Backdoor Defense on CIFAR-10 10% clean data (train)
93.73AccuracyBefore
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BeforeBackbone=ResNet-182021.10 | 93.73 | 99.97 | |
| ANPBackbone=ResNet-182021.10 | 93.01 | 3.34 | |
| MCRBackbone=ResNet-18, Hyperparameter=0.32021.10 | 89.17 | 5.7 | |
| Fine-tuning (FT)Backbone=ResNet-18, Hyperparameter=0.012021.10 | 87.74 | 2.6 | |
| Fine-pruning (FP)Backbone=ResNet-182021.10 | 86.95 | 1.83 | |
| Fine-tuning (FT)Backbone=ResNet-18, Hyperparameter=0.022021.10 | 85.56 | 2.13 |