Geoff Pleiss, Columbia University
Deep learning excels with large-scale unstructured data - common across many modern application domains - while probabilistic modeling offers the ability to encode prior knowledge and quantify uncertainty - necessary for safety-critical applications and downstream decision-making tasks. I will discuss examples from my research that bridge the gap between these two learning paradigms. The first half will show that insights from deep learning can improve the practicality of probabilistic models. I will discuss work that scales Gaussian process regression, a common probabilistic model, to datasets two orders of magnitude larger than previously reported. The second half will show that probabilistic methods can improve our understanding of deep learning. I will demonstrate that Gaussian process theory uncovers new insights about the effects of width and depth in neural networks. I will conclude with ongoing efforts to quantify neural network uncertainty, develop new inductive biases, and other work at the intersection of deep learning and probabilistic modeling.
Short Bio: Geoff Pleiss is a postdoctoral researcher at Columbia University, hosted by John Cunningham, with affiliations in the Department of Statistics and the Zuckerman Institute. He obtained his Ph.D. in Computer Science from Cornell University, advised by Kilian Weinberger, and his B.Sc. from Olin College of Engineering. His research interests are broadly situated in machine learning, including neural networks, Gaussian processes, uncertainty quantification, and scalability. Geoff is also the co-founder and maintainer of the GPyTorch software framework.