Searching for changing-state AGNs in massive datasets — I: applying deep learning and anomaly detection techniques to find AGNs with anomalous variability behaviours. (arXiv:2106.07660v2 [astro-ph.IM] UPDATED)
<a href="http://arxiv.org/find/astro-ph/1/au:+Sanchez_Saez_P/0/1/0/all/0/1">P. Sánchez-Sáez</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Lira_H/0/1/0/all/0/1">H. Lira</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Marti_L/0/1/0/all/0/1">L. Martí</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Sanchez_Pi_N/0/1/0/all/0/1">N. Sánchez-Pi</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Arredondo_J/0/1/0/all/0/1">J. Arredondo</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Bauer_F/0/1/0/all/0/1">F. E. Bauer</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Bayo_A/0/1/0/all/0/1">A. Bayo</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Cabrera_Vives_G/0/1/0/all/0/1">G. Cabrera-Vives</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Donoso_Oliva_C/0/1/0/all/0/1">C. Donoso-Oliva</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Estevez_P/0/1/0/all/0/1">P. A. Estévez</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Eyheramendy_S/0/1/0/all/0/1">S. Eyheramendy</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Forster_F/0/1/0/all/0/1">F. Förster</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Hernandez_Garcia_L/0/1/0/all/0/1">L. Hernández-García</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Arancibia_A/0/1/0/all/0/1">A. M. Muñoz Arancibia</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Perez_Carrasco_M/0/1/0/all/0/1">M. Pérez-Carrasco</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Sepulveda_M/0/1/0/all/0/1">M. Sepúlveda</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Vergara_J/0/1/0/all/0/1">J. R. Vergara</a>
The classic classification scheme for Active Galactic Nuclei (AGNs) was
recently challenged by the discovery of the so-called changing-state
(changing-look) AGNs (CSAGNs). The physical mechanism behind this phenomenon is
still a matter of open debate and the samples are too small and of
serendipitous nature to provide robust answers. In order to tackle this
problem, we need to design methods that are able to detect AGN right in the act
of changing-state. Here we present an anomaly detection (AD) technique designed
to identify AGN light curves with anomalous behaviors in massive datasets. The
main aim of this technique is to identify CSAGN at different stages of the
transition, but it can also be used for more general purposes, such as cleaning
massive datasets for AGN variability analyses. We used light curves from the
Zwicky Transient Facility data release 5 (ZTF DR5), containing a sample of
230,451 AGNs of different classes. The ZTF DR5 light curves were modeled with a
Variational Recurrent Autoencoder (VRAE) architecture, that allowed us to
obtain a set of attributes from the VRAE latent space that describes the
general behaviour of our sample. These attributes were then used as features
for an Isolation Forest (IF) algorithm, that is an anomaly detector for a “one
class” kind of problem. We used the VRAE reconstruction errors and the IF
anomaly score to select a sample of 8,809 anomalies. These anomalies are
dominated by bogus candidates, but we were able to identify 75 promising CSAGN
candidates.
The classic classification scheme for Active Galactic Nuclei (AGNs) was
recently challenged by the discovery of the so-called changing-state
(changing-look) AGNs (CSAGNs). The physical mechanism behind this phenomenon is
still a matter of open debate and the samples are too small and of
serendipitous nature to provide robust answers. In order to tackle this
problem, we need to design methods that are able to detect AGN right in the act
of changing-state. Here we present an anomaly detection (AD) technique designed
to identify AGN light curves with anomalous behaviors in massive datasets. The
main aim of this technique is to identify CSAGN at different stages of the
transition, but it can also be used for more general purposes, such as cleaning
massive datasets for AGN variability analyses. We used light curves from the
Zwicky Transient Facility data release 5 (ZTF DR5), containing a sample of
230,451 AGNs of different classes. The ZTF DR5 light curves were modeled with a
Variational Recurrent Autoencoder (VRAE) architecture, that allowed us to
obtain a set of attributes from the VRAE latent space that describes the
general behaviour of our sample. These attributes were then used as features
for an Isolation Forest (IF) algorithm, that is an anomaly detector for a “one
class” kind of problem. We used the VRAE reconstruction errors and the IF
anomaly score to select a sample of 8,809 anomalies. These anomalies are
dominated by bogus candidates, but we were able to identify 75 promising CSAGN
candidates.
http://arxiv.org/icons/sfx.gif
