Systematic exploration of the LISA mass spectrum and training‑prior coverage

Systematically explore the full LISA‑accessible mass spectrum of quasi‑circular, spin‑aligned massive black‑hole binaries to ascertain the implications of training‑prior coverage for Dingo’s posterior accuracy and importance‑sampling efficiency, especially in sparsely represented regions such as low chirp masses and high primary spins.

Background

The study finds that sampling efficiency decreases at high SNR and is sensitive to how well different regions of parameter space are represented in the training prior. In particular, cases with low chirp mass and high primary spin show degraded performance due to sparse training coverage and increased waveform complexity.

To address these limitations, the authors indicate that a comprehensive, systematic exploration of the full LISA mass spectrum and its interaction with prior coverage is needed but has not yet been performed.