dc.contributor.author | Xing, M | en_US |
dc.contributor.author | Feng, Z | en_US |
dc.contributor.author | Su, Y | en_US |
dc.contributor.author | Oh, C | en_US |
dc.contributor.author | AAAI Conference on Artificial Intelligence 2024 | en_US |
dc.date.accessioned | 2024-03-15T09:05:29Z | |
dc.date.available | 2023-12-09 | en_US |
dc.identifier.uri | https://qmro.qmul.ac.uk/xmlui/handle/123456789/95364 | |
dc.description.abstract | Detecting OOD inputs is crucial to deploy machine learning models to the real world safely. However, existing OOD detection methods require an in-distribution (ID) dataset to retrain the models. In this paper, we propose a Deep Generative Models (DGMs) based transferable OOD detection that does not require retraining on the new ID dataset. We first establish and substantiate two hypotheses on DGMs: DGMs exhibit a predisposition towards acquiring low-level features, in preference to semantic information; the lower bound of DGM's log-likelihoods is tied to the conditional entropy between the model input and target output. Drawing on the aforementioned hypotheses, we present an innovative image-erasing strategy, which is designed to create distinct conditional entropy distributions for each individual ID dataset. By training a DGM on a complex dataset with the proposed image-erasing strategy, the DGM could capture the discrepancy of conditional entropy distribution for varying ID datasets, without re-training. We validate the proposed method on the five datasets and show that, without retraining, our method achieves comparable performance to the state-of-the-art group-based OOD detection methods. The project codes will be open-sourced on our project website. | |
dc.rights | This is a pre-copyedited, author-produced version accepted for publication in Proceedings of the AAAI Conference on Artificial Intelligence in Play following peer review. The version of record is available at https://ojs.aaai.org/index.php/AAAI/article/view/28444 | |
dc.rights | This item is distributed under the terms of the Creative Commons Attribution 4.0 Unported License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. | |
dc.title | Learning by Erasing: Conditional Entropy Based Transferable Out-of-Distribution Detection | en_US |
dc.type | Conference Proceeding | |
dc.rights.holder | © 2024, Association for the Advancement of Artificial Intelligence | |
dc.rights.holder | © 2024, The Author(s) | |
pubs.notes | Not known | en_US |
pubs.publication-status | Accepted | en_US |
dcterms.dateAccepted | 2023-12-09 | en_US |
rioxxterms.funder | Default funder | en_US |
rioxxterms.identifier.project | Default project | en_US |