To Join via Zoom: To join this seminar virtually, please request Zoom connection details from headsec [at] stat.ubc.ca
Title: Statistical Imaging of Black Holes using the Event Horizon Telescope
Abstract: In 2019 and 2021, the Event Horizon Telescope produced the first-ever images of the black holes M87* and Sgr A*, respectively. However, the Event Horizon Telescope is not a regular camera. It does not directly measure the on-sky image, and the high computational cost is converting the estimated sparse data to an actual image. As a result, imaging requires high-performance computing and statistical modeling. In this presentation, I will present the different statistical techniques used by the EHT to analyze the data. The focus will be on recent advances in applying computational Bayesian inference to the imaging problem. I will introduce the statistical model we use, which requires modeling the instrument and the image using non-linear and often weakly non-identifiable models. To sample from this posterior required using novel statistical inference techniques, such as the recently developed non-reversible optimal parallel tempering algorithm developed at UBC. These results demonstrate the potential collaboration between computational statistics and radio imaging and how it can benefit both communities. This collaboration will become more critical shortly with the advent of more powerful telescopes, such as the next-generation Event Horizon telescope, that will increase the data volume and model complexity by 2-3 orders of magnitude.