{"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/efficient-counterfactual-learning-from-bandit","title":"Efficient Counterfactual Learning from Bandit Feedback","arxiv_id":"1809.03084","date":"2018-09-10","proceeding":null,"authors":["Yusuke Narita","Shota Yasui","Kohei Yata"],"abstract":"What is the most statistically efficient way to do off-policy evaluation and\noptimization with batch data from bandit feedback? For log data generated by\ncontextual bandit algorithms, we consider offline estimators for the expected\nreward from a counterfactual policy. Our estimators are shown to have lowest\nvariance in a wide class of estimators, achieving variance reduction relative\nto standard estimators. We then apply our estimators to improve advertisement\ndesign by a major advertisement company. Consistent with the theoretical\nresult, our estimators allow us to improve on the existing bandit algorithm\nwith more statistical confidence compared to a state-of-the-art benchmark.","url_abs":"http://arxiv.org/abs/1809.03084v3","url_pdf":"http://arxiv.org/pdf/1809.03084v3.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":[],"tasks":[{"task_slug":"causal-inference","task_name":"Causal Inference"},{"task_slug":"off-policy-evaluation","task_name":"Off-policy evaluation"},{"task_slug":"visual-object-tracking","task_name":"Visual Object Tracking"},{"task_slug":null,"task_name":"counterfactual"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/causal-inference-on-idhp","task":"Causal Inference","dataset":"IDHP","model":"","rank_in_archive_order":3,"of":3,"metrics":{"Average Treatment Effect Error":"-0.225"},"uses_additional_data":false},{"leaderboard":"/sota/visual-object-tracking-on-vot2014","task":"Visual Object Tracking","dataset":"VOT2014","model":"","rank_in_archive_order":1,"of":1,"metrics":{"Expected Average Overlap (EAO)":"1.047"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1809.03084","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}