{"pk":27023,"title":"Learning to reinforcement learn","subtitle":null,"abstract":"In recent years deep reinforcement learning (RL) systems have attained superhuman performance in a number ofchallenging task domains, but are constrained by a demand for large training sets. A critical present objective is thus to developdeep RL methods that can adapt rapidly to new tasks. In the present work we introduce a novel approach to this challenge,which we refer to as deep meta-reinforcement learning. Previous work has shown that recurrent networks can support meta-learning in a fully supervised context. We extend this approach to the RL setting. What emerges is a system that is trainedusing one RL algorithm, but whose recurrent dynamics implement a second, quite separate RL procedure. This second, learnedRL algorithm can differ from the original one in arbitrary ways and exploit structure in the training domain. We unpack thesepoints in five proof-of-concept experiments to examine key aspects of deep meta-RL.","language":"eng","license":{"name":"","short_name":"","text":null,"url":""},"keywords":[],"section":"Talks: Papers","is_remote":true,"remote_url":"https://escholarship.org/uc/item/1tn6q2t7","frozenauthors":[{"first_name":"Jane","middle_name":"","last_name":"Wang","name_suffix":"","institution":"DeepMind","department":""},{"first_name":"Zeb","middle_name":"","last_name":"Kurth-Nelson","name_suffix":"","institution":"DeepMind","department":""},{"first_name":"Hubert","middle_name":"","last_name":"Soyer","name_suffix":"","institution":"DeepMind","department":""},{"first_name":"Joel","middle_name":"","last_name":"Leibo","name_suffix":"","institution":"DeepMind","department":""},{"first_name":"Dhruva","middle_name":"","last_name":"Tirumala","name_suffix":"","institution":"DeepMind","department":""},{"first_name":"Remi","middle_name":"","last_name":"Munos","name_suffix":"","institution":"DeepMind","department":""},{"first_name":"Charles","middle_name":"","last_name":"Blundell","name_suffix":"","institution":"DeepMind","department":""},{"first_name":"Dharshan","middle_name":"","last_name":"Kumaran","name_suffix":"","institution":"DeepMind","department":""},{"first_name":"Matt","middle_name":"","last_name":"Botvinick","name_suffix":"","institution":"DeepMind","department":""}],"date_submitted":null,"date_accepted":null,"date_published":"2017-01-01T21:00:00+03:00","render_galley":null,"galleys":[{"label":"PDF","type":"pdf","path":"https://journalpub.escholarship.org/cognitivesciencesociety/article/27023/galley/16659/download/"}]}