In contrast to the GRNN used to estimate the values of
continuous variables, the probabilistic neural network
(PNN) finds decision boundaries between categories of
patterns. Therefore, the PNN is mainly used for classifi-
cation problems. The PNN is a parallel implementation of a
standard Bayesian classifier and has a four-layer network
that can perform pattern classification. It is based essentially
on the estimation of probability density functions for
various classes learned from training samples.