{"type":"ScholarlyArticle","title":"Using Predictive Representations to Improve Generalization in Reinforcement Learning","authors":[{"name":"Eddie J. Rafols","slug":"eddie-j-rafols","given":"Eddie J.","family":"Rafols","affiliation":""},{"name":"Mark B. Ring","slug":"mark-b-ring","given":"Mark B.","family":"Ring","affiliation":""},{"name":"Richard S. Sutton","slug":"richard-s-sutton","given":"Richard S.","family":"Sutton","affiliation":""},{"name":"Brian Tanner","slug":"brian-tanner","given":"Brian","family":"Tanner","affiliation":""}],"abstract":"The predictive representations hypothesis holds that particularly good generalization will result from representing the state of the world in terms of predictions about possible future experience. This hypothesis has been a central motivation behind recent research in, for example, PSRs and TD networks. In this paper we present the first explicit investigation of this hypothesis. We show in a reinforcement-learning example (a grid-world navigation task) that a predictive representation in tabular form can learn much faster than both the tabular explicit-state representation and a tabular history-based method.","doi":null,"url":"https://ijcai.org/proceedings/2005/227","pdf":"https://ijcai.org/proceedings/2005/0227.pdf","markdown":"https://ijcai.org/proceedings/2005/227.md","bibtex":"https://ijcai.org/proceedings/2005/bibtex/227","volume":{"slug":"2005","title":"Proceedings of the Nineteenth International Joint Conference on Artificial Intelligence","short_title":"IJCAI 2005","url":"https://ijcai.org/proceedings/2005/","year":2005,"location":"Edinburgh, Scotland, UK","dates":"","editors":"","isbn":"","publisher":"International Joint Conferences on Artificial Intelligence","doi_prefix":""},"section":null,"track":"","number":227,"pages":{"first":835,"last":839},"keywords":[],"videos":[],"references":[],"license_note":"Copyright International Joint Conferences on Artificial Intelligence."}