曲线递增策略的自适应粒子群算法研究  

Research on adaptive particle swarm optimization algorithm with curve increasing strategy

作　　者：吴凡[1] 洪思 杨冰 胡贤夫 Wu Fan;Hong Si;Yang Bing;Hu Xianfu(School of Management Science&Engineering,Anhui University of Technology,Ma’anshan Anhui 243032,China;Qiutai Microelectronics Technology Co.Ltd.,Kunshan Jiangsu 215300,China)

机构地区：[1]安徽工业大学管理科学与工程学院,安徽马鞍山243032 [2]丘钛微电子科技有限公司,江苏昆山215300

出　　处：《计算机应用研究》2021年第6期1653-1656,1661,共5页Application Research of Computers

基　　金：国家自然科学基金资助项目(61702006)。

摘　　要：群智能算法以其动态寻优能力强、实现途径简单等特点不断成为进化算法领域的研究热点。控参的选择对算法寻优性能有着极大影响,首先从数学推导角度对粒子群参数进行深入研究,接着提出一种契合粒子本身进化公式的,且具有反向思维的曲线递增策略的改进算法。最后验证该算法具备以下两点突出优势:a)有效避免早熟问题,在处理维度灾难问题上,寻优性能更强,且具备良好的平衡全局与局部寻优性能;b)算法控参简单,可有效解决鲁棒性低且繁琐的人工调参问题。Swarm intelligence algorithm has become a hotspot in the field of evolutionary algorithm because of its strong dynamic optimization ability and simple implementation approach.The choice of control parameters has a great influence on the optimization performance of the algorithm.Firstly,this paper studied the particle swarm parameters from the point of mathematical derivation,then proposed an improved algorithm with inverse thinking curve increment strategy,which fitted the evolution formula of the particle itself.Finally,it verified that the algorithm has the following two outstanding advantages.It can effectively avoid the problem of precocity.In dealing with the problem of dimensional disaster,the optimization performance is stronger and has a good balance between global and local optimization performance.The algorithm is simple and can effectively solve the problem of low robustness and tedious manual parameter adjustment.

关 键 词：控参 数学推导 曲线递增 粒子群算法 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]