Our group at SLAC National Accelerator Laboratory is leading R&D of machine learning applications for in the area of experimental neutrino physics and a wider community of High Energy Physics. Modern neutrino experiments employ a big (100 to 10,000 tonnes), high-resolution (~mm/pixel) particle imaging detectors that records meters-long particle trajectories produced from a neutrino interaction. We address fundamental challenges in modeling these detectors, analyzing particle images, and inferring physics from big data using machine learning and advanced computing techniques. Our research has potential to accelerate physics discovery process by orders of magnitude and to maximize physics information extracted from the big, high-recision particle imaging detectors.
Areas of technical R&D include:
- Development and optimization of end-to-end deep-learning based data analysis chain
- Scaling applications at massively parallel computing infrastructure
- Quantification of uncertainty/confidence-interval for deep-learning models
- Mitigation and quantification of domain shift (e.g. discrepancies between data and simulation)
- Differentiable software for modeling physics and solving the inverse problems
Areas of physics research include:
- Neutrino oscillation analysis and cross-section measurements at the Short Baseline Neutrino program and Deep Underground Neutrino Experiment
- Physics of neutrino-nucleus interaction including modeling of many-body interactions inside a large (~40 nucleons) nucleus
- Modeling of detector physics processes using machine learning and differentiable simulation softwares
For details, feel free to contact Kazuhiro Terao.