Abstract

Contributed Talk - Splinter Learning

Wednesday, 23 September 2020, 11:15   (virtual room G)

Neural Network assisted population synthesis studies

Vos J., Bobrick A. and Vuckovic M.
Universität Potsdam, Lund University, University of Valparaiso

Currently there are two main ways of studying an observed sample of binary stars. One can use 1D stellar/binary evolution models as e.g. MESA or a binary population synthesis code as e.g. Binstar. The first is accurate but very slow, while the second one is very fast, but typically not very accurate. We present a new approach, where accurate MESA models are used to train a machine learning model, that is then used as a population synthesis code. This allows us to combine the accuracy of 1D stellar evolution codes with the speed of population synthesis codes. We created the NNaPS python package to simplify these kind of studies. We also present the first results booked with this method, focused on hot subdwarf binaries.