On the exploration of symbolic regression analysis in nuclear physics.

The use of machine learning (ML) techniques in nuclear physics has been an increasing subject of interest for some years. Regression analysis is a subset of ML techniques that is designed to learn the connection between independent and dependent variables. Symbolic regression is a special class of regression analysis in that there is no explicit form given to model a system from the outset of the analysis. Given a target distribution, mathematical operations, and input quantities, symbolic regression can discover the expression that best describes the target distribution. We have chosen to explore the use of this method in the arena of nuclear density functional theory with the goal of constraining nucleonic densities. Applying the method to charge densities belonging to spherical nuclei has given encouraging results regarding the determined expression's profile as compared to the target distribution and the ability of the determined expression to predict physical attributes of the system (i.e., charge radius). We plan to continue the investigation of symbolic regression in nuclear physics through density functional theory into non-spherical and exotic nuclei.