Multi-block Nonconvex Nonsmooth Proximal ADMM: Convergence and Rates Under Kurdyka-Lojasiewicz Property

被引:14
作者
Yashtini, Maryam [1 ]
机构
[1] Georgetown Univ, Dept Math & Stat, 327A St Marys Hall 37th & O St NW, Washington, DC 20057 USA
关键词
Nonconvex nonsmooth optimization; Proximal ADMM; Kurdyka-Lojasiewicz property; Convergence; Convergence rates; ALTERNATING DIRECTION METHOD; VARIABLE SELECTION; MINIMIZATION; OPTIMIZATION; ALGORITHMS; REGULARIZATION; DECOMPOSITION; MULTIPLIERS; STEPSIZE;
D O I
10.1007/s10957-021-01919-7
中图分类号
C93 [管理学]; O22 [运筹学];
学科分类号
070105 ; 12 ; 1201 ; 1202 ; 120202 ;
摘要
We study the convergence and convergence rates of a multi-block proximal alternating direction method of multipliers (PADMM) for solving linearly constrained separable nonconvex nonsmooth optimization problems. This algorithm is an important variant of the alternating direction method of multipliers (ADMM) which includes a proximal term in each subproblem, to cancel out complicated terms in applications where subproblems are not easy to solve or do not admit a simple closed form solution. We consider an over-relaxation step size in the dual update and provide a detailed proof of the convergence for any step size beta is an element of(0,2). We prove the convergence of the sequence generated by the PADMM by showing that it has a finite length and it is Cauchy. Under the powerful Kurdyka-Lojasiewicz (KL) property, we establish the convergence rates for the values and the iterates, and we show that various values of KL-exponent associated with the objective function can raise PADMM with three different convergence rates. More precisely, we show that if the (KL) exponent theta=0, the sequence generated by PADMM converges in a finite numbers of iterations. If theta is an element of(0,1/2], then the sequential rate of convergence is cQk where c>0, Q is an element of(0,1), and k is an element of N is the iteration number. If theta is an element of(1/2,1], then O(1/kr) rate where r=(1-theta)/(2 theta-1) is achieved.
引用
收藏
页码:966 / 998
页数:33
相关论文
共 70 条
[31]   Fast Alternating Direction Optimization Methods [J].
Goldstein, Tom ;
O'Donoghue, Brendan ;
Setzer, Simon ;
Baraniuk, Richard .
SIAM JOURNAL ON IMAGING SCIENCES, 2014, 7 (03) :1588-1623
[32]  
Gu Y., ARXIV150602221
[33]   Convergence of alternating direction method for minimizing sum of two nonconvex functions with linear constraints [J].
Guo, K. ;
Han, D. R. ;
Wu, T. T. .
INTERNATIONAL JOURNAL OF COMPUTER MATHEMATICS, 2017, 94 (08) :1653-1669
[34]   An Alternating Direction Approximate Newton Algorithm for Ill-Conditioned Inverse Problems with Application to Parallel MRI [J].
Hager W. ;
Ngo C. ;
Yashtini M. ;
Zhang H.-C. .
Journal of the Operations Research Society of China, 2015, 3 (2) :139-162
[35]   AN O(1/k) CONVERGENCE RATE FOR THE VARIABLE STEPSIZE BREGMAN OPERATOR SPLITTING ALGORITHM [J].
Hager, William W. ;
Yashtini, Maryam ;
Zhang, Hongchao .
SIAM JOURNAL ON NUMERICAL ANALYSIS, 2016, 54 (03) :1535-1556
[36]   A Note on the Alternating Direction Method of Multipliers [J].
Han, Deren ;
Yuan, Xiaoming .
JOURNAL OF OPTIMIZATION THEORY AND APPLICATIONS, 2012, 155 (01) :227-238
[37]  
He B., 2010, ALTERNATING DIRECTIO
[38]   ON FULL JACOBIAN DECOMPOSITION OF THE AUGMENTED LAGRANGIAN METHOD FOR SEPARABLE CONVEX PROGRAMMING [J].
He, Bingsheng ;
Hou, Liusheng ;
Yuan, Xiaoming .
SIAM JOURNAL ON OPTIMIZATION, 2015, 25 (04) :2274-2312
[39]   On non-ergodic convergence rate of Douglas-Rachford alternating direction method of multipliers [J].
He, Bingsheng ;
Yuan, Xiaoming .
NUMERISCHE MATHEMATIK, 2015, 130 (03) :567-577
[40]   ALTERNATING DIRECTION METHOD WITH GAUSSIAN BACK SUBSTITUTION FOR SEPARABLE CONVEX PROGRAMMING [J].
He, Bingsheng ;
Tao, Min ;
Yuan, Xiaoming .
SIAM JOURNAL ON OPTIMIZATION, 2012, 22 (02) :313-340