Abstract: An orgbozing radial basis function network is proposed, and a hybrid learning algorithm, which holds both non-supervised orgedzing function and supervised learning function, is introduced tO it. Some illustrative examples are given, showing the network might be applied to function approximating and pattern classification.