Image Classification on Small CIFAR-5 (test)
99.96Retention Accuracy (%)Re-train
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Re-trainBackbone=ResNet-182026.01 | 99.96 | 8.33 | 94.8 | 27 | 0 | 3.33 | |
| OriginalBackbone=ResNet-182026.01 | 99.93 | 0 | 95.37 | 4.67 | 7.82 | — | |
| NTKBackbone=ResNet-182026.01 | 99.93 | 7 | 95.37 | 16 | 3.23 | 4.67 | |
| CRBackbone=ResNet-182026.01 | 99.56 | 14 | 91.8 | 58.17 | 10.06 | — | |
| RURKBackbone=ResNet-182026.01 | 99.52 | 5.67 | 93.83 | 33.33 | 2.6 | 2 | |
| FisherBackbone=ResNet-182026.01 | 92.67 | 12.67 | 88.8 | 47.33 | 9.49 | 3 |