一种基于粒子优势分析的异步混合粒子群算法

A Hybrid Particle Swarm Optimization Algorithm Based on Asynchronous Advantage Analysis of Particle

针对标准粒子群算法收敛速度较慢、求解精度不高等缺陷,引入了均值漂移与球隙迁移算法的思想,提出一种混合算法.该算法结合最优粒子与自己的祖先粒子来对粒子进行优势分析,根据其优势分析结果确定粒子的更新速度级别,将速度进行分解,分配到粒子的不同维中以达到异步更新的目的;为每个粒子设置一个淘汰概率的属性,当粒子被淘汰时会被自动替换;算法还引入了扰动机制和随机重启策略.显然,改进后的算法增加了粒子搜索的多样性和明智性,从而加快了收敛速度.最后,将混合算法用于求解高维TSP问题,实验结果表明改进后的算法是可行的、有效的.

In order to overcome the shortcoming of the standard particle swarm optimization algorithm such as bad convergence and ac- curacy, an hybrid PSO based on mean shift and sphere-gap transferring is proposed. The algorithm combines particle optimal particle and their ancestors to analysis of particle, according to its advantage analysis results to determine particle update speed level, the speed of decomposition, assigned to different dimensional particle in order to achieve the purpose of asynchronous update; attribute set one out probabilities for each particle, is automatically replaced when the particles are eliminated; algorithm also introduces disturbance mechanism and random restart strategy. Obviously, the hybrid algorithm can increase the diversity of particles and wise search, thus speeding up the convergence speed. Finally, the hybrid algorithm is used to solve the high dimensional TSP problem;the experimental results show that the hybrid algorithm is feasible and effective.