Pith. sign in

REVIEW 3 major objections 7 minor 58 references

An Adaptive Framework for Autoregressive Forecasting in CFD Using Hybrid Modal Decomposition and Deep Learning

T0 review · 3 major / 7 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read Periodic retraining of a POD-deep-learning forecaster can stabilize long-horizon flow predictions and cut CFD solver work by 30 to 95 percent.

desk verdict A legitimate proof of concept for adaptive POD-DL forecasting with real code and data, but the 30-95% cost reduction is a snapshot-counting estimate, not a measured speedup, and the abstract overstates it. read the letter →

arxiv 2505.01531 v1 pith:Q7QPDHMK submitted 2025-05-02 physics.flu-dyn cs.AI

Rodrigo Abadía-Heredia , Manuel Lopez-Martin , Soledad Le Clainche This is my paper · ORCID
classification physics.flu-dyncs.AI
keywords adaptiveforecastingproperorthogonaldecompositiondeeplearningautoregressivemodelscomputationalfluiddynamicsreduced-ordertransferlong-horizonprediction
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 claims that a deep-learning model forecasting fluid flow can be kept accurate over long time horizons by periodically retraining it on fresh simulation data instead of letting it run open-loop. The framework alternates between predicting a block of future snapshots with a hybrid POD-deep-learning (POD-DL) model and pulling a small set of new snapshots to retrain that model using transfer learning. Because proper orthogonal decomposition reduces each flow field to a handful of mode amplitudes, each retraining session is cheap, which is what makes the adaptive loop affordable. On a laminar cylinder wake, a turbulent cylinder wake, and a subsonic jet, the paper reports theoretical reductions of 30% to 95% in the number of snapshots a numerical solver would need to compute, without losing physical consistency as measured by prediction-error probabilities, percentiles, and turbulent kinetic energy.

What carries the argument

The load-bearing mechanism is the POD-DL hybrid: proper orthogonal decomposition via the method of snapshots ($\tilde{D} = U \Sigma \Gamma^*$) compresses the high-dimensional velocity field into $k$ POD coefficients $a_i(t) = \sigma_i \gamma_{t,i}$, and an LSTM followed by three dense layers forecasts the next coefficient vector from a window of $m=10$ past vectors. The adaptive loop is the second mechanism: after $P$ autoregressive predictions, the model is retrained on $S_1$ new snapshots with transfer-learning weight initialization, keeping the same POD truncation $k$ (6 modes for the laminar flow, 13 for the turbulent flow, 5 for the jet). The paper measures success with the probability of prediction error $\Pr_\alpha(t_S) \equiv \Pr(-\alpha \le v_s - \hat{v}_s \le \alpha)$ and with the theoretical saving formula $\mathrm{TCS}$, and it checks physical consistency through vorticity percentiles and turbulent kinetic energy.

What would settle it

Run the adaptive loop with a live CFD solver replacing the precomputed dataset, and compare total wall-clock time against the solver alone computing every snapshot. If the adaptive time (training, retraining, POD, and solver spin-up after each retraining) is not lower than the solver-only time on the same problem, the claimed computational saving fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that an adaptive retraining schedule stabilizes autoregressive POD-DL forecasts over long horizons and thereby reduces the computational cost of CFD simulations. The framework trains a POD-DL model on $S_0$ snapshots, predicts the next $P$ snapshots, then retrains on $S_1$ new snapshots, initializing each retraining with the previous optimal weights via transfer learning. The theoretical computational saving is $\mathrm{TCS} = (1 - (S_0 + n S_1)/S) \times 100\%$, where $n$ is the number of retraining updates and $S$ the total number of snapshots, a measure that counts only the snapshots the solver never has to compute. The authors further claim this is, to their knowledge, the first generalizable and fully data-driven adaptive framework of this kind applied to the POD-DL hybrid model, and that it works across regimes from laminar to turbulent, with the largest savings (up to 95%) in the statistically steady jet case.

Load-bearing premise

The claimed savings assume that the time spent training and retraining the POD-DL model and computing the POD is negligible compared with the CFD solver; the paper reports roughly 1.2-1.76 minutes per training session and does not fold those timings into the 30-95% saving figures.

Editorial extensions

If this is right

  • If the framework holds up, autoregressive deep-learning models no longer need to be abandoned after a fixed number of steps; retraining on fresh snapshots can restore accuracy after a distributional shift, as shown when the laminar predictions begin to diverge near snapshot 250 and then reconverge after retraining.
  • For the laminar cylinder wake, predicting 299 of 499 snapshots gives a theoretical saving of about 60%, and predicting 399 snapshots gives about 80%; for the turbulent wake the saving is 30-70% depending on block sizes, and for the subsonic jet it reaches 65-95%.
  • Because the framework is fully data-driven, the component that supplies retraining snapshots could be an experimental measurement campaign or any other source of high-fidelity snapshots, not only a CFD solver.
  • The hyperparameters $S_0$, $S_1$, and $P$ directly set the accuracy-cost trade-off reported by the TCS metric, so tuning them is the practical lever for applying the framework to a new flow.
  • The framework is positioned as a proof of concept for accelerating CFD, with the code released openly and planned for integration into the ModelFLOWs-app.

Reading between the lines

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

  • The 30-95% figures are theoretical in a specific sense: they count snapshots not computed by the CFD solver, but exclude the measured training times (about 1.2-1.76 minutes per session) and the POD computation; a wall-clock comparison with a live solver could show smaller net savings, especially when retraining is frequent.
  • The framework currently uses a fixed prediction horizon $P$; a natural untested extension is to trigger retraining from an uncertainty estimate or PDE residual when the model begins to diverge, which would remove the need to choose $P$ in advance and could improve the turbulent case.
  • Because the POD basis is frozen after the initial training, any dynamics that become energetic only later must be represented in the fixed low-dimensional space; adaptively updating the basis or growing $k$ is a testable extension that could matter for flows with strong transient events.
  • The same adaptive loop should apply beyond fluid dynamics to any high-dimensional time-dependent simulation or experimental measurement where snapshots can be obtained on demand, so the impact could extend to other fields with expensive forward models.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 7 minor

Summary. The manuscript proposes an adaptive framework for CFD forecasting in which a POD-LSTM model alternately predicts flow fields over P snapshots and is retrained on newly acquired CFD snapshots, using transfer learning from the previous weights. The framework is tested on three precomputed datasets: a laminar three-dimensional cylinder wake (Re=280), a turbulent cylinder wake (Re=4000), and a statistically steady isothermal subsonic jet. Accuracy is assessed through a probability-of-prediction-error (PPE) metric, percentile comparisons, and turbulent kinetic energy, while computational savings are reported through a theoretical computational saving (TCS) metric based on snapshot counts. The paper claims reductions in computational cost from 30% to 95% without compromising accuracy, and presents the approach as the first generalizable, fully data-driven adaptive framework for stabilizing autoregressive DL forecasts in CFD. The code is provided open-source and the datasets are publicly available.

Significance. The paper addresses a real and timely problem: autoregressive deep-learning surrogates for CFD tend to drift over long horizons, and periodic retraining with fresh high-fidelity data is a natural remedy. The authors provide a clear proof of concept on three flow regimes, and the open-source implementation is a concrete strength. The demonstration that retraining can recover accuracy after a distributional shift, especially in the laminar case (T2), is valuable. However, the central quantitative claim—a 30–95% reduction in computational cost—currently rests on an accounting identity over snapshot counts, not on measured wall-clock time or a cost model that includes training, SVD, inference, and solver-restart overhead. The absence of a non-adaptive baseline also leaves the stabilization benefit unquantified. If the cost claim were supported by proper timings or a realistic cost model, the framework would be a useful contribution; as it stands, the paper is a credible proof-of-concept with overstated headline claims.

major comments (3)
  1. [§4, Eq. (18); §5.1–§5.3; Abstract] The headline 30–95% computational-cost reduction is computed from snapshot counts alone: TCS = (1 - (S0 + nS1)/S) × 100% counts only the snapshots requested from the data source and assigns zero cost to DL training/retraining, POD/SVD, inference, and solver restart. Section 5.1 reports 1.2 minutes per training session, Section 5.2 reports 1.76 minutes, and Section 5.3 reports 1.18 minutes; with n retraining sessions these times are not negligible relative to a CFD solve, yet they are excluded from the reported savings. Section 6 also concedes that invoking a live solver requires restart/convergence steps that reduce the speedup, without quantifying them. The abstract therefore overstates the result: what is demonstrated is a reduction in the number of solver-generated snapshots, not a measured reduction in computational cost. Please either report wall-clock timings for the full pipeline against the solver, include a cost model with training/SVD/restart terms, or revise the abstract and conclusions to state the snapshot-count interpretation.
  2. [§5.1, Figs. 5 and 6; §6] The stabilization claim is not tested against a non-adaptive baseline. In test cases T1 and T3 the predictions diverge at the distributional shift around snapshot 250 and only recover after a later retraining (Figs. 5(a) and 5(c)); the paper does not show the corresponding one-shot POD-DL forecast that would quantify how much the adaptive retraining extends the useful horizon. Without this comparison, the evidence supports “periodic retraining can eventually recover accuracy after a distribution shift” but not “the adaptive framework stabilizes autoregressive forecasts over long horizons.” Please add a no-retraining baseline and report the divergence horizon and integrated error for both settings.
  3. [Abstract; §6–§7] The claims of “first generalizable and fully data-driven adaptive framework” and “without compromising physical consistency or accuracy” are stronger than the experiments support. The adaptive schedule is manual: S0, S1, P, E, k, m, B, and the PPE thresholds α are all chosen per problem (Tables 3–5), and the framework has no uncertainty-based trigger for deciding when to retrain; Section 6 lists uncertainty quantification as future work. This is acceptable as a proof of concept, but the manuscript should either add an automatic retraining criterion and a sensitivity study over these hyperparameters, or explicitly restrict the claims to the tested configurations.
minor comments (7)
  1. [§5.3, Fig. 13] The caption of Fig. 13 says “laminar flow past a cylinder” but the figure shows the isothermal subsonic jet; please correct the caption.
  2. [§5.3, Figs. 14 and 15] The captions of Figs. 14 and 15 refer to “vorticity” prediction error and vorticity percentiles, but the isothermal jet case is evaluated directly on the streamwise velocity field; please replace “vorticity” with “velocity” in those captions.
  3. [Appendix A, Algorithm 1] The algorithm returns the sum of bins 0 to index-1, where index = floor(err_thresh/bin_width); when err_thresh is not a bin edge this sum equals Pr(|error| <= floor(err_thresh/bin_width)*bin_width), not Pr(|error| <= err_thresh), and when err_thresh < bin_width it returns an empty sum (0). Please redefine the binning so that the reported PPE matches the stated definition in Eq. (17), or state the approximation explicitly.
  4. [§5.1–§5.3] The phrase “theoretical computational saving” in Sections 5.1–5.3 is inconsistent with the abstract’s “demonstrating … reduction in computational cost”; please unify the terminology throughout the manuscript.
  5. [§5.2] A table reporting per-session training time, inference time, SVD/POD time, and solver or dataset-generation time would make the cost discussion reproducible; currently only training and prediction times are given, without a wall-clock comparison.
  6. [§4, Eq. (17); §5.1–§5.3] The PPE thresholds α are selected from the observed range of vorticity or velocity values in each dataset, so the accuracy claims are threshold-dependent; please report the sensitivity of Prα to α and to the histogram bin width.
  7. [§3.2, Eq. (24)] There is a typographical error in the sentence containing “Here, is ˜vx represents”; the text also switches between “vorticity” and “velocity” in Section 5.3, which should be proofread.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation: the POD-DL forecasts are validated against external precomputed data, and the cost-saving metric TCS is an explicitly defined accounting identity rather than a derived prediction.

full rationale

The paper's derivation chain is self-contained. POD coefficients are obtained by SVD from precomputed snapshots, the LSTM forecasts the coefficients autoregressively, and accuracy is measured with PPE, percentiles, TKE, and visual field comparisons against ground-truth snapshots that were not used for training at the time of prediction. The claimed 30-95% computational saving is not a model prediction derived from the framework; it is the definition TCS = (1 - (S0 + nS1)/S) x 100% applied to chosen hyperparameters S0, S1, and P. That is a transparent accounting identity over snapshot counts, not a circular reduction: the physical/accuracy claim stands or falls on the independent PPE and TKE comparisons. The paper itself flags the relevant caveats, stating in Section 2 that 'instead of employing a live numerical solver, it is substituted with a precomputed dataset' and in Section 6 that 'incorporating a numerical solver within the adaptive framework may reduce the computational speedup achieved.' These are correctness and assumption concerns about unmeasured wall-clock costs, not circularity. The only notable self-citations are to Abadia-Heredia et al. (2025) for the window size m = 10 and for the benefits of transfer learning, but those are design choices imported from prior empirical work, and they do not determine the PPE/TKE outcomes or force the central claim. No uniqueness theorem, fitted parameter renamed as a prediction, or definitional equivalence between input and output is present. Therefore the paper exhibits no significant circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim rests on standard POD/LSTM mathematics plus several domain assumptions about the representativeness of the datasets, the predictability of POD coefficients, and the validity of TCS as a cost measure. The free parameters are hyperparameters of the framework (k, m, B, E, S0/S1/P, and PPE thresholds); none is fitted to force the target result, but the reported savings range is sensitive to the chosen S0/S1/P values. No new physical entities are introduced.

free parameters (6)
  • Number of retained POD modes k = 6 (laminar), 13 (turbulent), 5 (jet)
    Set by cumulative energy thresholds (90% for laminar, 30% for turbulent and jet) from the first S0 snapshots; fixed across retrainings; directly controls the forecasting problem dimensionality.
  • LSTM input window length m = 10
    Taken from the authors' prior work Abadia-Heredia et al. (2025), not tuned here; determines the temporal memory of the forecaster.
  • Batch size B = 8
    Standard multiple-of-2 choice; minor effect on training.
  • Training epochs E = 1500
    Fixed across all test cases; authors state performance is not sensitive to E.
  • Adaptive hyperparameters S0, S1, P = Varies per test: laminar S0=100, S1=50-100, P=100-400; turbulent S0=200, S1=100-200, P=50-500; jet S0=100, S1=50-100…
    Chosen by the authors to explore the framework; the reported TCS savings (30-95%) are a direct function of these choices via Eq. (18), so the headline savings depend on the tested configurations.
  • PPE error thresholds alpha = 0.5/1.0 (laminar vorticity), 0.2/0.5 (turbulent vorticity), 0.01/0.02 (jet velocity)
    Chosen based on the observed range of the reference fields; the accuracy claims (e.g., Pr_alpha > 70%) depend on these thresholds, which differ across cases.
assumptions (5)
  • standard math SVD/POD provides a valid reduced basis and the first k modes capture the scales of interest.
    Invoked in Section 3.1; the entire method depends on the POD truncation representing the flow for forecasting and for recovery via Eq. (2).
  • domain assumption An LSTM trained with one-step-ahead MSE, Adam, and cosine annealing can capture the temporal evolution of POD coefficients in an autoregressive loop.
    Section 3.2; no proof of forecasting capability beyond the empirical results; central to the method.
  • domain assumption The three datasets (laminar wake, turbulent wake, isothermal jet) are representative and sufficient to establish generalizability across flow regimes.
    Section 5; the paper's claim of 'generalizable' is based on only these three, with the jet being statistically steady, which the authors note makes prediction easier.
  • domain assumption Retraining on the newest S1 snapshots with transfer learning improves or maintains accuracy without catastrophic forgetting of previously learned dynamics.
    Section 4 and Fig. 2; the adaptive cycle assumes the model can be updated quickly and effectively on new data.
  • ad hoc to paper The theoretical computational saving TCS in Eq. (18) is a meaningful proxy for actual CFD cost reduction.
    This is the paper's own metric; it counts only skipped snapshots and excludes training time, POD computation, and solver startup costs, so it is an assumption rather than a measured quantity.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An Adaptive Framework for Autoregressive Forecasting in CFD Using Hybrid Modal Decomposition and Deep Learning." pith.science (2026). https://pith.science/paper/Q7QPDHMK

@misc{pith2026250501531,
  author       = {Pith},
  title        = {Pith review of: An Adaptive Framework for Autoregressive Forecasting in CFD Using Hybrid Modal Decomposition and Deep Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Q7QPDHMK}},
  note         = {Machine review of arXiv:2505.01531}
}
read the original abstract

This work presents, to the best of the authors' knowledge, the first generalizable and fully data-driven adaptive framework designed to stabilize deep learning (DL) autoregressive forecasting models over long time horizons, with the goal of reducing the computational cost required in computational fluid dynamics (CFD) simulations.The proposed methodology alternates between two phases: (i) predicting the evolution of the flow field over a selected time interval using a trained DL model, and (ii) updating the model with newly generated CFD data when stability degrades, thus maintaining accurate long-term forecasting. This adaptive retraining strategy ensures robustness while avoiding the accumulation of predictive errors typical in autoregressive models. The framework is validated across three increasingly complex flow regimes, from laminar to turbulent, demonstrating from 30 \% to 95 \% reduction in computational cost without compromising physical consistency or accuracy. Its entirely data-driven nature makes it easily adaptable to a wide range of time-dependent simulation problems. The code implementing this methodology is available as open-source and it will be integrated into the upcoming release of the ModelFLOWs-app.

Figures

Figures reproduced from arXiv: 2505.01531 by the authors.

Figure 1
Figure 1. Variation of the learning rate using the CosineAnnealingWarmRestarts scheduler [PITH_FULL_IMAGE:figures/full_fig_p014_1.png] view at source ↗
Figure 2
Figure 2. Visual representation of the adaptive framework proposed in this work. Initially, [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. Percentage of energy contained, Eg(k), when the truncation is done at the k, up to 20, most energetic modes in the laminar flow past a cylinder. This truncation is kept for all subsequent trainings within the adaptive framework. For this reason, it is required that S1 ≥ k; otherwise, the number of POD modes computed from the new S1 snapshots would be insufficient to match the selected truncation level k. The reliabi… view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Snapshots comparing the ground truth (GT) vorticity, of the laminar flow past [PITH_FULL_IMAGE:figures/full_fig_p020_4.png]
Figure 5
Figure 5. Figure 5: The top row of all figures displays the probability of achieving an absolute [PITH_FULL_IMAGE:figures/full_fig_p021_5.png]
Figure 6
Figure 6. Figure 6: Temporal evolution of the 25th (bottom), 50th (middle), and 75th (top) vorticity [PITH_FULL_IMAGE:figures/full_fig_p022_6.png]
Figure 7
Figure 7. Figure 7: Percentage of energy contained, Eg(k), when the truncation is done at the k, up to 40, most energetic modes in the turbulent flow past a cylinder. This truncation is maintained for all training iterations in the adaptive framework to maintain consistency with the POD-D…
Figure 8
Figure 8. Figure 8: Snapshots comparing the ground truth (GT) vorticity, of the turbulent flow past [PITH_FULL_IMAGE:figures/full_fig_p026_8.png]
Figure 9
Figure 9. Figure 9: The top row of all figures displays the probability of achieving an absolute vor [PITH_FULL_IMAGE:figures/full_fig_p026_9.png]
Figure 10
Figure 10. Figure 10: Turbulent kinetic energy (TKE) from the ground truth (left) and adaptive [PITH_FULL_IMAGE:figures/full_fig_p028_10.png]
Figure 11
Figure 11. Figure 11: Temporal evolution of the 25th (bottom), 50th (middle), and 75th (top) vortic [PITH_FULL_IMAGE:figures/full_fig_p029_11.png]
Figure 12
Figure 12. Figure 12: Percentage of energy contained, Eg(k), when the truncation is done at the n, up to 20, most energetic modes in the isothermal jet. This truncation to 5 modes is consistently applied throughout all training iterations within the adaptive framework to maintain architect…
Figure 13
Figure 13. Figure 13: Snapshots comparing the ground truth (GT) vorticity, of the laminar flow past [PITH_FULL_IMAGE:figures/full_fig_p032_13.png]
Figure 14
Figure 14. Figure 14: The top row of all figures displays the probability of achieving an absolute [PITH_FULL_IMAGE:figures/full_fig_p032_14.png]
Figure 15
Figure 15. Figure 15: Temporal evolution of the 25th (bottom), 50th (middle), and 75th (top) vortic [PITH_FULL_IMAGE:figures/full_fig_p034_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 19 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  3. [3]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  4. [4]

    , author Corrochano, A

    author Abadía-Heredia, R. , author Corrochano, A. , author Lopez-Martin, M. , author Le Clainche, S. , year 2025 . title Generalization capabilities and robustness of hybrid models grounded in physics compared to purely deep learning models . journal Physics of Fluids volume 37 , pages 035149 . https://doi.org/10.1063/5.0253876, :10.1063/5.0253876

    work page doi:10.1063/5.0253876 2025

  5. [5]

    , author Lopez-Martín, M

    author Abadía-Heredia, R. , author Lopez-Martín, M. , author le Clainche , S. , year 2024 . title Evaluating forecast divergence in fluid dynamics: A metric for autoregressive models in the absence of true solution data , in: booktitle ECCOMAS 2024 , publisher Scipedia . https://www.scipedia.com/public/Abadia-Heredia_et_al_2024a

    work page 2024

  6. [6]

    , author López-Martín, M

    author Abadía-Heredia, R. , author López-Martín, M. , author Carro, B. , author Arribas, J. , author Pérez, J. , author Le Clainche , S. , year 2022 . title A predictive hybrid reduced order model based on proper orthogonal decomposition combined with deep learning architectures . journal Expert Systems with Applications volume 187 , pages 115910 . https:...

    arXiv 2022

  7. [7]

    , year 1991

    author Aubry, N. , year 1991 . title On the hidden beauty of the proper orthogonal decomposition . journal Theoret. Comput. Fluid Dynamics volume 2 , pages 339--352 . :10.1007/BF00271473

    work page doi:10.1007/bf00271473 1991

  8. [8]

    , author Gauch, M

    author Auer, A. , author Gauch, M. , author Klotz, D. , author Hochreiter, S. , year 2023 . title Conformal prediction for time series with modern hopfield networks . https://arxiv.org/abs/2303.12783, http://arxiv.org/abs/2303.12783 arXiv:2303.12783

    arXiv 2023

Show all 58 references
  1. [9]

    , author Pöppel, K

    author Beck, M. , author Pöppel, K. , author Spanring, M. , author Auer, A. , author Prudnikova, O. , author Kopp, M. , author Klambauer, G. , author Brandstetter, J. , author Hochreiter, S. , year 2024 . title xlstm: Extended long short-term memory . https://arxiv.org/abs/240...

    2024 arXiv

  2. [10]

    , author Clainche, S.L

    author Beltr \'a n, V. , author Clainche, S.L. , author Vega, J.M. , year 2020 . title A data-driven rom based on hodmd , in: editor Mart \'i nez \'A lvarez, F. , editor Troncoso Lora, A. , editor S \'a ez Mu \ n oz, J.A. , editor Quinti \'a n, H. , editor Corchado, E. (Eds.),...

    2020

  3. [11]

    , author Le Clainche, S

    author Beltrán, V. , author Le Clainche, S. , author Vega, J.M. , year 2022 . title An adaptive data-driven reduced order model based on higher order dynamic mode decomposition . journal Journal of Scientific Computing volume 92 . http://dx.doi.org/10.1007/s10915-022-01855-2, ...

    2022 doi

  4. [12]

    , author Girfoglio, M

    author Besabe, L. , author Girfoglio, M. , author Quaini, A. , author Rozza, G. , year 2025 . title Data-driven reduced order modeling of a two-layer quasi-geostrophic ocean model . journal Results in Engineering volume 25 , pages 103691 . https://www.sciencedirect.com/science...

    2025

  5. [13]

    , author Jordan, P

    author Brès, G.A. , author Jordan, P. , author Jaunet, V. , author Le Rallic, M. , author Cavalieri, A.V.G. , author Towne, A. , author Lele, S.K. , author Colonius, T. , author Schmidt, O.T. , year 2018 . title Importance of the nozzle-exit boundary-layer state in subsonic tu...

    2018 doi

  6. [14]

    , author Barone, M

    author Carlberg, K. , author Barone, M. , author Antil, H. , year 2017 . title Galerkin v. least-squares petrov–galerkin projection in nonlinear model reduction . journal Journal of Computational Physics volume 330 , pages 693--734 . https://www.sciencedirect.com/science/artic...

    2017 doi

  7. [15]

    , author Freitas, R.S.M

    author Corrochano, A. , author Freitas, R.S.M. , author Parente, A. , author Clainche, S.L. , year 2023 . title A predictive physics-aware hybrid reduced order model for reacting flows . https://arxiv.org/abs/2301.09860, http://arxiv.org/abs/2301.09860 arXiv:2301.09860

    2023 arXiv

  8. [16]

    , author Jati, A

    author Ekambaram, V. , author Jati, A. , author Nguyen, N. , author Sinthong, P. , author Kalagnanam, J. , year 2023 . title Tsmixer: Lightweight mlp-mixer model for multivariate time series forecasting , in: booktitle Proceedings of the 29th ACM SIGKDD Conference on Knowledge...

    2023

  9. [17]

    , author Kaltenbach, S

    author Gao, H. , author Kaltenbach, S. , author Koumoutsakos, P. , year 2024 . title Generative learning for forecasting the dynamics of high-dimensional complex systems . journal Nature Communications volume 15 , pages 8904 . https://doi.org/10.1038/s41467-024-53165-w, :10.10...

    2024 doi

  10. [18]

    , author Bengio, Y

    author Glorot, X. , author Bengio, Y. , year 2010 . title Understanding the difficulty of training deep feedforward neural networks , in: editor Teh, Y.W. , editor Titterington, M. (Eds.), booktitle Proceedings of the Thirteenth International Conference on Artificial Intellige...

    2010

  11. [19]

    , author Bengio, Y

    author Goodfellow, I. , author Bengio, Y. , author Courville, A. , year 2016 . title Deep Learning . publisher MIT Press . note http://www.deeplearningbook.org

    2016

  12. [20]

    , author Wang, Y

    author Han, R. , author Wang, Y. , author Zhang, Y. , author Chen, G. , year 2019 . title A novel spatial-temporal prediction method for unsteady wake flows based on hybrid deep neural network . journal Physics of Fluids volume 31 , pages 127101 . https://doi.org/10.1063/1.512...

    2019 doi

  13. [21]

    , author Fukami, K

    author Hasegawa, K. , author Fukami, K. , author Murata, T. , author Fukagata, K. , year 2020 . title Machine-learning-based reduced-order modeling for unsteady flows around bluff bodies of various shapes . journal Theoretical and Computational Fluid Dynamics volume 34 , pages...

    2020 doi

  14. [22]

    , author Zhang, X

    author He, K. , author Zhang, X. , author Ren, S. , author Sun, J. , year 2015 . title Delving deep into rectifiers: Surpassing human-level performance on imagenet classification , in: booktitle 2015 IEEE International Conference on Computer Vision (ICCV) , pp. pages 1026--103...

    2015 doi

  15. [23]

    , author Schmidhuber, J

    author Hochreiter, S. , author Schmidhuber, J. , year 1997 . title Long Short-Term Memory . journal Neural Computation volume 9 , pages 1735--1780 . https://doi.org/10.1162/neco.1997.9.8.1735, :10.1162/neco.1997.9.8.1735

    1997 doi

  16. [24]

    , author Lumley, J.L

    author Holmes, P. , author Lumley, J.L. , author Berkooz, G. , year 1996 . title Turbulence, Coherent Structures, Dynamical Systems and Symmetry . Cambridge Monographs on Mechanics, publisher Cambridge University Press

    1996

  17. [25]

    , year 2018

    author Jiménez, J. , year 2018 . title Coherent structures in wall-bounded turbulence . journal Journal of Fluid Mechanics volume 842 , pages P1 . :10.1017/jfm.2018.144

    2018 doi

  18. [26]

    , author Ba, J

    author Kingma, D.P. , author Ba, J. , year 2017 . title Adam: A method for stochastic optimization . https://arxiv.org/abs/1412.6980, http://arxiv.org/abs/1412.6980 arXiv:1412.6980

    2017 arXiv

  19. [27]

    , author Vlachas, P.R

    author Kičić, I. , author Vlachas, P.R. , author Arampatzis, G. , author Chatzimanolakis, M. , author Guibas, L. , author Koumoutsakos, P. , year 2023 . title Adaptive learning of effective dynamics for online modeling of complex systems . journal Computer Methods in Applied M...

    2023

  20. [28]

    o wer, M. , author Lottes, J. , author Rasp, S. , author D \

    author Kochkov, D. , author Yuval, J. , author Langmore, I. , author Norgaard, P. , author Smith, J. , author Mooers, G. , author Kl \"o wer, M. , author Lottes, J. , author Rasp, S. , author D \"u ben, P. , author Hatfield, S. , author Battaglia, P. , author Sanchez-Gonzalez,...

    2024 doi

  21. [29]

    , author Pritzel, A

    author Lakshminarayanan, B. , author Pritzel, A. , author Blundell, C. , year 2017 . title Simple and scalable predictive uncertainty estimation using deep ensembles . https://arxiv.org/abs/1612.01474, http://arxiv.org/abs/1612.01474 arXiv:1612.01474

    2017 arXiv

  22. [30]

    , author Pérez, J.M

    author Le Clainche , S. , author Pérez, J.M. , author Vega, J.M. , year 2018 . title Spatio-temporal flow structures in the three-dimensional wake of a circular cylinder . journal Fluid Dynamics Research volume 50 , pages 051406 . https://dx.doi.org/10.1088/1873-7005/aab2f1, :...

    2018 doi

  23. [31]

    , author Varas, F

    author Le Clainche, S. , author Varas, F. , author Vega, J.M. , year 2017 . title Accelerating oil reservoir simulations using pod on the fly . journal International Journal for Numerical Methods in Engineering volume 110 , pages 79--100 . https://onlinelibrary.wiley.com/doi/a...

    2017 doi

  24. [32]

    , author Vega, J.M

    author Le Clainche, S. , author Vega, J.M. , year 2017 . title Higher order dynamic mode decomposition to identify and extrapolate flow patterns . journal Physics of Fluids volume 29 , pages 084102 . https://doi.org/10.1063/1.4997206, :10.1063/1.4997206, http://arxiv.org/abs/h...

    2017 doi

  25. [33]

    , author Li, Z

    author Lin, L. , author Li, Z. , author Li, R. , author Li, X. , author Gao, J. , year 2024 . title Diffusion models for time-series applications: a survey . journal Frontiers of Information Technology & Electronic Engineering volume 25 , pages 19--41 . https://doi.org/10.1631...

    2024 doi

  26. [34]

    , author Hutter, F

    author Loshchilov, I. , author Hutter, F. , year 2017 . title Sgdr: Stochastic gradient descent with warm restarts . https://arxiv.org/abs/1608.03983, http://arxiv.org/abs/1608.03983 arXiv:1608.03983

    2017 arXiv

  27. [35]

    , author Noack, B

    author Luchtenburg, D. , author Noack, B. , author Schlegel, M. , year 2009 . title Galerkin method for fluid flows with analytical examples and matlab source codes. journal Berlin Institute of Technology. Report 01

    2009

  28. [36]

    , year 1967

    author Lumley, J.L. , year 1967 . title The structure of inhomogeneous turbulent flows , in: editor Yaglom, A.M. , editor Tartarsky, V.I. (Eds.), booktitle Atmospheric Turbulence and Radio Wave Propagation , publisher Nauka , address Moscow . pp. pages 166--178

    1967

  29. [37]

    , author Spiliotis, E

    author Makridakis, S. , author Spiliotis, E. , author Assimakopoulos, V. , year 2018 . title The m4 competition: Results, findings, conclusion and way forward . journal International Journal of Forecasting volume 34 , pages 802--808 . https://www.sciencedirect.com/science/arti...

    2018 doi

  30. [38]

    , author López, E

    author Martínez-Sánchez, A. , author López, E. , author Le Clainche, S. , author Lozano-Durán, A. , author Srivastava, A. , author Vinuesa, R. , year 2023 . title Causality analysis of large-scale structures in the flow around a wall-mounted square cylinder . journal Journal o...

    2023 doi

  31. [39]

    , author Hess, D

    author Mendez, M.A. , author Hess, D. , author Watz, B.B. , author Buchlin, J.M. , year 2020 . title Multiscale proper orthogonal decomposition (mpod) of tr-piv data—a case study on stationary and transient cylinder wake flows . journal Measurement Science and Technology volum...

    2020 doi

  32. [40]

    , author Gaitonde, D.V

    author Mohan, A.T. , author Gaitonde, D.V. , year 2018 . title A deep learning based approach to reduced order modeling for turbulent flow control using lstm neural networks . https://arxiv.org/abs/1804.09269, http://arxiv.org/abs/1804.09269 arXiv:1804.09269

    2018 arXiv

  33. [41]

    , author Fukami, K

    author Nakamura, T. , author Fukami, K. , author Hasegawa, K. , author Nabae, Y. , author Fukagata, K. , year 2021 . title Convolutional neural network and long short-term memory based reduced order surrogate for minimal turbulent channel flow . journal Physics of Fluids volum...

    2021 doi

  34. [42]

    , author Morzynski, M

    author Noack, B.R. , author Morzynski, M. , author Tadmor, G. , year 2011 . title Reduced-order modelling for flow control. journal New York: Springer

    2011

  35. [44]

    , author Carlberg, K.T

    author Parish, E.J. , author Carlberg, K.T. , year 2020 b. title Time-series machine-learning error models for approximate solutions to parameterized dynamical systems . journal Computer Methods in Applied Mechanics and Engineering volume 365 , pages 112990 . https://www.scien...

    2020

  36. [45]

    , author Rahman, S.M

    author Pawar, S. , author Rahman, S.M. , author Vaddireddy, H. , author San, O. , author Rasheed, A. , author Vedula, P. , year 2019 . title A deep learning enabler for nonintrusive reduced order modeling of fluid flows . journal Physics of Fluids volume 31 , pages 085101 . ht...

    2019 doi

  37. [46]

    , author Manzoni, A

    author Quarteroni, A. , author Manzoni, A. , author Negri, F. , year 2016 . title Reduced basis methods for partial differential equations. journal Springer

    2016

  38. [47]

    , author Terragni, F

    author Rapún, M.L. , author Terragni, F. , author Vega, J.M. , year 2015 . title Adaptive pod-based low-dimensional modeling supported by residual estimates . journal International Journal for Numerical Methods in Engineering volume 104 , pages 844--868 . https://onlinelibrary...

    2015 doi

  39. [48]

    , author Dedè, L

    author Regazzoni, F. , author Dedè, L. , author Quarteroni, A. , year 2019 . title Machine learning for fast and reliable solution of time-dependent differential equations . journal Journal of Computational Physics volume 397 , pages 108852 . https://www.sciencedirect.com/scie...

    2019 doi

  40. [49]

    , author Schulze, P

    author Reiss, J. , author Schulze, P. , author Sesterhenn, J. , author Mehrmann, V. , year 2018 . title The shifted proper orthogonal decomposition: A mode decomposition for multiple transport phenomena . journal SIAM Journal on Scientific Computing volume 40 , pages A1322--A1...

    2018 doi

  41. [50]

    , year 2000

    author Rempfer, D. , year 2000 . title On low-dimensional galerkin models for fluid flow . journal Theoretical and Computational Fluid Dynamics volume 14 , pages 75--88 . https://doi.org/10.1007/s001620050131, :10.1007/s001620050131

    2000 doi

  42. [51]

    , author Abadía-Heredia, R

    author Sengupta, A. , author Abadía-Heredia, R. , author Hetherington, A. , author Pérez, J.M. , author Clainche, S.L. , year 2025 . title Hybrid machine learning models based on physical patterns to accelerate cfd simulations: a short guide on autoregressive models . https://...

    2025 arXiv

  43. [52]

    , year 1987

    author Sirovich, L. , year 1987 . title Turbulence and the dynamics of coherent structures: I, ii, iii . journal Quart. Appl. Math. volume 45 , pages 561--590

    1987

  44. [53]

    , author Sapsis, T.P

    author Stamatelopoulos, S. , author Sapsis, T.P. , year 2025 . title Can diffusion models capture extreme event statistics? journal Computer Methods in Applied Mechanics and Engineering volume 435 , pages 117589 . https://www.sciencedirect.com/science/article/pii/S004578252400...

    2025

  45. [54]

    , year 2001

    author Tsay, R. , year 2001 . title Analysis of Financial Time Series . publisher John Wiley & Sons

    2001

  46. [55]

    , author Le Clainche , S

    author Vega, J.M. , author Le Clainche , S. , year 2021 . title Higher Order Dynamic Mode Decomposition and Its Applications . publisher Academic Press . :https://doi.org/10.1016/B978-0-12-819743-1.00007-0

    2021 doi

  47. [56]

    , author Byeon, W

    author Vlachas, P.R. , author Byeon, W. , author Wan, Z.Y. , author Sapsis, T.P. , author Koumoutsakos, P. , year 2018 . title Data-driven forecasting of high-dimensional chaotic systems with long short-term memory networks . journal Proceedings of the Royal Society A: Mathema...

    2018

  48. [57]

    , year 2004

    author Vreman, A.W. , year 2004 . title An eddy-viscosity subgrid-scale model for turbulent shear flow: Algebraic theory and applications . journal Physics of Fluids volume 16 , pages 3670--3681 . https://doi.org/10.1063/1.1785131, :10.1063/1.1785131, http://arxiv.org/abs/http...

    2004 doi

  49. [58]

    , author Heaney, C

    author Xiao, D. , author Heaney, C. , author Mottet, L. , author Fang, F. , author Lin, W. , author Navon, I. , author Guo, Y. , author Matar, O. , author Robins, A. , author Pain, C. , year 2019 . title A reduced order model for turbulent flows in the urban environment using ...

    2019 doi

  50. [59]

    , author Zhou, G

    author Xu, L. , author Zhou, G. , author Zhao, F. , author Guo, Z. , author Zhang, K. , year 2022 . title A data-driven reduced order modeling for fluid flow analysis based on series forecasting intelligent algorithm . journal IEEE Access volume 10 , pages 60163--60176 . :10.1...

    2022

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.