Wei Hu

Wei Hu is an Assistant Professor in the Department of Computer Science and Engineering at the University of Michigan's College of Engineering. His research focuses on uncovering the theoretical and scientific foundations of deep learning, aiming to open the black box of neural networks through a combination of theoretical and empirical approaches.

Dr. Hu received his PhD in Computer Science from Princeton University, where he was advised by Sanjeev Arora. Prior to his PhD, he completed his undergraduate studies at Tsinghua University as a member of the prestigious Yao Class. He also served as a FODSI postdoc at UC Berkeley before joining the University of Michigan faculty.

His research interests center around understanding the fundamental mechanisms of deep learning, particularly focusing on training dynamics, generalization properties, and the theoretical underpinnings of neural networks. His work spans both clean, controlled problems and complex real-world models, with recent emphasis on transformer architectures, grokking phenomena, and the implicit biases in neural network training.

Analysis of his recent publications reveals a strong focus on theoretical deep learning with particular attention to transformer models, grokking phenomena, and generalization theory. His work bridges the gap between theoretical understanding and practical deep learning applications, with significant contributions to understanding abrupt learning transitions, representation learning, and the dynamics of neural network training.

• AAAI New Faculty Highlights, 2024
• Google Research Scholar Award, 2023
• Siebel Scholar, 2021
• Best paper award at ICML Workshop on Modern Trends in Nonconvex Optimization for Machine Learning, 2018
• Gordon Y.S. Wu Fellowship, 2016
• Gold medal (1st place), The 27th Chinese Mathematical Olympiad, 2012

Dr. Hu currently advises three PhD students: Pulkit Gopalani, Zhiwei Xu (co-advised with Yixin Wang), and Yongyi Yang. His research group has received substantial funding through awards including the Google Research Scholar Award. He teaches courses including Introduction to Machine Learning (EECS 445) and specialized topics in machine learning theory and large language models (EECS 598/CSE 598).

His research group maintains an active presence in top machine learning conferences, with publications appearing regularly in venues such as NeurIPS, ICML, ICLR, and others. The group's work has gained significant recognition in the theoretical machine learning community for its rigorous approach to understanding deep learning phenomena.