基于信息熵对遗传算法中杂交概率的研究

针对现有杂交概率的计算方法复杂且不利于种群摆脱局部优现象,提出了基于信息熵的杂交概率计算方法。利用种群熵和种群方差来分析杂交算子在种群进化中的作用,充分考虑了种群的整体情况和进化潜力,从而确定杂交概率的计算,以更好地控制遗传算法的进化过程。数值实验表明,新提出的杂交概率计算方式不仅便于求解,而且能有效地增强算法的稳定性、全局收敛性,加快算法收敛速度,使算法易于摆脱局部优现象。

Study on crossover probability based on information entropy in genetic algorithms

To overcome the shortcomings of complex computation and easily falling into local optimums for the existing approaches to compute crossover probability,a novel method based on information entropy is presented.The entropy of the population shows the individuals’ evolving potential,and the variance of the population reflects the whole population distribution.The role of the crossover operator in the procedure of population evolving through the entropy of the population and the variance of the population is analyzed.Then,a new computing method is designed for crossover probability so as to get the better control of the evolving process of genetic algorithms.Numerical examples show the method of computing crossover probability can not only be used conveniently but also enhance the robustness,global optimization ability and the convergence speed of the genetic algorithm.The genetic algorithm with the new crossover probability can escape local optimums easily.