Pith. sign in

REVIEW 4 major objections 6 minor 53 references

A Decentralized Framework for Radio-interferometric Image Reconstruction

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a decentralized framework, which partitions radio-interferometric visibilities by spatial frequency and exchanges filtered deconvolved residuals between nodes, reconstructs images of similar quality to serial imaging…

desk verdict A genuinely new decentralized spatial-frequency parallelization scheme with direct wall-clock evidence of speedups, but the 'similar quality' claim is under-supported because most multi-scale CLEAN runs never converged. read the letter →

arxiv 2507.22128 v1 pith:D6EK2PGP submitted 2025-07-29 astro-ph.IM

S. Wang , S. Mignot , S. Prunet , L. Di Mascolo , M. Spinelli , A. Ferrari This is my paper
classification astro-ph.IM
keywords radiointerferometryimagereconstructionparallelcomputingspatialfrequencypartitioningmajor-minorloopsdeconvolutionmulti-scaleCLEANsparsityregularization
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

Radio-interferometric image reconstruction usually needs all measured spatial frequencies together, which makes parallelizing by spatial frequency difficult. The paper proposes a decentralized scheme that splits the visibilities into two spatial-frequency partitions, each handled by its own compute node, and then has the nodes exchange filtered deconvolved residual images after each major cycle. Each node supplements its local visibilities with the other node's residual, forming a pseudo-full-resolution problem, and the final image is a weighted average of the two node reconstructions. Across four datasets and two reconstruction algorithms, the authors report images of similar quality to serial reconstruction in substantially less wall time, with the largest datasets approaching the theoretical 2x speedup. If it scales, this gives next-generation array imaging pipelines a new axis of parallelization.

What carries the argument

The load-bearing mechanism is the one-major-cycle-delayed exchange of filtered deconvolved residuals between nodes. The filters $\Gamma_L$ and $\Gamma_H$ restrict each residual to its own spatial-frequency domain, normalize it, and make the weights sum to one in the overlap region; for the receiving node, the surrogate residual $\rho^{n-1}_H = \sum_j \Phi^j_H - \Gamma_H \sum_j \bar{\imath}^j_{V_L}$ subtracts the node's own previous estimate, preventing double-counting. The local objective then gains a second data-fidelity term that ties the reconstruction to the other node's filtered residual, so the serial major-minor loop becomes two parallel pseudo-full-resolution problems whose outputs are averaged.

What would settle it

Run serial and parallel multi-scale CLEAN to convergence, not stopping at five major cycles, on a dataset with known ground truth, and compare the final signal-to-noise ratios. A persistent quality gap would mean the surrogate residual does not fully capture the missing visibility information.

Watch

Extended reading notes

Core claim

The central claim is that the missing visibility information at one node can be represented by the other node's filtered deconvolved residual, so each node can solve a full-resolution deconvolution problem despite holding only part of the data. In the two-partition case, the low-spatial-frequency node and the high-spatial-frequency node each start with their local visibilities; from the second major cycle onward they exchange filtered deconvolved residuals, and the receiving node subtracts its own previous estimate to build a surrogate residual. The authors apply this scheme to multi-scale CLEAN and to sparsity-regularized convex reconstruction, and find that the parallel versions match serial image quality while taking substantially less time, with large datasets approaching the optimal 2x speedup.

Load-bearing premise

The framework's speedup claim depends on the filtered deconvolved residual from the other node being a faithful stand-in for the missing visibilities, so that each node's pseudo-full-resolution problem converges to the same image the serial full-data problem would produce.

Editorial extensions

If this is right

  • For two partitions and large datasets, the framework reaches close to the theoretical 2x speedup while producing images of similar quality to the serial reconstruction.
  • Because the exchange happens only once per major cycle, the framework is agnostic to the deconvolution algorithm; the paper demonstrates it with both multi-scale CLEAN and sparsity-regularized convex reconstruction.
  • A full-resolution image is only available after two major cycles, so the gain is largest when the serial method needs many major cycles, which is precisely the regime of large and complex datasets.
  • Scaling beyond two partitions will make inter-node image transfer a bottleneck, but the transferred images are band-limited and can be losslessly compressed, keeping transfer costs roughly linear in the number of partitions.
  • At higher image resolution, deconvolution becomes the dominant cost, so further scaling will require parallel or sparse deconvolution methods.

Reading between the lines

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

  • The pseudo-full-resolution surrogate fits a broader design pattern: any major-minor imaging loop with band-limited residual estimates can be spatially partitioned by frequency, so the framework could extend naturally to many partitions or to facet-based imaging.
  • The two-major-cycle latency implies the method is best suited to deep, high-dynamic-range imaging runs rather than quick-look reconstructions, where serial methods may finish before the parallel one produces a full-resolution image.
  • For arrays with dense central uv coverage, pairing frequency partitioning with baseline-dependent averaging could balance node workloads and reduce idle time, a combination the paper identifies but does not test.
  • The five-major-cycle stopping rule leaves open whether parallel and serial CLEAN converge to the same fixed point; testing convergence would separate a speed-at-equal-quality result from a true equivalence of the surrogate problem.
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

4 major / 6 minor

Summary. The manuscript proposes a decentralized framework for radio-interferometric image reconstruction that parallelizes by spatial frequency across two nodes. Each node reconstructs a full-resolution image using its local visibilities and uses filtered, deconvolved residuals received from the other node as a surrogate for the missing visibilities. The framework is instantiated for multi-scale CLEAN and for sparsity-regularized convex reconstruction (FISTA), and is compared with the serial versions on two simulated and two real datasets. The authors report that the parallel versions achieve significantly better wall-clock times at comparable image quality, with close to a factor-two speedup on the larger datasets, and they provide a sub-process time breakdown and a 10k by 10k scaling test.

Significance. If the result holds, the paper offers a useful additional axis of parallelization for SKA-era imaging, and the general major-minor loop formulation makes it potentially applicable to a range of deconvolution methods. The strengths of the paper are its direct wall-clock measurements on four datasets, the detailed time breakdowns, the application to two different reconstruction algorithms, and the inclusion of both simulated and real data. The open-code commitment and the Wasserstein-distance quality metric for real data are also positive features. The main caveat is that the 'similar quality' claim is made at a fixed, mostly non-converged stopping point, and the equivalence of the decentralized pseudo-full-resolution problem to the serial problem at convergence is not established. These issues are openly acknowledged in Section 6.1 but they directly affect the central claim.

major comments (4)
  1. [§6.1, Figure 5] The claim that the framework produces 'images of similar quality' is not supported for p-msc. The paper states that 'neither msc nor p-msc converged after our tested number of major-cycles for most of our datasets' and that 'it is not clear whether the final converged images will have similar quality.' Because all methods are stopped after exactly five major cycles, the observed equal-time advantage may simply reflect that the parallel method produces good images earlier, while the serial method would catch up or surpass it at convergence. The only converged comparison (HL Tau) is a single dataset. Please add experiments with a larger major-cycle budget or a proper convergence criterion for both serial and parallel methods, and restrict the quality claim to the early-termination regime if convergence cannot be demonstrated.
  2. [§4, Eqs. (5) and (7)] The convergence equivalence of the pseudo-full-resolution surrogate is the load-bearing assumption of the paper. The surrogate term rho^{n-1}_H is built from filtered deconvolved residuals Phi^j_H = Gamma_H i-bar^j_{VH} rather than from the true missing visibilities or residual images. If the minor-loop CLEAN stops early, or if regularization changes the deconvolved residual, the coupled fixed point of the two nodes need not coincide with the serial full-data solution. I do not see a proof of convergence or a consistency argument in the manuscript. Please provide a formal analysis for the linear/regularized case, or an empirical study comparing serial and parallel reconstructions at convergence across all datasets, not only HL Tau.
  3. [§5.2 and §6.1] Several free parameters are tuned per dataset and per node: the constants a and b in lambda_n = a * ||residual_n||_2 * b^n, the multi-scale CLEAN absolute threshold, the minor-cycle budget K=2000, and the fixed five-major-cycle stopping rule. Since both the speedup and quality comparisons depend on these choices, a sensitivity analysis or a table of chosen values is needed to show that the reported advantage is not an artifact of tuning. In particular, the statement that p-L1 required different regularization parameters for the VL and VH nodes suggests a potential extra tuning burden for the parallel method.
  4. [§6.2, Table 3] The measured speedups are partly driven by prototype overhead unrelated to the parallelization strategy. The paper notes that most of the 'disk read' time is actually overhead conforming visibilities to RASCIL's format, and that the parallel methods show 'more than the optimal 2x speedup' on Cygnus A because of this overhead. Repeated runs with error bars, and a separation of read/format overhead from the computational sub-processes, would make the 'close to optimal 2x speedup' claim quantitative and robust. As written, the reader cannot distinguish algorithmic speedup from implementation artifacts.
minor comments (6)
  1. [References] Cai et al. (2019) and Connor et al. (2022) are cited as 'NMRAS'; the journal abbreviation should be MNRAS.
  2. [Eq. (7)] The weights gamma_n and mu_n are only described qualitatively as 'weights corresponding to re-normalizing the PSF and dirty images globally'; please give their explicit definitions.
  3. [Table 3] The note says the wall times are per major cycle averaged across four major cycles, but the text says 'for a single major cycle'; clarify whether the reported numbers are per-cycle averages or totals, and state the number of runs used for the average.
  4. [Figure 5] The quality metric for real datasets is written as ||W(r,n)_5||_2; the subscript 5 is not defined in the text, although it appears to refer to the 5x5 window used for the Wasserstein distance. Define this notation where it is first used.
  5. [Section 5.3] The repository link is only 'provided upon publication'; for a reproducible methods submission, include an anonymized link or a DOI in the manuscript.
  6. [Section 6.1] The sentence 'msc always reconstructed a slightly better image than the p-msc' and the later statement that p-msc 'generally reconstructs images that are of higher quality' are both correct only in different senses, namely per-major-cycle quality versus equal-time quality; please make this distinction explicit to avoid an apparent contradiction.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: speedup is directly wall-clock measured and image quality is benchmarked against ground truth or noise residuals; acknowledged non-convergence limits the quality claim but does not make it circular.

full rationale

The paper's central claims are empirical: improved reconstruction times and similar image quality. The speedup numbers are measured wall-clock times (Section 6.2), not derived from the framework's equations. Quality is evaluated against ground-truth images for simulated datasets (S/N) and against noise-reference residuals via Wasserstein distance for real data (Section 6.1), i.e., external benchmarks independent of the method's fitted parameters. The only parameter choices justified by self-citation are δ=1 (from Wang et al. 2024) and per-dataset λ schedules from preliminary experiments; neither is a fitted version of the target conclusion, and the δ choice is additionally supported by an efficiency argument. The pseudo-full-resolution surrogate (Eqs. 5 and 7) is an approximation, and the paper explicitly states that for multi-scale CLEAN 'neither msc nor p-msc converged after our tested number of major-cycles for most of our datasets' and 'it is not clear whether the final converged images will have similar quality.' That is an honest limitation of the evidence, not a circular step: the claim is not made true by definition, nor does any equation reduce the output to the input. No uniqueness theorem, ansatz-smuggling, or renaming of known results is present. Therefore no circularity is found.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The central claim rests on several hand-chosen parameters (partition geometry, regularization schedule, CLEAN thresholds, stopping rules) and on the modeling assumption that exchanged deconvolved residuals represent missing spatial-frequency information. No new physical entities are introduced.

free parameters (7)
  • Partition radius l = 20, 25, 60, 40 pixels for Sgr B2, Sgr C, HL Tau, Cygnus A
    Chosen to roughly balance visibility counts between the two nodes (Table 2); affects wall time and reconstruction quality.
  • Overlap half-width delta = 1 pixel
    Set to 1 based on the authors' earlier finding in Wang et al. (2024) that delta has little effect on quality; affects filter design and frequency spillage.
  • Regularization parameters a and b in lambda_n = a * ||residual_n||_2 * pow(b,n) = Not reported; dataset- and node-dependent
    Selected through preliminary experiments (Section 5.2); lambda_VL and lambda_VH can differ, and VL needed a more aggressive schedule. This tuning affects the quality comparison.
  • Multi-scale CLEAN absolute threshold = 1e-4 to 2.5e-5 depending on dataset
    Set per dataset (Section 5.2); affects the number of minor iterations and the final image quality.
  • Fixed stopping rules = 5 major cycles, 50 FISTA iterations, K=2000 CLEAN iterations
    Fixed for all methods; multi-scale CLEAN often had not converged after 5 major cycles, so equal-cycle comparisons may not reflect final converged quality (Section 6.1).
  • CLEAN scale set S = {0, 1, 2, 4, 6, 10, 30}
    Chosen to keep reconstruction resolution similar to the sparsity-regularized method (Section 5.2).
  • Image combination weights = 0.5 and 0.5
    Equal weights were chosen after preliminary experiments showed inverse variance weighting gave similar results (Section 5.2).
assumptions (6)
  • domain assumption The radio measurement equation (Eq. 1) and its simplification to a 2D Fourier relation after ignoring D and w terms.
    Standard interferometric imaging model, cited to Smirnov (2011); used throughout the de/gridding step in Eq. 2.
  • domain assumption Filtered deconvolved residuals from other nodes can serve as a surrogate for missing visibility information.
    Core of the framework in Section 3 and Eqs. 5 and 7; not proven, only empirically tested.
  • ad hoc to paper The filters Gamma_L and Gamma_H, normalized by variance with weights in L intersect H summing to one, make the data-fidelity terms commensurable.
    Design choice in Section 4.1; no derivation that this weighting is optimal or unbiased.
  • domain assumption Uniform weighting and the improved w-gridder correctly represent the measurement operator.
    Used for all datasets and inherited from RASCIL and Ye et al. (2022).
  • domain assumption For real datasets, the residual of a perfect reconstruction should be statistically indistinguishable from the jackknife noise reference.
    Quality metric for HL Tau and Cygnus A in Section 6.1; assumes the noise model and jackknife estimate are unbiased.
  • domain assumption The PSF convolution operator H is a sufficient surrogate for the full measurement operator inside the minor loop.
    Standard major-minor loop assumption used in Eqs. 4 through 7.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Decentralized Framework for Radio-interferometric Image Reconstruction." pith.science (2026). https://pith.science/paper/D6EK2PGP

@misc{pith2026250722128,
  author       = {Pith},
  title        = {Pith review of: A Decentralized Framework for Radio-interferometric Image Reconstruction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D6EK2PGP}},
  note         = {Machine review of arXiv:2507.22128}
}
abstract

The advent of large aperture arrays, such as the ones currently under construction for the SKA project, allows for observing the Universe in the radio-spectrum at unprecedented resolution and sensitivity. To process the enormous amounts of data produced by these telescopes, scalable software pipelines are required. This paper helps address this by proposing a framework that allows for decentralized radio-interferometric image reconstruction, parallelizing by spatial frequency. This is achieved by creating pseudo-full-resolution problems for each node by using the local visibilities together with previous major cycle reconstructed images from the other nodes. We apply the proposed framework to both multiscale CLEAN and sparsity regularized convex reconstruction and compare them to their serial counterparts across four different data sets of varying properties in the context of two visibility partitions. We found that the parallelization framework allows for significantly improved reconstruction times for images of similar quality. This was especially the case for our larger datasets where we were able to achieve close to the optimal $2\times$ speedup.

Figures

Figures reproduced from arXiv: 2507.22128 by the authors.

Figure 1
Figure 1. A high-level overview of the radio-interferometric pipeline. A reconstructed image ˆı n is compared to the measured visibilities v in the de/gridding step, which outputs the residual ˜ı n in the spatial domain. This is passed to the estimation algorithm which generates ¯ı n , an estimate of ˜ı n without the effects of the mea￾surement operator F †GG†F. The current reconstructed image ˆı n is then updated to ˆı n+1 b… view at source ↗
Figure 2
Figure 2. An illustration of our decentralized framework in the context of two partitions VL and VH . We reconstruct two full￾resolution images in parallel, achieved by sharing the filtered de￾convolved residuals, Φn H and Φn L , between different nodes after each major cycle, denoted as MC in the figure. The deconvolved residu￾als are filtered both for normalization purposes, as well as to ensure that only spatial frequencie… view at source ↗
Figure 4
Figure 4. Dirty images of the four datasets we use for our experi￾ments. The top two are simulated, whereas the bottom two are real observations performed with ALMA and the VLA, respectively [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figures from the paper (3 more)
Figure 5
Figure 5. Figure 5: Time vs. accuracy for all four of our datasets, with each circle representing a major-cycle. The left two simulated datasets use S/N in dB for accuracy, whereas the two right real datasets use the L 2 norm of the per-pixel Wasserstein distances between the noisy refere…
Figure 6
Figure 6. Figure 6: The breakdown of computation time taken by the various sub-processes for our four methods. ing software, such as WSClean (Offringa et al. 2014), and also allow us to obtain results for much larger datasets. We also note that the overall conclusions of our results shoul…
Figure 7
Figure 7. Figure 7: A percentage breakdown of processing time used by the various sub-processes of the L1 method across our four datasets. the image sizes of 800MB, as opposed to roughly 24MB for the original test case of 1728 × 1728 pixels [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

53 extracted references · 47 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter doi edition editor eprint howpublished institution journal key month number organization pages publisher school series title misctitle type volume year version url label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts ...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION format.url url empty "" new.block "" url * "" * if FUNCTION format.eprint eprint empty "" archivePrefix empty "" archivePrefix "arXiv" = new.block " " eprint * " " * new.block " " eprint * " " * if if if FUNCTION format.doi doi empty "" " " doi * " " * if FUNCTION format.pid doi empty eprint empty ur...

  3. [3]

    +;!᥿23If& . & o&敕 @ȯ 0 S)

    thebibliography [1] 20pt to REFERENCES 6pt =0pt -12pt 10pt plus 3pt =0pt =0pt =1pt plus 1pt =0pt =0pt -12pt =13pt plus 1pt =20pt =13pt plus 1pt \@M =10000 =-1.0em =0pt =0pt 0pt =0pt =1.0em @enumiv\@empty 10000 10000 `\.\@m \@noitemerr \@latex@warning Empty `thebibliography' environment \@ifnextchar \@reference \@latexerr Missing key on reference command E...

    2021

  4. [4]

    2024, ApJS, 273, 3

    Aghabiglou, A., Chu, C., Dabbech, A., & Wiaux, Y. 2024, ApJS, 273, 3

    work page 2024

  5. [5]

    H., Magland, J., & Af Klinteberg, L

    Barnett, A. H., Magland, J., & Af Klinteberg, L. 2019, SJSC, 41, C479

    work page 2019

  6. [6]

    2009, SIIMS, 2, 183–202

    Beck, A., & Teboulle, M. 2009, SIIMS, 2, 183–202

    work page 2009

  7. [7]

    Bezanson, J., Edelman, A., Karpinski, S., & Shah, V. B. 2017, SIAMR, 59, 65

    work page 2017

  8. [8]

    Cai, X., Pratley, L., & McEwen, J. D. 2019, NMRAS, 485, 4559

    work page 2019

Show all 53 references
  1. [9]

    E., McEwen, J

    Carrillo, R. E., McEwen, J. D., & Wiaux, Y. 2014, MNRAS, 439, 3591

    2014

  2. [10]

    L., Ravi, V., & Hallinan, G

    Connor, L., Bouman, K. L., Ravi, V., & Hallinan, G. 2022, NMRAS, 514, 2614

    2022

  3. [11]

    Cornwell, T. J. 2008, IEEE JSTSP, 2, 793

    2008

  4. [12]

    J., Golap, K., & Bhatnagar, S

    Cornwell, T. J., Golap, K., & Bhatnagar, S. 2008, ISTSP, 2, 647

    2008

  5. [13]

    J., & Perley, R

    Cornwell, T. J., & Perley, R. A. 1992, A & A, 261, 353

    1992

  6. [14]

    J., Wortmann, P., Nikolic, B., Wang, F., & Stolyarov, V

    Cornwell, T. J., Wortmann, P., Nikolic, B., Wang, F., & Stolyarov, V. 2020, Radio Astronomy Simulation, Calibration and Imaging Library

    2020

  7. [15]

    2015, A & A, 576, A7

    Dabbech, A., Ferrari, C., Mary, D., et al. 2015, A & A, 576, A7

    2015

  8. [16]

    2005, JPDC, 65, 1108

    Dalc \' n, L., Paz, R., & Storti, M. 2005, JPDC, 65, 1108

    2005

  9. [17]

    2023, Nature, 615, 809

    Di Mascolo, L., Saro, A., Mroczkowski, T., et al. 2023, Nature, 615, 809

    2023

  10. [18]

    N., Starck, J.-L., et al

    Garsden, H., Girard, J. N., Starck, J.-L., et al. 2015, A & A, 575, A90

    2015

  11. [19]

    2023, RASTI, 2, 91

    Gheller, C., Taffoni, G., & Goz, D. 2023, RASTI, 2, 91

    2023

  12. [20]

    Godard, S. 2024. https://github.com/sysstat/sysstat

    2024

  13. [21]

    2022, ApJ, 925, 165

    Heywood, I., Rammala, I., Camilo, F., et al. 2022, ApJ, 925, 165

    2022

  14. [22]

    H \"o gbom, J. A. 1974, A & AS, 15, 417

    1974

  15. [23]

    2025, A & A, 693, A225

    Jarret, A., Kashani, S., Rué-Queralt, J., et al. 2025, A & A, 693, A225

    2025

  16. [24]

    G., Waterson, M., Alachkar, B., et al

    Labate, M. G., Waterson, M., Alachkar, B., et al. 2022, JATIS, 8, 011024

    2022

  17. [25]

    G., & Zhifeng, Z

    Mallat, S. G., & Zhifeng, Z. 1993, ITSP, 41, 3397

    1993

  18. [26]

    B., & Whitney, D

    Mann, H. B., & Whitney, D. R. 1947, The annals of mathematical statistics, 50

    1947

  19. [27]

    2016, A & C, 16, 140

    Merry, B. 2016, A & C, 16, 140

    2016

  20. [28]

    2022, in Proceedings of the 2022 IEEE Workshop on Signal Processing Systems

    Monnier, N., Guibert, D., Tasse, C., et al. 2022, in Proceedings of the 2022 IEEE Workshop on Signal Processing Systems

    2022

  21. [29]

    R., McKinley, B., Hurley-Walker, N., et al

    Offringa, A. R., McKinley, B., Hurley-Walker, N., et al. 2014, MNRAS, 444, 606

    2014

  22. [30]

    E., Repetti, A., et al

    Onose, A., Carrillo, R. E., Repetti, A., et al. 2016, MNRAS, 462, 4314

    2016

  23. [31]

    D., d'Avezac, M., et al

    Pratley, L., McEwen, J. D., d'Avezac, M., et al. 2019, arXiv preprint arXiv:1903.04502

    2019 arXiv

  24. [32]

    Rau, U., & Cornwell, T. J. 2011, A & A, 532, A71

    2011

  25. [33]

    Romein, J. W. 2012, in Proceedings of the 26th ACM international conference on Supercomputing, 321

    2012

  26. [34]

    2018, A & A, 612, A39

    Romero, C., McWilliam, M., Mac \' as-P \'e rez, J.-F., et al. 2018, A & A, 612, A39

    2018

  27. [35]

    2022, A & A, 664, A134

    Schmidt, K., Geyer, F., Fröse, S., et al. 2022, A & A, 664, A134

    2022

  28. [36]

    Sebokolodi, M. L. L., Perley, R., Eilek, J., et al. 2020, ApJ, 903, 36

    2020

  29. [37]

    Smirnov, O. M. 2011, A & A, 527, A106

    2011

  30. [38]

    P., Dewdney, P., & Cremonini, A

    Swart, G. P., Dewdney, P., & Cremonini, A. 2022, JATIS, 8, 011021

    2022

  31. [39]

    2018, A & A, 611, A87

    Tasse, C., Hugo, B., Mirmont, M., et al. 2018, A & A, 611, A87

    2018

  32. [40]

    2022, MNRAS, 518, 604–622

    Terris, M., Dabbech, A., Tang, C., & Wiaux, Y. 2022, MNRAS, 518, 604–622

    2022

  33. [41]

    2023, MNRAS, 521, 1–19

    Thouvenin, P.-A., Abdulaziz, A., Dabbech, A., Repetti, A., & Wiaux, Y. 2023, MNRAS, 521, 1–19

    2023

  34. [42]

    Van der Tol, S., Veenboer, B., & Offringa, A. R. 2018, A & A, 616, A27

    2018

  35. [43]

    2025, arXiv e-prints arXiv:2501.03150

    Van Marrewijk, J., Kaasinen, M., Popping, G., et al. 2025, arXiv e-prints arXiv:2501.03150

    2025 arXiv

  36. [44]

    J., Williams, W

    Van Weeren, R. J., Williams, W. L., Hardcastle, M. J., et al. 2016, ApJS, 223, 2

    2016

  37. [45]

    Veenboer, B., Petschow, M., & Romein, J. W. 2017, in IEEE International Parallel and Distributed Processing Symposium, 545

    2017

  38. [46]

    2009, Optimal transport: old and new, Vol

    Villani, C., et al. 2009, Optimal transport: old and new, Vol. 338 (Springer)

    2009

  39. [47]

    2024, A & A, 692, A61

    Wang, S., Prunet, S., Mignot, S., & Ferrari, A. 2024, A & A, 692, A61

    2024

  40. [48]

    2009, ApJ, 707, 1201

    Wei , A., Kov \'a cs, A., Coppin, K., et al. 2009, ApJ, 707, 1201

    2009

  41. [49]

    Wiaux, Y., Jacques, L., Puy, G., Scaife, A. M. M., & Vandergheynst, P. 2009, MNRAS, 395, 1733

    2009

  42. [50]

    J., Willis, A

    Wijnholds, S. J., Willis, A. G., & Salvini, S. 2018, MNRAS, 476, 2029

    2018

  43. [51]

    J., Dijkema, T

    Wijnholds, S. J., Dijkema, T. J., Groot, H., et al. 2023, in XXXVth General Assembly and Scientific Symposium of the International Union of Radio Science (URSI GASS) (IEEE), 1–4

    2023

  44. [52]

    2024, A & A, 687, A152

    Wortmann, P., Kent, J., & Nikolic, B. 2024, A & A, 687, A152

    2024

  45. [53]

    F., Tan, S

    Ye, H., Gull, S. F., Tan, S. M., & Nikolic, B. 2022, MNRAS, 510, 4110

    2022

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.