Back to search

University of Southampton

A data-driven study of variability in X-ray binaries using machine learning

Abstract

dc:description.abstract

The field of time series data mining has become increasingly important in light of the impending era of “Big Data”. With new astronomical sky surveys set to generate data at an unprecedented rate, the need for automated methods of data analysis is becoming increasingly urgent. This thesis makes a contribution to the toolbox of automated characterisation methods for astronomical time series, specifically light curves. Two end-to-end pipelines drawing on concepts from the fields of signal processing, data science, machine learning, and deep learning are presented, with examples showing how they can streamline the process of light curve analysis. The first pipeline is designed to extract and aggregate features from light curve segments, using a neural network and clustering algorithms to combine them into feature vectors of fixed length. These “light curve fingerprints” can be used as input for downstream machine learning algorithms, providing a way to tackle the problem of constructing a standard representation for light curves of variable length. I demonstrate the proposed method on the X-ray data of the Galactic low-mass black hole X-ray binary system GRS 1915+105, showcasing how it can be used to study the variability of the source on time scales ranging from seconds to hours, and to quantify the similarity of different light curves. The outcomes highlight the problem of the popular classification system of GRS 1915+105 observations, which does not account for intermediate class behaviour. I also explore the extension to another source, and outline the direction of further development. The fourth chapter of the thesis presents a periodicity detection and characterisation pipeline, which utilises the wavelet transform for timing-frequency analysis. The pipeline is applied to a sample of photometric light curves of Small Magellanic Cloud Be X-ray binaries, demonstrating its ability to facilitate analysis of the sources’ behaviour over time scales ranging from days to decades. Using this methodology, I detect several periodic signals for the first time, and provide the interpretation of their physical origin. The pipeline is a useful tool in the analysis of the binary orbital behaviour of the sources, the evolution of Be star non-radial pulsations, and their possible link with the formation mechanism of the circumstellar decretion discs.

Degree

thesis:*
Name dc:type.qualificationname
Ph.D.
Level dc:type.qualificationlevel
doctoral
Grantor dc:publisher.institution
University of Southampton
Year dc:date.issued
2024

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Orwat-Kapola, Jakub Kacper
Advisors dc:contributor.advisor
  • Bird, Tony
  • Hill, Adam
  • Altamirano, Diego

Rights

Language dc:language
English, eng

Chain of custody

source
Harvested from
University of Southampton
Base URL
eprints.soton.ac.uk/cgi/oai2
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Orwat-Kapola, Jakub Kacper. A data-driven study of variability in X-ray binaries using machine learning. doctoral thesis, University of Southampton, 2024.