Dj. Livingstone et al., DATA MODELING WITH NEURAL NETWORKS - ADVANTAGES AND LIMITATIONS, Journal of computer-aided molecular design, 11(2), 1997, pp. 135-142
The origins and operation of artificial neural networks are briefly de
scribed and their early application to data modelling in drug design i
s reviewed. Four problems in the use of neural networks in data modell
ing are discussed, namely overfitting, chance effects, overtraining an
d interpretation, and examples are given of the means by which the fir
st three of these may be avoided. The use of neural networks as a vari
able selection tool is shown and the advantage of networks as a nonlin
ear data modelling device is discussed. The display of multivariate da
ta in two dimensions employing a neural network is illustrated using e
xperimental and theoretical data for a set of charge transfer complexe
s.