求全局极小的“塌方”方法

A"Landslip"Method for Finding A Global Minimum of Multimodal Functions

对求全局极小的方法进行了研究，构造出寻优算法，把寻找全局极小的过程分解为局部寻优过程和突破过程的反复。局部寻优过程可采用一般寻局部极小的任一种优化算法，突破过程通过引入广义实力势函数和寻优过程的结合而实现，实例计算结果表明算法有效可行。

This paper studies one of the most important problems in nonlinear programing to find cutglobal minimum of a given objective function.In this paper,a new'landslip'method is proposed.The'landslip'method is composed of two algorithms.One is the local optopmization algorithm.The other isalgorithm which breaks through local minimum point.The global minimum point can be obtained by usingthetwo algorithms repeatedly.In former algorithm Broyden-Fletcher-Goldfarb-Shanno(BFGS)orDavidon-Fletcher-Powell(DPF)method can be used.In later algorithm,the general potential function (notobjective function-potentail function)is minimized,The Results obtained imply the'landslip'method iseffective and stable.