Machine learning improves neutrino event simulations using real data
Machine-learning-based neutrino-event generation constrained by experimental measurements
Machine Learning
Summary
Simulating how neutrinos interact with matter is important for physics experiments but can be inaccurate if only based on theory. The authors developed a machine learning model that creates simulated neutrino events and then adjusts itself using actual experimental measurements, even without detailed data on individual events. This approach improves the simulation’s accuracy for various measurements and works across different types of neutrino beams. Their method helps connect experimental results directly to better event simulations without needing to retrain the model completely.
What this means in practice
- •For particle physics simulation teams: Improve neutrino interaction simulations by calibrating generative models directly with measured cross sections without needing experimental event-level data.
- •For medical imaging developers: Use calibrated generative models trained on measured data to improve simulations of particle interactions in tissue for enhanced imaging techniques.
Authors
Jose Luis Bonilla, Krzysztof M. Graczyk, Rwik Dharmapal Banerjee, Beata E. Kowal, Hemant Prasad, Jan T. Sobczyk
Abstract
We present a generative surrogate for charged-current \(ν_μ\)-nucleus scattering that can be calibrated to measured differential cross sections without event-level experimental samples. Starting from a theoretical interaction model, we fine-tune the surrogate to a MicroBooNE measurement and show that this also improves the description of observables not used in the fit. The calibrated model can also be applied to measurements from different neutrino fluxes without retuning, while retaining event-level predictions of outgoing-muon kinematics. Our results show that experimental cross sections can directly constrain generative event models while preserving their applicability across different observables and beam configurations.