多样性制导分段进化的基因表达式编程

Gene Expression Programming Based on Diversity-Guided Grading Evolution

为了解决基于传统基因表达式编程(GEP)的函数挖掘及其改进算法仍然存在局部优化的缺陷这一问题,提出了以基因组多样性制导的分阶段进化挖掘算法DG-GEP。给出了GEP进化阶段和基因组多样性评估模式的定义;提出了描述进化阶段的进化因子概念和分阶段进化策略;采用动态遗传算子设计和群体规模控制方法,使进化更快速跳出局部最优。实验表明了新算法的有效性,能减少进化停滞代数65%以上,使群体的平均适应度提高12%以上。

Function mining algorithm based on traditional Gene Expression Programming （GEP） and other improved algorithm may still lead to local optimum trap. To solve this problem, a new algorithm based on Genome Diversity-Guided （DG-GEP） in grading evolution was proposed. The definition of GEP evolution phase and genome diversity evaluation model were given. The concept of anagenesis factor to describe evolution phases and strategy of grading evolution were proposed. The means of dynamic genetic operators and population control were used to make the evolution escape from localization trap quickly. The experiment showed that the new algorithm decreases the generations-stagnancy over 65% and increases the average fitness of colony over 12%.