Method Analysis Of Matrix Completion Theory

Matrix completion following the compressive sensing is a remarkable new field and it occurs in many applications of science and engineering such as collaborative filtering, image inpainting, machine learning, control, computer vision and predicting missing data in sensor net-works, etc. The problem is to complete a data matrix from a few observed entries. The researchers summarize many actual situations and establishe the theoretical model of the matrix completion:when a matrix that is low-rank or approximately low-rank satisfies some appropriate conditions including control of the sampled number, the marix coherence and so on. it can be completed exactly or inexactly by a convex optimization problem in high probability. The convex optimization is the smallest nuclear norm problem or the nuclear norm regularized linear least squares problem.Matrix completion started late and there are many problems and research directions worthy of our in-depth research. At present, many researchers have focused on recovery algorithms. Recovery algorithm is the core of matrix completion, which is of great significance to recover matrix.This paper introduces the basics of matrix completion, and then studys and analyzes the existing recovery algorithms as the main content. We give some types of Lagrange multiplier algorithm, a quadratc approximate algorithm and a series of algorithms over the Grassmann manifold, algorithm analysis, data implementation and experimental results.