Machine learning n/γ discrimination in CLYC scintillators

Abstract Two machine learning techniques, one supervised (Artificial Neural Network) and the other unsupervised ( k -means + + ) have been applied to the task of n/ γ discrimination in 7 Li-enriched CLYC detectors, and compared to traditional charge-comparison methods. The results show that a very basic artificial neural network can provide very good discrimination in the energy range investigated, and the k -means + + algorithm is capable of separating neutrons and gamma-rays in CLYC scintillators as well as suggesting reasonable window parameters for charge comparison methods.