Use of machine learning techniques to investigate trends in AGN/Quasar activity and evolution
In this study, we investigate how a machine learning algorithm, Uniform Manifold Approximation and Projection (UMAP), classifies and gives insight about active galactic nuclei (AGN) and quasar activity of over 200,000 low-redshift galaxies with spectra from the Sloan Digital Sky Survey. Using traditional broad-line and narrow-line diagnostics on the sample, we find that AGN (broad line and narrow line) and quasars occupy distinct branches on the UMAP space. We further investigate trends with dust reddening and AGN luminosity to understand whether the algorithm reveals paths within the broader galaxy evolution context. Lastly, we create a new UMAP projection for the AGN and quasars alone to search for trends more closely related to black hole growth. This work demonstrates the power of unsupervised ML tools to reveal physically meaningful patterns in large astronomical datasets, offering new ways to study black hole growth and AGN diversity.
