Pith. sign in

REVIEW 3 major objections 6 minor 143 references

Hybrid Lagrangian-Eulerian Model for Lagrangian Fluid Simulation

T0 review · 3 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Hybrid grid anchor cuts drift in Lagrangian fluid simulators

desk verdict Solid hybrid architecture with consistent benchmark gains, but the 'stable anchor' story outruns the evidence; send to review but ask for uncertainty quantification and a long-rollout cross-attention ablation. read the letter →

arxiv 2608.01164 v1 pith:2T7TUFEI submitted 2026-08-02 cs.CE cs.AI

Ruoyan Li , Wei Wang , Yizhou Sun This is my paper
classification cs.CEcs.AI
keywords hybridLagrangian-Eulerianneuralfluidsimulationcross-attentionEuleriangridanchoradaptivedownsamplingrolloutstabilityparticle-basederroraccumulation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to show that the two known weaknesses of pure Lagrangian neural fluid simulators—wasted capacity in nearly uniform regions and drift over long rollouts—can both be fixed by adding a fixed Eulerian grid as a spatial anchor. The proposed hybrid model downsamples kinematically redundant particles, aggregates their features onto Eulerian nodes, and uses cross-attention so each Lagrangian update queries the grid as a stable reference. Across 2D and 3D benchmarks, including high-resolution, fluid-solid, and deforming domains, the model reports lower trajectory error, lower kinetic-energy error, and substantially lower Sinkhorn divergence than the baselines, and it holds up over 200-step rollouts. A sympathetic reader would take the paper as evidence that particle-grid hybrids are a viable path to stable, accurate neural surrogates for Lagrangian fluids.

What carries the argument

Cross-attention to a fixed Eulerian graph, implemented inside an Encoder–Downsampler–Processor–Upsampler–Decoder. Adaptive downsampling (SAGPool-style self-attention pooling) condenses redundant particles and Eulerian nodes; a learnable aggregation kernel $\mathbf{W}$ projects downsampled Lagrangian features onto Eulerian nodes ($\mathbf{Z}_{\mathrm{lag}} = \mathbf{W}^\top \mathbf{q}_{\mathrm{lag}}$); multi-head self-attention on $\mathbf{Z}_{\mathrm{lag}}$ plays the role of a coarse field solver; then multi-head cross-attention uses the Eulerian features as keys and values (with $\mathbf{Z}_{\mathrm{lag}}$ as queries) to inject stable field information; the combined feature $\mathbf{F} = \m

What would settle it

Take the trained model and replace the Eulerian keys/values in cross-attention with constant or randomly initialized fixed grid embeddings while keeping everything else identical, then measure MSE and Sinkhorn divergence over a 200-step rollout. If error does not grow substantially, the fixed-grid-anchor explanation is wrong. Alternatively, hold cross-attention fixed and replace the downsampler with random particle dropping; if drift remains low, the stability gains come from the regularization of downsampling rather than re-grounding.

Watch

Extended reading notes

Core claim

The central discovery is that a hybrid Lagrangian–Eulerian neural architecture can suppress the error accumulation that plagues pure particle-based surrogates. The model builds a fixed Eulerian graph with far fewer nodes than particles, aggregates downsampled particle features onto it through a learnable kernel, applies multi-head self-attention as a coarse 'field solver,' and then lets Lagrangian tokens query the Eulerian features via cross-attention before scattering back to particles. Because the Eulerian nodes are tied to fixed spatial locations, the paper argues, they re-ground the Lagrangian update at every timestep and correct small trajectory deviations before they compound. Empirica

Load-bearing premise

The load-bearing premise is that a fixed Eulerian graph is a stable spatial reference and that cross-attention to it—not the downsampling, the larger receptive field, or some other architectural choice—is what suppresses error accumulation; the paper demonstrates the combination works but never isolates the anchoring mechanism.

Editorial extensions

If this is right

  • Long autoregressive rollouts of Lagrangian neural simulators become feasible without auxiliary corrective losses; the model stays accurate at 200 rollout steps on water, sand, goop, and multi-material data.
  • Neural surrogates can scale to tens of thousands of particles because the downsampler removes kinematic redundancy before the processor runs.
  • Particle-grid transfer no longer requires a rectangular voxel grid: because the Eulerian branch is a general mesh graph, irregular and deforming domains are handled, where voxelization-based approaches fail.
  • Benchmarked against both Lagrangian and Eulerian mesh models, the architecture reports the strongest numbers on the considered 2D and 3D fluid benchmarks, especially in Sinkhorn divergence.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • This paper's stability argument would be stronger if cross-attention to fixed-position Eulerian features were shown to beat a control that receives the same global information without fixed coordinates (e.g., a learned global token); such an ablation would isolate 'anchoredness' from mere long-range mixing. The paper does not provide that control.
  • The learned aggregation kernel $\mathbf{W}$ may itself be learning a discrete transfer operator reminiscent of classical particle-in-cell interpolation; a natural extension is to initialize or regularize it with classical SPH kernels and test whether physics-based priors reduce the data needed.
  • If the anchor mechanism generalizes, a promising next step is an adaptive Eulerian mesh that deforms or refines with the flow, preserving the anchor's benefits while avoiding fixed-grid resolution loss.
  • The paper itself concedes that over extreme temporal horizons trajectories still diverge; this means the anchor suppresses but does not eliminate drift, so the method's promise depends on how far the horizon can be pushed before unphysical distributions appear.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper proposes a hybrid Lagrangian-Eulerian graph-network simulator for weakly-compressible Lagrangian fluid flows. It augments a Lagrangian encoder-processor-decoder with adaptive SAGPool-based downsampling and a cross-attention mechanism that queries features aggregated onto Eulerian grid nodes. The authors argue this addresses a 'spatial bottleneck' and 'temporal drift' in pure-Lagrangian models, and they report consistent improvements over several baselines on 2D/3D LagrangeBench tasks, higher-resolution TGV, fluid-solid interaction, deformable domains, extended water rollouts, and additional materials. The central claim is a new state-of-the-art in accuracy and rollout stability, with the stability gain attributed to the Eulerian 'spatial anchor'.

Significance. If the empirical claims are reproducible, the proposed architecture is a useful contribution: it provides broad benchmark coverage (Tables 1-8), consistently outperforms a wide set of baselines, and contains useful ablations showing the contribution of downsampling, cross-attention, and Eulerian anchoring points on LDC. However, because the stability mechanism is not isolated and the results lack uncertainty quantification, the manuscript currently overstates its causal explanation. The work is of interest to the neural-PDE/CFD community but needs additional evidence before the SOTA claim can be accepted.

major comments (3)
  1. [§4 (Downsampler and Processor), Eq. (3), Eq. (7)] The 'stable spatial anchor' premise is not supported by the architecture as described. The Eulerian velocities in Eq. (3) are computed by kernel-averaging the current particle positions/velocities, which during autoregressive rollout are the model's own outputs. Moreover, the set of Eulerian nodes actually used in cross-attention is not the fixed grid: the downsampler applies SAGPooling to the Eulerian graph, so q_eu in Eq. (7) is a data-dependent subset of grid nodes selected from features that depend on the current (predicted) state. Consequently, the anchor coordinates can change from step to step, and any Lagrangian error enters the Eulerian features directly. The paper provides no error-contraction or perturbation analysis showing that Eq. (3)–(7) re-grounds the trajectory. This is a load-bearing issue for the abstract's causal claim that the cross-attended design 'substantially sup
  2. [§5.5, Table 6; Appendix E, Fig. 4] The extended-rollout evidence does not isolate the cross-attention mechanism. Table 6 compares the full model only against GraphUnet; the cross-attention ablation in Appendix E is on Lid-Driven Cavity at 5 steps, not in the extended regime where the stability claim is made. Thus the large gains at 200 steps could come from the downsampler, extra capacity, or the auxiliary training used in §5.5. The authors should report an extended-rollout ablation that removes cross-attention (or replaces it with a fixed-grid aggregation without attention) and, ideally, measure how errors in the Eulerian features evolve under perturbation.
  3. [Tables 1–8; Appendix E] All numerical comparisons are point estimates from a single run; no error bars, standard deviations, or repeated-seed statistics are reported. The hyperparameters (grid size s=32, λ=0.8, M=2, L=4) are selected on Lid-Driven Cavity and applied to all other datasets without a per-dataset sensitivity check. Given the claim of 'consistent' improvements and 'state-of-the-art' performance, the absence of uncertainty quantification makes it impossible to assess whether the relative gains (e.g., 11–57%) are statistically meaningful.
minor comments (6)
  1. [Appendix C] Numerical typos in the dataset descriptions, e.g., '40×10^{-3}55' and '100×10^{-3}5' should be corrected.
  2. [Appendix E] The paragraph 'Ablation Study on Aggregating Lagrangian Nodes to Eulerian Nodes' contains a duplicated sentence: 'The full model... achieves stronger performance across all metrics.' appears twice.
  3. [§4 Downsampler] Typo: 'downsmpling rate' should be 'downsampling rate'.
  4. [§5.5] The statement that auxiliary training strategies are applied in the extended-rollout experiment should be reconciled with §5's statement that such strategies are excluded in the main results; as written, a reader may think the main results also use them.
  5. [§4 Remark] The claim that the Eulerian representation is 'defined on a general mesh graph rather than a fixed grid' sits in tension with the repeated 'fixed grid' terminology in the abstract and §4; clarify whether the anchor is a fixed grid or a mesh that can be chosen per domain.
  6. [Appendix F] The comparison with Eulerian mesh simulators is useful, but the table does not state which of the listed baselines already use aggregation/downsampling; the discussion would benefit from a column indicating architecture family.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: central SOTA claim rests on external empirical benchmarks; self-citations are minor and non-load-bearing. The 'stable Eulerian anchor' explanation is mechanistically under-supported but that is an ablation/confound issue, not a circular derivation.

full rationale

The paper's central claim—that the hybrid Lagrangian–Eulerian model achieves a new state-of-the-art for accuracy and rollout stability—is established by empirical comparisons against external baselines (GNS, SEGNN, GraphUnet, PhysicsNFP, NeuralMPM, etc.) on LagrangeBench and additional datasets (Tables 1–6, Appendices F–G). No step of that comparison reduces to the model's own assumptions by construction. Hyperparameters such as grid size 32, downsampling ratio 0.8, and number of layers are tuned on LDC and transferred to other benchmarks; this is a mild form of architecture selection but not a fitted-input-called-prediction, because the reported numbers on held-out datasets are genuine out-of-sample evaluations. The self-citations (FD-Bench for dataset setup, DTA-GNN as a baseline) are not load-bearing: the empirical conclusions stand without them. The only candidate for a circular step is the Processor's mechanistic claim that 'the Eulerian branch acts as a spatial anchor' and that 'cross-attention re-grounds the Lagrangian update at every step against this stable, field-based reference.' However, Eq. (3) defines Eulerian velocities as kernel averages of the same predicted particle velocities, so the Eulerian field is not an independent reference during autoregressive rollout; it is a smoothed function of the model's own outputs. This undermines the causal explanation but does not make any reported result equivalent to an input by construction: cross-attention remains a learnable transformation, and the accuracy/stability numbers are measured against external ground truth. The lack of a long-rollout cross-attention ablation (Table 6 compares the full model only against GraphUnet) is an experimental isolation problem, not circular reasoning. Therefore no step exhibits Eq. X = Eq. Y by construction or a fitted parameter renamed as a prediction, and the circularity score is low.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The paper does not introduce new physical entities. The Eulerian nodes and the learnable aggregation kernel are computational constructs rather than postulated physical objects. The central load-bearing assumptions are the domain assumptions about the benchmark data and the ad hoc architectural assumptions about downsampling and spatial anchoring.

free parameters (5)
  • Eulerian grid resolution s = 32 nodes per dimension
    Chosen via ablation on LDC (Appendix E, Figure 6); applied to all other datasets.
  • Downsampling ratio lambda = 0.8
    Tuned on LDC (Appendix E, Figure 7); shared for both Lagrangian and Eulerian graphs.
  • Number of downsampling layers M = 2
    Tuned on LDC (Appendix E, Figure 8); used across all experiments.
  • Processor steps L = 4
    Architecture hyperparameter chosen at design time; no sensitivity analysis reported.
  • Learnable blending coefficient alpha = learned during training
    Part of the model; trained rather than hand-set, but can be viewed as a free parameter.
assumptions (5)
  • domain assumption The ground truth simulations (LagrangeBench SPH data) accurately represent the target physics
    The paper uses these datasets as ground truth for training and evaluation; any errors in the data would propagate to all results.
  • domain assumption Weakly-compressible Navier-Stokes is the governing equation and the datasets are solved with SPH using a quintic kernel
    Section 3 and Appendix C state this setup; the model is trained to mimic this solver, so the solver's assumptions become the model's assumptions.
  • ad hoc to paper SAGPool-based top-k selection identifies kinematically redundant particles without discarding information critical for dynamics
    Section 4 Downsampler justifies the capacity savings with this claim, but it is not proven; only empirical results support it.
  • ad hoc to paper A fixed Eulerian graph serves as a stable reference frame that corrects trajectory drift via cross-attention
    Section 4 Processor asserts this mechanism; it is supported only by ablation results, not by analysis.
  • ad hoc to paper Learnable aggregation kernel K with row normalization preserves enough information for the processor
    Section 4 Processor defines this without theoretical guarantees about information retention.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hybrid Lagrangian-Eulerian Model for Lagrangian Fluid Simulation." pith.science (2026). https://pith.science/paper/2T7TUFEI

@misc{pith2026260801164,
  author       = {Pith},
  title        = {Pith review of: Hybrid Lagrangian-Eulerian Model for Lagrangian Fluid Simulation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2T7TUFEI}},
  note         = {Machine review of arXiv:2608.01164}
}
read the original abstract

Pure Lagrangian neural simulators offer geometric flexibility and exact advection, making them well-suited for modeling moving domains and free surfaces. However, the absence of a fixed global reference frame introduces two severe limitations: a spatial bottleneck, in which model capacity is wasted on uniform regions because the dense particle neighborhoods required for stable gradients are applied indiscriminately, and rapid temporal drift, caused by purely local message passing that lacks a global anchor. Inspired by classical hybrid numerical solvers, we propose a Hybrid Lagrangian-Eulerian neural simulator that augments Lagrangian dynamics with an Eulerian representation. To address the spatial bottleneck, we introduce adaptive downsampling that eliminates kinematic redundancy, preserving micro-scale details on particles while aggregating compressed features onto Eulerian nodes to resolve large-scale dynamics. To counter temporal drift, we employ a cross-attention mechanism that queries these Eulerian features, using the fixed grid as a stable spatial anchor to correct trajectory deviations at every timestep. Comprehensive experiments show that this hierarchical, cross-attended design substantially suppresses error accumulation, establishing a new state-of-the-art for accuracy and rollout stability in Lagrangian fluid simulation.

Figures

Figures reproduced from arXiv: 2608.01164 by the authors.

Figure 1
Figure 1. Overall framework of the proposed model. We employ an Encoder – Downsampler – [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Visualization of datasets with irregular and deforming domains. [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Five-step rollout results of the ablation study on the downsampler and upsampler, reported [PITH_FULL_IMAGE:figures/full_fig_p017_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Five-step rollout results of the ablation study on the combined self-cross attention feature, [PITH_FULL_IMAGE:figures/full_fig_p017_4.png]
Figure 5
Figure 5. Figure 5: Five-step rollout results of the ablation study on the role of Eulerian feature in Lagrangian [PITH_FULL_IMAGE:figures/full_fig_p018_5.png]
Figure 6
Figure 6. Figure 6: Five-step rollout results of varying the discretization resolution of Eulerian nodes, reported [PITH_FULL_IMAGE:figures/full_fig_p018_6.png]
Figure 7
Figure 7. Figure 7: Five-step rollout results of varying the downsampling ratio, reported on the Lid-Driven [PITH_FULL_IMAGE:figures/full_fig_p019_7.png]
Figure 8
Figure 8. Figure 8: Five-step rollout results of varying the number of downsampling layers, reported on the [PITH_FULL_IMAGE:figures/full_fig_p019_8.png]
Figure 9
Figure 9. Figure 9: Analysis of the runtime and memory usage of models. We evaluate the runtime and memory consump￾tion of different models. Although our approach processes both Lagrangian and Eulerian repre￾sentations, it is naturally amenable to paralleliza￾tion at inference time. In pa…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

143 extracted references · 47 canonical work pages

  1. [3]

    T., Duvenaud, D., and Jacobsen, J.-H

    Behrmann, J., Grathwohl, W., Chen, R. T., Duvenaud, D., and Jacobsen, J.-H. Invertible residual networks. In International Conference on Machine Learning, pp.\ 573--582. PMLR, 2019

    2019

  2. [4]

    Brackbill, J. U. and Ruppel, H. M. FLIP : A method for adaptively zoned, particle-in-cell calculations of fluid flows in two dimensions. Journal of Computational Physics, 65 0 (2): 0 314--343, 1986. doi:10.1016/0021-9991(86)90211-1

    doi:10.1016/0021-9991(86)90211-1 1986

  3. [5]

    Geometric and physical quantities improve e(3) equivariant message passing

    Brandstetter, J., Hesselink, R., van der Pol, E., Bekkers, E., and Welling, M. Geometric and physical quantities improve e(3) equivariant message passing. 2021

    2021

  4. [6]

    E., and Welling, M

    Brandstetter, J., Worrall, D. E., and Welling, M. Message passing neural PDE solvers. In International Conference on Learning Representations, 2022. URL https://openreview.net/forum?id=vSix3HPYKSU

    2022

  5. [7]

    Choose a transformer: Fourier or galerkin, 2021

    Cao, S. Choose a transformer: Fourier or galerkin, 2021. URL https://arxiv.org/abs/2105.14995

    arXiv 2021

  6. [8]

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

    Cao, Y., Chai, M., Li, M., and Jiang, C. Efficient learning of mesh-based physical simulation with bi-stride multi-scale graph neural network. In International Conference on Machine Learning, 2023. URL https://openreview.net/forum?id=2Mbo7IEtZW

    2023

  7. [11]

    and Cani, M.-P

    Desbrun, M. and Cani, M.-P. Smoothed particles: a new paradigm for animating highly deformable bodies. In Proceedings of Eurographics Workshop on Computer Animation and Simulation, Poitiers, France, August 1996

    1996

  8. [12]

    Dwivedi, V. P. and Bresson, X. A generalization of transformer networks to graphs, 2021. URL https://arxiv.org/abs/2012.09699

    arXiv 2021

Show all 143 references
  1. [13]

    Drivaernet++: A large-scale multimodal car dataset with computational fluid dynamics simulations and deep learning benchmarks

    Elrefaie, M., Morar, F., Dai, A., and Ahmed, F. Drivaernet++: A large-scale multimodal car dataset with computational fluid dynamics simulations and deep learning benchmarks. In Globerson, A., Mackey, L., Belgrave, D., Fan, A., Paquet, U., Tomczak, J., and Zhang, C. (eds.), Ad...

    2024

  2. [15]

    Francis Harvey Harlow, Martha Evans, R. D. R. A machine calculation method for hydrodynamic problems. Report, Los Alamos Scientific Laboratory of the University of California, 1955

    1955

  3. [16]

    and Ji, S

    Gao, H. and Ji, S. Graph u-nets. In International Conference on Machine Learning, pp.\ 2083--2092, 2019

  4. [17]

    Gingold, R. A. and Monaghan, J. J. Smoothed particle hydrodynamics: Theory and application to non-spherical stars . Mon. Not. Roy. Astron. Soc., 181: 0 375, 1977

    1977

  5. [18]

    Learning to optimize multigrid PDE solvers

    Greenfeld, D., Galun, M., Basri, R., Yavneh, I., and Kimmel, R. Learning to optimize multigrid PDE solvers. In International Conference on Machine Learning, pp.\ 2415--2423. PMLR, 2019

    2019

  6. [22]

    Poseidon: Efficient foundation models for PDE s

    Herde, M., Raonic, B., Rohner, T., K \"a ppeli, R., Molinaro, R., de Bezenac, E., and Mishra, S. Poseidon: Efficient foundation models for PDE s. In The Thirty-eighth Annual Conference on Neural Information Processing Systems, 2024. URL https://openreview.net/forum?id=JC1VKK3UXk

    2024

  7. [23]

    Hoover, W. G. Smooth Particle Applied Mechanics: The State of the Art. World Scientific, Singapore, 2006

    2006

  8. [24]

    Huang, J., Yang, G., Wang, Z., and Park, J. J. Diffusionpde: Generative pde-solving under partial observation, 2024. URL https://arxiv.org/abs/2406.17763

    2024 arXiv

  9. [25]

    Topology-aware neural flux prediction guided by physics

    Jiang, H., Wang, J., Zhu, X., and He, Y. Topology-aware neural flux prediction guided by physics. In Forty-second International Conference on Machine Learning, 2025. URL https://openreview.net/forum?id=8Um2YotdbD

    2025

  10. [26]

    An airflow velocity field reconstruction method with sparse or incomplete data using physics-informed neural network

    Jing, G., Wang, H., Li, X., Wang, G., and Yang, Y. An airflow velocity field reconstruction method with sparse or incomplete data using physics-informed neural network. Journal of Building Engineering, 88: 0 109231, July 2024. Published: 1 July 2024

    2024

  11. [27]

    Johnson, N. L. The legacy and future of CFD at Los Alamos . In Proceedings of the 1996 Canadian CFD Conference. Office of Scientific and Technical Information (OSTI), 1996. OSTI 244662

    1996

  12. [28]

    Kingma, D. P. and Ba, J. Adam: A method for stochastic optimization, 2017. URL https://arxiv.org/abs/1412.6980

    2017 arXiv

  13. [29]

    Benchmarking invertible architectures on inverse problems

    Kruse, J., Ardizzone, L., Rother, C., and K \"o the, U. Benchmarking invertible architectures on inverse problems. arXiv preprint, 2021

    2021

  14. [30]

    Self-attention graph pooling

    Lee, J., Lee, I., and Kang, J. Self-attention graph pooling. In Proceedings of the 36th International Conference on Machine Learning, 09--15 Jun 2019

    2019

  15. [31]

    Flow field reconstruction with sensor placement policy learning

    Li, R., Wan, G., Huang, Z., Liu, Z., Wang, H., Luo, X., Wang, W., and Sun, Y. Flow field reconstruction with sensor placement policy learning. In The Thirty-ninth Annual Conference on Neural Information Processing Systems, 2025. URL https://openreview.net/forum?id=1esFEjMUBS

    2025

  16. [32]

    Neural operator: Graph kernel network for partial differential equations, 2020

    Li, Z., Kovachki, N., Azizzadenesheli, K., Liu, B., Bhattacharya, K., Stuart, A., and Anandkumar, A. Neural operator: Graph kernel network for partial differential equations, 2020. URL https://arxiv.org/abs/2003.03485

    2020 arXiv

  17. [33]

    B., Azizzadenesheli, K., liu, B., Bhattacharya, K., Stuart, A., and Anandkumar, A

    Li, Z., Kovachki, N. B., Azizzadenesheli, K., liu, B., Bhattacharya, K., Stuart, A., and Anandkumar, A. Fourier neural operator for parametric partial differential equations. In International Conference on Learning Representations, 2021. URL https://openreview.net/forum?id=c8P9NQVtmnO

    2021

  18. [34]

    B., Choy, C., Li, B., Kossaifi, J., Otta, S

    Li, Z., Kovachki, N. B., Choy, C., Li, B., Kossaifi, J., Otta, S. P., Nabian, M. A., Stadler, M., Hundt, C., Azizzadenesheli, K., and Anandkumar, A. Geometry-informed neural operator for large-scale 3d PDE s. In Thirty-seventh Conference on Neural Information Processing System...

    2023

  19. [35]

    Physics-informed neural operator for learning partial differential equations, 2023 b

    Li, Z., Zheng, H., Kovachki, N., Jin, D., Chen, H., Liu, B., Azizzadenesheli, K., and Anandkumar, A. Physics-informed neural operator for learning partial differential equations, 2023 b . URL https://arxiv.org/abs/2111.03794

    2023 arXiv

  20. [36]

    Learned turbulence modelling with differentiable fluid solvers: physics-based loss functions and optimisation horizons

    List, B., Chen, L.-W., and Thuerey, N. Learned turbulence modelling with differentiable fluid solvers: physics-based loss functions and optimisation horizons. Journal of Fluid Mechanics, 949: 0 A25, 2022

    2022

  21. [37]

    Lu, L., Jin, P., Pang, G., Zhang, Z., and Karniadakis, G. E. Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators. Nature Machine Intelligence, 3 0 (3): 0 218--229, 2021

    2021

  22. [38]

    Transolver++: An accurate neural solver for PDE s on million-scale geometries

    Luo, H., Wu, H., Zhou, H., Xing, L., Di, Y., Wang, J., and Long, M. Transolver++: An accurate neural solver for PDE s on million-scale geometries. In Forty-second International Conference on Machine Learning, 2025. URL https://openreview.net/forum?id=AM7iAh0krx

    2025

  23. [39]

    R.-S., Parker, L

    McCabe, M., Blancard, B. R.-S., Parker, L. H., Ohana, R., Cranmer, M., Bietti, A., Eickenberg, M., Golkar, S., Krawezik, G., Lanusse, F., Pettee, M., Tesileanu, T., Cho, K., and Ho, S. Multiple physics pretraining for spatiotemporal surrogate models. In The Thirty-eighth Annua...

    2024

  24. [40]

    and Magri, L

    Mo, Y. and Magri, L. Reconstructing unsteady flows from sparse, noisy measurements with a physics-constrained convolutional neural network, 2024. URL https://arxiv.org/abs/2409.00260

    2024 arXiv

  25. [42]

    Monaghan, J. J. Smoothed particle hydrodynamics. Annual Review of Astronomy and Astrophysics, 30: 0 543--574, 1992. doi:10.1146/annurev.aa.30.090192.002551

    1992

  26. [43]

    Nairn, J. A. Material point method calculations with explicit cracks. Computer Modeling in Engineering & Sciences, 4 0 (6): 0 649--664, 2003. doi:10.3970/cmes.2003.004.649

    2003 doi

  27. [44]

    Pytorch: An imperative style, high-performance deep learning library, 2019

    Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., Desmaison, A., Köpf, A., Yang, E., DeVito, Z., Raison, M., Tejani, A., Chilamkurthy, S., Steiner, B., Fang, L., Bai, J., and Chintala, S. Pytorch: An impera...

    2019 arXiv

  28. [45]

    Pfaff, T., Fortunato, M., Sanchez-Gonzalez, A., and Battaglia, P. W. Learning mesh-based simulation with graph networks. In International Conference on Learning Representations, 2021

    2021

  29. [46]

    Guaranteed conservation of momentum for learning particle-based fluid dynamics

    Prantl, L., Ummenhofer, B., Koltun, V., and Thuerey, N. Guaranteed conservation of momentum for learning particle-based fluid dynamics. In Oh, A. H., Agarwal, A., Belgrave, D., and Cho, K. (eds.), Advances in Neural Information Processing Systems, 2022. URL https://openreview....

    2022

  30. [49]

    A neural material point method for particle-based emulation

    Rochman-Sharabi, O., Lewin, S., and Louppe, G. A neural material point method for particle-based emulation. Transactions on Machine Learning Research, 2025. ISSN 2835-8856. URL https://openreview.net/forum?id=zSK81A2hxQ

    2025

  31. [50]

    Sanchez-Gonzalez, A., Godwin, J., Pfaff, T., Ying, R., Leskovec, J., and Battaglia, P. W. Learning to simulate complex physics with graph networks, 2020. URL https://arxiv.org/abs/2002.09405

    2020 arXiv

  32. [51]

    Deep learning of preconditioners for conjugate gradient solvers in urban water related problems

    Sappl, J., Seiler, L., Harders, M., and Rauch, W. Deep learning of preconditioners for conjugate gradient solvers in urban water related problems. arXiv preprint, 2019

    2019

  33. [52]

    UPS : Efficiently building foundation models for PDE solving via cross-modal adaptation

    Shen, J., Marwah, T., and Talwalkar, A. UPS : Efficiently building foundation models for PDE solving via cross-modal adaptation. Transactions on Machine Learning Research, 2024. ISSN 2835-8856. URL https://openreview.net/forum?id=0r9mhjRv1E

    2024

  34. [53]

    Sulsky, D., Chen, Z., and Schreyer, H. L. A particle method for history-dependent materials. Computer Methods in Applied Mechanics and Engineering, 118 0 (1): 0 179--196, 1994. doi:10.1016/0045-7825(94)90112-0

    1994 doi

  35. [54]

    A neural PDE solver with temporal stencil modeling

    Sun, Z., Yang, Y., and Yoo, S. A neural PDE solver with temporal stencil modeling. arXiv preprint, 2023

    2023

  36. [56]

    and Choromanska, A

    Teng, Y. and Choromanska, A. Invertible autoencoder for domain adaptation. Computation, 7 0 (2): 0 20, 2019

    2019

  37. [57]

    Lagrangebench: A lagrangian fluid mechanics benchmarking suite

    Toshev, A., Galletti, G., Fritz, F., Adami, S., and Adams, N. Lagrangebench: A lagrangian fluid mechanics benchmarking suite. Advances in Neural Information Processing Systems, 36, 2024 a

    2024

  38. [58]

    P., Galletti, G., Brandstetter, J., Adami, S., and Adams, N

    Toshev, A. P., Galletti, G., Brandstetter, J., Adami, S., and Adams, N. A. Learning lagrangian fluid mechanics with e( 3 )-equivariant graph neural networks, 2023

    2023

  39. [59]

    P., Erbesdobler, J

    Toshev, A. P., Erbesdobler, J. A., Adams, N. A., and Brandstetter, J. Neural sph: Improved neural modeling of lagrangian fluid dynamics, 2024 b . URL https://arxiv.org/abs/2402.06275

    2024 arXiv

  40. [60]

    Tran, A., Mathews, A., Xie, L., and Ong, C. S. Factorized fourier neural operators. In The Eleventh International Conference on Learning Representations, 2023. URL https://openreview.net/forum?id=tmIiMPl4IPa

    2023

  41. [61]

    Chapter 6: The particle-in-cell method

    Tskhakaya, D. Chapter 6: The particle-in-cell method. In Fehske, H., Schneider, R., and Wei e, A. (eds.), Computational Many-Particle Physics, volume 739 of Lecture Notes in Physics, pp.\ 271--?,. Springer, Berlin, Heidelberg, 2008. ISBN 978-3-540-74685-0. doi:10.1007/978-3-54...

    2008 doi

  42. [63]

    Recent advances on machine learning for computational fluid dynamics: A survey, 2024

    Wang, H., Cao, Y., Huang, Z., Liu, Y., Hu, P., Luo, X., Song, Z., Zhao, W., Liu, J., Sun, J., Zhang, S., Wei, L., Wang, Y., Wu, T., Ma, Z.-M., and Sun, Y. Recent advances on machine learning for computational fluid dynamics: A survey, 2024. URL https://arxiv.org/abs/2408.12171

    2024 arXiv

  43. [64]

    Fd-bench: A modular and fair benchmark for data-driven fluid simulation, 2025

    Wang, H., Li, R., Xu, F., Sun, F., Han, K., Huang, Z., Wan, G., Chang, C., Luo, X., Wang, W., and Sun, Y. Fd-bench: A modular and fair benchmark for data-driven fluid simulation, 2025. URL https://arxiv.org/abs/2505.20349

    2025 arXiv

  44. [65]

    and Wang, C

    Wang, T. and Wang, C. Latent neural operator for solving forward and inverse pde problems, 2024. URL https://arxiv.org/abs/2406.03923

    2024 arXiv

  45. [66]

    C., Doucet, A., and Syed, S

    Williams, C., Falck, F., Deligiannidis, G., Holmes, C. C., Doucet, A., and Syed, S. A unified framework for u-net design and analysis. In Thirty-seventh Conference on Neural Information Processing Systems, 2023. URL https://openreview.net/forum?id=43ruO2fMjq

    2023

  46. [67]

    Solving high-dimensional pdes with latent spectral models

    Wu, H., Hu, T., Luo, H., Wang, J., and Long, M. Solving high-dimensional pdes with latent spectral models. In International Conference on Machine Learning, 2023

    2023

  47. [68]

    Transolver: A fast transformer solver for pdes on general geometries

    Wu, H., Luo, H., Wang, H., Wang, J., and Long, M. Transolver: A fast transformer solver for pdes on general geometries. In International Conference on Machine Learning, 2024

    2024

  48. [69]

    Wu, Q., Yang, C., Zeng, K., and Bronstein, M. M. Supercharging graph transformers with advective diffusion. In Forty-second International Conference on Machine Learning, 2025. URL https://openreview.net/forum?id=MaOYl3P84E

    2025

  49. [70]

    Hybrid neural-mpm for interactive fluid simulations in real-time, 2025

    Xu, J., Huang, H., Zou, C., Savva, M., Wei, Y., and Chen, W. Hybrid neural-mpm for interactive fluid simulations in real-time, 2025. URL https://arxiv.org/abs/2505.18926

    2025 arXiv

  50. [71]

    Rf-pinns: Reactive flow physics-informed neural networks for field reconstruction of laminar and turbulent flames using sparse data

    Yadav, V., Casel, M., and Ghani, A. Rf-pinns: Reactive flow physics-informed neural networks for field reconstruction of laminar and turbulent flames using sparse data. Journal of Computational Physics, 524: 0 113698, March 2025. Published: 1 March 2025

    2025

  51. [72]

    PDE former: Towards a foundation model for one-dimensional partial differential equations

    Ye, Z., Huang, X., Chen, L., Liu, H., Wang, Z., and Dong, B. PDE former: Towards a foundation model for one-dimensional partial differential equations. In ICLR 2024 Workshop on AI4DifferentialEquations In Science, 2024. URL https://openreview.net/forum?id=GLDMCwdhTK

    2024

  52. [73]

    Sparse sensor reconstruction of vortex-impinged airfoil wake with machine learning

    Zhong, Y., Fukami, K., An, B., et al. Sparse sensor reconstruction of vortex-impinged airfoil wake with machine learning. Theor. Comput. Fluid Dyn., 37: 0 269--287, 2023. doi:10.1007/s00162-023-00657-y

    2023 doi

  53. [74]

    Unisolver: Pde-conditional transformers towards universal neural pde solvers

    Zhou, H., Ma, Y., Wu, H., Wang, H., and Long, M. Unisolver: Pde-conditional transformers towards universal neural pde solvers. In Forty-second International Conference on Machine Learning

  54. [75]

    2025 , eprint=

    FD-Bench: A Modular and Fair Benchmark for Data-driven Fluid Simulation , author=. 2025 , eprint=

    2025

  55. [76]

    Advances in Neural Information Processing Systems , volume=

    Lagrangebench: A lagrangian fluid mechanics benchmarking suite , author=. Advances in Neural Information Processing Systems , volume=

  56. [77]

    2006 , publisher =

    Smooth Particle Applied Mechanics: The State of the Art , author =. 2006 , publisher =

    2006

  57. [78]

    Annual Review of Astronomy and Astrophysics , year =

    Smoothed Particle Hydrodynamics , author =. Annual Review of Astronomy and Astrophysics , year =

  58. [79]

    Proceedings of Eurographics Workshop on Computer Animation and Simulation , year =

    Smoothed Particles: a new paradigm for animating highly deformable bodies , author =. Proceedings of Eurographics Workshop on Computer Animation and Simulation , year =

  59. [80]

    2020 , eprint=

    Learning to Simulate Complex Physics with Graph Networks , author=. 2020 , eprint=

    2020

  60. [81]

    2021 , eprint=

    Geometric and Physical Quantities improve E(3) Equivariant Message Passing , author=. 2021 , eprint=

    2021

  61. [82]

    International Conference on Machine Learning , pages=

    Graph U-Nets , author=. International Conference on Machine Learning , pages=

  62. [83]

    Forty-second International Conference on Machine Learning , year=

    Topology-aware Neural Flux Prediction Guided by Physics , author=. Forty-second International Conference on Machine Learning , year=

  63. [84]

    2021 , eprint=

    A Generalization of Transformer Networks to Graphs , author=. 2021 , eprint=

    2021

  64. [85]

    Forty-second International Conference on Machine Learning , year=

    Supercharging Graph Transformers with Advective Diffusion , author=. Forty-second International Conference on Machine Learning , year=

  65. [86]

    International Conference on Machine Learning , year =

    Efficient Learning of Mesh-Based Physical Simulation with Bi-Stride Multi-Scale Graph Neural Network , author =. International Conference on Machine Learning , year =

  66. [87]

    International Conference on Learning Representations , year=

    Fourier Neural Operator for Parametric Partial Differential Equations , author=. International Conference on Learning Representations , year=

  67. [88]

    Worrall and Max Welling , booktitle=

    Johannes Brandstetter and Daniel E. Worrall and Max Welling , booktitle=. Message Passing Neural. 2022 , url=

    2022

  68. [89]

    Thirty-seventh Conference on Neural Information Processing Systems , year=

    A Unified Framework for U-Net Design and Analysis , author=. Thirty-seventh Conference on Neural Information Processing Systems , year=

  69. [90]

    Jimack , keywords =

    Maria Luisa Taccari and Jonathan Nuttall and Xiaohui Chen and He Wang and Bennie Minnema and Peter K. Jimack , keywords =. Attention U-Net as a surrogate model for groundwater prediction , journal =. 2022 , issn =. doi:https://doi.org/10.1016/j.advwatres.2022.104169 , url =

    2022

  70. [91]

    2023 , eprint=

    Physics-Informed Neural Operator for Learning Partial Differential Equations , author=. 2023 , eprint=

    2023

  71. [92]

    International Conference on Learning Representations , year=

    Learning Mesh-Based Simulation with Graph Networks , author=. International Conference on Learning Representations , year=

  72. [93]

    International Conference on Machine Learning , year=

    Transolver: A Fast Transformer Solver for PDEs on General Geometries , author=. International Conference on Machine Learning , year=

  73. [94]

    Transolver++: An Accurate Neural Solver for

    Huakun Luo and Haixu Wu and Hang Zhou and Lanxiang Xing and Yichen Di and Jianmin Wang and Mingsheng Long , booktitle=. Transolver++: An Accurate Neural Solver for. 2025 , url=

    2025

  74. [95]

    Geometry-Informed Neural Operator for Large-Scale 3D

    Zongyi Li and Nikola Borislavov Kovachki and Chris Choy and Boyi Li and Jean Kossaifi and Shourya Prakash Otta and Mohammad Amin Nabian and Maximilian Stadler and Christian Hundt and Kamyar Azizzadenesheli and Anima Anandkumar , booktitle=. Geometry-Informed Neural Operator fo...

    2023

  75. [96]

    Transactions on Machine Learning Research , issn=

    A Neural Material Point Method for Particle-based Emulation , author=. Transactions on Machine Learning Research , issn=. 2025 , url=

    2025

  76. [97]

    2024 , eprint=

    Neural SPH: Improved Neural Modeling of Lagrangian Fluid Dynamics , author=. 2024 , eprint=

    2024

  77. [98]

    2023 , eprint=

    Learning Lagrangian Fluid Mechanics with E( 3 )-Equivariant Graph Neural Networks , author=. 2023 , eprint=

    2023

  78. [99]

    2024 , eprint=

    Recent Advances on Machine Learning for Computational Fluid Dynamics: A Survey , author=. 2024 , eprint=

    2024

  79. [100]

    2007 , isbn =

    LeVeque, Randall , title =. 2007 , isbn =

    2007

  80. [101]

    , biburl =

    LeVeque, Randall J. , biburl =

  81. [102]

    1989 , issue_date =

    BELYTSCHKO, TED , title =. 1989 , issue_date =. doi:10.1111/j.1467-8667.1989.tb00025.x , journal =

    1989

  82. [103]

    Gingold, R. A. and Monaghan, J. J. Smoothed particle hydrodynamics: Theory and application to non-spherical stars. Mon. Not. Roy. Astron. Soc. 1977

    1977

  83. [104]

    Lucy, L B , title =. Astron. J.; (United States) , issn =. 1977 , month =. doi:10.1086/112164 , url =

    1977 doi

  84. [105]

    Koshizuka and Y

    S. Koshizuka and Y. Oka , title =. Nuclear Science and Engineering , volume =. 1996 , publisher =. doi:10.13182/NSE96-A24205 , URL =

    1996 doi

  85. [106]

    Computer Methods in Applied Mechanics and Engineering , year=

    A particle method for history-dependent materials , author=. Computer Methods in Applied Mechanics and Engineering , year=

  86. [107]

    Advances in Neural Information Processing Systems , editor=

    Guaranteed Conservation of Momentum for Learning Particle-based Fluid Dynamics , author=. Advances in Neural Information Processing Systems , editor=. 2022 , url=

    2022

  87. [108]

    Proceedings of the 36th International Conference on Machine Learning , year =

    Self-Attention Graph Pooling , author =. Proceedings of the 36th International Conference on Machine Learning , year =

  88. [109]

    1955 , institution =

    A Machine Calculation Method for Hydrodynamic Problems , author =. 1955 , institution =

    1955

  89. [110]

    Computational Many-Particle Physics , editor =

    Tskhakaya, David , title =. Computational Many-Particle Physics , editor =. 2008 , doi =

    2008

  90. [111]

    Particle simulation of plasmas , author =. Rev. Mod. Phys. , volume =. 1983 , month =. doi:10.1103/RevModPhys.55.403 , url =

    1983 doi

  91. [112]

    Monaghan , abstract =

    J.J. Monaghan , abstract =. Simulating Free Surface Flows with SPH , journal =. 1994 , issn =. doi:https://doi.org/10.1006/jcph.1994.1034 , url =

    1994

  92. [113]

    Fundamental differences between SPH and grid methods: Simulating fluids using SPH and grid techniques , volume=

    Agertz, Oscar and Moore, Ben and Stadel, Joachim and Potter, Doug and Miniati, Francesco and Read, Justin and Mayer, Lucio and Gawryszczak, Artur and Kravtsov, Andrey and Nordlund, Ake and Pearce, Frazer and Quilis, Vicent and Rudd, Douglas and Springel, Volker and Stone, Jame...

    2007

  93. [114]

    Price, Daniel J. , year=. Smoothed particle hydrodynamics and magnetohydrodynamics , volume=. Journal of Computational Physics , publisher=. doi:10.1016/j.jcp.2010.12.011 , number=

    2010 doi

  94. [115]

    Improving convergence in smoothed particle hydrodynamics simulations without pairing instability: SPH without pairing instability , volume=

    Dehnen, Walter and Aly, Hossam , year=. Improving convergence in smoothed particle hydrodynamics simulations without pairing instability: SPH without pairing instability , volume=. Monthly Notices of the Royal Astronomical Society , publisher=. doi:10.1111/j.1365-2966.2012.214...

    2012

  95. [116]

    2025 , eprint=

    CORGI: GNNs with Convolutional Residual Global Interactions for Lagrangian Simulation , author=. 2025 , eprint=

    2025

  96. [117]

    2025 , eprint=

    Hybrid Neural-MPM for Interactive Fluid Simulations in Real-Time , author=. 2025 , eprint=

    2025

  97. [118]

    , booktitle =

    Johnson, Norman L. , booktitle =. The legacy and future of. 1996 , note =

    1996

  98. [119]

    and Ruppel, Hans M

    Brackbill, Jeremiah U. and Ruppel, Hans M. , journal =. 1986 , doi =

    1986

  99. [120]

    Computer Methods in Applied Mechanics and Engineering , volume =

    A particle method for history-dependent materials , author =. Computer Methods in Applied Mechanics and Engineering , volume =. 1994 , doi =

    1994

  100. [121]

    Computer Modeling in Engineering & Sciences , volume =

    Material Point Method Calculations with Explicit Cracks , author =. Computer Modeling in Engineering & Sciences , volume =. 2003 , doi =

    2003

  101. [122]

    The Thirty-ninth Annual Conference on Neural Information Processing Systems , year=

    Flow Field Reconstruction with Sensor Placement Policy Learning , author=. The Thirty-ninth Annual Conference on Neural Information Processing Systems , year=

  102. [123]

    2023 , eprint=

    Geometry-Informed Neural Operator for Large-Scale 3D PDEs , author=. 2023 , eprint=

    2023

  103. [124]

    The Eleventh International Conference on Learning Representations , year=

    Factorized Fourier Neural Operators , author=. The Eleventh International Conference on Learning Representations , year=

  104. [125]

    International Conference on Machine Learning , year=

    Solving High-Dimensional PDEs with Latent Spectral Models , author=. International Conference on Machine Learning , year=

  105. [126]

    2024 , eprint=

    DiffusionPDE: Generative PDE-Solving Under Partial Observation , author=. 2024 , eprint=

    2024

  106. [127]

    2024 , eprint=

    Latent Neural Operator for Solving Forward and Inverse PDE Problems , author=. 2024 , eprint=

    2024

  107. [128]

    2021 , eprint=

    Choose a Transformer: Fourier or Galerkin , author=. 2021 , eprint=

    2021

  108. [129]

    arXiv preprint arXiv:2302.14376 , year=

    GNOT: A General Neural Operator Transformer for Operator Learning , author=. arXiv preprint arXiv:2302.14376 , year=

    arXiv

  109. [130]

    Learning nonlinear operators via

    Lu, Lu and Jin, Pengzhan and Pang, Guofei and Zhang, Zhongqiang and Karniadakis, George Em , journal =. Learning nonlinear operators via

  110. [131]

    arXiv preprint arXiv:2402.12365 , year=

    Universal Physics Transformers , author=. arXiv preprint arXiv:2402.12365 , year=

    arXiv

  111. [132]

    arXiv preprint arXiv:2403.03542 , year=

    DPOT: Auto-Regressive Denoising Operator Transformer for Large-Scale PDE Pre-Training , author=. arXiv preprint arXiv:2403.03542 , year=

    arXiv

  112. [133]

    The Thirty-eighth Annual Conference on Neural Information Processing Systems , year=

    Multiple Physics Pretraining for Spatiotemporal Surrogate Models , author=. The Thirty-eighth Annual Conference on Neural Information Processing Systems , year=

  113. [134]

    2024 , url=

    Zhanhong Ye and Xiang Huang and Leheng Chen and Hongsheng Liu and Zidong Wang and Bin Dong , booktitle=. 2024 , url=

    2024

  114. [135]

    Poseidon: Efficient Foundation Models for

    Maximilian Herde and Bogdan Raonic and Tobias Rohner and Roger K. Poseidon: Efficient Foundation Models for. The Thirty-eighth Annual Conference on Neural Information Processing Systems , year=

  115. [136]

    2024 , url=

    Junhong Shen and Tanya Marwah and Ameet Talwalkar , journal=. 2024 , url=

    2024

  116. [137]

    Unisolver: PDE-Conditional Transformers Towards Universal Neural PDE Solvers , author=

  117. [138]

    Flow completion network: Inferring the fluid dynamics from incomplete flow information using graph neural networks , volume=

    He, Xiaodong and Wang, Yinan and Li, Juan , year=. Flow completion network: Inferring the fluid dynamics from incomplete flow information using graph neural networks , volume=. Physics of Fluids , publisher=. doi:10.1063/5.0097688 , number=

    doi

  118. [139]

    2024 , eprint=

    Reconstructing unsteady flows from sparse, noisy measurements with a physics-constrained convolutional neural network , author=. 2024 , eprint=

    2024

  119. [140]

    Journal of Building Engineering , volume =

    An airflow velocity field reconstruction method with sparse or incomplete data using physics-informed neural network , author =. Journal of Building Engineering , volume =. 2024 , month = jul, note =

    2024

  120. [141]

    Journal of Computational Physics , volume =

    Vikas Yadav and Mario Casel and Abdulla Ghani , title =. Journal of Computational Physics , volume =. 2025 , month = mar, note =

    2025

  121. [142]

    DrivAerNet: A Parametric Car Dataset for Data-Driven Aerodynamic Design and Prediction , volume=

    Elrefaie, Mohamed and Dai, Angela and Ahmed, Faez , year=. DrivAerNet: A Parametric Car Dataset for Data-Driven Aerodynamic Design and Prediction , volume=. Journal of Mechanical Design , publisher=. doi:10.1115/1.4068104 , number=

    doi

  122. [143]

    DrivAerNet++: A Large-Scale Multimodal Car Dataset with Computational Fluid Dynamics Simulations and Deep Learning Benchmarks , volume =

    Elrefaie, Mohamed and Morar, Florin and Dai, Angela and Ahmed, Faez , booktitle =. DrivAerNet++: A Large-Scale Multimodal Car Dataset with Computational Fluid Dynamics Simulations and Deep Learning Benchmarks , volume =

  123. [144]

    Journal of Fluid Mechanics , volume =

    Learned turbulence modelling with differentiable fluid solvers: physics-based loss functions and optimisation horizons , author =. Journal of Fluid Mechanics , volume =. 2022 , publisher =

    2022

  124. [145]

    and Yang, Y

    Sun, Z. and Yang, Y. and Yoo, S. , journal =. A neural

  125. [146]

    Learning to optimize multigrid

    Greenfeld, Daniel and Galun, Meirav and Basri, Ronen and Yavneh, Irad and Kimmel, Ron , booktitle =. Learning to optimize multigrid. 2019 , organization =

    2019

  126. [147]

    arXiv preprint , year =

    Deep learning of preconditioners for conjugate gradient solvers in urban water related problems , author =. arXiv preprint , year =

  127. [148]

    International Conference on Machine Learning , pages =

    Invertible residual networks , author =. International Conference on Machine Learning , pages =. 2019 , organization =

    2019

  128. [149]

    Computation , volume =

    Invertible autoencoder for domain adaptation , author =. Computation , volume =

  129. [150]

    arXiv preprint , year =

    Benchmarking invertible architectures on inverse problems , author =. arXiv preprint , year =

  130. [151]

    2019 , eprint=

    PyTorch: An Imperative Style, High-Performance Deep Learning Library , author=. 2019 , eprint=

    2019

  131. [152]

    2017 , eprint=

    Adam: A Method for Stochastic Optimization , author=. 2017 , eprint=

    2017

  132. [153]

    Sparse sensor reconstruction of vortex-impinged airfoil wake with machine learning , author=. Theor. Comput. Fluid Dyn. , volume=. 2023 , doi=

    2023

  133. [154]

    2020 , eprint=

    Neural Operator: Graph Kernel Network for Partial Differential Equations , author=. 2020 , eprint=

    2020

  134. [155]

    arXiv preprint arXiv:2407.03925 , year=

    Reduced-Order Neural Operators: Learning Lagrangian Dynamics on Highly Sparse Graphs , author=. arXiv preprint arXiv:2407.03925 , year=

  135. [156]

    and Menon, Dileep and Govind, Rahul and Sanka, Suraj and Sebastian, Amal S

    Ramachandran, Prabhu and Bhosale, Aditya and Puri, Kunal and Negi, Pawan and Muta, Abhinav and Dinesh, A. and Menon, Dileep and Govind, Rahul and Sanka, Suraj and Sebastian, Amal S. and Sen, Ananyo and Kaushik, Rohan and Kumar, Anshuman and Kurapati, Vikas and Patil, Mrinalgou...

    doi

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.