RBF神经网络中心选取算法

Research on Center Selection Algorithm of RBF Neural Network

径向基函数网络构造中的关键问题是网络中心的选取,最小二乘算法采用正交化方法,独立计算回归算子对输出的贡献,故使中心的选择步骤简单有效。给出了最小二乘算法及其应用函数逼近的实例,结果证明,由于计算过程中应用了这一算法的正交化性质,所以网络调整时对已有模式的扰动达到最小。这说明最小二乘算法不仅简单有效,而且性能优越,并有较强的实用性,在许多领域有广泛应用。

The selecting of the network center is the leading problem in making radial basic function network. Because of the orthogonal of OLS algorithm and the contribution of independence calculate,it makes the process simplicity. At the end of the paper an example of the application approximation of functions is givento explain the arithmetic. The example indecates that the OLS is an excellent algorthm because of its orthogonal quality. It makes fewer disturbances to the former model, so it has wide application.