Probabilistic Neural Networks for Rule Based Systems

Abstract

In this paper we introduce a theoretical approach for representing rule based system using Probabilistic Neural Networks (PNN). The
proposed scheme is inspired from the statistical algorithm kernel discriminant analysis where there are four layers and works with feed forward
network. This proposed approach is implemented in an algorithm called PNN-RBS. The purpose of the algorithm is to represent rule based
systems using probabilistic neural networks. This mechanism is used to have a rule based system machine learning approach and to be able to
produce results for unknown cases in the knowledge base. Also the approach should be capable of adding or removing new rules without
retraining for the neural network.

Keywords: Probabilistic Neural Networks, Neural Networks, Machine Learning