Speaker
Description
We investigate a hybrid approach of parametric and nonparametric
regression techniques to analyse the phase boundary between the
confined and deconfined phases of two-flavour quark matter. Data
derived from the Nambu-Jona-Lasinio (NJL) and Polyakov-loop extended
NJL (PNJL) models are trained and further used for the prediction of phase
transition boundaries with enhanced accuracy. We also observe the
classification of the order of phase transition and interpret the
model's predictions with extreme gradient boosting and SHapley Additive
exPlanations(SHAP) values. Our findings demonstrate that this learning
method effectively captures the complex behaviour of quark matter
transitions and is also useful to balance interpretability and
flexibility.
