Abstract
Kasra Mokhtarpour
Florida Atlantic University
kmokhtarpour2019@fau.edu
Machine-Learning Identification of Changing-State AGN Candidates in Gaia DR3 We use Gaia Data Release 3 (DR3) light curves to identify changing-state active galactic nuclei (CSAGNs). My research is primarily focused on objects already categorized as variable AGN in the vari_agn epoch photometry data, which contain 872,228 AGN light curves. For this purpose, we employ unsupervised deep learning and anomaly detection methods to identify unusual patterns of change in the light curves. We find that 8,504 light curves exhibit anomalous behavior, of which 682 sources show drastic flux changes, especially abrupt increases or decreases between two plateau-like flux levels, that are consistent with a changing-state transition. The novelty of this study is that narrowing the initial set of 8,504 anomalies to the final set of 682 CSAGN candidates, specifically through identification of the two distinct flux states, is also carried out using a machine-learning method called Gaussian mixture modeling (GMM), which minimizes the need for human visual inspection in selecting the CSAGN candidates. A spectroscopic follow-up is underway to confirm the nature of these candidates in Gaia DR3. In addition to building a large sample of CSAGN candidates while reducing human bias, this work also discusses the possible drivers of the changingstate transition, which can provide important insight into the underlying physics of the accretion disk and supermassive black hole (SMBH) fueling.