Adjuncts and alternatives to neural networks for supervised classification

While multilayer neural networks (NNs) are a powerful tool for supervised classification, their intrinsic nonlinearity often leads to slow convergence or divergence when the training sets include multimodal and/or overlapping classes. Well-known optimization techniques improve classification performance and convergence rate and reduce the tendency for divergence. Optimization techniques are also applied to the development of a noniterative perceptron-like algorithm, called the vector valued perceptron (VVP). A comparison of the VVP and the backpropagation (BP) algorithms for supervised classification indicates that the performance of VVPs is comparable to BP. VVPs are capable of solving multiclass classification problems such as the exclusive-or-problem, but require significantly less time than BP, especially for sample data with overlapping classes. VVPs applied as an adjunct and preprocessor for NNs in such cases result in improved NN classification performance and reduction in computational time