Pith. sign in

REVIEW 1 major objections 5 minor 1 cited by

A global Lipschitz stability perspective for understanding approximate approaches in Bayesian sequential learning

T0 review · 1 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The prior-to-posterior map is pointwise globally Lipschitz under total variation, Hellinger, and 1-Wasserstein distances, so bounded per-step approximation errors yield a bounded learning error in sequential Bayesian learning.

desk verdict Genuinely new global Lipschitz stability results under Hellinger and Wasserstein distances, but Theorem 16 as stated needs the assumption Q0=P0 (or an initial-error term) before the recursive error bounds are correct. read the letter →

arxiv 2507.20379 v1 pith:7ROTUSG7 submitted 2025-07-27 math.ST cs.NAmath.NAstat.TH

Liliang Wang , Alex A. Gorodetsky This is my paper · ORCID
classification math.STcs.NAmath.NAstat.TH MSC 62F1562M2060B10
keywords BayesiansequentiallearningglobalLipschitzstabilityposteriortotalvariationdistanceHellingerWassersteinonlinevariationalinferenceerrorbounds
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 proves that the Bayesian update used at each step of sequential learning — taking a prior distribution to the posterior — is globally stable: the distance between two posteriors grows at most linearly with the distance between the two priors, uniformly over all priors, for total variation, Hellinger, and 1-Wasserstein distances. The stability constant depends only on the base prior's evidence and on problem constants, so it stays finite even when the approximate posterior is arbitrarily far from the true posterior. From this, the paper derives the first general, non-asymptotic error analysis for approximate Bayesian sequential methods: the learning error at time $k$ is bounded by a weighted sum of per-step incremental approximation errors, and a second bound can be computed from the approximate posteriors alone. It also identifies sufficient conditions under which assimilating a new datum shrinks the learning error. If correct, this turns error analysis of any new sequential algorithm into a per-step verification task, with a finite accuracy certificate for the whole run.

What carries the argument

The load-bearing object is the prior-to-posterior map $F_k$ together with its pointwise global Lipschitz constant $K(\mu;y_k) = C(y_k)/Z_k(\mu)$, where $Z_k(\mu)$ is the evidence (marginal likelihood) of the base prior and $C(y_k)$ is a supremum of the likelihood, or of a product of the observation and transition models, or (for the Wasserstein statements) a Lipschitz constant of those models. The recursive mechanism is a triangle-inequality decomposition of the learning error at step $k$ into a propagated part, $d(F_k(P_{k-1}),F_k(Q_{k-1}))$, which the Lipschitz inequality compresses to a multiple of the previous step's error, plus the incremental approximation error $d(F_k(Q_{k-1}),Q_k)$. Recursing from step $k$ down to step 1 telescopes these factors into the weighted sums of Theorem 16, in which each incremental error is multiplied by a product of constants over the evidence, and one of the two resulting bounds replaces the intractable evidence of the true posteriors with evidence of the approximate posteriors, making it estimable in practice. The bounded-diameter assumption on the underlying space is what keeps the Wasserstein constants finite.

What would settle it

On $X=\mathbb{R}$ with a Gaussian observation model (bounded and Lipschitz likelihood), fix a base prior $\mu$ and take $\mu'$ to be Gaussian with the same mean and variance $\sigma'^2\to\infty$; compute the ratio $W_1(F_k(\mu),F_k(\mu'))/W_1(\mu,\mu')$ at a fixed step $k$. If the ratio is unbounded as $\sigma'^2$ grows, then the pointwise global Lipschitz claim under $W_1$ genuinely needs a bounded space, confirming the role of Assumption IP-SE-PS.4; if the ratio stays bounded, the bounded-diameter assumption could be relaxed and the Wasserstein results would extend to unbounded settings.

Watch

Extended reading notes

Core claim

The central discovery is the pointwise global Lipschitz continuity of the prior-to-posterior map $F_k$: for every admissible prior $\mu$, there is a constant $K(\mu;y_k)$ such that $d(F_k(\mu),F_k(\mu')) \le K(\mu;y_k)\,d(\mu,\mu')$ for every other admissible prior $\mu'$, where $d$ is total variation, Hellinger, or 1-Wasserstein distance, in inverse problems, state estimation, and parameter-state estimation. The constant has the form $C(y_k)/Z_k(\mu)$, where $Z_k(\mu)$ is the evidence of the base prior and $C(y_k)$ collects boundedness or Lipschitz constants of the likelihood and transition models; it never involves $\mu'$. Because the constant does not depend on $\mu'$, the bound applies no matter how different the two priors are — precisely the regime of approximate sequential learning, where the approximate posterior $Q_{k-1}$ can be far from the true $P_{k-1}$. The theorem also halves the best known total-variation bound under the same assumptions. Theorem 16 converts this stability into an error analysis: $d(P_k,Q_k)$ is at most a sum over steps of the incremental approximation errors $d(Q_j,Q^*_j)$, each weighted by a product of ratios of likelihood constants to evidence terms, giving one linear-in-error stability bound and one bound computable from the approximate posteriors alone. Theorems 19–23 then give sufficient conditions under which the propagated part of the error shrinks, $d(P_k,Q^*_k) \le d(P_{k-1},Q_{k-1})$, offering a first general explanation of learning error decay.

Load-bearing premise

For the 1-Wasserstein results the space on which the unknown state or parameter lives must be bounded, with a finite maximum distance $D$ between any two points; this fails for common unbounded models such as linear Gaussian state-space systems, while the total-variation and Hellinger results do not need it.

Editorial extensions

If this is right

  • Bounded-input, bounded-output stability: if every step's incremental approximation error is bounded, the overall learning error $d(P_k,Q_k)$ stays bounded in total variation, Hellinger, or 1-Wasserstein distance, regardless of how large the prior error was.
  • Two runs of any approximate method whose per-step incremental errors are both bounded produce approximate posteriors $Q_k$ and $\tilde Q_k$ that remain within a bounded distance of each other and of the true posterior.
  • The second set of bounds is computable online: it depends only on the approximate posteriors and on single-datum evidences $Z_i(Q_{i-1})$, never on the intractable true posteriors, so practitioners can certify a finite accuracy envelope during learning.
  • Under the sufficient conditions of Theorems 19–23, assimilating the next datum can strictly shrink the learning error, giving a first general theoretical account of the error decay observed empirically in sequential Monte Carlo and other learners.
  • In state estimation with transition-observation products bounded above and below, composing the Lipschitz stability with forgetting results implies that a well-chosen alternative initial prior can beat the true prior, and that a larger immediate incremental error can be worthwhile if it improves later steps.

Reading between the lines

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

  • The bounded-diameter assumption (IP-SE-PS.4) does real work: on unbounded spaces such as linear Gaussian state-space models the 1-Wasserstein constants become vacuous ($D=\infty$), so the Wasserstein half of Theorem 14 is, as stated, a bounded-space result; the total-variation and Hellinger claims carry the unbounded settings.
  • Read as a template, the framework reduces the analysis of any new sequential algorithm to one method-specific task, bounding the per-step error $d(Q_j,Q^*_j)$, after which Theorem 16 supplies the whole-run certificate; the paper's online-VI corollaries show exactly how such per-step bounds are produced.
  • Because the Lipschitz constant is $C(y_k)/Z_k(\mu)$, an informative datum that raises the evidence automatically shrinks the factor propagating prior error, suggesting that learning error decay is generic whenever observations are informative, beyond the specific sufficient conditions the paper states.
  • A testable extension suggested by the framework: on a bounded state space, run an ensemble or particle filter from deliberately corrupted initial priors, record per-step incremental errors, and compare the simulated learning error with the estimable bound of Equation (21) to gauge how tight the certificate is in a realistic regime.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

1 major / 5 minor

Summary. The paper develops a non-asymptotic error analysis framework for approximate Bayesian sequential learning (BSL) in inverse problems, state estimation, and parameter-state estimation. It formalizes the prior-to-posterior map F_k, the exact intermediate posterior Q_k^\ast, the approximate posterior Q_k, and the incremental approximation error d(Q_k,Q_k^\ast). The main results are: Theorem 14 establishes pointwise global Lipschitz stability of F_k under total variation, Hellinger, and 1-Wasserstein distances with explicit constants in Table 1; Theorem 16 propagates bounded incremental errors into two sets of bounds on the learning error d(P_k,Q_k); Theorems 19–23 give sufficient conditions for learning-error decay under data assimilation; and Corollaries 25–26 apply the framework to online variational inference. The proofs are collected in appendices, and a one-dimensional inverse problem is used to validate the TV and Hellinger bounds numerically.

Significance. If Theorem 16 is corrected as described below, the framework is a genuinely useful general tool: the Lipschitz constants in Table 1 are explicit functions of likelihood bounds and evidence terms, no coefficients are fitted, and the TV bound improves the corresponding Sprungk (2020) bound by a factor of 1/2 under the same assumptions. The second set of bounds in Theorem 16 is estimable from approximate posteriors and single-step evidence terms, which is attractive for online variational inference and particle-type methods. The paper also gives the first, to my knowledge, global Lipschitz statements under Hellinger and Wasserstein distances, and the VI corollaries show how the framework can be specialized. The main limitation is that all 1-Wasserstein results require the metric space to be bounded (Assumption IP-SE-PS.4), so standard linear-Gaussian state-space models on unbounded state spaces are excluded; this is disclosed in the manuscript but should be stated more prominently in the abstract and conclusions.

major comments (1)
  1. [Section 3.3, Theorem 16 (proof in Appendix C)] Theorem 16 is false as stated because the assumption Q0=P0 is missing. In the base case of the induction, the proof uses the identity Q1^\ast = F1(P0) = P1 to conclude d(P1,Q1) \le d(Q1^\ast,Q1); this identity holds only when Q0=P0, whereas the theorem assumes only Pi\in\bar P_{i+1} for i\in[0,k-1] and Qi\in\bar P_{i+1} for i\in[1,k-1]. A concrete counterexample within the paper's assumptions is the inverse problem on X=[0,1] with Lebesgue density h(y_k,x)=1, so that F_k is the identity map, and the choice Q_k=Q_{k-1} for every k. Then every incremental error d(Q_j^\ast,Q_j) vanishes and Equation (20) gives d(P_k,Q_k)\le 0, while the true learning error is d(P_k,Q_k)=d(P_0,Q_0)>0 whenever Q0\neq P0. The fix is local: add Q0=P0 (and Q0\in\bar P_1) to the assumptions, or add the propagated initial-error term to the bounds. Corollaries 25 and 26 inherit the same missing assumption from Theorem 16 and need the same correction.
minor comments (5)
  1. [Section 3.1, Assumption IP-SE-PS.4] The bounded-diameter condition D<\infty is load-bearing for every 1-Wasserstein statement in Theorem 14, including the rows for state estimation and parameter-state estimation in Table 1. This is explicitly disclosed, but the abstract and conclusions should state that the Wasserstein results apply only under bounded state/parameter spaces; the standard linear-Gaussian state-space model on an unbounded space is not covered.
  2. [Theorem 16 statements (20)–(21)] The theorem is stated only for k\ge 2 and does not give the k=1 case. With the corrected assumption Q0=P0, the k=1 bound is simply d(P1,Q1)\le d(Q1^\ast,Q1); please add this case for completeness.
  3. [Appendix B.3.1, proof of Lemma 48] There is a typo: 'Tonelli's thoerem' should read 'Tonelli's theorem'.
  4. [References] The citation 'Freitas and Neil (2001)' is incomplete; Sequential Monte Carlo Methods in Practice is edited by Doucet, de Freitas, and Gordon, and the author list and year should be corrected accordingly.
  5. [Section 6, numerical example] The experiment uses the same observation y at every update step, which is not a standard sequential data regime. Please state explicitly that this is a repeated-observation stress test for validating the bounds rather than a sequential-data benchmark.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the Lipschitz stability theorem and the recursive learning-error bounds are derived from explicit assumptions and the triangle inequality, with no fitted parameters renamed as predictions.

full rationale

The derivation chain is self-contained. Theorem 14 states explicit Lipschitz constants K(mu; y_k) that are functions of likelihood/transition upper bounds and the evidence Z_k(mu); when mu = P_{k-1}, Z_k(P_{k-1}) equals the fixed predictive density p(y_k | Y_{1:k-1}), not a fitted or tunable quantity. Theorem 16 recursively applies the triangle inequality and Theorem 14 from step k down to 1, and both bound formulas (20) and (21) are constructed by multiplying per-step constants; the second set depends only on approximate posteriors Q_j and their evidence terms Z_i(Q_{i-1}), which are model integrals, not parameters chosen to force the bound. The Wasserstein part imports the evidence Lipschitz bound from Sprungk (2020) but reproduces a complete proof, and it is an external, independently stated mathematical result rather than a self-citation. The paper's self-citations (e.g., Wang and Gorodetsky 2025) are used only as examples of online variational inference methods, not as load-bearing premises. The reviewer-identified omission in Theorem 16 that Q0 = P0 is not explicitly stated is a correctness gap in the theorem statement, not a circularity: with Q0 different from P0 the stated bound omits the propagated initial error and is false, but the derivation itself does not reduce to its own inputs. No fitted-input-called-prediction, self-definitional, or uniqueness-imported-from-authors pattern is present.

Assumptions & free parameters 0 free parameters · 8 assumptions · 0 invented entities

The framework introduces no fitted free parameters and no new entities such as particles, forces, or mediators. All constants are explicit model-dependent quantities such as likelihood bounds, evidence terms, and metric-space diameters. The axioms are standard measure-theoretic tools plus stated regularity assumptions on likelihoods and transition kernels.

assumptions (8)
  • domain assumption Assumption IP.1: sup_x h(y_k, x) is finite for each data point y_k.
    Used in Theorem 14 for inverse problems under total variation and Hellinger distances; bounds the likelihood in Propositions 27 and 30.
  • domain assumption Assumption SE.1: C_T^h(y_k; k) = sup_{x_{k-1}} integral h_k(y_k, x) T_k(x, x_{k-1}) dx is finite.
    Provides the total variation and Hellinger Lipschitz constant for state estimation in Propositions 28 and 31.
  • domain assumption Assumption PS.1: tilde C_T^h(y_k; k) as defined in Equation (12) is finite.
    Provides the total variation and Hellinger Lipschitz constant for parameter-state estimation in Propositions 29 and 32.
  • domain assumption Assumption IP.2: the likelihood h(y_k, x) is globally Lipschitz in x with best constant ||h||_Lip(y_k).
    Required for the 1-Wasserstein evidence Lipschitz bound in Lemma 46 and for Proposition 49.
  • domain assumption Assumptions SE.2 and PS.2 and PS.3: transition and observation models satisfy the displayed Lipschitz and metric-product conditions.
    Required for the 1-Wasserstein evidence bounds in Lemmas 47 and 48 and for Propositions 50 and 51.
  • domain assumption Assumption IP-SE-PS.4: the metric space is bounded, sup d(x, x') = D < infinity.
    Used in all 1-Wasserstein proofs to bound |f(x)| and to keep the Wasserstein Lipschitz constants finite; without it the Wasserstein results are not established.
  • domain assumption Exact and approximate posterior sequences remain admissible: P_i in bar P_{i+1} and Q_i in bar P_{i+1} for Theorem 16.
    Needed for the recursive application of Theorem 14 at every step; if an approximate posterior fails this admissibility condition, the recursion cannot continue.
  • standard math Standard analytic facts: Radon-Nikodym derivatives, Tonelli and Fubini theorems, Holder's inequality, and Kantorovich-Rubinstein duality.
    These are the main tools in the appendix proofs and are standard background results, not contributions of the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A global Lipschitz stability perspective for understanding approximate approaches in Bayesian sequential learning." pith.science (2026). https://pith.science/paper/7ROTUSG7

@misc{pith2026250720379,
  author       = {Pith},
  title        = {Pith review of: A global Lipschitz stability perspective for understanding approximate approaches in Bayesian sequential learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7ROTUSG7}},
  note         = {Machine review of arXiv:2507.20379}
}
read the original abstract

We establish a general, non-asymptotic error analysis framework for understanding the effects of incremental approximations made by practical approaches for Bayesian sequential learning (BSL) on their long-term inference performance. Our setting covers inverse problems, state estimation, and parameter-state estimation. In these settings, we bound the difference-termed the learning error-between the unknown true posterior and the approximate posterior computed by these approaches, using three widely used distribution metrics: total variation, Hellinger, and Wasserstein distances. This framework builds on our establishment of the global Lipschitz stability of the posterior with respect to the prior across these settings. To the best of our knowledge, this is the first work to establish such global Lipschitz stability under the Hellinger and Wasserstein distances and the first general error analysis framework for approximate BSL methods. Our framework offers two sets of upper bounds on the learning error. The first set demonstrates the stability of general approximate BSL methods with respect to the incremental approximation process, while the second set is estimable in many practical scenarios. Furthermore, as an initial step toward understanding the phenomenon of learning error decay, which is sometimes observed, we identify sufficient conditions under which data assimilation leads to learning error reduction.

Figures

Figures reproduced from arXiv: 2507.20379 by the authors.

Figure 1
Figure 1. Process of BSL. At each step k, BSL updates the prior Pk−1 using the data yk to obtain the posterior Pk. The symbol Fk denotes the map from a prior to its corresponding posterior at step k. 𝑃0 … … ≈ ≈ ≈ 𝑃1 𝑃2 𝑃𝑘−1 𝑃𝑘 ≈ 𝑄1 𝑄2 𝑄𝑘−1 𝑄𝑘 𝑦1 𝑦2 𝑦3 𝑦𝑘−1 𝑦𝑘 𝐹2 𝐹1 𝐹3 𝐹𝑘−1 𝐹𝑘 𝐹෠1 𝐹෠2 𝐹෠𝑘−1 𝐹෠3 𝐹෠𝑘 [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Procedure of approximate approaches for BSL. The exact posterior [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Break-down procedure of approximate approaches in BSL. The exact posterior [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: TV distance between posteriors, dT V (µk, µ′ k ), and its upper bound provided by Theorem 14. Because of the symmetry of the TV distance, Theorem 14 yields the bound Ch(y) Zk(µk−1)∨Zk(µ ′ k−1 ) dT V (µk−1, µ′ k−1 ). A zoom-in view of steps 12 through 20 is shown in pan…
Figure 5
Figure 5. Figure 5: Hellinger distance between posteriors, dH(µk, µ′ k ), and its upper bound from Theorem 14. Because the Hellinger distance is symmetric, Theorem 14 provides the bound 2 r Ch(y) Zk(µk−1)∨Zk(µ ′ k−1 ) dH(µk−1, µ′ k−1 ). A zoom-in view of steps 5 through 20 is shown in pan…

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Sequential Bayesian parameter-state estimation in dynamical systems with noisy and incomplete observations via a variational framework

    stat.ME 2025-12 conditional novelty 6.0 of 10

    FBOVI sequentially approximates the joint state-parameter posterior by a parameter marginal times a conditional Gaussian state distribution, with error bounds and demonstrations on pendulum, Lorenz-96, and a 51D conve...

Reference graph

Works this paper leans on

43 extracted references · 31 canonical work pages · cited by 1 Pith paper

  1. [1]

    On some Lipschitz -type functions

    Manisha Aggarwal and Stefan Cobzaș. On some Lipschitz -type functions. Journal of Mathematical Analysis and Applications, 517 0 (2): 0 126631, 2023. doi:10.1016/j.jmaa.2022.126631

    work page arXiv 2023

  2. [2]

    Data-driven approximation of stationary nonlinear filters with optimal transport maps

    Mohammad Al-Jarrah, Bamdad Hosseini, and Amirhossein Taghvaei. Data-driven approximation of stationary nonlinear filters with optimal transport maps. In IEEE 63rd Conference on Decision and Control (CDC), pages 2727--2733, 2024

    work page 2024

  3. [3]

    Fast filtering of non-Gaussian models using Amortized Optimal Transport Maps

    Mohammad Al-Jarrah, Bamdad Hosseini, and Amirhossein Taghvaei. Fast filtering of non-Gaussian models using amortized optimal transport maps. arXiv:2503.12633, 2025

    work page Pith review arXiv 2025

  4. [4]

    Stability and infinitesimal robustness of posterior distribution and posterior quantities

    Sanjib Basu. Stability and infinitesimal robustness of posterior distribution and posterior quantities. Statistics & Probability Letters, 71: 0 151--162, 1998. doi:10.1016/S0378-3758(98)00090-1

    work page doi:10.1016/s0378-3758(98)00090-1 1998

  5. [5]

    Uniform stability of posteriors

    Sanjib Basu. Uniform stability of posteriors. Statistics & Probability Letters, 46: 0 53--58, 2000. doi:10.1016/S0167-7152(99)00086-3

    work page doi:10.1016/s0167-7152(99)00086-3 2000

  6. [6]

    Berger and Elías Moreno

    James O. Berger and Elías Moreno. An overview of robust Bayesian analysis. Test, 3: 0 5--124, 1994. doi:10.1007/BF02562676

    doi:10.1007/bf02562676 1994

  7. [7]

    Bayesian robustness with more than one class of contamination

    Sudip Bose. Bayesian robustness with more than one class of contamination. J. Statist. Planning and Inference, 40: 0 177--187, 1994. doi:10.1016/0378-3758(94)90119-8

    work page doi:10.1016/0378-3758(94)90119-8 1994

  8. [8]

    Burt, Carl Edward Rasmussen, and Mark van der Wilk

    David R. Burt, Carl Edward Rasmussen, and Mark van der Wilk. Convergence of sparse variational inference in gaussian processes regression. Journal of Machine Learning Research, 21 0 (131): 0 1--63, 2020

    work page 2020

Show all 43 references
  1. [9]

    Online variational filtering and parameter learning

    Andrew Campbell, Yuyang Shi, Tom Rainforth, and Arnaud Doucet. Online variational filtering and parameter learning. In 35th Conference on Neural Information Processing Systems, 2021

    2021

  2. [10]

    Carrillo, Franca Hoffmann, Andrew M

    José A. Carrillo, Franca Hoffmann, Andrew M. Stuart, and Urbain Vaes. Statistical accuracy of approximate filtering methods. arXiv:2402.01593, 2024

    2024 arXiv

  3. [11]

    Measuring dependence in the Wasserstein distance for Bayesian nonparametric models

    Marta Catalano, Antonio Lijoi, and Igor Prünster. Measuring dependence in the Wasserstein distance for Bayesian nonparametric models. Annals of Statistics, 49 0 (5): 0 2916--2947, 2021. doi:10.1214/21-AOS2065

    2021 doi

  4. [12]

    Additive smoothing error in backward variational inference for general state-space models

    Mathis Chagneux, Élisabeth Gassiat, Pierre Gloaguen, and Sylvain Le Corff. Additive smoothing error in backward variational inference for general state-space models. Journal of Machine Learning Research, 25 0 (28): 0 1--33, 2024

    2024

  5. [13]

    A survey of convergence results on particle filtering

    Dan Crisan and Arnaud Doucet. A survey of convergence results on particle filtering. IEEE Transactions on Signal Processing, 50 0 (3): 0 736--746, 2002. doi:10.1109/78.984773

    2002 doi

  6. [14]

    Stable approximation schemes for optimal filters

    Dan Crisan, Alberto López-Yela, and Joaquin Miguez. Stable approximation schemes for optimal filters. SIAM/ASA Journal on Uncertainty Quantification, 8 0 (1): 0 483--509, 2020. doi:10.1137/19M1255410

    2020 doi

  7. [15]

    Deep composition of tensor-trains using squared inverse Rosenblatt transport

    Tiangang Cui and Sergey Dolgov. Deep composition of tensor-trains using squared inverse Rosenblatt transport. Foundations of Computational Mathematics, 22 0 (6): 0 1863--1922, 2022. doi:10.1007/s10208-021-09537-5

    1922 doi

  8. [16]

    Memming Park

    Matthew Dowling, Yuan Zhao, and II. Memming Park. Real-time variational method for learning neural trajectory and its dynamics. In 11th International Conference on Learning Representations, 2023

    2023

  9. [17]

    On the performance of particle filters with adaptive number of particles

    Víctor Elvira, Joaquín Miguez, and Petar Djurić. On the performance of particle filters with adaptive number of particles. Statistics and Computing, 31 0 (81), 2021. doi:10.1007/s11222-021-10056-0

    2021 doi

  10. [18]

    Sequential Monte Carlo Methods in Practice , volume 1

    Nando Freitas and Gordon Neil. Sequential Monte Carlo Methods in Practice , volume 1. New York: springer, 2001

    2001

  11. [19]

    Bayesian posterior perturbation analysis with integral probability metrics

    Alfredo Garbuno-Inigo, Tapio Helin, Franca Hoffmann, and Bamdad Hosseini. Bayesian posterior perturbation analysis with integral probability metrics. arXiv:2303.01512, 2023

    2023 arXiv

  12. [20]

    Gibbs and Su Francis Edward

    Alison L. Gibbs and Su Francis Edward. On choosing and bounding probability metrics. Internation Statistical Reivew, 70 0 (3): 0 419--435, 2002. doi:10.2307/1403865

    2002 doi

  13. [21]

    Transport map coupling filter for state-parameter estimation

    Jan Grashorn, Matteo Broggi, Ludovic Chamoin, and Michael Beer. Transport map coupling filter for state-parameter estimation. In Advances in Reliability, Safety and Security ESREL, 2024

    2024

  14. [22]

    Gustafson and Wasserman Larry

    Paul. Gustafson and Wasserman Larry. Local sensitivity diagnostics for Bayesian inference. Annals of Statistics, 23 0 (6): 0 2153--2167, 1995. doi:10.1214/aos/1034713652

    1995

  15. [23]

    Gaussian filters for nonlinear filtering problems

    Kazufumi Ito and Kaiqi Xiong. Gaussian filters for nonlinear filtering problems. IEEE Transactions on Automatic Control, 45 0 (5): 0 910--927, 2000. doi:10.1109/9.855552

    doi

  16. [24]

    On the approximation accuracy of gaussian variational inference

    Anya Katsevich and Philippe Rigollet. On the approximation accuracy of gaussian variational inference. Annals of Statistics, 52 0 (4): 0 1384--1409, 2024. doi:10.1214/24-AOS2393

    2024 doi

  17. [25]

    Data Assimilation: A Mathematical Introduction, volume 62

    Kody Law, Andrew Stuart, and Konstantinos Zygalakis. Data Assimilation: A Mathematical Introduction, volume 62. Springer, Cham, 2015

    2015

  18. [26]

    Stability and uniform approximation of nonlinear filters using the Hilbert metric and application to particle filters

    François Le Gland and Nadia Oudjane. Stability and uniform approximation of nonlinear filters using the Hilbert metric and application to particle filters. Annals of Applied Probability, 14 0 (1): 0 144--187, 2004. doi:10.1214/aoap/1075828050

    2004

  19. [27]

    Large Sample Asymptotics for the Ensemble Kalman Filter

    François Le Gland, Valérie Monbet, and Vu-Duc Tran. Large Sample Asymptotics for the Ensemble Kalman Filter . Diss. INRIA, 2009

    2009

  20. [28]

    Edward E. Leamer. Sets of posterior means with bounded variance prior. Econometrica, 50: 0 725--736, 1982. doi:10.2307/1912610

    1982 doi

  21. [29]

    Probabilistic filter and smoother for variational inference of Bayesian linear dynamical system

    Julian Neri, Roland Badeau, and Philippe Depalle. Probabilistic filter and smoother for variational inference of Bayesian linear dynamical system. In 2020 IEEE International Conference on Acoustics, Speech and Signal Processing, 2020

    2020

  22. [30]

    Ning Ning and Edward L. Ionides. Iterated block particle filter for high-dimensional parameter learning: beating the curse of dimensionality. Journal of Machine Learning Research, 24 0 (82): 0 1--76, 2023

    2023

  23. [31]

    Can local particle filters beat the curse of dimensionality? Annals of Applied Probability, 25 0 (5): 0 2809--2866, 2015

    Patrick Rebesschini and Ramon van Handel. Can local particle filters beat the curse of dimensionality? Annals of Applied Probability, 25 0 (5): 0 2809--2866, 2015. doi:10.1214/14-AAP1061

    2015 doi

  24. [32]

    Wasserstein convergence in Bayesian and frequentist deconvolution models

    Judith Rousseau and Catia Scricciolo. Wasserstein convergence in Bayesian and frequentist deconvolution models. Annals of Statistics, 52 0 (4): 0 1691--1715, 2024. doi:10.1214/24-AOS2413

    2024 doi

  25. [33]

    Posterior ranges of functions of parameters under priors with specified quantiles

    Fabrizio Ruggeri. Posterior ranges of functions of parameters under priors with specified quantiles. Communications in Statistics - Theory and methods, 19: 0 127--144, 1990. doi:10.1080/03610929008830192

    1990 doi

  26. [34]

    Sanz-Alonso, A

    D. Sanz-Alonso, A. Stuart, and A. Taeb. Inverse Problems and Data Assimilation, volume 107. Cambridge University Press, 2023

    2023

  27. [35]

    Measuring local sensitivity in Bayesian inference using a new class of metrics

    Tabassom Sedighi, Amin Hosseinian-Far, and Alireza Daneshkhah. Measuring local sensitivity in Bayesian inference using a new class of metrics. Communications in Statistics - Theory and Methods, 52 0 (11): 0 3581--3597, 2021. doi:10.1080/03610926.2021.1977956

    2021

  28. [36]

    Bounds on posterior expectation for density bounded classes with constant bandwith

    Siva Sivaganesan. Bounds on posterior expectation for density bounded classes with constant bandwith. Journal of Statistical Planning and Inference, 40: 0 331--343, 1994. doi:10.1016/0378-3758(94)90129-5

    1994 doi

  29. [37]

    On the local lipschitz stability of Bayesian inverse problems

    Björn Sprungk. On the local lipschitz stability of Bayesian inverse problems. Inverse Problems, 36 0 (5): 0 055015, 2020. doi:10.1088/1361-6420/ab6f43

    2020 doi

  30. [38]

    Bayesian Filtering and Smoothing

    Simo Säkkä. Bayesian Filtering and Smoothing. Cambridge University Press, 2013

    2013

  31. [39]

    Optimal Transport: Old and New

    Cédric Vallani. Optimal Transport: Old and New. Grundlehren der mathematischen Wissenschaften. Berlin, Germany: Springer, 2009

    2009

  32. [40]

    Factorization-based online variational inference for state-parameter estimation of partially observable nonlinear dynamical systems

    Liliang Wang and Alex Gorodetsky. Factorization-based online variational inference for state-parameter estimation of partially observable nonlinear dynamical systems. In AIAA Scitech 2025 Forum, 2025. doi:10.2514/6.2025-1960

    2025 doi

  33. [41]

    Optimal transport and Wasserstein distance

    Larry Wasserman. Optimal transport and Wasserstein distance. https://www.stat.cmu.edu/ larry/=sml/Opt.pdf, 2019

    2019

  34. [42]

    Convergence rates of variational posterior distributions

    Fengshuo Zhang and Chao Gao. Convergence rates of variational posterior distributions. Annals of Statistics, 48 0 (4): 0 2180--2207, 2020. doi:10.1214/19-AOS1883

    2020 doi

  35. [43]

    Tensor-train methods for sequential state and parameter learning in state-space models

    Yiran Zhao and Tiangang Cui. Tensor-train methods for sequential state and parameter learning in state-space models. Journal of Machine Learning Research, 25 0 (244): 0 1--51, 2024

    2024

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.