Out-of-Distribution Detection on FashionMNIST (ID) vs EMNIST (OoD) (test)
0.907AUROCLIN-LA
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| LIN-LAPrior Precision=Not Optimized2023.06 | 0.907 | — | — | — | — | |
| LIN-RIEM-LAPrior Precision=Not Optimized, Exponential Maps Source=Subset of Training Set2023.06 | 0.891 | — | — | — | — | |
| LIN-RIEM-LAPrior Precision=Not Optimized, Exponential Maps Source=Full Training Set2023.06 | 0.881 | — | — | — | — | |
| RIEM-LAPrior Precision=Optimized, Exponential Maps Source=Full Training Set2023.06 | 0.857 | — | — | — | — | |
| RIEM-LAPrior Precision=Optimized, Exponential Maps Source=Subset of Training Set2023.06 | 0.851 | — | — | — | — | |
| MC-DOptimizer=MC-D2025.11 | 0.821 | 0.244 | 0.25 | 0.752 | 0.885 | |
| MC-DOptimizer=MC-D2025.11 | 0.821 | 0.244 | — | — | 0.885 | |
| IVONOptimizer=IVON2025.11 | 0.82 | 0.236 | 0.249 | 0.743 | 0.883 | |
| RIEM-LAPrior Precision=Not Optimized, Exponential Maps Source=Subset of Training Set2023.06 | 0.814 | — | — | — | — | |
| MC-DOptimizer=MC-D2025.11 | 0.813 | 0.268 | 0.246 | 0.734 | 0.875 | |
| uCBOptvartheta=8 * 10^{-6}2025.11 | 0.81 | 0.245 | 0.259 | 0.72 | 0.879 | |
| uCBOptOptimizer=uCBOpt2025.11 | 0.81 | 0.245 | — | — | 0.879 | |
| LIN-LAPrior Precision=Optimized2023.06 | 0.806 | — | — | — | — | |
| IVONOptimizer=IVON2025.11 | 0.804 | 0.249 | 0.265 | 0.719 | 0.873 | |
| IVONOptimizer=IVON2025.11 | 0.804 | 0.249 | — | — | 0.873 | |
| IVONOptimizer=IVON@mean2025.11 | 0.796 | 0.261 | 0.268 | 0.708 | 0.866 | |
| uCBOptOptimizer=uCBOpt (ϑ → 0+)2025.11 | 0.796 | 0.266 | 0.267 | 0.7 | 0.865 | |
| uCBOptvartheta=-> 0+2025.11 | 0.796 | 0.262 | 0.27 | 0.708 | 0.867 | |
| LIN-RIEM-LAPrior Precision=Optimized, Exponential Maps Source=Subset of Training Set2023.06 | 0.795 | — | — | — | — | |
| RIEM-LAPrior Precision=Not Optimized, Exponential Maps Source=Full Training Set2023.06 | 0.794 | — | — | — | — | |
| uCBOptOptimizer=uCBOpt (ϑ = 8 ⋅ 10−6)2025.11 | 0.794 | 0.266 | 0.271 | 0.694 | 0.865 | |
| IVON@meanOptimizer=IVON@mean2025.11 | 0.787 | 0.266 | 0.279 | 0.694 | 0.861 | |
| IVON@meanOptimizer=IVON@mean2025.11 | 0.787 | 0.266 | — | — | 0.861 | |
| lCBOpt-adaptOptimizer=lCBOpt-adapt2025.11 | 0.784 | 0.269 | 0.278 | 0.674 | 0.859 | |
| lCBOpt-adaptOptimizer=lCBOpt-adapt2025.11 | 0.784 | 0.269 | 0.278 | 0.674 | 0.859 | |
| lCBOpt-adaptOptimizer=lCBOpt-adapt2025.11 | 0.784 | 0.269 | — | — | 0.859 | |
| uCBOpt-adaptOptimizer=uCBOpt-adapt2025.11 | 0.773 | 0.272 | 0.29 | 0.655 | 0.857 | |
| uCBOpt-adaptOptimizer=uCBOpt-adapt2025.11 | 0.773 | 0.272 | 0.29 | 0.655 | 0.857 | |
| uCBOpt-adaptOptimizer=uCBOpt-adapt2025.11 | 0.773 | 0.272 | — | — | 0.857 | |
| LaplaceOptimizer=Laplace2025.11 | 0.77 | 0.296 | 0.29 | 0.675 | 0.848 | |
| LaplaceOptimizer=Laplace2025.11 | 0.77 | 0.296 | — | — | 0.848 | |
| SGDOptimizer=SGD2025.11 | 0.767 | 0.297 | 0.293 | 0.676 | 0.845 | |
| SGDOptimizer=SGD2025.11 | 0.767 | 0.297 | — | — | 0.845 | |
| Vanilla LAPrior Precision=Optimized2023.06 | 0.765 | — | — | — | — | |
| LIN-RIEM-LAPrior Precision=Optimized, Exponential Maps Source=Full Training Set2023.06 | 0.75 | — | — | — | — | |
| AdamWOptimizer=AdamW2025.11 | 0.745 | 0.321 | 0.312 | 0.627 | 0.834 | |
| AdamWOptimizer=AdamW2025.11 | 0.745 | 0.321 | — | — | 0.627 | |
| SWAGOptimizer=SWAG2025.11 | 0.731 | 0.378 | 0.324 | 0.601 | 0.82 | |
| SWAGOptimizer=SWAG2025.11 | 0.731 | 0.378 | 0.324 | 0.601 | 0.82 | |
| SWAGOptimizer=SWAG2025.11 | 0.731 | 0.378 | — | — | 0.82 | |
| LaplaceOptimizer=Laplace2025.11 | 0.68 | 0.519 | 0.363 | 0.517 | 0.784 | |
| SGDOptimizer=SGD2025.11 | 0.674 | 0.52 | 0.368 | 0.508 | 0.78 | |
| MAPPrior Precision=Not Optimized2023.06 | 0.649 | — | — | — | — | |
| MAPPrior Precision=Optimized2023.06 | 0.649 | — | — | — | — | |
| AdamWOptimizer=AdamW2025.11 | 0.647 | 0.686 | 0.369 | 0.429 | 0.761 | |
| Vanilla LAPrior Precision=Not Optimized2023.06 | 0.495 | — | — | — | — |