This document discusses radial basis functions and splines for neural network modeling. It presents the architecture of a radial basis function network, which uses radial basis functions as activations in a single hidden layer. The network approximates functions as a linear combination of radial basis functions centered at different locations. The document then shows how to classify the output of an XOR gate using such a network with 4 radial basis functions in the hidden layer and computes the weights between the hidden and output layers to perform the classification.