Classification on MNIST-10 (test)
87.66AccuracyDeep Ensemble
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Deep Ensemble2026.06 | 87.66 | 0.2028 | 0.0976 | 0.4582 | |
| Laplace FullParameter Setting=Full-parameter, Curvature Approximation=PSD-GGN2026.06 | 85.85 | 0.2201 | 0.0791 | 0.4863 | |
| Ribbon FullSelection Criterion=Brier, Parameter Setting=Full-parameter, Curvature Approximation=PSD-GGN2026.06 | 85.85 | 0.2204 | 0.0799 | 0.4866 | |
| Ribbon FullSelection Criterion=ECE, Parameter Setting=Full-parameter, Curvature Approximation=PSD-GGN2026.06 | 85.84 | 0.2204 | 0.0797 | 0.4867 | |
| Ribbon FullSelection Criterion=NLL, Parameter Setting=Full-parameter, Curvature Approximation=PSD-GGN2026.06 | 85.82 | 0.2204 | 0.0795 | 0.4868 | |
| MC Dropout2026.06 | 85.41 | 0.2417 | 0.1302 | 0.538 | |
| VAE+WDDWCCEvaluation protocol=KNN classification, Number of runs=52025.07 | 78.6 | — | — | — | |
| VAE+WDIWCDEvaluation protocol=KNN classification, Number of runs=52025.07 | 75.5 | — | — | — | |
| VAE+CELEvaluation protocol=KNN classification, Number of runs=52025.07 | 75.1 | — | — | — | |
| Only CELEvaluation protocol=KNN classification, Number of runs=52025.07 | 73.8 | — | — | — | |
| Only WDIWCDEvaluation protocol=KNN classification, Number of runs=52025.07 | 72.6 | — | — | — | |
| SPCEvaluation protocol=KNN classification, Number of runs=52025.07 | 71.4 | — | — | — | |
| Only WDDWCCEvaluation protocol=KNN classification, Number of runs=52025.07 | 71 | — | — | — | |
| LDAEvaluation protocol=KNN classification, Number of runs=52025.07 | 69.7 | — | — | — | |
| Only VAEEvaluation protocol=KNN classification, Number of runs=52025.07 | 68.8 | — | — | — |