一种基于距离度量的自适应粒子群优化算法  

Distance Measurement Based Adaptive Particle Swarm Optimization

作　　者：李太勇[1,2] 吴江[1,2] 朱波[3] 方冰[4] 

机构地区：[1]西南财经大学经济信息工程学院,成都610074 [2]西南财经大学中国支付体系研究中心,成都610074 [3]西南财经大学金融学院,成都610074 [4]贵阳学院计算机科学系,贵阳550005

出　　处：《计算机科学》2010年第10期214-216,共3页Computer Science

基　　金：国家自然科学基金(60773169);西南财经大学金融智能与金融工程重点实验室校内公开项目(FIFE2010-P02)资助

摘　　要：惯性权值对粒子群优化(Particle Swarm Opti mization,PSO)算法的性能起着重要作用。基本的PSO算法未考虑各粒子的差异而在一次迭代中所有粒子采用固定的惯性权值。为了体现各粒子相对于已知最优解的差异,提出了一种基于距离度量的自适应PSO算法DMAPSO(Distance Measurement-based Adaptive PSO)。算法采用欧式距离计算粒子与已知全局最优粒子的差异,然后根据差异自适应调整各粒子的惯性权值。通过基准测试函数对算法进行了实验,结果表明,对于连续函数优化问题,提出的DMAPSO算法优于经典PSO算法,DMAPSO收敛到最优解的迭代次数比PSO平均减少了约60%。The inertia weight plays an important role in Particle Swarm Optimization（PSO）.The classical PSO used a fixed inertia weight for all particles in an iteration and ignored the difference among the particles.To cope with this issue,a Distance Measurement based Adaptive Particle Swarm Optimization（DMAPSO）was proposed.The Euclidean distance was used to calculate the difference between a particle and the known best global particle,and the particle tuned adaptively the value of the inertia weight according to the difference.Several classical benchmark functions were used to evaluate the strategy.The experimental results show that for continuous optimization problems,the DMAPSO outperforms the classical PSO.The iteration times for finding the best solutions in the DMAPSO decrease about 60% averagely compared with that in the classical PSO.

关 键 词：粒子群 优化算法 惯性权值 距离度量 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]