{"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/sketching-linear-classifiers-over-data","title":"Sketching Linear Classifiers over Data Streams","arxiv_id":"1711.02305","date":"2017-11-07","proceeding":null,"authors":["Kai Sheng Tai","Vatsal Sharan","Peter Bailis","Gregory Valiant"],"abstract":"We introduce a new sub-linear space sketch---the Weight-Median Sketch---for\nlearning compressed linear classifiers over data streams while supporting the\nefficient recovery of large-magnitude weights in the model. This enables\nmemory-limited execution of several statistical analyses over streams,\nincluding online feature selection, streaming data explanation, relative\ndeltoid detection, and streaming estimation of pointwise mutual information.\nUnlike related sketches that capture the most frequently-occurring features (or\nitems) in a data stream, the Weight-Median Sketch captures the features that\nare most discriminative of one stream (or class) compared to another. The\nWeight-Median Sketch adopts the core data structure used in the Count-Sketch,\nbut, instead of sketching counts, it captures sketched gradient updates to the\nmodel parameters. We provide a theoretical analysis that establishes recovery\nguarantees for batch and online learning, and demonstrate empirical\nimprovements in memory-accuracy trade-offs over alternative memory-budgeted\nmethods, including count-based sketches and feature hashing.","url_abs":"http://arxiv.org/abs/1711.02305v2","url_pdf":"http://arxiv.org/pdf/1711.02305v2.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":"sketching-linear-classifiers-over-data","repo_url":"https://github.com/stanford-futuredata/wmsketch","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"feature-selection","task_name":"feature selection"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}