Efficient Recovery of Low Rank Tensor via Triple Nonconvex Nonsmooth Rank Minimization

TL;DR

This paper introduces a novel double weighted nonconvex nonsmooth rank relaxation method for third order tensor recovery, improving accuracy and efficiency over existing tensor nuclear norm approaches.

Contribution

It proposes a flexible rank relaxation function derived from a triple nonconvex nonsmooth rank, with an inertial smoothing proximal gradient method and convergence guarantees.

Findings

Achieves higher accuracy in tensor recovery tasks.

Demonstrates superior performance on synthetic and real data.

Provides theoretical convergence analysis.

Abstract

A tensor nuclear norm (TNN) based method for solving the tensor recovery problem was recently proposed, and it has achieved state-of-the-art performance. However, it may fail to produce a highly accurate solution since it tends to treats each frontal slice and each rank component of each frontal slice equally. In order to get a recovery with high accuracy, we propose a general and flexible rank relaxation function named double weighted nonconvex nonsmooth rank (DWNNR) relaxation function for efficiently solving the third order tensor recovery problem. The DWNNR relaxation function can be derived from the triple nonconvex nonsmooth rank (TNNR) relaxation function by setting the weight vector to be the hypergradient value of some concave function, thereby adaptively selecting the weight vector. To accelerate the proposed model, we develop the general inertial smoothing proximal gradient method. Furthermore, we prove that any limit point of the generated subsequence is a critical point. Combining the Kurdyka-Lojasiewicz (KL) property with some milder assumptions, we further give its global convergence guarantee. Experimental results on a practical tensor completion problem with both synthetic and real data, the results of which demonstrate the efficiency and superior performance of the proposed algorithm.