Machine learning applications to DUNE prototype data
Kam-Biu Luk, Professor
Physics
Applications for Fall 2025 are closed for this project.
This project will focus on applying contrastive learning and anomaly detection techniques to various DUNE prototype simulated and real datasets in order to identify deficiencies in our understanding of how neutrinos interact. The work will involve: preparing inputs from the DUNE simulation and data suitable for ML-libraries; development of appropriate data augmentation to apply to events in the prototype datasets; development of an encoder model and the choice of an appropriate contrastive loss function for the problem; optimization of hyper-parameters and training to simulated data; interpretation of results.
Role: Students would have the opportunity to learn software tools commonly used in nuclear, particle physics, and cosmology. The student will be provided with ample guidance during this process and will have opportunities to shift focus to different aspects of the project - high performance computing, Bayesian statistical analysis, and/or neutrino physics - depending on their interests.
Qualifications: Junior or senior physics majors with an interest in computation. Some familiarity with programming on UNIX/LINUX and C++/Python is highly desirable.
Day-to-day supervisor for this project: Cheng-Ju Lin, Callum Wilkinson, Dan Dwyer, Staff Researcher
Hours: to be negotiated
Off-Campus Research Site: Physics Division, LBNL
Mathematical and Physical Sciences