Presentation 2004/1/22
A Study on the State Space Reconfiguration in the Agent's Action Space
Naoya ARIFUKU, Yoshinobu KAJIKAWA, Yasuo NOMURA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Agent learning in the environment where agent's position information is continuous values needs much time if the position values are used as state recognition of the agents. Hence, the agent's programers have divided the position information into arbitrary discrete states in advance. However, it may not guarantee that the number of state space is optimum for agents. This paper therefore proposes dynamic state space generation from state space divided in advance by using the Cluster Analysis. At the same time, the timing of state space generation should be determined by information obtainable during actions. We therefore use the Redisual Entropy evolving action uncertainty.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Reinforcement learning / Multivariate analysis / State space reconfiguration / Labyrinth problem
Paper # AI2003-65
Date of Issue

Conference Information
Committee AI
Conference Date 2004/1/22(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 Study on the State Space Reconfiguration in the Agent's Action Space
Sub Title (in English)
Keyword(1) Reinforcement learning
Keyword(2) Multivariate analysis
Keyword(3) State space reconfiguration
Keyword(4) Labyrinth problem
1st Author's Name Naoya ARIFUKU
1st Author's Affiliation Faculty of Engineering, Kansai University()
2nd Author's Name Yoshinobu KAJIKAWA
2nd Author's Affiliation Faculty of Engineering, Kansai University:Frontier Science Center, Kansai University
3rd Author's Name Yasuo NOMURA
3rd Author's Affiliation Faculty of Engineering, Kansai University
Date 2004/1/22
Paper # AI2003-65
Volume (vol) vol.103
Number (no) 623
Page pp.pp.-
#Pages 6
Date of Issue