Image Classification on MNIST-scale (test)
1.42Classification ErrorSESN Vector
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SESN VectorResolution=56x56, Data Augmentation=true, # Params=495 K2019.10 | 1.42 | — | |
| SESN ScalarResolution=56x56, Data Augmentation=true, # Params=495 K2019.10 | 1.5 | — | |
| DSS ScalarResolution=56x56, Data Augmentation=true, # Params=494 K2019.10 | 1.57 | — | |
| DSS VectorResolution=56x56, Data Augmentation=true, # Params=494 K2019.10 | 1.57 | — | |
| SiCNNResolution=56x56, Data Augmentation=true, # Params=497 K2019.10 | 1.59 | — | |
| SI-ConvNetResolution=56x56, Data Augmentation=true, # Params=495 K2019.10 | 1.59 | — | |
| CNNResolution=56x56, Data Augmentation=true, # Params=495 K2019.10 | 1.6 | — | |
| SEVF ScalarResolution=56x56, Data Augmentation=true, # Params=494 K2019.10 | 1.62 | — | |
| SESN VectorResolution=56x56, Data Augmentation=false, # Params=495 K2019.10 | 1.68 | — | |
| SESN ScalarResolution=56x56, Data Augmentation=false, # Params=495 K2019.10 | 1.74 | — | |
| SS-CNNResolution=56x56, Data Augmentation=true, # Params=494 K2019.10 | 1.76 | — | |
| SESN VectorResolution=28x28, Data Augmentation=true, # Params=495 K2019.10 | 1.76 | — | |
| SESN ScalarResolution=28x28, Data Augmentation=true, # Params=495 K2019.10 | 1.79 | — | |
| SEVF VectorResolution=56x56, Data Augmentation=true, # Params=475 K2019.10 | 1.81 | — | |
| SI-ConvNetResolution=56x56, Data Augmentation=false, # Params=495 K2019.10 | 1.82 | — | |
| SS-CNNResolution=56x56, Data Augmentation=false, # Params=494 K2019.10 | 1.84 | — | |
| SiCNNResolution=28x28, Data Augmentation=true, # Params=497 K2019.10 | 1.86 | — | |
| SEVF ScalarResolution=56x56, Data Augmentation=false, # Params=494 K2019.10 | 1.87 | — | |
| DSS ScalarResolution=56x56, Data Augmentation=false, # Params=494 K2019.10 | 1.92 | — | |
| SI-ConvNetResolution=28x28, Data Augmentation=true, # Params=495 K2019.10 | 1.94 | — | |
| DSS VectorResolution=28x28, Data Augmentation=true, # Params=494 K2019.10 | 1.95 | — | |
| CNNResolution=28x28, Data Augmentation=true, # Params=495 K2019.10 | 1.96 | — | |
| SEVF ScalarResolution=28x28, Data Augmentation=true, # Params=494 K2019.10 | 1.96 | — | |
| DSS VectorResolution=56x56, Data Augmentation=false, # Params=494 K2019.10 | 1.97 | — | |
| CNNResolution=56x56, Data Augmentation=false, # Params=495 K2019.10 | 2.02 | — | |
| SiCNNResolution=56x56, Data Augmentation=false, # Params=497 K2019.10 | 2.02 | — | |
| DSS ScalarResolution=28x28, Data Augmentation=true, # Params=494 K2019.10 | 2.04 | — | |
| SESN VectorResolution=28x28, Data Augmentation=false, # Params=495 K2019.10 | 2.08 | — | |
| SS-CNNResolution=28x28, Data Augmentation=true, # Params=494 K2019.10 | 2.1 | — | |
| SESN ScalarResolution=28x28, Data Augmentation=false, # Params=495 K2019.10 | 2.1 | — | |
| SEVF VectorResolution=56x56, Data Augmentation=false, # Params=475 K2019.10 | 2.12 | — | |
| SEVF VectorResolution=28x28, Data Augmentation=true, # Params=475 K2019.10 | 2.23 | — | |
| SEVF ScalarResolution=28x28, Data Augmentation=false, # Params=494 K2019.10 | 2.3 | — | |
| SS-CNNResolution=28x28, Data Augmentation=false, # Params=494 K2019.10 | 2.32 | — | |
| SiCNNResolution=28x28, Data Augmentation=false, # Params=497 K2019.10 | 2.4 | — | |
| SI-ConvNetResolution=28x28, Data Augmentation=false, # Params=495 K2019.10 | 2.4 | — | |
| DSS ScalarResolution=28x28, Data Augmentation=false, # Params=494 K2019.10 | 2.53 | — | |
| CNNResolution=28x28, Data Augmentation=false, # Params=495 K2019.10 | 2.56 | — | |
| DSS VectorResolution=28x28, Data Augmentation=false, # Params=494 K2019.10 | 2.58 | — | |
| SEVF VectorResolution=28x28, Data Augmentation=false, # Params=475 K2019.10 | 2.63 | — | |
| Augerino2023.06 | — | 97.53 | |
| CNNNumber of training samples=50002023.11 | — | 94.32 | |
| CNNNumber of training samples=25002023.11 | — | 93.89 | |
| CNNNumber of training samples=10002023.11 | — | 85.77 | |
| Constrainedinvariance constraint level=0.12023.06 | — | 97.92 | |
| DISCONumber of training samples=50002023.11 | — | 97.94 | |
| DISCONumber of training samples=25002023.11 | — | 96.65 | |
| DISCONumber of training samples=10002023.11 | — | 94.57 | |
| DSSNumber of training samples=50002023.11 | — | 96.54 | |
| DSSNumber of training samples=25002023.11 | — | 94.01 | |
| DSSNumber of training samples=10002023.11 | — | 92.81 | |
| Fourier CNNNumber of training samples=50002023.11 | — | 95.67 | |
| Fourier CNNNumber of training samples=25002023.11 | — | 94.19 | |
| Fourier CNNNumber of training samples=10002023.11 | — | 89.1 | |
| Per Res. CNNNumber of training samples=50002023.11 | — | 91.18 | |
| Per Res. CNNNumber of training samples=25002023.11 | — | 83.92 | |
| Per Res. CNNNumber of training samples=10002023.11 | — | 58.15 | |
| Resilientinvariance constraint level=0.12023.06 | — | 98.2 | |
| Scale-Equivariant Fourier NetworksNumber of training samples=50002023.11 | — | 98.35 | |
| Scale-Equivariant Fourier NetworksNumber of training samples=25002023.11 | — | 97.67 | |
| Scale-Equivariant Fourier NetworksNumber of training samples=10002023.11 | — | 96.06 | |
| SESNNumber of training samples=50002023.11 | — | 96.38 | |
| SESNNumber of training samples=25002023.11 | — | 94.02 | |
| SESNNumber of training samples=10002023.11 | — | 92.07 | |
| SI-CovNetNumber of training samples=50002023.11 | — | 96.41 | |
| SI-CovNetNumber of training samples=25002023.11 | — | 94.37 | |
| SI-CovNetNumber of training samples=10002023.11 | — | 92.8 | |
| SS-CNNNumber of training samples=50002023.11 | — | 94.77 | |
| SS-CNNNumber of training samples=25002023.11 | — | 92.59 | |
| SS-CNNNumber of training samples=10002023.11 | — | 91.76 | |
| Unconstrained2023.06 | — | 97.47 |