[ascl:2303.020]
HaloGraphNet: Predict halo masses from simulations
Villanueva-Domingo, Pablo;
Villaescusa-Navarro, Francisco;
Anglés-Alcázar, Daniel;
Genel, Shy;
Marinacci, Federico;
Spergel, David N.;
Hernquist, Lars;
Vogelsberger, Mark;
Dave, Romeel;
Narayanan, Desika
HaloGraphNet predicts halo masses from simulations using Graph Neural Networks. Given a dark matter halo and its galaxies, this software creates a graph with information about the 3D position, stellar mass and other properties. It then trains a Graph Neural Network to predict the mass of the host halo. Data are taken from the CAMELS hydrodynamic simulations.