Ab Initio Real-Time Gravitational-Wave Parameter Estimation

概要

We present a specialised GPU-native nested sampling kernel targeting rapid parameter estimation for gravitational wave inference problems. Building upon a Slice-within-Gibbs (SwiG) structure for rapid mixing, we investigate how far we can push baseline stochastic sampling techniques on modern GPU hardware. We demonstrate that for typical long-duration binary neutron star signals observed by the LIGO and Virgo detectors, we can achieve well calibrated posterior inference on the full uncompressed…