MMET: A Multi-Input and Multi-Scale Transformer for Efficient PDEs Solving
Authors: Yichen Luo, Jia Wang, Dapeng Lan, Yu Liu, Zhibo Pang
IJCAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental evaluations on diverse benchmarks spanning different physical fields demonstrate that MMET outperforms SOTA methods in both accuracy and computational efficiency. |
| Researcher Affiliation | Collaboration | Yichen Luo1 , Jia Wang2 , Dapeng Lan3 , Yu Liu3 and Zhibo Pang ,1,4 1Division of ISE, KTH Royal Institute of Technology, Stockholm, Sweden 2School of Advanced Technology, Xi an Jiaotong-Liverpool University, Suzhou, China 3Techforgood AS, Oslo, Norway 4Department of Automation Technology, ABB Corporate Research Sweden, Vasteras, Sweden |
| Pseudocode | No | No explicit pseudocode or algorithm blocks are present in the paper. The methodology is described in text and architectural diagrams like Figure 2 and Figure 3. |
| Open Source Code | Yes | This work is open-sourced at https://github.com/Yichen Luo-0/MMET. |
| Open Datasets | Yes | We select six representative datasets spanning elasticity, fluid mechanics, and thermodynamics. [...] Darcy Flow [Takamoto et al., 2024]: The 2D Darcy flow problem from the PDEBench datasets... Shape-Net Car (A) [Umetani and Bickel, 2018]... Heat2d (A, B) [Hao et al., 2023]... Beam2d (B, C) [Bai et al., 2023]... |
| Dataset Splits | Yes | For the Beam2d dataset, in the training stage, our model uses an unstructured mesh with 5404 nodes generated by Ansys Workbench. The query sequence is a regular point matrix with a resolution of [50 20]. |
| Hardware Specification | Yes | The model is trained on two NVIDIA H20 GPUs. |
| Software Dependencies | No | The paper mentions using Adam W [Loshchilov and Hutter, 2019] or L-BFGS [Liu and Nocedal, 1989] optimizers, but does not specify software dependencies with version numbers (e.g., Python, PyTorch, CUDA versions). |
| Experiment Setup | No | The paper mentions using Adam W or L-BFGS optimizers and states 'The model scale and training configuration are maintained consistently in all settings' but does not provide specific hyperparameter values like learning rate, batch size, or number of epochs. |