A0201
Title: A scalable algorithm for estimating a low-rank plus sparse matrix
Authors: Eric Chi - Rice University (United States) [presenting]
Abstract: The problem of estimating a low-rank plus sparse matrix from noisy and potentially incomplete measurements are considered. Examples of this problem include robust matrix recovery, compressive principal component pursuit, and LORS regression in analysing expression quantitative trait loci. Prior formulations of this problem involve solving a convex optimization problem, which requires repeatedly computing a singular value decomposition. We will discuss a non-convex formulation that admits more scalable algorithms that avoid computing an expensive singular value decomposition.
