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 (4): 9-13.

• 学术论文 • 上一篇    下一篇

基于神经网络的最短路算法及其应用

马传水, 阮传概   

  1. 北京邮电大学信息工程系, 北京100876
  • 收稿日期:1999-03-24 出版日期:1999-11-10

Neural Network for Shortest Path Algorithm and Its Application

Ma Chuanshui, Ruan Chuangai   

  1. Department of Information Engineering, Beijing University of Posts and Telecommunications, Beijing 100876
  • Received:1999-03-24 Online:1999-11-10

摘要: 提出了一种基于递归神经网络(recurrent neural network)的最短路算法, 降低了神经元之间的连接复杂度, 能稳定收敛到准最优解, 并将其应用于基于流量的动态分布式路由.仿真表明, 该算法收敛稳定.

关键词: 神经网络, 最短路, 最优化

Abstract: A shortest path algorithm based on recurrent neural network (RSP) is proposed, which decreases complexity of connections among units, compared with Hopfield neural network. Simulations show that the algorithm converges to near optimal solutions steadily. While it can be applied to dynamic distributed routingalgorithm based on flow.

Key words: neural networks, shortest path, optimization algorithms

中图分类号: