Pith. sign in

REVIEW 4 major objections 6 minor 291 references

A neural simulation-based inference pipeline recovers supernova population parameters and compares competing dust/scatter models without an explicit likelihood, finding log10 Bayes factor 1.9 against a two-population model.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-03 00:32 UTC pith:WVVYWFPV

load-bearing objection Solid SBI pipeline undermined by an NRE model-comparison section that doesn't compute the Bayes factors it claims. the 4 major comments →

arxiv 2607.28725 v1 pith:WVVYWFPV submitted 2026-07-30 astro-ph.CO

A Simulation Based Inference Approach to Modelling of Type Ia Supernova Populations

B. Popovic , M. Grayling , M. O'Callaghan , M. Smith , B. M. Boyd , K. Mandel , Philliam Wiseman , B. Carreres
show 6 more authors
N. Shiamtanis D. Scolnic E. Charleton J. Erceg A. Smith Y. Murakami
This is my paper
classification astro-ph.CO
keywords Type Ia supernovaedust extinctionR_Vsimulation-based inferenceneural posterior estimationneural ratio estimationBayes factormass step
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper argues that a neural-network-based simulation inference pipeline can recover supernova population parameters and compare competing dust/scatter models without an explicit likelihood. Built around the code Stjörnumál, it trains neural networks on simulated survey data to approximate posteriors and to estimate Bayes factors directly between proposed models. Applied to a high-redshift sample of 2,201 Type Ia supernovae, the pipeline finds moderate evidence against a two-population model (log10 Bayes factor = 1.9, mixture fraction 0.8) and a mass-dependent dust-law difference ΔE(R_V)=0.8 in the nominal model—smaller than earlier SALT-based estimates. The key payoff: if this approach holds, systematic model testing of supernova scatter becomes routine, which matters because the remaining standardization scatter is a dominant source of systematic uncertainty in cosmology.

Core claim

Stjörnumál demonstrates that an intractable SN Ia population likelihood—including survey selection, dust reddening, intrinsic scatter, and a host-mass step—can be emulated by neural posterior estimation trained on importance-sampled simulations, with a paired neural ratio estimator supplying Bayes factors between seven proposed models. On the high-redshift sample, the posterior for the nominal two-dust-population model is well calibrated, the high/low-mass host R_V difference shrinks to ΔE(R_V)=0.8, and the two-population Banana Split model is disfavoured at log10 Bayes factor +1.9, corresponding to a mixture fraction of 0.8 that favours a predominantly single population at high redshift. Th

What carries the argument

The work is carried by a two-stage simulation-based inference architecture. A large simulation bank of survey-realistic mock supernovae is generated with flat and exponential distributions for the underlying parameters, then reweighted by importance sampling to target any model; the network conditions on per-SN data including stretch, colour, brightness, their errors, redshift, host mass, and a precomputed residual distance μ_RES. Neural posterior estimation (NPE) then learns the mapping from these summaries to population-level parameters (c_int, β_int, R_V, E(B-V), γ, σ_int, x1 moments), while neural ratio estimation (NRE) trains a binary classifier to approximate the Bayes factor p(data|M1

Load-bearing premise

The standardisation coefficients α=0.145 and β=3.1 are fixed in advance when building the residual-distance summary; Appendix B shows that shifting β by 0.2 moves the inferred intrinsic colour-luminosity slope and mean R_V outside their 1σ ranges, so all population parameters and Bayes factors are conditional on these nuisance values.

What would settle it

Re-run the pipeline on the same supernova sample while allowing β_SALT (and α_SALT) to vary—for example, by recomputing μ_RES for each proposed value in an outer MCMC or by building the summary inside the simulation loop—and check whether the log10 Bayes factor against Banana Split drops below ~1 or the nominal ΔR_V shifts by more than ~0.1–0.2. If it does, the central conclusions are not robust to the fixed-nuisance choice. A second, complementary check: run the same model on a low-redshift sample and see whether the mixture fraction f_mix moves away from 0.8.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Population-level SN Ia scatter models can now be fit and compared in hours instead of days or weeks (100,000 mock datasets simulated in ~5 min, network trained in ~4 h), making routine systematic checking feasible.
  • With the current high-redshift sample, two-population models are disfavoured (log10 BF = 1.9, f_mix = 0.8), so existing high-z data do not require a second SN Ia population to explain the stretch-colour relation.
  • The nominal model's host-mass-dependent dust law difference shrinks to ΔE(R_V)=0.8, and the logistic-R_V variant gives 0.44, moving SALT-based estimates closer to values from full light-curve fitting approaches.
  • The inferred achromatic mass step γ=0.015±0.01 is within 2σ of zero, suggesting the observed mass step is largely a colour-dependent (dust) effect under the nominal model.
  • Results are robust to a mismatch of |Δw|<0.10 in dark-energy equation of state, meaning the population parameters are not strongly sensitive to the assumed cosmology within that range.

Where Pith is reading between the lines

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

  • Because the code is public, the same NRE classifier could be run on any proposed new model class, turning the pipeline into a fast anomaly-detection device: any model that beats the nominal one at high Bayes factor would immediately flag where current parametrisations fail.
  • The strong β_SALT dependence of β_int and R_V means the quoted Bayes factors are conditional on the fixed standardisation coefficients; a fully joint analysis that marginalises over α and β could plausibly shift the 1.9 log-unit preference for the single-population model.
  • The tension between χ²/ν and Bayes-factor rankings hints that the two metrics are sensitive to different features (e.g., the single-R_V model fares badly on χ² but not on BF); a deliberately composite diagnostic—χ² plus BF—would be more informative than either alone.
  • Adding a large low-redshift sample, where the stretch distribution is broader and bimodality signatures are stronger, should sharpen the two-population test; the paper notes its infrastructure already supports redshift-dependent parameter evolution, so this is directly testable.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. Stjörnumál is presented as a simulation-based inference (SBI) pipeline for constraining intrinsic and extrinsic SN Ia population parameters from SALT light-curve fits, replacing the MCMC in Dust2Dust with neural posterior estimation (NPE) and adding neural ratio estimation (NRE) for model comparison. The pipeline is trained on SNANA simulations of the DES-Dovekie sample (2201 SNe), with importance-sampled simulation banks, and is applied to seven models (Nominal, No Mass Step, Logistic RV, Single RV, 2 Colour, Banana Split, Banana Less Split). The paper reports well-calibrated posteriors from TARP/SBC/PPC, a claimed Bayes factor log10(BF)=1.9 (3.4 for Banana Less Split) against two-population models, a nominal-model ΔE(R_V)=0.8, and robustness to |Δw|<0.10. Public code is provided.

Significance. If the methodological claims held, Stjörnumál would be a valuable, fast and reproducible framework for comparing SN Ia scatter models, with a strong validation suite (TARP, SBC, PPC) and public code. The use of realistic SNANA simulations and the explicit comparison of a broad model set are strengths. However, the central model-comparison result rests on an NRE objective that does not estimate the claimed Bayes factor, and the cosmology/β_SALT robustness claim is not supported for β_int. These are load-bearing issues that need correction before the model-comparison and ΔR_V conclusions can be accepted.

major comments (4)
  1. [Section 6.2, Eq. (14)-(15)] The NRE objective is not a Bayes factor. Eq. (14) defines r_phi(θ,S_ω(d)) ≈ p(d|M1)/p(d|M2), but the right-hand side is a marginal likelihood ratio independent of θ. Eq. (15) trains a binary classifier on joint samples p1(d,θ)=p(θ|M1)p(d|θ,M1) and p2(d,θ)=p(θ|M2)p(d|θ,M2). The Bayes-optimal logit is log[p1(d,θ)/p2(d,θ)], a joint-density ratio that still depends on θ. No marginalization over θ is described; temperature scaling and repeating the run five times do not convert this into an evidence ratio. Consequently Table 3's recovery checks and Table 4's log(Z) values, including log10BF=1.9 and 3.4, are not Bayes factors as claimed. This is load-bearing for the model-comparison conclusions. The authors must either implement a correct estimator (e.g., a data-only classifier or explicit marginalization) or relabel the results and soften the claims accordingly.
  2. [Section 7.3 / Appendix B] The claimed robustness to |Δw|<0.10 is contradicted by the reported strong dependence of β_int μ on w (Figure 10) and by the β_SALT sensitivity tests in Appendix B, where β_int μ and R_V μ are inferred outside the 1σ value at Δβ_SALT=0.2. Since β_SALT is fixed to 3.1 when constructing μ_RES (Section 4), the inferred population parameters and the headline ΔE(R_V)=0.8 are conditional on this nuisance choice. The abstract's unqualified 'robust to |Δw|<0.10' should be restricted to parameters other than β_int, and the systematic error from the assumed β_SALT should be propagated into the quoted uncertainties.
  3. [Section 8.2] The headline ΔE(R_V)=0.8 is the difference between truncated-normal expectation values, with the high-mass R_V distribution truncated at R_V=1.2. Using the untruncated posterior means in Table 4 gives a different ΔR_V. Because the truncation choice substantially affects the result, the paper should quantify how ΔE(R_V) varies with the lower limit and justify the 1.2 cutoff before comparing with BayeSN values. As written, the comparison is not apples-to-apples and the 'discrepant ΔR_V remains' discussion rests on a convention that is partly ad hoc.
  4. [Section 7.2 / Figure 8] The SNANA recovery test shows a >2σ bias in c_int σ, consistently overestimated by 0.008, and a ~1.5σ bias in β_int μ. While the paper acknowledges this, the claim of 'fast and accurate posterior inference' is weakened by a realistic-simulation test that fails for one of the key intrinsic-scatter parameters. This bias should be included in the error budget or mitigated before drawing astrophysical conclusions from the posterior means, particularly for the c_int distribution.
minor comments (6)
  1. [Table 4] The table is difficult to parse: rows contain multiple values per model, some entries appear misaligned (e.g., the σ_βint row contains '2.259(0.14)'), and the use of parentheses for uncertainties is inconsistent with the caption. Please reformat with clear column separations and a note on what each pair represents.
  2. [Table 3 caption] 'A negative value indicates preference for the correct model' is confusing. Specify the sign convention for ln(Z) and define whether positive log(Z) always means preference for the first model.
  3. [Section 4] The abbreviation 'NSE' is used for the neural summary statistic but is not defined. Define it at first use.
  4. [Section 8.2] 'The Logistic model gives a ΔR_V =0.44, with an overall R_V =0.44' appears to contain a typo; clarify which quantity is 0.44.
  5. [Section 7.3 / Appendix B captions] The captions refer to 'the simulated cosmology' when varying β and α; these are nuisance parameters, not cosmology. Use 'assumed β_SALT' and 'assumed α_SALT'.
  6. [Introduction] 'Carreres et al. in prep.' is cited in the text but not included in the reference list; add the reference or remove the citation.

Circularity Check

0 steps flagged

No significant circularity: population parameters and Bayes factors are fitted/compared on simulations, not derived from the target quantities; the NRE joint-vs-marginal ratio concern is a correctness issue, not input-output circularity.

full rationale

The paper's derivation chain is not circular in the sense of the rubric. Stjörnumál trains NPE on SNANA simulations with parameter sets drawn from priors, then fits the DES-Dovekie data; the posterior values in Table 4 are simulation-based fits, not quantities defined in terms of the same data summary they predict. The μ_RES summary is constructed with fixed α_SALT=0.145 and β_SALT=3.1, and Appendix B explicitly shows that inferred β_int and R_V shift with that assumption; this is an acknowledged conditional dependence, not a fitted parameter renamed as a prediction. Self-citations to Dust2Dust (Popovic et al. 2023) and to the SALT surface (Popovic et al. 2025a) supply prior methodology/components, but the new inference and model comparisons do not reduce to those citations by construction; validation includes external SNANA recovery and data-level χ² metrics. The NRE concern raised by the skeptic is substantive but is a mathematical-support issue: Eq. (15) trains a classifier on joint p(d,θ|M) while Eq. (14) claims a marginal p(d|M1)/p(d|M2); if the code does not marginalize over θ, the reported log10 BF values are not the claimed Bayes factors. That is an internal correctness/calibration problem, not a case of an output being equivalent to its input by definition. Minor self-citations are present but none is load-bearing, so the circularity score is low.

Axiom & Free-Parameter Ledger

8 free parameters · 8 axioms · 0 invented entities

The central inference rests on the fidelity of SNANA simulations and the adequacy of the SALT-based summary statistics. The freedom to choose model parameterisations (logistic R_V, mixture splits) is exercised but transparently described. No new physical entities are introduced.

free parameters (8)
  • c_int μ, σ = μ=-0.085, σ=0.041
    Nominal-model intrinsic colour distribution (Table 4).
  • R_V μ (low/high mass) = 3.081, 1.769
    Nominal-model dust law mean split at 10 M_sun (Table 4).
  • R_V σ (low/high mass) = 0.618, 0.965
    Nominal-model dust law scatter split at 10 M_sun (Table 4).
  • β_int μ, σ = μ=2.371, σ=0.171
    Nominal-model intrinsic colour-luminosity slope (Table 4).
  • E(B-V) τ = 0.153
    Exponential dust column scale (Table 4).
  • γ (mass step) = 0.015
    Achromatic magnitude step between host-mass bins (Table 4).
  • σ_int (grey scatter) = 0.055
    Additional intrinsic grey scatter (Table 4).
  • x1 μ, σ = μ=-0.264, σ=1.057
    Nominal stretch distribution (Table 4).
axioms (8)
  • domain assumption SALT surface (Popovic et al. 2025a) accurately models SN Ia spectral energy distributions.
    Used to generate SALT parameters in simulations and data; if wrong, all population inferences are biased.
  • domain assumption SNANA DES simulations (Vincenzi et al. 2024) reproduce survey selection effects, cadence, and noise.
    Section 2.2; training simulations must match the real survey for the neural posterior to be valid.
  • domain assumption μ_RES computed with fixed α=0.145, β=3.1 is a sufficient summary for the population parameters.
    Section 4; the architecture cannot re-sample α and β, and Appendix B shows parameter shifts with β_SALT.
  • domain assumption The dust model parameterisation (Gaussian c_int, exponential E(B-V), Gaussian R_V) describes SN Ia scatter.
    Equations 3-5; from Brout & Scolnic (2021), Popovic et al. (2023); adopted as the nominal model.
  • domain assumption Host galaxy stellar mass split at 10 M_sun separates dust populations.
    Table 1; standard in the literature but not derived in this paper.
  • ad hoc to paper Logistic R_V parametrisation (Eq. 9) with L, k, σ is a useful description of the R_V population.
    Introduced in Popovic et al. (2024a) and adopted here; no physical derivation.
  • ad hoc to paper Mixture models require non-overlapping x1 priors ([-2,0) and (0,2]) to avoid degeneracy.
    Section 5.5; this constrains the population split and may bias f_mix.
  • ad hoc to paper γ added ad hoc to m_B and μ_RES after importance sampling is a valid implementation of the mass step.
    Section 4.1; acknowledged in Section 9.4 as 'ad-hoc addition'.

pith-pipeline@v1.3.0-alltime-deepseek · 23555 in / 12436 out tokens · 117828 ms · 2026-08-03T00:32:56.265282+00:00 · methodology

0 comments
read the original abstract

Type Ia Supernovae (SNe Ia) are prominent cosmological probes, utilising a standardisation process to reduce their observed scatter to $\sim0.15$ mag. A growing number of models seek to explain this remaining intrinsic scatter, based on a diversity of dust properties and possible connections to the progenitor systems. Inference of new models has been limited due to the cost of simulations and attendant complexity. Here, we present Stj\"ornum\'al, a simulation based inference pipeline to infer intrinsic and extrinsic parameters of SNe Ia, an upgrade to previous SN Ia modelling attempts with SALT, e.g. Dust2Dust. Stj\"ornum\'al provides fast and accurate posterior inference via Neural Posterior Estimation, integrated model comparison with Neural Ratio Estimation, and overall significant speed and quality-of-life upgrades. We fit the Dark Energy Survey (DES) 5-year SN sample, finding good agreement with previously-published dust model parameters for DES5YR. We test 7 models of SN Ia behaviour, finding that more data is needed to break degeneracies between $R_V$ models, but sufficient to evidence ($\log(10)~\textrm{Bayes Factor} = +1.9$, $f_{\rm mix} = 0.8$) against two populations of SNe Ia at high-redshift. We employ a combination of frequentist $\chi^2$ metrics and Bayesian model comparison to make model determinations, finding neither are sufficient on their own to properly compare models. For our nominal model, we find a smaller $\Delta R_V = 0.8$ for our nominal model than previous SALT-based attempts. We test our model for consistency against our assumed cosmology, and find our results are robust to $|\Delta w| < 0.10$. The code is publicly available at https://github.com/bap37/Stjornumal, and presents an opportunity to flexibly and rapidly test potential models of SNe Ia scatter in a common framework.

Figures

Figures reproduced from arXiv: 2607.28725 by A. Smith, B. Carreres, B. M. Boyd, B. Popovic, D. Scolnic, E. Charleton, J. Erceg, K. Mandel, M. Grayling, M. O'Callaghan, M. Smith, N. Shiamtanis, Philliam Wiseman, Y. Murakami.

Figure 1
Figure 1. Figure 1: The DES-Dovekie mass step as a function of colour (orange points) and the ‘achromatic’/‘grey’ value . 7 8 9 10 11 12 Host Galaxy Mass M 2.25 2.50 2.75 3.00 3.25 R V k = -1.0 k = 1.0 [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: An example of varying 𝑘 within the logistic function. When 𝑘 = 1, the logistic function reduces to a sigmoid. 5.1 No Mass Step The originalBrout & Scolnic (2021) and Popovic et al.(2023) models ascribed the entirety of the observed mass step to Δ𝑅𝑉 across the host galaxy mass, and therefore did not allow for any achromatic difference in host galaxy magnitude. We show in [PITH_FULL_IMAGE:figures/full_fig_p… view at source ↗
Figure 3
Figure 3. Figure 3: Frequentist calibration done for Stjörnumál over 200 simulations, showing how well the input parameters are inferred. Black points show indi￾vidual realisations, and orange points are the mean of the 200 simulations. Orange rectangle shows the 1 − 𝜎 model uncertainty. A black line is plotted to show 𝑁 𝜎 = 0. explicit likelihood function (see Equations 20,21 in Popovic et al. (2023)). Rather, our neural net… view at source ↗
Figure 5
Figure 5. Figure 5: Rank histograms (orange) for the Stjörnumál fiducial model. The coverage expected under uniformity is shown in grey fill; a flat histogram within the bounds of this grey fill indicates well-calibrated posteriors. Bell or U-shaped distributions indicate a posterior that is too wide or narrow, respectively. Our posteriors are generally well-calibrated. 0.00 0.25 0.50 0.75 1.00 Posterior Rank Empirical CDF ci… view at source ↗
Figure 6
Figure 6. Figure 6: Stjörnumál frequentist calibration for each model parameter, re￾peated 200 times with 4,000 draws from the posterior. The grey shaded area is the 95% confidence band under perfect calibration. We see excellent pos￾terior calibration. case of posteriors that are skewed to one side of the distribution, the posterior for that given parameter is likely outside the prior range. To perform SBC, we make use of th… view at source ↗
Figure 7
Figure 7. Figure 7: shows the results; we find Wasserstein distances consis￾tently under 0.1, and see good agreement between the data and our posterior predictions. 7.2 Model Recovery on SNANA Simulations Of course, our previous tests are only as valid as the Stjörnumál simulation framework, which is not used in cosmology analyses. We calculate the recovery of our input parameters by sampling from our priors 200 times, simula… view at source ↗
Figure 8
Figure 8. Figure 8: Frequentist calibration done for Stjörnumál over 200 SNANA sim￾ulations, showing how well the input parameters are inferred. Black points show individual realisations, and orange points are the mean of the 200 sim￾ulations. Orange rectangle shows the 1 − 𝜎 model uncertainty. A black line is plotted to show 𝑁 𝜎 = 0. 0.0 0.2 0.4 0.6 0.8 1.0 Credibility Level 0.0 0.2 0.4 0.6 0.8 1.0 Expected Coverage Probabil… view at source ↗
Figure 9
Figure 9. Figure 9 [PITH_FULL_IMAGE:figures/full_fig_p009_9.png] view at source ↗
Figure 11
Figure 11. Figure 11: Rank histograms as a function of assumed cosmology for Stjörnumál. We colour-code the different assumed cosmologies, and see good model agreement for most of our parameters. 8 RESULTS [PITH_FULL_IMAGE:figures/full_fig_p010_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Posteriors from the nominal Stjörnumál model, on the DES-Dovekie data. Orange histograms show the 1D posteriors, and correlations are shown for each parameter. In general, the mean values are well-constrained, though some of the 𝜎 values do not converge. MNRAS 000, 1–18 (2026) [PITH_FULL_IMAGE:figures/full_fig_p011_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Inferred Δ E(𝑅𝑉 ) vs 𝛾. Points are taken from the posterior draws of [PITH_FULL_IMAGE:figures/full_fig_p013_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Δ𝜇 between the Nominal model employed in this paper and each other tested model. Orange points show the difference in median 𝜇 in each bin, black line shows Δ𝜇 = 0. We see 𝑑𝜇/𝑑𝑧 < 0.025 in all cases. 9 DISCUSSION & CONCLUSIONS In this work, we have presented a novel simulation based inference approach to inferring intrinsic SNe Ia and extrinsic dust populations, providing more flexibility and significant … view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

291 extracted references · 12 canonical work pages · 5 internal anchors

  1. [1]

    arXiv e-prints , keywords =

    Accounting for Selection Effects in Supernova Cosmology with Simulation-Based Inference and Hierarchical Bayesian Modelling. arXiv e-prints , keywords =. doi:10.48550/arXiv.2407.15923 , archivePrefix =. 2407.15923 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2407.15923

  2. [2]

    The Extinction Law in SN Ia Hosts from Background Galaxy Measurements: Toward a 1% Determination of $H_0$

    The Extinction Law in SN Ia Hosts from Background Galaxy Measurements: Toward a 1\. arXiv e-prints , keywords =. doi:10.48550/arXiv.2503.09702 , archivePrefix =. 2503.09702 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2503.09702

  3. [3]

    arXiv e-prints , keywords =

    Two-population model of type Ia supernovae and their associations with host galaxies in ZTF DR2. arXiv e-prints , keywords =

  4. [4]

    , keywords =

    Reduction of the type Ia supernova host galaxy step in the outer regions of galaxies. , keywords =. doi:10.1093/mnras/staf248 , archivePrefix =. 2408.03749 , primaryClass =

    Pith/arXiv arXiv

  5. [5]

    , keywords =

    The Dark Energy Bedrock All-sky Supernova Program: Cross Calibration, Simulations, and Cosmology Forecasts. , keywords =. doi:10.3847/1538-4357/ae1e78 , archivePrefix =. 2508.10877 , primaryClass =

    Pith/arXiv arXiv

  6. [6]

    skysurvey: a pure python package to simulate the transient sky

    skysurvey: a pure python package to simulate the transient sky. arXiv e-prints , keywords =. doi:10.48550/arXiv.2605.25840 , archivePrefix =. 2605.25840 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2605.25840

  7. [7]

    arXiv e-prints , keywords =

    On Calibration of Modern Neural Networks. arXiv e-prints , keywords =. doi:10.48550/arXiv.1706.04599 , archivePrefix =. 1706.04599 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.1706.04599

  8. [8]

    arXiv e-prints , keywords =

    Deep Sets. arXiv e-prints , keywords =. doi:10.48550/arXiv.1703.06114 , archivePrefix =. 1703.06114 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.1703.06114

  9. [9]

    arXiv e-prints , keywords =

    Misspecification-robust amortised simulation-based inference using variational methods. arXiv e-prints , keywords =. doi:10.48550/arXiv.2509.05724 , archivePrefix =. 2509.05724 , primaryClass =

    work page doi:10.48550/arxiv.2509.05724

  10. [10]

    Monthly Notices of the Royal Astronomical Society , volume =

    O’Callaghan, Matthew and Mandel, Kaisey S and Gilmore, Gerry , title =. Monthly Notices of the Royal Astronomical Society , volume =. 2026 , month =. doi:10.1093/mnras/staf2147 , url =

    work page doi:10.1093/mnras/staf2147 2026

  11. [11]

    , keywords =

    FlowSN: Neural Simulation-Based Inference under Realistic Selection Effects applied to Supernova Cosmology. , keywords =. doi:10.1093/mnras/stag1046 , archivePrefix =. 2603.11165 , primaryClass =

    Pith/arXiv arXiv

  12. [12]

    Nature Astronomy , keywords =

    CIGaRS I: combined simulation-based inference from type Ia supernovae and host photometry. Nature Astronomy , keywords =. doi:10.1038/s41550-026-02842-5 , archivePrefix =. 2508.15899 , primaryClass =

    Pith/arXiv arXiv

  13. [13]

    , keywords =

    SIDE-real: Supernova Ia Dust Extinction with truncated marginal neural ratio estimation applied to real data. , keywords =. doi:10.1093/mnras/stae995 , archivePrefix =. 2403.07871 , primaryClass =

    Pith/arXiv arXiv

  14. [14]

    arXiv e-prints , keywords =

    Likelihood-free MCMC with Amortized Approximate Ratio Estimators. arXiv e-prints , keywords =. doi:10.48550/arXiv.1903.04057 , archivePrefix =. 1903.04057 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.1903.04057 1903

  15. [15]

    arXiv e-prints , keywords =

    Variational Inference with Normalizing Flows. arXiv e-prints , keywords =. doi:10.48550/arXiv.1505.05770 , archivePrefix =. 1505.05770 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.1505.05770

  16. [16]

    arXiv e-prints , keywords =

    On the origin of the environmental step: A BayeSN view of the ZTF SN Ia DR2. arXiv e-prints , keywords =

  17. [17]

    Proceedings of the National Academy of Science , keywords =

    The frontier of simulation-based inference. Proceedings of the National Academy of Science , keywords =. doi:10.1073/pnas.1912789117 , archivePrefix =. 1911.01429 , primaryClass =

    Pith/arXiv arXiv 1911

  18. [18]

    , keywords =

    The dependence of the Type Ia Supernova colour─luminosity relation on their host galaxy properties. , keywords =. doi:10.1093/mnras/staf1570 , archivePrefix =. 2509.12035 , primaryClass =

    arXiv

  19. [19]

    arXiv e-prints , keywords =

    Banana Split: Improved Cosmological Constraints with Two Light-Curve-Shape and Color Populations Using Union3.1+UNITY1.8. arXiv e-prints , keywords =. doi:10.48550/arXiv.2601.19854 , archivePrefix =. 2601.19854 , primaryClass =

    doi:10.48550/arxiv.2601.19854

  20. [20]

    , keywords =

    Measuring Cosmological Parameters with Type Ia Supernovae in redMaGiC Galaxies. , keywords =. doi:10.3847/1538-4357/ac8b82 , archivePrefix =. 2202.10480 , primaryClass =

    Pith/arXiv arXiv

  21. [21]

    , keywords =

    The Dark Energy Survey: Cosmology Results with 1500 New High-redshift Type Ia Supernovae Using the Full 5 yr Data Set. , keywords =. doi:10.3847/2041-8213/ad6f9f , archivePrefix =. 2401.02929 , primaryClass =

    Pith/arXiv arXiv 2041

  22. [22]

    arXiv e-prints , keywords =

    Validating Bayesian Inference Algorithms with Simulation-Based Calibration. arXiv e-prints , keywords =. doi:10.48550/arXiv.1804.06788 , archivePrefix =. 1804.06788 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.1804.06788

  23. [23]

    40th International Conference on Machine Learning , keywords =

    Sampling-Based Accuracy Testing of Posterior Estimators for General Inference. 40th International Conference on Machine Learning , keywords =. doi:10.48550/arXiv.2302.03026 , archivePrefix =. 2302.03026 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2302.03026

  24. [24]

    Jan Boelts and Michael Deistler and Manuel Gloeckler and Álvaro Tejero-Cantero and Jan-Matthis Lueckmann and Guy Moss and Peter Steinbach and Thomas Moreau and Fabio Muratore and Julia Linhart and Conor Durkan and Julius Vetter and Benjamin Kurt Miller and Maternus Herold and Abolfazl Ziaeemehr and Matthijs Pals and Theo Gruner and Sebastian Bischoff and ...

    2025

  25. [25]

    arXiv e-prints , keywords =

    Neural Spline Flows. arXiv e-prints , keywords =. doi:10.48550/arXiv.1906.04032 , archivePrefix =. 1906.04032 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.1906.04032 1906

  26. [26]

    arXiv e-prints , keywords =

    Normalizing Flows for Probabilistic Modeling and Inference. arXiv e-prints , keywords =. doi:10.48550/arXiv.1912.02762 , archivePrefix =. 1912.02762 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.1912.02762 1912

  27. [27]

    Communications on Pure and Applied Mathematics , year=

    A Family of Nonparametric Density Estimation Algorithms , author=. Communications on Pure and Applied Mathematics , year=

  28. [28]

    Communications in Mathematical Sciences , year=

    DENSITY ESTIMATION BY DUAL ASCENT OF THE LOG-LIKELIHOOD ∗ , author=. Communications in Mathematical Sciences , year=

  29. [29]

    , keywords =

    Two-population Bayesian hierarchical model of Type Ia supernovae. , keywords =. doi:10.1093/mnras/stad2590 , archivePrefix =. 2302.01906 , primaryClass =

    Pith/arXiv arXiv

  30. [30]

    Peregrine: Sequential simulation-based inference for gravitational wave signals

    Bhardwaj, Uddipta and Alvey, James and Miller, Benjamin Kurt and Nissanke, Samaya and Weniger, Christoph. Peregrine: Sequential simulation-based inference for gravitational wave signals. 2023. arXiv:2304.02035

    Pith/arXiv arXiv 2023

  31. [31]

    arXiv e-prints , keywords =

    BAO miscalibration cannot rescue late-time solutions to the Hubble tension. arXiv e-prints , keywords =. doi:10.48550/arXiv.2510.01974 , archivePrefix =. 2510.01974 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2510.01974

  32. [32]

    arXiv e-prints , keywords =

    The Local Distance Network: a community consensus report on the measurement of the Hubble constant at 1\. arXiv e-prints , keywords =. doi:10.48550/arXiv.2510.23823 , archivePrefix =. 2510.23823 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2510.23823

  33. [33]

    arXiv e-prints , keywords =

    The tension in the absolute magnitude of Type Ia supernovae. arXiv e-prints , keywords =. doi:10.48550/arXiv.2307.02434 , archivePrefix =. 2307.02434 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2307.02434

  34. [34]

    , keywords =

    Calibrating the cosmic distance scale ladder: the role of the sound-horizon scale and the local expansion rate as distance anchors. , keywords =. doi:10.1093/mnras/stv261 , archivePrefix =. 1411.1094 , primaryClass =

    Pith/arXiv arXiv

  35. [35]

    , keywords =

    A Tale of Many H _ 0. , keywords =. doi:10.1146/annurev-astro-052622-033813 , archivePrefix =. 2311.13305 , primaryClass =

    Pith/arXiv arXiv

  36. [36]

    arXiv e-prints , keywords =

    Implications for dark energy of cosmic transparency in light of DESI data. arXiv e-prints , keywords =. doi:10.48550/arXiv.2506.22599 , archivePrefix =. 2506.22599 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2506.22599

  37. [37]

    , keywords =

    Detecting strongly lensed type Ia supernovae with LSST. , keywords =. doi:10.1093/mnras/stae1356 , archivePrefix =. 2312.04621 , primaryClass =

    Pith/arXiv arXiv

  38. [38]

    , keywords =

    GW170817: Observation of Gravitational Waves from a Binary Neutron Star Inspiral. , keywords =. doi:10.1103/PhysRevLett.119.161101 , archivePrefix =. 1710.05832 , primaryClass =

    Pith/arXiv arXiv

  39. [39]

    arXiv e-prints , keywords =

    Cosmology with supernova Encore in the strong lensing cluster MACS J0138-2155: Time delays & Hubble constant measurement. arXiv e-prints , keywords =. doi:10.48550/arXiv.2509.12301 , archivePrefix =. 2509.12301 , primaryClass =

    doi:10.48550/arxiv.2509.12301

  40. [40]

    Science , keywords =

    Constraints on the Hubble constant from supernova Refsdal's reappearance. Science , keywords =. doi:10.1126/science.abh1322 , archivePrefix =. 2305.06367 , primaryClass =

    Pith/arXiv arXiv

  41. [41]

    Alignment with DESI BAO and signs of a non-accelerating universe

    Strong progenitor age bias in supernova cosmology ─ II. Alignment with DESI BAO and signs of a non-accelerating universe. , keywords =. doi:10.1093/mnras/staf1685 , archivePrefix =. 2510.13121 , primaryClass =

    arXiv

  42. [42]

    arXiv e-prints , keywords =

    Still Accelerating: Type Ia supernova cosmology is robust to host galaxy age evolution. arXiv e-prints , keywords =. doi:10.48550/arXiv.2601.13785 , archivePrefix =. 2601.13785 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2601.13785

  43. [43]

    Contemporary Physics , keywords =

    Bayes in the sky: Bayesian inference and model selection in cosmology. Contemporary Physics , keywords =. doi:10.1080/00107510802066753 , archivePrefix =. 0803.4089 , primaryClass =

    Pith/arXiv arXiv

  44. [44]

    , keywords =

    Cosmological Model Insensitivity of Local H _ 0 from the Cepheid Distance Ladder. , keywords =. doi:10.3847/1538-4357/ab7fb0 , archivePrefix =. 2001.09260 , primaryClass =

    Pith/arXiv arXiv 2001

  45. [45]

    software , year = 2022, month = dec, eid =

    joshspeagle/dynesty: v2.0.3. software , year = 2022, month = dec, eid =. doi:10.5281/zenodo.7388523 , version =

    doi:10.5281/zenodo.7388523 2022

  46. [46]

    , keywords =

    DYNESTY: a dynamic nested sampling package for estimating Bayesian posteriors and evidences. , keywords =. doi:10.1093/mnras/staa278 , archivePrefix =. 1904.02180 , primaryClass =

    Pith/arXiv arXiv 1904

  47. [47]

    Statistics and Computing , keywords =

    Dynamic nested sampling: an improved algorithm for parameter estimation and evidence calculation. Statistics and Computing , keywords =. doi:10.1007/s11222-018-9844-0 , archivePrefix =. 1704.03459 , primaryClass =

    Pith/arXiv arXiv

  48. [48]

    Nested Sampling. Bayesian Inference and Maximum Entropy Methods in Science and Engineering: 24th International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering , year = 2004, editor =. doi:10.1063/1.1835238 , adsurl =

    doi:10.1063/1.1835238 2004

  49. [49]

    arXiv e-prints , keywords =

    The Dark Energy Survey Supernova Program: A Reanalysis Of Cosmology Results And Evidence For Evolving Dark Energy With An Updated Type Ia Supernova Calibration. arXiv e-prints , keywords =. doi:10.48550/arXiv.2511.07517 , archivePrefix =. 2511.07517 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2511.07517

  50. [50]

    , keywords =

    CMB constraints on the early Universe independent of late-time cosmology. , keywords =. doi:10.1103/PhysRevD.107.103505 , archivePrefix =. 2302.12911 , primaryClass =

    Pith/arXiv arXiv

  51. [51]

    , keywords =

    Dark energy and curvature from a future baryonic acoustic oscillation survey using the Lyman- forest. , keywords =. doi:10.1103/PhysRevD.76.063009 , archivePrefix =. astro-ph/0607122 , primaryClass =

    Pith/arXiv arXiv

  52. [52]

    , keywords =

    Baryon oscillations as a cosmological probe. , keywords =. doi:10.1103/PhysRevD.68.083504 , archivePrefix =. astro-ph/0304001 , primaryClass =

    Pith/arXiv arXiv

  53. [53]

    , keywords =

    Probing Dark Energy with Baryonic Acoustic Oscillations from Future Large Galaxy Redshift Surveys. , keywords =. doi:10.1086/379122 , archivePrefix =. astro-ph/0307460 , primaryClass =

    Pith/arXiv arXiv

  54. [54]

    , keywords =

    Probing Dark Energy Using Baryonic Oscillations in the Galaxy Power Spectrum as a Cosmological Ruler. , keywords =. doi:10.1086/376983 , archivePrefix =. astro-ph/0301632 , primaryClass =

    Pith/arXiv arXiv

  55. [55]

    Classical and Quantum Gravity , keywords =

    Jerk, snap and the cosmological equation of state. Classical and Quantum Gravity , keywords =. doi:10.1088/0264-9381/21/11/006 , archivePrefix =. gr-qc/0309109 , primaryClass =

    Pith/arXiv arXiv

  56. [56]

    , keywords =

    The Dark Energy Survey Supernova Program: an updated measurement of the Hubble constant using the inverse distance ladder. , keywords =. doi:10.1093/mnras/staf122 , archivePrefix =. 2406.05049 , primaryClass =

    Pith/arXiv arXiv

  57. [57]

    , keywords =

    First cosmological results using Type Ia supernovae from the Dark Energy Survey: measurement of the Hubble constant. , keywords =. doi:10.1093/mnras/stz978 , archivePrefix =. 1811.02376 , primaryClass =

    Pith/arXiv arXiv

  58. [58]

    , keywords =

    The clustering of galaxies in the SDSS-III Baryon Oscillation Spectroscopic Survey: baryon acoustic oscillations in the Data Releases 10 and 11 Galaxy samples. , keywords =. doi:10.1093/mnras/stu523 , archivePrefix =. 1312.4877 , primaryClass =

    Pith/arXiv arXiv

  59. [59]

    , keywords =

    A buyer's guide to the Hubble constant. , keywords =. doi:10.1007/s00159-021-00137-4 , archivePrefix =. 2109.01161 , primaryClass =

    Pith/arXiv arXiv

  60. [60]

    The Tip of the Red Giant Branch Distance Ladder and the Hubble Constant

    The Tip of the Red Giant Branch Distance Ladder and the Hubble Constant. arXiv e-prints , keywords =. doi:10.48550/arXiv.2403.17048 , archivePrefix =. 2403.17048 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2403.17048

  61. [61]

    , keywords =

    Connecting Infrared Surface Brightness Fluctuation Distances to Type Ia Supernova Hosts: Testing the Top Rung of the Distance Ladder. , keywords =. doi:10.3847/1538-4357/ace04b , archivePrefix =. 2204.12060 , primaryClass =

    Pith/arXiv arXiv

  62. [62]

    , keywords =

    A new measurement of the Hubble constant using Type Ia supernovae calibrated with surface brightness fluctuations. , keywords =. doi:10.1051/0004-6361/202039196 , archivePrefix =. 2008.07754 , primaryClass =

    Pith/arXiv arXiv 2008

  63. [63]

    Optical SBFs of shell galaxies

    Optical SBFs of shell galaxies. , keywords =. doi:10.48550/arXiv.0812.2346 , archivePrefix =. 0812.2346 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.0812.2346

  64. [64]

    , keywords =

    A First Comparison of the Surface Brightness Fluctuation Survey Distances with the Galaxy Density Field: Implications for H _ 0 and. , keywords =. doi:10.1086/312404 , archivePrefix =. astro-ph/9910340 , primaryClass =

    Pith/arXiv arXiv

  65. [65]

    , keywords =

    A New Technique for Measuring Extragalactic Distances. , keywords =. doi:10.1086/114847 , adsurl =

    doi:10.1086/114847

  66. [66]

    , keywords =

    CATS: The Hubble Constant from Standardized TRGB and Type Ia Supernova Measurements. , keywords =. doi:10.3847/2041-8213/ace978 , archivePrefix =. 2304.06693 , primaryClass =

    Pith/arXiv arXiv 2041

  67. [67]

    , keywords =

    Comparing Tip of the Red Giant Branch Distance Scales: An Independent Reduction of the Carnegie-Chicago Hubble Program and the Value of the Hubble Constant. , keywords =. doi:10.3847/1538-4357/ac68df , archivePrefix =. 2108.00007 , primaryClass =

    Pith/arXiv arXiv

  68. [68]

    , keywords =

    Measurements of the Hubble Constant: Tensions in Perspective. , keywords =. doi:10.3847/1538-4357/ac0e95 , archivePrefix =. 2106.15656 , primaryClass =

    Pith/arXiv arXiv

  69. [69]

    , keywords =

    A Comprehensive Measurement of the Local Value of the Hubble Constant with 1 km s ^ -1 Mpc ^ -1 Uncertainty from the Hubble Space Telescope and the SH0ES Team. , keywords =. doi:10.3847/2041-8213/ac5c5b , archivePrefix =. 2112.04510 , primaryClass =

    Pith/arXiv arXiv 2041

  70. [70]

    , keywords =

    The Pantheon+ Analysis: Cosmological Constraints. , keywords =. doi:10.3847/1538-4357/ac8e04 , archivePrefix =. 2202.04077 , primaryClass =

    Pith/arXiv arXiv

  71. [71]

    , keywords =

    DAmodel: hierarchical Bayesian modelling of DA white dwarfs for spectrophotometric calibration. , keywords =. doi:10.1093/mnras/staf629 , archivePrefix =. 2412.08809 , primaryClass =

    Pith/arXiv arXiv

  72. [72]

    arXiv e-prints , keywords =

    A Reassessment of the Pantheon+ and DES 5YR Calibration Uncertainties: Dovekie. arXiv e-prints , keywords =. doi:10.48550/arXiv.2506.05471 , archivePrefix =. 2506.05471 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2506.05471

  73. [73]

    arXiv e-prints , keywords =

    ZTF SN Ia DR2: The secondary maximum in Type Ia supernovae. arXiv e-prints , keywords =. doi:10.48550/arXiv.2406.19460 , archivePrefix =. 2406.19460 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2406.19460

  74. [74]

    , keywords =

    Using rest-frame optical and NIR data from the RAISIN survey to explore the redshift evolution of dust laws in SN Ia host galaxies. , keywords =. doi:10.1093/mnras/stae1111 , archivePrefix =. 2402.18624 , primaryClass =

    Pith/arXiv arXiv

  75. [75]

    , keywords =

    Secondary Maximum in the Near-Infrared Light Curves of Type Ia Supernovae. , keywords =. doi:10.1086/506588 , archivePrefix =. astro-ph/0606449 , primaryClass =

    Pith/arXiv arXiv

  76. [77]

    arXiv e-prints , keywords =

    The axis of systematic bias in SN -0.5ex cosmology and implications for DESI 2024 results. arXiv e-prints , keywords =. doi:10.48550/arXiv.2409.18668 , archivePrefix =. 2409.18668 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2409.18668 2024

  77. [78]

    , keywords =

    The Zwicky Transient Facility: Surveys and Scheduler. , keywords =. doi:10.1088/1538-3873/ab0c2a , archivePrefix =. 1905.02209 , primaryClass =

    Pith/arXiv arXiv 1905

  78. [81]

    , keywords =

    On the Dependence of Type Ia SNe Luminosities on the Metallicity of Their Host Galaxies. , keywords =. doi:10.3847/2041-8205/818/1/L19 , archivePrefix =. 1511.05348 , primaryClass =

    Pith/arXiv arXiv 2041

  79. [82]

    , keywords =

    Using the local gas-phase oxygen abundances to explore a metallicity dependence in SNe Ia luminosities. , keywords =. doi:10.1093/mnras/stw1706 , archivePrefix =. 1607.05526 , primaryClass =

    Pith/arXiv arXiv

  80. [83]

    , keywords =

    Strong dependence of Type Ia supernova standardization on the local specific star formation rate. , keywords =. doi:10.1051/0004-6361/201730404 , archivePrefix =. 1806.03849 , primaryClass =

    Pith/arXiv arXiv

Showing first 80 references.