Do semidefinite relaxations solve sparse PCA up to the information limit?

Estimating the leading principal components of data, assuming they are sparse, is a central task in modern high-dimensional statistics. Many algorithms were developed for this sparse PCA problem, from simple diagonal thresholding to sophisticated semidefinite programming (SDP) methods. A key theoretical question is under what conditions can such algorithms recover the sparse principal components? We study this question for a single-spike model with an -sparse eigenvector, in the asymptotic regime as dimension and sample size both tend to infinity. Amini and Wainwright [Ann. Statist. 37 (2009) 2877-2921] proved that for sparsity levels , no algorithm, efficient or not, can reliably recover the sparse eigenvector. In contrast, for , diagonal thresholding is consistent. It was further conjectured that an SDP approach may close this gap between computational and information limits. We prove that when , the proposed SDP approach, at least in its standard usage, cannot recover the sparse spike. In fact, we conjecture that in the single-spike model, no computationally-efficient algorithm can recover a spike of -sparsity . Finally, we present empirical results suggesting that up to sparsity levels , recovery is possible by a simple covariance thresholding algorithm.
View on arXiv