HelmFluid: Learning Helmholtz Dynamics for Interpretable Fluid Prediction

Authors: Lanxiang Xing, Haixu Wu, Yuezhou Ma, Jianmin Wang, Mingsheng Long

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimentally, Helm Fluid achieves consistent state-of-the-art in both numerical simulated and real-world observed benchmarks, even for scenarios with complex boundaries.
Researcher Affiliation Academia 1School of Software, BNRist, Tsinghua University.
Pseudocode No The paper describes methods like Runge-Kutta and BFECC in text and equations, but does not provide structured pseudocode or algorithm blocks.
Open Source Code Yes Code is available at https://github.com/thuml/Helm Fluid.
Open Datasets Yes ERA5 Z500 with known boundary This dataset is processed from the fifth generation of ECMWF reanalysis (ERA5) data (Hersbach et al., 2020). ... Sea Temperature with unknown boundary This dataset consists of the reanalysis ocean temperature data (MDS) provided by ECMWF.
Dataset Splits Yes Table 8. A summary of experiment datasets. Note that the Spreading Ink dataset is different from other benchmarks, which only contains three video sequences. We strictly follow the Vortex (Deng et al., 2023) to split the video for training, validation and test. For example, in the training phase of video 1, we use the first 70 frames for training and the subsequent 30 frames for validation. As for the test, we use the first 100 frames as input and predict the following 50 frames.
Hardware Specification Yes All the experiments are implemented in Py Torch(Paszke et al., 2019), and conducted on a single NVIDIA A100 40GB GPU.
Software Dependencies No The paper mentions "Py Torch(Paszke et al., 2019)" and "Adam optimizer (Kingma & Ba, 2015)", but it does not provide specific version numbers for PyTorch or other ancillary software components like Python or CUDA.
Experiment Setup Yes Table 9. Experiment configurations in Helm Fluid for different benchmarks. ... BENCHMARK LEARNING RATE BATCH SIZE ... Table 10. Hyperparameter configurations of Helm Fluid for different resolutions.