Speaker
Description
Bayesian frameworks that combine theory, nuclear experiments, and astrophysical observations, such as GW170817 and NICER, have successfully placed strong constraints on the neutron star equation of state (EOS). However, the computational demands, which require days to weeks, have limited progress. Recent breakthroughs in machine learning, differential programming, and GPU acceleration transform this bottleneck. Normalizing flow-enhanced samplers analyze binary neutron star mergers in 20 minutes, while gradient-based methods enable complete Bayesian equation of state (EOS) inference in under one hour without the need for pre-trained emulators. This computational revolution enables previously impractical investigations, including high-dimensional parameter scaling, direct EOS breakdown density determination, and the revelation of nuclear parameter degeneracies. Our framework positions the field to fully exploit current and future data while opening a new window for studying supranuclear matter.
| Presentation Style | Oral Presentation |
|---|