Generalizing the multiple measurement setting from sparse vector recovery to low-rank matrix recovery

摘要

We discuss a novel multiple measurement setting for low-rank matrix recovery in analogy to the ap­proach taken in the sparse vector case. This setting is related to the so-called Generalized Low-Rank Approximation of Matrices (GLRAM) and relies on assuming a joint row and column space model of the matrices. We study the well-posedness of the corresponding optimization problem and propose for its solution an efficient algorithm, based on the classical Alternating Direction Method of multipliers (ADM) algorithm.