Image Classification on MNIST standard (test) (Class Disparity Metrics)
95.16Macro AccuracyBounded Adaptive
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Bounded Adaptiveepsilon=4.095*, DP-HPO=true2025.06 | 95.16 | 80.39 | |
| Constant Clippingepsilon=4.095*, DP-HPO=true2025.06 | 95.15 | 77.54 | |
| Unbounded Adaptiveepsilon=4.095*, DP-HPO=true2025.06 | 94.56 | 73.77 | |
| Constant Clippingepsilon=3.509, DP-HPO=true2025.06 | 94.35 | 69.9 | |
| Constant Clippingepsilon=2.916, DP-HPO=true2025.06 | 94.15 | 69.5 | |
| AUTOepsilon=4.095*, DP-HPO=true2025.06 | 93.95 | 70.27 | |
| Bounded Adaptiveepsilon=3.509, DP-HPO=true2025.06 | 93.88 | 73.88 | |
| Unbounded Adaptiveepsilon=3.509, DP-HPO=true2025.06 | 93.87 | 70.3 | |
| AUTOepsilon=3.509, DP-HPO=true2025.06 | 93.84 | 68.27 | |
| Bounded Adaptiveepsilon=2.916, DP-HPO=true2025.06 | 93.66 | 73.2 | |
| AUTOepsilon=2.916, DP-HPO=true2025.06 | 92.86 | 65.27 | |
| Unbounded Adaptiveepsilon=2.916, DP-HPO=true2025.06 | 92.72 | 66.14 | |
| Unbounded Adaptiveepsilon=2.311, DP-HPO=true2025.06 | 92.56 | 62.95 | |
| Constant Clippingepsilon=2.311, DP-HPO=true2025.06 | 92.23 | 54.5 | |
| Bounded Adaptiveepsilon=2.311, DP-HPO=true2025.06 | 91.77 | 60.23 | |
| AUTOepsilon=2.311, DP-HPO=true2025.06 | 91.74 | 60.75 | |
| AUTOepsilon=1.743, DP-HPO=true2025.06 | 89.43 | 52.62 | |
| Bounded Adaptiveepsilon=1.743, DP-HPO=true2025.06 | 83.2 | 40.23 | |
| Unbounded Adaptiveepsilon=1.743, DP-HPO=true2025.06 | 81.97 | 42.87 | |
| Constant Clippingepsilon=1.743, DP-HPO=true2025.06 | 72.9 | 20.11 |