摘要
通过分析径向基函数网络与支持向量机之间的关系,将结构风险最小化原则应用于径向基函数网络学习中,与传统的基于经验风险最小化原则的径向基函数网络相比,它充分保证了模型的泛化能力,能够弥补学习方法本身的缺陷。最后,将该算法应用于非线性时间序列预测,并与传统的径向基函数网络预测结果进行了比较,实验结果表明本算法提高了径向基函数网络的泛化能力。
Applied the structural risk minimization （SRM） principle to the study of RBF networks by analyzing of the relationships between the radial basis function （RBF） networks and support vector machines （ SVM ）. Compared with traditional RBF networks based on empirical risk minimization （ERM） principle, it fully assure model generalization and can remedy the shortcomings of single learning method. Finally, the proposed new algorithm was applied to non-linear time series forecast and was compared with the predicted outcome of traditional RBF networks. Experiments show that the proposed new algorithm imoroves the generalization ability of RBF networks.
出处
《化工自动化及仪表》
CAS
北大核心
2009年第3期34-37,共4页
Control and Instruments in Chemical Industry
基金
陕西省教育厅科学研究项目资助(07JK192)
