西南石油大学学报(自然科学版) ›› 1994, Vol. 16 ›› Issue (3): 117-122.DOI: 10.3863/j.issn.1000-2634.1994.03.020

• 基础研究 • 上一篇    下一篇

神经网络BP算法用于三类基本波形分类时训练样本输入方式的研究

胡泽 吴宁 陈伟

  

  1. 1.石油开发系 2.重庆大学电机系 3.石油开发系
  • 收稿日期:1992-10-31 修回日期:1900-01-01 出版日期:1994-08-20 发布日期:1994-08-20
  • 通讯作者: 胡泽

Researches on Input Modes of Training Samples of Backp ropagation Algorithm of Neural Networks Applied to Classifying Three Basic Wave Modes

Hu Ze Wu Lin Chen Wei
  

  1. 1.Southwestern Petroleum Institute 2.Department of Electic Engineering, Chongqing University 3.Southwestern Petroleum Institute
  • Received:1992-10-31 Revised:1900-01-01 Online:1994-08-20 Published:1994-08-20
  • Contact: Hu Ze

摘要: 为了进一步研究神径网络BP葬法用于摸式分类时网络的性能,本文采用了四种方式向多层
神经网络提供训练样本,并相互进行比较,结果衣明:随机输入法和部分输入法的输入方式为较佳,从而说明了训练样本的翰入方式对神经网络的性能有着直接的影响.

关键词: 神经网络, BP葬法, 模式分类, 样本, 训练样本

Abstract: In order to further research the properties of neural networks when their BP algorithm is used in mode classification,this paper proposes four ways of offering training samples to multi一layer neural networks, and makes a comparison between them. The results show that the better ways are random input mode and partial input mode,therefore the input mod-
es of traing samples have direct influence on the properties of neural network

Key words: Neural networks, Backpropagation algorithm, Mode classification, SampleTraining sample

中图分类号: