{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/analysis-of-confident-classifiers-for-out-of","title":"Analysis of Confident-Classifiers for Out-of-distribution Detection","arxiv_id":"1904.12220","date":"2019-04-27","proceeding":null,"authors":["Sachin Vernekar","Ashish Gaurav","Taylor Denouden","Buu Phan","Vahdat Abdelzad","Rick Salay","Krzysztof Czarnecki"],"abstract":"Discriminatively trained neural classifiers can be trusted, only when the\ninput data comes from the training distribution (in-distribution). Therefore,\ndetecting out-of-distribution (OOD) samples is very important to avoid\nclassification errors. In the context of OOD detection for image\nclassification, one of the recent approaches proposes training a classifier\ncalled \"confident-classifier\" by minimizing the standard cross-entropy loss on\nin-distribution samples and minimizing the KL divergence between the predictive\ndistribution of OOD samples in the low-density regions of in-distribution and\nthe uniform distribution (maximizing the entropy of the outputs). Thus, the\nsamples could be detected as OOD if they have low confidence or high entropy.\nIn this paper, we analyze this setting both theoretically and experimentally.\nWe conclude that the resulting confident-classifier still yields arbitrarily\nhigh confidence for OOD samples far away from the in-distribution. We instead\nsuggest training a classifier by adding an explicit \"reject\" class for OOD\nsamples.","url_abs":"http://arxiv.org/abs/1904.12220v1","url_pdf":"http://arxiv.org/pdf/1904.12220v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"analysis-of-confident-classifiers-for-out-of","repo_url":"https://github.com/sverneka/ConfidentClassifierICLR19","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}}],"tasks":[{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"ood-detection","task_name":"Out of Distribution (OOD) Detection"},{"task_slug":"out-of-distribution-detection","task_name":"Out-of-Distribution Detection"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1904.12220","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}