Abstract

Contributed Talk - Splinter EScience

Tuesday, 16 September 2025, 15:14

Photometric decomposition of AGN variability via Machine Learning: Spectral reconstruction and lag characterization

Iliana Isabel Cortés Pérez, Nikolaos Gianniotis, and Kai Lars Polsterer
Astroinformatics - HITS gGmbH

Disentangling the interplay between the accretion disk (AD) continuum and the reprocessed radiation from the broad line region (BLR) is critical for understanding the physical processes occurring in the inner region of active galactic nuclei (AGN). In this work, we present a machine learning framework based on Gaussian Processes (GPs) to isolate the AD and BLR contributions in photometric time series, while simultaneously reconstructing the underlying ultraviolet-optical spectra and estimating BLR time lags with quantified uncertainties. Furthermore, our method can deal with irregular sampling and accounts for observational noise. In preparation for upcoming time domain surveys such as LSST, we apply our method to SDSS archival data and examine its robustness and efficacy in extracting physical insights for AGN from photometry alone.