Presentation 2003/9/22
Local Linearized Least Squares Algorithm based on Penalty Function Method
Masahiro YOSHIDA, Hiroshi NINOMIYA, Hideki ASAI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this research, we introduce the local training method for feedforward neural networks. The local trainingmethods improve the computational complexity compared to the global training methods. Our method also can improve thecomputational complexity, because of learning for each neuron in neural networks. In addition, the proposed training method yields results with high convergence rates by using penalty term which is derived from Hessian matrix of the cost functionwith respect to weight.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Feedforward Neural Networks / Local Training / Penalty Function / Recursive Least Squares
Paper # MLP2003-63
Date of Issue

Conference Information
Committee NLP
Conference Date 2003/9/22(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 Nonlinear Problems (NLP)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Local Linearized Least Squares Algorithm based on Penalty Function Method
Sub Title (in English)
Keyword(1) Feedforward Neural Networks
Keyword(2) Local Training
Keyword(3) Penalty Function
Keyword(4) Recursive Least Squares
1st Author's Name Masahiro YOSHIDA
1st Author's Affiliation Department of Systems Engineering, Shizuoka University()
2nd Author's Name Hiroshi NINOMIYA
2nd Author's Affiliation Dept. of Inf. Science, Shonan Institute of Technology
3rd Author's Name Hideki ASAI
3rd Author's Affiliation Department of Systems Engineering, Shizuoka University
Date 2003/9/22
Paper # MLP2003-63
Volume (vol) vol.103
Number (no) 335
Page pp.pp.-
#Pages 6
Date of Issue