计算机工程Issue(11)：155-159,166,6.DOI:10.3969/j.issn.1000-3428.2014.11.031
基于M-H采样的快速反向微分进化算法
Fast Opposition Differential Evolution Algorithm Based on M-H Sampling
摘要
Abstract
In Differential Evolution ( DE ) algorithm, the population initialization is updated by using opposition optimization rule,so as to maintain the population diversity. However,the reverse individuals are easy to deviate from the global optimal solution,has slow convergence speed and easy to fall into local optimum in function optimization. This paper proposes a fast Opposition Differential Evolution( ODE) algorithm based on M-H( Metropolis-Hastings) sampling method. M-H sampling is used in the mutation operation of ODE. M-H sampling satisfies Markov Chain reversible conditions. One step transition probability of Markov Chain is calculated according to the selecting probability of individual ranking-assignment,not only chooses the best individual,but also searches for the optimal direction and keeps the population diversity. Simulation results show these individuals from M-H sampling have Markov stationary distribution property. The algorithm can accelerate the speed of convergence in unimodal functions and multimodal functions,balance the performance of global searching and local searching,and has higher precision and better robustness.关键词
微分进化算法/反向微分进化算法/转移概率/平稳分布/马尔可夫链蒙特卡洛/反向学习Key words
Differential Evolution ( DE ) algorithm/Opposition Differential Evolution ( ODE ) algorithm/transition probability/stationary distribution/Markov Chain Monte Carlo( MCMC)/Opposition-based Learning( OBL)分类
信息技术与安全科学引用本文复制引用
涂维维,葛洪伟,杨金龙,袁运浩..基于M-H采样的快速反向微分进化算法[J].计算机工程,2014,(11):155-159,166,6.基金项目
国家自然科学基金资助项目(61305017) （61305017）
江苏省自然科学基金资助项目(BK20130154) （BK20130154）
江苏高校优势学科建设工程基金资助项目。 （）
