I am Kangbo Lyu, a Ph.D. student at IIIS, Tsinghua University, adviced by Prof. Tao Du.
My research interests lie in scientific computing and physics-based
simulation, with a focus on large-scale numerical methods for
partial differential equations (PDEs). I am particularly interested
in combining numerical algorithms with machine learning to develop
efficient solvers and preconditioners for large-scale PDE systems. I may explore broader scope neural simulators in future projects.
My recent work explores neural multigrid methods and learned
preconditioners for large-scale sparse linear systems arising from
physics-based simulation. I am interested in bridging classical
numerical methods, such as multigrid and Krylov methods, with
modern machine learning techniques to enable fast and scalable
simulation.
Publications
Learning Sparse Approximate Inverse Preconditioners for Conjugate Gradient Solvers on GPUs Zherui Yang, Zhehao Li, Kangbo Lyu, Yixuan Li, Tao Du, Ligang Liu. NeurIPS 2025 [Project][Paper][Code]
A Multigrid-Inspired Neural Iterative Solver for Poisson Equations on Large Voxel Grids Kangbo Lyu, Ruihong Cen, Yushen Wu, Tao Du. ICLR 2026 Workshop on AI and Partial Differential Equations (Oral) [Project][Paper]
Computational Design of Terrestrial Robots with Anisotropic Friction Hang Hu, Kangbo Lyu, Changyu Hu, Zihan Li, Peiwen Yang, Minchen Li, Shuguang Li, Tao Du. SIGGRAPH North America 2026 (Conference track) [Paper]