Presentation 2001/4/6
A weight minimization approach to fault tolerant multi-layer neural networks
Haruhiko TAKASE, Hidehiko KITA, Terumine HAYASHI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We propose a new learning algorithm to enhance fault tolerance of multi-layer neural networks (MLN). This method is based on the idea that strong weights make MLN sensitive to faults. To decrease the number of strong weights, we introduce new evaluation function for new learning algorithm. It consists of not only the output error but also the sum of square of weights. With the new evaluation function the learning algorithm works not only minimize output error but also minimize weights. The value of parameter to balance effects of these two terms is decided actively during MLN's training. Next, to perform the proposed method, we apply it to pattern recognition problems. As a result, it is shown that miss recognition ratio and the activity of hidden units are improved.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Multi-layer neural networks / Fault tolerance / BP algorithm
Paper # CPSY2001-2,FTS2001-2
Date of Issue

Conference Information
Committee CPSY
Conference Date 2001/4/6(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 Computer Systems (CPSY)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) A weight minimization approach to fault tolerant multi-layer neural networks
Sub Title (in English)
Keyword(1) Multi-layer neural networks
Keyword(2) Fault tolerance
Keyword(3) BP algorithm
1st Author's Name Haruhiko TAKASE
1st Author's Affiliation Department of Electrical and Electronic Engineering, Faculty of Engineering, Mie University()
2nd Author's Name Hidehiko KITA
2nd Author's Affiliation Department of Electrical and Electronic Engineering, Faculty of Engineering, Mie University
3rd Author's Name Terumine HAYASHI
3rd Author's Affiliation Department of Electrical and Electronic Engineering, Faculty of Engineering, Mie University
Date 2001/4/6
Paper # CPSY2001-2,FTS2001-2
Volume (vol) vol.101
Number (no) 2
Page pp.pp.-
#Pages 7
Date of Issue