融合均值榜样的反向互学习水母搜索算法

Oppositional-mutual learning jellyfish search algorithm based on mean-value example

为解决水母搜索算法(jellyfish search algorithm,JS)的洋流运动缺乏多样性、群内运动缺乏引导性、种群间信息无交流,造成搜索速度慢、稳定性差及易早熟的问题,构建了一种融合均值榜样的反向互学习水母搜索算法(oppositional-mutual learning jellyfish search algorithm based on mean-value example,OMLJS).首先在水母跟随洋流运动(全局搜索)部分,利用前两代水母的平均位置代替只考虑上一代水母的平均位置来引导水母个体的位置更新,提高算法的全局搜索能力;其次在水母的群内主动运动(局部搜索)部分,利用最优个体代替随机个体来引导水母进行更有效的搜索,加快算法的收敛速度;然后在水母进入下一次迭代前增加对水母种群进行动态反向互学习步骤,增加种群多样性及增强种群间的信息交流,达到互补另外两个策略,提高算法的整体优化性能.选用12个经典的基准测试优化函数,将OMLJS与5个对比算法从解的平均值、最优值及方差进行对比分析,并用于求解最小生成树问题,OMLJS能够更快地找到最小生成树.实验结果表明,OMLJS的收敛速度、求解精度明显提高.

In order to solve the problems that the lack of divercity of the jellyfish search algorithm(JS)in ocean current motion,the lack of guidance in intra-swarm motion,and no exchange of information between populations,which result in slow search speed,poor stability,and susceptibility to precocious maturation,an oppositional-mutual learning jellyfish search algorithm based on mean-value example(OMLJS)is constructed.Firstly,in the part of the jellyfish following ocean currents(global search),the average position of the previous two generations of jellyfish is used instead of only considering the average position of the previous generation to guide the position update of individual jellyfish,which improves the global search ability of the algorithm;secondly,in the part of jellyfish's intra-swarm active movement(local search),the optimal individual is used instead of random individual to guide the jellyfish to perform more effective search,which accelerates the convergence speed of the algorithm;Then,before the jellyfish enters the next iteration,a dynamic reverse mutual learning step is added to the jellyfish population to increase the diversity of the population and enhance the information exchange between the populations,so as to complement the other two strategies and improve the overall optimization performance of the algorithm.Twelve classical benchmark test optimization functions are selected to compare and analyze OMLJS with five comparative algorithms in terms of the mean value,optimal value and variance of the solutions,and are used to solve the minimum spanning tree problem,where OMLJS is able to find the minimum spanning tree faster.The experimental results show that the convergence speed and solution accuracy of OMLJS are significantly improved.