{"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/offline-multi-action-policy-learning","title":"Offline Multi-Action Policy Learning: Generalization and Optimization","arxiv_id":"1810.04778","date":"2018-10-10","proceeding":null,"authors":["Zhengyuan Zhou","Susan Athey","Stefan Wager"],"abstract":"In many settings, a decision-maker wishes to learn a rule, or policy, that\nmaps from observable characteristics of an individual to an action. Examples\ninclude selecting offers, prices, advertisements, or emails to send to\nconsumers, as well as the problem of determining which medication to prescribe\nto a patient. While there is a growing body of literature devoted to this\nproblem, most existing results are focused on the case where data comes from a\nrandomized experiment, and further, there are only two possible actions, such\nas giving a drug to a patient or not. In this paper, we study the offline\nmulti-action policy learning problem with observational data and where the\npolicy may need to respect budget constraints or belong to a restricted policy\nclass such as decision trees. We build on the theory of efficient\nsemi-parametric inference in order to propose and implement a policy learning\nalgorithm that achieves asymptotically minimax-optimal regret. To the best of\nour knowledge, this is the first result of this type in the multi-action setup,\nand it provides a substantial performance improvement over the existing\nlearning algorithms. We then consider additional computational challenges that\narise in implementing our method for the case where the policy is restricted to\ntake the form of a decision tree. We propose two different approaches, one\nusing a mixed integer program formulation and the other using a tree-search\nbased algorithm.","url_abs":"http://arxiv.org/abs/1810.04778v2","url_pdf":"http://arxiv.org/pdf/1810.04778v2.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":"offline-multi-action-policy-learning","repo_url":"https://github.com/grf-labs/policyTree","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1810.04778","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}