Anomaly Detection in Astrophysics – VAE Separates Interacting Binary Stars from Normal Red Giants

Symbiotic binary stars, in which a white dwarf accretes from a red-giant companioncan be difficult to detect yet are key to understanding binary stellar evolution andsupernova progenitors. We present a novel application of a variational autoen-coder (VAE) applied to optical and infrared colors of red giants measured in theSkyMapper, 2MASS, and AllWISE surveys. By learning the latent structure ofred giants, the VAE flags anomalies, some of which align with known symbioticbinaries, successfully recovering most above the 90th percentile in anomaly scores.This analysis demonstrates the efficacy of unsupervised anomaly detection foruncovering hidden interacting binaries—and possibly other populations of astro-nomical objects with small training sets—in forthcoming large surveys such asRubin Observatory’s LSST.