Knowledge simplification of hierarchical neural network for multidimensional pattern recognition problems

Satoru Suzuki, Yasue Mitsukura

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

Abstract

The purpose of this study is to delete the redundant connections of hierarchical neural network constructed for solving pattern recognition problem with images. The performance of neural network changes depending on the number of hidden units. For example, a lot of hidden units cause the over-fitting problem and make it difficult to understand the role of hidden units. In order to diminish the redundant connections, we propose the connection elimination method by using genetic algorithm. Firstly, walsh-hadamard transform is applied to images for feature extraction. Secondly, neural network is trained with extracted features based on back-propagation algorithm. Finally, redundant connections are eliminated by optimization processing with genetic algorithm. In order to show the effectiveness of the proposed method, computer simulation is performed for face recognition examples. From the simulation results, it was confirmed that our proposed method was useful for eliminating redundant connections of neural network, maintaining recognition performance at high level.

Original languageEnglish
Title of host publicationProceedings of the SICE Annual Conference
Pages1050-1054
Number of pages5
Publication statusPublished - 2010
Externally publishedYes
EventSICE Annual Conference 2010, SICE 2010 - Taipei, Taiwan, Province of China
Duration: 2010 Aug 182010 Aug 21

Other

OtherSICE Annual Conference 2010, SICE 2010
CountryTaiwan, Province of China
CityTaipei
Period10/8/1810/8/21

Fingerprint

Pattern recognition
Neural networks
Genetic algorithms
Walsh transforms
Hadamard transforms
Backpropagation algorithms
Face recognition
Feature extraction
Computer simulation
Processing

Keywords

  • Genetic algorithm
  • Neural network
  • Walsh-hadamard transform

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Control and Systems Engineering
  • Computer Science Applications

Cite this

Suzuki, S., & Mitsukura, Y. (2010). Knowledge simplification of hierarchical neural network for multidimensional pattern recognition problems. In Proceedings of the SICE Annual Conference (pp. 1050-1054). [5602964]

Knowledge simplification of hierarchical neural network for multidimensional pattern recognition problems. / Suzuki, Satoru; Mitsukura, Yasue.

Proceedings of the SICE Annual Conference. 2010. p. 1050-1054 5602964.

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

Suzuki, S & Mitsukura, Y 2010, Knowledge simplification of hierarchical neural network for multidimensional pattern recognition problems. in Proceedings of the SICE Annual Conference., 5602964, pp. 1050-1054, SICE Annual Conference 2010, SICE 2010, Taipei, Taiwan, Province of China, 10/8/18.
Suzuki S, Mitsukura Y. Knowledge simplification of hierarchical neural network for multidimensional pattern recognition problems. In Proceedings of the SICE Annual Conference. 2010. p. 1050-1054. 5602964
@inproceedings{d45ab5b2311f41b9b4f632df6549a43f,
title = "Knowledge simplification of hierarchical neural network for multidimensional pattern recognition problems",
abstract = "The purpose of this study is to delete the redundant connections of hierarchical neural network constructed for solving pattern recognition problem with images. The performance of neural network changes depending on the number of hidden units. For example, a lot of hidden units cause the over-fitting problem and make it difficult to understand the role of hidden units. In order to diminish the redundant connections, we propose the connection elimination method by using genetic algorithm. Firstly, walsh-hadamard transform is applied to images for feature extraction. Secondly, neural network is trained with extracted features based on back-propagation algorithm. Finally, redundant connections are eliminated by optimization processing with genetic algorithm. In order to show the effectiveness of the proposed method, computer simulation is performed for face recognition examples. From the simulation results, it was confirmed that our proposed method was useful for eliminating redundant connections of neural network, maintaining recognition performance at high level.",
keywords = "Genetic algorithm, Neural network, Walsh-hadamard transform",
author = "Satoru Suzuki and Yasue Mitsukura",
year = "2010",
language = "English",
isbn = "9784907764364",
pages = "1050--1054",
booktitle = "Proceedings of the SICE Annual Conference",

}

TY - GEN

T1 - Knowledge simplification of hierarchical neural network for multidimensional pattern recognition problems

AU - Suzuki, Satoru

AU - Mitsukura, Yasue

PY - 2010

Y1 - 2010

N2 - The purpose of this study is to delete the redundant connections of hierarchical neural network constructed for solving pattern recognition problem with images. The performance of neural network changes depending on the number of hidden units. For example, a lot of hidden units cause the over-fitting problem and make it difficult to understand the role of hidden units. In order to diminish the redundant connections, we propose the connection elimination method by using genetic algorithm. Firstly, walsh-hadamard transform is applied to images for feature extraction. Secondly, neural network is trained with extracted features based on back-propagation algorithm. Finally, redundant connections are eliminated by optimization processing with genetic algorithm. In order to show the effectiveness of the proposed method, computer simulation is performed for face recognition examples. From the simulation results, it was confirmed that our proposed method was useful for eliminating redundant connections of neural network, maintaining recognition performance at high level.

AB - The purpose of this study is to delete the redundant connections of hierarchical neural network constructed for solving pattern recognition problem with images. The performance of neural network changes depending on the number of hidden units. For example, a lot of hidden units cause the over-fitting problem and make it difficult to understand the role of hidden units. In order to diminish the redundant connections, we propose the connection elimination method by using genetic algorithm. Firstly, walsh-hadamard transform is applied to images for feature extraction. Secondly, neural network is trained with extracted features based on back-propagation algorithm. Finally, redundant connections are eliminated by optimization processing with genetic algorithm. In order to show the effectiveness of the proposed method, computer simulation is performed for face recognition examples. From the simulation results, it was confirmed that our proposed method was useful for eliminating redundant connections of neural network, maintaining recognition performance at high level.

KW - Genetic algorithm

KW - Neural network

KW - Walsh-hadamard transform

UR - http://www.scopus.com/inward/record.url?scp=78649266008&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78649266008&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:78649266008

SN - 9784907764364

SP - 1050

EP - 1054

BT - Proceedings of the SICE Annual Conference

ER -