{"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/lossgrad-automatic-learning-rate-in-gradient","title":"LOSSGRAD: automatic learning rate in gradient descent","arxiv_id":"1902.07656","date":"2019-02-20","proceeding":null,"authors":["Bartosz Wójcik","Łukasz Maziarka","Jacek Tabor"],"abstract":"In this paper, we propose a simple, fast and easy to implement algorithm\nLOSSGRAD (locally optimal step-size in gradient descent), which automatically\nmodifies the step-size in gradient descent during neural networks training.\nGiven a function $f$, a point $x$, and the gradient $\\nabla_x f$ of $f$, we aim\nto find the step-size $h$ which is (locally) optimal, i.e. satisfies: $$\nh=arg\\,min_{t \\geq 0} f(x-t \\nabla_x f). $$ Making use of quadratic\napproximation, we show that the algorithm satisfies the above assumption. We\nexperimentally show that our method is insensitive to the choice of initial\nlearning rate while achieving results comparable to other methods.","url_abs":"http://arxiv.org/abs/1902.07656v1","url_pdf":"http://arxiv.org/pdf/1902.07656v1.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":"lossgrad-automatic-learning-rate-in-gradient","repo_url":"https://github.com/bartwojcik/lossgrad","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}