Pith. sign in

REVIEW 3 major objections 5 minor 73 references

FlowDAS: A Stochastic Interpolant-based Framework for Data Assimilation

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read FlowDAS claims a learned stochastic interpolant of adjacent states, conditioned on observations at each step, replaces the physical simulator in data assimilation.

desk verdict FlowDAS applies stochastic interpolants to data assimilation with broad empirical success, but the observation conditioning in Algorithm 2 is a hand-tuned DPS-style step that does not discretize the derived conditional SDE. read the letter →

arxiv 2501.16642 v2 pith:6DXWIVZN submitted 2025-01-13 eess.SP cs.LGeess.IV

Siyi Chen , Yixuan Jia , Qing Qu , He Sun , Jeffrey A Fessler This is my paper · ORCID
classification eess.SPcs.LGeess.IV
keywords dataassimilationstochasticinterpolantsstatetransitionlearningobservationconditioninggenerativeSDEscore-baseddiffusionweatherforecastingLorenz-63
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

The paper tries to establish that a generative model of the one-step state transition, rather than the global noise-to-data map used by diffusion models, is the right engine for data assimilation when the underlying dynamics are unknown. It introduces FlowDAS, which trains a stochastic interpolant to reproduce $p(x_{k+1}\mid x_k)$ from trajectory data and then biases each interpolation step toward the incoming measurement with a Bayes-derived observation term. On Lorenz-63, incompressible Navier-Stokes super-resolution and sparse-observation tasks, and SEVIR weather forecasting, FlowDAS reports lower RMSE and better physical fidelity than score-based diffusion DA, neural-operator DA, and, where feasible, a particle filter. A sympathetic reader would care because the method removes the need to know the true dynamical equations while still delivering an ensemble of forecasts consistent with observations.

What carries the argument

The central object is the stochastic interpolant path $X_s=\alpha_s X_0+\beta_s X_1+\sigma_s W_s$ between consecutive states, which converts the unknown transition $p(x_{k+1}\mid x_k)$ into a drift-learning problem. The learned drift $b_s(X_s,X_0)$ is trained by regression against the path velocity $R_s$. Observation conditioning enters through the augmented drift in Equation (7), and the intractable term $\nabla\log p(y\mid X_s,X_0)$ is evaluated by Monte Carlo marginalization over $X_1$ using Milstein or stochastic Runge-Kutta integrators, softmax weights, and a gradient-descent step on $\sum_j w_j\|y-A(\hat X_1^{(j)})\|^2$. This mechanism lets a surrogate SDE act as the forward model in an otherwise standard DA loop.

What would settle it

Run FlowDAS and a bootstrap particle filter that uses the true Lorenz equations on identical 15-step observation windows and compare ensemble calibration. If FlowDAS's credible intervals cover the truth far below the nominal rate while the particle filter's intervals are on target, the approximate conditioning is not producing the claimed posterior.

Watch

Extended reading notes

Core claim

FlowDAS claims that data assimilation can be run with a learned stochastic transition model. The state transition $p(x_{k+1}\mid x_k)$ is approximated by an interpolant SDE whose drift $b_s(X_s,X_0)$ is learned by matching the path velocity $\dot{\alpha}_s X_0+\dot{\beta}_s X_1+\dot{\sigma}_s W_s$. At inference, the drift is augmented to $b_s(X_s,y,X_0)=b_s(X_s,X_0)+\nabla\log p(y\mid X_s,X_0)/(\lambda_s\beta_s)$, so the generated state is pulled toward the observation. Because the observation only links through $X_1=x_{k+1}$, the conditional score is approximated by Monte Carlo samples of $X_1$ drawn with low-order stochastic integrators, weighted by $p(y\mid X_1)$, and applied as a gradient step on the squared observation misfit. The paper reports that this step-by-step, observation-conditioned rollout outperforms diffusion, neural-operator, and model-based baselines on all three benchmark families.

Load-bearing premise

The results stand on the assumption that the approximate gradient-descent conditioning step in Algorithm 2 really samples from the observation-conditioned distribution of the learned SDE, with a step size chosen by hand.

Editorial extensions

If this is right

  • Data assimilation can proceed without the true transition model; only trajectory data and the observation operator are needed.
  • Forecasts come as ensembles that are already consistent with the incoming observation, because conditioning happens at every interpolation step rather than in a separate post-hoc filter update.
  • The autoregressive rollout makes long-horizon prediction feasible in high-dimensional systems, where global diffusion generation degrades.
  • On the reported benchmarks, FlowDAS improves RMSE and CSI scores over SDA, FNO-DA, and Transolver-DA, and is competitive with a true-dynamics particle filter on Lorenz-63.
  • Conditioning on several previous states lets the framework handle non-Markovian dynamics, which the weather-forecasting experiment exploits.

Reading between the lines

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

  • The conditioning step is a hand-tuned gradient descent on the observation misfit, so the claimed posterior fidelity rests on the tuning of $\zeta_n$ and on the accuracy of the low-order integrators used to draw $\hat X_1^{(j)}$.
  • Calibration is the decisive test not reported in the paper: comparing FlowDAS ensemble coverage against a true-dynamics particle filter would show whether the approximate conditioning is doing Bayesian work or merely steering samples toward observations.
  • The same observation-guidance mechanism should extend to non-Gaussian likelihoods by replacing the squared misfit with the true log-likelihood, since the derivation only needs $p(y\mid X_1)$.
  • Treating $\zeta_n$ as an annealed likelihood-tempering schedule rather than a constant may improve stability on chaotic systems, which the paper does not explore.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes FlowDAS, a data assimilation framework built on stochastic interpolants. It learns the one-step state transition p(x_{k+1}|x_k) by training a drift network on paired consecutive states, then assimilates observations at inference time by modifying the interpolant drift with a term derived from Bayes' rule, Eq. (7). The implementation (Algorithm 2) replaces that conditional drift with a separate gradient-descent step that penalizes observation misfit of Monte Carlo forward samples. The framework is evaluated on Lorenz-63, incompressible Navier-Stokes super-resolution and sparse-observation tasks, particle image velocimetry, and weather forecasting on SEVIR, where it reports lower RMSE and better physical plausibility than neural-operator, score-based, and model-driven baselines.

Significance. If the posterior property claimed in Eq. (7) were faithfully realized, FlowDAS would be a valuable contribution: it offers a learnable, stochastic, autoregressive surrogate that can incorporate observations without retraining or a separate filtering step, and the experiments cover a credible range of low- and high-dimensional problems. The paper is generally well written, ships source code, and the empirical comparison is extensive, including ablations on the Monte Carlo sample count and on the order of the posterior sampler. The main limitation is that the implemented inference algorithm does not implement the derived conditional SDE, so the central claim of 'measurement-consistent forecasts' currently rests on a tuned heuristic rather than on the Bayes-conditioning derivation.

major comments (3)
  1. [Section 3.1, Eq. (7), and Algorithm 2, lines 8–10] The implemented observation guidance is not the conditional drift of Eq. (7). After the unconditional SDE step, Algorithm 2 applies a separate gradient step −ζ_n ∇_{X_{s_n}} Σ_j w_j ∥y − A(X̂_1^(j))∥², with a step size ζ_n that is hand-set per experiment (Table S.4: 0.0002 for Lorenz, 1–2 for Navier-Stokes, 0.1 for weather), while the observation noise γ varies only from 0.05 to 0.25. The step size is not tied to the theoretical scale 1/(λ_s β_s), to γ, or to the integrator step (Δs)_n, and the gradient is evaluated after the SDE increment rather than as an infinitesimal modification of the drift. Consequently, the argument that the generated states approximate p(x_{k+1}|x_k, y_{k+1}) is not supported by the derivation. Please either implement a discretization that faithfully matches Eq. (7), or provide empirical evidence that the heuristic produces samples from the target posterior, for example by comparing with a particle filter or analytic posterior on a low-dimensional problem and by reporting rank histograms or coverage intervals.
  2. [Section 3.1, Eqs. (9)–(10), and Section C.2] The Monte Carlo samples X̂_1^(j) used to compute the likelihood weights w_j are drawn from low-order integrators with bias O((1−s)²) and O((1−s)³). Near s ≈ 0, where most of the interpolation steps in the autoregressive rollout occur, the bias is largest and may substantially mis-weight the samples in Eq. (8). The ablation in Table S.7 shows that the 2nd-order method outperforms the 1st-order one, but it does not quantify the remaining bias relative to the conditional posterior. Please add an experiment with a higher-order integrator (e.g., RK4 or a finer grid) to show that the results are not an artifact of the biased sampler.
  3. [Table S.7 and Section B.1, Algorithm 2] The large improvement from 'No correction' (RMSE 0.206) to the guided result (RMSE 0.038) on the Navier-Stokes task is presented as evidence of the benefit of observation conditioning. Because the guidance is a separately tuned gradient step, this improvement could in principle arise from the particular choice of ζ_n rather than from any Bayes-consistent conditioning. Please provide a sensitivity analysis with respect to ζ (e.g., sweeping it over a range or setting it according to a theoretically derived schedule) and, if possible, compare against a variant that implements Eq. (7) directly; this would clarify whether the reported gains reflect faithful posterior conditioning or step-size tuning.
minor comments (5)
  1. [Table 2 caption] The word 'Naiver-Stokes' should be 'Navier-Stokes'.
  2. [Section 4.1, after Eq. (12)] The typo 'wehre' should be 'where'.
  3. [Appendix A.2, Eq. (S.20)] The notation ∇ log p(X_s, X_0) should be ∇ log p(X_s | X_0), consistent with the text that introduces this equation.
  4. [Algorithm 2, line 9] The softmax weights are computed from ∥y − A(X̂_1^(j))∥², which corresponds to exp(−∥y − A(X)∥²) rather than the Gaussian likelihood exp(−∥y − A(X)∥²/(2γ²)). Since γ varies across experiments (0.05 to 0.25), the relative weighting is not consistent with the stated observation model; please include the missing 1/(2γ²) factor or justify its omission.
  5. [Section 4.3, paragraph after 'Baselines and metrics'] The header 'InplementationDetails' contains a typo; it should be 'Implementation Details'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the observation-conditioned drift is derived from Bayes rule applied to the learned stochastic interpolant, and no fitted parameter is renamed as a prediction.

full rationale

The paper's central derivation is self-contained. The unconditional drift b_s is learned from consecutive-state pairs by regression to the interpolant velocity R_s (Eq. 5; Appendix B.1). The conditional drift in Eq. (7) follows by Bayes rule from the learned unconditional drift and the observation score, as detailed in Appendix A.1 with the drift-score identity in Eq. (6)/S.20. The conditional score is approximated by Monte Carlo marginalization over X_1 with softmax weights (Eq. 8 and Appendix A.3), and the samples X_hat_1^(j) come from standard low-order SDE integrators (Eqs. 9–10). No step defines its target in terms of its own input: predicted states are evaluated against ground-truth simulated states, not against the training pairs or the loss constants that define the method. The observation-guidance step size zeta_n in Table S.4 is a hand-set inference hyperparameter; it is tuned per task but is not a fitted parameter renamed as a prediction, and the paper does not claim to predict that quantity. The only author-overlapping citations are [4] for the minor remark that fine-tuning zeta_n can slightly improve performance, and [14, 65] cited in passing; none is load-bearing, and there is no imported uniqueness theorem or ansatz smuggled in through self-citation. The discrepancy between Eq. (7) and Algorithm 2's split-step gradient update is an implementation-approximation concern about posterior fidelity, which belongs under correctness risk rather than circularity.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The free parameters are inference hyperparameters tuned per experiment; the axioms are standard interpolant theory plus the stated DA data-generation assumptions and the empirical assumption of generalization. No new physical entities are introduced; the observation guidance is an algorithmic procedure, not a new mediator or latent dimension.

free parameters (4)
  • Monte Carlo sample count J = 21 (Lorenz), 25 (NS), 25 (weather)
    Number of posterior samples in the conditional score estimate; Table S.4; Table S.6 shows RMSE varies with J.
  • Guidance step size ζ_n = 0.0002 (Lorenz), 1-1.75 (NS), 0.1 (weather)
    Strength of observation gradient; tuned per task (Table S.4). Its scaling is not derived from Eq (7), so the observation guidance is partly heuristic.
  • Number of conditioning states L = 1 (Lorenz), 10 (NS), 6 (weather)
    Context length for non-Markovian dynamics; chosen per task.
  • Interpolant coefficients α_s, β_s, σ_s = not specified in the paper; standard choices from [16]
    Path shape between states; chosen by hand and affects the learned drift. The paper only states boundary conditions (Section 2.2).
assumptions (5)
  • standard math Stochastic interpolants (SDE (4) with drift from (5)) correctly transport X_0 to X_1 with the distribution of the data
    Foundational guarantee of the generative surrogate; cited from [1,2,16], used throughout Section 3.1.
  • domain assumption The data obey the state-space model x_{k+1}=Ψ(x_k)+ξ_k, y_{k+1}=A(x_{k+1})+η with Gaussian noise and known A and γ
    Defines the DA problem and the likelihood p(y|X_1) used in Eq (8).
  • standard math The drift-score relation b_s(X_s,X_0)=c_s/β_s+∇ log p(X_s|X_0)/(λ_s β_s) holds
    Bridges the learned drift to the posterior score in Eq (7); taken from [16].
  • domain assumption Low-order SDE integrators (Eqs 9-10) sample p(X_1|X_s,X_0) accurately enough
    The MC estimate (8) and the whole guidance procedure rely on these samples; the paper acknowledges the resulting bias in Section C.2.
  • domain assumption The training trajectories are representative of the test distribution and the neural drift generalizes
    For weather, the dynamics are unknown, so the entire forecast quality rests on the generalization of the learned drift to unseen storm events.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FlowDAS: A Stochastic Interpolant-based Framework for Data Assimilation." pith.science (2026). https://pith.science/paper/6DXWIVZN

@misc{pith2026250116642,
  author       = {Pith},
  title        = {Pith review of: FlowDAS: A Stochastic Interpolant-based Framework for Data Assimilation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6DXWIVZN}},
  note         = {Machine review of arXiv:2501.16642}
}
read the original abstract

Data assimilation (DA) integrates observations with a dynamical model to estimate states of PDE-governed systems. Model-driven methods (e.g., Kalman, particle) presuppose full knowledge of the true dynamics, which is not always satisfied in practice, while purely data-driven solvers learn a deterministic mapping between observations and states and therefore miss the intrinsic stochasticity of real processes. Recently, score-based diffusion models learn a global diffusion prior and provide a good modeling of the stochastic dynamics, showing new potential for DA. However, their all-at-once generation rather than step-by-step transition limits their performance when dealing with highly complex stochastic processes and lacks physical interpretability. To tackle these drawbacks, we introduce FlowDAS, a generative DA framework that uses stochastic interpolants to directly learn state transition dynamics and achieve step-by-step transition to better model the real dynamics. We also improve the framework by combining the observation, better suiting the DA settings. Directly learning the underlying dynamics from collected data removes restrictive dynamical assumptions, and conditioning on observations at each interpolation step yields stable, measurement-consistent forecasts. Experiments on Lorenz-63, Navier-Stokes super-resolution/sparse-observation scenarios, and large-scale weather forecasting -- where dynamics are partly or wholly unknown -- show that FlowDAS surpasses model-driven methods, neural operators, and score-based baselines in accuracy and physical plausibility.

Figures

Figures reproduced from arXiv: 2501.16642 by the authors.

Figure 1
Figure 1. An overview of FlowDAS. We introduce a stochastic interpolant-based framework for data assimilation, named FlowDAS, to estimate states xL:K from the noisy (sparse or low￾resolution) observations yL:K. FlowDAS models the stochastic dynamics of the system with a flow-based stochastic differential equation (SDE) and incorporates the observations to improve the prediction accuracy. On the right, we show a conceptual ill… view at source ↗
Figure 2
Figure 2. Data assimilation of Lorenz 1963 system. FlowDAS achieved results comparable to the state-of-the-art model-based BPF method, significantly outperforming the data-driven SDA method in recovering the underlying dynamics of this chaotic system. This highlights the efficiency and robustness of FlowDAS in capturing complex, nonlinear dynamics while maintaining accuracy and stability. The variables x1, x2 and x3 correspon… view at source ↗
Figure 3
Figure 3. Data assimilation of incompressible Navier-Stokes flow. The positive values (red) of the state, i.e., vorticity field, indicate clockwise rotation and negative values (blue) indicate counter-clockwise rotation. FlowDAS achieved results with more accurate details and higher ac￾curacy than all baselines, showing the efficiency of FlowDAS in tackling DA tasks with highly non-linear complex systems. Additionally, FlowDA… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Data assimilation of weather forecasting on SEVIR Vertical Integrated Liquid dataset under sparse observations. All DA models take previous six states (displayed in the first row; t−50 min to t min) as conditions and estimate the future state at t + 10 min. FlowDAS pre…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

73 extracted references · 59 canonical work pages

  1. [1]

    M. S. Albergo and E. Vanden-Eijnden. Building normalizing flows with stochastic inter- polants.arXiv preprint arXiv:2209.15571, 2022

    arXiv 2022

  2. [2]

    Stochasticinterpolants: Aunifyingframe- work for flows and diffusions.arXiv preprint arXiv:2303.08797, 2023

    M.S.Albergo,N.M.Boffi,andE.Vanden-Eijnden. Stochasticinterpolants: Aunifyingframe- work for flows and diffusions.arXiv preprint arXiv:2303.08797, 2023

    arXiv 2023

  3. [3]

    M. Asch, M. Bocquet, and M. Nodet.Data assimilation: methods, algorithms, and applications. SIAM, 2016

    work page 2016

  4. [4]

    W. Bai, S. Chen, W. Chen, and H. Sun. Blind inversion using latent diffusion priors, 2024. URLhttps://arxiv.org/abs/2407.01027. 12

    arXiv 2024

  5. [5]

    Banerjee, Y

    A. Banerjee, Y. van Dinther, and F. C. Vossepoel. On parameter bias in earthquake sequence models using data assimilation.Nonlinear Processes in Geophysics, 30(2):101–115, 2023

    work page 2023

  6. [6]

    R. N. Bannister. A review of operational methods of variational and ensemble-variational data assimilation.Quarterly Journal of the Royal Meteorological Society, 143(703):607–633, 2017

    work page 2017

  7. [7]

    F. Bao, Z. Zhang, and G. Zhang. A score-based nonlinear filter for data assimilation, 2023. URLhttps://arxiv.org/abs/2306.09282

    arXiv 2023

  8. [8]

    Bickel, B

    P. Bickel, B. Li, and T. Bengtsson. Sharp failure rates for the bootstrap particle filter in high dimensions. InPushing the limits of contemporary statistics: Contributions in honor of Jayanta K. Ghosh, volume 3, pages 318–330. Institute of Mathematical Statistics, 2008

    work page 2008

Show all 73 references
  1. [9]

    Beyondgaussianstatisticalmodelingingeophysicaldata assimilation.Monthly Weather Review, 138(8):2997–3023, 2010

    M.Bocquet,C.A.Pires,andL.Wu. Beyondgaussianstatisticalmodelingingeophysicaldata assimilation.Monthly Weather Review, 138(8):2997–3023, 2010

    2010

  2. [10]

    Bocquet, H

    M. Bocquet, H. Elbern, H. Eskes, M. Hirtl, R. Žabkar, G. Carmichael, J. Flemming, A. Inness, M. Pagowski, J. Pérez Camaño, et al. Data assimilation in atmospheric chemistry models: current status and future prospects for coupled chemistry meteorology models.Atmospheric chemist...

    2015

  3. [11]

    S. L. Brunton and B. R. Noack. Closed-loop turbulence control: Progress and challenges. Applied Mechanics Reviews, 67(5):050801, 2015

    2015

  4. [12]

    Carrassi, M

    A. Carrassi, M. Bocquet, L. Bertino, and G. Evensen. Data assimilation in the geosciences: An overview of methods, issues, and perspectives.Wiley Interdisciplinary Reviews: Climate Change, 9(5):e535, 2018

    2018

  5. [13]

    C.-Y. Chen, R. Antón, M.-y. Hung, P. Menon, E. A. Finol, and K. Pekkan. Effects of intralumi- nalthrombusonpatient-specificabdominalaorticaneurysmhemodynamicsviastereoscopic particle image velocity and computational fluid dynamics modeling.Journal of biomechanical engineering,...

    2014

  6. [14]

    J. Chen, P. Li, Y. Wang, P.-C. Ku, and Q. Qu. Sim2Real in reconstructive spectroscopy: Deep learning with augmented device-informed data simulation.APL Machine Learning, 2(3): 036106, 08 2024. ISSN 2770-9019. doi: 10.1063/5.0209339. URLhttps://doi.org/10.1063/ 5.0209339

    2024 doi

  7. [15]

    K. Chen, P. Ye, H. Chen, T. Han, W. Ouyang, T. Chen, L. BAI, et al. Fnp: Fourier neural processesforarbitrary-resolutiondataassimilation.AdvancesinNeuralInformationProcessing Systems, 37:137847–137872, 2024

    2024

  8. [16]

    Y. Chen, M. Goldstein, M. Hua, M. S. Albergo, N. M. Boffi, and E. Vanden-Eijnden. Proba- bilistic forecasting with stochastic interpolants and Föllmer processes. In R. Salakhutdinov, Z. Kolter, K. Heller, A. Weller, N. Oliver, J. Scarlett, and F. Berkenkamp, editors,Proceeding...

    2024

  9. [17]

    Chung, J

    H. Chung, J. Kim, M. T. Mccann, M. L. Klasky, and J. C. Ye. Diffusion posterior sampling for general noisy inverse problems.arXiv preprint arXiv:2209.14687, 2022. 13

    2022 arXiv

  10. [18]

    J. A. Cummings. Operational multivariate ocean data assimilation.Quarterly Journal of the RoyalMeteorologicalSociety: Ajournaloftheatmosphericsciences,appliedmeteorologyandphysical oceanography, 131(613):3583–3604, 2005

    2005

  11. [19]

    J. A. Cummings and O. M. Smedstad. Variational data assimilation for the global ocean. InData assimilation for atmospheric, oceanic and hydrologic applications (Vol. II), pages 303–343. Springer, 2013

    2013

  12. [20]

    Duraisamy, G

    K. Duraisamy, G. Iaccarino, and H. Xiao. Turbulence modeling in the age of data.Annual review of fluid mechanics, 51(1):357–377, 2019

    2019

  13. [21]

    F. G. Ergin, B. B. Watz, and N. F. Gade-Nielsen. A review of planar PIV systems and image processing tools for lab-on-chip microfluidics.Sensors, 18(9):3090, 2018

    2018

  14. [22]

    G. Evensen. The ensemble Kalman filter: Theoretical formulation and practical implementa- tion.Ocean dynamics, 53:343–367, 2003

    2003

  15. [23]

    S. J. Fletcher.Data assimilation for the geosciences: From theory to application. Elsevier, 2017

    2017

  16. [24]

    Prediff: Pre- cipitation nowcasting with latent diffusion models.Advances in Neural Information Processing Systems, 36:78621–78656, 2023

    Z.Gao,X.Shi,B.Han,H.Wang,X.Jin,D.Maddix,Y.Zhu,M.Li,andY.B.Wang. Prediff: Pre- cipitation nowcasting with latent diffusion models.Advances in Neural Information Processing Systems, 36:78621–78656, 2023

    2023

  17. [25]

    A. J. Geer, K. Lonitz, P. Weston, M. Kazumori, K. Okamoto, Y. Zhu, E. H. Liu, A. Collard, W.Bell,S.Migliorini,etal. All-skysatellitedataassimilationatoperationalweatherforecast- ing centres.Quarterly Journal of the Royal Meteorological Society, 144(713):1191–1217, 2018

    2018

  18. [26]

    N. J. Gordon, D. J. Salmond, and A. F. Smith. Novel approach to nonlinear/non-Gaussian Bayesianstateestimation. InIEEproceedingsF(radarandsignalprocessing),volume140,pages 107–113. IET, 1993

    1993

  19. [27]

    Gustafsson, T

    N. Gustafsson, T. Janjić, C. Schraff, D. Leuenberger, M. Weissmann, H. Reich, P. Brousseau, T. Montmerle, E. Wattrelot, A. Bučánek, et al. Survey of data assimilation methods for convective-scale numerical weather prediction at operational centres.Quarterly Journal of the Roya...

    2018

  20. [28]

    Huang, L

    L. Huang, L. Gianinazzi, Y. Yu, P. D. Dueben, and T. Hoefler. DiffDA: a diffusion model for weather-scaledataassimilation. InR.Salakhutdinov,Z.Kolter,K.Heller,A.Weller,N.Oliver, J. Scarlett, and F. Berkenkamp, editors,Proceedings of the 41st International Conference on Ma- chi...

    2024

  21. [29]

    Julier and J

    S. Julier and J. Uhlmann. Unscented filtering and nonlinear estimation.Proceedings of the IEEE, 92(3):401–422, 2004. doi: 10.1109/JPROC.2003.823141

    2004

  22. [30]

    PIVmeasurementsofflowwithinplugsinamicrochannel

    C.King,E.Walsh,andR.Grimes. PIVmeasurementsofflowwithinplugsinamicrochannel. Microfluidics and Nanofluidics, 3:463–472, 2007

    2007

  23. [31]

    Koschatzky, P

    V. Koschatzky, P. Moore, J. Westerweel, F. Scarano, and B. Boersma. High speed PIV applied to aerodynamic noise investigation.Experiments in fluids, 50:863–876, 2011

    2011

  24. [32]

    Lagemann, K

    C. Lagemann, K. Lagemann, S. Mukherjee, and W. Schröder. Deep recurrent optical flow learning for particle image velocimetry data.Nature Machine Intelligence, 3:641 – 651, 2021. URLhttps://api.semanticscholar.org/CorpusID:237869288. 14

    2021

  25. [33]

    Z. Li, N. Kovachki, K. Azizzadenesheli, B. Liu, K. Bhattacharya, A. Stuart, and A. Anand- kumar. Fourier neural operator for parametric partial differential equations, 2021. URL https://arxiv.org/abs/2010.08895

    2021 arXiv

  26. [34]

    E. N. Lorenz. Deterministic nonperiodic flow.Journal of the Atmospheric Sciences, 20:130–141,

  27. [35]

    Loshchilov and F

    I. Loshchilov and F. Hutter. Decoupled weight decay regularization. InInternational Confer- ence on Learning Representations, 2017. URLhttps://api.semanticscholar.org/CorpusID: 53592270

    2017

  28. [36]

    N.A.Z.M.Noar,N.I.A.Apandi,andN.Rosli. AcomparativestudyofTaylormethod,fourth order Runge-Kutta method and Runge-Kutta Fehlberg method to solve ordinary differential equations.AIPConferenceProceedings,2895(1):020003,032024. ISSN0094-243X. doi: 10.1063/ 5.0192085. URLhttps://doi...

    doi

  29. [37]

    Özcan, Ö

    C. Özcan, Ö. Kocatürk, C. Işlak, and C. Öztürk. Integrated particle image velocimetry and fluid–structure interaction analysis for patient-specific abdominal aortic aneurysm studies. BioMedical Engineering OnLine, 22(1):113, 2023

    2023

  30. [38]

    Peyret.Spectral methods for incompressible viscous flow, volume 148

    R. Peyret.Spectral methods for incompressible viscous flow, volume 148. Springer, 2002

    2002

  31. [39]

    Y. Qu, J. Nathaniel, S. Li, and P. Gentine. Deep generative data assimilation in multimodal setting. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 449–459, 2024

    2024

  32. [40]

    F. Rabier. Overview of global data assimilation developments in numerical weather- prediction centres.Quarterly Journal of the Royal Meteorological Society: A journal of the atmo- spheric sciences, applied meteorology and physical oceanography, 131(613):3215–3233, 2005

    2005

  33. [41]

    Reich and C

    S. Reich and C. Cotter.Probabilistic forecasting and Bayesian data assimilation. Cambridge Uni- versity Press, 2015

    2015

  34. [42]

    Dataassimilationmethodsintheearthsciences.Advancesinwaterresources, 31 (11):1411–1418, 2008

    R.H.Reichle. Dataassimilationmethodsintheearthsciences.Advancesinwaterresources, 31 (11):1411–1418, 2008

    2008

  35. [43]

    M. I. Ribeiro. Kalman and extended kalman filters: Concept, derivation and properties.In- stitute for Systems and Robotics, 43(46):3736–3741, 2004

    2004

  36. [44]

    Rodell, P

    M. Rodell, P. Houser, U. Jambor, J. Gottschalck, K. Mitchell, C.-J. Meng, K. Arsenault, B. Cos- grove, J. Radakovich, M. Bosilovich, et al. The global land data assimilation system.Bulletin of the American Meteorological society, 85(3):381–394, 2004

    2004

  37. [45]

    Rozet and G

    F. Rozet and G. Louppe. Score-based data assimilation.Advances in Neural Information Pro- cessing Systems, 36:40521–40541, 2023

    2023

  38. [46]

    G. S. Seabra, N. T. Mücke, V. L. S. Silva, D. Voskov, and F. C. Vossepoel. AI enhanced data assimilation and uncertainty quantification applied to geological carbon storage.ArXiv, abs/2402.06110, 2024. URLhttps://api.semanticscholar.org/CorpusID:267616732

    2024

  39. [47]

    Stochasticprocessesandfilteringtheory.IEEETransactionsonAutomaticControl,17 (5):752–753, 1972

    K.Senne. Stochasticprocessesandfilteringtheory.IEEETransactionsonAutomaticControl,17 (5):752–753, 1972. doi: 10.1109/TAC.1972.1100136. 15

    1972

  40. [48]

    Y. Shi, V. D. Bortoli, G. Deligiannidis, and A. Doucet. Conditional simulation using diffusion schrödinger bridges, 2022. URLhttps://arxiv.org/abs/2202.13460

    2022 arXiv

  41. [49]

    Shysheya, C

    A. Shysheya, C. Diaconu, F. Bergamin, P. Perdikaris, J. M. Hernández-Lobato, R. Turner, and E. Mathieu. On conditional diffusion models for pde simulations.Advances in Neural Infor- mation Processing Systems, 37:23246–23300, 2024

    2024

  42. [50]

    S. Simic. On a global upper bound for jensen’s inequality.Journal of mathematical analysis and applications, 343(1):414–419, 2008

    2008

  43. [51]

    E.J.Stamhuis.Basicsandprinciplesofparticleimagevelocimetry(PIV)formappingbiogenic and biologically relevant flows.Aquatic Ecology, 40(4):463–479, 2006

    2006

  44. [52]

    Süli and D

    E. Süli and D. F. Mayers.An introduction to numerical analysis. Cambridge university press, 2003

    2003

  45. [53]

    Taira, S

    Y. Taira, S. Sagara, and M. Oya. Model-based motion control for underwater vehicle- manipulatorsystemswithoneofthethreetypesofservosubsystems.Artificiallifeandrobotics, 25:133–148, 2020

    2020

  46. [54]

    Analysisofflowpatternsina patient-specific thoracic aortic aneurysm model.Computers & Structures, 87(11-12):680–690, 2009

    F.Tan,A.Borghi,R.Mohiaddin,N.Wood,S.Thom,andX.Xu. Analysisofflowpatternsina patient-specific thoracic aortic aneurysm model.Computers & Structures, 87(11-12):680–690, 2009

    2009

  47. [55]

    Z. J. Taylor, R. Gurka, G. A. Kopp, and A. Liberzon. Long-duration time-resolved PIV to studyunsteadyaerodynamics.IEEETransactionsonInstrumentationandMeasurement,59(12): 3262–3269, 2010

    2010

  48. [56]

    Nonlineardataassimilationingeosciences: anextremelyefficientparticle filter.Quarterly Journal of the Royal Meteorological Society, 136(653):1991–1999, 2010

    P.J.VanLeeuwen. Nonlineardataassimilationingeosciences: anextremelyefficientparticle filter.Quarterly Journal of the Royal Meteorological Society, 136(653):1991–1999, 2010

    1991

  49. [57]

    Van Oudheusden

    B. Van Oudheusden. PIV-based pressure measurement.Measurement Science and Technology, 24(3):032001, 2013

    2013

  50. [58]

    Veillette, S

    M. Veillette, S. Samsi, and C. Mattioli. Sevir: A storm event imagery dataset for deep learn- ing applications in radar and satellite meteorology.Advances in Neural Information Processing Systems, 33:22009–22019, 2020

    2020

  51. [59]

    Villani.The Wasserstein distances, pages 93–111

    C. Villani.The Wasserstein distances, pages 93–111. Springer Berlin Heidelberg, Berlin, Hei- delberg, 2009. ISBN 978-3-540-71050-9. doi: 10.1007/978-3-540-71050-9_6. URLhttps: //doi.org/10.1007/978-3-540-71050-9_6

    2009 doi

  52. [60]

    Voleti, A

    V. Voleti, A. Jolicoeur-Martineau, and C. Pal. Mcvd-masked conditional video diffusion for prediction,generation,andinterpolation.Advancesinneuralinformationprocessingsystems,35: 23371–23385, 2022

    2022

  53. [61]

    Dataassimilationanditsapplications.ProceedingsoftheNational Academy of Sciences, 97(21):11143–11144, 2000

    B.Wang,X.Zou,andJ.Zhu. Dataassimilationanditsapplications.ProceedingsoftheNational Academy of Sciences, 97(21):11143–11144, 2000

    2000

  54. [62]

    Wanner and I

    M. Wanner and I. Mezić. Robust approximation of the stochastic koopman operator, 2022. URLhttps://arxiv.org/abs/2011.00078. 16

    2022 arXiv

  55. [63]

    M. J. Werner, K. Ide, and D. Sornette. Earthquake forecasting based on data assimilation: Sequentialmontecarlomethodsforrenewalprocesses,2009. URLhttps://arxiv.org/abs/ 0908.1516

    2009 arXiv

  56. [64]

    H. Wu, H. Luo, H. Wang, J. Wang, and M. Long. Transolver: A fast transformer solver for pdes on general geometries, 2024. URLhttps://arxiv.org/abs/2402.02366

    2024 arXiv

  57. [65]

    Xiang, W

    Y. Xiang, W. Jin, H. Dong, M. Bai, Z. Fang, P. Zhao, H. Sun, K. Thambiratnam, Q. Zhang, and X. Huang. Adaf: An artificial intelligence data assimilation framework for weather forecast- ing.arXiv preprint arXiv:2411.16807, 2024

    2024 arXiv

  58. [66]

    X. Xu, X. Sun, W. Han, X. Zhong, L. Chen, Z. Gao, and H. Li. Fuxi-da: A generalized deep learning data assimilation framework for assimilating satellite observations.npj Climate and Atmospheric Science, 8(1):156, 2025

    2025

  59. [67]

    Yasuda and R

    Y. Yasuda and R. Onishi. Spatio-temporal super-resolution data assimilation (srda) utiliz- ing deep neural networks with domain generalization.Journal of Advances in Modeling Earth Systems, 15(11):e2023MS003658, 2023

    2023

  60. [68]

    L. Yu, Y. Cheng, K. Sohn, J. Lezama, H. Zhang, H. Chang, A. G. Hauptmann, M.-H. Yang, Y. Hao, I. Essa, and L. Jiang. Magvit: Masked generative video transformer. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 10459– 10469, June 2023

    2023

  61. [69]

    Zhang, J

    Z. Zhang, J. Hu, W. Cheng, D. Paudel, and J. Yang. Extdm: Distribution extrapolation diffu- sionmodelforvideoprediction. InProceedingsoftheIEEE/CVFConferenceonComputerVision and Pattern Recognition (CVPR), pages 19310–19320, June 2024

    2024

  62. [70]

    Zhao and L

    M. Zhao and L. Jiang. Data-driven probability density forecast for stochastic dynamical sys- tems, 2022. URLhttps://arxiv.org/abs/2210.03418. 17 Appendices A Mathematical Derivation A.1 Conditional Drift Stochastic interpolants approximate the state transitionp(xk+1 |x k)by in...

    2022 arXiv

  63. [72]

    1st-order

    Thestatetransitionswithineachtrajectoryfollow the dynamics defined in Equation (1). 2.Consecutive state pairs formation.For each trajectoryt, form two aligned sequences: (a)x t 0:K−1: The original sequence of states with each last statext K discarded. (b)x t 1:K: The sequence ...

    1963

  64. [73]

    The model is trained using the Adam optimizer with a base learning rate of0.005, and a linear learning rate scheduler is applied

    For the conditionX0 and timesteps, we use embeddings of dimension 4. The model is trained using the Adam optimizer with a base learning rate of0.005, and a linear learning rate scheduler is applied. Training is conducted over23000epochs. 27 Figure S.7:Forecaseting dynamics mod...

    1963

  65. [1963]

    URLhttps://api.semanticscholar.org/CorpusID:15359559

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.