摘要
针对粒子群算法易陷入局部极值和早熟收敛的缺陷,提出了基于q-高斯分布的自适应变异粒子群算法.采用q-高斯作为变异算子对粒子的全局最优位置进行q-高斯变异,克服了因种群遗失多样性所导致的早熟收敛缺陷,随着种群的进化,非广延熵指数q的自适应调整平衡了算法的全局搜索能力和局部开发能力.测试了4个标准复杂函数和优化BP神经网络参数,结果表明,基于q-高斯分布的自适应变异粒子群算法的优化性能最好,收敛速度快.
Aiming at the disadvantages that the particle swarm algorithm is easy to run into the local extremum and premature convergence,a particle swarm algorithm with adaptive mutation based on q-Gaussian distribution was proposed.q-Gaussian was taken as the mutation operator to carry out the q-Gaussian mutation for the global optimal position of particles.Thus,the premature convergence caused by the loss of population diversity is overcome.With the evolution of population,the adaptive adjustment of non-extensive entropic index q balances the global searching ability and local development ability of the algorithm.In addition,four standard complex functions were tested,and the parameters of BP neural network were optimized.The results show that the particle swarm algorithm with adaptive mutation based on q-Gaussian distribution has the best optimization performance and fast convergence speed.
出处
《沈阳工业大学学报》
EI
CAS
北大核心
2012年第3期354-360,共7页
Journal of Shenyang University of Technology
基金
国家自然科学基金资助项目(60474069)
关键词
粒子群算法
自适应变异
q-高斯分布
数值优化
神经网络参数优化
种群多样性
全局搜索能力
局部搜索能力
particle swarm algorithm
adaptive mutation
q-Gaussian distribution
numerical optimization
neural network parameter optimization
population diversity
global searching ability
local searching ability
