Machine Unlearning on TinyImageNet (50% Random Removal)
99.99Retain AccuracyRetrain
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| RetrainModel=ResNet-182025.09 | 99.99 | 43.01 | 57.28 | 71.22 | 0 | 121.93 | |
| FTModel=ResNet-182025.09 | 99.99 | 36.78 | 60.59 | 66.28 | 3.62 | 33.5 | |
| NegGrad+Model=ResNet-182025.09 | 99.99 | 47.62 | 58.85 | 66.43 | 2.74 | 33.5 | |
| ℓ1-sparseModel=ResNet-182025.09 | 99.99 | 38.83 | 60.25 | 65.82 | 3.14 | 33.5 | |
| NoTModel=ResNet-182025.09 | 99.99 | 40.94 | 58.27 | 66.23 | 2.01 | 33.5 | |
| CoUnModel=ResNet-182025.09 | 99.98 | 43.21 | 55.75 | 66.59 | 1.59 | 44.66 | |
| SalUnModel=ResNet-182025.09 | 93.59 | 44.74 | 45.53 | 70.41 | 5.17 | 25.01 |