Abstract:Latent low-rank representation (LLRR) is a critical method to uncover the low-dimensional subspace structure embedded in high-dimensional data, especially when the data sampling is insufficient and/or extremely corrupted. However, LLRR uses the nuclear norm as a convex relaxation of the rank function, and thus the rank may not be well approximated in practice since the nuclear norm treats all singular values equally. To overcome this deficiency, in this paper, we propose a new model, TS1-LLRR, which extends LLRR along with its corresponding subspace clustering algorithm. We first adopt the non-convex transformed Schatten-1 (TS1) function, a better approximation to the rank of the matrix than the nuclear norm, to model the coefficient matrix. The proposed model works well in learning intrinsic rank information and thus has a stronger ability to capture the subspace structure. Furthermore, to solve the proposed non-convex model, we design an efficient iterative algorithm by using the alternating direction method of multipliers (ADMM) framework. This algorithm performs well in both accuracy and convergence speed, and the resulting subproblems have closed-form solutions. Last, we construct the affinity matrix for spectral clustering by further exploiting the angular information of the principal directions of the low-rank representation, which enhances the capability of investigating the inherent correlation between data points. Extensive experimental results on image clustering and motion segmentation show that our proposed subspace clustering method outperforms the compared methods.

LRR for Subspace Segmentation Via Tractable Schatten- $p$ Norm Minimization and Factorization.

LatLRR for subspace clustering via reweighted Frobenius norm minimization

A Fast Generalized Low Rank Representation Framework Based on $l_{2,p}$ Norm Minimization for Subspace Clustering

Robust Subspace Segmentation Via Nonconvex Low Rank Representation.

Robust Subspace Clustering Via Joint Weighted Schatten-p Norm and Lq Norm Minimization.

A Group Norm Regularized Factorization Model for Subspace Segmentation

Robust Subspace Clustering Based on Latent Low-rank Representation with Weighted Schatten-<i>p</i> Norm Minimization

Latent Low-Rank Representation for Subspace Segmentation and Feature Extraction

Restricted -Isometry Properties of Nonconvex Matrix Recovery

Robust Subspace Segmentation Via Low-Rank Representation.

Scalable Algorithms for Tractable Schatten Quasi-Norm Minimization

Subspace Clustering Based on Latent Low Rank Representation with Weighted Nuclear Norm Minimization.

Subspace Clustering Based on Latent Low-Rank Representation with Transformed Schatten-1 Penalty Function

Low-Rank Approximation via Generalized Reweighted Iterative Nuclear and Frobenius Norms

LRSR: Low-Rank-Sparse Representation for Subspace Clustering.

Analysis and Improvement of Low Rank Representation for Subspace Segmentation

Tractable and Scalable Schatten Quasi-Norm Approximations for Rank Minimization

Fast Universal Low Rank Representation

Improving Generalization of Double Low-Rank Representation Using Schatten-p Norm

Robust Subspace Segmentation by Low-Rank Representation.

Fast Algorithm for Large-Scale Subspace Clustering by LRR