Radial Basis Neural Networks

During the last two decades, there is a tremendous growth in neural computing field with the invention of large number of new theories, new network topologies and learning algorithms. The overall objective of these discoveries was to develop/design a neural network model which has the ability to learn, to adapt and to generalize for a given input-output mapping with further enhancements for existing models. Architecture of neural networks have been developed from its simplest form, perceptron and to the other types such as multilayer feed forward networks, recurrent networks, Hopfield networks etc. There are various weaknesses exploit in these types of neural networks such as having complex architecture, having slow training procedure etc. In order to overcome these problems, a neural network type called Radial Basis Function Neural Network (RBFNN/RBNN) was developed recently. Radial basis neural networks have advantage of being simpler than other types of neural networks while providing faster training time. Also it has the property of universal approximation of functions.