Image Classification on CIFAR-10 (test) (Adversarial Robustness: PGD/DeepFool/AutoAttack)
93.27Accuracy (Clean)Standard Training
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Standard Training2021.05 | 93.27 | 0 | 0.02 | 0 | 0 | 0 | 0 | |
| Dense Orthogonal Classifierα=12021.05 | 92.62 | 78.68 | 32.22 | 0 | 61.36 | 12 | 0 | |
| FGSM-AT2021.05 | 88.23 | 0.01 | 0.17 | 0 | 0 | 0 | 0 | |
| TRADES1/λ=12021.05 | 82.22 | 39.93 | 41.83 | 34.62 | 10.7 | 17.29 | 6.15 | |
| ALP2021.05 | 80.7 | 48.27 | 49.54 | 43.23 | 19.7 | 27.22 | 12.27 | |
| Dense Orthogonal Classifierα=0.152021.05 | 80.43 | 48.83 | 51.31 | 44.15 | 27.5 | 32.24 | 12.31 | |
| PGD-AT2021.05 | 79.42 | 48.44 | 48.7 | 42.99 | 19.28 | 25.5 | 10.92 | |
| TRADES1/λ=62021.05 | 74.04 | 45.9 | 44.73 | 40.78 | 21.06 | 23.8 | 14.49 |