In this talk, I will give an overview of recent and ongoing research efforts in the visual computing group at JGU Mainz. Our work focuses on applying statistical learning methods to computer graphics as well as several more interdisciplinary research problems. I will start with a discussion of classic computer graphics problems, such as generative image modeling, where we have observed, as many of use, how in particular “deep” representation learning methods can lead to dramatical improvements. I will then outline some ideas for applications of such methods to various interdisciplinary research problems, such as speeding-up simulations in soft-matter physics, medical image analysis, or using concepts from statistical learning theory in social and behavioral science. Finally, if time permits, I would also like to describe some recent efforts in using non-standard representations in order to improve the efficiency of deep neural networks. A lot of the talk covers l work-in-progress, and will hopefully serve as a start into an interesting and open discussion on visual computing in the age of deep learning.