Machine Unlearning on CIFAR-10 50% random data removal
100Retain AccuracyRetrain
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| RetrainModel=ResNet-182025.09 | 100 | 7.29 | 92.28 | 17.33 | 0 | 15.24 | |
| NegGrad+Model=ResNet-182025.09 | 100 | 4.06 | 93.81 | 9.05 | 0.9 | 5.58 | |
| NoTModel=ResNet-182025.09 | 99.98 | 5.95 | 92.84 | 13.9 | 1.34 | 3.35 | |
| CoUnModel=ResNet-182025.09 | 99.97 | 6.19 | 92.36 | 16.94 | 0.4 | 3.35 | |
| ℓ1-sparseModel=ResNet-182025.09 | 99.77 | 9.02 | 90.66 | 16.05 | 1.22 | 4.19 | |
| FTModel=ResNet-182025.09 | 99.38 | 6.32 | 91.91 | 12.64 | 1.66 | 2.51 | |
| SalUnModel=ResNet-182025.09 | 98.7 | 3.74 | 92.37 | 16.4 | 1.47 | 7.82 |