摘要
在分析量子行为粒子群算法中吸引子指导作用的基础上,引入两种精英学习策略,提出了基于精英学习的量子粒子群算法(QPSO-EL).采用动态逼近学习策略对精英个体进行局部更新,协助其跳出自身局部极值点,引导种群进行有效搜索;借鉴群体早熟判断机制对停滞状态下的精英个体空间进行变尺度混沌扰动,增大种群全局搜索空间,有效平衡了算法的局部和全局搜索能力.典型函数的仿真结果表明,该算法具有收敛速度快、求解精度高的特点.
The local attractor point in the quantum-behaved particle swarm optimization algorithm plays an important role in determining the convergence process of population.Therefore,a quantum-behaved particle swarm optimization algorithm based on two elitist learning strategys（QPSO-EL） is presented.In this method,the dynamic-approximation search strategy is exerted on the elitist particles to avoid them running into local optima and provides a good guidance for the population.While the algorithm is found to be in a dead state according to the premature judgment mechanism,the mutative-scale chaotic perturbation is used to exhibit a wide range exploration and keep the balance of exploration and exploitation.The experiment results on classic functions demonstrate the global convergence ability and the search accuracy of the proposed method.
出处
《控制与决策》
EI
CSCD
北大核心
2013年第9期1341-1348,共8页
Control and Decision
基金
国家科技支撑计划项目(2009BAC56B03)
关键词
量子行为粒子群
精英学习
动态逼近搜索
变尺度混沌扰动
quantum-behaved particle swarm optimization
elitist learning
dynamic-approximation research
mutative scale chaotic perturbation
