The rise of computer vision in galaxy evolution
Modern galaxy surveys yield incredible amount of information, detecting millions to even billions of galaxies from the present-day Universe to the very early Universe, covering most of the cosmic history. One of the most widely used data format is imaging at various wavelengths. By deciphering the information encoded in the multi-wavelength images, astronomers can infer a great deal on the physical properties of galaxies and how they evolve over cosmic time. In the last 10 years or so, there has been an exponential growth of astronomical publications using deep learning techniques, in particular convolutional neural networks, for computer vision tasks. In this talk, I will first give a brief overview of applications of deep learning in various astronomical tasks such as galaxy morphological classification, detection of gravitationally lensed galaxies, and emulating realistic galaxy images. Then I will focus on the recent activities from my research group on classifying colliding galaxies using data obtained from the most powerful telescopes available today.