Multiway Spectral Clustering: A Margin-Based Perspective
Zhang, Zhihua ; Jordan, Michael I.
Statist. Sci., Tome 23 (2008) no. 1, p. 383-403 / Harvested from Project Euclid
Spectral clustering is a broad class of clustering procedures in which an intractable combinatorial optimization formulation of clustering is “relaxed” into a tractable eigenvector problem, and in which the relaxed solution is subsequently “rounded” into an approximate discrete solution to the original problem. In this paper we present a novel margin-based perspective on multiway spectral clustering. We show that the margin-based perspective illuminates both the relaxation and rounding aspects of spectral clustering, providing a unified analysis of existing algorithms and guiding the design of new algorithms. We also present connections between spectral clustering and several other topics in statistics, specifically minimum-variance clustering, Procrustes analysis and Gaussian intrinsic autoregression.
Publié le : 2008-08-15
Classification:  Spectral clustering,  spectral relaxation,  graph partitioning,  reproducing kernel Hilbert space,  large-margin classification,  Gaussian intrinsic autoregression
@article{1233153065,
     author = {Zhang, Zhihua and Jordan, Michael I.},
     title = {Multiway Spectral Clustering: A Margin-Based Perspective},
     journal = {Statist. Sci.},
     volume = {23},
     number = {1},
     year = {2008},
     pages = { 383-403},
     language = {en},
     url = {http://dml.mathdoc.fr/item/1233153065}
}
Zhang, Zhihua; Jordan, Michael I. Multiway Spectral Clustering: A Margin-Based Perspective. Statist. Sci., Tome 23 (2008) no. 1, pp.  383-403. http://gdmltest.u-ga.fr/item/1233153065/