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

研究成果: Conference contribution

7 被引用数 (Scopus)

抄録

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.

本文言語English
ホスト出版物のタイトルModeling Decisions for Artificial Intelligence - 10th International Conference, MDAI 2013, Proceedings
ページ152-165
ページ数14
DOI
出版ステータスPublished - 2013 12 1
イベント10th International Conference on Modeling Decisions for Artificial Intelligence, MDAI 2013 - Barcelona, Spain
継続期間: 2013 11 202013 11 22

出版物シリーズ

名前Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
8234 LNAI
ISSN(印刷版)0302-9743
ISSN(電子版)1611-3349

Conference

Conference10th International Conference on Modeling Decisions for Artificial Intelligence, MDAI 2013
国/地域Spain
CityBarcelona
Period13/11/2013/11/22

ASJC Scopus subject areas

  • 理論的コンピュータサイエンス
  • コンピュータ サイエンス(全般)

フィンガープリント

「Generalization of quadratic regularized and standard fuzzy c-means clustering with respect to regularization of hard c-means」の研究トピックを掘り下げます。これらがまとまってユニークなフィンガープリントを構成します。

引用スタイル