Pith.

Computation · arXiv:2611.17008 v1 · submitted 12 Nov 2026 · 278 references

Sticky Manifold Piecewise Deterministic Monte Carlo for Bayesian Model Averaging

Hirofumi Shiba, Luke Hardcastle, Kengo Kamatani

The paper says

“Assigning positive prior probability to lower-dimensional submodels enriches the range of uncertainty that Bayesian models can represent. The resulting posterior distributions combine continuous parameter uncertainty with discrete model uncertainty. Efficient movement between these submodels, however, can be difficult to design.”

more of the abstract

We propose the sticky manifold construction as a principled, modular framework for combining piecewise deterministic Markov processes (PDMPs) across different submodels. The integration is rejection-free, requires no additional likelihood evaluations, and retains the non-reversible and ballistic nature of the component PDMP dynamics. We prove that the resulting process leaves the intended mixed-dimensional target invariant. We also develop a manifold PDMP algorithm based on coordinate charts to provide the within-submodel dynamics and a blow-up procedure that resolves singular sets into codimension-one sticky boundaries. This geometric construction makes the remaining design choices explicit; component dynamics and unsticking kernels can be chosen freely, subject to easily verifiable velocity compatibility conditions. Finally, we demonstrate the sticky manifold algorithms in Bayesian model averaging examples, propagating uncertainty over pure versus mixed states in quantum state tomography, rank in Bayesian principal component analysis, and parametric families in survival extrapolation.

from the abstract

Pith says

Machine review · deepseek-v4-flash · 13 Nov 2026 · what was read was not recorded · not checked by a person

It can be wrong. Challenge this review or reply to it.

Holds up, with conditions. A principled, modular framework for trans-dimensional PDMP that generalises sticky PDMPs to curved submanifolds; the core invariance proof is solid, but the velocity compatibility condition is a genuine practical bottleneck that the paper acknowledges but does not resolve. This line is wrong

4 major objections · 7 minor

How often objections from reviews like this one are right: not yet measured

Are you an author? What you can do.

Not checked: figures, code, what the 278 references say.

The paper says, at Section 3.1, Eq. (3.2)

The review did not quote this passage.

Pith says: Objection 1 of 4 · major

The velocity compatibility condition (3.2) is central to the invariance proof but is only verified for Gaussian and unit-velocity cases. The paper acknowledges this gap in Section 3.5 but does not provide a general construction for Q_stick and Q_unstick when the ambient and manifold velocity distributions differ, e.g., under non-isometric charts. This limits the claimed generality of the framework. The authors should either provide a general method to construct such kernels or clarify the scope of applicability.

The paper says, at Section 3.2, Assumption 3.4

The review did not quote this passage.

Pith says: Objection 2 of 4 · major

The assumption that M is of codimension one, orientable, and closed in R^d is restrictive. While the blow-up construction in Section 3.5 and Remark 3.12 extend to nested strata and overlapping manifolds, the paper does not fully address higher codimensions or non-orientable manifolds. The authors should discuss the practical implications of these restrictions and whether they can be relaxed.

The paper says, at Section 4.2, Table 1

The review did not quote this passage.

A classifier read this sentence of the paper as acknowledging the concern (probability 0.69):

“The smaller sphere is reached rarely and produces longer residence episodes. marginal model-index process can be limited by the longer sojourn, even if the mixture weights of the target distribution are equalω= 1−ω= 1/2.”

Pith says: Objection 3 of 4 · major

The RMSE for R=1 and d=10,20,50 shows that the sampler fails to make any trans-dimensional jumps, resulting in RMSE=0.5 (the prior weight). This is a significant limitation: when the submanifold is in a low-density region, the process may never visit it, leading to reducible dynamics. The authors acknowledge this but do not propose a remedy. This issue should be discussed more prominently as a practical limitation.

The paper says, at Section 3.5, Example 3.13

The review did not quote this passage.

A classifier read this sentence of the paper as acknowledging the concern (probability 0.83):

“If, instead, a prior does not diverge sufficiently rapidly in the original space, a na¨ ıve application of PDMP dynamics on the blow-up coordinates will have zero probability of hitting the front face, producing reducible dynamics.”

Pith says: Objection 4 of 4 · major

The blow-up construction for the cone apex is elegant, but the paper does not verify the velocity compatibility condition for the polar coordinate chart. The Jacobian factor in Remark 3.14 suggests that the prior must diverge sufficiently rapidly for the process to hit the apex. This is a non-trivial condition that may not hold in practice. The authors should provide a concrete example where this condition is satisfied or discuss how to handle cases where it is not.

7 minor objections

The paper says, at Abstract

Pith says: minor objection 1 of 7

The abstract states 'easily verifiable velocity compatibility conditions', but the paper does not provide a general verification procedure. Consider rephrasing to reflect the current state of the examples.

The paper says, at Section 2.5

Pith says: minor objection 2 of 7

The discussion of the stereographic projection for S^1 is clear, but the extension to higher-dimensional spheres (Remark 2.4) notes that no isometric chart exists. The paper could mention that the geodesic flow approach is used instead, which is fine, but the transition from chart-based to geodesic-based dynamics could be smoother.

show 5 more

The paper says, at Section 3.1, Example 3.3

Pith says: minor objection 3 of 7

The SM-Zigzag with velocity refreshment is described, but the paper does not specify how the refreshment rate is chosen or whether it affects the invariance. A brief comment on the rate selection would be helpful.

The paper says, at Section 4.3, Table 2

Pith says: minor objection 4 of 7

The RMSE comparison uses different budgets (CPU time vs. number of evaluations). While both are informative, the CPU time comparison may be hardware-dependent. The authors should clarify that the conclusions are robust across both metrics.

The paper says, at Section 4.4, Table 3

Pith says: minor objection 5 of 7

The RJ-HMC results show high sensitivity to step size, as noted. The paper could include a brief discussion on how to tune HMC in practice for such problems, or recommend using PDMP dynamics as a more robust alternative.

The paper says, at Appendix C.5.1

Pith says: minor objection 6 of 7

The discussion of length-scale differences and HMC acceptance rates is insightful. Consider moving some of this material to the main text to support the claim that PDMP dynamics are more robust to multiscale structure.

The paper says, at General

Pith says: minor objection 7 of 7

The paper is well-written and the notation is consistent. However, the length is substantial; some technical details in the appendices could be streamlined or referenced more concisely.

What would settle it

Take a simple non-isometric chart, e.g., the stereographic projection of S^2 onto R^2 (Section 2.5). Specify the ambient PDMP with standard Gaussian velocity and the manifold PDMP with the pushforward of the Gaussian through the chart (which is not Gaussian). Construct explicit Q_stick and Q_unstick kernels that satisfy (3.2). If no such construction is provided, or if the construction requires resampling the velocity (thereby losing the persistence property claimed as an advantage), then the generality of the framework is limited to isometric charts or trivial velocity distributions.

from the skeptic's reading

How the review went

4 readings by deepseek-v4-flash · 13 Nov 2026

  1. Reader holds up, with conditions confidence high
  2. Skeptic agrees 9 s · 71,932 tokens read, 815 written
  3. Editor wrote the verdict above 9 s · 72,910 tokens read, 570 written
  4. Referee major revision: 4 major objections, 7 minor 13 s · 72,314 tokens read, 1,364 written

Not run on this paper: the simulated authors' defence.

Pith did not check this review's objections against the paper.

The review in full

The review in brief, machine-written

Hi,

This paper generalises the sticky PDMP idea (Bierkens et al. 2023a) from coordinate-aligned linear subspaces to general embedded submanifolds of codimension one. The core contribution is the sticky manifold PDMP (SM-PDMP), a modular framework that combines a Euclidean PDMP in the ambient space with a manifold PDMP on the submanifold, with sticking and unsticking kernels that preserve the mixed-dimensional target. The invariance proof (Theorem 3.5) is rigorous, using a blow-up construction and a flow-wise Gauss–Green formula. The paper also develops a coordinate-chart-based manifold PDMP (Algorithm 1) and a Rao–Blackwellised unsticking rate that can improve efficiency. The numerical examples are well-chosen and demonstrate the method on several problems (quantum state tomography, Bayesian PCA, survival extrapolation).

The main soft spot is the velocity compatibility condition (3.2). The paper states it as an assumption and gives examples where it holds (Gaussian velocities, unit velocities), but these are essentially the same distribution on both sides. The paper acknowledges in Section 3.5 that for non-isometric charts (e.g., stereographic projection) the pushforward of the ambient Gaussian is not Gaussian on the manifold, and does not provide a general construction for Q_stick/Q_unstick that satisfies (3.2) in that setting. The stress-test note is correct: this is a load-bearing concern because the claimed generality of the framework depends on these conditions being "easily verifiable" (abstract). The paper also restricts to codimension-one, orientable, closed submanifolds, which is a significant limitation for many practical problems. The blow-up procedure for higher codimensions is discussed but not fully developed, and the paper notes that it may limit the range of priors (Remark 3.14).

Overall, this is a solid methodological contribution with a clean mathematical core. The limitations are acknowledged, but they are real and not minor. The paper deserves a serious referee; the authors should be asked to address the velocity compatibility condition more thoroughly, either by providing general constructions or by clarifying the scope of applicability.

Best, [Your name]

The strongest objection

The invariance proof (Theorem 3.5) rests on the velocity compatibility condition (3.2), which requires that Q_stick maps the incoming ambient velocity distribution to the prescribed joint distribution of the manifold velocity and the stored normal component, and Q_unstick does the reverse. The paper gives examples where this holds (Gaussian velocities, unit velocities), but these are essentially the same distribution on both sides. In the more general setting where the ambient and manifold velocity distributions differ—for instance, when the manifold PDMP uses a coordinate chart that is not isometric (Section 2.5, stereographic projection), the pushforward of the ambient Gaussian is not Gaussian on the manifold—the paper does not provide a construction of Q_stick/Q_unstick that satisfies (3.2). The authors acknowledge this gap in Section 3.5 ("two charts of the quadratic cone") but do not resolve it. Without a general method to construct these kernels, the applicability of the framework to non-isometric charts is unsubstantiated. This is a load-bearing concern because the central claim of the paper is that the SM-PDMP leaves the target invariant for any choice of component dynamics and unsticking kernels, subject to "easily verifiable velocity compatibility conditions" (abstract). If those conditions cannot be verified for a practically relevant class of manifold PDMPs, the claimed generality is overstated.

The reader's weakest_assumption correctly identifies the velocity compatibility condition (3.2) as a non-trivial assumption that must be verified for each specific choice of dynamics and kernels. The paper provides examples only for cases where the ambient and manifold velocity distributions are essentially the same (Gaussian or uniform on spheres). The concern I raise is more specific: even when the manifold PDMP is derived from a coordinate chart (Algorithm 1), the velocity distribution on the manifold is the pushforward of the ambient velocity distribution through the chart derivative. If the chart is not isometric, this pushforward distribution differs from the ambient distribution, and the paper does not show how to construct Q_stick/Q_unstick satisfying (3.2) in that case. This is a concrete gap in the claimed generality. The paper acknowledges the issue implicitly (Section 3.5 discusses the cone example but does not provide a general solution) and states that the velocity compatibility condition is 'easily verifiable' (abstract), but no verification is given for non-isometric charts. The concrete test I propose would settle whether this gap is real or merely apparent. I agree with the reader's assessment that this is a significant concern, hence 'agree'. The verdict should remain CONDITIONAL because the paper is otherwise sound and the numerical examples are convincing, but this gap needs to be addressed before the claim of general applicability can be accepted.

The referee's report

The paper addresses a challenging problem in Bayesian model averaging: efficient trans-dimensional sampling without reversible-jump proposals. The sticky manifold construction is principled, modular, and rejection-free, requiring no additional likelihood evaluations. The invariance proof is rigorous and general, accommodating non-linear flows and non-differentiable test functions via a flow-wise Gauss-Green formula. The numerical examples demonstrate competitive performance against reversible-jump methods, particularly in settings with varying posterior geometry. The paper also introduces Rao-Blackwellisation techniques to improve efficiency. The work is likely to be of interest to researchers in computational statistics and Bayesian inference.

Pith's reading of the paper (a machine's summary, not the authors')

Bayesian model averaging often involves sampling from posterior distributions that combine continuous parameters with discrete model uncertainty. Traditional reversible jump MCMC can be inefficient because designing good proposals for moving between models is difficult. This paper proposes a new method based on piecewise deterministic Markov processes (PDMPs). The key idea is to construct a 'sticky manifold' where the process moves in a high-dimensional ambient space, but when it hits a lower-dimensional submanifold (representing a simpler model), it 'sticks' to it and evolves on that submanifold for a random time before 'unsticking' and returning to the ambient space. The rate at which it unsticks is given by a simple formula that depends only on the prior and the normal component of velocity, not on the likelihood. This makes the transitions rejection-free and computationally cheap. The authors prove that this process leaves the correct posterior distribution invariant. They also develop a manifold PDMP algorithm based on coordinate charts for within-submodel dynamics and a blow-up procedure to handle singular sets. The method is demonstrated on several examples: quantum state tomography (distinguishing pure vs mixed states), Bayesian principal component analysis (selecting the rank), and survival analysis (selecting parametric families). The numerical results show that the sticky manifold PDMP can be more efficient than reversible jump MCMC, especially when combined with Rao-Blackwellisation techniques that reduce the variance of the sticking times.

Core claim. The sticky manifold PDMP, defined by the process in Algorithm 2 and the generator in (3.1), leaves the mixed-dimensional target measure Π (defined in (1.1) and lifted to the state space E) invariant. This is proven in Theorem 3.5 under Assumption 3.4 and the velocity compatibility condition (3.2). The proof rests on a flow-wise Gauss-Green formula (Lemma B.5) that accommodates non-linear flows and non-differentiable test functions.

The premise it rests on. The velocity compatibility condition (3.2) requires that the sticking kernel Q_stick maps the incoming ambient velocity distribution to the prescribed joint distribution of the manifold velocity and the stored normal component, and Q_unstick does the reverse. This condition is stated in Section 3.1 and is crucial for the invariance proof. While the paper provides examples where this holds (e.g., Gaussian velocities, unit velocities), it is a non-trivial condition that must be verified for each specific choice of PDMP dynamics and kernels. The paper acknowledges that this condition may not hold in general (e.g., for the zig-zag sampler with non-linear submanifolds, Example 3.3 requires velocity refreshment). The assumption that the submanifold M is of codimension one, orientable, and closed in R^d (Assumption 3.4) is also a significant restriction, as many practical problems involve higher codimensions or non-orientable manifolds.

Check this review

Pith stored no record of this review's passes and took no fingerprint of it. Reviews made by code from 13 Nov 2026 on carry both.

How often reviews like this one are right

Reviews like this one: written by deepseek-v4-flash, where what was read was not recorded.

How often objections from reviews like this one are right: not yet measured

Every value above, in the review's public record (JSON)

Reviews by people

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review

How Pith checks papers · The full audit of this paper

The paper

Abstract

Assigning positive prior probability to lower-dimensional submodels enriches the range of uncertainty that Bayesian models can represent. The resulting posterior distributions combine continuous parameter uncertainty with discrete model uncertainty. Efficient movement between these submodels, however, can be difficult to design. We propose the sticky manifold construction as a principled, modular framework for combining piecewise deterministic Markov processes (PDMPs) across different submodels. The integration is rejection-free, requires no additional likelihood evaluations, and retains the non-reversible and ballistic nature of the component PDMP dynamics. We prove that the resulting process leaves the intended mixed-dimensional target invariant. We also develop a manifold PDMP algorithm based on coordinate charts to provide the within-submodel dynamics and a blow-up procedure that resolves singular sets into codimension-one sticky boundaries. This geometric construction makes the remaining design choices explicit; component dynamics and unsticking kernels can be chosen freely, subject to easily verifiable velocity compatibility conditions. Finally, we demonstrate the sticky manifold algorithms in Bayesian model averaging examples, propagating uncertainty over pure versus mixed states in quantum state tomography, rank in Bayesian principal component analysis, and parametric families in survival extrapolation.

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.

abstract compiled · 2026-11-13

The authors' main claim, as the classifier picked it out 0.52

“We propose the sticky manifold construction as a principled, modular framework for combining piecewise deterministic Markov processes (PDMPs) across different submodels.”

How the main claim is stated: asserted 1.00

What the abstract itself states, as probabilities: claims first 0.05 · states limitation 0.25 · human participants 0.03 · controlled comparison 0.03 · randomized assignment 0.01 · states quantitative result 0.04 · compares with prior results 0.11 · mentions code or data availability 0.02

The abstract, sentence by sentence

This table labels the abstract only (arXiv listing abstract v1). The number is the probability the classifier gave the label.

SentenceRoleThe authors' own workStates the main conclusionStrength
Assigning positive prior probability to lower-dimensional submodels enriches the range of uncertainty that Bayesian models can represent. background 1.00 0.08 0.05 asserted 1.00
The resulting posterior distributions combine continuous parameter uncertainty with discrete model uncertainty. background 0.99 0.11 0.05 asserted 1.00
Efficient movement between these submodels, however, can be difficult to design. gap 0.69 0.05 0.02 asserted 0.82
We propose the sticky manifold construction as a principled, modular framework for combining piecewise deterministic Markov processes (PDMPs) across different submodels. contribution 0.96 0.94 0.32 asserted 1.00
The integration is rejection-free, requires no additional likelihood evaluations, and retains the non-reversible and ballistic nature of the component PDMP dynamics. contribution 0.52 0.89 0.16 asserted 1.00
We prove that the resulting process leaves the intended mixed-dimensional target invariant. result 0.75 0.96 0.22 asserted 1.00
We also develop a manifold PDMP algorithm based on coordinate charts to provide the within-submodel dynamics and a blow-up procedure that resolves singular sets into codimension-one sticky boundaries. method 0.87 0.95 0.09 asserted 1.00
This geometric construction makes the remaining design choices explicit; component dynamics and unsticking kernels can be chosen freely, subject to easily verifiable velocity compatibility conditions. method 0.59 0.84 0.12 asserted 0.99
Finally, we demonstrate the sticky manifold algorithms in Bayesian model averaging examples, propagating uncertainty over pure versus mixed states in quantum state tomography, rank in Bayesian principal component analysis, and parametric families in survival extrapolation. contribution 0.65 0.96 0.19 asserted 1.00

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from proof or derivation 0.83

What was examined, and what was not

Sources read

  • arXiv listing abstract v1

Not examined

  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers

Levels run

  • P abstract

Not run

  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

How the labels were made

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-11-13: 1 request, 6,842 input tokens.

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON

Figures (18)
Figure 1 of the paper
Figure 1. An illustration of the sticky manifold piecewise deterministic Markov process, whose trajectory is
Figure 2 of the paper
Figure 2. Representative trajectories on the common posterior geometry: (a) sticky manifold forward event
Figure 3 of the paper
Figure 3. Two parametrisations for the one-dimensional unit hypersphere. (a) represents the arc length
Figure 4 of the paper
Figure 4. Radial real blow-up of a smooth embedded submanifold
Figure 5 of the paper
Figure 5. Blow-up of a half cone, or equivalently the space of all symmetric semi-definite rank-one matrices.
Figure 6 of the paper
Figure 6. (a) Trajectory of the sticky boomerang sampler for
Figure 7 of the paper
Figure 7. Model-indicator autocorrelation. Only RB-SM-FECMC with fixed sticking times (gray) plunges
Figure 8 of the paper
Figure 8. An illustrative SM-FECMC posterior trajectory in scenario E. (a) The trajectory of the Schur
Figure 9 of the paper
Figure 9. Relative RMSEs across the six PPCA scenarios. Left: RJ-HMC relative to RJ-FECMC; middle:
Figure 10 of the paper
Figure 10. Sensitivity of RJ-HMC to hyperparameters across six scenarios. Blue symbols use four common
Figure 11 of the paper
Figure 11. (a) Full SM-FECMC trajectory for the joint AFT model with uniform prior model probabilities
Figure 12 of the paper
Figure 12. (a) A trajectory of the sticky zig-zag sampler with
Figure 13 of the paper
Figure 13. Histograms of sphere residence times for FECMC (blue) and FECMC+N (orange). Top:
Figure 14 of the paper
Figure 14. Autocorrelation of the model-indicator process for the four FECMC variants as in Table
Figure 15 of the paper
Figure 15. Step-size sensitivity of RJ-HMC with T = 3 integration steps. ϵ = 0.08, 0.09 used in
Figure 16 of the paper
Figure 16. Variation of the shortest length scale (left) and the condition number (right) within the main
Figure 17 of the paper
Figure 17. Estimated posterior rank probabilities pbk (k = 0, . . . , 20) from the long reference runs for the six scenarios A–E and O. For scenarios A–E, both initial positions and reference values were obtained using independent long RB￾SM-FECMC runs from dispersed starts. Pilot runs are used to determine burn-in time Tburn-in that satisfies (1) rank-normalised split R <b 1.01 (Vehtari et al., 2021), (2) pooled effective sample sizes were at least 1,000, (3) time averages between first and second halves of [Tburn-in, 2Tburn-in] match with a prescribed tolerance, and (4) each rank with…
Figure 18 of the paper
Figure 18. Relative RMSEs across the six PPCA scenarios. Left: RJ-HMC relative to RJ-FECMC; middle:

Figures are the authors', reproduced from the paper.

References (278)
  1. [1] Survival models based on the Ornstein-Uhlenbeck process , author=. Lifetime data analysis , volume=. 2004 , publisher= on Pith
  2. [2] Large Sample Scaling Analysis of the Zig-Zag Algorithm for Bayesian Inference , author=. arXiv preprint arXiv:2411.14983 , year= on PithPith's reviewarXiv
  3. [3] Bayesian inference for spatio-temporal spike-and-slab priors , author=. Journal of Machine Learning Research , volume= on Pith
  4. [4] Automated Techniques for Efficient Sampling of Piecewise-Deterministic Markov Processes , author=. 2024 , eprint= on Pith
  5. [5] Peskun-Tierney Ordering for Markovian Monte Carlo: Beyond the Reversible Scenario , author=. The Annals of Statistics , volume=. 2021 , publisher=. doi:10.1214/20-AOS2008 , url = on Pithdoi:10.1214/20-aos2008
  6. [6] A tutorial on adaptive MCMC , author=. Statistics and computing , volume=. 2008 , publisher= on Pith
  7. [7] Stable Survival Extrapolation via Transfer Learning , author=. arXiv preprint arXiv:2409.16044 , year= on PitharXiv
  8. [8] Entropy contraction of the Gibbs sampler under log-concavity , author=. arXiv preprint arXiv:2410.00858 , year= on PitharXiv
  9. [9] Dimension-free mixing times of Gibbs samplers for Bayesian hierarchical models , author=. The Annals of Statistics , volume=. 2024 , publisher= on Pith
  10. [10] Adrian Bagust and Sophie Beale , title =. Medical Decision Making , volume =. 2014 , doi =. https://doi.org/10.1177/0272989X13497998 , abstract = on Pithdoi:10.1177/0272989x13497998
  11. [11] Bayesian variable selection and shrinkage strategies in a complicated modelling setting with missing data: A case study using multistate models , author=. Statistical Modelling , volume=. 2021 , publisher= on Pith
  12. [12] A review of survival analysis methods used in NICE technology appraisals of cancer treatments: consistency, limitations, and areas for improvement , author=. Medical Decision Making , volume=. 2019 , publisher= on Pith
  13. [13] Adaptive Stereographic MCMC , author=. 2025 , eprint= on Pith
  14. [14] Survival extrapolation in the presence of cause specific hazards , author=. Statistics in Medicine , volume=. 2015 , publisher= on Pith
  15. [15] Bayesian analysis for the poly-Weibull distribution , author=. Journal of the American Statistical Association , volume=. 1993 , publisher= on Pith
  16. [16] Piecewise deterministic sampling with splitting schemes , author=. arXiv preprint arXiv:2301.02537 , year= on PitharXiv
  17. [17] Monmarché, Pierre and Weisman, Jérémy and Lagardère, Louis and Piquemal, Jean-Philip , title =. The Journal of Chemical Physics , volume =. 2020 , month =. doi:10.1063/5.0005060 , url = on Pithdoi:10.1063/5.0005060
  18. [18] Adaptive Schemes for Piecewise Deterministic Monte Carlo Algorithms , author=. Bernoulli , volume=. 2022 , publisher=. doi:10.3150/21-BEJ1423 , URL = on Pithdoi:10.3150/21-bej1423
  19. [19] Bierkens, Joris and Grazzi, Sebastiano and Meulen, Frank van der and Schauer, Moritz , date =. Statistics and Computing , number =. 2023 , bdsk-url-1 =. doi:10.1007/s11222-022-10180-5 , id = on Pithdoi:10.1007/s11222-022-10180-5
  20. [20] Methods and Applications of PDMP Samplers with Boundary Conditions , author=. arXiv preprint arXiv:2303.08023 , year= on PithPith's reviewarXiv
show 80 more
  1. [21] Scaling of Piecewise Deterministic Monte Carlo for Anisotropic Targets , author =. Bernoulli , number =. 2025 , doi = on Pith
  2. [22] Joris Bierkens and Kengo Kamatani and Gareth O. Roberts , doi =. The Annals of Applied Probability , keywords =. 2022 , bdsk-url-1 = on Pith
  3. [23] Joris Bierkens and Paul Fearnhead and Gareth Roberts , journal =. 2019 , pages = on Pith
  4. [24] autoMALA: Locally adaptive Metropolis-adjusted Langevin algorithm , author=. International Conference on Artificial Intelligence and Statistics , pages=. 2024 , organization= on Pith
  5. [25] Spatial and spatio-temporal models with R-INLA , author=. Spatial and spatio-temporal epidemiology , volume=. 2013 , publisher= on Pith
  6. [26] GIST: Gibbs self-tuning for locally adaptive Hamiltonian Monte Carlo , author=. arXiv preprint arXiv:2404.15253 , year= on PithPith's reviewarXiv
  7. [27] Nawaf Bou-Rabee and Bob Carpenter and Tore Selland Kleppe and Sifan Liu , title =. Journal of Machine Learning Research , year = on Pith
  8. [28] Nawaf Bou-Rabee and Jes. The Annals of Applied Probability , number =. 2017 , doi = on Pith
  9. [29] Alexandre Bouchard-Côté and Sebastian J. Vollmer and Arnaud Doucet , year =. Journal of the American Statistical Association , volume = on Pith
  10. [30] Bayesian survival analysis using the rstanarm R package , author=. arXiv preprint arXiv:2002.09633 , year= on PithPith's reviewarXiv
  11. [31] Brooks, S. P. and Giudici, P. and Roberts, G. O. , title =. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume =. 2003 , month =. doi:10.1111/1467-9868.03711 , url = on Pithdoi:10.1111/1467-9868.03711
  12. [32] A flexible, random histogram kernel for discrete-time Hawkes processes , author=. arXiv preprint arXiv:2208.02921 , year= on PithPith's reviewarXiv
  13. [33] Geodesic Monte Carlo on Embedded Manifolds , author=. Scandinavian Journal of Statistics , volume=. 2013 , publisher=. doi:https://doi.org/10.1111/sjos.12036 , url = on Pithdoi:10.1111/sjos.12036
  14. [34] Girolami, Mark and Calderhead, Ben , title =. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume =. doi:https://doi.org/10.1111/j.1467-9868.2010.00765.x , url =. https://rss.onlinelibrary.wiley.com/doi/pdf/10.1111/j.1467-9868.2010.00765.x , a... on Pithdoi:10.1111/j.1467-9868.2010.00765.x
  15. [35] Bayesian inference in a piecewise Weibull proportional hazards model with unknown change points , author=. Journal of Animal Breeding and Genetics , volume=. 2007 , publisher= on Pith
  16. [36] A novel Bayesian continuous piecewise linear log-hazard model, with estimation and inference via reversible jump Markov chain Monte Carlo , author=. Statistics in medicine , volume=. 2020 , publisher= on Pith
  17. [37] Blended survival curves: a new approach to extrapolation for time-to-event outcomes from clinical trials in health technology assessment , author=. Medical Decision Making , volume=. 2023 , publisher= on Pith
  18. [38] Bayesian shared parameter joint models for heterogeneous populations , author=. arXiv preprint arXiv:2410.22534 , year= on PithPith's reviewarXiv
  19. [39] Reversible Jump PDMP Samplers for Variable Selection , author=. Journal of the American Statistical Association , volume=. 2023 , publisher= on Pith
  20. [40] PDMP Monte Carlo Methods for Piecewise Smooth Densities , volume=. Advances in Applied Probability , author=. 2024 , pages=. doi:10.1017/apr.2023.61 , number= on Pithdoi:10.1017/apr.2023.61
  21. [41] Bayesian blockwise inference for joint models of longitudinal and multistate data with application to longitudinal multimorbidity analysis , author=. Statistical Methods in Medical Research , pages=. 2024 , publisher= on Pith
  22. [42] MCMC using textit \ bouncy \ Hamiltonian dynamics: A unifying framework for Hamiltonian Monte Carlo and piecewise deterministic Markov process samplers , author=. arXiv preprint arXiv:2405.08290 , year= on PitharXiv
  23. [43] Nishimura, Akihiko and Dunson, David B and Lu, Jianfeng , title =. Biometrika , volume =. 2020 , month =. doi:10.1093/biomet/asz083 , url = on Pithdoi:10.1093/biomet/asz083
  24. [44] Smoothing Out Sticking Points: Sampling from Discrete-Continuous Mixtures with Dynamical Monte Carlo by Mapping Discrete Mass into a Latent Universe , author=. arXiv preprint arXiv:2511.07340 , year= on PitharXiv
  25. [45] Dynamic survival analysis: Modelling the hazard function via ordinary differential equations , author=. Statistical Methods in Medical Research , pages=. 2024 , publisher= on Pith
  26. [46] On harmonic oscillator hazard functions , author=. arXiv preprint arXiv:2408.15964 , year= on PithPith's reviewarXiv
  27. [47] Philip Cooney and Arthur White , keywords =. Extending Beyond Bagust and Beale: Fully Parametric Piecewise Exponential Models for Extrapolation of Survival Outcomes in Health Technology Assessment , journal =. 2023 , issn =. doi:https://doi.org/10.1016/j.jval.2023.06.007 , url = on Pithdoi:10.1016/j.jval.2023.06.007
  28. [48] Corbella, Alice and Spencer, Simon E. F. and Roberts, Gareth O. , date =. Statistics and Computing , number =. 2022 , bdsk-url-1 =. doi:10.1007/s11222-022-10142-x , id = on Pithdoi:10.1007/s11222-022-10142-x
  29. [49] A class of dynamic piecewise exponential models with random time grid , author=. Journal of Statistical Planning and Inference , volume=. 2012 , publisher= on Pith
  30. [50] Free knot splines with RJMCMC in survival data analysis , author=. Communications in Statistics—Theory and Methods , volume=. 2010 , publisher= on Pith
  31. [51] Bayesian methods for nonlinear classification and regression , author=. 2002 , publisher= on Pith
  32. [52] Sampling from a Manifold , author=. Advances in modern statistical theory and applications: a Festschrift in honor of Morris L. Eaton , volume=. 2013 , publisher=. doi:10.1214/12-IMSCOLL1006 , url = on Pithdoi:10.1214/12-imscoll1006
  33. [53] Bayesian inference for generalized additive mixed models based on Markov random field priors , author=. Journal of the Royal Statistical Society Series C: Applied Statistics , volume=. 2001 , publisher= on Pith
  34. [54] A closed-form filter for binary time series , author=. Statistics and Computing , volume=. 2021 , publisher= on Pith
  35. [55] Sampling algorithms in statistical physics: a guide for statistics and machine learning , author=. Statistical Science , volume=. 2024 , publisher= on Pith
  36. [56] Stochastic Gradient Piecewise Deterministic Monte Carlo Samplers , author=. arXiv preprint arXiv:2406.19051 , year= on PithPith's reviewarXiv
  37. [57] Scalable Monte Carlo for Bayesian Learning , author=. arXiv preprint arXiv:2407.12751 , year= on PithPith's reviewarXiv
  38. [58] Paul Fearnhead and Christopher Nemeth and Chris J. Oates and Chris Sherlock , year =. Scalable Monte Carlo for Bayesian Learning , series =. doi:https://doi.org/10.1017/9781009288460 , publisher = on Pithdoi:10.1017/9781009288460
  39. [59] Piecewise deterministic Markov processes for continuous-time Monte Carlo , author=. Statistical Science , volume=. 2018 , publisher= on Pith
  40. [60] Estimation of exponential survival probabilities with concomitant information , author=. Biometrics , pages=. 1965 , publisher= on Pith
  41. [61] A unified view on Bayesian varying coefficient models , author= on Pith
  42. [62] Adaptive Monte Carlo augmented with normalizing flows , author=. Proceedings of the National Academy of Sciences , volume=. 2022 , publisher= on Pith
  43. [63] Visualization in Bayesian workflow , author=. Journal of the Royal Statistical Society Series A: Statistics in Society , volume=. 2019 , publisher= on Pith
  44. [64] Nonreversible Jump Algorithms for Bayesian Nested Model Selection , author=. Journal of Computational and Graphical Statistics , volume=. 2020 , publisher= on Pith
  45. [65] Dynamic Bayesian models for survival data , author=. Journal of the Royal Statistical Society Series C: Applied Statistics , volume=. 1991 , publisher= on Pith
  46. [66] A weakly informative default prior distribution for logistic and other regression models , author= on Pith
  47. [67] Piecewise deterministic Markov processes for Bayesian neural networks , author=. Uncertainty in Artificial Intelligence , pages=. 2023 , organization= on Pith
  48. [68] Spatiotemporal blocking of the bouncy particle sampler for efficient inference in state-space models , author=. Statistics and Computing , volume=. 2021 , publisher= on Pith
  49. [69] Reversible Jump Markov Chain Monte Carlo Computation and Bayesian Model Determination , author=. Biometrika , volume=. 1995 , publisher=. doi:https://doi.org/10.1093/biomet/82.4.711 , url = on Pithdoi:10.1093/biomet/82.4.711
  50. [70] Enhanced secondary analysis of survival data: reconstructing the data from published Kaplan-Meier survival curves , author=. BMC medical research methodology , volume=. 2012 , publisher= on Pith
  51. [71] Diffusion Piecewise Exponential Models for Survival Extrapolation using Piecewise Deterministic Monte Carlo , author=. arXiv preprint arXiv:2505.05932 , year= on PithPith's reviewarXiv
  52. [72] Luke Hardcastle and Samuel Livingstone and Gianluca Baio , title =. The Annals of Applied Statistics , number =. 2025 , doi = on Pith
  53. [73] Application of the Bayesian dynamic survival model in medicine , author=. Statistics in Medicine , volume=. 2010 , publisher= on Pith
  54. [74] A parametric dynamic survival model applied to breast cancer survival times , author=. Journal of the Royal Statistical Society Series C: Applied Statistics , volume=. 2002 , publisher= on Pith
  55. [75] A fresh take on ‘Barker dynamics’ for MCMC , author=. International Conference on Monte Carlo and Quasi-Monte Carlo Methods in Scientific Computing , pages=. 2020 , organization= on Pith
  56. [76] Strong convergence of an explicit numerical method for SDEs with nonglobally Lipschitz continuous coefficients , author=. The Annals of Applied Probability , volume=. 2012 , publisher= on Pith
  57. [77] Bayesian survival analysis , author=. 2001 , publisher= on Pith
  58. [78] survextrap: a package for flexible and transparent survival extrapolation , author=. BMC Medical Research Methodology , volume=. 2023 , publisher= on Pith
  59. [79] Christopher Jackson and John Stevens and Shijie Ren and Nick Latimer and Laura Bojke and Andrea Manca and Linda Sharples , title =. Medical Decision Making , volume =. 2017 , doi =. https://doi.org/10.1177/0272989X16639900 , abstract = on Pithdoi:10.1177/0272989x16639900
  60. [80] Markov Chain Monte Carlo Methods and the Label Switching Problem in Bayesian Mixture Modeling , author=. Statistical Science , volume= on Pith
  61. [81] Xiongwen Ke and Yanan Fan and Qingping Zhou , year=. Fused L_. 2409.07874 , archivePrefix= on PitharXiv
  62. [82] Dynamic and Flexible Survival Models for Extrapolation of Relative Survival: A Case Study and Simulation Study , author=. Medical Decision Making , volume=. 2022 , publisher= on Pith
  63. [83] Generalized linear models for flexible parametric modeling of the hazard function , author=. Medical Decision Making , volume=. 2019 , publisher= on Pith
  64. [84] Bayesian Analysis of the Generalized Additive Proportional Hazards Model: Asymptotic Studies , author=. Bayesian Analysis , volume=. 2023 , publisher= on Pith
  65. [85] Zig-Zag Sampling for Discrete Structures and Nonreversible Phylogenetic MCMC , author=. Journal of Computational and Graphical Statistics , volume=. 2022 , publisher=. doi:https://doi.org/10.1080/10618600.2022.2032722 , url = on Pithdoi:10.1080/10618600.2022.2032722
  66. [86] Posterior analysis of latent competing risk models by parallel tempering , author=. Computational statistics & data analysis , volume=. 2004 , publisher= on Pith
  67. [87] tfp. mcmc: Modern Markov chain Monte Carlo tools built for modern hardware , author=. arXiv preprint arXiv:2002.01184 , year= on PithPith's reviewarXiv
  68. [88] The random timestep Euler method and its continuous dynamics , author=. arXiv preprint arXiv:2408.01409 , year= on PithPith's reviewarXiv
  69. [89] Lewis, P. A. W and Shedler, G. S. , doi =. Simulation of Nonhomogeneous Poisson Processes by Thinning , url =. 1979 , bdsk-url-1 =. https://onlinelibrary.wiley.com/doi/pdf/10.1002/nav.3800260304 , journal = on Pithdoi:10.1002/nav.3800260304
  70. [90] On the effect of prior assumptions in Bayesian model averaging with applications to growth regression , author=. Journal of applied econometrics , volume=. 2009 , publisher= on Pith
  71. [91] Structure Learning with Adaptive Random Neighborhood Informed MCMC , author=. Advances in Neural Information Processing Systems , volume= on Pith
  72. [92] Adaptive mcmc for bayesian variable selection in generalised linear models and survival models , author=. Entropy , volume=. 2023 , publisher= on Pith
  73. [93] Adaptive random neighbourhood informed Markov chain Monte Carlo for high-dimensional Bayesian variable selection , author=. Statistics and Computing , volume=. 2022 , publisher= on Pith
  74. [94] Bags: A Bayesian adaptive group sequential trial design with subgroup-specific survival comparisons , author=. Journal of the American Statistical Association , volume=. 2021 , publisher= on Pith
  75. [95] Skew-symmetric schemes for stochastic differential equations with non-Lipschitz drift: an unadjusted Barker algorithm , author=. arXiv preprint arXiv:2405.14373 , year= on PithPith's reviewarXiv
  76. [96] The Barker proposal: combining robustness and efficiency in gradient-based MCMC , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 2022 , publisher= on Pith
  77. [97] Almost sure stability of the Euler--Maruyama method with random variable stepsize for stochastic differential equations , author=. Numerical Algorithms , volume=. 2017 , publisher= on Pith
  78. [98] Polyhazard models for lifetime data , author=. Biometrics , volume=. 1999 , publisher= on Pith
  79. [99] Robust Approximate Sampling via Stochastic Gradient Barker Dynamics , author=. International Conference on Artificial Intelligence and Statistics , pages=. 2024 , organization= on Pith
  80. [100] Bayesian inference for polyhazard models in the presence of covariates , author=. Computational statistics & data analysis , volume=. 2001 , publisher= on Pith

The first 100 of 278 references are listed here. All of them are in this paper's citation record.

Challenges and discussion

Disagree with the review, or know something it missed? Leave a comment, or challenge this review formally: a challenge becomes a permanent entry in this paper's record, and its decision is published here.

No comments yet.

Sign in to comment