基于退火策略的暂态混沌神经网络算法及其在FDP中的应用

Neural Network with Transient Chaos for Solving FDP Problem

影片递送问题(简称FDP)是组合优化的一个新问题,它比旅行商问题(TSP)复杂的多,它可以推广到一大类路径和排序问题。文章给出了一种解FDP问题的混沌神经网络算法,该算法首先将FDP问题转换成TSP问题,然后利用神经元的自抑制反馈产生混沌动态,构造具有暂态混沌特性的神经网络算法(TCNN)。由于混沌的遍历性和随机搜索性有效地克服了Hopfield神经网络(HNN)极易陷入局部极小的缺陷;同时利用一时变参数控制混沌行为,使网络在经过一个短暂的倍周期倒分岔后逐渐趋于一般的神经网络,从而收敛到一个最优或近似最优的稳定平衡点。仿真表明,该算法具有更强的全局搜索能力和更高的搜索效率。

The Film Deliverer problem(FDP),a new problem in the combination optimization is much more complicated than the Traveling Salesman Problem(TSP).In this paper,a new neural network is presented to solve the FDP.First convert the FDP in-to TSP,and then given a neural network mode with transient chaos(TCNN)by introducing chaos which is generated by nega-tive self-feedback into HNN,TCNN would not be stuck into lo-cal Minimum.With a time -variant parameter to control the chaos,TCNN goes through an inverse bifurcation process and gradually approaches to HNN with converges to a stable equilib-rium point.Numerical simulation shows that TCNN has higher ability of searching for globally optimal to FDP prob lem thanHNN and higher efficiency of searching.