计算机工程 ›› 2012, Vol. 38 ›› Issue (12): 172-175.doi: 10.3969/j.issn.1000-3428.2012.12.051

• 人工智能及识别技术 • 上一篇    下一篇

基于留一准则的多尺度径向基函数网络

张 猛 1,付丽华 2,刘智慧 2,何婷婷 1,魏志成 3   

  1. (1. 华中师范大学计算机科学系,武汉 430079;2. 中国地质大学数学与物理学院,武汉 430074; 3. 河北师范大学物理科学与信息工程学院,石家庄 050016)
  • 收稿日期:2011-07-06 出版日期:2012-06-20 发布日期:2012-06-20
  • 作者简介:张 猛(1977-),男,副教授、博士,主研方向:机器学习,自然语言处理;付丽华(通讯作者),副教授、博士;刘智慧,讲师、硕士;何婷婷,教授、博士;魏志成,副教授、博士
  • 基金项目:
    国家自然科学基金资助项目(11026145, 61071188, 61102103);中央高校基本科研业务费专项基金资助项目(CUG090112, CUG110407, CCNU10A01013);湖北省自然科学基金资助项目(2010CDB04205, 2009CDB077);河北省教育厅自然科学青年基金资助项目 (2010258)

Multi-scale RBF Network Based on Leave One Out Criterion

ZHANG Meng 1, FU Li-hua 2, LIU Zhi-hui 2, HE Ting-ting 1, WEI Zhi-cheng 3   

  1. (1. Department of Computer Science, Central China Normal University, Wuhan 430079, China; 2. School of Mathematics and Physics, Chinese University of Geosciences, Wuhan 430074, China; 3. School of Physics Science and Information Engineering, Hebei Normal University, Shijiazhuang 050016, China)
  • Received:2011-07-06 Online:2012-06-20 Published:2012-06-20

摘要: 针对传统径向基函数(RBF)网络难以确定迭代停止条件的缺点,提出采用最小化留一误差来训练多尺度RBF网络的算法。分别使用全局k均值聚类算法和经验选择方法,构造RBF节点的中心和尺度参数备选项集合,利用正交前向选择方法逐步最小化留一误差,从而确定网络的每一项中心和尺度参数。实验结果显示,该算法能够自动终止新网络节点选择,不需要额外的迭代终止条件,与传统的RBF网络相比,能够产生稀疏性更高且泛化能力更好的径向基网络。

关键词: 径向基函数网络, 多尺度, 留一准则, 正交前向选择, 全局k均值聚类

Abstract: In order to circumvent the difficulty of pre-assigning a threshold to terminate the iterations in training traditional Radial Basis Function(RBF) network, a novel RBF network training algorithm is proposed. A global k-means clustering algorithm and empirical method are utilized to construct the candidate sets for centre and scales of regressors. At each regressor stage, the parameters of each term are selected by minimizing Leave One Out(LOO) criterion using orthogonal forward selection. Simulation results show that the new algorithm can be terminated fully automatically. Compared with the other RBF networks, this scheme is capable of producing sparser RBF network with much better generality.

Key words: Radial Basis Function(RBF) network, multi-scale, Leave One Out(LOO) criterion, orthogonal forward selection, global k-means clustering

中图分类号: