DELIGHT: Deep Learning Identification of Galaxy Hosts of Transients using Multi-resolution Images

作     者：Förster, Francisco Muñoz Arancibia, Alejandra M. Reyes, Ignacio Gagliano, Alexander Britt, Dylan Cuellar-Carrillo, Sara Figueroa-Tapia, Felipe Polzin, Ava Yousef, Yara Arredondo, Javier Rodríguez-Mancini, Diego Correa-Orellana, Javier Bayo, Amelia Bauer, Franz E. Catelan, Márcio Cabrera-Vives, Guillermo Dastidar, Raya Estévez, Pablo A. Pignata, Giuliano Hernandez-Garcia, Lorena Huijse, Pablo Reyes, Esteban Sánchez-Sáez, Paula Ramirez, Mauricio Grandón, Daniela Pineda-García, Jonathan Chabour-Barra, Francisca Silva-Farfán, Javier 

作者机构： Faculty of Physical and Mathematical Sciences University of Chile Chile  Nuncio Monseñor Sótero Sanz 100 Providencia Santiago Chile  Universidad de Chile Beauchef 851 Santiago8320000 Chile Department of Astronomy University of Chile Las Condes Santiago Chile Department of Astronomy University of Illinois at Urbana-Champaign 1002 W. Green St. IL61801 United States Center for Astrophysical Surveys National Center for Supercomputing Applications UrbanaIL61801 United States National Science Foundation Graduate Research United States Department of Physics Stanford University 382 Via Pueblo Mall StanfordCA94305 United States Kavli Institute for Particle Astrophysics & Cosmology Stanford University P.O. Box 2450 StanfordCA94305 United States Escuela de Ingeniería Eléctrica Pontificia Universidad Católica de Valparaíso Valparaíso Chile Instituto de Física y Astronomía Facultad de Ciencias Universidad de Valparaíso Gran Bretana 1111 Valparaíso Chile Department of Astronomy and Astrophysics The University of Chicago ChicagoIL60637 United States Barcelona School of Informatics Polytechnic University of Catalonia Barcelona08034 Spain Faculty of Physics University of Barcelona Barcelona08028 Spain Data Observatory Foundation Santiago Chile Instituto de Astrofísica Facultad de Física Pontificia Universidad Católica de Chile Santiago Chile European Southern Observatory Karl-Schwarzschild-Strasse 2 Garching bei München85748 Germany Centro de Astroingeniería Facultad de Física Pontificia Universidad Católica de Chile Santiago Chile Department of Computer Science University of Concepción Concepción Chile Instituto de Astrofisica Facultad de Ciencias Exactas Universidad Andres Bello Las Condes Santiago Chile Department of Electrical Engineering University of Chile Santiago Chile Instituto de Informática Universidad Austral Valdivia Chile Fintual Administradora General de Fondos S.A. Santiago Chile Departamento de Física Facultad de Ciencias Físicas y Matemáticas Universidad de Chile Santiago Chile 

出 版 物：《arXiv》 (arXiv)

年 卷 期：2022年

核心收录：

主　　题：Galaxies 

摘      要：We present DELIGHT, or Deep Learning Identification of Galaxy Hosts of Transients, a new algorithm designed to automatically and in real-time identify the host galaxies of extragalactic transients. The proposed algorithm receives as input compact, multi-resolution images centered at the position of a transient candidate and outputs two-dimensional offset vectors that connect the transient with the center of its predicted host. The multi-resolution input consists of a set of images with the same number of pixels, but with progressively larger pixel sizes and fields of view. A sample of 16,791 galaxies visually identified by the ALeRCE broker team was used to train a convolutional neural network regression model. We show that this method is able to correctly identify both relatively large (1000 00) and small (r ≤ 1000) apparent size host galaxies using much less information (32 kB) than with a large, single-resolution image (920 kB). The proposed method has fewer catastrophic errors in recovering the position and is more complete and has less contamination ( © 2022, CC BY-SA.