{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,8,7]],"date-time":"2025-08-07T09:13:34Z","timestamp":1754558014459},"publisher-location":"California","reference-count":0,"publisher":"International Joint Conferences on Artificial Intelligence Organization","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2021,8]]},"abstract":"<jats:p>Formally verifying that reinforcement learning systems act\n\nsafely is increasingly important, but existing methods\n\nonly verify over finite time.\n\nThis is of limited use for dynamical systems that run indefinitely.\n\nWe introduce the first method for verifying the time-unbounded\n\nsafety of neural networks controlling dynamical systems.\n\nWe develop a novel abstract interpretation method which,\n\nby constructing adaptable template-based polyhedra using MILP and interval\n\narithmetic, yields sound---safe and invariant---overapproximations\n\nof the reach set.\n\nThis provides stronger safety guarantees\n\nthan previous time-bounded methods and shows whether\n\nthe agent has generalised beyond the length of its training episodes.\n\nOur method supports ReLU activation functions\n\nand systems with linear, piecewise linear and non-linear dynamics\n\ndefined with polynomial and transcendental functions.\n\nWe demonstrate its efficacy on a range of benchmark control problems.<\/jats:p>","DOI":"10.24963\/ijcai.2021\/297","type":"proceedings-article","created":{"date-parts":[[2021,8,11]],"date-time":"2021-08-11T11:00:49Z","timestamp":1628679649000},"page":"2154-2160","source":"Crossref","is-referenced-by-count":15,"title":["Verifying Reinforcement Learning up to Infinity"],"prefix":"10.24963","author":[{"given":"Edoardo","family":"Bacci","sequence":"first","affiliation":[{"name":"University of Birmingham"}]},{"given":"Mirco","family":"Giacobbe","sequence":"additional","affiliation":[{"name":"University of Oxford"}]},{"given":"David","family":"Parker","sequence":"additional","affiliation":[{"name":"University of Birmingham"}]}],"member":"10584","event":{"number":"30","sponsor":["International Joint Conferences on Artificial Intelligence Organization (IJCAI)"],"acronym":"IJCAI-2021","name":"Thirtieth International Joint Conference on Artificial Intelligence {IJCAI-21}","start":{"date-parts":[[2021,8,19]]},"theme":"Artificial Intelligence","location":"Montreal, Canada","end":{"date-parts":[[2021,8,27]]}},"container-title":["Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence"],"original-title":[],"deposited":{"date-parts":[[2021,8,11]],"date-time":"2021-08-11T11:02:29Z","timestamp":1628679749000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.ijcai.org\/proceedings\/2021\/297"}},"subtitle":[],"proceedings-subject":"Artificial Intelligence Research Articles","short-title":[],"issued":{"date-parts":[[2021,8]]},"references-count":0,"URL":"https:\/\/doi.org\/10.24963\/ijcai.2021\/297","relation":{},"subject":[],"published":{"date-parts":[[2021,8]]}}}