A Comparative Study of Convolutional Neural Networks for the Detection of Strong Gravitational Lensing

TL;DR

This paper evaluates convolutional neural networks for detecting strong gravitational lensing in large astronomical surveys, demonstrating high accuracy and efficiency compared to traditional methods.

Contribution

It introduces a new CNN-based framework, LEXACTUM, and benchmarks its performance on gravitational lens detection tasks, showing competitive results.

Findings

High AUC scores of 0.9343 and 0.9870 for space and ground datasets

Fast processing times per image, 0.0061s and 0.0594s

CNNs outperform traditional visual inspection methods

Abstract

As we enter the era of large-scale imaging surveys with the up-coming telescopes such as LSST and SKA, it is envisaged that the number of known strong gravitational lensing systems will increase dramatically. However, these events are still very rare and require the efficient processing of millions of images. In order to tackle this image processing problem, we present Machine Learning techniques and apply them to the Gravitational Lens Finding Challenge. The Convolutional Neural Networks (CNNs) presented have been re-implemented within a new modular, and extendable framework, LEXACTUM. We report an Area Under the Curve (AUC) of 0.9343 and 0.9870, and an execution time of 0.0061s and 0.0594s per image, for the Space and Ground datasets respectively, showing that the results obtained by CNNs are very competitive with conventional methods (such as visual inspection and arc finders) for detecting gravitational lenses.