Machine Unlearning on SVHN 100 In Class Random Forgetting
99.999RARetrained
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| RetrainedBackbone=VGG16-BN2026.04 | 99.999 | 95.09 | 0.2 | 0.5 | — | — | — | |
| FTBackbone=ResNet-182026.04 | 99.99 | 4.33 | 0.17 | 0.04 | 0.27 | 0 | 0.43 | |
| Retrained (Baseline)Backbone=ResNet-182026.04 | 99.99 | 95.12 | 0.2 | 0.5 | — | — | — | |
| FTBackbone=VGG16-BN2026.04 | 99.99 | 4.47 | 0.22 | 0.03 | 0.27 | 0 | 0.43 | |
| SalUnBackbone=VGG16-BN2026.04 | 99.94 | 5.04 | 0.16 | 0.03 | 0.14 | 0.03 | 0.5 | |
| SalUnBackbone=ResNet-182026.04 | 99.9 | 3.14 | 0.13 | 0.04 | 0.17 | 0.03 | 0.5 | |
| SCRUBBackbone=ResNet-182026.04 | 99.79 | 4.44 | 0.18 | 0.05 | 0.03 | 0.03 | 0.5 | |
| SCRUBBackbone=VGG16-BN2026.04 | 99.79 | 9.55 | 0.36 | 0.03 | 0.02 | 0.03 | 0.5 | |
| CF-kBackbone=ResNet-182026.04 | 99.76 | 4.47 | 0.18 | 0.05 | 0.23 | 0.03 | 0.5 | |
| CF-kBackbone=VGG16-BN2026.04 | 99.76 | 4.53 | 0.23 | 0.04 | 0.24 | 0.03 | 0.5 | |
| RELOADBackbone=ResNet-182026.04 | 99.68 | 0.25 | 0.01 | 0 | 0.12 | 0.06 | 0.21 | |
| RELOADBackbone=VGG16-BN2026.04 | 99.67 | 0.93 | 0.05 | 0.01 | 0.14 | 0.06 | 0.21 | |
| EU-kBackbone=ResNet-182026.04 | 99.63 | 4.46 | 0.18 | 0.05 | 0.23 | 0.05 | 0.47 | |
| EU-kBackbone=VGG16-BN2026.04 | 99.63 | 4.54 | 0.23 | 0.04 | 0.24 | 0.05 | 0.47 | |
| GABackbone=ResNet-182026.04 | 99.57 | 4.37 | 0.17 | 0.05 | 0 | 0.05 | 0.52 | |
| FisherBackbone=ResNet-182026.04 | 99.57 | 0.09 | 0 | 0.01 | 2.17 | 0.05 | 0.47 | |
| GABackbone=VGG16-BN2026.04 | 99.57 | 4.46 | 0.22 | 0.03 | 0 | 0.05 | 0.51 | |
| FisherBackbone=VGG16-BN2026.04 | 99.48 | 0.09 | 0 | 0 | 1.42 | 0.05 | 0.47 | |
| SSDBackbone=VGG16-BN2026.04 | 14.55 | 84.19 | 2.05 | 0 | 0.01 | 8.51 | 7.84 | |
| SSDBackbone=ResNet-182026.04 | 12.75 | 82.52 | 2.12 | 0.01 | 0.01 | 8.55 | 7.88 |