Neural Network Fast-Tracks Galaxy Formation express
Studying the birth of galaxies requires simulating billions of years of dark matter growth and star formation — even supercomputers handle this slowly. The solution came from machine learning: a neural network, like a seasoned pastry chef, from the 'recipe' of dark matter mergers instantly evaluates the future galaxy — its mass, brightness, chemical composition. Training on thousands of virtual universes led to an almost perfect match with real calculations.
An unexpected finding: the details of internal processes turned out to be secondary. A galaxy's fate is almost entirely written in its dark matter history. Accuracy reaches 97%, and speed allows testing thousands of expansion of the universe scenarios in hours, bringing us closer to understanding cosmic evolution.
Original abstract
as written by the authors arXiv:2604.23147
Understanding how galaxy populations emerge and evolve from the growth of dark matter structure is a central challenge in galaxy formation theory. Semi-analytic models (SAMs) provide an efficient framework to address this problem, but exploring large ensembles of merger trees across broad parameter spaces remains computationally demanding. We develop a conditional graph neural network surrogate model that combines merger tree information with SAM parameters to predict galaxy properties across cosmic time. Using merger trees of dark matter halos from the Uchuu simulation and the Galacticus SAM, the model predicts stellar mass, luminosity, angular momentum, gas metal mass, and specific star formation rate across the wide redshift range of 0 <= z <= 5. For instance, the model can predict stellar mass at 0 <= z <= 3 with a scatter of 0.19-0.28 dex and coefficient of determination R^2 of 0.946-0.973 (R^2 close to 1 indicates prediction closely matching the truth). The results show that a single graph based model can reproduce these galaxy properties with good accuracy over multiple SAM realizations, merger trees and redshifts. This catalog-level model provides a practical route for accelerating SAM based studies of galaxy formation to enable a more detailed investigation of the model parameter space. The inference code, trained models, and example data products are publicly available at https://github.com/MutongCat/sam2galaxy-gnn.
The most distant galaxies formed when the Universe was less than a billion years old. Their fate was sealed only by the mergers of invisible dark matter — visible stars simply followed this script.
Like Asimov's 'psychohistory,' the neural network predicts the fate of galaxies through mathematical patterns — only here, it's dark matter instead of civilizations.