Pith.

Astrophysics of Galaxies · arXiv:2611.16286 v1 · submitted 12 Nov 2026 · 79 references

Machine-Learning Reconstruction of Intergalactic Medium Column-Density Maps from Observed Lyα Emission Cubes

Xihan Deng, Christopher Martin, Aaron Smith, Sultan Hassan, Drew M. Miles, Nikolaus Z. Prusinski, Vincent Picouet, Mateusz Matuszewski

show 2 more authors

Zeren Lin, Jake S. Bennett

The paper says

“Understanding the large-scale structure of the intergalactic medium (IGM) is essential for studying galaxy evolution, particularly at z ∼ 2–6, when cold filaments in the cosmic web are considered a primary mechanism for fueling galaxy growth.”

more of the abstract

At these redshifts, Lyα emission shifts into the optical regime and serves as the primary tracer of diffuse IGM for ground-based integral field units (IFUs). However, detections remain limited by low signal-to-noise ratios (S/N) and the complexities of resonant scattering. We present a proof-of-concept machine-learning framework to reconstruct IGM gas column-density maps from noisy Lyα IFU emission cubes. Using Keck Cosmic Web Imager (KCWI) observations of z ∼ 2.3 protoclusters as a case study, we train a custom convolutional neural network (CNN), the Column Density Estimator (CDE), on a simulated dataset combining IllustrisTNG100 with simplified radiative-transfer calculations and an approximate treatment of KCWI noise. The CDE successfully reconstructs broad filamentary structures in the simulated test set under these adopted assumptions. Applied to observed KCWI data, the CDE predicts gas column-density maps containing filamentary structures extending over ≳ 40′′ (∼ 1 comoving Mpc), consistent with an IGM interpretation. The observed reconstructions remain conditional on the adopted models. These results illustrate the potential of the CDE framework for connecting diffuse Lyα observations with simulations of the IGM.

from the abstract

Pith says

Machine review · deepseek-v4-flash · 13 Nov 2026 · what was read was not recorded · not checked by a person

It can be wrong. Challenge this review or reply to it.

Holds up, with conditions.

4 major objections · 10 minor

How often objections from reviews like this one are right: not yet measured

Are you an author? What you can do.

Not checked: figures, code, what the 79 references say.

The paper says, at §B.4.2, Fig. 15, Fig. 16

The review did not quote this passage.

A classifier read this sentence of the paper as acknowledging the concern (probability 0.87):

“However, such a local correction cannot repro- duce the nonlocal spatial and spectral redistribution of photons and has not been independently validated using pointing centers beyond the five used for its calibration.”

Pith says: Objection 1 of 4 · major

The empirical resonant-scattering correction is calibrated on only 30 matched intrinsic/full-RS cube pairs from five pointing centers (indices 3, 8, 13, 18, 23). The correction is assessed only on subcubes from the same five pointing centers used for calibration (Fig. 16).

more of objection 1

As the authors state, 'such a local correction cannot reproduce the nonlocal spatial and spectral redistribution of photons' (§6). The paper acknowledges this limitation but does not quantify the potential impact on the final column-density predictions. This is the most load-bearing concern because the entire training set depends on this correction; if the true scattering behavior in the observed fields differs from the calibration sample, the corrected intrinsic cubes—and hence the entire training set—will be systematically offset. The CDE would then learn a mapping from biased inputs to column density, and the observed reconstructions could be artifacts of that bias. The authors should either (a) provide a quantitative estimate of how errors in the correction propagate to the final column-density maps, or (b) demonstrate that the correction is robust by testing on independent pointing centers not used in the calibration.

The paper says, at §3.1, §6

The review did not quote this passage.

A classifier read this sentence of the paper as acknowledging the concern (probability 0.96):

“However, these measures cannot completely remove correlations between subcubes derived from the same pointing center, which may lead to optimistic es- timates of CDE performance.”

Pith says: Objection 2 of 4 · major

The training, validation, and test sets share the same 25 pointing centers. As the authors note (§6), 'these measures cannot completely remove correlations between subcubes derived from the same pointing center, which may lead to optimistic estimates of CDE performance.'

more of objection 2

This is a significant concern because the test set is not independent of the training set at the pointing-center level. The paper should quantify the degree of correlation (e.g., by reporting performance on a truly independent set of pointing centers, or by showing that performance degrades when pointing centers are held out). Without this, the reported test-set metrics (median image SSIM 0.635) may be inflated.

The paper says, at §4.5, Fig. 11

The review did not quote this passage.

A classifier read this sentence of the paper as acknowledging the concern (probability 0.88):

“Because the comparison uses selected subcubes from one F ABLE volume, it provides only a limited as- sessment of transferability.”

Pith says: Objection 3 of 4 · major

The cross-simulation stress test with FABLE is a valuable addition, but it is limited: only one FABLE volume is used, and subcubes are selected to match TNG column-density ranges. The noisy predictions yield median image SSIM 0.477, substantially lower than on the TNG test set (0.635).

more of objection 3

The authors acknowledge that 'this test does not isolate their individual effects' and 'provides only a limited assessment of transferability.' However, the paper does not discuss whether the FABLE results are consistent with the null hypothesis that the CDE fails to generalize. A more rigorous transferability test would involve applying the CDE to FABLE subcubes without the column-density selection and reporting the fraction of false-positive filament detections.

The paper says, at §5.1, Fig. 12

The review did not quote this passage.

Pith says: Objection 4 of 4 · major

The observed empty controls are a good check, but they cover only two wavelength ranges (3870–3889 Å and 4260–4279 Å) that are far from the Lyα emission lines. The paper should also test the CDE on observed subcubes at wavelengths where Lyα emission is expected but no signal is detected (e.g., the 'empty' regions within the same protocluster fields but at different redshifts). This would provide a stronger test of false-positive filament detection.

10 minor objections

The paper says, at §2.1

Pith says: minor objection 1 of 10

The definition of LU (line unit) is given as 1 LU = 1 photon s^{-1} cm^{-2} sr^{-1}, with a fixed conversion to erg units. However, the conversion factor depends on the wavelength of Lyα. The paper should state the assumed wavelength for this conversion.

The paper says, at §3.1, Fig. 4

Pith says: minor objection 2 of 10

The flowchart in Fig. 4 is helpful but somewhat cluttered. Consider using a larger font or splitting into two figures for clarity.

show 8 more

The paper says, at §3.2

Pith says: minor objection 3 of 10

The bagging ensemble uses 10 base models. The paper should justify this choice (e.g., by showing that performance saturates beyond 10 models).

The paper says, at §4.2, Fig. 8

Pith says: minor objection 4 of 10

The median image SSIM of 0.635 is reported with asymmetric error bars (+0.072/-0.088). The paper should clarify whether these are interquartile ranges or standard deviations.

The paper says, at §5.2, Fig. 13

Pith says: minor objection 5 of 10

The Monte Carlo uncertainty estimation uses only Gaussian noise perturbations. The paper should note that this does not capture systematic uncertainties from the empirical RS correction or the simulation assumptions.

The paper says, at §5.3, Fig. 14

Pith says: minor objection 6 of 10

The SB–δ relation is fitted with a quadratic. The paper should report the goodness of fit (e.g., χ² or R²) and the uncertainty on the fitted parameters.

The paper says, at §6

Pith says: minor objection 7 of 10

The discussion mentions that the CDE framework is 'not limited to Lyα emission or to studies of the IGM.' This is a strong claim; consider tempering it or providing a specific example.

The paper says, at Appendix B.4.2

Pith says: minor objection 8 of 10

The empirical correction is applied over a surface-brightness range of 10^{-21}–10^{-17} erg s^{-1} cm^{-2} arcsec^{-2} Å^{-1}. The paper should justify why this range is chosen and what fraction of voxels fall outside it.

The paper says, at Appendix B.5.1

Pith says: minor objection 9 of 10

The sky level is approximated as constant across each 30 Å window. The paper should discuss the impact of this approximation on the noise model.

The paper says, at Throughout

Pith says: minor objection 10 of 10

The paper uses 'CDE' to refer to both the neural network and the full framework. This is occasionally confusing; consider using 'CDE framework' consistently for the latter.

What would settle it

Generate full-RS cubes for at least two additional pointing centers not used in the calibration (e.g., indices 1 and 15). Extract 20 matched subcubes from these new full-RS cubes and their intrinsic counterparts. Compare the CDE predictions trained on (a) the standard corrected intrinsic cubes and (b) the full-RS cubes directly. If the median image SSIM on these new subcubes drops by more than 0.1 when using the standard correction relative to the full-RS-trained model, the correction is insufficient for independent environments and the observed reconstructions are unreliable.

from the skeptic's reading

How the review went

3 readings by deepseek-v4-flash · 13 Nov 2026

  1. Reader holds up, with conditions confidence moderate
  2. Skeptic agrees 14 s · 44,755 tokens read, 672 written
  3. Referee major revision: 4 major objections, 10 minor 16 s · 45,114 tokens read, 1,757 written

Not run on this paper: the editor, the simulated authors' defence.

Pith did not check this review's objections against the paper.

The review in full

The strongest objection

The paper's central claim—that the CDE reconstructs physically meaningful IGM filament maps from observed KCWI data—rests on the fidelity of the simulated training inputs. The most fragile link in that chain is the voxel-level empirical resonant-scattering correction (Section B.4.2). It is calibrated on only 30 matched intrinsic/full-RS cube pairs from five pointing centers (indices 3, 8, 13, 18, 23) and applied as a per-voxel multiplicative factor. As the authors state, 'such a local correction cannot reproduce the nonlocal spatial and spectral redistribution of photons' (Section 6). The correction is assessed only on subcubes from the same five pointing centers used for calibration (Figure 16), so its generalization to other environments is untested. If the true scattering behavior in the observed fields differs from the calibration sample, the corrected intrinsic cubes—and hence the entire training set—will be systematically offset from the true emission. The CDE would then learn a mapping from biased inputs to column density, and the observed reconstructions could be artifacts of that bias. The paper acknowledges this limitation but does not quantify the potential impact on the final column-density predictions. This is the single most load-bearing concern because it directly undermines the physical interpretability of the observed maps, which is the headline result.

The reader's weakest_assumption correctly identifies the empirical resonant-scattering correction as the most fragile link. I agree that this is the load-bearing concern. The paper states the limitation clearly, but the central claim—that the observed filament maps are physically meaningful—depends on the correction being adequate. The concrete test I propose would directly assess whether the correction generalizes to independent pointing centers. If it fails, the observed reconstructions could be artifacts of the correction bias. Therefore, the verdict remains CONDITIONAL, as the reader recommended, because the concern is unresolved but not fatal. The paper is a careful proof-of-concept, and the authors are transparent about the limitation. No change to the verdict is needed; the condition is simply that the correction must be validated on independent data before the observed maps can be treated as physical measurements.

The referee's report

If the central claim holds—that a simulation-trained CNN can recover physically meaningful IGM filament maps from low-signal Lyα IFU data—this would be a valuable proof of concept for connecting diffuse Lyα observations with simulations. The paper is transparent about limitations and provides a modular framework that can be retrained with improved simulations. The code and trained models are released. However, the significance is tempered by the acknowledged conditional nature of the results and the limited validation against independent simulations or observations.

Pith's reading of the paper (a machine's summary, not the authors')

Galaxies at redshift 2 to 3 sit inside a web of thin gas filaments that feed them. That gas glows faintly in a specific color, Lyα, but the glow is so weak that ordinary images show mostly noise. This paper builds a machine-learning model, the Column Density Estimator, that tries to undo the noise and estimate how much gas lies along each line of sight. The model was trained on a large set of fake observations: the authors took gas from the IllustrisTNG100 simulation, computed the expected Lyα glow with a simplified radiative-transfer code, added noise that mimics the Keck Cosmic Web Imager, and then asked the network to recover the known gas columns. On held-out simulated data the network recovered the broad filaments, though it smoothed over dense knots. When applied to real KCWI observations of two protoclusters, the model produced long filament-like structures spanning about one million light-years, and two sky regions with no detected emission produced no filaments. The authors are careful: the real-sky results are only as trustworthy as the simulation and noise model used for training, and the test data partly overlap with the training regions in the simulation, which can flatter the performance scores.

Core claim. The CDE successfully reconstructs broad filamentary structures in the simulated test set under the adopted assumptions, and applied to observed KCWI data it predicts gas column-density maps containing filamentary structures extending over about 40 arcseconds, roughly 1 comoving megaparsec, consistent with an IGM interpretation. If correct, this means a simulation-trained CNN can recover physically meaningful IGM filament maps from low-signal Lyα IFU cubes, conditional on the TNG100 and simplified radiative-transfer model.

The premise it rests on. The voxel-level empirical resonant-scattering correction, calibrated on only 30 matched intrinsic/full-RS cube pairs from five pointing centers, is applied to all production cubes as a proxy for full radiative transfer. Because it does not model nonlocal spatial and spectral redistribution of Lyα photons and has not been validated on independent pointing centers, any error in this correction propagates into the training inputs and hence into the observed reconstructions. If the true scattering behavior differs from this correction, the predicted filament maps could be artifacts. The paper states this limitation in Section B.4.2 and in the discussion.

Check this review

Pith stored no record of this review's passes and took no fingerprint of it. Reviews made by code from 13 Nov 2026 on carry both.

How often reviews like this one are right

Reviews like this one: written by deepseek-v4-flash, where what was read was not recorded.

How often objections from reviews like this one are right: not yet measured

Every value above, in the review's public record (JSON)

Reviews by people

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review

How Pith checks papers · The full audit of this paper

The paper

Abstract

Understanding the large-scale structure of the intergalactic medium (IGM) is essential for studying galaxy evolution, particularly at z ∼ 2–6, when cold filaments in the cosmic web are considered a primary mechanism for fueling galaxy growth. At these redshifts, Lyα emission shifts into the optical regime and serves as the primary tracer of diffuse IGM for ground-based integral field units (IFUs). However, detections remain limited by low signal-to-noise ratios (S/N) and the complexities of resonant scattering. We present a proof-of-concept machine-learning framework to reconstruct IGM gas column-density maps from noisy Lyα IFU emission cubes. Using Keck Cosmic Web Imager (KCWI) observations of z ∼ 2.3 protoclusters as a case study, we train a custom convolutional neural network (CNN), the Column Density Estimator (CDE), on a simulated dataset combining IllustrisTNG100 with simplified radiative-transfer calculations and an approximate treatment of KCWI noise. The CDE successfully reconstructs broad filamentary structures in the simulated test set under these adopted assumptions. Applied to observed KCWI data, the CDE predicts gas column-density maps containing filamentary structures extending over ≳ 40′′ (∼ 1 comoving Mpc), consistent with an IGM interpretation. The observed reconstructions remain conditional on the adopted models. These results illustrate the potential of the CDE framework for connecting diffuse Lyα observations with simulations of the IGM.

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.

abstract compiled · 2026-11-13

The authors' main claim, as the classifier picked it out 0.88

“These results illustrate the potential of the CDE framework for connecting diffuse Ly$\alpha$ observations with simulations of the IGM.”

How the main claim is stated: hedged possibility 0.71

What the abstract itself states, as probabilities: claims first 0.05 · states limitation 0.98 · human participants 0.02 · controlled comparison 0.05 · randomized assignment 0.01 · states quantitative result 0.95 · compares with prior results 0.08 · mentions code or data availability 0.03

The abstract, sentence by sentence

This table labels the abstract only (arXiv listing abstract v1). The number is the probability the classifier gave the label.

SentenceRoleThe authors' own workStates the main conclusionStrength
Understanding the large-scale structure of the intergalactic medium (IGM) is essential for studying galaxy evolution, particularly at $z \sim 2$--$6$, when cold filaments in the cosmic web are considered a primary mechanism for fueling galaxy growth. background 1.00 0.04 0.02 asserted 0.98
At these redshifts, Ly$\alpha$ emission shifts into the optical regime and serves as the primary tracer of diffuse IGM for ground-based integral field units (IFUs). background 1.00 0.05 0.02 asserted 1.00
However, detections remain limited by low signal-to-noise ratios (S/N) and the complexities of resonant scattering. background 0.54 0.07 0.02 asserted 1.00
We present a proof-of-concept machine-learning framework to reconstruct IGM gas column-density maps from noisy Ly$\alpha$ IFU emission cubes. contribution 1.00 0.95 0.11 asserted 1.00
Using Keck Cosmic Web Imager (KCWI) observations of $z \sim 2.3$ protoclusters as a case study, we train a custom convolutional neural network (CNN), the Column Density Estimator (CDE), on a simulated dataset combining IllustrisTNG100 with simplified radiative-transfer calculations and an approximate treatment of KCWI noise. method 1.00 0.96 0.05 asserted 0.99
The CDE successfully reconstructs broad filamentary structures in the simulated test set under these adopted assumptions. result 1.00 0.95 0.18 asserted 1.00
Applied to observed KCWI data, the CDE predicts gas column-density maps containing filamentary structures extending over $\gtrsim 40''$ ($\sim 1$ comoving Mpc), consistent with an IGM interpretation. result 0.99 0.96 0.42 hedged possibility 0.50
The observed reconstructions remain conditional on the adopted models. limitation 1.00 0.66 0.08 asserted 0.98
These results illustrate the potential of the CDE framework for connecting diffuse Ly$\alpha$ observations with simulations of the IGM. interpretation 0.99 0.59 0.56 hedged possibility 0.71

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from analysis of existing data 0.50

What was examined, and what was not

Sources read

  • arXiv listing abstract v1

Not examined

  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers

Levels run

  • P abstract

Not run

  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

How the labels were made

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-11-13: 1 request, 7,125 input tokens.

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON

Figures (20)
Figure 1 of the paper
Figure 1. Overview of the CDE framework. Gas-cell and star-particle snapshot data from IllustrisTNG are used to generate both two-dimensional column-density maps (serving as training ground truth) and simulated CDE inputs with two spatial dimensions and one spectral dimension (serving as training input). Each input–ground-truth pair forms a training example, and the collection of such examples is used to train and test the CDE. Once trained and validated, the CDE is applied to the observed CDE inputs to predict gas column-density maps. The digital dimensions and the corresponding physical…
Figure 2 of the paper
Figure 2. CDE architecture. The model uses a shallow U-Net configuration combining residual blocks with downsampling and upsampling pathways. Features from two earlier stages are concatenated with the upsampled features on a 40 × 40 pixel grid. The rightmost 80 × 80 skip branch is reduced to 40 × 40 using 2 × 2 max pooling before concatenation. Numbers (e.g., ×2) denote the count of consecutive blocks in each section. 3. DATA PREPARATION AND TRAINING 3.1. Data Preparation Overview To train the CDE, we generate a large set of simulated CDE inputs that resemble the observed data. It is…
Figure 3 of the paper
Figure 3. CDE-noiseless. Representative test-set example and performance statistics across all 1200 noiseless test-set regions. Top row: (a) Lyα surface-brightness map integrated over the 20 spectral channels of the noiseless CDE input; (b) CDE-noiseless-predicted gas column-density map; and (c) unmasked ground-truth gas column-density map. Red filament spines are obtained using SCMS. Bottom row: (d) linear RMSE; (e) RMSE in dex; and (f) image SSIM. In all three histograms, the vertical dashed line marks the value for the illustrated example. CDE-noiseless reconstructs filamentary morphology…
Figure 4 of the paper
Figure 4. Flowchart for data generation. The left column illustrates the reduction of observed KCWI data into CDE inputs, the middle column shows the generation of simulated CDE inputs, and the right column depicts the generation of column-density maps used as ground truth. The array dimensions of the data at each step are also indicated. The example images are included only to illustrate the workflow and do not represent the exact data products used in the final analysis
Figure 5 of the paper
Figure 5. Prediction error as a function of the loss
Figure 7 of the paper
Figure 7. Simple Structure Test: tests of adaptive smoothing and the CDE on toy filament subcubes. Rows show (top to bottom): the noiseless surface-brightness map after PSF and LSF convolution, integrated over the central 20 spectral channels; the noise-injected, adaptively smoothed surface-brightness map integrated over the same channels; and the CDE-predicted gas column-density map. Columns represent three toy cases: Filament 1, a narrow filament (1000 LU brightness, 20 pixel width); Filament 2, the same structure with doubled surface brightness; and Filament 3, the same structure with…
Figure 8 of the paper
Figure 8. CDE on Test Set: representative test-set example for the CDE with noisy, masked inputs and error statistics over the full test set. Top row: (a) the input Lyα surface-brightness map; (b) the CDE-predicted gas column-density map; (c) the masked ground-truth column-density map. Red filament spines are calculated using the SCMS algorithm. Bottom row: (d) the RMSE histogram for the full test set; (e) the RMSE-in-dex histogram for the full test set; (f) the image SSIM histogram for the full test set. In all three histograms, the vertical red dashed line marks the value for the example…
Figure 9 of the paper
Figure 9. Empirical Scaling Model on Test Set: representative test-set example for the empirical scaling model with noisy, masked inputs and error statistics over the full test set. Top row: (a) the input Lyα surface-brightness map; (b) the gas column-density map predicted by the empirical model; (c) the masked ground-truth column-density map. Bottom row: (d) the RMSE histogram for the full test set; (e) the RMSE-in-dex histogram for the full test set; (f) the image SSIM histogram for the full test set. In all three histograms, the vertical red dashed line marks the value for the example…
Figure 10 of the paper
Figure 10. CDE-DM on Test Set: representative test-set example for the CDE trained to predict dark matter with noisy, masked inputs and error statistics over the full test set. Top row: (a) the input Lyα surface-brightness map; (b) the CDE-predicted DM column-density map; (c) the masked ground-truth DM column-density map. The simulated CDE input shown here is identical to that used for the gas reconstruction in
Figure 11 of the paper
Figure 11. Cross-simulation stress test with FABLE.
Figure 12 of the paper
Figure 12. CDE predictions and filament identification in observed KCWI data. Top row:
Figure 13 of the paper
Figure 13. MC Uncertainty: fractional uncertainty of CDE predictions estimated from paired Gaussian noise per￾turbations of observed KCWI subcubes. Points show the me￾dian pixel-wise fractional uncertainty within each region, and error bars show the corresponding interquartile range. The horizontal dashed line marks the median of the ten region medians (∼ 7.4%), and the shaded band indicates their in￾terquartile range. The predicted column densities show only modest sensitivity to the applied Gaussian noise perturbations. such perturbation pairs, yielding a total of 20 realiza￾tions per…
Figure 14 of the paper
Figure 14. Gas overdensity normalization. Top: Spa￾tially averaged Lyα surface brightness integrated over the retained 20 spectral channels versus CDE-predicted 1 + δgas for the simulated test set (blue) and observed regions (or￾ange). The black dashed curve shows the quadratic relation fitted in logarithmic coordinates exclusively to the simulated data. Bottom: SB-normalized overdensities δ norm gas for the ten KCWI regions (R00–R09). The horizontal dashed line marks δ norm gas = 0, corresponding to agreement with the sim￾ulation-based normalization at the measured surface bright￾ness. The…
Figure 15 of the paper
Figure 15. Empirical resonant-scattering correction.
Figure 16 of the paper
Figure 16. Empirical resonant-scattering correction comparison.
Figure 17 of the paper
Figure 17. Noise-statistics comparison. Comparison of noise statistics between the low-surface-brightness observed subcube from Field A and the noise-injected simulated empty subcube after variance correction, the first smoothing round, and boundary trimming. Columns show, from left to right, the voxel-value distributions of (1) S/N, (2) surface brightness, and (3) corrected variance. The smoothing uses a spatial Gaussian kernel with FWHM = 5 pixels followed by a spectral boxcar kernel with a width of four channels. The simulated and observed distributions show broadly similar noise…
Figure 18 of the paper
Figure 18. Spatial ACF comparison. Both panels use spatial ACFs evaluated at zero spectral lag by combining the within-channel spatial correlations across all 20 retained channels and normalizing to unity at zero spatial lag. (a) 2D spatial ACF of the noise-injected simulated empty subcube before smoothing, showing the characteristic plus-sign pattern induced by coadding exposures at PA= 0◦ and PA= 90◦ . (b) Central 1D cuts through the spatial ACFs after the first smoothing round and boundary trimming, comparing observed and simulated subcubes. Each cut holds the lag in the orthogonal spatial…
Figure 19 of the paper
Figure 19. More representative test-set examples for the CDE
Figure 20 of the paper
Figure 20. Atlas of CDE reconstructions and filament detections for all ten KCWI regions (left column: regions 0–4; right column: regions 5–9). Each panel shows the adaptively smoothed Lyα surface-brightness map integrated over the retained 20 spectral channels and the corresponding CDE-predicted column density with SCMS filaments overlaid in red

Showing 19 of the paper's 20 figures. Figures are the authors', reproduced from the paper.

References (79)
  1. [1] Abadi, M., Agarwal, A., Barham, P., et al. 2015, TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems, https://www.tensorflow.org/ on Pith
  2. [2] Angulo, R. E., Springel, V., White, S. D. M., et al. 2012, MNRAS, 426, 2046, doi: 10.1111/j.1365-2966.2012.21830.x on Pithdoi:10.1111/j.1365-2966.2012.21830.x
  3. [3] Bacon, R., Accardo, M., Adjali, L., et al. 2010, in SPIE Conf. Ser., Vol. 7735, Ground-based and Airborne Instrumentation for Astronomy III, ed. I. S. McLean, S. K. Ramsay, & H. Takami, 773508, doi: 10.1117/12.856027 on PithPith's reviewdoi:10.1117/12.856027
  4. [4] Bacon, R., Mary, D., Garel, T., et al. 2021, A&A, 647, A107, doi: 10.1051/0004-6361/202039887 on Pithdoi:10.1051/0004-6361/202039887
  5. [5] Bennett, J. S., Smith, A., Arrigoni-Battaia, F., et al. 2026, The Open Journal of Astrophysics, 9, 65188, doi: 10.33232/001c.165188 on Pithdoi:10.33232/001c.165188
  6. [6] Bertone, S., & Schaye, J. 2012, MNRAS, 419, 780, doi: 10.1111/j.1365-2966.2011.19742.x on Pithdoi:10.1111/j.1365-2966.2011.19742.x
  7. [7] Breiman, L. 1996, Machine Learning, 24, 123, doi: 10.1007/BF00058655 on Pithdoi:10.1007/bf00058655
  8. [8] Byrohl, C., & Nelson, D. 2023, MNRAS, 523, 5248, doi: 10.1093/mnras/stad1779 on Pithdoi:10.1093/mnras/stad1779
  9. [9] Byrohl, C., & Nelson, D. 2025, THOR: a GPU-accelerated and MPI-parallel radiative transfer code, doi: 10.48550/arXiv.2507.11603 on PithPith's reviewarXivdoi:10.48550/arxiv.2507.11603
  10. [10] Byrohl, C., Nelson, D., Behrens, C., et al. 2021, MNRAS, 506, 5129, doi: 10.1093/mnras/stab1958 on Pithdoi:10.1093/mnras/stab1958
  11. [11] Hennawi, J. F., & Madau, P. 2014, Nature, 506, 63, doi: 10.1038/nature12898 on Pithdoi:10.1038/nature12898
  12. [12] Dubois, Y. 2020, MNRAS, 497, 279, doi: 10.1093/mnras/staa1832 on Pithdoi:10.1093/mnras/staa1832
  13. [13] Cattaneo, A., Faber, S. M., Binney, J., et al. 2009, Nature, 460, 213, doi: 10.1038/nature08135 on Pithdoi:10.1038/nature08135
  14. [14] Caucci, S., Colombi, S., Pichon, C., et al. 2008, MNRAS, 386, 211, doi: 10.1111/j.1365-2966.2008.13016.x on Pithdoi:10.1111/j.1365-2966.2008.13016.x
  15. [15] Cautun, M., van de Weygaert, R., Jones, B. J. T., & Frenk, C. S. 2014, MNRAS, 441, 2923, doi: 10.1093/mnras/stu768 on Pithdoi:10.1093/mnras/stu768
  16. [16] Chabrier, G. 2003, PASP, 115, 763, doi: 10.1086/376392 on PithPith's reviewdoi:10.1086/376392
  17. [17] Wasserman, L. 2015, MNRAS, 454, 1140, doi: 10.1093/mnras/stv1996 on PithPith's reviewdoi:10.1093/mnras/stv1996
  18. [18] Connor, L., Ravi, V., Sharma, K., et al. 2025, Nature Astronomy, 9, 1226, doi: 10.1038/s41550-025-02566-y on Pithdoi:10.1038/s41550-025-02566-y
  19. [19] Cucciati, O., Lemaux, B. C., Zamorani, G., et al. 2018, A&A, 619, A49, doi: 10.1051/0004-6361/201833655 on Pithdoi:10.1051/0004-6361/201833655
  20. [20] Darvish, B., Scoville, N. Z., Martin, C., et al. 2020, ApJ, 892, 8, doi: 10.3847/1538-4357/ab75c3 Dav´ e, R., Angl´ es-Alc´ azar, D., Narayanan, D., et al. 2019, MNRAS, 486, 2827, doi: 10.1093/mnras/stz937 on Pithdoi:10.3847/1538-4357/ab75c3
show 59 more
  1. [21] Dekel, A., & Birnboim, Y. 2006, MNRAS, 368, 2, doi: 10.1111/j.1365-2966.2006.10145.x on Pithdoi:10.1111/j.1365-2966.2006.10145.x
  2. [22] Dekel, A., Birnboim, Y., Engel, G., et al. 2009, Nature, 457, 451, doi: 10.1038/nature07648 on PithPith's reviewdoi:10.1038/nature07648
  3. [23] Deng, X. 2026, CDE: Ly-alpha data preparation, column-density estimation, and validation code, 0.8.0 Zenodo, doi: 10.5281/zenodo.23328516 on Pithdoi:10.5281/zenodo.23328516
  4. [24] Dijkstra, M. 2014, PASA, 31, e040, doi: 10.1017/pasa.2014.33 Faucher-Gigu` ere, C.-A., Kereˇ s, D., Dijkstra, M., Hernquist, L., & Zaldarriaga, M. 2010, ApJ, 725, 633, doi: 10.1088/0004-637X/725/1/633 on Pithdoi:10.1017/pasa.2014.33
  5. [25] Gallego, S. G., Cantalupo, S., Lilly, S., et al. 2018, MNRAS, 475, 3854, doi: 10.1093/mnras/sty037 on Pithdoi:10.1093/mnras/sty037
  6. [26] Harris, C. R., Millman, K. J., van der Walt, S. J., et al. 2020, Nature, 585, 357, doi: 10.1038/s41586-020-2649-2 on Pithdoi:10.1038/s41586-020-2649-2
  7. [27] He, K., Zhang, X., Ren, S., & Sun, J. 2016, arXiv e-prints, arXiv:1603.05027, doi: 10.48550/arXiv.1603.05027 on PithPith's reviewarXivdoi:10.48550/arxiv.1603.05027
  8. [28] Henden, N. A., Puchwein, E., Shen, S., & Sijacki, D. 2018, MNRAS, 479, 5385, doi: 10.1093/mnras/sty1780 on PithPith's reviewdoi:10.1093/mnras/sty1780
  9. [29] Horowitz, B., Lee, K.-G., White, M., Krolewski, A., & Ata, M. 2019, ApJ, 887, 61, doi: 10.3847/1538-4357/ab4d4c on Pithdoi:10.3847/1538-4357/ab4d4c
  10. [30] Hunter, J. D. 2007, Computing in Science & Engineering, 9, 90, doi: 10.1109/MCSE.2007.55 on Pithdoi:10.1109/mcse.2007.55
  11. [31] Ozbek, M. 2014, ApJ, 788, 49, doi: 10.1088/0004-637X/788/1/49 on Pithdoi:10.1088/0004-637x/788/1/49
  12. [32] Liepold, E. R., Ma, C.-P., & Walsh, J. L. 2023, ApJL, 945, L35, doi: 10.3847/2041-8213/acbbcf on Pithdoi:10.3847/2041-8213/acbbcf
  13. [33] Lilly, S. J., Le F` evre, O., Renzini, A., et al. 2007, ApJS, 172, 70, doi: 10.1086/516589 on Pithdoi:10.1086/516589
  14. [34] Liu, Y., Gao, L., Liao, S., et al. 2026, arXiv e-prints, arXiv:2601.22677, doi: 10.48550/arXiv.2601.22677 on PithPith's reviewarXivdoi:10.48550/arxiv.2601.22677
  15. [35] Macquart, J. P., Prochaska, J. X., McQuinn, M., et al. 2020, Nature, 581, 391, doi: 10.1038/s41586-020-2300-2 on Pithdoi:10.1038/s41586-020-2300-2
  16. [36] Madau, P., & Dickinson, M. 2014, ARA&A, 52, 415, doi: 10.1146/annurev-astro-081811-125615 on PithPith's reviewdoi:10.1146/annurev-astro-081811-125615
  17. [37] Marinacci, F., Vogelsberger, M., Pakmor, R., et al. 2018, MNRAS, 480, 5113, doi: 10.1093/mnras/sty2206 on PithPith's reviewdoi:10.1093/mnras/sty2206
  18. [38] Martin, D. C., Chang, D., Matuszewski, M., et al. 2014, ApJ, 786, 106, doi: 10.1088/0004-637X/786/2/106 on PithPith's reviewdoi:10.1088/0004-637x/786/2/106
  19. [39] Martin, D. C., Darvish, B., Lin, Z., et al. 2023, Nature Astronomy, 7, 1390, doi: 10.1038/s41550-023-02054-1 on Pithdoi:10.1038/s41550-023-02054-1
  20. [40] Martin, D. C., Matuszewski, M., Morrissey, P., et al. 2016, ApjL, 824, L5, doi: 10.3847/2041-8205/824/1/L5 Column Density Estimator35 on Pithdoi:10.3847/2041-8205/824/1/l5
  21. [41] Martin, D. C., O’Sullivan, D., Matuszewski, M., et al. 2019, Nature Astronomy, 3, 822, doi: 10.1038/s41550-019-0791-2 on Pithdoi:10.1038/s41550-019-0791-2
  22. [42] McClymont, W., Smith, A., & Tacchella, S. 2025, arXiv e-prints, arXiv:2510.13952, doi: 10.48550/arXiv.2510.13952 on Pithdoi:10.48550/arxiv.2510.13952
  23. [43] Michel-Dansac, L., Blaizot, J., Garel, T., et al. 2020, RASCAS: Resonant line transfer in AMR simulations,, ASCL, ascl:2002.002 http://ascl.net/2002.002 on Pith
  24. [44] Morrissey, P., Matuszewski, M., Martin, D. C., et al. 2018, ApJ, 864, 93, doi: 10.3847/1538-4357/aad597 on PithPith's reviewdoi:10.3847/1538-4357/aad597
  25. [45] Mortlock, A., Conselice, C. J., Bluck, A. F. L., et al. 2011, MNRAS, 413, 2845, doi: 10.1111/j.1365-2966.2011.18357.x on Pithdoi:10.1111/j.1365-2966.2011.18357.x
  26. [46] Naiman, J. P., Pillepich, A., Springel, V., et al. 2018, MNRAS, 477, 1206, doi: 10.1093/mnras/sty618 on PithPith's reviewdoi:10.1093/mnras/sty618
  27. [47] Rizzi, L. 2023, KCWI DRP: Keck Cosmic Web Imager Data Reduction Pipeline in Python,, ASCL, ascl:2301.019 on Pith
  28. [48] Nelson, D., Pillepich, A., Springel, V., et al. 2018, MNRAS, 475, 624, doi: 10.1093/mnras/stx3040 on PithPith's reviewdoi:10.1093/mnras/stx3040
  29. [49] Nelson, D., Springel, V., Pillepich, A., et al. 2019, Computational Astrophysics and Cosmology, 6, 2, doi: 10.1186/s40668-019-0028-x on Pithdoi:10.1186/s40668-019-0028-x
  30. [50] Nicastro, F., Kaastra, J., Krongold, Y., et al. 2018, Nature, 558, 406, doi: 10.1038/s41586-018-0204-1 on Pithdoi:10.1038/s41586-018-0204-1
  31. [51] Ouchi, M., Ono, Y., & Shibuya, T. 2020, ARA&A, 58, 617, doi: 10.1146/annurev-astro-032620-021859 on Pithdoi:10.1146/annurev-astro-032620-021859
  32. [52] Park, H., Song, H., Byrohl, C., et al. 2025, MNRAS, 544, 4246, doi: 10.1093/mnras/staf1971 on Pithdoi:10.1093/mnras/staf1971
  33. [53] Patat, F. 2003, A&A, 400, 1183, doi: 10.1051/0004-6361:20030030 on Pithdoi:10.1051/0004-6361:20030030
  34. [54] Petitjean, P. 2001, MNRAS, 326, 597, doi: 10.1046/j.1365-8711.2001.04595.x on Pithdoi:10.1046/j.1365-8711.2001.04595.x
  35. [55] Pillepich, A., Springel, V., Nelson, D., et al. 2018, MNRAS, 473, 4077, doi: 10.1093/mnras/stx2656 on PithPith's reviewdoi:10.1093/mnras/stx2656
  36. [56] Qezlou, M., Newman, A. B., Rudie, G. C., & Bird, S. 2022, ApJ, 930, 109, doi: 10.3847/1538-4357/ac6259 on Pithdoi:10.3847/1538-4357/ac6259
  37. [57] Rauch, M. 1998, ARA&A, 36, 267, doi: 10.1146/annurev.astro.36.1.267 on Pithdoi:10.1146/annurev.astro.36.1.267
  38. [58] Ronneberger, O., Fischer, P., & Brox, T. 2015, arXiv e-prints, arXiv:1505.04597, doi: 10.48550/arXiv.1505.04597 on PithPith's reviewarXivdoi:10.48550/arxiv.1505.04597
  39. [59] Rosdahl, J., & Blaizot, J. 2012, MNRAS, 423, 344, doi: 10.1111/j.1365-2966.2012.20883.x on Pithdoi:10.1111/j.1365-2966.2012.20883.x
  40. [60] Santhanam, G. K. 2017, MNRAS, 467, L110, doi: 10.1093/mnrasl/slx008 on Pithdoi:10.1093/mnrasl/slx008
  41. [61] Schaye, J., Chaikin, E., Schaller, M., et al. 2026, MNRAS, 548, stag375, doi: 10.1093/mnras/stag375 on Pithdoi:10.1093/mnras/stag375
  42. [62] Sembach, K. R., & Tonry, J. L. 1996, AJ, 112, 797, doi: 10.1086/118055 on Pithdoi:10.1086/118055
  43. [63] Shull, J. M., Smith, B. D., & Danforth, C. W. 2012, ApJ, 759, 23, doi: 10.1088/0004-637X/759/1/23 on Pithdoi:10.1088/0004-637x/759/1/23
  44. [64] Smith, A., Ma, X., Bromm, V., et al. 2019, MNRAS, 484, 39, doi: 10.1093/mnras/sty3483 on Pithdoi:10.1093/mnras/sty3483
  45. [65] Smith, A., Safranek-Shrader, C., Bromm, V., & Milosavljevi´ c, M. 2015, MNRAS, 449, 4336, doi: 10.1093/mnras/stv565 on Pithdoi:10.1093/mnras/stv565
  46. [66] Smith, A., Kannan, R., Tacchella, S., et al. 2022, MNRAS, 517, 1, doi: 10.1093/mnras/stac2641 on Pithdoi:10.1093/mnras/stac2641
  47. [67] Smith, M. J., & Geach, J. E. 2023, Royal Society Open Science, 10, 221454, doi: 10.1098/rsos.221454 on Pithdoi:10.1098/rsos.221454
  48. [68] Springel, V., Pakmor, R., Pillepich, A., et al. 2018, MNRAS, 475, 676, doi: 10.1093/mnras/stx3304 on Pithdoi:10.1093/mnras/stx3304
  49. [69] Stanway, E. R., & Eldridge, J. J. 2018, MNRAS, 479, 75, doi: 10.1093/mnras/sty1353 on Pithdoi:10.1093/mnras/sty1353
  50. [70] Tornotti, D., Fumagalli, M., Fossati, M., et al. 2025, Nature Astronomy, 9, 577, doi: 10.1038/s41550-024-02463-w on Pithdoi:10.1038/s41550-024-02463-w
  51. [71] Umehata, H., Fumagalli, M., Smail, I., et al. 2019, Science, 366, 97, doi: 10.1126/science.aaw5949 van der Wel, A., Franx, M., van Dokkum, P. G., et al. 2014, ApJ, 788, 28, doi: 10.1088/0004-637X/788/1/28 on Pithdoi:10.1126/science.aaw5949
  52. [72] Villaescusa-Navarro, F. 2018, Pylians: Python libraries for the analysis of numerical simulations,, ASCL, ascl:1811.008 http://ascl.net/1811.008 on Pith
  53. [73] Virtanen, P., Gommers, R., Oliphant, T. E., et al. 2020, Nature Methods, 17, 261, doi: 10.1038/s41592-019-0686-2 on Pithdoi:10.1038/s41592-019-0686-2
  54. [74] Vojtekova, A., Lieu, M., Valtchanov, I., et al. 2021, MNRAS, 503, 3204, doi: 10.1093/mnras/staa3567 on Pithdoi:10.1093/mnras/staa3567
  55. [75] Perreault-Levasseur, L. 2021, ApJ, 916, 42, doi: 10.3847/1538-4357/ac033a on Pithdoi:10.3847/1538-4357/ac033a
  56. [76] Wang, Z., Bovik, A., Sheikh, H., & Simoncelli, E. 2004, IEEE Transactions on Image Processing, 13, 600, doi: 10.1109/TIP.2003.819861 on Pithdoi:10.1109/tip.2003.819861
  57. [77] Weilbacher, P. M., Palsa, R., Streicher, O., et al. 2020, A&A, 641, A28, doi: 10.1051/0004-6361/202037855 on Pithdoi:10.1051/0004-6361/202037855
  58. [78] Yang, K. B., Wu, Q., & Wang, F. Y. 2022, ApJL, 940, L29, doi: 10.3847/2041-8213/aca145 on Pithdoi:10.3847/2041-8213/aca145
  59. [79] Zhao, H., Gallo, O., Frosio, I., & Kautz, J. 2015, arXiv e-prints, arXiv:1511.08861, doi: 10.48550/arXiv.1511.08861 on PithPith's reviewarXivdoi:10.48550/arxiv.1511.08861

Challenges and discussion

Disagree with the review, or know something it missed? Leave a comment, or challenge this review formally: a challenge becomes a permanent entry in this paper's record, and its decision is published here.

No comments yet.

Sign in to comment