Selective Unlearning on CIFAR-10
16.29Test ErrorCF-k
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| CF-kBackbone=All-CNN, Application=Removing Biases (RB)2023.02 | 16.29 | 0 | 0 | — | — | |
| OriginalBackbone=All-CNN, Application=Removing Biases (RB)2023.02 | 16.43 | 0 | 0 | — | — | |
| RetrainBackbone=All-CNN, Application=Removing Biases (RB)2023.02 | 16.47 | 25.67 | 0 | — | — | |
| FinetuneBackbone=All-CNN, Application=Removing Biases (RB)2023.02 | 16.5 | 0 | 0 | — | — | |
| SCRUBBackbone=All-CNN, Application=Removing Biases (RB)2023.02 | 16.55 | 29.33 | 0 | — | — | |
| RetrainArchitecture=All-CNN2023.02 | 16.71 | 26.67 | 0 | 51.33 | — | |
| OriginalArchitecture=All-CNN2023.02 | 16.71 | 0 | 0 | 68.67 | — | |
| CF-kArchitecture=All-CNN2023.02 | 16.82 | 0 | 0 | 69.67 | — | |
| FinetuneArchitecture=All-CNN2023.02 | 16.86 | 0 | 0 | 69.33 | — | |
| SCRUB+RArchitecture=All-CNN2023.02 | 16.88 | 26.33 | 0 | 49.33 | — | |
| SCRUBArchitecture=All-CNN2023.02 | 17.01 | 33 | 0 | 51 | — | |
| OriginalBackbone=ResNet, Application=Removing Biases (RB)2023.02 | 17.36 | 0 | 0 | — | — | |
| FinetuneBackbone=ResNet, Application=Removing Biases (RB)2023.02 | 17.37 | 0 | 0 | — | — | |
| RetrainBackbone=ResNet, Application=Removing Biases (RB)2023.02 | 17.4 | 29.67 | 0 | — | — | |
| CF-kBackbone=ResNet, Application=Removing Biases (RB)2023.02 | 17.4 | 0 | 0 | — | — | |
| EU-kBackbone=All-CNN, Application=Removing Biases (RB)2023.02 | 17.62 | 0.33 | 0.11 | — | — | |
| SCRUBBackbone=ResNet, Application=Removing Biases (RB)2023.02 | 18.04 | 70.33 | 0 | — | — | |
| EU-kArchitecture=All-CNN2023.02 | 18.44 | 0.33 | 0.32 | 66 | — | |
| FisherBackbone=ResNet, Application=Removing Biases (RB)2023.02 | 21.23 | 3 | 2.88 | — | — | |
| NegGrad+Backbone=All-CNN, Application=Removing Biases (RB)2023.02 | 21.36 | 45.33 | 3.23 | — | — | |
| FisherBackbone=All-CNN, Application=Removing Biases (RB)2023.02 | 21.39 | 13 | 4 | — | — | |
| NegGrad+Architecture=All-CNN2023.02 | 21.65 | 47 | 4.54 | 73 | — | |
| EU-kBackbone=ResNet, Application=Removing Biases (RB)2023.02 | 21.8 | 23.7 | 0.4 | — | — | |
| Bad-TArchitecture=All-CNN2023.02 | 22.43 | 1.67 | 10.13 | 77.67 | — | |
| Bad-TBackbone=All-CNN, Application=Removing Biases (RB)2023.02 | 22.43 | 1.67 | 10.13 | — | — | |
| NegGrad+Backbone=ResNet, Application=Removing Biases (RB)2023.02 | 22.7 | 53.7 | 4.1 | — | — | |
| Bad-TBackbone=ResNet, Application=Removing Biases (RB)2023.02 | 23.47 | 34.67 | 14.53 | — | — | |
| Bad-TBackbone=ResNet2023.02 | — | — | — | 59.67 | 4.19 | |
| Bad-TBackbone=All-CNN2023.02 | — | — | — | 77.67 | 4.11 | |
| CF-kBackbone=ResNet2023.02 | — | — | — | 65 | 0 | |
| CF-kBackbone=All-CNN2023.02 | — | — | — | 72.93 | 1.06 | |
| EU-kBackbone=ResNet2023.02 | — | — | — | 53 | 3.27 | |
| EU-kBackbone=All-CNN2023.02 | — | — | — | 51.6 | 1.22 | |
| FinetuneBackbone=ResNet2023.02 | — | — | — | 64 | 0.82 | |
| FinetuneBackbone=All-CNN2023.02 | — | — | — | 74.97 | 1.27 | |
| NegGrad+Backbone=ResNet2023.02 | — | — | — | 66.67 | 1.7 | |
| NegGrad+Backbone=All-CNN2023.02 | — | — | — | 70.03 | 1.92 | |
| OriginalBackbone=ResNet2023.02 | — | — | — | 65.33 | 0.47 | |
| OriginalBackbone=All-CNN2023.02 | — | — | — | 71.4 | 0.7 | |
| RetrainBackbone=ResNet2023.02 | — | — | — | 54 | 1.63 | |
| RetrainBackbone=All-CNN2023.02 | — | — | — | 48.73 | 0.24 | |
| SCRUBBackbone=ResNet2023.02 | — | — | — | 78 | 2.45 | |
| SCRUBBackbone=All-CNN2023.02 | — | — | — | 54.3 | 2.24 | |
| SCRUB+RBackbone=ResNet2023.02 | — | — | — | 58.67 | 1.89 | |
| SCRUB+RBackbone=All-CNN2023.02 | — | — | — | 54.3 | 2.24 |