Robust Field-level Likelihood-free Inference with Galaxies

TL;DR

This paper develops graph neural networks for likelihood-free inference of cosmological parameters from galaxy catalogs, demonstrating robustness across diverse simulations and astrophysical models, and highlighting the importance of using both galaxy positions and velocities.

Contribution

The paper introduces a robust, scale-invariant graph neural network approach for field-level likelihood-free inference using galaxy data, effective across multiple simulation codes and astrophysical parameters.

Findings

Models achieve ~12% precision in inferring mbda_m

Robustness demonstrated across different simulation codes and physics models

Effective extrapolation over a wide parameter space

Abstract

We train graph neural networks to perform field-level likelihood-free inference using galaxy catalogs from state-of-the-art hydrodynamic simulations of the CAMELS project. Our models are rotational, translational, and permutation invariant and do not impose any cut on scale. From galaxy catalogs that only contain $3$D positions and radial velocities of $\sim 1, 000$ galaxies in tiny $(25~h^{-1}{\rm Mpc})^3$ volumes our models can infer the value of $\Omega_{\rm m}$ with approximately $12$ % precision. More importantly, by testing the models on galaxy catalogs from thousands of hydrodynamic simulations, each having a different efficiency of supernova and AGN feedback, run with five different codes and subgrid models - IllustrisTNG, SIMBA, Astrid, Magneticum, SWIFT-EAGLE -, we find that our models are robust to changes in astrophysics, subgrid physics, and subhalo/galaxy finder. Furthermore, we test our models on $1,024$ simulations that cover a vast region in parameter space - variations in $5$ cosmological and $23$ astrophysical parameters - finding that the model extrapolates really well. Our results indicate that the key to building a robust model is the use of both galaxy positions and velocities, suggesting that the network have likely learned an underlying physical relation that does not depend on galaxy formation and is valid on scales larger than $\sim10~h^{-1}{\rm kpc}$.