Presentation 2009-03-13
Reinforcement Learning with Internal Rewards Based on Error in a Grid-based Map
Yoshifumi TANAKA, Masumi ISHIKAWA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) The present paper proposes to make reinforcement learning efficient by using internal rewards based on curiosity in addition to external rewards at goal in a goal reaching task. Here, curiosity is defined by the decrease in prediction error, which is defined by the difference between grid-based map and the sensory information at each grid. Simulation experiments indicate that the performance of the proposed method is superior to the conventional reinforcement learning in terms of the number of goals reached and the number of actions needed to reach the goal in a transient state. How parameter values affect the performance and learning of the environment is also analyzed.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) reinforcement learning / grid-based map / curiosity / internal reward
Paper # NC2008-151
Date of Issue

Conference Information
Committee NC
Conference Date 2009/3/4(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 with Internal Rewards Based on Error in a Grid-based Map
Sub Title (in English)
Keyword(1) reinforcement learning
Keyword(2) grid-based map
Keyword(3) curiosity
Keyword(4) internal reward
1st Author's Name Yoshifumi TANAKA
1st Author's Affiliation Department of Brain Science and Engineering, Graduate School of Life Science & Engineering, Kyushu Institute of Technology()
2nd Author's Name Masumi ISHIKAWA
2nd Author's Affiliation Department of Brain Science and Engineering, Graduate School of Life Science & Engineering, Kyushu Institute of Technology
Date 2009-03-13
Paper # NC2008-151
Volume (vol) vol.108
Number (no) 480
Page pp.pp.-
#Pages 6
Date of Issue