Sparse Convolution Transformers for DUNE FD Event and Particle Classification

The Deep Underground Neutrino Experiment (DUNE) is a long-baseline neutrino oscillation experiment that will measure the fundamental neutrino oscillation parameters using Fermilab's enhanced PIP-II beam. The far detector modules will consist of several liquid argon time projection chambers (LArTPCs) at SURF 1300 km away and will generate high resolution images of neutrino interactions. Recent advances in AI image generation such as stable diffusion XL can optimize ML reconstruction approaches that use large and sparse LArTPC images. We introduce a transformer network, which is often used for large language models, for simultaneous particle and neutrino interaction event classification at the DUNE horizontal-drift far detector. This presentation will highlight the classification accuracy improvements of this network and the potential applications for the DUNE oscillation analysis.