Using radial basis functions to approximate a function and its error bounds

A novel network called the validity index network (VI net) is presented. The VI net, derived from radial basis function networks, fits functions and calculates confidence intervals for its predictions, indicating local regions of poor fit and extrapolation.