Presentation 2008-06-27
GA-based geometrical learning of binary neural networks and its generalization capability
Syuhei SHIMADA, Hidehiro NAKANO, Arata MIYAUCHI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) GA-based geometrical learning of Binary Neural Network (BNN) is discussed in this paper. To apply BNN to the multi class separation problems, we propose two learning methods for multi bit output BNN. The first method learns the teacher signals corresponding to each class in parallel. The second method learns them seqencially. Performing numerical experiments for a basic multi class separation problem, the leaning performances of the generalization capability and the number of hidden-layer neurons in each method are compared.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Binary Neural Network / Genetic algorithm / Supervised learning / Geometrical learning / Generalization capability
Paper # NC2008-27
Date of Issue

Conference Information
Committee NC
Conference Date 2008/6/19(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) GA-based geometrical learning of binary neural networks and its generalization capability
Sub Title (in English)
Keyword(1) Binary Neural Network
Keyword(2) Genetic algorithm
Keyword(3) Supervised learning
Keyword(4) Geometrical learning
Keyword(5) Generalization capability
1st Author's Name Syuhei SHIMADA
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 2008-06-27
Paper # NC2008-27
Volume (vol) vol.108
Number (no) 101
Page pp.pp.-
#Pages 6
Date of Issue