Deep learning in astronomy: a tutorial perspective

Article Type

Research Article

Publication Title

European Physical Journal: Special Topics


Astronomy is a branch of science that covers the study and analysis of all extraterrestrial objects and their phenomena. The study brings together the aspects of mathematics, physics, and chemistry to elucidate the origin, evolution, and functions of the Universe and the celestial bodies contained within it. It is a highly complex and challenging task to study and analyze astrophysical phenomena in real-life scenarios by human beings. We need some systems and approaches as our helping hands, that should work instinctively as per the requirements, such as machines. However, machine demands multifaceted information-rich data sets and their processes to work as per the requirement. This aspiration is eventually fulfilled with the advancement of technology, and machine-based analysis of the plethora of available data sets changed our understanding of the Universe and astronomical discoveries. It led to a modern exploring environment that helps to understand observational astronomy (OA) better. The advanced approach of OA has become virtual day by day, which requires a panchromatic approach to the Universe for uncovering more precise and comprehensive physical images. These aspects make us realize that astronomical phenomena’ investigation is mostly data-dependent and necessitates a thorough understanding of complex data analysis and visualization methodologies. For improved analysis of the information obtained from the ancient stages of astronomical research—particularly the observed phenomena; mostly mathematical and statistical techniques were being used. Eventually, modern data analysis techniques played a role in analyzing these massive data and became the essential components of the era of the virtual observatory (VO). At present, the size and analysis of these data sets with complexities are at the challenging stage for VO and the bottleneck for major scientific and technology. Such scenarios of model-driven data analysis paradigm are experiencing a valuable companion of data-driven science for extracting the hidden and novel information from the data sets. A technology-enabled solution is the need of the present scenario in sizeable astronomical data analysis. In this regard, recently, the popularity of ML models is increasing very rapidly within the astronomical domain because of their ability to provide handy solutions. The present tutorial summarizes various aspects of ML that could rightly offer appreciable information and answers to astronomical phenomena. We have also aimed to accommodate different learning aspects of ML and deep learning (DL), including selecting, extracting, and preprocessing the input information.

First Page


Last Page




Publication Date


This document is currently not available here.