Evaluation of distance measures for speciated evolutionary neural networks in pattern classification problems

Kyung Joong Kim, Sung Bae Cho

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

3 Citations (Scopus)

Abstract

Recently, evolutionary neural networks are hot topics in a neural network community because of their flexibility and good performance. However, they suffer from a premature convergence problem caused by the genetic drift of evolutionary algorithms. The genetic diversity in a population decreases quickly and it loses an exploration capability. Based on the inspiration of diversity in nature, a number of speciation algorithms are proposed to maintain diverse solutions from the population. One problem arising from this approach is lack of information on the distance measures among neural networks to penalize or discard similar solutions. In this paper, a comparison is conducted for six distance measures (genotypic, phenotypic, and behavioral types) with representative speciation algorithms (fitness sharing and deterministic crowding genetic algorithms) on six UCI benchmark datasets. It shows that the choice of distance measures is important in the neural network evolution.

Original languageEnglish
Title of host publicationNeural Information Processing - 16th International Conference, ICONIP 2009, Proceedings
Pages630-637
Number of pages8
EditionPART 2
DOIs
Publication statusPublished - 2009 Dec 1
Event16th International Conference on Neural Information Processing, ICONIP 2009 - Bangkok, Thailand
Duration: 2009 Dec 12009 Dec 5

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 2
Volume5864 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other16th International Conference on Neural Information Processing, ICONIP 2009
CountryThailand
CityBangkok
Period09/12/109/12/5

Fingerprint

Evolutionary Neural Networks
Pattern Classification
Distance Measure
Classification Problems
Pattern recognition
Speciation
Neural Networks
Neural networks
Evaluation
Genetic Drift
Network Evolution
Premature Convergence
Deterministic Algorithm
Fitness
Evolutionary Algorithms
Sharing
Flexibility
Genetic Algorithm
Evolutionary algorithms
Benchmark

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Kim, K. J., & Cho, S. B. (2009). Evaluation of distance measures for speciated evolutionary neural networks in pattern classification problems. In Neural Information Processing - 16th International Conference, ICONIP 2009, Proceedings (PART 2 ed., pp. 630-637). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5864 LNCS, No. PART 2). https://doi.org/10.1007/978-3-642-10684-2_70
Kim, Kyung Joong ; Cho, Sung Bae. / Evaluation of distance measures for speciated evolutionary neural networks in pattern classification problems. Neural Information Processing - 16th International Conference, ICONIP 2009, Proceedings. PART 2. ed. 2009. pp. 630-637 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 2).
@inproceedings{7055ede017cd4b3c8a37edd4478411a9,
title = "Evaluation of distance measures for speciated evolutionary neural networks in pattern classification problems",
abstract = "Recently, evolutionary neural networks are hot topics in a neural network community because of their flexibility and good performance. However, they suffer from a premature convergence problem caused by the genetic drift of evolutionary algorithms. The genetic diversity in a population decreases quickly and it loses an exploration capability. Based on the inspiration of diversity in nature, a number of speciation algorithms are proposed to maintain diverse solutions from the population. One problem arising from this approach is lack of information on the distance measures among neural networks to penalize or discard similar solutions. In this paper, a comparison is conducted for six distance measures (genotypic, phenotypic, and behavioral types) with representative speciation algorithms (fitness sharing and deterministic crowding genetic algorithms) on six UCI benchmark datasets. It shows that the choice of distance measures is important in the neural network evolution.",
author = "Kim, {Kyung Joong} and Cho, {Sung Bae}",
year = "2009",
month = "12",
day = "1",
doi = "10.1007/978-3-642-10684-2_70",
language = "English",
isbn = "364210682X",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
number = "PART 2",
pages = "630--637",
booktitle = "Neural Information Processing - 16th International Conference, ICONIP 2009, Proceedings",
edition = "PART 2",

}

Kim, KJ & Cho, SB 2009, Evaluation of distance measures for speciated evolutionary neural networks in pattern classification problems. in Neural Information Processing - 16th International Conference, ICONIP 2009, Proceedings. PART 2 edn, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 2, vol. 5864 LNCS, pp. 630-637, 16th International Conference on Neural Information Processing, ICONIP 2009, Bangkok, Thailand, 09/12/1. https://doi.org/10.1007/978-3-642-10684-2_70

Evaluation of distance measures for speciated evolutionary neural networks in pattern classification problems. / Kim, Kyung Joong; Cho, Sung Bae.

Neural Information Processing - 16th International Conference, ICONIP 2009, Proceedings. PART 2. ed. 2009. p. 630-637 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5864 LNCS, No. PART 2).

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

TY - GEN

T1 - Evaluation of distance measures for speciated evolutionary neural networks in pattern classification problems

AU - Kim, Kyung Joong

AU - Cho, Sung Bae

PY - 2009/12/1

Y1 - 2009/12/1

N2 - Recently, evolutionary neural networks are hot topics in a neural network community because of their flexibility and good performance. However, they suffer from a premature convergence problem caused by the genetic drift of evolutionary algorithms. The genetic diversity in a population decreases quickly and it loses an exploration capability. Based on the inspiration of diversity in nature, a number of speciation algorithms are proposed to maintain diverse solutions from the population. One problem arising from this approach is lack of information on the distance measures among neural networks to penalize or discard similar solutions. In this paper, a comparison is conducted for six distance measures (genotypic, phenotypic, and behavioral types) with representative speciation algorithms (fitness sharing and deterministic crowding genetic algorithms) on six UCI benchmark datasets. It shows that the choice of distance measures is important in the neural network evolution.

AB - Recently, evolutionary neural networks are hot topics in a neural network community because of their flexibility and good performance. However, they suffer from a premature convergence problem caused by the genetic drift of evolutionary algorithms. The genetic diversity in a population decreases quickly and it loses an exploration capability. Based on the inspiration of diversity in nature, a number of speciation algorithms are proposed to maintain diverse solutions from the population. One problem arising from this approach is lack of information on the distance measures among neural networks to penalize or discard similar solutions. In this paper, a comparison is conducted for six distance measures (genotypic, phenotypic, and behavioral types) with representative speciation algorithms (fitness sharing and deterministic crowding genetic algorithms) on six UCI benchmark datasets. It shows that the choice of distance measures is important in the neural network evolution.

UR - http://www.scopus.com/inward/record.url?scp=76249090816&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=76249090816&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-10684-2_70

DO - 10.1007/978-3-642-10684-2_70

M3 - Conference contribution

AN - SCOPUS:76249090816

SN - 364210682X

SN - 9783642106828

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 630

EP - 637

BT - Neural Information Processing - 16th International Conference, ICONIP 2009, Proceedings

ER -

Kim KJ, Cho SB. Evaluation of distance measures for speciated evolutionary neural networks in pattern classification problems. In Neural Information Processing - 16th International Conference, ICONIP 2009, Proceedings. PART 2 ed. 2009. p. 630-637. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 2). https://doi.org/10.1007/978-3-642-10684-2_70