An application of neural networks in chemistry. Prediction of¹³C NMR chemical shifts

Abstract

Basic definitions of neural networks are given in terms of oriented graphs. Partial derivatives of an objective function with respect to the weight and threshold coefficients are derived. These derivatives are very important for the adaptation process, carried out by a version of the gradient method of We neural network considered. The stability of the adapted neural network toward small changes — “perturbation” — of input activities is described by sensitivities. The theory is illustrated by application of simple neural networks that reflect the topology of molecules to the classification of¹³C NMR chemical shifts of secondary carbons in acyclic alkanes.