基于语义聚类的遗传规划算法比较
Comparison of genetic programming algorithms based on semantic clustering
王菁 1徐赐文 1吕林旺1
作者信息
- 1. 中央民族大学理学院,北京 100081
- 折叠
摘要
针对遗传规划算法容易陷入局部最优解与局部搜索过慢的问题,提出一种基于语义聚类的遗传规划算法(genetic programming algorithm based on semantic clustering,SCGP),比较不同聚类算法对SCGP表现的影响.同时提出一种基于子种群规模的自适应适应度函数,提高局部搜索能力.在多个基准问题上对比标准遗传规划、几何语义遗传规划、K均值聚类遗传规划与SCGP,实验结果表明,SCGP算法在拟合能力和泛化能力上都有较大改善.在诸多聚类方法中,层次聚类嵌入的SCGP算法在基准问题上的泛化能力最优,与标准遗传规划、几何语义遗传规划、K均值聚类遗传规划相比,分别提高了 32.36％、61.29％、20.53％.
Abstract
To address the problem that genetic programming algorithms tend to fall into local optimal solutions with slow local search,a genetic programming algorithm based on semantic clustering(SCGP)was proposed and the effects of different cluste-ring algorithms were compared on the performance of SCGP.An adaptive fitness function based on subpopulation size was also proposed to improve the local search ability.Experimental results show that the SCGP algorithm improves the fitting ability and generalization ability.Among the clustering methods,the SCGP algorithm with hierarchical clustering embedding has the best generalization ability on the benchmark problem and improves the performance compared with the standard genetic planning,geometric semantic genetic planning,K-mean clustering genetic planning by 32.36％,61.29％,and 20.53％,respectively.
关键词
遗传规划/聚类算法/进化算法/语义/自适应/子种群/算法比较Key words
genetic programming/clustering algorithm/evolutionary algorithm/semantics/adaptive/subpopulation/algorith-mic comparison引用本文复制引用
出版年
2024
NETL
NSTL
万方数据