Deep Learning for Image Processing in Astrophysical Experiments
Maxim Borisyak Yandex School of Data Analysis, Moscow, Russia National Research University Higher School of Economics, Moscow, Russia (pdf of presentation is available here)
In recent years, Deep Learning has become a powerful tool for Data Analysis including image processing. Notably, it became the first Machine Learning algorithm that surpassed human performance in visual pattern recognition. Today usage of Deep Learning methods in natural sciences, such as High Energy Physics and Astrophysics, is rapidly growing. In this talk, we cover methods for image processing in the astrophysical experiment. Two particular methods, namely, track recognition and learning read-out model from real data, are explained in detail. For illustration purposes, we consider an astrophysical experiment: Cosmic Rays Found In Smartphones, which proposes usage of private mobile phones as a ground detector for Ultra High Energy Cosmic Rays. Unusual structure of the detector and unknown properties of individual sensors lead to a number of challenges which can be bypassed with the help of Deep Learning methods.