Exploring the halo-galaxy connection with probabilistic approaches

Paper highlights:

• Compares different probabilistic machine learning methods to model uncertainty in the halo–galaxy connection using the IllustrisTNG300 simulation.
• Evaluates multivariate Gaussian distribution, a multilayer perceptron classifier, normalizing flows (NFs), finding comparable performance with flows performing best in most regimes.
• Shows that different halo masses and galaxy populations exhibit varying levels of stochasticity, with important implications for large-scale structure analyses.

You can read the paper here.