{"external_id":{"arxiv":["2109.06234"]},"citation":{"ieee":"A. Tornede, V. Bengs, and E. Hüllermeier, “Machine Learning for Online Algorithm Selection under Censored Feedback,” <i>Proceedings of the 36th AAAI Conference on Artificial Intelligence</i>. AAAI, 2022.","short":"A. Tornede, V. Bengs, E. Hüllermeier, Proceedings of the 36th AAAI Conference on Artificial Intelligence (2022).","bibtex":"@article{Tornede_Bengs_Hüllermeier_2022, title={Machine Learning for Online Algorithm Selection under Censored Feedback}, journal={Proceedings of the 36th AAAI Conference on Artificial Intelligence}, publisher={AAAI}, author={Tornede, Alexander and Bengs, Viktor and Hüllermeier, Eyke}, year={2022} }","chicago":"Tornede, Alexander, Viktor Bengs, and Eyke Hüllermeier. “Machine Learning for Online Algorithm Selection under Censored Feedback.” <i>Proceedings of the 36th AAAI Conference on Artificial Intelligence</i>. AAAI, 2022.","apa":"Tornede, A., Bengs, V., &#38; Hüllermeier, E. (2022). Machine Learning for Online Algorithm Selection under Censored Feedback. In <i>Proceedings of the 36th AAAI Conference on Artificial Intelligence</i>. AAAI.","mla":"Tornede, Alexander, et al. “Machine Learning for Online Algorithm Selection under Censored Feedback.” <i>Proceedings of the 36th AAAI Conference on Artificial Intelligence</i>, AAAI, 2022.","ama":"Tornede A, Bengs V, Hüllermeier E. Machine Learning for Online Algorithm Selection under Censored Feedback. <i>Proceedings of the 36th AAAI Conference on Artificial Intelligence</i>. Published online 2022."},"author":[{"first_name":"Alexander","id":"38209","last_name":"Tornede","full_name":"Tornede, Alexander"},{"full_name":"Bengs, Viktor","id":"76599","first_name":"Viktor","last_name":"Bengs"},{"full_name":"Hüllermeier, Eyke","first_name":"Eyke","id":"48129","last_name":"Hüllermeier"}],"abstract":[{"lang":"eng","text":"In online algorithm selection (OAS), instances of an algorithmic problem\r\nclass are presented to an agent one after another, and the agent has to quickly\r\nselect a presumably best algorithm from a fixed set of candidate algorithms.\r\nFor decision problems such as satisfiability (SAT), quality typically refers to\r\nthe algorithm's runtime. As the latter is known to exhibit a heavy-tail\r\ndistribution, an algorithm is normally stopped when exceeding a predefined\r\nupper time limit. As a consequence, machine learning methods used to optimize\r\nan algorithm selection strategy in a data-driven manner need to deal with\r\nright-censored samples, a problem that has received little attention in the\r\nliterature so far. In this work, we revisit multi-armed bandit algorithms for\r\nOAS and discuss their capability of dealing with the problem. Moreover, we\r\nadapt them towards runtime-oriented losses, allowing for partially censored\r\ndata while keeping a space- and time-complexity independent of the time\r\nhorizon. In an extensive experimental evaluation on an adapted version of the\r\nASlib benchmark, we demonstrate that theoretically well-founded methods based\r\non Thompson sampling perform specifically strong and improve in comparison to\r\nexisting methods."}],"project":[{"_id":"1","name":"SFB 901: SFB 901"},{"name":"SFB 901 - B: SFB 901 - Project Area B","_id":"3"},{"_id":"10","name":"SFB 901 - B2: SFB 901 - Subproject B2"}],"status":"public","_id":"30867","publisher":"AAAI","publication":"Proceedings of the 36th AAAI Conference on Artificial Intelligence","user_id":"38209","type":"preprint","date_updated":"2022-08-24T12:44:27Z","language":[{"iso":"eng"}],"date_created":"2022-04-12T11:58:56Z","year":"2022","department":[{"_id":"34"},{"_id":"7"},{"_id":"26"}],"title":"Machine Learning for Online Algorithm Selection under Censored Feedback"}