Karim Carrion (National Autonomous University of Mexico)

Stage IV galaxy surveys, such as Euclid and the Vera C. Rubin Observatory’s Legacy Survey of Space and Time (LSST), will generate datasets of unprecedented size and precision. These surveys hold the potential to place strong constraints on models of the governing physics on cosmological scales. However, analyzing these vast datasets requires the exploration of high-dimensional and complex parameter spaces to accurately model systematic effects, particularly on non-linear scales, which often demand computationally intensive theoretical prediction. 
In this work, we use a novel framework for cosmological likelihood inference designed to accelerate Bayesian inference from cosmological surveys to derive constraints on beyond-standard models of gravity. We employ recent advances in machine learning, starting with CosmoPower to efficiently emulate the non-linear matter power spectrum based on the halo model reaction.
We focus on constraining an Interacting Dark Energy model, called Dark Scattering. Our first analysis combines KiDS-1000 data with information from CMB+BAO, constraining the dark energy—dark matter interaction parameter, A_{ds} =10.6^{+4.5}_{−7.3} b/GeV at 68% confidence, while also alleviating the S8 tension. A second analysis will also be presented, showing forecasts for Stage IV cosmic shear surveys, obtained through an automatically differentiable inference pipeline to further accelerate the Bayesian analysis.