Application of the Probabilistic RBF Neural Network in Multidimensional Classification Problems

Abstract

Neural network is an universal classifier and with the proper choosing of its architecture it can solve any, even very complicated, classification task. The main problem in neural network applications lies in the fact, that their learning process is complicated and time-consuming. It concerns especially multidimensional tasks for which neural network architecture is very extended. The probabilistic RBF neural network does not possess all of the mentioned disadvantages. It has only one coefficient to tune so its learning is very easy and much faster than a feedforward multilayer network.

The paper describes some experiments with the probabilistic RBF neural network used in multidimensional classification problems.