Searching for anomalous variability in the LSST era and beyond

by Dr Olexandra Razim (SMASH Fellow at the University of Nova Gorica)

Thursday 15 Jan 2026, 11:30 → 12:30 Europe/Rome

Sala Jappelli (Osservatorio astronomico di Padova)

Astronomy reached petabyte scales, and traditional discovery methods, such as manual inspection or classical statistics, aren't applicable to such data volumes. Machine learning is now central to object classification and characterisation, yet anomaly detection remains a persistent challenge. In the absence of ground truth, separating astrophysically meaningful outliers from instrumental artefacts is particularly difficult. While image-based approaches have advanced rapidly, anomaly detection in light curves remains comparatively underdeveloped.

High-cadence space missions have shown that roughly 60% of stars are variable at millimagnitude precision. Upcoming surveys such as the Legacy Survey of Space and Time (LSST) and the Roman Space Telescope will observe billions of sources with comparable photometric accuracy, dramatically increasing the volume and complexity of variability data. Without robust automated methods, many rare or unexpected phenomena are likely to remain unnoticed.

In this talk, we discuss the possibility of creating a semi-automatic, model-agnostic pipeline for anomaly detection in variable sources, including rare stellar types and novel subclasses of transient phenomena. We'll also talk about identifying the regions of parameter space most promising for discovery with next-generation surveys, and consider how these findings may advance our understanding of stellar and Galactic evolution.

 

Join Zoom Meeting

 

https://unipd.zoom.us/j/82482883288?pwd=MVvmMWSGaiVrGMyAFOdyQG61OAcI7D.1

Meeting ID: 824 8288 3288

Passcode: 072472