부산대학교 도서관

This paper describes the main challenges encountered when implementing new approaches for handling very large data sets in the area of Deep Cosmological Fields.The ultimate goal pursuit is the knowledge discovery in the area of galaxies surface brightness and by that the improvement of our current knowledge regarding galaxies and their role in the Universe. The specific problem addressed here is the de-blending of very faint and blurred galaxies in crowded images.We introduce a set of artificial intelligence methodologies for the effective implementation of a proof-of-concept, which can then be imported into Big Data Computing platforms, such as XSEDE. This paper is structured as follows: chapter 1 - Introduction, describing the current state-of-the-art and motivation related to the field of Big Data Mining, focussing on astronomical data; chapter 2 - Description of the difficulty of de-blending blurred and faint galaxies, one of the main challenges in the area of complex computing and very large data sets; chapter 3 - New approach, detailing the methodologies used for the problem of de blending blurred and overlapped sources; chapter 4 - Results and Discussion presenting preliminary results, advantages, disadvantages and future lines of work in this area.