A comparison of artificial neural network performance: The case of neutron/gamma pulse shape discrimination

Pulse shape discrimination is investigated using artificial neural networks, namely linear vector quantization and self organizing maps which are employed for classifying neutron and gamma rays at a variety of energies and for different relative sizes of the training and test sets. While classification performance confirms that both approaches are capable of excellent discrimination, some differences between the approaches are observed: linear vector quantization is particularly accurate in classifying the training set; the self organizing map, on the other hand, demonstrates higher prediction accuracy, with its clustering capabilities rendering it less sensitive to classification errors. Comparisons with existing analytical as well as artificial neural network approaches are made.