Dark Matter profiles of "in silico" galaxies: deep learning inference

Abstract

Machine learning has the potential to improve the reconstruction of the dark matter profile of galaxies with respect to traditional methods, like rotation curves. We demonstrate on the simulation suite Illustris-TNG that a steerable equivariant convolutional neural network (CNN) is able to infer the dark matter profiles within and around individual galaxies from photometric and interferometric data, improving on a standard CNN. Within the in silico environment of the simulations, our architecture is able to capture the dark matter distribution within galaxies without a parametrization of the profile. We perform an interpretability analysis to understand the internal mechanisms of the trained model and the most important data features used to estimate the dark matter profiles. The equivariant CNN recovers the dark matter profile of galaxies within the stellar mass range $[10^{10}-10^{12}]$ $M_{\odot }$ with excellent precision and accuracy: the mean squared error is reduced by a factor of ~ 3 from its value under the training distribution, demonstrating that the network has learnt from the data features. While this holds within the controlled 'in silico' environment of the simulation, we argue that few additional steps are needed before this method can be reliably applied to galaxies in the real field observations.

Published: May 27, 2026

References in corpus (41)

• Planck 2015 results. XIII. Cosmological parameters
• Stellar population synthesis at the resolution of 2003
• The EAGLE project: Simulating the evolution and assembly of galaxies and their environments
• A direct empirical proof of the existence of dark matter
• A Comprehensive Measurement of the Local Value of the Hubble Constant with 1 km/s/Mpc Uncertainty from the Hubble Space Telescope and the SH0ES Team
• A 2.4% Determination of the Local Value of the Hubble Constant
• Simulating Galaxy Formation with the IllustrisTNG Model
• First results from the IllustrisTNG simulations: matter and galaxy clustering
• First results from the IllustrisTNG simulations: the stellar mass content of groups and clusters of galaxies
• Sloan Digital Sky Survey IV: Mapping the Milky Way, Nearby Galaxies, and the Distant Universe
• First results from the IllustrisTNG simulations: the galaxy color bimodality
• THINGS: The HI Nearby Galaxy Survey
• First results from the IllustrisTNG simulations: A tale of two elements -- chemical evolution of magnesium and europium
• A History of Dark Matter
• Simba: Cosmological Simulations with Black Hole Growth and Feedback
• DESI 2024 VI: Cosmological Constraints from the Measurements of Baryon Acoustic Oscillations
• First results from the IllustrisTNG simulations: radio haloes and magnetic fields
• Direct constraints on the dark matter self-interaction cross-section from the merging galaxy cluster 1E0657-56
• New Interstellar Dust Models Consistent with Extinction, Emission, and Abundance Constraints
• The non-gravitational interactions of dark matter in colliding galaxy clusters
• The optical morphologies of galaxies in the IllustrisTNG simulation: a comparison to Pan-STARRS observations
• Cosmological simulations of black hole growth: AGN luminosities and downsizing
• The Origin of Disks and Spheroids in Simulated Galaxies
• Evidence for dark matter in the inner Milky Way
• Non-circular motions and the diversity of dwarf galaxy rotation curves
• Transfer learning for galaxy morphology from one survey to another
• The Hubble Sequence at $z\sim 0$ in the IllustrisTNG simulation with deep learning
• From lenticulars to blue compact dwarfs: the stellar mass fraction is regulated by disc gravitational instability
• A robust estimate of the Milky Way mass from rotation curve data
• SKIRT: hybrid parallelization of radiative transfer simulations
• The many reasons that the rotation curves of low-mass galaxies can fail as tracers of their matter distributions
• KURVS: The outer rotation curve shapes and dark matter fractions of $z\sim 1.5$ star-forming galaxies
• An estimate of the DM profile in the Galactic bulge region
• ROGER: Reconstructing Orbits of Galaxies in Extreme Regions using machine learning techniques
• MARTINI: Mock Array Radio Telescope Interferometry of the Neutral ISM
• Total and dark mass from observations of galaxy centers with Machine Learning
• Galaxy mass profiles with convolutional neural networks
• Cosmic Kite: Auto-encoding the Cosmic Microwave Background
• Bayesian technique to combine independently-trained Machine-Learning models applied to direct dark matter detection
• Machine-Learning Analysis of Radiative Decays to Dark Matter at the LHC
• On the effects of parameters on galaxy properties in CAMELS and the predictability of ${\Omega }_{\mathrm{m}}$