About Me
I am currently a second-year PhD student at Dartmouth College. I am honored to be advised by Prof. Yaoqing Yang. Prior to this, I earned my Bachelor’s degree in Mathematics and Applied Mathematics from South China University of Technology in 2024. My primary research interests lie in learning theory 📊 and optimization 📈.
✨ Research Interests
I am interested in both classical learning theory problems (Rademacher Complexity, Covering Number, PAC-Bayesian, etc.) and recently emerging theory problems (primarily focusing on the statistical and optimization properties of large-scale deep neural networks). If you have any interesting related problems, feel free to discuss with me anytime! I am very open to cooperation.
- Analyze the optimization properties (Non-Smoothness, Hessian, River-Valley, etc.) of large-scale neural networks, the behavior of optimization algorithms under specific loss landscapes, and develop scalable optimization algorithms based on these optimization properties and algorithms’ behavior.
- Investigate non-vacuous and theoretically interpretable metrics for data/model statistical complexity in neural networks training, and determine how the interplay between data complexity and model complexity can lead to improved generalization and robustness performance.
📖 Education
| 2024.9 - Present | Ph.D in Computer Science Dartmouth College |
| 2020.9 - 2024.6 | B.S. in Mathematics and Applied Mathematics South China University of Technology |
🏆 Selected Awards
📢 News
| Date | Update |
|---|---|
| 2026.06 | 💰 Honored to receive the ISIT 2026 Travel Grant. |
| 2026.06 | 🎉 Our paper, How Does Orthogonalization Adapt to the Neural-Network Hessian Structure? A Gradient Self Outer-Product Analysis at Initialization, has been accepted to the ICML 2026 HILD Workshop. |
| 2026.05 | 🎙️ Gave a talk about RMNP at Dartmouth--Berkeley--Rice AI Reading Group (Recording). |
| 2026.05 | ✨ New blog: A Proof of Orthogonalized and Row-Normalized Descent Directions is now online. |
| 2026.05 | 🎉 Honored to be recognized as an ICML 2026 Golden Reviewer. Grateful for the recognition from the community. |
| 2026.05 | ✨ New blog in understanding Row normalization and orthogonalization for NN online. Welcome to discuss! |
| 2026.05 | 🎉Two paper has been accepted to ICML 2026, including RMNP! We will soon release two blog posts to discuss the rich intuitions behind RMNP. |
| 2026.03 | 🎉 Our paper about Hessian Spectral Bifurcation has been accepted to ISIT 2026. |
| 2026.03 | 📄 New preprint about Matrix-Based Optimization: RMNP: Row-Momentum Normalized Preconditioning for Scalable Matrix-Based Optimization |
| 2026.02 | 🏆 My first paper as first author, Suspicious Alignment of SGD, received Best Student Paper Award at the 37th International Conference on Algorithmic Learning Theory (ALT) 2026! |
| 2026.02 | 🎙️ Giving a talk at Fields Institute on Feb 26th about our ALT paper (Recording) |
| 2026.02 | 📄 New short paper preprint about Hessian: Depth, Not Data: An Analysis of Hessian Spectral Bifurcation |
| 2026.02 | 📝 Serving as a reviewer for ICML 2026 |
| 2025.12 | 🎉 Suspicious Alignment of SGD has been accepted to ALT 2026. |
| 2025.12 | 💰 Honored to receive a $2,000 grant from Lambda AI |
| 2025.10 | 📝 Serving as a reviewer for ICLR 2026 |
| 2025.10 | 👋 First post on this website - welcome! |
📚 Selected Research
* indicates equal contribution.
How Does Orthogonalization Adapt to the Neural-Network Hessian Structure? A Gradient Self Outer-Product Analysis at Initialization
ICML 2026 Workshop on High-dimensional Learning Dynamics
RMNP: Row-Momentum Normalized Preconditioning for Scalable Matrix-Based Optimization
The 43rd International Conference on Machine Learning (ICML 2026)
Suspicious Alignment of SGD: A Fine-Grained Step Size Condition Analysis
The 37th International Conference on Algorithmic Learning Theory (ALT 2026)
Best Student Paper Award
Depth, Not Data: An Analysis of Hessian Spectral Bifurcation
IEEE International Symposium on Information Theory (ISIT 2026) (Extended version in preparation for IEEE Transactions on Information Theory)
🎙️ Recent Talks
Dartmouth--Berkeley--Rice AI Reading Group. Recording
Fields Institute. Recording
Dartmouth--Berkeley--Rice AI Reading Group. Recording
📝 Academic Service
🍽️ Miscellaneous
I am a food enthusiast who loves both eating and cooking. I enjoy preparing a hearty dinner after a busy work schedule and then rewarding myself with a game of StarCraft II 🎮. Here are some photos of the dishes I have made: See Gallery →