Presentation 2007-06-15
Learning of Neural Networks with D ichotomic Random Teacher Signals
Yoshifusa Ito, Cidambi Srinivasan, Hiroyuki Izumi,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Learning with dichotomic random teacher signals is a hard task for neural networks, because the learning cannot be completed by simply decreasing the distance between the output and teacher signals. The network meets more often the problems of local minima and overlearnings. To overcome this difficulty we have constructed a three layer neural network having hidden-layer units with less degree of freedom. It causes increment in the number of the units but the total number of constants to be trained can be decreased. The neural network is applied to learning of the Bayesian discriminant functions. Its performance was better than that of ordinary neural networks.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Dichotomic random variables / approximation / hidden-layer unit / Bayesian discriminant function / Layered neural netwoks
Paper # NC2007-21
Date of Issue

Conference Information
Committee NC
Conference Date 2007/6/7(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) Learning of Neural Networks with D ichotomic Random Teacher Signals
Sub Title (in English)
Keyword(1) Dichotomic random variables
Keyword(2) approximation
Keyword(3) hidden-layer unit
Keyword(4) Bayesian discriminant function
Keyword(5) Layered neural netwoks
1st Author's Name Yoshifusa Ito
1st Author's Affiliation Department of Information and Policy Studies Aichi-Gakuin University()
2nd Author's Name Cidambi Srinivasan
2nd Author's Affiliation Department of Statistics University of Kentucky
3rd Author's Name Hiroyuki Izumi
3rd Author's Affiliation Department of Information and Policy Studies Aichi-Gakuin University
Date 2007-06-15
Paper # NC2007-21
Volume (vol) vol.107
Number (no) 92
Page pp.pp.-
#Pages 6
Date of Issue