Beyond DC and MCMC: alternative algorithms and approaches to fitting light curves

Abstract

The parameter space of binary star light curve models is highly complex and degenerate, thus basic fitting approaches often fail to yield a good (and correct) estimate of the parameter values and their uncertainties. On the other hand, we have an increasingly large number of fitting and sampling algorithms available that can be relatively easily interfaced with open-source eclipsing binary packages, like PHOEBE 2. We showcase several fitting methods, including local and global minimizers, nested sampling and machine learning methods, and evaluate their performance on fitting a light curve model with PHOEBE 2.