Finding signatures of the nuclear symmetry energy in heavy-ion collisions with deep learning

A deep convolutional neural network (CNN) is developed to study symmetry energy $E_{\rm sym}(\rho)$ effects by learning the mapping between and two-dimensional (transverse momentum rapidity) distributions of protons neutrons in heavy-ion collisions. Supervised training performed with labelled data-set from ultrarelativistic quantum molecular dynamics (UrQMD) model simulation. It found that, using proton spectra on event-by-event basis as input, accuracy for classifying soft stiff about 60% due large fluctuations, while setting event-summed classification increases 98%. The 5-label (5 different sym}(\rho)$) task are 58% 72% neutron spectra, respectively. For regression task, mean absolute error (MAE) which measures average magnitude differences predicted actual $L$ (the slope parameter 20.4 14.8 MeV Fingerprints density-dependent nuclear transverse rapidity can be identified algorithm.