Machine learning algorithms have become essential statistical tools in experimental particle physics, providing the basis for the reduction and classification of big data sets, and the precision measurement of physical processes. In recent years, a collective effort has begun to harness these techniques to meet the challenges of the next generation of long-baseline neutrino oscillation experiments. One major neutrino project currently in its technical design phase is the Deep Underground Neutrino Experiment (DUNE), which will perform precision measurements of neutrino oscillations using a Megawatt neutrino beam produced at the Fermi Laboratory, USA, and a 40 kiloton Liquid Argon Time Projection Chamber (LAr-TPC), constructed 1300 km away at the Homestake mine. The superb imaging capabilities of
LAr-TPC detectors will enable DUNE to capture complex neutrino event topologies with high levels of spatial and calorimetric precision. However, new techniques of pattern recognition will be required to fully exploit the detailed information available in each LAr-TPC image.
In this PhD project, the student will apply techniques of machine learning to the analysis of LAr-TPC images, with the goal of precisely reconstructing the final-state particle tracks and electromagnetic showers in each image and classifying the overall neutrino event topology.
The techniques will be demonstrated on cosmic-ray and test-beam data from the DUNE prototype detectors, and will then be applied to the reconstruction of simulated neutrino data from DUNE to support both technical design and oscillation physics studies.