Abstract:Recovering a large matrix from a small subset of its entries is a challenging problem arising in many real applications, such as image inpainting and recommender systems. Many existing approaches formulate this problem as a general low-rank matrix approximation problem. Since the rank operator is nonconvex and discontinuous, most of the recent theoretical studies use the nuclear norm as a convex relaxation. One major limitation of the existing approaches based on nuclear norm minimization is that all the singular values are simultaneously minimized, and thus the rank may not be well approximated in practice. In this paper, we propose to achieve a better approximation to the rank of matrix by truncated nuclear norm, which is given by the nuclear norm subtracted by the sum of the largest few singular values. In addition, we develop a novel matrix completion algorithm by minimizing the Truncated Nuclear Norm. We further develop three efficient iterative procedures, TNNR-ADMM, TNNR-APGL, and TNNR-ADMMAP, to solve the optimization problem. TNNR-ADMM utilizes the alternating direction method of multipliers (ADMM), while TNNR-AGPL applies the accelerated proximal gradient line search method (APGL) for the final optimization. For TNNR-ADMMAP, we make use of an adaptive penalty according to a novel update rule for ADMM to achieve a faster convergence rate. Our empirical study shows encouraging results of the proposed algorithms in comparison to the state-of-the-art matrix completion algorithms on both synthetic and real visual datasets.

A Close Look at Matrix Completion with Noisy Data

A Non-Local Method For Robust Noisy Image Completion

Low Rank Matrix Recovery with Adversarial Sparse Noise

Matrix Completion with Noisy Entries and Outliers

Matrix Completion with Sparse Noisy Rows

Matrix Completion with Nonuniform Sampling: Theories and Methods.

Noisy Matrix Completion: Understanding Statistical Guarantees for Convex Relaxation via Nonconvex Optimization

Matrix Completion with Deterministic Sampling: Theories and Methods

Accelerated singular value thresholding for matrix completion.

Robust matrix completion with complex noise

Fast and Accurate Matrix Completion Via Truncated Nuclear Norm Regularization

Matrix Completion Based on Low-Rank and Local Features Applied to Images Recovery and Recommendation Systems.

Locally Linear Approximation Approach for Incomplete Data.

A Nearly Unbiased Matrix Completion Approach.

Robust Matrix Completion with Heavy-tailed Noise

A New Theory for Matrix Completion.

Adaptive Rank-One Matrix Completion using Sum of Outer Products

Robust Matrix Completion via Maximum Correntropy Criterion and Half Quadratic Optimization

Leveraged Matrix Completion with Noise

Partial Matrix Completion

Matrix Completion from Non-Uniformly Sampled Entries.