Presentation 1994/3/25
AUTO-SELECTION OF ACTIVATION FUNCTIONS IN MULTILAYER NEURAL NETWORKS APPLIED TO PATTERN CLASSIFICATION
Yoshinori Kimura, Kenji Nakayama,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Minimization of hidden units by auto-selecting optimum activation functions is discussed for multilayer neural networks applied to pattern classification.Three typical functions, including sigmoid,sinusoidal and Gaussian functions are employed. They have unique property for space division.Thus,a wide range of problems can be solved by combining these functions.Three functions are used in a hidden layer.One hidden unit has one of three functions.Efficiency of each function is evaluated using three criteria,similar to the conventional.The useful functions are selected in the learning process.Parity problem,counting 1 in bit-patterns and other two-dimensional classification problems can be solved with the minimum hidden units.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Multilayer neural network / Activation functions / Pattern classification / Minimum realization / Optimization
Paper # NC93-116
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 ENG
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) AUTO-SELECTION OF ACTIVATION FUNCTIONS IN MULTILAYER NEURAL NETWORKS APPLIED TO PATTERN CLASSIFICATION
Sub Title (in English)
Keyword(1) Multilayer neural network
Keyword(2) Activation functions
Keyword(3) Pattern classification
Keyword(4) Minimum realization
Keyword(5) Optimization
1st Author's Name Yoshinori Kimura
1st Author's Affiliation Department of Electrical and Computer Engineering,Faculty of Technology,Kanazawa University()
2nd Author's Name Kenji Nakayama
2nd Author's Affiliation Department of Electrical and Computer Engineering,Faculty of Technology,Kanazawa University
Date 1994/3/25
Paper # NC93-116
Volume (vol) vol.93
Number (no) 537
Page pp.pp.-
#Pages 8
Date of Issue