Presentation 2007-03-16
A learning algorithm of multi-bit Binary Neural Networks and its application to recognition systems
SATOSHI Murayama, Hidehiro NAKANO, Arata MIYAUCHI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) A geometrical learning has been proposed as a learning method of BNNs. However, geometrical learning is the learning method basically for 1-bit output BNN. That is, the study on learning method of multi-bit BNNs is not sufficient so far. In this paper, we propose a GA-based learning algorithm of multi-bit BNNs. The proposal method use a GA for evaluation of hidden neuron parameters of multi-bit BNNs. We consider an evaluation method of individuals of GA in order to suppress increase of redundant hidden neurons. In the numerical experiments, we evaluate the number of hidden neurons and recognition performance for multi-class classification problem.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Binary Neural Networks / Genetic algorithm / Supervised learning / Geometrical learning
Paper # NC2006-212
Date of Issue

Conference Information
Committee NC
Conference Date 2007/3/9(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 learning algorithm of multi-bit Binary Neural Networks and its application to recognition systems
Sub Title (in English)
Keyword(1) Binary Neural Networks
Keyword(2) Genetic algorithm
Keyword(3) Supervised learning
Keyword(4) Geometrical learning
1st Author's Name SATOSHI Murayama
1st Author's Affiliation Musashi Institute of Technology()
2nd Author's Name Hidehiro NAKANO
2nd Author's Affiliation Musashi Institute of Technology
3rd Author's Name Arata MIYAUCHI
3rd Author's Affiliation Musashi Institute of Technology
Date 2007-03-16
Paper # NC2006-212
Volume (vol) vol.106
Number (no) 590
Page pp.pp.-
#Pages 6
Date of Issue