Presentation 2001/3/16
A study for the acquisition of reinforcement learning procedure in navigation problems
Akitoshi Ogawa, Takashi Omori,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this paper, we propose a hypothesis that explains an evolution of animal intelligence seen from information processing. In the hypothesis, we assume asset of neural processing functions and a searching system, and expect thinking processes and action procedures are acquired in animals to execute a task when the function set is changed. To verify this hypothesis, we prepare two agents that have different structures. One acquires Q-learning procedure, and the other acquires prediction-based strategy when an environment model is added to the function set. By compare them, we show that the latter can acquires more powerful search algorithm than the former. VVe also propose a procedure representation model in brain like system to realize the procedure acquisition task.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Intelligence Evolution / Procedure Acquisition / Navigation / Reinforcement Learning / Attention System
Paper # NC2000-167
Date of Issue

Conference Information
Committee NC
Conference Date 2001/3/16(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) A study for the acquisition of reinforcement learning procedure in navigation problems
Sub Title (in English)
Keyword(1) Intelligence Evolution
Keyword(2) Procedure Acquisition
Keyword(3) Navigation
Keyword(4) Reinforcement Learning
Keyword(5) Attention System
1st Author's Name Akitoshi Ogawa
1st Author's Affiliation Hokkaido University()
2nd Author's Name Takashi Omori
2nd Author's Affiliation Hokkaido University
Date 2001/3/16
Paper # NC2000-167
Volume (vol) vol.100
Number (no) 688
Page pp.pp.-
#Pages 8
Date of Issue