{"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/scalable-learning-of-non-decomposable","title":"Scalable Learning of Non-Decomposable Objectives","arxiv_id":"1608.04802","date":"2016-08-16","proceeding":null,"authors":["Elad ET. Eban","Mariano Schain","Alan Mackey","Ariel Gordon","Rif A. Saurous","Gal Elidan"],"abstract":"Modern retrieval systems are often driven by an underlying machine learning\nmodel. The goal of such systems is to identify and possibly rank the few most\nrelevant items for a given query or context. Thus, such systems are typically\nevaluated using a ranking-based performance metric such as the area under the\nprecision-recall curve, the $F_\\beta$ score, precision at fixed recall, etc.\nObviously, it is desirable to train such systems to optimize the metric of\ninterest.\n  In practice, due to the scalability limitations of existing approaches for\noptimizing such objectives, large-scale retrieval systems are instead trained\nto maximize classification accuracy, in the hope that performance as measured\nvia the true objective will also be favorable. In this work we present a\nunified framework that, using straightforward building block bounds, allows for\nhighly scalable optimization of a wide range of ranking-based objectives. We\ndemonstrate the advantage of our approach on several real-life retrieval\nproblems that are significantly larger than those considered in the literature,\nwhile achieving substantial improvement in performance over the\naccuracy-objective baseline.","url_abs":"http://arxiv.org/abs/1608.04802v2","url_pdf":"http://arxiv.org/pdf/1608.04802v2.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":"scalable-learning-of-non-decomposable","repo_url":"https://github.com/tensorflow/models","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"scalable-learning-of-non-decomposable","repo_url":"https://github.com/tensorflow/models/tree/master/research/global_objectives","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"retrieval","task_name":"Retrieval"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1608.04802","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}