Nonlinear Discriminant Analysis Using Feed-forward Neural Networks : An Application to Diabetes Classification

抄録

This paper is concerned with the techniques for nonlinear discriminant analysis from the standpoint of comparison with hidden-layer feed-forward neural networks. We discuss learning algorithms by use of maximum likelihood method through Kullback-Leibler measure. Akaike's information criterion provides us the decision as to which of several competing network architectures is "best" for a given problem. We contrast the merits of hidden-layer feed-forward neural networks with those of ordinary discriminat analysis.