pith. machine review for the scientific record. sign in

arxiv: 2605.10719 · v1 · submitted 2026-05-11 · 🌌 astro-ph.CO · astro-ph.IM

Recognition: 2 theorem links

· Lean Theorem

Machine Learning Techniques for Astrophysics and Cosmology: Simulation-Based Inference

Leander Thiele
Authors on Pith no claims yet

Pith reviewed 2026-05-12 04:48 UTC · model grok-4.3

classification 🌌 astro-ph.CO astro-ph.IM
keywords simulation-based inferenceneural networkscosmologyastrophysicsparameter estimationlikelihood-free inferencemachine learning
0
0 comments X

The pith

Neural networks trained on forward simulations enable parameter inference in cosmology and astrophysics even when likelihoods cannot be computed directly, yet limited simulation budgets remain the central practical obstacle.

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

The paper offers a pedagogical overview of simulation-based inference for astrophysics and cosmology applications. It explains how neural networks can be trained to perform posterior estimation, likelihood estimation, or ratio estimation directly from simulated data. The review covers sequential variants, learned summary statistics, and the essential role of diagnostic checks to catch subtle failures. It surveys recent uses in the literature before identifying the scarcity of training simulations as the main barrier to broader adoption.

Core claim

Simulation-based inference enables parameter inference by training neural networks on forward simulations. It is useful both for intractable likelihoods and when posterior sampling must be fast. The basic techniques are posterior estimation, likelihood estimation, and ratio estimation. Alternatives, sequential versions, and learned summaries are available. Because failures can be subtle, diagnostics are required to validate results. Training with limited simulation budgets is the critical problem for applications to cosmology and astrophysics.

What carries the argument

The three core SBI techniques (posterior estimation, likelihood estimation, and ratio estimation) in which a neural network is trained to map simulated data to the desired statistical quantity.

If this is right

  • Diagnostic checks are mandatory because undetected failures would invalidate any downstream scientific conclusions.
  • Sequential and amortized variants of the techniques can stretch a fixed simulation budget farther than single-pass training.
  • Learned summary statistics become necessary when the raw simulated data are high-dimensional.
  • Existing applications in the cosmology and astrophysics literature already demonstrate practical feasibility.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Further reductions in required simulation count would immediately expand the range of models that can be constrained with current computing resources.
  • Cross-checks against analytic limits or semi-analytic mocks could expose hidden mismatches between simulated and observed systematics.
  • The same training-budget bottleneck is likely to appear in any domain that must rely on expensive forward models rather than closed-form likelihoods.

Load-bearing premise

Forward simulations capture enough of the relevant physics and systematics that a network trained on them will yield reliable inferences when applied to real observations.

What would settle it

An SBI model that passes all recommended diagnostics yet returns parameter constraints on real telescope data that differ substantially from those obtained by independent methods or from controlled mock tests with known inputs.

Figures

Figures reproduced from arXiv: 2605.10719 by Leander Thiele.

Figure 1
Figure 1. Figure 1: Schematic illustration of the setup. In the first situation, the likelihood itself is intractable. Let us contrast this with more traditional explicit-likelihood inference. Schematically, we can write the like￾lihood as p(x|θ) = Z Dη Dζ δ[x−m(θ,η,ζ )], (2) where m is the forward process which depends on parameters θ, nuisance param￾eters η, and initial conditions ζ (see [PITH_FULL_IMAGE:figures/full_fig_p… view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of posterior estimation. The normalizing flow directly predicts the solid red posteriors. 2.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 2.0 parameter 1.00 0.75 0.50 0.25 0.00 0.25 0.50 0.75 1.00 d ata x prior p( ) ground-truth posterior p( xo) posterior q( xo) Neural posterior estimation (NPE) targets the posterior p(θ|x) directly [16, 39, 46]. In practice, we introduce a surrogate density qφ (θ|x) which is… view at source ↗
Figure 3
Figure 3. Figure 3: Illustration of likelihood estimation. The normalizing flow predicts the orange curves from which the posterior follows after MCMC sampling. 2.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 2.0 parameter 1.00 0.75 0.50 0.25 0.00 0.25 0.50 0.75 1.00 d ata x prior p( ) ground-truth posterior p( xo) posterior q( xo) learned likelihood q(x 0) vertical orange. Thus, in contrast to NPE, MCMC sampling is needed to obtain the post… view at source ↗
Figure 4
Figure 4. Figure 4: Illustration of ratio estimation. The classifier predicts the orange curves from which the posterior follows after MCMC sampling. 2.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 2.0 parameter 1.00 0.75 0.50 0.25 0.00 0.25 0.50 0.75 1.00 d ata x prior p( ) ground-truth posterior p( xo) posterior q( xo) learned ratio r( , xo) Neural ratio estimation (NRE) targets the likelihood-to-evidence ratio r(θ,x) ≡ p(x|θ)/p(x) [20, 11… view at source ↗
Figure 5
Figure 5. Figure 5: Imperfect approxima￾tion of a box prior leads to bi￾ased inference if NPE is used in an i.i.d. situation. Multi￾plication of NPE posteriors amplifies the approximation error. It may seem daunting to choose among the collection of techniques that comprise SBI. How to decide between NPE, NLE, NRE, how to set the hyperparameters, how to compress? We only give a brief starting point and refer the reader to mor… view at source ↗
Figure 6
Figure 6. Figure 6: Example of posterior predictive checks. The observed xo should be “typical” of q(x|xo). 1.0 0.5 0.0 0.5 data x q(x xo) 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 density xo = 0.70 1.0 0.5 0.0 0.5 1.0 data x q(x xo) xo = +0.00 Many tests carry over from explicit-likelihood analyses. Data censoring can often uncover problems. In cosmological inferences, this can for example mean remov￾ing some small-scale infor… view at source ↗
Figure 7
Figure 7. Figure 7: Illustration of calibration tests. Left: ranks test, including illustrations of typical failure modes. Right: coverage tests (highest posterior density, HPD, and distance to random point, TARP). Calibration tests isolate the second failure mode, namely undertraining of the neural network. They are thus internal diagnostics without reference to the actual observed data and cannot probe for simulator misspec… view at source ↗
read the original abstract

Simulation-based inference (SBI) enables parameter inference by training neural networks on forward simulations. It is being applied both for intractable likelihoods as well as under time constraints on the posterior sampling. After motivating situations in which SBI is useful, we give a pedagogical description of the basic techniques. These are posterior, likelihood, and ratio estimation. Alternatives, sequential versions, and learned summaries are discussed briefly. We provide a brief guide to choosing among the techniques in practical scenarios. SBI needs to be verified through diagnostics since failures can be subtle but would invalidate the inference result. We explain the most common diagnostic techniques. We briefly list some recent SBI applications in the cosmology and astrophysics literature. Before concluding, we discuss current methodological challenges. We identify training with limited simulation budgets as the critical problem for applications to cosmology and astrophysics.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 2 minor

Summary. This manuscript is a pedagogical review of simulation-based inference (SBI) techniques for astrophysics and cosmology. It motivates SBI for intractable likelihoods or time-limited posterior sampling, describes the core methods of posterior estimation, likelihood estimation, and ratio estimation, briefly covers alternatives including sequential SBI and learned summaries, provides guidance on technique selection, explains standard diagnostics, lists recent applications from the literature, and identifies limited simulation budgets as the central methodological challenge.

Significance. If the prioritization of simulation-budget constraints holds after addressing the noted gaps, the review could usefully serve as an accessible entry point for cosmologists adopting SBI, synthesizing standard techniques and pointing to existing applications. Its significance is reduced by the descriptive focus without new quantitative benchmarks or explicit tests of simulation fidelity assumptions in the cited use cases.

major comments (2)
  1. [Discussion of methodological challenges] In the discussion of methodological challenges (final section before conclusion): the claim that limited simulation budgets constitute the critical problem for cosmology and astrophysics applications is not supported by any comparative analysis against other obstacles such as forward-model fidelity, baryonic physics approximations, or selection effects. The review presupposes that neural networks trained on current simulations will generalize reliably to observations once the budget is increased, without citing counter-examples or quantitative evidence from the applications listed in the preceding section.
  2. [Guide to choosing among techniques] In the section providing a guide to choosing among techniques: the guidance remains at a high level and is not explicitly linked to the specific cosmology and astrophysics applications discussed later, leaving unclear how practitioners should weigh posterior vs. ratio estimation in regimes with known simulation limitations.
minor comments (2)
  1. [Abstract] The abstract states the scope clearly but could explicitly note that the work is a review synthesizing existing methods rather than introducing new derivations or results.
  2. [Applications] Ensure that all applications referenced in the applications section include complete citations to the original papers for traceability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed comments, which will help strengthen the manuscript. We address each major comment below and indicate planned revisions.

read point-by-point responses

  1. Referee: In the discussion of methodological challenges (final section before conclusion): the claim that limited simulation budgets constitute the critical problem for cosmology and astrophysics applications is not supported by any comparative analysis against other obstacles such as forward-model fidelity, baryonic physics approximations, or selection effects. The review presupposes that neural networks trained on current simulations will generalize reliably to observations once the budget is increased, without citing counter-examples or quantitative evidence from the applications listed in the preceding section.

    Authors: We agree that the manuscript would benefit from a more explicit discussion of why limited simulation budgets are prioritized as the central challenge. While other issues such as forward-model fidelity are important, they are typically mitigated by improvements to the simulations themselves; SBI methods specifically address the computational expense of generating enough simulations for reliable inference. In the revised version we will expand the methodological challenges section to briefly compare these obstacles, cite literature quantifying their relative impacts in SBI contexts, and clarify the assumption of adequate simulation fidelity as a prerequisite (consistent with the reviewed applications). revision: yes

  2. Referee: In the section providing a guide to choosing among techniques: the guidance remains at a high level and is not explicitly linked to the specific cosmology and astrophysics applications discussed later, leaving unclear how practitioners should weigh posterior vs. ratio estimation in regimes with known simulation limitations.

    Authors: We accept that the guide would be more useful if it included explicit connections to the applications. The current guidance is deliberately general to remain broadly applicable, but we will revise the section to add cross-references and brief examples drawn from the applications listed later in the paper. These will illustrate how simulation-budget constraints influenced choices between posterior and ratio estimation in specific cosmology and astrophysics studies. revision: yes

Circularity Check

0 steps flagged

Descriptive review paper with no derivations or self-referential reductions

full rationale

This is a pedagogical review that motivates SBI use cases, describes standard techniques (posterior/likelihood/ratio estimation), covers diagnostics and applications from external literature, and states an observational conclusion about simulation budgets as the critical challenge. No original equations, fitted parameters, uniqueness theorems, or ansatzes are introduced. The central identification of the budget problem is presented as a field-level observation rather than derived from any internal construction or self-citation chain. All methods are attributed to prior work, satisfying the self-contained criterion with no load-bearing circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This review paper introduces no new free parameters, axioms, or invented entities; it describes established machine learning methods applied to cosmology.

pith-pipeline@v0.9.0 · 5429 in / 1039 out tokens · 52912 ms · 2026-05-12T04:48:38.029478+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

64 extracted references · 64 canonical work pages · 3 internal anchors

  1. [1]

    M. S. Albergo, N. M. Boffi, and E. Vanden-Eijnden. Stochastic Interpolants: A Unifying Framework for Flows and Diffusions.arXiv e-prints, page arXiv:2303.08797, Mar. 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  2. [2]

    Alsing, T

    J. Alsing, T. Charnock, S. Feeney, and B. Wandelt. Fast likelihood-free cos- mology with neural density estimators and active learning.Mon. Not. Roy. Astron. Soc., 488(3):4440–4458, Sept. 2019

    work page 2019

  3. [3]

    Alsing, B

    J. Alsing, B. Wandelt, and S. Feeney. Massive optimal data compression and density estimation for scalable, likelihood-free inference in cosmology.Mon. Not. Roy. Astron. Soc., 477(3):2874–2885, July 2018

    work page 2018

  4. [4]

    Bairagi, B

    A. Bairagi, B. Wandelt, and F. Villaescusa-Navarro. How many simu- lations do we need for simulation-based inference in cosmology?, 2025. arXiv:2503.13755

    work page arXiv 2025

  5. [5]

    Charnock, G

    T. Charnock, G. Lavaux, and B. D. Wandelt. Automatic physical inference with information maximizing neural networks.Phys. Rev. D, 97(8):083004, Apr. 2018

    work page 2018

  6. [6]

    Y . Chen, M. Goldstein, M. Hua, M. S. Albergo, N. M. Boffi, and E. Vanden- Eijnden. Probabilistic Forecasting with Stochastic Interpolants and F ¨ollmer Processes.arXiv e-prints, page arXiv:2403.13724, Mar. 2024

    work page arXiv 2024

  7. [7]

    M. Dax, S. R. Green, J. Gair, J. H. Macke, A. Buonanno, and B. Sch ¨olkopf. Real-Time Gravitational Wave Science with Neural Posterior Estimation. Phys. Rev. Lett., 127(24):241103, Dec. 2021

    work page 2021

  8. [8]

    Goncalves, and Jakob H

    M. Deistler, J. Boelts, P. Steinbach, G. Moss, T. Moreau, M. Gloeckler, P. L. C. Rodrigues, J. Linhart, J. K. Lappalainen, B. K. Miller, P. J. Gonc ¸alves, J.-M. Lueckmann, C. Schr ¨oder, and J. H. Macke. Simulation-Based Inference: A Practical Guide.arXiv e-prints, page arXiv:2508.12939, Aug. 2025

    work page arXiv 2025

  9. [9]

    Delaunoy, J

    A. Delaunoy, J. Hermans, F. Rozet, A. Wehenkel, and G. Louppe. Towards Reliable Simulation-Based Inference with Balanced Neural Ratio Estimation. arXiv e-prints, page arXiv:2208.13624, Aug. 2022

    work page arXiv 2022

  10. [10]

    Delaunoy, B

    A. Delaunoy, B. K. Miller, P. Forr ´e, C. Weniger, and G. Louppe. Balancing Simulation-based Inference for Conservative Posteriors.arXiv e-prints, page arXiv:2304.10978, Apr. 2023

    work page arXiv 2023

  11. [11]

    Durkan, I

    C. Durkan, I. Murray, and G. Papamakarios. On Contrastive Learning for Likelihood-free Inference.arXiv e-prints, page arXiv:2002.03712, Feb. 2020

    work page arXiv 2002

  12. [12]

    Falkiewicz, N

    M. Falkiewicz, N. Takeishi, I. Shekhzadeh, A. Wehenkel, A. Delaunoy, G. Louppe, and A. Kalousis. Calibrating Neural Simulation-Based In- ference with Differentiable Coverage Probability.arXiv e-prints, page arXiv:2310.13402, Oct. 2023

    work page arXiv 2023

  13. [13]

    Fluri, T

    J. Fluri, T. Kacprzak, A. Lucchi, A. Schneider, A. Refregier, and T. Hofmann. Full w CDM analysis of KiDS-1000 weak lensing maps using deep learning. Phys. Rev. D, 105(8):083518, Apr. 2022

    work page 2022

  14. [14]

    Gatti, G

    M. Gatti, G. Campailla, N. Jeffrey, L. Whiteway, A. Porredon, J. Prat, J. Williamson, M. Raveri, B. Jain, V . Ajani, G. Giannini, M. Yamamoto, C. Zhou, J. Blazek, D. Anbajagane, S. Samuroff, T. Kacprzak, et al. Dark 16 Leander Thiele Energy Survey Year 3 results: Simulation-based cosmological inference with wavelet harmonics, scattering transforms, and mo...

    work page 2025

  15. [15]

    Grazian and Y

    C. Grazian and Y . Fan. A review of Approximate Bayesian Computation meth- ods via density estimation: inference for simulator-models.arXiv e-prints, page arXiv:1909.02736, Sept. 2019

    work page arXiv 1909

  16. [16]

    Greenberg, M

    D. Greenberg, M. Nonnenmacher, and J. Macke. Automatic posterior transfor- mation for likelihood-free inference. InProceedings of the 36th International Conference on Machine Learning, pages 2404–2414, 2019

    work page 2019

  17. [17]

    Gretton, K

    A. Gretton, K. M. Borgwardt, M. J. Rasch, B. Sch ¨olkopf, and A. Smola. A kernel two-sample test.Journal of Machine Learning Research, 13(25):723– 773, 2012

    work page 2012

  18. [18]

    C. Hahn, P. Lemos, L. Parker, B. R ´egaldo-Saint Blancard, M. Eickenberg, S. Ho, J. Hou, E. Massara, C. Modi, A. Moradinezhad Dizgah, and D. Spergel. Cosmological constraints from non-Gaussian and nonlinear galaxy clustering using the SIMBIG inference framework.Nature Astronomy, 8:1457–1467, Nov. 2024

    work page 2024

  19. [19]

    A. F. Heavens, R. Jimenez, and O. Lahav. Massive lossless data compression and multiple parameter estimation from galaxy spectra.Mon. Not. Roy. Astron. Soc., 317(4):965–972, Oct. 2000

    work page 2000

  20. [20]

    Hermans, V

    J. Hermans, V . Begy, and G. Louppe. Likelihood-free MCMC with amortized approximate ratio estimators. InProceedings of the 37th International Con- ference on Machine Learning, pages 4239–4248, 2020

    work page 2020

  21. [21]

    Hermans, A

    J. Hermans, A. Delaunoy, F. Rozet, A. Wehenkel, V . Begy, and G. Louppe. A trust crisis in simulation-based inference? Beware, your posterior approxima- tions can be unfaithful.Transactions on Machine Learning Research, 2022

    work page 2022

  22. [22]

    Hikida, A

    Y . Hikida, A. Bharti, N. Jeffrey, and F.-X. Briol. Multilevel neural simulation- based inference.arXiv e-prints, page arXiv:2506.06087, June 2025

    work page arXiv 2025

  23. [23]

    Homer, O

    J. Homer, O. Friedrich, and D. Gruen. Simulation-based inference has its own Dodelson-Schneider effect (but it knows that it does), 2024. arXiv:2412.02311

    work page arXiv 2024

  24. [24]

    Jeffrey, J

    N. Jeffrey, J. Alsing, and F. Lanusse. Likelihood-free inference with neural compression of DES SV weak lensing map statistics.Mon. Not. Roy. Astron. Soc., 501(1):954–969, Feb. 2021

    work page 2021

  25. [25]

    Jeffrey, L

    N. Jeffrey, L. Whiteway, M. Gatti, J. Williamson, J. Alsing, A. Porredon, J. Prat, C. Doux, B. Jain, C. Chang, T.-Y . Cheng, T. Kacprzak, P. Lemos, et al. Dark energy survey year 3 results: likelihood-free, simulation-based wCDM inference with neural compression of weak-lensing map statistics.Mon. Not. Roy. Astron. Soc., 536(2):1303–1322, Jan. 2025

    work page 2025

  26. [26]

    H. Jia. Cosmological analysis with calibrated neural quantile estimation and approximate simulators, 2024. arXiv:2411.14748

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  27. [27]

    H. Jia. Simulation-based inference with quantile regression. InProceedings of the 41st International Conference on Machine Learning, pages 21731–21752, 2024. Simulation-Based Inference 17

    work page 2024

  28. [28]

    Variational Inference with Normalizing Flows

    D. Jimenez Rezende and S. Mohamed. Variational Inference with Normalizing Flows.arXiv e-prints, page arXiv:1505.05770, May 2015

    work page Pith review arXiv 2015

  29. [29]

    Karchev, M

    K. Karchev, M. Grayling, B. M. Boyd, R. Trotta, K. S. Mandel, and C. Weniger. SIDE-real: Supernova Ia Dust Extinction with truncated marginal neural ratio estimation applied to real data.Mon. Not. Roy. Astron. Soc., 530(4):3881–3896, June 2024

    work page 2024

  30. [30]

    A. N. Krouglova, H. R. Johnson, B. Confavreux, M. Deistler, and P. J. Gonc ¸alves. Multifidelity simulation-based inference for computationally ex- pensive simulators, 2025. arXiv:2502.08416

    work page arXiv 2025

  31. [31]

    Lemos, A

    P. Lemos, A. Coogan, Y . Hezaveh, and L. Perreault-Levasseur. Sampling- Based Accuracy Testing of Posterior Estimators for General Inference.40th International Conference on Machine Learning, 202:19256–19273, Jan. 2023

    work page 2023

  32. [32]

    Lemos, L

    P. Lemos, L. Parker, C. Hahn, S. Ho, M. Eickenberg, J. Hou, E. Massara, C. Modi, A. M. Dizgah, B. R.-S. Blancard, D. Spergel, and SimBIG Col- laboration. Field-level simulation-based inference of galaxy clustering with convolutional neural networks.Phys. Rev. D, 109(8):083536, Apr. 2024

    work page 2024

  33. [33]

    Lemos, S

    P. Lemos, S. Sharief, N. Malkin, S. Salhi, C. Stone, L. Perreault-Levasseur, and Y . Hezaveh. PQMass: Probabilistic Assessment of the Quality of Generative Models using Probability Mass Estimation. InThe Thirteenth International Conference on Learning Representations, page 7567, Jan. 2025

    work page 2025

  34. [34]

    Linhart, A

    J. Linhart, A. Gramfort, and P. L. C. Rodrigues. L-C2ST: Local Diagnostics for Posterior Approximations in Simulation-Based Inference.arXiv e-prints, page arXiv:2306.03580, June 2023

    work page arXiv 2023

  35. [35]

    Lipman, R

    Y . Lipman, R. T. Q. Chen, H. Ben-Hamu, M. Nickel, and M. Le. Flow match- ing for generative modeling. InThe Eleventh International Conference on Learning Representations, 2023

    work page 2023

  36. [36]

    X. Liu, C. Gong, and Q. Liu. Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow.arXiv e-prints, page arXiv:2209.03003, Sept. 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  37. [37]

    Revisiting Classifier Two-Sample Tests

    D. Lopez-Paz and M. Oquab. Revisiting classifier two-sample tests, 2018. arXiv:1610.06545

    work page Pith review arXiv 2018

  38. [38]

    Lueckmann, G

    J.-M. Lueckmann, G. Bassetto, T. Karaletsos, and J. H. Macke. Likelihood- free inference with emulator networks. InProceedings of the 1st Symposium on Advances in Approximate Bayesian Inference, pages 32–53, 2019

    work page 2019

  39. [39]

    Lueckmann, P

    J.-M. Lueckmann, P. J. Goncalves, G. Bassetto, K. ¨Ocal, M. Nonnenmacher, and J. H. Macke. Flexible statistical inference for mechanistic models of neural dynamics. InAdvances in Neural Information Processing Systems, volume 30, pages 1289–1299, 2017

    work page 2017

  40. [40]

    Marin, P

    J.-M. Marin, P. Pudlo, C. P. Robert, and R. Ryder. Approximate Bayesian Computational methods.arXiv e-prints, page arXiv:1101.0955, Jan. 2011

    work page arXiv 2011

  41. [41]

    Marjoram, J

    P. Marjoram, J. Molitor, V . Plagnol, and S. Tavar ´e. Markov chain monte carlo without likelihoods.Proceedings of the National Academy of Sciences, 100(26):15324–15328, 2003. 18 Leander Thiele

    work page 2003

  42. [42]

    Mishra-Sharma and K

    S. Mishra-Sharma and K. Cranmer. Neural simulation-based inference ap- proach for characterizing the Galactic Centerγ-ray excess.Phys. Rev. D, 105(6):063017, Mar. 2022

    work page 2022

  43. [43]

    Modi and O

    C. Modi and O. H. E. Philcox. Hybrid SBI or how I learned to stop worrying and learn the likelihood, 2023. arXiv:2309.10270

    work page arXiv 2023

  44. [44]

    N. A. Montel. Basics of Simulation-Based Inference.https: //wwwmpa.mpa-garching.mpg.de/˜komatsu/lecturenotes/ Noemi_Anau_Montel_on_SBI.pdf, Nov. 2025. accessed: 4/30/2026

    work page 2025

  45. [45]

    C. P. Novaes, L. Thiele, J. Armijo, S. Cheng, J. A. Cowell, G. A. Marques, E. G. M. Ferreira, M. Shirasaki, K. Osato, and J. Liu. Cosmology from HSC Y1 weak lensing data with combined higher-order statistics and simulation- based inference.Phys. Rev. D, 111(8):083510, Apr. 2025

    work page 2025

  46. [46]

    Papamakarios and I

    G. Papamakarios and I. Murray. Fastε-free inference of simulation models with Bayesian conditional density estimation. InAdvances in Neural Informa- tion Processing Systems, volume 29, pages 1036–1044, 2016

    work page 2016

  47. [47]

    Papamakarios, E

    G. Papamakarios, E. Nalisnick, D. J. Rezende, S. Mohamed, and B. Laksh- minarayanan. Normalizing flows for probabilistic modeling and inference. Journal of Machine Learning Research, 22(57):1–64, 2021

    work page 2021

  48. [48]

    Papamakarios, D

    G. Papamakarios, D. Sterratt, and I. Murray. Sequential neural likelihood: Fast likelihood-free inference with autoregressive flows. InProceedings of the 22nd International Conference on Artificial Intelligence and Statistics, pages 837–848, 2019

    work page 2019

  49. [49]

    J. K. Pritchard, M. T. Seielstad, A. Perez-Lezaun, and M. W. Feldman. Popula- tion growth of human Y chromosomes: a study of Y chromosome microsatel- lites.Mol Biol Evol, 16(12):1791–1798, Dec. 1999

    work page 1999

  50. [50]

    Ruhlmann, P

    P.-L. Ruhlmann, P. L. C. Rodrigues, M. Arbel, and F. Forbes. Flow Matching for Robust Simulation-Based Inference under Model Misspecification.arXiv e-prints, page arXiv:2509.23385, Sept. 2025

    work page arXiv 2025

  51. [51]

    S. A. Sisson, Y . Fan, and M. A. Beaumont. Overview of Approximate Bayesian Computation.arXiv e-prints, page arXiv:1802.09720, Feb. 2018

    work page arXiv 2018

  52. [52]

    E. G. Tabak and C. V . Turner. A Family of Nonparametric Density Estimation Algorithms.Commun. Pure Appl. Math., 66(2):145–164, 2013

    work page 2013

  53. [53]

    E. G. Tabak and E. Vanden-Eijnden. Density estimation by dual ascent of the log-likelihood.Commun. Math. Sci., 8(1):217–233, 2010

    work page 2010

  54. [54]

    Scuola Medica Salernitana,

    S. Talts, M. Betancourt, D. Simpson, A. Vehtari, and A. Gelman. Validating Bayesian Inference Algorithms with Simulation-Based Calibration.arXiv e- prints, page arXiv:1804.06788, Apr. 2018

    work page arXiv 2018

  55. [55]

    Tejero-Cantero, J

    A. Tejero-Cantero, J. Boelts, M. Deistler, J.-M. Lueckmann, C. Durkan, P. J. Gonc ¸alves, D. S. Greenberg, and J. H. Macke. sbi: A toolkit for simulation- based inference.Journal of Open Source Software, 5(52):2505, 2020

    work page 2020

  56. [56]

    Thiele, A

    L. Thiele, A. E. Bayer, and N. Takeishi. Simulation-Efficient Cosmological Inference with Multi-Fidelity SBI.arXiv e-prints, page arXiv:2507.00514, July 2025. Simulation-Based Inference 19

    work page arXiv 2025

  57. [57]

    Thiele, E

    L. Thiele, E. Massara, A. Pisani, C. Hahn, D. N. Spergel, S. Ho, and B. Wan- delt. Neutrino Mass Constraint from an Implicit Likelihood Analysis of BOSS V oids.Astrophys. J., 969(2):89, July 2024

    work page 2024

  58. [58]

    von Wietersheim-Kramsta, K

    M. von Wietersheim-Kramsta, K. Lin, N. Tessore, B. Joachimi, A. Loureiro, R. Reischke, and A. H. Wright. KiDS-SBI: Simulation-based inference anal- ysis of KiDS-1000 cosmic shear.Astron. Astrophys., 694:A223, Feb. 2025

    work page 2025

  59. [59]

    B. Wang, J. Leja, V . A. Villar, and J. S. Speagle. SBI ++: Flexible, Ultra-fast Likelihood-free Inference Customized for Astronomical Applications.Astro- phys. J. Lett., 952(1):L10, July 2023

    work page 2023

  60. [60]

    B. Y . Wang and L. Thiele. Set-based Implicit Likelihood Inference of Galaxy Cluster Mass.arXiv e-prints, page arXiv:2507.20378, July 2025

    work page arXiv 2025

  61. [61]

    Wehenkel, J

    A. Wehenkel, J. L. Gamella, O. Sener, J. Behrmann, G. Sapiro, J.-H. Jacobsen, and M. Cuturi. Addressing Misspecification in Simulation-based Inference through Data-driven Calibration.arXiv e-prints, page arXiv:2405.08719, May 2024

    work page arXiv 2024

  62. [62]

    J. B. Wildberger, M. Dax, S. Buchholz, S. R. Green, J. Macke, and B. Sch ¨olkopf. Flow Matching for Scalable Simulation-Based Inference. In Machine Learning for Astrophysics, page 34, July 2023

    work page 2023

  63. [63]

    Zhang, C

    G. Zhang, C. Modi, and O. H. E. Philcox. Modeling galaxy surveys with hybrid SBI, 2025. arXiv:2505.13591

    work page arXiv 2025

  64. [64]

    Zhang, T

    K. Zhang, T. Jayasinghe, and J. Bloom. Stellar Spectra Fitting with Amortized Neural Posterior Estimation and nbi. InMachine Learning for Astrophysics, page 39, July 2023

    work page 2023