Abstract
This paper proposes a Neural Network model using Generalised kernel functions for the hidden layer of a feed forward network. These functions are Generalised Radial Basis Functions (GRBF), and the architecture, weights and node topology are learned through an evolutionary algorithm. The proposed model is compared with the corresponding standard hidden-node models: Product Unit (PU) neural networks, Multilayer Perceptrons (MLP) with Sigmoidal Units (SUs) and the RBF neural networks. The proposed methodology is tested using twelve benchmark classification datasets from well-known machine learning problems. GRBFs are found to perform better than other standard basis functions at the classification task.