Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

A Proximal Newton Adaptive Importance Sampler

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A scaled Newton proximal step lets importance samplers adapt proposals to targets that split into a smooth part and a convex nonsmooth part, extending gradient-based adaptation to nondifferentiable Bayesian models.

desk verdict Plausible extension of adaptive importance sampling to nonsmooth targets via a proximal Newton step; the experiments support it, but missing error bars, code, and the closest baseline leave the case conditional. read the letter →

arxiv 2412.16558 v2 pith:KLCYHVP7 submitted 2024-12-21 stat.CO stat.ME

Víctor Elvira , Émilie Chouzenoux , O. Deniz Akyildiz This is my paper · ORCID
classification stat.COstat.ME MSC 65C0562F1590C2565K05
keywords adaptiveimportancesamplingproximalmethodsNewtonmethodnonsmoothtargetdistributionsproposaladaptationpopulationMonteCarlosparsepriorspreconditioning
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

Adaptive importance samplers improve their proposals over time, but most gradient-based adaptation requires a differentiable target. This paper proposes PNAIS, which adapts both the means and the covariances of a population of proposals using a scaled Newton proximal gradient step, so that the target need only split into a smooth part and a convex nonsmooth part. Experiments on a simplex-constrained Gaussian mixture and a Gaussian likelihood with a sparse $\ell^1$ prior show that PNAIS attains substantially lower relative mean squared error than non-adaptive and first-order ablated variants. If the result holds, gradient-based proposal adaptation becomes available for a wide class of nonsmooth Bayesian models, including sparsity-promoting priors and hard constraints.

What carries the argument

The load-bearing device is the scaled proximal Newton update in equation (10), where the proximity operator of $g$ is taken with respect to the metric $A^{-1}$ with $A = \theta\,\Gamma$; this simultaneously plays the role of a gradient step for the smooth part $f$ and a proximal step for the nonsmooth part $g$. The matrix $\Gamma$ is the inverse Hessian of $f$ when it is positive definite, and the previous proposal covariance otherwise, and $\theta$ is a damped factor tuned by backtracking to keep the step stable. The covariance of each proposal is then set equal to $A$, coupling the proposal spread to the local curvature. When the proximal step has no closed form, a dual forward-backward algorithm computes it, and the 'glocal' resampling scheme (GLR) interleaves local and global resampling to form the next generation of proposals.

What would settle it

Take a unimodal target of the form $\exp(-f-g)$ with a unique minimizer, run PNAIS with GLR resampling, and compare the adapted means and estimator MSE against the same update with resampling disabled; if the resampled means drift away from the minimizer or the MSE fails to decrease while the no-resampling variant converges, the transfer assumption is violated.

Watch

Extended reading notes

Core claim

The central claim is that for a target of the form $\tilde{\pi}(x) \propto \exp(-f(x)-g(x))$ with $f$ differentiable and $g$ convex nonsmooth, the mean update $\mu^{(t+1)}_n = \mathrm{prox}_{A^{-1},g}\bigl(\tilde{\mu}^{(t)}_n - A(\tilde{\mu}^{(t)}_n)\nabla f(\tilde{\mu}^{(t)}_n)\bigr)$ with $A(\tilde{\mu}) = \theta\,\Gamma(\tilde{\mu})$, where $\Gamma$ is the inverse Hessian of $f$ when positive definite and the current proposal covariance otherwise, together with the covariance update $\Sigma^{(t+1)}_n = A(\tilde{\mu}^{(t)}_n)$, produces a sequence of proposals that efficiently concentrate mass on the target. The proximal operator of $g$ handles the nonsmooth part (projection for constraints, soft-thresholding for $\ell^1$), while the Newton scaling supplies local second-order information of the smooth part. The paper demonstrates this by showing that, on the tested examples, PNAIS achieves the lowest relative MSE among the variants considered, particularly for the normalization constant estimate.

Load-bearing premise

The load-bearing premise is that the convergence and stability of the proximal Newton iteration, which the paper only cites for the no-resampling setting, carry over unchanged when the same update is embedded inside the stochastic resampling loop of the sampler; Section III-B explicitly states the convergence result 'without resampling' and offers no analysis of the resampled iteration.

Editorial extensions

If this is right

  • Sparsity-inducing priors such as $\ell^1$ and hard constraints such as simplex or $\ell^2$-ball domains can be handled by AIS without smoothing or variable transformation.
  • The covariance rule (13) provides a data-dependent proposal scale, removing the need to hand-tune a fixed isotropic covariance.
  • The algorithm's core only needs a computable proximity operator for $g$, so new nonsmooth terms can be plugged in without changing the machinery.
  • The reported computational time is comparable to first-order and non-adaptive variants, suggesting second-order information comes at little extra cost.
  • The GLR scheme's combination of diversity-preserving local resampling with global resampling every $\Delta$ iterations appears to keep proposals from collapsing while still letting optimization move them.

Reading between the lines

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

  • A natural ablation the paper does not run is to disable resampling entirely; comparing PNAIS with and without GLR would isolate whether the optimization step's convergence guarantees survive the resampling loop, and would directly test the weakest assumption.
  • If the Hessian-inverse covariance choice is read as a local Laplace approximation, PNAIS can be seen as building an adaptive Gaussian approximation of a nonsmooth posterior; this interpretation is not developed in the paper.
  • Because the update only requires a prox for $g$, the same scheme could in principle accommodate nonconvex regularizers with well-defined prox (e.g., MCP or SCAD), though the safe rules and convergence arguments would need revisiting.
  • A benchmark against other sampling schemes that handle nonsmooth targets, such as proximal MCMC with Moreau-Yosida smoothing, would clarify where PNAIS sits relative to existing tools; the current comparisons are only against ablated self-variants.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes PNAIS (Proximal Newton Adaptive Importance Sampler), an adaptive importance sampling method for target densities of the form exp(-f-g), where f is differentiable and g is convex but nonsmooth (e.g., an L1 penalty or an indicator of a constraint set). At each iteration, after a 'glocal' resampling step, each proposal mean is updated by one scaled proximal Newton step (Eq. 10) that uses the gradient and Hessian of f, and each proposal covariance is set to the same scaling matrix (Eq. 13) used in that step. The paper reports an ablation study on two 2D examples (a Gaussian mixture truncated to a simplex, and a Gaussian likelihood with an L1 prior) plus a high-dimensional constrained example in the supplement, showing that PNAIS often attains lower relative mean squared error than ablated variants. The central claim is that the proximal Newton adaptation improves efficiency for non-smooth targets without breaking the validity of the importance sampling estimators.

Significance. If the empirical claims hold, PNAIS is a useful extension of gradient-based adaptive importance sampling to nonsmooth targets, with potential impact in Bayesian signal processing and sparse regression. The use of proximal operators is principled, and the ablation design (comparing with no adaptation, no covariance adaptation, robust covariance adaptation, and gradient-only mean adaptation) is sensible. The paper's main strength is the algorithmic combination of a damped proximal Newton step with population Monte Carlo resampling, which is plausible and clearly explained. A notable weakness is the absence of any theoretical analysis of the full resampled algorithm, and the empirical evidence is reported without uncertainty quantification. The paper also stops short of addressing a potential tail-heaviness problem in the proposal adaptation when the damping parameter is below one. These issues limit the strength of the general efficiency claim, but they are addressable in a revision.

major comments (3)
  1. [Section III-A/B] The covariance update (13) sets Σ_n^(t+1) = A(µ~_n^(t)) = θ_n^(t) Γ(µ~_n^(t)). For a target of the form (2) with a convex quadratic f and g = α||x||_1, the Gaussian proposal with covariance θΓ is lighter-tailed than the target whenever θ<1, because the proposal precision is θ^{-1}∇²f. Then log(π/q) contains the positive quadratic term (1/(2θ)-1/2)(x-µ)^T∇²f(µ)(x-µ), which tends to +∞ as ||x||→∞, so the importance weights are unbounded and the IS variance can be infinite. The paper reports no accepted θ values, no effective sample sizes, and no tail diagnostics for Example B, so the low relative MSE in Table II may reflect a favorable regime where the backtracking keeps θ near 1 rather than a robust property of the adaptation. Please provide an analysis or a safeguard (e.g., decouple the damping for the covariance so that the proposal is never lighter-tailed than the smooth part, or use a heavier-tailed proposal family), and report θ/ESS statistics in the experiments.
  2. [Section III-A/B] The convergence guarantees cited for the mean update (10) apply to deterministic iterates without resampling (see [42], [6]). In PNAIS, the sequence fed into (10) is the resampled mean µ~_n^(t) from the GLR step (Section III-A), which depends on the weighted particle pool and is therefore stochastic; the descent properties or the assumptions of [42] (e.g., Lipschitz continuity of the Hessian along the iterates) are not established for the resampled sequence. The paper also gives no argument that the self-normalized estimators built from the final weighted sample pool are consistent as (N,K,T) grow. Since the central claim is about estimator accuracy, the paper should either provide a rigorous justification of the full resampled algorithm under stated conditions, or explicitly frame PNAIS as a heuristic whose properties are only empirical.
  3. [Section IV] All relative MSE results are reported as point estimates over 100 independent runs, with no standard errors, confidence intervals, or significance tests. For example, in Table II (right), the relative MSE for E[X] is 1.56e-5 for PNAIS(13) versus 1.10e-5 for PNAIS-grad(13), a difference of only about 40% that may not be statistically significant at 100 runs; the text's claim that PNAIS is 'superior' in this example is not supported without uncertainty quantification. Please provide standard errors, paired statistical comparisons, or boxplots for the main tables so that the superiority claims can be properly assessed.
minor comments (5)
  1. [Section III-B] The notation for the proximity operator in Eq. (10) is inconsistent with Eq. (6) and with the supplementary material (which uses prox_{A^{-1},g}). Please clarify whether the metric matrix is A or A^{-1} both in the main text and in the inner-solver description.
  2. [Section IV] The ground-truth integrals for Example A are said to be 'determined by numerical integration with a rough grid.' Please specify the grid resolution and provide an accuracy check (e.g., convergence under grid refinement) so that the benchmark values can be considered reliable.
  3. [Section IV] The sentence 'PNAIS shows superior performance with the Hessian-based covariance adaptation and without metric acceleration in the proximal gradient mean adaptation' is ambiguous: in Table II, the PNAIS-grad variant with Eq. (13) actually achieves lower MSE for E[X] and E[X^2] than the full PNAIS, while PNAIS is better for Z. Please rephrase to avoid contradicting the table.
  4. [Section III] The paper would benefit from a brief discussion of the choice of Gaussian proposals in the light of the tail-heaviness concern raised above; a Student-t proposal with finite degrees of freedom would be a safer default for nonsmooth targets.
  5. No code is released. Given that the algorithm is described in full, providing reference code (even a MATLAB/Python snippet) would substantially improve reproducibility and uptake.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the proposal adaptation consumes the target's gradient and Hessian rather than the estimated integrals, and all ground-truth values are external to the algorithm.

full rationale

The paper's central claim is that the scaled proximal-Newton adaptation (10)-(13) improves importance-sampling efficiency for targets of the form exp(-f-g). The quantities defining the adaptation are the target's smooth part f (via its gradient and Hessian) and the nonsmooth part g (via its proximity operator); none of these is defined in terms of the expectations being estimated or in terms of the reported MSE. The ground-truth integrals for Examples A and B are obtained by numerical integration (e.g., E~pi[X]=[0.2369,0.3023]^T and Z=0.5398) or from the known Gaussian-Laplace structure, and Example C discards a coordinate precisely because no ground truth is available. The ablation competitors are alternative algorithms rather than fitted versions of PNAIS, so comparing against them is not a fitted-input-called-prediction. The paper's convergence statements are explicitly qualified: 'The convergence of the scheme (10), without resampling..., has been established...' and the paper does not claim a proof for the resampled loop; that is an unproved transfer of a known optimization result, which is a robustness or correctness gap rather than circularity. The citations to [6], [7], [18], and [42] provide background for the proximal-Newton and Langevin inspiration; they do not by themselves force the numerical outcomes, and the main numerical evidence is self-contained against external benchmarks. Overall, no equation in the paper reduces to its own input, and no fitted parameter is renamed as a prediction.

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

The central claim rests on a standard target decomposition, cited optimization theory, and the practical computability of proximal operators. No data-fitted constants appear. The key unexamined load is the transfer of the no-resampling convergence theory to the stochastic resampled loop.

assumptions (4)
  • domain assumption The target distribution factorizes as exp(-f(x)-g(x)) with f differentiable and g proper lower semicontinuous convex (Eq. 2, Section II-A).
    This structure is the problem class the algorithm is designed for; it excludes targets where the nonsmooth part is non-convex or not separable.
  • standard math The scaled proximal gradient and variable metric proximal Newton iterations converge to a minimizer of f+g under stated conditions (Section II-D and III-B, citing [6,42,14,50]).
    The paper relies on these cited convergence results without reproving them, and then assumes they justify the mean adaptation when combined with resampling.
  • domain assumption The proximal operator of g can be evaluated in closed form or via the dual forward-backward inner solver (Section III-C and Table S-I).
    The numerical examples use projection and soft-thresholding; for general convex g, the inner solver's convergence to the required tolerance is assumed.
  • ad hoc to paper The ground truth integrals for Examples A and B, obtained by numerical integration with a rough grid, are accurate enough to serve as benchmarks (Section IV-A).
    The reported relative MSE values depend on the accuracy of these ground truths; for Example C the first coordinate has no ground truth and is excluded.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Proximal Newton Adaptive Importance Sampler." pith.science (2026). https://pith.science/paper/KLCYHVP7

@misc{pith2026241216558,
  author       = {Pith},
  title        = {Pith review of: A Proximal Newton Adaptive Importance Sampler},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KLCYHVP7}},
  note         = {Machine review of arXiv:2412.16558}
}
read the original abstract

Adaptive importance sampling (AIS) algorithms are a rising methodology in signal processing, statistics, and machine learning. An effective adaptation of the proposals is key for the success of AIS. Recent works have shown that gradient information about the involved target density can greatly boost performance, but its applicability is restricted to differentiable targets. In this paper, we propose a proximal Newton adaptive importance sampler for the estimation of expectations with respect to non-smooth target distributions. We implement a scaled Newton proximal gradient method to adapt the proposal distributions, enabling efficient and optimized moves even when the target distribution lacks differentiability. We show the good performance of the algorithm in two scenarios: one with convex constraints and another with non-smooth sparse priors.

Figures

Figures reproduced from arXiv: 2412.16558 by the authors.

Figure 1
Figure 1. (left) Target function for Example A. The Target equa [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗

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. MCMC Importance Sampling via Moreau-Yosida Envelopes

    stat.CO 2025-01 accept novelty 6.0 of 10

    Using the Moreau-Yosida envelope density as an importance distribution yields an asymptotically normal, finite-variance Markov chain importance sampling estimator that often beats proximal MALA and HMC.

Reference graph

Works this paper leans on

57 extracted references · 52 canonical work pages · cited by 1 Pith paper

  1. [53]

    Shukla, D

    A. Shukla, D. V ats, and E. C. Chi. Mcmc importance sampli ng via moreau-yosida envelopes. arXiv preprint arXiv:2501.02228, 2025

    arXiv 2025

  2. [42]

    J. D. Lee, Y . Sun, and M. A. Saunders. Proximal newton-ty pe methods for minimizing composite functions. SIAM Journal on Optimization , 24(3):1420–1443, 2014

    work page 2014

  3. [6]

    Becker and J

    S. Becker and J. Fadili. A quasi-newton proximal splitti ng method. In Advances in Neural Information Processing Systems , volume 25, 2012

    work page 2012

  4. [1]

    Abboud, M

    F. Abboud, M. Stamm, E. Chouzenoux, J.-C. Pesquet, and H. Talbot. Distributed Algorithms for Scalable Proximity Operator Co mputation and Application to Video Denoising. Digital Signal Processing , 128:103610, Aug. 2022

    work page 2022

  5. [2]

    O. D. Akyildiz. Global convergence of optimized adaptiv e importance samplers. F oundations of Data Science , 2024

    work page 2024

  6. [3]

    ¨O. D. Akyildiz and J. M´ ıguez. Convergence rates for optimis ed adaptive importance samplers. Statistics and Computing , 31:1–17, 2021

    work page 2021

  7. [4]

    Attouch and J

    H. Attouch and J. Bolte. On the convergence of the proxima l algorithm for nonsmooth functions involving analytic features. Mathematical Programming, 116:5–16, 2009

    work page 2009

  8. [5]

    Attouch, J

    H. Attouch, J. Bolte, and B. F. Svaiter. Convergence of de scent methods for semi-algebraic and tame problems: proximal algorithms , forward– backward splitting, and regularized gauss–seidel methods . Mathematical Programming, 137(1):91–129, 2013

    work page 2013

Show all 57 references
  1. [7]

    Bonettini, I

    S. Bonettini, I. Loris, F. Porta, and M. Prato. V ariable m etric inexact line-search-based methods for nonsmooth optimization. SIAM Journal on Optimization , 26(2):891–921, 2016

    2016

  2. [8]

    J. F. Bonnans, J. C. Gilbert, C. Lemar´ echal, and C. A. Sag astiz´ abal. A family of variable metric proximal methods. Mathematical Program- ming, 68:15–47, 1995

    1995

  3. [9]

    M. F. Bugallo, V . Elvira, L. Martino, D. Luengo, J. Miguez , and P . M. Djuric. Adaptive Importance Sampling: The past, the presen t, and the future. IEEE Signal Processing Magazine , 34(4):60–79, 2017

    2017

  4. [10]

    J. V . Candy. Bayesian Signal Processing: Classical, Modern, and Particle Filtering Methods , volume 54. John Wiley & Sons, 2016

    2016

  5. [11]

    Capp´ e, R

    O. Capp´ e, R. Douc, A. Guillin, J.-M. Marin, and C. P . Rob ert. Adaptive importance sampling in general mixture classes. Statistics and Comput- ing, 18:447–459, 2008

    2008

  6. [12]

    Capp´ e, A

    O. Capp´ e, A. Guillin, J.-M. Marin, and C. P . Robert. Pop ulation Monte Carlo. Journal of Computational and Graphical Statistics , 13(4):907– 929, 2004

    2004

  7. [13]

    G. H. Chen and R. T. Rockafellar. Convergence rates in fo rward– backward splitting. SIAM Journal on Optimization , 7(2):421–444, 1997

    1997

  8. [14]

    Chouzenoux, J.-C

    E. Chouzenoux, J.-C. Pesquet, and A. Repetti. V ariable metric forward- backward algorithm for minimizing the sum of a differentiab le function and a convex function. Journal of Optimization Theory and Applications , 162(1):107–132, July 2014

    2014

  9. [15]

    P . L. Combettes, D. Dung, and B. C. Vu. Proximity for sums of com- posite functions. Journal of Mathematical Analysis and Applications , 380(2):680–688, Aug. 2011

    2011

  10. [16]

    P . L. Combettes and J.-C. Pesquet. Proximal splitting m ethods in signal processing. In Fixed-point Algorithms for Inverse Problems in Science and Engineering , pages 185–212. Springer, 2011

    2011

  11. [17]

    P . L. Combettes and B. C. V˜ u. V ariable metric forward–b ackward split- ting with applications to monotone inclusions in duality. Optimization, 63(9):1289–1318, 2014

    2014

  12. [18]

    Corbineau, D

    M.-C. Corbineau, D. Kouam´ e, E. Chouzenoux, J.-Y . Tourneret, and J.-C. Pesquet. Preconditioned P-ULA for Joint Deconvolution-Se gmentation of Ultrasound Images. IEEE Signal Processing Letters , 26(10):1456– 1460, Oct. 2019

    2019

  13. [19]

    P . M. Djuric, J. H. Kotecha, J. Zhang, Y . Huang, T. Ghirma i, M. F. Bugallo, and J. Miguez. Particle filtering. IEEE Signal Processing Magazine, 20(5):19–38, 2003

    2003

  14. [20]

    Durmus, S

    A. Durmus, S. Majewski, and B. Miasojedow. Analysis of l angevin monte carlo via convex optimization. Journal of Machine Learning Research, 20(73):1–46, 2019

    2019

  15. [21]

    El-Laham and M

    Y . El-Laham and M. F. Bugallo. Stochastic gradient popu lation monte carlo. IEEE Signal Processing Letters , 27:46–50, 2019

    2019

  16. [22]

    El-Laham and M

    Y . El-Laham and M. F. Bugallo. Policy gradient importan ce sampling for bayesian inference. IEEE Trans. on Sig. Proc. , 69:4245–4256, 2021

    2021

  17. [23]

    El-Laham, P

    Y . El-Laham, P . M. Djuri´ c, and M. F. Bugallo. A variatio nal adap- tive population importance sampler. In ICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal P rocessing (ICASSP), pages 5052–5056. IEEE, 2019

    2019

  18. [24]

    El-Laham, V

    Y . El-Laham, V . Elvira, and M. Bugallo. Robust covarian ce adaptation in adaptive importance sampling. IEEE Signal Processing Letters , 25(8):1049–1053, 2018

    2018

  19. [25]

    Elvira and ´E

    V . Elvira and ´E. Chouzenoux. Langevin-based strategy for efficient proposal adaptation in population monte carlo. In Proceedings of the IEEE International Conference on Acoustics, Speech and Sig nal Proc. (ICASSP 2019) , pages 5077–5081, 2019

    2019

  20. [26]

    Elvira and E

    V . Elvira and E. Chouzenoux. Optimized population Mont e Carlo. IEEE Transactions on Signal Processing , 70:2489–2501, 2022

    2022

  21. [27]

    Elvira, E

    V . Elvira, E. Chouzenoux, ¨O. D. Akyildiz, and L. Martino. Gradient- based adaptive importance samplers. Journal of the Franklin Institute , 360(13):9490–9514, 2023

    2023

  22. [28]

    Elvira and L

    V . Elvira and L. Martino. Advances in importance sampli ng. arXiv preprint arXiv:2102.05407, 2021

    2021 arXiv

  23. [29]

    Elvira, L

    V . Elvira, L. Martino, D. Luengo, and M. F. Bugallo. Effic ient multiple importance sampling estimators. IEEE Signal Processing Letters , 22(10):1757–1761, 2015

    2015

  24. [30]

    Elvira, L

    V . Elvira, L. Martino, D. Luengo, and M. F. Bugallo. Here tical multiple importance sampling. IEEE Sig. Proc. Letters , 23(10):1474–1478, 2016

    2016

  25. [31]

    Elvira, L

    V . Elvira, L. Martino, D. Luengo, and M. F. Bugallo. Mult iple importance sampling with overlapping sets of proposals. In Proceedings of the IEEE W orkshop on Statistical Signal Processing (SSP 2016), 2016

    2016

  26. [32]

    Elvira, L

    V . Elvira, L. Martino, D. Luengo, and M. F. Bugallo. Impr oving population monte carlo: Alternative weighting and resampl ing schemes. Signal Processing, 131:77–91, 2017

    2017

  27. [33]

    Elvira, L

    V . Elvira, L. Martino, D. Luengo, and M. F. Bugallo. Popu lation Monte Carlo schemes with reduced path degeneracy. In Proceedings of the 7th IEEE International W orkshop on Computational Advances in Multi- Sensor Adaptive Processing (CAMSAP 2017) , pages 1–5, 2017

    2017

  28. [34]

    Elvira, L

    V . Elvira, L. Martino, D. Luengo, M. F. Bugallo, et al. Ge neralized multiple importance sampling. Statistical Science, 34(1):129–155, 2019

    2019

  29. [35]

    Elvira, L

    V . Elvira, L. Martino, D. Luengo, and J. Corander. A grad ient adaptive population importance sampler. In Proceedings of the IEEE International Conference on Acoustics, Speech and Signal P rocessing (ICASSP 2015) , pages 4075–4079, 2015

    2015

  30. [36]

    P . C. Encinar, F. R. Crucinio, and O. D. Akyildiz. Proxim al interacting particle langevin algorithms. arXiv preprint arXiv:2406.14292, 2024

    2024 arXiv

  31. [37]

    Fort and E

    G. Fort and E. Moulines. Stochastic variable metric pro ximal gradient with variance reduction for non-convex composite optimiza tion. Statis- tics and Computing , 33(3), Apr. 2023

    2023

  32. [38]

    J. V . Goldman, T. Sell, and S. S. Singh. Gradient-based m arkov chain Monte Carlo for Bayesian inference with non-differen tiable priors. Journal of the American Stat. Association , 117(540):2182–2193, 2022

    2022

  33. [39]

    Habring, M

    A. Habring, M. Holler, and T. Pock. Subgradient langevi n methods for sampling from nonsmooth potentials. SIAM Journal on Mathematics of Data Science , 6(4):897–925, 2024

    2024

  34. [40]

    Huang, E

    Y . Huang, E. Chouzenoux, V . Elvira, and J.-C. Pesquet. E fficient Bayes Inference in Neural Networks through Adaptive Importance S ampling. Journal of The Franklin Institute , 360(16):12125–12149, Sept. 2023

    2023

  35. [41]

    Kviman, H

    O. Kviman, H. Melin, H. Koptagel, V . Elvira, and J. Lager gren. Multiple importance sampling elbo and deep ensembles of var iational approximations. In International Conference on Artificial Intelligence and Statistics , pages 10687–10702. PMLR, 2022

    2022

  36. [43]

    H. Liao, X. Qian, J. Z. Huang, and P . Li. Rare event detect ion by acquisition-guided sampling. IEEE Transactions on Automation Science and Engineering , 2024

    2024

  37. [44]

    J. S. Liu and J. S. Liu. Monte Carlo strategies in scientific computing , volume 10. Springer, 2001

    2001

  38. [45]

    Luengo, L

    D. Luengo, L. Martino, M. Bugallo, V . Elvira, and S. S¨ ar kk¨ a. A survey of Monte Carlo methods for parameter estimation. EURASIP Journal on Advances in Signal Processing , 2020:1–62, 2020

    2020

  39. [46]

    Mateos, J

    G. Mateos, J. A. Bazerque, and G. B. Giannakis. Distribu ted sparse linear regression. IEEE Trans. on Sig. Proc. , 58(10):5262–5276, 2010

    2010

  40. [47]

    A. B. Owen. Monte Carlo theory, methods and examples . https://artowen.su.domains/mc/, 2013

    2013

  41. [48]

    Parikh and S

    N. Parikh and S. Boyd. Proximal algorithms. F oundations and Trends® in Optimization , 1(3):127–239, 2014

    2014

  42. [49]

    M. Pereyra. Proximal Markov chain Monte Carlo algorith ms. Statistics and Computing , 26(4):745–760, 2016

    2016

  43. [50]

    Repetti and Y

    A. Repetti and Y . Wiaux. V ariable metric forward-backw ard algorithm for composite minimization problems. SIAM Journal on Optimization , 31(2):1215–1241, 2021

    2021

  44. [51]

    C. P . Robert and G. Casella. Monte Carlo Statistical Methods . Springer Texts in Statistics. Springer-V erlag New Y ork, 1999

    1999

  45. [52]

    Schuster

    I. Schuster. Gradient importance sampling. arXiv preprint arXiv:1507.05781, 2015

    2015 arXiv

  46. [54]

    P . Tseng. A modified forward-backward splitting method for maximal monotone mappings. SIAM Journal on Control and Optimization , 38(2):431–446, 2000. 6 SUPLEMENTARY MATERIAL Dual forward-backward algorithm to compute the proximal st ep We summarize in Table S-I the iterations ...

    2000

  47. [55]

    Set iteration number J > 0

    [Initialization]: Set A ∈ Rdx×dx SDP , ˜ξ ∈ Rdx , and g ∈ Γ 0(Rdx ). Set iteration number J > 0. Set L = A1/ 2, ˜ζ = L−1 ˜ξ, ρ = ‖L‖2, ζ1 = L ˜ξ

  48. [56]

    , J]: ξj = ˜ζ − L⊤ζj ˜ζj = ζj + ρ−1Lξj ζj+1 = ˜ζj − ρ−1proxρg (ρ ˜ζj )

    [For j = 1, . . . , J]: ξj = ˜ζ − L⊤ζj ˜ζj = ζj + ρ−1Lξj ζj+1 = ˜ζj − ρ−1proxρg (ρ ˜ζj )

  49. [57]

    fig1.png

    [Output]: Return LξJ Additional results In Tables S-II and S-III, we provide the results for Examples A and B, respectively, u sing LR resampling strategy, instead of the GLR one, in step c)i) of PNAIS algorithm. Again, the propo sed method displays excellent performance on bo...

    arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.