Equivalence of a Radial Basis Function NN and a Perceptron

摘要

This article deals with a transformation of conditional average estimator, which is utilized in the description of a radial basis function NN, to the multi-layer perceptron equation. The exact equivalence of both paradigms is demonstrated for a one-dimensional case with symmetric triangular activation functions. The transformation provides a simple interpretation of perceptron parameters in terms of statistical samples of input data.