A Supervised Radial Basis Function Neural Network

摘要

We introduce a radial basis function neural network, in which the variances of the radial basis function are corrected by a supervised feedback, with very little cost in training time. A significant decrease of the generalization error is observed in two test problems.