多模态函数聚类后再创种群的并行搜索佳点集萤火虫算法  

A Parallel Search Good-Point Set Glowworm Swarm Optimization of Re-created Population after Clustering of Multi-Modal Functions

作　　者：方贤 铁治欣[1] 李敬明[2] 高雄 

机构地区：[1]浙江理工大学信息学院,杭州310018 [2]合肥工业大学管理学院,合肥230009

出　　处：《浙江理工大学学报（自然科学版）》2017年第6期843-850,共8页Journal of Zhejiang Sci-Tech University(Natural Sciences)

基　　金：国家自然科学基金项目(61170110);浙江省公益技术应用研究项目(2014C31G2060072);安徽省教育厅自然科学研究重点项目(KJ2016A308)

摘　　要：萤火虫算法在求解多模态函数时,随着峰值个数的增加,往往需要更大的种群规模才能得到较为理想的结果,而且初始种群是否均匀分布对结果也有很大影响。针对萤火虫算法的这些不足,提出了一种多模态函数的聚类后再创种群的并行搜索佳点集萤火虫算法。该算法首先以数论佳点集的思想将萤火虫均匀分布于搜索空间中,在粗糙搜索完成后,通过密度聚类算法进行捕峰操作,重新构造等同于峰值点数的各个平行空间;然后在各空间中继续加入少量佳点集生成的萤火虫并行精细搜索,最终可获得各个平行空间的局部最优解以及整个空间的全局最优解。与其他算法在12个典型多模态函数中的测试结果进行对比,该算法总体上缩小了种群规模,加快了收敛速度,搜索精度更高,时间成本更低,稳定性能更好。In the event that the glowworm swarm optimization algorithm is used to solve multi-modal function,a larger population size is usually needed to obtain an ideal result as the number of peak values increases.In addition,the uniform distribution of the initial population also has a great influence on the results.In view of this,aparallel search good-point set glowworm swarm optimization based on re-created population after clustering of multi-modal functions（PGSGSO）is proposed in this paper.Firstly,the glowworms are evenly distributed in the search space based on the good-point set theory,and the function peaks are captured with density-based clustering algorithm after rough search is finished,to recreate parallel spaces with the same number of peak values;a small amount of glowworms are added into the spaces for fine search to obtain the locally optimal solution of the parallel spaces and the globally optimal solution of the whole space.The comparison with the test results of other algorithms in 12 typical multimodal functions show that the proposed algorithm is superior in respect of population size,convergence speed,searching precision,time cost and stability.

关 键 词：萤火虫算法 多模态函数 佳点集 密度聚类算法 并行搜索 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]