作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2006, Vol. 32 ›› Issue (20): 210-212. doi: 10.3969/j.issn.1000-3428.2006.20.078

• 人工智能及识别技术 • 上一篇    下一篇

神经网络集成在图书剔旧分类中的应用

徐 敏   

  1. (南通大学计算机科学与技术学院,南通 226007)
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2006-10-20 发布日期:2006-10-20

Application of Neural Network Ensembles in Book Weeding

XU Min   

  1. (College of Computer Science & Technology, Nantong University, Nantong 226007)
  • Received:1900-01-01 Revised:1900-01-01 Online:2006-10-20 Published:2006-10-20

摘要: 在分析图书剔旧工作的基础上,指出用智能的方法解决图书剔旧问题的必要性。提出了可以用神经网络集成技术来解决该问题,并给出一种动态构建神经网络集成的方法,该方法在训练神经网络集成成员网络时不仅调整网络的连接权值,而且动态地构建神经网络集成中各成员神经网络的结构,从而在提高单个网络精度的同时,增加了各网络成员之间的差异度,减小了集成的泛化误差。实验证明该方法可以有效地用于图书剔旧分类。

关键词: 图书剔旧, 神经网络集成, 负相关学习

Abstract: Through analyzing the work of book weeding, it is known that to use the intelligent method in book weeding is very important. The neural network ensembles technology is a good idea for dealing with these problems. This paper presents a constructive method for training neural network ensembles. In training individual networks in ensembles, the proposed method not only adjusts the connection weights, but also makes the network architecture of the individual networks. So it improves the accuracy of the individual networks while increasing the diversity among the individual networks and decreasing the generalization error of network ensembles. Experiments prove that this ensemble is very effect in weeding books.

Key words: Book weeding, Neural network ensembles, Negative correlation learning