Identifying galaxy mergers in observations and simulations with deep learning

Abstract

Context. Mergers are an important aspect of galaxy formation and evolution. With large upcoming surveys, such as Euclid and LSST, accurate techniques that are fast and efficient are needed to identify galaxy mergers for further study. Aims. We aim to test whether deep learning techniques can be used to reproduce visual classification of observations, physical classification of simulations and highlight any differences between these two classifications. With one of the main difficulties of merger studies being the lack of a truth sample, we can use our method to test biases in visually identified merger catalogues. Methods. A convolutional neural network architecture was developed and trained in two ways: one with observations from SDSS and one with simulated galaxies from EAGLE, processed to mimic the SDSS observations. The SDSS images were also classified by the simulation trained network and the EAGLE images classified by the observation trained network.Results. The observationally trained network achieves an accuracy of 91.5% while the simulation trained network achieves 65.2% on the visually classified SDSS and physically classified EAGLE images respectively. Classifying the SDSS images with the simulation trained network was less successful, only achieving an accuracy of 64.6%, while classifying the EAGLE images with the observation network was very poor, achieving an accuracy of only 53.0% with preferential assignment to the non-merger classification. This suggests that most of the simulated mergers do not have conspicuous merger features and visually identified merger catalogues from observations are incomplete and biased towards certain merger types. Conclusions. The networks trained and tested with the same data perform the best, with observations performing better than simulations, a result of the observational sample being biased towards conspicuous mergers. Classifying SDSS observations with the simulation trained network has proven to work, providing tantalizing prospects for using simulation trained networks for galaxy identification in large surveys.

Deep learning predictions of galaxy merger stage and the importance of observational realism

“…Ackermann et al 2018;Walmsley et al 2019). Following the approaches adopted for quantitative morphologies by calibrating on hydrodynamical simulations, Pearson et al (2019) train a CNN on synthetic Sloan Digital Sky Survey (SDSS) images of galaxies from the EA-GLE simulation (Schaye et al 2015) and examine biases from redshift, star-formation rates, and apparent brightness on merger and non-merger classifications -though with poor classification performance (65.5% in a binary classification). Nonetheless, one of the key elements of their synthetic SDSS images is that they were inserted into a handful of SDSS survey fields in an attempt to match observational biases in real images (realistic skies, resolution, and crowding by nearby sources).…”

Physical drivers of the cosmic star formation history

“…All ages and times are in Gyr. 3.50, 4.00, 4.50, 5.00, 5.50, 6.00, 6.50, 7.00, 7.50, 8.00, 8.50, 9.00, 9.50, 10.00, 10.50, 11.00, 12.00, 13.00 Table 5.8: Terms used when describing neural network performance from Pearson et al (2019) Term Definition Positive (P) An object classified in the catalogues or identified by a network as a merger. Negative (N) An object classified in the catalogues or identified by a network as a non-merger.…”

“…For the SDSS data, we use the network trained on SDSS images from Pearson et al (2019). The merging and non-merging galaxies used to train this network were collected following .…”

“…A complete and thorough description of CNNs is beyond the scope of this paper but further details are explained in . This paper uses the definitions of Pearson et al (2019) for the terms to describe the properties of CNNs. These terms may be an alternate nomenclature to other works or may be unfamiliar.…”

“…For this work, we use the architecture developed in Pearson et al (2019) and use this to train on data from CANDELS and KiDS (the network trained in Pearson et al (2019) is Figure 5.8: Schematic representation of the architecture of the CNN used with input (three or one colour, 64×64 pixel image) on the left and output (binary classification of merger or non-merger) on the right. The sizes of the kernels (red) and layers are shown with the input layer having a depth of three for the SDSS (gri) and CANDELS (1.6 µm, 1.25 µm, and 814 nm) data and one for the KiDS (r-band) data.…”