pith. machine review for the scientific record. sign in

arxiv: 2604.23840 · v1 · submitted 2026-04-26 · 🌌 astro-ph.EP · astro-ph.IM

Recognition: unknown

Probabilistic Spectral Reconstruction of Trans-Neptunian Objects from Sparse Photometry: A Framework for Taxonomy, Survey Optimization, and Outlier Detection

Hsing Wen Lin , Larissa Markwardt , Kevin J. Napier , Fred C. Adams , Renu Malhotra , David W. Gerdes
Authors on Pith no claims yet

Pith reviewed 2026-05-08 05:02 UTC · model grok-4.3

classification 🌌 astro-ph.EP astro-ph.IM
keywords trans-Neptunian objectsspectral reconstructionprincipal component analysisphotometrynear-infraredtaxonomyBayesian inferencesurvey optimization
0
0 comments X

The pith

A low-dimensional principal component model reconstructs trans-Neptunian object spectra from photometry with 95 percent credible interval coverage.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents a probabilistic framework that uses principal component analysis trained on near-IR spectra to model the spectral variability of trans-Neptunian objects. It shows that only 4 to 5 components are needed to capture the main features for taxonomic classification, while 8 to 10 provide good reconstruction of full spectra from sparse photometry along with uncertainty estimates. A sympathetic reader would care because this allows turning abundant photometric data into compositional insights that are otherwise limited by scarce spectroscopy, and it identifies optimal filters for surveys like JWST. The work also shows how to flag unusual objects by seeing where photometry points to low-probability parts of the spectral manifold.

Core claim

Using a principal component representation trained on a sample of near-IR spectra, we model the spectral manifold of TNOs and perform Bayesian inference in this reduced space to reconstruct full spectra from photometry while propagating uncertainties. Leave-one-out cross-validation demonstrates that the dominant modes of spectral variability are low-dimensional: 4 to 5 principal components capture the structure relevant for taxonomic classification, while 8-10 components improve spectral reconstruction fidelity and uncertainty calibration. For most objects, the reconstructed spectra achieve empirical credible-interval coverage of 95 percent across wavelength. This suggests the diversity of近红

What carries the argument

Principal component analysis trained on near-IR spectra, enabling Bayesian inference in the reduced latent space to reconstruct spectra from photometry.

Load-bearing premise

The training sample of near-IR spectra is representative of the full diversity of TNO spectral shapes.

What would settle it

Full spectroscopy of an object whose photometry-based reconstruction lies outside the reported 95 percent credible intervals across wavelengths would contradict the coverage result.

Figures

Figures reproduced from arXiv: 2604.23840 by David W. Gerdes, Fred C. Adams, Hsing Wen Lin, Kevin J. Napier, Larissa Markwardt, Renu Malhotra.

Figure 1
Figure 1. Figure 1: The architecture and workflow of the TNO spectral reconstructor and classifier, divided into three operational phases. (Left) Preparing Phase: Observational seed spectra are compressed via Principal Component Analysis (PCA) to define the initial latent manifold, which is subsequently augmented using Kernel Density Estimation (KDE) to map a continuous probability distribution. (Middle) Training Phase: The a… view at source ↗
Figure 2
Figure 2. Figure 2: An illustration of synthetic training data (mock) and the observational seed spectra distributions in a 5-dimensional manifolds. 2.5. The Posterior: Reconstructed Spectra The final reconstruction step acts as the generative decoder of our pipeline, synthesizing the full spectra from the predicted latent variables. By applying the inverse PCA transform to the coefficients output by the AutoML regressor, we … view at source ↗
Figure 3
Figure 3. Figure 3: Fractional and cumulative explained variance of the observational seed spectra as a function of the number of principal components. The rapid decay of individual variance (solid blue line) demonstrates that the dominant modes of TNO spectral diversity are intrinsically low-dimensional, with the first five components capturing the vast majority of the variance. that the model does not merely memorize the lo… view at source ↗
Figure 4
Figure 4. Figure 4: Empirical coverage as a function of wavelength for nominal 95% posterior credible intervals under LOOCV. Coverage is computed as the fraction of held-out spectra whose true reflectance lies within the reconstructed interval at each wavelength. Coverage remains near nominal across most wavelengths, with localized decreases near the strong absorption features, particu￾larly around 3µm where no direct photome… view at source ↗
Figure 5
Figure 5. Figure 5: Leave-one-out cross-validation (LOOCV) reconstruction residuals (reconstructed minus reference re view at source ↗
Figure 6
Figure 6. Figure 6: Reconstructed spectra for four distinct TNO spectral types. The spectra of the two Neptune Trojans, 2010 TS191 and 2013 VX30, were withheld from the training seeds to test generalization. The two TNOs 174567 Varda and 469705 }Ka, ga, ra were reconstructed using a Leave-One-Out Cross-Validation (LOOCV) framework. The shaded regions indicate the 95% pos￾terior credible intervals, while the gray lines represe… view at source ↗
Figure 7
Figure 7. Figure 7: Confusion matrices for four selected 2-band photometry classification scenarios. predictive distribution, yielding the final reconstructed spectrum and realistic uncertainty bounds tailored to the planned observation. A critical question therefore arises for future survey design: which filters yield the highest information content? In other words, what are the minimal optimal filter sets required to distin… view at source ↗
Figure 8
Figure 8. Figure 8: Example of a water-type spectrum reconstruction using only [F090W, F360M] photometry. and becomes narrow at this location. Consequently, a targeted 2-band reconstruction can actually outperform a 4-band configuration at specific wavelengths if the latter omits a local filter, as the direct photometric constraint successfully collapses the local posterior variance. Thus, while two-band reconstruction lacks … view at source ↗
Figure 9
Figure 9. Figure 9: Confusion matrices for four selected 4-band photometry classification scenarios. The rightmost panel corresponds to the filter set of JWST GO #7248. H2O CO2 organic methanol Predicted label H2O CO2 organic methanol True label 1 0 0 0 0 0.98 0.085 0.023 0 0.015 0.88 0.044 0 0.00031 0.035 0.93 ['F090W', 'F115W', 'F150W', 'F360M', 'F410M', 'F460M'] H2O CO2 organic methanol Predicted label H2O CO2 organic meth… view at source ↗
Figure 10
Figure 10. Figure 10: Confusion matrices for four selected 6-band photometry classification scenarios. 4.1.3. Six-Filter Combinations (Three Exposures) Although one might expect that increasing the number of photometric bands to six would substantially refine the TNO taxonomy, we find the marginal gains to be minimal view at source ↗
Figure 11
Figure 11. Figure 11: Reconstructed spectra of the Neptune Trojan outliers 2006 RJ103 and 2011 SO277 derived from the 4-band filter configuration [F090W, F115W, F410M, F460M]. 3. Two-Band Constraints (Slope Selection): The addition of a second filter imposes a strong constraint on the allowed region of latent space. For example, a large color value for (F090W - F360M) immediately renders Water-type spectra (which are typically… view at source ↗
Figure 12
Figure 12. Figure 12: Reconstruction spectra of 2006 RJ103 and 2011 SO277 with the 6-band photometry filter configuration [F090W, F115W, F150W, F360M, F410M, F460M]. The extra filters pinch the predictive reconstruction uncertainties. 5.2. Why Some Bands Are More Informative Than Others? Our survey optimization analysis (Section 4.1) yielded a clear hierarchy of filter importance, with short-wavelength (e.g., F090W) and long-w… view at source ↗
Figure 13
Figure 13. Figure 13: The squared Principal Component Feature Loadings (L 2 k(λ)) of the TNO training spectra across the 0.7–5.0 µm range. We plot the squared loadings because L 2 k(λ) mathematically represents the spectral density of variance for each principal component. As detailed in Section 5.2, this variance is directly proportional to the reduction in Shannon entropy. Therefore, the color intensity serves as a map of in… view at source ↗
read the original abstract

Near-infrared (near-IR) spectroscopy provides critical constraints on the surface composition of trans-Neptunian objects (TNOs), but spectroscopic observations remain limited compared to broadband photometry. We develop a probabilistic latent-space framework to quantify how much spectral information is retained in sparse photometric measurements. Using a principal component representation trained on a sample of near-IR spectra, we model the spectral manifold of TNOs and perform Bayesian inference in this reduced space to reconstruct full spectra from photometry while propagating uncertainties. Leave-one-out cross-validation demonstrates that the dominant modes of spectral variability are low-dimensional: 4 to 5 principal components capture the structure relevant for taxonomic classification, while 8-10 components improve spectral reconstruction fidelity and uncertainty calibration. For most objects, the reconstructed spectra achieve empirical credible-interval coverage of 95 percent across wavelength. This suggests the diversity of near-IR spectral shapes is governed by structured, correlated surface processes rather than stochastic variation. Practically, we apply this framework to survey optimization, quantifying the information content of JWST/NIRCam filters to identify optimal configurations (e.g., F090W, F115W, F410M, F460M) for TNO taxonomy. Additionally, we demonstrate the pipeline's capability to detect and reconstruct rare spectral types, such as the peculiar Neptune Trojans 2006 RJ103 and 2011 SO277, by allowing constraining photometry to select low-probability intermediate models from the continuous topological manifold. Ultimately, this framework bridges the gap between sparse photometry and spectroscopy, providing a statistically rigorous tool to map the compositional structure of minor planets in upcoming large-scale surveys.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper develops a probabilistic latent-space framework for reconstructing near-IR spectra of trans-Neptunian objects from sparse photometry. It trains PCA on a sample of near-IR spectra to define a low-dimensional manifold, then performs Bayesian inference over the PCA coefficients given photometric constraints while propagating uncertainties. Leave-one-out cross-validation on the training spectra is used to argue that 4-5 principal components suffice for taxonomy and 8-10 yield 95% empirical credible-interval coverage; the framework is further applied to JWST/NIRCam filter optimization and to outlier reconstruction for two Neptune Trojans.

Significance. If externally validated, the approach would provide a statistically grounded way to extract spectral information from the much larger photometric datasets expected from LSST and other surveys, while highlighting the low-dimensional structure of TNO spectral variability. The filter-optimization results and outlier-handling capability are practically useful, but the strength of these claims depends on whether the PCA basis generalizes beyond the training spectra.

major comments (3)
  1. [Results on LOO CV and spectral reconstruction] The leave-one-out cross-validation (reported in the results section on dimensionality and coverage) holds out spectra that were already used to construct the PCA basis. This internal validation cannot confirm that the Bayesian posterior over PCA coefficients remains well-calibrated when applied to photometry of TNOs whose near-IR spectral shapes lie outside the training manifold, which is load-bearing for the reconstruction-fidelity and 95% coverage claims.
  2. [Methods on PCA training and data sample] The representativeness assumption for the training spectral sample is stated but not tested with external data or selection-effect analysis. No quantitative assessment of how well the sampled TNOs cover compositional or orbital phase space is provided, directly affecting whether the reported dimensionality (4-5 PCs for taxonomy) and uncertainty calibration generalize to new objects.
  3. [Application to outlier detection] The two Neptune Trojan examples (2006 RJ103 and 2011 SO277) are presented as demonstrations of outlier detection via the continuous manifold, but two cases do not constitute a systematic external test of generalization or robustness to objects with rare spectral types.
minor comments (2)
  1. [Data and methods] Clarify the precise number of spectra in the training set and any cuts applied (e.g., wavelength coverage, S/N thresholds) so readers can assess sample size relative to the number of retained principal components.
  2. [Survey optimization results] In the JWST filter-optimization discussion, explicitly state the assumed photometric uncertainties and how they propagate into the information-content metric.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We agree that the validation aspects require clarification regarding the scope of the claims. We will revise the paper to better articulate the limitations of the internal validation and the representativeness of the training sample, while reframing the outlier examples as illustrative. Our responses to each major comment are provided below.

read point-by-point responses

  1. Referee: The leave-one-out cross-validation (reported in the results section on dimensionality and coverage) holds out spectra that were already used to construct the PCA basis. This internal validation cannot confirm that the Bayesian posterior over PCA coefficients remains well-calibrated when applied to photometry of TNOs whose near-IR spectral shapes lie outside the training manifold, which is load-bearing for the reconstruction-fidelity and 95% coverage claims.

    Authors: We acknowledge that the leave-one-out cross-validation provides an internal assessment within the training spectral sample and does not directly test generalization to spectra outside the observed manifold. The 95% credible interval coverage is an empirical result for the held-out training spectra. For new TNOs, if their spectral shapes fall outside the training distribution, the model may underperform, which the framework can flag through elevated reconstruction uncertainties or outlier probabilities. In the revised manuscript, we will add explicit language in the results and discussion sections stating that the coverage claims apply to objects consistent with the training distribution and discuss the implications for applying the method to the broader TNO population. This revision will strengthen the manuscript by transparently addressing the generalization limits. revision: partial

  2. Referee: The representativeness assumption for the training spectral sample is stated but not tested with external data or selection-effect analysis. No quantitative assessment of how well the sampled TNOs cover compositional or orbital phase space is provided, directly affecting whether the reported dimensionality (4-5 PCs for taxonomy) and uncertainty calibration generalize to new objects.

    Authors: The training sample is limited to the available published near-IR spectra of TNOs, which we acknowledge may not fully represent the entire population due to observational biases toward brighter or more accessible objects. We agree that a quantitative assessment of coverage is warranted. In the revised version, we will include an analysis of the orbital parameter space (e.g., semi-major axis vs. inclination) covered by the training objects compared to the known TNO catalog, along with a discussion of potential selection effects. This will provide readers with a better understanding of the sample's representativeness. However, performing a full external validation would require additional independent spectra, which are not currently available in sufficient numbers. revision: yes

  3. Referee: The two Neptune Trojan examples (2006 RJ103 and 2011 SO277) are presented as demonstrations of outlier detection via the continuous manifold, but two cases do not constitute a systematic external test of generalization or robustness to objects with rare spectral types.

    Authors: We concur that two specific cases do not amount to a systematic test of the outlier detection capability. These examples were intended to demonstrate how the continuous manifold allows reconstruction of intermediate spectral types for objects that deviate from the primary modes. In the revised manuscript, we will modify the relevant section to present these as illustrative case studies of the framework's application to potential outliers, rather than as validation of the method's robustness. We will also note that systematic testing would benefit from larger photometric samples expected from future surveys like LSST. revision: yes

Circularity Check

0 steps flagged

No circularity in derivation chain

full rationale

The framework trains PCA on an external near-IR spectral sample, performs Bayesian inference over latent coefficients given photometry, and reports LOO cross-validation results on held-out members of that same sample. No equation reduces a claimed output (e.g., 4–5 PCs for taxonomy or 95 % coverage) to a fitted quantity by construction, nor does any load-bearing premise rest on a self-citation whose content is itself unverified. The derivation remains self-contained against the external spectral training data and the photometric observations to which the model is applied.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 1 invented entities

The framework rests on dimensionality reduction of spectral data and probabilistic inference; the number of components is selected via cross-validation rather than derived from first principles.

free parameters (1)
  • Number of principal components = 4-5 or 8-10
    Selected via leave-one-out cross-validation as 4-5 for taxonomy and 8-10 for reconstruction fidelity.
axioms (2)
  • domain assumption TNO near-IR spectral variability is captured by a low-dimensional linear manifold via PCA.
    Invoked to reduce the spectral space for Bayesian reconstruction.
  • domain assumption Photometric measurements provide sufficient constraints to perform inference in the latent space.
    Central to the reconstruction step.
invented entities (1)
  • Continuous topological manifold of TNO spectra no independent evidence
    purpose: To enable interpolation and outlier detection via probabilistic selection of intermediate models.
    Constructed from the PCA of the training spectra.

pith-pipeline@v0.9.0 · 5628 in / 1440 out tokens · 58352 ms · 2026-05-08T05:02:51.832508+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

50 extracted references · 46 canonical work pages · 1 internal anchor

  1. [1]

    Fast likelihood-free cosmology with neural density estimators and active learning , ISSN=

    Alsing, J., Charnock, T., Feeney, S., & Wandelt, B. 2019, MNRAS, 488, 4440, doi: 10.1093/mnras/stz1960

    work page doi:10.1093/mnras/stz1960 2019

  2. [2]

    Massive optimal data compression and density estimation for scalable, likelihood-free inference in cosmology

    Alsing, J., Wandelt, B., & Feeney, S. 2018, MNRAS, 477, 2874, doi: 10.1093/mnras/sty819 Arsalan Bashir, S. M., Wang, Y., Khan, M., & Niu, Y. 2021, arXiv e-prints, arXiv:2102.09351, doi: 10.48550/arXiv.2102.09351 Astropy Collaboration, Robitaille, T. P., Tollerud, E. J., et al. 2013, A&A, 558, A33, doi: 10.1051/0004-6361/201322068 Astropy Collaboration, Pr...

    work page doi:10.1093/mnras/sty819 2018

  3. [3]

    A., Belskaya, I

    Barucci, M. A., Belskaya, I. N., Fulchignoni, M., & Birlan, M. 2005, AJ, 130, 1291, doi: 10.1086/431957

    work page doi:10.1086/431957 2005

  4. [4]

    R., Milby, Z., Wong, I., & Brown, M

    Belyakov, M., Davis, M. R., Milby, Z., Wong, I., & Brown, M. E. 2024, PSJ, 5, 119, doi: 10.3847/PSJ/ad3d55

    work page doi:10.3847/psj/ad3d55 2024

  5. [5]

    H., Bernstein, G

    Bernardinelli, P. H., Bernstein, G. M., Jindal, N., et al. 2023, ApJS, 269, 18, doi: 10.3847/1538-4365/acf6bf

    work page doi:10.3847/1538-4365/acf6bf 2023

  6. [6]

    H., Bernstein, G

    Bernardinelli, P. H., Bernstein, G. M., Abbott, T. M. C., et al. 2025, AJ, 169, 305, doi: 10.3847/1538-3881/adc459

    work page doi:10.3847/1538-3881/adc459 2025

  7. [7]

    H., & Little, E

    Bizhani, M., Ardakani, O. H., & Little, E. 2022, Scientific Reports, 12, 4264, doi: 10.1038/s41598-022-08170-8

    work page doi:10.1038/s41598-022-08170-8 2022

  8. [8]

    2019 b , , 874, 106, 10.3847/1538-4357/ab06c1

    Brout, D., Sako, M., Scolnic, D., et al. 2019, ApJ, 874, 106, doi: 10.3847/1538-4357/ab06c1

    work page doi:10.3847/1538-4357/ab06c1 2019

  9. [9]

    A., Dotto, E., & Strazzulla, G

    Brunetto, R., Barucci, M. A., Dotto, E., & Strazzulla, G. 2006, ApJ, 644, 646, doi: 10.1086/503359

    work page doi:10.1086/503359 2006

  10. [10]

    2025, ApJL, 982, L8, doi: 10.3847/2041-8213/adb977

    Brunetto, R., Hénault, E., Cryan, S., et al. 2025, ApJL, 982, L8, doi: 10.3847/2041-8213/adb977

    work page doi:10.3847/2041-8213/adb977 2025

  11. [11]

    J., & Binzel, R

    Bus, S. J., & Binzel, R. P. 2002a, Icarus, 158, 106, doi: 10.1006/icar.2002.6857

    work page doi:10.1006/icar.2002.6857 2002

  12. [12]

    J., Binzel, R

    Bus, S. J., & Binzel, R. P. 2002b, Icarus, 158, 146, doi: 10.1006/icar.2002.6856

    work page doi:10.1006/icar.2002.6856 2002

  13. [13]

    2023, JWST Calibration Pipeline, 1.11.4 Zenodo, doi: 10.5281/zenodo.8247246 24

    Bushouse, H., Eisenhamer, J., Dencheva, N., et al. 2023, JWST Calibration Pipeline, 1.11.4 Zenodo, doi: 10.5281/zenodo.8247246 24

    work page doi:10.5281/zenodo.8247246 2023

  14. [14]

    XGBoost: A Scalable Tree Boosting System

    Chen, T., & Guestrin, C. 2016, in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (ACM), 785–794, doi: 10.1145/2939672.2939785

    work page doi:10.1145/2939672.2939785 2016

  15. [15]

    P., Moroz, L

    Cruikshank, D. P., Moroz, L. V., & Clark, R. N. 2019, Spectroscopy of Ices, Volatiles, and Organics in the Visible and Infrared Regions, ed. J. L. Bishop, I. Bell, James F., & J. E. Moersch, 102–119, doi: 10.1017/9781316888872.007 Dalle Ore, C. M., Barucci, M. A., Emery, J. P., et al. 2015, Icarus, 252, 311, doi: 10.1016/j.icarus.2015.01.014

    work page doi:10.1017/9781316888872.007 2019

  16. [16]

    2026, Gaia DR3 supervised classification of asteroid reflectance spectra, https://arxiv.org/abs/2602.22816

    Delbo, M., Dyer, T., Bhat, U., et al. 2026, Gaia DR3 supervised classification of asteroid reflectance spectra, https://arxiv.org/abs/2602.22816

    work page arXiv 2026

  17. [17]

    E., Binzel, R

    DeMeo, F. E., Binzel, R. P., Slivan, S. M., & Bus, S. J. 2009, Icarus, 202, 160, doi: 10.1016/j.icarus.2009.02.005

    work page doi:10.1016/j.icarus.2009.02.005 2009

  18. [18]

    E., & Carry, B

    DeMeo, F. E., & Carry, B. 2013, Icarus, 226, 723, doi: 10.1016/j.icarus.2013.06.027

    work page doi:10.1016/j.icarus.2013.06.027 2013

  19. [19]

    Nature 505, 629–634

    DeMeo, F. E., & Carry, B. 2014, Nature, 505, 629, doi: 10.1038/nature12908

    work page doi:10.1038/nature12908 2014

  20. [20]

    P., Wong, I., Brunetto, R., et al

    Emery, J. P., Wong, I., Brunetto, R., et al. 2024, Icarus, 414, 116017, doi: 10.1016/j.icarus.2024.116017

    work page doi:10.1016/j.icarus.2024.116017 2024

  21. [21]

    AutoGluon-Tabular: Robust and Accurate AutoML for Structured Data

    Erickson, N., Mueller, J., Shirkov, A., et al. 2020, arXiv preprint arXiv:2003.06505

    work page internal anchor Pith review arXiv 2020

  22. [22]

    C., Pike, R

    Fraser, W. C., Pike, R. E., Marsset, M., et al. 2023, PSJ, 4, 80, doi: 10.3847/PSJ/acc844

    work page doi:10.3847/psj/acc844 2023

  23. [23]

    He, K., Chen, X., Xie, S., et al. 2022, in 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 15979–15988, doi: 10.1109/CVPR52688.2022.01553 Hénault, E., Brunetto, R., Pinilla-Alonso, N., et al. 2025, A&A, 694, A126, doi: 10.1051/0004-6361/202452321

    work page doi:10.1109/cvpr52688.2022.01553 2022

  24. [24]

    Data analysis recipes: Fitting a model to data

    Hogg, D. W., Bovy, J., & Lang, D. 2010, arXiv e-prints, arXiv:1008.4686, doi: 10.48550/arXiv.1008.4686

    work page Pith review doi:10.48550/arxiv.1008.4686 2010

  25. [25]

    J., Brunetto, R., Cruikshank, D

    Holler, B. J., Brunetto, R., Cruikshank, D. P., et al. 2025, Research Notes of the American Astronomical Society, 9, 241, doi: 10.3847/2515-5172/ae03a2

    work page doi:10.3847/2515-5172/ae03a2 2025

  26. [26]

    2015, in 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 5197–5206, doi: 10.1109/CVPR.2015.7299156 Ivezić, Ž., Connolly, A

    Huang, J.-B., Singh, A., & Ahuja, N. 2015, in 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 5197–5206, doi: 10.1109/CVPR.2015.7299156 Ivezić, Ž., Connolly, A. J., VanderPlas, J. T., & Gray, A. 2014, Statistics, Data Mining, and Machine Learning in Astronomy: A Practical Python Guide for the Analysis of Survey Data, doi: 10.1515/9...

    work page doi:10.1109/cvpr.2015.7299156 2015

  27. [27]

    Jeffrey, J

    Jeffrey, N., Alsing, J., & Lanusse, F. 2021, MNRAS, 501, 954, doi: 10.1093/mnras/staa3594

    work page doi:10.1093/mnras/staa3594 2021

  28. [28]

    J., et al

    Licandro, J., Pinilla-Alonso, N., Holler, B. J., et al. 2025, Nature Astronomy, 9, 245, doi: 10.1038/s41550-024-02417-2

    work page doi:10.1038/s41550-024-02417-2 2025

  29. [29]

    2024, AJ, 167, 13, doi: 10.3847/1538-3881/ad0b7a

    Luo, N., Wang, X., Gu, S., et al. 2024, AJ, 167, 13, doi: 10.3847/1538-3881/ad0b7a

    work page doi:10.3847/1538-3881/ad0b7a 2024

  30. [30]

    W., Holler, B

    Markwardt, L., Lin, H. W., Holler, B. J., et al. 2025, PSJ, 6, 154, doi: 10.3847/PSJ/addecd Nesvorný, D., Vokrouhlický, D., Alexand ersen, M., et al. 2020, AJ, 160, 46, doi: 10.3847/1538-3881/ab98fb

    work page doi:10.3847/psj/addecd 2025

  31. [31]

    2011, Journal of Machine Learning Research, 12, 2825

    Pedregosa, F., Varoquaux, G., Gramfort, A., et al. 2011, Journal of Machine Learning Research, 12, 2825

    2011

  32. [32]

    2015, A&A, 577, A35, doi: 10.1051/0004-6361/201425436

    Peixinho, N., Delsanti, A., & Doressoundiram, A. 2015, A&A, 577, A35, doi: 10.1051/0004-6361/201425436

    work page doi:10.1051/0004-6361/201425436 2015

  33. [33]

    2012, A&A, 546, A86, doi: 10.1051/0004-6361/201219057 Penttilä, A., Fedorets, G., & Muinonen, K

    Peixinho, N., Delsanti, A., Guilbert-Lepoutre, A., Gafeira, R., & Lacerda, P. 2012, A&A, 546, A86, doi: 10.1051/0004-6361/201219057 Penttilä, A., Fedorets, G., & Muinonen, K. 2022, Frontiers in Astronomy and Space Sciences, 9, 816268, doi: 10.3389/fspas.2022.816268 Penttilä, A., Hietala, H., & Muinonen, K. 2021, A&A, 649, A46, doi: 10.1051/0004-6361/202038545

    work page doi:10.1051/0004-6361/201219057 2012

  34. [34]

    E., Fraser, W

    Pike, R. E., Fraser, W. C., Schwamb, M. E., et al. 2017, AJ, 154, 101, doi: 10.3847/1538-3881/aa83b1

    work page doi:10.3847/1538-3881/aa83b1 2017

  35. [35]

    N., et al

    Pinilla-Alonso, N., Brunetto, R., De Prá, M. N., et al. 2025, Nature Astronomy, 9, 230, doi: 10.1038/s41550-024-02433-2

    work page doi:10.1038/s41550-024-02433-2 2025

  36. [36]

    E., Fraser, W

    Schwamb, M. E., Fraser, W. C., Bannister, M. T., et al. 2019, ApJS, 243, 12, doi: 10.3847/1538-4365/ab2194 Souza Feliciano, A. C., Holler, B., Grundy, W., et al. 2025, in EPSC-DPS Joint Meeting 2025, Vol. 2025, EPSC–DPS2025–848, doi: 10.5194/epsc-dps2025-848

    work page doi:10.3847/1538-4365/ab2194 2019

  37. [37]

    A., Fraser, W

    Stansberry, J. A., Fraser, W. C., Trilling, D. E., et al. 2021, An Ultra-Sensitive Pencil Beam Search for 10 km Trans-Neptunian Objects„ JWST Proposal. Cycle 1, ID. #1568

    2021

  38. [38]

    Kutulakos, K. N. 2025, Super Resolved Imaging with Adaptive Optics, https://arxiv.org/abs/2508.04648

    work page arXiv 2025

  39. [39]

    S., & Lin, S

    Tai, Y.-W., Liu, S., Brown, M. S., & Lin, S. 2010, in 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2400–2407, doi: 10.1109/CVPR.2010.5539933

    work page doi:10.1109/cvpr.2010.5539933 2010

  40. [40]

    2025, AJ, 169, 201, doi: 10.3847/1538-3881/adb710

    Tang, Y., Jiang, Y., Feng, Y., Zhang, X., & Jiang, X. 2025, AJ, 169, 201, doi: 10.3847/1538-3881/adb710

    work page doi:10.3847/1538-3881/adb710 2025

  41. [41]

    1987, Inverse problem theory

    Tarantola, A. 1987, Inverse problem theory. Methods for data fitting and model parameter estimation

    1987

  42. [42]

    C., & Romanishin, W

    Tegler, S. C., & Romanishin, W. 1998, Nature, 392, 49, doi: 10.1038/32108

    work page doi:10.1038/32108 1998

  43. [43]

    Tsiganis, K., Gomes, R., Morbidelli, A., & Levison, H. F. 2005, Nature, 435, 459, doi: 10.1038/nature03539

    work page doi:10.1038/nature03539 2005

  44. [44]

    2017, in Super-Resolution Imaging (CRC Press), 155–186

    Vandewalle, P., Sbaiz, L., & Vetterli, M. 2017, in Super-Resolution Imaging (CRC Press), 155–186

    2017

  45. [45]

    E., et al

    Virtanen, P., Gommers, R., Oliphant, T. E., et al. 2020, Nature Methods, 17, 261, doi: 10.1038/s41592-019-0686-2

    work page doi:10.1038/s41592-019-0686-2 2020

  46. [46]

    2015, Deep Networks for Image Super-Resolution with Sparse Prior, https://arxiv.org/abs/1507.08905

    Wang, Z., Liu, D., Yang, J., Han, W., & Huang, T. 2015, Deep Networks for Image Super-Resolution with Sparse Prior, https://arxiv.org/abs/1507.08905

    work page arXiv 2015

  47. [47]

    J., Fraser, W

    Wong, I., Holler, B. J., Fraser, W. C., & Brown, M. E. 2025, PSJ, 6, 271, doi: 10.3847/PSJ/ae1d73

    work page doi:10.3847/psj/ae1d73 2025

  48. [48]

    Star: A benchmark for astronomical star fields super-resolution.arXiv preprint arXiv:2507.16385,

    Wu, K.-C., Zhuang, G., Huang, J., et al. 2025, STAR: A Benchmark for Astronomical Star Fields Super-Resolution, https://arxiv.org/abs/2507.16385

    work page arXiv 2025

  49. [49]

    P., Burbine, T

    Xu, S., Binzel, R. P., Burbine, T. H., & Bus, S. J. 1995, Icarus, 115, 1, doi: 10.1006/icar.1995.1075

    work page doi:10.1006/icar.1995.1075 1995

  50. [50]

    M., et al

    Zhang, C., Zhu, C., Turner, A. M., et al. 2023, Science Advances, 9, eadg6936, doi: 10.1126/sciadv.adg6936

    work page doi:10.1126/sciadv.adg6936 2023