Roger Huang (LBNL), Machine Learning Based Reweighting Methods in Neutrino Physics

Thursday 27 Aug 2026, 15:00 → 17:00 US/Eastern

Next-generation long-baseline neutrino oscillation experiments like DUNE and Hyper-Kamiokande will achieve orders of magnitude improvements in statistics compared to current long-baseline experiments. For them to make their desired measurements of parameters like δ_CP, this increase in statistics will have to be accompanied by much stronger control of many systematic uncertainties around neutrino-nucleus interactions. Machine learning-based reweighting methods offer better ways to probe high-dimensional effects in our simulation and data, providing a path to better understanding these uncertainties. This talk will present two such applications in neutrino physics. The first is with OmniFold, a ML-based unfolding method that allows extraction of unbinned, high-dimensional results, applied to the T2K experiment. The second is with ML-based reweighting of the predictions from neutrino event generators, facilitating easier study of the effects of different theoretical models when analyzing neutrino data in general.