Deep Learning for Liquid Scintillator-Based Double-Beta Decay Searches

Liquid scintillator-based (LS) detectors are one of the leading technologies in the search for neutrinoless double-beta decay. They are currently limited by naturally occurring and spallation-induced  backgrounds. In the future, they will be limited by the neutrino-electron scattering of Boron-8 solar neutrinos. Here we use a convolutional neural network, a common algorithm from computer vision, to distinguish between signal and background events that would have made it through existing cuts. The network was trained on the MC events for different levels of photocathode coverage and quantum efficiency to demonstrate its performance under these parameters. The ultimate goal is to apply sophisticated machine learning techniques to reject backgrounds in real detector data.