Pixel-by-pixel Image Classification on MNIST permuted
96.3AccuracyLipschitz RNN using Euler
Evaluation results
| Method | Links | |
|---|---|---|
| Lipschitz RNN using EulerN=128, # params=≈31K2020.06 | 96.3 | |
| Exponential RNNN=360, # params=≈69K2020.06 | 96.2 | |
| Lipschitz RNN using RK2N=128, # params=≈34K2020.06 | 96.2 | |
| Antisymmteric RNNN=128, # params=≈10K2020.06 | 95.8 | |
| Incremental RNNN=128, # params=≈4K/8K2020.06 | 95.6 | |
| MomentumLSTMN=256, # params=≈270K2020.06 | 94.7 | |
| Kronecker RNNN=512, # params=≈11K2020.06 | 94.5 | |
| Lipschitz RNN using EulerN=64, # params=≈9K2020.06 | 94.2 | |
| Lipschitz RNN using RK2N=64, # params=≈9K2020.06 | 94.2 | |
| Full Capacity Unitary RNNN=512, # params=≈270K2020.06 | 94.1 | |
| LSTM baselineN=128, # params=≈68K2020.06 | 92.7 | |
| Unitary RNNN=512, # params=≈9K2020.06 | 91.4 | |
| Soft orth. RNNN=128, # params=≈18K2020.06 | 91.4 | |
| Sequential NAIS-NetN=128, # params=≈18K2020.06 | 90.8 |