Image Classification on CIFAR-10 (val) (Adversarial Robustness)
95.01AccuracyNatural
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| NaturalTraining=Standard Natural Training2019.04 | 95.01 | 780 | 0 | 0 | 0 | 0 | |
| Freem=22019.04 | 91.45 | 816 | 33.92 | 33.2 | 34.57 | 33.41 | |
| Freem=42019.04 | 87.83 | 800 | 41.15 | 40.35 | 41.96 | 40.73 | |
| 7-PGD trainedBackbone=Wide-Resnet 32-102019.04 | 87.25 | 5,418 | 45.84 | 45.29 | 46.52 | 45.53 | |
| Freem=82019.04 | 85.96 | 785 | 46.82 | 46.19 | 46.6 | 46.33 | |
| Freem=102019.04 | 83.94 | 785 | 46.31 | 45.79 | 45.86 | 45.94 |