摘要
提出了一种新型的PSO算法———含维变异算子的粒子群算法(PSODMO),该算法在变异开始后到迭代结束之前的每一次迭代中,计算每一维的收敛度,以一定的概率对收敛度最小的维进行变异:让所有粒子在该维上的位置重新均匀分布在可行区域上.通过对4个多峰的测试函数所做的对比实验,表明所提出的PSODMO增强了全局搜索能力,搜索成功率大为提高,克服了原始的PSO算法易于收敛到局部最优的缺点.也明显优于对原始PSO进行传统变异的方法.
A new particle swarm optimization with dimension mutation operator (PSODMO) is presented. According to this algorithm, the degrees of convergence of every dimension are calculated in every iteration from the beginning of mutation; the dimension of minimal convergent degree is mutated according to some probability; the positions of all particles in this dimension are distributed in the range [-xmax,xmax] evenly. Comparative experiments on four multi-peak testing functions indicate that the PSODMO enhances the global searching ability and the probability of successful searching, and overcomes the original PSO's liability to convergence to local optimum. It is also superior to PSO with traditional mutation.
出处
《武汉大学学报(工学版)》
CAS
CSCD
北大核心
2005年第4期79-83,共5页
Engineering Journal of Wuhan University
基金
交通部博士基金项目(200332581106).
关键词
粒子群优化算法
维变异算子
全局最优
particle swarm optimization(PSO)
dimension mutation
global optimum