English

应用科学学报 ›› 2002, Vol. 20 ›› Issue (3): 287-290.

• 论文 • 上一篇    下一篇

模糊反向传播神经网络的函数逼近能力研究

王士同1,2, 朱晓铭1   

  1. 1. 华东船舶工业学院计算机系, 江苏镇江 212003;
    2. 江南大学信息学院, 江苏无锡 214036
  • 收稿日期:2001-03-19 修回日期:2001-08-14 出版日期:2002-09-30 发布日期:2002-09-30
  • 作者简介:王士同(1964-),男,江苏邗江人,教授,博导.
  • 基金资助:
    国家自然科学基金资助项目(69840003);江苏省青年科技基金课题(BQ2000003)

Research on the Function Approximation Capabilities of Fuzzy Backpropagation Neural Networks

WANG Shi-tong1,2, ZHU Xiao-ming1   

  1. 1. Department of Computer, East China Shipbuilding Institute, Zhenjiang 212003, China;
    2. School of Information, Southern Yangtsi University, Wuxi 214036, China
  • Received:2001-03-19 Revised:2001-08-14 Online:2002-09-30 Published:2002-09-30

PDF

113

被引次数

2

摘要： 研究了模糊反向传播神经网络的函数逼近能力.研究结果给出了单调连续函数的FBP按序单调特性、连续映射定理以及非函数一致逼近定理,并且说明了FBP虽然能保持连续性映射,但不如原神经网络具有函数逼近能力.

关键词: 连续函数, 模糊神经网络, 函数逼近能力, 反向传播

Abstract: This paper studies the function approximation capabilities of FBP. The results here show the ordered monotony property, the continuous mapping theorem and the non-function approximation theorem. It is proved that although FBP still retain the continuous mapping property, their capabilities to approximate function are worse when compared with traditional neural networks.

Key words: function approximation capabilities, continuous function, fuzzy neural networks, backpropagation

中图分类号: