Berlin 2024 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

DY: Fachverband Dynamik und Statistische Physik

DY 5: Machine Learning in Dynamics and Statistical Physics I

DY 5.13: Vortrag

Montag, 18. März 2024, 12:45–13:00, BH-N 243

Stellar evolution forecasting with a timescale-adapted evolutionary coordinate and machine learning — •Kiril Maltsev and et al. — Heidelberger Institut für Theoretische Studien, Schloss-Wolfsbrunnenweg 35, 69118 Heidelberg

Many astrophysical applications require efficient yet reliable forecasts of stellar evolution tracks. One example is population synthesis, which generates forward predictions of models for statistical comparison with observations. The majority of state-of-the-art rapid population synthesis methods are based on approximate analytic fitting formulae to stellar evolution tracks that are computationally cheap to sample statistically over a continuous parameter range. The computational costs of running detailed stellar evolution codes over wide and densely sampled parameter grids are prohibitive, while stellar-age based interpolation in-between sparsely sampled grid points leads to intolerably large systematic prediction errors. In this work, we use supervised learning methods to construct an emulator of stellar evolution at a satisfactory trade-off between cost-efficiency and accuracy. We construct a timescale-adapted evolutionary coordinate and use it in a two-step interpolation scheme that traces the evolution of stars from zero age main sequence all the way to the end of core helium burning while covering a mass range from red dwarfs to very massive Wolf-Rayet stars. The feedforward neural network regression model that we train to predict stellar surface variables can make millions of predictions within tens of seconds on a 4-core CPU, with a mean prediction error that is an order of magnitude lower than typical observational uncertainties.