Table 2.
OOD detection for regression on YearPredictionMSD + Boston Housing.
| Model | #Parameters | RMSE | TNR at TPR 95% | AUROC | Detection Accuracy |
AUPR In |
AUPR Out |
|---|---|---|---|---|---|---|---|
| BBP | 30.0K | 9.5 ± 0.2 | 9.0 ± 1.4 | 56.8 ± 0.9 | 52.1 ± 0.7 | 45.3 ± 1.3 | 1.3 ± 0.1 |
| MC-dropout | 14.9K | 8.8 ± 0.0 | 6.1 ± 0.5 | 53.0 ± 1.2 | 53.7 ± 0.6 | 99.2 ± 0.2 | 1.1 ± 0.1 |
| ANPs–SDE-Net | 288.5K | 8.8 ± 0.0 | 44.4 ± 4.2 | 84.2 ± 1.6 | 75.2 ± 1.7 | 99.8 ± 0.0 | 28.8 ± 2.0 |
| CNPs–SDE-Net | 141.0K | 8.9 ± 0.1 | 7.9 ± 1.2 | 59.3 ± 1.5 | 58.6 ± 0.9 | 99.2 ± 0.1 | 1.3 ± 0.1 |
| SDE-Net | 12.4K | 8.7 ± 0.1 | 64.3 ± 0.6 | 84.1 ± 1.1 | 80.6 ± 0.5 | 99.7 ± 0.0 | 24.7 ± 1.0 |