Identification of AGN in JWST with Machine Learning

Understanding the link between supermassive black hole growth and host galaxy evolution requires assessing not just high-luminosity active galactic nuclei (AGN), but also previously unseen low-luminosity AGN. The James Webb Space Telescope (JWST) pushes further down the AGN luminosity function than ever before, allowing the identification of elusive low-luminosity AGN. Existing AGN color selection methods make use of at most 4 photometric filters, but JWST/MIRI and JWST/NIRCam collectively have 19 broadband mid-infrared photometric filters, demonstrating a need for improved AGN selection methods that make use of all available information. Such improved methods are particularly needed for the identification of AGN in strongly star forming galaxies, known as composite galaxies, which can appear quite similar to typical star forming galaxies, and to identify heavily obscured AGN. I am developing a new tool for the identification of AGN in JWST/MIRI, utilizing XGBoostLSS, an extension of the popular Extreme Gradient Boosting (XGBoost) algorithm that models all moments of a parametric distribution (i.e., mean, location, scale and shape) instead of the conditional mean only.