Physics of Eclipsing Binaries. V. General Framework for Solving the Inverse Problem

• Published in: The Astrophysical journal. Supplement series Vol. 250; no. 2; pp. 34 - 50
• Main Authors: Conroy, Kyle E., Kochoska, Angela, Hey, Daniel, Pablo, Herbert, Hambleton, Kelly M., Jones, David, Giammarco, Joseph, Abdul-Masih, Michael, Prša, Andrej
• Format: Journal Article
• Language: English
• Published: Saskatoon The American Astronomical Society 01.10.2020; IOP Publishing
• Subjects: Algorithms; Eclipsing binary stars; Inverse problems; Mathematical models; Optimization; Parameter estimation; Parameter robustness; Samplers; Stellar systems
• ISSN: 0067-0049; 1538-4365; 1538-4365
• DOI: 10.3847/1538-4365/abb4e2

PHOEBE 2 is a Python package for modeling the observables of eclipsing star systems, but until now it has focused entirely on the forward model-that is, generating a synthetic model given fixed values of a large number of parameters describing the system and the observations. The inverse problem, obtaining orbital and stellar parameters given observational data, is more complicated and computationally expensive as it requires generating a large set of forward models to determine which set of parameters and uncertainties best represents the available observational data. The process of determining the best solution and also of obtaining reliable and robust uncertainties on those parameters often requires the use of multiple algorithms, including both optimizers and samplers. Furthermore, the forward model of PHOEBE has been designed to be as physically robust as possible, but it is computationally expensive compared to other codes. It is useful, therefore, to use whichever code is most efficient given the reasonable assumptions for a specific system, but learning the intricacies of multiple codes presents a barrier to doing this in practice. Here we present release 2.3 of PHOEBE (publicly available from http://phoebe-project.org), which introduces a general framework for defining and handling distributions on parameters and utilizing multiple different estimation, optimization, and sampling algorithms. The presented framework supports multiple forward models, including the robust model built into PHOEBE itself.