Jake Levinson, University of Washington

Friday, October 4, 2019 - 3:30pm to 4:30pm

GWN 201

From July 2018 to July 2019, I worked at Google Research on two projects in machine learning -- one in computer vision, using a mix of geometric and "deep" neural network-based approaches for 3D (triangle mesh) models of people; and the other in what I would call "statistical theory of deep learning", applying random matrix theory to large neural networks.

I'll talk a bit about the work I did, and I'll compare the day-to-day experience of doing research at Google to doing research as a pure mathematician in combinatorics and algebraic geometry.

Related Links: