{"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/slang-fast-structured-covariance","title":"SLANG: Fast Structured Covariance Approximations for Bayesian Deep Learning with Natural Gradient","arxiv_id":"1811.04504","date":"2018-11-11","proceeding":"NeurIPS 2018 12","authors":["Aaron Mishkin","Frederik Kunstner","Didrik Nielsen","Mark Schmidt","Mohammad Emtiyaz Khan"],"abstract":"Uncertainty estimation in large deep-learning models is a computationally\nchallenging task, where it is difficult to form even a Gaussian approximation\nto the posterior distribution. In such situations, existing methods usually\nresort to a diagonal approximation of the covariance matrix despite, the fact\nthat these matrices are known to result in poor uncertainty estimates. To\naddress this issue, we propose a new stochastic, low-rank, approximate\nnatural-gradient (SLANG) method for variational inference in large, deep\nmodels. Our method estimates a \"diagonal plus low-rank\" structure based solely\non back-propagated gradients of the network log-likelihood. This requires\nstrictly less gradient computations than methods that compute the gradient of\nthe whole variational objective. Empirical evaluations on standard benchmarks\nconfirm that SLANG enables faster and more accurate estimation of uncertainty\nthan mean-field methods, and performs comparably to state-of-the-art methods.","url_abs":"http://arxiv.org/abs/1811.04504v2","url_pdf":"http://arxiv.org/pdf/1811.04504v2.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":"slang-fast-structured-covariance","repo_url":"https://github.com/aaronpmishkin/SLANG","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"slang-fast-structured-covariance","repo_url":"https://github.com/lamantinushka/StructuredCovariance","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"variational-inference","task_name":"Variational Inference"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1811.04504","atlas_url":"https://app.syntology.ai/?focus=1811.04504","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}