Real-time gravitational-wave data analysis with machine
learning
Name
marx-emarx-phd-physics-2026-thesis.pdf
Size
19.4 MB
Format
Adobe PDF
Checksum (MD5)
558ec28138199d5bca025f96b1142257
Author(s)
Marx, Ethan Jacob
Advisor(s)
Katsavounidis, Erotokritos
Date Issued
May 2026
Publisher
Massachusetts Institute of Technology
Abstract
Real-time gravitational-wave data analysis is critical for maximizing the potential of multi-messenger astronomy. During the LIGO-Virgo-KAGRA (LVK) collaboration's first four observing runs, classical statistical methods of matched filtering and Bayesian inference enabled nearly 200 high-significance candidates. Despite this success, these methods suffer from computational burdens that will be exacerbated by increasing detection rates. In this thesis, I develop a complementary solution to real-time data analysis based on machine learning. First, I motivate the need for machine learning in the field of gravitational-wave astronomy. I then describe machine-learning algorithms Aframe, for detecting gravitational-wave signals from compact binary coalescences, and AMPLFI, for estimating the parameters of the source that produced them, and demonstrate performance on par with their classical algorithm counterparts. Next, I detail a software system for deploying these algorithms for real-time analysis. Finally, I apply the algorithms to search for astrophysical gravitational waves in data during the third and fourth LVK observing runs.
MIT Department
Massachusetts Institute of Technology. Department of Physics
Terms of Use
In Copyright - Educational Use Permitted
Copyright retained by author(s)
Persistent DSpace Link
