参数型Topkis-Veinott方法及其数值计算

A parameter-type Topkis-Veinott Method and its numerical experiment

讨论一种解决不等式约束优化问题的T opk is-V einott方法的变化形式。在每次迭代时,这种方法都利用一种线性约束半正定二次问题来产生一个合理的下降方向;同时,用半光滑牛顿方法去解出产生于非线性补充问题(NCP函数)的子问题。根据给定算法用MATLAB语言编写程序。初步的数值结果表明,参数c影响着算法的速度。

In this paper, a variant of the Topkis-Veinott method for solving inequality constrained optimization problem is discussed. At each iteration, the method uses a linearly constrained positive semidefinite quadratic problem to generate a feasible descent direction. We use the Semismooth Newton Method to solve the subproblem which is derived from computing nonlinear complementary problem（NCP-function）. According to the given algorithm, we give the procedure which is written by MATLAB. The convergence velocity of the method is affeeted by the chosen of parameterc.