一种高效自学习性回溯搜索优化算法

Effective Self-learning Backtracking Search Optimization Algorithm

针对回溯搜索优化算法收敛速度慢的问题,文中在理论分析的基础上,分别对变异算子和交叉算子进行了改进。设计了双种群引导形式的变异算子,并引入一种基于麦克斯韦-玻尔兹曼分布变异尺度系数,有效地增加了变异方程的搜索效率,并设计了一种带有自学习性的交叉策略,使得回溯搜索优化算法的收敛速度和全局搜索性能均得到了一定的提高,通过数值实验,说明了改进的有效性。

For slow convergence speedof Backtracking Search Optimization Algorithm（ BSA）,this paper makes some improvements on mutation operator and crossover operator on base of theoretical analysis. Firstly,a mutation operator with two-population guided form is designed,anda novel mutation scale factor based on Maxwell-Boltzmann distribution is introduced, which enhance search efficiency of mutation equation effectively. Secondly, crossover strategy is designed with self-learning property,both them enhance the performance of BSA,and numericalexperiments for testing the improved BSA are given in the end.