Radial basis function neural network (RBFN) has the power of the unive
rsal function approximation But how to construct an RBFN to solve a gi
ven problem is usually not straightforward. This paper describes a met
hod to construct an RBFN classifier efficiently and effectively. The m
ethod determines the middle layer neurons by a fast clustering algorit
hm and computes the optimal weights between the middle and the output
layers statistically. We applied the proposed method to construct an R
BFN classifier for an unconstrained handwritten digit recognition. The
experiment showed that the method could construct an RBFN classifier
quickly and the performance of the classifier was better than the best
result previously reported. (C) 1997 Elsevier Science Ltd.