Pith. sign in

REVIEW 2 major objections 6 minor 55 references

Weakly supervised machine learning for model-agnostic searches of new phenomena in the $\gamma$-ray sky

T0 review · 2 major / 6 minor · reviewed 2026-07-09 · glm-5.2

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

Pith's one-line read Train classifiers on unlabeled mixtures to find exotic gamma-ray signals

desk verdict Honest proof-of-concept for BvM weak supervision in gamma-ray astronomy; math is sound, case studies are well-designed, but the same-background assumption is untested under realistic conditions read the letter →

arxiv 2607.07158 v1 pith:L4WZ4ZAL submitted 2026-07-08 astro-ph.HE astro-ph.IMhep-ph

Michael Krämer , Silvia Manconi , Kathrin Nippel This is my paper · ORCID
classification astro-ph.HEastro-ph.IMhep-ph PACS 95.75.Pq95.85.Pw95.30.Cq
keywords signalsupervisedgammaclassificationfullyphenomenasearchesweakly
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 proposes that classifiers trained to distinguish a pure background sample from a mixed sample containing an unknown fraction of signal can recover signal-like events without ever being shown labeled signal examples. The mathematical core is simple: if the ratio of mixed-sample to background-sample densities is f times the signal-to-background likelihood ratio plus (1-f), then for any nonzero signal fraction f, this ratio is monotonically related to the true signal likelihood ratio. A classifier trained on the two samples therefore learns a score that ranks events by how signal-like they are, even though no individual event was ever labeled as signal. The authors test this background-versus-mixture approach on three gamma-ray astrophysics problems: separating pulsars from active galactic nuclei (a clean benchmark where the method approaches fully supervised performance), identifying dark matter subhalos (where performance is competitive but degraded by spectral overlap with astrophysical sources), and detecting spectral wiggles induced by axion-photon oscillations (where both supervised and weakly supervised methods struggle when the modulation amplitude is small). In each case, the signal fraction in the mixed sample controls the tradeoff between true positive rate and false positive rate: larger fractions improve sensitivity but admit more background contamination. The method is positioned not as a replacement for model-specific likelihood searches but as a model-agnostic candidate-selection and anomaly-ranking tool that can flag interesting sources for follow-up without committing to a particular signal hypothesis during training.

What carries the argument

The background-versus-mixture (BvM) setup: two training samples are constructed, one pure background B drawn from p_A(x) and one mixture M drawn from f*p_E(x) + (1-f)*p_A(x). A boosted decision tree is trained to distinguish B from M. Because the background component is identically distributed in both samples, the classifier implicitly learns the signal-to-background density ratio p_E(x)/p_A(x). The signal fraction f and the relative sample sizes control the conservatism of the decision boundary. Normalizing flows are used to augment background samples when the catalog contains too few real sources for controlled mixture studies.

What would settle it

If the background distributions in the pure background sample and the mixed sample differ for reasons other than signal admixture (e.g., selection effects, exposure differences, or correlations between mixture-defining variables and input features), the classifier score is no longer monotone in the signal-to-background likelihood ratio, and the method produces biased rankings. The paper acknowledges this but validates only on controlled samples where the same-background assumption holds by construction.

Watch

Extended reading notes

Core claim

A classifier trained to separate a pure background sample from a mixed sample containing an unknown signal fraction learns a decision score that is monotonically related to the optimal signal-versus-background likelihood ratio, because the density ratio p_M(x)/p_B(x) = f * p_E(x)/p_A(x) + (1-f) is strictly increasing in p_E(x)/p_A(x) for any f in (0,1]. This means weakly supervised training on unlabeled mixtures can rank gamma-ray sources by how exotic they are, approaching fully supervised performance when signal and background are well separated, without requiring labeled signal examples during training.

Load-bearing premise

The astrophysical background component must be distributed identically in the pure background sample and the mixed sample. All three case studies satisfy this by construction using simulated or carefully matched data, but in a real application to unassociated Fermi-LAT sources, selection effects such as differing exposure, Galactic latitude, or detection-significance thresholds between identified and unidentified sources would likely violate this assumption and bias the clasS

Editorial extensions

If this is right

  • Weakly supervised classifiers could be applied directly to unassociated Fermi-LAT sources, using identified astrophysical sources as background and unassociated sources as the mixture, to produce a ranked candidate list for follow-up observations without assuming a specific dark matter or new physics model.
  • The same-background assumption means that any selection effects differing between identified and unassociated source populations (exposure, Galactic latitude, flux thresholds) would bias the classifier, so real-data application requires careful matching or reweighting of the background sample to the mixture sample.
  • The method generalizes to any spectral anomaly search where a reference population of normal sources can be defined, including TeV gamma-ray spectra from Cherenkov telescopes or X-ray observations, provided the energy binning is fine enough to capture the relevant spectral features.
  • The tradeoff between signal fraction and false positive rate implies that optimal candidate selection may require scanning over multiple mixture constructions or combining weakly supervised scores from several signal-fraction settings.
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 / 6 minor

Summary. The manuscript explores weakly supervised classification in a background-versus-mixture (BvM) setup for model-agnostic searches of new phenomena in Fermi-LAT gamma-ray data. The central mathematical claim—that the BvM classifier score is monotone in the signal-to-background likelihood ratio p_E(x)/p_A(x) when the background distribution is identical across samples—is a standard CWoLa result and is correctly stated. Three case studies of increasing difficulty are presented: pulsar-AGN separation (controlled benchmark), dark-matter subhalo identification, and ALP-induced spectral irregularities. In each case, weakly supervised performance is compared against a supervised baseline, and the dependence on signal fraction and sample composition is characterized. The approach is positioned not as a replacement for likelihood-based discovery, but as a candidate-selection and anomaly-ranking strategy. The presentation is clear and the physics motivations are well developed.

Significance. The application of weakly supervised (CWoLa-type) methods to gamma-ray source classification is timely and has not, to the authors' knowledge, been previously attempted in this context. The mathematical foundation (Eq. 2.1 and the monotonicity argument in §2.2) is sound and clearly stated. The three case studies are well-chosen, spanning a controlled benchmark, a model-driven exotic search, and a subtle spectral-deformation scenario, which collectively illustrate both the potential and the limitations of the method. The honest reporting of performance degradation relative to supervised baselines—particularly in the ALP case where the signal is subtle—adds credibility. The normalizing-flow augmentation of the AGN background (Appendix A) is a useful technical contribution, validated with a classifier-discrimination test. The work is a reasonable proof-of-concept that connects developments in collider-physics anomaly detection to high-energy astrophysics.

major comments (2)
  1. §2.2 and §6: The same-background assumption—that p_A(x) is identically distributed in the background sample B and the mixed sample M—is the load-bearing condition for the monotonicity result. All three case studies satisfy this by construction: the pulsar-AGN benchmark uses flow-augmented AGN from the same learned distribution; the DM subhalo case injects signal into a disjoint subset of the same background pool; the ALP case applies modulations to randomly chosen spectra from the same simulated pulsar population. The authors acknowledge that realistic applications to unassociated Fermi-LAT sources would face selection-effect violations (§2.2, §4.4, §6) and defer this to future work. This is an honest framing, but it means the paper demonstrates the method under idealized conditions only. The central claim—that weak supervision 'can identify anomalous or signal-like subsets of data' (§6,
  2. §3.4: The background sample B is augmented with normalizing-flow-generated AGN spectra, while the mixed sample M contains real AGN (plus pulsars). The same-background assumption requires that the flow-generated and real AGN spectra follow the same distribution. Appendix A validates the flow with a classifier-discrimination test (accuracy 52.6%, AUC 0.53), which is reassuring. However, the flow is trained on the same 4FGL AGN population used to construct M, so any subtle distributional mismatch between generated and real spectra would directly bias the BvM classifier in a way that is hard to detect from the discrimination test alone. The authors should discuss whether the flow-generated AGN in B and the real AGN in M are drawn from statistically independent realizations, or whether there is overlap, and whether the discrimination test is sensitive enough to detect the level of mismatch
minor comments (6)
  1. §3.3: The statement 'Adding positional or flux-history information does not further improve performance for BDTs' could benefit from a quantitative comparison (e.g., TPR/FPR values) to support the claim, since the preceding paragraph gives specific numbers for the flux-band-only and all-features cases.
  2. Table 1: For g_aγ = 50×10^{-11} GeV^{-1}, the TPR values (0.157 and 0.108) are very low. The text in §5.3 notes this is expected, but it would help to state explicitly in the table caption that these values indicate the classifier is essentially failing to identify modulated spectra at this coupling, rather than underperforming.
  3. §5.2: The choice m_a = 1 neV is mentioned without justification in the main text; the reader must consult Appendix B. A one-sentence motivation in §5.2 would improve readability.
  4. Figure 3: The arrow convention (tail = supervised, tip = weakly supervised) is explained in the caption but is somewhat unusual. Consider adding a small legend or making the convention more visually intuitive.
  5. §4.4, last paragraph: The sentence 'Improved separation may be expected for annihilation channels with harder spectra, for example τ+τ−' is reasonable but reads as speculation without supporting evidence. A brief quantitative comparison or a reference would strengthen it.
  6. The abstract states 'in favourable cases, the method approaches the performance of fully supervised classifiers.' Given that this is true primarily for the pulsar-AGN benchmark and not for the more physically motivated DM or ALP cases, a slightly more qualified phrasing would be more precise.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the central mathematical claim is a standard probability result, and self-citations serve only as reference benchmarks.

full rationale

The paper's load-bearing mathematical claim (§2.2) is that p_M(x)/p_B(x) = f·p_E(x)/p_A(x) + (1−f) is strictly increasing in the likelihood ratio p_E(x)/p_A(x) for f ∈ (0,1]. This is a standard, self-contained probability identity that does not depend on the authors' prior work. The supervised benchmarks in §3.3 and §4.3 are compared against the authors' own prior results ([14], [20]), but these are used as performance reference points, not as inputs to the weak-supervision derivation. The normalizing flow (Appendix A) is used only for data augmentation of background samples and is validated independently (a classifier distinguishing real from generated spectra achieves 52.6% accuracy, consistent with random guessing). The three case studies (pulsar-AGN, dark matter subhalos, ALP modulations) each construct controlled test scenarios where the same-background assumption is satisfied by construction, and the authors explicitly acknowledge this as a limitation for real-world application (§2.2, §4.4, §6). No step in the derivation chain reduces to its own inputs by definition, no prediction is a renamed fit, and no uniqueness claim is smuggled via self-citation. The self-citations present are normal scholarly practice and do not form a load-bearing circular chain.

Assumptions & free parameters 10 free parameters · 5 assumptions · 0 invented entities

The paper does not invent new particles, forces, or entities. It uses established physics models (WIMPs, ALPs, pulsars, AGN) and established ML methods (BDTs, normalizing flows, weak supervision). The free parameters are either ML hyperparameters (selected by grid search), physics benchmark points (m_a, g_agamma), or experimental control variables (signal fraction, sample ratio). None are fitted to produce the central claim; they are scanned to characterize performance.

free parameters (10)
  • Signal fraction f in mixed sample = varied: 0.04-0.14 (pulsar-AGN), 0.04-0.14 (DM), 0.04-0.14 (ALP)
    The fraction of exotic/signal sources in the mixed sample M. Not fitted to data but scanned as a control parameter. Performance is reported as a function of f.
  • Background-to-mixture ratio |B|:|M| = 1:1 (balanced) and 2:1 (imbalanced)
    Relative sample sizes for B and M in the weakly supervised setup. Chosen to study the effect of training imbalance.
  • BDT number of estimators = 200 (pulsar-AGN, DM), 400 (ALP)
    Number of boosted trees. Selected by grid search on validation performance.
  • BDT max depth = 3
    Maximum tree depth. Selected by grid search.
  • BDT learning rate = not specified numerically
    Selected by grid search but exact value not reported.
  • Classifier score threshold = 0.5
    Fixed threshold for computing TPR/FPR. Authors state qualitative behavior is unchanged for other reasonable thresholds.
  • ALP mass m_a = 1 neV
    Fixed benchmark value for ALP-photon oscillation simulations. Choice discussed in Appendix B but not optimized.
  • ALP-photon coupling g_agamma = 50e-11 and 500e-11 GeV^-1
    Two benchmark coupling strengths chosen to represent weak and strong modulation regimes.
  • Normalizing flow: number of knots K = 10
    Spline knots for coupling transformations (Table 4). Architecture hyperparameter.
  • Normalizing flow: number of transformations = 8
    Number of sequential coupling transformations (Table 4). Architecture hyperparameter.
assumptions (5)
  • domain assumption Same-background assumption: p_A(x) is distributed identically in background sample B and mixed sample M
    Stated in §2.2: 'Under the same-background assumption (the p_A component is distributed identically in B and M).' This is the foundational assumption of the BvM framework. All three case studies satisfy it by construction using simulated or matched data.
  • domain assumption The classifier is sufficiently expressive to learn the optimal decision boundary
    Implicit in §2.2: 'in the limit of sufficient training data and an expressive classifier, thresholds on the classifier score correspond to increasing evidence for an exotic origin.' BDTs with max depth 3 and 200-400 trees may not fully satisfy this in practice.
  • domain assumption Gamma-ray source spectra in fixed energy bins are sufficient features for classification
    §3.2: the baseline input is seven flux-band values from 4FGL. For ALP case, 50 energy bins are used (§5.2). The choice to exclude variability and positional information is motivated by model-agnostic considerations but limits achievable performance.
  • domain assumption The Jansson-Farrar GMF model adequately represents the Galactic magnetic field for ALP oscillation calculations
    §5.2: 'we use the Jansson-Farrar model for the regular GMF, with parameters as given in [54].' Updated models [55] and turbulent components [56] are deferred to future work.
  • domain assumption Simulated dark-matter subhalo spectra from N-body distributions are representative of real subhalo signals
    §4.2: subhalo population modeled using cosmological distributions from Aquarius simulations [4]. The detectability assessment uses fermipy and standard Fermi ScienceTools.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Weakly supervised machine learning for model-agnostic searches of new phenomena in the $\gamma$-ray sky." pith.science (2026). https://pith.science/paper/L4WZ4ZAL

@misc{pith2026260707158,
  author       = {Pith},
  title        = {Pith review of: Weakly supervised machine learning for model-agnostic searches of new phenomena in the $\gamma$-ray sky},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L4WZ4ZAL}},
  note         = {Machine review of arXiv:2607.07158}
}
abstract

The $\gamma$-ray sky, as observed by the Fermi Large Area Telescope, contains a significant number of unassociated sources that may point to new astrophysical populations or more exotic phenomena. Machine-learning methods are widely used for source classification and searches for new physics, but most existing approaches rely on fully supervised training and therefore on explicit signal models. We explore weakly supervised classification as a less model-dependent strategy for analysing $\gamma$-ray source spectra. In a background-versus-mixture setup, classifiers are trained on samples with different signal admixtures rather than on fully labelled signal and background events. We study three representative scenarios: pulsar-active galactic nuclei separation as a controlled benchmark, the identification of dark-matter subhalos, and spectral irregularities induced by axion-photon oscillations. In each case we investigate the impact of signal fraction and sample composition on classification performance. Our results show that weak supervision can identify anomalous or signal-like subsets of data while reducing the reliance on detailed signal templates during training. In favourable cases, the method approaches the performance of fully supervised classifiers, while remaining applicable in situations where the signal model is uncertain or only partially specified. Weakly supervised learning therefore provides a complementary candidate-selection and anomaly-ranking strategy for $\gamma$-ray data analysis and searches for new phenomena.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

55 extracted references · 55 canonical work pages

  1. [1]

    The Large Area Telescope on the Fermi Gamma-ray Space Telescope Mission

    W.B. Atwood, A.A. Abdo, M. Ackermann, W. Althouse, B. Anderson et al.,The Large Area Telescope on the Fermi Gamma-Ray Space Telescope Mission,Astrophys. J.697(2009) 1071 [0902.1089]. [2]Fermi-LATcollaboration,F ermiLarge Area Telescope Fourth Source Catalog,Astrophys. J. Suppl.247(2020) 33 [1902.10045]. [3]Fermi-LATcollaboration,Fermi Large Area Telescope...

    work page Pith review arXiv 2009

  2. [2]

    The Aquarius Project: the subhalos of galactic halos

    V. Springel, J. Wang, M. Vogelsberger, A. Ludlow, A. Jenkins, A. Helmi et al.,The Aquarius Project: the subhalos of galactic halos,Mon. Not. Roy. Astron. Soc.391(2008) 1685 [0809.0898]

    work page Pith review arXiv 2008

  3. [3]

    Dark matter subhalos and the dwarf satellites of the Milky Way

    P. Madau, J. Diemand and M. Kuhlen,Dark matter subhalos and the dwarf satellites of the Milky Way,Astrophys. J.679(2008) 1260 [0802.2265]. [6]Fermi-LATcollaboration,Search for dark matter satellites using the fermi-lat,Astrophys. J. 747(2012) 121 [1201.2691]

    work page Pith review arXiv 2008

  4. [4]

    Detecting Axion-Like Particles With Gamma Ray Telescopes

    D. Hooper and P.D. Serpico,Detecting Axion-Like Particles With Gamma Ray Telescopes, Phys. Rev. Lett.99(2007) 231102 [0706.3203]. [8]Fermi-LATcollaboration,Search for Spectral Irregularities due to Photon–Axionlike-Particle Oscillations with the Fermi Large Area Telescope,Phys. Rev. Lett.116(2016) 161101 [1603.06978]

    work page Pith review arXiv 2007

  5. [5]

    Fermi's Sibyl: Mining the gamma-ray sky for dark matter subhaloes

    N. Mirabal, V. Frias-Martinez, T. Hassan and E. Frias-Martinez,Fermi’s Sibyl: Mining the gamma-ray sky for dark matter subhaloes,Mon. Not. Roy. Astron. Soc.424(2012) L64 [1205.4825]

    work page Pith review arXiv 2012

  6. [6]

    Classification and Ranking of Fermi LAT Gamma-ray Sources from the 3FGL Catalog using Machine Learning Techniques

    P.M. Saz Parkinson, H. Xu, P.L.H. Yu, D. Salvetti, M. Marelli and A.D. Falcone,Classification and Ranking of Fermi LAT Gamma-ray Sources from the 3FGL Catalog using Machine Learning Techniques,Astrophys. J.820(2016) 8 [1602.00385]

    work page Pith review arXiv 2016

  7. [7]

    Blazar Flaring Patterns (B-FlaP): Classifying Blazar Candidates of Uncertain type in the third Fermi-LAT catalog by Artificial Neural Networks

    G. Chiaro, D. Salvetti, G. La Mura, M. Giroletti, D.J. Thompson and D. Bastieri,Blazar flaring patterns (B-FlaP) classifying blazar candidate of uncertain type in the third Fermi-LAT catalogue by artificial neural networks,Mon. Not. Roy. Astron. Soc.462(2016) 3180 [1607.07822]

    work page Pith review arXiv 2016

  8. [8]

    3FGLzoo. Classifying 3FGL Unassociated Fermi-LAT Gamma-ray Sources by Artificial Neural Networks

    D. Salvetti, G. Chiaro, G. La Mura and D.J. Thompson,3FGLzoo: classifying 3FGL unassociated Fermi-LATγ-ray sources by artificial neural networks,Mon. Not. Roy. Astron. Soc.470(2017) 1291 [1705.09832]

    work page Pith review arXiv 2017

Show all 55 references
  1. [9]

    Finke, M

    T. Finke, M. Krämer and S. Manconi,Classification of Fermi-LAT sources with deep learning using energy and time spectra,Mon. Not. Roy. Astron. Soc.507(2021) 4061 [2012.05251]

    2021 arXiv

  2. [10]

    Butter, T

    A. Butter, T. Finke, F. Keil, M. Krämer and S. Manconi,Classification of Fermi-LAT blazars with Bayesian neural networks,JCAP04(2022) 023 [2112.01403]

    2022 arXiv

  3. [11]

    Mirabal, E

    N. Mirabal, E. Charles, E.C. Ferrara, P.L. Gonthier, A.K. Harding, M.A. Sánchez-Conde et al., 3FGL Demographics Outside the Galactic Plane using Supervised Machine Learning: Pulsar and Dark Matter Subhalo Interpretations,Astrophys. J.825(2016) 69 [1605.00711]

    2016 arXiv

  4. [12]

    Coronado-Blazquez, M.A

    J. Coronado-Blazquez, M.A. Sanchez-Conde, A. Dominguez, A. Aguirre-Santaella, M. Di Mauro, N. Mirabal et al.,Unidentified Gamma-ray Sources as Targets for Indirect Dark Matter Detection with the Fermi-Large Area Telescope,JCAP07(2019) 020 [1906.11896]

    2019 arXiv

  5. [13]

    Mirabal and A

    N. Mirabal and A. Bonaca,Machine-learned dark matter subhalo candidates in the 4FGL-DR2: search for the perturber of the GD-1 stream,JCAP11(2021) 033 [2105.12131]. – 25 –

    2021 arXiv

  6. [14]

    Gammaldi, B

    V. Gammaldi, B. Zaldívar, M.A. Sánchez-Conde and J. Coronado-Blázquez,A search for dark matter among Fermi-LAT unidentified sources with systematic features in machine learning, Mon. Not. Roy. Astron. Soc.520(2023) 1348 [2207.09307]

    2023 arXiv

  7. [15]

    Amerio, D

    A. Amerio, D. Malyshev, B. Zaldívar, V. Gammaldi and M.Á. Sánchez-Conde,Search for dark matter subhalos among unassociated fermi-lat sources in presence of dataset shift,2503.14584

    arXiv

  8. [16]

    Butter, M

    A. Butter, M. Krämer, S. Manconi and K. Nippel,Searching for dark matter subhalos in the Fermi-LAT catalog with Bayesian neural networks,JCAP07(2023) 033 [2304.00032]

    2023 arXiv

  9. [17]

    Hernández-González, I

    J. Hernández-González, I. Inza and J.A. Lozano,Weak supervision and other non-standard classification problems: A taxonomy,Pattern Recognit. Lett.69(2016) 49

    2016

  10. [18]

    Zhou,A brief introduction to weakly supervised learning,National Science Review5 (2018) 44

    Z.-H. Zhou,A brief introduction to weakly supervised learning,National Science Review5 (2018) 44

    2018

  11. [19]

    Metodiev, B

    E.M. Metodiev, B. Nachman and J. Thaler,Classification without labels: Learning from mixed samples in high energy physics,JHEP10(2017) 174 [1708.02949]

    2017 arXiv

  12. [20]

    Collins, K

    J.H. Collins, K. Howe and B. Nachman,Extending the search for new resonances with machine learning,Phys. Rev. D99(2019) 014038 [1902.02634]

    2019 arXiv

  13. [21]

    Karagiorgi, G

    G. Karagiorgi, G. Kasieczka, S. Kravitz, B. Nachman and D. Shih,Machine learning in the search for new fundamental physics,Nature Rev. Phys.4(2022) 399

    2022

  14. [22]

    Belis, P

    V. Belis, P. Odagiu and T.K. Aarrestad,Machine learning for anomaly detection in particle physics,Rev. Phys.12(2024) 100091 [2312.14190]

    2024 arXiv

  15. [23]

    Shih, M.R

    D. Shih, M.R. Buckley, L. Necib and J. Tamanas,via machinae: Searching for stellar streams using unsupervised machine learning,Mon. Not. Roy. Astron. Soc.509(2021) 5992 [2104.12789]

    2021 arXiv

  16. [24]

    Shih, M.R

    D. Shih, M.R. Buckley and L. Necib,Via Machinae 2.0: Full-sky, model-agnostic search for stellar streams in Gaia DR2,Mon. Not. Roy. Astron. Soc.529(2024) 4745 [2303.01529]

    2024 arXiv

  17. [25]

    Pettee, S

    M. Pettee, S. Thanvantri, B. Nachman, D. Shih, M.R. Buckley and J.H. Collins,Weakly supervised anomaly detection in the Milky Way,Mon. Not. Roy. Astron. Soc.527(2023) 8459 [2305.03761]

    2023 arXiv

  18. [26]

    Hallin, D

    A. Hallin, D. Shih, C. Krause and M.R. Buckley,Via Machinae 3.0: A search for stellar streams in Gaia with the CATHODE algorithm,2509.08064

    arXiv

  19. [27]

    Cerruti,Characterization and classification ofγ-ray bursts from blazars,Astron

    M. Cerruti,Characterization and classification ofγ-ray bursts from blazars,Astron. Astrophys. 698(2025) A101 [2410.21974]

    2025 arXiv

  20. [28]

    Friedman,Greedy function approximation: A gradient boosting machine.,Annals of Statistics29(2001) 1189

    J.H. Friedman,Greedy function approximation: A gradient boosting machine.,Annals of Statistics29(2001) 1189

    2001

  21. [29]

    Finke, M

    T. Finke, M. Hein, G. Kasieczka, M. Krämer, A. Mück, P. Prangchaikul et al.,Tree-based algorithms for weakly supervised anomaly detection,Phys. Rev. D109(2024) 034033 [2309.13111]

    2024 arXiv

  22. [30]

    Grinsztajn, E

    L. Grinsztajn, E. Oyallon and G. Varoquaux,Why do tree-based models still outperform deep learning on tabular data?,ArXiv(2022) [2207.08815]

    2022 arXiv

  23. [31]

    Hallin, J

    A. Hallin, J. Isaacson, G. Kasieczka, C. Krause, B. Nachman, T. Quadfasel et al.,Classifying anomalies through outer density estimation,Phys. Rev. D106(2022) 055006 [2109.00546]

    2022 arXiv

  24. [32]

    Rezende and S

    D.J. Rezende and S. Mohamed,Variational inference with normalizing flows,ArXiv(2015) [1505.05770]

    2015 arXiv

  25. [33]

    Dinh, J.N

    L. Dinh, J.N. Sohl-Dickstein and S. Bengio,Density estimation using real nvp,ArXiv(2016) [1605.08803]

    2016 arXiv

  26. [34]

    Durkan, A

    C. Durkan, A. Bekasov, I. Murray and G. Papamakarios,Neural spline flows,1906.04032. – 26 –

    1906 arXiv

  27. [35]

    Cirelli, G

    M. Cirelli, G. Corcella, A. Hektor, G. Hutsi, M. Kadastik, P. Panci et al.,PPPC 4 DM ID: A Poor Particle Physicist Cookbook for Dark Matter Indirect Detection,JCAP03(2011) 051 [1012.4515]

    2011 arXiv

  28. [36]

    M.J. Wood, R. Caputo, E. Charles, M. di Mauro, J.D. Magill and J.S. Perkins,Fermipy: An open-source python package for analysis of fermi-lat data,arXiv: Instrumentation and Methods for Astrophysics(2017)

    2017

  29. [37]

    Di Luzio, M

    L. Di Luzio, M. Giannotti, E. Nardi and L. Visinelli,The landscape of QCD axion models, Phys. Rept.870(2020) 1 [2003.01100]

    2020 arXiv

  30. [38]

    Chadha-Day, J

    F. Chadha-Day, J. Ellis and D.J.E. Marsh,Axion dark matter: What is it and why now?,Sci. Adv.8(2022) abj3618 [2105.01406]

    2022 arXiv

  31. [39]

    O’Hare,Cosmology of axion dark matter,PoSCOSMICWISPers(2024) 040 [2403.17697]

    C.A.J. O’Hare,Cosmology of axion dark matter,PoSCOSMICWISPers(2024) 040 [2403.17697]

    2024 arXiv

  32. [40]

    Majumdar, F

    J. Majumdar, F. Calore and D. Horns,Search for gamma-ray spectral modulations in galactic pulsars,Journal of Cosmology and Astroparticle Physics2018(2018) 048–048 [1801.08813]

    2018 arXiv

  33. [41]

    Z.-Q. Xia, C. Zhang, Y.-F. Liang, L. Feng, Q. Yuan, Y.-Z. Fan et al.,Searching for spectral oscillations due to photon-axionlike particle conversion using the Fermi-LAT observations of bright supernova remnants,Phys. Rev. D97(2018) 063003 [1801.01646]

    2018 arXiv

  34. [42]

    Liang, C

    Y.-F. Liang, C. Zhang, Z.-Q. Xia, L. Feng, Q. Yuan and Y.-Z. Fan,Constraints on axion-like particle properties with TeV gamma-ray observations of Galactic sources,JCAP06(2019) 042 [1804.07186]

    2019 arXiv

  35. [43]

    Pallathadka et al.,Reconciling hints on axion-like-particles from high-energy gamma rays with stellar bounds,JCAP11(2021) 036 [2008.08100]

    G.A. Pallathadka et al.,Reconciling hints on axion-like-particles from high-energy gamma rays with stellar bounds,JCAP11(2021) 036 [2008.08100]

    2021 arXiv

  36. [44]

    Batkovic, A

    I. Batkovic, A. De Angelis, M. Doro and M. Manganaro,Axion-like particle searches with iacts, Universe7(2021) 185 [2106.03424]

    2021 arXiv

  37. [45]

    Reynolds, M.C.D

    C.S. Reynolds, M.C.D. Marsh, H.R. Russell, A.C. Fabian, R. Smith, F. Tombesi et al., Astrophysical limits on very light axion-like particles from Chandra grating spectroscopy of NGC 1275,Astrophys. J.890(2020) 59 [1907.05475]

    2020 arXiv

  38. [46]

    Kachelriess and J

    M. Kachelriess and J. Tjemsland,Detecting ALP wiggles at TeV energies,JCAP01(2024) 044 [2305.03604]

    2024 arXiv

  39. [47]

    Schiavone, L

    F. Schiavone, L. Di Venere and F. Giordano,Projected sensitivity of CTAO to axion-like particles from blazars with a machine learning approach,2512.19259

    arXiv

  40. [48]

    Mirizzi and D

    A. Mirizzi and D. Montanino,Stochastic conversions of tev photons into axion-like particles in extragalactic magnetic fields,Journal of Cosmology and Astroparticle Physics2009(2017) 004–004 [0911.0015]

    2017 arXiv

  41. [49]

    Jaffe,Practical Modeling of Large-Scale Galactic Magnetic Fields: Status and Prospects, Galaxies7(2019) 52 [1904.12689]

    T.R. Jaffe,Practical Modeling of Large-Scale Galactic Magnetic Fields: Status and Prospects, Galaxies7(2019) 52 [1904.12689]

    2019 arXiv

  42. [50]

    Jansson and G.R

    R. Jansson and G.R. Farrar,A New Model of the Galactic Magnetic Field,ApJ, Volume 757, Issue 1, article id. 14, 13 pp. (2012)(2012) [1204.3662]

    2012 arXiv

  43. [51]

    Unger and G.R

    M. Unger and G.R. Farrar,The Coherent Magnetic Field of the Milky Way,Astrophys. J.970 (2024) 95 [2311.12120]

    2024 arXiv

  44. [52]

    Carenza, C

    P. Carenza, C. Evoli, M. Giannotti, A. Mirizzi and D. Montanino,Turbulent axion-photon conversions in the Milky Way,Phys. Rev. D104(2021) 023003 [2104.13935]

    2021 arXiv

  45. [53]

    Meyer, J

    M. Meyer, J. Davies and J. Kuhlmann,gammaalps: An open-source python package for computing photon-axion-like-particle oscillations in astrophysical environments,2108.02061. [58]Fermi-LATcollaboration,The Third Fermi Large Area Telescope Catalog of Gamma-Ray Pulsars,Astrophys. ...

    2023 arXiv

  46. [54]

    Bhattacharjee, C

    P. Bhattacharjee, C. Eckner, G. Zaharijas, G. Kluge and G. D’Amico,Probing the Parameter Space of Axion-Like Particles Using Simulation-Based Inference,2509.20578

  47. [55]

    Durkan, A

    C. Durkan, A. Bekasov, I. Murray and G. Papamakarios,Cubic-spline flows,ArXiv(2019) [1906.02145]. – 28 –

    2019 arXiv

Pith tools

Reviewed July 9, 2026 · model on record in the stance chip above.