Robust marginalization of baryonic effects for cosmological inference at the field level

TL;DR

This paper develops neural network-based likelihood-free inference methods that extract cosmological parameters from 2D mass maps, effectively marginalizing over baryonic physics and utilizing all resolved scales.

Contribution

The authors introduce a neural network approach for field-level inference that robustly marginalizes baryonic effects, outperforming traditional summary statistic methods.

Findings

Networks infer mbda_m and  with 4% and 2.5% accuracy.

Method extracts information beyond power spectra and one-point functions.

Approach generalizes to different simulations, demonstrating robustness.

Abstract

We train neural networks to perform likelihood-free inference from $(25\,h^{-1}{\rm Mpc})^2$ 2D maps containing the total mass surface density from thousands of hydrodynamic simulations of the CAMELS project. We show that the networks can extract information beyond one-point functions and power spectra from all resolved scales ($\gtrsim 100\,h^{-1}{\rm kpc}$) while performing a robust marginalization over baryonic physics at the field level: the model can infer the value of $\Omega_{\rm m} (\pm 4\%)$ and $\sigma_8 (\pm 2.5\%)$ from simulations completely different to the ones used to train it.