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.

Low-Rank Tensor Completion Based on Log-Det Rank Approximation and Matrix Factorization

Tensor Completion Via a Multi-Linear Low-N-rank Factorization Model

Robust Approximations of Low-Rank Minimization for Tensor Completion

Beyond Low Rank: A Data-Adaptive Tensor Completion Method

A Novel Tensor Factorization-Based Method with Robustness to Inaccurate Rank Estimation

Hybrid Singular Value Thresholding For Tensor Completion

Low rank tensor completion with sparse regularization in a transformed domain

An Efficient Tensor Completion Method Via New Latent Nuclear Norm

Imbalanced low-rank tensor completion via latent matrix factorization

Tensor Factorization for Low-Rank Tensor Completion.

Tensor completion via nonconvex tensor ring rank minimization with guaranteed convergence

Low-Rank Tensor Completion by Truncated Nuclear Norm Regularization

Robust Low Tubal Rank Tensor Completion via Factor Tensor Norm Minimization

Low Multilinear Rank Tensor Completion with Missing Data

Low-Rank Tensor Completion Using Matrix Factorization Based on Tensor Train Rank and Total Variation

Matrix completion by Truncated Nuclear Norm Regularization

Tensor Ring Decomposition with Rank Minimization on Latent Space: An Efficient Approach for Tensor Completion

A Low-Rank Tensor Completion Method via Strassen–Ottaviani Flattening

A Scale-Invariant Relaxation in Low-Rank Tensor Recovery with an Application to Tensor Completion

Fast and Accurate Matrix Completion Via Truncated Nuclear Norm Regularization

Robust Low-tubal-rank Tensor Completion based on Tensor Factorization and Maximum Correntopy Criterion