Training artificial neural networks using Taguchi methods

Abstract

This paper shows how the process optimization methods known as Taguchi methods may be applied to the training of Artificial Neural Networks. A comparison is made between the efficiency of training using Taguchi methods and the efficiency of conventional training methods; attention is drawn to the advantages of Taguchi methods. Further, it is shown that Taguchi methods offer potential benefits in evaluating network behaviour such as the ability to examine interaction of weights and neurons within a network.