混沌粒子群优化及其分析

通过分析了经典的粒子群优化中单个粒子模型,发现其具有混沌Hopfield神经网络的特点.提出了一种新的粒子群优化模型,该模型不像以往的粒子群算法那样包含随机参数,而是一个确定性的混沌Hopfield神经网络群,其搜索轨道展现了从混沌到周期分岔再到汇的逆周期分岔演化过程.初始混沌式搜索模式展宽了搜索范围,逆周期分岔演化过程决定了搜索的稳定性和收敛性.另外,理论上给出了新的粒子群优化的收敛性结论.最后,通过数值仿真给出了与经典的粒子群优化结果的不同点,并且说明了混沌粒子群优化的有效性.

Chaotic Particle Swarm Optimization and Analysis

A single particle structure of the classical particle swarm optimization was analyzed which was found to have some properties of a Chaos-Hopfield neural network.A new model of the particle swarm optimization was proposed.The model is a deterministic Chaos-Hopfield neural network swarm which is different from the existing one with stochastic parameters.Its search orbits show an evolution process of inverse period bifurcation from chaos to periodic orbits then to sink.In this evolution process,the initial chaos-like search expands the optimal scope,and inverse period bifurcation determines the stability and convergence of the search.Moreover,the convergence is theoretically analyzed.Finally,the numerical simulation shows the difference between the chaotic particle swarm optimization and the classical particle swarm optimization;and it also demonstrates the efficiency of the presented technique.