动态概率粒子群优化模型及实验分析  

Experiment and Analysis on Dynamic Probabilistic Particle Swarm Optimization Model

作　　者：倪庆剑[1] 邢汉承[1] 张志政[1,2] 王蓁蓁[1] 

机构地区：[1]东南大学计算机科学与工程学院,南京210096 [2]南京大学计算机软件新技术国家重点实验室,南京210093

出　　处：《计算机科学》2009年第2期222-226,共5页Computer Science

基　　金：国家自然科学基金重大研究计划项目(90412014)资助

摘　　要：对传统PSO算法中种群产生方式的特点,结合根据历史信息直接取样生成新种群的思想,抽象出动态概率粒子群优化(DPPSO)模型,并给出了该模型的形式化描述;同时提出了DPPSO模型可以采用的几种动态概率进化算子,最后通过常用Benchmark函数优化问题对DPPSO模型采用不同进化算子时的性能进行了实验分析。实验结果验证了DPPSO模型及所提出的进化算子的有效性,同时根据实验结果提出了进化算子设计与选择的指导性建议,并对相关参数设置也做了分析和讨论。The dynamic probabilistic particle optimization model （DPPSO） was constructed based on the new inspiration of population generation method according to historical information about particles, and the formal description of DPPSO model was proposed, several dynamic probabilistie evolution operators were also designed for realization of DPPSO model. DPPSO algorithms with several evolution operators were tested on five benchmark functions which are common in evolutionary computation. Experiment results demonstrate the effectiveness of DPPSO model and the proposed evolution operators. The suggestion on designing evolution operators was provided based on experimental data and some configurations of relative parameter setting were also discussed and recommended.

关 键 词：群智能 动态概率粒子群优化模型 种群生成方式 速度属性 动态概率进化算子 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]