I am a postdoc in the AMS department at JHU, working with Prof. Yannis Kevrekidis. Before that, I was a Research Fellow at the Simons Institute at UC Berkeley for the MPG program in fall 2024. I obtained my PhD in Mathematics at Georgia Institute of Technology, advised by Prof. Molei Tao.
I will be on the academic job market in the 2026-2027 cycle.
Research
I am interested in developing the mathematical foundations of deep learning theory in all its aspects, especially from a dynamical perspective. My research lies at the intersection of machine learning and applied math, combining tools from optimization, (stochastic) dynamics, computational math, analysis, topology, and sampling. I am currently also interested in large language models and diffusion models.
Specifically, my work includes studying the effects of the following:
- Training dynamics
- Large learning rate:
Good regularity creates large learning rate implicit biases: edge of stability, balancing, and catapult with Zhenghao Xu, Tuo Zhao, Molei Tao
Large Learning Rate Tames Homogeneity: Convergence and Balancing Effect with Minshuo Chen, Tuo Zhao, Molei Tao - Regular learning rate (for a family of neural network architectures):
Convergence of Gradient Descent for General Neural Network Architectures Beyond the NTK Regime - Infinitesimal learning rate (gradient flow/NTK regime), and other hyperparameters in diffusion model:
Evaluating the design space of diffusion-based generative models with Ye He, Molei Tao
- Large learning rate:
- Data
Data Uniformity Improves Training Efficiency and More, with a Convergence Framework Beyond the NTK Regime with Shangding Gu - Architecture
Why Do Deep Residual Networks Generalize Better than Deep Feedforward Networks? — A Neural Tangent Kernel Perspective with Kaixuan Huang, Molei Tao, Tuo Zhao
Preprints and Publications
Preprint
Preprint
NeurIPS 2024
NeurIPS 2024
JMLR
IEEE Control Systems Letters 2023
ICLR 2023
ICLR 2022
NeurIPS 2020
Recent talks
- 01/2026: MPG reunion workshop, Simons Institute
- 01/2026: JMM
- 11/2025: SIAM NNP
- 07/2025: Optimization Meets Generative AI: Insights and New Designs, International Conference on Continuous Optimization (ICCOPT)
- 05/2025: Math Machine Learning seminar, MPI MiS + UCLA
- 05/2025: Dynamical Systems for Machine Learning (MS141), SIAM Conference on Applications of Dynamical Systems (DS25)
- 04/2025: CMX seminar, Caltech
- 04/2025: AMS postdoc seminar, JHU
Teaching
Instructor in AMS department, JHU:
- EN 553.432 Bayesian Statistics
Fall 2025, Fall 2026 - EN 553.361 Introduction to Optimization I
Spring 2025, Spring 2026
Teaching Assistant in School of Mathematics, Georgia Tech:
- MATH 2552 Differential Equation
Fall 2018, Spring 2019, Fall 2019, Fall 2020, Spring 2021, Spring 2022, Spring 2023, Fall 2023 - MATH 2550 Introduction to Multivariable Calculus
Summer 2019, Summer 2024 - MATH 2551 Multivariable Calculus
Spring 2020