ResearchSpace

What good are actions? Accelerating learning using learned action priors

Show simple item record

dc.contributor.author Rosman, Benjamin S
dc.contributor.author Ramamoorthy, S
dc.date.accessioned 2013-01-28T08:48:42Z
dc.date.available 2013-01-28T08:48:42Z
dc.date.issued 2012-11
dc.identifier.citation Rosman, B.S. and Ramamoorthy, S. 2012. What good are actions? Accelerating learning using learned action priors. IEEE Conference on Development and Learning and Epigenetic Robotics (ICDL-EpiRob 2012), San Diego, California, USA, 7-9 November 2012 en_US
dc.identifier.uri http://homepages.inf.ed.ac.uk/s0896970/papers/icdl12.pdf
dc.identifier.uri http://hdl.handle.net/10204/6475
dc.description IEEE Conference on Development and Learning and Epigenetic Robotics (ICDL-EpiRob 2012), San Diego, California, USA, 7-9 November 2012 en_US
dc.description.abstract The computational complexity of learning in sequential decision problems grows exponentially with the number of actions available to the agent at each state. We present a method for accelerating this process by learning action priors that express the usefulness of each action in each state. These are learned from a set of different optimal policies from many tasks in the same state space, and are used to bias exploration away from less useful actions. This is shown to improve performance for tasks in the same domain but with different goals. We extend our method to base action priors on perceptual cues rather than absolute states, allowing the transfer of these priors between tasks with differing state spaces and transition functions, and demonstrate experimentally the advantages of learning with action priors in a reinforcement learning context. en_US
dc.language.iso en en_US
dc.relation.ispartofseries Workflow;9988
dc.subject Artificial intelligence en_US
dc.subject Learning actions en_US
dc.subject Markov Decision Process en_US
dc.subject MDP en_US
dc.title What good are actions? Accelerating learning using learned action priors en_US
dc.type Conference Presentation en_US
dc.identifier.apacitation Rosman, B. S., & Ramamoorthy, S. (2012). What good are actions? Accelerating learning using learned action priors. http://hdl.handle.net/10204/6475 en_ZA
dc.identifier.chicagocitation Rosman, Benjamin S, and S Ramamoorthy. "What good are actions? Accelerating learning using learned action priors." (2012): http://hdl.handle.net/10204/6475 en_ZA
dc.identifier.vancouvercitation Rosman BS, Ramamoorthy S, What good are actions? Accelerating learning using learned action priors; 2012. http://hdl.handle.net/10204/6475 . en_ZA
dc.identifier.ris TY - Conference Presentation AU - Rosman, Benjamin S AU - Ramamoorthy, S AB - The computational complexity of learning in sequential decision problems grows exponentially with the number of actions available to the agent at each state. We present a method for accelerating this process by learning action priors that express the usefulness of each action in each state. These are learned from a set of different optimal policies from many tasks in the same state space, and are used to bias exploration away from less useful actions. This is shown to improve performance for tasks in the same domain but with different goals. We extend our method to base action priors on perceptual cues rather than absolute states, allowing the transfer of these priors between tasks with differing state spaces and transition functions, and demonstrate experimentally the advantages of learning with action priors in a reinforcement learning context. DA - 2012-11 DB - ResearchSpace DP - CSIR KW - Artificial intelligence KW - Learning actions KW - Markov Decision Process KW - MDP LK - https://researchspace.csir.co.za PY - 2012 T1 - What good are actions? Accelerating learning using learned action priors TI - What good are actions? Accelerating learning using learned action priors UR - http://hdl.handle.net/10204/6475 ER - en_ZA


Files in this item

This item appears in the following Collection(s)

Show simple item record