Presentation 2000/1/13
A recognization method of environmental change on reinforcement learning
Shinya Yamamoto, Fumihiko Yamaguchi, Hiroaki Saito, Masakazu Nakanishi,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) There are some methods that resolve problems of reinforcement learning in non Marokov Decision Process(non-MDP)environment on environment changes. The efficient method of recognizing environmental change has not yet been proposed. This paper proposes a method for recognizing environmental changes on Stochastic Gradient Ascent(SGA)which is a major learning engine in non-MDP environment. It uses the change of an internal variable W of SGA. Our method can be easily put in SGA and it is available for all SGA-applicable problems. We had a simulation to show the efficiency of our method and succeeded to reduce the recognition time to almost half of the conventional method.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) reinforcement learning / Stochastic Gradient Ascent / environmental change
Paper # AI99-81
Date of Issue

Conference Information
Committee AI
Conference Date 2000/1/13(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 Artificial Intelligence and Knowledge-Based Processing (AI)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) A recognization method of environmental change on reinforcement learning
Sub Title (in English)
Keyword(1) reinforcement learning
Keyword(2) Stochastic Gradient Ascent
Keyword(3) environmental change
1st Author's Name Shinya Yamamoto
1st Author's Affiliation Keio University()
2nd Author's Name Fumihiko Yamaguchi
2nd Author's Affiliation Keio University
3rd Author's Name Hiroaki Saito
3rd Author's Affiliation Keio University
4th Author's Name Masakazu Nakanishi
4th Author's Affiliation Keio University
Date 2000/1/13
Paper # AI99-81
Volume (vol) vol.99
Number (no) 534
Page pp.pp.-
#Pages 6
Date of Issue