Abstract
Given a pre-trained in-distribution (ID) model, the inference-time out-of-distribution (OOD) detection aims to recognize OOD data during the inference stage. However, some representative methods share an unproven assumption that the probability that OOD data belong to every ID class should be the same, i.e., these OOD-to-ID probabilities actually form a uniform distribution. In this paper, we show that this assumption makes the above methods incapable when the ID model is trained with class-imbalanced data.Fortunately, by analyzing the causal relations between ID/OOD classes and features, we identify several common scenarios where the OOD-to-ID probabilities should be the ID-class-prior distribution and propose two strategies to modify existing inference-time detection methods: 1) replace the uniform distribution with the ID-class-prior distribution if they explicitly use the uniform distribution; 2) otherwise, reweight their scores according to the similarity between the ID-class-prior distribution and the softmax outputs of the pre-trained model. Extensive experiments show that both strategies can improve the OOD detection performance when the ID model is pre-trained with imbalanced data, reflecting the importance of ID-class prior in OOD detection.
Original language | English |
---|---|
Title of host publication | Proceedings of the 40th International Conference on Machine Learning, ICML 2023 |
Editors | Andreas Krause, Emma Brunskill, Kyunghyun Cho, Barbara Engelhardt, Sivan Sabato, Jonathan Scarlett |
Publisher | ML Research Press |
Number of pages | 22 |
Publication status | Published - Jul 2023 |
Event | 40th International Conference on Machine Learning, ICML 2023 - Honolulu, United States Duration: 23 Jul 2023 → 29 Jul 2023 https://icml.cc/Conferences/2023 https://proceedings.mlr.press/v202/ https://openreview.net/group?id=ICML.cc/2023/Conference |
Publication series
Name | Proceedings of Machine Learning Research |
---|---|
Volume | 202 |
ISSN (Print) | 2640-3498 |
Conference
Conference | 40th International Conference on Machine Learning, ICML 2023 |
---|---|
Country/Territory | United States |
City | Honolulu |
Period | 23/07/23 → 29/07/23 |
Internet address |