Computer Science – Information Theory
Scientific paper
2011-06-08
Computer Science
Information Theory
5 pages, 1 figure, to be presented at ISIT
Scientific paper
We consider the problem of identifying the sparse principal component of a rank-deficient matrix. We introduce auxiliary spherical variables and prove that there exists a set of candidate index-sets (that is, sets of indices to the nonzero elements of the vector argument) whose size is polynomially bounded, in terms of rank, and contains the optimal index-set, i.e. the index-set of the nonzero elements of the optimal solution. Finally, we develop an algorithm that computes the optimal sparse principal component in polynomial time for any sparsity degree.
Asteris Megasthenis
Karystinos George N.
Papailiopoulos Dimitris S.
No associations
LandOfFree
Sparse Principal Component of a Rank-deficient Matrix does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with Sparse Principal Component of a Rank-deficient Matrix, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Sparse Principal Component of a Rank-deficient Matrix will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-578333