摘要
讨论了一种新的、正弦型径向基函数 (SRBF)神经网络 ,并用来逼近 n维连续函数。该 SRBF所采用的 n维正弦型的基函数是光滑的 ,并且是致密的。该 SRBF网络的权因子是输入的低阶多项式函数。本文给出的一种简单计算程序 ,显著地降低了网络训练和计算时间。并且由于 SRBF的基函数可以非均匀的量化格点为中心 ,因而降低了网络所需存储的样本数 ,网络的输出及其一阶导数都是连续的。对于非线性系统 ,该 SRBF网络在系统定义域内的逼近是精确的 ,并且在存储参数的个数上是最优的。通过实例仿真 ,证明该方法步骤简单 ,训练速度快 。
A new Sine radial basis function(SRBF) neural network which is used to approximate a continuous function ofn vari- ables is presented.The SRBF uses an n dimensional raised sine type ofthat RBF is smooth,yethas compactsupport.The SRBF network coefficients are low order polynomial functions of the input.A simple computational procedure is presented which signifi- cantly reduces the network training and evaluation time.Storage space is also reduced by allowing for a nonuniform grid of points about which the SRBFs are centered.The network outputis shown to be continuous and have a continuous firstderivative.Forthe nonlinear system,the SRBF network repersentation is exacton the domain overwhich itis defined,and itisoptimal in terms of the number of distinctstorage parameters required.Several examples are presented which illustrate the algorithm is concise,effective and accurate.
出处
《现代电子技术》
2003年第15期38-41,共4页
Modern Electronics Technique
