On Bezdek-type possibilistic clustering for spherical data, its kernelization, and spectral clustering approach

研究成果: Conference contribution

抄録

In this study, a Bezdek-type fuzzified possibilistic clustering algorithm for spherical data (bPCS), its kernelization (K-bPCS), and spectral clustering approach (sK-bPCS) are proposed. First, we propose the bPCS by setting a fuzzification parameter of the Tsallis entropy-based possibilistic clustering optimization problem for spherical data (tPCS) to infinity, and by modifying the cosine correlationbased dissimilarity between objects and cluster centers. Next, we kernelize bPCS to obtain K-bPCS, which can be applied to non-spherical data with the help of a given kernel, e.g., a Gaussian kernel. Furthermore, we propose a spectral clustering approach to K-bPCS called sK-bPCS, which aims to solve the initialization problem of bPCS and K-bPCS. Furthermore, we demonstrate that this spectral clustering approach is equivalent to kernelized principal component analysis (K-PCA). The validity of the proposed methods is verified through numerical examples.

本文言語English
ホスト出版物のタイトルModeling Decisions for Artificial Intelligence - 13th International Conference, MDAI 2016, Proceedings
出版社Springer Verlag
ページ178-190
ページ数13
9880 LNAI
ISBN(印刷版)9783319456553
DOI
出版ステータスPublished - 2016
イベント13th International Conference on Modeling Decisions for Artificial Intelligence, MDAI 2016 - Sant Juliadeloria, Andorra
継続期間: 2016 9 192016 9 21

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)
9880 LNAI
ISSN(印刷版)03029743
ISSN(電子版)16113349

Other

Other13th International Conference on Modeling Decisions for Artificial Intelligence, MDAI 2016
国/地域Andorra
CitySant Juliadeloria
Period16/9/1916/9/21

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

「On Bezdek-type possibilistic clustering for spherical data, its kernelization, and spectral clustering approach」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