Pith. sign in

REVIEW 3 major objections 6 minor 84 references

Efficient sampling for sparse Bayesian learning using hierarchical prior normalization

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

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

Pith's one-line read A coordinate change speeds up sparse Bayesian sampling

desk verdict The analytic KR prior-normalization map for SBL is a clean, correct, and useful contribution, but the abstract overclaims efficiency gains that the numerical evidence only partially supports. read the letter →

arxiv 2505.23753 v2 pith:YIOH4XHQ submitted 2025-05-29 math.NA cs.NA

Jan Glaubitz , Youssef Marzouk This is my paper
classification math.NAcs.NA MSC 62F1565C0565C4068U10
keywords sparseBayesianlearninghierarchicalpriornormalizationtransportmapsKnothe–RosenblattrearrangementMCMCsamplinginverseproblemsgeneralizedgammahyper-priorellipticalslice
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

Sparse Bayesian learning (SBL) priors produce high-dimensional posteriors that are hard for MCMC: they are non-log-concave, can be multimodal, and have strong correlations between the unknown $x$ and the variance hyperparameters $\theta$. The paper shows that an analytically constructed change of coordinates, using the product-like form of the SBL prior and the Knothe--Rosenblatt rearrangement, turns the SBL prior into a standard normal prior. Sampling the resulting prior-normalized posterior with ordinary MCMC samplers and pulling samples back through the inverse map is markedly more efficient than sampling the original posterior, especially for heavy-tailed hyper-priors. The point is that the coordinate change itself, not a new sampling algorithm, is what makes sparse Bayesian sampling tractable.

What carries the argument

The central object is the Knothe--Rosenblatt (KR) rearrangement: a triangular, monotone transport map that sends one density to another by composing univariate cumulative distribution functions. Applied to each factor $\pi_0^i(x_i,\theta_i)$, it gives the 2D map $s(x,\theta) = (s_\theta(\theta),\, x/\sqrt{\theta})$, where $s_\theta = (\Phi_0)^{-1}\circ P_\theta$ sends the generalized gamma hyper-prior to a standard normal and $x/\sqrt{\theta}$ sends the conditional Gaussian to a standard normal. Its inverse, $t(u,\tau) = (P_\theta^{-1}\circ\Phi_0(\tau),\, \sqrt{t_\tau(\tau)}\, u)$, evaluated stably through the inverse incomplete gamma function, is what maps posterior samples back to the original coordinates.

What would settle it

Apply the same block-diagonal map to an SBL model with a single shared hyperparameter $\theta$ controlling all coordinates, then check whether the pushforward of the joint prior is standard normal; it will not be, and the MCMC efficiency gains demonstrated here should disappear.

Watch

Extended reading notes

Core claim

The paper establishes that the SBL posterior $\pi_y(x,\theta) \propto f(x;y)\,\pi_0(x|\theta)\pi_0(\theta)$, with conditionally Gaussian $x|\theta$ and a generalized gamma hyper-prior on $\theta$, can be transformed exactly into a posterior with a standard normal prior by a block-diagonal Knothe--Rosenblatt map. Because the joint prior factors as $\prod_i \pi_0^i(x_i,\theta_i)$, the map decouples into $n$ independent 2D maps with the closed form $s(x,\theta) = ((\Phi_0)^{-1}\circ P_\theta(\theta),\, x/\sqrt{\theta})$. Sampling the pushforward posterior $\phi_y(u,\tau) = g(u,\tau;y)\,\phi_0(u,\tau)$ with standard MCMC and pulling back via the inverse map yields samples of the original posterior; numerically, this samples more efficiently than targeting the original posterior directly, with the largest gains for heavy-tailed hyper-priors ($r=-1/2$ and $r=-1$) where the original posterior is multimodal and anisotropic.

Load-bearing premise

The closed-form map exists only because the prior factors into independent $(x_i,\theta_i)$ pairs; if hyperparameters are shared across coordinates or the components are coupled, the block-diagonal Knothe--Rosenblatt map no longer normalizes the prior and the derived sampling gains do not follow.

Editorial extensions

If this is right

  • Adaptive Metropolis and MALA achieve better mixing and higher effective sample sizes on the normalized posterior than on the original one, with the gap widening as the hyper-prior becomes heavier tailed.
  • In the impulse-image experiment, elliptical slice sampling on the normalized posterior produced more accurate reconstructions in about 15% of the runtime of a Gibbs sampler applied to the original posterior.
  • MALA step sizes that were tuned for one value of the hyper-prior parameter $r$ continue to work across other values in the normalized coordinates, whereas the original posterior required step sizes an order of magnitude smaller as $r$ decreased.
  • The same construction applies to any scale-mixture-of-normals prior, including Laplace and horseshoe priors, since it relies only on the conditional Gaussian structure and the hyper-prior CDF.

Reading between the lines

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

  • If the improvement persists in higher dimensions, prior normalization could replace Gibbs sampling as the default workhorse for hierarchical Bayesian inverse problems with sparsity priors, since the normalized posterior is compatible with dimension-robust and likelihood-informed samplers.
  • A natural stress test is to break the product structure: with a shared hyperparameter across coordinates, the block-diagonal map no longer exactly normalizes the prior, so the analytic derivation would need a coupled transport map and the closed form would be lost.
  • The paper's diagnostics suggest that MPSRF and ESS can be blind to local trapping in multimodal SBL posteriors; reporting chain-level sample paths or individual posterior samples, as done here, may be the more informative check for such problems.
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. The paper proposes a change of variables for sparse Bayesian learning (SBL) posteriors: a Knothe–Rosenblatt (KR) transport map that normalizes the hierarchical SBL prior to a standard normal prior. Because the SBL prior is a product over components, the high-dimensional map decouples into identical two-dimensional maps; the paper derives these maps analytically and gives a stable implementation of the inverse map in terms of generalized gamma quantile functions. The authors then compare standard MCMC algorithms (adaptive Metropolis, MALA, Gibbs, elliptical slice) on the original and prior-normalized posteriors for a toy problem, a signal deconvolution problem, a Burgers inverse problem, and an impulse-image recovery problem, and claim significant efficiency gains from prior normalization.

Significance. If the efficiency claim is supported, the paper would provide a simple, exact, and essentially parameter-free reparameterization that could improve MCMC sampling for a widely used class of hierarchical sparsity-promoting priors. The analytic derivation in Section 3 is clean and correct, the inverse map has a practical stable implementation (Appendix A), and the code is publicly available. The method does not fit any quantity to data or rely on the paper's own numerical results, so circularity is not a concern. The main open question is whether the empirical evidence actually establishes the claimed gains, and that is where the manuscript currently falls short.

major comments (3)
  1. [Section 4.4, Figure 17] The impulse-image experiment simultaneously changes the sampler (Gibbs for the original posterior vs. elliptical slice for the prior-normalized posterior) and the chain length (1e5 vs. 1e4 samples per chain). The reported runtime advantage and the visibly better x-mean for the prior-normalized case therefore cannot be attributed to prior normalization; they could also be due to the different sampler. To make the central claim load-bearing, please also apply the same base sampler (e.g., adaptive Metropolis or MALA) to both parameterizations in this problem, or at least report an equivalent same-sampler comparison; otherwise the abstract's claim of 'significant performance improvements' is not supported by this experiment.
  2. [Section 4.2, Table 3] For r = -1, the reported ESS per second is lower for the prior-normalized posterior (7.4e-1) than for the original (1.2e+0), while the MPSRF is lower for the prior-normalized posterior. The text then argues that both metrics are unreliable in multimodal settings. This is an internal tension: the only quantitative efficiency metric that favors the prior-normalized posterior is declared unreliable, and the metric that is unfavorable is also declared unreliable. The qualitative trace plots and per-chain means are suggestive but do not provide a quantitative measure of between-mode mixing. Please add a quantitative multimodal diagnostic, such as the fraction of time each chain spends in each mode, the number of transitions between the two high-density regions, or a mode-aware effective sample size, and report it for both parameterizations. Without such a diagnostic, the claim that the prior-normalized posterior is sampled more efficiently is not quantitatively established.
  3. [Section 4.1, Table 2] For r = 1, where the original posterior is log-concave, the original posterior yields better MPSRF and ESS than the prior-normalized posterior for both AM and MALA (e.g., MALA MPSRF 1.8e-3 vs. 2.8e-3; AM ESS/s 1.3e+5 vs. 1.0e+5). The abstract and introduction state without qualification that standard MCMC techniques sample the prior-normalized posterior more efficiently. The claim should be qualified to the non-log-concave, multimodal cases (for instance r = -1) for which the paper is arguing, and the r = 1 case should be discussed as a limitation or as evidence that prior normalization is not universally beneficial.
minor comments (6)
  1. [Figure 17 caption] The caption says the ES sampler generated chains 'from the original posterior' with roughly 3 minutes runtime, but the text and context indicate it sampled the prior-normalized posterior; please correct this typo.
  2. [Table 2 caption] The column labeled 'MPSRF' is described in the text as 'MPSRF minus one multiplied by the time'; the caption should state this explicitly so the reader knows the reported quantity is not the raw MPSRF.
  3. [Section 3.2, Eq. (3.2)] The product-like form relies on the assumptions that the x_i are conditionally independent given their own θ_i and that the θ_i are independent. This is exact for the SBL priors considered, but the extension to general scale mixtures of normals in Section 5 would require stating how this product structure is preserved or relaxed; otherwise the decoupling into 2D maps does not immediately apply.
  4. [Section 4.2] The manuscript thins the MCMC chains by keeping every 1000th sample, but it does not state whether the MPSRF and ESS values in Table 3 are computed on the thinned chains; if so, this should be reported because thinning can materially reduce the estimated ESS.
  5. [Figure 3] The y-axis tick labels in Figure 3 appear to be missing or unclear; adding proper axis labels or a description in the caption would improve readability.
  6. [Section 4.4] The runtime comparison between Gibbs and ES would be more informative if per-sample computational costs and any burn-in or adaptation period were reported, since the two samplers have very different per-iteration costs.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation; the prior-normalizing map is an explicit change of variables, and the only self-citations are background references that are not load-bearing.

full rationale

The paper's central derivation is a change of variables, not a fit or a self-referential prediction. The prior-normalizing map S is defined by the exact Knothe--Rosenblatt construction: the first component is s_theta = (Phi_0)^{-1} o P_theta (Eq. 3.6), the second is s_{x|theta} = x/sqrt(theta) (Eq. 3.5), and the inverse is t_tau = (P_theta)^{-1} o Phi_0 (Eq. 3.9) with t_{u|tau} = sqrt(t_tau(tau)) u. These formulas are explicit and derive solely from the SBL product-form prior (3.2) and standard normal CDFs; no parameter is fitted to data and no prediction is obtained from the quantity it claims to predict. The posterior identity phi_y = S#pi_y in (1.3) is the exact pushforward under the same map, so the sampling target is a genuine transformed posterior rather than a disguised input. The efficiency comparisons in Section 4 are empirical and may be confounded (for example, the Gibbs-versus-elliptical-slice comparison in Section 4.4 changes both sampler and coordinates), but confounding is a correctness or experimental-design concern, not circularity. Citations to the authors' earlier transport-map work [50, 57, 84] are background references and are not load-bearing: the KR map used here is textbook (Rosenblatt--Knothe) and is derived in the paper's own equations, with no uniqueness theorem or ansatz imported from a self-citation. The product-form assumption in (3.2) is stated explicitly and is exact for the SBL priors considered, so it is not a hidden re-import of the conclusion. On this basis, no specific circular step can be exhibited, and the appropriate finding is a low score reflecting only a minor self-citation footprint.

Assumptions & free parameters 0 free parameters · 2 assumptions · 0 invented entities

The method introduces no new entities and no fitted parameters; it combines standard transport theory with the product structure of SBL priors.

assumptions (2)
  • standard math Knothe-Rosenblatt rearrangements exist and are unique for absolutely continuous probability measures.
    Used in Section 3.2 to justify existence of the triangular map S.
  • domain assumption The SBL prior is a product of independent bivariate factors (x_i, θ_i), with conditionally Gaussian x_i|θ_i and generalized gamma θ_i.
    Section 2 and Eq. (3.2): this product structure is what enables the block-diagonal TM and its analytic derivation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Efficient sampling for sparse Bayesian learning using hierarchical prior normalization." pith.science (2026). https://pith.science/paper/YIOH4XHQ

@misc{pith2026250523753,
  author       = {Pith},
  title        = {Pith review of: Efficient sampling for sparse Bayesian learning using hierarchical prior normalization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YIOH4XHQ}},
  note         = {Machine review of arXiv:2505.23753}
}
read the original abstract

We introduce an approach for efficient Markov chain Monte Carlo (MCMC) sampling for challenging high-dimensional distributions in sparse Bayesian learning (SBL). The core innovation involves using hierarchical prior-normalizing transport maps (TMs), which are deterministic couplings that transform the sparsity-promoting SBL prior into a standard normal one. We analytically derive these prior-normalizing TMs by leveraging the product-like form of SBL priors and Knothe--Rosenblatt (KR) rearrangements. These transform the complex target posterior into a simpler reference distribution equipped with a standard normal prior that can be sampled more efficiently. Specifically, one can leverage the standard normal prior by using more efficient, structure-exploiting samplers. Our numerical experiments on various inverse problems -- including signal deblurring, inverting the non-linear inviscid Burgers equation, and recovering an impulse image -- demonstrate significant performance improvements for standard MCMC techniques.

Figures

Figures reproduced from arXiv: 2505.23753 by the authors.

Figure 1
Figure 1. Normalized densities of the univariate generalized gamma hyper-prior [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Contour plots of the posterior density π y (x, θ) in Example 2.1 with y = 0.2 and σ 2 = 10−2.8 . We use the parameter combinations (r, β, ϑ) as detailed in [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. First components of the KR map s and its inverse t = s −1 for the parameters in [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (14 more)
Figure 4
Figure 4. Figure 4: Contour plots of the prior-normalized posterior for the univariate toy problem in [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: True signal and noisy, blurred observational data [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: Top: MAP estimates (x MAP, θ MAP) of the original posterior. Bottom: Pullback of the MAP estimates (u MAP, τ MAP) of the prior-normalized posterior. Figure 8b shows that the samples from the prior-normalized posterior differ visibly from one another, indicating that th…
Figure 7
Figure 7. Figure 7: Mean and 90% quantile ranges of the x- and θ-samples of the original and prior-normalized posterior for r = ±1. We used the AM algorithm initialized with the respective MAP estimate. estimates for the means and quantile ranges. Noticeable differences appear for r = −1,…
Figure 8
Figure 8. Figure 8: Individual x-samples of the first MCMC chain from the original and prior-normalized posterior for r = −1. We used the AM algorithm initialized with the respective MAP estimate. 0 2.50×10 6 5.00×10 6 7.50×10 6 1.00×10 7 Iterations 0.25 0.00 0.25 0.50 0.75 1.00 1.25 x[2 …
Figure 9
Figure 9. Figure 9: Traces for the x22-samples of the posteriors for r = −1. Note that t22 ≈ 0.172 is close to the first jump discontinuity and yields g(t22) = 1. We use the AM algorithm initialized with the MAP estimates [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]
Figure 10
Figure 10. Figure 10: Mean and 90% quantile ranges of the original and prior-normalized posterior for [PITH_FULL_IMAGE:figures/full_fig_p015_10.png]
Figure 11
Figure 11. Figure 11: Individual x-samples of the first MCMC chain from the original and prior-normalized posterior for r = −1. We used the AM algorithm initialized with random prior draws. 0 2.50×10 6 5.00×10 6 7.50×10 6 1.00×10 7 Iterations 0.25 0.00 0.25 0.50 0.75 1.00 1.25 x[2 2] (a) x…
Figure 12
Figure 12. Figure 12: Traces for the x22-samples of the original and prior-normalized posterior for r = −1. Note that t22 ≈ 0.172 is close to the first jump discontinuity and yields g(t22) = 1. We used the AM algorithm initialized with random draws from the respective prior. with periodic …
Figure 13
Figure 13. Figure 13: Mean and 90% quantile range of the x- and θ-samples for each of the first three MCMC chains of the original and prior-normalized posterior for r = −1. We used the AM algorithm initialized with random draws from the respective prior. is (4.7) b = F(u) + e, where u = [u…
Figure 14
Figure 14. Figure 14: Setup for the inverse Burgers problem, including the initial data (dotted blue), the reference (solid [PITH_FULL_IMAGE:figures/full_fig_p018_14.png]
Figure 15
Figure 15. Figure 15: Mean and 90% quantile ranges of the original and prior-normalized posterior with [PITH_FULL_IMAGE:figures/full_fig_p019_15.png]
Figure 16
Figure 16. Figure 16: Traces for the u26-samples of the original and prior-normalized posterior for r = −1. Note that x26 ≈ 0.252 is close to the first jump discontinuity and yields u0(x26) = 1. We used the AM algorithm initialized with the respective MAP estimate and random draws from the…
Figure 17
Figure 17. Figure 17: (Log-)Means of the x- and θ-samples from the original and prior-normalized posterior correspond￾ing to the impulse image. The Gibbs sampler generated four chains with 105 samples each from the original posterior (≈ 20 minutes runtime). The ES sampler generated four ch…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

84 extracted references · 76 canonical work pages

  1. [1]

    Agapiou, J

    S. Agapiou, J. M. Bardsley, O. Papaspiliopoulos, and A. M. Stuart , Analysis of the Gibbs sampler for hierarchical inverse problems, SIAM/ASA Journal on Uncertainty Quantification, 2 (2014), pp. 511–544

    2014

  2. [2]

    Ambrosio and N

    L. Ambrosio and N. Gigli , A user’s guide to optimal transport , Modelling and Optimisation of Flows on Networks, (2013), p. 1

    2013

  3. [3]

    D. F. Andrews and C. L. Mallows , Scale mixtures of normal distributions , Journal of the Royal Statistical Society: Series B (Methodological), 36 (1974), pp. 99–102

    1974

  4. [4]

    Andrieu and J

    C. Andrieu and J. Thoms , A tutorial on adaptive MCMC , Statistics and Computing, 18 (2008), pp. 343–373

    2008

  5. [5]

    Ascolani, G

    F. Ascolani, G. O. Roberts, and G. Zanella , Scalability of Metropolis-within-Gibbs schemes for high- dimensional Bayesian models , arXiv preprint arXiv:2403.09416, (2024)

    arXiv 2024

  6. [6]

    Atchad´e and G

    Y. Atchad´e and G. Fort , Limit theorems for some adaptive MCMC algorithms with subgeometric kernels , Bernoulli, 16 (2010)

    work page 2010

  7. [7]

    Y. F. Atchad ´e, An adaptive version for the Metropolis adjusted Langevin algorithm with a truncated drift , Methodology and Computing in Applied Probability, 8 (2006), pp. 235–254

    work page 2006

  8. [8]

    K. B. Athreya and S. N. Lahiri , Measure Theory and Probability Theory , vol. 19, Springer, 2006

    work page 2006

Show all 84 references
  1. [9]

    S. D. Babacan, R. Molina, and A. K. Katsaggelos , Bayesian compressive sensing using Laplace priors , IEEE Transactions on Image Processing, 19 (2009), pp. 53–63. EFFICIENT SAMPLING FOR SBL USING PRIOR NORMALIZATION 23

    2009

  2. [10]

    Baptista, Y

    R. Baptista, Y. Marzouk, and O. Zahm, On the representation and learning of monotone triangular transport maps, Foundations of Computational Mathematics, (2023), pp. 1–46

    2023

  3. [11]

    E. M. L. Beale and C. L. Mallows , Scale mixing of symmetric distributions with zero means , The Annals of Mathematical Statistics, (1959), pp. 1145–1151

    1959

  4. [12]

    Beskos, M

    A. Beskos, M. Girolami, S. Lan, P. E. F arrell, and A. M. Stuart , Geometric MCMC for infinite- dimensional inverse problems , Journal of Computational Physics, 335 (2017), pp. 327–351

    2017

  5. [13]

    Beskos, N

    A. Beskos, N. Pillai, G. Roberts, J.-M. Sanz-Serna, and A. Stuart, Optimal tuning of the hybrid Monte Carlo algorithm , Bernoulli, 19 (2013), pp. 1501–1534

    2013

  6. [14]

    Bezanson, A

    J. Bezanson, A. Edelman, S. Karpinski, and V. B. Shah , Julia: A fresh approach to numerical computing , SIAM Review, 59 (2017), pp. 65–98

    2017

  7. [15]

    V. I. Bogachev, A. V. Kolesnikov, and K. V. Medvedev, Triangular transformations of measures, Sbornik: Mathematics, 196 (2005), p. 309

    2005

  8. [16]

    Brooks, A

    S. Brooks, A. Gelman, G. Jones, and X.-L. Meng , Handbook of Markov Chain Monte Carlo , CRC press, 2011

    2011

  9. [17]

    S. P. Brooks and A. Gelman , General methods for monitoring convergence of iterative simulations , Journal of Computational and Graphical Statistics, 7 (1998), pp. 434–455

    1998

  10. [18]

    Cabezas and C

    A. Cabezas and C. Nemeth , Transport elliptical slice sampling , in International Conference on Artificial Intelligence and Statistics, PMLR, 2023, pp. 3664–3676

    2023

  11. [19]

    Calvetti, M

    D. Calvetti, M. Pragliola, and E. Somersalo, Sparsity promoting hybrid solvers for hierarchical Bayesian inverse problems, SIAM Journal on Scientific Computing, 42 (2020), pp. A3761–A3784

    2020

  12. [20]

    Calvetti, M

    D. Calvetti, M. Pragliola, E. Somersalo, and A. Strang , Sparse reconstructions from few noisy data: analysis of hierarchical Bayesian models with generalized gamma hyperpriors , Inverse Problems, 36 (2020), p. 025010

    2020

  13. [21]

    Calvetti and E

    D. Calvetti and E. Somersalo, A Gaussian hypermodel to recover blocky objects, Inverse Problems, 23 (2007), p. 733

    2007

  14. [22]

    Calvetti and E

    D. Calvetti and E. Somersalo , Bayesian Scientific Computing , vol. 215, Springer Nature, 2023

    2023

  15. [23]

    Calvetti and E

    D. Calvetti and E. Somersalo , Computationally efficient sampling methods for sparsity promoting hierar- chical Bayesian models , SIAM/ASA Journal on Uncertainty Quantification, 12 (2024), pp. 524–548

    2024

  16. [24]

    Calvetti, E

    D. Calvetti, E. Somersalo, and A. Strang , Hierachical Bayesian models and sparsity: ℓ2-magic, Inverse Problems, 35 (2019), p. 035003

    2019

  17. [25]

    C. M. Carvalho, N. G. Polson, and J. G. Scott, Handling sparsity via the horseshoe, in Artificial Intelligence and Statistics, PMLR, 2009, pp. 73–80

    2009

  18. [26]

    N. K. Chada, M. A. Iglesias, L. Roininen, and A. M. Stuart , Parameterizations for ensemble Kalman inversion, Inverse Problems, 34 (2018), p. 055009

    2018

  19. [27]

    V. Chen, M. M. Dunlop, O. Papaspiliopoulos, and A. M. Stuart , Robust MCMC sampling with non- Gaussian and hierarchical priors in high dimensions , arXiv preprint arXiv:1803.03344, 3 (2018)

    2018 arXiv

  20. [28]

    Churchill and A

    V. Churchill and A. Gelb , Sampling-based spotlight SAR image reconstruction from phase history data for speckle reduction and uncertainty quantification , SIAM/ASA Journal on Uncertainty Quantification, 10 (2022), pp. 1225–1249

    2022

  21. [29]

    Cotter, G

    S. Cotter, G. Roberts, A. Stuart, and D. White, MCMC methods for functions: Modifying old algorithms to make them faster , Statistical Science, 28 (2013), pp. 424–446

    2013

  22. [30]

    T. Cui, X. T. Tong, and O. Zahm , Prior normalization for certified likelihood-informed subspace detection of Bayesian inverse problems , Inverse Problems, 38 (2022), p. 124002

    2022

  23. [31]

    https://dlmf.nist.gov/, Release 1.2.0 of 2024-03-15

    NIST Digital Library of Mathematical Functions . https://dlmf.nist.gov/, Release 1.2.0 of 2024-03-15. F. W. J. Olver, A. B. Olde Daalhuis, D. W. Lozier, B. I. Schneider, R. F. Boisvert, C. W. Clark, B. R. Miller, B. V. Saunders, H. S. Cohl, and M. A. McClain, eds

    2024

  24. [32]

    Dong and M

    Y. Dong and M. Pragliola, Inducing sparsity via the horseshoe prior in imaging problems , Inverse Problems, 39 (2023), p. 074001

    2023

  25. [33]

    M. M. Dunlop, T. Helin, and A. M. Stuart , Hyperparameter estimation in Bayesian MAP estimation: parameterizations and consistency, The SMAI Journal of Computational Mathematics, 6 (2020), pp. 69–100

    2020

  26. [34]

    M. A. Figueiredo, J. M. Bioucas-Dias, and R. D. Nowak , Majorization–minimization algorithms for wavelet-based image restoration, IEEE Transactions on Image Processing, 16 (2007), pp. 2980–2991

    2007

  27. [35]

    Fleischer , Transformations for accelerating MCMC simulations with broken ergodicity , in 2007 Winter Simulation Conference, IEEE, 2007, pp

    M. Fleischer , Transformations for accelerating MCMC simulations with broken ergodicity , in 2007 Winter Simulation Conference, IEEE, 2007, pp. 658–666

    2007

  28. [36]

    Flock, Y

    R. Flock, Y. Dong, F. Uribe, and O. Zahm , Continuous Gaussian mixture solution for linear Bayesian 24 J. GLAUBITZ AND Y. MARZOUK inversion with application to Laplace priors , Inverse Problems, (2024)

    2024

  29. [37]

    Gelman, J

    A. Gelman, J. B. Carlin, H. S. Stern, and D. B. Rubin, Bayesian Data Analysis, Chapman and Hall/CRC, 2003

    2003

  30. [38]

    Gelman and D

    A. Gelman and D. B. Rubin , Inference from iterative simulation using multiple sequences , Statistical Science, 7 (1992), pp. 457–472

    1992

  31. [39]

    Glaubitz and A

    J. Glaubitz and A. Gelb , Leveraging joint sparsity in hierarchical Bayesian learning , SIAM/ASA Journal on Uncertainty Quantification, 12 (2024), pp. 442–472

    2024

  32. [40]

    Glaubitz, A

    J. Glaubitz, A. Gelb, and G. Song , Generalized sparse Bayesian learning and application to image recon- struction, SIAM/ASA Journal on Uncertainty Quantification, 11 (2023), pp. 262–284

    2023

  33. [41]

    Gom `es, C

    O. Gom `es, C. Combes, and A. Dussauchoy , Parameter estimation of the generalized gamma distribution , Mathematics and Computers in Simulation, 79 (2008), pp. 955–963

    2008

  34. [42]

    Haario, E

    H. Haario, E. Saksman, and J. Tamminen , An adaptive Metropolis algorithm , Bernoulli, 7 (2001), pp. 223– 242

    2001

  35. [43]

    Kaipio and E

    J. Kaipio and E. Somersalo , Statistical inverse problems: discretization, model reduction and inverse crimes , Journal of Computational and Applied Mathematics, 198 (2007), pp. 493–504

    2007

  36. [44]

    Knothe, Contributions to the theory of convex bodies

    H. Knothe, Contributions to the theory of convex bodies. , Michigan Mathematical Journal, 4 (1957), pp. 39–52

    1957

  37. [45]

    Lindbloom, J

    J. Lindbloom, J. Glaubitz, and A. Gelb , Efficient sparsity-promoting MAP estimation for Bayesian linear inverse problems, Inverse Problems, 41 (2025), p. 025001

    2025

  38. [46]

    Lindbloom, M

    J. Lindbloom, M. Pasha, J. Glaubitz, and Y. Marzouk , Priorconditioned sparsity-promoting projection methods for deterministic and Bayesian linear inverse problems , arXiv preprint arXiv:2505.01827, (2025)

    2025 arXiv

  39. [47]

    J. S. Liu , Monte Carlo Strategies in Scientific Computing , Springer Science & Business Media, 2013

    2013

  40. [48]

    Markkanen, L

    M. Markkanen, L. Roininen, J. M. Huttunen, and S. Lasanen, Cauchy difference priors for edge-preserving Bayesian inversion , Journal of Inverse and Ill-Posed Problems, 27 (2019), pp. 225–240

    2019

  41. [49]

    Marshall and G

    T. Marshall and G. Roberts , An adaptive approach to Langevin MCMC , Statistics and Computing, 22 (2012), pp. 1041–1057

    2012

  42. [50]

    Marzouk, T

    Y. Marzouk, T. Moselhy, M. Parno, and A. Spantini , Sampling via measure transport: An introduction , Handbook of Uncertainty Quantification, 1 (2016), p. 2

    2016

  43. [51]

    Murray, R

    I. Murray, R. Adams, and D. MacKay , Elliptical slice sampling , in Proceedings of the 13th International Conference on Artificial Intelligence and Statistics, 2010, pp. 541–548

    2010

  44. [52]

    Nishihara, I

    R. Nishihara, I. Murray, and R. P. Adams , Parallel MCMC with generalized elliptical slice sampling , The Journal of Machine Learning Research, 15 (2014), pp. 2087–2112

    2014

  45. [53]

    Nocedal and S

    J. Nocedal and S. J. Wright , Numerical Optimization, Springer, 2006

    2006

  46. [54]

    Papaspiliopoulos, G

    O. Papaspiliopoulos, G. O. Roberts, and M. Sk¨old, Non-centered parameterisations for hierarchical models and data augmentation , Bayesian Statistics, 7 (2003), pp. 307–326

    2003

  47. [55]

    Papaspiliopoulos, G

    O. Papaspiliopoulos, G. O. Roberts, and M. Sk ¨old, A general framework for the parametrization of hier- archical models, Statistical Science, (2007), pp. 59–73

    2007

  48. [56]

    Park and G

    T. Park and G. Casella , The Bayesian Lasso , Journal of the American Statistical Association, 103 (2008), pp. 681–686

    2008

  49. [57]

    M. D. Parno and Y. M. Marzouk, Transport map accelerated Markov chain Monte Carlo, SIAM/ASA Journal on Uncertainty Quantification, 6 (2018), pp. 645–682

    2018

  50. [58]

    F. J. Pinski, G. Simpson, A. M. Stuart, and H. Weber , Algorithms for Kullback–Leibler approximation of probability measures in infinite dimensions , SIAM Journal on Scientific Computing, 37 (2015), pp. A2733– A2757

    2015

  51. [59]

    Rackauckas and Q

    C. Rackauckas and Q. Nie , DifferentialEquations.jl – A performant and feature-rich ecosystem for solving differential equations in Julia , Journal of Open Research Software, 5 (2017), p. 15

    2017

  52. [60]

    Ranocha, M

    H. Ranocha, M. Schlottke-Lakemper, A. R. Winters, E. F aulhaber, J. Chan, and G. J. Gassner , Adaptive numerical simulations with Trixi.jl: A case study of Julia for scientific computing , Proceedings of the JuliaCon Conferences, 1 (2022), p. 77

    2022

  53. [61]

    Ranocha, A

    H. Ranocha, A. R. Winters, M. Schlottke-Lakemper, P. ¨Offner, J. Glaubitz, and G. J. Gassner, On the robustness of high-order upwind summation-by-parts methods for nonlinear conservation laws , Journal of Computational Physics, 520 (2025), p. 113471

    2025

  54. [62]

    Rezende and S

    D. Rezende and S. Mohamed , Variational inference with normalizing flows , in International Conference on Machine Learning, PMLR, 2015, pp. 1530–1538

    2015

  55. [63]

    Robert and G

    C. Robert and G. Casella , Monte Carlo Statistical Methods , Springer Science & Business Media, 2013

    2013

  56. [64]

    G. O. Roberts and J. S. Rosenthal , Optimal scaling of discrete approximations to Langevin diffusions , EFFICIENT SAMPLING FOR SBL USING PRIOR NORMALIZATION 25 Journal of the Royal Statistical Society: Series B (Statistical Methodology), 60 (1998), pp. 255–268

    1998

  57. [65]

    Rosenblatt, Remarks on a multivariate transformation , The Annals of Mathematical Statistics, 23 (1952), pp

    M. Rosenblatt, Remarks on a multivariate transformation , The Annals of Mathematical Statistics, 23 (1952), pp. 470–472

    1952

  58. [66]

    Santambrogio, Optimal Transport for Applied Mathematicians , Springer, 2015

    F. Santambrogio, Optimal Transport for Applied Mathematicians , Springer, 2015

    2015

  59. [67]

    Sanz-Alonso and O

    D. Sanz-Alonso and O. Al-Ghattas, A first course in Monte Carlo methods, arXiv preprint arXiv:2405.16359, (2024)

    2024 arXiv

  60. [68]

    Sanz-Alonso and N

    D. Sanz-Alonso and N. W aniorek , Hierarchical Bayesian inverse problems: A high-dimensional statistics viewpoint, arXiv preprint arXiv:2401.03074, (2024)

    2024 arXiv

  61. [69]

    Z. Si, Y. Liu, and A. Strang , Path-following methods for maximum a posteriori estimators in Bayesian hierarchical models: How estimates depend on hyperparameters , SIAM Journal on Optimization, 34 (2024), pp. 2201–2230

    2024

  62. [70]

    A. M. Stuart, Inverse problems: a Bayesian perspective , Acta Numerica, 19 (2010), pp. 451–559

    2010

  63. [71]

    Suuronen, N

    J. Suuronen, N. K. Chada, and L. Roininen , Cauchy Markov random field priors for Bayesian inversion , Statistics and Computing, 32 (2022), p. 33

    2022

  64. [72]

    X. Tan, J. Li, and P. Stoica, Efficient sparse bayesian learning via Gibbs sampling, in 2010 IEEE International Conference on Acoustics, Speech and Signal Processing, IEEE, 2010, pp. 3634–3637

    2010

  65. [73]

    M. E. Tipping , Sparse Bayesian learning and the relevance vector machine , Journal of Machine Learning Re- search, 1 (2001), pp. 211–244

    2001

  66. [74]

    Uribe, J

    F. Uribe, J. M. Bardsley, Y. Dong, P. C. Hansen, and N. A. Riis , A hybrid Gibbs sampler for edge- preserving tomographic reconstruction with uncertain view angles, SIAM/ASA Journal on Uncertainty Quan- tification, 10 (2022), pp. 1293–1320

    2022

  67. [75]

    Uribe, Y

    F. Uribe, Y. Dong, and P. C. Hansen, Horseshoe priors for edge-preserving linear Bayesian inversion , SIAM Journal on Scientific Computing, 45 (2023), pp. B337–B365

    2023

  68. [76]

    Vehtari, A

    A. Vehtari, A. Gelman, D. Simpson, B. Carpenter, and P.-C. B ¨urkner, Rank-normalization, folding, and localization: An improved ˆR for assessing convergence of MCMC (with discussion) , Bayesian Analysis, 16 (2021), pp. 667–718

    2021

  69. [77]

    Villani, Optimal Transport: Old and New , vol

    C. Villani, Optimal Transport: Old and New , vol. 338, Springer, 2009

    2009

  70. [78]

    W ang, J

    Z. W ang, J. M. Bardsley, A. Solonen, T. Cui, and Y. M. Marzouk , Bayesian inverse problems with ℓ1 priors: a randomize-then-optimize approach , SIAM Journal on Scientific Computing, 39 (2017), pp. S140– S166

    2017

  71. [79]

    West, On scale mixtures of normal distributions , Biometrika, 74 (1987), pp

    M. West, On scale mixtures of normal distributions , Biometrika, 74 (1987), pp. 646–648

    1987

  72. [80]

    D. P. Wipf and B. D. Rao , Sparse Bayesian learning for basis selection , IEEE Transactions on Signal Pro- cessing, 52 (2004), pp. 2153–2164

    2004

  73. [81]

    Wolff and A

    U. Wolff and A. Collaboration , Monte Carlo errors with less errors , Computer Physics Communications, 156 (2004), pp. 143–153

    2004

  74. [82]

    Xiao and J

    Y. Xiao and J. Glaubitz , Sequential image recovery using joint hierarchical Bayesian learning , Journal of Scientific Computing, 96 (2023), p. 4

    2023

  75. [83]

    J. Yang, G. O. Roberts, and J. S. Rosenthal , Optimal scaling of random-walk Metropolis algorithms on general target distributions , Stochastic Processes and their Applications, 130 (2020), pp. 6094–6132

    2020

  76. [84]

    O. Zahm, T. Cui, K. Law, A. Spantini, and Y. Marzouk , Certified dimension reduction in nonlinear Bayesian inverse problems , Mathematics of Computation, 91 (2022), pp. 1789–1835

    2022

Pith tools

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