{"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/manifold-identification-for-ultimately","title":"Manifold Identification for Ultimately Communication-Efficient Distributed Optimization","arxiv_id":null,"date":"2020-01-01","proceeding":"ICML 2020 1","authors":["Yu-Sheng Li","Wei-Lin Chiang","Ching-pei Lee"],"abstract":"The expensive inter-machine communication is the bottleneck of\ndistributed optimization.\nExisting study tackles this problem by shortening the communication\nrounds, but the reduction of per-round communication cost is not\nwell-studied.\nThis work proposes a progressive manifold identification approach with \nsound theoretical justifications to greatly reduce both the\ncommunication rounds and the bytes communicated per round for partly\nsmooth regularized problems, which include many large-scale machine\nlearning tasks such as the training of $\\ell_1$- and\ngroup-LASSO-regularized models.\nOur method uses an inexact proximal quasi-Newton method to iteratively\nidentify a sequence of low-dimensional smooth manifolds in which the\nfinal solution lies, and restricts the model update within the current\nmanifold to lower significantly the per-round communication cost.\nAfter identifying the final manifold within which the problem is\nsmooth, we take superlinear-convergent truncated semismooth\nNewton steps obtained through preconditioned conjugate gradient to\nlargely reduce the communication rounds.\nExperiments show that when compared with the state of the art,\nthe communication cost of our method is significantly lower\nand the running time is up to $10$ times faster.","url_abs":"https://proceedings.icml.cc/static/paper_files/icml/2020/3524-Paper.pdf","url_pdf":"https://proceedings.icml.cc/static/paper_files/icml/2020/3524-Paper.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":"manifold-identification-for-ultimately","repo_url":"https://github.com/leepei/madpqn","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"distributed-optimization","task_name":"Distributed Optimization"}],"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}