Image Classification on CIFAR-10 (Adversarial Robustness Analysis - PGD/AutoAttack)
93.27Clean AccuracyStandard Training
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Standard Training2021.05 | 93.27 | 0.62 | 0.69 | 0.01 | 0 | 0 | |
| Dense Orthogonal Classifierα=12021.05 | 92.62 | 89.83 | 84.35 | 89.61 | 0.92 | 0 | |
| FGSM-AT2021.05 | 88.23 | 2.08 | 29.54 | 0.01 | 0.01 | 0 | |
| TRADES1/λ=12021.05 | 82.22 | 58.17 | 18.26 | 54.68 | 51.88 | 51.26 | |
| ALP2021.05 | 80.7 | 55.68 | 22.16 | 56.22 | 54.26 | 53.34 | |
| Dense Orthogonal Classifierα=0.152021.05 | 80.43 | 60.41 | 30.64 | 59.88 | 54.59 | 53.48 | |
| PGD-AT2021.05 | 79.42 | 57.25 | 23.08 | 56.7 | 54.67 | 53.31 | |
| TRADES1/λ=62021.05 | 74.04 | 54.95 | 25.24 | 54.29 | 51.32 | 51.1 |