可拓聚类适应度共享小生境遗传算法研究

Research of fitness sharing niche genetic algorithms based on extension clustering

摘要:

针对遗传算法易陷入早熟收敛和全局搜索能力差等缺点，提出一种基于可拓理论的小生境遗传算法.算法首先构造了遗传编码物元和可拓遗传算子，然后通过可拓聚类方法实现小生境群体的划分，结合适应度共享技术和聚类代表个体保存策略，维持稳定多样的小生境.仿真实验表明，该算法能可靠、快速地收敛到全局最优解，有效避免早熟收敛，其收敛速度和求解精度均优于简单遗传算法和常规小生境算法.

Abstract:

To solve the problems of premature convergence and weak ability in global search of the genetic algorithm, a fitness sharing niche genetic algorithm based on extenics is proposed.The algorithm build the matter-element code and extension genetic operator, create niche groups by extension clustering, and preserve the stability of niche groups by combining fitness sharing mechanism and elitist retention strategy. Experiments show that the algorithm can solve the optimal performance with global search ability and fast convergence rate. It is proved to be more effective and accurate than standard geneic algorithm and normal niche genetic algorithm.