Our primary research goal is to advance understanding of the inner workings of deep learning, with a particular emphasis on representation learning. Our research spans a broad range of topics in deep learning.