计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (18): 41-43.

• 理论研究 • 上一篇    下一篇

构造性层状神经网络集成方法

徐 敏   

  1. 南通大学 计算机科学与技术学院,江苏 南通 226019
  • 收稿日期:2007-09-18 修回日期:2007-12-19 出版日期:2008-06-21 发布日期:2008-06-21
  • 通讯作者: 徐 敏

Constructive algorithm for training nappe neural network ensembles

XU Min   

  1. College of Computer Science,Nantong University,Nantong,Jiangsu 226019,China
  • Received:2007-09-18 Revised:2007-12-19 Online:2008-06-21 Published:2008-06-21
  • Contact: XU Min

摘要: 在分析构造性神经网络集成和层状神经网络集成方法的基础上,提出了一种构造性层状神经网络集成方法。该方法自动确定神经网络集成中成员神经网络的数目,以及成员神经网络的结构等。集成在保证成员神经网络精度的同时,又保证了成员网络之间的差异度。用户只需要简单定义一些参数,就可以构造出性能较好的神经网络集成。

Abstract: Through the analysis the work of constructive algorithm and nappe neural networks,the constructive algorithm for training nappe neural network ensembles is proposed.The proposed method constructs the ensembles automatically.So it improves the accuracy of the individual networks while increasing the diversity among the individual networks.And this method helps the users with little experience in using neural networks to design the ensembles architecture.Compared with standard ensembles in solving several problems this methods with excellent performance.