Pith. sign in

REVIEW 4 major objections 6 minor 50 references

Functional normalizing flow for statistical inverse problems of partial differential equations

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read This paper establishes conditions under which normalizing flows can be defined in infinite-dimensional function spaces for Bayesian inverse problems of PDEs, yielding a mesh-independent variational inference algorithm with an explicit…

desk verdict A genuinely useful infinite-dimensional normalizing flow framework for Bayesian inverse problems, but the printed Radon–Nikodym formula in the central theorem has a sign error that contradicts its own appendix and the Cameron–Martin theorem. read the letter →

arxiv 2411.13277 v3 pith:XZN3YY57 submitted 2024-11-20 math.NA cs.NA

classification math.NAcs.NA MSC 62F1565N21
keywords inverseproblemsinfinite-dimensionalvariationalinferencefunctionalnormalizingflowBayesiananalysisforfunctionspartialdifferentialequationsdiscretizationinvarianceconditional
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

This paper proposes a way to run variational inference for Bayesian inverse problems of PDEs directly in infinite-dimensional function space, so the algorithm behaves consistently as the numerical mesh is refined. The authors construct normalizing flows, compositions of simple transformations, that map a Gaussian prior measure into a rich family of approximate posterior measures while remaining equivalent to the prior, which is what makes a Radon-Nikodym derivative and a KL-divergence loss exist. They state conditions on each transformation: a finite-rank perturbation with image in the Cameron-Martin space, plus a spectral condition that keeps each layer invertible. Four concrete flows are given, and a conditional variant learns to map measurement data to flow parameters so new datasets do not require full retraining. Numerical experiments on an elliptic source problem and steady-state Darcy flow match preconditioned Crank-Nicolson posteriors and confirm the claimed mesh independence.

What carries the argument

The central object is the layer $f^{(n)}_\theta(u)=u+F^{(n)}_\theta(u)$ on a separable Hilbert space, with the image of $F^{(n)}$ contained in the Cameron-Martin space $H$ of the Gaussian prior, the operator $F^{(n)}$ of finite rank, and the point spectrum of $DF^{(n)}$ avoiding $(-\infty,-1]$. The key identity is the Radon-Nikodym formula of Theorem 2.4, which expresses the density of the transformed measure through a product of Fredholm-Carleman determinants, here determinants for operators of the form $I+K$ with $K$ finite rank, multiplied by a Cameron-Martin-type exponential. This identity is what converts KL-divergence minimization into a Monte-Carlo-estimable loss, and the spectral condition is what guarantees each layer is invertible and the measures stay mutually absolutely continuous.

What would settle it

For the single-layer constant shift $f(u)=u+m$ with $m$ in the Cameron-Martin space, compare the printed Theorem 2.4 density with the classical Cameron-Martin density $\exp(\langle m,u\rangle_H - \tfrac12\|m\|_H^2)$; this fixes the sign of the inner-product term and decides whether the loss used by Algorithms 1 and 3 is the correct Radon-Nikodym derivative.

Watch

Extended reading notes

Core claim

The paper claims that for a composition $f_\theta = (I+F^{(N)})\circ\cdots\circ(I+F^{(1)})$ acting on a separable Hilbert space with Gaussian prior $\mu_0$, if each $F^{(n)}$ is finite rank with image in the Cameron-Martin space, each layer is bijective, and no point spectrum of $DF^{(n)}$ lies in $(-\infty,-1]$, then the push-forward measure is equivalent to the prior and its Radon-Nikodym derivative is $$\frac{d\mu_{f_\$\theta$}}{d\mu_0}(f_\$\theta$(u)) = \prod_{n=1}^N \left|\det_1\big($Df^{{(n)}}$(u_{n-1})\big)\right|^{-1} \exp\!\left(\tfrac12\langle f_\$\theta$(u)-u,\,f_\$\theta$(u)-u\rangle_H + \langle u,\,u-f_\$\theta$(u)\rangle_H\right).$$ This explicit density turns the infinite-dimensional change of measure into a computable loss, so the flow parameters can be optimized by stochastic gradient methods to approximate the posterior. The paper then constructs four layers, functional Householder, projected transformation, planar, and Sylvester flows, that satisfy the conditions, and a neural-network-controlled conditional version that produces an approximate posterior for new measurement data after a single forward pass.

Load-bearing premise

The load-bearing premise is that the true posterior differs from the prior mainly through the first $M$ eigenmodes of the prior covariance, because every flow layer is a finite-rank perturbation built from those modes and leaves all higher modes untouched.

Editorial extensions

If this is right

  • If the conditions hold, the KL divergence between the flow-induced measure and the posterior can be estimated from prior samples, making posterior approximation a tractable stochastic optimization problem in function space.
  • The discretization-invariance theorem means the same trained flow can be evaluated on any mesh, so the method belongs to the Bayesianize-then-discretize family and avoids retraining as grids are refined.
  • The conditional variant CNF-iVI lets a new measurement dataset produce an approximate posterior immediately, with an additional warm-start retraining step for higher accuracy.
  • Numerical comparisons with pCN on an elliptic source problem and steady-state Darcy flow show similar posterior means and covariance functions while yielding effectively independent samples with much higher effective sample sizes than the MCMC baseline.

Reading between the lines

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

  • Editorial extension: because every concrete flow layer is built from the first $M$ prior eigenvectors, the method's expressive range is confined to a finite-dimensional subspace, and problems where the data chiefly inform higher modes would require a different construction.
  • Editorial extension: the conditional network encodes measurement information by projecting onto the same prior eigenbasis, so the amortized variant inherits the mesh dependence of that truncation rather than being fully mesh-independent.
  • Editorial extension: replacing the fixed prior eigenbasis with a learned or wavelet basis could extend the same conditions to non-Gaussian or Besov priors, but the paper does not demonstrate such an extension.
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

4 major / 6 minor

Summary. The paper develops an infinite-dimensional variational inference method for Bayesian inverse problems of PDEs. Starting from a Gaussian prior on a separable Hilbert space, the authors define functional normalizing flows as compositions of maps f^(n)(u)=u+F^(n)(u), give conditions under which the push-forward is equivalent to the prior, and derive a Radon-Nikodym derivative formula that is then used as the KL loss in Algorithms 1 and 3. Four concrete layers are proposed (functional Householder, projected transformation, planar, and Sylvester flows), a discretization-invariance theorem is stated, and a conditional variant (CNF-iVI) is introduced to amortize training across measurement data. Numerical experiments for a linear elliptic inverse problem and a nonlinear Darcy-flow inverse problem compare the approximate posteriors with pCN samples and report mesh-independent behavior.

Significance. If the main formula and the supporting theory are corrected, the contribution is valuable to the PDE-constrained Bayesian inversion community: it provides one of the few infinite-dimensional normalizing-flow constructions with an explicit RN derivative, together with concrete layer families and a public code repository. The paper is also to be credited for benchmarking against pCN, for demonstrating quantitative covariance comparisons, and for attempting to verify discretization invariance numerically rather than only claiming it. However, the central RN-derivative formula contains a sign error that makes the printed theorem false, the discretization-invariance theorem is stated under hypotheses that the numerical examples do not satisfy, and the conditional training gradient in Algorithm 2 appears to miss a score term. These issues currently block acceptance.

major comments (4)
  1. [Theorem 2.4 and Appendix 7.3] The printed Radon-Nikodym derivative has the wrong sign in the linear term. The theorem states the exponent ½⟨fθ(u)−u, fθ(u)−u⟩_H + ⟨u, u−fθ(u)⟩_H, but the one-layer derivation in Appendix 7.3 gives Λ_F(u) = |det1(I+DF(u))| exp(−½∥F(u)∥²_H − ⟨u,F(u)⟩_H), so the reciprocal density must contain +⟨u,F(u)⟩_H = +⟨u, fθ(u)−u⟩_H. The final display of Appendix 7.3 repeats the incorrect sign instead of correcting it. For the constant shift f(u)=u+m with m∈H, the printed formula yields exp(½∥m∥²_H − ⟨u,m⟩_H), whereas the Cameron-Martin formula gives exp(½∥m∥²_H + ⟨u,m⟩_H). Since the same expression is used as the loss in Section 2.2 and in Algorithms 1 and 3, the optimization objective as written is not the KL divergence to the posterior. The sign must be flipped throughout Theorem 2.4, Theorems 2.5–2.8, and the algorithms.
  2. [Theorem 2.10 and Sections 4.1, 4.2] The discretization-invariance theorem assumes Hu is continuously embedded in C(D), but the numerical examples use Hu = L²(Ω) in Section 4.1 and Hu = L∞(Ω) in Section 4.2. Neither space embeds continuously into C(D) for the domains considered, so the stated theorem does not apply to the examples that are used to verify discretization invariance. The proof of Theorem 2.10 in Appendix 7.8 works under different, partially unstated regularity conditions (it invokes pointwise evaluation of a∈K⊂C(D) and Riemann-sum convergence). The paper needs to either restrict the claims to settings satisfying the theorem's hypotheses, or extend the theorem and the examples consistently.
  3. [Sections 2.3.2, 2.4.2, and Algorithms 1–3] All four concrete flow layers are parameterized so that Im(F^(n)) ⊂ span{φ_1,…,φ_M}, the span of the first M prior eigenvectors. Consequently fθ(u)−u always lies in that M-dimensional subspace, and every component of u orthogonal to span{φ_1,…,φ_M} is left unchanged by the push-forward. The paper never states this structural limitation: any posterior deviation in the orthogonal complement is invisible to the approximation, regardless of training time. The role of M as a modeling assumption should be made explicit, and the paper should discuss whether the chosen numerical problems are ones where the data mainly inform the leading modes.
  4. [Algorithm 2, line 7] The gradient estimator for conditional training appears to be missing the score-function term. The objective is q(λ)=E_{ν_λ(v)}[log(dν_λ/dµ0)−log(dµ/dµ0)], and line 6 draws samples from ν_{λ_k}(v_i). If samples are drawn from the measure being optimized, the unbiased gradient of E_{u∼ν_λ}[f_λ(u)] is E[∇_λ f_λ(u) + f_λ(u)∇_λ log p_λ(u)], where p_λ=dν_λ/dµ0. The displayed formula only differentiates the log-ratio at the sampled points and omits the second term. Unless the authors intend a reparameterized gradient (which is not stated), the update is biased and the CNF-iVI training in Section 5 is not minimizing the stated objective.
minor comments (6)
  1. [Section 2.4.1] The sentence introducing eq. (2.10) says 'Each layer of functional Householder flow employs the transformation'; it should say functional planar flow.
  2. [Section 2.4.2] In the definition of functional Sylvester flow, the domains of A_n and B_n are transposed: A_n should map R^M to H_u and B_n should map H_u to R^M, consistent with the subsequent parameterization and with Theorem 2.9. As printed, the composition A_n h(B_n u + b_n) is not well-defined.
  3. [Appendix 7.5 and 7.6] The proofs of Theorems 2.7, 2.8, 2.5, and 2.6 twice state the eigenvalue condition as 'not in (−∞,1]' instead of 'not in (−∞,−1]'. This is a typo but should be fixed because the numerical condition imposed on the diagonal elements is > −1.
  4. [Table 9] The entry c(x1,x2) = −0.455003 for the 50×50 discretization level differs from all other rows by two orders of magnitude and is almost certainly a decimal-point typo. As printed, this figure contradicts the discretization-invariance claim and should be corrected.
  5. [Table 2 and surrounding text] There are several typographical errors, including '0.0.02755' in Table 2 and 'discreate' in Tables 2, 3, 5, 6, 8, 9, 11, and 12. These should be corrected.
  6. [Section 4.1.2] The text states that M=20 is 'in Subsection 2.4', but functional projected transformation flow is introduced in Subsection 2.3.2.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the central Radon-Nikodym-derivative derivation is self-contained from standard Gaussian measure theory, and the self-citations are background only.

full rationale

The claimed derivation chain is not circular. The paper's central theoretical object is the Radon-Nikodym derivative in Theorem 2.4, obtained by applying Bogachev's Theorems 5.8.3, 6.6.7 and Corollary 6.6.8 to the finite-rank, Cameron-Martin-valued perturbation F(u) (Appendix 7.3), and the equivalence statement Theorem 2.3 is justified via Example 10.27 of Bogachev. None of these inputs assumes the target posterior or the final loss; the posterior enters only through the standard likelihood exp(-Phi) in Bayes' formula. The four flow layers are checked against the conditions of Theorem 2.4 in Appendices 7.5 and 7.6 rather than being tuned to reproduce the numerical outputs. The numerical validation compares posterior means and covariances against pCN samples, an external benchmark independent of the fitted flow parameters. The authors' own prior work (e.g., [21,23,24,44]) is cited only for the general infinite-dimensional variational inference background and does not supply the load-bearing RN formula or the measure-equivalence conditions. There is therefore no fitted input renamed as a prediction and no self-citation chain forcing the result. One non-circular correctness concern should be flagged separately: the printed exponent in Theorem 2.4, +<u, u-f(u)>_H, differs in sign from the final line of Appendix 7.3, which derives +<u, f(u)-u>_H; for a constant shift this contradicts the Cameron-Martin formula. That inconsistency is a mathematical error in the statement, not circularity, and does not change the circularity score.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new particles or physical entities; the only new items are the flow architectures, which are mathematical constructions. The main assumptions are Gaussian priors, measure equivalence as a requirement, and a continuity/compactness setup for discretization invariance. The finite-rank choice M is a hand-selected parameter, not derived from first principles.

free parameters (1)
  • M (projection dimension) = 20 (in main experiments), 10 (conditional flow)
    Choice of rank for the finite-rank perturbations; the approximation quality of the posterior depends on this hand-picked dimension. Introduced in Sections 2.3.2 and 2.4.2.
assumptions (4)
  • domain assumption Prior is a Gaussian measure µ0 = N(0,C) on a separable Hilbert space, and the posterior has density (2.3) against µ0.
    Invoked throughout Section 2.1 from [47]; the entire method is restricted to Gaussian priors.
  • domain assumption The approximating measures must be equivalent to the prior for the KL divergence to be well-defined, as required in [38,39].
    Section 2.1, the motivation for Theorem 2.3.
  • standard math Bogachev's theorems on Radon-Nikodym derivatives of Gaussian measures under shifts (Bogachev, 1998, Theorems 5.8.3, 6.6.7, Corollary 6.6.8) are correct and valid under the stated finite-rank and Cameron-Martin conditions.
    Foundation of Theorem 2.4 proof in Appendix 7.3.
  • domain assumption Point evaluations of candidate functions are well-defined and Riemann-sum approximations converge for functions in a compact set K of C(D); the theorem additionally assumes Hu embeds continuously into C(D).
    Theorem 2.10 proof in Appendix 7.8; the numerical examples use Hu = L2(Ω), which does not embed continuously into C(Ω), so the theoretical condition is not met in the applications.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Functional normalizing flow for statistical inverse problems of partial differential equations." pith.science (2026). https://pith.science/paper/XZN3YY57

@misc{pith2026241113277,
  author       = {Pith},
  title        = {Pith review of: Functional normalizing flow for statistical inverse problems of partial differential equations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XZN3YY57}},
  note         = {Machine review of arXiv:2411.13277}
}
read the original abstract

Inverse problems of partial differential equations are ubiquitous across various scientific disciplines and can be formulated as statistical inference problems using Bayes' theorem. To address large-scale problems, it is crucial to develop discretization-invariant algorithms, which can be achieved by formulating methods directly in infinite-dimensional space. We propose a novel normalizing flow based infinite-dimensional variational inference method (NF-iVI) to extract posterior information efficiently. Specifically, by introducing well-defined transformations, the prior in Bayes' formula is transformed into post-transformed measures that approximate the posterior. To circumvent the issue of mutually singular probability measures, we formulate general conditions for the employed transformations. As guiding principles, these conditions yield four concrete transformations. Additionally, to minimize computational demands, we have developed a conditional normalizing flow variant, termed CNF-iVI, which is adapt at processing measurement data of varying dimensions while requiring minimal computational resources. We apply the proposed algorithms to three typical inverse problems governed by the simple smooth equation, the steady-state Darcy flow equation, and the electric impedance tomography. Numerical results confirm our theoretical findings, illustrate the efficiency of our algorithms, and verify the discretization-invariant property.

Figures

Figures reproduced from arXiv: 2411.13277 by the authors.

Figure 1
Figure 1. A comparison of the approximate posterior obtained by functional Householder flow with 24 layers and the posterior generated by the pCN algorithm. The blue shade area represents the 95% credibility region of estimated posterior. (a): Result of the approximate posterior generated by functional Householder flow compared with the background truth of u. (b): Result of the posterior generated by pCN algorithm compared wi… view at source ↗
Figure 2
Figure 2. The estimated variance and covariance functions obtained by the pCN algorithm (blue solid line), and NF-iVI algorithm (red dashed line). (a): The covariance function c(x, y) on all the mesh point pairs {(xi , xi)} Ng i=1; (b): The covariance function c(x, y) on the mesh points {(xi , xi+10)} Ng−10 i=1 ; (c): The covari￾ance function c(x, y) on the mesh points {(xi , xi+20)} Ng−20 i=1 . (a) NF-iVI (b) pCN (c) Differe… view at source ↗
Figure 3
Figure 3. The comparison of covariance obtained by two method. (a): The covariance given by NF-iVI method; (b): The covariance given by the pCN method; (c): The difference between the covariance obtained by NF-iVI and pCN sampling methods. 24 [PITH_FULL_IMAGE:figures/full_fig_p024_3.png] view at source ↗
Figures from the paper (14 more)
Figure 4
Figure 4. Figure 4: The comparison of approximate posterior generated by functional Householder flow with different mesh size n = {50, 75, 100, 200, 300} respectively. The blue shade area represents the 95% credibility region of approximate posterior [PITH_FULL_IMAGE:figures/full_fig_p02…
Figure 5
Figure 5. Figure 5: , we present the results obtained from the 5-layer functional projected transformation flow. We observe that these results are consistent with the posterior generated by the pCN algorithm, as shown in subfigure (b) of [PITH_FULL_IMAGE:figures/full_fig_p027_5.png]
Figure 6
Figure 6. Figure 6: The comparison of posterior covariance of u with the mesh size n = 100 respectively. (a): The covariance given by NF-iVI method; (b): The covariance given by the pCN sampling method; (c): The difference between the covariance obtained by NF-iVI and pCN sampling methods…
Figure 7
Figure 7. Figure 7: The estimated variance and covariance functions obtained by the pCN method (blue solid line), Householder NF-iVI method (green dashed line) and projected transformation NF-iVI (red dashed line). (a)(d): The covariance function c(x, y) on all the mesh point pairs {(xi ,…
Figure 8
Figure 8. Figure 8: The comparison of the approximate posterior obtained by functional planar flow with 24 layers and the background truth. (a)(d): The background truth of u. (b): The mean of approximate posterior obtained by functional planar flow with 24 layers. (c):The variance functio…
Figure 9
Figure 9. Figure 9: The estimated variance and covariance functions obtained by the pCN algorithm (blue solid line), and NF-iVI algorithm (red dashed line). (a): The covariance function c(x, y) on all the mesh point pairs {(xi , xi)} Ng i=1; (b): The covariance function c(x, y) on the mes…
Figure 10
Figure 10. Figure 10: The comparison of approximation ability between functional planar flow and functional Sylvester flow. We use 5 layers for both flow models to approximate the posterior. (a)(d): The background truth of u. (b): The mean of approximate posterior obtained by functional Sy…
Figure 11
Figure 11. Figure 11: The estimated variance and covariance functions obtained by the pCN method (blue solid line), planar NF-iVI method (green dashed line) and Sylvester NF-iVI (red dashed line). (a)(d): The covariance function c(x, y) on all the mesh point pairs {(xi , xi)} Ng i=1; (b)(e…
Figure 12
Figure 12. Figure 12: The comparison of the approximate posterior obtained by CNF-iVI method and the pCN method. The estimated variance and covariance functions obtained by the pCN algorithm are drawn in blue solid line, and by CNF-iVI algorithm the red dashed line. (a): The background tru…
Figure 13
Figure 13. Figure 13: The comparison of the approximate posterior obtained by CNF-iVI method and the pCN method. The estimated variance and covariance functions obtained by the pCN algorithm are drawn in blue solid line, and by CNF-iVI algorithm the red dashed line. (a): The background tru…
Figure 14
Figure 14. Figure 14: The comparison of the approximate posterior obtained by CNF-iVI method and the pCN method. The estimated variance and covariance functions obtained by the pCN algorithm are drawn in blue solid line, and by CNF-iVI algorithm the red dashed line. (a): The background tru…
Figure 15
Figure 15. Figure 15: The comparison of the approximate posterior obtained by retrained NF-iVI method and the pCN method. The estimated variance and covariance functions obtained by the pCN algorithm are drawn in blue solid line, and by retrained NF-iVI algorithm the red dashed line. (a): …
Figure 16
Figure 16. Figure 16: The comparison of the approximate posterior obtained by retrained NF-iVI method and the pCN method. The estimated variance and covariance functions obtained by the pCN algorithm are drawn in blue solid line, and by retrained NF-iVI algorithm the red dashed line. (a): …
Figure 17
Figure 17. Figure 17: The comparison of the approximate posterior obtained by retrained NF-iVI method and the pCN method. The estimated variance and covariance functions obtained by the pCN algorithm are drawn in blue solid line, and by retrained NF-iVI algorithm the red dashed line. (a): …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 49 canonical work pages

  1. [1]

    Agapiou, O

    S. Agapiou, O. Papaspiliopoulos, D. Sanz-Alonso, and A. M. Stuart, Importance sampling: intrinsic dimension and computational cost , Statist. Sci. 32 (2017), no. 3, 405–431. MR 3696003

  2. [2]

    27 (2018), 1–111

    Martin Benning and Martin Burger, Modern regularization methods for inverse problems , Acta Numer. 27 (2018), 1–111. MR 3826506 50

  3. [3]

    Farrell, and Andrew M

    Alexandros Beskos, Mark Girolami, Shiwei Lan, Patrick E. Farrell, and Andrew M. Stuart, Geometric MCMC for infinite-dimensional inverse problems , J. Comput. Phys. 335 (2017), 327–351. MR 3612501

  4. [4]

    Muzaffer, and Andrew M

    Alexandros Beskos, Ajay Jasra, Ege A. Muzaffer, and Andrew M. Stuart, Sequential Monte Carlo methods for Bayesian elliptic inverse problems , Stat. Comput. 25 (2015), no. 4, 727–737. MR 3360488

  5. [5]

    Bishop, Pattern Recognition and Machine Learning, Information Science and Statistics, Springer, New York, 2006

    Christopher M. Bishop, Pattern Recognition and Machine Learning, Information Science and Statistics, Springer, New York, 2006. MR 2247587

  6. [6]

    Blei, Alp Kucukelbir, and Jon D

    David M. Blei, Alp Kucukelbir, and Jon D. McAuliffe, Variational inference: a review for statisticians , J. Amer. Statist. Assoc. 112 (2017), no. 518, 859–877. MR 3671776

  7. [7]

    Bogachev, Gaussian Measures, American Mathematical Society, Providence, RI, 1998

    Vladimir I. Bogachev, Gaussian Measures, American Mathematical Society, Providence, RI, 1998. MR 1642391

  8. [8]

    MR 2663405

    , Differentiable Measures and the Malliavin Calculus , Mathematical Surveys and Monographs, American Mathematical Society, Providence, RI, 2010. MR 2663405

Show all 50 references
  1. [9]

    Tan Bui-Thanh, Omar Ghattas, James Martin, and Georg Stadler, A computational framework for infinite- dimensional Bayesian inverse problems Part I: The linearized case, with application to global seismic inversion , SIAM J. Sci. Comput. 35 (2013), no. 6, A2494–A2523. MR 3126997

  2. [10]

    Nguyen, FEM-based discretization-invariant MCMC methods for PDE-constrained Bayesian inverse problems , Inverse Probl

    Tan Bui-Thanh and Quoc P. Nguyen, FEM-based discretization-invariant MCMC methods for PDE-constrained Bayesian inverse problems , Inverse Probl. Imaging 10 (2016), no. 4, 943–975. MR 3610747

  3. [11]

    2, 025008, 38

    Daniela Calvetti, Matthew Dunlop, Erkki Somersalo, and Andrew Stuart, Iterative updating of model error for Bayesian inversion , Inverse Problems 34 (2018), no. 2, 025008, 38. MR 3751052

  4. [12]

    MR 4628023

    Daniela Calvetti and Erkki Somersalo, Bayesian Scientific Computing , Springer, Cham, 2023. MR 4628023

  5. [13]

    Peng Chen and Omar Ghattas, Stein variational reduced basis Bayesian inversion , SIAM J. Sci. Comput. 43 (2021), no. 2, A1163–A1193. MR 4235177

  6. [14]

    Cotter, Masoumeh Dashti, James C

    Simon L. Cotter, Masoumeh Dashti, James C. Robinson, and Andrew M. Stuart, Bayesian inverse problems for functions and applications to fluid mechanics , Inverse Problems 25 (2009), no. 11, 115008, 43. MR 2558668

  7. [15]

    Cotter, Gareth O

    Simon L. Cotter, Gareth O. Roberts, Andrew M. Stuart, and David White, MCMC methods for functions: modifying old algorithms to make them faster , Statist. Sci. 28 (2013), no. 3, 424–446. MR 3135540

  8. [16]

    Imaging 6 (2012), no

    Masoumeh Dashti, Stephen Harris, and Andrew Stuart, Besov priors for Bayesian inverse problems , Inverse Probl. Imaging 6 (2012), no. 2, 183–200. MR 2942737

  9. [17]

    Stuart, The Bayesian approach to inverse problems, Handbook of uncertainty quantification

    Masoumeh Dashti and Andrew M. Stuart, The Bayesian approach to inverse problems, Handbook of uncertainty quantification. Vol. 1, 2, 3, Springer, Cham, 2017, pp. 311–428. MR 3839555

  10. [18]

    Laurent Dinh, Jascha Sohl-Dickstein, and Samy Bengio, Density estimation using Real NVP , International Conference on Learning Representations, 2017

  11. [19]

    Andrzej Granas and James Dugundji, Fixed point theory, Springer Monographs in Mathematics, Springer-Verlag, New York, 2003

  12. [20]

    Mallick,A variational Bayesian approach for inverse problems with skew-t error distributions , J

    Nilabja Guha, Xiaoqing Wu, Yalchin Efendiev, Bangti Jin, and Bani K. Mallick,A variational Bayesian approach for inverse problems with skew-t error distributions , J. Comput. Phys. 301 (2015), 377–393. MR 3402736

  13. [21]

    Junxiong Jia, Peijun Li, and Deyu Meng, Stein variational gradient descent on infinite-dimensional space and applications to statistical inverse problems , SIAM J. Numer. Anal. 60 (2022), no. 4, 2225–2252. MR 4469515

  14. [22]

    Imaging 15 (2021), no

    Junxiong Jia, Jigen Peng, and Jinghuai Gao, Posterior contraction for empirical Bayesian approach to inverse problems under non-diagonal assumption , Inverse Probl. Imaging 15 (2021), no. 2, 201–228. MR 4214068

  15. [23]

    Junxiong Jia, Yanni Wu, Peijun Li, and Deyu Meng, Variational inverting network for statistical inverse prob- lems of partial differential equations , J. Mach. Learn. Res. 24 (2023), paper no. 201, 60. MR 4633590

  16. [24]

    Junxiong Jia, Qian Zhao, Zongben Xu, Deyu Meng, and Yee Leung, Variational Bayes’ method for functions with applications to some inverse problems , SIAM J. Sci. Comput. 43 (2021), no. 1, A355–A383. MR 4205093

  17. [25]

    Bangti Jin, A variational Bayesian method to inverse problems with impulsive noise , J. Comput. Phys. 231 (2012), no. 2, 423–435. MR 2872083

  18. [26]

    Com- put

    Bangti Jin and Jun Zou, Hierarchical Bayesian inference for ill-posed problems via variational method , J. Com- put. Phys. 229 (2010), no. 19, 7317–7343. MR 2677781

  19. [27]

    MR 2102218

    Jari Kaipio and Erkki Somersalo, Statistical and Computational Inverse Problems , Applied Mathematical Sci- ences, Springer-Verlag, New York, 2005. MR 2102218

  20. [28]

    MR 3025302

    Andreas Kirsch, An Introduction to the Mathematical Theory of Inverse Problems , second ed., Applied Mathe- matical Sciences, Springer, New York, 2011. MR 3025302

  21. [29]

    Nikola Kovachki, Zongyi Li, Burigede Liu, Kamyar Azizzadenesheli, Kaushik Bhattacharya, Andrew Stuart, and Anima Anandkumar, Neural operator: learning maps between function spaces with applications to PDEs , J. Mach. Learn. Res. 24 (2023), paper no. 89, 97. MR 4582511

  22. [30]

    Blei, Automatic differentiation variational inference, J

    Alp Kucukelbir, Dustin Tran, Rajesh Ranganath, Andrew Gelman, and David M. Blei, Automatic differentiation variational inference, J. Mach. Learn. Res. 18 (2017), paper no. 14, 45. MR 3634881 51

  23. [31]

    Kushner and G

    Harold J. Kushner and G. George Yin, Stochastic Approximation Algorithms and Applications , Applications of Mathematics, Springer-Verlag, New York, 1997. MR 1453116

  24. [32]

    5, 1537–1563

    Matti Lassas and Samuli Siltanen, Can one use total variation prior for edge-preserving Bayesian inversion? , Inverse Problems 20 (2004), no. 5, 1537–1563. MR 2109134

  25. [33]

    Zongyi Li, Nikola Kovachki, Kamyar Azizzadenesheli, Burigede Liu, Kaushik Bhattacharya, Andrew Stuart, and Anima Anandkumar, Fourier neural operator for parametric partial differential equations , International Conference on Learning Representations, 2021

  26. [34]

    Wenyuan Liao and Ou Wei, A fourth-order compact numerical scheme for three-dimensional acoustic wave equation with variable velocity , Recent advances in mathematical and statistical methods, Springer Proc. Math. Stat., Springer, Cham, 2018, pp. 279–289. MR 3879571

  27. [35]

    GuoJun Liu, Yang Liu, MaoZu Guo, Peng Li, and MingYu Li, Variational inference with gaussian mixture model and householder flow , Neural Networks 109 (2019), 43–55

  28. [36]

    Roberts, and Martin Sk¨ old, Non-Centered Parameterizations for Hierar- chical Models and Data Augmentation , Bayesian statistics, 7, Oxford Univ

    Omiros Papaspiliopoulos, Gareth O. Roberts, and Martin Sk¨ old, Non-Centered Parameterizations for Hierar- chical Models and Data Augmentation , Bayesian statistics, 7, Oxford Univ. Press, New York, 2003, pp. 307–326. MR 2003180

  29. [37]

    Pillai, Andrew M

    Natesh S. Pillai, Andrew M. Stuart, and Alexandre H. Thi´ ery,Noisy gradient flow from a random walk in Hilbert space, Stoch. Partial Differ. Equ. Anal. Comput. 2 (2014), no. 2, 196–232. MR 3249584

  30. [38]

    Pinski, Geoffrey Simpson, Andrew M

    Frederick J. Pinski, Geoffrey Simpson, Andrew M. Stuart, and Harald Weber, Algorithms for Kullback-Leibler approximation of probability measures in infinite dimensions , SIAM J. Sci. Comput. 37 (2015), no. 6, A2733– A2757. MR 3424069

  31. [39]

    Pinski, Geoffrey Simpson, Andrew M

    Frederick J. Pinski, Geoffrey Simpson, Andrew M. Stuart., and Harald Weber, Kullback-Leibler approximation for probability measures on infinite dimensional spaces , SIAM J. Math. Anal. 47 (2015), no. 6, 4091–4122. MR 3419882

  32. [40]

    Danilo Jimenez Rezende and Shakir Mohamed, Variational inference with normalizing flows , International Conference on Machine Learning 37 (2015), 1550–1558

  33. [41]

    Herbert Robbins and Sutton Monro, A stochastic approximation method , Ann. Math. Statistics 22 (1951), 400–407. MR 42668

  34. [42]

    Robert and George Casella, Monte Carlo Statistical Methods , Springer-Verlag, New York, 1999

    Christian P. Robert and George Casella, Monte Carlo Statistical Methods , Springer-Verlag, New York, 1999. MR 1707311

  35. [43]

    Daniel Rudolf and Bj¨ orn Sprungk, On a generalization of the preconditioned Crank-Nicolson metropolis algo- rithm, Found. Comput. Math. 18 (2018), no. 2, 309–343. MR 3777781

  36. [44]

    Jiaming Sui and Junxiong Jia, Non-centered parametric variational Bayes’ approach for hierarchical inverse problems of partial differential equations , Math. Comp. 93 (2024), no. 348, 1715–1760. MR 4730247

  37. [45]

    Jakub M Tomczak and Max Welling, Improving variational auto-encoders using householder flow , IEEE Trans- actions on Image Processing 29 (2020), 2487–2499

  38. [46]

    Rianne Van Den Berg, Leonard Hasenclever, Jakub M Tomczak, and Max Welling, Sylvester normalizing flows for variational inference , 34th Conference on Uncertainty in Artificial Intelligence, 2018, pp. 393–402

  39. [47]

    Weglein, Fernanda V

    Arthur B. Weglein, Fernanda V. Ara´ ujo, Paulo M. Carvalho, Robert H. Stolt, Kenneth H. Matson, Richard T. Coates, Dennis Corrigan, Douglas J. Foster, Simon A. Shaw, and Haiyan Zhang, Inverse scattering series and seismic exploration, Inverse Problems 19 (2003), no. 6, R27–R83...

  40. [48]

    Max Welling and Yee Whye Teh, Bayesian learning via stochastic gradient langevin dynamics , Proceedings of the 28th International Conference on Machine Learning, 2011, pp. 681–688

  41. [49]

    Pattern Analysis and Machine Intelligence 41 (2018), no

    Cheng Zhang, Judith B¨ utepage, Hedvig Kjellstr¨ om, and Stephan Mandt, Advances in variational inference , IEEE Trans. Pattern Analysis and Machine Intelligence 41 (2018), no. 8, 2008–2026

  42. [50]

    Imaging Sci

    Qingping Zhou, Tengchao Yu, Xiaoqun Zhang, and Jinglai Li, Bayesian inference and uncertainty quantification for medical image reconstruction with Poisson data , SIAM J. Imaging Sci. 13 (2020), no. 1, 29–52. MR 4048008 52

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.