Lucia Papalini | Searcharxiv

arXiv2025

Can Transformers help us perform parameter estimation of overlapping signals in gravitational wave detectors?

Overlapping signals represent one of the major data analysis challenges in next-generation gravitational wave detectors. We leverage Transformers and Normalizing Flows, state-of-the-art machine learning algorithms, to address the parameter estimation of overlapping binary black hole mergers in the Einstein Telescope (ET). Our proposed model combines a Transformer-based "Knowledge Extractor Neural Network" (KENN) with a Normalizing Flow (HYPERION) to perform rapid and unbiased inference over multiple overlapping black hole binary events. The choice of architecture leverages the strength of Transformers in capturing complex and long-range temporal structures in the strain time series data, while Normalizing Flows provide a powerful framework to sample posterior distributions. We demonstrate the effectiveness and robustness of our model over simulated gravitational wave signals, showing that it maintains the same level of accuracy regardless of the correlation level in the data. Moreover our model provides estimates of chirp mass and coalescence times within <10-20% from the true simulated value. The results obtained are promising and show how this approach might represent a first step toward a deep-learning based inference pipeline for ET and other future gravitational wave detectors.

gr-qc↗

arXiv2024

Deep learning to detect gravitational waves from binary close encounters: Fast parameter estimation using normalizing flows

A yet undetected class of GW signals is represented by the close encounters between compact objects in highly-eccentric e~1 orbits, that can occur in binary systems formed in dense environments such as globular clusters. The expected gravitational signals are short-duration pulses that would repeat over a much longer time scale in case of multiple passages at periastron. These sources represent a unique opportunity of exploring astrophysical formation channels as well as a different way of testing GR. In the case of binary systems containing neutron stars, the observation of these sources could help to constrain the EOS, thanks to the signature left in the GW signal by the f-modes excitation that can occur during the encounter. The detection and PE of these signals is challenging given the short duration of expected signals and the sensitivities of current ground-based GW interferometers. We present a novel approach that exploits Probabilistic ML. We have used Conditional Normalizing Flows to model complex probability distributions and therefore infer posterior distributions for the source parameters. Fast detection and PE is very important as it could trigger electromagnetic follow-up campaigns and offer the possibility to study these events in a multimessenger context. To develop and test the algorithm, we have focused on the simulations of single bursts emission obtained using the Effective Fly-by formalism and embedded in the noise of aLIGO and Virgo during O3. Our proposed model outperforms standard Bayesian methods in accuracy and is 5 orders of magnitude faster, being able to produce 5x10^4 posterior samples in just 0.5s. The results are extremely promising and constitute the first successful attempt for a fast and complete parameter estimation of binary CEs using deep learning, offering a new approach to study the evolution of orbital parameters of compact binary systems.

gr-qc↗