Scientific goals
In order to maximize the scientific return of cosmological missions such as CMB-S4, DESI, eROSITA, Euclid, Roman Observatory, Rubin Observatory, Simons Observatory, PFS, and SKA, accurate theoretical predictions in the non-linear regime, for generic statistics, are needed. Furthermore, these predictions need to incorporate the uncertainity associated to our poor knowledge of baryonic effects such as AGN feedback.
CAMELS is a large suite of N-body and state-of-the-art (magneto-)hydrodynamic simulations designed to tackle this problem. At its core, CAMELS represents a large dataset to train machine learning algorithms. The main scientific goals of CAMELS are these:
Provide theory predictions for summary statistics and full 3D fields as a function of cosmology and astrophysics.
Train neural networks to extract cosmological information while marginalizing over baryonic effects.
Develop machine learning techniques to find the mapping between N-body simulations and hydrodynamic simulations with full baryonic physics.
Quantify the dependence of galaxy formation and evolution on astrophysical and cosmological parameters.
Use machine learning to efficiently calibrate subgrid parameters in cosmological hydrodynamic simulations to match a set of observations.