Massively parallel transdimensional sampling with data assimilation: an application to LISA galactic binaries

Abstract

Transdimensional Bayesian inference is becoming a key ingredient in gravitational-wave astronomy. In many relevant applications, the number of components needed to describe the data is not known a priori and must be inferred jointly with their parameters. We present a transdimensional Sequential Monte Carlo (SMC) framework designed to exploit massive parallelism. The method evolves a population of particles through a sequence of tempered distributions and uses the No-U-Turn Sampler for the explo…