Campus, Varian 355
In this talk, I will give an overview of how to do field-level likelihood-free inference with galaxies catalogs. More specifically, only using galaxy phase-space information, not imposing a scale cutoff, I will show how to convert these into graphs. Also, I will briefly explain how to use graph neural networks to infer the Omega matter of these simulations. I will show that the model is robust across 5 different hydrodynamic simulations (Astrid, IllustrisTNG, SIMBA, Magneticum, and SWIFT-EAGLE), i.e., 5 different sub-grid physics models, being one of the first able to do this task. Moreover, this model can take into account different halo/subhalo finders and do fairly good predictions while considering many more variations in the parameter space of cosmology and astrophysics. I will chat a bit about the ability of this model to take into account errors in the peculiar velocities and selection effects.