Novel back propagation algorithm for reduction of hidden units and acceleration of convergence using artificial selection

Research output: Chapter in Book/Report/Conference proceedingConference contribution

27 Citations (Scopus)

Abstract

A novel back-propagation algorithm with artificial selection is proposed. It is effective for both fast convergence and reduction of the number of hidden units. The main feature of the proposed algorithm is detection of the worst hidden unit. This is done by using the proposed badness factor, which indicates the badness of each hidden unit. It is the sum of back-propagated error components over all patterns for each hidden unit. For fast convergence, all the weights connected to the detected worst unit are reset to small random values at a suitable time. As for the reduction of hidden units, detected bad units are erased by precendent. Computer simulation results show the effectiveness of the proposed algorithm; for example, the numbers of hidden units in the EX-OR problems converge to 2 (theoretical number).

Original languageEnglish
Title of host publication90 Int Jt Conf Neural Networks IJCNN 90
PublisherPubl by IEEE
Pages625-630
Number of pages6
Publication statusPublished - 1990
Event1990 International Joint Conference on Neural Networks - IJCNN 90 - San Diego, CA, USA
Duration: 1990 Jun 171990 Jun 21

Other

Other1990 International Joint Conference on Neural Networks - IJCNN 90
CitySan Diego, CA, USA
Period90/6/1790/6/21

Fingerprint

Backpropagation algorithms
Computer simulation

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Novel back propagation algorithm for reduction of hidden units and acceleration of convergence using artificial selection. / Hagiwara, Masafumi.

90 Int Jt Conf Neural Networks IJCNN 90. Publ by IEEE, 1990. p. 625-630.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Hagiwara, M 1990, Novel back propagation algorithm for reduction of hidden units and acceleration of convergence using artificial selection. in 90 Int Jt Conf Neural Networks IJCNN 90. Publ by IEEE, pp. 625-630, 1990 International Joint Conference on Neural Networks - IJCNN 90, San Diego, CA, USA, 90/6/17.
@inproceedings{7afc62ef98d04a09ad4c1443654e755a,
title = "Novel back propagation algorithm for reduction of hidden units and acceleration of convergence using artificial selection",
abstract = "A novel back-propagation algorithm with artificial selection is proposed. It is effective for both fast convergence and reduction of the number of hidden units. The main feature of the proposed algorithm is detection of the worst hidden unit. This is done by using the proposed badness factor, which indicates the badness of each hidden unit. It is the sum of back-propagated error components over all patterns for each hidden unit. For fast convergence, all the weights connected to the detected worst unit are reset to small random values at a suitable time. As for the reduction of hidden units, detected bad units are erased by precendent. Computer simulation results show the effectiveness of the proposed algorithm; for example, the numbers of hidden units in the EX-OR problems converge to 2 (theoretical number).",
author = "Masafumi Hagiwara",
year = "1990",
language = "English",
pages = "625--630",
booktitle = "90 Int Jt Conf Neural Networks IJCNN 90",
publisher = "Publ by IEEE",

}

TY - GEN

T1 - Novel back propagation algorithm for reduction of hidden units and acceleration of convergence using artificial selection

AU - Hagiwara, Masafumi

PY - 1990

Y1 - 1990

N2 - A novel back-propagation algorithm with artificial selection is proposed. It is effective for both fast convergence and reduction of the number of hidden units. The main feature of the proposed algorithm is detection of the worst hidden unit. This is done by using the proposed badness factor, which indicates the badness of each hidden unit. It is the sum of back-propagated error components over all patterns for each hidden unit. For fast convergence, all the weights connected to the detected worst unit are reset to small random values at a suitable time. As for the reduction of hidden units, detected bad units are erased by precendent. Computer simulation results show the effectiveness of the proposed algorithm; for example, the numbers of hidden units in the EX-OR problems converge to 2 (theoretical number).

AB - A novel back-propagation algorithm with artificial selection is proposed. It is effective for both fast convergence and reduction of the number of hidden units. The main feature of the proposed algorithm is detection of the worst hidden unit. This is done by using the proposed badness factor, which indicates the badness of each hidden unit. It is the sum of back-propagated error components over all patterns for each hidden unit. For fast convergence, all the weights connected to the detected worst unit are reset to small random values at a suitable time. As for the reduction of hidden units, detected bad units are erased by precendent. Computer simulation results show the effectiveness of the proposed algorithm; for example, the numbers of hidden units in the EX-OR problems converge to 2 (theoretical number).

UR - http://www.scopus.com/inward/record.url?scp=0025547727&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0025547727&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:0025547727

SP - 625

EP - 630

BT - 90 Int Jt Conf Neural Networks IJCNN 90

PB - Publ by IEEE

ER -