Machine Unlearning on ImageNet-100 (test)
25.77UASFRon
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SFRonBackbone=ViT, Forgetting Ratio=20%2025.05 | 25.77 | 78.5 | 77.86 | 59.4 | 20.64 | 87.65 | |
| GABackbone=ViT, Forgetting Ratio=20%2025.05 | 12.42 | 87.68 | 84.99 | 12.12 | 2.58 | 59.57 | |
| NegGrad+Backbone=ViT, Forgetting Ratio=20%2025.05 | 12.12 | 91.67 | 86.86 | 15.63 | 0.74 | 85.96 | |
| RTBackbone=ViT, Forgetting Ratio=20%2025.05 | 11.89 | 92.08 | 88.04 | 14.47 | 0 | 837.96 | |
| MCUBackbone=ViT, Forgetting Ratio=20%2025.05 | 11.78 | 91.06 | 87.22 | 14.89 | 0.59 | 150.57 | |
| MCUβBackbone=ViT, Forgetting Ratio=20%2025.05 | 10.98 | 92.06 | 87.4 | 14.58 | 0.42 | 149.88 | |
| NegTVBackbone=ViT, Forgetting Ratio=20%2025.05 | 10.06 | 91.47 | 87.11 | 13.07 | 1.19 | 22.29 | |
| RLBackbone=ViT, Forgetting Ratio=20%2025.05 | 9.54 | 91.85 | 87.83 | 29.43 | 4.44 | 245.6 | |
| FTBackbone=ViT, Forgetting Ratio=20%2025.05 | 8.87 | 92.32 | 87.75 | 10.53 | 1.87 | 84.88 | |
| SalUnBackbone=ViT, Forgetting Ratio=20%2025.05 | 8.85 | 91.67 | 87.75 | 22.65 | 2.98 | 225.75 |