北京邮电大学学报

  • EI核心期刊

北京邮电大学学报 ›› 1999, Vol. 22 ›› Issue (2): 58-61.

• 研究报告 • 上一篇    下一篇

对Hopfield神经网络求解TSP的研究

陈萍, 郭金锋   

  1. 北京邮电大学电信工程学院, 北京 100876
  • 收稿日期:1998-10-22 出版日期:1999-03-10

Analysis on Hopfield Neural Networks Solution to TSP

Chen Ping, Guo Jinfeng   

  1. School of Telecommunication Engineering, Beijing University of Posts and Telecommunications, Beijing 100876
  • Received:1998-10-22 Online:1999-03-10

摘要: 为了研究Hopfield神经网络解决TSP(travelling salesman problem)的算法中网络参数和城市拓扑对网络性能的影响, 利用计算机模拟Hopfield神经网络求解TSP的思路、 算法.依据大量运算结果对参数和城市拓扑进行了分析, 并得出以下结论: (1) Hopfield对网络参数的假定值并不是最佳值, 参数D对于网络的性能有比较明显的影响, D取100时,网络的收敛率大大提高; 而参数A, B, C对网络性能的影响并无明显规律, A, B可以定义在400~700的范围内, C可以在150~250之间.(2) 城市拓扑结构是影响神经网络性能最重要的因素之一.

关键词: 神经网络, 城市拓扑, 网络参数

Abstract: To study the effection of network parameters and topology of city on Hopfield NN on solving TSP, computer simulation is used and theidea and algorithm of Hopfield NN on solving TSP is carried out.According to the calculating results on parameters and topology of city, the correspondent conclusion have been given.First, the premised values to network parameters made byHopfield are not the best values.The parameter D has a major influence on the performance on network.When D equals to 100, the convergence ration of network enhanced largely.There are no obvious rules to tell the influence on performance of network made by parameters A, B, C.A and B could be defined between 400 and 700.C could be set among150 to 250.Second, the topology of a city is one of the most important factorsto the influence on network performance, which mean both the convergence ratio and the quality of the path.

Key words: neural networks, topology of city, network parameters

中图分类号: