Image Classification on MNIST binarized (test)
99.43Last-10 AccuracyCNN 3×3
Evaluation results
| Method | Links | |
|---|---|---|
| CNN 3×3Config=L=6, Train params=1.60M, Deploy params=1.60M float, Inference=∼30M MACs2026.05 | 99.43 | |
| CNN 2×2Config=L=6, Train params=712k, Deploy params=712k float, Inference=∼10M MACs2026.05 | 98.9 | |
| CNN 3×3Config=L=3, Train params=94k, Deploy params=94k float, Inference=∼2M MACs2026.05 | 98.86 | |
| M-CovJacConfig=L=6, Train params=1.54M, Deploy params=0 float, Inference=0 MACs2026.05 | 98.3 | |
| MLPConfig=L=6, h=128, Train params=184k, Deploy params=184k float, Inference=184k MACs2026.05 | 97.63 | |
| MLPConfig=L=3, h=64, Train params=59k, Deploy params=59k float, Inference=59k MACs2026.05 | 96.95 |