Reconstructing neutrino interactions in the liquid argon DUNE Near Detector using machine learning

The next-generation DUNE experiment will depend on a combination of liquid argon time projection chambers (LArTPCs) and other detectors to make measurements of neutrino oscillation parameters with unprecedented precision. A LArTPC close to the neutrino beam source, the Near Detector LArTPC (ND-LAr), will be used to constrain uncertainties arising from beam flux and neutrino interaction modeling. This detector will use modular TPCs and pixel readout technology to help cope with challenges associated with the extremely high rate of neutrino interactions expected in ND-LAr. In conjunction, a new approach to event reconstruction that leverages recent advances in machine learning (ML) and the unprecedented detail afforded by LArTPC technology is also being pursued. A fully ML-based end-to-end reconstruction chain called DeepLearnPhysics has recently been demonstrated to cope successfully with analogous challenges on previous LArTPC experiments. We present the status of efforts to adapt DeepLearnPhysics for use in ND-LAr, including initial work towards reconstruction of ν_μ interactions in simulation.