Abstract

Poster - Splinter LargeScale   (MW-2235)

Simulation-based inference of Cosmological parameters from the Higher-order statistics of galaxy distribution

Phu Huy Nguyen, Jiamin Hou
University of Rome "Tor Vergata", USM LMU

The galaxy power spectrum captures information in the cosmic density field if the field is Gaussian; late–time gravitational evolution or deviations from standard single-field slow-roll inflation lead to leakage into higher-order statistics. In this work, we build a simulation-based inference (SBI) pipeline. Using a large suite of N-body mock catalogs, we infer cosmological parameters from the summary statistics via neural posterior estimation. This SBI inference avoids the analytic likelihood assumptions of the traditional analyses. We examine how the choice of summary statistics, their dynamical range, and their compression shape what the network can learn at a fixed simulation budget. In the current work, we focus on the power spectrum and bispectrum. In the next step, we will extend the work to four-point functions and incorporate modified initial conditions.