Predicting Physical Parameters of Cepheid and RR Lyrae variables in an Instant with Machine Learning

Aug 3, 2026

Upload by Logic

A machine learning framework estimates physical parameters of classical pulsating stars, focusing on RR Lyrae and Cepheid variables, by automatically matching theoretical and observed light-curve properties across multiple wavelengths. Artificial neural networks are trained on theoretical pulsation model grids to predict fundamental parameters including mass, radius, luminosity, and effective temperature from stellar period and light-curve amplitude/phase structure. Incorporating light-curve structure improves accuracy up to 60% relative to using period alone, and the method enables rapid catalog generation for large samples from the Magellanic Clouds.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages3

Reading Time8 min

AccessFree to download

Introduction

Analysis and results

Model grid and data sources

Prediction method and validation

Conclusions

arXiv :2303 . 13692v1 [ astro-ph . SR] 23 Mar 2023 Machine Learning in Astronomy: Possibilities and Pitfalls Proceedings IAU Symposium No. , A. Mahabal, C. Fluke & J. Mclver, eds. doi:10.1017/xxxxx Predicting Physical Parameters of Cepheid and RR Lyrae variables in an Instant with Machine Learning A. Bhardwaj 1, E. P. Bellinger2 , S. M. Kanbur3 & M. Marconi 1 1INAF-Osservatorio Astronomico di Capodimonte, Salita Moiariello 16, 80131, Naples, Italy email: [anupam.bhardwaj@inaf.it](anupam.bhardwaj@inaf.it) 2Max Planck Institute for Astrophysics, Karl-Schwarzschild-Straße 1, 85748, Garching, Germany 3Department of Physics, State University of New York, Oswego, NY 13126, USA Abstract. We present a machine learning method to estimate the physical parameters of classical pulsating stars such as RR Lyrae and Cepheid variables based on an automated comparison of their theoretical and observed light curve parameters at multiple wavelengths. We train arti􀀂cial neural networks (ANNs) on theoretical pulsation models to predict the fundamental parameters (mass, radius, luminosity, and effective temperature) of Cepheid and RR Lyrae stars based on their period and light-curve parameters. The fundamental parameters of these stars can be estimated up to 60 percent more accurately when the lightcurve parameters are taken into consideration. This method was applied to the observations of hundreds of Cepheids and thousands of RR Lyrae in the Magellanic Clouds to produce catalogs of estimated masses, radii, luminosities, and other parameters of these stars. Keywords. stars: variable, Cepheid, RR Lyrae-galaxies: Magellanic Clouds-cosmology: distance scale 1. Introduction Classical Cepheids and RR Lyrae variable stars are well known distance indicators and are also excellent tracers of young and old age stellar populations, respectively (Subramanianet al. 2017; Beaton et al. 2018; Bhardwaj 2020, 2022) . Modern pulsation codes can reproduce observed pulsation periods, light and radial velocity curves, and peak-to-peak amplitude variations for these classical pulsating stars at all wavelengths(Bono, Marconi & Stellingwerf 2000; Marconi et al. 2013, 2015) . In the era of large variability surveys, a quantitative comparison of the predicted and observed pulsation properties of Cepheid and RR Lyrae variables provides stringent constraints on their intrinsic evolutionary parameters, and subsequently, provides new challenges for the stellar evolution and pulsation theory (Bhardwaj et al. 2015, 2017; Das et al. 2018) . We aim to employ modern automated methods for comparing observed and predicted light curve structure, which is de􀀂ned by its amplitude and phase parameters, of Cepheid and RR Lyrae variables and generate catalogs of physical parameters of observed variables in the Galaxy and the Magellanic Clouds. 2. Analysis and results Our theoretical model grid includes 390 models with composition (Y = 0.25 and Z = 0.008) representative of Cepheids in the Large Magellanic Cloud (LMC) computed by Marconi et al.(2013), and a total of 270 models representative of RR Lyrae metallicities ranging from Z = 0.0001 to Z = 0.02 and helium abundances ranging from Y = 0.245 to Y = 0.27 (Marconi et al. 2015) . The pulsation models provide multiband theoretical light curves for a broad range of stellar masses, luminosity levels, and chemical composition. The observational V 􀀀 and I􀀀band light curves were taken from the Optical Gravitational Lensing Experiment survey (Soszy´nski et al. 2018) . We use ANNs trained using scikit-learn on the grid of theoretical © International Astronomical Union, 2021 2 Bhardwaj A. et al. Figure 1. Predicted versus actual luminosities for estimates based on the period (left) and period and lightcurve parameters (right). The trends in errors (left) vanish when the light curve structure is included (right) . models to predict physical parameters of observed stars (see Bellinger et al. 2020, for more detai

FAQs

What is the main goal of the machine learning approach in this study?

To estimate physical parameters of Cepheid and RR Lyrae stars by comparing observed and theoretical light-curve properties automatically, across multiple wavelengths.

Which inputs does the model use to predict stellar physical parameters?

It uses stellar period and light-curve parameters that describe the light-curve structure through amplitude and phase information, trained from theoretical pulsation models.

How does adding light-curve structure affect prediction accuracy?

Including light-curve structure significantly improves performance, with fundamental parameters estimated up to about 60% more accurately than using period alone.

Aug 3, 2026

Upload by Logic