Optimal estimation and rank detection for sparse spiked covariance matrices

Abstract

This paper considers a sparse spiked covariancematrix model in the high-dimensional setting and studies the minimax estimation of the covariance matrix and the principal subspace as well as the minimax rank detection. The optimal rate of convergence for estimating the spiked covariance matrix under the spectral norm is established, which requires significantly different techniques from those for estimating other structured covariance matrices such as bandable or sparse covariance matrices. We also establish the minimax rate under the spectral norm for estimating the principal subspace, the primary object of interest in principal component analysis. In addition, the optimal rate for the rank detection boundary is obtained. This result also resolves the gap in a recent paper by Berthet and Rigollet [2] where the special case of rank one is considered.

Information-Theoretic Bounds and Phase Transitions in Clustering, Sparse PCA, and Submatrix Localization

“…Independent of the present work, and building on our preprint [10], Perry et al [43] obtained the same tight threshold in this limit with a smaller error term. Previous work [16] had determined that threshold scales as λ = Θ( √ −γ log γ) up to a constant factor.…”

Detection of Correlations With Adaptive Sensing

“…There is, also for this problem, a growing literature, see [10], [12], [26]. Note that when the coordinates of u are constrained in {0, 1/ √ k}, we recover a problem akin to that of detection of positive correlations, but with unnormalized variances over the contaminated set.…”

“…Such problems have recently received a lot of attention in the literature, where different models and choices of non-diagonal covariance alternatives were considered [4], [5], [10], [12], [20]. We consider the detection of sparse positive correlations, which has been treated in the case of a unique multivariate sample [4], or of multiple samples [5].…”

“…Besides anomaly detection, detection of correlations is also of interest to assess to what extent dimensionality reduction can be performed on a data stream. Reduction of dimensionality is a workhorse of data analysis, and there has been a strong recent interest in modifying principal component analysis to deal with high-dimensional data [10], [12], [26]. Testing when this type of transformation is justified is thus an important problem.…”

Video Denoising via Empirical Bayesian Estimation of Space-Time Patches

“…Several authors [18,28,45] studied the relation between the sample covariance matrix and the population one, showing that the sample eigenvectors and eigenvalues are inconsistent estimators. Cai et al [9], impose additionally a group-sparsity constraint on the signal eigenvectors, and proposed estimators with an optimal minimax risk convergence rate. The proposed estimators are mainly of theoretical interest, since require a global search for the support of the eigenvectors.…”