My Favorite Deep Learning Papers of 2017

Even with so many deep learning papers coming out this year, there were a few publications I felt managed to rise above the rest. Here are the five papers that impacted my mental models the most over the last year. For each, I state the "goal" of the paper, briefly summarize the work, and explain why I found it so interesting. Rather than describe exactly what the authors did here, I'll let some of the incredible results stand on their own: These stunning images are from the CycleGAN paper, in which the authors learn a pair of translation networks capable of translating between unpaired sets of images. These stunning images are from the CycleGAN paper, in which the authors learn a pair of translation networks capable of translating between unpaired sets of images.