JIN Qinglong,GAO Qianming.A partially parallel ADMM algorithm with relaxed parameters[J].Journal of Yanbian University,2021,47(03):216-221.
一种放松参数的部分并行ADMM算法
- Title:
- A partially parallel ADMM algorithm with relaxed parameters
- 文章编号:
- 1004-4353(2021)03-0216-06
- Keywords:
- separable convex optimization; alternating direction multiplier method; partial parallel splitting; variational inequality; multiple blocks
- 分类号:
- O159
- 文献标志码:
- A
- 摘要:
- 为了求解多块线性约束可分凸优化问题,提出了一种放松参数的部分并行交替方向乘子法(ADMM)算法—PPADMMR算法.该算法在子问题中引入带参数的临近项,放松了临近参数范围.数值实验表明,PPADMMR算法的收敛速度优于部分并行ADMM(PPADMM)算法,因此提出的PPADMMR算法可为研究快速ADMM算法提供参考.
- Abstract:
- To solve multi - block linearly constrained separable convex optimization problems, we propose a partially parallel alternating direction multiplier method with relaxed parameters(PPADMMR). The algorithm introduces the proximal point terms with parameters into the sub - problems and relaxes the range of parameters. Numerical experiments show that the convergence speed of the new algorithm is better than that of PPADMM. Therefore, the proposed PPADMMR algorithm in this paper can provide an excellent reference to investigate the faster ADMM algorithm.
参考文献/References:
[1] BOYD S, PARIKH N, CHU E, et al.Distributed optimization and statistical learning via the alternating direction method of multipliers[J].Found Trends Mach Learn, 2010,3(1):1-122.
[2] BOSE N K, BOO K J.High - resolution image reconstruction with multisensors[J].Int J Imag Syst Tech, 1998,9(4):294-304.
[3] RIBSHIRANI R, SAUNDERS M, ROSSET S, et al.Sparsity and smoothness via the fused lasso[J].J Roy Stat Soc, 2010,67(1):91-108.
[4] YE Y Y, CHEN C H, HE B S, et al.The direct extension of ADMM for multi - block convex minimization problems is not necessarily convergent[J].Math Program, 2016,155(1/2):57-79.
[5] 陈彩华.多块交替方向乘子法不收敛反例的几点注记[J].运筹学学报,2019,23(3):135-140.
[6] HE B S, TAO M,YUAN X M.A splitting method for separable convex programming[J].Ima J Numer Anal, 2013(1):394-426.
[7] HOU L S, HE H J, YANG J F.A partially parallel splitting method for multiple - block separable convex programming with applications to robust PCA[J].Comput Optim Appl, 2016,63(1):273-303.
备注/Memo
收稿日期: 2021-07-06
基金项目: 江苏省研究生科研与实践创新计划(KYCX20_1321)
作者简介: 金青龙(1996—),男,硕士,研究方向为最优化算法理论与设计.