Presentation 1998/3/19
A Neural Network Model Which Obtains Active Perception using Reinforcement Learning
Mitsuo Takano, Yutaka Sakaguchi,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We human beings recognize objects through processing only appropriate sensory information dependent on our purpose and context. How do we obtain such active perception function ? To answer this question, the authors tried to construct a neural network model using Temporal Difference (TD) learning, one of the reinforcement algorithms. This model learns how to choose sensory information for getting recognition result efficiently. The result of computer simulation showed that the model obtained such an ability through trial and error. The authors also discussed future approaches to expand the proposed model to a more general model of human recognition mechanism.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) active perception / neural network model / reinforcement learning / TD learning / attention
Paper #
Date of Issue

Conference Information
Committee NC
Conference Date 1998/3/19(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) A Neural Network Model Which Obtains Active Perception using Reinforcement Learning
Sub Title (in English)
Keyword(1) active perception
Keyword(2) neural network model
Keyword(3) reinforcement learning
Keyword(4) TD learning
Keyword(5) attention
1st Author's Name Mitsuo Takano
1st Author's Affiliation Graduate School of Information Systems, University of Electro-Communications()
2nd Author's Name Yutaka Sakaguchi
2nd Author's Affiliation Graduate School of Information Systems, University of Electro-Communications
Date 1998/3/19
Paper #
Volume (vol) vol.97
Number (no) 623
Page pp.pp.-
#Pages 6
Date of Issue