CIRCULAR BACKPROPAGATION NETWORKS FOR CLASSIFICATION

The class of mapping networks is a general family of tools to perform a wide variety of tasks; however, no unifying framework exists to desc ribe their theoretical and practical properties. This paper presents a standardized, uniform representation for this class of networks, and introduces a simple modification of the multilayer perceptron with int eresting practical properties, especially well suited to cope with pat tern classification tasks. The proposed model unifies the two main rep resentation paradigms found in the class of mapping networks for class ification, namely, the surface-based and the prototype-based schemes, while retaining the advantage of being trainable by backpropagation. T he enhancement in the representation properties and the generalization performance are assessed through results about the worst-case require ment in terms of hidden units and about the Vapnik-Chervonenkis dimens ion and Cover capacity. The theoretical properties of the network also suggest that the proposed modification to the multilayer perceptron i s in many senses optimal. A number of experimental verifications also confirm theoretical results about the model's increased performances, as compared with the multilayer perceptron and the Gaussian radial bas is Functions network.