Identifying Multiply-Imaged Regions in Gravitationally Lensed Galaxies Using Machine Learning Techniques

Sammanfattning

The aim of this paper was to explore the possibility of introducing machine learning techniques into the field of gravitational lensing analysis. More specifically, clustering algorithms were used to identify multiply-imaged regions in a gravitationally lensed galaxy, with the hope of in the future being able to improve constraints on gravitational lensing models. To do this, a software package for fitting emission lines in integral field unit data cubes from James Webb Space Telescope observations with the Near Infrared Spectrograph was used to retrieve emission line fluxes and kinematic properties of the galaxy SGAS J1723+3411. These properties were then used as parameters in a multi-dimensional parameter space, where each spaxel in the data cube represents one point in the parameter space. The clustering algorithms then labeled these points based on how they clustered together. The results showed that, while the algorithms correctly labeled several identical regions as the same cluster, they were not sufficiently clear in doing so, with the classification being quite noisy in some areas. However, the method did still show some promise and is a potential avenue for future work.