计算机应用 ›› 2012, Vol. 32 ›› Issue (12): 3339-3342.DOI: 10.3724/SP.J.1087.2012.03339

• 人工智能 • 上一篇    下一篇

基于矩阵运算的单隐层Madaline网络批量学习

张银川,白书奎   

  1. 河海大学 计算机与信息学院,南京 211100
  • 收稿日期:2012-06-26 修回日期:2012-08-05 发布日期:2012-12-29 出版日期:2012-12-01
  • 通讯作者: 张银川
  • 作者简介:张银川(1988-),男,江苏新沂人,硕士研究生,主要研究方向:人工神经网络、模式识别、数据挖掘;〓白书奎(1989-),男,河南南阳人,硕士研究生,主要研究方向:模式分类、数据挖掘。

Matrix computing-based batch learning for Madaline network with one hidden layer

ZHANG Yin-chuan,BAI Shu-kui   

  1. College of Computer and Information, Hohai University, Nanjing Jiangsu 211100,China
  • Received:2012-06-26 Revised:2012-08-05 Online:2012-12-29 Published:2012-12-01
  • Contact: ZHANG Yin-chuan

摘要: 针对前向离散型单隐层Madaline网络建立了以矩阵为基础的数学模型,结合高维空间超平面划分理论,通过对表示样本的矩阵与代表网络性质的矩阵进行分析运算,在输入样本维度较低的情况下给出了Madaline网络的批量学习方法。该方法可有效地解决离散数据的两类分类问题。

关键词: Madaline网络, 矩阵运算, 批量学习, 两类分类

Abstract: In this paper, a matrix computing-based mathematic model was established for the feedforward discrete Madaline network with one hidden layer. By analyzing the matrix representing samples and the matrix representing attributes of the network, and combining the hyperplane division theory in high dimensional space, a batch learning method was proposed for Madaline network with the input of lower dimensional samples. This method can effectively solve the problem of two-category classification of discrete data.

Key words: Madaline network, matrix computing, batch learning, two-category classification

中图分类号: