Pith. sign in

REVIEW 4 major objections 4 minor 73 references

Improving Data Fidelity via Diffusion Model-based Correction and Super-Resolution

T0 review · 4 major / 4 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 A two-step diffusion pipeline can lift low-fidelity, low-resolution data to high-fidelity, high-resolution output without the governing equations or paired training examples.

desk verdict Genuinely useful IPD idea, but test-set hyperparameter tuning undermines the headline numbers; worth a serious referee after a validation-set redo. read the letter →

arxiv 2505.08526 v2 pith:BIDFLH3R submitted 2025-05-13 math.NA cs.NA

Wuzhe Xu , Yulong Lu , Sifan Wang , Tong-Rui Liu This is my paper · ORCID
classification math.NAcs.NA MSC 60H1065M9968T07
keywords diffusionmodelssuper-resolutionbiascorrectionscore-basedgenerativeSDEditimbalancedperturbinganddenoisingscientificcomputingclimatedata
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 low-fidelity, low-resolution data can be corrected and upscaled by a two-step diffusion pipeline that never uses the governing physics or paired low- and high-fidelity examples. Correction runs a forward diffusion for a short time $t_1$ to let added noise dominate unknown biases, then denoises with the backward probability-flow ODE for an independent time $t_2$, using a score model trained only on clean high-fidelity low-resolution samples; keeping $t_1 < t_2$ is what allows large-scale structure to survive while bias is removed. Super-resolution then chains three conditional diffusion models (cascaded SR3) to reinsert fine-scale detail. The paper supports the correction step with an $L_2$ error bound and an exact variance-addition argument for Gaussian biases, and tests the full pipeline on three PDE benchmarks and a climate reanalysis dataset.

What carries the argument

The central mechanism is imbalanced perturbing and denoising (IPD): perturb the low-fidelity input forward for time $t_1$, then run the backward probability-flow ODE for the larger time $t_2$, with the score model trained on clean high-fidelity samples. The load-bearing identity is distribution alignment, $p(u_l(t_1)) \approx p(\tilde{u}_h(t_2))$, and the inequality $t_1 < t_2$ is what lets a small perturbation erase bias while a moderate denoising interval keeps large-scale structure intact. In practice, the two times are selected on test data by minimizing a distributional metric, and the experiments adopt a weighted mean energy log ratio. The super-resolution half is cascaded SR3: three conditional diffusion models, each trained on pairs created by a known downsampling operator at magnification $\times 2$, refine the corrected $32\times32$ field back to $256\times256$.

What would settle it

Take a fixed clean high-fidelity dataset and add a strongly non-Gaussian bias, for instance sparse localized spikes with heavy-tailed amplitudes. If no pair $t_1<t_2$ within the search interval makes the alignment metric small while $t_2$ stays below the structure-preserving threshold, then IPD corrections will fail to approach the high-fidelity data; comparing IPD against balanced perturbing-denoising on that bias family and checking whether the Theorem 1 bound is violated would settle whether the method removes diverse biases as claimed.

Watch

Extended reading notes

Core claim

The central claim is that a score-based diffusion model trained on the target high-fidelity distribution can remove diverse, unknown biases from low-fidelity data if forward perturbation and backward denoising run for two independent times $t_1$ and $t_2$ instead of one shared time. For $0 < t_1 < t_2 < 1$, perturbing $u_l = \tilde{u}_h + e$ for time $t_1$ gives $u_l(t_1)$, and the paper argues the distribution $p(u_l(t_1))$ aligns with $p(\tilde{u}_h(t_2))$; running the probability-flow ODE from $t_2$ back to $0$ then maps the perturbed sample toward $\tilde{u}_h$. Theorem 1 states that, with probability at least $1-\lambda$, $\|\hat{u}_l(t_1,t_2)-\tilde{u}_h\|_2^2 \le e^{2L_s t_2}[\|e\|_2^2 + \sigma^2(t_2)\delta + (\sigma^2(t_1)+\sigma^2(t_2))C_\lambda]$, where $\delta$ is the score training loss and $C_\lambda$ is a dimension-dependent concentration constant. Proposition 1 shows that for Gaussian bias $e\sim N(0,\gamma^2 I)$ with $\sigma^2(t_2)=\sigma^2(t_1)+\gamma^2$, the expected correction error is at most $e^{2L_s t_2}\sigma^2(t_2)\delta$, vanishing as the training loss goes to zero. This is the mechanism by which IPD can beat balanced perturbing-denoising, whose single time cannot simultaneously screen out the bias and preserve structure.

Load-bearing premise

The correction step rests on the premise that some perturbation time $t_1$ and some larger denoising time $t_2$ make the noise-perturbed low-fidelity distribution align closely with the noise-perturbed high-fidelity distribution while $t_2$ is still small enough that backward denoising preserves large-scale structure, and the paper gives no constructive condition guaranteeing such a pair exists for a given bias.

Editorial extensions

If this is right

  • One correction model, trained only on clean high-fidelity low-resolution snapshots, handles numerical errors from different PDE solvers and white, pink, and brown noise without retraining or knowing the bias type.
  • Because bias is removed before upscaling, the super-resolution stage sees data already consistent with the target distribution, so systematic error is not amplified by the finer-grid model.
  • In the Gaussian-bias case, the correction error bound tends to zero as the score training loss tends to zero, meaning the bias is removed completely in the ideal score limit.
  • Reported experiments show IPD consistently gives lower total variation distance and narrower variability than balanced perturbing and denoising across metrics and search horizons.
  • The cascaded SR3 chain makes the method applicable to arbitrary resolution gaps by adding more $\times2$ stages, as demonstrated by the $32\times32$ to $256\times256$ example.

Reading between the lines

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

  • Going beyond the paper: the alignment condition $p(u_l(t_1)) \approx p(\tilde{u}_h(t_2))$ presumes the bias can be mimicked by injected Gaussian noise at some stage of the forward SDE; biases with strong spatial correlation or heavy tails may require a learned bias model or repeated correction passes rather than a single IPD pair.
  • Going beyond the paper: the grid search for $(t_1,t_2)$ is a calibration step on the data itself; a natural extension is to learn a mapping from dataset statistics to the optimal times, removing the test-time search and reducing inference cost.
  • Going beyond the paper: IPD's logic is distribution-translation rather than physics-specific, so the same correction step could serve as a general-purpose purifier for any target domain where clean samples exist, such as transferring bias between numerical solvers or observational products.
  • Going beyond the paper: the error bound contains the factor $e^{2L_s t_2}$, so the practical robustness seen on 2D fields is not automatically guaranteed for high-dimensional or strongly nonlinear 3D problems, where the Lipschitz constant of the learned score can be large.
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 / 4 minor

Summary. The manuscript proposes a two-stage diffusion-based pipeline, DCSR, for correcting and super-resolving low-fidelity, low-resolution (LFLR) data. In the correction stage, an "imbalanced perturbing and denoising" (IPD) variant of SDEdit perturbs the LFLR input for a time t1 and then denoises with a score model trained on high-fidelity low-resolution (HFLR) data for an independent time t2. In the super-resolution stage, a cascaded SR3 model upsamples the corrected field. The method is designed to require no physics knowledge and only HFHR/HFLR reference data for training, not paired LFLR data. Algorithm 2 selects t1 and t2 by minimizing a distribution-alignment metric between perturbed LFLR and HFLR sets; Section 4.1 additionally fixes the metric and terminal search time Te empirically. The theoretical results, Proposition 1 and Theorem 1, provide conditional L2 error bounds for the correction step. Numerical experiments on 1D advection, 2D linear elasticity, 2D Navier-Stokes, and ERA5 climate data compare DCSR with interpolation and SR3-only baselines.

Significance. The idea of a physics-free, unpaired correction-then-super-resolution pipeline is timely and potentially useful, and the IPD variant of SDEdit is a plausible and clearly presented contribution. The paper also provides explicit theoretical statements and an ablation of IPD versus balanced perturbing and denoising (BPD), which are strengths. However, the central empirical claim is weakened by two issues: the main hyperparameters t1, t2, Te, and the metric M are selected on the test LFLR datasets, and the reported errors are computed on the same samples; and the theoretical bounds do not, as stated, guarantee that the corrected output is closer to the target than the biased input. If the authors can address the test-set tuning issue with a proper validation protocol and compare against the closest unpaired-debiasing baselines [49] and [60], the contribution would be valuable. As it stands, the evidence for the headline "versatile fidelity improvement" claim is not yet conclusive.

major comments (4)
  1. [Algorithm 2 and Section 4.1] The hyperparameters t1, t2, Te, and the metric M are selected using the test LFLR and HFLR datasets, and the reported errors are computed on the same samples. In Section 4.1, Figure 3 is a boxplot of TVD on the test LF solutions, and the text states that "Consequently, IPD with MELRw metric and Te=0.2 is adopted in Algorithm 3 for the rest numerical examples." This is data leakage: Algorithm 2 searches for the pair (t1,t2) that minimizes a distributional metric on the test data, and the final evaluation uses those same test samples. The method may still work, but the current protocol cannot distinguish genuine correction capability from tuning to the test set. Please introduce a validation split or an explicit tuning protocol that does not touch the test samples, and report performance on truly held-out LFLR data.
  2. [Theorem 1 (Section 3.1)] Theorem 1 bounds the correction error by e^{2 Ls t2} [ ||e||^2 + sigma^2(t2) delta + (sigma^2(t1)+sigma^2(t2)) C_lambda ], where e is the bias. This is an upper bound that grows with ||e||^2; it does not establish that the corrected output is closer to the target HFLR data than the input, nor does it show that the bias is removed. The text in Section 3.1 says the theorem characterizes a trade-off, but the abstract and contributions claim that robustness and effectiveness are "validated theoretically." Please either weaken the theoretical claim to a conditional stability bound or add a result that compares the correction error with ||e||^2 and shows a contraction regime.
  3. [Section 3.1, paragraph on IPD and Algorithm 2] The load-bearing premise of IPD is that there exist t1 < t2 such that p(ul(t1)) aligns closely with p(uh(t2)) while t2 remains small enough to preserve large-scale structure. The manuscript asserts this alignment qualitatively and then searches for such times in Algorithm 2, but provides no constructive condition or quantitative characterization of when such times exist. Since Algorithm 2 will always find a minimizer on a finite grid even if the distributions do not genuinely align, the reader cannot tell whether the reported gains reflect a real alignment or overfitting to the test noise realizations. Please add either a theoretical criterion for alignability, or a sensitivity/ablation study (e.g., comparing the selected times with random or fixed times, and plotting the alignment metric as a function of t1 and t2) that demonstrates the mechanism is operative beyond the grid-search minimum.
  4. [Sections 1.3 and 4] The two closest works, [49] and [60], perform unpaired debiasing at low resolution followed by probabilistic super-resolution, which is exactly the same two-stage architecture as DCSR. The paper differentiates itself by claiming versatility and bias removal, but no numerical comparison against these methods is provided in Section 4; the only baselines are LFLR+Interp and LFLR+SR, neither of which performs a debiasing step. A direct comparison on at least one of the PDE or climate benchmarks is needed to support the claim that DCSR is preferable to or more general than these existing approaches. If such a comparison is infeasible, the paper should state the obstacle explicitly and restrict its claims accordingly.
minor comments (4)
  1. [Algorithm 4 and Section 3.3] The third paired dataset is written as {˜uh_{32,i}, ˜uh_{64,i}} in Algorithm 4 but as {˜uh_i, ˜uh_{64,i}} in Section 3.3. Please make the notation consistent, since ˜uh normally denotes the 32x32 HFLR data.
  2. [Proof of Theorem 1] In Eq. (27) of the proof, the term sigma^2(t) delta should presumably be sigma^2(t2) delta, since the reconstruction bound is stated with t2 in Proposition 1 and Lemma 2. This is a typographical inconsistency that should be fixed.
  3. [Throughout] There are several typographical errors, including "isotrophic" instead of "isotropic", "repling" instead of "relying", "Algotirhm" instead of "Algorithm", and "userspecific" instead of "user-specific". A careful proofreading pass is recommended.
  4. [Section 4.1] The text says "The study is conducted on all six types of LF data," but Figure 3 and the surrounding discussion show results only for the LW solver with pink noise, with other cases deferred to the appendix. Please either include the additional cases in the main text or state clearly that Figure 3 is representative and that the appendix contains the full ablation.

Circularity Check

1 steps flagged · score 4.0 of 10

IPD hyperparameters (t1, t2, M, Te) are selected on the same test LFLR/HFLR samples used for evaluation, so part of the reported fidelity gain reflects test-set tuning rather than an independent prediction.

  1. fitted input called prediction [Section 3.1 ('Selection of t1 and t2'), Algorithm 2, and Section 4.1 (Figures 3–5)]
    "Both the metric M and the terminal searching time Te are selected empirically, and the criteria for choosing M will be discussed in detail in Section 4.1. ... Consequently, IPD with MELRw metric and Te = 0.2 is adopted in Algorithm 3 for the rest numerical examples."

    Algorithm 2 selects the key correction parameters t1* and t2* by minimizing a distributional metric M between perturbed LFLR and perturbed HFLR samples. In the experiments, those samples are the same test LFLR and HFLR datasets that are later corrected and scored. Moreover, the choice of M and Te is made in Figure 3 by inspecting the resulting TVD between IPD corrections and the reference on those same 1D advection test samples. Thus the headline empirical claim that IPD 'removes diverse biases' is evaluated after its central hyperparameters have been tuned against that same evaluation data.

full rationale

The core theoretical derivation is not circular: Proposition 1 is a conditional expectation bound under an explicit Gaussian-bias assumption with the score-matching loss δ as an input, and Theorem 1 is a stability-style upper bound rather than a fitted claim. The diffusion models are trained on HFLR/HFHR reference data independently of the LFLR test data, and no load-bearing uniqueness theorem is imported from the authors' prior work. The cited prior work [60] is presented as a different approach, not as justification for the central mechanism. The main circularity concern is empirical rather than definitional: t1*, t2*, M, and Te are selected using the same test LFLR and HFLR distributions against which correction quality is measured, and M/Te are chosen using the evaluation criterion on those same samples. This makes the performance comparison with BPD partially a fitted result. However, because the corrected samples are generated by a learned diffusion score model and are not directly equal to the target by construction, the circularity is partial, warranting a score of 4 rather than 6 or higher.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The method rests on a score model trained on high-fidelity reference data and on the empirical selection of t1, t2, Te, and the metric. No physically invented entities are introduced. The theoretical bounds depend on the score-matching accuracy delta and an assumed Lipschitz constant rather than on standalone predictions.

free parameters (3)
  • t1* and t2* (perturbation and denoising times) = not reported; selected on test LFLR and HFLR sets via Algorithm 2
    Central hyperparameters of the IPD correction step. They are chosen by matching perturbed empirical distributions on the test data, and the same test data are then corrected and evaluated.
  • Te (terminal search time) = 0.2
    Chosen from the 1D advection ablation in Section 4.1 and used for all later examples; no principled derivation is given.
  • Metric M for distribution alignment = MELRw
    Selected after comparing MMD, MELRu, MELRw, and W2 in Section 4.1; an empirical choice that affects the selected t1 and t2.
assumptions (4)
  • domain assumption The approximate score function S_theta is Ls-Lipschitz continuous in its input for t in [0,1] and has training loss L(theta) < delta.
    Required by Theorem 1 and Proposition 1 in Appendix C. No construction or verification of Ls is given, and delta is an empirical training loss rather than a certified quantity.
  • domain assumption There exist t1 < t2 in [0, Te] such that p(u_l(t1)) is close to p(tilde_u_h(t2)) under the chosen metric.
    Core alignment premise of IPD stated in Section 3.1 and Figure 2. The existence is not proved; Algorithm 2 searches for such times on the test data.
  • ad hoc to paper For Proposition 1, the bias e is Gaussian N(0, gamma^2 I) and t2 satisfies sigma^2(t2) = sigma^2(t1) + gamma^2.
    Toy model used to demonstrate the IPD advantage. The bias variance gamma is assumed known, which conflicts with the claimed error-source-agnostic setting.
  • domain assumption The bias enters additively as u_l = tilde_u_h + e.
    This is definitional for the difference, but for structured solver errors it is not shown that e is independent of tilde_u_h or that a single unconditional diffusion model trained on tilde_u_h can remove it.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Improving Data Fidelity via Diffusion Model-based Correction and Super-Resolution." pith.science (2026). https://pith.science/paper/BIDFLH3R

@misc{pith2026250508526,
  author       = {Pith},
  title        = {Pith review of: Improving Data Fidelity via Diffusion Model-based Correction and Super-Resolution},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BIDFLH3R}},
  note         = {Machine review of arXiv:2505.08526}
}
read the original abstract

We propose a unified diffusion model-based correction and super-resolution method to enhance the fidelity and resolution of diverse low-quality data through a two-step pipeline. First, the correction step employs a novel enhanced stochastic differential editing technique based on an imbalanced perturbation and denoising process, ensuring robust and effective bias correction at the low-resolution level. The robustness and effectiveness of this approach are validated theoretically and experimentally. Next, the super-resolution step leverages cascaded conditional diffusion models to iteratively refine the corrected data to high-resolution. Numerical experiments on three PDE problems and a climate dataset demonstrate that the proposed method effectively enhances low-fidelity, low-resolution data by correcting numerical errors and noise while simultaneously improving resolution to recover fine-scale structures.

Figures

Figures reproduced from arXiv: 2505.08526 by the authors.

Figure 1
Figure 1. Diagram of the two-step pipeline. The Correction step (blue arrow) removes various biases from LFLR [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Diagram of the imbalanced perturbing and denoising (IPD) process. The process begins with perturbing the [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Box plot of comparison the TVD between correction obtained by Algotirhm 3 under different settings and [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Comparison of correction obtained by Algorithm 3 using BPD (select [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: Corrections using IPD with the MELRw metric and [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]
Figure 6
Figure 6. Figure 6: Strain field ε11(x) distribution inside a RVE microstructure. The top row shows the results of low-fidelity data obtained using the continuous Green’s operator, while the bottom row shows results of low-fidelity data obtained using the discrete Green’s operator. The le…
Figure 7
Figure 7. Figure 7: Strain field ε22(x) distribution inside a RVE microstructure. The top row shows the results of low-fidelity data obtained using the continuous Green’s operator, while the bottom row shows results of low-fidelity data obtained using the discrete Green’s operator. The le…
Figure 8
Figure 8. Figure 8: Numerical results for the Navier-Stokes equation. The top row shows the results of LFLR data obtained [PITH_FULL_IMAGE:figures/full_fig_p016_8.png]
Figure 9
Figure 9. Figure 9: Top row presents results of low resolution potential vorticity data polluted with White noise and the bottom [PITH_FULL_IMAGE:figures/full_fig_p017_9.png]
Figure 10
Figure 10. Figure 10: Top row presents results of low resolution potential vorticity data polluted with Pink noise and the Bottom [PITH_FULL_IMAGE:figures/full_fig_p027_10.png]
Figure 11
Figure 11. Figure 11: Top row presents results of low resolution potential vorticity data polluted with Brown noise and the Bottom [PITH_FULL_IMAGE:figures/full_fig_p027_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

73 extracted references · 40 canonical work pages

  1. [49]

    Debias coarsely, sample conditionally: Statistical downscaling through optimal transport and probabilis- tic diffusion models

    Zhong Yi Wan, Ricardo Baptista, Anudhyan Boral, Yi-Fan Chen, John Anderson, Fei Sha, and Leonardo Zepeda- N´u˜nez. Debias coarsely, sample conditionally: Statistical downscaling through optimal transport and probabilis- tic diffusion models. Advances in Neural Information Processing Systems, 36:47749–47763, 2023

    work page 2023

  2. [60]

    Diffusion-based models for unpaired super-resolution in fluid dynamics

    Wuzhe Xu, Yulong Lu, Anqing Xuan, Ali Barzegari, et al. Diffusion-based models for unpaired super-resolution in fluid dynamics. arXiv preprint arXiv:2504.05443, 2025

    arXiv 2025

  3. [1]

    A new generative adversarial network for medical images super resolution

    Waqar Ahmad, Hazrat Ali, Zubair Shah, and Shoaib Azmat. A new generative adversarial network for medical images super resolution. Scientific Reports, 12(1):9533, 2022

    work page 2022

  4. [2]

    Large scale gan training for high fidelity natural image synthesis

    Andrew Brock. Large scale gan training for high fidelity natural image synthesis. arXiv preprint arXiv:1809.11096, 2018

    arXiv 2018

  5. [3]

    Strong trends in the skill of the era-40 and ncep–ncar reanalyses in the high and midlatitudes of the southern hemisphere, 1958–2001

    David H Bromwich and Ryan L Fogt. Strong trends in the skill of the era-40 and ncep–ncar reanalyses in the high and midlatitudes of the southern hemisphere, 1958–2001. Journal of Climate, 17(23):4603–4619, 2004

    work page 1958

  6. [4]

    A super-resolution diffusion model for recovering bone microstructure from ct images

    Trevor J Chan and Chamith S Rajapakse. A super-resolution diffusion model for recovering bone microstructure from ct images. Radiology: Artificial Intelligence, 5(6):e220251, 2023

    work page 2023

  7. [5]

    Deep generative image models using a laplacian pyramid of adversarial networks

    Emily L Denton, Soumith Chintala, Rob Fergus, et al. Deep generative image models using a laplacian pyramid of adversarial networks. Advances in neural information processing systems, 28, 2015

    work page 2015

  8. [6]

    Image super-resolution using deep convolutional networks

    Chao Dong, Chen Change Loy, Kaiming He, and Xiaoou Tang. Image super-resolution using deep convolutional networks. IEEE transactions on pattern analysis and machine intelligence, 38(2):295–307, 2015. 17

    work page 2015

Show all 73 references
  1. [7]

    Overview and meteorological validation of the wind integration national dataset toolkit

    C Draxl, BM Hodge, A Clifton, and J McCaa. Overview and meteorological validation of the wind integration national dataset toolkit. Technical report, National Renewable Energy Lab.(NREL), Golden, CO (United States), 2015

    2015

  2. [8]

    Smith, Michael P

    Gideon Dresdner, Dmitrii Kochkov, Peter Norgaard, Leonardo Zepeda-N´u˜nez, Jamie A. Smith, Michael P. Bren- ner, and Stephan Hoyer. Learning to correct spectral methods for simulating turbulent flows. 2022

    2022

  3. [9]

    Spectral neural operators

    VS Fanaskov and Ivan V Oseledets. Spectral neural operators. In Doklady Mathematics, volume 108, pages S226–S232. Springer, 2023

    2023

  4. [10]

    Pot: Python optimal transport

    R ´emi Flamary, Nicolas Courty, Alexandre Gramfort, Mokhtar Z Alaya, Aur´elie Boisbunon, Stanislas Chambon, Laetitia Chapel, Adrien Corenflos, Kilian Fatras, Nemo Fournier, et al. Pot: Python optimal transport. Journal of Machine Learning Research, 22(78):1–8, 2021

    2021

  5. [11]

    A physics-informed variational deep- onet for predicting crack path in quasi-brittle materials

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

    2022

  6. [12]

    Climalign: Unsupervised statistical downscaling of cli- mate variables via normalizing flows

    Brian Groenke, Luke Madaus, and Claire Monteleoni. Climalign: Unsupervised statistical downscaling of cli- mate variables via normalizing flows. In Proceedings of the 10th International Conference on Climate Informat- ics, pages 60–66, 2020

    2020

  7. [13]

    Adaptive fourier neural operators: Efficient token mixers for transformers

    John Guibas, Morteza Mardani, Zongyi Li, Andrew Tao, Anima Anandkumar, and Bryan Catanzaro. Adaptive fourier neural operators: Efficient token mixers for transformers. arXiv preprint arXiv:2111.13587, 2021

    2021 arXiv

  8. [14]

    Gnot: A general neural operator transformer for operator learning

    Zhongkai Hao, Zhengyi Wang, Hang Su, Chengyang Ying, Yinpeng Dong, Songming Liu, Ze Cheng, Jian Song, and Jun Zhu. Gnot: A general neural operator transformer for operator learning. In International Conference on Machine Learning, pages 12556–12569. PMLR, 2023

    2023

  9. [15]

    Hard-constrained deep learning for climate downscaling

    Paula Harder, Alex Hernandez-Garcia, Venkatesh Ramesh, Qidong Yang, Prasanna Sattegeri, Daniela Szwarc- man, Campbell Watson, and David Rolnick. Hard-constrained deep learning for climate downscaling. Journal of Machine Learning Research, 24(365):1–40, 2023

    2023

  10. [16]

    A generative deep learning approach to stochastic downscaling of precipitation forecasts

    Lucy Harris, Andrew TT McRae, Matthew Chantry, Peter D Dueben, and Tim N Palmer. A generative deep learning approach to stochastic downscaling of precipitation forecasts. Journal of Advances in Modeling Earth Systems, 14(10):e2022MS003120, 2022

    2022

  11. [17]

    The era5 global reanalysis

    Hans Hersbach, Bill Bell, Paul Berrisford, Shoji Hirahara, Andr´as Hor´anyi, Joaqu´ın Mu˜noz-Sabater, Julien Nico- las, Carole Peubey, Raluca Radu, Dinand Schepers, et al. The era5 global reanalysis. Quarterly Journal of the Royal Meteorological Society, 146(730):1999–2049, 2020

    1999

  12. [18]

    Denoising diffusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. Advances in neural information processing systems, 33:6840–6851, 2020

    2020

  13. [19]

    The ncep/ncar 40-year reanalysis project

    Eugenia Kalnay, Masao Kanamitsu, Robert Kistler, William Collins, Dennis Deaven, Lev Gandin, Mark Iredell, Suranjana Saha, Glenn White, John Woollen, et al. The ncep/ncar 40-year reanalysis project. In Renewable energy, pages V ol1146–V ol1194. Routledge, 2018

    2018

  14. [20]

    Unpaired image-to-image translation via neural schr\” odinger bridge

    Beomsu Kim, Gihyun Kwon, Kwanyoung Kim, and Jong Chul Ye. Unpaired image-to-image translation via neural schr\” odinger bridge. arXiv preprint arXiv:2305.15086, 2023

    2023 arXiv

  15. [21]

    Ma- chine learning–accelerated computational fluid dynamics

    Dmitrii Kochkov, Jamie A Smith, Ayya Alieva, Qing Wang, Michael P Brenner, and Stephan Hoyer. Ma- chine learning–accelerated computational fluid dynamics. Proceedings of the National Academy of Sciences , 118(21):e2101784118, 2021

    2021

  16. [22]

    Neural optimal transport

    Alexander Korotin, Daniil Selikhanovych, and Evgeny Burnaev. Neural optimal transport. arXiv preprint arXiv:2201.12220, 2022

    2022 arXiv

  17. [23]

    Adaptive estimation of a quadratic functional by model selection

    Beatrice Laurent and Pascal Massart. Adaptive estimation of a quadratic functional by model selection. Annals of statistics, pages 1302–1338, 2000

    2000

  18. [24]

    Patch complexity, finite pixel correlations and optimal denoising

    Anat Levin, Boaz Nadler, Fredo Durand, and William T Freeman. Patch complexity, finite pixel correlations and optimal denoising. In Computer Vision–ECCV 2012: 12th European Conference on Computer Vision, Florence, Italy, October 7-13, 2012, Proceedings, Part V 12, pages 73–86....

    2012

  19. [25]

    Generative adversarial models for extreme downscaling of climate datasets

    Guiye Li and Guofeng Cao. Generative adversarial models for extreme downscaling of climate datasets. arXiv preprint arXiv:2402.14049, 2024

    2024 arXiv

  20. [26]

    Srdiff: Single image super-resolution with diffusion probabilistic models

    Haoying Li, Yifan Yang, Meng Chang, Shiqi Chen, Huajun Feng, Zhihai Xu, Qi Li, and Yueting Chen. Srdiff: Single image super-resolution with diffusion probabilistic models. Neurocomputing, 479:47–59, 2022

    2022

  21. [27]

    Adbm: Adversarial diffusion bridge model for reliable adversarial purification

    Xiao Li, Wenxuan Sun, Huanran Chen, Qiongxiu Li, Yining Liu, Yingzhe He, Jie Shi, and Xiaolin Hu. Adbm: Adversarial diffusion bridge model for reliable adversarial purification. arXiv preprint arXiv:2408.00315, 2024. 18

    2024 arXiv

  22. [28]

    Transformer for partial differential equations’ operator learning

    Zijie Li, Kazem Meidani, and Amir Barati Farimani. Transformer for partial differential equations’ operator learning. arXiv preprint arXiv:2205.13671, 2022

    2022 arXiv

  23. [29]

    Fourier neural operator for parametric partial differential equations

    Zongyi Li, Nikola Borislavov Kovachki, Kamyar Azizzadenesheli, Burigede liu, Kaushik Bhattacharya, An- drew Stuart, and Anima Anandkumar. Fourier neural operator for parametric partial differential equations. In International Conference on Learning Representations, 2021

    2021

  24. [30]

    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/JMS Journal of Data Science, 1(3):1–27, 2024

    2024

  25. [31]

    Enhanced deep residual networks for single image super-resolution

    Bee Lim, Sanghyun Son, Heewon Kim, Seungjun Nah, and Kyoung Mu Lee. Enhanced deep residual networks for single image super-resolution. In Proceedings of the IEEE conference on computer vision and pattern recog- nition workshops, pages 136–144, 2017

    2017

  26. [32]

    B. A. Lippmann and Julian Schwinger. Variational principles for scattering processes. I. Physical Review, 79(3):469–480, 1950

    1950

  27. [33]

    Dpm-solver++: Fast solver for guided sampling of diffusion probabilistic models

    Cheng Lu, Yuhao Zhou, Fan Bao, Jianfei Chen, Chongxuan Li, and Jun Zhu. Dpm-solver++: Fast solver for guided sampling of diffusion probabilistic models. arXiv preprint arXiv:2211.01095, 2022

    2022 arXiv

  28. [34]

    Learning nonlinear oper- ators via DeepONet based on the universal approximation theorem of operators

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

    2021

  29. [35]

    Generative downscaling of pde solvers with physics-guided diffusion models.Journal of Scientific Computing, 101(71), 2024

    Yulong Lu and Wuzhe Xu. Generative downscaling of pde solvers with physics-guided diffusion models.Journal of Scientific Computing, 101(71), 2024

    2024

  30. [36]

    Evaluation of era-40, ncep-1, and ncep-2 reanalysis air temperatures with ground-based measurements in china

    Lijuan Ma, Tingjun Zhang, Qingxiang Li, Oliver W Frauenfeld, and Dahe Qin. Evaluation of era-40, ncep-1, and ncep-2 reanalysis air temperatures with ground-based measurements in china. Journal of Geophysical Research: Atmospheres, 113(D15), 2008

    2008

  31. [37]

    Image restoration using very deep convolutional encoder- decoder networks with symmetric skip connections

    Xiaojiao Mao, Chunhua Shen, and Yu-Bin Yang. Image restoration using very deep convolutional encoder- decoder networks with symmetric skip connections. Advances in neural information processing systems , 29, 2016

    2016

  32. [38]

    Multiple physics pre- training for physical surrogate models

    Michael McCabe, Bruno R ´egaldo-Saint Blancard, Liam Holden Parker, Ruben Ohana, Miles Cranmer, Alberto Bietti, Michael Eickenberg, Siavash Golkar, Geraud Krawezik, Francois Lanusse, et al. Multiple physics pre- training for physical surrogate models. arXiv preprint arXiv:2310...

    arXiv

  33. [39]

    Generation of random distribution of fibres in long-fibre reinforced composites

    Ant ´onio Rui Melro, Pedro Ponces Camanho, and Silvestre Taveira Pinho. Generation of random distribution of fibres in long-fibre reinforced composites. Composites Science and Technology, 68(9):2092–2102, 2008

    2008

  34. [40]

    Sdedit: Guided image synthesis and editing with stochastic differential equations

    Chenlin Meng, Yutong He, Yang Song, Jiaming Song, Jiajun Wu, Jun-Yan Zhu, and Stefano Ermon. Sdedit: Guided image synthesis and editing with stochastic differential equations. arXiv preprint arXiv:2108.01073 , 2021

    2021 arXiv

  35. [41]

    A numerical method for computing the overall response of nonlinear compos- ites with complex microstructure

    Herv ´e Moulinec and Pierre Suquet. A numerical method for computing the overall response of nonlinear compos- ites with complex microstructure. Computer Methods in Applied Mechanics and Engineering , 157(1-2):69–94, 1998

    1998

  36. [42]

    Diffusion models for adversarial purification

    Weili Nie, Brandon Guo, Yujia Huang, Chaowei Xiao, Arash Vahdat, and Anima Anandkumar. Diffusion models for adversarial purification. arXiv preprint arXiv:2205.07460, 2022

    2022 arXiv

  37. [43]

    A robust generative adversarial network approach for climate downscaling and weather generation

    Neelesh Rampal, Peter B Gibson, Steven Sherwood, Gab Abramowitz, and Sanaa Hobeichi. A robust generative adversarial network approach for climate downscaling and weather generation. Authorea Preprints, 2024

    2024

  38. [44]

    Im- age super-resolution via iterative refinement

    Chitwan Saharia, Jonathan Ho, William Chan, Tim Salimans, David J Fleet, and Mohammad Norouzi. Im- age super-resolution via iterative refinement. IEEE transactions on pattern analysis and machine intelligence , 45(4):4713–4726, 2022

    2022

  39. [45]

    On the effectiveness of neural operators at zero-shot weather downscaling

    Saumya Sinha, Brandon Benton, and Patrick Emami. On the effectiveness of neural operators at zero-shot weather downscaling. arXiv preprint arXiv:2409.13955, 2024

    2024 arXiv

  40. [46]

    Denoising diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502, 2020

    2010 arXiv

  41. [47]

    Score- based generative modeling through stochastic differential equations

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. Score- based generative modeling through stochastic differential equations. arXiv preprint arXiv:2011.13456, 2020

    2011 arXiv

  42. [48]

    Dual diffusion implicit bridges for image-to-image translation

    Xuan Su, Jiaming Song, Chenlin Meng, and Stefano Ermon. Dual diffusion implicit bridges for image-to-image translation. arXiv preprint arXiv:2203.08382, 2022. 19

    2022 arXiv

  43. [50]

    Guided diffusion model for adversarial purifi- cation

    Jinyi Wang, Zhaoyang Lyu, Dahua Lin, Bo Dai, and Hongfei Fu. Guided diffusion model for adversarial purifi- cation. arXiv preprint arXiv:2205.14969, 2022

    2022 arXiv

  44. [51]

    Micrometer: Micromechanics transformer for predicting mechanical responses of heterogeneous materials

    Sifan Wang, Tong-Rui Liu, Shyam Sankaran, and Paris Perdikaris. Micrometer: Micromechanics transformer for predicting mechanical responses of heterogeneous materials. arXiv preprint arXiv:2410.05281, 2024

    2024 arXiv

  45. [52]

    Long-time integration of parametric evolution equations with physics-informed deeponets

    Sifan Wang and Paris Perdikaris. Long-time integration of parametric evolution equations with physics-informed deeponets. Journal of Computational Physics, 475:111855, 2023

    2023

  46. [53]

    Bridging operator learning and conditioned neural fields: A unifying perspective

    Sifan Wang, Jacob H Seidman, Shyam Sankaran, Hanwen Wang, George J Pappas, and Paris Perdikaris. Bridging operator learning and conditioned neural fields: A unifying perspective. arXiv preprint arXiv:2405.13998, 2024

    2024 arXiv

  47. [54]

    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(40):eabi8605, 2021

    2021

  48. [55]

    Improved architectures and training algorithms for deep operator networks

    Sifan Wang, Hanwen Wang, and Paris Perdikaris. Improved architectures and training algorithms for deep operator networks. Journal of Scientific Computing, 92(2):35, 2022

    2022

  49. [56]

    Generative diffusion-based downscaling for climate

    Robbie A Watt and Laura A Mansfield. Generative diffusion-based downscaling for climate. arXiv preprint arXiv:2404.17752, 2024

    2024 arXiv

  50. [57]

    Super-resolution neural operator

    Min Wei and Xuesong Zhang. Super-resolution neural operator. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18247–18256, 2023

    2023

  51. [58]

    U-fno—an en- hanced fourier neural operator-based deep-learning model for multiphase flow

    Gege Wen, Zongyi Li, Kamyar Azizzadenesheli, Anima Anandkumar, and Sally M Benson. U-fno—an en- hanced fourier neural operator-based deep-learning model for multiphase flow. Advances in Water Resources, 163:104180, 2022

    2022

  52. [59]

    Climate variable downscaling with conditional normalizing flows

    Christina Winkler, Paula Harder, and David Rolnick. Climate variable downscaling with conditional normalizing flows. arXiv preprint arXiv:2405.20719, 2024

    2024 arXiv

  53. [61]

    Diffusion models: A comprehensive survey of methods and applications

    Ling Yang, Zhilong Zhang, Yang Song, Shenda Hong, Runsheng Xu, Yue Zhao, Wentao Zhang, Bin Cui, and Ming-Hsuan Yang. Diffusion models: A comprehensive survey of methods and applications. ACM Computing Surveys, 56(4):1–39, 2023

    2023

  54. [62]

    Fourier neural operators for arbitrary resolution climate data downscaling

    Qidong Yang, Alex Hernandez-Garcia, Paula Harder, Venkatesh Ramesh, Prasanna Sattegeri, Daniela Szwar- cman, Campbell D Watson, and David Rolnick. Fourier neural operators for arbitrary resolution climate data downscaling. arXiv preprint arXiv:2305.14452, 2023

    2023 arXiv

  55. [63]

    Jiaxiang Yi and Miguel A. Bessa. Rvesimulator: An automated representative volume element simulator for data-driven material discovery. In AI for Accelerated Materials Design - NeurIPS 2023 Workshop, 2023

    2023

  56. [64]

    Latent diffusion model-based mri superresolution enhances mild cognitive impairment prognostication and alzheimer’s disease classification

    Dan Yoon, Youho Myong, Young Gyun Kim, Yongsik Sim, Minwoo Cho, Byung-Mo Oh, and Sungwan Kim. Latent diffusion model-based mri superresolution enhances mild cognitive impairment prognostication and alzheimer’s disease classification. NeuroImage, page 120663, 2024

    2024

  57. [65]

    Adversarial purification with score-based generative models

    Jongmin Yoon, Sung Ju Hwang, and Juho Lee. Adversarial purification with score-based generative models. In International Conference on Machine Learning, pages 12062–12072. PMLR, 2021

    2021

  58. [66]

    Learning deep implicit fourier neural operators (ifnos) with applications to heterogeneous material modeling

    Huaiqian You, Quinn Zhang, Colton J Ross, Chung-Hao Lee, and Yue Yu. Learning deep implicit fourier neural operators (ifnos) with applications to heterogeneous material modeling. Computer Methods in Applied Mechanics and Engineering, 398:115296, 2022

    2022

  59. [67]

    Beyond a gaussian denoiser: Residual learning of deep cnn for image denoising

    Kai Zhang, Wangmeng Zuo, Yunjin Chen, Deyu Meng, and Lei Zhang. Beyond a gaussian denoiser: Residual learning of deep cnn for image denoising. IEEE transactions on image processing, 26(7):3142–3155, 2017

    2017

  60. [68]

    Ffdnet: Toward a fast and flexible solution for cnn-based image denoising

    Kai Zhang, Wangmeng Zuo, and Lei Zhang. Ffdnet: Toward a fast and flexible solution for cnn-based image denoising. IEEE Transactions on Image Processing, 27(9):4608–4622, 2018

    2018

  61. [69]

    Fast sampling of diffusion models with exponential integrator

    Qinsheng Zhang and Yongxin Chen. Fast sampling of diffusion models with exponential integrator. arXiv preprint arXiv:2204.13902, 2022

    2022 arXiv

  62. [70]

    Unpaired image-to-image translation using cycle- consistent adversarial networks

    Jun-Yan Zhu, Taesung Park, Phillip Isola, and Alexei A Efros. Unpaired image-to-image translation using cycle- consistent adversarial networks. In Proceedings of the IEEE international conference on computer vision, pages 2223–2232, 2017. 20

    2017

  63. [71]

    Fourier-deeponet: Fourier-enhanced deep operator networks for full waveform inversion with improved accuracy, generalizability, and robustness

    Min Zhu, Shihang Feng, Youzuo Lin, and Lu Lu. Fourier-deeponet: Fourier-enhanced deep operator networks for full waveform inversion with improved accuracy, generalizability, and robustness. Computer Methods in Applied Mechanics and Engineering, 416:116300, 2023. 21 A Notation ...

    2023

  64. [72]

    ˆuh(t∗ 1,t∗

    The correction of ul using IPD with optimalt∗ 1 andt∗ 2. ˆuh(t∗ 1,t∗

  65. [73]

    The enhancement of DCSR for ulcorrection using IPD with optimalt∗ 1 andt∗ 2. N Number of training samples M Number of testing samples ODEsolve(x(t1),t 1,t 2;v) The solution of ODE driven by velocity field− 1 2 d[σ2(t)] dt v(x(t),t ), starting from initial condition x(t1) att1 ...

Pith tools

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