Speaker
Description
Determining the neutrino mass ordering remains one of the central open questions in particle physics. Although upcoming long baseline, atmospheric, and reactor neutrino experiments are expected to provide a definitive answer in the next decade, current data offer limited sensitivity, and the small spectral differences between the two hypotheses challenge traditional chi squared and likelihood based analyses, especially in the presence of parameter degeneracies.
In this talk, we will present a study exploring machine learning techniques as an alternative strategy for tackling the mass ordering problem. Using synthetic datasets constructed under realistic experimental assumptions, including matter effects and statistical fluctuations, we train a feedforward neural network to classify the normal and inverted neutrino mass ordering and to estimate the oscillation parameter Δm_31^2. The machine learning approach is benchmarked against standard likelihood based fits and other multivariate classifiers to enable a detailed comparison of methodologies.
We will describe the network architecture, training procedure, and validation workflow, as well as the criteria used to assess performance relative to established oscillation analysis techniques. The talk will also highlight how this framework interfaces with current analysis tools and how it can be extended as a pedagogical introduction to machine learning within neutrino physics.
| I read the instructions above | Yes |
|---|