Exploring the Variable Sky: Anomaly Detection with SNAD Pipelines

par Konstantin Malanchev (University of Illinois at Urbana-Champaign)

vendredi 15 mars 2024, 14:30 → 16:30 Europe/Paris

9109 (Salle)

SNAD, an international consortium dedicated to machine learning in astronomy, primarily focuses on anomaly detection within large light-curve catalogs. Our approach, which involves a machine-human interaction loop, has enabled us to identify and study hundreds of transients and variable stars within the Zwicky Transient Facility.

In this talk, I will present these findings and discuss the suite of tools we have developed to facilitate them. This suite ranges from data pre-processing and machine learning applications to expert analysis instruments. Our team has developed a new time-series feature extraction tool in Python and Rust, “light-curve”, prioritizing performance. This tool is currently employed by several ZTF/LSST alert brokers. Additionally, we have created a Python library, “coniferest”, which enables a user-friendly way to build custom active anomaly discovery pipelines.

We have launched the SNAD Viewer, a public web portal designed for expert analysis of ZTF data release light curves. This portal, widely used by the community beyond our team, provides access to cross-matched results from various transient surveys and static catalogs. It also integrates with several other web portals, including Fink and Astro-COLIBRI.

 

Illustration from Malanchev et al., 2023