Pith. sign in

REVIEW 3 major objections 5 minor 47 references

Emulating CO Line Radiative Transfer with Deep Learning

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read COEmuNet, a 3D convolutional neural network, emulates CO line radiative transfer for stellar-wind models with a median relative error around 7 percent and a roughly 1000-fold speedup over a classical solver.

desk verdict A well-executed first CNN emulator for CO line maps, but its headline accuracy is likely inflated by a view-level split that lets the network see every wind model in training. read the letter →

arxiv 2507.11398 v1 pith:HGQRFRES submitted 2025-07-15 astro-ph.IM astro-ph.GA

Shiqi Su , Frederik De Ceuster , Jaehoon Cha , Mark I. Wilkinson , Jeyan Thiyagalingam , Jeremy Yates , Yi-Hang Zhu , Jan Bolte This is my paper · ORCID
classification astro-ph.IMastro-ph.GA MSC 68T0785A25
keywords radiativetransfersurrogatemodelconvolutionalneuralnetworkCOlineemissionAGBstarssyntheticobservationsmachinelearninghydrodynamics
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

Computing the CO line emission from an evolved stellar wind is expensive because every sightline requires solving the radiative transfer equation through a three-dimensional, structured outflow. This paper claims that a 3D convolutional network trained on 2.18 million pairs of hydrodynamic models and solver-generated images can learn that mapping well enough to act as a surrogate: a median relative error of 6.9 percent against the Pomme solver, a structural similarity of 0.91, and a GPU inference time about a thousand times shorter than the solver. The emulator takes the line-of-sight velocity, kinetic temperature, and CO number density as inputs and returns a multi-frequency intensity image cube, turning a radiative-transfer solve into a single forward pass. Such a speedup would make parameter-space fitting to ALMA and JWST observations, real-time simulation visualization, and the inclusion of line transfer in larger simulations computationally practical. The paper is explicitly a proof of concept, trained on LTE line transfer for the CO J=2-1 transition over seven frequency channels and arbitrary viewing directions.

What carries the argument

The load-bearing object is the learned map $\tilde{I}_{\theta}(\boldsymbol{\xi})$ from the three physical fields — line-of-sight velocity $v_z(\boldsymbol{x})$, kinetic temperature $T(\boldsymbol{x})$, and CO number density $n_{\rm CO}(\boldsymbol{x})$, each on a $64^3$ grid — to the synthetic observation $I_\nu(\boldsymbol{\xi}, \hat{\boldsymbol{n}})$, a $7 \times 64 \times 64$ intensity cube spanning seven frequency channels and one viewing direction. The architecture is an asymmetric encoder-decoder: three parallel encoder branches of residual convolution blocks compress each field from resolution 64 to 4, a flattened latent code is refined by two fully connected layers, and a decoder upsamples back to the image cube using nearest-neighbour interpolation and 3D convolutions. The training objective mixes mean-squared error with a discrete-Fourier-transform frequency loss (Zhang et al. 2020) to sharpen texture reconstruction, and the dataset is built by rotating each of 10,903 AMRVAC wind models into 200 orientations, re-gridding them into a common spherical domain, and solving the formal line integral along the z-axis with the Pomme solver to produce ground truth.

What would settle it

Hold out entire hydrodynamic models instead of individual views: train on a random 80 percent of the 10,903 AMRVAC wind models, validate on 10 percent, and test on the remaining 10 percent, using all 200 rotated views of each test model only in the test set. If the median MaxRel on that model-level test set exceeds the paper's 10 percent acceptance threshold (or rises well above the reported 6.9 percent), the claim that COEmuNet generalizes to new wind models is falsified; the same experiment on a single wind model from an unseen binary-parameter regime would test extrapolation.

Watch

Extended reading notes

Core claim

The central discovery is that the multi-frequency position-velocity image cube $I_\nu(\boldsymbol{\xi}, \hat{\boldsymbol{n}})$ — the specific intensity at seven frequencies along an arbitrary line of sight — is a function of the three volumetric fields entering the radiative transfer problem, and that a neural network can represent this function accurately and cheaply. COEmuNet encodes the three $64^3$ fields through separate residual-branch encoders, merges them through a shared latent code, and decodes to a $7 \times 64 \times 64$ intensity cube, trained with a combined pixel-space mean-squared-error and Fourier-domain loss. On a randomly split 10 percent test set of the 2,180,600 sample pairs, the network achieves a median MaxRel of 6.9 percent, ZNCC of 0.97, and SSIM of 0.91, while running about 1000 times faster than the Pomme solver on GPU for the seven-frequency case. The authors frame the result as a proof of concept: LTE is assumed to match the training-data solver, scattering is neglected, and the geometry is restricted to binary-perturbed AGB outflows observed through 200 rotations of each wind model.

Load-bearing premise

The reported accuracy is measured on a test set formed by randomly splitting the 2.18 million data pairs, not by holding out entire hydrodynamic models: because each of the 10,903 wind models contributes 200 rotated views, the same stellar wind can appear in both training and test data, so the 6.9 percent median error may not reflect performance on a truly new wind model.

Editorial extensions

If this is right

  • Fitting models to ALMA and JWST observations of AGB wind structures can shift from expensive repeated forward solves to near-instant neural evaluations, making large parameter-space searches practical.
  • Simulation studies can visualize synthetic line cubes in real time, rotating a wind model and seeing its predicted emission at any angle during an interactive session.
  • Where radiative transfer is currently replaced by cheap sub-grid recipes inside larger simulations, CO line emission can now be computed on the fly at a fraction of the previous cost.
  • The accuracy of the emulator degrades on a tail of hard cases the paper characterizes — 20.9 percent of test samples have MaxRel above 20 percent — so applications needing faithful spiral-arm or high-gradient detail must budget for those outliers.
  • Because the CPU inference time is slower than the numerical solver, the practical speedup is tied to GPU availability; on CPU-only environments the surrogate offers no advantage.

Reading between the lines

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

  • If the network generalizes to genuinely unseen wind models, the same encoder-decoder pipeline should transfer to non-LTE CO line transfer and to other molecules, since the architecture learns from data and the LTE assumption is not baked into the network structure.
  • The random split almost certainly inflates the reported accuracy, because each of the 10,903 wind models contributes 200 viewing angles to the pool; a model-level holdout test would give a truer estimate of performance on new winds and is a natural prerequisite before using the emulator to fit observations.
  • A fairer speed comparison would pit the GPU emulator against a GPU-optimized solver or a CPU emulator against the CPU solver; the 1000x figure specifically compares GPU neural inference with CPU numeric integration.
  • A direct extension with observational payoff would be to condition the network on binary orbital parameters or to train one model across multiple CO transitions, converting the proof of concept into a reusable tool for the ATOMIUM survey.
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

3 major / 5 minor

Summary. The paper presents COEmuNet, a 3D convolutional encoder-decoder neural network trained to emulate CO J=2-1 line radiative transfer solutions from the Pomme solver for 3D hydrodynamic models of binary-perturbed AGB star winds. It generates synthetic position-velocity images at seven frequency channels for arbitrary viewing directions. The authors report a median MaxRel error of 6.9%, ZNCC 0.97, and SSIM 0.91 on a held-out test set, together with a roughly 1000x GPU speedup over Pomme on CPU. The intended application is rapid model fitting to ALMA/JWST observations and possible integration into larger simulations.

Significance. If the reported accuracy persists for genuinely unseen hydrodynamic models, COEmuNet would be a practically useful surrogate, since the paper identifies a real computational bottleneck in spectral-line radiative transfer. The paper is clearly written, provides public code, gives substantial detail on data preparation and training, and candidly discusses important limitations (LTE assumption, restricted to one CO line, specific binary morphology). The speedup claim is plausible for GPU deployment. However, the headline accuracy figure is currently not backed by a generalization test to new stellar-wind models, which is essential for the stated application.

major comments (3)
  1. [Section 2.2.1 and Table 1] The reported accuracy (median MaxRel 6.9%) is measured on a test set formed by randomly splitting the 2,180,600 sample pairs, but each of the 10,903 AMRVAC models contributes 200 rotated views of the same underlying wind structure. A random sample-level split therefore places the same physical wind model in both training and test partitions, and the network can interpolate between views of a seen model rather than generalize to a new wind. Since the stated application is interpreting observations of unseen winds, this is the central evaluation metric and it is not sufficient. Please retrain or re-evaluate the model under a model-level split (e.g., hold out all 200 views of a set of whole hydrodynamic models) and report MaxRel, ZNCC, and SSIM for that protocol. This is directly testable with the released code.
  2. [Section 2.5.1 and Equations (12)-(16)] The global preprocessing statistics (mean, variance, median, minimum, maximum) used for standardization and normalization are computed from the entire 7TB dataset, including the test partition, before the train/test split is applied. This leaks information from the test set into the training process through the input normalization and can slightly inflate the reported accuracy. The statistics should be estimated on the training split only and then applied unchanged to validation and test splits.
  3. [Section 3.2 and Table 2] The headline '1000 times speedup' compares COEmuNet on an A100 GPU against Pomme on an Intel Xeon CPU, even though the paper itself notes that COEmuNet on CPU is slower than Pomme. The abstract should state the hardware comparison explicitly (GPU surrogate versus CPU solver), and ideally the paper should include a CPU-to-CPU comparison or a clear statement that the speedup applies only to GPU deployment. Without this qualifier, the claim is misleading.
minor comments (5)
  1. [Equation (30)] The MaxRel equation is ambiguous: it applies a 'Median' to an expression that already sums over all pixels and frequencies, leaving unclear whether the median is over test samples, pixels, or frequency channels. Please specify the aggregation order.
  2. [Section 3 and Figure 3 caption] The text in Section 3 refers to the 'validation dataset' for the cumulative MaxRel distribution, while the caption of Figure 3 says 'test dataset'. These should be reconciled.
  3. [Equations (31) and (36)] The ZNCC formula should include the summation over pixels in the numerator; as written, it resembles a pointwise ratio. In Equation (36), 'σ_{I_ν y}' appears to be a typo for the cross-covariance term σ_{xy}.
  4. [Section 2.2.1] The paper describes 200 viewing directions as 'arbitrary viewing directions'; it would be more precise to say that the model is trained and evaluated on a finite set of 200 rotations sampling the viewing sphere, since continuous arbitrary angles are not demonstrated.
  5. [Data Availability] The code is publicly available, but the data are only available 'on reasonable request'. For reproducibility, consider releasing at least a subset of the dataset or a derived downsampled version.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: COEmuNet is trained and evaluated against the external Pomme radiative-transfer solver, and the paper explicitly distinguishes emulating the solver from reproducing observations.

full rationale

The paper's central claim is that COEmuNet approximates the output of the Pomme radiative-transfer solver. The training targets are Pomme-computed intensities, and the test targets are held-out Pomme-computed intensities generated with the same forward model; this is standard surrogate-model validation, not a circular derivation. The authors state explicitly: 'We do not have a ground truth dataset derived from observational data. Instead, we generate our training dataset using numerical simulations.' Pomme is cited as an external, open-source solver with stated assumptions (LTE, no scattering), and the emulator's accuracy is established by direct comparison of its outputs to Pomme outputs on a held-out split, not by invoking Pomme's correctness as a premise. The only self-citation is the reference to Pomme (De Ceuster et al. 2024), which is not load-bearing: the conclusion is that the emulator matches the solver, and the solver itself is independently reproducible code. A potential limitation is that the dataset is split randomly at the sample level, so the 200 rotated views of a single hydrodynamic model can appear in both training and test sets; this affects generalization claims about unseen wind models, but it is an evaluation-protocol concern, not circularity, because no training target is identical to a test target by construction and the network must still learn the radiative-transfer mapping. The derivation chain is therefore self-contained and no circular step is present.

Assumptions & free parameters 9 free parameters · 5 assumptions · 0 invented entities

The central accuracy result is an empirical fit: COEmuNet's weights are fitted to Pomme outputs on a large training set. The listed hyperparameters and preprocessing choices are additional hand-set degrees of freedom. No new physical entities are postulated.

free parameters (9)
  • COEmuNet weights = not reported
    Learned by Adam on the training split; these are the central fitted parameters of the emulator.
  • loss mixing weight lambda = 0.8
    Chosen by grid search; balances MSE and frequency-domain loss.
  • number of frequency bins = 7
    Hand-picked to balance spectral coverage vs dataset size; no sensitivity study.
  • grid resolution = 64^3
    Chosen as a trade-off between fidelity and computational cost.
  • network width schedule = 8-16-32 channels
    Hand-designed architecture; no ablation study.
  • learning rates = 4e-3 pretrain, 1e-4 fine-tune
    Grid-searched for pretrain; fixed for fine-tune.
  • batch sizes = 512 pretrain, 2048 fine-tune
    Set by GPU memory and throughput.
  • preprocessing statistics = global mean, std, min, median
    Computed from training data chunks and used to normalize inputs and targets.
  • MaxRel acceptability threshold = 10%
    Adopted as a performance target for observational use.
assumptions (5)
  • domain assumption LTE assumption for level populations
    Level populations follow Boltzmann at local kinetic temperature (Eq. 10); authors note LTE is not correct for circumstellar CO but use it for data availability.
  • domain assumption No scattering in the RT equation
    Equation 1 omits scattering, justified as negligible at these wavelengths.
  • domain assumption Gaussian line profile from Doppler broadening
    Line profile is assumed Gaussian, dominated by thermal and turbulent motions (Eq. 6).
  • domain assumption Hydrodynamic models are representative
    AMRVAC binary-perturbed AGB models are assumed to capture the structures needed for training; no independent validation against observed AGB winds.
  • domain assumption Pomme solver is the ground truth
    The Pomme RT solver output is treated as the reference; no comparison to other RT codes or observations in this paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Emulating CO Line Radiative Transfer with Deep Learning." pith.science (2026). https://pith.science/paper/HGQRFRES

@misc{pith2026250711398,
  author       = {Pith},
  title        = {Pith review of: Emulating CO Line Radiative Transfer with Deep Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HGQRFRES}},
  note         = {Machine review of arXiv:2507.11398}
}
read the original abstract

Modelling carbon monoxide (CO) line radiation is computationally expensive for traditional numerical solvers, especially when applied to complex, three-dimensional stellar atmospheres. We present COEmuNet, a 3D convolutional neural network (CNN)-based surrogate model that emulates CO line radiation transport with high accuracy and efficiency. It consists of an asymmetric encoder-decoder design that takes 3D hydrodynamical models as inputs and generates synthetic observations of evolved stellar atmospheres. The model is trained on data from hydrodynamic simulations of Asymptotic Giant Branch (AGB) stars perturbed by a companion. Given a set of input parameters, including velocity fields, kinetic temperature distribution, and CO molecular number densities, the COEmuNet model emulates spectral line observations with a median relative error of ~7% compared to a classical numerical solver of the radiative transfer equation, measured over seven frequency channels and arbitrary viewing directions. Besides, COEmuNet delivers a 1000 times speedup, enabling efficient model fitting to observational datasets, real-time visualization of simulations and progress toward integration in large-scale cosmological simulations.

Figures

Figures reproduced from arXiv: 2507.11398 by the authors.

Figure 1
Figure 1. Overview of the COEmuNet architecture. The model consists of an encoder and a decoder. The encoder has three separate branches, each processing a specific spectral feature: 𝑉𝑧 (𝒙), 𝑇 (𝒙), 𝑛𝐶𝑂 (𝒙)). Each branch applies convolutional blocks with residual connections, progressively increasing feature channels while reducing spatial resolution. The latent space concatenates the encoded features and refines them through … view at source ↗
Figure 2
Figure 2. Training history of COEmuNet in terms of the combined frequency (Equation 28) and MSE (Equation 27) loss function (left panel), along with validation metrics (middle and right panels) on validation set. The left panel shows the steady decrease in the training and validation losses, which stabilize around epoch 50, with the training loss converging to approximately 0.11. The middle and right panels present the valida… view at source ↗
Figure 3
Figure 3. Cumulative distribution of MaxRel values across the test dataset. The steep rise for MaxRel less than 10% indicates that the majority (69.9%) of the test dataset has low MaxRel values, while the flatness of the curve beyond 30% shows that only a small fraction of samples exhibit high errors. This distribution highlights the overall accuracy of COEmuNet while also indicating the presence of some outliers with signifi… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Visualization of intensity maps across multiple frequencies. The top row shows the target intensity maps from frequency 𝜈13 to 𝜈19, while the middle row displays corresponding COEmuNet reconstructions across the same frequency range. The bottom row shows the MaxRel val…
Figure 5
Figure 5. Figure 5: Zoom-in of central 32 by 32 pixels of the target and predicted images from [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Target and predicted image pairs for 3 frequencies as in [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Target and predicted pairs for the validation sample with the highest MaxRel value of 53.10%, revealing significant reconstruction discrepancies, particularly in the spiral arms and high-intensity regions. been explored and may require retraining or fine-tuning on doma…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 21 canonical work pages

  1. [1]

    F., 2018, @doi [arXiv preprint arXiv:1803.08375] 10.48550/arXiv.1803.08375

    Agarap A. F., 2018, @doi [arXiv preprint arXiv:1803.08375] 10.48550/arXiv.1803.08375

    arXiv doi:10.48550/arxiv.1803.08375 2018

  2. [2]

    Altay G., Croft R. A. C., Pelupessy I., 2008, @doi [Monthly Notices of the Royal Astronomical Society] 10.1111/j.1365-2966.2008.13212.x , 386, 1931

    arXiv 2008

  3. [3]

    W., Pratt V., Rivest R

    Blum M., Floyd R. W., Pratt V., Rivest R. L., Tarjan R. E., 1973, @doi [Journal of Computer and System Sciences] 10.1016/S0022-0000(73)80033-9 , 7, 448

    doi:10.1016/s0022-0000(73)80033-9 1973

  4. [4]

    Ceulemans T., De Ceuster F., Decin L., Yates J., 2024, @doi [Astronomy and Computing] https://doi.org/10.1016/j.ascom.2024.100889 , 49, 100889

    work page arXiv 2024

  5. [5]

    F., Golub G

    Chan T. F., Golub G. H., LeVeque R. J., 1979, Technical Report CS-TR-79-773, Updating formulae and a pairwise algorithm for computing sample variances, http://i.stanford.edu/pub/cstr/reports/cs/tr/79/773/CS-TR-79-773.pdf . Department of Computer Science, Stanford University, http://i.stanford.edu/pub/cstr/reports/cs/tr/79/773/CS-TR-79-773.pdf

    work page 1979

  6. [6]

    De Ceuster F., Homan W., Yates J., Decin L., Boyle P., Hetherington J., 2019, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stz3557 , 492, 1812

    work page doi:10.1093/mnras/stz3557 2019

  7. [7]

    De Ceuster F., et al., 2020, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/staa3199 , 499, 5194

    work page doi:10.1093/mnras/staa3199 2020

  8. [8]

    De Ceuster F., et al., 2022, @doi [Journal of Open Source Software] 10.21105/joss.03905 , 7, 3905

    doi:10.21105/joss.03905 2022

Show all 47 references
  1. [9]

    De Ceuster F., Ceulemans T., Decin L., Danilovich T., Yates J., 2024, @doi [arXiv preprint arXiv:2402.18525] 10.48550/arXiv.2402.18525

    2024 doi arXiv

  2. [10]

    Decin L., et al., 2020, @doi [Science] 10.1126/science.abb1229 , 369, 1497

    2020 doi

  3. [11]

    Di Stefano L., Mattoccia S., Tombari F., 2005, @doi [Pattern Recognition Letters] 10.1016/j.patrec.2005.03.022 , 26, 2129

    2005 doi

  4. [12]

    Dullemond C. P., Juhasz A., Pohl A., Sereshti F., Shetty R., Peters T., Commercon B., Flock M., 2012, RADMC-3D: A multi-purpose radiative transfer tool, Astrophysics Source Code Library ( @eprint ascl 1202.015 ), http://ascl.net/1202.015

    2012

  5. [13]

    Frostholm T., Haugb lle T., Grassi T., 2018, @doi [arXiv preprint arXiv:1809.05541] 10.48550/arXiv.1809.05541

    2018 doi arXiv

  6. [14]

    Fukushima K., Miyake S., Ito T., 1983, @doi [IEEE Transactions on Systems, Man, and Cybernetics] 10.1109/TSMC.1983.6313076 , SMC-13, 826

    1983

  7. [15]

    A., et al., 2022, @doi [Astronomy & Astrophysics] 10.1051/0004-6361/202140431 , 660, A94

    Gottlieb C. A., et al., 2022, @doi [Astronomy & Astrophysics] 10.1051/0004-6361/202140431 , 660, A94

    2022 doi

  8. [16]

    IEEE, Las Vegas, NV, USA, pp 770--778, @doi 10.1109/CVPR.2016.90 , http://ieeexplore.ieee.org/document/7780459/

    He K., Zhang X., Ren S., Sun J., 2016, in 2016 IEEE Conference on Computer Vision and Pattern Recognition ( CVPR ). IEEE, Las Vegas, NV, USA, pp 770--778, @doi 10.1109/CVPR.2016.90 , http://ieeexplore.ieee.org/document/7780459/

    2016

  9. [17]

    Hornik K., Stinchcombe M., White H., 1989, @doi [Neural Networks] 10.1016/0893-6080(89)90020-8 , 2, 359

    1989 doi

  10. [18]

    Ioffe S., Szegedy C., 2015, @doi [arXiv preprint arXiv:1502.03167] 10.48550/arXiv.1502.03167

    2015 doi arXiv

  11. [19]

    Keppens R., Nool M., Tóth G., Goedbloed J., 2003, @doi [Computer Physics Communications] 10.1016/S0010-4655(03)00139-5 , 153, 317

    2003 doi

  12. [20]

    Keppens R., Popescu Braileanu B., Zhou Y., Ruan W., Xia C., Guo Y., Claes N., Bacchini F., 2023, @doi [Astronomy & Astrophysics] 10.1051/0004-6361/202245359 , 673, A66

    2023 doi

  13. [21]

    P., Ba J., 2014, Adam: A Method for Stochastic Optimization , @doi 10.48550/ARXIV.1412.6980

    Kingma D. P., Ba J., 2014, Adam: A Method for Stochastic Optimization , @doi 10.48550/ARXIV.1412.6980

    2014 doi arXiv

  14. [22]

    IEEE Comput

    Krattenthaler W., Mayer K., Zeiller M., 1994, in Proceedings of 1st International Conference on Image Processing . IEEE Comput. Soc. Press, Austin, TX, USA, pp 208--212, @doi 10.1109/ICIP.1994.413305 , http://ieeexplore.ieee.org/document/413305/

    1994

  15. [23]

    R., Matzner C

    Krumholz M. R., Matzner C. D., 2009, @doi [The Astrophysical Journal] 10.1088/0004-637X/703/2/1352 , 703, 1352

    2009 doi

  16. [24]

    L., 2020, @doi [Journal of Quantitative Spectroscopy and Radiative Transfer] 10.1016/j.jqsrt.2020.106928 , 246

    Le T., Liu C., Yao B., Natraj V., Yung Y. L., 2020, @doi [Journal of Quantitative Spectroscopy and Radiative Transfer] 10.1016/j.jqsrt.2020.106928 , 246

    2020

  17. [25]

    Lecun Y., Bottou L., Bengio Y., Haffner P., 1998, @doi [Proceedings of the IEEE] 10.1109/5.726791 , 86, 2278

    1998 doi

  18. [26]

    Lu X., Wang Y., 2024, @doi [Proceedings of the Combustion Institute] 10.1016/j.proci.2024.105282 , 40, 105282

    2024

  19. [27]

    Mishra S., Molinaro R., 2021, @doi [Journal of Quantitative Spectroscopy and Radiative Transfer] 10.1016/j.jqsrt.2021.107705 , 270, 107705

    2021

  20. [28]

    M., Sim S

    Noebauer U. M., Sim S. A., 2019, @doi [Living Reviews in Computational Astrophysics] https://doi.org/10.1007/s41115-019-0004-9 , 5, 1

    2019 doi

  21. [29]

    V., Schafer R

    Oppenheim A. V., Schafer R. W., eds, 2010, Discrete-time signal processing, 3. ed edn. Prentice Hall signal processing series, Pearson, Upper Saddle River Munich

    2010

  22. [30]

    Paszke A., et al., 2019, @doi [Advances in neural information processing systems] 10.48550/arXiv.1912.01703 , 32

    2019 doi arXiv

  23. [31]

    Springer Berlin Heidelberg, Berlin, Heidelberg, pp 53--67, @doi 10.1007/978-3-642-35289-8_5 , https://doi.org/10.1007/978-3-642-35289-8_5

    Prechelt L., 2012, Early Stopping --- But When?. Springer Berlin Heidelberg, Berlin, Heidelberg, pp 53--67, @doi 10.1007/978-3-642-35289-8_5 , https://doi.org/10.1007/978-3-642-35289-8_5

    2012 doi

  24. [32]

    P., Mour \ a o A

    Rino-Silvestre J., Gonz \'a lez-Gait \'a n S., Stalevski M., Smole M., Guilherme-Garcia P., Carvalho J. P., Mour \ a o A. M., 2023, @doi [Neural Computing and Applications] 10.1007/s00521-022-08071-x , 35, 7719

    2023 doi

  25. [33]

    Schmidt V., 1992, @doi [Reports on Progress in Physics] 10.1088/0034-4885/55/9/003 , 55, 1483

    1992 doi

  26. [34]

    L., van der Tak F

    Schöier F. L., van der Tak F. F. S., van Dishoeck E. F., Black J. H., 2005, @doi [Astronomy and Astrophysics] 10.1051/0004-6361:20041729 , 432, 369

    2005 doi

  27. [35]

    S., Cochrane R

    Sethuram S. S., Cochrane R. K., Hayward C. C., Acquaviva V., Villaescusa-Navarro F., Popping G., Wise J. H., 2023, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stad2524 , 526, 4520

    2023 doi

  28. [36]

    Simonyan K., 2014, @doi [arXiv preprint arXiv:1409.1556] 10.48550/arXiv.1409.1556

    2014 doi arXiv

  29. [37]

    S., 2017, Proceedings of the IEEE, 105, 2295

    Sze V., hsin Chen Y., Yang T.-J., Emer J. S., 2017, Proceedings of the IEEE, 105, 2295

    2017

  30. [38]

    Teyssier R., Commer c on B., 2019, @doi [Frontiers in Astronomy and Space Sciences] 10.3389/fspas.2019.00051 , 6, 51

    2019

  31. [39]

    N., Preisig H

    Thombre M. N., Preisig H. A., Addis M. B., 2015, in Gernaey K. V., Huusom J. K., Gani R., eds, Computer Aided Chemical Engineering, Vol. 37, 12th International Symposium on Process Systems Engineering and 25th European Symposium on Computer Aided Process Engineering. Elsevier,...

    2015 doi

  32. [40]

    Vandenbroucke B., Wood K., 2018, @doi [Astronomy and Computing] 10.1016/j.ascom.2018.02.005 , 23, 40

    2018 doi

  33. [41]

    Verstocken S., Van De Putte D., Camps P., Baes M., 2017, @doi [Astronomy and Computing] 10.1016/j.ascom.2017.05.003 , 20, 16

    2017 doi

  34. [42]

    Wang Z., Bovik A., Sheikh H., Simoncelli E., 2004, @doi [IEEE Transactions on Image Processing] 10.1109/TIP.2003.819861 , 13, 600

    2004

  35. [43]

    Wang C., Duan Q., Gong W., Ye A., Di Z., Miao C., 2014, @doi [Environmental Modelling & Software] 10.1016/j.envsoft.2014.05.026 , 60, 167

    2014 doi

  36. [44]

    Woitke P., Krueger D., Sedlmayr E., 1996, , https://ui.adsabs.harvard.edu/abs/1996A&A...311..927W 311, 927

    1996

  37. [45]

    E., Chané E., Keppens R., 2018, @doi [The Astrophysical Journal Supplement Series] 10.3847/1538-4365/aaa6c8 , p

    Xia C., Teunissen J., Mellah I. E., Chané E., Keppens R., 2018, @doi [The Astrophysical Journal Supplement Series] 10.3847/1538-4365/aaa6c8 , p. 30

    2018 doi

  38. [46]

    IEEE, Kunming, China, pp 1--3, @doi 10.1109/ICSICT49897.2020.9278168 , https://ieeexplore.ieee.org/document/9278168/

    Zhang Y., Jing M., Fan Y., Zeng X., 2020, in 2020 IEEE 15th International Conference on Solid - State & Integrated Circuit Technology ( ICSICT ). IEEE, Kunming, China, pp 1--3, @doi 10.1109/ICSICT49897.2020.9278168 , https://ieeexplore.ieee.org/document/9278168/

    2020

  39. [47]

    write newline

    " write newline "" before.all 'output.state := FUNCTION fin.entry write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence 'output.stat...

Pith tools

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