pith. sign in

arxiv: 2606.19562 · v1 · pith:VAM6YCXUnew · submitted 2026-06-17 · 💻 cs.LG · physics.flu-dyn

Advances in Scientific Machine Learning for Coupled Fluid Flow and Transport

Gabriel F. Barros , R\^omulo M. Silva , Alvaro L. G. A. Coutinho This is my paper

Pith reviewed 2026-06-26 20:51 UTC · model grok-4.3

classification 💻 cs.LG physics.flu-dyn
keywords Scientific Machine LearningPhysics-Informed Neural Networksβ-Variational AutoencodersTurbidity CurrentsRayleigh-Bénard ConvectionSurrogate ModelingNavier-Stokes EquationsReduced-Order Models
0
0 comments X

The pith

Scientific machine learning builds fast surrogate models for coupled fluid flow and transport within tested regimes.

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

The paper reviews recent SciML methods for approximating systems governed by the incompressible Navier-Stokes equations coupled to scalar transport. These systems arise in turbidity currents and thermal convection and are expensive to simulate at high fidelity because of nonlinear coupling and multiscale effects. The review covers linear reduced-order techniques such as Dynamic Mode Decomposition and nonlinear approaches including Physics-Informed Neural Networks and β-Variational Autoencoders, together with their combination with adaptive mesh refinement and data compression. It presents two concrete contributions: PINN-based surrogates for lock-exchange turbidity currents and β-VAE extraction of disentangled modes from Rayleigh-Bénard convection. Within the data regimes and modeling assumptions examined, the methods produce accurate approximations at substantially lower cost than full-order simulations.

Core claim

The chapter demonstrates that SciML methods, including PINNs for surrogate modeling of turbidity currents and β-VAEs for disentangled nonlinear modes from thermal flows, enable fast and accurate approximations of coupled incompressible Navier-Stokes and scalar transport equations, substantially lowering computational costs relative to traditional full-order simulations when applied inside the specific data regimes considered.

What carries the argument

Physics-Informed Neural Networks (PINNs) and β-Variational Autoencoders (β-VAEs) combined with adaptive mesh refinement/coarsening and scientific data compression, used to construct reduced-order surrogates of the Navier-Stokes and transport system.

If this is right

  • PINN surrogates can reproduce lock-exchange turbidity current dynamics at reduced cost.
  • β-VAEs can extract disentangled nonlinear modes from Rayleigh-Bénard convection simulations.
  • The combined SciML and HPC workflow lowers computational expense for coupled flow-transport problems relative to full-order models.
  • Accuracy holds for the nonlinear coupling and multiscale features present in the studied benchmarks.

Where Pith is reading between the lines

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

  • Real-time prediction capability is stated to depend strongly on the concrete problem and data regime.
  • Uncertainty quantification for these surrogates is identified as an active research direction rather than a delivered capability.
  • The same surrogate-construction pattern could be tested on other coupled transport problems that share the same governing equations.

Load-bearing premise

The surrogate models remain accurate and useful only inside the specific data regimes and modeling assumptions considered in the two example problems.

What would settle it

Compare PINN surrogate predictions for turbidity currents against full-order Navier-Stokes solutions on parameter values lying outside the training ranges used in the lock-exchange benchmarks.

Figures

Figures reproduced from arXiv: 2606.19562 by Alvaro L. G. A. Coutinho, Gabriel F. Barros, R\^omulo M. Silva.

Figure 1
Figure 1. Figure 1: Lock-exchange problem at 𝑆𝑐 = 1.0 and 𝐺𝑟 = 5 × 106 . (a) Schematic of the computational domain showing the boundary conditions: no-slip walls for velocity (u = 0) on all boundaries and zero-flux conditions for sediment concentration. The heavier fluid (𝜌2, with 𝑐 = 1) initially occupies the left column of dimensions 𝐻 × 𝑆, while the lighter fluid (𝜌1, with 𝑐 = 0) fills the remainder. (b) Temporal evolution… view at source ↗
Figure 2
Figure 2. Figure 2: Buoyancy plots for Rayleigh-Benard convection at different values of ´ 𝑅𝑎 and 𝑃𝑟 = 1.0. The top row shows a late-time snapshot at 𝑅𝑎 = 106 , where the flow exhibits organized large-scale convective rolls with relatively smooth temperature gradients. The distinction between these values is important because it directly impacts the feasibility of reduced-order modeling: the slow singular-value decay observed… view at source ↗
Figure 4
Figure 4. Figure 4: Visualization of POD spatial modes for the 2D lock-exchange problem. From top [PITH_FULL_IMAGE:figures/full_fig_p016_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Singular values for the Y1 matrix containing the 2D lock-exchange snapshots and the values for 𝜅 for each 𝑟 = {50, 100, 150, 200, 250}. Singular values were limited to a maximum of 𝑟 = 400 for proper visualization. The rapid initial decay (roughly one order of magnitude for 𝑟 ≤ 25) indicates that most of the flow’s energy is concentrated in a small number of dominant modes. The retained energy 𝜅 quantifies… view at source ↗
Figure 7
Figure 7. Figure 7: Singular values, 𝜅 and relative Frobenius norm between the truncated approxi￾mation and the ground truth data for the Rayleigh-Benard dataset. ´ [PITH_FULL_IMAGE:figures/full_fig_p020_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: Geometry and initial conditions for the 3D lock-exchange simulation (top) and [PITH_FULL_IMAGE:figures/full_fig_p022_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: This figure contrasts the singular value decay for the uncompressed baseline data [PITH_FULL_IMAGE:figures/full_fig_p023_10.png] view at source ↗
Figure 10
Figure 10. Figure 10: Singular values for the tested cases. The dashed line represents the number of [PITH_FULL_IMAGE:figures/full_fig_p024_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: DMD eigenvalues for the 3D lock-exchange simulation plotted in the complex [PITH_FULL_IMAGE:figures/full_fig_p025_11.png] view at source ↗
Figure 13
Figure 13. Figure 13: Schematic of a PINN used to solve a 2D density-driven gravity flow prob [PITH_FULL_IMAGE:figures/full_fig_p029_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Mean predicted concentration at three different times and datasets. [PITH_FULL_IMAGE:figures/full_fig_p031_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Mean predicted 𝑥−velocity at three different times and datasets. 0 y t = 0 s Dataset 1 t = 15 s t = 30 s 0 Dataset 2 y 0.0 4.5 9.0 13.5 18.0 x 0 Dataset 3 y 0.0 4.5 9.0 13.5 18.0 x 0.0 4.5 9.0 13.5 18.0 x −0.629 −0.329 −0.029 0.272 0.572 [PITH_FULL_IMAGE:figures/full_fig_p031_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Mean predicted 𝑦−velocity at three different times and datasets. 0 y t = 0 s Dataset 1 t = 15 s t = 30 s 0 Dataset 2 y 0.0 4.5 9.0 13.5 18.0 x 0 Dataset 3 y 0.0 4.5 9.0 13.5 18.0 x 0.0 4.5 9.0 13.5 18.0 x 0.00 0.25 0.50 0.75 1.00 [PITH_FULL_IMAGE:figures/full_fig_p031_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Mean pointwise error for the concentration at three different times and datasets. [PITH_FULL_IMAGE:figures/full_fig_p031_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: Mean pointwise error for the 𝑥−velocity at three different times and datasets. 0 y t = 0 s Dataset 1 t = 15 s t = 30 s 0 Dataset 2 y 0.0 4.5 9.0 13.5 18.0 x 0 Dataset 3 y 0.0 4.5 9.0 13.5 18.0 x 0.0 4.5 9.0 13.5 18.0 x 0.0000 0.1081 0.2162 0.3243 0.4324 [PITH_FULL_IMAGE:figures/full_fig_p032_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: Mean pointwise error for the 𝑦−velocity at 3 different times and datasets. 0 100000 200000 300000 400000 Iterations 100 2 × 10−1 3 × 10−1 4 × 10−1 6 × 10−1 Relative L 2 Error x−velocity Error 0 100000 200000 300000 400000 Iterations 100 2 × 10−1 3 × 10−1 4 × 10−1 6 × 10−1 y−velocity Error 0 100000 200000 300000 400000 Iterations 10−2 10−1 Concentration Error Dataset 1 - Mean Dataset 1 - Mean ± Std Dataset… view at source ↗
Figure 20
Figure 20. Figure 20: 𝐿 2 error for multiple runs using the datasets in [PITH_FULL_IMAGE:figures/full_fig_p032_20.png] view at source ↗
Figure 21
Figure 21. Figure 21: SVD-reconstruction of the concentration using different mode-truncation cri [PITH_FULL_IMAGE:figures/full_fig_p033_21.png] view at source ↗
Figure 22
Figure 22. Figure 22: Evolution of the collocation points for two different sampling techniques. [PITH_FULL_IMAGE:figures/full_fig_p034_22.png] view at source ↗
Figure 23
Figure 23. Figure 23: Snapshots of the reconstruction of the entire concentration. [PITH_FULL_IMAGE:figures/full_fig_p035_23.png] view at source ↗
Figure 24
Figure 24. Figure 24: Snapshots of the reconstruction of the 𝑢−velocity. differential equations (𝑤𝑒1 , 𝑤𝑒2 , and 𝑤𝑒3 ), all methods exhibit similar behavior, except for the MC+RAR scheme for the momentum equations, where it starts to show a growing for the weights until it becomes constant after some point. Regarding the data-related dynamic loss weights, the MC and MC+RAR schemes show higher weights as training progresses, bu… view at source ↗
Figure 25
Figure 25. Figure 25: Concentration (left), 𝑢−velocity (middle), and 𝑣−velocity (right) errors over time. 10−2 Total Loss 10−2 Data Loss 10−3 PDE Loss 0 200000 400000 Epoch 10−1 Concentration Error 0 200000 400000 Epoch 10−1 100 u−velocity Error 0 200000 400000 Epoch 10−1 100 v−velocity Error Fixed MC RAR MC + RAR (a) The loss components and errors are shown for all the sampling techniques for a case where the concentration ou… view at source ↗
Figure 26
Figure 26. Figure 26: Evolution of the losses, errors, and dynamic loss weights for different collocation [PITH_FULL_IMAGE:figures/full_fig_p037_26.png] view at source ↗
Figure 27
Figure 27. Figure 27: Snapshots showing the selected region for the training dataset used to estimate [PITH_FULL_IMAGE:figures/full_fig_p039_27.png] view at source ↗
Figure 28
Figure 28. Figure 28: Convergence of the Grashof number (left), loss function (middle) and relative error (right) for several runs. Section Summary This section demonstrated the application of Physics-Informed Neural Networks to density-driven gravity flows, addressing both field reconstruction from sparse data and inverse parameter estimation. For the forward reconstruction problem, we showed that PINNs can recover velocity, … view at source ↗
Figure 29
Figure 29. Figure 29: Variational Autoencoder (VAE) architecture. The encoder maps the high [PITH_FULL_IMAGE:figures/full_fig_p041_29.png] view at source ↗
Figure 30
Figure 30. Figure 30: Comparison of reconstructed fields and error metrics across different Rayleigh [PITH_FULL_IMAGE:figures/full_fig_p043_30.png] view at source ↗
read the original abstract

This chapter reviews recent advances in Scientific Machine Learning (SciML) for modeling coupled fluid flow and transport phenomena governed by the incompressible Navier-Stokes and scalar transport equations. Such systems, found in applications like turbidity currents and thermal convection, feature strong nonlinear coupling and multiscale behavior that make high-fidelity simulations computationally expensive. To address this, the chapter surveys state-of-the-art SciML methods for building efficient surrogate models, including linear reduced-order techniques based on Singular Value Decomposition (such as Dynamic Mode Decomposition) and nonlinear neural network approaches like Physics-Informed Neural Networks (PINNs) and $\beta$-Variational Autoencoders ($\beta$-VAEs). It first covers the authors' work combining these models with High Performance Computing strategies, including Adaptive Mesh Refinement/Coarsening (AMR/C) and scientific floating-point data compression. It then presents two new contributions: surrogate modeling of turbidity currents via PINNs, and the extraction of disentangled nonlinear modes from thermal flows using $\beta$-VAEs. Governing equations and representative benchmarks, including lock-exchange flows and Rayleigh-B\'enard convection, illustrate these methodologies. The chapter is intentionally long, covering both the mathematical and physical foundations of coupled fluid flow and the computational aspects of state-of-the-art modeling. Overall, it demonstrates how SciML enables fast, accurate approximations of complex coupled systems within the specific data regimes and modeling assumptions considered, while substantially reducing computational cost relative to full-order simulations. Broader capabilities such as real-time prediction and uncertainty quantification remain active research directions whose feasibility depends strongly on the problem at hand.

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

1 major / 0 minor

Summary. This chapter reviews SciML methods (DMD, PINNs, β-VAEs) for surrogate modeling of incompressible Navier-Stokes plus scalar transport problems, with applications to turbidity currents and Rayleigh-Bénard convection. It integrates these with HPC techniques (AMR/C, data compression) and presents two new contributions: a PINN surrogate for lock-exchange turbidity currents and a β-VAE for disentangled nonlinear mode extraction in thermal convection. The central claim is that the resulting surrogates deliver fast, accurate approximations inside the specific data regimes and modeling assumptions considered, while substantially lowering cost relative to full-order simulations; broader capabilities such as real-time prediction and UQ are noted as problem-dependent.

Significance. If the quantitative validations and cost comparisons for the two new contributions hold, the chapter would supply a useful, self-contained survey of SciML techniques for strongly coupled multiscale fluid-transport systems together with concrete HPC integrations. The explicit scoping to considered regimes avoids over-claiming generality.

major comments (1)
  1. [Abstract] Abstract: the statements that the PINN and β-VAE surrogates are 'fast and accurate' and 'substantially reducing computational cost' are presented without any quantitative error metrics, validation protocols, baseline comparisons, or timing data. Because these two new contributions are the primary novel elements, the absence of such evidence is load-bearing for the central claim.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their thoughtful review and for highlighting the need to strengthen the abstract's presentation of our new contributions. We address the single major comment below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the statements that the PINN and β-VAE surrogates are 'fast and accurate' and 'substantially reducing computational cost' are presented without any quantitative error metrics, validation protocols, baseline comparisons, or timing data. Because these two new contributions are the primary novel elements, the absence of such evidence is load-bearing for the central claim.

    Authors: We agree that the abstract should include quantitative support for the claims regarding the PINN turbidity-current surrogate and the β-VAE mode extraction. The body of the manuscript (Sections 4.2 and 5.3) already reports relative L2 errors, validation against high-fidelity AMR/C simulations, and wall-clock timing comparisons demonstrating order-of-magnitude speedups within the considered regimes. We will revise the abstract to incorporate concise quantitative statements (e.g., typical error levels and observed speedups) together with a brief reference to the validation protocols, while preserving the existing scoping language that limits generality claims. revision: yes

Circularity Check

0 steps flagged

Minor self-citations in review context; no load-bearing circularity

full rationale

This is a review chapter surveying SciML methods for fluid flow and transport, with two scoped example applications (PINN surrogate for turbidity currents; β-VAE for Rayleigh-Bénard). The central claim is explicitly limited to accuracy and cost reduction 'within the specific data regimes and modeling assumptions considered,' and flags that broader capabilities 'depend strongly on the problem at hand.' No derivation reduces by construction to fitted inputs, no uniqueness theorem is imported from self-citation, and self-citations to prior author work are not load-bearing for the scoped demonstration. The paper is self-contained against external benchmarks within its stated scope.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The document is a review chapter; no new free parameters, axioms, or invented entities are introduced beyond the standard governing equations and SciML architectures already present in the cited literature.

pith-pipeline@v0.9.1-grok · 5829 in / 1125 out tokens · 22932 ms · 2026-06-26T20:51:06.946056+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

80 extracted references · 35 canonical work pages

  1. [1]

    Research in Engineering Design31(3), 275–298 (2020)

    Alizadeh, R., Allen, J.K., Mistree, F.: Managing computational complexity using surrogate models: a critical review. Research in Engineering Design31(3), 275–298 (2020). DOI 10.1007/s00163- 020-00336-7

    work page doi:10.1007/s00163- 2020

  2. [2]

    Proceedings of the Royal Society of London

    Allen, C.M.: Numerical simulation of contaminant dispersion in estuary flows. Proceedings of the Royal Society of London. A. Mathematical and Physical Sciences381(1780), 179–194 (1982)

    1982

  3. [3]

    Nat Methods15(6), 399–400 (2018)

    Altman, N., Krzywinski, M.: The curse (s) of dimensionality. Nat Methods15(6), 399–400 (2018)

    2018

  4. [4]

    In: High Performance Computing in Science and Engineering´ 16: Transac- tions of the High Performance Computing Center, Stuttgart (HLRS) 2016, pp

    Axtmann, G., Rist, U.: Scalability of openfoam with large eddy simulations and dns on high- performance systems. In: High Performance Computing in Science and Engineering´ 16: Transac- tions of the High Performance Computing Center, Stuttgart (HLRS) 2016, pp. 413–424. Springer (2017)

    2016

  5. [5]

    Physics of fluids17(10) (2005)

    Bae, J.H., Yoo, J.Y., Choi, H.: Direct numerical simulation of turbulent supercritical flows with heat transfer. Physics of fluids17(10) (2005)

    2005

  6. [6]

    Dissertation, Federal University of Rio de Janeiro (COPPE/UFRJ), Rio de Janeiro, Brazil (2023)

    Barros, G.F.: Contributions to the application of snapshot-based data-driven methods in computa- tional science and engineering. Dissertation, Federal University of Rio de Janeiro (COPPE/UFRJ), Rio de Janeiro, Brazil (2023). Awarded the 2024 ABMEC Best Doctoral Thesis Prize

    2023

  7. [7]

    CILAMCE 2020 - Proceedings of the XLI Ibero-Latin-American Congress on Computational Methods in Engineering (2020)

    Barros, G.F., C ˆortes, A.M.A., Coutinho, A.L.: Dynamic mode decomposition for density-driven gravity current simulations. CILAMCE 2020 - Proceedings of the XLI Ibero-Latin-American Congress on Computational Methods in Engineering (2020)

    2020

  8. [8]

    Engineering with Computers40(1), 455–476 (2024)

    Barros, G.F., Grave, M., Camata, J.J., Coutinho, A.L.: Enhancing dynamic mode decomposition workflow with in situ visualization and data compression. Engineering with Computers40(1), 455–476 (2024). DOI 10.1007/s00366-023-01805-y Advances in Scientific Machine Learning for Coupled Fluid Flow and Transport 47

    work page doi:10.1007/s00366-023-01805-y 2024

  9. [9]

    Engineering with Computers38(5), 4241–4268 (2022)

    Barros, G.F., Grave, M., Viguerie, A., Reali, A., Coutinho, A.L.: Dynamic mode decomposition in adaptive mesh refinement and coarsening simulations. Engineering with Computers38(5), 4241–4268 (2022). DOI 10.1007/s00366-021-01485-6

    work page doi:10.1007/s00366-021-01485-6 2022

  10. [10]

    Bischof, M

    Bischof, R., Kraus, M.A.: Multi-objective loss balancing for physics-informed deep learn- ing. Computer Methods in Applied Mechanics and Engineering439, 117914 (2025). DOI 10.1016/j.cma.2025.117914

    work page doi:10.1016/j.cma.2025.117914 2025

  11. [11]

    In: Proceedings of the 20th SIGNLL conference on computational natural language learning, pp

    Bowman, S., Vilnis, L., Vinyals, O., Dai, A., Jozefowicz, R., Bengio, S.: Generating sentences from a continuous space. In: Proceedings of the 20th SIGNLL conference on computational natural language learning, pp. 10–21 (2016)

    2016

  12. [12]

    arXiv preprint arXiv:1804.03599 (2018)

    Burgess, C.P., Higgins, I., Pal, A., Matthey, L., Watters, N., Desjardins, G., Lerchner, A.: Understanding disentangling in𝛽-V AE. arXiv preprint arXiv:1804.03599 (2018). URL https://arxiv.org/abs/1804.03599

    Pith/arXiv arXiv 2018

  13. [13]

    The International Journal of High Performance Computing Applications33(6), 1201–1220 (2019)

    Cappello, F., Di, S., Li, S., Liang, X., Gok, A.M., Tao, D., Yoon, C.H., Wu, X.C., Alexeev, Y., Chong, F.T.: Use cases of lossy compression for floating-point data in scientific data sets. The International Journal of High Performance Computing Applications33(6), 1201–1220 (2019)

    2019

  14. [14]

    CRC Press (1997)

    Carey, G.F.: Computational Grids: Generations, Adaptation & Solution Strategies. CRC Press (1997)

    1997

  15. [15]

    Journal of Fluid Mechanics204, 1–30 (1989)

    Castaing, B., Gunaratne, G., Heslot, F., Kadanoff, L., Libchaber, A., Thomae, S., Wu, X.Z., Zaleski, S., Zanetti, G.: Scaling of hard thermal turbulence in rayleigh-b´enard convection. Journal of Fluid Mechanics204, 1–30 (1989). DOI 10.1017/S0022112089001643

    work page doi:10.1017/s0022112089001643 1989

  16. [16]

    arXiv preprint arXiv:2603.05225 (2026)

    Chen, D., Cong, J., Mirhoseini, A., Kozyrakis, C., Mitra, S., Xiong, J., Young, C., Anandkumar, A., Littman, M., Kirschen, A., Shao, S., Leef, S., Shanbhag, N., Milojicic, D., Schulte, M., Cauwenberghs, G., Chow, J.M., Dao, T., Gopalakrishnan, K., Ho, R., Kim, H., Olukotun, K., Pan, D.Z., Ren, M., Roth, D., Singh, A., Sun, Y., Wang, Y., LeCun, Y., Puri, R...

    arXiv 2035

  17. [17]

    Engineering Applications of Computational Fluid Mechanics18(1), 2407005 (2024)

    Chen, X., Wang, Z., Deng, L., Yan, J., Gong, C., Yang, B., Wang, Q., Zhang, Q., Yang, L., Pang, Y., Liu, J.: Towards a new paradigm in intelligence-driven computational fluid dynamics simulations. Engineering Applications of Computational Fluid Mechanics18(1), 2407005 (2024). DOI 10.1080/19942060.2024.2407005

    work page doi:10.1080/19942060.2024.2407005 2024

  18. [18]

    arXiv preprint arXiv:1711.02257 (2017)

    Chen, Z., Badrinarayanan, V., Lee, C.Y., Rabinovich, A.: Gradnorm: Gradient normalization for adaptive loss balancing in deep multitask networks. arXiv preprint arXiv:1711.02257 (2017). URL https://arxiv.org/abs/1711.02257

    Pith/arXiv arXiv 2017

  19. [19]

    In: Stochastic processes and functional analysis, pp

    Chipman, J.S.: “proofs” and proofs of the eckart–young theorem. In: Stochastic processes and functional analysis, pp. 71–83. CRC Press (2020)

    2020

  20. [20]

    arXiv preprint arXiv:2505.22904 (2025)

    Choi, Y., Cheung, S.W., Kim, Y., Tsai, P.H., Diaz, A.N., Zanardi, I., Chung, S.W., Copeland, D.M., Kendrick, C., Anderson, W., Iliescu, T., Heinkenschloss, M.: Defining foundation models for computational science: A call for clarity and rigor. arXiv preprint arXiv:2505.22904 (2025). URL https://arxiv.org/abs/2505.22904

    arXiv 2025

  21. [21]

    In: Variational Methods for Machine Learning with Applications to Deep Networks

    Cinelli, L.P., Marins, M.A., da Silva, E.A.B., Netto, S.L.: Variational methods for machine learn- ing with applications to deep networks. In: Variational Methods for Machine Learning with Applications to Deep Networks. Springer (2021). DOI 10.1007/978-3-030-70679-1

    work page doi:10.1007/978-3-030-70679-1 2021

  22. [22]

    Colbrook, M.J.: Chapter 4 - the multiverse of dynamic mode decomposition algorithms. In: S. Mishra, A. Townsend (eds.) Numerical Analysis Meets Machine Learning,Handbook of Nu- merical Analysis, vol. 25, pp. 127–230. Elsevier (2024). DOI 10.1016/bs.hna.2024.05.004

    work page doi:10.1016/bs.hna.2024.05.004 2024

  23. [23]

    Scientific Reports14(1), 3826 (2024)

    Coscia, D., Demo, N., Rozza, G.: Generative adversarial reduced order modelling. Scientific Reports14(1), 3826 (2024). DOI 10.1038/s41598-024-54067-z

    work page doi:10.1038/s41598-024-54067-z 2024

  24. [24]

    Scientific machine learning through physics-informed neural networks: Where we are and what’s next

    Cuomo, S., Cola, V.S.D., Giampaolo, F., Rozza, G., Raissi, M., Piccialli, F.: Scientific machine learning through physics–informed neural networks: Where we are and what’s next. Journal of Scientific Computing92(3) (2022). DOI 10.1007/s10915-022-01939-z

    work page doi:10.1007/s10915-022-01939-z 2022

  25. [25]

    Applied Thermal Engineering112, 841–854 (2017)

    Dbouk, T.: A review about the engineering design of optimal heat transfer systems using topology optimization. Applied Thermal Engineering112, 841–854 (2017)

    2017

  26. [26]

    Chemical engineering science81, 329–344 (2012) 48 Gabriel F

    Deen, N.G., Kriebitzsch, S.H., van der Hoef, M.A., Kuipers, J.: Direct numerical simulation of flow and heat transfer in dense fluid–particle systems. Chemical engineering science81, 329–344 (2012) 48 Gabriel F. Barros, R ˆomulo M. Silva and Alvaro L. G. A. Coutinho

    2012

  27. [27]

    SIAM Journal on Numerical Analysis52(5), 2415–2439 (2014)

    Deteix, J., Jendoubi, A., Yakoubi, D.: A coupled prediction scheme for solving the navier–stokes and convection-diffusion equations. SIAM Journal on Numerical Analysis52(5), 2415–2439 (2014). DOI 10.1137/130942516

    work page doi:10.1137/130942516 2014

  28. [28]

    ACM computing surveys57(11), 1–38 (2025)

    Di, S., Liu, J., Zhao, K., Liang, X., Underwood, R., Zhang, Z., Shah, M., Huang, Y., Huang, J., Yu, X., et al.: A survey on error-bounded lossy compression for scientific datasets. ACM computing surveys57(11), 1–38 (2025)

    2025

  29. [29]

    Expert Systems with Applications202, 117038 (2022)

    Eivazi, H., Le Clainche, S., Hoyas, S., Vinuesa, R.: Towards extraction of orthogonal and parsi- monious non-linear modes from turbulent flows. Expert Systems with Applications202, 117038 (2022). DOI 10.1016/j.eswa.2022.117038

    work page doi:10.1016/j.eswa.2022.117038 2022

  30. [30]

    Gavish, M., Donoho, D.L.: The optimal hard threshold for singular values is 4/ √

  31. [31]

    DOI 10.1109/TIT.2014.2323359

    IEEE Trans- actions on Information Theory60(8), 5040–5053 (2014). DOI 10.1109/TIT.2014.2323359

    work page doi:10.1109/tit.2014.2323359 2014

  32. [32]

    International Journal for Numerical Methods in Fluids73(8), 729–749 (2013)

    Guerra, G.M., Zio, S., Camata, J.J., Rochinha, F.A., Elias, R.N., Paraizo, P.L., Coutinho, A.L.: Numerical simulation of particle-laden flows by the residual-based variational multiscale method. International Journal for Numerical Methods in Fluids73(8), 729–749 (2013)

    2013

  33. [33]

    SIAM (2018)

    Heath, M.T.: Scientific computing: an introductory survey, revised second edition. SIAM (2018)

    2018

  34. [34]

    arXiv preprint arXiv:1912.12355 (2019)

    Heydari, A.A., Thompson, C.A., Mehmood, A.: Softadapt: Techniques for adaptive loss weighting of neural networks with multi-part loss functions. arXiv preprint arXiv:1912.12355 (2019). URL https://arxiv.org/abs/1912.12355

    arXiv 1912

  35. [35]

    Springer Berlin Heidelberg (2009)

    Jiji, L.M.: Heat Convection. Springer Berlin Heidelberg (2009). DOI 10.1007/978-3-642-02971-4

    work page doi:10.1007/978-3-642-02971-4 2009

  36. [36]

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

    Jin, X., Cai, S., Li, H., Karniadakis, G.E.: Nsfnets (navier-stokes flow nets): Physics-informed neural networks for the incompressible navier-stokes equations. Journal of Computational Physics 426, 109951 (2021). DOI 10.1016/j.jcp.2020.109951

    work page doi:10.1016/j.jcp.2020.109951 2021

  37. [37]

    Nature Reviews Physics3(6), 422–440 (2021)

    Karniadakis, G.E., Kevrekidis, I.G., Lu, L., Perdikaris, P., Wang, S., Yang, L.: Physics-informed machine learning. Nature Reviews Physics3(6), 422–440 (2021)

    2021

  38. [38]

    Kingma, D.P., Salimans, T., Welling, M.: Variational dropout and the local reparameterization trick. In: C. Cortes, N. Lawrence, D. Lee, M. Sugiyama, R. Garnett (eds.) Advances in Neural Information Processing Systems, vol. 28. Curran Associates, Inc. (2015)

    2015

  39. [39]

    In: 2nd International Confer- ence on Learning Representations, ICLR 2014 - Conference Track Proceedings (2014)

    Kingma, D.P., Welling, M.: Auto-encoding variational bayes. In: 2nd International Confer- ence on Learning Representations, ICLR 2014 - Conference Track Proceedings (2014). DOI 10.61603/ceas.v2i1.33

    work page doi:10.61603/ceas.v2i1.33 2014

  40. [40]

    Foundations and Trends® in Machine Learning12(4), 307–392 (2019)

    Kingma, D.P., Welling, M.: An introduction to variational autoencoders. Foundations and Trends® in Machine Learning12(4), 307–392 (2019)

    2019

  41. [41]

    Journal of Machine Learning Research24, 1–97 (2023)

    Kovachki, N., Zongyi Li, N., Burigede Liu, C., Azizzadenesheli, K., Kaushik Bhattacharya, N., Stuart, A., Anandkumar, A., Editor, C., Rosasco, L., Li, Z., Liu, B., Bhattacharya, K.: Neural Operator: Learning Maps Between Function Spaces With Applications to PDEs. Journal of Machine Learning Research24, 1–97 (2023)

    2023

  42. [42]

    Visual Informatics5(3), 15–27 (2021)

    Krake, T., Reinhardt, S., Hlawatsch, M., Eberhardt, B., Weiskopf, D.: Visualization and selection of dynamic mode decomposition components for unsteady flow. Visual Informatics5(3), 15–27 (2021). DOI 10.1016/j.visinf.2021.06.003

    work page doi:10.1016/j.visinf.2021.06.003 2021

  43. [43]

    Journal of scientific computing3(1), 25–43 (1988)

    Leonard, A.D., Hill, J.C.: Direct numerical simulation of turbulent flows with chemical reaction. Journal of scientific computing3(1), 25–43 (1988)

    1988

  44. [44]

    In: ICLR 2021 - 9th International Conference on Learning Representations (2021)

    Li, Z., Kovachki, N., Azizzadenesheli, K., Liu, B., Bhattacharya, K., Stuart, A., Anandkumar, A.: Fourier Neural Operator for Parametric Partial Differential Equations. In: ICLR 2021 - 9th International Conference on Learning Representations (2021)

    2021

  45. [45]

    IEEE transactions on visualization and computer graphics20(12), 2674–2683 (2014)

    Lindstrom, P.: Fixed-rate compressed floating-point arrays. IEEE transactions on visualization and computer graphics20(12), 2674–2683 (2014)

    2014

  46. [46]

    Computational Mechanics46(4), 545–557 (2010)

    Lins, E.F., Elias, R.N., Rochinha, F.A., Coutinho, A.L.: Residual-based variational multiscale simulation of free surface flows. Computational Mechanics46(4), 545–557 (2010)

    2010

  47. [47]

    Lecture notes in computational science and engineering (2012)

    Logg, A., Mardal, K.A., Wells, G.: Automated solution of differential equations by the finite element method: the fenics book. Lecture notes in computational science and engineering (2012)

    2012

  48. [48]

    Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators https://doi.org/10.1038/s42256-021-00302-5

    Lu, L., Jin, P., Pang, G., Zhang, Z., Karniadakis, G.E.: Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators. Nature Machine Intelligence (2021). DOI 10.1038/s42256-021-00302-5

    work page doi:10.1038/s42256-021-00302-5 2021

  49. [49]

    SIAM Journal on Scientific Computing43(6), B1105–B1132 (2021)

    Lu, L., Pestourie, R., Yao, W., Wang, Z., Verdugo, F., Johnson, S.G.: Physics-informed neural networks with hard constraints for inverse design. SIAM Journal on Scientific Computing43(6), B1105–B1132 (2021). DOI 10.1137/21m1397908 Advances in Scientific Machine Learning for Coupled Fluid Flow and Transport 49

    work page doi:10.1137/21m1397908 2021

  50. [50]

    In: Atmospheric Turbulence and Radio Wave Propagation, edited by A

    Lumley, J.L.: The structure of inhomogeneous turbulence. In: Atmospheric Turbulence and Radio Wave Propagation, edited by A. M. Yaglom and V. I. Tatarski (Nauka, Moscow), pp. 166–178 (1967)

    1967

  51. [51]

    SSRN Electronic Journal (2022)

    McClenny, L., Braga-Neto, U.: Self-adaptive physics-informed neural networks. SSRN Electronic Journal (2022). DOI 10.2139/ssrn.4086448

    work page doi:10.2139/ssrn.4086448 2022

  52. [52]

    Annual Review of Fluid Mechanics 42(Volume 42, 2010), 135–156 (2010)

    Meiburg, E., Kneller, B.: Turbidity currents and their deposits. Annual Review of Fluid Mechanics 42(Volume 42, 2010), 135–156 (2010). DOI 10.1146/annurev-fluid-121108-145618

    work page doi:10.1146/annurev-fluid-121108-145618 2010

  53. [53]

    Journal of Fluid Mechanics545, 339–372 (2005)

    Necker, F., Hartel, C., Kleiser, L., Meiburg, E.: Mixing and dissipation in particle-driven gravity currents. Journal of Fluid Mechanics545, 339–372 (2005). DOI 10.1017/S0022112005006932

    work page doi:10.1017/s0022112005006932 2005

  54. [54]

    International Journal of Multiphase Flow28(2), 279–300 (2002)

    Necker, F., H ¨artel, C., Kleiser, L., Meiburg, E.: High-resolution simulations of particle-driven gravity currents. International Journal of Multiphase Flow28(2), 279–300 (2002). DOI 10.1016/S0301-9322(01)00065-9

    work page doi:10.1016/s0301-9322(01)00065-9 2002

  55. [55]

    Advances in Neural Information Processing Systems37, 44989–45037 (2024)

    Ohana, R., McCabe, M., Meyer, L., Morel, R., Agocs, F., Beneitez, M., Berger, M., Burkhart, B., Dalziel, S., Fielding, D., et al.: The well: a large-scale collection of diverse physics simulations for machine learning. Advances in Neural Information Processing Systems37, 44989–45037 (2024)

    2024

  56. [56]

    Proceedings in Applied Mathematics and Mechanics24(4), e202400210 (2024)

    Padula, G., Girfoglio, M., Rozza, G.: A brief review of reduced order models using intrusive and non-intrusive techniques. Proceedings in Applied Mathematics and Mechanics24(4), e202400210 (2024). DOI 10.1002/pamm.202400210

    work page doi:10.1002/pamm.202400210 2024

  57. [57]

    Advances in environmental research3(1), 45–69 (2014)

    Patil, S., Chore, H.: Contaminant transport through porous media: An overview of experimental and numerical studies. Advances in environmental research3(1), 45–69 (2014)

    2014

  58. [58]

    Computer Methods in Applied Mechanics and Engineering306, 196–215 (2016)

    Peherstorfer, B., Willcox, K.: Data-driven operator inference for nonintrusive projection-based model reduction. Computer Methods in Applied Mechanics and Engineering306, 196–215 (2016). DOI 10.1016/j.cma.2016.03.025

    work page doi:10.1016/j.cma.2016.03.025 2016

  59. [59]

    In: SIAM Review (2018)

    Peherstorfer, B., Willcox, K., Gunzburger, M.: Survey of multifidelity methods in uncertainty propagation, inference, and optimization. In: SIAM Review (2018). DOI 10.1137/16M1082469

    work page doi:10.1137/16m1082469 2018

  60. [60]

    Journal of Computational Physics477, 111902 (2023)

    Psaros, A.F., Meng, X., Zou, Z., Guo, L., Karniadakis, G.E.: Uncertainty quantification in scientific machine learning: Methods, metrics, and comparisons. Journal of Computational Physics477, 111902 (2023). DOI 10.1016/j.jcp.2022.111902

    work page doi:10.1016/j.jcp.2022.111902 2023

  61. [61]

    Springer (2015)

    Quarteroni, A., Manzoni, A., Negri, F.: Reduced basis methods for partial differential equations: an introduction. Springer (2015)

    2015

  62. [62]

    Journal of Computational Physics378, 686–707 (2019)

    Raissi, M., Perdikaris, P., Karniadakis, G.E.: Physics-informed neural networks: A deep learn- ing framework for solving forward and inverse problems involving nonlinear partial differential equations. Journal of Computational Physics378, 686–707 (2019)

    2019

  63. [63]

    arXiv preprint arXiv:2301.00942 (2023)

    Ray, D., Pinti, O., Oberai, A.A.: Deep learning and computational physics (lecture notes). arXiv preprint arXiv:2301.00942 (2023). URL https://arxiv.org/abs/2301.00942

    arXiv 2023

  64. [64]

    arXiv preprint arXiv:1810.00597 (2018)

    Rezende, D.J., Viola, F.: Taming vaes. arXiv preprint arXiv:1810.00597 (2018). URL https://arxiv.org/abs/1810.00597

    Pith/arXiv arXiv 2018

  65. [65]

    In: Proceedings of the international conference for high performance computing, networking, storage and analysis, pp

    Rudi, J., Malossi, A.C.I., Isaac, T., Stadler, G., Gurnis, M., Staar, P.W., Ineichen, Y., Bekas, C., Curioni, A., Ghattas, O.: An extreme-scale implicit solver for complex pdes: Highly heterogeneous flow in earth’s mantle. In: Proceedings of the international conference for high performance computing, networking, storage and analysis, pp. 1–12 (2015)

    2015

  66. [66]

    Journal of Mathematical Imaging and Vision62(3), 352–364 (2020)

    Ruthotto, L., Haber, E.: Deep neural networks motivated by partial differential equations. Journal of Mathematical Imaging and Vision62(3), 352–364 (2020)

    2020

  67. [67]

    Journal of Fluid Mechanics656, 5–28 (2010)

    Schmid, P.J.: Dynamic mode decomposition of numerical and experimental data. Journal of Fluid Mechanics656, 5–28 (2010). DOI 10.1017/S0022112010001217

    work page doi:10.1017/s0022112010001217 2010

  68. [68]

    In: Annual Review of Fluid Mechanics, vol

    Schmid, P.J.: Dynamic Mode Decomposition and Its Variants. In: Annual Review of Fluid Mechanics, vol. 54, pp. 225–254 (2021). DOI 10.1146/annurev-fluid-030121-015835

    work page doi:10.1146/annurev-fluid-030121-015835 2021

  69. [69]

    Dissertation, Federal University of Rio de Janeiro (COPPE/UFRJ), Rio de Janeiro, Brazil (2023)

    Silva, R.M.: Advances in physics-informed neural networks for parameter estimation and re- construction of multiphase fluid flows. Dissertation, Federal University of Rio de Janeiro (COPPE/UFRJ), Rio de Janeiro, Brazil (2023)

    2023

  70. [70]

    Nature Communications15(1) (2024)

    Solera-Rico, A., Sanmiguel Vila, C., G ´omez-L´opez, M., Wang, Y., Almashjary, A., Dawson, S.T., Vinuesa, R.:𝛽-Variational autoencoders and transformers for reduced-order modelling of fluid flows. Nature Communications15(1) (2024). DOI 10.1038/s41467-024-45578-4

    work page doi:10.1038/s41467-024-45578-4 2024

  71. [71]

    Proceedings Of The Royal Society A-Mathematical Physical And Engineering Sciences (2026) 50 Gabriel F

    Sreekumar, A., Komis, G., Barman, S., Astolfi, A., Chronopoulos, D.: From snapshots to manifolds: A practical roadmap for interpolatory reduced-order-modeling. Proceedings Of The Royal Society A-Mathematical Physical And Engineering Sciences (2026) 50 Gabriel F. Barros, R ˆomulo M. Silva and Alvaro L. G. A. Coutinho

    2026

  72. [72]

    AIAA Journal55(12), 4013–4041 (2017)

    Taira, K., Brunton, S.L., Dawson, S.T., Rowley, C.W., Colonius, T., McKeon, B.J., Schmidt, O.T., Gordeyev, S., Theofilis, V., Ukeiley, L.S.: Modal analysis of fluid flows: An overview. AIAA Journal55(12), 4013–4041 (2017). DOI 10.2514/1.J056060

    work page doi:10.2514/1.j056060 2017

  73. [73]

    International Journal for numerical methods in fluids47(3), 201–231 (2005)

    Valli, A., Carey, G., Coutinho, A.: Control strategies for timestep selection in finite element simu- lation of incompressible flows and coupled reaction–convection–diffusion processes. International Journal for numerical methods in fluids47(3), 201–231 (2005)

    2005

  74. [74]

    Communications in Numerical Methods in Engineering 24(12), 1941–1952 (2008)

    Valli, A.M.P., Carey, G.F., Coutinho, A.L.G.A.: On decoupled time step/subcycling and iteration strategies for multiphysics problems. Communications in Numerical Methods in Engineering 24(12), 1941–1952 (2008). DOI 10.1002/cnm.1085

    work page doi:10.1002/cnm.1085 1941

  75. [75]

    Finite Elements in Analysis and Design248(2025)

    Velho, R.M., C ˆortes, A.M., Barros, G.F., Rochinha, F.A., Coutinho, A.L.: Advances in data-driven reduced order models using two-stage dimension reduction for coupled viscous flow and transport. Finite Elements in Analysis and Design248(2025). DOI 10.1016/j.finel.2025.104355

    work page doi:10.1016/j.finel.2025.104355 2025

  76. [76]

    Symmetry14(6), 1219 (2022)

    Wang, C., Qiang, X., Xu, M., Wu, T.: Recent advances in surrogate modeling methods for uncer- tainty quantification and propagation. Symmetry14(6), 1219 (2022)

    2022

  77. [77]

    Computer Methods in Applied Mechanics and Engineering403, 115671 (2023)

    Wu, C., Zhu, M., Tan, Q., Kartha, Y., Lu, L.: A comprehensive study of non-adaptive and residual- based adaptive sampling for physics-informed neural networks. Computer Methods in Applied Mechanics and Engineering403, 115671 (2023). DOI 10.1016/j.cma.2022.115671

    work page doi:10.1016/j.cma.2022.115671 2023

  78. [78]

    Applied Ocean Research39, 53–71 (2013)

    Xing, Y., Ai, C., Jin, S.: A three-dimensional hydrodynamic and salinity transport model of estuarine circulation with an application to a macrotidal estuary. Applied Ocean Research39, 53–71 (2013). DOI 10.1016/j.apor.2012.10.003

    work page doi:10.1016/j.apor.2012.10.003 2013

  79. [79]

    In: The Fourteenth International Conference on Learning Represen- tations (2026)

    Zandieh, A., Daliri, M., Hadian, M., Mirrokni, V.: Turboquant: Online vector quantization with near-optimal distortion rate. In: The Fourteenth International Conference on Learning Represen- tations (2026)

    2026

  80. [80]

    Measurement Science and Technology36(1), 012003 (2025)

    Zhao, X., Sun, Y., Li, Y., Jia, N., Xu, J.: Applications of machine learning in real-time control systems: a review. Measurement Science and Technology36(1), 012003 (2025)

    2025