嵌入共轭梯度法的混合粒子群优化算法  

Hybrid PSO Algorithm with Conjugate Gradient Methods

作　　者：梁昔明[1] 李德生[1] 

机构地区：[1]北京建筑大学理学院,北京100044

出　　处：《小型微型计算机系统》2014年第4期835-839,共5页Journal of Chinese Computer Systems

基　　金：北京自然科学基金项目(4122022)资助;北京市属高等学校人才强教计划项目(201107123)资助;北京建筑工程学院博士启动基金项目(331612018)资助

摘　　要：针对标准粒子群算法收敛速度慢和精度低的问题,提出了一种嵌入共轭梯度法的混合粒子群优化算法.算法有效结合了粒子群优化算法较强的全局搜索能力和共轭梯度法快速精细的局部搜索能力,在基本粒子群算法得到的最优解的基础上引入共轭梯度法,加快了算法的收敛速度,克服了基本粒子群算法收敛慢的弊端.相比于基本粒子群算法,它能够以较高精度和较快速度收敛到所求无约束优化问题的全局最优解.数值实验结果表明,所得混合算法是一种求解高维多峰连续函数无约束优化问题的高效方法.A kind of hybrid Particle Swarm Optimization（PSO） algorithm coupling with conjugate gradient method is proposed to o- vercome the slow speed convergence and low accuracy computation in basic PSO. The algorithm effectively combines the strongly global search ability of PSO with the fast local search ability of conjugate gradient method. Quoting conjugate gradient method based on optimal solution of basic PSO to overcome the slow speed convergence in basic PSO. Compared to the basic PSO, the proposed al- gorithm can take a higher precision and faster convergence speed to obtain the global minimization of a given unconstrained optimiza- tion problem. Numerical experiment results show that the hybrid PSO algorithm is a kind of efficient method for solving the uncon- strationed optimization problem on high-dimension multi-modal continuous function.

关 键 词：混合粒子群优化算法 共轭梯度法 无约束优化问题 数值实验 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]