Bayesian atomic structure calculations for collisional problems

Abstract

Synopsis The calculations of collisional processes require an accurate description of the target. In general, the atomic structure is obtained through tedious iterations in which a variety of configurations and parameters are chosen to minimize the differences between the numerical and experimental values of the energies and the oscillator strengths. Using a Bayesian machine learning analysis through a Tree–structured Parzen Estimator, we can reproduce the experimental atomic structure with high accuracy. Results for neutral beryllium are presented.