Generalization of quadratic regularized and standard fuzzy c-means clustering with respect to regularization of hard c-means

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

5 Citations (Scopus)

Abstract

In this paper, the quadratic regularized and standard fuzzy c-means clustering algorithms (qFCM and sFCM) are generalized with respect to hard c-means (HCM) regularization. First, qFCM is generalized from quadratic regularization to power regularization. The relation between this generalization and sFCM is then compared to the relation between other pairs of methods from the perspective of HCM regularization, and, based on this comparison, sFCM is generalized through the addition of a fuzzification parameter. In this process, we see that other methods can be constructed by combining HCM and a regularization term that can either be weighted by data-cluster dissimilarity or not. Furthermore, we see numerically that the existence or nonexistence of this weighting determines the property of these methods' classification rules for an extremely large datum. We also note that the problem of non-convergence in some methods can be avoided through further modification.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages152-165
Number of pages14
Volume8234 LNAI
DOIs
Publication statusPublished - 2013
Event10th International Conference on Modeling Decisions for Artificial Intelligence, MDAI 2013 - Barcelona
Duration: 2013 Nov 202013 Nov 22

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume8234 LNAI
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other10th International Conference on Modeling Decisions for Artificial Intelligence, MDAI 2013
CityBarcelona
Period13/11/2013/11/22

Fingerprint

Clustering algorithms

Keywords

  • fuzzy c-means clustering
  • regularization

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Kanzawa, Y. (2013). Generalization of quadratic regularized and standard fuzzy c-means clustering with respect to regularization of hard c-means. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 8234 LNAI, pp. 152-165). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8234 LNAI). https://doi.org/10.1007/978-3-642-41550-0_14

Generalization of quadratic regularized and standard fuzzy c-means clustering with respect to regularization of hard c-means. / Kanzawa, Yuchi.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 8234 LNAI 2013. p. 152-165 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8234 LNAI).

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

Kanzawa, Y 2013, Generalization of quadratic regularized and standard fuzzy c-means clustering with respect to regularization of hard c-means. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 8234 LNAI, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 8234 LNAI, pp. 152-165, 10th International Conference on Modeling Decisions for Artificial Intelligence, MDAI 2013, Barcelona, 13/11/20. https://doi.org/10.1007/978-3-642-41550-0_14
Kanzawa Y. Generalization of quadratic regularized and standard fuzzy c-means clustering with respect to regularization of hard c-means. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 8234 LNAI. 2013. p. 152-165. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-642-41550-0_14
Kanzawa, Yuchi. / Generalization of quadratic regularized and standard fuzzy c-means clustering with respect to regularization of hard c-means. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 8234 LNAI 2013. pp. 152-165 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{9ac81c6f26d94d8b8625331d33b17d25,
title = "Generalization of quadratic regularized and standard fuzzy c-means clustering with respect to regularization of hard c-means",
abstract = "In this paper, the quadratic regularized and standard fuzzy c-means clustering algorithms (qFCM and sFCM) are generalized with respect to hard c-means (HCM) regularization. First, qFCM is generalized from quadratic regularization to power regularization. The relation between this generalization and sFCM is then compared to the relation between other pairs of methods from the perspective of HCM regularization, and, based on this comparison, sFCM is generalized through the addition of a fuzzification parameter. In this process, we see that other methods can be constructed by combining HCM and a regularization term that can either be weighted by data-cluster dissimilarity or not. Furthermore, we see numerically that the existence or nonexistence of this weighting determines the property of these methods' classification rules for an extremely large datum. We also note that the problem of non-convergence in some methods can be avoided through further modification.",
keywords = "fuzzy c-means clustering, regularization",
author = "Yuchi Kanzawa",
year = "2013",
doi = "10.1007/978-3-642-41550-0_14",
language = "English",
isbn = "9783642415494",
volume = "8234 LNAI",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "152--165",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Generalization of quadratic regularized and standard fuzzy c-means clustering with respect to regularization of hard c-means

AU - Kanzawa, Yuchi

PY - 2013

Y1 - 2013

N2 - In this paper, the quadratic regularized and standard fuzzy c-means clustering algorithms (qFCM and sFCM) are generalized with respect to hard c-means (HCM) regularization. First, qFCM is generalized from quadratic regularization to power regularization. The relation between this generalization and sFCM is then compared to the relation between other pairs of methods from the perspective of HCM regularization, and, based on this comparison, sFCM is generalized through the addition of a fuzzification parameter. In this process, we see that other methods can be constructed by combining HCM and a regularization term that can either be weighted by data-cluster dissimilarity or not. Furthermore, we see numerically that the existence or nonexistence of this weighting determines the property of these methods' classification rules for an extremely large datum. We also note that the problem of non-convergence in some methods can be avoided through further modification.

AB - In this paper, the quadratic regularized and standard fuzzy c-means clustering algorithms (qFCM and sFCM) are generalized with respect to hard c-means (HCM) regularization. First, qFCM is generalized from quadratic regularization to power regularization. The relation between this generalization and sFCM is then compared to the relation between other pairs of methods from the perspective of HCM regularization, and, based on this comparison, sFCM is generalized through the addition of a fuzzification parameter. In this process, we see that other methods can be constructed by combining HCM and a regularization term that can either be weighted by data-cluster dissimilarity or not. Furthermore, we see numerically that the existence or nonexistence of this weighting determines the property of these methods' classification rules for an extremely large datum. We also note that the problem of non-convergence in some methods can be avoided through further modification.

KW - fuzzy c-means clustering

KW - regularization

UR - http://www.scopus.com/inward/record.url?scp=84893270345&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84893270345&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-41550-0_14

DO - 10.1007/978-3-642-41550-0_14

M3 - Conference contribution

SN - 9783642415494

VL - 8234 LNAI

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

SP - 152

EP - 165

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

ER -