中文摘要：

基于Glowinski的交替方向法和何炳生教授的改善步长的收缩算法。提出一个求解结构型变分不等式的加速随机方法．新方法的优势在于利用独立同分布的随机数来扩张步长，克服了传统的交替方向法中固定步长因子的缺点，证明了新方法的下降方向是可行的．在适当的假设条件下，给出新方法的性质，并证明新方法依概率收敛．通过对来自于金融和统计中问题的一系列数值试验，验证新方法的可行性和有效性．

英文摘要：

In this paper, we proposed an accelerated stochastic method for solving a class of structured variational inequalities, based on the alternating directions method of Glowinski Roland and the extended contraction method of HE Bingsheng. The new method resolved the inadequacy of fixed steplength in the classic alternating directions method and showed that descent direction is feasible. Meanwhile, the properties and convergence of the new method are proved under suitable conditions. Numerical results gained through solving the finance and statistics problem are provided to demonstrate that our method is feasible and effective.