Inferring galaxy dark halo properties from visible matter with machine learning

Abstract

Next-generation surveys will provide photometric and spectroscopic data of millions to billions of galaxies with unprecedented precision. This offers a unique chance to improve our understanding of the galaxy evolution and the unresolved nature of dark matter (DM). At galaxy scales, the density distribution of DM is strongly affected by feedback processes, which are difficult to fully account for in classical techniques to derive galaxy masses. We explore the capability of supervised Machine Learning (ML) algorithms to predict the DM content of galaxies from ‘luminous’ observational-like parameters, using the TNG100 simulation. In particular, we use, Photometric (magnitudes in different bands), Structural (the stellar half-mass radius and three different baryonic masses), and Kinematic (1D velocity dispersion and the maximum rotation velocity) parameters to predict the total DM mass, DM half-mass radius, DM mass inside one and two stellar half-mass radii. We adopt the coefficient of determination, R2, as a metric to evaluate the accuracy of these predictions. We find that using all observational quantities together (Photometry, Structural, and Kinematics) we reach high accuracy for all DM quantities (up to R2 ∼ 0.98). This first test shows that ML tools are promising to predict the DM in real galaxies. The next steps will be to implement the observational realism of the training sets, by closely selecting samples which accurately reproduce the typical observed ‘luminous’ scaling relations. The so-trained pipelines will be suitable for real galaxy data collected from Rubin/LSST, Euclid, CSST, 4MOST, DESI, to derive, e.g. the properties of their central DM fractions.

Robust Field-level Likelihood-free Inference with Galaxies

“…In particular, likelihood-free inference methods work by taking data directly from the simulations (without the need for summary statistics and, thus, model comparison), and many papers have shown competitive results compared with the usual statistical inference methods (Ravanbakhsh et al 2017;Hassan et al 2020;Mangena et al 2020;Ntampaka et al 2020;Villaescusa-Navarro et al 2021a;Shao et al 2022b;Cole et al 2022;Makinen et al 2022;. At a level closer to the observations and simulations, many papers exploring the halogalaxy connection are able to make predictions that are comparable to the output of numerical/analytical methods (Kamdar et al 2016;Jo & Kim 2019;Yip et al 2019;Zhang et al 2019;Kasmanoff et al 2020;Wadekar et al 2020;Moster et al 2021;Villanueva-Domingo et al 2021;Shao et al 2022a;Jespersen et al 2022;Delgado et al 2022;Lovell et al 2022;McGibbon & Khochfar 2022;von Marttens et al 2022;Rodrigues et al 2023). Furthermore, a clear advantage of ML models is that, once they are trained, they typically make predictions much faster than traditional methods (Jespersen et al 2022); a disadvantage arises when these models fail to extrapolate their predictions across different data sets, from the ones with which they were trained (Villaescusa-Navarro et al 2021a.…”