Presentation 1993/9/22
Evaluation of the Capability of multilayer perceptron by its geometrical properties
Tetsuya Matsumoto, Jun-ichiro Toriwaki, Tatsuhiro Yonekura,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Multilayer perceptron(MLP)can be regarded as the mapping from the input layer to the output layer.From this point of view,MLP learning is a process of searching the ″optimai″ mapping from the input space to the output space.However,the meaning of optimality and features of the generalization have not been made clear in various application areas such as pattern recognition.In this paper we propose a new procedure for evaluating the capability of MLP and show that the ″total curvature″ of the hypersurface spann ed by the output vectors of MLP is an efficient measure of the capability.We justify the procedure by simulation experiments to evaluate the capability of several kinds of MLP using total curvature as the evaluation functional.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Multilayer perceptron / Generalization / Curvature / Ill-posed problem / Geometrical Property
Paper # NC93-38
Date of Issue

Conference Information
Committee NC
Conference Date 1993/9/22(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 Neurocomputing (NC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Evaluation of the Capability of multilayer perceptron by its geometrical properties
Sub Title (in English)
Keyword(1) Multilayer perceptron
Keyword(2) Generalization
Keyword(3) Curvature
Keyword(4) Ill-posed problem
Keyword(5) Geometrical Property
1st Author's Name Tetsuya Matsumoto
1st Author's Affiliation Education center for information processing,Nagoya university()
2nd Author's Name Jun-ichiro Toriwaki
2nd Author's Affiliation Faculty of engineering,Nagoya university
3rd Author's Name Tatsuhiro Yonekura
3rd Author's Affiliation Faculty of engineering,Ibaraki university
Date 1993/9/22
Paper # NC93-38
Volume (vol) vol.93
Number (no) 247
Page pp.pp.-
#Pages 8
Date of Issue