Page History

...

1. Augmenting GNNs with morphological information.  One promising avenue is to include the morphological information of galaxies while trying to learn the galaxy-halo-environment connection, since galaxy mergers and evolutionary processes leave imprints on galaxy appearances. There are several ways to quantify galaxy morphology, and in this project, we will compare a baseline GNN against GNN models augmented with morphological information. All data here will come from publicly accessible cosmological simulations.
2. Applying GNNs to observations of galaxy clusters. We can also apply trained GNN models to observations of galaxy clusters in order to estimate the dark matter mass profiles. Many lensing clusters have been imaged by HST, JWST, and ground-based telescopes, providing an orthogonal constraints on the cluster masses. In this case, GNNs trained on simulation data would be adapted to galaxy catalogs taken from real observations.
3. Using symbolic regression to interpret scaling laws. By using symbolic regression (e.g., PySR) to derive analytic formulas in place of neural networks, we can learn an interpretable form of GNN outputs, e.g., the galaxy stellar dark matter halo mass as a function of dark matter halo galaxy properties. 

Student Work

Planned work:

...