A1257
Title: Beyond James-Stein estimation for PCA
Authors: Alexander Shkolnik - University of California, Santa Barbara (United States) [presenting]
Abstract: Recent progress on James-Stein estimation of principal components and, more generally, of singular vectors of random data matrices is surveyed. Connections to Bayesian methods, regularization, and Ledoit-Wolf covariance estimation are discussed. Several results on the convergence properties of the James-Stein estimator for the leading singular vector pair are presented. The topic of the inadmissibility of Principal Component Analysis (PCA), as well as the inadmissibility of the James-Stein estimator for PCA, are also discussed.
