Pith. sign in

REVIEW 3 major objections 6 minor 42 references

Simulation-based inference with neural posterior estimation applied to X-ray spectral fitting II -- High-resolution spectroscopy with the X-ray Integral Field Unit

T0 review · 3 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read Summary statistics make X-ray fitting 10 to 100 times faster

desk verdict Useful, mostly honest empirical comparison of summary statistics for high-resolution X-ray SBI, but the 'well-calibrated' claim outruns the evidence and needs a coverage test. read the letter →

arxiv 2506.05911 v1 pith:5K6EBWA6 submitted 2025-06-06 astro-ph.IM astro-ph.COastro-ph.HE

classification astro-ph.IMastro-ph.COastro-ph.HE
keywords simulation-basedinferenceneuralposteriorestimationX-rayspectralfittingsummarystatisticsX-IFUhigh-resolutionspectroscopynormalisingflowscalibration
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

Simulation-based inference with neural posterior estimation can fit high-resolution X-ray spectra quickly, but only if the spectrum is compressed first. This paper argues that hand-crafted summary statistics, such as total and mean counts, standard deviation, counts in 50 logarithmic energy bins, hardness and differential ratios between bins, plus count-weighted mean energies in selected emission-line complexes, beat both the raw spectrum and automatically learned embeddings for the high-resolution X-ray Integral Field Unit. With those summaries, single- and multi-round neural inference recovers posterior distributions matching exact nested-sampling computation for smooth Comptonised spectra, relativistic reflection models, and two-temperature plasma emission models. Inference becomes about 10 to 100 times faster than nested sampling, and a trained single-round network is amortized: once trained, it fits many spectra almost instantly. A dedicated line-complex statistic is required for redshift and velocity, which otherwise are poorly constrained.

What carries the argument

The central object is the hand-crafted summary-statistic vector: the total, mean, and standard deviation of the counts; counts in 50 logarithmically spaced energy bins; hardness ratios and differential ratios between adjacent bins; and, for line-rich models, count-weighted mean energies in four selected line complexes. This vector replaces the full 20k-to-24k channel spectrum before being passed to a neural posterior estimator built from masked autoregressive normalising flows. The compression does the main work: it reduces the mapping the flow must learn, removes redundant highly-correlated information, and lets the network interpolate instead of overfitting. Multi-round inference uses truncated proposals to concentrate simulations near the observation, while single-round inference builds an amortized network that answers many spectra at once.

What would settle it

Take a simulated X-IFU observation whose spectrum has a narrow line or a shape change in an energy band not represented by the 50 logarithmic bins or the four line complexes, and run the SBI pipeline with the paper's summary statistics. If the resulting posterior is offset from the true parameter or its credible intervals are systematically too narrow compared with nested sampling, the claim that hand-crafted summaries are sufficient would be refuted. A simpler version: repeat the two-plasma test with one line complex removed and check redshift and velocity calibration.

Watch

Extended reading notes

Core claim

The paper's central claim is that a high-resolution X-ray spectrum can be replaced by a small set of physically motivated summary statistics without losing the information needed for parameter inference, and that neural posterior estimation on those summaries matches exact nested-sampling results at a fraction of the cost. The authors demonstrate this on three regimes: a smooth Comptonised continuum, a relativistic reflection model, and a two-temperature plasma model with emission lines. In each case, the compressed summary vector, roughly one hundred numbers compared with tens of thousands of spectral channels, produces posteriors compatible with nested sampling, in the line-free cases with about 10 to 100 times less computation. The paper also finds that single-round amortized training, while expensive to build, can be reused to constrain many spectra and to map which parameters an observation can actually measure. Finally, they show that line information needs dedicated summary statistics: count-weighted mean energies in selected line complexes are necessary to recover redshift and velocity.

Load-bearing premise

The load-bearing premise is that the handful of hand-picked statistics preserves all the information the parameters can imprint on the spectrum; the paper's own result shows this is fragile, since redshift and velocity are badly recovered until dedicated line-complex statistics are added, and those statistics were chosen after examining the test problems.

Editorial extensions

If this is right

  • Multi-round inference with about 25k simulated spectra matches nested-sampling posteriors for smooth models, replacing millions of response convolutions with a 10 to 100 times faster workflow.
  • Single-round amortized inference, though trained on roughly 200k simulations, can be reused on many spectra and enables feasibility studies that map which parameters an observational setup can constrain.
  • For models with emission lines, the summary statistics must include line-aware quantities such as count-weighted mean energies per line complex; otherwise redshift and velocity posteriors are biased and too broad.
  • The surrogate posterior distributions are close to, but slightly wider than, the exact posteriors, and can be used to initialise or propose for exact methods to make those methods faster.
  • Adding a background spectrum is straightforward in this likelihood-free setup, since a Poisson background realisation can be added during simulation without any marginalisation step.
  • The surrogate posterior distributions are close to, but slightly wider than, the exact posteriors, and can be used to initialise or propose for exact methods to make those methods faster.

Reading between the lines

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

  • Because the summary statistics were chosen after seeing the test models, the claimed efficiency is tied to those models; a parameter whose effect appears only in a part of the spectrum not covered by the summaries would likely be missed, as the paper itself shows for redshift and velocity before adding line-complex statistics.
  • A natural extension is to apply the same recipe to other high-resolution instruments, or to replace hand-picked line complexes with data-driven summaries such as wavelet scattering transforms, which the authors mention as future work.
  • At very high signal-to-noise the posterior volume shrinks and the amortized network's mapping degrades; restricting the training prior to the relevant count range, as the paper suggests, is a testable way to recover performance.
  • The success of simple summaries suggests a general heuristic for likelihood-free inference on high-dimensional spectra: invest in domain-motivated compression before increasing network capacity.
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

3 major / 6 minor

Summary. This manuscript extends the authors' earlier SBI-NPE work to high-resolution X-IFU spectroscopy. The central idea is to compress simulated spectra before feeding them to a normalizing-flow posterior estimator. Two compression schemes are compared: hand-crafted summary statistics (global count moments, counts in 50 logarithmically spaced bins, hardness and differential ratios, and weighted energies in selected line complexes) and learned embeddings with fully connected networks, with raw spectra as a baseline. The method is benchmarked against BXA nested sampling on three simulated spectral models (tbabs*(comptt+powerlaw), tbabs*relxillNS, and tbabs*(bapec+bapec)), using both multi-round and single-round inference, and an additional real-data application to an XMM-Newton spectrum is included in Appendix C. The main claims are that simple summary statistics are much more efficient than full spectra or learned embeddings, that multi-round inference converges quickly to results comparable to nested sampling, that single-round inference enables amortized feasibility studies, and that the method "delivers well calibrated posteriors."

Significance. If the central claims hold, this is a practically useful contribution for X-IFU science: a 10-100x speedup over nested sampling, an open implementation (the SIXSA fork bsixsa is publicly available), and a real-data validation are all positive features. The paper also gives a clear demonstration that emission-line information requires dedicated summary statistics, which is an important and honest finding. However, the strongest advertised conclusion, that SBI "delivers well calibrated posteriors," is not supported by the evidence presented. The kT example in Fig. 1 is a concrete case of overconfidence on an unidentifiable parameter, and no coverage or simulation-based calibration test is reported. The contribution is therefore significant but needs additional validation before the calibration claim can be accepted.

major comments (3)
  1. [Sec. 3.1, Fig. 1] The paper's own Fig. 1 shows that for tbabs*(comptt+powerlaw), BXA produces an essentially flat marginal for kT, while the MRI summary-statistics posterior is narrow. The authors attribute this to a known limitation of normalizing flows in expressing discontinuous distributions (Sec. 3.1). This is not a benign discrepancy: on an unidentifiable parameter, the surrogate posterior is overconfident, which is exactly the failure mode that the abstract's and Sec. 7's "well calibrated posteriors" claim must exclude. No coverage or simulation-based calibration check (e.g. simulation-based calibration) is reported anywhere. I request either a quantitative calibration/coverage analysis on all three models and all parameters, or a revised, more limited statement that does not claim calibrated posteriors. This is load-bearing because the main advertised benefit is replacing exact inference.
  2. [Sec. 5, Fig. D.3, Sec. 6.3] The summary statistics are not shown to be sufficient for the parameters of interest, and the paper itself admits in Sec. 6.3 that the choice of summaries is motivated a posteriori. Fig. D.3 demonstrates that without the line-complex weighted energies, redshift and velocity are poorly recovered. Thus the central performance claim is conditional on hand-picked statistics that were selected after seeing the benchmark problems. The active-subspace sensitivity analysis in Sec. 6.3 is interpretability, not a sufficiency or robustness check. To make the central claim defensible, please either demonstrate robustness on at least one model not used to design the summaries, or provide a principled procedure for constructing and validating summaries for new models.
  3. [Sec. 6.2, Fig. 7] Fig. 7 shows that the posterior log-probability distributions produced by SBI are systematically wider and lower than those from BXA, and Sec. 6.2 concedes that the surrogate distributions include parameter values worse than those obtained with exact techniques. Combined with Fig. 2, where the lower energy score of MRI with summary statistics is attributed to the artificially narrow kT distribution, the evidence does not support "posterior distributions comparable to those from exact computation" as a general statement. The comparison is based primarily on visual corner plots from a single simulated realization per model. Please report quantitative discrepancy measures (e.g. coverage across repeated simulations, maximum mean discrepancy, or proper scoring rules) and clarify what "comparable" means in the presence of the observed systematic broadening and the kT overconfidence.
minor comments (6)
  1. [Sec. 7] The first sentence contains a typo: "In tpaper,aper" should be "In this paper, we".
  2. [Sec. 5] The text contains typos: "suing" should be "using" and "sumarry statistics" should be "summary statistics".
  3. [Appendix C] The real-data validation uses an XMM-Newton EPIC-PN spectrum, not a high-resolution X-IFU spectrum; the text should state this caveat explicitly, since the paper's headline is high-resolution X-IFU performance.
  4. [Sec. 6.2, Acknowledgements] "cstatminimisation" should be "Cstat minimisation" in Sec. 6.2, and "scypi" should be "scipy" in the Acknowledgements.
  5. [Fig. 2] The numeric labels above the heatmap are not defined clearly; please state in the caption what the three values (0.23, 0.17, 0.38) refer to.
  6. [References] The Kaastra & Bleeker (2016) reference is malformed; the journal is Astronomy & Astrophysics, 587, A151.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central performance claim is benchmarked against external exact inference (BXA) and real XMM-Newton data; self-citations and a-posteriori summary-statistic choices are not load-bearing in the derivation.

full rationale

The paper's derivation chain is anchored externally rather than to its own outputs. The claim that summary statistics yield posteriors comparable to exact computation is tested directly against BXA nested-sampling runs on the same simulated X-IFU observations (Sec. 3.1, Fig. 1; Sec. 4, Fig. 3; Sec. 5, Fig. D.2), and against a real XMM-Newton spectrum of NGC 7793 ULX-4 in Appendix C. These are independent references, not quantities constructed from the SBI posteriors. The line-complex weighted energies in Sec. 5 are fixed functions of the spectrum (mean energies weighted by counts), and the evidence for their benefit is an ablation (Fig. D.3) rather than a fitted parameter renamed as a prediction. The paper explicitly concedes in Sec. 6.3 that the summaries were 'motivate[d] a posteriori' and in Sec. 6.2 that the surrogate posteriors 'include parameter values that are worse than the one obtained using exact techniques'; these are in-sample-selection and calibration limitations, not circular reductions, because the summary statistics are not derived from the target parameters by construction and the mapping from summaries to parameters is learned from simulations. Self-citations to Paper I (Barret & Dupourqué 2024) supply the SIXSA codebase and prior low-resolution validation, but the high-resolution conclusions do not rest on any asserted theorem or ansatz from that paper. Finally, the abstract's 'well calibrated posteriors' claim is unsupported by a coverage test and is tensioned by the admitted normalising-flow smoothing of an unconstrained kT (Sec. 3.1); that is a correctness risk, which per the analysis rules is not counted as circularity. No quoted equation reduces one of the paper's predictions to its inputs by definition.

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

The method introduces no new physical entities. It relies on the simulator being correct, the hand-picked summary statistics being sufficient, the neural network converging, and several hand-chosen hyperparameters (bins, line bands, rounds, simulations, architecture). The summary statistics and training budgets are tuned to the test problems, which is the main source of circularity burden.

free parameters (5)
  • Number of coarse energy bins for sum statistics = 50
    The spectrum is split into 50 logarithmically spaced bins to compute sum counts; this number is chosen by hand and impacts the information content of the summary statistics.
  • Summary statistic set composition = mean, sum, standard deviation, hardness ratios, differential ratios
    Chosen by hand as simple shape and normalization indicators; the paper motivates the choice a posteriori in Sec. 6.3, meaning it is tuned to the test problems.
  • Line complex energy bands = O VIII 0.62-0.66 keV; Fe XXIV+Ne X 0.97-1.2 keV; Fe XXIV+Mg XII 1.4-1.55 keV; Fe XXV 6.4-6.65 keV
    Hand-selected to match known atomic lines in the bapec model; these bands are designed for the specific science case and not general.
  • Number of rounds and simulations = 5 rounds x 5k (comptt), 20 rounds x 10k (bapec), 200k (SRI)
    Chosen empirically as 'a good compromise between speed and accuracy' (Sec. 3.1); directly affects posterior quality and is tuned to each problem.
  • MAF architecture = 10 transforms, 100 hidden units
    Hyperparameters selected as a default set, increased from Paper I to handle higher resolution data.
assumptions (5)
  • domain assumption XSPEC spectral models (comptt, relxillNS, bapec) provide an accurate description of the physical emission and of X-IFU observations.
    All simulations are generated with XSPEC models; if the models or the instrument response are wrong, the learned posteriors inherit the error. Invoked throughout Sec. 2.
  • domain assumption The Poisson realization of model-predicted counts is sufficient to reproduce the measurement distribution, making SBI asymptotically equivalent to Cstat inference.
    Stated in Sec. 2; the paper does not derive this equivalence and excludes Gaussian-statistics reduction software (footnote 1).
  • ad hoc to paper The chosen summary statistics are approximately sufficient statistics for the parameters of interest.
    The method's success rests on this; the paper itself shows it fails for redshift and velocity without the added line-complex weighted energies (Sec. 5, Fig. D.3).
  • ad hoc to paper The neural density estimator (MAF) converges to a reliable approximation of the posterior.
    Training curves (Fig. 6) show early overfitting in early rounds; the paper assumes final validation loss stability indicates convergence, but there is no formal guarantee or calibration test.
  • standard math Standard probability theory and normalizing-flow properties as described in the cited literature.
    Background from Greenberg et al. (2019), Papamakarios et al. (2017), and the sbi package documentation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Simulation-based inference with neural posterior estimation applied to X-ray spectral fitting II -- High-resolution spectroscopy with the X-ray Integral Field Unit." pith.science (2026). https://pith.science/paper/5K6EBWA6

@misc{pith2026250605911,
  author       = {Pith},
  title        = {Pith review of: Simulation-based inference with neural posterior estimation applied to X-ray spectral fitting II -- High-resolution spectroscopy with the X-ray Integral Field Unit},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5K6EBWA6}},
  note         = {Machine review of arXiv:2506.05911}
}
read the original abstract

X-ray spectral fitting in high-energy astrophysics can be reliably accelerated using Machine Learning. In particular, Simulation-based Inference (SBI) produces accurate posterior distributions in the Gaussian and Poisson regime for low-resolution spectra, much faster than other exact approaches such as Monte Carlo Markov Chains or Nested Sampling. We now aim to highlight the capabilities of SBI for high-resolution spectra, as what will be provided by the newAthena X-ray Integral Field Unit (X-IFU). The large number of channels encourages us to use compressed representations of the spectra, taking advantage of the likelihood-free inference aspect of SBI. Two compression schemes are explored, using either simple summary statistics, such as the counts in arbitrary bins or ratios between these bins. We benchmark the efficiency of these approaches using simulated X-IFU spectra with various spectral models, including smooth comptonised spectra, relativistic reflexion models and plasma emission models. We find that using simple and meaningful summary statistics is much more efficient than working directly with the full spectrum, and can derive posterior distributions comparable to those from exact computation using nested sampling. Multi-round inference converges quickly to the good solution. Amortized single round inference requires more simulations, hence longer training time, but can be used to infer model parameters from many observations afterwards. Information from the emission lines must be accounted for using dedicated summary statistics. SBI for X-ray spectral fitting is a robust technique that delivers well calibrated posteriors. This approach shows great promises for high-resolution spectra, offering its potential for the scientific exploitation of the X-IFU. We now plan to apply it to the current era of high-resolution telescopes, and further challenge this approach with real data.

Figures

Figures reproduced from arXiv: 2506.05911 by the authors.

Figure 1
Figure 1. Left: Comparison of the posterior distributions obtained from a BXA run with 1k live-points and a MRI run with 5 rounds and 5k simulations each using summary statistics for a tbabs*(comptt + powerlaw) model. Right: Posterior predictive spectra associated with the posterior distribution obtained using summary statistics. The bands detail the (16-84) and (2.5-97.5) percentiles of the posterior predictive spectra. wher… view at source ↗
Figure 2
Figure 2. Energy score for posterior distributions as produced by BXA, MRI using summary statistics, MRI using raw spectra, and a grid of MRI using a fully connected embedding net with a varying number of layers and hidden parameters. and spread of a multidimensional distribution against a reference parameter value. It is defined as ES n p(θ|x), θˆ o = E∥θ − θˆ∥ − 1 2 E∥θ − θ ′ ∥ where θ and θ ′ are independent random samples… view at source ↗
Figure 3
Figure 3. Left: Comparison of the posterior distributions obtained from a BXA run with 2k live-points and SRI run using the whole spectrum or summary stats, trained both with 200k simulations for a tbabs*relxillNS model. Right: Posterior predictive spectra associated with the posterior distribution obtained using summary statistics. The bands detail the (16-84) and (2.5-97.5) percentiles of the posterior predictive spectra. o… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Left : Energy score obtained from the reconstruction of the spin a, reflection fraction f and ionisation log ξ for varying values of these parameters in a fixed observational setup. Right : Marginalized constraints for the spin a and the ionisation log ξ, compared to t…
Figure 5
Figure 5. Figure 5: Posterior predictive spectra of the 4 line complexes added as extra summary statistics for a MRI run with a tbabs*(bapec+bapec) model for 20 rounds of 10k simulations. The brightest emission lines are highlighted in each plot. of the training. To check whether the infe…
Figure 6
Figure 6. Figure 6: Evolution of the loss for the training and validation set depending on the training epochs and inference round for two spectral models used in this paper. Each round adds a set of 5k simulations to the global set for the left panel and a set of 10k simulations for the …
Figure 8
Figure 8. Figure 8: Sensitivity of each summary statistics to the parameters of a tbabs*(bapec+bapec). The sensitivity arrays are normalized to allow qualitative comparison between the summary statistics. proposal. Multiple-round inference can only fit a single spec￾trum, but it provides …
Figure 7
Figure 7. Figure 7: Comparison of the distribution of posterior probability for the samples produced using BXA and the various SBI inference pipelines used in this paper. posterior distributions for the parameters. In contrast to our pre￾vious work, we focus here on high-energy resolution…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

42 extracted references · 25 canonical work pages

  1. [1]

    Abadi, M., Agarwal, A., Barham, P., et al. 2015, TensorFlow: Large-Scale Ma- chine Learning on Heterogeneous Systems, software available from tensor- Article number, page 10 of 16 Simon Dupourqué and Didier Barret: Neural networks for X-ray spectral fitting flow.org

  2. [2]

    2020, Journal of Machine Learn- ing Research, 21, 1

    Andreux, M., Angles, T., Exarchakis, G., et al. 2020, Journal of Machine Learn- ing Research, 21, 1

  3. [3]

    Arnaud, K. A. 1996, in Astronomical Society of the Pacific Conference Series, V ol. 101, Astronomical Data Analysis Software and Systems V , ed. G. H. Jacoby & J. Barnes, 17

  4. [4]

    K., & Talib, R

    Ayesha, S., Hanif, M. K., & Talib, R. 2020, Information Fusion, 59, 44

  5. [5]

    d., et al

    Barret, D., Albouys, V ., Herder, J.-W. d., et al. 2023, Experimental Astronomy, 55, 373

  6. [6]

    & Cappi, M

    Barret, D. & Cappi, M. 2019, A&A, 628, A5

  7. [7]

    & Dupourqué, S

    Barret, D. & Dupourqué, S. 2024, A&A, 686, A133

  8. [8]

    Betancourt, M. J. & Girolami, M. 2013, Hamiltonian Monte Carlo for Hierarchi- cal Models

Show all 42 references
  1. [9]

    & Gallo, L

    Bonson, K. & Gallo, L. C. 2016, MNRAS, 458, 1927

  2. [10]

    & Boorman, P

    Buchner, J. & Boorman, P. 2024, Statistical Aspects of X-ray Spectral Analy- sis, ed. C. Bambi & A. Santangelo (Singapore: Springer Nature Singapore), 5403–5451

  3. [11]

    2014, A&A, 564, A125

    Buchner, J., Georgakakis, A., Nandra, K., et al. 2014, A&A, 564, A125

  4. [12]

    A., Steiner, J

    Choudhury, K., Garcia, J. A., Steiner, J. F., & Bambi, C. 2017, ApJ, 851, 57, arXiv:1711.02416 [astro-ph]

  5. [13]

    G., Dow, E., & Wang, Q

    Constantine, P. G., Dow, E., & Wang, Q. 2014, SIAM Journal on Scientific Com- puting, 36, A1500

  6. [14]

    2025, Nature Astronomy, 9, 36 de Plaa, J., Kaastra, J

    Cruise, M., Guainazzi, M., Aird, J., et al. 2025, Nature Astronomy, 9, 36 de Plaa, J., Kaastra, J. S., Gu, L., Mao, J., & Raassen, T. 2019, SPEX: High- Resolution Spectral Modeling and Fitting for X-ray Astronomy

  7. [15]

    H., & Gonçalves, P

    Deistler, M., Macke, J. H., & Gonçalves, P. J. 2022, Proceedings of the National Academy of Sciences, 119, e2207632119 Dupourqué, S., Barret, D., Diez, C. M., Guillot, S., & Quintin, E. 2024, A&A, 690, A317 García, J. A., Dauser, T., Ludlam, R., et al. 2022, ApJ, 926, 13

  8. [16]

    & Raftery, A

    Gneiting, T. & Raftery, A. E. 2007, Journal of the American Statistical Associa- tion, 102, 359

  9. [17]

    S., Nonnenmacher, M., & Macke, J

    Greenberg, D. S., Nonnenmacher, M., & Macke, J. H. 2019, Automatic Posterior Transformation for Likelihood-Free Inference

  10. [18]

    R., Millman, K

    Harris, C. R., Millman, K. J., van der Walt, S. J., et al. 2020, Nature, 585, 357

  11. [19]

    Hastings, W. K. 1970, Biometrika, 57, 97

  12. [20]

    2015, Deep Residual Learning for Image Recognition

    He, K., Zhang, X., Ren, S., & Sun, J. 2015, Deep Residual Learning for Image Recognition

  13. [21]

    Hinton, S. R. 2016, The Journal of Open Source Software, 1, 00045

  14. [22]

    Houck, J. C. & Denicola, L. A. 2000, in Astronomical Society of the Pacific Con- ference Series, V ol. 216, Astronomical Data Analysis Software and Systems IX, ed. N. Manset, C. Veillet, & D. Crabtree, 591

  15. [23]

    Hunter, J. D. 2007, Computing in science & engineering, 9, 90

  16. [24]

    2018, MNRAS, 475, 4739 Kaastra & Bleeker

    Ichinohe, Y ., Yamada, S., Miyazaki, N., & Saito, S. 2018, MNRAS, 475, 4739 Kaastra & Bleeker. 2016, antike und abendland, 587, A151

  17. [25]

    Kingma, D. P. & Ba, J. 2017, Adam: A Method for Stochastic Optimization

  18. [26]

    2010, in Proceedings of the 9th Python in Science Conference, ed

    McKinney, W. 2010, in Proceedings of the 9th Python in Science Conference, ed. S. van der Walt & J. Millman, 51 – 56

  19. [27]

    2013, arXiv e-prints, arXiv:1306.2307

    Nandra, K., Barret, D., Barcons, X., et al. 2013, arXiv e-prints, arXiv:1306.2307

  20. [28]

    2017, in Advances in Neural Infor- mation Processing Systems, V ol

    Papamakarios, G., Pavlakou, T., & Murray, I. 2017, in Advances in Neural Infor- mation Processing Systems, V ol. 30 (Curran Associates, Inc.)

  21. [29]

    L., Lieu, M., & Matzeu, G

    Parker, M. L., Lieu, M., & Matzeu, G. A. 2022, MNRAS, 514, 4061

  22. [30]

    2017, in NIPS AutodiffWorkshop

    Paszke, A., Gross, S., Chintala, S., et al. 2017, in NIPS AutodiffWorkshop

  23. [31]

    2011, Journal of Machine Learning Research, 12, 2825

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

  24. [32]

    2025, Experimental Astronomy, 59, 18

    Peille, P., Barret, D., Cucchetti, E., et al. 2025, Experimental Astronomy, 59, 18

  25. [33]

    A., Gúrpide, A., Bachetti, M., & Fürst, F

    Quintin, E., Webb, N. A., Gúrpide, A., Bachetti, M., & Fürst, F. 2021, MNRAS, 503, 5485

  26. [34]

    Rezende, D. J. & Mohamed, S. 2016, Variational Inference with Normalizing Flows, arXiv:1505.05770 [stat]

  27. [35]

    M., et al

    Siemiginowska, A., Burke, D., Günther, H. M., et al. 2024, ApJ Supplement Series, 274, 43, publisher: The American Astronomical Society

  28. [36]

    2006, Bayesian Analysis, 1, 833

    Skilling, J. 2006, Bayesian Analysis, 1, 833

  29. [37]

    K., Brickhouse, N

    Smith, R. K., Brickhouse, N. S., Liedahl, D. A., & Raymond, J. C. 2001, ApJ, 556, L91, arXiv: astro-ph/0106478

  30. [38]

    2020, The Journal of Open Source Software, 5, 2505

    Tejero-Cantero, A., Boelts, J., Deistler, M., et al. 2020, The Journal of Open Source Software, 5, 2505

  31. [39]

    1994, ApJ, 434, 570

    Titarchuk, L. 1994, ApJ, 434, 570

  32. [40]

    2025, X-ray spectral fitting with Monte Carlo Dropout Neural Networks

    Tutone, A., Anitra, A., Ambrosi, E., et al. 2025, X-ray spectral fitting with Monte Carlo Dropout Neural Networks

  33. [41]

    E., et al

    Virtanen, P., Gommers, R., Oliphant, T. E., et al. 2020, Nature Methods XRISM Science Team. 2020, Science with the X-ray Imaging and Spectroscopy Mission (XRISM)

  34. [42]

    & Allen, S

    Zanetta, F. & Allen, S. 2024, Scoringrules: a python library for probabilistic forecast evaluation Article number, page 11 of 16 A&A proofs:manuscript no. aa55215-25 Appendix A: Including the background Including a background spectrum in the inference using SBI is straightforw...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.