Pith. sign in

REVIEW 5 major objections 7 minor 94 references

Partitioning data and structured latent fields inside INLA yields distributed and recursive Bayesian posteriors that stay close to full-data inference while cutting cost.

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

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

T0 review · grok-4.5

2026-07-30 23:15 UTC pith:UFTHCTKP

load-bearing objection Solid INLA-native distributed/recursive pipeline with exact Gaussian data-split identities; latent-field partition log-det surrogates are uncontrolled truncations that the examples do not isolate. the 5 major comments →

arxiv 2607.23396 v1 pith:UFTHCTKP submitted 2026-07-25 stat.ME

Distributed and recursive Bayesian inference for Big Data and complex spatio-temporal models

Mario Figueira , David Conesa , Antonio L\'opez-Qu\'ilez , H{\aa}vard Rue This is my paper
classification stat.ME MSC 62F1562M3065C60
keywords distributed inferencerecursive inferencelatent Gaussian modelsINLAspatio-temporal modelsBig Datafederated learningGMRF partitioning
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Massive spatio-temporal datasets, streaming observations, and privacy rules that block raw-data sharing make ordinary Bayesian updates impractical. This paper shows how to run those updates inside the INLA framework by splitting both the observations and the structured pieces of the latent field (spatial, temporal, or graph components) into smaller blocks, computing local Gaussian approximations, and recombining them. For Gaussian likelihoods the recombination is exact; for non-Gaussian likelihoods it is an approximation whose quality the authors monitor. The same machinery supports recursive (online) updating when new batches arrive and federated analysis when sites cannot exchange raw records. Case studies on a simulated spatio-temporal field and on U.S. PM2.5 monitoring data show that latent-field summaries remain close to the centralized answer even when hyperparameter geometry drifts, giving practitioners a practical route to scalable, privacy-aware Bayesian inference without rewriting their models from scratch.

Core claim

Inside INLA, both the data likelihood and the structured components of the latent Gaussian field can be partitioned so that the product (or recursive update) of the resulting local Gaussian approximations recovers, to usable accuracy, the same latent-field marginals and hyperparameter summaries that full-data INLA would produce, while the smaller sparse matrices sharply reduce memory and factorization cost.

What carries the argument

Block-independent and block-correlated partitions of the latent-field precision matrix (Eqs. 42–48), together with the product-of-Gaussians recombination for the conditional latent posterior (Eq. 13) and the recursive precision/mean updates (Eqs. 35–38).

Load-bearing premise

For non-Gaussian likelihoods the product of the local Gaussian approximations stays accurate enough that hyperparameter geometry and latent-field marginals do not degrade materially when the blocks are recombined.

What would settle it

On a non-Gaussian spatio-temporal model whose full-data INLA solution is known, increase the number or heterogeneity of partitions until the recombined hyperparameter CCD or the latent-field credible intervals diverge beyond a pre-chosen tolerance from the centralized answer.

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

If this is right

  • Large spatio-temporal models whose precision matrices no longer fit in memory can be analyzed by automatic graph partitioning without redesigning the likelihood.
  • Streaming environmental or epidemiological series can be updated recursively by feeding each new batch into the previous latent posterior as prior.
  • Privacy-sensitive consortia can exchange only local posterior summaries (means, precisions, marginal likelihoods) and still obtain a joint INLA analysis.
  • Meta-analysis of published studies becomes a special case of the same recursive product of approximate posteriors.

Where Pith is reading between the lines

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

  • The same partition-and-recombine algebra could be applied to other sparse latent-Gaussian engines beyond INLA, provided they expose the conditional precision and mode.
  • When hyperparameter posteriors already differ across blocks (as seen in the PM2.5 example), a hierarchical hyperprior that itself varies by partition may reduce the observed CCD drift.
  • Federated settings with non-overlapping latent structures (different covariates or meshes) are a direct extension once the mean and precision are expanded to a common super-field.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 7 minor

Summary. The manuscript develops a framework for distributed, recursive, and federated Bayesian inference for latent Gaussian models built on INLA/R-INLA. For data partitioning it derives exact identities for the conditional latent-field posterior under Gaussian likelihoods and conditional independence (Eqs. 9–11, 30–31), and Gaussian-approximation combinations for non-Gaussian likelihoods (Eq. 13, recursive updates Eqs. 35–38). It then constructs distributed approximations to the hyperparameter posterior (Eqs. 16–21) and several strategies for hyperparameter marginals (interpolation, asymmetric Gaussian, Laplace, integration-free, consensus MC). Section 5 partitions the latent field itself via block-independent (Eqs. 43–45) and block-correlated/halo (Eqs. 46–48) schemes with log-determinant surrogates. Two examples are given: a Gaussian simulated spatio-temporal case with 2 partitions, and a 873,762-observation PM2.5 Gamma model with space–time interaction. The simulation matches full-data inference well; the PM2.5 case recovers latent fields but shows substantial divergence in the CCD and hyperparameter marginals (Figs. 10, 13), attributed to non-Gaussianity and hyperparameter heterogeneity.

Significance. If the approximation layers hold up, this is a practically significant contribution: it makes INLA viable for massive spatio-temporal datasets, streaming updates, and privacy-constrained (federated) settings within the widely used R-INLA stack. Explicit strengths: (i) the data-partitioning and recursive identities (Eqs. 9–11, 30–31) are exact and parameter-free for Gaussian likelihoods under conditional independence; (ii) the authors are candid about non-Gaussian error accumulation (Eq. 13) and recursive non-invariance under partition reordering (§4.1); (iii) the lifted-space determinant identity (Eq. 48) makes the latent-field partitioning surrogate falsifiable, since the neglected term is computable; (iv) a real 873k-observation PM2.5 case study is attempted. The main limitation is that the two ingredients most central to the headline claim — the determinant surrogates and the complexity reduction — are asserted rather than measured.

major comments (5)
  1. [§5, Eqs. (43)–(48)] §5.1–5.2, Eqs. (43)–(48): the latent-field partitioning rests on two uncontrolled log-determinant surrogates: truncation of the expansion in Eq. (45) after noting tr(D^{-1}R)=0, and dropping the interface Schur-complement term log|B Q~^{-1} B^T| from the exact identity in Eq. (48). These determinants enter the hyperparameter posterior via Eq. (14), so any θ-dependent error tilts the hyperparameter geometry, the CCD support points (Eq. 21), and everything downstream. No bound, no scaling analysis in partition count or interface size, and no empirical isolation is given. This is testable: the neglected term in Eq. (48) is computable at interface dimension; please report its magnitude relative to log|Q| across θ, or run the block-correlated scheme with exact determinants to isolate the surrogate's contribution. As stated, 'preserving accuracy' for latent-field partitioning is unestablished
  2. [§6, Figs. 3, 10, 13] §6.1 and §6.2: the empirical section cannot currently support the partitioning claims. §6.1 uses only 2 temporal partitions with a minimal interface, precisely the regime where the neglected Schur term is small; §6.2 shows substantial divergence in the CCD and hyperparameter marginals (Figs. 10, 13), which the text attributes to non-Gaussianity and hyperparameter heterogeneity — but the determinant surrogate is an unexamined confound in that attribution. Please add a study varying partition count/interface size (e.g., 2, 4, 8 temporal cuts) in the Gaussian simulation where the rest of the machinery is exact, so the surrogate's error can be isolated, and disentangle the two error sources in §6.2.
  3. [§3.3, Eqs. (18)–(21)] Eqs. (18)–(21): the global Gaussian approximation sums per-partition Hessians H_{θ*_i} evaluated at partition-specific modes θ*_i (Eq. 19). A Hessian of the global log-posterior evaluated at the global mode is not the sum of Hessians at differing local modes unless the modes coincide; the product-of-Gaussians combination in Eqs. (20)–(21) is a consensus-style heuristic, not an identity. This is load-bearing because the CCD integration design (and hence the marginals in Eq. 22) is built from it. At minimum, state the regularity condition (proximity of the θ*_i) under which this is controlled, quantify the error when it fails (Fig. 10 suggests §6.2 is such a case), and consider evaluating all partition Hessians at a common reference point.
  4. [Abstract, §6, §7] Abstract and §7: the central claim 'reducing computational complexity while preserving accuracy' is not quantified on either axis. No runtimes, memory footprints, or factorization costs are reported anywhere in §6, despite the complexity reduction being the paper's stated motivation; and §6.2 itself exhibits clear accuracy loss in the hyperparameter posterior (Figs. 10, 13). Please (i) report timing/memory comparisons against full-data R-INLA for both examples, and (ii) temper the abstract/conclusions to match the demonstrated scope: latent-field marginals recover well, hyperparameter geometry does not in the non-Gaussian case.
  5. [§3.4, Eq. (29)] Eq. (29): consensus Monte Carlo with weights w_i = π̃(y_i) (per-partition marginal likelihoods) is nonstandard — the Scott et al. (2016) algorithm uses precision (inverse-covariance) weights, which is what makes it exact for Gaussian subposteriors. Marginal-likelihood weighting lacks justification here and can be dominated by partition size. Please either justify this choice or use standard precision weighting, and clarify how Eq. (29) interacts with the asymmetric-Gaussian subposteriors from the integration-free algorithm.
minor comments (7)
  1. [§6] Broken cross-references: §6 first paragraph cites 'Section ??' and §6.2 cites 'Algorithm ??' for the partitioning procedure; the Supplementary Material reference target must be restored.
  2. [§3.1, Eq. (12)] Eq. (12): the canonical-form notation is unclear and the line 'The mean is computed as b_i = w_i Q_i^{-1}(θ)µ_i' appears dimensionally inconsistent — b_i should be the linear coefficient vector (canonical potential), not a scaled mean. Please rewrite Eq. (12) with explicit canonical parameters and check the sign of the linear term.
  3. [§2–3] Notation: n denotes both the number of observations (Eq. 1) and the number of partitions (Eq. 6 onward); |y| in the weights below Eq. (10) is undefined; the asymmetric-Gaussian variances in Eqs. (24) and (26) are written ambiguously ((σ^-_j)^{-2} z^2 vs (z/σ)^2) and lack normalization constants.
  4. [Throughout] Typos and garbled text: 'previosly' (§3.2), 'hyperparamters' (§3.2), 'block-independeent', 'propoerties' (§5), 'Dendogram', 'matriz', 'highlited' (Fig. 1 caption), 'could themselves varying' (§7); the sentence defining the simulation model in §6.1 ('βcat us is a spatial component...') is broken.
  5. [§6, Figs. 3–13] Figures: Figs. 3 and 10 are difficult to read at print size and captions are sparse (no indication of which hyperparameter axis is which in Fig. 3). Quantitative accuracy metrics (e.g., distances between marginal posteriors, coverage comparisons) would strengthen §6 beyond visual overlay.
  6. [§5.2] Positioning: the block-correlated scheme is essentially a Schwarz/domain-decomposition approximation for GMRF determinants; there is existing literature on parallel and domain-decomposed inference for latent Gaussian models that should be cited and contrasted, beyond the Toselli–Widlund reference.
  7. [§4.3] §4.3 honestly acknowledges that integration support points cannot in general be reused across recursive steps; a brief summary of what is actually reused and the resulting per-step cost would help readers assess the practical value of the recursive scheme relative to a fresh R-INLA run.

Circularity Check

0 steps flagged

No significant circularity: distributed/recursive identities are algebraic rearrangements of the INLA/GMRF posterior, and latent-field log-det surrogates are explicit truncations benchmarked against full-data INLA.

full rationale

The paper’s load-bearing steps are (i) rewriting the conditional latent-field posterior under conditional independence (Eqs. 6–11, 30–31), which for Gaussian likelihoods is an exact algebraic identity, not a fit or a definitional loop; (ii) product-of-Gaussians and recursive mode updates for non-Gaussian likelihoods (Eqs. 12–13, 34–38), presented and used as approximations with acknowledged error accumulation; and (iii) block-independent / block-correlated latent-field partitions whose log-determinant surrogates (Eqs. 43–48) are derived as zeroth-order truncations of known matrix expansions and justified by external domain-decomposition references (Toselli & Widlund 2005; Zhang 2005), not by self-citation uniqueness theorems. Empirical claims are checked against centralized full-data INLA on the same models (§6), an external benchmark relative to the distributed procedure. Self-citations (Figueira et al. 2025a,b; Van Niekerk & Rue; Rue et al.) supply INLA background and prior sequential-consensus tooling; they do not force the distributed identities or the accuracy claim by construction. Correctness concerns about uncontrolled determinant truncations or non-Gaussian hyperparameter divergence are validity issues, not circularity. No fitted-input-as-prediction, self-definitional, or uniqueness-import patterns appear.

Axiom & Free-Parameter Ledger

3 free parameters · 4 axioms · 2 invented entities

The framework rests on standard LGM/INLA assumptions plus several modeling choices about how to split priors, approximate non-Gaussian conditionals, and surrogate log-determinants of partitioned precision matrices. No new physical entities are postulated; the ‘invented’ pieces are algorithmic partition schemes.

free parameters (3)
  • prior weights w_i on partitions = user-chosen; examples use equal or size-based splits
    Chosen by the analyst (uniform, size-based, or other schemes with sum w_i=1); they alter each partition posterior and the combined hyperparameter Gaussian.
  • number and cut locations of latent-field partitions = two temporal partitions in examples; PM2.5 uses 14+1 temporal nodes
    Graph cuts (temporal precision blocks, spatial meshes) are design choices that control both cost and approximation quality; not derived from data.
  • CCD / integration support design
    Support points and weights Δ_k for hyperparameter integration are constructed from the approximate modal Gaussian; design choices affect marginal accuracy.
axioms (4)
  • domain assumption Observations are conditionally independent given the latent field and hyperparameters (Eq. 1), allowing likelihood factorization across partitions.
    Standard LGM assumption; load-bearing for all distributed product formulas in §3.1.
  • domain assumption INLA’s Gaussian / low-rank VB approximation to the conditional latent-field posterior is adequate at each partition or recursive step.
    Inherited from Rue et al. (2009) and Van Niekerk & Rue (2024); errors accumulate in non-Gaussian distributed/recursive products.
  • ad hoc to paper Block-independent log|Q|≈∑log|Q_i^{(0)}| (zeroth-order perturbation) or block-correlated ∑log|Q_i| (neglecting interface Schur complement) are acceptable surrogates for the global log-determinant.
    Justified via matrix perturbation / domain-decomposition arguments in §5 but not error-bounded for the models used.
  • standard math Sparse GMRF precision structure and standard INLA hyperparameter exploration (CCD/grid, numerical-integration-free marginals) remain valid after recombination.
    Background computational linear algebra and INLA practice.
invented entities (2)
  • Block-independent latent-field partition scheme no independent evidence
    purpose: Sever cross-boundary edges so precision becomes block-diagonal and determinants factor, cutting factorization cost.
    Methodological construct; alters conditional independence of the original GMRF by design.
  • Block-correlated (halo / extended) latent-field partition scheme no independent evidence
    purpose: Keep boundary correlations via overlapping re-scaled local precisions that sum to global Q, while still approximating log|Q| by ∑log|Q_i|.
    Domain-decomposition-inspired construction specific to this paper’s distributed INLA pipeline.

pith-pipeline@v1.2.0-grok45-kimik3 · 28892 in / 3195 out tokens · 62035 ms · 2026-07-30T23:15:41.338748+00:00 · methodology

0 comments
read the original abstract

The rapid growth of massive and complex datasets in fields such as econometrics, environmental sciences, risk management, and public policy has reshaped statistical modeling while introducing significant computational and methodological challenges. These challenges arise not only from data scale and model complexity, but also from the sequential or streaming nature of modern applications and from data-privacy constraints that prevent sharing raw data and thus limit joint analysis. To address these challenges, we introduce a novel and comprehensive Bayesian framework for distributed and recursive inference, grounded in the Integrated Nested Laplace Approximations (INLA) methodology and implemented using the R-INLA software. Our contributions include the partitioning both data and structured model components, reducing computational complexity while preserving accuracy relative to centralized full-data inference. We demonstrate the effectiveness of the proposed framework through case studies that highlight its applicability in large-scale, streaming, and privacy-sensitive settings. By integrating distributed, federated, and recursive paradigms, this work offers scalable, adaptive, and generalizable tools for modern Bayesian inference.

Figures

Figures reproduced from arXiv: 2607.23396 by Antonio L\'opez-Qu\'ilez, David Conesa, H{\aa}vard Rue, Mario Figueira.

Figure 1
Figure 1. Figure 1: Example with two graphs—a dendrogram and a grid—with the diagonal blocks [PITH_FULL_IMAGE:figures/full_fig_p023_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Plots of the temporal and spatial effects, together with the spatial patterns of [PITH_FULL_IMAGE:figures/full_fig_p030_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: CCD for the full data inference (red) and from the combination of the dis [PITH_FULL_IMAGE:figures/full_fig_p031_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Posterior distributions of the intercept and covariate parameters from the [PITH_FULL_IMAGE:figures/full_fig_p031_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Posterior distributions of the categorical variable parameters from the full [PITH_FULL_IMAGE:figures/full_fig_p032_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Posterior mean and 95% credible intervals of the temporal effect under the full-data (red), distributed (blue), and recursive (green) analyses [PITH_FULL_IMAGE:figures/full_fig_p032_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of the posterior mean, standard deviation, and 2.5th and 97.5th [PITH_FULL_IMAGE:figures/full_fig_p033_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Posterior distributions of the hyperparameters on the internal scale, obtained [PITH_FULL_IMAGE:figures/full_fig_p034_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: PM [PITH_FULL_IMAGE:figures/full_fig_p035_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: CCD plot of the integration scheme in the hyperparameter space for full [PITH_FULL_IMAGE:figures/full_fig_p036_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Comparison of posterior mean and standard deviation of the spatial effect [PITH_FULL_IMAGE:figures/full_fig_p037_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Posterior mean and 95% credible intervals of the temporal effect under the full-data (red) and distributed (blue) analyses [PITH_FULL_IMAGE:figures/full_fig_p038_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Posterior distributions of the hyperparameters on the internal scale and the [PITH_FULL_IMAGE:figures/full_fig_p038_13.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

94 extracted references · 24 canonical work pages · 2 internal anchors

  1. [1]

    SIAM Journal on Discrete Mathematics , author =

    Planar. SIAM Journal on Discrete Mathematics , author =. 1994 , pages =. doi:10.1137/S0895480191198768 , language =

    doi:10.1137/s0895480191198768 1994

  2. [2]

    International Data Privacy Law , author =

    Practical approaches to big data privacy over time , volume =. International Data Privacy Law , author =. 2018 , pages =. doi:10.1093/idpl/ipx027 , language =

    work page doi:10.1093/idpl/ipx027 2018

  3. [3]

    SIAM Journal on Matrix Analysis and Applications , author =

    An. SIAM Journal on Matrix Analysis and Applications , author =. 1996 , note =. doi:10.1137/s0895479894278952 , language =

    doi:10.1137/s0895479894278952 1996

  4. [4]

    Spatial modeling with R-INLA: A review , journal =

    Bakka, Haakon and Rue, H. Spatial modeling with R-INLA: A review , journal =. doi:https://doi.org/10.1002/wics.1443 , year =

    doi:10.1002/wics.1443

  5. [5]

    Acta Numerica , author =

    Numerical solution of saddle point problems , volume =. Acta Numerica , author =. 2005 , pages =. doi:10.1017/S0962492904000212 , language =

    doi:10.1017/s0962492904000212 2005

  6. [6]

    , year =

    Bishop, Christopher M. , year =. Pattern recognition and machine learning , isbn =

  7. [7]

    Journal of Statistical Mechanics: Theory and Experiment , author =

    Fast unfolding of communities in large networks , volume =. Journal of Statistical Mechanics: Theory and Experiment , author =. 2008 , pages =. doi:10.1088/1742-5468/2008/10/P10008 , number =

    doi:10.1088/1742-5468/2008/10/p10008 2008

  8. [8]

    Bernstein - von

    Bochkina, Natalia , year =. Bernstein - von. doi:10.1007/978-3-031-30114-8_10 , pages =

    work page doi:10.1007/978-3-031-30114-8_10

  9. [9]

    Spectral clustering and biclustering: learning large graphs and contingency tables , isbn =

    Bolla, Marianna , year =. Spectral clustering and biclustering: learning large graphs and contingency tables , isbn =

  10. [10]

    Streaming

    Broderick, Tamara and Boyd, Nicholas and Wibisono, Andre and Wilson, Ashia C and Jordan, Michael I , editor =. Streaming. Advances in

  11. [11]

    Computational Statistics & Data Analysis , author =

    A novel method of marginalisation using low discrepancy sequences for integrated nested. Computational Statistics & Data Analysis , author =. 2021 , pages =. doi:10.1016/j.csda.2020.107147 , language =

    arXiv 2021

  12. [12]

    Statistical Science , author =

    Particle. Statistical Science , author =. doi:10.1214/10-STS325 , number =

    doi:10.1214/10-sts325

  13. [13]

    The Annals of Statistics , author =

    On the. The Annals of Statistics , author =. doi:10.1214/14-AOS1246 , number =

    work page doi:10.1214/14-aos1246

  14. [14]

    , year =

    Chung, Fan Rong K. , year =. Spectral graph theory , isbn =

  15. [15]

    Diggle and Paula Moraga and Barry Rowlingson and Benjamin M

    Peter J. Diggle and Paula Moraga and Barry Rowlingson and Benjamin M. Taylor , title =. Statistical Science , number =. 2013 , doi =

    2013

  16. [16]

    Duff, I. S. , collaborator =. Direct

  17. [17]

    Engineering Applications of Artificial Intelligence , author =

    A comprehensive survey of clustering algorithms:. Engineering Applications of Artificial Intelligence , author =. 2022 , note =. doi:10.1016/j.engappai.2022.104743 , language =

    arXiv 2022

  18. [18]

    BioScience , author =

    Situating. BioScience , author =. 2018 , pages =. doi:10.1093/biosci/biy068 , language =

    work page doi:10.1093/biosci/biy068 2018

  19. [19]

    Foundations of Data Science , author =

    Smart. Foundations of Data Science , author =. 2022 , pages =. doi:10.3934/fods.2021037 , number =

    work page doi:10.3934/fods.2021037 2022

  20. [20]

    2022 , journal =

    Approximate Bayesian Inference for the Interaction Types 1, 2, 3 and 4 with Application in Disease Mapping , author=. 2022 , journal =. 2206.09287 , archivePrefix=

    Pith/arXiv arXiv 2022

  21. [21]

    Czechoslovak Mathematical Journal , author =

    Algebraic connectivity of graphs , volume =. Czechoslovak Mathematical Journal , author =. 1973 , pages =. doi:10.21136/CMJ.1973.101168 , language =

    arXiv 1973

  22. [22]

    Environmental and Ecological Statistics , author =

    A computationally efficient procedure for combining ecological datasets by means of sequential consensus inference , volume =. Environmental and Ecological Statistics , author =. 2025 , pages =. doi:10.1007/s10651-025-00653-x , language =

    work page doi:10.1007/s10651-025-00653-x 2025

  23. [23]

    2025 , eprint=

    Integrating Expert Knowledge and Recursive Bayesian Inference: A Framework for Spatial and Spatio-Temporal Data Challenges , author=. 2025 , eprint=

    2025

  24. [24]

    Physics Reports , author =

    Community detection in graphs , volume =. Physics Reports , author =. 2010 , note =. doi:10.1016/j.physrep.2009.11.002 , number =

    doi:10.1016/j.physrep.2009.11.002 2010

  25. [25]

    Physics Reports , author =

    Community detection in networks:. Physics Reports , author =. 2016 , pages =. doi:10.1016/j.physrep.2016.09.002 , language =

    doi:10.1016/j.physrep.2016.09.002 2016

  26. [26]

    The Annals of Statistics , author =

    Wald. The Annals of Statistics , author =. doi:10.1214/aos/1017938917 , number =

    arXiv

  27. [27]

    Statistical Theory and Related Fields , author =

    A review of distributed statistical inference , volume =. Statistical Theory and Related Fields , author =. 2022 , pages =. doi:10.1080/24754269.2021.1974158 , language =

    arXiv 2022

  28. [28]

    SIAM Journal on Numerical Analysis , author =

    Nested. SIAM Journal on Numerical Analysis , author =. 1973 , note =. doi:10.1137/0710032 , language =

    doi:10.1137/0710032 1973

  29. [29]

    SIAM Journal on Numerical Analysis , author =

    An. SIAM Journal on Numerical Analysis , author =. 1976 , note =

    1976

  30. [30]

    Numerische Mathematik , author =

    The analysis of a nested dissection algorithm , volume =. Numerische Mathematik , author =. 1986 , note =. doi:10.1007/bf01396660 , language =

    work page doi:10.1007/bf01396660 1986

  31. [31]

    Journal of Computer and System Sciences , author =

    Planar. Journal of Computer and System Sciences , author =. 1995 , pages =. doi:10.1006/jcss.1995.1076 , language =

    arXiv 1995

  32. [32]

    Geometric approximation algorithms , isbn =

    Har-Peled, Sariel , year =. Geometric approximation algorithms , isbn =

  33. [33]

    Journal of Visual Communication and Image Representation , author =

    Graph morphology , volume =. Journal of Visual Communication and Image Representation , author =. 1992 , pages =. doi:10.1016/1047-3203(92)90028-R , language =

    work page doi:10.1016/1047-3203(92)90028-r 1992

  34. [34]

    Stochastic variational inference , journal =

    Hoffman, Matthew D and Blei, David M and Wang, Chong and Paisley, John , year =. Stochastic variational inference , journal =

  35. [35]

    The American Statistician , author =

    Making. The American Statistician , author =. 2021 , pages =. doi:10.1080/00031305.2019.1665584 , number =

    Pith/arXiv arXiv 2021

  36. [36]

    Sampling for

    Huang, Zaijing and Gelman, Andrew , year =. Sampling for

  37. [37]

    Illian and Sigrunn H

    Janine B. Illian and Sigrunn H. S. The Annals of Applied Statistics , number =. 2012 , doi =

    2012

  38. [38]

    Pattern Recognition Letters , author =

    Data clustering: 50 years beyond. Pattern Recognition Letters , author =. 2010 , pages =. doi:10.1016/j.patrec.2009.09.011 , language =

    doi:10.1016/j.patrec.2009.09.011 2010

  39. [39]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , author =

    Normalized cuts and image segmentation , volume =. IEEE Transactions on Pattern Analysis and Machine Intelligence , author =. 2000 , pages =. doi:10.1109/34.868688 , number =

    doi:10.1109/34.868688 2000

  40. [40]

    Psychometrika , author =

    Hierarchical. Psychometrika , author =. 1967 , note =. doi:10.1007/bf02289588 , language =

    doi:10.1007/bf02289588 1967

  41. [41]

    Karypis, George and Kumar, Vipin , month = sep, year =

  42. [42]

    , month = mar, year =

    Kaufman, Leonard and Rousseeuw, Peter J. , month = mar, year =. Finding. doi:10.1002/9780470316801 , note =

    doi:10.1002/9780470316801

  43. [43]

    Entropy , author =

    On. Entropy , author =. 2023 , pages =. doi:10.3390/e25060884 , language =

    work page doi:10.3390/e25060884 2023

  44. [44]

    Statistics in Medicine , doi =

    Knorr-Held, Leonhard , title =. Statistics in Medicine , doi =

  45. [45]

    Misspecified Bernstein-Von Mises theorem for hierarchical models

    Koers, Geerten and Szabó, Botond and Vaart, Aad van der , month = jun, year =. Misspecified. doi:10.48550/arXiv.2308.07803 , urldate =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2308.07803

  46. [46]

    , year =

    Kullback, S. , year =. Information

  47. [47]

    Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume =

    Lindgren, Finn and Rue, H. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume =. doi:https://doi.org/10.1111/j.1467-9868.2011.00777.x , year =

    arXiv 2011

  48. [48]

    Journal of Statistical Software , author =

    Bayesian. Journal of Statistical Software , author =. 2015 , pages =. doi:10.18637/jss.v063.i19 , number =

    doi:10.18637/jss.v063.i19 2015

  49. [49]

    SIAM Journal on Computing , author =

    Applications of a. SIAM Journal on Computing , author =. 1980 , pages =. doi:10.1137/0209046 , language =

    work page doi:10.1137/0209046 1980

  50. [50]

    SIAM Journal on Numerical Analysis , author =

    Comparative. SIAM Journal on Numerical Analysis , author =. 1976 , note =

    1976

  51. [51]

    ACM Transactions on Mathematical Software , author =

    Modification of the minimum-degree algorithm by multiple elimination , volume =. ACM Transactions on Mathematical Software , author =. 1985 , note =. doi:10.1145/214392.214398 , language =

    arXiv 1985

  52. [52]

    SORT , author =

    Leave-group-out cross-validation for latent gaussian models , volume =. SORT , author =. 2025 , pages =. doi:10.57645/20.8080.02.25 , language =

    work page doi:10.57645/20.8080.02.25 2025

  53. [53]

    Journal of the Royal Statistical Society Series C: Applied Statistics , author =

    Fully. Journal of the Royal Statistical Society Series C: Applied Statistics , author =. 2013 , pages =. doi:10.1111/rssc.12007 , number =

    work page doi:10.1111/rssc.12007 2013

  54. [54]

    Computers & Structures , author =

    Algorithms for reducing the bandwidth and profile of a sparse matrix , volume =. Computers & Structures , author =. 1992 , pages =. doi:10.1016/0045-7949(92)90386-E , language =

    work page doi:10.1016/0045-7949(92)90386-e 1992

  55. [55]

    Communication-

    McMahan, Brendan and Moore, Eider and Ramage, Daniel and Hampson, Seth and Arcas, Blaise Aguera y , editor =. Communication-. Proceedings of the 20th. 2017 , pages =

    2017

  56. [56]

    Journal of Statistical Planning and Inference , author =

    Approximate. Journal of Statistical Planning and Inference , author =. 2007 , pages =. doi:10.1016/j.jspi.2006.07.016 , language =

    work page doi:10.1016/j.jspi.2006.07.016 2007

  57. [57]

    and Simpson, Daniel and Lindgren, Finn and Rue, H

    Martins, Thiago G. and Simpson, Daniel and Lindgren, Finn and Rue, H. Bayesian computing with. 2013 , pages =. doi:10.1016/j.csda.2013.04.014 , journal =

    work page doi:10.1016/j.csda.2013.04.014 2013

  58. [58]

    Parallel sparse

    Monien, Burkhard and Schulze, Jürgen , isbn =. Parallel sparse. Lecture. 1997 , doi =

    1997

  59. [59]

    , year =

    Myers, Raymond H. , year =. Response

  60. [60]

    WIREs Data Mining and Knowledge Discovery , author =

    Algorithms for hierarchical clustering: an overview , volume =. WIREs Data Mining and Knowledge Discovery , author =. 2012 , note =. doi:10.1002/widm.53 , number =

    doi:10.1002/widm.53 2012

  61. [61]

    Combinatorial scientific computing , isbn =

    Naumann, Uwe and Schenk, Olaf , year =. Combinatorial scientific computing , isbn =

  62. [62]

    A new avenue for

    Van Niekerk, Janet and Krainski, Elias and Rustand, Denis and Rue, H. A new avenue for. 2023 , pages =. doi:10.1016/j.csda.2023.107692 , journal =

    arXiv 2023

  63. [63]

    Journal of Machine Learning Research , year =

    Van Niekerk, Janet and Rue, H. Journal of Machine Learning Research , year =

  64. [64]

    Big problems in spatio-temporal disease mapping:

    Orozco-Acosta, Erick and Adin, Aritz and Ugarte, María Dolores , month = apr, year =. Big problems in spatio-temporal disease mapping:. doi:10.1016/j.cmpb.2023.107403 , journal =

    arXiv 2023

  65. [65]

    A first course in random matrix theory: for physicists, engineers and data scientists , isbn =

    Potters, Marc and Bouchaud, Jean-Philippe , year =. A first course in random matrix theory: for physicists, engineers and data scientists , isbn =. doi:10.1017/9781108768900 , language =

    doi:10.1017/9781108768900

  66. [66]

    Journal of Computational and Graphical Statistics , author =

    Global. Journal of Computational and Graphical Statistics , author =. 2021 , pages =. doi:10.1080/10618600.2020.1811105 , number =

    arXiv 2021

  67. [67]

    Online structured laplace approximations for overcoming catastrophic forgetting , publisher =

    Ritter, Hippolyt and Botev, Aleksandar and Barber, David , year =. Online structured laplace approximations for overcoming catastrophic forgetting , publisher =

  68. [68]

    Gaussian

    Rue, H. Gaussian. 2005 , doi =

    2005

  69. [69]

    Journal of the Royal Statistical Society

    Approximate. Journal of the Royal Statistical Society. Series B: Statistical Methodology , author =. doi:10.1111/j.1467-9868.2008.00700.x , language =

    arXiv 2008

  70. [70]

    and Boley, Daniel L

    Savaresi, Sergio M. and Boley, Daniel L. , month = apr, year =. On the performance of bisecting. Proceedings of the 2001. doi:10.1137/1.9781611972719.5 , urldate =

    work page doi:10.1137/1.9781611972719.5 2001

  71. [71]

    Bayesian

    Särkkä, Simo , month = sep, year =. Bayesian

  72. [72]

    IEEE Signal Processing Magazine , author =

    Signal. IEEE Signal Processing Magazine , author =. 2013 , pages =. doi:10.1109/MSP.2013.2259911 , number =

    arXiv 2013

  73. [73]

    Frontiers in Environmental Science , author =

    The. Frontiers in Environmental Science , author =. 2021 , pages =. doi:10.3389/fenvs.2021.619092 , urldate =

    arXiv 2021

  74. [74]

    International Journal of Management Science and Engineering Management , author =

    Bayes and big data: the consensus. International Journal of Management Science and Engineering Management , author =. 2016 , pages =. doi:10.1080/17509653.2016.1142191 , number =

    arXiv 2016

  75. [75]

    and Illian, J

    Simpson, D. and Illian, J. B. and Lindgren, F. and Sørbye, S. H. and Rue, H. , title =. Biometrika , volume =. 2016 , month =

    2016

  76. [76]

    Smith, Adam and Thakurta, Abhradeep and Upadhyay, Jalaj , month = may, year =. Is. 2017. doi:10.1109/SP.2017.35 , urldate =

    work page doi:10.1109/sp.2017.35 2017

  77. [77]

    Journal of the Royal Statistical Society Series C: Applied Statistics , author =

    Careful. Journal of the Royal Statistical Society Series C: Applied Statistics , author =. 2019 , pages =. doi:10.1111/rssc.12321 , language =

    work page doi:10.1111/rssc.12321 2019

  78. [78]

    Bernstein - von

    Spokoiny, Vladimir , month = apr, year =. Bernstein - von. doi:10.48550/arXiv.1302.3430 , publisher =

    Pith/arXiv arXiv doi:10.48550/arxiv.1302.3430

  79. [79]

    Journal of Machine Learning Research , author =

    Scalable. Journal of Machine Learning Research , author =. 2018 , pages =

    2018

  80. [80]

    , year =

    Toselli, Andrea and Widlund, Olof B. , year =. Domain decomposition methods--algorithms and theory , isbn =

Showing first 80 references.