A new GPU-accelerated global fitting algorithm for LISA data analysis, Michael Katz, AEI Potsdam

The global fit of LISA data is one of the most challenging analysis tasks in the field of gravitational waves. In this talk, I will review the status of the LISA global fit problem and discuss progress on our new global fit implementation designed to examine the LDC2A (Sangria) data set from the LISA Data Challenges (LDC) Working Group. The sampling methods we use rely on a combination of Markov Chain Monte Carlo (MCMC) techniques, including reversible-jump MCMC (RJMCMC) to sample the uncertain number of sources in the data. Additionally, we employ the technique of “multiple-try’’ MCMC, a novel method for gravitational wave analyses where we harness the parallel power of GPUs by suggesting multiple proposal destinations for a single move of an MCMC walker.

Our algorithm is built from a combination of publicly available and soon-to-be publicly available codes including Galactic binary waveforms, a generic (not gravitational-wave specific) global fit sampler based on emcee, and a LISA-based toolkit that brings together waveform and sampler codes to accomplish a wide array of LISA-specific analyses. I will briefly mention these codes including where to find them and how to think about using them, as well as our main goals for and work towards creating a community-driven set of tools for LISA data analysis.