Image Classification on CIFAR-100 (Noise Scale Robustness Evaluation)
787,542Noise ScaleFull DP
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Full DPepsilon=0.2, Model Architecture=ResNet-182026.01 | 787,542 | 61.54 | 68.06 | |
| LDPepsilon=0.2, Model Architecture=ResNet-182026.01 | 762,930 | 60.3 | 65.79 | |
| Full DPepsilon=0.3, Model Architecture=ResNet-182026.01 | 543,256 | 69.1 | — | |
| LDPepsilon=0.3, Model Architecture=ResNet-182026.01 | 538,719 | 68.84 | — | |
| Time-Varying DPepsilon=0.2, Model Architecture=ResNet-182026.01 | 532,786 | 43.16 | 40.49 | |
| Full DPepsilon=0.4, Model Architecture=ResNet-182026.01 | 436,562 | 76.81 | — | |
| LDPepsilon=0.4, Model Architecture=ResNet-182026.01 | 399,175 | 74.63 | — | |
| DPAepsilon=0.2, Model Architecture=ResNet-182026.01 | 379,128 | 20.12 | 37.66 | |
| Time-Varying DPepsilon=0.3, Model Architecture=ResNet-182026.01 | 324,836 | 48.32 | — | |
| LaDP-FLepsilon=0.2, Model Architecture=ResNet-182026.01 | 302,859 | — | — | |
| Sensitive DPepsilon=0.2, Model Architecture=ResNet-182026.01 | 281,766 | -7.49 | -12.4 | |
| DPAepsilon=0.3, Model Architecture=ResNet-182026.01 | 274,038 | 38.74 | — | |
| Full DPepsilon=0.5, Model Architecture=ResNet-182026.01 | 267,569 | 64.77 | — | |
| DPAepsilon=0.4, Model Architecture=ResNet-182026.01 | 239,767 | 57.77 | — | |
| LDPepsilon=0.5, Model Architecture=ResNet-182026.01 | 232,067 | 59.39 | — | |
| Time-Varying DPepsilon=0.4, Model Architecture=ResNet-182026.01 | 192,879 | 47.5 | — | |
| LaDP-FLepsilon=0.3, Model Architecture=ResNet-182026.01 | 167,884 | — | — | |
| DPAepsilon=0.5, Model Architecture=ResNet-182026.01 | 142,834 | 34.01 | — | |
| Sensitive DPepsilon=0.3, Model Architecture=ResNet-182026.01 | 127,458 | -31.72 | — | |
| Time-Varying DPepsilon=0.5, Model Architecture=ResNet-182026.01 | 122,365 | 22.97 | — | |
| LaDP-FLepsilon=0.4, Model Architecture=ResNet-182026.01 | 101,256 | — | — | |
| Sensitive DPepsilon=0.4, Model Architecture=ResNet-182026.01 | 97,145 | -4.23 | — | |
| LaDP-FLepsilon=0.5, Model Architecture=ResNet-182026.01 | 94,253 | — | — | |
| Sensitive DPepsilon=0.5, Model Architecture=ResNet-182026.01 | 88,766 | -6.18 | — | |
| Full DPepsilon=0.2, Model Architecture=CNN2026.01 | 77,523 | 70.37 | 69.68 | |
| LDPepsilon=0.2, Model Architecture=CNN2026.01 | 72,781 | 68.44 | 66.85 | |
| LDPepsilon=0.3, Model Architecture=CNN2026.01 | 56,700 | 65.97 | — | |
| Full DPepsilon=0.3, Model Architecture=CNN2026.01 | 52,652 | 63.35 | — | |
| Time-Varying DPepsilon=0.2, Model Architecture=CNN2026.01 | 50,803 | 54.79 | 53.11 | |
| LDPepsilon=0.4, Model Architecture=CNN2026.01 | 44,469 | 79.7 | — | |
| Full DPepsilon=0.4, Model Architecture=CNN2026.01 | 42,236 | 78.62 | — | |
| DPAepsilon=0.3, Model Architecture=CNN2026.01 | 37,329 | 48.31 | — | |
| DPAepsilon=0.2, Model Architecture=CNN2026.01 | 35,603 | 35.49 | 41.35 | |
| Time-Varying DPepsilon=0.3, Model Architecture=CNN2026.01 | 34,114 | 43.43 | — | |
| DPAepsilon=0.4, Model Architecture=CNN2026.01 | 27,668 | 67.37 | — | |
| Full DPepsilon=0.5, Model Architecture=CNN2026.01 | 25,786 | 66.35 | — | |
| Sensitive DPepsilon=0.2, Model Architecture=CNN2026.01 | 23,854 | 3.71 | 5.33 | |
| Time-Varying DPepsilon=0.4, Model Architecture=CNN2026.01 | 23,578 | 61.71 | — | |
| LaDP-FLepsilon=0.2, Model Architecture=CNN2026.01 | 22,968 | — | — | |
| Sensitive DPepsilon=0.3, Model Architecture=CNN2026.01 | 21,271 | 9.28 | — | |
| LaDP-FLepsilon=0.3, Model Architecture=CNN2026.01 | 19,297 | — | — | |
| LDPepsilon=0.5, Model Architecture=CNN2026.01 | 18,574 | 53.28 | — | |
| Time-Varying DPepsilon=0.5, Model Architecture=CNN2026.01 | 18,270 | 52.51 | — | |
| DPAepsilon=0.5, Model Architecture=CNN2026.01 | 10,119 | 14.25 | — | |
| Sensitive DPepsilon=0.4, Model Architecture=CNN2026.01 | 9,548 | 5.45 | — | |
| LaDP-FLepsilon=0.4, Model Architecture=CNN2026.01 | 9,028 | — | — | |
| Sensitive DPepsilon=0.5, Model Architecture=CNN2026.01 | 8,933 | 2.87 | — | |
| LaDP-FLepsilon=0.5, Model Architecture=CNN2026.01 | 8,677 | — | — |