Scientific Discovery with Foundation Models

The next generation of telescopes such as the SKA and the Vera C. Rubin Observatory will produce enormous data sets, far too large for traditional analysis techniques. Machine learning has proven invaluable in handling massive data volumes and automating many tasks traditionally done by human scientists. In this talk, I will explore the use of machine learning for automating the discovery and follow-up of interesting astronomical phenomena, both in the image and time domains. I will discuss how the human-machine interface plays a critical role in maximising scientific discovery with automated tools, demonstrating applications of the active anomaly detection framework, Astronomaly, on a variety of datasets. Finally, I will investigate the role foundation models play in enabling scientific discovery in massive surveys.

Bio: 

Born in South Africa with a PhD from the University of Cape Town, Prof. Michelle Lochner is an Associate Professor at the University of the Western Cape. Her research focus is on using artificial intelligence for scientific discovery in large astronomical datasets from next-generation optical and radio telescopes. These include the Vera C. Rubin Observatory in Chile, and the Square Kilometre Array and its precursor, MeerKAT, located in South Africa. She is the lead developer of Astronomaly—an open-source active learning platform for anomaly detection in astronomical data—which has been applied to millions of sources from multiple large-scale sky surveys. Committed to equity in physics, she is also the founder and director of the Supernova Foundation, an international mentoring programme for women and gender minorities in physics.