AN INFORMATION CRITERION FOR OPTIMAL NEURAL NETWORK SELECTION

Neural networks have been used to resolve a variety of classification problems. The computational properties of many of the possible network designs have been analyzed, but the decision as to which of several competing network architectures is "best" for a given problem remains subjective. A relationship between optimal network design and statistical model identification is described. A derivative of Akaike's information criterion (AIC) is given. This modification yields an information statistic which can be used to objectively select a "best" network for binary classification problems. The technique can be extended to problems with an arbitrary number of classes.