{"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/simple-bayesian-algorithms-for-best-arm","title":"Simple Bayesian Algorithms for Best Arm Identification","arxiv_id":"1602.08448","date":"2016-02-26","proceeding":null,"authors":["Daniel Russo"],"abstract":"This paper considers the optimal adaptive allocation of measurement effort\nfor identifying the best among a finite set of options or designs. An\nexperimenter sequentially chooses designs to measure and observes noisy signals\nof their quality with the goal of confidently identifying the best design after\na small number of measurements. This paper proposes three simple and intuitive\nBayesian algorithms for adaptively allocating measurement effort, and\nformalizes a sense in which these seemingly naive rules are the best possible.\nOne proposal is top-two probability sampling, which computes the two designs\nwith the highest posterior probability of being optimal, and then randomizes to\nselect among these two. One is a variant of top-two sampling which considers\nnot only the probability a design is optimal, but the expected amount by which\nits quality exceeds that of other designs. The final algorithm is a modified\nversion of Thompson sampling that is tailored for identifying the best design.\nWe prove that these simple algorithms satisfy a sharp optimality property. In a\nfrequentist setting where the true quality of the designs is fixed, one hopes\nthe posterior definitively identifies the optimal design, in the sense that\nthat the posterior probability assigned to the event that some other design is\noptimal converges to zero as measurements are collected. We show that under the\nproposed algorithms this convergence occurs at an exponential rate, and the\ncorresponding exponent is the best possible among all allocation","url_abs":"http://arxiv.org/abs/1602.08448v4","url_pdf":"http://arxiv.org/pdf/1602.08448v4.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":"simple-bayesian-algorithms-for-best-arm","repo_url":"https://github.com/PlaytikaOSS/pybandits","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"none","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[{"task_slug":"thompson-sampling","task_name":"Thompson Sampling"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1602.08448","atlas_url":"https://app.syntology.ai/?focus=1602.08448","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}