Image Classification and Robustness on MNIST (ID) vs KMNIST (OOD) (test)
99.98ID AccuracyEDL++
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| EDL++Uncertainty Strategy=Monte Carlo2025.06 | 99.98 | 91.94 | 11.67 | 27.79 | |
| C-EDLUncertainty Strategy=Monte Carlo2025.06 | 99.98 | 90.2 | 3.69 | 12.4 | |
| EDL++Uncertainty Strategy=Metamorphic Transformation2025.06 | 99.98 | 93.73 | 2.21 | 4.2 | |
| C-EDLUncertainty Strategy=Metamorphic Transformation2025.06 | 99.98 | 92.42 | 1.9 | 3.01 | |
| Posterior Network2025.06 | 99.97 | 94.72 | 3.78 | 23.91 | |
| EDL2025.06 | 99.97 | 95.42 | 3.23 | 20.88 | |
| S-EDL2025.06 | 99.97 | 95.08 | 3.16 | 14.96 | |
| H-EDL2025.06 | 99.97 | 95.59 | 3.33 | 20.03 | |
| I-EDL2025.06 | 99.96 | 95.47 | 3.54 | 19.57 | |
| R-EDL2025.06 | 99.96 | 95.32 | 3.21 | 16.4 | |
| DA-EDL2025.06 | 99.9 | 94.12 | 3.37 | 10.06 |