pith. sign in

arxiv: 2605.23712 · v1 · pith:NOKBQG57new · submitted 2026-05-22 · 💻 cs.CE · cs.LG

Operator Learning for Reconstructing Flow Fields from Sparse Measurements: a Language Model Approach

Qian Zhang , George Em Karniadakis This is my paper

Pith reviewed 2026-05-25 02:22 UTC · model grok-4.3

classification 💻 cs.CE cs.LG
keywords operator learningflow field reconstructionsparse measurementslanguage modelssequence-to-sequencemesh-free methodsfluid mechanics
0
0 comments X

The pith

Language models reconstruct flow fields from sparse measurements by treating the task as sequence-to-sequence learning.

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

The paper proposes an operator learning method that uses language model architectures to reconstruct full flow fields from very sparse input data in a mesh-free way. Sparse measurements are cast as context tokens and the remaining field locations as query tokens in a sequence-to-sequence setup, allowing the model to learn spatial correlations and long-range dependencies directly from data. The approach is demonstrated on four benchmark cases spanning two-dimensional vortex streets, continental temperature fields, three-dimensional blood flow, and turbulent jet measurements. Competitive accuracy is reported even when fewer than 10 percent of the points are observed.

Core claim

Flow field reconstruction can be performed mesh-free by reformulating the problem as a sequence-to-sequence task inside a language model architecture, with sparse measurements supplied as context and unobserved locations supplied as queries; the resulting operator learns the necessary spatial correlations and long-range dependencies without an explicit mesh or physical constraints.

What carries the argument

Language-model architecture adapted to sequence-to-sequence operator learning, where sparse measurements act as context and full-field queries produce the reconstructed flow values.

If this is right

  • Accurate reconstruction holds for highly incomplete inputs below 10 percent observed across all tested cases.
  • The same architecture processes both two-dimensional and three-dimensional data from simulations and experiments without modification.
  • No mesh generation or enforcement of governing equations is required during inference.
  • The results point toward language models as a scalable route for scientific data completion tasks.

Where Pith is reading between the lines

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

  • The same context-query framing could be applied to other inverse problems that supply scattered sensor readings, such as reconstructing pressure or temperature fields from limited probes.
  • If the model truly learns general operators, fine-tuning on one fluid regime might transfer to related but unseen regimes with minimal additional data.
  • Hybrid training that adds soft physical penalties to the language-model loss could be tested to check whether accuracy improves on extrapolation cases.

Load-bearing premise

Reformulating sparse flow reconstruction as a sequence-to-sequence task with language model architecture will capture the necessary spatial correlations and long-range dependencies without a mesh or explicit physical constraints.

What would settle it

Reconstruction error on a held-out flow dataset with under 10 percent observed points that substantially exceeds error from standard interpolation or mesh-based methods would falsify competitive accuracy.

Figures

Figures reproduced from arXiv: 2605.23712 by George Em Karniadakis, Qian Zhang.

Figure 1
Figure 1. Figure 1: The architecture of the reconstruction framework. Observation tokens (orange, carrying both position [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: True flow field (left), observed flow field (middle), and reconstructed field (right) for snapshot index 90. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Ablation study on the vortex-street dataset. Error bars denote one standard deviation across test [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Map of weather stations across the contiguous United States. Orange dots represent observed stations, [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: True temperature field (left) and reconstructed temperature field (right) on two different days in 2024. [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Reconstruction results for the blood flow dataset at the mid-plane. The figure compares sparse observa [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Time-resolved relative RMSE for the u, v components and total in-plane velocity field across the validation snapshots. Our method is among the best-performing methods across all snapshots. 3.4 3D Turbulent Jet Flow Turbulent jet flow experiments are fundamental to understanding the evolution of engine exhaust dynamics, which play a critical role in the design of efficient propulsion systems. Accurate recon… view at source ↗
Figure 8
Figure 8. Figure 8: Reconstruction results for snapshot index 180 at the mid-plane [PITH_FULL_IMAGE:figures/full_fig_p013_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Energy-spectrum comparison for the 3D turbulent-jet case. The x-axis represents spatial wavenumber [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗
read the original abstract

Reconstructing flow fields from sparse measurements is a fundamental problem in fluid mechanics with broad implications for modeling, control, and design. In this work, we propose a novel operator learning framework that leverages the architecture of language models to perform flow reconstruction in a mesh-free manner. We reformulate flow field reconstruction as a sequence-to-sequence learning task, where sparse measurements are treated as context and unobserved locations as queries. Our model learns to reconstruct the full flow field from sparse inputs, effectively capturing spatial correlations and long-range dependencies. We evaluate the proposed approach on four benchmark datasets: (1) two-dimensional vortex street simulations, (2) daily average temperature data across the contiguous United States, (3) three-dimensional blood flow simulations based on dissipative particle dynamics, and (4) three-dimensional turbulent jet flow measurements obtained via particle tracking velocimetry. Across all cases, our method demonstrates competitive reconstruction accuracy, even with highly incomplete data (less than 10\% observed), and achieves efficient performance. The results highlight the potential of language models as robust and scalable tools for scientific data reconstruction, and suggest a promising direction toward the development of foundation models for scientific and engineering applications.

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

2 major / 1 minor

Summary. The manuscript proposes a novel operator learning framework that reformulates sparse flow field reconstruction as a sequence-to-sequence task using language model architectures, treating sparse measurements as context tokens and unobserved locations as query targets. It evaluates the mesh-free approach on four benchmark datasets (2D vortex street simulations, contiguous US daily temperature data, 3D blood flow via dissipative particle dynamics, and 3D turbulent jet PTV measurements), claiming competitive reconstruction accuracy even with less than 10% observed data while capturing spatial correlations and long-range dependencies.

Significance. If the performance claims hold with proper quantitative support, the work would indicate that transformer-style models can serve as scalable, mesh-free tools for scientific data reconstruction tasks in fluid mechanics, potentially advancing toward foundation models for engineering applications without explicit physical constraints.

major comments (2)
  1. [Abstract] Abstract: the central claim of 'competitive reconstruction accuracy' on four datasets with <10% observed data is asserted without any quantitative metrics, baselines, error bars, or ablation studies, preventing assessment of whether the seq2seq LM approach actually outperforms or matches existing methods.
  2. [Abstract] Abstract: the assertion that the language model 'effectively captur[es] spatial correlations and long-range dependencies' without a mesh or explicit physical constraints lacks discussion of how coordinate information is encoded or whether the model preserves key properties (e.g., divergence-free structure in vortex-street or jet cases); standard attention may rely solely on data patterns, risking non-physical outputs.
minor comments (1)
  1. [Abstract] Abstract: the four benchmark datasets are named but lack citations to their original sources or simulation parameters, which would aid reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their detailed review and constructive feedback on the abstract. We address each major comment below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim of 'competitive reconstruction accuracy' on four datasets with <10% observed data is asserted without any quantitative metrics, baselines, error bars, or ablation studies, preventing assessment of whether the seq2seq LM approach actually outperforms or matches existing methods.

    Authors: We agree that the abstract would be strengthened by including quantitative support. The full manuscript reports relative L2 errors (with standard deviations from multiple runs), direct comparisons to baselines including kriging, RBF interpolation, and other operator learning methods (FNO, DeepONet variants), and ablation studies on sparsity levels and model components in Sections 4.1–4.4 and the supplementary material. We will revise the abstract to include representative quantitative metrics (e.g., average relative errors across the four benchmarks at <10% observation) while keeping it concise. revision: yes

  2. Referee: [Abstract] Abstract: the assertion that the language model 'effectively captur[es] spatial correlations and long-range dependencies' without a mesh or explicit physical constraints lacks discussion of how coordinate information is encoded or whether the model preserves key properties (e.g., divergence-free structure in vortex-street or jet cases); standard attention may rely solely on data patterns, risking non-physical outputs.

    Authors: The abstract is space-constrained, but Section 3.1 details the coordinate encoding: spatial coordinates are tokenized via sinusoidal positional encodings and concatenated with measurement values in the input sequence, enabling the transformer to learn spatial relationships without an explicit mesh. The approach is intentionally data-driven without physics constraints to maintain generality across benchmarks. On the vortex-street and turbulent-jet cases, visualizations and quantitative metrics in Section 4 demonstrate preservation of coherent structures (vortex shedding, jet spreading); we do not claim strict divergence-free enforcement. We will add a short clause to the abstract noting the coordinate-aware tokenization. We acknowledge that data-driven models can in principle produce non-physical outputs, though this was not observed in our evaluations. revision: yes

Circularity Check

0 steps flagged

No significant circularity; claims rest on external benchmarks

full rationale

The paper reformulates sparse flow reconstruction as a sequence-to-sequence task using a language-model architecture and reports competitive accuracy on four independent benchmark datasets (vortex street, US temperature, blood flow, turbulent jet). No equations, self-citations, or fitted parameters are shown that reduce the central performance claims to the inputs by construction. The derivation chain is self-contained against external data and does not invoke uniqueness theorems or ansatzes from prior author work as load-bearing justification.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No concrete free parameters, axioms, or invented entities can be extracted from the abstract alone.

pith-pipeline@v0.9.0 · 5733 in / 1046 out tokens · 27631 ms · 2026-05-25T02:22:42.680508+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

46 extracted references · 46 canonical work pages · 4 internal anchors

  1. [1]

    Andersson, Andrew El-Kadi, Dominic Masters, Timo Ewalds, Jacklynn Stott, Shakir Mohamed, Peter Battaglia, Remi Lam, and Matthew Willson

    Ilan Price, Alvaro Sanchez-Gonzalez, Ferran Alet, Tom R. Andersson, Andrew El-Kadi, Dominic Masters, Timo Ewalds, Jacklynn Stott, Shakir Mohamed, Peter Battaglia, Remi Lam, and Matthew Willson. Proba- bilistic weather forecasting with machine learning.Nature, Dec 2024

    work page 2024

  2. [2]

    Alhussein Fawzi, Matej Balog, Aja Huang, Thomas Hubert, Bernardino Romera-Paredes, Mohammadamin Barekatain, Alexander Novikov, Francisco J. R. Ruiz, Julian Schrittwieser, Grzegorz Swirszcz, David Silver, Demis Hassabis, and Pushmeet Kohli. Discovering faster matrix multiplication algorithms with reinforcement learning.Nature, 610(7930):47–53, 2022

    work page 2022

  3. [3]

    Physics-informed neural networks (pinns) for fluid mechanics: A review.Acta Mechanica Sinica, 37(12):1727–1738, 2021

    Shengze Cai, Zhiping Mao, Zhicheng Wang, Minglang Yin, and George Em Karniadakis. Physics-informed neural networks (pinns) for fluid mechanics: A review.Acta Mechanica Sinica, 37(12):1727–1738, 2021

    work page 2021

  4. [4]

    Solving the quantum many-body problem with artificial neural networks.Science, 355(6325):602–606, 2017

    Giuseppe Carleo and Matthias Troyer. Solving the quantum many-body problem with artificial neural networks.Science, 355(6325):602–606, 2017

    work page 2017

  5. [5]

    Sympnets: Intrinsic structure-preserving symplectic networks for identifying hamiltonian systems.Neural Networks, 132:166– 179, 2020

    Pengzhan Jin, Zhen Zhang, Aiqing Zhu, Yifa Tang, and George Em Karniadakis. Sympnets: Intrinsic structure-preserving symplectic networks for identifying hamiltonian systems.Neural Networks, 132:166– 179, 2020

    work page 2020

  6. [6]

    Sms: Spiking marching scheme for efficient long time integration of differential equations.Journal of Computational Physics, 516:113363, 2024

    Qian Zhang, Adar Kahana, George Em Karniadakis, and Panos Stinis. Sms: Spiking marching scheme for efficient long time integration of differential equations.Journal of Computational Physics, 516:113363, 2024

    work page 2024

  7. [7]

    Artificial to spiking neural networks conversion with calibration in scientific machine learning.SIAM Journal on Scientific Computing, 47(3):C559–C577, 2025

    Qian Zhang, Chenxi Wu, Adar Kahana, George Em Karniadakis, Youngeun Kim, Yuhang Li, and Priyadarshini Panda. Artificial to spiking neural networks conversion with calibration in scientific machine learning.SIAM Journal on Scientific Computing, 47(3):C559–C577, 2025

    work page 2025

  8. [8]

    Theilman, Qian Zhang, Adar Kahana, Eric C

    Bradley H. Theilman, Qian Zhang, Adar Kahana, Eric C. Cyr, Nathaniel Trask, James B. Aimone, and George Em Karniadakis. Spiking physics-informed neural networks on loihi 2.2024 Neuro Inspired Compu- tational Elements Conference (NICE), 2024

    work page 2024

  9. [9]

    Chenxi Wu, Min Zhu, Qinyang Tan, Yadhu Kartha, and Lu Lu. A comprehensive study of non-adaptive and residual-based adaptive sampling for physics-informed neural networks.Computer Methods in Applied Mechanics and Engineering, 403:115671, 2023

    work page 2023

  10. [10]

    Randomized forward mode gradient for spiking neural networks in scientific machine learning.arXiv:2411.07057, 2024

    Ruyin Wan, Qian Zhang, and George Em Karniadakis. Randomized forward mode gradient for spiking neural networks in scientific machine learning.arXiv:2411.07057, 2024

    work page arXiv 2024

  11. [11]

    Highly accurate protein structure prediction with alphafold.Nature, 596(7873):583–589, 2021

    John Jumper, Richard Evans, Alexander Pritzel, Tim Green, Michael Figurnov, Olaf Ronneberger, Kathryn Tunyasuvunakool, Russ Bates, Augustin ˇZ´ ıdek, Anna Potapenko, et al. Highly accurate protein structure prediction with alphafold.Nature, 596(7873):583–589, 2021

    work page 2021

  12. [12]

    Scaling deep learning for materials discovery.Nature, 624(80):80–85, 2023

    Amil Merchant, Simon Batzner, Samuel S Schoenholz, Muratahan Aykol, Gowoon Cheon, and Ekin Dogus Cubuk. Scaling deep learning for materials discovery.Nature, 624(80):80–85, 2023

    work page 2023

  13. [13]

    Laplace neural operator for solving differ- ential equations.Nature Machine Intelligence, 6(6):631–640, 2024

    Qianying Cao, Somdatta Goswami, and George Em Karniadakis. Laplace neural operator for solving differ- ential equations.Nature Machine Intelligence, 6(6):631–640, 2024

    work page 2024

  14. [14]

    Systems biology: Identifiability analysis and parameter identification via systems-biology-informed neural networks

    Mitchell Daneker, Zhen Zhang, George Em Karniadakis, and Lu Lu. Systems biology: Identifiability analysis and parameter identification via systems-biology-informed neural networks. InComputational Modeling of Signaling Networks, pages 87–105. Springer, 2023

    work page 2023

  15. [15]

    Sun, and George Em Karniadakis

    Qian Zhang, Konstantina Sampani, Mengjia Xu, Shengze Cai, Yixiang Deng, He Li, Jennifer K. Sun, and George Em Karniadakis. AOSLO-net: A Deep Learning-Based Method for Automatic Segmentation of Reti- nal Microaneurysms From Adaptive Optics Scanning Laser Ophthalmoscopy Images.Translational Vision Science & Technology, 11(8):7–7, 08 2022. 15

    work page 2022

  16. [16]

    Sheng Zhang, Joan Ponce, Zhen Zhang, Guang Lin, and George Karniadakis. An integrated framework for building trustworthy data-driven epidemiological models: Application to the covid-19 outbreak in new york city.PLOS Computational Biology, 17(9):1–29, 09 2021

    work page 2021

  17. [17]

    Ehsan Kharazmi, Min Cai, Xiaoning Zheng, Zhen Zhang, Guang Lin, and George Em Karniadakis. Iden- tifiability and predictability of integer- and fractional-order epidemiological models using physics-informed neural networks.Nature Computational Science, 1(11):744–753, 2021

    work page 2021

  18. [18]

    Learning nonlin- ear operators via deeponet based on the universal approximation theorem of operators.Nature Machine Intelligence, 3(3):218–229, 2021

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

    work page 2021

  19. [19]

    A physics-informed variational deeponet for predicting crack path in quasi-brittle materials.Computer Methods in Applied Mechanics and Engineering, 391:114587, 2022

    Somdatta Goswami, Minglang Yin, Yue Yu, and George Em Karniadakis. A physics-informed variational deeponet for predicting crack path in quasi-brittle materials.Computer Methods in Applied Mechanics and Engineering, 391:114587, 2022

    work page 2022

  20. [20]

    Spiking neural operators for scientific machine learning.arXiv:2205.10130, 2022

    Adar Kahana, Qian Zhang, Leonard Gleyzer, and George Em Karniadakis. Spiking neural operators for scientific machine learning.arXiv:2205.10130, 2022

    work page arXiv 2022

  21. [21]

    Blending neural operators and relaxation methods in pde numerical solvers.Nature Machine Intelligence, pages 1–11, 2024

    Enrui Zhang, Adar Kahana, Alena Kopaniˇ c´ akov´ a, Eli Turkel, Rishikesh Ranade, Jay Pathak, and George Em Karniadakis. Blending neural operators and relaxation methods in pde numerical solvers.Nature Machine Intelligence, pages 1–11, 2024

    work page 2024

  22. [22]

    Fourier Neural Operator for Parametric Partial Differential Equations

    Zongyi Li, Nikola Kovachki, Kamyar Azizzadenesheli, Burigede Liu, Kaushik Bhattacharya, Andrew Stuart, and Anima Anandkumar. Fourier neural operator for parametric partial differential equations. arXiv:2010.08895, 2020

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  23. [23]

    Vito: Vision transformer-operator.Computer Methods in Applied Mechanics and Engineering, 428:117109, 2024

    Oded Ovadia, Adar Kahana, Panos Stinis, Eli Turkel, Dan Givoli, and George Em Karniadakis. Vito: Vision transformer-operator.Computer Methods in Applied Mechanics and Engineering, 428:117109, 2024

    work page 2024

  24. [24]

    Brunton, J

    Krithika Manohar, Bingni W. Brunton, J. Nathan Kutz, and Steven L. Brunton. Data-driven sparse sensor placement for reconstruction: Demonstrating the benefits of exploiting known patterns.IEEE Control Systems Magazine, 38(3):63–86, 2018

    work page 2018

  25. [25]

    Benjamin Erichson, Lionel Mathelin, Zhewei Yao, Steven L

    N. Benjamin Erichson, Lionel Mathelin, Zhewei Yao, Steven L. Brunton, Michael W. Mahoney, and J. Nathan Kutz. Shallow neural networks for fluid flow reconstruction with limited sensors.Proceedings of the Royal Society A, 476(2239):20200097, 2020

    work page 2020

  26. [26]

    Global field recon- struction from sparse sensors with Voronoi tessellation-assisted deep learning.Nature Machine Intelligence, 3(11):945–951, 2021

    Kai Fukami, Romit Maulik, Nesar Ramachandra, Koji Fukagata, and Kunihiko Taira. Global field recon- struction from sparse sensors with Voronoi tessellation-assisted deep learning.Nature Machine Intelligence, 3(11):945–951, 2021

    work page 2021

  27. [27]

    Kosiorek, Seungjin Choi, and Yee Whye Teh

    Juho Lee, Yoonho Lee, Jungtaek Kim, Adam R. Kosiorek, Seungjin Choi, and Yee Whye Teh. Set transformer: A framework for attention-based permutation-invariant neural networks. InInternational Conference on Machine Learning, 2019

    work page 2019

  28. [28]

    Perceiver IO: A General Architecture for Structured Inputs & Outputs

    Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, Olivier H´ enaff, Matthew M. Botvinick, Andrew Zisserman, Oriol Vinyals, and Jo˜ ao Carreira. Perceiver IO: A general architecture for structured inputs & outputs.arXiv:2107.14795, 2021

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  29. [29]

    Torr, and Vladlen Koltun

    Hengshuang Zhao, Li Jiang, Jiaya Jia, Philip H.S. Torr, and Vladlen Koltun. Point transformer. In IEEE/CVF International Conference on Computer Vision, 2021

    work page 2021

  30. [30]

    A physics-informed diffusion model for high-fidelity flow field reconstruction.Journal of Computational Physics, 478:111972, 2023

    Dule Shu, Zijie Li, and Amir Barati Farimani. A physics-informed diffusion model for high-fidelity flow field reconstruction.Journal of Computational Physics, 478:111972, 2023

    work page 2023

  31. [31]

    Operator learning for reconstructing flow fields from sparse measurements: an energy transformer approach.arXiv:2501.08339, 2025

    Qian Zhang, Dmitry Krotov, and George Em Karniadakis. Operator learning for reconstructing flow fields from sparse measurements: an energy transformer approach.arXiv:2501.08339, 2025. 16

    work page arXiv 2025

  32. [32]

    GPT-4 Technical Report

    OpenAI. GPT-4 technical report.arXiv:2303.08774, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  33. [33]

    DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning

    DeepSeek-AI. DeepSeek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning. arXiv:2501.12948, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  34. [34]

    In-context operator learning with data prompts for differential equation problems.Proceedings of the National Academy of Sciences, 120(39):e2310142120, 2023

    Liu Yang, Siting Liu, Tingwei Meng, and Stanley J Osher. In-context operator learning with data prompts for differential equation problems.Proceedings of the National Academy of Sciences, 120(39):e2310142120, 2023

    work page 2023

  35. [35]

    Fine-tune language models as multi-modal differential equation solvers.Neural Networks, page 107455, 2025

    Liu Yang, Siting Liu, and Stanley J Osher. Fine-tune language models as multi-modal differential equation solvers.Neural Networks, page 107455, 2025

    work page 2025

  36. [36]

    Pde generalization of in-context operator networks: A study on 1d scalar nonlinear conservation laws.Journal of Computational Physics, 519:113379, 2024

    Liu Yang and Stanley J Osher. Pde generalization of in-context operator networks: A study on 1d scalar nonlinear conservation laws.Journal of Computational Physics, 519:113379, 2024

    work page 2024

  37. [37]

    Vicon: Vision in-context operator networks for multi-physics fluid dynamics prediction.arXiv:2411.16063, 2024

    Yadi Cao, Yuxuan Liu, Liu Yang, Rose Yu, Hayden Schaeffer, and Stanley Osher. Vicon: Vision in-context operator networks for multi-physics fluid dynamics prediction.arXiv:2411.16063, 2024

    work page arXiv 2024

  38. [38]

    Maddison, Tiago Ramalho, David Saxton, Murray Shana- han, Yee Whye Teh, Danilo J

    Marta Garnelo, Dan Rosenbaum, Christopher J. Maddison, Tiago Ramalho, David Saxton, Murray Shana- han, Yee Whye Teh, Danilo J. Rezende, and S. M. Ali Eslami. Conditional neural processes. InInternational Conference on Machine Learning, 2018

    work page 2018

  39. [39]

    Attentive neural processes

    Hyunjik Kim, Andriy Mnih, Jonathan Schwarz, Marta Garnelo, Ali Eslami, Dan Rosenbaum, Oriol Vinyals, and Yee Whye Teh. Attentive neural processes. InInternational Conference on Learning Representations, 2019

    work page 2019

  40. [40]

    D. T. Lee and B. J. Schachter. Two algorithms for constructing a Delaunay triangulation.International Journal of Computer & Information Sciences, 9:219–242, 1980

    work page 1980

  41. [41]

    M. A. Oliver and R. Webster. Kriging: a method of interpolation for geographical information systems. International Journal of Geographical Information Systems, 4(3):313–332, 1990

    work page 1990

  42. [42]

    Karhunen–loeve procedure for gappy data.Journal of the Optical Society of America A, 12(8):1657–1664, 1995

    Richard Everson and Lawrence Sirovich. Karhunen–loeve procedure for gappy data.Journal of the Optical Society of America A, 12(8):1657–1664, 1995

    work page 1995

  43. [43]

    Transformer neural processes: Uncertainty-aware meta learning via se- quence modeling

    Tung Nguyen and Aditya Grover. Transformer neural processes: Uncertainty-aware meta learning via se- quence modeling. InProceedings of the 39th International Conference on Machine Learning, volume 162 of Proceedings of Machine Learning Research, pages 16569–16594. PMLR, 2022

    work page 2022

  44. [44]

    Attention is all you need.Advances in Neural Information Processing Systems, 2017

    A Vaswani. Attention is all you need.Advances in Neural Information Processing Systems, 2017

    work page 2017

  45. [45]

    Language models are unsupervised multitask learners.OpenAI blog, 1(8):9, 2019

    Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al. Language models are unsupervised multitask learners.OpenAI blog, 1(8):9, 2019

    work page 2019

  46. [46]

    Physics-informed neural networks enhanced par- ticle tracking velocimetry: An example for turbulent jet flow.IEEE Transactions on Instrumentation and Measurement, 2024

    Shengze Cai, Callum Gray, and George Em Karniadakis. Physics-informed neural networks enhanced par- ticle tracking velocimetry: An example for turbulent jet flow.IEEE Transactions on Instrumentation and Measurement, 2024. 17 A Baseline Methods and Hyperparameters All baseline experiments use the same data interface as RFormer, with observed coordinates an...

    work page 2024