针对具有可分结构的单调变分不等式问题,基于邻近点算法和文献[12]提出的下降型算法构造了一个新的下降方向,并利用下降量的下界来选择最优步长,提出一种下降型邻近点交替方向乘子法;证明了算法的收敛性;并将该方法与文献[11]中算法的下降量下界进行比较,从理论上说明了算法的优越性.
The problem of variational inequalities with separable structure was studied. A new descent direction was designed based on proximal point algorithm and decent algorithm hy reference [12] ,and appropriate step size in light of lower bound of slippage was selected. A descent proximal alternating direction method of multipliers was presen ted. In addition, the convergence of the algorithm was proved. The new method was compared on the lower bound of slippage with the one by reference [11]. The result shows that the new one is superior in theoretical senses.