An efficient K-hyperplane clustering algorithm and its application to sparse component analysis

Zhaoshui He, Andrzej Cichocki

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

6 Citations (Scopus)

Abstract

Based on eigenvalue decomposition, a novel efficient K-HPC algorithm is developed in this paper, which is easy to implement. And it enables us to detect the number of hyperplanes and helps to avoid local minima by overestimating the number of hyperplanes. A confidence index is proposed to evaluate which estimated hyperplanes are most significant and which are spurious. So we can choose those significant hyperplanes with high rank priority and remove the spurious hyperplanes according to their corresponding confidence indices. Furthermore, a multilayer clustering framework called "multilayer K-HPC" is proposed to further improve the clustering results. The K-HPC approach can be directly applied to sparse component analysis (SCA) to develop efficient SCA algorithm. Two examples including a sparse component analysis example demonstrate the proposed algorithm.

Original languageEnglish
Title of host publicationAdvances in Neural Networks - ISNN 2007 - 4th International Symposium on Neural Networks, ISNN 2007, Proceedings
PublisherSpringer Verlag
Pages1032-1041
Number of pages10
EditionPART 2
ISBN (Print)9783540723929
DOIs
Publication statusPublished - 2007
Externally publishedYes
Event4th International Symposium on Neural Networks, ISNN 2007 - Nanjing, China
Duration: 3 Jun 20077 Jun 2007

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 2
Volume4492 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference4th International Symposium on Neural Networks, ISNN 2007
Country/TerritoryChina
CityNanjing
Period3/06/077/06/07

Fingerprint

Dive into the research topics of 'An efficient K-hyperplane clustering algorithm and its application to sparse component analysis'. Together they form a unique fingerprint.

Cite this