Prediction of Atomic Ground State Relaxation Rate from Fluorescence Spectra Using Machine Learning

We consider the implementation of machine learning methods to retrieve the values of the physical parameters of experimentally studied atomic systems from the registered spectra. The specific task was to predict the relaxation rate of the ground state sublevels of Rb atomic vapor from the measured fluorescence spectra, which is a typical regression problem. Linear and nonlinear machine learning methods have been studied as promising methods for processing and predicting physical behavior. An optimal regression model is presented, which is characterized by high accuracy and short modeling time for the key indicators of the function.