Pith. sign in

REVIEW 4 major objections 5 minor 44 references

Sensitivity-Constrained Fourier Neural Operators for Forward and Inverse Problems in Parametric Differential Equations

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The paper claims that training Fourier Neural Operators with an additional loss on parameter sensitivities yields surrogates whose gradients are accurate and whose parameter-inversion performance is far better than plain FNO or FNO-PINN…

desk verdict A practical, well-executed study showing that supervising parameter Jacobians makes FNOs dramatically better at inversion and robustness; the main gap is unvalidated PDE sensitivity labels, which is fixable. read the letter →

arxiv 2505.08740 v3 pith:WC3BSQL3 submitted 2025-05-13 cs.LG cs.CE

Abdolmehdi Behroozi , Chaopeng Shen and , Daniel Kifer This is my paper · ORCID
classification cs.LGcs.CE
keywords FourierNeuralOperatorsparameterinversionsensitivityanalysisJacobiansupervisionparametricdifferentialequationsphysics-informedregularizationdifferentiablesolvers
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 argues that a neural operator can learn solution paths $u(t)$ very accurately while still misrepresenting how those solutions depend on the parameters $p$, and that this hidden gradient error is what makes plain Fourier Neural Operators poor at inverse problems. To fix this, the authors add a sensitivity loss that compares the model's Jacobian $\partial \hat u/\partial p$, obtained by automatic differentiation, with true Jacobians $\partial u/\partial p$ computed once from a differentiable solver or by finite differences. The resulting Sensitivity-Constrained FNO (SC-FNO) keeps forward accuracy and lifts sensitivity $R^2$ values from roughly 0.2-0.8 to 0.91-0.99 across ODEs, a damped wave equation, forced Burgers, Navier-Stokes vorticity, and Allen-Cahn test cases. It also improves multi-parameter inversion (for example, from $R^2=0.657$ to $R^2=0.987$ for one PDE1 parameter), works with up to 82 parameters, needs less training data, and stays accurate when test parameters are perturbed beyond the training range. A sympathetic reader would take this as evidence that explicit supervision of parameter sensitivities is a simple and broadly applicable regularizer for learned surrogate models.

What carries the argument

The load-bearing object is the sensitivity loss $L_s = \frac{1}{M}\sum_{j=1}^{M} \|\partial \hat u(x_j,t_j;p)/\partial p - \partial u(x_j,t_j;p)/\partial p\|^2$, where $\partial \hat u/\partial p$ is the Jacobian of the network output with respect to the input parameters, obtained by automatic differentiation through the operator, and $\partial u/\partial p$ is a precomputed true Jacobian from a differentiable numerical solver or a fourth-order finite-difference approximation. The model itself is a standard Fourier Neural Operator, a network that learns maps between function spaces using learnable convolutions in the Fourier domain; the only change is that training minimizes the usual solution loss plus this Jacobian-matching term, with the equation-based PINN loss optional. The sensitivity term is what forces the network to use the parameters $p$ correctly, and the paper's experiments isolate it as the source of the inversion and extrapolation improvements.

What would settle it

Take one of the paper's PDE cases, say PDE2, and recompute the Jacobian labels with two independent methods, such as Richardson-extrapolated finite differences on the original solver and an adjoint solver; if SC-FNO's advantage over FNO in sensitivity $R^2$ or inversion $R^2$ largely disappears when the labels change, then the improvement is an artifact of the label-generation pipeline rather than of the sensitivity constraint itself.

Watch

Extended reading notes

Core claim

The central claim is that Fourier Neural Operators can match solution paths while badly mispredicting the parameter sensitivities $\partial u/\partial p$, and this sensitivity mismatch is the main reason they fail at parameter inversion and under extrapolated parameters. Adding the sensitivity loss $L_s$ fixes that: SC-FNO keeps solution accuracy comparable to FNO while raising sensitivity $R^2$ from about 0.2-0.8 to 0.91-0.99 in the paper's test beds, and in simultaneous multi-parameter inversion it raises $R^2$ from 0.657 to 0.987 for the parameter c of PDE1 and keeps all PDE2 parameters above 0.96 where FNO hovers near 0.85. The paper also reports that the physics-informed equation loss in FNO-PINN gives only small sensitivity gains, with $R^2$ below 0.52 for most PDE2 gradients, so the sensitivity term is doing the work. The same pattern holds for ODEs, for Navier-Stokes vorticity, for Allen-Cahn near bifurcation, for other neural operator architectures, and for sensitivity labels produced by finite differences rather than automatic differentiation.

Load-bearing premise

The training targets that teach the model its gradients are the Jacobians supplied by a differentiable solver or finite differences, and for the PDE cases the paper assumes these Jacobians are correct rather than verifying them against an independent analytical or high-accuracy reference.

Editorial extensions

If this is right

  • Surrogates trained this way can be placed inside gradient-based optimization and data-assimilation loops: backpropagating through SC-FNO recovers parameters with inversion $R^2$ mostly above 0.9 and relative $L^2$ errors roughly 1/6 to 1/2.8 those of FNO or FNO-PINN.
  • Sensitivity supervision substitutes for training data: with 100 to 500 samples SC-FNO keeps the accuracy that FNO reaches only with more data, and in the 82-parameter zoned Burgers test, SC-FNO trained on 100 samples beats FNO trained on 500.
  • The improvement transfers to other neural operators, including Wavelet, Multiwavelet, and DeepONet architectures, so the conclusion is about the loss function rather than about Fourier networks specifically.
  • Because finite-difference sensitivity labels work nearly as well as automatic differentiation labels, the method applies to existing simulation codes that are not differentiable.
  • Gradient-aware models tolerate concept drift: at 40% parameter perturbation beyond the training range, SC-FNO keeps solution $R^2$ near 0.91-0.93 where FNO drops to 0.53-0.73.

Reading between the lines

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

  • If the paper is right, solution-path accuracy is a poor diagnostic for a surrogate's reliability in inverse or optimization settings; benchmarks for learned surrogates should report $\partial u/\partial p$ accuracy, because that is what inversion and data assimilation actually consume.
  • The archived Jacobian labels are exact in the paper, but the argument implies that approximate labels from coarser solvers or adjoint estimators might suffice to steer learning, which would remove the main data-preparation cost in practice.
  • The near-ineffectiveness of the PINN loss suggests a design rule worth testing: when the quantity of interest depends on parameters, supervise the parameter-gradient directly rather than relying on PDE residuals, which contain no information about $\partial u/\partial p$.
  • A natural extension is to apply the same constraint to sensitivities with respect to initial and boundary conditions; the ODE results already show gains for initial-condition sensitivities, so boundary-condition sensitivities are the obvious next test.
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

4 major / 5 minor

Summary. This paper proposes Sensitivity-Constrained Fourier Neural Operators (SC-FNO), which augment the standard FNO data loss with a sensitivity loss L_s (Eq. 6). The loss penalizes the difference between the model's AD-computed Jacobian \partial \hat{u}/\partial p and a 'true' Jacobian \partial u/\partial p prepared offline by a differentiable torchdiffeq-based solver or by finite differences. The authors evaluate four loss configurations on two ODEs and four PDEs, reporting that sensitivity supervision improves sensitivity R^2, parameter inversion accuracy, robustness to out-of-range parameter perturbations, and data efficiency relative to FNO and FNO-PINN, while adding 30-130% training time per epoch. The framework is also demonstrated on WNO, MWNO, DeepONet, and on an 82-parameter zoned Burgers equation. Section 3.5 and Appendix D.3 validate the Jacobian generation for ODE1 against an analytical solution, but not for the PDE cases.

Significance. If the reported gains are genuine, the contribution is significant and practical: a simple additive loss that makes neural operator surrogates usable for sensitivity analysis and gradient-based inversion, with apparent applicability across multiple operator architectures. The paper is commendably broad empirically, releases code, and includes a direct analytical validation for ODE1. Because the central mechanism is supervised learning of solver-provided Jacobians, however, the scientific weight of the PDE results hinges on the correctness of those Jacobian labels, and this is currently not established. The paper also provides machine-checkable pseudocode and reproducible experiment structure, which strengthens the presentation.

major comments (4)
  1. [Section 2.3 and Appendix D.3] The Jacobian labels for PDE1-PDE4 are never validated. Table D.13 compares AD and finite-difference sensitivities only against the analytical solution of ODE1; no analogous check exists for any PDE. Since the sensitivity loss trains the model to reproduce exactly these labels, any solver error (discretization, AD through time-stepping, stiffness effects) is learned as ground truth, and the reported sensitivity R^2 improvements could measure agreement with an inaccurate label rather than with the true PDE sensitivity. Please add a grid-convergence study for \partial u/\partial p and a comparison against an independent adjoint or high-order finite-difference Jacobian for at least PDE1-PDE3.
  2. [Section 3.5 and Table 5] The AD-versus-FD comparison for PDE1 reports downstream SC-FNO model performance, not agreement between the two Jacobian labeling methods. If both labels share a common discretization error, both trained models can perform well on the surrogate metrics while the underlying sensitivities are still wrong. Please report direct agreement statistics between the AD and FD Jacobians on the same PDE1 trajectories, such as per-parameter R^2 and relative L^2, before training any surrogate.
  3. [Section 3.1 and Figures 1-2] The inversion experiments use synthetic observations generated by the same differentiable solver that produced the training labels. This creates a potential circularity: an SC-FNO that faithfully matches the solver's Jacobians can appear to invert parameters successfully even if those Jacobians deviate from the true PDE sensitivities. Please re-run at least one inversion benchmark with observations generated by an independent solver or a different discretization (or with the analytical solution in the ODE case) and confirm that the inversion gains persist.
  4. [Tables 1-5 and Appendix D] All quantitative comparisons in the main tables are based on single training runs, without error bars or repeated seeds. The reported margins are often large, so seed variability may not overturn the main conclusion, but the claim that SC-FNO outperforms FNO and FNO-PINN would be more robust with mean and standard deviation over at least 3-5 seeds, particularly for the low-data regimes (N=100) and for the high-dimensional PDE2-zoned case.
minor comments (5)
  1. [Algorithms 2 and 3] The pseudocode states that c1, c2, and c3 are 'learnable coefficients', but the paper gives no details on how these coefficients are updated or which multi-task learning weighting scheme is used; please clarify or cite the method.
  2. [Section 2.4] The sentence 'we randomly select a subset of spatial-temporal points in each epoch (n < Nspatial points × t < Ttime points)' is not a well-formed inequality; please rewrite it to define the sampled index sets clearly.
  3. [Abstract and Section 3.6] The abstract claims the approach 'decreases training time while maintaining accuracy', but the reported per-epoch training times in Table C.8 all increase relative to FNO; the intended claim is presumably that fewer data or epochs are needed to reach a given accuracy, and this should be stated precisely.
  4. [Table C.8] The training-time measurements are reported without specifying the hardware and software versions; please state the computing environment used for the per-epoch times.
  5. [Appendix D.1, Tables D.9-D.10] The DeepONet baselines show very low sensitivity R^2 values (for example, 0.117 for \partial u/\partial c in D.9), which raises the question of whether the DeepONet architecture and hyperparameters are comparable in capacity and training effort to the FNO baselines; please comment on this.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: sensitivity labels come from an external solver and inversion is evaluated on held-out parameters.

full rationale

The paper's central mechanism is supervised learning of parameter sensitivities. Equation (6) defines L_s as the discrepancy between the predicted Jacobian ∂û/∂p and the target ∂u/∂p, where the targets are produced by an external differentiable solver or finite differences (Section 2.3), not by the SC-FNO model itself. The reported sensitivity R² values (Tables 1–4 and Appendix D) are computed on held-out test parameter values, and the paper states that 'validation and test sets contained parameter values not encountered during training' (Section 3.1), so the sensitivity gains are not statistically forced by construction. The inversion experiments (Section 3.1) optimize parameters by matching surrogate predictions to synthetic observations generated by the same solver, but they evaluate against known ground-truth parameters on a held-out split containing unseen parameter values; no fitted constant is later relabeled as a prediction. The paper cites several works by its own authors (e.g., Shen et al. 2023; Song et al. 2024a) in the introduction, but only as examples of differentiable modeling; none of these citations carries the load-bearing claim, and there is no imported uniqueness theorem or ansatz that forces the SC-FNO result. The genuine weakness is that the Jacobian labels for PDE1–PDE4 are validated against an analytical solution only for ODE1 (Table D.13); if the torchdiffeq-based solver's Jacobians were inaccurate for the PDE cases, the reported sensitivity and inversion improvements would overstate accuracy relative to the true PDE. That is an external-validity or correctness risk, not circularity: the targets still come from an independent numerical procedure rather than being defined in terms of the model's own output. Therefore, no step in the derivation is equivalent to its inputs by construction, and the paper is best characterized as self-contained against its chosen benchmarks rather than circular.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The method introduces no new physical entity; it adds a supervised loss term on the model's Jacobian. The free parameters are limited to learned loss weights and a gradient sampling design choice, neither of which is a physical constant fitted to support a derivation.

free parameters (2)
  • Loss weights c1, c2, c3 = Learned during training via multi-task learning (Liebel & Koerner 2018)
    Algorithms 2-3 set total loss as c1*Lu + c2*Ls (+ c3*Leq) with learnable coefficients; these are trained rather than hand-selected, so they are not ad hoc fitted constants, but they are additional trainable parameters of the method.
  • Gradient sampling counts (n spatial, t temporal points per epoch)
    Section 2.4: a random subset of n < N_spatial and t < T_time points is used to compute L_s each epoch; this is a design choice rather than an empirical constant fitted to data.
assumptions (3)
  • domain assumption The differentiable numerical solver (torchdiffeq extended to PDEs) produces sufficiently accurate solutions and Jacobians for PDE1-PDE4.
    Section 2.3 describes the solver; only ODE1 is validated against analytical gradients (Table D.13). The correctness of PDE Jacobian labels is assumed.
  • standard math Automatic differentiation through the FNO yields stable and meaningful Jacobians with respect to input parameters.
    AD is a standard tool, but its stability through the FNO and the sensitivity loss computation is assumed.
  • domain assumption The parameter distributions and initial condition sampling used to generate training data are representative and without leakage between train, validation, and test sets.
    Section 3.1 states a 70/15/15 split with disjoint parameter values; this is an assumption about the data generation pipeline.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Sensitivity-Constrained Fourier Neural Operators for Forward and Inverse Problems in Parametric Differential Equations." pith.science (2026). https://pith.science/paper/WC3BSQL3

@misc{pith2026250508740,
  author       = {Pith},
  title        = {Pith review of: Sensitivity-Constrained Fourier Neural Operators for Forward and Inverse Problems in Parametric Differential Equations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WC3BSQL3}},
  note         = {Machine review of arXiv:2505.08740}
}
read the original abstract

Parametric differential equations of the form du/dt = f(u, x, t, p) are fundamental in science and engineering. While deep learning frameworks such as the Fourier Neural Operator (FNO) can efficiently approximate solutions, they struggle with inverse problems, sensitivity estimation (du/dp), and concept drift. We address these limitations by introducing a sensitivity-based regularization strategy, called Sensitivity-Constrained Fourier Neural Operators (SC-FNO). SC-FNO achieves high accuracy in predicting solution paths and consistently outperforms standard FNO and FNO with physics-informed regularization. It improves performance in parameter inversion tasks, scales to high-dimensional parameter spaces (tested with up to 82 parameters), and reduces both data and training requirements. These gains are achieved with a modest increase in training time (30% to 130% per epoch) and generalize across various types of differential equations and neural operators. Code and selected experiments are available at: https://github.com/AMBehroozi/SC_Neural_Operators

Figures

Figures reproduced from arXiv: 2505.08740 by the authors.

Figure 1
Figure 1. Inversion of the parameter α in PDE1 using FNO and SC-FNO models (a) single parameter inversion, (b) simultaneous multi-parameter inversion. (a) PDE1 (b) PDE2 [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 3
Figure 3. Sample prediction of models for ODEs and PDE1 and PDE2. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 5
Figure 5. Performance of models for PDE1 for perturbed datasets, (a) [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figures from the paper (1 more)
Figure 6
Figure 6. Figure 6: Sample prediction of models for PDE 3 generalization capabilities of the models under limited training data scenarios. The remaining portion of the dataset, not used for training, served to test and measure the models’ performance [PITH_FULL_IMAGE:figures/full_fig_p00…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 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 format.date year duplicate empty "emp...

  2. [2]

    A differentiable, physics-informed ecosystem modeling and learning framework for large-scale inverse problems: demonstration with photosynthesis simulations

    Doaa Aboelyazeed, Chonggang Xu, Forrest Hoffman, Jiangtao Liu, Alex Jones, Chris Rackauckas, Kathryn Lawson, and Chaopeng Shen. A differentiable, physics-informed ecosystem modeling and learning framework for large-scale inverse problems: demonstration with photosynthesis simulations. Biogeosciences, 20 0 (13): 0 2671--2692, 2023. doi:10.5194/bg-20-2671-2...

    doi:10.5194/bg-20-2671-2023 2023

  3. [3]

    Managing computational complexity using surrogate models: a critical review

    Reza Alizadeh, Janet K Allen, and Farrokh Mistree. Managing computational complexity using surrogate models: a critical review. Research in Engineering Design, 31 0 (3): 0 275--298, 2020

    2020

  4. [4]

    A surrogate-model-based method for constrained optimization

    Charles Audet, J Denni, Douglas Moore, Andrew Booker, and Paul Frank. A surrogate-model-based method for constrained optimization. In 8th symposium on multidisciplinary analysis and optimization, pp.\ 4891, 2000

    work page 2000

  5. [5]

    D. M. Barker, W. Huang, Y-R. Guo, A. J. Bourgeois, and Q. N. Xiao. A three-dimensional variational data assimilation system for mm5: Implementation and initial results. Monthly Weather Review, 132 0 (4): 0 897 -- 914, 2004. doi:10.1175/1520-0493(2004)132<0897:ATVDAS>2.0.CO;2. URL https://journals.ametsoc.org/view/journals/mwre/132/4/1520-0493_2004_132_089...

    doi:10.1175/1520-0493(2004)132 2004

  6. [6]

    Spherical Fourier neural operators: Learning stable dynamics on the sphere

    Boris Bonev, Thorsten Kurth, Christian Hundt, Jaideep Pathak, Maximilian Baust, Karthik Kashinath, and Anima Anandkumar. Spherical Fourier neural operators: Learning stable dynamics on the sphere. In Proceedings of the 40th International Conference on Machine Learning , volume 202 of ICML '23 , pp.\ 2806--2823, Honolulu, Hawaii, USA, July 2023. JMLR.org. ...

    work page arXiv 2023

  7. [7]

    Ricky T. Q. Chen, Yulia Rubanova, Jesse Bettencourt, and David Duvenaud. Neural ordinary differential equations. In Proceedings of the 32nd International Conference on Neural Information Processing Systems , NIPS '18, pp.\ 6572--6583, Montreal, Canada, December 2018. Curran Associates Inc

    work page 2018

  8. [8]

    Sobolev training for neural networks

    Wojciech M Czarnecki, Simon Osindero, Max Jaderberg, Grzegorz Swirszcz, and Razvan Pascanu. Sobolev training for neural networks. Advances in neural information processing systems, 30, 2017

    2017

Show all 44 references
  1. [9]

    Differentiable, learnable, regionalized process-based models with multiphysical outputs can approach state-of-the-art hydrologic prediction accuracy

    Dapeng Feng, Jiangtao Liu, Kathryn Lawson, and Chaopeng Shen. Differentiable, learnable, regionalized process-based models with multiphysical outputs can approach state-of-the-art hydrologic prediction accuracy. Water Resources Research, 58 0 (10): 0 e2022WR032404, 2022. ISSN ...

    2022 doi

  2. [10]

    The suitability of differentiable, physics-informed machine learning hydrologic models for ungauged regions and climate change impact assessment

    Dapeng Feng, Hylke Beck, Kathryn Lawson, and Chaopeng Shen. The suitability of differentiable, physics-informed machine learning hydrologic models for ungauged regions and climate change impact assessment. Hydrology and Earth System Sciences, 27 0 (12): 0 2357--2373, June 2023...

    2023 doi

  3. [11]

    Model-parallel fourier neural operators as learned surrogates for large-scale parametric pdes

    Thomas J Grady, Ravi Khan, Mathias Louboutin, Ziyi Yin, Philipp A Witte, Ravi Chandra, et al. Model-parallel fourier neural operators as learned surrogates for large-scale parametric pdes. Computational Geosciences, 178: 0 105402, 2023. doi:10.1016/j.cageo.2023.105402

    2023

  4. [12]

    Multiwavelet-based Operator Learning for Differential Equations

    Gaurav Gupta, Xiongye Xiao, and Paul Bogdan. Multiwavelet-based Operator Learning for Differential Equations . In Advances in Neural Information Processing Systems , November 2021

    2021

  5. [13]

    Tartakovsky

    QiZhi He, Panos Stinis, and Alexandre M. Tartakovsky. Physics-constrained deep neural network method for estimating parameters in a redox flow battery. Journal of Power Sources, 528: 0 231147, April 2022. ISSN 0378-7753. doi:10.1016/j.jpowsour.2022.231147

    2022

  6. [14]

    Shah, and Will Tebbutt

    Mike Innes, Alan Edelman, Keno Fischer, Chris Rackauckas, Elliot Saba, Viral B. Shah, and Will Tebbutt. A Differentiable Programming System to Bridge Machine Learning and Scientific Computing , July 2019

    2019

  7. [15]

    Peishi Jiang, Zhao Yang, Jiali Wang, Chenfu Huang, Pengfei Xue, T. C. Chakraborty, Xingyuan Chen, and Yun Qian. Efficient Super-Resolution of Near-Surface Climate Modeling Using the Fourier Neural Operator . Journal of Advances in Modeling Earth Systems, 15 0 (7): 0 e2023MS003...

    2023 doi

  8. [16]

    NSFnets ( Navier-Stokes flow nets): Physics-informed neural networks for the incompressible Navier-Stokes equations

    Xiaowei Jin, Shengze Cai, Hui Li, and George Em Karniadakis. NSFnets ( Navier-Stokes flow nets): Physics-informed neural networks for the incompressible Navier-Stokes equations. Journal of Computational Physics, 426: 0 109951, February 2021. ISSN 0021-9991. doi:10.1016/j.jcp.2...

    2021

  9. [17]

    Kevrekidis, Lu Lu, Paris Perdikaris, Sifan Wang, and Liu Yang

    George Em Karniadakis, Ioannis G. Kevrekidis, Lu Lu, Paris Perdikaris, Sifan Wang, and Liu Yang. Physics-informed machine learning. Nature Reviews Physics, 3 0 (6): 0 422--440, June 2021. ISSN 2522-5820. doi:10.1038/s42254-021-00314-5

    2021 doi

  10. [18]

    Neural operator: Learning maps between function spaces

    Nikola B Kovachki, Zongyi Li, Burigede Liu, Kamyar Azizzadenesheli, Kaushik Bhattacharya, Andrew Stuart, et al. Neural operator: Learning maps between function spaces. ArXiv Preprint ArXiv:210808481, 2021

    2021

  11. [19]

    Fourier Neural Operator for Parametric Partial Differential Equations , May 2021

    Zongyi Li, Nikola Kovachki, Kamyar Azizzadenesheli, Burigede Liu, Kaushik Bhattacharya, Andrew Stuart, and Anima Anandkumar. Fourier Neural Operator for Parametric Partial Differential Equations , May 2021

    2021

  12. [20]

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

    Zongyi Li, Hongkai Zheng, Nikola Kovachki, David Jin, Haoxuan Chen, Burigede Liu, Kamyar Azizzadenesheli, and Anima Anandkumar. Physics-informed neural operator for learning partial differential equations, 2023

    2023

  13. [21]

    Physics- Informed Neural Operator for Learning Partial Differential Equations

    Zongyi Li, Hongkai Zheng, Nikola Kovachki, David Jin, Haoxuan Chen, Burigede Liu, Kamyar Azizzadenesheli, and Anima Anandkumar. Physics- Informed Neural Operator for Learning Partial Differential Equations . ACM / IMS Journal of Data Science, February 2024. doi:10.1145/3648506

    2024 doi

  14. [22]

    Auxiliary tasks in multi-task learning

    Lukas Liebel and Marco K \"o rner. Auxiliary tasks in multi-task learning. arXiv preprint arXiv:1805.06334, 2018

    2018 arXiv

  15. [23]

    Gradient-enhanced neural network response surface approximations

    Weiyu Liu and S Batill. Gradient-enhanced neural network response surface approximations. In 8th Symposium on Multidisciplinary Analysis and Optimization, pp.\ 4923, 2000

    2000

  16. [24]

    Discretize- Optimize vs

    Derek Onken and Lars Ruthotto. Discretize- Optimize vs. Optimize-Discretize for Time-Series Regression and Continuous Normalizing Flows , July 2020

    2020

  17. [25]

    Differential equations, mechanics, and computation, volume 51

    Richard S Palais and Robert Andrew Palais. Differential equations, mechanics, and computation, volume 51. American Mathematical Soc., 2009

    2009

  18. [26]

    Raissi, P

    M. Raissi, P. Perdikaris, and G. E. Karniadakis. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. Journal of Computational Physics, 378: 0 686--707, 2019. ISSN 0021-9991. do...

    2019

  19. [27]

    On the usefulness of overparameterized ecological models

    Peter Reichert and Martin Omlin. On the usefulness of overparameterized ecological models. Ecological Modelling, 95 0 (2): 0 289--299, February 1997. ISSN 0304-3800. doi:10.1016/S0304-3800(96)00043-9

    1997 doi

  20. [28]

    PhyCRNet : Physics-informed convolutional-recurrent network for solving spatiotemporal PDEs

    Pu Ren, Chengping Rao, Yang Liu, Jian-Xun Wang, and Hao Sun. PhyCRNet : Physics-informed convolutional-recurrent network for solving spatiotemporal PDEs . Computer Methods in Applied Mechanics and Engineering, 389: 0 114399, February 2022. ISSN 0045-7825. doi:10.1016/j.cma.2021.114399

    2022

  21. [29]

    Why so many published sensitivity analyses are false: A systematic review of sensitivity analysis practices

    Andrea Saltelli, Ksenia Aleksankina, William Becker, Pamela Fennell, Federico Ferretti, Niels Holst, Sushan Li, and Qiongli Wu. Why so many published sensitivity analyses are false: A systematic review of sensitivity analysis practices. Environmental modelling & software, 114:...

    2019

  22. [30]

    JAX MD : A Framework for Differentiable Physics

    Samuel Schoenholz and Ekin Dogus Cubuk. JAX MD : A Framework for Differentiable Physics . In Advances in Neural Information Processing Systems , volume 33, pp.\ 11428--11441. Curran Associates, Inc., 2020

    2020

  23. [31]

    Appling, Pierre Gentine, Toshiyuki Bandai, Hoshin Gupta, Alexandre Tartakovsky, Marco Baity-Jesi , Fabrizio Fenicia, Daniel Kifer, Li Li, Xiaofeng Liu, Wei Ren, Yi Zheng, Ciaran J

    Chaopeng Shen, Alison P. Appling, Pierre Gentine, Toshiyuki Bandai, Hoshin Gupta, Alexandre Tartakovsky, Marco Baity-Jesi , Fabrizio Fenicia, Daniel Kifer, Li Li, Xiaofeng Liu, Wei Ren, Yi Zheng, Ciaran J. Harman, Martyn Clark, Matthew Farthing, Dapeng Feng, Praveen Kumar, Doa...

    2023

  24. [32]

    When ancient numerical demons meet physics-informed machine learning: adjoint-based gradients for implicit differentiable modeling

    Yalan Song, Wouter Knoben, Martyn Clark, Dapeng Feng, Kathryn Lawson, Kamlesh Sawadekar, and Chaopeng Shen. When ancient numerical demons meet physics-informed machine learning: adjoint-based gradients for implicit differentiable modeling. Hydrology and Earth System Sciences, ...

    2024 doi

  25. [33]

    Improving physics-informed, differentiable hydrologic models for capturing unseen extreme events

    Yalan Song, Kamlesh Sawadekar, Jonathan M Frame, Ming Pan, Martyn Clark, Wouter JM Knoben, Andrew W Wood, Trupesh Patel, and Chaopeng Shen. Improving physics-informed, differentiable hydrologic models for capturing unseen extreme events. ESS Open Archive, 2024 b . doi:10.22541...

    2024

  26. [34]

    A. M. Tartakovsky, C. Ortiz Marrero, Paris Perdikaris, G. D. Tartakovsky, and D. Barajas-Solano. Physics-informed deep neural networks for learning parameters and constitutive relationships in subsurface flow problems. Water Resources Research, 56 0 (5): 0 e2019WR026731, 2020....

    2020 doi

  27. [35]

    Wavelet Neural Operator for solving parametric partial differential equations in computational mechanics problems

    Tapas Tripura and Souvik Chakraborty. Wavelet Neural Operator for solving parametric partial differential equations in computational mechanics problems. Computer Methods in Applied Mechanics and Engineering, 404: 0 115783, February 2023. ISSN 0045-7825. doi:10.1016/j.cma.2022.115783

    2023

  28. [36]

    From calibration to parameter learning: Harnessing the scaling effects of big data in geoscientific modeling

    Wen-Ping Tsai, Dapeng Feng, Ming Pan, Hylke Beck, Kathryn Lawson, Yuan Yang, Jiangtao Liu, and Chaopeng Shen. From calibration to parameter learning: Harnessing the scaling effects of big data in geoscientific modeling. Nature Communications, 12 0 (1): 0 5988, October 2021. IS...

    2021 doi

  29. [37]

    Fully probabilistic deep models for forward and inverse problems in parametric pdes

    Arnaud Vadeboncoeur, Ömer Deniz Akyildiz, Ieva Kazlauskaite, Mark Girolami, and Fehmi Cirak. Fully probabilistic deep models for forward and inverse problems in parametric pdes. Journal of Computational Physics, 491: 0 112369, 2023. ISSN 0021-9991. doi:https://doi.org/10.1016/...

    2023

  30. [38]

    Learning the solution operator of parametric partial differential equations with physics-informed deeponets

    Sifan Wang, Hanwen Wang, and Paris Perdikaris. Learning the solution operator of parametric partial differential equations with physics-informed deeponets. Science advances, 7 0 (40): 0 eabi8605, 2021

    2021

  31. [39]

    Brenowitz, Spencer K

    Oliver Watt-Meyer , Noah D. Brenowitz, Spencer K. Clark, Brian Henn, Anna Kwa, Jeremy McGibbon, W. Andre Perkins, Lucas Harris, and Christopher S. Bretherton. Neural Network Parameterization of Subgrid-Scale Physics From a Realistic Geography Global Storm-Resolving Simulation ...

    2024 doi

  32. [40]

    O'Gorman

    Janni Yuval and Paul A. O'Gorman. Stable machine-learning parameterization of subgrid processes for climate modeling at a range of resolutions. Nature Communications, 11 0 (1): 0 3295, July 2020. ISSN 2041-1723. doi:10.1038/s41467-020-17142-3

    2020 doi

  33. [41]

    Differentiable Modeling and Optimization of Battery Electrolyte Mixtures Using Geometric Deep Learning , November 2023

    Shang Zhu, Bharath Ramsundar, Emil Annevelink, Hongyi Lin, Adarsh Dave, Pin-Wen Guan, Kevin Gering, and Venkatasubramanian Viswanathan. Differentiable Modeling and Optimization of Battery Electrolyte Mixtures Using Geometric Deep Learning , November 2023

    2023

  34. [42]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should ...

  35. [43]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@firs...

  36. [44]

    end-to-end

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibset...

    1907

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.