论文标题
压缩参数和非参数逼近引力波的可能性
Compressed Parametric and Non-Parametric Approximations to the Gravitational Wave Likelihood
论文作者
论文摘要
准紧凑型二进制合并的重力波观测意味着其参数的后验测量复杂。尽管高斯与相关可能性的高斯近似在该领域具有数十年的历史，但是这些近似值的相对普遍性和实用性尚未得到理解，因此专注于仔细，全面的通用贝叶斯参数推断。在我们以前的三维工作的基础上，我们以示例证明了有界的多元正常可能性近似值是观察到的重力波源的完全可能性的足够准确的表示。适合在https://gitlab.com/xevra/nal-data上发表在Gravitatatoal-Wave瞬态目录中的每个事件，以及https://gitlab.com/xevra/gwalk上的代码发布。我们认为我们的近似值足以用于播种推理，并且与波形模型系统学相比，引入的错误要小得多。为了证明这些近似值作为参数模型的实用性，即单个重力波源的可能性，我们展示了它们在对观察到的重力波源和低延迟参数参数推理的应用中进行建模的示例。
Gravitational-wave observations of quasicircular compact binary mergers imply complicated posterior measurements of their parameters. Though Gaussian approximations to the pertinent likelihoods have decades of history in the field, the relative generality and practical utility of these approximations hasn't been appreciated, given focus on careful, comprehensive generic Bayesian parameter inference. Building on our previous work in three dimensions, we demonstrate by example that bounded multivariate normal likelihood approximations are a sufficiently accurate representation of the full likelihood of observed gravitational-wave sources. Fits for each event published in the Gravitatinoal-Wave Transient Catalogs at https://gitlab.com/xevra/nal-data, along with a code release at https://gitlab.com/xevra/gwalk. We argue our approximations are more than accurate enough for popultion inference and introduce much smaller errors than waveform model systematics. To demonstrate the utility of these approximations as parametric models for the likelihood of individual gravitational-wave sources, we show examples of their application to modeling the population of observed gravitational-wave sources as well as low-latency parameter inference.
