{"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/matrix-factorization-on-gpus-with-memory","title":"Matrix Factorization on GPUs with Memory Optimization and Approximate Computing","arxiv_id":"1808.03843","date":"2018-08-11","proceeding":null,"authors":["Wei Tan","Shiyu Chang","Liana Fong","Cheng Li","Zijun Wang","Liangliang Cao"],"abstract":"Matrix factorization (MF) discovers latent features from observations, which\nhas shown great promises in the fields of collaborative filtering, data\ncompression, feature extraction, word embedding, etc. While many\nproblem-specific optimization techniques have been proposed, alternating least\nsquare (ALS) remains popular due to its general applicability e.g. easy to\nhandle positive-unlabeled inputs, fast convergence and parallelization\ncapability. Current MF implementations are either optimized for a single\nmachine or with a need of a large computer cluster but still are insufficient.\nThis is because a single machine provides limited compute power for large-scale\ndata while multiple machines suffer from the network communication bottleneck.\n  To address the aforementioned challenge, accelerating ALS on graphics\nprocessing units (GPUs) is a promising direction. We propose the novel approach\nin enhancing the MF efficiency via both memory optimization and approximate\ncomputing. The former exploits GPU memory hierarchy to increase data reuse,\nwhile the later reduces unnecessary computing without hurting the convergence\nof learning algorithms. Extensive experiments on large-scale datasets show that\nour solution not only outperforms the competing CPU solutions by a large margin\nbut also has a 2x-4x performance gain compared to the state-of-the-art GPU\nsolutions. Our implementations are open-sourced and publicly available.","url_abs":"http://arxiv.org/abs/1808.03843v1","url_pdf":"http://arxiv.org/pdf/1808.03843v1.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":"matrix-factorization-on-gpus-with-memory","repo_url":"https://github.com/cuMF/cumf_als","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":null,"task_name":"CPU"},{"task_slug":"collaborative-filtering","task_name":"Collaborative Filtering"},{"task_slug":"data-compression","task_name":"Data Compression"},{"task_slug":null,"task_name":"GPU"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}