求解随机线性互补问题的Levenberg-Marquardt型算法

Levenberg-Marquardt Type Method for Stochastic Linear Complementarity Problem

针对随机线性互补问题的期望残差极小化模型,利用蒙特卡罗方法将其转化为有限个样本的近似问题.基于投影Levenberg-Marquardt算法,给出了求解近似问题的1种Levenberg-Marquardt型算法,证明了算法在一定条件下是全局收敛的.数值实验表明算法是有效的.

In this paper, we consider the expected residual minimization formulation of stochastic linear complementarity problem. By employing the Monte Carlo method, the expected residual minimization problem has been formulated as an approximate problem. Based on the projected Levenberg-Marquardt method, we propose a Levenberg Marquardt type method for sol ving the approximate problem. The global convergence of the proposed algorithm is proved under mild condition. Numerical re- suits show that our algorithm is efficient.