Multi-label Image Classification on CelebA (test)
90.3Accuracynon-DP
Evaluation results
| Method | Links | |
|---|---|---|
| non-DPModel=ResNet9, epsilon=inf2022.06 | 90.3 | |
| AUTO-S clippingModel=ResNet9, epsilon=8, delta=5e-62022.06 | 87.58 | |
| Abadi's clippingModel=ResNet9, epsilon=8, delta=5e-62022.06 | 87.52 | |
| AUTO-S clippingModel=ResNet9, epsilon=3, delta=5e-62022.06 | 87.05 | |
| Abadi's clippingModel=ResNet9, epsilon=3, delta=5e-62022.06 | 86.81 |