Presentation 1994/3/25
Error analysis and training data arrangement for neural networks
Kenji Fukumizu, Sumio Watanabe,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) This paper discusses the error of function approximation neural networks,when they approximate an unknown system which can give a response to an arbitrary input,such as in some system identification problems.In this case,we can choose an input distribution to obtain training data different from the one giving input data in the actual environment.Applying the statistical asymptotic theory,we elucidate which input distribution for training minimizes the mean squared error in the actual environment.Based on this analysis a novel algorithm to obtain training data is proposed,and its effect is verified by computer simulations.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) neural network / parametric estimation / statistical learning / statistical asymptotic theory
Paper # NC93-126
Date of Issue

Conference Information
Committee NC
Conference Date 1994/3/25(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) Error analysis and training data arrangement for neural networks
Sub Title (in English)
Keyword(1) neural network
Keyword(2) parametric estimation
Keyword(3) statistical learning
Keyword(4) statistical asymptotic theory
1st Author's Name Kenji Fukumizu
1st Author's Affiliation Information and Communication Research and Development Center, Ricoh Co.,Ltd.()
2nd Author's Name Sumio Watanabe
2nd Author's Affiliation Information and Communication Research and Development Center, Ricoh Co.,Ltd.
Date 1994/3/25
Paper # NC93-126
Volume (vol) vol.93
Number (no) 537
Page pp.pp.-
#Pages 8
Date of Issue