变测度的积分型全局优化算法

Variable Measure of an Integral Global Optimization Algorithm

郑权等首先提出积分-水平集求总极值的方法,实现算法中采用Monte-Carlo 随机投点产生近似水平集来缩小搜索区域范围,但这一算法可能失去总极值点．此后,邬 冬华等给出了一种修正的积分－水平集的方法,一种区域不收缩的分箱方法以保证总极 值点不被丢失．本文在此基础上采取对不同的箱子采用不同的测度这一策略,使水平值 更充分的下降,更快的达到全局极小值,以提高修正算法的计算效率．最后给出的数值算 例说明了算法是有效的．

Professor Zhengquan firstly presented an integral-level set method to solve the global optimization. In implemental algorithm, he construct approximate level-set to reduce the field of search by Monte-Carlo Method , but this algorithm may loss the global optimization. Later, the modified algorithm was put forward by Dong-hua Wu and so on, which is a method of disparting the field of search and ensure that the optimal value couldn＇t be lost. This paper is based on the modified algorithm ,and takes different measure in different sub-box to make the level value reduce enough,then reachs the global optimization more quickly, so improving the effciency of the algorithm. Finally, the numerical value of example show that the algorithm is effictive.