{"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/balancing-speed-and-quality-in-online","title":"Balancing Speed and Quality in Online Learning to Rank for Information Retrieval","arxiv_id":"1711.09446","date":"2017-11-26","proceeding":null,"authors":["Oosterhuis Harrie","de Rijke Maarten"],"abstract":"In Online Learning to Rank (OLTR) the aim is to find an optimal ranking model\nby interacting with users. When learning from user behavior, systems must\ninteract with users while simultaneously learning from those interactions.\nUnlike other Learning to Rank (LTR) settings, existing research in this field\nhas been limited to linear models. This is due to the speed-quality tradeoff\nthat arises when selecting models: complex models are more expressive and can\nfind the best rankings but need more user interactions to do so, a requirement\nthat risks frustrating users during training. Conversely, simpler models can be\noptimized on fewer interactions and thus provide a better user experience, but\nthey will converge towards suboptimal rankings. This tradeoff creates a\ndeadlock, since novel models will not be able to improve either the user\nexperience or the final convergence point, without sacrificing the other. Our\ncontribution is twofold. First, we introduce a fast OLTR model called Sim-MGD\nthat addresses the speed aspect of the speed-quality tradeoff. Sim-MGD ranks\ndocuments based on similarities with reference documents. It converges rapidly\nand, hence, gives a better user experience but it does not converge towards the\noptimal rankings. Second, we contribute Cascading Multileave Gradient Descent\n(C-MGD) for OLTR that directly addresses the speed-quality tradeoff by using a\ncascade that enables combinations of the best of two worlds: fast learning and\nhigh quality final convergence. C-MGD can provide the better user experience of\nSim-MGD while maintaining the same convergence as the state-of-the-art MGD\nmodel. This opens the door for future work to design new models for OLTR\nwithout having to deal with the speed-quality tradeoff.","url_abs":"http://arxiv.org/abs/1711.09446v1","url_pdf":"http://arxiv.org/pdf/1711.09446v1.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":"balancing-speed-and-quality-in-online","repo_url":"https://github.com/HarrieO/BalancingSpeedQualityOLTR","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"information-retrieval","task_name":"Information Retrieval"},{"task_slug":"learning-to-rank","task_name":"Learning-To-Rank"},{"task_slug":"retrieval","task_name":"Retrieval"}],"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}