Machine Unlearning on CIFAR-10 (10% random removal)
5.48AccuracyCU
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CUBackbone=ResNet-18, Comp. Cost (PFLOPs)=3.362025.09 | 5.48 | 67 | 0.61 | |
| ℓ1-sparseBackbone=ResNet-18, Comp. Cost (PFLOPs)=6.922025.09 | 5.4 | 59 | 0.38 | |
| NegGrad+Backbone=ResNet-18, Comp. Cost (PFLOPs)=6.022025.09 | 4.82 | 1 | 0.58 | |
| RetrainBackbone=ResNet-18, Comp. Cost (PFLOPs)=27.372025.09 | 4.81 | 0 | 0 | |
| SalUnBackbone=ResNet-18, Comp. Cost (PFLOPs)=8.662025.09 | 4.31 | 50 | 0.59 | |
| NoTBackbone=ResNet-18, Comp. Cost (PFLOPs)=7.522025.09 | 4.19 | 62 | 0.3 | |
| CoUnBackbone=ResNet-18, Comp. Cost (PFLOPs)=8.022025.09 | 4.12 | 69 | 0.25 | |
| SCRUBBackbone=ResNet-18, Comp. Cost (PFLOPs)=8.932025.09 | 3.93 | 88 | 0.62 | |
| FTBackbone=ResNet-18, Comp. Cost (PFLOPs)=6.322025.09 | 3.76 | 105 | 0.65 | |
| BadTBackbone=ResNet-18, Comp. Cost (PFLOPs)=9.892025.09 | 0.07 | 474 | 1.58 | |
| SSDBackbone=ResNet-18, Comp. Cost (PFLOPs)=0.062025.09 | 0.02 | 479 | 3.83 |