Image Classification on CIFAR-10 (test) (Machine Unlearning Evaluation)
100Retain AccuracyOriginal
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| OriginalBackbone=ResNet-182026.01 | 100 | 0 | 94.76 | 0.07 | 8.34 | — | |
| Re-trainBackbone=ResNet-182026.01 | 100 | 9.47 | 93.3 | 22.5 | 0 | 17.33 | |
| CF-kBackbone=ResNet-182026.01 | 100 | 0 | 94.38 | 0.42 | 8.16 | 1 | |
| GDBackbone=ResNet-182026.01 | 99.98 | 0 | 94.29 | 0.1 | 8.22 | 0.2 | |
| SCRUBBackbone=ResNet-182026.01 | 99.61 | 12.45 | 92.7 | 7.1 | 4.84 | 30 | |
| RURKBackbone=ResNet-182026.01 | 99.55 | 14.63 | 92.6 | 18.2 | 2.65 | 30 | |
| EU-kBackbone=ResNet-182026.01 | 99.34 | 1.12 | 92.97 | 0.87 | 7.74 | 4.33 | |
| NegGrad+Backbone=ResNet-182026.01 | 99.28 | 14 | 92.02 | 18.18 | 2.71 | 1 | |
| NGDBackbone=ResNet-182026.01 | 97.53 | 10.67 | 90.7 | 3.7 | 6.27 | 20.67 | |
| SSDBackbone=ResNet-182026.01 | 96.49 | 66.45 | 88.59 | 5.12 | 20.65 | 7.33 | |
| GABackbone=ResNet-182026.01 | 95.41 | 61.37 | 85.98 | 0 | 21.25 | 1 |