论文部分内容阅读
基于支持向量回归和RBF(Radial Basis Function)神经网络,研究了带有未知但有界噪声的非线性系统的集员辨识问题.推导了噪声界以及支持向量个数与ε-不敏感参数之间的关系,给出了利用噪声界选择ε-不敏感参数的方法.描述了通过支持向量回归选择RBF神经网络规模的方法.该方法以Gaussian核函数作为径向基函数,支持向量作为径向基函数的中心构建RBF神经网络.运用改进的OBE(Optimal Bounding Ellipsoid)算法对RBF神经网络的权值进行辨识,得到与给定输入输出数据和噪声界序列一致的一类RBF神经网络.仿真算例验证了算法的有效性.
Based on the support vector regression (RBF) and Radial Basis Function (RBF) neural networks, the collective identification problem of nonlinear systems with unknown and bounded noises is studied. The noise bounds, the number of support vectors and ε-insensitive parameters are derived , A method of selecting ε-insensitive parameters by using the noise boundary is given.The method of selecting RBF neural networks by support vector regression is described.The method takes Gaussian kernel as radial basis function and support vector as radial basis Function to construct RBF neural network.An improved algorithm of Optimal Bounding Ellipsoid (OBE) is used to identify the weights of RBF neural networks, and a RBF neural network is obtained which is consistent with given input and output data and noise boundary sequences. The effectiveness of the algorithm is verified.