{"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/predictive-uncertainty-estimation-via-prior","title":"Predictive Uncertainty Estimation via Prior Networks","arxiv_id":"1802.10501","date":"2018-02-28","proceeding":"NeurIPS 2018 12","authors":["Andrey Malinin","Mark Gales"],"abstract":"Estimating how uncertain an AI system is in its predictions is important to\nimprove the safety of such systems. Uncertainty in predictive can result from\nuncertainty in model parameters, irreducible data uncertainty and uncertainty\ndue to distributional mismatch between the test and training data\ndistributions. Different actions might be taken depending on the source of the\nuncertainty so it is important to be able to distinguish between them.\nRecently, baseline tasks and metrics have been defined and several practical\nmethods to estimate uncertainty developed. These methods, however, attempt to\nmodel uncertainty due to distributional mismatch either implicitly through\nmodel uncertainty or as data uncertainty. This work proposes a new framework\nfor modeling predictive uncertainty called Prior Networks (PNs) which\nexplicitly models distributional uncertainty. PNs do this by parameterizing a\nprior distribution over predictive distributions. This work focuses on\nuncertainty for classification and evaluates PNs on the tasks of identifying\nout-of-distribution (OOD) samples and detecting misclassification on the MNIST\ndataset, where they are found to outperform previous methods. Experiments on\nsynthetic and MNIST and CIFAR-10 data show that unlike previous non-Bayesian\nmethods PNs are able to distinguish between data and distributional\nuncertainty.","url_abs":"http://arxiv.org/abs/1802.10501v4","url_pdf":"http://arxiv.org/pdf/1802.10501v4.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":"predictive-uncertainty-estimation-via-prior","repo_url":"https://github.com/asharakeh/probdet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1802.10501","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}