切比雪夫神经网络权值与结构确定及其模式分类应用.docVIP

  切比雪夫神经网络权值与结构确定及其模 式分类应用# 殷勇华，张雨浓** 5 10 15 20 25 30 35 40 （中山大学信息科学与技术学院，广州 510006） 摘 要 ： 本 文 提 出 一 种 可 用 于 模 式 分 类 的 新 型 的 单 输 出 切 比 雪 夫 （ Single-Output Chebyshev-Polynomial）神经网络（简称 SOCP 网络）。基于 SOCP 网络，本文进而提出另 一种可用于模式分类的多输出切比雪夫（Multi-Output Chebyshev-Polynomial）神经网络（简 称 MOCP 网络）。为避免冗长的迭代学习过程，本文采用权值直接确定法一步获得网络的 最优权值。另外，为使网络同时具备较好的拟合和泛化能力，本文提出四折交叉验证法用于 确定网络适当的隐层神经元数目。结合权值直接确定法和四折交叉验证法，本文最终提出 SOCP 网络和 MOCP 网络相对应的权值与结构确定法。数值实验结果验证了所提出的 SOCP 网络、MOCP 网络以及相对应权值与结构确定法的有效性，并且由该算法所确定的网络在 模式分类的应用中具有很高的准确率和很强的鲁棒性。 关键词：切比雪夫多项式；神经网络；权值与结构确定；模式分类；鲁棒性 中图分类号：TP183 Weights and Structure Determination of Chebyshev-Polynomial Neural Networks for Pattern Classification YIN Yonghua, ZHANG Yunong (School of Information Science and Technology, Sun Yat-sen University, GuangZhou 510006) Abstract: This paper ?rstly proposes a new type of single-output Chebyshev-polynomial feed-forward neural network (SOCPNN) for pattern classification. Then, based on such a SOCPNN, another new type of multi-output Chebyshev-polynomial feed-forward neural network (MOCPNN) is proposed for pattern classification in this paper. In order to avoid lengthy iterative-learning procedure, the weights-direct-determination (WDD) method is applied to obtaining the optimal weights of the SOCPNN and MOCPNN. In addition, the 4-fold cross-validation (4FCV) method is used to determine the appropriate numbers of the SOCPNN and MOCPNN hidden-layer neurons such that they can achieve good performances (i.e., approximation and generalization). By combining the presented WDD with 4FCV methods, two weights-and-structure-determination (WASD) algorithms, one for the SOCPNN and the other for the MOCPNN, are thus proposed for pattern classi?cation. Furthermore, experiment results substantiate the high accuracy and strong robustness of the proposed SOCPNN and MOCPNN equipped with the WASD algorithms for pattern classi?cation. Keywords: Chebyshev polynomial; neural networks; weights and structure determination (WASD); pattern clas