Presentation 2004/5/21
A Forward-propagation Learning Rule in Consideration of the Correlation of Propagated-error
Yoshihiro OHAMA, Naohiro FUKUMURA, Yoji UNO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We have proposed that an inverse model can be acquired in an artificial multi-layered neural network using a forward-propagation (FP) rule but the algorithm of this rule is complex. In this research, FP rule is regarded as a method of the maximum likelihood estimation. A gradient of its logarithmic likelihood function includes the correlation of forward-propagated signals. It is confirmed by computer simulation that a newly devised FP rule achives locally convergence from random states of a multi-layered neural network in contrast to the former FP rule.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) forward-propagation rule / inverse model / motor learning / learning algorithm / maximum likelihood estimation
Paper # NC2004-4
Date of Issue

Conference Information
Committee NC
Conference Date 2004/5/21(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 Forward-propagation Learning Rule in Consideration of the Correlation of Propagated-error
Sub Title (in English)
Keyword(1) forward-propagation rule
Keyword(2) inverse model
Keyword(3) motor learning
Keyword(4) learning algorithm
Keyword(5) maximum likelihood estimation
1st Author's Name Yoshihiro OHAMA
1st Author's Affiliation Department of Information and Computer Sciences, Toyohashi University of Technology()
2nd Author's Name Naohiro FUKUMURA
2nd Author's Affiliation Department of Information and Computer Sciences, Toyohashi University of Technology
3rd Author's Name Yoji UNO
3rd Author's Affiliation Department of Information and Computer Sciences, Toyohashi University of Technology
Date 2004/5/21
Paper # NC2004-4
Volume (vol) vol.104
Number (no) 99
Page pp.pp.-
#Pages 6
Date of Issue