A deep-learning based charged-current electron neutrino interaction identification in the ArgoNeuT experiment

Identification of electron neutrino interactions in liquid argon time projection chambers is essential to seeking answers to questions of the fundamental nature of neutrinos. These analyses include determining the ordering of the mass states and the value of the CP-violating phase in the neutrino sector in the Deep Underground Neutrino Experiment (DUNE) and performing neutrino oscillation measurements and beyond the Standard Model searches in the Fermilab Short-Baseline Neutrino Program. The ArgoNeuT experiment has collected GeV-scale neutrino/antineutrino data, which has been used to investigate the deep-learning based identification of charged-current neutrino interactions that forms a key part of neutrino oscillation analysis sensitivities in DUNE. We have been testing the network performance on real neutrino data. We will show the reconstruction performance using different readout planes and compare that with the boosted-decision-tree based electron neutrino classification method developed in the ArgoNeuT experiment.