Neural network extraction of ρ(770) pole position in the ππ invariant mass distribution

Abstract

The typical way to extract resonance parameters is by using the standard fitting. In this study, we demonstrate that the pole position of a resonance can be found through neural network (NN) inference.  We used a NN trained on synthetic data to predict the mass and width of the ρ(770) resonance based on available phase shift data for ππ scattering. The synthetic data was generated using K-matrix parameters and the corresponding pole positions were found using the Newton-Raphson method. We compared the performance of two NNs trained on two datasets with differing pole position distributions. The NN trained on the more restrictive dataset performed better in predicting the ρ(770) resonance parameters.