AI for LISA
The Laser Interferometer Space Antenna (LISA) is an ESA-NASA mission adopted in 2024 and expected to be launched in 2035. LISA is the definitive probe of the millihertz gravitational-wave spectrum, populated by massive black hole binaries, extreme mass-ratio inspirals, compact Galactic binaries, and perhaps cosmological sources. Extracting all of these signals from the signal-rich, confused-limited LISA data stream presents major computational and statistical challenges, both fundamental and practical. AI and ML techniques (such as simulation-based inference and neural emulators) have recently attracted significant interest in our community because of their potential efficiency, generality, and flexibility. I will discuss how these methods could be applied to LISA data analysis for tasks of detection, parameter estimation, detector characterization, and waveform generation. I will focus on how AI can complement traditional matched-filtering and MCMC-based pipelines, and I will highlight potential pitfalls and limitations.