Presentation 2004/1/19
An experiential learning algorithm for solving maze tasks with episodic memory (Neurocomputing)
Yoshito AOTA, Yoko YAMAGUCHI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) It is known that Catastrophic interference of learning is one of big problem in the field of learning theory. Although many researchers have tried to conquer this problem, it was imperfect. We consider that this problem is based on natural character of existing learning theories. Because existing methods use some kind of optimization process, sometime it causes interference when the same one situation has not one answer but some answers. In this paper, we propose context dependent action selection system for distinguishing correct situation with episodic memory. The experiential learning algorithm with episodic memory for association of and learning correct contexts is proposed and applied to maze tasks.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Catastrophic interference / episodic memory / learning theory / intelligence / maze task
Paper # NC2003-117
Date of Issue

Conference Information
Committee NC
Conference Date 2004/1/19(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) An experiential learning algorithm for solving maze tasks with episodic memory (Neurocomputing)
Sub Title (in English)
Keyword(1) Catastrophic interference
Keyword(2) episodic memory
Keyword(3) learning theory
Keyword(4) intelligence
Keyword(5) maze task
1st Author's Name Yoshito AOTA
1st Author's Affiliation Japan Science and Technology Agency (JST), CREST "Creating the Brain"()
2nd Author's Name Yoko YAMAGUCHI
2nd Author's Affiliation RIKEN, Brain Science Institute
Date 2004/1/19
Paper # NC2003-117
Volume (vol) vol.103
Number (no) 601
Page pp.pp.-
#Pages 6
Date of Issue