Out-of-distribution Detection on 20 Newsgroup (test)
0.76FPR@95Deep Abstaining Classifier (DAC)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Deep Abstaining Classifier (DAC)Model Architecture=2 Layered GRU, Out-of-distribution Dataset=Yelp2021.05 | 0.76 | 99.61 | |
| Deep Abstaining Classifier (DAC)Model Architecture=2 Layered GRU, Out-of-distribution Dataset=Multi30K2021.05 | 0.8 | 99.52 | |
| Deep Abstaining Classifier (DAC)Model Architecture=2 Layered GRU, Out-of-distribution Dataset=WMT162021.05 | 1.4 | 99.33 | |
| Deep Abstaining Classifier (DAC)Model Architecture=2 Layered GRU, Out-of-distribution Dataset=IMDB2021.05 | 1.78 | 99.17 | |
| Outlier Exposure (OE)Model Architecture=2 Layered GRU, Out-of-distribution Dataset=WMT162021.05 | 2 | 98.8 | |
| Outlier Exposure (OE)Model Architecture=2 Layered GRU, Out-of-distribution Dataset=Multi30K2021.05 | 3.2 | 97.3 | |
| Deep Abstaining Classifier (DAC)Model Architecture=2 Layered GRU, Out-of-distribution Dataset=SNLI2021.05 | 3.9 | 98.32 | |
| Outlier Exposure (OE)Model Architecture=2 Layered GRU, Out-of-distribution Dataset=Yelp2021.05 | 3.9 | 97.8 | |
| Outlier Exposure (OE)Model Architecture=2 Layered GRU, Out-of-distribution Dataset=SNLI2021.05 | 12.5 | 95.1 | |
| Outlier Exposure (OE)Model Architecture=2 Layered GRU, Out-of-distribution Dataset=IMDB2021.05 | 18.6 | 93.5 |