All publications

FluxMC: Rapid and High-Fidelity Inference for Space-Based Gravitational-Wave Observations

Bo Liang, Chang Liu, Hanlin Song, …, He Wang, et al.

All 14 authors

Bo Liang, Chang Liu, Hanlin Song, Tianyu Zhao, Minghui Du, He Wang, Haohao Gu, Sensen He, Yuxiang Xu, Wei-Liang Qian, Li-e Qiang, Peng Xu, Ziren Luo, Mingming Sun

PreprintarXiv:2604.049572026

First public:

  • Parameter inference
  • AI for science

Research summary

Flow matching guides parallel-tempering Monte Carlo toward distant regions of the posterior. Benchmarks for space-based massive-black-hole-binary signals examine how this hybrid sampler handles multimodality and costly waveform evaluations, with improved convergence and posterior accuracy in experiments spanning different waveform models.

Extended research summary

FluxMC combines flow matching with parallel-tempering Monte Carlo to guide Bayesian sampling across difficult parameter spaces. Tests on massive black-hole binaries show improved convergence and posterior accuracy for space-based gravitational-wave observations. With higher-mode waveforms, the method converges in under five hours in the reported experiments, addressing the computational cost and local trapping that can hinder conventional samplers.

Extended notes

Benchmarks cover multimodal Gaussian targets and massive-black-hole-binary signals with IMRPhenomD and IMRPhenomHM waveforms. Flow-generated global proposals complement parallel-tempering exploration while the Monte Carlo step preserves asymptotic sampling correctness.