Neutrino interaction event filtering at liquid argon time projection chambers using neural networks with minimal input model bias

In current and future neutrino oscillation experiments using liquid argon time projection chambers (LAr-TPCs), a key challenge is identifying neutrino interactions from the pervading cosmic-ray background. Rejection of such background is often possible using traditional cut-based selections, but this typically requires the prior use of computationally expensive reconstruction algorithms. This work demonstrates an alternative approach of using 3D Convolutional Neural Networks (CNNs) trained on low-level timing information from only the scintillation light signal of interactions inside LAr-TPCs. We further present a means of mitigating biases from imperfect simulations by applying Domain Adversarial Neural Networks (DANNs). These techniques are applied to example simulations from the ICARUS detector, the far detector of the Short Baseline Neutrino experiment at Fermilab. The results show that cosmic background is reduced by up to 74% whilst neutrino interaction selection efficiency remains over 94%, even in cases where the simulation poorly describes the data.