Sub-class machine unlearning on CIFAR-20 Sea sub-class
85.09Retained Accuracy (RA)Pretrain
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| PretrainBackbone=ResNet-182024.02 | 85.09 | 97.66 | 85.21 | 5.94 | 91.81 | 6,910 | |
| SCRUBBackbone=ResNet-182024.02 | 84.86 | 88.17 | 84.86 | 2.58 | 57.07 | 113 | |
| SSDBackbone=ResNet-182024.02 | 84.79 | 78 | 84.61 | 1.24 | 8 | 7 | |
| RetrainBackbone=ResNet-182024.02 | 84.6 | 80.93 | 84.61 | 0 | 51.61 | 4,298 | |
| MU-MisBackbone=ResNet-182024.02 | 84.35 | 81 | 84.33 | 0.2 | 1.25 | 10 | |
| SalUnBackbone=ResNet-182024.02 | 82.85 | 81 | 83.1 | 1.11 | 13.4 | 63 | |
| BTBackbone=ResNet-182024.02 | 82.51 | 81 | 82.63 | 1.38 | 15 | 47 | |
| FTBackbone=ResNet-182024.02 | 82.36 | 88 | 82.43 | 3.83 | 58.08 | 417 | |
| DUCKBackbone=ResNet-182024.02 | 80.95 | 66.45 | 80.77 | 7.34 | 54.92 | 68 | |
| NGBackbone=ResNet-182024.02 | 80.95 | 75 | 80.84 | 4.45 | 60 | 3 | |
| MUNBaBackbone=ResNet-182024.02 | 80.64 | 84 | 80.66 | 3.66 | 60 | 564 | |
| RLBackbone=ResNet-182024.02 | 80.48 | 77 | 80.34 | 4.11 | 48.7 | 3 | |
| SCARBackbone=ResNet-182024.02 | 76.3 | 77.4 | 76.12 | 6.77 | 51.84 | 434 | |
| LoTusBackbone=ResNet-182024.02 | 73.12 | 81 | 73.39 | 7.59 | 61.2 | 16 | |
| JiTBackbone=ResNet-182024.02 | 51.48 | 7.2 | 51.04 | 46.81 | 32.2 | 4 |