Presentation 1994/5/19
Speedup of Learning in 3-layer Neural Networks using Second-order Method
Kazumi Saito, Ryohei Nakano,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) By using nonlinear optimization techniques such as quasi-Newton methods or conjugate gradient methods,speedup of learning in three- layer neural networks has been expected.However,since existing methods employ heavy direct searches in order to calculate the optimal step lengths,their improvement has not been remarkable.In this paper,we propose two methods:by calculating the optimal step lengths as the minimal points of the second-order approximations, they can perform one iteration efficiently.In the experiments,the proposed methods worked much better than the existing methods.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) 3-layer-neural networks / back propagation / quesi-Newton′s meth od / Conjugate gradient method
Paper # NC94-7
Date of Issue

Conference Information
Committee NC
Conference Date 1994/5/19(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) Speedup of Learning in 3-layer Neural Networks using Second-order Method
Sub Title (in English)
Keyword(1) 3-layer-neural networks
Keyword(2) back propagation
Keyword(3) quesi-Newton′s meth od
Keyword(4) Conjugate gradient method
1st Author's Name Kazumi Saito
1st Author's Affiliation NTT Communication Science Laboratories()
2nd Author's Name Ryohei Nakano
2nd Author's Affiliation NTT Communication Science Laboratories
Date 1994/5/19
Paper # NC94-7
Volume (vol) vol.94
Number (no) 40
Page pp.pp.-
#Pages 8
Date of Issue