Pith. sign in

REVIEW 4 major objections 4 minor 63 references

Learning from galactic rotation curves: a neural network approach

T0 review · 4 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Trained neural networks can take an observed dwarf-galaxy rotation curve and return dark matter halo and baryon parameters that describe the data as well as MCMC chains do.

desk verdict A clean proof-of-concept that NNs can match MCMC for ULDM rotation-curve fitting, but parameter recovery on known truths is never demonstrated. read the letter →

arxiv 2412.03547 v2 pith:XBB77WH5 submitted 2024-12-04 astro-ph.CO hep-ph

Bihag Dave , Gaurav Goswami This is my paper · ORCID
classification astro-ph.COhep-ph
keywords galacticrotationcurvesSPARCgalaxiesultra-lightdarkmatterneuralnetworksparameterinferenceheteroscedasticlossMCMCcomparisondwarf
topics Dark Matter
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

The paper asks whether a neural network can do what a Bayesian MCMC analysis does for galactic rotation curves: take an observed rotation curve and return the parameters of the dark matter halo and the stellar disk. Using seven dark-matter-dominated dwarf galaxies from the SPARC catalog, the authors generate about half a million simulated rotation curves per galaxy under an ultra-light dark matter (ULDM) core-plus-NFW-halo model, train a separate network for each galaxy, and then feed the observed curves to the trained nets. Their central claim is that the networks extract parameter values that describe the observations well: point estimates mostly fall within the 1-$\sigma$ intervals of emcee MCMC, and a heteroscedastic loss function yields large uncertainties precisely for the parameters, $r_s$ and $\Upsilon_*$, to which these rotation curves are insensitive. The paper itself cautions that these learned uncertainties are not equivalent to Bayesian posteriors and that training is still slower than MCMC, so the contribution is a complementary, likelihood-free route to parameter estimation rather than a replacement.

What carries the argument

The load-bearing object is a fully connected neural network with two hidden layers of 200 ReLU neurons, trained per galaxy on $5\times10^5$ simulated rotation curves generated from a ULDM core-halo density profile, namely a Schive-style soliton core matched continuously to an NFW envelope at a transition radius $r_t$. The input is the vector of observed velocities at the galaxy's $N_{\rm obs}$ radius bins, and the target is the five-parameter vector $P$. Two additions carry the argument: adding Gaussian noise drawn from the SPARC error bars to each simulated input during training, which the paper equates with Tikhonov regularization and which makes the network cope with non-smooth observed curves; and replacing the mean-squared-error loss with a heteroscedastic loss that learns a Gaussian variance for each parameter alongside its mean, so uncertainties are produced at inference time without writing a likelihood. The MCMC comparison uses the same model and uniform priors, giving the reference posteriors that the neural outputs are tested against.

What would settle it

Generate a set of mock galaxies whose true parameters lie deliberately outside the Table 1 ranges, for example $r_t < 1$ kpc or $\Upsilon_* > 0.8$, feed their rotation curves to the trained networks, and check whether the inferred parameters track the true values or instead collapse to the boundary of the training ranges; the latter would confirm that the reported agreement with MCMC is conditional on the chosen parameter box.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that a per-galaxy feed-forward neural network trained on simulated rotation curves can serve as a direct map from observed rotation-curve velocities to the five ULDM-plus-baryon parameters $P=(m, s, r_t, r_s, \Upsilon_*)$. The key empirical results are that adding Gaussian noise drawn from the observed velocity errors to the training inputs substantially improves the network's fit to real SPARC data, that point estimates from both the multiple-realizations method and the heteroscedastic-loss network lie within the 1-$\sigma$ MCMC intervals for most parameters and galaxies, and that the heteroscedastic network automatically assigns large uncertainties to $r_s$ and $\Upsilon_*$, which are the parameters the rotation curves barely constrain. The paper frames this as evidence that neural networks can be a useful complementary approach to likelihood-based inference for this class of problem, while noting that the uncertainty from the heteroscedastic loss need not match the Bayesian posterior and that the approach is not yet faster than MCMC.

Load-bearing premise

The inference stands or falls on whether the hand-picked uniform parameter ranges in Table 1, chosen by eye to make simulated curves resemble each observed rotation curve, actually cover all plausible parameter values, since a network can never return parameters outside its training distribution.

Editorial extensions

If this is right

  • For the seven SPARC dwarf galaxies studied, adding noise to training inputs turns networks whose predicted curves miss the observed rotation curve badly into networks whose reconstructed curves have low reduced $\chi^2$.
  • Multiple realizations of the observed rotation curve, pushed through a trained network, produce parameter chains whose medians and correlations (for example, a positive $m$--$s$ correlation and an anticorrelation of $r_t$ with $m$ and $s$) resemble MCMC posteriors for $m$, $s$, and $r_t$.
  • A heteroscedastic-loss network returns usefully large uncertainties for $r_s$ and $\Upsilon_*$, the parameters to which dwarf-galaxy rotation curves are insensitive, correcting the overconfidence of the multiple-realizations method.
  • If these results hold, rotation-curve parameter inference can be done without an explicit likelihood, which matters for dark matter models whose rotation-curve predictions are expensive or hard to evaluate.
  • The trained networks predict galaxy-specific parameters rather than a single global ULDM particle mass, and the paper identifies pooling multiple galaxies into one network as the natural next step.

Reading between the lines

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

  • A natural test the paper does not run is to push mock rotation curves generated from parameters outside the Table 1 ranges through the trained networks; because inference is conditional on the training distribution, such a test would show how much of the reported agreement with MCMC is a consequence of sharing the same hand-picked priors.
  • The heteroscedastic network's large uncertainties on $r_s$ and $\Upsilon_*$ suggest the loss function is implicitly learning a sensitivity measure, close to the Fisher information of the rotation curve with respect to each parameter, which could be formalized and compared with the Cramer-Rao bound on mock data.
  • Because the networks ignore radius values entirely, they cannot distinguish two rotation curves with identical velocities at different radii; feeding shuffled or rescaled radii to the trained networks is a cheap way to quantify how much radial information the current pipeline discards.
  • The per-galaxy training strategy means the method does not yet produce a single constraint on the fundamental ULDM particle mass $m$ across the sample; pooling galaxies into one network with radii as an additional input is the obvious extension and is explicitly named as future work in the paper.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper trains per-galaxy feed-forward neural networks to map rotation curves of seven SPARC dwarf galaxies to five parameters of an ultra-light dark matter (ULDM) model: particle mass m, scale parameter s, transition radius r_t, NFW scale radius r_s, and stellar mass-to-light ratio Upsilon_*. Training data are simulated rotation curves drawn from uniform parameter ranges chosen galaxy by galaxy in Table 1. The authors compare MSE training with and without injected observational noise, and they develop two uncertainty quantification schemes: multiple realizations of the observed curve passed through the network, and a heteroscedastic loss that predicts per-parameter variances. Validation consists of reduced chi-squared values of reconstructed rotation curves and a comparison of point estimates and 1-sigma intervals with emcee MCMC. The paper concludes that the trained networks can extract parameters that describe the observed rotation curves well.

Significance. If the claimed accuracy were established, the method would offer a useful fast emulator-like alternative to MCMC for this five-parameter ULDM model, and the heteroscedastic uncertainty treatment would be an interesting lightweight addition to the astrophysical machine-learning toolbox. The paper's strengths are its explicit internal consistency check against MCMC, its careful treatment of training-set noise, and its honest discussion of the limitations of the heteroscedastic loss (Section 5.1 and the caveats in Section 6). However, the MCMC comparison shares the same forward model, the same uniform priors, and the same Gaussian noise model, so agreement between the neural network and MCMC is a self-consistency test rather than an external accuracy test. The absence of quantitative ground-truth validation on the held-out simulated test set is the main weakness and is directly relevant to the central claim.

major comments (4)
  1. [Sections 2.4 and 3.1] The test set of 5x10^4 simulated curves with known target parameters is split but never used to report quantitative recovery accuracy for the five parameters. Figure 2 shows only three example reconstructed curves, and the text reports only loss values and a qualitative 'visually similar' statement. Since the abstract's central claim is that parameters are 'extracted', the paper should report test-set metrics — for example, median bias, RMSE, and per-parameter correlation between predicted and true values — for each of m, s, r_t, r_s, and Upsilon_*. This is especially important for r_s and Upsilon_*, which the paper says have little influence on the rotation curves and for which the network uncertainties are large.
  2. [Section 5, Eq. (5.1)] The MCMC comparison uses the same forward model (eqs. 2.1-2.6), the same uniform priors in Table 1, and the same Gaussian noise model as the neural network training. Therefore the agreement shown in Figure 9 and Table 2 is a consistency check between two methods that share the same generative assumptions; it does not establish that the network recovers the true parameter values. A network trained under the same priors is expected to approximate the prior-constrained posterior mean. The paper should explicitly frame the MCMC agreement as a self-consistency test and add a validation against simulated curves with known parameters, reporting coverage of the 1-sigma intervals for both the multiple-realizations and heteroscedastic methods.
  3. [Section 2.3, step 1 and Table 1] The uniform ranges for each parameter are chosen by visually comparing simulated curves with the observed rotation curve, as stated in step 1 of Section 2.3. This makes the training distribution data-informed. Consequently, the low reduced chi-squared values in Figures 3, 4, 6, and 7 are partly guaranteed by construction, and parameters outside the chosen ranges are unrecoverable by the network. The paper should state in the abstract and Section 6 that the inferred values and uncertainties are conditional on these hand-picked ranges, and ideally test the sensitivity of the results to reasonable variations of the Table 1 ranges, particularly for r_s and Upsilon_*.
  4. [Section 4.2, Eq. (4.1), and Figure 5] The heteroscedastic loss assumes that the five parameters are independent and Gaussian, a point the paper acknowledges. The calibration check in Figure 5 is qualitative: the text says that 'for most samples' the difference lies within three times the predicted uncertainty, but no fraction is quoted and no 1-sigma coverage statistic is computed. As a result, it is unclear whether the reported uncertainties have the claimed frequentist meaning. The authors should report, on the test set, the fraction of true parameter values contained within the predicted 1-sigma interval for each parameter and each uncertainty method, and compare these fractions with the expected 68% for calibrated Gaussian uncertainties.
minor comments (4)
  1. [Throughout] There are several typographical errors: 'T able 1' in Section 2.3, 'obatined' in Section 4.1, 'discreet' in Section 2.2.1, and 'UGC 5271' in the caption of Figure 6 (should be UGC 5721).
  2. [Table 2, UGCA 444 row] The MSE/multiple-realizations row for UGCA 444 reports 1-sigma uncertainties of about 0.01 kpc for r_t and 0.02 kpc for r_s, which are implausibly small compared with the other galaxies and contradict the paper's statement that the multiple-realizations method under-estimates uncertainties for insensitive parameters. This outlier should either be explained or explicitly flagged as a known failure of that method.
  3. [Section 4.2, Eq. (4.1)] The notation for the heteroscedastic loss would benefit from clarification: the text defines s_ik as the log of the predicted variance, but the equation uses e^{-s_ik} without explicitly connecting the two in the displayed formula. Adding a short sentence defining s_ik in the equation would improve readability.
  4. [Section 6, caveat (c)] The authors note in the discussion that the input vector contains only velocities and not the associated radii. This is an important limitation for the generality of the method and should also be mentioned in the abstract, so that readers do not overstate the approach's transferability to other galaxies or to rotation curves with different radial sampling.

Circularity Check

1 steps flagged · score 4.0 of 10

NN parameter extraction is validated partly against its own inputs: per-galaxy training priors are tuned to the observed rotation curves, and the MCMC comparison uses the same forward model and the same uniform priors, making it a consistency check rather than an independent accuracy test.

  1. fitted input called prediction [Sec. 2.3 (Table 1) and Sec. 5 (MCMC comparison)]
    "First, we define a uniform distribution for each parameter. While any random combination of parameters can form a velocity curve, not all of them will be visually similar to the observed one. Hence, choosing sensible ranges for the uniform distributions for each galaxy is important. We do this by examining how the numerically generated curves vary with each parameter in comparison to the observed rotation curve and choose the lower and upper limits accordingly. ... We also use the same uniform priors as those defined in Table 1."

    The per-galaxy uniform ranges in Table 1 are selected by visually matching simulated curves to the observed rotation curve, so the training distribution is already anchored to the target observation. A network trained on curves drawn from these ranges is therefore predisposed to output parameters whose forward-model curves resemble the observed one; the reported low chi2_red between predicted-parameter curves and observed curves is partially guaranteed by construction of the training prior, not an independent test. The MCMC comparison in Sec. 5 uses the same forward model (eqs. 2.1-2.6) and 'the same uniform priors as those defined in Table 1', with a Gaussian likelihood.

full rationale

The paper's derivation chain is: define a ULDM-plus-baryons rotation-curve model, choose per-galaxy uniform parameter ranges by visual comparison with the observed rotation curve, generate 5e5 simulated curves from those ranges, train a per-galaxy neural network, feed the observed rotation curve to the network, and validate the inferred parameters via reduced chi2 and via comparison with emcee MCMC that uses the same model and the same uniform priors. The circular element is the prior selection: because the training distribution is chosen to make simulated curves visually similar to the observed curve, the network's ability to produce curves with low chi2_red is partly built into the training data rather than independently demonstrated. The MCMC comparison is a meaningful test that the neural network has learned the posterior implied by the shared generative model and priors, but it is not an external accuracy test; no test on simulated rotation curves with known true parameters, and no bias or coverage statistics, are reported. This makes the headline claim 'the trained neural networks can extract parameters that describe observations well' validated largely against the paper's own inputs. The paper is candid about important limitations (Sec. 5.1: heteroscedastic uncertainties may not equal Bayesian posterior; Sec. 6: the ANN approach cannot surpass or replace MCMC; caveat (c): radius information is unused), which reduces the severity. Self-citations [38,39] are not load-bearing for the central inference result. Overall, this is moderate circularity in the validation logic: score 4, because the neural-network implementation, noise treatment, and uncertainty quantification retain independent technical content, but the central validation reduces partly to a data-informed prior and a self-consistent MCMC comparison.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The central claim rests primarily on the chosen ULDM profile, the rotation curve model, and the hand-picked per-galaxy priors. The free parameters are the prior range boundaries and the neural network hyperparameters. No new physical entities are introduced by this paper.

free parameters (2)
  • Galaxy-specific uniform prior ranges (Table 1) = m in [1,10] x 10^-23 eV; s in [1,9] x 10^3; r_t lower bound 1 kpc, upper bound largest observed radius; r_s in [1,15]…
    Chosen by visually comparing simulated curves to observed rotation curves (Sec. 2.3, step 1). The posterior estimates depend directly on these hand-picked ranges.
  • Neural network hyperparameters (architecture, learning rate, batch size, dropout, epochs) = 2 hidden layers, 200 neurons per layer, ReLU, Adam with LR 1e-4, batch size 32, dropout 0.2, 250 epochs
    Selected via grid search on 4 galaxies and trial-and-error (Sec. 2.5). A different hyperparameter choice could change point estimates and uncertainties.
assumptions (5)
  • domain assumption ULDM density profile in eqs. (2.2) to (2.4): soliton core with Schive+2014 fitting function plus NFW envelope with continuity at r_t
    Used to generate all training simulations; if this profile is not the true dark matter distribution, the inferred parameters are not physically meaningful.
  • domain assumption Rotation curve model in eq. (2.1): quadrature sum of DM, gas, disk and bulge velocities, spherical symmetry, and eq. (2.6) circular velocity
    Standard modeling assumptions, but deviations such as disk geometry or non-spherical halos would bias the inference.
  • domain assumption Observed velocity errors are Gaussian and independent, used both for noise injection in training and for the MCMC likelihood in eq. (5.1)
    SPARC errors are not necessarily Gaussian or independent; this affects both the NN noise injection and the MCMC baseline.
  • ad hoc to paper Uniform priors over the Table 1 ranges are an acceptable representation of the posterior
    The ranges are hand-picked per galaxy (Sec. 2.3), not derived from physical priors, so the posterior is conditional on this choice.
  • ad hoc to paper Heteroscedastic loss in eq. (4.1) assumes parameters are independent and Gaussian
    Acknowledged in Sec. 4.2 and caveat (a) in Sec. 6; this assumption may not hold for correlated model parameters.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning from galactic rotation curves: a neural network approach." pith.science (2026). https://pith.science/paper/XBB77WH5

@misc{pith2026241203547,
  author       = {Pith},
  title        = {Pith review of: Learning from galactic rotation curves: a neural network approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XBB77WH5}},
  note         = {Machine review of arXiv:2412.03547}
}
abstract

For a galaxy, given its observed rotation curve, can one directly infer parameters of the dark matter density profile (such as dark matter particle mass $m$, scaling parameter $s$, core-to-envelope transition radius $r_t$ and NFW scale radius $r_s$), along with Baryonic parameters (such as the stellar mass-to-light ratio $\Upsilon_*$)? In this work, using simulated rotation curves, we train neural networks, which can then be fed observed rotation curves of dark matter dominated dwarf galaxies from the SPARC catalog, to infer parameter values and their uncertainties. Since observed rotation curves have errors, we also explore the very important effect of noise in the training data on the inference. We employ two different methods to quantify uncertainties in the estimated parameters, and compare the results with those obtained using Bayesian methods. We find that the trained neural networks can extract parameters that describe observations well for the galaxies we studied.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 20 canonical work pages

  1. [1]

    R. L. Workmanet al.[Particle Data Group],Review of Particle Physics,PTEP2022(2022), 083C01 doi:10.1093/ptep/ptac097

    doi:10.1093/ptep/ptac097 2022

  2. [2]

    Aghanimet al.[Planck],Planck 2018 results

    N. Aghanimet al.[Planck],Planck 2018 results. VI. Cosmological parameters,Astron. Astrophys.641(2020), A6 [erratum: Astron. Astrophys.652(2021), C4] doi:10.1051/0004-6361/201833910 [arXiv:1807.06209 [astro-ph.CO]]

    arXiv 2020

  3. [3]

    Ivezi´ cet al.[LSST],LSST: from Science Drivers to Reference Design and Anticipated Data Products,Astrophys

    ˇZ. Ivezi´ cet al.[LSST],LSST: from Science Drivers to Reference Design and Anticipated Data Products,Astrophys. J.873(2019) no.2, 111 doi:10.3847/1538-4357/ab042c [arXiv:0805.2366 [astro-ph]]

    arXiv 2019

  4. [4]

    Abazajianet al.[CMB-S4],Snowmass 2021 CMB-S4 White Paper, [arXiv:2203.08024 [astro-ph.CO]]

    K. Abazajianet al.[CMB-S4],Snowmass 2021 CMB-S4 White Paper, [arXiv:2203.08024 [astro-ph.CO]]

    arXiv 2021

  5. [5]

    A. G. Adameet al.[DESI],DESI 2024 VI: cosmological constraints from the measurements of baryon acoustic oscillations,JCAP02(2025), 021 doi:10.1088/1475-7516/2025/02/021 [arXiv:2404.03002 [astro-ph.CO]]

    arXiv 2025

  6. [6]

    Mehta, M

    P. Mehta, M. Bukov, C. H. Wang, A. G. R. Day, C. Richardson, C. K. Fisher and D. J. Schwab, A high-bias, low-variance introduction to Machine Learning for physicists,Phys. Rept.810 (2019), 1-124 doi:10.1016/j.physrep.2019.03.001 [arXiv:1803.08823 [physics.comp-ph]]

    arXiv 2019

  7. [7]

    Alzubaidi et al.,Review of deep learning: concepts, CNN architectures, challenges, applications, future directions,J Big Data8(2021), 53 https://doi.org/10.1186/s40537-021-00444-8

    L. Alzubaidi et al.,Review of deep learning: concepts, CNN architectures, challenges, applications, future directions,J Big Data8(2021), 53 https://doi.org/10.1186/s40537-021-00444-8

    doi:10.1186/s40537-021-00444-8 2021

  8. [8]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser and I. Polosukhin,Attention Is All You Need, [arXiv:1706.03762 [cs.CL]]

    arXiv

Show all 63 references
  1. [9]

    Graff, F

    P. Graff, F. Feroz, M. P. Hobson and A. Lasenby,BAMBI: blind accelerated multimodal Bayesian inference,Mon. Not. Roy. Astron. Soc.421(2012), 169-180 doi:10.1111/j.1365-2966.2011.20288.x [arXiv:1110.2997 [astro-ph.IM]]

    2012

  2. [10]

    G. J. Wang, X. J. Ma, S. Y. Li and J. Q. Xia,Reconstructing Functions and Estimating Parameters with Artificial Neural Networks: A Test with a Hubble Parameter and SNe Ia, Astrophys. J. Suppl.246(2020) no.1, 13 doi:10.3847/1538-4365/ab620b [arXiv:1910.03636 [astro-ph.CO]]

    2020 arXiv

  3. [11]

    G. J. Wang, S. Y. Li and J. Q. Xia,ECoPANN: A Framework for Estimating Cosmological Parameters using Artificial Neural Networks,Astrophys. J. Suppl.249(2020) no.2, 25 doi:10.3847/1538-4365/aba190 [arXiv:2005.07089 [astro-ph.CO]]

    2020 arXiv

  4. [12]

    Fluri, A

    J. Fluri, A. Lucchi, T. Kacprzak, A. Refregier and T. Hofmann,Cosmological parameter estimation and inference using deep summaries,Phys. Rev. D104(2021) no.12, 123526 doi:10.1103/PhysRevD.104.123526 [arXiv:2107.09002 [astro-ph.CO]]

    2021 arXiv

  5. [13]

    M. Ho, A. Farahi, M. M. Rau and H. Trac,Approximate Bayesian Uncertainties on Deep Learning Dynamical Mass Estimates of Galaxy Clusters,Astrophys. J.908(2021) no.2, 204 doi:10.3847/1538-4357/abd101 [arXiv:2006.13231 [astro-ph.CO]]

    2021 arXiv

  6. [14]

    Gmez-Vargas, R

    I. Gmez-Vargas, R. M. Esquivel, R. Garca-Salcedo and J. A. Vzquez,Neural network within a Bayesian inference framework,J. Phys.: Conf. Ser.1723(2021), 012022 doi:10.1088/1742-6596/1723/1/012022

    2021 doi

  7. [15]

    Pal and R

    S. Pal and R. Saha,ParamANN: a neural network to estimate cosmological parameters for ΛCDM Universe using Hubble measurements,Phys. Scripta99(2024) no.11, 115007 doi:10.1088/1402-4896/ad804d [arXiv:2309.15179 [astro-ph.CO]]

    2024 arXiv

  8. [16]

    S. Pal, P. Chanda and R. Saha,Estimation of the Full-sky Power Spectrum between Intermediate and Large Angular Scales from Partial-sky CMB Anisotropies Using an Artificial – 25 – Neural Network,Astrophys. J.945(2023) no.1, 77 doi:10.3847/1538-4357/acb4ee [arXiv:2203.14060 [ast...

    2023 arXiv

  9. [17]

    Pal and R

    S. Pal and R. Saha,Reconstruction of full sky CMB E and B modes spectra removing E-to-B leakage from partial sky using deep learning,J. Astrophys. Astron.44(2023) no.2, 84 doi:10.1007/s12036-023-09974-4 [arXiv:2211.09112 [astro-ph.CO]]

    2023 arXiv

  10. [18]

    R. Shah, S. Saha, P. Mukherjee, U. Garain and S. Pal,LADDER: Revisiting the Cosmic Distance Ladder with Deep Learning Approaches and Exploring Its Applications,Astrophys. J. Suppl.273(2024) no.2, 27 doi:10.3847/1538-4365/ad5558 [arXiv:2401.17029 [astro-ph.CO]]

    2024 arXiv

  11. [19]

    Garcia-Arroyo, I

    G. Garcia-Arroyo, I. G´ omez-Vargas and J. A. V´ azquez,Reconstructing rotation curves with artificial neural networks, [arXiv:2404.05833 [astro-ph.GA]]

    arXiv

  12. [20]

    S. Pal, S. K. Yadav, R. Saha and T. Souradeep,Accurate and Unbiased Reconstruction of CMB B Mode using Deep Learning, [arXiv:2404.18100 [astro-ph.CO]]

    arXiv

  13. [21]

    Hagimoto, A

    R. Hagimoto, A. J. Long and M. A. Amin,Extracting axion string network parameters from simulated CMB birefringence maps using convolutional neural networks,JCAP03(2025), 001 doi:10.1088/1475-7516/2025/03/001 [arXiv:2411.05002 [astro-ph.CO]]

    2025 arXiv

  14. [22]

    Artola, S

    A. Artola, S. E. I. Bosman, P. Gaikwad, F. B. Davies, F. Nasir, E. P. Farina, K. Protuˇ sov´ a, E. Puchwein and B. Spina,Signatures of warm dark matter in the cosmological density fields extracted using Machine Learning, [arXiv:2411.17853 [astro-ph.CO]]

    arXiv

  15. [23]

    Garuda, J

    N. Garuda, J. F. Wu, D. Nelson and A. Pillepich,Estimating Dark Matter Halo Masses in Simulated Galaxy Clusters with Graph Neural Networks, [arXiv:2411.12629 [astro-ph.GA]]

    arXiv

  16. [24]

    Bada-Nerin, O

    R. Bada-Nerin, O. Bulashenko, O. Gramaxo Freitas and J. A. Font,Parameter estimation of microlensed gravitational waves with conditional variational autoencoders,Phys. Rev. D111 (2025) no.8, 084067 doi:10.1103/PhysRevD.111.084067 [arXiv:2412.00566 [gr-qc]]

    2025 arXiv

  17. [25]

    Lelli, S

    F. Lelli, S. S. McGaugh and J. M. Schombert,SPARC: Mass Models for 175 Disk Galaxies with Spitzer Photometry and Accurate Rotation Curves,Astron. J.152(2016), 157 doi:10.3847/0004-6256/152/6/157 [arXiv:1606.09251 [astro-ph.GA]]

    2016 arXiv

  18. [26]

    Salucci,The distribution of dark matter in galaxies,Astron

    P. Salucci,The distribution of dark matter in galaxies,Astron. Astrophys. Rev.27(2019) no.1, 2 doi:10.1007/s00159-018-0113-1 [arXiv:1811.08843 [astro-ph.GA]]

    2019 arXiv

  19. [27]

    Profumo, L

    S. Profumo, L. Giani and O. F. Piattella,An Introduction to Particle Dark Matter,Universe5 (2019) no.10, 213 doi:10.3390/universe5100213 [arXiv:1910.05610 [hep-ph]]

    2019 arXiv

  20. [28]

    H. Y. Schive, T. Chiueh and T. Broadhurst,Cosmic Structure as the Quantum Interference of a Coherent Dark Wave,Nature Phys.10(2014), 496-499 doi:10.1038/nphys2996 [arXiv:1406.6586 [astro-ph.GA]]

    2014 arXiv

  21. [29]

    E. G. M. Ferreira,Ultra-light dark matter,Astron. Astrophys. Rev.29(2021) no.1, 7 doi:10.1007/s00159-021-00135-6 [arXiv:2005.03254 [astro-ph.CO]]

    2021 arXiv

  22. [30]

    Hui,Wave Dark Matter,Ann

    L. Hui,Wave Dark Matter,Ann. Rev. Astron. Astrophys.59(2021), 247-289 doi:10.1146/annurev-astro-120920-010024 [arXiv:2101.11735 [astro-ph.CO]]

    2021 arXiv

  23. [31]

    D. F. J. Kimball and K. van Bibber,The Search for Ultralight Bosonic Dark Matter, Springer, 2023 [ISBN:978-3-0309-5852-7]

    2023

  24. [32]

    D. J. C. McKay,Information Theory, Inference and Learning Algorithms, Cambrige University Press, 2003 [ISBN: 978-0-5216-4298-9]

    2003

  25. [33]

    Bernal, L

    T. Bernal, L. M. Fern´ andez-Hern´ andez, T. Matos and M. A. Rodr ´ ıguez-Meza,Rotation curves of high-resolution LSB and SPARC galaxies with fuzzy and multistate (ultralight boson) scalar field dark matter,Mon. Not. Roy. Astron. Soc.475(2018) no.2, 1447-1468 doi:10.1093/mnras...

    2018 arXiv

  26. [34]

    Delgado and A

    V. Delgado and A. Mu˜ noz Mateo,Self-interacting superfluid dark matter droplets,Mon. Not. Roy. Astron. Soc.518(2022) no.3, 4064-4072 doi:10.1093/mnras/stac3386 [arXiv:2201.12418 [astro-ph.CO]]

    2022 arXiv

  27. [35]

    Khelashvili, A

    M. Khelashvili, A. Rudakovskyi and S. Hossenfelder,Dark matter profiles of SPARC galaxies: a challenge to fuzzy dark matter,Mon. Not. Roy. Astron. Soc.523(2023) no.3, 3393-3405 doi:10.1093/mnras/stad1595 [arXiv:2207.14165 [astro-ph.CO]]

    2023 arXiv

  28. [36]

    Ba˜ nares-Hern´ andez, A

    A. Ba˜ nares-Hern´ andez, A. Castillo, J. Martin Camalich and G. Iorio,Confronting fuzzy dark matter with the rotation curves of nearby dwarf irregular galaxies,Astron. Astrophys.676 (2023), A63 doi:10.1051/0004-6361/202346686 [arXiv:2304.05793 [astro-ph.GA]]

    2023 arXiv

  29. [37]

    J. F. Navarro, C. S. Frenk and S. D. M. White,The Structure of cold dark matter halos, Astrophys. J.462(1996), 563-575 doi:10.1086/177173 [arXiv:astro-ph/9508025 [astro-ph]]

    1996 arXiv

  30. [38]

    Chakrabarti, B

    S. Chakrabarti, B. Dave, K. Dutta and G. Goswami,Constraints on the mass and self-coupling of ultra-light scalar field dark matter using observational limits on galactic central mass,JCAP 09(2022), 074 doi:10.1088/1475-7516/2022/09/074 [arXiv:2202.11081 [astro-ph.CO]]

    2022 arXiv

  31. [39]

    Dave and G

    B. Dave and G. Goswami,Self-interactions of ULDM to the rescue?,JCAP07(2023), 015 doi:10.1088/1475-7516/2023/07/015 [arXiv:2304.04463 [astro-ph.CO]]

    2023 arXiv

  32. [40]

    N. Bar, K. Blum and C. Sun,Galactic rotation curves versus ultralight dark matter: A systematic comparison with SPARC data,Phys. Rev. D105(2022) no.8, 083015 doi:10.1103/PhysRevD.105.083015 [arXiv:2111.03070 [hep-ph]]

    2022 arXiv

  33. [41]

    ´A. O. F. de Almeida, L. Amendola and V. Niro,Galaxy rotation curves in modified gravity models,JCAP08(2018), 012 doi:10.1088/1475-7516/2018/08/012 [arXiv:1805.11067 [astro-ph.GA]]

    2018 arXiv

  34. [42]

    Bhatia, S

    E. Bhatia, S. Chakrabarti and S. Chakraborty,Phenomenology of renormalization group improved gravity from the kinematics of SPARC galaxies,Phys. Rev. D110(2024) no.12, 124014 doi:10.1103/PhysRevD.110.124014 [arXiv:2403.00531 [gr-qc]]

    2024 arXiv

  35. [43]

    Benisty, D

    D. Benisty, D. Vasak, J. Struckmeier and H. Stoecker,Bounding the cosmological constant using galactic rotation curves from SPARC dataset,Phys. Rev. D110(2024) no.6, 063028 doi:10.1103/PhysRevD.110.063028 [arXiv:2405.16650 [astro-ph.CO]]

    2024 arXiv

  36. [44]

    Strang,Linear Algebra and Learning from Data, Wellesley Cambrige Press, 2018 [ISBN: 978-0-6921-9638-0]

    G. Strang,Linear Algebra and Learning from Data, Wellesley Cambrige Press, 2018 [ISBN: 978-0-6921-9638-0]

    2018

  37. [45]

    Hornik, M

    K. Hornik, M. Stinchcombe, H. White,Universal approximation of an unknown mapping and its derivatives using multilayer feedforward networks,Neural Networks3(1990) 5, 551560 doi:10.1016/0893-6080(90)90005-6

    1990 doi

  38. [46]

    S. J. Prince,Understanding Deep Learning, The MIT Press, 2023

    2023

  39. [47]

    Mhaskar, Q

    H. Mhaskar, Q. Liao, and T. PoggioWhen and why are deep networks better than shallow ones?,In Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence(2017) 23432349

    2017

  40. [48]

    Y. Meir, O. Tevet, Y. Tzach, S. Hodassman, R. D. Gross and I. Kanter,Efficient shallow learning as an alternative to deep learning Sci. Rep.13(2023) 5423 doi:10.1038/s41598-023-32559-8 [arXiv:2211.11106 [cs.LG]]

    2023 arXiv

  41. [49]

    D. E. Rumelhart, G. E. Hinton, and R. J. Williams,Learning representations by back-propagating errors,Nature323, (1986) 533

    1986

  42. [50]

    Eisert, A

    L. Eisert, A. Pillepich, D. Nelson, R. S. Klessen, M. Huertas-Company, and V. Rodriguez-Gomez,ERGO-ML I: Inferring the assembly histories of IllustrisTNG galaxies from integral observable properties via invertible neural networks,Mon. Not. Roy. Astron. Soc. 519(2023) no.2, 219...

    2023 arXiv

  43. [51]

    D. P. Kingma and J. Ba,Adam: A Method for Stochastic Optimization, [arXiv:1412.6980 [cs.LG]]

    arXiv

  44. [52]

    Srivastava, G

    N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov,Dropout: A simple way to prevent neural networks from overfitting,Journal of Machine Learning Research 15(2014) 1929

    2014

  45. [53]

    C. M. Bishop,Training with Noise is Equivalent to Tikhonov Regularization,Neural Computation,7, (1995) 1, 108-116 doi:10.1162/neco.1995.7.1.108

    1995 doi

  46. [54]

    Kendall and Y

    A. Kendall and Y. Gal,What uncertainties do we need in Bayesian deep learning for computer vision? In Proceedings of the 31st International Conference on Neural Information Processing Systems (NIPS’17) Curran Associates Inc.(2017) 55805590

    2017

  47. [55]

    Fluri, T

    J. Fluri, T. Kacprzak, A. Lucchi, A. Refregier, A. Amara, T. Hofmann and A. Schneider, Cosmological constraints with deep learning from KiDS-450 weak lensing maps,Phys. Rev. D 100(2019) no.6, 063514 doi:10.1103/PhysRevD.100.063514 [arXiv:1906.03156 [astro-ph.CO]]

    2019 arXiv

  48. [56]

    Andr´ es-Carcasona, M

    M. Andr´ es-Carcasona, M. Martinez and L. M. Mir,Fast Bayesian gravitational wave parameter estimation using convolutional neural networks,Mon. Not. Roy. Astron. Soc.527(2023) no.2, 2887-2894 doi:10.1093/mnras/stad3448 [arXiv:2309.04303 [gr-qc]]

    2023 arXiv

  49. [57]

    Dvorkin, S

    C. Dvorkin, S. Mishra-Sharma, B. Nord, V. A. Villar, C. Avestruz, K. Bechtol, A. ´Ciprijanovi´ c, A. J. Connolly, L. H. Garrison and G. Narayan,et al. Machine Learning and Cosmology, [arXiv:2203.08056 [hep-ph]]

    arXiv

  50. [58]

    M. A. Amin, F. Y. Cyr-Racine, T. Eifler, R. Flauger, M. M. Ivanov, M. LoVerde, C. B. d. S. Nascimento, A. H. G. Peter, M. Vogelsberger and S. Watson,et al. Snowmass2021 Theory Frontier White Paper: Data-Driven Cosmology, [arXiv:2203.07946 [astro-ph.CO]]

    arXiv

  51. [59]

    H. J. Horta, R. Volpi, D. Marinelli, and L. Malag,Parameter estimation for the cosmic microwave background with bayesian neural networks,Phys. Rev. D102, (2020) 103509 doi:10.1103/PhysRevD.102.103509 [arXiv:1911.08508 [astro-ph.IM]]

    2020 arXiv

  52. [60]

    H. J. Hort´ ua, L. ´A. Garc ´ ıa and L. Casta˜ neda C.,Constraining cosmological parameters from N-body simulations with variational Bayesian neural networks,Front. Astron. Space Sci.10 (2023), 1139120 doi:10.3389/fspas.2023.1139120 [arXiv:2301.03991 [astro-ph.IM]]

    2023

  53. [61]

    Stachurski, C

    F. Stachurski, C. Messenger and M. Hendry,Cosmological inference using gravitational waves and normalizing flows,Phys. Rev. D109(2024) no.12, 123547 doi:10.1103/PhysRevD.109.123547 [arXiv:2310.13405 [gr-qc]]

    2024 arXiv

  54. [62]

    Kolmus, J

    A. Kolmus, J. Janquart, T. Baka, T. van Laarhoven, C. Van Den Broeck and T. Heskes,Tuning neural posterior estimation for gravitational wave inference, [arXiv:2403.02443 [astro-ph.IM]]

    arXiv

  55. [63]

    Di Valentinoet al.[CosmoVerse Network],The CosmoVerse White Paper: Addressing observational tensions in cosmology with systematics and fundamental physics,Phys

    E. Di Valentinoet al.[CosmoVerse Network],The CosmoVerse White Paper: Addressing observational tensions in cosmology with systematics and fundamental physics,Phys. Dark Univ.49(2025), 101965 doi:10.1016/j.dark.2025.101965 [arXiv:2504.01669 [astro-ph.CO]]. A Monitoring the loss...

    2025

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.