基于信息熵的变异算子设计

Design on a Mutation Operator Based on Entropy

基因多样性是影响算法收敛性能的重要因素之一,为了维持种群足够的多样性以保证演化高效进行,从变异算子设计思想入手,分析了传统变异算子存在的不足,即固定的、与问题无关的变异算子的设计思想存在不足。提出了基于信息熵的变异算子,并讨论了它在克服早熟收敛方面的作用;最后,通过实验证明了该方法的有效性。

The diversity of genes is one of important factors that affect the convergence property of genetic algorithm. To maintain sufficient diversity among individuals in order to enable further evolvement, this paper analyzes the traditional mutation operator of GAs in design idea of mutation operator, and the analysis shows that the design idea has some disadvantages. That is, the design idea of mutation operator that is stochastically independent and occurs with fixed probability is not perfect. The mutation operator based on information entropy is presented to take the place of the traditional one. The contribution of mutation operator based on information entropy on overcoming premature convergence is also discussed. The experimental results are given to demonstrate the effectiveness of the proposed method.