Indexed by:
Abstract:
Sparse clustering, which aims at finding a proper partition of extremely high dimensional data set with fewest relevant features, has been attracted more and more attention. Most researches model the problem through minimizing weighted feature contributions subject to a l(1) constraint. However, the l(0) constraint is the essential constraint for sparse modeling while the l(1) constraint is only a convex relaxation of it. In this article, we bridge the gap between the l(0) constraint and the l(1) constraint through development of two new sparse clustering models, which are the sparse k-means with the l(q) (0 < q < 1) constraint and the sparse k-means with the l(0) constraint. By proving the certain forms of the optimal solution of particular l(q) (0 <= q <1) non-convex optimizations, two efficient iterative algorithms are proposed. We conclude with experiments on both synthetic data and the Allen Developing Mouse Brain Atlas data that the l(q) (0 <= q <1) models exhibit the advantages compared with the standard k-means and sparse k-means with the l(1) constraint.
Keyword:
Reprint Author's Address:
Source :
2013 IEEE 13TH INTERNATIONAL CONFERENCE ON DATA MINING (ICDM)
ISSN: 1550-4786
Year: 2013
Page: 797-806
Language: English
Cited Count:
WoS CC Cited Count: 2
SCOPUS Cited Count: 4
ESI Highly Cited Papers on the List: 0 Unfold All
WanFang Cited Count:
Chinese Cited Count:
30 Days PV: 6
Affiliated Colleges: