{"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/accelerating-generalized-linear-models-with","title":"Accelerating Generalized Linear Models with MLWeaving: A One-Size-Fits-All System for Any-precision Learning (Technical Report)","arxiv_id":"1903.03404","date":"2019-03-08","proceeding":null,"authors":["Zeke Wang","Kaan Kara","Hantian Zhang","Gustavo Alonso","Onur Mutlu","Ce Zhang"],"abstract":"Learning from the data stored in a database is an important function\nincreasingly available in relational engines. Methods using lower precision\ninput data are of special interest given their overall higher efficiency but,\nin databases, these methods have a hidden cost: the quantization of the real\nvalue into a smaller number is an expensive step. To address the issue, in this\npaper we present MLWeaving, a data structure and hardware acceleration\ntechnique intended to speed up learning of generalized linear models in\ndatabases. ML-Weaving provides a compact, in-memory representation enabling the\nretrieval of data at any level of precision. MLWeaving also takes advantage of\nthe increasing availability of FPGA-based accelerators to provide a highly\nefficient implementation of stochastic gradient descent. The solution adopted\nin MLWeaving is more efficient than existing designs in terms of space (since\nit can process any resolution on the same design) and resources (via the use of\nbit-serial multipliers). MLWeaving also enables the runtime tuning of\nprecision, instead of a fixed precision level during the training. We\nillustrate this using a simple, dynamic precision schedule. Experimental\nresults show MLWeaving achieves up to16 performance improvement over\nlow-precision CPU implementations of first-order methods.","url_abs":"http://arxiv.org/abs/1903.03404v2","url_pdf":"http://arxiv.org/pdf/1903.03404v2.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":"accelerating-generalized-linear-models-with","repo_url":"https://github.com/fpgasystems/MLWeaving","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"all","task_name":"All"},{"task_slug":null,"task_name":"CPU"},{"task_slug":"quantization","task_name":"Quantization"},{"task_slug":"retrieval","task_name":"Retrieval"}],"methods":[{"method_slug":"speed","method_name":"SPEED"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}