Optimal transport regularized dynamic radio interferometric reconstruction

Abstract

We propose a new technique for dynamic video reconstruction of radio interferometric data using the optimal transport (OT) distance as a regularizer within the regularized maximum likelihood (RML) framework. Compared to other commonly used regularization terms in RML and related prior-based approaches, the OT distance more naturally captures information including physical motion and promotes coherent dynamics. This property makes the OT regularization particularly suitable for dynamic imaging of astrophysical phenomena on minute timescales. We perform a series of tests on simulated very long baseline interferometric (VLBI) observations of Sagittarius A$^\ast$ (SgrA*), the supermassive black hole at the Galactic Center, with the Event Horizon Telescope (EHT), using both geometric and general relativistic magnetohydrodynamic models. We demonstrate that adding an OT regularization term yields significant improvements in dynamic reconstructions compared to those without it, as quantified by better recovery of orbital motion and higher dynamic cross-correlation metrics. Continued developments in OT theory across various fields, proposed extensions to the EHT network, and broad applicability to other VLBI networks make it so that OT regularization for dynamic radio interferometric reconstructions will continue to be utilized and improved in the future.