http://arxiv.org/abs/2210.11428
Understanding the formation and evolution of ring galaxies, galaxies with an atypical ring-like structure, will improve understanding of black holes and galaxy dynamics as a whole. Current catalogs of rings are extremely limited: manual analysis takes months to accumulate an appreciable sample of rings and existing computational methods are vastly limited in terms of accuracy and detection rate. Without a sizable sample of rings, further research into their properties is severely restricted. This project investigates the usage of a convolutional neural network (CNN) to identify rings from unclassified samples of galaxies. A CNN was trained on a sample of 100,000 simulated galaxies, transfer learned to a sample of real galaxies and applied to a previously unclassified dataset to generate a catalog of rings which was then manually verified. Data augmentation with a generative adversarial network (GAN) to simulate images of galaxies was also used. A catalog of 1151 rings was extracted with 7.4 times the precision and 15.4 times the detection rate of conventional algorithms. The properties of these galaxies were then estimated from their photometry and compared to the Galaxy Zoo 2 catalog of rings. With upcoming surveys such as the Vera Rubin Observatory Legacy Survey of Space and Time obtaining images of billions of galaxies, similar models could be crucial in classifying large populations of rings to better understand the peculiar mechanisms by which they form and evolve.
H. Krishakumar and J. Kalmbach
Fri, 21 Oct 22
44/76
Comments: 15 pages, 16 figures. Submitted to AJ
You must be logged in to post a comment.