Abstract:The nuclear equation of state cannot be directly measured in the heavy-ion collision experiments; it is usually inferred from the comparison between transport model simulations and experimental measurements. The in-medium correction factor (F , which is defined as the ratio of the cross section in the nuclear medium to that in free space) on the nucleon-nucleon elastic cross section is one of the important inputs of the transport model, and its magnitude is still debated. The advent of Machine Learning (ML) has profoundly influenced the way scientists study the natural world and has provided a new paradigm that merges traditional investigation with advanced datadriven techniques. The aim of this work is to present a ML-based method to study the in-medium correction factor F . The ultra-relativistic quantum molecular dynamics (UrQMD) transport model is used to simulate 132Sn + 124Sn collisions at beam energy of 0.27 GeV/nucleon with different impact parameters. The value of F is randomly selected from 0.4 to 0.8 for the simulation of each event. Several observables simulated by the UrQMD model with different F , which are thought to be probably sensitive to the in-medium nucleon-nucleon cross sections, are fed into the LightGBM (Light Gradient Boosting Machine, which is a modern decision tree-based ML algorithm) to establish the mapping between the observables and F . The mean absolute error (MAE), which is the absolute difference between the true and the predicted F , is about 0.080 and 0.021 by using event-by-event and 40-event summed observables, respectively. It indicates that ML can recognize information about the F factor. Furthermore, according to the results of Shapley Additive exPlanations (SHAP), an interpretability analysis method in ML, features that have the greatest effect on the F are identified. ML combined with the transport model may open a new venue to study the F factor. In addition, the interpretability analysis of the ML algorithm may also offer valuable insights for subsequent research.

Inference of Parameters for Back-shifted Fermi Gas Model using Feedback Neural Network

Inference of parameters for the back-shifted Fermi gas model using a feedforward neural network

Prediction of Nuclear Charge Density Distribution with Feedback Neural Network

Uncertainties of Nuclear Level Density estimated by Bayesian Neural Networks

Nuclear mass predictions with multi-hidden-layer feedforward neural network

Global prediction of nuclear charge density distributions using deep neural network

Extraction of fissile isotope antineutrino spectra using feedforward neural network

Predictions of Nuclear Charge Radii and Physical Interpretations Based on the Naive Bayesian Probability Classifier

A Data-Driven Density Functional Model for Nuclear Systems

CoLFI: Cosmological Likelihood-free Inference with Neural Density Estimators

Application of a Neural Network to Store and Compute the Optical Properties of Non-Spherical Particles

Physics-informed neural networks for predicting the asymptotic outcome of fast neutrino flavor conversions

Deep Learning: Extrapolation Tool for Ab Initio Nuclear Theory.

Neural Network Emulation of Spontaneous Fission

Calibration of nuclear charge density distribution by back-propagation neural networks

Taming nucleon density distributions with deep neural network method

Machine learning the in-medium correction factor on nucleon-nucleon elastic cross section

From NS observations to nuclear matter properties: a machine learning approach

Refining mass formulas for astrophysical applications: a Bayesian neural network approach

Nuclear masses learned from a probabilistic neural network

Direct local parametrization of nuclear state densities using the back-shifted Bethe formula