一般约束最优化强收敛的拟乘子-强次可行方向法

A QUASI-MULTIPLIER-STRONGLY SUBFEASIBLE DIRECTIONS METHOD WITH STRONG CONVERGENCE FOR GENERAL CONSTRAINED OPTIMIZATION

本文讨论一般等式和不等式约束优化问题 ,利用广义投影技术和强次可行方向法思想 ,结合拟 K-T点和拟乘子法 [1] 两个新概念 ,建立问题一个初始点任意的有显式搜索方向的新算法 .证明算法不仅收敛到原问题的拟 K- T点 ,且具有更好的强收敛性 .对算法进行了一定的数值试验 .

In this paper,optimization problem with general equality and inequality constraints are discussed.A new algorithm with arbitrary initial point is presented by using the generalized projection technique and the idea of strongly subfeasible directions method and combining with the two concepts of quasi K T point and quasi multipier methods.The algorithm is proved to a quasi K T point of the problem,and strongly subfeasible directions method is abstained.