arXiv subjects

Robin S. H. Yuen

Publications and source records attributed to Robin S. H. Yuen.

2 recordsLinked to original sources

arXiv2021

Identification of Lensed Gravitational Waves with Deep Learning

Similar to light, gravitational waves (GWs) can be lensed. Such lensing phenomena can magnify the waves, create multiple images observable as repeated events, and superpose several waveforms together, inducing potentially discernible patterns on the waves. In particular, when the lens is small, $\lesssim 10^5 M_\odot$, it can produce lensed images with time delays shorter than the typical gravitational-wave signal length that conspire together to form ``beating patterns''. We present a proof-of-principle study utilizing deep learning for identification of such a lensing signature. We bring the excellence of state-of-the-art deep learning models at recognizing foreground objects from background noises to identifying lensed GWs from noise present spectrograms. We assume the lens mass is around $10^3 M_\odot$ -- $10^5 M_\odot$, which can produce the order of millisecond time delays between two images of lensed GWs. We discuss the feasibility of distinguishing lensed GWs from unlensed ones and estimating physical and lensing parameters. Suggested method may be of interest to the study of more complicated lensing configurations for which we do not have accurate waveform templates.

gr-qc↗

arXiv2019

Ranking Candidate Signals with Machine Learning in Low-Latency Search for Gravitational-Waves from Compact Binary Mergers

In the multi-messenger astronomy era, accurate sky localization and low latency time of gravitational-wave (GW) searches are keys in triggering successful follow-up observations on the electromagnetic counterpart of GW signals. We, in this work, focus on the latency time and study the feasibility of adopting supervised machine learning (ML) method for ranking candidate GW events. We consider two popular ML methods, random forest and neural networks. We observe that the evaluation time of both methods takes tens of milliseconds for $\sim$ 45,000 evaluation samples. We compare the classification efficiency between the two ML methods and a conventional low-latency search method with respect to the true positive rate at given false positive rate. The comparison shows that about 10\% improved efficiency can be achieved at lower false positive rate $\sim 2 \times 10^{-5}$ with both ML methods. We also present that the search sensitivity can be enhanced by about 18\% at $\sim 10^{-11}$Hz false alarm rate. We conclude that adopting ML methods for ranking candidate GW events is a prospective approach to yield low latency and high efficiency in searches for GW signals from compact binary mergers.

astro-ph.IM↗