Random Subset Unlearning on CIFAR-10
100Retention Accuracy (RA)Pretrain
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| PretrainBackbone=ResNet-18, Unlearning ratio=10%, Task Type=Random subset unlearning2024.02 | 100 | 99.96 | 94.68 | 1.84 | 92.64 | 14,322 | |
| RetrainBackbone=ResNet-18, Unlearning ratio=10%, Task Type=Random subset unlearning2024.02 | 100 | 94.51 | 94.75 | 0 | 84.77 | 12,800 | |
| FTBackbone=ResNet-18, Unlearning ratio=10%, Task Type=Random subset unlearning2024.02 | 100 | 94.73 | 93.17 | 0.51 | 68.66 | 395 | |
| SCRUBBackbone=ResNet-18, Unlearning ratio=10%, Task Type=Random subset unlearning2024.02 | 100 | 95.79 | 93.43 | 0.87 | 77.56 | 207 | |
| SalUnBackbone=ResNet-18, Unlearning ratio=10%, Task Type=Random subset unlearning2024.02 | 100 | 94.31 | 93.19 | 0.59 | 26.52 | 247 | |
| MunBaBackbone=ResNet-18, Unlearning ratio=10%, Task Type=Random subset unlearning2024.02 | 100 | 94.57 | 93.19 | 0.54 | 58.2 | 420 | |
| SSDBackbone=ResNet-18, Unlearning ratio=10%, Task Type=Random subset unlearning2024.02 | 99.99 | 99.98 | 94.66 | 1.86 | 92.54 | 18 | |
| BTBackbone=ResNet-18, Unlearning ratio=10%, Task Type=Random subset unlearning2024.02 | 99.64 | 93.06 | 93.01 | 1.18 | 7.2 | 50 | |
| SCARBackbone=ResNet-18, Unlearning ratio=10%, Task Type=Random subset unlearning2024.02 | 98.63 | 98.64 | 92.43 | 2.61 | 48.36 | 197 | |
| DUCKBackbone=ResNet-18, Unlearning ratio=10%, Task Type=Random subset unlearning2024.02 | 98.04 | 97.9 | 92.38 | 2.57 | 90.59 | 21 | |
| MU-MisBackbone=ResNet-18, Unlearning ratio=10%, Task Type=Random subset unlearning2024.02 | 97.76 | 97.43 | 91.5 | 2.8 | 33.04 | 116 | |
| LoTusBackbone=ResNet-18, Unlearning ratio=10%, Task Type=Random subset unlearning2024.02 | 96.9 | 96.73 | 90.55 | 3.17 | 56.1 | 172 | |
| NGBackbone=ResNet-18, Unlearning ratio=10%, Task Type=Random subset unlearning2024.02 | 96.46 | 96.15 | 90.52 | 3.13 | 88.21 | 25 | |
| RLBackbone=ResNet-18, Unlearning ratio=10%, Task Type=Random subset unlearning2024.02 | 96.17 | 93.99 | 88.29 | 4.27 | 83.11 | 26 | |
| JiTBackbone=ResNet-18, Unlearning ratio=10%, Task Type=Random subset unlearning2024.02 | 95.45 | 95.46 | 89.63 | 3.54 | 86 | 255 |