Machine learning for gravitational waves

The direct detection of gravitational waves by the LIGO–Virgo Collaboration marked a major advance in modern astrophysics, confirming a key prediction of Einstein’s general theory of relativity and opening a new observational window on the Universe and fundamental physics [1]. Current detections of compact-object mergers involving neutron stars and black holes will soon be complemented by next-generation ground- and space-based interferometers, most notably the Einstein Telescope and LISA [2]. Extracting robust physical and astrophysical information from gravitational-wave signals is a computationally intensive task, particularly in the context of large parameter spaces and precision tests of fundamental physics. Machine learning methods offer powerful tools to address these challenges by enabling fast detection, waveform modelling, and parameter inference [3,4]. This PhD project focuses on the development and application of advanced machine learning techniques for gravitational-wave data analysis. The research will target efficient detection and rapid parameter inference for signals from astrophysical sources, with the broader goal of probing fundamental physics. The project will combine state-of-the-art deep learning algorithms with high-performance and GPU-based computing for waveform simulation and network training. Part of the work will be carried out within the international collaborations of LISA and the Einstein Telescope. The successful candidate will therefore interact with a wide range of researchers, including theorists, phenomenologists, data analysts, and software engineers, and will gain experience working in large scientific collaborations.

Expected qualifications: Strong programming skills in Python, or clear evidence of the ability to rapidly develop them, are essential. Prior experience with machine learning and/or gravitational-wave theory or modelling is advantageous but not required.

References:

I. D. Saltas et al (2018) — arXiv: 1812.03969
E. Barausse et al (2020) — arXiv: 2001.09793
E. Cuoco et al (202) — arXiv: 2005.03745
G. Ventagli & I. D. Saltas (2024) — arXiv: 2405.17908