pith. sign in

arxiv: 2505.20349 · v2 · pith:EPWCS272new · submitted 2025-05-25 · ⚛️ physics.flu-dyn · cs.LG

FD-Bench: A Modular and Fair Benchmark for Data-driven Fluid Simulation

Haixin Wang , Ruoyan Li , Fred Xu , Fang Sun , Kaiqiao Han , Zijie Huang , Ching Chang , Xiao Luo
show 2 more authors
Wei Wang Yizhou Sun
This is my paper

Pith reviewed 2026-05-22 12:55 UTC · model grok-4.3

classification ⚛️ physics.flu-dyn cs.LG
keywords data-driven fluid simulationbenchmarkneural PDE solversreproducibilitymodular evaluationgeneralization analysisfluid dynamicsnumerical solver comparison
0
0 comments X

The pith

FD-Bench supplies a modular benchmark that ranks 85 data-driven fluid models across 10 scenarios with standardized protocols.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces FD-Bench to address fragmented evaluation practices in neural PDE solvers for fluid dynamics. It supplies unified datasets, evaluation protocols, and a modular breakdown that isolates spatial, temporal, and loss-function choices. The benchmark runs 85 models through 10 representative flow scenarios under one experimental setup. This setup also includes direct head-to-head tests against traditional numerical solvers plus tests of generalization across resolutions, initial conditions, and time windows. A reader would care because reproducible leaderboards remove the main obstacle to steady progress in the field.

Core claim

FD-Bench systematically evaluates 85 baseline models across 10 representative flow scenarios under a unified experimental setup. It provides a modular design that enables fair comparisons across spatial, temporal, and loss function modules, the first systematic framework for direct comparison with traditional numerical solvers, fine-grained generalization analysis across resolutions, initial conditions, and temporal windows, and a user-friendly extensible codebase.

What carries the argument

The modular design that separates spatial, temporal, and loss modules together with ten representative flow scenarios under one experimental protocol.

If this is right

  • Model developers can isolate the effect of any single module without setup differences confounding the results.
  • The leaderboard supplies the first consistent ordering of architectures that can be compared directly to classical solvers.
  • Generalization tests across resolution and time window reveal which models remain stable when conditions change.
  • The open codebase lets researchers add new models or scenarios without rebuilding the evaluation stack.
  • Future work can extend the same modular protocol to three-dimensional or multi-physics problems.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The benchmark could serve as a template for standardized testing in related areas such as solid mechanics or combustion modeling.
  • Hybrid neural-numerical solvers might be ranked more reliably once the same modular splits are applied to them.
  • Engineering teams could use the leaderboard to select models for real-time control tasks where both accuracy and speed matter.
  • Community extensions might add uncertainty quantification or inverse-problem benchmarks on top of the existing structure.

Load-bearing premise

The ten chosen flow scenarios and the splits between spatial, temporal, and loss modules are enough to produce rankings that reflect real performance differences outside the tested cases.

What would settle it

A previously unseen flow scenario or a shift in initial conditions in which currently lower-ranked models outperform the current leaders would show that the benchmark rankings do not generalize.

Figures

Figures reproduced from arXiv: 2505.20349 by Ching Chang, Fang Sun, Fred Xu, Haixin Wang, Kaiqiao Han, Ruoyan Li, Wei Wang, Xiao Luo, Yizhou Sun, Zijie Huang.

Figure 1
Figure 1. Figure 1: Limitations identified in three key areas. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: A schematic illustration of common approaches for each key module in data-driven neural [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: We collect and generate 10 representative fluid flow scenarios that span a diverse range of physical conditions. We also present the corresponding visualizations. This approach captures the full distribution of flow fields, enabling stochastic sampling and modeling of complex outcomes. Its drawbacks include significantly higher computational cost due to multiple diffusion time steps and the need for carefu… view at source ↗
Figure 4
Figure 4. Figure 4: We compare FNO against traditional solver ( [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of rollout performance between FNO (grid data), MeshGraphNets (mesh data), [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Runtime of FNO against pseudo-spectral solver operating at lower resolutions on the incompressible N-S. Solver x implies pseudo-spectral solver operating at x × x resolution. We present the results for incompressible N-S in [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Extrapolation evaluation on zero-shot generalization across diverse initial conditions, [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: We compare FNO against pseudo-spectral solver operating at lower resolutions on Burgers’ [PITH_FULL_IMAGE:figures/full_fig_p021_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: We compare FNO against pseudo-spectral solver operating at lower resolutions on Advec [PITH_FULL_IMAGE:figures/full_fig_p021_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: We compare the runtime of the FNO against pseudo-spectral solver operating at lower [PITH_FULL_IMAGE:figures/full_fig_p022_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: RMSE versus model capacity for five architectures at four scaling levels. [PITH_FULL_IMAGE:figures/full_fig_p023_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Visualization of Fourier+Next-step+Variable on Compressible N-S. I Limitations and Broad Impact Despite our efforts to cover a diverse set of architectures, our benchmark is limited by time constraints and thus does not yet include all possible representation paradigms (e.g., higher-order spectral methods, graph-wavelet embeddings, or learned Lagrangian bases). Expanding to these and other emerging modali… view at source ↗
Figure 13
Figure 13. Figure 13: Visualization of Fourier+Next-step+Variable on Compressible N-S [PITH_FULL_IMAGE:figures/full_fig_p029_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Visualization of Fourier+Next-step+Variable on Compressible N-S. 29 [PITH_FULL_IMAGE:figures/full_fig_p029_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Visualization of Fourier+Next-step+Variable on Diffusion-Reaction. Prediction Ground Truth Residua l [PITH_FULL_IMAGE:figures/full_fig_p030_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Visualization of Conv+Next-step+Variable on Stochastic N-S . 30 [PITH_FULL_IMAGE:figures/full_fig_p030_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Visualization of Fourier+Next-step+Variable on Stochastic N-S . Prediction Ground Truth Residua l [PITH_FULL_IMAGE:figures/full_fig_p031_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: Visualization of Latent+Next-step+Variable on Stochastic N-S . 31 [PITH_FULL_IMAGE:figures/full_fig_p031_18.png] view at source ↗
read the original abstract

Data-driven modeling of fluid dynamics has advanced rapidly with neural PDE solvers, yet a fair and strong benchmark remains fragmented due to the absence of unified PDE datasets and standardized evaluation protocols. Although architectural innovations are abundant, fair assessment is further impeded by the lack of clear disentanglement between spatial, temporal and loss modules. In this paper, we introduce FD-Bench, the first fair, modular, comprehensive and reproducible benchmark for data-driven fluid simulation. FD-Bench systematically evaluates 85 baseline models across 10 representative flow scenarios under a unified experimental setup. It provides four key contributions: (1) a modular design enabling fair comparisons across spatial, temporal, and loss function modules; (2) the first systematic framework for direct comparison with traditional numerical solvers; (3) fine-grained generalization analysis across resolutions, initial conditions, and temporal windows; and (4) a user-friendly, extensible codebase to support future research. Through rigorous empirical studies, FD-Bench establishes the most comprehensive leaderboard to date, resolving long-standing issues in reproducibility and comparability, and laying a foundation for robust evaluation of future data-driven fluid models. The code is open-sourced at https://github.com/WillDreamer/FD-Bench.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 3 minor

Summary. The manuscript introduces FD-Bench, a modular benchmark for data-driven fluid simulation. It evaluates 85 baseline models across 10 representative flow scenarios under a unified experimental setup, with contributions including a design that disentangles spatial, temporal, and loss modules; direct comparisons to traditional numerical solvers; generalization tests across resolutions, initial conditions, and temporal windows; and an open-source extensible codebase. The central claim is that this framework resolves long-standing reproducibility and comparability issues and establishes the most comprehensive leaderboard to date.

Significance. If the 10 scenarios and modular protocol produce rankings that reflect intrinsic model differences rather than benchmark-specific choices, FD-Bench would provide a useful standardized platform for the community and support more reliable evaluation of neural PDE solvers. The open code release is a clear strength that aids reproducibility. The significance is tempered by the need to verify that performance differences generalize beyond the selected flows and module combinations.

major comments (3)
  1. [Section 4.1] Section 4.1 and Table 1: The ten flow scenarios are presented as representative, but the manuscript lacks quantitative coverage analysis (e.g., range of Reynolds numbers, dimensionality, or boundary complexity). This is load-bearing for the claim that the leaderboard rankings are generalizable and resolve comparability issues.
  2. [Section 5.3] Section 5.3: The generalization analysis reports metrics across resolutions and initial conditions but omits statistical significance tests, error bars from multiple runs, or cross-validation details. Without these, it is difficult to assess whether observed differences are robust or could be artifacts of the specific data splits.
  3. [Section 3.2] Section 3.2: The modular framework is described as enabling fair isolation of spatial, temporal, and loss modules, yet the reported experiments do not include full ablations or interaction tests. This weakens the assertion that rankings reflect disentangled module contributions rather than combined effects.
minor comments (3)
  1. [Figure 2] Figure 2: The modular architecture diagram would benefit from explicit labels on data flow between spatial and temporal modules to improve readability.
  2. [Introduction] The abstract and introduction cite the absence of unified datasets, but a brief comparison table to prior fluid benchmarks (e.g., in related work) would strengthen the novelty claim.
  3. [Appendix A] Appendix A: Hyperparameter ranges and exact data preprocessing steps should be listed explicitly to complement the open-source code.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive and detailed comments. We address each major comment point by point below, clarifying our approach and noting the revisions incorporated to improve the manuscript.

read point-by-point responses

  1. Referee: [Section 4.1] Section 4.1 and Table 1: The ten flow scenarios are presented as representative, but the manuscript lacks quantitative coverage analysis (e.g., range of Reynolds numbers, dimensionality, or boundary complexity). This is load-bearing for the claim that the leaderboard rankings are generalizable and resolve comparability issues.

    Authors: We agree that explicit quantitative coverage metrics would strengthen the justification for the selected scenarios. In the revised manuscript we have expanded Section 4.1 with a new table (Table 2) that reports the Reynolds-number range (10^0 to 10^6), dimensionality (2-D and 3-D), and boundary-condition types (periodic, no-slip, inflow/outflow) for each of the ten flows. This analysis shows that the benchmark spans laminar-to-turbulent regimes and a variety of boundary complexities, thereby supporting the generalizability of the reported rankings. revision: yes

  2. Referee: [Section 5.3] Section 5.3: The generalization analysis reports metrics across resolutions and initial conditions but omits statistical significance tests, error bars from multiple runs, or cross-validation details. Without these, it is difficult to assess whether observed differences are robust or could be artifacts of the specific data splits.

    Authors: We acknowledge the value of statistical rigor. The revised Section 5.3 now includes error bars computed from five independent runs that differ in random seeds for data splitting and initialization. We have also added a 3-fold cross-validation over initial conditions and temporal windows, together with paired Wilcoxon signed-rank tests. The tests confirm that the performance gaps between leading models remain statistically significant (p < 0.01), reducing the likelihood that the observed trends are artifacts of particular splits. revision: yes

  3. Referee: [Section 3.2] Section 3.2: The modular framework is described as enabling fair isolation of spatial, temporal, and loss modules, yet the reported experiments do not include full ablations or interaction tests. This weakens the assertion that rankings reflect disentangled module contributions rather than combined effects.

    Authors: The referee is correct that the primary experiments emphasize end-to-end leaderboard construction rather than exhaustive module ablations. To address this concern we have added a new Section 5.4 that presents systematic one-at-a-time ablations (fixing two modules while varying the third) on a representative subset of scenarios, together with an analysis of pairwise interaction effects. The results indicate that module contributions are largely additive, with only modest interactions in a few cases; these findings are now reported to support the claim that the modular design enables disentangled evaluation. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical benchmark with direct experimental rankings

full rationale

The paper introduces FD-Bench as an empirical evaluation framework that runs 85 models on 10 fixed flow scenarios under a unified modular protocol for spatial, temporal, and loss components. Leaderboard rankings arise from direct execution on provided datasets and open-sourced code rather than any derivation, first-principles prediction, or fitted parameter that reduces to the paper's own inputs by construction. No equations, uniqueness theorems, or self-citation chains appear in the load-bearing claims; the work is self-contained against external reproduction and does not rename known results or smuggle ansatzes. This matches the default expectation for non-derivational benchmark papers.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The benchmark rests on the domain assumption that the selected 10 flow scenarios adequately represent the space of fluid dynamics problems and that modular decomposition isolates the effects of each component without hidden interactions.

axioms (2)
  • domain assumption The ten chosen flow scenarios are representative of the broader class of fluid dynamics problems.
    Invoked when claiming the leaderboard is comprehensive.
  • domain assumption Modular separation of spatial, temporal, and loss modules allows fair attribution of performance differences.
    Central to the first listed contribution.

pith-pipeline@v0.9.0 · 5771 in / 1225 out tokens · 27117 ms · 2026-05-22T12:55:41.976868+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

140 extracted references · 140 canonical work pages · 2 internal anchors

  1. [1]

    Universal physics transformers: A framework for efficiently scaling neural operators

    Benedikt Alkin, Andreas Fürst, Simon Lucas Schmid, Lukas Gruber, Markus Holzleitner, and Johannes Brandstetter. Universal physics transformers: A framework for efficiently scaling neural operators. InThe Thirty-eighth Annual Conference on Neural Information Processing Systems, 2024

    work page 2024

  2. [2]

    Physics-informed diffusion models

    Jan-Hendrik Bastek, WaiChing Sun, and Dennis Kochmann. Physics-informed diffusion models. InThe Thirteenth International Conference on Learning Representations, 2025

    work page 2025

  3. [3]

    Accurate medium-range global weather forecasting with 3d neural networks.Nature, 619(7970):533– 538, 2023

    Kaifeng Bi, Lingxi Xie, Hengheng Zhang, Xin Chen, Xiaotao Gu, and Qi Tian. Accurate medium-range global weather forecasting with 3d neural networks.Nature, 619(7970):533– 538, 2023

    work page 2023

  4. [4]

    Enhancing the inductive biases of graph neural ODE for modeling physical systems

    Suresh Bishnoi, Ravinder Bhattoo, Jayadeva Jayadeva, Sayan Ranu, and N M Anoop Krishnan. Enhancing the inductive biases of graph neural ODE for modeling physical systems. InThe Eleventh International Conference on Learning Representations, 2023

    work page 2023

  5. [5]

    Magnet: Mesh agnostic neural pde solver.Advances in Neural Information Processing Systems, 35:31972– 31985, 2022

    Oussama Boussif, Yoshua Bengio, Loubna Benabbou, and Dan Assouline. Magnet: Mesh agnostic neural pde solver.Advances in Neural Information Processing Systems, 35:31972– 31985, 2022

    work page 2022

  6. [6]

    Envisioning better benchmarks for machine learning pde solvers

    Johannes Brandstetter. Envisioning better benchmarks for machine learning pde solvers. Nature Machine Intelligence, 7(1):2–3, 2025

    work page 2025

  7. [7]

    Clifford neural layers for PDE modeling

    Johannes Brandstetter, Rianne van den Berg, et al. Clifford neural layers for PDE modeling. InThe Eleventh International Conference on Learning Representations, 2023

    work page 2023

  8. [8]

    Worrall, and Max Welling

    Johannes Brandstetter, Daniel E. Worrall, and Max Welling. Message passing neural PDE solvers. InInternational Conference on Learning Representations, 2022

    work page 2022

  9. [9]

    HAMLET: Graph transformer neural operator for partial differential equations

    Andrey Bryutkin, Jiahao Huang, Zhongying Deng, Guang Yang, Carola-Bibiane Schönlieb, and Angelica I Aviles-Rivero. HAMLET: Graph transformer neural operator for partial differential equations. InInternational Conference on Machine Learning, 2024

    work page 2024

  10. [10]

    Yousuff Hussaini, Alfio Quarteroni, and Thomas A

    Claudio Canuto, M. Yousuff Hussaini, Alfio Quarteroni, and Thomas A. Zang.Spectral Methods in Fluid Dynamics. Springer-Verlag, Berlin, Heidelberg, 1988

    work page 1988

  11. [11]

    Lno: Laplace neural operator for solving differential equations.Arxiv, 2023

    Qianying Cao, Somdatta Goswami, and George Em Karniadakis. Lno: Laplace neural operator for solving differential equations.Arxiv, 2023

    work page 2023

  12. [12]

    Choose a transformer: Fourier or galerkin

    Shuhao Cao. Choose a transformer: Fourier or galerkin. In A. Beygelzimer, Y . Dauphin, P. Liang, and J. Wortman Vaughan, editors,Advances in Neural Information Processing Systems, 2021

    work page 2021

  13. [13]

    Spectral-refiner: Accurate fine-tuning of spatiotemporal fourier neural operator for turbulent flows

    Shuhao Cao, Francesco Brarda, Ruipeng Li, and Yuanzhe Xi. Spectral-refiner: Accurate fine-tuning of spatiotemporal fourier neural operator for turbulent flows. InThe Thirteenth International Conference on Learning Representations, 2025

    work page 2025

  14. [14]

    Efficient learning of mesh- based physical simulation with bi-stride multi-scale graph neural network

    Yadi Cao, Menglei Chai, Minchen Li, and Chenfanfu Jiang. Efficient learning of mesh- based physical simulation with bi-stride multi-scale graph neural network. InInternational Conference on Machine Learning, pages 3541–3558. PMLR, 2023

    work page 2023

  15. [15]

    A liquid plug moving in an annular pipe–heat transfer analysis

    Yadi Cao, Xuan Gao, and Ri Li. A liquid plug moving in an annular pipe–heat transfer analysis. International Journal of Heat and Mass Transfer, 139:1065–1076, 2019

    work page 2019

  16. [16]

    A liquid plug moving in an annular pipe—flow analysis.Physics of Fluids, 30(9), 2018

    Yadi Cao and Ri Li. A liquid plug moving in an annular pipe—flow analysis.Physics of Fluids, 30(9), 2018

    work page 2018

  17. [17]

    Implicit neural spatial representations for time-dependent pdes

    Honglin Chen, Rundi Wu, Eitan Grinspun, Changxi Zheng, and Peter Yichen Chen. Implicit neural spatial representations for time-dependent pdes. InInternational Conference on Machine Learning, pages 5162–5177. PMLR, 2023

    work page 2023

  18. [18]

    Neural ordinary differential equations.Advances in neural information processing systems, 31, 2018

    Ricky TQ Chen, Yulia Rubanova, Jesse Bettencourt, and David K Duvenaud. Neural ordinary differential equations.Advances in neural information processing systems, 31, 2018. 10

    work page 2018

  19. [19]

    Physics-informed learning of governing equations from scarce data.Nature communications, 12(1):6136, 2021

    Zhao Chen, Yang Liu, and Hao Sun. Physics-informed learning of governing equations from scarce data.Nature communications, 12(1):6136, 2021

    work page 2021

  20. [20]

    Fourier neural operator for fluid flow in small-shape 2d simulated porous media dataset.Algorithms, 16(1):24, 2023

    Abouzar Choubineh, Jie Chen, David A Wood, Frans Coenen, and Fei Ma. Fourier neural operator for fluid flow in small-shape 2d simulated porous media dataset.Algorithms, 16(1):24, 2023

    work page 2023

  21. [21]

    Prediction of transonic flow over supercritical airfoils using geometric-encoding and deep-learning strategies.Physics of Fluids, 35(7), 2023

    Zhiwen Deng, Jing Wang, Hongsheng Liu, Hairun Xie, BoKai Li, Miao Zhang, Tingmeng Jia, Yi Zhang, Zidong Wang, and Bin Dong. Prediction of transonic flow over supercritical airfoils using geometric-encoding and deep-learning strategies.Physics of Fluids, 35(7), 2023

    work page 2023

  22. [22]

    Multiscale meshgraphnets.Arxiv, 2022

    Meire Fortunato, Tobias Pfaff, Peter Wirnsberger, Alexander Pritzel, and Peter Battaglia. Multiscale meshgraphnets.Arxiv, 2022

    work page 2022

  23. [23]

    Transformers for modeling physical systems.Neural Networks, 146:272–289, 2022

    Nicholas Geneva and Nicholas Zabaras. Transformers for modeling physical systems.Neural Networks, 146:272–289, 2022

    work page 2022

  24. [24]

    Learning physics informed neural odes with partial measurements

    Paul Ghanem, Ahmet Demirkaya, Tales Imbiriba, Alireza Ramezani, Zachary Danziger, and Deniz Erdogmus. Learning physics informed neural odes with partial measurements. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 16799–16807, 2025

    work page 2025

  25. [25]

    Gingold and J.J

    R.A. Gingold and J.J. Monaghan. Smoothed particle hydrodynamics: theory and application to non-spherical stars.Monthly Notices of the Royal Astronomical Society, 181(3):375–389, 1977

    work page 1977

  26. [26]

    Efficient token mixing for transformers via adaptive fourier neural operators

    John Guibas, Morteza Mardani, Zongyi Li, Andrew Tao, Anima Anandkumar, and Bryan Catanzaro. Efficient token mixing for transformers via adaptive fourier neural operators. In International conference on learning representations, 2021

    work page 2021

  27. [27]

    Multiwavelet-based operator learning for differential equations

    Gaurav Gupta, Xiongye Xiao, and Paul Bogdan. Multiwavelet-based operator learning for differential equations. In A. Beygelzimer, Y . Dauphin, P. Liang, and J. Wortman Vaughan, editors,Advances in Neural Information Processing Systems, 2021

    work page 2021

  28. [28]

    Brainode: Dynamic brain signal analysis via graph-aided neural ordinary differential equations

    Kaiqiao Han, Yi Yang, Zijie Huang, Xuan Kan, Ying Guo, Yang Yang, Lifang He, Liang Zhan, Yizhou Sun, Wei Wang, et al. Brainode: Dynamic brain signal analysis via graph-aided neural ordinary differential equations. In2024 IEEE EMBS International Conference on Biomedical and Health Informatics (BHI), pages 1–8. IEEE, 2024

    work page 2024

  29. [29]

    Predicting physics in mesh-reduced space with temporal attention

    XU HAN, Han Gao, Tobias Pfaff, Jian-Xun Wang, and Liping Liu. Predicting physics in mesh-reduced space with temporal attention. InInternational Conference on Learning Representations, 2022

    work page 2022

  30. [30]

    Dpot: Auto-regressive denoising operator transformer for large-scale pde pre-training.ICML, 2024

    Zhongkai Hao, Chang Su, Songming Liu, Julius Berner, Chengyang Ying, Hang Su, Anima Anandkumar, Jian Song, and Jun Zhu. Dpot: Auto-regressive denoising operator transformer for large-scale pde pre-training.ICML, 2024

    work page 2024

  31. [31]

    Gnot: A general neural operator transformer for operator learning

    Zhongkai Hao, Zhengyi Wang, Hang Su, Chengyang Ying, Yinpeng Dong, Songming Liu, Ze Cheng, Jian Song, and Jun Zhu. Gnot: A general neural operator transformer for operator learning. InInternational Conference on Machine Learning, pages 12556–12569. PMLR, 2023

    work page 2023

  32. [32]

    Pinnacle: A comprehensive benchmark of physics- informed neural networks for solving pdes.Arxiv, 2023

    Zhongkai Hao, Jiachen Yao, Chang Su, Hang Su, Ziao Wang, Fanzhi Lu, Zeyu Xia, Yichi Zhang, Songming Liu, Lu Lu, et al. Pinnacle: A comprehensive benchmark of physics- informed neural networks for solving pdes.Arxiv, 2023

    work page 2023

  33. [33]

    Flow completion network: Inferring the fluid dynamics from incomplete flow information using graph neural networks.Physics of Fluids, 34(8), 2022

    Xiaodong He, Yinan Wang, and Juan Li. Flow completion network: Inferring the fluid dynamics from incomplete flow information using graph neural networks.Physics of Fluids, 34(8), 2022

    work page 2022

  34. [34]

    Group equivariant fourier neural operators for partial differential equations.ICML, 2023

    Jacob Helwig, Xuan Zhang, Cong Fu, Jerry Kurtin, Stephan Wojtowytsch, and Shuiwang Ji. Group equivariant fourier neural operators for partial differential equations.ICML, 2023. 11

    work page 2023

  35. [35]

    Poseidon: Efficient foundation models for pdes

    Maximilian Herde, Bogdan Raonic, Tobias Rohner, Roger Käppeli, Roberto Molinaro, Em- manuel de Bézenac, and Siddhartha Mishra. Poseidon: Efficient foundation models for pdes. Advances in Neural Information Processing Systems, 37:72525–72624, 2024

    work page 2024

  36. [36]

    Denoising diffusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020

    work page 2020

  37. [37]

    DiffusionPDE: Generative PDE-solving under partial observation

    Jiahe Huang, Guandao Yang, Zichen Wang, and Jeong Joon Park. DiffusionPDE: Generative PDE-solving under partial observation. InThe Thirty-eighth Annual Conference on Neural Information Processing Systems, 2024

    work page 2024

  38. [38]

    Meta-auto-decoder for solving parametric partial differential equations.Advances in Neural Information Processing Systems, 35:23426–23438, 2022

    Xiang Huang, Zhanhong Ye, Hongsheng Liu, Shi Ji, Zidong Wang, Kang Yang, Yang Li, Min Wang, Haotian Chu, Fan Yu, et al. Meta-auto-decoder for solving parametric partial differential equations.Advances in Neural Information Processing Systems, 35:23426–23438, 2022

    work page 2022

  39. [39]

    Learning continuous system dynamics from irregularly-sampled partial observations

    Zijie Huang, Yizhou Sun, and Wei Wang. Learning continuous system dynamics from irregularly-sampled partial observations. InAdvances in Neural Information Processing Systems, 2020

    work page 2020

  40. [40]

    Physics-informed regularization for domain-agnostic dynamical system modeling, 2024

    Zijie Huang, Wanjia Zhao, Jingdong Gao, Ziniu Hu, Xiao Luo, Yadi Cao, Yuanzhou Chen, Yizhou Sun, and Wei Wang. Physics-informed regularization for domain-agnostic dynamical system modeling, 2024

    work page 2024

  41. [41]

    Pac-fno: Parallel-structured all-component fourier neural operators for recognizing low-quality images.Arxiv, 2024

    Jinsung Jeon, Hyundong Jin, Jonghyun Choi, Sanghyun Hong, Dongeun Lee, Kookjin Lee, and Noseong Park. Pac-fno: Parallel-structured all-component fourier neural operators for recognizing low-quality images.Arxiv, 2024

    work page 2024

  42. [42]

    Zhongyi Jiang, Min Zhu, and Lu Lu. Fourier-mionet: Fourier-enhanced multiple-input neural operators for multiphase modeling of geological carbon sequestration.Reliability Engineering & System Safety, 251:110392, 2024

    work page 2024

  43. [43]

    Mionet: Learning multiple-input operators via tensor product.SIAM Journal on Scientific Computing, 44(6):A3490–A3514, 2022

    Pengzhan Jin, Shuai Meng, and Lu Lu. Mionet: Learning multiple-input operators via tensor product.SIAM Journal on Scientific Computing, 44(6):A3490–A3514, 2022

    work page 2022

  44. [44]

    Nsfnets (navier-stokes flow nets): Physics-informed neural networks for the incompressible navier-stokes equations

    Xiaowei Jin, Shengze Cai, Hui Li, and George Em Karniadakis. Nsfnets (navier-stokes flow nets): Physics-informed neural networks for the incompressible navier-stokes equations. Journal of Computational Physics, 426:109951, 2021

    work page 2021

  45. [45]

    Klaasen and W.C

    G.A. Klaasen and W.C. Troy. Stationary wave solutions of a system of reaction-diffusion equations derived from the fitzhugh–nagumo equations.SIAM Journal on Applied Mathematics, 44(1):96–110, 1984

    work page 1984

  46. [46]

    Benchmarking autoregressive conditional diffusion models for turbulent flow simulation.arXiv preprint arXiv:2309.01745, 2023

    Georg Kohl, Li-Wei Chen, and Nils Thuerey. Benchmarking autoregressive conditional diffusion models for turbulent flow simulation.arXiv preprint arXiv:2309.01745, 2023

    work page arXiv 2023

  47. [47]

    Learning in latent spaces improves the predictive accuracy of deep neural operators.Arxiv, 2023

    Katiana Kontolati, Somdatta Goswami, George Em Karniadakis, and Michael D Shields. Learning in latent spaces improves the predictive accuracy of deep neural operators.Arxiv, 2023

    work page 2023

  48. [48]

    Learning skillful medium-range global weather forecasting.Science, 382(6677):1416–1421, 2023

    Remi Lam, Alvaro Sanchez-Gonzalez, et al. Learning skillful medium-range global weather forecasting.Science, 382(6677):1416–1421, 2023

    work page 2023

  49. [49]

    HyperdeepONet: learning operator with complex target function space using the limited resources via hypernetwork

    Jae Yong Lee, SungWoong CHO, and Hyung Ju Hwang. HyperdeepONet: learning operator with complex target function space using the limited resources via hypernetwork. InThe Eleventh International Conference on Learning Representations, 2023

    work page 2023

  50. [50]

    Data-driven prediction of unsteady flow over a circular cylinder using deep learning.Journal of Fluid Mechanics, 879:217–254, 2019

    Sangseung Lee and Donghyun You. Data-driven prediction of unsteady flow over a circular cylinder using deep learning.Journal of Fluid Mechanics, 879:217–254, 2019

    work page 2019

  51. [51]

    Two-stage fourth order: temporal-spatial coupling in computational fluid dynamics (cfd).Advances in Aerodynamics, 1:1–36, 2019

    Jiequan Li. Two-stage fourth order: temporal-spatial coupling in computational fluid dynamics (cfd).Advances in Aerodynamics, 1:1–36, 2019. 12

    work page 2019

  52. [52]

    Synthetic lagrangian turbulence by generative diffusion models.Nature Machine Intelligence, pages 1–11, 2024

    Tianyi Li, Luca Biferale, Fabio Bonaccorso, Martino Andrea Scarpolini, and Michele Buzzi- cotti. Synthetic lagrangian turbulence by generative diffusion models.Nature Machine Intelligence, pages 1–11, 2024

    work page 2024

  53. [53]

    Graph neural network-accelerated lagrangian fluid simula- tion.Computers & Graphics, 103:201–211, 2022

    Zijie Li and Amir Barati Farimani. Graph neural network-accelerated lagrangian fluid simula- tion.Computers & Graphics, 103:201–211, 2022

    work page 2022

  54. [54]

    Scalable transformer for pde surrogate modeling

    Zijie Li, Dule Shu, and Amir Barati Farimani. Scalable transformer for pde surrogate modeling. Advances in Neural Information Processing Systems, 36, 2024

    work page 2024

  55. [55]

    Fourier neural operator with learned deformations for pdes on general geometries.Journal of Machine Learning Research, 24(388):1–26, 2023

    Zongyi Li, Daniel Zhengyu Huang, Burigede Liu, and Anima Anandkumar. Fourier neural operator with learned deformations for pdes on general geometries.Journal of Machine Learning Research, 24(388):1–26, 2023

    work page 2023

  56. [56]

    Multipole graph neural operator for parametric partial differential equations.Advances in Neural Information Processing Systems, 33:6755–6766, 2020

    Zongyi Li, Nikola Kovachki, Kamyar Azizzadenesheli, Burigede Liu, Andrew Stuart, Kaushik Bhattacharya, and Anima Anandkumar. Multipole graph neural operator for parametric partial differential equations.Advances in Neural Information Processing Systems, 33:6755–6766, 2020

    work page 2020

  57. [57]

    Fourier neural operator for parametric partial differential equations

    Zongyi Li, Nikola Borislavov Kovachki, Kamyar Azizzadenesheli, Kaushik Bhattacharya, Andrew Stuart, Anima Anandkumar, et al. Fourier neural operator for parametric partial differential equations. InInternational Conference on Learning Representations

    work page

  58. [58]

    Fourier neural operator for parametric partial differential equations

    Zongyi Li, Nikola Borislavov Kovachki, et al. Fourier neural operator for parametric partial differential equations. InInternational Conference on Learning Representations, 2021

    work page 2021

  59. [59]

    Geometry-informed neural operator for large- scale 3d PDEs

    Zongyi Li, Nikola Borislavov Kovachki, et al. Geometry-informed neural operator for large- scale 3d PDEs. InThirty-seventh Conference on Neural Information Processing Systems, 2023

    work page 2023

  60. [60]

    Physics-informed neural operator for learning partial differential equations.Arxiv, 2021

    Zongyi Li, Hongkai Zheng, Nikola Kovachki, David Jin, Haoxuan Chen, Burigede Liu, Kamyar Azizzadenesheli, and Anima Anandkumar. Physics-informed neural operator for learning partial differential equations.Arxiv, 2021

    work page 2021

  61. [61]

    B-deeponet: An enhanced bayesian deep- onet for solving noisy parametric pdes using accelerated replica exchange sgld.Journal of Computational Physics, 473:111713, 2023

    Guang Lin, Christian Moya, and Zecheng Zhang. B-deeponet: An enhanced bayesian deep- onet for solving noisy parametric pdes using accelerated replica exchange sgld.Journal of Computational Physics, 473:111713, 2023

    work page 2023

  62. [62]

    Multi-scale rotation- equivariant graph neural networks for unsteady eulerian fluid dynamics.Physics of Fluids, 34(8), 2022

    Mario Lino, Stathi Fotiadis, Anil A Bharath, and Chris D Cantwell. Multi-scale rotation- equivariant graph neural networks for unsteady eulerian fluid dynamics.Physics of Fluids, 34(8), 2022

    work page 2022

  63. [63]

    Flow Matching for Generative Modeling

    Yaron Lipman, Ricky TQ Chen, Heli Ben-Hamu, Maximilian Nickel, and Matt Le. Flow matching for generative modeling.arXiv preprint arXiv:2210.02747, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  64. [64]

    Fast fluid simulation via dynamic multi-scale gridding

    Jinxian Liu, Ye Chen, Bingbing Ni, Wei Ren, Zhenbo Yu, and Xiaoyang Huang. Fast fluid simulation via dynamic multi-scale gridding. InProceedings of the AAAI Conference on Artificial Intelligence, volume 37, pages 1675–1682, 2023

    work page 2023

  65. [65]

    Domain agnostic fourier neural operators

    Ning Liu, Siavash Jafarzadeh, and Yue Yu. Domain agnostic fourier neural operators. In Thirty-seventh Conference on Neural Information Processing Systems, 2023

    work page 2023

  66. [66]

    Learning nonlinear operators via deeponet based on the universal approximation theorem of operators

    Lu Lu, Pengzhan Jin, Guofei Pang, Zhongqiang Zhang, and George Em Karniadakis. Learning nonlinear operators via deeponet based on the universal approximation theorem of operators. Nature machine intelligence, 3(3):218–229, 2021

    work page 2021

  67. [67]

    Pgode: Towards high-quality system dynamics modeling, 2024

    Xiao Luo, Yiyang Gu, Huiyu Jiang, Hang Zhou, Jinsheng Huang, Wei Ju, Zhiping Xiao, Ming Zhang, and Yizhou Sun. Pgode: Towards high-quality system dynamics modeling, 2024

    work page 2024

  68. [68]

    Care: Modeling interacting dynamics under temporal environmental variation

    Xiao Luo, Haixin Wang, Zijie Huang, Huiyu Jiang, Abhijeet Sadashiv Gangan, Song Jiang, and Yizhou Sun. Care: Modeling interacting dynamics under temporal environmental variation. InThirty-seventh Conference on Neural Information Processing Systems, 2023. 13

    work page 2023

  69. [69]

    HOPE: High-order graph ODE for modeling interacting dynamics

    Xiao Luo, Jingyang Yuan, Zijie Huang, Huiyu Jiang, Yifang Qin, Wei Ju, Ming Zhang, and Yizhou Sun. HOPE: High-order graph ODE for modeling interacting dynamics. In Andreas Krause, Emma Brunskill, Kyunghyun Cho, Barbara Engelhardt, Sivan Sabato, and Jonathan Scarlett, editors,Proceedings of the 40th International Conference on Machine Learning, volume 202 ...

    work page 2023

  70. [70]

    Cfdbench: A comprehensive benchmark for machine learning methods in fluid dynamics.Arxiv, 2023

    Yining Luo, Yingfa Chen, and Zhen Zhang. Cfdbench: A comprehensive benchmark for machine learning methods in fluid dynamics.Arxiv, 2023

    work page 2023

  71. [71]

    V orticity and incompressible flow

    Andrew J Majda, Andrea L Bertozzi, and A Ogawa. V orticity and incompressible flow. cambridge texts in applied mathematics.Appl. Mech. Rev., 55(4):B77–B78, 2002

    work page 2002

  72. [72]

    Weak baselines and reporting biases lead to overoptimism in machine learning for fluid-related partial differential equations.Nature Machine Intelligence, 6(10):1256–1269, 2024

    Nick McGreivy and Ammar Hakim. Weak baselines and reporting biases lead to overoptimism in machine learning for fluid-related partial differential equations.Nature Machine Intelligence, 6(10):1256–1269, 2024

    work page 2024

  73. [73]

    Mikulevicius and B

    R. Mikulevicius and B. L. Rozovskii. Stochastic navier–stokes equations for turbulent flows. SIAM Journal on Mathematical Analysis, 35(5):1250–1310, 2004

    work page 2004

  74. [74]

    Moukalled, L

    F. Moukalled, L. Mangani, and M. Darwish.The Finite Volume Method in Computational Fluid Dynamics. Springer, 1 edition, 2016

    work page 2016

  75. [75]

    Shock wave prediction in transonic flow fields using domain-informed probabilistic deep learning.Physics of Fluids, 36(1), 2024

    Bilal Mufti, Anindya Bhaduri, Sayan Ghosh, Liping Wang, and Dimitri N Mavris. Shock wave prediction in transonic flow fields using domain-informed probabilistic deep learning.Physics of Fluids, 36(1), 2024

    work page 2024

  76. [76]

    Particle-based fluid simulation for inter- active applications

    Matthias Müller, David Charypar, and Markus Gross. Particle-based fluid simulation for inter- active applications. InProceedings of the 2003 ACM SIGGRAPH/Eurographics symposium on Computer animation, pages 154–159. Citeseer, 2003

    work page 2003

  77. [77]

    The well: a large-scale collection of diverse physics simulations for machine learning.Advances in Neural Information Processing Systems, 37:44989–45037, 2024

    Ruben Ohana, Michael McCabe, Lucas Meyer, Rudy Morel, Fruzsina Agocs, Miguel Beneitez, Marsha Berger, Blakesly Burkhart, Stuart Dalziel, Drummond Fielding, et al. The well: a large-scale collection of diverse physics simulations for machine learning.Advances in Neural Information Processing Systems, 37:44989–45037, 2024

    work page 2024

  78. [78]

    Neural-fly enables rapid learning for agile flight in strong winds.Science Robotics, 7(66), 2022

    Michael O’Connell, Guanya Shi, Xichen Shi, et al. Neural-fly enables rapid learning for agile flight in strong winds.Science Robotics, 7(66), 2022

    work page 2022

  79. [79]

    Fourcastnet: A global data-driven high- resolution weather model using adaptive fourier neural operators.Arxiv, 2022

    Jaideep Pathak, Shashank Subramanian, et al. Fourcastnet: A global data-driven high- resolution weather model using adaptive fourier neural operators.Arxiv, 2022

    work page 2022

  80. [80]

    Learning mesh-based simulation with graph networks

    Tobias Pfaff, Meire Fortunato, Alvaro Sanchez-Gonzalez, and Peter Battaglia. Learning mesh-based simulation with graph networks. InInternational Conference on Learning Repre- sentations, 2021

    work page 2021

Showing first 80 references.