Astronomical transient event recognition with machine learning

"The study of astronomical transient events will occur on unprecedented scale with the next generation of multi-epoch and multi-band astronomical surveys.  The scientific success of these surveys rely on the automatized recognition and classification of such events. In this project we propose and test a methodology to perform transient event classification.  The method takes as an input observed brightness time series, also known as light curves, to produce simple statistical descriptors which are then fed into multiple machine learning models to finally detect and classify between different types of transient objects. We apply this method to observational data from the Catalina Real Time Transient survey and perform five different binary and multi-class classification experiments.  We find that Random Forests are the best performing models, scoring a recall of 89 % on binary (transient & non-transient) classification. Six-class transient classification scored a 77% recall, and a 66% recall when including ambiguous sources class with new samples; including non-transient sources as an additional class yields similar results. These results are similar to the ones presented in the astronomical literature, thus becoming a feasible alternative to be used in upcoming astronomical surveys."--Tomado del Formato de Documento de Grado

• Autores: Gómez Mosquera, Diego Alejandro
• Tipo de recurso: Trabajo de grado de pregrado
• Fecha de publicación: 2017
• Institución: Universidad de los Andes
• Repositorio: Séneca: repositorio Uniandes
• Idioma: eng
• OAI Identifier: oai:repositorio.uniandes.edu.co:1992/39577
• Acceso en línea: http://hdl.handle.net/1992/39577
• Palabra clave: Astrofísica relativística; Erupciones de rayos gamma; Astronomía con rayos X; Fenómenos transitorios (Dinámica); Aprendizaje automático (Inteligencia artificial); Phyton (Lenguaje de programación de computadores); Ingeniería
• Rights: openAccess
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/