This project will develop a flow-matching approach to detect and characterize galaxy clusters directly from telescope images. Their method will bypass intermediate galaxy catalogs and instead deliver probabilistic estimates of cluster properties such as mass, redshift and richness, with the goal of producing Bayesian galaxy cluster catalogs from Dark Energy Survey imaging and preparing software for future use with the Vera C. Rubin Observatory’s Legacy Survey of Space and Time.
2026
Flow Matching for Next-Generation Galaxy Cluster Cosmology
(top) (a) Simulated galaxy cluster center; (b) image of a galaxy cluster; (c) a galaxy cluster in the Rubin first-look images. (bottom) Confusion matrices illustrating well-calibrated posterior samples from dual-stream CNF applied to galaxy detection.
Other Researchers
Jeffrey Regier, Assistant Professor, Statistics
Camille Avestruz, Assistant Professor, Physics