Machine Learning for New Physics Searches in B → K*0µ+µ− Decays

Abstract

We report the status of a neural network regression model trained to extract new physics (NP) parameters in Monte Carlo (MC) simulation data. We utilize a new EvtGen NP MC generator to generate B → K*0µ+µ− events according to the deviation of the Wilson Coefficient C9 from its SM value, δC9. We train a three-dimensional ResNet regression model, using images built from the angular observables and the invariant mass of the di-muon system, to extract values of δ C9 directly from the MC data samples. This work is intended for future analyses at the Belle II experiment but may also find applicability at other experiments.