Wei Hu

Wei Hu 

Wei Hu (胡威)
Ph.D. Student
Department of Computer Science
Princeton University

Email: huwei [at] cs [dot] princeton [dot] edu

[Google Scholar]


I am a fourth year PhD student in the Department of Computer Science at Princeton University. I am very fortunate to be advised by Sanjeev Arora. Previously, I did my undergrad at Tsinghua University, where I was a member of Yao Class. I have also spent time at research labs of Google and Microsoft.

I am interested in the theoretical foundation of modern machine learning and optimization.


Authors are ordered alphabetically, except for papers marked with “(#)”


Conference publications

Blog posts about my work


Understanding Deep Learning via Analyzing Trajectories of Gradient Descent

IIIS-Haihua Frontier Seminar Series at Tsinghua University, December 2019, Beijing, China

Provable Benefit of Orthogonal Initialization in Optimizing Deep Linear Networks

Workshop on Optimization for Machine Learning, December 2019, Vancouver, BC, Canada

On the Connection between Overparameterized Neural Networks and Kernels, and How to Make That Useful

Google, July 2019, New York, NY, USA
Institute for Interdisciplinary Information Sciences at Tsinghua University, June 2019, Beijing, China
The 2nd Machine Learning Theory Workshop at Peking University, June 2019, Beijing, China

Width Provably Matters in Optimization for Deep Linear Neural Networks

ICML, June 2019, Long Beach, CA, USA

Fine-Grained Analysis of Optimization and Generalization for Overparameterized Two-Layer Neural Networks

The 13th Annual Machine Learning Symposium at New York Academy of Sciences, March 2019, New York, NY, USA

On the Dynamics of Gradient Descent for Training Deep Neural Networks

Princeton-IAS Theoretical Machine Learning Seminar, October 2018, Princeton, NJ, USA

Algorithmic Regularization in Learning Deep Homogeneous Models: Layers are Automatically Balanced

ICML 2018 Workshop on Modern Trends in Nonconvex Optimization for Machine Learning, July 2018, Stockholm, Sweden

An Analysis of the t-SNE Algorithm for Data Visualization

COLT, July 2018, Stockholm, Sweden

Linear Convergence of a Frank-Wolfe Type Algorithm over Trace-Norm Balls

NIPS, December 2017, Long Beach, CA, USA

New Characterizations in Turnstile Streams with Applications

CCC, May 2016, Tokyo, Japan

Combinatorial Multi-Armed Bandit with General Reward Functions

Microsoft Research Asia Theory Group Seminar, March 2016, Beijing, China


Princeton University


Selected awards