Pith. sign in

REVIEW 3 major objections 5 minor 44 references

A Distributed Plug-and-Play MCMC Algorithm for High-Dimensional Inverse Problems

T0 review · 3 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read A distributed plug-and-play MCMC sampler scales Bayesian image reconstruction to 4096^2 color images on multiple GPUs while preserving the exact posterior of the serial run.

desk verdict Useful distributed PnP-ULA extension with real scaling results, but the machine-precision equivalence claim rests on an unproven block-separability assumption and there is no code. read the letter →

arxiv 2511.00870 v2 pith:PKDBBIDA submitted 2025-11-02 stat.ME cs.DCeess.SP

Maxime Bouton , Pierre-Antoine Thouvenin , Audrey Repetti , Pierre Chainais This is my paper · ORCID
classification stat.MEcs.DCeess.SP MSC 62F1565C0568U10
keywords MCMCplug-and-playpriorsLangevinalgorithmsdistributedcomputingimaginginverseproblemsuncertaintyquantificationconvolutionalneuralnetworkssplitGibbssampler
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 argues that plug-and-play MCMC imaging—Bayesian inversion with a neural-network denoiser as prior—can be made to scale to very large images by distributing every operator, including the denoiser, across multiple workers. The proposed sampler splits the posterior into local blocks using an augmented auxiliary variable, applies a Langevin update with a CNN denoiser, and coordinates workers with small boundary exchanges. The central claim is that the distributed sampler reproduces the serial sampler's outputs to machine precision for any number of workers, while running fast enough to handle 4096^2 color images with pixel-wise uncertainty maps. If true, it removes the memory and runtime barriers that currently confine plug-and-play MCMC to moderate-size problems.

What carries the argument

The load-bearing mechanism is the localized-operator decomposition: linear operators such as convolutions are expressed as blocks that depend only on local selections of the input plus small 'ghost' regions exchanged between neighboring workers. For the CNN denoiser, each convolutional layer is distributed with an overlap-save scheme, so the network evaluates on each worker up to boundary communication. The associated log-prior is assumed block-additively separable with respect to the same partition (Assumption A5); this is what lets the PnP-ULA update (Eq. 24) and PSGLA auxiliary update (Eq. 25) factor across workers without changing the target distribution.

What would settle it

Take a trained CNN denoiser, compute (or estimate by finite differences) the score function of the prior it induces—for instance via the Tweedie relation—and check whether log p(x) separates as the sum of per-block terms on a 2-worker partition; a single partition where separability fails would falsify the core equivalence. Alternatively, compare serial and distributed chains on the same problem with many random seeds and test whether the empirical distributions of the samples agree beyond machine-precision arithmetic.

Watch

Extended reading notes

Core claim

The paper's central discovery is that a PnP-ULA transition and its auxiliary-variable Gibbs step can be implemented in a distributed SPMD fashion with a localized CNN denoiser, so that the posterior remains unchanged. Concretely, Algorithm 1 decomposes the gradient of the data-fidelity term, the proximal map of the regularizer, and the denoiser itself via overlap-save convolutions and ghost-region communications; under block-additive separability of the log-prior, each worker updates its own slice of x and z. The authors report that for B in {1,2,4} the operation outputs coincide down to machine precision (modulo random-number generation), and that the sampler enables sampling of 4096^2 colo

Load-bearing premise

The whole equivalence rests on the assertion, stated without proof or empirical check in Section II-C, that the prior density induced by the distributed CNN denoiser is block-additively separable with respect to the same partition as the observation operators; if that separability fails, Algorithm 1 samples a different posterior than the serial PnP-ULA.

Editorial extensions

If this is right

  • Plug-and-play MCMC can be applied to images at 4096^2 pixels, producing MMSE estimates and pixel-wise variance maps that a single-GPU sampler cannot fit in memory.
  • Users can increase the number of workers without re-tuning the sampler: because the distributed and serial outputs coincide to machine precision, the target posterior is the same for any B.
  • Lightweight CNNs with few layers become attractive for distributed sampling: they preserve reconstruction quality while keeping communication phases and message sizes small.
  • The same distributed treatment extends naturally to other inverse problems where the observation operator is local, covering common Gaussian and Poisson imaging models.

Reading between the lines

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

  • Editorial extension: the block-separability of the denoiser-induced prior (Assumption A5) is asserted rather than proved; a denoiser with a global receptive field or one whose boundary handling couples blocks would break the machine-precision equivalence, so the claim is only as strong as that separability.
  • Editorial extension: a direct test of the equivalence claim is to compare serial and distributed chains under several random seeds; since only pseudo-random generation differs, statistical agreement rather than bitwise agreement is the appropriate criterion.
  • Editorial extension: the overlap-save distribution strategy suggests a recipe for scaling other plug-and-play samplers, such as diffusion-based priors, provided their non-local operations can be localized; communicating the full receptive field across layers is a costly alternative the paper explicitly leaves open.
  • Editorial extension: the scaling results imply the method pays off when per-worker compute dominates communication; for cheap priors like total variation, distribution can slow the sampler, so the practical benefit is conditional on network cost.
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 a distributed MCMC sampler for high-dimensional Bayesian imaging inverse problems. It combines the split Gibbs / approximate data augmentation (AXDA) framework with Plug-and-Play unadjusted Langevin algorithm (PnP-ULA), and distributes all operators—including convolutional neural network denoisers—over multiple GPUs using an SPMD architecture. The central claims are (i) the distributed sampler is numerically equivalent to its serial counterpart, producing identical outputs for B=1,2,4 workers down to machine precision (Sec. IV-B); (ii) it scales to images as large as 4096^2 color images (Sec. IV-C); and (iii) reconstruction quality is 'comparable to other PnP methods' despite the distributed implementation. The paper reports communication and computation overheads for three denoiser architectures (DDFB, DnCNN, DRUNet) and gives strong/weak scaling experiments on inpainting and deconvolution under Gaussian and Poisson noise.

Significance. If the central equivalence claim is correct, this is a practically valuable contribution: it would be the first distributed PnP-MCMC scheme that retains the statistical guarantees of serial PnP-ULA while scaling to very high resolution. The paper has genuine engineering strengths: the communication-aware treatment of CNN layers (overlap-save with ghost regions, Eqs. (19)-(20)), the detailed FLOP/message-size analysis in Table IV, and the scalability experiments on a real multi-GPU system (Tables VI-VII) are nontrivial and useful. The authors also make a good faith effort to compare three different CNN priors and a TV baseline. However, the load-bearing claim of machine-precision equivalence rests on an unproven structural assumption about the denoiser-induced prior, and the experimental support for the 'comparable to other PnP methods' statement is currently limited to a comparison with a TV baseline, not with other PnP methods or even serial PnP-ULA. These gaps need to be addressed before the central claims can be accepted.

major comments (3)
  1. [Section II-C, after Eq. (25); Assumption (A5), Eq. (6)] The paper asserts that after distributing D_epsilon with the overlap-save strategy of Section II-B2, 'the associated prior p satisfies Assumption (A5)' (block-additive separability of log p with respect to the same partition as the operators). This is the key step that allows Algorithm 1 to be interpreted as a distributed implementation of serial PnP-ULA. No proof is provided, and the assertion is not obvious. For a PnP prior defined through a denoiser, log p is not explicitly known; even if each convolution layer is local with a halo, the composition of convolutions, nonlinearities, and (for DRUNet) down/up-sampling does not automatically yield a log-density that is exactly block-additive in the sense of Eq. (6). Moreover, the halo defined in Eq. (18) is one-sided (eB_b extends only in the +Ny direction by L_y-1). For a symmetric convolution kernel, computing output rows at the top of b
  2. [Section IV-B, Table V and text near 'operation outputs coincide'] The statement that 'Algorithm 1 produces estimates of the same quality for B in {1,2,4}. In particular, operation outputs coincide for any number of workers B down to machine precision, except the generation of pseudo-random numbers' is central to the paper's contribution, but it is not supported by any reported numerical comparison. Table V reports a single value per metric and prior, with no indication of how the B=1,2,4 runs were aggregated, no standard deviations, and no repeated-seed variability. There is also no release of the code or data (the link in Section III is a placeholder 'https://repo/to/code'), so the reader cannot reproduce or verify the equivalence claim. In addition, the abstract's 'reconstruction performance comparable to other PnP methods' is not tested: the experiments compare the proposed sampler with a TV baseline and with three denoiser choices, but there is no
  3. [Section IV-B, DRUNet discussion] The authors themselves note that DRUNet 'performance deteriorates sharply' in Poisson deconvolution and that the poor results 'suggest instability during sampling, likely caused by an underestimation of L_D'. This is an acknowledged limitation, but it interacts with the paper's broad claim that the sampler achieves performance 'comparable to other PnP methods'. If the estimated Lipschitz constant is unreliable for a state-of-the-art denoiser such as DRUNet, the practical utility of the method for general CNNs is diminished. The authors should either (i) provide a more careful Lipschitz estimation procedure (e.g., spectral normalization or power iteration) and rerun the affected experiments, or (ii) clearly delimit the claim to the architectures for which the sampler is stable (here, DDFB appears to be the only consistently stable learned prior). The current text leaves the reader uncerta
minor comments (5)
  1. [Section III, code availability] The code URL is a placeholder ('https://repo/to/code'). If the paper is intended for publication, a working repository should be provided or the statement should be removed.
  2. [Table V] Table V reports metrics for B in {1,2,4} but the table has no B dimension. Please clarify whether the values are identical across B by construction, or report them separately. If the equivalence claim is established, a sentence stating that values are identical to machine precision would be sufficient.
  3. [Section II-C, second paragraph] Typo: 'auxilliary' should be 'auxiliary'.
  4. [Section II-B2a, Eq. (18)] The notation for the extended block is slightly confusing: eB_b is defined as a set of indices, but later eNy,b is used for its cardinality. Please make the distinction between the set and its size explicit.
  5. [Reference [34]] Formatting issue in the reference: 'F oundations' should be 'Foundations'.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation: the sampler is assembled from previously published, independently verifiable components; the main weakness is an unproven block-separability assertion (A5), which is a correctness gap, not a circular step.

full rationale

I walked the derivation chain. The proposed sampler is constructed by combining AXDA/SGS [10], the distributed SGS framework of [11], and PnP-ULA [3]; no fitted parameter is renamed as a prediction, and reconstruction/scaling performance is evaluated against external TV baselines and standard metrics (SNR/PSNR/SSIM). The only novel load-bearing assertion is at the end of Section II-C: "Further, following the proposed distributed strategy for D_epsilon described in Section II-B2, the associated prior p satisfies Assumption (A5)." This block-additivity claim is asserted without proof or empirical verification, and the one-sided ghost region of Eq. (18) (eN_{y,b} = N_{y,b} + L_y - 1) raises a genuine boundary question for symmetric kernels. That is a missing-support/correctness risk, not circularity: the paper does not define p in terms of (A5), does not derive a prediction from a fit, and does not import a uniqueness or ansatz result from the authors' prior work. The Section IV-B statement that "operation outputs coincide for any number of workers B down to machine precision" is an implementation claim about the overlap-save distributed operator (Eqs. 19-20); equivalence is by construction of the distributed convolution and is empirically checkable by comparing Algorithm 1 runs for B in {1,2,4}. Self-citations are present ([10] Vono-Dobigeon-Chainais, [11] Thouvenin-Repetti-Chainais, [26] the authors' own preliminary workshop paper), and [11] is a structural foundation, but those are peer-reviewed, externally published results whose assumptions do not include the present equivalence target; per the review rules that is real evidence and does not raise the circularity score. No definitional cycle, no fitted-input-renamed-as-prediction, no load-bearing self-citation chain, and no renaming of a known result were found.

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

The central claim rests on standard assumptions inherited from [3] and [11], plus one unproven structural assertion: that the CNN-induced prior's log-density is block-separable after distributed evaluation. The main free parameters are the AXDA coupling, the denoising level, and the TV weight, all hand-tuned per experiment. No new physical entities are introduced.

free parameters (5)
  • AXDA coupling parameter rho = rho=1e-5 (Gaussian TV), rho1=10, rho2=1e-3 (Poisson)
    Hand-set per experiment; controls the bias of the approximate posterior pi_rho in Eq. (21).
  • Denoising level epsilon = epsilon=sigma for Gaussian tasks; epsilon=0.05 for Poisson
    Chosen per task; the authors note tuning is difficult for Poisson and causes DRUNet artifacts (Section IV-B).
  • TV regularization beta = beta=40 (Gaussian), beta=13 (Poisson)
    Hand-set for the TV baseline; affects all TV comparisons.
  • Estimated Lipschitz constant L_D = DDFB K=4: 2; DDFB K=20: 2; DnCNN: 3; DRUNet: 7 (Table II)
    Estimate from [40]; used in stepsize conditions (26). Underestimation is suggested as a cause of DRUNet instability.
  • Step-size safety factors = alpha=1; 0.99 factors in lambda and gamma
    Ad hoc choices to satisfy strict inequalities in (26).
assumptions (4)
  • domain assumption There exists a proper prior density p whose log-density is associated with the denoiser D_epsilon, under the condition that I_N - D_epsilon is Lipschitz with constant L_D (from [3, Sec. 3.2]).
    Used to justify the PnP-ULA update (24) in Section II-C.
  • domain assumption Assumptions (A1)-(A5): f1∘H1 Lipschitz-differentiable, f2 proper l.s.c. convex, H1 and H2 localized, and f1, f2, log p block-additively separable.
    Required by the distributed decomposition in Section II-A and Algorithm 1.
  • domain assumption AXDA constructs an approximate posterior pi_rho (Eq. 21) with fixed rho>0; the sampler targets pi_rho rather than the exact posterior.
    Bias is controlled only asymptotically as rho→0; the paper fixes rho per experiment and does not quantify the resulting UQ error.
  • ad hoc to paper After distributing D_epsilon with overlap-save communications, the prior density p satisfies block-additive separability (A5).
    Asserted in Section II-C ('the associated prior p satisfies Assumption (A5)') but no proof or empirical check is provided; this is the structural premise for exact equivalence between distributed and serial samplers.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Distributed Plug-and-Play MCMC Algorithm for High-Dimensional Inverse Problems." pith.science (2026). https://pith.science/paper/PKDBBIDA

@misc{pith2026251100870,
  author       = {Pith},
  title        = {Pith review of: A Distributed Plug-and-Play MCMC Algorithm for High-Dimensional Inverse Problems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PKDBBIDA}},
  note         = {Machine review of arXiv:2511.00870}
}
read the original abstract

Markov Chain Monte Carlo (MCMC) algorithms are standard approaches to solve imaging inverse problems and quantify estimation uncertainties, a key requirement in absence of ground-truth data. To improve estimation quality, Plug-and-Play MCMC algorithms, such as PnP-ULA, have been recently developed to accommodate priors encoded by a denoising neural network. Designing scalable samplers for high-dimensional imaging inverse problems remains a challenge: drawing and storing high-dimensional samples can be prohibitive, especially for high-resolution images. To address this issue, this work proposes a distributed sampler based on approximate data augmentation and PnP-ULA to solve very large problems. The proposed sampler uses lightweight denoising convolutional neural network, to efficiently exploit multiple GPUs on a Single Program Multiple Data architecture. Reconstruction performance and scalability are evaluated on several imaging problems. Communication and computation overheads due to the denoiser are carefully discussed. The proposed distributed approach noticeably combines three very precious qualities: it is scalable, enables uncertainty quantification, for a reconstruction performance comparable to other PnP methods.

Figures

Figures reproduced from arXiv: 2511.00870 by the authors.

Figure 1
Figure 1. Illustration of a 2D linear convolution operator [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Communications and operations underlying the proposed distributed [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Inpainting results – Restoration results with Algorithm 1 using B = 4 workers. First row: ground truth x and observations y. Last 4 rows, top to bottom: MMSE (left) and pixel variance of the red channel (right) using either TV, DDFB, DnCNN, or DRUNet. Dashed lines on the ground truth image show the partition considered for x. C. Scalability experiments Both strong and weak scaling results are influenced by the overa… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 3 linked inside Pith

  1. [1]

    C. P. Robert,The Bayesian Choice: From Decision-Theoretic F ounda- tions to Computational Implementation, 2nd ed., ser. Springer Texts in Statistics. Springer, 2007

    2007

  2. [2]

    Bayesian imaging with data-driven priors encoded by neural networks,

    M. Holden, M. Pereyra, and K. C. Zygalakis, “Bayesian imaging with data-driven priors encoded by neural networks,”SIAM J. Imaging Sci., vol. 15, no. 2, pp. 892–924, Jun. 2022

    2022

  3. [3]

    Bayesian imaging using Plug & Play priors: When Langevin meets Tweedie,

    R. Laumont, V . de Bortoli, A. Almansa, J. Delon, A. Durmus, and M. Pereyra, “Bayesian imaging using Plug & Play priors: When Langevin meets Tweedie,”SIAM J. Imaging Sci., vol. 15, no. 2, pp. 701–737, Jun. 2022

    2022

  4. [4]

    Regularization by denois- ing: Bayesian model and Langevin-within-split Gibbs sampling,

    E. C. Faye, M. D. Fall, and N. Dobigeon, “Regularization by denois- ing: Bayesian model and Langevin-within-split Gibbs sampling,”IEEE Trans. Image Process., vol. 34, pp. 221–234, 2025

    2025

  5. [5]

    Scalable Bayesian uncertainty quantification with data- driven priors for radio interferometric imaging,

    T. I. Liaudat, M. Mars, M. A. Price, M. Pereyra, M. M. Betcke, and J. D. McEwen, “Scalable Bayesian uncertainty quantification with data- driven priors for radio interferometric imaging,”RAS Tech. and Instrum., vol. 3, no. 1, pp. 505–534, Jan. 2024

    2024

  6. [6]

    Playing with duality: an overview of recent primal-dual approaches for solving large-scale optimization problems,

    N. Komodakis and J.-C. Pesquet, “Playing with duality: an overview of recent primal-dual approaches for solving large-scale optimization problems,”IEEE Signal Process. Mag., vol. 32, no. 3, pp. 31–54, Nov. 2015

    2015

  7. [7]

    Non-log-concave and nonsmooth sampling via Langevin Monte Carlo algorithms,

    T. T.-K. Lau, H. Liu, and T. Pock, “Non-log-concave and nonsmooth sampling via Langevin Monte Carlo algorithms,” inProc. Adv. Tech. in Optim. for Mach. Learning and Imaging, A. Benfenati, F. Porta, T. A. Bubba, and M. Viola, Eds. Singapore: Springer Nature, 2024, pp. 83–149

    2024

  8. [8]

    Analy- sis of primal-dual Langevin algorithms,

    M. Burger, M. J. Ehrhardt, L. Kuger, and L. Weigand, “Analy- sis of primal-dual Langevin algorithms,” Nov. 2024, arXiv preprint, 10.48550/arXiv.2405.18098

    arXiv doi:10.48550/arxiv.2405.18098 2024

Show all 44 references
  1. [9]

    Nonlinear image recovery with half-quadratic regularization,

    D. Geman and C. Yang, “Nonlinear image recovery with half-quadratic regularization,”IEEE Trans. Image Process., vol. 4, no. 7, pp. 932–946, 1995

    1995

  2. [10]

    Asymptotically exact data augmentation: Models, properties, and algorithms,

    M. V ono, N. Dobigeon, and P. Chainais, “Asymptotically exact data augmentation: Models, properties, and algorithms,”J. Comput. Graph. Stat., vol. 30, no. 2, pp. 335–348, 2021. 11

    2021

  3. [11]

    A distributed split- Gibbs sampler with hypergraph structure for high-dimensional inverse problems,

    P.-A. Thouvenin, A. Repetti, and P. Chainais, “A distributed split- Gibbs sampler with hypergraph structure for high-dimensional inverse problems,”J. Comput. Graph. Stat., vol. 33, no. 3, pp. 814–832, Oct. 2024

    2024

  4. [12]

    Stochastic proximal Langevin algorithm: Potential splitting and nonasymptotic rates,

    A. Salim, D. Koralev, and P. Richtarik, “Stochastic proximal Langevin algorithm: Potential splitting and nonasymptotic rates,” inAdv. in Neural Information Process. Systems, vol. 32. Curran Associates, Inc., 2019, pp. 6649–6661

    2019

  5. [13]

    The SPMD model: Past, present and future,

    F. Darema, “The SPMD model: Past, present and future,” inRecent Advances in Parallel Virtual Machine and Message Passing Interface, Y . Cotronis and J. Dongarra, Eds., Heidelberg, 2001, p. 1

    2001

  6. [14]

    Deep learning techniques for inverse problems in imaging,

    G. Ongie, A. Jalal, C. A. Metzler, R. G. Baraniuk, A. G. Dimakis, and R. Willett, “Deep learning techniques for inverse problems in imaging,”IEEE J. Sel. Inf. Theory, vol. 1, no. 1, pp. 39–56, May 2020. [Online]. Available: https://ieeexplore.ieee.org/document/ 9084378/?arnumb...

    2020

  7. [15]

    Plug- and-Play priors for model based reconstruction,

    S. V . Venkatakrishnan, C. A. Bouman, and B. Wohlberg, “Plug- and-Play priors for model based reconstruction,” inProc. IEEE Glob. Conf. on Sig. and Inf. Processing (GlobalSIP), Dec. 2013, pp. 945–948. [Online]. Available: https://ieeexplore.ieee.org/document/ 6737048/?arnumber=6737048

    2013

  8. [16]

    Plug-and-Play image restoration with deep denoiser prior,

    K. Zhang, Y . Li, W. Zuo, L. Zhang, L. Van Gool, and R. Timofte, “Plug-and-Play image restoration with deep denoiser prior,”IEEE Trans. Pattern Anal. Mach. Intell., vol. 44, no. 10, pp. 6360–6376,

  9. [17]

    NF-ULA: Normalizing flow-based unadjusted Langevin algorithm for imaging inverse problems,

    Z. Cai, J. Tang, S. Mukherjee, J. Li, C. Sch ¨onlieb, and X. Zhang, “NF-ULA: Normalizing flow-based unadjusted Langevin algorithm for imaging inverse problems,”SIAM J. Imaging Sci., vol. 17, no. 2, pp. 820–860, Jun. 2024

    2024

  10. [18]

    Plug-and-Play split Gibbs sampler: Embedding deep generative priors in Bayesian inference,

    F. Coeurdoux, N. Dobigeon, and P. Chainais, “Plug-and-Play split Gibbs sampler: Embedding deep generative priors in Bayesian inference,”IEEE Trans. Image Process., vol. 33, pp. 3496– 3507, 2024. [Online]. Available: https://ieeexplore.ieee.org/document/ 10541919?source=authoralert

    2024

  11. [19]

    Principled probabilistic imaging using diffusion models as Plug-and-Play priors,

    Z. Wu, Y . Sun, Y . Chen, B. Zhang, Y . Yue, and K. L. Bouman, “Principled probabilistic imaging using diffusion models as Plug-and-Play priors,”Adv. in Neural Information Process. Systems, vol. 37, pp. 118 389–118 427, Dec. 2024. [Online]. Available: https://proceedings.neuri...

    2024

  12. [20]

    The little book of deep learning,

    F. Fleuret, “The little book of deep learning,” 2024. [Online]. Available: https://fleuret.org/public/lbdl.pdf

    2024

  13. [21]

    Normalizing flows: An introduction and review of current methods,

    I. Kobyzev, S. J. D. Prince, and M. A. Brubaker, “Normalizing flows: An introduction and review of current methods,”IEEE Trans. Pattern Anal. Mach. Intell., vol. 43, no. 11, pp. 3964–3979, Nov. 2021

    2021

  14. [22]

    A survey on diffusion models for inverse problems,

    G. Daras, H. Chung, C.-H. Lai, Y . Mitsufuji, J. C. Ye, P. Milanfar, A. G. Dimakis, and M. Delbracio, “A survey on diffusion models for inverse problems,” Sep. 2024, arXiv preprint, 10.48550/arXiv.2410.00083

    2024 doi arXiv

  15. [23]

    Algorithm unrolling: Interpretable, efficient deep learning for signal and image processing,

    V . Monga, Y . Li, and Y . C. Eldar, “Algorithm unrolling: Interpretable, efficient deep learning for signal and image processing,”IEEE Signal Process. Mag., vol. 38, no. 2, pp. 18–44, Mar. 2021. [Online]. Available: https://ieeexplore.ieee.org/document/9363511

    2021

  16. [24]

    Dual forward- backward unfolded network for flexible Plug-and-Play,

    A. Repetti, M. Terris, Y . Wiaux, and J.-C. Pesquet, “Dual forward- backward unfolded network for flexible Plug-and-Play,” inProc. Eu- ropean Signal Process. Conf. (EUSIPCO). Belgrade, Serbia: IEEE, Aug. 2022, pp. 957–961

    2022

  17. [25]

    Unfolded proximal neural networks for robust image Gaussian denoising,

    H. T. V . Le, A. Repetti, and N. Pustelnik, “Unfolded proximal neural networks for robust image Gaussian denoising,”IEEE Trans. Image Process., vol. 33, pp. 4475–4487, 2024. [Online]. Available: https://ieeexplore.ieee.org/document/10630640/?arnumber=10630640

    2024

  18. [26]

    Multi-GPU distributed PnP-ULA for high-dimensional imaging inverse problems,

    M. Bouton, P.-A. Thouvenin, A. Repetti, and P. Chainais, “Multi-GPU distributed PnP-ULA for high-dimensional imaging inverse problems,” inProc. IEEE Workshop Stat. Sign. Proc., Edinburgh, U. K., Jun. 2025, pp. 66–70

    2025

  19. [27]

    Beyond a Gaussian denoiser: Residual learning of deep CNN for image denoising,

    K. Zhang, W. Zuo, Y . Chen, D. Meng, and L. Zhang, “Beyond a Gaussian denoiser: Residual learning of deep CNN for image denoising,” IEEE Trans. Image Process., vol. 26, no. 7, pp. 3142–3155, Jul. 2017. [Online]. Available: https://ieeexplore.ieee.org/document/7839189

    2017

  20. [28]

    Proximit ´e et dualit ´e dans un espace hilbertien,

    J. J. Moreau, “Proximit ´e et dualit ´e dans un espace hilbertien,”Bulletin de la soci ´et´e math ´ematique de France, vol. 93, p. 273, 1965. [Online]. Available: https://hal.science/hal-01740635

    1965

  21. [29]

    The science of computing,

    V . Eijkhout, “The science of computing,” 2023. [Online]. Available: https://github.com/VictorEijkhout/TheArtofHPC pdfs

    2023

  22. [30]

    Efficient Bayesian compu- tation by proximal Markov chain Monte Carlo: When Langevin meets Moreau,

    A. Durmus, E. Moulines, and M. Pereyra, “Efficient Bayesian compu- tation by proximal Markov chain Monte Carlo: When Langevin meets Moreau,”SIAM J. Imaging Sci., vol. 11, no. 1, pp. 473–506, Jan. 2018

    2018

  23. [31]

    Nonlinear total variation based noise removal algorithms,

    L. I. Rudin, S. Osher, and E. Fatemi, “Nonlinear total variation based noise removal algorithms,”Phys. D, vol. 60, no. 1-4, pp. 259–268, Nov. 1992

    1992

  24. [32]

    Dualization of signal recovery problems,

    P. L. Combettes, D. D ˜ung, and B. C. V ˜u, “Dualization of signal recovery problems,”Set-V alued Anal, vol. 18, no. 3, pp. 373–404, Dec. 2010

    2010

  25. [33]

    Proximal splitting methods in signal processing,

    P. L. Combettes and J.-C. Pesquet, “Proximal splitting methods in signal processing,” inFixed-Point Algorithms for Inverse Problems in Science and Engineering. Springer, May 2011, vol. 49, pp. 185–212

    2011

  26. [34]

    Vetterli, J

    M. Vetterli, J. Kova ˇcevi´c, and V . K. Goyal,F oundations of Signal Processing. Cambridge: Cambridge University Press, 2014

    2014

  27. [35]

    Scaling painting style transfer,

    B. Galerne, L. Raad, J. Lezama, and J.-M. Morel, “Scaling painting style transfer,”Computer Graphics F orum, vol. 43, no. 4, p. e15155, 2024

    2024

  28. [36]

    Split-and-augmented Gibbs sampler–Application to large-scale inference problems,

    M. V ono, N. Dobigeon, and P. Chainais, “Split-and-augmented Gibbs sampler–Application to large-scale inference problems,”IEEE Trans. Signal Process., vol. 67, no. 6, pp. 1648–1661, Mar. 2019

    2019

  29. [37]

    Accelerating proximal Markov chain Monte Carlo by using an explicit stabilized method,

    M. Pereyra, L. V . Mieles, and K. C. Zygalakis, “Accelerating proximal Markov chain Monte Carlo by using an explicit stabilized method,” SIAM J. Imaging Sci., vol. 13, no. 2, pp. 905–935, Jan. 2020

    2020

  30. [38]

    ImageNet large scale visual recognition challenge,

    O. Russakovsky, J. Deng, H. Su, J. Krause, S. Satheesh, S. Ma, Z. Huang, A. Karpathy, A. Khosla, M. Bernstein, A. C. Berg, and L. Fei- Fei, “ImageNet large scale visual recognition challenge,”Int. J. Comput. Vis., vol. 115, no. 3, pp. 211–252, Dec. 2015

    2015

  31. [39]

    Adam: A Method for Stochastic Optimization,

    D. P. Kingma and J. Ba, “Adam: A Method for Stochastic Optimization,” Jan. 2017, arXiv preprint, 10.48550/arXiv.1412.6980

    2017 doi arXiv

  32. [40]

    Learning maximally monotone operators for image recovery,

    J.-C. Pesquet, A. Repetti, M. Terris, and Y . Wiaux, “Learning maximally monotone operators for image recovery,”SIAM J. Imaging Sci., vol. 14, no. 3, pp. 1206–1237, 2021

    2021

  33. [41]

    NTIRE 2017 challenge on single image super-resolution: Dataset and study,

    E. Agustsson and R. Timofte, “NTIRE 2017 challenge on single image super-resolution: Dataset and study,” inProc. IEEE Conf. Comput. Vis. and Pattern Recog. Workshops (CVPRW), Jul. 2017, pp. 1122–1131

    2017

  34. [42]

    Image quality assessment: From error visibility to structural similarity,

    Z. Wang, A. Bovik, H. Sheikh, and E. Simoncelli, “Image quality assessment: From error visibility to structural similarity,”IEEE Trans. Image Process., vol. 13, no. 4, pp. 600–612, Apr. 2004. [Online]. Available: http://ieeexplore.ieee.org/document/1284395/

    2004

  35. [43]

    Mean squared error: Love it or leave it? A new look at signal fidelity measures,

    Z. Wang and A. C. Bovik, “Mean squared error: Love it or leave it? A new look at signal fidelity measures,”IEEE Signal Process. Mag., vol. 26, no. 1, pp. 98–117, Jan. 2009

    2009

  36. [2021]

    Available: https://ieeexplore.ieee.org/abstract/document/ 9454311

    [Online]. Available: https://ieeexplore.ieee.org/abstract/document/ 9454311

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.