Speaker
Markus Bachlechner
(RWTH Aachen University)
Description
The improved sensitivity of the Einstein Telescope increases the observable volume of compact binary systems and extends the time window in which the inspiral phase is measurable. Neural Networks (NNs) can efficiently analyze the vast amount of data by reducing computational costs and runtime. This talk presents a fast Binary Black Hole parameter reconstruction by applying a conventional convolutional NN, which conditions a subsequent Normalizing Flow (NF). NFs can learn arbitrarily complex multimodal distributions in multiple dimensions and manifolds. Using the NF, an approximated posterior parameter distribution on an event-by-event basis is obtained faster than in real-time.
Primary authors
Markus Bachlechner
(RWTH Aachen University)
Achim Stahl
(RWTH Aachen University)
