For EDITORS

For READERS

All Issues

Vol.10, 2020
Vol.9, 2019
Vol.8, 2018
Vol.7, 2017
Vol.6, 2016
Vol.5, 2015
Vol.4, 2014
Vol.3, 2013
Vol.2, 2012
Vol.1, 2011
Volume 7, Number 2, 2017, Pages 600-616                                                                DOI:10.11948/2017037
A new model for sparse and low-rank matrix decomposition
Zisheng Liu,Jicheng Li,Guo Li,Jianchao Bai,Xuenian Liu
Keywords:Robust principal component analysis, sparse matrix, low-rank matrix, nuclear norm, matrix decomposition.
Abstract:
      The robust principal component analysis (RPCA) model is a popular method for solving problems with the nuclear norm and $\ell_1$ norm. However, it is time-consuming since in general one has to use the singular value decomposition in each iteration. In this paper, we introduce a novel model to reformulate the existed model by making use of low-rank matrix factorization to surrogate the nuclear norm for the sparse and low-rank decomposition problem. In such case we apply the Penalty Function Method (PFM) and Augmented Lagrangian Multipliers Method (ALMM) to solve this new non-convex optimization problem. Theoretically, corresponding to our methods, the convergence analysis is given respectively. Compared with classical RPCA, some practical numerical examples are simulated to show that our methods are much better than RPCA.
PDF      Download reader