Image Classification on CIFAR-10 noise padded
59AccuracyLipschitz RNN using Euler
Evaluation results
| Method | Links | |
|---|---|---|
| Lipschitz RNN using EulerN=256, # params=134K/158K2020.06 | 59 | |
| Lipschitz RNN using RK2N=256, # params=134K/158K2020.06 | 58.9 | |
| Lipschitz RNN using EulerN=128, # params=34K/46K2020.06 | 57.4 | |
| Lipschitz RNN using RK2N=128, # params=34K/46K2020.06 | 57.3 | |
| AntisymmetricRNN w/ gating# units=256, # params=37k2019.02 | 54.7 | |
| Incremental RNNN=1282020.06 | 54.5 | |
| Antisymmetric RNNN=256, # params=36K2020.06 | 48.3 | |
| AntisymmetricRNN# units=256, # params=36k2019.02 | 48.3 | |
| Ablation model# units=196, # params=42k2019.02 | 46.2 | |
| LSTM baselineN=128, # params=69K2020.06 | 11.6 | |
| LSTM# units=128, # params=69k2019.02 | 11.6 |