
Some Progress on Galaxy-Halo Connection
Yen-Ting Lin of the Institute for Advanced Study presents recent research on how galaxies populate dark matter halos. He covers four projects: detecting assembly bias signals using the constrained simulation Elucid, a graph neural network that predicts galaxy stellar mass and color from redshift 0 to 5 using only dark matter subhalo merger-tree properties, a new method for tracking how the stellar mass-halo mass relation evolves with redshift, and a technique for estimating host halo mass by counting a galaxy's neighboring galaxies. The talk is a working astrophysics seminar aimed at researchers already familiar with halo occupation and large-scale structure concepts, walking through methodology and results from each project rather than offering a general introduction. It reflects current work combining simulations and machine learning to connect observable galaxy properties to the underlying dark matter structure they inhabit.