01 1.13361932e-01 -1.49085090e-01
-5.62993288e-01 1.24940789e+00 2.79766858e-01 -7.29734004e-01
9.04652923e-02 9.40078944e-02 1.90369084e-01 -3.49322945e-01
-3.99568707e-01 2.76991367e-01 2.49008894e-01 -3.94605249e-01
3.75020385e-01 -9.68750194e-02 -1.35865641e+00 -8.35241318e-01
-6.34363890e-01 1.57142580e-01 1.21498966e+00 4.69603986e-01
9.06949759e-01 3.70594978e-01 4.89566118e-01 -5.34180284e-01
-1.19244373e+00 -1.29136527e+00 -6.27580404e-01 3.07263192e-02
4.97818083e-01 1.04836658e-01 -2.89778680e-01 1.55588146e-02]
|
[14.277769088745117, 6.270174980163574]
|
ff8ab487-dcc6-47b2-a4e2-ac50a8395026
|
neural-view-synthesis-and-matching-for-semi
|
2110.14213
| null |
https://arxiv.org/abs/2110.14213v1
|
https://arxiv.org/pdf/2110.14213v1.pdf
|
Neural View Synthesis and Matching for Semi-Supervised Few-Shot Learning of 3D Pose
|
We study the problem of learning to estimate the 3D object pose from a few labelled examples and a collection of unlabelled data. Our main contribution is a learning framework, neural view synthesis and matching, that can transfer the 3D pose annotation from the labelled to unlabelled images reliably, despite unseen 3D views and nuisance variations such as the object shape, texture, illumination or scene context. In our approach, objects are represented as 3D cuboid meshes composed of feature vectors at each mesh vertex. The model is initialized from a few labelled images and is subsequently used to synthesize feature representations of unseen 3D views. The synthesized views are matched with the feature representations of unlabelled images to generate pseudo-labels of the 3D pose. The pseudo-labelled data is, in turn, used to train the feature extractor such that the features at each mesh vertex are more invariant across varying 3D views of the object. Our model is trained in an EM-type manner alternating between increasing the 3D pose invariance of the feature extractor and annotating unlabelled data through neural view synthesis and matching. We demonstrate the effectiveness of the proposed semi-supervised learning framework for 3D pose estimation on the PASCAL3D+ and KITTI datasets. We find that our approach outperforms all baselines by a wide margin, particularly in an extreme few-shot setting where only 7 annotated images are given. Remarkably, we observe that our model also achieves an exceptional robustness in out-of-distribution scenarios that involve partial occlusion.
|
['Adam Kortylewski', 'Alan Yuille', 'Shenxiao Mei', 'Angtian Wang']
|
2021-10-27
| null |
http://proceedings.neurips.cc/paper/2021/hash/3a61ed715ee66c48bacf237fa7bb5289-Abstract.html
|
http://proceedings.neurips.cc/paper/2021/file/3a61ed715ee66c48bacf237fa7bb5289-Paper.pdf
|
neurips-2021-12
|
['3d-pose-estimation']
|
['computer-vision']
|
[ 3.07077080e-01 3.34848315e-01 -8.18925202e-02 -5.43977380e-01
-1.06425953e+00 -9.21115100e-01 8.27417612e-01 -3.86718243e-01
-2.31552467e-01 2.61924148e-01 6.01136610e-02 2.52501577e-01
3.16061288e-01 -3.24523985e-01 -1.28883600e+00 -7.69666910e-01
2.51947552e-01 1.04495502e+00 2.62383252e-01 1.51913360e-01
2.50037521e-01 8.68447304e-01 -1.86235058e+00 1.86585486e-01
6.58129454e-02 1.17962611e+00 5.05140014e-02 7.89725244e-01
1.77219734e-01 2.41506323e-01 -4.56241280e-01 -2.30557010e-01
7.92766511e-01 -1.72705278e-01 -5.90995133e-01 9.58953738e-01
9.53720272e-01 -4.74624544e-01 -2.19053701e-01 8.58518541e-01
5.78191459e-01 -3.58305592e-03 1.13371742e+00 -1.11541235e+00
-1.63322732e-01 -2.74648905e-01 -7.01452255e-01 -2.60513514e-01
6.27511919e-01 1.52629495e-01 7.85382569e-01 -1.36646187e+00
1.09940171e+00 1.37122464e+00 4.00134057e-01 6.99089229e-01
-1.32839847e+00 -2.67737567e-01 8.68623331e-02 -2.63335288e-01
-1.16877604e+00 -5.04032135e-01 9.05024171e-01 -6.10542953e-01
7.95403183e-01 8.37704837e-02 5.29570639e-01 1.27007961e+00
1.55987486e-01 7.06806183e-01 1.10454798e+00 -5.37115574e-01
4.05618876e-01 1.87565804e-01 -2.96169102e-01 7.95564651e-01
-4.38414840e-03 2.40749776e-01 -5.27072430e-01 -2.59312898e-01
9.03443933e-01 8.80817771e-02 -5.39299138e-02 -1.45209908e+00
-1.30981135e+00 6.61775887e-01 2.82100469e-01 -4.69939977e-01
-3.50985020e-01 -1.26041211e-02 2.86309272e-01 1.22264557e-01
5.71141303e-01 2.16637820e-01 -6.84050620e-01 2.58717656e-01
-5.17613947e-01 4.01391953e-01 7.96842277e-01 1.31253362e+00
8.75266969e-01 -2.04551965e-01 1.12260422e-02 7.47424841e-01
3.04553717e-01 8.50231171e-01 1.69873595e-01 -1.28372216e+00
3.95227909e-01 6.73346639e-01 2.27650777e-01 -5.65233529e-01
-2.33261779e-01 -2.35908434e-01 -3.85825902e-01 4.08722490e-01
4.81255203e-01 1.14952959e-01 -1.27506649e+00 1.64734232e+00
9.46467936e-01 -1.76657185e-01 9.88403428e-03 9.23622012e-01
7.50951529e-01 3.11790854e-01 -3.65260988e-01 -8.21865574e-02
1.26721787e+00 -6.96292043e-01 -1.07188612e-01 -3.34156662e-01
4.28338617e-01 -8.88752222e-01 6.94924116e-01 1.82714418e-01
-1.20647049e+00 -7.02310801e-01 -9.68290687e-01 -4.26332057e-02
-2.36808851e-01 7.32320622e-02 1.18432522e-01 3.51175576e-01
-7.14013338e-01 4.30145174e-01 -6.61692441e-01 -3.70883375e-01
4.74955946e-01 4.52867478e-01 -7.63983309e-01 -2.53421754e-01
-5.18853486e-01 1.01790822e+00 4.80892152e-01 1.10189505e-02
-1.43977046e+00 -4.34000820e-01 -1.13038397e+00 -4.62076753e-01
5.84549069e-01 -7.69355118e-01 1.17316246e+00 -9.23955679e-01
-1.31398451e+00 1.57390594e+00 -5.82634471e-02 9.22952369e-02
6.39163673e-01 -8.04600492e-02 1.28300384e-01 8.88736695e-02
1.36672303e-01 9.54145789e-01 1.25505269e+00 -1.69984722e+00
-5.52426815e-01 -8.74327421e-01 1.50039857e-02 5.35622120e-01
4.37937081e-01 -2.86469281e-01 -7.33273089e-01 -3.74047458e-01
5.13297319e-01 -1.30126560e+00 -1.83460906e-01 2.34879225e-01
-4.45117474e-01 1.07957162e-02 9.17959690e-01 -1.93419933e-01
9.11321715e-02 -2.07072234e+00 6.24191165e-01 1.35045126e-01
6.36079162e-02 -6.77671656e-03 -1.81823134e-01 2.90067255e-01
-2.49475449e-01 -3.38331044e-01 -5.49308956e-02 -3.80085051e-01
6.19806349e-03 3.43348235e-01 -1.14374876e-01 8.50212038e-01
3.76715571e-01 9.17519271e-01 -7.64917970e-01 -2.99209893e-01
5.37974596e-01 4.16344225e-01 -4.04771805e-01 6.69591308e-01
-4.08556730e-01 6.18668854e-01 -3.70648921e-01 7.04671085e-01
6.79863155e-01 -1.49056360e-01 -5.52705824e-02 -2.82564998e-01
2.93139637e-01 -1.01086102e-01 -1.33150673e+00 1.89265573e+00
-3.86165589e-01 1.03193559e-01 -1.06631249e-01 -8.76266778e-01
1.03146029e+00 3.27047199e-01 3.72294039e-01 -2.71820664e-01
1.92448363e-01 2.87349969e-01 -3.79422426e-01 -5.70572972e-01
6.29274175e-02 -1.18992761e-01 -2.43768260e-01 5.81758320e-01
4.55692649e-01 -7.54154682e-01 -1.70386627e-01 9.68626738e-02
8.19576859e-01 6.09325528e-01 3.85082483e-01 4.30081077e-02
4.19248283e-01 -2.03440875e-01 1.86256915e-01 4.34020311e-01
2.98035163e-02 9.94988024e-01 4.06227261e-01 -6.48375690e-01
-1.58819342e+00 -1.27820337e+00 -2.88306803e-01 7.97427654e-01
6.10685609e-02 7.13408068e-02 -6.79288626e-01 -1.31675828e+00
1.55972853e-01 4.29763019e-01 -7.88322926e-01 -1.16684750e-01
-3.45320314e-01 -1.80179864e-01 6.78501616e-04 5.09963870e-01
9.70246568e-02 -9.34943199e-01 -7.96122074e-01 -1.29599750e-01
1.76473916e-01 -1.32173598e+00 -4.80333328e-01 4.95960265e-01
-8.73219848e-01 -1.14428806e+00 -6.89502478e-01 -8.95898640e-01
1.15897882e+00 2.57284880e-01 1.12833035e+00 -2.01500908e-01
-3.09984654e-01 6.35532081e-01 -3.00594300e-01 -3.88396472e-01
-5.27860224e-01 -2.54402786e-01 2.18825653e-01 1.58381730e-01
2.39329875e-01 -4.67402101e-01 -5.45189321e-01 5.44702768e-01
-7.72907257e-01 3.57793388e-03 5.05251229e-01 8.91891658e-01
9.38282311e-01 -3.17644328e-01 2.28599742e-01 -1.03577197e+00
-2.78994799e-01 -2.09558845e-01 -6.54948473e-01 4.89305854e-02
3.69255170e-02 1.42326981e-01 3.24980855e-01 -6.18118882e-01
-9.34461057e-01 8.05609405e-01 1.97226137e-01 -8.51986408e-01
-5.88835418e-01 -1.59260541e-01 -4.85356778e-01 -1.51592478e-01
8.47021103e-01 -1.92913469e-02 5.34058847e-02 -3.98257971e-01
4.69713479e-01 6.54491782e-01 6.28621876e-01 -6.74125731e-01
1.05593443e+00 6.58684790e-01 2.40193963e-01 -6.91264093e-01
-1.11597574e+00 -4.88970846e-01 -1.23107374e+00 -3.00781876e-01
8.70422959e-01 -1.07208490e+00 -3.34971577e-01 3.97193700e-01
-1.18095124e+00 -1.06541790e-01 -4.15387958e-01 4.93308276e-01
-1.11528420e+00 1.02965288e-01 -2.19051048e-01 -7.16706216e-01
4.74928971e-03 -1.28452837e+00 1.94034231e+00 -9.00922939e-02
-2.85600394e-01 -8.04659486e-01 2.01405883e-02 5.29955566e-01
-2.68818945e-01 5.55792034e-01 8.99447143e-01 -7.87037730e-01
-5.55285573e-01 -5.73198736e-01 -1.44839268e-02 5.01816094e-01
-2.27754600e-02 -1.37889892e-01 -1.35103524e+00 -3.51344168e-01
1.74187973e-01 -8.66322994e-01 3.73647243e-01 2.81161070e-01
9.69172895e-01 8.18468928e-02 -2.59611517e-01 6.32053673e-01
1.21568573e+00 -1.23105906e-02 2.32645690e-01 -4.85209152e-02
7.56764948e-01 8.88510048e-01 7.12071598e-01 3.04958433e-01
7.25830197e-02 8.27939808e-01 7.94890583e-01 -5.39889932e-02
-2.08242051e-02 -3.84152293e-01 1.88712820e-01 4.11132663e-01
-7.73155987e-02 -8.93672481e-02 -6.92913592e-01 2.54830509e-01
-1.54564273e+00 -6.19259715e-01 1.71380132e-01 2.45974207e+00
5.15061677e-01 3.12137544e-01 1.27459764e-01 -7.44913146e-02
5.88918149e-01 4.39022928e-02 -8.51796389e-01 -1.83389068e-01
9.46016759e-02 7.08530024e-02 3.98492068e-01 3.33237171e-01
-1.08559954e+00 6.53426528e-01 5.54901171e+00 4.05693948e-01
-7.79829860e-01 -2.11268052e-01 5.79607248e-01 -1.31596714e-01
-1.18783295e-01 -2.39675306e-02 -8.29395294e-01 1.07247800e-01
4.25371319e-01 3.04281086e-01 2.77195007e-01 9.16776419e-01
-1.83036909e-01 -4.92692750e-04 -1.67656136e+00 1.01407659e+00
4.65483934e-01 -1.00862765e+00 1.27889171e-01 2.27227584e-01
9.77225244e-01 -4.45810296e-02 5.13020623e-03 2.14369327e-01
1.35103673e-01 -7.80486345e-01 8.20908308e-01 4.05741870e-01
9.62816775e-01 -7.06297696e-01 6.62048936e-01 6.75938368e-01
-7.69950330e-01 1.71095401e-01 -3.68566215e-01 2.09439829e-01
1.27925485e-01 2.03350067e-01 -9.84272838e-01 4.51633960e-01
5.28112113e-01 5.42840004e-01 -5.43301225e-01 6.89212322e-01
-1.93403676e-01 -1.41709559e-02 -2.68924087e-01 2.91853547e-01
1.31492332e-01 -4.27007638e-02 6.30972803e-01 7.36522555e-01
1.17435962e-01 -2.27715038e-02 2.36611128e-01 7.19780445e-01
-2.09941000e-01 -2.04750821e-01 -9.80325699e-01 3.39606196e-01
2.65832514e-01 1.29701030e+00 -7.53147781e-01 -2.01339543e-01
-3.42949390e-01 1.06360412e+00 3.69042575e-01 3.73890132e-01
-5.04434526e-01 1.17554352e-01 1.35754153e-01 1.50959328e-01
6.91167653e-01 7.70969409e-03 1.74374625e-01 -1.17701948e+00
2.18779653e-01 -8.65400374e-01 1.39969662e-01 -1.08027363e+00
-1.37936938e+00 4.91114378e-01 3.88952643e-01 -1.38350868e+00
-6.31979644e-01 -7.79512882e-01 -2.68619567e-01 6.96429729e-01
-8.83557141e-01 -1.29632485e+00 -2.66890347e-01 3.00960481e-01
7.59649038e-01 -1.13321811e-01 9.35199380e-01 -1.45899445e-01
5.29398629e-03 2.05989495e-01 -1.39876604e-01 -7.26947039e-02
6.32746339e-01 -1.32622302e+00 5.82436562e-01 3.56265306e-01
3.91036928e-01 8.70447159e-02 6.86119199e-01 -4.58618522e-01
-1.59379387e+00 -1.10267293e+00 5.59291601e-01 -1.09781051e+00
1.58942804e-01 -9.86113906e-01 -6.79941952e-01 7.65993893e-01
-2.02227682e-01 5.98721385e-01 2.57731527e-01 -2.38478377e-01
-4.34165657e-01 1.15817055e-01 -1.20084321e+00 3.94349843e-01
1.20003223e+00 -5.86073935e-01 -6.97641313e-01 4.08100903e-01
5.06206393e-01 -8.82943034e-01 -7.72584260e-01 5.98546863e-01
7.24310875e-01 -9.09358382e-01 1.09174073e+00 -7.51462758e-01
3.64308119e-01 -2.52115190e-01 -5.85982978e-01 -1.29109323e+00
4.74407636e-02 -3.09940815e-01 -5.14412597e-02 9.61200535e-01
2.35009417e-01 -1.90014824e-01 9.84840930e-01 5.24752498e-01
4.43819612e-02 -9.02463436e-01 -9.35756385e-01 -6.34193361e-01
-1.08355775e-01 -2.84178197e-01 3.30001593e-01 5.37615061e-01
-6.51918650e-01 7.07470775e-01 -2.85291344e-01 1.86322406e-01
8.86789739e-01 3.26225787e-01 1.32399881e+00 -1.14297414e+00
-3.81635815e-01 1.75387472e-01 -7.76592433e-01 -1.17369306e+00
3.48054230e-01 -9.31687117e-01 1.59032956e-01 -9.68845606e-01
4.71643597e-01 -1.23674452e-01 2.93129951e-01 1.81934163e-01
6.11412041e-02 3.87874663e-01 5.51559143e-02 6.62309350e-03
-6.13881528e-01 4.77879882e-01 1.51368749e+00 4.05718274e-02
-5.89626729e-02 1.84639931e-01 -7.67603442e-02 9.08415675e-01
2.79897273e-01 -4.74792719e-01 -4.17865932e-01 -3.47596049e-01
-6.95429817e-02 2.12113664e-01 6.77440107e-01 -5.89484036e-01
-1.89792827e-01 -5.71785262e-03 9.21723604e-01 -8.72404993e-01
6.53160214e-01 -1.13754964e+00 2.69459963e-01 1.25499770e-01
-3.53279114e-01 -9.79560018e-02 -4.13579755e-02 6.93069756e-01
2.19915628e-01 -1.87057808e-01 7.98008621e-01 -3.47949326e-01
-4.05492127e-01 5.21502614e-01 1.61771670e-01 3.75585139e-01
1.19206285e+00 -5.11679649e-01 1.08886324e-01 -3.52123976e-01
-8.39360416e-01 1.04247235e-01 9.45921838e-01 4.22173232e-01
6.22854888e-01 -1.52655649e+00 -6.59947336e-01 7.42771387e-01
4.97668296e-01 5.88446558e-01 1.86950609e-01 3.95544857e-01
-2.62775838e-01 1.90995201e-01 -2.15107530e-01 -1.21316171e+00
-1.17166448e+00 7.95476794e-01 3.82066220e-01 1.93575948e-01
-5.09199858e-01 7.88437307e-01 4.80931044e-01 -9.55379069e-01
5.02487004e-01 -9.67735201e-02 3.28288555e-01 -3.04869618e-02
2.29287669e-01 9.87252370e-02 2.46072233e-01 -1.06060898e+00
-2.35735834e-01 1.05107296e+00 -1.63770169e-01 -1.14914507e-01
1.31857502e+00 -2.22889353e-02 3.36888671e-01 6.50505304e-01
1.59202385e+00 -1.46295354e-01 -1.76108789e+00 -3.17792773e-01
-2.99868226e-01 -5.92842102e-01 -3.52569818e-01 -5.19621789e-01
-8.88164699e-01 7.74675727e-01 6.19547546e-01 -1.85722619e-01
6.99584365e-01 6.55334532e-01 2.71006435e-01 3.74313504e-01
5.69133759e-01 -8.15068483e-01 3.44945759e-01 3.87834251e-01
9.79506731e-01 -1.40228844e+00 1.13413684e-01 -2.66702473e-01
-7.12351084e-01 9.95241225e-01 6.56011403e-01 -4.28853184e-01
4.18573976e-01 9.76568684e-02 2.39430726e-01 -5.19022465e-01
-7.41724014e-01 -5.77749945e-02 6.51021540e-01 8.23569655e-01
-2.05946201e-03 -2.14486673e-01 5.20311415e-01 5.81226824e-03
-1.69172749e-01 -3.82197738e-01 1.52113631e-01 9.09389198e-01
-3.22479665e-01 -9.28601265e-01 -4.67561483e-01 3.56999934e-01
-2.15360895e-01 4.82732862e-01 -6.79073036e-01 1.00136089e+00
1.41627237e-01 3.09687376e-01 1.57144755e-01 -2.50960410e-01
7.14860976e-01 2.26726711e-01 9.41066206e-01 -1.05550873e+00
-1.76783577e-01 3.91035348e-01 -1.15697823e-01 -6.53155386e-01
-5.65916479e-01 -7.81670809e-01 -9.94044840e-01 3.36457223e-01
-4.72194552e-01 -1.99401602e-01 6.01588130e-01 9.39247370e-01
1.45072550e-01 3.78495231e-02 9.94110107e-01 -1.66661417e+00
-7.74741352e-01 -7.33279705e-01 -6.72364771e-01 7.68790841e-01
4.69389111e-01 -1.04466867e+00 -6.73192024e-01 1.83067858e-01]
|
[8.072380065917969, -2.870124101638794]
|
40d3becc-2d56-4c9f-a8b8-994c092a5b82
|
sign-language-production-a-review
|
2103.15910
| null |
https://arxiv.org/abs/2103.15910v1
|
https://arxiv.org/pdf/2103.15910v1.pdf
|
Sign Language Production: A Review
|
Sign Language is the dominant yet non-primary form of communication language used in the deaf and hearing-impaired community. To make an easy and mutual communication between the hearing-impaired and the hearing communities, building a robust system capable of translating the spoken language into sign language and vice versa is fundamental. To this end, sign language recognition and production are two necessary parts for making such a two-way system. Sign language recognition and production need to cope with some critical challenges. In this survey, we review recent advances in Sign Language Production (SLP) and related areas using deep learning. This survey aims to briefly summarize recent achievements in SLP, discussing their advantages, limitations, and future directions of research.
|
['Mohammad Sabokrou', 'Sergio Escalera', 'Kourosh Kiani', 'Razieh Rastgoo']
|
2021-03-29
| null | null | null | null |
['sign-language-production']
|
['natural-language-processing']
|
[-9.82526466e-02 -9.08412486e-02 -1.08401880e-01 -3.23268026e-01
-6.20652497e-01 -4.17202920e-01 4.15657848e-01 -7.97689676e-01
-4.23282295e-01 7.04266191e-01 6.01467848e-01 -2.75520325e-01
8.46611559e-02 -6.97049499e-01 -2.76852876e-01 -8.50499451e-01
-6.48494661e-02 3.47035602e-02 1.25193298e-01 -4.52655762e-01
8.68153274e-02 6.65153980e-01 -1.87339735e+00 3.81853968e-01
8.34525526e-01 6.25889421e-01 4.49718028e-01 9.12499487e-01
-5.73142111e-01 9.10817623e-01 -4.53392416e-01 6.63655624e-02
5.47876321e-02 -7.30975688e-01 -8.16547155e-01 7.11144879e-02
2.99758315e-01 -7.85956860e-01 -3.25609744e-01 8.22717786e-01
9.70669389e-01 -2.07818002e-01 6.53439939e-01 -1.12651098e+00
-6.66945457e-01 4.06260788e-01 1.05781779e-01 -3.68690252e-01
6.10998690e-01 9.38174650e-02 6.85418487e-01 -9.74057078e-01
5.88438213e-01 1.12175548e+00 2.54240751e-01 1.01057827e+00
-4.72885877e-01 -8.40843439e-01 3.69430155e-01 1.84361175e-01
-1.22676969e+00 -7.06236601e-01 5.48079729e-01 -5.17000735e-01
9.07851398e-01 1.28074229e-01 1.11823440e+00 7.22881615e-01
-1.92526683e-01 1.32205248e+00 1.16550469e+00 -9.26280200e-01
-3.90768088e-02 -2.84947932e-01 2.06505880e-01 5.00380695e-01
2.29693323e-01 1.76503107e-01 -9.09744561e-01 3.16696912e-01
7.59892404e-01 -4.04273868e-01 -5.99045396e-01 -1.53460577e-01
-1.16501760e+00 3.16163510e-01 1.89972743e-01 7.81604588e-01
-3.76263857e-01 3.17686290e-01 1.06217004e-01 6.92848444e-01
-4.41336900e-01 -7.98317324e-03 -2.73570240e-01 -5.44005156e-01
-6.01323724e-01 5.11768274e-02 9.02399778e-01 9.05729175e-01
-1.10924982e-01 2.14859888e-01 -2.02832937e-01 1.12537754e+00
9.76725519e-01 8.91287863e-01 4.98280972e-01 -5.27541459e-01
3.11077803e-01 4.30205286e-01 -6.27446687e-03 -8.44832435e-02
-2.50006258e-01 2.49257293e-02 -7.25758672e-01 1.01120806e+00
6.34814620e-01 -1.73436478e-01 -1.55831254e+00 1.50638473e+00
-9.17769074e-02 -4.21121329e-01 1.96749359e-01 8.42814207e-01
1.31538820e+00 3.75591993e-01 1.21107846e-01 -2.65610605e-01
1.20168412e+00 -8.57640088e-01 -1.03691781e+00 -9.88132656e-02
2.91640341e-01 -9.93670464e-01 9.61946309e-01 3.69803905e-01
-1.04191327e+00 -2.52291918e-01 -7.34927952e-01 -3.46537441e-01
-2.98601151e-01 9.75642800e-02 4.90249962e-01 5.46126783e-01
-1.20566165e+00 -1.32305309e-01 -7.69862413e-01 -6.86143517e-01
3.61640543e-01 3.53561521e-01 -5.17956913e-01 -1.60631746e-01
-1.06936145e+00 1.12725472e+00 -2.86160479e-03 2.70985782e-01
-1.03397481e-01 -2.55667508e-01 -6.73841059e-01 -4.64385092e-01
-3.80788803e-01 -5.39214432e-01 1.87534976e+00 -5.55181563e-01
-2.03619695e+00 1.20650792e+00 -3.48107606e-01 1.39020219e-01
8.95811081e-01 -4.58258241e-01 -4.72305924e-01 -1.02103844e-01
-2.27276891e-01 4.27029371e-01 5.17859519e-01 -1.25950551e+00
-1.08949184e+00 -1.26882091e-01 -4.47501242e-01 1.57332227e-01
2.47880414e-01 6.67635083e-01 -3.62564623e-01 -5.66981196e-01
6.49589896e-01 -5.53357601e-01 3.22801650e-01 8.33378315e-01
-1.39622703e-01 -4.25842762e-01 7.56368101e-01 -9.78698313e-01
9.72253025e-01 -2.08549953e+00 -2.69342549e-02 8.95111114e-02
2.22518429e-01 6.93248093e-01 -1.62945509e-01 8.16656888e-01
2.86099702e-01 -2.99727052e-01 -2.55180985e-01 -7.30851963e-02
4.62825596e-02 5.03572881e-01 -4.64203835e-01 2.69824058e-01
-1.45906493e-01 1.00367260e+00 -9.32344854e-01 -3.57213557e-01
3.53337586e-01 6.79522216e-01 -7.70295039e-02 3.11630875e-01
1.29111290e-01 7.21306443e-01 -3.00139844e-01 1.15722907e+00
6.19708359e-01 1.92428276e-01 5.67588396e-02 3.54700893e-01
-7.09327579e-01 4.32791263e-01 -1.12592399e+00 1.15026319e+00
-6.00694239e-01 9.41371560e-01 5.24763286e-01 -6.34601414e-01
9.25419331e-01 6.45177484e-01 1.76977068e-01 -7.23588645e-01
2.81347007e-01 1.14342749e+00 3.14418614e-01 -1.00840175e+00
-2.41737053e-01 -4.56936926e-01 2.73288816e-01 5.68687081e-01
-3.72329988e-02 -4.87794787e-01 2.52719492e-01 -3.13104630e-01
6.47968769e-01 -2.32014418e-01 5.66957951e-01 2.07074329e-01
6.01583302e-01 -5.27925670e-01 2.53806442e-01 5.79068720e-01
-4.76250499e-01 5.96390367e-01 -1.20965987e-01 -2.25094110e-01
-5.00199854e-01 -1.35497761e+00 -5.18750101e-02 9.06184077e-01
-9.26222503e-02 3.32989693e-01 -3.76689941e-01 -1.59872502e-01
8.60665273e-03 3.29886109e-01 9.13986415e-02 3.03803235e-01
-7.78988540e-01 -5.11300415e-02 6.59000874e-01 6.87771261e-01
8.90195727e-01 -1.84509706e+00 -4.06178653e-01 2.98962951e-01
-2.50216454e-01 -9.26934183e-01 -2.68670768e-01 -2.40884587e-01
-4.37560141e-01 -1.19478667e+00 -1.41338897e+00 -1.94895005e+00
7.35308468e-01 2.44360328e-01 3.17078441e-01 3.67423475e-01
1.10298991e-02 4.14368331e-01 -6.55927181e-01 -5.81285834e-01
-8.09514046e-01 -3.98043066e-01 2.33477637e-01 -3.05762142e-01
4.89185452e-01 -7.79364467e-01 -2.32141554e-01 -5.12598343e-02
-6.34087920e-01 2.65930146e-01 9.20561135e-01 8.81130040e-01
1.60398036e-01 -5.38028300e-01 3.80513459e-01 2.76005387e-01
8.47595513e-01 3.60671252e-01 -4.54896092e-01 5.35419345e-01
-1.65280208e-01 -5.43469600e-02 1.57516167e-01 -4.57428187e-01
-8.34532857e-01 1.04602069e-01 -5.93175650e-01 7.32720315e-01
-2.16822028e-01 3.87784451e-01 -4.34371352e-01 -5.35861552e-01
2.36847162e-01 8.26092660e-01 3.80236179e-01 -7.39695847e-01
4.24501210e-01 1.58069432e+00 7.43561029e-01 -1.63537025e-01
8.14734519e-01 3.30521017e-01 -2.94777781e-01 -1.22295725e+00
-2.14547291e-01 -5.45519769e-01 -8.42856348e-01 -5.57552099e-01
5.23786545e-01 -6.08119488e-01 -8.08201551e-01 1.68052828e+00
-1.55294871e+00 -5.20335793e-01 -2.91642368e-01 9.19161201e-01
-6.27184808e-01 4.09504235e-01 -4.73792493e-01 -1.03002763e+00
-3.98908883e-01 -1.03241479e+00 8.08798432e-01 3.60217303e-01
-2.36053899e-01 -3.80848557e-01 2.81132877e-01 2.08830595e-01
6.52753174e-01 -1.79586723e-01 5.84838986e-01 -3.06589548e-02
-6.55322075e-01 -3.56726140e-01 -4.64949012e-01 7.97441900e-01
5.41595876e-01 6.55861571e-02 -9.44808125e-01 1.79180000e-02
-4.69182581e-01 -3.38699758e-01 8.21222901e-01 5.02542615e-01
2.58079141e-01 -1.78664714e-01 -9.86401215e-02 5.45322597e-01
8.81031632e-01 6.27754569e-01 7.72404432e-01 4.15914170e-02
2.65862048e-01 7.14584649e-01 5.20043075e-02 -4.12374474e-02
5.51257074e-01 4.70258266e-01 -2.52597541e-01 -1.22548617e-01
-1.00183809e+00 -5.83011687e-01 4.12343651e-01 1.42715681e+00
-3.89724404e-01 -6.52636811e-02 -1.11855924e+00 6.34911478e-01
-1.53748930e+00 -9.68896806e-01 -1.90475404e-01 1.88599634e+00
9.27573144e-01 -3.62666368e-01 -8.41048285e-02 7.40190387e-01
5.54624438e-01 -2.20159113e-01 -2.47684240e-01 -3.08668137e-01
-4.28119153e-01 4.08286154e-01 -4.96282196e-03 9.07984793e-01
-8.84751439e-01 1.03833055e+00 7.52231932e+00 1.59627691e-01
-1.64952564e+00 -2.82197982e-01 -5.15568554e-01 3.00750673e-01
4.36505862e-02 -3.70398670e-01 -6.45936191e-01 2.50366420e-01
2.82696337e-02 -9.94580761e-02 3.93465191e-01 5.56691408e-01
5.46525955e-01 -1.61137015e-01 -8.80710721e-01 1.33790851e+00
2.92727388e-02 -1.05421948e+00 1.75660729e-01 -6.32632449e-02
4.96977538e-01 2.52888113e-01 -2.35479712e-01 2.17656881e-01
2.76896507e-01 -9.48684990e-01 8.71444821e-01 6.81630611e-01
1.11000025e+00 -1.40881035e-02 6.78914905e-01 2.06643060e-01
-1.26618826e+00 -7.44416416e-02 4.78601098e-01 -5.46695590e-01
8.11761081e-01 2.15620771e-01 -4.58455861e-01 -7.17688203e-02
5.44168651e-01 2.72161424e-01 1.77932128e-01 1.70938027e+00
-1.00911558e+00 4.40155417e-01 -3.85892153e-01 -4.99286473e-01
-1.28345296e-01 4.92590442e-02 5.93531132e-01 1.19412148e+00
5.38060308e-01 3.12685370e-01 -5.81380539e-02 3.68306607e-01
7.08670318e-02 3.89249623e-01 -5.94627440e-01 -1.53400108e-01
3.68319809e-01 3.43563855e-01 -2.23881707e-01 -2.04592720e-01
-6.80603743e-01 9.60222125e-01 -1.94701076e-01 5.14438629e-01
9.92636010e-03 -7.54805267e-01 1.07812047e+00 6.60817623e-02
-1.75932776e-02 -7.12545514e-01 -4.28534120e-01 -1.03307652e+00
5.79119921e-01 -7.10790455e-01 -1.04828291e-01 -7.51116216e-01
-1.12995577e+00 2.86757469e-01 -2.79847592e-01 -1.56234956e+00
-3.53012860e-01 -1.07764697e+00 -5.38028121e-01 9.91843462e-01
-1.90286410e+00 -1.75998318e+00 -5.07026732e-01 3.89805347e-01
4.80864257e-01 -1.96625397e-01 9.46122289e-01 3.47254008e-01
9.95675474e-02 5.66339970e-01 -9.60217975e-03 6.07478619e-01
4.32141393e-01 -7.87971079e-01 2.25171179e-01 8.62318754e-01
-1.03128124e-02 4.66180593e-01 5.48286617e-01 -4.90851730e-01
-1.02035975e+00 -3.67328614e-01 1.75988889e+00 6.78201914e-02
5.27071357e-01 -1.95188195e-01 -5.56037009e-01 3.21117699e-01
-2.80895401e-02 -2.14359269e-01 4.48135048e-01 -3.93654495e-01
-3.13682944e-01 8.82135630e-02 -1.06873667e+00 9.04563308e-01
1.56094873e+00 -8.07809353e-01 -9.04088140e-01 2.46660456e-01
3.89138043e-01 -2.84695387e-01 -2.07224175e-01 2.73475021e-01
1.39449942e+00 -6.00365162e-01 6.31874144e-01 -3.75446826e-01
-2.97923163e-02 -4.85731870e-01 -1.91414282e-01 -1.03730941e+00
1.24324933e-01 -8.53814602e-01 -5.14824921e-03 9.63174164e-01
1.83113158e-01 -9.85090554e-01 5.17482698e-01 6.17475510e-01
-1.92521125e-01 -4.21110302e-01 -1.16753066e+00 -1.03320742e+00
2.84962803e-01 -7.49124885e-01 4.12004709e-01 3.98421496e-01
4.09650058e-01 -1.25623763e-01 -3.46805602e-01 -8.90412629e-02
3.49054039e-01 -1.12363640e-02 8.97317410e-01 -1.29113781e+00
1.68909699e-01 -1.00122046e+00 -6.36980712e-01 -1.42580044e+00
-1.78964421e-01 -7.47578442e-01 3.54747653e-01 -2.44647694e+00
-3.17772985e-01 4.17930745e-02 4.95539233e-02 7.02452302e-01
4.45710152e-01 -1.35469036e-02 3.47592801e-01 2.09666997e-01
2.08110765e-01 4.84459609e-01 1.71897829e+00 -1.44307569e-01
-5.58543742e-01 4.31481004e-01 -6.20022535e-01 7.55075216e-01
9.71351385e-01 1.14473246e-01 -3.43236178e-02 -5.59547603e-01
-1.43499404e-01 -3.28287840e-01 2.58631766e-01 -8.15467119e-01
4.60816830e-01 -3.55892569e-01 -3.25155735e-01 -7.37413406e-01
6.53848499e-02 -4.47035193e-01 -4.91794407e-01 8.15053105e-01
-3.64800096e-02 -5.38495064e-01 -1.80861861e-01 -1.73348114e-01
-7.22113311e-01 4.38378491e-02 9.83867824e-01 -8.96479413e-02
-1.17780185e+00 -7.00673461e-02 -1.06507540e+00 -2.91393995e-01
8.16556633e-01 -5.18795967e-01 -1.17655344e-01 -7.49031782e-01
-7.67065942e-01 2.74309695e-01 -4.70721833e-02 6.63275301e-01
9.16103363e-01 -1.29686427e+00 -8.21530581e-01 7.14752555e-01
3.62472475e-01 -1.40309900e-01 2.48756446e-02 6.64162397e-01
-9.47108686e-01 7.08149970e-01 -4.10574317e-01 1.17817079e-03
-1.53797090e+00 -4.15905952e-01 5.84092975e-01 2.37222150e-01
-7.70264506e-01 1.08383858e+00 -2.48615876e-01 -6.69021606e-01
7.32398927e-01 -7.41851032e-01 -3.58276308e-01 -1.59354016e-01
9.56949532e-01 2.49969646e-01 -1.60728723e-01 -7.94868648e-01
-5.12992620e-01 9.49771702e-01 3.31210613e-01 -5.58250129e-01
1.08614612e+00 -8.10763910e-02 -3.93399119e-01 3.17240477e-01
9.20849264e-01 1.04737274e-01 -6.85335934e-01 -1.60324737e-01
-7.89790675e-02 -3.62603813e-01 -7.51811266e-02 -1.08693349e+00
-7.95214117e-01 1.08116484e+00 7.66432643e-01 -2.73902029e-01
1.15187252e+00 2.11160600e-01 1.02931213e+00 5.78014433e-01
7.83955514e-01 -1.32123971e+00 -9.03892741e-02 8.22560728e-01
1.78868973e+00 -1.15014303e+00 -5.50122857e-01 -4.41486895e-01
-2.57499486e-01 1.26522613e+00 4.49753672e-01 2.44942248e-01
1.15402544e+00 3.83274674e-01 9.62748289e-01 1.07833005e-01
-8.76036100e-03 -8.09334815e-01 3.57895523e-01 1.22025406e+00
8.37284088e-01 2.58441359e-01 -9.55759346e-01 3.81198764e-01
-4.98172671e-01 6.47316217e-01 1.73949108e-01 1.35817730e+00
-8.37769091e-01 -1.55317974e+00 -4.88600194e-01 2.47065380e-01
2.38804564e-01 1.31433576e-01 -5.00925839e-01 6.97907567e-01
3.06266218e-01 1.02410436e+00 -3.53031427e-01 -2.22834080e-01
6.87297285e-01 3.04142773e-01 5.95318556e-01 -3.77370983e-01
8.91080499e-02 -1.94735736e-01 1.48577243e-01 -1.52934223e-01
-4.10946131e-01 -6.54824317e-01 -1.55133152e+00 -3.04518286e-02
6.12258650e-02 -6.00225270e-01 8.01538527e-01 1.09748578e+00
-1.18447468e-01 1.78202331e-01 1.25756711e-01 -5.46427190e-01
-5.16529739e-01 -1.07389390e+00 -8.24348092e-01 -3.49338017e-02
7.70097017e-01 -5.06674767e-01 -3.39486897e-01 4.73381318e-02]
|
[9.100822448730469, -6.411412715911865]
|
ae7a7753-6b6a-467c-a056-80d6141f6eee
|
cm3-a-causal-masked-multimodal-model-of-the
|
2201.07520
| null |
https://arxiv.org/abs/2201.07520v1
|
https://arxiv.org/pdf/2201.07520v1.pdf
|
CM3: A Causal Masked Multimodal Model of the Internet
|
We introduce CM3, a family of causally masked generative models trained over a large corpus of structured multi-modal documents that can contain both text and image tokens. Our new causally masked approach generates tokens left to right while also masking out a small number of long token spans that are generated at the end of the string, instead of their original positions. The casual masking object provides a type of hybrid of the more common causal and masked language models, by enabling full generative modeling while also providing bidirectional context when generating the masked spans. We train causally masked language-image models on large-scale web and Wikipedia articles, where each document contains all of the text, hypertext markup, hyperlinks, and image tokens (from a VQVAE-GAN), provided in the order they appear in the original HTML source (before masking). The resulting CM3 models can generate rich structured, multi-modal outputs while conditioning on arbitrary masked document contexts, and thereby implicitly learn a wide range of text, image, and cross modal tasks. They can be prompted to recover, in a zero-shot fashion, the functionality of models such as DALL-E, GENRE, and HTLM. We set the new state-of-the-art in zero-shot summarization, entity linking, and entity disambiguation while maintaining competitive performance in the fine-tuning setting. We can generate images unconditionally, conditioned on text (like DALL-E) and do captioning all in a zero-shot setting with a single model.
|
['Luke Zettlemoyer', 'Mike Lewis', 'Gargi Ghosh', 'Mandar Joshi', 'Dmytro Okhonko', 'Naman Goyal', 'Hu Xu', 'Vladimir Karpukhin', 'Candace Ross', 'Bernie Huang', 'Armen Aghajanyan']
|
2022-01-19
| null | null | null | null |
['entity-disambiguation']
|
['natural-language-processing']
|
[ 6.77669704e-01 7.11996853e-01 -3.35838675e-01 -8.57481435e-02
-1.33138943e+00 -9.46843505e-01 1.40792644e+00 -4.94441539e-02
-1.11369535e-01 8.38324666e-01 9.40198123e-01 -2.26506457e-01
3.38960111e-01 -9.22395289e-01 -1.35557878e+00 -6.88657403e-01
8.45967010e-02 6.84952855e-01 3.28018889e-02 -7.91761428e-02
-1.75849959e-01 -1.40134752e-01 -1.44383800e+00 8.39527845e-01
7.48611033e-01 4.84203905e-01 3.39988202e-01 9.02739286e-01
-3.04851592e-01 8.72105956e-01 -7.53219306e-01 -8.06972980e-01
4.90981564e-02 -6.79232478e-01 -5.65607846e-01 3.62781078e-01
8.47427964e-01 -3.86986464e-01 -5.21714270e-01 5.67099750e-01
4.96100336e-01 -2.34878436e-01 8.32985342e-01 -1.17508245e+00
-1.28278649e+00 1.38301849e+00 -8.13937962e-01 -1.40768364e-01
4.09681320e-01 4.52695400e-01 1.23761010e+00 -8.67999077e-01
1.22999716e+00 1.46681416e+00 1.46379381e-01 6.94314897e-01
-1.51987457e+00 -4.65480775e-01 2.59872854e-01 -1.65914521e-01
-8.69176149e-01 -3.73143166e-01 3.96477640e-01 -6.55193210e-01
8.64078939e-01 2.56839633e-01 1.91813290e-01 1.87675703e+00
-3.57302986e-02 7.92328775e-01 8.28063965e-01 -6.01861358e-01
-5.46292029e-02 3.44418250e-02 -1.60776496e-01 7.97535956e-01
4.34433967e-02 -1.23558030e-01 -8.38302493e-01 -2.27664560e-01
4.71041322e-01 -2.58153915e-01 -2.74702489e-01 -1.51488379e-01
-1.58588517e+00 8.96938205e-01 2.27405280e-01 1.52814195e-01
-2.46428564e-01 4.28635538e-01 1.46114305e-01 -1.11905426e-01
4.80457038e-01 3.41911912e-01 -1.30887449e-01 3.11878473e-01
-1.21276021e+00 3.12702030e-01 7.63306201e-01 1.24800611e+00
7.48869002e-01 1.47304431e-01 -9.76825535e-01 4.86112744e-01
-1.48277041e-02 6.89954996e-01 2.71025062e-01 -7.37187326e-01
7.90672243e-01 3.12682748e-01 3.29514332e-02 -3.76807153e-01
2.11826146e-01 -4.57648009e-01 -8.69134128e-01 -2.91830301e-02
2.92717908e-02 -3.42983127e-01 -1.51775229e+00 2.17794180e+00
1.59825608e-01 1.58899903e-01 1.84575215e-01 4.41454083e-01
8.43601644e-01 1.06181550e+00 2.36614078e-01 -8.05943906e-02
1.59411144e+00 -7.83828259e-01 -5.49944162e-01 -4.13163722e-01
2.69707233e-01 -7.29299307e-01 1.19796109e+00 5.92356771e-02
-1.25291276e+00 -3.87526155e-01 -9.55039859e-01 -4.84598905e-01
-3.82787466e-01 3.48446704e-02 1.75351098e-01 4.55467179e-02
-1.03937697e+00 3.38776499e-01 -4.31658059e-01 -1.72816277e-01
3.97067815e-01 -2.36683071e-01 -3.72937024e-01 -2.91210055e-01
-1.36991215e+00 5.87230086e-01 3.04709017e-01 -3.45475733e-01
-1.23734283e+00 -8.36009860e-01 -9.54547882e-01 1.58146992e-01
5.06283641e-01 -1.32268262e+00 1.14175558e+00 -7.39723623e-01
-7.44838774e-01 1.01370323e+00 -3.31727117e-01 -6.96062863e-01
7.21948624e-01 -2.39505649e-01 -2.68490851e-01 2.98796922e-01
6.95667028e-01 1.20861793e+00 1.14253211e+00 -1.59661925e+00
-3.80806565e-01 -2.26984844e-01 -1.55720543e-02 1.39936060e-01
-6.07783273e-02 -1.80630997e-01 -6.00635529e-01 -8.93254042e-01
-5.66707194e-01 -7.18544543e-01 -9.13766921e-02 -3.43268573e-01
-1.02835894e+00 -2.94379629e-02 7.46352971e-01 -7.65003502e-01
1.01130390e+00 -2.08559918e+00 4.14665759e-01 -1.49002448e-01
1.12208977e-01 -2.54445761e-01 -3.83199364e-01 6.52487397e-01
-1.18971623e-01 7.11407900e-01 -4.75409538e-01 -5.97802758e-01
2.62108564e-01 3.07035297e-01 -9.10176516e-01 -6.28577098e-02
4.90568101e-01 1.36622739e+00 -9.18553531e-01 -7.21228540e-01
-6.41346797e-02 5.24481714e-01 -5.18770456e-01 1.43537596e-01
-9.33137476e-01 1.59537092e-01 -1.22450352e-01 2.40983576e-01
1.82380214e-01 -4.93845046e-01 7.81604126e-02 -2.47849002e-01
-2.58105285e-02 2.36661166e-01 -1.04080796e+00 1.81927729e+00
-6.04534626e-01 6.67802036e-01 -1.76771328e-01 -3.72893602e-01
3.57018709e-01 5.85027695e-01 2.23662391e-01 -3.87550563e-01
-3.23767185e-01 -2.08504587e-01 -4.96401191e-01 -4.32473570e-01
7.22239554e-01 -1.23996831e-01 -5.11308372e-01 6.22921526e-01
4.41058040e-01 -3.05544119e-02 4.66454983e-01 1.10362411e+00
1.14377832e+00 2.53583997e-01 -2.10149819e-03 2.32633233e-01
-1.62085950e-01 4.93911393e-02 -1.78041379e-03 1.04393780e+00
7.76768446e-01 1.03182709e+00 8.79819214e-01 2.69743770e-01
-1.58373737e+00 -1.33124232e+00 2.35481903e-01 1.16311562e+00
-1.93228662e-01 -4.82459575e-01 -6.73083544e-01 -6.15498781e-01
4.32368033e-02 1.26175559e+00 -9.67739105e-01 -5.02972007e-02
-5.36030352e-01 -5.60809314e-01 7.75133252e-01 4.67721850e-01
1.78476319e-01 -1.07220829e+00 -1.59489512e-01 1.26251534e-01
-5.46248078e-01 -1.24215913e+00 -8.03267360e-01 2.42515311e-01
-3.08714092e-01 -7.59671509e-01 -7.92225122e-01 -5.86061776e-01
5.78473568e-01 1.23814149e-02 1.46034610e+00 -3.09882373e-01
-4.22437549e-01 4.20107633e-01 -1.10431902e-01 -2.33321041e-01
-8.04473460e-01 9.50901434e-02 -5.22962093e-01 1.64561912e-01
-3.08434844e-01 -5.04035354e-01 -3.88036400e-01 -1.82950437e-01
-1.40339410e+00 3.92639995e-01 7.87827730e-01 9.70752954e-01
6.71435237e-01 -4.22542989e-01 3.70924979e-01 -1.26853406e+00
3.65096927e-01 -8.12511027e-01 -2.39200771e-01 3.86903912e-01
-3.16053689e-01 5.42074978e-01 3.99180621e-01 -6.24185145e-01
-1.29957843e+00 -1.32559478e-01 1.55051827e-01 -5.00388682e-01
-1.81177124e-01 3.92362833e-01 -4.09761280e-01 8.75972450e-01
8.65219593e-01 3.32115531e-01 -3.65997791e-01 -5.21495879e-01
1.30717874e+00 3.46160203e-01 1.24842393e+00 -6.44091189e-01
1.03629804e+00 6.00681722e-01 -1.16586946e-01 -5.00117779e-01
-1.02263951e+00 -2.01986730e-01 -3.45365644e-01 8.37383419e-02
1.10570598e+00 -1.20462883e+00 3.01868338e-02 1.90061480e-01
-1.37580585e+00 -3.42743814e-01 -7.12593734e-01 -1.82599887e-01
-4.25395042e-01 1.81895986e-01 -5.65956891e-01 -3.80123317e-01
-4.29377109e-01 -7.17081189e-01 1.46803331e+00 -9.28570926e-02
-2.78371751e-01 -7.06744611e-01 -1.50737256e-01 2.38685459e-01
1.99749898e-02 4.72349107e-01 1.36473787e+00 -5.49934447e-01
-8.71097684e-01 -1.66557953e-02 -1.75383225e-01 3.95900160e-02
-2.60407567e-01 2.37612054e-01 -9.23444211e-01 7.07878396e-02
-6.06708646e-01 -3.87368023e-01 1.39075077e+00 2.60177851e-01
8.61131012e-01 -9.17596042e-01 -3.68138790e-01 4.12982762e-01
1.39259422e+00 -2.79216141e-01 7.01386392e-01 2.14239284e-02
8.68990183e-01 4.57143724e-01 9.18543711e-03 3.04165453e-01
3.66487563e-01 3.52769285e-01 6.44593537e-01 -3.02966595e-01
-6.24511302e-01 -8.99209321e-01 4.89983588e-01 3.12417835e-01
3.59067827e-01 -8.91615331e-01 -5.77514231e-01 9.34503973e-01
-1.75664377e+00 -1.47179747e+00 -2.03834489e-01 1.80738127e+00
1.26850808e+00 1.70590112e-03 -5.96482940e-02 -4.38923448e-01
8.56854439e-01 5.78765154e-01 -4.62057292e-01 -1.23860784e-01
-6.89945102e-01 2.17001766e-01 5.34729123e-01 6.15230858e-01
-9.77822006e-01 9.97044146e-01 6.17998648e+00 8.66923749e-01
-8.94601345e-01 3.59519213e-01 6.75857842e-01 -5.69940031e-01
-1.15434384e+00 2.05418453e-01 -9.97072160e-01 5.38420081e-01
8.63610864e-01 -1.26086593e-01 3.17701489e-01 6.04052842e-01
5.83630754e-03 -4.03810255e-02 -1.19123995e+00 5.43230891e-01
2.36013696e-01 -1.88515818e+00 6.39566362e-01 1.59016758e-01
1.11176860e+00 -2.57925224e-02 2.17940420e-01 2.02845439e-01
8.88864756e-01 -1.09837782e+00 1.13490903e+00 3.91962707e-01
1.31404269e+00 -3.35165709e-01 2.23777160e-01 3.32540154e-01
-6.79388940e-01 9.69012231e-02 4.45668548e-02 4.26158041e-01
4.84491080e-01 8.41335833e-01 -8.76017213e-01 5.77154815e-01
3.62767875e-01 4.10071820e-01 -4.94606614e-01 3.47522885e-01
-5.99490881e-01 7.20052242e-01 -1.56040341e-01 1.80862695e-01
3.14192742e-01 2.43799865e-01 7.83419073e-01 1.46094298e+00
4.27671134e-01 3.65994424e-02 -4.08382267e-02 1.31298554e+00
-6.59093082e-01 -3.43978405e-01 -7.38892198e-01 -2.37921193e-01
5.07675886e-01 1.21833539e+00 -3.78092527e-01 -6.59547985e-01
-3.78317803e-01 1.10664475e+00 5.77377640e-02 6.02500200e-01
-9.22010481e-01 -5.47825657e-02 2.83837110e-01 3.92784625e-01
5.30963838e-01 -3.00376099e-02 -3.06764841e-01 -1.23630238e+00
5.47654815e-02 -8.41759682e-01 4.87780601e-01 -1.36811399e+00
-1.31746244e+00 5.59975386e-01 2.40876198e-01 -7.29870319e-01
-6.81741953e-01 -6.68562949e-02 -6.08409584e-01 9.53610003e-01
-1.28264058e+00 -1.51278174e+00 -1.72744483e-01 5.64516068e-01
6.91902518e-01 9.49597731e-02 6.84395790e-01 3.07840910e-02
-3.57768983e-01 3.55380118e-01 -1.17427386e-01 2.83092111e-01
1.01565731e+00 -1.41752326e+00 6.73964441e-01 1.31540847e+00
5.14889479e-01 5.13011217e-01 8.65079105e-01 -8.62567544e-01
-1.08173752e+00 -1.43174148e+00 9.10157502e-01 -5.84657967e-01
8.00451159e-01 -7.46162176e-01 -7.31668174e-01 1.02252829e+00
8.90309393e-01 -4.87999260e-01 3.56334418e-01 -1.84344798e-01
-7.36998737e-01 2.68909097e-01 -6.62890196e-01 8.62649918e-01
1.03852761e+00 -5.42938769e-01 -7.51645923e-01 4.46043849e-01
1.29114544e+00 -3.23860526e-01 -4.13082480e-01 6.19495399e-02
1.73944294e-01 -6.49861813e-01 9.75604355e-01 -6.45953476e-01
1.26443851e+00 -1.24932833e-01 -1.05085604e-01 -1.32101142e+00
-1.78359240e-01 -9.50830698e-01 -3.93425584e-01 1.71869254e+00
7.55474448e-01 -2.05095753e-01 4.98414606e-01 2.61708766e-01
-2.43504435e-01 -3.97867799e-01 -7.50587106e-01 -5.06443679e-01
4.34813499e-02 -2.57254660e-01 5.83715141e-01 7.22328961e-01
-3.80631834e-01 7.85181880e-01 -5.55039585e-01 5.68563826e-02
7.97674775e-01 2.77561992e-01 6.90212905e-01 -7.63858795e-01
-6.40742719e-01 -2.97888041e-01 2.55458564e-01 -9.55573082e-01
1.28077552e-01 -1.03648365e+00 1.49365291e-01 -1.93880379e+00
5.39175749e-01 2.26595365e-02 2.44758904e-01 8.44759941e-01
-3.74921381e-01 3.09285820e-01 2.16635495e-01 2.53797531e-01
-2.83660650e-01 4.19420302e-01 1.18659890e+00 -4.57677424e-01
2.66048968e-01 -4.88660306e-01 -1.07149136e+00 4.47229594e-01
2.98056632e-01 -5.17182112e-01 -4.46002096e-01 -6.91773474e-01
3.73490065e-01 2.22643867e-01 9.09958363e-01 -4.84107047e-01
1.02199152e-01 -8.56335238e-02 5.94986916e-01 -6.66150868e-01
2.78793663e-01 -1.84300229e-01 2.64034063e-01 -7.80959800e-02
-7.45945990e-01 -8.43739063e-02 1.33606875e-02 7.57990181e-01
-8.48187730e-02 5.17414771e-02 4.71782655e-01 -5.35567880e-01
-5.73617697e-01 2.20601529e-01 -1.31689176e-01 4.55758572e-01
7.88583159e-01 3.70303728e-02 -8.72554481e-01 -5.89732170e-01
-8.32955718e-01 2.24862814e-01 5.31518996e-01 5.43540835e-01
3.19587618e-01 -1.15789843e+00 -9.68529463e-01 -3.29319015e-02
8.95525292e-02 2.17751086e-01 3.48974824e-01 3.60347390e-01
1.80538699e-01 3.08680564e-01 1.09719381e-01 -5.49239516e-01
-9.60481882e-01 7.38880754e-01 -1.20849237e-01 -5.59692264e-01
-6.07228160e-01 7.88176239e-01 5.75818360e-01 -2.04317681e-02
2.82379277e-02 5.40961586e-02 1.88343361e-01 3.90032113e-01
3.77191961e-01 -1.44570485e-01 -2.04217687e-01 -3.63701880e-01
1.11968219e-01 2.07255155e-01 2.01988116e-01 -7.86820471e-01
1.23123169e+00 -1.70274332e-01 -6.81351945e-02 4.14740294e-01
1.03045273e+00 3.86144549e-01 -1.37121558e+00 -2.47474894e-01
-3.69903117e-01 -4.44902405e-02 -1.34973630e-01 -1.15280461e+00
-7.46430278e-01 9.03458118e-01 -3.36694345e-02 8.27665851e-02
7.89538801e-01 7.10153222e-01 6.63111031e-01 -1.67006701e-01
-2.00411696e-02 -6.71124935e-01 3.68083745e-01 1.90761849e-01
1.00260842e+00 -9.20971453e-01 -1.00048319e-01 -3.50152552e-01
-9.26437855e-01 6.13929749e-01 3.99947703e-01 6.17229342e-02
-3.10574993e-02 6.09687150e-01 -2.53650129e-01 -9.81599689e-02
-1.14946079e+00 -2.61913180e-01 2.70694315e-01 6.30514801e-01
3.16181988e-01 3.27529944e-02 8.75284970e-02 6.18690610e-01
-2.94372320e-01 -4.20001030e-01 6.19751155e-01 6.51678145e-01
-2.59135723e-01 -9.73952234e-01 -3.96865070e-01 4.29223835e-01
-5.51948130e-01 -7.40289092e-01 -5.43542504e-01 8.12904477e-01
1.98220253e-01 6.84633195e-01 2.40729228e-01 2.36184634e-02
3.50118838e-02 4.91847664e-01 3.56185377e-01 -8.43260527e-01
-4.15810555e-01 2.93014944e-01 2.80592769e-01 -2.65943408e-01
-1.66172221e-01 -6.91615760e-01 -1.21459591e+00 -5.42872399e-02
8.23766738e-02 7.83579648e-02 5.77957034e-01 8.94557059e-01
6.83762848e-01 7.42939174e-01 3.59202683e-01 -7.42128015e-01
-4.42608833e-01 -9.96957302e-01 -1.71062812e-01 7.30867445e-01
3.79004627e-01 -2.38880947e-01 -2.84924835e-01 7.21306443e-01]
|
[11.152073860168457, 0.7152987122535706]
|
3b49e948-5e2c-4fef-a0f6-31298085474b
|
combining-attention-module-and-pixel-shuffle
|
2210.16836
| null |
https://arxiv.org/abs/2210.16836v1
|
https://arxiv.org/pdf/2210.16836v1.pdf
|
Combining Attention Module and Pixel Shuffle for License Plate Super-Resolution
|
The License Plate Recognition (LPR) field has made impressive advances in the last decade due to novel deep learning approaches combined with the increased availability of training data. However, it still has some open issues, especially when the data come from low-resolution (LR) and low-quality images/videos, as in surveillance systems. This work focuses on license plate (LP) reconstruction in LR and low-quality images. We present a Single-Image Super-Resolution (SISR) approach that extends the attention/transformer module concept by exploiting the capabilities of PixelShuffle layers and that has an improved loss function based on LPR predictions. For training the proposed architecture, we use synthetic images generated by applying heavy Gaussian noise in terms of Structural Similarity Index Measure (SSIM) to the original high-resolution (HR) images. In our experiments, the proposed method outperformed the baselines both quantitatively and qualitatively. The datasets we created for this work are publicly available to the research community at https://github.com/valfride/lpr-rsr/
|
['David Menotti', 'William Robson Schwartz', 'Jorge de A. Lambert', 'Rayson Laroca', 'Valfride Nascimento']
|
2022-10-30
| null | null | null | null |
['license-plate-recognition']
|
['computer-vision']
|
[ 2.31859952e-01 -4.98942435e-01 8.42330419e-03 -1.39143273e-01
-1.36703908e+00 -4.61998224e-01 6.00866377e-01 -6.50046766e-01
-1.76954314e-01 6.72432184e-01 2.64185339e-01 1.60706118e-01
4.91849445e-02 -7.14878798e-01 -9.62357998e-01 -6.76049709e-01
3.18108618e-01 1.44102126e-01 4.19349641e-01 -2.37713158e-01
4.45405215e-01 6.01796269e-01 -1.51844978e+00 5.62857032e-01
7.80859709e-01 9.13016975e-01 5.47042251e-01 3.90291005e-01
3.45633060e-01 1.00964940e+00 -1.98373422e-01 -5.96377015e-01
7.03337371e-01 -3.01095061e-02 -4.17235851e-01 1.71394140e-01
6.81285918e-01 -5.34001172e-01 -8.01616132e-01 1.13212538e+00
3.99065942e-01 2.37906307e-01 3.58454823e-01 -8.03190112e-01
-1.07562804e+00 1.51411250e-01 -8.78615141e-01 4.32362616e-01
2.70290405e-01 1.54216200e-01 8.51570427e-01 -1.27124965e+00
7.56861985e-01 1.10493982e+00 7.89727509e-01 4.27616656e-01
-1.09422410e+00 -7.37964749e-01 -2.38607764e-01 4.05465961e-01
-1.56598258e+00 -6.34044111e-01 7.74446666e-01 -3.25289011e-01
7.86635160e-01 2.32767295e-02 -5.14069945e-02 1.23466110e+00
1.13444664e-01 6.56771541e-01 1.29230297e+00 -1.68784052e-01
-8.84925425e-02 3.03598851e-01 -1.25281349e-01 4.61943209e-01
2.12699443e-01 2.36508176e-01 -2.61821806e-01 3.19313109e-01
1.18314087e+00 2.85478920e-01 -4.04049754e-01 1.08813025e-01
-9.52513337e-01 5.94527483e-01 4.11543906e-01 3.12866300e-01
-2.58325785e-01 -2.15100512e-01 -2.43297848e-03 4.58773375e-02
4.81554300e-01 3.55853200e-01 -1.29814669e-01 2.18914807e-01
-1.10707188e+00 5.89807779e-02 2.90425897e-01 7.89937198e-01
5.84124684e-01 2.82797039e-01 -8.07711333e-02 1.33915639e+00
2.56254196e-01 4.58425105e-01 5.24861515e-01 -1.08439171e+00
7.54843414e-01 1.55229032e-01 3.80369186e-01 -9.99017358e-01
-5.34230396e-02 -5.11178076e-01 -1.02490306e+00 3.31788272e-01
9.98709500e-02 2.07976118e-01 -8.09067667e-01 1.21684515e+00
-2.30323792e-01 3.72737765e-01 2.57593870e-01 1.16102767e+00
8.74950886e-01 9.56909537e-01 -2.16424108e-01 -4.83494066e-02
1.19708025e+00 -1.15199113e+00 -6.04515374e-01 -2.21991166e-01
1.86790153e-02 -7.83892095e-01 1.01610005e+00 3.69531959e-01
-1.10245633e+00 -7.62088001e-01 -1.12214279e+00 -1.93482354e-01
-2.42025390e-01 5.13845086e-01 -1.21723644e-01 3.74676675e-01
-1.10071087e+00 4.91221637e-01 -4.37902957e-01 -1.90629259e-01
7.44950294e-01 1.42346367e-01 -5.78820646e-01 -5.95414639e-01
-1.02257526e+00 8.21485400e-01 3.95903625e-02 7.48102292e-02
-1.03037345e+00 -6.62542582e-01 -6.14896119e-01 8.07457864e-02
3.19138139e-01 -1.49673015e-01 8.33731174e-01 -8.97344410e-01
-1.54468334e+00 8.45375717e-01 2.45817706e-01 -5.02291560e-01
4.71926838e-01 -2.69878834e-01 -5.28225720e-01 1.78458273e-01
-1.29583418e-01 4.62361962e-01 8.15161705e-01 -1.45637059e+00
-5.64255357e-01 -4.49288338e-01 4.45806719e-02 2.12206885e-01
5.90493120e-02 4.51934129e-01 -4.33429211e-01 -8.05063367e-01
-8.02582204e-02 -7.71055222e-01 -1.31755859e-01 -3.76176715e-01
-2.47421190e-01 2.82278746e-01 6.91565871e-01 -1.18385363e+00
5.99460304e-01 -2.34206581e+00 -1.91896483e-01 -3.38550538e-01
1.50178045e-01 5.71739972e-01 -2.24055052e-01 2.76838005e-01
-2.22670168e-01 1.16845384e-01 -2.43678853e-01 -4.23476338e-01
-3.10119867e-01 -2.71819919e-01 -5.97818613e-01 6.62070870e-01
2.27138489e-01 6.97506011e-01 -3.47682148e-01 -3.37710351e-01
3.75981927e-01 8.33543718e-01 -1.74252644e-01 3.32172632e-01
8.82472917e-02 5.79987764e-01 -2.78080136e-01 6.60143435e-01
1.06228268e+00 -3.09689850e-01 -2.73867637e-01 -3.28569740e-01
-3.99416983e-01 -1.31956056e-01 -9.27971482e-01 1.34319687e+00
-5.46395183e-01 9.21529770e-01 2.95707621e-02 -7.64055371e-01
1.26972377e+00 3.06982011e-01 4.14612919e-01 -1.11682379e+00
-1.42205536e-01 2.64531404e-01 -4.92935628e-01 -4.53182101e-01
4.91987526e-01 -1.84299514e-01 2.64520228e-01 5.82517870e-02
-1.24997362e-01 2.77634174e-01 1.36152087e-02 -1.14192426e-01
8.22219729e-01 1.91063896e-01 -4.41552252e-02 1.00017361e-01
9.69107985e-01 -1.81235492e-01 6.04563355e-01 5.04454434e-01
-6.58728257e-02 1.16409039e+00 9.46531519e-02 -4.82542872e-01
-1.60469067e+00 -1.08585536e+00 -2.23813474e-01 7.37924695e-01
2.51902521e-01 1.08378157e-01 -4.65082943e-01 -2.24960774e-01
-4.21659052e-01 5.91313958e-01 -3.24817508e-01 2.77264267e-01
-7.38863230e-01 -6.99919164e-01 5.00635505e-01 5.62100232e-01
1.03053880e+00 -9.85551775e-01 -1.07868593e-02 -3.86226736e-02
-2.95565516e-01 -1.62525344e+00 -3.68692040e-01 -4.95133400e-01
-8.03480268e-01 -8.09366763e-01 -9.88152385e-01 -7.96660244e-01
4.98596460e-01 6.00111008e-01 8.98594618e-01 -2.82458335e-01
-2.42651835e-01 1.53760090e-01 -4.85624194e-01 6.28231838e-02
-4.09138530e-01 -3.24908227e-01 -1.49730459e-01 4.05165255e-01
3.88411395e-02 -5.16476750e-01 -5.80925465e-01 5.09285212e-01
-1.11204863e+00 1.65828720e-01 1.09890032e+00 6.20716751e-01
7.77331114e-01 3.11969459e-01 5.40465951e-01 -5.05024850e-01
4.40051109e-01 -4.09738034e-01 -1.06695509e+00 1.58095986e-01
-3.65631312e-01 -2.26482540e-01 9.41960752e-01 -1.06960900e-01
-1.49205434e+00 -1.06218189e-01 -3.53131324e-01 -7.87760556e-01
-4.31797922e-01 3.18047032e-02 -1.98174387e-01 -2.41745874e-01
4.19176549e-01 6.78516984e-01 -8.74596760e-02 -7.30306506e-01
2.20714379e-02 8.97961080e-01 6.41202986e-01 -2.55240917e-01
9.44244921e-01 6.12288535e-01 -1.79975435e-01 -9.35415268e-01
-6.63379431e-01 -3.32408220e-01 -4.10953164e-01 -2.90912271e-01
7.79114306e-01 -1.22118771e+00 -5.81998527e-01 7.00559676e-01
-8.31896901e-01 -1.50295898e-01 -7.39336833e-02 5.46979070e-01
-6.34711862e-01 4.97675776e-01 -8.52136552e-01 -7.91140318e-01
-2.32480973e-01 -1.30135250e+00 1.06148839e+00 3.86304438e-01
8.09855878e-01 -5.49628258e-01 7.69100040e-02 1.00740552e+00
7.79953361e-01 1.16828434e-01 3.60261530e-01 -6.16070867e-01
-1.12375283e+00 -3.40160191e-01 -6.68489456e-01 7.38099277e-01
-2.20397875e-01 -1.62342027e-01 -9.85283017e-01 -3.10071379e-01
8.15088674e-02 -4.52564508e-01 1.05782580e+00 4.96472090e-01
1.17477238e+00 -4.38267142e-01 1.58944294e-01 7.81744421e-01
1.91943502e+00 1.60675570e-02 1.12565398e+00 5.99557400e-01
6.64471507e-01 3.98459464e-01 5.62583745e-01 3.84224415e-01
3.59489530e-01 1.02476263e+00 3.77510577e-01 -8.34128782e-02
-4.17436570e-01 -1.67949960e-01 7.03644335e-01 4.98698562e-01
-4.87986803e-01 -2.92273432e-01 -8.60714018e-01 2.98725039e-01
-1.57284248e+00 -1.25916648e+00 4.33936566e-02 2.32103610e+00
4.40871507e-01 -1.14440277e-01 -9.38155781e-03 -1.91703573e-01
9.81985152e-01 3.60891849e-01 -4.66187686e-01 -4.90898602e-02
-4.96705860e-01 -2.97053546e-01 6.92909420e-01 3.90653968e-01
-1.14830220e+00 8.32124949e-01 5.19844103e+00 9.06095743e-01
-1.09037912e+00 2.05700144e-01 9.40931082e-01 2.20106423e-01
-4.14802209e-02 -5.27050734e-01 -9.19432342e-01 5.11612535e-01
9.33589995e-01 -5.88678978e-02 6.05624497e-01 6.66034281e-01
2.83555180e-01 1.92662358e-01 -6.99230373e-01 1.17246985e+00
2.93134123e-01 -1.49168551e+00 2.98813470e-02 1.36269763e-01
8.21073174e-01 5.08654952e-01 7.00363636e-01 2.34430328e-01
4.24542725e-02 -1.24552679e+00 4.28934664e-01 7.18236268e-01
1.05831301e+00 -7.40010381e-01 1.11132145e+00 3.01396579e-01
-1.00566566e+00 -4.19871539e-01 -6.83106840e-01 4.14362013e-01
-9.85791534e-02 2.13382944e-01 -6.14072084e-01 5.91185033e-01
7.84229517e-01 1.04555416e+00 -6.48414850e-01 1.10576558e+00
1.37688920e-01 4.60833699e-01 -1.81707479e-02 6.06642962e-01
1.21725090e-01 -4.67274845e-01 7.28946924e-01 1.21129048e+00
4.87251580e-01 6.35100380e-02 -1.31640986e-01 1.11403418e+00
-3.42060059e-01 4.12315801e-02 -6.13698483e-01 2.15074912e-01
2.43699625e-01 1.33542287e+00 -4.13815260e-01 -1.45716239e-02
-8.60317528e-01 7.38607168e-01 -1.02807336e-01 4.79220778e-01
-9.33414221e-01 1.22049619e-02 4.14406717e-01 5.22893965e-01
5.83654284e-01 -1.27785578e-01 4.97748069e-02 -1.43813384e+00
1.76992908e-01 -8.20178390e-01 1.45836711e-01 -1.08072174e+00
-1.40899241e+00 8.14879239e-01 -2.00763240e-01 -1.65739489e+00
2.23258939e-02 -6.08710527e-01 -3.87309432e-01 8.40586126e-01
-1.88710523e+00 -1.19159126e+00 -3.19685847e-01 5.07923126e-01
9.37979996e-01 -4.42447275e-01 4.08645898e-01 5.60763717e-01
-7.36885428e-01 5.26655376e-01 6.21667802e-01 4.67095822e-01
6.66468322e-01 -7.34948039e-01 2.62793511e-01 1.11200738e+00
-1.22270904e-01 2.23843127e-01 6.23489022e-01 -3.72027934e-01
-1.29716980e+00 -1.49340498e+00 2.85133719e-01 -2.70803034e-01
4.06334847e-01 -1.02916077e-01 -1.08596551e+00 7.53010631e-01
1.52828306e-01 2.92261422e-01 5.07625163e-01 -6.06976926e-01
-5.74873209e-01 -4.21721816e-01 -1.40336561e+00 3.60915244e-01
5.54390013e-01 -2.79259771e-01 -5.08039415e-01 1.34558752e-01
5.93119144e-01 -1.33843571e-01 -1.03928781e+00 4.19592679e-01
4.15483892e-01 -1.14003587e+00 1.27301979e+00 -6.65076077e-02
7.26161063e-01 -5.18688023e-01 -5.34960985e-01 -9.42061603e-01
-5.24804413e-01 -1.03537505e-02 1.95012629e-01 1.21727014e+00
3.49635184e-01 -4.99891192e-01 4.72064048e-01 5.38494349e-01
-1.39435560e-01 -5.60411811e-01 -8.19453061e-01 -9.23364341e-01
-4.60542440e-02 -1.02559030e-01 3.14962476e-01 6.11112177e-01
-5.88288486e-01 9.69296098e-02 -8.49267483e-01 4.64452922e-01
1.06877053e+00 1.30754843e-01 4.20883805e-01 -9.71730947e-01
-2.51055777e-01 -3.39637578e-01 -5.86558342e-01 -8.34295928e-01
3.17525193e-02 -6.98164463e-01 2.42510103e-02 -1.45286572e+00
4.80501831e-01 -5.44609316e-02 -3.32148433e-01 1.61310688e-01
1.71881422e-01 7.44117856e-01 4.90496814e-01 5.84030211e-01
-9.67576265e-01 5.45030713e-01 1.01836050e+00 -1.21674031e-01
-2.22218353e-02 1.48373052e-01 -5.22189438e-01 6.48085177e-01
1.02170682e+00 -2.07856357e-01 5.83551750e-02 -4.22601610e-01
1.42889917e-01 4.24967647e-01 4.42183942e-01 -1.20520210e+00
1.55166596e-01 1.97787464e-01 6.01158500e-01 -4.93344188e-01
5.43065071e-01 -7.55573511e-01 2.40703493e-01 5.70839979e-02
-5.08280218e-01 -1.38333082e-01 1.55049145e-01 6.35922790e-01
-4.30034995e-01 -2.47901119e-02 1.39315784e+00 -2.44071513e-01
-8.14062595e-01 3.99337500e-01 -1.90308243e-01 -1.08478129e-01
9.20429766e-01 -2.99324453e-01 -5.78027189e-01 -3.23596179e-01
-3.65252852e-01 -2.81582952e-01 7.23596811e-01 6.13253474e-01
1.07943511e+00 -1.28968394e+00 -1.25272322e+00 2.36175418e-01
9.01501477e-02 -1.97158366e-01 6.48535430e-01 7.36268699e-01
-6.69649959e-01 5.48048913e-01 -6.72079921e-01 -2.45838314e-01
-1.04733503e+00 6.82338953e-01 4.01137948e-01 -3.34713578e-01
-9.57428098e-01 4.88182873e-01 3.51760119e-01 -2.67069489e-01
5.47861643e-02 2.02924848e-01 -5.85884452e-01 -3.40268135e-01
1.00705934e+00 5.32949567e-01 1.50867570e-02 -1.12116170e+00
-1.79769367e-01 8.04830968e-01 -3.05515140e-01 1.62756711e-03
1.92584610e+00 -2.99187779e-01 1.11673050e-01 1.06883802e-01
1.20234513e+00 -1.71210784e-02 -1.50181532e+00 -3.99921775e-01
-2.55491406e-01 -9.48894441e-01 1.55487314e-01 -7.27970243e-01
-1.35106266e+00 6.83217764e-01 6.73902214e-01 -9.65080410e-02
1.19584584e+00 -2.46906132e-02 7.74281383e-01 2.53630221e-01
3.78239155e-01 -9.09717977e-01 1.27393097e-01 2.67586470e-01
1.15542579e+00 -1.46912742e+00 -1.79205552e-01 -1.34459242e-01
-8.95457625e-01 1.13313174e+00 4.39147234e-01 -5.35840809e-01
3.92093718e-01 3.06300789e-01 -1.06524281e-01 2.29148895e-01
-7.00473309e-01 -7.96594191e-03 8.76127556e-02 5.36216676e-01
4.30278242e-01 -1.68241307e-01 9.14058015e-02 5.98176003e-01
2.05356836e-01 3.82083505e-02 8.99083614e-01 2.20020846e-01
-4.34087873e-01 -7.59132624e-01 -6.38151348e-01 3.74229580e-01
-7.37870216e-01 -9.91013348e-02 1.13817751e-01 6.16132081e-01
-4.42418158e-02 8.43559980e-01 -2.37645097e-02 -3.49785358e-01
2.24527225e-01 -3.02150875e-01 1.99381039e-01 -2.43773326e-01
-7.84265324e-02 8.69127363e-02 -1.78757757e-01 -5.97822666e-01
-4.06981409e-01 -6.81373060e-01 -8.17368448e-01 -2.20664695e-01
-1.05091944e-01 -2.23892346e-01 7.17767477e-01 5.28278887e-01
4.17076916e-01 3.85564089e-01 9.35779989e-01 -9.76736188e-01
-4.78161395e-01 -9.05963838e-01 -8.17122936e-01 3.30853105e-01
4.35790330e-01 -4.27031606e-01 -3.56358767e-01 2.87849218e-01]
|
[11.108196258544922, -2.1879682540893555]
|
3e13ff0e-62f6-40d7-ba8f-4a265fc21114
|
sequence-to-backward-and-forward-sequences-a
|
1607.00970
| null |
http://arxiv.org/abs/1607.00970v2
|
http://arxiv.org/pdf/1607.00970v2.pdf
|
Sequence to Backward and Forward Sequences: A Content-Introducing Approach to Generative Short-Text Conversation
|
Using neural networks to generate replies in human-computer dialogue systems
is attracting increasing attention over the past few years. However, the
performance is not satisfactory: the neural network tends to generate safe,
universally relevant replies which carry little meaning. In this paper, we
propose a content-introducing approach to neural network-based generative
dialogue systems. We first use pointwise mutual information (PMI) to predict a
noun as a keyword, reflecting the main gist of the reply. We then propose
seq2BF, a "sequence to backward and forward sequences" model, which generates a
reply containing the given keyword. Experimental results show that our approach
significantly outperforms traditional sequence-to-sequence models in terms of
human evaluation and the entropy measure, and that the predicted keyword can
appear at an appropriate position in the reply.
|
['Zhi Jin', 'Lili Mou', 'Ge Li', 'Yiping Song', 'Rui Yan', 'Lu Zhang']
|
2016-07-04
|
sequence-to-backward-and-forward-sequences-a-2
|
https://aclanthology.org/C16-1316
|
https://aclanthology.org/C16-1316.pdf
|
coling-2016-12
|
['short-text-conversation']
|
['natural-language-processing']
|
[ 1.73573241e-01 6.15075409e-01 -1.80025715e-02 -5.07646203e-01
-5.13601661e-01 -5.01843810e-01 9.06071246e-01 -3.27904485e-02
-5.44043899e-01 1.16385102e+00 8.92982781e-01 -3.32091957e-01
8.94466490e-02 -8.37404191e-01 -2.17021137e-01 -4.00300831e-01
3.89878243e-01 6.28729999e-01 2.81618349e-02 -9.05731916e-01
5.63803852e-01 1.10235505e-01 -9.88199115e-01 6.35618985e-01
7.24998713e-01 4.68686968e-01 2.97323525e-01 8.95875573e-01
-4.65089738e-01 1.22948718e+00 -1.12354243e+00 -7.70300627e-01
-1.56778574e-01 -1.15734637e+00 -1.42852843e+00 -5.35065770e-01
-5.03108561e-01 -4.73303497e-01 -3.57391089e-01 7.96518207e-01
6.74232244e-01 3.27065170e-01 8.55711579e-01 -8.96828413e-01
-6.31750524e-01 1.28430283e+00 1.17144257e-01 -1.21620350e-01
9.40111578e-01 2.86482006e-01 1.27302778e+00 -7.45883763e-01
7.08864152e-01 1.47759366e+00 7.35752881e-01 9.90538061e-01
-1.00110865e+00 -1.95987955e-01 -3.16204160e-01 4.59645577e-02
-1.05027652e+00 -2.76922971e-01 7.59315729e-01 -1.04168564e-01
1.21909654e+00 5.10918796e-01 6.34987175e-01 1.13510430e+00
4.56811517e-01 9.96697247e-01 8.25533807e-01 -6.10350788e-01
8.08495060e-02 3.80167156e-01 -1.08362891e-01 4.36664373e-01
-4.21905577e-01 4.42418829e-02 -5.80168366e-01 -3.44883949e-01
4.58193272e-01 -5.31718493e-01 -3.61513764e-01 3.17814738e-01
-1.07243311e+00 1.42886019e+00 3.09365988e-01 5.54131329e-01
-6.61518395e-01 -7.64228478e-02 4.74943548e-01 3.00210476e-01
5.22512317e-01 7.98640668e-01 -4.98222113e-01 -5.21476328e-01
-6.63091600e-01 7.83544302e-01 1.62527359e+00 8.63021195e-01
4.87727314e-01 1.06513098e-01 -4.84867483e-01 9.38710928e-01
2.73225576e-01 2.23730192e-01 7.70993471e-01 -1.10531640e+00
1.73227534e-01 4.64270681e-01 2.56583899e-01 -1.15482748e+00
-3.20472091e-01 -3.77312899e-02 -7.60769725e-01 -3.40964347e-01
3.76187772e-01 -7.80647576e-01 -1.42909318e-01 1.73038220e+00
5.70097417e-02 -5.53962767e-01 5.37048817e-01 7.80209363e-01
1.14596915e+00 1.00833690e+00 -1.02976806e-01 -2.74464607e-01
8.81120741e-01 -7.35660374e-01 -1.00250137e+00 -1.12103261e-01
8.01734865e-01 -7.25984812e-01 9.72898245e-01 8.83698016e-02
-1.32996428e+00 -2.93847322e-01 -6.16994977e-01 6.29438683e-02
-1.23118915e-01 -2.09732980e-01 3.15342933e-01 4.77076501e-01
-1.20551097e+00 7.16454148e-01 5.34637384e-02 -2.03077763e-01
-3.78707975e-01 2.56088316e-01 -1.17567240e-03 3.92282099e-01
-1.70634031e+00 1.17821956e+00 8.31878185e-01 2.66779703e-03
-3.15073967e-01 -3.43603671e-01 -8.26913714e-01 5.08835204e-02
1.40492201e-01 -8.12643886e-01 1.72383070e+00 -8.63156021e-01
-1.94546449e+00 3.48167181e-01 -1.70936808e-01 -6.54440105e-01
5.23628354e-01 -2.42951401e-02 1.10199116e-02 -2.69715041e-02
-1.78927571e-01 1.11767638e+00 3.88457656e-01 -1.36036670e+00
-5.86601794e-01 9.70137715e-02 8.73778015e-02 4.78651553e-01
-3.04672476e-02 2.02568606e-01 7.81652853e-02 -7.14280725e-01
-3.34277958e-01 -8.79264534e-01 -3.16851079e-01 -7.03851342e-01
-6.55456543e-01 -8.70700181e-01 3.75773460e-01 -7.29222894e-01
1.36252594e+00 -1.68288946e+00 1.49569958e-01 1.83528483e-01
4.21396345e-02 3.48849237e-01 1.24997338e-02 9.05892134e-01
1.47287965e-01 1.64959550e-01 -2.23875269e-01 1.02484733e-01
3.35733406e-02 3.12989295e-01 -3.17570418e-01 -1.95707753e-01
9.84586626e-02 1.12831724e+00 -9.15579498e-01 -3.08957428e-01
-2.63283581e-01 3.05388093e-01 -7.41445184e-01 6.65922225e-01
-4.62744147e-01 2.23513603e-01 -2.29291648e-01 -2.40123540e-01
5.05415238e-02 -2.40982827e-02 3.20142448e-01 -2.06437651e-02
1.56295031e-01 7.47661471e-01 -4.66790646e-01 1.34698415e+00
-3.51825953e-01 6.99245036e-01 -2.90843993e-01 -4.55211848e-01
1.28468120e+00 5.12915611e-01 1.15305096e-01 -5.47335267e-01
3.18431526e-01 1.52653113e-01 3.51457775e-01 -5.39567530e-01
1.19973004e+00 -2.21669763e-01 -4.55798447e-01 9.66776371e-01
-1.81056373e-03 -4.18574214e-01 1.89735785e-01 5.50413072e-01
7.82741964e-01 8.66343677e-02 4.56456274e-01 -1.02898732e-01
5.52200317e-01 5.47214858e-02 2.82570750e-01 9.88863230e-01
7.08479360e-02 3.91053230e-01 7.04948545e-01 -3.56816947e-01
-1.11386549e+00 -5.28946579e-01 5.96000910e-01 1.15746057e+00
-1.96927890e-01 -2.86925763e-01 -1.18677068e+00 -7.30266213e-01
-3.74453992e-01 1.42087126e+00 -4.53003675e-01 -2.97179610e-01
-7.29448557e-01 -3.64267617e-01 8.90847147e-01 2.33273268e-01
3.85861725e-01 -1.81513596e+00 -4.80493724e-01 5.54029703e-01
-7.43399560e-01 -6.93824470e-01 -6.01478517e-01 6.83275312e-02
-6.08080685e-01 -4.88036007e-01 -9.45075929e-01 -7.80759871e-01
2.93720961e-01 -1.04743913e-01 1.23460233e+00 2.02489078e-01
2.24505633e-01 1.45534724e-01 -5.59744477e-01 -3.40008169e-01
-1.28173757e+00 1.77641734e-01 -1.54314071e-01 -1.79821357e-01
4.45910543e-01 -2.68942863e-01 -2.99628079e-01 2.23638728e-01
-9.16302323e-01 1.54950485e-01 2.91688353e-01 9.39141631e-01
-3.20499539e-01 -3.09707522e-01 9.07009423e-01 -8.76916528e-01
1.89626062e+00 -4.96607274e-01 1.41055092e-01 1.38605252e-01
-5.19838572e-01 3.32932830e-01 6.63310528e-01 -3.21385205e-01
-1.23628414e+00 -8.93146843e-02 -5.43839633e-01 3.94037098e-01
-1.89507440e-01 7.37792134e-01 8.89455080e-02 2.87302524e-01
1.03508198e+00 6.39282823e-01 3.38937670e-01 -2.37042427e-01
4.84401494e-01 9.10773098e-01 5.66926539e-01 -4.06169832e-01
2.46875718e-01 -3.40322196e-01 -6.14325345e-01 -7.17891514e-01
-5.53394675e-01 -2.05788761e-01 -3.33219260e-01 -4.51063663e-01
5.65773308e-01 -3.06171149e-01 -1.00106478e+00 2.86268532e-01
-1.84695971e+00 -2.51563996e-01 -2.02883080e-01 2.48308957e-01
-7.95414984e-01 5.10064542e-01 -9.28457201e-01 -1.08149827e+00
-8.35873425e-01 -9.50403392e-01 4.75013137e-01 4.00365651e-01
-1.02837074e+00 -9.92336214e-01 7.44644105e-02 1.20749615e-01
6.58702433e-01 -2.68866628e-01 1.09501886e+00 -1.29446793e+00
2.30660215e-02 -1.37916774e-01 -6.54119775e-02 3.61449361e-01
-3.89922708e-02 -3.06384563e-01 -4.72626895e-01 2.74600118e-01
1.77107349e-01 -3.48259211e-01 4.36937749e-01 1.93237707e-01
5.70200801e-01 -1.18163466e+00 6.94756582e-02 -1.47373825e-01
8.20572555e-01 5.60897946e-01 8.16296756e-01 -2.22869664e-02
1.74452677e-01 1.20612168e+00 3.86868238e-01 7.26505280e-01
6.34550452e-01 5.77026844e-01 7.81912580e-02 1.76686600e-01
9.99002829e-02 -5.05707204e-01 3.67910355e-01 1.01902688e+00
2.32531458e-01 -7.76067376e-01 -7.27511525e-01 5.39652526e-01
-1.76773608e+00 -1.03971291e+00 -1.73128992e-01 1.67813110e+00
1.36104500e+00 -1.35905528e-03 2.71815002e-01 1.12597421e-02
6.76917076e-01 -8.62397850e-02 -1.47869736e-01 -8.55178177e-01
-6.82724267e-02 -1.55152483e-02 8.02256837e-02 7.97989011e-01
-4.06462848e-01 1.11263156e+00 6.96689892e+00 6.84941590e-01
-8.01212549e-01 -1.89673990e-01 6.14114761e-01 4.35605377e-01
-6.37513041e-01 -8.56125355e-02 -7.93165565e-01 5.42021036e-01
1.26073539e+00 -5.87431312e-01 3.62773091e-01 6.98921323e-01
3.53070796e-01 -1.03176668e-01 -9.14348602e-01 4.61858332e-01
2.92012423e-01 -1.38523602e+00 3.39441121e-01 -7.06930608e-02
3.46723169e-01 -4.08318549e-01 -2.67378509e-01 2.93798774e-01
7.72755265e-01 -1.00014770e+00 4.69677359e-01 7.11806774e-01
2.14297533e-01 -1.04820776e+00 9.62049842e-01 8.82893622e-01
-3.83172989e-01 2.13452518e-01 -3.21578920e-01 -1.05835930e-01
3.92270625e-01 5.23943454e-02 -1.75146377e+00 2.95913458e-01
1.84973419e-01 8.20816308e-02 -1.48127273e-01 7.82543302e-01
-4.34431314e-01 5.95884264e-01 -3.84799093e-02 -1.15429509e+00
5.52538157e-01 2.80417129e-02 6.32435560e-01 1.47955787e+00
1.98869884e-01 2.87224799e-01 -8.62223003e-03 9.63532209e-01
-4.47567813e-02 4.56586659e-01 -7.87377357e-01 -7.00363368e-02
4.71986294e-01 8.44784737e-01 -3.20907116e-01 -4.50096458e-01
1.67543650e-01 1.14195502e+00 -1.98086891e-02 1.44046485e-01
-5.27430475e-01 -6.93883061e-01 1.11076757e-02 -2.82125682e-01
2.75180675e-02 9.98309851e-02 -1.96178138e-01 -5.85338473e-01
-2.30209023e-01 -1.24921501e+00 5.76062985e-02 -9.51682925e-01
-1.23076725e+00 1.02553082e+00 1.22550331e-01 -7.29115486e-01
-1.28379822e+00 -1.08232923e-01 -5.29105604e-01 1.16130972e+00
-7.64419258e-01 -7.65894234e-01 1.14907488e-01 3.35839003e-01
7.66821861e-01 -3.65400732e-01 1.17405856e+00 -1.60603747e-01
-8.01720843e-02 6.53061032e-01 -2.10764199e-01 3.85548949e-01
4.73014921e-01 -1.05805361e+00 7.06292033e-01 2.38721833e-01
-3.22126932e-02 8.79002988e-01 1.11985517e+00 -8.47908914e-01
-9.35654163e-01 -6.18626058e-01 1.68635905e+00 -1.70195922e-01
4.52203184e-01 3.18311062e-03 -9.73665714e-01 2.88772494e-01
8.58016133e-01 -1.10471141e+00 7.18725204e-01 -2.38727912e-01
3.25010531e-02 6.32010698e-01 -1.16541040e+00 8.72117937e-01
4.19493616e-01 -3.21307510e-01 -9.47371423e-01 5.33297837e-01
9.45670187e-01 -3.10111016e-01 -6.47822440e-01 9.31360871e-02
4.95771706e-01 -9.54235077e-01 6.14262342e-01 -7.58475125e-01
6.86096191e-01 2.11724490e-01 1.71096865e-02 -1.67567432e+00
-1.88393027e-01 -1.04948485e+00 3.86100970e-02 9.93026257e-01
8.56349051e-01 -4.49366331e-01 8.14044952e-01 8.43606353e-01
-5.99549972e-02 -7.74355352e-01 -7.48459816e-01 -4.49776679e-01
2.58218437e-01 -1.03875950e-01 6.61178768e-01 6.81389213e-01
5.44572771e-01 8.39829862e-01 -9.08169091e-01 -7.35896051e-01
-2.31025815e-02 -2.67227173e-01 7.76916146e-01 -1.11280513e+00
-3.27464908e-01 -6.78994417e-01 1.07027225e-01 -1.46274376e+00
2.77497381e-01 -7.52662122e-01 6.26943767e-01 -1.51764345e+00
-2.99183149e-02 -4.09603044e-02 4.64820206e-01 2.17048973e-01
-2.18880489e-01 -1.41475350e-01 2.45310888e-01 1.09463362e-02
-2.36802667e-01 7.01662183e-01 9.84100997e-01 -1.32758245e-01
-5.21661937e-01 1.56631231e-01 -7.90658593e-01 6.13341331e-01
1.00666714e+00 -5.15123665e-01 -1.74555898e-01 -3.22934776e-03
4.38696116e-01 6.31720364e-01 -3.64665352e-02 -4.02431846e-01
3.79435420e-01 -1.41372919e-01 7.83333480e-02 -5.65987170e-01
3.53374362e-01 -4.03449178e-01 -5.61175160e-02 6.46288395e-01
-1.20639074e+00 3.31402361e-01 -1.47755042e-01 3.55705827e-01
-1.47675186e-01 -8.79283905e-01 4.32418823e-01 -4.92468774e-01
-2.82489806e-01 -3.00478399e-01 -9.67769921e-01 -1.12113805e-04
5.65709531e-01 -6.81256577e-02 7.74752796e-02 -1.30796623e+00
-3.22568029e-01 -3.29500474e-02 1.02144450e-01 5.63022792e-01
9.54470456e-01 -1.12302577e+00 -1.11627293e+00 -2.03932926e-01
-1.91191956e-01 -4.58346188e-01 -3.67882252e-02 2.14409575e-01
-4.79862005e-01 6.58994734e-01 1.94395259e-02 -1.55071169e-01
-1.17100239e+00 9.29932818e-02 3.49292099e-01 -3.64945680e-01
-4.71479386e-01 9.11572456e-01 -2.20727175e-01 -7.51139224e-01
2.91887462e-01 2.85838604e-01 -6.68229878e-01 -1.77235194e-02
6.10202909e-01 1.73267797e-01 -2.16104582e-01 -7.93386459e-01
1.64894670e-01 -4.04258877e-01 -3.29525799e-01 -6.43637598e-01
1.10471010e+00 -4.38450947e-02 -3.02482307e-01 3.13149780e-01
1.03627908e+00 -2.53433049e-01 -6.18043780e-01 -4.36278373e-01
1.97447360e-01 -2.25280717e-01 -3.11949432e-01 -1.05844796e+00
-5.37917614e-01 5.56524694e-01 1.89405754e-02 6.17850959e-01
6.06774986e-01 -2.15490401e-01 1.18292630e+00 8.58375430e-01
1.96199775e-01 -1.22290111e+00 1.51696205e-01 1.18066502e+00
1.10785770e+00 -7.65462160e-01 -3.96533668e-01 -3.00094578e-03
-1.18544316e+00 1.24959290e+00 6.49194360e-01 1.08437292e-01
1.19773045e-01 1.67410046e-01 1.66530013e-01 -3.48433964e-02
-1.04757011e+00 1.94672510e-01 8.13800320e-02 4.39465880e-01
8.15053225e-01 1.22545101e-02 -8.17019403e-01 3.44421953e-01
-7.74095356e-01 -1.28027648e-01 8.58600497e-01 8.74907494e-01
-6.33955896e-01 -1.23065400e+00 -1.84438989e-01 3.99393350e-01
-5.35746098e-01 -3.73874307e-01 -1.16607904e+00 3.59290570e-01
-5.14989614e-01 1.27632499e+00 -2.75547594e-01 -7.19692051e-01
1.58285201e-01 3.77110690e-01 8.06463733e-02 -5.63166857e-01
-1.33563709e+00 -1.80232674e-01 5.76683998e-01 -1.12849772e-01
-2.98216879e-01 -3.48315924e-01 -1.20084798e+00 -4.49663013e-01
-5.04505396e-01 7.30064273e-01 7.17540860e-01 1.03014123e+00
2.58594453e-01 2.21380666e-01 8.05323303e-01 -6.58497274e-01
-8.84244978e-01 -1.40056860e+00 -4.28587496e-02 2.24053025e-01
-1.05289459e-01 -2.10920274e-02 -1.92270800e-01 -1.45385966e-01]
|
[12.671626091003418, 8.323009490966797]
|
fff41ae3-f5b7-46e8-af34-b23f3eaa2e9d
|
trigger-gnn-a-trigger-based-graph-neural
|
2204.05518
| null |
https://arxiv.org/abs/2204.05518v2
|
https://arxiv.org/pdf/2204.05518v2.pdf
|
Trigger-GNN: A Trigger-Based Graph Neural Network for Nested Named Entity Recognition
|
Nested named entity recognition (NER) aims to identify the entity boundaries and recognize categories of the named entities in a complex hierarchical sentence. Some works have been done using character-level, word-level, or lexicon-level based models. However, such researches ignore the role of the complementary annotations. In this paper, we propose a trigger-based graph neural network (Trigger-GNN) to leverage the nested NER. It obtains the complementary annotation embeddings through entity trigger encoding and semantic matching, and tackle nested entity utilizing an efficient graph message passing architecture, aggregation-update mode. We posit that using entity triggers as external annotations can add in complementary supervision signals on the whole sentences. It helps the model to learn and generalize more efficiently and cost-effectively. Experiments show that the Trigger-GNN consistently outperforms the baselines on four public NER datasets, and it can effectively alleviate the nested NER.
|
['Liang Zhang', 'Wei Yan', 'Yingting Hu', 'Fanyang Bu', 'Yuan Sui']
|
2022-04-12
| null | null | null | null |
['nested-named-entity-recognition']
|
['natural-language-processing']
|
[-2.22548321e-01 3.45140755e-01 -1.37611836e-01 -4.49814230e-01
-4.22774762e-01 -6.35493696e-01 1.82832330e-01 5.41472852e-01
-7.36181915e-01 4.73721057e-01 8.18714321e-01 -1.61425203e-01
1.04964405e-01 -9.99491692e-01 -4.78749603e-01 -1.42970428e-01
-1.88207433e-01 2.23093748e-01 4.23375875e-01 -2.17618346e-01
-6.64182454e-02 1.71193376e-01 -8.80111516e-01 2.75733680e-01
9.64800835e-01 8.37758183e-01 3.58995982e-02 4.62166935e-01
-6.80850267e-01 1.18657696e+00 -5.84321737e-01 -8.63632023e-01
-3.63054536e-02 -2.51660347e-01 -9.57354724e-01 -3.70122373e-01
1.58153132e-01 -5.46840280e-02 -6.83330894e-01 1.04926932e+00
7.86327422e-01 3.18932265e-01 3.59459013e-01 -9.89049792e-01
-1.00152588e+00 1.17010200e+00 -2.25504890e-01 2.41395250e-01
3.76056850e-01 -1.23600028e-01 1.43352318e+00 -9.06826437e-01
8.11484933e-01 1.14197254e+00 1.04155529e+00 5.17110765e-01
-7.33928919e-01 -5.06240487e-01 3.86606336e-01 1.20908268e-01
-1.40723753e+00 -2.17580311e-02 7.79916108e-01 -5.59458211e-02
1.22684848e+00 -4.21925858e-02 3.69176835e-01 1.03824925e+00
-2.09635049e-02 9.29523647e-01 6.19223595e-01 -1.77084446e-01
6.36663437e-02 -4.03826982e-01 7.34607458e-01 9.56461787e-01
4.10029233e-01 -3.40679973e-01 -3.00953925e-01 -6.47334680e-02
4.18456554e-01 6.74065799e-02 -1.81386203e-01 8.39642510e-02
-9.45818603e-01 8.42056274e-01 8.73562038e-01 6.18633628e-01
-4.00993556e-01 1.26096189e-01 7.90497661e-01 1.23894460e-01
4.60137606e-01 6.02808654e-01 -8.60216439e-01 1.97994485e-01
-5.86348116e-01 -3.90470594e-01 1.16459966e+00 1.11527324e+00
7.40453601e-01 9.82452631e-02 -4.41508740e-01 6.60908163e-01
3.97924393e-01 8.56986120e-02 5.70302427e-01 -2.70118445e-01
7.29295492e-01 1.31819558e+00 -3.92123282e-01 -1.15759552e+00
-9.87684965e-01 -5.13561189e-01 -9.18047845e-01 -8.45276535e-01
-5.10017276e-02 -5.56872308e-01 -9.39483881e-01 1.83840811e+00
5.40692985e-01 6.03646398e-01 2.32660890e-01 5.91420054e-01
1.61128438e+00 5.67200005e-01 6.11937940e-01 7.42200762e-02
1.61967432e+00 -1.14511192e+00 -1.06578481e+00 -3.49821776e-01
1.34305704e+00 -3.59480083e-01 1.00031388e+00 -3.48729670e-01
-4.89937752e-01 -4.42909330e-01 -9.69498873e-01 -4.58675891e-01
-9.25150454e-01 3.12515676e-01 7.53099203e-01 5.74135780e-01
-9.65865016e-01 5.19040883e-01 -6.23878419e-01 -3.42456877e-01
2.71469831e-01 3.02830219e-01 -5.53632975e-01 4.44452763e-02
-1.89013505e+00 7.89433360e-01 1.06124544e+00 3.93233657e-01
-4.58656847e-01 -6.38603032e-01 -1.39414263e+00 3.00412953e-01
4.66543823e-01 -6.44890428e-01 9.12595987e-01 -3.46607566e-01
-1.06218922e+00 7.40582526e-01 -1.35162339e-01 -4.77946728e-01
-1.62582174e-01 -1.65219069e-01 -7.61722386e-01 -8.18643644e-02
2.51544565e-01 4.48985457e-01 -1.36803521e-03 -8.10251176e-01
-4.31970090e-01 -3.69290501e-01 2.85348862e-01 2.51828641e-01
-6.16076887e-01 1.39062390e-01 -4.85692978e-01 -6.45372629e-01
2.38339398e-02 -5.19168615e-01 -4.80750144e-01 -9.23992217e-01
-7.41621733e-01 -9.02416825e-01 5.57569981e-01 -8.91465247e-01
1.87475944e+00 -2.13959408e+00 -1.86513476e-02 -1.76023155e-01
4.57432985e-01 1.05459109e-01 -2.92872757e-01 8.02063644e-01
7.98086300e-02 6.16875112e-01 -6.67431727e-02 -4.53847915e-01
2.15006188e-01 2.55210608e-01 -1.72554612e-01 2.58233361e-02
4.61964846e-01 1.29069197e+00 -9.90006506e-01 -7.50237703e-01
-3.06296706e-01 4.53698575e-01 -4.20347750e-01 3.41881365e-01
-1.00839011e-01 1.61651179e-01 -5.07335305e-01 4.52485234e-01
4.49194849e-01 -5.03438711e-01 4.91161227e-01 -4.54965472e-01
6.17270991e-02 5.60811937e-01 -1.22859442e+00 1.74625599e+00
-3.49260837e-01 3.03764373e-01 -2.83321351e-01 -6.99037492e-01
9.76216137e-01 4.74210918e-01 5.99305630e-02 -5.25404572e-01
1.29113600e-01 1.31757483e-01 -3.14036131e-01 -5.26413321e-01
6.48371637e-01 1.95785463e-01 -5.94504297e-01 1.20764129e-01
4.04813737e-01 5.40076792e-01 3.19507688e-01 4.96499300e-01
1.57483828e+00 -9.13188532e-02 4.35784668e-01 6.94412217e-02
5.80918014e-01 -1.33299097e-01 1.01751494e+00 6.28194928e-01
-1.69511452e-01 8.78063962e-02 5.66878617e-01 -3.59586209e-01
-6.42774403e-01 -8.37277055e-01 1.51184618e-01 1.41074157e+00
1.98421091e-01 -7.21548617e-01 -8.20611715e-01 -1.29771233e+00
-1.45600170e-01 7.02991724e-01 -6.04047358e-01 -2.62818277e-01
-9.36784983e-01 -5.89680791e-01 1.05459440e+00 9.35663521e-01
6.94799304e-01 -1.16560292e+00 2.59958595e-01 4.89560872e-01
-3.02416831e-01 -1.34587252e+00 -7.34570324e-01 2.82909900e-01
-7.90855408e-01 -9.61218357e-01 -1.30267531e-01 -1.18834758e+00
6.15573943e-01 -1.56536043e-01 1.20953619e+00 1.29003510e-01
1.03946039e-02 3.17796588e-01 -5.89361548e-01 -1.05011687e-01
-4.03491706e-02 6.39270186e-01 -1.18828043e-01 -2.20524389e-02
6.14197016e-01 -3.98536891e-01 -4.44946289e-01 1.55845493e-01
-8.79378915e-01 -2.47897789e-01 5.55059373e-01 8.57354283e-01
4.54750389e-01 -5.23379073e-02 8.54417324e-01 -1.21949363e+00
7.66186774e-01 -7.70515263e-01 -2.79129207e-01 4.85763878e-01
-5.23177683e-01 3.08348596e-01 9.90984797e-01 -2.43429825e-01
-1.15780795e+00 -3.68767194e-02 -3.05651546e-01 -1.04348451e-01
-2.80251384e-01 9.73535299e-01 -6.16600811e-01 9.84390974e-02
3.52811098e-01 6.76392168e-02 -1.00424433e+00 -7.01512218e-01
7.86360443e-01 6.44483685e-01 4.46163863e-01 -4.52416837e-01
6.23370588e-01 1.24421842e-01 -2.16740623e-01 -5.34354389e-01
-1.35254109e+00 -7.75020003e-01 -8.13938975e-01 1.07266955e-01
1.26109684e+00 -9.90130723e-01 -5.05695701e-01 4.70541157e-02
-1.55214906e+00 -4.60870638e-02 -1.14428647e-01 3.42464417e-01
1.08570606e-01 4.99369889e-01 -1.03776407e+00 -5.24018347e-01
-5.33871472e-01 -6.60722673e-01 1.01151848e+00 6.20627820e-01
9.37559307e-02 -1.43026292e+00 1.71485350e-01 3.16752166e-01
1.86016396e-01 1.86818123e-01 8.23234499e-01 -1.57135201e+00
-4.49553043e-01 -3.36018920e-01 -4.10438508e-01 1.10211805e-01
4.99351285e-02 -4.70420361e-01 -8.36987317e-01 -1.81656986e-01
-3.16935301e-01 -2.01313227e-01 9.51544106e-01 -1.17208079e-01
7.34075904e-01 -5.31503797e-01 -4.84543115e-01 6.35545254e-01
1.46480000e+00 6.24931837e-03 4.52706307e-01 1.66296139e-01
1.20558572e+00 5.33402681e-01 2.20988974e-01 2.55047828e-01
1.05883598e+00 1.33021593e-01 2.03828856e-01 -1.61858559e-01
-2.57685512e-01 -7.28789330e-01 2.30684668e-01 1.57695293e+00
1.88150600e-01 -5.04712701e-01 -1.11898112e+00 7.02256918e-01
-1.79177856e+00 -7.01038957e-01 -2.91037738e-01 1.54614592e+00
8.73466313e-01 2.65403222e-02 -2.96800911e-01 -2.99265832e-01
1.07922208e+00 3.29654217e-01 -5.18111348e-01 -2.60478139e-01
-4.38671112e-01 1.06461629e-01 5.55176914e-01 2.44919419e-01
-1.38273752e+00 1.26522768e+00 5.49545813e+00 7.85415053e-01
-6.83109105e-01 3.96247774e-01 2.82033384e-01 6.93777561e-01
-2.87413716e-01 2.02285305e-01 -1.36834598e+00 4.08307552e-01
1.14637363e+00 -2.57649750e-01 4.95343423e-03 8.80093813e-01
-1.93183348e-01 6.29731119e-01 -1.03661358e+00 7.54427016e-01
1.66075453e-01 -1.26880646e+00 1.29843682e-01 -2.22000510e-01
6.61677122e-01 3.09573323e-01 -6.47524118e-01 8.90392065e-01
5.96649408e-01 -7.47312307e-01 2.15159819e-01 6.04048967e-01
4.33363348e-01 -6.17504120e-01 1.21734607e+00 2.29589418e-01
-2.03632474e+00 -1.59329087e-01 -3.14994633e-01 4.02923003e-02
3.14504534e-01 4.72610384e-01 -7.46743619e-01 1.14620984e+00
4.14376199e-01 5.89549839e-01 -8.34762812e-01 9.23205972e-01
-6.96072876e-01 8.87508810e-01 -2.55746424e-01 -5.10755002e-01
3.18728626e-01 3.09559554e-02 3.38804096e-01 1.69270980e+00
-9.09306705e-02 1.86443388e-01 5.33514142e-01 5.26295066e-01
-7.51271129e-01 5.93278587e-01 -5.61565161e-01 -3.91720653e-01
8.02006602e-01 1.46376312e+00 -7.22014785e-01 -3.48237365e-01
-6.05238795e-01 1.12671733e+00 9.90156531e-01 2.84188002e-01
-8.00302446e-01 -1.03982604e+00 1.90911338e-01 -4.52495486e-01
4.92762417e-01 -3.40619594e-01 -2.25327477e-01 -1.49482250e+00
2.03848146e-02 -4.24216926e-01 1.03127575e+00 -5.14066100e-01
-1.75355327e+00 6.14503205e-01 -4.04576272e-01 -7.64497697e-01
2.98939496e-01 -6.36302531e-01 -7.87240505e-01 3.60677540e-01
-1.42715859e+00 -1.30934465e+00 -1.27664804e-01 5.06843507e-01
2.64296234e-01 7.31499791e-02 8.73562217e-01 6.31876290e-01
-1.03816903e+00 8.73481214e-01 -1.90457448e-01 1.18738651e+00
4.70317423e-01 -1.53204060e+00 6.34024203e-01 9.64337170e-01
3.92912745e-01 9.17457223e-01 1.91916171e-02 -9.38803196e-01
-1.34322190e+00 -1.43833470e+00 1.38215148e+00 -4.41661745e-01
9.34438467e-01 -4.96744245e-01 -9.71011877e-01 8.90158772e-01
4.84112799e-01 5.06949946e-02 9.89887178e-01 3.42814177e-01
-6.45184338e-01 3.66540276e-03 -8.40332031e-01 6.04552627e-01
1.41373444e+00 -7.81594276e-01 -1.08842933e+00 3.27838928e-01
1.61580467e+00 -2.63151020e-01 -1.20699525e+00 4.68864709e-01
-1.27519205e-01 -3.61637801e-01 6.90224767e-01 -1.18471658e+00
4.83221374e-02 -3.81548554e-01 -3.74015681e-02 -1.21311998e+00
-4.35109705e-01 -3.63127470e-01 -3.30521256e-01 1.88948429e+00
6.67553365e-01 -6.93602085e-01 5.66323519e-01 4.18612242e-01
-4.25795376e-01 -7.28822172e-01 -9.20028329e-01 -8.66528332e-01
-1.18837886e-01 -3.82298410e-01 8.16190362e-01 1.50478518e+00
1.04833998e-01 9.25143838e-01 -8.02643970e-02 5.99877179e-01
3.84661943e-01 -1.41886294e-01 2.96905100e-01 -1.34275913e+00
1.58187076e-01 -1.54956818e-01 -6.75186694e-01 -1.15812838e+00
5.70497334e-01 -1.33812535e+00 -1.43384799e-01 -2.04545426e+00
4.99874875e-02 -2.75531054e-01 -6.82042122e-01 9.32632089e-01
-5.52551568e-01 -2.08740130e-01 -3.64953130e-02 -1.10157594e-01
-1.23702371e+00 5.56027412e-01 8.08290243e-01 -1.81596771e-01
-6.51396438e-02 -6.75295830e-01 -7.13402390e-01 6.66667998e-01
8.79295290e-01 -6.70663953e-01 -1.49439216e-01 -7.74086177e-01
6.15241766e-01 -1.88133582e-01 2.64290906e-02 -9.22061861e-01
8.18668187e-01 1.83014959e-01 2.43634075e-01 -6.27607703e-01
-1.43281043e-01 -6.00678146e-01 -1.85652897e-01 1.63173929e-01
-4.40135837e-01 2.69221723e-01 1.24712372e-02 8.82631004e-01
-3.47929925e-01 -3.59787077e-01 1.75132066e-01 -1.06145941e-01
-1.12311065e+00 5.78647852e-01 3.09449006e-02 6.96751475e-01
7.27375269e-01 1.28725097e-01 -4.75516737e-01 1.25291765e-01
-7.16332138e-01 6.14343584e-01 -5.54979257e-02 5.50497711e-01
4.89437789e-01 -1.56589389e+00 -5.66969872e-01 -1.47949770e-01
2.41032407e-01 -8.59279837e-03 3.74294102e-01 5.84927976e-01
-3.59970182e-01 3.15005004e-01 3.91949594e-01 -1.42924845e-01
-8.79445195e-01 5.88496745e-01 2.41878718e-01 -8.33508193e-01
-4.40491229e-01 1.02478004e+00 -3.39294947e-03 -1.12615895e+00
2.44015604e-01 -1.31540596e-01 -7.28634655e-01 2.87048012e-01
3.04846406e-01 3.93828422e-01 3.02981716e-02 -5.41317344e-01
-4.25178647e-01 4.31164891e-01 -5.62563613e-02 3.51362556e-01
1.19634354e+00 -1.27211690e-01 -4.13926691e-01 3.35581511e-01
1.31795692e+00 1.78005904e-01 -5.37881613e-01 -4.46454883e-01
7.93917418e-01 1.75863579e-01 -8.25652331e-02 -5.52750409e-01
-8.85045946e-01 8.01302075e-01 2.09107503e-01 4.67518687e-01
9.48464096e-01 9.97592732e-02 1.23148108e+00 7.92403817e-01
3.18482131e-01 -1.02398264e+00 -9.23818201e-02 1.07687247e+00
4.59910423e-01 -1.03972471e+00 -4.18490350e-01 -5.62929392e-01
-5.72815359e-01 1.12439370e+00 9.58655059e-01 4.19930853e-02
6.06902361e-01 1.62593141e-01 6.15128502e-02 -2.98812866e-01
-6.47525907e-01 -5.82120061e-01 4.20107007e-01 4.08157527e-01
6.47146702e-01 -1.81219280e-02 -6.87503040e-01 1.27343178e+00
6.22257143e-02 -3.19125056e-01 3.02608013e-01 7.81711340e-01
-4.31278735e-01 -1.03108728e+00 2.40582108e-01 4.03608143e-01
-5.10265589e-01 -7.05755711e-01 -5.50084114e-01 6.46921813e-01
2.47755766e-01 1.09355879e+00 -1.88423112e-01 -6.17021441e-01
6.53993249e-01 4.03880805e-01 -1.97077751e-01 -9.05023515e-01
-9.72272635e-01 -5.27091324e-01 4.96955454e-01 -4.62474048e-01
-1.77742869e-01 -9.18380320e-02 -1.87085855e+00 5.29811345e-02
-7.01386809e-01 5.71675420e-01 3.66825432e-01 8.38714361e-01
7.85290003e-01 6.84260905e-01 7.18742192e-01 -1.14866510e-01
-4.77460265e-01 -1.02607369e+00 -5.20589888e-01 5.06429493e-01
-1.58600792e-01 -3.70849520e-01 -5.42071760e-01 -2.17294946e-01]
|
[9.54372787475586, 9.416807174682617]
|
4302ae02-6edc-411a-ae0c-d4ef8475d177
|
blind-omnidirectional-image-quality-1
|
2302.12393
| null |
https://arxiv.org/abs/2302.12393v1
|
https://arxiv.org/pdf/2302.12393v1.pdf
|
Blind Omnidirectional Image Quality Assessment: Integrating Local Statistics and Global Semantics
|
Omnidirectional image quality assessment (OIQA) aims to predict the perceptual quality of omnidirectional images that cover the whole 180$\times$360$^{\circ}$ viewing range of the visual environment. Here we propose a blind/no-reference OIQA method named S$^2$ that bridges the gap between low-level statistics and high-level semantics of omnidirectional images. Specifically, statistic and semantic features are extracted in separate paths from multiple local viewports and the hallucinated global omnidirectional image, respectively. A quality regression along with a weighting process is then followed that maps the extracted quality-aware features to a perceptual quality prediction. Experimental results demonstrate that the proposed S$^2$ method offers highly competitive performance against state-of-the-art methods.
|
['Zhou Wang', 'Wei Zhou']
|
2023-02-24
| null | null | null | null |
['image-quality-assessment']
|
['computer-vision']
|
[-3.22008617e-02 -5.02548873e-01 1.64921612e-01 -6.49499893e-01
-9.70424533e-01 -1.38930872e-01 3.83258909e-01 -2.87331045e-01
-3.25883567e-01 4.77262437e-01 6.15499735e-01 -2.18168095e-01
-5.70159018e-01 -8.52068543e-01 -2.26599887e-01 -5.15756428e-01
-1.52833775e-01 -4.64460790e-01 -1.50452092e-01 -2.84899652e-01
4.54850376e-01 3.51955503e-01 -1.92692113e+00 5.31920791e-02
1.09819925e+00 1.55698860e+00 3.91518295e-01 7.02930093e-01
2.21974656e-01 9.08950686e-01 -4.27287668e-01 -4.07049835e-01
5.05474389e-01 -1.95768625e-01 -4.44595039e-01 2.77600586e-01
5.95491588e-01 -3.89476180e-01 -5.98575890e-01 1.42837405e+00
7.25271523e-01 1.95611522e-01 3.64231288e-01 -9.85221088e-01
-6.66627109e-01 -3.07443291e-01 -3.19385380e-01 3.59847307e-01
8.09019744e-01 3.40311736e-01 1.13270128e+00 -1.21053302e+00
4.09424394e-01 1.03222537e+00 1.71757787e-01 6.15036376e-02
-6.71897411e-01 -4.25678015e-01 1.42196774e-01 5.61614692e-01
-1.37538946e+00 -5.94283640e-01 7.13956118e-01 -2.38904715e-01
7.70125508e-01 3.47575843e-01 8.24405313e-01 4.81799126e-01
3.38164777e-01 7.26155400e-01 1.51979625e+00 -1.66363567e-01
4.16286558e-01 2.77978331e-02 -2.82186031e-01 9.08572555e-01
-6.68856278e-02 4.43405807e-01 -7.01206505e-01 2.19978243e-01
5.95282376e-01 -1.10031441e-01 -6.57428741e-01 -7.81291842e-01
-1.18035257e+00 4.76413012e-01 7.32135177e-01 -7.06054419e-02
-4.78690624e-01 -2.44612604e-01 8.35799202e-02 5.63822269e-01
2.57113397e-01 4.27813619e-01 -2.81831801e-01 -6.93677068e-02
-6.15049958e-01 7.27119297e-02 1.81478068e-01 1.12334454e+00
7.12360382e-01 2.50685394e-01 -1.06086005e-02 1.12297654e+00
4.25051272e-01 9.25922871e-01 1.27894089e-01 -1.23319054e+00
3.21675628e-01 4.71781909e-01 5.61766982e-01 -1.22491169e+00
-2.46796653e-01 -7.95864284e-01 -6.66220307e-01 6.89963579e-01
2.60385312e-03 2.63262928e-01 -6.94337368e-01 1.35918462e+00
8.53294730e-02 -3.08375359e-01 1.54307038e-01 1.38506532e+00
1.00742650e+00 5.60041010e-01 -8.62112194e-02 -1.87191129e-01
1.19179153e+00 -9.09490705e-01 -4.59066957e-01 -2.46920586e-01
-1.91938519e-01 -8.21982265e-01 1.33797419e+00 7.51576900e-01
-1.28812671e+00 -9.56991315e-01 -1.19484997e+00 1.28019005e-01
-1.30500525e-01 7.90240020e-02 5.68502247e-01 8.73022377e-01
-1.46358502e+00 1.97759606e-02 -1.28498882e-01 -7.40826130e-02
1.68947026e-01 2.20289513e-01 -5.13495386e-01 -8.86743546e-01
-8.85796189e-01 7.76560843e-01 -2.58172572e-01 -5.55506535e-02
-1.15124154e+00 -4.46218014e-01 -1.11019802e+00 -2.15365812e-01
3.55626971e-01 -8.11698735e-01 7.92597115e-01 -6.00579441e-01
-1.49187148e+00 8.35020542e-01 -5.19808114e-01 1.65862933e-01
3.05295318e-01 -1.58026457e-01 -1.03702283e+00 1.89858645e-01
3.12475234e-01 4.91608858e-01 8.92183423e-01 -1.81346381e+00
-1.21303225e+00 -7.23721206e-01 3.90320957e-01 7.67775893e-01
3.37137654e-02 -2.58063227e-02 -8.07657242e-01 -3.92112762e-01
6.15185380e-01 -3.24197769e-01 -1.55909657e-01 1.47795692e-01
-1.52804524e-01 1.26197487e-01 3.66387129e-01 -8.10050607e-01
1.00818098e+00 -2.05474877e+00 4.84518707e-02 3.40207338e-01
1.71151862e-01 -5.70528023e-02 -2.80162364e-01 5.19443043e-02
2.48444468e-01 -3.60670716e-01 8.22323933e-02 -1.67560250e-01
-7.59245381e-02 -1.35807797e-01 1.49994403e-01 6.29968941e-01
-2.86461055e-01 5.77086151e-01 -1.17923391e+00 -2.92772710e-01
6.76699042e-01 3.80511373e-01 -4.80353057e-01 4.66566473e-01
3.50446194e-01 6.34610772e-01 -1.58281371e-01 1.14669991e+00
9.97176528e-01 -4.09121141e-02 2.18904694e-03 -3.65925014e-01
-3.74102145e-01 3.08005542e-01 -1.19805336e+00 2.13026142e+00
-8.56183231e-01 3.06034386e-01 8.13924968e-02 -7.76874840e-01
1.03947079e+00 1.13381088e-01 3.80894959e-01 -1.62809205e+00
7.75482226e-03 3.60968888e-01 -3.04918557e-01 -5.10605037e-01
6.41755700e-01 -6.34180978e-02 -8.94063339e-02 -1.99398592e-01
2.01778457e-01 -5.76587200e-01 -1.54671937e-01 -2.50849247e-01
8.76432240e-01 8.87036473e-02 5.59787393e-01 -3.26708138e-01
8.86637688e-01 -4.00348753e-01 6.60519540e-01 9.28938508e-01
-5.72950363e-01 6.47511363e-01 -2.88493097e-01 -4.10676539e-01
-1.09144497e+00 -1.72453177e+00 -2.02823412e-02 8.61578107e-01
8.83629918e-01 -1.39104098e-01 -3.61009836e-01 -4.77438688e-01
-3.70450139e-01 6.44200623e-01 -2.75616497e-01 -1.17137052e-01
-8.30603838e-02 -4.03750658e-01 -2.14526653e-01 2.58004338e-01
9.27287459e-01 -9.15557683e-01 -5.47881305e-01 -5.83124161e-02
-5.78611314e-01 -1.10421348e+00 -2.75151402e-01 -2.38408804e-01
-5.77573657e-01 -6.93061531e-01 -6.38721824e-01 -6.41554058e-01
4.74158823e-01 9.77576375e-01 1.34467030e+00 -2.52761781e-01
-2.55274087e-01 5.76754391e-01 -3.64613861e-01 -2.11189836e-01
2.47440964e-01 -7.83218265e-01 1.96457744e-01 1.91692740e-01
2.74193764e-01 -6.20312989e-01 -1.25646853e+00 5.85359693e-01
-6.49076700e-01 -9.94977206e-02 6.73804462e-01 8.01926613e-01
8.53456140e-01 4.46600914e-01 3.22685540e-01 6.92672729e-02
4.53600109e-01 -2.70522773e-01 -6.43415272e-01 6.02638908e-02
-6.97431147e-01 -2.38074452e-01 6.59100890e-01 4.00043190e-01
-1.24603009e+00 -5.74823737e-01 -2.87519455e-01 -1.31409854e-01
-1.98163047e-01 3.22846830e-01 -8.46917510e-01 -2.07443714e-01
3.77898842e-01 4.80216265e-01 -3.33112538e-01 -4.23584640e-01
4.26859140e-01 7.27654457e-01 8.83192897e-01 -1.30213782e-01
7.62155056e-01 8.85093331e-01 -1.64075419e-01 -8.52265000e-01
-6.89694226e-01 -9.70276833e-01 -6.96649849e-02 -6.49677753e-01
6.45031154e-01 -1.33486748e+00 -7.32897460e-01 5.18244088e-01
-6.10666275e-01 1.70092136e-01 -4.60646719e-01 8.67206931e-01
-8.87913227e-01 3.29626501e-01 -2.19738722e-01 -8.72508407e-01
-3.33401859e-02 -1.45501018e+00 1.02974319e+00 4.20534670e-01
3.21142495e-01 -4.45502937e-01 -2.02947333e-01 7.39216685e-01
3.10712606e-01 -2.39240408e-01 8.38483989e-01 3.81867737e-01
-8.33643556e-01 2.29590293e-02 -8.69478345e-01 6.35225236e-01
3.69872540e-01 -7.44423509e-01 -9.21412289e-01 -3.76733720e-01
2.23023087e-01 -2.96862811e-01 5.52358568e-01 6.20432019e-01
1.17818844e+00 -2.45036557e-01 3.22658479e-01 1.00231445e+00
1.83228076e+00 5.81592739e-01 1.04343510e+00 4.14539486e-01
4.59750324e-01 4.48343188e-01 9.86479402e-01 6.77515149e-01
6.46440446e-01 8.62885296e-01 9.17262912e-01 -2.78714657e-01
-2.31006309e-01 -3.86202872e-01 2.66650766e-01 7.97752261e-01
-2.47458309e-01 -2.39925504e-01 -5.86782575e-01 7.96345413e-01
-1.24905646e+00 -8.35585237e-01 4.09259140e-01 2.33574390e+00
3.09166640e-01 -6.78093210e-02 -8.08353871e-02 4.47728068e-01
3.89474928e-01 4.71531302e-01 -5.72658360e-01 -3.51278603e-01
-3.75762731e-01 2.04443187e-01 4.69669163e-01 6.64546669e-01
-9.45302844e-01 6.39890134e-01 5.96212101e+00 7.88829923e-01
-8.89075458e-01 -6.44737408e-02 3.88101488e-01 4.12739590e-02
-6.52426243e-01 -1.90212429e-01 -1.52443260e-01 1.05785891e-01
5.53129554e-01 1.03411950e-01 6.01948857e-01 8.95169437e-01
5.11658192e-01 -3.11181605e-01 -6.07660174e-01 1.53529608e+00
3.65785897e-01 -8.91576648e-01 1.56106859e-01 2.12712243e-01
8.06531131e-01 1.38337258e-02 6.64247811e-01 -1.75885156e-01
4.66755241e-01 -1.12888908e+00 9.78207111e-01 8.27327728e-01
1.24921155e+00 -8.20563495e-01 7.24783599e-01 1.56446919e-03
-1.28255820e+00 -5.61673105e-01 -4.07061189e-01 8.93063024e-02
1.56793624e-01 5.09002566e-01 -3.49151313e-01 1.10489559e+00
1.33626246e+00 7.31812358e-01 -5.74034214e-01 1.24279404e+00
-9.47201625e-02 1.68394111e-02 3.11784387e-01 2.54165024e-01
9.91606340e-02 -3.93913835e-01 9.13049877e-01 8.49090755e-01
7.45959938e-01 2.86089838e-01 -1.46762595e-01 4.35743749e-01
1.55528188e-01 2.63724923e-01 -7.30513871e-01 6.04204774e-01
1.10260762e-01 7.98520565e-01 2.58191470e-02 -1.17087938e-01
-7.09464312e-01 1.11034131e+00 -1.06295526e-01 4.25283045e-01
-4.20631886e-01 -3.23244274e-01 1.04204011e+00 -3.67761776e-02
2.68539369e-01 -2.90091932e-01 -3.59348595e-01 -1.40759802e+00
2.33924419e-01 -1.03516793e+00 1.43867865e-01 -1.45315969e+00
-1.18979752e+00 6.93811178e-01 -4.23282325e-01 -1.75730729e+00
1.15298880e-02 -6.32204652e-01 -6.13493938e-03 9.08185959e-01
-2.05695415e+00 -9.61037040e-01 -8.49307418e-01 1.04696596e+00
6.33243084e-01 -3.96379977e-01 8.03758264e-01 2.12888718e-01
1.67371914e-01 4.74270880e-01 2.78841913e-01 -1.93950795e-02
3.01034451e-01 -1.08289611e+00 7.54055008e-02 1.03511965e+00
-7.27477670e-02 3.96865726e-01 7.61318624e-01 -1.22214153e-01
-1.47788811e+00 -7.75870919e-01 6.74526215e-01 -3.96941677e-02
1.47804886e-01 1.42583057e-01 -2.56753981e-01 -1.35440260e-01
6.80872193e-03 3.33965510e-01 6.35447323e-01 -2.21754983e-01
-5.60833752e-01 -6.56558633e-01 -1.44123971e+00 8.42404962e-01
1.61933708e+00 -7.76142359e-01 -2.73523480e-01 -4.06059831e-01
4.70004141e-01 6.58370107e-02 -9.72445667e-01 7.75991321e-01
8.29057336e-01 -1.75375867e+00 1.37281024e+00 -1.06708549e-01
2.09625334e-01 -6.24326766e-01 -1.09750462e+00 -1.47848403e+00
-4.17044550e-01 -1.18244790e-01 3.53922069e-01 5.21390140e-01
1.04918644e-01 -3.91827643e-01 4.33336169e-01 2.96242535e-02
-3.02018821e-01 -3.80428702e-01 -1.09000850e+00 -6.63919151e-01
-4.96285111e-01 -6.67856336e-01 6.28418148e-01 4.58074868e-01
-1.24892592e-01 -1.37178019e-01 -5.17520368e-01 4.02102321e-01
1.04142046e+00 7.36058205e-02 5.42708099e-01 -1.03230762e+00
-2.21140668e-01 -2.33015582e-01 -8.62591147e-01 -1.38634479e+00
-2.96196729e-01 -3.48441064e-01 9.11836401e-02 -1.80915642e+00
4.49981578e-02 -6.34745240e-01 -7.38300204e-01 -3.16244125e-01
-1.22361399e-01 3.82369697e-01 7.23381490e-02 7.37978220e-02
-8.88911366e-01 8.32981229e-01 1.41393602e+00 -4.82162297e-01
-5.24842851e-02 5.34432046e-02 -9.19382215e-01 6.09292209e-01
4.91864055e-01 1.32026523e-01 -6.77850604e-01 -5.58310330e-01
2.89819807e-01 6.68638349e-01 4.63863432e-01 -1.38416934e+00
-8.31728950e-02 -3.22777808e-01 4.07231092e-01 -4.10002887e-01
5.50120533e-01 -9.40094352e-01 -5.17812558e-02 1.47069246e-01
-7.02738911e-02 1.61751196e-01 -3.56629461e-01 6.70176625e-01
-6.20727718e-01 3.06323886e-01 1.06712782e+00 -3.49267691e-01
-1.25264871e+00 4.67348903e-01 -1.04430318e-01 -1.13741741e-01
4.33307022e-01 -4.28603649e-01 -2.08265811e-01 -9.31083500e-01
-6.02441251e-01 -2.35705376e-01 6.41437531e-01 5.91028810e-01
1.31280446e+00 -1.31560290e+00 -6.61361992e-01 3.02804172e-01
7.64719665e-01 -4.20503527e-01 5.87540030e-01 4.75410551e-01
-3.76295060e-01 5.28004408e-01 -6.38265252e-01 -5.72200119e-01
-9.66391742e-01 5.73096514e-01 4.02490228e-01 -5.77049926e-02
-4.61467177e-01 8.28144670e-01 1.87527999e-01 -3.83614033e-01
8.45549852e-02 5.58812520e-04 -5.02989888e-01 -4.47949260e-01
6.77085876e-01 4.70001608e-01 2.39057451e-01 -1.06454599e+00
-3.68680328e-01 7.24102139e-01 3.87002558e-01 -6.13950372e-01
1.09709013e+00 -9.86268997e-01 8.01579952e-02 2.54093707e-01
1.29257298e+00 3.29804569e-01 -1.40460503e+00 -5.66812694e-01
-4.41913843e-01 -1.23287439e+00 2.56811619e-01 -1.09320736e+00
-1.14751172e+00 7.73198187e-01 1.25887811e+00 -7.16991127e-02
1.69565618e+00 -1.41779989e-01 4.40327585e-01 -8.22044462e-02
9.93883133e-01 -1.06006968e+00 2.15575308e-01 1.79044157e-01
8.87279332e-01 -1.58459258e+00 -1.44260272e-01 -2.74663419e-01
-6.00344837e-01 6.00390851e-01 5.39152741e-01 1.56945825e-01
6.95156455e-01 -3.35034698e-01 4.43374544e-01 -1.07494138e-01
-2.84460604e-01 -5.23492336e-01 5.90068579e-01 1.11539137e+00
6.39373958e-02 2.33804986e-01 -1.45728111e-01 2.73296446e-01
-4.06473339e-01 -5.05416170e-02 3.07667464e-01 6.27614617e-01
-5.42044938e-01 -6.77682221e-01 -3.14599633e-01 2.44011730e-01
-2.85080492e-01 -2.13845327e-01 3.71355087e-01 3.18856627e-01
4.15032625e-01 1.70043409e+00 -7.94403404e-02 -6.84617758e-01
5.88414967e-01 -4.11095679e-01 4.26962763e-01 -8.54516402e-02
5.22614308e-02 -6.82447925e-02 1.04573928e-01 -1.11852586e+00
-5.30835330e-01 -4.46658045e-01 -7.72501647e-01 -2.19076484e-01
1.86261505e-01 5.68213835e-02 9.37625766e-01 6.54080570e-01
1.45160586e-01 3.77209723e-01 1.07358241e+00 -6.81139648e-01
-5.92418537e-02 -5.65472543e-01 -8.99873555e-01 4.55192387e-01
6.96546376e-01 -6.93755805e-01 -5.02070189e-01 -2.09048718e-01]
|
[11.80144214630127, -1.9180090427398682]
|
6e4469c9-07b8-4ea4-a2ea-8a5f4df868c3
|
e-2tad-an-energy-efficient-tracking-based
|
2204.04416
| null |
https://arxiv.org/abs/2204.04416v4
|
https://arxiv.org/pdf/2204.04416v4.pdf
|
E^2TAD: An Energy-Efficient Tracking-based Action Detector
|
Video action detection (spatio-temporal action localization) is usually the starting point for human-centric intelligent analysis of videos nowadays. It has high practical impacts for many applications across robotics, security, healthcare, etc. The two-stage paradigm of Faster R-CNN inspires a standard paradigm of video action detection in object detection, i.e., firstly generating person proposals and then classifying their actions. However, none of the existing solutions could provide fine-grained action detection to the "who-when-where-what" level. This paper presents a tracking-based solution to accurately and efficiently localize predefined key actions spatially (by predicting the associated target IDs and locations) and temporally (by predicting the time in exact frame indices). This solution won first place in the UAV-Video Track of 2021 Low-Power Computer Vision Challenge (LPCVC).
|
['Gang Hua', 'Zhangyang Wang', 'Zhou Ren', 'Yi Wu', 'Pengcheng Pi', 'Zhenyu Hu', 'Taiyu Long', 'Siqi Fan', 'Hao-Yu Miao', 'Zhenyu Wu', 'Xin Hu']
|
2022-04-09
| null | null | null | null |
['fine-grained-action-detection', 'action-localization', 'spatio-temporal-action-localization']
|
['computer-vision', 'computer-vision', 'computer-vision']
|
[ 4.00193453e-01 -3.28375340e-01 -3.24127227e-01 7.30910227e-02
-4.32372630e-01 -4.97361273e-01 7.13369071e-01 -1.43635496e-01
-6.14105940e-01 5.33397496e-01 3.20504844e-01 -4.78931889e-02
-1.65608555e-01 -6.55161917e-01 -4.63525593e-01 -7.57283449e-01
-2.65984178e-01 9.83796865e-02 5.79750419e-01 -1.00993320e-01
2.84339935e-01 9.72867727e-01 -1.83109653e+00 5.00624299e-01
5.66445366e-02 9.66896713e-01 2.23433226e-01 9.87636149e-01
4.47357774e-01 1.34254742e+00 -5.41049957e-01 -8.52085948e-02
4.52535510e-01 -4.77013022e-01 -6.53567791e-01 1.38147488e-01
2.12660402e-01 -5.85628569e-01 -5.76683223e-01 1.07449651e+00
2.83293426e-01 2.59035647e-01 2.19293579e-01 -1.57947159e+00
-2.57652491e-01 3.35799187e-01 -6.40277505e-01 7.23078609e-01
6.57235384e-01 3.61506492e-01 5.66675901e-01 -5.54019034e-01
7.62806535e-01 1.16136003e+00 7.37767518e-01 6.74019575e-01
-4.11917299e-01 -4.57346171e-01 2.16922030e-01 7.83880949e-01
-1.49968934e+00 -2.52220809e-01 5.00084758e-01 -5.06139517e-01
1.08071280e+00 3.10475022e-01 8.86236846e-01 1.18201590e+00
1.72427237e-01 1.05234182e+00 5.82517266e-01 -1.90331504e-01
7.82739222e-02 -3.64179671e-01 -2.23998159e-01 7.43824601e-01
1.72120720e-01 2.62418032e-01 -5.35583258e-01 7.76225030e-02
1.04074562e+00 4.69419241e-01 -1.12668283e-01 -1.24401405e-01
-1.72579348e+00 4.81091827e-01 3.04191619e-01 4.89025772e-01
-7.93002903e-01 4.77550000e-01 5.83775461e-01 -1.69032425e-01
2.96348959e-01 1.69543222e-01 -6.03521645e-01 -4.70884323e-01
-1.08119905e+00 3.34166020e-01 7.17857629e-02 9.65807319e-01
3.14248949e-01 1.10462099e-01 -6.33459687e-01 -2.97419764e-02
1.37663379e-01 4.53033864e-01 3.09626818e-01 -1.08362281e+00
3.15396488e-01 7.25544930e-01 6.08004093e-01 -1.03282619e+00
-6.85321867e-01 4.01135487e-03 -8.59552920e-01 1.89885437e-01
5.16599655e-01 -3.08317423e-01 -6.81311429e-01 1.26341927e+00
4.87838387e-01 5.16538084e-01 -2.62997329e-01 1.22455263e+00
5.80848277e-01 6.17051601e-01 4.30697590e-01 -2.14633346e-01
1.53765881e+00 -9.36197937e-01 -6.02818668e-01 -5.97983859e-02
6.97604179e-01 -4.03566450e-01 4.02191848e-01 4.83654469e-01
-8.48035216e-01 -7.94723451e-01 -6.87914789e-01 1.20878555e-01
-5.47896147e-01 6.42218113e-01 6.41286671e-01 5.87299168e-01
-1.12136221e+00 4.31627780e-01 -8.22178900e-01 -6.85823917e-01
5.25422215e-01 3.20954889e-01 -5.41982353e-01 1.67655535e-02
-9.59707320e-01 8.69406641e-01 5.45227885e-01 1.78330868e-01
-1.28339791e+00 -3.80827248e-01 -5.78639567e-01 -1.03009887e-01
5.48887372e-01 -4.83151704e-01 1.07931435e+00 -1.04801393e+00
-1.12956119e+00 7.76381850e-01 -7.85815120e-02 -9.86495018e-01
6.72600865e-01 -3.67913276e-01 -4.49697673e-01 3.28503549e-01
8.82495269e-02 9.93681848e-01 8.55850458e-01 -5.48280120e-01
-1.37179756e+00 -3.37539673e-01 2.89348245e-01 1.93571493e-01
-1.34233445e-01 6.33101463e-01 -5.77267766e-01 -5.86990654e-01
-1.84101984e-01 -8.02955568e-01 -3.85647863e-01 1.33835748e-01
-1.62304416e-01 -5.83698630e-01 1.02629066e+00 -8.39710474e-01
1.04163110e+00 -2.01116252e+00 1.20237753e-01 -1.88049212e-01
1.54800788e-01 4.96069103e-01 1.86795872e-02 1.45881251e-01
-1.05441727e-01 -2.54316956e-01 5.24170101e-01 9.70270578e-03
-2.15524226e-01 -2.93145269e-01 -1.22180670e-01 8.69919062e-01
1.15140148e-01 9.64323401e-01 -1.16489244e+00 -5.91294110e-01
7.49805748e-01 5.24565935e-01 -3.49996328e-01 -8.03002641e-02
-1.66802779e-01 5.01129389e-01 -5.54080486e-01 9.16315436e-01
2.04718336e-01 -5.58137475e-03 4.26965440e-03 -4.23935354e-01
-5.77619493e-01 -2.55425066e-01 -1.24957025e+00 1.34942269e+00
5.78574687e-02 8.86034787e-01 -4.39606141e-03 -1.05554032e+00
5.50613821e-01 4.84843016e-01 1.14111638e+00 -5.89423478e-01
2.72966474e-01 -1.30661443e-01 -1.50060490e-01 -4.60113049e-01
6.20398283e-01 4.40718830e-01 -1.39505312e-01 1.38587520e-01
-1.34584486e-01 7.31920242e-01 4.13000464e-01 2.76920944e-02
1.47536111e+00 5.64961553e-01 5.48437119e-01 -1.38999410e-02
6.76239729e-01 3.91744405e-01 5.66880822e-01 8.58011901e-01
-7.94203997e-01 3.67199540e-01 3.75951640e-02 -1.01884711e+00
-9.47920799e-01 -6.45534098e-01 5.12115538e-01 1.19446981e+00
2.33143181e-01 -3.00567150e-01 -8.88164818e-01 -8.33254278e-01
-4.03584272e-01 3.41537356e-01 -5.59823036e-01 -1.44644499e-01
-7.84252882e-01 -2.91391283e-01 6.62881613e-01 7.18411088e-01
7.95717061e-01 -1.56011653e+00 -1.35345221e+00 2.86060691e-01
-3.04185390e-01 -1.28587544e+00 -3.35103273e-01 -5.65551557e-02
-5.11195064e-01 -1.29135048e+00 -7.13786006e-01 -5.29354393e-01
5.60687900e-01 6.58984482e-01 7.44799256e-01 4.32198346e-02
-6.27403378e-01 7.23485410e-01 -6.00276113e-01 -5.13455927e-01
9.33564156e-02 -3.20868254e-01 3.86443049e-01 1.14235044e-01
5.88618755e-01 1.54977500e-01 -8.27701271e-01 4.02219087e-01
-4.45531636e-01 -5.06809466e-02 6.33040190e-01 1.86161712e-01
5.44511974e-01 4.83259439e-01 1.06112547e-01 -1.82986632e-01
1.07971191e-01 -2.69278497e-01 -7.65848458e-01 4.74108756e-01
1.06284760e-01 -5.74387193e-01 5.07215023e-01 -4.46509123e-01
-7.82225370e-01 6.69188678e-01 4.99648303e-02 -6.32311821e-01
-5.25301576e-01 -1.64444283e-01 -1.24574944e-01 -4.74617854e-02
4.95914876e-01 4.17521030e-01 -4.56873655e-01 -9.57278013e-02
3.29483598e-01 4.21472877e-01 6.55437469e-01 -2.87735998e-03
7.18038082e-01 7.70728409e-01 7.89235681e-02 -8.49575400e-01
-7.62081742e-01 -8.66895497e-01 -9.25147474e-01 -8.94968510e-01
1.44295764e+00 -1.05244076e+00 -1.06296885e+00 5.22114754e-01
-1.48895216e+00 -1.60869792e-01 -3.16823751e-01 4.94688183e-01
-6.80820465e-01 2.51410335e-01 -3.59082162e-01 -1.03231728e+00
-3.22413385e-01 -9.32431161e-01 1.34204412e+00 2.86481142e-01
-6.54978827e-02 -5.94474852e-01 -2.21950859e-01 3.55951339e-01
2.90706784e-01 3.00752789e-01 1.09095667e-02 -2.91814208e-01
-1.03172195e+00 -4.50657129e-01 -3.14794928e-01 2.24332884e-02
-1.22712059e-02 -4.28331224e-03 -6.04430318e-01 -1.11610614e-01
-2.14636981e-01 2.32422441e-01 6.13668859e-01 8.16120744e-01
1.18545914e+00 -2.14456528e-01 -6.13447011e-01 3.53367925e-01
1.18756330e+00 5.26548505e-01 8.09472919e-01 3.88458848e-01
7.48271286e-01 4.61330861e-01 1.20311165e+00 6.47036731e-01
9.07805562e-02 1.03025699e+00 5.22394061e-01 -6.65650144e-02
-2.22899824e-01 -9.53621268e-02 7.54697144e-01 -1.85714200e-01
-6.86194539e-01 -2.71791607e-01 -8.55878770e-01 5.10017216e-01
-2.18387270e+00 -1.74601138e+00 -2.42756099e-01 1.99033904e+00
2.60740165e-02 9.54478383e-02 5.40419638e-01 1.65236384e-01
9.76594746e-01 -2.36700382e-02 -3.81337494e-01 1.30337015e-01
1.13329954e-01 -1.87133193e-01 9.09897268e-01 -1.89331740e-01
-1.69970810e+00 1.14752424e+00 6.24991703e+00 8.00122797e-01
-8.81746769e-01 2.50899494e-01 2.98578739e-01 -3.82750072e-02
8.02762747e-01 -3.29094321e-01 -1.11891198e+00 4.04152304e-01
9.45227921e-01 1.69670954e-02 3.94563496e-01 1.16939187e+00
7.51430035e-01 -2.81461686e-01 -8.94391596e-01 1.14933991e+00
1.21112898e-01 -1.57568908e+00 -9.42438841e-02 -9.63057578e-02
4.55883741e-01 -8.44743252e-02 -2.82638282e-01 2.77720213e-01
2.36545086e-01 -7.92709410e-01 9.45121825e-01 7.06726670e-01
6.71676576e-01 -7.33178198e-01 7.11656153e-01 4.20470536e-01
-1.67661834e+00 -5.41848719e-01 -3.14323217e-01 -9.77059901e-02
3.99852306e-01 3.64596471e-02 -7.38137305e-01 4.16900545e-01
1.09291720e+00 1.04206824e+00 -4.84569192e-01 1.18090904e+00
-1.17945835e-01 2.66837716e-01 -5.82210124e-02 -2.43498206e-01
2.42207929e-01 -9.11836773e-02 4.78377432e-01 1.18753994e+00
5.09192109e-01 2.72053868e-01 3.69474262e-01 2.64703512e-01
2.74823278e-01 -2.16975793e-01 -6.53763592e-01 -7.67725408e-02
2.22634777e-01 1.40271652e+00 -1.30135846e+00 -4.19537127e-01
-5.00574470e-01 1.04282427e+00 -2.12578535e-01 -9.67466459e-02
-1.28382015e+00 -8.74416307e-02 7.38248408e-01 3.86163145e-01
4.96787429e-01 -4.16870147e-01 3.61595452e-01 -7.89780438e-01
-2.09933862e-01 -7.47165143e-01 4.60048229e-01 -9.74987328e-01
-6.13569140e-01 2.47688890e-01 4.97322567e-02 -1.80374026e+00
-2.59955466e-01 -8.84772599e-01 -3.58273298e-01 2.09025323e-01
-9.43544030e-01 -1.41561770e+00 -5.72989404e-01 8.93848836e-01
8.64754438e-01 -2.89265722e-01 5.90497017e-01 5.65570831e-01
-5.70691228e-01 1.18019611e-01 -3.21325928e-01 4.71683353e-01
2.55794704e-01 -8.72870326e-01 2.54056126e-01 1.36247361e+00
3.81228656e-01 2.23128006e-01 7.16907561e-01 -6.71852648e-01
-1.52032781e+00 -1.41876531e+00 7.18523502e-01 -7.36421406e-01
6.60560131e-01 -6.12137914e-02 -1.31601244e-01 7.12519348e-01
-7.86340162e-02 1.53675258e-01 5.80330715e-02 -4.38356251e-01
1.43276975e-01 -1.75649002e-01 -1.10800242e+00 7.02312946e-01
1.25724924e+00 -3.58317226e-01 -1.50480390e-01 7.90249944e-01
4.75706458e-01 -2.59501368e-01 -4.85911012e-01 3.35849732e-01
5.35648167e-01 -1.07610714e+00 1.26271462e+00 -6.15915000e-01
5.97965531e-02 -7.38482296e-01 -1.28831621e-02 -5.60363054e-01
-4.35505658e-01 -5.34796715e-01 -3.60066086e-01 8.45537245e-01
-2.88156360e-01 9.94545817e-02 1.01409090e+00 1.66617036e-01
-9.20728892e-02 -3.95262718e-01 -1.05347037e+00 -7.11914062e-01
-8.63557279e-01 -5.24609089e-01 2.47176841e-01 5.39315760e-01
-2.64491528e-01 -2.20364496e-01 -7.54052281e-01 3.47192466e-01
5.70287645e-01 -1.69438869e-01 6.93977892e-01 -1.05106974e+00
7.40207955e-02 -4.53355521e-01 -1.06845093e+00 -9.44472075e-01
-2.81746000e-01 -2.81954736e-01 1.87411875e-01 -1.65144980e+00
2.03563973e-01 1.82052284e-01 -3.53316039e-01 4.84913379e-01
1.07934847e-01 4.35575396e-01 2.04582646e-01 1.12406932e-01
-1.34083951e+00 7.78706595e-02 8.35946560e-01 -1.04761660e-01
7.69878626e-02 1.92097470e-01 -1.32544637e-01 7.78620481e-01
8.04909945e-01 -3.57819736e-01 -1.98059037e-01 -6.72471821e-02
-9.84129310e-02 1.03266969e-01 8.62923324e-01 -1.61635613e+00
5.34426868e-01 -4.96482641e-01 5.54747522e-01 -9.26303327e-01
4.22530442e-01 -1.02172899e+00 2.33446926e-01 8.84568095e-01
-1.21139914e-01 2.62009025e-01 -9.28219482e-02 6.76256061e-01
-3.15045640e-02 7.06917942e-02 7.28220463e-01 -4.01642531e-01
-1.52975261e+00 4.07837808e-01 -7.99420714e-01 -3.98491800e-01
1.76993155e+00 -4.00329292e-01 -4.01005208e-01 -2.14494348e-01
-5.70101976e-01 -2.69178934e-02 1.32103220e-01 7.36941516e-01
6.48431301e-01 -1.23845696e+00 -5.40582538e-01 -1.55012915e-02
3.46602760e-02 -4.30654585e-01 4.43117142e-01 1.02917433e+00
-6.65047050e-01 7.85620034e-01 -5.38843215e-01 -5.44221997e-01
-1.66694808e+00 8.27902317e-01 3.15212905e-01 -2.85505176e-01
-6.87252581e-01 8.76866221e-01 -1.82585508e-01 1.59855574e-01
4.53404218e-01 -2.72854082e-02 -5.27537525e-01 -8.44893511e-03
1.01549935e+00 6.81152284e-01 -2.63239503e-01 -1.04042292e+00
-6.84919894e-01 4.23268795e-01 3.03925693e-01 1.79248437e-01
1.20917881e+00 6.68550730e-02 1.16437189e-01 -1.26426974e-02
6.02739513e-01 -5.07768989e-01 -1.55632234e+00 1.58115402e-01
3.54082018e-01 -3.33291471e-01 7.31177703e-02 -4.95526940e-01
-1.04336631e+00 7.35362172e-01 9.19850826e-01 2.15780824e-01
1.12523878e+00 1.87702160e-02 5.79042912e-01 4.76982683e-01
9.60478544e-01 -1.33642936e+00 2.17451975e-01 3.51964980e-01
6.77021682e-01 -1.07082665e+00 1.74994409e-01 -2.64524519e-01
-6.45258784e-01 1.11771178e+00 7.18276560e-01 -1.51845589e-02
4.05090243e-01 2.69363850e-01 -2.99726456e-01 -2.88722187e-01
-4.35377985e-01 -5.95849276e-01 2.61648655e-01 9.29455400e-01
1.81927532e-01 6.46116212e-02 -5.08024581e-02 3.08464706e-01
3.84647250e-01 1.53507099e-01 2.68238097e-01 1.04281163e+00
-6.49710119e-01 -6.62680089e-01 -4.41484004e-01 3.46007943e-01
-3.96229833e-01 1.62684619e-01 -3.15519750e-01 8.00706744e-01
6.10340953e-01 1.04770434e+00 9.69703346e-02 -6.48768187e-01
3.41193080e-01 -1.72720030e-01 4.42206264e-01 -3.58482480e-01
-4.77906227e-01 -2.03343347e-01 9.68722440e-03 -1.11636555e+00
-9.45045233e-01 -1.02865303e+00 -1.07324076e+00 -2.03509659e-01
6.68421537e-02 -2.77444601e-01 5.38867056e-01 9.81261671e-01
4.12090182e-01 5.02889633e-01 3.95091414e-01 -1.32932210e+00
-1.41496167e-01 -8.16929936e-01 -3.15073639e-01 2.68830121e-01
9.75461602e-02 -7.82130957e-01 -6.18510544e-02 3.05478841e-01]
|
[8.346537590026855, 0.4420221745967865]
|
9a5f2956-a38b-4201-a622-a179a4c84c74
|
point-set-voting-for-partial-point-cloud
|
2007.04537
| null |
https://arxiv.org/abs/2007.04537v2
|
https://arxiv.org/pdf/2007.04537v2.pdf
|
Point Set Voting for Partial Point Cloud Analysis
|
The continual improvement of 3D sensors has driven the development of algorithms to perform point cloud analysis. In fact, techniques for point cloud classification and segmentation have in recent years achieved incredible performance driven in part by leveraging large synthetic datasets. Unfortunately these same state-of-the-art approaches perform poorly when applied to incomplete point clouds. This limitation of existing algorithms is particularly concerning since point clouds generated by 3D sensors in the real world are usually incomplete due to perspective view or occlusion by other objects. This paper proposes a general model for partial point clouds analysis wherein the latent feature encoding a complete point clouds is inferred by applying a local point set voting strategy. In particular, each local point set constructs a vote that corresponds to a distribution in the latent space, and the optimal latent feature is the one with the highest probability. This approach ensures that any subsequent point cloud analysis is robust to partial observation while simultaneously guaranteeing that the proposed model is able to output multiple possible results. This paper illustrates that this proposed method achieves state-of-the-art performance on shape classification, part segmentation and point cloud completion.
|
['Jun-ming Zhang', 'Yu-Ping Wang', 'Matthew Johnson-Roberson', 'Ram Vasudevan', 'Weijia Chen']
|
2020-07-09
| null | null | null | null |
['point-cloud-completion']
|
['computer-vision']
|
[ 2.84595013e-01 4.77121584e-03 -4.31529731e-02 -3.92606914e-01
-9.11812603e-01 -6.50486052e-01 6.48795009e-01 3.79067242e-01
5.89920208e-03 3.16198051e-01 -5.85643470e-01 -3.45755517e-02
-1.20237850e-01 -1.05122209e+00 -7.84373522e-01 -7.20553517e-01
3.43981311e-02 1.18310189e+00 4.84436631e-01 1.43025309e-01
4.29845244e-01 1.20496678e+00 -1.96872771e+00 6.27615675e-02
8.10168445e-01 8.19166005e-01 3.57913703e-01 5.06883740e-01
-5.11769652e-01 -3.02511722e-01 -4.81940448e-01 -3.32735837e-01
5.24285555e-01 2.53491968e-01 -4.34045553e-01 5.63891053e-01
3.84884030e-01 -6.07117545e-04 2.42711812e-01 1.11156476e+00
1.67816922e-01 -1.90712452e-01 6.22514963e-01 -1.53351021e+00
1.69445667e-02 5.09526394e-02 -4.82602715e-01 -5.10021865e-01
4.42096502e-01 -5.83132580e-02 8.22497249e-01 -1.13367260e+00
6.24775112e-01 1.22171140e+00 5.19529939e-01 1.99715346e-01
-1.42488217e+00 -4.76602912e-01 8.94506350e-02 -2.46710911e-01
-1.54590392e+00 -4.18585688e-02 1.08178699e+00 -5.06425738e-01
6.43674791e-01 5.27155399e-01 8.69183242e-01 5.58170617e-01
-1.28031090e-01 7.49464095e-01 1.02709067e+00 -3.76113236e-01
5.47734618e-01 1.30424067e-01 -4.49422076e-02 3.53502601e-01
4.21981961e-01 -6.67561814e-02 -3.36943418e-01 -7.02933311e-01
7.48999834e-01 3.84912640e-01 5.64842038e-02 -1.07636452e+00
-1.06879687e+00 5.64853370e-01 2.05400050e-01 7.41996542e-02
-4.56716865e-01 7.79693434e-03 -5.09971306e-02 1.17583230e-01
5.52854419e-01 -2.12117583e-02 -4.19789523e-01 8.34277496e-02
-1.28471684e+00 6.25318527e-01 6.13569319e-01 1.13812685e+00
1.00410259e+00 -2.23408312e-01 4.26486611e-01 5.83642244e-01
6.13615155e-01 8.52199733e-01 -2.77180791e-01 -1.03249133e+00
5.28516412e-01 1.08023775e+00 2.64556319e-01 -9.63211179e-01
-8.14567730e-02 -3.80603760e-01 -6.78161085e-01 7.77799189e-01
2.26195037e-01 5.79557955e-01 -8.86557341e-01 1.12587738e+00
6.88079178e-01 1.29266828e-01 -2.83218902e-02 6.77287042e-01
3.56069565e-01 5.34645677e-01 -2.51549333e-01 -7.38372132e-02
1.04762578e+00 -1.57688707e-01 -2.75897056e-01 -3.46224047e-02
1.24362506e-01 -8.92907917e-01 8.39145601e-01 6.64166033e-01
-9.85215902e-01 -5.33632755e-01 -9.44953740e-01 4.19495523e-01
-1.03622809e-01 -1.05460826e-02 4.10580695e-01 6.46665633e-01
-8.08351398e-01 5.29855847e-01 -1.17004287e+00 -2.69846618e-01
5.03275633e-01 6.42107606e-01 -3.42432022e-01 -2.04133570e-01
-3.33157420e-01 5.03421485e-01 2.24410638e-01 -7.43335113e-04
-5.91569185e-01 -5.60436308e-01 -4.55835700e-01 -1.06395349e-01
2.19418362e-01 -8.17585051e-01 9.50106919e-01 -5.83717585e-01
-1.05288303e+00 1.07490897e+00 -3.73745650e-01 -3.45181525e-01
7.75222898e-01 -1.02994718e-01 -7.35478178e-02 2.21211344e-01
1.30813673e-01 6.34519935e-01 1.06436682e+00 -1.82826042e+00
-6.79208457e-01 -8.50037277e-01 -2.06105724e-01 1.47408143e-01
1.70446500e-01 -2.62466937e-01 -6.29481614e-01 -2.64406204e-01
1.00244641e+00 -1.12862825e+00 -3.83405477e-01 2.03407049e-01
-3.72045636e-01 -2.71151096e-01 1.04201818e+00 4.76040058e-02
5.89726806e-01 -2.03723669e+00 8.19898620e-02 4.71795887e-01
1.83560606e-02 -9.20118913e-02 2.29174957e-01 5.37284493e-01
-1.60273891e-02 8.70244354e-02 -6.64349377e-01 -8.78880620e-01
3.19422111e-02 5.07308364e-01 -6.39107525e-01 7.73625910e-01
9.12528336e-02 4.85600978e-01 -6.85654223e-01 -5.90050876e-01
7.26305723e-01 3.45741987e-01 -2.96583891e-01 7.78484792e-02
-4.74125326e-01 4.84962970e-01 -7.20034063e-01 9.45206523e-01
1.34052324e+00 -5.68862669e-02 -9.44216102e-02 8.55059475e-02
-2.94465333e-01 -2.97360748e-01 -1.50827384e+00 1.84987903e+00
-9.92996097e-02 7.82655999e-02 1.80118456e-01 -6.59663558e-01
1.35943961e+00 4.25078750e-01 9.96675193e-01 1.26692846e-01
-1.55596375e-01 4.84708965e-01 -5.91319144e-01 -8.02569687e-02
5.16969442e-01 -1.76336840e-01 -3.51627059e-02 1.54352784e-01
-4.34986442e-01 -9.25565720e-01 -3.16692412e-01 4.94874120e-02
7.63626754e-01 2.91850179e-01 1.39301389e-01 -2.47422252e-02
5.39646566e-01 3.15084577e-01 4.94958460e-01 6.42657876e-01
7.97718763e-02 9.08971548e-01 2.19301268e-01 -2.93439031e-01
-1.21451414e+00 -1.22737825e+00 -4.05105263e-01 7.53791444e-03
4.29026753e-01 -8.16853642e-02 -5.88408351e-01 -3.77423763e-01
3.11613560e-01 7.69742131e-01 -3.01401705e-01 2.24275172e-01
-4.05900210e-01 -2.03481168e-01 2.06115693e-01 1.58529565e-01
2.67521948e-01 -8.11690390e-01 -8.66049826e-01 2.05530971e-01
6.81303740e-02 -1.09064436e+00 3.34721744e-01 -1.94069728e-01
-1.66323149e+00 -1.09938610e+00 -5.79873681e-01 -4.14144009e-01
9.99382079e-01 5.40383577e-01 1.09788525e+00 1.58891022e-01
3.30991261e-02 4.76489931e-01 -4.75026369e-01 -4.07884389e-01
-4.33922946e-01 1.20920397e-03 -4.38854471e-02 2.42343679e-01
4.62486356e-01 -6.93110824e-01 -3.97646993e-01 3.71279538e-01
-1.09206653e+00 -6.22845441e-02 3.48953456e-01 3.86383504e-01
1.07171714e+00 1.60682142e-01 -1.16507955e-01 -6.93818629e-01
8.64255577e-02 -2.72346407e-01 -1.04997957e+00 8.96704942e-02
-4.28052425e-01 -2.87294369e-02 2.18711257e-01 7.35985413e-02
-9.11223412e-01 6.20243073e-01 -1.98872797e-02 -9.40558791e-01
-6.07659340e-01 1.92979172e-01 -4.09758359e-01 -2.89731640e-02
4.20866847e-01 2.58626491e-01 5.74158877e-03 -6.88595653e-01
2.43206710e-01 5.30987680e-01 3.55462253e-01 -6.54746890e-01
1.19580019e+00 1.15310431e+00 5.84619164e-01 -9.50171292e-01
-3.05871993e-01 -7.90486574e-01 -1.06737709e+00 -3.71411294e-01
6.99057341e-01 -7.29180515e-01 -6.45601392e-01 5.11824012e-01
-1.36744320e+00 3.36849093e-01 -5.02750099e-01 3.89917046e-01
-6.96181834e-01 5.55245221e-01 1.56393498e-01 -1.21739268e+00
8.06206986e-02 -1.41607654e+00 1.52105415e+00 -2.06168622e-01
-1.05289242e-03 -6.25866175e-01 9.33545232e-02 2.60641932e-01
-2.33326852e-01 6.16126060e-01 6.64863706e-01 -3.45516860e-01
-1.28210366e+00 -7.42929757e-01 1.34879932e-01 1.36702746e-01
-3.46041545e-02 3.75315934e-01 -1.08024728e+00 -2.78294325e-01
2.57083654e-01 2.57106740e-02 5.03889918e-01 4.90787655e-01
1.06750476e+00 2.78078586e-01 -5.67246974e-01 5.50736427e-01
1.62795353e+00 -1.18812388e-02 4.40624058e-01 1.47417769e-01
5.13944387e-01 6.47521436e-01 1.07056391e+00 5.24624586e-01
1.64422110e-01 8.14999402e-01 1.06999648e+00 5.34829721e-02
2.71423787e-01 -3.24307293e-01 -1.64669737e-01 5.48343897e-01
-1.85179815e-01 -4.74861301e-02 -1.18400109e+00 5.67856669e-01
-1.84452534e+00 -7.41026640e-01 -6.09964490e-01 2.60548925e+00
5.09306118e-02 2.73702830e-01 -1.59953281e-01 5.70911348e-01
6.58114672e-01 6.16712682e-02 -4.85804647e-01 -4.60557931e-04
-2.22213507e-01 1.13345899e-01 5.87421715e-01 5.14417887e-01
-9.03138816e-01 6.72109485e-01 5.53884888e+00 7.36636937e-01
-8.03348780e-01 1.05972309e-02 1.41686171e-01 3.18497032e-01
-6.79853499e-01 4.32494104e-01 -7.78280020e-01 2.35596433e-01
2.17808068e-01 3.58670838e-02 -7.14922398e-02 1.03757036e+00
7.68447742e-02 -2.90853053e-01 -9.39531326e-01 1.03001487e+00
-4.69920486e-02 -1.17680991e+00 1.84136406e-01 4.83967513e-01
8.40129554e-01 1.50040090e-01 9.57357511e-02 -3.50305825e-01
2.66931113e-02 -6.08750880e-01 8.92963707e-01 6.23365879e-01
7.22985685e-01 -7.23396361e-01 4.48249876e-01 8.80616426e-01
-1.12386453e+00 1.06507532e-01 -4.49662447e-01 1.29001155e-01
2.88853824e-01 9.26013291e-01 -1.14410317e+00 7.43780375e-01
4.28006589e-01 4.11769420e-01 -3.26550007e-01 1.30462325e+00
-1.89253002e-01 4.41632956e-01 -7.73521185e-01 2.19717950e-01
2.20844001e-02 -5.60419738e-01 1.09479141e+00 7.51295507e-01
5.25245428e-01 1.96651015e-02 3.69487554e-01 8.76130044e-01
2.51113296e-01 -7.03298673e-02 -8.68692815e-01 4.20123577e-01
6.62540078e-01 9.77930009e-01 -1.12378705e+00 -2.67881900e-01
-3.16224903e-01 6.40429676e-01 -2.82413270e-02 1.59298778e-01
-4.56403762e-01 2.28900343e-01 5.69442272e-01 2.60687560e-01
2.91569203e-01 -6.80044830e-01 -8.14262331e-01 -9.69040036e-01
1.76407680e-01 -2.69833386e-01 1.22258961e-01 -7.92683661e-01
-1.10757875e+00 4.98603523e-01 3.15088123e-01 -1.77869463e+00
-1.77479044e-01 -4.33688939e-01 -3.05060685e-01 8.93103004e-01
-1.39755726e+00 -1.29523492e+00 -3.78878534e-01 5.37560582e-01
7.08250582e-01 5.36861680e-02 9.05490875e-01 6.21643439e-02
1.70942411e-01 -1.04857065e-01 8.79263803e-02 -3.88095975e-01
1.50852934e-01 -1.15697575e+00 5.95916212e-01 8.42488229e-01
2.76867837e-01 3.98610860e-01 9.33206856e-01 -9.66579437e-01
-1.42798209e+00 -8.70099127e-01 7.35414088e-01 -6.93935752e-01
8.18945318e-02 -4.48919922e-01 -8.94437134e-01 4.73506540e-01
-5.28219163e-01 1.72489211e-01 2.11192131e-01 -1.56358302e-01
-7.42553174e-02 -1.44527629e-01 -1.36448884e+00 3.05113345e-01
8.34522188e-01 -3.63699973e-01 -6.55178607e-01 3.85629147e-01
4.25596923e-01 -4.59257960e-01 -6.30937338e-01 6.74235821e-01
2.34791547e-01 -1.10917425e+00 1.12308598e+00 -1.95434466e-01
1.14416927e-01 -6.18374884e-01 -4.62687999e-01 -9.29171324e-01
4.33217809e-02 -3.54346991e-01 2.88926531e-02 9.92669284e-01
8.80782902e-02 -4.85612512e-01 1.38300288e+00 6.37274146e-01
-2.62771755e-01 -5.82464457e-01 -1.37328124e+00 -7.34876275e-01
-2.57137150e-01 -8.74895036e-01 1.00742257e+00 7.36584365e-01
-6.59648120e-01 -3.28632325e-01 5.11964411e-02 5.60402453e-01
1.18056798e+00 7.03320920e-01 1.22739744e+00 -1.78459668e+00
1.14340529e-01 -2.33886689e-01 -7.50396848e-01 -1.02802658e+00
6.80446848e-02 -7.98390329e-01 -9.63547602e-02 -1.57269752e+00
-6.85155839e-02 -1.00622618e+00 2.25986019e-01 -3.93329188e-04
1.49038956e-01 2.17136398e-01 3.26447517e-01 6.99185491e-01
-2.00819969e-01 3.95205647e-01 1.10052598e+00 -4.79618236e-02
-2.46665955e-01 6.57476127e-01 8.37669224e-02 9.58713293e-01
6.46914721e-01 -7.74160385e-01 -3.04649621e-01 -4.06300455e-01
3.13011438e-01 3.24260324e-01 5.92256188e-01 -1.09972548e+00
3.28861892e-01 -1.69673130e-01 2.85433650e-01 -1.43910539e+00
1.03869891e+00 -1.49546385e+00 7.22200036e-01 3.92577678e-01
1.47863954e-01 -2.47285496e-02 -9.18637738e-02 7.21657991e-01
-2.84777761e-01 -3.41418505e-01 6.10139072e-01 -3.15744072e-01
-3.17407370e-01 5.35941005e-01 6.24777824e-02 -5.72585464e-01
1.16744423e+00 -9.01390791e-01 3.86214197e-01 -9.04144943e-02
-7.33151436e-01 5.83474375e-02 1.03253388e+00 2.28170380e-01
8.98194313e-01 -1.19989789e+00 -7.12650597e-01 5.24485469e-01
3.18454146e-01 6.22038007e-01 2.15069607e-01 3.76843810e-01
-6.42768562e-01 3.32215995e-01 6.55972138e-02 -1.34498549e+00
-1.47284102e+00 3.78979683e-01 5.67891635e-02 -4.12053466e-02
-7.60280192e-01 5.99117279e-01 -1.13228038e-01 -6.46868110e-01
-1.71068162e-02 -3.75392258e-01 1.45176813e-01 -9.34030861e-02
-1.56478480e-01 4.82309133e-01 4.04243171e-01 -8.38657558e-01
-3.55730712e-01 9.39045608e-01 3.58646840e-01 -3.81312221e-01
1.32571101e+00 -9.42087080e-03 -3.01668607e-02 6.13235772e-01
8.96510005e-01 2.71743000e-01 -1.29532278e+00 -2.30822816e-01
-6.99817538e-02 -1.05547917e+00 -9.39359739e-02 -3.28876227e-01
-7.43470728e-01 9.38317537e-01 5.44796526e-01 3.56858462e-01
8.36628258e-01 1.77817687e-01 3.98990601e-01 1.96397096e-01
9.78724897e-01 -7.37696469e-01 -4.59792197e-01 1.68332055e-01
8.17198396e-01 -9.95570660e-01 3.01480889e-01 -8.45635295e-01
-3.17376703e-01 1.10691750e+00 1.04245469e-01 -3.13281626e-01
7.44037688e-01 1.56738758e-01 -1.46349251e-01 -5.36356747e-01
-3.86707038e-01 2.40897406e-02 1.33724764e-01 5.09689212e-01
-2.66979337e-01 3.49813908e-01 4.25958224e-02 -9.39507857e-02
-2.56509930e-01 -3.24389599e-02 2.11040691e-01 1.08183718e+00
-6.02885306e-01 -1.37547088e+00 -1.08542740e+00 3.41224939e-01
-3.25009674e-02 5.84747732e-01 -2.78255880e-01 7.44652331e-01
1.51184812e-01 7.46254563e-01 3.02759141e-01 -2.76366416e-02
5.66351056e-01 4.93682548e-02 4.24237102e-01 -6.31066263e-01
-2.09109634e-01 2.36789376e-01 -4.75915402e-01 -4.53649968e-01
-6.01788461e-01 -1.14457488e+00 -1.31088412e+00 1.21404435e-02
-5.01507223e-01 2.29063213e-01 1.21131372e+00 5.60665429e-01
3.02646607e-01 -2.52314545e-02 8.50343585e-01 -1.09760237e+00
-5.93834817e-01 -5.99206567e-01 -7.26947010e-01 4.70924795e-01
1.52006775e-01 -7.65554726e-01 -3.10979337e-01 -4.45214212e-02]
|
[8.231078147888184, -3.103618860244751]
|
d5ca28d4-c101-420a-ab88-508c3ede4663
|
wave-u-net-discriminator-fast-and-lightweight
|
2303.13909
| null |
https://arxiv.org/abs/2303.13909v1
|
https://arxiv.org/pdf/2303.13909v1.pdf
|
Wave-U-Net Discriminator: Fast and Lightweight Discriminator for Generative Adversarial Network-Based Speech Synthesis
|
In speech synthesis, a generative adversarial network (GAN), training a generator (speech synthesizer) and a discriminator in a min-max game, is widely used to improve speech quality. An ensemble of discriminators is commonly used in recent neural vocoders (e.g., HiFi-GAN) and end-to-end text-to-speech (TTS) systems (e.g., VITS) to scrutinize waveforms from multiple perspectives. Such discriminators allow synthesized speech to adequately approach real speech; however, they require an increase in the model size and computation time according to the increase in the number of discriminators. Alternatively, this study proposes a Wave-U-Net discriminator, which is a single but expressive discriminator with Wave-U-Net architecture. This discriminator is unique; it can assess a waveform in a sample-wise manner with the same resolution as the input signal, while extracting multilevel features via an encoder and decoder with skip connections. This architecture provides a generator with sufficiently rich information for the synthesized speech to be closely matched to the real speech. During the experiments, the proposed ideas were applied to a representative neural vocoder (HiFi-GAN) and an end-to-end TTS system (VITS). The results demonstrate that the proposed models can achieve comparable speech quality with a 2.31 times faster and 14.5 times more lightweight discriminator when used in HiFi-GAN and a 1.90 times faster and 9.62 times more lightweight discriminator when used in VITS. Audio samples are available at https://www.kecl.ntt.co.jp/people/kaneko.takuhiro/projects/waveunetd/.
|
['Shogo Seki', 'Kou Tanaka', 'Hirokazu Kameoka', 'Takuhiro Kaneko']
|
2023-03-24
| null | null | null | null |
['speech-synthesis']
|
['speech']
|
[ 1.90680072e-01 2.24931791e-01 2.88086325e-01 -5.08813784e-02
-1.14180231e+00 -6.37432992e-01 4.75883752e-01 -6.17483020e-01
-9.71711148e-03 7.46105194e-01 2.47081682e-01 -4.30949479e-01
4.14639354e-01 -8.59692514e-01 -6.23208284e-01 -8.18147719e-01
2.57015705e-01 6.43520802e-02 -1.13556378e-01 -3.48859757e-01
-3.45562428e-01 3.76510113e-01 -1.41180098e+00 1.73417479e-01
1.02978992e+00 1.06512380e+00 4.74992633e-01 9.97646630e-01
1.85888514e-01 5.24209142e-01 -1.12677956e+00 -3.80117029e-01
2.76883900e-01 -9.72175002e-01 -1.63179114e-01 -2.79519141e-01
1.90402672e-01 -2.87833095e-01 -7.33536005e-01 1.09399939e+00
1.09890676e+00 1.57984048e-01 7.58431971e-01 -1.26640511e+00
-6.13324225e-01 8.01432490e-01 -1.57969639e-01 7.23820776e-02
1.03780165e-01 4.95423555e-01 7.70266652e-01 -8.04298878e-01
1.86719194e-01 1.17059314e+00 3.90362561e-01 8.26627254e-01
-8.95827889e-01 -1.14961255e+00 -4.05775398e-01 1.08899567e-02
-1.26995718e+00 -8.03413689e-01 9.59938705e-01 -2.00302765e-01
7.28261769e-01 4.96972054e-01 5.18252194e-01 1.50976479e+00
4.86586206e-02 6.25567973e-01 9.27463472e-01 -3.17566723e-01
2.02694103e-01 2.71381084e-02 -5.75721323e-01 3.62117141e-01
-2.61359304e-01 5.63390970e-01 -4.41422999e-01 2.19393134e-01
7.88842797e-01 -4.59346741e-01 -4.67005998e-01 6.77709162e-01
-8.20541263e-01 7.96939254e-01 3.37924510e-01 4.19665307e-01
-3.01073760e-01 2.63436526e-01 3.53287011e-01 5.24459362e-01
2.73221493e-01 4.00953948e-01 4.79919352e-02 -4.37519222e-01
-1.16529775e+00 1.83551475e-01 5.88143289e-01 1.04913950e+00
3.10534686e-01 1.22729897e+00 -2.78072149e-01 9.15952206e-01
2.09592298e-01 9.83029842e-01 8.79389763e-01 -7.48413742e-01
6.13156199e-01 -5.22743799e-02 -3.65368336e-01 -6.36358798e-01
-2.01749235e-01 -7.27991641e-01 -1.24823308e+00 5.23040891e-01
6.53824285e-02 -5.40651143e-01 -1.04057705e+00 1.85900092e+00
7.27175847e-02 3.14201534e-01 4.21613514e-01 7.67257631e-01
9.56079662e-01 1.18367088e+00 -2.74853736e-01 -1.64738178e-01
1.21559000e+00 -9.70560372e-01 -9.09803569e-01 -1.86986521e-01
1.77665830e-01 -1.03500617e+00 1.12585247e+00 4.13781762e-01
-1.39148355e+00 -1.15315795e+00 -1.24973309e+00 9.11015943e-02
-1.76033199e-01 3.99706155e-01 -1.07632592e-01 8.81839693e-01
-1.08362877e+00 3.44467103e-01 -3.92225176e-01 1.20070428e-01
1.54362489e-02 6.50469363e-02 -7.09369630e-02 4.94183421e-01
-1.56680322e+00 5.78733563e-01 3.34584713e-01 -1.42757129e-02
-1.32211041e+00 -5.77100337e-01 -7.77083218e-01 2.08988100e-01
-9.58327949e-03 -5.78975379e-01 1.40824342e+00 -1.02289677e+00
-2.12631083e+00 2.07317874e-01 9.10502449e-02 -5.77234507e-01
5.43622613e-01 5.99407703e-02 -9.06211615e-01 1.95707351e-01
-2.29515314e-01 4.99234319e-01 1.17448783e+00 -1.02482808e+00
-5.34596324e-01 8.41509104e-02 -1.91249087e-01 1.36056378e-01
-1.00898124e-01 -6.50688633e-02 -4.38236305e-03 -1.19188654e+00
-1.40715182e-01 -8.73423934e-01 1.46125242e-01 -3.38792354e-01
-6.20858133e-01 -9.29580908e-03 9.79660094e-01 -9.70498502e-01
1.30128658e+00 -2.15792298e+00 8.47941414e-02 -9.13564414e-02
3.70753668e-02 7.98567414e-01 -2.18600750e-01 6.71778262e-01
-7.24516660e-02 1.42359182e-01 -3.03976506e-01 -4.34013247e-01
-1.26801273e-02 -4.55728061e-02 -4.61418480e-01 2.29750693e-01
2.83631653e-01 8.66945684e-01 -6.87625408e-01 -1.74449846e-01
2.27276847e-01 5.86873710e-01 -3.65841866e-01 6.47407472e-01
-2.08289642e-02 6.87414289e-01 -1.23987682e-01 4.30152327e-01
5.39105177e-01 4.13697392e-01 -1.16583325e-01 -2.55469173e-01
-1.69620827e-01 4.88590628e-01 -1.06229711e+00 1.47777808e+00
-8.26850116e-01 8.44803512e-01 1.65834427e-01 -8.60745192e-01
1.38787079e+00 9.21705246e-01 -7.03844875e-02 -7.58081257e-01
1.45597652e-01 5.29631138e-01 1.67893037e-01 -4.31452483e-01
2.50498444e-01 -2.56591022e-01 1.35035749e-04 2.67162919e-01
3.48175704e-01 -5.12410522e-01 -1.42965931e-02 -6.47554919e-02
8.74623179e-01 -9.25538093e-02 1.54451370e-01 -1.01592895e-02
5.34452856e-01 -7.29209900e-01 5.67434371e-01 3.33993137e-01
1.51403487e-01 8.06833982e-01 2.25523934e-01 2.03868702e-01
-1.35786688e+00 -1.25615728e+00 1.49839237e-01 6.39922678e-01
-2.46632025e-01 -2.04436958e-01 -8.81266475e-01 -1.94518507e-01
-4.08483416e-01 9.50081527e-01 -1.21875718e-01 -4.14910823e-01
-6.58923388e-01 -2.00182319e-01 1.29740453e+00 4.83976901e-01
6.07081711e-01 -1.18253362e+00 -2.49453977e-01 3.84045631e-01
-1.42744005e-01 -9.91038859e-01 -7.42830694e-01 1.49341315e-01
-5.14810145e-01 -3.78563732e-01 -1.04860485e+00 -8.05817664e-01
3.14074486e-01 -2.60275573e-01 7.16795146e-01 -3.15561324e-01
1.08963847e-01 -2.27991819e-01 -3.92576277e-01 -4.65311646e-01
-1.07539582e+00 1.89480290e-01 2.89621949e-01 1.55803964e-01
-3.22590292e-01 -9.90059316e-01 -4.29321975e-01 3.22006196e-01
-1.04746890e+00 2.54525781e-01 6.36749923e-01 9.32939291e-01
4.05407548e-01 1.21567510e-01 1.10097158e+00 -4.94572401e-01
8.79203200e-01 -4.55401838e-01 -6.19924426e-01 -2.49594357e-02
-3.96805108e-01 -5.72209843e-02 1.44975901e+00 -6.81863129e-01
-1.13839340e+00 -3.70307803e-01 -7.58633971e-01 -6.42178833e-01
-6.35336563e-02 2.95102924e-01 -6.06006384e-01 1.17369600e-01
6.53762460e-01 5.48560143e-01 -6.80331588e-02 -4.46021765e-01
2.89872825e-01 1.17580378e+00 7.51817226e-01 -3.25194150e-01
1.16918027e+00 -2.67720103e-01 -2.29649484e-01 -9.45711732e-01
-1.56146869e-01 1.72342390e-01 8.42638314e-03 -2.49060005e-01
6.44569516e-01 -9.81800437e-01 -5.91691732e-01 7.88753748e-01
-1.19253016e+00 -4.49575126e-01 -3.41020018e-01 7.54663825e-01
-6.78610086e-01 1.95039492e-02 -7.03507900e-01 -8.52531731e-01
-7.58275628e-01 -1.30436647e+00 7.85470843e-01 5.35181940e-01
-1.55716315e-02 -6.43520594e-01 -1.08403862e-02 2.72810400e-01
6.88277662e-01 2.80582845e-01 7.63158560e-01 -5.45502543e-01
-2.93314189e-01 5.61506972e-02 2.36730620e-01 8.79260242e-01
2.73622274e-01 1.62513152e-01 -1.34915578e+00 -3.42533946e-01
2.21810654e-01 -1.42476201e-01 4.24151421e-01 2.68135041e-01
8.37839007e-01 -6.56356156e-01 2.60867238e-01 7.69636571e-01
1.13137758e+00 9.20506775e-01 7.95311332e-01 -2.77400613e-01
6.39419496e-01 2.70696640e-01 1.17926516e-01 2.46560469e-01
-9.47673693e-02 8.21738899e-01 2.36344799e-01 -1.91128090e-01
-7.94593692e-01 -6.03206873e-01 7.11520672e-01 1.55550683e+00
3.28157805e-02 -6.62606955e-01 -5.68703532e-01 3.54378849e-01
-1.25476229e+00 -1.03592241e+00 1.55262619e-01 2.00337410e+00
1.01993418e+00 1.30548343e-01 1.45301953e-01 5.77059507e-01
8.43708634e-01 2.20730931e-01 -6.74098849e-01 -7.40525544e-01
-1.49935663e-01 5.62835813e-01 1.95018113e-01 4.78782684e-01
-5.11703134e-01 7.16976166e-01 5.01956749e+00 1.43431914e+00
-1.45196426e+00 1.67555958e-01 5.91145575e-01 -1.49912030e-01
-4.16898578e-01 -4.06001747e-01 -5.01446664e-01 8.19321811e-01
1.39649808e+00 -4.12349969e-01 7.16329157e-01 6.84733748e-01
3.45319599e-01 4.57126528e-01 -8.69893134e-01 1.03365290e+00
-1.22204348e-01 -1.15671754e+00 4.60356101e-02 -1.01837307e-01
7.09931612e-01 -2.73099691e-01 3.65226865e-01 3.94095421e-01
-1.05037764e-01 -1.21602833e+00 1.09476542e+00 3.09297413e-01
1.38884747e+00 -9.67468679e-01 6.41585112e-01 3.43536049e-01
-1.42533088e+00 9.77153108e-02 -1.15047187e-01 8.74754116e-02
3.91553551e-01 5.53952277e-01 -9.55328226e-01 7.06886649e-01
2.76398361e-01 8.72667059e-02 -1.19038848e-02 6.67748749e-01
-4.96109724e-01 1.04849422e+00 -9.68568623e-02 -2.17413250e-02
3.53487074e-01 -1.84852451e-01 7.67596960e-01 1.15567601e+00
7.87717581e-01 -2.73330398e-02 -3.36592227e-01 9.18159127e-01
-3.36224288e-01 -1.17486835e-01 -5.66345274e-01 -2.48822108e-01
8.41425180e-01 1.04632068e+00 -1.25757620e-01 -2.95664996e-01
-1.45326763e-01 7.22124517e-01 -1.72691852e-01 5.06509602e-01
-1.13537085e+00 -8.68660390e-01 6.81261241e-01 1.35429818e-02
2.12022290e-01 -1.38060823e-01 -3.45365293e-02 -7.02958167e-01
-1.45619232e-02 -1.31361151e+00 -2.36779779e-01 -9.96289909e-01
-1.13011944e+00 1.19895387e+00 -2.98619181e-01 -1.48295867e+00
-7.13853240e-01 -3.19646060e-01 -9.97246742e-01 1.30973160e+00
-1.22684455e+00 -1.02241445e+00 -3.32103580e-01 5.53096712e-01
8.22659373e-01 -5.53678274e-01 7.43100822e-01 4.81327862e-01
-4.96742129e-01 1.01598966e+00 1.40884832e-01 1.89255252e-01
4.57148492e-01 -9.78793859e-01 8.21405470e-01 1.14556623e+00
1.51333183e-01 2.50402749e-01 7.07652688e-01 -3.21946561e-01
-1.27388096e+00 -1.28266585e+00 6.13440096e-01 3.31816763e-01
3.76879215e-01 -4.79791731e-01 -8.25461328e-01 3.49153906e-01
5.26842773e-01 -2.17922226e-01 4.13052440e-01 -7.35254645e-01
-1.27042279e-01 -4.15302575e-01 -1.24320698e+00 6.77923203e-01
8.64489198e-01 -7.53350675e-01 -2.76317835e-01 -1.82317480e-01
1.00006211e+00 -6.41644418e-01 -7.50458956e-01 2.13454619e-01
5.20232081e-01 -9.19993937e-01 6.49869621e-01 -3.23555209e-02
4.79783773e-01 -5.20993948e-01 -2.63276607e-01 -1.75112021e+00
-7.57205039e-02 -1.16678810e+00 -1.47750005e-01 1.54335988e+00
5.43238103e-01 -8.67915571e-01 2.79723734e-01 -9.39462259e-02
-5.95961392e-01 -7.91803420e-01 -1.20912707e+00 -1.15174878e+00
8.48026350e-02 -3.80279124e-01 8.74829113e-01 5.07510006e-01
-2.39111021e-01 2.78503031e-01 -6.17741883e-01 2.61061639e-01
2.92329937e-01 -2.47510359e-01 6.68743610e-01 -7.17660666e-01
-5.30511141e-01 -5.28246641e-01 -2.94351786e-01 -1.01192474e+00
-2.01622061e-02 -9.83418286e-01 7.35226572e-02 -1.29816592e+00
-5.64035594e-01 -3.90192837e-01 -4.88968790e-02 1.64629012e-01
1.60188563e-02 1.56681269e-01 4.52700615e-01 -1.53876349e-01
4.96398509e-01 8.47869217e-01 1.40499854e+00 -2.79954225e-01
-1.80767179e-01 3.60352457e-01 -4.25457954e-01 3.83136064e-01
1.19864941e+00 -4.59411144e-01 -6.66435659e-01 -2.65558749e-01
-2.98776746e-01 4.25866574e-01 1.27772421e-01 -1.36765099e+00
2.35191360e-01 2.13979274e-01 1.57944173e-01 -3.32625866e-01
6.81239963e-01 -5.40228069e-01 6.16770685e-01 5.12969136e-01
-3.03819060e-01 3.42861153e-02 1.94942057e-01 1.34313375e-01
-5.95663309e-01 -3.07023674e-01 9.76503968e-01 1.50881112e-01
-2.33628556e-01 3.00144732e-01 -4.17213172e-01 2.19067872e-01
8.28525424e-01 -1.20294392e-01 -3.60210270e-01 -7.38177299e-01
-4.02929574e-01 -2.84969956e-01 -4.91137877e-02 3.82298112e-01
6.89519107e-01 -1.66785216e+00 -9.46708381e-01 5.08346081e-01
-2.63260990e-01 -1.33544207e-01 4.77498889e-01 3.92060399e-01
-2.95105100e-01 4.02132869e-01 -2.17640102e-01 -2.00005174e-01
-1.11515749e+00 2.82716691e-01 4.48816001e-01 -4.64397222e-02
-5.16032159e-01 8.20965827e-01 3.16231638e-01 -1.71437770e-01
2.37752661e-01 -3.55807066e-01 5.74673153e-02 -5.92907891e-03
4.96422827e-01 5.46916485e-01 5.14452010e-02 -5.79749584e-01
-2.22985558e-02 4.11708415e-01 5.00524580e-01 -4.34365064e-01
1.01509964e+00 1.76040828e-01 3.87132227e-01 3.75987381e-01
1.33504725e+00 2.41572231e-01 -1.11879992e+00 -3.51704148e-05
-7.08287656e-01 -8.17018524e-02 1.30592540e-01 -8.07489157e-01
-1.45930219e+00 1.08740008e+00 5.10161221e-01 3.73469114e-01
1.45866144e+00 -2.78895289e-01 1.23759758e+00 -2.37257585e-01
3.07224303e-01 -8.49981248e-01 1.07808970e-02 2.50153542e-01
1.27922010e+00 -6.24683261e-01 -6.19320989e-01 -1.84755586e-02
-7.86150098e-01 1.21743381e+00 5.14543533e-01 6.88687488e-02
3.89443487e-01 7.68856287e-01 1.64944813e-01 3.93704116e-01
-7.14245677e-01 -7.72765577e-02 2.62126356e-01 7.07590640e-01
2.64667958e-01 2.92811453e-01 -1.58040687e-01 7.68835604e-01
-7.82696903e-01 -2.88811445e-01 5.67941546e-01 3.05543989e-01
-2.26915866e-01 -1.00440395e+00 -3.87771398e-01 1.92326710e-01
-4.08206463e-01 -3.41604888e-01 4.83064987e-02 4.00431365e-01
1.45805776e-01 1.35329247e+00 1.22610159e-01 -7.83745527e-01
4.85077620e-01 3.60809341e-02 2.03579158e-01 -3.36452395e-01
-8.23599279e-01 1.56387538e-01 1.11490116e-01 -2.56022155e-01
1.35735944e-01 -2.09938407e-01 -1.24775910e+00 -3.47037107e-01
-2.42910609e-01 2.70743102e-01 8.11430335e-01 4.98195052e-01
3.01541835e-01 1.06346881e+00 1.16416991e+00 -6.95738554e-01
-7.14163125e-01 -1.25346637e+00 -7.10994005e-01 -4.54779118e-02
4.62459803e-01 -2.63496161e-01 -5.93448699e-01 3.47780772e-02]
|
[15.372417449951172, 6.191558361053467]
|
9914e777-debc-4c41-ab0f-b35fcb7a334a
|
10000-times-accelerated-robust-subset
|
1409.3660
| null |
http://arxiv.org/abs/1409.3660v4
|
http://arxiv.org/pdf/1409.3660v4.pdf
|
10,000+ Times Accelerated Robust Subset Selection (ARSS)
|
Subset selection from massive data with noised information is increasingly
popular for various applications. This problem is still highly challenging as
current methods are generally slow in speed and sensitive to outliers. To
address the above two issues, we propose an accelerated robust subset selection
(ARSS) method. Specifically in the subset selection area, this is the first
attempt to employ the $\ell_{p}(0<p\leq1)$-norm based measure for the
representation loss, preventing large errors from dominating our objective. As
a result, the robustness against outlier elements is greatly enhanced.
Actually, data size is generally much larger than feature length, i.e. $N\gg
L$. Based on this observation, we propose a speedup solver (via ALM and
equivalent derivations) to highly reduce the computational cost, theoretically
from $O(N^{4})$ to $O(N{}^{2}L)$. Extensive experiments on ten benchmark
datasets verify that our method not only outperforms state of the art methods,
but also runs 10,000+ times faster than the most related method.
|
['Feiyun Zhu', 'Xinliang Zhu', 'Ying Wang', 'Shiming Xiang', 'Chunhong Pan', 'Bin Fan']
|
2014-09-12
| null | null | null | null |
['temporal-action-proposal-generation', 'music-modeling', 'nested-named-entity-recognition']
|
['computer-vision', 'music', 'natural-language-processing']
|
[ 1.60365283e-01 -4.48900670e-01 -1.51268477e-02 -1.95744321e-01
-8.82427812e-01 -3.08123082e-01 -8.93339664e-02 5.49515665e-01
-5.25725245e-01 8.72369707e-01 -3.30220908e-01 -9.95468870e-02
-3.23275596e-01 -8.89365017e-01 -5.44125617e-01 -8.15534413e-01
-1.82496831e-01 2.50213593e-01 3.53258431e-01 -2.03090400e-01
5.35615563e-01 3.57865214e-01 -1.79982615e+00 -3.36186022e-01
1.28294003e+00 1.50103664e+00 -1.30570650e-01 -3.39781903e-02
-9.31394398e-02 3.82273704e-01 -5.34506023e-01 -3.28675956e-01
5.13241351e-01 -3.54734123e-01 -4.28113729e-01 1.51937753e-01
2.91601688e-01 -1.18543021e-01 -2.65402272e-02 1.43084049e+00
5.72783351e-01 4.52302486e-01 2.29294822e-01 -1.10793829e+00
-2.02658009e-02 5.85263431e-01 -1.10294139e+00 4.17138971e-02
1.35434240e-01 -7.20836967e-02 1.03031611e+00 -1.06440222e+00
3.07023615e-01 9.28541958e-01 5.80180645e-01 -7.03154951e-02
-9.95722473e-01 -9.12040472e-01 5.62073708e-01 4.63247709e-02
-1.85184276e+00 -2.74063259e-01 5.86686254e-01 -5.59723042e-02
4.83961940e-01 6.74235761e-01 4.50014502e-01 4.44811493e-01
-1.23590954e-01 7.93764472e-01 9.11097884e-01 -3.18708390e-01
4.12879199e-01 6.90522715e-02 2.78794616e-01 4.98508722e-01
8.40588331e-01 -3.06498170e-01 -4.65566725e-01 -5.15275717e-01
4.23253179e-01 2.04027742e-01 -4.68417555e-01 -2.19195411e-01
-1.01455605e+00 7.91365266e-01 9.03883874e-02 5.97647540e-02
-3.02152336e-01 6.33944720e-02 6.41252100e-01 3.88795674e-01
5.42595565e-01 3.78132343e-01 -4.64118034e-01 -1.56440169e-01
-9.36791956e-01 3.82819235e-01 4.63825583e-01 1.07334065e+00
7.68186688e-01 5.49062267e-02 1.14440490e-02 1.00457036e+00
2.76246555e-02 3.06370795e-01 2.77930766e-01 -7.08656371e-01
7.22084045e-01 9.60996687e-01 1.42915130e-01 -1.30633676e+00
-3.85259986e-01 -7.35096574e-01 -1.18355262e+00 -8.95139948e-02
4.03675824e-01 -1.67151801e-02 -4.79215413e-01 1.62014401e+00
3.88532698e-01 1.82650283e-01 -3.84438694e-01 6.02605164e-01
2.60831714e-01 4.36273009e-01 -1.21821120e-01 -7.83979833e-01
1.17644346e+00 -3.49835992e-01 -5.81529319e-01 -1.72454610e-01
6.72876656e-01 -7.85276473e-01 1.09272289e+00 6.21931374e-01
-9.97307301e-01 -3.52718562e-01 -1.05561256e+00 4.97027069e-01
-1.81126650e-02 1.04070157e-01 7.15480804e-01 5.87515473e-01
-5.98131537e-01 5.62691808e-01 -6.71317458e-01 -7.49254748e-02
5.22421658e-01 6.53787553e-01 -2.31941164e-01 -1.47953406e-01
-9.45682466e-01 2.95921117e-01 4.19548512e-01 3.37279916e-01
-1.97882757e-01 -5.93514800e-01 -6.86632395e-01 -7.05084950e-02
8.17676365e-01 -2.33143404e-01 6.79421246e-01 -4.62678611e-01
-1.04835534e+00 3.06170821e-01 -1.24426834e-01 -2.96437919e-01
5.75622380e-01 -3.52423608e-01 -3.72491598e-01 -1.30417779e-01
1.45233795e-01 -7.68322721e-02 7.39632845e-01 -1.13866723e+00
-8.55171859e-01 -7.51903176e-01 -2.57920146e-01 7.28617534e-02
-7.49090075e-01 2.39443943e-01 -5.22746146e-01 -7.98441947e-01
6.06111705e-01 -9.75640416e-01 -6.31459296e-01 -1.48117870e-01
-5.69881797e-01 -2.62150019e-01 5.35025775e-01 -2.41419911e-01
1.75308871e+00 -2.31774449e+00 -2.33390242e-01 7.80734837e-01
3.33637625e-01 3.92762154e-01 1.94726661e-01 3.56243610e-01
1.24548532e-01 1.46556973e-01 -4.11377877e-01 -1.52755067e-01
-1.63341418e-01 -1.68438092e-01 -2.08048329e-01 7.01786935e-01
-2.31565177e-01 2.74671227e-01 -6.88494503e-01 -3.26112539e-01
-7.76904598e-02 1.14152759e-01 -5.54503024e-01 -4.53938544e-02
1.17169701e-01 1.81257948e-01 -5.97595870e-01 7.77980208e-01
9.86237347e-01 -2.51584560e-01 8.17679092e-02 -1.88533083e-01
-2.44803712e-01 -1.53122276e-01 -1.93488216e+00 1.41470528e+00
-2.22497344e-01 7.68673196e-02 -3.21313366e-02 -1.02361977e+00
1.24468541e+00 6.81275427e-02 6.92299664e-01 -4.99669582e-01
2.59956211e-01 7.31391370e-01 -8.48315582e-02 -1.47640526e-01
5.39101601e-01 -2.35506054e-02 -3.53986651e-01 3.58362496e-01
-5.92695475e-01 1.46769658e-01 3.70453507e-01 -3.41031849e-02
1.31085420e+00 -3.11838657e-01 4.64197457e-01 -3.52452159e-01
6.39561117e-01 -1.34794816e-01 1.18627012e+00 8.15522552e-01
-2.03172058e-01 5.60474813e-01 6.17818713e-01 -3.99000674e-01
-6.59214795e-01 -6.72514558e-01 -2.24367723e-01 8.71381283e-01
4.90089804e-01 -6.41203225e-01 -5.72008014e-01 -4.62467074e-01
9.77810249e-02 6.87158644e-01 -4.26299304e-01 -2.46326074e-01
-7.89484143e-01 -1.24121785e+00 3.63629729e-01 5.47836900e-01
4.77331191e-01 -8.18626404e-01 -3.90839905e-01 3.48209202e-01
4.67264280e-02 -8.51999223e-01 -4.45442796e-01 1.60659611e-01
-1.03714252e+00 -8.88666630e-01 -5.03538847e-01 -6.06151819e-01
9.48199868e-01 4.15341169e-01 8.79975677e-01 2.85270780e-01
-3.49328905e-01 -3.56908292e-01 -7.20483899e-01 -5.46654999e-01
1.59452036e-01 2.59000286e-02 4.57983345e-01 1.33613750e-01
4.96496916e-01 -6.14572763e-01 -7.76528060e-01 4.95697886e-01
-9.74645674e-01 -4.23192948e-01 6.10470295e-01 8.88580501e-01
1.05209780e+00 5.09374380e-01 6.90048873e-01 -1.06831038e+00
5.49861252e-01 -3.68955433e-01 -8.92968297e-01 1.41855434e-01
-7.85594344e-01 -1.57000363e-01 1.05282223e+00 -2.33954981e-01
-6.95536852e-01 2.07125232e-01 -1.37979880e-01 -3.40783060e-01
4.70179096e-02 5.25572240e-01 -2.97984093e-01 -1.24936506e-01
3.93635780e-01 3.45446795e-01 -1.12229563e-01 -5.74370682e-01
-2.42631193e-02 6.84238970e-01 2.94301957e-01 -4.28215355e-01
8.28929007e-01 6.25742018e-01 3.40525843e-02 -6.26363516e-01
-6.55644000e-01 -4.78285998e-01 -3.96617025e-01 2.61039108e-01
6.95265904e-02 -8.19142520e-01 -9.15223122e-01 3.09670568e-01
-6.18817091e-01 2.76932120e-01 -2.94316918e-01 6.28654659e-01
-3.18978220e-01 4.87549871e-01 -3.07816893e-01 -9.46117640e-01
-5.01850724e-01 -1.21772337e+00 6.76401913e-01 8.46965015e-02
-1.18541315e-01 -3.01973104e-01 -2.10413650e-01 1.67443633e-01
3.29140335e-01 4.46888894e-01 8.58641088e-01 -8.75819147e-01
-4.93971974e-01 -6.96961105e-01 -3.07810545e-01 3.31889510e-01
1.25891015e-01 -1.55683354e-01 -3.92538428e-01 -5.81896305e-01
1.80392861e-01 -1.32427132e-02 7.47391522e-01 9.72461700e-02
1.60388136e+00 -2.71751225e-01 -2.87645638e-01 6.08360708e-01
1.54417288e+00 2.01695725e-01 6.38415396e-01 3.73222470e-01
6.42762899e-01 3.79073828e-01 1.20536292e+00 1.13653815e+00
-7.01754773e-03 6.71002507e-01 4.56635863e-01 -8.72131623e-03
4.00233895e-01 3.71699259e-02 2.52153426e-01 9.13647056e-01
5.61886169e-02 -2.73259044e-01 -8.24883401e-01 5.11442244e-01
-1.88698399e+00 -5.10421395e-01 -3.37934136e-01 2.83489180e+00
7.48710275e-01 3.05582792e-01 6.09421581e-02 4.49733943e-01
8.31648171e-01 2.54010912e-02 -6.61946356e-01 -2.33008042e-02
-1.53079659e-01 2.96858579e-01 6.20771587e-01 -3.78067270e-02
-1.02087176e+00 6.47402167e-01 4.95197010e+00 1.33764350e+00
-7.18091071e-01 -3.09461296e-01 7.25625575e-01 -3.94758970e-01
-1.58711657e-01 7.64432102e-02 -9.34799612e-01 8.47304523e-01
5.75052440e-01 -3.11128557e-01 1.65502876e-01 8.80097628e-01
2.76394010e-01 -2.76163638e-01 -8.29420507e-01 1.26353157e+00
7.74272606e-02 -9.73857403e-01 -6.23514764e-02 1.93038404e-01
6.36041760e-01 -2.29705110e-01 1.93616729e-02 7.11021200e-02
9.24702659e-02 -8.16753209e-01 5.69088578e-01 -2.56202221e-02
8.65432084e-01 -1.29016125e+00 1.00061321e+00 5.22039115e-01
-1.51873791e+00 -2.52465546e-01 -6.98861778e-01 1.38160408e-01
1.05418630e-01 1.16800380e+00 -2.27288887e-01 6.98272288e-01
9.49753881e-01 4.47588503e-01 -4.90233183e-01 1.17666006e+00
1.97088376e-01 3.97047341e-01 -6.74844980e-01 -2.81807669e-02
-6.64121807e-02 -2.86677569e-01 6.65168226e-01 8.49801779e-01
6.75806403e-01 2.53181756e-01 4.86978173e-01 2.90915787e-01
-2.30282784e-01 6.65933430e-01 -4.10392135e-01 3.91384184e-01
8.43629241e-01 1.08952332e+00 -9.77019489e-01 -1.46644622e-01
-3.56193811e-01 7.26614058e-01 2.05819443e-01 -6.69835787e-03
-7.72800028e-01 -1.02017963e+00 7.04373717e-01 2.54124850e-01
3.81106853e-01 -1.21375062e-01 -4.43980634e-01 -1.09184635e+00
4.42636251e-01 -1.09037435e+00 5.55266142e-01 1.05402343e-01
-1.27262473e+00 6.93593562e-01 -2.54288316e-01 -1.80422175e+00
1.41424537e-01 -2.78846800e-01 -2.24040985e-01 5.86408198e-01
-1.16814566e+00 -4.37753677e-01 -4.20195669e-01 5.21632552e-01
3.48422199e-01 -9.75582525e-02 5.76102793e-01 6.14256084e-01
-9.79513466e-01 1.03535974e+00 4.49575305e-01 -1.14205457e-01
7.17015445e-01 -8.24535966e-01 1.60518140e-01 9.71872628e-01
-2.00497255e-01 8.38206768e-01 7.07084239e-01 -5.72806537e-01
-1.42814004e+00 -1.11430025e+00 5.61243236e-01 -3.04767899e-02
4.78918701e-01 -2.50852764e-01 -9.56401467e-01 2.13717282e-01
-4.03489888e-01 2.62149781e-01 9.29843724e-01 2.06053555e-02
-2.06526890e-01 -5.97011089e-01 -1.28894651e+00 6.84434056e-01
1.18141067e+00 2.56661698e-02 9.28769410e-02 2.84774840e-01
4.65793014e-01 -2.98483938e-01 -1.08705366e+00 8.10389280e-01
3.20216119e-01 -1.19034588e+00 8.43855202e-01 -1.76469490e-01
-1.67614341e-01 -5.34030020e-01 -3.24585974e-01 -8.99587512e-01
-8.94876122e-02 -7.96591759e-01 -1.88751742e-01 1.23910487e+00
3.35802108e-01 -8.81003559e-01 7.51924157e-01 5.36149263e-01
-4.56616608e-03 -1.10680234e+00 -9.99302089e-01 -8.68001640e-01
-1.53517097e-01 -4.43492711e-01 7.35388935e-01 7.68290162e-01
-2.14389995e-01 -1.36659458e-01 -4.89000231e-01 3.69748808e-02
7.46605039e-01 2.91224360e-01 7.57208049e-01 -1.42371786e+00
-2.42684409e-01 -3.52614850e-01 -5.34985185e-01 -9.88332927e-01
-2.93998241e-01 -5.29625297e-01 1.27949908e-01 -9.72844779e-01
1.58354759e-01 -1.01692009e+00 -6.88577831e-01 3.86508137e-01
-5.66259265e-01 2.68059760e-01 -2.56000403e-02 4.04859304e-01
-8.39243948e-01 5.82855165e-01 7.75754511e-01 3.45232397e-01
-3.64561200e-01 2.36110657e-01 -1.02552104e+00 9.87256169e-01
8.45430017e-01 -6.46314263e-01 -3.02891135e-01 -2.24841505e-01
5.21815717e-01 -8.39804038e-02 -1.80531472e-01 -1.08104491e+00
1.97465107e-01 -2.73062259e-01 1.90558240e-01 -7.21635401e-01
1.00696482e-01 -8.30228031e-01 4.65941355e-02 3.99965912e-01
-7.70271942e-02 1.37180522e-01 2.23868750e-02 7.46131957e-01
-3.80628467e-01 -1.61672950e-01 7.57682264e-01 -1.41992941e-01
-5.35903037e-01 5.65894902e-01 6.38490841e-02 -1.70871988e-02
1.20802951e+00 -2.82235205e-01 -2.08560482e-01 -1.30063519e-01
-3.30600500e-01 3.55155677e-01 5.22055626e-01 -3.59527357e-02
6.36562526e-01 -1.31680858e+00 -7.63046384e-01 1.98538125e-01
2.77561694e-01 4.03533220e-01 2.59060204e-01 1.14337599e+00
-5.19066274e-01 1.29935369e-01 2.18847215e-01 -4.57133502e-01
-1.18374228e+00 4.83614296e-01 -2.74613827e-01 -3.36532682e-01
-5.74927568e-01 1.17465317e+00 -1.02512576e-01 -1.65373385e-01
3.84554118e-01 5.23441583e-02 6.34959489e-02 1.66370898e-01
6.95409000e-01 8.64964962e-01 2.37070516e-01 -4.22441512e-01
-5.92170835e-01 5.27328670e-01 -3.96300107e-01 1.76990882e-01
1.33802021e+00 -2.29884595e-01 -4.13252950e-01 2.64940768e-01
1.25300121e+00 3.45555544e-01 -1.00073278e+00 -3.30815643e-01
1.59472048e-01 -8.70415509e-01 -2.09151074e-01 -1.48301095e-01
-1.10330880e+00 4.31725860e-01 5.40314734e-01 3.34119767e-01
1.53190267e+00 -2.54521161e-01 1.07493985e+00 3.68380427e-01
7.23176003e-01 -1.42024899e+00 -1.49802491e-01 2.57085681e-01
6.47511661e-01 -1.17101514e+00 4.91639227e-01 -7.07033396e-01
-5.21867812e-01 8.16071272e-01 7.29606092e-01 -3.26816499e-01
5.70935845e-01 1.11603409e-01 -1.79084376e-01 -3.28740943e-03
-4.17678714e-01 -5.78709319e-02 -5.60027622e-02 -2.29395181e-02
3.48798126e-01 -2.02851519e-02 -7.34699249e-01 7.23140836e-01
-3.90621871e-02 -1.09854095e-01 2.98666030e-01 1.07412350e+00
-5.75526595e-01 -1.33493567e+00 -5.00490427e-01 9.91280258e-01
-5.98325133e-01 -1.04400016e-01 1.83331072e-02 6.34081602e-01
2.03609139e-01 1.00407934e+00 -1.07871614e-01 -4.67508882e-01
5.40904403e-01 -2.86678255e-01 3.02222986e-02 -4.68896627e-01
-4.93758589e-01 2.08957702e-01 -2.37803310e-01 -6.75186038e-01
-5.60137555e-02 -8.54593217e-01 -1.32092714e+00 -4.25237298e-01
-6.55359566e-01 3.04104865e-01 3.95532161e-01 6.08915687e-01
4.18677956e-01 2.48675242e-01 1.02935028e+00 -3.75508398e-01
-9.51058090e-01 -6.13681376e-01 -8.89740765e-01 3.70231837e-01
-4.04802077e-02 -7.22247362e-01 -5.55138826e-01 -5.47594011e-01]
|
[6.900694370269775, 4.596887588500977]
|
44e16135-e83f-45d7-a183-e5f5a6a76ce7
|
understanding-unfairness-via-training-concept
|
2306.17828
| null |
https://arxiv.org/abs/2306.17828v1
|
https://arxiv.org/pdf/2306.17828v1.pdf
|
Understanding Unfairness via Training Concept Influence
|
Knowing the causes of a model's unfairness helps practitioners better understand their data and algorithms. This is an important yet relatively unexplored task. We look into this problem through the lens of the training data - one of the major sources of unfairness. We ask the following questions: how would a model's fairness performance change if, in its training data, some samples (1) were collected from a different (e.g. demographic) group, (2) were labeled differently, or (3) some features were changed? In other words, we quantify the fairness influence of training samples by counterfactually intervening and changing samples based on predefined concepts, i.e. data attributes such as features (X), labels (Y), or sensitive attributes (A). To calculate a training sample's influence on the model's unfairness w.r.t a concept, we first generate counterfactual samples based on the concept, i.e. the counterfactual versions of the sample if the concept were changed. We then calculate the resulting impact on the unfairness, via influence function, if the counterfactual samples were used in training. Our framework not only helps practitioners understand the observed unfairness and repair their training data, but also leads to many other applications, e.g. detecting mislabeling, fixing imbalanced representations, and detecting fairness-targeted poisoning attacks.
|
['Yang Liu', 'Yuanshun Yao']
|
2023-06-30
| null | null | null | null |
['fairness', 'fairness']
|
['computer-vision', 'miscellaneous']
|
[ 3.60324055e-01 2.85381675e-01 -4.77242500e-01 -5.97122848e-01
-2.89879918e-01 -6.15954161e-01 5.47774434e-01 3.11390191e-01
-6.53481781e-01 1.16260850e+00 5.19530118e-01 -4.83059168e-01
-1.59612596e-01 -8.26275349e-01 -8.78123939e-01 -8.17840159e-01
2.23394230e-01 3.10129702e-01 -4.50406432e-01 8.39882791e-02
7.20712483e-01 3.52805227e-01 -1.46241653e+00 2.72092491e-01
1.08859813e+00 6.19741797e-01 -6.84076846e-01 1.85950980e-01
2.47234628e-02 7.70955086e-01 -7.88111627e-01 -1.04452515e+00
5.25087357e-01 -4.97800797e-01 -9.02016103e-01 -4.50805873e-02
4.97594982e-01 -3.48300934e-01 -5.38380928e-02 1.55428123e+00
2.65929222e-01 1.37316465e-01 5.79018891e-01 -1.65350425e+00
-5.09454072e-01 8.90024304e-01 -6.54407859e-01 1.31967440e-01
8.33142772e-02 5.56800544e-01 1.09048152e+00 -1.97895199e-01
4.34116304e-01 1.47077334e+00 5.52170038e-01 6.84712648e-01
-1.62426484e+00 -1.15615273e+00 1.67504191e-01 8.22084770e-02
-9.47558761e-01 -7.05058455e-01 5.66094935e-01 -5.58367431e-01
1.98005602e-01 6.31600320e-01 4.45278406e-01 1.01913106e+00
3.21671218e-01 3.88322562e-01 1.31663764e+00 -2.72199661e-01
5.72925031e-01 3.36365342e-01 1.62584081e-01 2.65352845e-01
8.46671939e-01 7.45290041e-01 -2.49422163e-01 -6.22331977e-01
1.39825165e-01 7.39051774e-02 -2.55319625e-01 -3.17392975e-01
-9.15794551e-01 1.11961043e+00 5.77596843e-01 8.68873894e-02
-4.75039095e-01 1.34461269e-01 4.62101877e-01 4.40586567e-01
5.40669918e-01 1.11835325e+00 -6.51769698e-01 1.01901755e-01
-9.24835861e-01 6.03441715e-01 4.63300049e-01 1.92657709e-01
9.35803831e-01 -2.17418075e-01 -4.95639145e-01 3.59715313e-01
-7.56932646e-02 3.52403611e-01 6.26127481e-01 -1.08299923e+00
7.47904956e-01 5.28734863e-01 4.66111541e-01 -1.08597922e+00
-1.73919410e-01 -3.50142628e-01 -6.57824457e-01 3.45416874e-01
7.45618939e-01 -3.56609851e-01 -6.10449433e-01 2.19480968e+00
3.56715471e-01 -8.39122534e-02 -7.49438778e-02 9.72182453e-01
8.80832225e-02 2.46196240e-01 4.63150173e-01 -4.66581941e-01
8.89561176e-01 -3.56548041e-01 -6.91268265e-01 -1.65571108e-01
8.66576374e-01 -5.18535674e-01 1.17200780e+00 8.06968436e-02
-9.71086562e-01 -1.96779430e-01 -5.69305480e-01 4.34754968e-01
-3.16307694e-01 -5.79036653e-01 7.41633713e-01 1.16519845e+00
-6.40011609e-01 1.07775867e+00 3.43440287e-02 -9.29772556e-02
8.82681191e-01 2.32254863e-01 -3.92145365e-01 -1.66717350e-01
-1.52982140e+00 7.77146876e-01 2.42837653e-01 -4.06665653e-01
-7.88563311e-01 -1.07174003e+00 -5.95770180e-01 4.48982745e-01
6.41572535e-01 -8.55970144e-01 8.55083048e-01 -1.54694831e+00
-9.05249119e-01 8.56008232e-01 1.33670643e-01 -6.51218772e-01
8.46777737e-01 2.12222133e-02 -3.16817045e-01 -4.84461218e-01
3.77721727e-01 3.44905317e-01 9.65432942e-01 -1.27215827e+00
-5.51631331e-01 -8.74194801e-01 7.96934664e-02 1.72858492e-01
-1.47598743e-01 5.69342896e-02 8.18271101e-01 -8.42244029e-01
-3.77156049e-01 -9.09485519e-01 -3.17022324e-01 5.69917746e-02
-8.44200253e-01 2.77385920e-01 3.75787437e-01 -3.30579609e-01
1.30462909e+00 -2.11952353e+00 -3.71500283e-01 4.35918212e-01
2.82213390e-01 5.26109755e-01 -7.44162500e-02 -9.01601240e-02
-5.16462266e-01 8.64285588e-01 -3.85450184e-01 1.48527414e-01
4.46639098e-02 -8.67756307e-02 -4.19545323e-01 5.43932319e-01
5.59059419e-02 5.70232511e-01 -8.81491721e-01 -8.24816227e-02
1.64554864e-01 -2.09677622e-01 -9.72406685e-01 2.08017305e-01
8.55845362e-02 1.45531178e-01 -2.20599979e-01 1.12713374e-01
8.96342576e-01 3.73771578e-01 8.52939561e-02 -2.37061962e-01
1.24609924e-03 3.76868725e-01 -1.02180564e+00 6.81892931e-01
-1.17070504e-01 3.13819885e-01 -3.78653675e-01 -1.10080349e+00
7.12440968e-01 1.05836786e-01 2.45875776e-01 -6.84588075e-01
7.93897733e-02 2.09568053e-01 4.18450475e-01 -5.07026196e-01
4.61074203e-01 -9.69351530e-01 -8.19936246e-02 8.09101224e-01
-4.25100833e-01 1.06947333e-01 -2.43243769e-01 9.20123458e-02
9.58015442e-01 -6.23847544e-01 6.76103830e-01 -3.36233407e-01
2.86961883e-01 -1.94456622e-01 9.32292402e-01 9.18838978e-01
-6.66728973e-01 2.77730405e-01 8.08119237e-01 -5.62692523e-01
-9.81494606e-01 -1.06864715e+00 -2.46867537e-01 9.21933949e-01
-1.54817253e-01 1.55749783e-01 -6.70469224e-01 -1.07352591e+00
6.16737604e-01 1.31842864e+00 -9.90116477e-01 -9.75835621e-01
-2.60611232e-02 -9.58136380e-01 6.03947997e-01 9.28439498e-02
5.05369008e-01 -8.43205512e-01 -7.24233210e-01 -9.79252309e-02
-1.89141244e-01 -3.75736833e-01 -6.53106213e-01 -2.51698703e-01
-1.00227499e+00 -1.26180875e+00 -2.09935665e-01 2.72799402e-01
6.77004695e-01 2.33666942e-01 1.15138042e+00 1.73272386e-01
6.48960285e-03 -1.37591541e-01 -1.46291569e-01 -6.58126175e-01
-5.08511841e-01 -4.46593344e-01 2.91145176e-01 2.30289862e-01
3.97364944e-01 -2.53420025e-01 -4.91576612e-01 2.39563718e-01
-1.06251705e+00 -3.28986973e-01 3.15991610e-01 8.71462226e-01
1.09531954e-01 2.44025141e-01 7.59541094e-01 -1.64099503e+00
9.25883472e-01 -5.99370241e-01 -3.17693323e-01 4.38766271e-01
-9.99800205e-01 1.54630467e-01 7.23602712e-01 -4.38686460e-01
-1.20892310e+00 -4.06178057e-01 3.03244203e-01 -3.30604702e-01
-2.78488368e-01 1.91873819e-01 -5.60309649e-01 4.24647778e-01
9.35530543e-01 -4.22115207e-01 -3.92509578e-03 -2.55385935e-01
4.82923329e-01 6.00876033e-01 2.25786135e-01 -7.31098354e-01
6.71056449e-01 4.28616583e-01 -1.08770490e-01 -1.04453303e-01
-1.02865756e+00 -1.07510127e-02 -1.99899539e-01 2.32311040e-02
4.50141728e-01 -3.91489714e-01 -6.61098182e-01 2.98570096e-01
-8.77433956e-01 -5.84508367e-02 -6.76919580e-01 4.58445340e-01
-2.98170954e-01 7.80320540e-02 1.02602385e-01 -9.95381355e-01
-1.46004662e-01 -1.00729549e+00 4.21394885e-01 2.25544646e-01
-5.79537749e-01 -9.52206969e-01 1.19687319e-01 5.67245662e-01
4.12006825e-01 3.00285935e-01 1.26001310e+00 -9.59728777e-01
1.39678633e-02 -3.78730178e-01 -2.21309230e-01 5.19999385e-01
3.50555360e-01 1.28826797e-01 -1.15901518e+00 -2.90774643e-01
1.76242217e-01 -2.96125084e-01 6.32355392e-01 6.04286194e-01
1.39364672e+00 -1.19668770e+00 -7.01177716e-02 3.76188397e-01
1.16068959e+00 1.81888565e-01 7.68834054e-01 8.44270736e-02
4.21964467e-01 1.04476964e+00 6.44068480e-01 6.84934556e-01
-2.94835325e-02 6.65683091e-01 5.39246559e-01 4.59565111e-02
4.45118278e-01 -6.12670004e-01 3.51771504e-01 -1.03904486e-01
2.10713074e-01 1.53339580e-01 -6.22538328e-01 3.79517704e-01
-1.48581445e+00 -1.36102378e+00 -6.50727898e-02 2.75246501e+00
9.42661166e-01 4.43655215e-02 1.69087306e-01 3.44204545e-01
1.15100706e+00 1.80492669e-01 -8.85069132e-01 -9.00415182e-01
1.34476542e-01 7.40858912e-02 5.32681942e-01 6.00686014e-01
-8.08246017e-01 8.04858804e-01 5.43524313e+00 6.50813580e-01
-1.05285621e+00 -1.65771410e-01 1.28520310e+00 -4.14343506e-01
-9.26421404e-01 3.63591492e-01 -2.52176434e-01 7.84645319e-01
8.05833936e-01 -9.69478250e-01 5.96608698e-01 5.65743685e-01
5.43344796e-01 -1.84415519e-01 -1.20188463e+00 5.61311781e-01
-1.66886553e-01 -1.14802039e+00 3.54153812e-01 2.80458331e-01
7.80599535e-01 -5.97304165e-01 2.08712861e-01 2.92806476e-01
6.89275444e-01 -1.17858243e+00 7.89206505e-01 4.88639593e-01
7.98274338e-01 -8.53188634e-01 1.02073753e+00 4.34035420e-01
-1.00479968e-01 -2.23741665e-01 -4.26845342e-01 -4.61330086e-01
-2.89512575e-01 1.31671274e+00 -7.76663780e-01 2.72366792e-01
4.37584668e-01 2.97256261e-01 -4.46882218e-01 9.14959967e-01
-1.66299596e-01 7.55172968e-01 2.02924430e-01 2.67577648e-01
-6.95581958e-02 -1.56536743e-01 5.38069844e-01 6.36717558e-01
-2.26252507e-02 -9.50421300e-03 -4.17900622e-01 1.26596189e+00
-5.80790162e-01 5.86136542e-02 -7.70452976e-01 -8.98346230e-02
7.92105317e-01 8.19827795e-01 -1.18827254e-01 -3.73743623e-01
1.71763286e-01 4.96749192e-01 1.67071447e-01 3.76566947e-01
-7.14023888e-01 -2.18371987e-01 1.22436047e+00 9.18402821e-02
-6.00740492e-01 8.42955291e-01 -5.63127160e-01 -1.14772117e+00
-1.85286820e-01 -1.06670845e+00 5.99813700e-01 -5.72136104e-01
-1.63997114e+00 -6.44946769e-02 -2.80108869e-01 -9.91403162e-01
-1.78353544e-02 -1.31087840e-01 -7.50193715e-01 1.07955635e+00
-1.27493942e+00 -2.66347677e-01 1.34843871e-01 4.67798561e-01
-1.02623656e-01 -1.22556105e-01 7.70312428e-01 1.58209652e-02
-6.62722170e-01 8.90484810e-01 -4.76288535e-02 5.37870005e-02
9.40175712e-01 -1.05631065e+00 1.46751881e-01 8.41241598e-01
-2.71373689e-02 7.10465789e-01 8.12598109e-01 -5.98907411e-01
-4.07917470e-01 -1.17264926e+00 1.15562224e+00 -3.96395534e-01
4.38413382e-01 1.66448250e-01 -1.01628935e+00 5.82029283e-01
-3.80789936e-01 -7.38689080e-02 9.59282815e-01 2.35224724e-01
-7.28488386e-01 -1.91870496e-01 -2.08018422e+00 5.92220962e-01
9.92754459e-01 -4.33388352e-01 -6.77662432e-01 1.25843482e-02
6.93685770e-01 4.02573347e-02 -3.93433779e-01 2.45342746e-01
5.95058560e-01 -1.38937545e+00 7.36317873e-01 -1.63873982e+00
7.32258379e-01 9.29814577e-03 -2.74040103e-01 -1.84658575e+00
-4.62236345e-01 -3.23308319e-01 4.20032471e-01 1.29409754e+00
3.86047423e-01 -8.99775684e-01 6.70520484e-01 1.28095341e+00
2.80602306e-01 -4.11644608e-01 -1.04219604e+00 -4.40991789e-01
4.26906943e-01 -2.85664260e-01 1.32637119e+00 1.55822289e+00
2.17520576e-02 5.55892102e-02 -4.67298418e-01 -9.65353549e-02
7.01569140e-01 1.90306142e-01 6.87924504e-01 -1.26235282e+00
7.75760412e-02 -4.62082177e-01 -2.49389708e-01 -6.83287345e-03
4.68938887e-01 -9.27318156e-01 -2.87913322e-01 -7.67957747e-01
5.80849111e-01 -6.41530752e-01 -1.99744076e-01 5.44548094e-01
-7.43321240e-01 -2.57780582e-01 5.94438195e-01 1.27950236e-01
-6.32247627e-02 5.18960059e-01 1.07445467e+00 9.21707042e-03
-1.44254584e-02 2.28587881e-01 -1.47053969e+00 6.46901250e-01
9.72511709e-01 -9.07697320e-01 -2.55970001e-01 -1.18869767e-01
1.72246501e-01 -5.21019148e-03 6.34883046e-01 -5.71537614e-01
-2.25368381e-01 -8.34956348e-01 2.37201810e-01 4.26938832e-01
-3.57816488e-01 -9.57023799e-01 2.30624691e-01 9.85328674e-01
-1.08949661e+00 3.36232521e-02 -8.90991613e-02 3.85975301e-01
7.31232464e-02 -3.85109544e-01 9.88137603e-01 -2.56755650e-01
2.63193063e-02 9.76250768e-02 -9.32180509e-02 5.31496823e-01
9.99797404e-01 1.52658805e-01 -8.13145339e-01 -2.80740291e-01
-2.89996713e-01 4.36164439e-02 7.50016809e-01 2.78366655e-01
4.08947557e-01 -1.40399504e+00 -8.78835857e-01 1.57170489e-01
1.35748401e-01 -4.22345459e-01 2.21838981e-01 4.57912058e-01
1.42772570e-01 1.44826591e-01 -5.02241910e-01 8.16801041e-02
-1.07108974e+00 6.51660800e-01 5.16216576e-01 -2.84844220e-01
1.32532701e-01 6.17517829e-01 4.78482276e-01 -5.30928373e-01
-1.44269928e-01 1.27823791e-02 9.31907669e-02 1.94221318e-01
5.58278322e-01 6.40370190e-01 -9.03263167e-02 -5.36441147e-01
-2.17698574e-01 4.38158177e-02 -1.37831777e-01 5.20126484e-02
1.08461082e+00 4.83688749e-02 -2.14077592e-01 3.35318029e-01
1.18181825e+00 9.93751585e-02 -9.67248559e-01 -8.96489546e-02
4.13110182e-02 -1.47553813e+00 -5.56831434e-02 -1.07999575e+00
-1.21034086e+00 7.90786147e-01 5.43767989e-01 1.58009589e-01
9.08574760e-01 -3.47230524e-01 2.50685900e-01 -2.89385952e-02
4.12598550e-01 -9.95194137e-01 -1.82126865e-01 -1.29106864e-01
7.76779234e-01 -1.15548062e+00 1.59824520e-01 -1.67306304e-01
-8.76547337e-01 5.64759672e-01 3.92085910e-01 7.07744434e-02
3.80951852e-01 -2.35661462e-01 3.43748555e-03 2.22265888e-02
-7.66912222e-01 2.05195665e-01 1.87951829e-02 5.78762174e-01
4.56177235e-01 7.43994951e-01 -5.03786564e-01 7.15060890e-01
-4.31744665e-01 1.95394576e-01 7.78936088e-01 2.32533038e-01
-2.27340326e-01 -8.95095646e-01 -6.57893300e-01 1.20559251e+00
-4.20626760e-01 -1.73937172e-01 -6.66177154e-01 4.27202702e-01
4.84631151e-01 8.18151951e-01 1.44875720e-01 -5.37259877e-01
5.75190067e-01 1.49284050e-01 7.71370083e-02 -6.10759854e-01
-8.33496094e-01 -7.84251511e-01 9.39406604e-02 -7.17014015e-01
-3.45030755e-01 -9.89543915e-01 -8.01597297e-01 -9.93238091e-01
-2.83183843e-01 3.29596817e-01 2.49136016e-01 9.93477106e-01
3.37985992e-01 1.00273937e-01 1.05685067e+00 -3.48391026e-01
-1.07970858e+00 -7.41270423e-01 -6.43880129e-01 1.22062099e+00
3.16252232e-01 -7.17047811e-01 -9.54084992e-01 -3.51105273e-01]
|
[8.907381057739258, 5.37070369720459]
|
6534b4c5-a6dc-47ee-86f6-606d9178e8bd
|
mixture-of-graphs-zero-shot-relational
| null | null |
https://openreview.net/forum?id=RdQXgI1pXt1
|
https://openreview.net/pdf?id=RdQXgI1pXt1
|
Mixture-of-Graphs: Zero-shot Relational Learning for Knowledge Graph by Fusing Ontology and Textual Experts
|
Knowledge Graph Embedding (KGE) have been proposed and succeed utilized to knowledge Graph Completion (KGC). But dominant KGE models often fail in zero-shot relational learning because they cannot learn effective representations for unseen relations. Previous studies mainly separately utilize the textual description of relation and its neighbor relations to represent unseen relations. In fact, the semantics of a relation can be expressed by three kinds of graphs: factual graph, ontology graph and textual description graph, and they can complement and enhance each other. Therefore, to obtain more accurate representation of relation in zero-shot learning, we propose the mixture-of-graphs (MoG) experts to improve the effect of current KGE for unseen relations. We build multi-aspect associations between seen and unseen relations which will be used directly to guide previous KGE methods such as TransE and RotatE on zero-shot relational learning. The experiments on multiple public datasets verify the effectiveness of the proposed method, which improves the state-of-the-art zero-shot relational learning method by 12.84% in Hits@10 on average.
|
['Anonymous']
|
2021-11-16
| null | null | null |
acl-arr-november-2021-11
|
['relational-reasoning']
|
['natural-language-processing']
|
[-3.97191107e-01 5.60240686e-01 -5.01522303e-01 -1.89503849e-01
-3.13326985e-01 -8.99230912e-02 6.16123915e-01 4.29908007e-01
-7.11189508e-02 6.26734257e-01 5.14728844e-01 -5.93783893e-02
-6.23781800e-01 -1.44744217e+00 -5.53866684e-01 -3.82333606e-01
-7.39983469e-03 8.19437742e-01 6.99549496e-01 -5.60069323e-01
-3.54068577e-01 8.21067765e-02 -1.47147882e+00 3.74944955e-01
6.96261644e-01 6.38448715e-01 -2.13325948e-01 2.24711463e-01
-5.04725039e-01 1.27694964e+00 -4.23411787e-01 -1.03672767e+00
-3.10082555e-01 -9.75005329e-02 -9.77871358e-01 -2.55763561e-01
-2.46163429e-04 -6.83477819e-02 -1.18416250e+00 1.11960673e+00
4.80640948e-01 4.70266581e-01 7.50482798e-01 -1.51121366e+00
-1.36979616e+00 1.01805592e+00 -2.88961321e-01 2.01208994e-01
4.85762864e-01 -4.71134722e-01 1.29159999e+00 -1.04087031e+00
9.45677757e-01 1.69872141e+00 4.68981951e-01 2.98373044e-01
-8.20979416e-01 -6.26592636e-01 -6.83639059e-03 9.13479269e-01
-1.81142628e+00 -2.72785157e-01 6.95202410e-01 -2.35491902e-01
1.24799490e+00 1.35978907e-01 6.19828820e-01 9.95260358e-01
-2.21736133e-01 6.76518321e-01 4.48424041e-01 -4.59330529e-01
-6.79592714e-02 1.69895947e-01 4.96728063e-01 7.64631867e-01
7.06130624e-01 -2.13710859e-01 -6.19034708e-01 -8.34748596e-02
5.38454592e-01 1.80271342e-01 -4.31328297e-01 -5.54319024e-01
-8.46041679e-01 7.59135544e-01 7.94753969e-01 3.45581114e-01
-1.67803228e-01 -3.37044783e-02 4.40834790e-01 2.69656241e-01
4.29399103e-01 1.73187748e-01 -2.46509299e-01 2.30386063e-01
2.67307982e-02 -8.67839379e-04 8.95983219e-01 1.42916155e+00
9.62515533e-01 -1.71748653e-01 -2.88777351e-01 1.08224583e+00
3.39278489e-01 4.43833843e-02 2.80473471e-01 -3.85657549e-01
6.38592839e-01 1.32956326e+00 -2.63380021e-01 -1.48816526e+00
-1.47120461e-01 -2.79664427e-01 -8.48084986e-01 -6.86674654e-01
-3.70896906e-01 2.28123665e-01 -9.61033165e-01 1.20892668e+00
5.25743127e-01 6.77322030e-01 5.08666635e-01 6.57734275e-01
1.81822658e+00 5.81974089e-01 1.69993669e-01 -1.54429451e-01
1.42287350e+00 -8.94533277e-01 -1.24154651e+00 -1.23336084e-01
9.04988110e-01 -3.69606733e-01 7.57832885e-01 -2.31883407e-01
-3.60681236e-01 -4.37834799e-01 -1.04991269e+00 -2.40277514e-01
-1.06934428e+00 -2.95235574e-01 1.06643248e+00 4.72327977e-01
-6.50410116e-01 5.31943262e-01 -5.31345665e-01 -5.87651849e-01
5.75104415e-01 1.06105015e-01 -5.37630200e-01 -6.05989873e-01
-1.89913166e+00 1.00255716e+00 1.21823728e+00 -1.55800926e-02
-8.07213843e-01 -5.51780283e-01 -1.24474931e+00 2.63919920e-01
1.16481602e+00 -5.59023380e-01 4.68319386e-01 2.04625562e-01
-6.65360391e-01 7.28775203e-01 -3.47909294e-02 -1.67026028e-01
-2.17421174e-01 -2.25616574e-01 -1.07850242e+00 1.71567544e-01
2.34877206e-02 1.44673452e-01 3.41705024e-01 -1.48430562e+00
-4.15467858e-01 -4.03489590e-01 4.73958045e-01 2.96405017e-01
-5.97589016e-01 -2.51022428e-01 -9.00960743e-01 -4.80200648e-01
3.02432895e-01 -5.80343962e-01 1.89053580e-01 -5.61444879e-01
-4.43097413e-01 -7.37503231e-01 1.01350248e+00 -4.57454532e-01
1.54951704e+00 -1.94405282e+00 1.41317070e-01 8.71233046e-02
6.61452413e-01 4.43246335e-01 -1.63575754e-01 7.26460397e-01
-2.23056078e-01 2.94920236e-01 2.83075720e-01 5.05050682e-02
-8.29262212e-02 8.49730372e-01 -1.79832876e-01 1.12188896e-02
2.87965708e-03 1.42358112e+00 -1.37554407e+00 -9.68500435e-01
1.61296636e-01 5.43463767e-01 1.37805985e-02 1.25150949e-01
8.30260441e-02 -3.29086900e-01 -5.46969354e-01 8.81596446e-01
4.13486928e-01 -4.12242681e-01 5.95794976e-01 -7.10440755e-01
8.01846743e-01 6.14313036e-03 -1.11869645e+00 1.53785181e+00
-1.11354113e-01 1.35319635e-01 -7.00496674e-01 -8.94849241e-01
1.13703489e+00 5.08293033e-01 2.91749656e-01 -3.73833895e-01
2.01062977e-01 -1.67129621e-01 -7.35856742e-02 -6.88412726e-01
5.52109540e-01 -1.60220519e-01 2.66099554e-02 5.19980043e-02
7.18481362e-01 1.11045688e-01 2.49520645e-01 9.46422875e-01
1.17103612e+00 8.87906253e-02 5.17574549e-01 1.49875388e-01
3.83979023e-01 -1.99327633e-01 7.52162099e-01 4.64953870e-01
9.32124723e-03 1.44671664e-01 7.21996784e-01 -3.78160685e-01
-5.38155973e-01 -1.13252592e+00 1.89460829e-01 9.10360038e-01
6.35007322e-01 -1.12522471e+00 -3.76320556e-02 -9.96223629e-01
1.18892193e-01 8.87064040e-01 -6.96041703e-01 -7.84843206e-01
-1.43753290e-01 -6.95884466e-01 4.55234200e-01 6.01596713e-01
4.14424986e-01 -1.02010465e+00 3.60065639e-01 2.51964867e-01
-3.14244300e-01 -1.37421060e+00 3.58510539e-02 -6.69543147e-02
-4.18420553e-01 -1.52728188e+00 -1.00113429e-01 -8.34168971e-01
4.97230589e-01 5.16031206e-01 1.30761445e+00 3.08903456e-01
-2.04444841e-01 4.72032249e-01 -9.92072821e-01 -1.63812369e-01
-8.73003975e-02 3.22938040e-02 1.33105159e-01 -2.93160379e-02
8.65110457e-01 -7.64443576e-01 -8.03206265e-02 2.42922723e-01
-8.84516120e-01 -2.78659910e-01 3.46351415e-01 7.89500475e-01
5.99278450e-01 7.61794508e-01 5.97288966e-01 -1.44095421e+00
7.86979079e-01 -6.78097963e-01 -6.12115394e-03 9.35344696e-01
-8.79698217e-01 8.27394873e-02 2.28678599e-01 -4.02809739e-01
-1.22919750e+00 -4.62692052e-01 4.18776512e-01 -1.04758584e+00
3.71747375e-01 9.35673475e-01 -4.82524931e-01 -5.87257445e-02
6.55359626e-01 -2.59285718e-02 -5.58703780e-01 -2.55138129e-01
8.95736039e-01 4.54005927e-01 5.01138031e-01 -6.17814481e-01
9.24222827e-01 2.07161158e-01 1.36724457e-01 -6.01525128e-01
-1.22967994e+00 -7.89275169e-01 -7.15927184e-01 -1.41536370e-01
7.92837918e-01 -9.67571437e-01 -7.57375240e-01 -1.11582480e-01
-1.11827290e+00 4.09131169e-01 -2.76062846e-01 4.63623524e-01
-3.19007672e-02 4.56591010e-01 -6.90448821e-01 -7.89518356e-01
-1.74823463e-01 -4.82369840e-01 7.97484457e-01 2.87058830e-01
-3.17839608e-02 -1.15735734e+00 7.01186284e-02 4.28094834e-01
2.58240793e-02 3.04538816e-01 1.39419389e+00 -8.54089975e-01
-6.76495194e-01 -2.72227615e-01 -4.65131134e-01 -1.91120543e-02
2.77382672e-01 -1.67597786e-01 -8.57118130e-01 -5.44475429e-02
-4.89584535e-01 -5.79668522e-01 1.12551272e+00 -3.13351631e-01
6.69772685e-01 -2.71907032e-01 -7.69998670e-01 4.33560431e-01
1.61925006e+00 1.19730935e-01 9.10425961e-01 3.76652628e-02
1.38357604e+00 6.22672498e-01 8.35930526e-01 2.94417083e-01
8.10819387e-01 4.60544050e-01 3.95923823e-01 3.72059166e-01
-2.69196749e-01 -7.20781386e-01 -2.15807036e-01 1.28359854e+00
-4.90246385e-01 -3.32772136e-01 -9.09247220e-01 7.02813029e-01
-2.19871306e+00 -1.03330338e+00 -2.86815226e-01 1.88427746e+00
7.72681057e-01 1.27815321e-01 -3.73240858e-01 6.28794953e-02
9.49935615e-01 3.30396861e-01 -2.62452215e-01 8.02500695e-02
-2.90899992e-01 1.56548440e-01 2.17530906e-01 4.75781620e-01
-1.07052398e+00 1.51559544e+00 5.27660275e+00 1.13985384e+00
-2.54477244e-02 2.05040693e-01 -5.53858504e-02 2.74772555e-01
-6.26666963e-01 3.45444143e-01 -8.74496758e-01 -9.97685120e-02
6.50601804e-01 -5.88390350e-01 3.65850061e-01 9.92717862e-01
-7.97477305e-01 2.59504080e-01 -1.02160728e+00 1.33562958e+00
2.35830635e-01 -1.29816508e+00 5.74340522e-01 -1.57372534e-01
7.82543838e-01 -4.93482798e-01 -5.01759648e-01 1.22907317e+00
7.13498116e-01 -1.00851083e+00 -2.13107347e-01 7.25151598e-01
6.85011923e-01 -8.94517303e-01 1.12631893e+00 1.73660114e-01
-1.78561747e+00 1.49935246e-01 -8.26669037e-01 1.68174088e-01
8.66897479e-02 4.95792329e-01 -8.81776273e-01 1.52697968e+00
6.24219775e-01 1.05659354e+00 -6.84436560e-01 3.88350457e-01
-6.69967711e-01 3.32430214e-01 5.60742803e-02 -3.68278176e-02
-1.43191013e-02 -9.35517401e-02 3.14745605e-01 7.89728820e-01
3.96666154e-02 7.40300596e-01 2.60413527e-01 7.58658528e-01
-3.65901798e-01 1.17890343e-01 -7.76788712e-01 -5.11061192e-01
9.17587101e-01 1.27019083e+00 -4.49622273e-01 -5.39425850e-01
-7.13654935e-01 7.75404990e-01 8.74078095e-01 5.04721224e-01
-6.59551799e-01 -6.69385672e-01 3.53801578e-01 -1.86204091e-01
2.54009098e-01 1.32664785e-01 4.47711647e-01 -1.43458235e+00
-2.44393777e-02 -4.50279325e-01 1.06571984e+00 -9.27861392e-01
-1.65356481e+00 5.44285297e-01 2.41451621e-01 -9.22200620e-01
2.40924326e-03 -2.69587398e-01 -3.32022727e-01 6.56909466e-01
-1.34338844e+00 -1.45579576e+00 -5.43865502e-01 8.28685105e-01
6.34315759e-02 -3.77590477e-01 1.06970751e+00 4.27141666e-01
-5.04399061e-01 4.43134248e-01 -3.79817337e-01 4.01120663e-01
4.59030211e-01 -1.13288653e+00 2.09574714e-01 5.67278028e-01
4.44593549e-01 7.76089609e-01 2.69745797e-01 -1.27873170e+00
-1.51066041e+00 -1.26880801e+00 9.84248817e-01 -5.78792393e-01
8.92824590e-01 -9.96733159e-02 -1.29041946e+00 1.11781442e+00
-1.31359056e-01 4.72436637e-01 7.26320267e-01 8.33733261e-01
-7.20752954e-01 -2.58655936e-01 -6.94728136e-01 6.37479663e-01
1.49034333e+00 -8.89603972e-01 -1.16531289e+00 3.75536114e-01
1.32115412e+00 -1.53868243e-01 -1.24363399e+00 7.50258327e-01
1.50321975e-01 -6.19368076e-01 1.16435361e+00 -1.00667918e+00
2.91498840e-01 -3.46134216e-01 -4.05056447e-01 -1.20693052e+00
-5.49327075e-01 -1.68419331e-01 -8.73683214e-01 1.64160979e+00
1.84913680e-01 -4.80440021e-01 7.93261349e-01 3.91941220e-01
-2.99047888e-03 -9.17043209e-01 -6.59509778e-01 -1.02142322e+00
-5.59352577e-01 -2.88828373e-01 7.83217192e-01 1.59020281e+00
3.57034534e-01 1.01623988e+00 -4.53789771e-01 3.78589600e-01
6.11258626e-01 3.26773226e-02 6.96599185e-01 -1.37857795e+00
-1.45440206e-01 5.20485416e-02 -1.01363742e+00 -4.85306054e-01
3.21694285e-01 -1.26260459e+00 -5.10971546e-01 -2.09730482e+00
4.49516535e-01 -3.14259708e-01 -5.93443334e-01 6.74892068e-01
-6.65154636e-01 -3.35849494e-01 -9.23746899e-02 1.22065336e-01
-8.79615426e-01 1.07667065e+00 1.34138060e+00 -4.53174263e-01
-3.26667167e-02 -5.50605893e-01 -6.88107848e-01 4.73795146e-01
3.01203161e-01 -5.20102859e-01 -1.09841537e+00 -9.20153558e-02
4.43129808e-01 1.22874171e-01 2.62930006e-01 -6.63864613e-01
5.43521106e-01 -2.28402321e-03 6.90426156e-02 -5.85087717e-01
6.11900985e-01 -8.65227938e-01 3.51965636e-01 3.58649790e-02
-1.06857769e-01 -5.15748620e-01 -1.40623108e-01 1.09850621e+00
-4.99298245e-01 -2.76564956e-01 1.42506376e-01 -1.92021087e-01
-1.30877304e+00 6.33398771e-01 6.30490959e-01 3.51361215e-01
9.82757449e-01 5.73772639e-02 -7.88734555e-01 -4.71220702e-01
-9.72878635e-01 4.26694304e-01 -9.08327550e-02 6.94978118e-01
1.08417666e+00 -1.89655113e+00 -6.18366539e-01 -2.75738001e-01
7.50526607e-01 4.88914922e-02 4.83681172e-01 4.60515440e-01
-1.71013221e-01 1.71227843e-01 1.77760139e-01 -6.94743171e-02
-1.18947148e+00 1.22468686e+00 4.65258360e-02 -3.78171235e-01
-7.99207270e-01 1.00511789e+00 8.20232630e-02 -3.62655133e-01
1.39595076e-01 1.93735316e-01 -6.60116017e-01 2.33576804e-01
4.43605870e-01 3.69690001e-01 -7.57527351e-03 -5.72611809e-01
-5.26811481e-01 4.47368801e-01 -2.89171219e-01 3.16834927e-01
1.18831384e+00 7.54889920e-02 -4.07676399e-01 6.34096265e-01
1.07013381e+00 -2.92322129e-01 -3.96749735e-01 -7.95727909e-01
1.28611058e-01 -6.74133718e-01 8.42157286e-03 -4.42879647e-01
-1.03078997e+00 6.16232753e-01 -4.84883673e-02 2.64473408e-01
6.36699975e-01 4.80535030e-01 6.92869604e-01 7.02241838e-01
6.99781537e-01 -8.23966920e-01 8.06988627e-02 3.98464590e-01
7.62194514e-01 -1.12055969e+00 3.59855831e-01 -1.16144812e+00
-7.67652631e-01 8.84867549e-01 8.08748603e-01 8.49181339e-02
8.46074104e-01 -1.53573856e-01 -3.42416257e-01 -7.98940420e-01
-8.77870619e-01 -7.60179460e-01 6.16571784e-01 9.03470039e-01
1.79086283e-01 2.54906058e-01 -1.85585111e-01 8.64598572e-01
-1.81031395e-02 -2.22930402e-01 2.28500098e-01 8.64793718e-01
-3.40315610e-01 -1.00280666e+00 2.05477372e-01 5.32903373e-01
1.51572645e-01 -3.23246032e-01 -5.85724771e-01 9.84977186e-01
2.08542630e-01 1.14410067e+00 -4.25411552e-01 -9.01384890e-01
4.66270953e-01 1.85322627e-01 3.82599503e-01 -9.11645651e-01
4.68993559e-02 -4.06638324e-01 6.27084374e-01 -4.19175297e-01
-4.03738737e-01 3.36786844e-02 -1.33559239e+00 -4.10783559e-01
-5.88009298e-01 2.92216450e-01 -2.25356609e-01 9.46591556e-01
1.80303767e-01 9.25539136e-01 1.37422845e-01 2.04151385e-02
-4.90167290e-02 -9.66887653e-01 -1.04735351e+00 7.79574811e-01
-4.73206490e-01 -1.22745359e+00 -2.58992821e-01 -1.63600951e-01]
|
[8.813823699951172, 7.969298839569092]
|
35688835-c1b9-4e50-9db5-5febe096b430
|
native-language-identification-with-attention
| null | null |
https://aclanthology.org/2020.icon-main.35
|
https://aclanthology.org/2020.icon-main.35.pdf
|
Native-Language Identification with Attention
|
The paper explores how an attention-based approach can increase performance on the task of native-language identification (NLI), i.e., to identify an author’s first language given information expressed in a second language. Previously, Support Vector Machines have consistently outperformed deep learning-based methods on the TOEFL11 data set, the de facto standard for evaluating NLI systems. The attention-based system BERT (Bidirectional Encoder Representations from Transformers) was first tested in isolation on the TOEFL11 data set, then used in a meta-classifier stack in combination with traditional techniques to produce an accuracy of 0.853. However, more labelled NLI data is now available, so BERT was also trained on the much larger Reddit-L2 data set, containing 50 times as many examples as previously used for English NLI, giving an accuracy of 0.902 on the Reddit-L2 in-domain test scenario, improving the state-of-the-art by 21.2 percentage points.
|
['Björn Gambäck', 'Stian Steinbakken']
| null | null | null | null |
icon-2020-12
|
['native-language-identification']
|
['natural-language-processing']
|
[-6.73528090e-02 8.21059048e-02 -3.04251820e-01 -1.80645838e-01
-1.12965059e+00 -9.36147451e-01 9.17515755e-01 1.29051089e-01
-7.08630323e-01 7.74079800e-01 2.54350811e-01 -8.26028347e-01
6.79452792e-02 -2.10087985e-01 -4.59252447e-01 -1.79436043e-01
2.86716312e-01 9.87323046e-01 -3.52448016e-01 -1.17454417e-02
3.51055078e-02 4.27246422e-01 -1.16702425e+00 2.96462119e-01
8.46541703e-01 8.60343099e-01 -1.15336820e-01 8.90201092e-01
-4.18480545e-01 9.26781416e-01 -9.21032965e-01 -6.50413215e-01
1.32778287e-01 5.71619906e-02 -1.24231684e+00 -5.51547229e-01
8.55370402e-01 -2.60179073e-01 -5.64434528e-01 7.85580099e-01
6.13257885e-01 -1.47331759e-01 7.88096666e-01 -9.17354226e-01
-1.08720231e+00 1.11385441e+00 -3.24963868e-01 4.79865044e-01
5.92434406e-01 3.47141288e-02 1.24633002e+00 -1.01027906e+00
5.61024964e-01 1.31859601e+00 6.96962893e-01 6.09928250e-01
-1.24898255e+00 -1.04706061e+00 -1.75203934e-01 9.91623476e-02
-1.47070992e+00 -7.72750437e-01 3.20266217e-01 -4.00576472e-01
1.56675482e+00 1.20859794e-01 1.72787577e-01 1.48908019e+00
-3.16089272e-01 1.25566006e+00 1.18600452e+00 -7.39449620e-01
-3.97233605e-01 5.85378289e-01 5.12080431e-01 2.58715719e-01
2.26714343e-01 1.20007254e-01 -4.94780034e-01 7.64136463e-02
2.32313141e-01 -6.36695921e-01 -3.12062889e-01 3.31480771e-01
-1.64894664e+00 9.06440914e-01 1.66551620e-01 9.67557669e-01
-7.79127851e-02 -1.89386100e-01 5.97995937e-01 5.57585895e-01
5.68743527e-01 1.04412282e+00 -8.84284019e-01 -3.12952191e-01
-1.12888801e+00 1.97688863e-01 1.13176894e+00 9.63656902e-01
2.96764463e-01 1.91862851e-01 -2.61751086e-01 9.41891491e-01
-9.56084877e-02 3.21143270e-01 9.28397238e-01 -5.96276939e-01
7.49583900e-01 5.59808850e-01 2.40611807e-02 -3.73732656e-01
-2.42682427e-01 -7.01296687e-01 -7.02236056e-01 5.11434898e-02
9.36149061e-01 -2.86179662e-01 -7.96347439e-01 1.65832794e+00
-4.13612127e-01 -3.00944746e-01 3.66698533e-01 4.90140617e-01
9.78281319e-01 9.26684618e-01 1.05475262e-01 2.23644838e-01
1.13577497e+00 -8.06940138e-01 -4.61836904e-01 -2.16335863e-01
1.09421337e+00 -7.32626379e-01 1.01205456e+00 4.77986276e-01
-8.60220015e-01 -7.74966538e-01 -8.66435170e-01 -3.29908907e-01
-7.05114365e-01 5.50133049e-01 5.27722478e-01 8.26220095e-01
-1.29115212e+00 3.25907558e-01 -8.30640420e-02 -5.01375258e-01
3.67187977e-01 4.14647639e-01 -6.44136429e-01 -2.58225612e-02
-1.46611714e+00 9.58402276e-01 5.04785717e-01 -3.81062448e-01
-7.60447562e-01 -9.97696280e-01 -7.30531216e-01 1.89721569e-01
-1.90640625e-03 1.49691096e-02 1.28240275e+00 -1.11018693e+00
-1.36023760e+00 1.40038669e+00 3.41185741e-02 -7.53609776e-01
5.67353904e-01 -1.57346144e-01 -7.11092532e-01 -3.13147873e-01
3.00102562e-01 8.01637173e-01 3.48881215e-01 -9.29619431e-01
-7.52851427e-01 -2.03009948e-01 1.94720477e-02 -7.54210055e-02
-5.96273065e-01 3.61411363e-01 -1.54382512e-01 -4.43103760e-01
-6.63001239e-01 -8.15217197e-01 4.42322314e-01 -5.97647727e-01
-6.47601247e-01 -7.64370084e-01 6.53250575e-01 -1.25032198e+00
1.31433630e+00 -2.02950168e+00 3.01691983e-02 -3.34098786e-02
2.40405872e-01 8.36663902e-01 -2.52892435e-01 1.52306721e-01
-4.87830907e-01 5.76292276e-01 -6.10542074e-02 -4.79502052e-01
-1.13627245e-03 -3.04773182e-01 -1.78143084e-01 1.55917034e-01
2.61369437e-01 1.01353610e+00 -7.84388304e-01 -1.75765082e-01
-2.81833974e-03 2.13336855e-01 -2.15644404e-01 1.98599532e-01
6.96097985e-02 3.61356765e-01 1.27367005e-01 5.89902520e-01
2.16712669e-01 -2.26588205e-01 1.29802348e-02 1.19374223e-01
-3.81091863e-01 6.12758517e-01 -7.08714664e-01 1.58847809e+00
-8.46571803e-01 1.36544847e+00 -4.65860078e-03 -1.00638473e+00
1.17408574e+00 6.33514762e-01 2.30266735e-01 -5.83615363e-01
9.00658369e-02 5.38126349e-01 6.35639489e-01 5.00788502e-02
3.21104497e-01 3.34151149e-01 -3.01482111e-01 3.34530354e-01
3.19586247e-01 3.16363156e-01 2.25157693e-01 2.24791706e-01
1.12594664e+00 -1.51548550e-01 2.96704084e-01 -6.84916496e-01
8.97245288e-01 3.70512456e-02 1.78366348e-01 9.54091012e-01
-1.42634988e-01 2.21901447e-01 3.95597547e-01 -2.69986629e-01
-1.27534091e+00 -5.62730491e-01 -3.50776851e-01 1.11617446e+00
-7.15638161e-01 -2.58996606e-01 -8.20531666e-01 -8.37915599e-01
3.08743864e-01 1.26393950e+00 -5.17533600e-01 8.26128945e-02
-3.85034531e-01 -3.79379332e-01 9.11961198e-01 5.44313550e-01
4.06804085e-01 -1.10516751e+00 -4.53592837e-03 3.68234634e-01
-6.58257380e-02 -1.11654937e+00 -3.19960505e-01 3.69831830e-01
-4.88014929e-02 -7.91063488e-01 -1.16924977e+00 -8.87909830e-01
1.59662385e-02 -2.12237686e-01 1.40542281e+00 4.39706780e-02
-2.81410158e-01 4.21820015e-01 -1.61623478e-01 -4.44022745e-01
-8.86356473e-01 8.18794310e-01 1.52032793e-01 -9.44410637e-02
8.28132808e-01 1.20965138e-01 1.71162277e-01 -1.79269552e-01
-2.85344630e-01 -8.73174891e-02 4.80588585e-01 1.10095632e+00
-1.02248259e-01 -1.94004580e-01 6.84769273e-01 -8.63839388e-01
6.42925560e-01 -3.08210373e-01 -6.51622057e-01 3.13615054e-01
-6.10734284e-01 4.75402772e-02 8.55841458e-01 -7.78173506e-01
-9.20401931e-01 -1.77448764e-01 -2.45094880e-01 -2.81360865e-01
-3.76726091e-01 4.93781179e-01 -4.18451816e-01 -1.75257787e-01
4.38845664e-01 2.25511983e-01 -1.79970562e-01 -7.94217348e-01
1.84347987e-01 1.36858475e+00 6.47909462e-01 -3.41101706e-01
5.40783346e-01 -5.02132893e-01 -5.42403817e-01 -9.66614664e-01
-8.06696177e-01 -5.47404945e-01 -7.76929796e-01 2.72463262e-01
6.79279804e-01 -1.02997971e+00 -9.36124742e-01 6.27783895e-01
-1.25234926e+00 -6.41167641e-01 9.24469728e-04 4.76847440e-01
-2.46977463e-01 8.51640925e-02 -6.94085598e-01 -7.99933434e-01
-4.82803047e-01 -1.27553618e+00 8.96112680e-01 -2.00893462e-01
-7.21894681e-01 -1.10252523e+00 -4.86307293e-02 4.52211499e-01
4.00744408e-01 -2.66365290e-01 1.08529890e+00 -1.49232090e+00
1.45190032e-02 -5.49729347e-01 -5.09982526e-01 5.08000433e-01
-2.57700592e-01 -1.98275700e-01 -1.32146001e+00 -4.08040255e-01
-4.30609763e-01 -4.96236831e-01 7.40665793e-01 -7.60137737e-02
1.16197836e+00 -3.10062528e-01 -3.93563777e-01 5.27329803e-01
1.22574413e+00 3.65104020e-01 4.03426111e-01 5.78773439e-01
8.31517100e-01 4.30148870e-01 3.56095051e-03 -1.48065621e-03
2.85400718e-01 8.29374194e-01 -2.54753232e-01 -4.64032330e-02
-4.39590812e-01 -2.42720813e-01 3.76411378e-01 6.72981799e-01
4.19899702e-01 -7.07352161e-01 -1.58475757e+00 7.78174222e-01
-1.25111103e+00 -8.59119892e-01 -2.59537194e-02 2.21959782e+00
9.47816968e-01 1.71058923e-01 1.52197316e-01 4.73652273e-01
5.93520641e-01 -1.81085199e-01 -3.73139888e-01 -6.87675357e-01
-1.89713463e-01 2.76935041e-01 8.34011853e-01 8.39823484e-01
-1.29059565e+00 1.27226365e+00 6.53504229e+00 1.00478184e+00
-1.24926209e+00 1.49597660e-01 9.39114332e-01 2.08130047e-01
-2.95463532e-01 -2.32117280e-01 -1.47078645e+00 6.86415970e-01
1.54115117e+00 -4.77290303e-01 5.87401986e-01 6.12554193e-01
-5.08411944e-01 2.21184105e-01 -1.35212290e+00 1.18166852e+00
1.51678920e-01 -1.10536134e+00 6.62754327e-02 3.37728828e-01
5.50890148e-01 1.99112162e-01 1.45048335e-01 7.92313576e-01
3.92768323e-01 -1.39115870e+00 6.68802857e-01 1.41004845e-01
1.41330159e+00 -8.28206599e-01 8.44043612e-01 4.80639786e-01
-7.08618045e-01 -2.56513506e-01 -2.08856776e-01 1.20086875e-02
-2.75033653e-01 3.55909348e-01 -1.30102110e+00 2.65714139e-01
5.93156517e-01 7.52244294e-01 -7.47631848e-01 6.53995395e-01
-1.47000179e-01 1.10511291e+00 -3.30375642e-01 -3.92346054e-01
5.24802208e-01 3.79457802e-01 6.99457586e-01 1.50194836e+00
2.62540489e-01 -3.42524081e-01 -5.67455366e-02 7.81675160e-01
-3.67507428e-01 2.61506498e-01 -8.89176369e-01 -4.99570370e-01
4.76407886e-01 1.19704199e+00 -3.94769795e-02 -6.38336182e-01
-6.38789415e-01 1.17498386e+00 6.49556458e-01 3.06589931e-01
-5.09430707e-01 -7.10937142e-01 5.87267637e-01 3.03668845e-02
7.41223693e-02 -1.00802153e-01 -1.21973939e-01 -1.13830435e+00
-4.52028573e-01 -1.26161897e+00 2.66609579e-01 -7.48570621e-01
-1.31821334e+00 9.21841025e-01 -1.72121972e-01 -7.73350656e-01
-6.55461311e-01 -1.19488776e+00 -3.09868246e-01 1.48217475e+00
-1.28068852e+00 -1.18622243e+00 1.42761812e-01 2.97217965e-01
5.82182527e-01 -9.79988039e-01 1.17092216e+00 4.24070090e-01
-7.57623672e-01 1.14772475e+00 5.53070068e-01 8.40584576e-01
7.76822925e-01 -1.56476188e+00 7.91500568e-01 8.08230817e-01
6.35312498e-01 5.79255521e-01 2.52019316e-01 -4.10871327e-01
-1.09308922e+00 -8.40619981e-01 1.55261457e+00 -6.79996133e-01
9.11520064e-01 -4.80904758e-01 -8.36324155e-01 8.94073606e-01
3.89462113e-01 -4.86689121e-01 6.99539065e-01 3.45015764e-01
-4.37424242e-01 -4.37803678e-02 -1.06116140e+00 4.98855948e-01
9.75412250e-01 -8.00064325e-01 -5.29398859e-01 4.19425637e-01
7.78222263e-01 -1.80681899e-01 -1.11219561e+00 1.08746760e-01
3.56484652e-01 -5.10956049e-01 1.03703225e+00 -7.06023157e-01
2.60990858e-01 1.93466589e-01 -1.36228621e-01 -1.46618366e+00
-7.34623313e-01 -5.62908411e-01 1.16682112e-01 1.69400489e+00
7.78885543e-01 -7.36554027e-01 3.73403698e-01 5.44740021e-01
5.63173518e-02 -4.47998434e-01 -1.09082019e+00 -9.19937909e-01
5.92155755e-01 -4.93027538e-01 6.42174482e-01 1.07654464e+00
-6.70733452e-02 7.17075348e-01 -1.53570965e-01 -2.83213228e-01
3.65523070e-01 -2.67544150e-01 8.14715981e-01 -1.45845079e+00
-1.05862625e-01 -8.87475371e-01 -4.28642035e-01 -8.07538927e-01
8.87548208e-01 -1.54492939e+00 -3.05918515e-01 -1.12829912e+00
1.57830164e-01 -4.20805514e-01 -2.67078161e-01 7.13974774e-01
-8.94114226e-02 1.84775993e-01 3.57078105e-01 2.36719832e-01
-2.54288256e-01 1.40466364e-02 6.13720238e-01 -6.06232584e-01
1.45749927e-01 -2.46648550e-01 -9.10058022e-01 3.76778483e-01
7.44928181e-01 -1.30304545e-01 -9.13592987e-03 -5.74882030e-01
-2.07855001e-01 -1.41610816e-01 -2.11941190e-02 -1.25434887e+00
1.08857468e-01 5.42169988e-01 7.66419828e-01 -4.57739949e-01
1.70715019e-01 -7.35290408e-01 -2.53430009e-01 4.14518535e-01
-7.72036433e-01 1.30303010e-01 5.89635253e-01 1.62367392e-02
-1.46687120e-01 -4.12919432e-01 6.24264657e-01 -1.18275248e-01
-9.64067042e-01 3.04841667e-01 -6.37609124e-01 4.05753702e-01
6.52520657e-01 -1.54020414e-01 -1.94721110e-02 -3.94095629e-01
-4.20235515e-01 -7.16137737e-02 3.17948729e-01 5.69979787e-01
1.25894412e-01 -1.26905203e+00 -1.16581357e+00 6.22312009e-01
3.00075591e-01 -5.08522034e-01 -3.73469710e-01 3.90225559e-01
-3.73002291e-01 1.07427585e+00 -8.92617926e-02 -4.08339649e-01
-1.19422245e+00 6.12144649e-01 4.04690653e-01 -5.88678539e-01
-4.91520345e-01 1.09614646e+00 2.26937793e-02 -7.28510261e-01
3.76987100e-01 -2.37364750e-02 -2.39517331e-01 8.97674188e-02
6.60647035e-01 2.73707002e-01 7.83439726e-02 -8.59062493e-01
-6.30951941e-01 2.42743507e-01 -4.10654306e-01 -4.85262960e-01
8.01372409e-01 1.25094444e-01 1.00037418e-01 6.39656663e-01
1.70197368e+00 1.01549178e-01 -3.99748385e-01 -2.95682520e-01
3.21928978e-01 -1.91400588e-01 3.73258859e-01 -1.32639933e+00
-7.17354238e-01 1.12026799e+00 4.69214350e-01 9.17724073e-02
4.95905578e-01 -2.19157502e-01 7.12982893e-01 3.90778631e-01
2.28975669e-01 -9.47178841e-01 -4.67367887e-01 8.50860119e-01
1.03213775e+00 -1.50354815e+00 -4.50661212e-01 2.08996728e-01
-5.81712782e-01 1.18223417e+00 4.42148864e-01 3.12727422e-01
4.30003524e-01 3.92547667e-01 8.42459053e-02 3.25132757e-01
-5.56683481e-01 5.27558252e-02 4.67774242e-01 6.17887497e-01
1.05060709e+00 2.99785316e-01 2.08840996e-01 6.26788616e-01
-3.88881475e-01 -6.05070926e-02 2.37048477e-01 1.94372535e-01
3.69960885e-03 -9.16381776e-01 -3.85053515e-01 6.54962897e-01
-6.77312493e-01 -4.97752190e-01 -7.24965811e-01 9.71939027e-01
-8.94525871e-02 9.69397783e-01 3.44123960e-01 -3.42702210e-01
1.39770778e-02 5.70411146e-01 1.82159826e-01 -5.22762001e-01
-9.63442087e-01 -4.63791132e-01 3.96178484e-01 -2.11487100e-01
1.56534970e-01 -7.34019101e-01 -6.17528260e-01 -4.35816646e-01
-1.73069850e-01 1.65714517e-01 5.11800885e-01 7.91698158e-01
1.79521576e-01 4.52410430e-01 3.31634969e-01 -6.09133601e-01
-6.77221358e-01 -1.36397314e+00 -2.98911899e-01 4.48989928e-01
2.61531591e-01 -4.32436973e-01 -4.93140191e-01 -3.89382541e-01]
|
[10.374920845031738, 10.54114818572998]
|
a7e7b315-df76-4cb6-ab2b-f13a31546e0e
|
bridging-the-gap-between-training-and-2
|
2005.09343
| null |
https://arxiv.org/abs/2005.09343v1
|
https://arxiv.org/pdf/2005.09343v1.pdf
|
Bridging the Gap Between Training and Inference for Spatio-Temporal Forecasting
|
Spatio-temporal sequence forecasting is one of the fundamental tasks in spatio-temporal data mining. It facilitates many real world applications such as precipitation nowcasting, citywide crowd flow prediction and air pollution forecasting. Recently, a few Seq2Seq based approaches have been proposed, but one of the drawbacks of Seq2Seq models is that, small errors can accumulate quickly along the generated sequence at the inference stage due to the different distributions of training and inference phase. That is because Seq2Seq models minimise single step errors only during training, however the entire sequence has to be generated during the inference phase which generates a discrepancy between training and inference. In this work, we propose a novel curriculum learning based strategy named Temporal Progressive Growing Sampling to effectively bridge the gap between training and inference for spatio-temporal sequence forecasting, by transforming the training process from a fully-supervised manner which utilises all available previous ground-truth values to a less-supervised manner which replaces some of the ground-truth context with generated predictions. To do that we sample the target sequence from midway outputs from intermediate models trained with bigger timescales through a carefully designed decaying strategy. Experimental results demonstrate that our proposed method better models long term dependencies and outperforms baseline approaches on two competitive datasets.
|
['Hong-Bin Liu', 'Ickjai Lee']
|
2020-05-19
| null | null | null | null |
['spatio-temporal-forecasting']
|
['time-series']
|
[ 5.34284592e-01 -1.54413328e-01 -2.81595200e-01 -5.17970502e-01
-5.36779046e-01 -4.14290577e-01 6.72232032e-01 1.11388721e-01
-5.63464820e-01 1.22032809e+00 3.63083124e-01 -4.87946242e-01
-1.49916625e-02 -9.18259025e-01 -7.49475241e-01 -8.40925038e-01
-1.81611568e-01 4.87400770e-01 5.55627942e-01 -3.44621927e-01
2.75830030e-01 7.95156062e-02 -1.64728987e+00 6.06301129e-01
1.17576528e+00 7.49191582e-01 4.98945564e-01 9.62860107e-01
-5.96861184e-01 1.09100735e+00 -5.04093885e-01 -1.63873378e-02
-2.32532844e-02 -5.94279170e-01 -6.61167860e-01 -2.41665646e-01
4.96422350e-02 -1.49804831e-01 5.73866107e-02 8.37913692e-01
3.86635184e-01 4.41919625e-01 4.28488851e-01 -1.10725820e+00
-1.43896297e-01 6.03062689e-01 -5.10138392e-01 6.54393137e-01
1.30802527e-01 3.01578015e-01 6.21189058e-01 -6.45839572e-01
3.70639354e-01 1.20112252e+00 8.79244387e-01 5.11529744e-01
-7.56601214e-01 -8.39351475e-01 7.18001127e-01 1.72674924e-01
-1.01469421e+00 -2.56848097e-01 4.00857478e-01 -5.95856130e-01
1.09529567e+00 3.10749561e-01 7.48101294e-01 1.12507010e+00
2.74424195e-01 6.98762774e-01 1.09838355e+00 -1.36847407e-01
3.39043200e-01 7.11576790e-02 -2.68224925e-01 3.28141242e-01
-4.75624055e-01 3.75100523e-01 -3.94296139e-01 -2.10936040e-01
5.13239861e-01 2.10744247e-01 -1.89727053e-01 3.48413140e-01
-1.16596639e+00 6.65863156e-01 2.43444517e-01 4.40651983e-01
-4.22668308e-01 1.31535351e-01 4.71349120e-01 2.05902562e-01
8.91315520e-01 2.21834332e-02 -6.88604712e-01 -3.46385628e-01
-1.29278779e+00 4.49729323e-01 4.54427242e-01 7.89715588e-01
6.21558070e-01 9.34065059e-02 -5.10303557e-01 6.38301611e-01
8.96959230e-02 4.73919749e-01 6.77286088e-01 -6.08177006e-01
9.48814809e-01 3.21659416e-01 3.91621768e-01 -8.57084632e-01
-2.65452176e-01 -5.01529038e-01 -1.12979794e+00 -1.47055909e-01
4.80066508e-01 -5.69148540e-01 -8.81441116e-01 1.62839985e+00
6.68861210e-01 1.05373061e+00 -6.09804038e-03 7.65174448e-01
5.13237894e-01 1.24161255e+00 4.21445608e-01 -2.81986952e-01
9.94956195e-01 -1.12106252e+00 -5.74724376e-01 -3.45133245e-01
8.41949642e-01 -4.02112246e-01 7.21972883e-01 1.77385867e-01
-8.79560292e-01 -7.39342630e-01 -6.34927869e-01 2.88355649e-01
-4.69726980e-01 -1.98690236e-01 1.51658937e-01 4.94020730e-01
-8.35661948e-01 7.55902350e-01 -8.05337667e-01 -1.08696677e-01
3.89030576e-01 -1.91374160e-02 7.57561624e-02 7.99546838e-02
-1.67059183e+00 7.53974378e-01 7.16535151e-01 4.28495705e-01
-6.83916390e-01 -1.07812023e+00 -7.21423388e-01 -5.92206270e-02
1.56359717e-01 -5.65506339e-01 1.31189716e+00 -1.05547440e+00
-1.48010910e+00 2.20335484e-01 -5.66921413e-01 -6.62429571e-01
7.51165092e-01 -2.01249450e-01 -4.91927952e-01 -3.76570374e-01
6.00926727e-02 6.18661523e-01 7.61668503e-01 -7.85871148e-01
-1.24054623e+00 -1.01723388e-01 -2.66339302e-01 2.44696468e-01
-9.37594194e-03 -9.48798060e-02 3.54521000e-03 -7.31412590e-01
-5.41788280e-01 -8.00312579e-01 -7.27043331e-01 -3.54991943e-01
-4.49993722e-02 -4.00740176e-01 5.64350367e-01 -7.58449614e-01
1.41786420e+00 -1.86190367e+00 -1.96141124e-01 2.21434072e-01
-2.32033759e-01 6.75644338e-01 -2.59499192e-01 6.46261692e-01
5.97384246e-03 -3.51509377e-02 -5.63325822e-01 -3.29663038e-01
-3.81104231e-01 3.34987879e-01 -7.75591075e-01 2.48820469e-01
3.32095265e-01 8.39849114e-01 -1.53640223e+00 -3.43731344e-01
2.40740851e-01 4.11209583e-01 -4.58463311e-01 2.92767346e-01
-7.20422924e-01 1.10958052e+00 -4.95388806e-01 -3.61926667e-02
7.52089858e-01 -2.36057490e-01 2.11696327e-01 6.93043411e-01
-5.45734942e-01 5.51888168e-01 -1.09652472e+00 1.66547275e+00
-6.57155812e-01 4.25207078e-01 -4.93100733e-01 -1.22223985e+00
9.58259165e-01 5.54797709e-01 3.13401908e-01 -8.89405668e-01
-2.84011424e-01 8.74022022e-02 -2.43967831e-01 -8.10670853e-01
4.91218388e-01 -3.05987149e-01 2.01878458e-01 3.47435981e-01
-5.61964571e-01 2.03514993e-01 5.88603541e-02 -1.76890999e-01
7.92348862e-01 4.44739550e-01 1.53883055e-01 -2.38538929e-03
8.56533587e-01 2.55716383e-01 8.77628267e-01 7.21099555e-01
-3.98618244e-02 3.69974971e-01 2.51993269e-01 -8.09486151e-01
-1.18604159e+00 -5.39010108e-01 2.17809960e-01 1.26831353e+00
-2.04120234e-01 -1.02163039e-01 -4.83972460e-01 -7.25849390e-01
-2.06063628e-01 8.34943473e-01 -6.85925364e-01 1.61995351e-01
-1.01761734e+00 -7.80324578e-01 4.92539972e-01 7.30286241e-01
5.50765157e-01 -1.35722852e+00 -7.28380561e-01 5.72103024e-01
-3.94652963e-01 -9.39632058e-01 -5.14534652e-01 -1.41744539e-01
-1.04219115e+00 -5.69487572e-01 -7.89298475e-01 -6.41934812e-01
4.36989725e-01 1.26270428e-01 1.13861787e+00 1.66899726e-01
-3.62233892e-02 -4.44466144e-01 -3.05240631e-01 -4.08710748e-01
-3.75961840e-01 3.27928156e-01 -4.06741917e-01 -2.39442550e-02
2.58426696e-01 -6.76834106e-01 -7.14844167e-01 2.19133765e-01
-8.18843603e-01 2.97838390e-01 4.61076826e-01 8.04257214e-01
4.24342811e-01 2.07253844e-01 1.05512226e+00 -1.12685227e+00
3.81281257e-01 -9.48922753e-01 -6.69375956e-01 2.29143232e-01
-3.43554884e-01 1.78534761e-01 1.00418425e+00 -4.63496089e-01
-1.39636624e+00 2.20937915e-02 -4.47434127e-01 -1.31685734e-01
-3.48793387e-01 7.12256312e-01 3.93219978e-01 6.61437035e-01
4.37492013e-01 7.20362842e-01 -2.19220683e-01 -3.46013010e-01
1.79489568e-01 6.56024039e-01 2.46674255e-01 -2.51698941e-01
6.46333337e-01 3.96745294e-01 -2.36179247e-01 -8.08463991e-01
-1.01180923e+00 -6.14560306e-01 -4.97952819e-01 -2.04365909e-01
8.73449385e-01 -9.76985872e-01 -6.13372087e-01 5.20531595e-01
-1.31131530e+00 -7.51357079e-01 -1.92682713e-01 3.46691668e-01
-1.82416692e-01 1.61741123e-01 -2.99886048e-01 -1.17107916e+00
-3.25925589e-01 -7.16536760e-01 1.14090097e+00 1.40014365e-01
-2.99101025e-01 -1.23672247e+00 5.81098378e-01 8.59685689e-02
5.57379246e-01 3.13353747e-01 7.09793746e-01 -3.83406907e-01
-5.43997824e-01 2.23637000e-01 -4.00963575e-02 8.15443620e-02
-1.90333817e-02 -2.14249104e-01 -9.04302835e-01 -2.04708472e-01
-1.48981765e-01 -1.39308572e-01 8.68332624e-01 2.43907779e-01
1.26589739e+00 -5.84224224e-01 -5.01700222e-01 4.27513868e-01
1.34030175e+00 3.51735801e-01 7.27672756e-01 2.82790899e-01
5.47955334e-01 8.21804225e-01 8.98588598e-01 5.30330718e-01
5.51461101e-01 3.95683348e-01 1.47262514e-01 5.05026020e-02
1.38217121e-01 -4.91396010e-01 3.99904490e-01 7.80734479e-01
8.40243921e-02 -5.04284024e-01 -1.00759923e+00 9.14023995e-01
-2.15607810e+00 -1.39033818e+00 -3.27775925e-01 2.16977692e+00
1.03520715e+00 2.02609785e-02 1.60663143e-01 2.06294313e-01
6.30460680e-01 4.31320935e-01 -4.27515626e-01 -3.10998768e-01
2.34457240e-01 2.58656114e-01 4.84431475e-01 7.93096483e-01
-9.76256847e-01 8.55014980e-01 5.65665054e+00 8.69526625e-01
-1.40259647e+00 8.48290324e-02 6.84520602e-01 -1.21705890e-01
-3.90042365e-01 -1.58582419e-01 -9.52316403e-01 1.04593170e+00
1.37653422e+00 2.51699746e-01 8.45444053e-02 3.62943500e-01
8.11229944e-01 -1.32580414e-01 -8.35213363e-01 5.41802108e-01
-4.65153158e-01 -1.49206007e+00 -1.25394151e-01 -3.45416844e-01
1.00636172e+00 1.00162946e-01 -4.54084352e-02 3.96612376e-01
4.22085971e-01 -1.18374169e+00 6.11651599e-01 7.20451295e-01
5.85846424e-01 -7.62070417e-01 7.62339652e-01 1.04950786e+00
-1.53487492e+00 -7.29756802e-02 -2.07848638e-01 -3.51489305e-01
3.19219112e-01 8.53235662e-01 -1.13077247e+00 6.70435309e-01
6.03524804e-01 8.11099470e-01 -8.75737593e-02 1.01746917e+00
-2.37098023e-01 8.92004251e-01 -1.27589643e-01 -2.79172242e-01
5.72061777e-01 -2.71738291e-01 3.97681922e-01 1.41150224e+00
4.63029921e-01 1.87583193e-01 3.28976691e-01 5.71037114e-01
3.55847955e-01 -2.67015137e-02 -5.28263330e-01 3.32644761e-01
5.85938990e-01 7.03967929e-01 -4.20542210e-01 -5.83773553e-01
-3.24126631e-01 7.73198009e-01 3.29347342e-01 4.26757753e-01
-1.10465991e+00 -1.94517195e-01 4.57424551e-01 2.25036249e-01
6.63891256e-01 -1.69674903e-01 -2.70894349e-01 -8.29250574e-01
3.39882188e-02 -6.68611705e-01 4.00544912e-01 -5.44180334e-01
-1.07583058e+00 5.56672990e-01 -6.85887635e-02 -1.28960788e+00
-5.39899886e-01 7.51020014e-02 -8.92112195e-01 1.22396803e+00
-2.08884263e+00 -8.91450584e-01 -1.26289874e-01 3.07245106e-01
8.39928389e-01 2.50407636e-01 5.52990258e-01 4.62260127e-01
-5.11489868e-01 2.54115522e-01 2.52488792e-01 -2.13020407e-02
4.37065005e-01 -1.04548120e+00 8.32562983e-01 8.68852198e-01
-3.16559583e-01 1.88916907e-01 7.77557790e-01 -8.95697236e-01
-6.65056765e-01 -1.77982461e+00 1.53552616e+00 -1.14955150e-01
4.97996777e-01 -2.23334402e-01 -1.19193971e+00 5.33227623e-01
1.52767688e-01 1.04212947e-03 4.04670417e-01 -1.91680938e-01
4.42281105e-02 -7.40416124e-02 -8.62957418e-01 3.48300993e-01
1.02850342e+00 -2.43634179e-01 -4.25702065e-01 2.87193358e-01
6.67306900e-01 -5.38374305e-01 -5.62019110e-01 3.15877020e-01
4.75515246e-01 -8.79790783e-01 7.10313082e-01 -7.45937526e-01
6.61456406e-01 -4.73329037e-01 2.93438166e-01 -1.46514845e+00
-2.26954103e-01 -5.97684264e-01 1.20351408e-02 1.19049394e+00
5.75361073e-01 -7.87863672e-01 9.54029799e-01 1.35125488e-01
-9.06142890e-02 -8.92516851e-01 -8.68388593e-01 -8.36898744e-01
1.83112890e-01 -4.64815825e-01 9.51211691e-01 8.90765727e-01
-3.07065904e-01 1.24083161e-01 -7.06002116e-01 3.58338922e-01
3.38525295e-01 3.16709459e-01 6.04727983e-01 -9.67166603e-01
-2.28915438e-01 -1.75700843e-01 1.15710773e-01 -1.55247068e+00
9.58227217e-02 -6.69910491e-01 4.20424551e-01 -1.37274897e+00
-8.57661292e-02 -8.53128493e-01 -4.65775914e-02 2.10813880e-01
-5.86413324e-01 -1.43300623e-01 -1.59224838e-01 5.54232411e-02
-5.24284542e-01 6.57266974e-01 1.35903323e+00 6.99692443e-02
-4.03539777e-01 3.39401484e-01 -2.02497363e-01 3.53449047e-01
8.78185570e-01 -7.32305288e-01 -7.14698374e-01 -4.76737678e-01
3.78596097e-01 3.50262552e-01 2.18757361e-01 -8.65059555e-01
4.19537008e-01 -5.83701253e-01 2.97630787e-01 -8.45460355e-01
-5.64516820e-02 -6.55668080e-01 3.31580639e-01 5.97807109e-01
-5.43508410e-01 1.14614308e-01 2.20079020e-01 9.32005823e-01
-2.19633937e-01 -5.58994748e-02 5.69383502e-01 -3.04521799e-01
-7.86024094e-01 4.17308986e-01 -6.17302716e-01 3.36407542e-01
1.02399266e+00 -1.96895003e-01 -1.54044226e-01 -2.75557756e-01
-5.60854852e-01 6.96120918e-01 -1.47736385e-01 3.88072282e-01
3.90376538e-01 -1.15992212e+00 -9.24328208e-01 2.12751374e-01
-2.40181401e-01 4.85707402e-01 4.98781383e-01 8.81396711e-01
-2.81069219e-01 6.35277510e-01 9.97924805e-02 -7.82846451e-01
-9.84646440e-01 3.82211715e-01 3.22253853e-01 -7.88043320e-01
-5.51175416e-01 8.98323298e-01 9.14528295e-02 -6.63766503e-01
2.70274818e-01 -5.59723079e-01 -4.19426024e-01 -4.28235941e-02
8.50028813e-01 4.86515492e-01 1.10532470e-01 -3.74299437e-01
-1.84091672e-01 3.55650932e-01 1.28039062e-01 -9.50596556e-02
1.44107199e+00 -2.40012765e-01 1.07375540e-01 7.04201043e-01
9.52069819e-01 -2.66767919e-01 -1.54928160e+00 -3.79338801e-01
1.37194425e-01 -5.00366688e-01 -1.18346192e-01 -7.63392329e-01
-8.47774208e-01 9.75390971e-01 4.12324667e-01 1.72366023e-01
9.73684430e-01 -4.54164803e-01 1.31366134e+00 1.36518041e-02
8.51278678e-02 -9.74243462e-01 -1.83024839e-01 8.81331205e-01
5.69387138e-01 -1.19237351e+00 -3.98893565e-01 -2.33279973e-01
-7.03796983e-01 8.56663704e-01 5.04797041e-01 -7.46456012e-02
5.76662183e-01 2.87174016e-01 -8.56468081e-03 1.28938183e-01
-1.15753913e+00 -8.20912942e-02 1.08671620e-01 5.40788710e-01
5.64256251e-01 -2.81860624e-02 -4.14541289e-02 2.32339084e-01
-1.50730133e-01 3.94816786e-01 -4.52266866e-03 9.60670948e-01
-5.55377305e-01 -1.20169556e+00 -3.36516887e-01 3.88928831e-01
-3.35647881e-01 -3.08241397e-01 1.95728093e-01 4.11107957e-01
4.26293939e-01 8.11479270e-01 3.78328115e-01 -8.08541104e-02
1.29448846e-01 3.01239252e-01 9.33169797e-02 -5.41054189e-01
-8.19856346e-01 -2.15470657e-01 1.46748975e-01 -3.49155873e-01
-5.82368493e-01 -8.74759912e-01 -1.31524837e+00 -3.36353272e-01
-4.05482501e-02 3.00764889e-01 3.55435163e-01 1.39836705e+00
4.36259419e-01 6.54730856e-01 7.20018804e-01 -7.73492813e-01
-4.15820122e-01 -9.82517958e-01 -5.69444709e-02 2.53694087e-01
7.33099103e-01 -2.21886963e-01 -1.14861421e-01 2.13267490e-01]
|
[6.853796482086182, 2.490302324295044]
|
85ca518a-195c-4fd5-9c07-b528a4919af1
|
towards-less-generic-responses-in-neural
| null | null |
https://aclanthology.org/D18-1297
|
https://aclanthology.org/D18-1297.pdf
|
Towards Less Generic Responses in Neural Conversation Models: A Statistical Re-weighting Method
|
Sequence-to-sequence neural generation models have achieved promising performance on short text conversation tasks. However, they tend to generate generic/dull responses, leading to unsatisfying dialogue experience. We observe that in the conversation tasks, each query could have multiple responses, which forms a 1-to-n or m-to-n relationship in the view of the total corpus. The objective function used in standard sequence-to-sequence models will be dominated by loss terms with generic patterns. Inspired by this observation, we introduce a statistical re-weighting method that assigns different weights for the multiple responses of the same query, and trains the common neural generation model with the weights. Experimental results on a large Chinese dialogue corpus show that our method improves the acceptance rate of generated responses compared with several baseline models and significantly reduces the number of generated generic responses.
|
['Yahui Liu', 'Xiaojiang Liu', 'Jian Yao', 'Jun Gao', 'Shuming Shi', 'Wei Bi']
|
2018-10-01
| null | null | null |
emnlp-2018-10
|
['short-text-conversation']
|
['natural-language-processing']
|
[ 3.47302914e-01 2.44362324e-01 2.44155200e-03 -7.11813331e-01
-9.89782870e-01 -4.27641213e-01 7.44023561e-01 -3.58200938e-01
-3.64889681e-01 1.10750663e+00 7.62661219e-01 -1.69874191e-01
3.48045021e-01 -7.67206728e-01 -4.85920906e-02 -5.60176969e-01
5.41504622e-01 7.26237953e-01 8.84836689e-02 -8.87766242e-01
3.77466619e-01 -3.18233997e-01 -1.03186309e+00 7.66017616e-01
9.91134465e-01 6.19228780e-01 4.55221385e-01 9.09021735e-01
-4.78527844e-01 1.00628972e+00 -1.26797438e+00 -9.12017524e-01
-8.15581903e-02 -1.05194438e+00 -1.19360256e+00 -1.47967130e-01
9.89459455e-02 -4.55064565e-01 -3.63120705e-01 8.50696683e-01
7.91075945e-01 5.36514103e-01 6.00585818e-01 -1.11203015e+00
-7.69166946e-01 9.78892624e-01 -2.65088469e-01 -1.37309149e-01
7.96216965e-01 6.30550608e-02 1.09664321e+00 -8.32607329e-01
5.19369781e-01 1.57409167e+00 3.43052238e-01 1.18839669e+00
-1.07515848e+00 -4.35246378e-01 1.16086334e-01 -5.83840068e-03
-1.00121105e+00 -2.89513052e-01 4.51694846e-01 -2.57248320e-02
1.15298676e+00 4.43865925e-01 2.53211409e-01 1.46793997e+00
1.28302917e-01 1.03022313e+00 5.99616349e-01 -4.07480925e-01
-1.41635641e-01 1.81866959e-01 6.04808033e-02 1.55559376e-01
-4.57791418e-01 -1.97216526e-01 -6.78028524e-01 -3.69649202e-01
4.34487283e-01 -2.57955730e-01 -3.80524576e-01 4.37103182e-01
-9.41370010e-01 1.26022625e+00 9.36968997e-02 2.97030449e-01
-3.98794025e-01 -1.18939556e-01 4.15773809e-01 7.15113044e-01
4.91896987e-01 6.56301916e-01 -4.34903949e-01 -5.35210371e-01
-5.62878251e-01 8.20950150e-01 1.12683237e+00 1.04359245e+00
5.73990047e-01 3.71848568e-02 -6.39967442e-01 1.40108407e+00
-1.28556684e-01 1.78170294e-01 9.35876310e-01 -8.09110403e-01
7.07897246e-01 5.55617273e-01 4.50717509e-01 -8.98397982e-01
-2.57437915e-01 -1.99623629e-01 -1.04017258e+00 -3.98796350e-01
4.36558664e-01 -5.98369241e-01 -4.06707317e-01 2.07040501e+00
5.36294058e-02 -3.84691358e-01 3.95936638e-01 8.70321989e-01
7.70108223e-01 8.79214108e-01 -1.07109375e-01 -2.49057859e-01
9.05090988e-01 -1.21723545e+00 -9.45443213e-01 -3.17780793e-01
8.03497314e-01 -9.77706909e-01 1.42814815e+00 2.00585485e-01
-1.28166199e+00 -7.22500682e-01 -5.97785056e-01 9.15512890e-02
-3.51543874e-02 1.37206102e-02 3.91196668e-01 5.00258327e-01
-1.08081722e+00 4.25912023e-01 -3.52022499e-02 -2.36167237e-01
-3.92058522e-01 1.67867839e-01 4.71963659e-02 4.80048582e-02
-1.59858561e+00 1.12401676e+00 4.12819266e-01 5.11020645e-02
-5.03647625e-01 -3.23363692e-01 -6.79873586e-01 8.23858678e-02
3.43897939e-01 -8.43144774e-01 1.89772880e+00 -9.07360673e-01
-1.78333950e+00 5.15960932e-01 -4.85368520e-01 -3.21932346e-01
5.49883664e-01 -3.15679491e-01 -3.49667609e-01 -1.55678466e-01
-9.31863207e-04 7.27991462e-01 5.71815670e-01 -1.06092858e+00
-7.99838066e-01 1.54436320e-01 2.87311614e-01 5.18101037e-01
-1.44120723e-01 2.21023470e-01 -1.54768899e-01 -7.14198768e-01
-2.15921283e-01 -9.03984070e-01 -3.73170078e-01 -8.59352589e-01
-3.63503903e-01 -7.55028248e-01 3.06044579e-01 -4.76471931e-01
1.50670254e+00 -1.69116044e+00 2.51574159e-01 -1.81336567e-01
-4.52759042e-02 1.81932092e-01 -5.20343304e-01 9.08908367e-01
1.97313383e-01 6.14758097e-02 -6.47847056e-02 -2.85300791e-01
1.40498772e-01 4.17095311e-02 -5.45804799e-01 -4.04840440e-01
1.42115921e-01 9.73425925e-01 -1.11905611e+00 -2.17367247e-01
-2.32743144e-01 -4.32095937e-02 -5.86413085e-01 8.74767542e-01
-3.35268289e-01 1.80561453e-01 -3.86915863e-01 9.60009992e-02
4.37447757e-01 -3.12381834e-01 2.99133360e-01 3.31173211e-01
2.57715464e-01 7.24331498e-01 -5.05481064e-01 1.62150085e+00
-6.50589347e-01 2.77254671e-01 -8.72239023e-02 -5.64005315e-01
1.03637660e+00 5.21737218e-01 -2.04640487e-03 -6.34294689e-01
-5.83179705e-02 2.23486692e-01 2.71017432e-01 -6.02934778e-01
1.05630744e+00 -3.23207051e-01 -6.00176275e-01 8.89255285e-01
6.20749183e-02 -3.93992215e-01 3.47626984e-01 5.28089762e-01
8.08157682e-01 -1.44107908e-01 9.83922035e-02 1.31620526e-01
7.53110826e-01 -1.79175392e-01 4.81535107e-01 1.14771986e+00
8.52122903e-02 6.22852266e-01 6.16596162e-01 -2.90388614e-01
-9.25740659e-01 -6.91472292e-01 5.43638766e-01 1.52356458e+00
4.64354977e-02 -2.90567219e-01 -8.89920413e-01 -6.22226417e-01
-4.02197093e-01 1.07395041e+00 -3.37729514e-01 -3.91036093e-01
-7.54181802e-01 -8.30250442e-01 7.70987809e-01 3.84663075e-01
2.67021120e-01 -1.51394904e+00 -1.72793627e-01 6.67481303e-01
-9.67459798e-01 -8.53437901e-01 -1.06061637e+00 -3.14613611e-01
-4.39074427e-01 -7.02486813e-01 -9.62555289e-01 -9.25450921e-01
4.32204485e-01 2.20260233e-01 1.35912085e+00 1.93029195e-01
1.80975527e-01 1.21656587e-04 -6.24807715e-01 -2.56804436e-01
-8.94387126e-01 3.65418255e-01 4.59959358e-02 1.21800825e-02
5.27259111e-01 -1.58833310e-01 -5.04397750e-01 4.84598219e-01
-7.88874745e-01 3.06817554e-02 2.72962332e-01 1.39946342e+00
-2.34418944e-01 -4.54627991e-01 1.22889733e+00 -8.96734774e-01
1.77046072e+00 -3.75341922e-01 4.21263725e-02 5.19747376e-01
-6.00257039e-01 3.27085033e-02 8.49930167e-01 -7.32180536e-01
-1.49057209e+00 -4.79358733e-01 -3.26914251e-01 1.45041212e-01
9.48038511e-03 5.67568064e-01 -1.25449458e-02 3.49332958e-01
6.94058001e-01 4.73445177e-01 7.74213672e-02 -2.70863920e-01
4.06131983e-01 9.39279199e-01 3.96068156e-01 -6.03079021e-01
3.43966126e-01 -3.47135067e-01 -7.40444183e-01 -5.73800385e-01
-8.34743202e-01 -4.16875869e-01 -8.75716805e-02 -2.55680233e-01
4.72915322e-01 -6.84307396e-01 -6.28112137e-01 7.48002589e-01
-1.66714656e+00 -5.23594320e-01 2.77770553e-02 3.27387273e-01
-5.28896987e-01 5.07703424e-01 -1.13062656e+00 -1.05262959e+00
-6.17281556e-01 -1.03084803e+00 7.19022930e-01 5.29234290e-01
-9.08663571e-01 -8.76568258e-01 3.58110592e-02 3.74380380e-01
7.69367874e-01 -3.94352645e-01 1.09847212e+00 -8.76315117e-01
-1.49294958e-01 -1.46966800e-01 5.63032813e-02 4.13761288e-01
8.93419683e-02 -3.34269077e-01 -6.24112904e-01 -1.37088850e-01
3.00923526e-01 -8.31347466e-01 6.35606408e-01 -6.20363047e-03
9.73238051e-01 -7.84793794e-01 7.17245191e-02 -3.63299027e-02
9.14866745e-01 5.54686666e-01 7.40019917e-01 -1.41939893e-01
3.24750125e-01 1.04491079e+00 7.02998161e-01 5.81455827e-01
4.58708912e-01 6.49326861e-01 1.79760419e-02 5.42199314e-02
3.08522910e-01 -4.92118329e-01 4.38362747e-01 1.11660063e+00
2.31029421e-01 -8.17810237e-01 -3.62059504e-01 6.43780768e-01
-2.09954643e+00 -1.12675786e+00 -1.46154702e-01 2.14560294e+00
1.25699341e+00 -2.11385433e-02 2.09005281e-01 -3.84781241e-01
8.59879971e-01 2.43107766e-01 -2.21467584e-01 -8.91601741e-01
-2.48177603e-01 2.93821752e-01 -1.58603564e-01 8.26571226e-01
-4.31974798e-01 1.10673797e+00 7.12933111e+00 1.06790364e+00
-9.85206902e-01 -7.08025927e-03 6.07267559e-01 -1.92012712e-01
-5.85433483e-01 -1.64495677e-01 -8.25331211e-01 6.07517302e-01
9.02596414e-01 -6.56700611e-01 5.34990013e-01 5.58022618e-01
2.49317050e-01 1.84872355e-02 -1.00423300e+00 6.54893517e-01
1.97080404e-01 -1.07357967e+00 4.83654827e-01 -3.43911648e-01
7.94554055e-01 -3.56964439e-01 -8.21700469e-02 6.70257270e-01
7.31878221e-01 -1.04393852e+00 3.16770047e-01 4.82087046e-01
5.59534669e-01 -8.42052341e-01 1.01178181e+00 7.18798578e-01
-6.59630358e-01 7.95878544e-02 -5.90333283e-01 -1.79789200e-01
6.25633121e-01 1.48383647e-01 -1.21749425e+00 4.88284469e-01
1.72692910e-01 -1.00237466e-01 -7.48209096e-03 5.59146762e-01
-3.24244857e-01 4.38375056e-01 1.32962346e-01 -6.82471275e-01
4.49579567e-01 -3.18487644e-01 3.56856763e-01 1.19228208e+00
4.49719787e-01 1.75575912e-01 1.36763155e-01 8.00413787e-01
-2.92049885e-01 2.63227820e-01 -4.19068366e-01 -3.49468105e-02
5.12155592e-01 1.07369375e+00 6.94130315e-03 -5.65795481e-01
-2.63883531e-01 1.19114506e+00 3.37031454e-01 4.49166238e-01
-5.86312234e-01 -6.93404794e-01 4.68914181e-01 -2.28260443e-01
-8.45344514e-02 2.01591223e-01 -3.42567042e-02 -9.48803723e-01
6.32982031e-02 -1.25740325e+00 3.02008808e-01 -8.18272412e-01
-1.60430968e+00 9.48856771e-01 -1.32180586e-01 -9.31474030e-01
-1.05823696e+00 -1.16482861e-01 -9.29591537e-01 1.34488189e+00
-1.02809906e+00 -7.87266970e-01 -3.04676406e-03 4.14114594e-01
1.11706400e+00 -4.06102091e-01 1.06575823e+00 6.77519888e-02
-3.60273659e-01 9.97805357e-01 -1.43039495e-01 8.42898339e-02
1.00569808e+00 -1.16243517e+00 7.16178238e-01 4.03897852e-01
-3.53062004e-01 9.14288580e-01 6.69662774e-01 -5.03803194e-01
-6.78498864e-01 -8.19853663e-01 1.61699665e+00 -3.52857143e-01
4.92793143e-01 -2.49394700e-01 -9.73163188e-01 3.46969396e-01
6.96546197e-01 -1.03971124e+00 9.78022873e-01 1.17947116e-01
-3.25095542e-02 3.11714500e-01 -9.32057381e-01 9.40158069e-01
8.31899285e-01 -6.09531105e-01 -8.46323192e-01 4.03780818e-01
9.83080983e-01 -4.71375316e-01 -5.82050502e-01 2.86477804e-01
6.09513402e-01 -9.09324765e-01 6.79305494e-01 -8.93624544e-01
7.99830019e-01 1.62053257e-01 1.13640890e-01 -1.73461938e+00
-2.89135754e-01 -1.13495481e+00 4.80389297e-02 1.09883225e+00
7.79223442e-01 -4.92004484e-01 7.50642240e-01 8.29585552e-01
-1.59699276e-01 -8.66185248e-01 -6.87406361e-01 -5.15661955e-01
2.78871685e-01 1.66227639e-01 8.22608769e-01 7.29596913e-01
4.53940898e-01 9.45051014e-01 -9.64158118e-01 -5.87083280e-01
2.86498293e-02 1.13018572e-01 9.56848681e-01 -8.67699862e-01
-5.16041815e-01 -5.74802577e-01 3.62696469e-01 -1.75547850e+00
2.75509924e-01 -5.63268006e-01 4.03221309e-01 -1.37682760e+00
1.15497783e-01 -2.55231798e-01 1.19855963e-01 4.90602143e-02
-8.09397161e-01 -1.16371349e-01 1.64095759e-01 -4.79625352e-02
-5.55353284e-01 8.39026392e-01 1.33855379e+00 -1.49489239e-01
-3.12789738e-01 3.56004864e-01 -8.94315302e-01 3.19171488e-01
8.88670444e-01 -3.71364355e-01 -5.46534896e-01 -4.74443585e-01
2.84432590e-01 5.91444373e-01 -2.93919683e-01 -2.99286574e-01
1.88606098e-01 -3.49695414e-01 -1.69407189e-01 -6.02023840e-01
5.00642478e-01 -9.67455059e-02 -1.44252509e-01 3.47058654e-01
-1.09013689e+00 3.24301183e-01 -2.32128486e-01 3.40334266e-01
-4.79594290e-01 -6.22030258e-01 3.26631606e-01 -5.48864305e-01
-2.64458656e-01 3.02612074e-02 -6.84695721e-01 2.92819679e-01
4.76511121e-01 -8.93894359e-02 -2.90623248e-01 -1.22351968e+00
-3.49669099e-01 6.06049836e-01 5.39112762e-02 8.19567502e-01
6.96259916e-01 -1.33737969e+00 -1.11437070e+00 -9.13238749e-02
1.60112560e-01 -2.19052820e-03 4.78253990e-01 2.59816229e-01
-1.86131537e-01 4.87428546e-01 1.68574657e-02 -1.86491027e-01
-1.31940234e+00 1.03611343e-01 3.98710191e-01 -7.29007840e-01
-8.23006481e-02 1.27972353e+00 3.91877204e-01 -8.95736337e-01
3.30623478e-01 1.24880567e-01 -2.99286366e-01 4.96149063e-02
6.59162879e-01 2.24854439e-01 -1.53730169e-01 -4.20803010e-01
1.14174560e-01 -3.55017781e-02 -5.60687602e-01 -4.89818037e-01
8.58228803e-01 -1.80805847e-01 -1.17914401e-01 3.72505903e-01
1.01762283e+00 -1.14287250e-01 -8.32463264e-01 -4.71282870e-01
-9.59112942e-02 -4.55882788e-01 -8.81337583e-01 -8.69660854e-01
-6.24561906e-01 8.17926466e-01 -1.86880946e-01 4.57130849e-01
8.55448365e-01 -3.59821945e-01 1.29999959e+00 6.27458751e-01
3.90035480e-01 -1.38015926e+00 4.15692776e-01 1.14020586e+00
1.17702639e+00 -1.03319466e+00 -5.54227471e-01 -1.46017149e-01
-1.18487382e+00 9.85966384e-01 1.10723317e+00 2.13051066e-01
-1.53435692e-01 8.02351460e-02 4.25716847e-01 1.95262909e-01
-1.29266751e+00 1.54677182e-01 -1.24147885e-01 5.03961742e-01
7.76597679e-01 1.60441190e-01 -7.79652119e-01 6.53362036e-01
-6.26437843e-01 -1.73634335e-01 7.84186602e-01 5.41912735e-01
-3.65311205e-01 -1.71171010e+00 -3.44344899e-02 4.43842620e-01
-5.54234982e-01 -3.18127036e-01 -9.04932976e-01 2.60110348e-01
-5.02045333e-01 1.45591605e+00 -1.10208400e-01 -7.69255877e-01
3.45434427e-01 4.39521790e-01 1.67939290e-01 -6.90263450e-01
-1.13013303e+00 2.84535997e-02 5.62296987e-01 -1.13366045e-01
-1.66946143e-01 -2.74799496e-01 -1.00523019e+00 -4.56270009e-01
-5.90017438e-01 5.75860262e-01 1.01438031e-01 8.49489927e-01
2.43414059e-01 3.70179981e-01 9.75040555e-01 -4.06434089e-01
-1.29033089e+00 -1.60764873e+00 -2.65538484e-01 7.21357167e-01
1.84132993e-01 -1.14588909e-01 -3.31836194e-01 -2.76323348e-01]
|
[12.587491035461426, 8.29146957397461]
|
b2ea3727-5f02-4728-96b3-71fa0e46cbf7
|
denseattentionseg-segment-hands-from
|
1903.12368
| null |
http://arxiv.org/abs/1903.12368v2
|
http://arxiv.org/pdf/1903.12368v2.pdf
|
DenseAttentionSeg: Segment Hands from Interacted Objects Using Depth Input
|
We propose a real-time DNN-based technique to segment hand and object of
interacting motions from depth inputs. Our model is called DenseAttentionSeg,
which contains a dense attention mechanism to fuse information in different
scales and improves the results quality with skip-connections. Besides, we
introduce a contour loss in model training, which helps to generate accurate
hand and object boundaries. Finally, we propose and release our InterSegHands
dataset, a fine-scale hand segmentation dataset containing about 52k depth maps
of hand-object interactions. Our experiments evaluate the effectiveness of our
techniques and datasets, and indicate that our method outperforms the current
state-of-the-art deep segmentation methods on interaction segmentation.
|
['Hao Zhang', 'Zihao Bo', 'Junhai Yong', 'Feng Xu']
|
2019-03-29
| null | null | null | null |
['hand-segmentation']
|
['computer-vision']
|
[-9.28481966e-02 -3.13286595e-02 -2.10970327e-01 -3.56653929e-01
-6.11430466e-01 -4.27480817e-01 9.16893110e-02 -7.86804557e-01
-4.39900041e-01 5.15116692e-01 4.53027874e-01 7.60730579e-02
1.61920682e-01 -5.27378440e-01 -7.16598392e-01 -6.11014426e-01
2.60967463e-01 7.55258024e-01 8.72483373e-01 -1.81464791e-01
7.73790479e-02 7.30177581e-01 -1.02747703e+00 5.72959900e-01
8.51286113e-01 9.17620182e-01 4.87472385e-01 9.87077296e-01
-7.37936944e-02 7.67106831e-01 -7.52792597e-01 -3.99820775e-01
3.13663125e-01 -2.57881165e-01 -1.39958143e+00 -9.34364274e-02
6.03839576e-01 -1.21568406e+00 -7.68216193e-01 8.01556349e-01
1.01756346e+00 2.33180866e-01 6.39590502e-01 -1.04171133e+00
-6.98359907e-01 7.21058190e-01 -8.25608432e-01 2.38500834e-01
3.55651200e-01 5.05457103e-01 6.94547296e-01 -6.78647101e-01
9.03820932e-01 1.70008719e+00 5.30463099e-01 7.96241224e-01
-1.02865660e+00 -6.71213984e-01 6.50639415e-01 1.89624950e-01
-1.19750059e+00 -3.05569079e-02 6.52720034e-01 -5.61335921e-01
1.24820781e+00 -7.56307095e-02 9.00380313e-01 1.45580006e+00
-2.22210819e-03 1.72693598e+00 4.90084052e-01 -5.95122315e-02
-1.20720878e-01 -7.18532205e-01 2.86671400e-01 6.17217898e-01
-2.04995468e-01 9.10440832e-02 -3.67760241e-01 2.04547375e-01
1.67281175e+00 -1.10931963e-01 -4.82166886e-01 -1.44671008e-01
-1.23696983e+00 4.62860852e-01 8.77370656e-01 2.60050029e-01
-4.75253135e-01 4.26733106e-01 3.38110626e-01 -4.73759174e-01
2.31722653e-01 1.47196114e-01 -7.16425419e-01 -2.41241753e-01
-8.20841372e-01 6.04670942e-01 4.52695549e-01 1.18889761e+00
9.50953811e-02 -2.49265477e-01 -7.35261381e-01 5.63309491e-01
2.21852690e-01 3.01104635e-01 1.38540447e-01 -1.43605578e+00
7.45318294e-01 4.73738700e-01 1.79874659e-01 -6.38569474e-01
-5.28600454e-01 8.12845901e-02 -7.39354432e-01 2.91886657e-01
7.33100355e-01 -2.32755587e-01 -1.38344479e+00 1.46450448e+00
2.66452491e-01 6.38376027e-02 -3.13344419e-01 1.38970220e+00
1.16284609e+00 4.35519993e-01 1.75873101e-01 3.57315481e-01
1.10243356e+00 -1.67538834e+00 -1.00997937e+00 -7.76295438e-02
1.78151235e-01 -4.29065555e-01 1.32117760e+00 6.09310627e-01
-1.30525661e+00 -9.24356997e-01 -4.99436289e-01 -5.07690430e-01
-6.69587180e-02 1.93288215e-02 9.40608263e-01 1.87077209e-01
-1.10317051e+00 7.95356095e-01 -1.15793025e+00 -4.89064865e-02
8.59454751e-01 4.47506100e-01 -1.20449036e-01 1.99912056e-01
-9.27270830e-01 4.86049056e-01 6.36144161e-01 5.39298236e-01
-9.17720079e-01 -6.58995032e-01 -7.11973608e-01 -1.14096478e-01
3.56197894e-01 -7.30210543e-01 1.39370191e+00 -5.35881877e-01
-1.68103087e+00 6.53064787e-01 -1.88206196e-01 -9.26038250e-02
8.08744967e-01 -7.39061356e-01 2.79634625e-01 3.53341222e-01
-7.62252212e-02 1.42003751e+00 4.62818742e-01 -1.44109333e+00
-6.11033082e-01 -4.80559975e-01 3.93050462e-01 1.21371239e-01
1.35312229e-01 5.18776178e-02 -1.02336752e+00 -1.05555105e+00
-2.25686305e-03 -6.76149189e-01 -3.24153960e-01 -6.11630343e-02
-7.35540509e-01 -4.36010867e-01 9.04311359e-01 -1.14300668e+00
1.13189280e+00 -1.85004735e+00 7.37115860e-01 5.00623928e-03
4.60566252e-01 3.63573641e-01 -3.89684826e-01 -1.79199725e-02
2.60213494e-01 8.65800232e-02 -2.58430272e-01 -5.09434879e-01
-5.71884774e-02 1.84858248e-01 -2.70427108e-01 -1.36989325e-01
4.16251533e-02 1.59920073e+00 -7.10881829e-01 -7.52715528e-01
3.20922107e-01 6.72575533e-01 -6.27697825e-01 4.87770617e-01
-6.63119674e-01 8.22083890e-01 -5.19709587e-01 8.12227607e-01
5.96796989e-01 -2.68598258e-01 -1.58722952e-01 -3.95914704e-01
1.63460553e-01 3.79718281e-02 -9.73285198e-01 2.42998505e+00
-8.39397684e-02 5.11386395e-01 -1.40185654e-02 -2.92431355e-01
3.23932797e-01 2.86145329e-01 2.10878789e-01 -3.77482235e-01
5.82085431e-01 -1.85405955e-01 -9.53963846e-02 -5.35243154e-01
3.70233685e-01 5.91507733e-01 3.24288279e-01 3.96305263e-01
9.17245895e-02 -2.74683088e-01 -3.90429311e-02 7.95801803e-02
7.58937657e-01 5.36999762e-01 -4.18746859e-01 -5.75481653e-02
1.61544174e-01 -2.48740941e-01 3.96819830e-01 6.37646496e-01
-4.92150158e-01 1.05663133e+00 5.21034837e-01 -4.49519187e-01
-8.26479971e-01 -1.07102835e+00 1.24527946e-01 1.22395945e+00
4.49054807e-01 5.76395057e-02 -1.34283996e+00 -1.00382519e+00
-1.72524646e-01 1.98111609e-01 -7.55753756e-01 3.46992910e-01
-9.35857117e-01 -3.29488903e-01 5.42880774e-01 1.54752600e+00
9.59126890e-01 -1.79515421e+00 -5.13311148e-01 1.58706397e-01
-4.63581771e-01 -1.09709108e+00 -1.00257587e+00 -1.03226565e-01
-1.00560701e+00 -1.23355865e+00 -1.36858380e+00 -1.03132117e+00
4.45903450e-01 -9.95381176e-02 1.12298882e+00 -5.57014570e-02
-4.82958198e-01 1.44617110e-01 -2.83536226e-01 -4.12114225e-02
1.60400793e-01 5.04196882e-01 -2.84805954e-01 -5.41886330e-01
2.06995383e-01 -5.02492368e-01 -9.54105556e-01 3.39509070e-01
-5.99433362e-01 1.28018767e-01 4.35333937e-01 6.62751436e-01
6.51687384e-01 -4.43903387e-01 2.84357369e-01 -5.21713018e-01
6.63297713e-01 4.07396942e-01 -4.30001646e-01 3.32339823e-01
6.97247162e-02 1.07066922e-01 4.03977260e-02 -5.72672784e-01
-1.28157341e+00 3.93947005e-01 -4.07519341e-01 -6.15212262e-01
-3.49450231e-01 -2.19386756e-01 -5.28997540e-01 -1.33185923e-01
4.18553412e-01 -1.07459508e-01 -3.96112263e-01 -7.13797450e-01
6.72899306e-01 6.17993414e-01 9.73801076e-01 -6.93988085e-01
1.82283372e-01 5.19340396e-01 -5.19510388e-01 -3.61484468e-01
-8.55084300e-01 -2.27480367e-01 -1.29313493e+00 -1.33791476e-01
1.45004332e+00 -6.21046185e-01 -1.04704976e+00 1.24219894e+00
-1.87342834e+00 -1.18155551e+00 -1.73586935e-01 3.36391389e-01
-7.03693628e-01 3.26280534e-01 -1.23227584e+00 -6.19500041e-01
-4.68642771e-01 -1.36273527e+00 1.47650814e+00 4.51812476e-01
-3.63311619e-01 -8.28014374e-01 -1.08691417e-02 4.95646358e-01
-1.19509995e-01 2.48527989e-01 5.16964316e-01 -1.86693028e-01
-8.60659719e-01 1.99034274e-01 -6.55817509e-01 2.92552620e-01
1.38249576e-01 2.37337407e-02 -1.10800040e+00 -2.22125471e-01
-6.45376801e-01 -4.46859509e-01 1.06927216e+00 1.01847327e+00
1.81181180e+00 3.73955108e-02 -5.94874918e-01 8.11932445e-01
7.09642589e-01 4.69038516e-01 8.87784421e-01 -9.82592162e-03
1.31626701e+00 5.69768727e-01 3.91404688e-01 3.20374131e-01
4.14423019e-01 6.78361237e-01 2.85228759e-01 -4.48492527e-01
-5.34111440e-01 -1.92072496e-01 -2.71125376e-01 4.50572431e-01
-6.42856419e-01 -4.12756056e-01 -7.56370962e-01 6.07296467e-01
-1.96037936e+00 -7.95163035e-01 5.57844639e-02 1.43545687e+00
1.03183651e+00 2.44536743e-01 5.81818163e-01 -6.36137202e-02
7.05836952e-01 1.27403721e-01 -9.58650291e-01 4.90586050e-02
-7.37533718e-02 5.26334643e-01 2.62184024e-01 8.32631707e-01
-1.21237826e+00 1.62854087e+00 7.40650940e+00 8.64080846e-01
-5.92656493e-01 -2.97966413e-02 4.48511541e-01 -8.90885815e-02
-2.94204336e-02 -6.16778076e-01 -8.43849480e-01 2.63106167e-01
2.91030314e-02 5.36438942e-01 5.34863949e-01 8.89504313e-01
2.51235515e-01 -9.67188925e-02 -1.19950831e+00 1.05892754e+00
-2.57743448e-02 -1.12861013e+00 1.32641181e-01 -6.95777535e-02
9.28489804e-01 -1.67453110e-01 -1.03498608e-01 -3.64177078e-02
5.72655797e-01 -1.24115705e+00 7.23797441e-01 5.88650167e-01
5.62264860e-01 -8.33954453e-01 8.04215133e-01 2.48445272e-02
-1.55419230e+00 1.26731828e-01 1.15292720e-01 -4.21686321e-02
5.69000781e-01 1.25024453e-01 -2.80180633e-01 1.88275382e-01
9.61799741e-01 6.59934461e-01 -3.85777324e-01 8.73643279e-01
-5.15195966e-01 2.11681291e-01 -1.55702353e-01 2.59221613e-01
2.83761144e-01 1.15967952e-01 1.23903841e-01 1.26073515e+00
-2.51332939e-01 5.54904222e-01 3.62605721e-01 1.21436751e+00
2.83374079e-03 -4.55563515e-01 -1.30551457e-01 9.92133375e-03
2.88056105e-01 8.20463121e-01 -9.79792297e-01 -7.23318398e-01
2.23027300e-02 1.69017112e+00 1.81621522e-01 6.99644089e-01
-8.54705691e-01 -7.59724557e-01 8.73065233e-01 -7.41698593e-02
4.98607576e-01 -3.81812125e-01 -5.57903230e-01 -8.53082776e-01
1.21651836e-01 -6.81439102e-01 1.33529663e-01 -1.01619339e+00
-1.26824045e+00 8.20837855e-01 1.04364760e-01 -5.70680320e-01
-1.94876686e-01 -7.79947579e-01 -3.17415893e-01 9.28451419e-01
-9.84449565e-01 -1.34635007e+00 -7.40695119e-01 8.19193244e-01
8.61873567e-01 1.91482410e-01 5.03900290e-01 1.98666021e-01
-4.97574449e-01 5.65535665e-01 -4.50733930e-01 7.43087649e-01
4.62315947e-01 -1.34817767e+00 1.15538478e+00 4.71519142e-01
3.22522931e-02 4.53392565e-01 1.94985464e-01 -1.00970244e+00
-6.75032735e-01 -8.73685479e-01 3.58115643e-01 -7.24773586e-01
1.52816981e-01 -3.81297827e-01 -7.88930357e-01 9.39105690e-01
3.21627349e-01 1.86453965e-02 1.73421293e-01 -5.62381074e-02
-2.28863224e-01 4.87897187e-01 -1.16914356e+00 7.81651676e-01
1.94496346e+00 -5.51881373e-01 -7.81744301e-01 3.29691589e-01
1.10157073e+00 -1.03505182e+00 -6.12622142e-01 4.69821066e-01
9.32643592e-01 -1.05594754e+00 1.16430223e+00 -7.95708895e-01
6.35451853e-01 -5.60453087e-02 1.70893356e-01 -1.08496344e+00
-4.31285888e-01 -3.74070704e-01 -3.58062595e-01 8.97718966e-01
5.56665622e-02 -1.44027546e-01 1.26767075e+00 7.36330926e-01
-1.70693696e-01 -8.86856377e-01 -5.75910270e-01 -6.51527047e-01
2.16110602e-01 -4.32603121e-01 6.77263916e-01 3.90907228e-01
-2.58776903e-01 2.44983379e-02 -1.27467737e-01 1.32352402e-02
6.26662135e-01 2.96801776e-02 7.88163304e-01 -1.12029767e+00
-2.98127592e-01 -6.69055581e-01 -2.16812342e-01 -1.94304562e+00
4.85154480e-01 -3.26010525e-01 2.88536131e-01 -1.95896995e+00
3.35639149e-01 -1.37527026e-02 6.86104447e-02 6.22452617e-01
-2.99230307e-01 4.42200571e-01 3.09163868e-01 7.64352903e-02
-5.20847678e-01 5.28730631e-01 2.07604122e+00 -4.25785959e-01
-7.14491844e-01 -9.60901827e-02 -1.41944334e-01 9.38296020e-01
4.55909789e-01 -8.82256776e-03 -4.14692461e-01 -7.09303558e-01
-6.15519762e-01 4.55977358e-02 6.27425551e-01 -9.65736628e-01
-5.06853051e-02 -3.24620217e-01 6.30928874e-01 -1.23777831e+00
4.11829889e-01 -4.85081673e-01 -3.06023598e-01 4.13290024e-01
-4.43756700e-01 -1.50667697e-01 2.91174084e-01 2.45160326e-01
-6.32244572e-02 2.30690271e-01 6.03604615e-01 -2.43373930e-01
-8.36184978e-01 6.93712890e-01 -3.16019729e-02 2.51084715e-01
7.92810440e-01 -3.36705536e-01 -4.78176177e-02 -2.58594841e-01
-1.01028407e+00 4.58369821e-01 1.83282718e-01 5.33617377e-01
6.71790957e-01 -1.33388698e+00 -4.40801203e-01 -2.03376058e-02
-3.86893958e-01 8.30957651e-01 4.80656803e-01 4.29737508e-01
-7.33411014e-01 6.00363731e-01 -3.73134613e-01 -6.15484118e-01
-1.15759134e+00 3.63543540e-01 4.26180840e-01 -5.34741543e-02
-9.66422796e-01 1.56272435e+00 6.09936833e-01 -3.32616806e-01
8.77694845e-01 -9.81394053e-01 -1.77846532e-02 -2.05320701e-01
7.22451389e-01 5.24673223e-01 -3.11502814e-01 -3.81842494e-01
-3.34091872e-01 9.24149394e-01 -1.49290353e-01 -3.10566008e-01
9.43085492e-01 1.76870957e-01 7.94979036e-02 2.57220536e-01
8.93481374e-01 -4.40227777e-01 -2.04531574e+00 2.06363618e-01
-4.61491108e-01 -6.77822053e-01 -1.46179363e-01 -1.15351796e+00
-1.49288809e+00 1.11954784e+00 6.25064433e-01 -4.17330027e-01
1.02748382e+00 2.65077561e-01 1.43168902e+00 3.86122227e-01
4.01706129e-01 -1.02266014e+00 4.91498053e-01 6.54068708e-01
1.30364919e+00 -1.18087590e+00 -4.33673173e-01 -5.92808664e-01
-8.03172350e-01 9.21412528e-01 1.24780250e+00 -2.03231514e-01
4.69050080e-01 6.27861321e-01 1.50056869e-01 -1.77372649e-01
6.47457391e-02 -3.09019446e-01 3.76405984e-01 8.35802436e-01
3.35235178e-01 -4.87017483e-02 -1.62489042e-02 9.60830986e-01
-1.82032242e-01 4.66894239e-01 -2.00279027e-01 8.50603461e-01
-5.46552613e-02 -9.54391241e-01 -2.39531964e-01 1.99334130e-01
-1.08492903e-01 -2.42382735e-02 -8.17363739e-01 8.40289891e-01
2.38111958e-01 5.45071483e-01 1.39316276e-01 -6.95589840e-01
6.01633608e-01 -7.92339891e-02 8.89666677e-01 -4.52851415e-01
-6.31415606e-01 1.75929829e-01 -4.48808521e-01 -9.84346926e-01
-4.76830482e-01 -1.43520549e-01 -1.80505109e+00 -3.30944180e-01
-3.74108225e-01 -3.35321844e-01 8.66805986e-02 1.10198152e+00
3.57833534e-01 1.01772249e+00 -1.15512744e-01 -1.83017743e+00
-9.18629318e-02 -1.18141651e+00 -6.89346313e-01 2.01805279e-01
3.36885363e-01 -9.52054739e-01 7.41264969e-02 2.08995178e-01]
|
[6.705166339874268, -0.5973857045173645]
|
01d2dd34-019b-470c-a390-72650fe29910
|
real-time-end-to-end-video-text-spotter-with
|
2207.08417
| null |
https://arxiv.org/abs/2207.08417v3
|
https://arxiv.org/pdf/2207.08417v3.pdf
|
Real-time End-to-End Video Text Spotter with Contrastive Representation Learning
|
Video text spotting(VTS) is the task that requires simultaneously detecting, tracking and recognizing text in the video. Existing video text spotting methods typically develop sophisticated pipelines and multiple models, which is not friend for real-time applications. Here we propose a real-time end-to-end video text spotter with Contrastive Representation learning (CoText). Our contributions are three-fold: 1) CoText simultaneously address the three tasks (e.g., text detection, tracking, recognition) in a real-time end-to-end trainable framework. 2) With contrastive learning, CoText models long-range dependencies and learning temporal information across multiple frames. 3) A simple, lightweight architecture is designed for effective and accurate performance, including GPU-parallel detection post-processing, CTC-based recognition head with Masked RoI. Extensive experiments show the superiority of our method. Especially, CoText achieves an video text spotting IDF1 of 72.0% at 41.0 FPS on ICDAR2015video, with 10.5% and 32.0 FPS improvement the previous best method. The code can be found at github.com/weijiawu/CoText.
|
['Ping Luo', 'Zhongyuan Wang', 'Size Li', 'Hong Zhou', 'Chunhua Shen', 'Jiahong Li', 'Zhuang Li', 'Wejia Wu']
|
2022-07-18
| null | null | null | null |
['text-spotting']
|
['computer-vision']
|
[ 2.26697668e-01 -8.38872194e-01 -1.35478795e-01 1.77911390e-02
-9.71342921e-01 -4.41111982e-01 3.72491211e-01 -1.59341604e-01
-4.16108906e-01 -9.37451199e-02 8.64812210e-02 -3.68251830e-01
5.78817368e-01 -1.86906219e-01 -6.32291853e-01 -5.84662199e-01
4.58508968e-01 2.46148601e-01 7.04526365e-01 2.15301856e-01
3.77296895e-01 2.35272855e-01 -1.42980182e+00 6.89677298e-01
5.78481138e-01 1.03624594e+00 5.11030614e-01 1.38398170e+00
-7.00633004e-02 1.09632611e+00 -3.97318780e-01 -8.91358182e-02
9.35129896e-02 -1.37399167e-01 -3.54848176e-01 4.05645631e-02
6.36534452e-01 -8.89942765e-01 -6.99017167e-01 5.90510011e-01
5.78987718e-01 -1.47353128e-01 4.22194511e-01 -1.08838773e+00
-1.82916194e-01 3.33428502e-01 -1.01130354e+00 2.81758577e-01
4.31425661e-01 3.34043682e-01 7.02832282e-01 -1.23720884e+00
3.09040338e-01 9.74350274e-01 7.39771724e-01 6.57989979e-01
-7.41867483e-01 -7.37259686e-01 1.05841450e-01 1.38135269e-01
-1.35310471e+00 -7.34973252e-01 1.09294802e-01 -4.99822915e-01
1.00330341e+00 4.62974638e-01 3.67785335e-01 1.21628213e+00
3.69662434e-01 1.44751823e+00 7.41181254e-01 -4.44645494e-01
-1.21615075e-01 -2.76907295e-01 1.20640449e-01 8.26517224e-01
1.19839534e-01 -1.67264476e-01 -9.91125762e-01 1.92084849e-01
7.23407209e-01 4.62129503e-01 -2.57676750e-01 3.68284792e-01
-1.45491898e+00 3.73500019e-01 -1.65713131e-01 2.44070023e-01
-3.63832042e-02 5.14066994e-01 8.18141341e-01 1.76125124e-01
6.31262243e-01 -4.00091231e-01 -4.39555198e-01 -6.01311207e-01
-1.49391735e+00 -2.53640503e-01 4.79237825e-01 1.06871557e+00
3.07751089e-01 1.99865118e-01 -5.58123231e-01 6.60364866e-01
4.52962965e-01 1.07723665e+00 7.03609943e-01 -1.29368499e-01
8.66759896e-01 3.84588271e-01 1.58609301e-02 -7.38000453e-01
-2.92136997e-01 9.63668525e-02 -6.10775173e-01 -5.08162081e-02
3.35405767e-01 -4.06840414e-01 -1.04492652e+00 8.48657966e-01
2.56028235e-01 5.79962850e-01 -4.08874780e-01 9.39209461e-01
9.86681104e-01 9.19445813e-01 -3.26745994e-02 -2.79944558e-02
1.42650092e+00 -1.33966088e+00 -8.14361095e-01 -4.19644177e-01
7.81187594e-01 -1.24276721e+00 1.10930443e+00 3.91832590e-01
-8.34855914e-01 -3.24679673e-01 -8.59813273e-01 -3.69062155e-01
-1.51560858e-01 8.28400135e-01 1.28768593e-01 7.35390007e-01
-1.03942513e+00 2.22442359e-01 -1.30104220e+00 -4.91932064e-01
4.20455009e-01 6.14581227e-01 5.74814295e-03 1.71106793e-02
-5.92740595e-01 2.42380992e-01 2.54455842e-02 1.28705099e-01
-8.03414166e-01 -4.61571455e-01 -4.32682991e-01 -6.92835227e-02
6.68874621e-01 -4.07739401e-01 1.30196059e+00 -1.09364426e+00
-1.73842013e+00 7.36684740e-01 -6.01893783e-01 -3.00953984e-01
8.78924906e-01 -7.13339329e-01 -3.14549476e-01 3.48025352e-01
9.55273211e-02 2.96283305e-01 1.47971010e+00 -4.74956244e-01
-8.45666885e-01 -1.67875335e-01 -6.44752562e-01 2.79852360e-01
-5.64255774e-01 6.09643519e-01 -1.18652356e+00 -8.95113945e-01
-1.41147420e-01 -8.65192831e-01 3.10372114e-01 3.03342283e-01
-4.69621509e-01 -1.73569188e-01 1.41464198e+00 -8.62473190e-01
1.36392438e+00 -2.18919873e+00 -2.04651877e-01 -3.74906808e-01
3.90344143e-01 5.84646642e-01 5.96684627e-02 3.59443456e-01
2.09634915e-01 -6.55549839e-02 3.68890405e-01 -6.52115345e-01
-5.35370409e-02 -4.72144514e-01 -4.53625679e-01 6.79249704e-01
-4.98938225e-02 1.06691253e+00 -4.89363492e-01 -7.02398896e-01
7.18277216e-01 5.63645840e-01 -2.27043867e-01 2.46234477e-01
-1.52182803e-01 2.67304573e-02 -5.76129615e-01 1.04923809e+00
5.37643373e-01 -4.74673390e-01 -1.17510520e-01 -1.09304145e-01
-3.11935246e-01 -2.74651162e-02 -1.12517560e+00 1.50443316e+00
-2.55850554e-01 1.32380295e+00 3.77521291e-02 -3.52696329e-01
5.38940012e-01 2.58542478e-01 4.02870506e-01 -6.52713895e-01
4.91694927e-01 1.05284199e-01 -7.53977299e-01 -7.27776766e-01
8.24516714e-01 4.77483273e-01 2.11034760e-01 5.58182776e-01
-3.30734104e-01 5.15374541e-01 7.72227123e-02 2.83206165e-01
1.28534269e+00 3.84785980e-01 -1.43194526e-01 5.59546836e-02
4.56408828e-01 -1.33458421e-01 2.74440318e-01 8.32574606e-01
-3.09598327e-01 8.44180703e-01 3.16649228e-01 -4.27496225e-01
-8.77444386e-01 -5.59253812e-01 2.46022239e-01 1.52588463e+00
2.05185786e-01 -7.75018573e-01 -7.15478539e-01 -6.24873519e-01
-2.46394679e-01 1.53140277e-01 -5.14919102e-01 3.42544943e-01
-8.29552114e-01 -5.74901342e-01 8.07589591e-01 7.32004881e-01
4.70693082e-01 -7.80181289e-01 -7.98647404e-01 -4.69074398e-03
-2.53484964e-01 -1.42680275e+00 -1.22891688e+00 -1.33177876e-01
-8.48429918e-01 -1.06072891e+00 -1.00031936e+00 -7.68351912e-01
4.10830051e-01 7.91250288e-01 5.76027334e-01 2.25719169e-01
-6.39733016e-01 5.39072096e-01 -5.00828087e-01 -2.91521894e-03
-5.88977560e-02 -1.27713485e-02 -2.27510020e-01 3.26105058e-01
4.08012092e-01 2.37696037e-01 -6.68845475e-01 5.76345265e-01
-7.93248177e-01 5.48468947e-01 5.62792122e-01 7.95210183e-01
5.26438415e-01 -3.87251526e-01 -3.01253032e-02 -5.94471693e-01
1.77556559e-01 -1.11730501e-01 -8.38488042e-01 3.12714040e-01
-3.23583484e-01 -4.38277960e-01 6.55787170e-01 -6.16324961e-01
-9.42473054e-01 4.36975360e-01 -2.75856908e-02 -8.30824137e-01
-4.06149738e-02 1.18248947e-01 2.58500934e-01 -7.32042193e-02
4.06093806e-01 5.89964032e-01 -1.29265442e-01 -3.02208751e-01
-3.35326046e-03 1.09287000e+00 3.14651370e-01 -2.26716727e-01
5.84894896e-01 6.01432383e-01 -4.72445846e-01 -1.17321610e+00
-4.35917586e-01 -9.03468013e-01 -6.09745860e-01 -4.04829651e-01
8.91587794e-01 -1.33593655e+00 -9.39303935e-01 8.82541060e-01
-1.02341473e+00 -7.92134821e-01 4.35528219e-01 4.62810934e-01
-5.20179152e-01 7.30608344e-01 -8.17584574e-01 -8.81571829e-01
-8.72738540e-01 -9.59285855e-01 1.78613341e+00 2.12952450e-01
5.70401773e-02 -7.24768341e-01 -2.27259144e-01 5.92263579e-01
4.68141854e-01 -1.53930141e-02 -7.08661526e-02 -3.92348439e-01
-8.35548937e-01 -4.17088270e-01 -5.55882454e-01 -1.56041980e-01
-9.07107219e-02 3.90900791e-01 -1.08789229e+00 -4.69707578e-01
-2.72550970e-01 -2.33164907e-01 1.09409964e+00 5.28531730e-01
1.08522308e+00 -7.19119310e-02 -5.31246841e-01 6.49873912e-01
1.37249911e+00 2.66521633e-01 6.07479751e-01 2.17057973e-01
1.09552979e+00 4.02947981e-03 8.61659408e-01 6.48022056e-01
1.84014350e-01 7.86346912e-01 4.83435206e-02 -1.77342072e-01
-4.24258083e-01 -3.09150666e-02 9.53283191e-01 8.01325321e-01
2.76259840e-01 -5.75052023e-01 -1.07217932e+00 2.56982625e-01
-2.20490456e+00 -9.93068278e-01 -4.81395692e-01 2.09383130e+00
3.39105010e-01 4.48890701e-02 3.61018151e-01 8.47219527e-02
1.17934978e+00 1.78471506e-01 -6.76978111e-01 1.46816730e-01
3.92856970e-02 -7.48355240e-02 7.55814135e-01 3.74242157e-01
-1.46052742e+00 1.27996480e+00 5.37361288e+00 1.05837560e+00
-1.43219674e+00 1.91819131e-01 5.72756231e-01 -6.00419343e-01
4.71168667e-01 -3.20842028e-01 -1.10378528e+00 7.69884706e-01
1.06879449e+00 7.11701214e-02 3.07180047e-01 5.54327071e-01
5.85895300e-01 -1.11364037e-01 -9.26458597e-01 1.51307034e+00
3.96161139e-01 -1.50626206e+00 -2.10780770e-01 -1.76475257e-01
3.80494088e-01 4.44024086e-01 7.15559125e-02 7.27094263e-02
-1.92045331e-01 -8.20364058e-01 1.03548110e+00 3.45875800e-01
1.32989740e+00 -4.45717752e-01 4.95916843e-01 1.04012638e-01
-1.68856311e+00 4.32721339e-02 -5.24095166e-03 3.26027811e-01
8.86844322e-02 4.17770863e-01 -8.96551669e-01 3.24210316e-01
7.64886022e-01 1.21523547e+00 -7.48692393e-01 1.04189706e+00
1.67558212e-02 9.47814703e-01 -4.54717815e-01 -5.16291857e-01
1.25681609e-01 2.60433674e-01 3.28642309e-01 1.78713202e+00
2.92171657e-01 -1.86589912e-01 1.96577281e-01 3.21763575e-01
-1.04149505e-01 1.47656828e-01 -1.81984127e-01 -2.41310954e-01
2.53312200e-01 1.27598512e+00 -1.19255519e+00 -3.94158483e-01
-6.88718021e-01 1.37123215e+00 -1.78209394e-01 1.66688770e-01
-1.25350952e+00 -7.23298073e-01 1.59663945e-01 1.09435715e-01
6.71705008e-01 -2.79928625e-01 -3.67527783e-01 -1.49928117e+00
2.69418001e-01 -8.53429854e-01 3.43530416e-01 -7.61167705e-01
-8.24204803e-01 3.63305926e-01 -6.74911320e-01 -1.37193739e+00
1.62128955e-01 -8.65570366e-01 -5.96279383e-01 5.54960549e-01
-1.31957054e+00 -1.38024628e+00 -6.94181025e-01 8.03225756e-01
1.16136682e+00 -1.72027946e-01 4.19098258e-01 5.17019808e-01
-1.16982341e+00 8.76165271e-01 5.06716430e-01 4.69055653e-01
1.07456470e+00 -8.93086195e-01 8.40835392e-01 1.17401028e+00
5.76318316e-02 1.02456056e-01 3.79028857e-01 -8.27218711e-01
-2.04542661e+00 -1.36956394e+00 6.10458374e-01 -5.88173687e-01
7.73601294e-01 -6.63365602e-01 -7.25114822e-01 7.59886265e-01
8.94599333e-02 2.00038984e-01 3.90269548e-01 -2.72733033e-01
-3.53652507e-01 7.58042466e-03 -6.73388481e-01 6.25219762e-01
7.01815426e-01 -6.16534531e-01 -6.11250587e-02 6.05950773e-01
4.66967165e-01 -7.31450081e-01 -2.42239594e-01 -1.89844459e-01
8.03226829e-01 -8.74208391e-01 5.67516387e-01 4.15652022e-02
2.55226493e-01 -4.37978059e-01 -2.15195436e-02 -1.98960364e-01
5.89569695e-02 -1.14496291e+00 -4.41797495e-01 1.02699018e+00
1.28649995e-01 -4.89583969e-01 1.08617067e+00 2.92413652e-01
-7.46830776e-02 -5.86155057e-01 -1.04222441e+00 -6.15840614e-01
-4.24879789e-01 -6.41164720e-01 1.31168738e-01 6.78774655e-01
2.92233098e-02 4.34541732e-01 -8.17533672e-01 1.53259307e-01
3.45162660e-01 9.00997445e-02 8.63893569e-01 -7.81589448e-01
-2.64286667e-01 -3.65074992e-01 -2.95358241e-01 -1.72284043e+00
-1.70279115e-01 -4.73854691e-01 2.56026983e-01 -1.37158287e+00
3.34337503e-01 -3.85767780e-02 4.48707454e-02 5.53812325e-01
-2.51771688e-01 2.09001094e-01 3.75273854e-01 5.14890611e-01
-1.24572635e+00 3.90799195e-01 8.02666187e-01 -1.08404428e-01
-5.71979135e-02 -6.97074458e-02 -2.11974651e-01 5.99236369e-01
7.68162429e-01 -4.98211145e-01 1.92540474e-02 -7.27773964e-01
-6.90472722e-02 1.61286384e-01 3.41953367e-01 -9.52522695e-01
7.77473330e-01 8.65402073e-02 5.61599135e-01 -1.04928982e+00
3.43925029e-01 -6.42881632e-01 -3.57758969e-01 6.21879458e-01
-1.18121952e-01 7.83878267e-02 4.94580120e-01 6.43593788e-01
2.16395900e-01 -1.89706117e-01 6.69890463e-01 4.50126380e-01
-8.16424668e-01 4.11147445e-01 -6.75959945e-01 -3.41415964e-03
1.08884454e+00 -3.27698410e-01 -5.38507402e-01 -1.40258878e-01
-1.87362894e-01 1.78569108e-01 4.42965984e-01 5.55795074e-01
8.40904951e-01 -8.04998279e-01 -7.71163285e-01 2.24998325e-01
-2.29814332e-02 -2.29402706e-01 3.44236851e-01 1.06043136e+00
-9.15952086e-01 6.60506248e-01 2.15026617e-01 -9.71929789e-01
-1.82271981e+00 2.79717982e-01 2.47651562e-01 -4.14580256e-02
-9.64921296e-01 8.10858250e-01 8.22248980e-02 2.77110398e-01
5.06726623e-01 -3.19178492e-01 6.42951503e-02 -2.74882764e-02
1.06477427e+00 6.07390285e-01 1.44809276e-01 -4.74705935e-01
-5.12033045e-01 8.73873413e-01 -4.64138448e-01 -4.89603467e-02
9.28004503e-01 -2.39500374e-01 3.78242403e-01 4.52911317e-01
9.57447767e-01 -1.17449192e-02 -1.57765937e+00 -9.29187164e-02
-5.16753383e-02 -7.92441308e-01 3.04884076e-01 -6.93073034e-01
-1.02454174e+00 8.95006657e-01 8.61758530e-01 -1.73756212e-01
1.18031216e+00 -3.43453467e-01 1.00631094e+00 3.88317913e-01
-1.58122426e-03 -1.24545586e+00 3.78742039e-01 5.90174317e-01
5.26443601e-01 -1.34518003e+00 8.91945139e-02 -2.88938612e-01
-6.42107427e-01 1.42576241e+00 5.17800331e-01 2.75748577e-02
2.29316071e-01 7.00775385e-01 1.62163422e-01 5.50220609e-02
-1.02973413e+00 -1.45686761e-01 1.62076801e-01 1.56802714e-01
5.89312315e-01 -1.17466375e-01 2.47535884e-01 1.28902853e-01
4.96338636e-01 8.49953443e-02 4.95328128e-01 1.02080715e+00
-4.65252876e-01 -5.94446421e-01 -5.72195530e-01 6.07318163e-01
-6.39792860e-01 -3.14260155e-01 -3.45579296e-01 5.37176132e-01
-4.28839326e-01 1.13364244e+00 7.75004998e-02 -5.17089844e-01
1.26390487e-01 -1.82042062e-01 9.68946591e-02 -3.95017534e-01
-7.03936100e-01 7.40228176e-01 -1.45648777e-01 -6.94202483e-01
-7.66017437e-02 -7.87723064e-01 -1.31751895e+00 -6.49332106e-01
-6.35691285e-01 -2.61495173e-01 7.24660218e-01 8.28449488e-01
5.54503679e-01 4.90726352e-01 6.46907568e-01 -9.43974972e-01
-4.90258187e-02 -7.79765368e-01 -3.11324179e-01 -9.63389575e-02
6.66734457e-01 -2.45138451e-01 -3.13120008e-01 4.27047104e-01]
|
[11.982329368591309, 2.2038843631744385]
|
f1db9ec6-d237-486b-88ec-8b08feb7c638
|
dct-dual-channel-training-of-action
|
2306.15913
| null |
https://arxiv.org/abs/2306.15913v1
|
https://arxiv.org/pdf/2306.15913v1.pdf
|
DCT: Dual Channel Training of Action Embeddings for Reinforcement Learning with Large Discrete Action Spaces
|
The ability to learn robust policies while generalizing over large discrete action spaces is an open challenge for intelligent systems, especially in noisy environments that face the curse of dimensionality. In this paper, we present a novel framework to efficiently learn action embeddings that simultaneously allow us to reconstruct the original action as well as to predict the expected future state. We describe an encoder-decoder architecture for action embeddings with a dual channel loss that balances between action reconstruction and state prediction accuracy. We use the trained decoder in conjunction with a standard reinforcement learning algorithm that produces actions in the embedding space. Our architecture is able to outperform two competitive baselines in two diverse environments: a 2D maze environment with more than 4000 discrete noisy actions, and a product recommendation task that uses real-world e-commerce transaction data. Empirical results show that the model results in cleaner action embeddings, and the improved representations help learn better policies with earlier convergence.
|
['Harshad Khadilkar', 'Hardik Meisheri', 'Pranavi Pathakota']
|
2023-06-28
| null | null | null | null |
['product-recommendation']
|
['miscellaneous']
|
[ 1.62967399e-01 1.60888478e-01 -3.65787297e-01 -2.98880577e-01
-9.02262866e-01 -5.02054393e-01 7.61027694e-01 -2.03561574e-01
-7.19630837e-01 7.82893896e-01 7.84725845e-01 -3.03867310e-01
-7.69769354e-03 -6.50902212e-01 -7.66746759e-01 -6.84153795e-01
-2.55204916e-01 4.78165716e-01 -1.51600257e-01 -3.27173591e-01
2.01925397e-01 1.52367681e-01 -1.32390392e+00 1.49922952e-01
5.47292113e-01 9.08737183e-01 2.17831731e-01 9.82732356e-01
2.86083579e-01 1.13973868e+00 -2.85826176e-01 -2.62667030e-01
5.77013135e-01 -3.34374756e-01 -7.00106442e-01 5.07798605e-02
1.33018553e-01 -7.91891873e-01 -8.53532732e-01 7.86143124e-01
4.52711523e-01 5.43536067e-01 5.37518024e-01 -1.14693964e+00
-9.83527482e-01 5.45736790e-01 1.03052638e-01 7.62666017e-02
2.21695691e-01 7.47503698e-01 1.28127396e+00 -2.34856695e-01
7.86288202e-01 1.53733230e+00 3.89580607e-01 9.90608990e-01
-1.51278055e+00 -1.97792962e-01 4.97013122e-01 2.54632026e-01
-6.11452281e-01 -4.65569109e-01 3.62882525e-01 -4.80933189e-02
1.59285009e+00 -3.70185822e-01 5.25668681e-01 1.81559825e+00
6.31488025e-01 1.15809858e+00 8.36740434e-01 4.61062193e-02
6.93996668e-01 -1.43404022e-01 -2.79380053e-01 4.80427742e-01
3.85516062e-02 5.32198310e-01 -4.34684515e-01 -7.12365136e-02
6.34398818e-01 3.24779183e-01 2.91963238e-02 -7.29932785e-01
-1.29773343e+00 9.54515100e-01 1.81477875e-01 5.18908277e-02
-7.30834126e-01 8.19319785e-01 4.70734298e-01 7.15663552e-01
8.82489160e-02 9.47499394e-01 -8.13614130e-01 -9.29328263e-01
-2.92447478e-01 6.72128618e-01 8.00646067e-01 7.35350311e-01
3.52521181e-01 3.46568078e-01 -2.91020185e-01 4.95209306e-01
2.78844595e-01 5.58919489e-01 7.02301621e-01 -1.59824169e+00
6.20120585e-01 2.52847165e-01 6.63051188e-01 -6.13556802e-01
-3.23022127e-01 -2.19438061e-01 -2.39260137e-01 4.20241684e-01
3.99897486e-01 -4.27509069e-01 -8.71397793e-01 2.09781241e+00
1.69738516e-01 2.40994304e-01 5.57971120e-01 6.67520285e-01
-2.18741536e-01 7.58156717e-01 1.23541065e-01 2.40127340e-01
7.30065525e-01 -1.32347620e+00 -8.65327418e-01 -6.92163944e-01
8.66697729e-01 -1.47802979e-01 1.06158018e+00 6.06678247e-01
-9.74866927e-01 -4.58813697e-01 -1.14493334e+00 -1.87714905e-01
-3.10361922e-01 -1.66622382e-02 8.22529197e-01 3.16083252e-01
-9.84267831e-01 1.06247020e+00 -1.47761738e+00 -2.95725018e-01
4.66577858e-01 3.41993779e-01 -3.97896945e-01 -1.57677293e-01
-1.17936599e+00 1.29591763e+00 3.78028035e-01 -1.98235705e-01
-1.43191314e+00 -3.28363925e-01 -1.06362247e+00 1.23505823e-01
4.88570511e-01 -3.40298921e-01 1.75728834e+00 -7.50649989e-01
-2.11539197e+00 -3.25788818e-02 1.87807500e-01 -1.03713000e+00
3.77945513e-01 -5.81731200e-01 -4.37410176e-01 1.31260790e-02
4.77180779e-02 6.61696196e-01 7.60453045e-01 -5.73354363e-01
-6.92139745e-01 -4.63417292e-01 2.48818114e-01 3.94721538e-01
-3.53321910e-01 -5.00563979e-01 -9.90042910e-02 -4.02087957e-01
-3.29053789e-01 -1.08645201e+00 -7.15039790e-01 6.60439879e-02
8.19267631e-02 -1.70291990e-01 5.57735324e-01 -6.37417674e-01
9.39541280e-01 -2.08851814e+00 5.67831933e-01 -2.24107206e-01
-4.23190705e-02 1.92423284e-01 -6.57776237e-01 6.82904422e-01
1.45836681e-01 -2.34777570e-01 7.24021569e-02 -3.89812142e-01
5.51474452e-01 7.21584797e-01 -5.61754465e-01 4.01848227e-01
3.41500640e-01 1.11087883e+00 -1.19363260e+00 2.79308528e-01
2.41531372e-01 2.97170103e-01 -9.58231330e-01 1.87356830e-01
-6.80888355e-01 3.85034978e-01 -7.59706140e-01 2.71819174e-01
6.14483505e-02 -3.61241400e-02 5.52463591e-01 4.22006369e-01
1.70008957e-01 6.56007826e-01 -1.17454660e+00 2.15108848e+00
-5.93254983e-01 2.96705693e-01 -5.01470007e-02 -9.33594525e-01
7.74928033e-01 2.60903358e-01 6.11633122e-01 -1.06407511e+00
7.62927458e-02 -4.29820502e-03 4.85054553e-02 -4.90822345e-01
5.55454373e-01 -5.75032122e-02 -3.45608622e-01 7.63500035e-01
2.85290360e-01 -1.04838744e-01 1.04353093e-01 1.40012845e-01
1.53519869e+00 6.17269397e-01 1.40371054e-01 2.42499352e-01
-1.20521290e-02 -1.60469800e-01 7.05646157e-01 8.62793624e-01
-6.16155982e-01 2.17586607e-02 9.17585015e-01 -6.40825212e-01
-1.20911729e+00 -1.03744984e+00 3.31477672e-01 1.24626625e+00
-7.02188835e-02 -5.28247952e-01 -3.73501122e-01 -8.75809848e-01
3.90942872e-01 1.21470797e+00 -7.61834323e-01 -7.18037546e-01
-5.34528315e-01 -3.07601571e-01 4.25747842e-01 9.24957335e-01
3.70465100e-01 -1.03528583e+00 -6.49481297e-01 4.66729939e-01
1.99731931e-01 -8.89873922e-01 -5.14032185e-01 5.71250677e-01
-8.35248172e-01 -8.68564367e-01 -2.83362150e-01 -5.16984999e-01
1.27723739e-01 -1.69398114e-01 9.31708634e-01 -5.29211104e-01
-8.14505271e-04 5.61065853e-01 -3.53023916e-01 -1.90376014e-01
-6.36178493e-01 -1.49340834e-02 4.59380984e-01 -2.48284221e-01
6.76372945e-01 -4.91750807e-01 -6.07938945e-01 8.88222829e-02
-8.44687760e-01 -1.83052972e-01 5.55287540e-01 1.13901567e+00
3.51307899e-01 -1.79660693e-01 5.30248880e-01 -5.72311044e-01
8.04325938e-01 -5.41653752e-01 -6.32748067e-01 1.15941398e-01
-8.00644696e-01 7.63321936e-01 8.61019313e-01 -4.35528547e-01
-1.07076037e+00 1.95542619e-01 -2.08158448e-01 -2.79279798e-01
-1.36421487e-01 5.02999462e-02 1.62378338e-03 4.24735606e-01
7.53328681e-01 3.26522738e-01 2.94719279e-01 -6.86037421e-01
8.60552490e-01 6.69209182e-01 2.62533754e-01 -5.33814490e-01
4.36678171e-01 2.30408996e-01 -2.75590628e-01 -2.98191458e-01
-5.80309808e-01 -1.69903040e-02 -4.33807701e-01 3.09945852e-01
8.16626012e-01 -8.98062110e-01 -7.08017826e-01 1.05921514e-01
-7.94494569e-01 -8.75489414e-01 -6.21745467e-01 8.13885570e-01
-1.25516617e+00 1.25631750e-01 -8.99978995e-01 -6.85246944e-01
4.44230326e-02 -1.34783781e+00 1.06467617e+00 5.26531264e-02
-1.12099126e-01 -9.31884825e-01 4.50442225e-01 3.93715277e-02
4.19614047e-01 -2.03159340e-02 7.57852316e-01 -8.40520442e-01
-5.65688431e-01 -1.26946807e-01 3.55067372e-01 6.78533912e-01
8.64382014e-02 -3.58387500e-01 -7.18077600e-01 -4.13978666e-01
-4.28438298e-02 -8.53362620e-01 9.31062460e-01 2.82439649e-01
1.10895157e+00 -2.89786130e-01 -1.20561831e-01 5.07829547e-01
1.15041685e+00 5.19911647e-01 6.70463800e-01 4.51063484e-01
1.90342858e-01 1.69102460e-01 8.09814870e-01 6.25676513e-01
3.59155983e-01 5.16497314e-01 5.76622069e-01 5.62181830e-01
1.44993693e-01 -8.23859930e-01 9.05170083e-01 5.53240597e-01
2.62788206e-01 -2.44602412e-01 -5.82480848e-01 5.02821386e-01
-2.10644484e+00 -1.11537063e+00 8.01259100e-01 1.94498658e+00
8.24824214e-01 2.34826192e-01 8.05894360e-02 -2.47501686e-01
1.14478730e-02 4.77938652e-01 -1.17803741e+00 -8.63379300e-01
3.51971090e-01 3.39447647e-01 6.91428959e-01 6.20265782e-01
-1.20134056e+00 1.17055261e+00 7.20405531e+00 4.06867653e-01
-8.88539016e-01 2.39306204e-02 3.40666384e-01 -5.14071941e-01
-2.42179126e-01 -1.87624782e-01 -6.56359613e-01 4.09138173e-01
1.40118778e+00 -4.70099486e-02 8.63409460e-01 1.14581656e+00
1.71900794e-01 1.45421281e-01 -1.40315557e+00 7.24050283e-01
-1.56987891e-01 -1.25789952e+00 -2.72277325e-01 1.76515535e-01
8.39764535e-01 3.20744812e-01 4.42155093e-01 8.69521260e-01
1.09303892e+00 -1.04778540e+00 5.31197727e-01 3.96964848e-01
5.28178811e-01 -6.39776289e-01 3.99901509e-01 4.36064333e-01
-7.29274929e-01 -7.58217871e-01 -4.80779827e-01 -3.72640491e-01
1.67819276e-01 -3.25881541e-01 -8.38768661e-01 8.41735974e-02
2.76171356e-01 1.26010072e+00 -2.18280628e-01 5.90388954e-01
-3.13542813e-01 4.10180062e-01 -4.18007411e-02 -3.31536472e-01
7.60615170e-01 -2.70520061e-01 3.33702415e-01 6.47099853e-01
1.92160442e-01 -8.03899020e-02 2.05224782e-01 7.25419939e-01
-3.27235609e-02 -4.22336489e-01 -9.56529200e-01 -7.29430497e-01
2.28462577e-01 7.17049241e-01 -1.75825849e-01 -2.32045874e-01
-4.98047113e-01 1.23564756e+00 5.38621843e-01 6.08720362e-01
-9.45724726e-01 -2.93696940e-01 1.53036034e+00 -4.39072490e-01
7.65698254e-01 -5.99796832e-01 -4.58337506e-03 -1.28969228e+00
-1.71638042e-01 -1.12355030e+00 6.02381192e-02 -4.86763567e-01
-9.76594865e-01 2.16275483e-01 -4.08297092e-01 -1.17177081e+00
-7.33045995e-01 -9.84526098e-01 -1.93934768e-01 5.61671555e-01
-1.51332653e+00 -5.03750682e-01 5.47016561e-01 2.93842316e-01
7.23216176e-01 -3.11791182e-01 1.09356201e+00 1.27164170e-01
-5.72162986e-01 5.59810460e-01 9.07792151e-01 2.76748464e-02
5.83619893e-01 -1.50896001e+00 7.76290536e-01 6.29941046e-01
1.74165845e-01 5.92632115e-01 7.20763803e-01 -5.33823252e-01
-1.91141772e+00 -1.12112617e+00 4.86180633e-01 -8.97997081e-01
8.36376846e-01 -3.73632014e-01 -4.23051745e-01 1.28201687e+00
1.11237004e-01 -2.45497897e-02 4.97922808e-01 1.33092478e-01
-4.73738343e-01 -1.21215448e-01 -1.06457448e+00 8.59763741e-01
1.13963330e+00 -6.60914898e-01 -7.64650106e-01 3.87395233e-01
9.73861516e-01 -3.61989647e-01 -8.31621587e-01 -3.26844394e-01
6.11294687e-01 -6.11187994e-01 9.28711534e-01 -1.44760799e+00
5.72504401e-01 5.23974076e-02 -4.22719836e-01 -1.78000259e+00
-4.75036204e-01 -7.34938025e-01 -4.22708184e-01 3.52687746e-01
4.86923307e-01 -6.43611848e-01 6.60817206e-01 8.47533047e-01
-4.65091430e-02 -9.22423542e-01 -8.90282154e-01 -9.14068282e-01
2.69719213e-01 -4.91688520e-01 5.59051871e-01 2.97265291e-01
2.33046815e-01 2.13996395e-01 -6.70650065e-01 -8.99988785e-02
2.56049544e-01 -8.23401138e-02 7.26813436e-01 -6.44211233e-01
-6.82545125e-01 -1.53347820e-01 -5.22413075e-01 -1.61065066e+00
3.13292354e-01 -7.28339136e-01 1.97650626e-01 -1.52546465e+00
-1.33734047e-01 -4.61682454e-02 -5.72916985e-01 6.24046445e-01
1.21370897e-01 -3.50914955e-01 2.11678594e-01 -3.42382997e-01
-1.01905334e+00 1.23855722e+00 1.28292835e+00 -2.07439095e-01
-2.12304115e-01 -8.98277462e-02 -8.32417607e-01 5.12064159e-01
7.51254797e-01 -4.63591009e-01 -7.52112508e-01 -7.86269963e-01
1.51948586e-01 4.16221330e-03 8.26257467e-02 -8.98297369e-01
-6.83320463e-02 -4.00006682e-01 3.61424893e-01 4.73843850e-02
6.63988888e-01 -7.73812354e-01 -4.07415211e-01 6.55294895e-01
-9.72417235e-01 1.99174836e-01 7.10945651e-02 1.18989289e+00
9.77653041e-02 9.21256021e-02 6.03531957e-01 -9.09949616e-02
-9.24401462e-01 2.62349695e-01 -5.10098755e-01 1.43760160e-01
1.24056125e+00 5.86365424e-02 -2.41856948e-01 -5.96460342e-01
-9.24881995e-01 6.00669682e-01 4.15455520e-01 9.16250050e-01
5.34175038e-01 -1.52078974e+00 -3.71294796e-01 2.79424220e-01
5.96465357e-02 -6.65347576e-01 6.43269941e-02 5.20019710e-01
-2.56536603e-01 5.99626541e-01 -4.86837059e-01 -8.07597041e-02
-6.28537476e-01 6.69769287e-01 3.95655751e-01 -4.86588597e-01
-8.20848525e-01 7.43070662e-01 -2.53067583e-01 -6.98365808e-01
5.94636917e-01 -7.86863983e-01 -3.32977884e-02 -2.32198507e-01
8.18829417e-01 3.22251409e-01 -2.91066200e-01 -1.52487323e-01
-7.69402161e-02 -6.14004955e-02 -2.67574281e-01 -4.01044428e-01
1.63960731e+00 -2.97613256e-02 5.96221745e-01 4.13509786e-01
1.26366735e+00 -6.30638480e-01 -2.10076666e+00 -3.77393067e-01
3.64957266e-02 -7.04734623e-01 1.14967264e-01 -9.18496251e-01
-6.87298000e-01 8.38360310e-01 6.95855498e-01 3.77753153e-02
6.29382193e-01 -3.00849229e-01 1.04297423e+00 9.26003635e-01
5.16329408e-01 -1.72733235e+00 4.04116333e-01 8.38591397e-01
6.16023540e-01 -1.16665566e+00 -3.94818813e-01 6.78820968e-01
-1.22558880e+00 9.70010996e-01 6.06035173e-01 -3.84166211e-01
4.80911583e-01 1.67296186e-01 -1.70285493e-01 6.03049509e-02
-1.46009862e+00 -2.31311336e-01 -3.64177465e-01 7.77330458e-01
1.28117139e-02 1.14349984e-01 -8.94162133e-02 5.96418142e-01
2.08390191e-01 1.42106444e-01 4.35716540e-01 1.12974977e+00
-5.09271026e-01 -1.35552502e+00 1.84070960e-01 5.12101054e-01
-1.82771683e-01 2.21007064e-01 -8.20343271e-02 4.98203158e-01
-1.89556986e-01 8.67748082e-01 1.29504114e-01 -5.90211868e-01
4.37225014e-01 4.24017519e-01 4.97063458e-01 -6.08838081e-01
-2.47933000e-01 -2.74638474e-01 2.11295128e-01 -1.46810460e+00
5.31767309e-02 -7.46526480e-01 -1.32388270e+00 -2.08131194e-01
2.58812129e-01 -2.05400586e-02 5.14727294e-01 8.28472614e-01
7.87146389e-01 6.18143678e-01 7.45988011e-01 -7.90002644e-01
-1.69888079e+00 -8.89618754e-01 -7.98499644e-01 5.47654629e-01
5.96379340e-01 -7.79800594e-01 -1.43853590e-01 -2.31043488e-01]
|
[4.14881706237793, 1.6179198026657104]
|
6178682e-29d4-4852-a687-70d76e7a1153
|
easytransfer-a-simple-and-scalable-deep
|
2011.09463
| null |
https://arxiv.org/abs/2011.09463v3
|
https://arxiv.org/pdf/2011.09463v3.pdf
|
EasyTransfer -- A Simple and Scalable Deep Transfer Learning Platform for NLP Applications
|
The literature has witnessed the success of leveraging Pre-trained Language Models (PLMs) and Transfer Learning (TL) algorithms to a wide range of Natural Language Processing (NLP) applications, yet it is not easy to build an easy-to-use and scalable TL toolkit for this purpose. To bridge this gap, the EasyTransfer platform is designed to develop deep TL algorithms for NLP applications. EasyTransfer is backended with a high-performance and scalable engine for efficient training and inference, and also integrates comprehensive deep TL algorithms, to make the development of industrial-scale TL applications easier. In EasyTransfer, the built-in data and model parallelism strategies, combined with AI compiler optimization, show to be 4.0x faster than the community version of distributed training. EasyTransfer supports various NLP models in the ModelZoo, including mainstream PLMs and multi-modality models. It also features various in-house developed TL algorithms, together with the AppZoo for NLP applications. The toolkit is convenient for users to quickly start model training, evaluation, and online deployment. EasyTransfer is currently deployed at Alibaba to support a variety of business scenarios, including item recommendation, personalized search, conversational question answering, etc. Extensive experiments on real-world datasets and online applications show that EasyTransfer is suitable for online production with cutting-edge performance for various applications. The source code of EasyTransfer is released at Github (https://github.com/alibaba/EasyTransfer).
|
['Wei Lin', 'Deng Cai', 'Yaliang Li', 'Xianyan Jia', 'Hanjie Pan', 'Chengyu Wang', 'Ang Wang', 'Jun Huang', 'Cen Chen', 'Peng Li', 'Minghui Qiu']
|
2020-11-18
| null | null | null | null |
['compiler-optimization']
|
['computer-code']
|
[-4.61083233e-01 -2.09992141e-01 -4.02074635e-01 -4.74244148e-01
-1.15577257e+00 -7.47231662e-01 3.73808503e-01 -1.29386485e-01
-2.03194946e-01 5.26455283e-01 4.96482924e-02 -6.83529496e-01
1.21523537e-01 -7.34853148e-01 -6.44822299e-01 -2.83759356e-01
8.72298703e-02 1.01014912e+00 -9.21136886e-02 -4.03466254e-01
-4.26184051e-02 1.52074739e-01 -1.22901154e+00 8.68459940e-01
8.16758811e-01 1.06389058e+00 5.25673330e-01 8.80926073e-01
-7.24258900e-01 9.13897753e-01 -2.21268833e-01 -3.97133470e-01
2.80638337e-01 -4.11139131e-02 -1.11138034e+00 -4.93393034e-01
2.26334974e-01 -2.66550720e-01 1.74544901e-01 5.12568533e-01
7.21746683e-01 6.11567236e-02 1.75084770e-01 -1.44379568e+00
-6.28776848e-01 5.47266901e-01 -5.74997775e-02 -9.00310799e-02
5.83246171e-01 3.49922329e-01 1.23365378e+00 -9.85454023e-01
5.26123881e-01 1.20463634e+00 6.73491657e-01 3.52196544e-01
-7.94721425e-01 -9.05687928e-01 -2.01281339e-01 2.01867566e-01
-1.11094999e+00 -5.38723946e-01 3.74927580e-01 -1.37197956e-01
1.51884472e+00 2.96417147e-01 2.81672001e-01 1.15778279e+00
1.62902668e-01 1.37094474e+00 1.06459320e+00 -6.88557982e-01
-8.61123204e-02 5.71004808e-01 2.09433019e-01 7.96118617e-01
-4.85840619e-01 -3.05326462e-01 -5.75788856e-01 -4.17058825e-01
3.58747512e-01 -1.96842805e-01 -1.29091024e-01 -3.71337007e-03
-1.32964778e+00 9.01791513e-01 3.06324512e-01 5.07082045e-01
-3.07357937e-01 -5.75428456e-02 6.17494226e-01 5.87115467e-01
6.09843254e-01 4.45246816e-01 -1.13865614e+00 -4.79348838e-01
-6.26011074e-01 3.05385441e-01 1.37875390e+00 1.25922751e+00
8.88247967e-01 -3.45820844e-01 2.17238199e-02 1.26684093e+00
4.05946702e-01 4.38981533e-01 8.36211562e-01 -9.69053924e-01
7.50771165e-01 7.08584726e-01 -1.61826789e-01 -7.05366135e-01
-2.98466444e-01 -8.63044581e-04 -8.11289608e-01 -5.97589254e-01
4.03034210e-01 -2.98653424e-01 -5.06457150e-01 1.39775431e+00
4.42082047e-01 1.08652353e-01 1.57858416e-01 6.80398822e-01
1.05239451e+00 1.28658509e+00 2.52444834e-01 5.74909039e-02
1.39853716e+00 -1.40681171e+00 -4.97712255e-01 -4.64972287e-01
1.23047733e+00 -9.89809573e-01 1.75243294e+00 5.55478811e-01
-1.15902650e+00 -5.54504037e-01 -5.25114715e-01 -5.68606138e-01
-7.18974590e-01 2.34352186e-01 9.43969309e-01 3.48318547e-01
-8.76049161e-01 1.30154639e-01 -7.24202752e-01 -7.76975095e-01
1.11413924e-02 3.40852201e-01 -2.92882621e-01 -3.86349231e-01
-1.50229204e+00 8.30664933e-01 4.26201224e-01 7.56554231e-02
-4.47938204e-01 -9.46192920e-01 -6.98819399e-01 -7.99552575e-02
3.39682817e-01 -8.53971720e-01 1.66840327e+00 -1.12496388e+00
-1.97787678e+00 8.67059529e-01 -2.45075777e-01 -4.10011172e-01
2.09306300e-01 -3.37213725e-01 -5.89412093e-01 -1.05892755e-01
3.58218029e-02 6.86261714e-01 4.19849545e-01 -5.59001863e-01
-4.88865167e-01 3.59650701e-02 2.26126425e-02 3.59708786e-01
-6.34732783e-01 5.72009146e-01 -7.85828650e-01 -2.54989266e-01
-6.29705966e-01 -9.52238142e-01 -2.89325491e-02 -1.76778764e-01
-1.84910610e-01 -6.40645921e-01 7.13037252e-01 -8.58915746e-01
1.24432230e+00 -1.92846155e+00 -1.21669680e-01 -7.94346929e-02
-1.73369154e-01 5.17577469e-01 -3.62016827e-01 1.02187395e+00
1.92199394e-01 3.14751454e-02 -1.35854736e-03 -3.32362294e-01
3.79368752e-01 2.35427678e-01 -4.48264271e-01 -7.22722858e-02
-6.62117377e-02 1.28498781e+00 -6.74613059e-01 -5.64370513e-01
2.44239986e-01 1.12441920e-01 -6.28800452e-01 5.18383086e-01
-6.85524166e-01 1.05776377e-01 -5.50426722e-01 6.89609826e-01
4.69953179e-01 -5.59030712e-01 3.51931691e-01 5.74107021e-02
-6.04291670e-02 7.91755140e-01 -8.68675709e-01 1.97029436e+00
-8.98428500e-01 5.53988814e-01 3.54624957e-01 -8.04198027e-01
7.68914819e-01 4.60256487e-01 3.06204528e-01 -6.27923131e-01
2.98009850e-02 2.85949558e-01 -2.22107425e-01 -7.28857577e-01
5.22527993e-01 1.40901431e-01 -2.05488697e-01 8.77820194e-01
2.41010711e-01 -2.84685373e-01 1.26944199e-01 3.86474401e-01
8.75280619e-01 1.58000767e-01 3.76726687e-01 -1.36702165e-01
5.56294799e-01 3.62760097e-01 2.88923442e-01 3.77163053e-01
9.97077450e-02 8.42818841e-02 7.48467073e-02 -3.42948645e-01
-7.07534134e-01 -7.87443101e-01 -2.43470948e-02 1.80376399e+00
-5.08917212e-01 -7.36982703e-01 -6.32662296e-01 -5.70606172e-01
3.73994708e-02 8.93538415e-01 1.39373347e-01 1.81364715e-01
-4.87809539e-01 -6.29092574e-01 7.01333284e-01 3.26199293e-01
8.48583579e-01 -1.57435608e+00 2.09527269e-01 2.50314862e-01
-5.96574605e-01 -1.22862232e+00 -5.40537357e-01 -1.88498870e-01
-6.12036884e-01 -7.01135039e-01 -3.83715570e-01 -1.07714927e+00
2.58951306e-01 1.13963336e-01 1.25909269e+00 -1.70045644e-02
-2.27422014e-01 4.87849057e-01 -4.81712610e-01 -5.30193448e-01
-5.22459745e-01 5.52869976e-01 -7.56891593e-02 -2.03043833e-01
6.40869796e-01 -2.51640022e-01 -2.80173630e-01 2.85379112e-01
-8.19803059e-01 2.36436993e-01 4.74458396e-01 7.38097250e-01
2.42215067e-01 -3.04105908e-01 7.64525890e-01 -1.03820825e+00
9.83933032e-01 -7.47015655e-01 -4.41140473e-01 6.09785140e-01
-5.51030576e-01 -3.35590214e-01 7.61617899e-01 -5.08677423e-01
-1.23121023e+00 -3.44678134e-01 -5.19077480e-01 -7.87418038e-02
-3.07904840e-01 9.93761361e-01 -1.41697973e-01 1.42493993e-01
5.16484916e-01 2.00147659e-01 2.65411404e-03 -7.57582068e-01
5.91158450e-01 1.22078013e+00 3.52986008e-02 -7.52318740e-01
3.82744402e-01 -1.45434812e-01 -8.39816868e-01 -7.74338901e-01
-8.78049970e-01 -5.83524406e-01 -3.70782137e-01 1.87526524e-01
4.90567952e-01 -1.02488434e+00 -6.89341068e-01 4.45153832e-01
-1.07117498e+00 -8.83929610e-01 1.57552436e-01 3.95857573e-01
-4.08890694e-01 3.22769254e-01 -1.06752121e+00 -3.96760762e-01
-1.02494383e+00 -9.93123829e-01 9.40745652e-01 4.04689871e-02
-3.76355231e-01 -1.34952223e+00 7.89282843e-02 9.22738492e-01
5.78739643e-01 -6.37249708e-01 1.06899953e+00 -9.71616089e-01
-5.73413730e-01 -2.97423959e-01 -1.86735764e-01 6.27245128e-01
-2.49766499e-01 -3.72681692e-02 -9.11359012e-01 -2.85534650e-01
-2.88087398e-01 -8.79750848e-01 1.54193982e-01 1.48125082e-01
1.31716025e+00 -4.23704863e-01 -4.28785324e-01 5.73197007e-01
1.11182082e+00 -1.83830753e-01 5.58577180e-01 4.36628044e-01
4.88803983e-01 6.05031312e-01 7.91258216e-01 2.96762995e-02
8.19150388e-01 6.68944716e-01 -2.51907498e-01 -9.05731767e-02
5.57549037e-02 -3.67851973e-01 8.21908534e-01 1.55977678e+00
4.25818145e-01 -3.71091604e-01 -1.31951988e+00 3.59503090e-01
-1.99415302e+00 -6.33053839e-01 -1.42340213e-01 1.82312202e+00
1.12846506e+00 -2.80016422e-01 -4.65680249e-02 -6.72304273e-01
2.66819715e-01 -2.25324765e-01 -3.59963179e-01 -7.09124088e-01
1.58385560e-01 3.10603410e-01 7.63793066e-02 4.27501321e-01
-8.14674020e-01 1.33595109e+00 6.26524448e+00 1.32716167e+00
-1.24901640e+00 4.79496986e-01 5.68424702e-01 1.15128756e-02
-1.41532049e-01 -6.24980330e-02 -9.86604750e-01 4.15762812e-01
1.42906404e+00 -4.97960418e-01 6.03233337e-01 9.93871868e-01
1.60023674e-01 1.14065900e-01 -9.83242035e-01 7.98693180e-01
6.65456930e-04 -1.50846899e+00 -1.44630298e-01 -1.21260442e-01
3.94918174e-01 8.19743097e-01 -1.33000821e-01 1.07119977e+00
3.42566550e-01 -8.86007726e-01 1.87757164e-01 1.82365566e-01
6.99552119e-01 -5.67619383e-01 8.08583081e-01 1.00431049e+00
-1.02690315e+00 -6.98876232e-02 -4.21023309e-01 -1.53046906e-01
1.93196654e-01 4.25612658e-01 -1.24474192e+00 5.50635874e-01
7.44549513e-01 5.93272865e-01 -2.95508683e-01 7.05546021e-01
-1.17992461e-01 9.66434121e-01 -4.90621775e-01 -1.85306072e-01
2.25633919e-01 -2.76689082e-01 1.53764844e-01 1.48938644e+00
2.23748356e-01 -1.08230282e-02 4.44403142e-01 6.92029536e-01
-2.22635895e-01 7.46807218e-01 -5.58175027e-01 -2.74414003e-01
5.01615942e-01 1.73862255e+00 -2.22457990e-01 -5.35330951e-01
-5.45026004e-01 8.38756323e-01 5.32590866e-01 3.04695010e-01
-8.69500935e-01 -2.93276489e-01 5.49173832e-01 -2.26313937e-02
-1.18026808e-01 -3.08108717e-01 8.47899690e-02 -1.37448823e+00
-1.21983908e-01 -1.39808691e+00 5.75825274e-01 -1.12932599e+00
-1.61354053e+00 6.92269742e-01 8.37998688e-02 -9.57858086e-01
-4.48195815e-01 -7.47418225e-01 -5.59043944e-01 9.77310836e-01
-1.59232831e+00 -1.50953782e+00 -5.46301678e-02 7.27517843e-01
7.66943812e-01 -3.13650191e-01 1.26610363e+00 5.15868664e-01
-7.64145911e-01 5.93818665e-01 1.38616249e-01 2.01084688e-01
1.09196043e+00 -8.24040532e-01 4.43618536e-01 3.04707676e-01
2.22487748e-01 9.13638175e-01 1.57946810e-01 -3.21607143e-01
-1.86449099e+00 -1.13730478e+00 1.36743009e+00 -6.74908757e-01
1.08948541e+00 -7.50818551e-01 -8.72959554e-01 1.16320324e+00
5.17911673e-01 -2.81710565e-01 9.69364762e-01 4.09761012e-01
-2.67022640e-01 -2.31038079e-01 -9.52367246e-01 4.99368936e-01
5.43861985e-01 -7.52038240e-01 -6.57465696e-01 1.03894722e+00
8.68527114e-01 -4.45404530e-01 -1.02224970e+00 2.22154722e-01
4.62403178e-01 -4.91415083e-01 9.20662761e-01 -6.41759336e-01
3.39870691e-01 1.12707943e-01 -2.68335622e-02 -1.07283187e+00
-1.86023563e-01 -8.47175837e-01 -1.53792664e-01 1.52050054e+00
7.18594074e-01 -1.07453382e+00 5.63485920e-01 8.55983913e-01
-7.02580735e-02 -8.13077450e-01 -7.08754480e-01 -5.27974963e-01
2.60832369e-01 -9.96494293e-01 5.79345405e-01 1.09737194e+00
3.38660568e-01 7.75200605e-01 -1.90296695e-01 -1.60358567e-02
7.46645182e-02 4.50791597e-01 1.16418910e+00 -7.12539256e-01
-6.26237452e-01 -2.23957956e-01 4.27436471e-01 -1.43369007e+00
4.10584688e-01 -1.38855505e+00 -1.98228151e-01 -1.48421764e+00
6.26882240e-02 -8.14603925e-01 -6.33879751e-02 9.91900980e-01
1.59190252e-01 9.20100957e-02 2.66005188e-01 3.49472791e-01
-6.35996163e-01 4.08917248e-01 1.04719031e+00 1.03679458e-02
-2.41438612e-01 3.77772301e-02 -7.00577259e-01 5.98371685e-01
9.64907825e-01 -2.88144916e-01 -3.38379741e-01 -6.89513683e-01
2.90590793e-01 -7.16824755e-02 1.01404823e-01 -4.82824922e-01
3.41890544e-01 -7.84215182e-02 -3.02282013e-02 -4.23768163e-01
3.28541368e-01 -6.93543196e-01 -8.08808431e-02 1.07147858e-01
-4.10183787e-01 1.09778680e-01 5.57203293e-01 -1.43878996e-01
-2.39240021e-01 -2.05069169e-01 3.02538991e-01 -2.79700190e-01
-8.25801373e-01 4.02135372e-01 -3.07896465e-01 6.13164268e-02
8.88913453e-01 2.85437852e-01 -4.07357901e-01 -5.60421884e-01
-3.04218590e-01 6.50088072e-01 5.07796109e-02 6.26810670e-01
2.40755111e-01 -1.05584371e+00 -7.22754180e-01 3.67454946e-01
1.70657977e-01 -1.46648645e-01 2.16690034e-01 9.55952168e-01
-7.23863423e-01 7.34312236e-01 1.72605440e-01 -4.16436553e-01
-1.22235978e+00 3.61971021e-01 1.41300187e-01 -6.53522491e-01
-4.98358339e-01 9.24194813e-01 1.94905233e-02 -1.39757061e+00
1.11095861e-01 -5.66135347e-01 3.01180601e-01 -2.39838585e-01
7.01758027e-01 1.71543397e-02 1.47403434e-01 -1.16464891e-01
-2.28646159e-01 1.82080254e-01 -8.83894637e-02 3.54846604e-02
1.31119263e+00 -4.93394993e-02 -5.05780220e-01 6.57601893e-01
1.37578261e+00 -3.13282982e-02 -3.93989980e-01 -2.85738647e-01
-4.05871868e-02 -4.14012298e-02 1.18775152e-01 -1.32176983e+00
-6.27377450e-01 9.42573369e-01 5.56622781e-02 1.04226261e-01
1.02494133e+00 -6.43069670e-02 1.35024500e+00 7.73187280e-01
6.42084837e-01 -1.05626774e+00 -2.30487987e-01 7.35873818e-01
1.02017355e+00 -1.20182168e+00 -1.13471337e-01 -3.20555598e-01
-8.97017777e-01 9.89240348e-01 5.63440979e-01 1.43419251e-01
7.23464489e-01 3.73863786e-01 3.66680115e-01 3.78964120e-03
-1.36381423e+00 1.37512207e-01 2.78795213e-01 2.21648365e-01
7.35962808e-01 1.44764990e-01 -4.61045317e-02 7.37531722e-01
-2.82608330e-01 4.54171479e-01 -1.48250267e-01 8.27877343e-01
-1.05231307e-01 -1.56248200e+00 -1.68083921e-01 3.61771733e-01
-2.79746652e-01 -5.13705254e-01 -3.20648819e-01 9.44694877e-01
-8.63133371e-02 9.10261273e-01 -1.65632129e-01 -1.35871485e-01
9.26630571e-02 5.18718243e-01 2.29525626e-01 -7.82494724e-01
-1.17058313e+00 5.74794523e-02 5.00664234e-01 -5.63053548e-01
-1.22122146e-01 -3.96788627e-01 -1.25485277e+00 -6.18068039e-01
-2.96842128e-01 3.54218215e-01 7.72248268e-01 8.47636938e-01
7.92938054e-01 -6.32344782e-02 2.34321073e-01 -5.53075194e-01
-4.08386618e-01 -1.26853991e+00 -3.24084997e-01 1.55800521e-01
-3.11366230e-01 -9.59600434e-02 -1.90730467e-01 -1.74337178e-02]
|
[10.922798156738281, 8.569759368896484]
|
d83c7fc7-a33b-4946-85f3-4e9cd356aad9
|
understanding-bias-in-anomaly-detection-a
| null | null |
https://openreview.net/forum?id=sAzh_FTFDxz
|
https://openreview.net/pdf?id=sAzh_FTFDxz
|
Understanding Bias in Anomaly Detection: A Semi-Supervised View with PAC Guarantees
|
Anomaly detection presents a unique challenge in machine learning, due to the scarcity of labeled anomaly data. Existing work attempts to mitigate such problems via semi-supervised learning, i.e., augmenting unsupervised anomaly detection models with additional labeled anomaly samples. However, the labeled data often does not align with the target distribution and therefore introduces harmful bias to the trained model. In this paper, we aim to understand the effect of a biased anomaly set for anomaly detection. In particular, we formally state the anomaly detection problem as a semi-supervised learning task. We focus on the anomaly detector's recall at a given false positive rate as the main performance metric. Given two different anomaly score functions, we formally define their difference in performance as the relative scoring bias of the anomaly detectors. We establish the first finite sample rates for estimating the relative scoring bias for semi-supervised anomaly detection. We then empirically validate our theoretical results on both synthetic and real-world datasets. Furthermore, we provide extensive empirical study on how a biased training anomaly set affects the anomaly score function and therefore the resulting detection performance. Our case study demonstrates scenarios in which the biased anomaly set can be useful, and provides a solid benchmark for future research.
|
['Haitao Zheng', 'Yuxin Chen', 'Ziyu Ye']
|
2021-01-01
| null | null | null | null |
['supervised-anomaly-detection', 'semi-supervised-anomaly-detection']
|
['computer-vision', 'computer-vision']
|
[ 2.83895522e-01 1.60373822e-02 -8.40967819e-02 -6.46671891e-01
-7.99260318e-01 -4.71707314e-01 5.15038252e-01 4.27169085e-01
-3.01023275e-01 3.06661308e-01 -2.25329533e-01 -3.62441540e-01
-1.20694109e-04 -5.46710074e-01 -6.06916606e-01 -5.52806377e-01
-3.80511791e-01 4.29033637e-01 2.15967372e-01 2.61431485e-01
3.49233985e-01 4.54993635e-01 -1.58532655e+00 -1.23799287e-01
1.07412124e+00 1.01727712e+00 -6.09663904e-01 5.58634758e-01
-1.33592367e-01 4.96527880e-01 -8.63017797e-01 -2.23283812e-01
6.44331098e-01 -5.51712155e-01 -5.58198512e-01 4.45261389e-01
5.77913046e-01 -3.96657735e-01 1.13059759e-01 1.39520788e+00
8.51266906e-02 9.74819586e-02 8.79810035e-01 -1.92302346e+00
-1.07916020e-01 4.65619326e-01 -7.24848747e-01 6.40480101e-01
1.60380378e-01 9.93769020e-02 1.12379193e+00 -7.86198139e-01
1.42444611e-01 9.06799972e-01 5.28575718e-01 5.65768480e-01
-1.28603077e+00 -7.28106976e-01 2.65312314e-01 -5.88093176e-02
-1.22010076e+00 -3.52456361e-01 8.03677261e-01 -4.54584748e-01
4.52793837e-01 2.31644958e-01 2.02712446e-01 9.39170480e-01
-3.16513772e-03 7.81305969e-01 8.08681667e-01 -4.82009053e-01
6.39575124e-01 1.33984864e-01 5.80702841e-01 4.30519104e-01
7.81915963e-01 2.27298960e-01 -2.11313680e-01 -7.07180858e-01
3.74251068e-01 2.26909757e-01 2.04382259e-02 -6.34945691e-01
-6.16185486e-01 9.27106738e-01 -3.65914069e-02 4.39232588e-02
-3.03207070e-01 -6.40009940e-02 5.33121228e-01 6.37083769e-01
5.45650780e-01 5.79333305e-01 -4.03417379e-01 -5.13621159e-02
-7.59878874e-01 2.34366983e-01 8.27089310e-01 6.82485342e-01
5.53392828e-01 2.65626699e-01 5.88950841e-03 8.95884633e-01
4.48217541e-01 5.24209321e-01 3.24056655e-01 -6.01957977e-01
2.81217307e-01 6.35713041e-01 1.72141969e-01 -8.08736861e-01
-2.50986874e-01 -3.00745070e-01 -4.19594467e-01 6.28963709e-02
8.56680095e-01 -2.86575496e-01 -8.43415797e-01 1.85329592e+00
2.33549953e-01 3.37456077e-01 3.03496979e-02 6.81222558e-01
-1.86333191e-02 3.05269420e-01 7.34105408e-02 -4.10381109e-01
8.95974994e-01 -7.88851917e-01 -6.41632438e-01 -3.66998911e-01
1.12903666e+00 -5.43333232e-01 1.17607176e+00 2.84481883e-01
-6.83931708e-01 1.80057898e-01 -9.85070944e-01 8.63839507e-01
-1.14963740e-01 -3.43404353e-01 4.06009853e-01 8.30581188e-01
-5.77176690e-01 3.72588784e-01 -1.10327017e+00 -5.52675605e-01
3.89563173e-01 1.11923963e-02 -1.70585662e-01 4.04744130e-03
-9.23094034e-01 7.19877481e-01 3.50690514e-01 -2.89489567e-01
-6.52299821e-01 -5.98883152e-01 -9.15442228e-01 -6.38234466e-02
5.62737226e-01 8.36318135e-02 1.54900014e+00 -1.13391829e+00
-7.38591790e-01 7.98160434e-01 -1.40865535e-01 -8.36291134e-01
4.61277515e-01 -2.92077333e-01 -6.89578235e-01 -1.02080800e-01
1.93111569e-01 -5.86043000e-02 8.84563267e-01 -1.40417683e+00
-9.83929336e-01 -4.68883932e-01 -3.29883367e-01 -2.06135541e-01
-3.31819594e-01 1.15810603e-01 -2.98649967e-02 -8.31452906e-01
2.92358816e-01 -8.13990712e-01 -4.40494180e-01 -2.65742272e-01
-5.80438197e-01 -5.07765636e-02 1.05907845e+00 -8.53642300e-02
1.40840602e+00 -2.43522739e+00 -6.31302118e-01 9.18732762e-01
2.42510736e-01 1.87264204e-01 -7.59022832e-02 2.51728565e-01
-2.90883034e-01 1.19457550e-01 -8.03127587e-01 -1.75656840e-01
-1.83820784e-01 4.03961301e-01 -6.35746777e-01 6.37625575e-01
4.34134930e-01 3.75327945e-01 -1.01917410e+00 -1.55085757e-01
-3.44433412e-02 -2.48666540e-01 -7.82887161e-01 3.88862759e-01
6.91772327e-02 3.07074308e-01 -4.15602475e-01 8.43606174e-01
6.52113676e-01 -8.44627395e-02 3.33680660e-02 4.75902945e-01
3.20495039e-01 -1.48304726e-03 -1.36534488e+00 8.76675546e-01
-1.23700134e-01 5.02107322e-01 -2.17356652e-01 -1.34591496e+00
1.12775910e+00 -7.10047558e-02 5.85036337e-01 -4.89951342e-01
-1.70240194e-01 4.91687298e-01 3.36884528e-01 -3.03378373e-01
2.29059428e-01 -1.57783523e-01 -1.28444269e-01 9.28898096e-01
-1.38897359e-01 2.98752785e-01 1.45020068e-01 9.69086960e-02
1.43046451e+00 -4.91740525e-01 4.92474765e-01 -3.25122833e-01
4.15774077e-01 -1.16946928e-01 6.93393528e-01 1.10255229e+00
-5.34694612e-01 4.77781624e-01 8.49895000e-01 -4.43247169e-01
-9.50680494e-01 -1.40655184e+00 -4.79590178e-01 9.68402326e-01
6.90544844e-02 -2.22619444e-01 -6.68007076e-01 -1.22077274e+00
2.06833735e-01 8.96261692e-01 -4.65421051e-01 -6.63491070e-01
-2.93527156e-01 -1.10272908e+00 7.15231419e-01 5.92915595e-01
2.22903222e-01 -8.97116661e-01 -6.98508263e-01 -1.12492003e-01
3.68885547e-02 -1.00155795e+00 -3.45698029e-01 2.94806600e-01
-1.00055003e+00 -1.37072372e+00 -1.58392310e-01 -3.49796772e-01
1.16083169e+00 1.18773296e-01 1.17812097e+00 2.88364083e-01
-1.15895003e-01 5.83352625e-01 -5.26317239e-01 -7.18743742e-01
-5.30773222e-01 -2.05442369e-01 5.67430079e-01 1.89643160e-01
9.03837860e-01 -4.52726692e-01 -2.83575296e-01 4.88278210e-01
-1.17271376e+00 -7.82938659e-01 3.79838645e-01 8.90451491e-01
4.72931623e-01 8.29013884e-02 7.73178995e-01 -1.22070694e+00
7.41537571e-01 -7.99527049e-01 -7.56686270e-01 -1.59936473e-02
-9.39626515e-01 1.92714229e-01 4.81221229e-01 -4.31120187e-01
-6.93829954e-01 -2.18482073e-02 1.04906842e-01 -3.55283111e-01
-3.93477559e-01 3.88700128e-01 -5.30758463e-02 1.46728233e-01
8.64287198e-01 1.02731593e-01 1.21618353e-01 -3.59508574e-01
-8.33883509e-02 6.42064393e-01 3.99825752e-01 -6.47130907e-01
1.11846888e+00 4.41650748e-01 -1.19554974e-01 -7.83382058e-01
-1.05681860e+00 -6.46707356e-01 -5.97848177e-01 -4.87308726e-02
1.25028506e-01 -5.34849405e-01 -7.94831943e-03 5.28026640e-01
-6.02388263e-01 -2.27535009e-01 -5.97725689e-01 3.41015369e-01
-3.73030692e-01 5.38841724e-01 -3.05612087e-01 -1.20578218e+00
-1.43876255e-01 -9.22641993e-01 9.15769696e-01 -1.16536327e-01
-3.65039468e-01 -1.04919720e+00 3.44458282e-01 -1.78789422e-01
3.09551448e-01 3.67006421e-01 7.69281507e-01 -1.70944715e+00
-5.86508438e-02 -6.17168367e-01 -2.21495345e-01 6.35994077e-01
2.77939916e-01 3.26001607e-02 -9.94874775e-01 -3.57734144e-01
4.64835316e-02 -1.42487347e-01 7.99608707e-01 1.32557437e-01
1.33149004e+00 -3.54304492e-01 -1.03229634e-01 3.51220846e-01
1.12598920e+00 1.83246627e-01 4.02526855e-01 4.26543981e-01
5.92464387e-01 5.56688130e-01 9.39860046e-01 5.80502212e-01
-1.04633667e-01 4.22005594e-01 3.87068450e-01 1.20256312e-01
6.41323268e-01 -1.86910123e-01 3.74648452e-01 4.81385350e-01
3.83026958e-01 -7.38046877e-03 -1.35095370e+00 4.95834500e-01
-1.84212565e+00 -8.16198826e-01 1.08367885e-02 2.86429119e+00
5.51209629e-01 4.18662697e-01 5.18547475e-01 5.42459130e-01
8.54559124e-01 -1.05287090e-01 -7.05928445e-01 -3.29612702e-01
5.76854572e-02 1.38846394e-02 4.46258307e-01 3.49693298e-01
-1.20743954e+00 5.75836480e-01 7.08288383e+00 4.01936740e-01
-9.73782599e-01 -2.55939454e-01 5.76183081e-01 -7.67137483e-02
-4.32073185e-03 4.04885039e-02 -4.96691465e-01 6.05628669e-01
1.10918951e+00 -3.53149474e-01 -2.38078404e-02 1.08757675e+00
-3.72565910e-02 -4.85034510e-02 -1.41891193e+00 7.00762510e-01
7.46316556e-03 -5.73072076e-01 2.20094219e-01 2.83979923e-01
5.84910750e-01 -5.61118908e-02 3.07822190e-02 3.73495966e-01
1.84080765e-01 -7.34505296e-01 2.19784588e-01 2.20944151e-01
5.20191967e-01 -9.54420447e-01 1.04592180e+00 3.38269889e-01
-7.99366653e-01 -3.24041396e-01 -9.97576639e-02 -1.45910919e-01
-2.01268062e-01 8.31710935e-01 -8.02239954e-01 -1.10061662e-02
5.19360304e-01 5.02436161e-01 -7.02592850e-01 1.13014042e+00
3.56039181e-02 1.11798024e+00 -6.04098618e-01 2.74778068e-01
1.07748426e-01 -1.20888621e-01 8.05193841e-01 1.03694582e+00
2.27061436e-01 -2.20877498e-01 3.01543355e-01 6.88298523e-01
1.97514836e-02 1.35847360e-01 -8.26518536e-01 -8.28391984e-02
7.50727952e-01 8.28670681e-01 -7.37385273e-01 -1.94312811e-01
-5.45785964e-01 5.65286100e-01 2.21348152e-01 4.06525493e-01
-6.42150939e-01 -3.68315905e-01 8.43442380e-01 2.32564986e-01
-6.85333461e-02 2.29481667e-01 -3.98349047e-01 -1.33693326e+00
1.55090019e-01 -9.03559864e-01 7.74374366e-01 4.43163179e-02
-1.71347880e+00 1.92417949e-01 2.03291982e-01 -1.53398442e+00
-5.19386828e-01 -4.75387186e-01 -1.04545343e+00 4.57368731e-01
-1.10940528e+00 -4.67854649e-01 -2.01434582e-01 3.41193408e-01
4.23556060e-01 -3.06030214e-01 7.48718917e-01 2.03261152e-01
-9.07831252e-01 1.07603943e+00 1.51721641e-01 5.12279212e-01
8.09525967e-01 -1.49764061e+00 6.28215492e-01 1.34741414e+00
2.07650319e-01 4.65858668e-01 8.89316022e-01 -6.31816030e-01
-9.62766290e-01 -1.28679085e+00 3.32780361e-01 -7.07057357e-01
8.41554046e-01 -1.71562806e-01 -1.36185551e+00 9.13305879e-01
-6.58821106e-01 4.74183887e-01 9.83177662e-01 3.24331611e-01
-5.59600532e-01 4.65303101e-02 -1.58870029e+00 6.03711009e-01
9.44294810e-01 -3.33604604e-01 -6.31981075e-01 1.56227380e-01
4.33540910e-01 -1.94716483e-01 -5.11972964e-01 6.94428623e-01
2.82559007e-01 -9.67292011e-01 6.20239735e-01 -9.88338351e-01
2.44695514e-01 -1.49602085e-01 -2.89234012e-01 -1.31605816e+00
8.10895413e-02 -1.65769264e-01 -5.35225868e-01 1.08712065e+00
5.72891593e-01 -1.04845572e+00 8.18643749e-01 7.82585561e-01
3.33290160e-01 -7.46917963e-01 -8.32104027e-01 -1.08566713e+00
5.48004173e-02 -8.86738002e-01 4.90704745e-01 1.06307518e+00
4.96735126e-02 -3.33185904e-02 -1.58670038e-01 4.80914563e-01
9.11698103e-01 -1.06121659e-01 8.59372377e-01 -1.51498258e+00
-3.14630345e-02 -3.89207810e-01 -8.51951122e-01 -5.75792789e-01
1.13792211e-01 -6.35963619e-01 8.20060521e-02 -5.89674532e-01
2.06664503e-01 -5.11662126e-01 -5.90201259e-01 3.82534504e-01
-3.95564049e-01 1.53377637e-01 -2.20304325e-01 3.02355409e-01
-6.84055924e-01 3.21352959e-01 4.33222890e-01 1.70744151e-01
-3.59923214e-01 3.95967215e-01 -6.68977201e-01 1.04050040e+00
1.06402862e+00 -6.00107491e-01 -3.99843007e-01 9.37417299e-02
-6.74147680e-02 -3.86926919e-01 2.04093382e-01 -8.83762538e-01
-3.79301272e-02 -2.27918893e-01 1.33631304e-01 -2.75630414e-01
-4.04550344e-01 -9.18976307e-01 -6.79374278e-01 5.06373227e-01
-5.00941634e-01 2.10655302e-01 1.92338631e-01 9.09037948e-01
-2.85378188e-01 -3.15957159e-01 9.26916897e-01 3.08852613e-01
-6.05891585e-01 2.74234354e-01 -4.84056085e-01 5.59715748e-01
1.23242545e+00 -8.20277557e-02 -1.31504431e-01 -3.98884326e-01
-4.47689027e-01 4.79351193e-01 5.58300734e-01 3.84120792e-01
6.04515135e-01 -1.41850650e+00 -5.48807561e-01 6.08494997e-01
7.93839455e-01 -6.77180067e-02 -1.97110489e-01 7.65937746e-01
-2.39135534e-01 -4.75528166e-02 -1.17563441e-01 -9.00069416e-01
-1.08174872e+00 4.51692760e-01 3.95687401e-01 -1.56804547e-01
-6.32021070e-01 4.17414218e-01 2.84375045e-02 -5.00226378e-01
5.37256539e-01 -1.80545911e-01 1.10639572e-01 -2.63729542e-01
6.86550915e-01 3.68018985e-01 -1.45955523e-02 -4.56216872e-01
-3.68935138e-01 9.36803967e-02 -4.43614423e-01 -8.78165662e-02
7.78379381e-01 1.49233844e-02 1.23749994e-01 7.34496891e-01
8.69965911e-01 1.32396175e-02 -9.00295913e-01 -5.64295352e-01
6.25765741e-01 -7.56396353e-01 -2.97903299e-01 -4.26833361e-01
-9.86547112e-01 5.27328014e-01 6.90420926e-01 5.76248288e-01
1.16251540e+00 -6.53705299e-02 4.96742487e-01 6.97486699e-01
2.81745732e-01 -1.04185343e+00 1.57122716e-01 5.08008301e-01
4.68922675e-01 -1.64765441e+00 -1.56472430e-01 -3.26423526e-01
-6.56688869e-01 8.16049635e-01 9.81221497e-01 -3.77487689e-01
7.53540456e-01 2.95013517e-01 3.12430143e-01 -2.34153181e-01
-5.87729454e-01 -3.38201560e-02 2.40778685e-01 5.45039177e-01
2.51669049e-01 6.69722483e-02 -5.95230125e-02 6.83128417e-01
-1.92785591e-01 -4.80225056e-01 6.42847538e-01 1.21462059e+00
-5.26685178e-01 -8.31514835e-01 -4.82834727e-01 1.10513747e+00
-6.45464301e-01 1.82295591e-01 -6.04925931e-01 6.38824284e-01
-4.32749927e-01 9.15555596e-01 5.07771730e-01 -4.18524116e-01
5.01617312e-01 4.04313952e-01 -7.28992447e-02 -7.87216067e-01
3.03073018e-03 -2.66418099e-01 -1.43943399e-01 -6.18270159e-01
-1.02385595e-01 -7.97893286e-01 -1.20650387e+00 -8.86836424e-02
-4.70313579e-01 3.23208213e-01 2.13369727e-01 1.18775880e+00
1.19393811e-01 2.31918186e-01 9.21550930e-01 -3.03003877e-01
-1.07572198e+00 -9.56535637e-01 -8.24824631e-01 7.08780587e-01
5.74894786e-01 -7.62365222e-01 -9.78137076e-01 -4.15694118e-01]
|
[7.605988502502441, 2.5022377967834473]
|
41776d66-c6a7-4710-a512-583e933ced9a
|
imaginarynet-learning-object-detectors
|
2210.06886
| null |
https://arxiv.org/abs/2210.06886v1
|
https://arxiv.org/pdf/2210.06886v1.pdf
|
ImaginaryNet: Learning Object Detectors without Real Images and Annotations
|
Without the demand of training in reality, humans can easily detect a known concept simply based on its language description. Empowering deep learning with this ability undoubtedly enables the neural network to handle complex vision tasks, e.g., object detection, without collecting and annotating real images. To this end, this paper introduces a novel challenging learning paradigm Imaginary-Supervised Object Detection (ISOD), where neither real images nor manual annotations are allowed for training object detectors. To resolve this challenge, we propose ImaginaryNet, a framework to synthesize images by combining pretrained language model and text-to-image synthesis model. Given a class label, the language model is used to generate a full description of a scene with a target object, and the text-to-image model deployed to generate a photo-realistic image. With the synthesized images and class labels, weakly supervised object detection can then be leveraged to accomplish ISOD. By gradually introducing real images and manual annotations, ImaginaryNet can collaborate with other supervision settings to further boost detection performance. Experiments show that ImaginaryNet can (i) obtain about 70% performance in ISOD compared with the weakly supervised counterpart of the same backbone trained on real data, (ii) significantly improve the baseline while achieving state-of-the-art or comparable performance by incorporating ImaginaryNet with other supervision settings.
|
['WangMeng Zuo', 'Kailai Feng', 'Zitong Huang', 'Minheng Ni']
|
2022-10-13
| null | null | null | null |
['weakly-supervised-object-detection']
|
['computer-vision']
|
[ 4.69525933e-01 4.67755169e-01 -7.68000931e-02 -2.97588527e-01
-8.17713201e-01 -6.00776613e-01 8.60343933e-01 -2.76373416e-01
-6.20714247e-01 3.00226361e-01 -2.67657369e-01 -2.43391633e-01
6.38998747e-01 -6.66937411e-01 -1.04277980e+00 -5.07315695e-01
3.31598550e-01 5.52870095e-01 4.17115539e-01 2.24986132e-02
-1.32501826e-01 3.16268533e-01 -1.56060910e+00 3.93869609e-01
4.98583823e-01 1.02269685e+00 7.62961864e-01 6.20908141e-01
7.52668828e-02 1.19493246e+00 -5.83599210e-01 -4.00906593e-01
5.91509223e-01 -1.63149163e-01 -5.56395829e-01 6.92356825e-01
7.42373168e-01 -8.42611432e-01 -4.08725828e-01 1.07093835e+00
4.06642914e-01 -5.96526675e-02 4.88724440e-01 -1.25354898e+00
-6.76702023e-01 4.65022385e-01 -7.02462673e-01 -6.82191923e-02
3.90641481e-01 7.68636465e-01 8.11283171e-01 -1.38203812e+00
6.05074048e-01 1.37325859e+00 3.90872717e-01 7.42747247e-01
-1.38861597e+00 -6.64711773e-01 3.76910508e-01 -9.50549543e-02
-1.39206517e+00 -4.13636327e-01 6.57517612e-01 -3.56999606e-01
7.62133181e-01 -2.16940984e-01 5.72833776e-01 1.34923410e+00
-4.51316833e-01 1.10577738e+00 8.76493752e-01 -3.93917441e-01
6.52791187e-02 4.57501620e-01 -1.99908808e-01 9.58414197e-01
3.18587631e-01 3.09528172e-01 -3.06565613e-01 2.34563932e-01
7.34432638e-01 2.47915871e-02 -2.21337363e-01 -3.80702406e-01
-1.38303137e+00 5.04875541e-01 8.41519952e-01 -1.12755433e-01
-4.01272446e-01 2.13213533e-01 1.14935450e-01 2.53484827e-02
2.37441897e-01 4.17222649e-01 -2.80056179e-01 5.36371112e-01
-7.74804115e-01 1.02264903e-01 5.54373801e-01 1.25752890e+00
7.87831604e-01 1.89447865e-01 -1.82761267e-01 4.77506727e-01
3.08531880e-01 9.32749093e-01 1.37371391e-01 -1.02434015e+00
5.61938226e-01 7.33678401e-01 3.15967828e-01 -4.84542400e-01
-2.05819041e-01 -7.26190925e-01 -7.83669889e-01 4.19809490e-01
5.48090816e-01 2.21410636e-02 -1.05890667e+00 1.67279637e+00
4.47039008e-01 3.02682132e-01 2.94913888e-01 1.16923618e+00
7.64855385e-01 7.71533847e-01 4.01621945e-02 1.65145025e-01
1.44109654e+00 -1.32605493e+00 -1.08859338e-01 -7.16951132e-01
4.61942464e-01 -7.02639341e-01 1.28066707e+00 2.84785092e-01
-9.78283286e-01 -8.68666410e-01 -1.02560210e+00 -1.93319604e-01
-7.84662664e-02 5.94996929e-01 3.80200624e-01 2.83363521e-01
-1.04346025e+00 -2.15675291e-02 -7.69462824e-01 -3.41020793e-01
6.34751976e-01 5.49214855e-02 -4.00399745e-01 -5.08058608e-01
-7.66650200e-01 7.39058673e-01 6.98979080e-01 -1.80930980e-02
-1.72284758e+00 -6.66193604e-01 -8.96292329e-01 -1.63940430e-01
7.82596290e-01 -1.01145208e+00 1.36347389e+00 -1.21290851e+00
-1.20387340e+00 1.25356936e+00 2.06558391e-01 -6.92508042e-01
9.30443108e-01 -1.38591275e-01 -1.62889779e-01 3.84049147e-01
4.77627158e-01 1.19112551e+00 1.05097938e+00 -1.57428133e+00
-7.93799758e-01 -3.31312679e-02 4.17252481e-01 2.67696917e-01
-9.99622121e-02 -9.28685665e-02 -8.32061589e-01 -3.28405082e-01
-4.76874895e-02 -9.55134988e-01 -1.79073542e-01 7.20889211e-01
-6.89527273e-01 -3.46569121e-01 8.99844825e-01 -3.22745472e-01
2.09399402e-01 -2.15161705e+00 -1.80828989e-01 -2.63922662e-01
4.71880347e-01 5.08988857e-01 -5.89381993e-01 6.50044829e-02
2.19571367e-01 -2.47877836e-01 -2.06304267e-01 -8.63842905e-01
-1.14349529e-01 2.35126600e-01 -6.68292284e-01 4.96376455e-01
7.14783609e-01 1.13361323e+00 -1.13165236e+00 -5.09398818e-01
5.02438486e-01 4.46404964e-01 -3.69023502e-01 6.13084555e-01
-6.53291404e-01 4.02830571e-01 -2.58083373e-01 6.14624143e-01
6.58058286e-01 -6.63138926e-01 7.69596174e-02 -3.07951838e-01
1.05385192e-01 3.62456366e-02 -1.06792510e+00 1.53890479e+00
-5.77779353e-01 7.09032834e-01 5.29760458e-02 -1.00832510e+00
8.41569602e-01 2.78701745e-02 1.57918837e-02 -6.88441157e-01
1.09209113e-01 8.23553801e-02 -1.17844485e-01 -4.46887910e-01
6.26570508e-02 1.62345439e-01 5.90175129e-02 5.74713826e-01
3.07781160e-01 -2.55998582e-01 1.99089944e-01 4.80249554e-01
9.30477440e-01 3.96191657e-01 1.56164095e-02 1.80234790e-01
5.94621062e-01 3.70907336e-02 4.02751923e-01 1.03006148e+00
-1.50625184e-01 6.80325210e-01 7.82104358e-02 -7.24880040e-01
-1.26124430e+00 -1.28969574e+00 6.72404766e-02 1.17601430e+00
4.37414497e-01 -7.12874904e-02 -7.48334229e-01 -8.23454797e-01
-1.19314402e-01 6.52105272e-01 -5.51653922e-01 -1.27167314e-01
-4.45813388e-01 -4.31012362e-01 6.23844862e-01 5.70960879e-01
9.40140188e-01 -1.06025350e+00 -7.25230336e-01 5.51278479e-02
-2.75771171e-01 -1.65607357e+00 -4.60625798e-01 1.69171728e-02
-4.38463241e-01 -9.95485425e-01 -6.25858605e-01 -9.96386647e-01
9.27811742e-01 8.25452507e-01 1.26478636e+00 2.60337949e-01
-4.20191675e-01 4.35263664e-01 -1.00139931e-01 -4.39673215e-01
-6.83204055e-01 -7.62743205e-02 2.69208867e-02 2.16208354e-01
7.58376643e-02 -2.68671304e-01 -8.20840955e-01 3.75476718e-01
-1.00438428e+00 5.32867730e-01 1.05153251e+00 7.73929536e-01
5.53974688e-01 -2.06348047e-01 4.85211849e-01 -7.25686908e-01
-1.00734189e-01 -2.38460049e-01 -8.39087963e-01 2.56286561e-01
-5.08223951e-01 -1.95706580e-02 6.45985663e-01 -7.44527400e-01
-1.20213151e+00 4.79832411e-01 1.57845229e-01 -7.10990608e-01
-2.16207400e-01 1.44043967e-01 -1.83347836e-01 -6.22982681e-02
1.00290883e+00 5.18524170e-01 -1.64578438e-01 -2.18144804e-01
6.65646017e-01 6.07989252e-01 9.73290682e-01 -5.02547085e-01
1.20199597e+00 7.77294517e-01 -2.90203005e-01 -4.83035982e-01
-1.50992405e+00 -5.57341278e-01 -6.62789822e-01 -2.58126259e-01
8.94466043e-01 -1.50799823e+00 -5.49824059e-01 4.28712189e-01
-1.44564629e+00 -6.60661459e-01 -3.30135971e-01 3.05457473e-01
-4.58699197e-01 5.27376644e-02 -4.03020591e-01 -8.02793384e-01
-1.97868511e-01 -9.65247035e-01 1.60366106e+00 1.28274217e-01
4.22285110e-01 -6.20335877e-01 -5.83777428e-01 6.49413824e-01
2.21215829e-01 6.58346266e-02 3.62802416e-01 -4.35782492e-01
-1.13056004e+00 -3.02598149e-01 -7.56286144e-01 5.54534256e-01
-1.91225693e-01 -3.46288919e-01 -1.26068354e+00 -4.16957021e-01
-1.06475063e-01 -7.79804468e-01 8.81906033e-01 -1.96525216e-01
1.06503892e+00 -3.69363278e-01 -3.17628294e-01 5.54780602e-01
1.21903670e+00 -2.20029131e-01 3.31597537e-01 1.21573448e-01
8.84918749e-01 6.22554898e-01 5.83074868e-01 2.64540106e-01
5.58028877e-01 7.17970371e-01 6.19613588e-01 -5.82419455e-01
-7.45012522e-01 -6.48499429e-01 5.95092416e-01 1.17533945e-01
2.30587304e-01 -2.71275580e-01 -1.02422702e+00 4.86135900e-01
-1.74373507e+00 -8.72438550e-01 4.88235988e-02 1.99176204e+00
8.92097890e-01 4.39623564e-01 1.00453362e-01 -1.78061560e-01
7.98437655e-01 -5.35087734e-02 -9.89299774e-01 5.18715739e-01
-2.58069009e-01 -2.85376877e-01 5.78585505e-01 4.25722778e-01
-1.07936382e+00 1.17879558e+00 5.15001154e+00 5.81910193e-01
-1.24920762e+00 3.10613275e-01 7.09537208e-01 1.21125288e-01
1.05615087e-01 1.36067700e-02 -9.12290573e-01 2.17187688e-01
5.39353728e-01 2.43833810e-02 4.14490879e-01 1.07134056e+00
1.57617599e-01 8.08240622e-02 -1.35155189e+00 1.09293187e+00
3.96851003e-01 -1.22003663e+00 3.03206801e-01 -2.90895347e-02
8.46614182e-01 3.82666439e-01 1.14103563e-01 4.12295610e-01
4.84428763e-01 -8.35567176e-01 1.07252121e+00 2.66197622e-01
9.84516561e-01 -1.28222689e-01 6.49664342e-01 8.18170965e-01
-1.16362417e+00 -1.98522076e-01 -4.78757054e-01 -9.63949114e-02
1.75456151e-01 5.10210037e-01 -1.35503411e+00 2.08150342e-01
5.83600044e-01 8.02161455e-01 -7.60027111e-01 8.59574318e-01
-6.95866525e-01 6.02528691e-01 -4.01636243e-01 1.97374195e-01
2.73008794e-01 1.13364654e-02 4.69174296e-01 1.05847871e+00
-5.76091988e-04 -1.34307042e-01 7.63958454e-01 1.31761050e+00
-3.47077936e-01 -2.31626198e-01 -5.83699763e-01 2.22916424e-01
4.55118716e-01 1.31706262e+00 -7.16409683e-01 -6.19730890e-01
-4.68939930e-01 1.14984739e+00 3.02120805e-01 5.21931469e-01
-8.69830608e-01 -9.52432230e-02 1.78875968e-01 2.41078034e-01
2.61857599e-01 -1.39873356e-01 1.15963973e-01 -1.29636908e+00
1.95020482e-01 -7.72931516e-01 1.36667207e-01 -1.26043177e+00
-1.24036765e+00 7.97730148e-01 -2.36734971e-01 -1.26560545e+00
-1.83899865e-01 -7.90229440e-01 -4.53334421e-01 6.54854357e-01
-1.70674574e+00 -1.77208245e+00 -7.25125909e-01 6.13662302e-01
7.19507456e-01 -1.42660856e-01 4.57850784e-01 2.17629924e-01
-4.11550879e-01 4.52675909e-01 -4.72478539e-01 4.48764443e-01
7.21465647e-01 -1.33915555e+00 5.12991190e-01 9.77786124e-01
4.14260983e-01 2.94900745e-01 4.71969157e-01 -4.62835133e-01
-1.28420627e+00 -1.76192284e+00 3.57272625e-01 -6.71433449e-01
5.54110646e-01 -8.09522152e-01 -6.60641849e-01 7.00046837e-01
-1.30287275e-01 6.77848220e-01 1.22916661e-01 -4.87187266e-01
-7.41178930e-01 -2.62883633e-01 -9.80784595e-01 6.11197770e-01
1.25035059e+00 -7.78849602e-01 -5.05520761e-01 7.67656088e-01
1.17365479e+00 -4.08931375e-01 -3.46155137e-01 2.61070579e-01
3.06810051e-01 -7.14946330e-01 1.29040980e+00 -4.13957268e-01
3.76958579e-01 -8.10587287e-01 -1.13028944e-01 -8.07275653e-01
7.30304793e-02 -4.00961369e-01 -5.26390001e-02 1.10694039e+00
4.18671846e-01 -5.01421154e-01 7.01692164e-01 3.54397982e-01
-1.36756703e-01 -4.95313108e-01 -6.58345819e-01 -8.22638750e-01
-2.95589566e-01 -5.46700418e-01 3.90351951e-01 6.88503802e-01
-7.03330278e-01 6.11975372e-01 -4.16557431e-01 5.53414404e-01
9.86969233e-01 1.61475316e-01 1.27599216e+00 -1.02953374e+00
-5.61868906e-01 -1.33151442e-01 -5.07077277e-01 -1.47187793e+00
1.79710984e-01 -9.35305476e-01 3.32568526e-01 -1.50256217e+00
3.53445470e-01 -4.78086501e-01 -6.80883750e-02 6.44733548e-01
-1.95104495e-01 8.36120963e-01 4.40000117e-01 4.39408302e-01
-1.13119161e+00 5.10152578e-01 1.37864304e+00 -4.74265009e-01
-5.60013323e-05 -4.87778895e-02 -7.14455545e-01 7.94496417e-01
5.09964108e-01 -4.75685716e-01 -5.25860727e-01 -6.27121508e-01
4.56048697e-02 1.13187847e-03 9.55760837e-01 -1.03227687e+00
3.68026793e-01 3.11719831e-02 5.86207867e-01 -4.67111081e-01
3.84658664e-01 -7.78556645e-01 -3.11800241e-01 5.56458652e-01
-4.19770658e-01 -3.51519883e-01 -8.70425894e-04 7.98778892e-01
-8.02617446e-02 2.11846873e-05 7.63525724e-01 -7.29751140e-02
-8.15096021e-01 3.90673995e-01 -1.10697396e-01 -4.55213487e-02
1.13554990e+00 -6.26720265e-02 -5.29845834e-01 -3.82982343e-01
-5.85607111e-01 4.46390629e-01 5.00618339e-01 5.70058107e-01
7.53352642e-01 -1.11419976e+00 -8.54468822e-01 2.14165777e-01
6.59449220e-01 5.93961835e-01 -1.49725871e-02 5.21790147e-01
-3.89086992e-01 1.23337783e-01 1.36805162e-01 -1.09454656e+00
-9.23791826e-01 8.27405512e-01 5.47018528e-01 -5.16147092e-02
-7.04702258e-01 8.17704201e-01 9.36171412e-01 -6.36247098e-01
4.77904886e-01 -2.35954016e-01 2.64066279e-01 -3.61499995e-01
7.79174566e-01 -1.55384243e-01 -1.39719948e-01 -5.43907940e-01
-2.01320410e-01 3.21990728e-01 -1.57591939e-01 -6.15769997e-02
1.18692732e+00 -1.61085546e-01 2.47398153e-01 2.71373302e-01
9.98200238e-01 -3.90783459e-01 -1.95424819e+00 -7.21335292e-01
-2.01584727e-01 -4.06232923e-01 6.31233305e-02 -1.00986302e+00
-1.07447302e+00 9.30035710e-01 7.15377748e-01 -1.73982620e-01
1.00488198e+00 3.75876755e-01 5.75324059e-01 8.60348940e-01
5.63671887e-01 -6.70624435e-01 8.81185830e-01 2.09546939e-01
8.44547451e-01 -1.65777886e+00 -2.92533159e-01 -4.39888954e-01
-6.99776888e-01 9.50429380e-01 9.19481158e-01 -1.76782370e-01
2.25294709e-01 2.49623924e-01 2.54447460e-01 -6.27592281e-02
-7.46558726e-01 -5.19502640e-01 2.64773637e-01 7.43748784e-01
-9.48608145e-02 -1.90527253e-02 6.44642174e-01 2.82226205e-01
9.93548259e-02 -5.20409122e-02 4.54583049e-01 5.46160281e-01
-4.87319380e-01 -8.62577558e-01 -3.93861860e-01 2.48307809e-01
-3.10943108e-02 -1.38618186e-01 -3.96674961e-01 6.66483283e-01
3.95085514e-01 1.00695395e+00 4.72398773e-02 -1.56395152e-01
2.49067172e-01 -3.62715840e-01 2.95455426e-01 -8.87867093e-01
-1.02627590e-01 -1.39001548e-01 -1.35820523e-01 -5.45890808e-01
-4.08309996e-01 -4.55585092e-01 -1.19860208e+00 1.81352422e-01
-5.02910733e-01 -2.52045363e-01 6.08139396e-01 8.82048070e-01
2.79564738e-01 4.02704507e-01 6.21299922e-01 -1.14040077e+00
-7.12492883e-01 -9.06513393e-01 -1.61959335e-01 6.00062788e-01
5.63510001e-01 -5.96474826e-01 -3.43836337e-01 4.67348129e-01]
|
[9.744929313659668, 1.4526995420455933]
|
dafa9a3a-9f42-45e7-b9fd-7fd326f63c76
|
handsformer-keypoint-transformer-for
|
2104.14639
| null |
https://arxiv.org/abs/2104.14639v2
|
https://arxiv.org/pdf/2104.14639v2.pdf
|
Keypoint Transformer: Solving Joint Identification in Challenging Hands and Object Interactions for Accurate 3D Pose Estimation
|
We propose a robust and accurate method for estimating the 3D poses of two hands in close interaction from a single color image. This is a very challenging problem, as large occlusions and many confusions between the joints may happen. State-of-the-art methods solve this problem by regressing a heatmap for each joint, which requires solving two problems simultaneously: localizing the joints and recognizing them. In this work, we propose to separate these tasks by relying on a CNN to first localize joints as 2D keypoints, and on self-attention between the CNN features at these keypoints to associate them with the corresponding hand joint. The resulting architecture, which we call "Keypoint Transformer", is highly efficient as it achieves state-of-the-art performance with roughly half the number of model parameters on the InterHand2.6M dataset. We also show it can be easily extended to estimate the 3D pose of an object manipulated by one or two hands with high performance. Moreover, we created a new dataset of more than 75,000 images of two hands manipulating an object fully annotated in 3D and will make it publicly available.
|
['Vincent Lepetit', 'Mahdi Rad', 'Sayan Deb Sarkar', 'Shreyas Hampali']
|
2021-04-29
| null |
http://openaccess.thecvf.com//content/CVPR2022/html/Hampali_Keypoint_Transformer_Solving_Joint_Identification_in_Challenging_Hands_and_Object_CVPR_2022_paper.html
|
http://openaccess.thecvf.com//content/CVPR2022/papers/Hampali_Keypoint_Transformer_Solving_Joint_Identification_in_Challenging_Hands_and_Object_CVPR_2022_paper.pdf
|
cvpr-2022-1
|
['3d-pose-estimation']
|
['computer-vision']
|
[-2.54058987e-01 -6.40183091e-02 -1.57195795e-02 -1.51038155e-01
-7.57078111e-01 -6.05803668e-01 4.16493118e-01 -3.00807148e-01
-5.49122214e-01 2.71142393e-01 -1.48012534e-01 1.29958794e-01
1.90665592e-02 -1.77787647e-01 -9.93705630e-01 -4.58529860e-01
3.81219499e-02 1.03935528e+00 4.69568461e-01 -6.87504606e-03
1.73445463e-01 8.08648169e-01 -1.37129426e+00 -1.32036224e-01
3.39059234e-01 9.36690331e-01 2.34016076e-01 8.10092092e-01
1.70625538e-01 3.56142163e-01 -6.85813308e-01 -4.72583383e-01
3.46048683e-01 -1.09034814e-02 -9.61351097e-01 1.28530338e-01
8.44275832e-01 -6.90625846e-01 -4.25396532e-01 7.43855178e-01
5.42260766e-01 5.85678630e-02 6.34120166e-01 -1.54786718e+00
-3.60984653e-01 1.43707708e-01 -8.89056027e-01 -2.52394646e-01
5.56827784e-01 6.00317940e-02 9.63760436e-01 -9.06601369e-01
7.72240043e-01 1.59113050e+00 4.81194049e-01 4.60006505e-01
-1.18809032e+00 -5.65223575e-01 3.12407792e-01 2.20018908e-01
-1.48946881e+00 -1.68005407e-01 7.74306297e-01 -5.34437180e-01
8.38403702e-01 9.59735066e-02 8.66133809e-01 1.21180236e+00
1.66298404e-01 1.00124192e+00 9.61914480e-01 -3.90988737e-01
-8.32856521e-02 -3.55513692e-01 4.47637662e-02 1.08005512e+00
-1.18435182e-01 -2.05053210e-01 -4.99028236e-01 -1.21178143e-01
1.34590328e+00 2.26756617e-01 -9.46339294e-02 -8.63447011e-01
-1.43126738e+00 5.19971907e-01 7.40308642e-01 2.83202995e-02
-2.98744828e-01 5.32252908e-01 3.84947993e-02 6.74509183e-02
1.05893269e-01 2.43356302e-01 -5.16236484e-01 -2.05307230e-01
-4.97444481e-01 6.76201284e-01 9.15302396e-01 1.12273681e+00
5.75838864e-01 -7.56001294e-01 -1.23281620e-01 6.55641079e-01
4.57913697e-01 4.17572945e-01 -5.16197421e-02 -1.33144224e+00
4.98769701e-01 4.36551183e-01 3.94384712e-01 -9.20182586e-01
-5.14985085e-01 -1.55732512e-01 -6.03866875e-01 5.11628449e-01
7.54804790e-01 3.58371623e-03 -1.25714540e+00 1.74007440e+00
5.75871706e-01 -7.70690590e-02 -5.93954086e-01 1.17551768e+00
4.03579980e-01 2.56286651e-01 -2.44901791e-01 4.06715274e-01
1.63711405e+00 -1.36311400e+00 -6.83301806e-01 -3.17826390e-01
4.28574085e-02 -8.81256402e-01 9.87877965e-01 5.11519492e-01
-1.11809027e+00 -7.73008049e-01 -8.89433205e-01 -4.36460972e-01
-3.67747486e-01 5.04484594e-01 7.59082377e-01 1.06116906e-01
-9.05841351e-01 6.60371780e-01 -1.34200966e+00 -4.53785151e-01
1.35135338e-01 5.82951128e-01 -6.58284068e-01 6.79640621e-02
-7.57294595e-01 1.19659197e+00 -1.18113942e-02 6.22893393e-01
-8.43776643e-01 -2.12140203e-01 -8.45562637e-01 -1.72727212e-01
5.73958516e-01 -6.59903467e-01 1.22672832e+00 -3.13438445e-01
-1.66130137e+00 9.46731091e-01 -1.65139109e-01 2.17583776e-01
7.87446618e-01 -8.93608272e-01 1.44207537e-01 -2.24327482e-03
1.91680893e-01 8.23732018e-01 9.84086037e-01 -1.47257102e+00
-4.98473287e-01 -7.98867464e-01 2.32923627e-01 8.89482647e-02
7.73467198e-02 4.43023629e-03 -1.29503727e+00 -5.89824796e-01
5.06628573e-01 -1.42725515e+00 -1.17658965e-01 6.17969573e-01
-7.59332776e-01 -4.27377909e-01 7.98936188e-01 -7.00787723e-01
6.82733178e-01 -1.88014328e+00 7.74419785e-01 2.30208486e-01
3.38280469e-01 6.98104203e-02 -2.00378612e-01 1.49014756e-01
1.15162954e-01 -3.30515683e-01 5.32846898e-02 -6.00966513e-01
2.06125036e-01 1.81044132e-01 -7.48149529e-02 5.99462092e-01
2.24082708e-01 1.09411705e+00 -7.55122006e-01 -4.27655011e-01
2.55932480e-01 7.18507946e-01 -2.40662515e-01 6.11342013e-01
-9.93629824e-03 4.22930062e-01 -3.41574997e-01 6.51610732e-01
5.76345444e-01 -2.55309194e-01 2.03267664e-01 -4.86318946e-01
2.38433421e-01 -3.94993648e-02 -1.47478473e+00 2.15890861e+00
-3.27227712e-01 4.54319209e-01 3.32101673e-01 -5.91280520e-01
7.52290249e-01 2.79602051e-01 3.18137646e-01 -3.46345566e-02
1.23224624e-01 7.41567090e-02 -2.81889498e-01 -4.66375113e-01
1.79746553e-01 2.75180250e-01 -5.62017001e-02 4.22697604e-01
3.55718583e-01 -3.72650743e-01 1.03688585e-02 -1.14307038e-01
9.17027652e-01 5.11981487e-01 2.39081867e-02 1.71606898e-01
2.54335850e-01 -4.26393539e-01 2.59957999e-01 6.43118501e-01
-1.66123688e-01 7.58247614e-01 6.63989663e-01 -5.18193305e-01
-9.95133758e-01 -1.27448153e+00 2.57607222e-01 9.84272122e-01
2.68592089e-01 -3.63755703e-01 -7.43743837e-01 -8.53963733e-01
3.89239699e-01 -7.90553167e-02 -8.03311348e-01 -1.44389104e-02
-8.45237076e-01 -7.77861476e-02 2.43993238e-01 9.94965076e-01
4.20158625e-01 -7.86414742e-01 -6.18635654e-01 -5.79921827e-02
-2.23019004e-01 -1.20043612e+00 -6.36954308e-01 3.05487335e-01
-6.38778389e-01 -1.37119138e+00 -1.21313572e+00 -8.86953533e-01
7.83796072e-01 1.54971614e-01 9.98195231e-01 1.56053156e-02
-4.49980646e-01 3.53397965e-01 -1.60116255e-02 -2.42645428e-01
1.87404945e-01 3.18310916e-01 1.01519451e-01 -1.62997052e-01
9.25657526e-02 -4.22868013e-01 -5.66237688e-01 6.01648450e-01
-3.98757160e-01 -9.43963081e-02 7.47042060e-01 6.70395613e-01
5.16864538e-01 -2.54301667e-01 -1.65102944e-01 -4.07200396e-01
3.27258885e-01 2.38666683e-01 -5.64122498e-01 3.93518001e-01
5.48722334e-02 2.64014035e-01 2.31422588e-01 -5.15743375e-01
-8.11813176e-01 7.44533777e-01 -7.71377087e-02 -6.84832931e-01
-3.12966198e-01 -4.59605306e-02 -1.92129418e-01 -1.81179300e-01
2.60440886e-01 -3.64967078e-01 2.10566707e-02 -8.11764598e-01
6.42795146e-01 5.29542327e-01 7.60436237e-01 -7.83992231e-01
9.52213764e-01 5.37951469e-01 2.23122165e-01 -5.39872587e-01
-9.29794908e-01 -5.55555761e-01 -1.29436886e+00 -2.24169895e-01
1.03233898e+00 -7.92237759e-01 -1.29657686e+00 7.62765884e-01
-1.42453134e+00 -4.40702111e-01 1.48529097e-01 4.80787575e-01
-6.17247939e-01 4.40738201e-01 -7.74396718e-01 -5.79278052e-01
-6.60175383e-02 -1.41376722e+00 1.67877841e+00 1.01111293e-01
-3.33111703e-01 -5.57409108e-01 8.64430442e-02 3.73326719e-01
1.36262868e-02 2.58398980e-01 7.25221038e-01 -2.22569004e-01
-5.83069384e-01 -5.05641937e-01 -2.26117969e-01 7.71692302e-03
1.13152467e-01 1.05567619e-01 -9.01173413e-01 -5.14433563e-01
-3.58805597e-01 -4.98621464e-01 7.71224260e-01 2.69832224e-01
1.17922497e+00 1.20863989e-01 -5.63540280e-01 3.68541837e-01
8.56926203e-01 -1.28065154e-01 3.13735515e-01 1.20146066e-01
1.09526134e+00 6.57926977e-01 6.37696326e-01 1.78258315e-01
4.08678204e-01 1.27909541e+00 5.33178329e-01 -1.71632886e-01
-6.79239854e-02 -2.45346025e-01 6.94961334e-03 5.45153260e-01
-6.57398701e-01 2.13802960e-02 -7.58811653e-01 2.53832042e-01
-2.13791132e+00 -3.42427284e-01 8.60488489e-02 2.13837004e+00
7.13330984e-01 2.39280403e-01 3.04615617e-01 -5.95831648e-02
6.72458231e-01 3.79089266e-02 -5.76629162e-01 2.03671202e-01
4.05908465e-01 3.53933781e-01 2.92762250e-01 5.96997440e-01
-1.37927985e+00 1.06775630e+00 6.54363871e+00 3.88891399e-01
-8.26352954e-01 -1.23685099e-01 1.25471801e-01 -5.73044717e-02
3.73852730e-01 -9.59497541e-02 -8.31053495e-01 1.02860890e-01
2.77109802e-01 6.66414976e-01 5.40045440e-01 8.67617130e-01
-2.53677636e-01 -1.45721108e-01 -1.42740715e+00 1.04723692e+00
1.18199006e-01 -6.76067650e-01 -8.76109377e-02 1.46991303e-02
2.83188790e-01 -2.63774097e-01 -9.67663005e-02 -1.00919297e-02
3.07966828e-01 -8.62972617e-01 8.55565786e-01 6.43636882e-01
5.79203904e-01 -5.65147877e-01 5.46485305e-01 2.18845993e-01
-1.39482796e+00 1.29013389e-01 -2.62950987e-01 -1.16297035e-02
1.45205468e-01 2.92084068e-01 -4.76453543e-01 2.18347058e-01
8.99320424e-01 4.72371072e-01 -6.42718494e-01 9.72438157e-01
-7.31026530e-01 -1.85622618e-01 -4.32858199e-01 1.53264076e-01
-4.99355756e-02 1.20460898e-01 2.58567750e-01 9.08266664e-01
1.36387259e-01 -1.93257451e-01 3.33319813e-01 8.87145519e-01
-9.70726237e-02 -4.27882493e-01 -3.35150063e-01 1.86518043e-01
2.81954169e-01 1.32598197e+00 -9.14373279e-01 -3.00646901e-01
-1.60168007e-01 1.63227081e+00 7.16572642e-01 4.56583440e-01
-9.00822341e-01 -6.50667608e-01 8.14471304e-01 -1.24845609e-01
4.00524020e-01 -7.42230177e-01 2.40716159e-01 -1.18206525e+00
4.82321888e-01 -5.64198911e-01 3.49275172e-02 -9.28037465e-01
-1.21621811e+00 4.25893188e-01 -3.52143049e-02 -8.95465195e-01
-1.95881173e-01 -1.08817673e+00 -1.91868663e-01 8.77749503e-01
-1.03130996e+00 -1.40043664e+00 -7.28571773e-01 6.90912426e-01
3.81347895e-01 3.56308609e-01 9.61114526e-01 1.67859241e-01
-4.77177590e-01 4.53946322e-01 -2.54256397e-01 5.25306344e-01
1.02489173e+00 -1.57690406e+00 7.92758286e-01 3.49146545e-01
2.90070653e-01 7.08667397e-01 3.75161558e-01 -4.74051565e-01
-1.63884532e+00 -5.32167196e-01 7.51897037e-01 -9.11228299e-01
4.40537930e-01 -8.79511297e-01 -6.71104372e-01 1.02935958e+00
-1.10726200e-01 2.91259497e-01 7.17138872e-02 3.10948998e-01
-5.47810256e-01 -7.80247450e-02 -7.75615454e-01 5.74698746e-01
1.21612847e+00 -6.67890549e-01 -6.24603510e-01 6.01350129e-01
4.93541837e-01 -9.40274298e-01 -8.07262897e-01 1.46782517e-01
1.01210582e+00 -7.33865023e-01 1.23767996e+00 -6.85281336e-01
3.39629352e-01 -4.06484485e-01 7.02495500e-02 -1.11773455e+00
-3.52579713e-01 -5.30936718e-01 -4.14801717e-01 8.96952569e-01
8.28876644e-02 -2.25649044e-01 9.13504004e-01 7.32694209e-01
2.98821718e-01 -8.21869552e-01 -9.65911746e-01 -6.81835532e-01
-2.58369386e-01 -2.00453386e-01 3.36647451e-01 4.53703284e-01
-1.34535626e-01 3.73457283e-01 -5.56636035e-01 2.75660932e-01
7.29433775e-01 2.23596424e-01 1.20154190e+00 -1.29612029e+00
-3.65101367e-01 -3.07321817e-01 -5.68180561e-01 -1.51519656e+00
3.24936897e-01 -3.64828885e-01 3.54495376e-01 -1.35162795e+00
3.72328162e-01 -2.07775608e-01 -5.91137856e-02 7.10648119e-01
-2.90344924e-01 3.52425337e-01 4.39178765e-01 2.17089877e-01
-5.50609112e-01 3.46834421e-01 1.37979579e+00 -2.61051506e-01
-1.35688663e-01 1.59856170e-01 -6.13108240e-02 8.28411579e-01
3.00354928e-01 -2.42082968e-01 1.19600505e-01 -5.66178501e-01
-1.39970765e-01 1.73664764e-01 8.11673164e-01 -1.06987762e+00
1.88506782e-01 6.38697669e-02 7.32733011e-01 -7.73432255e-01
7.03988075e-01 -1.06186652e+00 7.61778504e-02 5.12203693e-01
-4.29839969e-01 1.75884500e-01 3.72762792e-02 5.18853128e-01
7.14378953e-02 2.92424969e-02 5.56553245e-01 -2.97030896e-01
-6.02559865e-01 3.82284701e-01 -6.75610900e-02 -2.73874104e-01
1.02203214e+00 1.68038517e-01 3.90884802e-02 -3.51512641e-01
-1.05068231e+00 2.74807811e-01 3.74422491e-01 6.90294743e-01
5.74221611e-01 -1.45762360e+00 -3.13076258e-01 2.77446300e-01
-6.23637671e-03 2.19204620e-01 4.82650511e-02 7.36346543e-01
-5.77674925e-01 4.36283976e-01 -2.75009751e-01 -8.84536266e-01
-1.46594715e+00 5.34070253e-01 3.29520553e-01 -1.28458217e-01
-5.88352561e-01 9.89834487e-01 -2.29552779e-02 -6.68465316e-01
7.15394735e-01 -5.31665802e-01 1.46352634e-01 -1.75205041e-02
3.14493358e-01 3.58348101e-01 4.01825495e-02 -5.87659419e-01
-5.69464147e-01 1.16521740e+00 -1.47002459e-01 -1.57929838e-01
1.14692795e+00 8.75521749e-02 -2.48028502e-01 4.35847372e-01
1.41214573e+00 -2.27596954e-01 -1.70254636e+00 -1.64838582e-01
-2.52579927e-01 -6.83194578e-01 -3.30228508e-01 -7.09854543e-01
-1.19250476e+00 1.09689558e+00 6.96352065e-01 -2.49931142e-01
6.04449213e-01 4.30920780e-01 7.68638432e-01 7.32038498e-01
4.10036862e-01 -9.43454564e-01 4.64086831e-01 5.49882352e-01
1.11699069e+00 -1.29383731e+00 4.58833575e-02 -5.09935617e-01
-2.98882544e-01 1.24997079e+00 8.19383860e-01 -2.36636445e-01
5.59416533e-01 1.55581445e-01 1.93377167e-01 -3.42795134e-01
-2.25689366e-01 -1.73535511e-01 5.75572491e-01 5.04916787e-01
4.46215332e-01 7.30482414e-02 1.44455627e-01 3.01546067e-01
2.23111939e-02 -1.03874646e-01 -8.00156593e-02 1.23695982e+00
-1.02263063e-01 -1.15429974e+00 -4.48434263e-01 4.89810258e-02
-1.50993764e-01 3.71689111e-01 -7.30876625e-01 9.90500987e-01
-6.75964504e-02 5.49708188e-01 1.22023724e-01 -4.32980478e-01
7.61926532e-01 6.07245229e-02 9.97375965e-01 -4.63999510e-01
-1.66962385e-01 2.58059621e-01 -3.38102698e-01 -1.08999109e+00
-5.17843783e-01 -4.23531175e-01 -1.06651640e+00 -1.51656389e-01
-4.32703137e-01 -1.91305354e-01 6.71623111e-01 9.66162622e-01
2.51946598e-01 4.40243274e-01 2.36896664e-01 -1.64071536e+00
-7.02475548e-01 -9.01037097e-01 -6.73801005e-01 4.70966280e-01
3.68619084e-01 -1.16999328e+00 -5.97344674e-02 -1.42449737e-01]
|
[6.629334449768066, -0.833489716053009]
|
97165113-7b27-488e-829f-d3d94adcf7e3
|
learning-to-accelerate-partial-differential
|
2206.07681
| null |
https://arxiv.org/abs/2206.07681v2
|
https://arxiv.org/pdf/2206.07681v2.pdf
|
Learning to Accelerate Partial Differential Equations via Latent Global Evolution
|
Simulating the time evolution of Partial Differential Equations (PDEs) of large-scale systems is crucial in many scientific and engineering domains such as fluid dynamics, weather forecasting and their inverse optimization problems. However, both classical solvers and recent deep learning-based surrogate models are typically extremely computationally intensive, because of their local evolution: they need to update the state of each discretized cell at each time step during inference. Here we develop Latent Evolution of PDEs (LE-PDE), a simple, fast and scalable method to accelerate the simulation and inverse optimization of PDEs. LE-PDE learns a compact, global representation of the system and efficiently evolves it fully in the latent space with learned latent evolution models. LE-PDE achieves speed-up by having a much smaller latent dimension to update during long rollout as compared to updating in the input space. We introduce new learning objectives to effectively learn such latent dynamics to ensure long-term stability. We further introduce techniques for speeding-up inverse optimization of boundary conditions for PDEs via backpropagation through time in latent space, and an annealing technique to address the non-differentiability and sparse interaction of boundary conditions. We test our method in a 1D benchmark of nonlinear PDEs, 2D Navier-Stokes flows into turbulent phase and an inverse optimization of boundary conditions in 2D Navier-Stokes flow. Compared to state-of-the-art deep learning-based surrogate models and other strong baselines, we demonstrate up to 128x reduction in the dimensions to update, and up to 15x improvement in speed, while achieving competitive accuracy.
|
['Jure Leskovec', 'Takashi Maruyama', 'Tailin Wu']
|
2022-06-15
| null | null | null | null |
['weather-forecasting']
|
['miscellaneous']
|
[-2.10677743e-01 -4.74366128e-01 4.24883842e-01 1.78301811e-01
-6.73749804e-01 -5.46774268e-01 4.78431523e-01 -1.05315529e-01
-3.85757685e-01 1.04375458e+00 1.08466409e-01 -4.50144202e-01
-2.03217417e-01 -6.20154798e-01 -6.57082200e-01 -8.90585721e-01
-4.89125907e-01 8.46729636e-01 -2.05061048e-01 -1.17473386e-01
2.48942729e-02 7.59437442e-01 -1.04825294e+00 -2.18924806e-01
8.95640671e-01 1.02208972e+00 -3.85879189e-01 1.01514411e+00
2.06144691e-01 6.64417386e-01 -1.92166358e-01 4.18573201e-01
3.48812878e-01 -5.69436550e-01 -7.01759815e-01 -1.50755490e-03
2.92806864e-01 -5.88537335e-01 -4.70854372e-01 7.35503078e-01
7.19715297e-01 6.73804402e-01 8.95823836e-01 -1.01941013e+00
-3.80316854e-01 -7.54014105e-02 -3.87892187e-01 3.35061729e-01
-6.11400232e-02 7.40226388e-01 7.08736539e-01 -9.21228766e-01
6.63970828e-01 1.24731183e+00 1.28981984e+00 5.48030615e-01
-1.51506138e+00 -5.07831514e-01 -2.00319495e-02 -2.64672995e-01
-1.17279184e+00 -2.94230312e-01 5.61717570e-01 -6.84069514e-01
1.28243279e+00 2.46805191e-01 9.95566547e-01 9.23927903e-01
6.52142167e-01 4.34336245e-01 1.06619990e+00 2.06142381e-01
4.93008912e-01 -1.84882671e-01 7.40995118e-03 8.66217971e-01
8.60170126e-02 4.46976751e-01 -2.81957328e-01 -7.18229890e-01
1.25880075e+00 -2.78914034e-01 -4.41132128e-01 -2.53679454e-01
-9.49240863e-01 1.06286049e+00 2.03968123e-01 -5.55698350e-02
-4.85109419e-01 3.99918616e-01 3.43400419e-01 3.03512931e-01
9.15950179e-01 9.14132953e-01 -6.66368604e-01 -4.12815005e-01
-1.15017402e+00 7.78815091e-01 1.34143603e+00 4.67451394e-01
5.09705842e-01 5.11779308e-01 -8.87699798e-02 4.10693944e-01
3.22349936e-01 6.29370570e-01 4.27313179e-01 -1.60317671e+00
1.51231084e-02 2.59282112e-01 3.98846209e-01 -8.63118172e-01
-4.13165659e-01 -7.66514182e-01 -1.41073418e+00 5.18371463e-01
1.71761975e-01 -7.69845128e-01 -9.77062404e-01 1.65429688e+00
7.09140599e-01 8.43542099e-01 1.28241450e-01 1.01420879e+00
3.32581729e-01 1.28461480e+00 -2.65435904e-01 -5.20379841e-01
8.24366212e-01 -1.26545238e+00 -5.84698975e-01 -1.95310831e-01
9.38150406e-01 -5.89777946e-01 7.00967431e-01 5.59197292e-02
-1.37179255e+00 -3.84169042e-01 -8.69643986e-01 -1.11146390e-01
-5.39670475e-02 -3.11486870e-01 5.98461568e-01 -1.00645106e-02
-1.28753030e+00 1.53107142e+00 -1.41311204e+00 2.76591897e-01
2.58458942e-01 3.81636471e-01 -1.40360445e-02 3.26021254e-01
-1.15959859e+00 7.74016261e-01 -3.85245711e-01 2.25644499e-01
-1.15695310e+00 -1.68247569e+00 -9.58537281e-01 1.20535731e-01
-1.47812650e-01 -1.20238912e+00 1.18609738e+00 -5.30182958e-01
-1.96814251e+00 5.88332355e-01 -2.08503440e-01 -5.59133649e-01
9.67047870e-01 -2.82164395e-01 -8.99891369e-03 -1.82013914e-01
1.17002174e-01 1.63339108e-01 8.36346149e-01 -9.42671418e-01
-2.84886062e-02 1.22774385e-01 -3.47013652e-01 3.25130999e-01
-1.61530584e-01 -4.60859209e-01 -2.05311567e-01 -7.39299774e-01
5.01585305e-02 -1.27096391e+00 -7.25654304e-01 4.00444210e-01
9.61382315e-02 1.98324263e-01 1.06416810e+00 -7.95156658e-01
1.07377601e+00 -1.76370811e+00 5.96042395e-01 -9.38365310e-02
3.70456070e-01 3.84765953e-01 1.03256226e-01 3.27696562e-01
-9.44810957e-02 -1.26779407e-01 -8.71186435e-01 -8.33909214e-01
1.67255044e-01 4.40321922e-01 -7.79223859e-01 8.11755061e-01
1.25077754e-01 1.08281541e+00 -9.13487434e-01 -1.92670614e-01
4.21049371e-02 8.83696079e-01 -8.65948081e-01 3.52997750e-01
-1.46339983e-01 9.91588295e-01 -3.49339008e-01 7.14016780e-02
7.26234555e-01 -6.93130016e-01 -1.74584389e-01 -4.21260633e-02
-3.05990309e-01 2.04815984e-01 -1.34179044e+00 1.57492673e+00
-9.15470898e-01 5.77360153e-01 6.97163820e-01 -1.13084793e+00
3.96170795e-01 4.61369336e-01 6.65548325e-01 -4.72625107e-01
6.50522113e-02 4.41646278e-01 -2.48959109e-01 -2.88684547e-01
3.48309278e-01 -5.09263396e-01 4.06517796e-02 7.61251688e-01
-2.62804300e-01 -7.39317179e-01 1.35002837e-01 -1.25295194e-02
1.18221188e+00 1.74787477e-01 -1.30766600e-01 -6.46881044e-01
5.72890699e-01 1.49707243e-01 5.53764701e-01 5.74215412e-01
1.55015171e-01 3.58832926e-01 3.57993066e-01 -6.95735872e-01
-1.16484547e+00 -7.22247124e-01 -3.51730406e-01 5.31562686e-01
-5.64460680e-02 -7.81331956e-02 -5.67367792e-01 8.41138512e-02
4.27109569e-01 5.33812821e-01 -6.18655145e-01 -3.11082929e-01
-1.22059751e+00 -1.12680197e+00 4.86366659e-01 5.66820264e-01
5.85161090e-01 -8.30680668e-01 -6.34230316e-01 6.03861332e-01
3.59743685e-01 -9.37734425e-01 -4.76025283e-01 1.36757538e-01
-1.31260490e+00 -6.39903367e-01 -9.97877419e-01 -5.70723176e-01
5.92956364e-01 -5.43070436e-01 1.23241448e+00 2.53376774e-02
-4.88146693e-01 2.99187720e-01 5.07087290e-01 3.70912731e-01
-6.05628669e-01 4.65687215e-02 3.13424200e-01 -2.35394910e-01
-6.31720722e-01 -9.96783614e-01 -6.39589906e-01 8.95004645e-02
-6.89755261e-01 8.32219273e-02 1.03898026e-01 1.06687963e+00
6.48691297e-01 6.36086091e-02 5.03659733e-02 -7.37722635e-01
8.27194393e-01 -5.17938137e-01 -9.69103932e-01 -2.97975302e-01
-5.98408580e-01 3.65426153e-01 9.27672207e-01 -7.04322636e-01
-9.73637104e-01 -1.62421614e-01 -3.69659007e-01 -8.39045167e-01
3.71609628e-01 6.34922802e-01 5.23417354e-01 -4.94626343e-01
5.12506604e-01 2.29789734e-01 -6.50182087e-03 -6.38650298e-01
-4.85025644e-02 -2.35721376e-02 3.99547815e-01 -9.78518844e-01
8.14357758e-01 5.95479012e-01 5.42053580e-01 -7.74118066e-01
-7.91328907e-01 -3.35724503e-01 -3.72466236e-01 1.19124196e-01
6.12260222e-01 -7.93881834e-01 -8.57380867e-01 9.19506669e-01
-1.12779415e+00 -1.10789013e+00 -6.86881065e-01 3.88969719e-01
-6.09781206e-01 2.21695393e-01 -1.28394711e+00 -5.81989288e-01
-5.01789570e-01 -1.03213453e+00 1.19077682e+00 -6.60665631e-02
-3.26981753e-01 -1.73042488e+00 5.46902955e-01 -2.91678816e-01
9.74852204e-01 6.56106472e-01 7.33391404e-01 2.84380972e-01
-4.09227282e-01 -6.12570047e-02 1.61180156e-03 2.11171299e-01
2.36847345e-02 -3.50965746e-02 -5.27300358e-01 -6.89488828e-01
7.11974323e-01 -8.44338313e-02 9.38971698e-01 6.36751950e-01
8.37967217e-01 -5.95819294e-01 -5.91075242e-01 1.40766108e+00
1.35639191e+00 -1.97890848e-01 -1.91366822e-02 -2.96179950e-01
7.98077345e-01 1.47967339e-01 7.72904307e-02 6.63828552e-01
-3.10443081e-02 3.77255946e-01 6.93859998e-03 -4.11248982e-01
1.87450759e-02 3.10318798e-01 3.48212898e-01 1.19423163e+00
3.20107378e-02 -1.26981661e-01 -1.09717560e+00 4.78336006e-01
-1.74452627e+00 -4.57656085e-01 -2.77880490e-01 1.80022275e+00
1.15478098e+00 7.48152360e-02 -4.15120602e-01 -7.47385994e-02
1.74662992e-01 2.02120468e-01 -1.25544107e+00 -4.87892240e-01
9.75139663e-02 3.92092377e-01 5.82296669e-01 1.18350685e+00
-1.12661242e+00 7.37089992e-01 6.85188389e+00 5.14914870e-01
-1.60056651e+00 3.38752329e-01 7.24883556e-01 -2.92548835e-01
-1.02066234e-01 -1.85570866e-01 -8.71200562e-01 2.69044042e-01
1.23831654e+00 -3.66002083e-01 5.75126290e-01 6.50894165e-01
3.56284082e-01 2.57152438e-01 -9.69470024e-01 7.85215974e-01
-3.50621670e-01 -1.77730107e+00 -1.96268767e-01 8.72988701e-02
1.41567552e+00 2.92337835e-01 2.18285412e-01 2.73936152e-01
4.56582546e-01 -1.16670835e+00 3.44181776e-01 4.63525593e-01
7.93663859e-01 -4.46738899e-01 4.38255847e-01 4.77729082e-01
-1.20166564e+00 2.03631595e-01 -1.24437392e-01 -5.33974349e-01
6.80805385e-01 6.74102187e-01 -1.15993522e-01 -2.06925347e-02
6.01162910e-01 1.04805243e+00 1.97340369e-01 6.71941161e-01
1.03880748e-01 7.77686179e-01 -7.63135552e-01 3.44296306e-01
6.49127364e-01 -6.08691990e-01 9.38226104e-01 1.02344680e+00
5.16093612e-01 3.99062335e-01 2.27471322e-01 1.24407935e+00
5.92915043e-02 -3.29576343e-01 -2.56974071e-01 -1.27648547e-01
-9.24515501e-02 1.05723023e+00 -7.02065885e-01 -6.00714087e-01
1.35121614e-01 9.94696498e-01 -4.04532859e-03 7.17058599e-01
-1.15852737e+00 4.49091382e-02 1.05750525e+00 2.00511217e-01
4.25443649e-01 -7.32097507e-01 -2.55071819e-01 -1.19191074e+00
-1.37441590e-01 -5.22617459e-01 2.39126161e-01 -4.27417576e-01
-1.33159065e+00 6.06957793e-01 -1.09683394e-01 -1.00364816e+00
-3.44091475e-01 -6.71668172e-01 -6.19500935e-01 1.19321537e+00
-1.57487333e+00 -6.74010396e-01 -1.95754990e-01 3.26029688e-01
2.74686635e-01 2.56636292e-01 9.05745327e-01 4.11644459e-01
-5.15946865e-01 2.28198841e-01 7.38734245e-01 -5.15099987e-02
3.10055405e-01 -1.16679454e+00 8.50259423e-01 6.73330963e-01
-5.60891628e-01 5.74498057e-01 9.69081879e-01 -7.63476670e-01
-1.66401005e+00 -1.11199784e+00 5.06908774e-01 -2.63750374e-01
8.43120873e-01 -1.41688034e-01 -1.31174290e+00 6.35502815e-01
-5.88336447e-03 6.16590381e-01 2.16998979e-01 -3.67449611e-01
2.27048635e-01 2.07843155e-01 -9.11835015e-01 3.93371642e-01
9.51195598e-01 -4.57064331e-01 -2.05800861e-01 5.22265315e-01
5.14824212e-01 -1.14312744e+00 -9.98085201e-01 5.28965592e-01
3.46082449e-01 -5.25336385e-01 1.23482132e+00 -6.27348125e-01
5.89323997e-01 -1.95068210e-01 4.65535313e-01 -1.33563566e+00
-3.87201488e-01 -1.30478764e+00 -7.02648580e-01 6.24014795e-01
1.30725861e-01 -9.15294051e-01 8.05622458e-01 7.68152714e-01
-4.07665312e-01 -1.27017093e+00 -1.09783792e+00 -7.46010900e-01
5.51240087e-01 -1.73920095e-01 3.12601954e-01 1.21257567e+00
-6.39833510e-01 1.33773535e-01 -4.00677860e-01 3.78281862e-01
7.22632289e-01 4.67164904e-01 4.90740955e-01 -9.80674148e-01
-6.32579803e-01 -5.93398571e-01 2.12276012e-01 -1.39188302e+00
3.55292618e-01 -6.87215984e-01 1.19849525e-01 -1.43441999e+00
-4.51483637e-01 -6.25541747e-01 -7.14674741e-02 5.74205723e-03
-1.10113785e-01 4.09751505e-01 -3.73288244e-01 2.54635781e-01
-1.87373623e-01 8.42894077e-01 1.46209562e+00 1.40628424e-02
-4.98380899e-01 -1.51041001e-01 6.44578710e-02 7.49376237e-01
4.76821333e-01 -6.72364235e-01 -1.85401976e-01 -6.44264698e-01
2.04288572e-01 4.13239330e-01 4.17540729e-01 -1.14950585e+00
5.11815250e-01 -1.02659583e-01 2.92112827e-01 -2.94485807e-01
7.37000883e-01 -3.82505804e-01 3.54344219e-01 8.85013103e-01
-1.78093761e-01 1.85017690e-01 7.58326411e-01 2.68932581e-01
-1.76214069e-01 1.53477550e-01 1.07060027e+00 -3.38752866e-01
-3.42626989e-01 7.21681774e-01 -7.30945230e-01 5.66573441e-01
7.00378060e-01 -5.51741105e-03 1.39126748e-01 -3.27200770e-01
-6.64843738e-01 3.50708604e-01 5.61338782e-01 -1.38254836e-01
2.85200804e-01 -1.22862649e+00 -7.95949340e-01 5.98555386e-01
-8.38998675e-01 3.89451444e-01 4.47579563e-01 9.98455405e-01
-1.12379062e+00 1.05953000e-01 -1.68415040e-01 -5.97363353e-01
-8.38546276e-01 1.26746505e-01 8.38998616e-01 -8.17731977e-01
-8.63594174e-01 1.33073390e+00 2.87029028e-01 -5.83837032e-01
-1.15033828e-01 -5.36291480e-01 2.88900137e-01 -8.83050188e-02
1.09976433e-01 7.19261646e-01 -8.48637596e-02 -5.54160595e-01
-2.22753286e-01 9.69251335e-01 4.36201036e-01 -2.16976792e-01
1.61308169e+00 2.60520458e-01 -3.85673076e-01 3.64395916e-01
1.63451922e+00 -2.09415048e-01 -1.88240469e+00 -1.05022833e-01
-4.66864526e-01 1.42123383e-02 4.78251308e-01 -3.65360320e-01
-1.18379962e+00 9.97303367e-01 2.99263090e-01 1.13711454e-01
7.93431878e-01 -5.09158373e-01 1.46163130e+00 3.67339164e-01
-2.05367088e-01 -8.81792426e-01 -1.51278421e-01 1.02060258e+00
8.23793411e-01 -9.09523368e-01 3.94990385e-01 -1.52771082e-02
-7.75423199e-02 1.13622499e+00 3.69155407e-01 -6.09409273e-01
1.08401966e+00 8.11125755e-01 1.03996195e-01 -1.90445125e-01
-9.43098187e-01 6.03346646e-01 3.64171654e-01 -2.70607084e-01
1.96865231e-01 -3.70096147e-01 1.47198558e-01 8.92865136e-02
-1.25982329e-01 1.82777047e-02 1.77917659e-01 1.01948190e+00
1.27004623e-01 -8.83800030e-01 -1.78486824e-01 2.43629888e-01
-2.04544708e-01 -2.10875478e-02 1.88226491e-01 6.14319921e-01
-2.16324478e-01 1.80754393e-01 2.86395371e-01 3.00965309e-01
3.30592357e-02 7.30822757e-02 2.82204777e-01 -4.13960546e-01
-6.37586713e-01 2.43488904e-02 -2.37434953e-01 -6.83493912e-01
-1.12237886e-01 -7.34560192e-01 -1.42135370e+00 -5.67684710e-01
-3.50983217e-02 5.33589244e-01 3.21551085e-01 9.97723758e-01
5.99040151e-01 7.97590315e-01 3.26360136e-01 -1.53578877e+00
-7.08008587e-01 -7.10526407e-01 -3.60695839e-01 2.86583573e-01
7.88562417e-01 -6.48545563e-01 -9.21949565e-01 1.10973045e-02]
|
[6.529872894287109, 3.4285295009613037]
|
c79a1607-4d58-4c45-a38c-feace353db5a
|
aiir-mix-multi-agent-reinforcement-learning
|
2302.09531
| null |
https://arxiv.org/abs/2302.09531v1
|
https://arxiv.org/pdf/2302.09531v1.pdf
|
AIIR-MIX: Multi-Agent Reinforcement Learning Meets Attention Individual Intrinsic Reward Mixing Network
|
Deducing the contribution of each agent and assigning the corresponding reward to them is a crucial problem in cooperative Multi-Agent Reinforcement Learning (MARL). Previous studies try to resolve the issue through designing an intrinsic reward function, but the intrinsic reward is simply combined with the environment reward by summation in these studies, which makes the performance of their MARL framework unsatisfactory. We propose a novel method named Attention Individual Intrinsic Reward Mixing Network (AIIR-MIX) in MARL, and the contributions of AIIR-MIX are listed as follows:(a) we construct a novel intrinsic reward network based on the attention mechanism to make teamwork more effective. (b) we propose a Mixing network that is able to combine intrinsic and extrinsic rewards non-linearly and dynamically in response to changing conditions of the environment. We compare AIIR-MIX with many State-Of-The-Art (SOTA) MARL methods on battle games in StarCraft II. And the results demonstrate that AIIR-MIX performs admirably and can defeat the current advanced methods on average test win rate. To validate the effectiveness of AIIR-MIX, we conduct additional ablation studies. The results show that AIIR-MIX can dynamically assign each agent a real-time intrinsic reward in accordance with their actual contribution.
|
['Shiyi Huang', 'Shitong Shao', 'Weiyan Liu', 'Wei Li']
|
2023-02-19
| null | null | null | null |
['starcraft-ii', 'starcraft']
|
['playing-games', 'playing-games']
|
[-3.15712154e-01 3.67646031e-02 -1.13745943e-01 1.91797227e-01
-3.66671830e-01 -3.11042279e-01 4.74019766e-01 2.37864424e-02
-9.64299321e-01 1.07347333e+00 6.62656575e-02 7.66427591e-02
-4.68974769e-01 -7.03986704e-01 -4.68601704e-01 -8.49448562e-01
-3.03546578e-01 6.10877216e-01 5.30581355e-01 -1.19587111e+00
4.11570072e-01 1.77713871e-01 -1.49824345e+00 -3.24327387e-02
1.16861033e+00 5.89853048e-01 4.34168816e-01 6.67707384e-01
2.42015719e-01 1.33356869e+00 -1.07727861e+00 -1.19380094e-01
4.59578395e-01 -7.55818963e-01 -7.27666259e-01 -4.43428993e-01
-5.50687850e-01 -4.99296010e-01 -2.54438788e-01 7.77136803e-01
8.87453914e-01 2.30678484e-01 4.28387195e-01 -1.64116096e+00
-4.90505874e-01 1.31785536e+00 -8.23173523e-01 2.38300413e-01
2.05964237e-01 2.72765905e-01 1.05797505e+00 -2.10435554e-01
2.75176525e-01 1.27217650e+00 3.35479021e-01 8.59869659e-01
-9.91833866e-01 -7.08982766e-01 5.18276513e-01 3.92637581e-01
-9.75462914e-01 -1.41710020e-03 8.80514860e-01 -8.50510523e-02
7.12497532e-01 2.39141047e-01 8.36875916e-01 9.53941882e-01
3.22032779e-01 1.04028618e+00 1.18500233e+00 -3.60009938e-01
6.07599378e-01 -2.66162425e-01 -2.18822300e-01 4.15405899e-01
2.18060076e-01 5.90362549e-01 -5.94884574e-01 -1.85088485e-01
8.83043945e-01 -2.55684942e-01 8.93577039e-02 -3.50624442e-01
-1.18990433e+00 8.37725759e-01 6.39774799e-01 3.65485013e-01
-6.91247046e-01 4.71371353e-01 3.45681578e-01 6.99061871e-01
1.15506634e-01 9.66445565e-01 -2.16574013e-01 -2.97999024e-01
-3.33352417e-01 6.28045917e-01 4.70596939e-01 2.86970079e-01
5.28342903e-01 3.08897704e-01 -4.70044315e-01 7.68065035e-01
2.02216089e-01 3.48044515e-01 7.27624118e-01 -1.19084060e+00
2.00696811e-01 4.86932278e-01 3.36656839e-01 -8.49210620e-01
-6.22956455e-01 -5.15084863e-01 -5.04381955e-01 9.07289088e-01
3.27321172e-01 -6.39153421e-01 -2.65255749e-01 1.97910547e+00
2.02427030e-01 1.13255538e-01 3.80742311e-01 1.18522024e+00
6.59160376e-01 4.34546560e-01 -1.05409222e-02 -2.50803649e-01
8.79536986e-01 -1.22954977e+00 -6.28339946e-01 -2.93109924e-01
5.21001101e-01 -1.93151042e-01 1.01203144e+00 4.53439891e-01
-1.05424166e+00 -3.42654228e-01 -1.11927140e+00 8.33858013e-01
1.98292270e-01 -2.01201841e-01 8.27985764e-01 5.27754188e-01
-9.79729354e-01 6.33476913e-01 -5.50165534e-01 1.22127779e-01
7.72246867e-02 5.89255512e-01 -2.70070210e-02 3.23955148e-01
-1.42108190e+00 1.03705788e+00 4.15978253e-01 -3.28886230e-03
-1.27428401e+00 -1.93709880e-01 -5.04656732e-01 1.70627937e-01
8.56977284e-01 -3.87752801e-01 1.40497911e+00 -1.25563836e+00
-1.99921799e+00 1.53185025e-01 6.54189825e-01 -4.75637048e-01
4.76400256e-01 -1.29814520e-01 7.26893991e-02 4.29550037e-02
1.34004921e-01 7.62361288e-01 7.41066873e-01 -1.53864396e+00
-7.91260660e-01 5.75693743e-03 6.99233055e-01 7.09968865e-01
-3.44191521e-01 -8.05614442e-02 1.26706123e-01 -6.87359512e-01
-5.70204198e-01 -8.43421519e-01 -6.99405372e-01 -6.26155317e-01
1.58079743e-01 -4.38959569e-01 4.52289969e-01 -9.10658538e-02
1.19796515e+00 -1.86207545e+00 5.44551611e-01 -1.18349701e-01
4.59987730e-01 4.49139237e-01 -7.72842407e-01 5.43589711e-01
1.78818166e-01 -2.46578917e-01 5.49734347e-02 5.28511107e-02
8.26468319e-02 3.90567869e-01 -1.46064656e-02 2.13051841e-01
1.36714242e-02 8.53482425e-01 -1.46396804e+00 -3.53473932e-01
-6.97348192e-02 -3.66142578e-02 -8.72937143e-01 5.08719146e-01
-2.99229175e-01 4.25147653e-01 -6.54733717e-01 5.02200663e-01
4.51228619e-01 2.43591860e-01 2.94475317e-01 3.97474706e-01
-1.24597810e-01 1.04277078e-02 -1.16214561e+00 1.35709107e+00
-1.89734071e-01 1.20082460e-01 1.90998048e-01 -1.03323686e+00
1.11031079e+00 2.76312768e-01 8.37704122e-01 -9.47530568e-01
4.09038275e-01 1.59035757e-01 5.97083986e-01 -2.39558220e-01
7.31641412e-01 -2.15496980e-02 -3.47260207e-01 8.63324344e-01
-6.08208170e-03 -1.08441599e-02 3.87872338e-01 3.66259187e-01
1.29932880e+00 3.26064080e-01 2.15107456e-01 -1.31137535e-01
4.60079819e-01 -1.41811371e-02 9.55556750e-01 1.21393526e+00
-6.06362402e-01 -3.22685167e-02 8.32818687e-01 -5.05476058e-01
-8.06768179e-01 -7.51831055e-01 5.47188699e-01 1.47802794e+00
5.40978968e-01 -2.72414535e-01 -6.25168324e-01 -1.05245209e+00
-1.32739162e-02 5.42291880e-01 -8.02758276e-01 -4.26606268e-01
-5.62379301e-01 -9.06963348e-01 4.38101918e-01 3.38391125e-01
5.86243749e-01 -1.71670306e+00 -1.06650949e+00 4.96732533e-01
-3.01258236e-01 -5.97463071e-01 -4.09829080e-01 3.38834763e-01
-3.43264967e-01 -9.56762612e-01 -6.44491553e-01 -3.87074083e-01
4.37010735e-01 2.98724771e-01 9.75021601e-01 2.98466802e-01
-2.86332052e-02 4.09992427e-01 -8.50319684e-01 -4.20542300e-01
-4.42721695e-01 8.81992429e-02 2.95894295e-01 -2.13290781e-01
1.56335786e-01 -6.11398101e-01 -4.12158132e-01 5.35675108e-01
-7.93899119e-01 -1.29462853e-01 7.82323360e-01 1.12552881e+00
-2.35974882e-02 5.70565388e-02 1.13701034e+00 -4.40813899e-01
1.45891368e+00 -4.88532364e-01 -7.00273156e-01 1.26105249e-01
-6.08318150e-01 3.66755545e-01 6.99645877e-01 -7.97152102e-01
-8.45299423e-01 -2.92520821e-01 -4.09483425e-02 -2.79528916e-01
3.25278491e-01 4.21844661e-01 1.72710001e-01 -1.78217694e-01
7.82891870e-01 3.08455117e-02 3.20310652e-01 -9.21595376e-03
1.87115997e-01 5.85233331e-01 2.09100559e-01 -9.90205228e-01
5.91076314e-01 -8.64175856e-02 4.90121245e-02 -8.40828791e-02
-7.04174161e-01 -1.24868397e-02 -7.33336359e-02 -6.30478263e-01
3.81196469e-01 -7.29919910e-01 -1.41706347e+00 5.74153423e-01
-8.44586551e-01 -7.93694854e-01 -6.28558576e-01 5.56670845e-01
-7.93464839e-01 2.38171563e-01 -6.82201684e-01 -9.67367649e-01
-2.36776114e-01 -1.31419110e+00 3.56388509e-01 4.38749075e-01
4.94460687e-02 -6.17338717e-01 4.12556946e-01 1.57276064e-01
6.56303465e-01 4.09677215e-02 3.88527095e-01 -4.87972826e-01
-2.98218250e-01 2.06722900e-01 1.10533722e-01 1.97214320e-01
3.38417590e-02 -3.17658395e-01 -3.98136914e-01 -4.46466118e-01
-2.65501618e-01 -8.25643182e-01 6.31856740e-01 3.68438691e-01
6.18760884e-01 -2.83543348e-01 1.01229340e-01 6.88544940e-03
1.18982124e+00 4.70081359e-01 6.63675964e-01 8.53679657e-01
2.55753070e-01 4.95672047e-01 1.08839345e+00 9.15915728e-01
4.49579805e-01 8.10072660e-01 1.16808701e+00 1.28271673e-02
1.42200902e-01 -2.06090249e-02 8.26476574e-01 6.29179657e-01
-5.43781400e-01 -4.34121564e-02 -4.02954519e-01 4.04354990e-01
-2.50732923e+00 -1.35337603e+00 2.13338047e-01 2.10525703e+00
8.78955960e-01 2.71144897e-01 6.07751071e-01 1.62049327e-02
5.98598123e-01 9.12988856e-02 -7.16799974e-01 -6.21464729e-01
-1.04766555e-01 -2.64485031e-02 4.19076174e-01 5.71583927e-01
-7.27748930e-01 1.15422928e+00 6.83236647e+00 9.90080416e-01
-8.35102856e-01 1.36363924e-01 3.58235478e-01 -2.49645784e-01
-1.98988244e-01 1.30491003e-01 -5.10142922e-01 3.27056885e-01
4.68573302e-01 -3.87244850e-01 9.72941935e-01 8.67290616e-01
1.87738732e-01 -3.13437015e-01 -5.76487422e-01 6.23859644e-01
-2.88194418e-02 -1.03586805e+00 -1.75856054e-01 -1.81010459e-02
5.35033762e-01 -8.77347868e-03 -1.27690569e-01 9.04538691e-01
1.16515338e+00 -9.29441392e-01 7.57040262e-01 2.96800524e-01
2.51846373e-01 -1.04996550e+00 9.53533232e-01 5.98918200e-01
-9.78332281e-01 -5.52188814e-01 -3.84288341e-01 -6.15215302e-01
-1.14429735e-01 -5.42077329e-03 -5.44643641e-01 7.03890562e-01
6.69013917e-01 4.64693487e-01 -3.68378818e-01 9.43117917e-01
-5.41642606e-01 3.19113910e-01 -1.00382872e-01 -4.62814361e-01
5.30023992e-01 -2.61202693e-01 7.49209642e-01 4.70636278e-01
1.52254589e-02 8.90301093e-02 3.89934093e-01 7.37416267e-01
7.42340386e-02 2.92210896e-02 -2.51087576e-01 4.12991494e-02
4.82739210e-01 1.39673769e+00 -4.66883272e-01 -6.14440925e-02
1.53118774e-01 6.67301536e-01 7.11215556e-01 1.30586430e-01
-9.56289470e-01 -3.51546705e-01 6.91929162e-01 -1.71247795e-01
1.47017568e-01 2.93837860e-02 1.17757648e-01 -9.30796444e-01
-2.71791011e-01 -1.07706928e+00 2.31788293e-01 -7.17188358e-01
-1.26727474e+00 9.10508990e-01 -9.34322998e-02 -1.35327184e+00
-3.10963988e-01 -2.13364899e-01 -6.77170932e-01 4.72160786e-01
-1.55910540e+00 -7.68621802e-01 -2.44723961e-01 5.85527241e-01
3.92174512e-01 -7.00708449e-01 6.52243972e-01 1.01094536e-01
-5.83006978e-01 6.66888177e-01 -1.12281874e-01 -8.88674632e-02
6.82106674e-01 -1.22382796e+00 -3.72360664e-05 5.62786877e-01
-3.67346585e-01 4.59043495e-02 7.68392265e-01 -5.30594468e-01
-1.13190043e+00 -5.75368881e-01 1.88903198e-01 -5.54972850e-02
6.32504880e-01 -8.19838718e-02 -4.37396199e-01 2.38284424e-01
5.01951277e-01 -1.11607857e-01 4.62488443e-01 2.11410061e-01
4.80467379e-02 -2.41048694e-01 -1.10894704e+00 9.17240262e-01
7.91289330e-01 3.02401811e-01 -5.80262482e-01 -1.67802021e-01
8.35346341e-01 -2.53680438e-01 -6.17810786e-01 3.81466329e-01
3.69711429e-01 -1.21680689e+00 7.40891695e-01 -5.97843647e-01
5.53997219e-01 -5.20165026e-01 1.68078855e-01 -1.94023848e+00
-6.87919199e-01 -8.68389904e-01 4.95554805e-02 8.26707184e-01
1.20447196e-01 -7.02425778e-01 7.41580427e-01 1.70963973e-01
-9.95557979e-02 -8.33821893e-01 -1.00410700e+00 -9.24050629e-01
1.36783361e-01 8.03010091e-02 6.81775630e-01 8.86086524e-01
5.44071317e-01 4.49308544e-01 -9.03387904e-01 -1.82831347e-01
5.79363883e-01 -6.58857748e-02 9.60640252e-01 -1.08946991e+00
-7.71763623e-01 -7.99726546e-01 -7.47450739e-02 -8.34880471e-01
6.13607578e-02 -5.33368170e-01 1.59655303e-01 -1.54740655e+00
2.05611289e-01 -1.09368765e+00 -6.44090593e-01 8.35405648e-01
-4.38496649e-01 -8.36867094e-02 5.96671581e-01 2.01045230e-01
-1.03244793e+00 9.93366241e-01 1.56633246e+00 -1.35089666e-01
-4.90777403e-01 1.57749560e-02 -1.00711739e+00 3.81485611e-01
1.02356911e+00 -6.35608792e-01 -3.90544325e-01 -1.41658992e-01
4.10645455e-01 4.22620177e-01 1.41306752e-02 -9.89683807e-01
1.65207013e-01 -7.45948017e-01 -1.72638074e-01 -1.16419330e-01
2.30301619e-01 -6.18560553e-01 -2.60280877e-01 7.95466185e-01
-4.51140195e-01 1.70547664e-01 4.61295918e-02 3.40748250e-01
-1.48921326e-01 -5.18289983e-01 4.60545421e-01 -1.72850579e-01
-5.78459620e-01 1.08204193e-01 -6.96494460e-01 5.36620803e-02
1.30963957e+00 2.98982821e-02 -5.23039460e-01 -6.06102645e-01
-3.33894283e-01 7.15684235e-01 1.57039031e-01 5.09934068e-01
6.46812379e-01 -1.29901469e+00 -1.07324636e+00 -1.09184831e-01
2.52783466e-02 -3.27078491e-01 2.26659238e-01 7.27143943e-01
-2.78092206e-01 -1.07708707e-01 -8.80218446e-01 -1.08797580e-01
-1.02837873e+00 7.14593232e-01 5.50990582e-01 -8.09096634e-01
-4.08436090e-01 5.74191630e-01 7.54990205e-02 -5.62457681e-01
3.05050492e-01 2.53195584e-01 -6.23145640e-01 -2.39139870e-01
5.00970721e-01 1.40830830e-01 -4.53971237e-01 -2.86616415e-01
-2.26451010e-01 2.45764673e-01 -1.82178214e-01 -3.03670973e-01
1.46911156e+00 1.63415834e-01 3.92170399e-02 4.26193625e-02
-2.65287445e-03 -3.74520235e-02 -1.58884370e+00 -1.13858700e-01
-3.77054989e-01 -4.39155370e-01 2.46200673e-02 -8.87391090e-01
-1.26213050e+00 3.89551699e-01 3.78235847e-01 1.52968839e-01
1.26713371e+00 -2.62692690e-01 4.33358580e-01 3.77859294e-01
7.05101788e-01 -1.38330245e+00 7.56408036e-01 7.00280011e-01
8.77515256e-01 -1.11123872e+00 -3.79796177e-02 1.29755765e-01
-1.21403801e+00 8.36316645e-01 1.11362231e+00 -4.98428464e-01
2.00032011e-01 3.94028693e-01 6.61930665e-02 -5.07398359e-02
-1.08558404e+00 -4.47136551e-01 -4.14555907e-01 6.93286061e-01
6.96185082e-02 7.29303136e-02 -7.43709147e-01 7.32738912e-01
6.89353198e-02 -1.12356097e-01 9.05344844e-01 1.10547590e+00
-6.67885900e-01 -1.55462968e+00 -5.50493956e-01 1.54801995e-01
-2.24616244e-01 2.08504722e-01 -3.41080785e-01 5.39917529e-01
2.98230480e-02 1.16408551e+00 -1.73108280e-01 -6.23755395e-01
2.67982721e-01 -7.66872942e-01 4.94587958e-01 -1.98549598e-01
-1.22855079e+00 1.88903604e-02 -1.78141017e-02 -4.83713508e-01
-5.13506472e-01 -3.41274381e-01 -1.40666080e+00 -3.57971638e-01
-3.03959757e-01 6.57489777e-01 3.59458655e-01 8.28601897e-01
2.38068432e-01 1.10319269e+00 9.90987301e-01 -9.57451999e-01
-8.72498155e-01 -9.21136081e-01 -6.42328382e-01 3.25584024e-01
1.13394864e-01 -1.15071690e+00 -2.80334741e-01 -7.98113406e-01]
|
[3.737269163131714, 1.9153109788894653]
|
eef1cb44-e92b-4538-b41e-0ecf73371be9
|
analyzing-an-imitation-learning-network-for
|
1912.10837
| null |
https://arxiv.org/abs/1912.10837v1
|
https://arxiv.org/pdf/1912.10837v1.pdf
|
Analyzing an Imitation Learning Network for Fundus Image Registration Using a Divide-and-Conquer Approach
|
Comparison of microvascular circulation on fundoscopic images is a non-invasive clinical indication for the diagnosis and monitoring of diseases, such as diabetes and hypertensions. The differences between intra-patient images can be assessed quantitatively by registering serial acquisitions. Due to the variability of the images (i.e. contrast, luminosity) and the anatomical changes of the retina, the registration of fundus images remains a challenging task. Recently, several deep learning approaches have been proposed to register fundus images in an end-to-end fashion, achieving remarkable results. However, the results are difficult to interpret and analyze. In this work, we propose an imitation learning framework for the registration of 2D color funduscopic images for a wide range of applications such as disease monitoring, image stitching and super-resolution. We follow a divide-and-conquer approach to improve the interpretability of the proposed network, and analyze both the influence of the input image and the hyperparameters on the registration result. The results show that the proposed registration network reduces the initial target registration error up to 95\%.
|
['Nishant Ravikumar', 'Siming Bayer', 'Weilin Fu', 'Xia Zhong', 'Andreas Maier']
|
2019-12-19
| null | null | null | null |
['image-stitching']
|
['computer-vision']
|
[ 3.02106626e-02 -1.02491021e-01 -4.19991463e-02 -3.57484967e-01
-2.99331874e-01 -4.04075027e-01 2.21123606e-01 -1.16593942e-01
-4.83840764e-01 5.14923096e-01 -7.90105090e-02 -8.91760662e-02
-1.60651475e-01 -4.18037593e-01 -4.50929403e-01 -8.28546822e-01
-4.05215919e-02 -4.24069352e-02 1.69703782e-01 1.03880249e-01
1.95606917e-01 3.58744770e-01 -1.50557721e+00 -1.31435454e-01
1.31207633e+00 9.21444476e-01 6.12811707e-02 3.54820490e-01
1.06474116e-01 3.93442363e-01 -4.30572093e-01 -3.99329305e-01
4.72231805e-01 -6.51846051e-01 -3.71094584e-01 2.81957030e-01
7.51644075e-01 -4.55474854e-01 -2.57668197e-01 1.46283555e+00
6.95785880e-01 -5.79899810e-02 3.80679965e-01 -6.99088037e-01
-3.55168104e-01 1.65214032e-01 -7.29509354e-01 3.35988700e-01
-6.36056736e-02 3.12932342e-01 3.90988767e-01 -3.02723795e-01
5.42324305e-01 9.55957294e-01 3.77427727e-01 2.74569243e-01
-1.18283248e+00 -5.71625412e-01 -1.79545611e-01 1.63903385e-01
-1.14577281e+00 -4.70365971e-01 6.32469416e-01 -7.12730646e-01
1.16253152e-01 1.34530649e-01 7.69639194e-01 5.60564280e-01
3.51506680e-01 2.61815101e-01 1.55449188e+00 -3.18900615e-01
-1.35983407e-01 1.44084454e-01 -2.48333603e-01 6.48125827e-01
5.24952829e-01 1.75135702e-01 1.06140219e-01 6.24074936e-02
1.27955353e+00 3.10205109e-03 -4.78184253e-01 -3.31267625e-01
-9.94013667e-01 5.35157263e-01 6.95079207e-01 3.43098819e-01
-2.51788437e-01 -1.63103063e-02 2.97527909e-01 2.03642085e-01
3.81636709e-01 5.64698875e-01 -1.07864037e-01 1.14544235e-01
-6.58898890e-01 -1.96379855e-01 2.65807867e-01 3.31339568e-01
2.89897740e-01 -1.19615003e-01 -1.42864838e-01 6.20297253e-01
3.67362589e-01 3.01571399e-01 7.95704067e-01 -9.40554380e-01
3.17553014e-01 5.68342686e-01 3.71602803e-01 -9.35155153e-01
-6.66285038e-01 -6.37117147e-01 -1.06881666e+00 5.17509222e-01
7.60946810e-01 -2.99065858e-01 -6.31387532e-01 1.46563804e+00
4.88728940e-01 2.62138844e-01 -1.47003680e-01 1.33649623e+00
8.00255597e-01 -3.64658423e-02 3.75824124e-02 -4.59673345e-01
1.24030614e+00 -8.90503645e-01 -7.66271889e-01 2.87779979e-02
4.89986330e-01 -8.82162809e-01 9.74064112e-01 1.50863826e-01
-1.29835212e+00 -7.24503756e-01 -8.43074799e-01 -5.50679564e-02
1.94243833e-01 5.28668940e-01 4.05795515e-01 3.88312876e-01
-9.15046453e-01 6.95341945e-01 -1.00356781e+00 -2.45511666e-01
4.97185409e-01 4.75333482e-01 -3.60895514e-01 2.07643002e-01
-7.86863685e-01 8.19579780e-01 1.64798364e-01 5.11006117e-01
-1.08430944e-01 -6.85204029e-01 -5.11344612e-01 -1.00548401e-01
7.56947398e-02 -9.11639094e-01 6.53869212e-01 -1.11918926e+00
-1.72977901e+00 1.14579999e+00 -5.88639267e-02 -2.86025673e-01
8.88702035e-01 -7.71319270e-02 -4.15111303e-01 1.92670360e-01
-1.79991722e-01 3.61630499e-01 6.96380138e-01 -7.15260684e-01
-4.44025248e-01 -6.82246685e-01 -3.99501733e-02 1.70940220e-01
6.96825236e-02 1.34806156e-01 -3.90952379e-01 -4.43412662e-01
3.78436074e-02 -1.00342262e+00 -2.53955424e-01 4.39539522e-01
-4.18007493e-01 2.73801297e-01 1.04347847e-01 -6.22298062e-01
8.46097469e-01 -2.25465560e+00 1.16809078e-01 5.15329279e-02
5.36182463e-01 4.66481537e-01 -8.29857886e-02 -4.39132482e-01
-6.01941980e-02 9.16205999e-03 6.62688166e-02 -1.38067439e-01
-6.24198437e-01 -3.05555642e-01 1.60613939e-01 8.75164807e-01
-1.54693183e-02 8.43132377e-01 -7.41757393e-01 -5.39866984e-01
3.20432305e-01 5.43539226e-01 -3.18798184e-01 4.57489610e-01
8.34006220e-02 1.13642538e+00 -4.32539940e-01 2.80466318e-01
6.74653590e-01 -4.16808188e-01 3.50686193e-01 -6.40237451e-01
-3.13649207e-01 -1.20306220e-02 -9.20630038e-01 1.52689743e+00
-3.68854135e-01 8.24605286e-01 -4.74690944e-02 -9.13736403e-01
9.27789569e-01 3.41254234e-01 5.12825251e-01 -9.45620477e-01
4.96855617e-01 3.72644216e-01 6.07741535e-01 -8.86141777e-01
-8.08400214e-02 -9.64728221e-02 5.90841234e-01 2.74117708e-01
-4.14279342e-01 1.51921660e-01 1.86110228e-01 -6.58750892e-01
4.26468939e-01 6.49223030e-02 4.35496777e-01 -9.64457020e-02
6.73571527e-01 -2.43873477e-01 3.88074189e-01 2.31141284e-01
-4.08290774e-01 7.48766065e-01 6.72883868e-01 -6.63539648e-01
-1.01811540e+00 -6.67845428e-01 -4.77175921e-01 1.37483314e-01
5.06423235e-01 3.67104858e-01 -6.63349450e-01 -4.47018981e-01
6.16183765e-02 -3.19983922e-02 -6.69100285e-01 -1.07790038e-01
-5.09885192e-01 -9.30624127e-01 3.41360778e-01 2.15757772e-01
7.31960475e-01 -7.54478514e-01 -8.80605042e-01 -1.44835347e-02
-1.13415003e-01 -1.28072262e+00 -4.42841530e-01 -5.79507232e-01
-1.22725725e+00 -1.39885247e+00 -9.38918173e-01 -7.62929142e-01
8.28641355e-01 2.25719482e-01 7.32630312e-01 3.96100692e-02
-4.86300737e-01 -1.78983405e-01 5.97253926e-02 3.49093415e-02
-4.59746271e-01 -1.16124958e-01 -7.03789368e-02 4.53614742e-01
-5.21908887e-02 -7.03265250e-01 -1.15319872e+00 5.49889743e-01
-6.34117484e-01 3.80729698e-02 8.52763653e-01 6.72115564e-01
8.39080930e-01 -1.28832206e-01 2.32509881e-01 -7.25853324e-01
6.37560308e-01 5.18454537e-02 -1.12818265e+00 3.81249309e-01
-5.44577658e-01 4.24048454e-02 4.66128916e-01 -5.32641888e-01
-7.95402229e-01 -1.53884783e-01 1.78644717e-01 -4.66478050e-01
-9.87176821e-02 3.71098638e-01 2.47784585e-01 -5.58310449e-01
6.83073759e-01 -1.82819273e-02 7.01762795e-01 -3.30798000e-01
1.69154882e-01 5.69703639e-01 4.79834020e-01 -4.91707250e-02
5.06727517e-01 5.88159800e-01 2.41355985e-01 -6.03320479e-01
-6.69064999e-01 -3.02275807e-01 -5.97308457e-01 -2.28546426e-01
9.45547044e-01 -8.15812469e-01 -9.89472985e-01 5.84003806e-01
-1.02878249e+00 -2.83622503e-01 -1.32368222e-01 8.48417401e-01
-3.91398251e-01 5.92886090e-01 -4.24228132e-01 -3.11825752e-01
-5.35249770e-01 -1.55659688e+00 5.72175801e-01 6.38214231e-01
9.32974294e-02 -1.08675218e+00 9.05579552e-02 3.07420522e-01
5.86713254e-01 5.90286493e-01 1.00687408e+00 -1.71000496e-01
-6.27478063e-01 -5.80527931e-02 -5.92511952e-01 3.71739924e-01
3.20623398e-01 -1.90724141e-03 -7.53802717e-01 -2.98899263e-01
7.40230158e-02 3.17437425e-02 5.10951281e-01 9.08668995e-01
9.94906783e-01 -1.49793029e-01 -5.73583096e-02 1.04891682e+00
1.37478423e+00 2.00551376e-01 9.16844487e-01 3.17640364e-01
5.85632801e-01 7.93126643e-01 4.18463171e-01 1.62713706e-01
1.34069428e-01 8.61632407e-01 4.59057450e-01 -5.63172281e-01
-3.19569647e-01 2.58005351e-01 2.13282183e-02 5.15109599e-01
-5.32132328e-01 1.49053603e-01 -3.96529496e-01 2.39484221e-01
-1.60816765e+00 -7.32936382e-01 -2.40117446e-01 2.66859293e+00
8.55189204e-01 -1.97751954e-01 1.31439045e-01 -5.09261012e-01
9.08739507e-01 -3.88869457e-02 -7.31779039e-01 7.86801800e-02
-3.12117785e-02 -6.69053569e-02 4.95832682e-01 4.17599171e-01
-9.13865864e-01 4.88609642e-01 5.78343105e+00 2.13165939e-01
-1.73546505e+00 9.93545726e-03 8.01871121e-01 -7.60576967e-03
-4.55766879e-02 -2.09252238e-01 -4.17222977e-01 6.32490158e-01
4.31552351e-01 1.19726323e-02 5.04119992e-01 2.63427585e-01
5.33438861e-01 -1.71018794e-01 -9.09626842e-01 1.25031126e+00
-1.87302623e-02 -1.13947141e+00 -2.37805888e-01 1.52864531e-01
7.36078918e-01 7.24598393e-02 3.37242663e-01 -3.81981224e-01
-4.59550440e-01 -9.66957808e-01 7.61561468e-02 8.03988218e-01
9.84647453e-01 -3.40278149e-01 8.74395072e-01 -1.91252217e-01
-6.32611692e-01 7.66364858e-02 -3.24443728e-01 2.27205977e-01
1.54975215e-02 5.58283865e-01 -3.93895954e-01 2.88494974e-01
5.46233296e-01 6.57379866e-01 -7.05946684e-01 1.60456777e+00
-1.43269971e-01 7.43431747e-02 1.00599244e-01 2.39343166e-01
-2.47755915e-01 -8.91265750e-01 5.60754716e-01 5.27807415e-01
4.11585897e-01 -8.92447457e-02 -2.32604668e-01 1.21411622e+00
2.60951072e-02 2.45385960e-01 -1.85442507e-01 1.17356099e-01
2.57070605e-02 1.28194141e+00 -4.65525478e-01 -1.25627741e-01
-3.37293714e-01 6.25301003e-01 5.17374277e-02 5.17711639e-01
-8.44261825e-01 -3.22118670e-01 4.45321202e-01 4.34863240e-01
4.56167795e-02 -2.92813666e-02 -1.32465094e-01 -1.38242328e+00
3.74912411e-01 -7.95046866e-01 6.83058277e-02 -8.00073028e-01
-9.42822993e-01 7.54954994e-01 -2.22990870e-01 -1.58330500e+00
-5.56786694e-02 -4.46021914e-01 -6.92499459e-01 9.04718935e-01
-1.90196860e+00 -6.80776954e-01 -8.18959653e-01 5.07883072e-01
-2.03420997e-01 -1.54356867e-01 4.18853611e-01 4.70390350e-01
-7.78716087e-01 6.81487083e-01 2.63931125e-01 1.23572685e-01
1.03348327e+00 -1.02242005e+00 -4.85347174e-02 7.39829540e-01
-2.63125598e-01 5.89090765e-01 5.51901698e-01 -3.01969141e-01
-9.40756202e-01 -1.03114963e+00 4.33318347e-01 1.19002527e-02
4.81350273e-01 2.72620171e-01 -7.21649289e-01 4.03928787e-01
1.65347144e-01 5.91383576e-01 3.84742260e-01 -2.16468424e-01
-6.42528832e-02 -6.00742996e-01 -1.18617606e+00 6.30344331e-01
5.37028193e-01 -2.72208512e-01 -1.68157950e-01 2.35710785e-01
2.85033703e-01 -8.09815288e-01 -1.06749988e+00 2.10869461e-01
8.47449958e-01 -1.23376453e+00 8.54509354e-01 -3.84219140e-01
5.70878685e-01 -3.18594724e-01 5.50964594e-01 -1.19523942e+00
-8.52457210e-02 -7.54296005e-01 2.95982778e-01 8.04225683e-01
2.66524881e-01 -1.05773032e+00 4.41386729e-01 6.11207306e-01
1.09709352e-01 -5.36000192e-01 -8.53555977e-01 -3.83159041e-01
-2.37339422e-01 3.74117345e-01 3.25146228e-01 6.90726578e-01
-2.34457895e-01 6.86621889e-02 -2.26076737e-01 2.11932003e-01
6.33026838e-01 4.46679890e-01 7.74478316e-01 -1.28910971e+00
-2.58713961e-01 -5.82065761e-01 -6.61540389e-01 -8.48084033e-01
-2.22957179e-01 -7.00110674e-01 -2.08141252e-01 -1.12048352e+00
5.78253120e-02 -4.84489590e-01 -1.90067485e-01 -7.36964401e-03
-3.09713840e-01 3.13196331e-01 -2.60606874e-02 5.76415002e-01
-3.17135125e-01 3.70279104e-01 1.86436582e+00 3.89944091e-02
-5.84945619e-01 2.94724077e-01 -4.60714251e-01 5.77360749e-01
8.77012908e-01 -2.46560350e-01 -2.02814445e-01 -4.16528583e-01
9.61577520e-02 3.66547018e-01 5.52189767e-01 -8.44259202e-01
1.74543738e-01 1.98211774e-01 2.63277829e-01 2.55936496e-02
-6.22716062e-02 -7.37149060e-01 1.36182144e-01 5.50928116e-01
-3.32402766e-01 -3.91476378e-02 7.46349543e-02 3.70809734e-01
-4.60485250e-01 -5.10822088e-02 1.13397729e+00 -9.24555957e-02
-1.45524219e-01 3.44845325e-01 1.84275419e-01 2.78745025e-01
8.85743856e-01 -1.77311867e-01 -5.91427326e-01 -2.56634623e-01
-6.78666294e-01 -2.19500139e-02 4.58006144e-01 1.18604109e-01
4.56341952e-01 -8.90075266e-01 -6.57450140e-01 3.04716706e-01
1.79402158e-02 2.98953149e-02 5.01147985e-01 1.83530557e+00
-7.37843692e-01 1.71398178e-01 -5.46739221e-01 -6.85242653e-01
-1.23659623e+00 3.70271713e-01 1.03544188e+00 -1.68279737e-01
-6.57487154e-01 2.96038479e-01 2.47409746e-01 7.68564045e-02
1.82105049e-01 -6.57426953e-01 -5.34468234e-01 -1.19897395e-01
4.60895211e-01 3.17570180e-01 -4.91290838e-02 -4.78739530e-01
4.59187664e-02 1.23459947e+00 -8.35285708e-02 4.31541383e-01
1.07074273e+00 -3.83689135e-01 -2.90411770e-01 5.20998649e-02
1.04002929e+00 -8.05743262e-02 -1.32009816e+00 -2.07087442e-01
-4.57463026e-01 -6.50429666e-01 2.29770362e-01 -6.31074309e-01
-1.55510163e+00 9.68411505e-01 1.18699741e+00 9.71543491e-02
1.09556460e+00 -3.24161291e-01 5.87852240e-01 -5.17038330e-02
3.22138220e-02 -6.74076974e-01 -1.74764082e-01 -2.74173349e-01
7.68589020e-01 -1.49954307e+00 1.33863568e-01 -4.14672226e-01
-6.05791509e-01 1.30857420e+00 4.01205391e-01 -1.42089143e-01
3.71864915e-01 -1.70752987e-01 4.83171016e-01 -1.35463011e-02
-1.14801452e-02 -2.38005653e-01 6.12924397e-01 3.60024661e-01
5.03118098e-01 -1.34032026e-01 -5.57540119e-01 1.32188886e-01
1.32953525e-01 2.84194112e-01 6.03479207e-01 1.29078552e-01
-5.61221456e-03 -8.71204674e-01 -1.08525611e-01 2.58818418e-01
-5.79201281e-01 9.31393504e-02 1.91592410e-01 9.41787958e-01
5.98281585e-02 6.77147806e-01 2.27128640e-01 9.47708413e-02
3.64549965e-01 -4.17135447e-01 6.73411429e-01 -2.27310911e-01
-4.03086901e-01 3.73752922e-01 -1.42705366e-01 -5.00371456e-01
-8.25973451e-01 -4.28684115e-01 -8.88099968e-01 -2.81775352e-02
-2.57372946e-01 -2.20086545e-01 4.99357671e-01 8.05005372e-01
5.44972122e-01 4.27861184e-01 8.22710752e-01 -4.67752159e-01
-5.95368922e-01 -8.39158773e-01 -6.05745077e-01 5.19371271e-01
4.78034645e-01 -5.17267406e-01 -4.77610588e-01 3.79581265e-02]
|
[15.742307662963867, -3.926579236984253]
|
3986ed8b-3e91-4fce-80b1-76c44f795e49
|
knowledge-distillation-for-multi-target
|
2205.06237
| null |
https://arxiv.org/abs/2205.06237v2
|
https://arxiv.org/pdf/2205.06237v2.pdf
|
Knowledge Distillation for Multi-Target Domain Adaptation in Real-Time Person Re-Identification
|
Despite the recent success of deep learning architectures, person re-identification (ReID) remains a challenging problem in real-word applications. Several unsupervised single-target domain adaptation (STDA) methods have recently been proposed to limit the decline in ReID accuracy caused by the domain shift that typically occurs between source and target video data. Given the multimodal nature of person ReID data (due to variations across camera viewpoints and capture conditions), training a common CNN backbone to address domain shifts across multiple target domains, can provide an efficient solution for real-time ReID applications. Although multi-target domain adaptation (MTDA) has not been widely addressed in the ReID literature, a straightforward approach consists in blending different target datasets, and performing STDA on the mixture to train a common CNN. However, this approach may lead to poor generalization, especially when blending a growing number of distinct target domains to train a smaller CNN. To alleviate this problem, we introduce a new MTDA method based on knowledge distillation (KD-ReID) that is suitable for real-time person ReID applications. Our method adapts a common lightweight student backbone CNN over the target domains by alternatively distilling from multiple specialized teacher CNNs, each one adapted on data from a specific target domain. Extensive experiments conducted on several challenging person ReID datasets indicate that our approach outperforms state-of-art methods for MTDA, including blending methods, particularly when training a compact CNN backbone like OSNet. Results suggest that our flexible MTDA approach can be employed to design cost-effective ReID systems for real-time video surveillance applications.
|
['Eric Granger', 'Rafael M. O. Cruz', 'Le Thanh Nguyen-Meidine', 'Sajjad Abdoli', 'Djebril Mekhazni', 'Félix Remigereau']
|
2022-05-12
| null | null | null | null |
['multi-target-domain-adaptation']
|
['computer-vision']
|
[ 4.06328700e-02 -2.78942913e-01 2.61445786e-03 -5.17267466e-01
-5.20237625e-01 -7.14751244e-01 7.34759927e-01 -1.88118573e-02
-7.57899761e-01 8.34254563e-01 9.48929116e-02 -1.00786041e-03
-9.22620296e-02 -5.85498214e-01 -7.65260875e-01 -5.79676926e-01
6.43808320e-02 7.93094933e-01 3.03702712e-01 -2.00884163e-01
-1.59147903e-01 6.38666511e-01 -1.49083650e+00 -1.24350101e-01
9.70418155e-01 4.50683177e-01 4.62770574e-02 3.62341344e-01
-2.72357196e-01 3.32655966e-01 -1.13441610e+00 -7.92526662e-01
6.78821743e-01 -3.12770605e-01 -7.98149347e-01 -1.26311988e-01
8.02685201e-01 -5.49574733e-01 -4.66785789e-01 9.56752121e-01
8.17449868e-01 4.25509036e-01 5.56221247e-01 -1.41831505e+00
-6.46200359e-01 3.40028465e-01 -5.98276615e-01 3.76207232e-01
4.66649264e-01 8.19311589e-02 1.11620635e-01 -4.86334920e-01
6.11680746e-01 1.39304042e+00 8.91632557e-01 9.71258402e-01
-1.32181311e+00 -1.23788536e+00 1.86466232e-01 2.61047751e-01
-1.50795376e+00 -2.99866349e-01 7.52487898e-01 -3.16176742e-01
7.20015407e-01 -9.78613719e-02 4.48915958e-01 1.63134575e+00
-1.67024434e-01 8.06001663e-01 9.54740644e-01 -5.97685203e-02
1.67771652e-01 2.94688225e-01 2.02516578e-02 1.49291545e-01
4.95843977e-01 -3.99477407e-03 -3.96041483e-01 -2.41646022e-01
7.29806960e-01 9.86184459e-03 -1.91503853e-01 -5.14201701e-01
-1.06350470e+00 5.22731245e-01 5.39588809e-01 6.82313964e-02
-2.55687177e-01 -3.27414244e-01 6.95301235e-01 5.75794578e-01
5.18999755e-01 4.57807213e-01 -3.56807798e-01 1.31972460e-02
-9.75619256e-01 7.00948417e-01 7.86108971e-01 1.21561444e+00
7.80244768e-01 -1.58014536e-01 -1.16284840e-01 9.56701875e-01
-1.81101620e-01 4.49851632e-01 4.20870602e-01 -5.20646930e-01
4.85346109e-01 6.50640726e-01 2.14783892e-01 -7.04749107e-01
-3.91356498e-01 -3.33327055e-01 -1.04719007e+00 -2.62537338e-02
7.76445568e-01 -2.38336563e-01 -1.09245849e+00 1.93042254e+00
6.13039434e-01 3.70011717e-01 1.88779414e-01 9.10210788e-01
8.92690837e-01 2.84198195e-01 3.87379020e-01 3.20076585e-01
1.26760399e+00 -7.04605043e-01 -3.49177539e-01 -2.24802762e-01
4.84169841e-01 -5.96146703e-01 4.60567117e-01 1.76220581e-01
-7.24621356e-01 -8.22282672e-01 -9.41923797e-01 1.72754917e-02
-7.45834291e-01 -6.95743710e-02 6.30814508e-02 7.55244613e-01
-1.05914259e+00 2.47777194e-01 -4.75126892e-01 -1.10537267e+00
4.10874188e-01 8.09946835e-01 -7.57830501e-01 -4.73163575e-01
-1.19118798e+00 9.21647489e-01 5.73940873e-01 -1.20574147e-01
-7.47151375e-01 -8.48729134e-01 -8.79798710e-01 -5.28558120e-02
2.98552662e-01 -1.07636344e+00 1.21975088e+00 -1.21236992e+00
-1.44932020e+00 9.48770583e-01 -7.22932369e-02 -5.52226484e-01
8.11373353e-01 -3.81797731e-01 -5.30266166e-01 -5.50661124e-02
2.76085526e-01 1.04392505e+00 9.90108311e-01 -1.28107882e+00
-6.55985117e-01 -3.59188974e-01 1.89630389e-01 1.45522714e-01
-4.66222584e-01 2.48028740e-01 -4.92024183e-01 -6.33402586e-01
-5.33420742e-01 -9.78586018e-01 -4.71156964e-05 -1.91434130e-01
-1.91011861e-01 -5.01467526e-01 1.03383243e+00 -5.82456768e-01
6.60380661e-01 -2.05956745e+00 1.24500498e-01 2.15077177e-02
6.38447627e-02 8.13557863e-01 -5.08905053e-01 3.44527334e-01
-2.52222270e-01 -2.45201662e-01 -8.99679959e-02 -6.33773208e-01
-2.28900522e-01 2.84018427e-01 4.74978331e-03 5.20097375e-01
3.40911537e-01 6.63500607e-01 -1.04595661e+00 -3.86426330e-01
8.44408944e-02 5.36120772e-01 -2.82316238e-01 5.90686083e-01
1.33062914e-01 7.59536266e-01 -1.76479369e-01 5.77443361e-01
1.19416142e+00 6.61658645e-02 -7.59351673e-03 -7.48086274e-02
5.23998402e-02 -2.37625018e-02 -1.29207623e+00 1.65303290e+00
-5.46813942e-02 5.72953463e-01 3.16825733e-02 -1.24260521e+00
1.09263492e+00 2.25625977e-01 3.97065014e-01 -6.86944783e-01
9.93689522e-02 1.27283290e-01 -1.50150433e-01 -4.32893008e-01
7.26589322e-01 -7.29470281e-03 -1.34688184e-01 2.24846765e-01
2.84184724e-01 3.83907825e-01 1.66349545e-01 1.10885456e-01
1.08036518e+00 7.01939017e-02 1.74566492e-01 -2.52811998e-01
6.51194632e-01 2.87796170e-01 7.29009867e-01 9.29986238e-01
-5.14192283e-01 6.90690935e-01 1.09317683e-01 -8.11775208e-01
-1.21977854e+00 -1.16660702e+00 6.73192069e-02 1.07956290e+00
3.79591852e-01 3.04031409e-02 -8.40528905e-01 -9.47044373e-01
1.62558571e-01 2.16083199e-01 -5.56882381e-01 -1.96882457e-01
-9.34231281e-01 -7.34103382e-01 8.69730532e-01 7.36885071e-01
1.00187433e+00 -7.20622420e-01 -2.40909085e-01 2.95480102e-01
-2.23868534e-01 -1.30924904e+00 -5.45260370e-01 -2.18389183e-02
-5.42518139e-01 -1.14244759e+00 -1.44904912e+00 -8.10913086e-01
7.74996996e-01 6.70176029e-01 9.53590512e-01 -1.24918170e-01
-9.50332358e-02 6.00565374e-01 -3.85367304e-01 -5.30935764e-01
-3.95160764e-01 4.19364661e-01 4.44565773e-01 9.71336216e-02
8.13725770e-01 -3.56341302e-01 -5.83997846e-01 5.71563363e-01
-9.25678492e-01 -2.51626223e-01 5.30630052e-01 9.56607819e-01
4.67466228e-02 -6.07309602e-02 8.73588085e-01 -6.96937919e-01
7.60573030e-01 -6.72226012e-01 -6.32036328e-01 2.16233999e-01
-2.33898133e-01 -1.36340082e-01 6.81575358e-01 -9.25557554e-01
-1.33741689e+00 -4.23162319e-02 1.54109061e-01 -6.38068020e-01
-6.05222464e-01 1.44673750e-01 -1.79451436e-01 -1.51379660e-01
7.77492464e-01 2.64965534e-01 2.08328828e-01 -4.99000371e-01
5.31207956e-02 7.51532495e-01 9.99302208e-01 -6.56199098e-01
1.13322616e+00 3.64061952e-01 -3.33804846e-01 -7.02586710e-01
-3.98784250e-01 -8.27294230e-01 -8.48024249e-01 -1.43621519e-01
9.30805147e-01 -1.31041396e+00 -4.81537819e-01 8.92472863e-01
-1.35966349e+00 -3.20303172e-01 1.39034037e-02 3.99617344e-01
-1.83088660e-01 5.04677951e-01 -2.38612741e-01 -4.41071957e-01
-3.48546386e-01 -1.07784820e+00 8.92272174e-01 6.64298713e-01
-3.54814619e-01 -9.65675294e-01 1.58897564e-01 3.65998477e-01
5.66778421e-01 5.87527044e-02 5.17800927e-01 -1.16371214e+00
-3.97048175e-01 -4.00093198e-01 -4.98151481e-01 2.19694555e-01
2.92817682e-01 -4.98648465e-01 -9.77555633e-01 -7.81975269e-01
-4.49332595e-01 -2.64305532e-01 6.48552954e-01 1.42999530e-01
8.91008794e-01 -7.84749016e-02 -6.14291131e-01 6.81336582e-01
1.06182575e+00 1.18949123e-01 4.89106715e-01 8.46298575e-01
6.58321679e-01 6.38613880e-01 5.36112487e-01 2.05012113e-01
7.95334518e-01 7.47968495e-01 1.53280556e-01 -2.76506841e-01
-2.71964490e-01 -1.92983881e-01 3.85933489e-01 1.50909111e-01
-1.36628196e-01 -2.90262669e-01 -1.05341482e+00 9.12765622e-01
-1.90585029e+00 -9.52207685e-01 2.83754736e-01 2.39919853e+00
5.53991735e-01 -1.75165325e-01 7.56216288e-01 -4.26743358e-01
1.02442348e+00 -2.75137246e-01 -7.80250072e-01 -1.52088344e-01
-1.07077666e-01 -5.35021443e-03 6.96376324e-01 -8.61445069e-02
-1.25354826e+00 8.32521439e-01 6.04804277e+00 5.69128931e-01
-1.05277169e+00 1.17099807e-01 1.17894284e-01 6.51988238e-02
3.54299992e-01 -3.36973816e-01 -1.04608500e+00 5.61898649e-01
8.82558107e-01 -2.42738053e-01 8.97941142e-02 9.17815030e-01
-1.69669747e-01 7.09011406e-03 -1.27606070e+00 1.12320817e+00
9.66034904e-02 -1.04935050e+00 1.46456182e-01 -2.66712811e-02
7.56608546e-01 1.32255871e-02 -7.94876441e-02 5.73016942e-01
4.08775508e-01 -6.66753232e-01 3.83420378e-01 1.17111824e-01
6.19643927e-01 -8.10682893e-01 1.06325054e+00 3.89326274e-01
-1.26432276e+00 -1.76298216e-01 -6.15392208e-01 1.91698894e-01
-4.19835411e-02 3.05746254e-02 -1.10779834e+00 7.95369506e-01
1.17808950e+00 6.57026649e-01 -5.70603609e-01 1.39459050e+00
2.52175778e-01 8.82225409e-02 -3.07961345e-01 3.48187327e-01
3.72564979e-02 1.25439525e-01 6.88215077e-01 1.48896849e+00
2.52644360e-01 -7.47359395e-02 2.90442497e-01 7.57023394e-01
-1.80582047e-01 -2.65901774e-01 -7.25369632e-01 3.86300653e-01
7.36050606e-01 8.96026492e-01 -4.35777962e-01 -5.12858450e-01
-5.62001765e-01 1.25745666e+00 4.04059231e-01 6.24358654e-01
-7.40769923e-01 -2.66534090e-01 1.10910213e+00 9.52002481e-02
2.74380922e-01 -2.33335152e-01 1.36382788e-01 -1.28645217e+00
8.30296576e-02 -9.74456131e-01 7.40712583e-01 -3.10479105e-01
-1.79849958e+00 5.88079691e-01 5.96583426e-01 -1.49261713e+00
-2.48014688e-01 -3.46484780e-01 -6.75038517e-01 1.09672141e+00
-1.85732543e+00 -1.32992852e+00 -5.61434805e-01 1.10377312e+00
6.33293390e-01 -4.94632244e-01 6.87931359e-01 4.64231580e-01
-8.74690354e-01 1.18500233e+00 7.09679574e-02 5.11647284e-01
1.35826600e+00 -1.21220350e+00 7.13796616e-01 9.65677202e-01
-5.23186445e-01 7.00105488e-01 7.10851312e-01 -7.60192871e-01
-1.31596518e+00 -1.32497334e+00 6.32452667e-01 -3.73527199e-01
2.87740022e-01 -3.33934397e-01 -1.25048661e+00 7.56566703e-01
2.97519505e-01 4.25205901e-02 7.16002405e-01 1.95603333e-02
-7.16433942e-01 -3.55701685e-01 -1.47473538e+00 4.60034758e-01
1.06062555e+00 -4.31673825e-01 -7.46290922e-01 1.41206598e-02
3.22893143e-01 -5.35663605e-01 -8.44471514e-01 3.38846654e-01
4.66044277e-01 -8.18728626e-01 1.16621363e+00 -5.78867078e-01
-8.92640799e-02 -3.28559548e-01 3.82162482e-01 -1.48684394e+00
-2.98583090e-01 -5.20823359e-01 7.68804550e-02 1.54766190e+00
-1.86531663e-01 -9.34088349e-01 8.36122334e-01 9.78840888e-01
5.36204353e-02 1.64534978e-03 -1.07166350e+00 -1.33029652e+00
2.60861725e-01 1.04031757e-01 9.62231576e-01 1.10683322e+00
-3.32899779e-01 -1.79507881e-02 -4.87025291e-01 5.78219175e-01
8.45983744e-01 -3.03386301e-01 1.53000069e+00 -1.45490742e+00
1.08995326e-01 -3.93182397e-01 -6.74085438e-01 -1.13724053e+00
3.60663474e-01 -6.15348458e-01 2.42632274e-02 -1.17632556e+00
3.47186685e-01 -5.08587718e-01 -2.63726026e-01 5.81701636e-01
-2.86608934e-01 1.48570910e-01 3.03974360e-01 3.33062977e-01
-3.70279193e-01 3.59974235e-01 8.46972048e-01 -4.44067836e-01
-3.73108327e-01 -6.39942214e-02 -5.64458787e-01 4.27947283e-01
7.13622987e-01 -5.37540555e-01 -3.63582462e-01 -6.40244424e-01
-4.13966596e-01 -2.63015509e-01 5.95840871e-01 -1.24308527e+00
6.95783913e-01 3.04664392e-02 6.87829673e-01 -4.97442365e-01
2.10801363e-01 -8.67987275e-01 2.62859344e-01 2.48618141e-01
7.60039315e-02 1.39983609e-01 5.25876641e-01 5.40058792e-01
-2.25972101e-01 7.29370266e-02 8.66135538e-01 -7.79236779e-02
-1.05174804e+00 3.13963205e-01 -2.46142328e-01 -1.46116748e-01
1.07750297e+00 -6.00174546e-01 -5.44292510e-01 -1.95565507e-01
-3.95690769e-01 5.23947358e-01 5.52074015e-01 8.02131295e-01
5.30112028e-01 -1.27051580e+00 -8.00933480e-01 4.00855124e-01
3.70888919e-01 2.81832010e-01 4.51874495e-01 5.25348365e-01
-1.96486786e-01 2.83456326e-01 -5.82048297e-01 -6.99750543e-01
-1.38659453e+00 6.40680671e-01 4.91646945e-01 -1.96914047e-01
-6.70789123e-01 7.62491703e-01 3.77295285e-01 -6.47389829e-01
3.61471206e-01 1.42968684e-01 -1.34969220e-01 5.64619079e-02
7.08778143e-01 4.19035226e-01 3.66908461e-02 -7.74993062e-01
-4.56043601e-01 5.88455379e-01 -5.71747899e-01 4.29679751e-01
1.18196762e+00 -2.69924849e-01 2.19941095e-01 -2.75285959e-01
1.07271671e+00 -5.38049817e-01 -1.34253550e+00 -4.44607496e-01
-1.65291708e-02 -4.15263414e-01 -5.06554961e-01 -6.87237024e-01
-7.81565607e-01 4.90106553e-01 7.30387807e-01 -7.85563737e-02
1.28380466e+00 -4.43329066e-02 9.42309320e-01 5.15394449e-01
4.85886633e-01 -1.01425660e+00 2.82069147e-02 3.72037649e-01
6.59460008e-01 -1.59964073e+00 -1.54430568e-01 -1.35004506e-01
-4.14267570e-01 1.04081082e+00 1.13674998e+00 -5.04169464e-02
3.44648331e-01 -8.69493559e-02 2.89037645e-01 5.54488599e-02
-5.89026101e-02 -1.94378823e-01 1.95675939e-01 1.14205718e+00
-1.45153925e-02 -1.41139209e-01 5.28562255e-02 4.06330317e-01
-3.87542807e-02 1.74651384e-01 5.06793082e-01 9.26958323e-01
1.24422535e-01 -1.51945329e+00 -7.36141741e-01 4.92546149e-02
-2.54328758e-01 4.17977750e-01 -3.90996814e-01 1.17524803e+00
3.52344811e-01 8.73183489e-01 1.58264756e-01 -3.57534617e-01
5.52689731e-01 8.03180709e-02 3.86748940e-01 -5.42567730e-01
-8.02391291e-01 -3.29190731e-01 -1.59336776e-01 -2.53900141e-01
-8.06586742e-01 -7.75531054e-01 -6.78457499e-01 -5.64483285e-01
-4.46235053e-02 -1.36113435e-01 4.03849691e-01 1.01609802e+00
4.56172407e-01 6.39235601e-02 2.85362542e-01 -1.10096729e+00
-4.72188056e-01 -8.86555254e-01 -2.88164824e-01 7.00195968e-01
5.15513062e-01 -8.75279546e-01 5.94296828e-02 7.37644732e-02]
|
[14.69869613647461, 1.0414929389953613]
|
e809de63-c891-42bc-bb47-3b920d7d9370
|
gadformer-an-attention-based-model-for-group
|
2303.09841
| null |
https://arxiv.org/abs/2303.09841v1
|
https://arxiv.org/pdf/2303.09841v1.pdf
|
GADFormer: An Attention-based Model for Group Anomaly Detection on Trajectories
|
Group Anomaly Detection (GAD) reveals anomalous behavior among groups consisting of multiple member instances, which are, individually considered, not necessarily anomalous. This task is of major importance across multiple disciplines, in which also sequences like trajectories can be considered as a group. However, with increasing amount and heterogenity of group members, actual abnormal groups get harder to detect, especially in an unsupervised or semi-supervised setting. Recurrent Neural Networks are well established deep sequence models, but recent works have shown that their performance can decrease with increasing sequence lengths. Hence, we introduce with this paper GADFormer, a GAD specific BERT architecture, capable to perform attention-based Group Anomaly Detection on trajectories in an unsupervised and semi-supervised setting. We show formally and experimentally how trajectory outlier detection can be realized as an attention-based Group Anomaly Detection problem. Furthermore, we introduce a Block Attention-anomaly Score (BAS) to improve the interpretability of transformer encoder blocks for GAD. In addition to that, synthetic trajectory generation allows us to optimize the training for domain-specific GAD. In extensive experiments we investigate our approach versus GRU in their robustness for trajectory noise and novelties on synthetic and real world datasets.
|
['Peer Kröger', 'Darpan Malik', 'Andreas Lohrer']
|
2023-03-17
| null | null | null | null |
['group-anomaly-detection']
|
['methodology']
|
[ 4.17461395e-02 -6.18763044e-02 2.01689288e-01 -2.58144021e-01
-5.17759860e-01 -5.96662760e-01 7.23843396e-01 5.67865610e-01
-3.08945268e-01 4.80667114e-01 2.08114281e-01 -4.74760890e-01
-2.16774195e-01 -7.22324550e-01 -8.79234493e-01 -5.78282118e-01
-5.25698960e-01 5.04675448e-01 1.91862896e-01 -2.09319666e-01
2.61651129e-01 6.75557852e-01 -1.55475056e+00 1.69190198e-01
1.17659080e+00 7.47289240e-01 -2.45339379e-01 7.37168372e-01
1.11254275e-01 7.73133993e-01 -7.17810631e-01 -2.95726717e-01
2.73865074e-01 -6.32852077e-01 -5.92508674e-01 3.47268760e-01
3.66757601e-01 -2.94892251e-01 -3.16950351e-01 1.03497005e+00
3.38519961e-01 4.80808109e-01 7.46788204e-01 -1.49920988e+00
-4.06178623e-01 6.45400524e-01 -2.26696000e-01 6.97438538e-01
1.94531813e-01 4.99987155e-01 1.22187829e+00 -6.62559927e-01
3.94321352e-01 1.03127944e+00 6.06661916e-01 2.90122300e-01
-1.01586390e+00 -3.75373900e-01 4.84517157e-01 5.94657183e-01
-1.04011452e+00 -1.12317197e-01 6.83245718e-01 -4.64757890e-01
1.02781630e+00 3.89156997e-01 6.39535367e-01 1.47729325e+00
-1.36588454e-01 9.21270967e-01 1.79988593e-01 -8.61452669e-02
4.58513439e-01 -3.73025268e-01 2.26630300e-01 2.89493799e-01
4.27836210e-01 -1.17036596e-01 -1.19889930e-01 -1.54891312e-01
3.66238266e-01 3.28047544e-01 -1.86105430e-01 -1.46497965e-01
-1.33638668e+00 7.42579758e-01 4.50616002e-01 6.21769011e-01
-5.40125608e-01 3.44193339e-01 7.83340275e-01 6.98897004e-01
6.24597490e-01 5.96669853e-01 -8.90030190e-02 -4.33605850e-01
-6.67966664e-01 4.83201236e-01 4.72388595e-01 7.87305355e-01
4.17089850e-01 3.89433473e-01 -4.60310429e-01 5.32439768e-01
-7.79407695e-02 9.52064767e-02 8.45679939e-01 -4.65246856e-01
5.56327999e-01 9.18406427e-01 5.27061969e-02 -1.14564097e+00
-6.06116295e-01 -7.38643467e-01 -1.02536690e+00 -1.24747872e-01
5.89047074e-01 9.27519053e-03 -6.84117377e-01 1.86661422e+00
5.54478131e-02 6.95436835e-01 -6.22492880e-02 8.80956352e-01
2.03611270e-01 5.04372180e-01 -4.27891128e-02 -8.23694095e-02
9.44031656e-01 -9.62336242e-01 -6.05943084e-01 -7.01860264e-02
1.15931308e+00 -8.41137990e-02 1.20869386e+00 5.04484177e-01
-8.24731469e-01 -3.85314465e-01 -8.78558040e-01 1.84049845e-01
-4.10583228e-01 -6.95501864e-02 3.16837221e-01 6.38347983e-01
-8.82311225e-01 9.22771752e-01 -1.08611524e+00 -6.02592766e-01
5.39390862e-01 2.26877749e-01 -2.00219050e-01 1.66205376e-01
-1.00450897e+00 4.82963651e-01 6.01868570e-01 8.10176581e-02
-9.10110593e-01 -5.34297109e-01 -6.26033068e-01 1.81141913e-01
3.06284398e-01 -3.71953636e-01 9.06848133e-01 -9.23063695e-01
-9.66966569e-01 5.81312656e-01 -1.52458340e-01 -1.09986305e+00
7.18021452e-01 -2.52043635e-01 -7.67690778e-01 7.20079849e-03
-7.01475814e-02 1.25600174e-02 8.77800941e-01 -7.44616270e-01
-6.53149962e-01 -3.14544946e-01 3.33524309e-02 -9.21562314e-02
-4.70349669e-01 -1.43153682e-01 1.18222848e-01 -7.98351347e-01
7.97961652e-03 -8.20365071e-01 -3.80939960e-01 -4.80620652e-01
-5.64805329e-01 -4.04053181e-01 9.71493125e-01 -6.11687779e-01
1.55136633e+00 -2.21291900e+00 2.38730147e-01 4.70759273e-01
3.15839946e-01 4.18467700e-01 -9.29478705e-02 6.43151224e-01
-2.64668405e-01 1.69897854e-01 -5.58235884e-01 -3.90046537e-01
9.63925049e-02 3.02094400e-01 -4.71222073e-01 6.13764763e-01
4.42234010e-01 9.63393450e-01 -1.10905981e+00 1.92676529e-01
2.88398527e-02 -2.62295544e-01 -7.96991944e-01 2.83284158e-01
-3.33907753e-01 8.37967515e-01 -3.22973520e-01 5.21928608e-01
3.62915099e-01 -1.87123492e-01 -1.51523963e-01 3.76050115e-01
5.75308269e-03 8.79677907e-02 -8.45940828e-01 1.54767799e+00
-1.02696382e-01 6.42781436e-01 -5.62067747e-01 -1.50306439e+00
9.31650400e-01 2.94035137e-01 4.29132849e-01 -4.83415991e-01
-6.72394782e-02 5.49369454e-01 4.64746922e-01 -5.03552437e-01
6.77477419e-01 2.67427117e-01 -9.61234793e-02 7.68214226e-01
-4.61265892e-02 5.61863661e-01 4.58747804e-01 1.91370428e-01
1.52753842e+00 -4.69088584e-01 1.67789117e-01 3.10823694e-02
7.13437617e-01 -2.94462532e-01 3.67106378e-01 9.13037241e-01
-8.21840540e-02 7.71141648e-01 8.77592742e-01 -5.49757659e-01
-1.33722913e+00 -8.00359309e-01 2.44409651e-01 8.96446824e-01
-1.77203253e-01 -4.08262104e-01 -6.91601396e-01 -1.00719726e+00
-1.83397625e-02 7.83097684e-01 -5.46392143e-01 -5.87212801e-01
-8.59970272e-01 -8.92052650e-01 7.80052662e-01 5.24138331e-01
1.94080561e-01 -1.32280457e+00 -5.24629056e-01 3.21306616e-01
-2.83126421e-02 -9.92537677e-01 -3.51116061e-01 4.45352569e-02
-7.32120633e-01 -1.05700588e+00 -5.45331299e-01 -2.92681307e-01
5.37206233e-01 -5.37218116e-02 1.01725507e+00 3.33669662e-01
-3.30235772e-02 4.58814740e-01 -7.31386900e-01 -3.57664347e-01
-6.68232083e-01 4.08280849e-01 3.54967326e-01 5.89636505e-01
4.62602556e-01 -1.07513416e+00 -6.19862020e-01 3.05528849e-01
-1.23174798e+00 -5.23302674e-01 3.78621727e-01 6.38037264e-01
2.66193092e-01 -1.89295992e-01 9.58012283e-01 -6.86956882e-01
7.97474682e-01 -9.59104657e-01 -3.03233474e-01 -1.00706965e-01
-2.52942026e-01 -3.36732306e-02 1.06860495e+00 -2.54830241e-01
-4.59390163e-01 -5.18398941e-01 -4.90224302e-01 -7.79485464e-01
-4.80042815e-01 4.21823382e-01 -1.91242233e-01 3.77889842e-01
6.93863511e-01 3.76456916e-01 -2.63694320e-02 -5.01646578e-01
1.97388902e-01 5.12460172e-01 3.99547249e-01 -4.49149400e-01
7.85194218e-01 4.24276650e-01 -6.98021194e-03 -9.14188921e-01
-5.16328275e-01 -6.07533216e-01 -7.25987554e-01 -2.04661876e-01
6.27445936e-01 -5.76274395e-01 -6.41575992e-01 5.09718657e-01
-9.93390322e-01 -4.13246393e-01 -4.67338204e-01 2.96658337e-01
-6.77113354e-01 6.56379342e-01 -4.55131292e-01 -7.99832880e-01
-1.66881740e-01 -1.03902209e+00 8.64413142e-01 -2.18382820e-01
-4.59328115e-01 -1.15655851e+00 1.20688446e-01 -2.18119428e-01
3.79639000e-01 5.28141558e-01 9.57610965e-01 -1.62702084e+00
-6.47409439e-01 -3.14326078e-01 5.03762159e-03 4.61730450e-01
7.04529211e-02 -2.67969370e-01 -8.35558414e-01 -4.96005148e-01
-2.04978168e-01 1.18399173e-01 8.06971610e-01 2.13444699e-02
1.49817789e+00 -5.19108117e-01 -2.52925783e-01 4.93361920e-01
9.78869975e-01 1.76908791e-01 6.95936382e-01 3.02410543e-01
9.55716252e-01 4.44573432e-01 6.16985857e-01 5.95505774e-01
-8.59867930e-02 8.40558171e-01 7.31869519e-01 2.25484475e-01
2.97093391e-01 -1.24049835e-01 5.79835474e-01 9.22659278e-01
-2.54714280e-01 -6.68456495e-01 -9.57012594e-01 8.13506544e-01
-2.12993956e+00 -1.26154757e+00 -4.09788102e-01 2.28470445e+00
1.83072090e-01 4.28740866e-03 7.30943918e-01 4.44158882e-01
5.62280893e-01 2.14017645e-01 -6.83580577e-01 -4.66290176e-01
-1.89899325e-01 -1.30813122e-01 1.55776143e-01 5.78527600e-02
-1.10468018e+00 6.76817179e-01 5.18357706e+00 8.82924020e-01
-9.04516935e-01 1.24202706e-01 5.88914692e-01 -1.72717437e-01
-4.04037595e-01 -3.48816812e-01 -3.96181643e-01 9.08554554e-01
1.23302639e+00 -3.40507597e-01 7.33028203e-02 7.38070190e-01
3.08227479e-01 5.16067445e-01 -1.44962871e+00 8.63053322e-01
-2.97448728e-02 -9.78898644e-01 1.57689705e-01 1.89785317e-01
6.30095422e-01 9.95299444e-02 2.06360623e-01 4.32852715e-01
-4.84903865e-02 -1.07169807e+00 5.61301708e-01 4.46236819e-01
1.97159231e-01 -1.01737773e+00 7.71721065e-01 4.45419759e-01
-8.87101233e-01 -3.66452307e-01 -3.10009152e-01 -3.05155385e-03
3.05543214e-01 6.05463386e-01 -7.71391809e-01 8.40980709e-01
5.13450444e-01 1.19828057e+00 -7.49739349e-01 1.15090084e+00
4.35751975e-02 9.21622992e-01 -3.78047913e-01 8.99275467e-02
7.37446249e-01 -5.73380768e-01 1.21097863e+00 1.24993205e+00
7.89829016e-01 -2.82658964e-01 3.94137017e-02 9.27619338e-01
1.87140680e-03 6.74867723e-03 -9.77856398e-01 -3.05052340e-01
1.58388034e-01 9.17509556e-01 -6.67592227e-01 -2.23224461e-01
-2.38484472e-01 1.20245028e+00 3.11124325e-01 3.97679150e-01
-9.12280917e-01 -2.09890187e-01 1.05791771e+00 1.58724621e-01
3.80209804e-01 -1.39947593e-01 1.05357431e-01 -1.24558949e+00
1.96283713e-01 -1.03648436e+00 5.08913040e-01 -2.20486283e-01
-1.38673627e+00 5.31124949e-01 -1.82966918e-01 -1.71862185e+00
-7.03637600e-01 -4.21965837e-01 -9.92001712e-01 3.74788553e-01
-9.81885314e-01 -7.96654820e-01 -1.96729600e-01 4.42294568e-01
5.92674434e-01 -3.78347129e-01 5.13496637e-01 5.46987414e-01
-8.68786752e-01 7.94138849e-01 3.40530932e-01 2.24711016e-01
3.54837358e-01 -1.43991292e+00 1.02676785e+00 1.25005615e+00
3.02269876e-01 3.68818402e-01 7.77599692e-01 -6.18814349e-01
-8.55976939e-01 -1.61081433e+00 5.33097029e-01 -6.14201903e-01
8.88800919e-01 -4.55727369e-01 -1.43709993e+00 1.04360020e+00
-8.62598047e-02 -3.71721089e-02 5.10188282e-01 7.54178241e-02
-6.53858781e-02 2.28107482e-01 -8.73909235e-01 8.62742126e-01
1.60455108e+00 -3.41987967e-01 -2.75910050e-01 4.56696123e-01
8.34783614e-01 -2.50874668e-01 -6.61576390e-01 3.03815305e-01
1.26371440e-02 -1.33987725e+00 7.79610932e-01 -1.03293872e+00
3.16964477e-01 -4.06068057e-01 5.62365837e-02 -1.61729622e+00
-1.25255167e-01 -6.73316836e-01 -5.68870842e-01 1.12040484e+00
1.89733282e-01 -8.68150532e-01 7.31329501e-01 1.12888813e-02
-6.37906075e-01 -7.03296244e-01 -9.39222753e-01 -1.16122687e+00
-5.27574262e-03 -7.60357976e-01 8.89803350e-01 9.23632145e-01
3.94494459e-02 -8.17392096e-02 -4.82440233e-01 2.43240133e-01
2.35565096e-01 -9.02028754e-02 7.87876964e-01 -1.16927791e+00
-2.63319939e-01 -7.39071190e-01 -1.00387800e+00 -9.79711652e-01
1.16821624e-01 -1.07729173e+00 -2.79067695e-01 -1.03651488e+00
-4.75976616e-01 -3.41351807e-01 -3.02991509e-01 1.36269495e-01
-1.96304768e-01 1.07448265e-01 -1.11156069e-02 9.80185121e-02
-8.16784441e-01 6.95132852e-01 6.14484251e-01 6.91202134e-02
-2.33021915e-01 3.12714785e-01 -2.19910964e-01 6.02063417e-01
9.20704007e-01 -3.14724505e-01 -3.87966752e-01 -9.79730412e-02
4.38450903e-01 -4.34376597e-01 4.99898225e-01 -1.35802901e+00
3.74137536e-02 2.88148612e-01 -1.60377264e-01 -7.03837931e-01
5.27035743e-02 -7.51770914e-01 3.93089047e-03 4.13810492e-01
-3.67236733e-01 5.23806095e-01 5.25579266e-02 9.62387562e-01
-3.85457724e-01 -1.75949737e-01 2.88460970e-01 6.92152828e-02
-6.47486448e-01 5.52525640e-01 -7.51750410e-01 1.25904322e-01
1.12516296e+00 -2.37907067e-01 -1.10346884e-01 -7.37097800e-01
-7.66798139e-01 3.48258495e-01 4.39698160e-01 4.93414342e-01
4.91339713e-01 -1.54522502e+00 -7.63827682e-01 3.50836247e-01
3.72513115e-01 8.63319263e-02 5.04667282e-01 1.27664900e+00
-4.71865743e-01 4.06963617e-01 -9.84057486e-02 -8.24474931e-01
-9.24316704e-01 6.96496487e-01 3.93507689e-01 -4.12533015e-01
-8.65421534e-01 3.93530130e-01 5.96516095e-02 -4.05296743e-01
2.60149151e-01 -6.77516758e-01 -1.88041419e-01 7.96155035e-02
5.79382896e-01 7.19018042e-01 3.30185384e-01 -4.79625583e-01
-5.81485704e-02 1.38541758e-01 -1.06183067e-01 2.75574684e-01
1.12929082e+00 -5.84180877e-02 -1.19179729e-02 7.87294924e-01
1.12611842e+00 -2.55078733e-01 -9.67039704e-01 -1.98310092e-02
5.49103796e-01 -2.97863334e-01 -6.46950185e-01 -1.89904377e-01
-8.49001348e-01 9.06208098e-01 3.49371701e-01 7.80467689e-01
9.40321803e-01 -1.32840872e-01 8.83087516e-01 4.64944184e-01
1.30914673e-01 -7.39044130e-01 2.00299859e-01 7.27133334e-01
8.24160874e-01 -1.03534508e+00 -5.72466612e-01 -2.22031269e-02
-6.04766667e-01 9.68684733e-01 4.54767078e-01 -3.48839343e-01
3.09100181e-01 -2.72815764e-01 -3.67344052e-01 -1.99865416e-01
-6.73569500e-01 -3.20151597e-01 3.29287320e-01 4.97047067e-01
9.99215767e-02 -1.58935532e-01 -2.23120481e-01 5.57933450e-01
-1.22870982e-01 -4.26017463e-01 6.14410281e-01 4.80609298e-01
-3.54399651e-01 -1.02221799e+00 -1.45725176e-01 7.58937597e-01
-5.29740751e-01 1.67843640e-01 -2.48703748e-01 7.45835483e-01
6.48132861e-02 6.48507655e-01 5.92846036e-01 -4.02118146e-01
3.88714701e-01 3.68116289e-01 3.10535077e-03 -5.26025355e-01
-7.62137175e-01 -5.07069170e-01 9.12024304e-02 -6.69710696e-01
-8.33773538e-02 -7.49815822e-01 -1.04226303e+00 -2.80973732e-01
1.14938259e-01 2.28006482e-01 1.45600125e-01 1.13784778e+00
5.98557472e-01 6.58319712e-01 4.90469456e-01 -6.01388693e-01
-4.64165956e-01 -1.15590024e+00 -6.34028912e-01 9.65905547e-01
6.01577163e-01 -4.77574408e-01 -6.41170621e-01 -3.35393727e-01]
|
[7.496887683868408, 2.4433155059814453]
|
158e92f0-6123-4963-afed-b57bbee0567b
|
spiking-fer-spiking-neural-network-for-facial
|
2304.10211
| null |
https://arxiv.org/abs/2304.10211v1
|
https://arxiv.org/pdf/2304.10211v1.pdf
|
Spiking-Fer: Spiking Neural Network for Facial Expression Recognition With Event Cameras
|
Facial Expression Recognition (FER) is an active research domain that has shown great progress recently, notably thanks to the use of large deep learning models. However, such approaches are particularly energy intensive, which makes their deployment difficult for edge devices. To address this issue, Spiking Neural Networks (SNNs) coupled with event cameras are a promising alternative, capable of processing sparse and asynchronous events with lower energy consumption. In this paper, we establish the first use of event cameras for FER, named "Event-based FER", and propose the first related benchmarks by converting popular video FER datasets to event streams. To deal with this new task, we propose "Spiking-FER", a deep convolutional SNN model, and compare it against a similar Artificial Neural Network (ANN). Experiments show that the proposed approach achieves comparable performance to the ANN architecture, while consuming less energy by orders of magnitude (up to 65.39x). In addition, an experimental study of various event-based data augmentation techniques is performed to provide insights into the efficient transformations specific to event-based FER.
|
['Chaabane Djéraba', 'José Mennesson', 'Amel Aissaoui', 'Benjamin Allaert', 'Sami Barchid']
|
2023-04-20
| null | null | null | null |
['facial-expression-recognition']
|
['computer-vision']
|
[ 4.92759913e-01 -2.69495577e-01 6.11150935e-02 -3.44495714e-01
-2.06993267e-01 -5.72512411e-02 5.60249388e-01 8.79141465e-02
-6.73973739e-01 7.62662828e-01 -2.65032589e-01 1.45896778e-01
8.76265690e-02 -8.47448230e-01 -9.38832343e-01 -7.98804700e-01
-2.14889999e-02 -1.84856519e-01 2.14541912e-01 -6.03874736e-02
-2.70191988e-04 8.33128870e-01 -1.84132564e+00 3.05816382e-01
3.95927817e-01 1.44771254e+00 -1.47752851e-01 1.97203115e-01
-1.78067684e-01 1.14549291e+00 -6.46510720e-01 -6.09895825e-01
-7.45598301e-02 -4.85228688e-01 -3.43261659e-01 -2.02270716e-01
9.74618271e-02 -2.07998157e-01 -5.72711706e-01 1.05726767e+00
5.86891651e-01 1.35311306e-01 1.48840398e-01 -1.73792040e+00
-1.15576491e-01 4.83159423e-01 -5.64803421e-01 3.88177276e-01
1.92210197e-01 -1.68987498e-01 3.32476050e-01 -8.01229715e-01
5.70578814e-01 9.93233383e-01 8.13749731e-01 8.10370922e-01
-1.03767574e+00 -1.03534269e+00 -6.21110685e-02 5.97026110e-01
-1.43779039e+00 -8.58869076e-01 9.38683152e-01 1.27624407e-01
1.24694800e+00 -9.13234614e-03 9.46792066e-01 1.69189858e+00
2.61646301e-01 8.01577210e-01 1.06725168e+00 -2.71654606e-01
7.27881789e-01 -2.37859219e-01 9.15751681e-02 5.47883630e-01
4.60008472e-01 -1.25787377e-01 -1.09465551e+00 -1.00821815e-01
5.94893575e-01 2.86119103e-01 -1.76872849e-01 8.84508416e-02
-7.71469235e-01 4.46094692e-01 3.50655705e-01 3.40096951e-01
-5.85547805e-01 7.07041562e-01 6.89436257e-01 9.42885950e-02
3.09153020e-01 -1.55200019e-01 -1.02342717e-01 -4.50129986e-01
-9.62650001e-01 5.77514321e-02 7.53948092e-01 7.66200483e-01
4.70436215e-01 5.18513322e-01 -4.27620001e-02 5.56843817e-01
7.92100281e-02 2.59352893e-01 5.52903533e-01 -8.12430859e-01
9.33383256e-02 6.51648998e-01 -2.67502367e-01 -7.72355556e-01
-4.20415699e-01 5.68884984e-02 -1.28066599e+00 2.77287662e-01
2.98362941e-01 -1.51516438e-01 -7.93067098e-01 1.86662519e+00
1.04267649e-01 7.62435615e-01 1.14138134e-01 5.88182449e-01
8.11893046e-01 7.81353176e-01 4.27655101e-01 -4.85587865e-01
1.43372750e+00 -6.51167810e-01 -9.23051059e-01 -6.02968447e-02
1.94169596e-01 -3.39693367e-01 7.20131576e-01 5.38731635e-01
-1.01549351e+00 -3.14452499e-01 -1.05614936e+00 9.88529250e-02
-5.09123504e-01 3.13483834e-01 9.03102875e-01 8.41287494e-01
-1.11517727e+00 4.75268960e-01 -1.36590946e+00 -6.78019881e-01
8.79407525e-01 6.59870088e-01 -2.75328934e-01 3.16339880e-01
-1.00203073e+00 5.33824384e-01 2.48621792e-01 1.37891546e-01
-8.25142384e-01 -5.27561367e-01 -7.69792974e-01 3.13849241e-01
-5.30544855e-02 -3.10111344e-01 1.15238476e+00 -1.43246698e+00
-1.87191451e+00 7.56987810e-01 -3.69023204e-01 -8.43990743e-01
1.39479324e-01 -1.62920654e-02 -6.18434489e-01 2.65219390e-01
-5.55186212e-01 7.07723141e-01 7.21880436e-01 -6.60073578e-01
-4.58856702e-01 -4.46316689e-01 1.12955406e-01 -2.78910547e-01
-9.23349202e-01 5.86700320e-01 -2.44181380e-01 -7.15649009e-01
-4.47469294e-01 -8.54925573e-01 -1.16636194e-01 3.60183001e-01
2.64805127e-02 -3.31132233e-01 1.04036188e+00 -2.00969875e-01
1.30462337e+00 -2.31388950e+00 -1.58305541e-01 -3.07840053e-02
1.56387359e-01 2.48933092e-01 5.64458147e-02 3.44961286e-02
-1.68324798e-01 -1.16856068e-01 -1.94449246e-01 -4.99954164e-01
-1.97691515e-01 3.80249858e-01 -1.73693255e-01 4.33491439e-01
5.95739424e-01 7.61188567e-01 -5.87465346e-01 -3.51339996e-01
9.27741528e-02 8.03192675e-01 -3.86953771e-01 -1.04342692e-01
2.39807926e-02 3.08592767e-01 -2.43649393e-01 8.50611269e-01
5.95511913e-01 -3.51309359e-01 1.97648585e-01 -5.34196913e-01
-1.44985482e-01 -2.10030869e-01 -1.06524563e+00 1.69941187e+00
-3.08384567e-01 1.00300097e+00 -1.69906229e-01 -1.10165966e+00
1.06756270e+00 4.13603932e-01 7.10754693e-01 -9.23803091e-01
6.84365690e-01 2.66561985e-01 -2.47392327e-01 -2.99934059e-01
3.64797592e-01 8.15579966e-02 5.76788522e-02 1.30590409e-01
4.14184570e-01 6.97988510e-01 3.25368404e-01 -1.98266268e-01
1.42972267e+00 2.56342143e-01 3.47843587e-01 -2.18852516e-02
2.85830796e-01 -3.62225384e-01 7.21482217e-01 2.54029155e-01
-3.02641124e-01 2.52335697e-01 4.44072247e-01 -6.70368552e-01
-6.78342342e-01 -8.20063710e-01 -4.28122990e-02 7.02681661e-01
3.25222313e-01 -4.52405870e-01 -1.02546394e+00 -3.16831946e-01
-5.13558269e-01 4.70546633e-01 -5.52137911e-01 -3.25500548e-01
-6.66428328e-01 -1.18474615e+00 1.07898962e+00 8.36552620e-01
9.93687093e-01 -1.18065166e+00 -1.38068545e+00 3.54503959e-01
-1.48016512e-01 -1.52259243e+00 2.74518311e-01 4.98092979e-01
-8.10804665e-01 -8.02481353e-01 -6.88243151e-01 -7.37064183e-01
5.52051783e-01 -2.34007344e-01 8.84534299e-01 -1.76911160e-01
-4.71222430e-01 2.01471150e-01 -4.22186315e-01 -6.99453533e-01
-2.58945432e-02 -1.38854995e-01 9.85568687e-02 6.38528287e-01
8.01275849e-01 -9.34352338e-01 -5.43765247e-01 1.89057849e-02
-1.19743216e+00 -5.87455230e-03 5.41002572e-01 6.15706325e-01
7.93015778e-01 -1.94669157e-01 5.66056430e-01 -5.46311736e-01
3.42041999e-01 -5.11601567e-01 -6.76307380e-01 2.59430885e-01
-3.15285891e-01 -5.08909971e-02 8.16109359e-01 -6.66913152e-01
-1.26017404e+00 4.65401500e-01 -2.67903864e-01 -6.94997251e-01
-2.17863053e-01 1.80617481e-01 -2.47814730e-01 -4.80533510e-01
6.62638843e-01 1.88283220e-01 -2.24978656e-01 -1.80375680e-01
-2.47629046e-01 5.02648950e-01 5.66927254e-01 -3.96020353e-01
1.39416561e-01 7.73983777e-01 2.37697765e-01 -7.45355725e-01
-2.86321729e-01 1.80651620e-02 -1.40582815e-01 -4.19481069e-01
7.68082321e-01 -9.97072339e-01 -8.56550455e-01 9.35555518e-01
-1.47183490e+00 -3.65616471e-01 -4.84030098e-01 4.30962056e-01
-5.11611760e-01 -2.88431626e-02 -8.00067365e-01 -7.78185248e-01
-5.07709265e-01 -9.37462032e-01 1.09673631e+00 7.37340808e-01
-4.76728864e-02 -5.57971716e-01 -8.67649764e-02 -2.22244263e-01
5.77598214e-01 7.73236096e-01 4.33330983e-01 -6.46572113e-01
-5.09584427e-01 -3.77802253e-02 -2.46643245e-01 1.99997604e-01
-5.46890013e-02 1.79241583e-01 -1.34116602e+00 -6.95330352e-02
1.41833633e-01 -4.54474062e-01 7.26187646e-01 2.54438400e-01
1.69058752e+00 4.54209708e-02 -4.84893858e-01 8.01052272e-01
1.65272343e+00 4.15934205e-01 1.08980954e+00 1.43112540e-01
2.20891684e-01 1.83984805e-02 1.85654387e-01 9.44605649e-01
2.30774015e-01 6.03590012e-01 4.87209886e-01 -1.65158957e-01
-1.99612319e-01 4.10274677e-02 6.82711959e-01 6.60107493e-01
-4.87408519e-01 -5.81494510e-01 -6.90671742e-01 3.55533510e-01
-1.76286662e+00 -9.33779955e-01 -7.07225204e-02 1.83781958e+00
5.57507515e-01 -5.81920035e-02 -1.02082096e-01 4.15633112e-01
6.67204320e-01 8.81548300e-02 -6.94233000e-01 -4.76058960e-01
-4.38014090e-01 1.07264447e+00 3.41230720e-01 -5.22860110e-01
-9.87288117e-01 6.42386436e-01 5.58023787e+00 9.81790423e-01
-1.62930405e+00 2.60334373e-01 7.33042121e-01 -2.44131148e-01
1.62110955e-01 -2.96389669e-01 -9.13456798e-01 7.26269007e-01
1.56909084e+00 -1.66756496e-01 3.43254477e-01 6.66636229e-01
1.59389123e-01 -1.84914231e-01 -1.10956931e+00 1.64063883e+00
2.01071993e-01 -1.41959190e+00 1.43006012e-01 -1.95905373e-01
5.39932966e-01 -1.19157687e-01 -3.37318301e-01 1.83783814e-01
-3.20752561e-02 -9.26039875e-01 7.20212340e-01 5.93091965e-01
9.24506962e-01 -9.29462373e-01 7.31760502e-01 -8.51191133e-02
-1.51694715e+00 -7.34276697e-02 -4.14583504e-01 1.58704408e-02
2.23228514e-01 5.81480861e-01 -1.91225260e-02 3.59736383e-01
1.21199644e+00 7.68367589e-01 -3.66943866e-01 1.04463792e+00
1.42843917e-01 7.27274418e-01 -6.77706361e-01 -1.75320357e-01
-1.46471024e-01 1.84422210e-01 2.22018927e-01 1.20008671e+00
7.97210276e-01 1.80648103e-01 -3.92480463e-01 8.34466755e-01
-7.08004832e-01 -6.19826354e-02 -4.90389585e-01 -1.70864910e-01
3.72892886e-01 1.52003443e+00 -9.87521529e-01 -2.16212705e-01
-6.18284047e-01 1.39120078e+00 1.94180906e-01 1.87113062e-01
-1.45473325e+00 -5.64288318e-01 5.69106698e-01 -1.13112271e-01
1.28351972e-01 9.83221736e-03 4.50205021e-02 -1.08480203e+00
1.47732675e-01 -6.08983040e-01 1.94290832e-01 -8.46815109e-01
-9.39135313e-01 7.81210184e-01 -2.87627459e-01 -1.15499389e+00
-1.37321828e-02 -6.60279453e-01 -5.20902574e-01 1.97806403e-01
-1.80910420e+00 -1.08604097e+00 -8.78943503e-01 9.87064898e-01
4.55531180e-01 -9.44320112e-02 1.01404142e+00 8.73682916e-01
-8.95628750e-01 8.43484223e-01 -4.40088697e-02 1.12101331e-01
6.00748360e-01 -6.19364798e-01 1.41666502e-01 8.21140409e-01
2.33857378e-01 4.71233316e-02 2.50625432e-01 -2.20411241e-01
-1.67800760e+00 -1.42100036e+00 6.29056334e-01 8.32979903e-02
5.57618797e-01 -3.98027688e-01 -6.91974103e-01 6.26939595e-01
3.62114012e-01 5.73208213e-01 7.02712715e-01 -6.76209807e-01
-8.15957412e-02 -5.17582357e-01 -1.27255416e+00 5.32365978e-01
1.33118916e+00 -3.27760756e-01 2.51965318e-02 2.38809995e-02
3.73877674e-01 -3.47345650e-01 -7.40867734e-01 5.57550788e-01
4.43035930e-01 -9.87668395e-01 6.49824739e-01 -2.66874939e-01
3.64781559e-01 -2.84737796e-01 -1.42098844e-01 -8.91807675e-01
1.64170593e-01 -7.30551839e-01 -5.63440442e-01 1.39402783e+00
-5.75502738e-02 -3.83584023e-01 1.03899693e+00 2.55076647e-01
1.78335905e-02 -7.42665768e-01 -1.19859910e+00 -8.55266929e-01
-6.21779621e-01 -3.92569184e-01 4.96299088e-01 8.01692367e-01
-3.39498878e-01 -8.19719583e-02 -2.38550231e-01 4.90936358e-03
4.20346141e-01 -1.86918318e-01 4.03043956e-01 -1.31610394e+00
8.27838331e-02 -2.86310434e-01 -9.27919030e-01 -7.16343164e-01
2.89476752e-01 -7.10603774e-01 -2.98758298e-02 -9.70514953e-01
1.87245920e-01 -7.03502074e-03 -5.09710193e-01 7.14338601e-01
3.25342208e-01 6.43362999e-01 -2.66237091e-02 -4.35294025e-02
-7.45150328e-01 7.16177642e-01 3.05869520e-01 5.82466386e-02
1.03250258e-02 -4.40497607e-01 -3.02804559e-01 1.02213788e+00
7.90149033e-01 -5.55359602e-01 -1.78782374e-01 -3.85542899e-01
2.45742589e-01 -1.57542303e-01 5.85557401e-01 -1.64185560e+00
7.66478777e-01 9.65271145e-02 5.20022929e-01 -1.45833090e-01
4.52692270e-01 -1.04324591e+00 5.82926393e-01 3.18791360e-01
6.03829995e-02 1.79514840e-01 6.75063848e-01 5.67879677e-01
-5.21479249e-01 1.75563917e-02 9.01494265e-01 2.95334011e-01
-1.03880906e+00 3.71645808e-01 -7.40655780e-01 -1.74549505e-01
1.41692173e+00 -4.61886525e-01 -1.70142382e-01 -1.55026503e-02
-5.97415745e-01 -3.27306062e-01 1.10012926e-01 1.59928665e-01
6.58335030e-01 -1.40844524e+00 -3.03048313e-01 2.26065889e-01
1.31240293e-01 -2.47105852e-01 2.24700660e-01 8.99884880e-01
-4.74648744e-01 6.95342422e-02 -7.34302759e-01 -5.82361042e-01
-1.33403361e+00 1.79415107e-01 4.09983903e-01 -2.93530613e-01
-3.58010739e-01 6.87782407e-01 -2.08215743e-01 2.37501681e-01
3.76173526e-01 -4.33351427e-01 -1.66889310e-01 1.19901992e-01
6.10303164e-01 5.17434657e-01 4.28452700e-01 -2.98957944e-01
-4.57920909e-01 3.97941440e-01 4.14460927e-01 2.79502094e-01
1.59611189e+00 1.41506031e-01 -1.69553921e-01 2.36372173e-01
9.77557361e-01 -3.46452713e-01 -1.20802045e+00 -4.27219979e-02
9.77850258e-02 9.98491347e-02 5.67368716e-02 -4.76365685e-01
-1.56153941e+00 9.97489691e-01 1.07663095e+00 -1.20944433e-01
1.87312698e+00 -2.90730864e-01 1.02366507e+00 4.70804304e-01
6.79146886e-01 -9.45501745e-01 1.22191958e-01 2.49830529e-01
5.08950651e-01 -9.53609347e-01 -4.45260584e-01 -3.69914919e-01
-1.05746306e-01 1.47220242e+00 7.47398138e-01 -1.86525583e-01
6.64136767e-01 8.51597786e-01 -4.92566407e-01 -2.15296909e-01
-7.18095303e-01 -9.10967514e-02 -3.94051582e-01 2.54340321e-01
2.06819490e-01 -2.90305167e-01 -2.66514003e-01 9.93254364e-01
3.16042423e-01 9.81246591e-01 3.75014305e-01 1.14545071e+00
4.35797833e-02 -1.06876361e+00 -1.02003619e-01 4.58838046e-01
-7.58507431e-01 -2.49841399e-02 -2.34998688e-01 6.71417356e-01
2.80119330e-01 6.00175858e-01 4.37192142e-01 -2.97950089e-01
4.59827155e-01 2.32687905e-01 5.93765318e-01 -1.74916368e-02
-6.86119199e-01 -2.53514111e-01 -5.13098389e-02 -8.95183384e-01
-1.02683747e+00 -3.82306218e-01 -1.45336068e+00 -4.70352799e-01
-1.22167930e-01 -3.13623101e-01 7.50265241e-01 7.95859635e-01
6.95326447e-01 6.99582636e-01 3.50309879e-01 -8.52157831e-01
1.54604227e-03 -5.18096507e-01 -4.68553960e-01 3.27844530e-01
-6.96569532e-02 -6.43338084e-01 -2.25669503e-01 4.33230489e-01]
|
[8.253815650939941, 2.425004720687866]
|
99f26656-2d98-47d8-b99f-dbc052083bf3
|
discriminatively-trained-sparse-code
| null | null |
http://papers.nips.cc/paper/4787-discriminatively-trained-sparse-code-gradients-for-contour-detection
|
http://papers.nips.cc/paper/4787-discriminatively-trained-sparse-code-gradients-for-contour-detection.pdf
|
Discriminatively Trained Sparse Code Gradients for Contour Detection
|
Finding contours in natural images is a fundamental problem that serves as the basis of many tasks such as image segmentation and object recognition. At the core of contour detection technologies are a set of hand-designed gradient features, used by most existing approaches including the state-of-the-art Global Pb (gPb) operator. In this work, we show that contour detection accuracy can be significantly improved by computing Sparse Code Gradients (SCG), which measure contrast using patch representations automatically learned through sparse coding. We use K-SVD and Orthogonal Matching Pursuit for efficient dictionary learning and encoding, and use multi-scale pooling and power transforms to code oriented local neighborhoods before computing gradients and applying linear SVM. By extracting rich representations from pixels and avoiding collapsing them prematurely, Sparse Code Gradients effectively learn how to measure local contrasts and find contours. We improve the F-measure metric on the BSDS500 benchmark to 0.74 (up from 0.71 of gPb contours). Moreover, our learning approach can easily adapt to novel sensor data such as Kinect-style RGB-D cameras: Sparse Code Gradients on depth images and surface normals lead to promising contour detection using depth and depth+color, as verified on the NYU Depth Dataset. Our work combines the concept of oriented gradients with sparse representation and opens up future possibilities for learning contour detection and segmentation.
|
['Ren Xiaofeng', 'Liefeng Bo']
|
2012-12-01
| null | null | null |
neurips-2012-12
|
['contour-detection']
|
['computer-vision']
|
[ 3.41064811e-01 -2.65151083e-01 -4.04753983e-01 -2.38565743e-01
-7.01761603e-01 -6.30233347e-01 3.08740973e-01 5.25983512e-01
-5.23715198e-01 2.71777868e-01 -1.16870932e-01 -4.33321223e-02
3.66408825e-02 -1.06733930e+00 -6.45811796e-01 -7.75377095e-01
-3.76477391e-01 2.67747253e-01 6.62553549e-01 -2.15149760e-01
5.57505846e-01 8.01335633e-01 -1.75570321e+00 2.38933697e-01
6.56608880e-01 1.49792814e+00 7.16279745e-02 6.54030800e-01
-3.66189271e-01 4.40315425e-01 -1.44982979e-01 -2.25894794e-01
2.97796011e-01 -3.42192203e-01 -4.63028163e-01 2.74871439e-01
7.62372613e-01 -1.28295168e-01 2.98558436e-02 1.22427213e+00
4.85728025e-01 -1.06832527e-01 6.57514989e-01 -7.59248197e-01
-5.76057971e-01 2.31333263e-02 -8.58252227e-01 2.20200062e-01
2.48864502e-01 -2.48517275e-01 1.03271377e+00 -1.02663052e+00
7.31857717e-01 8.45721483e-01 8.47707391e-01 3.59711260e-01
-1.19912291e+00 -2.14982465e-01 -1.18715465e-02 1.54617950e-01
-1.35802817e+00 1.33226179e-02 1.06423211e+00 -6.06183350e-01
7.90720940e-01 3.91439229e-01 8.46723437e-01 3.84705245e-01
1.59656614e-01 9.13639426e-01 1.15713477e+00 -5.92476964e-01
5.54149091e-01 -2.01612059e-02 9.94779915e-02 1.11360109e+00
2.91542679e-01 1.25885040e-01 -6.42805696e-01 9.22276080e-02
1.02841270e+00 1.06817875e-02 -5.86004794e-01 -7.77179658e-01
-1.02601278e+00 1.16648138e+00 6.84872925e-01 3.71915430e-01
-2.98697889e-01 -4.41347063e-02 1.28709853e-01 2.46427059e-02
3.46227884e-01 3.19033638e-02 -4.08677429e-01 2.66652033e-02
-1.06257975e+00 6.92083165e-02 6.05696142e-01 5.99612713e-01
1.18770528e+00 -9.55046043e-02 1.96709707e-02 9.98144686e-01
1.65421486e-01 6.94671094e-01 7.68224776e-01 -9.59649980e-01
4.35349196e-02 8.19902778e-01 -3.52517843e-01 -1.37689769e+00
-5.42478859e-01 -7.79112354e-02 -8.69477808e-01 6.14342153e-01
6.97872758e-01 2.03290761e-01 -1.08213043e+00 1.06241381e+00
4.00505096e-01 1.16848633e-01 -1.78165182e-01 1.10327220e+00
6.04830861e-01 5.39857686e-01 -3.15428615e-01 2.89443647e-04
1.27180600e+00 -6.56608999e-01 -3.19103599e-01 -3.09893876e-01
5.01477122e-01 -7.23317862e-01 9.83110428e-01 6.18848801e-01
-9.26874161e-01 -4.18759495e-01 -1.24282014e+00 -1.72048464e-01
-4.96799678e-01 1.32085204e-01 1.02017856e+00 8.31411123e-01
-1.03495467e+00 7.77612031e-01 -9.14113164e-01 -4.61308360e-02
6.38545990e-01 2.46573955e-01 -3.35411340e-01 -3.64810228e-01
-7.05888152e-01 5.05918443e-01 1.64781556e-01 1.41329430e-02
-4.21757758e-01 -4.13470984e-01 -1.18766272e+00 -4.86780368e-02
6.65698946e-02 -1.04188295e-02 4.38438863e-01 -8.24464321e-01
-1.45058000e+00 1.12989712e+00 -8.81567001e-02 -4.57747400e-01
1.97759807e-01 -3.98020782e-02 2.82757785e-02 6.39168203e-01
1.59039497e-02 8.41275871e-01 1.04745281e+00 -1.14398038e+00
-6.17307961e-01 -4.97178465e-01 -4.58277285e-01 -3.89805958e-02
-5.49058795e-01 -2.72105545e-01 -4.81184512e-01 -7.71460295e-01
7.39757359e-01 -6.81416929e-01 -3.11513036e-01 4.39040512e-01
-1.17725711e-02 -6.29819855e-02 7.47657597e-01 -7.69322276e-01
8.95596921e-01 -2.06335521e+00 2.43394330e-01 7.44945288e-01
-4.08292690e-04 6.82540387e-02 -2.03723032e-02 -1.86048850e-01
1.00079156e-01 -1.49488747e-01 -8.32809865e-01 -2.19464660e-01
-3.14836055e-01 2.37071961e-01 -8.47499967e-02 6.10792696e-01
4.70727086e-01 8.50015342e-01 -6.39624715e-01 -6.39080286e-01
3.12130660e-01 5.51963091e-01 -5.64593971e-01 -1.31785288e-01
-8.67662728e-02 6.79265037e-02 -1.70554608e-01 1.01998317e+00
7.83241570e-01 -1.84905887e-01 -1.53057292e-01 -1.27657071e-01
-2.25723490e-01 -1.26661703e-01 -1.42542160e+00 2.04297280e+00
-2.38171205e-01 7.90087283e-01 1.17932200e-01 -1.28468359e+00
1.34862339e+00 -1.26496375e-01 5.75559497e-01 -9.37194407e-01
1.38806507e-01 4.35849220e-01 -4.39111710e-01 -1.55660480e-01
1.47065163e-01 -1.18067525e-01 1.28582552e-01 7.78448060e-02
1.95226967e-01 -5.22884488e-01 2.27454498e-01 -2.19419263e-02
9.31112945e-01 -5.60293719e-02 1.77052319e-01 -4.27927613e-01
6.37773693e-01 1.68048561e-01 5.04634261e-01 4.15973574e-01
-1.11549407e-01 1.05481851e+00 4.45540518e-01 -5.15732050e-01
-7.01884389e-01 -9.39238369e-01 -4.86235797e-01 8.03879917e-01
3.40501368e-01 -1.89768851e-01 -6.23536825e-01 -4.90645140e-01
2.08991647e-01 -5.71246631e-02 -6.30764067e-01 9.41373408e-02
-7.03183889e-01 -7.50518680e-01 1.45607263e-01 6.56277537e-01
4.67524290e-01 -1.08457351e+00 -9.94363070e-01 2.97291607e-01
2.20187441e-01 -9.67805982e-01 -3.19760501e-01 5.06385982e-01
-1.20228028e+00 -9.26916718e-01 -1.08913851e+00 -1.12155294e+00
6.52905345e-01 1.26250803e-01 8.89074147e-01 1.29681334e-01
-8.72985303e-01 4.66648668e-01 -4.76829410e-01 7.13701770e-02
1.51357993e-01 -1.77719012e-01 -5.11729062e-01 1.85908675e-01
8.08354393e-02 -4.66005266e-01 -6.81777537e-01 1.43031523e-01
-7.91707039e-01 -2.44907692e-01 7.15780079e-01 8.58094573e-01
9.77296233e-01 -4.53766026e-02 4.72330749e-02 -6.75214529e-01
2.04077721e-01 -2.09535751e-02 -8.88635635e-01 1.69522643e-01
-5.09994805e-01 8.01184475e-02 2.16065198e-01 -3.63163739e-01
-5.65983176e-01 4.31796551e-01 -2.65442848e-01 -3.88988703e-01
-2.61823386e-02 5.20840168e-01 1.26076937e-01 -5.76514900e-01
7.15373039e-01 4.81171161e-01 3.51371504e-02 -3.64969432e-01
3.58308911e-01 3.62601608e-01 5.90645552e-01 -5.68648040e-01
5.18596590e-01 8.49485159e-01 1.41736522e-01 -1.14722931e+00
-4.55013037e-01 -7.72536218e-01 -6.41227782e-01 -1.43862754e-01
8.63540888e-01 -8.68031383e-01 -6.22496068e-01 5.70343435e-01
-9.33129191e-01 -4.72297162e-01 -2.99057215e-01 1.33327425e-01
-5.47855377e-01 5.30547678e-01 -7.42766082e-01 -7.74437368e-01
-3.36681992e-01 -9.10852611e-01 1.27925479e+00 4.66043442e-01
1.71166778e-01 -9.88840163e-01 9.88416076e-02 1.37145773e-01
4.68258142e-01 5.10010839e-01 7.16339469e-01 1.68038368e-01
-5.18829465e-01 -2.61759639e-01 -1.95025042e-01 4.81793880e-01
4.58183549e-02 -1.09967217e-01 -9.09034312e-01 -2.66790956e-01
-4.39023674e-02 -4.02864873e-01 1.32823431e+00 5.65862954e-01
1.10965967e+00 4.02704924e-02 -2.06920013e-01 8.58167946e-01
1.76142097e+00 -2.33508553e-02 6.63500845e-01 2.46576071e-01
7.90840089e-01 6.86731815e-01 4.82680172e-01 4.66610879e-01
1.29379451e-01 4.88734961e-01 2.28185177e-01 -2.60724068e-01
-2.62406886e-01 -4.13505733e-02 1.67821318e-01 4.91056800e-01
-3.59622464e-02 4.88106549e-01 -9.47429657e-01 6.74466670e-01
-1.57523644e+00 -5.90901673e-01 -1.82240203e-01 2.09994841e+00
1.00744188e+00 3.02779496e-01 -1.08756684e-03 5.80723166e-01
4.70762163e-01 -8.31086375e-03 -3.34780186e-01 -3.30648929e-01
-4.61660564e-01 8.85761559e-01 6.89354658e-01 6.84275448e-01
-1.33641756e+00 1.01829672e+00 5.33852816e+00 8.82205963e-01
-1.49391592e+00 -4.20939438e-02 7.94393539e-01 3.18117172e-01
-2.11227283e-01 -1.33890674e-01 -6.33887827e-01 2.29525357e-01
3.77315462e-01 3.34561080e-01 3.77603531e-01 9.69564617e-01
-3.86002660e-01 -4.04351443e-01 -6.09934330e-01 1.29394019e+00
2.36717701e-01 -1.52060688e+00 -2.80065060e-01 -7.27983117e-02
9.18767571e-01 2.03480627e-02 1.10150732e-01 -7.81411752e-02
-3.38979140e-02 -9.28510368e-01 7.75885284e-01 1.95440322e-01
7.80060530e-01 -6.23901784e-01 5.61574697e-01 1.07711807e-01
-1.43341088e+00 -1.13575302e-01 -7.42301822e-01 -1.33454020e-03
-6.57633916e-02 1.04537416e+00 -3.35197777e-01 9.52187739e-03
8.52706552e-01 8.84358644e-01 -5.33957839e-01 1.10478461e+00
-2.57641464e-01 5.28651953e-01 -6.71095133e-01 -1.28354818e-01
1.49286866e-01 -3.81023765e-01 7.69358873e-02 1.32967579e+00
1.98137432e-01 1.81380078e-01 1.29409865e-01 7.80635715e-01
9.78710875e-02 4.01056141e-01 -4.63958025e-01 2.12255433e-01
1.00391641e-01 1.36173368e+00 -1.34353340e+00 -2.38330558e-01
-4.29202467e-01 1.19483364e+00 2.71867931e-01 1.38209015e-01
-3.59946758e-01 -4.23834950e-01 5.44045746e-01 7.69998506e-02
7.47293890e-01 -4.67627406e-01 -7.75447845e-01 -1.00548434e+00
6.30508959e-02 -6.15985572e-01 2.96389461e-01 -2.97015488e-01
-1.01589942e+00 3.58412743e-01 -5.25021076e-01 -1.12916017e+00
1.97848022e-01 -9.66355145e-01 -5.28565764e-01 4.86591697e-01
-1.86024010e+00 -9.49302733e-01 -5.04874945e-01 7.98037887e-01
3.77224833e-01 1.32900327e-01 8.47996593e-01 1.32413238e-01
-1.04535319e-01 3.69559586e-01 4.62626182e-02 4.39467192e-01
2.82864660e-01 -1.43135393e+00 1.86689496e-01 7.79501796e-01
7.17480123e-01 2.47968748e-01 2.08097607e-01 -5.17435610e-01
-1.21929872e+00 -6.79041147e-01 5.12346566e-01 -1.56519916e-02
4.48523045e-01 -5.04187942e-01 -1.00785470e+00 2.03206353e-02
-2.45005891e-01 6.21562898e-01 4.76412982e-01 -2.55011290e-01
-2.85474569e-01 -1.69296458e-01 -1.10690427e+00 1.44540384e-01
8.07247102e-01 -3.97112548e-01 -3.41568083e-01 1.90404579e-01
2.70893395e-01 -4.93158281e-01 -7.97169924e-01 3.65703613e-01
4.56381768e-01 -1.32900596e+00 1.13461220e+00 2.60937680e-02
2.99190730e-01 -1.83538407e-01 -2.39886433e-01 -9.64766920e-01
-3.85853052e-02 -2.51537681e-01 -1.42820934e-02 9.01382565e-01
1.87600598e-01 -4.15169656e-01 1.29591537e+00 2.25623682e-01
-2.28759926e-03 -8.98607194e-01 -1.19538903e+00 -5.45263946e-01
1.24689676e-01 -5.26222467e-01 1.03077121e-01 8.64203215e-01
1.12477489e-01 -1.50804445e-01 1.73116103e-01 9.85779464e-02
8.74331594e-01 5.17141283e-01 3.31570596e-01 -1.37763202e+00
-2.59167135e-01 -7.21627116e-01 -8.10856819e-01 -1.30153406e+00
-2.12958492e-02 -9.26822662e-01 1.10539027e-01 -1.27736759e+00
-1.56974316e-01 -7.15400219e-01 -2.38280967e-01 5.51343620e-01
-5.29849976e-02 6.84303582e-01 1.66741163e-01 2.41708551e-02
-2.24358931e-01 5.23344040e-01 1.21105862e+00 -4.08940613e-01
-3.10750127e-01 -1.78528637e-01 -4.14631933e-01 7.39900887e-01
6.60575509e-01 -4.33673382e-01 2.05200370e-02 -1.34501442e-01
1.58353403e-01 -8.23067687e-03 3.99370790e-01 -1.26350284e+00
4.16235656e-01 8.36098120e-02 5.85429370e-01 -3.28767866e-01
3.03319126e-01 -7.55629241e-01 -6.54609203e-01 6.73105896e-01
1.02183796e-01 -2.76419640e-01 3.40174198e-01 5.38877010e-01
-5.10395885e-01 -2.30004862e-01 9.26088750e-01 -1.22648560e-01
-1.06417930e+00 9.04648155e-02 -2.00666294e-01 8.96654055e-02
9.43098545e-01 -4.16967183e-01 4.86245044e-02 -6.90091252e-02
-6.89155161e-01 -1.10175155e-01 4.80137855e-01 1.08813390e-01
1.17488551e+00 -1.19159603e+00 -5.09600341e-01 7.23964930e-01
4.80195886e-05 2.64970541e-01 -2.50190273e-02 7.14403450e-01
-1.00703824e+00 7.64738917e-02 -3.80376935e-01 -1.07549095e+00
-1.07053399e+00 2.60969728e-01 3.19011897e-01 2.60470361e-02
-7.59067416e-01 1.34026933e+00 -1.43983275e-01 -6.46967092e-05
2.57884115e-01 -6.79051399e-01 -1.67602032e-01 2.24582657e-01
5.05736709e-01 1.82004556e-01 3.65235269e-01 -4.93269265e-01
-5.02019405e-01 1.44785702e+00 2.84346074e-01 -1.14904888e-01
1.35729730e+00 2.58583695e-01 -1.36690482e-01 1.73653394e-01
1.42638087e+00 5.78170121e-02 -1.47485793e+00 -4.74706814e-02
2.75171876e-01 -5.77046216e-01 3.70882869e-01 -5.57621539e-01
-1.51521933e+00 1.12444878e+00 1.01886964e+00 -2.00773571e-02
1.23443842e+00 4.26332727e-02 8.23842406e-01 2.03713998e-01
5.86470902e-01 -1.03762388e+00 3.50882143e-01 3.33879650e-01
6.75011396e-01 -1.44913030e+00 1.78771596e-02 -6.14995897e-01
-6.31099164e-01 1.24458003e+00 6.84590712e-02 -5.22737324e-01
8.98097396e-01 4.24880147e-01 2.25844219e-01 -2.76018739e-01
-1.16771050e-02 -5.88636875e-01 3.38443011e-01 6.49639487e-01
1.36731386e-01 1.09596260e-01 -2.69682378e-01 3.42490464e-01
2.76251361e-02 -2.60411084e-01 3.58740479e-01 1.11168945e+00
-8.24896693e-01 -1.01033020e+00 -5.60591400e-01 3.22030514e-01
-1.28288761e-01 -9.86361206e-02 -1.88306808e-01 5.15448332e-01
2.08927691e-01 6.52039707e-01 1.88199252e-01 -1.45954967e-01
2.46888205e-01 -7.78130721e-03 5.79810798e-01 -4.71237004e-01
-2.64576048e-01 2.00853050e-01 -5.67516565e-01 -7.25349844e-01
-4.83854651e-01 -7.50490129e-01 -1.59610665e+00 2.07267597e-01
-2.66243398e-01 -1.12223104e-01 8.36108804e-01 6.14979029e-01
4.76043411e-02 9.49079171e-02 4.82388139e-01 -1.19122839e+00
-2.02763408e-01 -4.77999717e-01 -8.48327458e-01 5.59480309e-01
2.52205670e-01 -7.38072515e-01 -4.55310792e-01 3.30937356e-01]
|
[9.499054908752441, 0.09624890238046646]
|
5c9cd15f-f74f-46ac-893c-5b11a8c80c86
|
learning-discriminative-and-robust-time
|
1912.06808
| null |
https://arxiv.org/abs/1912.06808v3
|
https://arxiv.org/pdf/1912.06808v3.pdf
|
Environmental Sound Classification with Parallel Temporal-spectral Attention
|
Convolutional neural networks (CNN) are one of the best-performing neural network architectures for environmental sound classification (ESC). Recently, temporal attention mechanisms have been used in CNN to capture the useful information from the relevant time frames for audio classification, especially for weakly labelled data where the onset and offset times of the sound events are not applied. In these methods, however, the inherent spectral characteristics and variations are not explicitly exploited when obtaining the deep features. In this paper, we propose a novel parallel temporal-spectral attention mechanism for CNN to learn discriminative sound representations, which enhances the temporal and spectral features by capturing the importance of different time frames and frequency bands. Parallel branches are constructed to allow temporal attention and spectral attention to be applied respectively in order to mitigate interference from the segments without the presence of sound events. The experiments on three environmental sound classification (ESC) datasets and two acoustic scene classification (ASC) datasets show that our method improves the classification performance and also exhibits robustness to noise.
|
['Dading Chong', 'Wenwu Wang', 'Yuexian Zou', 'Helin Wang']
|
2019-12-14
| null | null | null | null |
['environmental-sound-classification', 'sound-classification']
|
['audio', 'audio']
|
[ 3.19663316e-01 -6.31951213e-01 3.84456456e-01 -4.57421660e-01
-5.43413281e-01 -3.49927992e-01 2.98041612e-01 2.10559249e-01
-5.50095201e-01 3.52588803e-01 3.40411514e-01 1.34299740e-01
-3.31166625e-01 -5.33825457e-01 -4.20318872e-01 -8.88458550e-01
-2.86663741e-01 -4.65169758e-01 4.86529827e-01 -1.13909684e-01
9.44952965e-02 4.11864340e-01 -2.02055335e+00 2.96902895e-01
6.62450731e-01 1.15366149e+00 4.39345181e-01 8.82397175e-01
-2.20293164e-01 5.91452956e-01 -6.93092644e-01 1.99085101e-01
-4.93346862e-02 -5.29313803e-01 -4.83153671e-01 -3.21548283e-01
2.64234036e-01 -3.50679271e-02 -4.32824343e-01 1.03287446e+00
8.57832789e-01 5.50649583e-01 3.76662940e-01 -1.05514419e+00
-2.47068450e-01 6.11805260e-01 -3.11725855e-01 6.36061192e-01
7.89688006e-02 7.88187236e-02 1.14142370e+00 -6.60668254e-01
-2.57957727e-02 1.10400975e+00 6.53465450e-01 3.96548033e-01
-7.74337232e-01 -8.02022398e-01 2.96538740e-01 7.14235783e-01
-1.11576986e+00 -6.20393455e-01 1.28700471e+00 -3.63125771e-01
8.53029072e-01 3.68933946e-01 5.89061201e-01 1.13401675e+00
-8.41781721e-02 7.68534720e-01 5.87348104e-01 -4.49408144e-01
2.09161103e-01 -2.97633231e-01 3.43611985e-01 6.33473545e-02
-4.90018547e-01 1.92007184e-01 -7.13582158e-01 -9.60744843e-02
4.51972038e-01 -7.91867673e-02 -4.30799544e-01 3.46410424e-01
-8.82138550e-01 5.62960148e-01 6.08816147e-01 6.05770111e-01
-4.99906003e-01 3.26338559e-01 8.50689650e-01 1.03744544e-01
4.91152197e-01 3.54530066e-01 -4.99095321e-01 -3.14679652e-01
-6.59545422e-01 8.24458674e-02 2.22107008e-01 5.30489385e-01
5.37963033e-01 5.98837078e-01 -2.44272500e-01 1.15925765e+00
1.05491340e-01 2.70556927e-01 6.66432321e-01 -6.93805039e-01
3.41292948e-01 6.56194389e-02 -5.22583351e-02 -1.07149994e+00
-6.19734526e-01 -7.71576941e-01 -8.42239082e-01 -2.61367142e-01
1.48884088e-01 -1.02876790e-01 -8.14694881e-01 1.85034180e+00
1.04859754e-01 8.19510698e-01 8.54447111e-03 9.74863529e-01
9.93672490e-01 9.14948642e-01 2.69962251e-01 -2.88601428e-01
1.41736889e+00 -6.85895562e-01 -9.64454293e-01 -1.09421477e-01
1.90025806e-01 -8.54256094e-01 1.10070455e+00 2.32287332e-01
-7.09134340e-01 -1.10421550e+00 -9.46963370e-01 4.70461547e-02
-3.29114586e-01 3.22889835e-01 3.80394608e-01 3.47156733e-01
-5.52880466e-01 6.28253698e-01 -8.68581831e-01 -3.47305536e-02
2.11945698e-01 2.04503417e-01 1.71457771e-02 3.60803097e-01
-1.55988669e+00 3.47946584e-01 2.18449473e-01 6.89430833e-01
-1.01599872e+00 -6.17076516e-01 -6.67594433e-01 3.12762022e-01
-1.65380351e-02 -6.25197291e-02 1.29313362e+00 -1.25316608e+00
-1.59481299e+00 2.51872450e-01 -1.85258582e-01 -4.74735856e-01
7.59406090e-02 -4.82348680e-01 -6.37208402e-01 3.11326683e-01
-1.04907602e-01 3.94947618e-01 9.13540781e-01 -7.84928918e-01
-5.63106239e-01 -1.28482744e-01 -2.23082327e-03 4.40625548e-02
-5.56483030e-01 3.52707148e-01 -1.75933644e-01 -9.04788315e-01
1.64956301e-01 -7.34274328e-01 -1.89602733e-01 -2.44728789e-01
-3.02977085e-01 -3.16358238e-01 9.99236703e-01 -7.17616498e-01
1.26087868e+00 -2.63688231e+00 9.79177281e-02 -2.11025611e-03
-2.56143570e-01 5.48894167e-01 -2.59066552e-01 3.03972661e-01
-7.06048980e-02 -1.48979723e-01 -2.71803409e-01 -1.65094748e-01
-1.03297792e-01 2.61784613e-01 -4.69937295e-01 2.60573119e-01
3.45961213e-01 4.59809512e-01 -9.52650666e-01 -2.23396540e-01
1.97230667e-01 6.90382421e-01 -4.97772068e-01 3.25499564e-01
-1.23839051e-01 6.71880901e-01 -4.65838403e-01 2.39403740e-01
4.71453011e-01 4.49449092e-01 -1.90186247e-01 -3.65606397e-01
-2.47294664e-01 5.84989429e-01 -1.23937917e+00 1.50796497e+00
-6.18020475e-01 8.99116933e-01 1.57316744e-01 -9.35482979e-01
9.76733804e-01 6.95421219e-01 4.53853250e-01 -7.71135032e-01
1.64658219e-01 2.04978675e-01 2.88491130e-01 -8.20551276e-01
4.07171071e-01 8.50995630e-03 4.67379950e-02 4.33371700e-02
8.07035714e-02 -1.65437628e-02 -4.01420072e-02 -3.79209012e-01
7.97614217e-01 -2.97981687e-02 -6.42870441e-02 -2.06523523e-01
7.40048349e-01 -6.31725192e-01 9.01223481e-01 4.56306636e-01
-3.74180317e-01 6.89670086e-01 1.97506458e-01 -5.51014185e-01
-7.23410666e-01 -6.93959177e-01 -2.41411358e-01 1.48208678e+00
2.67478880e-02 -4.69027847e-01 -5.88827610e-01 -3.49468172e-01
-3.65696251e-01 4.81482714e-01 -4.81913447e-01 -3.72191966e-01
-8.16491902e-01 -7.32500255e-01 8.12954843e-01 8.74709725e-01
5.12982190e-01 -1.29315221e+00 -8.54588568e-01 5.45240462e-01
-2.82706052e-01 -1.04549539e+00 -3.90465021e-01 6.74922705e-01
-5.68009436e-01 -8.62269998e-01 -5.29070795e-01 -7.20180213e-01
1.34980604e-01 1.51630685e-01 5.92151463e-01 -1.56860963e-01
-2.77012378e-01 7.64743313e-02 -5.91141343e-01 -6.25429392e-01
1.20582164e-03 5.95616214e-02 1.62960328e-02 4.48241174e-01
1.32611528e-01 -8.89666080e-01 -6.14139199e-01 2.25256085e-01
-8.72564316e-01 -8.39239284e-02 1.68364629e-01 9.03871179e-01
4.59121048e-01 2.60100394e-01 8.73274207e-01 -3.61624390e-01
4.77363974e-01 -2.94676423e-01 -3.25389028e-01 -1.97998941e-01
2.09534094e-01 -1.26213938e-01 9.14860964e-01 -7.08165646e-01
-1.07093906e+00 6.64481893e-02 -5.54616153e-01 -4.12326723e-01
-4.95782346e-01 3.83466601e-01 -1.04786187e-01 2.31380478e-01
5.63623190e-01 3.31625938e-01 -6.00290775e-01 -8.59717190e-01
-1.60153717e-01 7.87374258e-01 5.29951811e-01 -5.19171000e-01
3.93888354e-01 2.73773342e-01 -1.25149444e-01 -1.10301769e+00
-9.16039884e-01 -6.20527148e-01 -5.68703949e-01 -2.74333030e-01
8.88022244e-01 -7.78020084e-01 -4.46795285e-01 6.10716641e-01
-1.23946273e+00 -7.42220804e-02 -2.01563254e-01 8.26293647e-01
-2.75020033e-01 2.30011493e-01 -5.93887150e-01 -1.13032389e+00
-3.96353632e-01 -1.26170099e+00 1.01698184e+00 3.98833543e-01
-1.02550939e-01 -5.94347239e-01 -1.17193768e-02 -2.09024087e-01
5.71140468e-01 1.73874676e-01 8.93382370e-01 -5.50139427e-01
-2.28965685e-01 -2.78715789e-02 -2.08985768e-02 4.66083109e-01
1.95850044e-01 3.62063378e-01 -1.61493814e+00 -2.01555081e-02
-5.39368279e-02 2.74516791e-02 1.08332968e+00 6.06028974e-01
1.45283842e+00 -1.50576383e-01 3.41835618e-02 5.86765528e-01
1.03020239e+00 6.58785462e-01 5.46253800e-01 3.91999409e-02
6.40573740e-01 6.53563857e-01 4.60186929e-01 5.29969096e-01
-1.02584235e-01 7.93507814e-01 5.81241131e-01 -7.96581898e-03
-1.34033918e-01 -5.82903549e-02 3.77119511e-01 1.12343633e+00
-2.25704655e-01 -1.22302368e-01 -6.80780232e-01 8.89197946e-01
-1.71285379e+00 -8.97002697e-01 -3.59665304e-01 2.02785873e+00
8.33143055e-01 -2.24859845e-02 -4.11255434e-02 6.73270822e-01
7.26872802e-01 4.19540107e-01 -3.22973549e-01 -3.28443915e-01
-1.32847905e-01 5.35048604e-01 -7.27355108e-02 1.83664888e-01
-1.44888258e+00 6.89207733e-01 6.09029484e+00 8.86052608e-01
-1.37981451e+00 1.70859322e-01 2.47062251e-01 -1.30486324e-01
-1.78682655e-02 -1.78573862e-01 -5.64696431e-01 5.32092392e-01
1.13411939e+00 3.04784119e-01 1.06382690e-01 6.57047570e-01
3.85124356e-01 1.63164258e-01 -8.11372876e-01 9.02641594e-01
-4.72644001e-01 -9.37463343e-01 -1.24622770e-01 -4.11380857e-01
4.82467890e-01 -1.31090537e-01 1.27694711e-01 2.50939339e-01
-1.98556110e-01 -7.77519405e-01 9.14528370e-01 5.73359430e-01
4.08689618e-01 -1.02733946e+00 7.91096628e-01 1.86773404e-01
-1.64538789e+00 -3.10461700e-01 -4.74462509e-01 -9.93726328e-02
-1.93164423e-02 5.23893774e-01 -4.83747989e-01 5.89518964e-01
1.08854163e+00 8.15859556e-01 -3.86698067e-01 1.16879439e+00
-2.47096181e-01 1.05667675e+00 -3.12419832e-01 -1.27781123e-01
4.37101096e-01 1.92949444e-01 6.84989095e-01 1.48224568e+00
3.01598400e-01 1.92760732e-02 -9.06602144e-02 5.68417132e-01
1.94479197e-01 2.05804095e-01 -2.79866010e-01 6.18123403e-03
3.73464853e-01 1.02082455e+00 -5.62877417e-01 1.54595776e-02
-2.52876848e-01 5.13775229e-01 -4.89487424e-02 4.96929407e-01
-8.50537360e-01 -7.96073794e-01 8.62878382e-01 -2.65545309e-01
4.23430771e-01 -2.85644531e-01 -3.27919908e-02 -7.27867484e-01
-1.15736062e-02 -5.78092992e-01 4.37758476e-01 -7.08001554e-01
-1.04138041e+00 8.06150615e-01 -1.71970502e-01 -1.40213490e+00
-1.17637873e-01 -2.80001372e-01 -9.01900411e-01 8.12600911e-01
-1.81715107e+00 -9.31385934e-01 -3.13951433e-01 6.48875535e-01
7.38270104e-01 -4.44861278e-02 7.26753652e-01 5.34156621e-01
-6.50635064e-01 4.60624516e-01 7.58432597e-02 2.46495634e-01
6.12727582e-01 -9.94815588e-01 7.54178390e-02 8.54675293e-01
3.16786647e-01 3.99846107e-01 6.55716002e-01 -1.70662850e-01
-9.21233952e-01 -1.13781846e+00 7.31113851e-01 2.90634930e-01
5.47736764e-01 -3.08264464e-01 -1.15822136e+00 2.53123611e-01
1.81506664e-01 1.43801272e-01 8.08668137e-01 3.48970816e-02
-4.45264101e-01 -5.91049671e-01 -5.53198099e-01 2.43208557e-01
7.66890645e-01 -8.93402934e-01 -5.94219327e-01 -3.98491398e-02
9.10510540e-01 -1.42609060e-01 -4.49404657e-01 5.55567980e-01
5.62448621e-01 -9.11987126e-01 9.07496274e-01 -4.51560408e-01
2.59013325e-01 -5.68639278e-01 -2.71478176e-01 -1.37164617e+00
-2.75851727e-01 -6.08588755e-01 7.14457035e-02 1.36033726e+00
3.92281543e-03 -4.08766240e-01 1.08840227e-01 -1.27524808e-01
-6.15583718e-01 -3.69756818e-01 -1.17481494e+00 -6.32208884e-01
-4.06757057e-01 -1.00088763e+00 7.39902139e-01 7.77926207e-01
-3.83592129e-01 1.39438048e-01 -5.61953187e-01 4.41985995e-01
1.89613491e-01 2.17165813e-01 3.84949088e-01 -1.26559174e+00
-2.41851181e-01 -4.70997334e-01 -5.17687738e-01 -8.88230741e-01
2.09636956e-01 -5.45541525e-01 3.74845773e-01 -1.06149745e+00
-2.51080483e-01 -2.11909145e-01 -8.01540792e-01 5.49893260e-01
-2.75561571e-01 2.19431296e-01 7.57807642e-02 6.09565340e-03
-3.37398320e-01 8.29216361e-01 1.09725964e+00 -1.30191147e-01
-3.68100941e-01 2.82605320e-01 -1.91268966e-01 8.12968791e-01
6.82328224e-01 -3.97095263e-01 -4.11746442e-01 -5.18024564e-01
2.74644903e-04 -1.25087332e-02 5.14894962e-01 -1.32012534e+00
3.53486031e-01 -1.09740347e-01 3.15721899e-01 -7.46641815e-01
4.06813323e-01 -8.05332839e-01 2.22786576e-01 3.05952340e-01
-6.41625822e-01 -3.18737209e-01 5.69942892e-01 7.11558521e-01
-7.41594672e-01 -2.80536681e-01 9.76085305e-01 2.17335131e-02
-8.43353033e-01 2.89609909e-01 -5.17388225e-01 -2.18849242e-01
5.77687740e-01 7.83632398e-02 4.16085646e-02 -3.10340226e-01
-9.00452495e-01 -9.19035673e-02 -5.64615846e-01 5.03307223e-01
6.04823887e-01 -1.49285221e+00 -6.94074273e-01 3.36656302e-01
4.58391309e-02 -1.11216232e-01 8.44418466e-01 7.43822515e-01
-1.90677255e-01 3.74779880e-01 -2.23489612e-01 -6.29707277e-01
-1.29581165e+00 3.17382842e-01 6.18485332e-01 1.92497298e-01
-5.57186544e-01 1.04889989e+00 4.27748859e-01 1.37354815e-02
6.48511946e-01 -7.63679385e-01 -6.38608456e-01 3.06364775e-01
6.62375510e-01 3.99463415e-01 1.00720689e-01 -8.09614003e-01
-3.63193423e-01 7.45137155e-01 2.48387158e-01 1.49322137e-01
1.59403408e+00 -1.01835489e-01 6.64640144e-02 6.73066437e-01
1.30651951e+00 4.78074513e-02 -1.35112071e+00 -4.29621041e-01
2.24701893e-02 -2.72578925e-01 2.95459419e-01 -4.91811067e-01
-1.29788768e+00 1.49054110e+00 6.56748652e-01 4.63553816e-01
1.51319194e+00 -3.13206971e-01 9.34729874e-01 -5.85402586e-02
1.36415809e-02 -1.05857813e+00 1.44706517e-01 7.68723667e-01
9.65017378e-01 -8.40112448e-01 -5.71449161e-01 -1.68472812e-01
-3.88281584e-01 1.53448546e+00 5.24502933e-01 -6.60440400e-02
7.74942756e-01 3.06892961e-01 4.24724221e-02 2.50186538e-03
-6.25741363e-01 -5.74243844e-01 4.39679444e-01 5.89500904e-01
4.26204324e-01 -4.39166017e-02 6.29036874e-02 9.55316365e-01
-2.21307784e-01 -5.04554033e-01 1.02138318e-01 7.59296060e-01
-4.84988689e-01 -7.21023023e-01 -3.86600554e-01 5.47806034e-03
-7.07687020e-01 -6.85684606e-02 -2.02628270e-01 2.77178973e-01
3.47147316e-01 1.03893352e+00 1.52339429e-01 -4.75765586e-01
5.26452541e-01 2.99112320e-01 2.29780953e-02 -4.59910601e-01
-8.61921072e-01 5.92675328e-01 -5.98633476e-02 -4.56602365e-01
-5.70765018e-01 -4.67939466e-01 -1.30834603e+00 4.03617054e-01
-4.90386039e-01 3.25968087e-01 5.81162393e-01 7.86535144e-01
3.15425336e-01 1.19823241e+00 8.61891627e-01 -9.71450746e-01
-1.64231643e-01 -1.14564157e+00 -6.35447741e-01 3.29972208e-01
8.08679640e-01 -6.15095019e-01 -5.14312387e-01 1.66800797e-01]
|
[15.185385704040527, 5.23579740524292]
|
edf86f71-1b48-4e6e-8ba4-6313f87c0943
|
real-time-polyp-detection-localisation-and
|
2011.07631
| null |
https://arxiv.org/abs/2011.07631v2
|
https://arxiv.org/pdf/2011.07631v2.pdf
|
Real-Time Polyp Detection, Localization and Segmentation in Colonoscopy Using Deep Learning
|
Computer-aided detection, localisation, and segmentation methods can help improve colonoscopy procedures. Even though many methods have been built to tackle automatic detection and segmentation of polyps, benchmarking of state-of-the-art methods still remains an open problem. This is due to the increasing number of researched computer vision methods that can be applied to polyp datasets. Benchmarking of novel methods can provide a direction to the development of automated polyp detection and segmentation tasks. Furthermore, it ensures that the produced results in the community are reproducible and provide a fair comparison of developed methods. In this paper, we benchmark several recent state-of-the-art methods using Kvasir-SEG, an open-access dataset of colonoscopy images for polyp detection, localisation, and segmentation evaluating both method accuracy and speed. Whilst, most methods in literature have competitive performance over accuracy, we show that the proposed ColonSegNet achieved a better trade-off between an average precision of 0.8000 and mean IoU of 0.8100, and the fastest speed of 180 frames per second for the detection and localisation task. Likewise, the proposed ColonSegNet achieved a competitive dice coefficient of 0.8206 and the best average speed of 182.38 frames per second for the segmentation task. Our comprehensive comparison with various state-of-the-art methods reveals the importance of benchmarking the deep learning methods for automated real-time polyp identification and delineations that can potentially transform current clinical practices and minimise miss-detection rates.
|
['Håvard D. Johansen', 'Nikhil Kumar Tomar', 'Pål Halvorsen', 'Michael A. Riegler', 'Jens Rittscher', 'Dag D. Johansen', 'Sharib Ali', 'Debesh Jha']
|
2020-11-15
| null | null | null | null |
['medical-object-detection']
|
['computer-vision']
|
[ 1.79710060e-01 5.58420718e-02 -2.52822861e-02 -5.95568540e-03
-9.88091826e-01 -5.68804801e-01 2.64060885e-01 8.12538743e-01
-8.03719699e-01 2.77004987e-01 -1.69780537e-01 -6.14180446e-01
-2.17583831e-02 -6.43172681e-01 -3.84778678e-01 -8.50340247e-01
-4.56250340e-01 2.87136436e-01 6.31819129e-01 4.14032489e-02
7.97803253e-02 2.78603584e-01 -9.79520679e-01 4.47428107e-01
8.64943683e-01 8.96349430e-01 3.59557658e-01 1.34885418e+00
4.40859437e-01 2.28940219e-01 -3.75274003e-01 -4.07372981e-01
3.03040773e-01 -4.52484697e-01 -7.25848079e-01 -2.72995323e-01
3.71425331e-01 -1.53987691e-01 -9.96050239e-02 9.13243651e-01
9.62113321e-01 -1.89826950e-01 5.69361150e-01 -2.40340859e-01
-1.55328736e-01 3.59194607e-01 -3.84425133e-01 7.35173106e-01
3.78636628e-01 2.53312498e-01 3.06671917e-01 -1.98263586e-01
4.59390700e-01 5.58616281e-01 1.11357164e+00 2.52417564e-01
-9.27009106e-01 -1.52447581e-01 -3.84787917e-01 -2.77372181e-01
-9.98106599e-01 8.87331367e-02 -1.28423169e-01 -6.57327235e-01
9.30456758e-01 3.70242834e-01 1.00434184e+00 5.69914699e-01
5.84979415e-01 6.91132605e-01 9.10815060e-01 -6.77462041e-01
-2.66079083e-02 2.60623127e-01 -2.60627896e-01 1.13912463e+00
8.00762594e-01 3.35340410e-01 1.51332378e-01 -1.06575437e-01
8.35267663e-01 -1.88312799e-01 -4.02128190e-01 -4.49721485e-01
-1.62564409e+00 9.70508575e-01 7.52133548e-01 5.84092438e-01
-4.26563025e-01 5.65800928e-02 8.56921136e-01 -2.76398822e-03
1.07261941e-01 8.76206636e-01 -2.41391003e-01 -1.64615393e-01
-1.23898077e+00 -4.49353531e-02 1.17172456e+00 5.34844697e-01
-9.51696038e-02 -4.04179305e-01 -1.83325186e-01 5.43088734e-01
3.50271106e-01 2.53341198e-01 1.01259136e+00 -5.59198678e-01
1.56570971e-02 7.20116556e-01 3.30187045e-02 -1.04197228e+00
-6.90901935e-01 -6.68721378e-01 -8.87620091e-01 1.78340122e-01
7.76418686e-01 -1.01757310e-01 -9.54278946e-01 7.22661376e-01
2.78313369e-01 -2.68779490e-02 6.19119368e-02 9.10431087e-01
1.00720394e+00 1.52798459e-01 -7.91888461e-02 9.76586267e-02
1.61108434e+00 -1.07694161e+00 -2.51955867e-01 -1.87658027e-01
9.71085787e-01 -1.28151584e+00 5.03195822e-01 6.61619663e-01
-1.26258898e+00 -5.08675992e-01 -1.26738012e+00 1.23975158e-01
-2.37204134e-01 5.06567597e-01 6.10534072e-01 1.07753360e+00
-1.10486245e+00 5.08891284e-01 -1.33549392e+00 -4.84127581e-01
4.14051205e-01 5.17246902e-01 -1.53417766e-01 -7.94665515e-02
-6.49150968e-01 1.11082315e+00 5.72970450e-01 5.56820491e-03
-8.08969021e-01 -8.15071762e-01 -8.47764015e-01 -3.36567283e-01
1.28097564e-01 -9.77244735e-01 1.52490509e+00 -6.88829303e-01
-1.26264632e+00 1.35947657e+00 3.64023834e-01 -1.14952123e+00
9.71607208e-01 -1.06907479e-01 -1.60828382e-01 4.41086978e-01
-8.43597800e-02 7.55478799e-01 3.66088301e-01 -7.26860344e-01
-1.09494865e+00 2.25888062e-02 -5.01807332e-02 1.38681322e-01
1.09150521e-01 -2.05310017e-01 -6.46280050e-01 -4.56652343e-01
3.35394964e-02 -1.19446826e+00 -6.93517089e-01 3.20915788e-01
-1.77398026e-01 3.91177773e-01 2.36070454e-01 -8.72286916e-01
1.34309816e+00 -2.10642695e+00 -2.06240162e-01 2.99311001e-02
1.01232395e-01 8.83724988e-01 2.56736010e-01 1.60566438e-02
2.11261794e-01 -2.02543125e-03 -3.41756940e-01 5.87764084e-02
-3.93017352e-01 -5.47903962e-02 5.19107461e-01 9.78811026e-01
-2.42144078e-01 9.55038190e-01 -1.20484853e+00 -6.69777274e-01
9.77664113e-01 7.64089882e-01 -5.84742725e-01 1.53259151e-02
1.58193603e-01 4.71699476e-01 -1.83366954e-01 5.64431787e-01
4.97262001e-01 -2.95667648e-01 1.93339139e-01 -2.16373429e-01
-3.74883682e-01 1.01015545e-01 -1.05998039e+00 1.98488557e+00
-6.14629209e-01 6.95267379e-01 2.24390533e-02 -9.43710566e-01
6.12988651e-01 5.98299801e-01 5.28689027e-01 -5.08398652e-01
4.46783423e-01 8.11961174e-01 5.07850647e-01 -6.87490702e-01
1.63675576e-01 1.38272122e-01 3.09718907e-01 -5.38036376e-02
-9.43519473e-02 -4.41635489e-01 6.50561810e-01 -2.09314436e-01
1.03893471e+00 -3.01379990e-02 9.98928428e-01 -5.76827943e-01
8.62673938e-01 3.13056409e-01 -1.50878266e-01 9.65640783e-01
-6.45511568e-01 7.52296388e-01 1.28661811e-01 -7.28364110e-01
-9.66855466e-01 -7.31294215e-01 -4.96989518e-01 4.02308702e-01
2.13848740e-01 1.67507425e-01 -8.79755437e-01 -6.53815329e-01
-3.10427368e-01 5.52059561e-02 -6.62233591e-01 1.42773360e-01
-7.69293785e-01 -1.14611638e+00 5.66830099e-01 2.79835790e-01
4.70134407e-01 -9.96881068e-01 -1.63118255e+00 6.16632402e-01
-1.59406185e-01 -9.53663290e-01 -1.99609682e-01 -2.55921260e-02
-1.05209124e+00 -1.53164566e+00 -1.32763934e+00 -1.34887278e+00
6.76108539e-01 1.63896084e-01 1.41698539e+00 2.84549713e-01
-9.49859440e-01 2.08107039e-01 -4.45747137e-01 -4.87999499e-01
-7.60507166e-01 1.72680840e-01 -6.99294090e-01 -7.35286117e-01
2.55763650e-01 2.55903378e-02 -1.38470829e+00 2.34242678e-01
-8.41999888e-01 -7.16204718e-02 7.05619097e-01 9.01162326e-01
8.08321655e-01 -4.01858687e-01 -6.73592240e-02 -7.97771811e-01
4.70159322e-01 -1.31470099e-01 -7.88910449e-01 2.25110441e-01
-6.13619924e-01 -2.30263799e-01 2.79741585e-01 -3.12054396e-01
-6.50579512e-01 4.39291000e-01 -3.22308958e-01 2.99351811e-01
-2.77762681e-01 4.95587051e-01 1.11996424e+00 -6.35562062e-01
1.10189903e+00 9.12507847e-02 1.39340296e-01 -1.16093792e-01
2.13121027e-01 4.79036510e-01 6.03465676e-01 6.54782206e-02
7.64154829e-03 6.92806661e-01 1.22744687e-01 -7.38848746e-01
-6.13954246e-01 -1.20604277e+00 -5.65059185e-01 1.45911887e-01
1.04798973e+00 -9.41310406e-01 -4.46119517e-01 3.80695850e-01
-8.29142690e-01 -9.33399796e-02 -2.96239704e-01 7.12107658e-01
-4.78190243e-01 6.32578969e-01 -8.25127959e-01 -4.49846715e-01
-9.16374385e-01 -1.71697855e+00 9.08398092e-01 3.54210347e-01
-3.21056724e-01 -1.37696648e+00 3.34134340e-01 3.36472951e-02
6.38131440e-01 8.71295631e-01 3.57662916e-01 -6.43512666e-01
-3.89981985e-01 -6.88324153e-01 -3.75899822e-01 2.36362651e-01
7.16665015e-02 -1.27951428e-01 -6.58914804e-01 -6.03025019e-01
7.51929060e-02 1.96706548e-01 9.62314785e-01 1.17056954e+00
9.08256114e-01 7.11498931e-02 -6.85019135e-01 8.32549036e-01
1.67588210e+00 4.69522029e-01 4.01933134e-01 6.89149261e-01
1.40160069e-01 2.41895691e-01 6.03145719e-01 2.44421512e-01
1.39450468e-02 3.97371680e-01 5.78098595e-01 -5.86076200e-01
-3.76412272e-01 2.11094707e-01 -3.70957434e-01 5.68956256e-01
-4.55477461e-02 -8.52971151e-02 -1.21988034e+00 8.38704467e-01
-1.59180748e+00 -4.50713843e-01 -4.89623457e-01 2.32768726e+00
6.44913852e-01 1.45031109e-01 1.57550320e-01 2.02031344e-01
4.49896574e-01 -3.25688452e-01 -1.18060544e-01 -3.70531052e-01
2.72992402e-01 3.68011206e-01 7.36723304e-01 4.00788248e-01
-1.67061126e+00 2.01439470e-01 6.17085361e+00 6.26381397e-01
-1.19918263e+00 1.24417081e-01 6.84445918e-01 1.26717642e-01
5.89823544e-01 -4.76007521e-01 -5.25059640e-01 3.05285692e-01
8.88134599e-01 3.93612683e-02 -8.04953203e-02 7.96087205e-01
-5.95800616e-02 -4.95591551e-01 -1.08535635e+00 8.82643878e-01
9.66426730e-02 -1.44896853e+00 -4.63997632e-01 -2.65498366e-02
1.03582704e+00 2.83810556e-01 1.62913017e-02 1.69841666e-03
-1.48524836e-01 -1.04868281e+00 1.13531716e-01 3.08432013e-01
7.95303404e-01 -4.81907010e-01 1.32378161e+00 1.00994915e-01
-1.15506005e+00 1.93360925e-01 -2.95546412e-01 2.12969437e-01
1.40750423e-01 6.56346440e-01 -1.39208233e+00 5.53267241e-01
6.46242976e-01 6.10607445e-01 -6.04745507e-01 2.24077868e+00
-1.63875706e-02 2.95016885e-01 -4.10454959e-01 -2.09281310e-01
6.60691559e-01 5.46293035e-02 4.69629079e-01 2.08258128e+00
4.47040975e-01 -2.51943529e-01 2.16948345e-01 1.89089909e-01
4.02443171e-01 3.98390710e-01 -1.97883606e-01 2.45666966e-01
6.73535690e-02 1.38034785e+00 -1.21898246e+00 -5.37087560e-01
-3.92117918e-01 7.74208486e-01 -4.04214621e-01 -1.57189637e-01
-7.09281206e-01 -3.11379820e-01 -6.66899085e-02 1.51095554e-01
2.96955854e-01 1.09732226e-01 -3.37865859e-01 -7.27417171e-01
-1.01841502e-01 -9.27038908e-01 7.10068047e-01 -6.99944794e-02
-9.82811272e-01 6.81977570e-01 -2.06710711e-01 -1.40402317e+00
-1.94025844e-01 -1.09702003e+00 -4.79914367e-01 6.47973895e-01
-1.66969228e+00 -9.21529710e-01 -4.89704460e-01 -5.61902998e-03
6.47139251e-01 7.52064213e-02 1.15417337e+00 3.13327491e-01
1.28784059e-02 6.04054630e-01 2.51651347e-01 1.50101572e-01
4.97633040e-01 -1.55804873e+00 4.28026140e-01 8.41734350e-01
-1.35025959e-02 4.12978858e-01 5.72489738e-01 -3.30171615e-01
-9.02958155e-01 -8.24065387e-01 5.72696388e-01 -2.96147823e-01
5.21200716e-01 4.11044985e-01 -5.36090136e-01 2.42820457e-01
1.91001669e-01 1.37390345e-01 6.59934998e-01 -5.12642920e-01
2.84968764e-01 3.77504319e-01 -1.40714204e+00 4.00729626e-01
3.39344949e-01 1.04478121e-01 -4.80697572e-01 5.83072960e-01
4.42170620e-01 -1.19732749e+00 -9.58362818e-01 5.21468580e-01
6.97862208e-01 -1.41665459e+00 1.31023633e+00 2.29097188e-01
3.10913950e-01 -4.90788855e-02 3.97849768e-01 -1.12288177e+00
-6.34578690e-02 -4.19279605e-01 8.14781860e-02 2.42945135e-01
6.83531165e-01 -5.74362636e-01 8.09273899e-01 4.36986461e-02
-3.32390040e-01 -1.07717431e+00 -9.04591739e-01 -3.03466350e-01
6.63706884e-02 -1.41731799e-01 -1.71877503e-01 6.35077238e-01
-8.53867829e-02 -5.33527076e-01 2.13496149e-01 3.59392888e-03
3.88933599e-01 1.05306823e-02 3.40999782e-01 -8.81719768e-01
-3.03220570e-01 -7.70599186e-01 -7.24986315e-01 -7.99857974e-01
-9.36279714e-01 -8.91163886e-01 3.65800746e-02 -1.99766636e+00
3.48210782e-02 -2.76767731e-01 -2.69519836e-01 1.73370279e-02
-2.28311241e-01 4.49285090e-01 -7.80143961e-02 8.40248093e-02
-4.93227124e-01 -6.16570950e-01 1.57243264e+00 4.73006442e-02
-4.15707886e-01 3.99689436e-01 -3.94034892e-01 9.35664773e-01
8.45529616e-01 -4.57446307e-01 3.66844684e-02 -2.32804358e-01
7.04457313e-02 4.05943617e-02 4.18605804e-01 -1.44206762e+00
2.97104031e-01 6.54552042e-01 3.96792471e-01 -5.25951028e-01
-2.64710076e-02 -6.42472982e-01 -6.72127604e-02 1.52446616e+00
-1.94053978e-01 -3.19145471e-02 3.70803475e-01 3.72585505e-01
-4.18198854e-01 -5.31069338e-01 1.15998638e+00 -7.58023798e-01
-4.81781870e-01 2.35428344e-02 -3.25717568e-01 2.00574458e-01
1.12593186e+00 -5.22418261e-01 1.42167229e-02 2.28195116e-01
-8.24452281e-01 5.58345132e-02 2.76623368e-01 9.10074785e-02
4.45958257e-01 -6.00530565e-01 -9.85090494e-01 2.26021305e-01
-4.86731995e-03 3.30036283e-01 2.37471923e-01 1.44828832e+00
-1.62096214e+00 7.48842895e-01 7.59461969e-02 -9.74567711e-01
-1.31346130e+00 4.51964706e-01 9.28177118e-01 -6.94358528e-01
-7.63803601e-01 1.22843719e+00 -6.73911348e-02 -2.28052855e-01
2.52138823e-01 -8.12934577e-01 -2.27927193e-01 -1.10316798e-01
5.57652175e-01 4.41899121e-01 4.68589693e-01 -1.74377605e-01
-3.49245012e-01 5.28633058e-01 -2.20459908e-01 2.75474101e-01
8.25318813e-01 6.19323626e-02 1.71157613e-01 1.64102152e-04
9.45850611e-01 -2.84154743e-01 -9.22125041e-01 9.95985046e-02
4.63234866e-03 -4.31128860e-01 3.05339694e-01 -1.22694969e+00
-8.45591545e-01 8.34552824e-01 1.37134826e+00 4.10784751e-01
1.06379974e+00 -4.77182239e-01 7.15063989e-01 -1.35439694e-01
1.95872426e-01 -5.85041046e-01 -1.72727406e-01 9.61710662e-02
5.36441267e-01 -1.67005563e+00 3.08885187e-01 -4.95265782e-01
-4.11427915e-01 1.23416722e+00 1.47855267e-01 -4.72750455e-01
5.63579559e-01 4.03673798e-01 4.60202068e-01 -1.18788309e-01
3.58097665e-02 -5.43037467e-02 7.32048810e-01 6.52741015e-01
1.04154181e+00 3.22561204e-01 -8.02418470e-01 2.46790666e-02
-1.86850548e-01 2.31330648e-01 4.21306491e-01 9.80218053e-01
-5.70611656e-01 -8.67680371e-01 -3.26775551e-01 4.93508697e-01
-1.11329901e+00 -2.68115610e-01 2.31877133e-01 1.13785958e+00
2.29200795e-01 8.73697877e-01 -3.69050801e-01 5.61886251e-01
4.16872054e-01 -4.79071468e-01 6.01203620e-01 -5.74369311e-01
-1.21766889e+00 3.57180476e-01 5.27634993e-02 -4.00066108e-01
-5.51157475e-01 -5.25953293e-01 -1.05393815e+00 1.81040958e-01
-3.51176292e-01 3.36313434e-02 9.47298110e-01 5.78396976e-01
-1.20350100e-01 8.43573213e-01 -1.35201976e-01 -7.96998620e-01
-4.53058958e-01 -7.74085104e-01 -1.10737942e-01 3.23271126e-01
4.66325939e-01 -8.27470869e-02 -3.31239879e-01 2.12369174e-01]
|
[14.259086608886719, -3.0056605339050293]
|
edacbac9-a5d8-4d88-8cf6-881fe9965bba
|
multilingual-sentence-transformer-as-a
|
2301.12140
| null |
https://arxiv.org/abs/2301.12140v1
|
https://arxiv.org/pdf/2301.12140v1.pdf
|
Multilingual Sentence Transformer as A Multilingual Word Aligner
|
Multilingual pretrained language models (mPLMs) have shown their effectiveness in multilingual word alignment induction. However, these methods usually start from mBERT or XLM-R. In this paper, we investigate whether multilingual sentence Transformer LaBSE is a strong multilingual word aligner. This idea is non-trivial as LaBSE is trained to learn language-agnostic sentence-level embeddings, while the alignment extraction task requires the more fine-grained word-level embeddings to be language-agnostic. We demonstrate that the vanilla LaBSE outperforms other mPLMs currently used in the alignment task, and then propose to finetune LaBSE on parallel corpus for further improvement. Experiment results on seven language pairs show that our best aligner outperforms previous state-of-the-art models of all varieties. In addition, our aligner supports different language pairs in a single model, and even achieves new state-of-the-art on zero-shot language pairs that does not appear in the finetuning process.
|
['Yun Chen', 'Yue Han', 'Hanqing Wang', 'Guanhua Chen', 'Weikang Wang']
|
2023-01-28
| null | null | null | null |
['word-alignment', 'xlm-r']
|
['natural-language-processing', 'natural-language-processing']
|
[-2.44323581e-01 -2.80696660e-01 -4.81552660e-01 -3.71293575e-01
-1.18012071e+00 -7.23076403e-01 6.22340798e-01 1.70081079e-01
-8.17523837e-01 6.15442932e-01 3.89418066e-01 -9.13863838e-01
3.19262385e-01 -5.54592133e-01 -8.79284441e-01 -3.29608172e-01
1.31769672e-01 9.97418821e-01 -1.87662855e-01 -8.34089816e-01
-4.94416319e-02 6.04496710e-02 -6.30546808e-01 -2.15802491e-02
1.16923845e+00 -1.39121398e-01 4.15262282e-01 5.57159007e-01
-5.78227222e-01 2.94055879e-01 -3.01041245e-01 -8.00995171e-01
4.13473904e-01 -2.89679766e-01 -7.46010959e-01 -5.22787511e-01
1.18902528e+00 -1.98668893e-02 -7.40007758e-02 1.09659421e+00
6.42406285e-01 -4.03626710e-01 5.06875813e-01 -1.05259836e+00
-1.14213347e+00 1.50072253e+00 -6.15518868e-01 3.98315966e-01
2.38659382e-01 2.45084390e-01 1.55443120e+00 -1.21627855e+00
6.04957402e-01 1.20370150e+00 8.34516048e-01 4.67821211e-01
-1.21641922e+00 -9.32150185e-01 3.90147269e-01 4.34427351e-01
-1.40463555e+00 -5.08516014e-01 5.82709193e-01 -3.80096048e-01
1.47779250e+00 5.68742529e-02 5.33229411e-01 1.17527854e+00
4.73593950e-01 7.74361134e-01 1.28724241e+00 -9.23916042e-01
-3.16746652e-01 -3.64390910e-02 3.49416435e-01 6.92367613e-01
3.17629546e-01 -2.74429172e-01 -4.59493786e-01 1.30710438e-01
3.70292693e-01 -5.54559946e-01 4.28273492e-02 -1.28623590e-01
-1.53363621e+00 8.30309689e-01 -3.65505926e-02 7.22437203e-01
-9.01381969e-02 -1.49726540e-01 5.78546882e-01 7.89220631e-01
5.11467874e-01 8.60817552e-01 -6.91445291e-01 -1.14909656e-01
-9.50836480e-01 7.37325102e-02 7.41868794e-01 1.04109442e+00
8.58469307e-01 9.73470584e-02 -1.72289982e-01 1.09661949e+00
2.93309242e-02 7.38301516e-01 7.57570505e-01 -2.88579345e-01
8.92687738e-01 3.48026216e-01 -3.92528087e-01 -4.07008499e-01
-2.11920500e-01 -5.08147836e-01 -7.47163534e-01 -3.30456913e-01
3.86556000e-01 -6.27242252e-02 -7.17632115e-01 2.02636695e+00
-5.73988408e-02 4.75032181e-02 1.75190628e-01 6.25968099e-01
4.35802072e-01 6.76648915e-01 1.87229320e-01 2.87146140e-02
1.41900003e+00 -1.57128990e+00 -7.55635560e-01 -8.64363253e-01
1.14088953e+00 -1.39730215e+00 1.58097863e+00 8.24416280e-02
-1.04454696e+00 -8.13310087e-01 -1.20102000e+00 -3.41467947e-01
-4.85727459e-01 2.25790232e-01 6.35389805e-01 6.07793450e-01
-1.16223526e+00 2.22062916e-01 -6.22121394e-01 -6.41957045e-01
-3.09715033e-01 1.57692432e-01 -6.26836956e-01 -3.36238980e-01
-1.67707634e+00 1.53237617e+00 2.17777342e-01 -7.21403658e-02
-5.28761387e-01 -6.61155641e-01 -1.08686423e+00 -1.89239025e-01
-4.83274758e-02 -7.53969729e-01 1.20609689e+00 -6.31854951e-01
-1.56797504e+00 1.35453641e+00 -2.91890264e-01 -5.88278234e-01
8.62078816e-02 -5.73963344e-01 -5.74819386e-01 -5.79454601e-01
3.92746538e-01 7.06367075e-01 3.39600563e-01 -7.18626201e-01
-5.76083362e-01 -7.07764030e-02 -7.80570060e-02 2.42525145e-01
-5.83017290e-01 4.01634187e-01 -3.08459818e-01 -7.53916383e-01
-2.62193829e-01 -1.01937044e+00 -4.86511946e-01 -7.43945956e-01
-4.49448287e-01 -4.46502030e-01 9.30113941e-02 -9.24233556e-01
1.46672475e+00 -1.63589191e+00 2.72473276e-01 -1.16879575e-01
-2.90047437e-01 5.53371966e-01 -9.60345924e-01 7.70979047e-01
-2.63789356e-01 2.16335177e-01 -7.32280836e-02 -5.88350594e-01
3.18297863e-01 2.95366049e-01 -2.58411855e-01 4.04889226e-01
3.38247895e-01 1.10355854e+00 -9.41737473e-01 -6.08287573e-01
2.16837898e-01 1.88847378e-01 -4.99554455e-01 2.95378238e-01
-8.44931155e-02 3.64920527e-01 1.29626989e-01 5.20093441e-01
6.73080087e-01 3.75617862e-01 4.44145232e-01 -1.63696721e-01
-4.52509373e-01 1.01788747e+00 -6.60677016e-01 2.16040349e+00
-1.06425285e+00 5.24446011e-01 -1.84808865e-01 -9.84729946e-01
9.18846965e-01 1.71608463e-01 1.84088707e-01 -7.96119452e-01
-5.99917062e-02 6.92196369e-01 5.65940440e-01 -2.81238884e-01
8.47571909e-01 9.35338885e-02 -5.65009654e-01 5.60803056e-01
4.80256438e-01 -2.16525957e-01 4.65012759e-01 1.23413466e-01
9.00827110e-01 2.66843408e-01 6.31866455e-01 -5.83757281e-01
6.42866969e-01 1.37886368e-02 5.70855319e-01 6.40204310e-01
-3.02754212e-02 1.74009055e-01 -9.48848426e-02 -3.95406723e-01
-1.30228651e+00 -1.13722992e+00 -1.65184096e-01 1.51940167e+00
-1.49284005e-01 -7.19738901e-01 -7.80180395e-01 -6.62000775e-01
-3.39010894e-01 9.89885211e-01 -1.42976090e-01 4.28167395e-02
-1.28642869e+00 -8.27080190e-01 7.21769512e-01 2.91914225e-01
1.06148131e-01 -8.47997844e-01 4.33382660e-01 5.60080051e-01
-4.45317060e-01 -1.40893543e+00 -9.75590408e-01 1.38610631e-01
-4.03265327e-01 -7.00668097e-01 -4.83629733e-01 -1.27987659e+00
4.29359972e-01 -2.36168001e-02 1.66617787e+00 -3.08867872e-01
3.34766768e-02 1.55941755e-01 -3.60121876e-01 -3.40912819e-01
-7.71929324e-01 9.24154103e-01 4.54588652e-01 -3.75842839e-01
1.09758699e+00 -7.53486693e-01 -2.95888409e-02 2.46519269e-03
-4.17595476e-01 -6.37141913e-02 7.65793324e-01 8.85333598e-01
6.03277445e-01 -7.01828480e-01 5.11298835e-01 -9.54668283e-01
7.72763789e-01 -2.75237322e-01 -5.79808652e-01 4.84272957e-01
-5.28771520e-01 3.33766550e-01 9.27368045e-01 -5.90721130e-01
-5.14983058e-01 -1.08766131e-01 -6.29424393e-01 2.29513515e-02
5.12363156e-03 5.75793564e-01 -4.24944967e-01 1.43569008e-01
5.15292704e-01 2.31575415e-01 -3.59020352e-01 -6.56574547e-01
9.01999593e-01 7.76273489e-01 7.14822948e-01 -9.02691960e-01
1.10970771e+00 -3.36698860e-01 -4.10542488e-01 -5.38731813e-01
-9.90697801e-01 -5.31443596e-01 -9.55816627e-01 1.73040166e-01
8.27821732e-01 -1.18111110e+00 8.14721510e-02 3.78771603e-01
-1.48805356e+00 -3.70944351e-01 -1.01140089e-01 5.63039243e-01
-2.44269639e-01 2.54702747e-01 -9.19337571e-01 -2.38890812e-01
-6.28057837e-01 -1.36026454e+00 1.00537586e+00 -1.96878001e-01
-5.57448566e-01 -1.23862374e+00 7.53836811e-01 3.46866637e-01
5.61324179e-01 -4.64507610e-01 1.13201857e+00 -1.02576435e+00
-3.59636486e-01 2.39396840e-02 1.41319200e-01 4.55124199e-01
2.12634638e-01 -1.52986839e-01 -6.91356838e-01 -4.06744659e-01
-3.43097836e-01 -2.58202940e-01 7.00071275e-01 1.57876596e-01
5.35541356e-01 -2.87683815e-01 -1.64158996e-02 7.40663111e-01
1.28574562e+00 -4.32190955e-01 4.46872294e-01 7.11752117e-01
9.95115280e-01 3.21448714e-01 5.25173068e-01 -3.47581983e-01
8.56644928e-01 9.33894873e-01 1.51582316e-01 -4.61396635e-01
-3.80719155e-01 -3.67823154e-01 9.84288633e-01 2.28378129e+00
2.22487062e-01 6.86783809e-03 -1.07802856e+00 9.93504524e-01
-1.66248465e+00 -6.60868049e-01 -2.16204807e-01 2.06873775e+00
1.43540537e+00 -9.34022367e-02 -1.54343516e-01 -3.59632522e-01
5.58592856e-01 2.72418201e-01 3.71085145e-02 -1.08244443e+00
-4.95728552e-01 7.68723786e-01 6.56380832e-01 9.25632000e-01
-1.05245626e+00 1.82138133e+00 6.05854940e+00 9.60146964e-01
-1.05821049e+00 5.14641881e-01 2.24306732e-01 2.61219263e-01
-5.39091289e-01 2.34198421e-01 -1.37039137e+00 2.45146617e-01
1.02020180e+00 -1.51252910e-01 4.97810990e-01 7.65499830e-01
-1.75167277e-01 4.52302605e-01 -1.41808653e+00 7.16410756e-01
2.40015015e-01 -9.25349355e-01 2.65494943e-01 -8.90859365e-02
8.77737403e-01 5.64145446e-01 -4.46102917e-02 8.74030173e-01
7.94580340e-01 -1.20061398e+00 4.50866640e-01 1.04738109e-01
9.01313901e-01 -7.83545971e-01 9.39554751e-01 1.97608605e-01
-1.17048299e+00 4.13022250e-01 -6.26614094e-01 -5.83549961e-02
4.78971422e-01 6.19735777e-01 -9.07260180e-01 5.95191836e-01
2.22342208e-01 5.92166901e-01 -8.04640651e-01 6.41335249e-01
-5.47539532e-01 8.95036578e-01 -1.41520247e-01 9.45019722e-02
4.91170436e-01 -4.32239473e-01 6.98815703e-01 1.76135433e+00
5.21883547e-01 -6.64864480e-01 6.24285519e-01 3.39766800e-01
-9.23441127e-02 7.69958556e-01 -5.96222818e-01 -7.94548094e-02
6.30097866e-01 1.28952909e+00 7.95567408e-02 -2.16275603e-01
-6.88594103e-01 1.10259128e+00 9.37753856e-01 1.76409911e-02
-6.48965001e-01 -2.68196732e-01 1.09967220e+00 -6.47555068e-02
1.40273660e-01 -5.10130763e-01 -4.12859134e-02 -1.37919378e+00
-9.32326689e-02 -1.39413989e+00 1.99852318e-01 -3.48279893e-01
-1.75707114e+00 8.44387054e-01 -2.99164683e-01 -1.15426159e+00
-5.29977679e-01 -9.36388373e-01 -6.33716881e-01 1.22072232e+00
-1.81159377e+00 -1.73485446e+00 3.59585285e-01 4.40480709e-01
7.81358600e-01 -6.86631203e-01 1.29045928e+00 6.22674167e-01
-5.75810969e-01 1.06700993e+00 -7.36611113e-02 4.51606125e-01
1.37669063e+00 -1.41463566e+00 1.04350102e+00 1.18540287e+00
8.32165837e-01 8.56103301e-01 6.52493358e-01 -3.91867012e-01
-1.38510048e+00 -1.14335787e+00 1.93294597e+00 -6.41584396e-01
1.33382630e+00 -5.81285477e-01 -7.84924090e-01 1.02001774e+00
1.00471354e+00 -1.73892453e-01 8.76126111e-01 7.64596283e-01
-7.05979109e-01 -2.20252991e-01 -3.13891232e-01 1.11053550e+00
1.09614789e+00 -8.07785332e-01 -9.38075066e-01 4.62977618e-01
8.64797890e-01 -2.68630266e-01 -9.02755499e-01 3.90609086e-01
3.28527153e-01 -3.75708729e-01 9.26732659e-01 -8.97190630e-01
4.14975435e-01 -1.36629373e-01 -3.41255218e-01 -2.01079893e+00
-5.48579276e-01 -6.89050853e-01 4.23065394e-01 1.57722366e+00
7.79797316e-01 -7.18588471e-01 6.01896979e-02 -3.88105631e-01
-4.52827483e-01 -5.58403969e-01 -9.45548058e-01 -1.15923309e+00
9.39368248e-01 -4.48900193e-01 7.76418626e-01 1.46110678e+00
9.83931050e-02 9.07087147e-01 -3.88900191e-01 2.04345644e-01
4.34334010e-01 7.99465179e-02 1.00062847e+00 -7.93908954e-01
-5.08619785e-01 -6.00560367e-01 -2.89734125e-01 -1.02302432e+00
8.17588568e-01 -1.44577205e+00 2.61616614e-02 -1.38753068e+00
2.46734366e-01 -6.05255723e-01 -4.80491608e-01 5.55279195e-01
-6.18515670e-01 2.29846090e-01 1.04999967e-01 -1.46262988e-01
-3.84888500e-01 4.15919304e-01 9.86412823e-01 -4.91795123e-01
-1.68831879e-03 -3.51813644e-01 -5.14867365e-01 5.62902033e-01
1.04965031e+00 -5.92892230e-01 -3.78597863e-02 -9.75931406e-01
3.77548277e-01 -4.83372897e-01 -3.48599076e-01 -7.40291119e-01
1.63498238e-01 -6.23013377e-02 -3.14064026e-01 -3.82985204e-01
4.51588891e-02 -3.80889714e-01 -2.09866270e-01 1.64225966e-01
-2.19065115e-01 7.70565808e-01 3.30856889e-01 -3.69633287e-01
-4.03518319e-01 -4.48452264e-01 4.45105165e-01 -1.30257845e-01
-6.16626740e-01 3.43222380e-01 -3.05264145e-01 4.71614301e-01
3.62361282e-01 3.71695101e-01 -2.81273037e-01 -4.50168811e-02
-2.42925689e-01 2.76309907e-01 4.04503256e-01 7.93972373e-01
-1.71025842e-01 -1.65732706e+00 -1.42319894e+00 1.83350906e-01
4.41831917e-01 -3.61842006e-01 -2.49882355e-01 8.06249082e-01
-3.52009922e-01 4.99938071e-01 -2.22577482e-01 -5.23423493e-01
-1.22971439e+00 4.89480972e-01 1.09439626e-01 -9.90270376e-01
-2.14059040e-01 9.89117146e-01 -7.19604194e-02 -1.35876215e+00
-8.09008330e-02 -2.15347216e-01 -3.95309851e-02 -2.27789003e-02
2.84213573e-01 -1.36102349e-01 3.07065994e-01 -8.77965629e-01
-4.81235147e-01 8.31019402e-01 -5.41981399e-01 -2.11527094e-01
1.18995118e+00 -4.04861057e-03 -5.54576159e-01 4.71285611e-01
9.62981582e-01 6.90347254e-01 -4.80212361e-01 -4.26258534e-01
2.58185536e-01 3.38320136e-02 -2.51760364e-01 -6.56060576e-01
-7.16232777e-01 1.13790751e+00 2.70894557e-01 -4.45307642e-01
7.63810515e-01 -1.01945549e-01 1.14513540e+00 4.54359770e-01
3.87370318e-01 -1.20927691e+00 -3.00344855e-01 1.17617190e+00
6.98277473e-01 -1.27344477e+00 -1.19801454e-01 -2.11220771e-01
-4.15350735e-01 9.49969828e-01 8.78927827e-01 -1.51555538e-01
3.16644013e-01 6.83846593e-01 5.38410127e-01 3.15379381e-01
-6.90589070e-01 -3.32269907e-01 3.89398247e-01 4.81847525e-01
9.73532081e-01 4.00758386e-01 -7.78806686e-01 6.43566132e-01
-8.86116743e-01 -6.44912839e-01 2.22227111e-01 3.55338007e-01
-2.40996018e-01 -2.09751964e+00 -2.45640397e-01 1.50520122e-02
-4.60439742e-01 -9.94112909e-01 -3.39571625e-01 9.57028329e-01
1.53023645e-01 5.58322787e-01 2.08914921e-01 -4.16293234e-01
1.33859813e-01 3.52180868e-01 7.33102798e-01 -6.74778342e-01
-7.08417475e-01 -1.83212116e-01 3.98301005e-01 -3.42898846e-01
-1.17716566e-01 -5.62898695e-01 -7.09721148e-01 -2.60511041e-01
-1.80049077e-01 6.11247346e-02 5.83061278e-01 1.13768470e+00
3.62139866e-02 3.65634799e-01 5.33163011e-01 -7.08250225e-01
-6.28456414e-01 -1.37521267e+00 3.35731059e-02 2.30717659e-01
1.30787939e-01 -2.32908782e-02 -1.05324462e-01 -1.85721368e-01]
|
[11.129159927368164, 10.123950958251953]
|
8bbb079a-d199-41b1-ad17-0d0279d0eb4a
|
irnlp-daiict-lt-edi-eacl2021-hope-speech
| null | null |
https://aclanthology.org/2021.ltedi-1.15
|
https://aclanthology.org/2021.ltedi-1.15.pdf
|
IRNLP_DAIICT@LT-EDI-EACL2021: Hope Speech detection in Code Mixed text using TF-IDF Char N-grams and MuRIL
|
This paper presents the participation of the IRNLP_DAIICT team from Information Retrieval and Natural Language Processing lab at DA-IICT, India in LT-EDI@EACL2021 Hope Speech Detection task. The aim of this shared task is to identify hope speech from a code-mixed data-set of YouTube comments. The task is to classify comments into Hope Speech, Non Hope speech or Not in language, for three languages: English, Malayalam-English and Tamil-English. We use TF-IDF character n-grams and pretrained MuRIL embeddings for text representation and Logistic Regression and Linear SVM for classification. Our best approach achieved second, eighth and fifth rank with weighted F1 score of 0.92, 0.75 and 0.57 in English, Malayalam-English and Tamil-English on test dataset respectively
|
['Prasenjit Majumder', 'Shripad Bhat', 'Bhargav Dave']
| null | null | null | null |
eacl-ltedi-2021-4
|
['hope-speech-detection']
|
['natural-language-processing']
|
[-4.90908533e-01 1.61005482e-01 -3.87103021e-01 -6.63555190e-02
-1.08126783e+00 -8.10289800e-01 7.80072391e-01 5.68060517e-01
-4.66212988e-01 5.95767796e-01 9.30541635e-01 -8.56966376e-01
9.52489600e-02 -2.92694181e-01 -6.99924529e-02 -1.14502341e-01
8.57926682e-02 4.07571137e-01 -1.68989137e-01 -1.82553485e-01
4.69528288e-01 2.04506099e-01 -1.03450620e+00 8.66064906e-01
6.50704503e-01 7.68645823e-01 2.05005437e-01 1.43204248e+00
-2.54290164e-01 1.45432258e+00 -9.79635939e-02 -5.17962694e-01
-7.88340159e-03 -6.45169690e-02 -1.26691449e+00 -1.90647170e-01
1.54882014e-01 4.30939160e-02 -7.18610883e-01 1.10567522e+00
4.43830729e-01 2.71973610e-01 1.04744530e+00 -1.08612525e+00
-1.08531153e+00 9.42172527e-01 -2.78991193e-01 4.99272048e-01
5.03630579e-01 -4.20801699e-01 1.20050383e+00 -1.38167107e+00
5.72055340e-01 1.26496208e+00 5.48772156e-01 4.79020238e-01
-5.36897421e-01 -5.68589389e-01 -5.82541466e-01 3.44016366e-02
-1.26614404e+00 -5.40993392e-01 4.50786442e-01 -6.44660115e-01
1.50121498e+00 3.84236813e-01 -2.73240115e-02 9.99500930e-01
3.77646655e-01 1.16013420e+00 1.07869136e+00 -6.19622231e-01
-3.18722092e-02 8.48311067e-01 2.83766776e-01 8.24899733e-01
-5.20569801e-01 -5.20915806e-01 -4.97901797e-01 -2.25474089e-01
-2.63304971e-02 -8.33934546e-02 4.03998010e-02 5.15193462e-01
-1.05897570e+00 1.36552739e+00 -1.39640287e-01 4.83503640e-01
-2.87320375e-01 -3.71944189e-01 8.77077520e-01 4.76485938e-01
7.95314968e-01 3.23250182e-02 -7.80952930e-01 -6.71586037e-01
-9.35772300e-01 -2.59799529e-02 7.79390991e-01 1.01502502e+00
1.18146747e-01 1.32019818e-01 -3.34065259e-02 1.58937550e+00
7.92016149e-01 6.38725996e-01 1.20694482e+00 -4.02199984e-01
7.46818662e-01 5.00431001e-01 -1.27397671e-01 -9.50713933e-01
-1.97090775e-01 1.46546690e-02 -5.78367531e-01 -3.33065540e-01
-2.90130526e-02 -4.32794392e-01 -8.63097608e-01 9.83904839e-01
-2.77578652e-01 -2.83384204e-01 8.00572455e-01 3.88849735e-01
1.36133897e+00 1.35350049e+00 9.79561433e-02 -1.55710250e-01
1.50019670e+00 -1.59193909e+00 -6.61912382e-01 -1.99128151e-01
7.74943054e-01 -1.44777703e+00 1.06920636e+00 3.59139532e-01
-1.08593595e+00 -3.48943979e-01 -6.46851718e-01 -2.65835136e-01
-7.63135850e-01 7.11894035e-01 3.50961596e-01 7.97642946e-01
-9.68071043e-01 3.80819440e-02 -3.27888101e-01 -6.92643046e-01
3.16937655e-01 6.86713606e-02 -5.49599946e-01 -4.09130864e-02
-1.00851214e+00 6.21507943e-01 2.07348838e-01 -6.04211390e-01
-7.79262304e-01 -4.78976935e-01 -8.90754640e-01 -1.77212194e-01
-5.91129005e-01 5.28927803e-01 1.20752048e+00 -7.48951554e-01
-1.13232768e+00 1.22963357e+00 -2.78312325e-01 -6.49299204e-01
2.76428461e-01 5.62484667e-04 -1.02728391e+00 7.45849535e-02
1.17570013e-01 5.75272381e-01 5.23613214e-01 -5.93594790e-01
-8.04723382e-01 -2.60045350e-01 -2.56776303e-01 1.72633678e-01
-7.78532386e-01 6.36647820e-01 3.43716778e-02 -6.39739990e-01
-1.40544310e-01 -8.94964516e-01 2.65683353e-01 -5.82226753e-01
-4.45656389e-01 -7.06229925e-01 1.09536266e+00 -1.25362265e+00
1.36989892e+00 -2.22851324e+00 -3.40022296e-01 -2.99436331e-01
2.03263592e-02 4.00901228e-01 -1.10486843e-01 8.77096772e-01
-2.05815166e-01 3.95461529e-01 3.18390310e-01 -3.15791517e-01
2.88374200e-02 1.22520059e-01 -4.93693322e-01 3.92276555e-01
4.93674949e-02 5.31738043e-01 -7.98397005e-01 -4.66086090e-01
1.02312647e-01 5.70805073e-01 -2.22268611e-01 1.60512328e-01
3.32484782e-01 -2.56658524e-01 -2.56748557e-01 5.54739118e-01
5.09579241e-01 2.13541016e-01 -2.71217436e-01 3.50986645e-02
-4.63855147e-01 7.03432918e-01 -7.46087193e-01 9.84758019e-01
-7.89000452e-01 1.18044400e+00 -1.18688583e-01 -9.20022368e-01
1.06604362e+00 8.14383268e-01 1.93722337e-01 -4.22038823e-01
3.12868863e-01 3.98815989e-01 -2.05545183e-02 -8.89904499e-01
9.34403479e-01 5.28170317e-02 -1.29470304e-01 2.67891437e-01
3.04762512e-01 1.10018283e-01 1.28022119e-01 5.13521671e-01
9.72638249e-01 -7.77224958e-01 4.05750930e-01 -4.98643726e-01
8.39502215e-01 -8.47616792e-02 5.74567765e-02 6.02310181e-01
-5.24056792e-01 6.87671363e-01 4.20177430e-01 -2.42284119e-01
-1.36280465e+00 -8.87863219e-01 -4.63680297e-01 1.42763329e+00
-5.59210539e-01 -6.53771281e-01 -4.01589870e-01 -9.92371917e-01
-2.72136480e-01 1.01452982e+00 -2.59305626e-01 -1.06350593e-01
-2.86041349e-01 -2.72542953e-01 8.57824445e-01 1.03728406e-01
4.12608445e-01 -1.08200645e+00 2.44891010e-02 1.07460789e-01
-2.63373196e-01 -1.13860321e+00 -8.25281501e-01 3.66180152e-01
-5.80169797e-01 -5.29716790e-01 -8.71202528e-01 -1.33428502e+00
4.03739125e-01 -1.54803008e-01 7.20821440e-01 -2.77674407e-01
-3.57841700e-01 3.60980958e-01 -7.97508955e-01 -3.03812921e-01
-6.99791074e-01 -3.37692583e-03 -5.38425483e-02 -1.32347986e-01
7.54445910e-01 1.04088537e-01 -1.38878733e-01 -7.59510770e-02
-5.15367687e-01 -1.99134871e-01 2.60849357e-01 8.86226356e-01
-7.77845010e-02 8.26689750e-02 6.17258549e-01 -8.01628172e-01
1.07785952e+00 -9.60214734e-01 -1.70577709e-02 1.53897330e-01
-4.89433795e-01 -1.21134505e-01 7.80026019e-01 -4.76747513e-01
-8.28087151e-01 1.30555108e-01 -5.96194327e-01 5.76463267e-02
-1.50996670e-01 8.58365834e-01 2.53849775e-01 4.05117482e-01
6.00106359e-01 6.34069383e-01 -5.22557855e-01 -4.20678079e-01
1.21075988e-01 1.89820886e+00 2.61887461e-01 -2.46172398e-01
3.30672354e-01 -7.63845071e-02 -8.88465464e-01 -1.45553136e+00
-4.64214534e-01 -1.08595431e+00 -4.77772862e-01 -8.85340497e-02
1.14632452e+00 -1.14887905e+00 -1.42248780e-01 5.10541260e-01
-1.34742773e+00 1.28968194e-01 1.72100663e-01 5.67186534e-01
-1.08807527e-01 4.26496893e-01 -9.92435753e-01 -1.25933444e+00
-6.55937731e-01 -1.18350697e+00 7.74458826e-01 1.20811313e-01
-3.68893564e-01 -1.18162489e+00 3.07056159e-02 8.36439252e-01
4.25356984e-01 -4.37595159e-01 1.02130866e+00 -1.47974896e+00
4.87634748e-01 -5.36231875e-01 -3.76301438e-01 7.29052722e-01
1.72648236e-01 2.10571662e-01 -9.22662795e-01 -1.45819739e-01
-1.11451469e-01 -7.81650186e-01 6.43339336e-01 8.52628201e-02
8.01875591e-01 -8.06460798e-01 1.90588295e-01 -4.49231826e-02
1.35885608e+00 4.20821637e-01 5.55651546e-01 1.76723614e-01
3.68155926e-01 3.07035357e-01 3.60274732e-01 6.47422373e-01
5.54941535e-01 1.37142181e-01 -3.54058519e-02 5.60830653e-01
-9.94237661e-02 -4.16537046e-01 7.79819369e-01 1.55268264e+00
6.03383601e-01 -6.24510109e-01 -1.39226842e+00 9.95797276e-01
-1.42390263e+00 -7.99909711e-01 -6.62867069e-01 1.79360807e+00
8.92924488e-01 1.96148939e-02 2.12889746e-01 2.00045094e-01
4.11334276e-01 1.29298791e-01 2.47956634e-01 -1.20992970e+00
1.11889079e-01 3.84411365e-02 5.95753312e-01 7.11674392e-01
-1.30111778e+00 9.77375090e-01 4.85781717e+00 1.10985529e+00
-1.14213240e+00 4.60401297e-01 6.92480505e-01 3.11137110e-01
-1.77170455e-01 -1.54591098e-01 -9.17442977e-01 6.03899240e-01
1.82460225e+00 -4.06897217e-01 3.26596200e-01 1.03992248e+00
1.66651845e-01 1.30350217e-01 -5.66363931e-01 1.10636890e+00
3.99303555e-01 -1.30933201e+00 -1.91374242e-01 -2.13569582e-01
7.88288355e-01 8.07198226e-01 2.59674013e-01 6.67319655e-01
4.99455005e-01 -1.34459829e+00 5.43337345e-01 4.71938998e-02
6.20021582e-01 -1.01346958e+00 8.88356805e-01 7.37308502e-01
-1.12306762e+00 -2.94297874e-01 -5.43186486e-01 1.66360125e-01
-2.33074710e-01 3.31621319e-01 -1.10126758e+00 7.99928675e-04
6.89921916e-01 8.38491499e-01 -4.61615473e-01 6.56181872e-01
1.07168339e-01 9.36811745e-01 -1.45263858e-02 -6.82394207e-01
6.86928809e-01 3.68181616e-01 3.60409051e-01 1.91672552e+00
2.79736757e-01 -1.02834605e-01 2.87838131e-01 2.63518602e-01
-2.96239614e-01 4.87299353e-01 -6.40331566e-01 -6.94115937e-01
5.07985890e-01 1.33348548e+00 -6.56990469e-01 -3.69547814e-01
-4.56231296e-01 8.62175405e-01 4.16500643e-02 1.38255702e-02
-5.41794658e-01 -9.21488166e-01 2.91289359e-01 -1.67850792e-01
5.22299409e-02 -3.98099303e-01 -3.79784293e-02 -1.14091551e+00
-3.85920554e-01 -1.17911386e+00 2.14585319e-01 -6.59923375e-01
-1.40294981e+00 9.74738836e-01 -3.36933494e-01 -9.79787469e-01
-1.56220749e-01 -8.47069204e-01 -6.38231158e-01 9.69682693e-01
-1.22005272e+00 -1.06654072e+00 2.80852854e-01 5.75372100e-01
1.46813238e+00 -9.95263040e-01 8.08953345e-01 6.57664001e-01
-3.82645249e-01 7.78372049e-01 6.86107397e-01 6.08992815e-01
4.33409184e-01 -1.24693012e+00 8.22171271e-02 5.01385450e-01
3.16752642e-01 5.05246520e-01 6.67273641e-01 -5.16095519e-01
-1.35371137e+00 -1.15116334e+00 1.96176887e+00 -5.03325224e-01
1.14262342e+00 -5.57658195e-01 -4.66262400e-01 4.95825082e-01
7.57727146e-01 -4.03174162e-02 7.34685540e-01 -3.27862442e-01
-3.62820774e-01 3.84095967e-01 -1.35643613e+00 4.07227337e-01
1.00959629e-01 -1.20642972e+00 -6.54093921e-01 9.28321898e-01
5.25188029e-01 -8.30017924e-02 -9.13612306e-01 -3.92542571e-01
3.07243168e-01 -4.32324171e-01 6.81117833e-01 -8.84893894e-01
9.57843006e-01 1.88172728e-01 -7.57071972e-01 -1.10812056e+00
1.62825420e-01 -3.40335369e-01 3.90165210e-01 1.65585518e+00
8.77872527e-01 -3.24675828e-01 6.28400922e-01 1.65445507e-01
-1.85468063e-01 -5.07508337e-01 -1.09523869e+00 -6.34323418e-01
5.66034019e-01 -6.58691585e-01 -3.95592779e-01 1.16816509e+00
4.68199402e-01 6.36308670e-01 -4.73500133e-01 -2.83670537e-02
2.15860978e-01 -7.93616831e-01 3.74505550e-01 -9.11787391e-01
-1.10504348e-02 -8.27075392e-02 -4.21494961e-01 -9.98376906e-01
5.39680719e-01 -1.24199486e+00 1.19102091e-01 -1.52784240e+00
2.17142627e-01 -1.55189887e-01 6.52463213e-02 5.80361068e-01
4.53500479e-01 3.02531142e-02 6.42133802e-02 1.90749258e-01
-3.90727460e-01 3.95672500e-01 3.85713518e-01 -4.71863925e-01
2.52471417e-01 1.72810424e-02 -3.54027450e-01 6.00306213e-01
7.11769164e-01 -8.25325727e-01 -5.76488435e-01 -2.92250603e-01
-8.74062069e-04 3.02434146e-01 -1.13100313e-01 -7.30344713e-01
2.98921704e-01 5.72039969e-02 -1.74083970e-02 -8.00295472e-01
1.95273444e-01 -6.74028099e-01 -5.87287366e-01 4.36870366e-01
-9.84231532e-01 4.77084219e-01 8.22748914e-02 2.21908912e-01
-4.25648630e-01 -9.06921387e-01 7.29389250e-01 -2.38831043e-01
-4.55935031e-01 -2.99380859e-03 -1.18182540e+00 3.10614139e-01
7.11482823e-01 1.79359205e-02 -4.97255355e-01 -5.66593230e-01
-5.53193986e-01 2.25812327e-02 -1.21489391e-01 9.61535811e-01
9.36645091e-01 -1.18289733e+00 -1.05991387e+00 2.67756879e-01
2.23688573e-01 -9.95716512e-01 -2.54332013e-02 8.22121680e-01
-7.50146568e-01 7.00086236e-01 1.75576210e-01 -1.66193604e-01
-1.78432178e+00 2.46877838e-02 -6.93533570e-02 -1.67218298e-01
-1.42719164e-01 1.10640502e+00 -5.59716761e-01 -9.75925684e-01
3.35706174e-01 -1.29307285e-01 -6.41683877e-01 5.62526397e-02
5.55775881e-01 2.29009807e-01 1.88728541e-01 -1.22376752e+00
-4.65695947e-01 1.91843361e-01 -3.45282525e-01 -1.96707264e-01
1.44208157e+00 -1.67398065e-01 -2.79292643e-01 6.65147483e-01
2.08622980e+00 4.44331527e-01 -1.57392815e-01 1.74651533e-01
3.82392228e-01 -2.54743099e-01 4.16658819e-01 -6.86098456e-01
-7.39969611e-01 1.13061368e+00 9.39075172e-01 5.34354746e-01
4.50764030e-01 1.35613516e-01 9.60744500e-01 4.12237793e-01
-2.96609342e-01 -1.31812966e+00 6.73881322e-02 1.17992783e+00
7.69874692e-01 -1.42430973e+00 -4.57417339e-01 1.91152796e-01
-1.09502065e+00 1.61136973e+00 1.54999167e-01 -7.70050809e-02
1.08440065e+00 1.05889857e-01 2.42568284e-01 1.12631306e-01
-1.03955925e+00 3.01731735e-01 4.50399578e-01 6.93874955e-01
1.21230781e+00 -8.39065239e-02 -4.63598043e-01 4.59899247e-01
-4.11703497e-01 -2.27471814e-01 9.14528012e-01 6.56930745e-01
-7.20361829e-01 -8.03722858e-01 1.07984081e-01 7.44759440e-01
-1.18443382e+00 -5.28425157e-01 -3.82290989e-01 5.20686150e-01
-2.17892259e-01 1.50696528e+00 -4.38780263e-02 -6.04543090e-01
-1.25591457e-01 3.10900331e-01 -5.23160160e-01 -7.48782158e-01
-8.82083178e-01 1.47451580e-01 5.29175818e-01 3.21292847e-01
-1.29665723e-02 -6.90599144e-01 -1.45645189e+00 -3.11689436e-01
-2.00728074e-01 8.88032377e-01 1.06925833e+00 8.14688623e-01
1.04710951e-01 1.89271405e-01 7.57894576e-01 -8.64387602e-02
-4.46685046e-01 -1.08374059e+00 -5.35992503e-01 2.28157729e-01
4.69732523e-01 6.86318427e-02 -6.93439901e-01 1.13347836e-01]
|
[9.367328643798828, 10.686226844787598]
|
b929e4da-dde7-40e1-9799-25181f8118d0
|
finding-strength-in-weakness-learning-to
|
1911.02182
| null |
https://arxiv.org/abs/1911.02182v2
|
https://arxiv.org/pdf/1911.02182v2.pdf
|
Finding Strength in Weakness: Learning to Separate Sounds with Weak Supervision
|
While there has been much recent progress using deep learning techniques to separate speech and music audio signals, these systems typically require large collections of isolated sources during the training process. When extending audio source separation algorithms to more general domains such as environmental monitoring, it may not be possible to obtain isolated signals for training. Here, we propose objective functions and network architectures that enable training a source separation system with weak labels. In this scenario, weak labels are defined in contrast with strong time-frequency (TF) labels such as those obtained from isolated sources, and refer either to frame-level weak labels where one only has access to the time periods when different sources are active in an audio mixture, or to clip-level weak labels that only indicate the presence or absence of sounds in an entire audio clip. We train a separator that estimates a TF mask for each type of sound event, using a sound event classifier as an assessor of the separator's performance to bridge the gap between the TF-level separation and the ground truth weak labels only available at the frame or clip level. Our objective function requires the classifier applied to a separated source to assign high probability to the class corresponding to that source and low probability to all other classes. The objective function also enforces that the separated sources sum up to the mixture. We benchmark the performance of our algorithm using synthetic mixtures of overlapping events created from a database of sounds recorded in urban environments. Compared to training a network using isolated sources, our model achieves somewhat lower but still significant SI-SDR improvement, even in scenarios with significant sound event overlap.
|
['Fatemeh Pishdadian', 'Jonathan Le Roux', 'Gordon Wichern']
|
2019-11-06
| null | null | null | null |
['audio-source-separation']
|
['audio']
|
[ 5.92010617e-01 -1.94351777e-01 1.27366811e-01 -1.14016376e-01
-1.31962633e+00 -6.97574615e-01 3.58146131e-01 3.49789768e-01
-2.55643517e-01 6.59882247e-01 1.66568771e-01 -6.12076893e-02
-2.15199724e-01 -5.94164312e-01 -6.08156264e-01 -8.99050355e-01
-2.98456281e-01 2.72068024e-01 3.12243074e-01 1.61688030e-01
-3.26556772e-01 4.65674073e-01 -1.81580734e+00 5.59740901e-01
5.93940735e-01 1.15591991e+00 1.28985539e-01 1.01219821e+00
5.45649789e-03 7.42388070e-01 -1.25950515e+00 2.43771181e-01
3.40938449e-01 -7.33259797e-01 -3.51888925e-01 -1.00829370e-01
5.15048265e-01 -4.15735282e-02 -1.09244332e-01 8.66868377e-01
6.97881341e-01 2.13097915e-01 5.72460592e-01 -1.27007771e+00
2.56979674e-01 8.48097861e-01 -3.08387250e-01 3.54434311e-01
3.95772040e-01 -8.78497362e-02 1.03727734e+00 -6.67886674e-01
4.78612632e-02 8.70480359e-01 6.60875678e-01 7.08521530e-02
-1.39862597e+00 -7.82885075e-01 1.68883875e-02 3.20515186e-02
-1.24267077e+00 -8.11593473e-01 8.59954536e-01 -4.53801662e-01
7.39445508e-01 4.07000303e-01 3.61532837e-01 1.06856644e+00
-3.21728230e-01 6.64596677e-01 7.03622580e-01 -5.32393038e-01
4.36621636e-01 2.03753188e-02 7.94250667e-02 -1.09646313e-01
-1.78356543e-01 1.40909582e-01 -7.64392257e-01 -2.51082987e-01
4.07739818e-01 -3.93122733e-01 -3.97440553e-01 1.69615015e-01
-1.09492207e+00 4.79073018e-01 1.29561365e-01 5.76898754e-01
-4.03274000e-01 1.04591928e-01 2.30627909e-01 4.19487923e-01
4.81948525e-01 3.44900042e-01 -3.89608204e-01 1.71577446e-02
-1.49890888e+00 2.79263198e-01 1.03089130e+00 5.23928106e-01
6.20587885e-01 5.45461714e-01 -1.99248359e-01 1.05082119e+00
-2.78806090e-02 4.96996999e-01 2.13301033e-01 -1.08252454e+00
2.52179295e-01 1.09743187e-02 2.70520598e-01 -7.82690346e-01
-2.38283232e-01 -9.36924756e-01 -5.40791273e-01 4.13034618e-01
7.17488348e-01 -3.24245691e-01 -8.59133899e-01 2.01663303e+00
2.07572982e-01 6.93406284e-01 1.04189478e-01 8.25464129e-01
5.73752463e-01 8.52013171e-01 -1.68300107e-01 -4.77305651e-01
1.06653750e+00 -5.37090659e-01 -5.76203287e-01 -3.02488506e-01
-1.02538072e-01 -9.15815175e-01 7.38043785e-01 6.10105932e-01
-1.03065109e+00 -8.89477551e-01 -9.99603093e-01 5.71713090e-01
-1.56585723e-01 1.69074491e-01 5.40004820e-02 7.37566292e-01
-6.61279678e-01 4.95043755e-01 -7.25436389e-01 5.57903014e-02
5.37560619e-02 2.43391126e-01 -6.11432828e-02 3.35490733e-01
-1.32678437e+00 4.87607002e-01 6.81446120e-02 3.08584832e-02
-1.33740258e+00 -8.22303057e-01 -6.12995684e-01 4.67620373e-01
4.20625895e-01 -7.79298618e-02 1.39766324e+00 -1.40566742e+00
-1.17408550e+00 4.57317293e-01 -2.68064767e-01 -6.24495208e-01
3.19480717e-01 -1.50195181e-01 -9.76875305e-01 2.98674375e-01
2.08873808e-01 3.80957484e-01 1.04886854e+00 -1.33378744e+00
-9.69334781e-01 1.40357852e-01 2.94581018e-02 8.82170051e-02
-6.65552467e-02 3.04405153e-01 -1.75745159e-01 -7.12458968e-01
8.35703313e-02 -7.17370093e-01 1.51297405e-01 -1.93236366e-01
-4.28656757e-01 4.93442304e-02 7.60144949e-01 -6.35202169e-01
1.14149690e+00 -2.44196081e+00 -6.17615394e-02 2.66088873e-01
-2.18912754e-02 1.62140161e-01 -1.55498177e-01 1.91762954e-01
-3.13601643e-01 -2.69497752e-01 -4.28061545e-01 -4.54854369e-01
9.05797258e-02 3.20875235e-02 -6.32232726e-01 3.19058686e-01
2.54368067e-01 5.01489826e-02 -9.44568217e-01 -2.69398838e-01
8.80612433e-02 5.72598994e-01 -3.55699658e-01 2.87614405e-01
-2.27268517e-01 4.65269744e-01 1.08463638e-01 4.57061201e-01
4.55238163e-01 1.93144754e-01 -4.43855859e-02 -1.53990671e-01
-9.50717106e-02 6.10457063e-01 -1.92413163e+00 1.37101984e+00
-4.59188133e-01 8.10359538e-01 6.96147978e-01 -1.03134108e+00
8.06384444e-01 7.99847841e-01 6.40226424e-01 -2.21349105e-01
-1.06830464e-03 4.35735255e-01 3.03156108e-01 -3.38326901e-01
2.76203692e-01 -3.40493023e-01 -1.05130114e-01 3.58494848e-01
2.97726631e-01 -1.78373829e-01 2.82808036e-01 -1.41400829e-01
1.21353090e+00 -8.56962577e-02 -1.49534255e-01 -2.69092247e-02
4.46695775e-01 -4.35099244e-01 6.04805589e-01 8.59865606e-01
-1.24154717e-01 7.66827524e-01 3.43959540e-01 1.88587219e-01
-5.58087289e-01 -1.47444439e+00 -2.47594714e-01 1.37924802e+00
-7.40763247e-02 -3.07960838e-01 -6.46490216e-01 -3.41520041e-01
-7.67207295e-02 8.41683626e-01 -2.07112208e-01 -1.74394831e-01
-5.72446406e-01 -6.71966195e-01 9.66144443e-01 2.49622315e-01
8.63777772e-02 -1.05496037e+00 -5.77149510e-01 4.02642459e-01
-3.92108947e-01 -9.27480400e-01 -3.57484490e-01 9.24995244e-01
-3.64806920e-01 -8.52060437e-01 -5.47185540e-01 -6.19536221e-01
1.54088140e-01 7.14802742e-02 1.05868757e+00 -4.47968394e-01
-9.02095512e-02 2.19890967e-01 -3.49450469e-01 -6.39576972e-01
-5.85532129e-01 -1.44993648e-01 2.03696892e-01 4.33900326e-01
9.27724987e-02 -8.14536333e-01 -3.52225035e-01 4.44250971e-01
-8.96203518e-01 -4.91697252e-01 8.69376957e-02 5.08328259e-01
3.88135612e-01 5.68748772e-01 9.23689783e-01 -2.36699954e-01
4.28004056e-01 -6.54342115e-01 -3.71436924e-01 -1.07719466e-01
8.73712152e-02 -2.86662936e-01 7.52205551e-01 -8.32259417e-01
-8.84943545e-01 7.98496529e-02 -2.71808386e-01 -4.60451692e-01
-5.57174921e-01 4.86258298e-01 -3.97201210e-01 4.80389476e-01
8.53461862e-01 1.18420422e-01 -4.18665737e-01 -6.27024889e-01
-3.00178789e-02 8.26925635e-01 8.60379696e-01 -6.61030233e-01
5.66117942e-01 3.31903428e-01 -2.09351256e-01 -9.97057021e-01
-8.88752818e-01 -5.12437999e-01 -3.62357706e-01 -3.31163853e-01
5.61561108e-01 -9.03947294e-01 -2.38174632e-01 4.25608486e-01
-9.63742137e-01 -2.16320261e-01 -7.14887679e-01 7.41223931e-01
-3.29558313e-01 -2.80897748e-02 -3.13308805e-01 -1.21691227e+00
1.80490002e-01 -1.03515506e+00 1.09440506e+00 8.29457268e-02
-4.82244223e-01 -7.15295315e-01 7.31076449e-02 -7.83679336e-02
2.39669442e-01 2.42341995e-01 6.18654132e-01 -1.00199044e+00
-1.15588605e-01 -4.03132319e-01 3.63787562e-01 6.95117176e-01
4.08691823e-01 6.41388968e-02 -1.63493979e+00 -7.19717965e-02
2.69402772e-01 -1.55848503e-01 9.28426027e-01 6.32101059e-01
7.81328499e-01 -9.09992680e-02 -2.75421232e-01 3.33152622e-01
9.92833316e-01 5.37095547e-01 4.18455124e-01 -1.51964530e-01
4.50101465e-01 6.65535092e-01 1.07128225e-01 1.54455513e-01
-2.92240649e-01 8.11837792e-01 1.98966101e-01 -3.44821244e-01
-4.06962186e-01 -6.58179075e-02 6.70867264e-01 4.71200645e-01
2.51309574e-01 -6.08032107e-01 -5.82997441e-01 7.12276518e-01
-1.44876587e+00 -1.31242883e+00 -6.93878606e-02 2.61781883e+00
1.06030440e+00 4.99574870e-01 3.90812039e-01 8.78984511e-01
9.00619626e-01 2.43426755e-01 -4.76269186e-01 -1.03614517e-01
-3.81195933e-01 5.51859081e-01 6.80268481e-02 6.74219310e-01
-1.33072603e+00 3.35950553e-01 6.39336634e+00 9.57884431e-01
-1.26615477e+00 1.05478078e-01 3.08388650e-01 -6.68989539e-01
-4.56839949e-02 -1.34674117e-01 -4.80632961e-01 6.70928359e-01
1.24124765e+00 5.02397195e-02 4.52493280e-01 3.97441387e-01
3.57934803e-01 -3.29911351e-01 -1.34872448e+00 7.91382611e-01
-9.03388113e-02 -7.91013300e-01 -3.88385564e-01 -1.77109674e-01
4.68245447e-01 3.72545198e-02 -3.88420676e-03 1.90733686e-01
2.48338461e-01 -1.06304765e+00 1.25222743e+00 2.81521469e-01
6.19881690e-01 -7.09653616e-01 4.04779792e-01 5.49549520e-01
-1.44047880e+00 -1.87233418e-01 1.37817133e-02 -1.30148917e-01
2.12606400e-01 9.55516636e-01 -8.38733613e-01 6.38512611e-01
5.92970610e-01 2.85627812e-01 -1.87895279e-02 1.22949576e+00
-1.86189130e-01 1.20486569e+00 -7.35640168e-01 5.12289405e-01
-7.20949695e-02 1.28831342e-01 9.59163547e-01 1.53956223e+00
4.86900508e-01 -2.99843252e-01 4.12191361e-01 8.19815218e-01
1.51910901e-01 -2.64073312e-01 -3.97386521e-01 2.12942928e-01
6.82288527e-01 1.07746208e+00 -8.30513000e-01 -3.52468163e-01
-2.90432721e-01 5.42573571e-01 -2.73679763e-01 5.67121565e-01
-8.83145213e-01 -5.47825813e-01 8.03426445e-01 2.05965698e-01
3.23473513e-01 -6.06737882e-02 -1.44700840e-01 -8.19576263e-01
-2.00963132e-02 -8.69086266e-01 3.68580163e-01 -7.48070002e-01
-1.16949093e+00 6.20047569e-01 -9.66765918e-03 -1.46175075e+00
-4.44945276e-01 -2.88720161e-01 -8.22085202e-01 1.12694657e+00
-1.23999977e+00 -6.67174637e-01 2.05979906e-02 5.79949081e-01
5.42030215e-01 -4.65111881e-02 8.30887914e-01 6.19298398e-01
-4.13689315e-01 4.05345082e-01 5.56541122e-02 1.64053142e-01
8.52572381e-01 -1.23376453e+00 -1.48655757e-01 9.77037847e-01
6.66320503e-01 2.53337651e-01 1.09650946e+00 -4.16628569e-01
-6.85181022e-01 -1.01324236e+00 7.30779827e-01 -3.35379124e-01
5.53951621e-01 -4.84532982e-01 -9.53599989e-01 3.45263958e-01
6.63539171e-02 2.81501766e-02 7.62445450e-01 -1.59800630e-02
-3.67514104e-01 -4.23989087e-01 -8.61815095e-01 8.12088996e-02
6.42775178e-01 -7.82821357e-01 -7.63272345e-01 1.94788694e-01
5.69045007e-01 -1.62680075e-01 -4.55287725e-01 3.49674821e-01
5.44939399e-01 -1.03971994e+00 1.07186949e+00 -3.61404717e-01
2.73145914e-01 -8.33784461e-01 -4.38955277e-01 -1.41976011e+00
-1.41456321e-01 -5.94097674e-01 -8.28274786e-02 1.58929396e+00
5.63308060e-01 -2.49893159e-01 3.80556732e-01 8.21643546e-02
-3.09296906e-01 -8.55570436e-02 -1.14882338e+00 -1.03853142e+00
-1.27666920e-01 -9.08832788e-01 5.36690235e-01 8.80209446e-01
-1.27130434e-01 2.82880247e-01 -3.11477304e-01 6.09872162e-01
4.69360262e-01 1.54636681e-01 4.55526173e-01 -1.42840636e+00
-7.54730105e-01 -5.18310070e-01 -2.67946720e-02 -7.09172368e-01
2.35873610e-01 -8.22008073e-01 5.06673396e-01 -1.26811659e+00
-3.60079497e-01 -4.20529395e-01 -7.24732399e-01 3.79171818e-01
-8.15869868e-02 2.98605949e-01 8.09908211e-02 2.13989690e-01
-2.31120288e-01 1.18677549e-01 5.32184660e-01 -3.82549137e-01
-4.56282049e-01 4.34897006e-01 -4.13022041e-01 9.07692194e-01
5.33550322e-01 -6.88594401e-01 -4.39657658e-01 -3.00499082e-01
5.31444661e-02 2.50646740e-01 5.71603119e-01 -1.55780160e+00
2.24337280e-01 -9.16334018e-02 5.11744976e-01 -3.56860995e-01
5.16303062e-01 -9.65597808e-01 5.60801625e-01 1.71220407e-01
-5.73572695e-01 -7.51388490e-01 3.53215575e-01 3.79514515e-01
-5.19998372e-01 -3.09489667e-01 8.22020352e-01 3.33326198e-02
-1.53968960e-01 -1.84481546e-01 -5.97945154e-01 6.59641065e-03
5.81392348e-01 -1.87942564e-01 3.51404473e-02 -5.67738771e-01
-1.07004273e+00 -1.93289503e-01 -4.52457170e-05 2.00056434e-01
2.02467054e-01 -1.18848550e+00 -8.40206981e-01 3.08433473e-01
-1.56379908e-01 -6.65792003e-02 1.08388968e-01 6.17575288e-01
1.57365322e-01 7.72549510e-02 8.60123560e-02 -6.56612337e-01
-1.18355978e+00 3.49332690e-01 6.20762944e-01 -2.89203264e-02
-3.17766100e-01 9.33310986e-01 3.85805339e-01 -1.01573929e-01
5.69130242e-01 -4.85004187e-01 -2.05548272e-01 4.62202877e-01
6.50896132e-01 5.48099697e-01 2.99013942e-01 -7.86534250e-01
-3.82845014e-01 3.08358639e-01 6.08025074e-01 -5.24874508e-01
1.14485347e+00 6.90079331e-02 6.29945099e-02 1.03458166e+00
1.03487647e+00 6.16883099e-01 -1.39087319e+00 -2.61554480e-01
-1.36347771e-01 -2.23837391e-01 1.37241513e-01 -1.00852919e+00
-1.00392640e+00 9.14424598e-01 7.02610672e-01 7.86044300e-01
1.38202441e+00 6.67061284e-02 5.75440049e-01 5.57510965e-02
1.06842235e-01 -9.22625422e-01 -1.07927509e-01 4.49097395e-01
6.86974704e-01 -7.05077529e-01 -3.95057410e-01 -9.32337344e-02
-2.19006404e-01 9.61490452e-01 1.86057717e-01 -1.57258645e-01
6.75749958e-01 8.53998542e-01 1.57203749e-01 1.45962089e-01
-5.47150970e-01 -3.17498088e-01 4.37902719e-01 5.66472292e-01
3.60104710e-01 1.77154504e-02 4.36099648e-01 8.00253272e-01
-3.33789796e-01 -1.43424243e-01 2.46280640e-01 6.95606887e-01
-6.76512420e-01 -9.79490280e-01 -8.56770277e-01 4.81850743e-01
-7.14236379e-01 -1.42183930e-01 -2.79025465e-01 2.70103037e-01
5.19395292e-01 1.49033248e+00 3.37795824e-01 -2.72383034e-01
4.90070552e-01 5.22728086e-01 2.73260087e-01 -6.75135672e-01
-7.45157421e-01 8.78345549e-01 3.28597724e-01 -1.61241993e-01
-5.61932623e-01 -7.37497151e-01 -1.23248804e+00 2.45093897e-01
-4.00751233e-01 4.08157885e-01 3.88653845e-01 8.90187204e-01
-4.61869501e-02 9.58828270e-01 6.26569808e-01 -1.03361821e+00
-2.05023929e-01 -1.05579817e+00 -8.47314358e-01 2.26578817e-01
9.06900823e-01 -5.14472246e-01 -8.25606704e-01 5.57907939e-01]
|
[15.216172218322754, 5.475631237030029]
|
2781f26f-476a-4a0b-85cd-3359d71c6205
|
deep-learning-systems-for-advanced-driving
|
2304.06041
| null |
https://arxiv.org/abs/2304.06041v1
|
https://arxiv.org/pdf/2304.06041v1.pdf
|
Deep Learning Systems for Advanced Driving Assistance
|
Next generation cars embed intelligent assessment of car driving safety through innovative solutions often based on usage of artificial intelligence. The safety driving monitoring can be carried out using several methodologies widely treated in scientific literature. In this context, the author proposes an innovative approach that uses ad-hoc bio-sensing system suitable to reconstruct the physio-based attentional status of the car driver. To reconstruct the car driver physiological status, the author proposed the use of a bio-sensing probe consisting of a coupled LEDs at Near infrared (NiR) spectrum with a photodetector. This probe placed over the monitored subject allows to detect a physiological signal called PhotoPlethysmoGraphy (PPG). The PPG signal formation is regulated by the change in oxygenated and non-oxygenated hemoglobin concentration in the monitored subject bloodstream which will be directly connected to cardiac activity in turn regulated by the Autonomic Nervous System (ANS) that characterizes the subject's attention level. This so designed car driver drowsiness monitoring will be combined with further driving safety assessment based on correlated intelligent driving scenario understanding.
|
['Francesco Rundo']
|
2023-04-05
| null | null | null | null |
['photoplethysmography-ppg']
|
['medical']
|
[ 1.91539135e-02 3.43364179e-01 1.15946926e-01 -2.32657567e-01
2.97029167e-01 -2.62294292e-01 1.85121551e-01 -1.49301067e-01
-3.13286722e-01 9.11795914e-01 2.64304150e-02 2.49009044e-03
1.58877119e-01 -4.75537837e-01 -1.03298277e-01 -9.79830325e-01
5.69796443e-01 -1.74351439e-01 -1.63207635e-01 -3.54549289e-01
4.13753510e-01 7.01784968e-01 -2.14368773e+00 -2.47459233e-01
8.60679746e-01 1.12660241e+00 3.28445733e-02 5.65991759e-01
8.44873115e-02 7.22776949e-01 -4.82292742e-01 1.60325035e-01
2.24372879e-01 -8.04336071e-01 -9.17443447e-03 -3.91153127e-01
-1.71148777e-01 -1.26023039e-01 -2.11845487e-01 9.55046475e-01
4.50833380e-01 1.61241248e-01 4.80875224e-01 -1.26513815e+00
-3.28699261e-01 3.46520059e-02 6.06943145e-02 5.64626813e-01
2.34005675e-01 6.55076742e-01 -4.06716056e-02 -3.87244850e-01
1.90232247e-01 8.60399306e-01 3.28178823e-01 8.97724628e-01
-9.15820777e-01 -6.71002269e-01 -5.50959706e-01 1.01142800e+00
-1.28592515e+00 -5.65311730e-01 1.16583157e+00 -4.54889923e-01
5.55505693e-01 4.18479770e-01 8.28367352e-01 8.59548271e-01
8.34408045e-01 -1.41945288e-01 1.63081038e+00 -9.22542140e-02
2.85272866e-01 9.38056946e-01 5.20101249e-01 4.99841630e-01
3.88558567e-01 5.32196760e-01 -3.82623494e-01 1.47684500e-01
-2.05920771e-01 -5.07081598e-02 -3.48008752e-01 3.65058064e-01
-6.17350698e-01 2.95892954e-01 1.56076774e-01 5.68409204e-01
-7.10507154e-01 6.46374151e-02 3.18370193e-01 1.55079529e-01
1.13390967e-01 3.32844287e-01 -2.29181778e-02 -1.56412512e-01
-5.01188397e-01 -1.82493702e-01 9.67249393e-01 1.73973054e-01
7.84551561e-01 3.12519260e-02 -2.35350147e-01 2.65837222e-01
5.42547762e-01 8.29668939e-01 7.79416263e-01 -1.08573651e+00
-4.98293340e-01 7.24745274e-01 1.37382135e-01 -9.97373164e-01
-4.80667174e-01 -2.98118860e-01 -2.46850222e-01 5.77506125e-01
2.07541153e-01 -3.66001546e-01 -3.04517262e-02 1.47859168e+00
5.53616822e-01 5.30017793e-01 2.44617358e-01 1.07343221e+00
9.58815694e-01 4.99269694e-01 3.61943275e-01 -6.32429779e-01
1.74631023e+00 -2.60748267e-01 -1.19347394e+00 8.74344036e-02
1.87156364e-01 -3.26677054e-01 5.80536664e-01 4.08865124e-01
-8.21332455e-01 -9.16212440e-01 -1.29303992e+00 -1.36517745e-03
-7.95684040e-01 -1.55960962e-01 1.87019203e-02 1.08838034e+00
-8.72194231e-01 3.41041178e-01 -4.24201250e-01 -5.11910915e-01
9.38348398e-02 5.64025640e-02 -2.70261526e-01 8.41806307e-02
-1.22136915e+00 1.43412924e+00 1.49397656e-01 5.53189933e-01
-7.68501937e-01 -7.83510864e-01 -4.21837419e-01 7.40657523e-02
4.75125853e-03 -7.93205976e-01 5.10989606e-01 -8.06693912e-01
-2.04896593e+00 9.91626024e-01 -3.54490131e-01 -5.12522042e-01
2.02144340e-01 2.19853580e-01 -7.56053925e-01 3.84493589e-01
-3.18423718e-01 3.09584081e-01 6.54785395e-01 -9.92412567e-01
1.02634415e-01 -7.56034613e-01 -6.09297097e-01 2.30298728e-01
-1.72521979e-01 -6.95433840e-02 5.56470752e-01 3.09405297e-01
-1.09315313e-01 -8.17736864e-01 1.82856768e-01 -3.15443873e-01
-1.14003502e-01 -4.86067593e-01 9.30250823e-01 -4.16797280e-01
8.09790671e-01 -2.16621256e+00 -4.02237207e-01 6.54672310e-02
2.31512964e-01 6.11765921e-01 5.53565025e-02 3.10006738e-01
-2.36895561e-01 -4.01235789e-01 9.43564549e-02 9.12351608e-02
-4.93716896e-02 -4.78382744e-02 3.12750880e-03 8.75796974e-01
-9.77035910e-02 7.47371972e-01 -6.46930456e-01 -3.03895533e-01
6.67236447e-01 5.48307955e-01 2.81596601e-01 3.34258765e-01
1.61973968e-01 7.30761409e-01 -4.13100779e-01 4.41321194e-01
7.75739610e-01 8.15435052e-01 -2.77533352e-01 -4.89775777e-01
-5.48945546e-01 -1.55918375e-01 -4.73796725e-01 1.14630342e+00
-1.60626695e-01 7.73736060e-01 1.89212784e-01 -1.15418065e+00
1.67815614e+00 4.03658718e-01 5.78685343e-01 -1.16630757e+00
5.64483345e-01 1.85111202e-02 1.08413361e-01 -1.27090549e+00
-1.42813757e-01 -5.55408478e-01 5.86518943e-01 2.38575861e-01
-3.64288837e-01 1.17931748e-02 -1.03018194e-01 -3.18387926e-01
1.11251450e+00 2.37581745e-01 1.35291070e-01 -4.95898992e-01
1.23777437e+00 -2.00567320e-01 1.97928533e-01 1.38511166e-01
-9.93733168e-01 -3.63806903e-01 4.26780969e-01 -4.48891193e-01
-4.77007985e-01 -4.24790323e-01 -4.55483824e-01 4.24429744e-01
4.01857048e-01 4.35182691e-01 -1.04596627e+00 3.39651793e-01
3.09100598e-02 1.07275379e+00 -6.24324262e-01 -7.83218682e-01
-8.87152180e-02 -4.19426084e-01 4.00425911e-01 -2.46494099e-01
5.01198530e-01 -1.38162291e+00 -1.23750579e+00 2.49280706e-01
3.14367488e-02 -1.03754818e+00 4.74368691e-01 -3.37161086e-02
-7.14916527e-01 -1.09410679e+00 -1.33628830e-01 -1.02800615e-02
-2.34305896e-02 1.84313908e-01 6.41427636e-01 -2.11313292e-02
-6.97017729e-01 5.32092929e-01 -3.48472558e-02 -1.06297612e+00
-7.03799963e-01 -3.08669329e-01 -2.11482011e-02 5.41301727e-01
1.17953598e+00 -6.99932635e-01 -9.55456913e-01 1.21872053e-01
-2.72255957e-01 -4.30407554e-01 4.14019257e-01 -2.62234449e-01
2.60118812e-01 -3.95251632e-01 9.89246786e-01 -4.60195810e-01
5.97736359e-01 -7.71158993e-01 -6.72199368e-01 -2.22986892e-01
-8.48647058e-01 -1.13890648e-01 5.79670012e-01 -6.35474771e-02
-1.19747484e+00 -2.76205063e-01 3.88479680e-02 -2.93603390e-01
-8.75473857e-01 5.41651957e-02 -2.49484077e-01 -5.50138541e-02
8.56920123e-01 3.08369994e-01 6.53617978e-01 -1.05118090e-02
-9.62795168e-02 8.64829779e-01 7.80991316e-01 1.60763636e-01
3.39389563e-01 4.90952909e-01 5.95985830e-01 -1.07320035e+00
-4.99075085e-01 -5.07256866e-01 -2.10963920e-01 -1.02338088e+00
1.08354843e+00 -7.02116072e-01 -1.77918375e+00 3.20926160e-01
-9.24788177e-01 6.58325329e-02 -6.58038855e-02 8.78068984e-01
-3.73536348e-01 4.06323254e-01 -1.94243565e-01 -1.45750582e+00
-4.58772928e-01 -7.12949812e-01 2.04741761e-01 4.77619857e-01
4.67437021e-02 -7.55961180e-01 4.05348092e-01 8.52223217e-01
6.82989597e-01 5.89682400e-01 4.77874339e-01 -2.33245924e-01
-3.25305939e-01 -6.18411660e-01 7.90263936e-02 6.29552484e-01
-9.73558053e-02 -2.35235631e-01 -1.46228480e+00 4.85097378e-01
7.05242455e-01 1.37678161e-01 5.12446046e-01 6.35310411e-01
9.84712243e-01 2.71988481e-01 -2.94415116e-01 4.39841151e-01
1.56059778e+00 5.46974719e-01 1.30277145e+00 -7.32027218e-02
2.60706574e-01 1.04789221e+00 6.71371162e-01 3.49923253e-01
2.62059331e-01 2.89386958e-01 6.87070072e-01 3.21976468e-02
-1.46864727e-01 3.31371903e-01 5.91136873e-01 9.04512927e-02
-3.10424596e-01 -1.14171609e-01 -3.53142679e-01 -1.77660864e-02
-1.28184605e+00 -1.32874787e+00 -9.10118401e-01 2.19533730e+00
5.40876031e-01 -2.22511232e-01 9.61833447e-02 2.53212512e-01
8.13867211e-01 -4.88673419e-01 -8.40861917e-01 -9.17897820e-01
-9.95015912e-03 4.23487604e-01 5.56947768e-01 4.54836845e-01
-1.30502403e-01 3.15280348e-01 6.01792049e+00 -9.23123881e-02
-1.32062924e+00 1.46323130e-01 3.51610750e-01 -1.26303881e-01
-2.76759714e-01 -2.51467694e-02 -5.89662790e-01 9.37604547e-01
1.75184488e+00 -2.27609500e-01 3.84819061e-01 5.80429554e-01
9.61028814e-01 -7.20088124e-01 -7.15842187e-01 1.09783721e+00
3.99428368e-01 -7.46255517e-01 -7.23539770e-01 2.31750891e-01
-1.65039763e-01 -1.71578407e-01 -1.35271296e-01 4.72712852e-02
-9.64532375e-01 -7.93628335e-01 1.84193224e-01 1.47423613e+00
6.78325951e-01 -6.63975775e-01 8.13455284e-01 3.02312136e-01
-6.35742068e-01 -3.37959021e-01 -3.40932786e-01 -3.09123993e-01
4.14252505e-02 7.67487228e-01 -6.43006384e-01 -4.10162210e-02
2.96805412e-01 6.28549099e-01 -3.00502747e-01 7.00038493e-01
1.19871937e-01 6.35778666e-01 -8.56127813e-02 -3.80215764e-01
-2.18006134e-01 -7.79170454e-01 7.49873996e-01 6.61571085e-01
9.27753821e-02 5.72119117e-01 -6.79197669e-01 1.64345479e+00
4.46293294e-01 -6.31106570e-02 -8.83297384e-01 1.59255907e-01
1.48738369e-01 1.64444327e+00 -4.64538962e-01 -3.89985651e-01
-1.02494210e-01 3.50787550e-01 -5.01030982e-01 3.06726217e-01
-8.45331430e-01 -5.19116521e-01 8.78465593e-01 3.81860495e-01
-4.05883104e-01 1.66747123e-01 -2.99027890e-01 -4.91992950e-01
-2.50445187e-01 1.35076061e-01 -1.08752541e-01 -1.04459631e+00
-8.07351947e-01 3.65979075e-01 -7.21964166e-02 -8.82278860e-01
2.49663427e-01 -6.24951541e-01 -6.45015001e-01 1.41901028e+00
-1.88095462e+00 -4.05235171e-01 -1.12836635e+00 7.34348416e-01
6.78179860e-02 -1.14831038e-01 7.02913165e-01 5.80056347e-02
-8.91200185e-01 9.53074917e-02 -5.61048985e-01 -6.76832139e-01
9.22902167e-01 -8.15761268e-01 -8.17193985e-01 7.23768175e-01
-1.18873489e+00 6.03140108e-02 9.38687623e-01 -4.24840569e-01
-1.52657688e+00 -6.35009468e-01 8.76556933e-01 -3.61836314e-01
3.06986928e-01 2.96237264e-02 -7.40701139e-01 6.17258325e-02
7.22381711e-01 2.33583704e-01 9.06216025e-01 -6.88374579e-01
2.52093047e-01 -7.55591214e-01 -1.56830263e+00 1.49698436e-01
4.80679840e-01 -4.08564746e-01 -6.87493980e-01 1.60548985e-01
2.19823584e-01 2.20936283e-01 -1.03644824e+00 4.76037785e-02
5.66873312e-01 -1.35131383e+00 4.96744514e-01 -3.55677485e-01
-6.22963905e-02 -4.35851842e-01 4.39677387e-01 -8.29888940e-01
-1.64021611e-01 -6.50470316e-01 1.10881478e-01 7.32214451e-01
-1.00561656e-01 -1.34533978e+00 4.18771565e-01 1.05606830e+00
-5.51200211e-01 -1.72190323e-01 -8.26361954e-01 -2.29820147e-01
-3.84978622e-01 -1.92259505e-01 3.13011765e-01 5.58437824e-01
6.75907552e-01 2.66259789e-01 4.51183990e-02 4.63427119e-02
6.37502253e-01 -4.48340148e-01 3.79764766e-01 -1.52309847e+00
2.64893740e-01 -2.66856909e-01 -8.12867463e-01 7.30841830e-02
4.33742732e-01 -6.77487791e-01 6.87237903e-02 -1.06921899e+00
-1.88190922e-01 1.28218532e-01 -5.67057312e-01 -1.09796487e-01
5.97592592e-02 3.51617783e-01 -2.30246052e-01 -2.99154967e-01
4.99117374e-02 3.33478928e-01 1.23396552e+00 4.06027406e-01
-3.91950458e-01 -1.22158900e-01 -6.78092480e-01 1.88213456e-02
8.67145658e-01 -2.93359518e-01 -4.86230850e-01 7.14086175e-01
1.52219310e-01 4.10576940e-01 6.83800936e-01 -1.56896412e+00
2.82142371e-01 -3.16808745e-02 2.37171769e-01 -5.26277661e-01
3.71930957e-01 -1.15035152e+00 4.71614629e-01 9.37680006e-01
-6.08227327e-02 -2.67058581e-01 1.02066197e-01 5.37184119e-01
-1.04296163e-01 -2.93147802e-01 1.04726195e+00 -2.30668426e-01
-4.53283191e-01 -2.23758236e-01 -9.68067288e-01 -5.46377659e-01
1.62816870e+00 -9.32353854e-01 -5.91622174e-01 -6.12906925e-02
-5.07865727e-01 -1.84631705e-01 2.45281775e-02 5.79551943e-02
3.19353670e-01 -8.85250747e-01 -4.82477605e-01 3.76915932e-01
1.61629647e-01 -8.99226010e-01 6.73916578e-01 1.50058043e+00
-4.95779246e-01 5.01450956e-01 -7.07479000e-01 -4.89287496e-01
-1.05547059e+00 8.53372872e-01 9.82354999e-01 7.73310125e-01
-3.67989361e-01 3.26088011e-01 -4.07394439e-01 5.51310956e-01
-1.62548915e-01 5.47137950e-03 -8.62226129e-01 9.76564512e-02
6.48180485e-01 1.06267500e+00 3.47944684e-02 -6.38169765e-01
-4.56040770e-01 6.22131467e-01 7.32565045e-01 1.37705356e-01
8.38093340e-01 -5.30902803e-01 -5.21844983e-01 7.02865243e-01
1.20326400e+00 -3.13028365e-01 -1.05571699e+00 4.14874852e-01
-4.69009250e-01 -1.06717974e-01 3.71759415e-01 -1.05315065e+00
-8.80203664e-01 8.24322999e-01 1.20614398e+00 5.61088920e-01
1.30129242e+00 -3.00995827e-01 7.48355746e-01 2.03035325e-01
-1.25699684e-01 -1.38416600e+00 -2.21974075e-01 -3.06563944e-01
6.28166199e-01 -1.04746377e+00 -3.23464423e-01 -2.75152624e-01
-5.52889287e-01 1.46723914e+00 4.52689558e-01 -2.44314611e-01
9.40155268e-01 2.57357806e-01 1.92704126e-01 -3.50469917e-01
-8.30228865e-01 -4.66809154e-01 1.12292275e-01 6.96010172e-01
2.85388768e-01 1.60029024e-01 -8.34862828e-01 4.58983660e-01
-5.46049662e-02 4.56992596e-01 7.02562332e-01 1.53899193e-01
-7.90861070e-01 -4.47145522e-01 -5.86045206e-01 1.93601429e-01
-6.01001568e-02 3.77616525e-01 -2.15416029e-01 2.40356907e-01
6.56570137e-01 1.47725201e+00 1.57324046e-01 -1.80591017e-01
5.46620071e-01 6.23372793e-01 3.93728316e-01 -1.98648110e-01
-3.79155278e-01 -2.15642363e-01 -2.54111886e-01 -7.72391975e-01
-8.82947147e-01 -6.70369446e-01 -1.19354355e+00 -2.21324429e-01
9.01835859e-02 1.47355244e-01 1.32569265e+00 1.10367906e+00
4.55203295e-01 4.11177754e-01 8.35672140e-01 -4.71411526e-01
-1.02628298e-01 -1.15470254e+00 -9.24981833e-01 3.33964318e-01
5.32273352e-01 -7.44935334e-01 -8.01162183e-01 -1.53236479e-01]
|
[13.594582557678223, 2.9362919330596924]
|
c24ccfbb-c823-43d0-bd17-4e0a6c92d01b
|
generalized-few-shot-semantic-segmentation
|
2010.05210
| null |
https://arxiv.org/abs/2010.05210v4
|
https://arxiv.org/pdf/2010.05210v4.pdf
|
Generalized Few-shot Semantic Segmentation
|
Training semantic segmentation models requires a large amount of finely annotated data, making it hard to quickly adapt to novel classes not satisfying this condition. Few-Shot Segmentation (FS-Seg) tackles this problem with many constraints. In this paper, we introduce a new benchmark, called Generalized Few-Shot Semantic Segmentation (GFS-Seg), to analyze the generalization ability of simultaneously segmenting the novel categories with very few examples and the base categories with sufficient examples. It is the first study showing that previous representative state-of-the-art FS-Seg methods fall short in GFS-Seg and the performance discrepancy mainly comes from the constrained setting of FS-Seg. To make GFS-Seg tractable, we set up a GFS-Seg baseline that achieves decent performance without structural change on the original model. Then, since context is essential for semantic segmentation, we propose the Context-Aware Prototype Learning (CAPL) that significantly improves performance by 1) leveraging the co-occurrence prior knowledge from support samples, and 2) dynamically enriching contextual information to the classifier, conditioned on the content of each query image. Both two contributions are experimentally shown to have substantial practical merit. Extensive experiments on Pascal-VOC and COCO manifest the effectiveness of CAPL, and CAPL generalizes well to FS-Seg by achieving competitive performance. Code is available at https://github.com/dvlab-research/GFS-Seg.
|
['Hengshuang Zhao', 'Michelle Shu', 'Shu Liu', 'Jiaya Jia', 'Li Jiang', 'Xin Lai', 'Zhuotao Tian']
|
2020-10-11
| null |
http://openaccess.thecvf.com//content/CVPR2022/html/Tian_Generalized_Few-Shot_Semantic_Segmentation_CVPR_2022_paper.html
|
http://openaccess.thecvf.com//content/CVPR2022/papers/Tian_Generalized_Few-Shot_Semantic_Segmentation_CVPR_2022_paper.pdf
|
cvpr-2022-1
|
['generalized-few-shot-semantic-segmentation']
|
['computer-vision']
|
[ 3.95427078e-01 1.44880027e-01 -2.83973724e-01 -5.34239352e-01
-1.01611805e+00 -5.73991895e-01 3.36423367e-01 3.81046012e-02
-4.36020613e-01 4.82429534e-01 -1.95677832e-01 -1.26087949e-01
2.59146422e-01 -5.65680981e-01 -9.76503134e-01 -5.65612376e-01
1.75099462e-01 4.93614107e-01 9.70412314e-01 -9.76892412e-02
1.59024149e-01 -2.37585567e-02 -1.67973375e+00 2.85406232e-01
1.21593928e+00 1.06261098e+00 5.13538241e-01 4.35603440e-01
-2.04069912e-01 1.61649913e-01 -4.49555278e-01 -1.93960935e-01
3.49279493e-01 -5.61784208e-01 -9.11896706e-01 1.98273569e-01
4.94480580e-01 -1.40965477e-01 1.61116049e-02 1.11567366e+00
2.04128131e-01 5.05996644e-01 4.62708145e-01 -1.16528285e+00
-7.62351036e-01 5.19602001e-01 -4.54060197e-01 1.85157865e-01
1.54019669e-01 2.66216934e-01 1.13095164e+00 -1.14382148e+00
7.02819705e-01 1.15689075e+00 6.01867855e-01 8.33541632e-01
-1.04770195e+00 -4.65459347e-01 6.60238564e-01 4.06622320e-01
-1.33606040e+00 -1.71932980e-01 6.28333271e-01 -3.00555646e-01
8.00272763e-01 1.04390100e-01 6.25078022e-01 1.16790473e+00
-4.44302291e-01 1.30922139e+00 9.87075984e-01 -4.26280409e-01
6.61412477e-01 4.75652739e-02 6.70577049e-01 4.94799584e-01
1.45228148e-01 -1.05984174e-01 -3.13511759e-01 9.68581513e-02
3.76435846e-01 -1.31257311e-01 -4.16122049e-01 -4.16499138e-01
-8.82037044e-01 8.38901818e-01 5.37581265e-01 3.49330753e-01
1.70622263e-02 7.28600367e-04 3.29350382e-01 -4.13721465e-02
4.86352175e-01 3.88581902e-01 -5.53914249e-01 -1.34641349e-01
-1.13386989e+00 1.05020724e-01 6.60602510e-01 1.40592015e+00
7.87349284e-01 -2.05185071e-01 -4.06714439e-01 9.46499109e-01
-9.83870029e-02 2.99745411e-01 6.41063690e-01 -8.53935480e-01
2.14472756e-01 4.72623944e-01 -7.47903213e-02 -3.73042107e-01
-1.62745789e-01 -4.81983930e-01 -2.73270339e-01 -3.22215796e-01
2.84461886e-01 -1.08166982e-03 -1.48365414e+00 1.83441782e+00
4.63026047e-01 5.99757671e-01 -7.47918189e-02 9.01993573e-01
7.87711501e-01 7.06571102e-01 2.51355529e-01 -1.03101470e-01
1.14557302e+00 -1.46058130e+00 -3.86441141e-01 -5.34103453e-01
7.02489674e-01 -3.39283019e-01 1.60119140e+00 8.86351168e-02
-6.22614980e-01 -6.12374246e-01 -1.08250129e+00 5.71415201e-02
-5.69867194e-01 -2.27693677e-01 6.26259863e-01 5.04731357e-01
-9.11451936e-01 5.69843709e-01 -9.52021420e-01 -5.44447780e-01
7.54906416e-01 1.16246693e-01 9.51562971e-02 -4.37453210e-01
-1.15237665e+00 3.85882139e-01 7.48223186e-01 -1.78381562e-01
-8.86091769e-01 -7.73587227e-01 -9.66949582e-01 1.26909971e-01
8.83449554e-01 -5.55285037e-01 1.43990660e+00 -1.06014419e+00
-1.24117756e+00 7.00871944e-01 -2.19477996e-01 -3.87995988e-01
6.03033304e-01 -1.86802879e-01 -1.90090507e-01 3.48423064e-01
3.96985292e-01 1.06173694e+00 6.28150165e-01 -1.30274355e+00
-7.30621219e-01 -2.62524694e-01 1.31865060e-02 1.05194040e-01
-2.71533966e-01 -2.40889519e-01 -1.00650120e+00 -6.53353512e-01
1.23068787e-01 -9.39579904e-01 -3.32853019e-01 -2.13624984e-01
-4.79632020e-01 -4.39023912e-01 7.66119361e-01 -3.60361576e-01
1.19377470e+00 -2.25035596e+00 -8.23721141e-02 -1.20615795e-01
-2.22957060e-01 6.33046806e-01 -3.26189578e-01 1.74326241e-01
3.34162980e-01 2.04627559e-01 -7.58697093e-01 -4.84002113e-01
2.33174395e-02 3.95325899e-01 -2.22668171e-01 2.82526761e-02
2.88523734e-01 1.20564413e+00 -1.12354398e+00 -6.06857121e-01
5.97614720e-02 9.01743025e-02 -5.93027771e-01 8.91762972e-02
-5.88815272e-01 4.37446862e-01 -5.08575618e-01 9.67917681e-01
5.60174167e-01 -4.31456655e-01 -8.60950425e-02 9.81352702e-02
1.25403330e-01 -8.09136033e-02 -9.35985625e-01 1.95084798e+00
-1.07557930e-01 1.95108876e-01 -3.07348877e-01 -1.12512946e+00
6.73129916e-01 -6.65243119e-02 1.53603941e-01 -7.64408231e-01
2.47326314e-01 3.60397279e-01 -2.72934854e-01 -4.46366429e-01
4.19718206e-01 -8.66190046e-02 -2.75763363e-01 7.35844001e-02
4.74600941e-01 -7.24685639e-02 5.22146821e-01 2.95470357e-01
8.34201694e-01 2.66055226e-01 2.33019903e-01 -2.30029836e-01
2.33852759e-01 8.69279057e-02 9.74229813e-01 9.73953426e-01
-7.08205879e-01 8.45961869e-01 3.22434723e-01 -3.07423957e-02
-7.74649799e-01 -1.13517189e+00 -1.65119544e-01 1.18160462e+00
6.60768628e-01 -3.12116325e-01 -1.16566992e+00 -9.94471610e-01
-1.14724718e-01 1.05564511e+00 -6.75230503e-01 -1.87718734e-01
-3.70095551e-01 -6.40257120e-01 2.58784771e-01 7.82341063e-01
4.58681345e-01 -1.03238308e+00 -5.65057695e-01 2.15609923e-01
-1.11847617e-01 -1.35127866e+00 -6.33344352e-01 7.12647513e-02
-8.19117904e-01 -1.10903478e+00 -9.19134080e-01 -9.89697576e-01
6.23219073e-01 4.51813757e-01 9.01104748e-01 5.81849031e-02
-3.02593708e-01 3.20859343e-01 -8.25232267e-01 -3.88698220e-01
-8.29525888e-02 1.26799077e-01 -3.58924955e-01 4.78887223e-02
4.54302132e-01 -4.37619716e-01 -5.45740247e-01 3.80387425e-01
-8.65228951e-01 2.55846828e-01 4.35336292e-01 8.40096295e-01
1.06906688e+00 -3.58101785e-01 7.47223079e-01 -1.27692652e+00
1.32151708e-01 -5.71298540e-01 -3.92929614e-01 4.39645827e-01
-6.02437675e-01 -3.02418679e-01 6.06393218e-01 -3.68739426e-01
-9.06226039e-01 1.22947477e-01 -1.38208926e-01 -6.13401055e-01
-3.62081051e-01 4.17786658e-01 -2.97935367e-01 1.20994233e-01
5.29954016e-01 1.89363986e-01 -3.71678770e-01 -6.37798607e-01
5.67865014e-01 7.23425686e-01 7.74148762e-01 -6.70014620e-01
4.45928305e-01 4.54979420e-01 -5.07826447e-01 -8.83209646e-01
-1.34989238e+00 -8.98122072e-01 -6.73717201e-01 -2.78857350e-02
8.02040279e-01 -8.46118629e-01 1.85460329e-01 5.77515662e-01
-8.25899661e-01 -6.92506731e-01 -4.83717680e-01 1.65528446e-01
-6.45433962e-01 5.12356520e-01 -5.02138615e-01 -6.18665159e-01
-2.65777975e-01 -1.08101785e+00 1.11071157e+00 5.38557827e-01
5.18035553e-02 -8.32810581e-01 -1.43289655e-01 4.28510219e-01
9.44114104e-02 3.63948882e-01 6.41188383e-01 -9.66037512e-01
-6.33174777e-01 -1.33646697e-01 -2.78447866e-01 4.85441267e-01
-6.39042035e-02 -2.20123857e-01 -1.07079506e+00 -3.38073820e-01
-1.00296512e-01 -4.09647405e-01 1.26879942e+00 3.86785924e-01
1.33710921e+00 -5.81065677e-02 -3.89185429e-01 7.87446737e-01
1.49383819e+00 2.88102835e-01 4.64443475e-01 2.40424573e-01
8.35198283e-01 4.04630184e-01 1.12700331e+00 1.58634275e-01
4.97563273e-01 6.22382939e-01 2.13487580e-01 5.07595576e-02
-4.11375254e-01 -3.34375620e-01 1.48970529e-01 8.69330525e-01
2.53810793e-01 -3.16367328e-01 -9.43977356e-01 9.23078418e-01
-2.08316517e+00 -4.93600965e-01 1.02634318e-01 2.02164149e+00
8.12270701e-01 3.25418591e-01 8.12547877e-02 -9.40501094e-02
9.73037362e-01 4.48101051e-02 -8.19628000e-01 -1.32373691e-01
-1.80544369e-02 3.34801614e-01 3.11624199e-01 4.34754252e-01
-1.28775287e+00 1.51186776e+00 5.09788990e+00 1.17610633e+00
-8.68953288e-01 4.59588408e-01 5.93866944e-01 -1.30008325e-01
-2.14906707e-01 2.16300353e-01 -9.66001391e-01 6.20721579e-01
7.32311904e-01 -2.38911718e-01 1.29688770e-01 1.06253314e+00
-1.27128378e-01 -1.99239805e-01 -1.02671206e+00 7.70517230e-01
2.76395947e-01 -1.00982773e+00 1.85929045e-01 -3.70521426e-01
1.00081599e+00 2.27137193e-01 -1.00306451e-01 6.00407541e-01
2.80598607e-02 -5.93854010e-01 9.28218126e-01 1.58139810e-01
9.01793122e-01 -4.72370654e-01 6.54694438e-01 3.34245712e-01
-1.17168653e+00 -1.43334225e-01 -4.88756269e-01 2.64063597e-01
2.98334002e-01 4.40578520e-01 -7.19769895e-01 6.49799049e-01
6.89599156e-01 8.61716747e-01 -7.70844400e-01 1.32413411e+00
-5.13785720e-01 8.94074917e-01 -2.79607713e-01 -6.06741849e-03
6.56838715e-01 8.84256735e-02 6.23301208e-01 1.34154046e+00
2.51757413e-01 3.06915283e-01 4.88568336e-01 9.64772880e-01
-1.15614971e-02 1.66499868e-01 -1.16750032e-01 -5.81413731e-02
5.75819135e-01 9.69922304e-01 -1.22747838e+00 -6.02818310e-01
-4.28931236e-01 1.07880354e+00 3.40285361e-01 4.43855047e-01
-8.39549720e-01 -4.37646210e-01 4.13725227e-01 -4.52098958e-02
6.95993125e-01 7.63095841e-02 -2.83208311e-01 -1.20233786e+00
6.74781427e-02 -5.12274146e-01 6.55458510e-01 -5.53325117e-01
-1.21503329e+00 6.09139621e-01 9.65303853e-02 -1.07948554e+00
5.57820350e-02 -4.22215074e-01 -5.86238563e-01 3.57753724e-01
-1.74813890e+00 -1.33907485e+00 -4.34675276e-01 4.47583586e-01
1.08835661e+00 2.07356974e-01 5.96322060e-01 2.05420747e-01
-7.89800048e-01 8.32102358e-01 -1.14679663e-03 -8.99782255e-02
5.98160565e-01 -1.31265450e+00 6.86325252e-01 1.04692078e+00
2.15140820e-01 2.47035325e-01 5.68967938e-01 -6.59729064e-01
-7.97895849e-01 -1.39190733e+00 6.14833713e-01 -3.68616194e-01
4.73407835e-01 -5.14532506e-01 -1.26675141e+00 6.82458460e-01
-2.98855245e-01 2.50145972e-01 5.77453077e-01 9.54051595e-03
-5.14697373e-01 2.16683149e-01 -9.86138046e-01 5.21342576e-01
1.33760595e+00 -3.46553653e-01 -7.76370347e-01 3.63518715e-01
1.25195777e+00 -3.82682711e-01 -4.02479082e-01 4.92918938e-01
1.32193521e-01 -8.28139842e-01 7.13544369e-01 -5.35916090e-01
2.68975794e-01 -3.45009476e-01 -2.85881490e-01 -1.18333662e+00
-1.69370279e-01 -4.21796560e-01 -1.44634292e-01 1.31820357e+00
3.93973410e-01 -5.82626700e-01 7.20733523e-01 5.38035810e-01
-6.95260823e-01 -1.05186427e+00 -9.14259076e-01 -1.35783756e+00
1.70911461e-01 -5.30587137e-01 5.25248230e-01 1.05498314e+00
-1.00465015e-01 1.45049721e-01 3.08995415e-02 8.27298686e-02
6.00735128e-01 3.35603505e-01 5.61341584e-01 -1.02154350e+00
-5.01238763e-01 -4.03007597e-01 -4.18483675e-01 -1.11888945e+00
9.06425044e-02 -1.00358284e+00 4.48324531e-01 -1.66436362e+00
3.12051177e-01 -4.99480277e-01 -6.01965010e-01 6.85493469e-01
-7.36258030e-01 2.75586456e-01 5.17189980e-01 2.10806698e-01
-1.15061319e+00 6.75947785e-01 1.34918475e+00 -8.25719237e-02
-4.14622545e-01 -6.08647875e-02 -7.74613917e-01 7.91356564e-01
7.09675312e-01 -4.21410322e-01 -5.97445905e-01 -3.41142178e-01
-3.04964602e-01 -4.23862547e-01 1.53513983e-01 -1.18419123e+00
1.51211157e-01 -1.25875115e-01 -1.22680277e-01 -4.82654721e-01
1.95622578e-01 -3.75465304e-01 -2.87619889e-01 2.25255042e-01
-2.23573416e-01 -6.86625361e-01 2.21993446e-01 8.65367532e-01
-2.67183632e-01 -4.54305977e-01 9.96281505e-01 -1.45038337e-01
-1.49876547e+00 5.58872104e-01 1.12088434e-01 6.19913280e-01
1.31564200e+00 -4.14169848e-01 -2.97670305e-01 1.00100428e-01
-7.86016881e-01 5.47629178e-01 6.85997427e-01 5.51616430e-01
4.22703147e-01 -1.10260952e+00 -3.80712897e-01 1.42262774e-02
5.39691865e-01 3.46275657e-01 5.41008234e-01 7.61156619e-01
-3.34845930e-01 2.80907482e-01 1.02803692e-01 -7.06240773e-01
-1.02481019e+00 7.78146386e-01 5.49472719e-02 3.72209139e-02
-7.59375811e-01 1.35474360e+00 5.53556323e-01 -2.43876338e-01
3.10606807e-01 -3.66509706e-01 1.58827996e-03 -9.21851583e-03
5.18251956e-01 1.48856610e-01 -4.41297740e-02 -5.41565895e-01
-3.01469147e-01 5.91263175e-01 -2.50351161e-01 5.40128686e-02
1.16271412e+00 -1.99797228e-01 3.44059914e-01 5.42416871e-01
1.13395035e+00 -4.94062006e-01 -1.71026635e+00 -4.07938987e-01
2.03095481e-01 -4.34783727e-01 -1.42473891e-01 -8.84045124e-01
-9.48164523e-01 9.89662051e-01 4.47850585e-01 -5.80896512e-02
1.16313088e+00 3.13751936e-01 1.23522508e+00 2.28634924e-01
5.76903403e-01 -1.36285651e+00 1.93181753e-01 5.87235391e-01
5.72165906e-01 -1.26692748e+00 -3.47448796e-01 -8.30410659e-01
-7.92204797e-01 6.95878565e-01 7.35711396e-01 -9.24581736e-02
5.46633124e-01 -1.93774939e-01 -6.96718991e-02 -1.15661122e-01
-4.17976618e-01 -5.62563777e-01 2.49521971e-01 6.22602284e-01
-1.53133914e-01 1.99287921e-01 -3.89147341e-01 1.15442348e+00
-1.51843121e-02 3.40901837e-02 4.43016499e-01 1.13696599e+00
-7.71735907e-01 -9.79245365e-01 5.62604144e-02 3.87557834e-01
-1.90644935e-01 -1.04688264e-01 -4.32455212e-01 7.01759517e-01
3.02559793e-01 8.80021751e-01 7.09626228e-02 -1.54911205e-01
3.92675459e-01 2.78471291e-01 3.16805243e-01 -1.05519760e+00
-1.94759667e-01 1.45810038e-01 -9.77132693e-02 -7.96664894e-01
-2.38017812e-01 -7.34362543e-01 -1.49634409e+00 2.76153505e-01
-4.53396022e-01 1.15360238e-01 4.48638260e-01 1.02407742e+00
4.59443361e-01 4.96195465e-01 3.91143829e-01 -7.83123672e-01
-4.92942214e-01 -8.21615815e-01 -6.13690078e-01 5.53655207e-01
9.05064791e-02 -7.92998552e-01 -4.14478302e-01 8.63080025e-02]
|
[9.56368350982666, 1.231203317642212]
|
0fec5321-cfe2-4023-8714-be8bb191a485
|
robust-full-fov-depth-estimation-in-tele-wide
|
1909.03375
| null |
https://arxiv.org/abs/1909.03375v2
|
https://arxiv.org/pdf/1909.03375v2.pdf
|
Robust Full-FoV Depth Estimation in Tele-wide Camera System
|
Tele-wide camera system with different Field of View (FoV) lenses becomes very popular in recent mobile devices. Usually it is difficult to obtain full-FoV depth based on traditional stereo-matching methods. Pure Deep Neural Network (DNN) based depth estimation methods can obtain full-FoV depth, but have low robustness for scenarios which are not covered by training dataset. In this paper, to address the above problems we propose a hierarchical hourglass network for robust full-FoV depth estimation in tele-wide camera system, which combines the robustness of traditional stereo-matching methods with the accuracy of DNN. More specifically, the proposed network comprises three major modules: single image depth prediction module infers initial depth from input color image, depth propagation module propagates traditional stereo-matching tele-FoV depth to surrounding regions, and depth combination module fuses the initial depth with the propagated depth to generate final output. Each of these modules employs an hourglass model, which is a kind of encoder-decoder structure with skip connections. Experimental results compared with state-of-the-art depth estimation methods demonstrate that our method not only produces robust and better subjective depth quality on wild test images, but also obtains better quantitative results on standard datasets.
|
['Tae-ui Kim', 'Seungmin Han', 'Irina Kim', 'Soonkeun Chang', 'Seongwook Song', 'Kai Guo']
|
2019-09-08
| null | null | null | null |
['stereo-matching']
|
['computer-vision']
|
[ 1.14621907e-01 -2.34776095e-01 3.30659486e-02 -4.04371619e-01
-4.93533820e-01 -1.37054265e-01 2.19595060e-01 -6.19171560e-01
-4.33959454e-01 7.28772223e-01 1.84616849e-01 -7.55551755e-02
2.12766320e-01 -1.11078918e+00 -5.97896516e-01 -7.64500201e-01
5.72442293e-01 -1.32041156e-01 5.77802181e-01 6.16918551e-03
3.11148614e-01 2.67782003e-01 -1.43175900e+00 2.85551786e-01
7.25306153e-01 1.33882368e+00 4.83713180e-01 7.57674456e-01
1.14365868e-01 1.03959370e+00 -3.08547199e-01 -1.07827067e-01
3.64065915e-01 -1.77737296e-01 -4.09166396e-01 1.22785635e-01
7.88201869e-01 -1.42348301e+00 -1.01577628e+00 1.13869429e+00
6.35440111e-01 -1.22627005e-01 2.61439055e-01 -9.44421828e-01
-3.50148350e-01 1.13236651e-01 -8.50366294e-01 2.61824012e-01
4.36821967e-01 2.93682188e-01 2.65359938e-01 -6.03049278e-01
5.02679527e-01 1.25078976e+00 6.18665457e-01 5.91623485e-01
-4.62690413e-01 -7.98748076e-01 1.22601561e-01 6.66683465e-02
-1.22610307e+00 -2.74344355e-01 6.51256263e-01 -2.63363302e-01
8.29899788e-01 -3.29605848e-01 7.92367339e-01 8.08639824e-01
6.25629187e-01 8.12546909e-01 9.36637938e-01 -1.42804071e-01
-1.75068825e-02 -1.18636876e-01 -1.59706473e-01 6.94099665e-01
1.46393985e-01 4.80255365e-01 -4.53207493e-01 5.44035673e-01
1.43124437e+00 2.55988508e-01 -8.74807358e-01 -1.84539452e-01
-1.23411286e+00 4.35537517e-01 5.93674004e-01 1.16113089e-01
3.36409695e-02 2.43375361e-01 1.21491998e-01 1.66039482e-01
4.54990596e-01 -2.76029259e-01 -3.08095843e-01 -4.62935008e-02
-9.06057060e-01 8.67552683e-02 1.58293873e-01 1.09952223e+00
9.37701583e-01 4.22398560e-02 1.09336294e-01 6.67690635e-01
4.63779867e-01 5.84004045e-01 5.96573591e-01 -1.16894925e+00
8.67919683e-01 4.74003106e-01 2.42124824e-03 -6.43779933e-01
-3.47727090e-01 -2.51191944e-01 -1.04870880e+00 4.79423285e-01
5.03230751e-01 -3.54391396e-01 -7.88248777e-01 1.33685207e+00
2.08756089e-01 6.40721247e-02 1.94740921e-01 1.10475469e+00
1.16734457e+00 8.24887276e-01 -6.27564371e-01 -9.06649977e-02
8.95982444e-01 -1.01773310e+00 -6.02113068e-01 -5.84689736e-01
5.05325198e-01 -7.00923562e-01 7.89761662e-01 7.08017349e-01
-1.36732996e+00 -7.26785600e-01 -1.34268010e+00 -5.66874564e-01
3.65061965e-03 1.81577608e-01 4.70873952e-01 5.92902601e-01
-1.27873170e+00 3.40830594e-01 -4.56762701e-01 -1.66048065e-01
2.68238485e-01 4.21115935e-01 -4.77827102e-01 -4.75927770e-01
-1.16796231e+00 4.36067075e-01 4.46664214e-01 3.39074731e-01
-1.02715719e+00 -6.21833980e-01 -1.00452483e+00 -2.27702200e-01
1.31363794e-01 -1.00861144e+00 1.41314566e+00 -8.58897746e-01
-1.68022418e+00 6.22582912e-01 -2.61975855e-01 -7.15938881e-02
5.84460258e-01 -1.60138756e-01 -1.06995113e-01 2.76958138e-01
2.71408893e-02 1.06727195e+00 6.18928671e-01 -1.19003856e+00
-1.18687022e+00 -6.77290201e-01 3.94350082e-01 4.61976141e-01
-2.00325564e-01 -4.39836621e-01 -8.45239103e-01 -1.50092110e-01
4.78639007e-01 -5.35333455e-01 4.31293584e-02 4.14836466e-01
-3.35031122e-01 2.06439465e-01 9.55037296e-01 -4.70124960e-01
1.15100622e+00 -1.94692683e+00 -7.38366321e-02 -1.51600897e-01
4.27072287e-01 -3.30729634e-02 2.11833566e-01 3.63320932e-02
6.69451654e-02 -4.64798033e-01 1.23900436e-02 -3.34783167e-01
-6.49904907e-01 -1.08272493e-01 -4.63059321e-02 5.28552890e-01
-4.61849749e-01 5.57153225e-01 -9.48671520e-01 -6.64448380e-01
8.02220643e-01 5.77302337e-01 -4.95319307e-01 3.52928191e-01
1.39721315e-02 4.84439522e-01 -1.90743804e-01 7.87724853e-01
1.10748315e+00 -1.50125906e-01 -1.47497132e-01 -3.82633328e-01
-2.61845350e-01 -2.04224750e-01 -1.10921097e+00 1.95641148e+00
-7.06979394e-01 1.03733063e+00 2.27642246e-02 -2.29733527e-01
7.53647745e-01 1.29081565e-03 1.86944872e-01 -9.70863521e-01
4.21901375e-01 3.59664619e-01 -2.11996168e-01 -6.18928611e-01
6.14949405e-01 -1.42867357e-01 2.10475504e-01 4.76097874e-02
-6.17520399e-02 -4.05013144e-01 -5.60520515e-02 4.63126488e-02
8.54124069e-01 2.43634477e-01 2.18445882e-02 2.34327748e-01
7.93289125e-01 -4.10402089e-01 6.99121892e-01 2.16243982e-01
-3.10574025e-01 1.17165816e+00 3.23002607e-01 -3.24825555e-01
-1.17162585e+00 -1.06239712e+00 -2.06015751e-01 2.57740349e-01
8.72332215e-01 4.41253465e-03 -7.82783568e-01 -4.29216206e-01
-3.59148383e-01 8.61664936e-02 -5.37788391e-01 9.34748575e-02
-4.62749004e-01 -2.97960043e-01 3.34715694e-01 7.66167581e-01
1.52975643e+00 -7.17776537e-01 -5.73334992e-01 -4.66127470e-02
-5.23684025e-01 -1.30687523e+00 -5.92930317e-01 -6.86122775e-02
-1.18988144e+00 -1.14908612e+00 -1.14630222e+00 -9.08340335e-01
4.65742081e-01 8.80335271e-01 9.04482603e-01 -7.68900588e-02
-1.15402648e-02 -1.10689942e-02 6.04105974e-03 4.82188985e-02
7.65934885e-02 -1.94237649e-01 -3.32866937e-01 -6.03535362e-02
3.23268622e-01 -5.94570577e-01 -1.29607487e+00 5.25002897e-01
-1.05284595e+00 3.33395422e-01 5.39193034e-01 6.74020052e-01
4.79010314e-01 2.82147735e-01 1.26992604e-02 -5.80591440e-01
1.58270150e-01 -3.35018307e-01 -9.36193883e-01 -1.32219613e-01
-4.66804832e-01 -2.07742736e-01 5.74420750e-01 1.80886628e-03
-1.40363145e+00 3.70305235e-04 -2.42267206e-01 -5.93884408e-01
3.56614962e-02 -8.86935741e-02 -5.46907544e-01 -3.13949704e-01
4.51861501e-01 3.96174192e-01 -6.64035231e-02 -1.71330392e-01
-9.50965360e-02 1.05872464e+00 7.32643425e-01 -3.34027670e-02
5.53219378e-01 9.51681435e-01 -4.38301601e-02 -6.61153018e-01
-7.35531271e-01 -5.33371925e-01 -5.44874489e-01 -5.24007797e-01
1.11155915e+00 -1.58437467e+00 -7.63164759e-01 1.04908669e+00
-1.18292320e+00 -5.06660104e-01 3.00412059e-01 7.56295264e-01
-4.45261300e-01 6.86783075e-01 -8.34626317e-01 -4.17214483e-01
-2.39406034e-01 -1.36474216e+00 1.39005065e+00 7.10467696e-01
4.14555222e-01 -1.08803773e+00 -2.19019294e-01 6.35441601e-01
1.40254229e-01 -3.27116251e-02 5.57444632e-01 7.73359239e-01
-1.16069639e+00 1.45767862e-02 -6.74392223e-01 5.12654841e-01
1.79402437e-02 -9.92142856e-02 -1.44028032e+00 -2.33058244e-01
1.00831583e-01 -2.16734946e-01 9.20572817e-01 8.51919651e-01
1.40036738e+00 2.91804940e-01 -2.45884240e-01 1.25292003e+00
1.88028455e+00 4.24852580e-01 1.26034749e+00 7.38124371e-01
1.02286029e+00 5.23539484e-01 8.38998079e-01 4.50370967e-01
5.97145796e-01 5.74370682e-01 8.05390120e-01 -1.86687469e-01
-2.97225058e-01 -2.28241295e-01 4.20779616e-01 5.13815701e-01
6.66568130e-02 -5.30291736e-01 -5.08194208e-01 3.81413579e-01
-1.62576866e+00 -9.18769836e-01 -3.77776295e-01 2.28814793e+00
5.17747343e-01 3.33317459e-01 -2.41027176e-01 2.48509109e-01
5.72930574e-01 2.74414361e-01 -7.69010365e-01 -1.03567839e-01
-3.41730624e-01 -2.90481329e-01 8.59329581e-01 6.19042099e-01
-9.75812912e-01 6.95877552e-01 5.47902250e+00 9.06785965e-01
-1.14932072e+00 4.03243266e-02 8.15379560e-01 -4.07656312e-01
-3.64280432e-01 -1.94992751e-01 -9.17752981e-01 5.90032637e-01
3.53713810e-01 3.46983194e-01 1.48426428e-01 8.18674207e-01
4.52527821e-01 -6.84784949e-01 -1.16470337e+00 1.69461322e+00
1.71663523e-01 -1.30790675e+00 -8.88325274e-02 2.88214654e-01
1.15765560e+00 8.88859108e-02 -3.76029522e-03 -2.94858962e-02
-1.20223753e-01 -8.53884280e-01 5.63978255e-01 4.56343561e-01
1.18987167e+00 -8.00169289e-01 1.03454781e+00 3.40866208e-01
-1.22195685e+00 -3.02566290e-01 -6.24551415e-01 -9.03014392e-02
2.45752379e-01 8.33624721e-01 -1.33943081e-01 4.74924833e-01
9.20412958e-01 1.15867412e+00 -3.68224591e-01 1.22373009e+00
-2.60693252e-01 -1.89918533e-01 -1.66527368e-02 3.41880769e-01
3.79061401e-01 -1.62077114e-01 1.27278626e-01 7.31669843e-01
4.97375369e-01 3.20529416e-02 -2.87642002e-01 5.98755479e-01
-1.29856274e-01 -2.81893730e-01 -8.81376922e-01 7.44323730e-01
3.76020014e-01 1.00486720e+00 -3.89224857e-01 -3.68326426e-01
-7.87111163e-01 1.16220105e+00 3.37517038e-02 3.18699032e-01
-7.34257281e-01 -6.64856970e-01 5.69391489e-01 3.80516917e-01
8.11274275e-02 -1.52892293e-02 -3.28847140e-01 -1.34724510e+00
1.46270320e-01 -5.81806421e-01 9.60426182e-02 -1.58985639e+00
-7.67804861e-01 5.62183440e-01 -1.18247852e-01 -1.79474294e+00
-1.71849936e-01 -7.57630765e-01 -6.56614363e-01 9.03883696e-01
-1.91847837e+00 -8.96772325e-01 -1.23638868e+00 8.85783732e-01
8.76450896e-01 -2.41036229e-02 1.00160450e-01 5.69735527e-01
-6.78643227e-01 5.13599932e-01 3.44161898e-01 4.63859178e-02
8.10640395e-01 -9.96633053e-01 8.96899179e-02 9.79694426e-01
-6.69377506e-01 3.34354937e-01 9.65964645e-02 -5.48155248e-01
-1.16273510e+00 -1.14107788e+00 5.77540636e-01 -1.14114597e-01
7.58597553e-02 -9.39216763e-02 -4.57655132e-01 5.91382504e-01
2.51770169e-01 7.24399015e-02 2.77468539e-03 -5.02760351e-01
-5.26016243e-02 -6.05419457e-01 -1.12580633e+00 5.22005260e-01
1.10371017e+00 -5.52706480e-01 -9.27281082e-02 1.04718588e-01
7.54622161e-01 -7.63920128e-01 -6.55984163e-01 3.82495850e-01
7.08502054e-01 -1.79981279e+00 9.54946935e-01 6.38118446e-01
9.85500455e-01 -5.16269982e-01 -2.60051131e-01 -9.50301170e-01
2.23925952e-02 -2.96793610e-01 1.11506075e-01 9.46925223e-01
-5.35892732e-02 -5.82362115e-01 1.24171019e+00 2.83833802e-01
-4.29979503e-01 -8.29196870e-01 -7.91328132e-01 -5.00118196e-01
-1.68202773e-01 -3.78011942e-01 4.48507696e-01 4.39023763e-01
-1.71293154e-01 2.99104899e-01 -4.70860004e-01 2.08333194e-01
9.18957710e-01 -1.46269605e-01 9.34500277e-01 -9.52794135e-01
-2.65596330e-01 -1.80839166e-01 -6.23468041e-01 -1.81309295e+00
-1.56840354e-01 -2.48492673e-01 3.98129635e-02 -1.87177038e+00
3.37166399e-01 -2.23677590e-01 2.44025096e-01 -1.41970038e-01
1.15689531e-01 5.05095541e-01 -2.08223030e-01 1.44138679e-01
-4.24298376e-01 4.96118575e-01 1.72833574e+00 -1.52347341e-01
-1.55434430e-01 -9.62878466e-02 -4.00306821e-01 9.70690608e-01
5.25630116e-01 -3.21942903e-02 -9.93072391e-01 -9.34142828e-01
3.13831329e-01 6.07267559e-01 3.70603472e-01 -1.55550551e+00
4.05171394e-01 6.81138709e-02 8.23079824e-01 -1.00078678e+00
6.04458690e-01 -7.73970842e-01 -1.28157884e-01 4.58855510e-01
1.75566480e-01 -1.17939383e-01 9.12485048e-02 6.28867924e-01
-5.55629909e-01 -1.97120443e-01 9.75637317e-01 -2.88009018e-01
-1.08912277e+00 6.02786541e-01 -2.24134609e-01 9.99708287e-03
9.80266273e-01 -9.83359933e-01 -5.54368556e-01 -7.10051417e-01
-1.24306165e-01 2.13745669e-01 6.61288798e-01 3.06735247e-01
1.17767036e+00 -1.23591900e+00 -4.03048784e-01 4.00086880e-01
2.36658975e-01 5.75625360e-01 7.40313053e-01 6.95717812e-01
-1.13324296e+00 6.82493508e-01 -2.62760043e-01 -8.73248160e-01
-1.18758380e+00 2.86698192e-01 8.66319239e-01 -3.30551667e-03
-7.10492373e-01 1.04789102e+00 9.80984330e-01 -1.18104503e-01
4.27291781e-01 -6.36463225e-01 -1.07744634e-01 -4.09920722e-01
6.35381043e-01 3.32606345e-01 -5.95770665e-02 -4.28936839e-01
-9.82055515e-02 1.18823659e+00 -8.60622972e-02 -2.28412613e-01
9.17540669e-01 -6.59534574e-01 1.58648163e-01 3.39393675e-01
1.45557678e+00 -2.30124012e-01 -1.79656434e+00 -1.72525421e-01
-1.08460426e+00 -1.07573140e+00 5.21765590e-01 -4.17315543e-01
-1.46903002e+00 1.23003542e+00 9.27483559e-01 -5.07397592e-01
1.48575771e+00 -4.86198455e-01 1.29568517e+00 2.40198806e-01
5.80297828e-01 -1.11526489e+00 2.00699031e-01 4.63232756e-01
2.94211149e-01 -1.63385141e+00 -9.85023230e-02 -3.69241476e-01
-2.56321400e-01 1.45705605e+00 1.26887107e+00 1.74366049e-02
5.65781653e-01 2.46919796e-01 1.03802070e-01 -2.86135599e-02
-6.13463461e-01 -4.60896417e-02 -1.86726674e-01 6.81636572e-01
2.85583407e-01 -5.38024187e-01 9.88528579e-02 7.23410696e-02
-1.72645915e-02 2.65294641e-01 9.04549181e-01 5.85363984e-01
-5.78296721e-01 -8.01828563e-01 -4.97120351e-01 2.34727859e-01
-2.68474579e-01 -2.36902446e-01 8.00197348e-02 9.08412576e-01
5.11628628e-01 1.02805841e+00 2.96544880e-01 -6.06606007e-01
1.73961952e-01 -7.92195261e-01 6.72449946e-01 -3.09245735e-01
-2.93994993e-01 7.70856366e-02 -1.37523441e-02 -7.72336245e-01
-3.28068256e-01 -3.23102921e-01 -1.18292093e+00 -6.53657854e-01
-2.45287567e-01 -3.92192364e-01 6.21641040e-01 8.09779346e-01
-9.42572802e-02 5.47543406e-01 7.47846663e-01 -9.87799168e-01
1.11497059e-01 -9.16217148e-01 -8.15168381e-01 2.80951299e-02
5.84651709e-01 -4.27797973e-01 -5.43632984e-01 -1.37364030e-01]
|
[8.93398666381836, -2.4525365829467773]
|
f5d67f8c-f133-47a7-91b9-552cfb92b315
|
ssegcn-syntactic-and-semantic-enhanced-graph
| null | null |
https://aclanthology.org/2022.naacl-main.362
|
https://aclanthology.org/2022.naacl-main.362.pdf
|
SSEGCN: Syntactic and Semantic Enhanced Graph Convolutional Network for Aspect-based Sentiment Analysis
|
Aspect-based Sentiment Analysis (ABSA) aims to predict the sentiment polarity towards a particular aspect in a sentence. Recently, graph neural networks based on dependency tree convey rich structural information which is proven to be utility for ABSA. However, how to effectively harness the semantic and syntactic structure information from the dependency tree remains a challenging research question. In this paper, we propose a novel Syntactic and Semantic Enhanced Graph Convolutional Network (SSEGCN) model for ABSA task. Specifically, we propose an aspect-aware attention mechanism combined with self-attention to obtain attention score matrices of a sentence, which can not only learn the aspect-related semantic correlations, but also learn the global semantics of the sentence. In order to obtain comprehensive syntactic structure information, we construct syntactic mask matrices of the sentence according to the different syntactic distances between words. Furthermore, to combine syntactic structure and semantic information, we equip the attention score matrices by syntactic mask matrices. Finally, we enhance the node representations with graph convolutional network over attention score matrices for ABSA. Experimental results on benchmark datasets illustrate that our proposed model outperforms state-of-the-art methods.
|
['Yanna Wang', 'Zili Zhou', 'Zheng Zhang']
| null | null | null | null |
naacl-2022-7
|
['aspect-based-sentiment-analysis']
|
['natural-language-processing']
|
[ 5.74755110e-02 1.12637289e-01 3.65230814e-02 -7.04113901e-01
-2.54206836e-01 -3.06177646e-01 2.91598797e-01 3.53083104e-01
-1.03413165e-01 2.14776322e-01 7.00057268e-01 -2.39730418e-01
-1.34473532e-01 -1.12612998e+00 -5.09832442e-01 -5.33940971e-01
2.55066544e-01 1.23865813e-01 -1.28546834e-01 -7.21813381e-01
4.10114229e-01 1.72638260e-02 -1.04531145e+00 4.38361585e-01
9.89544690e-01 1.08857012e+00 1.97229579e-01 2.50405759e-01
-7.17097104e-01 9.01629150e-01 -6.07886314e-01 -6.75685823e-01
-3.73588383e-01 -6.96307123e-01 -8.55378270e-01 1.98133171e-01
2.50204746e-02 2.16213062e-01 -1.18571430e-01 1.27204931e+00
2.23969832e-01 8.05586427e-02 4.17289466e-01 -1.02270508e+00
-1.01646674e+00 8.52639735e-01 -6.34901345e-01 2.97013670e-01
3.13985258e-01 1.00919470e-01 1.78594673e+00 -8.97590101e-01
4.48604465e-01 1.30891466e+00 5.32261491e-01 1.29039317e-01
-6.30517185e-01 -4.06617761e-01 8.36837232e-01 4.70578045e-01
-8.49706948e-01 1.55113325e-01 1.44055176e+00 -1.53083250e-01
1.18651366e+00 2.07742378e-01 1.09275413e+00 7.08057463e-01
3.97354096e-01 1.01441288e+00 7.76668131e-01 -7.56746083e-02
-4.69811261e-02 -1.17201127e-01 7.05412328e-01 9.49664176e-01
3.05586964e-01 -6.88668072e-01 -4.69658434e-01 1.33849695e-01
1.07233033e-01 3.31416875e-02 -2.41456270e-01 -2.41949350e-01
-1.00523365e+00 1.09735441e+00 8.66647959e-01 3.63080919e-01
-5.64476967e-01 2.05828082e-02 5.68300903e-01 2.84435272e-01
7.05741167e-01 5.93590379e-01 -7.25832283e-01 2.28682846e-01
-1.03602313e-01 -3.71114649e-02 6.75490558e-01 8.82152677e-01
1.01738250e+00 2.13707551e-01 -3.55979294e-01 7.02587366e-01
6.71897411e-01 5.34267247e-01 5.27014792e-01 -3.25502157e-01
7.32163668e-01 1.43082571e+00 -6.38722479e-01 -1.63809121e+00
-6.97241962e-01 -8.26976597e-01 -9.02713120e-01 -3.73538136e-01
-9.18484926e-02 -1.56782165e-01 -7.30057478e-01 1.60050511e+00
3.57629597e-01 -1.18404821e-01 1.01456083e-01 9.54393208e-01
1.17147934e+00 6.95075035e-01 7.11807981e-02 -8.92108604e-02
1.75998044e+00 -1.15658617e+00 -8.67164493e-01 -6.49302721e-01
8.23389232e-01 -5.42097569e-01 1.21314275e+00 -1.32320672e-01
-8.29783976e-01 -4.65569139e-01 -1.11033154e+00 -1.97379395e-01
-5.80809593e-01 -9.88123938e-02 9.85411406e-01 2.78279930e-01
-9.24138367e-01 3.46838266e-01 -5.14532745e-01 -2.98877925e-01
6.77772105e-01 4.83630776e-01 -2.18011767e-01 -6.00747019e-02
-1.46122289e+00 4.76377934e-01 3.52530360e-01 3.98544401e-01
-3.45951796e-01 -4.61531699e-01 -1.44857001e+00 4.33394313e-01
4.61430132e-01 -9.23267603e-01 7.83840537e-01 -1.15384090e+00
-1.20482111e+00 6.74240291e-01 -3.01551819e-01 -9.99227092e-02
-5.43914616e-01 9.72240418e-03 -3.94004494e-01 1.21348225e-01
3.02071482e-01 1.89114839e-01 7.04578161e-01 -1.02078640e+00
-3.84485751e-01 -7.90253162e-01 5.70803046e-01 5.65326214e-01
-6.88955247e-01 6.46404549e-02 -3.92246246e-01 -6.30904734e-01
1.08907886e-01 -4.73524809e-01 -4.69348580e-01 -6.67081714e-01
-6.07339799e-01 -3.05885017e-01 7.01074719e-01 -7.31429696e-01
1.36438179e+00 -2.09607530e+00 3.95407438e-01 2.86502302e-01
4.50934470e-01 1.34365991e-01 -4.46512938e-01 3.13539326e-01
-5.65836169e-02 1.05733179e-01 -4.70518440e-01 -1.47017658e-01
-1.35802209e-01 1.29202753e-01 -1.07572086e-01 1.32969040e-02
6.31527960e-01 1.23895025e+00 -1.13295996e+00 -4.71378982e-01
-1.23479940e-01 4.73208308e-01 -7.91066110e-01 1.89341918e-01
-3.34019989e-01 2.90225148e-01 -9.07462120e-01 6.21931255e-01
6.13406241e-01 -5.15564501e-01 2.59679973e-01 -4.66887623e-01
3.48746121e-01 5.81474721e-01 -5.49826324e-01 1.66496277e+00
-7.06224740e-01 3.47565293e-01 -7.89992884e-02 -1.21793187e+00
1.22206664e+00 -1.32857755e-01 2.34824896e-01 -8.05315495e-01
5.28985381e-01 -5.47523461e-02 3.58398229e-01 -5.43951213e-01
5.18982828e-01 -1.64660037e-01 -2.31132805e-01 4.71151441e-01
2.87980527e-01 -5.89712150e-03 2.40752503e-01 3.90538722e-01
9.47203934e-01 -1.00596808e-01 3.38673651e-01 -4.30551648e-01
1.06750453e+00 -1.71766564e-01 6.08072102e-01 7.80234411e-02
-6.54144436e-02 4.32929158e-01 9.79496360e-01 -5.40827215e-01
-5.95978856e-01 -5.15115023e-01 4.05636042e-01 1.05632806e+00
2.74602056e-01 -7.62009263e-01 -7.61801004e-01 -1.05552685e+00
-3.06594610e-01 5.84414840e-01 -9.03771102e-01 -5.29388726e-01
-6.12861633e-01 -1.07498169e+00 -2.30766565e-01 7.18969107e-01
6.48478985e-01 -1.40575635e+00 1.56201810e-01 1.17684580e-01
-1.33618623e-01 -1.02036262e+00 -7.15449691e-01 -1.32787168e-01
-7.25040495e-01 -1.18850935e+00 -2.22223118e-01 -1.07401109e+00
8.55241001e-01 4.75063533e-01 1.36721849e+00 3.52463245e-01
9.42671746e-02 2.86348164e-01 -5.69339991e-01 -3.56420517e-01
6.62606508e-02 4.76592213e-01 -4.90916461e-01 5.38567841e-01
7.36922324e-01 -6.40967369e-01 -7.95801997e-01 -4.20630760e-02
-7.62420774e-01 1.18450001e-02 7.55026817e-01 7.45445311e-01
7.19779432e-01 -1.26963720e-01 6.98591292e-01 -1.40472853e+00
1.06222618e+00 -6.72283232e-01 -2.38891706e-01 1.20004699e-01
-5.86029530e-01 -3.51016363e-03 8.27399552e-01 2.52515703e-01
-1.07952654e+00 -2.39235193e-01 -3.95101458e-01 -8.05264711e-02
1.90619826e-01 1.16690493e+00 -6.78473651e-01 2.11708605e-01
6.29041791e-02 4.26252156e-01 -7.85655975e-02 -1.53401211e-01
3.90052468e-01 4.07849491e-01 5.41494451e-02 -2.82320738e-01
4.63601738e-01 4.35173333e-01 1.61555469e-01 -6.39949679e-01
-1.55252528e+00 -4.57395673e-01 -5.39008737e-01 -8.02659094e-02
1.02055407e+00 -8.26999485e-01 -5.95803976e-01 4.38556075e-01
-1.25221717e+00 1.96586907e-01 -1.52577221e-01 2.16053367e-01
-3.03112090e-01 4.19347852e-01 -5.24522901e-01 -4.26434100e-01
-7.83710301e-01 -1.11535156e+00 1.10703361e+00 3.11023384e-01
-5.88564388e-02 -1.41983044e+00 -4.68183681e-02 5.51129818e-01
3.55335176e-01 8.24450105e-02 1.06394804e+00 -8.95722389e-01
-5.27315974e-01 -2.55385697e-01 -4.44217414e-01 3.30373377e-01
2.40826353e-01 -3.51865411e-01 -7.65988529e-01 -1.22689731e-01
2.16720864e-01 7.33547062e-02 1.02770877e+00 2.44596913e-01
1.15922284e+00 -4.40312564e-01 -5.45965880e-02 6.18039966e-01
1.20890617e+00 -7.95342624e-02 4.04419810e-01 4.24299896e-01
1.35108495e+00 7.56241262e-01 5.82290053e-01 2.25426674e-01
8.25127900e-01 1.53940558e-01 6.08680367e-01 -7.90671539e-03
-1.27840236e-01 -2.83837378e-01 2.30976835e-01 1.70510995e+00
1.36732655e-02 -1.53779581e-01 -7.23605633e-01 6.22332811e-01
-1.68413043e+00 -5.36175370e-01 -4.59768564e-01 1.51162767e+00
3.81847650e-01 2.80095160e-01 -2.49375999e-01 -2.33280077e-03
6.05846047e-01 7.44634151e-01 -4.75816131e-01 -5.47109663e-01
-2.76172876e-01 4.36311699e-02 -4.30201590e-02 4.71984863e-01
-1.04181039e+00 1.02202916e+00 4.89839792e+00 6.73727334e-01
-8.16067636e-01 7.62792081e-02 6.23856068e-01 3.05174142e-01
-1.01522422e+00 3.87502611e-02 -5.15711546e-01 3.85043025e-01
7.06739843e-01 -4.30281520e-01 1.49517000e-01 9.51903224e-01
-1.34922445e-01 5.16212404e-01 -6.18101716e-01 5.79860151e-01
5.03908575e-01 -1.16385782e+00 5.68242550e-01 -1.80975050e-01
7.30495512e-01 -5.26163019e-02 1.23323975e-02 5.65281630e-01
6.83551133e-02 -7.12584794e-01 1.57187775e-01 4.00899321e-01
2.34301388e-01 -1.04473650e+00 1.27852333e+00 -1.59177676e-01
-1.63583684e+00 6.17366843e-02 -5.40568173e-01 -2.18567953e-01
1.27400039e-02 5.95175982e-01 -3.68913084e-01 9.23275590e-01
5.01019776e-01 1.43176186e+00 -8.33685219e-01 5.96825719e-01
-6.97167456e-01 5.93980491e-01 2.80016512e-01 -5.35515189e-01
5.72541058e-01 -4.89495963e-01 5.73461413e-01 9.79809821e-01
1.14735670e-01 5.02482690e-02 -7.32626468e-02 7.54529834e-01
-3.61788690e-01 6.93458796e-01 -7.18926430e-01 -2.42623761e-01
-5.17685451e-02 1.64530218e+00 -7.16788471e-01 -2.16688022e-01
-7.80842423e-01 8.80951345e-01 6.81218088e-01 3.26073557e-01
-4.56367135e-01 -6.84167802e-01 7.98377454e-01 -1.54575378e-01
4.52506214e-01 6.41440181e-03 -4.17676896e-01 -1.36465669e+00
7.88295791e-02 -6.79663837e-01 5.91885865e-01 -8.70336950e-01
-1.42136848e+00 8.89648795e-01 -5.19526482e-01 -1.04296470e+00
1.70406476e-01 -6.82117105e-01 -1.09459591e+00 8.78082871e-01
-1.72618532e+00 -1.50176787e+00 -3.49169761e-01 4.83997554e-01
6.16527319e-01 -2.86253631e-01 6.32558346e-01 -2.25756746e-02
-5.74324787e-01 3.90206099e-01 -3.82270873e-01 4.11611885e-01
7.55457655e-02 -1.27462339e+00 7.61985719e-01 6.07317448e-01
3.22719067e-02 7.78676629e-01 2.92384058e-01 -6.44010842e-01
-1.62608039e+00 -1.28002179e+00 1.05369699e+00 -3.61553729e-01
1.01703584e+00 -2.46662170e-01 -1.14335215e+00 7.88983941e-01
4.82719183e-01 -1.74820498e-01 9.09398913e-01 6.52667463e-01
-4.54339385e-01 -2.85763264e-01 -5.95187545e-01 5.64564466e-01
9.95725274e-01 -6.04021549e-01 -7.86068916e-01 3.07616353e-01
1.35207808e+00 -1.16337657e-01 -6.88917994e-01 3.93696904e-01
1.29218638e-01 -8.29755366e-01 6.64079905e-01 -8.76295388e-01
8.41704130e-01 -2.06714183e-01 -6.41015917e-02 -1.63777912e+00
-4.73633140e-01 -5.89132793e-02 -2.18803622e-03 1.39181364e+00
6.87080801e-01 -7.65210390e-01 7.13534236e-01 1.22657865e-01
-4.14021850e-01 -1.14399409e+00 -5.75954378e-01 -2.86023319e-01
-2.62509398e-02 -4.34577644e-01 1.00204277e+00 1.05780792e+00
2.08069488e-01 1.11220944e+00 -7.28354231e-02 1.53590918e-01
3.42533499e-01 6.37405515e-01 4.14773315e-01 -1.12496996e+00
-1.00823924e-01 -7.48661637e-01 -5.68681955e-01 -7.06728935e-01
4.82221305e-01 -1.30566108e+00 -3.21888506e-01 -1.99572515e+00
4.48875159e-01 1.79936409e-01 -6.81045413e-01 2.84842521e-01
-8.16030562e-01 4.69704308e-02 6.54326677e-02 -2.72815973e-01
-1.01466131e+00 1.03726935e+00 1.57972860e+00 -5.04344583e-01
9.35242325e-02 -2.23062247e-01 -1.26126122e+00 8.43551517e-01
8.60235691e-01 -2.89403468e-01 -5.47821105e-01 -5.44634581e-01
9.14276183e-01 -2.17137560e-01 -1.32620269e-02 -5.84811211e-01
5.34154698e-02 -1.53175090e-02 1.47414193e-01 -5.62658668e-01
1.77464500e-01 -7.90686071e-01 -6.36874974e-01 4.34357762e-01
-2.35225350e-01 1.78099573e-01 3.46661694e-02 8.29771101e-01
-5.90411782e-01 -9.85862240e-02 2.39912480e-01 -2.46317163e-01
-6.22024000e-01 5.55025458e-01 -1.36967301e-01 2.80460030e-01
6.94568694e-01 3.04597914e-01 -3.35089654e-01 -3.19142908e-01
-5.73748946e-01 5.55838525e-01 5.92044778e-02 5.91489434e-01
8.64825964e-01 -1.55253065e+00 -6.55810773e-01 2.73725301e-01
3.73234540e-01 1.33811787e-01 5.87504327e-01 7.46952236e-01
-3.56697887e-01 2.99774677e-01 -1.15373209e-01 -2.67526597e-01
-1.19694960e+00 7.58474112e-01 1.14593580e-01 -5.00843108e-01
-4.17121679e-01 8.92204165e-01 5.21283567e-01 -7.39860654e-01
-4.87691849e-01 -3.49246174e-01 -9.07404602e-01 2.24831134e-01
4.88468558e-01 -1.23133615e-01 1.62808914e-02 -8.37485850e-01
-4.50078189e-01 8.72766197e-01 -2.55231112e-01 4.26873952e-01
1.45007837e+00 -2.48015210e-01 -5.61121941e-01 3.25000137e-01
1.33121634e+00 7.85643682e-02 -6.49875462e-01 -2.95669377e-01
-7.05692247e-02 -3.31865549e-01 7.21040964e-02 -4.26967800e-01
-1.62633824e+00 1.16103160e+00 -2.03117400e-01 4.07025665e-01
1.22160470e+00 1.05962232e-01 9.50999320e-01 4.51435983e-01
-1.13791414e-01 -8.09097409e-01 2.04310536e-01 7.08732486e-01
1.03055692e+00 -1.23316169e+00 -2.82456465e-02 -6.19873226e-01
-9.18370485e-01 9.26067829e-01 8.64348173e-01 -3.27889085e-01
8.54807675e-01 -2.16422468e-01 1.05793387e-01 -7.29626954e-01
-6.33550763e-01 -4.65558559e-01 4.98648167e-01 5.11903405e-01
5.14301181e-01 2.26642061e-02 -4.77503031e-01 1.24344647e+00
-4.15761322e-01 -5.60938358e-01 3.54805082e-01 6.53910816e-01
-3.96737516e-01 -8.65857840e-01 1.61833942e-01 5.97674370e-01
-4.42285746e-01 -5.50544262e-01 -5.24783015e-01 3.64981771e-01
-2.29635522e-01 9.29829836e-01 1.71161536e-02 -5.81388235e-01
5.53132534e-01 -2.42121592e-02 4.09003645e-02 -7.28948891e-01
-8.15164387e-01 -1.14344738e-01 2.41575867e-01 -4.77700621e-01
-4.30878967e-01 -5.09387493e-01 -1.33407295e+00 -4.40501980e-03
-1.55250683e-01 2.92596728e-01 6.67977154e-01 1.12206316e+00
5.61586618e-01 1.20720780e+00 8.53508532e-01 -2.87565500e-01
4.49179448e-02 -1.04343259e+00 -5.60038924e-01 5.99986315e-01
1.36574805e-01 -3.22311670e-01 -2.86529481e-01 -4.54804927e-01]
|
[11.497591972351074, 6.605571746826172]
|
6e7c799e-7808-47ec-bf82-1ecb3b6da057
|
collaborative-training-of-medical-artificial
|
2211.13606
| null |
https://arxiv.org/abs/2211.13606v2
|
https://arxiv.org/pdf/2211.13606v2.pdf
|
Collaborative Training of Medical Artificial Intelligence Models with non-uniform Labels
|
Due to the rapid advancements in recent years, medical image analysis is largely dominated by deep learning (DL). However, building powerful and robust DL models requires training with large multi-party datasets. While multiple stakeholders have provided publicly available datasets, the ways in which these data are labeled vary widely. For Instance, an institution might provide a dataset of chest radiographs containing labels denoting the presence of pneumonia, while another institution might have a focus on determining the presence of metastases in the lung. Training a single AI model utilizing all these data is not feasible with conventional federated learning (FL). This prompts us to propose an extension to the widespread FL process, namely flexible federated learning (FFL) for collaborative training on such data. Using 695,000 chest radiographs from five institutions from across the globe - each with differing labels - we demonstrate that having heterogeneously labeled datasets, FFL-based training leads to significant performance increase compared to conventional FL training, where only the uniformly annotated images are utilized. We believe that our proposed algorithm could accelerate the process of bringing collaborative training methods from research and simulation phase to the real-world applications in healthcare.
|
['Daniel Truhn', 'Sven Nebelung', 'Christiane Kuhl', 'Jakob Nikolas Kather', 'Firas Khader', 'Gustav Mueller-Franzes', 'Marwin Saehn', 'Peter Isfort', 'Soroosh Tayebi Arasteh']
|
2022-11-24
| null | null | null | null |
['privacy-preserving-deep-learning', 'privacy-preserving-deep-learning']
|
['methodology', 'natural-language-processing']
|
[ 1.07131399e-01 5.67964315e-02 -3.64717156e-01 -5.46762168e-01
-1.07357347e+00 -5.12009442e-01 2.10701182e-01 1.98452279e-01
-2.84460515e-01 7.69015193e-01 2.78660268e-01 -6.21329486e-01
-3.22513431e-01 -8.68440688e-01 -5.96787333e-01 -7.90000439e-01
1.42971026e-02 6.01751208e-01 -2.04779655e-01 3.97775739e-01
-3.88295650e-01 6.12758398e-01 -1.09075475e+00 6.81290388e-01
7.04980552e-01 1.07986009e+00 -2.62767244e-02 6.61564708e-01
-2.08901063e-01 1.36639488e+00 -5.07434428e-01 -4.08840716e-01
5.63658178e-01 -2.59825379e-01 -8.75807762e-01 1.34177774e-01
3.48139495e-01 -5.00821352e-01 -3.15028876e-01 8.20020318e-01
7.37090290e-01 -3.18681687e-01 3.38637918e-01 -1.33883429e+00
-7.01637089e-01 7.22439110e-01 -4.01135683e-01 3.90768498e-02
-2.52915052e-04 2.80094653e-01 8.28211963e-01 -5.73685229e-01
8.38080406e-01 5.87475717e-01 8.13391626e-01 5.41427195e-01
-5.53611517e-01 -8.24264705e-01 -1.21078074e-01 5.50876558e-02
-1.01540995e+00 -7.40848184e-02 5.88508070e-01 -4.80834872e-01
3.67495924e-01 4.05484706e-01 3.55501354e-01 1.18424332e+00
1.89927623e-01 8.74342918e-01 1.19792545e+00 -4.00584459e-01
2.38590196e-01 1.08695574e-01 -4.26978059e-02 1.00613320e+00
4.82133985e-01 -7.13587925e-02 -3.52163196e-01 -5.67148626e-01
5.26614070e-01 7.14516103e-01 -1.50015637e-01 -5.67994356e-01
-1.55965412e+00 8.22522402e-01 6.41938150e-01 4.02845532e-01
-5.95320940e-01 -1.67925134e-01 5.40032029e-01 2.05735609e-01
3.67547721e-01 3.56994778e-01 -5.47865927e-01 2.22273961e-01
-1.05726242e+00 -1.06965221e-01 8.04454923e-01 7.55868495e-01
6.19313538e-01 -2.80412406e-01 -2.39301875e-01 6.00215554e-01
1.92747667e-01 2.30013549e-01 6.98801696e-01 -1.24017179e+00
2.79914379e-01 9.78541672e-01 -8.41443986e-02 -7.49414921e-01
-3.99105191e-01 -5.54231644e-01 -1.17273653e+00 3.09943501e-02
4.40531313e-01 -5.63018680e-01 -8.97031188e-01 1.45800781e+00
5.14873862e-01 2.45722100e-01 5.85934594e-02 8.24803472e-01
7.33073950e-01 1.09907933e-01 1.19237788e-01 5.61902970e-02
1.14850128e+00 -1.01722968e+00 -6.04458928e-01 1.34541497e-01
8.58806908e-01 -4.56396461e-01 7.17960596e-01 5.06913364e-01
-7.48149633e-01 -2.70711243e-01 -7.64549553e-01 3.16795707e-01
-3.92526835e-01 2.83609144e-02 8.89249325e-01 6.17847443e-01
-1.01178837e+00 2.36540571e-01 -9.10595953e-01 -4.20965970e-01
9.67126667e-01 4.02155042e-01 -6.31422043e-01 -5.32814860e-01
-9.67619121e-01 6.53058708e-01 1.23236990e-02 -6.36412054e-02
-1.18245077e+00 -8.73075306e-01 -4.64955211e-01 -1.33203804e-01
2.29125842e-01 -1.06125522e+00 1.25474262e+00 -9.67332065e-01
-7.93895125e-01 9.67969120e-01 3.70098919e-01 -5.96185982e-01
7.67809689e-01 5.97016588e-02 -4.96553212e-01 1.76952824e-01
2.16842145e-01 5.61501801e-01 5.63622534e-01 -1.32219803e+00
-7.89378822e-01 -3.44315618e-01 1.47606358e-01 -2.26162560e-02
-6.38728857e-01 5.72542362e-02 -2.35627100e-01 -5.75398982e-01
-1.93861634e-01 -9.90210533e-01 -5.54212570e-01 3.42917174e-01
-2.86414146e-01 -6.12579733e-02 1.12817824e+00 -4.58556920e-01
1.08484101e+00 -2.09091783e+00 -3.83613169e-01 1.45360634e-01
7.52240479e-01 1.95400134e-01 1.84987649e-01 3.07720959e-01
3.66288312e-02 2.05589101e-01 -2.55249560e-01 -1.69775963e-01
-2.42185295e-01 4.84974653e-01 8.21482465e-02 5.79152763e-01
-2.07628652e-01 8.91420960e-01 -9.93470430e-01 -9.69504356e-01
1.06018156e-01 3.43878478e-01 -6.12960696e-01 3.92638832e-01
-4.57421169e-02 6.34072125e-01 -7.96833098e-01 1.11736667e+00
2.99507558e-01 -1.09218144e+00 3.21701795e-01 -2.75939345e-01
1.21352300e-01 -4.18854445e-01 -8.32072139e-01 1.86788619e+00
-4.62827206e-01 3.79698694e-01 2.17826784e-01 -9.37641740e-01
5.63398540e-01 7.78058648e-01 1.30105376e+00 -2.88107157e-01
2.33627930e-01 3.19345623e-01 1.43455938e-01 -6.13548696e-01
-2.17100307e-02 -1.06540054e-01 7.22642988e-02 8.88219357e-01
2.27973387e-01 3.76550496e-01 -1.64344788e-01 4.24503118e-01
1.74216127e+00 -2.01686203e-01 1.10624939e-01 1.69242755e-01
3.41752797e-01 3.41890872e-01 5.36053598e-01 7.19442189e-01
-7.19978809e-01 5.37925661e-01 -2.77485177e-02 -6.45103693e-01
-8.91136408e-01 -8.44339252e-01 -2.55000532e-01 9.91177380e-01
-2.78455317e-01 -1.40849784e-01 -5.56544900e-01 -1.14047694e+00
1.65820107e-01 1.64704815e-01 -7.65327156e-01 5.53912222e-02
-4.65209067e-01 -7.10052490e-01 6.93002582e-01 4.65034604e-01
5.62068880e-01 -1.19242787e+00 -9.61367607e-01 1.45432070e-01
-1.66838422e-01 -9.03562546e-01 -2.34506488e-01 3.62010539e-01
-8.17783058e-01 -1.32617998e+00 -8.93701911e-01 -7.12411642e-01
8.33616614e-01 1.37207866e-01 1.17196822e+00 1.84371233e-01
-6.29501581e-01 6.80071235e-01 -4.37403798e-01 -6.29901826e-01
-5.40606022e-01 -3.64483856e-02 -1.69299528e-01 1.87595546e-01
1.92655772e-01 -1.58373192e-01 -9.97687042e-01 2.04183757e-01
-1.00281286e+00 1.36594579e-01 8.30534399e-01 8.80674839e-01
5.95640719e-01 1.42336916e-03 6.92048550e-01 -1.65142679e+00
4.64231879e-01 -1.10780513e+00 5.41940518e-02 5.71506858e-01
-6.58653021e-01 -2.53941655e-01 6.51954114e-01 -3.07385534e-01
-1.07302094e+00 2.27891326e-01 1.01371877e-01 -7.09352016e-01
-3.24130088e-01 6.58184290e-01 2.92120844e-01 -1.70629084e-01
8.05746138e-01 -2.04800352e-01 1.27049834e-01 -2.59317786e-01
2.27954298e-01 1.01417029e+00 5.58589041e-01 -6.26142979e-01
5.29481292e-01 6.01056516e-01 -1.50304154e-01 -2.62801815e-02
-1.05845809e+00 -6.85671329e-01 -4.42093760e-01 -3.63950670e-01
7.83209741e-01 -8.80872190e-01 -2.69986272e-01 1.33418098e-01
-6.49772942e-01 -1.26627520e-01 -5.25550902e-01 5.30519962e-01
-3.64130080e-01 2.44486518e-02 -5.92222273e-01 -3.38121921e-01
-6.55404210e-01 -1.19047117e+00 8.23746562e-01 1.15399044e-02
-1.75330132e-01 -1.05738854e+00 2.07249910e-01 7.81649351e-01
7.24722147e-01 5.78013361e-01 9.57235157e-01 -9.49663758e-01
-5.18634737e-01 -4.09236461e-01 -8.39245692e-02 3.05225492e-01
7.05354512e-01 -9.10182074e-02 -1.00092208e+00 -4.09658790e-01
1.78783014e-01 -6.65870130e-01 2.47040644e-01 3.28725934e-01
1.53042352e+00 -3.08546335e-01 -5.01523316e-01 6.00464702e-01
1.45008433e+00 2.00665489e-01 -2.50423420e-02 4.67883319e-01
6.55528426e-01 2.64103562e-01 5.20675957e-01 6.20131910e-01
4.28270847e-01 -8.05727392e-02 5.81291497e-01 -5.89573443e-01
-2.68032521e-01 6.55753240e-02 -9.54252928e-02 8.55520487e-01
1.53425291e-01 -1.58263311e-01 -1.33716869e+00 5.62162280e-01
-1.78889608e+00 -7.34355509e-01 2.09997416e-01 1.86588550e+00
8.39077294e-01 -3.02102417e-01 -1.17484927e-01 -7.53503591e-02
6.73326731e-01 -1.05904214e-01 -9.63194072e-01 1.47801898e-02
1.86454691e-02 2.24044859e-01 6.27858102e-01 -2.55072445e-01
-1.09173954e+00 1.93803087e-01 6.76232529e+00 3.15434575e-01
-1.33831286e+00 4.98977929e-01 1.00866413e+00 -9.81929749e-02
-2.28031039e-01 -3.79622430e-01 -2.80945808e-01 4.41056609e-01
1.16034257e+00 -3.43420208e-01 -5.81966490e-02 9.80853260e-01
-2.90754512e-02 1.89610451e-01 -1.11148858e+00 9.19110537e-01
-5.56589365e-02 -1.77345526e+00 1.36398017e-01 2.15488061e-01
1.06325424e+00 5.16070724e-01 -2.75653377e-02 2.07325265e-01
7.94506550e-01 -1.09652317e+00 1.83343634e-01 6.00495040e-01
9.26729023e-01 -5.59098840e-01 8.46376777e-01 6.55367136e-01
-8.54803324e-01 -3.87739003e-01 -7.13104531e-02 4.50649172e-01
-2.76552260e-01 4.44403052e-01 -1.11193132e+00 8.35061073e-01
7.96683371e-01 5.81111550e-01 -6.43981457e-01 1.08503604e+00
3.36472332e-01 9.02970672e-01 -1.19291671e-01 3.59475285e-01
2.69923031e-01 1.07950971e-01 -1.57594532e-01 1.01730108e+00
2.97187001e-01 1.23743951e-01 4.02511597e-01 3.55870157e-01
-5.23786604e-01 1.29279211e-01 -7.15285838e-01 -1.62963942e-02
5.01801550e-01 1.57696378e+00 -5.10330498e-01 -4.27521139e-01
-9.10276592e-01 6.16797030e-01 2.36043245e-01 -4.32641022e-02
-9.00125802e-01 1.58781022e-01 2.64900923e-01 9.07006785e-02
-6.32849038e-02 2.29524344e-01 -1.45072088e-01 -9.72452581e-01
-3.67759764e-01 -1.44105947e+00 9.17950332e-01 -5.48181832e-01
-1.78644502e+00 8.29175949e-01 -3.77599388e-01 -1.45022225e+00
-2.24945828e-01 -4.02955890e-01 -3.73636723e-01 6.48334086e-01
-1.47607732e+00 -1.52819979e+00 -5.89333534e-01 1.15513122e+00
5.03512204e-01 -4.24926698e-01 1.07540977e+00 6.93868101e-01
-5.89139819e-01 4.19175178e-01 7.31344670e-02 4.57247049e-01
9.69003558e-01 -1.06587791e+00 -2.80503362e-01 6.01070166e-01
1.54346198e-01 6.54698849e-01 1.09204940e-01 -4.80537504e-01
-1.55167162e+00 -1.40968156e+00 4.79679167e-01 -4.55556840e-01
7.38156557e-01 1.74141362e-01 -6.99073493e-01 9.06735241e-01
3.37682098e-01 7.74735749e-01 1.32637310e+00 -6.31808117e-02
-9.98366773e-02 -1.17477894e-01 -1.51052344e+00 1.35067046e-01
8.26224208e-01 -3.77600908e-01 -2.74544805e-01 8.16308558e-01
5.73338747e-01 -3.56480032e-01 -1.25787306e+00 5.16541123e-01
2.98005909e-01 -7.46638894e-01 7.33241379e-01 -7.76100278e-01
3.39913368e-01 -1.73626676e-01 -2.00743705e-01 -1.09443104e+00
-2.60962099e-01 -2.61933237e-01 -2.81128347e-01 8.43200684e-01
2.07658127e-01 -6.60161078e-01 1.09589124e+00 9.94631529e-01
-2.66822666e-01 -1.10229814e+00 -7.98117042e-01 -3.39216232e-01
1.08640276e-01 -3.61401081e-01 7.88360596e-01 1.46263623e+00
-3.49089712e-01 -2.32728824e-01 -2.43110389e-01 1.83626890e-01
6.72558904e-01 3.24721217e-01 6.47582114e-01 -1.23404014e+00
-2.91915298e-01 -1.32208364e-02 -1.19772851e-01 -3.73730510e-01
-2.74009369e-02 -1.12460184e+00 -3.54383081e-01 -1.79256189e+00
5.37556291e-01 -1.12598777e+00 -8.84984672e-01 9.25406694e-01
1.31883305e-02 2.46464133e-01 5.16353436e-02 6.25145733e-01
-8.02186906e-01 -7.89224431e-02 1.40904510e+00 -2.51164854e-01
4.53651637e-01 -2.93630064e-02 -8.38280141e-01 6.95964098e-01
7.24912167e-01 -5.89170098e-01 -3.79054546e-01 -5.53512454e-01
-5.92443123e-02 2.86323518e-01 2.46654227e-01 -1.14851499e+00
7.68137395e-01 -1.54119238e-01 5.15732050e-01 -4.42970842e-01
-1.53709710e-01 -1.30079675e+00 5.47503769e-01 5.99448264e-01
-3.90230507e-01 4.03668061e-02 -9.71909538e-02 5.83186209e-01
-4.40433353e-01 2.38764584e-02 5.65634012e-01 -6.02892637e-01
-6.15883768e-01 7.05504000e-01 4.21262495e-02 -6.42637834e-02
1.34229767e+00 -1.56051502e-01 -4.66971397e-01 -1.52209207e-01
-7.15404153e-01 2.50980914e-01 4.19850826e-01 6.72786906e-02
4.82066542e-01 -1.09632123e+00 -8.10947001e-01 2.33954072e-01
7.43927509e-02 2.50901014e-01 3.41312379e-01 8.19065213e-01
-5.89989543e-01 3.52666825e-01 -1.98221922e-01 -6.34226501e-01
-1.26587152e+00 5.47853768e-01 4.56571430e-01 -4.84438151e-01
-7.00717509e-01 7.68531919e-01 6.00411966e-02 -8.09575558e-01
4.06404167e-01 1.21397696e-01 1.60946697e-01 -7.54055604e-02
1.68702319e-01 2.89054960e-01 4.42116737e-01 -3.33240658e-01
-3.83787185e-01 -4.09849435e-02 -1.38973832e-01 4.07174379e-01
1.58044958e+00 1.56711951e-01 -5.44523224e-02 4.67680246e-01
1.14634252e+00 -3.99868228e-02 -1.10405338e+00 -2.66925097e-01
-1.02645546e-01 -3.92914355e-01 1.95173919e-01 -1.16065216e+00
-1.52104199e+00 4.44792002e-01 8.15565765e-01 -7.09065050e-02
1.22714674e+00 -9.57233757e-02 8.73677313e-01 3.77886802e-01
8.34428251e-01 -6.95937514e-01 1.17836915e-01 5.04414085e-03
3.98923159e-01 -1.53992140e+00 1.27169266e-01 1.04121864e-01
-8.54750872e-01 9.54651177e-01 5.98948538e-01 5.44230901e-02
7.93678880e-01 4.98751342e-01 4.93329823e-01 -3.66956711e-01
-9.66920495e-01 2.24927917e-01 -4.27548252e-02 3.39834899e-01
5.96913636e-01 1.99337170e-01 8.02642107e-02 3.61050725e-01
1.67801380e-01 4.39654112e-01 3.01356733e-01 1.32266057e+00
-1.81050971e-02 -1.34094787e+00 -4.45067048e-01 1.02851462e+00
-6.79341316e-01 1.29283875e-01 -2.34495640e-01 7.54054606e-01
5.12427807e-01 7.97458768e-01 -1.10203184e-01 -1.51183203e-01
1.07184388e-01 1.48879811e-01 1.65226310e-01 -7.94040263e-01
-9.89471555e-01 -3.15589428e-01 -1.66293323e-01 -4.59042609e-01
-7.16942370e-01 -5.57335615e-01 -1.34322870e+00 -2.68411845e-01
-1.94673717e-01 1.75253645e-01 6.18617475e-01 6.45915389e-01
5.27250230e-01 6.49547637e-01 8.43026638e-01 -3.00164908e-01
-5.12447119e-01 -6.02967858e-01 -5.16012967e-01 6.30481064e-01
3.68294716e-01 -4.18758780e-01 -2.91334152e-01 1.74146622e-01]
|
[6.099012851715088, 6.463272571563721]
|
36c78981-5f5c-4d28-a56c-a69e61cacf94
|
multimodal-pre-training-framework-for
|
2303.11879
| null |
https://arxiv.org/abs/2303.11879v1
|
https://arxiv.org/pdf/2303.11879v1.pdf
|
Multimodal Pre-training Framework for Sequential Recommendation via Contrastive Learning
|
Sequential recommendation systems utilize the sequential interactions of users with items as their main supervision signals in learning users' preferences. However, existing methods usually generate unsatisfactory results due to the sparsity of user behavior data. To address this issue, we propose a novel pre-training framework, named Multimodal Sequence Mixup for Sequential Recommendation (MSM4SR), which leverages both users' sequential behaviors and items' multimodal content (\ie text and images) for effectively recommendation. Specifically, MSM4SR tokenizes each item image into multiple textual keywords and uses the pre-trained BERT model to obtain initial textual and visual features of items, for eliminating the discrepancy between the text and image modalities. A novel backbone network, \ie Multimodal Mixup Sequence Encoder (M$^2$SE), is proposed to bridge the gap between the item multimodal content and the user behavior, using a complementary sequence mixup strategy. In addition, two contrastive learning tasks are developed to assist M$^2$SE in learning generalized multimodal representations of the user behavior sequence. Extensive experiments on real-world datasets demonstrate that MSM4SR outperforms state-of-the-art recommendation methods. Moreover, we further verify the effectiveness of MSM4SR on other challenging tasks including cold-start and cross-domain recommendation.
|
['Zhiqi Shen', 'Xin Zhou', 'Lingzi Zhang']
|
2023-03-21
| null | null | null | null |
['sequential-recommendation']
|
['miscellaneous']
|
[ 2.75474578e-01 -8.52718532e-01 -6.98009014e-01 -4.59383041e-01
-6.69747472e-01 -6.35828614e-01 3.92084718e-01 -3.49349707e-01
-4.05919611e-01 2.85639226e-01 5.29495299e-01 -1.77081779e-01
8.54885057e-02 -3.82902920e-01 -8.38493049e-01 -6.55452371e-01
2.67850488e-01 1.32244870e-01 -1.70542896e-01 -3.09298724e-01
3.24123055e-01 -1.44651815e-01 -1.52747142e+00 1.05101216e+00
8.99568200e-01 1.21309090e+00 6.66818738e-01 4.06784356e-01
-2.99096763e-01 5.99473655e-01 -1.38486728e-01 -5.16828239e-01
1.17952630e-01 -5.28766274e-01 -3.78949255e-01 4.64064270e-01
4.69839275e-01 -7.26365268e-01 -5.65911651e-01 8.24176311e-01
4.94715869e-01 7.17653155e-01 6.51808739e-01 -8.92955303e-01
-1.34794962e+00 8.62019300e-01 -7.15169668e-01 4.54304303e-04
6.31114423e-01 1.40539750e-01 1.25251281e+00 -1.23602307e+00
4.78012592e-01 1.25602877e+00 2.98707813e-01 6.26028299e-01
-1.09319305e+00 -7.05895901e-01 7.66164660e-01 2.59449512e-01
-1.11136699e+00 -2.04877183e-01 5.96438229e-01 -2.83609688e-01
7.02184975e-01 2.79504091e-01 2.67081887e-01 1.63322914e+00
-3.22029918e-01 1.52610219e+00 6.35654151e-01 -9.38171446e-02
-2.77417928e-01 3.65603060e-01 3.93082321e-01 7.19818115e-01
-2.22882122e-01 -8.01546872e-02 -6.00557923e-01 5.91563620e-02
7.19367206e-01 7.27981567e-01 -4.21982586e-01 -1.67558998e-01
-1.39440894e+00 7.40337133e-01 3.32320809e-01 1.80407077e-01
-2.27163509e-01 -2.24208534e-01 5.13473213e-01 4.93976295e-01
3.05171281e-01 1.50357962e-01 -4.09098327e-01 3.07710856e-01
-6.29706979e-01 -1.63346790e-02 4.03755277e-01 1.11037111e+00
5.25087714e-01 -3.26276966e-03 -5.05183995e-01 1.28759265e+00
5.58537781e-01 7.07934618e-01 6.85924411e-01 -4.17738467e-01
8.24564159e-01 4.92880553e-01 2.73923963e-01 -1.01322222e+00
-9.97051969e-02 -4.17019159e-01 -8.17706168e-01 -7.65443087e-01
1.98043928e-01 -1.63151920e-01 -9.36132073e-01 1.67377496e+00
4.78018038e-02 2.03556582e-01 5.54756038e-02 1.26022899e+00
1.26655722e+00 1.01679909e+00 7.41472989e-02 -3.18521947e-01
1.28870547e+00 -1.42387879e+00 -5.76190591e-01 -1.12693213e-01
5.04931033e-01 -7.24802792e-01 1.44314730e+00 3.33962649e-01
-1.04435396e+00 -8.29091311e-01 -7.40177095e-01 -1.00367785e-01
-2.05281287e-01 7.94183850e-01 4.47567135e-01 2.49948338e-01
-6.22969687e-01 1.70764938e-01 -4.01178539e-01 -1.50391191e-01
2.41010278e-01 4.85748798e-01 -1.27044633e-01 -5.12306392e-01
-1.25694358e+00 4.49786156e-01 1.31777138e-01 2.77003109e-01
-7.88108706e-01 -4.95152593e-01 -9.54238355e-01 2.59797573e-02
6.52844131e-01 -5.78469515e-01 1.22560573e+00 -1.45788133e+00
-1.57847559e+00 4.56903726e-01 -3.22747737e-01 1.57586530e-01
3.74138122e-03 -3.08995068e-01 -8.54847968e-01 6.98129088e-02
-8.56537670e-02 5.45277476e-01 1.10702825e+00 -1.61341166e+00
-9.23822343e-01 -2.77611643e-01 3.07947919e-02 5.25907040e-01
-7.32113063e-01 5.42829745e-03 -1.13887119e+00 -1.04910970e+00
-2.66355753e-01 -9.30442452e-01 -2.38910779e-01 -5.87863922e-01
-2.79686958e-01 -2.18979359e-01 6.45457089e-01 -7.15146959e-01
1.40174758e+00 -2.30863047e+00 5.50779760e-01 4.25117254e-01
-9.92656127e-02 2.16935292e-01 -8.38506162e-01 2.90987730e-01
9.92110595e-02 -2.59316534e-01 1.58917248e-01 -6.26340508e-01
1.40392676e-01 2.48608559e-01 -5.05700171e-01 2.49417454e-01
-3.46182674e-01 1.08977532e+00 -8.96225393e-01 -2.90744394e-01
2.19282553e-01 4.20514196e-01 -6.84316397e-01 5.98303080e-01
-4.38156247e-01 4.87592369e-01 -5.34975052e-01 6.08958244e-01
6.29161358e-01 -5.90287089e-01 4.37116295e-01 -6.31813586e-01
-1.31538184e-02 1.97530031e-01 -9.61006820e-01 1.98775756e+00
-6.07886970e-01 -6.42813966e-02 -1.05008543e-01 -9.52211261e-01
5.88966608e-01 3.25365543e-01 5.70586085e-01 -1.19097126e+00
1.06968895e-01 -5.76977944e-03 -4.03645396e-01 -7.36790955e-01
7.75974929e-01 9.69772488e-02 -1.94070399e-01 7.34825194e-01
3.27887267e-01 6.04549170e-01 1.64078847e-01 3.95191997e-01
5.57327569e-01 2.79500365e-01 -1.33771643e-01 2.92804122e-01
7.04206109e-01 -4.06280160e-01 3.85146737e-01 8.43972206e-01
2.33242601e-01 5.06752074e-01 -7.98768457e-03 -2.02514529e-01
-7.61658311e-01 -8.90257716e-01 3.41754645e-01 1.76091325e+00
6.63848817e-01 -4.70376164e-01 -2.79049784e-01 -1.23863125e+00
-1.14485942e-01 5.91311932e-01 -5.91383159e-01 -3.17886233e-01
-5.17585337e-01 -6.10671341e-01 -4.73288493e-03 6.06334090e-01
1.64101526e-01 -1.15193510e+00 2.12841555e-01 1.08552143e-01
-5.86575031e-01 -1.13131082e+00 -1.33079386e+00 -3.10594976e-01
-6.17990911e-01 -7.29590952e-01 -9.63237405e-01 -8.99819970e-01
9.43167567e-01 1.03557146e+00 9.93099988e-01 2.28356376e-01
1.61949918e-01 8.03943098e-01 -9.60754097e-01 3.05536985e-01
4.66849767e-02 -3.23573593e-03 3.33935232e-03 4.23563957e-01
3.40403855e-01 -2.04567537e-01 -9.48203206e-01 7.03249872e-01
-1.11113715e+00 1.32027492e-01 8.20903778e-01 9.51923728e-01
5.17377019e-01 -3.23394656e-01 4.87411201e-01 -1.00274050e+00
5.72290361e-01 -9.45192456e-01 -1.17973506e-01 5.18704712e-01
-5.07153928e-01 -2.55307823e-01 1.03631377e+00 -8.26732278e-01
-1.33656597e+00 4.76362323e-03 -2.29700699e-01 -6.15244210e-01
-2.14797676e-01 7.63627350e-01 -1.64256275e-01 1.81943893e-01
6.85497597e-02 4.98510242e-01 -1.76419556e-01 -7.74979115e-01
6.61082447e-01 7.21100330e-01 4.75080937e-01 -4.20626909e-01
5.01865208e-01 3.18517357e-01 -7.31968582e-01 -5.08204758e-01
-1.24058378e+00 -9.17741835e-01 -2.77379185e-01 -2.75189489e-01
6.39523089e-01 -9.68019128e-01 -7.72252977e-01 2.61630088e-01
-8.32668722e-01 -1.44208014e-01 1.29028678e-01 6.83068275e-01
-2.52739638e-01 6.92877054e-01 -8.25917721e-01 -5.98474920e-01
-3.92918110e-01 -1.21575522e+00 1.03646791e+00 1.74269795e-01
1.56690210e-01 -8.71856689e-01 -1.84612140e-01 7.26845264e-01
1.09901257e-01 -7.09306479e-01 9.05612826e-01 -6.47080600e-01
-3.55405957e-01 -2.92009152e-02 -5.42955995e-01 3.84461045e-01
1.27028152e-01 -3.89459789e-01 -5.13909400e-01 -4.87553716e-01
-2.80403882e-01 -5.73319137e-01 1.11623538e+00 2.66668022e-01
1.44368446e+00 -4.59701896e-01 -3.09577972e-01 5.10094047e-01
1.25402200e+00 3.73412579e-01 4.70496893e-01 1.49154723e-01
1.24044788e+00 5.64114034e-01 7.85089970e-01 6.78824663e-01
6.86290085e-01 8.02948356e-01 3.08400363e-01 -1.43513471e-01
3.16763669e-02 -4.75239187e-01 8.79266918e-01 1.02750158e+00
-9.95762721e-02 -6.29225194e-01 -1.34916633e-01 2.74651378e-01
-2.20153904e+00 -1.10002315e+00 2.82098856e-02 2.21164918e+00
6.39747620e-01 -3.96413743e-01 3.79719347e-01 -5.27315795e-01
7.04152048e-01 1.06581599e-01 -6.15883827e-01 -6.85502589e-02
-1.19558714e-01 -1.98493257e-01 3.27817827e-01 2.56945312e-01
-1.06929421e+00 8.40012908e-01 5.16329622e+00 1.01263404e+00
-1.09072971e+00 8.20459574e-02 4.73763585e-01 -3.55958730e-01
-7.39022017e-01 -5.15149891e-01 -8.39351714e-01 9.36429381e-01
5.71438491e-01 4.54242766e-01 7.28499234e-01 5.24057150e-01
2.93913603e-01 2.48678699e-01 -1.03930712e+00 1.13590670e+00
6.61692739e-01 -1.16923845e+00 3.59297544e-01 -1.49061516e-01
9.12802279e-01 -1.56527981e-01 5.52993059e-01 7.08151579e-01
3.47698122e-01 -7.57225335e-01 7.06615806e-01 6.32699013e-01
7.71836102e-01 -7.11790800e-01 7.26188779e-01 2.13577256e-01
-1.40032506e+00 -4.81307954e-01 -2.83825427e-01 5.80684006e-01
4.19590235e-01 1.49823785e-01 -2.44287774e-01 8.69099498e-01
7.16349602e-01 1.35171974e+00 -3.84511888e-01 6.70636177e-01
4.07376960e-02 6.38338566e-01 9.53962132e-02 2.66462769e-02
3.63872141e-01 -6.66295767e-01 4.13875520e-01 1.54626667e+00
4.39554870e-01 2.58133292e-01 5.46165943e-01 3.51827264e-01
-2.99925864e-01 4.20731515e-01 -1.58829808e-01 -3.09258729e-01
6.77865222e-02 1.25309932e+00 -3.05996090e-01 -3.81742626e-01
-1.06214249e+00 1.38338912e+00 4.10484374e-01 7.05006719e-01
-8.89416218e-01 1.33275241e-01 4.73960251e-01 -3.69152337e-01
8.50136459e-01 -1.52700037e-01 5.61835021e-02 -1.69838667e+00
-1.85545132e-01 -1.36212289e+00 7.16521919e-01 -5.52078724e-01
-1.71253455e+00 2.14627802e-01 -2.51218826e-01 -1.58490276e+00
1.84049293e-01 -5.79571724e-01 -3.07447970e-01 5.39965212e-01
-1.51679552e+00 -1.49770665e+00 -9.22434777e-02 1.06344628e+00
9.12232161e-01 -3.67404819e-01 5.40429175e-01 7.69788086e-01
-9.16208327e-01 1.01883364e+00 4.00438637e-01 -1.40953362e-01
9.15002763e-01 -9.47479129e-01 -1.68329135e-01 5.40826142e-01
2.75304407e-01 8.71864438e-01 3.04243475e-01 -5.51989615e-01
-1.99027514e+00 -1.10740519e+00 3.33490849e-01 -1.87151164e-01
3.92360061e-01 -2.15955436e-01 -7.55506575e-01 8.14961731e-01
2.42799178e-01 -3.55817199e-01 1.10823655e+00 2.11125448e-01
-5.58647692e-01 -3.99082899e-03 -5.71870923e-01 8.53685915e-01
1.02850580e+00 -4.79071587e-01 -4.72263336e-01 2.34131604e-01
7.49579847e-01 -1.80523574e-01 -7.21449256e-01 3.26730430e-01
8.61632526e-01 -5.93654096e-01 1.19681609e+00 -8.15345883e-01
6.98129535e-01 -1.19470209e-01 -3.91236633e-01 -1.38555980e+00
-5.99732101e-01 -3.12295169e-01 -3.80083501e-01 1.17144513e+00
5.39260566e-01 -5.21620326e-02 5.13269365e-01 2.74711102e-01
-3.09486359e-01 -6.27637923e-01 -1.20870702e-01 -3.34546745e-01
-3.96461606e-01 -4.54793990e-01 4.40181822e-01 1.07806015e+00
2.57438064e-01 6.36017561e-01 -1.17422950e+00 1.14770621e-01
3.68006289e-01 6.45016968e-01 6.28382623e-01 -5.34732163e-01
-5.66185534e-01 -3.32784802e-01 3.94964039e-01 -2.00533795e+00
2.68140405e-01 -8.75098169e-01 1.61639556e-01 -1.43092906e+00
6.35315001e-01 -2.47289211e-01 -7.74455905e-01 3.91323566e-01
-4.59935993e-01 4.22093332e-01 1.45190060e-01 3.69255155e-01
-1.22573185e+00 7.14895487e-01 1.71459413e+00 -1.56358153e-01
-2.70229340e-01 1.58587888e-01 -8.17959249e-01 4.97737765e-01
2.65723437e-01 -7.01589463e-03 -6.33570790e-01 -6.30374789e-01
3.41685891e-01 9.08055380e-02 -2.91097593e-02 -1.26775742e-01
1.33763999e-01 -3.90407711e-01 4.49233443e-01 -8.36508989e-01
2.83529252e-01 -8.88522565e-01 -2.71885633e-01 -1.39782563e-01
-7.93636322e-01 -2.52972860e-02 -5.60010560e-02 8.39857757e-01
-1.63113713e-01 -1.37296826e-01 4.22915787e-01 -2.09006384e-01
-1.09833872e+00 7.08668590e-01 -3.14248621e-01 -4.22694325e-01
6.99761450e-01 7.13563487e-02 -2.48529524e-01 -5.58489740e-01
-8.49443197e-01 6.08670592e-01 1.77274078e-01 7.74390399e-01
1.00870526e+00 -1.54375553e+00 -4.28660989e-01 7.36000463e-02
3.64554942e-01 -5.25204957e-01 9.06944036e-01 1.02438056e+00
1.74824476e-01 3.51479530e-01 -2.11702622e-02 -3.45456153e-01
-1.36013126e+00 9.49976861e-01 3.13310176e-02 -1.65631995e-01
-4.77107435e-01 9.89680350e-01 5.50413609e-01 -4.78275299e-01
4.93191659e-01 3.32120582e-02 -6.52665198e-01 2.88975596e-01
7.47574031e-01 1.40181527e-01 -2.93419093e-01 -7.55942464e-01
2.36974303e-02 4.67648745e-01 -6.86567366e-01 -8.24963748e-02
1.24277902e+00 -7.57924914e-01 5.34412451e-02 4.08212990e-01
1.32790017e+00 5.38951717e-03 -1.13382578e+00 -7.14288235e-01
-4.33880389e-01 -7.10942209e-01 1.66270137e-02 -8.50689232e-01
-1.31688893e+00 5.98438740e-01 5.23569643e-01 -3.92960459e-02
1.22197974e+00 -6.66895211e-02 1.35775757e+00 4.48768765e-01
5.06654792e-02 -1.09552550e+00 5.06212056e-01 4.07588899e-01
7.68588364e-01 -1.52814996e+00 -2.72279888e-01 -2.04248458e-01
-1.29353583e+00 9.54905987e-01 8.16616416e-01 1.80054419e-02
4.95035172e-01 -3.04213881e-01 1.94549933e-03 9.45995450e-02
-7.19092548e-01 -2.41083547e-01 8.99398208e-01 2.05527857e-01
5.51469445e-01 -7.53488988e-02 -3.71299356e-01 1.25480831e+00
5.13330817e-01 2.90714763e-02 -5.96455373e-02 7.25595474e-01
-2.71461457e-01 -1.27975750e+00 -1.04053199e-01 5.89029729e-01
-2.86909699e-01 -3.06590468e-01 -1.97682738e-01 1.77374348e-01
2.72241920e-01 1.00052893e+00 -1.33297697e-01 -8.72418642e-01
3.66894513e-01 -3.23490620e-01 4.32874948e-01 -6.24591410e-01
-6.30608857e-01 4.70953643e-01 6.84728846e-02 -6.02409244e-01
-5.53562641e-01 -6.14913821e-01 -1.00390399e+00 -1.31116346e-01
-3.12839925e-01 2.07254186e-01 3.41977984e-01 1.20583665e+00
3.12114537e-01 3.05221200e-01 1.08714759e+00 -1.08030796e+00
-4.60625261e-01 -7.61320770e-01 -5.38616121e-01 9.02813554e-01
3.76271695e-01 -5.99356234e-01 -7.72877559e-02 2.87910789e-01]
|
[10.177943229675293, 5.593347549438477]
|
c4592ada-776a-454e-b0b0-36b306664570
|
graph-convolution-for-semi-supervised
|
2102.06966
| null |
https://arxiv.org/abs/2102.06966v4
|
https://arxiv.org/pdf/2102.06966v4.pdf
|
Graph Convolution for Semi-Supervised Classification: Improved Linear Separability and Out-of-Distribution Generalization
|
Recently there has been increased interest in semi-supervised classification in the presence of graphical information. A new class of learning models has emerged that relies, at its most basic level, on classifying the data after first applying a graph convolution. To understand the merits of this approach, we study the classification of a mixture of Gaussians, where the data corresponds to the node attributes of a stochastic block model. We show that graph convolution extends the regime in which the data is linearly separable by a factor of roughly $1/\sqrt{D}$, where $D$ is the expected degree of a node, as compared to the mixture model data on its own. Furthermore, we find that the linear classifier obtained by minimizing the cross-entropy loss after the graph convolution generalizes to out-of-distribution data where the unseen data can have different intra- and inter-class edge probabilities from the training data.
|
['Aukosh Jagannath', 'Kimon Fountoulakis', 'Aseem Baranwal']
|
2021-02-13
| null | null | null | null |
['stochastic-block-model']
|
['graphs']
|
[ 2.80713171e-01 4.94832873e-01 -4.26699594e-02 -5.06386995e-01
-3.88431787e-01 -5.39766967e-01 6.74440324e-01 4.89538491e-01
-3.17654669e-01 4.93452579e-01 -3.56789947e-01 -5.63821077e-01
-2.89743632e-01 -1.01078534e+00 -6.72796369e-01 -1.16196799e+00
-4.95937288e-01 7.04092503e-01 1.17159434e-01 3.31855536e-01
-7.16797411e-02 6.31553054e-01 -1.32773674e+00 -1.14874065e-01
7.79840171e-01 9.33847606e-01 -1.77354842e-01 8.15584838e-01
-1.26207069e-01 4.01759595e-01 -4.27864432e-01 -5.99558890e-01
1.86994940e-01 -3.04096073e-01 -7.75693893e-01 1.94110602e-01
4.47379500e-01 3.54521126e-01 -7.17898965e-01 1.44275963e+00
8.26236606e-02 5.54619767e-02 1.27718568e+00 -1.53857279e+00
-4.30627227e-01 4.72938329e-01 -5.60899913e-01 6.02185130e-02
-1.54267982e-01 -3.30711126e-01 1.09808743e+00 -3.03948551e-01
3.20995390e-01 1.14617455e+00 5.73968589e-01 2.86564738e-01
-1.69209313e+00 -5.56663334e-01 1.09996267e-01 -1.15319848e-01
-1.55565250e+00 6.37506098e-02 7.78840601e-01 -7.19129086e-01
6.46229267e-01 1.71812415e-01 2.91083395e-01 8.17302823e-01
2.47745693e-01 7.12955773e-01 9.96720433e-01 -5.87420166e-01
3.89116436e-01 2.92614996e-01 6.15150988e-01 8.70762944e-01
4.68517661e-01 2.22369432e-02 8.13868120e-02 -4.00970459e-01
4.34729993e-01 -3.22617334e-03 -1.70879990e-01 -8.46007466e-01
-4.97779489e-01 9.44097459e-01 2.91060537e-01 2.51708001e-01
5.79076000e-02 1.47152528e-01 1.74945578e-01 2.08004102e-01
7.18691289e-01 -1.04120284e-01 -3.55360955e-01 2.91224808e-01
-9.64096546e-01 -1.87606022e-01 1.07425177e+00 8.85332286e-01
9.16230142e-01 -2.27949053e-01 3.44719470e-01 5.86611807e-01
4.43095684e-01 3.43298405e-01 5.82904592e-02 -4.06943440e-01
4.21982378e-01 6.55046582e-01 -2.34673783e-01 -5.50408602e-01
-3.76037508e-01 -7.60651469e-01 -1.13688517e+00 4.75792885e-01
7.40565777e-01 -2.72777736e-01 -1.11374938e+00 1.92869592e+00
1.11096926e-01 -2.32825056e-02 8.74604955e-02 1.68508619e-01
4.66773689e-01 3.68551135e-01 3.79055142e-02 -3.47871594e-02
8.58514547e-01 -5.66771805e-01 -3.31866771e-01 -8.74804556e-02
7.22152650e-01 -3.53324771e-01 6.00089014e-01 3.18976372e-01
-6.96852207e-01 -4.33480024e-01 -1.05799580e+00 3.90933067e-01
-5.02113283e-01 -1.37627795e-01 5.35363376e-01 1.04193103e+00
-1.21354389e+00 8.37130666e-01 -8.78328085e-01 -3.35689574e-01
4.56534386e-01 6.55914724e-01 -5.26316404e-01 -3.26727539e-01
-9.05621350e-01 6.60459459e-01 3.42738539e-01 -1.25000507e-01
-5.59311092e-01 -3.88444155e-01 -8.97581875e-01 2.11025149e-01
1.68073833e-01 -4.09740090e-01 7.48143613e-01 -9.44229484e-01
-8.44057083e-01 8.21632683e-01 -4.15098257e-02 -3.27582121e-01
3.06156784e-01 5.37388384e-01 -4.74029779e-01 -8.91888607e-03
-1.17618322e-01 2.04882771e-01 9.20561612e-01 -1.23995006e+00
-4.90745425e-01 -7.08552182e-01 -2.95911115e-02 -4.14404497e-02
-1.13503225e-01 -3.05104703e-01 -2.96549827e-01 -4.00043011e-01
1.02720849e-01 -1.18030655e+00 -3.13929349e-01 -1.27468020e-01
-5.84344089e-01 -2.74228901e-01 7.58940637e-01 -3.62527788e-01
1.02325940e+00 -2.46493077e+00 1.08425185e-01 6.93294048e-01
6.12444520e-01 7.45632872e-02 1.53391391e-01 5.39300203e-01
-5.40445209e-01 2.50562429e-01 -4.68165219e-01 -3.93293560e-01
6.65401444e-02 2.37304255e-01 1.87491789e-01 7.87472427e-01
1.18166313e-01 4.62100685e-01 -6.25268817e-01 -3.70384216e-01
-1.06108814e-01 4.50241655e-01 -2.05021217e-01 -1.03808925e-01
4.07983996e-02 1.60373282e-02 -3.83309931e-01 3.16610634e-01
9.55020905e-01 -7.68613160e-01 3.40869397e-01 1.52542755e-01
3.97553772e-01 4.37834719e-03 -1.30103052e+00 1.16929567e+00
-1.02694318e-01 7.78011680e-01 2.97857016e-01 -1.41946161e+00
5.82356572e-01 2.17673883e-01 3.88559371e-01 9.89778154e-03
3.14896405e-01 -4.76164110e-02 3.86964083e-01 -4.31173667e-02
-7.18167946e-02 -3.04242074e-01 -2.44761948e-02 4.34795648e-01
4.18470979e-01 2.58536518e-01 3.03492785e-01 3.86950135e-01
1.36345732e+00 -3.62593234e-01 2.23332047e-01 -4.46615338e-01
1.36632785e-01 -4.14052278e-01 1.84574604e-01 8.48037124e-01
-4.09968533e-02 4.12730992e-01 9.81345534e-01 3.15692909e-02
-7.79062212e-01 -1.29956520e+00 -3.56413096e-01 5.33601344e-01
6.01784475e-02 -2.12587908e-01 -9.75938559e-01 -1.02868187e+00
1.92930177e-01 6.63204610e-01 -8.92486930e-01 -3.24661553e-01
5.29378280e-03 -1.32915950e+00 3.55612069e-01 5.02159834e-01
2.61438876e-01 -5.43935180e-01 1.06876642e-01 -1.20181210e-01
2.15595156e-01 -9.25954640e-01 -3.18029791e-01 7.58654356e-01
-8.87728453e-01 -1.23956978e+00 -5.37130892e-01 -7.28810728e-01
1.02808964e+00 -1.32321522e-01 9.71455812e-01 -3.20384018e-02
-3.21355641e-01 4.13844764e-01 -8.77154171e-02 -1.92032814e-01
-5.73754907e-01 1.30662948e-01 -1.22352809e-01 2.23328158e-01
4.44303453e-01 -7.00656056e-01 -2.63074696e-01 8.58750716e-02
-8.05771589e-01 -1.04068182e-01 6.81137383e-01 6.28433347e-01
2.24366993e-01 6.66553199e-01 1.45196632e-01 -1.06512690e+00
4.22527492e-01 -6.59731328e-01 -5.43779850e-01 3.62488836e-01
-7.79503763e-01 4.24143702e-01 4.05198723e-01 -4.04272348e-01
-6.51303589e-01 1.00211419e-01 1.07815862e-01 -1.68738365e-01
-4.53898907e-01 4.80693996e-01 -4.07560527e-01 -2.65972018e-01
4.62146044e-01 6.89661577e-02 2.27679312e-01 -5.46798289e-01
2.95911610e-01 7.75104105e-01 1.91911981e-01 -3.85326743e-01
7.41474271e-01 4.18489456e-01 5.54210544e-01 -1.12933528e+00
-4.34504986e-01 -3.46407145e-01 -7.98955321e-01 -9.69909653e-02
7.56289601e-01 -5.37321687e-01 -5.94440997e-01 6.73223317e-01
-8.13035131e-01 -5.06500721e-01 -2.95142949e-01 6.05259538e-01
-3.78745914e-01 5.40659249e-01 -6.36873782e-01 -1.04357338e+00
3.01301897e-01 -9.76134896e-01 6.27383769e-01 1.28024831e-01
7.87343532e-02 -1.45829141e+00 -6.53462484e-02 -8.12293380e-04
-3.60012762e-02 2.56023258e-01 1.45715296e+00 -1.12058651e+00
-3.34894508e-01 -8.27460468e-01 -2.45650530e-01 7.12333322e-01
1.26742437e-01 6.32046610e-02 -7.76742458e-01 -3.79798591e-01
-8.41963887e-02 7.23340735e-02 8.94710243e-01 4.39839959e-01
1.14252257e+00 -1.09042928e-01 -8.07476461e-01 3.45259994e-01
1.45491612e+00 1.31037876e-01 3.51204365e-01 -4.71460283e-01
6.14906490e-01 4.82933611e-01 -1.12107284e-01 -1.08828582e-02
8.61009881e-02 4.03596163e-01 4.06248361e-01 -2.70423681e-01
9.84018892e-02 -1.70657098e-01 2.25867368e-02 3.47218037e-01
1.77288949e-01 -6.96945429e-01 -9.11980391e-01 2.96959907e-01
-1.66065609e+00 -7.20032334e-01 -3.48215431e-01 2.59699345e+00
3.36599737e-01 4.99248266e-01 2.08522692e-01 3.41453850e-01
1.03696787e+00 1.83425657e-02 -4.35227096e-01 -1.13360181e-01
-7.61439949e-02 4.82708603e-01 7.70654738e-01 8.15271676e-01
-1.27862680e+00 2.99241275e-01 6.68167353e+00 1.08947062e+00
-8.55909526e-01 -1.22601070e-01 8.18942189e-01 3.38096440e-01
-7.33424872e-02 6.65378422e-02 -7.33955979e-01 7.25079536e-01
1.06302738e+00 2.90943459e-02 1.99853450e-01 7.43178666e-01
-3.66190553e-01 -3.15103531e-01 -1.36091852e+00 6.05215192e-01
1.40351295e-01 -8.29422891e-01 -1.97272390e-01 7.26302028e-01
5.34049749e-01 5.28517850e-02 1.68436468e-01 1.92038596e-01
6.99003458e-01 -1.04801500e+00 3.03222656e-01 3.71267527e-01
7.23364592e-01 -6.54673994e-01 6.24322236e-01 7.71485806e-01
-1.10121191e+00 7.48811439e-02 -2.21850686e-02 -6.22292086e-02
-1.63898468e-01 6.82136893e-01 -7.53774226e-01 4.95100141e-01
3.56899321e-01 3.47344637e-01 -6.80931807e-01 1.11924350e+00
1.36182696e-01 6.03347600e-01 -6.37908876e-01 -6.57009110e-02
9.16670486e-02 -4.60571378e-01 3.97318900e-01 1.13617849e+00
2.44344160e-01 -3.10045816e-02 1.03306435e-01 6.21584833e-01
-1.24605395e-01 -1.17394090e-01 -5.72760522e-01 -3.13629180e-01
-1.66225818e-03 1.17266405e+00 -1.23024070e+00 -2.82171965e-01
-6.71635866e-01 8.65359902e-01 5.67353725e-01 5.07361114e-01
-4.77397293e-01 -5.06528556e-01 3.56462538e-01 3.38534087e-01
5.36953986e-01 -3.31604958e-01 -2.31470596e-02 -9.29063439e-01
-1.20227657e-01 -2.72804826e-01 5.04604220e-01 -3.67888153e-01
-1.48719215e+00 5.32719314e-01 2.75085449e-01 -8.01563501e-01
-1.50887743e-01 -1.05690038e+00 -5.60452640e-01 1.02162516e+00
-9.59977984e-01 -9.55632746e-01 6.60524517e-02 2.54617035e-01
-1.83411390e-01 -6.19210973e-02 9.09659326e-01 2.16653779e-01
-4.60616648e-01 5.81278861e-01 5.53504944e-01 4.95636314e-01
2.47163191e-01 -1.53480220e+00 3.28669995e-01 7.54184186e-01
3.85804713e-01 4.47737783e-01 8.35294306e-01 -5.47431827e-01
-8.12391758e-01 -8.53361130e-01 8.37947369e-01 -3.54822278e-01
6.87675178e-01 -7.49715447e-01 -9.59851623e-01 6.74097717e-01
5.21656265e-03 4.61809069e-01 1.02168369e+00 2.21651584e-01
-3.66283923e-01 1.11097686e-01 -1.42675221e+00 3.08189332e-01
8.78322721e-01 -5.88252127e-01 -1.82530805e-01 3.05092037e-01
1.14798740e-01 2.09903941e-02 -6.51534319e-01 3.48482132e-01
3.23270380e-01 -9.63734448e-01 5.94141066e-01 -9.14477229e-01
2.68179607e-02 -5.26410230e-02 -3.29943687e-01 -1.34388554e+00
-3.15123200e-01 -3.57996672e-01 -4.42458391e-02 9.99356329e-01
6.90561891e-01 -6.12924278e-01 1.15984344e+00 6.94067478e-01
2.56869256e-01 -7.73798943e-01 -9.75187480e-01 -8.02680254e-01
2.23496705e-01 -3.60392064e-01 1.10841937e-01 6.31169856e-01
7.50642195e-02 4.28582340e-01 1.06437758e-01 3.29433471e-01
9.84569967e-01 5.53969517e-02 4.01107132e-01 -1.71311104e+00
-5.92959583e-01 -5.10883629e-01 -8.34916234e-01 -1.09903872e+00
4.09958959e-01 -1.25761783e+00 -2.15110570e-01 -1.29959202e+00
3.73813033e-01 -5.40751994e-01 -3.39942843e-01 9.37505439e-02
-3.92192677e-02 1.70934066e-01 -1.79386467e-01 -1.67260885e-01
-4.07214612e-01 1.52104706e-01 7.41354465e-01 -1.33452818e-01
1.44965142e-01 5.42175055e-01 -6.78979933e-01 7.99269736e-01
5.36262810e-01 -5.57948411e-01 -2.97499985e-01 1.32188603e-01
1.44866526e-01 1.45629689e-01 4.28421110e-01 -9.58679676e-01
1.16465122e-01 1.60965368e-01 2.95500875e-01 -3.86553496e-01
3.73692244e-01 -9.83648241e-01 2.88532555e-01 2.49634326e-01
-3.20237160e-01 -2.54597157e-01 -6.19276762e-02 1.08324087e+00
-2.94351578e-02 -5.67189813e-01 8.82815301e-01 -2.91397516e-02
6.17775731e-02 5.35352468e-01 -4.83791530e-01 8.74225199e-02
1.01784337e+00 -1.23083293e-01 -1.61528841e-01 -5.65248549e-01
-1.16663337e+00 -6.68859780e-02 4.19150352e-01 -1.47256896e-01
4.46617939e-02 -1.04826164e+00 -3.95289749e-01 3.65614891e-01
8.02658796e-02 -1.09188788e-01 1.95817754e-01 7.99319029e-01
-1.82220742e-01 4.27722000e-02 2.61217982e-01 -4.59096372e-01
-1.41644490e+00 7.59027660e-01 4.61280793e-01 -3.08582693e-01
-2.40813300e-01 9.06641304e-01 2.97804534e-01 -9.51414704e-02
2.39665389e-01 -1.22953303e-01 2.18449697e-01 -2.33071367e-03
4.68539679e-03 4.41368967e-01 1.96062163e-01 -6.63706064e-01
-1.58020020e-01 1.93267971e-01 -1.74972445e-01 -1.49908081e-01
1.04646063e+00 4.80780527e-02 -1.24035500e-01 7.04358578e-01
1.53928602e+00 6.71266690e-02 -1.26918352e+00 -3.85132790e-01
-1.27681820e-02 -2.28621081e-01 -5.29161654e-04 -5.31373203e-01
-9.52086747e-01 1.09712410e+00 5.14964223e-01 1.04495203e+00
6.33323848e-01 4.62760270e-01 -2.63399128e-02 4.51517925e-02
1.16137467e-01 -8.11892927e-01 -2.92825073e-01 2.13274524e-01
2.66690850e-01 -1.07807052e+00 -1.56241134e-01 -7.40192473e-01
-1.32042795e-01 1.08858097e+00 1.06669113e-01 -3.97067100e-01
1.37818444e+00 5.55359364e-01 -3.92398894e-01 -2.32214808e-01
-3.82909656e-01 -1.68844163e-01 3.41784924e-01 8.40109527e-01
3.16641361e-01 2.64876336e-01 6.41137362e-02 5.44524968e-01
-8.57931152e-02 -3.89967352e-01 3.39569330e-01 7.46288240e-01
-2.93117017e-01 -1.26326215e+00 3.55721489e-02 9.27749276e-01
-4.36811626e-01 -1.05873153e-01 -4.83834088e-01 9.49671924e-01
7.28977695e-02 7.28181720e-01 2.39309564e-01 -3.09688240e-01
3.46332509e-03 5.43259680e-01 6.46158099e-01 -6.54828787e-01
7.32619092e-02 -5.63349128e-02 -1.70450330e-01 7.18913451e-02
-1.94240347e-01 -7.99639583e-01 -1.06571174e+00 -2.40393728e-01
-5.61228096e-01 4.02037501e-01 5.67780256e-01 1.10797644e+00
1.69790983e-02 3.10995966e-01 5.35452247e-01 -6.58108771e-01
-5.00505030e-01 -9.94207919e-01 -1.17731392e+00 9.49381739e-02
3.75107259e-01 -6.37174964e-01 -9.87748981e-01 -1.73749372e-01]
|
[6.948781967163086, 5.6425089836120605]
|
5aa463cf-5668-4594-8860-58bebc7f4139
|
active-fine-tuning-from-gmad-examples
|
2003.03849
| null |
https://arxiv.org/abs/2003.03849v2
|
https://arxiv.org/pdf/2003.03849v2.pdf
|
Active Fine-Tuning from gMAD Examples Improves Blind Image Quality Assessment
|
The research in image quality assessment (IQA) has a long history, and significant progress has been made by leveraging recent advances in deep neural networks (DNNs). Despite high correlation numbers on existing IQA datasets, DNN-based models may be easily falsified in the group maximum differentiation (gMAD) competition with strong counterexamples being identified. Here we show that gMAD examples can be used to improve blind IQA (BIQA) methods. Specifically, we first pre-train a DNN-based BIQA model using multiple noisy annotators, and fine-tune it on multiple subject-rated databases of synthetically distorted images, resulting in a top-performing baseline model. We then seek pairs of images by comparing the baseline model with a set of full-reference IQA methods in gMAD. The resulting gMAD examples are most likely to reveal the relative weaknesses of the baseline, and suggest potential ways for refinement. We query ground truth quality annotations for the selected images in a well controlled laboratory environment, and further fine-tune the baseline on the combination of human-rated images from gMAD and existing databases. This process may be iterated, enabling active and progressive fine-tuning from gMAD examples for BIQA. We demonstrate the feasibility of our active learning scheme on a large-scale unlabeled image set, and show that the fine-tuned method achieves improved generalizability in gMAD, without destroying performance on previously trained databases.
|
['Kede Ma', 'Zhihua Wang']
|
2020-03-08
| null | null | null | null |
['blind-image-quality-assessment']
|
['computer-vision']
|
[ 2.70546854e-01 2.67350413e-02 -1.33251818e-02 -4.88624185e-01
-1.70954096e+00 -6.87525809e-01 3.32910091e-01 -1.18705772e-01
-5.45111299e-01 6.80331349e-01 5.61064184e-01 -1.74564108e-01
-3.80614191e-01 -4.00797904e-01 -5.27330637e-01 -7.28909016e-01
-2.14809760e-01 6.19328082e-01 1.99186485e-02 -3.04681640e-02
1.70996264e-01 4.49478924e-01 -1.23241961e+00 4.99466985e-01
1.03593850e+00 9.77118015e-01 -1.66648611e-01 6.96697116e-01
3.60563725e-01 7.09729016e-01 -9.71388578e-01 -8.85864079e-01
6.97172701e-01 -3.77433389e-01 -1.02286351e+00 3.19057882e-01
1.05206800e+00 -8.49829495e-01 -2.52429307e-01 1.04546189e+00
1.09008622e+00 8.96836668e-02 5.13207018e-01 -1.24181139e+00
-9.19017911e-01 1.60728231e-01 -3.45489442e-01 4.70685273e-01
2.99446791e-01 8.20706367e-01 1.08687305e+00 -9.40839350e-01
4.17318463e-01 1.28338623e+00 7.82537639e-01 7.22031415e-01
-1.53933084e+00 -9.06727016e-01 -1.33659467e-01 4.58208859e-01
-1.11352956e+00 -9.01659966e-01 3.93247753e-01 -4.15853441e-01
1.02041519e+00 -2.87731234e-02 5.82736373e-01 1.28572440e+00
-3.07499349e-01 8.84790361e-01 1.26282227e+00 -2.72611350e-01
4.41207409e-01 -2.82478034e-01 -1.89661443e-01 6.61050379e-01
1.60028204e-01 3.68899852e-01 -7.74181604e-01 -4.69637126e-01
7.43449092e-01 -5.76873064e-01 -5.49357414e-01 -6.01164341e-01
-1.42486989e+00 7.57536232e-01 6.05680406e-01 -2.20457122e-01
-3.63618314e-01 -9.63787660e-02 2.12603450e-01 5.60663640e-01
4.86875176e-01 1.11431265e+00 -2.79961109e-01 6.19537346e-02
-1.46517670e+00 3.15766215e-01 4.30424005e-01 5.90568304e-01
7.13187337e-01 -1.05254225e-01 -4.16424245e-01 9.66625333e-01
2.45361373e-01 4.04552072e-01 4.51954991e-01 -1.62721467e+00
4.40837473e-01 4.95587379e-01 4.27526563e-01 -8.59277248e-01
-9.56080705e-02 -1.38912022e-01 -5.93666852e-01 9.17481303e-01
6.76414967e-01 2.73313165e-01 -1.19707549e+00 1.49302900e+00
-3.95041890e-02 -7.38667250e-02 -1.27044722e-01 1.14956236e+00
6.87485814e-01 4.51390147e-01 2.12714642e-01 -3.88142169e-01
8.24747503e-01 -9.30714726e-01 -5.31950533e-01 -2.07046773e-02
3.49770069e-01 -6.59360230e-01 1.36950684e+00 8.99549901e-01
-1.28828192e+00 -5.84945858e-01 -1.26779211e+00 -3.34696211e-02
1.88104622e-02 -1.33247823e-01 2.80601025e-01 6.14902794e-01
-1.55983186e+00 5.17441034e-01 -7.58721292e-01 -1.21132217e-01
1.00424147e+00 5.60757875e-01 -4.24003989e-01 -3.31634611e-01
-1.11738849e+00 8.79709661e-01 7.71794319e-02 1.63147390e-01
-1.71010017e+00 -7.10450828e-01 -6.54608786e-01 -2.75234580e-01
3.45726699e-01 -8.79680157e-01 1.35772598e+00 -1.33486748e+00
-1.26004589e+00 1.13923442e+00 1.19141310e-01 -4.60019588e-01
6.67461336e-01 -1.61198914e-01 -6.38409317e-01 6.58767760e-01
2.10234225e-01 1.09717774e+00 9.69302416e-01 -1.39388394e+00
-3.36530000e-01 -2.04494357e-01 4.23588693e-01 3.14078540e-01
-2.50185132e-01 2.73836344e-01 -2.63302088e-01 -7.00130403e-01
-1.25466138e-01 -7.16613054e-01 -3.02470088e-01 5.92171133e-01
-2.69321650e-01 -4.52401936e-02 3.14073414e-01 -6.25476539e-01
8.97610962e-01 -2.12110853e+00 -1.61082983e-01 6.58002794e-02
5.80078959e-01 6.14462256e-01 -6.64672554e-01 -5.13782762e-02
-1.88060433e-01 1.94626912e-01 -4.84155387e-01 -4.00334805e-01
-1.30263358e-01 -5.28448308e-03 -1.96683392e-01 4.57565099e-01
6.68667316e-01 9.57652271e-01 -1.10360968e+00 -6.72998369e-01
-1.13222532e-01 1.25943273e-01 -6.79408908e-01 4.77471411e-01
-1.59885406e-01 4.69600677e-01 1.60865039e-01 7.57801831e-01
5.59768617e-01 -5.14518678e-01 -1.36655748e-01 -4.70340937e-01
2.54508495e-01 3.10760647e-01 -1.02870023e+00 1.71884942e+00
-5.88637069e-02 6.23373151e-01 -3.26391123e-02 -9.51588690e-01
6.63035691e-01 4.86705571e-01 4.72752869e-01 -1.06303561e+00
-1.86202168e-01 1.81615874e-01 1.93591103e-01 -5.86629152e-01
2.96625674e-01 -6.52218312e-02 3.44421327e-01 6.74524009e-01
5.84312022e-01 -1.83298457e-02 2.40856081e-01 3.89071286e-01
1.20386839e+00 -1.57032292e-02 1.84486791e-01 -1.20919399e-01
1.91585481e-01 2.57032186e-01 5.03649354e-01 9.32294309e-01
-1.07313919e+00 1.12933028e+00 2.08130583e-01 -3.21682900e-01
-1.44867480e+00 -1.37658405e+00 -1.17171310e-01 9.64784145e-01
-8.43143743e-03 -1.86675355e-01 -6.22419119e-01 -1.01252007e+00
-2.65207171e-01 2.37791732e-01 -7.32257962e-01 -1.54140487e-01
-3.57682139e-01 -1.03194642e+00 7.20604479e-01 4.28434908e-01
6.63142502e-01 -1.08980608e+00 -9.73741338e-02 4.92079705e-02
-3.42810631e-01 -8.03588986e-01 -2.16068402e-01 -2.45809881e-03
-7.13253021e-01 -1.18868101e+00 -1.04134154e+00 -6.33484066e-01
5.17847240e-01 2.06375763e-01 1.72651911e+00 2.15714008e-01
-3.19620110e-02 3.76629949e-01 -2.04756185e-01 -1.14067696e-01
-6.40066564e-01 -4.80922908e-01 2.22433582e-01 -1.65411666e-01
2.53490448e-01 -3.14449310e-01 -9.71292138e-01 5.02452970e-01
-9.30724680e-01 -1.43814832e-01 6.09432340e-01 1.05080235e+00
6.93605661e-01 -8.84946510e-02 8.26212227e-01 -6.34689212e-01
7.64149070e-01 -2.08528116e-01 -6.76205099e-01 4.02295083e-01
-8.78860235e-01 -1.24889284e-01 2.53265321e-01 -5.22051990e-01
-1.01348293e+00 -1.46196899e-03 -3.74520943e-02 -4.63790387e-01
-8.51088986e-02 2.38322213e-01 -4.75374162e-01 -2.65779048e-01
1.30132794e+00 -2.59908468e-01 -1.00039713e-01 -5.72373010e-02
5.26224256e-01 5.57340562e-01 8.95109832e-01 -5.96497536e-01
7.99434841e-01 4.26870316e-01 -4.39012378e-01 3.41545534e-03
-1.05911136e+00 -3.19526970e-01 -4.02158976e-01 -2.23093092e-01
7.10741282e-01 -1.10210645e+00 -3.34540665e-01 5.05223691e-01
-8.28940153e-01 -6.36562228e-01 -3.67690891e-01 4.88829792e-01
-8.24300706e-01 3.81809831e-01 -6.99629784e-01 -5.05538166e-01
-3.97908866e-01 -1.52046514e+00 1.01643527e+00 -4.71228473e-02
-3.93539220e-01 -7.45810747e-01 3.90500665e-01 9.36106741e-01
2.82661140e-01 -2.06061825e-01 8.82711351e-01 -6.87183678e-01
-7.99727023e-01 1.28214508e-01 -3.81286293e-01 7.71436810e-01
-4.17378582e-02 -1.36153683e-01 -1.24507523e+00 -4.86566007e-01
-1.44352093e-01 -1.03372753e+00 7.79023588e-01 2.93784022e-01
9.86096799e-01 -3.28209132e-01 2.82185912e-01 5.33963084e-01
1.34541714e+00 1.40267298e-01 8.57005477e-01 5.46012938e-01
3.70473146e-01 2.93320179e-01 5.58858216e-01 -1.14263155e-01
1.19717993e-01 4.79544431e-01 5.04966915e-01 -3.74092996e-01
-5.43478727e-01 1.03812851e-01 3.83759111e-01 7.55685985e-01
2.59671453e-03 -2.98731148e-01 -1.09216821e+00 8.43245745e-01
-1.40156162e+00 -9.98336852e-01 2.60059208e-01 2.24006844e+00
1.16765845e+00 2.10343167e-01 3.12529951e-01 3.43605608e-01
6.07871830e-01 -5.90340830e-02 -7.46324778e-01 -1.96782649e-02
-4.17880356e-01 2.80035764e-01 2.29018748e-01 3.81399065e-01
-1.17790985e+00 4.76640701e-01 7.73080969e+00 6.12296522e-01
-7.83320665e-01 1.46748722e-01 1.01562619e+00 -3.32295656e-01
-1.43551320e-01 -1.34281695e-01 -2.47282594e-01 2.49567792e-01
9.58391309e-01 -1.30964905e-01 5.33318341e-01 5.61983347e-01
2.89467812e-01 5.90299396e-03 -1.34641385e+00 1.15394580e+00
2.21662208e-01 -1.25932992e+00 2.38068357e-01 1.26679353e-02
1.12504184e+00 1.82128936e-01 2.91787267e-01 2.21867144e-01
7.16951668e-01 -1.24297488e+00 5.27451932e-01 4.50391024e-01
9.88062322e-01 -4.90543276e-01 7.42436290e-01 -4.80564609e-02
-3.04141909e-01 -2.73442894e-01 -4.17582422e-01 2.66508818e-01
1.93064101e-02 5.60125113e-01 -9.86745238e-01 3.35617006e-01
1.06866968e+00 3.66364330e-01 -9.58056331e-01 1.49738014e+00
-1.21865712e-01 9.62588549e-01 6.83970600e-02 5.33281803e-01
-5.75970933e-02 1.20902844e-01 5.93160212e-01 8.12963843e-01
6.54770061e-02 1.41067088e-01 -1.95232853e-01 1.00261378e+00
-3.48318577e-01 -1.72069311e-01 -2.10805476e-01 -1.89509198e-01
2.47931629e-01 1.16295516e+00 -4.24323082e-01 -6.39225066e-01
-4.97836679e-01 1.03668499e+00 4.95842159e-01 5.12241483e-01
-5.03804326e-01 -4.83000204e-02 5.23577452e-01 -1.04484029e-01
9.00851861e-02 6.44696131e-02 -1.55674145e-01 -1.25184143e+00
-1.97578669e-01 -1.69198012e+00 4.76576984e-01 -1.34652352e+00
-1.82949638e+00 7.10124016e-01 -1.58431619e-01 -1.53967535e+00
-2.66058415e-01 -5.10273993e-01 -1.33976817e-01 1.06726694e+00
-1.25964487e+00 -7.27492690e-01 -4.46890503e-01 7.86113799e-01
4.04819757e-01 -3.00376594e-01 9.41043973e-01 6.00924194e-01
-3.33474666e-01 8.56178522e-01 -1.08642898e-01 3.96579266e-01
1.25763297e+00 -1.29367805e+00 4.89819646e-01 1.16250503e+00
3.44276577e-01 5.96475482e-01 5.71039975e-01 -4.91615534e-01
-7.14483202e-01 -9.48393941e-01 4.04588878e-01 -7.50988364e-01
5.59112549e-01 -1.84718102e-01 -1.08243835e+00 4.22658414e-01
2.16051400e-01 1.37741506e-01 6.16825759e-01 2.48804204e-02
-6.53009772e-01 -1.14825591e-01 -1.41323304e+00 4.56957489e-01
1.08595073e+00 -8.65444899e-01 -6.83605969e-01 3.11493546e-01
6.22583032e-01 -2.01417983e-01 -7.94667363e-01 6.07981443e-01
3.26108277e-01 -1.19594312e+00 1.02265537e+00 -7.33724058e-01
4.75840002e-01 -4.65603173e-01 -4.28981520e-02 -1.54949737e+00
-5.77326238e-01 -4.13827896e-01 -2.40881648e-02 1.13970280e+00
4.98928636e-01 -6.20842502e-02 8.57197940e-01 5.91658354e-01
-2.50774801e-01 -5.53371906e-01 -7.94895470e-01 -7.17074215e-01
6.67167306e-02 -4.66171414e-01 4.91915852e-01 1.03280258e+00
-3.08066666e-01 1.99584991e-01 -3.33119571e-01 1.62770629e-01
8.74051988e-01 -2.62032300e-01 6.23683155e-01 -9.58914042e-01
-4.43944395e-01 -3.56310695e-01 -5.30353069e-01 -8.75229836e-01
-2.69209981e-01 -6.56378269e-01 3.22453499e-01 -1.63990963e+00
4.77534622e-01 -3.72778624e-01 -5.05949676e-01 5.69002926e-01
-4.59822118e-01 9.57613766e-01 -6.96840463e-03 5.50768554e-01
-9.63969707e-01 3.75957757e-01 1.51255333e+00 -5.53121388e-01
-7.43506302e-04 -4.16139543e-01 -7.33517110e-01 3.65720987e-01
5.62116504e-01 -3.86939108e-01 -5.82011282e-01 -5.76139271e-01
4.79086518e-01 -1.07221700e-01 6.07883275e-01 -1.15333462e+00
1.44667476e-02 5.11357673e-02 7.55237639e-01 -2.52247989e-01
2.83360124e-01 -5.51168799e-01 -1.27301840e-02 4.29860577e-02
-7.13686228e-01 9.09427330e-02 5.43993786e-02 4.24299389e-01
-2.78522342e-01 -5.81575409e-02 9.95963097e-01 -2.63533294e-01
-7.83950329e-01 4.91246223e-01 -1.10115416e-01 2.22421423e-01
3.31210375e-01 -3.20742637e-01 -6.17533326e-01 -5.81762135e-01
-8.48102987e-01 1.73569545e-02 6.11895561e-01 2.64143765e-01
5.79540312e-01 -1.43853223e+00 -8.40462208e-01 -2.34158207e-02
4.08811808e-01 5.91062047e-02 2.67782599e-01 7.00400770e-01
-5.03505707e-01 -6.54443130e-02 -4.51051861e-01 -6.94556534e-01
-9.91708636e-01 7.63632476e-01 7.07987368e-01 -1.81724146e-01
-2.70431548e-01 8.72260451e-01 3.76583546e-01 -3.42486471e-01
3.76266003e-01 5.91293834e-02 -1.00481533e-01 -7.85377845e-02
8.36768508e-01 2.81993955e-01 5.12356699e-01 -6.15153491e-01
-2.77947247e-01 2.61401329e-02 3.27622034e-02 -3.78378183e-01
1.26130915e+00 -2.88718015e-01 1.53520405e-01 1.60337850e-01
1.09384084e+00 5.39795570e-02 -1.47507071e+00 -2.34114200e-01
-8.30909237e-02 -4.82816696e-01 1.20741248e-01 -1.50875306e+00
-1.24261069e+00 7.31421053e-01 1.23239398e+00 -1.07209042e-01
1.42910779e+00 -5.74478023e-02 4.75546211e-01 4.06518608e-01
2.83987433e-01 -8.60829592e-01 6.87427819e-01 -6.60341457e-02
1.00057065e+00 -1.56136751e+00 1.69375166e-01 1.18496731e-01
-6.36982620e-01 8.04185748e-01 6.68816030e-01 -1.71980157e-01
1.25176847e-01 2.74262764e-02 7.54458785e-01 -2.63186663e-01
-7.39151001e-01 -2.78803352e-02 4.11907583e-01 9.80144858e-01
1.60438925e-01 -2.16845006e-01 1.56805381e-01 3.47152233e-01
-2.30965778e-01 2.75730103e-01 5.45052767e-01 4.43062693e-01
-1.52382001e-01 -9.24863636e-01 -5.20130038e-01 5.13942838e-01
-4.97413546e-01 -3.24407399e-01 -3.79704148e-01 5.77112019e-01
3.10452819e-01 1.17527890e+00 1.89158786e-02 -2.65698522e-01
2.96774536e-01 -1.25595659e-01 6.41547561e-01 -3.59649837e-01
-5.35469353e-01 -7.89217651e-02 3.04854959e-01 -9.12300110e-01
-7.82817483e-01 -4.94296640e-01 -7.15260148e-01 -1.30455256e-01
-3.16501141e-01 -5.41369691e-02 2.05478609e-01 7.79515266e-01
2.70508021e-01 8.99051428e-02 4.80260104e-01 -7.15435088e-01
-4.60032642e-01 -9.33066964e-01 -2.52663195e-01 8.97618890e-01
5.89622259e-01 -6.50069416e-01 -5.84092259e-01 2.81631052e-01]
|
[11.887497901916504, -1.8003774881362915]
|
b3137b8b-2e73-417d-a638-b65f125ebc42
|
towards-better-evaluation-for-dynamic-link
|
2207.10128
| null |
https://arxiv.org/abs/2207.10128v2
|
https://arxiv.org/pdf/2207.10128v2.pdf
|
Towards Better Evaluation for Dynamic Link Prediction
|
Despite the prevalence of recent success in learning from static graphs, learning from time-evolving graphs remains an open challenge. In this work, we design new, more stringent evaluation procedures for link prediction specific to dynamic graphs, which reflect real-world considerations, to better compare the strengths and weaknesses of methods. First, we create two visualization techniques to understand the reoccurring patterns of edges over time and show that many edges reoccur at later time steps. Based on this observation, we propose a pure memorization baseline called EdgeBank. EdgeBank achieves surprisingly strong performance across multiple settings because easy negative edges are often used in the current evaluation setting. To evaluate against more difficult negative edges, we introduce two more challenging negative sampling strategies that improve robustness and better match real-world applications. Lastly, we introduce six new dynamic graph datasets from a diverse set of domains missing from current benchmarks, providing new challenges and opportunities for future research. Our code repository is accessible at https://github.com/fpour/DGB.git.
|
['Reihaneh Rabbany', 'Kellin Pelrine', 'Shenyang Huang', 'Farimah Poursafaei']
|
2022-07-20
| null | null | null | null |
['dynamic-link-prediction']
|
['graphs']
|
[ 7.31337145e-02 1.67941898e-01 -5.39990664e-01 -1.46757767e-01
-2.42890999e-01 -8.05429339e-01 7.04308271e-01 3.92812163e-01
1.08026387e-03 8.00185680e-01 1.29694179e-01 -6.32740617e-01
-2.08880082e-01 -8.66943836e-01 -5.31106472e-01 -2.51131505e-01
-9.36636984e-01 3.53271544e-01 4.78148669e-01 -2.84972519e-01
7.37267956e-02 3.17921728e-01 -1.22159624e+00 1.86873704e-01
8.48098099e-01 4.56865788e-01 -1.69376627e-01 6.91897154e-01
1.84077099e-01 7.11981177e-01 -5.03288507e-01 -6.63903594e-01
4.01208669e-01 -3.03918630e-01 -8.91136944e-01 -1.85795650e-01
7.03297257e-01 1.17817082e-01 -5.80476403e-01 1.00740111e+00
4.94714648e-01 2.32004687e-01 5.20804942e-01 -1.68326747e+00
-7.69828200e-01 9.68876421e-01 -7.73757041e-01 7.56076813e-01
6.20686293e-01 3.43945831e-01 1.54676914e+00 -5.78728795e-01
1.05052364e+00 1.22021818e+00 9.62086260e-01 3.51582736e-01
-1.37643909e+00 -6.18950307e-01 7.99288809e-01 4.58942950e-01
-1.15743291e+00 -4.04777139e-01 9.31723118e-01 -3.68837982e-01
1.05884206e+00 4.41542417e-01 7.71971643e-01 1.43267035e+00
3.07112455e-01 6.38182998e-01 9.82422590e-01 -1.89646035e-01
-1.83480561e-01 -1.29877016e-01 4.62767333e-01 8.06249201e-01
4.79012161e-01 -8.48414525e-02 -5.77494442e-01 -3.35389495e-01
3.95370454e-01 -1.28785476e-01 -5.90426505e-01 -7.26812661e-01
-1.06520462e+00 7.00570405e-01 4.75599378e-01 3.36053789e-01
-8.82147476e-02 8.35563689e-02 4.55339402e-01 6.46014333e-01
6.77789688e-01 4.47733581e-01 -4.88500357e-01 -2.67478704e-01
-5.39228022e-01 1.86507761e-01 1.21795654e+00 9.45556223e-01
6.96428180e-01 -1.57832280e-01 -1.97669163e-01 6.37483180e-01
1.50493011e-01 -4.19025775e-03 1.59514859e-01 -4.64713335e-01
6.49383247e-01 5.97337544e-01 -2.52270073e-01 -1.38957930e+00
-4.56386030e-01 -7.89261103e-01 -7.38165021e-01 -1.75829202e-01
7.51282871e-01 -5.74673302e-02 -9.72168386e-01 1.90597808e+00
2.49646768e-01 3.80111545e-01 -2.91445076e-01 4.81366813e-01
7.29064226e-01 3.83754343e-01 1.39254639e-02 -1.29933611e-01
9.84543264e-01 -1.12556779e+00 -5.40296972e-01 -4.79429930e-01
7.78923988e-01 -5.11206150e-01 1.41029131e+00 2.34740451e-01
-8.42713416e-01 -3.37252058e-02 -9.96437848e-01 1.72225013e-01
-5.64037323e-01 -3.82744581e-01 9.91742074e-01 5.10609865e-01
-1.27962899e+00 9.47718561e-01 -9.37785506e-01 -7.29799032e-01
2.25938901e-01 6.25662580e-02 -1.51551425e-01 -4.77666669e-02
-1.34817326e+00 8.27859521e-01 2.25525737e-01 -2.31520534e-02
-6.27031028e-01 -8.84790361e-01 -8.57390046e-01 5.76568255e-03
7.10046291e-01 -6.33600056e-01 1.14252937e+00 -6.86490893e-01
-8.78525257e-01 9.11597967e-01 3.65421847e-02 -4.68194723e-01
8.63609552e-01 -4.25722189e-02 -5.87634861e-01 -1.13021500e-01
-2.79402863e-02 1.42149091e-01 4.41146672e-01 -1.21704257e+00
-2.49530345e-01 -2.06779197e-01 1.51270345e-01 -8.88543669e-03
-5.31874299e-01 -3.68104398e-01 -7.54082263e-01 -8.41814816e-01
-2.83224195e-01 -9.75077271e-01 -1.38603240e-01 -2.40737468e-01
-8.21051419e-01 -2.81707525e-01 7.93352962e-01 -4.73066062e-01
1.76733422e+00 -1.79025626e+00 1.02150247e-01 5.82369566e-01
7.45389581e-01 1.42424144e-02 -2.97657877e-01 7.82388389e-01
-4.24931139e-01 4.58248287e-01 -1.71240479e-01 -3.19749445e-01
-8.30732659e-02 -4.20053750e-02 -8.21350962e-02 3.50217491e-01
1.86183691e-01 1.01216543e+00 -1.20028293e+00 -4.21347678e-01
-2.02848390e-01 2.13687703e-01 -4.14290547e-01 -2.99002349e-01
-2.20968202e-01 1.27250895e-01 -2.45894089e-01 8.92023385e-01
4.05630469e-01 -8.33435953e-01 6.55355394e-01 -1.64850473e-01
3.22671056e-01 4.15178448e-01 -1.14897072e+00 1.36391103e+00
1.56745370e-02 8.12781990e-01 -4.95300479e-02 -9.77198005e-01
7.40210652e-01 -1.84831411e-01 4.97952908e-01 -7.60424554e-01
-1.98930934e-01 -6.44195005e-02 3.25531363e-01 -3.92948389e-01
5.54277182e-01 3.30927312e-01 2.85947233e-01 6.38347745e-01
4.12473604e-02 1.81711227e-01 7.76270986e-01 7.46734321e-01
1.69999075e+00 -6.70423880e-02 3.27204257e-01 -2.74447262e-01
9.39179733e-02 -1.50067568e-01 5.78415811e-01 7.56058037e-01
-4.37180936e-01 3.65279764e-01 8.99888694e-01 -5.39327204e-01
-7.56068170e-01 -1.07444465e+00 6.07748851e-02 1.13466752e+00
1.44762620e-01 -1.00862300e+00 -1.64987043e-01 -1.11898375e+00
3.71924907e-01 5.16126513e-01 -8.03587139e-01 -1.66442722e-01
-5.93787670e-01 -9.89996731e-01 2.19794691e-01 4.52531427e-01
-3.56006250e-02 -9.09655869e-01 1.77419893e-02 4.12069075e-02
-8.84950086e-02 -9.93809342e-01 -4.62596953e-01 4.71465401e-02
-7.95066953e-01 -1.50470936e+00 -4.09338266e-01 -6.66562617e-01
5.16392589e-01 3.63197803e-01 1.98320746e+00 5.25461018e-01
-3.38113964e-01 7.04243720e-01 -4.13511157e-01 -2.10185602e-01
-2.57905900e-01 5.59450984e-01 -1.46695867e-01 -4.26014185e-01
1.94253013e-01 -9.57789123e-01 -6.61438704e-01 2.26803571e-01
-4.78844941e-01 3.11270263e-02 3.16937268e-01 7.19115317e-01
3.24233294e-01 -6.48122579e-02 6.34810090e-01 -1.36096740e+00
1.06532371e+00 -8.53834331e-01 -3.71239364e-01 3.70588392e-01
-1.05408847e+00 -2.62143929e-03 5.29416144e-01 -5.37329435e-01
-5.92620194e-01 -4.99443233e-01 2.34360233e-01 -1.50739953e-01
3.50556552e-01 8.20332706e-01 2.70810783e-01 -2.26097014e-02
9.03694332e-01 -1.72400445e-01 5.38805239e-02 -3.45717937e-01
3.78248394e-01 6.09958731e-02 3.05584311e-01 -7.29221165e-01
9.30225730e-01 3.38173777e-01 -2.12765247e-01 -4.50886697e-01
-5.95078230e-01 -3.41439784e-01 -2.65265763e-01 -4.15374309e-01
2.85118092e-02 -6.96507454e-01 -4.95820671e-01 3.08528543e-01
-7.38056064e-01 -9.59610462e-01 -2.24815011e-01 -1.05625959e-02
-2.33567372e-01 6.96927488e-01 -7.17391193e-01 -5.55471718e-01
-5.35590708e-01 -7.35389411e-01 5.62821686e-01 3.01636904e-02
-4.92892355e-01 -1.54427338e+00 3.23413372e-01 -2.00922757e-01
3.46336693e-01 5.81835568e-01 9.18905258e-01 -7.21482933e-01
-5.63219547e-01 -8.67602080e-02 -1.98545352e-01 -6.00595064e-02
1.99234098e-01 6.61114454e-01 -5.87801814e-01 -7.78868854e-01
-7.63000250e-01 -2.06231833e-01 1.07048047e+00 4.62016053e-02
1.07372749e+00 -2.28940934e-01 -8.29757035e-01 4.87098068e-01
1.27371442e+00 -2.32246593e-01 4.47373241e-01 3.59850764e-01
8.25750709e-01 5.73412299e-01 5.67291260e-01 3.49163443e-01
7.96764493e-01 5.15802622e-01 4.88649040e-01 -1.30228311e-01
-3.34583580e-01 -2.92468250e-01 3.58239591e-01 8.45186114e-01
-8.56806859e-02 -7.02532709e-01 -1.35562289e+00 8.00559402e-01
-1.98476839e+00 -9.75500166e-01 -2.90182203e-01 2.10131264e+00
8.73024702e-01 6.40504360e-01 4.78427976e-01 3.79641056e-02
7.61287689e-01 4.58366156e-01 -6.61324978e-01 8.26382786e-02
5.57423830e-02 1.16105735e-01 3.92479926e-01 4.74189997e-01
-9.34820950e-01 7.98687518e-01 6.61436987e+00 6.06416881e-01
-1.20403636e+00 -2.09842026e-01 6.67460084e-01 -8.15172344e-02
-6.77876532e-01 1.78838283e-01 -4.74039942e-01 4.95601118e-01
7.47932196e-01 -6.61423087e-01 4.35557067e-01 7.10832775e-01
-8.81998539e-02 1.74467981e-01 -1.28958416e+00 7.00304866e-01
-1.39006987e-01 -1.34596407e+00 -1.67048723e-01 -5.54241650e-02
6.00781679e-01 4.73223150e-01 8.51114839e-02 4.82773721e-01
7.89260089e-01 -8.10814083e-01 3.79941881e-01 4.03603792e-01
6.73935711e-01 -4.06234771e-01 3.17174971e-01 5.76988347e-02
-1.55147791e+00 3.13351154e-02 -1.24638225e-03 -1.90861672e-01
9.09924060e-02 8.20955634e-01 -9.53018963e-01 7.74124801e-01
9.04652655e-01 1.34972596e+00 -1.15062129e+00 1.21330082e+00
-2.14481473e-01 7.05318213e-01 -3.88934523e-01 1.26275076e-02
-2.12701663e-01 -7.78875649e-02 7.12113202e-01 1.40971220e+00
1.41501248e-01 -3.96608204e-01 2.74358213e-01 6.42355800e-01
-2.07617953e-01 1.93013877e-01 -9.95750546e-01 -3.53146315e-01
7.29121804e-01 1.42216194e+00 -9.83437777e-01 -2.46222317e-01
-4.86505687e-01 5.64435124e-01 8.84704590e-01 6.98155820e-01
-8.10010552e-01 -3.09804320e-01 6.48525417e-01 3.05458605e-01
7.72621715e-03 -4.46865052e-01 -3.01084742e-02 -1.47549784e+00
2.21195593e-01 -9.75168169e-01 9.67576325e-01 -3.55206609e-01
-1.67611134e+00 4.92707521e-01 3.31369378e-02 -1.11754441e+00
-2.71186769e-01 -4.97266978e-01 -9.99974132e-01 4.86655623e-01
-1.52805161e+00 -9.41776395e-01 -6.63967669e-01 4.26585853e-01
2.42630422e-01 1.95462123e-01 6.21307790e-01 4.47608799e-01
-8.80416811e-01 7.69836724e-01 -2.15287611e-01 1.61293402e-01
9.85183358e-01 -1.49701262e+00 1.07839143e+00 9.91138339e-01
3.75363529e-01 6.59774721e-01 8.46238136e-01 -9.69918609e-01
-1.30661488e+00 -9.69221354e-01 5.98856688e-01 -5.76405764e-01
1.20132232e+00 -6.05728567e-01 -1.13433313e+00 1.11900294e+00
3.10261548e-01 1.37852162e-01 5.93935788e-01 7.02703536e-01
-4.11247224e-01 -6.68595731e-03 -8.33840370e-01 8.81117821e-01
1.68909299e+00 -3.32016736e-01 -7.11919460e-03 4.86565143e-01
6.06023669e-01 -3.93503666e-01 -9.69778478e-01 6.96720660e-01
5.13156474e-01 -8.88633907e-01 9.51202691e-01 -8.51998985e-01
1.57819077e-01 2.95502730e-02 3.61682624e-01 -1.63982368e+00
-4.79460448e-01 -1.09450746e+00 -6.57161355e-01 1.34621990e+00
7.67413259e-01 -1.04710269e+00 9.79598343e-01 3.80006045e-01
5.38302772e-02 -1.07364547e+00 -4.57852900e-01 -9.51175511e-01
-1.08719438e-01 -1.78692147e-01 3.38892162e-01 1.43964088e+00
2.93289900e-01 4.31524217e-01 -2.20621258e-01 -2.04384506e-01
5.82578480e-01 1.97008401e-01 9.84713376e-01 -1.35479355e+00
-4.75763977e-01 -6.38610959e-01 -4.63736862e-01 -8.91075253e-01
1.73333228e-01 -1.24564886e+00 -4.44225222e-01 -1.64668715e+00
2.70948321e-01 -6.09813035e-01 -4.06058937e-01 6.81175292e-01
-5.11082232e-01 2.28271484e-01 1.82239562e-01 1.18526220e-01
-7.81279087e-01 3.98707032e-01 1.31857109e+00 -2.18660742e-01
-2.02067390e-01 -1.81473233e-02 -7.49260664e-01 5.21772981e-01
9.85540867e-01 -2.71705568e-01 -6.70492887e-01 -2.88496763e-01
5.43301284e-01 -2.87827909e-01 9.64449272e-02 -7.45845973e-01
1.76954985e-01 -9.60365310e-02 1.86054781e-02 -4.47504729e-01
-2.02358682e-02 -4.60773766e-01 1.74741596e-01 5.07820368e-01
-1.37912408e-01 7.11148620e-01 2.27286622e-01 8.77822876e-01
5.72544523e-02 1.62188649e-01 4.07776088e-01 1.15358301e-01
-8.44335675e-01 5.32025456e-01 1.22760266e-01 5.60780168e-01
1.13598180e+00 -1.72806621e-01 -7.82023489e-01 -7.22377062e-01
-8.89964104e-01 7.57911742e-01 6.00355327e-01 6.05436802e-01
3.10188681e-01 -1.26164234e+00 -5.74856699e-01 -2.10058570e-01
2.73939371e-01 -3.15678298e-01 -4.41661179e-02 1.04141045e+00
-3.41334045e-01 -2.36465946e-01 -8.55484977e-02 -4.33404744e-01
-1.40328789e+00 5.10321736e-01 3.26882869e-01 -6.78596675e-01
-7.84178257e-01 7.95795321e-01 -1.83394253e-01 -3.63840789e-01
3.47985148e-01 -1.92725390e-01 -1.37507558e-01 2.01825783e-01
3.22696455e-02 5.33713639e-01 1.01156272e-01 -8.53621513e-02
-3.82688046e-01 1.48221388e-01 -3.65691632e-01 3.76820326e-01
1.41017175e+00 -1.15162209e-01 -9.24132578e-03 5.21776676e-01
9.25137699e-01 2.68147230e-01 -1.05628252e+00 -3.62045348e-01
4.27446008e-01 -4.60284263e-01 -4.84849393e-01 -8.19195092e-01
-1.26445842e+00 3.68032813e-01 2.60477334e-01 9.21977043e-01
9.64375854e-01 2.39162482e-02 4.45882887e-01 2.50724316e-01
2.28292599e-01 -9.19529200e-01 3.42784673e-01 4.78549898e-01
9.18739617e-01 -1.32420290e+00 3.21267933e-01 -8.89757752e-01
-3.84833932e-01 1.01362658e+00 8.81142199e-01 -1.81789305e-02
7.16430366e-01 2.97616959e-01 -8.07633996e-02 -5.72533786e-01
-1.13856483e+00 -1.95121765e-01 1.72659382e-01 7.01528132e-01
6.92974925e-01 7.40286754e-03 -4.78833586e-01 1.16011970e-01
-3.50727051e-01 -3.45986605e-01 5.62353194e-01 9.00794446e-01
5.78551739e-02 -1.29650247e+00 1.86251998e-01 7.87454784e-01
-2.91102231e-01 -1.41545147e-01 -8.10513735e-01 1.15861464e+00
-6.03954673e-01 6.77295148e-01 3.61145101e-02 -4.90992278e-01
4.08904076e-01 9.28821564e-02 6.07222140e-01 -7.18823612e-01
-3.90306801e-01 -4.28306043e-01 5.96698225e-01 -6.90676868e-01
-1.66142046e-01 -5.95666766e-01 -9.55553293e-01 -5.73267162e-01
-2.88357407e-01 1.02476524e-02 1.58028126e-01 2.49049544e-01
6.23833001e-01 7.13101506e-01 2.92150646e-01 -6.74248576e-01
-4.99211013e-01 -9.89759266e-01 -3.78015995e-01 7.65713274e-01
1.75111711e-01 -8.19656014e-01 -5.56882083e-01 -1.59255981e-01]
|
[7.030940532684326, 6.122877597808838]
|
3d03e6bd-4778-469b-9bd4-207cbf99778b
|
what-makes-convolutional-models-great-on-long
|
2210.09298
| null |
https://arxiv.org/abs/2210.09298v1
|
https://arxiv.org/pdf/2210.09298v1.pdf
|
What Makes Convolutional Models Great on Long Sequence Modeling?
|
Convolutional models have been widely used in multiple domains. However, most existing models only use local convolution, making the model unable to handle long-range dependency efficiently. Attention overcomes this problem by aggregating global information but also makes the computational complexity quadratic to the sequence length. Recently, Gu et al. [2021] proposed a model called S4 inspired by the state space model. S4 can be efficiently implemented as a global convolutional model whose kernel size equals the input sequence length. S4 can model much longer sequences than Transformers and achieve significant gains over SoTA on several long-range tasks. Despite its empirical success, S4 is involved. It requires sophisticated parameterization and initialization schemes. As a result, S4 is less intuitive and hard to use. Here we aim to demystify S4 and extract basic principles that contribute to the success of S4 as a global convolutional model. We focus on the structure of the convolution kernel and identify two critical but intuitive principles enjoyed by S4 that are sufficient to make up an effective global convolutional model: 1) The parameterization of the convolutional kernel needs to be efficient in the sense that the number of parameters should scale sub-linearly with sequence length. 2) The kernel needs to satisfy a decaying structure that the weights for convolving with closer neighbors are larger than the more distant ones. Based on the two principles, we propose a simple yet effective convolutional model called Structured Global Convolution (SGConv). SGConv exhibits strong empirical performance over several tasks: 1) With faster speed, SGConv surpasses S4 on Long Range Arena and Speech Command datasets. 2) When plugging SGConv into standard language and vision models, it shows the potential to improve both efficiency and performance.
|
['Debadeepta Dey', 'Deming Chen', 'Yi Zhang', 'Tianle Cai', 'Yuhong Li']
|
2022-10-17
| null | null | null | null |
['long-range-modeling']
|
['natural-language-processing']
|
[-8.26044828e-02 -4.82791632e-01 -1.13853946e-01 -3.41696650e-01
-1.50064141e-01 -7.35576093e-01 5.10304987e-01 -1.41617775e-01
-6.79767609e-01 3.12789947e-01 1.17943071e-01 -6.67363107e-01
-9.84608531e-02 -6.72300041e-01 -6.50647342e-01 -5.15689492e-01
-2.00181514e-01 -2.11695686e-01 5.68249047e-01 -3.64841968e-01
1.18529908e-01 5.46452522e-01 -1.38745987e+00 1.85720161e-01
8.26895058e-01 9.47144985e-01 7.48602867e-01 7.90909827e-01
-1.54286265e-01 8.35157752e-01 -5.40861070e-01 -1.31858677e-01
2.39875659e-01 -4.16606724e-01 -1.07129872e+00 -2.32565030e-01
1.48004100e-01 -3.43048215e-01 -3.26884806e-01 1.04454851e+00
5.04654288e-01 1.43963501e-01 3.46484542e-01 -1.05577755e+00
-8.21231008e-01 8.06421876e-01 -2.26880699e-01 4.30751234e-01
-1.89070538e-01 3.53803426e-01 1.23493481e+00 -5.80870271e-01
2.01017082e-01 1.10857618e+00 9.95233774e-01 4.78389859e-01
-1.14000475e+00 -4.89781022e-01 4.28639859e-01 3.05007190e-01
-1.25465190e+00 -2.26706296e-01 3.18563282e-01 -2.83468157e-01
1.38384593e+00 2.67275333e-01 6.35502458e-01 1.10295486e+00
4.54252549e-02 7.60708213e-01 8.81986797e-01 -2.56781489e-01
2.00427115e-01 -2.66267568e-01 4.65670735e-01 6.12385333e-01
1.01700716e-01 7.54627958e-02 -4.04235244e-01 -3.75689082e-02
1.11127639e+00 -1.52557939e-01 -3.53143066e-01 -1.24408618e-01
-1.32914960e+00 7.94727802e-01 5.89692473e-01 6.34146512e-01
-2.58781254e-01 5.90027213e-01 5.74687958e-01 5.27535498e-01
1.98887199e-01 5.84918320e-01 -7.10000813e-01 -4.51662570e-01
-6.30856812e-01 1.24255873e-01 5.44214129e-01 7.56363451e-01
6.57753170e-01 9.10600200e-02 -2.97600068e-02 9.15614605e-01
2.85679325e-02 3.67915243e-01 5.84660649e-01 -8.46722901e-01
3.92738700e-01 4.66870010e-01 -6.83723763e-02 -7.45559514e-01
-5.75599134e-01 -7.11842120e-01 -9.12094355e-01 9.75314900e-02
4.14251983e-01 -1.56230703e-01 -8.84006143e-01 2.08903956e+00
-2.93073714e-01 1.56652585e-01 5.73210754e-02 8.53106201e-01
6.00760639e-01 7.55735099e-01 1.38248488e-01 1.65435910e-01
1.30521166e+00 -1.05613148e+00 -5.21911263e-01 -4.35314894e-01
8.29002440e-01 -6.27395630e-01 1.20695448e+00 1.44020513e-01
-9.45201516e-01 -7.50207722e-01 -1.11295092e+00 -1.68527905e-02
-2.60560662e-01 1.68542519e-01 8.70260417e-01 5.28808832e-01
-1.34876382e+00 6.47960305e-01 -9.40592229e-01 -4.79006648e-01
4.67429124e-02 2.76886761e-01 -2.79948294e-01 2.44736359e-01
-1.43563521e+00 1.12700605e+00 6.88585699e-01 7.56861940e-02
-6.59220040e-01 -6.20773792e-01 -7.45437622e-01 3.97911727e-01
2.32181147e-01 -7.11558759e-01 1.35795760e+00 -1.03565598e+00
-1.51568007e+00 4.23493207e-01 -1.42033666e-01 -5.70308328e-01
2.60974497e-01 -4.01627749e-01 -1.88515589e-01 1.09275028e-01
-1.54231668e-01 6.12836897e-01 7.22760975e-01 -7.37613559e-01
-5.37818193e-01 -1.37634184e-02 1.74237266e-01 9.20203328e-02
-5.20028710e-01 1.42152056e-01 -6.06685102e-01 -7.33859777e-01
-6.64275587e-02 -9.46468711e-01 -2.61053801e-01 -1.20493665e-01
-5.53628206e-02 -3.56408954e-01 6.81159556e-01 -3.37539554e-01
1.30932820e+00 -2.26238275e+00 6.70026839e-02 7.20086619e-02
1.89662755e-01 7.21176565e-01 -5.46387374e-01 5.00752330e-01
-3.20834994e-01 2.75455385e-01 -1.81301996e-01 -4.55135293e-02
-7.47833913e-03 3.76092881e-01 -3.85018975e-01 3.09021354e-01
2.77376831e-01 1.34182346e+00 -9.55253541e-01 4.63227592e-02
1.68366745e-01 4.52880979e-01 -5.50438881e-01 1.06668711e-01
1.10176643e-02 7.48838708e-02 -3.13017070e-01 4.42713499e-02
5.40981293e-01 -6.05638325e-01 6.92055449e-02 -1.89437374e-01
-3.18994939e-01 3.81918073e-01 -1.01261711e+00 1.58504665e+00
-3.31800431e-01 6.84692085e-01 3.90429869e-02 -1.12631118e+00
8.18115413e-01 1.77230284e-01 3.06313276e-01 -5.46889067e-01
8.73766616e-02 2.38281891e-01 1.00471154e-01 -4.72136587e-01
3.32635343e-01 -1.02741145e-01 4.23042066e-02 3.62872958e-01
1.26685590e-01 -4.52054068e-02 1.35280974e-02 2.57296920e-01
1.20249474e+00 -2.77645402e-02 3.49582642e-01 -4.67800617e-01
4.21302885e-01 -3.68141890e-01 4.10171807e-01 9.88343418e-01
-1.89515963e-01 3.74374986e-01 5.65086007e-01 -4.13646042e-01
-1.07893145e+00 -8.13280761e-01 1.15171365e-01 1.22708511e+00
1.01500146e-01 -6.22394741e-01 -7.79221594e-01 -5.33147871e-01
-7.56900907e-02 3.94139290e-01 -6.42959774e-01 -2.91150540e-01
-5.73912263e-01 -6.31135464e-01 9.61982489e-01 9.38912868e-01
8.06948781e-01 -1.05722916e+00 -7.41252959e-01 2.48688683e-01
-2.26019040e-01 -1.15001154e+00 -7.02053666e-01 4.72450614e-01
-8.51591110e-01 -8.09804320e-01 -7.25325346e-01 -6.86688006e-01
2.17649028e-01 4.91963118e-01 9.02139306e-01 2.14142933e-01
-6.11139797e-02 1.92571357e-01 -5.24376333e-01 -2.30392411e-01
-2.29114130e-01 4.39449102e-01 1.04117878e-01 -9.38668549e-02
2.60104030e-01 -6.65482700e-01 -5.12395263e-01 4.00956988e-01
-9.60801721e-01 3.56893875e-02 7.59427190e-01 9.23580706e-01
1.49566578e-02 -8.33640397e-02 6.39670491e-01 -4.50641274e-01
7.84782946e-01 -3.25184733e-01 -4.85351473e-01 1.67990237e-01
-5.08317769e-01 3.84718060e-01 7.47027874e-01 -5.88284075e-01
-6.76741600e-01 -1.35782227e-01 -5.25908649e-01 -4.19718117e-01
7.39574507e-02 7.66553283e-01 1.90413877e-01 -5.40162437e-02
5.88627934e-01 4.78180915e-01 2.25070566e-01 -5.53862333e-01
2.88376033e-01 5.28694391e-01 4.82910275e-01 -5.34756422e-01
5.84340215e-01 3.10089827e-01 -2.70750076e-01 -1.06813002e+00
-6.64458871e-01 -4.44176316e-01 -5.77942729e-01 1.24372914e-01
8.26739728e-01 -7.61851609e-01 -1.00618279e+00 8.31254661e-01
-1.24144089e+00 -6.74899995e-01 -1.67854369e-01 5.47822833e-01
-5.36943257e-01 5.46732605e-01 -8.31165016e-01 -6.61194086e-01
-2.31880233e-01 -1.10278153e+00 6.06113434e-01 -2.13196129e-02
-2.38097116e-01 -1.03820372e+00 -1.63558900e-01 -1.82881862e-01
8.85210335e-01 -1.93036109e-01 9.26801026e-01 -5.26910484e-01
-3.64623070e-01 1.46143019e-01 -4.42615092e-01 7.64924228e-01
2.88938247e-02 -1.01919614e-01 -8.64765108e-01 -4.09859955e-01
2.25042924e-02 -2.19125152e-01 1.14644575e+00 3.30320001e-01
1.28526199e+00 -6.54706433e-02 -5.72075620e-02 8.98254156e-01
1.25532913e+00 8.64463747e-02 6.44863844e-01 4.52319145e-01
6.38254464e-01 2.87941784e-01 1.18760139e-01 2.41706669e-01
1.95291027e-01 6.20592237e-01 4.79399055e-01 -1.73525348e-01
-9.89870876e-02 -1.84532508e-01 6.21392906e-01 1.12537730e+00
-1.47033304e-01 -1.59198672e-01 -8.69540811e-01 5.36693215e-01
-1.99691463e+00 -1.08035874e+00 -7.70289302e-02 1.96720624e+00
7.92515814e-01 2.00552613e-01 1.08549315e-02 1.04523543e-02
4.64418977e-01 3.18640500e-01 -5.46302259e-01 -5.59806049e-01
-2.76291877e-01 2.73279995e-01 6.46336854e-01 4.33423936e-01
-1.03093278e+00 1.06050396e+00 7.23827648e+00 9.08606529e-01
-1.38814890e+00 5.78614473e-02 2.33598188e-01 6.91148639e-02
-1.62859231e-01 -4.09931578e-02 -8.88956249e-01 4.52604860e-01
1.03716910e+00 3.06288898e-02 4.91622150e-01 7.93396056e-01
1.12038165e-01 2.04436317e-01 -1.00783634e+00 9.79911208e-01
-2.64820516e-01 -1.44171691e+00 7.36818388e-02 -2.33093165e-02
4.21403468e-01 4.73875463e-01 8.03553835e-02 3.64739060e-01
4.54983354e-01 -1.26691270e+00 8.92161310e-01 1.31198779e-01
8.55083704e-01 -7.35759437e-01 6.62233591e-01 5.67781687e-01
-1.32777631e+00 -3.22976708e-01 -5.02865553e-01 -3.28071862e-01
1.46815598e-01 3.43145013e-01 -3.13157290e-01 3.90583128e-01
7.87160099e-01 7.55642653e-01 -5.13471723e-01 8.90103102e-01
-2.65369624e-01 7.80086458e-01 -3.61851543e-01 -1.18263245e-01
8.36074054e-01 8.05581827e-03 3.24719667e-01 1.57953537e+00
2.83402950e-01 7.55014867e-02 9.38508734e-02 7.91365802e-01
1.38880119e-01 -1.04376614e-01 -5.32665372e-01 -1.79387495e-01
4.01279837e-01 9.77019906e-01 -4.99784827e-01 -2.08414569e-01
-6.47301078e-01 1.01374733e+00 6.13137603e-01 5.42697251e-01
-9.17989552e-01 -6.17808282e-01 1.05573654e+00 -1.31779954e-01
7.23182917e-01 -6.48133337e-01 -2.77421683e-01 -1.22785902e+00
-1.57936171e-01 -8.65746677e-01 1.59481555e-01 -6.42463267e-01
-1.22054303e+00 8.20263147e-01 -2.16174826e-01 -9.76042151e-01
-7.55984783e-02 -7.68086314e-01 -5.42916894e-01 1.11530864e+00
-1.60754514e+00 -1.04114091e+00 -2.03512520e-01 7.75638878e-01
4.67403233e-01 4.39339355e-02 8.52396190e-01 2.30231598e-01
-4.33174908e-01 7.72913396e-01 -4.48520146e-02 3.01065862e-01
4.50904578e-01 -1.12682939e+00 8.97525251e-01 9.57711220e-01
5.18414415e-02 1.11067247e+00 3.94174546e-01 -4.15381968e-01
-1.18688405e+00 -1.04071295e+00 8.41982663e-01 -3.15107614e-01
8.78323853e-01 -3.25501293e-01 -1.08262742e+00 7.09356189e-01
1.98324263e-01 2.56673321e-02 4.38818693e-01 2.33375072e-01
-8.21843684e-01 2.77936645e-02 -5.20271063e-01 5.90711892e-01
1.22714520e+00 -7.21805692e-01 -4.59644616e-01 -1.24160625e-01
8.67598712e-01 -1.38192639e-01 -7.30425477e-01 3.44063073e-01
4.19512123e-01 -9.54974473e-01 9.96705234e-01 -5.95587730e-01
1.59632400e-01 -2.65681416e-01 -1.15946770e-01 -1.37184691e+00
-7.57147968e-01 -8.39430332e-01 -1.44288704e-01 6.52642906e-01
3.73500705e-01 -9.82429326e-01 2.80306995e-01 1.63436487e-01
-2.15981305e-01 -8.12995493e-01 -8.48046482e-01 -1.32367086e+00
3.16189498e-01 -6.54340446e-01 5.78317225e-01 9.31303740e-01
8.36105458e-03 4.57802325e-01 -3.95033926e-01 9.75549892e-02
1.97719574e-01 -6.92613646e-02 5.22354186e-01 -1.14274085e+00
-3.83453548e-01 -7.11692750e-01 -4.14911509e-01 -1.76353741e+00
2.51765568e-02 -8.93609881e-01 1.80718338e-03 -1.26949167e+00
2.05807209e-01 -5.96175194e-01 -4.96995389e-01 8.48991632e-01
-2.94759512e-01 1.84232086e-01 3.38389784e-01 3.02433342e-01
-2.73625672e-01 3.19775045e-01 1.12420440e+00 1.66319758e-01
-1.18644074e-01 -7.62110800e-02 -8.51247311e-01 6.86226845e-01
9.41846550e-01 -2.40114462e-02 -5.15293837e-01 -6.70042753e-01
4.35858727e-01 -2.97854096e-01 4.66128260e-01 -9.22569871e-01
2.55170375e-01 -1.71424776e-01 8.13454539e-02 -4.15834397e-01
2.21097395e-01 -6.25772595e-01 -1.72014404e-02 5.93349993e-01
-3.97091299e-01 2.45429605e-01 2.43189603e-01 4.85309303e-01
-1.76664636e-01 -2.56850004e-01 9.20286715e-01 -7.44456947e-02
-9.72516835e-01 2.53498793e-01 -5.92053473e-01 -1.05724536e-01
8.15613985e-01 -1.24285191e-01 -3.21928471e-01 -4.64407921e-01
-4.93363321e-01 1.47412986e-01 3.58582199e-01 6.54207408e-01
4.78796035e-01 -1.15082908e+00 -6.73213601e-01 3.31605881e-01
-2.02437676e-02 -2.82742739e-01 8.34798738e-02 8.47684681e-01
-5.87075830e-01 6.74624026e-01 -1.08183913e-01 -6.65205956e-01
-1.20819163e+00 6.84633136e-01 4.31463480e-01 -1.85129374e-01
-8.57348740e-01 9.78656709e-01 4.34611678e-01 -3.69583249e-01
2.64142960e-01 -7.92442083e-01 -1.17868856e-01 -5.26898243e-02
7.52581418e-01 8.15639719e-02 -9.47557762e-02 -3.95304948e-01
-4.15931344e-01 6.66449904e-01 -1.68208778e-01 2.35961066e-04
1.40627420e+00 -9.68735218e-02 -1.12360194e-01 3.58571559e-01
1.30890477e+00 -2.23306715e-01 -1.44864881e+00 -3.62835586e-01
-2.53333128e-04 -1.87710971e-01 5.57240695e-02 -6.12578928e-01
-1.02559400e+00 9.85938966e-01 2.14148924e-01 3.48317474e-01
1.10148692e+00 6.82436395e-03 9.10748959e-01 3.99383366e-01
1.96711406e-01 -8.47940326e-01 1.35155872e-01 1.26928425e+00
6.74363673e-01 -7.84446955e-01 -3.52287382e-01 -2.71149933e-01
-6.77200556e-01 1.25110638e+00 4.59383845e-01 -3.17252949e-02
5.59942901e-01 4.39421803e-01 2.86647547e-02 -7.94283375e-02
-8.07860494e-01 -3.90168697e-01 1.47976801e-01 4.53726083e-01
3.70705605e-01 -1.89205091e-02 -2.21250787e-01 5.17371297e-01
-9.08818692e-02 4.72068042e-02 2.30122745e-01 8.11166227e-01
-6.53755426e-01 -1.03167832e+00 -5.55971824e-02 1.15811072e-01
-3.69536072e-01 -3.24627817e-01 -1.42208308e-01 6.15513682e-01
1.65438913e-02 7.86014438e-01 7.53405243e-02 -5.26289463e-01
2.40514711e-01 7.49369487e-02 2.92565107e-01 -4.43543345e-01
-7.65694261e-01 -7.50098228e-02 -1.40642121e-01 -6.86345160e-01
-2.65738368e-01 -2.22995400e-01 -1.17032659e+00 -5.30661762e-01
-4.52762872e-01 9.12702829e-02 6.12440050e-01 9.06316876e-01
5.19129574e-01 6.27876282e-01 3.45743150e-01 -6.62378371e-01
-9.49799299e-01 -9.60605562e-01 -4.02147263e-01 2.46376976e-01
5.57944715e-01 -4.36614573e-01 -3.30886960e-01 -4.13000137e-02]
|
[10.80197525024414, 6.547518253326416]
|
a6c584d5-d50f-46e2-b095-942a86d75a20
|
mivolo-multi-input-transformer-for-age-and
|
2307.04616
| null |
https://arxiv.org/abs/2307.04616v1
|
https://arxiv.org/pdf/2307.04616v1.pdf
|
MiVOLO: Multi-input Transformer for Age and Gender Estimation
|
Age and gender recognition in the wild is a highly challenging task: apart from the variability of conditions, pose complexities, and varying image quality, there are cases where the face is partially or completely occluded. We present MiVOLO (Multi Input VOLO), a straightforward approach for age and gender estimation using the latest vision transformer. Our method integrates both tasks into a unified dual input/output model, leveraging not only facial information but also person image data. This improves the generalization ability of our model and enables it to deliver satisfactory results even when the face is not visible in the image. To evaluate our proposed model, we conduct experiments on four popular benchmarks and achieve state-of-the-art performance, while demonstrating real-time processing capabilities. Additionally, we introduce a novel benchmark based on images from the Open Images Dataset. The ground truth annotations for this benchmark have been meticulously generated by human annotators, resulting in high accuracy answers due to the smart aggregation of votes. Furthermore, we compare our model's age recognition performance with human-level accuracy and demonstrate that it significantly outperforms humans across a majority of age ranges. Finally, we grant public access to our models, along with the code for validation and inference. In addition, we provide extra annotations for used datasets and introduce our new benchmark.
|
['Irina Tolstykh', 'Maksim Kuprashevich']
|
2023-07-10
| null | null | null | null |
['age-and-gender-estimation', 'facial-attribute-classification', 'age-and-gender-classification', 'age-estimation', 'gender-prediction', 'age-estimation']
|
['computer-vision', 'computer-vision', 'computer-vision', 'computer-vision', 'computer-vision', 'miscellaneous']
|
[-5.35673602e-03 1.53653115e-01 8.28952566e-02 -7.18702972e-01
-4.44687873e-01 -4.24009234e-01 7.69716144e-01 1.13595210e-01
-7.15718687e-01 5.74137628e-01 1.15643412e-01 3.16893458e-01
1.04000099e-01 -5.71008623e-01 -4.54939127e-01 -5.89502454e-01
-5.73859662e-02 6.59867346e-01 -1.58264190e-01 1.20960660e-01
-9.32423119e-03 2.25445151e-01 -2.06798100e+00 8.22200328e-02
9.57521081e-01 1.26681399e+00 -4.17405605e-01 6.91122830e-01
3.99379313e-01 2.54350185e-01 -6.67419672e-01 -1.18836892e+00
2.65867889e-01 2.70394444e-01 -5.55980504e-01 -7.48822559e-03
1.35221148e+00 -5.50242245e-01 -7.35841244e-02 8.65802050e-01
8.05773854e-01 -2.80424058e-01 8.30838025e-01 -1.58118427e+00
-6.20025754e-01 3.54562283e-01 -6.84149861e-01 -2.54576027e-01
6.53762579e-01 2.02597678e-01 7.35820770e-01 -9.80537653e-01
5.80639362e-01 1.53625512e+00 7.04526722e-01 7.53135920e-01
-1.13760304e+00 -8.58648181e-01 1.55345365e-01 2.25737512e-01
-1.39598644e+00 -7.14952588e-01 3.69666606e-01 -4.58353370e-01
5.42121112e-01 3.54022115e-01 5.87490499e-01 1.60932672e+00
-2.72149235e-01 7.49631584e-01 1.41201949e+00 -3.12684715e-01
8.74791667e-02 2.26118460e-01 1.53774261e-01 8.52202654e-01
2.80556440e-01 2.86573339e-02 -6.70543373e-01 -1.58745095e-01
2.99355656e-01 -6.09832518e-02 1.62037797e-02 -2.97641575e-01
-9.12790358e-01 4.43814844e-01 1.62094012e-01 -1.78473130e-01
-1.18565209e-01 -6.68590963e-02 2.27423266e-01 9.11387652e-02
4.70084488e-01 1.00756571e-01 -3.32413405e-01 -1.18602403e-01
-1.02709174e+00 5.06131947e-01 8.55996847e-01 6.17149055e-01
4.50431764e-01 -2.34356686e-01 -5.07225513e-01 1.02153361e+00
1.80657789e-01 8.33908975e-01 1.82743251e-01 -1.19639850e+00
3.27813059e-01 7.50250936e-01 1.36244684e-01 -9.15510595e-01
-3.83845866e-01 -4.55428779e-01 -8.28317642e-01 4.03535604e-01
8.97274256e-01 5.27428277e-02 -1.11079597e+00 2.03259349e+00
3.56394917e-01 -1.19232178e-01 -3.22373062e-01 7.69184232e-01
1.02384317e+00 -9.49311629e-03 2.96876788e-01 -8.91497359e-02
1.62005329e+00 -7.25585938e-01 -4.87693906e-01 -3.34591925e-01
1.71663910e-01 -5.62897205e-01 8.18216145e-01 6.00972712e-01
-1.13489795e+00 -5.33544362e-01 -8.48002732e-01 3.75932939e-02
-5.95788181e-01 5.09733021e-01 6.59143746e-01 1.08052969e+00
-1.10379219e+00 3.81118923e-01 -4.92099136e-01 -6.20302141e-01
6.69492483e-01 6.92750692e-01 -8.79163325e-01 -2.15408176e-01
-8.65988672e-01 7.26070344e-01 -1.85144730e-02 1.01997003e-01
-6.66996837e-01 -7.65881538e-01 -9.24788594e-01 -1.33616835e-01
2.41465047e-01 -7.87975311e-01 1.09147465e+00 -8.98440182e-01
-9.61094737e-01 1.48187232e+00 -2.01932266e-01 -2.42285207e-01
9.22485352e-01 -3.41277599e-01 -3.87998849e-01 5.84377013e-02
8.53353813e-02 1.14952612e+00 9.23699021e-01 -1.05010760e+00
-5.54362237e-01 -9.81913388e-01 9.21740830e-02 -2.64625661e-02
-6.81463897e-01 2.38064095e-01 -6.98094070e-01 -4.65956718e-01
-3.76440138e-01 -9.10334587e-01 1.07587487e-01 2.96149731e-01
-1.15591444e-01 -2.42636502e-01 4.01267618e-01 -9.15725350e-01
1.30206394e+00 -2.19049168e+00 1.12468101e-01 1.28979057e-01
4.27405030e-01 1.38879538e-01 -2.30416134e-02 1.17153428e-01
3.99357453e-02 -5.31975441e-02 2.45897975e-02 -9.28958416e-01
2.46590763e-01 9.26040486e-03 2.17387572e-01 3.59987020e-01
2.03736592e-02 7.66314209e-01 -5.69921196e-01 -7.38098443e-01
6.51296005e-02 5.30611694e-01 -4.36849177e-01 1.79158613e-01
2.01175675e-01 3.71566772e-01 1.02658264e-01 1.21878231e+00
8.70176733e-01 1.59405246e-01 -2.94063259e-02 -1.80580005e-01
1.73841581e-01 -5.61562300e-01 -1.22471416e+00 1.39260888e+00
-4.17914182e-01 1.94002807e-01 2.96209455e-01 -4.69464958e-01
9.10289049e-01 1.26942262e-01 2.41000727e-01 -7.41244912e-01
2.55088478e-01 2.41787404e-01 -1.97962403e-01 -3.93081307e-01
4.96431231e-01 3.76700133e-01 -1.96405753e-01 1.21838927e-01
3.66063982e-01 1.58281118e-01 4.79783773e-01 1.88049495e-01
8.29250097e-01 -5.25084510e-03 7.88238868e-02 -4.39416915e-02
5.32301247e-01 -6.18096888e-01 5.69995999e-01 7.30621040e-01
-5.38210511e-01 8.56806099e-01 5.77428758e-01 -5.18432558e-01
-9.43273008e-01 -1.26403260e+00 -2.85439670e-01 1.27825797e+00
-2.92313099e-01 -5.02299547e-01 -1.00613201e+00 -7.05029666e-01
2.44166061e-01 2.23135296e-02 -1.22664249e+00 1.76826626e-01
-1.49926081e-01 -8.10501575e-01 6.98411047e-01 5.76845467e-01
4.65299368e-01 -7.92820573e-01 -3.85343134e-01 -3.80291104e-01
-1.38306767e-01 -1.40719378e+00 -4.15485561e-01 -4.58987862e-01
-5.34486353e-01 -1.35538387e+00 -9.89633858e-01 -4.92352694e-01
8.49214137e-01 -3.63822252e-01 1.32697940e+00 1.23144858e-01
-6.04751348e-01 6.70094132e-01 -5.41705452e-02 -6.84688389e-01
5.59221162e-03 2.21909672e-01 2.56630182e-01 3.32811534e-01
4.71882463e-01 -5.36602437e-01 -8.39797258e-01 3.51526797e-01
-4.96447414e-01 -8.19665045e-02 4.04517025e-01 6.43400013e-01
1.66302249e-01 -5.08241057e-01 4.93856251e-01 -8.24289024e-01
3.64439309e-01 -2.09988594e-01 -5.71623027e-01 2.37818763e-01
-6.76553786e-01 -1.54464319e-01 1.04592241e-01 -3.78061533e-01
-8.29746962e-01 2.81550199e-01 -1.22947603e-01 -3.04582715e-01
-2.52939373e-01 -1.12120390e-01 -2.98748493e-01 4.78789322e-02
6.05226278e-01 -2.46516630e-01 2.81019241e-01 -5.07054925e-01
1.97835892e-01 8.48728895e-01 9.54994798e-01 -7.18153894e-01
5.94748259e-01 5.29487252e-01 -4.80281524e-02 -5.69999397e-01
-8.88795555e-01 -1.89646393e-01 -7.59996235e-01 -5.54487884e-01
6.54407918e-01 -1.16883028e+00 -1.17437494e+00 9.83099937e-01
-8.60241473e-01 6.32015383e-03 1.07568227e-01 2.72951484e-01
-2.21845344e-01 7.97346309e-02 -5.17637193e-01 -1.02254057e+00
-5.74765265e-01 -1.01281512e+00 1.35423052e+00 5.34015536e-01
-4.71588880e-01 -7.99951911e-01 -3.17317486e-01 7.28810728e-01
2.59503692e-01 6.11138523e-01 4.24733758e-01 -4.70322847e-01
-1.07990719e-01 -3.89706373e-01 -4.43560749e-01 2.33883098e-01
-2.32261434e-01 3.22805703e-01 -1.44458640e+00 -3.28401148e-01
-6.97977066e-01 -6.29292786e-01 1.02627361e+00 2.15902239e-01
1.26108742e+00 5.67488968e-02 -2.38617614e-01 4.73332703e-01
1.08159888e+00 -5.88242531e-01 5.16537428e-01 1.19854778e-01
5.83247304e-01 1.09997261e+00 6.03758872e-01 6.10115886e-01
7.85573542e-01 7.07757890e-01 6.35113776e-01 -1.92675784e-01
-7.26411119e-02 -8.38492885e-02 2.42696613e-01 2.56985664e-01
-6.13649786e-01 2.15554327e-01 -9.09215212e-01 5.35008132e-01
-1.74912834e+00 -8.21979702e-01 -3.37695181e-01 2.38415909e+00
7.90752769e-01 1.49017461e-02 5.24963975e-01 2.57605821e-01
5.40686846e-01 5.51010184e-02 -2.53356636e-01 -5.24012506e-01
-1.05252028e-01 4.51263845e-01 3.62898290e-01 3.50590259e-01
-1.20693040e+00 5.93959212e-01 7.05997419e+00 7.13777542e-01
-9.85253513e-01 8.95032659e-02 8.75294507e-01 -4.03094769e-01
1.44679382e-01 -5.75997531e-01 -9.21150446e-01 6.11302495e-01
7.17483878e-01 7.59861395e-02 3.75303715e-01 8.02684665e-01
-1.70553699e-01 -4.14481223e-01 -1.32124424e+00 1.31590426e+00
4.19811070e-01 -6.63092494e-01 -3.99916202e-01 2.16918692e-01
5.61414063e-01 -3.97611499e-01 3.87806088e-01 4.19139057e-01
9.66688618e-02 -1.38623214e+00 8.84245515e-01 5.91469526e-01
1.15310860e+00 -7.11926699e-01 8.50481689e-01 1.69112664e-02
-9.78224337e-01 -2.96726048e-01 -1.31915566e-02 -2.55212069e-01
-3.61488163e-01 7.17376769e-01 -3.96442771e-01 4.92140800e-01
1.08778608e+00 4.89679158e-01 -1.35718858e+00 9.31893349e-01
-1.79835096e-01 2.09740356e-01 -3.72431606e-01 2.67448992e-01
-4.31784213e-01 1.87474862e-01 -3.89091186e-02 1.03164375e+00
2.57782847e-01 -2.11178422e-01 4.80352454e-02 5.28102934e-01
-2.10473984e-01 1.11251965e-01 -4.08672094e-01 3.31574380e-01
4.46042031e-01 1.58389091e+00 -3.09948832e-01 -1.60914943e-01
-4.53904688e-01 8.41803908e-01 4.78020281e-01 1.98690504e-01
-7.00854659e-01 -1.72535464e-01 7.75563419e-01 1.56380072e-01
8.06664303e-02 -5.58330119e-02 -2.56743431e-01 -9.56591606e-01
2.58925557e-01 -9.00422990e-01 6.18957520e-01 -6.74001873e-01
-1.32926261e+00 3.72246891e-01 1.21066801e-01 -8.15840364e-01
-3.40971798e-01 -7.89712727e-01 -4.15294141e-01 8.25639129e-01
-1.20088792e+00 -1.62890434e+00 -7.64098406e-01 5.68904042e-01
7.99979940e-02 -2.73682743e-01 9.86900449e-01 7.71738946e-01
-7.02622592e-01 1.18949318e+00 -3.68498713e-01 2.32057288e-01
1.21762311e+00 -1.31782043e+00 1.76494077e-01 5.86997628e-01
-2.26282835e-01 5.02682328e-01 7.27030694e-01 -3.94759655e-01
-1.16301692e+00 -9.01567459e-01 9.27604675e-01 -8.96479309e-01
3.72928619e-01 -8.06678534e-01 -5.51048279e-01 5.19717038e-01
2.13818159e-02 1.99628815e-01 6.66756153e-01 5.93254924e-01
-7.36365438e-01 -4.17039514e-01 -1.36935520e+00 4.54944372e-01
1.22760355e+00 -2.58816838e-01 -3.51102501e-01 -5.15272468e-02
1.43374965e-01 -6.14016771e-01 -9.61314917e-01 6.40091956e-01
1.26424515e+00 -1.29468071e+00 9.66389477e-01 -2.36360103e-01
6.55336261e-01 -1.26671372e-02 -3.17211449e-02 -9.14381087e-01
-8.10446159e-04 -9.63979512e-02 -3.86745572e-01 1.79737937e+00
2.39167839e-01 -5.16740263e-01 7.76252806e-01 9.33146775e-01
5.00563264e-01 -9.49780345e-01 -7.72248268e-01 -5.84614575e-01
-1.12888448e-01 -3.57595444e-01 7.67525971e-01 6.06295705e-01
-4.38547611e-01 -1.16646904e-02 -4.19934124e-01 1.68825641e-01
9.31352735e-01 -9.77420658e-02 1.01255274e+00 -1.46464670e+00
-7.73126557e-02 -3.41564387e-01 -6.06396437e-01 -1.93349078e-01
2.13687956e-01 -5.90169489e-01 -3.49166006e-01 -1.24518025e+00
5.27039826e-01 -3.10082704e-01 -1.48481905e-01 6.07930481e-01
-3.34397256e-01 9.78670061e-01 2.68554330e-01 -7.45407343e-02
-6.43162727e-01 3.04483891e-01 9.28644478e-01 -2.24140674e-01
2.87910283e-01 4.19988949e-03 -7.95248508e-01 7.25323856e-01
5.37090421e-01 -2.32347175e-01 -6.01086840e-02 -3.31687510e-01
2.24726066e-01 -3.42654765e-01 6.28880918e-01 -1.13944483e+00
5.39521575e-02 9.03531015e-02 9.39746797e-01 -3.70060146e-01
7.22392619e-01 -5.84753811e-01 8.25683028e-02 2.74980247e-01
6.10622913e-02 -8.00794661e-02 -1.11936992e-02 7.06490427e-02
-1.26138538e-01 6.12128936e-02 7.73465037e-01 5.08137308e-02
-4.67907399e-01 5.64715266e-01 2.47205421e-01 5.43511696e-02
9.60879326e-01 -3.66928786e-01 -5.16510844e-01 -2.90118873e-01
-8.96239996e-01 6.26109838e-01 7.87564874e-01 6.68033361e-01
1.60764188e-01 -1.32794070e+00 -9.34943438e-01 2.78912485e-01
4.68880504e-01 -4.46469337e-01 2.41807595e-01 8.62410486e-01
-2.03280136e-01 -8.44196416e-03 -4.74036366e-01 -7.07130194e-01
-1.88752806e+00 3.87187928e-01 2.07518980e-01 -1.40281498e-01
-1.99052133e-02 6.97742403e-01 3.56974676e-02 -6.81149244e-01
5.47636986e-01 6.70246258e-02 -4.32278961e-01 3.63042772e-01
7.77091742e-01 4.44837272e-01 4.41138223e-02 -6.06133640e-01
-4.92202073e-01 6.45135522e-01 -4.19055820e-02 -9.93200913e-02
1.23511505e+00 -5.64367846e-02 -1.19157836e-01 2.81532347e-01
7.47933924e-01 2.25607127e-01 -1.04819787e+00 -3.38457264e-02
-3.22335660e-01 -7.37999499e-01 -4.71315682e-01 -9.80980873e-01
-1.15846443e+00 8.62108529e-01 9.82323825e-01 -9.18512344e-02
1.12727594e+00 3.27651873e-02 3.89289767e-01 -6.20402284e-02
4.35562700e-01 -1.33308089e+00 9.43035483e-02 2.79315025e-01
7.48823345e-01 -1.64328742e+00 3.20947617e-01 -5.22165835e-01
-3.72673273e-01 7.17050970e-01 9.39104080e-01 3.05463791e-01
3.58100414e-01 1.07669473e-01 2.21214145e-01 8.24985057e-02
-6.55146956e-01 -2.18902588e-01 4.01819587e-01 9.25983906e-01
4.43452537e-01 1.91360101e-01 -4.26952541e-01 9.76830482e-01
-6.08939171e-01 2.56061908e-02 2.31462687e-01 4.65412080e-01
1.39975892e-02 -1.29289639e+00 -6.16312444e-01 6.14003181e-01
-7.27722049e-01 2.85784721e-01 -5.41639149e-01 6.70933783e-01
5.37088931e-01 8.21269691e-01 -2.08034660e-04 -2.72883058e-01
3.24215084e-01 9.52515155e-02 8.93229425e-01 -3.26979816e-01
-6.92744613e-01 -4.99894351e-01 2.56338865e-01 -6.58208132e-01
-2.53010958e-01 -8.08344364e-01 -7.80776024e-01 -5.33238947e-01
1.72574162e-01 -2.33769342e-01 8.42966855e-01 7.17577577e-01
3.31802249e-01 2.16904610e-01 4.93081421e-01 -8.12795818e-01
-3.39657336e-01 -1.04709399e+00 -6.09461010e-01 6.32431269e-01
1.33962473e-02 -8.51378858e-01 -1.81556687e-01 1.13689773e-01]
|
[13.53357982635498, 1.0014959573745728]
|
f26e9eb3-fb5b-4013-8ae3-236c0251831a
|
latent-space-semi-supervised-time-series-data
| null | null |
https://openreview.net/forum?id=0qbEq5UBfGD
|
https://openreview.net/pdf?id=0qbEq5UBfGD
|
Latent Space Semi-Supervised Time Series Data Clustering
|
Time series data is abundantly available in the real world, but there is a distinct lack of large, labeled datasets available for many types of learning tasks. Semi-supervised models, which can leverage small amounts of expert-labeled data along with a larger unlabeled dataset, have been shown to improve performance over unsupervised learning models. Existing semi-supervised time series clustering algorithms suffer from lack of scalability as they are limited to perform learning operations within the original data space. We propose an autoencoder-based semi-supervised learning model along with multiple semi-supervised objective functions which can be used to improve the quality of the autoencoder’s learned latent space via the addition of a small number of labeled examples. Experiments show that our methods can consistently improve k-Means clustering performance on a variety of datasets. Our methods achieve a maximum average ARI of 0.897, a 140% increase over an unsupervised CAE model. Our methods also achieve a maximum improvement of 44% over a semi-supervised model.
|
['Farnoush Kashani', 'Russell Bowler', 'Katerina Kechris', 'Andrew Hill']
|
2021-01-01
| null | null | null | null |
['time-series-clustering']
|
['time-series']
|
[-1.43172428e-01 -2.24580362e-01 -3.97127062e-01 -5.81767440e-01
-9.54581738e-01 -6.71248794e-01 6.61927342e-01 5.77173531e-02
-3.02935869e-01 3.31636637e-01 2.73122013e-01 -1.27402797e-01
-2.14650199e-01 -5.20722687e-01 -5.35926700e-01 -7.66128719e-01
-2.65307069e-01 6.19971812e-01 -7.65980035e-02 1.45461336e-01
-1.06897861e-01 7.17842206e-02 -1.48311496e+00 1.95348769e-01
7.57825136e-01 1.07643056e+00 -1.19508453e-01 3.23235333e-01
-3.89682464e-02 1.06113327e+00 -3.73697579e-01 3.45434934e-01
2.35127285e-01 -3.03902209e-01 -6.65678978e-01 4.81642812e-01
-1.85543716e-01 -2.08233863e-01 -2.82137066e-01 6.57975674e-01
1.97634876e-01 2.78821528e-01 7.02113569e-01 -1.51713884e+00
-4.61169392e-01 7.42068291e-01 -4.93301302e-01 1.04274035e-01
-1.73556805e-01 -1.62145942e-01 1.06111848e+00 -7.26073742e-01
4.05261487e-01 6.88066125e-01 8.18099439e-01 2.00321630e-01
-1.40716827e+00 -7.08646715e-01 -1.24781832e-01 7.51966387e-02
-1.35089445e+00 -3.70620191e-01 1.04739082e+00 -4.82090473e-01
1.20882177e+00 -1.08591162e-01 5.46041250e-01 7.49316514e-01
-2.91047931e-01 7.30336010e-01 1.31368613e+00 -5.71946144e-01
7.11019158e-01 1.35235965e-01 3.85751069e-01 2.77802110e-01
-1.81018621e-01 1.26479760e-01 -3.57202202e-01 -3.04541916e-01
4.41526353e-01 5.29208243e-01 1.79843023e-01 -4.65034217e-01
-1.14357305e+00 1.02711308e+00 1.61022782e-01 4.36507136e-01
-4.93862987e-01 -9.77652594e-02 3.70776176e-01 3.72275352e-01
7.89745152e-01 5.40734410e-01 -7.21512020e-01 -3.51178408e-01
-1.21456802e+00 -2.01829657e-01 7.48412907e-01 8.13531876e-01
9.86835361e-01 3.43462825e-01 3.95354688e-01 8.48433673e-01
4.78699543e-02 4.69408840e-01 9.14254189e-01 -1.18691039e+00
1.79193437e-01 9.01855290e-01 -8.69137719e-02 -6.50797427e-01
-4.16645288e-01 -2.78919518e-01 -6.14853501e-01 -1.29509345e-02
2.49053806e-01 -3.60465735e-01 -9.15017247e-01 1.57179308e+00
3.05136293e-01 3.83751750e-01 2.36437261e-01 4.72561151e-01
2.49253899e-01 9.71630275e-01 -1.85422584e-01 -6.29794896e-01
7.60933876e-01 -1.16857719e+00 -7.52370834e-01 -3.85938808e-02
8.32465112e-01 -4.47299898e-01 9.34043407e-01 4.31876272e-01
-6.76514864e-01 -4.27566677e-01 -8.97738695e-01 4.23923880e-01
-5.66893697e-01 1.33818910e-01 7.53693998e-01 4.97165710e-01
-8.74209166e-01 6.50422871e-01 -1.42272258e+00 -3.55829477e-01
4.49215144e-01 3.98824632e-01 -3.98482293e-01 6.99850544e-02
-8.23300540e-01 4.46589053e-01 6.66736603e-01 -3.60426217e-01
-8.63234818e-01 -7.40642965e-01 -8.10798824e-01 9.98801589e-02
4.44245577e-01 3.21750753e-02 1.21703696e+00 -1.10379267e+00
-1.18998814e+00 4.59489673e-01 -7.58797228e-02 -7.66803384e-01
-3.10903881e-02 -1.09034047e-01 -7.63353944e-01 3.98307085e-01
1.22137420e-01 6.04001999e-01 9.19413447e-01 -1.21756005e+00
-5.60972452e-01 -3.67938548e-01 -3.96500081e-01 6.69205487e-02
-1.03877246e+00 -1.58035103e-02 -3.25427413e-01 -8.97913873e-01
1.47104353e-01 -1.16979241e+00 -4.13573772e-01 -4.59042996e-01
-1.93220958e-01 -3.49664688e-01 1.21154308e+00 -4.85307127e-01
1.40714788e+00 -2.26586008e+00 -2.79724509e-01 4.32902813e-01
3.06258559e-01 1.98758263e-02 1.68917194e-01 6.12823665e-01
-3.81611794e-01 -4.80093658e-02 -3.47391129e-01 -3.79450351e-01
2.35809479e-02 4.15944487e-01 -4.56738472e-01 3.86501670e-01
-1.96268759e-03 6.85152411e-01 -9.74326551e-01 -3.70071173e-01
3.17008257e-01 2.90288419e-01 -3.86017323e-01 2.59619176e-01
-5.87925836e-02 2.29485050e-01 -9.42788646e-02 3.47723335e-01
9.03908834e-02 -8.46803546e-01 2.74676204e-01 7.87727535e-02
7.02802613e-02 1.90703407e-01 -1.16697311e+00 1.62827671e+00
-1.27621219e-01 6.39359176e-01 -4.45195377e-01 -1.18204522e+00
7.25977421e-01 5.40975869e-01 1.16283131e+00 -2.75480032e-01
8.29240605e-02 7.48632401e-02 -8.70242193e-02 -1.94083184e-01
2.14949623e-01 -1.25329956e-01 -2.92944722e-03 1.14497221e+00
2.56326377e-01 5.61794229e-02 3.24847698e-01 2.36449376e-01
1.13588738e+00 -1.29379705e-01 1.21301576e-01 -2.57710367e-01
-2.92228232e-03 3.87802660e-01 7.07307398e-01 4.54274058e-01
-1.37290031e-01 5.15362620e-01 5.61353192e-02 -6.31139278e-01
-1.25057638e+00 -8.17527592e-01 -1.05168536e-01 1.10877526e+00
-3.69644046e-01 -7.99619675e-01 -6.97358072e-01 -6.83778942e-01
-8.57730657e-02 5.14708459e-01 -6.98363185e-01 -1.82842955e-01
-2.26466179e-01 -9.13410008e-01 4.87635493e-01 1.09670889e+00
1.83944896e-01 -9.18632925e-01 -4.70759809e-01 2.61334240e-01
-1.10409237e-01 -1.26372993e+00 -2.64638990e-01 6.16249263e-01
-1.30118656e+00 -8.67081046e-01 -3.49914014e-01 -7.60164917e-01
8.43464196e-01 3.26552659e-01 9.40381706e-01 -2.98067898e-01
4.47343253e-02 4.98247802e-01 -6.09010041e-01 -4.75633830e-01
-3.65787029e-01 3.74316573e-02 5.86531937e-01 8.97572339e-02
8.74056041e-01 -7.81651974e-01 -3.85322690e-01 3.89764309e-01
-9.90386784e-01 -2.10212082e-01 9.06640887e-02 9.40514982e-01
6.84544444e-01 6.99535072e-01 7.76170790e-01 -8.91897142e-01
3.86191547e-01 -5.85666060e-01 -4.93820220e-01 6.31274804e-02
-1.27496994e+00 1.26507044e-01 8.21168959e-01 -7.29009688e-01
-9.09271598e-01 3.28030318e-01 4.10849094e-01 -9.20660913e-01
-3.73977572e-01 7.88812637e-01 3.90390694e-01 3.40896755e-01
7.66043603e-01 2.47504547e-01 1.24283455e-01 -4.89217818e-01
3.52351904e-01 8.91903818e-01 4.44552392e-01 -4.89744663e-01
9.23482180e-01 7.98845351e-01 -5.13251245e-01 -5.98295569e-01
-9.25368309e-01 -9.02480006e-01 -1.01308441e+00 -1.02296241e-01
7.03215480e-01 -1.24163640e+00 -2.14771971e-01 3.79819661e-01
-2.23426968e-01 -7.19548583e-01 -5.32843351e-01 8.91241014e-01
-5.49052775e-01 3.32010657e-01 -6.68815136e-01 -7.68279195e-01
-2.33087629e-01 -6.39523089e-01 8.45782340e-01 -4.43631113e-02
-4.68441129e-01 -1.25703216e+00 2.72283137e-01 3.26192856e-01
1.67701378e-01 1.96953192e-01 7.51235604e-01 -1.20208359e+00
1.77136399e-02 -3.73952180e-01 3.10264349e-01 4.47746992e-01
4.60352778e-01 -4.62798849e-02 -1.06532264e+00 -5.49649239e-01
1.55065089e-01 -6.42959058e-01 7.18136847e-01 3.92211318e-01
1.18679059e+00 -2.23008662e-01 -2.99202561e-01 3.37858200e-01
1.27686179e+00 4.63767469e-01 1.45848244e-01 1.50672838e-01
6.89298570e-01 4.77827102e-01 5.04336476e-01 7.05110729e-01
3.95510286e-01 1.82018340e-01 -1.24054417e-01 -1.12219438e-01
3.97648752e-01 -2.26414278e-01 3.37388724e-01 1.61432517e+00
6.37475401e-02 1.76112682e-01 -1.34913826e+00 8.53445292e-01
-2.02118826e+00 -1.05289257e+00 6.56562112e-03 2.05897355e+00
9.25146401e-01 1.60152987e-01 4.96700406e-01 7.20607817e-01
5.34701586e-01 2.39578579e-02 -7.74103224e-01 2.21934646e-01
-9.46708117e-03 6.94789216e-02 3.41757119e-01 8.98759160e-03
-1.37718189e+00 7.63592482e-01 7.42053032e+00 6.25214636e-01
-1.09173739e+00 6.08139597e-02 4.06161636e-01 -1.65770933e-01
2.75975689e-02 8.93419683e-02 -3.21165562e-01 5.06239235e-01
1.56973708e+00 -2.36506283e-01 5.88504434e-01 1.05710471e+00
2.49669492e-01 2.06148759e-01 -1.14015734e+00 1.18918204e+00
1.31113783e-01 -1.22926283e+00 -3.67835164e-01 1.28463447e-01
1.25833833e+00 2.05485985e-01 -3.75819877e-02 4.73062754e-01
8.29789639e-01 -8.03729117e-01 3.23395491e-01 1.37591824e-01
4.15687263e-01 -7.79137194e-01 4.09960955e-01 5.92735171e-01
-1.16407847e+00 -2.94259578e-01 -3.89058031e-02 -3.02016258e-01
-9.03500523e-03 6.48679614e-01 -6.92592323e-01 2.50638366e-01
9.85265255e-01 1.17149734e+00 -4.83899504e-01 6.56111836e-01
-4.62774783e-02 1.32863402e+00 -4.99490380e-01 2.78687507e-01
3.88312101e-01 -1.46460757e-01 8.49735215e-02 8.52990150e-01
8.91290978e-02 1.91308752e-01 6.01713002e-01 5.95952094e-01
7.99214765e-02 1.09099343e-01 -6.05403721e-01 -5.88099539e-01
6.83660328e-01 1.13644135e+00 -9.05657828e-01 -6.53929591e-01
-5.89723051e-01 7.60144234e-01 1.44756809e-01 5.17160892e-01
-6.47182584e-01 -2.89307952e-01 1.35464743e-01 -2.11531594e-01
4.26586956e-01 -4.11463022e-01 -1.76041856e-01 -1.34740245e+00
-1.18739277e-01 -1.03368807e+00 7.94686556e-01 -8.68734002e-01
-1.56072414e+00 4.23111528e-01 -3.17094009e-03 -1.62889087e+00
-7.15290546e-01 -3.78063947e-01 -4.12587583e-01 2.87418187e-01
-1.03489208e+00 -1.01032233e+00 -1.81553438e-01 8.64622235e-01
5.96546471e-01 -4.84086812e-01 1.04377592e+00 2.54811645e-01
-6.12319708e-01 3.53135884e-01 7.87032425e-01 5.80854535e-01
8.14238727e-01 -1.50802803e+00 2.92851981e-02 9.00165260e-01
5.16719699e-01 7.97587931e-01 3.98732901e-01 -5.34031808e-01
-1.36658978e+00 -1.23681676e+00 6.02005661e-01 -5.13787210e-01
9.05361891e-01 -1.86450899e-01 -9.20051455e-01 1.15540671e+00
1.79975644e-01 1.15574017e-01 1.39841855e+00 4.59858596e-01
-6.54721141e-01 -2.43197545e-01 -7.89990664e-01 2.71039724e-01
5.52134156e-01 -1.00805593e+00 -8.20812345e-01 4.49746609e-01
7.23143458e-01 3.19159552e-02 -1.34621310e+00 2.50960559e-01
3.36986363e-01 -5.93313515e-01 7.95131445e-01 -7.60294139e-01
4.45849836e-01 -3.98236126e-01 -1.30591944e-01 -1.39279532e+00
-2.69868791e-01 -4.97080833e-01 -5.91758907e-01 1.16652071e+00
4.01859254e-01 -4.08461094e-01 7.84375072e-01 7.98749626e-01
1.02025844e-01 -3.69951755e-01 -3.95422727e-01 -1.09154713e+00
-1.91197603e-03 -6.86278820e-01 4.42724973e-01 1.66922724e+00
5.30858457e-01 3.59738588e-01 -3.89978647e-01 5.08802272e-02
8.47842991e-01 4.82052356e-01 7.49158084e-01 -1.45163608e+00
-2.37692103e-01 -1.74952731e-01 -2.30122596e-01 -6.67707860e-01
2.94721186e-01 -8.74856889e-01 -9.89164412e-03 -1.25198662e+00
2.96540439e-01 -3.78123134e-01 -7.25592852e-01 8.60262573e-01
-3.31746712e-02 2.06601351e-01 -2.32828468e-01 7.36074090e-01
-8.90736222e-01 5.50820053e-01 3.83952290e-01 -7.69208595e-02
-5.84594786e-01 -2.53545761e-01 -4.30616468e-01 8.27879369e-01
8.64970922e-01 -6.32688403e-01 -7.95565605e-01 -2.88397104e-01
-2.10880712e-01 -4.66543064e-02 -7.00908229e-02 -1.17463756e+00
4.89034265e-01 -1.61374807e-01 4.44480985e-01 -6.80779696e-01
1.28353044e-01 -1.00159335e+00 1.73259303e-01 1.68461472e-01
-5.38231969e-01 1.70952559e-01 6.91073462e-02 8.15096200e-01
-4.94726926e-01 1.60914421e-01 5.60385287e-01 5.24285622e-03
-8.40039611e-01 2.72162914e-01 -4.69364852e-01 7.09466040e-02
1.02322221e+00 -1.46081761e-01 4.25315201e-02 -6.02658331e-01
-8.83191168e-01 2.92270422e-01 3.82819533e-01 4.19412613e-01
2.62047440e-01 -1.61097395e+00 -2.80219704e-01 1.63393050e-01
3.76474500e-01 -4.53433432e-02 4.82045449e-02 6.94945693e-01
2.45759916e-02 3.57991636e-01 -2.54049659e-01 -7.80279577e-01
-9.07665372e-01 9.18488801e-01 4.58708545e-03 -2.91754246e-01
-7.20563889e-01 1.88453108e-01 -2.49372929e-01 -7.47874022e-01
3.40536922e-01 -2.26279125e-01 -8.37227181e-02 6.58950880e-02
5.89206755e-01 3.98151129e-01 6.51677772e-02 -5.71181953e-01
-1.83911666e-01 3.67568523e-01 -7.42369890e-02 -3.08650374e-01
1.86278152e+00 -7.88355991e-02 1.19880565e-01 9.94143903e-01
1.33614528e+00 -3.36681187e-01 -1.36979544e+00 -6.71407938e-01
1.92646146e-01 -9.51586217e-02 2.33476087e-01 -5.10723472e-01
-1.00101268e+00 7.20299482e-01 4.68788266e-01 3.90032738e-01
1.40220201e+00 2.89827473e-02 7.33159423e-01 7.38712609e-01
3.22542280e-01 -1.57991445e+00 3.77937615e-01 3.47813398e-01
1.16240725e-01 -1.54706490e+00 -5.13262637e-02 3.00281271e-02
-8.63367379e-01 9.49008882e-01 4.04761285e-01 -2.28216544e-01
1.03548884e+00 4.97980058e-01 3.78238410e-01 -3.61694455e-01
-1.13237560e+00 -2.32750386e-01 2.69428045e-01 4.77320075e-01
3.02482873e-01 5.97882532e-02 3.37220728e-01 7.47902095e-01
-9.00222212e-02 7.15670288e-02 2.68692523e-01 1.15397322e+00
-3.12688321e-01 -8.69815767e-01 -1.78719744e-01 7.05477178e-01
-3.83739948e-01 1.44749746e-01 -3.15864146e-01 6.02326632e-01
-3.97699326e-01 1.35153306e+00 2.33948842e-01 -5.30740201e-01
-1.41641319e-01 5.11528969e-01 -1.61416352e-01 -6.90361261e-01
-4.26725179e-01 6.09098554e-01 -2.84750015e-01 -5.21611571e-01
-9.90623295e-01 -7.69829512e-01 -1.38104606e+00 -2.11902604e-01
-4.49846357e-01 4.94781882e-01 4.53838021e-01 1.14056921e+00
3.78133118e-01 3.02114695e-01 1.10345697e+00 -4.28677112e-01
-5.78773141e-01 -9.07666445e-01 -8.18233669e-01 8.36541891e-01
1.91574812e-01 -4.56128061e-01 -4.72503752e-01 7.60784149e-01]
|
[7.4112653732299805, 2.883622884750366]
|
ed528c1c-0921-48dd-8b2c-ee21eb6cd4ad
|
security-consideration-for-deep-learning
|
1803.11157
| null |
http://arxiv.org/abs/1803.11157v2
|
http://arxiv.org/pdf/1803.11157v2.pdf
|
Security Consideration For Deep Learning-Based Image Forensics
|
Recently, image forensics community has paied attention to the research on
the design of effective algorithms based on deep learning technology and facts
proved that combining the domain knowledge of image forensics and deep learning
would achieve more robust and better performance than the traditional schemes.
Instead of improving it, in this paper, the safety of deep learning based
methods in the field of image forensics is taken into account. To the best of
our knowledge, this is a first work focusing on this topic. Specifically, we
experimentally find that the method using deep learning would fail when adding
the slight noise into the images (adversarial images). Furthermore, two kinds
of strategys are proposed to enforce security of deep learning-based method.
Firstly, an extra penalty term to the loss function is added, which is referred
to the 2-norm of the gradient of the loss with respect to the input images, and
then an novel training method are adopt to train the model by fusing the normal
and adversarial images. Experimental results show that the proposed algorithm
can achieve good performance even in the case of adversarial images and provide
a safety consideration for deep learning-based image forensics
|
['Rongrong Ni', 'Pengpeng Yang', 'Yao Zhao', 'Wei Zhao', 'Haorui Wu']
|
2018-03-29
| null | null | null | null |
['image-forensics']
|
['computer-vision']
|
[-2.92144027e-02 -9.75949913e-02 1.95706934e-01 -1.23911098e-01
-3.17901224e-01 -1.98745847e-01 4.35744703e-01 -8.05215985e-02
-6.19191110e-01 4.27543849e-01 -1.81701198e-01 -4.93555456e-01
1.96369155e-03 -9.19541657e-01 -7.91503489e-01 -8.51426542e-01
7.58019760e-02 -1.84293464e-01 1.48360774e-01 -1.80582460e-02
4.42174524e-01 6.29676282e-01 -9.43189263e-01 3.80499437e-02
7.31944263e-01 9.93628144e-01 -6.63055032e-02 1.39077291e-01
-1.74885139e-01 9.13487673e-01 -6.94781721e-01 -8.54606032e-01
6.84089780e-01 -2.45942965e-01 -8.60361874e-01 1.71401560e-01
1.81780130e-01 -7.96855807e-01 -7.86739409e-01 1.40490377e+00
6.95940256e-01 8.80411193e-02 4.14647460e-01 -1.31362200e+00
-7.06171036e-01 1.47232056e-01 -8.83873880e-01 4.75922346e-01
-1.79730617e-02 4.45094526e-01 3.00326765e-01 -4.52042818e-01
3.23758900e-01 1.17943668e+00 4.96935546e-01 4.18563068e-01
-5.64969540e-01 -8.59075189e-01 -5.07974327e-02 7.67143130e-01
-1.11984682e+00 -2.12149411e-01 1.07636046e+00 -3.06010753e-01
2.81820655e-01 -8.19453299e-02 2.29872108e-01 9.24026012e-01
2.08932087e-01 9.25172210e-01 1.24987447e+00 -4.86835152e-01
-1.00512002e-02 3.20912063e-01 -3.06131262e-02 5.60553372e-01
2.30797276e-01 3.07323188e-01 -9.56134796e-02 -6.34318590e-02
5.82546175e-01 1.49054170e-01 -2.59668916e-01 -9.17612538e-02
-5.25078058e-01 8.87453079e-01 5.43298960e-01 3.52383614e-01
-4.31037903e-01 -6.19655512e-02 7.43659556e-01 2.37580881e-01
3.01630795e-01 9.20851007e-02 -2.91181933e-02 2.42574975e-01
-7.89493978e-01 3.12703126e-03 3.85536075e-01 3.90519053e-01
5.97411871e-01 4.09469873e-01 1.70227543e-01 4.05855745e-01
1.39054000e-01 2.09423825e-01 5.59153974e-01 -9.42183733e-01
4.66832578e-01 4.70353365e-01 -1.33434728e-01 -1.36858881e+00
1.80670112e-01 -3.17786127e-01 -7.93719709e-01 4.97359931e-01
4.16323900e-01 -1.77423477e-01 -5.87796688e-01 1.42672276e+00
3.43052685e-01 4.45526510e-01 9.87972766e-02 9.15970206e-01
5.81276715e-01 7.07667887e-01 9.17762145e-02 -3.67449313e-01
9.59792852e-01 -6.33224607e-01 -9.43332195e-01 1.68867916e-01
1.14480339e-01 -8.23969722e-01 9.75047171e-01 5.34831166e-01
-9.33889747e-01 -6.21876657e-01 -1.14373565e+00 2.37991944e-01
-4.20906603e-01 -3.67506415e-01 5.79380810e-01 8.52137506e-01
-5.59942424e-01 4.54162955e-01 -6.29404545e-01 -4.96672280e-02
7.01986730e-01 2.89940745e-01 -4.63572234e-01 -2.17930600e-01
-1.36025155e+00 8.92444789e-01 5.18751204e-01 2.65831649e-01
-9.94115651e-01 -4.07698959e-01 -5.09901166e-01 -3.86430733e-02
5.37485063e-01 -2.39870191e-01 7.59638190e-01 -1.13087869e+00
-1.21478057e+00 9.38215137e-01 3.78946364e-01 -7.07762182e-01
7.48238504e-01 -2.41309449e-01 -5.65096438e-01 4.40050840e-01
4.00404073e-02 1.17979601e-01 1.01895976e+00 -1.17684543e+00
-3.05773973e-01 -3.81454647e-01 3.23745906e-01 -2.02474788e-01
-7.07924783e-01 2.20139071e-01 -3.74456525e-01 -6.21361852e-01
-3.40633780e-01 -5.83163083e-01 -1.65545821e-01 2.37294301e-01
-2.61950403e-01 9.51894671e-02 1.33645713e+00 -9.41769838e-01
9.72072721e-01 -2.37033439e+00 -3.65493476e-01 1.99547336e-01
2.04989072e-02 9.11118627e-01 8.25629011e-02 2.73072928e-01
-1.26791030e-01 9.98314023e-02 -3.97369355e-01 1.40610412e-01
-2.30002463e-01 2.37943560e-01 -3.69155318e-01 7.99949110e-01
4.94466349e-02 5.59522212e-01 -7.91647375e-01 -7.09250867e-01
5.41991055e-01 4.16104823e-01 -3.21796060e-01 3.87219936e-01
1.24149576e-01 4.33931082e-01 -5.94754696e-01 3.64704788e-01
1.21166110e+00 1.43780172e-01 -8.57594088e-02 -2.64195442e-01
1.64484486e-01 -3.07013869e-01 -1.09432065e+00 1.13578618e+00
-2.47954175e-01 5.01440167e-01 1.70480520e-01 -1.44905674e+00
8.27139497e-01 3.40994567e-01 4.19627517e-01 -8.88748050e-01
4.32164282e-01 2.24189367e-02 1.01716511e-01 -8.96743536e-01
4.39669602e-02 -4.00459588e-01 2.68659323e-01 4.44426298e-01
-1.25995591e-01 2.55550623e-01 -1.21132545e-01 1.23785175e-01
9.28120852e-01 4.96183261e-02 -3.96061018e-02 1.30407035e-01
9.96684551e-01 -3.58966947e-01 6.40207410e-01 4.60865289e-01
-5.02047300e-01 2.28591159e-01 3.85601491e-01 -6.22550189e-01
-1.01330423e+00 -7.03621149e-01 -6.17592745e-02 6.54763043e-01
3.94216806e-01 1.17026642e-01 -1.00281131e+00 -9.56565738e-01
-5.32825403e-02 5.26459992e-01 -4.42926645e-01 -4.03678775e-01
-7.27313340e-01 -9.06817257e-01 9.31501150e-01 3.48363757e-01
1.15835702e+00 -1.18719172e+00 -4.13694680e-01 -7.12314397e-02
-5.75255379e-02 -1.20610154e+00 -2.42382318e-01 -2.70620167e-01
-8.19858491e-01 -1.41571784e+00 -4.80569303e-01 -5.61332047e-01
6.25896931e-01 2.67911315e-01 3.26465189e-01 3.16271424e-01
-2.91520685e-01 1.99884683e-01 -4.00105447e-01 -3.91572237e-01
-5.19334316e-01 -4.32855874e-01 -1.78914458e-01 3.87856483e-01
2.34379679e-01 -5.24676085e-01 -6.49612367e-01 1.82454944e-01
-1.33454800e+00 -3.59341741e-01 6.88390553e-01 8.34454238e-01
1.14879891e-01 6.33017659e-01 4.61589873e-01 -7.57441282e-01
6.47205353e-01 -5.39093852e-01 -4.68596637e-01 1.49006575e-01
-5.64915776e-01 -3.14854234e-02 8.27613711e-01 -3.98093522e-01
-1.25024045e+00 -3.28919828e-01 -2.79589236e-01 -7.19300389e-01
-2.46263072e-02 2.27824211e-01 -4.68536586e-01 -4.71858978e-01
3.28872949e-01 5.48190534e-01 1.87518507e-01 -4.89755571e-01
2.48771235e-01 6.39134645e-01 6.41695499e-01 -5.60853362e-01
9.21608865e-01 7.32276499e-01 1.64236918e-01 -5.80270171e-01
-5.87374091e-01 -3.20155144e-01 -2.68287390e-01 -3.73603493e-01
6.62891507e-01 -4.45471257e-01 -8.00296903e-01 8.78998399e-01
-1.19172585e+00 1.82511464e-01 1.00018635e-01 4.57822174e-01
-4.34405863e-01 1.14850879e+00 -7.09488392e-01 -9.54945207e-01
-5.08120775e-01 -1.24063027e+00 4.78122979e-01 1.21989205e-01
4.88527596e-01 -9.76050615e-01 -3.21249008e-01 5.30691862e-01
2.56809652e-01 5.84855676e-01 8.98471594e-01 -6.21840119e-01
-4.89441127e-01 -4.20981646e-01 -3.95402700e-01 9.87700582e-01
-6.73257606e-03 -2.08206207e-01 -9.50129688e-01 -3.25216085e-01
8.01220715e-01 -2.25246012e-01 6.91841543e-01 8.63043666e-02
1.51954770e+00 -6.08806968e-01 1.00510895e-01 7.72039950e-01
1.66624081e+00 3.89405161e-01 1.18528748e+00 7.90962338e-01
6.74739659e-01 5.68119109e-01 6.18242741e-01 3.49990249e-01
-9.55355838e-02 4.81108308e-01 9.15351152e-01 -1.80236459e-01
-3.01899426e-02 -5.90324067e-02 3.52694482e-01 3.95768046e-01
-6.81941658e-02 -1.84632003e-01 -6.56385601e-01 4.39910561e-01
-1.71881974e+00 -1.27218902e+00 -9.03560296e-02 2.20842838e+00
5.81571221e-01 2.99363315e-01 -5.17678633e-02 4.40788001e-01
1.06563210e+00 3.02128553e-01 -5.97253442e-01 -3.06436181e-01
-6.62743375e-02 1.19200684e-01 5.89404404e-01 1.76921651e-01
-1.13887572e+00 8.11824024e-01 5.79818726e+00 1.20070934e+00
-1.44320512e+00 2.85358995e-01 6.19808614e-01 1.42739654e-01
6.98175207e-02 6.01775609e-02 -1.94633573e-01 8.64655793e-01
6.05606973e-01 -1.18624773e-02 3.32936704e-01 8.03640425e-01
2.48331532e-01 1.20328590e-01 -3.60700727e-01 9.58096147e-01
1.47685066e-01 -1.04571772e+00 1.38561144e-01 3.09072495e-01
3.54509145e-01 -4.70901221e-01 2.85311669e-01 9.20044929e-02
-6.64933845e-02 -8.06201518e-01 5.46387494e-01 4.04881358e-01
3.56329352e-01 -1.08532929e+00 1.09869242e+00 4.85062510e-01
-6.12858415e-01 -1.49160981e-01 -5.08240581e-01 5.17088324e-02
1.63867190e-01 4.82798696e-01 -6.18944585e-01 7.62622297e-01
6.51831448e-01 3.24673295e-01 -3.33330154e-01 1.04228854e+00
-2.21294940e-01 7.42689848e-01 -3.73813254e-03 4.89399463e-01
5.49705565e-01 -1.22498021e-01 4.85466123e-01 1.01602316e+00
1.05457358e-01 4.44502085e-02 1.41321316e-01 7.25549638e-01
-1.10417433e-01 1.88635603e-01 -7.58718550e-01 -7.03679107e-05
1.77960649e-01 1.15808117e+00 -5.39563477e-01 -2.13735774e-01
-4.44993913e-01 8.49961221e-01 4.78132591e-02 2.16212437e-01
-8.20342243e-01 -3.39682817e-01 1.24636762e-01 3.38735044e-01
2.29138762e-01 -1.05786256e-01 -1.90400586e-01 -9.60504889e-01
1.43062890e-01 -9.65352654e-01 4.51800168e-01 -5.25286198e-01
-1.40318143e+00 3.93189341e-01 -1.39610618e-01 -1.23633337e+00
-3.28731909e-02 -5.54998815e-01 -1.03748441e+00 7.09762692e-01
-1.70109594e+00 -1.21839511e+00 -1.15023859e-01 9.05590713e-01
2.63458312e-01 -5.46263874e-01 3.13171655e-01 6.13408208e-01
-6.43655896e-01 6.83179498e-01 2.05425069e-01 5.47909379e-01
6.49443805e-01 -7.08879948e-01 -3.72914635e-02 1.20585680e+00
-2.56338805e-01 6.37903154e-01 7.42889166e-01 -5.55800855e-01
-1.24272394e+00 -9.03743327e-01 2.61688739e-01 1.05329923e-01
5.62519073e-01 1.33986652e-01 -1.11390054e+00 3.93276632e-01
4.11092758e-01 2.27127537e-01 6.36250079e-01 -4.98622537e-01
-4.37394708e-01 -2.19357610e-01 -1.60174263e+00 2.89391577e-01
4.31037039e-01 -5.59275389e-01 -6.25779033e-01 2.83296913e-01
5.92616320e-01 -2.54125316e-02 -6.62732482e-01 4.12964225e-01
3.66771519e-01 -1.11132312e+00 1.12329340e+00 -5.74250698e-01
2.84058630e-01 -4.26227242e-01 -4.38717976e-02 -6.75160468e-01
8.00380036e-02 -4.74037111e-01 -2.20603094e-01 1.24484801e+00
-4.63476896e-01 -4.42686021e-01 7.16761112e-01 2.24943459e-01
-2.96953376e-02 -5.62784910e-01 -8.36032748e-01 -9.85652506e-01
1.60115406e-01 -2.89968520e-01 5.63557088e-01 1.05238640e+00
-4.14313316e-01 -2.09687650e-01 -6.96912169e-01 3.55589956e-01
9.43023801e-01 -2.14321196e-01 7.21471965e-01 -7.94025540e-01
-1.81708857e-01 -1.97648242e-01 -6.70431614e-01 -4.99732643e-01
3.75079364e-01 -5.15000761e-01 -2.68028557e-01 -1.09949327e+00
2.00427920e-01 -2.55889654e-01 -4.58353758e-01 3.00824344e-01
-3.44527602e-01 1.95962802e-01 3.15114856e-01 3.26130778e-01
-3.73660415e-01 4.89837110e-01 1.27688646e+00 -2.59745568e-01
3.36822212e-01 8.51357579e-02 -7.72298098e-01 8.31785738e-01
7.63424814e-01 -5.87953746e-01 -3.75602990e-01 -3.80353272e-01
-3.40630174e-01 1.11346766e-02 5.97975373e-01 -9.24851120e-01
2.56025642e-01 -2.95109808e-01 2.38974184e-01 -3.77913237e-01
1.28009185e-01 -1.01543927e+00 -2.33320445e-02 6.71837091e-01
-3.41970846e-02 -1.98356628e-01 1.55878395e-01 6.29842639e-01
-3.68433833e-01 -5.46713233e-01 9.93633866e-01 -4.25832480e-01
-7.81463742e-01 3.84586751e-01 1.69192087e-02 -1.12769634e-01
1.17635906e+00 -3.37322891e-01 -1.59679934e-01 -2.87495553e-01
-3.05745661e-01 -9.02682468e-02 3.49069148e-01 1.64688468e-01
6.92604721e-01 -1.25561416e+00 -6.38807356e-01 -3.84839401e-02
-2.78562427e-01 -3.08500558e-01 5.11371732e-01 5.13270319e-01
-6.94730043e-01 2.20986046e-02 -5.17370939e-01 -7.89585784e-02
-1.10928392e+00 1.16871333e+00 2.79246747e-01 -2.84764796e-01
-5.73782802e-01 4.12637115e-01 2.72407144e-01 -1.35903046e-01
3.56533438e-01 6.19384527e-01 -2.27446213e-01 -1.49776161e-01
6.05179548e-01 6.27498507e-01 -3.97649072e-02 -7.87864149e-01
-2.63831198e-01 4.29651648e-01 -2.49282807e-01 7.06132576e-02
1.34282196e+00 9.84755438e-03 -2.33180031e-01 -1.93240598e-01
1.35522842e+00 2.54578218e-02 -1.17165256e+00 -2.23158047e-01
-1.24854989e-01 -8.84606838e-01 3.14652681e-01 -5.31842113e-01
-1.58790576e+00 1.23072588e+00 9.04589057e-01 1.41056329e-01
1.30542922e+00 -5.57477832e-01 1.23617065e+00 1.33298859e-01
2.52523363e-01 -1.11080480e+00 4.07564789e-01 8.35753158e-02
5.35270035e-01 -1.27670264e+00 9.05363336e-02 -9.72691476e-02
-5.22268832e-01 1.13937831e+00 6.85766280e-01 -3.95667106e-01
5.12003779e-01 2.82176249e-02 1.48058245e-02 3.07344669e-03
-8.60464573e-03 1.26070324e-02 -8.73846263e-02 7.23967791e-01
-5.76263629e-02 -3.89581144e-01 -6.53371751e-01 3.78811091e-01
2.59818256e-01 1.42945111e-01 4.26507980e-01 9.38361883e-01
-3.76854897e-01 -1.24362135e+00 -7.63813436e-01 3.80328670e-02
-9.60618317e-01 5.33773154e-02 -2.26480924e-02 7.16128409e-01
5.38793623e-01 7.88352430e-01 -2.96876878e-01 -2.64179945e-01
1.55238494e-01 -1.98064998e-01 3.30653936e-01 -6.43235222e-02
-5.35524309e-01 -1.59241185e-01 -4.45276350e-01 -3.10876071e-01
-4.91202772e-01 -4.60176796e-01 -1.05579662e+00 -7.01171815e-01
-4.33318257e-01 1.83744147e-01 6.75749362e-01 9.97835040e-01
1.43666249e-02 3.01807463e-01 8.84435654e-01 -5.28997302e-01
-7.04255521e-01 -5.69552302e-01 -5.67334116e-01 7.13185668e-01
1.30159438e-01 -5.17389357e-01 -4.94466066e-01 6.57665506e-02]
|
[12.331591606140137, 0.943594753742218]
|
c1ee7760-8476-4888-a586-818c8d782d5f
|
speaker-aware-bert-for-multi-turn-response
|
2004.03588
| null |
https://arxiv.org/abs/2004.03588v2
|
https://arxiv.org/pdf/2004.03588v2.pdf
|
Speaker-Aware BERT for Multi-Turn Response Selection in Retrieval-Based Chatbots
|
In this paper, we study the problem of employing pre-trained language models for multi-turn response selection in retrieval-based chatbots. A new model, named Speaker-Aware BERT (SA-BERT), is proposed in order to make the model aware of the speaker change information, which is an important and intrinsic property of multi-turn dialogues. Furthermore, a speaker-aware disentanglement strategy is proposed to tackle the entangled dialogues. This strategy selects a small number of most important utterances as the filtered context according to the speakers' information in them. Finally, domain adaptation is performed to incorporate the in-domain knowledge into pre-trained language models. Experiments on five public datasets show that our proposed model outperforms the present models on all metrics by large margins and achieves new state-of-the-art performances for multi-turn response selection.
|
['Zhen-Hua Ling', 'Jia-Chen Gu', 'Xiaodan Zhu', 'Quan Liu', 'Tianda Li', 'Zhiming Su', 'Si Wei']
|
2020-04-07
| null | null | null | null |
['conversational-response-selection']
|
['natural-language-processing']
|
[ 3.76609759e-03 1.02382727e-01 -2.79615968e-01 -6.60051048e-01
-1.41880476e+00 -6.44638896e-01 8.17309201e-01 -4.11559083e-03
-5.42290747e-01 8.22745740e-01 5.12288451e-01 -2.11979359e-01
-1.08190246e-01 -2.73067057e-01 -1.76982507e-02 -6.98281109e-01
2.76821941e-01 8.67268443e-01 4.05332327e-01 -8.69094670e-01
3.49080324e-01 1.71171010e-01 -9.14533556e-01 5.88324845e-01
1.04911256e+00 5.14210761e-01 4.73065823e-01 7.97980785e-01
-3.00950259e-01 9.54054654e-01 -6.90485060e-01 -4.62330878e-01
-7.24228024e-02 -8.32288682e-01 -1.43538678e+00 2.21021306e-02
-1.85133249e-01 -1.30489126e-01 -1.35795698e-01 6.61613345e-01
6.44657552e-01 5.30600607e-01 6.52658165e-01 -7.45592535e-01
-2.10247748e-02 1.05463111e+00 -2.56586075e-01 4.14775193e-01
6.05132997e-01 1.57035455e-01 1.49394941e+00 -9.62827027e-01
5.67714214e-01 1.75844014e+00 8.38075057e-02 8.45856071e-01
-1.22037852e+00 -3.78486365e-01 3.07843238e-01 4.11606610e-01
-9.15031672e-01 -7.07490563e-01 1.10575843e+00 -1.94837973e-01
8.06741059e-01 6.34602666e-01 2.43445829e-01 1.17133534e+00
-7.71368295e-03 1.13088906e+00 1.10443437e+00 -7.34951735e-01
3.89184952e-02 5.23507893e-01 3.07338297e-01 3.62584352e-01
-7.44150460e-01 -4.30767953e-01 -8.32007706e-01 -5.17178774e-01
1.42482862e-01 -5.50109923e-01 -5.09413220e-02 5.52802579e-03
-1.11828077e+00 1.19144762e+00 6.65334091e-02 3.82970959e-01
-4.30864573e-01 -4.80738997e-01 6.12798512e-01 6.08877838e-01
6.93046927e-01 7.19311714e-01 -4.59631622e-01 -4.24258739e-01
-2.90884763e-01 3.80008250e-01 1.10537672e+00 7.97854304e-01
6.85494781e-01 -4.47400957e-01 -6.16847932e-01 1.32035959e+00
3.79560173e-01 2.32278466e-01 3.50075722e-01 -7.98371315e-01
9.68787253e-01 6.13195360e-01 1.88748568e-01 -5.46733975e-01
-5.15117645e-01 -1.34038210e-01 -6.42223358e-01 -6.52930021e-01
4.69186395e-01 -2.15204999e-01 -3.82942080e-01 1.69330072e+00
7.71446705e-01 -3.19286495e-01 2.58151650e-01 9.31509495e-01
8.36615086e-01 4.76465851e-01 1.25905856e-01 -4.84891385e-01
1.49637544e+00 -1.10386527e+00 -7.89221287e-01 -3.09746981e-01
5.93449533e-01 -1.01928902e+00 1.21973097e+00 1.91965520e-01
-8.23301733e-01 -1.73901334e-01 -5.44408917e-01 1.10357277e-01
1.70077711e-01 1.53074160e-01 4.76199478e-01 5.01310170e-01
-6.39599562e-01 6.58567101e-02 -3.25831324e-01 -4.59030449e-01
-3.34254980e-01 2.76828706e-01 4.91333543e-04 2.41512477e-01
-1.75743544e+00 1.13776851e+00 1.41365463e-02 1.90778837e-01
-9.25284684e-01 -1.20072432e-01 -5.44279575e-01 -8.41657966e-02
6.97647154e-01 -2.85678327e-01 1.64929092e+00 -8.56990039e-01
-2.42603016e+00 7.15965033e-01 -4.22268927e-01 -3.05213720e-01
6.56089306e-01 -8.65916908e-02 -1.54957578e-01 4.42402393e-01
4.21725661e-02 2.27477685e-01 8.40994775e-01 -1.09076059e+00
-7.93657720e-01 -1.51551813e-01 3.71946961e-01 7.08978713e-01
-1.73532382e-01 4.09957796e-01 -2.27912426e-01 -1.16896592e-01
5.90500087e-02 -1.11191213e+00 -2.71067202e-01 -1.05449474e+00
-6.27403021e-01 -8.99330556e-01 4.96734619e-01 -5.34375846e-01
1.23389518e+00 -1.74364650e+00 4.55846369e-01 1.16742380e-01
2.61419833e-01 2.73659468e-01 -3.63939196e-01 9.71746147e-01
4.49017107e-01 -1.25584990e-01 1.15051165e-01 -5.60787141e-01
-2.25483328e-02 -1.37822274e-02 -2.13556513e-01 2.42847681e-01
1.97040275e-01 4.86667603e-01 -9.78829265e-01 -6.44987822e-01
2.20998466e-01 -1.36198983e-01 -5.18413067e-01 6.56677246e-01
-2.94234663e-01 1.07809031e+00 -9.22274411e-01 2.03622252e-01
4.07347262e-01 2.87260767e-02 6.31209016e-01 2.60473460e-01
-9.63010862e-02 1.13822663e+00 -7.09190130e-01 1.47971284e+00
-7.54966080e-01 2.93275326e-01 2.45427325e-01 -9.82756376e-01
1.00096238e+00 5.06269574e-01 1.82449907e-01 -5.87751567e-01
1.41553417e-01 1.27801031e-01 4.78377789e-01 -6.26648903e-01
7.61734962e-01 -9.32769701e-02 -5.85961461e-01 5.95277727e-01
1.11676469e-01 -2.13115022e-01 1.49886698e-01 5.93506694e-01
8.12019050e-01 -3.76232356e-01 4.53225702e-01 -2.95003027e-01
1.04313850e+00 -2.14474440e-01 6.35457277e-01 9.65132535e-01
-5.77217996e-01 2.10954443e-01 7.68714488e-01 -2.04694733e-01
-4.99391377e-01 -4.62062865e-01 1.80233061e-01 1.84902799e+00
2.95220137e-01 -2.79477566e-01 -7.20508993e-01 -9.19386208e-01
-3.85276675e-01 8.66039455e-01 -5.44515252e-01 -2.13092268e-01
-7.99773455e-01 -6.50891066e-01 3.98111105e-01 -1.00359537e-01
4.20721799e-01 -1.16155243e+00 -2.51407951e-01 4.17579472e-01
-9.57986593e-01 -1.08998466e+00 -6.12536550e-01 4.15659338e-01
-4.84508783e-01 -8.44446480e-01 -3.82504135e-01 -5.78980684e-01
8.83579552e-02 3.57285798e-01 9.75608170e-01 -1.89473540e-01
3.89831036e-01 2.10664049e-01 -8.44417572e-01 -2.31915876e-01
-1.01656353e+00 6.68483853e-01 -6.82242438e-02 3.55064183e-01
4.92830187e-01 -2.04039231e-01 -4.12673771e-01 6.09477699e-01
-2.83389300e-01 -9.27994177e-02 5.58430493e-01 1.37084293e+00
1.14113027e-02 -4.97332245e-01 1.20505178e+00 -1.05631101e+00
1.21499860e+00 -5.02575994e-01 -1.42407224e-01 5.89434922e-01
-4.32005286e-01 1.44758478e-01 5.85295916e-01 -6.13698781e-01
-1.62829936e+00 -1.22035466e-01 -1.32666782e-01 2.75719345e-01
-6.90310448e-02 5.06889045e-01 -3.03740710e-01 7.34331235e-02
5.95101476e-01 2.93499261e-01 -9.81368795e-02 -6.18655264e-01
4.55378979e-01 1.04830050e+00 -2.34629229e-01 -6.99814081e-01
2.76312053e-01 -5.77616990e-02 -5.69687545e-01 -8.95072222e-01
-8.34287643e-01 -1.00187814e+00 -7.85664678e-01 -3.30456525e-01
6.43611312e-01 -7.16292083e-01 -8.74391139e-01 3.85112107e-01
-1.47843373e+00 -4.15874332e-01 2.26593927e-01 5.44218838e-01
-5.49132586e-01 3.15508366e-01 -9.01381314e-01 -1.36695075e+00
-4.25204337e-01 -1.29783237e+00 7.86608338e-01 2.97010511e-01
-4.18097973e-01 -9.04875875e-01 2.71926135e-01 7.93337345e-01
2.64731616e-01 -5.61879635e-01 8.64234626e-01 -1.50070930e+00
-4.73281890e-01 -1.19083911e-01 1.65789545e-01 1.58814296e-01
9.97662544e-02 -3.34810346e-01 -9.29577589e-01 -2.86819730e-02
1.89364046e-01 -6.31188929e-01 8.42591643e-01 2.00257124e-03
4.97442991e-01 -3.93870294e-01 -2.77208120e-01 -2.66177148e-01
6.50792122e-01 3.04260045e-01 3.20011884e-01 2.31599361e-01
4.52504426e-01 9.80706513e-01 1.06443059e+00 6.20655656e-01
8.98669839e-01 9.31353271e-01 2.37678662e-01 2.67765135e-01
2.96869129e-01 -1.67098135e-01 3.60889584e-01 1.19187164e+00
1.39116704e-01 -4.37775165e-01 -6.68184936e-01 5.54035962e-01
-2.02783847e+00 -9.97779906e-01 -1.78197965e-01 2.00387001e+00
1.30779159e+00 8.59449133e-02 5.05000591e-01 -2.19715565e-01
7.97214746e-01 4.40579176e-01 -3.85987520e-01 -7.25114226e-01
1.62158273e-02 -3.24685089e-02 2.81302677e-03 8.72507393e-01
-1.03164124e+00 1.38374972e+00 5.65604019e+00 1.04032791e+00
-9.16284621e-01 4.05097485e-01 5.11936009e-01 1.13523319e-01
-2.53756374e-01 1.52198493e-01 -9.56915557e-01 2.59473622e-01
9.00746167e-01 -1.43167928e-01 2.90869206e-01 6.36424899e-01
4.91904527e-01 -3.74803156e-01 -9.69523311e-01 3.97316366e-01
5.48525378e-02 -8.20132256e-01 -2.13930398e-01 -2.36623704e-01
3.79596353e-01 -1.68061361e-01 -1.26763552e-01 6.83444202e-01
6.66702092e-01 -4.40083385e-01 5.00622690e-01 3.37276131e-01
3.45020890e-01 -7.21843302e-01 7.16161847e-01 8.62627804e-01
-7.36962974e-01 -1.60720065e-01 -2.44743720e-01 7.92010650e-02
1.27446100e-01 2.65028507e-01 -1.39377654e+00 6.32603347e-01
3.23523104e-01 2.59286314e-01 -1.95496723e-01 6.30189538e-01
-4.22466427e-01 9.67722416e-01 -1.08888336e-01 -6.85644984e-01
3.44456047e-01 -8.21391642e-02 8.94531786e-01 1.29282129e+00
-3.48269194e-01 3.43061179e-01 3.33996534e-01 5.17746031e-01
1.09441550e-02 3.39387447e-01 -1.52716205e-01 2.29998738e-01
7.59327710e-01 1.35716343e+00 -3.34070951e-01 -2.91971058e-01
-9.23884511e-02 9.29749310e-01 4.71213698e-01 2.99140722e-01
-5.09714067e-01 -3.45628768e-01 5.83781779e-01 -4.22313541e-01
3.81364562e-02 1.17755048e-01 8.08142200e-02 -1.15019143e+00
-1.50416747e-01 -1.22385061e+00 5.12658954e-01 7.82033056e-02
-1.30546308e+00 7.84072280e-01 -6.61512166e-02 -1.15572810e+00
-6.15096211e-01 -7.33604934e-03 -6.57096922e-01 9.68919337e-01
-1.52992201e+00 -1.18016243e+00 3.12129855e-01 4.73777205e-01
1.06731951e+00 -2.34161749e-01 9.70783710e-01 8.42973292e-02
-6.47683620e-01 8.67988527e-01 1.62065610e-01 7.23869773e-03
1.07112169e+00 -1.15052342e+00 -2.50176061e-02 5.24652481e-01
-1.59329362e-02 7.18888342e-01 8.15556228e-01 -3.39716077e-01
-1.22404265e+00 -5.35569370e-01 1.40503252e+00 -5.48953414e-01
6.30676270e-01 -5.36650240e-01 -9.09407973e-01 5.27939677e-01
4.12660897e-01 -6.48256719e-01 7.93578744e-01 6.04543388e-01
-2.56061286e-01 -6.53252229e-02 -1.04711986e+00 5.09489357e-01
6.03669703e-01 -7.57141471e-01 -9.05865967e-01 5.59802711e-01
7.66045451e-01 -6.40917718e-01 -4.43576097e-01 -2.93808598e-02
4.87338960e-01 -8.05396557e-01 7.96668112e-01 -8.01135063e-01
1.14966549e-01 2.98960030e-01 8.31327494e-03 -1.61895132e+00
-1.91192627e-01 -1.08812809e+00 1.64000779e-01 1.42839837e+00
5.59537113e-01 -7.08449841e-01 4.34736103e-01 6.97299182e-01
-4.13792245e-02 -5.04883468e-01 -1.19363642e+00 -4.84280854e-01
1.57703981e-01 -1.13799032e-02 4.24436480e-01 9.11594033e-01
6.43986523e-01 1.02894843e+00 -7.50527263e-01 -7.54598826e-02
2.47048497e-01 3.09109479e-01 9.64841843e-01 -1.10516524e+00
-2.27753565e-01 -4.12061781e-01 2.16186434e-01 -1.62702727e+00
4.56517190e-01 -5.85293949e-01 4.48205471e-01 -1.16869235e+00
2.56795466e-01 -5.40046275e-01 -2.83888012e-01 3.03140935e-02
-6.56983614e-01 -3.99033368e-01 3.18655401e-01 3.34136605e-01
-1.06238365e+00 9.37213361e-01 1.26196778e+00 -1.40988424e-01
-6.81351662e-01 7.16008663e-01 -6.89158618e-01 5.08032739e-01
7.33364463e-01 -7.33752131e-01 -2.25664690e-01 -6.19950593e-02
-2.66375601e-01 6.44008577e-01 -1.03650883e-01 -1.43293396e-01
3.00211877e-01 -6.28790557e-01 -5.23254931e-01 -5.73036432e-01
5.89380622e-01 -4.19863731e-01 -6.45874739e-01 2.17493221e-01
-1.22169924e+00 -3.18360060e-01 -2.67818898e-01 7.97164142e-01
-2.27466673e-01 -3.43907326e-01 5.46484828e-01 -1.82474732e-01
-3.93846750e-01 -1.48014829e-01 -7.70907462e-01 3.23749959e-01
6.53140843e-01 3.67694497e-01 -2.37286568e-01 -7.76728928e-01
-6.19013786e-01 4.83943701e-01 -3.95087332e-01 5.14447391e-01
3.60196352e-01 -9.35385764e-01 -8.74142051e-01 -2.32600018e-01
3.11518967e-01 -1.88453436e-01 5.14816344e-01 1.00980604e+00
1.79667085e-01 6.10594094e-01 3.15360039e-01 -4.67111528e-01
-1.61533821e+00 -2.97406781e-02 3.10166597e-01 -8.75291705e-01
-2.46942252e-01 1.12358391e+00 2.02844128e-01 -8.35027814e-01
8.54975730e-02 -1.08829113e-02 -5.72473705e-01 3.23383927e-01
4.56054598e-01 1.10599019e-01 1.42926844e-02 -7.79835999e-01
-5.14351249e-01 -1.46546468e-01 -4.84754473e-01 -4.94134158e-01
1.11969197e+00 -7.31779635e-01 -1.33990720e-01 7.23384380e-01
9.46880639e-01 2.35138893e-01 -9.17990029e-01 -8.70017529e-01
3.91824543e-01 -3.30743343e-01 -3.65755588e-01 -9.94961560e-01
-3.19683731e-01 8.63789022e-01 -5.51495254e-02 4.01152700e-01
7.66942322e-01 1.63048387e-01 6.50175035e-01 8.32921505e-01
3.60438645e-01 -1.36127591e+00 4.38749403e-01 1.02944624e+00
1.02518940e+00 -1.51155233e+00 -3.16348374e-01 -4.97235566e-01
-1.22148931e+00 9.75254834e-01 7.89642930e-01 2.79323220e-01
3.33797783e-01 -3.21415961e-01 4.72215116e-01 -1.04891613e-01
-1.24451733e+00 -3.70396584e-01 1.67079389e-01 3.16371411e-01
4.97292429e-01 3.18131179e-01 -7.29694963e-01 5.21720052e-01
-6.92901164e-02 -6.27641857e-01 4.67906803e-01 5.59033930e-01
-4.28816468e-01 -1.44029474e+00 -2.41022140e-01 2.31009245e-01
-3.43319207e-01 -1.62687898e-02 -9.31729257e-01 5.10663509e-01
-5.89141011e-01 1.58530581e+00 -5.27735710e-01 -5.72204947e-01
3.41125965e-01 3.47139597e-01 2.26283073e-02 -8.62747967e-01
-1.20246005e+00 1.68411434e-01 7.22942173e-01 -2.01718405e-01
-5.43260753e-01 -7.23467231e-01 -9.50755060e-01 -9.43209305e-02
-8.70445669e-01 6.47397041e-01 4.92900133e-01 1.29874659e+00
2.90727675e-01 3.55156302e-01 1.33185959e+00 -6.28962994e-01
-1.30242443e+00 -1.53677762e+00 -3.73357743e-01 3.32695276e-01
3.51238757e-01 -5.74116826e-01 -4.55794156e-01 -4.32619005e-01]
|
[12.601560592651367, 7.827559947967529]
|
d41c9f1a-be07-45a0-805b-fa114ee5a7d3
|
cem500k-a-large-scale-heterogeneous-unlabeled
| null | null |
https://www.biorxiv.org/content/10.1101/2020.12.11.421792v1
|
https://www.biorxiv.org/content/10.1101/2020.12.11.421792v1.full.pdf
|
CEM500K – A large-scale heterogeneous unlabeled cellular electron microscopy image dataset for deep learning
|
Automated segmentation of cellular electron microscopy (EM) datasets remains a challenge. Supervised deep learning (DL) methods that rely on region-of-interest (ROI) annotations yield models that fail to generalize to unrelated datasets. Newer unsupervised DL algorithms require relevant pre-training images, however, pre-training on currently available EM datasets is computationally expensive and shows little value for unseen biological contexts, as these datasets are large and homogeneous. To address this issue, we present CEM500K, a nimble 25 GB dataset of 500,000 unique cellular EM images curated from nearly 600 three-dimensional (3D) and 10,000 two-dimensional (2D) images from >100 unrelated imaging projects. We show that models pre-trained on CEM500K learn features that are biologically relevant and resilient to meaningful image augmentations. Critically, we evaluate transfer learning from these pre-trained models on six publicly available and one newly derived benchmark segmentation task and report state-of-the-art results on each. We release the CEM500K dataset, pre-trained models and curation pipeline for model building and further expansion by the EM community.
|
['Kedar Narayan', 'Ryan W Conrad']
|
2020-12-11
| null | null | null | null |
['electron-microscopy-image-segmentation']
|
['computer-vision']
|
[ 3.84130895e-01 -2.34202873e-02 5.66315770e-01 -4.03942853e-01
-1.14615667e+00 -4.53365058e-01 6.12756073e-01 1.48248509e-01
-1.02026725e+00 1.16178501e+00 -3.20696145e-01 -2.03489035e-01
2.25295037e-01 -2.82125592e-01 -8.52120757e-01 -9.39816296e-01
3.42393145e-02 1.03471208e+00 3.86559308e-01 3.11326385e-01
3.86137843e-01 8.30296457e-01 -8.73580873e-01 4.95889693e-01
3.08634371e-01 7.22788036e-01 7.11028278e-01 9.62887049e-01
4.91104461e-02 5.76866210e-01 -4.06866044e-01 3.33198793e-02
-1.10436291e-01 -4.20334756e-01 -1.46271598e+00 -1.12675555e-01
5.24216175e-01 -2.34132141e-01 -8.61927122e-02 6.56572402e-01
8.22341502e-01 -2.45296612e-01 9.39660668e-01 -9.30549443e-01
-8.91977668e-01 9.64744166e-02 -3.01066101e-01 9.17692363e-01
-4.24168229e-01 2.87929088e-01 4.68414634e-01 -8.15010905e-01
1.43347526e+00 7.88086832e-01 1.00940287e+00 9.42919910e-01
-2.01900291e+00 -2.59849936e-01 -1.77677467e-01 -1.86358437e-01
-1.00260496e+00 -5.07629752e-01 2.02667996e-01 -7.47548044e-01
1.33116925e+00 -5.41860722e-02 5.57133913e-01 1.17004871e+00
3.58279973e-01 5.70919037e-01 1.43244541e+00 -6.51430786e-02
2.88967073e-01 -2.57570177e-01 5.84473573e-02 5.35570145e-01
-2.77633294e-02 -3.39647323e-01 -1.08796231e-01 -2.97145158e-01
1.01242661e+00 -1.70483544e-01 -3.07363629e-01 -4.08878684e-01
-1.54166996e+00 4.80700791e-01 1.84456229e-01 4.13215846e-01
-9.75148380e-02 7.97586963e-02 4.34819549e-01 6.07037395e-02
5.86250544e-01 6.53340161e-01 -9.59187031e-01 6.89173788e-02
-1.11330080e+00 1.60620600e-01 4.42020208e-01 6.96816921e-01
9.03040111e-01 -2.62582660e-01 2.38080487e-01 8.87353361e-01
1.15083903e-01 8.20135400e-02 6.21448755e-01 -1.05535436e+00
-2.02296942e-01 5.18848538e-01 -1.13742761e-01 -6.38537824e-01
-1.00833523e+00 -1.76913232e-01 -8.23666930e-01 1.45180136e-01
5.78695178e-01 -7.31923729e-02 -1.23478591e+00 1.79094958e+00
2.93390214e-01 1.50898904e-01 -2.56127030e-01 6.32203579e-01
9.39961374e-01 2.39068523e-01 2.35769227e-01 3.11425012e-02
8.89836252e-01 -6.98757648e-01 -3.10538113e-01 -1.66093230e-01
1.00088263e+00 -6.90250456e-01 7.90438116e-01 2.23024264e-01
-1.01885676e+00 -2.34569192e-01 -8.75041723e-01 -4.40975964e-01
-8.37182403e-01 -4.29943651e-02 4.62142676e-01 1.88331939e-02
-1.53058338e+00 6.36906087e-01 -1.15051496e+00 -9.67686474e-01
1.05747950e+00 6.66996241e-01 -7.30168104e-01 8.28926638e-02
-3.75229657e-01 1.04317260e+00 2.70569265e-01 -2.44087055e-01
-1.43845475e+00 -1.03052545e+00 -3.92805755e-01 -4.23928708e-01
-4.35592294e-01 -8.64975393e-01 1.02493131e+00 -3.74329746e-01
-9.04293656e-01 1.88055050e+00 -1.05151109e-01 -5.77646196e-01
2.83112079e-01 2.56162316e-01 -2.64663622e-02 7.36825466e-01
2.90601909e-01 1.30652368e+00 2.57326424e-01 -1.59567869e+00
-3.43080640e-01 -4.13019419e-01 -5.89650691e-01 -4.57365245e-01
-9.92530212e-03 1.08193822e-01 -2.35636652e-01 -4.88521874e-01
1.40471965e-01 -7.97914028e-01 -3.82780939e-01 1.52785242e-01
-3.90340924e-01 2.83121347e-01 1.08721185e+00 -7.31162786e-01
4.62811500e-01 -1.69665039e+00 2.38210961e-01 -3.40810567e-01
5.08601248e-01 1.60656005e-01 -2.77148068e-01 9.88165066e-02
-1.41105220e-01 1.86856523e-01 -5.62549472e-01 -7.18470097e-01
-1.48969322e-01 1.40346825e-01 1.67847216e-01 8.46591890e-01
4.98737335e-01 1.17298079e+00 -8.32635522e-01 -8.07880402e-01
3.37807536e-01 5.64986944e-01 -4.45004582e-01 3.17640305e-01
-1.31011605e-01 1.09805083e+00 -8.75315070e-02 7.80503809e-01
6.34729028e-01 -7.95061707e-01 1.39614135e-01 -2.25843489e-01
1.97833866e-01 -1.15743212e-01 -3.16350669e-01 1.74687040e+00
-1.00976847e-01 8.85283947e-01 4.94299501e-01 -1.10614562e+00
6.13568723e-01 1.04620472e-01 7.75100470e-01 -3.92754555e-01
2.96950668e-01 2.23576695e-01 -2.36415431e-01 -2.48800114e-01
-1.24813765e-01 -3.22814673e-01 4.16905165e-01 4.03108537e-01
7.29808033e-01 -3.85111123e-01 1.90895215e-01 3.21058542e-01
1.48490226e+00 1.97453663e-01 -9.42813931e-04 -7.11361349e-01
2.36661896e-01 2.07344919e-01 5.93670666e-01 4.20990080e-01
-4.27280933e-01 1.17114735e+00 4.21049565e-01 -6.30589604e-01
-1.58220279e+00 -1.07306743e+00 -7.85414398e-01 8.48185718e-01
-9.73587260e-02 8.08535144e-02 -1.09486687e+00 -5.59962749e-01
-8.27886164e-02 -1.86462961e-02 -1.10915267e+00 1.65019676e-01
-4.14071530e-01 -1.38281775e+00 8.27944934e-01 5.47514141e-01
3.44243705e-01 -1.18264937e+00 -4.33741391e-01 3.64678204e-01
8.09755642e-03 -1.31061029e+00 -1.30781338e-01 7.69600093e-01
-9.28768933e-01 -1.27537775e+00 -9.28805232e-01 -1.21371055e+00
1.15906215e+00 -5.34292236e-02 1.49512875e+00 3.15273792e-01
-7.82342255e-01 4.06169266e-01 1.66703276e-02 -1.35664880e-01
-2.92924166e-01 1.17915183e-01 7.59994984e-03 -5.50746799e-01
5.54983914e-01 -7.75448203e-01 -6.26460493e-01 3.94107848e-01
-9.32554841e-01 1.38067827e-01 4.31395620e-01 1.18717957e+00
1.21115088e+00 -5.13794839e-01 8.90515447e-01 -1.14384127e+00
2.03830257e-01 -6.39724314e-01 -3.16854775e-01 7.66129494e-02
-2.54923284e-01 -3.80990058e-01 5.43573976e-01 -2.92676330e-01
-7.63821721e-01 1.64354056e-01 -3.33374351e-01 -1.93501741e-01
-5.18388510e-01 5.88714220e-02 -1.36336414e-02 -4.35677230e-01
5.92568099e-01 1.83701634e-01 7.40516707e-02 -5.36767781e-01
1.17572732e-01 4.18091267e-01 9.60193753e-01 -8.71111572e-01
2.96766579e-01 9.00908768e-01 2.20834956e-01 -7.90803194e-01
-5.91958284e-01 -3.98664296e-01 -1.17077637e+00 3.60059366e-02
1.22625446e+00 -6.44673884e-01 -3.55192542e-01 6.92525685e-01
-9.80123699e-01 -9.46710229e-01 -3.05314618e-03 3.88862044e-02
-9.55597162e-01 3.57619226e-01 -1.22753143e+00 -2.45171681e-01
-4.62523550e-01 -1.26031446e+00 1.34926820e+00 7.38637224e-02
-4.48337942e-01 -1.16941524e+00 3.63314211e-01 4.59799498e-01
3.40608984e-01 4.67013121e-01 1.28391314e+00 -8.07066143e-01
-5.16699672e-01 4.25320007e-02 -3.78415287e-01 1.98459908e-01
1.42763644e-01 8.28894451e-02 -1.11820352e+00 -4.90145683e-01
-2.87575990e-01 -9.04310405e-01 1.13652718e+00 5.50161541e-01
1.44126713e+00 2.74033964e-01 -6.27235174e-01 1.02409983e+00
1.37372792e+00 -6.13654172e-03 7.25168824e-01 6.25315964e-01
4.26355422e-01 5.14427722e-01 6.47136867e-02 2.34629679e-02
1.62907138e-01 1.62260517e-01 3.30083966e-01 -6.29847050e-01
-1.56045556e-01 1.88966215e-01 -3.63778532e-01 5.64098418e-01
1.05845034e-01 -8.48976150e-02 -1.08690488e+00 8.61649334e-01
-1.50932395e+00 -6.71038330e-01 -7.31717795e-02 1.59003425e+00
1.16508770e+00 9.69458371e-03 -1.52362464e-02 -4.41248119e-01
5.74978113e-01 -3.94235879e-01 -9.60777938e-01 7.25003630e-02
-7.60623813e-01 3.84953558e-01 2.52058089e-01 3.25063825e-01
-1.34912109e+00 9.72418070e-01 7.49691916e+00 6.89835131e-01
-8.84869277e-01 1.76640615e-01 1.42103124e+00 -1.30533561e-01
-3.16673890e-02 -1.14205226e-01 -7.41343915e-01 4.91797805e-01
9.03150380e-01 1.79064065e-01 2.34823689e-01 7.59412587e-01
-8.72747879e-03 -1.52203500e-01 -1.38751459e+00 9.09518778e-01
-2.06060842e-01 -1.90569615e+00 -8.27673525e-02 3.45324516e-01
9.37950552e-01 8.03318679e-01 6.84849918e-02 3.19877043e-02
5.57850301e-01 -1.36248183e+00 9.90980417e-02 6.21396005e-01
8.74050081e-01 -5.66672206e-01 9.32654679e-01 9.25277844e-02
-6.38940871e-01 5.29893517e-01 -7.92112708e-01 4.62988883e-01
1.98930770e-01 5.78986704e-01 -6.91139758e-01 -2.56645516e-03
9.95264530e-01 6.69318795e-01 -8.72266710e-01 9.40596223e-01
4.97911185e-01 4.17822063e-01 -2.74829209e-01 3.84618759e-01
1.43781245e-01 -9.30546895e-02 1.21721134e-01 1.56356609e+00
2.35694442e-02 -7.74613721e-03 -1.28583582e-02 1.31691027e+00
-4.55567151e-01 -1.72206387e-01 -4.98493433e-01 -2.45440945e-01
3.01187396e-01 1.76478982e+00 -1.40534306e+00 -3.56045097e-01
-2.28899449e-01 8.98304224e-01 8.01535904e-01 4.02313352e-01
-5.88653147e-01 -1.08382598e-01 7.97676623e-01 3.17305513e-02
3.28150034e-01 -1.51263401e-01 -2.98786283e-01 -8.68354142e-01
-6.56223118e-01 -6.30287051e-01 2.73699462e-01 -9.73827600e-01
-1.94158280e+00 5.28999746e-01 -5.18040240e-01 -4.47996408e-01
2.01995671e-01 -1.03088629e+00 -4.73394781e-01 7.20782638e-01
-1.28288364e+00 -1.28361964e+00 -1.50397360e-01 1.95464566e-01
2.55889535e-01 -9.61692184e-02 1.05646598e+00 2.41098151e-01
-7.12965846e-01 2.22965181e-01 4.67532337e-01 2.06155419e-01
8.06583047e-01 -1.46780443e+00 5.25772750e-01 3.81940663e-01
-1.51911139e-01 9.36261296e-01 3.89231443e-01 -6.23268366e-01
-9.69675601e-01 -1.30178678e+00 5.54220378e-01 -1.06212354e+00
5.37457824e-01 -4.01689321e-01 -1.08594990e+00 1.07261300e+00
2.18638390e-01 4.37520891e-01 9.47359920e-01 -3.06199282e-01
4.08891477e-02 5.93524158e-01 -1.66481686e+00 4.98030394e-01
9.42341030e-01 -5.53168833e-01 -5.47192216e-01 4.22185510e-01
4.76244152e-01 -1.04500867e-01 -1.36040151e+00 4.07305002e-01
2.12194085e-01 -9.06041622e-01 1.08911252e+00 -8.65629017e-01
4.68825221e-01 -2.83110291e-01 -2.19773725e-01 -1.16997480e+00
-4.35811669e-01 -4.03983176e-01 -1.18601033e-02 1.16837859e+00
5.33323705e-01 -4.20679480e-01 9.05440092e-01 4.36114490e-01
-4.94958490e-01 -1.07740128e+00 -9.12843347e-01 -4.46436316e-01
7.80774653e-01 9.47282314e-02 3.40284288e-01 1.14444625e+00
7.47444406e-02 2.88617969e-01 1.66922778e-01 -2.38234788e-01
7.46636569e-01 -1.32384121e-01 8.61446500e-01 -1.18070507e+00
-2.28037965e-02 -5.19369960e-01 -5.52003145e-01 -9.43385422e-01
2.67125398e-01 -1.05175245e+00 1.49232507e-01 -1.62043941e+00
8.78281474e-01 -1.03652105e-01 -4.43484128e-01 4.65598047e-01
-7.61236995e-02 7.87910998e-01 -4.57897127e-01 2.52727866e-01
-8.91717196e-01 2.81801492e-01 1.32054055e+00 -1.47431076e-01
4.21315789e-01 -7.63412058e-01 -4.48567808e-01 8.38534236e-01
7.90908694e-01 -3.78819227e-01 -6.77487552e-02 -5.47097802e-01
-2.35987172e-01 -5.97509325e-01 5.86412191e-01 -1.10459316e+00
2.05017120e-01 -3.09437588e-02 1.10364115e+00 -8.54932964e-01
3.15624237e-01 -3.95543456e-01 3.71575579e-02 4.75476263e-03
-2.50289142e-01 -1.04372203e-01 4.79913324e-01 4.24557388e-01
2.01667666e-01 2.71059275e-02 1.43763041e+00 -6.61507964e-01
-4.14063692e-01 5.01415670e-01 -7.25208879e-01 4.06108201e-01
8.94143701e-01 -3.66845876e-01 -5.81148088e-01 1.21158592e-01
-9.53048348e-01 2.41333991e-01 1.15571868e+00 -3.18928152e-01
4.78231549e-01 -1.02017701e+00 -4.85074401e-01 5.15117124e-03
9.95054245e-02 3.08461428e-01 3.59600604e-01 1.01429808e+00
-8.76206160e-01 5.11901796e-01 -6.52584195e-01 -8.34426999e-01
-9.90591407e-01 5.20467818e-01 6.20783627e-01 -2.96867728e-01
-6.51631594e-01 1.17057681e+00 5.35206556e-01 -9.46162939e-01
-1.04770266e-01 -3.00564349e-01 5.89274056e-02 -3.59068990e-01
4.49534982e-01 1.56302392e-01 6.14191666e-02 -6.89722657e-01
-3.26765329e-01 4.99596059e-01 -3.04965317e-01 2.24553764e-01
1.81523418e+00 -3.81630272e-01 -5.10577917e-01 3.75355750e-01
1.41719449e+00 -7.00081229e-01 -1.50649047e+00 1.75865572e-02
2.63671070e-01 -3.61191891e-02 2.13533845e-02 -8.25100601e-01
-1.10770237e+00 9.74351525e-01 4.11096692e-01 -1.79807782e-01
8.68379474e-01 2.89905548e-01 1.01352453e+00 4.57482755e-01
5.58820546e-01 -1.21680546e+00 1.69223458e-01 6.37014627e-01
3.49755257e-01 -1.19449997e+00 -1.65047288e-01 -6.89306762e-03
-1.37936071e-01 9.01406646e-01 9.12200809e-01 7.91483521e-02
6.36579335e-01 6.36964619e-01 1.83897719e-01 -4.08858389e-01
-9.55830336e-01 -1.35574177e-01 -3.02884609e-01 1.18917203e+00
5.67061841e-01 -4.55792665e-01 6.19320199e-02 7.86807477e-01
9.02538374e-02 6.83326498e-02 3.66312146e-01 1.20762157e+00
-5.46467245e-01 -8.51747513e-01 3.68064307e-02 6.79422200e-01
-7.20648229e-01 -8.34379867e-02 -7.12508917e-01 8.09498310e-01
1.36686444e-01 5.57682157e-01 1.97636023e-01 -3.39102969e-02
-2.42060721e-01 1.29292533e-01 7.47219622e-01 -6.95200026e-01
-2.63788909e-01 1.38455778e-01 -2.16456234e-01 -2.41771638e-01
-3.92435044e-01 -5.49691796e-01 -1.83604538e+00 -4.75211412e-01
-7.08162189e-02 -2.16347113e-01 2.39720017e-01 9.79117513e-01
5.65900922e-01 4.12393212e-01 2.81127170e-02 -1.31557107e+00
1.49385780e-01 -1.19657815e+00 -7.33852804e-01 6.45515263e-01
1.70655340e-01 -5.92993915e-01 -4.09805894e-01 7.45584548e-01]
|
[14.28128719329834, -3.1044530868530273]
|
c5ff7451-8493-4738-8f37-8b0b469a0198
|
multi-tasking-dialogue-comprehension-with
|
2110.03269
| null |
https://arxiv.org/abs/2110.03269v1
|
https://arxiv.org/pdf/2110.03269v1.pdf
|
Multi-tasking Dialogue Comprehension with Discourse Parsing
|
Multi-party dialogue machine reading comprehension (MRC) raises an even more challenging understanding goal on dialogue with more than two involved speakers, compared with the traditional plain passage style MRC. To accurately perform the question-answering (QA) task according to such multi-party dialogue, models have to handle fundamentally different discourse relationships from common non-dialogue plain text, where discourse relations are supposed to connect two far apart utterances in a linguistics-motivated way.To further explore the role of such unusual discourse structure on the correlated QA task in terms of MRC, we propose the first multi-task model for jointly performing QA and discourse parsing (DP) on the multi-party dialogue MRC task. Our proposed model is evaluated on the latest benchmark Molweni, whose results indicate that training with complementary tasks indeed benefits not only QA task, but also DP task itself. We further find that the joint model is distinctly stronger when handling longer dialogues which again verifies the necessity of DP in the related MRC.
|
['Hai Zhao', 'Zhuosheng Zhang', 'Yuchen He']
|
2021-10-07
| null |
https://aclanthology.org/2021.paclic-1.64
|
https://aclanthology.org/2021.paclic-1.64.pdf
|
paclic-2021-11
|
['discourse-parsing']
|
['natural-language-processing']
|
[ 4.89538640e-01 1.07588053e+00 4.68246222e-01 -3.98567498e-01
-1.31746209e+00 -7.54799724e-01 1.14499617e+00 3.39446336e-01
-3.30080837e-01 1.05684698e+00 9.37023580e-01 -7.15299487e-01
-1.18084468e-01 -5.05319059e-01 -5.04607379e-01 -5.15351593e-01
-5.84145226e-02 9.35201705e-01 4.15780604e-01 -1.12702084e+00
1.84844553e-01 -3.09584022e-01 -1.03400087e+00 8.06380749e-01
9.18471217e-01 5.02677619e-01 3.74377966e-01 1.06376255e+00
-1.84690848e-01 1.33313143e+00 -1.00026083e+00 -6.93815649e-01
-1.93939164e-01 -6.44973457e-01 -1.78421545e+00 3.87047976e-02
3.68713349e-01 -1.96495637e-01 1.06780946e-01 6.79625988e-01
5.13720810e-01 2.01444104e-01 7.47226000e-01 -6.23739362e-01
-2.39715129e-01 9.71526921e-01 -1.48345158e-01 3.29841197e-01
1.06518102e+00 1.33481696e-01 1.49351490e+00 -3.71576369e-01
6.27406776e-01 1.65274751e+00 4.77497756e-01 5.51482379e-01
-8.05431247e-01 3.64357471e-01 4.10525501e-02 3.94328892e-01
-6.64642811e-01 -5.59768081e-01 6.47751927e-01 -8.08462352e-02
1.18460083e+00 6.17339075e-01 1.71520822e-02 1.38253689e+00
-1.72451884e-02 9.20857370e-01 1.17718673e+00 -6.05282187e-01
-1.20418094e-01 -6.21572696e-02 5.54224014e-01 3.67043167e-01
-4.57960814e-01 -5.95769107e-01 -2.69683927e-01 -1.69761941e-01
-7.51184300e-02 -1.15326071e+00 -5.87895811e-01 2.03406945e-01
-1.45766878e+00 9.99650717e-01 -1.44206300e-01 5.89149415e-01
-1.65755585e-01 -3.94230038e-01 8.44958544e-01 7.65990257e-01
3.29325676e-01 7.60294616e-01 -8.67637694e-01 -4.50349092e-01
-2.51970619e-01 3.99545908e-01 1.43793023e+00 7.86753654e-01
2.40329549e-01 -5.83661377e-01 -5.55284917e-01 1.08753395e+00
2.19449848e-01 2.86782056e-01 3.14701170e-01 -1.11667883e+00
1.20392144e+00 4.83678311e-01 4.43957858e-02 -6.06393337e-01
-7.85505354e-01 -6.07382990e-02 -9.36121464e-01 -5.44390261e-01
1.18297851e+00 -4.94481325e-01 -3.43731195e-02 1.73127365e+00
4.59569216e-01 -5.58032513e-01 8.04858744e-01 5.55199742e-01
1.11824918e+00 7.95272827e-01 5.26592843e-02 -4.91268545e-01
1.84757745e+00 -1.12370515e+00 -9.92624760e-01 -1.38494551e-01
1.04773355e+00 -8.67033839e-01 1.21508741e+00 2.45271593e-01
-1.35363245e+00 -3.34551841e-01 -9.16970074e-01 -6.59810722e-01
-6.94094300e-02 -6.04034439e-02 3.15084398e-01 5.13199568e-01
-8.68968606e-01 3.11664253e-01 -2.28956208e-01 -2.17528790e-01
4.71773744e-02 -5.61954603e-02 -1.75957769e-01 2.04057004e-02
-1.72142124e+00 1.54826629e+00 3.24205428e-01 2.98135996e-01
-3.86539489e-01 -3.47501576e-01 -1.02139378e+00 1.97574664e-02
6.37455344e-01 -6.47494674e-01 1.79161966e+00 -6.27369821e-01
-2.03295374e+00 9.15928364e-01 -2.33046636e-01 -6.35749817e-01
6.59006357e-01 -3.65511656e-01 -1.23824932e-01 1.76705793e-01
-9.58894491e-02 3.16073298e-01 6.21309459e-01 -1.15046430e+00
-5.31580687e-01 -3.22815090e-01 5.85512698e-01 7.74566710e-01
2.46431336e-01 1.13944024e-01 8.57732967e-02 -2.61305291e-02
-2.47169867e-01 -6.82910204e-01 7.56973624e-02 -8.69332612e-01
-7.44884133e-01 -9.19542074e-01 6.05683506e-01 -1.06673563e+00
1.28572810e+00 -1.76391697e+00 7.13914573e-01 -3.63335758e-01
1.26114339e-01 3.73905063e-01 -1.69172615e-01 9.05772209e-01
7.00074434e-02 -1.40294522e-01 -5.16922474e-01 -4.94026005e-01
3.00189793e-01 4.62508649e-01 -3.54280382e-01 2.58740246e-01
4.19957638e-01 1.06400764e+00 -7.55616009e-01 -5.21770120e-01
-3.87498289e-02 -1.44855425e-01 -1.54264778e-01 4.83897060e-01
-7.75203049e-01 9.54844296e-01 -5.20206749e-01 2.67681807e-01
3.43212605e-01 -2.91832149e-01 5.28283477e-01 4.70319912e-02
-2.32715104e-02 8.72170627e-01 -6.44789696e-01 1.74226737e+00
-5.99048316e-01 7.19295919e-01 2.04890132e-01 -1.16050398e+00
5.82426488e-01 7.19193578e-01 -2.01333910e-01 -8.81564975e-01
2.72146493e-01 8.87176245e-02 7.91293681e-01 -9.02387083e-01
7.24848270e-01 -4.93684635e-02 -3.83397490e-01 5.90159178e-01
7.84071013e-02 -3.05724353e-01 2.36416444e-01 3.28215510e-01
1.14698887e+00 -9.35786068e-02 6.66430414e-01 -4.40059513e-01
1.30431116e+00 8.67097676e-02 -1.78706553e-02 7.24708915e-01
-1.72782093e-01 3.15603197e-01 1.02194583e+00 3.06010395e-02
-8.21731627e-01 -4.54332560e-01 -3.49738777e-01 1.37408602e+00
-1.57566424e-02 -2.10612312e-01 -9.09407556e-01 -9.75689828e-01
-4.79140818e-01 1.05367720e+00 -3.97106826e-01 3.66607666e-01
-1.21351695e+00 -8.65384817e-01 9.65409994e-01 -1.49547264e-01
6.43710673e-01 -1.15304029e+00 -5.52202523e-01 3.75829250e-01
-8.16127419e-01 -1.38299811e+00 6.37675598e-02 9.56087410e-02
-4.35593814e-01 -1.27111816e+00 -8.52308512e-01 -7.74112761e-01
-2.61058547e-02 -1.40685514e-01 1.50092757e+00 2.79405028e-01
4.47550654e-01 4.89836484e-01 -7.45918214e-01 -3.48150313e-01
-1.22802460e+00 5.73812485e-01 -5.31593204e-01 -3.54836583e-01
1.42559916e-01 -2.90736675e-01 -1.66211441e-01 2.23921746e-01
-6.03259444e-01 2.79230505e-01 2.73162663e-01 1.02451777e+00
-3.18251878e-01 -4.43906367e-01 1.06567812e+00 -1.08928025e+00
1.17324734e+00 -5.30694962e-01 -4.25041281e-02 7.15089262e-01
2.33683929e-01 2.16467574e-01 6.82763636e-01 -1.00748643e-01
-1.69728196e+00 -5.60527921e-01 -7.17100203e-01 9.39331532e-01
-3.03091198e-01 5.26916027e-01 -4.89430159e-01 4.17760998e-01
6.16155624e-01 -1.30680355e-03 2.65124030e-02 -3.70967090e-01
6.82767332e-01 7.02333987e-01 3.84130388e-01 -8.61548305e-01
5.60418963e-01 2.77969218e-03 -7.95702115e-02 -1.14346504e+00
-1.07154107e+00 -5.09647369e-01 -9.16155398e-01 4.57790541e-03
1.17879844e+00 -7.25163162e-01 -9.44397509e-01 4.12508518e-01
-1.84182227e+00 -5.90751588e-01 -8.22105184e-02 2.02644877e-02
-7.88461268e-01 9.28824365e-01 -8.42794240e-01 -9.01204884e-01
-3.13399196e-01 -1.01222932e+00 1.06186604e+00 -7.88875893e-02
-5.47748268e-01 -1.46212626e+00 2.79283673e-01 1.02554941e+00
2.54414797e-01 8.28070268e-02 1.47346282e+00 -1.24110126e+00
-2.76966393e-01 3.57039869e-01 -1.89611614e-01 3.61562222e-01
1.08531959e-01 -4.19494092e-01 -1.13688099e+00 4.01043780e-02
5.17264605e-01 -8.69193792e-01 5.13376832e-01 -1.23705432e-01
4.01649743e-01 -3.19151133e-01 2.84974068e-01 -3.60744059e-01
8.15023661e-01 -5.48857190e-02 7.76704252e-01 3.44173580e-01
5.04649818e-01 1.46399760e+00 8.60573888e-01 7.10214674e-02
9.93809581e-01 6.50244713e-01 2.63161004e-01 3.81556191e-02
-5.20435013e-02 1.68203041e-01 4.89457011e-01 1.47678292e+00
-8.82958770e-02 -6.03418589e-01 -1.04207373e+00 4.62084621e-01
-2.00577235e+00 -8.42027783e-01 -8.30579817e-01 1.73952317e+00
1.31081986e+00 -1.47719188e-02 -7.21917301e-02 6.39591217e-02
3.40498120e-01 5.49366474e-01 -7.49275833e-02 -6.85362101e-01
-4.88648117e-01 1.91773653e-01 -1.32758260e-01 8.93124104e-01
-1.00685740e+00 8.46419036e-01 5.46113682e+00 8.07572663e-01
-4.54063147e-01 2.99587071e-01 6.58542991e-01 6.04683876e-01
-1.12542748e-01 -8.34833980e-02 -6.67943001e-01 -5.21862619e-02
1.16104925e+00 -8.32328498e-02 -7.08173960e-03 3.02850157e-01
1.04748078e-01 -5.62951088e-01 -1.50120962e+00 3.28927487e-01
9.68723074e-02 -1.11412895e+00 -5.63173834e-03 -6.78643212e-02
3.59782666e-01 -2.43518487e-01 -2.97799826e-01 6.29258156e-01
2.44885892e-01 -1.16613865e+00 2.75961369e-01 2.28718475e-01
2.49092966e-01 -3.78859788e-01 1.07858813e+00 9.30603981e-01
-6.09360456e-01 1.27503738e-01 -2.16707438e-01 -2.89662570e-01
6.66085005e-01 -2.09012330e-02 -9.73993301e-01 1.06460381e+00
1.17825828e-01 1.78026482e-01 -5.18645108e-01 2.64384866e-01
-3.84680629e-01 5.54656506e-01 -3.73868719e-02 -1.96815342e-01
4.41852927e-01 -2.45783657e-01 8.68526518e-01 1.38111675e+00
-1.58530951e-01 2.87216157e-01 -1.51466295e-01 3.00385475e-01
1.93033461e-02 3.04457068e-01 -3.70606482e-01 2.59625584e-01
7.04666600e-02 1.02920401e+00 -2.40263537e-01 -4.55197990e-01
-5.82965732e-01 9.50320780e-01 4.77156132e-01 1.54018719e-02
-5.57975054e-01 3.28899249e-02 2.00794071e-01 -3.62734735e-01
2.23919287e-01 -3.05890769e-01 -8.50209743e-02 -1.09468448e+00
1.56725332e-01 -1.41472280e+00 3.36442739e-01 -3.14860553e-01
-1.46578705e+00 7.81962156e-01 3.93160433e-02 -7.96027601e-01
-6.08155310e-01 -5.17449856e-01 -7.11170077e-01 9.50625539e-01
-1.82526529e+00 -1.15089214e+00 1.12958081e-01 6.30413949e-01
1.03844774e+00 1.18178830e-01 1.28397310e+00 1.21571742e-01
-4.41378891e-01 3.18026066e-01 -2.29397276e-03 2.53147483e-02
8.16269279e-01 -1.72526455e+00 3.34666550e-01 4.65904146e-01
-1.99031103e-02 4.29082394e-01 9.12506402e-01 -3.20134073e-01
-1.32016885e+00 -5.22731423e-01 1.31633222e+00 -9.75383103e-01
9.32143688e-01 -4.50897783e-01 -1.37558103e+00 5.37483513e-01
1.03569138e+00 -7.83655345e-01 7.60017753e-01 4.76885021e-01
-2.66740676e-02 5.52154720e-01 -8.33402097e-01 5.15078008e-01
8.06375206e-01 -6.63684130e-01 -1.67626333e+00 7.93990076e-01
1.09246933e+00 -7.17740595e-01 -1.16977227e+00 3.36462229e-01
-3.95104326e-02 -9.77956057e-01 8.04331899e-01 -6.61617696e-01
6.12633109e-01 3.99348475e-02 -1.94991484e-01 -1.27367020e+00
5.20615637e-01 -8.34263206e-01 1.49287442e-02 1.23213029e+00
7.42823422e-01 -5.48762321e-01 1.08880691e-01 5.44702232e-01
-3.24064225e-01 -5.18769205e-01 -1.33424258e+00 -4.18503284e-01
6.92961574e-01 -2.48461112e-01 2.07927823e-01 9.42459404e-01
5.51198423e-01 1.28159189e+00 -3.73349965e-01 1.09057322e-01
2.47339770e-01 -1.55808598e-01 9.06789780e-01 -1.17602003e+00
-6.33900046e-01 -1.74175382e-01 3.59428018e-01 -1.62330985e+00
4.25361395e-01 -5.26417673e-01 2.32570738e-01 -1.50854552e+00
-3.45215917e-01 -3.50384653e-01 6.67183280e-01 -1.06016450e-01
-4.57431376e-01 -3.72964054e-01 2.45605707e-01 1.25611320e-01
-1.01839685e+00 8.69833827e-01 1.53399253e+00 2.63475459e-02
-1.86298713e-01 1.94579422e-01 -6.32941067e-01 8.31337273e-01
5.37552655e-01 2.16358360e-02 -4.55097467e-01 -4.91251409e-01
1.94685489e-01 8.24785471e-01 1.10905385e-02 -3.79824132e-01
3.35852087e-01 2.43158981e-01 -4.28689748e-01 -4.75240350e-01
4.55745608e-01 -4.16876793e-01 -5.58181465e-01 1.41467720e-01
-6.59010887e-01 7.77427014e-03 9.94842574e-02 4.14926857e-01
-5.25331616e-01 -5.23264349e-01 3.80155325e-01 -3.27725202e-01
-3.85383517e-01 -5.99402368e-01 -7.22192109e-01 5.82209408e-01
7.17608690e-01 1.59772351e-01 -9.31114018e-01 -6.12787664e-01
-9.32821691e-01 5.71663797e-01 -2.25930303e-01 3.30824196e-01
7.30307624e-02 -5.50775290e-01 -1.11903179e+00 -5.33265710e-01
-5.56975901e-02 4.42583710e-01 4.93359119e-01 1.04442894e+00
-3.36089045e-01 8.69126141e-01 5.20378463e-02 -6.04345560e-01
-1.46834528e+00 7.64041543e-02 2.84912020e-01 -1.05091715e+00
-6.44225121e-01 7.48975337e-01 1.44980460e-01 -7.54471004e-01
2.35171035e-01 -3.37013721e-01 -9.50077474e-01 4.88195121e-01
5.60056210e-01 3.75806451e-01 2.56723374e-01 -8.37450325e-01
8.49697068e-02 1.29526913e-01 -3.16926956e-01 -1.92045718e-01
1.10313928e+00 -6.85233712e-01 -4.14223343e-01 7.30559707e-01
9.73180771e-01 6.08943887e-02 -9.07648742e-01 -4.81809527e-01
6.10447228e-01 2.27749139e-01 -6.33087099e-01 -1.00281978e+00
4.36519273e-02 9.65256810e-01 -2.60192961e-01 8.64587009e-01
5.85795999e-01 1.54327944e-01 8.57997000e-01 8.15614939e-01
3.02645624e-01 -1.09299815e+00 6.62464276e-02 1.26920569e+00
1.41903281e+00 -1.40541935e+00 -7.17793629e-02 -5.76189935e-01
-1.09728730e+00 1.24493086e+00 4.37642813e-01 2.04282105e-01
1.82289660e-01 -2.95837134e-01 3.16657186e-01 -4.58902866e-01
-9.67253864e-01 -2.37335399e-01 2.23337635e-01 5.72980404e-01
6.27659440e-01 -3.66923623e-02 -5.98098516e-01 4.09509450e-01
-5.44013262e-01 -8.49637270e-01 7.32639194e-01 7.61546671e-01
-3.05738270e-01 -1.43505800e+00 -2.75193483e-01 2.96292454e-02
-4.55666989e-01 -2.31871292e-01 -6.50794983e-01 1.11006224e+00
-3.08887571e-01 1.36633682e+00 -8.39699358e-02 2.69735754e-01
4.25481617e-01 3.42283487e-01 4.70048308e-01 -6.99144542e-01
-1.12917948e+00 -2.36448288e-01 1.18470907e+00 -3.31405461e-01
-9.52157438e-01 -7.08483756e-01 -1.11855328e+00 -1.47786826e-01
-3.00019234e-01 3.46979767e-01 4.25182700e-01 1.58939433e+00
-2.22862326e-02 6.52705193e-01 3.73271376e-01 -4.92792994e-01
-9.85647619e-01 -1.50992811e+00 -1.09126821e-01 3.46494108e-01
5.67426980e-01 -2.36399740e-01 -3.98399651e-01 -2.58035690e-01]
|
[12.18153190612793, 8.13184928894043]
|
9ae156ac-de1a-4eaf-a236-7c96bf133641
|
topnet-structural-point-cloud-decoder
| null | null |
http://openaccess.thecvf.com/content_CVPR_2019/html/Tchapmi_TopNet_Structural_Point_Cloud_Decoder_CVPR_2019_paper.html
|
http://openaccess.thecvf.com/content_CVPR_2019/papers/Tchapmi_TopNet_Structural_Point_Cloud_Decoder_CVPR_2019_paper.pdf
|
TopNet: Structural Point Cloud Decoder
|
3D point cloud generation is of great use for 3D scene modeling and understanding. Real-world 3D object point clouds can be properly described by a collection of low-level and high-level structures such as surfaces, geometric primitives, semantic parts,etc. In fact, there exist many different representations of a 3D object point cloud as a set of point groups. Existing frameworks for point cloud genera-ion either do not consider structure in their proposed solutions, or assume and enforce a specific structure/topology,e.g. a collection of manifolds or surfaces, for the generated point cloud of a 3D object. In this work, we pro-pose a novel decoder that generates a structured point cloud without assuming any specific structure or topology on the underlying point set. Our decoder is softly constrained to generate a point cloud following a hierarchical rooted tree structure. We show that given enough capacity and allowing for redundancies, the proposed decoder is very flexible and able to learn any arbitrary grouping of points including any topology on the point set. We evaluate our decoder on the task of point cloud generation for 3D point cloud shape completion. Combined with encoders from existing frameworks, we show that our proposed decoder significantly outperforms state-of-the-art 3D point cloud completion methods on the Shapenet dataset
|
[' Silvio Savarese', ' Ian Reid', ' Hamid Rezatofighi', ' Vineet Kosaraju', 'Lyne P. Tchapmi']
|
2019-06-01
| null | null | null |
cvpr-2019-6
|
['point-cloud-completion']
|
['computer-vision']
|
[ 6.37236461e-02 2.46354535e-01 1.64792374e-01 -2.82831877e-01
-6.24749482e-01 -8.60856533e-01 7.46715665e-01 2.81824559e-01
3.22209716e-01 1.24819949e-02 -1.90580770e-01 -3.78190964e-01
1.74525052e-01 -1.17253733e+00 -1.37759626e+00 -3.68384391e-01
-7.25544989e-02 1.15410793e+00 2.72934198e-01 -3.12778056e-01
3.22239697e-01 1.15402842e+00 -1.60756731e+00 1.24634400e-01
7.18189478e-01 7.63045728e-01 5.05571842e-01 6.03676140e-01
-6.49454772e-01 -7.59378076e-02 -4.05864507e-01 -3.09043169e-01
6.16800606e-01 2.10977882e-01 -7.28445113e-01 6.81173086e-01
7.05988824e-01 -1.96539894e-01 -1.91298738e-01 9.07521307e-01
2.15397120e-01 -3.35872978e-01 8.11868310e-01 -1.24204803e+00
-4.42302167e-01 -4.84161952e-04 -3.96400541e-01 -6.14764810e-01
2.31688991e-01 -6.77608773e-02 7.81060994e-01 -1.28406501e+00
8.51842880e-01 1.34792280e+00 5.68340242e-01 4.10880953e-01
-1.18025589e+00 -4.08232957e-01 8.98890495e-02 -3.14346403e-01
-1.56682122e+00 -2.29022682e-01 1.02802634e+00 -5.64729750e-01
9.87436414e-01 2.66723573e-01 9.04198706e-01 4.40851539e-01
8.70860741e-02 4.36474949e-01 4.59594905e-01 -3.24062347e-01
3.40728343e-01 -9.70466062e-02 7.22605586e-02 8.02379608e-01
3.00509214e-01 -3.21432710e-01 -2.10774794e-01 -4.68371332e-01
1.19982123e+00 2.85629392e-01 6.76172748e-02 -9.86045718e-01
-1.17614031e+00 6.58088028e-01 6.72068238e-01 -9.54863336e-03
-2.66579658e-01 3.44710767e-01 1.15376934e-02 1.99275166e-01
5.50573766e-01 3.42847295e-02 -3.69608283e-01 3.43801409e-01
-6.78294599e-01 4.13179398e-01 8.60943019e-01 1.75804305e+00
1.26235247e+00 -1.08778849e-01 2.20007226e-01 3.85433227e-01
4.90015537e-01 6.83381617e-01 -3.53424698e-01 -1.06092834e+00
6.06173635e-01 9.78769422e-01 1.38744950e-01 -8.82480502e-01
-1.35804325e-01 -2.66094208e-01 -8.34403872e-01 4.05505687e-01
-1.70015186e-01 3.90654653e-01 -1.04750836e+00 1.42588305e+00
6.39513016e-01 4.80468780e-01 -1.34466542e-02 7.35149384e-01
9.11622226e-01 7.25305200e-01 -4.54957843e-01 1.94234893e-01
1.35219717e+00 -3.66642833e-01 -8.62641167e-03 9.81721357e-02
5.63244879e-01 -6.61468029e-01 6.36331081e-01 1.13564797e-01
-1.35278761e+00 -4.49083656e-01 -9.38907683e-01 -4.97905761e-01
-8.39122478e-03 -9.68159810e-02 5.69443762e-01 4.32669640e-01
-1.24903846e+00 4.80574518e-01 -8.78273606e-01 -1.75269186e-01
5.14397442e-01 5.20056188e-01 -5.04273713e-01 -1.54429391e-01
-3.33591312e-01 5.72546899e-01 1.52445257e-01 -1.43435910e-01
-8.84956717e-01 -7.95633972e-01 -8.05498481e-01 8.20350945e-02
1.81916431e-01 -1.49804437e+00 1.08063722e+00 -4.57844406e-01
-1.07100964e+00 1.21804714e+00 -3.76975149e-01 -3.15109789e-01
2.62978315e-01 8.39168280e-02 1.33414179e-01 1.93191811e-01
-2.00716443e-02 9.38035190e-01 9.37327087e-01 -1.77671993e+00
-3.66657674e-01 -6.41256690e-01 2.82602757e-01 3.89979035e-01
3.72764975e-01 -2.72505641e-01 -7.15464771e-01 -2.81170338e-01
7.79757380e-01 -1.11108053e+00 -2.86146224e-01 4.85895991e-01
-6.67067409e-01 -3.69494706e-01 1.02787650e+00 -2.25169837e-01
4.42677647e-01 -2.22258520e+00 2.77434617e-01 3.99528116e-01
4.64976490e-01 -5.96625097e-02 -3.14864554e-02 4.67833370e-01
-8.00528601e-02 4.90709037e-01 -6.67091072e-01 -8.16838384e-01
1.48102224e-01 5.83207548e-01 -4.36358452e-01 4.97939080e-01
3.53264093e-01 6.72996104e-01 -8.24517548e-01 -3.31209153e-01
5.13514936e-01 7.79667079e-01 -9.53658342e-01 8.43758658e-02
-6.36929691e-01 4.49982762e-01 -6.74265802e-01 5.37077010e-01
1.08091331e+00 -3.11499268e-01 -2.32145384e-01 -1.57304764e-01
-1.72202185e-01 4.86009300e-01 -1.28019822e+00 2.16572094e+00
-4.98718828e-01 1.77772924e-01 2.76194960e-01 -6.86004281e-01
1.23839688e+00 4.54112500e-01 7.49129057e-01 1.84689939e-01
-1.06359266e-01 4.10237670e-01 -2.78136432e-01 1.31562993e-01
5.85503101e-01 -6.57264888e-02 1.88530814e-02 8.99005085e-02
-1.25161380e-01 -1.01892066e+00 -4.87233847e-01 3.77262831e-01
1.10515225e+00 7.48188049e-02 -4.24185619e-02 -1.91789076e-01
2.93881893e-01 2.35619754e-01 2.59184986e-01 4.18829918e-01
6.55769527e-01 1.10202241e+00 3.01238656e-01 -2.50360548e-01
-1.52355862e+00 -1.40729475e+00 -3.63252014e-01 1.16020218e-02
4.24090356e-01 -5.36050916e-01 -5.71516752e-01 -2.35345617e-01
1.89836502e-01 6.49291754e-01 -1.21293195e-01 4.53621820e-02
-5.94472229e-01 -7.68689215e-02 1.44941524e-01 -1.97416861e-02
1.14277437e-01 -6.70819640e-01 -4.72062618e-01 1.58521995e-01
3.47729325e-02 -1.18719602e+00 -5.13545394e-01 9.89611726e-03
-1.43169558e+00 -1.01098526e+00 -3.80552471e-01 -1.02050018e+00
1.19401944e+00 7.77233303e-01 1.38290489e+00 3.69625121e-01
8.93506706e-02 5.29675126e-01 -2.37168297e-01 -4.10508573e-01
-5.36276221e-01 -4.63847183e-02 -5.91501743e-02 -5.19731827e-02
4.12097685e-02 -9.17222977e-01 -3.72711152e-01 1.54000625e-01
-1.11393559e+00 4.64013100e-01 2.40115523e-01 2.69279212e-01
1.18452835e+00 -1.04744518e-02 -6.95913658e-02 -7.10960865e-01
6.26739711e-02 -4.98369217e-01 -6.31713450e-01 -8.81803874e-03
1.37004599e-01 1.62526533e-01 3.78577888e-01 2.57329904e-02
-4.76043165e-01 4.67141837e-01 -2.98306465e-01 -9.90917623e-01
-2.39397511e-01 2.64956295e-01 -4.81559217e-01 -1.92467257e-01
4.21495080e-01 1.91218764e-01 -3.95529382e-02 -8.36746156e-01
5.55369914e-01 3.94588739e-01 4.07648712e-01 -8.13201606e-01
1.11497033e+00 9.29754555e-01 5.36828101e-01 -1.02961278e+00
-3.09533060e-01 -4.96637851e-01 -1.19982517e+00 -1.75814778e-02
8.12446237e-01 -1.08820069e+00 -5.87267041e-01 2.04372734e-01
-1.89884412e+00 3.54965031e-02 -3.64834934e-01 9.36731175e-02
-9.58182037e-01 4.72137570e-01 -1.38243124e-01 -4.63925213e-01
-3.48150223e-01 -1.25239122e+00 1.74370396e+00 -5.27040422e-01
8.43212306e-02 -7.95429170e-01 -1.48795530e-01 4.87041250e-02
-1.71313927e-01 5.95312297e-01 1.22351050e+00 -1.59887031e-01
-1.39998245e+00 -8.02006796e-02 3.63044068e-02 9.99289379e-02
1.62711024e-01 1.65900677e-01 -5.40430725e-01 -1.75782502e-01
1.03697419e-01 2.58045375e-01 5.35738587e-01 1.93684474e-01
1.27215052e+00 -2.85369635e-01 -4.66901094e-01 1.03153884e+00
1.57895029e+00 -1.34013057e-01 5.91069341e-01 -1.44511342e-01
1.02969766e+00 4.30263489e-01 1.63674019e-02 5.89845300e-01
6.14075065e-01 8.93123031e-01 1.02754819e+00 5.80644980e-02
-1.35588914e-01 -5.47857821e-01 1.15631185e-01 1.01944840e+00
-9.88922864e-02 -1.78894743e-01 -1.01570237e+00 4.15660799e-01
-1.57630765e+00 -6.09414577e-01 -6.99726999e-01 2.43170166e+00
4.72986907e-01 -6.70308173e-02 -9.58063528e-02 8.99088904e-02
7.96958447e-01 -1.21078789e-01 -5.38981497e-01 -2.64056772e-01
-1.39337897e-01 3.27765316e-01 5.98474562e-01 4.99684662e-01
-6.82861090e-01 8.72392237e-01 5.42233753e+00 5.45520484e-01
-8.55151832e-01 -9.66779236e-03 -2.63888147e-02 9.31820050e-02
-8.65402341e-01 3.74184191e-01 -7.66839385e-01 1.37079358e-01
4.25046861e-01 -2.82890767e-01 2.80038565e-01 7.55622566e-01
1.75418667e-02 4.90497589e-01 -1.38995111e+00 1.17300200e+00
-9.80493799e-02 -1.67046332e+00 6.12478137e-01 4.26366508e-01
6.44596100e-01 2.87956625e-01 -1.24269389e-01 -2.26997495e-01
3.60125512e-01 -8.52875590e-01 1.13018942e+00 4.25188720e-01
9.12855864e-01 -5.91198742e-01 1.65920798e-02 7.64527977e-01
-1.23879409e+00 5.19421458e-01 -6.98172212e-01 1.12969257e-01
3.12402129e-01 6.67643189e-01 -9.74257112e-01 9.41652238e-01
3.02468091e-01 8.15585911e-01 -1.95508182e-01 1.24176610e+00
-4.27512638e-02 3.13568294e-01 -7.51797259e-01 2.89427131e-01
1.81421474e-01 -4.84786093e-01 9.26353931e-01 6.62292480e-01
8.05113792e-01 3.57820243e-01 3.18859965e-01 1.02006054e+00
-2.11750165e-01 -4.75268811e-02 -1.12659895e+00 3.88915896e-01
6.88429892e-01 9.27984059e-01 -6.81197166e-01 -3.97405952e-01
-4.01620209e-01 7.00262725e-01 2.33741656e-01 9.30723399e-02
-5.71893454e-01 7.97774047e-02 9.14706469e-01 4.64337140e-01
3.66287649e-01 -1.00460434e+00 -7.61456251e-01 -1.14870715e+00
1.12238750e-01 -3.56896698e-01 -1.73157871e-01 -1.20778096e+00
-1.04959261e+00 4.27962691e-01 5.17960936e-02 -1.65451443e+00
1.28237113e-01 -5.39064765e-01 -4.39611048e-01 1.01194143e+00
-1.35767949e+00 -1.13558817e+00 -2.83847779e-01 7.33886063e-01
4.58637774e-01 1.38963759e-01 7.11377978e-01 9.87734273e-03
1.63569257e-01 -2.64235348e-01 -1.65480614e-01 -2.19292283e-01
-9.52677336e-03 -1.19817472e+00 1.04479730e+00 5.68319201e-01
3.68776143e-01 6.60306871e-01 4.10833508e-01 -7.96574950e-01
-2.05666161e+00 -1.36171651e+00 8.12964320e-01 -8.21383357e-01
1.76488236e-01 -8.04529130e-01 -1.00540781e+00 7.68654823e-01
-2.79277742e-01 -1.65666327e-01 1.86845660e-01 -1.93907589e-01
-3.06218922e-01 1.51474148e-01 -1.14744985e+00 6.04004443e-01
1.48583710e+00 -4.18262273e-01 -6.60050750e-01 7.55792081e-01
1.05499876e+00 -8.61349344e-01 -8.41021836e-01 4.99069273e-01
-4.13880199e-02 -7.57770360e-01 1.37101376e+00 -4.64522272e-01
4.36741829e-01 -6.66463554e-01 -3.31106007e-01 -1.20076203e+00
-2.58573294e-01 -5.82221806e-01 -1.07182883e-01 7.20000148e-01
1.35836482e-01 -2.44989291e-01 1.00982642e+00 3.98522437e-01
-8.95056427e-01 -4.60149139e-01 -1.11644423e+00 -7.04729319e-01
1.39686018e-01 -8.71729970e-01 1.02667665e+00 7.53815651e-01
-5.78294754e-01 3.19919467e-01 -1.12688933e-02 7.61372864e-01
8.88525784e-01 3.64641994e-01 1.28275490e+00 -1.49006701e+00
1.61300242e-01 -3.97364944e-01 -6.27161205e-01 -1.59758806e+00
1.45611227e-01 -1.40449393e+00 -1.43140525e-01 -1.96001196e+00
-3.24137419e-01 -1.05613077e+00 5.10124624e-01 2.56373882e-01
4.35316145e-01 1.98841188e-03 3.39558244e-01 4.74560231e-01
-8.68703723e-02 7.03209221e-01 1.55598891e+00 -1.18016377e-01
-2.33881518e-01 1.41441196e-01 -4.42755610e-01 6.51636183e-01
5.12772381e-01 -4.68868464e-01 -2.46697813e-01 -1.01222014e+00
2.85184801e-01 3.74641955e-01 7.05530643e-01 -8.04472506e-01
1.76713079e-01 -3.20173847e-03 -7.34688193e-02 -1.10685647e+00
9.57260489e-01 -1.24046636e+00 6.94964290e-01 2.75571018e-01
2.76262879e-01 7.16341585e-02 3.22402641e-02 5.06503999e-01
-4.42804694e-02 -2.00920224e-01 4.44858134e-01 -3.01535994e-01
-3.58804047e-01 8.87670159e-01 2.57624954e-01 -3.75287950e-01
9.73562717e-01 -6.13356471e-01 -3.76535058e-02 -1.18043646e-01
-5.83919942e-01 1.92359641e-01 1.11692739e+00 5.31566441e-01
1.00198865e+00 -1.50935030e+00 -8.10627639e-01 4.26099896e-01
2.54503161e-01 9.91994858e-01 5.85470088e-02 2.05707625e-01
-8.30072999e-01 3.74036252e-01 3.39955091e-02 -1.17579150e+00
-1.04488528e+00 4.32263404e-01 2.98677683e-01 4.09542441e-01
-1.06566954e+00 4.92495298e-01 5.59883058e-01 -6.98589027e-01
-1.17221177e-01 -8.59123409e-01 3.60089511e-01 -5.17160952e-01
-1.70818985e-01 -2.91720051e-02 2.56013572e-01 -9.82822180e-01
-2.05388695e-01 1.02924848e+00 4.30400431e-01 6.31218478e-02
1.33227706e+00 2.75779933e-01 -2.72710860e-01 3.74822468e-01
1.12075543e+00 -3.55225289e-03 -1.16912568e+00 -2.75795460e-01
-2.01892540e-01 -6.22522175e-01 -1.29792526e-01 -1.53385336e-02
-7.78365135e-01 9.46954668e-01 -1.33044660e-01 2.35333130e-01
7.06885874e-01 4.75769579e-01 6.42776728e-01 4.51337636e-01
1.08921456e+00 -2.59015560e-01 -2.55502164e-01 8.01947057e-01
1.18725264e+00 -7.40952313e-01 -1.34818211e-01 -1.20215845e+00
-1.11262493e-01 1.08613682e+00 7.64854178e-02 -5.58173716e-01
9.10864532e-01 6.67073727e-02 -5.44671237e-01 -4.83870924e-01
-8.52419972e-01 -2.20536456e-01 3.48006904e-01 7.19386458e-01
5.64137362e-02 2.23828480e-01 1.74511775e-01 8.25328529e-02
-6.04735494e-01 -1.99104220e-01 5.84153295e-01 8.05757821e-01
-6.17972732e-01 -1.29603577e+00 -6.01955414e-01 4.56114650e-01
1.23176269e-01 -3.28239352e-02 -2.92262554e-01 6.94144011e-01
2.29966745e-01 5.77290356e-01 5.52737057e-01 -3.19844753e-01
5.69087088e-01 -2.06910431e-01 7.15442479e-01 -1.23256361e+00
-1.71088785e-01 -5.58221899e-02 -2.79070407e-01 -1.90751553e-01
-4.28360790e-01 -8.93718481e-01 -1.52600706e+00 -4.43586618e-01
-1.41145587e-01 -1.67222153e-02 1.16554892e+00 6.76335454e-01
7.69302905e-01 5.07852808e-02 7.92377889e-01 -1.15554154e+00
-1.61860332e-01 -4.28124726e-01 -6.32752657e-01 4.53905344e-01
3.62747252e-01 -6.56004727e-01 -1.71343118e-01 2.42313281e-01]
|
[8.46279525756836, -3.501328945159912]
|
21271248-2aaa-45b5-93a5-0ca226043d65
|
infotec-centrogeo-at-semeval-2020-task-8-deep
| null | null |
https://aclanthology.org/2020.semeval-1.151
|
https://aclanthology.org/2020.semeval-1.151.pdf
|
Infotec + CentroGEO at SemEval-2020 Task 8: Deep Learning and Text Categorization approach for Memes classification
|
The information shared on social media is increasingly important; both images and text, and maybe the most popular combination of these two kinds of data are the memes. This manuscript describes our participation in Memotion task at SemEval 2020. This task is about to classify the memes in several categories related to the emotional content of them. For the proposed system construction, we used different strategies, and the best ones were based on deep neural networks and a text categorization algorithm. We obtained results analyzing the text and images separately, and also in combination. Our better performance was achieved in Task A, related to polarity classification.
|
['Mario Graff', "Tania Ram{\\'\\i}rez-delReal", "Sabino Miranda-Jim{\\'e}nez", 'Daniela Moctezuma', 'Eric S. Tellez', 'Guillermo Ruiz']
|
2020-12-01
| null | null | null |
semeval-2020
|
['text-categorization']
|
['natural-language-processing']
|
[-2.14756504e-01 2.92207003e-02 -2.62856446e-02 -2.40501717e-01
-1.17858753e-01 -5.52623272e-01 9.21424448e-01 7.27489591e-01
-6.44639492e-01 9.33624327e-01 2.69616038e-01 2.99418867e-01
1.67525575e-01 -8.79498422e-01 -2.51059443e-01 -5.17640650e-01
2.44427308e-01 3.83152425e-01 2.15954348e-01 -4.34774458e-01
8.45813572e-01 1.95892558e-01 -1.81650651e+00 7.12333500e-01
6.05317831e-01 1.26416230e+00 8.84999335e-02 5.20709813e-01
-6.93030715e-01 1.27510130e+00 -5.59212923e-01 -5.92836618e-01
-3.48064035e-01 -4.19352561e-01 -1.00509751e+00 -7.85286874e-02
4.02518928e-01 1.43829137e-01 -1.64156020e-01 9.83261883e-01
4.15064067e-01 2.13751748e-01 9.19426978e-01 -1.10309589e+00
-5.38232207e-01 7.67995894e-01 -4.66871679e-01 4.64668602e-01
4.10382181e-01 -5.33316255e-01 8.66796076e-01 -1.12162888e+00
1.11852348e+00 1.07812178e+00 4.35308337e-01 3.03181410e-01
-6.96959257e-01 -3.98603290e-01 -1.79117933e-01 6.74908578e-01
-1.14618504e+00 -4.09699559e-01 9.65488613e-01 -8.63573790e-01
7.03835368e-01 2.07242295e-01 9.34704244e-01 1.46849477e+00
6.18990600e-01 8.72631431e-01 1.48193991e+00 -3.89475018e-01
9.12600756e-02 9.14917529e-01 5.30152798e-01 3.78190219e-01
-5.48782274e-02 -7.32497334e-01 -6.17755830e-01 -7.90013969e-02
-2.00743750e-01 -1.86900690e-01 -1.40516609e-01 9.08212513e-02
-1.09496486e+00 8.21283221e-01 3.47498804e-01 1.02248740e+00
-2.00498030e-01 -8.97195712e-02 6.82673454e-01 5.68134248e-01
1.15403771e+00 6.17389858e-01 3.01478468e-02 -7.03979135e-02
-9.00261283e-01 -4.54600714e-02 1.08688068e+00 3.44400972e-01
7.82126784e-01 -3.57376695e-01 -1.15787715e-01 1.10574710e+00
2.86469944e-02 4.37229812e-01 7.80655146e-01 -3.94026041e-01
4.45941418e-01 6.39989018e-01 -6.48464859e-02 -1.84991479e+00
-8.56027186e-01 -2.98144072e-01 -8.39136183e-01 -8.88115168e-02
9.45905149e-02 -4.29442436e-01 -5.24575353e-01 1.42875874e+00
-6.91307262e-02 -4.51420009e-01 -1.78616211e-01 5.08514225e-01
1.28993344e+00 9.91321802e-01 4.78206240e-02 -4.52139556e-01
1.22735918e+00 -7.58468807e-01 -1.13071954e+00 -1.49260953e-01
4.56781596e-01 -7.91070640e-01 6.48352146e-01 4.10040945e-01
-1.22049642e+00 -5.40164948e-01 -1.20832384e+00 -8.80935639e-02
-1.01155496e+00 2.32217863e-01 1.62003949e-01 3.28575611e-01
-1.31809640e+00 7.56502986e-01 -3.32659893e-02 -7.33900726e-01
2.88876444e-01 1.02064706e-01 -5.14751732e-01 5.68911970e-01
-1.40086365e+00 1.18355441e+00 3.90087783e-01 -9.42617729e-02
-2.02614471e-01 1.64838403e-01 -4.41485763e-01 1.99963283e-02
1.13826342e-01 -3.90992045e-01 7.03709304e-01 -1.66379118e+00
-1.32725286e+00 1.62903225e+00 7.68273473e-02 -4.43944782e-01
4.92376596e-01 -1.18481345e-01 -5.98704875e-01 3.87328386e-01
-9.93478596e-02 7.50050426e-01 9.94201779e-01 -1.16313076e+00
-3.87363672e-01 -2.42213234e-01 -1.23986304e-01 1.94725007e-01
-1.15896201e+00 1.88879699e-01 -4.68348823e-02 -5.18798351e-01
-8.37236643e-02 -1.01258314e+00 2.47275159e-01 -3.77237946e-01
-3.89013261e-01 -3.82226825e-01 8.36366177e-01 -7.09472895e-01
1.19465566e+00 -2.14463162e+00 2.87068993e-01 1.97278172e-01
5.89100540e-01 -1.73060700e-01 1.21095173e-01 6.29901707e-01
-2.10453764e-01 3.12767357e-01 6.25002533e-02 -3.57564539e-01
-4.90956232e-02 -4.57100481e-01 -3.33445281e-01 3.64110887e-01
-3.72087397e-02 6.85522556e-01 -6.77030325e-01 -8.49187970e-01
-1.83557689e-01 3.13786715e-01 1.16863385e-01 -4.45538834e-02
-1.77019507e-01 3.74253392e-01 -2.85371929e-01 2.10350588e-01
4.46582705e-01 -2.28967175e-01 2.40662977e-01 -2.90621877e-01
-2.74318963e-01 1.38071731e-01 -4.97179478e-01 1.11524403e+00
-2.77675629e-01 1.45910037e+00 -1.21248916e-01 -9.59558249e-01
1.24030268e+00 3.96539003e-01 7.86914706e-01 -9.27637517e-01
7.15419471e-01 4.81613018e-02 -1.21224254e-01 -9.01356399e-01
1.00612152e+00 -8.24015215e-02 -1.36336043e-01 7.43373990e-01
4.15699817e-02 -3.36325802e-02 5.30909181e-01 3.60165745e-01
6.93227112e-01 -3.55890602e-01 4.12370086e-01 -7.11600840e-01
6.85132086e-01 -1.20638892e-01 -9.06697586e-02 2.80517280e-01
-3.62090707e-01 5.59898019e-01 1.07757092e+00 -5.47177732e-01
-1.03450322e+00 -3.82139713e-01 -2.91975111e-01 1.18037629e+00
2.97816217e-01 -3.58936787e-01 -6.46898031e-01 -6.14934027e-01
-3.67808074e-01 4.56708699e-01 -9.80758548e-01 -9.16684568e-02
-3.20584536e-01 -7.93879807e-01 3.35205086e-02 3.94481458e-02
5.72271287e-01 -1.41498733e+00 -3.17195356e-01 2.24631820e-02
-6.10162377e-01 -1.04876149e+00 -2.56720502e-02 1.93915799e-01
-5.90176165e-01 -9.40129101e-01 -8.90620828e-01 -9.50460255e-01
4.63380307e-01 9.92400572e-02 1.18449867e+00 -2.43951406e-04
6.52281195e-02 4.13995743e-01 -5.76267540e-01 -5.05855143e-01
-5.00030756e-01 3.67730051e-01 -1.47972569e-01 5.33489883e-01
1.53293997e-01 -4.03244674e-01 -3.49964321e-01 8.01912621e-02
-8.37857902e-01 6.06075265e-02 2.81119317e-01 5.65907061e-01
1.96616113e-01 -6.34568408e-02 5.81100702e-01 -1.09423709e+00
8.55113685e-01 -8.51724029e-01 1.07354499e-01 -4.91657481e-02
-2.87150353e-01 -2.63467997e-01 4.64081138e-01 -3.34351629e-01
-9.24266100e-01 -1.97365507e-01 -2.03678861e-01 3.36819999e-02
-1.22785382e-03 6.04571223e-01 2.31158048e-01 -1.67165939e-02
8.35549831e-01 -2.81748250e-02 -2.23003745e-01 -2.38066673e-01
-1.75379336e-01 9.61422563e-01 1.91247948e-02 -1.68957397e-01
1.08769417e-01 4.49654937e-01 -1.83765784e-01 -1.08228481e+00
-7.55413234e-01 -3.38473022e-01 -2.94271559e-01 -8.21958303e-01
1.00885820e+00 -6.47964954e-01 -5.50571203e-01 8.52827311e-01
-1.24463964e+00 -4.12032828e-02 -2.75086090e-02 2.53637522e-01
-3.77301127e-01 3.04633528e-01 -7.82529950e-01 -8.06567371e-01
-3.38184506e-01 -6.72929168e-01 5.16514301e-01 7.45352432e-02
-4.39524025e-01 -9.22634363e-01 4.10942346e-01 3.00660372e-01
4.19264644e-01 2.90465713e-01 9.47574914e-01 -7.78374016e-01
8.37614462e-02 -2.58691758e-01 -3.19921106e-01 2.86341012e-01
-2.90573001e-01 2.04819337e-01 -1.23464894e+00 -4.38955612e-02
2.06279784e-01 -5.91074288e-01 1.30585051e+00 2.60985583e-01
1.38132906e+00 -4.20754626e-02 -3.94372553e-01 -2.00767294e-01
1.17479920e+00 2.93565005e-01 8.59151781e-01 4.42925215e-01
4.97868627e-01 1.16848433e+00 4.91954565e-01 5.19199252e-01
3.58096063e-01 6.88009501e-01 4.23024416e-01 -1.34918243e-02
5.24121746e-02 2.44707197e-01 4.85763967e-01 1.28348732e+00
-2.08133548e-01 -7.11113572e-01 -1.11438119e+00 5.62408686e-01
-1.94296777e+00 -1.22367191e+00 -5.33320367e-01 1.66755712e+00
4.32204723e-01 1.46616057e-01 2.62931228e-01 2.58627355e-01
1.13494146e+00 4.18704629e-01 -1.13833047e-01 -6.71872437e-01
-6.47056878e-01 -2.29268923e-01 -3.77466157e-02 2.98108816e-01
-1.45350385e+00 4.61282164e-01 6.67802095e+00 8.69801700e-01
-1.41445529e+00 3.93742949e-01 1.02296638e+00 -1.70592546e-01
-1.20745704e-01 -5.62803745e-01 -3.33547711e-01 7.63771057e-01
1.08664858e+00 -2.36046046e-01 -2.21567415e-02 7.23999381e-01
-1.40873507e-01 -5.62293470e-01 -7.26588309e-01 1.10425091e+00
6.00090921e-01 -1.33813250e+00 -1.75043881e-01 -1.95713937e-01
9.36841190e-01 7.99379647e-02 1.72293916e-01 1.84527233e-01
-5.95772743e-01 -6.52391732e-01 9.63937581e-01 8.46001506e-01
3.65579039e-01 -5.33829033e-01 1.01599824e+00 1.46475956e-01
-7.29905367e-01 -6.54228851e-02 -1.40313610e-01 -1.91932410e-01
-1.30552799e-01 8.87606442e-01 -3.54020864e-01 8.17753077e-02
8.32427979e-01 1.23082304e+00 -7.45987773e-01 7.90284157e-01
-1.96839254e-02 3.41887653e-01 -2.49936841e-02 -8.00023019e-01
-6.62563741e-02 -1.89320240e-02 7.14702189e-01 1.55425191e+00
3.73233169e-01 -2.35074401e-01 -1.16881080e-01 5.13758004e-01
-3.89184237e-01 7.11189091e-01 -6.50717020e-01 -3.72653037e-01
-4.52965870e-02 1.81975329e+00 -1.46125603e+00 -6.65464878e-01
-4.29249406e-02 9.00439084e-01 4.08536911e-01 -1.68076623e-02
-8.67869914e-01 -5.85141361e-01 -2.03613281e-01 2.22257048e-01
-9.89720970e-02 2.84109637e-02 -3.79100442e-01 -1.32503450e+00
3.88025343e-02 -4.43893641e-01 3.44004691e-01 -1.13414633e+00
-1.53608394e+00 1.14823306e+00 -1.45610571e-01 -1.16208065e+00
1.29521519e-01 -6.90672159e-01 -4.90109980e-01 3.69374812e-01
-1.04745209e+00 -7.93500483e-01 -5.43040454e-01 5.59532166e-01
3.35449636e-01 -2.69315153e-01 5.95113099e-01 5.34547687e-01
-4.02590424e-01 7.97842368e-02 1.52471304e-01 -2.36306354e-01
9.90857780e-01 -1.17182100e+00 -3.05451930e-01 2.84190595e-01
5.99037036e-02 -1.24107614e-01 7.08923042e-01 -4.63401377e-01
-7.07984626e-01 -5.72418690e-01 1.32369030e+00 -4.08510834e-01
8.66539121e-01 -3.84702593e-01 -6.95804179e-01 3.00362796e-01
8.16921890e-01 -5.88266194e-01 7.33793020e-01 1.77735671e-01
-9.00218077e-03 5.07413112e-02 -9.70425725e-01 4.14658189e-01
6.43712342e-01 -6.08288109e-01 -6.27030969e-01 4.71807808e-01
2.14380711e-01 -9.94119346e-02 -6.54871702e-01 1.78030897e-02
4.59310442e-01 -1.19318819e+00 4.08349305e-01 -2.14098275e-01
1.30853784e+00 1.98879391e-02 -1.39428005e-02 -1.45726252e+00
-2.31706530e-01 -6.22904189e-02 -2.30169389e-02 1.23376584e+00
4.93560284e-01 -6.06353045e-01 6.75138652e-01 -1.97617393e-02
5.94603829e-02 -4.72963601e-01 -6.56560600e-01 -1.64437488e-01
-2.92218104e-03 -2.65317619e-01 -6.67422786e-02 1.40383840e+00
5.04358947e-01 8.98895621e-01 -5.18575728e-01 -7.34593928e-01
-2.08620448e-02 1.58444494e-01 3.41275185e-01 -1.49646747e+00
3.81665602e-02 -7.93457866e-01 -6.07498169e-01 -2.26956517e-01
4.23775971e-01 -9.69015181e-01 4.25129477e-03 -1.53935897e+00
6.15356684e-01 -2.13029627e-02 -4.30067092e-01 2.54399061e-01
8.91056582e-02 7.89610863e-01 3.69293213e-01 3.39833289e-01
-8.75556946e-01 3.04195613e-01 1.05091560e+00 -5.49044788e-01
-1.41065806e-01 -3.78854692e-01 -5.91295123e-01 8.34425449e-01
8.55899870e-01 -4.09301877e-01 -6.60457299e-04 -2.93978900e-02
1.17375600e+00 -3.32253985e-02 2.05577463e-01 -1.19776773e+00
1.06021501e-01 1.97886810e-01 4.38174009e-01 -8.75044644e-01
5.68765879e-01 -6.12462699e-01 2.61804760e-01 3.31657350e-01
-5.77765703e-01 -2.79032767e-01 1.89866960e-01 1.68538749e-01
-5.40531337e-01 -4.55096066e-01 1.04354644e+00 -3.82954143e-02
-7.36215651e-01 2.63560228e-02 -9.37912345e-01 -1.51663676e-01
1.06793523e+00 -9.09790583e-03 -6.77465558e-01 -7.97056556e-01
-1.41816497e+00 -1.85873061e-01 1.79132372e-01 4.94957477e-01
3.47004235e-01 -1.26996541e+00 -7.90710807e-01 -3.92243296e-01
1.55630633e-01 -1.02112710e+00 4.58082646e-01 1.19024754e+00
-5.46862781e-01 2.35099092e-01 -7.72097349e-01 -3.03867191e-01
-1.39610219e+00 5.39122820e-01 3.19854677e-01 -3.85091752e-01
-1.79852575e-01 5.77099979e-01 1.51237234e-01 -3.77094932e-02
3.97734977e-02 2.05299795e-01 -1.20144820e+00 1.09192848e+00
6.46654546e-01 5.28741717e-01 1.47226825e-01 -9.46913183e-01
-3.17941040e-01 7.09942460e-01 3.83811668e-02 -4.81215455e-02
1.40618992e+00 -3.76996845e-01 -7.06440270e-01 1.10517883e+00
1.31558740e+00 1.28824800e-01 -2.59424686e-01 1.58534586e-01
-6.39538467e-02 -5.94996288e-02 2.48264194e-01 -7.87523687e-01
-1.18074203e+00 1.06341982e+00 5.97281933e-01 1.21995568e+00
9.58116710e-01 -1.05164304e-01 6.65107012e-01 6.45947754e-01
1.70829356e-01 -1.72850919e+00 3.79772365e-01 7.89888740e-01
1.17990041e+00 -1.34213746e+00 1.90857023e-01 -3.25403154e-01
-7.21571088e-01 1.52879584e+00 4.95128632e-01 -1.94859937e-01
7.76937842e-01 -1.20946176e-01 5.37769943e-02 -4.54951704e-01
-7.17252195e-01 -1.26294447e-02 4.18315887e-01 7.87897781e-02
7.43178368e-01 -1.01783939e-01 -1.02006745e+00 6.42484486e-01
-4.68956158e-02 -1.51511177e-01 8.90931010e-01 7.71934152e-01
-6.14449203e-01 -7.62264490e-01 -4.51398194e-01 6.14439368e-01
-6.74530268e-01 1.95243955e-01 -1.04169178e+00 4.84038234e-01
2.71341711e-01 9.62174416e-01 3.48368198e-01 -8.05925250e-01
2.65928786e-02 1.33660674e-01 2.95443118e-01 -2.32306987e-01
-7.80152142e-01 -2.66963154e-01 5.90588272e-01 -2.65426010e-01
-9.06317174e-01 -6.24532223e-01 -8.09655190e-01 -6.21551752e-01
-1.69539511e-01 1.98570251e-01 1.01353061e+00 9.52785492e-01
1.62897557e-01 4.74438250e-01 9.65292931e-01 -1.14619136e+00
4.35863107e-01 -1.13288462e+00 -8.40730727e-01 6.15845680e-01
-5.16166398e-03 -5.92137039e-01 -4.98764604e-01 1.30216181e-01]
|
[8.513882637023926, 10.712678909301758]
|
2b629cfb-479f-4baf-9a21-f82abc0351d0
|
unifying-the-discrete-and-continuous-emotion
|
2210.16642
| null |
https://arxiv.org/abs/2210.16642v1
|
https://arxiv.org/pdf/2210.16642v1.pdf
|
Unifying the Discrete and Continuous Emotion labels for Speech Emotion Recognition
|
Traditionally, in paralinguistic analysis for emotion detection from speech, emotions have been identified with discrete or dimensional (continuous-valued) labels. Accordingly, models that have been proposed for emotion detection use one or the other of these label types. However, psychologists like Russell and Plutchik have proposed theories and models that unite these views, maintaining that these representations have shared and complementary information. This paper is an attempt to validate these viewpoints computationally. To this end, we propose a model to jointly predict continuous and discrete emotional attributes and show how the relationship between these can be utilized to improve the robustness and performance of emotion recognition tasks. Our approach comprises multi-task and hierarchical multi-task learning frameworks that jointly model the relationships between continuous-valued and discrete emotion labels. Experimental results on two widely used datasets (IEMOCAP and MSPPodcast) for speech-based emotion recognition show that our model results in statistically significant improvements in performance over strong baselines with non-unified approaches. We also demonstrate that using one type of label (discrete or continuous-valued) for training improves recognition performance in tasks that use the other type of label. Experimental results and reasoning for this approach (called the mismatched training approach) are also presented.
|
['Rita Singh', 'Bhiksha Raj', 'Hira Dhamyal', 'Roshan Sharma']
|
2022-10-29
| null | null | null | null |
['speech-emotion-recognition']
|
['speech']
|
[ 1.00630261e-01 -9.18231830e-02 -2.03957066e-01 -9.59565163e-01
-8.21106911e-01 -3.23218554e-01 6.35498226e-01 7.00095221e-02
-4.62053210e-01 2.66257137e-01 2.56565809e-01 1.88521463e-02
2.32047561e-04 -2.53924400e-01 -9.19509530e-02 -5.78420520e-01
9.37465578e-02 2.71506429e-01 -4.23971266e-01 -2.75722712e-01
-1.13910884e-02 3.48960042e-01 -1.87321019e+00 6.29739881e-01
5.12512445e-01 1.36334133e+00 -3.72890979e-01 6.33268118e-01
-2.27168396e-01 8.99580061e-01 -6.37998044e-01 -4.90498662e-01
6.39035832e-03 -4.16962415e-01 -8.54235709e-01 3.81007679e-02
2.29415104e-01 -5.82463667e-02 1.19669385e-01 8.64024460e-01
5.52456617e-01 3.33460659e-01 8.38892817e-01 -1.62145460e+00
-3.57591122e-01 2.94333071e-01 -2.53201485e-01 -1.41124263e-01
4.28249300e-01 -5.60162663e-01 1.17657471e+00 -8.97569716e-01
2.37487704e-01 1.31622291e+00 7.11150169e-01 6.76200092e-01
-9.71710920e-01 -6.49663627e-01 3.06603372e-01 3.07840437e-01
-1.28843677e+00 -6.71212554e-01 1.01568842e+00 -3.34194958e-01
1.16616964e+00 3.36805493e-01 3.31682056e-01 1.56377161e+00
-1.61681369e-01 9.54350352e-01 1.47595489e+00 -6.20158136e-01
7.56314695e-02 6.14469588e-01 3.58763307e-01 3.94969493e-01
-6.39959931e-01 1.23368502e-01 -6.90519929e-01 -2.47203544e-01
1.76492080e-01 -3.91614377e-01 1.53148007e-02 -8.49263091e-03
-8.47405493e-01 8.69443536e-01 -2.11995319e-01 5.79087675e-01
-4.06646878e-01 -1.61224604e-01 8.52997959e-01 6.21930659e-01
7.79784203e-01 2.70092875e-01 -5.12139440e-01 -6.23650610e-01
-8.69767308e-01 -2.68584132e-01 1.03986764e+00 7.04779029e-01
4.96342123e-01 3.63719881e-01 -1.58123419e-01 1.42060006e+00
2.68333495e-01 1.56446442e-01 7.09191144e-01 -8.28718305e-01
-7.83672035e-02 2.94329315e-01 1.43106543e-02 -1.16125488e+00
-6.01251304e-01 -1.64613679e-01 -6.51378512e-01 7.15107992e-02
1.31246120e-01 -2.18565241e-01 -7.19316244e-01 1.93512821e+00
9.00223032e-02 2.59023756e-01 5.22114098e-01 8.57936084e-01
9.06911671e-01 6.66565061e-01 3.58242542e-01 -2.75331438e-01
1.57533085e+00 -1.03930640e+00 -1.05841231e+00 -3.04990947e-01
7.05410779e-01 -6.16547048e-01 1.08962178e+00 6.89376593e-01
-8.22203100e-01 -5.98445117e-01 -8.80900562e-01 4.01780233e-02
-6.77698255e-01 4.02764410e-01 7.12292016e-01 1.09968722e+00
-8.87425184e-01 2.66763926e-01 -5.82712770e-01 -2.45152414e-01
-3.53743024e-02 1.52287900e-01 -2.98202902e-01 3.98289382e-01
-1.40406001e+00 1.13515306e+00 1.76403552e-01 1.13874830e-01
-5.53840458e-01 -2.48026028e-01 -8.55395675e-01 3.80488411e-02
1.78313196e-01 -1.39322385e-01 1.30304229e+00 -1.46604228e+00
-1.77952313e+00 1.04233134e+00 -3.04172605e-01 -2.76549220e-01
1.16828248e-01 -1.70675263e-01 -9.05108869e-01 2.84626693e-01
-1.89942718e-01 6.64369226e-01 8.87660861e-01 -1.25883377e+00
-5.83509982e-01 -2.73455262e-01 -1.00324444e-01 2.36917630e-01
-7.69950449e-01 5.81272662e-01 -8.59534517e-02 -5.44958234e-01
1.03996219e-02 -8.01452100e-01 1.75009519e-01 -3.38613778e-01
-6.52756616e-02 -6.60861790e-01 8.00802708e-01 -5.43751359e-01
1.11099315e+00 -2.31923079e+00 1.14627421e-01 2.19765261e-01
-2.78822947e-02 1.43561125e-01 -2.52226055e-01 4.08170849e-01
-3.13979626e-01 3.51197034e-01 4.35944684e-02 -7.46009052e-01
2.78221726e-01 3.62275749e-01 -1.78363845e-01 1.82571009e-01
2.21399233e-01 6.09012723e-01 -6.45683467e-01 -3.94306451e-01
2.26773441e-01 7.45354831e-01 -2.08459884e-01 3.05272758e-01
2.45377898e-01 3.45686734e-01 -2.02702388e-01 6.55297220e-01
3.17481399e-01 1.65421888e-01 1.94015279e-01 -3.72751832e-01
7.24422038e-02 3.41168195e-01 -1.04678822e+00 1.39421165e+00
-7.54932642e-01 6.34051919e-01 2.71001697e-01 -1.27625263e+00
1.28248441e+00 8.68972421e-01 6.45684838e-01 -6.12440705e-01
3.05843472e-01 8.15785378e-02 -2.90187113e-02 -6.56277597e-01
5.51611066e-01 -6.04811847e-01 -4.54894572e-01 3.25903356e-01
2.03027606e-01 -2.76359111e-01 -9.06778127e-02 -1.20312616e-01
7.95461655e-01 -4.76246737e-02 3.85564506e-01 1.27218217e-01
6.00313723e-01 -3.43265980e-01 8.30834627e-01 6.28035665e-01
-6.80421472e-01 2.88227290e-01 5.09306133e-01 -2.53329694e-01
-7.29887962e-01 -6.45286977e-01 -1.75378963e-01 1.50199044e+00
-2.11344555e-01 -4.27091032e-01 -3.78717542e-01 -5.49331963e-01
-2.08194599e-01 7.72827029e-01 -6.18849993e-01 -3.54545474e-01
-1.11603849e-01 -5.77013791e-01 1.09505570e+00 5.36114633e-01
2.34221488e-01 -1.10701525e+00 -6.66403592e-01 9.50682238e-02
-3.53637099e-01 -1.42518389e+00 1.23788364e-01 6.62599266e-01
-2.61054516e-01 -5.42960584e-01 -3.36626440e-01 -6.32875264e-01
1.23886973e-01 2.93588433e-02 9.81481075e-01 -1.87127590e-01
-3.47048044e-02 9.57793832e-01 -9.57692683e-01 -5.01705229e-01
-4.68582422e-01 -2.34964445e-01 1.88520998e-01 4.45364386e-01
6.76797926e-01 -5.27804196e-01 8.09118599e-02 3.09639961e-01
-8.78784537e-01 -7.75973797e-02 3.09462726e-01 9.68063116e-01
2.26734921e-01 -1.34014621e-01 1.03751314e+00 -5.79717457e-01
1.00668395e+00 -5.59183478e-01 5.91071695e-02 3.44669312e-01
-5.43271959e-01 -9.01896507e-02 4.73895341e-01 -5.81143439e-01
-1.18231213e+00 2.55128928e-02 -3.30094904e-01 -8.12259138e-01
-6.32508695e-01 7.43210912e-01 -1.91098481e-01 5.51524423e-02
1.82038873e-01 9.85510051e-02 1.15243718e-01 -2.59328485e-01
3.80982727e-01 1.14912605e+00 3.49940002e-01 -8.02824140e-01
1.52786346e-02 2.28245065e-01 -3.37284654e-01 -9.13660049e-01
-8.12234759e-01 -6.08438134e-01 -5.72709620e-01 -5.43900311e-01
8.50662470e-01 -9.29040790e-01 -7.96528399e-01 4.96529132e-01
-1.03112972e+00 -1.19826414e-01 -5.20041436e-02 6.60013199e-01
-6.57510877e-01 3.37966561e-01 -7.51461804e-01 -1.37215245e+00
-1.34604499e-01 -9.81833518e-01 1.14937472e+00 4.75348756e-02
-6.73329234e-01 -1.20601571e+00 -1.88114330e-01 3.87123466e-01
4.75996256e-01 1.30632773e-01 9.13888693e-01 -1.20179415e+00
5.99990964e-01 -2.48699337e-01 4.04486656e-02 6.74108028e-01
4.00455631e-02 8.06998909e-02 -1.35795391e+00 -6.57718927e-02
3.20974052e-01 -8.88317347e-01 7.46114850e-01 2.71389075e-02
1.05488384e+00 -3.99519987e-02 4.08730209e-02 3.18767667e-01
8.36801171e-01 4.50320840e-01 3.54583263e-01 2.32428104e-01
3.63422960e-01 1.09266424e+00 6.96386933e-01 6.60761833e-01
5.94702125e-01 8.13510537e-01 2.73342609e-01 -2.72096004e-02
2.17422947e-01 1.98589027e-01 5.95386744e-01 1.15491581e+00
1.94036573e-01 -2.16349795e-01 -8.61115694e-01 4.82214510e-01
-1.77107275e+00 -1.01196599e+00 -6.83250651e-02 1.76338184e+00
7.97850430e-01 -1.10380337e-01 2.54561394e-01 3.31043392e-01
6.61264002e-01 2.65482217e-01 -4.59368318e-01 -1.19305968e+00
-3.99937242e-01 1.70259833e-01 -1.97032869e-01 3.84677470e-01
-1.30638790e+00 1.02916861e+00 6.53270817e+00 7.38036871e-01
-1.28072822e+00 2.32084528e-01 6.26941919e-01 -2.02841431e-01
-1.23008564e-01 -3.48068118e-01 -5.14332712e-01 1.33851081e-01
1.32810640e+00 -1.09077089e-01 2.60358423e-01 8.92340899e-01
1.40865386e-01 7.83086121e-02 -1.21626937e+00 1.24157369e+00
4.16593492e-01 -4.56848294e-01 -3.55551571e-01 -1.17374450e-01
2.68523067e-01 -2.56859511e-01 1.42470568e-01 7.16210306e-01
1.44070521e-01 -9.49192643e-01 8.64261210e-01 4.75096256e-01
5.01682580e-01 -6.91726923e-01 7.03779995e-01 3.04140151e-01
-1.08869207e+00 -4.76545952e-02 -8.55841041e-02 -1.99875757e-01
1.53995544e-01 2.91716605e-01 -4.97401834e-01 6.60649061e-01
5.62168837e-01 7.53607035e-01 -2.55579412e-01 4.57583725e-01
-1.09228507e-01 6.51770830e-01 -1.72684506e-01 3.88064012e-02
2.24101126e-01 -3.19558561e-01 3.64872545e-01 1.56626093e+00
2.86997080e-01 6.62779883e-02 3.01671028e-01 6.71766162e-01
9.34038833e-02 2.89497405e-01 -5.36595941e-01 -2.99620390e-01
5.65616488e-01 1.52810216e+00 -5.55016458e-01 -4.75593925e-01
-6.25833273e-01 1.16340816e+00 3.41417372e-01 3.52113843e-01
-9.16142881e-01 -2.83207774e-01 8.86150002e-01 -7.99923420e-01
4.43593673e-02 -8.20456222e-02 -2.85823137e-01 -1.03743160e+00
-1.18792996e-01 -9.91433620e-01 4.82456058e-01 -7.30204880e-01
-1.58918309e+00 8.54635417e-01 -2.14386862e-02 -1.10334206e+00
-5.29513299e-01 -7.42057145e-01 -4.26994354e-01 7.37380981e-01
-1.49348271e+00 -1.20675874e+00 -1.62923485e-01 5.73509216e-01
4.53051656e-01 -1.39028952e-01 1.27809989e+00 3.58938426e-01
-6.85722053e-01 7.89932430e-01 -1.62904635e-01 2.06032529e-01
1.03050327e+00 -1.19362378e+00 -3.30284417e-01 3.95046800e-01
4.10104305e-01 3.15079510e-01 6.87993705e-01 -2.74530709e-01
-1.14196634e+00 -5.73928356e-01 9.48522925e-01 -2.14124784e-01
5.51797092e-01 -3.84492248e-01 -9.56243932e-01 5.59819996e-01
3.02229851e-01 -1.63042665e-01 1.28460348e+00 5.62627673e-01
-8.61460149e-01 -1.59480702e-02 -1.09379148e+00 2.22013414e-01
5.56337297e-01 -9.81891870e-01 -7.44352221e-01 4.88565266e-02
4.16018546e-01 -2.16196645e-02 -1.17430711e+00 6.46920383e-01
8.84111941e-01 -1.15920460e+00 7.54311085e-01 -7.68352687e-01
2.98919261e-01 2.34604135e-01 -4.99879926e-01 -1.53499174e+00
-5.03805932e-03 -3.04654479e-01 4.84635755e-02 1.35348320e+00
3.38085264e-01 -5.57790756e-01 4.08632070e-01 7.92115986e-01
-3.75948459e-01 -8.44635665e-01 -1.05659986e+00 -8.96995068e-01
1.28563136e-01 -9.21696067e-01 1.68418139e-01 1.34212446e+00
4.93213773e-01 5.26252210e-01 -6.30099833e-01 4.14651260e-02
-5.95724657e-02 1.73830427e-02 5.04682720e-01 -1.40629256e+00
-1.92165636e-02 -6.61910176e-01 -4.20124412e-01 -8.46842170e-01
8.44071805e-01 -8.56378317e-01 -9.47548356e-03 -1.21776807e+00
-1.59932166e-01 -3.63873303e-01 -5.97512722e-01 7.81807721e-01
-1.02462284e-02 2.24573649e-02 2.46542692e-01 6.34077266e-02
-5.67350745e-01 8.04264665e-01 4.34582144e-01 1.50371730e-01
-1.35536432e-01 -2.36613557e-01 -6.40694022e-01 8.14689875e-01
8.55255365e-01 -3.21841955e-01 -2.57271767e-01 -6.62234649e-02
2.23971158e-01 2.14346960e-01 2.34138951e-01 -8.73466015e-01
2.15082645e-01 8.03622510e-03 1.70975681e-02 -2.47995660e-01
1.12290740e+00 -9.41326439e-01 -2.49492452e-01 -9.62440223e-02
-6.56458735e-01 -4.40067379e-03 4.19397533e-01 3.88423115e-01
-6.40049994e-01 -2.03376427e-01 6.30199850e-01 1.99828863e-01
-7.64571786e-01 -2.00377867e-01 -7.40195274e-01 -1.16649777e-01
9.78042126e-01 -2.43357301e-01 -3.31485807e-03 -8.53654385e-01
-1.13670409e+00 1.32192388e-01 -1.69483587e-01 8.78773808e-01
7.28121281e-01 -1.41679335e+00 -6.31414771e-01 1.60108492e-01
2.65289128e-01 -1.02081740e+00 1.03860326e-01 9.99161601e-01
5.08085489e-01 2.55201757e-01 -3.15781564e-01 -4.00997818e-01
-1.62067795e+00 3.51738155e-01 5.00880241e-01 -1.63223341e-01
3.92329805e-02 8.64803970e-01 -1.54023945e-01 -7.24824548e-01
4.51191604e-01 -1.64609656e-01 -3.36438268e-01 7.43264019e-01
2.35450312e-01 2.10786492e-01 5.69624007e-02 -1.09561682e+00
-4.69017088e-01 3.16494465e-01 -8.07379931e-02 -5.25988042e-01
1.24809849e+00 -2.91527390e-01 -1.60207629e-01 1.24164653e+00
1.22327781e+00 -1.78421825e-01 -7.46407449e-01 -2.44849890e-01
2.90130317e-01 -2.34240338e-01 1.80300251e-01 -1.02025664e+00
-7.47827053e-01 1.13288522e+00 5.26060939e-01 5.00027478e-01
1.29610991e+00 -8.70431513e-02 3.87876749e-01 2.99482256e-01
2.09755406e-01 -1.32884812e+00 1.86791018e-01 6.66817725e-01
7.65627444e-01 -1.33950865e+00 -5.65404773e-01 -4.30049419e-01
-1.13753867e+00 1.23016727e+00 6.83110237e-01 3.29082936e-01
6.19192243e-01 3.19514006e-01 4.90857363e-01 -2.45118991e-01
-1.06916630e+00 -4.03765827e-01 3.90208751e-01 4.64834422e-01
8.50460351e-01 1.51111081e-01 -3.14707994e-01 9.59767878e-01
-9.60547477e-02 -3.62147331e-01 2.86465943e-01 8.37813139e-01
-3.11539948e-01 -1.34702957e+00 -3.04989398e-01 3.28628123e-01
-5.86578071e-01 -6.57925084e-02 -7.94407248e-01 6.39536679e-01
-1.88787188e-02 1.52960742e+00 7.01115653e-02 -6.97951913e-01
2.91318178e-01 7.85805345e-01 1.70535251e-01 -4.54542100e-01
-8.80730808e-01 3.15666318e-01 5.59426486e-01 -4.23347950e-01
-7.98111498e-01 -5.95347881e-01 -1.12117267e+00 1.59587383e-01
-2.44384065e-01 3.88768196e-01 8.27230334e-01 1.01497316e+00
3.49152476e-01 5.38583279e-01 7.57582664e-01 -6.45524204e-01
-5.53804874e-01 -1.11206079e+00 -6.73852682e-01 5.88063598e-01
1.65218487e-01 -7.30459690e-01 -5.96219838e-01 -4.76921983e-02]
|
[13.403241157531738, 5.747459411621094]
|
ad19bbb1-5e1c-4087-9882-7a6fb4969ce7
|
danna-sep-unite-to-separate-them-all
|
2112.03752
| null |
https://arxiv.org/abs/2112.03752v1
|
https://arxiv.org/pdf/2112.03752v1.pdf
|
Danna-Sep: Unite to separate them all
|
Deep learning-based music source separation has gained a lot of interest in the last decades. Most of the existing methods operate with either spectrograms or waveforms. Spectrogram based models learn suitable masks for separating magnitude spectrogram into different sources, and waveform-based models directly generate waveforms of individual sources. The two types of models have complementary strengths; the former is superior given harmonic sources such as vocals, while the latter demonstrates better results for percussion and bass instruments. In this work, we improved upon the state-of-the-art (SoTA) models and successfully combined the best of both worlds. The backbones of the proposed framework, dubbed Danna-Sep, are two spectrogram-based models including a modified X-UMX and U-Net, and an enhanced Demucs as the waveform-based model. Given an input of mixture, we linearly combined respective outputs from the three models to obtain the final result. We showed in the experiments that, despite its simplicity, Danna-Sep surpassed the SoTA models by a large margin in terms of Source-to-Distortion Ratio.
|
['Kin-Wai Cheuk', 'Chin-Yun Yu']
|
2021-12-07
| null | null | null | null |
['music-source-separation']
|
['music']
|
[ 9.54955444e-02 -5.21031916e-01 2.62342453e-01 3.65679823e-02
-1.07637775e+00 -6.57005072e-01 4.85343188e-01 -2.97589242e-01
3.51202935e-02 5.63147366e-01 3.08891237e-01 -4.50545922e-02
-4.30041790e-01 -5.31545281e-01 -5.19975722e-01 -8.38391125e-01
-2.20249236e-01 -1.43789008e-01 1.85799718e-01 -2.97494113e-01
1.00547634e-01 2.01985478e-01 -1.76069486e+00 3.47609997e-01
9.37920094e-01 1.23024631e+00 9.69767049e-02 8.09878409e-01
-2.14116782e-01 7.69645572e-01 -7.00664163e-01 -3.10218006e-01
4.34512705e-01 -8.76529872e-01 -1.68233603e-01 -2.80200034e-01
3.65783155e-01 -3.20922844e-02 -4.13715094e-01 1.04815364e+00
8.64432216e-01 2.27055237e-01 6.99663222e-01 -1.23592162e+00
-7.92568803e-01 1.24708343e+00 -5.69440365e-01 2.11048663e-01
7.46957511e-02 -1.08590961e-01 9.99142230e-01 -1.03460324e+00
-2.62117628e-02 1.17815042e+00 8.95289421e-01 2.63514012e-01
-1.04065585e+00 -9.14005280e-01 -1.35380045e-01 3.46812546e-01
-1.28925550e+00 -6.01031184e-01 1.19084299e+00 -3.41309309e-01
5.88840961e-01 3.35114449e-01 4.69796896e-01 1.00985909e+00
-1.77879661e-01 8.97974193e-01 1.22672534e+00 -4.71041322e-01
1.58284917e-01 2.42400263e-03 3.91080379e-02 2.72571385e-01
-1.33308426e-01 3.04073453e-01 -7.28254378e-01 -3.12611967e-01
7.27849722e-01 -3.93719733e-01 -5.13842940e-01 5.96460178e-02
-9.73163009e-01 4.91613328e-01 3.00367355e-01 5.88192046e-01
-2.63908356e-01 2.41278261e-01 1.22523718e-01 2.72950768e-01
2.65224159e-01 4.39261049e-01 -1.19628407e-01 -3.20013911e-01
-1.43333328e+00 2.97807723e-01 6.62920237e-01 6.55080438e-01
1.80972725e-01 8.57348263e-01 -1.53463170e-01 1.04310846e+00
1.77825153e-01 5.85830152e-01 7.58609593e-01 -7.44410276e-01
3.12907726e-01 -1.48085818e-01 1.47083744e-01 -1.05563116e+00
-3.33722889e-01 -9.87960696e-01 -8.69051874e-01 3.44846696e-01
4.34680313e-01 -2.83108860e-01 -9.79060650e-01 1.89956820e+00
-1.98786017e-02 8.46000433e-01 6.08808883e-02 1.07344639e+00
8.14946651e-01 1.01842582e+00 -4.04097766e-01 -2.16780856e-01
9.48996544e-01 -9.99197423e-01 -9.40510273e-01 1.23183638e-01
-2.63289511e-01 -1.23985612e+00 7.85502791e-01 7.97495365e-01
-1.46079111e+00 -1.05108297e+00 -1.42297554e+00 2.37565801e-01
-1.53814554e-01 3.86694402e-01 3.89948428e-01 9.89099920e-01
-1.11529195e+00 9.15961027e-01 -6.15461767e-01 1.25822768e-01
1.32605568e-01 4.00641449e-02 2.08725512e-01 4.78244781e-01
-1.27278149e+00 6.39348567e-01 1.32900864e-01 1.64332643e-01
-1.09304619e+00 -7.45888293e-01 -6.66510642e-01 4.08241332e-01
2.04942286e-01 -4.33555573e-01 1.36283183e+00 -7.37303674e-01
-2.05204034e+00 3.13679725e-01 -4.95819654e-03 -5.96791983e-01
3.66272211e-01 -4.41500545e-01 -9.61164355e-01 -4.13391814e-02
-3.77951980e-01 2.67412037e-01 1.17460668e+00 -1.39203894e+00
-6.68021381e-01 -7.12856427e-02 -1.89617977e-01 1.97295360e-02
-3.11668515e-01 6.66732639e-02 -1.09655388e-01 -1.08938551e+00
-6.80066005e-04 -7.30961084e-01 1.10993318e-01 -4.30128664e-01
-4.36394393e-01 -2.88277175e-02 6.73601091e-01 -6.53190255e-01
1.42629206e+00 -2.25422096e+00 3.05113077e-01 -7.83382282e-02
-4.86577787e-02 5.25514781e-01 -3.60911280e-01 6.14099264e-01
-3.39869589e-01 -4.86270264e-02 -2.88122386e-01 -4.67351824e-01
1.60450175e-01 -3.31352770e-01 -7.86820114e-01 3.39582056e-01
2.26138800e-01 4.23033655e-01 -8.19153309e-01 -1.13681436e-01
1.84646994e-01 6.88159704e-01 -3.73006344e-01 3.15104902e-01
-2.78542042e-02 1.48298547e-01 1.39939189e-01 3.36749673e-01
7.26935387e-01 3.09792757e-01 -1.31196936e-03 -2.50525922e-01
-2.13190511e-01 4.32998836e-01 -1.63621187e+00 1.89137590e+00
-5.08689046e-01 6.84585154e-01 3.93725753e-01 -6.40200019e-01
1.05521727e+00 8.63164604e-01 5.74189425e-01 -1.97182134e-01
2.27708548e-01 6.38688087e-01 3.74308825e-01 -3.45365614e-01
4.94780660e-01 -2.94325203e-01 1.80906400e-01 4.43105787e-01
5.10640740e-01 -3.58121783e-01 1.21639922e-01 -1.22080132e-01
6.41207218e-01 4.03871089e-01 1.81324288e-01 -1.11842304e-01
5.66457510e-01 -5.41252732e-01 4.44034368e-01 6.22333884e-01
-5.40501103e-02 9.36480224e-01 1.90990433e-01 1.54851511e-01
-6.77809775e-01 -1.42498052e+00 -4.51945961e-02 9.98997748e-01
6.68141395e-02 -3.32064509e-01 -7.93593287e-01 -1.84266806e-01
-7.95477852e-02 8.34743500e-01 -3.20114791e-01 -1.34062678e-01
-4.97478753e-01 -6.69207275e-01 1.03163493e+00 4.97989714e-01
2.80350983e-01 -9.28673804e-01 -3.22082907e-01 3.49040389e-01
-1.34070560e-01 -8.14404011e-01 -5.37274122e-01 2.97103494e-01
-5.38053691e-01 -7.19592154e-01 -1.20955670e+00 -6.27385974e-01
-1.63742065e-01 3.38045657e-01 6.95489049e-01 -5.84462285e-01
-1.60093829e-02 4.34319265e-02 -3.20066720e-01 -8.39498818e-01
-4.73312736e-01 -2.08031133e-01 4.08885419e-01 5.95883965e-01
-1.17776401e-01 -1.33963788e+00 -4.19831932e-01 1.44056126e-01
-9.49771345e-01 -2.11906433e-01 5.54573417e-01 5.34687936e-01
3.35428804e-01 3.90225440e-01 1.15981519e+00 -3.88395667e-01
9.49099541e-01 -5.15610635e-01 -3.52845222e-01 -1.12196654e-01
-3.63532484e-01 -8.20607468e-02 1.01740217e+00 -5.80814838e-01
-1.12890494e+00 -2.24126711e-01 -4.84041929e-01 -7.67332017e-01
-9.19552147e-02 5.40253997e-01 -2.64737725e-01 2.25861460e-01
7.28044748e-01 3.32566857e-01 -3.42898965e-01 -8.99172962e-01
5.15787125e-01 9.08278883e-01 1.11375856e+00 -5.18622220e-01
8.80749166e-01 2.47725502e-01 -2.84993649e-01 -7.51875639e-01
-6.60163224e-01 -4.82180089e-01 -3.04231465e-01 -1.39537752e-01
5.72764337e-01 -7.71791697e-01 -4.50324893e-01 8.23354900e-01
-1.18061697e+00 -1.57578103e-02 -3.53187501e-01 7.58932590e-01
-4.75353569e-01 1.78840339e-01 -6.65602684e-01 -1.24704587e+00
-4.14585322e-01 -9.60937977e-01 7.26169348e-01 4.52404022e-01
-6.97314516e-02 -7.45404601e-01 3.29183102e-01 -1.48821831e-01
6.24114752e-01 2.45441303e-01 8.54726374e-01 -6.75267875e-01
-4.43291329e-02 -1.13187708e-01 1.42148778e-01 6.43403530e-01
3.45280886e-01 -1.56399626e-02 -1.53119946e+00 -1.68434102e-02
4.11324084e-01 -1.45152465e-01 9.23425972e-01 5.50711095e-01
9.67026114e-01 3.45536172e-02 2.31992081e-01 6.76153600e-01
1.33846533e+00 6.20980620e-01 6.52517736e-01 -2.83969700e-01
5.00573397e-01 2.21815795e-01 -5.79225086e-03 3.43158364e-01
-8.21118280e-02 7.30960667e-01 3.36101145e-01 -1.25504121e-01
-5.95051169e-01 -3.66680205e-01 5.94657719e-01 1.37127125e+00
-3.06176513e-01 -4.54141796e-01 -3.80984783e-01 7.01819837e-01
-1.75331521e+00 -1.16294837e+00 -2.13489905e-01 2.13149643e+00
9.47670043e-01 1.38576567e-01 4.05561805e-01 7.50921607e-01
6.59245253e-01 4.10643935e-01 -4.45103168e-01 -3.60632211e-01
-2.22505897e-01 6.34389102e-01 2.42032539e-02 3.90544415e-01
-1.09606540e+00 3.75528216e-01 6.37648821e+00 1.26749229e+00
-1.25353956e+00 2.17389628e-01 3.37556563e-02 -2.42550194e-01
-3.13261539e-01 -3.08462977e-01 -2.61633366e-01 4.93507057e-01
9.64999795e-01 -3.40441853e-01 7.00311244e-01 7.22303689e-01
1.94749132e-01 3.07266235e-01 -1.06114733e+00 1.07501137e+00
2.15987355e-01 -1.09255147e+00 -5.00777215e-02 -3.24634314e-01
7.35278070e-01 -2.06452280e-01 3.29433709e-01 3.58931690e-01
1.29801184e-01 -1.11952829e+00 1.18399203e+00 3.91681612e-01
7.51373827e-01 -8.45465839e-01 5.88777125e-01 3.70553404e-01
-1.36146641e+00 -8.12835023e-02 -2.39748523e-01 -1.64708167e-01
3.00377697e-01 6.95708036e-01 -3.55157852e-01 9.95915055e-01
3.81018579e-01 5.12043893e-01 -8.88962522e-02 1.42098308e+00
-2.69150913e-01 1.18171883e+00 -2.80344814e-01 2.51071811e-01
3.05924177e-01 -1.62664250e-01 9.51720238e-01 1.38233626e+00
6.01569951e-01 -1.01265252e-01 -5.38826175e-03 1.05270934e+00
-7.32569173e-02 -4.79656346e-02 -1.62206247e-01 -7.85037428e-02
4.36455160e-01 1.38255131e+00 -4.88755375e-01 -2.36463144e-01
-1.72120005e-01 6.26555979e-01 -2.66353577e-01 3.94311279e-01
-1.11354029e+00 -7.82965064e-01 6.08620465e-01 -5.86363710e-02
3.84545535e-01 -1.55429929e-01 -3.46853763e-01 -1.02602255e+00
-1.89627275e-01 -1.02603829e+00 1.63412213e-01 -9.23820496e-01
-1.31195307e+00 8.72155547e-01 -1.17282964e-01 -1.57791913e+00
-2.13936150e-01 -5.21386087e-01 -8.83384824e-01 1.11493552e+00
-1.39064372e+00 -8.65942419e-01 -1.92560628e-01 4.63412404e-01
6.01616025e-01 -2.19974458e-01 9.03195739e-01 6.06948078e-01
-3.43832284e-01 6.43310487e-01 2.53423423e-01 3.96927521e-02
7.62073457e-01 -1.35094869e+00 4.25629705e-01 1.07426405e+00
7.65511215e-01 5.12956142e-01 7.63585210e-01 -2.38036811e-01
-1.16281044e+00 -9.43673909e-01 5.18716753e-01 1.22037334e-02
5.31898737e-01 -2.87171006e-01 -8.12253475e-01 2.30966598e-01
5.34696400e-01 -3.49300057e-01 7.25551248e-01 -3.36867683e-02
-4.46306854e-01 -3.23870271e-01 -7.84061551e-01 5.29033482e-01
8.14428806e-01 -4.41312879e-01 -8.52558672e-01 -1.50440574e-01
5.83678842e-01 -3.89424264e-01 -5.45451522e-01 1.80069804e-01
7.20750391e-01 -1.33293986e+00 9.95007157e-01 -4.06532884e-01
6.37718737e-01 -5.57460427e-01 -2.51889914e-01 -1.82317173e+00
-5.17270267e-01 -1.08460855e+00 -2.96744764e-01 1.54205871e+00
4.10789609e-01 -4.35113043e-01 1.81749970e-01 -2.79501051e-01
-5.21082222e-01 -6.67077720e-01 -8.13358247e-01 -1.13512492e+00
1.43867120e-01 -8.26391101e-01 8.02595854e-01 7.25229204e-01
1.20089129e-01 3.92249197e-01 -8.09920073e-01 2.86015838e-01
4.75757718e-01 3.91651899e-01 5.70381403e-01 -9.58271921e-01
-7.34661877e-01 -8.47904742e-01 -2.51940101e-01 -9.81002748e-01
-2.68957168e-01 -9.53947365e-01 2.15225995e-01 -1.45188975e+00
-2.48905286e-01 -2.11298391e-01 -8.09360623e-01 1.94613412e-01
-4.02413070e-01 3.51004809e-01 6.18667722e-01 -8.08528997e-03
-1.39042698e-02 5.94466507e-01 8.95201325e-01 -4.46459576e-02
-5.68342924e-01 3.71270835e-01 -7.79142976e-01 9.53061044e-01
7.23613560e-01 -4.84478354e-01 -6.79942191e-01 -4.59801644e-01
-3.13213542e-02 3.22468460e-01 2.91319519e-01 -1.54738820e+00
2.52234221e-01 1.49906024e-01 2.44249776e-01 -5.65853894e-01
7.49358058e-01 -6.43896341e-01 3.81609678e-01 1.45013809e-01
-4.60718185e-01 -5.10419428e-01 3.13893259e-01 5.10777652e-01
-5.39213777e-01 -1.92584127e-01 8.52187991e-01 1.84809372e-01
-1.61081746e-01 5.47347330e-02 -2.94693202e-01 -7.07766414e-02
5.52027345e-01 -3.41061801e-02 -2.20304668e-01 -7.66891062e-01
-5.47746062e-01 -3.34888786e-01 -2.87812471e-01 4.59727407e-01
5.70249736e-01 -1.54315555e+00 -1.02400446e+00 1.24971420e-01
-4.03258383e-01 -4.23822224e-01 2.70554990e-01 8.31485033e-01
-1.95662901e-02 3.49121630e-01 -1.91141933e-01 -3.05997819e-01
-9.36060965e-01 5.38033485e-01 4.11606163e-01 -1.49587572e-01
-3.11103016e-01 1.01319158e+00 2.76960492e-01 -1.76932424e-01
4.86980319e-01 -5.87101758e-01 -2.93286562e-01 4.17342894e-02
7.68437505e-01 6.92837775e-01 3.02381348e-02 -5.65007567e-01
-2.37552002e-01 6.21353388e-01 5.09563684e-01 -5.46541631e-01
1.35309482e+00 2.26849332e-01 1.38298362e-01 8.27798784e-01
9.66160774e-01 6.60405517e-01 -1.02613759e+00 -2.27822259e-01
-2.20620185e-01 -3.88773084e-01 8.20014179e-02 -1.04701054e+00
-1.10386074e+00 1.15389800e+00 4.99182194e-01 6.71060622e-01
1.58851779e+00 -5.23751438e-01 1.16414738e+00 -1.89681858e-01
3.02751869e-01 -7.03289211e-01 -7.91841224e-02 3.26047719e-01
9.29135919e-01 -5.08653045e-01 -2.97232062e-01 -2.94947296e-01
-5.01146376e-01 1.21038902e+00 1.67718247e-01 -3.53267193e-01
6.70742631e-01 5.68524659e-01 2.24360347e-01 2.71359533e-01
-5.02484024e-01 -3.20751637e-01 6.63184583e-01 6.36999846e-01
4.15578097e-01 -4.78819124e-02 -1.94288060e-01 1.43917394e+00
-5.19790530e-01 6.48265183e-02 5.04503191e-01 5.08622050e-01
-4.26081210e-01 -9.34057891e-01 -8.06450427e-01 1.05515607e-01
-7.57090569e-01 -2.48766437e-01 -3.20774406e-01 3.97507370e-01
3.51138324e-01 1.41845894e+00 -2.37727597e-01 -6.92597389e-01
4.38837141e-01 3.77292454e-01 5.17601371e-01 -3.95322829e-01
-7.91941166e-01 6.66262090e-01 -2.68992707e-02 -1.38692960e-01
-4.99976188e-01 -3.60837400e-01 -1.16458392e+00 7.79327899e-02
-4.48716313e-01 2.52929449e-01 6.04507864e-01 7.05842376e-01
2.16436222e-01 1.01294959e+00 8.50739598e-01 -1.21346104e+00
-7.44132042e-01 -1.25832248e+00 -9.20524359e-01 1.05005011e-01
5.26752353e-01 -4.06014383e-01 -5.01325250e-01 3.87131214e-01]
|
[15.435449600219727, 5.60034704208374]
|
71776043-4833-47a4-a2c3-25e729ce58de
|
video-inpainting-of-complex-scenes
|
1503.05528
| null |
http://arxiv.org/abs/1503.05528v2
|
http://arxiv.org/pdf/1503.05528v2.pdf
|
Video Inpainting of Complex Scenes
|
We propose an automatic video inpainting algorithm which relies on the
optimisation of a global, patch-based functional. Our algorithm is able to deal
with a variety of challenging situations which naturally arise in video
inpainting, such as the correct reconstruction of dynamic textures, multiple
moving objects and moving background. Furthermore, we achieve this in an order
of magnitude less execution time with respect to the state-of-the-art. We are
also able to achieve good quality results on high definition videos. Finally,
we provide specific algorithmic details to make implementation of our algorithm
as easy as possible. The resulting algorithm requires no segmentation or manual
input other than the definition of the inpainting mask, and can deal with a
wider variety of situations than is handled by previous work. 1. Introduction.
Advanced image and video editing techniques are increasingly common in the
image processing and computer vision world, and are also starting to be used in
media entertainment. One common and difficult task closely linked to the world
of video editing is image and video " inpainting ". Generally speaking, this is
the task of replacing the content of an image or video with some other content
which is visually pleasing. This subject has been extensively studied in the
case of images, to such an extent that commercial image inpainting products
destined for the general public are available, such as Photoshop's " Content
Aware fill " [1]. However, while some impressive results have been obtained in
the case of videos, the subject has been studied far less extensively than
image inpainting. This relative lack of research can largely be attributed to
high time complexity due to the added temporal dimension. Indeed, it has only
very recently become possible to produce good quality inpainting results on
high definition videos, and this only in a semi-automatic manner. Nevertheless,
high-quality video inpainting has many important and useful applications such
as film restoration, professional post-production in cinema and video editing
for personal use. For this reason, we believe that an automatic, generic video
inpainting algorithm would be extremely useful for both academic and
professional communities.
|
['Patrick Pérez', 'Matthieu Fradet', 'Andrés Almansa', 'Yann Gousseau', 'Alasdair Newson']
|
2015-03-18
| null | null | null | null |
['video-inpainting']
|
['computer-vision']
|
[ 6.41637325e-01 -1.58837855e-01 1.75279066e-01 -1.37414619e-01
-4.92964923e-01 -3.94876570e-01 3.74686331e-01 7.93465450e-02
-5.80470264e-01 8.42873633e-01 -2.06292018e-01 -7.56156817e-02
2.86831427e-02 -6.62306607e-01 -7.28991985e-01 -7.22738206e-01
9.43747833e-02 2.71650851e-01 4.73436028e-01 -2.80346990e-01
2.32916445e-01 6.91867173e-01 -1.84678340e+00 1.31768391e-01
6.46948397e-01 7.92105258e-01 4.76616144e-01 7.84893632e-01
-1.51999429e-01 5.96123934e-01 -5.19378245e-01 -4.32987869e-01
3.60508472e-01 -8.10991764e-01 -6.60221159e-01 7.32080162e-01
4.20861691e-01 -1.99654132e-01 -1.13883123e-01 1.23144150e+00
2.15171814e-01 2.16950238e-01 2.00386465e-01 -8.46482635e-01
1.71075948e-02 -2.82370448e-02 -6.98318005e-01 1.82063147e-01
6.09400272e-01 -1.08215371e-02 5.01368642e-01 -4.90213662e-01
9.61799562e-01 9.19531643e-01 4.64396149e-01 3.97496849e-01
-1.44570839e+00 -2.45043278e-01 -8.27757046e-02 2.33791828e-01
-1.36187744e+00 -4.69223231e-01 9.05480266e-01 -3.95668089e-01
5.20666718e-01 6.12123311e-01 7.89486110e-01 7.25637496e-01
3.41577947e-01 5.75653315e-01 1.11330664e+00 -7.35039532e-01
1.62805036e-01 1.85707256e-01 -4.35220361e-01 5.63613772e-01
-6.41037151e-02 -7.85243064e-02 -3.11606884e-01 1.07224174e-01
1.16803527e+00 -6.60205632e-02 -6.17636323e-01 -4.11581665e-01
-1.22721064e+00 5.39482117e-01 2.76608802e-02 7.51282036e-01
-5.36106944e-01 9.44671035e-02 5.06226540e-01 6.87482595e-01
7.67734826e-01 3.20033222e-01 -6.04678020e-02 -4.09641236e-01
-1.43797791e+00 4.48347569e-01 8.02326620e-01 7.07421660e-01
7.50346422e-01 -6.19193353e-03 4.60558534e-01 9.84053195e-01
-3.13006667e-03 1.81869790e-01 3.09501886e-01 -1.21773732e+00
2.61540025e-01 1.19911104e-01 2.82930374e-01 -1.30888629e+00
6.23834133e-02 -5.30869737e-02 -9.56719339e-01 7.13239431e-01
4.54913974e-01 1.73206523e-01 -6.17838919e-01 1.33875847e+00
3.85508716e-01 8.95258263e-02 -1.70663536e-01 9.31066930e-01
2.14512423e-01 9.42538261e-01 -2.21121266e-01 -8.17469418e-01
1.17257535e+00 -7.85542250e-01 -1.05614674e+00 -7.57320672e-02
7.02434629e-02 -1.35146272e+00 7.35519290e-01 8.16909790e-01
-1.38481236e+00 -5.65212727e-01 -9.32851493e-01 -3.13958153e-02
-1.33601353e-01 -2.80750066e-01 3.65276396e-01 6.13470376e-01
-1.11481059e+00 7.28996813e-01 -7.98685730e-01 -4.57584113e-01
6.78983703e-02 3.22052926e-01 -6.34198189e-01 -2.64923304e-01
-8.71181846e-01 9.93974268e-01 3.12213033e-01 6.56917021e-02
-3.96982908e-01 -5.23350835e-01 -7.63462245e-01 -9.43721011e-02
6.48329020e-01 -5.31214595e-01 1.10731983e+00 -1.68194389e+00
-1.57529020e+00 1.15091765e+00 -1.39141306e-01 -4.17751402e-01
9.65958118e-01 -1.07309818e-01 -4.25161093e-01 3.71866405e-01
-6.42205030e-02 5.28829396e-01 1.30551708e+00 -1.22773993e+00
-5.55707932e-01 -9.27347690e-02 4.46513705e-02 2.96014249e-01
-3.03339697e-02 2.14735955e-01 -8.92220914e-01 -1.04825735e+00
2.80084480e-02 -8.06014061e-01 -3.12136024e-01 4.57789332e-01
4.29068916e-02 3.22538495e-01 1.07463419e+00 -9.21533763e-01
1.12685263e+00 -2.37426877e+00 4.66230154e-01 -1.79411191e-02
1.03849083e-01 3.73828024e-01 1.25716969e-01 4.99227524e-01
-2.01807305e-01 -2.62904078e-01 -5.88044524e-01 -3.66053522e-01
-3.93874615e-01 3.76669407e-01 -1.56290293e-01 6.10207498e-01
8.02931003e-03 3.65099192e-01 -7.52664506e-01 -6.32956564e-01
6.77183628e-01 6.42721117e-01 -4.75492448e-01 1.26948491e-01
-2.69730985e-01 4.48987544e-01 -1.11560389e-01 4.13976461e-01
6.15827978e-01 1.70787260e-01 1.72497243e-01 -1.17534690e-01
-4.36982810e-01 -2.97167689e-01 -1.50453937e+00 1.76293969e+00
-4.02980536e-01 9.64937389e-01 6.81255877e-01 -1.03508866e+00
7.12705433e-01 5.12133598e-01 7.71694839e-01 -5.84584892e-01
1.62813351e-01 3.01400989e-01 -2.10882053e-01 -5.91742814e-01
8.28955054e-01 -4.03691679e-01 4.42263454e-01 2.76796341e-01
-2.87121803e-01 -5.52021325e-01 6.05612874e-01 -3.32911802e-03
8.62871051e-01 2.96705753e-01 3.60592335e-01 -1.47988677e-01
6.74226999e-01 1.82507187e-02 2.66410530e-01 2.90209442e-01
-3.07454765e-02 9.19391990e-01 4.43291992e-01 -3.32365453e-01
-1.22515440e+00 -6.28574014e-01 -1.74687549e-01 5.51995277e-01
2.45261431e-01 -2.52848774e-01 -1.04981411e+00 -2.06034213e-01
-4.23099935e-01 2.58946568e-01 -3.75142097e-01 3.83452475e-01
-7.35367894e-01 -4.31555569e-01 -7.65024647e-02 -3.86950709e-02
4.58897591e-01 -1.29935908e+00 -7.14459658e-01 7.30528474e-01
-1.88810185e-01 -1.24952173e+00 -4.21630144e-01 -1.42337263e-01
-1.00556254e+00 -1.10444820e+00 -1.08248746e+00 -8.72445345e-01
6.18508518e-01 4.47124213e-01 9.83803451e-01 3.26814324e-01
-4.47122991e-01 5.12153268e-01 -5.30726850e-01 -4.31402549e-02
-7.92053878e-01 -2.39371389e-01 -2.18629882e-01 2.93507785e-01
-1.64328739e-01 -7.31774628e-01 -4.62215871e-01 4.08498168e-01
-1.68676102e+00 2.00275227e-01 4.53756124e-01 6.75596297e-01
6.83487892e-01 4.29650754e-01 2.27740437e-01 -9.77855325e-01
4.60286230e-01 -7.88209960e-02 -7.44679749e-01 4.76637913e-04
-3.48805159e-01 -2.80195773e-01 7.01798260e-01 -3.07221055e-01
-9.71429527e-01 1.46179259e-01 -4.96204495e-01 -5.22448063e-01
-2.76489556e-01 3.52230161e-01 -1.74669579e-01 -4.14277256e-01
4.38764364e-01 3.27346236e-01 3.86790067e-01 -4.53435272e-01
9.92015675e-02 4.61244076e-01 6.64537966e-01 -2.54400194e-01
8.49200904e-01 5.53469062e-01 7.44298846e-02 -1.33664072e+00
-2.20578268e-01 -4.97041613e-01 -4.55437541e-01 -4.62944984e-01
8.52022290e-01 -5.13257384e-01 -2.15028837e-01 5.85202098e-01
-1.11119890e+00 -3.17321390e-01 -4.26926672e-01 3.18598390e-01
-8.15999269e-01 8.19276750e-01 -5.30855596e-01 -5.06161809e-01
1.08546294e-01 -1.31933963e+00 8.12712431e-01 1.28211021e-01
-1.25619009e-01 -1.11532223e+00 2.07229462e-02 3.51950765e-01
3.63974333e-01 5.09177923e-01 6.98339045e-01 2.88189888e-01
-7.10633457e-01 -2.43699193e-01 -3.06503810e-02 6.61835134e-01
2.46851310e-01 2.00813264e-01 -6.04490280e-01 -3.25276464e-01
2.66310543e-01 6.19772077e-02 5.53821921e-01 4.44733232e-01
9.47442710e-01 3.97994258e-02 -1.51413992e-01 2.93707043e-01
1.68719506e+00 3.23633969e-01 1.08628905e+00 3.62792909e-01
4.98956829e-01 8.16485882e-01 9.03547049e-01 2.54649669e-01
-2.02700868e-01 1.16991031e+00 3.90695781e-01 -2.93090671e-01
-1.50213301e-01 1.92935258e-01 2.34585002e-01 6.58841252e-01
-3.55432570e-01 -1.77675650e-01 -4.83369678e-01 6.12063289e-01
-1.81629884e+00 -1.27218902e+00 -3.50981086e-01 2.39036942e+00
7.92722583e-01 -2.36773267e-02 6.69818148e-02 6.88673198e-01
8.29959214e-01 1.90839112e-01 -5.34908362e-02 -4.93594617e-01
-3.27077806e-02 3.11395556e-01 3.73992234e-01 6.82121933e-01
-1.02668250e+00 6.61255479e-01 5.33423185e+00 9.77019131e-01
-1.27886140e+00 4.72776368e-02 6.12316430e-01 8.72807726e-02
-1.02102950e-01 -2.35154778e-02 -2.38258541e-01 6.46263182e-01
5.40427208e-01 1.29199950e-02 5.87359667e-01 5.56031287e-01
5.59348404e-01 -6.67827189e-01 -8.30671012e-01 1.30271864e+00
2.75922507e-01 -1.27899683e+00 -2.35526979e-01 1.57913491e-01
7.69230008e-01 -5.18728614e-01 -1.04597934e-01 -2.49367148e-01
-6.00375056e-01 -8.33757997e-01 7.38892436e-01 4.58166629e-01
7.38639057e-01 -8.83900702e-01 5.56598485e-01 2.51997560e-01
-1.04756391e+00 2.58571237e-01 -2.28417218e-01 -1.50619373e-02
7.06037521e-01 8.21950853e-01 -2.30922043e-01 6.58555984e-01
6.27598286e-01 6.08905971e-01 -2.42480308e-01 1.39165497e+00
4.01171669e-02 1.85928345e-01 -3.05510968e-01 3.92121822e-01
1.16735600e-01 -5.35007298e-01 7.32959032e-01 1.10139525e+00
3.37117940e-01 1.95495576e-01 6.28317334e-03 4.52105850e-01
7.65715614e-02 4.46911871e-01 -6.15673959e-01 4.35245223e-02
-2.66768396e-01 1.20990562e+00 -1.10060108e+00 -2.67797112e-01
-4.99464899e-01 1.41986167e+00 -1.41679034e-01 1.15083002e-01
-7.77319133e-01 -3.88922453e-01 4.71238106e-01 6.27244473e-01
4.47696179e-01 -4.03289378e-01 1.03944838e-01 -1.16018081e+00
1.37812182e-01 -1.09977067e+00 2.85277013e-02 -6.98464811e-01
-8.68268371e-01 7.50436246e-01 1.34953067e-01 -1.38394380e+00
-4.49352294e-01 -4.25202280e-01 -3.41389537e-01 5.89793265e-01
-1.32650089e+00 -7.30775952e-01 -1.64437428e-01 6.39412224e-01
9.82012987e-01 1.40397027e-01 6.93096817e-01 7.38096356e-01
-2.21155763e-01 6.70167506e-02 2.78940707e-01 -3.43224585e-01
7.95079529e-01 -8.85541618e-01 -1.27609938e-01 1.03534174e+00
2.16917381e-01 1.95704505e-01 1.20420098e+00 -4.18856621e-01
-1.27904189e+00 -7.74682701e-01 9.32704449e-01 -6.53119711e-03
4.15905118e-01 -8.73321220e-02 -1.03714228e+00 4.04140711e-01
3.45215976e-01 3.06062493e-03 2.08799750e-01 -5.18272161e-01
2.88279146e-01 -2.01177970e-01 -1.14967263e+00 5.17722666e-01
6.54670119e-01 -2.55128473e-01 -4.27715957e-01 3.38009834e-01
2.06398934e-01 -5.14824569e-01 -7.50875115e-01 3.16663124e-02
3.38829160e-01 -1.36778355e+00 8.14102650e-01 1.86837837e-01
5.08448422e-01 -6.59499049e-01 7.43458122e-02 -9.98582125e-01
9.23785269e-02 -9.71760631e-01 1.90140739e-01 1.19847977e+00
-3.56667712e-02 -3.63641590e-01 8.45504403e-01 5.02121091e-01
-6.02814592e-02 -5.71669340e-01 -1.04767394e+00 -5.61533391e-01
-4.55939651e-01 -4.41402733e-01 -1.95783116e-02 9.10278201e-01
-3.61335963e-01 -3.94955426e-02 -7.53755808e-01 -1.83668733e-01
5.78615665e-01 1.35034502e-01 8.00130010e-01 -1.07645476e+00
-4.89595503e-01 -4.21814531e-01 -6.08984888e-01 -1.01775098e+00
-7.62875155e-02 -2.38659292e-01 9.18271840e-02 -1.31971788e+00
-1.50655866e-01 -3.23750883e-01 1.88806742e-01 6.10890351e-02
8.57776254e-02 7.47555435e-01 3.39251459e-01 2.53288776e-01
-2.65893757e-01 2.09603488e-01 1.36597717e+00 3.05169784e-02
-2.38333255e-01 7.92368129e-02 -2.21360967e-01 6.58900559e-01
5.51478446e-01 -3.93808067e-01 -3.30027610e-01 -1.60857767e-01
1.33348927e-01 3.98192525e-01 3.20044726e-01 -1.05587816e+00
5.52892089e-02 -5.93183003e-02 1.64623782e-01 -2.75155723e-01
5.37479877e-01 -1.22188771e+00 7.38023400e-01 3.55590969e-01
1.03069074e-01 1.41122222e-01 1.51291862e-01 5.03116667e-01
-6.84724450e-01 -5.53011119e-01 1.13246644e+00 -3.45518261e-01
-8.39423358e-01 5.04480638e-02 -6.25267506e-01 -3.58220547e-01
1.23097110e+00 -6.09172523e-01 2.72582442e-01 -6.95722222e-01
-8.01218331e-01 -3.05308014e-01 9.29986358e-01 2.34144822e-01
6.14310443e-01 -9.14238036e-01 -5.29888332e-01 2.94303149e-01
-2.70608097e-01 -9.26001370e-02 4.66646641e-01 1.02514553e+00
-1.25000918e+00 -5.65052591e-03 -4.30957198e-01 -5.49662709e-01
-1.57541513e+00 6.05374753e-01 6.10639900e-02 -1.57082170e-01
-8.48176360e-01 4.07140672e-01 2.82003284e-02 4.58723247e-01
1.65441036e-01 -1.08621381e-01 -6.77034557e-02 9.44599956e-02
6.60530508e-01 3.55395466e-01 7.96322078e-02 -8.00574958e-01
-1.22724092e-02 6.43013537e-01 -1.03090242e-01 -2.24521935e-01
1.32611668e+00 -3.13243747e-01 -3.04186791e-01 2.53868908e-01
1.01631904e+00 1.86084270e-01 -1.34097564e+00 2.79612303e-01
-2.38370344e-01 -8.08884382e-01 1.21311154e-02 -3.52969378e-01
-1.13555038e+00 7.27647543e-01 5.61792791e-01 6.69680595e-01
1.51141000e+00 -4.10474181e-01 7.85009086e-01 -1.48606956e-01
4.91731197e-01 -1.12311530e+00 -1.72509909e-01 1.53488338e-01
9.82715964e-01 -1.01513684e+00 2.78168231e-01 -6.76561177e-01
-4.06729579e-01 1.18009472e+00 1.15319863e-01 -2.35155031e-01
3.78963739e-01 3.66668850e-01 1.69600725e-01 9.46116894e-02
-3.15625340e-01 -3.86904143e-02 -2.22331844e-02 4.88896549e-01
4.66127992e-01 -2.82895535e-01 -7.95209944e-01 -4.00667131e-01
2.10227326e-01 1.94489852e-01 7.08117366e-01 1.02229691e+00
-3.78713250e-01 -1.60071087e+00 -6.97116315e-01 1.86011404e-01
-7.74381697e-01 1.35618091e-01 -8.09435621e-02 1.01950920e+00
1.62416801e-01 6.80776417e-01 -1.71551973e-01 1.42092630e-01
3.49715680e-01 -1.49501935e-01 7.38358319e-01 -4.77808535e-01
-6.86409652e-01 4.67095971e-01 3.97214592e-02 -6.39750242e-01
-8.90713811e-01 -5.99666238e-01 -8.28873336e-01 -4.55952674e-01
-1.49907246e-01 1.75193250e-01 8.42368960e-01 7.43060470e-01
-6.12496212e-02 3.84802878e-01 3.75370264e-01 -1.32427311e+00
8.90528038e-02 -4.64538932e-01 -8.17302525e-01 5.48997462e-01
1.95309520e-01 -5.14564455e-01 -2.29914457e-01 5.61591685e-01]
|
[11.345494270324707, -2.217207670211792]
|
71f60a5e-6ee3-48bf-bfad-0aea90144138
|
deepadversaries-examining-the-robustness-of
|
2112.14299
| null |
https://arxiv.org/abs/2112.14299v3
|
https://arxiv.org/pdf/2112.14299v3.pdf
|
DeepAdversaries: Examining the Robustness of Deep Learning Models for Galaxy Morphology Classification
|
With increased adoption of supervised deep learning methods for processing and analysis of cosmological survey data, the assessment of data perturbation effects (that can naturally occur in the data processing and analysis pipelines) and the development of methods that increase model robustness are increasingly important. In the context of morphological classification of galaxies, we study the effects of perturbations in imaging data. In particular, we examine the consequences of using neural networks when training on baseline data and testing on perturbed data. We consider perturbations associated with two primary sources: 1) increased observational noise as represented by higher levels of Poisson noise and 2) data processing noise incurred by steps such as image compression or telescope errors as represented by one-pixel adversarial attacks. We also test the efficacy of domain adaptation techniques in mitigating the perturbation-driven errors. We use classification accuracy, latent space visualizations, and latent space distance to assess model robustness. Without domain adaptation, we find that processing pixel-level errors easily flip the classification into an incorrect class and that higher observational noise makes the model trained on low-noise data unable to classify galaxy morphologies. On the other hand, we show that training with domain adaptation improves model robustness and mitigates the effects of these perturbations, improving the classification accuracy by 23% on data with higher observational noise. Domain adaptation also increases by a factor of ~2.3 the latent space distance between the baseline and the incorrectly classified one-pixel perturbed image, making the model more robust to inadvertent perturbations.
|
['Stefan M. Wild', 'Sandeep Madireddy', 'Brian Nord', 'Kevin Pedro', 'Gabriel Nathan Perdue', 'F. Javier Sánchez', 'Gregory Snyder', 'Diana Kafkes', 'Aleksandra Ćiprijanović']
|
2021-12-28
| null | null | null | null |
['morphology-classification']
|
['computer-vision']
|
[ 6.32171929e-01 -7.23709911e-02 7.22103715e-01 -1.82740673e-01
-6.69513226e-01 -9.03421640e-01 7.78468549e-01 2.80881464e-01
-6.47418261e-01 2.73957372e-01 4.16838348e-01 -4.89572674e-01
-4.51030917e-02 -9.07437205e-01 -9.26015675e-01 -9.27097499e-01
-5.69421798e-02 2.96363860e-01 3.70998263e-01 9.98991281e-02
4.57841933e-01 6.84859693e-01 -1.34597039e+00 2.48978689e-01
7.27468550e-01 6.85297608e-01 -1.41069710e-01 8.43187749e-01
1.98772207e-01 3.76799583e-01 -1.07075334e+00 -7.02261627e-02
7.41825044e-01 -1.92686811e-01 -5.24569333e-01 1.14237323e-01
7.67741978e-01 -5.53017616e-01 -4.30094033e-01 1.22402132e+00
6.47719145e-01 2.75623888e-01 5.77529073e-01 -9.21922624e-01
-5.98419070e-01 2.63859481e-02 -4.48806703e-01 5.15614569e-01
-1.56228259e-01 1.01110137e+00 3.51266950e-01 -4.04329151e-01
6.84233904e-01 1.22495294e+00 8.10012579e-01 6.20344505e-02
-1.97552347e+00 -5.76305807e-01 -3.15096468e-01 -1.04758561e-01
-1.05165243e+00 -4.95179504e-01 5.03391147e-01 -1.05763769e+00
9.36799407e-01 1.81628451e-01 1.58287331e-01 1.04742444e+00
1.44893378e-01 -5.26296020e-01 1.13755119e+00 -3.44123960e-01
3.74333471e-01 8.80615786e-03 -2.01485291e-01 1.34784102e-01
3.07976395e-01 5.74483633e-01 -2.55143970e-01 -3.60339612e-01
7.84836411e-01 -5.24581313e-01 -7.83687159e-02 1.41352434e-02
-1.00474322e+00 6.66447222e-01 4.72397506e-01 -5.78662120e-02
-1.48299277e-01 -8.03277304e-04 4.10655439e-01 6.16719015e-02
6.00515485e-01 1.13718677e+00 -4.63289738e-01 -4.62683141e-02
-8.67057681e-01 3.77449363e-01 3.15604031e-01 3.65912616e-01
5.66306233e-01 2.75923580e-01 -7.29399128e-03 7.53135502e-01
-2.31226891e-01 4.96883392e-01 5.35467923e-01 -1.29592597e+00
2.60652781e-01 4.75370049e-01 1.74927011e-01 -1.17666304e+00
-4.29374486e-01 -5.76811910e-01 -6.59753740e-01 8.68288279e-01
8.43532860e-01 -2.86686718e-01 -1.04801905e+00 1.64545631e+00
1.84864134e-01 3.84535976e-02 -4.94332202e-02 7.71415710e-01
4.15814042e-01 5.81915557e-01 8.89548287e-02 -6.75561652e-02
9.89095867e-01 -3.37698847e-01 -2.50241250e-01 -3.67242754e-01
5.87423086e-01 -9.12492156e-01 1.27606654e+00 1.89493209e-01
-9.83837664e-01 -5.62941670e-01 -1.15455055e+00 -4.35671844e-02
-4.23803717e-01 -4.83677953e-01 -1.24540687e-01 4.77044761e-01
-8.60920668e-01 8.72168481e-01 -8.38818371e-01 -2.97311157e-01
3.47991467e-01 2.69100666e-01 -3.87073934e-01 3.06339085e-01
-7.93316305e-01 9.46110487e-01 1.02824502e-01 -3.97241652e-01
-7.14016080e-01 -1.21788323e+00 -5.54822326e-01 2.04809323e-01
7.21179247e-02 -3.07570130e-01 1.11191642e+00 -8.64547849e-01
-8.77718329e-01 7.73777425e-01 6.98297992e-02 -6.71749592e-01
6.20354533e-01 3.07212006e-02 -3.52751434e-01 -1.27407750e-02
-1.60259515e-01 4.29530263e-01 8.66957486e-01 -1.18637538e+00
-1.62635565e-01 -3.93238068e-01 -2.77167201e-01 -5.56598864e-02
-4.47515190e-01 2.16442317e-01 1.09522253e-01 -7.48084009e-01
9.40174535e-02 -1.02343869e+00 -1.93107560e-01 1.54293448e-01
-9.41282064e-02 6.46500647e-01 8.36526275e-01 -1.10928535e+00
8.01055789e-01 -2.34402394e+00 -1.68766111e-01 2.51240820e-01
2.99067050e-01 4.79927570e-01 -1.66318506e-01 1.03533290e-01
-3.34152222e-01 5.63832760e-01 -4.45343941e-01 -1.68482974e-01
-2.52222657e-01 -5.66668063e-02 -3.72136593e-01 4.93237466e-01
4.44105625e-01 4.59006280e-01 -5.34110487e-01 1.04925528e-01
1.32091045e-01 2.01784283e-01 -7.97795773e-01 2.09168956e-01
-1.96791381e-01 8.26764226e-01 3.84839356e-01 1.37271926e-01
8.16052914e-01 7.07393214e-02 -1.11679815e-01 -2.11569682e-01
-3.19041461e-01 4.26570535e-01 -1.03763020e+00 9.90503967e-01
-2.35771611e-01 1.00376439e+00 1.67892620e-01 -6.18406892e-01
9.29674089e-01 2.55690422e-03 7.34853223e-02 -7.43802845e-01
-6.65662140e-02 1.46581322e-01 6.02482259e-01 -4.70465899e-01
4.57786560e-01 -1.95668653e-01 3.42818737e-01 4.19287592e-01
-1.38385251e-01 -1.88036561e-01 -1.29094169e-01 3.44201736e-02
1.47426867e+00 -2.19736412e-01 -2.01499134e-01 -2.19085127e-01
-8.67695510e-02 2.10386127e-01 6.61213100e-01 9.31314230e-01
-3.58569950e-01 8.70159686e-01 7.02734292e-01 -2.75877655e-01
-1.82951272e+00 -1.03949201e+00 -2.52875000e-01 8.40047598e-01
-4.21758145e-01 -1.56462982e-01 -7.07515717e-01 -3.44681084e-01
2.74228483e-01 9.88165796e-01 -5.30933082e-01 -6.59358025e-01
-4.62264806e-01 -1.08444333e+00 8.24636638e-01 5.08262396e-01
5.22411346e-01 -7.99218774e-01 -5.79634726e-01 -7.30587244e-02
-5.33380508e-02 -1.06843066e+00 -2.43480265e-01 8.62389430e-02
-8.66728425e-01 -8.79220486e-01 -1.71418890e-01 -2.39306092e-01
7.19516933e-01 -6.84603257e-03 9.26648915e-01 1.69530481e-01
-4.64667439e-01 1.61059663e-01 -2.28480101e-01 -3.77939731e-01
-9.26872432e-01 -3.75853002e-01 2.40211174e-01 -2.77615786e-01
1.38910323e-01 -7.23369718e-01 -6.14616573e-01 2.59538144e-01
-1.15900004e+00 -2.68890798e-01 1.17449194e-01 8.56277406e-01
2.43527517e-01 5.02799451e-01 1.38946667e-01 -6.50486767e-01
4.45062906e-01 -4.82276291e-01 -8.66911411e-01 -2.35431388e-01
-5.10743618e-01 1.91113710e-01 5.61226070e-01 -7.01767147e-01
-1.14953351e+00 -2.44593695e-01 6.63531050e-02 -3.83412480e-01
-4.12943095e-01 3.28378826e-01 1.12548880e-01 -2.80007243e-01
1.50849366e+00 -1.17164157e-01 2.39563346e-01 -6.77255690e-01
4.24154587e-02 5.31984925e-01 8.77461195e-01 -4.11140800e-01
1.22271788e+00 5.71177602e-01 2.04566151e-01 -1.03803730e+00
-3.51137459e-01 -1.16291642e-01 -5.87302804e-01 -3.17082070e-02
8.60030234e-01 -6.85909569e-01 -1.68421328e-01 6.66123271e-01
-9.34922278e-01 -6.47949815e-01 -2.40677133e-01 3.19430053e-01
-1.89206794e-01 4.17692572e-01 -3.00439298e-01 -5.52575707e-01
2.46596523e-02 -1.36385572e+00 6.28773689e-01 4.85903099e-02
-2.97780633e-01 -7.73880601e-01 -4.86679263e-02 2.52059102e-01
5.82882643e-01 4.80789393e-01 1.38552856e+00 -7.67756462e-01
-3.60544115e-01 -1.26673952e-01 -1.57548383e-01 5.88886201e-01
4.18596230e-02 4.41154480e-01 -1.26298106e+00 -3.67455155e-01
2.05929905e-01 -9.89109501e-02 6.60503149e-01 3.35293472e-01
1.20952725e+00 -2.79536039e-01 1.15671918e-01 8.08050156e-01
1.22179806e+00 1.27103254e-01 7.69111276e-01 6.56779706e-01
3.94389957e-01 8.85219872e-01 1.78813040e-01 1.55066535e-01
-4.00776297e-01 6.77652955e-01 4.02899891e-01 -2.98096985e-01
-2.71136880e-01 -1.21691162e-02 6.47297949e-02 1.34028554e-01
2.98310399e-01 6.96749762e-02 -1.29112446e+00 4.21613693e-01
-1.32034695e+00 -9.65440094e-01 -2.45692387e-01 2.66295123e+00
8.07677388e-01 4.82257217e-01 1.84172750e-01 4.02297359e-03
6.58136845e-01 -9.43139866e-02 -7.77090251e-01 -4.38190967e-01
-3.15782994e-01 1.10403858e-01 8.43697906e-01 5.64177036e-01
-9.54166114e-01 5.43865025e-01 6.57972002e+00 2.08114758e-01
-1.24449193e+00 -9.10147727e-02 8.92174780e-01 -5.34437954e-01
-2.35585254e-02 -7.23838806e-02 -2.85965264e-01 6.58190548e-01
1.15117681e+00 -5.59238568e-02 6.67365849e-01 6.18136644e-01
5.66890955e-01 -2.72919059e-01 -8.81339073e-01 5.46180487e-01
-2.54220456e-01 -1.37385988e+00 -1.63652658e-01 3.84009093e-01
6.86694980e-01 2.23458111e-01 3.16537052e-01 3.76722887e-02
4.23149943e-01 -1.00429869e+00 6.41256094e-01 3.87261450e-01
8.58257174e-01 -6.17077172e-01 6.62154198e-01 3.56559604e-01
-3.39975297e-01 -1.65237933e-01 -5.85394025e-01 -3.88196588e-01
-2.35842377e-01 6.25362813e-01 -1.15424764e+00 -8.87626410e-02
8.57240796e-01 4.37937938e-02 -1.10054386e+00 9.94555116e-01
1.25618726e-01 1.04649258e+00 -4.45849955e-01 7.09125340e-01
-1.63678020e-01 -1.98480442e-01 7.98353434e-01 1.09849048e+00
2.39035070e-01 -6.13007434e-02 -2.91191459e-01 1.03143632e+00
1.84262209e-02 -4.51790005e-01 -7.70543098e-01 -9.68280584e-02
7.17925191e-01 9.48625326e-01 -6.24648690e-01 -1.47879064e-01
-2.83607841e-01 7.07995176e-01 2.91545898e-01 4.66689557e-01
-5.31975150e-01 -2.08427116e-01 1.06143272e+00 4.66873884e-01
1.34328216e-01 -4.73949432e-01 -9.36571181e-01 -7.03059733e-01
-1.70393083e-02 -9.98636067e-01 7.10374489e-02 -9.81312394e-01
-1.20780182e+00 1.66475073e-01 4.25413251e-02 -8.87230217e-01
-9.51638445e-02 -6.31133735e-01 -8.97088945e-01 1.19433665e+00
-8.26214671e-01 -5.03363729e-01 -3.36889893e-01 -7.21656159e-02
2.11009562e-01 -3.20506215e-01 6.54464006e-01 9.29898843e-02
-5.58262110e-01 4.73123461e-01 5.91136336e-01 1.81582019e-01
1.05428827e+00 -1.29046297e+00 8.97092640e-01 1.26446176e+00
-1.50196269e-01 7.72239149e-01 1.12747705e+00 -9.47162926e-01
-7.84326315e-01 -1.18729389e+00 3.78494412e-01 -6.19102299e-01
8.34501266e-01 -4.09397036e-01 -1.40569782e+00 5.20362198e-01
-1.33396119e-01 -9.79927257e-02 6.03938103e-01 -5.85189201e-02
-6.03031516e-01 3.79521027e-02 -1.51929319e+00 7.97408462e-01
9.34482574e-01 -6.20484114e-01 -5.01245975e-01 2.82970428e-01
8.37882698e-01 -2.71545380e-01 -8.66768837e-01 4.47283298e-01
2.70344555e-01 -9.92948174e-01 1.04991496e+00 -8.16149056e-01
7.16081381e-01 -3.81865919e-01 -1.34331420e-01 -1.46124792e+00
-6.72401845e-01 -3.28519285e-01 3.70026410e-01 1.31283236e+00
2.63808995e-01 -5.59616625e-01 4.98585224e-01 9.13406193e-01
1.87507924e-02 -8.84234235e-02 -9.26920712e-01 -7.80508399e-01
6.01371467e-01 -2.88031787e-01 4.93448317e-01 9.28498328e-01
-4.25678641e-01 -1.16624318e-01 1.16372392e-01 7.11895406e-01
5.19194961e-01 -5.28781354e-01 8.17066669e-01 -9.54483628e-01
-5.02767622e-01 -5.11261106e-01 -5.15535951e-01 -2.72271067e-01
-3.39388460e-01 -5.86250842e-01 1.70304924e-01 -9.53568041e-01
-3.50421928e-02 -3.08105677e-01 8.97567496e-02 1.48185253e-01
-3.54378760e-01 3.21083575e-01 3.53124201e-01 4.02201056e-01
2.32301608e-01 8.85792971e-02 7.14707553e-01 1.12671904e-01
1.40996312e-03 -3.34765136e-01 -4.81730163e-01 7.59680450e-01
8.70524943e-01 -5.19250453e-01 -3.72367539e-02 -6.44572616e-01
1.33354321e-01 -5.44638395e-01 8.35231304e-01 -1.20959210e+00
-1.05682686e-01 -1.08099498e-01 6.69681132e-01 -1.46929592e-01
1.74452394e-01 -6.05773270e-01 1.79823801e-01 5.08211732e-01
-3.96095693e-01 1.18701108e-01 7.01251566e-01 3.67087811e-01
3.90966862e-01 -3.44275445e-01 1.21555364e+00 -1.08628504e-01
-2.94193208e-01 -3.70967299e-01 -7.75558770e-01 7.50416741e-02
5.80016255e-01 1.00776292e-02 -6.69373453e-01 -3.05735201e-01
-6.86658204e-01 -2.09332973e-01 1.01284051e+00 1.16208486e-01
-3.87831740e-02 -9.19568539e-01 -6.93873584e-01 5.17059684e-01
-1.62929013e-01 -4.08844985e-02 1.92146361e-01 5.39981961e-01
-7.44057298e-01 -3.92230004e-02 -1.74062625e-01 -5.49248159e-01
-1.44008589e+00 4.33885217e-01 6.81650817e-01 -3.30832377e-02
-2.67338365e-01 1.01438606e+00 1.68420538e-01 -5.12949884e-01
1.62396356e-01 -2.36835942e-01 3.57166022e-01 -2.43325830e-01
4.23933357e-01 7.31228530e-01 3.41267407e-01 -4.69235599e-01
-3.25909778e-02 5.79303876e-02 -1.45875327e-02 -2.88452476e-01
1.16145468e+00 1.49907529e-01 2.08246503e-02 2.42986783e-01
9.80412662e-01 7.85723925e-02 -1.73589182e+00 -3.08137417e-01
-3.72731015e-02 -6.70250535e-01 2.21204087e-01 -1.17040038e+00
-5.61419129e-01 9.77256358e-01 1.00110245e+00 4.04232979e-01
1.10902250e+00 -3.54998410e-01 3.24218839e-01 1.39010787e-01
-2.38439307e-01 -1.00736153e+00 2.37331599e-01 4.25546557e-01
1.01149142e+00 -1.15548992e+00 1.53031096e-01 -1.71440672e-02
-4.13582504e-01 8.80539775e-01 8.40050578e-01 5.43816527e-03
2.97950387e-01 4.34125334e-01 2.48306319e-01 3.94203380e-04
-5.12080193e-01 1.83324859e-01 6.37599528e-02 9.11478400e-01
9.16023031e-02 -1.48855537e-01 5.97994924e-02 2.59712785e-01
-4.61397439e-01 -4.21475321e-01 7.87749827e-01 7.88095415e-01
-5.07710457e-01 -7.75685489e-01 -1.04374230e+00 7.16657579e-01
-2.43876845e-01 -1.89011157e-01 -5.66228211e-01 5.39617300e-01
2.99556345e-01 9.61169779e-01 6.13444507e-01 -4.07328159e-01
3.49908412e-01 3.33145350e-01 1.64408714e-01 -6.95960820e-01
-6.24926090e-01 1.01214675e-02 7.10239410e-02 -2.79562324e-01
2.12277174e-01 -1.04799080e+00 -1.04329073e+00 -5.73716044e-01
-1.28580973e-01 -4.36373949e-01 8.13904524e-01 8.98464501e-01
6.01637602e-01 7.98383474e-01 4.29627329e-01 -9.73685205e-01
-1.00775409e+00 -1.17430902e+00 -2.25812674e-01 7.30110049e-01
4.69176412e-01 -5.49316227e-01 -8.08329463e-01 3.18204314e-01]
|
[5.584811210632324, 7.879670143127441]
|
a3fa97ba-0719-4332-aa18-3c468bc104b8
|
software-vulnerability-prediction-knowledge
|
2303.06177
| null |
https://arxiv.org/abs/2303.06177v1
|
https://arxiv.org/pdf/2303.06177v1.pdf
|
Software Vulnerability Prediction Knowledge Transferring Between Programming Languages
|
Developing automated and smart software vulnerability detection models has been receiving great attention from both research and development communities. One of the biggest challenges in this area is the lack of code samples for all different programming languages. In this study, we address this issue by proposing a transfer learning technique to leverage available datasets and generate a model to detect common vulnerabilities in different programming languages. We use C source code samples to train a Convolutional Neural Network (CNN) model, then, we use Java source code samples to adopt and evaluate the learned model. We use code samples from two benchmark datasets: NIST Software Assurance Reference Dataset (SARD) and Draper VDISC dataset. The results show that proposed model detects vulnerabilities in both C and Java codes with average recall of 72\%. Additionally, we employ explainable AI to investigate how much each feature contributes to the knowledge transfer mechanisms between C and Java in the proposed model.
|
['Goksu Karadag', 'Basak Gencer Unsalver', 'Ramin F Fouladi', 'Khadija Hanifi']
|
2023-03-10
| null | null | null | null |
['vulnerability-detection']
|
['miscellaneous']
|
[-8.16964284e-02 -1.66061278e-02 -3.69738415e-02 -3.92094761e-01
-6.06866419e-01 -8.40759754e-01 3.79117936e-01 2.95742661e-01
1.25065774e-01 2.18385622e-01 1.19735315e-01 -6.91911697e-01
2.34656092e-02 -9.75006342e-01 -6.66785657e-01 8.34795609e-02
5.98820113e-02 -3.65310460e-01 3.43686402e-01 -3.01975995e-01
8.10892165e-01 7.98242390e-02 -1.20624804e+00 6.12085819e-01
1.09987259e+00 5.44793963e-01 -1.03586242e-01 5.30951142e-01
-3.73013318e-01 9.55832183e-01 -6.77493036e-01 -6.15709484e-01
4.02877063e-01 -3.47472169e-03 -1.08829212e+00 -6.32581592e-01
3.53764653e-01 -1.70791805e-01 -8.49960819e-02 1.38543952e+00
2.54332006e-01 -3.66940081e-01 6.02390885e-01 -1.44303393e+00
-1.05235875e+00 1.00616622e+00 -9.52366889e-01 3.12838912e-01
3.56308073e-01 7.77182430e-02 7.17425942e-01 -5.53225815e-01
3.20630401e-01 1.00921643e+00 8.47078264e-01 5.45583010e-01
-7.68528461e-01 -9.12743926e-01 -2.44200796e-01 2.34603882e-01
-1.10367954e+00 9.15022567e-02 1.03283501e+00 -8.68160188e-01
1.30024767e+00 -1.05415471e-01 -4.55683731e-02 9.32646871e-01
3.75532359e-01 4.44231480e-01 1.12694550e+00 -5.75407386e-01
9.97512117e-02 6.08867168e-01 7.31620014e-01 6.74892128e-01
4.22155380e-01 -1.61864348e-02 6.91924542e-02 -3.57584059e-01
7.23174214e-02 6.63781017e-02 -3.33706513e-02 1.37707954e-02
-5.30548871e-01 8.81179154e-01 3.66022110e-01 5.44753075e-01
-1.00045085e-01 -2.30538752e-02 7.54766643e-01 7.31135726e-01
-3.46529484e-02 3.23728263e-01 -7.67806768e-01 -3.04892153e-01
-5.08874834e-01 4.18675169e-02 8.11217606e-01 8.21666420e-01
9.72930908e-01 1.33171022e-01 4.15321022e-01 5.31173885e-01
6.03494287e-01 2.18125671e-01 4.65623170e-01 -3.53273928e-01
8.33363593e-01 1.36613834e+00 -2.96942621e-01 -1.38279819e+00
-8.56077969e-02 -2.19072416e-01 -2.60704666e-01 6.52644694e-01
2.52284259e-02 -1.51721060e-01 -6.04403436e-01 1.37881315e+00
7.52279386e-02 9.89234671e-02 3.64155054e-01 3.52924973e-01
8.58634174e-01 4.11845297e-01 1.56638011e-01 4.08526421e-01
1.18966687e+00 -8.53976369e-01 -9.73735563e-03 -1.14216335e-01
9.02914166e-01 -8.72892797e-01 9.61705148e-01 3.20681274e-01
-5.14153898e-01 -6.01701081e-01 -1.26892126e+00 6.38275594e-02
-7.64047027e-01 7.18166307e-02 5.09538531e-01 1.19025540e+00
-7.43586063e-01 3.33493501e-01 -5.97211480e-01 -4.43604529e-01
4.12678957e-01 2.31124282e-01 -4.22521859e-01 8.19941685e-02
-9.72846866e-01 6.97797239e-01 6.11049771e-01 -3.13494086e-01
-1.11410725e+00 -7.57551312e-01 -6.18887544e-01 3.47199887e-01
1.00018643e-01 1.06486544e-01 9.46831465e-01 -1.22289109e+00
-9.23077583e-01 6.39359474e-01 5.47128022e-01 -3.48289192e-01
4.76637147e-02 -2.05129609e-01 -7.59692788e-01 -3.31139416e-01
-4.64717895e-02 7.93377403e-03 4.74459440e-01 -1.22711301e+00
-5.18876076e-01 -2.23198965e-01 4.30925995e-01 -5.42205870e-01
-6.35040581e-01 4.69609290e-01 -8.84591416e-03 -4.43011791e-01
-5.50096512e-01 -8.23339820e-01 3.27680588e-01 -5.10619879e-01
-5.00127971e-01 -1.57507628e-01 9.49309826e-01 -1.00926709e+00
1.37325394e+00 -2.04148912e+00 -1.08925790e-01 4.23327118e-01
3.15822154e-01 6.21419847e-01 -3.27260464e-01 5.27587175e-01
-4.69892502e-01 4.50108647e-01 -4.98076051e-01 2.84747273e-01
-9.69847441e-02 -2.64192432e-01 -4.65880811e-01 6.27231738e-03
3.03325236e-01 5.99952102e-01 -4.69207674e-01 -2.39293307e-01
-1.16777368e-01 2.91946441e-01 -6.81663454e-01 2.35600263e-01
-1.23329058e-01 1.33225128e-01 -6.65842712e-01 8.66716266e-01
9.36727464e-01 1.42413467e-01 -5.30209728e-02 -1.01186022e-01
-1.94230348e-01 -1.30685838e-02 -1.07443082e+00 1.36229193e+00
-7.26859748e-01 5.81040025e-01 -3.75019044e-01 -9.25981462e-01
1.29222357e+00 3.20791453e-01 -1.18403807e-01 -6.26471281e-01
2.24486634e-01 3.11287493e-01 3.16629440e-01 -7.66428411e-01
9.89051759e-02 4.69979525e-01 -1.38987735e-01 8.35627079e-01
2.97763888e-02 3.72810185e-01 -2.82630503e-01 1.51999906e-01
1.47236037e+00 -9.55797674e-04 1.09902732e-01 -2.54683226e-01
1.04208040e+00 5.22211976e-02 4.88037407e-01 5.70109606e-01
-2.79247612e-01 1.15822680e-01 7.81338155e-01 -7.54368961e-01
-1.03602886e+00 -7.13467777e-01 1.82323456e-01 7.68600404e-01
-1.61687866e-01 -3.58160287e-01 -1.00000012e+00 -1.33917356e+00
-1.32640928e-01 7.00841010e-01 -8.67703617e-01 -4.97328281e-01
-5.56042433e-01 -4.32319432e-01 9.25034821e-01 7.12539613e-01
6.37231052e-01 -1.37215579e+00 -6.59369051e-01 -6.69453368e-02
2.57731110e-01 -6.48229718e-01 -2.36668587e-01 -8.31398219e-02
-6.10812187e-01 -1.55416214e+00 5.54785244e-02 -8.19637477e-01
5.41470289e-01 3.16896029e-02 1.09682572e+00 4.93237615e-01
-6.55140936e-01 1.89590544e-01 -6.00022316e-01 -4.66565788e-01
-9.11710203e-01 3.22876841e-01 -4.36008096e-01 -1.69033960e-01
8.94476175e-01 -4.84772652e-01 -3.18858981e-01 2.02189177e-01
-9.87426281e-01 -4.42092985e-01 5.82174420e-01 4.36773181e-01
-1.95488691e-01 7.19494298e-02 5.59933603e-01 -1.05646348e+00
7.79516041e-01 -1.11060214e+00 -8.59670877e-01 5.58131516e-01
-6.94711089e-01 1.62100196e-01 6.41475677e-01 -1.82401121e-01
-1.16930139e+00 -1.55010790e-01 -7.97246248e-02 -3.98127846e-02
-2.95602202e-01 8.27295601e-01 -2.61721879e-01 -3.59016389e-01
8.54415298e-01 1.25248194e-01 -2.33194083e-01 -5.70151746e-01
1.54725224e-01 9.78787243e-01 3.47456157e-01 -8.42591286e-01
1.09276187e+00 -6.47117794e-02 -4.70704556e-01 -3.77578825e-01
-1.88943595e-01 -6.59371018e-02 -3.24115276e-01 1.62924435e-02
9.24972475e-01 -5.25319576e-01 -5.23487270e-01 6.26724303e-01
-1.24757540e+00 -3.95535715e-02 4.67513442e-01 1.28857434e-01
-3.29149142e-02 4.44311738e-01 -4.06320393e-01 -6.43857658e-01
-7.16015577e-01 -1.55355644e+00 5.61518013e-01 2.90791601e-01
1.37235327e-02 -7.79515564e-01 6.10890329e-01 2.58954108e-01
8.29414308e-01 4.46980298e-01 1.46744335e+00 -8.79382551e-01
-5.76846778e-01 -3.62155467e-01 -4.64004010e-01 4.83493626e-01
3.15595180e-01 4.38728869e-01 -9.25568104e-01 -1.20734356e-01
-6.67627230e-02 -3.97983074e-01 5.11745274e-01 -2.35201001e-01
1.06155288e+00 -2.69147694e-01 -2.48577923e-01 4.46859539e-01
1.86659169e+00 5.94526231e-01 7.59223878e-01 6.02987528e-01
7.84947157e-01 5.98390639e-01 1.48424566e-01 1.42167538e-01
4.79509860e-01 3.75334710e-01 5.36013842e-01 3.67782772e-01
1.80752892e-02 -9.92621332e-02 3.51556182e-01 9.39411640e-01
1.10754952e-01 1.73614010e-01 -1.70081735e+00 6.27589464e-01
-1.58253908e+00 -6.90544784e-01 -2.28599310e-01 1.97585130e+00
7.40596354e-01 6.52500018e-02 -1.32862866e-01 1.78843692e-01
5.81931591e-01 -3.96952897e-01 -3.08523089e-01 -7.86627829e-01
3.14051896e-01 3.87338340e-01 1.88315421e-01 3.79065067e-01
-1.07400978e+00 7.94605911e-01 5.38921261e+00 3.97620112e-01
-1.22897828e+00 1.59835055e-01 4.40072834e-01 5.51063418e-01
-4.45519984e-01 3.49429429e-01 -5.65765381e-01 4.96825933e-01
1.23969805e+00 -3.19760025e-01 2.87826061e-01 1.25739670e+00
-4.36720878e-01 2.66472399e-01 -9.80553210e-01 4.39363629e-01
5.45465499e-02 -1.07748735e+00 -1.07079066e-01 -1.89057529e-01
8.08322906e-01 6.43639937e-02 6.05926216e-02 6.29397750e-01
4.84652847e-01 -1.11863637e+00 2.72003174e-01 5.18565476e-01
3.15264255e-01 -1.00047612e+00 1.02802622e+00 9.63139832e-02
-1.13767779e+00 -4.12018925e-01 -5.48513412e-01 7.52229542e-02
-6.35025501e-01 4.69336808e-02 -8.15118968e-01 4.90127981e-01
1.09068120e+00 3.98052335e-01 -1.15771580e+00 1.01064646e+00
2.45557167e-02 8.01209092e-01 2.04998031e-01 4.91483808e-02
9.04134437e-02 2.15434074e-01 2.03199182e-02 1.22897530e+00
3.71398687e-01 -3.95876169e-01 -7.12176831e-03 1.24689698e+00
-2.47006025e-02 1.72722399e-01 -7.39206553e-01 -2.20728651e-01
6.13455653e-01 1.24996781e+00 -5.21143973e-01 2.61060223e-02
-9.70636010e-01 4.41844106e-01 3.68054926e-01 1.02223769e-01
-8.22150946e-01 -9.85225022e-01 4.74490136e-01 -2.33244881e-01
2.29379132e-01 -4.78082988e-03 -3.08961898e-01 -1.00604689e+00
4.08646539e-02 -1.22717047e+00 3.51455927e-01 -4.71234322e-01
-1.20201373e+00 9.79128361e-01 4.20658961e-02 -1.16784275e+00
1.88011110e-01 -5.68479836e-01 -1.20580554e+00 1.09325767e+00
-1.56904590e+00 -1.42107797e+00 -7.09572136e-01 4.94838893e-01
3.02003801e-01 -9.10514951e-01 7.28770077e-01 6.25230789e-01
-6.11180246e-01 1.01307487e+00 -1.51774764e-01 6.23976469e-01
6.02905869e-01 -1.19813049e+00 8.63133073e-01 1.00926828e+00
-1.29799351e-01 1.14226043e+00 3.69446486e-01 -8.11353385e-01
-1.40483069e+00 -1.22431934e+00 3.73016119e-01 -7.14493036e-01
7.76960194e-01 -6.78043962e-02 -1.34772587e+00 7.70933211e-01
4.38967466e-01 5.54953469e-03 8.86979520e-01 -1.11199699e-01
-1.16002631e+00 -4.23299633e-02 -1.41951942e+00 1.23729475e-01
4.03613925e-01 -7.51339793e-01 -6.48629487e-01 -1.65911853e-01
6.90385163e-01 6.66624904e-02 -8.82195234e-01 1.14647061e-01
4.91624445e-01 -1.14757943e+00 6.34595573e-01 -6.90650284e-01
9.18734908e-01 -4.43902165e-01 -3.26219320e-01 -1.00101674e+00
6.45971745e-02 5.36651770e-03 -3.52445580e-02 1.68587470e+00
5.37803769e-01 -6.07286870e-01 7.16800809e-01 7.14203894e-01
-1.05555290e-02 -3.34108472e-01 -4.59795356e-01 -4.41728324e-01
6.28801823e-01 -3.44980031e-01 1.00009978e+00 1.46501267e+00
-1.75093375e-02 -1.15170918e-01 -1.27852159e-02 3.79279584e-01
6.00249290e-01 4.08397848e-03 7.79469252e-01 -1.19815469e+00
-5.65298617e-01 -3.28718901e-01 -6.55232072e-01 7.46502504e-02
2.32400060e-01 -9.41906095e-01 -3.04844052e-01 -9.86186564e-01
5.21298587e-01 -4.17189360e-01 -4.77286309e-01 7.80443132e-01
-1.43541992e-01 -2.25154385e-02 -1.84397176e-02 -8.87028407e-04
-1.05822653e-01 6.34224191e-02 1.98050678e-01 -4.13596570e-01
-1.68222319e-02 -1.75815463e-01 -8.44986558e-01 6.70382261e-01
9.97123957e-01 -6.54534161e-01 -5.35684109e-01 -7.47107685e-01
3.61203611e-01 -2.61974484e-02 2.80900687e-01 -1.20313954e+00
6.47026375e-02 -6.95147961e-02 -4.26732339e-02 -2.42797256e-01
-6.48240268e-01 -9.07689035e-01 5.11195660e-02 5.62711000e-01
-3.47460687e-01 3.19768876e-01 5.22014499e-01 2.69789100e-01
-1.18196316e-01 -6.97753131e-01 7.43589818e-01 -1.60081699e-01
-8.68346810e-01 1.14628725e-01 -6.54944852e-02 2.60460209e-02
1.29450715e+00 -1.35288745e-01 -7.78859794e-01 2.76941627e-01
-6.09789155e-02 6.33611381e-02 5.14115036e-01 1.10168660e+00
5.63249171e-01 -1.20892203e+00 -6.73500299e-01 2.81897902e-01
5.13273180e-01 -6.51830912e-01 2.52964079e-01 2.98196435e-01
-8.14394116e-01 2.77248949e-01 -7.42704630e-01 -1.47451460e-01
-1.36990106e+00 7.83869624e-01 3.49122882e-01 6.54696375e-02
-2.87014395e-01 5.91689229e-01 -1.50979623e-01 -9.77182567e-01
9.18243602e-02 -1.53678849e-01 -6.26043558e-01 -5.27398229e-01
6.86971366e-01 4.11041647e-01 1.59274228e-02 -4.70357746e-01
-6.09527588e-01 6.83267772e-01 -4.57665622e-01 4.48856920e-01
1.40772104e+00 4.92378771e-01 -4.12326574e-01 -6.51420932e-03
1.53127933e+00 1.11783490e-01 -7.26475894e-01 -2.34465469e-02
5.63660860e-01 -4.65730458e-01 -1.68618843e-01 -1.16061127e+00
-1.31220984e+00 1.14105821e+00 1.15729737e+00 1.98678955e-01
9.20864224e-01 -1.82227209e-01 5.09541273e-01 4.47909147e-01
4.31364864e-01 -5.87916136e-01 1.48759291e-01 4.23539728e-01
6.30266786e-01 -1.48429608e+00 -4.00666177e-01 -1.85219184e-01
-3.96988153e-01 1.34257579e+00 1.17883801e+00 -2.94148028e-01
7.62724161e-01 3.82607222e-01 2.52551407e-01 -2.47843027e-01
-5.02821505e-01 5.59388280e-01 1.97716668e-01 9.00299847e-01
7.46734262e-01 -1.74881324e-01 -1.94430143e-01 7.89664268e-01
1.26205355e-01 8.11128542e-02 9.36276913e-01 1.20432115e+00
-4.85981435e-01 -1.47306561e+00 -2.42060438e-01 3.52424443e-01
-6.41404390e-01 -2.01438248e-01 -5.25740325e-01 5.96761227e-01
2.29180798e-01 1.04771006e+00 -4.42157388e-01 -7.56581008e-01
1.46976992e-01 -1.65201649e-02 9.55557674e-02 -7.01497018e-01
-1.15328348e+00 -7.19403028e-01 -2.14399815e-01 -4.52416003e-01
-4.62556593e-02 -2.35502467e-01 -1.04030526e+00 -1.72692612e-01
-2.50290900e-01 3.67071927e-01 7.04877198e-01 7.43976235e-01
5.42889774e-01 5.31366527e-01 6.51322246e-01 -9.39048231e-02
-5.95948577e-01 -9.49736953e-01 3.09985522e-02 4.84430045e-01
2.03341886e-01 -5.28730690e-01 -1.44991949e-01 3.93662229e-02]
|
[7.08120584487915, 7.773388862609863]
|
9061b2e9-5e9a-4a2d-992d-9150411bdb5d
|
bayesian-modeling-of-lexical-resources-for
| null | null |
https://aclanthology.org/P17-1095
|
https://aclanthology.org/P17-1095.pdf
|
Bayesian Modeling of Lexical Resources for Low-Resource Settings
|
Lexical resources such as dictionaries and gazetteers are often used as auxiliary data for tasks such as part-of-speech induction and named-entity recognition. However, discriminative training with lexical features requires annotated data to reliably estimate the lexical feature weights and may result in overfitting the lexical features at the expense of features which generalize better. In this paper, we investigate a more robust approach: we stipulate that the lexicon is the result of an assumed generative process. Practically, this means that we may treat the lexical resources as observations under the proposed generative model. The lexical resources provide training data for the generative model without requiring separate data to estimate lexical feature weights. We evaluate the proposed approach in two settings: part-of-speech induction and low-resource named-entity recognition.
|
['Mark Dredze', 'Nicholas Andrews', 'Jason Eisner', 'Benjamin Van Durme']
|
2017-07-01
| null | null | null |
acl-2017-7
|
['low-resource-named-entity-recognition']
|
['natural-language-processing']
|
[ 1.03193866e-02 4.03929949e-01 -3.71528059e-01 -7.38510311e-01
-6.83477044e-01 -5.88450313e-01 7.38112569e-01 1.71854064e-01
-8.27262878e-01 8.13993216e-01 2.93528557e-01 -2.68476367e-01
2.12238163e-01 -7.15525925e-01 -6.94607615e-01 -6.53125823e-01
1.53111234e-01 5.40876925e-01 -5.39578050e-02 6.30578306e-03
4.39448096e-02 4.79030162e-01 -1.53631651e+00 -8.08292255e-02
8.48524332e-01 7.02183187e-01 3.67310047e-01 3.27931494e-01
-5.80741167e-01 5.07602751e-01 -8.09671044e-01 -5.06046176e-01
7.27280155e-02 -4.64614004e-01 -7.26705849e-01 3.20282578e-01
-2.63228118e-02 1.91100821e-01 7.56325608e-04 1.04067111e+00
3.64048481e-01 4.10385996e-01 7.67541230e-01 -8.77186060e-01
-4.75907445e-01 7.41354167e-01 -2.88817361e-02 3.52739662e-01
3.44574928e-01 -2.67731845e-01 1.35807300e+00 -1.04632235e+00
7.31630325e-01 9.13150907e-01 4.40950513e-01 5.01149058e-01
-1.23731685e+00 -4.80671376e-01 3.55731249e-01 7.90060461e-02
-1.70305204e+00 -9.41588581e-01 7.25856125e-01 -2.72167832e-01
1.04439604e+00 1.02303363e-02 2.50591964e-01 1.24543619e+00
-3.34581703e-01 9.54848111e-01 8.90911639e-01 -7.19249368e-01
1.57869697e-01 5.36384046e-01 1.50091410e-01 5.56047559e-01
4.32230681e-01 -1.91412065e-02 -6.75644875e-01 -1.36080086e-01
6.51886404e-01 -1.03840202e-01 -2.74032354e-01 -2.41006926e-01
-9.31599379e-01 8.79928708e-01 -8.26919749e-02 4.59045410e-01
-5.72529256e-01 -2.86635429e-01 3.44888777e-01 2.48997360e-01
5.83807528e-01 2.65672922e-01 -6.80788696e-01 -1.62247419e-01
-9.31195915e-01 -1.42443657e-01 1.07187378e+00 1.20648372e+00
1.11035335e+00 1.89996496e-01 2.41726890e-01 1.15278971e+00
5.51013529e-01 5.16799331e-01 8.02609384e-01 -1.98595196e-01
4.81079042e-01 4.76654887e-01 2.88291811e-03 -3.73928905e-01
-3.42575252e-01 -2.85893649e-01 -4.11160350e-01 -4.09538746e-01
4.47459012e-01 -2.28524312e-01 -9.18410897e-01 1.97266614e+00
5.09917617e-01 8.67277384e-02 1.54366985e-01 5.77770293e-01
7.57700145e-01 5.28413236e-01 3.92687649e-01 -4.70139235e-01
1.27996588e+00 -6.04103148e-01 -9.06845868e-01 -3.20643425e-01
6.65136874e-01 -7.24422991e-01 1.04839754e+00 -3.69986240e-03
-1.00224769e+00 -5.60634732e-01 -7.61488020e-01 1.34416446e-01
-5.78276634e-01 2.55162746e-01 4.58542824e-01 7.82548726e-01
-7.61778474e-01 2.47936010e-01 -8.47843111e-01 -4.98263389e-01
5.02991900e-02 2.79124349e-01 -4.96881276e-01 6.46340922e-02
-1.20204043e+00 1.03549802e+00 6.19943857e-01 6.39387816e-02
-8.30563307e-01 -2.09450439e-01 -1.16653597e+00 2.16941778e-02
4.50518459e-01 -4.47171658e-01 1.22556376e+00 -9.06723142e-01
-1.48595285e+00 1.03032839e+00 -3.45671654e-01 -1.65141717e-01
1.08833097e-01 -1.21135108e-01 -4.40852255e-01 -1.48725554e-01
7.36526400e-02 1.78132579e-01 9.29653227e-01 -9.70412612e-01
-6.29549265e-01 -2.97898024e-01 -4.31849100e-02 3.44948530e-01
-2.24847674e-01 2.16416657e-01 -4.35098708e-01 -6.43023014e-01
4.82169054e-02 -8.04586947e-01 1.03968985e-01 -6.25969052e-01
-4.57179785e-01 -3.63724828e-01 4.41151828e-01 -4.51128572e-01
1.19657564e+00 -2.10341167e+00 2.45507527e-03 2.59670556e-01
-1.44310027e-01 2.20850453e-01 -4.65079695e-02 3.69421691e-01
-1.36056736e-01 1.28439888e-01 1.53512023e-02 -5.17797828e-01
5.98921664e-02 5.54199994e-01 -2.55907863e-01 4.53513026e-01
5.06486654e-01 7.63984978e-01 -6.38550520e-01 -6.21442497e-01
1.42698899e-01 4.51841921e-01 -3.31188411e-01 3.40769082e-01
-4.57062945e-02 3.46225679e-01 -6.27168655e-01 4.40379471e-01
1.56439424e-01 -7.69585893e-02 3.41405779e-01 1.97616946e-02
-1.32506907e-01 8.86088610e-01 -1.26355672e+00 1.38980520e+00
-9.08478558e-01 3.07676256e-01 -7.44797438e-02 -1.17400086e+00
8.47370267e-01 4.49038237e-01 1.36160269e-01 -4.03813899e-01
2.81557977e-01 1.53611198e-01 1.42836437e-01 -3.51378709e-01
3.48200828e-01 -4.86464918e-01 -2.73182094e-01 4.99582112e-01
7.37558782e-01 1.44941971e-01 1.59409434e-01 -1.58304974e-01
7.82759368e-01 2.58301347e-01 8.76153111e-01 -1.11232832e-01
4.96468902e-01 -3.30047786e-01 5.62913954e-01 6.82214558e-01
4.39586416e-02 1.53931633e-01 1.12551168e-01 -9.88890529e-02
-9.63055968e-01 -9.09335852e-01 -4.41662550e-01 1.56069040e+00
-2.65260428e-01 -5.19272208e-01 -5.10049284e-01 -9.33876693e-01
-3.44313413e-01 9.82860923e-01 -4.58794564e-01 9.04564001e-03
-4.56360936e-01 -7.49883354e-01 6.16997719e-01 5.70972681e-01
-2.69201281e-03 -1.22464108e+00 -3.44316751e-01 3.43441099e-01
-5.96863292e-02 -1.10446107e+00 -5.30860484e-01 7.36176848e-01
-7.22415626e-01 -8.68774235e-01 -5.07971227e-01 -6.84731781e-01
7.82004833e-01 -2.16291860e-01 1.12596595e+00 6.60958374e-03
4.64564636e-02 3.07952940e-01 -5.45607090e-01 -3.54937464e-01
-6.21704817e-01 1.67622998e-01 2.18494132e-01 1.46942362e-01
6.52491033e-01 -5.52998126e-01 -5.64702079e-02 3.04918617e-01
-9.23306882e-01 -2.14732185e-01 8.60654831e-01 9.87895250e-01
6.68658197e-01 -2.39373654e-01 6.67390585e-01 -1.24091065e+00
3.99707973e-01 -4.75123048e-01 -5.20214736e-01 3.22985470e-01
-4.35857713e-01 5.48694849e-01 7.76889920e-01 -7.62201965e-01
-1.31505847e+00 2.03851461e-01 -3.79650444e-01 -2.39076346e-01
-6.62125170e-01 7.60323226e-01 -6.81238294e-01 1.16658680e-01
4.78359997e-01 4.50388789e-01 -4.15209651e-01 -6.47254765e-01
5.21190405e-01 7.61779189e-01 1.56820714e-01 -7.68750370e-01
8.07608247e-01 -9.20932833e-03 -3.56502533e-01 -1.18728662e+00
-1.06395292e+00 -7.67148435e-01 -8.36424351e-01 1.13988489e-01
6.28061414e-01 -8.23619306e-01 -3.22230697e-01 3.85121740e-02
-1.14387548e+00 -1.69776723e-01 -5.95680714e-01 6.51728868e-01
-5.29875517e-01 1.84732601e-01 -2.74768472e-01 -9.99521554e-01
-1.67050567e-02 -8.96780849e-01 1.13450241e+00 1.02087356e-01
-2.23148078e-01 -1.35824275e+00 2.05505222e-01 -5.73122278e-02
1.67502850e-01 -3.03830296e-01 8.63108754e-01 -1.51758587e+00
-2.21067399e-01 -3.73881757e-01 2.01311022e-01 5.16217768e-01
4.05336261e-01 -1.71193689e-01 -1.20334303e+00 7.39115402e-02
1.39773592e-01 -2.96772957e-01 5.81788063e-01 1.12849064e-01
6.73007846e-01 -4.37617004e-01 -1.09559521e-01 5.57762027e-01
1.13672352e+00 1.37736350e-01 4.14974272e-01 -6.50866926e-02
5.84892154e-01 6.50814354e-01 5.15730500e-01 5.02181530e-01
4.40211296e-01 5.86741626e-01 -2.34019697e-01 2.39436060e-01
6.46148771e-02 -4.02936995e-01 5.11385858e-01 1.35695016e+00
-2.75703031e-03 -2.80041069e-01 -1.07309663e+00 6.41982973e-01
-1.51579630e+00 -8.41671109e-01 1.76783815e-01 2.25861979e+00
1.23258424e+00 1.17671266e-01 6.05815202e-02 -1.63683787e-01
7.94785440e-01 7.27632493e-02 -2.97604442e-01 -1.23020269e-01
-4.50505540e-02 4.74027216e-01 3.20273638e-01 4.86746788e-01
-1.16436040e+00 9.94139552e-01 6.65522861e+00 7.95448780e-01
-1.07620800e+00 3.54069680e-01 1.70828462e-01 2.65249372e-01
-1.68465674e-01 1.10125072e-01 -1.37453282e+00 3.20245057e-01
1.20959067e+00 -2.48746023e-01 1.83770820e-01 9.08719242e-01
-2.32708622e-02 -5.05770035e-02 -1.27320397e+00 7.56416976e-01
1.83328658e-01 -7.13211179e-01 1.76442713e-01 1.94476917e-01
4.33923930e-01 7.04347938e-02 -1.49561748e-01 5.53795278e-01
3.16234916e-01 -6.98026538e-01 7.63203681e-01 4.26148146e-01
7.98140109e-01 -6.03528321e-01 7.09821165e-01 6.47171378e-01
-1.06956363e+00 3.19022566e-01 -4.58218873e-01 8.86637494e-02
1.64657310e-01 4.24110264e-01 -1.07943559e+00 4.83465761e-01
5.69499247e-02 2.98200190e-01 -3.68877292e-01 9.63373303e-01
-5.65042436e-01 9.31768298e-01 -4.59640443e-01 -1.75205648e-01
5.54593131e-02 -1.66173488e-01 4.00489509e-01 1.22156250e+00
1.34384692e-01 1.58842541e-02 3.98381561e-01 5.84079623e-01
-3.31139684e-01 6.45197511e-01 -6.77710354e-01 -4.01241571e-01
5.39923966e-01 1.20533776e+00 -6.94887280e-01 -4.55866516e-01
-8.39592576e-01 7.28302419e-01 4.91096258e-01 3.41532081e-01
-7.02745259e-01 -2.23736852e-01 5.38671494e-01 4.78688627e-02
5.79706550e-01 -2.72301018e-01 1.37200385e-01 -1.42527783e+00
-1.27297625e-01 -7.38472462e-01 2.05416664e-01 -2.87442088e-01
-1.46258366e+00 6.71689570e-01 1.94127962e-01 -8.76272082e-01
-7.17377365e-01 -6.96787000e-01 -4.52382058e-01 1.15182531e+00
-1.56948864e+00 -1.01975894e+00 8.99325013e-02 6.46260440e-01
6.81185305e-01 -1.63106024e-01 1.11634707e+00 1.63806826e-01
-6.11292541e-01 6.11767530e-01 -4.66114618e-02 3.78654152e-01
7.30800331e-01 -1.23295522e+00 2.67062515e-01 7.58895993e-01
7.49881864e-01 9.95348871e-01 7.11642206e-01 -7.06455767e-01
-1.04973245e+00 -1.14626813e+00 1.11009133e+00 -5.03568411e-01
9.31772232e-01 -5.59670746e-01 -9.37761128e-01 8.98847461e-01
1.45001728e-02 1.06405474e-01 1.25614297e+00 6.85145378e-01
-3.19489181e-01 1.79270029e-01 -7.88836122e-01 2.05863416e-01
9.62186217e-01 -9.03284609e-01 -9.90674436e-01 2.70098239e-01
4.25311863e-01 -1.04971826e-01 -7.81140745e-01 -8.73222798e-02
3.32174182e-01 -2.12249681e-01 5.59278190e-01 -7.70924389e-01
-1.96307600e-01 -2.27153972e-02 -3.07640433e-01 -1.41888082e+00
6.60521910e-03 -5.01246393e-01 -2.75434386e-02 1.43288350e+00
7.16726005e-01 -5.32821774e-01 4.70743835e-01 5.87647080e-01
-5.03088981e-02 -3.04659456e-01 -1.00064540e+00 -9.40285206e-01
-1.52656674e-01 -6.45108640e-01 4.02383745e-01 9.89272714e-01
-2.81246230e-02 8.42614770e-01 -5.12875728e-02 1.66299284e-01
3.95575643e-01 -5.15197515e-02 5.41856527e-01 -1.35597920e+00
-3.36692214e-01 -9.33095217e-02 -4.14845347e-01 -1.18577349e+00
8.53035092e-01 -8.71946335e-01 4.42622185e-01 -9.95058179e-01
1.13957778e-01 -6.35503411e-01 -3.02380770e-01 7.31102109e-01
-2.80779958e-01 -4.91439253e-02 -8.11592788e-02 1.61790818e-01
-6.60660684e-01 6.47711575e-01 6.55865073e-01 2.09443256e-01
-1.58079684e-01 3.08940828e-01 -5.70989966e-01 7.72823751e-01
6.28028512e-01 -6.51128292e-01 -3.43521297e-01 -2.04843864e-01
2.14110896e-01 -4.11669910e-02 6.79443553e-02 -3.48092347e-01
1.63139775e-01 -1.28463581e-01 2.25531876e-01 -2.50693411e-01
2.42677405e-01 -7.71813691e-01 1.26090914e-03 -2.47190624e-01
-3.17604214e-01 -1.32835925e-01 3.91102284e-02 5.20413935e-01
-3.92698497e-01 -7.46713758e-01 6.05629027e-01 -1.57337323e-01
-5.76446593e-01 1.05122350e-01 -4.46459264e-01 2.95791626e-01
7.21658409e-01 -2.25065321e-01 7.60833547e-02 -3.15456778e-01
-1.17039084e+00 -3.56553316e-01 4.37280387e-01 3.00153047e-01
3.81914884e-01 -1.16780579e+00 -5.61126709e-01 3.69084120e-01
2.66851902e-01 -1.76239178e-01 -1.76954418e-01 8.61961663e-01
8.61769840e-02 5.13007760e-01 6.16708212e-02 -3.66106480e-01
-9.96345401e-01 6.87882423e-01 2.02703718e-02 -2.58747548e-01
-2.08503440e-01 7.49875069e-01 4.86421138e-01 -5.50109386e-01
1.29919246e-01 -2.06710190e-01 -2.59616762e-01 4.08415586e-01
3.99834126e-01 -7.75033012e-02 3.02123338e-01 -1.11245167e+00
-3.96912485e-01 2.00192928e-01 -1.97921216e-01 -1.81423530e-01
1.40996253e+00 -3.37440401e-01 2.38425195e-01 8.84813905e-01
1.07786429e+00 3.39812517e-01 -7.96464562e-01 -5.68466842e-01
3.99011999e-01 -1.76922753e-01 7.94524699e-02 -4.45219070e-01
-6.75337195e-01 8.26815665e-01 1.22483715e-01 2.05380216e-01
1.00787270e+00 2.76150316e-01 2.67100424e-01 5.52843392e-01
6.24935627e-01 -1.08061790e+00 -4.36718613e-01 5.95725834e-01
4.53355998e-01 -1.17216468e+00 -2.64468044e-01 -3.72158796e-01
-7.10814655e-01 8.68105829e-01 3.27949226e-01 -3.70560549e-02
8.00897419e-01 2.77941495e-01 4.96689342e-02 -1.34877294e-01
-7.95140684e-01 -8.10716569e-01 5.70716500e-01 5.63076735e-01
7.15126634e-01 -7.43179172e-02 -2.74219215e-01 8.16214263e-01
-1.02746241e-01 -4.68455195e-01 2.60869205e-01 8.89007866e-01
-4.76298183e-01 -1.39663041e+00 -1.40963301e-01 3.63703817e-01
-5.88472426e-01 -4.44378436e-01 -3.74031395e-01 8.86316121e-01
3.64324041e-02 7.22378433e-01 3.00625060e-02 1.18096955e-01
4.47436571e-01 7.58239925e-01 4.64528710e-01 -1.20978594e+00
-4.50889677e-01 3.04375231e-01 3.14253479e-01 -4.50517088e-01
-7.27895916e-01 -8.34196448e-01 -1.07142770e+00 2.17586234e-01
-9.27129805e-01 4.18160349e-01 7.04434574e-01 1.20139277e+00
2.77872626e-02 2.47235507e-01 5.42908669e-01 -5.69847226e-01
-6.68446958e-01 -1.27550125e+00 -7.09755063e-01 4.98975724e-01
2.91741658e-02 -7.96718597e-01 -2.94386178e-01 2.80599773e-01]
|
[10.320990562438965, 9.621506690979004]
|
12fa1e2c-78dd-4592-b23a-52d443e70b59
|
deep-steganalysis-end-to-end-learning-with
|
1806.10443
| null |
http://arxiv.org/abs/1806.10443v1
|
http://arxiv.org/pdf/1806.10443v1.pdf
|
Deep Steganalysis: End-to-End Learning with Supervisory Information beyond Class Labels
|
Recently, deep learning has shown its power in steganalysis. However, the
proposed deep models have been often learned from pre-calculated noise
residuals with fixed high-pass filters rather than from raw images. In this
paper, we propose a new end-to-end learning framework that can learn
steganalytic features directly from pixels. In the meantime, the high-pass
filters are also automatically learned. Besides class labels, we make use of
additional pixel level supervision of cover-stego image pair to jointly and
iteratively train the proposed network which consists of a residual calculation
network and a steganalysis network. The experimental results prove the
effectiveness of the proposed architecture.
|
['Yinlong Qian', 'Jing Dong', 'Tieniu Tan', 'Wei Wang']
|
2018-06-27
| null | null | null | null |
['steganalysis']
|
['computer-vision']
|
[ 6.01948917e-01 9.48381126e-02 1.76785529e-01 -2.61577547e-01
-5.57227135e-01 9.39417332e-02 6.38934135e-01 -7.03805566e-01
-3.73948336e-01 5.69542468e-01 -2.41683573e-01 -4.31435287e-01
4.45570856e-01 -8.85407686e-01 -7.53863454e-01 -9.95059490e-01
8.56791586e-02 -2.69287348e-01 2.19841123e-01 -1.88377738e-01
2.47392491e-01 -2.05222759e-02 -1.12544060e+00 4.49104935e-01
8.16831052e-01 1.05430627e+00 2.55811006e-01 7.50766516e-01
2.99177617e-01 1.11252594e+00 -3.57812881e-01 -8.68086219e-02
5.54533482e-01 -9.45634067e-01 -3.00010830e-01 4.25291866e-01
1.30443409e-01 -5.96612275e-01 -7.44242847e-01 1.28278637e+00
4.24182862e-01 -2.65022218e-01 3.70996177e-01 -7.39001930e-01
-6.88711345e-01 7.04909742e-01 -6.44927979e-01 1.82438493e-01
-2.43037567e-01 6.13493621e-01 7.92235553e-01 -7.69677877e-01
3.44256282e-01 1.00484598e+00 6.47079110e-01 4.47847962e-01
-9.07937467e-01 -9.74990189e-01 -1.16328202e-01 2.71737307e-01
-1.03187978e+00 -5.14620841e-01 1.22854257e+00 -2.43607074e-01
7.43855774e-01 -1.32024378e-01 7.44872391e-01 1.01996183e+00
2.91064322e-01 7.81461358e-01 1.39130354e+00 -4.99987930e-01
-2.74667680e-01 2.11365610e-01 -3.84354830e-01 9.66328204e-01
2.67375320e-01 6.05735660e-01 1.26196861e-01 2.46533528e-01
9.26499546e-01 3.52198988e-01 -4.59076852e-01 -3.00781965e-01
-1.07997286e+00 9.78389740e-01 7.42131650e-01 4.68273640e-01
-2.45701402e-01 6.08161032e-01 2.27258965e-01 6.00559115e-01
3.75329643e-01 1.14068799e-01 -1.79588199e-01 4.81942624e-01
-1.01787841e+00 -4.63666558e-01 7.26460159e-01 6.53865814e-01
1.11488795e+00 5.27774572e-01 1.44540444e-01 4.27592576e-01
3.77591908e-01 4.47663635e-01 5.31963408e-01 -4.37277079e-01
3.86842668e-01 1.97700202e-01 -2.09025264e-01 -1.42858696e+00
-2.14223005e-02 -6.61545873e-01 -1.19921207e+00 5.16292810e-01
6.27428219e-02 -2.05852672e-01 -1.27047193e+00 1.37867236e+00
-5.12027889e-02 8.33668590e-01 2.77797461e-01 7.92633235e-01
3.81716996e-01 8.11512947e-01 -1.58285089e-02 -1.46498710e-01
7.77507722e-01 -1.24753654e+00 -7.30246067e-01 -2.84022927e-01
5.83589315e-01 -6.23252332e-01 5.84953070e-01 4.35257852e-01
-8.35655391e-01 -8.16927433e-01 -1.33294702e+00 4.88803983e-02
-2.93954939e-01 7.16909766e-02 3.45642239e-01 8.18821728e-01
-9.20855045e-01 6.80202961e-01 -7.50312805e-01 2.11291954e-01
6.00772619e-01 5.32117724e-01 -4.06321324e-02 -6.53094277e-02
-1.47429180e+00 6.06179953e-01 1.03621566e+00 3.71265560e-01
-1.32118666e+00 2.58357823e-02 -8.70683730e-01 2.58277923e-01
4.11365330e-01 -3.37798446e-01 7.28147864e-01 -1.65186131e+00
-1.71180320e+00 7.55019665e-01 4.44102228e-01 -6.09148383e-01
6.22833431e-01 8.43258873e-02 -6.82733297e-01 3.09335768e-01
-3.07207674e-01 3.23986173e-01 1.59958422e+00 -1.38952255e+00
-8.53800654e-01 2.79246289e-02 -1.40062630e-01 -2.92626947e-01
-4.74106342e-01 -3.68243277e-01 -2.82154977e-01 -7.15588212e-01
2.52828747e-02 -8.02492142e-01 -2.82032907e-01 -2.38597870e-01
-2.46753663e-01 2.55849421e-01 1.27748108e+00 -9.81942475e-01
1.04750013e+00 -2.27565002e+00 -3.41162384e-01 5.76835573e-01
3.58845264e-01 8.25285196e-01 -3.00549507e-01 3.02992426e-02
-2.47172743e-01 8.72063041e-02 -3.36689711e-01 1.35526627e-01
-2.49913439e-01 -2.90219098e-01 -1.93467394e-01 7.59002328e-01
3.86407554e-01 1.04958832e+00 -9.81167853e-01 -6.22144163e-01
6.02246523e-01 6.27696872e-01 -3.72268170e-01 1.94108099e-01
-2.01321661e-01 6.34180963e-01 -4.67463642e-01 3.99187595e-01
9.43826020e-01 -5.07799685e-01 3.85866165e-01 -1.38938770e-01
1.43764839e-01 -1.06141761e-01 -7.49736667e-01 1.47471070e+00
-4.73027498e-01 7.83559442e-01 9.81047153e-02 -1.32644820e+00
9.37320471e-01 3.85253698e-01 3.48842412e-01 -7.98269391e-01
6.32861018e-01 4.86624002e-01 9.41844285e-02 -6.83083057e-01
5.09435758e-02 -1.28994480e-01 1.92871928e-01 4.15047467e-01
6.93951622e-02 -4.30972911e-02 -2.55203575e-01 -2.21606195e-01
1.03639662e+00 2.94356700e-02 3.29461843e-01 9.42977145e-02
1.11963522e+00 -1.54424220e-01 4.95919049e-01 7.69839466e-01
-3.32196727e-02 4.11676228e-01 1.56722933e-01 -4.88849968e-01
-1.34987581e+00 -4.57670718e-01 1.53110370e-01 7.94650495e-01
4.59671915e-01 8.34667683e-03 -7.10648298e-01 -1.01945817e+00
-3.90793115e-01 2.80362576e-01 -4.54778492e-01 -3.44203681e-01
-9.56505060e-01 -4.69454736e-01 6.35281622e-01 7.01630935e-02
1.21644163e+00 -1.22055721e+00 -5.16882420e-01 3.70365411e-01
-5.12764305e-02 -1.11700368e+00 -4.98615384e-01 1.21637963e-01
-8.73004794e-01 -1.12443042e+00 -6.52299762e-01 -1.23957241e+00
7.48099625e-01 4.58383352e-01 7.05045164e-01 5.64185619e-01
-9.23543721e-02 -2.27132633e-01 -4.11839098e-01 -2.51983441e-02
-5.79054773e-01 4.62603904e-02 -4.43092346e-01 3.00820947e-01
2.03320384e-01 -5.69487631e-01 -7.60188937e-01 1.49754554e-01
-1.07883430e+00 1.85117200e-01 1.40622783e+00 1.31620502e+00
4.27948952e-01 6.10272825e-01 2.67418236e-01 -1.31802475e+00
1.98537230e-01 -4.17616189e-01 -6.98290229e-01 2.08163738e-01
-6.73578620e-01 1.52157888e-01 8.18148971e-01 -4.11512583e-01
-9.56029356e-01 9.23152417e-02 -1.58558711e-01 -6.93272412e-01
7.10791945e-02 4.39919651e-01 -1.79459855e-01 -4.73280668e-01
2.81070262e-01 6.92224741e-01 1.33933097e-01 -2.20269218e-01
1.51973128e-01 7.39640236e-01 4.99650389e-01 1.24126039e-01
1.31760418e+00 5.37819445e-01 -2.00207800e-01 -6.51159525e-01
-5.21042585e-01 -1.09478123e-01 -3.23294520e-01 7.55692795e-02
7.31428146e-01 -1.12404823e+00 -5.94637811e-01 8.73137593e-01
-8.99113119e-01 -3.27219337e-01 3.40904519e-02 4.67103064e-01
-4.82155412e-01 7.09526598e-01 -6.02994680e-01 -6.31830990e-01
-4.45845991e-01 -1.07059646e+00 5.68952918e-01 1.35109812e-01
6.78586304e-01 -1.20498133e+00 -2.88753808e-01 1.09406829e-01
4.93504763e-01 4.76334751e-01 6.15198612e-01 -5.79490542e-01
-8.41368377e-01 -3.57027620e-01 -5.90101540e-01 7.88435161e-01
3.22628736e-01 -4.37949061e-01 -1.01046789e+00 -7.03928888e-01
5.17179847e-01 -3.61898184e-01 1.52292383e+00 2.04154864e-01
1.27443266e+00 -4.71655220e-01 -3.38713408e-01 1.03812718e+00
1.84571183e+00 1.24500580e-01 9.13358569e-01 3.31751108e-01
8.59217584e-01 4.23682392e-01 2.70172834e-01 2.18934074e-01
-9.60379839e-02 5.63192032e-02 6.85437381e-01 -3.58005017e-01
-2.62945414e-01 -2.39683598e-01 5.16385615e-01 7.17953861e-01
1.21243760e-01 -3.79416436e-01 -6.52421176e-01 3.73238325e-01
-1.62623847e+00 -1.11527705e+00 1.28172552e-02 1.70225811e+00
6.65348351e-01 3.11061144e-01 -4.89608437e-01 2.47550741e-01
9.60983515e-01 6.04093432e-01 -5.34435332e-01 3.90697867e-02
-1.45388663e-01 1.80096135e-01 7.91838825e-01 4.73015398e-01
-1.39675629e+00 9.95813251e-01 6.00364685e+00 1.06771469e+00
-1.57526207e+00 1.39615208e-01 6.58723474e-01 5.29726207e-01
-2.02654541e-01 1.24285445e-01 -2.87026405e-01 7.19813406e-01
6.82383180e-01 4.52857524e-01 4.82750773e-01 7.05228448e-01
6.52184561e-02 3.19770932e-01 -7.04850674e-01 9.57989633e-01
1.85404807e-01 -1.21811223e+00 -8.40265155e-02 3.35329384e-01
1.04630065e+00 -1.50195375e-01 4.21178490e-01 1.95894212e-01
2.72773027e-01 -9.23282743e-01 3.34663719e-01 3.80555928e-01
1.08740664e+00 -7.44958878e-01 9.13156748e-01 2.70297378e-01
-9.56580341e-01 -4.20767367e-01 -3.91604424e-01 9.04026926e-02
-1.34009182e-01 6.05822682e-01 -1.00696003e+00 4.81300801e-01
1.97937101e-01 1.02921247e+00 -3.17341417e-01 7.55113125e-01
-5.24224520e-01 7.97814786e-01 7.61282891e-02 2.42675647e-01
5.82348585e-01 7.10578868e-03 2.50647068e-01 1.23054218e+00
4.11116630e-01 1.92892313e-01 2.15461448e-01 7.99701214e-01
-2.16939330e-01 -3.06142062e-01 -5.57115257e-01 -2.73124814e-01
7.66788572e-02 1.20265806e+00 -6.61816299e-01 -3.67240012e-01
-5.04417539e-01 1.03621745e+00 -7.38645270e-02 3.53926420e-01
-8.53113115e-01 -7.52615988e-01 3.08740079e-01 7.56349042e-02
1.00698495e+00 -7.44928122e-02 -1.28905103e-01 -1.33875239e+00
-4.25162703e-01 -9.90527987e-01 2.10757069e-02 -3.26983303e-01
-1.02191067e+00 4.01932627e-01 -6.35501385e-01 -1.48769403e+00
-1.83485538e-01 -6.88635647e-01 -8.10112476e-01 7.44928718e-01
-2.15453887e+00 -1.27827907e+00 -4.61111605e-01 7.74014831e-01
4.44307953e-01 -4.13700134e-01 3.08669716e-01 1.32569313e-01
-5.73212028e-01 5.73036313e-01 4.90723878e-01 6.88826382e-01
4.28594410e-01 -8.38915646e-01 3.74351740e-01 1.15193725e+00
-2.26685047e-01 2.41283298e-01 5.09184718e-01 -7.14913189e-01
-1.15221405e+00 -1.26170373e+00 3.31247151e-01 2.31720984e-01
5.71855068e-01 -2.18173847e-01 -8.31305563e-01 6.92357123e-01
4.95262146e-01 3.63183498e-01 1.18860722e-01 -6.08490169e-01
-3.22778106e-01 -2.68981159e-01 -1.22380781e+00 1.68260857e-01
7.83940613e-01 -4.55328822e-01 -1.95106491e-01 8.79713818e-02
5.34899771e-01 -2.19158456e-01 -3.35042506e-01 3.01029027e-01
4.60608602e-01 -9.11671460e-01 8.69510233e-01 -1.99378714e-01
5.25891900e-01 -3.10942292e-01 1.35570183e-01 -1.15204203e+00
-5.04519999e-01 -5.86098373e-01 -1.92788005e-01 7.79974103e-01
1.98313802e-01 -6.63938403e-01 9.65073764e-01 -2.28264943e-01
-6.10384382e-02 -5.22492349e-01 -5.51881850e-01 -5.01139045e-01
-2.90631413e-01 -2.04342883e-02 4.37376946e-01 9.33240950e-01
-3.36964786e-01 1.43795222e-01 -1.07343841e+00 3.16565305e-01
8.89936447e-01 6.71424046e-02 6.14300132e-01 -9.94277418e-01
-4.62314546e-01 -2.22854495e-01 -5.71309447e-01 -1.11120856e+00
2.14502826e-01 -7.15044379e-01 3.73922348e-01 -9.95072961e-01
9.09257531e-02 -4.98658299e-01 -7.27924883e-01 3.57212007e-01
-3.19032043e-01 5.33042252e-01 -1.13369361e-01 3.52611542e-01
-6.11582100e-01 4.63206738e-01 1.44908845e+00 -4.74173874e-01
8.17086548e-02 -1.62748650e-01 -5.36321402e-01 7.28932858e-01
9.81929302e-01 -7.29562879e-01 -3.54966104e-01 -5.97375214e-01
-8.32921639e-02 4.57977429e-02 4.59710449e-01 -1.18897784e+00
2.25368977e-01 -8.31090733e-02 7.29393899e-01 -3.69921803e-01
2.09817350e-01 -1.16924131e+00 -2.50233728e-02 1.05112040e+00
-1.58638850e-01 -6.96959555e-01 -4.43444639e-01 9.50193644e-01
-2.04728484e-01 -1.67979911e-01 1.08800864e+00 -3.94868731e-01
-9.35432732e-01 2.39183873e-01 -2.51349002e-01 -3.05570066e-01
1.19159865e+00 -5.78694582e-01 -3.44870649e-02 -2.94914514e-01
-4.35152620e-01 4.63175680e-03 4.58251894e-01 1.26003578e-01
1.01745629e+00 -1.15087128e+00 -7.68363953e-01 5.95066428e-01
-2.37224191e-01 -2.08700880e-01 2.90530145e-01 7.61958301e-01
-7.84236014e-01 3.52262259e-01 -4.22570407e-01 -3.64671767e-01
-9.98628139e-01 9.46590900e-01 4.29404497e-01 -5.24337769e-01
-7.59253025e-01 7.01253891e-01 1.17877372e-01 -4.83297035e-02
9.53413993e-02 -5.02477400e-02 -1.59248501e-01 -3.16721469e-01
4.95630592e-01 2.37558812e-01 -3.67520392e-01 -5.83431244e-01
3.02889906e-02 5.75443745e-01 -2.64869124e-01 1.43214703e-01
1.43293333e+00 -2.97187239e-01 -1.28781796e-01 -1.47749797e-01
1.45698833e+00 -2.43727177e-01 -1.54879022e+00 -5.48426032e-01
-1.23927385e-01 -6.48909569e-01 2.16076493e-01 -4.71049190e-01
-1.69434667e+00 9.90060568e-01 8.21251452e-01 1.74885273e-01
1.40223110e+00 -5.07450640e-01 1.20588398e+00 5.70080161e-01
2.62533247e-01 -1.03633213e+00 4.21044916e-01 3.05923134e-01
3.22082639e-01 -1.59939480e+00 -1.53774589e-01 -1.90770596e-01
-2.46036753e-01 1.17349291e+00 4.22019541e-01 -6.75750017e-01
7.44687915e-01 2.55777128e-02 8.20778012e-02 4.33566868e-02
-4.28357720e-01 -2.62049556e-01 -1.08479723e-01 6.24948204e-01
4.84374464e-02 -3.42935085e-01 -2.37340093e-01 7.47574195e-02
1.81529358e-01 1.36878401e-01 4.67643231e-01 8.68856728e-01
-8.04679334e-01 -9.99699771e-01 -3.01170737e-01 3.60341191e-01
-6.84179723e-01 -2.01346964e-01 3.59110832e-02 6.60048783e-01
4.21984881e-01 1.03896332e+00 -2.01908305e-01 -8.68628263e-01
-2.53688425e-01 -2.66216069e-01 2.28275895e-01 -4.38337535e-01
-5.18697798e-01 1.75071180e-01 -3.92812014e-01 -2.39364237e-01
-5.21925151e-01 -1.51462808e-01 -1.04047441e+00 -3.98982674e-01
-7.03850210e-01 1.34275690e-01 5.01911819e-01 9.02056456e-01
8.73028561e-02 6.87794089e-01 1.31831980e+00 -7.91177273e-01
-6.91282511e-01 -1.09309244e+00 -5.99288106e-01 3.46315205e-01
8.08260977e-01 -8.05373639e-02 -5.65945387e-01 2.64652312e-01]
|
[4.294366836547852, 8.057324409484863]
|
42f9f148-f980-4890-901e-77684b674242
|
making-better-use-of-edges-via-perceptual
| null | null |
http://openaccess.thecvf.com/content_cvpr_2015/html/Qi_Making_Better_Use_2015_CVPR_paper.html
|
http://openaccess.thecvf.com/content_cvpr_2015/papers/Qi_Making_Better_Use_2015_CVPR_paper.pdf
|
Making Better Use of Edges via Perceptual Grouping
|
We propose a perceptual grouping framework that organizes image edges into meaningful structures and demonstrate its usefulness on various computer vision tasks. Our grouper formulates edge grouping as a graph partition problem, where a learning to rank method is developed to encode probabilities of candidate edge pairs. In particular, RankSVM is employed for the first time to combine multiple Gestalt principles as cue for edge grouping. Afterwards, an edge grouping based object proposal measure is introduced that yields proposals comparable to state-of-the-art alternatives. We further show how human-like sketches can be generated from edge groupings and consequently used to deliver state-of-the-art sketch-based image retrieval performance. Last but not least, we tackle the problem of freehand human sketch segmentation by utilizing the proposed grouper to cluster strokes into semantic object parts.
|
['Timothy Hospedales', 'Tao Xiang', 'Yi-Zhe Song', 'Yi Li', 'Yonggang Qi', 'Honggang Zhang', 'Jun Guo']
|
2015-06-01
| null | null | null |
cvpr-2015-6
|
['sketch-based-image-retrieval']
|
['computer-vision']
|
[ 3.29008251e-01 7.55351186e-02 -2.74681836e-01 -2.83208072e-01
-6.02211893e-01 -6.49341345e-01 8.85040462e-01 3.87271434e-01
-2.73201764e-01 9.90740508e-02 -1.42688453e-01 -1.58213466e-01
-4.66681927e-01 -8.29383731e-01 -3.45309943e-01 -4.06128794e-01
-3.59694473e-02 5.81141829e-01 4.47698206e-01 1.52151454e-02
8.63294601e-01 8.48266065e-01 -1.60817933e+00 2.38974795e-01
8.52811635e-01 8.60182345e-01 4.41309124e-01 6.60638928e-01
-4.52824205e-01 1.92926988e-01 -4.24565583e-01 -4.42071229e-01
2.42585972e-01 -3.00014257e-01 -6.23990655e-01 5.13063192e-01
8.95258904e-01 -1.61836982e-01 -1.65022224e-01 1.10464334e+00
4.00395751e-01 3.63732666e-01 1.17937350e+00 -1.28258789e+00
-7.14234889e-01 5.16831458e-01 -5.21227539e-01 -4.10953671e-01
3.00534695e-01 -2.27803677e-01 1.53356385e+00 -1.17769039e+00
9.36995983e-01 1.19249141e+00 1.52811021e-01 3.67242128e-01
-1.44879401e+00 -3.85354519e-01 3.23494732e-01 1.99227363e-01
-1.23911560e+00 -1.95075855e-01 1.30059922e+00 -4.17422265e-01
6.98096633e-01 3.61502588e-01 6.50197744e-01 6.10295653e-01
-2.31996953e-01 1.33122778e+00 9.74494040e-01 -5.40891111e-01
3.83408874e-01 -1.61264673e-01 3.64446342e-01 1.20711386e+00
3.11383039e-01 -2.86754817e-01 -6.45063400e-01 -1.85662910e-01
1.01927650e+00 -2.78966781e-02 -1.12809815e-01 -8.64362001e-01
-1.11732256e+00 5.94718337e-01 6.88975811e-01 2.98021168e-01
-5.83731294e-01 3.09495240e-01 2.97734529e-01 -6.00131713e-02
2.59043813e-01 4.84188884e-01 5.14105260e-02 5.41477919e-01
-1.26194513e+00 1.74062163e-01 5.58530271e-01 1.01573241e+00
6.74430370e-01 -2.44706571e-01 -3.13425452e-01 1.02240598e+00
4.55843896e-01 1.63950607e-01 -1.05190560e-01 -1.01965547e+00
1.93467140e-01 6.54654026e-01 -8.55603367e-02 -1.36428595e+00
-9.49445441e-02 -2.16686070e-01 -7.41827786e-01 4.02062505e-01
3.49119455e-01 5.61457038e-01 -9.37274516e-01 1.45219159e+00
-2.22411123e-03 2.08239779e-01 -2.50847310e-01 8.19125056e-01
1.15161705e+00 3.61005962e-01 2.05168426e-01 2.29376983e-02
1.51052916e+00 -1.15671718e+00 -3.90253037e-01 2.93888953e-02
-1.22028247e-01 -7.59375215e-01 1.03834593e+00 7.15922117e-01
-1.22683692e+00 -9.17702317e-01 -9.87919271e-01 1.42568618e-01
-2.91433781e-01 4.67328370e-01 9.03604209e-01 6.15656316e-01
-1.05837357e+00 9.05453503e-01 -4.27364141e-01 -4.11384761e-01
5.51896453e-01 1.97857752e-01 -3.97246867e-01 2.61729322e-02
-4.60351914e-01 3.80439907e-01 4.33459699e-01 5.29851168e-02
-4.34597671e-01 -2.88268059e-01 -7.39693940e-01 1.95963368e-01
3.84104192e-01 -1.02220488e+00 5.81601441e-01 -5.11503041e-01
-1.38960814e+00 9.94320631e-01 -1.11465514e-01 -6.75614476e-02
5.30983865e-01 1.05827011e-01 -2.47310773e-02 6.84230745e-01
-1.15410297e-03 1.12660170e+00 1.18016052e+00 -1.79042971e+00
-5.53224623e-01 -1.51477516e-01 -3.35856616e-01 1.68527186e-01
-2.84045339e-01 -1.81380972e-01 -1.08168042e+00 -1.02714682e+00
5.40069044e-01 -6.84579730e-01 -3.23379666e-01 1.55342788e-01
-5.02626181e-01 -7.26112664e-01 4.70999569e-01 -4.79209751e-01
1.25949514e+00 -1.81484926e+00 4.23449218e-01 7.70416379e-01
4.28282499e-01 1.26826629e-01 -3.52131426e-01 5.48723876e-01
1.93460345e-01 2.39492476e-01 -3.41721237e-01 -5.94960511e-01
2.24739552e-01 2.29252595e-03 -2.46520266e-01 1.61125228e-01
2.76108980e-01 1.01825941e+00 -1.02943194e+00 -7.31572926e-01
4.35340196e-01 2.45570913e-01 -4.27582979e-01 2.17338100e-01
-1.71970934e-01 1.49971291e-01 -4.54678714e-01 6.85355067e-01
6.40950441e-01 -7.61701241e-02 2.88423777e-01 -5.62000096e-01
1.46668464e-01 -2.58693248e-01 -1.31635308e+00 1.84320819e+00
3.11821811e-02 4.98309463e-01 -2.05070406e-01 -8.72500777e-01
1.34539747e+00 -8.73037353e-02 3.91228408e-01 -2.80561298e-01
4.99843108e-03 1.71121702e-01 -2.75643140e-01 -2.59605855e-01
8.81437123e-01 1.92006826e-01 -9.71310288e-02 4.08570260e-01
2.65548378e-01 -4.38008428e-01 3.28387469e-01 3.35941643e-01
8.41518998e-01 2.40190357e-01 2.10171983e-01 -3.61005485e-01
6.08362317e-01 -3.16296935e-01 -1.70886156e-03 1.05297923e+00
-6.80185184e-02 7.96951830e-01 3.43701273e-01 -9.94189158e-02
-8.82313371e-01 -1.41415107e+00 7.10525513e-02 1.19836795e+00
4.81966406e-01 -4.06581938e-01 -6.83435798e-01 -8.49383950e-01
1.42142192e-01 5.26803851e-01 -3.92798185e-01 2.59038419e-01
-3.49409699e-01 -2.42887616e-01 2.76824057e-01 5.09458184e-01
3.77546579e-01 -1.47224927e+00 -4.54619169e-01 8.19272622e-02
1.40909001e-01 -8.75618100e-01 -5.82077324e-01 3.05129383e-02
-8.81498516e-01 -1.18330097e+00 -9.15595889e-01 -1.11008024e+00
1.08088660e+00 5.13421237e-01 1.17682326e+00 4.49902028e-01
-5.33056617e-01 9.48137462e-01 -3.93152386e-01 1.04423501e-01
-1.58922911e-01 -1.23909406e-01 -2.84000456e-01 1.63342252e-01
-6.87111989e-02 -4.22252804e-01 -8.73449981e-01 2.84484327e-01
-8.55817616e-01 2.68678486e-01 7.88780153e-01 7.77649045e-01
6.55548453e-01 -6.05818704e-02 5.14041305e-01 -8.19166899e-01
8.03102970e-01 1.27495080e-01 -3.80063862e-01 7.48510659e-01
-4.58560407e-01 2.76969165e-01 3.66522551e-01 -2.48178586e-01
-9.41132486e-01 1.80813596e-01 2.08911553e-01 -2.31926426e-01
-3.26024830e-01 1.61502063e-01 -2.51521409e-01 -3.70777577e-01
3.21376055e-01 1.32511914e-01 -2.18902722e-01 -3.31442893e-01
1.08608603e+00 5.34067571e-01 8.10904503e-01 -9.95373487e-01
5.07017195e-01 4.55460966e-01 2.76426822e-01 -1.24548542e+00
-4.32430767e-02 -8.24765325e-01 -8.34418595e-01 -6.18722916e-01
8.07767928e-01 -3.63437504e-01 -8.28344643e-01 3.41379046e-01
-1.26218855e+00 -2.70732760e-01 -8.78985152e-02 1.07044548e-01
-7.58384109e-01 9.33584630e-01 -5.91657341e-01 -1.01927662e+00
-4.19266164e-01 -9.89587665e-01 1.39472783e+00 2.88489074e-01
-1.39508799e-01 -7.68011689e-01 -3.15518647e-01 3.14627290e-01
-1.63186654e-01 9.33534279e-02 1.02445173e+00 -5.58724761e-01
-8.41551483e-01 -1.31380573e-01 -7.65519321e-01 1.15080051e-01
-4.13458375e-03 1.72467679e-01 -6.73449516e-01 -2.63580590e-01
-5.72295785e-01 8.32576975e-02 1.33574462e+00 3.47749949e-01
1.09904718e+00 1.54759303e-01 -5.36296070e-01 1.75529361e-01
1.45995641e+00 1.98891982e-01 5.27875721e-01 -1.91928387e-01
8.82133782e-01 7.48525858e-01 5.63351452e-01 3.46783936e-01
1.71132237e-01 5.73133469e-01 1.06064893e-01 -2.59963095e-01
-4.39234465e-01 -3.03958744e-01 -1.34241343e-01 8.40278745e-01
-1.78443357e-01 -4.96408194e-01 -5.52969992e-01 5.46708822e-01
-2.06609178e+00 -7.99411595e-01 -3.68312776e-01 2.13251734e+00
2.77859569e-01 3.07216495e-01 2.92771071e-01 1.88670512e-02
9.75568414e-01 3.04376394e-01 -1.22149549e-01 -3.89202386e-02
-3.08523644e-02 4.87565637e-01 3.35648268e-01 5.58037698e-01
-1.20718491e+00 1.42382979e+00 6.08816290e+00 1.08068883e+00
-6.35856926e-01 -4.76739109e-01 5.41031480e-01 7.76987791e-01
-5.29999614e-01 1.16577767e-01 -3.49761009e-01 1.44738317e-01
-3.75982583e-01 1.55319929e-01 3.89309764e-01 6.71564162e-01
-2.39710525e-01 -2.32950404e-01 -1.03742528e+00 1.21249056e+00
2.64070243e-01 -1.14371860e+00 4.73623306e-01 -1.78500772e-01
7.60599196e-01 -6.88485980e-01 7.13182017e-02 -1.76344320e-01
3.18356186e-01 -8.18681300e-01 7.33880937e-01 7.87584841e-01
5.07671535e-01 -6.19369149e-01 1.41398013e-01 -4.43523899e-02
-1.69577479e+00 1.77751422e-01 -3.93156052e-01 2.35544190e-01
1.77378833e-01 2.08935082e-01 -8.39466572e-01 7.68421471e-01
2.16458991e-01 5.53659379e-01 -7.55481124e-01 1.42199683e+00
-4.22597319e-01 2.36832529e-01 -1.27516299e-01 -3.25082183e-01
2.74184465e-01 -4.63973641e-01 5.69807768e-01 1.37260258e+00
2.78618783e-01 1.27731353e-01 4.90419358e-01 9.15305138e-01
-1.32870838e-01 3.59304279e-01 -5.38531065e-01 -4.36823256e-02
5.02578557e-01 1.57185149e+00 -1.72359228e+00 -5.63622653e-01
-2.55290698e-02 1.42514169e+00 3.20997417e-01 4.77767766e-01
-3.30675364e-01 -4.85285044e-01 2.69735038e-01 -3.04010928e-01
5.89632034e-01 -5.69455147e-01 -4.37627643e-01 -8.73294413e-01
-8.93249363e-02 -3.89548838e-01 2.59518385e-01 -8.62515748e-01
-1.58049953e+00 4.72144306e-01 2.45606396e-02 -1.01594758e+00
3.18622664e-02 -8.73572648e-01 -7.48293102e-01 4.74034935e-01
-1.21992469e+00 -1.34340644e+00 -3.50689679e-01 2.75915384e-01
7.85021961e-01 -3.07320029e-01 7.96018779e-01 1.02914140e-01
-1.13050871e-01 4.01088566e-01 -3.08631927e-01 2.47654080e-01
6.85755193e-01 -1.64659142e+00 7.38386393e-01 8.10593486e-01
9.03917491e-01 7.77922690e-01 4.66423631e-01 -8.85780811e-01
-1.12050033e+00 -7.51588464e-01 6.95717990e-01 -2.03023151e-01
4.92675036e-01 -3.77672374e-01 -7.59842038e-01 -1.16278984e-01
2.09411100e-01 -1.04859084e-01 3.41618538e-01 -2.08380688e-02
-5.29801965e-01 1.91172197e-01 -8.49591672e-01 9.25021231e-01
1.35472691e+00 -6.87806606e-01 -6.47110999e-01 -5.13775870e-02
1.74392834e-01 8.51867273e-02 -6.42393351e-01 2.93705255e-01
8.92951488e-01 -1.02198851e+00 1.34656107e+00 -6.25780821e-01
1.86829329e-01 -5.07718623e-01 -1.03709064e-01 -1.08093929e+00
-4.43179429e-01 -6.29834831e-01 1.70194939e-01 1.41040432e+00
7.76826888e-02 -2.70882338e-01 9.58943903e-01 2.79033542e-01
1.96159258e-02 -6.14226401e-01 -4.28239524e-01 -7.73113132e-01
-4.02835786e-01 -4.44364816e-01 1.03826113e-01 7.98296154e-01
-1.45416379e-01 2.74082810e-01 -1.25575975e-01 -6.20794408e-02
1.12451494e+00 4.23749536e-01 7.07530022e-01 -1.58427167e+00
-3.60891461e-01 -1.15539253e+00 -6.10497594e-01 -1.32195830e+00
3.95959496e-01 -1.17780876e+00 2.06199914e-01 -1.83730292e+00
1.08991794e-01 -6.58283412e-01 -4.30402756e-01 2.53829271e-01
-5.11233509e-01 5.38449883e-01 7.46572912e-01 4.52896617e-02
-8.13136637e-01 3.42100978e-01 1.14518332e+00 -3.94490719e-01
-1.99783117e-01 -1.07293308e-01 -4.69958812e-01 6.96792901e-01
5.88897407e-01 -2.54655719e-01 -2.91770577e-01 9.74535719e-02
8.45856965e-02 4.26953584e-02 6.48028255e-01 -9.70492899e-01
4.85865414e-01 2.64520962e-02 2.39362448e-01 -6.95729792e-01
8.98055881e-02 -8.14925671e-01 -9.60775092e-03 2.77500480e-01
-4.50724214e-01 -2.42424622e-01 -1.08934686e-01 7.27104306e-01
-1.09074628e-02 -4.81377393e-01 4.26863611e-01 -1.42627984e-01
-1.14690745e+00 2.13286981e-01 -2.67123878e-01 -2.22940281e-01
7.62757659e-01 -2.15943024e-01 -1.41475633e-01 -2.05172852e-01
-8.58157754e-01 4.80328649e-02 3.96317363e-01 4.44376647e-01
9.21546817e-01 -1.46816170e+00 -5.83599508e-01 9.70586017e-02
2.66989231e-01 -4.13491309e-01 2.72877753e-01 3.92817140e-01
-5.13445854e-01 -2.21086238e-02 -3.38725477e-01 -6.60588503e-01
-1.67646611e+00 6.16668224e-01 -3.36589664e-01 -6.18057661e-02
-5.74808657e-01 9.15802121e-01 2.05533743e-01 -9.68910754e-02
4.62126493e-01 -3.40680689e-01 -1.82907403e-01 4.10563767e-01
9.96938325e-04 4.57900554e-01 -1.47372231e-01 -4.11334068e-01
-1.86162338e-01 1.06424844e+00 2.47456998e-01 -3.65390360e-01
1.07331049e+00 1.71035811e-01 -1.23965941e-01 1.65064529e-01
9.78451490e-01 5.96879050e-02 -1.21867180e+00 -6.36523589e-02
2.69616097e-01 -5.78634918e-01 -1.60245761e-01 -5.53822339e-01
-7.91622162e-01 8.51930380e-01 4.67851192e-01 3.54032159e-01
1.01031244e+00 2.31791466e-01 6.13426685e-01 5.67095876e-01
5.05299270e-01 -1.13407600e+00 5.89633703e-01 7.23102763e-02
9.76152539e-01 -1.23389018e+00 2.19643936e-01 -8.03136230e-01
-5.66979349e-01 1.40581274e+00 3.56884785e-02 -6.06790423e-01
4.85154480e-01 -1.43974423e-01 -3.44756365e-01 -4.10278648e-01
-9.18411650e-03 -8.34147453e-01 1.09434605e+00 4.85054195e-01
2.09977612e-01 3.29830557e-01 -5.87053835e-01 4.63393718e-01
1.99632108e-01 -3.80660564e-01 -4.65043895e-02 4.93674189e-01
-6.27523601e-01 -1.48012078e+00 -1.70229077e-01 3.93117756e-01
3.64086516e-02 -3.86934280e-02 -7.91477025e-01 6.50477588e-01
3.82200442e-02 7.36051083e-01 -3.09035257e-02 -2.66953498e-01
3.34034771e-01 9.38946530e-02 8.83772254e-01 -4.66511995e-01
-6.01664543e-01 3.30117971e-01 -1.50620565e-01 -3.15175235e-01
-4.56551284e-01 -4.23847228e-01 -1.16951394e+00 3.07200044e-01
-4.78700787e-01 -5.65251783e-02 5.39704025e-01 6.16033196e-01
1.88936815e-01 5.38807511e-01 4.40170377e-01 -1.33108759e+00
1.14956178e-01 -4.74024564e-01 -5.47422409e-01 6.95403516e-01
-1.97130993e-01 -8.13737690e-01 -3.81972827e-03 1.87396724e-02]
|
[11.702627182006836, 0.436005562543869]
|
baa70a46-c872-4771-a902-6e214180ab97
|
answering-questions-over-knowledge-graphs
|
2303.02206
| null |
https://arxiv.org/abs/2303.02206v1
|
https://arxiv.org/pdf/2303.02206v1.pdf
|
Answering Questions Over Knowledge Graphs Using Logic Programming Along with Language Models
|
Question Answering over Knowledge Graphs (KGQA) is the task of answering natural language questions over a knowledge graph (KG). This task requires a model to reason over multiple edges of the KG to reach the right answer. In this work, we present a method to equip large language models (LLMs) with classic logical programming languages to provide an explainable solution to the problem. Our goal is to extract the representation of the question in the form of a Prolog query, which can then be used to answer the query programmatically. To demonstrate the effectiveness of this approach, we use the MetaQA dataset and show that our method finds the correct answer entities for all the questions in the test dataset.
|
['Kenneth Joseph', 'Navid Madani']
|
2023-03-03
| null | null | null | null |
['logical-reasoning']
|
['reasoning']
|
[-7.56997690e-02 1.14709818e+00 7.10578337e-02 -3.95819753e-01
-7.17583477e-01 -6.78256810e-01 2.74415016e-01 3.50923985e-01
2.37538457e-01 7.18960822e-01 -1.40294820e-01 -8.91368151e-01
-3.28888535e-01 -1.62625337e+00 -1.04013574e+00 2.41873294e-01
-1.00869024e-02 6.53020322e-01 7.43709207e-01 -3.66957664e-01
-2.03470681e-02 2.61275947e-01 -1.23706007e+00 6.90188110e-01
1.05414379e+00 8.18552315e-01 -9.26879123e-02 5.81935823e-01
-8.30615044e-01 1.72417438e+00 -3.03551048e-01 -8.26290309e-01
-1.02898143e-02 -4.41115409e-01 -1.73054826e+00 -2.57718325e-01
6.54841661e-01 -4.72509935e-02 -1.16962001e-01 1.13183546e+00
-3.65635604e-01 5.35822511e-02 4.77620631e-01 -1.49702525e+00
-4.35329020e-01 8.12694669e-01 1.74918383e-01 -1.34973943e-01
1.04799652e+00 -2.56527245e-01 1.53486288e+00 -7.59291649e-01
1.06410956e+00 1.48734105e+00 2.32033908e-01 5.77874541e-01
-1.03440118e+00 3.23465727e-02 -8.06625411e-02 7.30225444e-01
-1.36990356e+00 -1.13119006e-01 4.92416263e-01 -1.61780700e-01
1.29128897e+00 4.68046159e-01 5.69118023e-01 -1.23402745e-01
3.16745609e-01 4.87289220e-01 7.93717802e-01 -9.71225381e-01
2.77981728e-01 5.54582894e-01 7.96875715e-01 1.52677321e+00
5.39351940e-01 -3.44151765e-01 -5.95198214e-01 -4.02923763e-01
4.17100668e-01 -7.10754395e-01 -1.02459535e-01 -5.49797833e-01
-4.66046929e-01 1.06606865e+00 4.39593405e-01 1.82323188e-01
-2.00169384e-01 4.42264855e-01 -9.15010720e-02 5.88273466e-01
-1.27023607e-01 8.33146632e-01 -6.84260726e-01 3.64258736e-01
-2.82742351e-01 4.63582307e-01 1.76572514e+00 8.26184928e-01
1.06487930e+00 -4.22883749e-01 4.21630964e-02 -4.66499925e-02
6.21526599e-01 5.06047130e-01 -2.65193015e-01 -1.22724199e+00
4.56252575e-01 1.21081328e+00 2.52087355e-01 -1.16552544e+00
-8.59577581e-02 -5.30600138e-02 3.66948128e-01 -3.78672010e-03
6.23941898e-01 -7.45744854e-02 -4.47577745e-01 1.64061344e+00
6.74395680e-01 -1.09373406e-01 5.77967644e-01 5.31803429e-01
1.13408720e+00 7.11737752e-01 1.91667214e-01 1.12564586e-01
1.62974119e+00 -7.60476649e-01 -5.27301729e-01 -3.77435654e-01
1.05306649e+00 8.20208788e-02 8.16059053e-01 2.43361130e-01
-9.27443147e-01 -5.44771701e-02 -9.18720186e-01 -4.17713284e-01
-5.40612400e-01 -2.33742788e-01 6.15604043e-01 2.40371436e-01
-1.12013793e+00 -5.39753251e-02 -4.24171925e-01 -2.91622221e-01
8.52585509e-02 4.70165730e-01 -3.57312053e-01 -6.97099507e-01
-1.49484575e+00 1.16098571e+00 9.96951222e-01 -9.98585895e-02
-6.05899692e-01 -5.86445987e-01 -1.14302385e+00 3.95469069e-01
1.00099492e+00 -1.18033493e+00 1.05912828e+00 -5.76083958e-01
-9.27107275e-01 8.87892365e-01 -5.26426017e-01 -7.13585138e-01
-4.56281789e-02 1.04734302e-01 -4.18320894e-01 5.27992547e-01
-8.65104049e-03 5.28023601e-01 2.93324739e-01 -1.26081073e+00
-5.91509640e-01 -6.96327984e-01 9.83014405e-01 -2.30077535e-01
2.72119820e-01 6.02331571e-02 -5.26714623e-01 4.15776372e-01
3.61206472e-01 -6.63923562e-01 -6.12256408e-04 -3.72541010e-01
-5.28756380e-01 -5.26905000e-01 5.38561881e-01 -8.03232312e-01
1.10057521e+00 -1.67663538e+00 1.82434663e-01 6.65568709e-01
5.67615449e-01 -5.18310443e-02 8.35884288e-02 4.21519756e-01
8.31833854e-02 3.14457864e-01 -6.58055022e-02 5.34361720e-01
3.61672938e-01 6.40493751e-01 -6.33078158e-01 -1.87000513e-01
4.16412562e-01 1.34021115e+00 -7.38988101e-01 -8.18583429e-01
-1.96220800e-01 -2.75014371e-01 -7.69507170e-01 3.48747522e-01
-1.36550772e+00 -1.86227381e-01 -7.53291368e-01 5.03757179e-01
4.54944760e-01 -3.79507095e-01 5.66181481e-01 -1.35397971e-01
4.61128652e-01 3.96831602e-01 -1.42603219e+00 1.24813521e+00
-4.10806984e-01 4.10752803e-01 -7.78073668e-02 -7.32904077e-01
9.08654451e-01 9.96565223e-02 -1.69489354e-01 -4.89363164e-01
-2.32681111e-01 3.71896595e-01 -1.83890224e-01 -1.07132721e+00
3.19305331e-01 -4.76573259e-01 -1.88889146e-01 4.36758459e-01
-3.99526581e-03 -4.41839129e-01 3.63590002e-01 6.36865199e-01
1.21525013e+00 1.67243648e-02 2.25534037e-01 -2.02350989e-01
9.19887066e-01 8.00156713e-01 2.67784894e-01 8.21623027e-01
4.85142142e-01 -3.44750345e-01 1.06589448e+00 -6.99730694e-01
-4.85118002e-01 -9.36134577e-01 3.74503314e-01 6.38702750e-01
4.06952463e-02 -9.10569608e-01 -7.63187885e-01 -9.92499113e-01
1.19175196e-01 1.22094929e+00 -2.53341287e-01 -1.77064300e-01
-5.38654149e-01 3.06060184e-02 5.93179107e-01 2.76064277e-01
3.73499244e-01 -1.18834436e+00 -7.91540980e-01 -1.67932976e-02
-5.04290402e-01 -1.33500504e+00 4.76541221e-01 -1.12496808e-01
-7.16185451e-01 -1.77358043e+00 5.78503728e-01 -8.33945394e-01
9.42689359e-01 -3.43484133e-01 1.38266146e+00 5.48598409e-01
-1.18365204e-02 7.60272682e-01 -1.95755661e-01 -4.78188902e-01
-6.40326321e-01 -3.68868047e-03 -5.90737164e-01 -2.28313908e-01
7.68707514e-01 -8.96696821e-02 2.07851112e-01 -8.22725594e-02
-1.25306487e+00 -5.02852537e-02 3.47571634e-02 1.52270392e-01
6.97230697e-01 3.87767017e-01 5.13960540e-01 -1.24042547e+00
6.84665918e-01 -4.56122875e-01 -9.44275320e-01 1.10849583e+00
-3.71349782e-01 7.74988532e-01 7.41455793e-01 2.12927461e-01
-8.85165691e-01 5.70996590e-02 1.00460269e-01 1.58842862e-01
3.19586396e-02 1.30073071e+00 -5.61438203e-01 -2.99956352e-01
6.92963600e-01 -8.91877618e-03 -2.99790323e-01 -1.32747293e-01
6.54904544e-01 9.94673520e-02 5.61532795e-01 -1.07596421e+00
9.03031230e-01 2.19140396e-01 5.03752708e-01 -3.54730844e-01
-1.00236952e+00 -5.49479723e-02 -2.64872581e-01 3.13362740e-02
7.10821688e-01 -4.43703920e-01 -9.48943853e-01 -4.14881974e-01
-1.28593135e+00 -3.20297331e-01 -4.85580236e-01 8.30053985e-02
-6.27194285e-01 1.73078462e-01 -2.29343310e-01 -9.25374627e-01
-2.54486591e-01 -6.70175910e-01 6.23481214e-01 1.65745959e-01
-2.73245484e-01 -1.04521024e+00 1.45954311e-01 6.05898380e-01
1.61072537e-01 1.68007717e-01 1.94497705e+00 -9.94545281e-01
-9.94924307e-01 -2.53535956e-01 -2.86916435e-01 8.73568654e-02
-2.59656578e-01 -1.46808639e-01 -5.26954234e-01 3.11077178e-01
-7.85942841e-03 -3.83351982e-01 2.48478234e-01 -1.36925340e-01
8.69938850e-01 -5.92530251e-01 -1.62993446e-01 4.02625650e-02
1.74926245e+00 -1.31683514e-01 8.97280455e-01 1.64697558e-01
3.72825742e-01 8.20985496e-01 4.82283473e-01 -3.23057294e-01
9.61938560e-01 4.38446462e-01 3.83317590e-01 4.14723396e-01
1.30201936e-01 -7.49112189e-01 7.43563846e-02 1.96495384e-01
2.73189664e-01 2.11359374e-02 -1.39475930e+00 5.78906476e-01
-1.87750030e+00 -8.11109960e-01 -5.59529364e-01 1.78903925e+00
8.47936094e-01 -1.76714629e-01 -2.62818187e-01 -8.97005945e-02
1.82322919e-01 -3.38069379e-01 -2.99395561e-01 -6.93849325e-01
1.04017928e-01 5.12106359e-01 -2.77599357e-02 1.22664177e+00
-4.51367706e-01 1.18848443e+00 6.33025122e+00 4.15955544e-01
-4.37033862e-01 -2.09162936e-01 -1.13515623e-01 5.04380047e-01
-8.73789191e-01 7.31110573e-01 -6.79037690e-01 -5.22460520e-01
1.12013566e+00 -4.19725269e-01 6.85711980e-01 7.49151647e-01
-2.91364402e-01 -4.14520949e-01 -1.31030941e+00 3.34589481e-01
1.04474485e-01 -1.28840125e+00 4.35760885e-01 -2.46967241e-01
4.75978911e-01 -5.49370348e-01 -6.69686139e-01 6.03391111e-01
4.39620137e-01 -1.17493153e+00 5.36876380e-01 8.62988174e-01
3.35335493e-01 -6.93059444e-01 7.05078542e-01 6.37178779e-01
-1.02611685e+00 -2.22753137e-01 -3.91651630e-01 -5.22435419e-02
-1.80326942e-02 4.23863918e-01 -1.20408249e+00 1.10097992e+00
1.68713585e-01 -7.05293864e-02 -7.40733087e-01 7.49329686e-01
-9.06038940e-01 3.70722771e-01 -4.52744663e-01 -1.52289391e-01
-2.23466549e-02 -9.16364789e-02 3.11437458e-01 7.14641809e-01
1.67065173e-01 5.92671335e-01 6.48448095e-02 1.30571198e+00
-2.19844565e-01 1.38271600e-01 -6.43845379e-01 -2.47313976e-01
8.03124742e-04 1.00046337e+00 -1.61440969e-01 -6.21472836e-01
-4.28581089e-01 4.78869170e-01 6.88285172e-01 4.60437506e-01
-4.94843870e-01 -5.14891803e-01 -1.99909620e-02 1.96866512e-01
5.49401939e-02 -8.39103088e-02 4.39149737e-02 -1.21910822e+00
3.11318666e-01 -1.15608394e+00 9.23456967e-01 -1.40138030e+00
-8.10000002e-01 3.54498088e-01 2.10958555e-01 -9.75442156e-02
-5.96551776e-01 -6.92407846e-01 -5.71867049e-01 1.16990232e+00
-1.59704840e+00 -1.15517426e+00 -3.84440899e-01 7.82787502e-01
-2.39760011e-01 4.06861842e-01 1.06352103e+00 -1.21176071e-01
1.83683150e-02 -8.08907598e-02 -9.60656166e-01 4.91674468e-02
4.50243130e-02 -1.37314451e+00 -3.37513462e-02 8.81136477e-01
3.39625448e-01 9.37204301e-01 7.63723791e-01 -7.82423556e-01
-2.05810428e+00 -9.54088151e-01 1.64194214e+00 -8.16566706e-01
8.54580283e-01 6.57034591e-02 -1.11588538e+00 1.19958603e+00
-6.39402419e-02 -9.28345695e-02 5.44364512e-01 5.20885773e-02
-6.00398839e-01 -9.50367600e-02 -1.11734605e+00 3.25786263e-01
5.93829751e-01 -9.01060820e-01 -1.28211534e+00 2.09202111e-01
9.03659761e-01 -3.72755051e-01 -7.95078158e-01 2.46468127e-01
3.21290940e-01 -5.53227007e-01 4.96474743e-01 -1.43741643e+00
4.73220080e-01 -7.57656693e-01 -3.90312850e-01 -9.69887316e-01
3.64443988e-01 -2.90556431e-01 -5.90969265e-01 7.29468167e-01
9.06788051e-01 -7.92622745e-01 7.72275209e-01 1.43112290e+00
2.34534308e-01 -7.19629526e-01 -9.07602310e-01 -3.44866633e-01
-5.34196757e-03 -7.81063735e-01 8.25844586e-01 6.60518348e-01
6.40136600e-01 4.45855945e-01 2.09113047e-01 7.05133617e-01
4.31725353e-01 7.23351300e-01 7.41025448e-01 -1.31001902e+00
-4.74375755e-01 2.73893774e-01 -2.89188325e-01 -7.08868444e-01
5.47645450e-01 -1.20224428e+00 -6.77629784e-02 -2.32281637e+00
1.43114075e-01 -2.04352587e-01 3.45731616e-01 8.01996112e-01
5.62685877e-02 -4.49116170e-01 1.50723383e-01 -2.54735380e-01
-7.75187552e-01 -5.86760938e-02 1.11894691e+00 -3.46270233e-01
4.82231118e-02 -1.42001405e-01 -9.13127184e-01 6.62315369e-01
6.11166060e-01 -7.05229461e-01 -5.46290934e-01 -4.25241232e-01
1.26931620e+00 5.14716327e-01 7.50987828e-01 -5.59853375e-01
5.91430783e-01 -3.50419879e-01 -3.63844812e-01 -2.86493059e-02
1.92693416e-02 -1.02118909e+00 2.99164087e-01 6.43493950e-01
-3.73304129e-01 -1.16652116e-01 2.32854247e-01 2.33379364e-01
-5.23450732e-01 -8.19163144e-01 6.30464777e-03 -1.67877138e-01
-7.48270929e-01 -7.18870983e-02 -3.30902338e-02 4.09110755e-01
8.65987539e-01 2.78361320e-01 -5.66893578e-01 -4.79327887e-01
-9.37207818e-01 7.39596486e-01 2.81701744e-01 -1.01771355e-01
7.06716657e-01 -8.46115947e-01 -5.17164767e-01 -3.28738950e-02
3.11294407e-01 8.20225477e-02 -1.59902737e-01 5.45842826e-01
-9.09855187e-01 8.75999212e-01 2.44613051e-01 8.34423080e-02
-1.12490833e+00 5.67541420e-01 8.89545202e-01 -5.19522130e-01
-1.35903701e-01 3.56715113e-01 -4.13994044e-02 -6.33647680e-01
-2.08986811e-02 -5.48648536e-01 -3.00025344e-01 -3.92077535e-01
4.97726828e-01 1.62673201e-02 -2.82249860e-02 -1.94593400e-01
-5.67845404e-01 3.55019420e-01 3.58045578e-01 -1.44391403e-01
1.13732636e+00 1.57487646e-01 -7.66797185e-01 -4.65109870e-02
7.22357750e-01 3.12994719e-01 -9.65212211e-02 -3.98087561e-01
4.06659722e-01 -2.63236821e-01 -3.03997576e-01 -9.02821064e-01
-6.36941731e-01 6.35328948e-01 -2.26020604e-01 7.25393295e-01
7.48405874e-01 7.44474411e-01 4.46224421e-01 1.22774935e+00
6.59266174e-01 -6.62378967e-01 -3.17648470e-01 5.68747520e-01
8.02360058e-01 -8.38578641e-01 -1.66578621e-01 -9.01963472e-01
-3.76606643e-01 1.35739398e+00 6.28509939e-01 2.71760523e-01
2.57846415e-01 3.36016938e-02 -1.55459881e-01 -1.11267626e+00
-9.16413128e-01 -2.46107981e-01 2.10398793e-01 3.36631894e-01
-1.83862612e-01 -4.86743636e-02 -2.67792225e-01 7.40657389e-01
-3.43919694e-01 4.27953959e-01 5.30515552e-01 9.51019406e-01
-7.73582220e-01 -1.23499787e+00 -3.18476111e-01 2.04648495e-01
-4.66324598e-01 -1.14824861e-01 -9.27789152e-01 7.84152091e-01
7.40694394e-03 1.35228741e+00 -5.25586128e-01 -1.04059853e-01
5.91165900e-01 6.33573472e-01 7.92737842e-01 -8.36659193e-01
-3.37069482e-01 -9.98489499e-01 7.89467633e-01 -4.64688569e-01
-2.74214029e-01 1.76319405e-01 -1.86823153e+00 -3.00348669e-01
-2.96204627e-01 8.19180310e-01 4.38197970e-01 1.33392024e+00
2.53434300e-01 1.81838289e-01 1.95775200e-02 7.78398871e-01
-4.63837683e-01 -3.58576208e-01 -4.87235278e-01 2.67155290e-01
-9.32353884e-02 -1.10588953e-01 -2.58240581e-01 1.45275757e-01]
|
[10.112194061279297, 7.727481365203857]
|
6211f56d-abb3-4b24-898a-7f10a6075295
|
cppf-towards-robust-category-level-9d-pose
|
2203.03089
| null |
https://arxiv.org/abs/2203.03089v2
|
https://arxiv.org/pdf/2203.03089v2.pdf
|
CPPF: Towards Robust Category-Level 9D Pose Estimation in the Wild
|
In this paper, we tackle the problem of category-level 9D pose estimation in the wild, given a single RGB-D frame. Using supervised data of real-world 9D poses is tedious and erroneous, and also fails to generalize to unseen scenarios. Besides, category-level pose estimation requires a method to be able to generalize to unseen objects at test time, which is also challenging. Drawing inspirations from traditional point pair features (PPFs), in this paper, we design a novel Category-level PPF (CPPF) voting method to achieve accurate, robust and generalizable 9D pose estimation in the wild. To obtain robust pose estimation, we sample numerous point pairs on an object, and for each pair our model predicts necessary SE(3)-invariant voting statistics on object centers, orientations and scales. A novel coarse-to-fine voting algorithm is proposed to eliminate noisy point pair samples and generate final predictions from the population. To get rid of false positives in the orientation voting process, an auxiliary binary disambiguating classification task is introduced for each sampled point pair. In order to detect objects in the wild, we carefully design our sim-to-real pipeline by training on synthetic point clouds only, unless objects have ambiguous poses in geometry. Under this circumstance, color information is leveraged to disambiguate these poses. Results on standard benchmarks show that our method is on par with current state of the arts with real-world training data. Extensive experiments further show that our method is robust to noise and gives promising results under extremely challenging scenarios. Our code is available on https://github.com/qq456cvb/CPPF.
|
['Cewu Lu', 'Weiming Wang', 'Ruoxi Shi', 'Yang You']
|
2022-03-07
| null |
http://openaccess.thecvf.com//content/CVPR2022/html/You_CPPF_Towards_Robust_Category-Level_9D_Pose_Estimation_in_the_Wild_CVPR_2022_paper.html
|
http://openaccess.thecvf.com//content/CVPR2022/papers/You_CPPF_Towards_Robust_Category-Level_9D_Pose_Estimation_in_the_Wild_CVPR_2022_paper.pdf
|
cvpr-2022-1
|
['6d-pose-estimation-using-rgbd']
|
['computer-vision']
|
[ 1.39391527e-01 -1.38412297e-01 1.08224697e-01 -3.03496718e-01
-1.08267021e+00 -8.17354441e-01 5.66456258e-01 -4.56509031e-02
-1.89458296e-01 3.51211905e-01 -3.43263030e-01 -1.11248530e-02
8.24047625e-02 -6.71618164e-01 -9.49019611e-01 -5.81846952e-01
7.80514777e-02 8.58645260e-01 5.06896198e-01 -1.36823818e-01
3.76584440e-01 9.58707988e-01 -1.71077037e+00 -4.13967445e-02
5.18862665e-01 1.14354157e+00 -7.42780268e-02 5.48458457e-01
1.49553254e-01 1.04341343e-01 -5.56922376e-01 -4.11780089e-01
8.41268420e-01 -3.36035006e-02 -4.31117088e-01 3.55744839e-01
1.19477415e+00 -4.10303384e-01 1.70130525e-02 9.49342251e-01
4.74215508e-01 8.55211765e-02 7.61042058e-01 -1.34359837e+00
-1.81652997e-02 -2.61902630e-01 -7.82762527e-01 -2.56126344e-01
6.29038334e-01 4.09473509e-01 9.17990029e-01 -1.43105102e+00
8.09617043e-01 1.17888153e+00 8.28479230e-01 3.78788173e-01
-1.37981868e+00 -7.83061564e-01 1.37568384e-01 -2.52876610e-01
-1.51925290e+00 -2.05620691e-01 8.76160920e-01 -3.92221749e-01
6.91549540e-01 3.77907842e-01 8.09468091e-01 1.10337543e+00
6.42461330e-02 5.44482291e-01 1.02301681e+00 -2.43308499e-01
3.75229806e-01 -3.03690612e-01 -2.90045291e-01 4.82855827e-01
3.86352211e-01 1.70209423e-01 -6.18208289e-01 -3.09424192e-01
8.78599703e-01 2.33611450e-01 -1.85026079e-01 -1.06679749e+00
-1.51326501e+00 5.26497483e-01 8.79902840e-01 -2.83619165e-01
-2.90062904e-01 1.88922465e-01 -7.48187751e-02 5.15426788e-03
2.87899286e-01 5.75866580e-01 -5.26925862e-01 -9.55050904e-03
-7.89861143e-01 6.47501409e-01 5.55004954e-01 1.02705967e+00
9.84205842e-01 -2.87702948e-01 1.34805972e-02 5.87117076e-01
2.29578912e-01 9.76801276e-01 -4.91571836e-02 -1.31622255e+00
3.66918534e-01 6.25635564e-01 3.22154999e-01 -1.09389913e+00
-4.17366177e-01 -4.94385451e-01 -4.92593974e-01 5.57081103e-01
4.82687056e-01 2.67166287e-01 -1.16598630e+00 1.25966895e+00
8.19090366e-01 1.90520659e-01 -3.90783310e-01 1.15451145e+00
5.81967831e-01 2.83018798e-01 -3.26569200e-01 1.61595434e-01
1.04362547e+00 -4.28423136e-01 1.29942372e-01 -4.09783959e-01
3.25320214e-01 -9.21927452e-01 1.12921429e+00 4.81964201e-01
-7.48519242e-01 -4.21790212e-01 -8.72454762e-01 1.13935545e-01
-1.48020282e-01 9.23949033e-02 4.33949053e-01 4.52399552e-01
-5.81595600e-01 4.39539135e-01 -9.19387221e-01 -3.10770154e-01
4.78670180e-01 5.00913620e-01 -4.85308915e-01 -2.35204652e-01
-5.43523192e-01 7.45087802e-01 2.03062236e-01 1.32695019e-01
-6.04600370e-01 -7.32795954e-01 -8.98425758e-01 -5.08149087e-01
5.40706992e-01 -6.67227983e-01 1.30810308e+00 -5.91619432e-01
-1.26174045e+00 9.71935570e-01 1.25575848e-02 -1.02172447e-02
7.63521016e-01 -2.89909512e-01 1.09615482e-01 1.05172858e-01
2.60086894e-01 9.25024688e-01 9.23375130e-01 -1.67725813e+00
-5.34942269e-01 -6.30784988e-01 2.37975642e-02 2.06028223e-01
2.27378458e-01 -5.34622312e-01 -6.72143161e-01 -6.51330948e-01
7.51583338e-01 -1.24589860e+00 -3.05540681e-01 3.78758490e-01
-4.06836689e-01 2.55185515e-02 8.17136347e-01 -1.82883501e-01
4.57092136e-01 -2.22453856e+00 -9.62861553e-02 5.08779526e-01
1.58911124e-01 -2.91490778e-02 -1.79622509e-02 1.54758066e-01
1.13519393e-01 -1.62011713e-01 -3.00796926e-01 -4.15400565e-01
4.16475907e-02 3.06879669e-01 -2.96168029e-01 7.73557484e-01
3.86199236e-01 7.41728544e-01 -1.00820363e+00 -3.84964585e-01
6.10340357e-01 3.82719547e-01 -8.04889202e-01 6.10340051e-02
-3.10465723e-01 5.83175719e-01 -5.03400981e-01 1.16615772e+00
9.59358096e-01 -1.95717514e-01 -1.91354841e-01 -4.22858357e-01
1.31487980e-01 -5.24274893e-02 -1.52374029e+00 1.79354942e+00
-3.54578227e-01 2.22020507e-01 -1.26232073e-01 -4.68936503e-01
9.87442970e-01 -1.64095551e-01 4.44852710e-01 -3.73241603e-01
2.65055418e-01 4.36002403e-01 -2.09457085e-01 2.82388330e-02
5.58109164e-01 -4.07289937e-02 -3.26593757e-01 1.29494816e-01
1.80515833e-02 -8.95980537e-01 -5.98453954e-02 -1.97103582e-02
1.13462901e+00 3.78208250e-01 1.91341147e-01 7.11724088e-02
1.95244312e-01 3.36552113e-01 5.79079032e-01 6.91191614e-01
-1.48776531e-01 1.23444200e+00 6.63214400e-02 -4.28256303e-01
-1.03366959e+00 -1.43653274e+00 -3.11249673e-01 5.28171659e-01
4.87523943e-01 -2.84095675e-01 -3.80406350e-01 -7.62864113e-01
3.85676682e-01 4.35715050e-01 -4.18702036e-01 -3.25455479e-02
-5.86687922e-01 -5.21542609e-01 8.15453753e-02 5.29272258e-01
2.73330837e-01 -6.53228641e-01 -8.24983418e-01 3.94501351e-02
1.14847444e-01 -1.20137513e+00 -1.75318792e-01 2.74002165e-01
-8.38007689e-01 -1.21192765e+00 -6.42171979e-01 -4.97920871e-01
7.77795076e-01 4.09746140e-01 1.17628324e+00 -1.48272827e-01
-3.68424684e-01 4.43787783e-01 -4.92039919e-01 -3.10568988e-01
-3.40939276e-02 -1.07065924e-01 1.53886080e-01 -1.00918882e-01
2.39240229e-01 -4.27891225e-01 -8.20909441e-01 6.70795679e-01
-5.70162773e-01 -1.47395968e-01 6.35174870e-01 7.93932498e-01
9.52402711e-01 -3.29760760e-01 -1.25307078e-03 -4.30861115e-01
-4.67763506e-02 -1.86740443e-01 -6.92981958e-01 -1.09566867e-01
-2.10807711e-01 -7.26364851e-02 3.53494674e-01 -5.16222179e-01
-6.54944360e-01 5.85888386e-01 6.64872974e-02 -8.17235172e-01
-3.41301680e-01 5.40929940e-03 -8.33243728e-02 -4.71494555e-01
9.10915077e-01 -1.11835465e-01 -1.04587339e-01 -4.54432875e-01
2.80968219e-01 4.95948017e-01 6.52683198e-01 -6.78608418e-01
1.20962119e+00 7.56544530e-01 2.77035743e-01 -5.67846060e-01
-7.12287009e-01 -5.42081773e-01 -7.90838063e-01 -3.39446694e-01
5.10994136e-01 -1.01307118e+00 -6.46640539e-01 3.51543874e-01
-1.03912067e+00 -2.61190385e-01 -3.05380851e-01 4.71497029e-01
-6.05491638e-01 1.35319576e-01 -1.67529196e-01 -7.09344387e-01
3.99340503e-02 -1.20383644e+00 1.77648282e+00 3.03779114e-02
-2.95462281e-01 -3.85446936e-01 5.75046241e-02 3.30024064e-01
-4.67176735e-02 6.24322295e-01 2.43088707e-01 -3.15259635e-01
-8.84907544e-01 -5.67095876e-01 -6.32371232e-02 2.34611720e-01
-3.59606333e-02 2.24750862e-01 -8.28636706e-01 -3.25458407e-01
-1.86905473e-01 -4.80326980e-01 6.06913030e-01 9.84440446e-02
1.22808373e+00 2.31303036e-01 -3.28511298e-01 6.71384096e-01
1.29151702e+00 -4.02391940e-01 3.32731843e-01 2.58955002e-01
7.26946712e-01 3.48344862e-01 1.15704668e+00 4.73677725e-01
3.24831307e-01 8.62682164e-01 8.61588120e-01 4.45757732e-02
-1.66874319e-01 -3.35335225e-01 1.71901911e-01 2.40329772e-01
-1.77013189e-01 -1.62359737e-02 -1.10516679e+00 2.89063662e-01
-1.53560019e+00 -5.32910228e-01 -2.02998057e-01 2.44597220e+00
5.51001906e-01 3.71510535e-01 3.02988067e-02 1.52658328e-01
5.69766521e-01 1.93476658e-02 -5.96295297e-01 1.98532894e-01
1.50533384e-02 4.16817367e-01 8.07936311e-01 3.45627517e-01
-1.09422135e+00 9.00759816e-01 5.35464954e+00 7.17259109e-01
-1.23444223e+00 -1.28578901e-01 3.87533695e-01 -2.68101543e-01
-1.54023513e-01 1.19505852e-01 -8.11103344e-01 2.58952290e-01
2.48565271e-01 4.52366978e-01 1.39370248e-01 9.79438305e-01
-1.18425608e-01 -2.42735341e-01 -1.06873477e+00 1.28604543e+00
3.23933177e-02 -1.01218450e+00 -7.79016539e-02 9.91237760e-02
8.24972570e-01 2.76476920e-01 1.66919176e-02 1.06699832e-01
2.55208939e-01 -7.79925466e-01 9.56291378e-01 2.87552923e-01
8.07672143e-01 -5.98082006e-01 6.30300283e-01 4.70509291e-01
-1.16594708e+00 8.84341374e-02 -2.92120010e-01 -1.15071060e-02
8.92943703e-03 7.24756300e-01 -1.08357644e+00 6.08888090e-01
8.30467403e-01 6.51369095e-01 -8.19590747e-01 1.15245950e+00
-3.72410923e-01 1.62813470e-01 -7.88623691e-01 4.77280393e-02
5.40798455e-02 8.64585340e-02 6.55582845e-01 7.79509842e-01
5.65495193e-01 -9.55706611e-02 3.51566941e-01 7.04399407e-01
3.27521116e-02 -1.49488807e-01 -4.88250911e-01 6.14955366e-01
6.50562584e-01 1.30213463e+00 -1.00112283e+00 -4.43933569e-02
-8.47465172e-02 1.00445378e+00 2.00308189e-01 1.42362952e-01
-7.75931776e-01 -2.03672219e-02 6.72215641e-01 3.89275879e-01
4.73337352e-01 -4.72808033e-01 -3.72088313e-01 -1.24822974e+00
4.09764528e-01 -7.53518283e-01 1.48707822e-01 -8.70315850e-01
-1.39322317e+00 3.34810853e-01 1.01651669e-01 -1.80632269e+00
-2.23119691e-01 -8.89930189e-01 -3.79695326e-01 5.23558319e-01
-1.21286559e+00 -1.23972082e+00 -6.71691060e-01 4.95891899e-01
4.86815691e-01 3.65226686e-01 6.13147080e-01 2.55827587e-02
-7.05311671e-02 5.06743431e-01 -1.84702218e-01 9.95967463e-02
9.03341532e-01 -1.23773861e+00 4.80910301e-01 6.43925965e-01
3.42520207e-01 4.37538832e-01 7.72128224e-01 -6.47864103e-01
-1.67833233e+00 -1.11709762e+00 3.98898274e-01 -9.19494271e-01
4.26152200e-01 -6.39318764e-01 -6.01728499e-01 5.20691931e-01
-6.61012650e-01 7.14375436e-01 2.08431572e-01 -4.98565063e-02
-4.38822091e-01 -1.75423637e-01 -1.26627362e+00 4.99717772e-01
1.22492874e+00 -3.65521729e-01 -6.29347920e-01 3.53970617e-01
3.60340208e-01 -9.49133039e-01 -8.16573977e-01 8.38503599e-01
7.38663971e-01 -1.03823030e+00 1.36408615e+00 -4.98822853e-02
1.55673876e-01 -8.44344079e-01 -4.48484898e-01 -1.21939349e+00
-4.88657281e-02 -2.90286034e-01 -8.83260593e-02 8.16067576e-01
1.88696966e-01 -4.99368250e-01 1.08363712e+00 5.05008399e-01
-1.65210024e-01 -7.02995837e-01 -1.15171909e+00 -1.03864622e+00
-6.05176389e-02 -6.95217550e-01 5.56226611e-01 6.53991401e-01
-5.31941056e-01 2.22292133e-02 1.03831654e-02 5.12398303e-01
8.38873088e-01 3.92141402e-01 1.48908985e+00 -1.35105598e+00
-2.59973258e-01 -3.02625686e-01 -8.51992726e-01 -1.14806807e+00
-7.64993802e-02 -6.33097589e-01 2.26094782e-01 -1.17658675e+00
-1.61450073e-01 -6.79616153e-01 1.02222778e-01 3.43626559e-01
-1.05308756e-01 8.68669510e-01 2.69955665e-01 3.27798963e-01
-5.84681988e-01 4.76520002e-01 1.25202155e+00 9.95551143e-03
-2.57315729e-02 8.84011462e-02 -2.15046853e-01 8.29675496e-01
6.51986420e-01 -4.53541160e-01 -1.63338259e-02 -1.93178684e-01
1.77823976e-01 -1.16963483e-01 9.90569174e-01 -1.27995694e+00
1.04973711e-01 -1.97762966e-01 7.49458313e-01 -1.06605601e+00
7.87272811e-01 -1.07305765e+00 2.36439317e-01 4.19755429e-01
2.57350117e-01 -7.73182064e-02 -5.79375103e-02 4.75600243e-01
-3.54073569e-02 4.73383889e-02 7.30711997e-01 -1.87364683e-01
-7.38545358e-01 5.04512012e-01 4.09856588e-01 1.89620201e-02
1.08799565e+00 -5.78139305e-01 -1.21570811e-01 -1.02980234e-01
-6.15111113e-01 1.96517736e-01 1.12282336e+00 3.54374737e-01
6.04284525e-01 -1.40209830e+00 -5.98400593e-01 3.88370186e-01
6.80807531e-01 6.96056187e-01 4.05578129e-02 7.50850022e-01
-6.94238186e-01 -8.23627487e-02 9.24652815e-02 -1.24496818e+00
-1.03856266e+00 3.09266776e-01 2.46209651e-01 2.10632414e-01
-3.94822955e-01 8.70010316e-01 -1.44916354e-02 -8.61159861e-01
1.88023940e-01 -5.84644735e-01 4.77824450e-01 -7.65788928e-02
1.72057331e-01 2.25101262e-01 4.45570678e-01 -7.61584520e-01
-5.67851961e-01 9.85068202e-01 1.27002195e-01 -2.54584178e-02
1.26567042e+00 1.31632954e-01 2.37357751e-01 4.09066737e-01
1.14939415e+00 2.07099706e-01 -1.66002631e+00 -1.06731080e-01
-2.30005354e-01 -1.00636864e+00 -1.65819436e-01 -5.98448336e-01
-9.14575100e-01 6.53151572e-01 6.18434310e-01 -1.67174321e-02
7.11530626e-01 2.59179205e-01 5.53365290e-01 4.02984768e-01
7.85623491e-01 -9.06332076e-01 1.20500818e-01 4.99855518e-01
1.01069140e+00 -1.50007629e+00 2.15909466e-01 -6.64909959e-01
-2.72419959e-01 1.16992831e+00 7.36163139e-01 -5.23311913e-01
4.47502464e-01 2.04279259e-01 1.37118790e-02 -3.30276489e-01
-2.07577005e-01 -4.94196899e-02 3.65455389e-01 6.00291312e-01
7.26537220e-03 7.70673081e-02 1.83975562e-01 1.09187305e-01
-6.01344228e-01 -2.20168039e-01 2.37723336e-01 1.21279657e+00
-4.60071921e-01 -1.03392529e+00 -9.58310306e-01 3.89553249e-01
-6.57798126e-02 3.18998009e-01 -4.56436485e-01 8.81859183e-01
2.86040366e-01 5.77611566e-01 2.32321307e-01 -5.18462539e-01
5.74291050e-01 -2.58066028e-01 7.52107441e-01 -6.71501100e-01
-3.16027999e-01 1.56969856e-02 -1.26531914e-01 -8.75284612e-01
-4.35391337e-01 -8.30885291e-01 -1.20321918e+00 -1.44508660e-01
-4.07071739e-01 -2.79250771e-01 7.17242122e-01 5.97833753e-01
3.64540488e-01 1.12095676e-01 7.20123708e-01 -1.70664716e+00
-6.73668623e-01 -6.85342848e-01 -3.96701485e-01 5.98135293e-01
4.10473138e-01 -1.08998537e+00 -6.00980401e-01 -2.47904643e-01]
|
[7.576259613037109, -2.679096221923828]
|
a8b34e58-7baf-48c6-b2ea-2a0e4e65e347
|
tiling-and-stitching-segmentation-output-for
|
1805.12219
| null |
http://arxiv.org/abs/1805.12219v3
|
http://arxiv.org/pdf/1805.12219v3.pdf
|
Tiling and Stitching Segmentation Output for Remote Sensing: Basic Challenges and Recommendations
|
In this work we consider the application of convolutional neural networks
(CNNs) for pixel-wise labeling (a.k.a., semantic segmentation) of remote
sensing imagery (e.g., aerial color or hyperspectral imagery). Remote sensing
imagery is usually stored in the form of very large images, referred to as
"tiles", which are too large to be segmented directly using most CNNs and their
associated hardware. As a result, during label inference, smaller sub-images,
called "patches", are processed individually and then "stitched" (concatenated)
back together to create a tile-sized label map. This approach suffers from
computational ineffiency and can result in discontinuities at output
boundaries. We propose a simple alternative approach in which the input size of
the CNN is dramatically increased only during label inference. This does not
avoid stitching altogether, but substantially mitigates its limitations. We
evaluate the performance of the proposed approach against a vonventional
stitching approach using two popular segmentation CNN models and two
large-scale remote sensing imagery datasets. The results suggest that the
proposed approach substantially reduces label inference time, while also
yielding modest overall label accuracy increases. This approach contributed to
our wining entry (overall performance) in the INRIA building labeling
competition.
|
['Kyle Bradbury', 'Leslie M. Collins', 'Bohao Huang', 'Jordan M. Malof', 'Daniel Reichman']
|
2018-05-30
| null | null | null | null |
['segmentation-of-remote-sensing-imagery']
|
['miscellaneous']
|
[ 9.92736101e-01 1.09474942e-01 -1.92019101e-02 -3.93132031e-01
-6.14265680e-01 -9.90201533e-01 3.65557492e-01 2.27050290e-01
-5.36679804e-01 6.89338326e-01 -5.81733704e-01 -6.57798946e-01
7.51260743e-02 -1.07423007e+00 -7.31781244e-01 -9.15175319e-01
1.12825088e-01 1.61963359e-01 -8.57742911e-04 1.93867177e-01
2.25299820e-01 7.96252370e-01 -1.68127286e+00 1.59516141e-01
9.05502558e-01 1.25739026e+00 1.22014605e-01 5.17466486e-01
-3.31980228e-01 5.52405417e-01 -5.69326162e-01 3.79302017e-02
6.60097182e-01 -2.10683092e-01 -1.09776974e+00 6.08993769e-01
8.31364930e-01 -3.50751966e-01 3.82922649e-01 1.29857957e+00
1.57580972e-01 1.72808528e-01 3.67831141e-01 -9.01541293e-01
-1.02980271e-01 6.11316979e-01 -8.96080196e-01 -2.04154134e-01
-4.07119453e-01 7.04671293e-02 8.48617196e-01 -5.42584181e-01
6.01359308e-01 9.46804404e-01 1.02125525e+00 2.77904391e-01
-1.46452010e+00 -6.71314120e-01 7.92321563e-02 -5.26776373e-01
-1.61748493e+00 -2.92250574e-01 6.16697669e-01 -5.02972186e-01
7.12065697e-01 5.34974992e-01 6.94016218e-01 2.30004147e-01
-1.42710850e-01 5.61576128e-01 1.23323691e+00 -3.66210997e-01
5.09271383e-01 -1.72546819e-01 1.82328537e-01 7.27427185e-01
2.19621047e-01 -2.14234758e-02 1.95238575e-01 -6.97167739e-02
7.91618228e-01 1.69540551e-02 -8.81359875e-02 -8.04452300e-02
-1.11477447e+00 8.19508612e-01 8.40256095e-01 2.40604758e-01
-2.72767574e-01 3.62182587e-01 4.17066395e-01 8.07653219e-02
9.64503825e-01 4.67563748e-01 -3.79590660e-01 5.25230825e-01
-1.51128185e+00 1.46636024e-01 4.78561670e-01 6.87964141e-01
1.40298510e+00 4.51596603e-02 -2.86049470e-02 8.46671343e-01
-1.01866566e-01 1.63426042e-01 -4.46111225e-02 -8.80093396e-01
4.76385385e-01 7.12404191e-01 1.52506098e-01 -1.14918888e+00
-7.73514748e-01 -5.08473516e-01 -1.11319447e+00 4.31809157e-01
4.79369700e-01 -4.86771733e-01 -1.46214724e+00 1.32637870e+00
3.12573701e-01 2.32848868e-01 -1.57409266e-01 8.95212770e-01
5.59360921e-01 7.99524724e-01 3.33609104e-01 1.91271350e-01
1.35779917e+00 -9.48907316e-01 -2.56541878e-01 -4.25512761e-01
9.21382248e-01 -6.84878886e-01 7.00415015e-01 5.03854811e-01
-6.98512495e-01 -6.13471389e-01 -1.09990525e+00 6.78451881e-02
-6.47111654e-01 6.15998864e-01 7.76686788e-01 8.49045753e-01
-1.19536126e+00 7.33251095e-01 -7.00317323e-01 -3.47230464e-01
8.33074450e-01 4.96384263e-01 -1.58937693e-01 8.72950479e-02
-6.47641838e-01 3.86180043e-01 7.66316473e-01 4.67231363e-01
-7.49408185e-01 -6.20499372e-01 -6.81901097e-01 6.12914972e-02
5.21259189e-01 -7.30047151e-02 9.42150533e-01 -1.41013539e+00
-1.10230875e+00 1.06237042e+00 9.25450772e-02 -5.42865992e-01
3.26831311e-01 -5.71964569e-02 9.92479920e-03 7.10996538e-02
-3.70302275e-02 1.36204827e+00 7.67283797e-01 -1.49129379e+00
-8.42439711e-01 -3.09458107e-01 2.82205820e-01 1.92197729e-02
-1.46347865e-01 -1.26780599e-01 -9.70884692e-03 -8.22247267e-01
5.22767782e-01 -1.12853098e+00 -6.39595866e-01 1.90816179e-01
-6.32370412e-01 1.80073425e-01 9.47544515e-01 -5.78694820e-01
8.46086204e-01 -2.12930274e+00 -3.03378612e-01 4.41766709e-01
1.96646601e-01 4.50018853e-01 -1.21376403e-01 9.68374014e-02
-2.68059045e-01 4.91400689e-01 -8.25399995e-01 -2.52395928e-01
-4.10875678e-01 2.98975855e-01 -2.80825377e-01 6.30080938e-01
4.82127786e-01 9.51775312e-01 -7.67435908e-01 -3.94737124e-01
3.28129232e-01 2.34122857e-01 -2.63255328e-01 -1.40526235e-01
-5.89656353e-01 5.87646484e-01 -2.13194251e-01 9.00475979e-01
1.09990585e+00 -2.02527553e-01 8.49663764e-02 -3.37987453e-01
-6.33126020e-01 -3.24296914e-02 -1.21083820e+00 1.64167249e+00
-3.55351478e-01 6.49253666e-01 7.65817165e-02 -1.34003270e+00
1.16908741e+00 2.66240567e-01 5.09110928e-01 -3.26574415e-01
2.58452624e-01 2.22571999e-01 -3.94602507e-01 -1.67783368e-02
8.46952438e-01 -1.17005028e-01 6.10037558e-02 6.00138605e-01
-4.03275311e-01 -1.96554035e-01 1.91167161e-01 -3.09547246e-01
5.36254644e-01 1.66301057e-01 2.89503783e-01 -6.20681047e-01
3.25112551e-01 5.83375275e-01 4.61652875e-01 6.39105618e-01
-5.94933778e-02 7.52943337e-01 4.47367966e-01 -8.57615769e-01
-1.19087577e+00 -4.14745867e-01 -1.99339151e-01 8.75903428e-01
1.22677594e-01 6.89925775e-02 -1.07598126e+00 -6.40843153e-01
-1.65487617e-01 3.85546654e-01 -7.86894560e-01 3.14787716e-01
-5.35733163e-01 -8.75586510e-01 1.02545190e+00 4.36219513e-01
1.03795612e+00 -1.12482786e+00 -1.02942109e+00 4.08049911e-01
-1.89099923e-01 -9.60115254e-01 -1.63574353e-01 5.28563142e-01
-1.21001828e+00 -9.30496454e-01 -6.34832859e-01 -8.26606750e-01
8.75603437e-01 5.84462345e-01 8.95651460e-01 2.08027408e-01
-3.60015541e-01 -4.01324689e-01 -4.81724739e-01 -1.05512425e-01
-1.13551833e-01 3.37769508e-01 -5.68414867e-01 2.43084699e-01
1.60865515e-01 -3.06619793e-01 -5.30319929e-01 2.08305210e-01
-1.43335438e+00 3.10697705e-01 5.77648222e-01 8.00997078e-01
7.48174191e-01 4.71804529e-01 2.06279635e-01 -1.16732717e+00
8.80644321e-02 -2.38103345e-01 -1.00232434e+00 4.17790502e-01
-4.24341887e-01 -2.60594875e-01 6.86608195e-01 -2.36947164e-01
-9.44925606e-01 5.24118423e-01 -1.28514692e-02 -9.22368392e-02
-6.76603198e-01 6.72874212e-01 1.30119070e-01 -4.42049980e-01
5.79022348e-01 -2.72536445e-02 -7.68916607e-02 -5.07155716e-01
3.05853039e-01 8.01579356e-01 5.37761271e-01 -4.05378371e-01
6.51797831e-01 8.38873208e-01 2.71676984e-02 -8.34071219e-01
-8.35148215e-01 -4.75116730e-01 -6.75159514e-01 -2.36604452e-01
1.09351456e+00 -7.96756685e-01 -4.01375651e-01 7.05134571e-01
-1.14230561e+00 -5.32857001e-01 -2.64411867e-01 9.00539234e-02
-2.80207634e-01 3.39429408e-01 -5.94129205e-01 -7.24310994e-01
-3.66311818e-01 -1.15887916e+00 1.12022603e+00 3.71021181e-01
-9.01233591e-03 -8.03473949e-01 -2.88431376e-01 3.15981060e-01
3.94881517e-01 6.28177941e-01 7.58846462e-01 -1.94114715e-01
-4.99388605e-01 -3.45821725e-03 -8.41105640e-01 4.98815268e-01
1.62487566e-01 1.74580306e-01 -1.27301049e+00 -2.68250108e-01
-2.48565033e-01 -4.24326032e-01 1.00980604e+00 5.50621867e-01
1.33653188e+00 -1.45914957e-01 -1.80266231e-01 7.63738751e-01
1.80232501e+00 2.73366153e-01 8.23222280e-01 3.92976254e-01
9.06562090e-01 7.32504725e-01 6.04648292e-01 5.22308111e-01
-6.72506168e-02 4.98132616e-01 5.43719113e-01 -7.89906681e-01
-3.09885200e-02 1.69246554e-01 -1.79736704e-01 2.20556617e-01
-5.60577847e-02 -3.72578770e-01 -1.18216443e+00 5.81050098e-01
-1.61820757e+00 -5.92310667e-01 -3.72046947e-01 1.95463908e+00
7.44388223e-01 -2.42563337e-01 -2.02017725e-01 3.10434520e-01
9.79807556e-01 2.86555469e-01 -4.61855322e-01 -3.85324270e-01
-1.73269972e-01 5.86214066e-01 1.19519436e+00 4.28307027e-01
-1.71466291e+00 1.23920310e+00 5.48859024e+00 9.40664768e-01
-1.29520190e+00 -4.30597216e-02 1.08612144e+00 5.19110441e-01
6.33119047e-02 1.29352435e-01 -6.47499442e-01 1.13401942e-01
6.55207992e-01 5.78854501e-01 3.34688812e-01 6.52098656e-01
2.03692690e-01 -6.02396905e-01 -6.50641799e-01 6.37213707e-01
-3.09912682e-01 -1.46527374e+00 -4.69167950e-03 1.51695773e-01
1.09670401e+00 3.62506472e-02 3.81834470e-02 -2.26990268e-01
3.25191498e-01 -1.05396426e+00 7.58340478e-01 2.99919676e-02
1.13310850e+00 -9.20381546e-01 8.25626791e-01 1.82020605e-01
-1.50506818e+00 -7.24107921e-02 -4.89029646e-01 -2.68308744e-02
1.66874453e-02 8.18734646e-01 -6.75523639e-01 5.65259337e-01
6.00592494e-01 5.63300133e-01 -6.06044888e-01 9.81908560e-01
1.13568574e-01 7.04217076e-01 -4.21971649e-01 4.52658772e-01
9.02754426e-01 -4.49701816e-01 -3.49075324e-03 1.20739520e+00
3.03570628e-01 2.13405237e-01 4.12012190e-01 1.02960718e+00
-1.78407937e-01 -2.50793491e-02 -5.32733679e-01 -1.17487028e-01
3.12677532e-01 1.37296414e+00 -1.63541842e+00 -5.98409235e-01
-2.20923603e-01 8.28551412e-01 -3.39795984e-02 3.79635692e-01
-6.93271399e-01 -3.60033035e-01 3.79014403e-01 -2.40110576e-01
3.89168501e-01 -1.05124928e-01 -7.93744504e-01 -8.11381042e-01
-2.46726751e-01 -8.81650269e-01 1.84639275e-01 -6.95847809e-01
-7.51404166e-01 5.55350959e-01 -1.41194269e-01 -1.41772103e+00
3.02159399e-01 -3.45470011e-01 -2.54284680e-01 8.31762314e-01
-1.65825307e+00 -1.27891076e+00 -6.39244199e-01 2.98128873e-01
5.15622079e-01 4.53592032e-01 8.35118651e-01 3.19300085e-01
-5.57961583e-01 3.12403798e-01 1.20377816e-01 3.00757945e-01
2.04314455e-01 -1.06054497e+00 6.83003545e-01 9.46539760e-01
8.48823264e-02 2.34141961e-01 2.80471087e-01 -6.22863352e-01
-6.65817738e-01 -1.69642222e+00 6.97774112e-01 3.32228035e-01
3.38278949e-01 -1.69762835e-01 -8.33249867e-01 4.72200453e-01
-4.75350693e-02 2.36258730e-02 4.81357843e-01 -4.37126607e-01
-3.51154566e-01 -7.94695318e-02 -1.38010502e+00 3.50128621e-01
5.91911137e-01 -3.62341434e-01 -3.66678764e-03 3.17448705e-01
5.75382113e-01 -3.06219250e-01 -6.89226925e-01 4.48552221e-01
4.62711006e-01 -9.32490885e-01 8.04413795e-01 -7.74266347e-02
5.73209405e-01 -4.96728867e-01 -8.53459165e-02 -1.02543342e+00
-1.04917645e-01 -1.82120472e-01 6.69234633e-01 9.88802373e-01
3.77704680e-01 -6.36852503e-01 1.07480443e+00 4.05892134e-01
-1.71126634e-01 -3.37960929e-01 -7.10774124e-01 -7.77719796e-01
1.53412065e-02 -3.02444905e-01 7.94111252e-01 1.26854944e+00
-6.46844268e-01 -1.56635791e-01 -2.21018508e-01 3.52376044e-01
6.55457497e-01 2.46779546e-01 5.15135169e-01 -1.44746566e+00
5.31590246e-02 -4.68401879e-01 -9.35537368e-02 -1.00211203e+00
-8.02328512e-02 -7.99836278e-01 4.15078789e-01 -1.25775766e+00
-1.13583505e-01 -9.67410862e-01 1.77406718e-03 9.53692734e-01
8.64866674e-02 9.42834854e-01 1.25194758e-01 2.69047737e-01
-9.37031731e-02 -8.25463086e-02 1.02854002e+00 -3.78514141e-01
-3.32422316e-01 -1.41912952e-01 -3.64381760e-01 5.94765246e-01
1.25532532e+00 -5.96067131e-01 -1.41024634e-01 -5.59401035e-01
1.31921127e-01 -9.90979597e-02 6.49131894e-01 -1.17324221e+00
-3.43849175e-02 -1.50218025e-01 1.45620733e-01 -7.93156445e-01
-7.27805942e-02 -9.93377686e-01 5.19402981e-01 5.90347886e-01
-4.28583056e-01 -2.70776540e-01 3.61854136e-01 2.15404153e-01
-2.39951357e-01 -3.78580749e-01 1.03907728e+00 -3.31456512e-01
-7.69082427e-01 2.00001329e-01 -3.78470242e-01 -5.12791812e-01
1.00787461e+00 -6.02628708e-01 -2.96443999e-01 1.13567822e-01
-5.30350208e-01 -8.82599428e-02 5.64258516e-01 -1.34287387e-01
2.14977399e-01 -9.12192106e-01 -5.00726879e-01 3.46964091e-01
7.64619559e-02 4.38827336e-01 2.75101393e-01 7.40867794e-01
-1.29171228e+00 4.47977751e-01 -2.52797663e-01 -6.49286091e-01
-1.11818373e+00 2.04969972e-01 3.34734917e-01 -2.91738063e-01
-7.43634224e-01 9.94508922e-01 5.85304983e-02 -4.35115993e-01
1.58327103e-01 -7.12725937e-01 -1.29688114e-01 3.56162071e-01
3.87969673e-01 2.53707081e-01 3.43978643e-01 -5.82260013e-01
-1.21105783e-01 6.91041529e-01 1.08869471e-01 8.29459261e-03
1.27386951e+00 -4.23354432e-02 -4.76185858e-01 3.96889150e-02
1.15527833e+00 -4.84314620e-01 -1.41195726e+00 -1.54186025e-01
1.44449234e-01 -4.17156935e-01 3.81032616e-01 -6.38760388e-01
-1.51681471e+00 1.01793826e+00 6.72499180e-01 3.43657881e-01
1.31187928e+00 -4.41475034e-01 8.77040803e-01 5.14705539e-01
3.31368744e-01 -1.14789927e+00 -4.79636192e-01 3.48087728e-01
2.04745084e-01 -1.14538479e+00 5.40992878e-02 -5.53780496e-01
-3.01985741e-01 1.24981809e+00 2.94985652e-01 -1.43340126e-01
3.88920337e-01 3.94259185e-01 3.70393068e-01 -3.19212586e-01
-1.02188721e-01 -4.73605812e-01 -2.58430187e-02 4.33389932e-01
3.07194144e-01 3.44040543e-01 -2.79809028e-01 -1.08970709e-01
7.79793635e-02 2.29652189e-02 3.73595893e-01 1.15196395e+00
-7.32939243e-01 -9.31784093e-01 -7.36224592e-01 5.11583328e-01
-3.42931092e-01 -3.67893755e-01 -3.58861178e-01 5.87443590e-01
7.06578195e-01 9.30652380e-01 2.74595708e-01 -3.96420807e-01
-8.17956924e-02 -2.89705396e-02 1.23040248e-02 -6.06747389e-01
-9.38493729e-01 1.49739295e-01 3.81489135e-02 -3.89182240e-01
-8.79868507e-01 -3.10900629e-01 -1.01088583e+00 -3.19344223e-01
-5.15507340e-01 -1.58707872e-01 7.99072325e-01 8.91256332e-01
2.70306151e-02 3.51188123e-01 5.78755140e-01 -1.18575931e+00
-1.59638882e-01 -7.74441302e-01 -7.21992195e-01 6.65223300e-02
2.10089371e-01 -3.22041959e-01 -2.74849176e-01 4.08219635e-01]
|
[9.402003288269043, -1.21368408203125]
|
865853e3-486c-4bb5-9568-edf1209a3873
|
sydog-a-synthetic-dog-dataset-for-improved-2d
|
2108.00249
| null |
https://arxiv.org/abs/2108.00249v1
|
https://arxiv.org/pdf/2108.00249v1.pdf
|
SyDog: A Synthetic Dog Dataset for Improved 2D Pose Estimation
|
Estimating the pose of animals can facilitate the understanding of animal motion which is fundamental in disciplines such as biomechanics, neuroscience, ethology, robotics and the entertainment industry. Human pose estimation models have achieved high performance due to the huge amount of training data available. Achieving the same results for animal pose estimation is challenging due to the lack of animal pose datasets. To address this problem we introduce SyDog: a synthetic dataset of dogs containing ground truth pose and bounding box coordinates which was generated using the game engine, Unity. We demonstrate that pose estimation models trained on SyDog achieve better performance than models trained purely on real data and significantly reduce the need for the labour intensive labelling of images. We release the SyDog dataset as a training and evaluation benchmark for research in animal motion.
|
['Adrian Hilton', 'Charles Malleson', 'Moira Shooter']
|
2021-07-31
| null | null | null | null |
['animal-pose-estimation']
|
['computer-vision']
|
[-2.09404439e-01 8.05190429e-02 7.64931962e-02 -3.43171835e-01
-1.55452222e-01 -4.88454670e-01 2.09402189e-01 -1.72551591e-02
-7.80744016e-01 6.90368056e-01 -8.05439204e-02 5.36814518e-02
3.11598182e-01 -3.95193160e-01 -8.91101718e-01 -2.31523901e-01
-2.69111335e-01 5.66930354e-01 6.40464365e-01 -4.04103607e-01
-2.15007388e-03 5.16481102e-01 -1.64458287e+00 -2.63938725e-01
1.24951780e-01 7.13940620e-01 3.43145818e-01 8.51003766e-01
7.31834829e-01 5.55759788e-01 -3.41603100e-01 -1.64573893e-01
2.38797426e-01 -3.80220771e-01 -6.41651273e-01 -1.25635296e-01
2.91254789e-01 -5.27215004e-01 -2.22966164e-01 6.58458054e-01
4.48332667e-01 1.48277625e-01 5.35444915e-01 -1.50437236e+00
2.15967476e-01 1.12253800e-01 -5.87812662e-01 2.46214285e-01
4.32477176e-01 3.17437142e-01 7.61470318e-01 -3.85224551e-01
9.60126460e-01 1.11924160e+00 9.60479140e-01 7.32032776e-01
-1.44540954e+00 -4.26108718e-01 -3.59110653e-01 2.88977444e-01
-1.11106265e+00 -2.46798053e-01 4.88256693e-01 -6.80371761e-01
7.87434042e-01 1.89067516e-02 1.06536889e+00 1.21902370e+00
2.58393317e-01 6.30885303e-01 7.88295388e-01 -8.28316733e-02
2.62702852e-01 -3.35541427e-01 -1.64189517e-01 6.59740925e-01
3.66816312e-01 4.02521133e-01 -3.43061239e-01 -1.20940572e-03
1.10451460e+00 -3.41419786e-01 -1.79343164e-01 -1.14237785e+00
-1.50318730e+00 8.84733737e-01 6.48080468e-01 -2.95810610e-01
-3.89555335e-01 5.63020527e-01 4.51362222e-01 -5.16635776e-02
6.40748069e-02 7.71944404e-01 -6.74952507e-01 -3.63311291e-01
-5.29719055e-01 8.55134130e-01 1.01154780e+00 1.08386326e+00
3.47570330e-01 -2.36629218e-01 2.51817316e-01 6.30974233e-01
4.87437218e-01 4.52074111e-01 2.79982597e-01 -1.22809196e+00
1.81745768e-01 3.28186870e-01 2.48454556e-01 -1.15493143e+00
-8.84422421e-01 -4.03747633e-02 -3.25920582e-01 2.53309011e-01
6.10976875e-01 -1.10925911e-02 -8.92544389e-01 1.89311087e+00
7.04926372e-01 -5.08030690e-02 -1.19810522e-01 1.05233681e+00
9.76896048e-01 3.68480951e-01 4.02740538e-01 4.52753127e-01
1.49941313e+00 -8.50565493e-01 -2.86662668e-01 -7.34475553e-01
8.61571074e-01 -6.08065784e-01 6.29491925e-01 1.96288154e-01
-6.66982532e-01 -2.94452727e-01 -1.14807725e+00 -8.56173709e-02
-3.27028275e-01 2.09790453e-01 8.24850619e-01 3.38848352e-01
-6.92510128e-01 6.29635453e-01 -1.18228507e+00 -6.75493598e-01
5.70607603e-01 4.75927025e-01 -8.16449165e-01 3.12467605e-01
-8.16701353e-01 1.30260658e+00 4.16305155e-01 2.27483019e-01
-8.39026511e-01 -5.54983854e-01 -1.30724156e+00 -5.14263272e-01
2.05321535e-01 -6.76928997e-01 1.71252573e+00 -1.88309222e-01
-1.31447458e+00 1.28520656e+00 5.18256724e-01 -8.66121352e-01
8.45085561e-01 -4.68233109e-01 3.16088051e-01 3.31226103e-02
3.33089620e-01 1.51762307e+00 5.68509936e-01 -8.65360081e-01
-3.31292540e-01 -4.05229628e-01 -9.11679193e-02 6.89273607e-03
5.00351548e-01 -2.15480149e-01 -2.52362192e-01 -6.89959109e-01
8.73592943e-02 -1.54726577e+00 -6.55548871e-01 3.32823783e-01
-7.36743212e-02 1.86362296e-01 9.48923469e-01 -7.26411581e-01
6.56347692e-01 -1.81436467e+00 3.54761750e-01 -1.40829057e-01
1.10099658e-01 2.19334945e-01 1.99106131e-02 3.61385137e-01
6.24177866e-02 -3.88819516e-01 -1.33463219e-01 5.58404066e-02
-1.75108299e-01 4.25149530e-01 1.17281996e-01 7.70056188e-01
1.39891088e-01 9.74980474e-01 -8.01246226e-01 -6.42377675e-01
5.24023294e-01 2.45187864e-01 -9.93234277e-01 2.94362098e-01
-3.28686744e-01 7.86306083e-01 -3.35961461e-01 4.66453075e-01
4.23384041e-01 -2.08282173e-01 5.76266386e-02 -2.85403579e-01
2.57535905e-01 -3.81619036e-02 -9.12474811e-01 1.74331582e+00
-3.03790301e-01 6.74917698e-01 6.59029186e-02 -7.32772410e-01
5.92068434e-01 3.13100815e-02 6.89295590e-01 -2.74160832e-01
4.52895224e-01 2.20347673e-01 2.52415478e-01 -5.51176727e-01
5.07042766e-01 -7.39824250e-02 -5.07121384e-01 6.87719882e-02
8.39104205e-02 -8.93060923e-01 2.57288456e-01 -1.22770920e-01
1.08450413e+00 8.43497038e-01 5.29825509e-01 -2.84069300e-01
1.63907781e-01 6.20186925e-01 2.68801659e-01 1.56382203e-01
-4.49744403e-01 5.64649105e-01 2.09176898e-01 -6.69490933e-01
-1.38037229e+00 -9.17981565e-01 -2.18602106e-01 9.21598196e-01
3.45115960e-01 -3.62365484e-01 -1.08303583e+00 -2.41385773e-01
2.63504922e-01 2.87775815e-01 -8.22575212e-01 -2.78205872e-01
-8.56393158e-01 -5.54408014e-01 4.73390907e-01 8.49773109e-01
4.14697587e-01 -1.41968465e+00 -1.76026130e+00 3.34517330e-01
-1.80842564e-01 -1.68570673e+00 -1.97022989e-01 2.97842443e-01
-7.15598464e-01 -1.11250913e+00 -5.14150262e-01 -8.64454508e-01
5.01728714e-01 -2.31253468e-02 1.04332256e+00 -6.37649521e-02
-8.17571104e-01 6.47624880e-02 -3.06657434e-01 -6.96667373e-01
-3.44425648e-01 1.30373076e-01 1.97794922e-02 -9.89772379e-01
2.95762748e-01 -3.89674932e-01 -7.88671911e-01 7.61633039e-01
-6.10393047e-01 4.05226886e-01 4.36823130e-01 7.63000071e-01
4.40786153e-01 -6.87867165e-01 4.53487784e-01 -6.26840234e-01
1.35434538e-01 -3.52951109e-01 -8.38403285e-01 -3.82889003e-01
1.38999283e-01 9.99251381e-02 2.67273426e-01 -4.94590461e-01
-3.94883394e-01 5.45888901e-01 -2.44025603e-01 -5.61647788e-02
-3.05497527e-01 2.58959800e-01 2.01391727e-01 -2.46641055e-01
7.35284984e-01 -4.77844715e-01 3.18537593e-01 -4.07573134e-01
2.05717966e-01 2.11637765e-01 9.93188918e-01 -3.31056654e-01
4.67980355e-01 4.16347355e-01 3.69816840e-01 -8.37605178e-01
-5.74308634e-01 -5.58031023e-01 -8.49166870e-01 -4.04951841e-01
1.16120076e+00 -8.79701853e-01 -9.80410933e-01 5.26262522e-01
-1.17167795e+00 -5.03217041e-01 -2.12434962e-01 8.43814850e-01
-1.21344602e+00 7.44002908e-02 -5.07573366e-01 -3.89130235e-01
-1.40464827e-01 -1.24214911e+00 1.45082211e+00 3.07155997e-02
-9.04763222e-01 -6.21076763e-01 1.50513127e-01 4.50186998e-01
3.44271027e-02 9.67243373e-01 4.92072135e-01 -1.99042559e-01
-4.98980403e-01 -4.80154634e-01 1.14064775e-01 5.16646691e-02
-2.81074822e-01 -1.87711984e-01 -5.49436867e-01 -2.59037167e-01
-2.83671796e-01 -6.83112323e-01 4.19614166e-01 5.81330001e-01
7.10941136e-01 2.10441783e-01 -4.98042852e-01 4.35878962e-01
1.14376593e+00 -7.14486167e-02 4.31942314e-01 6.20491326e-01
6.73729777e-01 1.02180564e+00 9.16571975e-01 4.11630005e-01
4.06051725e-01 1.21955812e+00 6.80190742e-01 4.50953394e-02
-1.44530147e-01 -5.18800974e-01 -8.05364400e-02 4.01915967e-01
-4.89833131e-02 2.81504057e-02 -1.14061558e+00 7.39264190e-01
-1.91488671e+00 -7.91793764e-01 -2.32369304e-01 1.97033393e+00
5.76986969e-01 -2.56133117e-02 4.86800909e-01 1.00341596e-01
3.02613139e-01 -2.68047392e-01 -4.19523567e-01 -2.78332084e-01
3.55105907e-01 1.15666784e-01 8.77974868e-01 1.90970078e-01
-1.41569662e+00 1.04376352e+00 6.93683624e+00 3.42062145e-01
-8.76079679e-01 -2.46588439e-01 1.08774163e-01 9.43280831e-02
5.80544472e-01 -3.05741161e-01 -6.97277188e-01 2.59023964e-01
6.87402487e-01 3.82317640e-02 8.95176977e-02 1.25524199e+00
1.92236111e-01 -6.27234638e-01 -1.42392337e+00 9.11729276e-01
-1.73485801e-01 -9.00882244e-01 -6.01402938e-01 1.78899854e-01
4.86584604e-01 1.03854805e-01 -2.34924763e-01 7.84789920e-02
3.92881751e-01 -1.14621794e+00 9.90903974e-01 8.81778896e-02
4.16953266e-01 -6.12960041e-01 8.50572705e-01 6.62030697e-01
-1.11447191e+00 2.26947978e-01 -2.52307206e-01 -1.45774856e-01
2.34612882e-01 -2.55179197e-01 -1.31105614e+00 2.95856427e-02
9.32045758e-01 5.25961637e-01 -8.66908729e-01 1.32541442e+00
-9.43446681e-02 2.55755305e-01 -7.68751264e-01 -2.59917766e-01
1.67129755e-01 -1.69603035e-01 5.00082970e-01 9.13202167e-01
6.55588973e-03 -1.40233248e-01 2.39838697e-02 5.98622918e-01
1.82368055e-01 -1.01755351e-01 -7.04471350e-01 1.70881838e-01
8.50379020e-02 1.20751905e+00 -1.08765578e+00 1.03708707e-01
4.91010025e-02 7.30036199e-01 1.87228084e-01 -2.27462053e-01
-1.10663998e+00 -5.37465475e-02 6.21605814e-01 3.76070827e-01
4.21991438e-01 -5.19016266e-01 3.27686332e-02 -6.65947616e-01
-2.50716470e-02 -6.77140951e-01 -5.98968491e-02 -7.63144672e-01
-6.66011512e-01 4.57221389e-01 6.88009977e-01 -1.42096472e+00
-8.82017553e-01 -8.16879690e-01 -6.56890348e-02 4.24765259e-01
-7.51787066e-01 -1.19513786e+00 -5.15712559e-01 -7.77678471e-03
4.70005929e-01 2.98546314e-01 7.33730912e-01 1.86273679e-01
-1.91455305e-01 1.99867472e-01 -3.72132629e-01 9.40678641e-02
4.38582987e-01 -9.94144261e-01 8.65212977e-01 4.21257496e-01
1.83308497e-02 1.87209055e-01 1.30614269e+00 -5.77835202e-01
-1.12633514e+00 -9.27980065e-01 3.66196901e-01 -8.55631828e-01
5.50992787e-01 -4.35736537e-01 -5.47803760e-01 6.71033144e-01
-3.57565314e-01 3.31140041e-01 3.42665195e-01 -4.86138970e-01
1.55414790e-01 4.64914382e-01 -1.20704126e+00 7.10841477e-01
1.20635557e+00 -4.20418903e-02 -6.09979510e-01 1.96036920e-01
5.20161211e-01 -9.15594161e-01 -9.29089665e-01 7.77849376e-01
9.84597027e-01 -7.03056335e-01 1.11801338e+00 -6.22935295e-01
5.76154768e-01 -4.12678599e-01 -1.87348202e-02 -1.27384913e+00
-2.57711783e-02 -8.62491578e-02 1.66372448e-01 5.44036627e-01
2.24357188e-01 -1.57029212e-01 1.12472069e+00 3.52408797e-01
2.12939098e-01 -7.80208230e-01 -9.41653430e-01 -7.13221848e-01
-1.55001834e-01 -3.35580975e-01 4.27341610e-01 4.35860515e-01
-1.72650173e-01 3.87142628e-01 -4.60669786e-01 -1.16980553e-01
5.64763069e-01 2.55585872e-02 1.33934140e+00 -1.37406075e+00
-2.67419755e-01 -3.85322981e-02 -1.24805224e+00 -1.07045782e+00
1.48957491e-01 -3.97670835e-01 6.10062838e-01 -1.53045571e+00
-1.53863907e-01 -1.67637438e-01 6.96660757e-01 4.35357988e-01
1.06646493e-02 7.51030743e-01 1.53391540e-01 4.90996763e-02
-5.95164776e-01 3.21055740e-01 1.41050494e+00 2.99847096e-01
6.05218066e-03 4.20160033e-02 -3.84359509e-02 1.08911467e+00
6.23961806e-01 -5.17460644e-01 -3.16986442e-01 -1.71916083e-01
1.17937699e-01 1.77188292e-01 9.30041611e-01 -1.25201261e+00
-1.23453059e-03 -4.31367941e-02 4.53860104e-01 -7.24342346e-01
6.01148486e-01 -1.08168554e+00 4.97855902e-01 8.92384052e-01
-1.71495020e-01 1.78611040e-01 4.50770855e-01 6.32662237e-01
2.59086024e-02 -7.93740079e-02 9.87325609e-01 -2.29884461e-01
-9.17671859e-01 1.39699101e-01 -3.60860765e-01 2.16220975e-01
1.49078071e+00 -3.72719705e-01 -1.09988846e-01 -3.26995403e-01
-6.19016171e-01 2.91147858e-01 6.85566425e-01 6.77946448e-01
3.01743567e-01 -1.27163947e+00 -4.87334430e-01 2.39712745e-01
3.97142738e-01 3.79213005e-01 3.49829495e-02 9.06697989e-01
-1.49462521e+00 3.14754754e-01 -6.43727243e-01 -9.25999105e-01
-1.44754159e+00 4.46807116e-01 2.95666665e-01 -1.27492070e-01
-6.27387583e-01 7.48770893e-01 3.09247494e-01 -6.05006218e-01
-3.86978723e-02 -5.95272660e-01 -2.28005648e-01 -3.09359491e-01
1.02532677e-01 4.63723391e-01 3.02130450e-02 -1.20531094e+00
-5.62581360e-01 7.23277092e-01 2.20301360e-01 -7.41410106e-02
1.50631547e+00 1.50224790e-01 3.22388589e-01 1.02443710e-01
1.16171098e+00 -5.29693961e-01 -1.35704565e+00 3.48475426e-01
2.88179100e-01 -4.18268412e-01 -2.33462498e-01 -2.09735528e-01
-6.76793694e-01 7.20778823e-01 6.07697487e-01 -6.95057586e-02
4.29359049e-01 2.28661776e-01 8.67341876e-01 4.14889812e-01
8.67243052e-01 -1.18293679e+00 1.00855157e-01 2.81381547e-01
1.05890179e+00 -1.29456937e+00 1.97432935e-01 -5.43521225e-01
-7.46119440e-01 7.12465048e-01 8.79616201e-01 -4.07562107e-01
6.47880673e-01 5.39164782e-01 7.62324706e-02 -2.50297338e-01
-5.16199529e-01 -2.21596792e-01 2.64051497e-01 8.56969178e-01
4.34533238e-01 1.00477405e-01 -2.76446402e-01 2.32428282e-01
-8.89006793e-01 1.36752456e-01 3.76373947e-01 1.18728638e+00
-2.68130809e-01 -9.31021094e-01 -5.02707660e-01 2.68144906e-01
-4.87014472e-01 4.39382553e-01 -2.92620242e-01 9.87453282e-01
8.89251232e-02 5.92544436e-01 -1.59362063e-01 -2.86587059e-01
5.63451231e-01 -5.28560519e-01 8.52549016e-01 -7.81811893e-01
-3.66321266e-01 -1.35516614e-01 2.80816704e-01 -6.24239802e-01
-6.29271150e-01 -5.36845028e-01 -1.25518119e+00 -3.06575507e-01
-3.31326962e-01 3.91096473e-02 8.27793717e-01 7.02165365e-01
-3.15326080e-03 4.50337052e-01 7.43897445e-03 -1.49247885e+00
-5.32393456e-01 -9.80841935e-01 -3.76151979e-01 5.62324643e-01
2.31684238e-01 -1.05361116e+00 -1.46904901e-01 3.19725364e-01]
|
[7.561185836791992, -0.9850667715072632]
|
9bb2a102-6d9c-41ee-a260-9e04e542253a
|
partially-personalized-federated-learning
|
2305.18285
| null |
https://arxiv.org/abs/2305.18285v1
|
https://arxiv.org/pdf/2305.18285v1.pdf
|
Partially Personalized Federated Learning: Breaking the Curse of Data Heterogeneity
|
We present a partially personalized formulation of Federated Learning (FL) that strikes a balance between the flexibility of personalization and cooperativeness of global training. In our framework, we split the variables into global parameters, which are shared across all clients, and individual local parameters, which are kept private. We prove that under the right split of parameters, it is possible to find global parameters that allow each client to fit their data perfectly, and refer to the obtained problem as overpersonalized. For instance, the shared global parameters can be used to learn good data representations, whereas the personalized layers are fine-tuned for a specific client. Moreover, we present a simple algorithm for the partially personalized formulation that offers significant benefits to all clients. In particular, it breaks the curse of data heterogeneity in several settings, such as training with local steps, asynchronous training, and Byzantine-robust training.
|
['Samuel Horváth', 'Eduard Gorbunov', 'Rustem Islamov', 'Konstantin Mishchenko']
|
2023-05-29
| null | null | null | null |
['personalized-federated-learning']
|
['methodology']
|
[-5.39925396e-01 3.49871874e-01 -4.33619589e-01 -4.83210117e-01
-7.49745607e-01 -8.20283175e-01 1.97559923e-01 -3.68206799e-02
-3.18107367e-01 6.74926281e-01 2.87639052e-01 -1.15365386e-01
-6.20691955e-01 -7.85270512e-01 -8.24941516e-01 -1.11733079e+00
-1.44237980e-01 7.15605438e-01 -2.07798854e-01 6.53616488e-02
-2.97852904e-01 6.02625251e-01 -1.29460239e+00 3.45489115e-01
7.52266109e-01 9.33751941e-01 -3.60891223e-02 4.63283837e-01
-2.12096959e-01 6.47218525e-01 -5.08809328e-01 -6.42572761e-01
4.85710949e-01 6.11355267e-02 -1.02379918e+00 3.14940929e-01
2.09272265e-01 -4.01310742e-01 -3.51307839e-01 8.21473181e-01
3.14301878e-01 1.98391691e-01 -1.77353974e-02 -1.32781148e+00
-3.28615874e-01 1.14379644e+00 -2.08585307e-01 -2.25498870e-01
-2.43796468e-01 2.26400509e-01 1.13564825e+00 -1.93884671e-01
4.88604695e-01 1.02229083e+00 5.18144727e-01 6.48893654e-01
-1.44217741e+00 -3.94458085e-01 4.73651081e-01 -2.72485942e-01
-1.00548112e+00 -5.81017852e-01 4.69182402e-01 -2.82344371e-01
5.26288748e-01 4.64895517e-01 2.60349840e-01 1.11473739e+00
-2.68153995e-01 7.02995598e-01 8.23870480e-01 -2.18386918e-01
5.70028424e-01 5.45807123e-01 2.80793399e-01 4.52393442e-01
2.70359010e-01 -1.05163395e-01 -5.93014479e-01 -7.42884755e-01
6.52201176e-01 4.59239244e-01 -5.50558567e-01 -7.86629498e-01
-8.88215780e-01 8.30736399e-01 2.87924439e-01 2.63858616e-01
-5.03810048e-01 1.10442221e-01 4.25059706e-01 5.76601803e-01
3.45267922e-01 3.03082943e-01 -9.06411588e-01 4.79893200e-02
-5.80004692e-01 1.18897773e-01 1.05304742e+00 9.08588290e-01
1.27859008e+00 -4.11962450e-01 -1.94970593e-01 8.28818679e-01
-2.51931623e-02 2.30341926e-01 6.27509058e-01 -1.21008849e+00
6.37579143e-01 5.80900371e-01 3.56582880e-01 -5.50029993e-01
-2.53606409e-01 -5.77705324e-01 -6.59742653e-01 -1.23038873e-01
3.56532544e-01 -6.28142178e-01 -2.55838871e-01 2.23972750e+00
6.42499149e-01 -5.14640026e-02 2.57385187e-02 8.28461885e-01
1.33496493e-01 3.87986392e-01 -3.94641235e-02 -8.50032941e-02
1.08559191e+00 -1.27565372e+00 -3.82759392e-01 -2.31677800e-01
7.40490198e-01 -1.19656920e-01 9.05526102e-01 2.03409269e-01
-1.11858141e+00 3.17235338e-03 -5.82986236e-01 3.03302020e-01
-1.71321467e-01 -2.41069227e-01 8.11267316e-01 8.59391987e-01
-1.11429095e+00 9.00318027e-01 -8.03275585e-01 -3.68885905e-01
4.83775914e-01 6.05402648e-01 -5.98300099e-01 -9.70499665e-02
-8.23226988e-01 3.06490719e-01 3.35032851e-01 -3.52694899e-01
-9.98433173e-01 -7.27328777e-01 -1.49465501e-01 4.41250414e-01
5.63568234e-01 -1.08117640e+00 1.40571499e+00 -1.03737807e+00
-1.45454443e+00 7.10584283e-01 -1.12469755e-02 -2.41119683e-01
7.66267061e-01 6.78722467e-03 -4.12022881e-02 -1.95014719e-02
-2.29636669e-01 3.95707116e-02 7.26626694e-01 -1.35142255e+00
-8.47152293e-01 -6.42420352e-01 5.09271264e-01 2.07968667e-01
-1.01905942e+00 -2.09331408e-01 -7.06271470e-01 -9.52859595e-02
-2.11466439e-02 -7.45555878e-01 -3.41323227e-01 1.06839135e-01
-2.63393134e-01 -2.18771949e-01 5.18177211e-01 -2.66931176e-01
9.70612764e-01 -2.40891910e+00 2.45130450e-01 4.92479116e-01
5.29050946e-01 5.30349351e-02 -1.96847245e-01 5.93656063e-01
1.95644170e-01 3.23685735e-01 1.41788945e-01 -8.37029099e-01
3.18034530e-01 7.25922406e-01 -2.08305240e-01 3.83767813e-01
-5.40903807e-01 8.18217456e-01 -6.61983192e-01 -2.76955634e-01
-4.71548915e-01 2.21691608e-01 -8.09328735e-01 5.03783107e-01
-4.33349937e-01 3.87883842e-01 -9.82803404e-01 2.13804752e-01
7.22226501e-01 -6.13830805e-01 8.42189610e-01 5.68941236e-03
1.61416963e-01 2.50531763e-01 -1.48798120e+00 1.47958469e+00
-7.35799968e-01 -1.46115258e-01 9.10517514e-01 -6.66518509e-01
5.47582805e-01 4.66773957e-01 5.89778066e-01 -3.02592456e-01
-1.51101679e-01 2.18072936e-01 -7.82677233e-01 -4.55113143e-01
3.25654209e-01 7.55119696e-03 2.74568230e-01 1.15827620e+00
1.44396067e-01 8.76538217e-01 -2.76732475e-01 2.45700464e-01
9.81313586e-01 -2.64386058e-01 -1.64942816e-01 -2.00255603e-01
1.87992364e-01 -5.24233282e-01 7.99547791e-01 9.23548222e-01
-4.08259174e-03 2.45076999e-01 7.49469638e-01 -5.04797697e-01
-8.24090421e-01 -6.78900063e-01 2.69671351e-01 1.63254201e+00
-7.26258457e-02 -6.06642008e-01 -8.72488558e-01 -1.02922070e+00
3.09166074e-01 2.70917207e-01 -6.59788549e-01 8.86611342e-02
-3.97714466e-01 -5.65511346e-01 2.88024366e-01 1.74634784e-01
3.64529878e-01 -6.93115413e-01 -4.39037234e-01 2.72627503e-01
-1.88835755e-01 -8.76572549e-01 -7.56649137e-01 4.19901282e-01
-1.05667078e+00 -1.02124488e+00 -5.29074728e-01 -3.32316488e-01
6.42082095e-01 5.51162064e-01 9.38481212e-01 2.51175135e-01
2.29817837e-01 5.26269436e-01 -2.33104274e-01 2.30596468e-01
-1.33008361e-01 4.67265993e-01 -6.71493188e-02 5.76144814e-01
-8.27587768e-02 -1.02497578e+00 -7.31322885e-01 4.21342999e-01
-9.99202371e-01 -1.89323202e-01 3.62859517e-01 8.09476495e-01
3.43851984e-01 5.36586270e-02 5.44791460e-01 -1.38238978e+00
5.60218930e-01 -8.97055626e-01 -3.53643984e-01 6.81903064e-01
-7.62894571e-01 1.94285110e-01 9.83095109e-01 -4.00495112e-01
-1.10046756e+00 -1.08282410e-01 2.77718306e-01 -6.27082765e-01
-3.68758291e-02 2.01519430e-01 -4.79621023e-01 -2.76315153e-01
7.54239440e-01 9.52215791e-02 1.58731237e-01 -1.09833097e+00
6.16181910e-01 8.57119441e-01 2.97369212e-01 -1.35127294e+00
5.60499072e-01 4.75474805e-01 -4.34709966e-01 -2.76613712e-01
-7.12911248e-01 -1.43916309e-01 -2.27716193e-01 3.36910971e-02
1.67367578e-01 -7.01012969e-01 -1.08727205e+00 3.91508728e-01
-7.03297615e-01 -6.02850080e-01 -6.19884968e-01 7.50471652e-02
-5.39157331e-01 5.04768193e-01 -5.47727764e-01 -7.32764423e-01
-5.39191484e-01 -1.02939832e+00 5.31628311e-01 3.64607602e-01
2.34197527e-01 -1.16983128e+00 8.39124769e-02 4.83581662e-01
7.46568024e-01 -4.33149710e-02 9.55896616e-01 -1.00835502e+00
-7.01931059e-01 -1.03698291e-01 7.72656277e-02 3.05274248e-01
1.41040057e-01 -2.96362400e-01 -1.15444934e+00 -6.67886794e-01
-3.74024436e-02 -4.93710548e-01 5.11395335e-01 -2.13521108e-01
1.32301688e+00 -1.28302014e+00 -4.09022510e-01 1.15439773e+00
1.39302993e+00 -3.08496475e-01 1.59963280e-01 3.53212714e-01
4.28626299e-01 6.17519796e-01 3.37816067e-02 9.75447536e-01
5.25537908e-01 6.72002554e-01 4.95810241e-01 1.91778883e-01
4.10016418e-01 -3.68770421e-01 1.18150912e-01 2.78060704e-01
4.06947881e-02 -9.73427743e-02 -5.02163708e-01 4.83147740e-01
-2.22415709e+00 -6.81353986e-01 4.50942755e-01 2.49655461e+00
9.88474011e-01 -3.47221375e-01 4.43696976e-01 -2.93803692e-01
6.32667601e-01 8.28609839e-02 -7.80541241e-01 -3.15718651e-01
-1.10718742e-01 3.80672850e-02 6.57658219e-01 3.48388165e-01
-7.99531102e-01 6.38733149e-01 6.34460258e+00 5.72401047e-01
-1.02017701e+00 4.50944334e-01 6.48302197e-01 -3.73535037e-01
-7.56394863e-01 1.05016969e-01 -6.40828669e-01 6.10154867e-01
1.08605218e+00 -4.80581403e-01 1.06178141e+00 1.25795019e+00
-1.10319585e-01 3.75216872e-01 -1.36934900e+00 6.42634749e-01
-4.64709848e-01 -1.32103324e+00 -5.62794469e-02 3.32448184e-01
8.53014231e-01 1.78345546e-01 1.23058170e-01 8.38639960e-02
6.90517306e-01 -7.29846120e-01 5.21204352e-01 3.99749845e-01
3.86577666e-01 -9.53976154e-01 3.45845252e-01 7.19591498e-01
-5.62700450e-01 -5.88517249e-01 -3.90151739e-01 4.55036223e-01
-2.34839678e-01 5.30714989e-01 -4.17291850e-01 7.96656847e-01
7.02323377e-01 6.87549040e-02 -3.01837772e-01 9.80672777e-01
3.19760203e-01 3.66495132e-01 -5.89361906e-01 3.33572209e-01
1.50402650e-01 -2.97495365e-01 1.86945111e-01 9.53842223e-01
1.30725443e-01 5.40881790e-02 1.98953584e-01 6.02687955e-01
-4.25270468e-01 3.03144544e-01 -2.01298073e-01 -2.77708806e-02
8.02764833e-01 1.46773672e+00 1.66752085e-01 -2.74770945e-01
-2.00459629e-01 8.80285144e-01 9.82099354e-01 5.74203730e-01
-4.22141910e-01 7.65091879e-03 1.15755475e+00 -2.47116555e-02
3.85339200e-01 1.29651949e-01 -2.10549772e-01 -1.43327606e+00
2.15108022e-01 -1.09735012e+00 9.07099485e-01 -1.87141821e-01
-1.59274638e+00 5.57788610e-01 -3.59721780e-01 -6.03556335e-01
-1.05727732e-01 -1.89984575e-01 -6.63662672e-01 8.50139260e-01
-1.51310122e+00 -1.12600207e+00 -1.64721280e-01 1.02141988e+00
-2.37422407e-01 -8.81267413e-02 9.75072563e-01 3.77673596e-01
-8.69083345e-01 1.22309780e+00 4.96546060e-01 -3.25156003e-01
7.03591287e-01 -9.14862812e-01 -7.21518919e-02 4.26271796e-01
-4.51887250e-02 9.28875089e-01 4.84883964e-01 -1.63430169e-01
-1.70942760e+00 -1.18253887e+00 9.04585481e-01 -3.97562683e-02
4.99894470e-01 -1.50678992e-01 -1.09207797e+00 1.13333929e+00
-1.07600115e-01 1.23511404e-01 7.09128201e-01 4.21657413e-01
-7.59135902e-01 -6.08589649e-01 -1.45466888e+00 3.95085186e-01
8.89327109e-01 -4.85891789e-01 1.51552349e-01 9.23232973e-01
7.39727080e-01 -2.38694325e-01 -1.18328094e+00 -1.98281169e-01
3.27281445e-01 -1.11639106e+00 7.53651798e-01 -1.04604220e+00
-1.13016590e-01 2.58343697e-01 -2.75498182e-01 -1.15781772e+00
-4.78062272e-01 -1.24485254e+00 -4.27248359e-01 1.41113186e+00
2.50285298e-01 -1.28564930e+00 9.81859207e-01 1.22264242e+00
1.56698689e-01 -8.09924662e-01 -9.70160306e-01 -6.58427656e-01
2.86254317e-01 9.24573988e-02 1.37050235e+00 9.27497268e-01
1.91529796e-01 -1.58781439e-01 -5.31278491e-01 2.62163728e-01
8.17828059e-01 4.18200552e-01 9.11741674e-01 -1.08326519e+00
-9.82910097e-01 -3.32109451e-01 2.77835697e-01 -1.16540682e+00
1.41701981e-01 -8.16129744e-01 -3.54444921e-01 -1.17804635e+00
5.04659295e-01 -1.05181003e+00 -4.64656770e-01 9.78349030e-01
-3.55318636e-02 -3.66744250e-01 3.20843041e-01 5.53807735e-01
-7.11982906e-01 2.66738027e-01 1.00360715e+00 1.90733030e-01
-2.82217026e-01 3.54474843e-01 -1.23876750e+00 5.60206808e-02
8.37638795e-01 -4.55408603e-01 -3.73418927e-01 -5.60265541e-01
2.21444756e-01 3.15710276e-01 2.16175303e-01 -4.64739621e-01
5.38013458e-01 -6.16551518e-01 -1.37417719e-01 2.02117622e-01
1.49234205e-01 -9.07375813e-01 5.25288641e-01 4.32379842e-02
-5.17620146e-01 -3.91230524e-01 -4.56628889e-01 6.27973616e-01
2.55411297e-01 -1.51804432e-01 9.25146282e-01 -2.30601162e-01
1.03422277e-01 8.52561057e-01 2.82467961e-01 -4.56379494e-03
1.03957677e+00 2.08037168e-01 -5.81245124e-01 -5.80768049e-01
-6.87029779e-01 5.71552813e-01 7.96009481e-01 9.80260074e-02
-8.18005726e-02 -1.17434680e+00 -4.36255157e-01 2.25329995e-01
-2.97259912e-02 6.14938773e-02 6.70383930e-01 8.11575949e-01
4.06721011e-02 2.28911608e-01 -3.76363657e-03 -2.46535629e-01
-1.01525116e+00 6.84942722e-01 7.17235684e-01 -2.03621447e-01
-8.69537413e-01 6.60750329e-01 2.74405450e-01 -6.85992777e-01
7.28979409e-01 1.62280783e-01 1.52568415e-01 4.04612198e-02
5.52979827e-01 3.41847748e-01 1.80795476e-01 -1.30699039e-01
-1.33734152e-01 1.56833738e-01 -2.42095158e-01 5.00797480e-02
1.66863549e+00 -3.51828277e-01 -2.32366711e-01 1.43631235e-01
1.13233984e+00 -2.11012457e-02 -1.66703761e+00 -7.19751060e-01
-4.03353512e-01 -6.05753601e-01 -1.08742759e-01 -8.47757459e-01
-1.62403607e+00 4.59348261e-01 1.72170341e-01 4.47787941e-01
9.42405641e-01 1.32288530e-01 6.75640762e-01 5.56684434e-01
6.04845583e-01 -9.62230682e-01 -2.71738529e-01 1.72000006e-01
3.48991215e-01 -7.44210303e-01 -4.53908205e-01 4.08214442e-02
-7.08655298e-01 1.09287035e+00 5.42974949e-01 1.70714825e-01
4.04362738e-01 1.13368668e-01 -7.29594901e-02 4.98766862e-02
-1.33609581e+00 2.49531612e-01 -3.29540402e-01 5.09001851e-01
-1.07170813e-01 2.94944793e-01 4.43724133e-02 1.03527927e+00
6.75894842e-02 -1.03576563e-01 2.61305988e-01 7.99180984e-01
-4.16688591e-01 -1.48891199e+00 -3.19336683e-01 2.44245812e-01
-5.64287305e-01 3.76961380e-01 -1.50157809e-01 4.11221623e-01
-5.90112573e-03 8.69617462e-01 -1.71406373e-01 -2.92876959e-01
2.00553998e-01 4.13723916e-01 2.30310887e-01 -3.10823500e-01
-1.02866542e+00 -1.29801765e-01 6.23398237e-02 -7.61111081e-01
4.86815609e-02 -3.61620128e-01 -8.87688398e-01 -8.39393973e-01
-3.70218828e-02 5.72501600e-01 7.18163311e-01 8.26195121e-01
8.24529409e-01 1.05919860e-01 1.39363432e+00 -5.03108680e-01
-1.21461606e+00 -5.06445050e-01 -9.32065189e-01 3.58118206e-01
4.50581789e-01 -1.46367550e-01 -7.24671245e-01 -2.64946222e-01]
|
[5.827064037322998, 6.2793450355529785]
|
2e324bc2-262a-46c5-80f7-371862caa3ed
|
empirical-interpretation-of-speech-emotion
| null | null |
http://www.interspeech2020.org/index.php?m=content&c=index&a=show&catid=350&id=1146
|
http://www.interspeech2020.org/uploadfile/pdf/Thu-2-2-8.pdf
|
Empirical Interpretation of Speech Emotion Perception with Attention Based Model for Speech Emotion Recognition
|
Speech emotion recognition is essential for obtaining emotional intelligence which affects the understanding of context and meaning of speech. Harmonically structured vowel and consonant sounds add indexical and linguistic cues in spoken in- formation. Previous research argued whether vowel sound cues were more important in carrying the emotional context from a psychological and linguistic point of view. Other research also claimed that emotion information could exist in small over- lapping acoustic cues. However, these claims are not corroborated in computational speech emotion recognition systems. In this research, a convolution-based model and a long-short- term memory-based model, both using attention, are applied to investigate these theories of speech emotion on computational models. The role of acoustic context and word importance is demonstrated for the task of speech emotion recognition. The IEMOCAP corpus is evaluated by the proposed models, and 80.1% unweighted accuracy is achieved on pure acoustic data which is higher than current state-of-the-art models on this task. The phones and words are mapped to the attention vectors and it is seen that the vowel sounds are more important for defining emotion acoustic cues than the consonants, and the model can assign word importance based on acoustic context
|
['Thomas Hain Speech', 'Rosanna Milner', 'Md AsifJalal']
|
2020-10-28
| null | null | null |
interspeech-2020-10
|
['emotional-intelligence']
|
['natural-language-processing']
|
[-1.09853581e-01 1.15807407e-01 2.11675152e-01 -6.37105405e-01
-1.56163707e-01 -7.35161901e-02 1.63869202e-01 1.65868446e-01
-7.03059435e-01 3.12861294e-01 4.89592433e-01 -1.32741362e-01
-9.74334627e-02 -3.97576630e-01 -2.67260343e-01 -4.72700953e-01
1.30762577e-01 -1.68663844e-01 -3.66439372e-01 -3.90519142e-01
5.28213263e-01 8.17831904e-02 -2.10136962e+00 4.29242730e-01
7.92011082e-01 1.15951025e+00 5.89490294e-01 8.00567985e-01
-6.62603796e-01 7.05817044e-01 -8.31711709e-01 -4.70081687e-01
-6.07726932e-01 -4.01710153e-01 -7.23059118e-01 -2.25458235e-01
2.91738305e-02 1.45258814e-01 8.96243155e-02 1.03376842e+00
8.11302722e-01 5.38792312e-01 5.59728980e-01 -9.90286589e-01
-1.21775448e+00 7.29885101e-01 3.53299320e-01 4.88873452e-01
4.23365951e-01 -3.78187299e-01 9.85370219e-01 -1.28469121e+00
1.04598664e-01 1.31192207e+00 5.15654504e-01 7.67781794e-01
-5.83977759e-01 -6.54006541e-01 3.53443384e-01 7.37141550e-01
-1.15468860e+00 -5.29377341e-01 1.04669785e+00 -4.84330237e-01
1.61755002e+00 5.02784252e-01 9.11998153e-01 1.09983921e+00
1.58115357e-01 3.96267891e-01 1.10869062e+00 -6.12380803e-01
1.71601057e-01 6.13307059e-01 4.93977427e-01 1.87843770e-01
-4.74657834e-01 2.88272232e-01 -9.03110623e-01 1.56151548e-01
1.29369915e-01 -3.73446465e-01 -2.00916395e-01 7.00606585e-01
-6.76349342e-01 9.73190725e-01 1.65665790e-01 9.15860713e-01
-6.31471753e-01 4.50426042e-02 5.98210394e-01 3.28876525e-01
5.90443909e-01 3.59412163e-01 -5.28445661e-01 -6.05704010e-01
-4.77861017e-01 -4.60443795e-01 5.80683351e-01 1.75330505e-01
3.29064399e-01 6.39647186e-01 -5.67765869e-02 1.31439030e+00
6.84352338e-01 5.23561418e-01 8.60992610e-01 -6.99601948e-01
-6.52233511e-02 1.33624256e-01 -4.82875913e-01 -1.14174223e+00
-3.49419385e-01 -2.76269436e-01 -3.41454118e-01 3.19175012e-02
-1.62767828e-01 -1.13470599e-01 -5.74270844e-01 2.02325153e+00
6.19812533e-02 2.80122340e-01 5.32625854e-01 8.71932805e-01
1.34303010e+00 8.88713837e-01 4.04409826e-01 -6.45533681e-01
1.68868411e+00 -7.74535954e-01 -1.61647809e+00 -5.12932539e-01
3.63731831e-01 -8.54399741e-01 1.40311193e+00 4.50505257e-01
-9.72320676e-01 -9.26805794e-01 -1.04391682e+00 -6.59321696e-02
-8.02517414e-01 -9.97778624e-02 7.44357526e-01 1.17229831e+00
-9.44704115e-01 1.65230036e-01 -2.50274301e-01 -2.44237795e-01
2.33514104e-02 3.93492579e-02 -1.58370242e-01 5.50229490e-01
-1.72645652e+00 1.17751539e+00 2.16439694e-01 2.42466375e-01
-4.98338252e-01 -5.63766599e-01 -1.01797998e+00 1.76418334e-01
-2.89597809e-01 -1.15982972e-01 1.03014731e+00 -1.50294363e+00
-1.63024473e+00 7.53874719e-01 -4.70968723e-01 -1.53061196e-01
-6.47689283e-01 -1.24925144e-01 -1.00764155e+00 3.85113835e-01
-3.73373061e-01 4.34655041e-01 7.40264893e-01 -1.03557587e+00
-4.46956426e-01 -3.33728224e-01 -2.44866759e-01 4.13239628e-01
-8.91577542e-01 5.92916906e-01 9.99943540e-02 -6.42498434e-01
2.97396611e-02 -4.33099538e-01 2.18085364e-01 -5.07267475e-01
2.79546499e-01 -7.49354780e-01 6.46269500e-01 -9.03606951e-01
1.58990312e+00 -2.30466294e+00 -2.25368068e-01 7.26542249e-03
-2.14535952e-01 2.63093919e-01 3.87254581e-02 2.51259208e-01
-3.18026245e-01 3.51125181e-01 -2.18824092e-02 -2.39105642e-01
3.77641141e-01 3.86288464e-01 -2.15696439e-01 4.60981876e-02
2.67288476e-01 7.49352217e-01 -6.30052269e-01 -3.45806152e-01
2.45866910e-01 1.04649818e+00 -4.67018187e-01 2.38759756e-01
2.91934550e-01 9.93867684e-03 9.18179750e-02 4.11842346e-01
3.47678095e-01 6.44633532e-01 4.53274697e-03 -1.95666701e-02
-2.65878588e-01 6.38890326e-01 -9.75013256e-01 1.38091397e+00
-5.86412728e-01 1.03670526e+00 1.67186409e-01 -1.07431173e+00
1.23250902e+00 8.99166226e-01 -1.44708395e-01 -9.00685191e-01
3.94398481e-01 8.12047273e-02 5.07622898e-01 -1.08718729e+00
5.84680974e-01 -6.48406446e-01 -6.53659832e-03 1.39281809e-01
9.65780020e-02 -4.37289387e-01 -3.55362982e-01 -2.52071619e-01
2.73236275e-01 -4.10595179e-01 1.47781089e-01 -2.61188626e-01
6.18886530e-01 -7.94076681e-01 4.87340391e-01 2.22394705e-01
-5.35413921e-01 1.37301177e-01 1.96821496e-01 -4.16568555e-02
-4.14016604e-01 -6.48836553e-01 -4.44803923e-01 1.68251073e+00
-2.02474222e-01 -1.94604814e-01 -9.86667037e-01 -1.86143354e-01
-4.21217591e-01 1.36039901e+00 -6.20976031e-01 -4.25350279e-01
-3.36893536e-02 -4.68687892e-01 6.92867458e-01 5.06586373e-01
-1.14776760e-01 -1.75197613e+00 -6.44550204e-01 2.43030831e-01
-3.08508009e-01 -9.21935678e-01 -2.02969015e-01 3.83868426e-01
-2.06365302e-01 -4.99110341e-01 -1.62049592e-01 -1.15658915e+00
1.67703599e-01 -2.44612589e-01 9.76157606e-01 1.25761271e-01
-2.67539680e-01 6.93592131e-01 -4.27156895e-01 -1.06422269e+00
-4.12769198e-01 -5.26194334e-01 2.46436536e-01 6.07878789e-02
9.03024971e-01 -4.59115326e-01 -2.50388086e-01 -8.29934031e-02
-7.07476616e-01 -5.43601036e-01 2.28861764e-01 7.27703452e-01
3.92366230e-01 -1.60604026e-02 1.16851199e+00 -1.18955463e-01
1.12636018e+00 -5.40853918e-01 3.53264838e-01 6.05571130e-03
-4.49303031e-01 -4.07609820e-01 1.84783354e-01 -6.22323632e-01
-1.25499356e+00 -5.27970433e-01 -7.50011742e-01 -2.47193068e-01
-6.08301520e-01 6.72164798e-01 -2.41725430e-01 8.44417065e-02
3.40305716e-01 2.31306359e-01 1.13859110e-01 5.78604192e-02
1.26826555e-01 1.15155411e+00 2.16057271e-01 -6.11696959e-01
-2.85883754e-01 -2.02908784e-01 -6.83925211e-01 -1.40254295e+00
-4.92961824e-01 -1.86993539e-01 -2.19809145e-01 -6.20214403e-01
1.25786734e+00 -6.13790333e-01 -7.99812019e-01 3.66274714e-01
-1.23460662e+00 -2.76014879e-02 -2.13387176e-01 9.28556800e-01
-4.11185741e-01 2.52406806e-01 -7.21769989e-01 -1.57760704e+00
-4.91288930e-01 -1.13174140e+00 5.63702345e-01 2.49146119e-01
-4.44182158e-01 -1.11683202e+00 -2.15848669e-01 3.03245425e-01
6.20975494e-01 -3.81111056e-01 9.52002108e-01 -9.60445404e-01
5.24712026e-01 4.66845259e-02 2.44269550e-01 8.11226428e-01
-1.11539587e-01 3.95608246e-02 -1.58491576e+00 5.07626295e-01
5.69623649e-01 -5.03662527e-01 6.66687787e-01 4.46404815e-01
1.06260967e+00 -1.07593648e-01 4.10462856e-01 1.35968387e-01
9.95826483e-01 9.10030305e-01 7.11311221e-01 -1.42914936e-01
2.39013076e-01 1.10067344e+00 6.99977338e-01 3.64517480e-01
5.24426997e-01 2.71666080e-01 3.54327738e-01 1.56800240e-01
-1.14597902e-02 1.99510589e-01 5.80702662e-01 1.76677275e+00
1.69990927e-01 -6.47616610e-02 -9.03015256e-01 7.78195679e-01
-1.37401605e+00 -1.14158773e+00 -1.68963671e-01 1.82486606e+00
8.63490522e-01 -1.35500878e-02 -3.90151024e-01 4.28090632e-01
6.59544766e-01 1.39767513e-01 -2.26008981e-01 -1.45891511e+00
-2.31387213e-01 4.56537217e-01 -6.97865188e-01 8.11325252e-01
-7.88242877e-01 9.45389748e-01 6.47444439e+00 9.09716964e-01
-1.25393236e+00 3.19791734e-01 5.94724417e-01 -9.83764753e-02
-6.32425427e-01 -5.59335530e-01 -3.79157394e-01 4.31563824e-01
1.46622920e+00 -8.45546946e-02 3.76344830e-01 7.73286879e-01
2.65061647e-01 1.91247556e-02 -7.75581658e-01 1.09437704e+00
5.23159623e-01 -6.11884773e-01 -2.06561387e-01 -2.56368935e-01
6.21291026e-02 -4.19860214e-01 3.15889239e-01 6.03843391e-01
-5.56023121e-01 -1.27430832e+00 8.08737457e-01 8.19279134e-01
5.25455058e-01 -1.09927022e+00 8.65484774e-01 1.00121409e-01
-1.01993656e+00 -6.87417854e-03 -4.18885559e-01 -6.79360986e-01
1.59669563e-01 2.33823299e-01 -6.82447135e-01 -8.43355581e-02
7.48786569e-01 8.97742212e-02 -7.08253086e-02 2.75107086e-01
-2.75770575e-01 1.09222341e+00 -5.89963682e-02 -6.75865471e-01
4.96159308e-02 -1.03601307e-01 4.02018130e-01 1.54285526e+00
4.27059680e-01 3.64852101e-01 -2.36942247e-01 8.28513026e-01
3.66822630e-01 6.78019106e-01 -6.27419055e-01 -3.90919447e-01
7.31688201e-01 1.12986541e+00 -5.21902442e-01 -3.03461552e-01
-5.50970614e-01 7.49541998e-01 1.62023529e-01 3.63657355e-01
-9.32073772e-01 -5.74471056e-01 1.17333984e+00 -5.57577968e-01
1.52960315e-01 -3.02796275e-03 -2.61930108e-01 -4.32585001e-01
-2.37754524e-01 -6.87697768e-01 -5.82841830e-03 -1.08472407e+00
-1.29275298e+00 7.98413634e-01 -4.64067847e-01 -3.40116024e-01
-2.37457633e-01 -7.58992314e-01 -6.43289089e-01 1.10133731e+00
-1.43046892e+00 -7.50160515e-01 2.26217248e-02 5.19717693e-01
8.87713850e-01 -2.30021462e-01 1.37765884e+00 2.91455179e-01
-4.45174456e-01 6.21089399e-01 -4.83683705e-01 -1.13167733e-01
5.85498750e-01 -1.10465908e+00 -3.08077842e-01 6.31213069e-01
2.20782399e-01 6.62211895e-01 6.45130754e-01 -4.13590282e-01
-1.04669082e+00 -4.58227962e-01 1.46087658e+00 -2.70026803e-01
5.51435769e-01 -2.30841413e-01 -1.16508067e+00 1.48383662e-01
8.26310933e-01 -3.53225648e-01 1.59794891e+00 2.04774767e-01
-2.07070649e-01 1.43329442e-01 -1.10903549e+00 4.39239889e-01
6.72204196e-01 -1.05857301e+00 -1.22530842e+00 -1.28637582e-01
1.27577579e+00 1.89955458e-01 -9.59537983e-01 1.90594032e-01
5.69520593e-01 -1.01934671e+00 8.91608119e-01 -6.01013720e-01
4.36796427e-01 1.64461285e-01 -7.39910543e-01 -1.39589250e+00
-2.20629543e-01 -1.01580188e-01 1.17782829e-02 1.49498284e+00
5.51582098e-01 -7.40074933e-01 -8.14769883e-03 6.09919310e-01
-5.99083245e-01 -7.12005973e-01 -1.18342566e+00 -3.36150020e-01
9.93021727e-02 -1.39325333e+00 5.35465300e-01 1.24941599e+00
4.43932563e-01 4.77071881e-01 5.96722402e-03 2.12591335e-01
-1.69135496e-01 -5.22329628e-01 -4.65519764e-02 -1.03544891e+00
8.90583843e-02 -8.05562973e-01 -4.37392563e-01 -2.21551508e-01
6.25721037e-01 -8.46560538e-01 1.51118070e-01 -1.41089237e+00
-2.87405998e-01 8.07359219e-02 -7.17881203e-01 2.79358923e-01
-4.36368436e-01 -2.51726620e-03 2.77284890e-01 -6.56433105e-01
-1.49746269e-01 8.58547568e-01 7.82489061e-01 1.27219439e-01
-9.97955129e-02 -4.26629514e-01 -8.22054863e-01 9.70936179e-01
8.97654533e-01 -1.02344878e-01 -5.63135386e-01 -1.38276473e-01
2.55019754e-01 2.95872390e-02 8.54795352e-02 -8.76717269e-01
1.98836789e-01 -4.58641835e-02 2.21971095e-01 -3.02566200e-01
8.48647058e-01 -8.91599357e-01 -2.01246142e-01 2.15719506e-01
-4.88416076e-01 -1.15080259e-03 5.64857781e-01 8.29942599e-02
-4.81999993e-01 -6.53638661e-01 4.78962481e-01 9.24974680e-02
-1.01116335e+00 -2.48853624e-01 -8.86823952e-01 1.06160371e-02
8.11569035e-01 -2.72453725e-01 -4.91870344e-02 -5.02352178e-01
-1.03385329e+00 -3.18124950e-01 -5.16264856e-01 8.79063427e-01
1.04609108e+00 -1.43472600e+00 -5.25021017e-01 4.99457300e-01
-2.01536752e-02 -8.78340006e-01 5.62876463e-01 7.73693383e-01
1.40285995e-02 3.64845186e-01 2.94269379e-02 -2.36258179e-01
-1.53956234e+00 6.03474319e-01 3.55129987e-01 5.54814935e-01
7.30590522e-02 1.30332804e+00 1.89891487e-01 -4.58192855e-01
5.43475032e-01 -3.70622993e-01 -8.65434349e-01 7.32168138e-01
8.13888550e-01 2.28423551e-01 -9.21992734e-02 -1.01266885e+00
-4.69824880e-01 4.92225111e-01 4.09933478e-01 -3.86648893e-01
1.10142505e+00 -3.40169966e-01 -3.16031456e-01 1.16502643e+00
1.04673219e+00 1.93262786e-01 -4.09436733e-01 1.23691082e-01
-4.66685370e-02 -1.13088205e-01 2.35594839e-01 -1.01001108e+00
-7.87433326e-01 1.37562716e+00 8.84584963e-01 5.73560059e-01
1.18540573e+00 -6.45185122e-03 5.19165456e-01 2.07469717e-01
-1.48568004e-01 -1.68862033e+00 2.13194415e-01 9.49466884e-01
1.24518716e+00 -1.09961689e+00 -8.49367619e-01 -2.61865169e-01
-1.08177042e+00 1.07210553e+00 7.76563168e-01 2.72204041e-01
9.31783915e-01 4.93199944e-01 4.50660169e-01 -1.50632113e-01
-9.65088427e-01 -4.86789167e-01 4.99211460e-01 7.59450734e-01
9.55748975e-01 3.31881016e-01 -6.43565297e-01 1.61595583e+00
-6.46658301e-01 -6.03080451e-01 9.66408551e-02 4.78728056e-01
-7.99186826e-01 -6.35321438e-01 -5.99577129e-01 1.37231290e-01
-8.01159918e-01 -5.31816781e-01 -4.49152946e-01 3.56634736e-01
4.73014235e-01 1.55439734e+00 4.23083454e-01 -4.76484358e-01
3.04037631e-01 8.89031827e-01 4.01001684e-02 -4.81604427e-01
-8.99674892e-01 2.10087195e-01 3.15213978e-01 -3.20793539e-01
-5.73020816e-01 -5.00758410e-01 -1.71078849e+00 2.13054836e-01
-3.94538730e-01 4.51639473e-01 1.28616571e+00 1.07588160e+00
3.18595707e-01 1.05408597e+00 3.90313059e-01 -6.36733174e-01
1.79847181e-02 -1.28225362e+00 -4.94886905e-01 2.32765734e-01
8.30995739e-02 -5.14493585e-01 -6.79126680e-01 -1.27792820e-01]
|
[13.768638610839844, 5.789379596710205]
|
6dbe1494-b644-4cb5-8b95-10e53263c355
|
xnap-making-lstm-based-next-activity
|
2008.07993
| null |
https://arxiv.org/abs/2008.07993v3
|
https://arxiv.org/pdf/2008.07993v3.pdf
|
XNAP: Making LSTM-based Next Activity Predictions Explainable by Using LRP
|
Predictive business process monitoring (PBPM) is a class of techniques designed to predict behaviour, such as next activities, in running traces. PBPM techniques aim to improve process performance by providing predictions to process analysts, supporting them in their decision making. However, the PBPM techniques` limited predictive quality was considered as the essential obstacle for establishing such techniques in practice. With the use of deep neural networks (DNNs), the techniques` predictive quality could be improved for tasks like the next activity prediction. While DNNs achieve a promising predictive quality, they still lack comprehensibility due to their hierarchical approach of learning representations. Nevertheless, process analysts need to comprehend the cause of a prediction to identify intervention mechanisms that might affect the decision making to secure process performance. In this paper, we propose XNAP, the first explainable, DNN-based PBPM technique for the next activity prediction. XNAP integrates a layer-wise relevance propagation method from the field of explainable artificial intelligence to make predictions of a long short-term memory DNN explainable by providing relevance values for activities. We show the benefit of our approach through two real-life event logs.
|
['Jörg Becker', 'Martin Matzner', 'Sven Weinzierl', 'Jens Brunk', 'Kate Revoredo', 'Sandra Zilker']
|
2020-08-18
| null | null | null | null |
['activity-prediction', 'activity-prediction']
|
['computer-vision', 'time-series']
|
[ 4.14101422e-01 5.78585386e-01 -2.07506627e-01 -5.12943804e-01
-1.59715042e-02 2.39433974e-01 7.99615085e-01 5.60116768e-01
1.95578247e-01 5.90840638e-01 3.69758815e-01 -5.20575881e-01
-8.23016524e-01 -1.12558532e+00 -6.33780658e-01 -2.38256574e-01
-1.68724805e-01 7.04774439e-01 5.53921275e-02 1.13794133e-01
4.37871873e-01 6.02549851e-01 -1.41110265e+00 8.39981973e-01
6.43851280e-01 1.21831989e+00 7.56001174e-02 5.58167934e-01
-6.04835808e-01 1.89229929e+00 -3.79400194e-01 -3.13051641e-01
-1.64180547e-02 -3.86735380e-01 -9.36187685e-01 -1.32836953e-01
-3.40511292e-01 -1.10707618e-01 -1.82710573e-01 6.55138075e-01
-3.15578610e-01 1.82749316e-01 5.33326328e-01 -1.47905350e+00
-7.31633306e-01 1.17825902e+00 -1.11175947e-01 3.72313619e-01
3.74940842e-01 2.54145354e-01 1.07949018e+00 -4.98594582e-01
3.06798220e-01 1.22793996e+00 7.89984167e-01 6.72219694e-01
-1.17545354e+00 -2.72465080e-01 3.83471847e-01 6.19601905e-01
-7.01471090e-01 -5.75267486e-02 6.70302689e-01 -4.50711399e-01
1.24146914e+00 3.05637062e-01 8.24563563e-01 1.17091882e+00
6.28824294e-01 8.06913018e-01 8.53524208e-01 -4.28338706e-01
5.71584582e-01 -1.24217376e-01 3.65434319e-01 4.20861006e-01
2.96729833e-01 8.91752541e-02 -8.88437569e-01 -3.21916267e-02
7.48642504e-01 8.58349025e-01 7.38439634e-02 1.16016857e-01
-1.06178892e+00 4.69068795e-01 4.63596582e-01 6.11161351e-01
-9.93834376e-01 4.55320090e-01 1.77991390e-01 3.09992015e-01
2.81801969e-01 9.49562550e-01 -5.98559737e-01 -5.08040249e-01
-8.22551608e-01 2.35314876e-01 1.23361957e+00 6.69610679e-01
5.67616105e-01 1.08263977e-01 -6.54807210e-01 2.44052969e-02
4.97692257e-01 -3.55247259e-02 4.44917351e-01 -1.08391917e+00
3.87667179e-01 1.12954557e+00 9.67323035e-02 -8.91873896e-01
-5.08105338e-01 -2.53696322e-01 -8.98271203e-01 4.48449291e-02
3.53902310e-01 1.63956195e-01 -6.51871622e-01 1.17405510e+00
-3.15043747e-01 3.67932141e-01 -4.67093289e-02 6.10447168e-01
2.93926984e-01 9.07064974e-01 4.29775089e-01 -1.91685438e-01
1.25213897e+00 -1.14446092e+00 -9.61193621e-01 -3.18311244e-01
2.27063626e-01 -7.52385287e-03 7.56145179e-01 6.80558920e-01
-1.01233912e+00 -7.50977278e-01 -7.77961254e-01 4.11550492e-01
-8.76177624e-02 -5.32007456e-01 9.31167603e-01 2.68707961e-01
-8.28050852e-01 1.49127066e+00 -1.18047595e+00 -2.45335668e-01
6.50773525e-01 4.70714152e-01 -1.25427023e-01 6.52783811e-02
-1.19375443e+00 8.76171887e-01 4.74846900e-01 2.97708184e-01
-1.05523038e+00 -5.67634523e-01 -4.19805080e-01 9.02025402e-01
4.45819736e-01 -6.99464023e-01 1.42013705e+00 -1.01819372e+00
-1.22080696e+00 2.72730738e-02 -3.43349606e-01 -1.16271925e+00
4.69359934e-01 -4.06309068e-01 -5.49819529e-01 -1.22151233e-01
-2.14311287e-01 1.59366503e-01 6.74892485e-01 -8.49174678e-01
-8.33056867e-01 -3.18405867e-01 -2.21271053e-01 -1.71118066e-01
-2.32032761e-01 -1.42740324e-01 -1.98031999e-02 -6.26320615e-02
1.67283282e-01 -4.88416553e-01 -7.05989361e-01 -4.47318435e-01
-4.84962553e-01 -4.84081984e-01 5.86317480e-01 -7.04574347e-01
1.21811879e+00 -1.80287409e+00 -2.13305637e-01 3.48375887e-01
7.46730626e-01 8.16214904e-02 2.98821121e-01 6.06252253e-01
-1.61022749e-02 4.02999550e-01 1.22173846e-01 -3.48431379e-01
2.73142457e-01 3.40508819e-01 -3.06736678e-01 -1.35374621e-01
5.65767646e-01 1.07094932e+00 -7.28518784e-01 -5.26362769e-02
4.30895150e-01 2.36114919e-01 -9.48004499e-02 5.05050778e-01
-5.60100436e-01 4.91843343e-01 -4.56314743e-01 6.05419695e-01
9.89248306e-02 -4.41305906e-01 2.03460336e-01 4.25647765e-01
1.78845972e-02 3.99392426e-01 -8.15478325e-01 9.39702868e-01
-3.22426677e-01 7.38486290e-01 -4.60312724e-01 -9.86516595e-01
1.24096346e+00 4.57133889e-01 5.66170216e-01 -6.94241762e-01
-2.65795350e-01 -1.10615589e-01 1.26959220e-01 -4.09177661e-01
3.46799046e-01 -3.40087622e-01 1.90291733e-01 6.39464080e-01
-3.49007338e-01 5.40428400e-01 3.97738814e-02 -2.17980504e-01
1.70362878e+00 1.58060133e-01 5.27044952e-01 1.25521630e-01
6.53524935e-01 -1.19622335e-01 7.96019673e-01 9.50279474e-01
-3.62268299e-01 3.42044204e-01 9.13528979e-01 -1.09534085e+00
-9.29529071e-01 -7.47836292e-01 3.02540034e-01 1.12148082e+00
-1.55069172e-01 -4.54132497e-01 -4.92248118e-01 -7.41051853e-01
-4.13302705e-03 1.07191598e+00 -7.57027388e-01 -3.07003170e-01
-5.39891839e-01 -4.27637666e-01 4.26266998e-01 9.73045766e-01
5.32896042e-01 -1.68322504e+00 -5.96274257e-01 7.73743749e-01
9.44975913e-02 -8.90950918e-01 3.47566575e-01 6.62300408e-01
-1.25848722e+00 -1.05991149e+00 1.02509670e-01 -2.32607588e-01
3.01299274e-01 -3.45711917e-01 1.27766883e+00 -3.78270335e-02
2.91439354e-01 2.04946715e-02 -1.50048152e-01 -7.05350161e-01
-7.39691436e-01 6.76800907e-02 -5.85848391e-02 -6.15075529e-02
1.06096470e+00 -7.12007225e-01 -4.35687929e-01 1.41054973e-01
-5.85138559e-01 6.69899210e-02 7.96128511e-01 5.32582223e-01
5.21134377e-01 4.58973616e-01 6.66994691e-01 -9.64205205e-01
9.90980506e-01 -5.59104800e-01 -2.33180523e-01 2.46689945e-01
-1.33660603e+00 2.57559299e-01 7.64358103e-01 -2.13334411e-01
-1.18754148e+00 -2.64439940e-01 -4.74617258e-02 -3.57796103e-01
-5.79486191e-01 7.19649196e-01 -2.03039631e-01 6.24155700e-01
6.32237911e-01 4.14164156e-01 -2.60075986e-01 -4.33567852e-01
1.80104487e-02 2.20057532e-01 2.30183229e-01 -3.54654700e-01
3.42981100e-01 3.99428397e-01 2.58383751e-01 -1.35299250e-01
-7.81988621e-01 -3.49847555e-01 -4.12952453e-01 -3.34846973e-01
7.94935465e-01 -5.55673063e-01 -1.20088017e+00 -2.89700590e-02
-1.17764175e+00 -1.52714536e-01 -7.15247154e-01 2.89252430e-01
-7.19668329e-01 -1.00475766e-01 -1.00970542e+00 -1.05439627e+00
-6.38298273e-01 -6.11933827e-01 6.35719419e-01 2.26338401e-01
-8.86880517e-01 -1.21895504e+00 -9.94957238e-02 4.84999001e-01
5.20714462e-01 1.17125735e-01 1.03666174e+00 -1.42140758e+00
-8.74326527e-01 -4.67464149e-01 -1.19148999e-01 3.85972053e-01
2.14556500e-01 -1.97831258e-01 -9.42210257e-01 3.77826214e-01
3.18751991e-01 4.81089860e-01 5.84192276e-01 4.88021374e-01
1.53046989e+00 -5.53631425e-01 -3.38179827e-01 2.46977046e-01
1.21089232e+00 5.14779568e-01 8.52463841e-01 8.15122962e-01
7.60603726e-01 7.81801164e-01 5.16905308e-01 6.55395746e-01
3.22336495e-01 1.48390234e-01 6.36699975e-01 2.44144276e-01
2.57137626e-01 -4.71165121e-01 4.87723202e-01 4.68156070e-01
-7.67447293e-01 -1.14675798e-01 -1.22255611e+00 1.83559045e-01
-2.33004880e+00 -1.43974864e+00 -5.42763829e-01 1.65733051e+00
2.76258022e-01 6.48046196e-01 -1.64859086e-01 4.61781174e-01
5.90615928e-01 -1.87746674e-01 -5.46155334e-01 -8.35079730e-01
1.43448085e-01 -2.60217930e-03 2.53558725e-01 2.14054927e-01
-9.94326890e-01 5.08296788e-01 5.75959873e+00 3.09503227e-01
-5.74246347e-01 1.32591069e-01 7.76752532e-01 5.00971079e-02
-2.15734765e-01 2.64522862e-02 -9.20807421e-01 3.93713474e-01
1.72480869e+00 -1.97957292e-01 3.14970613e-01 1.18733990e+00
8.38435709e-01 1.61466405e-01 -1.70037925e+00 6.22914672e-01
-3.99435610e-01 -1.66298997e+00 1.31972462e-01 1.89461425e-01
6.96326911e-01 -3.28805983e-01 -3.17128628e-01 5.28512537e-01
4.38406736e-01 -1.49335349e+00 5.41059971e-01 1.21698081e+00
-2.63793111e-01 -9.59973454e-01 1.17085636e+00 4.74062353e-01
-9.99184310e-01 -6.73911572e-01 -4.32030141e-01 -7.97456384e-01
3.72502916e-02 8.57788861e-01 -1.33601344e+00 3.69954348e-01
6.24544263e-01 9.32417929e-01 -4.45830911e-01 7.51211166e-01
-5.54713964e-01 9.26719666e-01 1.90134227e-01 -1.57039508e-01
2.47144774e-01 -3.22810262e-02 2.21510753e-01 1.01877844e+00
3.55012923e-01 -4.30925876e-01 2.89696138e-02 1.35560048e+00
9.67645571e-02 -1.98858216e-01 -6.38251126e-01 -2.56670803e-01
2.73580015e-01 9.88228142e-01 -6.85334444e-01 -3.05430412e-01
-1.67490840e-01 7.82250345e-01 2.09245235e-01 1.65754989e-01
-6.79062665e-01 9.10430476e-02 5.71598649e-01 6.43949151e-01
1.84035495e-01 9.94935632e-02 -9.21338737e-01 -6.48428082e-01
2.38833558e-02 -8.28196347e-01 3.08756173e-01 -6.49082959e-01
-1.52468300e+00 6.28506720e-01 -4.49812204e-01 -9.31642830e-01
-5.22620082e-01 -5.06846964e-01 -8.40231478e-01 1.02666426e+00
-1.28732431e+00 -1.18030155e+00 -3.78246933e-01 3.17324311e-01
6.16845310e-01 -1.55981034e-01 7.78255224e-01 -2.49215558e-01
-5.74271560e-01 -3.41268748e-01 -1.60474509e-01 -8.68526548e-02
1.91883653e-01 -1.55730700e+00 5.28801024e-01 7.56150842e-01
-1.12952536e-03 7.14438498e-01 7.92610466e-01 -8.70241344e-01
-1.20721483e+00 -1.28088439e+00 1.36853600e+00 -6.95915639e-01
7.42902577e-01 1.91042647e-02 -1.24440253e+00 1.07793415e+00
1.04454510e-01 -4.13671553e-01 8.94505799e-01 4.63582426e-01
2.10857093e-01 -3.17901880e-01 -9.95419085e-01 4.76537198e-01
9.51986432e-01 -2.75022268e-01 -1.02099514e+00 9.31566730e-02
6.53286099e-01 3.38228971e-01 -1.10767901e+00 1.57860726e-01
2.44349122e-01 -1.18219328e+00 6.80881858e-01 -9.00186419e-01
8.96271408e-01 -1.47089601e-01 3.45694087e-02 -1.01406467e+00
-7.06621289e-01 -5.80166876e-01 -1.05183017e+00 1.09837162e+00
3.99168223e-01 -3.96089107e-01 1.17782569e+00 1.20877934e+00
-1.07891507e-01 -7.27619469e-01 -6.17501259e-01 -6.85364068e-01
-3.84878606e-01 -1.05357838e+00 1.05351079e+00 7.52487063e-01
2.79189795e-01 3.93078744e-01 -4.95885283e-01 6.89212754e-02
5.11489034e-01 -6.70107314e-03 5.09458899e-01 -1.84051871e+00
-4.55762029e-01 -5.41723967e-01 -5.40655017e-01 -6.84423983e-01
-6.04561064e-03 -7.09024310e-01 -9.83157828e-02 -2.01950145e+00
2.61928946e-01 -1.46902397e-01 -7.28485227e-01 7.16986656e-01
-9.56669152e-02 -5.89583218e-01 3.93060356e-01 5.35073698e-01
-6.98745191e-01 4.84632552e-01 1.16193056e+00 -1.23819925e-01
-3.85849446e-01 5.23855865e-01 -9.25993562e-01 9.38095212e-01
9.92511451e-01 -5.55181384e-01 -3.74548137e-01 2.24994924e-02
4.16545391e-01 2.77629644e-01 4.35872018e-01 -1.32795072e+00
4.57639426e-01 -4.35927182e-01 8.25320542e-01 -5.00556588e-01
1.46978587e-01 -1.09211624e+00 4.59060639e-01 8.99958134e-01
-6.64560020e-01 8.07321370e-02 -7.99555480e-02 8.19282770e-01
-3.94539595e-01 -2.42182627e-01 1.53809011e-01 -2.70393044e-01
-9.36852217e-01 1.94468379e-01 -9.64968145e-01 -6.88577890e-01
1.03135765e+00 -3.83544803e-01 -1.15336962e-01 -5.48700988e-01
-1.10760188e+00 2.64340751e-02 -8.68120715e-02 3.48785609e-01
8.01664829e-01 -1.07600498e+00 -4.71028388e-01 8.93059075e-02
-1.04692213e-01 7.02964365e-02 1.60852253e-01 7.96143889e-01
-5.40249050e-01 7.14545250e-01 -3.56312037e-01 -3.64323795e-01
-8.40787828e-01 6.75252557e-01 3.80344391e-01 -1.00301409e+00
-7.71196544e-01 5.28463304e-01 -1.57386586e-01 -2.62817949e-01
1.97518855e-01 -6.91313565e-01 -5.16281843e-01 -1.81624562e-01
8.81721199e-01 5.94853580e-01 1.11171618e-01 2.55820453e-01
-2.69497186e-01 -4.45493788e-01 -1.71098430e-02 1.48810551e-01
1.73416913e+00 -7.60374889e-02 -2.54021525e-01 7.34417498e-01
1.25496700e-01 -5.19202232e-01 -1.48066330e+00 -8.34873393e-02
1.02891898e+00 -3.11172932e-01 -1.44683972e-01 -8.95236909e-01
-8.13780844e-01 1.06219101e+00 2.40103111e-01 6.52173579e-01
9.79185462e-01 1.89290084e-02 5.85226655e-01 6.71642184e-01
1.89093441e-01 -1.03136325e+00 1.08762152e-01 5.68234742e-01
6.28898740e-01 -1.05080736e+00 -1.59248546e-01 -1.74540728e-01
-7.66970336e-01 1.43188643e+00 7.67080963e-01 1.70532122e-01
5.81809163e-01 1.46694884e-01 -3.02874744e-01 -3.89426261e-01
-1.23648429e+00 1.69822171e-01 1.66842565e-01 7.05769598e-01
4.87639278e-01 3.48508596e-01 2.60327637e-01 1.25202978e+00
-1.02617398e-01 4.12132829e-01 5.58255792e-01 8.79152060e-01
-6.30788684e-01 -8.64034951e-01 -4.09216464e-01 8.18131745e-01
-5.51867247e-01 -3.23946550e-02 -3.80070001e-01 4.49703753e-01
1.44215703e-01 1.12311423e+00 1.87874734e-01 -5.91266930e-01
5.79894841e-01 4.83109981e-01 -5.47795780e-02 -6.49395823e-01
-8.89127731e-01 -4.14995462e-01 2.53598899e-01 -7.32574999e-01
-1.93813175e-01 -6.86708510e-01 -1.34730792e+00 -8.70332897e-01
1.86548099e-01 2.09595058e-02 4.34362710e-01 1.34927332e+00
3.03893596e-01 1.23902786e+00 2.63472378e-01 -4.05431688e-01
-4.81674612e-01 -1.14796090e+00 -5.61567366e-01 5.06283760e-01
-1.40231460e-01 -2.48603031e-01 -3.26718353e-02 6.82905614e-02]
|
[8.581814765930176, 5.972146987915039]
|
a4c48e08-c8c1-4edb-806f-019ac8ffe400
|
iteratively-improving-biomedical-entity
|
2305.14645
| null |
https://arxiv.org/abs/2305.14645v1
|
https://arxiv.org/pdf/2305.14645v1.pdf
|
Iteratively Improving Biomedical Entity Linking and Event Extraction via Hard Expectation-Maximization
|
Biomedical entity linking and event extraction are two crucial tasks to support text understanding and retrieval in the biomedical domain. These two tasks intrinsically benefit each other: entity linking disambiguates the biomedical concepts by referring to external knowledge bases and the domain knowledge further provides additional clues to understand and extract the biological processes, while event extraction identifies a key trigger and entities involved to describe each biological process which also captures the structural context to better disambiguate the biomedical entities. However, previous research typically solves these two tasks separately or in a pipeline, leading to error propagation. What's more, it's even more challenging to solve these two tasks together as there is no existing dataset that contains annotations for both tasks. To solve these challenges, we propose joint biomedical entity linking and event extraction by regarding the event structures and entity references in knowledge bases as latent variables and updating the two task-specific models in a hard Expectation-Maximization (EM) fashion: (1) predicting the missing variables for each partially annotated dataset based on the current two task-specific models, and (2) updating the parameters of each model on the corresponding pseudo completed dataset. Experimental results on two benchmark datasets: Genia 2011 for event extraction and BC4GO for entity linking, show that our joint framework significantly improves the model for each individual task and outperforms the strong baselines for both tasks. We will make the code and model checkpoints publicly available once the paper is accepted.
|
['Lifu Huang', 'Zhiyang Xu', 'Minqian Liu', 'Xiaochu Li']
|
2023-05-24
| null | null | null | null |
['event-extraction', 'entity-linking']
|
['natural-language-processing', 'natural-language-processing']
|
[ 1.92791060e-01 3.66842270e-01 -4.58149195e-01 -1.31111011e-01
-1.16291308e+00 -6.00638509e-01 4.22131568e-01 7.83151627e-01
-6.12326562e-01 1.04571283e+00 4.03553188e-01 -2.27118328e-01
-1.19723544e-01 -4.88017946e-01 -8.01784039e-01 -7.43028224e-01
1.77284628e-02 6.27392411e-01 2.56064147e-01 3.57284665e-01
-7.90234283e-02 1.51404455e-01 -9.85279858e-01 3.38387638e-01
7.71482348e-01 8.02815020e-01 1.82227150e-01 5.18491924e-01
-2.08584681e-01 5.90678692e-01 -2.84275413e-01 -5.94508588e-01
-2.73297846e-01 -2.83279806e-01 -1.09933901e+00 -3.26232016e-01
-2.35828638e-01 8.29531997e-02 -1.98202148e-01 8.78352463e-01
6.20607913e-01 -9.97177362e-02 5.12079179e-01 -1.03449595e+00
-3.49617422e-01 7.55414605e-01 -6.41544342e-01 2.74942040e-01
2.22610846e-01 -6.53610080e-02 1.05100584e+00 -8.12749028e-01
9.84571338e-01 9.69771564e-01 6.21002853e-01 3.72338146e-01
-1.16029739e+00 -6.51287079e-01 9.53377709e-02 1.18385635e-01
-1.46360278e+00 -5.16280890e-01 3.61076027e-01 -4.02913779e-01
1.28349841e+00 1.06778353e-01 2.97657669e-01 1.14026034e+00
7.58418664e-02 7.13855863e-01 5.46786666e-01 -1.86071515e-01
2.11135924e-01 -5.41460738e-02 3.67876858e-01 7.26470113e-01
4.95804578e-01 -2.25093275e-01 -5.81407428e-01 -4.85085338e-01
1.95918798e-01 3.95417549e-02 -1.86130896e-01 1.33904040e-01
-1.52401161e+00 4.08724844e-01 5.18452898e-02 2.78948635e-01
-7.15890944e-01 2.06905790e-02 6.28004611e-01 -6.50847927e-02
4.88601536e-01 5.66186786e-01 -1.04220307e+00 1.29356429e-01
-8.87188494e-01 1.94196373e-01 9.80481744e-01 8.93616021e-01
5.90548337e-01 -7.37351537e-01 -3.30118775e-01 7.31002390e-01
4.45598841e-01 1.02015525e-01 4.83454466e-01 -5.47607660e-01
4.01328087e-01 7.13971317e-01 1.47116899e-01 -8.89029562e-01
-7.12677300e-01 -3.79353613e-01 -5.92137396e-01 -7.70747483e-01
6.02648556e-01 -3.82945091e-01 -9.20479953e-01 1.89905727e+00
7.12454498e-01 7.68386662e-01 2.22843438e-02 5.25849760e-01
1.30770314e+00 5.12017429e-01 7.22015500e-01 -2.47246057e-01
2.12304044e+00 -7.96006620e-01 -1.12535548e+00 -3.18918735e-01
7.42949307e-01 -8.22801888e-01 1.94532335e-01 1.08092777e-01
-1.02199554e+00 -1.55330136e-01 -1.01207721e+00 -4.57044065e-01
-6.59222186e-01 2.65125930e-01 7.38065183e-01 6.29443303e-02
-5.38699389e-01 3.31159085e-01 -1.27332103e+00 -3.47948700e-01
5.63222468e-01 3.22559267e-01 -4.44356263e-01 9.16693658e-02
-1.65056694e+00 9.97241139e-01 6.96317911e-01 1.31305501e-01
-6.54743075e-01 -9.85555172e-01 -8.53175581e-01 1.76203862e-01
5.71630597e-01 -9.51026440e-01 9.09055293e-01 -4.76346128e-02
-8.32222998e-01 1.05828476e+00 -4.84303713e-01 -3.91672522e-01
2.61014879e-01 -1.40698180e-01 -3.83756727e-01 9.83016193e-02
4.45902854e-01 7.86327481e-01 1.73280895e-01 -7.44954169e-01
-7.38015115e-01 -5.03075421e-01 -3.60914439e-01 -1.26383275e-01
-1.76236078e-01 1.13758847e-01 -1.14004707e+00 -6.82886362e-01
1.51920930e-01 -6.92395747e-01 -3.02640796e-01 -1.29999191e-01
-6.55716777e-01 -4.37677979e-01 3.55401218e-01 -9.72195089e-01
1.31435597e+00 -1.95224321e+00 2.94652164e-01 -1.28643021e-01
4.79977101e-01 7.20705166e-02 3.40696909e-02 2.09995478e-01
-2.91207254e-01 1.23755634e-01 -1.75405443e-01 -3.73891622e-01
-6.54650703e-02 1.80905774e-01 -5.62022477e-02 2.30006084e-01
4.42824423e-01 1.25886869e+00 -1.10723948e+00 -7.97816575e-01
-3.80758375e-01 5.57743728e-01 -4.01220262e-01 1.39144165e-02
-2.66507566e-01 5.79570234e-01 -7.30515599e-01 7.85945892e-01
3.52758288e-01 -7.33873725e-01 6.40411794e-01 -5.23148179e-01
3.20046335e-01 4.90142554e-01 -1.23842299e+00 1.74238336e+00
1.03505477e-01 2.05436409e-01 1.21872621e-02 -1.16068304e+00
4.85815316e-01 7.45392442e-01 8.24696779e-01 -2.23077759e-01
1.58994943e-02 1.37310565e-01 -3.39013785e-01 -7.96995580e-01
1.88971519e-01 -1.82246432e-01 -1.98601171e-01 3.61709595e-01
3.34830225e-01 4.74390745e-01 2.17585057e-01 3.75627041e-01
1.43217623e+00 2.46716425e-01 8.26212049e-01 1.00389920e-01
5.03418565e-01 -1.36692718e-01 1.10474908e+00 6.42643452e-01
-7.75876716e-02 1.74334273e-01 6.95981324e-01 -2.81132728e-01
-5.61507642e-01 -9.20223653e-01 -2.72943676e-01 8.61594617e-01
-1.03207693e-01 -7.06919014e-01 -4.65006888e-01 -9.62262750e-01
6.46784380e-02 5.41084647e-01 -7.67302275e-01 -3.22695673e-02
-4.93146390e-01 -1.25225556e+00 8.52574468e-01 6.13514543e-01
1.35649934e-01 -8.24713886e-01 -3.04555506e-01 4.43662673e-01
-6.77783191e-01 -1.46514809e+00 -4.10861045e-01 5.23497760e-01
-5.61050653e-01 -1.27615356e+00 -3.85376722e-01 -5.96404850e-01
6.76911354e-01 -3.06623846e-01 1.18550622e+00 1.44076109e-01
-6.05506659e-01 -5.91205508e-02 -4.29661348e-02 -6.67935967e-01
-2.53381461e-01 1.35071382e-01 -2.39047155e-01 -2.41617903e-01
6.88970864e-01 -3.64034921e-01 -6.18739545e-01 8.57763365e-02
-9.67278540e-01 1.81771427e-01 6.51109815e-01 8.76757681e-01
8.00683320e-01 -1.96624324e-02 7.80173957e-01 -1.30892825e+00
1.16151743e-01 -8.62105668e-01 -3.26787174e-01 3.72974426e-01
-6.73906744e-01 1.48485094e-01 1.39986739e-01 -3.21222723e-01
-1.04160893e+00 3.40606093e-01 -1.16639435e-01 1.10325299e-01
-1.85642779e-01 1.03235471e+00 -4.34834123e-01 7.26826489e-01
2.39739954e-01 5.75685836e-02 -3.93968374e-01 -5.41764736e-01
3.97382587e-01 3.93944383e-01 7.73377180e-01 -6.81880593e-01
4.39744174e-01 4.12860900e-01 6.31655753e-02 -3.34330797e-01
-1.15683115e+00 -7.00355291e-01 -7.12424338e-01 4.17185485e-01
1.27376103e+00 -1.11692584e+00 -7.79713690e-01 3.48469973e-01
-1.28467882e+00 5.68715334e-02 -1.05403373e-02 6.02880955e-01
-1.28352150e-01 3.91195953e-01 -8.39007616e-01 -3.86463016e-01
-3.97674292e-01 -1.13436472e+00 1.43986356e+00 2.03700170e-01
-4.54377383e-01 -1.05939674e+00 1.77390829e-01 4.31404948e-01
-1.08656891e-01 3.72109085e-01 1.22403848e+00 -1.04278672e+00
-4.87566531e-01 -2.70874798e-01 -3.07580680e-01 -3.72265100e-01
2.36743852e-01 -1.63132474e-01 -9.75460887e-01 1.14601828e-01
-2.35721931e-01 -5.79853542e-02 1.17435849e+00 3.55632961e-01
9.53159034e-01 -2.50349671e-01 -1.10258126e+00 5.43207347e-01
1.11659658e+00 -2.80007254e-02 4.63518262e-01 1.61813527e-01
5.83918571e-01 6.14531100e-01 4.46906000e-01 3.51435483e-01
7.10931599e-01 6.37678683e-01 7.15888515e-02 -3.35650504e-01
-2.00309195e-02 -2.15817034e-01 1.12282798e-01 6.75232589e-01
2.21941218e-01 -2.26071343e-01 -9.67305243e-01 7.48790205e-01
-2.04729199e+00 -7.56451547e-01 -1.31035924e-01 1.94473803e+00
1.65734994e+00 -1.25221943e-03 -2.07590163e-01 -2.69702017e-01
6.16566658e-01 -1.94017261e-01 -6.28178895e-01 3.75107229e-01
-1.00803368e-01 3.65467697e-01 3.60288054e-01 3.03055018e-01
-1.18311298e+00 9.49148297e-01 5.86543941e+00 6.79363430e-01
-8.36207330e-01 3.07815075e-01 6.05360270e-01 -1.63596228e-01
-3.84774655e-02 2.82123059e-01 -1.20931089e+00 5.61013401e-01
1.07587600e+00 -1.22882046e-01 2.40246067e-03 2.92662919e-01
1.70336083e-01 -1.21753305e-01 -1.43986619e+00 7.96363771e-01
-3.16969186e-01 -1.42210209e+00 -1.42169639e-01 5.39489388e-02
3.50492150e-01 1.28829092e-01 -3.40887845e-01 3.37326318e-01
5.44262171e-01 -9.03238654e-01 2.54220337e-01 8.41631114e-01
4.75657970e-01 -3.43951762e-01 8.31337929e-01 2.18983650e-01
-1.11141121e+00 2.61472642e-01 -9.47404206e-02 5.14855027e-01
2.55787522e-01 9.40983891e-01 -9.96577561e-01 9.11369681e-01
6.13112748e-01 8.87265086e-01 -5.28290510e-01 9.21643853e-01
-3.62504959e-01 7.74816930e-01 -1.86783031e-01 3.52433473e-01
-1.84148118e-01 2.73254871e-01 6.07497096e-01 1.47228777e+00
8.48362669e-02 2.62541592e-01 3.06257516e-01 1.09856045e+00
-4.58536208e-01 4.21307012e-02 -8.96429718e-02 -3.57210547e-01
6.58984721e-01 1.59439278e+00 -9.67363715e-01 -4.57859606e-01
-4.01427388e-01 7.47567475e-01 3.45358223e-01 3.03652912e-01
-1.00227737e+00 -4.53625858e-01 5.44176817e-01 -2.56184012e-01
2.63822585e-01 1.87818125e-01 -2.26541713e-01 -1.27682245e+00
-3.03173095e-01 -6.65451884e-01 9.40718532e-01 -5.70751369e-01
-1.40054679e+00 1.82886675e-01 -1.50199682e-01 -6.83960497e-01
-1.69197932e-01 -4.11030859e-01 -9.56704468e-02 9.78306830e-01
-1.58564937e+00 -1.24305224e+00 1.41901337e-02 2.05685362e-01
2.47576788e-01 3.26967299e-01 9.84786749e-01 6.82369530e-01
-9.30764794e-01 7.14076042e-01 -1.83444664e-01 3.42636496e-01
1.22315133e+00 -1.23401189e+00 2.81282961e-01 6.32742882e-01
-1.06288500e-01 9.68854666e-01 5.81725121e-01 -1.05633533e+00
-1.28210640e+00 -1.19780016e+00 1.42941868e+00 -8.48317742e-01
6.58068120e-01 -3.25186908e-01 -1.11038375e+00 9.07989740e-01
-8.09263512e-02 -1.36468694e-01 1.07967746e+00 2.75084317e-01
-2.79924244e-01 3.58368695e-01 -8.77326369e-01 2.92309463e-01
1.00919330e+00 -4.75196511e-01 -7.63826251e-01 5.71176291e-01
7.12172568e-01 -4.98788059e-01 -1.22177279e+00 3.49980712e-01
3.92827988e-01 -4.56288829e-02 1.20484924e+00 -1.04301405e+00
4.31020766e-01 -5.71677923e-01 -3.19207623e-03 -8.92053902e-01
-1.96255311e-01 -5.06858230e-01 -3.37852597e-01 1.50438726e+00
8.24209869e-01 -6.20186865e-01 3.87823701e-01 6.08446360e-01
5.38997948e-02 -7.34617949e-01 -7.11454391e-01 -2.51825571e-01
-2.38016859e-01 -3.04437876e-01 5.65154672e-01 1.35984397e+00
1.90555766e-01 6.76643968e-01 -1.96081519e-01 4.12208855e-01
3.38264883e-01 1.15310773e-01 4.50591922e-01 -1.37483072e+00
-3.56941909e-01 -1.91756770e-01 1.02434577e-02 -6.46757126e-01
2.79183090e-01 -1.28997600e+00 1.68965116e-01 -1.66731107e+00
7.30025709e-01 -3.08399111e-01 -5.79292655e-01 9.97989655e-01
-9.52883124e-01 -2.94344611e-02 -4.73778844e-01 1.50801197e-01
-9.17134643e-01 1.53443977e-01 6.96955800e-01 -1.76285744e-01
-1.45843402e-01 -2.89032310e-01 -1.12731481e+00 6.55774713e-01
1.91603720e-01 -8.38735163e-01 -8.53485614e-02 -2.46419594e-01
4.43003058e-01 2.62510091e-01 3.92168850e-01 -5.02547264e-01
5.90882361e-01 6.79726601e-02 6.25520706e-01 -7.50447750e-01
4.48632352e-02 -5.34798503e-01 3.65033746e-01 3.32206935e-01
-5.52124023e-01 -2.08551258e-01 3.78535062e-01 9.70607519e-01
-4.93282788e-02 -6.85531124e-02 5.50703704e-01 -1.09468408e-01
-4.33302462e-01 3.17110300e-01 -2.26966143e-01 2.69208878e-01
8.05875361e-01 1.88807845e-01 -4.49036717e-01 -3.44038829e-02
-1.24840927e+00 6.05936408e-01 8.18700865e-02 3.49682719e-01
1.92067951e-01 -1.03009224e+00 -8.45304132e-01 -1.44751504e-01
7.27588981e-02 7.36693591e-02 2.28369445e-01 1.29666030e+00
8.89378786e-02 5.99328816e-01 1.67666107e-01 -4.34748709e-01
-1.31034195e+00 6.52800441e-01 1.09081306e-01 -8.69971395e-01
-3.18375438e-01 9.62948024e-01 4.53099221e-01 -3.99898082e-01
1.80749267e-01 -1.98327884e-01 -4.10456300e-01 3.22791427e-01
5.08560300e-01 1.57955542e-01 2.02173829e-01 -3.36012691e-01
-6.64801776e-01 8.91430229e-02 -3.98254991e-01 1.11989550e-01
1.63473272e+00 -1.99350387e-01 -5.53896606e-01 1.94055244e-01
1.02963889e+00 3.38015109e-02 -8.59864533e-01 -5.59128463e-01
5.38707197e-01 2.20684320e-01 -1.92882121e-02 -1.14791656e+00
-8.65621090e-01 6.12569392e-01 1.45062611e-01 -2.75750756e-01
9.74364996e-01 3.18397135e-01 8.46487284e-01 1.61026806e-01
-7.72653148e-02 -9.42588866e-01 -2.29944319e-01 4.06575143e-01
4.52399313e-01 -9.41415846e-01 2.52742231e-01 -7.11620331e-01
-4.17217284e-01 8.11694086e-01 5.03303707e-01 5.59033692e-01
6.49666905e-01 4.96300757e-01 -2.96645880e-01 -4.73642677e-01
-1.16366172e+00 -1.70980081e-01 6.11335695e-01 2.06280470e-01
7.44445920e-01 -1.43802583e-01 -5.69239974e-01 1.26699769e+00
2.93905705e-01 3.07496011e-01 -1.34526834e-01 8.73840570e-01
-6.21771365e-02 -1.41069591e+00 -7.07303956e-02 5.52910447e-01
-1.01380014e+00 -3.94843459e-01 -3.96201462e-01 5.32523215e-01
2.96040416e-01 8.25495005e-01 -2.67996460e-01 2.45285891e-02
1.27973005e-01 5.86000025e-01 2.14372247e-01 -6.52808666e-01
-5.57030857e-01 5.20189166e-01 1.96735993e-01 -4.90854770e-01
-3.22892368e-01 -7.36977875e-01 -1.76025116e+00 1.78015277e-01
-3.94812673e-01 2.73846149e-01 3.91458899e-01 1.18358421e+00
7.92416453e-01 9.32203174e-01 -7.92297870e-02 -1.02990091e-01
-2.58702904e-01 -8.46926689e-01 -1.61244422e-01 3.78940225e-01
-6.94083646e-02 -6.98622108e-01 1.08084202e-01 5.12191653e-01]
|
[8.629146575927734, 8.836601257324219]
|
69e09c50-4479-4c0e-b845-e77bc2c425c7
|
two-stream-networks-for-weakly-supervised
| null | null |
http://openaccess.thecvf.com//content/CVPR2023/html/Wang_Two-Stream_Networks_for_Weakly-Supervised_Temporal_Action_Localization_With_Semantic-Aware_Mechanisms_CVPR_2023_paper.html
|
http://openaccess.thecvf.com//content/CVPR2023/papers/Wang_Two-Stream_Networks_for_Weakly-Supervised_Temporal_Action_Localization_With_Semantic-Aware_Mechanisms_CVPR_2023_paper.pdf
|
Two-Stream Networks for Weakly-Supervised Temporal Action Localization With Semantic-Aware Mechanisms
|
Weakly-supervised temporal action localization aims to detect action boundaries in untrimmed videos with only video-level annotations. Most existing schemes detect temporal regions that are most responsive to video-level classification, but they overlook the semantic consistency between frames. In this paper, we hypothesize that snippets with similar representations should be considered as the same action class despite the absence of supervision signals on each snippet. To this end, we devise a learnable dictionary where entries are the class centroids of the corresponding action categories. The representations of snippets identified as the same action category are induced to be close to the same class centroid, which guides the network to perceive the semantics of frames and avoid unreasonable localization. Besides, we propose a two-stream framework that integrates the attention mechanism and the multiple-instance learning strategy to extract fine-grained clues and salient features respectively. Their complementarity enables the model to refine temporal boundaries. Finally, the developed model is validated on the publicly available THUMOS-14 and ActivityNet-1.3 datasets, where substantial experiments and analyses demonstrate that our model achieves remarkable advances over existing methods.
|
['Hongbin Wang', 'Yadong Li', 'Yu Wang']
|
2023-01-01
| null | null | null |
cvpr-2023-1
|
['weakly-supervised-temporal-action', 'action-localization', 'action-recognition', 'multiple-instance-learning']
|
['computer-vision', 'computer-vision', 'computer-vision', 'methodology']
|
[ 4.19230431e-01 6.80630654e-02 -7.22160995e-01 -3.52034390e-01
-5.49992383e-01 -4.01900381e-01 5.90481162e-01 1.55572668e-01
-3.06727380e-01 5.95296264e-01 6.59558117e-01 4.23919737e-01
-1.81053177e-01 -3.12644213e-01 -6.95231378e-01 -8.67299318e-01
-2.77452081e-01 -6.08219206e-02 6.73457980e-01 1.12953432e-01
4.44234312e-01 1.70006808e-02 -1.66706848e+00 7.12869704e-01
5.07381141e-01 1.20970786e+00 1.36058435e-01 1.90436378e-01
3.14989865e-01 1.17506325e+00 -4.02670860e-01 1.42928436e-01
1.35685086e-01 -5.99496007e-01 -8.64091218e-01 3.56192797e-01
5.65013468e-01 -1.76464841e-01 -4.43595946e-01 9.52271760e-01
2.16646437e-02 4.21499312e-01 4.19791400e-01 -1.36537695e+00
-3.68616283e-01 6.00147426e-01 -5.96309245e-01 7.37973154e-01
4.06337023e-01 1.77088052e-01 1.22763062e+00 -6.81204200e-01
7.32818186e-01 1.03547585e+00 3.82596582e-01 3.85871500e-01
-9.46754038e-01 -3.43828648e-01 7.38815188e-01 6.59688234e-01
-1.39012516e+00 -5.93279362e-01 8.73343945e-01 -4.48077291e-01
7.52925992e-01 1.25560954e-01 6.99579835e-01 1.30808270e+00
-6.25842111e-03 8.99314940e-01 8.11449885e-01 -6.78061321e-02
3.38864893e-01 -2.33449578e-01 -7.29261562e-02 6.26487851e-01
-1.33544922e-01 -1.73979655e-01 -9.62899446e-01 2.41145015e-01
6.86821222e-01 2.91466087e-01 -3.90274078e-01 -4.86356735e-01
-1.64467824e+00 4.83547002e-01 3.53388846e-01 6.43259883e-01
-5.46145916e-01 3.23160827e-01 5.93568921e-01 -2.13152785e-02
3.98100466e-01 2.69963205e-01 -3.81397516e-01 -3.59981447e-01
-8.07724893e-01 -1.14215948e-02 2.67546207e-01 8.82818043e-01
6.73064530e-01 -3.15044999e-01 -5.90446293e-01 4.50736046e-01
2.02184975e-01 -2.10803807e-01 5.90554059e-01 -1.05171978e+00
4.53374833e-01 9.70146239e-01 1.98983401e-01 -1.33821869e+00
-2.14018449e-01 -3.32068980e-01 -4.75748688e-01 -1.50325283e-01
4.40237910e-01 2.16897070e-01 -6.79843247e-01 1.71526587e+00
3.29246610e-01 9.16317403e-01 3.52412984e-02 1.11654866e+00
5.61627269e-01 3.91196221e-01 2.90375739e-01 -2.39134833e-01
1.31323326e+00 -1.06523335e+00 -6.78590119e-01 -2.94816375e-01
6.36346579e-01 -3.18473995e-01 9.65744615e-01 2.11720899e-01
-8.41407537e-01 -7.92437732e-01 -9.05755281e-01 1.86596692e-01
-1.19468078e-01 1.85911149e-01 5.99593699e-01 -5.76439798e-02
-7.95527339e-01 7.96057224e-01 -8.48258555e-01 -5.00907660e-01
6.40313387e-01 7.88683891e-02 -3.52113456e-01 2.70826042e-01
-9.84325945e-01 5.18537879e-01 6.98672712e-01 2.30501190e-01
-1.23517299e+00 -4.11596417e-01 -8.16169739e-01 -5.57164066e-02
7.17350423e-01 -2.64741510e-01 9.40612376e-01 -1.54423094e+00
-1.13988972e+00 8.99112165e-01 -2.98390478e-01 -7.04297602e-01
3.85675490e-01 -2.70738274e-01 -4.49858576e-01 5.99273324e-01
3.79431725e-01 7.24646628e-01 9.55134034e-01 -1.01105928e+00
-1.16014063e+00 -2.38825604e-01 2.60625392e-01 4.53289568e-01
-4.38874662e-01 -1.48726618e-02 -5.59399068e-01 -6.42214358e-01
3.45547378e-01 -5.27760804e-01 -9.93707776e-02 -7.79570267e-02
-3.59198689e-01 -5.29933453e-01 8.84103835e-01 -3.89206916e-01
1.50069857e+00 -2.34641933e+00 2.84782350e-01 8.77107866e-03
2.71412134e-01 -4.55414392e-02 -2.12009139e-02 1.83441773e-01
-1.48542434e-01 -6.50970265e-02 1.55373424e-01 8.02080706e-03
-1.37614653e-01 2.50855386e-01 -2.66123742e-01 7.18473256e-01
1.93353668e-01 7.57735968e-01 -1.19517469e+00 -6.59595907e-01
2.91834295e-01 8.58414397e-02 -4.20803726e-01 1.63788587e-01
-3.05213004e-01 7.82543957e-01 -6.12028956e-01 7.38107204e-01
6.91817254e-02 -3.35884929e-01 1.74075991e-01 -5.17668426e-01
-1.00995123e-01 2.77149260e-01 -1.20029104e+00 1.86423886e+00
1.01096779e-01 5.82490683e-01 -3.05457473e-01 -1.38703907e+00
7.27378130e-01 3.53355289e-01 8.38887453e-01 -7.06738591e-01
3.74216260e-03 -3.20669636e-02 -9.99465734e-02 -9.00675297e-01
3.26258779e-01 1.31714985e-01 -1.11656055e-01 2.47549117e-01
1.97565064e-01 7.71659613e-01 3.57416898e-01 6.17148019e-02
1.12920165e+00 5.05883157e-01 3.07820290e-01 -1.53724954e-01
6.28500879e-01 -1.42569393e-01 9.91761148e-01 7.41591096e-01
-6.76760912e-01 4.87194449e-01 6.68596387e-01 -5.86117864e-01
-6.37625456e-01 -7.94709146e-01 1.08572990e-01 1.47981346e+00
6.98739529e-01 -5.54342985e-01 -6.65148318e-01 -1.05465841e+00
-3.14037055e-01 3.14895362e-01 -9.29817438e-01 -2.65132010e-01
-5.27993619e-01 -3.14991623e-01 3.78068596e-01 6.20062709e-01
5.85323811e-01 -1.25046527e+00 -8.82094324e-01 1.30077094e-01
-4.69461322e-01 -1.26835811e+00 -5.15772343e-01 1.46642193e-01
-7.97317743e-01 -1.41873956e+00 -4.63584721e-01 -9.23100173e-01
7.35260129e-01 3.36677939e-01 9.35375571e-01 9.46967527e-02
7.86641687e-02 4.02243525e-01 -7.27518678e-01 1.79219052e-01
-3.88414301e-02 -9.42066759e-02 7.80970082e-02 6.16937399e-01
7.43481040e-01 -5.88144779e-01 -1.05700040e+00 6.46667600e-01
-5.35988331e-01 1.19312890e-01 3.69774848e-01 5.40960908e-01
8.68629932e-01 1.23054825e-01 5.82822323e-01 -4.71339196e-01
7.32471645e-02 -6.57845855e-01 -1.82982638e-01 2.77235895e-01
-2.78352499e-01 -1.79672226e-01 5.42367101e-01 -5.91504455e-01
-9.01918590e-01 3.04936349e-01 3.53922665e-01 -6.07036054e-01
-5.86588264e-01 3.11125785e-01 -1.58290669e-01 2.87361920e-01
4.74599510e-01 4.26631808e-01 -4.41780776e-01 -3.48719031e-01
2.80031741e-01 3.24394614e-01 6.96785569e-01 -3.81155461e-01
5.08796871e-01 8.56331229e-01 -2.90578514e-01 -6.33900940e-01
-1.20627522e+00 -7.69106567e-01 -9.35349941e-01 -6.31975889e-01
1.11588526e+00 -1.02189779e+00 -5.71274817e-01 2.40649551e-01
-9.24835682e-01 -1.95251986e-01 -3.10684085e-01 5.59024692e-01
-7.99026608e-01 4.83253419e-01 -2.56609678e-01 -6.00642383e-01
1.68168724e-01 -9.10859704e-01 1.23334002e+00 4.05869097e-01
-3.95922303e-01 -8.06506336e-01 -5.82521781e-02 4.69170868e-01
-2.34129712e-01 3.83876860e-01 4.95967597e-01 -8.16006601e-01
-7.03487337e-01 -5.13881147e-02 -5.65754669e-03 1.02375969e-01
3.63380343e-01 -6.48393109e-02 -9.30219412e-01 -1.69033080e-01
-1.19566947e-01 -2.35813841e-01 9.37382817e-01 5.00616610e-01
1.43879199e+00 -3.69914889e-01 -5.75506568e-01 4.63994265e-01
1.02222288e+00 3.24437261e-01 5.38765728e-01 4.11211252e-01
7.09461629e-01 5.53817391e-01 9.12019908e-01 6.38934672e-01
1.58277795e-01 8.89369428e-01 6.69256687e-01 3.05046309e-02
9.45145339e-02 -4.86222267e-01 6.00283802e-01 4.52050567e-01
-2.41339549e-01 -5.38665988e-02 -6.25022769e-01 7.94607222e-01
-2.22865033e+00 -1.45163298e+00 1.45325800e-02 2.00251865e+00
6.72773838e-01 3.95935923e-01 2.63343006e-01 3.48066725e-02
8.33208084e-01 5.36448658e-01 -6.25990510e-01 2.00471073e-01
-8.06432292e-02 -2.77813524e-01 8.14800784e-02 7.85038695e-02
-1.61977470e+00 1.09453177e+00 5.54769278e+00 7.83689618e-01
-7.31176198e-01 1.93481505e-01 7.36753464e-01 -2.80646950e-01
1.40271395e-01 8.57825875e-02 -6.79183841e-01 7.08973527e-01
5.00787973e-01 3.34820058e-03 1.71778187e-01 8.03545356e-01
7.03355074e-01 -2.43868038e-01 -1.39846623e+00 7.99384832e-01
1.94145501e-01 -1.39317632e+00 -4.36304472e-02 -2.81594992e-01
6.88656390e-01 -1.66037768e-01 -2.04944223e-01 2.17275724e-01
-1.27494514e-01 -8.28706086e-01 9.12448227e-01 7.42900312e-01
4.02288437e-01 -5.43965101e-01 5.76121747e-01 2.61185527e-01
-1.59471738e+00 -3.34573030e-01 -2.82526881e-01 -1.65887102e-01
2.89473869e-02 1.37686059e-01 -6.02525532e-01 3.90863687e-01
8.28078866e-01 1.53008926e+00 -6.44475341e-01 1.01695919e+00
-3.17586899e-01 5.73100686e-01 9.03412178e-02 1.48617744e-01
4.74168062e-01 -1.65653259e-01 4.38647807e-01 1.01521027e+00
7.56855905e-02 9.37973186e-02 5.12463450e-01 6.44372761e-01
1.44133076e-01 -6.81225806e-02 -5.94300628e-01 -1.47567689e-01
3.05566847e-01 1.07026935e+00 -1.03562617e+00 -3.64471257e-01
-4.97678041e-01 1.07749975e+00 3.13182116e-01 5.19358575e-01
-1.20478714e+00 -1.26495734e-01 7.87106454e-01 -1.20509211e-02
3.78446311e-01 8.20793957e-03 -5.00522703e-02 -1.23415780e+00
1.03746518e-01 -7.22578704e-01 7.46269703e-01 -7.01779008e-01
-9.62312877e-01 4.20331448e-01 -1.44854680e-01 -1.68365085e+00
-3.91432494e-02 -2.10389644e-01 -6.63143277e-01 1.97005048e-01
-1.30428028e+00 -9.17361557e-01 -4.91987228e-01 7.59912312e-01
9.17703092e-01 -2.07652263e-02 4.75048512e-01 2.74798185e-01
-6.83075368e-01 3.28700095e-01 -1.45364881e-01 3.82784098e-01
5.55856109e-01 -1.00291896e+00 -1.58402532e-01 9.99288082e-01
4.88207459e-01 3.99063617e-01 6.07305527e-01 -7.15879917e-01
-9.81946468e-01 -1.27499950e+00 7.42425323e-01 -5.21036267e-01
8.23388159e-01 -1.62155882e-01 -8.99802923e-01 7.36402035e-01
-3.72201926e-03 1.12343244e-01 5.32244623e-01 -2.46814936e-02
-1.69149458e-01 -1.70298189e-01 -6.47447646e-01 5.52494049e-01
1.43057227e+00 -6.13067567e-01 -8.53790283e-01 4.75921571e-01
5.65328062e-01 -1.02810152e-01 -8.30479503e-01 2.83323109e-01
4.50034350e-01 -1.11822057e+00 9.18455064e-01 -7.98174739e-01
5.33756733e-01 -6.13442183e-01 -1.81027740e-01 -8.69770885e-01
-4.31689501e-01 -6.19366825e-01 -4.23607320e-01 1.30534053e+00
8.94694850e-02 -3.89696402e-03 7.10505545e-01 1.16385169e-01
-2.31606364e-01 -7.37632096e-01 -1.09505987e+00 -6.76238239e-01
-6.23231232e-01 -2.58478701e-01 2.93944389e-01 1.02892315e+00
3.44889998e-01 9.39162374e-02 -5.69605231e-01 2.08446473e-01
3.60445648e-01 2.82814473e-01 4.42540050e-01 -1.00488329e+00
-1.70739278e-01 -4.88891423e-01 -7.06772864e-01 -1.21422982e+00
3.49825561e-01 -7.99061954e-01 2.94385523e-01 -1.17236614e+00
3.89849156e-01 -1.48002580e-01 -8.59920382e-01 7.22046733e-01
-1.21544100e-01 3.89219642e-01 -6.81280419e-02 2.57595032e-01
-1.38670516e+00 6.18844151e-01 1.13322759e+00 -1.65722877e-01
-2.74910867e-01 3.49508077e-02 -5.70345819e-01 9.65343058e-01
6.70916736e-01 -3.80468994e-01 -5.98428607e-01 -2.05006406e-01
-5.97796310e-03 -1.16520941e-01 6.46967292e-01 -1.23696935e+00
1.66758329e-01 -4.59845722e-01 3.88963372e-01 -4.70794976e-01
1.47433653e-01 -8.94847989e-01 -1.21432833e-01 3.20091784e-01
-7.96118498e-01 -3.17442864e-01 -1.71085328e-01 1.03106511e+00
-3.44084084e-01 -5.29355519e-02 7.03322113e-01 -1.14255130e-01
-1.37412763e+00 4.51913565e-01 -4.50626075e-01 1.95740506e-01
1.39284229e+00 -5.22634566e-01 -1.81180522e-01 -2.19975531e-01
-9.14675117e-01 2.37656206e-01 2.03449845e-01 7.31685579e-01
5.83241582e-01 -1.39678478e+00 -3.60023439e-01 1.64237544e-01
4.08996880e-01 -1.97362557e-01 3.46604526e-01 1.11263859e+00
-1.92669705e-02 3.02910596e-01 -1.90335378e-01 -8.66684198e-01
-1.23123479e+00 5.86501539e-01 4.81019735e-01 1.56980053e-01
-8.92943144e-01 7.25169241e-01 4.17254418e-01 3.53175819e-01
5.48763990e-01 -5.36198556e-01 -5.49622238e-01 3.28304321e-01
6.09425783e-01 2.93536752e-01 -4.20253873e-01 -9.44537103e-01
-6.09236002e-01 4.99481440e-01 1.17483318e-01 4.27283317e-01
1.26126337e+00 -2.97355920e-01 8.09204429e-02 5.84300458e-01
1.02530777e+00 -4.80034500e-01 -1.89532232e+00 -2.74416804e-01
3.52022618e-01 -6.01566732e-01 -1.69913918e-01 -5.49575865e-01
-1.09649026e+00 6.40662909e-01 5.79991579e-01 5.98907843e-02
1.22857440e+00 3.69205534e-01 5.05858481e-01 1.68571308e-01
2.97531068e-01 -1.31585109e+00 5.23691118e-01 2.80821830e-01
6.39471233e-01 -1.18903840e+00 -1.36747643e-01 -2.46748790e-01
-8.62593293e-01 9.73419607e-01 9.94327009e-01 -1.21600449e-01
3.19424927e-01 -2.51556188e-01 -2.12020591e-01 -3.13783050e-01
-7.74902105e-01 -3.83539796e-01 3.93088579e-01 5.37314475e-01
2.05380112e-01 -1.56826392e-01 -3.03472400e-01 7.53922641e-01
4.70176965e-01 1.43305942e-01 2.87915587e-01 7.73057580e-01
-6.60731196e-01 -5.32453537e-01 -8.09888691e-02 2.70806074e-01
-3.16101521e-01 1.85485259e-01 -2.60541558e-01 4.83223289e-01
4.19885457e-01 9.17310834e-01 2.89034277e-01 -5.26878893e-01
1.04503885e-01 1.06212676e-01 2.30780736e-01 -6.06175125e-01
-3.92009944e-01 2.69190341e-01 2.46404987e-02 -1.00743532e+00
-9.18083429e-01 -7.92624116e-01 -1.37607706e+00 2.25018606e-01
7.83920810e-02 6.84273317e-02 -2.16173038e-01 1.14274120e+00
4.43600327e-01 5.69963813e-01 6.25071287e-01 -7.84571946e-01
-1.61565363e-01 -7.12806702e-01 -5.90725601e-01 7.27390349e-01
2.66558826e-01 -8.62559080e-01 -3.12769234e-01 5.72686613e-01]
|
[8.46787166595459, 0.6833968162536621]
|
8d150bd2-d59c-422d-bf13-5ecbda349837
|
refinements-in-motion-and-appearance-for
|
2003.07177
| null |
https://arxiv.org/abs/2003.07177v2
|
https://arxiv.org/pdf/2003.07177v2.pdf
|
Refinements in Motion and Appearance for Online Multi-Object Tracking
|
Modern multi-object tracking (MOT) system usually involves separated modules, such as motion model for location and appearance model for data association. However, the compatible problems within both motion and appearance models are always ignored. In this paper, a general architecture named as MIF is presented by seamlessly blending the Motion integration, three-dimensional(3D) Integral image and adaptive appearance feature Fusion. Since the uncertain pedestrian and camera motions are usually handled separately, the integrated motion model is designed using our defined intension of camera motion. Specifically, a 3D integral image based spatial blocking method is presented to efficiently cut useless connections between trajectories and candidates with spatial constraints. Then the appearance model and visibility prediction are jointly built. Considering scale, pose and visibility, the appearance features are adaptively fused to overcome the feature misalignment problem. Our MIF based tracker (MIFT) achieves the state-of-the-art accuracy with 60.1 MOTA on both MOT16&17 challenges.
|
['Donghaisheng Liu', 'En Yu', 'Shoudong Han', 'Piao Huang', 'Jun Zhao', 'Hongwei Wang', 'Alex ChiChung Kot']
|
2020-03-16
| null | null | null | null |
['online-multi-object-tracking']
|
['computer-vision']
|
[-2.9352039e-01 -5.7933170e-01 -1.2275130e-01 -1.6887167e-01
-4.4911668e-01 -4.4854426e-01 5.3680295e-01 -2.4011883e-01
-4.2688769e-01 3.8077718e-01 -3.0585003e-01 -1.6588545e-01
-4.7612853e-02 -4.1496795e-01 -5.1691121e-01 -9.2758739e-01
2.0010136e-01 2.5024909e-01 8.0788195e-01 1.3932478e-01
5.8082029e-02 4.2521188e-01 -1.6941105e+00 -2.5039068e-01
8.1915021e-01 9.7857338e-01 2.7981582e-01 8.1511211e-01
-1.5829921e-02 2.6877841e-01 -4.3836558e-01 -2.4868973e-01
3.3111286e-01 -4.0948298e-02 -1.9119577e-01 3.2674211e-01
8.6972797e-01 -2.9993096e-01 -2.9347646e-01 1.0335168e+00
4.1701272e-01 1.8687750e-01 4.1472769e-01 -1.7402283e+00
-4.7991532e-01 -1.2399966e-02 -1.0234046e+00 2.6976496e-01
1.6418736e-01 1.9924207e-01 4.8731351e-01 -1.0211250e+00
3.6138010e-01 1.3920676e+00 7.1267295e-01 4.8797646e-01
-8.2958597e-01 -6.8437696e-01 7.2191548e-01 5.0986952e-01
-1.6187428e+00 -2.5183854e-01 6.2100059e-01 -3.7419662e-01
4.2352295e-01 5.5967396e-01 8.4132391e-01 6.4133596e-01
2.2727060e-01 9.3796009e-01 6.5071130e-01 -2.7153328e-01
-2.4157625e-01 2.5249374e-01 3.5111237e-01 6.6389751e-01
4.7715187e-01 2.3438096e-01 -2.6092160e-01 -8.5230529e-02
7.5248080e-01 1.7317297e-01 4.6682455e-02 -5.9493554e-01
-1.1982207e+00 2.6969641e-01 1.8610956e-01 5.6739528e-02
5.1164996e-02 1.3271977e-01 4.6581026e-02 -2.0761250e-02
1.3281539e-01 -4.7879446e-01 -4.0982133e-01 1.3092297e-01
-8.4576726e-01 3.5408437e-01 2.2450058e-01 1.3784102e+00
7.4307233e-01 -1.0035976e-03 -2.8732729e-01 4.7507375e-01
9.8560292e-01 1.0783697e+00 1.5808567e-01 -8.0516243e-01
2.6208058e-01 6.7201233e-01 3.3585632e-01 -1.1802803e+00
-5.0888717e-01 -3.5354912e-01 -5.8167851e-01 3.2313624e-01
5.4120773e-01 -1.6399592e-01 -7.2887582e-01 1.5619333e+00
1.0610605e+00 5.0709188e-01 -2.5670493e-01 1.0834132e+00
8.7911367e-01 4.1789153e-01 1.5599832e-01 -3.4787092e-01
1.5387776e+00 -1.2055179e+00 -9.7445929e-01 -1.4611921e-01
5.6180543e-01 -1.0045087e+00 3.3527553e-01 2.5155443e-01
-9.7084534e-01 -8.8476300e-01 -1.1014302e+00 7.7497728e-02
-3.7933365e-01 5.1337212e-01 4.0927589e-01 8.6411834e-01
-9.4692856e-01 4.0841948e-02 -7.4898517e-01 -3.5519072e-01
7.9769656e-02 6.9257402e-01 -1.8346739e-01 9.7396150e-02
-8.5704988e-01 8.9230055e-01 1.0290483e-01 3.7794745e-01
-6.3149571e-01 -3.8668534e-01 -7.1406275e-01 -4.3875065e-01
2.0295343e-01 -8.7961960e-01 9.3091536e-01 -7.8891516e-01
-1.2903874e+00 6.5698838e-01 -4.2001793e-01 -1.9318765e-01
6.1566800e-01 -3.2312271e-01 -8.6559576e-01 -2.3800276e-01
2.6519833e-02 6.2865806e-01 1.0225333e+00 -1.2928476e+00
-1.2585008e+00 -4.2397344e-01 -2.7071849e-01 3.3856353e-01
-1.5460351e-01 2.4083264e-01 -1.0167416e+00 -4.7683585e-01
1.7566881e-01 -1.0632365e+00 -2.5559729e-01 3.2893130e-01
-6.6435345e-02 -1.5770595e-01 1.5261937e+00 -6.5781957e-01
1.4196875e+00 -2.1207328e+00 8.8749461e-02 6.4597085e-02
1.8164811e-01 3.1345820e-01 -1.3373612e-03 -2.4322534e-01
2.6790488e-01 -3.6725560e-01 3.0949548e-01 -5.7918710e-01
-4.7061957e-02 2.5969638e-02 9.8433554e-02 9.4332683e-01
-3.3154573e-02 8.2931900e-01 -8.4547061e-01 -9.2617649e-01
7.7846777e-01 5.2626717e-01 -2.6996896e-01 4.1597608e-02
-6.6187426e-02 6.4314705e-01 -5.3671241e-01 9.7735941e-01
1.4120322e+00 -6.8679027e-02 -1.2768890e-01 -4.3659550e-01
-6.6010588e-01 -5.7321715e-01 -1.6850497e+00 1.5357418e+00
-1.5913835e-02 3.3267522e-01 3.4582153e-01 -3.0698729e-01
7.7784443e-01 1.8398784e-01 7.6681179e-01 -4.6291101e-01
2.7938104e-01 -5.3827316e-02 -5.2311961e-02 -3.4807777e-01
8.2140899e-01 5.3019226e-01 1.3449499e-01 7.5192135e-03
-2.1471481e-01 5.1644307e-01 -5.9240870e-02 6.0744323e-03
6.0850239e-01 5.3639936e-01 4.8263956e-02 -1.5787123e-01
8.9278370e-01 1.3860375e-01 9.7360593e-01 5.2077121e-01
-7.2165197e-01 5.8445930e-01 -3.9317644e-01 -4.5801115e-01
-9.0092659e-01 -1.0156575e+00 -1.5713800e-01 8.3015543e-01
1.1307714e+00 -3.1646156e-01 -4.8600018e-01 -7.1689093e-01
6.7359306e-02 9.4511330e-02 -5.1514477e-01 1.0174499e-01
-9.1148680e-01 -8.2941836e-01 2.8741893e-01 4.5964447e-01
4.7278383e-01 -4.6066353e-01 -5.5843747e-01 1.6740456e-01
-7.5819895e-02 -1.0652418e+00 -9.4130337e-01 -4.4191337e-01
-6.0097021e-01 -1.1164362e+00 -5.3274596e-01 -5.9572607e-01
7.5457048e-01 1.0427103e+00 4.5827350e-01 3.8565481e-01
-3.2009515e-01 3.0377522e-01 -1.8537652e-01 -1.0507967e-01
3.1158842e-02 -3.4554914e-01 3.9038542e-01 3.8888031e-01
2.9653880e-01 -3.4397459e-01 -9.2206091e-01 7.8409636e-01
-4.3516076e-01 2.0897655e-01 7.2383344e-01 5.5380952e-01
6.2101740e-01 -1.9659257e-01 8.0125093e-02 -3.1982411e-03
-2.6097041e-01 -3.3455256e-01 -9.7713226e-01 6.1331064e-01
-4.5620820e-01 -3.2554796e-01 2.4932316e-01 -8.1160939e-01
-1.1887139e+00 5.1287872e-01 1.7652479e-01 -7.9618025e-01
-2.3971705e-01 -4.4210416e-01 -6.1911267e-01 -6.0799623e-01
6.1273772e-02 1.7725328e-01 -1.4571673e-01 -5.3821373e-01
5.9588796e-01 6.5911978e-01 5.7084608e-01 -2.9388186e-01
1.0731220e+00 7.9830605e-01 1.2586588e-01 -6.9169950e-01
-3.8468453e-01 -8.3984733e-01 -1.0201112e+00 -7.6333570e-01
9.4144344e-01 -1.0483013e+00 -9.2245477e-01 6.2444156e-01
-1.3046222e+00 2.4825546e-01 2.0889960e-01 6.5559870e-01
-1.6274939e-01 6.6102368e-01 -2.1726623e-01 -1.1257241e+00
-2.5225148e-01 -1.2339560e+00 1.2199433e+00 6.1784768e-01
2.1441115e-01 -8.5980225e-01 8.6895414e-03 2.2910425e-01
1.5569006e-01 4.6856496e-02 5.5720732e-02 -1.4412166e-01
-1.1335815e+00 -1.6799518e-01 -4.2889047e-01 -2.7798834e-01
1.0931164e-01 3.9322340e-01 -9.4185764e-01 -4.3045864e-01
-1.5472883e-01 3.4755194e-01 6.7788446e-01 6.7673099e-01
6.3107049e-01 9.7169012e-02 -9.2144698e-01 8.4930789e-01
1.2712538e+00 2.6764807e-01 3.1008172e-01 4.6623787e-01
9.9843436e-01 4.1997519e-01 1.2403760e+00 4.4220996e-01
5.9785444e-01 1.2129045e+00 4.9032468e-01 -1.0938096e-01
-1.9195423e-01 4.9277660e-02 4.1902950e-01 7.6492512e-01
-4.5711376e-02 -1.1776584e-01 -4.2528427e-01 4.0135711e-01
-2.3757560e+00 -9.1658366e-01 -7.0193326e-01 2.3131919e+00
1.6238922e-01 1.9960621e-02 4.0831205e-01 -4.0537921e-01
1.0750519e+00 3.4160916e-02 -4.5905101e-01 2.4131942e-01
-2.7126920e-01 -7.1524310e-01 9.0491515e-01 6.1121815e-01
-1.5932286e+00 8.0123723e-01 5.8265047e+00 9.9478853e-01
-6.9737947e-01 3.5474530e-01 9.2519529e-02 -2.7942024e-02
7.9170644e-02 1.6076499e-01 -1.5785112e+00 5.9906262e-01
5.3760386e-01 1.1710024e-01 4.1249629e-02 7.4684322e-01
3.0850619e-01 -1.5330714e-01 -5.2438706e-01 1.1623074e+00
7.1044214e-02 -1.0710313e+00 -1.0036577e-01 2.0504391e-01
5.1017767e-01 -3.1288967e-01 1.8767969e-01 1.2866564e-01
8.5056618e-02 -5.1350439e-01 1.1172701e+00 5.4271418e-01
4.4791862e-01 -5.2030969e-01 3.9050642e-01 3.0619514e-01
-2.0161207e+00 -7.0645697e-02 -2.6043755e-01 1.8861051e-01
5.3964746e-01 1.6055568e-01 -1.2938309e-01 1.1115255e+00
6.5858424e-01 7.8216159e-01 -7.9134184e-01 1.5571352e+00
2.6196399e-01 8.8798823e-03 -4.9496794e-01 2.0387287e-01
7.9516768e-02 -2.8037792e-01 8.0732876e-01 1.0637242e+00
3.2995471e-01 -7.1111105e-02 4.7987419e-01 5.0665134e-01
6.5605050e-01 -4.1862048e-02 -3.6193553e-01 7.9591715e-01
6.3752609e-01 1.5684545e+00 -6.7816657e-01 -3.0439806e-01
-8.0920547e-01 7.8795332e-01 -1.4428020e-02 1.9249628e-01
-1.4449906e+00 2.1854815e-01 8.6673999e-01 -5.8032252e-02
5.8552390e-01 -3.4008902e-01 -8.5630104e-02 -1.1780747e+00
9.1170482e-02 -4.0839732e-01 4.7510818e-01 -4.1551715e-01
-9.2369652e-01 3.7155199e-01 2.0178486e-02 -1.8339454e+00
2.6583317e-01 -5.1367581e-01 -7.1184427e-01 7.3451132e-01
-1.4092797e+00 -1.6634413e+00 -4.4429582e-01 6.6327310e-01
5.9671795e-01 1.3158475e-01 2.8228566e-01 8.7080151e-01
-1.0355155e+00 7.4929422e-01 1.0611489e-01 -1.0472479e-01
8.3490735e-01 -9.4728094e-01 2.1667784e-01 1.1697017e+00
-1.2403824e-01 4.8871809e-01 6.5150136e-01 -9.6490759e-01
-1.5822510e+00 -1.3813808e+00 5.3386652e-01 -6.8210310e-01
5.2650964e-01 -2.2695683e-01 -7.2415191e-01 6.5739965e-01
8.4751636e-02 3.5832682e-01 2.2932920e-01 -2.4468586e-01
5.7959251e-02 -1.6244486e-01 -9.8418921e-01 6.9657141e-01
1.1363955e+00 1.3716441e-01 -2.3276739e-01 1.3384870e-01
7.1915215e-01 -6.4644545e-01 -8.1533784e-01 4.5552656e-01
9.0244806e-01 -5.6318843e-01 1.3295640e+00 -2.2474381e-01
-8.9195377e-01 -1.1680496e+00 -1.5608065e-01 -4.9389285e-01
-5.6852901e-01 -7.2918010e-01 -5.0126112e-01 1.3697960e+00
7.7190571e-02 -3.1716928e-01 7.1237785e-01 5.8282655e-01
-1.6773510e-01 -3.6122096e-01 -1.1383778e+00 -8.7322807e-01
-4.9315438e-01 -2.0350428e-01 5.2729261e-01 7.3138922e-01
-4.1041696e-01 1.8394618e-01 -7.7976322e-01 7.3303360e-01
9.8816955e-01 1.9879489e-01 1.0496116e+00 -1.1727414e+00
-3.0298807e-02 -3.8254070e-01 -4.5183849e-01 -1.4829938e+00
-2.2742961e-01 -3.0612764e-01 1.2417866e-02 -1.0889794e+00
2.3994911e-01 -5.5633616e-01 -3.3968157e-01 -1.7920159e-02
-3.9827558e-01 3.4101598e-02 2.9322711e-01 2.9537734e-01
-1.1852460e+00 5.6768394e-01 1.2546041e+00 -1.3453405e-01
-1.9698103e-01 2.9434842e-01 -1.7840394e-01 6.3059896e-01
5.2415198e-01 -3.2145429e-01 -1.8659908e-01 -4.6549708e-01
-3.9382723e-01 -3.9484054e-02 6.6222942e-01 -1.2079792e+00
7.2636992e-01 -2.8622898e-01 5.2684003e-01 -1.4415010e+00
6.9319803e-01 -1.0770303e+00 6.0012656e-01 4.8321393e-01
3.9651734e-01 4.3919668e-01 2.9831865e-01 8.4886253e-01
6.9078766e-02 1.9000892e-01 7.0223999e-01 1.8964340e-01
-1.0661794e+00 6.9075501e-01 -1.7446125e-01 -4.5823014e-01
1.4091682e+00 -6.2461329e-01 -4.5031670e-01 1.0188286e-01
-6.8431765e-01 7.1546698e-01 6.2499869e-01 6.5606534e-01
6.5014577e-01 -1.6700909e+00 -5.1459295e-01 2.6651302e-01
4.5160402e-02 -1.0913231e-01 7.2858721e-01 1.4858299e+00
-1.8593436e-01 4.7987142e-01 2.1252533e-02 -9.3830836e-01
-1.8297524e+00 8.7788618e-01 3.0747113e-01 -1.5208052e-01
-6.4073086e-01 5.9554398e-01 3.8503927e-01 -2.4075863e-01
3.1018981e-01 -2.3854448e-01 -2.4771738e-01 -1.5010329e-01
7.4850714e-01 5.6441951e-01 -3.3388638e-01 -1.2933475e+00
-6.7949760e-01 1.1225158e+00 -3.2654315e-02 1.8820256e-02
6.1779648e-01 -8.4803122e-01 9.3733214e-02 1.8123910e-01
8.3596724e-01 4.9528256e-02 -1.5380838e+00 -9.8644719e-02
-7.2182730e-02 -7.7463096e-01 -1.3113761e-01 -4.2177910e-01
-1.0228839e+00 6.4957106e-01 1.0593632e+00 -1.1586190e-01
8.5209602e-01 -2.0284851e-01 8.2838106e-01 -1.5503684e-01
4.0485254e-01 -9.4280368e-01 -1.8420546e-01 3.8051730e-01
3.3102092e-01 -1.3385682e+00 1.0951892e-01 -5.5705452e-01
-4.5224550e-01 8.6707503e-01 1.1566699e+00 1.6361097e-01
6.4178550e-01 2.3495962e-01 1.1112457e-01 -3.3949967e-03
-5.8336008e-01 -3.7249509e-01 4.1835523e-01 7.7929538e-01
-9.0195514e-02 -8.6164579e-02 -3.1963280e-01 5.1914787e-01
5.0679237e-01 -4.3113065e-01 -5.8878977e-02 8.3587307e-01
-5.2088869e-01 -1.2142990e+00 -9.5333135e-01 -8.9239188e-02
-1.4898786e-01 2.2358918e-01 -2.5324855e-02 8.3572459e-01
5.8897084e-01 9.9885893e-01 1.5854079e-02 -6.9995397e-01
1.7102961e-01 -3.7347582e-01 5.6031030e-01 -7.5625524e-02
-5.4808348e-01 4.3505549e-01 -6.5267109e-03 -5.8050334e-01
-7.2708851e-01 -1.0245612e+00 -1.0936906e+00 -3.0791464e-01
-8.9469486e-01 -6.5985635e-02 5.8344072e-01 9.0429705e-01
3.0994946e-01 4.3573603e-01 6.0041893e-01 -9.6256500e-01
-1.8928659e-01 -7.4657845e-01 -4.1698167e-01 9.5625378e-02
5.1867384e-01 -9.7281718e-01 -1.0995353e-01 8.8100433e-02]
|
[6.503890037536621, -2.023082971572876]
|
407b8582-e8d2-458d-a7b4-e86f626770bb
|
spidercnn-deep-learning-on-point-sets-with
|
1803.11527
| null |
http://arxiv.org/abs/1803.11527v3
|
http://arxiv.org/pdf/1803.11527v3.pdf
|
SpiderCNN: Deep Learning on Point Sets with Parameterized Convolutional Filters
|
Deep neural networks have enjoyed remarkable success for various vision
tasks, however it remains challenging to apply CNNs to domains lacking a
regular underlying structures such as 3D point clouds. Towards this we propose
a novel convolutional architecture, termed SpiderCNN, to efficiently extract
geometric features from point clouds. SpiderCNN is comprised of units called
SpiderConv, which extend convolutional operations from regular grids to
irregular point sets that can be embedded in R^n, by parametrizing a family of
convolutional filters. We design the filter as a product of a simple step
function that captures local geodesic information and a Taylor polynomial that
ensures the expressiveness. SpiderCNN inherits the multi-scale hierarchical
architecture from classical CNNs, which allows it to extract semantic deep
features. Experiments on ModelNet40 demonstrate that SpiderCNN achieves
state-of-the-art accuracy 92.4% on standard benchmarks, and shows competitive
performance on segmentation task.
|
['Tianqi Fan', 'Yu Qiao', 'Mingye Xu', 'Yifan Xu', 'Long Zeng']
|
2018-03-30
|
spidercnn-deep-learning-on-point-sets-with-1
|
http://openaccess.thecvf.com/content_ECCV_2018/html/Yifan_Xu_SpiderCNN_Deep_Learning_ECCV_2018_paper.html
|
http://openaccess.thecvf.com/content_ECCV_2018/papers/Yifan_Xu_SpiderCNN_Deep_Learning_ECCV_2018_paper.pdf
|
eccv-2018-9
|
['3d-part-segmentation']
|
['computer-vision']
|
[-2.24567011e-01 5.01443967e-02 1.32402495e-01 -4.27994579e-01
-2.99347937e-01 -6.40022933e-01 6.74688339e-01 -2.54224092e-01
-4.93217438e-01 6.16784208e-02 -2.23347679e-01 -3.74679357e-01
-1.01058828e-02 -1.13583422e+00 -1.12221754e+00 -3.26834828e-01
-3.91375512e-01 2.57747948e-01 5.76030612e-01 -1.92484185e-01
1.75367981e-01 1.05460346e+00 -1.28271306e+00 1.68871018e-03
8.21213663e-01 1.25229502e+00 1.06106341e-01 4.35806185e-01
-2.55724877e-01 2.88432091e-01 -2.72152096e-01 -3.69782060e-01
4.89317536e-01 2.42101550e-01 -9.10700142e-01 7.29018066e-04
5.71050763e-01 -4.64263022e-01 -4.33216989e-01 1.01318765e+00
1.57181218e-01 7.08469823e-02 5.90389490e-01 -9.72131908e-01
-8.96868110e-01 9.60300341e-02 -5.22350371e-01 8.93227309e-02
-1.20738134e-01 1.34130105e-01 1.05193567e+00 -1.01896071e+00
4.30833161e-01 1.48149729e+00 1.16717100e+00 3.05198133e-01
-1.20889413e+00 -6.82580113e-01 1.98230986e-02 -2.57245630e-01
-1.30653298e+00 1.23152025e-01 5.85070491e-01 -4.17506516e-01
1.15340948e+00 -1.10627986e-01 8.72375965e-01 7.42299676e-01
9.87241492e-02 7.01711237e-01 7.14965343e-01 1.58896789e-01
1.13935165e-01 -6.21005952e-01 2.92907599e-02 9.23726976e-01
1.21312328e-01 6.45599291e-02 -6.25246242e-02 -6.21643811e-02
1.50540352e+00 1.51070446e-01 -1.00358635e-01 -4.63365018e-01
-1.05479491e+00 1.16515005e+00 1.35367596e+00 -5.00138365e-02
-1.79676622e-01 5.57097197e-01 2.98964471e-01 5.99288531e-02
4.52155054e-01 3.98685485e-01 -3.88515234e-01 2.62472242e-01
-7.45683730e-01 3.63583863e-01 5.03785789e-01 1.29687965e+00
9.52949703e-01 -1.11604415e-01 1.49557263e-01 6.84721828e-01
2.12984428e-01 4.58622962e-01 7.09347799e-02 -1.29681730e+00
1.36248976e-01 9.88815367e-01 -1.79768562e-01 -1.00794244e+00
-5.85012734e-01 -5.53676367e-01 -9.46560085e-01 3.86020750e-01
1.24771446e-01 -1.53021431e-02 -1.12806845e+00 1.33962750e+00
3.45360577e-01 4.20116276e-01 -2.55366117e-01 1.00259078e+00
1.08916092e+00 6.85808241e-01 -1.65288180e-01 7.04101741e-01
1.27481627e+00 -9.85200465e-01 6.56990781e-02 -9.34590548e-02
3.87971759e-01 -4.43674386e-01 8.91386628e-01 1.59301296e-01
-1.23089635e+00 -6.73644483e-01 -1.11631298e+00 -6.02656245e-01
-3.55690449e-01 2.18385942e-02 8.10421884e-01 1.76598847e-01
-1.34201491e+00 8.81856143e-01 -1.17553365e+00 -2.75766730e-01
8.83524716e-01 4.98001844e-01 -3.34249228e-01 9.54789817e-02
-8.13003540e-01 5.13408720e-01 2.70641416e-01 3.37862700e-01
-8.28221083e-01 -9.89292383e-01 -1.07393384e+00 2.20926449e-01
6.65853992e-02 -9.88370538e-01 1.29561651e+00 -5.82017303e-01
-1.48642361e+00 9.72231627e-01 7.54622892e-02 -6.33019865e-01
5.15842736e-01 -2.72942960e-01 4.88863550e-02 3.64587992e-01
2.48786688e-01 1.01002336e+00 8.55397463e-01 -1.04713285e+00
-7.06606686e-01 -3.81159037e-01 2.63883024e-01 -1.71853956e-02
-1.43445849e-01 -5.59350774e-02 -8.51861358e-01 -6.18917406e-01
2.86552310e-01 -9.13602650e-01 -6.16861403e-01 4.96661782e-01
-5.58021843e-01 -6.04888022e-01 1.01644123e+00 -8.05563554e-02
6.16875172e-01 -2.06152368e+00 -2.50058994e-03 3.95354182e-01
5.10190010e-01 3.56281519e-01 -2.05349460e-01 1.04996391e-01
7.00916201e-02 2.78798312e-01 -5.60307503e-01 -3.98901969e-01
6.00793883e-02 2.98304826e-01 -3.21403027e-01 5.03233016e-01
7.63304830e-01 1.32223427e+00 -8.02740216e-01 -2.24134162e-01
3.30180645e-01 6.66179955e-01 -6.85634792e-01 8.17269459e-02
-3.72449458e-01 2.01534435e-01 -6.80526912e-01 5.37573516e-01
1.04063737e+00 -6.30919456e-01 -4.66458589e-01 -9.11374614e-02
-2.26598755e-01 2.73282200e-01 -7.64660954e-01 2.00027180e+00
-3.49993885e-01 5.74407637e-01 2.55647629e-01 -9.77072060e-01
1.06962657e+00 -2.12621078e-01 4.23551440e-01 -4.73880649e-01
1.81798711e-01 2.10313335e-01 -2.40771398e-01 -3.59681621e-02
2.35347465e-01 2.51312673e-01 -2.57471818e-02 -8.07161536e-03
2.08891317e-01 -4.94621605e-01 -1.29738912e-01 6.13219850e-02
1.13764048e+00 1.31367862e-01 -1.03067279e-01 -5.19412935e-01
4.99307603e-01 6.22978806e-02 4.45524693e-01 6.65776610e-01
1.16986364e-01 8.62747371e-01 5.23825884e-01 -8.79252553e-01
-1.02542412e+00 -1.32587886e+00 -5.25924623e-01 6.80791438e-01
2.72008210e-01 -4.80437487e-01 -8.42081130e-01 -6.30979121e-01
3.17518532e-01 1.44507363e-01 -6.70062184e-01 -4.67699133e-02
-7.69709229e-01 -3.34762931e-01 6.90081179e-01 9.54936802e-01
1.01384234e+00 -1.02790201e+00 -8.03336442e-01 3.05827558e-01
2.48380214e-01 -1.45968485e+00 -4.33598042e-01 1.44456282e-01
-1.10128796e+00 -1.17394269e+00 -6.18105531e-01 -1.00665796e+00
6.01545215e-01 4.36081231e-01 1.34296811e+00 1.27128884e-01
-2.24325463e-01 2.00741544e-01 -1.75735399e-01 -4.89878267e-01
1.41805246e-01 3.79442662e-01 -3.98136914e-01 -1.69765070e-01
4.80635494e-01 -7.97517955e-01 -9.41848576e-01 3.49833131e-01
-9.85893726e-01 6.72148494e-03 5.68221748e-01 6.85890198e-01
8.15775931e-01 -2.80482590e-01 2.29686365e-01 -5.87961733e-01
3.16229671e-01 -3.15461874e-01 -9.06406105e-01 -2.83131182e-01
-6.05787821e-02 -5.43125793e-02 7.46733725e-01 1.02402285e-01
-5.45141220e-01 1.54358163e-01 -5.45122385e-01 -6.85681224e-01
-2.36706734e-01 2.64625609e-01 1.24586947e-01 -5.89735806e-01
4.66692835e-01 1.16213411e-01 8.21613893e-02 -5.05338907e-01
5.76012790e-01 3.41792852e-01 7.52081990e-01 -6.31195009e-01
8.31746697e-01 9.52569187e-01 2.80334771e-01 -9.79558647e-01
-1.04298532e+00 -4.59038109e-01 -8.71280849e-01 1.02508806e-01
1.23639441e+00 -9.58822787e-01 -8.95704627e-01 5.53337097e-01
-1.49955177e+00 -3.87488574e-01 -3.18803966e-01 1.51813507e-01
-7.18635976e-01 9.00710523e-02 -7.77285814e-01 -9.83779803e-02
-5.55183470e-01 -1.14790905e+00 1.47529459e+00 2.95408994e-01
1.53437391e-01 -9.05650198e-01 -2.70204455e-01 -3.05324979e-02
3.48091274e-01 5.67432702e-01 8.50873590e-01 -3.98302227e-01
-9.95601416e-01 -6.70949072e-02 -6.19539440e-01 6.30365312e-01
-8.18376318e-02 9.38828066e-02 -8.98492396e-01 -3.57059479e-01
-8.12182501e-02 -2.89307922e-01 1.15441728e+00 5.41520774e-01
1.79527152e+00 -4.73647527e-02 -3.73338461e-01 1.52066100e+00
1.49950039e+00 -1.56125069e-01 5.77163577e-01 3.98498982e-01
9.54145730e-01 1.58023939e-01 -5.46658486e-02 1.20461442e-01
4.03945923e-01 3.09714496e-01 9.72140312e-01 -4.06279624e-01
1.85092315e-02 -1.03874430e-01 -1.62508398e-01 6.80274785e-01
-2.94065058e-01 2.30451345e-01 -9.79320586e-01 5.57397425e-01
-1.78676391e+00 -5.65662742e-01 -2.25299343e-01 1.72160375e+00
4.15920228e-01 3.01625967e-01 -2.83451546e-02 -3.21988910e-01
4.78220522e-01 2.54504293e-01 -7.41950989e-01 -4.46302056e-01
-5.73882051e-02 6.79310262e-01 6.57014906e-01 1.78208902e-01
-1.45104611e+00 1.17133009e+00 6.42687845e+00 7.48945832e-01
-1.09311235e+00 -1.16702445e-01 4.45193589e-01 1.35095611e-01
-1.59551948e-01 -2.93020993e-01 -7.47741759e-01 2.08777368e-01
5.25296748e-01 2.27965906e-01 7.85567537e-02 9.85121071e-01
-4.23341803e-02 5.24174929e-01 -9.75959063e-01 8.58234644e-01
-3.47139210e-01 -1.77209413e+00 2.90304691e-01 5.03062606e-02
8.20219994e-01 7.67887414e-01 1.53718069e-01 4.73957360e-02
7.11167514e-01 -1.49862707e+00 6.07892990e-01 2.43583798e-01
9.12132502e-01 -9.54946101e-01 4.08847153e-01 2.53625005e-01
-1.46251047e+00 2.89533082e-02 -8.16198766e-01 8.14352259e-02
-3.16113569e-02 3.89858246e-01 -5.79204977e-01 4.60305959e-01
1.13359749e+00 1.17433202e+00 -3.21745008e-01 1.24817038e+00
-2.58208811e-01 4.65400279e-01 -5.88901937e-01 1.70261338e-01
1.00245428e+00 -4.95397359e-01 4.88709599e-01 1.38991106e+00
3.54207784e-01 1.29289389e-01 2.18920201e-01 1.39526272e+00
-4.02391285e-01 -1.51519910e-01 -7.91998744e-01 4.21309412e-01
3.77496421e-01 1.46219683e+00 -8.81569803e-01 -2.88108774e-02
-7.17035472e-01 8.97057533e-01 6.27116203e-01 3.91924024e-01
-6.88055336e-01 -5.85469544e-01 1.07027018e+00 4.96430695e-02
8.03683639e-01 -5.66985130e-01 -4.48939264e-01 -8.76374245e-01
4.60656211e-02 -4.99545634e-01 4.93458770e-02 -7.17652321e-01
-1.40018332e+00 7.95852602e-01 -2.40079835e-01 -1.09056175e+00
3.35590392e-01 -1.00010812e+00 -8.83907080e-01 8.83880615e-01
-1.72084689e+00 -1.40561724e+00 -5.56329370e-01 8.44043493e-01
3.87898475e-01 1.53479457e-01 5.64603984e-01 -1.01227701e-01
-2.00153872e-01 2.78684765e-01 8.11439287e-03 6.85997665e-01
-1.32423947e-02 -1.44709003e+00 1.28143299e+00 5.83570302e-01
5.53946085e-02 7.14336395e-01 -1.11319236e-02 -3.71992141e-01
-1.15826964e+00 -1.63234401e+00 4.36139375e-01 -2.74295896e-01
6.69730663e-01 -5.24278939e-01 -1.08896041e+00 7.82387435e-01
8.60899463e-02 5.96082389e-01 2.66656727e-01 -1.60968095e-01
-6.48898900e-01 -1.19709438e-02 -9.93147671e-01 5.77133179e-01
1.50356638e+00 -5.85402012e-01 -6.05998397e-01 3.92824292e-01
1.21337724e+00 -6.34897172e-01 -9.16226625e-01 5.94025493e-01
2.52336621e-01 -1.00861061e+00 1.37480032e+00 -7.80695975e-01
5.35803914e-01 -2.68700987e-01 -5.25424108e-02 -1.30395722e+00
-6.29687309e-01 -6.14221811e-01 2.72068270e-02 5.40902972e-01
1.38571382e-01 -6.63252771e-01 9.16093886e-01 2.49200165e-01
-6.65906966e-01 -1.02862382e+00 -9.11084056e-01 -9.34741378e-01
6.29872680e-01 -4.60430056e-01 8.67388070e-01 6.85849309e-01
-6.61762238e-01 1.12928204e-01 3.56277138e-01 3.11617315e-01
8.16334009e-01 3.30271751e-01 6.90861344e-01 -1.53369999e+00
2.30420992e-01 -7.44880855e-01 -7.54766583e-01 -1.69475412e+00
3.90231073e-01 -1.06708980e+00 -8.94243568e-02 -1.70482099e+00
-2.03509316e-01 -5.99167824e-01 -1.10350549e-01 4.24206764e-01
9.77555141e-02 4.73569661e-01 5.98143600e-02 1.33815721e-01
-5.49788654e-01 7.68021822e-01 1.61093140e+00 -1.64981186e-01
-1.39969647e-01 3.73369642e-02 -4.85432327e-01 1.12179339e+00
6.71436369e-01 -1.28604263e-01 -2.40892366e-01 -9.24156368e-01
9.08396915e-02 -2.92544156e-01 7.79904604e-01 -1.10406578e+00
3.18608582e-01 1.11773863e-01 4.43283170e-01 -8.11123967e-01
2.91240096e-01 -7.94208229e-01 -1.95057034e-01 2.43290842e-01
-9.74349976e-02 2.18572468e-01 2.64010131e-01 5.65817118e-01
-3.25576425e-01 1.20268993e-01 8.46966505e-01 -2.57842988e-01
-8.09518933e-01 1.00325203e+00 1.72561407e-01 1.13789521e-01
8.64892185e-01 -2.11353749e-01 -1.81551248e-01 5.74917495e-02
-5.89186072e-01 4.32179242e-01 5.57024360e-01 3.62013221e-01
9.39464331e-01 -1.54479003e+00 -6.20849013e-01 2.49256343e-01
-1.16802111e-01 1.04212773e+00 -3.69135961e-02 5.22866189e-01
-1.18332267e+00 4.00702268e-01 -2.44079992e-01 -1.01266205e+00
-6.29802525e-01 3.10612142e-01 6.55216515e-01 1.39043078e-01
-1.25676465e+00 1.12737119e+00 5.10028183e-01 -6.21600211e-01
1.59059659e-01 -8.28989446e-01 2.34861430e-02 -3.34757119e-01
2.59671301e-01 1.01214834e-01 2.11788848e-01 -5.27630568e-01
-3.21922958e-01 9.77961779e-01 -4.90713678e-03 3.12895358e-01
1.52497184e+00 2.05324247e-01 -3.60457331e-01 -4.48867818e-03
1.54898393e+00 -6.11810982e-01 -1.80706036e+00 -4.79273647e-01
3.75555381e-02 -2.66563088e-01 1.34425849e-01 -1.95872456e-01
-1.38063753e+00 1.06210589e+00 1.86519355e-01 3.24647605e-01
9.27480102e-01 1.96557760e-01 1.15441310e+00 5.57958007e-01
3.45789760e-01 -7.14512765e-01 7.02945963e-02 1.04779088e+00
1.02468300e+00 -1.13794756e+00 -3.70847225e-01 -5.50025463e-01
-3.15077044e-02 1.42991877e+00 6.24983191e-01 -9.94532168e-01
1.04359162e+00 2.24871516e-01 -1.61788866e-01 -5.32544672e-01
-3.92476141e-01 -2.82495409e-01 5.12445211e-01 6.25944257e-01
1.26215026e-01 6.88929483e-02 1.73832595e-01 6.38904750e-01
-4.29461390e-01 -1.12968750e-01 9.59006548e-02 6.80435598e-01
-5.48463225e-01 -5.95499158e-01 -5.38368039e-02 5.17359197e-01
-2.89637417e-01 -7.20247105e-02 -1.75703377e-01 9.97463226e-01
1.95405960e-01 4.99196917e-01 7.06495821e-01 -2.36112580e-01
4.93906796e-01 -4.68312204e-01 2.37032548e-01 -5.48390150e-01
-6.71434700e-01 -4.20374423e-02 -4.04183865e-01 -9.83963013e-01
-3.82926196e-01 -5.11602700e-01 -1.59166002e+00 -3.90837133e-01
3.15643325e-02 -6.34162948e-02 6.12777948e-01 7.80282557e-01
4.61292177e-01 5.11840820e-01 5.36407232e-01 -1.09981406e+00
-5.84032536e-01 -6.19454503e-01 -4.74102587e-01 3.08152348e-01
5.08383155e-01 -5.94681680e-01 -2.88185813e-02 -2.17410073e-01]
|
[7.943106174468994, -3.634932279586792]
|
594df015-8eee-4f29-9db9-da464e29f5fe
|
flexible-end-to-end-dialogue-system-for
|
1709.04264
| null |
http://arxiv.org/abs/1709.04264v1
|
http://arxiv.org/pdf/1709.04264v1.pdf
|
Flexible End-to-End Dialogue System for Knowledge Grounded Conversation
|
In knowledge grounded conversation, domain knowledge plays an important role
in a special domain such as Music. The response of knowledge grounded
conversation might contain multiple answer entities or no entity at all.
Although existing generative question answering (QA) systems can be applied to
knowledge grounded conversation, they either have at most one entity in a
response or cannot deal with out-of-vocabulary entities. We propose a fully
data-driven generative dialogue system GenDS that is capable of generating
responses based on input message and related knowledge base (KB). To generate
arbitrary number of answer entities even when these entities never appear in
the training set, we design a dynamic knowledge enquirer which selects
different answer entities at different positions in a single response,
according to different local context. It does not rely on the representations
of entities, enabling our model deal with out-of-vocabulary entities. We
collect a human-human conversation data (ConversMusic) with knowledge
annotations. The proposed method is evaluated on CoversMusic and a public
question answering dataset. Our proposed GenDS system outperforms baseline
methods significantly in terms of the BLEU, entity accuracy, entity recall and
human evaluation. Moreover,the experiments also demonstrate that GenDS works
better even on small datasets.
|
['Xuezheng Peng', 'Kaixiang Mo', 'Wenya Zhu', 'Zhangbin Zhu', 'Qiang Yang', 'Yu Zhang']
|
2017-09-13
| null | null | null | null |
['generative-question-answering']
|
['natural-language-processing']
|
[-6.46274490e-03 6.98831558e-01 2.36416951e-01 -2.83529103e-01
-1.12836671e+00 -6.98075116e-01 7.20457911e-01 -6.24395050e-02
-3.59209001e-01 1.15387988e+00 6.68990910e-01 9.26555321e-02
3.76002565e-02 -1.26981080e+00 -5.90500236e-01 -4.32384640e-01
4.64164734e-01 1.33786130e+00 5.82012653e-01 -8.73342752e-01
-3.62807252e-02 -3.62936884e-01 -1.30787337e+00 6.87183738e-01
1.06192148e+00 7.74258971e-01 2.85042584e-01 7.49849379e-01
-5.47241092e-01 1.12132382e+00 -1.08445740e+00 -7.73842216e-01
-2.39062622e-01 -7.81172693e-01 -1.50665760e+00 -1.77439526e-01
-1.31421402e-01 -2.96222772e-02 -1.06219292e-01 6.31297410e-01
9.05289412e-01 3.04264635e-01 5.86071551e-01 -1.10484958e+00
-8.96789968e-01 1.22687960e+00 3.50926638e-01 -5.99418469e-02
7.67332971e-01 -5.54907769e-02 1.06154478e+00 -7.90802300e-01
8.31244230e-01 1.42639506e+00 3.46928507e-01 8.34136367e-01
-7.65917480e-01 -3.92021507e-01 1.72459893e-02 4.11844343e-01
-1.49464560e+00 -3.29718858e-01 6.66049600e-01 -6.75318018e-02
8.55027616e-01 5.08418262e-01 4.48369116e-01 1.18548608e+00
-4.04009610e-01 7.49697626e-01 7.89795041e-01 -3.99884611e-01
2.19188020e-01 3.80972475e-01 2.00924009e-01 2.70526528e-01
-1.69465452e-01 -6.27911329e-01 -7.60524035e-01 -4.50583309e-01
4.66908276e-01 -5.59880793e-01 -5.54594636e-01 3.21239196e-02
-1.34579742e+00 1.03824246e+00 1.46271169e-01 3.14015806e-01
-5.05092740e-01 -2.15577185e-01 3.11801136e-01 2.54756272e-01
9.27989408e-02 6.11969292e-01 -4.24137741e-01 -2.67985374e-01
-2.04803154e-01 5.61919570e-01 1.46105099e+00 1.33262277e+00
7.78964520e-01 -3.21003258e-01 -6.29161537e-01 9.91732180e-01
-3.89051102e-02 5.56634545e-01 6.95370138e-01 -7.52898932e-01
6.45941675e-01 1.07011557e+00 4.35465157e-01 -1.04169929e+00
-2.89783716e-01 -3.86009961e-01 -8.45158637e-01 -9.15889561e-01
7.10756108e-02 -5.47796547e-01 -3.74715894e-01 1.76739240e+00
7.26451278e-01 -3.74849583e-03 8.20341885e-01 8.19358826e-01
1.77583289e+00 8.26670051e-01 8.41223374e-02 -2.39298984e-01
1.68607819e+00 -9.34331954e-01 -9.32648301e-01 -2.48018965e-01
4.90329117e-01 -5.83966434e-01 9.57338870e-01 5.57137951e-02
-8.57698619e-01 -4.28592443e-01 -5.30502081e-01 -1.16038024e-01
-3.19262683e-01 4.78001721e-02 3.54484111e-01 3.72335851e-01
-7.65703797e-01 -2.84432650e-01 -4.49692793e-02 -3.86556238e-01
-4.15865988e-01 1.79602295e-01 -1.05280489e-01 1.10088639e-01
-1.92736280e+00 8.19029093e-01 7.16272354e-01 1.17177576e-01
-8.00007463e-01 -2.73722053e-01 -7.31231689e-01 1.02651551e-01
7.38593757e-01 -1.04969394e+00 1.50534439e+00 -7.59087801e-01
-1.62043738e+00 5.09763122e-01 -1.11861609e-01 -4.49866444e-01
2.65773743e-01 -1.28362089e-01 -6.52372539e-01 2.28419676e-01
-2.85731740e-02 6.07790887e-01 4.04626071e-01 -1.20903170e+00
-6.44362926e-01 5.43953441e-02 6.93265140e-01 7.03315496e-01
-8.36298540e-02 -2.24141285e-01 -4.68167961e-01 -2.78518796e-01
-1.45615548e-01 -9.00956750e-01 5.32872044e-02 -1.11778152e+00
-5.25494754e-01 -6.23218000e-01 5.29958069e-01 -6.13352418e-01
1.41034865e+00 -1.64874244e+00 1.15916803e-01 -1.12440266e-01
-1.06467210e-01 3.20307642e-01 9.71092880e-02 9.14497137e-01
6.06699049e-01 2.96903588e-02 -1.11636175e-02 1.83805674e-01
9.90847945e-02 3.55590135e-01 -4.16325957e-01 -5.73511362e-01
5.06756790e-02 1.07437730e+00 -1.03919148e+00 -8.09213459e-01
-3.25742096e-01 3.01463574e-01 -6.03777289e-01 4.91007745e-01
-6.74718678e-01 5.46901047e-01 -8.53933096e-01 3.26119423e-01
2.59603322e-01 -4.84578907e-01 4.01399255e-01 -2.45710522e-01
3.69816571e-01 5.54045081e-01 -1.28357923e+00 1.68153787e+00
-6.79732442e-01 1.40042841e-01 -3.06560785e-01 -4.28623766e-01
9.84105051e-01 8.84589136e-01 -1.37174111e-02 -5.85909963e-01
1.37076378e-02 2.10739970e-01 -5.08995913e-02 -7.73747921e-01
1.01174045e+00 -1.24401167e-01 -5.70047081e-01 5.04864991e-01
2.34509885e-01 -1.28492445e-01 2.86064148e-01 5.89908719e-01
1.13221323e+00 -2.43778840e-01 3.50691527e-01 1.11519285e-01
7.81390548e-01 3.55965078e-01 5.08601904e-01 7.77239382e-01
3.26242775e-01 4.26722467e-01 4.63422775e-01 1.00126639e-01
-5.21079481e-01 -6.99243009e-01 2.40103245e-01 1.36962962e+00
4.68069583e-01 -4.76537704e-01 -8.63932669e-01 -7.84071445e-01
-3.93670976e-01 9.88390148e-01 -4.27394509e-01 -2.21367002e-01
-6.87224627e-01 -5.00092149e-01 6.76724970e-01 1.56856492e-01
7.61371791e-01 -1.56124318e+00 -2.99127251e-01 4.58243042e-01
-1.19931650e+00 -1.28774655e+00 -4.14014757e-01 -3.72752905e-01
-2.11813569e-01 -1.23886180e+00 -4.90943670e-01 -9.82920885e-01
4.42148566e-01 -4.05112095e-03 1.58180642e+00 -1.22760005e-01
3.95768881e-01 6.98169887e-01 -8.88098598e-01 -3.55608910e-01
-8.08340549e-01 5.73845387e-01 -4.74492788e-01 2.02672392e-01
6.39176130e-01 -3.25320095e-01 -7.50199735e-01 6.81298137e-01
-9.52262044e-01 1.98852450e-01 3.66676956e-01 8.80253315e-01
3.40058148e-01 -2.44193703e-01 1.29627621e+00 -1.09151661e+00
1.20104587e+00 -8.33436191e-01 9.32928696e-02 7.18918979e-01
-1.61863223e-01 1.41494945e-01 5.40942311e-01 -5.17867148e-01
-1.35154045e+00 -3.34694773e-01 -2.33812705e-01 2.07172513e-01
7.59577453e-02 6.23005033e-01 -3.80188972e-01 5.27806520e-01
1.04146779e+00 3.71275872e-01 -5.28313041e-01 -2.61929840e-01
5.64410269e-01 1.04578626e+00 5.66076636e-01 -8.31997633e-01
4.77367193e-01 -1.04994446e-01 -6.04422629e-01 -5.37457883e-01
-7.88710177e-01 -5.56573808e-01 -1.98855251e-01 -2.65205771e-01
8.28229368e-01 -1.03568745e+00 -8.91127944e-01 2.00326174e-01
-1.34761381e+00 -1.15811162e-01 -3.33285838e-01 3.10214877e-01
-5.28640211e-01 8.95923674e-02 -3.97177279e-01 -9.34232056e-01
-7.54968643e-01 -7.59847105e-01 1.01245487e+00 4.12836522e-01
-3.58857870e-01 -8.58669400e-01 1.71663895e-01 6.73799217e-01
3.53882134e-01 2.37643257e-01 9.66344178e-01 -1.29321837e+00
-6.26771450e-01 -1.35676473e-01 3.77623439e-01 6.51812479e-02
1.39392957e-01 -5.42990386e-01 -7.31206596e-01 9.84957293e-02
-2.31397510e-01 -7.30609655e-01 3.91130507e-01 -4.03196573e-01
4.89089727e-01 -8.28572094e-01 -1.32068694e-01 -2.42872030e-01
1.00329947e+00 2.98573643e-01 7.37136900e-01 6.47265539e-02
4.66471821e-01 7.86773086e-01 8.52894604e-01 3.58543634e-01
1.02755165e+00 8.14691246e-01 2.02912524e-01 4.76205051e-01
-5.83156347e-02 -5.73172867e-01 1.21027119e-01 1.18895197e+00
-2.14982759e-02 -7.98469543e-01 -6.98685706e-01 9.71161127e-01
-2.09960175e+00 -1.00174153e+00 -1.16049029e-01 2.02416420e+00
1.45684254e+00 -2.53073394e-01 1.25093743e-01 -2.31686771e-01
9.05571282e-01 -1.97658196e-01 -3.82546097e-01 -1.70579880e-01
-3.63768518e-01 2.15770379e-01 -1.58768788e-01 5.36891818e-01
-5.29561937e-01 1.13422740e+00 5.19640827e+00 8.76028478e-01
-6.99354053e-01 2.94386417e-01 1.56918630e-01 1.64492503e-01
-6.59300625e-01 5.55833802e-02 -1.08065140e+00 4.37435865e-01
9.34921205e-01 -6.07082367e-01 1.83017269e-01 7.82354116e-01
-1.97356477e-01 -1.02343030e-01 -8.74536932e-01 8.00286055e-01
1.87830195e-01 -1.35650980e+00 4.02283251e-01 -4.25564319e-01
6.17063940e-01 -4.68338937e-01 -4.56223100e-01 8.72111261e-01
6.65641665e-01 -8.88370395e-01 4.58305955e-01 7.34002173e-01
4.92646188e-01 -7.38521099e-01 1.01423299e+00 8.09549510e-01
-9.07527089e-01 1.43041790e-01 -2.84544855e-01 2.80191869e-01
4.10452276e-01 1.42713338e-01 -1.57491744e+00 8.72228682e-01
4.48482454e-01 -2.02060208e-01 -2.02652082e-01 6.47548974e-01
-5.33474207e-01 6.40002847e-01 -2.63440996e-01 -5.12410283e-01
6.53467774e-02 1.59584612e-01 4.83740628e-01 1.11238825e+00
4.18425441e-01 6.88448370e-01 9.73395854e-02 6.28547668e-01
-3.50619912e-01 6.17618084e-01 -2.48247653e-01 -2.89191445e-03
1.02301276e+00 1.21668661e+00 -2.35222787e-01 -5.36651015e-01
-1.36277840e-01 9.01062250e-01 1.37993291e-01 3.60805154e-01
-7.37818301e-01 -4.89949495e-01 9.05548856e-02 1.27753809e-01
3.03271830e-01 1.56714961e-01 4.55977052e-01 -1.12177718e+00
1.34885803e-01 -1.25304019e+00 5.85277319e-01 -9.58351314e-01
-1.11244988e+00 9.25211072e-01 1.49944976e-01 -1.02366889e+00
-1.00718188e+00 1.57365948e-01 -3.90104353e-01 7.38721013e-01
-1.22141182e+00 -1.24862671e+00 -6.62166774e-01 8.74020398e-01
6.40350580e-01 -8.72476473e-02 1.18071902e+00 3.22818846e-01
-2.77658105e-01 5.76783836e-01 -3.71569097e-01 2.99603850e-01
9.09115851e-01 -1.14157057e+00 2.73456007e-01 3.34339470e-01
2.21986488e-01 8.11291158e-01 8.09343338e-01 -7.53697217e-01
-1.21121466e+00 -9.84405875e-01 1.22737002e+00 -5.47219157e-01
2.44261220e-01 -3.12461287e-01 -1.06329644e+00 6.75806820e-01
5.82616150e-01 -5.58903873e-01 9.47574615e-01 1.48940623e-01
-2.39001550e-02 1.11741498e-01 -1.11518502e+00 5.23820698e-01
1.03755569e+00 -4.51220185e-01 -1.03234351e+00 4.76852924e-01
1.07121253e+00 -7.80282438e-01 -8.28415394e-01 5.29352725e-01
1.19499370e-01 -6.02675259e-01 6.50050998e-01 -7.62069941e-01
6.15461320e-02 -4.68342215e-01 -1.78864598e-01 -1.33380377e+00
1.05160959e-02 -8.35314095e-01 -4.82579976e-01 1.68471694e+00
7.36050129e-01 -4.58478302e-01 5.00968456e-01 7.13989198e-01
-2.10076362e-01 -5.27000725e-01 -9.31569040e-01 -4.38153535e-01
-3.02929103e-01 -2.86797471e-02 1.06281638e+00 8.93732250e-01
2.22103879e-01 1.03192103e+00 -5.03516078e-01 2.23792285e-01
-6.38880059e-02 3.73455435e-01 1.17378628e+00 -1.09078336e+00
-4.07411635e-01 2.38410294e-01 -1.17130116e-01 -1.32525313e+00
1.39798946e-03 -6.28190100e-01 1.15064777e-01 -1.89645112e+00
7.31239468e-02 -5.35449445e-01 2.02862933e-01 5.59689045e-01
-4.57499593e-01 -3.84468921e-02 -1.83274880e-01 6.88934848e-02
-9.32051539e-01 8.05739880e-01 1.26105809e+00 -9.68904644e-02
-3.73242587e-01 1.72514599e-02 -9.26105917e-01 3.84379953e-01
6.96013808e-01 -4.88107234e-01 -7.82532752e-01 -2.30528414e-01
5.66450834e-01 4.58628863e-01 2.94050485e-01 -6.72336698e-01
5.18946767e-01 -2.24133044e-01 -2.62677014e-01 -5.49816489e-01
4.47254181e-01 -4.48954970e-01 4.57686394e-01 -1.54804081e-01
-4.11468506e-01 -4.60720919e-02 -8.52044448e-02 4.86698002e-01
-5.40090680e-01 -4.81647849e-01 2.90726274e-01 -3.08333486e-01
-8.04536343e-01 -4.29592505e-02 -1.83773339e-01 6.62797868e-01
8.34715128e-01 -1.40758674e-03 -5.82901239e-01 -9.23707306e-01
-7.19168365e-01 4.27104890e-01 6.20870665e-02 7.10261583e-01
4.14929062e-01 -1.48764670e+00 -9.24193084e-01 -4.25159633e-01
4.09397066e-01 3.81190747e-01 6.32854640e-01 4.45536822e-01
-2.93857932e-01 4.76782799e-01 1.75807506e-01 -2.78871179e-01
-1.13595045e+00 2.99553216e-01 3.84672433e-01 -4.50759828e-01
-2.25247383e-01 9.03321087e-01 2.27206483e-01 -6.69319510e-01
9.09871161e-02 -1.74227171e-02 -6.49880111e-01 2.36064687e-01
5.13132155e-01 9.91075337e-02 1.50805963e-02 -7.17594147e-01
-6.99773580e-02 2.13497698e-01 2.87930132e-03 -3.18370700e-01
8.52325678e-01 -3.11296850e-01 -2.89622825e-02 4.21104610e-01
6.86568975e-01 3.37407470e-01 -6.06759846e-01 -6.13799453e-01
-1.66567802e-01 -1.22825049e-01 -5.44396579e-01 -1.07747722e+00
-5.95602214e-01 3.50567013e-01 7.22855106e-02 4.74513203e-01
9.69675362e-01 3.30402106e-01 1.12044394e+00 7.39204824e-01
6.76842749e-01 -1.10544789e+00 2.21349895e-01 7.36348033e-01
1.18139517e+00 -1.04625165e+00 -4.36394334e-01 -4.67358679e-01
-1.25503981e+00 7.25152552e-01 8.22754204e-01 4.22160417e-01
1.01685807e-01 -2.25242630e-01 2.68708408e-01 -3.23629618e-01
-9.69359517e-01 -4.62368876e-01 2.12276176e-01 6.21375620e-01
3.26070040e-01 1.76689401e-02 -5.96910000e-01 9.93634820e-01
-6.75794303e-01 -1.23343974e-01 4.36341465e-01 7.18615472e-01
-6.59837663e-01 -1.24594557e+00 -6.88633770e-02 -6.91393996e-03
-3.14206451e-01 -1.94969460e-01 -8.70701373e-01 7.81630099e-01
5.59520200e-02 1.46192193e+00 -2.58827627e-01 -4.42481190e-01
5.87735474e-01 4.53853428e-01 1.83473185e-01 -7.73341656e-01
-9.74237144e-01 -4.04293150e-01 7.90028870e-01 8.58559012e-02
-4.94860440e-01 -3.85583073e-01 -1.47099328e+00 -1.92453101e-01
-6.12071812e-01 1.04714990e+00 2.49300838e-01 7.63412654e-01
6.69027507e-01 3.91717762e-01 5.38802981e-01 1.60075694e-01
-2.85145313e-01 -1.38536441e+00 -2.11313739e-01 4.49317306e-01
-9.91885737e-02 -4.35586005e-01 -8.72303769e-02 9.70901847e-02]
|
[12.443760871887207, 8.071685791015625]
|
0182e204-f31d-41d7-8a34-23cb0bd87f13
|
semantic-neighborhoods-as-hypergraphs
| null | null |
https://aclanthology.org/P13-2040
|
https://aclanthology.org/P13-2040.pdf
|
Semantic Neighborhoods as Hypergraphs
| null |
['Pallavi Choudhury', 'Chris Quirk']
|
2013-08-01
| null | null | null |
acl-2013-8
|
['video-description']
|
['computer-vision']
|
[-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]
|
[-7.294008731842041, 3.632223129272461]
|
19efa00c-7b43-40c4-a012-7903e811af1e
|
transferring-procedural-knowledge-across
|
2304.13867
| null |
https://arxiv.org/abs/2304.13867v1
|
https://arxiv.org/pdf/2304.13867v1.pdf
|
Transferring Procedural Knowledge across Commonsense Tasks
|
Stories about everyday situations are an essential part of human communication, motivating the need to develop AI agents that can reliably understand these stories. Despite the long list of supervised methods for story completion and procedural understanding, current AI has no mechanisms to automatically track and explain procedures in unseen stories. To bridge this gap, we study the ability of AI models to transfer procedural knowledge to novel narrative tasks in a transparent manner. We design LEAP: a comprehensive framework that integrates state-of-the-art modeling architectures, training regimes, and augmentation strategies based on both natural and synthetic stories. To address the lack of densely annotated training data, we devise a robust automatic labeler based on few-shot prompting to enhance the augmented data. Our experiments with in- and out-of-domain tasks reveal insights into the interplay of different architectures, training regimes, and augmentation strategies. LEAP's labeler has a clear positive impact on out-of-domain datasets, while the resulting dense annotation provides native explainability.
|
['Kaixin Ma', 'Filip Ilievski', 'Yifan Jiang']
|
2023-04-26
| null | null | null | null |
['story-completion']
|
['natural-language-processing']
|
[ 4.32310164e-01 6.78189337e-01 -1.68748215e-01 -4.89893347e-01
-5.52713037e-01 -6.78569198e-01 1.12459612e+00 1.26284957e-01
-1.10942528e-01 6.84703708e-01 7.81138897e-01 -1.57404572e-01
1.09988086e-01 -9.77266610e-01 -7.20665753e-01 2.58776098e-01
2.13426694e-01 6.59093320e-01 2.12015957e-01 -5.30317664e-01
1.29126087e-01 1.45397279e-02 -1.32598436e+00 8.77065659e-01
7.49253213e-01 3.90561730e-01 1.03152677e-01 6.59943759e-01
-2.52770007e-01 1.43894994e+00 -5.75148344e-01 -5.42501926e-01
-9.50824618e-02 -6.64481640e-01 -1.08947825e+00 2.58446604e-01
-9.28551033e-02 -4.19620574e-01 -5.91219664e-01 2.41467476e-01
1.16824321e-01 1.29225358e-01 8.28634441e-01 -1.12190461e+00
-9.08280373e-01 1.18431270e+00 -7.83099309e-02 1.40694335e-01
5.71726680e-01 3.12784702e-01 9.56338644e-01 -5.95451355e-01
1.16167986e+00 8.44177127e-01 9.39537823e-01 9.06692088e-01
-1.45511281e+00 -3.05350929e-01 8.48807395e-02 1.78653955e-01
-1.01213920e+00 -6.07844293e-01 9.28535998e-01 -7.58151472e-01
8.33620548e-01 2.77409971e-01 9.61415827e-01 1.77977681e+00
-4.18845654e-01 9.31201994e-01 9.14452255e-01 -4.50688571e-01
2.01898143e-01 3.25537622e-01 2.89179459e-02 6.40437901e-01
4.74724174e-02 2.13154972e-01 -1.05332303e+00 1.46979094e-01
9.49902058e-01 -2.37801746e-01 -2.39046633e-01 -2.82252103e-01
-1.35885727e+00 7.35778451e-01 8.95359144e-02 4.40622181e-01
-1.72866091e-01 3.47616643e-01 5.27894795e-01 1.18139155e-01
5.12358367e-01 1.24940515e+00 -3.34168732e-01 -6.95670903e-01
-8.86191905e-01 6.20608807e-01 9.57521915e-01 1.13850725e+00
4.73527789e-01 1.13486357e-01 -4.68300492e-01 6.67766750e-01
-2.19561279e-01 -2.13826656e-01 2.31871784e-01 -1.15288389e+00
5.02171695e-01 9.22719836e-01 1.76961541e-01 -8.50951731e-01
-4.99351174e-01 -5.49048245e-01 -4.46939677e-01 -4.96337749e-02
3.90271962e-01 3.37504596e-02 -5.39436936e-01 1.82610011e+00
3.77828740e-02 1.49864620e-02 1.42603651e-01 5.98120272e-01
1.03429258e+00 5.03019214e-01 2.07101211e-01 2.03363970e-01
1.26229537e+00 -1.12224340e+00 -5.74609816e-01 -7.36016929e-01
1.03494275e+00 -1.98061943e-01 1.60198116e+00 8.44381005e-02
-1.13553953e+00 -3.95573676e-01 -1.05954349e+00 -2.61261344e-01
-2.94002742e-01 1.42301410e-01 9.46301639e-01 3.38163793e-01
-5.96074224e-01 5.80617428e-01 -8.19302142e-01 -4.45577562e-01
8.26717615e-01 -3.11143667e-01 -5.19719064e-01 -5.92722446e-02
-8.67218316e-01 1.12097394e+00 6.00405276e-01 -3.27882409e-01
-1.16859090e+00 -1.14954829e+00 -1.12949514e+00 4.19492014e-02
7.18970180e-01 -7.82179058e-01 1.65560973e+00 -9.43550229e-01
-1.27356219e+00 1.11590195e+00 5.64748906e-02 -5.64811766e-01
7.18735993e-01 -3.56570125e-01 -4.36602719e-02 -6.29966557e-02
2.69242793e-01 7.18177021e-01 2.74477214e-01 -1.40665817e+00
-1.54788584e-01 1.64835826e-01 4.13991481e-01 7.59627204e-04
-2.43092120e-01 -9.39077064e-02 2.72394251e-02 -9.07478809e-01
-1.62899703e-01 -7.18895316e-01 -8.98785144e-03 -1.75230931e-02
-4.53276396e-01 1.62313789e-01 4.62478936e-01 -5.67981184e-01
1.17797387e+00 -2.17136836e+00 1.64262459e-01 -4.70601857e-01
2.57632881e-01 -1.33237183e-01 -6.86477274e-02 6.81185067e-01
8.56946930e-02 3.40914100e-01 -5.40138245e-01 -5.42037845e-01
3.07541966e-01 2.50174254e-01 -5.20671308e-01 -5.92593811e-02
5.38442612e-01 1.09852755e+00 -1.00876343e+00 -5.05581915e-01
1.34257048e-01 2.92352676e-01 -8.45421612e-01 5.04227638e-01
-7.10414410e-01 6.69262528e-01 -3.73189032e-01 3.79238963e-01
-2.03539148e-01 -3.69633377e-01 6.37174845e-02 1.68096587e-01
9.04019028e-02 6.57496154e-01 -7.77454078e-01 2.17723680e+00
-4.94869053e-01 1.00551474e+00 -4.83465642e-01 -6.99600160e-01
8.03497851e-01 5.50829053e-01 8.05680826e-02 -5.16090751e-01
1.28308296e-01 -1.23848177e-01 -1.12227336e-01 -7.82701373e-01
8.00174177e-01 -2.78864473e-01 -4.98945445e-01 8.78844917e-01
1.69129491e-01 -6.25529587e-01 4.23773050e-01 3.99056345e-01
1.35958433e+00 5.27313650e-01 4.11387920e-01 2.66552605e-02
-8.94736871e-02 7.03745484e-01 2.84710169e-01 8.47769320e-01
-3.04129999e-02 7.87280738e-01 5.24295509e-01 -8.08000386e-01
-1.28456485e+00 -7.63810098e-01 7.54842535e-02 1.34589839e+00
-8.99294242e-02 -6.95281386e-01 -8.71539593e-01 -4.86054361e-01
-3.84665519e-01 1.30483341e+00 -1.04947519e+00 -3.13844830e-01
-5.06779373e-01 -3.84682119e-01 8.80055189e-01 6.90646708e-01
4.69893217e-01 -1.47848427e+00 -1.09845865e+00 3.99826407e-01
-4.81827557e-01 -1.38231325e+00 -4.19239961e-02 4.00176570e-02
-4.79391485e-01 -1.04043210e+00 -1.05985001e-01 -5.38049400e-01
4.24873203e-01 2.39514001e-02 1.62472069e+00 1.21620998e-01
-5.34829199e-02 4.10970181e-01 -6.32269263e-01 -6.04243100e-01
-8.64801466e-01 3.20763350e-01 -4.94096249e-01 -4.72605407e-01
1.80730924e-01 -9.08047259e-01 -1.09903455e-01 1.38956562e-01
-8.23059201e-01 1.04144239e+00 2.92144209e-01 6.61383927e-01
1.78104684e-01 -3.56202066e-01 5.18396735e-01 -1.37936711e+00
8.25785577e-01 -4.65266824e-01 -3.35657671e-02 1.31303728e-01
-1.93617702e-01 1.94246635e-01 3.73964489e-01 -5.61468482e-01
-1.29838848e+00 -1.69068888e-01 1.48780301e-01 1.21274248e-01
-3.86889577e-01 6.86092556e-01 3.62953506e-02 4.48535979e-01
1.12378645e+00 1.71945155e-01 -5.03893971e-01 -2.54546404e-01
8.63840997e-01 2.53903866e-01 9.32350993e-01 -9.42923486e-01
6.52225614e-01 4.85433072e-01 -4.31007415e-01 -6.33397639e-01
-1.24031043e+00 -6.76419772e-03 -6.28800929e-01 -3.28282654e-01
9.38732743e-01 -8.53765070e-01 -3.56373340e-01 6.96586892e-02
-1.37610984e+00 -8.69531691e-01 -8.70683074e-01 1.95937380e-01
-1.08201540e+00 -3.13596636e-01 -6.96886897e-01 -6.65996790e-01
-8.81216452e-02 -7.24033237e-01 8.58354807e-01 6.74427301e-02
-1.27459097e+00 -7.62670398e-01 3.00362796e-01 6.08390033e-01
3.28863114e-01 7.34651864e-01 1.11623406e+00 -8.51762116e-01
-6.48075879e-01 -1.51376843e-01 -8.28542784e-02 -2.04783440e-01
-1.76797628e-01 -2.04981774e-01 -1.12280810e+00 4.89304543e-01
-5.93186431e-02 -7.79622197e-01 5.42876780e-01 -2.36785803e-02
9.60576177e-01 -4.16626781e-01 -2.81328589e-01 5.17479777e-01
9.82259929e-01 -1.55737083e-02 6.89920545e-01 6.77725613e-01
7.16386795e-01 7.20231354e-01 3.55435699e-01 4.78051394e-01
7.03139722e-01 4.97328550e-01 1.10395022e-01 1.31958947e-01
-2.13225439e-01 -7.16706812e-01 -4.90895705e-03 5.66144645e-01
-9.04906169e-02 -1.91892415e-01 -1.18556345e+00 8.35595131e-01
-1.94805348e+00 -1.23262453e+00 6.53228685e-02 1.47009826e+00
1.24950933e+00 4.97899115e-01 -2.92096715e-02 2.51761764e-01
1.85504988e-01 3.94007266e-01 -3.02428633e-01 -4.60129231e-01
-1.28791898e-01 -4.23839362e-03 -1.58729717e-01 3.09392989e-01
-9.79038000e-01 1.08968127e+00 6.62446547e+00 5.23595273e-01
-7.04898834e-01 2.71051615e-01 5.82376003e-01 -2.13389471e-01
-5.46527565e-01 4.71186526e-02 -1.61472708e-01 7.30226263e-02
7.79978991e-01 -1.59515619e-01 5.31973898e-01 1.01327550e+00
1.99315190e-01 7.90319964e-02 -1.71084785e+00 6.15591109e-01
1.28959343e-01 -2.04995060e+00 1.18026160e-01 -2.50301242e-01
8.57832611e-01 -3.10457915e-01 -3.18690687e-01 6.48841202e-01
5.71942687e-01 -1.34638309e+00 1.31124890e+00 5.96785188e-01
8.69090259e-01 -3.38507742e-01 2.74657577e-01 6.47307932e-01
-8.45728874e-01 1.20209441e-01 2.56362557e-01 -6.33111596e-01
3.38297546e-01 2.31982432e-02 -1.00588548e+00 4.88380305e-02
2.91448474e-01 7.86523819e-01 -4.44513530e-01 6.10311866e-01
-6.68958127e-01 6.42515242e-01 1.68325044e-02 6.23521768e-02
1.70480266e-01 1.83658019e-01 5.94304979e-01 1.36377478e+00
-6.76063597e-02 4.95623112e-01 1.04971133e-01 1.44729149e+00
-2.91118652e-01 -2.92662859e-01 -8.36162210e-01 -5.21186888e-01
6.59870565e-01 8.88224959e-01 -6.06681406e-01 -3.28322709e-01
-3.29897910e-01 8.39800119e-01 6.47057831e-01 2.45141849e-01
-7.46984065e-01 1.65501624e-01 3.44192713e-01 4.76375520e-01
5.12868054e-02 -3.69329065e-01 -9.99878526e-01 -1.04364502e+00
2.61349771e-02 -1.01209617e+00 2.85536256e-02 -1.22374928e+00
-9.90180850e-01 6.07549369e-01 1.42846882e-01 -8.86140108e-01
-5.44696093e-01 -1.44582495e-01 -7.42901981e-01 3.66592467e-01
-1.14219081e+00 -1.54310632e+00 -6.01722300e-01 1.75239876e-01
9.32323456e-01 -6.75337240e-02 1.15467942e+00 -1.08424023e-01
-3.33927423e-01 3.33630145e-01 -4.29697245e-01 3.35577607e-01
4.16796267e-01 -1.05325127e+00 7.82627165e-01 7.23753095e-01
3.97193789e-01 3.68801028e-01 1.08562469e+00 -4.96326327e-01
-9.29876506e-01 -1.03074777e+00 6.66351736e-01 -1.07703960e+00
6.97013378e-01 -6.18243039e-01 -1.02302647e+00 1.12416768e+00
1.24068655e-01 -4.11537170e-01 8.25060010e-01 3.74932200e-01
-7.89609909e-01 5.06159067e-01 -8.50633740e-01 8.36976111e-01
1.40272498e+00 -9.10787046e-01 -1.05884326e+00 2.87987828e-01
1.03705752e+00 -4.62630719e-01 -6.21029019e-01 2.55341440e-01
6.07850194e-01 -9.43781495e-01 6.81980491e-01 -8.86723638e-01
1.43002987e+00 -2.81911902e-02 -3.48364897e-02 -1.30170274e+00
-1.27886087e-01 -6.72022760e-01 -1.72378063e-01 1.41250777e+00
7.25963175e-01 -6.47013960e-03 9.14200783e-01 8.75323951e-01
-5.31458020e-01 -7.54281759e-01 -4.47328150e-01 -4.12194341e-01
-1.58303186e-01 -8.70794833e-01 5.91956377e-01 1.17811501e+00
4.79464561e-01 6.24424398e-01 -3.46875519e-01 -3.26826692e-01
1.31058544e-01 -3.38823199e-02 1.17007446e+00 -1.21128678e+00
-5.12566686e-01 -4.50104624e-01 -9.31343883e-02 -6.95276201e-01
2.67753541e-01 -7.36613750e-01 9.66197699e-02 -1.91802025e+00
5.68445802e-01 -4.10173148e-01 4.55483496e-01 6.44630551e-01
5.01204059e-02 3.92226204e-02 1.21542171e-01 2.01158717e-01
-5.69540203e-01 5.94223917e-01 1.10441566e+00 -8.20145980e-02
-2.74382085e-01 -4.47489083e-01 -1.01940680e+00 1.06229675e+00
6.69078231e-01 -5.64547479e-01 -5.33638239e-01 -9.18512404e-01
5.12898207e-01 8.48032311e-02 5.26609540e-01 -1.21365893e+00
1.32616967e-01 -3.51071477e-01 2.60820128e-02 -1.63681477e-01
6.14343286e-01 -5.50486267e-01 3.16207737e-01 -3.64444777e-03
-8.31996262e-01 -8.62067640e-02 5.13098419e-01 5.44524133e-01
-1.27328426e-01 -2.17510462e-01 4.36894685e-01 -4.15947527e-01
-7.37540066e-01 -1.51579380e-01 -5.24379551e-01 4.96258914e-01
1.07384229e+00 -5.15312970e-01 -7.17262506e-01 -5.47741413e-01
-7.07493007e-01 8.86245910e-03 7.37846553e-01 4.19756770e-01
3.96336317e-01 -1.19933164e+00 -9.78073120e-01 -1.13590360e-02
5.68045497e-01 1.99532628e-01 1.60181776e-01 2.31467500e-01
-6.07417762e-01 1.00906610e-01 -3.78040940e-01 -3.09443921e-01
-7.99747646e-01 4.58854795e-01 2.53392488e-01 -3.86612386e-01
-7.52423823e-01 9.93820727e-01 2.42984295e-01 -5.35373092e-01
-2.61345115e-02 -3.00789297e-01 -1.66707262e-01 -1.53180778e-01
5.07772744e-01 6.70107827e-02 -2.23623589e-01 -3.66195470e-01
6.19160458e-02 -2.37075556e-02 -4.59970068e-03 -4.36905652e-01
1.70543540e+00 1.36488512e-01 2.81025410e-01 7.52924919e-01
3.92345309e-01 -2.05855250e-01 -1.51035273e+00 -1.23583764e-01
2.18183681e-01 -3.33632231e-01 -2.77661532e-01 -1.03641319e+00
-2.47170761e-01 8.52580428e-01 -3.23337793e-01 2.78423965e-01
6.04572535e-01 3.26483518e-01 7.45668828e-01 3.69900554e-01
5.38266361e-01 -8.11379313e-01 3.64373624e-01 6.49440885e-01
1.17203593e+00 -1.11259520e+00 -5.08326888e-02 -6.02758646e-01
-1.00563419e+00 7.48371303e-01 7.09095895e-01 -2.08032075e-02
-2.98233349e-02 4.16879505e-01 -2.22489983e-01 -4.64147568e-01
-9.72384691e-01 4.11126614e-02 7.61995465e-02 7.51724303e-01
5.72243929e-01 1.61241859e-01 1.34520039e-01 1.14742744e+00
-7.03353047e-01 1.97540760e-01 8.51122975e-01 1.12104833e+00
-1.70210466e-01 -7.68287480e-01 2.80741490e-02 3.37958366e-01
-1.65351167e-01 -1.61966160e-01 -7.80314207e-01 1.01372492e+00
6.12372458e-02 7.63492942e-01 -5.09737842e-02 -3.86703432e-01
5.38295388e-01 1.80949405e-01 6.07745349e-01 -1.15379262e+00
-7.00510561e-01 -6.52680576e-01 7.35425234e-01 -4.31288749e-01
-4.06327039e-01 -4.32240844e-01 -1.18166018e+00 -4.40755337e-01
-5.58412038e-02 -4.60495129e-02 2.36619711e-01 1.42794979e+00
2.96735346e-01 8.75894368e-01 -2.24798635e-01 -8.32590520e-01
-1.27972409e-01 -1.06606829e+00 -1.64553419e-01 7.33855844e-01
-8.54996499e-03 -6.65746152e-01 1.28354011e-02 6.15404665e-01]
|
[11.215782165527344, 8.756576538085938]
|
74883d61-c759-4a05-b300-4901fd54ea80
|
consistent-video-instance-segmentation-with
|
2206.07011
| null |
https://arxiv.org/abs/2206.07011v1
|
https://arxiv.org/pdf/2206.07011v1.pdf
|
Consistent Video Instance Segmentation with Inter-Frame Recurrent Attention
|
Video instance segmentation aims at predicting object segmentation masks for each frame, as well as associating the instances across multiple frames. Recent end-to-end video instance segmentation methods are capable of performing object segmentation and instance association together in a direct parallel sequence decoding/prediction framework. Although these methods generally predict higher quality object segmentation masks, they can fail to associate instances in challenging cases because they do not explicitly model the temporal instance consistency for adjacent frames. We propose a consistent end-to-end video instance segmentation framework with Inter-Frame Recurrent Attention to model both the temporal instance consistency for adjacent frames and the global temporal context. Our extensive experiments demonstrate that the Inter-Frame Recurrent Attention significantly improves temporal instance consistency while maintaining the quality of the object segmentation masks. Our model achieves state-of-the-art accuracy on both YouTubeVIS-2019 (62.1\%) and YouTubeVIS-2021 (54.7\%) datasets. In addition, quantitative and qualitative results show that the proposed methods predict more temporally consistent instance segmentation masks.
|
['Zicheng Liu', 'Andre Abrantes', 'Peng Chu', 'Jiang Wang', 'Quanzeng You']
|
2022-06-14
| null | null | null | null |
['video-instance-segmentation']
|
['computer-vision']
|
[ 2.77458489e-01 -5.66009842e-02 -7.51406193e-01 -5.96847415e-01
-9.21761155e-01 -4.06775773e-01 3.32331955e-01 -1.26308843e-01
-3.26419890e-01 4.72578049e-01 -1.02520131e-01 1.91583514e-01
1.79711133e-01 -3.18415135e-01 -8.79590333e-01 -3.45606357e-01
-2.43445709e-01 3.37194175e-01 7.51094341e-01 3.50054115e-01
3.09785575e-01 -5.06185517e-02 -1.58360779e+00 7.31235862e-01
7.00743616e-01 1.35645711e+00 4.65580106e-01 8.86051893e-01
-2.77354002e-01 1.05954564e+00 -3.29984277e-01 -1.48374617e-01
3.19761097e-01 -4.75817531e-01 -1.10133946e+00 6.72759175e-01
9.37107205e-01 -5.51604867e-01 -3.72719407e-01 8.16315055e-01
-4.30571847e-02 3.09319109e-01 1.78766906e-01 -1.30061674e+00
-3.87034088e-01 4.20412093e-01 -8.05549443e-01 6.79364800e-01
3.49502981e-01 4.85985875e-01 1.11441171e+00 -6.70425475e-01
8.77880216e-01 7.79534161e-01 3.65765989e-01 5.14695406e-01
-9.41314399e-01 -4.98308212e-01 9.11122203e-01 6.31554544e-01
-1.22379589e+00 -3.74222517e-01 5.04599333e-01 -4.62183207e-01
1.07524431e+00 3.88911277e-01 9.40308213e-01 8.10962498e-01
-7.82326087e-02 1.38050151e+00 7.28010058e-01 2.29010835e-01
3.47181186e-02 -4.21498567e-01 3.55016023e-01 5.86471081e-01
-3.96922499e-01 -1.40493140e-01 -6.54705346e-01 4.13129479e-01
8.05217206e-01 2.47080445e-01 -3.02033901e-01 4.73791324e-02
-1.49396467e+00 1.41538233e-01 3.01163495e-01 1.79425851e-01
-6.49181008e-01 4.68150318e-01 6.39352441e-01 -1.60909712e-01
6.17974818e-01 -8.92902613e-02 -6.96371078e-01 -5.30947566e-01
-1.54563105e+00 2.27877080e-01 5.09879231e-01 1.54258490e+00
5.36532640e-01 2.71242931e-02 -6.30636156e-01 7.01640666e-01
2.99608976e-01 1.64023638e-01 3.30964863e-01 -1.43071854e+00
6.04373097e-01 3.66670936e-01 1.89415053e-01 -7.70576596e-01
-6.35439381e-02 -1.82451561e-01 -4.15630370e-01 -2.99979925e-01
3.29961270e-01 2.52952933e-01 -1.23828948e+00 1.55433536e+00
3.68564755e-01 1.16152024e+00 -2.80043721e-01 1.15193129e+00
9.35035825e-01 9.64019358e-01 5.81192136e-01 -5.47391355e-01
1.17726898e+00 -1.59365273e+00 -7.99199641e-01 -3.26790899e-01
4.06270474e-01 -5.61842740e-01 9.08812344e-01 1.90340787e-01
-1.59021652e+00 -9.12555814e-01 -6.61962688e-01 -1.71482503e-01
2.41181821e-01 1.58004891e-02 3.19895715e-01 3.52940075e-02
-9.37185109e-01 5.23728311e-01 -1.15096486e+00 -4.47628647e-03
6.56215191e-01 3.49914163e-01 -6.54138625e-02 -1.35865957e-01
-6.79315686e-01 7.43880942e-02 5.43825686e-01 9.75398198e-02
-9.71297145e-01 -9.19494033e-01 -7.61447370e-01 -3.52345593e-02
6.51445687e-01 -3.91863644e-01 1.42991662e+00 -1.44198203e+00
-1.18941844e+00 7.93318927e-01 -6.99491143e-01 -6.24514222e-01
6.43075168e-01 -5.08848011e-01 -2.70496041e-01 3.59761745e-01
3.10600519e-01 1.19639504e+00 8.11058521e-01 -1.18084335e+00
-1.29535556e+00 -1.55153692e-01 1.02903852e-02 3.43961775e-01
9.53689069e-02 -9.62640494e-02 -1.52878308e+00 -7.67835379e-01
2.83075571e-01 -8.47012103e-01 -1.32816106e-01 3.10801975e-02
-4.45726037e-01 -2.76700020e-01 1.20783794e+00 -8.45485330e-01
1.38221419e+00 -2.09721160e+00 1.32607400e-01 -1.58560470e-01
9.11605656e-02 2.42267981e-01 -1.97010279e-01 -2.59667486e-01
-3.19465771e-02 2.74577588e-01 -2.71665961e-01 -6.19200706e-01
-1.80453226e-01 3.94377768e-01 -1.92692623e-01 2.25990549e-01
2.45335072e-01 1.13913310e+00 -1.03283107e+00 -7.17383564e-01
4.32070196e-01 3.63586456e-01 -7.50933051e-01 2.93447256e-01
-7.80817568e-01 5.62956631e-01 -3.16616446e-01 7.62723207e-01
3.32316250e-01 -5.41418314e-01 2.13879898e-01 -2.86748558e-01
6.64209276e-02 1.12176731e-01 -1.00374246e+00 1.93415856e+00
-3.26210000e-02 8.00848305e-01 -1.86248690e-01 -8.89122188e-01
3.26502711e-01 4.70879644e-01 1.05996585e+00 -1.06722021e+00
-1.60651267e-01 3.33846696e-02 -2.30837971e-01 -7.09823489e-01
8.06284845e-01 3.44012201e-01 1.85430959e-01 1.74523398e-01
-8.35634395e-02 3.60477716e-01 6.66635454e-01 2.79467970e-01
7.41863191e-01 5.60012400e-01 -3.41433883e-02 -7.32353702e-02
3.71827453e-01 1.09459192e-01 8.99554193e-01 5.78597903e-01
-5.33988357e-01 9.88881171e-01 3.34724754e-01 -6.33693814e-01
-1.19807053e+00 -7.80143917e-01 3.14351954e-02 1.04649484e+00
7.06082344e-01 -4.94534492e-01 -7.99093604e-01 -8.40569377e-01
-4.50331151e-01 4.84067976e-01 -4.57525253e-01 3.65486652e-01
-8.44706655e-01 -3.71562213e-01 1.20352119e-01 7.55171657e-01
6.62550509e-01 -1.12939036e+00 -5.31589568e-01 5.27213216e-01
-7.90920973e-01 -1.62289798e+00 -8.81872356e-01 -3.54805738e-01
-1.04524267e+00 -1.28012419e+00 -6.71017170e-01 -7.71305978e-01
5.23617864e-01 4.16967541e-01 1.29379332e+00 4.88374799e-01
-1.75857216e-01 3.87371361e-01 -4.66907114e-01 2.65540928e-01
-1.33692942e-04 -8.24235380e-02 -2.71939337e-01 4.49698307e-02
3.00606966e-01 -1.28225952e-01 -9.18790579e-01 6.87850118e-01
-9.78657484e-01 5.01137972e-01 -9.14018042e-03 5.32712042e-01
1.35839212e+00 -3.11963975e-01 3.61618400e-01 -8.46597314e-01
-2.17872545e-01 -4.69913572e-01 -4.02191550e-01 2.90290356e-01
-4.21485037e-01 -5.17565310e-01 2.92014956e-01 -4.34440643e-01
-8.31965625e-01 7.29818717e-02 2.31293998e-05 -8.13358188e-01
-3.33723903e-01 2.27152735e-01 1.18404970e-01 4.02583063e-01
7.09021464e-02 3.94566566e-01 -2.98048019e-01 -2.17045695e-01
2.00573653e-01 2.43021160e-01 6.93064630e-01 -4.36912417e-01
1.44491583e-01 3.40081394e-01 -5.05094886e-01 -5.27697563e-01
-9.09811318e-01 -7.73958921e-01 -6.10161722e-01 -5.16109586e-01
1.34290004e+00 -1.18498957e+00 -6.25242114e-01 4.05930996e-01
-1.00588119e+00 -9.47591007e-01 -1.76750571e-01 1.93656638e-01
-9.18917537e-01 5.28641582e-01 -8.65217209e-01 -5.05009353e-01
-2.11345434e-01 -1.46728945e+00 1.44966984e+00 1.90753385e-01
-2.73710668e-01 -7.07609355e-01 -4.20511097e-01 7.57658720e-01
-9.20313299e-02 1.45576805e-01 3.95126760e-01 -2.54524440e-01
-1.18049574e+00 3.12326383e-02 -3.64016742e-01 8.18340108e-02
-1.46540254e-02 3.26094717e-01 -6.01657629e-01 -9.36045125e-02
-4.34356272e-01 1.62916511e-01 8.25292885e-01 7.42498577e-01
1.76275826e+00 -4.74587142e-01 -2.99972117e-01 5.76164186e-01
1.38563955e+00 4.73540246e-01 8.38834524e-01 2.07922682e-01
1.06108809e+00 3.14745992e-01 1.13617599e+00 4.63980407e-01
3.74288291e-01 1.07975423e+00 4.95373428e-01 7.30647966e-02
-2.41590396e-01 6.65562898e-02 2.66206205e-01 5.68158746e-01
-2.05577031e-01 -5.21790624e-01 -8.65192235e-01 8.56138110e-01
-2.40899181e+00 -1.30902445e+00 -3.81738007e-01 1.87518799e+00
6.93840504e-01 2.85322785e-01 4.83221710e-01 -4.80393097e-02
8.29212964e-01 3.28641862e-01 -5.54419875e-01 -3.29998769e-02
1.77048519e-01 -2.11747527e-01 3.37569267e-01 5.92106938e-01
-1.20796406e+00 1.16242194e+00 5.87492228e+00 8.01503360e-01
-9.42427754e-01 1.39815167e-01 1.31698191e+00 -6.03309870e-01
5.79272695e-02 -1.54932231e-01 -6.62539184e-01 1.03424895e+00
7.46879339e-01 2.64320582e-01 1.91398084e-01 8.08036387e-01
5.44664502e-01 -1.77591562e-01 -1.22307003e+00 1.04925013e+00
-1.19099542e-01 -1.69662833e+00 2.68643647e-02 -1.84640184e-01
1.10661209e+00 1.33202691e-02 8.15992523e-03 2.01454550e-01
-3.88189644e-01 -9.20328617e-01 1.21156991e+00 7.03190923e-01
6.62008584e-01 -6.13342404e-01 5.27496994e-01 2.12059304e-01
-1.54043055e+00 3.63007449e-02 1.42157182e-01 1.22558326e-01
6.88028812e-01 2.65732110e-01 -5.82786977e-01 3.13277632e-01
1.05251706e+00 1.25278866e+00 -2.99257278e-01 1.14817464e+00
9.07337591e-02 6.91353381e-01 -2.82588631e-01 4.09151405e-01
5.90023935e-01 -1.30122349e-01 3.77622753e-01 1.29418707e+00
7.37465769e-02 5.46451747e-01 6.50265813e-01 7.70915806e-01
-5.86264804e-02 -2.19721481e-01 9.84395370e-02 -1.07404985e-01
2.97853231e-01 8.79374802e-01 -1.02663994e+00 -8.48592043e-01
-4.82779384e-01 1.22595620e+00 -1.55985700e-02 6.03623509e-01
-1.40477502e+00 3.28500509e-01 9.69767988e-01 1.56956419e-01
6.97004139e-01 -3.23444039e-01 -4.77888077e-01 -1.11786389e+00
3.36898774e-01 -7.06725776e-01 3.63262922e-01 -8.70111167e-01
-1.00046778e+00 4.76156175e-01 -8.64647999e-02 -1.28656435e+00
-2.82061070e-01 -1.58209264e-01 -4.09348041e-01 3.01540256e-01
-1.47038877e+00 -9.46752429e-01 -5.19562542e-01 5.58358312e-01
1.50583529e+00 2.21815616e-01 2.09775940e-01 4.62368786e-01
-8.56891990e-01 5.10625064e-01 -2.78526455e-01 1.91622272e-01
3.33898902e-01 -1.06449807e+00 4.62678254e-01 1.00304878e+00
4.02228594e-01 1.67939752e-01 5.81550121e-01 -7.98151076e-01
-1.10581267e+00 -1.39861798e+00 6.79146111e-01 -3.82993937e-01
2.40722761e-01 6.98351786e-02 -1.15304923e+00 7.25547671e-01
9.96853113e-02 5.06876290e-01 3.27343076e-01 -2.18927264e-01
-1.69438824e-01 2.34526880e-02 -9.28773999e-01 7.09331632e-01
1.39004743e+00 -4.08780545e-01 -9.62198898e-02 5.23430109e-01
9.01817620e-01 -7.81886220e-01 -1.02582800e+00 4.47846204e-01
3.68770748e-01 -9.91509974e-01 1.02136731e+00 -6.11892462e-01
9.17832196e-01 -4.95705932e-01 -1.64547086e-01 -3.23784262e-01
-3.85844558e-01 -5.26355803e-01 -6.62777126e-01 1.18579972e+00
2.80971020e-01 2.69041836e-01 1.16834283e+00 1.15465546e+00
-3.36386859e-01 -1.13211358e+00 -8.52767587e-01 -5.80947936e-01
-6.45142555e-01 -8.83717000e-01 3.16559911e-01 8.43567312e-01
-2.00959846e-01 -2.48915359e-01 -3.94020945e-01 3.04571778e-01
5.68356514e-01 3.50109726e-01 6.04750991e-01 -4.80015099e-01
-3.55265886e-01 -5.55873692e-01 -6.41056240e-01 -1.52611339e+00
2.53568619e-01 -6.09497905e-01 2.71023214e-01 -1.61981201e+00
3.12381953e-01 -4.33031797e-01 -4.05425876e-01 5.46067655e-01
-5.39852798e-01 6.67763710e-01 4.33219254e-01 5.12488186e-01
-1.49342537e+00 2.21387610e-01 1.19417405e+00 -2.13131875e-01
-3.02212566e-01 -1.78043470e-01 -1.08725391e-01 4.89800632e-01
5.01816034e-01 -2.68619210e-01 -5.45712590e-01 -6.72170579e-01
-2.70301014e-01 3.75169247e-01 4.97437179e-01 -1.05635238e+00
1.82543635e-01 -4.31829363e-01 3.66252840e-01 -7.62046576e-01
3.63844991e-01 -8.25071752e-01 3.55037987e-01 1.98098049e-01
-3.60456973e-01 -3.89460437e-02 3.00717324e-01 1.03527308e+00
-4.07800078e-01 1.27998263e-01 7.66541123e-01 -1.04722038e-01
-1.64485204e+00 9.73359942e-01 -1.60362601e-01 2.61045396e-01
1.42100692e+00 -6.75448835e-01 -8.16270430e-03 -2.19050273e-01
-1.07400215e+00 5.54967403e-01 4.97460693e-01 7.08262205e-01
6.75720274e-01 -1.13312995e+00 -5.24122596e-01 -4.22535837e-02
4.15591756e-03 2.52754152e-01 6.76429868e-01 1.08327127e+00
-5.71036518e-01 1.90411747e-01 -1.93508025e-02 -1.33954513e+00
-1.46405745e+00 4.76484805e-01 2.57262617e-01 -6.75053988e-03
-8.26159954e-01 1.03976095e+00 3.45152527e-01 3.81671458e-01
4.45492387e-01 -4.33725804e-01 9.11768973e-02 -1.65016785e-01
5.82635283e-01 1.94411501e-01 -2.71803707e-01 -8.32605183e-01
-2.08486661e-01 5.65078616e-01 -3.30069482e-01 2.05042642e-02
1.13982677e+00 -4.48395520e-01 2.25113302e-01 3.22093278e-01
1.18381369e+00 -7.08553672e-01 -2.02382016e+00 -1.98651418e-01
-1.83377545e-02 -8.43150258e-01 -1.73716620e-01 -6.38256788e-01
-1.53893995e+00 5.61363101e-01 4.46620524e-01 1.49379134e-01
1.18404663e+00 -2.19332147e-02 1.26612473e+00 -1.59828395e-01
4.59305525e-01 -1.10865700e+00 2.05015510e-01 4.80347604e-01
4.54397738e-01 -1.37013125e+00 -1.76195428e-01 -7.25791872e-01
-8.27039957e-01 8.68773639e-01 9.06148851e-01 -6.24530129e-02
3.28748643e-01 3.18925977e-02 -1.81808978e-01 -3.93123440e-02
-8.92469525e-01 -7.95258880e-02 6.24091804e-01 2.33269423e-01
6.05398178e-01 -1.38297722e-01 -2.62211025e-01 3.78767699e-01
4.34094220e-01 6.82855472e-02 2.40910962e-01 6.32163107e-01
-2.92838126e-01 -8.40939820e-01 3.91635969e-02 5.43793023e-01
-6.79250777e-01 -3.20346816e-03 1.21348746e-01 6.21999800e-01
1.85807809e-01 8.46317768e-01 3.73618215e-01 -1.40761241e-01
7.25942329e-02 4.90295514e-02 3.10434401e-01 -4.24137741e-01
-7.02647030e-01 4.32383925e-01 4.35353480e-02 -1.13503242e+00
-7.70497501e-01 -6.90789878e-01 -1.72447896e+00 -2.54217386e-01
3.54302600e-02 -1.44371092e-01 2.47842625e-01 1.04682517e+00
5.07955551e-01 7.88975060e-01 2.92442977e-01 -1.14787030e+00
3.63188148e-01 -5.95541596e-01 -3.05686921e-01 8.83132339e-01
2.87398607e-01 -2.89967924e-01 1.32373855e-01 6.82784081e-01]
|
[9.162773132324219, -0.09205283224582672]
|
54d241bb-c7a7-438e-89b5-1e2b3eb41d89
|
self-supervised-out-of-distribution-detection-1
|
2109.15222
| null |
https://arxiv.org/abs/2109.15222v3
|
https://arxiv.org/pdf/2109.15222v3.pdf
|
Natural Synthetic Anomalies for Self-Supervised Anomaly Detection and Localization
|
We introduce a simple and intuitive self-supervision task, Natural Synthetic Anomalies (NSA), for training an end-to-end model for anomaly detection and localization using only normal training data. NSA integrates Poisson image editing to seamlessly blend scaled patches of various sizes from separate images. This creates a wide range of synthetic anomalies which are more similar to natural sub-image irregularities than previous data-augmentation strategies for self-supervised anomaly detection. We evaluate the proposed method using natural and medical images. Our experiments with the MVTec AD dataset show that a model trained to localize NSA anomalies generalizes well to detecting real-world a priori unknown types of manufacturing defects. Our method achieves an overall detection AUROC of 97.2 outperforming all previous methods that learn without the use of additional datasets. Code available at https://github.com/hmsch/natural-synthetic-anomalies.
|
['Bernhard Kainz', 'Benjamin Hou', 'Jeremy Tan', 'Hannah M. Schlüter']
|
2021-09-30
| null | null | null | null |
['self-supervised-anomaly-detection', 'supervised-anomaly-detection']
|
['computer-vision', 'computer-vision']
|
[ 4.12689924e-01 1.93072096e-01 4.50845599e-01 -4.03635561e-01
-9.67546523e-01 -1.31989613e-01 6.25762880e-01 3.31352770e-01
8.56617615e-02 1.67101473e-01 -3.96988153e-01 -1.88642845e-01
2.70814270e-01 -6.19535267e-01 -1.03742945e+00 -6.06442332e-01
-1.99137285e-01 5.93838573e-01 3.47100168e-01 -1.22308157e-01
1.30341962e-01 4.13363725e-01 -1.54944015e+00 3.56045604e-01
1.10939598e+00 1.15916431e+00 -4.65765297e-02 7.64071405e-01
-5.79379313e-02 5.78876078e-01 -8.11167896e-01 -1.56383574e-01
5.82381725e-01 -3.53871763e-01 -5.09302199e-01 6.89707637e-01
9.95188773e-01 -2.66337782e-01 -1.99439913e-01 9.88915205e-01
1.43615738e-01 -1.49550706e-01 6.29298925e-01 -1.37286043e+00
-9.20958817e-01 -7.07967281e-02 -9.31854248e-01 6.30454421e-01
3.11735541e-01 5.15783131e-01 6.11262619e-01 -8.81350696e-01
3.43043119e-01 1.19123137e+00 8.86967182e-01 5.32060921e-01
-1.35217750e+00 -5.71644783e-01 1.08474337e-01 -6.95029423e-02
-1.15886021e+00 -1.44081786e-01 7.93435395e-01 -4.71276373e-01
7.59665966e-01 3.47085893e-01 3.10242862e-01 1.53332424e+00
3.74463052e-01 8.80209506e-01 9.32169676e-01 -4.55537200e-01
2.78233469e-01 -1.06189668e-01 -3.26779373e-02 7.23627627e-01
4.09734696e-01 -1.21500127e-01 -1.88907206e-01 -4.31350529e-01
7.10233450e-01 3.98096293e-01 7.01755509e-02 -3.50852728e-01
-1.19201112e+00 5.24467647e-01 3.04737926e-01 6.77870139e-02
-5.41491508e-01 9.38387811e-02 4.59298879e-01 5.23547232e-01
7.10767508e-01 6.34203732e-01 -4.22187239e-01 8.02441761e-02
-6.24821365e-01 1.83677673e-01 1.82257116e-01 9.15086329e-01
6.59250677e-01 6.79281235e-01 -1.53378725e-01 1.03494573e+00
2.69301217e-02 8.15207183e-01 3.72876793e-01 -1.04170537e+00
2.49608327e-02 6.15850449e-01 -1.24238376e-02 -7.95722902e-01
-2.95474529e-01 -4.93464947e-01 -9.85706151e-01 4.71212894e-01
5.54900885e-01 6.31353632e-02 -1.60970652e+00 1.40311539e+00
3.61194432e-01 5.46015739e-01 -2.48889928e-03 3.37025523e-01
3.38950187e-01 4.75114971e-01 -2.32503310e-01 2.13644370e-01
1.07094562e+00 -9.33308184e-01 -6.00128174e-01 -4.41444576e-01
4.90013003e-01 -8.13554168e-01 1.35864222e+00 6.93940878e-01
-9.06702995e-01 -4.99586612e-01 -9.67747033e-01 4.98702139e-01
-3.98939550e-01 -4.56137769e-02 2.07179844e-01 4.82775539e-01
-9.07995939e-01 3.38663071e-01 -1.16672194e+00 -5.00901282e-01
8.13814402e-01 -4.38484810e-02 -3.73377830e-01 -1.27739757e-01
-5.79284906e-01 3.46838534e-01 5.17354608e-02 -2.17621595e-01
-1.16355050e+00 -8.46674263e-01 -1.26348782e+00 -5.46380520e-01
4.91403699e-01 -5.13241053e-01 1.28643084e+00 -1.15169239e+00
-7.67824471e-01 1.09188008e+00 -1.09963536e-01 -7.29190171e-01
4.88166183e-01 -5.68983972e-01 -7.39627481e-01 7.61028677e-02
5.72553396e-01 4.34896469e-01 1.18958676e+00 -1.54737544e+00
-6.41093612e-01 -4.49698359e-01 -4.69963789e-01 -3.32408339e-01
-1.35439008e-01 -1.62600160e-01 -3.29405427e-01 -1.25415850e+00
2.22255602e-01 -7.74680614e-01 -4.75482047e-01 6.49653599e-02
-8.73015106e-01 1.24994233e-01 1.25767982e+00 -5.86347103e-01
7.88224399e-01 -2.30035329e+00 -4.85583574e-01 5.10852575e-01
1.67336360e-01 2.00974774e-02 -3.26601386e-01 2.02011541e-01
-4.66920644e-01 -2.12766871e-01 -9.11638558e-01 -6.02666438e-01
-3.39205712e-01 4.34855700e-01 -3.88884425e-01 4.75189835e-01
7.39667416e-01 5.09619653e-01 -9.07883763e-01 -2.20602527e-01
2.71175623e-01 3.81537154e-02 -4.86329824e-01 3.69220257e-01
-4.69011307e-01 7.63399363e-01 -2.03274369e-01 1.32344973e+00
8.83397639e-01 -2.85227060e-01 -6.34124160e-01 3.33121896e-01
3.55181396e-01 -2.84620672e-01 -9.97488678e-01 1.70615363e+00
-2.23269001e-01 4.75668162e-01 -6.80777356e-02 -1.06950831e+00
8.48378956e-01 -1.57740526e-02 5.57810485e-01 -7.66360223e-01
-1.74390003e-01 2.96620518e-01 -2.55528390e-01 -5.39768696e-01
9.49678048e-02 2.87928760e-01 -1.14654034e-01 2.09899366e-01
2.11940587e-01 -1.13210991e-01 6.59290180e-02 3.17665786e-01
1.72785842e+00 -1.75489888e-01 2.95841657e-02 -1.60681054e-01
3.83118004e-01 2.12873176e-01 6.42305553e-01 9.94104981e-01
-6.29586577e-02 1.20779335e+00 4.25285667e-01 -6.27600908e-01
-1.32843518e+00 -1.68451285e+00 -3.40318710e-01 5.83719552e-01
6.87321052e-02 -1.75324723e-01 -7.98401356e-01 -1.01906216e+00
2.69536942e-01 8.58756006e-01 -8.52725744e-01 -8.80129933e-02
-4.01640356e-01 -8.57817054e-01 5.71283638e-01 5.56136906e-01
4.69010562e-01 -1.10655117e+00 -3.12581897e-01 -1.47087425e-01
4.88947553e-04 -1.26997817e+00 -3.96095902e-01 -2.76695713e-02
-7.93518722e-01 -1.15024698e+00 -5.55529475e-01 -5.84751248e-01
1.06187916e+00 -4.55206484e-02 1.31386435e+00 1.43944383e-01
-1.06142354e+00 8.44597042e-01 -3.25152040e-01 -8.45838666e-01
-8.56480122e-01 -5.78391135e-01 2.49299273e-01 2.99677968e-01
3.48575205e-01 -7.25086212e-01 -5.99742115e-01 3.12560469e-01
-1.20982111e+00 -3.28305334e-01 6.04542315e-01 1.00798535e+00
8.85837138e-01 -1.49024308e-01 4.31506336e-01 -1.16791856e+00
2.59162128e-01 -7.23058403e-01 -5.26893437e-01 -1.29844978e-01
-6.33091271e-01 -1.06824145e-01 4.60071206e-01 -3.62579584e-01
-8.68032157e-01 1.79096162e-01 -1.25373751e-01 -8.67556512e-01
-9.46666837e-01 -1.28485307e-01 1.72972262e-01 1.64408043e-01
1.04146838e+00 2.91117698e-01 4.24697965e-01 -3.34591359e-01
1.51929781e-01 3.45568061e-01 9.77092206e-01 -4.76694912e-01
1.17640758e+00 8.47925782e-01 -1.20328471e-01 -1.04783213e+00
-7.83840358e-01 -6.47255003e-01 -4.25648659e-01 1.72730396e-03
7.65328228e-01 -1.06079519e+00 1.84291452e-01 9.18846011e-01
-7.91315794e-01 -3.79578888e-01 -8.49656641e-01 1.16584919e-01
-5.70457220e-01 5.94267547e-01 -5.78016818e-01 -7.57061243e-01
-3.14284772e-01 -8.17592204e-01 1.48008370e+00 -1.42001718e-01
-2.33522981e-01 -8.35063040e-01 1.33307025e-01 3.06628942e-01
4.75618839e-01 8.80782604e-01 6.34707987e-01 -9.53933239e-01
-5.68197727e-01 -4.67223346e-01 -1.67165715e-02 7.45489478e-01
6.15716696e-01 3.08907814e-02 -1.15022027e+00 -2.34318256e-01
4.62708287e-02 -2.98442870e-01 9.61887181e-01 3.49703103e-01
1.68261373e+00 -3.39877248e-01 -2.19133034e-01 5.74093640e-01
1.18811905e+00 1.63342908e-01 9.41369951e-01 5.23141265e-01
8.80788207e-01 2.10957050e-01 8.08797121e-01 4.89587694e-01
1.16984293e-01 3.26650500e-01 8.45138967e-01 -6.17136598e-01
-1.56020038e-02 1.09189861e-01 3.37204009e-01 3.33809733e-01
2.50424206e-01 -1.72980577e-01 -1.22583830e+00 9.95259702e-01
-1.65868890e+00 -9.08314824e-01 -1.56787887e-01 2.08322883e+00
4.72990632e-01 2.59646922e-01 2.08860785e-01 1.52287275e-01
5.21727026e-01 7.44332001e-02 -8.68944824e-01 -2.27953717e-01
-2.22637400e-01 2.74354458e-01 4.58005458e-01 1.89823583e-01
-1.49881673e+00 5.91098785e-01 6.14691210e+00 8.12507749e-01
-8.43421578e-01 2.13647615e-02 1.00083494e+00 -2.41214838e-02
-1.51085377e-01 -4.17672396e-01 -6.41736835e-02 5.25450349e-01
8.23692679e-01 3.85780334e-01 -1.19078256e-01 1.06005931e+00
-4.34087962e-02 -1.02451436e-01 -9.46177542e-01 8.87588859e-01
2.79342473e-01 -1.17700553e+00 2.15451300e-01 -1.37993380e-01
1.03239214e+00 2.80567080e-01 3.13254893e-01 3.59236784e-02
3.88200909e-01 -8.35395873e-01 3.64004970e-01 4.08693880e-01
6.64498568e-01 -5.26919544e-01 7.91759193e-01 5.45822568e-02
-7.09584057e-01 -2.10538208e-01 -9.53184068e-02 3.47252935e-01
-3.36664938e-03 8.89462292e-01 -8.77660632e-01 4.43101078e-01
1.19994116e+00 6.64605439e-01 -1.05892372e+00 8.98024917e-01
1.83603317e-01 8.45722914e-01 -5.64032078e-01 8.67537856e-01
1.71063259e-01 -1.04197279e-01 8.93333316e-01 1.19255376e+00
5.50157964e-01 -4.88808036e-01 9.93121490e-02 1.10235262e+00
1.63992479e-01 -7.19155520e-02 -1.01129270e+00 2.10974500e-01
8.67538229e-02 9.68785286e-01 -7.13707447e-01 -2.23015532e-01
-6.41479850e-01 1.42362833e+00 -7.18126744e-02 3.71745497e-01
-8.55112612e-01 -2.02605382e-01 8.24048042e-01 5.45734823e-01
2.63445139e-01 9.67681706e-02 -4.17744726e-01 -1.03324950e+00
3.93718868e-01 -1.16240823e+00 6.60628796e-01 -6.88921690e-01
-1.79747188e+00 6.54058456e-01 -3.43506336e-02 -1.64035928e+00
-3.48118305e-01 -7.66743898e-01 -1.02879286e+00 3.81168753e-01
-1.07813442e+00 -1.27534556e+00 -6.85081601e-01 6.51362836e-01
8.64420950e-01 -5.93300581e-01 8.86496723e-01 1.30013600e-01
-7.47261822e-01 7.83536494e-01 1.87993750e-01 2.69497812e-01
9.03577209e-01 -1.71365094e+00 1.09562373e+00 1.41126049e+00
2.42977291e-01 6.55861944e-03 7.93198228e-01 -7.80252159e-01
-8.13201010e-01 -1.51368535e+00 -1.59597069e-01 -9.39443886e-01
5.77596068e-01 -3.91030669e-01 -1.24690795e+00 8.68683100e-01
2.50696898e-01 5.50258517e-01 3.73273104e-01 -1.98271364e-01
-4.76565093e-01 -1.70726418e-01 -1.45801497e+00 4.48765188e-01
9.48447883e-01 -2.96387792e-01 -3.40212077e-01 6.41349137e-01
7.69452810e-01 -5.64971864e-01 -6.28126025e-01 8.34331393e-01
-2.11902767e-01 -1.06035900e+00 1.06453395e+00 -4.24748600e-01
5.54090738e-01 -4.12621975e-01 -9.40255672e-02 -1.32217574e+00
7.53384456e-02 -4.77669626e-01 -3.06558669e-01 9.06289399e-01
5.55481434e-01 -8.71801078e-01 6.90345287e-01 1.47823066e-01
-4.83078420e-01 -8.05971801e-01 -6.65696084e-01 -1.03624749e+00
-2.46039972e-01 -6.99692190e-01 3.25474292e-01 1.23957801e+00
-5.12807071e-01 -3.07111263e-01 -1.13816254e-01 8.00077379e-01
9.97416377e-01 -1.92982778e-01 9.29824114e-01 -1.19403541e+00
-1.68055832e-01 -9.45921019e-02 -1.00918698e+00 -3.75050038e-01
-2.72408485e-01 -5.14535189e-01 4.48266305e-02 -1.10647714e+00
-7.19000548e-02 -2.85784900e-01 -4.40345585e-01 5.72247624e-01
-2.30989680e-01 5.76272011e-01 -5.39502084e-01 4.72270437e-02
-6.69080317e-01 4.76400793e-01 8.62907588e-01 -2.58729070e-01
-1.56897884e-02 2.36876249e-01 -4.71506178e-01 9.86197710e-01
1.10623336e+00 -3.85393709e-01 -3.17571253e-01 -2.32970864e-01
-2.89654881e-01 -5.18681526e-01 5.56839705e-01 -1.34122694e+00
-2.04211324e-01 1.16071559e-01 4.96211529e-01 -6.21546805e-01
1.06967844e-01 -6.72107756e-01 -2.18211487e-01 3.60831857e-01
9.37243104e-02 5.53979218e-01 7.11820722e-01 1.01314282e+00
-3.55081826e-01 1.62488982e-01 8.21829617e-01 -1.73800010e-02
-7.13986933e-01 3.61267418e-01 -3.24982971e-01 2.99901515e-01
1.27104163e+00 -2.47011364e-01 -2.35726520e-01 -5.43805897e-01
-8.80609453e-01 2.65316099e-01 6.78439856e-01 7.01795816e-01
8.70604157e-01 -1.25477958e+00 -7.83910632e-01 6.47029817e-01
7.25920618e-01 6.14834368e-01 4.18397635e-01 6.76468253e-01
-8.81826758e-01 -2.90768266e-01 -2.69064695e-01 -1.28674388e+00
-1.07670534e+00 5.72297037e-01 4.12606806e-01 -1.13429531e-01
-1.02997231e+00 7.59523451e-01 3.83347720e-01 -6.27398252e-01
2.52989888e-01 -4.64908034e-01 3.98797035e-01 -7.14938521e-01
5.46383858e-01 2.55906373e-01 7.61951059e-02 -3.42881680e-01
-1.53605565e-01 3.63047242e-01 -2.99395800e-01 2.82212883e-01
1.27341092e+00 7.92075619e-02 -5.63672557e-02 5.81095159e-01
7.03635812e-01 1.28632337e-01 -1.23115063e+00 -4.53798413e-01
1.89757064e-01 -5.72211206e-01 -3.77607852e-01 -6.77313983e-01
-1.22369659e+00 5.11613190e-01 1.14520872e+00 2.61843294e-01
1.39312315e+00 2.82023907e-01 6.68690264e-01 3.99285018e-01
7.94006884e-03 -9.17737305e-01 7.00069010e-01 1.73392192e-01
1.10464275e+00 -1.73727667e+00 -2.02116802e-01 -4.72231299e-01
-8.65490198e-01 7.09838033e-01 1.20506704e+00 -4.49428916e-01
6.89686894e-01 5.50502777e-01 5.29090464e-01 -5.31757116e-01
-5.31297445e-01 -9.73240137e-02 2.15190738e-01 8.03487420e-01
-4.52115908e-02 -1.58200026e-01 5.53427637e-01 1.37177810e-01
7.31156990e-02 -5.73509812e-01 5.91597378e-01 9.12218928e-01
-6.19541481e-02 -8.47843409e-01 -6.61212802e-01 8.55463803e-01
-6.22806966e-01 1.22465856e-01 -2.00847954e-01 8.19566786e-01
1.90568835e-01 5.93237817e-01 6.51482880e-01 -2.72552699e-01
5.35436988e-01 2.10048735e-01 7.28155077e-02 -5.68370044e-01
-1.28518259e-02 -7.43377507e-02 -1.88181207e-01 -9.78853285e-01
-1.29238054e-01 -8.45388532e-01 -1.26949775e+00 2.04312325e-01
-3.88505422e-02 -3.61859798e-01 3.35799009e-01 7.44306743e-01
4.80385125e-01 7.83653319e-01 7.48751640e-01 -6.64603829e-01
-2.94308484e-01 -1.08652723e+00 -6.31720066e-01 1.04240882e+00
7.95322418e-01 -4.57814813e-01 -5.76428950e-01 9.61872712e-02]
|
[7.649716854095459, 2.1084816455841064]
|
0b487137-4b8d-4520-bab4-d9e02342455d
|
acceleration-of-federated-learning-with-1
|
2203.02645
| null |
https://arxiv.org/abs/2203.02645v1
|
https://arxiv.org/pdf/2203.02645v1.pdf
|
Acceleration of Federated Learning with Alleviated Forgetting in Local Training
|
Federated learning (FL) enables distributed optimization of machine learning models while protecting privacy by independently training local models on each client and then aggregating parameters on a central server, thereby producing an effective global model. Although a variety of FL algorithms have been proposed, their training efficiency remains low when the data are not independently and identically distributed (non-i.i.d.) across different clients. We observe that the slow convergence rates of the existing methods are (at least partially) caused by the catastrophic forgetting issue during the local training stage on each individual client, which leads to a large increase in the loss function concerning the previous training data at the other clients. Here, we propose FedReg, an algorithm to accelerate FL with alleviated knowledge forgetting in the local training stage by regularizing locally trained parameters with the loss on generated pseudo data, which encode the knowledge of previous training data learned by the global model. Our comprehensive experiments demonstrate that FedReg not only significantly improves the convergence rate of FL, especially when the neural network architecture is deep and the clients' data are extremely non-i.i.d., but is also able to protect privacy better in classification problems and more robust against gradient inversion attacks. The code is available at: https://github.com/Zoesgithub/FedReg.
|
['Tao Jiang', 'Minlie Huang', 'Zhiwei Hong', 'Chencheng Xu']
|
2022-03-05
|
acceleration-of-federated-learning-with
|
https://openreview.net/forum?id=541PxiEKN3F
|
https://openreview.net/pdf?id=541PxiEKN3F
|
iclr-2022-4
|
['distributed-optimization']
|
['methodology']
|
[-3.62859249e-01 -1.68012083e-01 -3.41980070e-01 -4.11883622e-01
-9.26070392e-01 -7.79750168e-01 2.26139814e-01 -2.05368742e-01
-6.31660581e-01 8.55996251e-01 -1.79142095e-04 -3.62611145e-01
1.65585894e-02 -8.02899361e-01 -1.04750133e+00 -1.13747215e+00
2.51835231e-02 3.91729772e-01 -1.78649038e-01 3.11121643e-01
-1.55237705e-01 4.93067205e-01 -1.19128776e+00 4.05827314e-01
7.78934598e-01 1.19215202e+00 -2.87527382e-01 4.98806208e-01
-5.53634427e-02 1.01183236e+00 -4.76283342e-01 -8.38454366e-01
6.44799352e-01 -1.94484234e-01 -6.71272933e-01 -4.23256695e-01
5.46565473e-01 -6.38235033e-01 -5.48492312e-01 1.25266194e+00
4.44802105e-01 5.48874028e-03 4.87610288e-02 -1.43938231e+00
-5.62505901e-01 6.41709328e-01 -3.78163129e-01 -2.93689340e-01
-3.77365172e-01 2.63591051e-01 7.88359642e-01 -7.92725801e-01
3.75756770e-01 9.76712048e-01 7.79506028e-01 7.06213355e-01
-1.15963328e+00 -1.02483213e+00 1.58331379e-01 3.03442806e-01
-1.46995342e+00 -5.42317986e-01 5.91110826e-01 -2.35486794e-02
3.93344104e-01 4.52603042e-01 1.24263540e-01 8.80388021e-01
8.79406556e-02 8.75162542e-01 9.01231945e-01 -4.44781259e-02
4.64186281e-01 5.44825256e-01 5.18772937e-02 7.50248075e-01
2.47233286e-01 4.33912314e-02 -7.34212041e-01 -9.31424320e-01
4.29745764e-01 4.81007934e-01 -4.45703983e-01 -6.21057272e-01
-4.97268498e-01 1.03068352e+00 4.44855332e-01 1.64886732e-02
-4.00637567e-01 1.22837082e-01 6.34140611e-01 6.61177993e-01
4.70640779e-01 -2.93587029e-01 -9.33323801e-01 2.15380043e-01
-7.11808562e-01 1.16246775e-01 1.13636541e+00 6.88494563e-01
1.08729553e+00 -1.16811946e-01 -3.80429439e-03 5.17977655e-01
1.51560530e-01 4.27794456e-01 5.89178562e-01 -1.04518330e+00
7.04067767e-01 4.57492501e-01 7.18380138e-02 -8.65535259e-01
7.68724009e-02 -6.09429181e-01 -1.08706546e+00 2.67841488e-01
4.62517381e-01 -4.82810706e-01 -2.85087436e-01 2.08988595e+00
6.46860242e-01 4.91911545e-03 1.27882034e-01 7.52393126e-01
1.78997949e-01 4.84234303e-01 -9.23766047e-02 -9.87588167e-02
9.91304636e-01 -1.14589989e+00 -5.20507991e-01 -1.92189850e-02
8.30661833e-01 -3.14415097e-01 8.25010598e-01 5.23556292e-01
-9.52505291e-01 -7.57270977e-02 -6.79620862e-01 -1.63754001e-01
-3.64854425e-01 3.67504656e-02 7.06741512e-01 7.91370213e-01
-1.22167504e+00 5.59981585e-01 -8.41108024e-01 2.67585963e-01
1.03610241e+00 6.54287398e-01 -6.91280663e-01 -2.38964215e-01
-1.15351534e+00 4.11368668e-01 2.26879582e-01 -9.13145766e-02
-8.97312582e-01 -9.63281631e-01 -4.61460173e-01 2.42981657e-01
2.32553989e-01 -6.75311565e-01 1.18248427e+00 -1.03187096e+00
-1.45160353e+00 5.52349746e-01 -1.02911770e-01 -7.75010407e-01
1.01395178e+00 -1.09701604e-01 -1.39452770e-01 -4.33545485e-02
-2.96998024e-01 3.34889591e-02 9.58814919e-01 -1.02683663e+00
-7.81827211e-01 -7.89461493e-01 -2.92418987e-01 -5.59648126e-02
-8.79997671e-01 -1.80793941e-01 -4.03451055e-01 -3.36772919e-01
-3.41933012e-01 -8.00289392e-01 -1.53827623e-01 4.30653989e-01
-1.63319319e-01 3.67556990e-04 1.07514048e+00 -7.81542718e-01
1.16563857e+00 -2.43884706e+00 -3.18515480e-01 4.22652245e-01
2.96315759e-01 6.61807716e-01 -1.07168019e-01 4.24641162e-01
8.15856308e-02 2.87330747e-02 -2.41722018e-01 -7.15980411e-01
6.37208670e-02 2.79171318e-01 -5.70120633e-01 8.70281041e-01
-6.20215952e-01 8.14507067e-01 -4.98010844e-01 -2.14792818e-01
-9.97530594e-02 7.03296721e-01 -6.70824170e-01 1.66976109e-01
-1.72500461e-01 5.38676023e-01 -5.13479292e-01 4.00137305e-01
1.06980121e+00 -4.61288065e-01 3.14804137e-01 7.16780648e-02
9.76897106e-02 2.33478487e-01 -1.24146903e+00 1.48848760e+00
-4.83986676e-01 2.63738990e-01 6.23751879e-01 -8.26206684e-01
6.00766420e-01 5.37715673e-01 4.44126427e-01 -4.51691419e-01
1.45219788e-01 3.00339133e-01 -5.36915302e-01 -1.63266569e-01
-1.29947960e-01 1.09486125e-01 2.52364218e-01 9.02726829e-01
-5.04735857e-02 6.56342208e-01 -4.38209802e-01 1.65303662e-01
1.04161108e+00 -3.69119018e-01 -1.16923578e-01 -1.47133088e-02
6.42590523e-01 -3.33330274e-01 8.58351231e-01 9.89525139e-01
-2.70485818e-01 3.21244061e-01 3.46679986e-01 -7.62223721e-01
-8.36775184e-01 -7.90810227e-01 1.58857927e-02 1.35779107e+00
-1.22961603e-01 -3.83176446e-01 -8.10033977e-01 -1.08190811e+00
5.08054376e-01 6.26105309e-01 -5.07187009e-01 -3.30746859e-01
-4.04124647e-01 -7.12589979e-01 7.61351645e-01 1.58599451e-01
7.54783392e-01 -5.82236648e-01 -2.13420123e-01 1.02214813e-01
-4.45087701e-02 -5.82648337e-01 -7.41279244e-01 2.48402834e-01
-1.00502336e+00 -9.23218131e-01 -4.07517940e-01 -4.28962111e-01
8.03212762e-01 2.75479555e-01 6.91855192e-01 7.93471113e-02
-1.72881588e-01 1.82281628e-01 2.40463763e-01 -3.30138713e-01
-2.59886861e-01 1.04388490e-01 7.63420463e-02 6.09437406e-01
2.72320926e-01 -6.93383038e-01 -7.40210176e-01 3.08638632e-01
-1.10244715e+00 -2.48918518e-01 3.46417427e-01 9.14093077e-01
4.96131867e-01 3.58241610e-02 4.71404344e-01 -1.32471037e+00
4.07543242e-01 -6.20020270e-01 -7.84067690e-01 4.64526683e-01
-8.77484143e-01 1.64208442e-01 1.11009479e+00 -5.56575358e-01
-1.17225814e+00 9.92062092e-02 2.50179116e-02 -7.89494634e-01
5.90859018e-02 3.05931605e-02 -3.17826778e-01 -4.32389826e-01
6.46569848e-01 3.00310135e-01 4.79332954e-01 -8.27672362e-01
4.14879918e-01 7.24075019e-01 3.24043483e-01 -5.58073163e-01
7.69148231e-01 7.71067441e-01 -2.09394932e-01 -2.20924437e-01
-5.53810596e-01 -1.75237477e-01 -3.96993198e-02 2.08005488e-01
1.67047665e-01 -1.04247510e+00 -1.08798110e+00 9.58680093e-01
-9.10548508e-01 -5.00820518e-01 -3.30992371e-01 3.82604003e-01
-1.83964103e-01 3.70946050e-01 -7.12101519e-01 -6.87012255e-01
-9.33433890e-01 -8.46481323e-01 3.56745809e-01 2.71030009e-01
1.98065445e-01 -1.12024283e+00 2.09478121e-02 5.16918004e-01
8.71349275e-01 -6.07116371e-02 7.68905640e-01 -9.23974395e-01
-7.15020478e-01 -5.92259169e-01 -1.11686881e-03 8.19322169e-01
1.08695917e-01 -4.24371779e-01 -1.14452839e+00 -6.86669528e-01
5.34749269e-01 -4.64583904e-01 6.68092608e-01 4.67990339e-03
1.44785798e+00 -1.24941623e+00 -5.50087243e-02 1.05800641e+00
1.48926365e+00 -2.40967900e-01 4.34999585e-01 6.11291155e-02
6.06498361e-01 3.54082137e-01 5.97817525e-02 8.38900268e-01
2.55110919e-01 2.14187145e-01 5.05763769e-01 -1.60624878e-03
2.93741196e-01 -2.97530204e-01 5.54557621e-01 5.35334408e-01
4.48514551e-01 -5.11213625e-03 -5.73869646e-01 3.92361790e-01
-2.03096151e+00 -8.53803992e-01 1.15269862e-01 2.48750782e+00
1.21006286e+00 -4.55014944e-01 -4.60100546e-02 -4.61072356e-01
6.92135930e-01 8.74035060e-02 -1.10539103e+00 -2.44634137e-01
-2.37905920e-01 1.93362162e-01 1.00717783e+00 3.92098188e-01
-8.94329131e-01 6.99054778e-01 5.06537294e+00 9.54963684e-01
-1.26545203e+00 6.49388313e-01 9.04387295e-01 -5.46650767e-01
-2.31616482e-01 6.20921366e-02 -7.70563841e-01 5.92112005e-01
8.89017344e-01 -3.51097018e-01 8.07031751e-01 1.24326122e+00
-2.77481824e-01 2.49005198e-01 -9.03537095e-01 9.65891004e-01
-2.44427666e-01 -1.35267448e+00 -9.21931211e-03 2.70509273e-01
8.32813978e-01 4.83748317e-01 4.28581685e-01 4.69818152e-02
5.82922697e-01 -6.74587250e-01 4.08744842e-01 5.27824819e-01
6.30196154e-01 -1.03829277e+00 4.99495178e-01 7.78888583e-01
-6.20232224e-01 -4.60307270e-01 -5.01987219e-01 2.19620302e-01
-4.40926164e-01 7.30542481e-01 -4.87283617e-01 4.28395748e-01
8.32253635e-01 8.07574987e-02 -3.62923265e-01 8.54711413e-01
-9.15346369e-02 7.14277148e-01 -7.14852154e-01 2.36900494e-01
8.68371725e-02 -3.39352101e-01 2.85035253e-01 7.29025304e-01
8.95622894e-02 -2.41512790e-01 -1.07772574e-01 6.99453831e-01
-5.65135717e-01 3.03417414e-01 -4.49745983e-01 2.87928134e-01
5.65415382e-01 1.17389429e+00 2.72537708e-01 -1.79099947e-01
-4.69937891e-01 1.04087532e+00 7.27646470e-01 5.02890170e-01
-6.90508366e-01 -3.66523802e-01 1.08057630e+00 -2.40730822e-01
4.21124607e-01 4.80280593e-02 -3.08360726e-01 -1.29168200e+00
3.53755325e-01 -1.01182628e+00 8.96992385e-01 -4.62113284e-02
-1.61335468e+00 3.10750365e-01 -5.84060133e-01 -7.66214669e-01
-1.40870944e-01 -2.98175305e-01 -5.80939591e-01 9.67006207e-01
-1.41476214e+00 -9.88933742e-01 1.55078977e-01 1.17061329e+00
-2.39263356e-01 -2.47599661e-01 9.78649855e-01 4.43764508e-01
-6.90678418e-01 1.32884276e+00 9.28072512e-01 2.15743229e-01
9.16810274e-01 -7.07426250e-01 5.46045639e-02 7.38731623e-01
2.25963723e-02 9.73706901e-01 2.23167703e-01 -3.72547835e-01
-1.68673384e+00 -1.36104274e+00 8.02907586e-01 -2.18583047e-01
4.66989696e-01 -4.98770624e-01 -1.21157622e+00 8.17933261e-01
-8.81067291e-02 4.06154841e-01 8.93762708e-01 -8.36284161e-02
-7.95818865e-01 -6.45565450e-01 -1.73714590e+00 3.91248167e-01
5.01060903e-01 -7.71255672e-01 2.75030792e-01 5.95829010e-01
5.45287609e-01 -2.64534265e-01 -8.20752621e-01 -7.47311041e-02
4.97337937e-01 -1.01674843e+00 7.61593401e-01 -7.34175026e-01
-2.21883118e-01 -1.11330606e-01 -2.85873972e-02 -8.68351698e-01
-8.36179182e-02 -1.02254641e+00 -6.12661302e-01 1.24237335e+00
3.41694951e-01 -1.44657278e+00 1.13972497e+00 1.20852304e+00
5.28526604e-01 -6.10716820e-01 -1.31300604e+00 -6.59571588e-01
1.25009298e-01 -2.12703690e-01 9.87985194e-01 9.52796519e-01
-3.43953669e-01 -3.00600678e-01 -6.07183814e-01 3.35842580e-01
9.31405723e-01 9.31416079e-02 1.01585448e+00 -8.80252957e-01
-4.54828650e-01 -8.90024975e-02 5.60107529e-02 -7.78603613e-01
2.48552725e-01 -9.82983887e-01 -3.30034643e-01 -7.05828190e-01
1.57118753e-01 -5.60036004e-01 -5.74074507e-01 1.13942802e+00
3.72600020e-03 -2.55409796e-02 1.11301042e-01 4.51620698e-01
-4.92575318e-01 5.88671327e-01 8.96305084e-01 1.21288508e-01
-1.58534288e-01 4.14910495e-01 -8.97989154e-01 3.33140761e-01
1.08099055e+00 -7.59556413e-01 -2.70152241e-01 -6.00305259e-01
7.20056742e-02 -1.38894454e-01 5.23444414e-01 -7.86168933e-01
6.83504760e-01 -5.83928898e-02 2.64509022e-01 -1.07272707e-01
-2.35058945e-02 -1.19909406e+00 5.67204237e-01 5.77960730e-01
-3.78001869e-01 -2.10410282e-01 -4.48626131e-02 6.47151828e-01
7.13624107e-03 -4.89260741e-02 9.80974615e-01 -4.10556085e-02
-7.79546574e-02 8.36032629e-01 4.94966283e-02 -2.53139026e-02
9.17665303e-01 3.60165864e-01 -4.41943228e-01 -4.54714179e-01
-4.06366229e-01 3.35257739e-01 7.01052904e-01 1.54808015e-02
2.65712261e-01 -1.14680243e+00 -5.81892490e-01 6.28922462e-01
-3.01031858e-01 1.14626452e-01 5.28944373e-01 7.89274395e-01
-3.01976264e-01 2.69202024e-01 9.79555994e-02 -1.22497134e-01
-1.15182197e+00 6.40758097e-01 4.70780730e-01 -3.00269276e-01
-6.12038732e-01 9.82790887e-01 2.43886203e-01 -7.19795287e-01
7.65582800e-01 2.56723642e-01 6.05809391e-01 -1.81999028e-01
8.21164250e-01 5.52616775e-01 2.62926191e-01 -3.14269513e-01
-2.70504773e-01 -8.61134753e-02 -5.71126938e-01 1.06881790e-01
1.48242140e+00 -1.47736520e-01 -4.54215139e-01 -4.52781692e-02
1.85169923e+00 7.27700293e-02 -1.48570132e+00 -7.48043299e-01
-4.51504409e-01 -5.98429143e-01 1.06242202e-01 -7.05181181e-01
-1.67682290e+00 6.65796101e-01 6.86070681e-01 -2.21131071e-01
1.20559859e+00 -3.19177508e-01 1.15359259e+00 5.39511025e-01
6.42681837e-01 -1.02257526e+00 -4.33844179e-01 4.61565703e-01
4.83283639e-01 -9.65036333e-01 -5.31611554e-02 1.29090756e-01
-6.41305089e-01 8.92423093e-01 3.07475477e-01 2.18727402e-02
7.85564899e-01 3.36861432e-01 1.71128646e-01 1.89816132e-01
-8.72416914e-01 6.39081001e-01 -3.02652389e-01 2.47546509e-01
-3.01144749e-01 -1.24478213e-01 4.37125824e-02 8.48338485e-01
8.95310119e-02 2.10873902e-01 -9.83312577e-02 1.08242345e+00
-5.61625836e-03 -1.44386601e+00 -4.28660214e-01 3.59131843e-01
-9.00924265e-01 7.35148862e-02 -2.55537808e-01 1.80230513e-01
-1.40306940e-02 6.28980458e-01 -2.20145553e-01 -2.73382485e-01
-2.52081566e-02 4.18194711e-01 -6.53740838e-02 -1.42684236e-01
-9.85491514e-01 -3.53904426e-01 -3.69043648e-01 -8.46927524e-01
2.75435746e-01 -5.95637023e-01 -1.07817054e+00 -8.13552022e-01
-2.64478713e-01 4.12545860e-01 8.24943244e-01 5.08942306e-01
9.92936373e-01 -3.12896818e-01 1.20414114e+00 -3.27723473e-01
-1.35748911e+00 -4.65362757e-01 -7.66693592e-01 3.94596517e-01
5.68320572e-01 4.00342271e-02 -8.58094633e-01 -1.52860895e-01]
|
[5.848702907562256, 6.462411880493164]
|
fa5bf22c-37e4-4257-bd14-8acc28bd4c18
|
a-bounded-operator-approach-to-technical
|
2009.08821
| null |
https://arxiv.org/abs/2009.08821v1
|
https://arxiv.org/pdf/2009.08821v1.pdf
|
A bounded operator approach to technical indicators without lag
|
In the framework of technical analysis for algorithmic trading we use a linear algebra approach in order to define classical technical indicators as bounded operators of the space $l^\infty(\mathbb{N})$. This more abstract view enables us to define in a very simple way the no-lag versions of these tools. Then we apply our results to a basic trading system in order to compare the classical Elder's impulse system with its no-lag version and the so-called Nyquist-Elder's impulse system.
|
['Frédéric Butin']
|
2020-09-18
| null | null | null | null |
['algorithmic-trading']
|
['time-series']
|
[-5.71016610e-01 4.82427627e-02 2.03714147e-01 7.61218462e-03
-2.84950107e-01 -8.83458972e-01 4.23861474e-01 -1.33933589e-01
-6.50393784e-01 7.52517998e-01 -3.58419478e-01 -7.91298449e-01
-4.19020683e-01 -6.14470720e-01 -3.34086835e-01 -8.17668736e-01
-5.01055598e-01 2.39925921e-01 -1.10367030e-01 -4.31058347e-01
1.57937005e-01 4.45615888e-01 -1.07695401e+00 -3.81034881e-01
6.56101763e-01 1.61202371e+00 -3.99312079e-01 7.52627611e-01
-1.64383456e-01 6.44197464e-01 -5.31027853e-01 -5.94863236e-01
9.17834938e-01 -6.79006040e-01 -5.19552290e-01 -4.16293085e-01
-3.66627485e-01 -3.82254049e-02 1.90180302e-01 1.38605773e+00
1.66277096e-01 1.01775035e-01 6.77294791e-01 -1.00928330e+00
-2.51560748e-01 8.44156384e-01 -3.81167561e-01 6.25067472e-01
8.88170451e-02 -1.04525471e-02 8.78243387e-01 -8.19844127e-01
3.81638765e-01 8.98252785e-01 9.15444493e-01 -8.88024569e-02
-1.52197456e+00 -6.61124945e-01 -3.40280503e-01 -2.20475987e-01
-1.25826752e+00 -9.81957689e-02 4.60069716e-01 -8.18069756e-01
3.39912891e-01 2.66570002e-01 8.80117834e-01 5.60249388e-01
5.84890187e-01 3.33502889e-01 1.72524703e+00 -4.02473658e-01
3.44145745e-01 3.71732235e-01 4.57366467e-01 7.19715595e-01
4.10186023e-01 4.54087228e-01 1.70672778e-02 -1.78244621e-01
8.78793120e-01 -2.39199828e-02 -1.28061816e-01 -1.50405779e-01
-1.19745994e+00 1.29934788e+00 1.43001471e-02 9.42060113e-01
-5.10290563e-01 1.27897099e-01 2.61657357e-01 7.87046969e-01
6.46495521e-01 5.11628270e-01 -2.43840694e-01 -1.94283977e-01
-8.97385299e-01 3.60866547e-01 1.46449864e+00 7.32948542e-01
4.37579960e-01 4.75414217e-01 -1.11823872e-01 5.20116352e-02
2.02056497e-01 6.64303124e-01 2.46242911e-01 -1.17663705e+00
1.55926734e-01 -1.45821080e-01 3.53747249e-01 -6.91714644e-01
-4.39715058e-01 -8.12563121e-01 -6.93183601e-01 5.99769473e-01
8.18563640e-01 -4.08700168e-01 -3.58381234e-02 1.59415257e+00
-1.61773950e-01 1.97851822e-01 9.36802384e-03 3.84297222e-01
-4.47566599e-01 4.74415213e-01 -3.90504748e-01 -9.81750011e-01
7.94116676e-01 -3.55279773e-01 -8.65980506e-01 8.57290566e-01
3.97510767e-01 -8.29397202e-01 8.61580849e-01 8.56726944e-01
-1.22411060e+00 -1.32513195e-01 -9.43933189e-01 7.56466210e-01
-4.83398855e-01 -2.33757906e-02 6.19868398e-01 8.53683472e-01
-1.26160920e+00 9.06526327e-01 -4.02340919e-01 -1.71444181e-03
-2.29338114e-03 8.40382725e-02 2.41258487e-01 8.33279788e-01
-1.05120206e+00 9.17607784e-01 3.11568290e-01 2.81724006e-01
-5.70939660e-01 -7.37740278e-01 -3.26160997e-01 5.06725386e-02
3.56600642e-01 -3.37813973e-01 1.45888722e+00 -7.66936004e-01
-1.74892914e+00 6.61941826e-01 4.30553913e-01 -7.27942228e-01
1.14169419e+00 -1.42695203e-01 -3.22684050e-01 6.00783192e-02
3.24252322e-02 -3.57973278e-01 1.02327740e+00 -6.85759604e-01
-4.94331896e-01 -1.44069225e-01 2.26972848e-01 -4.95290458e-01
5.66661581e-02 -1.89730749e-02 5.97044349e-01 -1.11642718e+00
1.45755839e-02 -9.43550766e-01 -2.34296367e-01 -4.70378220e-01
-9.87249836e-02 -1.49810597e-01 3.40804219e-01 -3.56139213e-01
1.18589628e+00 -2.06782651e+00 1.92678515e-02 5.36473632e-01
9.47441459e-02 1.99765116e-02 5.06208599e-01 4.68985856e-01
-3.88812900e-01 9.61070731e-02 -1.03641696e-01 5.38334139e-02
2.89230615e-01 -1.71442211e-01 -6.96809709e-01 7.48943627e-01
-3.94757390e-01 5.85951984e-01 -6.96853340e-01 -2.69928098e-01
2.14735866e-01 -4.65404168e-02 -4.55006450e-01 -6.79565147e-02
-2.21548066e-01 5.30397594e-01 -4.90352869e-01 2.86302924e-01
4.75481927e-01 7.75749460e-02 -2.19446659e-01 9.69210714e-02
-8.05198610e-01 -1.93456367e-01 -1.18680179e+00 1.12614036e+00
-5.94451368e-01 4.36848640e-01 4.89059895e-01 -1.28553462e+00
8.75843585e-01 3.75994414e-01 7.58089066e-01 -3.26802880e-01
5.43577850e-01 6.73973620e-01 -5.77115361e-03 -1.77015319e-01
5.07550733e-03 -8.50662947e-01 -2.79606879e-01 7.23652542e-01
-1.36757687e-01 5.53776696e-02 3.18086207e-01 -8.28039870e-02
8.91007960e-01 -1.97794750e-01 4.61171836e-01 -9.75845635e-01
6.57354116e-01 -3.20905060e-01 3.81386220e-01 7.35269010e-01
-8.01076069e-02 -2.66076803e-01 9.71319079e-01 -3.85473222e-01
-8.27515423e-01 -1.24790442e+00 -3.73013526e-01 7.45850325e-01
-3.15405369e-01 -5.16484454e-02 -5.13176680e-01 -3.91510159e-01
3.83776218e-01 7.58700430e-01 -8.07381988e-01 1.13883048e-01
-3.08317900e-01 -6.44104779e-01 6.35625303e-01 1.76525116e-01
4.87480372e-01 -7.51175880e-01 -9.11703527e-01 3.01950514e-01
5.49192488e-01 -2.98657089e-01 -2.64363557e-01 5.70098460e-01
-7.90491223e-01 -8.20898831e-01 -1.04584694e+00 -1.41396150e-01
-5.47461584e-02 -5.41433334e-01 9.92010713e-01 -5.56829870e-01
-1.40369713e-01 3.72558028e-01 -1.81155846e-01 -7.84616947e-01
-4.28804964e-01 -5.21883190e-01 4.13036168e-01 3.18361312e-01
2.73967832e-02 -8.00253451e-01 -4.02957976e-01 2.92706758e-01
-1.01130259e+00 -8.45737159e-01 3.57262999e-01 6.74530327e-01
2.55249262e-01 8.76278132e-02 4.34582949e-01 -6.25583768e-01
7.34825611e-01 -2.39186659e-01 -1.54359305e+00 2.44091243e-01
-1.03036499e+00 4.61578727e-01 7.36806870e-01 -5.66597700e-01
-6.02196157e-01 -3.86548907e-01 1.51633412e-01 -5.40772617e-01
9.08431649e-01 5.27864397e-01 9.68093947e-02 -3.35129350e-01
1.94579735e-01 -1.03739575e-01 7.89711531e-03 -7.01195896e-01
6.14795744e-01 4.01847005e-01 5.85779369e-01 -5.08865416e-01
8.35002720e-01 5.05246580e-01 5.28967500e-01 -5.68147361e-01
-6.51866257e-01 5.99002093e-03 -4.69946057e-01 -5.95828239e-03
8.34006369e-01 -3.01883429e-01 -1.43315637e+00 3.06539178e-01
-1.01026535e+00 -3.33783478e-01 -9.45435524e-01 8.87404084e-01
-9.59542036e-01 3.41336650e-04 -9.38945830e-01 -1.45570219e+00
2.21898761e-02 -7.97053456e-01 2.98074067e-01 -4.95216390e-03
1.12959594e-02 -1.26059210e+00 5.15088916e-01 -4.15984869e-01
4.61614758e-01 3.33159983e-01 7.50674129e-01 -8.41100931e-01
-1.75365552e-01 -2.39520639e-01 1.78908724e-02 7.27011323e-01
-2.08143324e-01 -1.47191331e-01 -4.87294108e-01 -2.79925968e-02
1.07567501e+00 3.92214358e-01 7.13948786e-01 6.13271475e-01
4.29807484e-01 -4.93896127e-01 -8.12119618e-02 7.87924647e-01
1.58585393e+00 6.25535727e-01 1.48713797e-01 3.15851063e-01
3.59136537e-02 3.81367087e-01 8.37275088e-01 6.18554354e-01
-4.06701267e-01 7.41840899e-01 5.84206427e-04 3.04127425e-01
8.42447460e-01 1.27269879e-01 4.16875809e-01 7.10968792e-01
-1.21119797e-01 3.25115055e-01 -6.48994446e-01 2.21289489e-02
-1.68546546e+00 -1.25398016e+00 -1.38033882e-01 2.37118959e+00
5.26165187e-01 3.01451683e-01 4.49732572e-01 -6.31044582e-02
7.01101661e-01 1.01537565e-02 -7.12919459e-02 -7.97348082e-01
-3.85364518e-02 6.35669649e-01 9.63471591e-01 7.14203954e-01
-1.05757535e+00 2.56040454e-01 7.37828350e+00 7.74285674e-01
-1.15689671e+00 4.08463091e-01 2.83079773e-01 -3.48829627e-02
-2.91987836e-01 2.88549453e-01 -6.13783717e-01 8.09677005e-01
1.23204696e+00 -7.85206378e-01 3.92004728e-01 8.34442794e-01
3.68476927e-01 -1.44498035e-01 -9.71987784e-01 1.13716662e+00
-1.75134838e-01 -1.05153131e+00 -5.57248294e-01 5.82543314e-01
6.07439697e-01 -7.26258874e-01 3.93843442e-01 2.08523810e-01
4.85172391e-01 -9.49736834e-01 7.86171436e-01 9.86535311e-01
6.43842995e-01 -7.42937207e-01 8.87816191e-01 1.90112844e-01
-1.25887084e+00 -2.40473568e-01 -7.88868442e-02 -3.85069609e-01
4.61315483e-01 8.63471091e-01 -3.84876847e-01 7.25849748e-01
3.44536662e-01 4.92398649e-01 -2.21320335e-02 7.91299939e-01
3.53831589e-01 4.17954415e-01 -5.99286258e-01 9.42385271e-02
4.31891829e-01 -1.02414989e+00 6.00019515e-01 9.13752496e-01
8.88219774e-01 1.42079040e-01 -1.86033383e-01 1.06191230e+00
2.92831331e-01 7.07709193e-02 -7.51323581e-01 -1.62931606e-01
1.85442194e-02 1.03846407e+00 -9.68828261e-01 -5.02778709e-01
-3.39335442e-01 6.36450648e-01 -5.51801324e-01 1.54642686e-01
-7.14438200e-01 -6.15228772e-01 5.96511781e-01 -6.41960874e-02
4.71395314e-01 -3.35641682e-01 -2.31589705e-01 -1.20149469e+00
-3.57991457e-02 -4.84995037e-01 4.68649149e-01 -2.34497592e-01
-1.08139813e+00 4.90597159e-01 2.31598526e-01 -1.46235478e+00
-6.45080447e-01 -7.78453708e-01 -5.98774254e-01 1.17584562e+00
-6.45610988e-01 -2.92313278e-01 6.37215912e-01 7.22745001e-01
3.08118109e-02 -4.17438656e-01 5.57925284e-01 1.35226578e-01
-3.46531242e-01 2.37003565e-01 7.86387444e-01 -1.24059342e-01
2.49470860e-01 -1.50983250e+00 -1.53127834e-01 9.33407545e-01
4.56594490e-02 5.34911454e-01 1.28521252e+00 -5.02625406e-01
-1.07482207e+00 -4.03562427e-01 7.30918765e-01 -3.86681229e-01
1.38215768e+00 -1.59022123e-01 -4.98555422e-01 8.67553353e-01
1.26820683e-01 -6.91048801e-02 3.01001936e-01 -2.62276888e-01
-1.90275103e-01 -6.18223369e-01 -1.03040326e+00 2.48624578e-01
5.84145129e-01 -4.93143976e-01 -8.35066915e-01 8.54731277e-02
5.72180927e-01 2.90352613e-01 -1.02326906e+00 2.01917037e-01
6.76080823e-01 -1.27186775e+00 6.09164894e-01 -3.86561692e-01
-3.56060266e-01 -2.86740214e-02 -1.46123633e-01 -1.01275694e+00
7.95105472e-02 -1.50648236e+00 2.50172257e-01 1.12920070e+00
4.46979888e-02 -1.35614908e+00 1.62677065e-01 1.35361716e-01
6.97199777e-02 -5.44129193e-01 -1.41973698e+00 -1.32511389e+00
3.44144911e-01 -4.28502053e-01 2.21635833e-01 7.70184815e-01
4.09749150e-01 -1.65430069e-01 -3.24885786e-01 -5.44613004e-01
8.57842624e-01 1.59013033e-01 1.71200082e-01 -1.48356974e+00
-6.36885405e-01 -9.13061976e-01 -4.43334103e-01 -5.66460907e-01
4.98969629e-02 -7.01196909e-01 2.66825110e-02 -3.74397069e-01
-2.62480527e-01 -4.25586313e-01 -6.85211718e-01 -3.21814507e-01
5.95630586e-01 3.02023813e-02 3.65681112e-01 3.38344961e-01
-3.28787088e-01 2.61828065e-01 5.67850411e-01 3.20465237e-01
-2.19593436e-01 5.32155633e-01 -4.43062425e-01 8.38530958e-01
4.48630035e-01 -3.33712369e-01 -2.79723585e-01 4.05471355e-01
4.08576429e-01 5.05814552e-01 3.62752467e-01 -8.12624335e-01
-6.85837939e-02 3.27207614e-04 -1.19825065e-01 -4.78818208e-01
3.22542340e-02 -7.05153883e-01 4.66590881e-01 9.11877990e-01
-4.11907762e-01 5.58190048e-01 -2.63691723e-01 3.80257785e-01
-7.57034048e-02 -4.73490536e-01 7.07462251e-01 -1.99463010e-01
7.26344585e-02 1.15551166e-01 -4.84741241e-01 5.19107059e-02
1.19931626e+00 1.98177904e-01 -1.66767668e-02 -5.86288810e-01
-1.12486589e+00 2.52216239e-04 4.41456407e-01 -4.38908577e-01
-3.35323848e-02 -1.25464702e+00 -2.20098168e-01 -4.82507003e-03
-5.84473848e-01 -7.81927884e-01 7.56813958e-02 1.67943943e+00
-7.80591786e-01 8.41249108e-01 -2.73765773e-01 -3.25578630e-01
-7.13470876e-01 9.14794624e-01 3.74939799e-01 -5.45496762e-01
-4.02635753e-01 4.87198681e-01 2.09510073e-01 4.55877662e-01
-2.23578978e-02 -4.50771004e-01 1.13717549e-01 3.85208070e-01
5.39336681e-01 7.33479619e-01 -3.17273021e-01 -3.62648636e-01
-3.20602834e-01 6.24613643e-01 5.64825118e-01 -8.99315715e-01
1.30939984e+00 -3.48559767e-01 -3.86418611e-01 1.25059104e+00
1.46989369e+00 3.30575824e-01 -1.18555307e+00 6.74601793e-02
5.37254214e-01 -8.68892968e-02 -3.87535244e-01 -2.57838130e-01
-7.17366576e-01 8.62062633e-01 6.75464332e-01 1.00780535e+00
9.62977648e-01 -1.04151227e-01 3.58394146e-01 3.81470293e-01
6.19114578e-01 -1.02728224e+00 -4.23230648e-01 3.37178558e-01
8.84657562e-01 -5.85414648e-01 -2.21659884e-01 -1.81246981e-01
-3.15801054e-01 1.45273387e+00 -4.81705755e-01 -5.63977003e-01
1.27415419e+00 4.50106144e-01 1.74672544e-01 7.44252652e-02
-4.11367297e-01 -4.23282236e-01 -1.46281004e-01 6.99723735e-02
2.95678407e-01 -2.03347858e-02 -9.66278970e-01 5.56495428e-01
-4.26517516e-01 2.92381912e-01 5.61545432e-01 6.81381702e-01
-3.19838524e-01 -1.13715255e+00 -4.48709935e-01 3.42894316e-01
-8.12561154e-01 1.40571427e-02 7.31844306e-02 1.13966227e+00
-2.93989312e-02 6.19072616e-01 8.84689111e-03 -2.60136336e-01
2.83110797e-01 5.50533175e-01 3.28821868e-01 -3.53556842e-01
-4.72867131e-01 4.16178435e-01 -3.50362748e-01 -6.32832587e-01
-4.82569575e-01 -8.88486207e-01 -6.00170255e-01 -5.45453548e-01
-6.83721080e-02 6.36997342e-01 6.35112822e-01 9.73342001e-01
-4.93622094e-01 1.81475088e-01 9.39336956e-01 -7.08210230e-01
-1.28803027e+00 -9.85631824e-01 -1.53017199e+00 -2.11121924e-02
4.32571679e-01 -6.71188235e-01 -7.65714884e-01 -2.61002749e-01]
|
[5.017591953277588, 4.030539512634277]
|
af84fc08-8ca1-4c40-b563-9b4929435ab0
|
a-new-evaluation-method-evaluation-data-and
|
2205.00217
| null |
https://arxiv.org/abs/2205.00217v1
|
https://arxiv.org/pdf/2205.00217v1.pdf
|
A New Evaluation Method: Evaluation Data and Metrics for Chinese Grammar Error Correction
|
As a fundamental task in natural language processing, Chinese Grammatical Error Correction (CGEC) has gradually received widespread attention and become a research hotspot. However, one obvious deficiency for the existing CGEC evaluation system is that the evaluation values are significantly influenced by the Chinese word segmentation results or different language models. The evaluation values of the same error correction model can vary considerably under different word segmentation systems or different language models. However, it is expected that these metrics should be independent of the word segmentation results and language models, as they may lead to a lack of uniqueness and comparability in the evaluation of different methods. To this end, we propose three novel evaluation metrics for CGEC in two dimensions: reference-based and reference-less. In terms of the reference-based metric, we introduce sentence-level accuracy and char-level BLEU to evaluate the corrected sentences. Besides, in terms of the reference-less metric, we adopt char-level meaning preservation to measure the semantic preservation degree of the corrected sentences. We deeply evaluate and analyze the reasonableness and validity of the three proposed metrics, and we expect them to become a new standard for CGEC.
|
['Shengyi Jiang', 'Ziyu Yang', 'Xiaotian Lin', 'Nankai Lin']
|
2022-04-30
| null | null | null | null |
['chinese-word-segmentation']
|
['natural-language-processing']
|
[ 1.60479862e-02 -4.41914111e-01 1.46924376e-01 -5.81464767e-01
-5.28349936e-01 -3.95098537e-01 2.51944274e-01 6.53443456e-01
-8.22731256e-01 6.34470344e-01 2.65366465e-01 -1.67066589e-01
5.69290072e-02 -8.44683647e-01 -2.49867842e-01 -3.82665336e-01
6.86876476e-01 -3.88846770e-02 5.30552924e-01 -2.93147832e-01
7.37881243e-01 7.14430064e-02 -1.28052568e+00 5.03392294e-02
1.67262971e+00 6.42659605e-01 4.31799114e-01 4.27758068e-01
-4.98560399e-01 4.52747196e-01 -8.89329493e-01 -5.18278301e-01
-3.82787406e-01 -9.02212262e-01 -9.25557137e-01 -2.29442462e-01
-1.11883305e-01 2.81826079e-01 2.11156338e-01 1.70680809e+00
4.71040547e-01 -5.72866909e-02 4.08990234e-01 -8.31404448e-01
-9.04226184e-01 7.20199347e-01 -2.62523770e-01 1.20712541e-01
3.04799169e-01 -8.07752460e-02 9.71051991e-01 -7.53039300e-01
5.43801367e-01 1.19113863e+00 6.29310250e-01 5.62797308e-01
-6.05089307e-01 -4.33941036e-01 2.82968491e-01 3.47816169e-01
-1.35096383e+00 -9.03046653e-02 4.67463642e-01 -1.82500914e-01
6.09453440e-01 3.64253700e-01 4.22929138e-01 4.14454192e-01
4.10023957e-01 7.10973501e-01 1.10362959e+00 -7.50190914e-01
1.53243095e-01 1.95388243e-01 4.85896319e-01 4.11899030e-01
5.84208727e-01 -4.62893426e-01 -1.42205562e-02 3.50678861e-01
2.45949686e-01 -3.27132493e-02 -5.80893338e-01 2.12131307e-01
-1.00401056e+00 6.98243141e-01 3.01938176e-01 7.98803091e-01
7.33262077e-02 -5.18590584e-02 5.49961984e-01 1.22065239e-01
4.06753987e-01 4.78469193e-01 -3.91110480e-01 -3.70315105e-01
-8.22160602e-01 3.20370018e-01 5.21721125e-01 1.03880501e+00
6.58799052e-01 -9.85804647e-02 -4.24980134e-01 8.62900257e-01
3.34532052e-01 4.23618287e-01 8.54727268e-01 -5.69609642e-01
4.24827278e-01 7.82056451e-01 8.46303776e-02 -1.34663188e+00
-3.61782908e-01 -5.27506828e-01 -7.17026353e-01 -2.64090210e-01
2.94715405e-01 -1.69478208e-02 -5.61183631e-01 1.75471938e+00
-2.69387756e-02 -3.35206300e-01 -8.31579138e-03 9.58105266e-01
1.03558373e+00 6.55972838e-01 2.05856428e-01 -4.56277996e-01
1.22671247e+00 -1.08523440e+00 -1.14494956e+00 -1.71323389e-01
1.00944102e+00 -1.06633377e+00 1.40791380e+00 1.89477473e-01
-9.21766162e-01 -5.71360052e-01 -1.13233292e+00 -1.70506135e-01
-3.47873896e-01 2.11124644e-01 1.72504917e-01 7.67465472e-01
-7.21617043e-01 5.50763965e-01 -5.66414833e-01 -4.99784172e-01
-2.20913455e-01 -1.92299768e-01 2.28877626e-02 -1.22304350e-01
-1.37597942e+00 1.08075821e+00 6.35692835e-01 2.48688221e-01
-1.01858638e-01 -2.47384384e-01 -6.23977780e-01 2.31768176e-01
2.07964003e-01 -2.36052975e-01 1.30874205e+00 -9.36316907e-01
-1.19921076e+00 6.80861056e-01 -2.84543335e-01 -2.85932329e-02
4.32251036e-01 -1.71545908e-01 -9.16193604e-01 -1.15616284e-01
2.21245691e-01 1.67115882e-01 1.93412811e-01 -1.20244324e+00
-8.53207707e-01 -2.51451313e-01 -1.42551079e-01 2.13860855e-01
-4.09221679e-01 3.38795811e-01 -7.71555126e-01 -8.63013208e-01
3.27158391e-01 -6.45977080e-01 -6.53234497e-02 -2.22578600e-01
-2.34350115e-01 -2.62931764e-01 5.68719566e-01 -8.92635345e-01
1.99843073e+00 -2.28527474e+00 -2.14413311e-02 7.94361755e-02
4.20706868e-02 4.77075279e-01 -1.64632604e-01 2.84445107e-01
2.50296623e-01 5.70468426e-01 -5.90677381e-01 -8.95487592e-02
-1.19776979e-01 7.13747963e-02 1.54837295e-01 9.43606198e-02
2.12719038e-01 7.05160141e-01 -1.14455950e+00 -6.78556919e-01
-1.00737840e-01 1.49859503e-01 -4.60744798e-01 1.23228997e-01
-1.37924522e-01 3.19841564e-01 -5.32158077e-01 4.22777146e-01
9.07924652e-01 2.13823579e-02 6.56543672e-02 -5.91078624e-02
-4.10538942e-01 4.33954448e-01 -1.14559400e+00 1.50123155e+00
-2.73060560e-01 3.70833009e-01 -3.08689296e-01 -7.12580919e-01
1.12537193e+00 4.60486189e-02 -8.70810896e-02 -9.59892154e-01
2.70794690e-01 6.91780210e-01 1.45295918e-01 -4.43391621e-01
1.19490600e+00 -2.55468152e-02 -2.29759440e-01 3.26682955e-01
-1.87916309e-01 -1.91272393e-01 3.83363903e-01 1.56596199e-01
7.64048576e-01 -2.13321283e-01 3.96036118e-01 -5.01732111e-01
7.82371700e-01 -2.97083296e-02 9.49440241e-01 4.62847263e-01
-4.72637773e-01 7.37851799e-01 5.56169331e-01 -3.93355303e-02
-8.41401935e-01 -6.63658202e-01 -2.76406348e-01 6.27424657e-01
6.25484943e-01 -5.59296966e-01 -1.12493455e+00 -7.86951661e-01
-3.98156673e-01 1.03905106e+00 -2.55849451e-01 -3.81344974e-01
-5.84625125e-01 -8.66372168e-01 7.13086903e-01 3.94586772e-01
7.33337164e-01 -1.07591045e+00 -3.49427313e-01 2.45301962e-01
-4.99447942e-01 -1.05213404e+00 -6.43776834e-01 -2.57598758e-01
-8.50117981e-01 -1.00670826e+00 -5.76743484e-01 -9.70674098e-01
8.03069413e-01 4.47608203e-01 1.07334232e+00 8.32333088e-01
3.13263148e-01 -6.61318526e-02 -1.03930783e+00 -5.38951159e-01
-6.86375082e-01 1.90673396e-01 -1.40722677e-01 -2.50509650e-01
5.41885674e-01 1.24593295e-01 -6.08375490e-01 2.32117206e-01
-1.14340246e+00 3.41167266e-04 4.08280611e-01 5.63736916e-01
4.75285202e-01 1.01738065e-01 5.24255097e-01 -9.94630158e-01
1.08658969e+00 -2.21072942e-01 -4.15987015e-01 6.70540273e-01
-1.08076322e+00 1.09358281e-01 6.14400029e-01 1.22494131e-01
-1.11613357e+00 -6.89273894e-01 -5.42714894e-01 3.48216087e-01
1.10348776e-01 8.54241252e-01 -3.86157572e-01 4.02032211e-02
4.25156176e-01 2.59471089e-01 -2.63002783e-01 -6.12442195e-01
1.09640256e-01 8.32217157e-01 4.45628524e-01 -4.79913086e-01
3.22879195e-01 -1.95245862e-01 -4.71564829e-01 -5.48046649e-01
-8.06555450e-01 -2.74418443e-01 -5.39076447e-01 -1.24485016e-01
9.43791628e-01 -5.61111927e-01 -3.07820827e-01 7.55899608e-01
-1.57147563e+00 2.17785671e-01 3.61738503e-02 4.82064873e-01
1.73113510e-01 9.87510741e-01 -5.60844362e-01 -5.86347878e-01
-4.96715039e-01 -1.31922543e+00 6.18399680e-01 5.77274024e-01
-1.07822388e-01 -9.86658096e-01 -7.57495388e-02 -2.48182379e-02
4.49723721e-01 3.04115489e-02 1.18236244e+00 -5.17478943e-01
-2.03120813e-01 -2.33751878e-01 -4.61502403e-01 6.97980821e-01
1.45939156e-01 2.24763468e-01 -5.21522999e-01 -1.16628870e-01
1.52041227e-01 2.05544010e-01 7.65324295e-01 1.46997869e-01
1.18028402e+00 -2.22656980e-01 1.09249083e-02 2.42078379e-01
1.58807003e+00 3.75570476e-01 9.38574612e-01 5.11635303e-01
5.51767170e-01 5.52275538e-01 8.41565728e-01 1.70687333e-01
4.37678844e-01 5.38549304e-01 2.09068581e-01 5.53568937e-02
-8.45562965e-02 -3.32124323e-01 3.58146101e-01 1.79120135e+00
7.90811926e-02 -5.22698879e-01 -9.28255856e-01 4.98550206e-01
-1.74586987e+00 -7.80333400e-01 -6.96904004e-01 2.24203038e+00
9.54856634e-01 1.84903085e-01 -3.08822423e-01 2.51455158e-01
1.06197751e+00 -1.02411481e-02 -7.46900365e-02 -7.16850817e-01
-4.17612851e-01 8.57657716e-02 1.55633792e-01 5.43955147e-01
-7.10977972e-01 1.08988738e+00 6.09159994e+00 1.05622363e+00
-1.14177990e+00 3.30057681e-01 4.83516186e-01 5.20121396e-01
-7.01517999e-01 2.31352881e-01 -8.53395998e-01 8.82295907e-01
7.04990685e-01 -5.07976294e-01 -9.17196926e-03 5.97943127e-01
2.96350956e-01 -2.50510871e-01 -6.17040813e-01 7.80307651e-01
1.78132191e-01 -9.24901485e-01 1.50418937e-01 -3.63237768e-01
7.48875022e-01 -3.57274115e-01 -3.36699456e-01 2.50293791e-01
-1.37410432e-01 -7.26726532e-01 8.84229481e-01 5.27760446e-01
6.10848069e-01 -7.62407303e-01 1.17915893e+00 2.86226660e-01
-1.07881308e+00 3.50424170e-01 -6.00017607e-01 -1.51254147e-01
2.21926183e-01 7.46473610e-01 -9.63924900e-02 8.40194881e-01
6.21199250e-01 5.57218373e-01 -7.86936879e-01 1.20343304e+00
-6.02120578e-01 6.31985247e-01 2.48001307e-01 -5.61248183e-01
2.21487612e-01 -4.32748109e-01 3.11414570e-01 1.49096406e+00
6.89943492e-01 3.24421190e-02 -7.95886368e-02 8.28501642e-01
-1.75954670e-01 6.10021710e-01 3.61925997e-02 -1.24559253e-01
6.57838166e-01 9.88100529e-01 -7.39770591e-01 -3.06257695e-01
-4.46442574e-01 1.06582677e+00 3.01629335e-01 1.15945004e-01
-7.87332058e-01 -8.72900069e-01 5.23515224e-01 -1.89696893e-01
4.18226933e-03 -2.15534911e-01 -7.77730346e-01 -1.39256585e+00
2.48525321e-01 -8.50150049e-01 1.48319453e-01 -5.70849240e-01
-1.05179739e+00 7.45432317e-01 -3.43489230e-01 -1.35289061e+00
3.01973015e-01 -3.97796422e-01 -7.35137939e-01 9.10726011e-01
-1.53033888e+00 -5.95578313e-01 -5.19755960e-01 1.85720682e-01
4.61928546e-01 5.84825389e-02 7.37515450e-01 5.50593674e-01
-8.74077916e-01 1.01420999e+00 2.68008143e-01 2.70691007e-01
7.38138616e-01 -9.95002687e-01 3.16918880e-01 1.49342000e+00
-1.26395136e-01 6.89928174e-01 7.69681633e-01 -6.61929429e-01
-7.18451381e-01 -1.11068296e+00 1.54135931e+00 -1.85329229e-01
3.48391950e-01 1.28513342e-03 -1.25138688e+00 2.32054710e-01
9.42260772e-02 -4.15321112e-01 4.05000478e-01 2.48433594e-02
-4.23733555e-02 -1.51682988e-01 -1.09774137e+00 6.63952291e-01
7.30125308e-01 -3.14403981e-01 -7.14045405e-01 3.45243774e-02
1.16073596e+00 -2.98041642e-01 -7.35340357e-01 3.82972956e-01
3.44723433e-01 -9.75440681e-01 2.01441184e-01 -2.48690233e-01
6.63067758e-01 -6.34850740e-01 -7.27543160e-02 -1.35470271e+00
-4.41469491e-01 -1.99518278e-02 4.54421312e-01 1.65574312e+00
5.10842562e-01 -6.20648444e-01 2.10737854e-01 5.36019266e-01
-5.09721756e-01 -6.00159943e-01 -8.10609341e-01 -8.18677783e-01
4.13430929e-01 -6.06110811e-01 1.06833899e+00 8.60149801e-01
6.22772649e-02 1.61832869e-01 -2.15664610e-01 -5.38154207e-02
8.18900093e-02 -4.02575545e-02 3.43950272e-01 -1.07868588e+00
2.54847351e-02 -7.07469523e-01 -2.32809797e-01 -1.02759075e+00
-3.96793745e-02 -8.06631446e-01 3.25302184e-01 -1.67562485e+00
3.29504043e-01 -5.27462125e-01 -4.73646402e-01 1.68943971e-01
-6.82705224e-01 8.62960592e-02 3.95266563e-01 2.88061708e-01
-6.27811134e-01 8.51351619e-01 1.30710590e+00 3.69889811e-02
-8.04445893e-02 -2.44157329e-01 -8.89407694e-01 6.76350355e-01
8.00288379e-01 -5.85414648e-01 -9.16112028e-03 -9.69473898e-01
4.36540872e-01 -4.20208037e-01 -8.82191658e-02 -9.80333865e-01
2.48212993e-01 -2.90098697e-01 -1.40407696e-01 -2.53551126e-01
-5.58001995e-01 -5.11366963e-01 -2.33022004e-01 6.11813724e-01
-2.40206286e-01 4.13181812e-01 -7.41168633e-02 3.03415179e-01
-6.16350830e-01 -7.04245567e-01 9.61705863e-01 -1.88541427e-01
-1.01906049e+00 -1.96340457e-02 -2.57756710e-01 3.45202863e-01
7.40317822e-01 -3.37741137e-01 -3.50460589e-01 -1.61523879e-01
-2.95305345e-02 2.16150835e-01 7.22801626e-01 5.00309885e-01
6.21880591e-01 -1.33841777e+00 -8.32802296e-01 -2.81267217e-03
4.31502044e-01 -8.10420364e-02 2.15599418e-01 7.85335839e-01
-9.60356236e-01 3.66897464e-01 -4.40617390e-02 -3.07203531e-01
-1.10202992e+00 2.18537569e-01 1.62272319e-01 -3.49618673e-01
-1.70891210e-01 6.15480602e-01 -1.04166806e-01 -5.20494163e-01
-4.77802306e-02 -5.43381989e-01 -4.32682544e-01 -3.34595650e-01
5.78987181e-01 3.95084590e-01 2.61641324e-01 -7.35800028e-01
-3.19382578e-01 5.47326028e-01 -8.03534538e-02 1.48738250e-01
8.98141623e-01 -4.32465047e-01 -6.21192217e-01 4.09980148e-01
8.40053618e-01 2.69032151e-01 -4.41143245e-01 -2.11032376e-01
4.30594355e-01 -5.20598114e-01 -8.70262533e-02 -9.11213636e-01
-9.41590190e-01 9.18960094e-01 4.86625254e-01 1.33325055e-01
1.31304991e+00 -3.79214346e-01 1.06058156e+00 1.14582174e-01
4.38923120e-01 -1.44226074e+00 -2.76571095e-01 9.99881923e-01
7.15951145e-01 -1.19954157e+00 -6.49488196e-02 -5.80774605e-01
-6.58278048e-01 1.23497832e+00 7.37057090e-01 1.59610018e-01
4.90582705e-01 -1.75964877e-01 2.14277774e-01 1.39662251e-01
-1.93974882e-01 -1.26876876e-01 4.31587636e-01 3.01073730e-01
1.05947888e+00 1.12249091e-01 -1.59302390e+00 9.44688201e-01
-3.29378724e-01 -1.47375867e-01 6.44054770e-01 6.65350795e-01
-7.72338033e-01 -1.38895273e+00 -8.96837935e-02 2.65268683e-01
-4.85297173e-01 -1.47471845e-01 -2.28280544e-01 6.77675545e-01
2.44607747e-01 1.22794425e+00 -1.75136909e-01 -5.12615860e-01
4.33991671e-01 -1.14308052e-01 1.74843863e-01 -7.13305831e-01
-6.54652894e-01 -2.56339788e-01 -1.07463770e-01 -2.06499085e-01
-4.55543995e-01 -5.10564625e-01 -1.59549880e+00 -5.45012057e-01
-5.98375201e-01 5.99223554e-01 5.55563867e-01 1.19270909e+00
2.90504336e-01 4.53200430e-01 3.94662380e-01 2.39695683e-02
-4.43433493e-01 -1.09276855e+00 -6.28964901e-01 6.17633343e-01
-2.00120687e-01 -2.99946368e-01 -2.33696610e-01 -3.10229480e-01]
|
[10.985759735107422, 10.76330852508545]
|
bf1653e5-490a-4458-9678-ed8e544465a5
|
convex-combination-consistency-between
|
2205.00400
| null |
https://arxiv.org/abs/2205.00400v1
|
https://arxiv.org/pdf/2205.00400v1.pdf
|
Convex Combination Consistency between Neighbors for Weakly-supervised Action Localization
|
In weakly-supervised temporal action localization (WS-TAL), the methods commonly follow the "localization by classification" procedure, which uses the snippet predictions to form video class scores and then optimizes a video classification loss. In this procedure, the snippet predictions (or snippet attention weights) are used to separate foreground and background. However, the snippet predictions are usually inaccurate due to absence of frame-wise labels, and then the overall performance is hindered. In this paper, we propose a novel C$^3$BN to achieve robust snippet predictions. C$^3$BN includes two key designs by exploring the inherent characteristics of video data. First, because of the natural continuity of adjacent snippets, we propose a micro data augmentation strategy to increase the diversity of snippets with convex combination of adjacent snippets. Second, we propose a macro-micro consistency regularization strategy to force the model to be invariant (or equivariant) to the transformations of snippets with respect to video semantics, snippet predictions and snippet features. Experimental results demonstrate the effectiveness of our proposed method on top of baselines for the WS-TAL tasks with video-level and point-level supervision.
|
['Zhilin Li', 'Ruoxi Chen', 'Zilei Wang', 'Qinying Liu']
|
2022-05-01
| null | null | null | null |
['weakly-supervised-action-localization', 'weakly-supervised-temporal-action', 'action-localization']
|
['computer-vision', 'computer-vision', 'computer-vision']
|
[ 4.69279349e-01 -5.82345650e-02 -6.27692461e-01 -2.81255215e-01
-5.23983836e-01 -1.56423911e-01 3.91251951e-01 -2.23289147e-01
-1.80747509e-01 4.89985019e-01 3.74810547e-01 1.74509570e-01
-5.75649068e-02 -2.28898153e-01 -1.08558786e+00 -7.29166329e-01
-2.37403676e-01 -2.02697188e-01 6.31929994e-01 7.39449561e-02
1.28316134e-01 1.18333399e-01 -1.61808550e+00 6.23034120e-01
7.77501941e-01 1.23251009e+00 3.38661432e-01 5.36566436e-01
-5.41538559e-02 1.21608496e+00 -9.20168385e-02 -2.17820808e-01
5.82107425e-01 -5.86866975e-01 -5.59493721e-01 4.20714259e-01
7.59427547e-01 -3.03895593e-01 -5.35514891e-01 1.15918612e+00
6.05115965e-02 3.62978965e-01 3.11438680e-01 -1.47273624e+00
-5.03970385e-01 3.12738627e-01 -8.24654818e-01 3.33895952e-01
4.45382625e-01 2.77297080e-01 1.24354708e+00 -1.19602287e+00
6.86282218e-01 1.29365242e+00 7.52726316e-01 4.68150109e-01
-1.14940095e+00 -6.08654499e-01 6.91793323e-01 4.32319850e-01
-1.48817325e+00 -6.14221036e-01 7.73385823e-01 -5.79345345e-01
6.96968496e-01 1.68622330e-01 7.31228113e-01 1.12165546e+00
7.17188120e-02 1.18643522e+00 7.53289819e-01 -2.33484104e-01
1.57479092e-01 8.49849954e-02 8.96369144e-02 9.88991678e-01
-2.12385148e-01 -1.08983278e-01 -9.32703555e-01 1.45919120e-03
8.66784573e-01 3.67342323e-01 -4.50708866e-01 -6.62188649e-01
-1.41617370e+00 5.16074479e-01 2.92354912e-01 3.07091456e-02
-2.12461248e-01 1.93750307e-01 4.40409005e-01 8.15039724e-02
4.86480504e-01 -1.08408004e-01 -4.18317348e-01 -1.84827134e-01
-9.05379653e-01 1.06473304e-01 2.51221120e-01 1.18590713e+00
6.90617621e-01 9.07079056e-02 -3.46072644e-01 7.67018974e-01
2.48368293e-01 3.38438869e-01 6.27826989e-01 -1.00801361e+00
8.22127581e-01 6.00022316e-01 1.39709681e-01 -1.24323463e+00
-2.82706134e-02 -1.53906792e-01 -7.21566379e-01 1.95836797e-02
3.54567766e-01 2.52338856e-01 -7.72172987e-01 1.78085339e+00
2.62384057e-01 7.06898332e-01 -1.32230148e-01 9.22761083e-01
4.52708066e-01 6.48461103e-01 1.54227287e-01 -2.68057197e-01
1.03282118e+00 -1.20335734e+00 -7.12250531e-01 -2.11963356e-01
6.05918348e-01 -4.16783333e-01 1.57959282e+00 6.07413985e-02
-1.06259692e+00 -7.24018276e-01 -9.36601996e-01 7.63200670e-02
7.31345778e-03 2.13296369e-01 2.05131769e-01 1.66524723e-01
-8.31293046e-01 5.84401786e-01 -1.02665865e+00 -3.52099389e-01
6.98670268e-01 1.69453233e-01 -4.01329041e-01 -5.93350902e-02
-8.46348464e-01 4.82839882e-01 5.68926871e-01 1.46128297e-01
-8.60305905e-01 -7.08638370e-01 -1.03222859e+00 -6.35677809e-03
7.68055737e-01 -4.00530130e-01 1.00160885e+00 -1.55160046e+00
-1.40155268e+00 5.65091729e-01 -4.63854879e-01 -5.16314745e-01
7.25267291e-01 -2.42576614e-01 -3.06685537e-01 3.28945786e-01
3.25781912e-01 7.19194770e-01 1.13202548e+00 -9.78961527e-01
-1.07533360e+00 -1.43686846e-01 -7.81000406e-02 3.76047701e-01
-5.08352876e-01 1.78608466e-02 -7.36799479e-01 -8.36484313e-01
2.97774434e-01 -8.23166966e-01 -1.34598807e-01 1.62052602e-01
-3.27123970e-01 -1.64210796e-01 1.02250326e+00 -7.20109820e-01
1.27819204e+00 -2.51121092e+00 1.89774826e-01 2.04309508e-01
2.12117046e-01 9.74264517e-02 -1.75382510e-01 -6.84626922e-02
-1.48147598e-01 -8.82226229e-02 -2.17186883e-01 -2.56298065e-01
-4.44883592e-02 1.80502832e-01 -2.54039109e-01 4.60099965e-01
2.85362512e-01 7.60462105e-01 -1.00225532e+00 -5.92886031e-01
3.63406330e-01 1.77080765e-01 -7.21903145e-01 1.50426075e-01
-4.25766051e-01 3.89196604e-01 -5.82440853e-01 7.62677133e-01
3.13743204e-01 -3.16219330e-01 -2.43314896e-02 -3.06311548e-01
-8.86275396e-02 9.38157365e-02 -1.21343493e+00 1.59376204e+00
1.19234040e-01 5.28144062e-01 -7.83475339e-02 -1.15103853e+00
5.79909205e-01 2.01945662e-01 7.44990885e-01 -4.35867280e-01
-1.85908109e-01 -9.53970384e-03 -3.33287627e-01 -6.58610344e-01
2.36198559e-01 1.23030264e-02 9.97399688e-02 1.66216809e-02
4.74159606e-02 4.72073317e-01 1.80322126e-01 3.36451918e-01
1.09093785e+00 5.68131447e-01 2.71977872e-01 -1.01652697e-01
5.73229074e-01 -1.11161038e-01 1.10893595e+00 6.08567059e-01
-5.29311657e-01 7.41200566e-01 6.24433994e-01 -4.19273853e-01
-8.86640549e-01 -1.08729184e+00 2.17102155e-01 1.27804208e+00
3.18568140e-01 -5.98218024e-01 -8.47387016e-01 -1.06776297e+00
-2.25107312e-01 2.88494051e-01 -6.07153654e-01 -2.94413775e-01
-6.67968988e-01 -4.07888919e-01 1.89337388e-01 8.95187259e-01
7.69552052e-01 -1.16316354e+00 -4.02277917e-01 1.96020737e-01
-2.77912706e-01 -1.39983881e+00 -1.06204057e+00 -7.97853395e-02
-7.79348671e-01 -1.09040797e+00 -4.62440759e-01 -8.79819751e-01
8.43402505e-01 4.74965513e-01 7.33969152e-01 5.54592535e-02
-3.08974404e-02 5.03392398e-01 -5.50659895e-01 -1.98665299e-02
6.91907182e-02 -3.67977589e-01 2.87550032e-01 6.46734536e-01
2.87728608e-01 -4.32580978e-01 -6.15469754e-01 5.76442301e-01
-8.74087095e-01 2.71973193e-01 5.19203424e-01 7.98175752e-01
1.02598262e+00 1.55033655e-02 4.69456285e-01 -6.47380948e-01
-9.07318667e-02 -2.54123896e-01 -4.65778798e-01 3.20075691e-01
-3.79619330e-01 -1.41537949e-01 7.17298150e-01 -6.19883418e-01
-9.21173036e-01 4.08757687e-01 2.24699184e-01 -9.64473188e-01
-5.87666687e-03 4.23229903e-01 -4.37843591e-01 -2.82472912e-02
2.37970531e-01 3.74075532e-01 1.71255786e-02 -2.11745411e-01
2.75199622e-01 2.24053085e-01 5.20911753e-01 -4.27976966e-01
5.50036907e-01 5.72260678e-01 -3.43956530e-01 -8.17213297e-01
-1.06621313e+00 -5.98019779e-01 -6.97430551e-01 -5.57414711e-01
1.12063360e+00 -9.19225037e-01 -4.22239453e-01 2.30471179e-01
-7.96058893e-01 -2.89264590e-01 -4.46405739e-01 5.91741383e-01
-7.79504836e-01 6.20897055e-01 -4.40209478e-01 -6.61131859e-01
1.19356669e-01 -1.19372845e+00 9.30819333e-01 5.28044514e-02
-1.11184917e-01 -7.53050685e-01 -3.05377513e-01 4.36048687e-01
-1.35714248e-01 1.36626139e-01 6.37121379e-01 -6.10534608e-01
-7.44322181e-01 -2.27685586e-01 -3.18207651e-01 6.32219732e-01
3.72432694e-02 -4.22125310e-02 -6.57388031e-01 -2.10812390e-01
-7.47259855e-02 -1.19597346e-01 8.79249394e-01 5.38076460e-01
1.60051477e+00 -5.52374363e-01 -3.39499474e-01 7.30966985e-01
1.04188275e+00 9.65065509e-02 4.80108112e-01 3.63848448e-01
1.07093966e+00 5.86904109e-01 7.91599393e-01 4.78604078e-01
3.94607127e-01 7.31459737e-01 4.39391613e-01 1.20842971e-01
-1.15735106e-01 -5.51200688e-01 8.28434885e-01 6.57859325e-01
-3.27510238e-01 2.59718485e-02 -5.98065257e-01 4.21319604e-01
-2.23634696e+00 -1.23704851e+00 3.90650593e-02 2.24671412e+00
5.26837528e-01 2.31784999e-01 2.43267536e-01 -1.79394826e-01
9.55372691e-01 3.48820567e-01 -5.28793395e-01 3.54505271e-01
-7.83976614e-02 -2.66913950e-01 6.49291039e-01 2.18494639e-01
-1.27578998e+00 1.05330539e+00 5.37807512e+00 9.76487160e-01
-7.88473368e-01 1.52992681e-01 4.84294742e-01 -2.34389469e-01
1.89215466e-01 1.42435282e-02 -8.64965618e-01 7.63243258e-01
3.98101598e-01 1.52449772e-01 3.88188779e-01 8.24690044e-01
5.98563910e-01 1.61102906e-01 -1.45354259e+00 9.04232621e-01
2.74238646e-01 -1.37686110e+00 3.24155614e-02 -1.89780444e-01
7.34797060e-01 -2.29852006e-01 1.41410939e-02 3.91295373e-01
1.22196034e-01 -5.34764230e-01 1.04720855e+00 5.16764522e-01
5.95041394e-01 -4.88076508e-01 4.28019941e-01 2.75468111e-01
-1.43752170e+00 -3.00553203e-01 -2.51246750e-01 1.51647404e-01
1.65772483e-01 4.18519676e-02 -4.38978255e-01 2.44616985e-01
1.01907325e+00 1.45275521e+00 -4.26899672e-01 9.18701887e-01
1.71125606e-02 6.09362483e-01 -1.69605732e-01 2.96389610e-01
3.61169219e-01 -2.94601053e-01 5.96864343e-01 1.19633377e+00
2.44537428e-01 2.14428052e-01 6.52183712e-01 6.26255214e-01
-5.80629893e-02 6.00980669e-02 -4.84676570e-01 8.61596987e-02
2.64907449e-01 9.16402459e-01 -6.55205667e-01 -4.74530131e-01
-7.78968036e-01 1.06528282e+00 3.07091177e-01 6.17904663e-01
-1.01759946e+00 1.09188378e-01 6.66971803e-01 3.89817417e-01
6.06198430e-01 -1.33888781e-01 8.99111852e-02 -1.43413484e+00
3.94265622e-01 -9.00528669e-01 5.64393103e-01 -7.28059888e-01
-1.29599023e+00 3.29523921e-01 8.43758434e-02 -1.64348900e+00
1.74568921e-01 -4.26713884e-01 -6.00835204e-01 8.11163429e-03
-1.29899025e+00 -1.06885362e+00 -2.76986271e-01 1.01081944e+00
9.93453205e-01 -2.12151408e-01 2.59621620e-01 2.55130053e-01
-8.53152812e-01 5.99474609e-01 1.17107995e-01 2.89755344e-01
6.25210226e-01 -1.00024784e+00 -1.06194608e-01 9.73941088e-01
2.35736609e-01 2.96608150e-01 4.81943518e-01 -7.34095335e-01
-1.15071106e+00 -1.46113133e+00 5.61977863e-01 -4.37502950e-01
9.28319275e-01 -4.76592183e-01 -8.72173846e-01 1.19380653e+00
-1.60097107e-01 4.60352778e-01 4.99745518e-01 -1.23579085e-01
-3.43581140e-01 -3.22233170e-01 -8.97949338e-01 8.56955111e-01
1.31529582e+00 -4.27357823e-01 -5.40922523e-01 5.52159309e-01
7.63107657e-01 -2.64318556e-01 -6.01254582e-01 4.85289007e-01
5.93137860e-01 -9.41471398e-01 1.09436858e+00 -6.23560131e-01
3.86282623e-01 -4.58673894e-01 -3.54752094e-01 -8.05009067e-01
-2.74246633e-01 -6.15694225e-01 -3.90216321e-01 1.15346837e+00
2.53860503e-01 -3.51990581e-01 9.65877652e-01 6.24935091e-01
-4.19699252e-01 -5.96222878e-01 -9.96831298e-01 -8.90488565e-01
-6.12564921e-01 -5.06105542e-01 7.14591891e-02 8.64154100e-01
4.24862690e-02 -7.84058422e-02 -7.74395823e-01 2.30107158e-01
5.05017638e-01 -1.65292323e-01 5.81965089e-01 -7.27795601e-01
-3.78616691e-01 -5.26272357e-01 -6.73176706e-01 -1.26685727e+00
2.04866409e-01 -7.38350451e-01 3.43645900e-01 -1.04392493e+00
3.44129711e-01 -2.80089498e-01 -6.15591347e-01 5.49600303e-01
-1.74015701e-01 2.73223966e-01 3.53051424e-01 4.70264643e-01
-1.25933576e+00 6.62328482e-01 1.02886236e+00 -2.89337616e-02
-3.47911775e-01 4.66714017e-02 -3.30452979e-01 1.20822084e+00
4.62187648e-01 -3.13704491e-01 -4.30225849e-01 -2.58589417e-01
-2.36620814e-01 -3.30277160e-02 6.46218479e-01 -1.01990485e+00
9.04928967e-02 -3.92504215e-01 4.08995658e-01 -4.92162555e-01
2.91714638e-01 -9.19277668e-01 -1.22873046e-01 6.01676218e-02
-6.90252602e-01 -6.33710399e-02 -1.96974024e-01 1.07957780e+00
-2.82462776e-01 -4.59892228e-02 6.87886417e-01 -2.86247227e-02
-1.06055343e+00 6.64993703e-01 -2.36533850e-01 4.09374088e-02
1.30918896e+00 -6.13146424e-01 4.82646078e-02 -1.41866997e-01
-8.80300939e-01 3.60557169e-01 3.46158296e-01 4.27075803e-01
8.19727659e-01 -1.54097700e+00 -4.51690078e-01 3.53637755e-01
2.47466788e-01 -2.52907369e-02 2.04057321e-01 1.21967387e+00
-1.95972264e-01 2.87862197e-02 -5.98703884e-02 -7.82078207e-01
-1.35252678e+00 5.57792723e-01 2.87665933e-01 3.88058200e-02
-1.07300222e+00 8.79131258e-01 6.55876815e-01 4.07394804e-02
6.71671331e-01 -5.02802670e-01 -9.56991017e-02 -1.68614268e-01
5.14058411e-01 2.67344415e-01 -3.14708889e-01 -7.39178836e-01
-3.18910182e-01 6.26637995e-01 -1.16484396e-01 1.40833780e-01
1.25842214e+00 -2.82411903e-01 1.60410270e-01 3.78115147e-01
1.06476092e+00 -1.78539693e-01 -1.97481847e+00 -3.94961089e-01
5.38317636e-02 -8.25313449e-01 -5.92835397e-02 -3.23678792e-01
-1.15964830e+00 4.99341637e-01 4.90854740e-01 -9.14476626e-03
1.17782092e+00 1.97281748e-01 7.28260279e-01 2.28135079e-01
1.44841105e-01 -1.29725373e+00 4.37277913e-01 3.07016313e-01
7.09412038e-01 -1.29997218e+00 -2.89130479e-01 -4.77552533e-01
-9.69969153e-01 7.63724625e-01 9.76154387e-01 -1.95907146e-01
4.80835497e-01 -1.92135833e-02 -2.84971625e-01 -8.38042516e-03
-6.31361008e-01 -1.13930650e-01 4.54347134e-01 4.55259383e-01
1.98394254e-01 -2.06815466e-01 -5.63774295e-02 6.67369783e-01
3.82719934e-01 3.08605786e-02 1.97590500e-01 1.00557625e+00
-6.02773428e-01 -6.80152297e-01 -2.62468904e-01 4.81075019e-01
-3.01114351e-01 -8.22895169e-02 -5.77645190e-02 7.21552908e-01
3.32325637e-01 7.26232588e-01 -5.98607734e-02 -4.72816676e-01
4.26501125e-01 6.19217716e-02 2.66556680e-01 -5.65288782e-01
-1.75713524e-01 2.86520928e-01 -1.87868327e-01 -9.69017327e-01
-6.96507275e-01 -9.86126959e-01 -1.31958675e+00 -1.14036789e-02
-2.59551674e-01 -7.50518441e-02 9.37604755e-02 1.11311758e+00
3.25675845e-01 3.93678039e-01 5.40850401e-01 -7.67082393e-01
-3.21781248e-01 -7.35945940e-01 -6.08861148e-01 6.54668510e-01
3.40705782e-01 -8.49981964e-01 -3.29045147e-01 5.91081381e-01]
|
[8.51846694946289, 0.675515353679657]
|
386b7ae5-2a67-46d9-8351-c18868148803
|
dynamic-kernel-distillation-for-efficient
|
1908.09216
| null |
https://arxiv.org/abs/1908.09216v1
|
https://arxiv.org/pdf/1908.09216v1.pdf
|
Dynamic Kernel Distillation for Efficient Pose Estimation in Videos
|
Existing video-based human pose estimation methods extensively apply large networks onto every frame in the video to localize body joints, which suffer high computational cost and hardly meet the low-latency requirement in realistic applications. To address this issue, we propose a novel Dynamic Kernel Distillation (DKD) model to facilitate small networks for estimating human poses in videos, thus significantly lifting the efficiency. In particular, DKD introduces a light-weight distillator to online distill pose kernels via leveraging temporal cues from the previous frame in a one-shot feed-forward manner. Then, DKD simplifies body joint localization into a matching procedure between the pose kernels and the current frame, which can be efficiently computed via simple convolution. In this way, DKD fast transfers pose knowledge from one frame to provide compact guidance for body joint localization in the following frame, which enables utilization of small networks in video-based pose estimation. To facilitate the training process, DKD exploits a temporally adversarial training strategy that introduces a temporal discriminator to help generate temporally coherent pose kernels and pose estimation results within a long range. Experiments on Penn Action and Sub-JHMDB benchmarks demonstrate outperforming efficiency of DKD, specifically, 10x flops reduction and 2x speedup over previous best model, and its state-of-the-art accuracy.
|
['Yuncheng Li', 'Xuecheng Nie', 'Linjie Luo', 'Jiashi Feng', 'Ning Zhang']
|
2019-08-24
|
dynamic-kernel-distillation-for-efficient-1
|
http://openaccess.thecvf.com/content_ICCV_2019/html/Nie_Dynamic_Kernel_Distillation_for_Efficient_Pose_Estimation_in_Videos_ICCV_2019_paper.html
|
http://openaccess.thecvf.com/content_ICCV_2019/papers/Nie_Dynamic_Kernel_Distillation_for_Efficient_Pose_Estimation_in_Videos_ICCV_2019_paper.pdf
|
iccv-2019-10
|
['2d-human-pose-estimation']
|
['computer-vision']
|
[-1.88965395e-01 -1.55367330e-01 -2.85829216e-01 -1.88706398e-01
-5.78258812e-01 -2.90028691e-01 1.16587162e-01 -4.41937417e-01
-7.34684825e-01 5.93264461e-01 3.30421269e-01 3.25280666e-01
1.82552367e-01 -6.38213634e-01 -9.20618951e-01 -7.18648851e-01
-2.10457727e-01 1.75595194e-01 4.37454611e-01 -1.33283541e-01
-4.42547292e-01 2.99078882e-01 -9.68327880e-01 -3.89491357e-02
5.96814752e-01 9.84592378e-01 6.53703213e-02 5.71426809e-01
3.51151496e-01 7.60251522e-01 -4.68447059e-01 -3.04037392e-01
4.31786984e-01 -2.27490157e-01 -5.49613357e-01 -2.99533159e-02
6.62225366e-01 -8.87145758e-01 -1.08977699e+00 8.42829823e-01
7.91342199e-01 3.54958147e-01 1.21075846e-01 -1.08779383e+00
-3.37925881e-01 3.75716954e-01 -6.69003367e-01 2.32677311e-01
5.18807292e-01 5.35136104e-01 4.88130540e-01 -6.70651138e-01
5.01701415e-01 1.28560638e+00 9.16195452e-01 7.44674027e-01
-1.08377254e+00 -8.04735303e-01 3.84606272e-01 1.61447108e-01
-1.49918509e+00 -2.71945626e-01 7.27013409e-01 -1.43217310e-01
7.18034565e-01 -1.00904116e-02 1.13064194e+00 1.07471454e+00
7.33033493e-02 1.04681253e+00 5.10316849e-01 1.92872621e-02
-1.06254913e-01 -8.32788467e-01 -3.29851568e-01 1.05351186e+00
3.67832482e-02 8.73319358e-02 -6.45979822e-01 1.26837492e-01
1.44363940e+00 3.01443160e-01 -5.48458993e-01 -4.24497843e-01
-1.39195704e+00 6.28018737e-01 7.42719829e-01 -1.34395972e-01
-4.81012821e-01 7.17990160e-01 7.15834677e-01 2.47140229e-01
3.19780558e-01 2.09647845e-02 -3.42875302e-01 -1.86236456e-01
-8.30750406e-01 4.57681179e-01 5.43163776e-01 8.12904477e-01
5.35087824e-01 1.68665633e-01 -4.20989066e-01 5.58103561e-01
2.27061242e-01 6.87379062e-01 4.55523878e-01 -1.04242349e+00
6.76515937e-01 3.94041657e-01 2.76768487e-02 -1.20516860e+00
-4.17068422e-01 -3.19077551e-01 -9.49212909e-01 -9.67534631e-02
5.54362595e-01 -2.94849664e-01 -9.98629987e-01 1.88856351e+00
8.85613441e-01 7.56051064e-01 -1.90849543e-01 1.41174316e+00
6.06982648e-01 5.98224163e-01 -5.36982261e-04 8.18045214e-02
1.43235028e+00 -1.32377338e+00 -4.66860324e-01 -3.10134262e-01
6.59091115e-01 -4.30784494e-01 8.95228922e-01 2.13596344e-01
-1.03978169e+00 -8.25418591e-01 -1.01472223e+00 -1.63547039e-01
2.49811366e-01 3.57242227e-01 9.59233999e-01 3.77701610e-01
-6.77177906e-01 6.32983148e-01 -1.48737574e+00 -6.63257465e-02
3.65854591e-01 6.23351812e-01 -5.40258408e-01 -1.26811936e-01
-1.26532662e+00 3.96621406e-01 5.34681380e-01 5.57256520e-01
-1.01669741e+00 -7.83241332e-01 -1.24277878e+00 -1.46640927e-01
7.72612035e-01 -9.42048728e-01 1.17509115e+00 -8.43404293e-01
-1.83220291e+00 3.47428620e-01 2.17670754e-01 -5.83201170e-01
7.12543786e-01 -8.83542418e-01 -2.66756237e-01 6.12449169e-01
-3.92248072e-02 7.84586668e-01 1.04229331e+00 -4.48022038e-01
-4.14859146e-01 -3.33941936e-01 3.10536623e-01 4.42477793e-01
-3.98388416e-01 -2.27475807e-01 -1.33838272e+00 -1.03841019e+00
7.29334578e-02 -1.19834924e+00 -3.98702145e-01 5.83071709e-01
-1.78550407e-01 -1.93844244e-01 7.44313061e-01 -9.48932409e-01
1.25354624e+00 -2.04709649e+00 4.13428187e-01 7.62209622e-03
3.72913748e-01 3.99522990e-01 -1.28774703e-01 6.13734638e-03
-9.29426809e-04 -6.59802496e-01 6.53959960e-02 -3.76706421e-01
-7.91831911e-02 3.97656441e-01 -1.47767249e-03 8.18597019e-01
1.05197892e-01 1.21052551e+00 -9.77981806e-01 -6.31294847e-01
3.60280931e-01 6.64030969e-01 -8.03493261e-01 3.04303139e-01
-1.12879455e-01 4.82238591e-01 -5.64312696e-01 5.15905142e-01
5.36725521e-01 -2.72537798e-01 4.58876900e-02 -7.52320945e-01
2.28954375e-01 -2.04314608e-02 -1.18869817e+00 2.47330403e+00
-3.98582131e-01 1.78674817e-01 1.46205677e-02 -9.45857167e-01
3.62108588e-01 2.66236603e-01 6.32524133e-01 -4.71145481e-01
1.39782220e-01 -8.77721682e-02 -1.67794779e-01 -3.01398695e-01
1.80080771e-01 9.48697329e-02 -2.17178062e-01 1.20022058e-01
2.33165994e-01 3.74632359e-01 7.38299340e-02 5.98850101e-02
1.14104402e+00 6.98792279e-01 7.26902261e-02 7.49306753e-02
5.57407558e-01 -3.28497201e-01 8.19302917e-01 3.93286765e-01
-3.79603028e-01 5.04905701e-01 1.98460475e-01 -7.52338946e-01
-7.42909729e-01 -1.24662805e+00 5.54469526e-01 1.05518281e+00
3.21915597e-01 -3.70245665e-01 -8.28669190e-01 -8.40671837e-01
3.80389132e-02 -2.82897294e-01 -7.28141129e-01 -4.29129064e-01
-1.17315400e+00 -3.35931897e-01 7.07664192e-01 9.89919960e-01
7.87792921e-01 -7.87352145e-01 -7.17623532e-01 3.03282052e-01
-2.73097664e-01 -1.20259869e+00 -1.10105336e+00 -2.33965412e-01
-8.71724606e-01 -9.07769620e-01 -1.21469140e+00 -9.11584854e-01
6.38808966e-01 4.49844562e-02 7.60245919e-01 -1.07074641e-01
-5.17770112e-01 1.46017164e-01 -1.83348820e-01 4.16085720e-01
2.06385598e-01 7.82432333e-02 2.63481468e-01 -1.48718476e-01
1.52381107e-01 -6.15636885e-01 -1.17918825e+00 3.30496609e-01
-8.21289361e-01 2.02787548e-01 6.75641775e-01 9.38240409e-01
5.38519681e-01 -1.92034706e-01 1.82706460e-01 -2.86329091e-01
1.11175403e-01 -1.51138902e-01 -4.40368980e-01 1.58109710e-01
-9.97118838e-03 1.14049248e-01 7.75708079e-01 -8.50714862e-01
-8.54752004e-01 3.96808952e-01 -2.53048390e-01 -1.05049479e+00
1.96552813e-01 2.62472600e-01 -8.90750438e-02 -2.16699943e-01
3.76575023e-01 4.41815615e-01 3.30564797e-01 -3.96523058e-01
4.77012217e-01 5.05838804e-02 1.03683376e+00 -7.74975598e-01
9.96146679e-01 5.72060883e-01 -7.62296617e-02 -3.57812971e-01
-9.47511196e-01 -4.37803566e-01 -5.93565702e-01 -2.63243049e-01
9.72865999e-01 -1.44287252e+00 -9.92283523e-01 7.12889969e-01
-9.10541475e-01 -4.42823261e-01 -1.94615170e-01 9.01321590e-01
-5.07342637e-01 6.13760829e-01 -1.05515897e+00 -1.76231697e-01
-5.32068193e-01 -9.53221500e-01 1.27485085e+00 3.45405012e-01
-7.33918995e-02 -8.11264277e-01 5.69646526e-03 3.11160415e-01
2.78702416e-02 4.33634222e-01 2.28337198e-01 -1.83654111e-02
-5.02069294e-01 -3.11403096e-01 -2.14164644e-01 4.28393066e-01
2.85957102e-02 -5.36307037e-01 -3.74043137e-01 -7.38104582e-01
-2.76549727e-01 -5.65966189e-01 7.66636014e-01 3.29451710e-01
1.14837873e+00 -3.10810715e-01 -3.13294888e-01 1.02336931e+00
1.12910676e+00 -3.06753933e-01 5.81840217e-01 1.34099811e-01
1.13413393e+00 7.97121301e-02 8.31957757e-01 5.52424490e-01
3.29096884e-01 9.06045973e-01 2.04319000e-01 -2.37950251e-01
-2.76171952e-01 -4.76009399e-01 5.22860289e-01 8.71565282e-01
-3.43259543e-01 2.20609665e-01 -4.36628342e-01 2.52563089e-01
-2.03300881e+00 -7.60795236e-01 5.06400824e-01 2.11677980e+00
1.01056039e+00 1.34206772e-01 4.18933868e-01 -2.60215342e-01
6.34926498e-01 4.34365869e-01 -7.14061201e-01 4.08020467e-01
4.04557496e-01 3.07654113e-01 4.47538882e-01 2.45710626e-01
-1.29474449e+00 1.00747156e+00 5.28618860e+00 1.11899054e+00
-1.22890484e+00 1.76474705e-01 2.48551548e-01 -4.71941352e-01
3.09381276e-01 -2.26875618e-01 -7.35084534e-01 5.08208454e-01
5.62663794e-01 9.19238478e-02 2.82479137e-01 9.58680212e-01
2.24392191e-01 7.79580474e-02 -9.53033388e-01 1.23591685e+00
7.71296769e-02 -1.16556287e+00 1.70594659e-02 -1.36442989e-01
4.73909110e-01 -1.06789038e-01 -8.51433575e-02 4.25066382e-01
8.17842185e-02 -6.98828042e-01 6.39304757e-01 4.66623634e-01
7.93146551e-01 -1.02086604e+00 5.39546311e-01 1.89360544e-01
-1.68799198e+00 6.18222952e-02 -4.45536494e-01 -1.98703274e-01
3.80167156e-01 3.44048768e-01 -3.94372225e-01 5.72137058e-01
8.06976199e-01 7.73292780e-01 -2.54848957e-01 8.41129839e-01
-3.40801686e-01 3.84238839e-01 -4.93389517e-01 2.77096093e-01
3.54616523e-01 -3.00178379e-02 3.31347227e-01 1.02728498e+00
2.25533947e-01 1.15641139e-01 8.07333291e-01 3.85793716e-01
-8.63598809e-02 -1.16640374e-01 -2.09401056e-01 9.78372321e-02
3.15726250e-01 1.16812670e+00 -5.70199966e-01 -4.32208568e-01
-4.31325644e-01 1.67712259e+00 4.85208094e-01 1.84490696e-01
-1.31031227e+00 -4.25650358e-01 8.03149760e-01 3.30803469e-02
4.91080433e-01 -4.97866094e-01 5.70890844e-01 -1.46531439e+00
3.32634211e-01 -9.14072156e-01 3.13289583e-01 -5.27482033e-01
-1.01442850e+00 2.79591471e-01 2.44756807e-02 -1.31688523e+00
-3.50725502e-01 -4.51113850e-01 -3.53277773e-01 7.13822782e-01
-1.15195048e+00 -1.45688951e+00 -5.56253910e-01 9.63065743e-01
4.82010841e-01 2.51444221e-01 5.78705132e-01 5.41448772e-01
-7.18919814e-01 8.81573141e-01 -2.85095900e-01 7.38139987e-01
9.30179000e-01 -1.03730869e+00 7.00593531e-01 8.93502653e-01
-8.42267871e-02 6.56193078e-01 3.20748448e-01 -7.68538237e-01
-1.53868508e+00 -1.28643668e+00 2.37330258e-01 -4.81261201e-02
5.60451448e-01 -3.12195957e-01 -6.19775772e-01 5.31027138e-01
-3.20986748e-01 7.12895989e-01 4.28758323e-01 -2.28761479e-01
-4.30188775e-01 -2.92059660e-01 -7.86257744e-01 7.83043921e-01
1.34392619e+00 -5.21550238e-01 -5.49352944e-01 3.34932685e-01
9.53375816e-01 -8.73058021e-01 -1.06385279e+00 4.20066386e-01
8.87235641e-01 -6.85139298e-01 1.29412115e+00 -4.85630095e-01
3.90214443e-01 -5.64139724e-01 1.80300981e-01 -8.60161185e-01
-3.89577061e-01 -8.26398134e-01 -6.90879941e-01 7.40691721e-01
-2.24014744e-01 -3.36602509e-01 1.16660440e+00 4.28927243e-01
-1.27923310e-01 -1.08099604e+00 -9.86602008e-01 -9.25395608e-01
-2.60575563e-01 -2.94738412e-01 3.82804871e-01 6.11430764e-01
-3.51459742e-01 -4.64775674e-02 -9.36701834e-01 3.30769479e-01
6.59266233e-01 8.58993083e-02 1.06102765e+00 -5.37407935e-01
-8.29902828e-01 1.03653416e-01 -8.28346491e-01 -1.81213510e+00
9.54869390e-02 -5.05468845e-01 1.53227866e-01 -1.11225343e+00
1.29074290e-01 -9.93485749e-02 -2.20482334e-01 6.76102400e-01
-4.96204257e-01 6.20835006e-01 3.19928586e-01 1.77147657e-01
-7.94052541e-01 5.94272256e-01 1.57051611e+00 -1.69728976e-02
-1.28497735e-01 -8.93198401e-02 -1.17651947e-01 9.35009360e-01
3.43772948e-01 -3.36247563e-01 -6.09806776e-01 -5.19740760e-01
-2.44266629e-01 3.13774347e-01 7.07004189e-01 -1.48897290e+00
2.87836164e-01 1.17933964e-02 7.72488415e-01 -4.80972409e-01
6.50121689e-01 -6.28508031e-01 1.43721223e-01 8.57269406e-01
-2.14906558e-02 5.26663922e-02 1.39773443e-01 8.39479446e-01
-1.73147008e-01 3.17714751e-01 5.96233070e-01 -1.45707786e-01
-1.04433990e+00 9.08398449e-01 1.57376528e-01 1.27672389e-01
1.13646114e+00 -2.46444672e-01 2.54758954e-01 -1.94494456e-01
-9.09607828e-01 3.11487049e-01 5.07585645e-01 4.34933156e-01
6.58699393e-01 -1.51845622e+00 -4.00785297e-01 2.28181764e-01
-1.87216654e-01 2.38367751e-01 6.96899652e-01 1.05712605e+00
-8.09586227e-01 1.35429665e-01 -1.99236497e-01 -7.03566015e-01
-1.25063038e+00 4.39618915e-01 3.47094893e-01 -4.59305495e-01
-1.02735424e+00 1.15519083e+00 4.46209699e-01 -1.10076219e-01
3.56056243e-01 -3.58215958e-01 1.91390648e-01 -2.23804876e-01
6.48358047e-01 2.68420935e-01 -3.14444125e-01 -4.94161338e-01
-3.72926950e-01 6.26805723e-01 -1.75560161e-01 4.77103479e-02
1.04972839e+00 -9.00771171e-02 1.64898276e-01 7.13353902e-02
1.41873956e+00 -1.71930447e-01 -1.94145858e+00 -3.68860334e-01
-6.55570507e-01 -5.98166287e-01 -2.69253571e-02 -4.48729247e-01
-1.42946255e+00 5.92393458e-01 6.40175521e-01 -7.07170486e-01
1.31700563e+00 -1.80639356e-01 1.44147384e+00 5.01113236e-01
3.71152222e-01 -1.11461461e+00 4.91945535e-01 2.71717846e-01
5.96514881e-01 -9.72733915e-01 9.09221172e-02 -5.17122209e-01
-4.23782736e-01 1.13220692e+00 9.56588745e-01 -4.62364584e-01
5.41056514e-01 2.36741439e-01 -2.09917743e-02 5.30940574e-03
-2.84612060e-01 -3.63716222e-02 4.09095228e-01 4.32335883e-01
2.72484511e-01 -1.22129269e-01 -2.26613835e-01 5.98663688e-01
1.09753102e-01 3.48055750e-01 -2.01533824e-01 1.02812374e+00
-2.00436294e-01 -9.39879775e-01 -2.12659359e-01 2.21496031e-01
-3.89108509e-01 -7.95420036e-02 2.59482145e-01 9.12870526e-01
2.56295830e-01 2.11965188e-01 -7.55084157e-02 -4.89200354e-01
2.83647627e-01 -2.48519748e-01 7.52222419e-01 -4.44292903e-01
-4.10211921e-01 2.92412043e-01 -6.47387728e-02 -1.14855921e+00
-5.23587704e-01 -3.92647088e-01 -1.33499503e+00 -2.86505342e-01
-3.31920177e-01 3.59612219e-02 -1.31737189e-02 9.54824805e-01
2.95157313e-01 6.53873861e-01 2.25467220e-01 -1.22840500e+00
-7.55530298e-01 -8.24823141e-01 -4.40240949e-01 4.70656157e-01
2.23306283e-01 -8.67630005e-01 1.25762016e-01 1.50102377e-01]
|
[7.17518424987793, -0.6010617613792419]
|
c0cad2bf-4c5a-43b8-b0da-1a6b60b650d3
|
reconet-real-time-coherent-video-style
|
1807.01197
| null |
http://arxiv.org/abs/1807.01197v2
|
http://arxiv.org/pdf/1807.01197v2.pdf
|
ReCoNet: Real-time Coherent Video Style Transfer Network
|
Image style transfer models based on convolutional neural networks usually
suffer from high temporal inconsistency when applied to videos. Some video
style transfer models have been proposed to improve temporal consistency, yet
they fail to guarantee fast processing speed, nice perceptual style quality and
high temporal consistency at the same time. In this paper, we propose a novel
real-time video style transfer model, ReCoNet, which can generate temporally
coherent style transfer videos while maintaining favorable perceptual styles. A
novel luminance warping constraint is added to the temporal loss at the output
level to capture luminance changes between consecutive frames and increase
stylization stability under illumination effects. We also propose a novel
feature-map-level temporal loss to further enhance temporal consistency on
traceable objects. Experimental results indicate that our model exhibits
outstanding performance both qualitatively and quantitatively.
|
['Derun Gu', 'Yizhou Yu', 'Fangjun Zhang', 'Chang Gao']
|
2018-07-03
| null | null | null | null |
['video-style-transfer']
|
['computer-vision']
|
[ 3.08923304e-01 -6.88444138e-01 4.34572771e-02 -3.37132424e-01
-1.10471763e-01 -5.19537449e-01 5.48755765e-01 -4.85901028e-01
-1.25359327e-01 6.98320329e-01 -1.20978185e-03 7.76337162e-02
-1.22748734e-02 -7.27012217e-01 -8.46729636e-01 -4.40282136e-01
3.11319202e-01 -4.94378984e-01 5.66968679e-01 -1.52573854e-01
1.56520247e-01 2.77951330e-01 -1.16597939e+00 3.07254732e-01
1.09362149e+00 1.09493983e+00 2.10043654e-01 4.81821120e-01
-7.70621002e-02 8.35275412e-01 -4.07776952e-01 -4.20502961e-01
4.55773175e-01 -7.22800553e-01 -3.74825865e-01 3.59186113e-01
7.30152369e-01 -6.68107808e-01 -7.76434302e-01 1.17659688e+00
4.54538375e-01 1.26316443e-01 4.21835214e-01 -1.33459318e+00
-1.12333560e+00 1.48524225e-01 -8.43610644e-01 1.77701917e-02
3.77539307e-01 4.35792804e-01 5.67208469e-01 -7.75908768e-01
6.81811690e-01 1.36855900e+00 5.58340549e-01 6.42095089e-01
-1.33037698e+00 -8.31321120e-01 2.48145267e-01 3.91266704e-01
-1.05734885e+00 -2.08602950e-01 1.15901041e+00 1.98935252e-02
3.86437207e-01 1.37000769e-01 8.01939249e-01 1.20662308e+00
5.59786618e-01 7.43862033e-01 1.20042098e+00 -1.46566644e-01
6.15803935e-02 -2.76915878e-01 -4.67923671e-01 5.86371183e-01
-9.71718654e-02 2.74456173e-01 -7.42105722e-01 3.64505708e-01
1.47307014e+00 5.40701784e-02 -4.09363717e-01 -2.99635202e-01
-1.26593399e+00 3.89448196e-01 5.43558359e-01 1.49665982e-01
-2.08963230e-01 2.81060576e-01 6.30829215e-01 6.00757360e-01
4.80288625e-01 7.79927969e-02 -2.11162910e-01 -1.21544696e-01
-8.23123813e-01 1.38290852e-01 2.00366274e-01 1.25780141e+00
5.28097332e-01 5.90037405e-01 -5.53298473e-01 8.42251956e-01
-4.34945375e-02 4.82116520e-01 3.51698816e-01 -1.27991974e+00
4.60891277e-01 2.87396878e-01 1.16971761e-01 -1.29274035e+00
8.27001408e-02 -1.70905650e-01 -1.09927297e+00 3.52680534e-01
1.28178224e-01 -5.72149195e-02 -7.06535578e-01 1.69890738e+00
7.88232312e-02 4.62089866e-01 -3.05590332e-01 1.01926780e+00
7.02317595e-01 1.01501656e+00 1.84298292e-01 -4.34038728e-01
1.04755020e+00 -8.81541967e-01 -1.10251403e+00 1.81036457e-01
-3.21298279e-02 -1.05240047e+00 1.35135031e+00 4.42907274e-01
-1.56420648e+00 -1.02118695e+00 -1.15703106e+00 -2.30690196e-01
2.38953754e-01 5.09595759e-02 3.60318631e-01 3.15325737e-01
-1.04328859e+00 7.34267890e-01 -5.92033148e-01 -1.62975445e-01
3.14961553e-01 2.03476623e-01 -1.45958811e-01 6.74175695e-02
-1.07053304e+00 4.73518461e-01 2.36461639e-01 5.33865541e-02
-4.98291701e-01 -7.38485098e-01 -8.19919050e-01 8.15195683e-03
1.54350787e-01 -8.20748270e-01 1.16919482e+00 -1.56421685e+00
-2.03976798e+00 4.69110221e-01 -1.47491425e-01 -1.17076814e-01
8.96641433e-01 -4.33547884e-01 -6.54103220e-01 2.24389091e-01
-8.68638903e-02 1.11426747e+00 1.08675551e+00 -1.27207279e+00
-6.24488056e-01 1.56534240e-01 -8.33758563e-02 1.63925722e-01
-6.84269965e-01 6.02441132e-02 -6.30955875e-01 -1.44850492e+00
-7.32367039e-02 -7.95447826e-01 1.76123008e-02 6.25264525e-01
4.45081107e-02 -5.70060872e-03 1.44831300e+00 -6.16679788e-01
1.18413186e+00 -2.18797708e+00 2.30782524e-01 -2.35464752e-01
2.70207711e-02 5.40061295e-01 -4.75703180e-01 5.43053076e-02
-2.31373422e-02 -2.16847435e-01 -1.44214287e-01 -1.01085551e-01
-3.76253426e-01 2.24841610e-01 -5.06138384e-01 1.63980916e-01
5.53366244e-01 1.01840007e+00 -9.90122616e-01 -4.93164748e-01
4.54191178e-01 5.45556307e-01 -7.20963299e-01 2.78634965e-01
-2.92293489e-01 6.23000979e-01 -4.27026242e-01 3.18992555e-01
9.16919589e-01 2.89970525e-02 -3.30360904e-02 -4.82567668e-01
-2.13092100e-02 -1.21982381e-01 -1.05764329e+00 1.81595385e+00
-4.46656495e-01 8.33222091e-01 -2.71855271e-03 -4.89216626e-01
1.10393727e+00 2.73325384e-01 5.56479275e-01 -9.94381785e-01
2.77356654e-01 7.53421932e-02 -1.65192977e-01 -3.17071348e-01
8.46729338e-01 -1.79222196e-01 1.71786532e-01 1.16387442e-01
-1.02521628e-01 -2.50606477e-01 -4.79251146e-03 1.34390816e-02
5.74653447e-01 6.80780649e-01 -1.18399426e-01 -4.27753121e-01
6.53929532e-01 -4.71712440e-01 9.79233563e-01 1.79771468e-01
-3.10091853e-01 9.12653267e-01 3.02400976e-01 -5.85211813e-01
-1.46771777e+00 -1.20299459e+00 1.97502196e-01 7.07625628e-01
6.69455826e-01 -3.04655641e-01 -6.48558617e-01 -4.79343891e-01
-2.91362762e-01 3.76686007e-01 -4.10057902e-01 -4.28129882e-01
-9.06979263e-01 -7.14478418e-02 4.60565895e-01 8.07886600e-01
1.13969731e+00 -1.04384172e+00 -4.07396495e-01 4.43790525e-01
-2.07005784e-01 -1.22709227e+00 -1.19672704e+00 -4.68140125e-01
-1.00293517e+00 -5.52045465e-01 -1.10782385e+00 -1.20216393e+00
5.14275074e-01 3.67229432e-01 8.54029477e-01 -9.50389430e-02
-7.53377974e-02 1.41246632e-01 -3.17719281e-01 -1.05572037e-01
-3.44033867e-01 -3.51910472e-01 2.41666332e-01 2.30020717e-01
-8.43005776e-02 -8.36856186e-01 -8.24785054e-01 5.77101707e-01
-1.35638511e+00 3.80446643e-01 2.66255438e-01 1.12748468e+00
4.55000430e-01 -1.68511644e-02 6.71158493e-01 -2.80209452e-01
5.34524560e-01 1.57224298e-01 -4.95403975e-01 1.94755226e-01
-3.64269167e-01 -7.61817172e-02 9.21141803e-01 -8.83468270e-01
-1.48093808e+00 -2.15660185e-01 1.22811526e-01 -1.02738738e+00
4.40618582e-02 -1.01472542e-01 -1.71153441e-01 -2.55758107e-01
3.96669567e-01 4.31575447e-01 1.62790209e-01 -3.24991435e-01
3.78892720e-01 1.76017255e-01 7.75770187e-01 -6.93221927e-01
9.49215293e-01 4.28095907e-01 7.39830285e-02 -6.38637900e-01
-4.90660608e-01 1.66196793e-01 -5.29905438e-01 -3.81062090e-01
7.50952005e-01 -9.91944492e-01 -6.73904538e-01 8.42498660e-01
-1.20295048e+00 -4.45642114e-01 -1.70822218e-01 3.83884043e-01
-9.06235337e-01 7.11164773e-01 -1.11015654e+00 -2.08205163e-01
-4.65599120e-01 -9.45795178e-01 8.29662859e-01 3.68791252e-01
-5.61362505e-02 -8.00500751e-01 -3.34273040e-01 -1.46174356e-01
5.88925779e-01 2.84486711e-01 6.52113974e-01 6.64193332e-01
-6.71051621e-01 4.27962959e-01 -6.32234991e-01 4.47544038e-01
5.12036264e-01 3.04092973e-01 -5.14912307e-01 -3.91009539e-01
-6.69363290e-02 -4.28315885e-02 8.11968148e-01 4.83926892e-01
1.52688289e+00 -3.65731984e-01 3.17014977e-02 8.40339184e-01
1.26856816e+00 5.73346794e-01 9.28424299e-01 1.73096314e-01
9.34578598e-01 4.70294505e-01 6.84475482e-01 2.91016638e-01
-2.27844894e-01 8.82126033e-01 -1.82896238e-02 -2.79540777e-01
-5.45148611e-01 -3.41583878e-01 6.52544558e-01 9.96901572e-01
-9.45967510e-02 -4.39555198e-01 -1.79083720e-01 4.50639993e-01
-1.87240565e+00 -9.74795163e-01 -8.50722119e-02 1.89454520e+00
9.37254131e-01 3.24597120e-01 1.96750060e-01 7.21378773e-02
7.82123029e-01 1.76082879e-01 -5.41139543e-01 -5.10164678e-01
-3.06521147e-01 -1.34016927e-02 2.34728992e-01 2.53855437e-01
-9.28091884e-01 9.97947395e-01 6.40497541e+00 1.05708063e+00
-1.37596989e+00 -4.43098806e-02 7.67906547e-01 -1.86819583e-01
-4.21269327e-01 -2.67865390e-01 -2.34258488e-01 6.84502006e-01
2.27931589e-01 -2.19095513e-01 2.42237985e-01 5.37507415e-01
5.08305728e-01 2.90322334e-01 -9.11131680e-01 1.17063582e+00
-2.89588887e-02 -1.31624591e+00 5.65700114e-01 -1.56713769e-01
1.03789127e+00 -7.17016518e-01 5.18432319e-01 -1.28230616e-01
4.19647135e-02 -6.51356876e-01 1.04043639e+00 5.05214572e-01
1.28524554e+00 -1.01296568e+00 2.50022858e-01 -2.29008734e-01
-1.38122809e+00 5.64542189e-02 -2.95017689e-01 -1.22797251e-01
4.06044930e-01 2.78257340e-01 -3.30500007e-02 6.24849379e-01
7.02018559e-01 1.33510411e+00 -4.68149155e-01 9.98931527e-01
-9.63535607e-02 4.16485846e-01 7.92384967e-02 7.85560459e-02
4.14072633e-01 -4.64718401e-01 7.19236016e-01 1.01417851e+00
5.89473069e-01 2.55654603e-01 4.33485880e-02 9.40714955e-01
-3.99690419e-02 -3.27835679e-02 -7.93165863e-01 2.93144554e-01
2.46358708e-01 9.31461573e-01 -6.21740222e-01 -3.20364267e-01
-3.76102209e-01 1.43892372e+00 -1.10945761e-01 4.47274894e-01
-1.10252154e+00 -3.32360178e-01 7.78357327e-01 4.68517728e-02
2.56761789e-01 -5.50442100e-01 -5.66213131e-01 -1.22194111e+00
2.42882505e-01 -6.89615965e-01 7.14421794e-02 -1.06439614e+00
-1.41584527e+00 4.70709652e-01 -2.02138171e-01 -1.92071092e+00
1.38425052e-01 -5.85470378e-01 -7.42406845e-01 5.22446096e-01
-1.32966220e+00 -1.13942885e+00 -4.59866017e-01 8.24057639e-01
1.02607036e+00 -1.27234474e-01 1.68526128e-01 4.05021399e-01
-3.84750038e-01 8.46523523e-01 5.17072268e-02 -1.08669661e-02
1.12915123e+00 -8.94168615e-01 2.92648613e-01 9.56993103e-01
-3.00983101e-01 4.65687484e-01 6.18997097e-01 -7.37369835e-01
-1.27439427e+00 -1.39952767e+00 2.74648607e-01 -1.90074891e-02
3.33407104e-01 -5.69442138e-02 -1.13421607e+00 4.35698897e-01
6.24643564e-01 -1.46116763e-01 7.52864704e-02 -5.68601847e-01
-3.56469780e-01 -4.57023144e-01 -9.98461485e-01 1.00429094e+00
1.24725437e+00 -4.23170269e-01 -3.58798802e-01 -1.96631268e-01
9.72247183e-01 -3.34230304e-01 -8.68430614e-01 4.88717437e-01
6.33569837e-01 -9.85031426e-01 8.62041414e-01 -1.86694130e-01
8.93746078e-01 -5.12557089e-01 1.10746384e-01 -1.21440637e+00
-6.95026278e-01 -1.01171505e+00 6.76547214e-02 1.51423776e+00
-1.70746461e-01 -4.09842759e-01 4.66203719e-01 2.95582175e-01
-2.84167789e-02 -4.57582504e-01 -6.81187391e-01 -1.15848207e+00
-2.26782672e-02 -9.66580883e-02 5.10659158e-01 8.78317893e-01
-1.94460571e-01 2.73844093e-01 -9.12276268e-01 -2.64075994e-01
6.68909609e-01 1.20715378e-02 6.09106302e-01 -7.47477114e-01
-9.25704762e-02 -5.83505571e-01 -3.10220271e-01 -1.21676791e+00
-4.84470055e-02 -2.27610052e-01 -1.43379066e-02 -1.05436838e+00
1.65260270e-01 -2.65997052e-01 -3.19580793e-01 1.77931115e-01
-2.18298316e-01 5.80370426e-01 3.41624856e-01 9.99845564e-02
-4.08316404e-01 1.01301968e+00 1.89998960e+00 -1.54708445e-01
-2.43883640e-01 -3.84187162e-01 -1.97474822e-01 6.78629160e-01
7.92010665e-01 2.51189303e-02 -5.90309918e-01 -7.82884061e-01
-1.07172579e-01 1.76296189e-01 3.58616352e-01 -1.14106750e+00
-6.22893311e-03 -4.13378268e-01 7.64920950e-01 -5.50171852e-01
3.29660982e-01 -7.67199397e-01 3.61100703e-01 4.96999145e-01
-3.97486091e-01 3.97120088e-01 3.94618630e-01 7.21491098e-01
-3.84103805e-01 3.80232990e-01 1.19046092e+00 2.21417412e-01
-9.27280962e-01 5.69190204e-01 -2.50218779e-01 -3.80197644e-01
1.15094304e+00 -3.45791668e-01 -1.13496661e-01 -5.42282999e-01
-3.58882606e-01 5.25197275e-02 8.85245800e-01 9.10945177e-01
8.18816245e-01 -1.96853864e+00 -7.06292689e-01 3.75218034e-01
-9.62245166e-02 -3.83076727e-01 4.86020058e-01 4.66570228e-01
-6.76785648e-01 1.71314001e-01 -9.31390524e-01 -6.45260513e-01
-1.23151147e+00 7.88525939e-01 2.38725111e-01 1.94851592e-01
-8.11947525e-01 6.02340579e-01 5.70086420e-01 1.10836126e-01
1.04369298e-01 -3.63091081e-01 1.03972830e-01 -3.36453199e-01
5.69130540e-01 3.25589329e-01 -3.44283611e-01 -4.73769069e-01
3.89325246e-02 9.86728787e-01 6.51387498e-04 -1.00746937e-01
1.10902834e+00 -4.53100175e-01 -2.16059405e-02 2.29589999e-01
1.15113127e+00 -2.60839045e-01 -1.94916737e+00 -4.55146104e-01
-4.22492206e-01 -1.01197529e+00 -9.51423272e-02 -4.43616509e-01
-1.36938810e+00 8.33524942e-01 6.85400009e-01 -9.12646800e-02
1.60789835e+00 -6.70269370e-01 1.41745019e+00 -2.06330836e-01
4.07077849e-01 -1.10974967e+00 7.42048919e-01 3.97691846e-01
1.02930093e+00 -9.29105759e-01 -1.84105024e-01 -6.36992276e-01
-6.18273497e-01 1.18553293e+00 9.84456897e-01 -3.96029323e-01
3.21618825e-01 1.78659767e-01 -5.06388023e-02 3.19315761e-01
-6.77015364e-01 3.37390780e-01 4.36314315e-01 6.34314120e-01
3.96872312e-01 -1.54393047e-01 -5.41733921e-01 1.93615502e-04
1.80528104e-01 2.74427474e-01 4.65944231e-01 6.69605315e-01
-2.15083733e-01 -1.19775081e+00 -2.52652705e-01 6.16991259e-02
-3.25635076e-01 1.72017589e-01 -5.10814525e-02 3.89498681e-01
3.30763906e-02 7.67169416e-01 1.60869643e-01 -4.92679447e-01
3.18472892e-01 -3.48123282e-01 7.36636221e-01 1.88989881e-02
-4.14071769e-01 4.43999857e-01 -2.06100941e-01 -6.42493308e-01
-5.34528017e-01 -4.85089362e-01 -9.15297389e-01 -5.75812519e-01
4.49198484e-02 -3.48619133e-01 4.97894585e-02 5.35120547e-01
4.58822936e-01 7.37263143e-01 8.63915563e-01 -9.67855453e-01
-1.51867181e-01 -7.12462544e-01 -7.02726722e-01 7.43644059e-01
2.25402877e-01 -5.68108737e-01 1.97131351e-01 5.74764967e-01]
|
[11.251672744750977, -0.8638311624526978]
|
f48a1609-0001-41d5-9d9f-773fc974a689
|
metaaudio-a-few-shot-audio-classification
|
2204.02121
| null |
https://arxiv.org/abs/2204.02121v2
|
https://arxiv.org/pdf/2204.02121v2.pdf
|
MetaAudio: A Few-Shot Audio Classification Benchmark
|
Currently available benchmarks for few-shot learning (machine learning with few training examples) are limited in the domains they cover, primarily focusing on image classification. This work aims to alleviate this reliance on image-based benchmarks by offering the first comprehensive, public and fully reproducible audio based alternative, covering a variety of sound domains and experimental settings. We compare the few-shot classification performance of a variety of techniques on seven audio datasets (spanning environmental sounds to human-speech). Extending this, we carry out in-depth analyses of joint training (where all datasets are used during training) and cross-dataset adaptation protocols, establishing the possibility of a generalised audio few-shot classification algorithm. Our experimentation shows gradient-based meta-learning methods such as MAML and Meta-Curvature consistently outperform both metric and baseline methods. We also demonstrate that the joint training routine helps overall generalisation for the environmental sound databases included, as well as being a somewhat-effective method of tackling the cross-dataset/domain setting.
|
['Mehrdad Yaghoobi', 'Timothy Hospedales', 'Sam Budgett', 'Calum Heggan']
|
2022-04-05
| null | null | null | null |
['few-shot-audio-classification']
|
['audio']
|
[ 5.03247261e-01 -4.77523357e-01 -6.41835332e-02 -2.63326973e-01
-1.48937237e+00 -2.84844190e-01 7.96966255e-01 5.26429750e-02
-5.23555756e-01 4.35743302e-01 3.45770657e-01 2.61408359e-01
-3.40366483e-01 -3.88533115e-01 -4.53273445e-01 -5.88097155e-01
-3.31547230e-01 2.73442000e-01 5.11121094e-01 -2.50348985e-01
2.32344866e-01 -5.89596340e-03 -2.23216486e+00 5.82002759e-01
1.60276055e-01 1.03685498e+00 1.06052183e-01 1.18184817e+00
-1.32791847e-02 7.63541877e-01 -5.49144208e-01 -2.62754291e-01
1.71279922e-01 -5.24243474e-01 -8.08859646e-01 5.40440902e-02
7.15535402e-01 6.26309635e-03 -4.93514948e-02 7.23391235e-01
1.30143058e+00 5.41429400e-01 7.47950196e-01 -1.33393705e+00
-2.16649354e-01 5.57180405e-01 -2.39324719e-01 4.88451242e-01
6.14035130e-01 4.84914482e-01 1.05811405e+00 -9.57875609e-01
6.32206380e-01 1.12736845e+00 9.89201307e-01 6.67491853e-01
-1.18254328e+00 -5.88603318e-01 -3.48753422e-01 6.33686841e-01
-1.23532438e+00 -9.68110919e-01 6.78760350e-01 -6.23885334e-01
1.19722092e+00 2.55978763e-01 3.63804281e-01 1.42086983e+00
-1.72200605e-01 6.14096284e-01 1.18743110e+00 -8.77250314e-01
6.29126787e-01 2.39455387e-01 -4.15552109e-02 2.34931171e-01
-3.99395317e-01 1.02256149e-01 -9.50761020e-01 -2.93713838e-01
8.15683305e-02 -4.42385256e-01 -1.38820097e-01 -4.72974718e-01
-1.16787684e+00 6.59270942e-01 -1.60521895e-01 4.96778429e-01
-1.88817456e-01 2.98082322e-01 1.09467912e+00 6.49708509e-01
6.98500752e-01 4.56234872e-01 -5.24969995e-01 -7.32781887e-01
-1.23285425e+00 2.67087340e-01 9.25373554e-01 8.92379284e-01
6.97912097e-01 2.98544824e-01 -1.78825736e-01 1.46608198e+00
-1.16163805e-01 4.45446931e-02 8.56459677e-01 -1.09070015e+00
3.25051576e-01 -2.01915175e-01 -2.49388650e-01 -5.29653251e-01
-2.71994919e-01 -1.37290820e-01 -4.21566188e-01 2.08308339e-01
2.06570998e-01 -6.52341992e-02 -8.14959526e-01 1.55465138e+00
3.05048853e-01 8.01268637e-01 1.21222563e-01 5.36724031e-01
1.07991433e+00 4.03657854e-01 2.59924173e-01 -3.24490994e-01
1.18119359e+00 -9.80878770e-01 -3.88676256e-01 -7.93660432e-03
7.22411215e-01 -8.72155428e-01 1.36177838e+00 5.18821001e-01
-1.03631032e+00 -8.73134494e-01 -1.13533521e+00 2.27674499e-01
-6.28746986e-01 -4.30546492e-01 3.84387165e-01 8.05888474e-01
-9.75512981e-01 8.69141817e-01 -5.88468015e-01 -7.45582223e-01
4.22440320e-01 1.39092490e-01 -3.74943316e-01 -1.31397977e-01
-1.13573241e+00 7.06054330e-01 3.25865000e-01 -6.23690784e-01
-1.22856379e+00 -1.07900047e+00 -8.56126308e-01 -1.89030185e-01
3.04072946e-01 -4.83146220e-01 1.56675160e+00 -8.70502412e-01
-1.54112530e+00 9.21883762e-01 2.16270566e-01 -8.03106427e-01
5.55186749e-01 -1.68230176e-01 -5.37319064e-01 2.09416553e-01
1.93783008e-02 8.87422919e-01 1.05653155e+00 -1.00744176e+00
-5.38900495e-01 -6.71776980e-02 -1.89622432e-01 1.83162645e-01
-4.80327070e-01 2.85349488e-01 -2.34343633e-01 -5.78362644e-01
-5.71187139e-01 -7.76557386e-01 -9.12077129e-02 -1.76423028e-01
1.06208190e-01 -4.27207761e-02 7.98642159e-01 -2.94978380e-01
1.17662787e+00 -2.33453441e+00 4.68260311e-02 -3.53367329e-01
-4.38956767e-01 2.25998297e-01 -4.31129456e-01 6.89392865e-01
-2.29560331e-01 -1.52889058e-01 -3.45766813e-01 -6.54128730e-01
1.20823495e-01 2.99120277e-01 -3.82280976e-01 4.24703836e-01
3.94147113e-02 5.29447198e-01 -9.84449804e-01 -7.94088006e-01
4.06570882e-01 4.05375957e-01 -5.81982017e-01 4.93644103e-02
-9.12409425e-02 2.02251703e-01 2.01764613e-01 7.92754173e-01
2.52421319e-01 2.96774387e-01 -4.11116809e-01 1.23632737e-01
4.55470383e-03 3.42097990e-02 -1.30606997e+00 2.26452255e+00
-8.38618815e-01 7.94628918e-01 -8.63346756e-02 -9.53489542e-01
6.60939336e-01 6.63406670e-01 7.38299370e-01 -5.55813611e-01
-7.23800361e-02 2.52207130e-01 -6.16526306e-02 -8.00057113e-01
2.17572734e-01 -2.62381822e-01 -2.50762235e-02 3.51288110e-01
9.28750217e-01 -4.18483824e-01 2.87435621e-01 -4.20157090e-02
1.18855202e+00 2.63534635e-01 4.79553461e-01 -1.53616071e-01
4.19634491e-01 -2.87998263e-02 1.33169562e-01 8.99422526e-01
-5.97212076e-01 9.54046428e-01 -4.57008556e-02 -2.38301784e-01
-1.13224673e+00 -8.44238281e-01 -4.68083680e-01 1.79697049e+00
-3.61473203e-01 -6.30224347e-01 -7.69944608e-01 -3.00467879e-01
-1.29673421e-01 7.44207263e-01 -7.10019767e-01 -8.29858184e-02
-1.26878262e-01 -8.76527667e-01 7.81302214e-01 4.28823173e-01
3.11549753e-01 -9.84743178e-01 -9.10970390e-01 4.86858785e-01
3.16151008e-02 -1.03970718e+00 -4.78203706e-02 5.95728517e-01
-7.84346163e-01 -9.38045979e-01 -7.66060829e-01 -4.92983043e-01
-3.77565622e-01 1.25894964e-01 1.17877746e+00 -3.32881808e-01
-8.05435717e-01 1.07943428e+00 -6.94673419e-01 -7.76026964e-01
-5.44631362e-01 1.99815333e-02 1.85104370e-01 -1.26342267e-01
5.02935112e-01 -9.11923766e-01 -4.52417135e-01 2.55558610e-01
-6.96483374e-01 -4.56266105e-01 1.03755221e-01 8.48623157e-01
4.48854923e-01 -2.52642244e-01 7.37194717e-01 -6.87309802e-01
6.98019803e-01 -6.32932186e-01 5.32213449e-02 4.64373194e-02
-4.57989067e-01 -3.53108555e-01 2.97911286e-01 -5.40769994e-01
-8.40805650e-01 1.21739343e-01 -3.72235358e-01 -5.82173467e-01
-7.14131117e-01 5.68953305e-02 1.18096024e-01 -1.52409792e-01
1.30298579e+00 2.40582764e-01 -9.72043201e-02 -6.19314492e-01
4.62495267e-01 9.19159770e-01 6.56522691e-01 -4.85829800e-01
4.55901802e-01 4.87130225e-01 -1.32329285e-01 -1.27222240e+00
-7.29544997e-01 -1.01090121e+00 -8.24483871e-01 -4.69443381e-01
7.54739821e-01 -1.04539490e+00 1.18952952e-01 2.95988470e-01
-6.82721734e-01 -4.43113357e-01 -7.53074706e-01 5.31104386e-01
-1.12718689e+00 3.41994852e-01 -3.27467978e-01 -9.46648538e-01
-2.31137455e-01 -9.94090319e-01 1.22356999e+00 -2.86383599e-01
-4.70852017e-01 -1.05308294e+00 7.65060425e-01 1.41607597e-01
3.46909553e-01 2.92292953e-01 6.50769532e-01 -9.86760974e-01
1.61010295e-01 -1.18043147e-01 2.72830725e-01 4.01652575e-01
2.81517347e-03 8.24690312e-02 -1.81251931e+00 -2.11722419e-01
-1.31349610e-02 -7.80606866e-01 1.16357350e+00 3.29260051e-01
9.44199264e-01 1.35024756e-01 1.00496067e-02 5.05665183e-01
1.48986578e+00 3.32710333e-02 5.06132841e-01 6.43252611e-01
2.49928534e-01 5.62230945e-01 7.30290949e-01 6.11561239e-01
-7.83410370e-02 8.12800467e-01 3.04822445e-01 3.02754104e-01
-4.33377683e-01 1.52125418e-01 3.98728997e-01 7.69453168e-01
-1.07645221e-01 1.28206775e-01 -8.90456557e-01 7.39332199e-01
-1.78677499e+00 -1.36520612e+00 1.69258118e-01 2.21542001e+00
8.72136950e-01 3.49028409e-02 5.93723953e-01 6.93611085e-01
5.27809918e-01 3.19748938e-01 -2.26953208e-01 -5.78115761e-01
1.43016383e-01 3.92004699e-01 7.63546228e-02 6.46959841e-02
-1.42612278e+00 6.67617023e-01 7.26080799e+00 1.10834277e+00
-8.48927259e-01 5.84659219e-01 2.75089353e-01 -5.57733119e-01
3.58893484e-01 -2.59572238e-01 -5.86811125e-01 4.29351896e-01
1.50956392e+00 2.18497273e-02 4.15392876e-01 1.18997562e+00
-1.70798786e-02 -4.55953963e-02 -1.26771712e+00 1.20362961e+00
3.33571225e-01 -1.25139296e+00 -2.87266254e-01 -2.59176433e-01
8.23379457e-01 4.80697513e-01 1.76442917e-02 6.63735390e-01
-6.47672117e-02 -7.21401691e-01 8.49211454e-01 2.64595479e-01
9.50821459e-01 -6.95581734e-01 4.72976536e-01 1.75786316e-01
-1.21213460e+00 -3.59182179e-01 -4.85928595e-01 -1.31623283e-01
1.71776446e-05 2.15795189e-01 -7.67445028e-01 4.33931142e-01
9.84437108e-01 8.16653848e-01 -6.48360014e-01 1.40736103e+00
3.72198492e-01 8.40917528e-01 -1.98468119e-01 1.29268900e-01
3.73336911e-01 3.60868096e-01 7.20927596e-01 1.91351128e+00
1.90533131e-01 -2.82455951e-01 1.64073333e-01 2.24630341e-01
2.25841850e-01 4.60075080e-01 -8.66406858e-01 1.40317753e-01
3.57181877e-01 1.23575974e+00 -5.96519768e-01 -4.22811568e-01
-5.19746661e-01 7.32497931e-01 1.08656362e-02 8.07530060e-02
-6.75572395e-01 -5.90546727e-01 7.41028309e-01 -9.81481224e-02
4.57380921e-01 1.05577640e-01 -1.67686179e-01 -9.42235947e-01
-2.84515053e-01 -9.89399672e-01 7.69941807e-01 -7.59745181e-01
-1.21340883e+00 4.58835512e-01 3.99471164e-01 -1.63857234e+00
-5.10435820e-01 -3.73575121e-01 -8.43756258e-01 2.08162293e-01
-1.42315090e+00 -1.01210129e+00 -2.76621103e-01 6.65717542e-01
1.10718191e+00 -5.49346149e-01 1.07649088e+00 4.87200528e-01
-1.98405787e-01 4.91074830e-01 1.35211900e-01 -2.92341053e-01
1.01240122e+00 -1.20865750e+00 1.75304845e-01 5.19940972e-01
7.38703907e-01 2.09920719e-01 1.03777075e+00 -4.69392017e-02
-1.21393085e+00 -9.84745800e-01 2.69688606e-01 -5.22790372e-01
9.07019198e-01 -3.01110625e-01 -9.16409910e-01 2.82418907e-01
3.95197809e-01 7.05387369e-02 1.07511163e+00 1.72644079e-01
-5.07262647e-01 -1.17993519e-01 -1.05934751e+00 2.74831802e-01
1.11216366e+00 -8.40402424e-01 -6.96337998e-01 4.54421580e-01
6.21428370e-01 -1.73096493e-01 -8.91370595e-01 3.62717271e-01
5.63400745e-01 -1.24560833e+00 1.10186684e+00 -7.43978977e-01
5.12871444e-01 1.14549659e-01 -6.19610071e-01 -1.36047244e+00
-6.84871897e-02 -7.12232828e-01 -2.06862405e-01 1.53270483e+00
1.64201677e-01 -1.12528563e-01 4.98091221e-01 -1.58360656e-02
-3.28441143e-01 -5.60688913e-01 -1.35736835e+00 -1.07060361e+00
-7.06945732e-02 -1.25761306e+00 2.94064015e-01 7.69173801e-01
1.54874370e-01 2.76076257e-01 -5.14823020e-01 -3.32905948e-01
6.31094337e-01 -2.25854889e-01 1.01348042e+00 -1.12336683e+00
-5.96146524e-01 -4.83527184e-01 -9.44653213e-01 -1.84593007e-01
7.88527727e-02 -8.78443778e-01 2.20865488e-01 -1.19272041e+00
1.03200585e-01 -2.83386055e-02 -5.58565140e-01 3.28505337e-01
1.75092369e-01 5.99509299e-01 9.61724073e-02 1.72656119e-01
-8.68754148e-01 3.96849334e-01 6.82189763e-01 -2.18181059e-01
-1.46036088e-01 -1.72758289e-02 -3.40020120e-01 7.32793510e-01
6.80455565e-01 -5.40937901e-01 -5.78470945e-01 -8.18228573e-02
-1.02010272e-01 -3.40318978e-01 4.62356538e-01 -1.65525830e+00
2.53421515e-01 6.93656653e-02 4.51862924e-02 -1.38177678e-01
5.85914791e-01 -6.03318095e-01 -3.00018955e-02 2.77630836e-01
-5.64031303e-01 -4.34472859e-01 2.85818130e-01 8.08209777e-01
-3.22583109e-01 -5.26505053e-01 1.14002359e+00 -3.99662405e-01
-1.16093457e+00 1.83153227e-01 -1.88374847e-01 4.56012905e-01
9.91563857e-01 -5.35818696e-01 1.23226725e-01 -4.12808985e-01
-9.77819622e-01 -1.18171051e-01 3.05766225e-01 5.46842575e-01
3.54781151e-01 -1.25788736e+00 -7.39877880e-01 -8.75978079e-03
6.56602383e-01 -5.84518373e-01 4.23003823e-01 9.69447196e-01
-1.67386383e-01 2.41429806e-01 -3.21594536e-01 -8.50489378e-01
-1.54909587e+00 5.56657493e-01 3.86612475e-01 2.19262272e-01
-7.24970818e-01 1.03318894e+00 -3.20773810e-01 -4.09343064e-01
6.37006044e-01 -1.74751550e-01 -1.52537048e-01 5.54274380e-01
7.71603107e-01 7.52621174e-01 2.24635571e-01 -5.80406547e-01
-2.88851291e-01 5.16473651e-01 2.88514197e-01 -4.71140265e-01
1.68818891e+00 2.26289071e-02 4.16723639e-01 1.19439197e+00
1.36383092e+00 -4.00030464e-01 -1.33782542e+00 -2.02853963e-01
6.33822307e-02 -4.56005067e-01 5.59769161e-02 -5.27874589e-01
-4.67232198e-01 1.32427311e+00 1.00743103e+00 2.46586189e-01
1.06413424e+00 -1.39422156e-02 4.27926660e-01 4.46576059e-01
3.86268795e-01 -1.48187828e+00 3.17193061e-01 3.09917003e-01
7.82005012e-01 -1.37991500e+00 6.13218471e-02 1.65943623e-01
-8.05870652e-01 1.16581047e+00 3.33542019e-01 -9.84768644e-02
7.31553793e-01 2.79964685e-01 -1.44937085e-02 -4.38581184e-02
-1.10902011e+00 -5.09540915e-01 3.58142674e-01 1.01269424e+00
4.18811709e-01 -3.38124335e-01 1.14419296e-01 3.22248489e-01
-2.07605496e-01 1.47719428e-01 3.41202676e-01 1.18578744e+00
-6.61035120e-01 -9.38006759e-01 -2.01641738e-01 3.41603220e-01
-6.63970172e-01 -3.77327800e-02 -3.49893481e-01 8.55335891e-01
2.69114137e-01 9.45433378e-01 -5.80619313e-02 -4.06027943e-01
3.43638688e-01 6.50978565e-01 5.99085689e-01 -9.27335560e-01
-7.19974458e-01 1.95255473e-01 2.21813053e-01 -5.25536060e-01
-8.75225663e-01 -8.56667101e-01 -7.09405303e-01 1.87534541e-01
-4.32243586e-01 1.82612636e-03 8.00066650e-01 8.86010349e-01
1.88732415e-01 5.22825181e-01 5.12468159e-01 -1.28170228e+00
-7.39534259e-01 -1.13762820e+00 -6.56866848e-01 4.54281956e-01
2.58989275e-01 -7.80283928e-01 -5.42483509e-01 3.71677816e-01]
|
[15.186585426330566, 5.075692176818848]
|
b5c28f86-bbd0-4eae-a1b9-b98f116f67dc
|
semantic-segmentation-of-anaemic-rbcs-using
|
2202.04650
| null |
https://arxiv.org/abs/2202.04650v1
|
https://arxiv.org/pdf/2202.04650v1.pdf
|
Semantic Segmentation of Anaemic RBCs Using Multilevel Deep Convolutional Encoder-Decoder Network
|
Pixel-level analysis of blood images plays a pivotal role in diagnosing blood-related diseases, especially Anaemia. These analyses mainly rely on an accurate diagnosis of morphological deformities like shape, size, and precise pixel counting. In traditional segmentation approaches, instance or object-based approaches have been adopted that are not feasible for pixel-level analysis. The convolutional neural network (CNN) model required a large dataset with detailed pixel-level information for the semantic segmentation of red blood cells in the deep learning domain. In current research work, we address these problems by proposing a multi-level deep convolutional encoder-decoder network along with two state-of-the-art healthy and Anaemic-RBC datasets. The proposed multi-level CNN model preserved pixel-level semantic information extracted in one layer and then passed to the next layer to choose relevant features. This phenomenon helps to precise pixel-level counting of healthy and anaemic-RBC elements along with morphological analysis. For experimental purposes, we proposed two state-of-the-art RBC datasets, i.e., Healthy-RBCs and Anaemic-RBCs dataset. Each dataset contains 1000 images, ground truth masks, relevant, complete blood count (CBC), and morphology reports for performance evaluation. The proposed model results were evaluated using crossmatch analysis with ground truth mask by finding IoU, individual training, validation, testing accuracies, and global accuracies using a 05-fold training procedure. This model got training, validation, and testing accuracies as 0.9856, 0.9760, and 0.9720 on the Healthy-RBC dataset and 0.9736, 0.9696, and 0.9591 on an Anaemic-RBC dataset. The IoU and BFScore of the proposed model were 0.9311, 0.9138, and 0.9032, 0.8978 on healthy and anaemic datasets, respectively.
|
['Israr Ahmed Shaikh', 'Syed Hamad Shirazi', 'Arif Iqbal Umar', 'Muhammad Shahzad']
|
2022-02-09
| null | null | null | null |
['morphological-analysis']
|
['natural-language-processing']
|
[-2.47316897e-01 -2.04392195e-01 3.71411920e-01 -4.66965675e-01
-3.48012924e-01 1.70911085e-02 7.99474046e-02 7.21077919e-01
-5.99305391e-01 8.97682369e-01 -1.72364116e-01 8.28280225e-02
5.07119019e-03 -1.23409367e+00 -2.74829000e-01 -8.89544010e-01
-7.50854984e-02 6.92295849e-01 3.63426775e-01 6.97825775e-02
3.99946064e-01 8.56407702e-01 -1.42019403e+00 4.60983306e-01
9.98814344e-01 1.29708815e+00 -3.10255259e-01 1.02363217e+00
-5.14983594e-01 9.45087612e-01 -8.40843499e-01 -5.26326954e-01
8.06252807e-02 -5.24859190e-01 -3.49475980e-01 1.86524749e-01
1.03309311e-01 -5.01535296e-01 -2.08283067e-01 7.94575155e-01
8.63459885e-01 -8.98590013e-02 6.98088467e-01 -8.90948057e-01
-6.10367358e-01 3.20421964e-01 -9.13790226e-01 5.91921031e-01
-1.34842023e-01 3.14281732e-01 1.66741744e-01 -5.05611718e-01
6.32156909e-04 1.19659472e+00 6.81236267e-01 2.80457467e-01
-8.04106653e-01 -5.53061724e-01 -8.12916994e-01 2.90239509e-02
-1.41148984e+00 -2.37830594e-01 4.40407962e-01 -5.57598531e-01
8.38447750e-01 2.37030044e-01 8.36557865e-01 -1.57978296e-01
4.48670834e-01 5.37157357e-01 1.24111474e+00 -4.58110094e-01
1.79622889e-01 4.22041826e-02 3.07136744e-01 1.00168562e+00
8.59486043e-01 -2.01114953e-01 -1.99436739e-01 2.59102911e-01
1.04804289e+00 5.75914979e-02 8.05219784e-02 9.58276615e-02
-7.03567207e-01 7.69610584e-01 4.35422391e-01 5.30353963e-01
-4.74292755e-01 -1.94423079e-01 4.61466521e-01 -8.96958709e-02
2.24745587e-01 8.51661488e-02 -2.01417163e-01 8.59115794e-02
-8.66766632e-01 -9.39991772e-02 6.51382565e-01 5.28239429e-01
7.47337639e-01 2.01678395e-01 -4.24201965e-01 9.44317043e-01
5.55163138e-02 5.71661294e-01 4.51721340e-01 -7.58006990e-01
-1.55801820e-02 9.18536425e-01 -4.35995758e-02 -1.11112607e+00
-7.41546810e-01 -3.57730865e-01 -1.20053613e+00 1.24493457e-01
6.30616605e-01 5.62137505e-03 -1.02090836e+00 1.21828759e+00
4.08746362e-01 -5.68511635e-02 2.63711154e-01 9.26749468e-01
1.21614015e+00 6.06989443e-01 2.45755419e-01 -2.25417674e-01
1.52516329e+00 -6.92096889e-01 -5.90183556e-01 4.62762356e-01
4.93289322e-01 -8.30223501e-01 7.77962148e-01 3.53973389e-01
-1.34738100e+00 -8.69063437e-01 -9.84800577e-01 -8.72264877e-02
-2.86056668e-01 5.25492847e-01 5.69429815e-01 7.79172897e-01
-1.04887497e+00 3.87321323e-01 -7.37200558e-01 -3.71592879e-01
7.92858601e-01 4.75951880e-01 -2.21819326e-01 -1.72897819e-02
-7.44852722e-01 7.19812214e-01 3.04570705e-01 1.92811608e-01
-6.42669141e-01 -6.99915409e-01 -6.43800974e-01 8.07448551e-02
-2.71636695e-01 -5.04598320e-01 6.33954346e-01 -5.51430881e-01
-1.21046412e+00 1.14998305e+00 1.79922625e-01 -4.20786142e-01
4.29973483e-01 1.52442083e-01 -4.46370721e-01 7.30286300e-01
1.50809631e-01 7.42089748e-01 6.90410957e-02 -1.00298369e+00
-8.03708732e-01 -5.40897548e-01 -2.83326864e-01 -1.91713378e-01
7.96582103e-02 1.00067616e-01 -4.68358070e-01 -3.97976786e-01
1.12701379e-01 -4.81751673e-02 1.19289972e-01 1.45899102e-01
-1.78868935e-01 -2.68786967e-01 3.83372784e-01 -9.31718886e-01
9.52354252e-01 -1.82428038e+00 -6.60056531e-01 1.45978585e-01
2.70655245e-01 4.70153242e-01 1.40207723e-01 -1.52068481e-01
1.99340940e-01 -1.00413345e-01 -3.80903691e-01 1.36998938e-02
-2.75066763e-01 7.36988261e-02 4.89290833e-01 6.37648880e-01
4.66331959e-01 9.07177746e-01 -6.00689828e-01 -1.12502670e+00
6.00198269e-01 7.52249479e-01 -2.85796136e-01 5.52317679e-01
2.27439836e-01 4.19356227e-01 -1.27514377e-01 8.89497578e-01
8.92779291e-01 -2.34869197e-01 -2.08338737e-01 -6.47387922e-01
-1.71225652e-01 -4.71958548e-01 -9.66246784e-01 1.02186203e+00
-3.31420213e-01 4.28547770e-01 -4.93166745e-02 -1.11560166e+00
1.37091672e+00 2.33080700e-01 7.85847127e-01 -1.01476276e+00
6.15170300e-01 3.68255526e-01 2.88533300e-01 -8.63905907e-01
1.05542421e-01 -4.09274727e-01 3.58082294e-01 3.96162122e-01
-1.00480162e-01 2.36953557e-01 6.05435789e-01 -7.24162906e-03
7.21599579e-01 -1.93595409e-01 2.78257877e-01 -5.70448339e-01
9.76162732e-01 1.80027321e-01 5.45178950e-01 4.14141148e-01
-4.31845546e-01 7.01259851e-01 4.29102987e-01 -8.40481281e-01
-1.01721776e+00 -9.52412188e-01 -5.50715387e-01 3.89418751e-01
1.26169950e-01 4.91113275e-01 -9.28972065e-01 -1.52661934e-01
1.15764879e-01 3.01606715e-01 -5.84500492e-01 6.73276857e-02
-8.51486027e-01 -1.15276086e+00 7.27749050e-01 6.04868829e-01
1.15897322e+00 -1.27832282e+00 -9.39336777e-01 2.96074003e-01
1.71907440e-01 -9.22064960e-01 1.23244397e-01 -2.83792883e-01
-9.06298399e-01 -1.54659498e+00 -8.27033043e-01 -7.29371846e-01
8.21089566e-01 -1.55083731e-01 1.14271879e+00 6.21444702e-01
-9.20496464e-01 -1.00757219e-01 -1.50445417e-01 -1.05560534e-01
-2.12607577e-01 -3.78786951e-01 -2.71267235e-01 2.28740633e-01
6.34282589e-01 -6.02139123e-02 -8.82618666e-01 1.66788459e-01
-8.50928783e-01 -1.29406586e-01 7.35901892e-01 7.21628129e-01
7.15263188e-01 2.44866624e-01 8.14599335e-01 -8.62378299e-01
2.27479890e-01 -2.00077340e-01 -5.72304189e-01 4.51068014e-01
-5.17310560e-01 -3.92852932e-01 6.65973544e-01 5.14898412e-02
-1.06708443e+00 -3.35332870e-01 -2.24473909e-01 -1.11631326e-01
-4.53279853e-01 1.24677092e-01 -9.30446982e-02 3.25868130e-02
3.25478256e-01 2.86981344e-01 8.50360319e-02 -3.00050378e-01
-1.45696804e-01 8.71508777e-01 8.47553432e-01 -5.88507652e-01
9.47579369e-02 5.82688570e-01 2.45667696e-01 -8.54094207e-01
-5.45417249e-01 -3.28546762e-01 -7.86725223e-01 -2.62854129e-01
1.08697474e+00 -7.53480971e-01 -1.01121652e+00 9.88758266e-01
-9.38758850e-01 -2.58744121e-01 -2.27120325e-01 4.31105375e-01
-2.40569711e-01 5.38175046e-01 -1.07676709e+00 -9.00228202e-01
-8.53720009e-01 -1.04797411e+00 7.32393265e-01 8.12133610e-01
1.13894336e-01 -1.11432564e+00 -2.33186245e-01 4.83245522e-01
6.55161202e-01 5.91488361e-01 1.06894708e+00 -7.06855476e-01
-3.65520239e-01 -3.04232568e-01 -8.57262909e-01 2.26684466e-01
3.81735831e-01 2.13206276e-01 -8.38811874e-01 -2.52355784e-01
1.11475714e-01 -4.07810837e-01 7.86184132e-01 9.58479881e-01
1.21592188e+00 -1.27338469e-01 8.85121822e-02 6.40057862e-01
1.64611483e+00 5.36387146e-01 7.79780447e-01 4.39116126e-03
5.48427045e-01 7.29034960e-01 5.29364586e-01 7.13510811e-01
5.27752817e-01 4.87368219e-02 -3.89070176e-02 -5.42859972e-01
-3.75704437e-01 2.67432958e-01 -3.94262224e-01 7.00484574e-01
5.25300689e-02 -2.08751872e-01 -9.38889802e-01 6.56820476e-01
-1.20354342e+00 -6.27758384e-01 -4.64802057e-01 1.99619389e+00
9.12806749e-01 3.16734612e-02 2.54023194e-01 2.40172759e-01
1.06233728e+00 -4.39046174e-01 -5.58022916e-01 -3.54963541e-01
-5.21955252e-01 8.34603310e-01 4.22563791e-01 3.80486101e-01
-9.28972542e-01 5.80156565e-01 4.84667444e+00 6.71212018e-01
-1.17373741e+00 2.11479347e-02 1.38743651e+00 1.47230076e-02
2.58397818e-01 -4.63311642e-01 -7.95434952e-01 6.76820457e-01
6.71170056e-01 2.24273205e-01 -1.07192360e-02 1.73309565e-01
2.92503357e-01 -7.46405602e-01 -7.85247982e-01 1.03609264e+00
1.39048859e-01 -1.31481183e+00 -5.58331050e-03 -3.20025444e-01
6.17941797e-01 -5.49442351e-01 -4.06559706e-01 1.45826921e-01
6.12392314e-02 -1.20354569e+00 1.34929657e-01 7.94602334e-01
1.03009510e+00 -6.23853028e-01 1.38500798e+00 -2.35134158e-02
-1.10568202e+00 3.26509267e-01 -8.12466800e-01 9.99087468e-02
-1.04941927e-01 7.95318186e-01 -4.98117924e-01 2.78959304e-01
7.23756015e-01 3.38223785e-01 -5.95446527e-01 1.19566929e+00
2.03862190e-01 3.95466328e-01 -2.40851685e-01 -1.37436269e-02
-1.06337041e-01 -2.99487084e-01 -3.69361728e-01 1.33303499e+00
2.01930478e-01 6.15910888e-01 -1.40088066e-01 8.12698364e-01
-5.89555316e-02 3.14687401e-01 8.80268142e-02 3.91250521e-01
3.28774184e-01 1.25175703e+00 -1.23739922e+00 -6.84368134e-01
-2.49370024e-01 3.33947182e-01 1.00123435e-01 9.17817131e-02
-7.91394413e-01 -7.00112045e-01 5.02280332e-02 1.44321904e-01
1.00833261e-02 2.95874804e-01 -8.30681264e-01 -7.60466993e-01
-3.44117731e-01 -3.54217052e-01 7.86465645e-01 -5.51911473e-01
-1.35216808e+00 5.17005503e-01 -1.33843854e-01 -5.29965162e-01
3.89614433e-01 -6.05341375e-01 -7.81390190e-01 1.05338168e+00
-1.58230436e+00 -9.07392263e-01 -8.19874823e-01 3.20572704e-01
3.01251084e-01 -4.00741905e-01 3.03505450e-01 4.67355639e-01
-8.78569782e-01 6.09119713e-01 -4.02628891e-02 6.09232724e-01
3.78575325e-01 -1.16894209e+00 -2.55391002e-01 7.53613174e-01
-5.82756996e-01 3.75143141e-01 2.77877331e-01 -5.74383795e-01
-9.94806349e-01 -1.04560745e+00 7.45548904e-01 1.39721587e-01
1.18901646e-02 2.52620071e-01 -1.06398320e+00 1.16724009e-02
8.11905041e-02 3.57043952e-01 7.29372919e-01 -7.04289854e-01
2.78283328e-01 -4.96509045e-01 -1.81855965e+00 2.52245646e-02
2.37694517e-01 6.85276687e-02 -2.30128363e-01 8.12433586e-02
2.12035060e-01 -5.91779292e-01 -1.19662344e+00 3.98494482e-01
7.30788887e-01 -1.27117240e+00 9.21891212e-01 -3.54935735e-01
5.93516171e-01 -2.76709288e-01 -1.81969404e-01 -5.59803605e-01
-8.54692236e-02 1.69268370e-01 -9.98068601e-03 1.29544175e+00
-7.33799934e-02 -6.69262230e-01 9.27482128e-01 4.14624006e-01
-3.47751707e-01 -1.08077681e+00 -7.92765915e-01 -3.64649922e-01
3.67352724e-01 1.89741179e-01 9.19550002e-01 7.55740166e-01
-5.32366216e-01 -3.46839875e-02 7.76064619e-02 -2.44072396e-02
7.73382008e-01 4.01713938e-01 6.65901482e-01 -1.11093700e+00
2.23895367e-02 -6.64520025e-01 -4.81068850e-01 -6.04219139e-01
-3.05771887e-01 -5.02800584e-01 -2.54153162e-01 -1.87048960e+00
3.99403244e-01 -8.21586311e-01 -5.00453711e-01 4.59292382e-01
-1.94013953e-01 6.37126923e-01 -2.93589830e-01 1.89028740e-01
-1.96646646e-01 1.42284259e-01 1.33679044e+00 -4.06336337e-02
-1.01950787e-01 -1.17193699e-01 -4.03026432e-01 2.45427564e-01
1.24875462e+00 -1.25731155e-01 -1.35195717e-01 -4.93003726e-01
-4.12455171e-01 2.26985335e-01 3.58353823e-01 -1.16650569e+00
2.04181314e-01 -1.07129313e-01 1.02553594e+00 -9.20494616e-01
8.13214481e-02 -3.22752088e-01 1.58316493e-02 1.05552316e+00
-1.86467886e-01 -1.19133055e-01 1.51061729e-01 -9.24597010e-02
-2.33745500e-01 -2.45446071e-01 1.36877096e+00 -2.79154450e-01
-4.90572482e-01 2.83963233e-01 3.53184529e-02 1.19409442e-01
1.12468874e+00 -7.33584106e-01 -6.14824653e-01 5.90843074e-02
-5.23393154e-01 2.49797225e-01 4.29896265e-01 -3.63232225e-01
9.08170283e-01 -9.56963837e-01 -1.04028022e+00 5.51723838e-01
-1.39204547e-01 2.24112108e-01 4.52593952e-01 1.01678193e+00
-1.43383181e+00 5.31294346e-01 -6.61782265e-01 -8.69700134e-01
-9.94865954e-01 3.67656708e-01 6.71972752e-01 -1.49222910e-01
-4.16015893e-01 8.40120971e-01 1.63256526e-01 -2.86707133e-02
-3.04062907e-02 -4.44361776e-01 -3.81559521e-01 -1.21868350e-01
4.06135231e-01 7.53228486e-01 8.77500474e-02 -7.73744881e-01
-2.82356381e-01 8.73843908e-01 3.49737287e-01 4.74631876e-01
1.10886037e+00 -5.92559576e-02 -4.25350875e-01 1.47929624e-01
1.04408967e+00 -2.24969164e-01 -9.68100250e-01 -1.01778448e-01
-3.00606459e-01 -5.61142802e-01 -1.17466249e-01 -8.86051953e-01
-1.65084207e+00 1.37395108e+00 9.80084121e-01 1.90216541e-01
1.33792114e+00 -1.20159090e-01 1.04019463e+00 -2.60494024e-01
2.73892671e-01 -8.50736618e-01 5.29963151e-03 -5.36254756e-02
3.48778695e-01 -1.31807899e+00 2.97471255e-01 -3.81986260e-01
-4.70710009e-01 1.38921583e+00 9.76783752e-01 3.92335653e-02
6.16191328e-01 3.26349050e-01 2.50579029e-01 -4.15933847e-01
-3.55444640e-01 -2.17550889e-01 -5.88812418e-02 6.82931662e-01
7.17370510e-01 1.99190035e-01 -4.59935457e-01 5.39033234e-01
-5.42211272e-02 2.07342312e-01 3.93801928e-01 8.39143455e-01
-8.85459125e-01 -5.82180500e-01 -4.50171620e-01 7.12809741e-01
-7.69875467e-01 2.24038884e-01 1.08708985e-01 8.53756666e-01
3.80355746e-01 1.05686724e+00 6.29682362e-01 -4.16486301e-02
1.86022267e-01 -2.59099394e-01 6.08588159e-01 -3.28706264e-01
-6.55467331e-01 1.43296897e-01 -2.02179134e-01 -3.54148336e-02
-4.41962242e-01 -3.55230063e-01 -1.78213882e+00 -5.74316502e-01
-1.91681579e-01 1.33538768e-01 2.82814592e-01 7.58452594e-01
-4.72086780e-02 4.86384183e-01 5.59870660e-01 -1.46997631e-01
-1.52033374e-01 -1.02813852e+00 -8.39046419e-01 5.67115128e-01
1.17760777e-01 -5.67213833e-01 -6.52826205e-02 1.38027981e-01]
|
[14.894109725952148, -3.1129302978515625]
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.