Probabilistic neural networks: a brief overview of theory, implementation, and application

Abstract Probabilistic neural networks (PNNs) offer a scalable alternative to the conventional back-propagation neural networks in classification problems without the need for massive forward and backward calculations that is associated with the ordinary neural networks. In addition, they can work with smaller sets of training data. However, this advantage may come at a cost of requiring large amounts of memory as the training data get larger. This chapter takes a look at the fundamental mathematics behind the modern PNNs, their application, and approaches that address some practical issues that come with them.