Image Classification on CIFAR-10 (test) (Epsilon Metrics Reporting)
0.94AccuracyALIBI
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ALIBIAccuracy level=High, Backbone=Wide-ResNet182021.06 | 0.94 | 8 | — | |
| PATE-FMAccuracy level=High, Backbone=Wide-ResNet18, δ=10^-52021.06 | 0.937 | 1.6 | — | |
| PATE-FMAccuracy level=Medium, Backbone=Wide-ResNet18, δ=10^-52021.06 | 0.869 | 0.29 | — | |
| ALIBIAccuracy level=Medium, Backbone=Wide-ResNet182021.06 | 0.842 | 2.1 | — | |
| PATE-FMAccuracy level=Low, Backbone=Wide-ResNet18, δ=10^-52021.06 | 0.734 | 0.18 | — | |
| ALIBIAccuracy level=Low, Backbone=Wide-ResNet182021.06 | 0.71 | 1 | — |