Scalable stellar evolution forecasting: Deep learning emulation vs. hierarchical nearest neighbor interpolation
arXiv:2309.12725 · doi:10.1051/0004-6361/202347118
Abstract
Many astrophysical applications require efficient yet reliable forecasts of stellar evolution tracks. One example is population synthesis, which generates forward predictions of models for comparison with observations. The majority of state-of-the-art rapid population synthesis methods are based on 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, such as MESA, 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 provide two solutions for automated interpolation methods that offer satisfactory trade-off points 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 ZAMS all the way to the end of core helium burning while covering a mass range from to . The feedforward neural network regression model (first solution) that we train to predict stellar surface variables can make millions of predictions, sufficiently accurate over the entire parameter space, within tens of seconds on a 4-core CPU. The hierarchical nearest-neighbor interpolation algorithm (second solution) that we hard-code to the same end achieves even higher predictive accuracy, the same algorithm remains applicable to all stellar variables evolved over time, but it is two orders of magnitude slower. Our methodological framework is demonstrated to work on the MIST (Choi et al. 2016) data set. Finally, we discuss the prospective applications of these methods and provide guidelines for generalizing them to higher dimensional parameter spaces.
arXiv commentAccepted at A&A
References in corpus (33)
- Modules for Experiments in Stellar Astrophysics (MESA)
- MESA Isochrones and Stellar Tracks (MIST). I: Solar-Scaled Models
- Binary interaction dominates the evolution of massive stars
- Evolution of binary stars and the effect of tides on binary populations
- MESA Isochrones and Stellar Tracks (MIST) 0: Methods for the construction of stellar isochrones
- First results from the IllustrisTNG simulations: matter and galaxy clustering
- Comprehensive analytic formulae for stellar evolution as a function of mass and metallicity
- A better measure of relative prediction accuracy for model selection and model estimation
- Progenitors of gravitational wave mergers: Binary evolution with the stellar grid-based code ComBinE
- An excess of massive stars in the local 30 Doradus starburst
- Gaia Data Release 3: Astrophysical parameters inference system (Apsis) I -- methods and content overview
- Rotating Massive Main-Sequence Stars II: Simulating a Population of LMC early B-type Stars as a Test of Rotational Mixing
- BONNSAI: a Bayesian tool for comparing stars with stellar evolution models
- A spectral synthesis code for rapid modelling of supernovae
- POSYDON: A General-Purpose Population Synthesis Code with Detailed Binary-Evolution Simulations
- Compact object mergers: exploring uncertainties from stellar and binary evolution with SEVN
- The VLT-FLAMES Tarantula Survey. XXIX. Massive star formation in the local 30 Doradus starburst
- Mining Gravitational-wave Catalogs To Understand Binary Stellar Evolution: A New Hierarchical Bayesian Framework
- Fundamental Parameters of Main-Sequence Stars in an Instant with Machine Learning
- Preparing the next gravitational million-body simulations: Evolution of single and binary stars in Nbody6++GPU, MOCCA and McLuster
- A Bayesian approach to the modelling of alpha Cen A
- Accuracy of atmospheric parameters of FGK dwarfs determined by spectrum fitting
- Fitting formulae for evolution tracks of massive stars under extreme metal poor environments for population synthesis calculations and star cluster simulations
- The fates of massive stars: exploring uncertainties in stellar evolution with METISSE
- Point forecasting and forecast evaluation with generalized Huber loss
- Stellar Parameter Determination from Photometry using Invertible Neural Networks
- Hierarchically modelling Kepler dwarfs and subgiants to improve inference of stellar properties with asteroseismology
- Dalek -- a deep-learning emulator for TARDIS
- Asteroseismology of Scuti stars: emulating model grids using a neural network
- Probabilistic Reconstruction of Type Ia Supernova SN 2002bo
- Active Learning for Computationally Efficient Distribution of Binary Evolution Simulations
- Modelling stars with Gaussian Process Regression: Augmenting Stellar Model Grid
- N-body Simulations with Live Stellar Evolution