Galaxy merger features concealed in the background
During their life and evolution, galaxies can approach each other and collide, becoming some of the most impressive views in the sky. These are galaxy mergers, as their interaction can result in two or more galaxies merging into one. These galaxies show a high variability of morphological distortions due to the tidal forces arising during the process, and their aspect depends on the relative masses, the stage within the process, and the camera resolution. In this talk, I will provide a view of galaxy merger identification, and explain my work that focuses on the combination of Machine Learning (ML) techniques making use of optical images and measurements. In fact, galaxy merger identification in large-scale surveys is one of the main areas of astronomy that is benefitting from the development of automatized ML classifications.
My PhD studies have centred mainly on training and characterizing ML techniques on datasets of mergers. An initial Neural Network was applied to the flux measurements from images in a training set of SDSS DR6 galaxies. The iteration as NN inputs of multiple combinations of these parameters led us to find how the SDSS DR6 sky background error is capable of tracing galaxy mergers with a test-set accuracy of up to 91 %. The interpretation of this result is that the sky background error is tracing low signal-to-noise features around the observed galaxies. We are currently developing a methodology to reproduce this sky error in Subaru/HSC images, with the goal of extending it to deep imaging and understanding better the parameterization. With this work, I want to stress the benefits of interpreting the results of ML models and how it gave us hints to unveil a potential new path for galaxy morphology classification.