Presentation | 2011-06-24 Q-learning in Continuous State-Action Space by Using a Selective Desensitization Neural Network Takaaki KOBAYASHI, Takeshi SHIBUYA, Fumihide TANAKA, Masahiko MORITA, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Value function approximation takes an important role for reinforcement learning in continuous state-action space. Conventional methods such as radial basis function networks need considerable amount of computation in its learning as well as optimal action selection. This paper proposes a novel representation of the output layer of selective desensitization neural networks. By using the method, the efficiency of learning is increased and amount of computation is decreased. The effectiveness of proposed method is confirmed through computer simulation experiments using acrobot. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Q-learning / continuous state-action space / function approximation / selective desensitization neural networks |
Paper # | NC2011-15 |
Date of Issue |
Conference Information | |
Committee | NC |
---|---|
Conference Date | 2011/6/16(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) | Q-learning in Continuous State-Action Space by Using a Selective Desensitization Neural Network |
Sub Title (in English) | |
Keyword(1) | Q-learning |
Keyword(2) | continuous state-action space |
Keyword(3) | function approximation |
Keyword(4) | selective desensitization neural networks |
1st Author's Name | Takaaki KOBAYASHI |
1st Author's Affiliation | Graduate School of System and Information Engineering, University of Tsukuba() |
2nd Author's Name | Takeshi SHIBUYA |
2nd Author's Affiliation | Graduate School of System and Information Engineering, University of Tsukuba |
3rd Author's Name | Fumihide TANAKA |
3rd Author's Affiliation | Graduate School of System and Information Engineering, University of Tsukuba |
4th Author's Name | Masahiko MORITA |
4th Author's Affiliation | Graduate School of System and Information Engineering, University of Tsukuba |
Date | 2011-06-24 |
Paper # | NC2011-15 |
Volume (vol) | vol.111 |
Number (no) | 96 |
Page | pp.pp.- |
#Pages | 5 |
Date of Issue |