基于庄家法则和信息熵的多目标进化算法

针对庄家算法的缺陷,提出了一种基于信息熵的庄家算法。其基本思想是：在使用庄家算法进行非支配解的选取前,先对群体的信息熵值进行计算。若熵值较低,即没有相对较好的分布度,则对群体进行遗传选择、交叉和变异操作,生成新的群体,直到熵值达到要求,再使用庄家法则进行计算。数值计算表明,这种新的算法既保持了庄家算法较高的收敛速度,又改善了群体的分布度,提高了种群的多样性,避免了过早收敛于局部最优解的现象。

Multi-Objective Evolutionary Algorithm Based on Dealer Principle and Information Entropy

The paper presents a multi-objective evolutionary algorithm based on dealer principle and information entropy for overcoming the defects of dealer principle.The basic idea of new method is to figure out the entropy value of each individual before the construction of non-dominated set.If the entropy value is small,which means the individuals are not well distributed,the population should be genetic select,crossover,mutate to make a new population.If the entropy value is high enough,the population should be deal by dealer principle.Numeric results show that this new algorithm method not only improves the rate of convergence but also avoids the premature convergence.