Learning, Solving and Optimizing PDEs with TensorGalerkin : an efficient high-performance Galerkin assembly algorithm
Shizheng Wen, Mingyuan Chi, Tianwei Yu, Ben Moseley, Mike Yan Michelis, Hao Sun, Siddhartha Mishra, [0.2cm] ETH Zurich, Switzerland Imperial College London, UK · 2026
Read Learning, Solving and Optimizing PDEs with TensorGalerkin : an efficient high-performance Galerkin assembly algorithm by Shizheng Wen, Mingyuan Chi, Tianwei Yu, Ben Moseley, Mike Yan Michelis, Hao Sun, Siddhartha Mishra, [0.2cm] ETH Zurich, Switzerland Imperial College London, UK (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-05052.txt.
Description
Solving partial differential equations the usual way means building large sparse matrices, a step that can eat most of the compute cost. Shizheng Wen and 7 co-authors attack that snag in this 2026 paper with TensorGalerkin, a toolkit built on the Galerkin method. Their design splits assembly into a map stage over mesh parts and a reduction done by sparse matrix multiplication.
Both stages live inside PyTorch's automatic differentiation, so the assembly graph keeps a fixed count of nodes no matter how big the mesh grows. That trait lets one engine serve three jobs: fast numerical solving, end-to-end differentiable tuning under PDE limits, and physics-informed operator learning. Tests on two- and three-dimensional tasks, covering elliptic, parabolic, and hyperbolic cases on messy meshes, showed clear gains in speed and accuracy over a range of rivals.
Catalog record
| CREATOR | Shizheng Wen, Mingyuan Chi, Tianwei Yu, Ben Moseley, Mike Yan Michelis, Hao Sun, Siddhartha Mishra, [0.2cm] ETH Zurich, Switzerland Imperial College London, UK |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 77 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-05052