Extracting cosmological parameters from N-body simulations using machine learning techniques

TL;DR

This paper demonstrates that machine learning models, including convolutional neural networks and random forest regressors, can accurately estimate key cosmological parameters from N-body simulation data and matter power spectra.

Contribution

It introduces a machine learning approach to extract multiple cosmological parameters from simulation data, showing competitive accuracy with traditional methods.

Findings

CNNs accurately estimate $\

Power spectrum-based methods are competitive with direct simulation analysis.

Scalar spectral index $n_s$ can be estimated, but with lower precision.

Abstract

We make use of snapshots taken from the Quijote suite of simulations, consisting of 2000 simulations where five cosmological parameters have been varied ($\Omega_m$, $\Omega_b$, $h$, $n_s$ and $\sigma_8$) in order to investigate the possibility of determining them using machine learning techniques. In particular, we show that convolutional neural networks can be employed to accurately extract $\Omega_m$ and $\sigma_8$ from the N-body simulations, and that these parameters can also be found from the non-linear matter power spectrum obtained from the same suite of simulations using both random forest regressors and deep neural networks. We show that the power spectrum provides competitive results in terms of accuracy compared to using the simulations and that we can also estimate the scalar spectral index $n_s$ from the power spectrum, at a lower precision.