Pith. sign in

REVIEW 2 major objections 4 minor 102 references

A robust neural determination of the source-count distribution of the Fermi-LAT sky at high latitudes

T0 review · 2 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash

Pith's one-line read Using simulation-based inference with a Gaussian-random-field-perturbed foreground, this paper shows that 14 years of Fermi-LAT sky maps can be reduced to a calibrated source-count distribution and a nearly complete source list.

desk verdict Careful, useful SBI pipeline for Fermi-LAT dN/dS, but the simulator-realism test is too weak to support the paper's central robustness claim. read the letter →

arxiv 2505.02906 v1 pith:HTK75BXE submitted 2025-05-05 astro-ph.HE astro-ph.IM

classification astro-ph.HEastro-ph.IM
keywords simulation-basedinferenceneuralratioestimationsource-countdistributionFermi-LATgamma-raypoint-sourcedetectionGaussianrandomfieldshigh-latitudeskydiffusebackground
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 claims that simulation-based inference with neural ratio estimation can do what catalog-based point-source counting cannot: detect individual gamma-ray emitters and reconstruct the full source-count distribution dN/dS below the catalog completeness threshold, all from 14 years of high-latitude Fermi-LAT data. The method recovers more than 98% of unflagged 4FGL-DR4 sources above S=3e-10 $cm^{-2}$ $s^{-1}$ and finds no candidate above that flux that is missing from the catalog. It then infers dN/dS both parametrically as a triple-broken power law and non-parametrically in flux bins, with results consistent at the 1-$\sigma$ level and in agreement with earlier 1p-PDF analyses. The paper handles foreground uncertainty by perturbing the Milky Way diffuse template with Gaussian random fields during training, so the networks learn to marginalize over plausible diffuse-background variations. If the claims hold, the approach extends statistical gamma-ray source counting to fainter fluxes while returning calibrated uncertainties and a concrete list of source candidates.

What carries the argument

The machine that carries the argument is a fast, GPU-based forward simulator of binned all-sky photon-count maps at HEALPix resolution Nside = 128. It draws point sources from a parametric dN/dS given by a triple-broken power law, scatters their photons through an energy-averaged Fermi-LAT PSF by direct Monte Carlo, and adds PSF-smoothed templates for the Galactic diffuse emission, isotropic background, and the Magellanic Clouds. In the robust variant, the diffuse template is multiplied by the exponential of a Gaussian random field with power spectrum P(k) = (k/2.5)^(-gamma), so the network learns to marginalize over foreground uncertainty; this is the ingredient that reduces non-catalog candidates from 685 to 387. Inference is done by neural ratio estimation, in which a classifier learns the likelihood-to-evidence ratio, with a U-Net on a spherical graph for pixel-wise source detection and autoregressive ratio estimators for the dN/dS parameters.

What would settle it

Take the pipeline and run it on a simulated sky where the faint source population below about S~1e-11 $cm^{-2}$ $s^{-1}$ is spatially clustered on degree scales instead of isotropic; if the Deep SVDD anomaly score still accepts the map while the inferred dN/dS or the candidate list moves outside the quoted credible intervals, the model-space assumption fails.

Watch

Extended reading notes

Core claim

The paper's central claim, stated on its own terms, is that a neural simulator trained on a realistically perturbed forward model closes the reality gap for high-latitude gamma-ray inference. Applied to 14 years of Fermi-LAT data with |b| >= 30 degrees in the 1-10 GeV band, the pipeline detects point sources through a pixel-wise neural ratio and recovers about 98% of all unflagged 4FGL-DR4 sources above S=3e-10 $cm^{-2}$ $s^{-1}$, roughly 70% over the full flux range, while keeping the number of non-catalog candidates to a few hundred and none above that flux. The same framework infers the source-count distribution in a triple-broken-power-law form and in 20 independent flux bins; the two reconstructions agree at the 1-$\sigma$ level, overlap the catalog's implied distribution, and match the 1p-PDF results of earlier likelihood analyses more closely when the diffuse Milky Way foreground is distorted by Gaussian random fields during training. The paper declares the non-parametric, GRF-trained dN/dS profile to be its best estimate for the high-latitude source-count distribution.

Load-bearing premise

The load-bearing premise is that the simulator's model space contains the real high-latitude Fermi-LAT sky, so a network trained only on simulated maps generalizes to the actual data; the anomaly test only shows the real map is not a statistical outlier, not that the model space is complete.

Editorial extensions

If this is right

  • Above S=3e-10 cm^-2 s^-1, the SBI detector reaches about 98% completeness relative to the 4FGL-DR4 catalog, and no non-catalog candidate appears at that flux, so the bright end of the dN/dS is consistent with the catalog.
  • The inferred parametric and non-parametric dN/dS agree with each other at the 1-sigma level and with the 1p-PDF results from the literature across the full flux range when GRFs are included, supporting a downturn in the number of faint sources below S~1e-10 cm^-2 s^-1.
  • Because the method yields amortized and well-calibrated posteriors, it can be extended to multiple energy bins, source spectra, and source classification without changing the inference principle.
  • Training with GRF-perturbed foregrounds reduces the non-catalog candidate population from 685 to 387, implying that part of the candidate excess in the no-GRF analysis was foreground fluctuation rather than genuine point sources.
  • The pipeline recovers about 70% of all unflagged 4FGL-DR4 sources over the full flux range, so it does not yet reach catalog-level sensitivity for faint sources; the paper attributes part of this gap to the catalog's own multi-step analysis pipeline.

Reading between the lines

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

  • If the central claim is right, the unresolved point-source fraction of the isotropic gamma-ray background below S~1e-10 cm^-2 s^-1 would be smaller than the original 1p-PDF estimates, a difference that could be tested with a blazar luminosity-function decomposition.
  • A testable extension the paper leaves implicit: the same GRF machinery could be replaced by a physically motivated gas-template alternative, and the inferred GRF parameters would then become measurements of missing or dark gas rather than effective distortions.
  • The Deep SVDD anomaly test is necessary but not sufficient; injecting a spatially clustered faint-source population into the simulator and checking whether the anomaly score still accepts the real map would settle whether the model space is wide enough.
  • Because the pipeline is amortized, rerunning it on independent early Fermi-LAT data selections or with a different PSF realization would provide a public, reproducible check of the 98% recovery claim.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. The paper develops a simulation-based inference (SBI) pipeline based on neural ratio estimation with spherical (DeepSphere) convolutional networks, and applies it to 14 years of Fermi-LAT high-latitude (|b|>30°) data in the 1–10 GeV band. The forward simulator comprises a triple-broken-power-law point-source population with uniform sky positions, Galactic diffuse, isotropic, LMC, and SMC backgrounds, energy-averaged exposure and PSF treatments, and optional Gaussian-random-field (GRF) distortions of the Galactic diffuse template applied after PSF convolution. Three inference tasks are performed: pixel-wise point-source detection with a U-Net; parametric inference of the TBPL and background parameters via autoregressive NRE with nested sampling; and non-parametric inference of the binned S^2 dN/dS. The main claims are: recovery of ~98% of un-flagged 4FGL-DR4 sources above S=3×10^-10 cm^-2 s^-1 with no non-4FGL candidate above that flux (for the GRF-trained detector); parametric and non-parametric dN/dS profiles consistent with each other and with the 1p-PDF literature results [13,17]; validated reconstruction on simulated data including the dim-flux regime; coverage tests showing generally well-calibrated posteriors; and a Deep SVDD anomaly test indicating that the real sky lies inside the simulator's model space. The non-parametric GRF-inclusive dN/dS is declared the best estimate (Sec. VII).

Significance. If correct, this is a substantial methodological advance: it is among the first applications of ML-based source detection and SBI parameter inference to a large ROI of genuine Fermi-LAT data rather than simulations, it reaches fluxes about a factor of three below the 4FGL completeness limit without producing bright spurious detections, and the reconstructed dN/dS agrees with independent likelihood-based 1p-PDF analyses. The paper is unusually thorough in its validation: simulated-data recovery with known ground truth (App. D), posterior calibration via coverage tests on 1000 samples (App. E1), explicit mis-modeling tests (App. F), and a fast simulator built on public Fermi Science Tools. The headline claims are falsifiable: the candidate source list and the absence of bright non-4FGL candidates can be checked by dedicated follow-up observations, and the declared "best estimate" dN/dS can be compared against future catalog-based measurements. These strengths make the paper a likely reference point for SBI in gamma-ray astrophysics, provided the reality-gap validation is strengthened as detailed in the major comments.

major comments (2)
  1. [Sec. III D, App. E2, App. F1, Sec. VII] The paper's central robustness claim — that training with GRF-augmented backgrounds protects the dN/dS and source-list inference against diffuse mis-modeling, culminating in the declaration of the non-parametric GRF-inclusive dN/dS as the best estimate (Sec. VII) — is not established for the small-scale regime by the validation actually presented. Sec. III D states that the GRF is multiplied into the Galactic diffuse template after PSF convolution, so the simulator can only produce unsmeared small-scale diffuse fluctuations, whereas the physical mis-modeling it is meant to absorb (gas clumps, template errors, dark gas) would be smoothed by the LAT PSF before appearing in the count maps. The two quantitative checks do not close this gap: the Deep SVDD anomaly test (App. E2) places the real 14-year data inside the simulated score distribution, but it equally places a sky generated with the FGMA diffuse template inside that distribution, and in the GRF-trained case even a pure Poisson noise map moves closer to the simulated distribution; both observations indicate that the SVDD score is insensitive to the specific kind of mis-modeling that would bias dN/dS, because the GRFs inflate the variance of the model space at all scales. App. F1's mis-modeling test only replaces the diffuse template by a large-scale variant missing the Fermi Bubbles and Loop I; it does not exercise the small-scale, pre-PSF regime. I recommend adding a closed-loop stress test in which diffuse mis-modeling is injected before PSF convolution (e.g., GRF perturbations applied to the unsmoothed template, or compact unresolved structures outside the GRF family), followed by verification that the GRF-trained parametric and non-parametric networks recover the injected dN/dS without bias; this test is feasible within the existing simulator framework and directly targets the weakest assumption flagged by the authors themselves.
  2. [App. E1 (Fig. 20), Sec. VII] The abstract and Sec. VII present "well-calibrated posterior distributions" as a headline achievement, but the coverage tests behind the declared best estimate show systematic overconfidence. In App. E1, Fig. 20 (non-parametric inference with GRFs) gives empirical 1σ coverages of 58–62% against the nominal 68% for the bright flux bins θS,18–θS,20, similarly reduced coverage for several dim bins (θS,15–17), and slight overconfidence for Aiso (63.5% at nominal 68%). Because the non-parametric GRF profile is the result the paper recommends for further use, the quoted credible intervals for that profile are narrower than their stated calibration; this should be either fixed (e.g., by increasing the training sample, as the authors suggest in App. E1) or explicitly disclosed wherever the profile is presented, including in the abstract's calibration claim.
minor comments (4)
  1. [Sec. III B a, Tab. II] The TBPL prior ranges are explicitly motivated by the 4FGL catalog values (Sec. III B a), so the agreement between the inferred dN/dS and the 4FGL histogram (Figs. 7–9) is partly by construction; the decisive independent cross-checks are the 1p-PDF results [13,17], and this asymmetry should be stated where the 4FGL agreement is invoked.
  2. [Sec. V A, Figs. 5, 11] The fluxes of non-4FGL candidates are computed as pixel counts minus best-fit background divided by exposure, with no uncertainty or systematic caveat attached; since these fluxes control the headline statement that no candidate exceeds 3×10^-10 cm^-2 s^-1, a one-sentence caveat on the uncertainty of this estimate would prevent over-interpretation.
  3. [Data Availability] The data availability statement promises only figure products "upon reasonable request"; given that the paper's core contribution is a validated SBI pipeline, releasing the trained networks, simulator code, or posterior samples would substantially strengthen reproducibility.
  4. [Sec. VI] The attribution of the dim-regime discrepancy with [28] to their fixed Galactic diffuse normalization is plausible, but it would be more persuasive if accompanied by the closed-loop test proposed in major comment 1, since the present analysis's own sensitivity to diffuse mis-modeling is precisely the point at issue.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the dN/dS inference is data-driven, trained on simulated maps without 4FGL labels, and validated against external literature and catalogs.

full rationale

The central inference is genuinely simulation-based: in every training sample, point-source fluxes and positions are drawn from the TBPL priors, and the neural networks learn ratios from simulated photon-count maps without any use of 4FGL labels. The real-data dN/dS posteriors come from applying those trained networks to the LAT map, not from re-inserting the catalog. Priors are wide and only loosely anchored to 4FGL values, and the non-parametric analysis explicitly drops the TBPL functional form, so the agreement with the catalog and with the independent 1p-PDF results of Refs. [13,17] is an empirical cross-check rather than a construction. The source-detection claim is a recovery/completeness statement against 4FGL, with the paper explicitly noting that 4FGL is assumed as ground truth only for evaluation; the detection network itself was trained on simulated sources only. The robustness of the GRF-enhanced simulator is supported by closed-loop recovery tests on simulated data with known ground truth, Bayesian coverage tests, explicit mis-modeling stress tests in App. F, and a Deep SVDD anomaly check that treats the real LAT map as a held-out sample relative to the simulator-trained network. The acknowledged approximation of applying GRFs after PSF convolution (Sec. III.D) is a modeling limitation that the authors disclose and motivate, but it does not identify any inferred quantity with an input by construction. No load-bearing step reduces to a fitted parameter renamed as a prediction, and no uniqueness or validity claim is imported solely from the authors' own prior work. The paper therefore exhibits no circular derivation chain.

Assumptions & free parameters 14 free parameters · 8 assumptions · 0 invented entities

The model introduces 14 free parameters (8 TBPL, 4 background normalizations, 2 GRF parameters), all inferred from data with broad priors. It relies on several domain assumptions about the sky model and simulator fidelity. No new physical entities are introduced; the GRF is a statistical nuisance model. The central claim depends on the simulator covering the real sky, an assumption only indirectly validated.

free parameters (14)
  • AS (dN/dS normalization) = 2.5 (posterior median, GRF case)
    Overall normalization of the triple-broken power law; inferred from the LAT data, central to the dN/dS result.
  • n1 (first slope) = 3.0
    Slope of the brightest segment; fitted to the source-count distribution.
  • n2 (second slope) = 1.9
    Intermediate-flux slope; fitted to the data.
  • n3 (third slope) = 1.8
    Intermediate-to-dim slope; fitted to the data.
  • n4 (last slope) = 0.0
    Dim-flux slope; weakly constrained, fitted with wide posterior.
  • Sb1 (first break flux) = 1.3e-8 cm^-2 s^-1
    Break between brightest segments; fitted.
  • Sb2 (second break flux) = 3.8e-10 cm^-2 s^-1
    Break near the completeness limit; fitted.
  • Sb3 (third break flux) = 4.2e-12 cm^-2 s^-1
    Break in the dim regime; fitted with wide uncertainty.
  • Adiff (Galactic diffuse normalization) = 0.81
    Scales the official Galactic diffuse template; inferred along with dN/dS.
  • Aiso (isotropic background normalization) = 0.9
    Scales the isotropic template; inferred and partly degenerate with the dim dN/dS.
  • ALMC (LMC normalization) = 0.85
    Scales the LMC template; inferred nuisance parameter.
  • ASMC (SMC normalization) = 0.56
    Scales the SMC template; loosely constrained, especially in non-parametric inference.
  • AGRF (GRF amplitude) = 2.0
    Amplitude of Gaussian random field modulations of the diffuse foreground; inferred in the GRF scenario.
  • gamma (GRF power-spectrum slope) = 2.4
    Spectral index of the GRF power spectrum; inferred and interpreted as foreground mis-modeling diagnostic.
assumptions (8)
  • domain assumption The high-latitude source-count distribution is well described by a triple-broken power law with parameters in the chosen prior ranges.
    Used for the parametric inference and to generate training data. The non-parametric method relaxes the functional form, but the simulator's training distribution still follows the TBPL.
  • domain assumption Point-like gamma-ray sources are uniformly distributed on the sky at |b|>30 degrees.
    Sec. III B b. This ignores possible large-scale structure or clustering of Galactic source populations; the authors argue known classes are compatible with uniformity.
  • domain assumption The combination of the Fermi diffuse model (gll_iem_v078), isotropic template, LMC/SMC templates, global normalizations, and GRF modulations spans the real high-latitude sky.
    The fidelity of all SBI posteriors depends on the simulator covering the true data distribution; validated only indirectly through anomaly detection and one alternative foreground test.
  • ad hoc to paper Gaussian random field distortions with power-law power spectrum, gamma in [1.5,3] and AGRF in [0,5], can absorb unresolved diffuse mis-modeling without biasing the source-count inference.
    Sec. III D. The GRF model and its prior ranges are phenomenological, chosen for computational convenience and flexibility, not derived from physics.
  • domain assumption Energy dispersion can be neglected in the single 1-10 GeV energy bin, affecting only a global rescaling of component fluxes.
    Sec. III A c. The authors argue high-latitude morphologies are smooth, but the approximation is not rigorously tested for all components.
  • domain assumption The energy-averaged PSF computed at (l,b)=(90,40) is representative for the whole ROI and is radially symmetric.
    Sec. III A b. They verify mild dependence on position but adopt a single effective PSF for all sources.
  • domain assumption Source fluxes are simulated only down to 1e-13 cm^-2 s^-1, with a single power-law extrapolation below; fainter sources are neglected.
    Sec. III B a. The behavior of the dN/dS below this flux is unmodeled and could affect the isotropic background degeneracy.
  • standard math Neural ratio estimation with the chosen architectures and training set sizes converges to the true likelihood ratio.
    Standard NRE result; finite network capacity and finite training data introduce approximations, partially validated by coverage tests.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A robust neural determination of the source-count distribution of the Fermi-LAT sky at high latitudes." pith.science (2026). https://pith.science/paper/HTK75BXE

@misc{pith2026250502906,
  author       = {Pith},
  title        = {Pith review of: A robust neural determination of the source-count distribution of the Fermi-LAT sky at high latitudes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HTK75BXE}},
  note         = {Machine review of arXiv:2505.02906}
}
abstract

Over the past 16 years, the Fermi Large Area Telescope (LAT) has significantly advanced our view of the GeV gamma-ray sky, yet several key questions remain - such as the composition of the isotropic gamma-ray background, the origin of the Fermi Bubbles or the potential presence of signatures from exotic physics like dark matter. Addressing these challenges requires sophisticated astrophysical modeling and robust statistical methods capable of handling high-dimensional parameter spaces. In this work, we analyze 14 years of high-latitude ($|b|\geq30^{\circ}$) Fermi-LAT data in the range from 1 to 10 GeV using simulation-based inference (SBI) via neural ratio estimation. This approach allows us to detect individual gamma-ray sources and derive a list of significant gamma-ray emitters containing more than 98\% of all sources listed in the Fermi-LAT Fourth Source Catalog (4FGL) with a flux $S>3\times10^{-10}\;\mathrm{cm}^{-2}\,\mathrm{s}^{-1}$ (about a factor of three larger than the flux above which 4FGL is nearly complete), without any non-4FGL source detected in that flux range. Additionally, we reconstruct the source-count distribution in both parametric and non-parametric forms, achieving large agreement with previous literature results as well as those sources detected by our SBI pipeline. We also quantitatively validate our gamma-ray emission simulator via an anomaly detection technique, demonstrating that the synthetic data closely reproduces the complexity of the real observations.

Figures

Figures reproduced from arXiv: 2505.02906 by the authors.

Figure 1
Figure 1. FIG. 1. High-latitude ( [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Comparison of the PSF profiles considered in this [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. Schematic overview of the Bayesian hierarchical model implemented in our gamma-ray simulator. The definitions for [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (24 more)
Figure 4
Figure 4. Figure 4: FIG. 4. Detection map ROC and threshold selection (see text for description). [PITH_FULL_IMAGE:figures/full_fig_p016_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5. Comparison of the source-count distribution of de [PITH_FULL_IMAGE:figures/full_fig_p017_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6 [PITH_FULL_IMAGE:figures/full_fig_p018_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7. Comparison of the differential source-count distribution d [PITH_FULL_IMAGE:figures/full_fig_p020_7.png]
Figure 8
Figure 8. Figure 8: FIG. 8. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p022_8.png]
Figure 9
Figure 9. Figure 9: FIG. 9. Comparison of the source-count distribution of the [PITH_FULL_IMAGE:figures/full_fig_p023_9.png]
Figure 10
Figure 10. Figure 10: FIG. 10. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p030_10.png]
Figure 11
Figure 11. Figure 11: FIG. 11. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p032_11.png]
Figure 12
Figure 12. Figure 12: FIG. 12. Comparison of the posterior probability density functions for the four background components of our gamma-ray [PITH_FULL_IMAGE:figures/full_fig_p033_12.png]
Figure 13
Figure 13. Figure 13: FIG. 13. Corner plot of the two-dimensional marginal posterior distributions obtained from applying the trained network of [PITH_FULL_IMAGE:figures/full_fig_p035_13.png]
Figure 14
Figure 14. Figure 14: Based on the intuition we get from applying the detection network on simulated data of the same un￾derlying dN/dS (1p-PDF best fit, see the right panel of [PITH_FULL_IMAGE:figures/full_fig_p036_14.png]
Figure 15
Figure 15. Figure 15: FIG. 15. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p037_15.png]
Figure 16
Figure 16. Figure 16: FIG. 16. Examples of [PITH_FULL_IMAGE:figures/full_fig_p039_16.png]
Figure 17
Figure 17. Figure 17: FIG. 17. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p040_17.png]
Figure 18
Figure 18. Figure 18: FIG. 18 [PITH_FULL_IMAGE:figures/full_fig_p042_18.png]
Figure 19
Figure 19. Figure 19: FIG. 19. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p043_19.png]
Figure 20
Figure 20. Figure 20: FIG. 20. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p044_20.png]
Figure 21
Figure 21. Figure 21: FIG. 21. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p045_21.png]
Figure 22
Figure 22. Figure 22: FIG. 22. Averaged SVDD score prediction [PITH_FULL_IMAGE:figures/full_fig_p047_22.png]
Figure 24
Figure 24. Figure 24: FIG. 24. Comparison of inferred mean source-count distribu [PITH_FULL_IMAGE:figures/full_fig_p048_24.png]
Figure 23
Figure 23. Figure 23: FIG. 23. Summary of the counts maps of the Galactic dif [PITH_FULL_IMAGE:figures/full_fig_p048_23.png]
Figure 25
Figure 25. Figure 25: FIG. 25 [PITH_FULL_IMAGE:figures/full_fig_p049_25.png]
Figure 26
Figure 26. Figure 26: FIG. 26. Examples of non-parametric inference results on test simulated data featuring a five-times broken source-count [PITH_FULL_IMAGE:figures/full_fig_p051_26.png]
Figure 27
Figure 27. Figure 27: FIG. 27. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p052_27.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

102 extracted references · 23 canonical work pages

  1. [1]

    7, we confronted the para- metric and non-parametric d N/dS inference results on real data with similar findings obtained with the 1p-PDF technique

    Recovery of the 1p-PDF best-fit dN/dS In the main text’s Fig. 7, we confronted the para- metric and non-parametric d N/dS inference results on real data with similar findings obtained with the 1p-PDF technique. A common observation between our two SBI inference methods was a reduced number of dim sources in comparison with the 1p-PDF best-fit TBPL (which ...

  2. [2]

    In this section, we showcase the obtained inference results for the large variety of dN/dS profiles allowed by the prior distribu- tions defined in Tab

    Performance on simulated sky maps In the previous section, we focused on a single dN/dS parameter realization and explored the perfor- mance of our parametric and non-parametric SBI ap- proach to recover these injected values. In this section, we showcase the obtained inference results for the large variety of dN/dS profiles allowed by the prior distribu-...

  3. [3]

    In the main text, we did not elaborate on all the information that we can extract from the autoregressive NRE approach and the paramet- ric dN/dS inference

    Parameter correlations in the parametric dN/dS inference approach Results without GRFs. In the main text, we did not elaborate on all the information that we can extract from the autoregressive NRE approach and the paramet- ric dN/dS inference. It provides us with the full joint posterior distributions and, thus, we can inspect the cor- relations among di...

  4. [4]

    Bayesian coverage of the trained networks In Bayesian statistics, it is possible to check the cali- bration of the method and of the posterior distributions it yields via so-called coverage tests. With coverage tests, we aim to verify that the obtained α% credible inter- vals Θˆp(ζ|d)(α) of the posteriors ˆp(ζ|d) 21 cover the true value of the respective ...

  5. [5]

    III that our gamma-ray simulator was designed in the spirit of generating realistic sam- ples of gamma-ray emission at high latitudes

    Assessing the simulator’s realism with anomaly detection We stressed in Sec. III that our gamma-ray simulator was designed in the spirit of generating realistic sam- ples of gamma-ray emission at high latitudes. While the results reported in Sec. V align well with previous litera- ture results – hence, they do not cast striking doubts on the accuracy of t...

  6. [6]

    The average SVDD is based on these re-normalized values. In the case w/o GRFs, we find that the real data is well within the variation expected from the simulator output, while the noise map is clearly an outlier that cannot be generated from the simulator. Interestingly, the sample featuring an alternative MW diffuse foreground model performs equally wel...

  7. [7]

    The un- certainties on these quantities will propagate into the predictions of the expected gamma-ray emission

    Alternative description of the Galactic diffuse emission As stressed in different places, the Galactic diffuse emission originates in high-energy charged cosmic rays colliding with the interstellar medium, thus tracing the structure and distribution of gas and ISRFs. The un- certainties on these quantities will propagate into the predictions of the expect...

  8. [8]

    J.-Q. Xia, A. Cuoco, E. Branchini and M. Viel, Tomography of the Fermi-lat γ-ray Diffuse Extragalactic Signal via Cross Correlations With Galaxy Catalogs , Astrophys. J. Suppl. 217 (2015) 15 [ 1503.05918]

Show all 102 references
  1. [9]

    As noted in [13], there is no evidence for the necessity of even three breaks of the d N/dS

    Shape of the source-count distribution Throughout this study, we worked with a simulator that generates high-latitude source populations follow- ing a TBPL source-count distribution. As noted in [13], there is no evidence for the necessity of even three breaks of the d N/dS. Y...

  2. [10]

    Fornasa and M

    M. Fornasa and M. A. S´ anchez-Conde,The nature of the Diffuse Gamma-Ray Background , Physics Reports 598 (2015) 1 [ 1502.02866]

  3. [11]

    Ackermann et al., Fermi Large Area Telescope Detection of Extended Gamma-Ray Emission from the Radio Galaxy Fornax A, Astrophys

    Fermi-LAT Collaboration, M. Ackermann et al., Fermi Large Area Telescope Detection of Extended Gamma-Ray Emission from the Radio Galaxy Fornax A, Astrophys. J. 826 (2016) 1 [ 1606.04905]

  4. [12]

    Ackermann et al., Anisotropies in the diffuse gamma-ray background measured by the Fermi LAT , Phys

    M. Ackermann et al., Anisotropies in the diffuse gamma-ray background measured by the Fermi LAT , Phys. Rev. D 85 (2012) 083007 [ 1202.2856]

  5. [13]

    Fornasa et al., Angular power spectrum of the diffuse gamma-ray emission as measured by the Fermi Large Area Telescope and constraints on its dark matter interpretation, Phys

    M. Fornasa et al., Angular power spectrum of the diffuse gamma-ray emission as measured by the Fermi Large Area Telescope and constraints on its dark matter interpretation, Phys. Rev. D 94 (2016) 123005 [1608.07289]

  6. [14]

    S. Ando, M. Fornasa, N. Fornengo, M. Regis and H.-S. Zechlin, Astrophysical interpretation of the anisotropies in the unresolved gamma-ray background , Phys. Rev. D 95 (2017) 123006 [ 1701.06988]

  7. [15]

    Ackermann et al., Unresolved Gamma-Ray Sky through its Angular Power Spectrum, Phys

    Fermi-LAT Collaboration, M. Ackermann et al., Unresolved Gamma-Ray Sky through its Angular Power Spectrum, Phys. Rev. Lett. 121 (2018) 241101 [1812.02079]

  8. [16]

    J.-Q. Xia, A. Cuoco, E. Branchini, M. Fornasa and M. Viel, A cross-correlation study of the Fermi-LAT γ-ray diffuse extragalactic signal , Mon. Not. Roy. Astron. Soc. 416 (2011) 2247 [ 1103.4861]

  9. [17]

    Manconi, M

    S. Manconi, M. Korsmeier, F. Donato, N. Fornengo, M. Regis and H. Zechlin, Testing gamma-ray models of blazars in the extragalactic sky , Phys. Rev. D 101 (2020) 103026 [ 1912.01622]

  10. [18]

    Branchini, S

    E. Branchini, S. Camera, A. Cuoco, N. Fornengo, M. Regis, M. Viel and J.-Q. Xia, Cross-correlating the γ-ray sky with Catalogs of Galaxy Clusters , Astrophys. J. Suppl. 228 (2017) 8 [ 1612.05788]

  11. [19]

    Cuoco, M

    A. Cuoco, M. Bilicki, J.-Q. Xia and E. Branchini, Tomographic imaging of the Fermi-LAT gamma-ray sky through cross-correlations: A wider and deeper look , Astrophys. J. Suppl. 232 (2017) 10 [ 1709.01940]

  12. [20]

    Ammazzalorso, N

    S. Ammazzalorso, N. Fornengo, S. Horiuchi and M. Regis, Characterizing the local gamma-ray Universe via angular cross-correlations, Phys. Rev. D 98 (2018) 103007 [1808.09225]

  13. [21]

    Ballet, P

    Fermi-LAT Collaboration, J. Ballet, P. Bruel, T. H. Burnett and B. Lott, Fermi Large Area Telescope Fourth Source Catalog Data Release 4 (4FGL-DR4) , 2307.12546

  14. [22]

    Zechlin, A

    H.-S. Zechlin, A. Cuoco, F. Donato, N. Fornengo and A. Vittino, Unveiling the Gamma-ray Source Count Distribution Below the Fermi Detection Limit with Photon Statistics, Astrophys. J. Suppl. 225 (2016) 18 [1512.07190]

  15. [23]

    Zechlin, A

    H.-S. Zechlin, A. Cuoco, F. Donato, N. Fornengo and M. Regis, Statistical Measurement of the Gamma-ray Source-count Distribution as a Function of Energy , Astrophys. J. Lett. 826 (2016) L31 [ 1605.04256]

  16. [24]

    S. K. Lee, M. Lisanti and B. R. Safdi, Distinguishing Dark Matter from Unresolved Point Sources in the Inner Galaxy with Photon Statistics , JCAP 05 (2015) 056 [1412.6099]

  17. [25]

    Malyshev and D

    D. Malyshev and D. W. Hogg, Statistics of gamma-ray point sources below the Fermi detection limit , Astrophysical Journal 738 (2011) 1

  18. [26]

    Manconi, F

    S. Manconi, F. Calore and F. Donato, Galactic Center excess at the highest energies: Morphology and photon-count statistics, Phys. Rev. D 109 (2024) 123042 [2402.04733]

  19. [27]

    Lisanti, S

    M. Lisanti, S. Mishra-Sharma, L. Necib and B. R. Safdi, Deciphering Contributions to the Extragalactic Gamma-Ray Background from 2 GeV to 2 TeV , Astrophys. J. 832 (2016) 117 [ 1606.04101]

  20. [28]

    J. J. Somalwar, L. J. Chang, S. Mishra-Sharma and M. Lisanti, Harnessing the Population Statistics of Subhalos to Search for Annihilating Dark Matter , Astrophys. J. 906 (2021) 57 [ 2009.00021]

  21. [29]

    Zechlin, S

    H.-S. Zechlin, S. Manconi and F. Donato, Constraining Galactic dark matter with gamma-ray pixel counts statistics, Phys. Rev. D 98 (2018) 083022 [ 1710.01506]

  22. [30]

    R. K. Leane and T. R. Slatyer, Revival of the Dark Matter Hypothesis for the Galactic Center Gamma-Ray Excess, Phys. Rev. Lett. 123 (2019) 241101 [1904.08430]

  23. [31]

    L. J. Chang, S. Mishra-Sharma, M. Lisanti, M. Buschmann, N. L. Rodd and B. R. Safdi, Characterizing the nature of the unresolved point sources in the Galactic Center: An assessment of systematic uncertainties, Phys. Rev. D 101 (2020) 023014 [1908.10874]

  24. [32]

    Buschmann, N

    M. Buschmann, N. L. Rodd, B. R. Safdi, L. J. Chang, S. Mishra-Sharma, M. Lisanti and O. Macias, Foreground Mismodeling and the Point Source Explanation of the Fermi Galactic Center Excess , Phys. Rev. D 102 (2020) 023023 [ 2002.12373]

  25. [33]

    R. K. Leane and T. R. Slatyer, The enigmatic Galactic Center excess: Spurious point sources and signal mismodeling, Phys. Rev. D 102 (2020) 063019 [2002.12371]

  26. [34]

    Calore, F

    F. Calore, F. Donato and S. Manconi, Dissecting the Inner Galaxy with γ-Ray Pixel Count Statistics , Phys. Rev. Lett. 127 (2021) 161102 [ 2102.12497]

  27. [35]

    Caron, C

    S. Caron, C. Eckner, L. Hendriks, G. J´ ohannesson, R. Ruiz de Austri and G. Zaharijas, Mind the gap: the discrepancy between simulation and reality drives interpretations of the Galactic Center Excess , JCAP 06 (2023) 013 [ 2211.09796]

  28. [36]

    Cranmer, J

    K. Cranmer, J. Brehmer and G. Louppe, The frontier of simulation-based inference, Proc. Nat. Acad. Sci. 117 (2020) 30055 [ 1911.01429]

  29. [37]

    Amerio, A

    A. Amerio, A. Cuoco and N. Fornengo, Extracting the gamma-ray source-count distribution below the Fermi-LAT detection limit with deep learning , JCAP 09 (2023) 029 [ 2302.01947]

  30. [38]

    Mishra-Sharma and K

    S. Mishra-Sharma and K. Cranmer, Neural simulation-based inference approach for characterizing the Galactic Center γ-ray excess, Phys. Rev. D 105 (2022) 063017 [ 2110.06931]

  31. [39]

    F. List, N. L. Rodd and G. F. Lewis, Extracting the Galactic Center excess’ source-count distribution with neural nets, Phys. Rev. D 104 (2021) 123022 [2107.09070]

  32. [40]

    Mishra-Sharma and K

    S. Mishra-Sharma and K. Cranmer, Semi-parametric γ-ray modeling with Gaussian processes and variational inference, in 34th Conference on Neural Information 54 Processing Systems, 10, 2020, 2010.10450

  33. [41]

    Christy, E

    K. Christy, E. J. Baxter and J. Kumar, Applying simulation-based inference to spectral and spatial information from the Galactic Center gamma-ray excess, JCAP 07 (2024) 066 [ 2402.04549]

  34. [42]

    Goodenough and D

    L. Goodenough and D. Hooper, Possible Evidence For Dark Matter Annihilation In The Inner Milky Way From The Fermi Gamma Ray Space Telescope, 0910.2998

  35. [43]

    Hooper and L

    D. Hooper and L. Goodenough, Dark Matter Annihilation in The Galactic Center As Seen by the Fermi Gamma Ray Space Telescope, Phys. Lett. B 697 (2011) 412 [ 1010.2752]

  36. [44]

    Mishra-Sharma, N

    S. Mishra-Sharma, N. L. Rodd and B. R. Safdi, NPTFit: A code package for Non-Poissonian Template Fitting, Astron. J. 153 (2017) 253 [ 1612.03173]

  37. [45]

    Hermans, V

    J. Hermans, V. Begy and G. Louppe, Likelihood-free MCMC with Amortized Approximate Ratio Estimators , 1903.04057

  38. [46]

    B. K. Miller, A. Cole, G. Louppe and C. Weniger, Simulation-efficient marginal posterior estimation with swyft: stop wasting your precious time , in 36th Conference on Neural Information Processing Systems , 11, 2020, 2011.13951

  39. [47]

    Fermi Science Support Development Team, Fermitools: Fermi Science Tools, Astrophysics Source Code Library, record ascl:1905.011, May, 2019

  40. [48]

    K. M. G´ orski, E. Hivon, A. J. Banday, B. D. Wand elt, F. K. Hansen, M. Reinecke and M. Bartelmann, HEALPix: A Framework for High-Resolution Discretization and Fast Analysis of Data Distributed on the Sphere, ApJ 622 (2005) 759 [ astro-ph/0409513]

  41. [49]

    Cowan, K

    G. Cowan, K. Cranmer, E. Gross and O. Vitells, Asymptotic formulae for likelihood-based tests of new physics, Eur. Phys. J. C 71 (2011) 1554 [ 1007.1727]. [Erratum: Eur.Phys.J.C 73, 2501 (2013)]

  42. [50]

    Linden, N

    T. Linden, N. L. Rodd, B. R. Safdi and T. R. Slatyer, High-energy tail of the Galactic Center gamma-ray excess, Phys. Rev. D 94 (2016) 103013 [ 1604.01026]

  43. [51]

    Fermi-LAT Collaboration, D. A. Smith et al., The Third Fermi Large Area Telescope Catalog of Gamma-Ray Pulsars, Astrophys. J. 958 (2023) 191 [2307.11132]

  44. [52]

    D. J. Bartlett, A. Kosti´ c, H. Desmond, J. Jasche and G. Lavaux, Constraints on dark matter annihilation and decay from the large-scale structure of the nearby Universe, Phys. Rev. D 106 (2022) 103526 [2205.12916]

  45. [53]

    T. A. Porter, G. Johannesson and I. V. Moskalenko, High-Energy Gamma Rays from the Milky Way: Three-Dimensional Spatial Models for the Cosmic-Ray and Radiation Field Densities in the Interstellar Medium, Astrophys. J. 846 (2017) 67 [ 1708.00816]

  46. [54]

    F. List, N. L. Rodd, G. F. Lewis and I. Bhat, The GCE in a New Light: Disentangling the γ-ray Sky with Bayesian Graph Convolutional Neural Networks , Phys. Rev. Lett. 125 (2020) 241102 [ 2006.12504]

  47. [55]

    M. Su, T. R. Slatyer and D. P. Finkbeiner, Giant Gamma-ray Bubbles from Fermi-LAT: AGN Activity or Bipolar Galactic Wind? , Astrophys. J. 724 (2010) 1044 [1005.5480]

  48. [56]

    Wolleben, A New Model For The Loop-I (The North Polar Spur) Region, Astrophys

    M. Wolleben, A New Model For The Loop-I (The North Polar Spur) Region, Astrophys. J. 664 (2007) 349 [0704.0276]

  49. [57]

    Fermi-LAT Collaboration, A. A. Abdo et al., Fermi Large Area Telescope Observations of Two Gamma-Ray Emission Components from the Quiescent Sun , ApJ 734 (2011) 116 [ 1104.2093]

  50. [58]

    Fermi-LAT Collaboration, A. A. Abdo et al., Fermi Observations of γ-Ray Emission from the Moon , ApJ 758 (2012) 140

  51. [59]

    Acero et al., Development of the Model of Galactic Interstellar Emission for Standard Point-Source Analysis of Fermi Large Area Telescope Data, Astrophys

    Fermi-LAT Collaboration, F. Acero et al., Development of the Model of Galactic Interstellar Emission for Standard Point-Source Analysis of Fermi Large Area Telescope Data, Astrophys. J. Suppl. 223 (2016) 26 [ 1602.07246]

  52. [60]

    Fermi-LAT Collaboration, A. A. Abdo et al., Detection of the Small Magellanic Cloud in gamma-rays with Fermi/LAT, Astrophys. J. 523 (2010) A46 [ 1008.2127]

  53. [61]

    Ackermann et al., Deep view of the Large Magellanic Cloud with six years of Fermi-LAT observations, Astron

    Fermi-LAT Collaboration, M. Ackermann et al., Deep view of the Large Magellanic Cloud with six years of Fermi-LAT observations, Astron. Astrophys. 586 (2016) A71 [1509.06903]

  54. [62]

    R. K. Leane and T. R. Slatyer, Spurious point source signals in the galactic center excess , Physical Review Letters 125 (2020) 121105

  55. [63]

    N. M. McClure-Griffiths, S. Stanimirovi´ c and D. R. Rybarczyk, Atomic Hydrogen in the Milky Way: A Stepping Stone in the Evolution of Galaxies , Annual Reviews of Astronomy and Astrophysics 61 (2023) 19 [2307.08464]

  56. [64]

    Mertsch and A

    P. Mertsch and A. Vittino, Bayesian inference of three-dimensional gas maps: I. Galactic CO , 2012.15770

  57. [65]

    S. C. Madden, D. Cormier, S. Hony, V. Lebouteiller, N. Abel, M. Galametz, I. De Looze, M. Chevance, F. L. Polles, M. Y. Lee, F. Galliano, A. Lambert-Huyghe, D. Hu and L. Ramambason, Tracing the total molecular gas in galaxies: [CII] and the CO-dark gas , Astron. Astrophys. 643...

  58. [66]

    D. J. Schlegel, D. P. Finkbeiner and M. Davis, Maps of dust IR emission for use in estimation of reddening and CMBR foregrounds, Astrophys. J. 500 (1998) 525 [astro-ph/9710327]

  59. [67]

    I. A. Grenier, J.-M. Casandjian and R. Terrier, Unveiling Extensive Clouds of Dark Gas in the Solar Neighborhood, Science 307 (2005) 1292

  60. [68]

    Dimitriou, C

    A. Dimitriou, C. Weniger and C. A. Correa, Towards reconstructing the halo clustering and halo mass function of N-body simulations using neural ratio estimation, 2206.11312

  61. [69]

    Donath et al., Gammapy: A python package for gamma-ray astronomy, A&A 678 (2023) A157

    A. Donath et al., Gammapy: A python package for gamma-ray astronomy, A&A 678 (2023) A157

  62. [70]

    G. H. Collin, N. L. Rodd, T. Erjavec and K. Perez, A Compound Poisson Generator Approach to Point-source Inference in Astrophysics, The Astrophysical Journal Supplement Series 260 (2022) 29

  63. [71]

    B. K. Miller, A. Cole, C. Weniger, F. Nattino, O. Ku and M. W. Grootes, swyft: Truncated Marginal Neural Ratio Estimation in Python , J. Open Source Softw. 7 (2022) 4205

  64. [72]

    R. K. Leane and T. R. Slatyer, The enigmatic Galactic Center excess: Spurious point sources and signal mismodeling, Physical Review D 102 (2020) 63019

  65. [73]

    Lueckmann, J

    J.-M. Lueckmann, J. Boelts, D. S. Greenberg, P. J. Gon¸ calves and J. H. Macke,Benchmarking simulation-based inference, 2101.04653. 55

  66. [74]

    Cranmer, J

    K. Cranmer, J. Pavez and G. Louppe, Approximating likelihood ratios with calibrated discriminative classifiers, 1506.02169

  67. [75]

    Papamakarios, E

    G. Papamakarios, E. Nalisnick, D. J. Rezende, S. Mohamed and B. Lakshminarayanan, Normalizing flows for probabilistic modeling and inference , Journal of Machine Learning Research 22 (2021) 1

  68. [76]

    Devroye, L

    L. Devroye, L. Gy¨ orfi and G. Lugosi,A Probabilistic Theory of Pattern Recognition, vol. 31 of Stochastic Modelling and Applied Probability . Springer, 1996, 10.1007/978-1-4612-0711-5

  69. [77]

    K. P. Murphy, Probabilistic Machine Learning: An introduction. MIT Press, 2022

  70. [78]

    A. Mao, M. Mohri and Y. Zhong, Cross-entropy loss functions: Theoretical analysis and applications , in International conference on Machine learning , pp. 23803–23828, PMLR, 2023, 2304.07288

  71. [79]

    B. K. Miller, A. Cole, P. Forr´ e, G. Louppe and C. Weniger, Truncated marginal neural ratio estimation, in 37th Conference on Neural Information Processing Systems, 07, 2021, 2107.01214, DOI

  72. [80]

    Defferrard, M

    M. Defferrard, M. Milani, F. Gusset and N. Perraudin, DeepSphere: a graph-based spherical CNN , in International Conference on Learning Representations, 2020, https://openreview.net/forum?id=B1e3OlStPB

  73. [81]

    Falcon and The PyTorch Lightning team, PyTorch Lightning, Mar., 2019

    W. Falcon and The PyTorch Lightning team, PyTorch Lightning, Mar., 2019. 10.5281/zenodo.3828935

  74. [82]

    Anau Montel, J

    N. Anau Montel, J. Alvey and C. Weniger, Scalable inference with autoregressive neural ratio estimation , Mon. Not. Roy. Astron. Soc. 530 (2024) 4107 [2308.08597]

  75. [83]

    Anau Montel and C

    N. Anau Montel and C. Weniger, Detection is truncation: studying source populations with truncated marginal neural ratio estimation , in 36th Conference on Neural Information Processing Systems: Workshop on Machine Learning and the Physical Sciences , 11, 2022, 2211.04291

  76. [84]

    Huang, D

    C.-W. Huang, D. Krueger, A. Lacoste and A. Courville, Neural autoregressive flows, 2018

  77. [85]

    Ronneberger, P

    O. Ronneberger, P. Fischer and T. Brox, U-net: Convolutional networks for biomedical image segmentation, in Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18 , pp....

  78. [86]

    Esteves, C

    C. Esteves, C. Allen-Blanchette, A. Makadia and K. Daniilidis, Learning so (3) equivariant representations with spherical cnns , in Proceedings of the european conference on computer vision (ECCV) , pp. 52–68, 2018

  79. [87]

    T. S. Cohen, M. Geiger, J. K¨ ohler and M. Welling, Spherical cnns, preprint (2018) [ 1801.10130]

  80. [88]

    Perraudin, M

    N. Perraudin, M. Defferrard, T. Kacprzak and R. Sgier, Deepsphere: Efficient spherical convolutional neural network with healpix sampling for cosmological applications, Astronomy and Computing 27 (2019) 130–146

  81. [89]

    Abdollahi et al., Fermi Large Area Telescope Fourth Source Catalog, Astrophys

    Fermi-LAT Collaboration, S. Abdollahi et al., Fermi Large Area Telescope Fourth Source Catalog, Astrophys. J. Suppl. 247 (2020) 33 [ 1902.10045]

  82. [90]

    K. He, X. Zhang, S. Ren and J. Sun, Deep Residual Learning for Image Recognition, 1512.03385

  83. [91]

    L. Ruff, R. Vandermeulen, N. Goernitz, L. Deecke, S. A. Siddiqui, A. Binder, E. M¨ uller and M. Kloft,Deep one-class classification, in Proceedings of the 35th International Conference on Machine Learning (J. Dy and A. Krause, eds.), vol. 80 of Proceedings of Machine Learning ...

  84. [92]

    Caron, L

    S. Caron, L. Hendriks and R. Verheyen, Rare and Different: Anomaly Scores from a combination of likelihood and out-of-distribution models to detect new physics at the LHC , SciPost Phys. 12 (2022) 077 [2106.10164]

  85. [93]

    with a lower angular resolution. Large-scale differ- ences will also be present since FGMA does not encom- pass any description of the FBs or Loop I, which are present in the 4FGL background model via re-injected fit residuals. These large-scale variations are easy to rec- ogn...

  86. [94]

    Ackermann et al., The spectrum of isotropic diffuse gamma-ray emission between 100 MeV and 820 GeV , Astrophys

    Fermi-LAT Collaboration, M. Ackermann et al., The spectrum of isotropic diffuse gamma-ray emission between 100 MeV and 820 GeV , Astrophys. J. 799 (2015) 86 [ 1410.3696]

  87. [95]

    CTA Consortium Collaboration, B. S. Acharya et al., Science with the Cherenkov Telescope Array . WSP, 11, 2018, 10.1142/10986, [ 1709.07997]

  88. [96]

    Panes, C

    B. Panes, C. Eckner, L. Hendriks, S. Caron, K. Dijkstra, G. J´ ohannesson, R. R. de Austri and G. Zaharijas, Identification of point sources in gamma rays using U-shaped convolutional neural networks and a data challenge , Astron. Astrophys. 656 (2021) A62 [2103.11068]

  89. [97]

    van den Oetelaar, S

    C. van den Oetelaar, S. Bhattacharyya, B. Panes, S. Caron, G. Zaharijas, R. R. de Austri and G. J´ ohannesson,Localisation and classification of gamma ray sources using neural networks , PoS ICRC2021 (2021) 663

  90. [98]

    M. Long, Z. Yu, W. Zoujian, W. Xiaotian and X. Jiarong, SiamVIT: A patchwise network for γ-ray point source detection, Astronomy & Astrophysics 692 (2024) A38

  91. [99]

    Loshchilov and F

    I. Loshchilov and F. Hutter, Decoupled Weight Decay Regularization, arXiv e-prints (2017) arXiv:1711.05101 [1711.05101]

  92. [100]

    A. Cole, B. K. Miller, S. J. Witte, M. X. Cai, M. W. Grootes, F. Nattino and C. Weniger, Fast and credible likelihood-free cosmology with truncated marginal neural ratio estimation, JCAP 09 (2022) 004 [ 2111.08030]

  93. [101]

    Ben Bekhti et al., HI4PI: A full-sky H I survey based on EBHIS and GASS , Astron

    HI4PI Collaboration, N. Ben Bekhti et al., HI4PI: A full-sky H I survey based on EBHIS and GASS , Astron. Astrophys. 594 (2016) A116 [ 1610.06175]

  94. [102]

    P. M. W. Kalberla, W. B. Burton, D. Hartmann, E. M. Arnal, E. Bajaja, R. Morras and W. G. L. Poppel, The Leiden/Argentine/Bonn (LAB) survey of Galactic HI: Final data release of the combined LDS and IAR surveys with improved stray-radiation corrections, Astron. Astrophys. 440 ...

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.