DocumentCode
3116225
Title
An Information Theoretic Perspective to Kernel K-Means
Author
Jenssen, Robert ; Eltoft, Torbjorn
Author_Institution
Dept. of Phys., Univ. of Tromso, Tromso
fYear
2006
fDate
6-8 Sept. 2006
Firstpage
161
Lastpage
166
Abstract
In this paper, we provide an information theoretic perspective to kernel K-means. We show that kernel K-means corresponds to maximizing an integrated squared error divergence measure between Parzen window estimated cluster probability density functions. Equivalently, this corresponds to a Bayes-like clustering rule in the input space, taking into account the Renyi entropies of the clusters.
Keywords
Bayes methods; information theory; mean square error methods; pattern clustering; probability; Bayes-like clustering; Parzen window; Renyi entropy; cluster probability density function; information theory; kernel K-means; squared error divergence measure; Clustering algorithms; Density measurement; Entropy; Independent component analysis; Kernel; Principal component analysis; Probability density function; Signal processing algorithms; Support vector machines; Time measurement;
fLanguage
English
Publisher
ieee
Conference_Titel
Machine Learning for Signal Processing, 2006. Proceedings of the 2006 16th IEEE Signal Processing Society Workshop on
Conference_Location
Arlington, VA
ISSN
1551-2541
Print_ISBN
1-4244-0656-0
Electronic_ISBN
1551-2541
Type
conf
DOI
10.1109/MLSP.2006.275541
Filename
4053640
Link To Document