About Me

I am Wanyu Zhang, a second-year PhD student at Stanford MS&E working with Prof. Madeleine Udell. Previously, I obtained my bachelor's degree at Shanghai University of Finance and Economics in 2025, where I was fortunate to work with Prof. Yinyu Ye, Prof.Dongdong Ge, and Prof. Huikang Liu.

My research interest lies in large-scale optimization, especially accelerating optimization algorithms with online learning and machine learning methods.

Publications

1. Data-driven Mixed Integer Optimization through Probabilistic Multi-variable Branching
   Yanguang Chen, Wenzhi Gao, Wanyu Zhang, Dongdong Ge, Huikang Liu, Yinyu Ye
   ICML, 2026.
2. Restarted Primal-Dual Hybrid Conjugate Gradient Method for Large-Scale Quadratic Programming
   Yicheng Huang, Wanyu Zhang, Hongpei Li, Dongdong Ge, Huikang Liu, Yinyu Ye
   INFORMS Journal on Computing, 2025.

Preprints

The following work leverages online learning methods to accelerate operator splitting methods (FBS, DRS, ADMM, PDHG) and stochastic gradient methods.

1. Operator Splitting Methods with Online Scaling
   Wanyu Zhang, Wenzhi Gao, Madeleine Udell
   arXiv preprint, 2026.
   The discussion when the envelope is nonsmooth is in the blog post.
2. Stochastic Gradient Methods with Online Scaling
   Wanyu Zhang, Wenzhi Gao, Yinyu Ye, Madeleine Udell
   arXiv preprint, 2026.
3. Enhanced PDHG for Linear Programming with Online Preconditioning
   Haihao Lu, Wanyu Zhang
   arXiv preprint, 2025.
   This work is empirical and the blog post provides a theoretical explanation on how and why online scaling works for PDHG on LP.