{"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/nostalgic-adam-weighting-more-of-the-past","title":"Nostalgic Adam: Weighting more of the past gradients when designing the adaptive learning rate","arxiv_id":"1805.07557","date":"2018-05-19","proceeding":null,"authors":["Haiwen Huang","Chang Wang","Bin Dong"],"abstract":"First-order optimization algorithms have been proven prominent in deep\nlearning. In particular, algorithms such as RMSProp and Adam are extremely\npopular. However, recent works have pointed out the lack of ``long-term memory\"\nin Adam-like algorithms, which could hamper their performance and lead to\ndivergence. In our study, we observe that there are benefits of weighting more\nof the past gradients when designing the adaptive learning rate. We therefore\npropose an algorithm called the Nostalgic Adam (NosAdam) with theoretically\nguaranteed convergence at the best known convergence rate. NosAdam can be\nregarded as a fix to the non-convergence issue of Adam in alternative to the\nrecent work of [Reddi et al., 2018]. Our preliminary numerical experiments show\nthat NosAdam is a promising alternative algorithm to Adam. The proofs, code and\nother supplementary materials can be found in an anonymously shared link.","url_abs":"http://arxiv.org/abs/1805.07557v2","url_pdf":"http://arxiv.org/pdf/1805.07557v2.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":"nostalgic-adam-weighting-more-of-the-past","repo_url":"https://github.com/andrehuang/NostalgicAdam-NosAdam","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"nostalgic-adam-weighting-more-of-the-past","repo_url":"https://github.com/andrehuang/NostalgicAdam","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[],"methods":[{"method_slug":"adam","method_name":"Adam"},{"method_slug":"rmsprop","method_name":"RMSProp"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1805.07557","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}