English

应用科学学报

• 论文 • 上一篇    下一篇

利用连续Hopfield神经网络求解聚类问题的模型构造

王茂芝 郭科 徐文皙 范安东   

  1. 成都理工大学信息管理学院信息与计算科学系，四川，成都，610059
  • 收稿日期:2004-09-15 修回日期:2004-12-03 出版日期:2006-01-31 发布日期:2006-01-31

Modeling Optimal Clustering Based on Continuous Hopfield Neural Network

WANG Mao-zhi, GUO Ke, XU Wen-xi, FAN An-dong   

  1. School of Management of Information, Chengdu University of Technology, Chengdu 610059, China
  • Received:2004-09-15 Revised:2004-12-03 Online:2006-01-31 Published:2006-01-31

被引次数

1

摘要： 在给出有中心聚类问题的数学描述后，把聚类问题转化为组合优化计算问题。利用连续Hopfield神经网络的优化计算能力，设计并构造了基于连续Hopfield神经网络求解聚类问题的模型。详细描述了模型的网络映射过程、推导了模型的状态转换方程以及能量函数，并通过引入竞争机制简化了状态转换方程和能量函数的迭代形式，最终给出了所构造模型的收敛性证明。同时通过在图像压缩编码中的应用验证了该模型的有效性和合理性。

关键词: 聚类, 组合优化, 神经网络, 能量函数, 收敛

Abstract: Clustering is a combinational optimal calculation based on its mathematical description. A model combined with continuous Hopfield neural network and used to solve optimal clustering is designed and constructed based on Hopfield’s optimal ability. Details of network mapping, energy function construction and nerve state changing equation are described. Their simplified formation is presented according to the winner-tale-all competitive mechanism. Convergence of the model is proved. Effectiveness and rationality of the model is verified in an application to image compression coding.

Key words:

cluster, combination optimal, neural network, energy function, convergence