Pith. sign in

REVIEW 3 major objections 4 minor 33 references

Zeroth-Order Langevin Monte Carlo via SPSA under Noisy Function Measurements

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Two noisy function evaluations per step are enough for Langevin Monte Carlo to converge in Wasserstein distance, with leading $W_2$ error $O(p^2)$ and oracle complexity $O(p/\epsilon^2+\delta^2p^3/\epsilon^3)$.

desk verdict The paper's advertised O(p^2) sharpening and oracle complexity rest on a false tensor bound and an inconsistent M1 scaling, but the SPSA-LMC algorithm and the diminishing-step analysis are worth a referee's time. read the letter →

arxiv 2608.07837 v1 pith:5ZRMADHV submitted 2026-08-08 math.OC math.STstat.COstat.TH

Hongbo Li , James C. Spall This is my paper · ORCID
classification math.OCmath.STstat.COstat.TH MSC 65C0590C56
keywords LMC-SPSAzeroth-ordersamplingsimultaneousperturbationstochasticapproximationnoisyfunctionevaluationsWassersteindistanceLangevinMonteCarlooraclecomplexitydimensiondependence
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper proposes LMC-SPSA, a sampler that replaces the gradient in Langevin Monte Carlo with a two-point simultaneous-perturbation estimate built from two noisy function evaluations per iteration. It tries to establish that, under strong convexity, smoothness, a third-derivative bound, and paired measurement noise, the sampler converges in 2-Wasserstein distance with an explicit non-asymptotic bound, including under a diminishing step-size and perturbation schedule. The central quantitative claim is a sharpened error bound whose dominant dimension dependence is $O(p^2)$ rather than the $O(p^4)$ of the earlier SPSA-LMC analysis, and a total noisy-oracle complexity of $O(p/\epsilon^2+\delta^2p^3/\epsilon^3)$ to reach $W_2$-accuracy $\epsilon$. If the claim is right, noisy black-box targets can be sampled almost as cheaply as targets with exact gradients, and the accuracy scaling improves over the one-point zeroth-order baseline while using only two oracle calls per step.

What carries the argument

The load-bearing object is the two-point SPSA gradient surrogate $\tilde G_k=\frac{\tilde f(X_k+c_k\Delta_k)-\tilde f(X_k-c_k\Delta_k)}{2c_k}\Delta_k^{-1}$, with $\Delta_k$ a Rademacher vector and $\tilde f$ the noisy oracle. It is conditionally nearly unbiased with bias $O(c_k^2)$, its projection-error second moment is exactly $(p-1)\mathbb{E}\|\nabla f(X_k)\|^2$ via the identity $\mathbb{E}[(\Delta\Delta^T-I)^2]=(p-1)I$, and its paired-noise variance is controlled by $\delta^2 p/(2c^2)$. The Wasserstein argument is carried by a one-step coupling between the chain and the Langevin diffusion: comparing $L_h-X_{k+1}$ with $L_0-X_k$ decomposes the error into drift contraction, Euler discretization remainder, SPSA projection error, third-derivative bias, and oracle noise. The explicit parameter choices then balance these five contributions equally so that each is $\epsilon/4$.

What would settle it

Take a strongly log-concave potential $f(x)=\frac{m}{2}\|x\|^2+\frac{c}{6}(\mathbf{1}^T x)^3$ on a bounded region where the Hessian stays positive definite, and consider the Rademacher perturbation with all coordinates equal to $1$. At that configuration the third-derivative contraction equals $c p^3$, whereas $M_3\|\Delta\|^3=M_3p^{3/2}$, so the inequality used to derive the $p^2$ term fails; a numerical version is to run LMC-SPSA on such a cubic-tilted target and plot $W_2/p^2$ versus $p$.

Watch

Extended reading notes

Core claim

The paper's central result is Theorem 3: under Assumptions A1-A5 the law of the LMC-SPSA chain satisfies $W_2(\nu_{K+1},\pi)\leq \prod_{j=0}^K(1-m h_j)W_2(\nu_0,\pi)+\frac{7\sqrt{2}M}{6m}\sqrt{h_K p}+\frac{M_3}{6m}c_K^2p^2+\sqrt{\frac{p-1}{m}}M_1\sqrt{h_K}+\sqrt{\frac{h_K p}{2m}}\frac{\delta}{c_K}$. The proof keeps the SPSA gradient estimator's projection error at size $(p-1)E\|\nabla f(X_k)\|^2$ through the exact identity $E[(\Delta\Delta^T-I)^2]=(p-1)I$ and compresses the third-derivative remainder to a $p^2$ term, replacing the $p^4$ term in the noise-free predecessor. Theorem 4 turns this bound into explicit choices of step size $h$, perturbation $c$, and iteration count $N$, giving the noisy-oracle complexity $O(p/\epsilon^2+\delta^2p^3/\epsilon^3)$; under perfectly paired noise $\delta=0$ this becomes $O(p/\epsilon^2)$. Theorems 5 and 6 extend the guarantee to diminishing schedules, with a balanced schedule yielding $W_2(\nu_k,\pi)=O(\delta^{2/3}pk^{-1/3})$ for fixed $\delta>0$.

Load-bearing premise

The $O(p^2)$ dimension improvement rests on the step in the proof of Theorem 3 that bounds the third-derivative remainder by $M_3\|\Delta\|^3$, as though Assumption A3's entrywise max-norm bound on third derivatives controlled the contraction norm of the tensor; if that step fails, the leading dimension dependence is larger than $p^2$.

Editorial extensions

If this is right

  • Under perfectly paired noise ($\delta=0$), the total oracle complexity for $W_2$-accuracy $\epsilon$ is $O(p/\epsilon^2)$, matching the iteration count of gradient LMC up to constants.
  • Under independent additive noise with variance $\sigma^2$ (so $\delta^2=\sigma^2$), the complexity is $O(p/\epsilon^2+\sigma^2p^3/\epsilon^3)$, improving the $\epsilon^{-4}$ scaling of the one-point ZO-LMC baseline at the cost of a larger dimension factor.
  • A diminishing schedule such as $h_k=a/(k+k_0)$ with balanced perturbations gives $W_2(\nu_k,\pi)=O(k^{-1/3})$ even when paired noise persists, so the sampler converges asymptotically without a constant-step bias.
  • Under a fixed oracle-call budget, the numerical section reports that LMC-SPSA achieves smaller mean-estimation errors and better dimension scaling than LMC-FDSA and one-point ZO-LMC, with an empirical mean-squared error scaling of roughly $p^{1.34}$.
  • The comparison with the one-point baseline is not uniform: LMC-SPSA wins on accuracy scaling but has higher dimension dependence, so which method is cheaper depends on the regime of $p$, $\epsilon$, and $\sigma$.

Reading between the lines

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

  • A stress test beyond the paper's quadratic example would run LMC-SPSA on a strongly log-concave target with a cubic component and plot $W_2/p^2$ against $p$; growth of the normalized error would indicate that the $O(p^2)$ dimension term is specific to the example rather than the general theorem.
  • The paired-noise assumption is the practical crux: when the two SPSA measurements cannot be paired, the $\delta^2p^3/\epsilon^3$ term dominates, so a concrete engineering follow-up is to impose common random numbers or antithetic noise across the two evaluations.
  • The same two-call SPSA surrogate should plug into kinetic Langevin or Hamiltonian Monte Carlo; the dimension-balancing logic here suggests the $O(p^2)$ term and the paired-noise benefit would carry over, but the paper does not analyze those samplers.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper proposes LMC-SPSA, an unadjusted Langevin Monte Carlo method that uses a two-point SPSA gradient surrogate built from two noisy function evaluations per iteration, for sampling from strongly log-concave targets when gradients are unavailable. The authors prove Wasserstein-2 convergence bounds under constant and diminishing step sizes, claim to sharpen the dominant dimension dependence in the error bound from O(p^4) to O(p^2), and derive an oracle complexity of O(p/epsilon^2 + delta^2 p^3/epsilon^3) under paired noise. They also provide numerical comparisons with finite-difference LMC and the one-point ZO-LMC of Roy et al. under matched oracle budgets.

Significance. If the theoretical claims were correct, the paper would contribute a query-efficient zeroth-order sampler with improved noise scaling relative to existing ZO-LMC methods, and the two-call-per-iteration SPSA estimator would be practically attractive. The paper also makes a good-faith effort at fair oracle-budget-matched numerical comparisons, and its diminishing-step-size analysis (Theorems 2, 5, 6) contains useful ideas. However, the two central advertised advances—the O(p^2) bound and the stated oracle complexity—are not supported by the proofs as written, and one complexity claim is inconsistent with the paper's own assumptions. The numerical experiment used to validate the O(p^2) sharpening does not actually exercise the third-derivative term that the sharpening concerns.

major comments (3)
  1. [Theorem 3, Eq. (42)] The key inequality |∇^3 f(X)(Δ⊗Δ⊗Δ)| ≤ M3 ||Δ||^3 is false under Assumption A3. A3 only bounds the max-norm of the third-derivative tensor, not its operator norm. For the tensor with every entry equal to M3, the i-th component of ∇^3 f(Δ⊗Δ⊗Δ) is M3(Σ_j Δ_j)^3; at Δ = (1,...,1) this equals M3 p^3, not M3 p^{3/2}. Thus the displayed bound in Eq. (42) is invalid, and the claimed O(p^2) third-derivative term is not established by the proof. A correct argument would need to control the L2 norm over the Rademacher randomness, for example via second moments of the Rademacher chaos of order three; no such estimate is supplied.
  2. [Theorem 4, Eqs. (48)-(50)] The stated oracle complexity O(p/epsilon^2 + delta^2 p^3/epsilon^3) is inconsistent with Assumption A4. The step-size choice in Eq. (48) contains the term m epsilon^2 / (16((p-1)M1^2 + ...)), so the iteration count N = Theta(1/(mh)) includes a contribution of order (p-1)M1^2/epsilon^2 from the projection-error term. Under the allowed scaling M1 = O(p), this term is Omega(p^3/epsilon^2); even under the natural M1 = O(sqrt(p)) it is Omega(p^2/epsilon^2). The proof of Theorem 4 explicitly writes 1/(mh) = O(1 + M^2p/(m^3 epsilon^2) + M1^2(p-1)/(m^2 epsilon^2) + delta^2 M3 p^3/(m^3 epsilon^3)) and then silently drops the M1 term when claiming Eq. (50). The paper's assertion that M1 = O(p) 'does not affect the leading-order dimension dependence' is therefore incorrect.
  3. [Section 6.1] The numerical validation of the O(p^2) dimension scaling uses the quadratic potential f(x) = (1/2)||x-2||^2, whose third derivative is identically zero. Consequently, the third-derivative remainder term being sharpened in Theorem 3 is absent in this experiment, so the observed flattening of W2/p^2 cannot provide evidence for the claimed sharpened bound. The experiment may only reflect the behavior of the projection-error and noise terms, which scale differently in p.
minor comments (4)
  1. [Section 4.4, Theorem 3 statement] The theorem statement says the bound holds for 'constant or shrinking h_k', but the displayed bound contains h_k and c_k only in the additive terms while the contraction factor is a product up to index k; for genuinely time-varying schedules the cumulative contribution of the remainders requires a summation argument as in Theorem 5. The statement should clarify that the displayed bound refers to a single-step or constant-step setting, or include the appropriate summed form.
  2. [Section 3] There is a duplicated sentence fragment: 'where∆k = (∆k1,...,∆kp)⊤ is a random perturbation vector.' appears twice in close succession. Please clean up the typesetting.
  3. [Section 5, Table 1 and surrounding text] The complexity comparison with Roy et al. suppresses the M1-dependent term in the LMC-SPSA row; since M1 may depend on p, the table's 'Total calls' column is misleading unless the dependence on M1 is shown or M1 is explicitly assumed constant.
  4. [Section 2, Fair-budget protocol] The numerical diagnostic is the squared error of the post-burn-in empirical mean (and moment errors), not a direct estimate of the Wasserstein-2 distance. The authors acknowledge this, but the abstract and introduction could be clearer that the empirical claims about W2 rely on the theoretical bounds, not on the plots.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: LMC-SPSA convergence is derived from stated assumptions via a standard one-step coupling argument; self-citations point to prior published lemmas and do not define the target result.

full rationale

The paper's central claims are a Wasserstein convergence bound and an oracle-complexity estimate for LMC-SPSA. The derivation chain is a standard LMC coupling argument: the SPSA gradient estimator is expanded by Taylor's theorem, its bias is controlled through the Rademacher expectation and Assumption A5, and the resulting one-step contraction is iterated with the help of known LMC lemmas. No parameter is fitted to data and then renamed a prediction; no quantity is defined in terms of the target Wasserstein error; no uniqueness theorem from the authors is invoked to rule out alternatives. The self-citations to [14] and [16] are to prior published work by Spall and coauthors, but their lemmas are stated as independent mathematical facts and are used as ingredients rather than as substitutes for the proof. The sharpened O(p^2) claim in Theorem 3 and the complexity claim in Theorem 4 may suffer from genuine mathematical gaps, such as the pointwise max-norm bound applied to the third-derivative tensor contraction and the treatment of M1=O(p) in the parameter choice; however, those are correctness concerns, not circularity. The reasoning does not reduce to its own inputs, so the appropriate circularity score is 0.

Assumptions & free parameters 3 free parameters · 9 assumptions · 0 invented entities

The central bounds rest on the strong-convexity, smoothness, and third-derivative assumptions, a uniform gradient-moment bound that the paper itself flags as nontrivial, paired-noise conditions, and several black-box lemmas from prior work. The p^2 sharpening additionally rests on an invalid tensor-contraction bound. No new physical entities are introduced.

free parameters (3)
  • stepsize h_k = constant h in Theorems 1/3/4; h_k = a/(k+k0) in Theorem 6; user-tuned in experiments
    Controls Euler discretization bias and contraction factor; not data-fitted, but the convergence and complexity statements depend on its schedule.
  • perturbation scale c_k = constant c, or c_k^2 = (9 m delta^2 h_k / (2 M3^2 p^3))^(1/3) in Theorem 6
    Balances SPSA bias of order c^2 against noise amplification of order delta^2/c^2; chosen by hand or by the balancing formula.
  • schedule constants a and k0 = a > 2/(3m); k0 >= max{1, a(m+M)/2}
    Additional degrees of freedom in the rate-optimal schedule; not informed by data.
assumptions (9)
  • domain assumption Strong convexity (A1)
    Used for the one-step contraction factor in W2 via Lemma 2 of [28]; stated in Section 4.1.
  • domain assumption Lipschitz gradient (A2)
    Used in the Euler discretization error bound (Lemma 4 of [28]); stated in Section 4.1.
  • domain assumption Bounded third derivative (A3)
    Used to bound SPSA bias; however Theorem 3's strengthening to a tensor-operator bound is not implied by this max-norm assumption.
  • domain assumption Uniform gradient expectation (A4)
    The paper itself remarks this is nontrivial; it feeds the projection-error term that the complexity claim appears to drop.
  • domain assumption Paired-noise condition (A5)
    Assumes mean-zero paired noise with bounded second moment and independence from the Gaussian innovation; defines delta in the complexity bound.
  • standard math Lemma 1 of [14]
    Black-box one-step Wasserstein recursion lemma used in Theorems 1, 3, 4; not reproved and co-authored by the present supervisor.
  • standard math Lemma 2 of [28]
    Contraction of exact LMC update under strong convexity; used in the one-step comparison.
  • standard math Lemma 4 of [28]
    Discretization error bound for the integrated drift along the diffusion; used in Eq (32).
  • ad hoc to paper Tensor-contraction bound in Theorem 3
    The inequality |nabla^3 f(X)(Delta (x) Delta (x) Delta)| <= M3 ||Delta||^3 is asserted to follow from A3; it is false for constant tensors and is the load-bearing step for the O(p^2) bound.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Zeroth-Order Langevin Monte Carlo via SPSA under Noisy Function Measurements." pith.science (2026). https://pith.science/paper/5ZRMADHV

@misc{pith2026260807837,
  author       = {Pith},
  title        = {Pith review of: Zeroth-Order Langevin Monte Carlo via SPSA under Noisy Function Measurements},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5ZRMADHV}},
  note         = {Machine review of arXiv:2608.07837}
}
abstract

In sampling problems, gradient-based schemes such as Langevin Monte Carlo (LMC) mix faster than non-gradient-based methods, but their applicability is limited by access to the gradient of the target log-density. In practice, gradients are often unavailable and function evaluations are noisy, e.g., stochastic simulators or black-box simulators, so we propose LMC-SPSA with noise, which approximates the gradient of the target log-density using two noisy function evaluations per iteration. We prove, under noisy gradient estimates, that LMC-SPSA converges in distribution by proving the convergence in Wasserstein distance. Furthermore, we construct a diminishing step-size schedule that still drives the Wasserstein error bound to convergence, extending convergence guarantees beyond the constant-step setting. Further, we sharpen the dominant dimension dependence of the Wasserstein error from $O(p^4)$ to $O(p^2)$ (with $p$ denoting the dimension), and support this analysis with numerical results. We show that LMC-SPSA achieves $W_2$-accuracy $\varepsilon$ with total noisy-oracle complexity of $O(p/\varepsilon^2+\delta^2p^3/\varepsilon^3)$, where $\delta$ is the paired-noise level. This improves the noise-dependent accuracy scaling relative to the ZO-LMC method of Roy et al. We further establish asymptotically vanishing Wasserstein error as the number of iterations $\to\infty$ under diminishing step-size and perturbation sequences and derive an explicit convergence rate for a balanced schedule under noisy zeroth-order feedback. Empirical experiments are conducted to verify the performance of LMC-SPSA with noise. We provide an oracle-budget-matched comparison with the ZO-LMC method, showing smaller empirical sampling errors under the same function-evaluation budget.

Figures

Figures reproduced from arXiv: 2608.07837 by the authors.

Figure 1
Figure 1. W2(νK, π)/p2 versus dimension p. Fluctuations at moderate p are due to finite￾iteration and noise effects. 6.2 Vanishing schedules and oracle efficiency under a fixed call budget A key implication of our refined nonasymptotic W2 bounds (Theorem 1/3) is that constant stepsize/perturbation generally induces a non-vanishing error floor (from Euler discretiza￾tion and gradient-estimation bias), whereas a properly chosen… view at source ↗
Figure 2
Figure 2. Oracle-budget comparison under shrinking schedules (truncated to [PITH_FULL_IMAGE:figures/full_fig_p024_2.png] view at source ↗
Figure 3
Figure 3. Scaling with dimension p under a fixed oracle-call budget B = 104 . For each method and dimension, the curves report the median over 20 independent runs after discarding the first half of each trajectory as burn-in. The first three panels show the absolute errors in the first/second/fourth moments of the first coordinate. The final panel shows the squared Euclidean error ∥µb − µ∥ 2 2 of the empirical mean. Across th… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 32 canonical work pages

  1. [1]

    Brooks, A

    S. Brooks, A. Gelman, G. L. Jones, and X.-L. Meng, Eds.,Handbook of Markov Chain Monte Carlo. Chapman & Hall/CRC, 2011. 30

    work page 2011

  2. [2]

    C. P. Robert and G. Casella,Monte Carlo Statistical Methods, 2nd ed. Springer, 2004

    2004

  3. [3]

    Equation of state calculations by fast computing machines,

    N. Metropolis, A. W. Rosenbluth, M. N. Rosenbluth, A. H. Teller, and E. Teller, “Equation of state calculations by fast computing machines,”Journal of Chemical Physics, vol. 21, no. 6, pp. 1087–1092, 1953

    work page 1953

  4. [4]

    Monte Carlo sampling methods using Markov chains and their applications,

    W. K. Hastings, “Monte Carlo sampling methods using Markov chains and their applications,”Biometrika, vol. 57, no. 1, pp. 97–109, 1970

    work page 1970

  5. [5]

    Simulation-based Inference via Langevin Dynamics with Score Matching

    H. Jiang, Y. Wang, and Y. Yang, “Simulation-based inference via Langevin dynamics with score matching,” arXiv preprint arXiv:2509.03853, 2025

    work page Pith review arXiv 2025

  6. [6]

    Particle Markov chain Monte Carlo methods,

    C. Andrieu, A. Doucet, and R. Holenstein, “Particle Markov chain Monte Carlo methods,”Journal of the Royal Statistical Society: Series B (Statistical Methodol- ogy), vol. 72, no. 3, pp. 269–342, 2010

    work page 2010

  7. [7]

    Stan: A probabilistic programming language,

    B. Carpenter, A. Gelman, M. D. Hoffman, D. Lee, B. Goodrich, M. Betancourt, M. Brubaker, J. Guo, P. Li, and A. Riddell, “Stan: A probabilistic programming language,”Journal of Statistical Software, vol. 76, no. 1, pp. 1–32, 2017

    work page 2017

  8. [8]

    MrBayes 3: Bayesian phylogenetic inference under mixed models,

    F. Ronquist and J. P. Huelsenbeck, “MrBayes 3: Bayesian phylogenetic inference under mixed models,”Bioinformatics, vol. 19, no. 12, pp. 1572–1574, 2003

    work page 2003

Show all 33 references
  1. [9]

    MCMC using Hamiltonian dynamics,

    R. M. Neal, “MCMC using Hamiltonian dynamics,” inHandbook of Markov Chain Monte Carlo, S. Brooks, A. Gelman, G. L. Jones, and X.-L. Meng, Eds. Chapman & Hall/CRC, 2011, ch. 5, pp. 113–162

    2011

  2. [10]

    Bayesian learning via stochastic gradient Langevin dynamics,

    M. Welling and Y. W. Teh, “Bayesian learning via stochastic gradient Langevin dynamics,” inProceedings of the 28th International Conference on Machine Learning (ICML), 2011, pp. 681–688

    2011

  3. [11]

    Exponential convergence of Langevin distribu- tions and their discrete approximations,

    G. O. Roberts and R. L. Tweedie, “Exponential convergence of Langevin distribu- tions and their discrete approximations,”Bernoulli, vol. 2, no. 4, pp. 341–363, 1996

    1996

  4. [12]

    Optimal scaling of discrete approximations to Langevin diffusions,

    G. O. Roberts and J. S. Rosenthal, “Optimal scaling of discrete approximations to Langevin diffusions,”Journal of the Royal Statistical Society: Series B (Statistical Methodology), vol. 60, no. 1, pp. 255–268, 1998

    1998

  5. [13]

    Underdamped Langevin MCMC: A non-asymptotic analysis,

    X. Cheng, N. S. Chatterji, P. L. Bartlett, and M. I. Jordan, “Underdamped Langevin MCMC: A non-asymptotic analysis,” inProceedings of the 31st Conference on Learn- ing Theory (COLT), 2018, pp. 300–323

    2018

  6. [14]

    Langevin Monte Carlo with SPSA-approximated gradients,

    S. Sun and J. C. Spall, “Langevin Monte Carlo with SPSA-approximated gradients,” inProceedings of the 57th Annual Conference on Information Sciences and Systems (CISS), 2023, pp. 1–6

    2023

  7. [15]

    Numerical differentiation of noisy, nonsmooth data,

    R. Chartrand, “Numerical differentiation of noisy, nonsmooth data,”ISRN Applied Mathematics, vol. 2011, Art. ID 164564, 11 pp., 2011

    2011

  8. [16]

    Multivariate stochastic approximation using a simultaneous perturba- tion gradient approximation,

    J. C. Spall, “Multivariate stochastic approximation using a simultaneous perturba- tion gradient approximation,”IEEE Transactions on Automatic Control, vol. 37, no. 3, pp. 332–341, 1992

    1992

  9. [17]

    A comparison of the finite difference and simultaneous per- turbation gradient estimation methods with noisy function evaluations,

    A. Blakney and J. Zhu, “A comparison of the finite difference and simultaneous per- turbation gradient estimation methods with noisy function evaluations,” inProceed- ings of the 53rd Annual Conference on Information Sciences and Systems (CISS), 2019, pp. 1–6. 31

    2019

  10. [18]

    Developments in stochastic optimization algorithms with gradient ap- proximations based on function measurements,

    J. C. Spall, “Developments in stochastic optimization algorithms with gradient ap- proximations based on function measurements,” inProceedings of the 1994 Winter Simulation Conference, 1994, pp. 207–214

    1994

  11. [19]

    J. C. Spall,Introduction to Stochastic Search and Optimization: Estimation, Simu- lation, and Control. John Wiley & Sons, 2003

    2003

  12. [20]

    Theoretical study and comparison of SPSA and RDSA algorithms,

    Y. Chen, “Theoretical study and comparison of SPSA and RDSA algorithms,” arXiv preprint arXiv:2107.12771, 2021

    2021 arXiv

  13. [21]

    Formal comparison of simultaneous perturba- tion stochastic approximation and random direction stochastic approximation,

    D. Peng, Y. Chen, and J. C. Spall, “Formal comparison of simultaneous perturba- tion stochastic approximation and random direction stochastic approximation,” in Proceedings of the 2023 American Control Conference (ACC), 2023, pp. 744–749

    2023

  14. [22]

    Langevin Monte Carlo: random coordinate descent and variance reduction,

    Z. Ding and Q. Li, “Langevin Monte Carlo: random coordinate descent and variance reduction,”Journal of Machine Learning Research, vol. 22, no. 205, pp. 1–51, 2021

    2021

  15. [23]

    A one-measurement form of simultaneous perturbation stochastic ap- proximation,

    J. C. Spall, “A one-measurement form of simultaneous perturbation stochastic ap- proximation,”Automatica, vol. 33, no. 1, pp. 109–112, 1997

    1997

  16. [24]

    One-point gradient estimators for zeroth-order stochastic gra- dient Langevin dynamics,

    L. Liu and Z. Wang, “One-point gradient estimators for zeroth-order stochastic gra- dient Langevin dynamics,” inOPT 2020: 12th Annual Workshop on Optimization for Machine Learning, 2020. [Online]. Available:https://opt-ml.org/oldopt/p apers/2020/paper_96.pdf

    2020

  17. [25]

    Derivative-free Bayesian inversion using multiscale dynamics,

    G. A. Pavliotis, A. M. Stuart, and U. Vaes, “Derivative-free Bayesian inversion using multiscale dynamics,”SIAM Journal on Applied Dynamical Systems, vol. 21, no. 1, pp. 284–326, 2022

    2022

  18. [26]

    Unlocking the power of differentially private zeroth-order optimization for fine-tuning LLMs,

    E. Bao, Y. Jiang, F. Wei, X. Xiao, Z. Li, Y. Li, and B. Ding, “Unlocking the power of differentially private zeroth-order optimization for fine-tuning LLMs,” inProceedings of the 34th USENIX Security Symposium (USENIX Security 25), 2025, pp. 1569– 1588

    2025

  19. [27]

    Generative modeling by estimating gradients of the data distribution,

    Y. Song and S. Ermon, “Generative modeling by estimating gradients of the data distribution,” inAdvances in Neural Information Processing Systems, vol. 32, 2019, pp. 11895–11907

    2019

  20. [28]

    User-friendly guarantees for the Langevin Monte Carlo with inaccurate gradient,

    A. S. Dalalyan and A. Karagulyan, “User-friendly guarantees for the Langevin Monte Carlo with inaccurate gradient,”Stochastic Processes and their Applications, vol. 129, no. 12, pp. 5278–5311, 2019

    2019

  21. [29]

    Hiruntiaranakul,SPSA-augmented Hamiltonian Monte Carlo, Master’s thesis, Johns Hopkins University, Baltimore, MD, USA, 2024

    S. Hiruntiaranakul,SPSA-augmented Hamiltonian Monte Carlo, Master’s thesis, Johns Hopkins University, Baltimore, MD, USA, 2024. [Online]. Available:https: //jscholarship.library.jhu.edu/bitstreams/fd72f4af-722d-4aed-afcb-94e a9d69616e/download

    2024

  22. [30]

    Stochastic zeroth-order discretizations of Langevin diffusions for Bayesian inference,

    A. Roy, L. Shen, K. Balasubramanian, and S. Ghadimi, “Stochastic zeroth-order discretizations of Langevin diffusions for Bayesian inference,”Bernoulli, vol. 28, no. 3, pp. 1810–1834, 2022

    2022

  23. [31]

    Preliminaryresultsonnon-Bernoullidistributionofperturbationsforsimul- taneous perturbation stochastic approximation,

    X.Cao, “Preliminaryresultsonnon-Bernoullidistributionofperturbationsforsimul- taneous perturbation stochastic approximation,” inProceedings of the 2011 Ameri- can Control Conference (ACC), 2011, pp. 2669–2670

    2011

  24. [32]

    Learning rules for neuro-controller via si- multaneous perturbation,

    Y. Maeda and R. J. P. de Figueiredo, “Learning rules for neuro-controller via si- multaneous perturbation,”IEEE Transactions on Neural Networks, vol. 8, no. 5, pp. 1119–1130, 1997. 32

    1997

  25. [33]

    Nonasymptotic convergence analysis for the un- adjusted Langevin algorithm,

    A. Durmus and E. Moulines, “Nonasymptotic convergence analysis for the un- adjusted Langevin algorithm,”The Annals of Applied Probability, vol. 27, no. 3, pp. 1551–1587, 2017. 33

    2017

Pith tools

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