Presentation 2009-01-19
Reinforcement Learning Using Selective Desensitization Neural Networks in the State Space with Redundant Dimensions
Tomoyuki SHIMBO, Ken YAMANE, Masahiko MORITA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Reinforcement learning has a problem that it requires a long time particularly when the state space is high dimensional with redundant dimensions. Here we report that a function approximator comprised of the selective desensitization neural network (SDNN) improves in the efficiency of reinforcement learning in the acrobot swing-up task, avoiding the explosive increase in learning time and computational costs when redundant variables are added.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Reinforcement Learning / Function Approximator / Selective Desensitization / Redundant Dimensions / Acrobot
Paper # NC2008-83
Date of Issue

Conference Information
Committee NC
Conference Date 2009/1/12(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Neurocomputing (NC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Reinforcement Learning Using Selective Desensitization Neural Networks in the State Space with Redundant Dimensions
Sub Title (in English)
Keyword(1) Reinforcement Learning
Keyword(2) Function Approximator
Keyword(3) Selective Desensitization
Keyword(4) Redundant Dimensions
Keyword(5) Acrobot
1st Author's Name Tomoyuki SHIMBO
1st Author's Affiliation Graduate School of Systems and Information Engineering, University of Tsukuba()
2nd Author's Name Ken YAMANE
2nd Author's Affiliation Graduate School of Systems and Information Engineering, University of Tsukuba
3rd Author's Name Masahiko MORITA
3rd Author's Affiliation Graduate School of Systems and Information Engineering, University of Tsukuba
Date 2009-01-19
Paper # NC2008-83
Volume (vol) vol.108
Number (no) 383
Page pp.pp.-
#Pages 6
Date of Issue