Sampling with prior knowledge for high-dimensional gravitational wave data analysis

Extracting knowledge from high-dimensional data has been notoriously difficult, primarily due to the so-called "curse of dimensionality" and complex joint distributions these dimensions. This is a particularly profound issue for gravitational wave analysis where one requires conduct Bayesian inference estimate posterior distributions. In this study, we incorporate prior physical by sampling desired interim develop training dataset. Accordingly, more relevant regions feature space are covered additional points, such that model can learn subtle but important details. We adapt normalizing flow method be expressive trainable, information effectively extracted represented transformation between target Once trained, our only takes approximately 1 s on V100 GPU generate thousands samples probabilistic purposes. The evaluation approach confirms efficacy efficiency inferences points promising direction similar research. source code, specifications, detailed procedures publicly accessible GitHub.