Machine Unlearning on CIFAR-10 1% random data forgetting
39.2Utility Retention (UA)SFRon
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SFRonBackbone=PreResNet-110, RTE=16.112025.05 | 39.2 | 62.41 | 62.88 | 32.1 | 26.86 | |
| NegGrad+Backbone=PreResNet-110, RTE=3.062025.05 | 20.47 | 98.37 | 89.68 | 24.8 | 4.67 | |
| RLBackbone=PreResNet-110, RTE=8.982025.05 | 17.2 | 99.91 | 90.45 | 46.6 | 8.72 | |
| MCUβBackbone=PreResNet-110, RTE=6.432025.05 | 11.02 | 99.97 | 91.02 | 19 | 0.3 | |
| RTBackbone=PreResNet-110, RTE=78.802025.05 | 11 | 99.98 | 90.66 | 18.2 | 0 | |
| MCUBackbone=PreResNet-110, RTE=6.432025.05 | 10.63 | 99.98 | 89.82 | 17.24 | 0.54 | |
| SalUnBackbone=PreResNet-110, RTE=6.172025.05 | 7.53 | 99.05 | 91.03 | 17.47 | 1.38 | |
| GABackbone=PreResNet-110, RTE=0.082025.05 | 1.4 | 99.18 | 90.12 | 3.53 | 6.4 | |
| FTBackbone=PreResNet-110, RTE=3.992025.05 | 0.47 | 99.93 | 90.96 | 3.2 | 6.47 | |
| NegTVBackbone=PreResNet-110, RTE=0.172025.05 | 0.13 | 99.98 | 90.97 | 1.26 | 7.03 |