A Sparse SVD Method for High-dimensional Data

Abstract

We present a new computational approach to approximating a large, noisy data table by a low-rank matrix with sparse singular vectors. The approximation is obtained from thresholded subspace iterations that produce the singular vectors simultaneously, rather than successively as in competing proposals. We introduce novel ways to estimate thresholding parameters which obviate the need for computationally expensive cross-validation. We also introduce a way to sparsely initialize the algorithm for computational savings that allow our algorithm to outperform the vanilla SVD on the full data table when the signal is sparse. A comparison with two existing sparse SVD methods suggests that our algorithm is computationally always faster and statistically always at least comparable to the better of the two competing algorithms.

References in corpus (3)

• Sparse principal component analysis and iterative thresholding
• Convex Sparse Matrix Factorizations
• Consistency of Sparse PCA in High Dimension, Low Sample Size Contexts

Cited by in corpus (5)

• Sparse CCA via Precision Adjusted Iterative Thresholding
• Minimax estimation in sparse canonical correlation analysis
• Sparse and Functional Principal Components Analysis
• Scalable Spectral Algorithms for Community Detection in Directed Networks
• Optimal Structured Principal Subspace Estimation: Metric Entropy and Minimax Rates