径向基函数神经网络及其在插值计算中的应用

Radial Basis Function Neural Network and Application to Data Interpolation

径向基函数神经网络(RBF)是Broomhead于1988年提出的一种新型前向神经网络,与传统的插值型神经网络BP网络相比,具有计算速度快、满足全局最优化要求的优点,所以近年来开始引起人们的重视,被引入到函数的逼近插值计算中,成为除BP网络外的另一种重要的插值神经网络。根据径向基函数神经网络(RBF)的原理,总结出了径向基函数网络的可用于复杂插值计算的一种实用插值算法。经对实例计算表明,该算法是快捷可靠的。

Neural network with radial basis function (RBF) is a new forward neural network proposed by Broomhead in 1988. It is of ac vantage of fast computation and meeting regional optimizing demand, compared with common error-back-propagation network. In recer years it becomes of more interests and is introduced into computation of approximating function interpolation. This paper presents the principle of neural network with RBF and the practicable interpolating computation. Case study shows it runs fast and reliably.