Image Classification on MNIST (Performance and Efficiency Metrics)
97.99AccuracyEuclidean
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| EuclideanBackbone=MLP2026.03 | 97.99 | 97.98 | 0.5221 | 0.358 | 9 | |
| Cosine (Unst.)Backbone=MLP2026.03 | 97.8 | 97.78 | 0.5264 | 0.382 | 10 | |
| Mahalanobis (Chol.)Backbone=MLP2026.03 | 97.74 | 97.71 | 0.5611 | 0.092 | 50 | |
| Cosine (Stable)Backbone=MLP2026.03 | 97.66 | 97.64 | 0.5266 | 0.4033 | 9.3333 | |
| BaselineBackbone=MLP2026.03 | 97.6 | 97.58 | 0.55 | 0.5657 | 10.3333 | |
| Chebyshev (Std.)Backbone=MLP2026.03 | 97.56 | 97.54 | 0.5881 | 0.7865 | 5.6667 |