Basic Operating Characteristics of Neural Networks When Applied to Structure-Activity Studies

Abstract

This paper describes the basic operating characteristics in perceptron-type neural networks, especially when applied to structure-activity relationships (SAR). It was shown that the neural network has outstanding abilities in both classification and fitting. The operation of the neural network is basically carried out in a nonlinear manner. This nonlinearity brings forth merits as well as a small number of demerits. Problems with the operation of the neural network are pointed out and discussions are aimed at solutions.