Back to search

University of British Columbia

Searching for strong gravitational lenses using Zernike polynomials.

Abstract

dc:description

Strong gravitational lensing is a direct consequence of general relativity. It refers to the bending of light from a distant source by a massive foreground galaxy, known as the lens. The lensed image of the background source, typically a young star-forming galaxy, appears as a ring or arcs around the lens. Such systems provide a powerful probe of the dark matter content of the lens galaxy. However, strong lenses are rare, so finding them requires surveying vast areas of the sky at high resolution. With its sharp imaging and wide field of view, the Euclid space telescope is expected to find more than 100,000 lenses among roughly 1.5 billion galaxies. Euclid's strong lensing team currently searches for them with computationally expensive deep-learning classifiers. In this thesis, we develop a lightweight alternative based on Zernike polynomials, a complete set of complex orthogonal polynomials on the unit disk, indexed by an order p and a repetition q. We show that the Einstein radius of a lens is inversely proportional to the order, p, of the most strongly excited mode. We decompose 6" x 6" cutouts of Euclid strong lenses and non-lenses into Zernike modes with 0 ≤ p ≤ 40 and q ∈{ 0, 1}, and use the moment amplitudes as features for simple machine-learning methods. Among these, a support vector machine (SVM) with a radial basis function kernel achieves the highest area under the ROC curve (AUC), while positive-unlabelled (PU) learning achieves the highest recall. Neural networks trained with a recall-weighted loss consistently outperform these classical methods. We combine SVM, PU learning, EasyEnsemble and neural networks into an ensemble classifier that reaches an AUC of 0.87 and a recall of 0.77 while requiring about 7000 times fewer operations than the classifiers currently used in the Euclid strong lensing discovery engine. Over the full survey, this would substantially reduce computing time, cost and energy consumption. This work is a step towards efficient and sustainable alternatives to the heavy algorithms used in large astronomical surveys.

Degree

thesis:*
Name thesis:degree_name
Master of Science - MSc
Level thesis:degree_level
master's
Discipline thesis:degree_discipline
Astronomy
Grantor dc:publisher
University of British Columbia
Year dc:date
2026

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Srivastava, Hrishabh
Contributors dc:contributor
  • Scott, Douglas

Rights

dc:rights
Statement dc:rights
  • Attribution-NonCommercial-NoDerivatives 4.0 International
Language dc:language
eng

Identifiers

dc:identifier.*
Handle dc:identifier
http://hdl.handle.net/2429/95746

Chain of custody

source
Harvested from
University of British Columbia
Base URL
circle.library.ubc.ca/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Srivastava, Hrishabh. Searching for strong gravitational lenses using Zernike polynomials.. master's thesis, University of British Columbia, 2026. http://hdl.handle.net/2429/95746