论文信息 - Numerical algorithms for low-rank matrix completion problems

Numerical algorithms for low-rank matrix completion problems

We consider a problem of recovering low-rank data matrix from sampling of its entries. Suppose that we observe m entries selected uniformly at random from an n1×n2 matrix M . One can hope that when enough entries are revealed (O(nrlogn)), it is possible to recover the matrix exactly. We downloaded eight solvers implemented in Matlab for low-rank matrix completion and tested them on different problems. The report includes brief description of the solvers used (based on the original papers) and results of the experiment carried out.

Marie Michenkova

[1] S. Yun,et al. An accelerated proximal gradient algorithm for nuclear norm regularized linear least squares problems , 2009 .

[2] Emmanuel J. Candès,et al. A Singular Value Thresholding Algorithm for Matrix Completion , 2008, SIAM J. Optim..

[3] Yi Ma,et al. The Augmented Lagrange Multiplier Method for Exact Recovery of Corrupted Low-Rank Matrices , 2010, Journal of structural biology.

[4] Yoram Bresler,et al. ADMiRA: Atomic Decomposition for Minimum Rank Approximation , 2009, IEEE Transactions on Information Theory.

[5] Shiqian Ma,et al. Fixed point and Bregman iterative methods for matrix rank minimization , 2009, Math. Program..

[6] Olgica Milenkovic,et al. Subspace Evolution and Transfer (SET) for Low-Rank Matrix Completion , 2010, IEEE Transactions on Signal Processing.

[7] Xiaoming Yuan,et al. Recovering Low-Rank and Sparse Components of Matrices from Incomplete and Noisy Observations , 2011, SIAM J. Optim..

[8] Yin Zhang,et al. Solving a low-rank factorization model for matrix completion by a nonlinear successive over-relaxation algorithm , 2012, Mathematical Programming Computation.

[9] Bingsheng He,et al. Linearized Alternating Direction Method with Gaussian Back Substitution for Separable Convex Programming , 2011 .