Spectral rotation versus K-Means in spectral clustering

Jin Huang, Feiping Nie, Heng Huang

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

158 Scopus citations

Abstract

Spectral clustering has been a popular data clustering algorithm. This category of approaches often resort to other clustering methods, such as K-Means, to get the final cluster. The potential flaw of such common practice is that the obtained relaxed continuous spectral solution could severely deviate from the true discrete solution. In this paper, we propose to impose an additional orthonormal constraint to better approximate the optimal continuous solution to the graph cut objective functions. Such a method, called spectral rotation in literature, optimizes the spectral clustering objective functions better than K-Means, and improves the clustering accuracy. We would provide efficient algorithm to solve the new problem rigorously, which is not significantly more costly than K-Means. We also establish the connection between our method and K-Means to provide theoretical motivation of our method. Experimental results show that our algorithm consistently reaches better cut and meanwhile outperforms in clustering metrics than classic spectral clustering methods.

Original languageEnglish
Title of host publicationProceedings of the 27th AAAI Conference on Artificial Intelligence, AAAI 2013
Pages431-437
Number of pages7
StatePublished - 2013
Externally publishedYes
Event27th AAAI Conference on Artificial Intelligence, AAAI 2013 - Bellevue, WA, United States
Duration: 14 Jul 201318 Jul 2013

Publication series

NameProceedings of the 27th AAAI Conference on Artificial Intelligence, AAAI 2013

Conference

Conference27th AAAI Conference on Artificial Intelligence, AAAI 2013
Country/TerritoryUnited States
CityBellevue, WA
Period14/07/1318/07/13

Fingerprint

Dive into the research topics of 'Spectral rotation versus K-Means in spectral clustering'. Together they form a unique fingerprint.

Cite this