AI and the World's Hard Problems · Astronomy & space · arXiv · 2026-10-20
Modeling Multi-band Active Galactic Nuclei Lightcurves Using an Self-supervised Attentive Neural Process Network
Tens of millions of quasars are expected to be observed by the LSST, so their variability needs to be modeled. The authors present an attentive neural process framework that reconstructs light curves in all six LSST bands and infers black hole mass, Eddington ratio and variability timescale, trained self-supervised on noisy observations alone. On a held-out test set it matched a near-optimal Gaussian process baseline and outperformed it in power spectrum reconstruction and damped-random-walk timescale estimation.