MIfA Colloquium - Argyro Sasli (UMN)

Title:  LISA Global-Fit: Current situation, improvements and the future

Abstract: LISA is a space-based gravitational-wave (GW) detector scheduled to "fly" in ~10 years. It will observe a wide variety of GW sources simultaneously, including millions of galactic binaries, supermassive black hole mergers, extreme mass-ratio inspirals, unmodeled bursts, and a stochastic GW background. Extracting information from this data stream requires a global-fit framework capable of modeling all signals and noise components together. Current pipelines, based on Gaussian likelihoods, have shown promising results in the LISA Data Challenges. However, recent studies indicate that the data will be non-stationary and non-Gaussian, which can bias parameter estimation. In this talk, I will review the current status of global-fit pipelines, outline limitations related to noise non-stationarity, glitches, and data gaps, and present new approaches to address them. In particular, I will introduce a statistical framework for robust inference in non-Gaussian regimes, which can serve both as a metric for evaluating global-fit pipelines and as a tool to probe the stochastic background. I will also discuss its integration into the global-fit framework. Finally, I will outline some future directions, including the use of machine learning to accelerate parameter estimation. These developments are a big step towards fully exploiting the mission’s astrophysical potential.