基于广义径向基函数神经网络的非线性时间序列预测器

NONLINEAR TIME SERIES PREDICTOR BASED ON GENERALIZED RADIAL BASIS FUNCTION NEURAL NETWORKS

该文对传统的径向基函数(RBF)神经网络的结构和学习算法进行了总结,并在此基础上提出了广义径向基函数模型概念,使这种网络具有更好的应用灵活性与可扩充性。文章基于Mackey-Glass造血模型方程的数值解数据,对此广义模型与现有的RBF模型和梯度径向基函数(GRBF)模型对非线性时间序列预测问题的应用结果进行了比较与讨论,显示出这种广义模型的应用有效性。

The architecture and learning algorithm of traditional radial basis function (RBF) neural networks are surveyed in this paper. A generalized radial basis function model is proposed, which is more flexible and extensible. Based on the numerical solution to Mackey-Glass hematopoietic model equation, the prediction results obtained by radial basis function (RBF) model, gradient radial basis function (GRBF) model, and the generalized radial basis function model are compared and discussed, which show the effectiveness of the generalized model.