Image Classification on MNIST (test) (Adversarial Robustness & Certification)
69Clean ErrorCROWN-IBP
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CROWN-IBPTraining Robustness level (epsilon_C^T)=0.3, Attack magnitude (epsilon_A)=0.0, Source=Current Study2023.05 | 69 | 100 | 100 | |
| CROWN-IBPTraining Robustness level (epsilon_C^T)=0.1, Attack magnitude (epsilon_A)=0.0, Source=Current Study2023.05 | 81 | 89.02 | 100 | |
| CROWN-IBPTraining Robustness level (epsilon_C^T)=0.1, Attack magnitude (epsilon_A)=0.1, Source=Current Study2023.05 | 92 | 2 | 5.41 | |
| CROWN-IBPTraining Robustness level (epsilon_C^T)=0.3, Attack magnitude (epsilon_A)=0.1, Source=Current Study2023.05 | 94 | 98.41 | 100 | |
| CROWN-IBPTraining Robustness level (epsilon_C^T)=0.1, Attack magnitude (epsilon_A)=0.1, Source=Reported in [42]2023.05 | 117 | — | 2.24 | |
| CROWN-IBPTraining Robustness level (epsilon_C^T)=0.3, Attack magnitude (epsilon_A)=0.2, Source=Current Study2023.05 | 130 | 99.62 | 100 | |
| CROWN-IBPTraining Robustness level (epsilon_C^T)=0.1, Attack magnitude (epsilon_A)=0.2, Source=Current Study2023.05 | 132 | 2.14 | 2.6 | |
| CROWN-IBPTraining Robustness level (epsilon_C^T)=0.3, Attack magnitude (epsilon_A)=0.3, Source=Current Study2023.05 | 141 | 7.06 | 9.98 | |
| CROWN-IBPTraining Robustness level (epsilon_C^T)=0.1, Attack magnitude (epsilon_A)=0.3, Source=Current Study2023.05 | 144 | 2.33 | 2.63 | |
| CROWN-IBPTraining Robustness level (epsilon_C^T)=0.3, Attack magnitude (epsilon_A)=0.3, Source=Reported in [42]2023.05 | 182 | — | 7.02 | |
| CROWN-IBPTraining Robustness level (epsilon_C^T)=0.1, Attack magnitude (epsilon_A)=0.4, Source=Current Study2023.05 | 242 | 3.36 | 3.41 | |
| CROWN-IBPTraining Robustness level (epsilon_C^T)=0.3, Attack magnitude (epsilon_A)=0.4, Source=Current Study2023.05 | 258 | 7.03 | 7.68 |