Professor Niharika Sravan
Drexel University
Autonomous Real-time Decision-making in the Era of Multi-messenger Astronomy
Location: Physics Research Building
Host: Jared Gdanski
Abstract: The detection of the electromagnetic counterparts to both gravitational waves and neutrinos in 2017 have heralded the era of multi-messenger astronomy. These discoveries were only possible due to data in all messengers being broadcast worldwide in real-time and cyberinfrastructure/algorithms for unifying and characterizing them. However, current protocols for discovery and inference still rely on human experts manually inspecting survey alert streams and intuiting optimal usage of limited follow-up resources. With increasing sensitivity of detectors in all messengers and the Rubin Observatory's Legacy Survey of Space and Time on the horizon, it is critical to replace existing human-centered infrastructure with autonomous systems strategizing and co-ordinating follow-up to maximize designated science objectives.
In this talk, I describe my efforts to design novel intelligent systems for strategizing optimal follow-up observations in real-time. These algorithms leverage AI planning and reinforcement learning approaches to evaluate the explore-exploit tradeoff and adaptively learn to make the optimal sequence of decisions under uncertainty. I discuss their application for two important and challenging science goals: optimal cosmological parameter estimation using Type Ia supernovae and identification of kilonovae among several contaminant transients. Finally, I assert that such solutions provide the leading edge necessary to secure access to limited resources, by yielding hard numbers to expected returns, and demonstrating the necessary agility for maximizing returns for difficult and novel science cases.
Bio: Dr. Niharika Sravan was awarded a PhD in Physics and Astronomy and a graduate certificate in Integrated Data Science from Northwestern University in 2018. Dr. Sravan was a postdoctoral researcher at Purdue University and California Institute of Technology before she moved to Drexel as an Assistant Professor in Spring 2023.
Professor Sravan is an expert in the use of machine learning methods for solving problems in astronomy. She has pioneered the development of the autonomous agents for strategizing real-time observations of astronomical transients. These systems enable us to constrain the physics of supernovae and discover elusive electromagnetic counterparts to gravitational wave sources. Her efforts leverage advances in reinforcement learning algorithms, specialized hardware for deep learning including TPUs and FPGAs, and data from astronomical surveys such as the Rubin Observatory and NASA’s current and future space facilities. Sravan is a member of the Zwicky Transient Facility collaboration.
