BIMSA > Relativistic Physics Seminar Relativistic Physics Seminar Rethinking Inference in Gravitational Wave Astronomy with Machine Learning
Rethinking Inference in Gravitational Wave Astronomy with Machine Learning
Organizers
Jahed Abedi , Dey Dipanjan , Puskar Mondal , Alejandro Torres-Orjuela
Speaker
He Wang
Time
Wednesday, May 6, 2026 2:00 PM - 3:00 PM
Venue
A3-2-301
Online
Zoom 928 682 9093 (BIMSA)
Abstract
Gravitational wave astronomy enables direct observations of compact objects such as black holes and neutron stars, but its data analysis relies on statistical inference under challenging conditions, including non-Gaussian and non-stationary noise. While traditional approaches such as matched filtering and Bayesian inference provide a solid foundation, they face increasing limitations as data complexity grows. Machine learning has recently emerged as a promising direction, offering new approaches to signal detection and inference. In this talk, I will focus on how machine learning reshapes the inference paradigm in gravitational wave data analysis. I will discuss recent progress using diffusion-based models to learn complex data distributions and support inference in realistic noise conditions, with an emphasis on interpretability and reliability. I will also briefly explore the use of large language models in algorithm design and scientific discovery. These developments point toward a broader shift, where machine learning becomes not only a computational tool, but an integral part of the scientific inference framework.