Data-driven Discovery of Transients in the New Era of Time-Domain Astronomy

Time-domain astronomy has reached an incredible new era where unprecedented amounts of data are becoming available. New large-scale astronomical surveys such as the Legacy Survey of Space and Time (LSST) are going to revolutionise transient astronomy, providing opportunities to discover entirely new classes of transients while also enabling a deeper understanding of known classes. LSST is expected to observe over 10 million transient alerts every night, at least two orders of magnitude more than any preceding survey. It has never been more important that astronomers develop fast and automated methods of identifying transient candidates for follow-up observations. In this thesis, I tackle two major challenges facing the future of transient astronomy: the early classification of transients and the detection of rare or previously unknown transients. I detail my development of a number of novel methods dealing with these issues. In the first chapter, I provide an introduction to the field of transient astronomy and motivate why new methods of transient identification are necessary. In the second chapter, I detail the development of a new photometric transient classifier, called RAPID, that is able to automatically classify a range of astronomical transients in real-time. My deep neural network architecture is the first method designed to provide early classifications of astronomical transients. In Chapter 3, I identify the issue that with such large data volumes, the astronomical community will struggle to identify rare and interesting anomalous transients that have previously been found serendipitously. I outline my novel method that uses a Bayesian parametric fit of light curves to identify anomalous transients in real-time. In Chapter 4, I highlight some issues with current photometric classifiers and improve upon RAPID so that it is capable of dealing with real data instead of just simulations. I present classifiers that perform effectively on real data from the Zwicky Transient Facility and the PanSTARRS surveys. Finally, in the last chapter, I discuss the conclusions of my work and highlight some future opportunities and work needed in preparing for discovery in the new era of time-domain astronomy.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2021

Author dc:creator

• Muthukrishna, Daniel

Advisor dc:contributor.advisor

• Mandel, Kaisey

• Astronomy
• Astrophysics
• Machine Learning
• Deep Learning
• Neural Networks
• Supernova
• Astronomical Transients
• Time-series
• Astronomical Surveys
• Classification

Licence

• https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/16411c5e-bc54-4798-be1d-3187a9f93b45/download
• https://www.rioxx.net/licenses/all-rights-reserved/

Language dc:language

eng

Author Identifier

0000-0001-9846-4417

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/330268