# American Institute of Mathematical Sciences

• Previous Article
An efficient adjoint computational method based on lifted IRK integrator and exact penalty function for optimal control problems involving continuous inequality constraints
• DCDS-S Home
• This Issue
• Next Article
Design of green bonds by double-barrier options

## An alternating minimization method for matrix completion problems

 1 School of Applied Mathematics, Nanjing University Of Finance & Economics, China 2 State Key Laboratory of Scientific and Engineering Computing, Academy of Mathematics and Systems Science, Chinese Academy of Sciences, University of Chinese Academy of Sciences, China

* Corresponding author

Received  September 2018 Revised  November 2018 Published  September 2019

Fund Project: The first author is supported by NSFC grants 11401295 and 11726618, and by Major Program of the National Social Science Foundation of China under Grant 12 & ZD114 and by National Social Science Foundation of China under Grant 15BGL158, 17BTQ063 and by Qinglan Project of Jiangsu Province, and Social Science Foundation of Jiangsu Province under Grant 18GLA002. The second author is supported by NSFC grants 11622112, 11471325, 91530204 and 11688101, the National Center for Mathematics and Interdisciplinary Sciences, CAS, and Key Research Program of Frontier Sciences QYZDJ-SSW-SYS010, CAS

Matrix completion problems have applications in various domains such as information theory, statistics, engineering, etc. Meanwhile, solving matrix completion problems is not a easy task since the nonconvex and nonsmooth rank operation is involved. Existing approaches can be categorized into two classes. The first ones use nuclear norm to take the place of rank operation, and any convex optimization algorithms can be used to solve the reformulated problem. The limitation of this class of approaches is singular value decomposition (SVD) is involved to tackle the nuclear norm which significantly increases the computational cost. The other ones factorize the target matrix by two slim matrices. Fast algorithms for solving the reformulated nonconvex optimization problem usually lack of global convergence, meanwhile convergence guaranteed algorithms require restricted stepsize. In this paper, we consider the matrix factorization model for matrix completion problems, and propose an alternating minimization method for solving it. The global convergence to a stationary point or local minimizer is guaranteed under mild conditions. We compare the proposed algorithm with some state-of-the-art algorithms in solving a bunch of testing problems. The numerical results illustrate the efficiency and great potential of our algorithm.

Citation: Yuan Shen, Xin Liu. An alternating minimization method for matrix completion problems. Discrete & Continuous Dynamical Systems - S, doi: 10.3934/dcdss.2020103
##### References:

show all references

##### References:
Relative error vs iteration number, $m$ = $n$ = 2000
Computing time vs sampling ratio, $m$ = $n$ = 2000
Computing time vs dimension
Results of speed performance test for synthetic data with $m$ = $n$ = $2000$
 Problem settings SVT LMaFit New Algorithm sampling ratio rank($L^*$) $\hbox{err}_{\Omega}(L)$ iter time $\hbox{err}_{\Omega}(L)$ iter time $\hbox{err}_{\Omega}(L)$ iter time 20% 5 8.947e-13 126.0 10.288 8.396e-13 67.7 1.746 8.880e-13 42.7 1.218 10 1.183e-7 190.2 22.593 9.005e-13 66.8 1.892 8.525e-13 57.3 1.821 40% 5 8.494e-13 87.0 12.286 7.022e-13 33.3 1.376 6.705e-13 24.2 1.100 10 8.702e-13 101.7 19.319 7.169e-13 39.2 1.722 7.001e-13 29.9 1.468 60% 5 8.470e-13 72.3 14.429 6.895e-13 33.2 1.821 5.106e-13 18.0 1.096 10 8.465e-13 80.4 18.717 7.573e-13 39.1 2.237 5.467e-13 21.0 1.334 80% 5 8.765e-13 62.3 15.112 6.556e-13 23.1 1.547 3.881e-13 15.0 1.100 10 8.180e-13 67.5 19.641 6.421e-13 25.1 1.801 4.308e-13 16.7 1.322
 Problem settings SVT LMaFit New Algorithm sampling ratio rank($L^*$) $\hbox{err}_{\Omega}(L)$ iter time $\hbox{err}_{\Omega}(L)$ iter time $\hbox{err}_{\Omega}(L)$ iter time 20% 5 8.947e-13 126.0 10.288 8.396e-13 67.7 1.746 8.880e-13 42.7 1.218 10 1.183e-7 190.2 22.593 9.005e-13 66.8 1.892 8.525e-13 57.3 1.821 40% 5 8.494e-13 87.0 12.286 7.022e-13 33.3 1.376 6.705e-13 24.2 1.100 10 8.702e-13 101.7 19.319 7.169e-13 39.2 1.722 7.001e-13 29.9 1.468 60% 5 8.470e-13 72.3 14.429 6.895e-13 33.2 1.821 5.106e-13 18.0 1.096 10 8.465e-13 80.4 18.717 7.573e-13 39.1 2.237 5.467e-13 21.0 1.334 80% 5 8.765e-13 62.3 15.112 6.556e-13 23.1 1.547 3.881e-13 15.0 1.100 10 8.180e-13 67.5 19.641 6.421e-13 25.1 1.801 4.308e-13 16.7 1.322
 [1] Zhengshan Dong, Jianli Chen, Wenxing Zhu. Homotopy method for matrix rank minimization based on the matrix hard thresholding method. Numerical Algebra, Control & Optimization, 2019, 9 (2) : 211-224. doi: 10.3934/naco.2019015 [2] Yangyang Xu, Ruru Hao, Wotao Yin, Zhixun Su. Parallel matrix factorization for low-rank tensor completion. Inverse Problems & Imaging, 2015, 9 (2) : 601-624. doi: 10.3934/ipi.2015.9.601 [3] Tao Wu, Yu Lei, Jiao Shi, Maoguo Gong. An evolutionary multiobjective method for low-rank and sparse matrix decomposition. Big Data & Information Analytics, 2017, 2 (1) : 23-37. doi: 10.3934/bdia.2017006 [4] Meijuan Shang, Yanan Liu, Lingchen Kong, Xianchao Xiu, Ying Yang. Nonconvex mixed matrix minimization. Mathematical Foundations of Computing, 2019, 2 (2) : 107-126. doi: 10.3934/mfc.2019009 [5] Yu-Ning Yang, Su Zhang. On linear convergence of projected gradient method for a class of affine rank minimization problems. Journal of Industrial & Management Optimization, 2016, 12 (4) : 1507-1519. doi: 10.3934/jimo.2016.12.1507 [6] Xianchao Xiu, Lingchen Kong. Rank-one and sparse matrix decomposition for dynamic MRI. Numerical Algebra, Control & Optimization, 2015, 5 (2) : 127-134. doi: 10.3934/naco.2015.5.127 [7] Yonggui Zhu, Yuying Shi, Bin Zhang, Xinyan Yu. Weighted-average alternating minimization method for magnetic resonance image reconstruction based on compressive sensing. Inverse Problems & Imaging, 2014, 8 (3) : 925-937. doi: 10.3934/ipi.2014.8.925 [8] Giuseppe Geymonat, Françoise Krasucki. Hodge decomposition for symmetric matrix fields and the elasticity complex in Lipschitz domains. Communications on Pure & Applied Analysis, 2009, 8 (1) : 295-309. doi: 10.3934/cpaa.2009.8.295 [9] Duo Wang, Zheng-Fen Jin, Youlin Shang. A penalty decomposition method for nuclear norm minimization with l1 norm fidelity term. Evolution Equations & Control Theory, 2019, 8 (4) : 695-708. doi: 10.3934/eect.2019034 [10] Vladimir Gaitsgory, Tanya Tarnopolskaya. Threshold value of the penalty parameter in the minimization of $L_1$-penalized conditional value-at-risk. Journal of Industrial & Management Optimization, 2013, 9 (1) : 191-204. doi: 10.3934/jimo.2013.9.191 [11] Yun Cai, Song Li. Convergence and stability of iteratively reweighted least squares for low-rank matrix recovery. Inverse Problems & Imaging, 2017, 11 (4) : 643-661. doi: 10.3934/ipi.2017030 [12] Yifu Feng, Min Zhang. A $p$-spherical section property for matrix Schatten-$p$ quasi-norm minimization. Journal of Industrial & Management Optimization, 2017, 13 (5) : 1-11. doi: 10.3934/jimo.2018159 [13] Jiying Liu, Jubo Zhu, Fengxia Yan, Zenghui Zhang. Compressive sampling and $l_1$ minimization for SAR imaging with low sampling rate. Inverse Problems & Imaging, 2013, 7 (4) : 1295-1305. doi: 10.3934/ipi.2013.7.1295 [14] Weinan E, Weiguo Gao. Orbital minimization with localization. Discrete & Continuous Dynamical Systems - A, 2009, 23 (1&2) : 249-264. doi: 10.3934/dcds.2009.23.249 [15] Yi Yang, Jianwei Ma, Stanley Osher. Seismic data reconstruction via matrix completion. Inverse Problems & Imaging, 2013, 7 (4) : 1379-1392. doi: 10.3934/ipi.2013.7.1379 [16] Quanyi Liang, Kairong Liu, Gang Meng, Zhikun She. Minimization of the lowest eigenvalue for a vibrating beam. Discrete & Continuous Dynamical Systems - A, 2018, 38 (4) : 2079-2092. doi: 10.3934/dcds.2018085 [17] Zhouchen Lin. A review on low-rank models in data analysis. Big Data & Information Analytics, 2016, 1 (2&3) : 139-161. doi: 10.3934/bdia.2016001 [18] Mehdi Badsi, Martin Campos Pinto, Bruno Després. A minimization formulation of a bi-kinetic sheath. Kinetic & Related Models, 2016, 9 (4) : 621-656. doi: 10.3934/krm.2016010 [19] M. Zuhair Nashed, Alexandru Tamasan. Structural stability in a minimization problem and applications to conductivity imaging. Inverse Problems & Imaging, 2011, 5 (1) : 219-236. doi: 10.3934/ipi.2011.5.219 [20] María Andrea Arias Serna, María Eugenia Puerta Yepes, César Edinson Escalante Coterio, Gerardo Arango Ospina. $(Q,r)$ Model with $CVaR_α$ of costs minimization. Journal of Industrial & Management Optimization, 2017, 13 (1) : 135-146. doi: 10.3934/jimo.2016008

2018 Impact Factor: 0.545

## Tools

Article outline

Figures and Tables