Robust, rapid, and simple gravitational-wave parameter estimation
稳健、快速、简便的引力波参数估计
Rapid and robust parameter estimation of gravitational-wave sources is a key component of modern multimessenger astronomy. We present a novel and straightforward method for rapid parameter estimation of gravitational-wave sources that uses metric-based importance sampling. The method enables robust parameter estimation of binary neutron star and binary black hole binaries and is trivially parallelized, enabling parameter estimation in seconds with modest resources. The algorithm achieves a median 35% effective sampling efficiency for a population of aligned-spin neutron star binaries sources. Surprisingly, this approach is also highly efficient for analyzing the full 15-dimensional parameter space of typical binary black holes, with a population median 20% efficiency achieved for a source detected primarily by the twin LIGO observatories and 9% for a network of three comparable sensitivity observatories. This method can serve immediate use to improve the low-latency data products of the gravitational-wave observatory network and may be a key component of how the millions of sources observed by next-generation observatories could be analyzed. The approach can also be broadly applied for problems where an approximate likelihood metric space can be constructed.
展开 ▾利用模板库在似然度量空间的铺砌高效构建提案分布,无需训练即可对BNS和BBH分别达到35%和9-20%的中位采样效率,且算法完全可解释,与机器学习方法相比更易排查偏差。
在传统贝叶斯参数估计(如Speagle 2020)耗时耗力的背景下,本研究提出一种基于度量铺砌的重要性采样方法,直接利用在似然空间构建的模板库(Kacanja+ 2024)来生成高效的提案分布,实现了秒级参数估计。相较于近期神经重要性采样方法(Dax+ 2023),本方法无需训练、完全基于第一性原理,因而行为可预测且效率相当。展望未来,该方法可集成到多信使天文学的低延迟管线中,为第三代探测器海量源的分析提供可行方案,其理念亦可推广到其他具有良好度量空间的物理问题。
预印本 2024-10-07 · 刊出 2025-07-22 · 收录 2026-10-13