粒子群优化算法位置更新方程的统一及简化  

Unification and simplification for position updating formulas in particle swarm optimization

作　　者：胡建[1] 胡旺[2] 殷锋[3] 

机构地区：[1]西南民族大学西南民族研究院,成都610041 [2]电子科技大学计算机科学与工程学院,成都610054 [3]西南民族大学计算机科学与技术学院,成都610041

出　　处：《中国科学：信息科学》2016年第11期1676-1692,共17页Scientia Sinica(Informationis)

基　　金：教育部人文社会科学研究青年基金项目(批准号:15YJC870010);四川省科技厅科技支撑项目(批准号:2014SZ0104);四川省教育厅资助科研项目重点项目(批准号:16ZA0012);西南民族大学研究生学位点建设项目(批准号:2016-XWD-B0304)资助

摘　　要：粒子群优化算法的理论分析和避免早熟一直是被重点研究的两个问题,但前者因复杂的动态性而不得不在简化的系统条件下进行,后者因不可避免地引入形式多样的操作算子而增加了算法复杂性,进而使得理论分析更加困难.对此,本文整理归纳出了大多数现有改进算法的位置更新方程之间的共性规律,给出了统一形式,并由多阶随机差分方程简化为一阶随机差分方程,使得粒子搜索行为控制和收敛性分析更为容易.实验在具有代表性的算法上进行,验证了对位置更新方程的统一和简化过程的合理性,并表明本文方法性能更具有竞争力.Particle swarm optimization（PSO） has been a popular research area in artificial intelligence technology,where the two issues of theoretical analysis and premature convergence have been the focus of attention.However,due to the complex dynamics of a particle swarm,the former was conducted only for simplified systems.The latter has only been addressed by introducing some additional operations,which inevitably increases the complexity of PSO and complicates the theoretical analysis.This paper proposes a unified and simplified rule for position updating in the existing algorithms as an attempt to solve the above-mentioned problems.This rule simplifies the multiple-order stochastic difference equation to a first-order stochastic difference equation,and facilitates the analysis of the convergence and control of the search behavior of particles.Experiments were conducted on some representative algorithms,and the results verified the correctness of the unification and simplification of position updating formulas,which also performed more competitively.

关 键 词：粒子群优化 群体智能 人工智能 进化计算 收敛性 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]