Image Classification on SVHN (test) (Clean, PGD, AA, C&W, DF, Speed up)
93.37Clean AccuracyAT
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| ATSparsity rate (sr)=0%, Backbone=ResNet18, Params=11.2M2025.04 | 93.37 | 56.27 | 50.14 | 58.85 | 59.8 | 1 | |
| HARPSparsity rate (sr)=95%, Backbone=ResNet18, Params=0.6M2025.04 | 92.75 | 55.21 | 45.95 | 51.3 | 47.4 | 3.19 | |
| EEDSparsity rate (sr)=95%, Backbone=ResNet18, Params=0.6M2025.04 | 90.76 | 52.74 | 47.19 | 47.23 | 46.99 | 3.3 | |
| HYDRASparsity rate (sr)=90%, Backbone=ResNet18, Params=1.1M2025.04 | 88.71 | 52.49 | 44.12 | 49.61 | 45.18 | 2.43 | |
| TwinRepSparsity rate (sr)=95%, Backbone=ResNet18, Params=0.6M2025.04 | 88.59 | 50.62 | 47.16 | 52.31 | 47.35 | 3.37 | |
| R-ADMMSparsity rate (sr)=90%, Backbone=ResNet18, Params=1.1M2025.04 | 84.34 | 51.91 | 38.4 | 49.46 | 42.87 | 2.23 |