Presentation 2000/6/15
NLP2000-22 / NC2000-16 An Upper Bound on the Size of Min-Max Modular Neural Networks Trainable in Polynomial Time
Bao-Liang Lu, Michinori Ichikawa,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this paper we present an upper bound on the size of milt-max modular (M^3) neural networks that can be trained in polynomial time for nonlinearly separable problems. Given an arbitrary training set with L training data for a K-class (K>2) pattern classification problem, we show that an M^3 network with [numerical formula] neurons, [numerical formula] MIN units, [numerical formula] MAX units. and ([numerical formula]) INV units, can exactly implement the training set in polynomial time without, learning, where L_i is the number of training data for class C_i for i = 1, …, K. We also show that this M^3 network can perform minimum-distance classifications with respect to finite training data.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Modular neurai network / upper bound on network size / minimum-distance classifer / piece wise linear machine
Paper # NLP2000-22,NC2000-16
Date of Issue

Conference Information
Committee NLP
Conference Date 2000/6/15(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 ENG
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) NLP2000-22 / NC2000-16 An Upper Bound on the Size of Min-Max Modular Neural Networks Trainable in Polynomial Time
Sub Title (in English)
Keyword(1) Modular neurai network
Keyword(2) upper bound on network size
Keyword(3) minimum-distance classifer
Keyword(4) piece wise linear machine
1st Author's Name Bao-Liang Lu
1st Author's Affiliation Lab.for Brain-Operative Device, RIKEN Brain Science Institute()
2nd Author's Name Michinori Ichikawa
2nd Author's Affiliation Lab.for Brain-Operative Device, RIKEN Brain Science Institute
Date 2000/6/15
Paper # NLP2000-22,NC2000-16
Volume (vol) vol.100
Number (no) 124
Page pp.pp.-
#Pages 8
Date of Issue