Zherui Yang (杨哲睿) | University of Science and Technology of China

About Me

I am a third-year Ph.D. student at the University of Science and Technology of China (USTC), advised by Prof. Ligang Liu. I also work closely with Assistant Professor Tao Du at Tsinghua University. Prior to that, I obtained my bachelor’s degree in Computer Science and Technology at Chien-Shiung Wu College, Southeast University. I am currently seeking an internship opportunity, particularly in multimodal models and world models.

Research Interests

Machine Learning for Numerical Linear Algebra: Developing data-driven methods to accelerate numerical linear algebra algorithms and provide efficient, effective alternatives to traditional approaches.
Neural Operators: Solving partial differential equations (PDEs) using fully learned, end-to-end, feed-forward neural networks.
Efficient Physics-Based Simulation: Developing efficient GPU-based simulation methods for elastodynamics and rigid body dynamics, including traditional, data-driven, and hybrid approaches.

Publications

NeurIPS

Learning Sparse Approximate Inverse Preconditioners for Conjugate Gradient Solvers on GPUs

Zherui Yang, Zhehao Li, Kangbo Lyu, Yixuan Li, Tao Du, Ligang Liu

The Thirty-Ninth Annual Conference on Neural Information Processing Systems (NeurIPS), 2025.

PDF Code

Open Source Projects

These are open-source projects I maintain on GitHub. Most are closely related to my research interests and are still under active development. If any of them interest you, feel free to contact me or visit the corresponding GitHub repositories for more details.

mathprim: A lightweight tensor library in C++20 with CUDA support, OpenMP parallelization, automatic dispatch across cuBLAS, cuSPARSE, and OpenBLAS, as well as built-in optimizer and sparse solver support.
ssim: A simple simulator for elastodynamics and mass-spring systems, featuring efficient GPU-based simulation methods.
tinygs: A lightweight 3D Gaussian Splatting library in C++ with CUDA. It is 10-th implementation for High-Quality and Fast 3DGS Reconstruction Challenge.