摘要
针对具有可分结构的凸极小化问题,提出了一种部分并行的可分方法.该方法是在预校正近似乘子法的基础之上,在极小化时采取了不同的格式,去掉了二次邻近项而直接用的增广项;在算法的迭代部分,预校正近似乘子法先计算x^(k+1),再计算z^(k+1),在部分并行的可分方法中,x^(k+1),z^(k+1)是并行计算的;通过数值算例得到的结果显示,该方法具有可行性.
In view of convex minimization problem with separable structure, this paper presents a separable method of partially parallel, and this method is evolved by the predictor-corrector proximal multiplier method. A different format is used in the process of minimization, the quadratic adjacent items are replaced but the augmented items are directly used in the method. Numerical example results show that this method is feasible.
出处
《重庆工商大学学报(自然科学版)》
2017年第2期16-21,共6页
Journal of Chongqing Technology and Business University:Natural Science Edition
基金
国家自然科学基金(61263020)
关键词
凸优化问题
交替方向乘子法
预校正近似乘子法
部分并行的可分方法
convex optimization problem
alternating direction multiplier method
predictor-corrector proximal multiplier method
partially parallel of separable method