Presentation 1993/11/24
On the novel method of designing a neural network for pattern classification.
Toshinori Yoshikawa, Satoshi Ookuma,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this report,we propose a 4-layered neural network and its learning algorithm for pattern classification,such as character recognition.The proposed network is basically equal to the 4- layered perceptron of which the output function of second layer units is changed.The network distributes the task of classification among the second layer units.Therefore,the network can make clear a role of the units,and the learning process becomes much faster than conventional one using the back- propagation method.In addition,the network obtains the classification boundary enough for pattern classification because the network assigns the task by making a hyperplane between the input pattern space.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) back-propagation method / output function / hyperplane / neural network / pattern classification
Paper # NC93-54
Date of Issue

Conference Information
Committee NC
Conference Date 1993/11/24(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) On the novel method of designing a neural network for pattern classification.
Sub Title (in English)
Keyword(1) back-propagation method
Keyword(2) output function
Keyword(3) hyperplane
Keyword(4) neural network
Keyword(5) pattern classification
1st Author's Name Toshinori Yoshikawa
1st Author's Affiliation Department of Electricity,Faculty of Engineering,Nagaoka University of Technology()
2nd Author's Name Satoshi Ookuma
2nd Author's Affiliation Department of Electricity,Faculty of Engineering,Nagaoka University of Technology
Date 1993/11/24
Paper # NC93-54
Volume (vol) vol.93
Number (no) 341
Page pp.pp.-
#Pages 7
Date of Issue