List your company
ResearchFormerly WizWand

Pixel-by-pixel Image Classification on Permuted Sequential MNIST (pMNIST) (test)

98.76AccuracyLSSL
32.449649.664866.8884.0952Mar 30, 2016Mar 5, 2017Feb 8, 2018Jan 14, 2019Dec 20, 2019Nov 24, 2020Oct 31, 2021
Updated 8mo ago

Evaluation results

MethodLinks
LSSLState Transition=Learned2021.1098.76
LSSL2021.1098.76
S42021.1098.7
LSSL-fixedState Transition=Fixed2021.1098.6
CKConv2021.1098.54
CKConv2021.1098.54
LMUFFT2021.1098.49
LMU-FFT2021.1098.49
HiPPO-LegS2020.0898.3
HiPPO-RNN2021.1098.3
HiPPO-RNN2021.1098.3
HiPPO-Laguerre2020.0898.24
TrellisNet2019.1098.13
TrellisNet2020.0898.13
TrellisNet2021.1098.13
TrellisNet2021.1098.13
HiPPO-LegT2020.0898.03
Transformer2019.1097.9
Transformer2020.0897.9
Transformer2021.1097.9
Transformer2021.1097.9
URLSTM + Zoneoutregularization=Zoneout2019.1097.58
URLSTMregularization=Zoneout2020.0897.58
LMUimplementation=ours2020.0897.29
Temporal ConvNet2019.1097.2
Temporal ConvNet2020.0897.2
TCN2021.1097.2
TCN2021.1097.2
LMUreference=[71]2020.0897.15
URLSTM2019.1096.96
URLSTMreference=[31]2020.0896.96
UR-LSTM2021.1096.96
URGRU + Zoneoutregularization=Zoneout2019.1096.51
URGRU2021.1096.51
UR-GRU2021.1096.51
LipschitzRNN2021.1096.3
LipschitzRNN2021.1096.3
Dilated RNN2020.0896.1
Dilated RNN2021.1096.1
IndRNN2019.1096
IndRNN2020.0896
IndRNNunits=1282019.1196
IndRNN2021.1096
BN-LSTMimplementation=ours, hidden units=1002016.0395.4
BN-LSTMunits=1002019.1195.4
LSTM w/f STARNumber of layers=8, units=1282019.1195.4
r-LSTMTraining Protocol=Full BP2018.0395.2
r-LSTMlayers=2, auxiliary loss=true2019.1095.2
r-LSTM2020.0895.2
r-LSTM2021.1095.2
r-LSTM2021.1095.2
LSTMImplementation=ours2019.1095.11
LSTM2020.0895.11
LSTM2021.1095.11
STARNumber of layers=12, units=1282019.1194.7
Dilated GRU2019.1094.6
FC URNNunits=5122019.1194.1
STARNumber of layers=8, units=1282019.1194.1
sTANH-RNN2016.0394
STANH-RNNunits=1282019.1194
GRUNumber of layers=2, units=1282019.1193.9
AntisymRNNunits=1282019.1193.1
r-LSTMTraining Protocol=Truncate3002018.0392.8
p-LSTMTraining Protocol=Full BP2018.0392.8
p-LSTMTraining Protocol=Truncate3002018.0392.5
LSTMNumber of layers=2, units=1282019.1191.9
uRNN2016.0391.4
uRNN2018.0391.4
uRNNunits=5122019.1191.4
Soft orthounits=1282019.1191.4
LSTMimplementation=ours, hidden units=1002016.0390.2
RHNNumber of layers=2, units=1282019.1189.5
LSTMTraining Protocol=Full BP2018.0389.4
LSTMTraining Protocol=Truncate3002018.0388.8
iRNN2016.0382
iRNN2018.0382
iRNNunits=1002019.1182
vRNNNumber of layers=1, units=1282019.1144
TANH-RNN2016.0335