Pith. sign in

REVIEW 3 major objections 3 minor 67 references

A tamed stochastic-gradient Hamiltonian Monte Carlo algorithm converges in Wasserstein-2 distance at rate λ^{1/4} for superlinearly growing, possibly discontinuous stochastic gradients, improving on the λ^{1/5} rate of its full-gradient tam

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

tSGHMC provably samples from strongly convex targets with superlinear, discontinuous stochastic gradients at a λ^{1/4} Wasserstein-2 rate.

T0 review reviewed 2026-08-02 challenge →

load-bearing objection New tamed SGHMC algorithm with real promise, but the advertised 1/4 Wasserstein rate does not follow from the paper's own theorem—the bound actually gives 3/8, which is still better than the 1/5 it improves on. the 3 major comments →

arxiv 2607.14862 v1 pith:3CXDX6KR submitted 2026-07-16 math.OC cs.NAmath.NAstat.ML

Tamed Stochastic Gradient Hamiltonian Monte Carlo

classification math.OC cs.NAmath.NAstat.ML MSC 60J2265C0590C15
keywords tamed stochastic gradient Hamiltonian Monte Carlosuperlinear gradientsdiscontinuous stochastic gradientsWasserstein-2 distancenon-asymptotic convergenceexpected excess riskdata-dependent strong convexity
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

This paper introduces tSGHMC, a tamed, stochastic-gradient version of Hamiltonian Monte Carlo aimed at sampling and optimization problems where gradients grow faster than linearly and may be discontinuous. Under a polynomial continuity-in-average condition and a data-dependent strong convexity condition, it proves a non-asymptotic Wasserstein-2 bound whose dominant error term is λ^{1/4}, improving the λ^{1/5} rate previously established for tamed kinetic Langevin Monte Carlo with exact gradients. The taming coefficient keeps iterates bounded without adding a full-gradient computation, so the method retains the low per-iteration cost of stochastic gradients. The same bound yields an expected-excess-risk estimate, giving an end-to-end guarantee for using tSGHMC samples to solve stochastic optimization problems. Experiments on posterior sampling, newsvendor, CVaR, and ReLU regression tasks show it outperforms the first-order tamed stochastic Langevin algorithm in most settings.

Core claim

The central claim is Theorem 1: for every ε>0 and β>0, as long as the step size λ is below a threshold λmax,γ and the friction γ is at least a problem-dependent γmin, the Wasserstein-2 distance between the law of tSGHMC's position θλ_n and the target πβ is bounded by Ċ(√(λγ)+λγ^{5/2}+γ^{-1}+exp(−λm/(2γ)n)W2(L(θ0,ν0),Πβ)+ε). When ε is set to λ^{1/4} and γ is chosen of order ε^{-1}, the bound becomes order λ^{1/4}, a faster rate than the λ^{1/5} obtained for tKLMC in [33]; equivalently, achieving ε precision requires Õ(ε^{-5}) iterations. This is the first such non-asymptotic guarantee for a stochastic-gradient kinetic Langevin sampler under superlinear and discontinuous gradients.

What carries the argument

The argument rests on the tamed stochastic gradient Hγ(θ,x) = mθ + (H(θ,x)−mθ)/(√(1+γ^{-1}|θ|^{4r})). This one coefficient simultaneously enforces a dissipativity condition, brings linear growth in θ, and keeps the deterministic drift close to the raw stochastic gradient H, which prevents the finite-time moment explosion that breaks plain Euler schemes. Around it the paper builds a chain: Moreau–Yosida regularization of u provides a globally Lipschitz, strongly convex surrogate; auxiliary continuous-time Langevin processes and a Lyapunov function control second and higher moments; and a contraction theorem for the regularized underdamped SDE supplies the exponential memory term in the bound.

Load-bearing premise

The proof collapses if Assumption 3 fails: the objective must be strongly convex with a known positive lower bound m on the smallest eigenvalue of E[A(X0)], because the taming coefficient uses that exact m; if m is unknown, overestimated, or the objective has a flat direction, the dissipativity, contraction, and moment bounds all cease to hold.

What would settle it

Run tSGHMC on a quartic potential u(θ)=|θ|^4 without a quadratic term, so E[A(X0)] has zero eigenvalue (m=0), and check whether W2(L(θλ_n),πβ) still decays; if it diverges or plateaus, Assumption 3 is essential. A milder check: in the newsvendor experiment (m=0.041, γ=1, r=1.5), λmax,γ is several orders of magnitude below the λ=0.05 used; test whether the log-log RMSE slope changes from the claimed λ^{1/4} once λ exceeds that threshold.

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

If this is right

  • With ε=λ^{1/4} and γ=O(ε^{-1}), Theorem 1 gives an Õ(ε^{-5}) iteration complexity to reach ε accuracy in Wasserstein-2 distance, improving on the Õ(ε^{-6}) complexity listed for tKLMC.
  • Theorem 2 converts the same bound into an expected excess risk guarantee E[u(θλ_n)]−u(θ*) ≤ ... + O(d/(mβ)); the β^{-1} term can be driven down by raising inverse temperature.
  • Stochastic-gradient usage means per-iteration cost is independent of dataset size, unlike tKLMC that uses exact gradients each step.
  • The assumptions cover discontinuous stochastic gradients (e.g., indicators in newsvendor and CVaR) and online data, so the algorithm is applicable to nonsmooth risk minimization.
  • Numerically, tSGHMC attains smaller RMSE and smaller expected excess risk than the first-order tamed stochastic Langevin algorithm across the tested problems, with comparable performance in the remaining cases.

Where Pith is reading between the lines

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

  • The rate gain comes specifically from scaling γ with λ (γ=O(λ^{-1/4})); an editor-level reading suggests the same parameter trade-off could be transplanted to other tamed kinetic samplers, including nonconvex variants, whenever a contraction estimate is available.
  • The dependency of the taming coefficient on the strong-convexity constant m means the proof does not currently cover the ReLU-network experiment, where m=10^{-10} is set arbitrarily; an adaptive or estimated m would be needed to bring that experiment under the theorem.
  • A testable prediction: in the regime λ ≪ λmax,γ the empirical Wasserstein-2 error should scale as λ^{1/4} on log-log plots; if instead the slope matches λ^{1/5} for small λ, the claimed improvement would not materialize in practice.
  • The bound is dimension- and temperature-dependent through Ċ=O((d/β)^{3r+1/2}), so the method is likely most attractive in moderate dimension or when β is large enough for concentration; extending the analysis to a log-Sobolev setting could decouple the rate from dimension.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 3 minor

Summary. The paper proposes tSGHMC (Eqs. (9)-(10)), a tamed stochastic-gradient kinetic Langevin algorithm for sampling and stochastic optimization with superlinearly growing and possibly discontinuous stochastic gradients. Under Assumptions 2-3 (polynomial continuity in average and data-dependent strong convexity), Theorem 1 gives a Wasserstein-2 error bound with terms sqrt(lambda gamma)+lambda gamma^(5/2)+gamma^{-1}+exp(...)+epsilon, and Theorem 2 converts this into an expected-excess-risk bound. The paper's central advertised claim is that choosing epsilon=lambda^(1/4) and gamma=O(lambda^{-1/4}) yields a W2 rate of lambda^(1/4), improving on the lambda^(1/5) rate of tKLMC [33]. Numerical experiments on penalized logistic regression, a quartic artificial example, newsvendor, CVaR minimization, and ReLU regression compare tSGHMC with TUSLA.

Significance. If the advertised rate were correct, this would be a valuable contribution: it extends tamed kinetic Langevin sampling to stochastic gradients under relaxed smoothness/curvature assumptions, provides explicit constants and a careful Moreau-Yosida and auxiliary-process proof architecture, and gives a non-asymptotic excess-risk guarantee. The paper is unusually explicit about constants and technical conditions, which is commendable. However, the central quantitative claim is internally inconsistent: the paper's own parameter choice in Theorem 1 (Eq. (17)) yields an Omega(lambda^(3/8)) bound, not O(lambda^(1/4)). The numerical experiments also violate the theorem's step-size restriction (Eq. (16)), and the ReLU regression experiment does not verify Assumption 3. The proof framework may still contain useful tools for a corrected analysis, but as written the main claim of an improved rate is not supported.

major comments (3)
  1. [Eq. (17), Remark 7] Substituting gamma=O(lambda^{-1/4}) and epsilon=lambda^(1/4) into (17) gives sqrt(lambda gamma)=Theta(lambda^(3/8)), lambda gamma^(5/2)=Theta(lambda^(3/8)), and gamma^{-1}=Theta(lambda^(1/4)). The right-hand side of (17) is therefore Omega(lambda^(3/8)), not O(lambda^(1/4)); the displayed bound in Remark 7, W2 <= 5Cdot(lambda^(1/4)+exp(...)W2(...)), simply drops these leading terms. The sentence in Section 1.1 that 'lambda gamma=O(lambda^(3/4)) ... leading to ... sqrt(lambda gamma)=O(lambda^(1/4))' is internally inconsistent, since sqrt(lambda gamma)=O(lambda^(3/8)) if lambda gamma=O(lambda^(3/4)). At best, Theorem 1 yields a lambda^(3/8) rate for this tuning, which is slower than the lambda^(1/5) rate of [33], reversing the claimed improvement. No cancellation between Propositions 4 and 5 can restore the lambda^(1/4) rate, since for this gamma Proposition 4 contains terms Theta(lambda^(
  2. [Sections 3.1-3.4, Eq. (16)] Every numerical experiment uses step sizes outside the theorem's validity domain lambda <= lambda_max,gamma. For example, in Section 3.1 (logistic regression, m=10^{-3}, gamma=1, r=0.5), q=12r=6 gives lambda_max,gamma = min(1/4, 3*10^{-3}/(16*6*5*2^4)) approximately 3.9*10^{-7}, while Figure 1 uses lambda up to 5*10^{-3}. In Section 3.3 lambda=0.05 with m=0.041, gamma=1, r=1.5 gives an even smaller lambda_max,gamma. Thus the RMSE plots and excess-risk trajectories cannot be cited as empirical validation of Theorem 1; they probe a regime in which the theorem is silent.
  3. [Section 3.4.2, Assumption 3] The ReLU nonlinear regression experiment sets m=10^{-10} but provides no argument that the objective (35) is m-strongly convex, or that E[A(X0)] has smallest eigenvalue at least m. For a one-hidden-layer ReLU network the objective is generically nonconvex; setting a tiny numerical value of m does not satisfy Assumption 3. Since m enters the taming coefficient (10), the step-size condition (16), and all moment/contraction estimates, Theorems 1 and 2 do not apply to this experiment as stated.
minor comments (3)
  1. [Section 3] Typo: 'vrious applications' should be 'various applications'.
  2. [Proposition 5 / Theorem 3 proof] The notation '16C0e/m' is easy to misread; use exp(1) or a different symbol to distinguish the constant e from the smoothing parameter epsilon.
  3. [Proof of Theorem 1] The step 'replace (n+1)T on both sides with n+1' is informal. Since lambda T >= 1/2, a short monotonicity/rounding argument would make the replacement rigorous.

Circularity Check

0 steps flagged

No circular derivation; the claimed λ^{1/4} rate is an internal arithmetic inconsistency rather than a circular reduction.

full rationale

Theorem 1's bound (17) is built from error terms √(λγ), λγ^{5/2}, γ^{-1}, the contraction term, and the Moreau-Yosida tolerance ε. Remark 7 chooses ε=λ^{1/4} and γ=O(ε^{-1}); substituting into (17) gives √(λγ)=Θ(λ^{3/8}), λγ^{5/2}=Θ(λ^{3/8}), and γ^{-1}=Θ(λ^{1/4}), so the right-hand side is Ω(λ^{3/8}) rather than O(λ^{1/4}). This is an internal mathematical inconsistency in the claimed rate, not a circular reduction: the rate is not manufactured by fitting outputs, and the proof's estimates do not define the target quantity in terms of itself. The heavy use of the authors' prior work ([33,38,39,44,49]) supplies technical lemmas and comparison baselines; those lemmas are used as tools (e.g., [38, Lemma 5.2] for continuous-time moment bounds) rather than as a substitute for the new discretization analysis. Assumption 3's m is both an assumption and an algorithm input, but that is a standard strong-convexity constant, not a fitted parameter renamed as a prediction. The experiments' step sizes violating λ≤λmax,γ and the unverified m for the ReLU example are correctness concerns, not circularity. Accordingly, no circular step is present; the paper is not self-referential in a load-bearing way, though the advertised 1/4 rate is not supported by its own equations.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 0 invented entities

The central proof rests on three domain assumptions and standard convex-analysis tools. The paper introduces no new particles, forces, or entities. The only hand-chosen numbers are algorithm hyperparameters (m,r,γ) and the proof parameter ε; none are fitted to numerical outputs to manufacture the rate. The step-size constraint λ≤λmax,γ is part of the theorem's hypotheses and is not satisfied in the experiments.

free parameters (4)
  • m (strong convexity constant) = 10^{-3}, 0.05, 0.041, 10^{-10} in different experiments
    Used inside the taming coefficient (10) and throughout the bounds. Chosen by hand per experiment; theory requires m to equal the smallest eigenvalue of E[A(X0)]. If mismatched, dissipativity (Lemma 14) and contraction (Theorem 3) collapse.
  • r (taming exponent) = 0.5, 1.5, 1 in experiments
    User-selected in (10); theory only requires 2r∈[ϑ,∞)∩N and r>0. The exponent determines moment orders and λmax scaling.
  • γ (friction coefficient) = 1, 3, 0.5 in experiments
    User-selected; theory requires γ≥γmin and the rate λ^{1/4} requires γ=O(λ^{-1/4}). Experiments use fixed γ=O(1), so the theoretical rate is not exercised in the numerics.
  • ε (Moreau-Yosida smoothing) = λ^{1/4}
    Chosen in Remark 7 to balance the regularization error with the discretization error. Not fitted to data; a pure proof parameter.
axioms (5)
  • domain assumption Assumption 1: finite 12r-th moments of (θ0,ν0) and X0; E[H(θ,X0)]=h(θ).
    Standard moment assumptions ensuring well-defined iterates and expectations; used in all moment estimates.
  • domain assumption Assumption 2: F is (2r-1)-polynomially Lipschitz in θ and ρ-polynomial in x; G satisfies a continuity-in-average condition with ϑ-growth.
    Core structural assumption that replaces global Lipschitzness; it drives the one-step error bounds and the taming effectiveness.
  • domain assumption Assumption 3: data-dependent strong convexity, E[A(X0)] ⪰ m I_d with m>0.
    Load-bearing: implies strong convexity of u (Remark 5), dissipativity (Remark 6), and exponential contraction of the underdamped SDE (Theorem 3). Without m>0 the central bound (17) is vacuous.
  • standard math Moreau-Yosida regularization properties (Lemma 1): 1/ε-Lipschitz gradient, preserves convexity and argmin.
    Borrowed from convex analysis; used to approximate non-globally Lipschitz h and to define πβ^ε.
  • standard math Contraction for kinetic Langevin with μ-strongly convex potential (Theorem 3, adapted from [18]).
    Provides the exponential decay term in the W2 bound under the MY-regularized potential.

reviewed 2026-08-02 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Tamed Stochastic Gradient Hamiltonian Monte Carlo." pith.science (2026). https://pith.science/paper/3CXDX6KR

@misc{pith2026260714862,
  author       = {Pith},
  title        = {Pith review of: Tamed Stochastic Gradient Hamiltonian Monte Carlo},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3CXDX6KR}},
  note         = {Machine review of arXiv:2607.14862}
}
Share X Bluesky LinkedIn Reddit HN
abstract

In this paper, we propose a novel tamed stochastic gradient Hamiltonian Monte Carlo (tSGHMC) algorithm for sampling and stochastic optimization problems with superlinearly growing stochastic gradients. Under a certain continuity in average condition and a strong convexity condition, we establish a non-asymptotic error bound in Wasserstein-2 distance for tSGHMC with the rate of convergence equal to $1/4$. Then, we derive an upper estimate for the associated expected excess risk, which provides a theoretical guarantee for the performance of tSGHMC. To illustrate the effectiveness of the proposed algorithm, we apply tSGHMC to practical examples, including a newsvendor problem and a Conditional Value-at-Risk minimization problem, using synthetic and real-world datasets. Numerical results support our theoretical findings. Furthermore, we compare tSGHMC with its first-order counterpart, namely, the tamed unadjusted stochastic Langevin algorithm. Simulation results demonstrate that tSGHMC achieves lower root mean square error and expected excess risk across a range of tasks.

discussion (0)

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

Reference graph

Works this paper leans on

67 extracted references · 1 canonical work pages

  1. [1]

    Nonasymptotic analysis of Stochastic Gradient Hamiltonian Monte Carlo under local conditions for nonconvex optimization.Journal of Machine Learning Research, 25(113):1–34, 2024

    O Deniz Akyildiz and Sotirios Sabanis. Nonasymptotic analysis of Stochastic Gradient Hamiltonian Monte Carlo under local conditions for nonconvex optimization.Journal of Machine Learning Research, 25(113):1–34, 2024

  2. [2]

    Shifted composition IV: toward ballistic acceleration for log-concave sampling

    Jason M Altschuler, Sinho Chewi, and Matthew S Zhang. Shifted composition IV: toward ballistic acceleration for log-concave sampling. InProceedings of the 58th Annual ACM Symposium on Theory of Computing, pages 1739–1750, 2026

  3. [3]

    Optimal inventory policy.Econometrica: Journal of the Econometric Society, pages 250–272, 1951

    Kenneth J Arrow, Theodore Harris, and Jacob Marschak. Optimal inventory policy.Econometrica: Journal of the Econometric Society, pages 250–272, 1951

  4. [4]

    Computing VaR and CVaR using stochas- tic approximation and adaptive unconstrained importance sampling.Monte Carlo Methods and Applications, 15(3):173–210, 2009

    Olivier Aj Bardou, Noufel Frikha, and Gilles Pagès. Computing VaR and CVaR using stochas- tic approximation and adaptive unconstrained importance sampling.Monte Carlo Methods and Applications, 15(3):173–210, 2009

  5. [5]

    On stochastic gradient Langevin dynamics with dependent data streams in the logconcave case.Bernoulli, 27(1):1–33, 2021

    Mathias Barkhagen, Ngoc Huy Chau, Éric Moulines, Miklós Rásonyi, Sotirios Sabanis, and Ying Zhang. On stochastic gradient Langevin dynamics with dependent data streams in the logconcave case.Bernoulli, 27(1):1–33, 2021

  6. [6]

    SIAM, 2017

    Amir Beck.First-order methods in optimization. SIAM, 2017

  7. [7]

    Stochastic C-stability and B-consistency of explicit and implicit Euler-type schemes.Journal of Scientific Computing, 67(3):955–987, 2016

    Wolf-Jürgen Beyn, Elena Isaak, and Raphael Kruse. Stochastic C-stability and B-consistency of explicit and implicit Euler-type schemes.Journal of Scientific Computing, 67(3):955–987, 2016

  8. [8]

    Stochastic gradient Hamiltonian Monte Carlo for non-convex learning.Stochastic Processes and their Applications, 149:341–368, 2022

    Huy N Chau and Miklós Rásonyi. Stochastic gradient Hamiltonian Monte Carlo for non-convex learning.Stochastic Processes and their Applications, 149:341–368, 2022

  9. [9]

    On stochastic gradient Langevin dynamics with dependent data streams: The fully nonconvex case.SIAM Journal on Mathematics of Data Science, 3(3):959–986, 2021

    Ngoc Huy Chau, Éric Moulines, Miklos Rásonyi, Sotirios Sabanis, and Ying Zhang. On stochastic gradient Langevin dynamics with dependent data streams: The fully nonconvex case.SIAM Journal on Mathematics of Data Science, 3(3):959–986, 2021

  10. [10]

    Stochastic gradient Hamiltonian Monte Carlo

    Tianqi Chen, Emily Fox, and Carlos Guestrin. Stochastic gradient Hamiltonian Monte Carlo. In International conference on machine learning, pages 1683–1691. PMLR, 2014

  11. [11]

    Convergence of Langevin MCMC in KL-divergence

    Xiang Cheng and Peter Bartlett. Convergence of Langevin MCMC in KL-divergence. InAlgorithmic learning theory, pages 186–211. PMLR, 2018

  12. [12]

    Sharp convergence rates for Langevin dynamics in the nonconvex setting.arXiv preprint arXiv:1805.01648, 2018

    Xiang Cheng, Niladri S Chatterji, Yasin Abbasi-Yadkori, Peter L Bartlett, and Michael I Jor- dan. Sharp convergence rates for Langevin dynamics in the nonconvex setting.arXiv preprint arXiv:1805.01648, 2018

  13. [13]

    Underdamped Langevin MCMC: A non-asymptotic analysis

    Xiang Cheng, Niladri S Chatterji, Peter L Bartlett, and Michael I Jordan. Underdamped Langevin MCMC: A non-asymptotic analysis. InConference on learning theory, pages 300–323. PMLR, 2018

  14. [14]

    Analysis of Langevin Monte Carlo from Poincare to log-Sobolev.Foundations of Computational Mathematics, 25(4):1345–1395, 2025

    Sinho Chewi, Murat A Erdogdu, Mufan Li, Ruoqi Shen, and Matthew S Zhang. Analysis of Langevin Monte Carlo from Poincare to log-Sobolev.Foundations of Computational Mathematics, 25(4):1345–1395, 2025

  15. [15]

    An explicit splitting SA V scheme for the kinetic Langevin dynamics.arXiv preprint arXiv:2509.04251, 2025

    Lei Dai, Yingsong Jiang, and Xiaojie Wang. An explicit splitting SA V scheme for the kinetic Langevin dynamics.arXiv preprint arXiv:2509.04251, 2025. 52 Z. W ANG AND Y . ZHANG

  16. [16]

    Further and stronger analogy between sampling and optimization: Langevin Monte Carlo and gradient descent

    Arnak Dalalyan. Further and stronger analogy between sampling and optimization: Langevin Monte Carlo and gradient descent. InConference on Learning Theory, pages 678–689. PMLR, 2017

  17. [17]

    Arnak S Dalalyan. Theoretical guarantees for approximate sampling from smooth and log-concave densities.Journal of the Royal Statistical Society Series B: Statistical Methodology, 79(3):651–676, 2017

  18. [18]

    Dalalyan and Lionel Riou-Durand

    Arnak S. Dalalyan and Lionel Riou-Durand. On Sampling from a Log-Concave Density Using Kinetic Langevin Diffusions.Bernoulli, 26(3), 2020. ISSN 1350-7265

  19. [19]

    Extreme M-quantiles as risk measures: FromL 1 toL p optimization.Bernoulli, 25(1):264–309, 2019

    Abdelaati Daouia, Stéphane Girard, and Gilles Stupfler. Extreme M-quantiles as risk measures: FromL 1 toL p optimization.Bernoulli, 25(1):264–309, 2019

  20. [20]

    Nonasymptotic convergence analysis for the unadjusted Langevin algorithm.The Annals of Applied Probability, 27(3):1551–1587, 2017

    Alain Durmus and Eric Moulines. Nonasymptotic convergence analysis for the unadjusted Langevin algorithm.The Annals of Applied Probability, 27(3):1551–1587, 2017

  21. [21]

    High-dimensional Bayesian inference via the unadjusted Langevin algorithm.Bernoulli, 25(4A):pp

    Alain Durmus and Eric Moulines. High-dimensional Bayesian inference via the unadjusted Langevin algorithm.Bernoulli, 25(4A):pp. 2854–2882, 2019

  22. [22]

    Efficient Bayesian computation by proximal Markov chain Monte Carlo: when Langevin meets Moreau.SIAM Journal on Imaging Sciences, 11 (1):473–506, 2018

    Alain Durmus, Eric Moulines, and Marcelo Pereyra. Efficient Bayesian computation by proximal Markov chain Monte Carlo: when Langevin meets Moreau.SIAM Journal on Imaging Sciences, 11 (1):473–506, 2018

  23. [23]

    CASdatasets: Insurance datasets, 2026

    Christophe Dutang and Arthur Charpentier. CASdatasets: Insurance datasets, 2026. URL https: //doi.org/10.57745/P0KHAG

  24. [24]

    Couplings and Quantitative Contraction Rates for Langevin Dynamics.The Annals of Probability, 47(4):1982–2010, 2019

    Andreas Eberle, Arnaud Guillin, and Raphael Zimmer. Couplings and Quantitative Contraction Rates for Langevin Dynamics.The Annals of Probability, 47(4):1982–2010, 2019

  25. [25]

    On the convergence of Langevin Monte Carlo: The interplay between tail growth and smoothness

    Murat A Erdogdu and Rasa Hosseinzadeh. On the convergence of Langevin Monte Carlo: The interplay between tail growth and smoothness. InConference on Learning Theory, pages 1776–1822. PMLR, 2021

  26. [26]

    A computer simulation of charged particles in solution

    Donald L Ermak. A computer simulation of charged particles in solution. I. Technique and equilibrium properties.The Journal of Chemical Physics, 62(10):4189–4196, 1975

  27. [27]

    Xuefeng Gao, Mert Gürbüzbalaban, and Lingjiong Zhu. Global convergence of stochastic gradient Hamiltonian Monte Carlo for nonconvex stochastic optimization: Nonasymptotic performance bounds and momentum-based acceleration.Operations Research, 70(5):2931–2947, 2022

  28. [28]

    Understanding the difficulty of training deep feedforward neural networks

    Xavier Glorot and Yoshua Bengio. Understanding the difficulty of training deep feedforward neural networks. InProceedings of the thirteenth international conference on artificial intelligence and statistics, pages 249–256. JMLR Workshop and Conference Proceedings, 2010

  29. [29]

    Approximating the nonlinear newsvendor and single-item stochastic lot-sizing problems when data is given by an oracle.Operations Research, 60 (2):429–446, 2012

    Nir Halman, James B Orlin, and David Simchi-Levi. Approximating the nonlinear newsvendor and single-item stochastic lot-sizing problems when data is given by an oracle.Operations Research, 60 (2):429–446, 2012

  30. [30]

    Martin Hutzenthaler, Arnulf Jentzen, and Peter E Kloeden. Strong and weak divergence in finite time of Euler’s method for stochastic differential equations with non-globally Lipschitz continuous coefficients.Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 467(2130):1563–1576, 2011

  31. [31]

    Martin Hutzenthaler, Arnulf Jentzen, and Peter E. Kloeden. Strong Convergence of an Explicit Numerical Method for SDEs with Nonglobally Lipschitz Continuous Coefficients.The Annals of Applied Probability, 22(4):1611–1641, 2012

  32. [32]

    Laplace’s method revisited: weak convergence of probability measures.The Annals of Probability, pages 1177–1182, 1980

    Chii-Ruey Hwang. Laplace’s method revisited: weak convergence of probability measures.The Annals of Probability, pages 1177–1182, 1980

  33. [33]

    Kinetic Langevin MCMC sampling without gradient Lipschitz continuity-the strongly convex case.Journal of Complexity, 85:101873, 2024

    Tim Johnston, Iosif Lytras, and Sotirios Sabanis. Kinetic Langevin MCMC sampling without gradient Lipschitz continuity-the strongly convex case.Journal of Complexity, 85:101873, 2024

  34. [34]

    Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980, 2014

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980, 2014

  35. [35]

    The newsboy problem with price-dependent demand distribution.IIE transactions, 20(2):168–175, 1988

    Amy Hing-Ling Lau and Hon-Shiang Lau. The newsboy problem with price-dependent demand distribution.IIE transactions, 20(2):168–175, 1988

  36. [36]

    Designing a quantity discount scheme for a newsvendor-type product with numerous heterogeneous retailers.European Journal of Operational Research, 180(2):585–600, 2007

    Amy Hing Ling Lau, Hon-Shiang Lau, and Jian-Cai Wang. Designing a quantity discount scheme for a newsvendor-type product with numerous heterogeneous retailers.European Journal of Operational Research, 180(2):585–600, 2007

  37. [37]

    Springer, 2016

    Jean-François Le Gall.Brownian motion, martingales, and stochastic calculus, volume 274. Springer, 2016. TAMED STOCHASTIC GRADIENT HAMILTONIAN MONTE CARLO 53

  38. [38]

    Luxu Liang, Ariel Neufeld, and Ying Zhang. Non-asymptotic convergence analysis of the sto- chastic gradient Hamiltonian Monte Carlo algorithm with discontinuous stochastic gradient with applications to training of ReLU neural networks.arXiv preprint arXiv:2409.17107, 2024

  39. [39]

    Dong-Young Lim, Ariel Neufeld, Sotirios Sabanis, and Ying Zhang. Non-asymptotic estimates for TUSLA algorithm for non-convex learning with applications to neural networks with ReLU activation function.IMA Journal of numerical analysis, 44(3):1464–1559, 2024

  40. [40]

    Langevin dynamics based algorithm e-THεO POULA for stochastic optimization problems with discontinuous stochastic gradient.Mathematics of Operations Research, 50(3):2333–2374, 2025

    Dong-Young Lim, Ariel Neufeld, Sotirios Sabanis, and Ying Zhang. Langevin dynamics based algorithm e-THεO POULA for stochastic optimization problems with discontinuous stochastic gradient.Mathematics of Operations Research, 50(3):2333–2374, 2025

  41. [41]

    An improved analysis of stochastic gradient descent with momentum.Advances in Neural Information Processing Systems, 33:18261–18271, 2020

    Yanli Liu, Yuan Gao, and Wotao Yin. An improved analysis of stochastic gradient descent with momentum.Advances in Neural Information Processing Systems, 33:18261–18271, 2020

  42. [42]

    Taming neural networks with TUSLA: Nonconvex learning via adaptive stochastic gradient Langevin algorithms.SIAM Journal on Mathematics of Data Science, 5(2):323–345, 2023

    Attila Lovas, Iosif Lytras, Miklós Rásonyi, and Sotirios Sabanis. Taming neural networks with TUSLA: Nonconvex learning via adaptive stochastic gradient Langevin algorithms.SIAM Journal on Mathematics of Data Science, 5(2):323–345, 2023

  43. [43]

    Contractive kinetic Langevin samplers beyond global Lipschitz continuity.arXiv preprint arXiv:2509.12031, 2025

    Iosif Lytras and Panayotis Mertikopoulos. Contractive kinetic Langevin samplers beyond global Lipschitz continuity.arXiv preprint arXiv:2509.12031, 2025

  44. [44]

    Taming under isoperimetry.Stochastic Processes and their Applications, 188:104684, 2025

    Iosif Lytras and Sotirios Sabanis. Taming under isoperimetry.Stochastic Processes and their Applications, 188:104684, 2025

  45. [45]

    Is there an analog of Nesterov acceleration for gradient-based MCMC?Bernoulli, 27(3): 1942–1992, 2021

    Yi-An Ma, Niladri S Chatterji, Xiang Cheng, Nicolas Flammarion, Peter L Bartlett, and Michael I Jordan. Is there an analog of Nesterov acceleration for gradient-based MCMC?Bernoulli, 27(3): 1942–1992, 2021

  46. [46]

    Estimating the tails of loss severity distributions using extreme value theory

    Alexander J McNeil. Estimating the tails of loss severity distributions using extreme value theory. ASTIN Bulletin: The Journal of the IAA, 27(1):117–137, 1997

  47. [47]

    Improved bounds for discretization of Langevin diffusions: Near-optimal rates without convexity.Bernoulli, 28(3): 1577–1601, 2022

    Wenlong Mou, Nicolas Flammarion, Martin J Wainwright, and Peter L Bartlett. Improved bounds for discretization of Langevin diffusions: Near-optimal rates without convexity.Bernoulli, 28(3): 1577–1601, 2022

  48. [48]

    Robust SGLD algorithm for solving non-convex distributionally robust optimisation problems.arXiv preprint arXiv:2403.09532, 2024

    Ariel Neufeld, Matthew Ng Cheng En, and Ying Zhang. Robust SGLD algorithm for solving non-convex distributionally robust optimisation problems.arXiv preprint arXiv:2403.09532, 2024

  49. [49]

    Non-asymptotic convergence bounds for modified tamed unadjusted Langevin algorithm in non-convex setting.Journal of Mathematical Analysis and Applications, 543(1):128892, 2025

    Ariel Neufeld, Matthew Cheng En Ng, and Ying Zhang. Non-asymptotic convergence bounds for modified tamed unadjusted Langevin algorithm in non-convex setting.Journal of Mathematical Analysis and Applications, 543(1):128892, 2025

  50. [50]

    Correlation functions and computer simulations.Nuclear Physics B, 180(3):378–384, 1981

    Giorgio Parisi. Correlation functions and computer simulations.Nuclear Physics B, 180(3):378–384, 1981

  51. [51]

    Stochastic processes and applications.Texts in applied mathematics, 60, 2014

    Grigorios A Pavliotis. Stochastic processes and applications.Texts in applied mathematics, 60, 2014

  52. [52]

    Pricing and the newsvendor problem: A review with extensions.Operations research, 47(2):183–194, 1999

    Nicholas C Petruzzi and Maqbool Dada. Pricing and the newsvendor problem: A review with extensions.Operations research, 47(2):183–194, 1999

  53. [53]

    Proximal mappings and Moreau envelopes of single-variable convex piecewise cubic functions and multivariable gauge functions

    Chayne Planiden and Xianfu Wang. Proximal mappings and Moreau envelopes of single-variable convex piecewise cubic functions and multivariable gauge functions. InNonsmooth optimization and its applications, pages 89–130. Springer, 2019

  54. [54]

    Non-convex learning via stochastic gradient Langevin dynamics: a nonasymptotic analysis

    Maxim Raginsky, Alexander Rakhlin, and Matus Telgarsky. Non-convex learning via stochastic gradient Langevin dynamics: a nonasymptotic analysis. InConference on Learning Theory, pages 1674–1703. PMLR, 2017

  55. [55]

    Springer, 1998

    R Tyrrell Rockafellar and Roger JB Wets.Variational analysis. Springer, 1998

  56. [56]

    A note on tamed Euler approximations.Electronic Communications in Probability, 18:1–10, 2013

    Sotirios Sabanis. A note on tamed Euler approximations.Electronic Communications in Probability, 18:1–10, 2013

  57. [57]

    A fully data-driven approach to minimizing CVaR for portfolio of assets via SGLD with discontinuous updating.arXiv preprint arXiv:2007.01672, 2020

    Sotirios Sabanis and Ying Zhang. A fully data-driven approach to minimizing CVaR for portfolio of assets via SGLD with discontinuous updating.arXiv preprint arXiv:2007.01672, 2020

  58. [58]

    Rapid convergence of the unadjusted Langevin algorithm: Isoperimetry suffices.Advances in neural information processing systems, 32, 2019

    Santosh Vempala and Andre Wibisono. Rapid convergence of the unadjusted Langevin algorithm: Isoperimetry suffices.Advances in neural information processing systems, 32, 2019

  59. [59]

    Springer, 2009

    Cédric Villani et al.Optimal transport: old and new, volume 338. Springer, 2009

  60. [60]

    Xiaojie Wang and Siqing Gan. The tamed Milstein method for commutative stochastic differential equations with non-globally Lipschitz continuous coefficients.Journal of Difference Equations and Applications, 19(3):466–490, 2013. 54 Z. W ANG AND Y . ZHANG

  61. [61]

    Bayesian learning via stochastic gradient Langevin dynamics

    Max Welling and Yee W Teh. Bayesian learning via stochastic gradient Langevin dynamics. In Proceedings of the 28th international conference on machine learning (ICML-11), pages 681–688, 2011

  62. [62]

    Global convergence of Langevin dynamics based algorithms for nonconvex optimization.Advances in Neural Information Processing Systems, 31, 2018

    Pan Xu, Jinghui Chen, Difan Zou, and Quanquan Gu. Global convergence of Langevin dynamics based algorithms for nonconvex optimization.Advances in Neural Information Processing Systems, 31, 2018

  63. [63]

    Modeling of strength of high-performance concrete using artificial neural networks

    I-C Yeh. Modeling of strength of high-performance concrete using artificial neural networks. Cement and Concrete research, 28(12):1797–1808, 1998

  64. [64]

    Langevin Monte Carlo for strongly log-concave distributions: Randomized midpoint revisited

    Lu Yu, Avetik Karagulyan, and Arnak Dalalyan. Langevin Monte Carlo for strongly log-concave distributions: Randomized midpoint revisited. InInternational Conference on Learning Representa- tions, volume 2024, pages 45515–45551, 2024

  65. [65]

    Dimension-Independent Convergence of Underdamped Langevin Monte Carlo in KL Divergence.arXiv preprint arXiv:2603.02429, 2026

    Shiyuan Zhang, Qiwei Di, Xuheng Li, and Quanquan Gu. Dimension-Independent Convergence of Underdamped Langevin Monte Carlo in KL Divergence.arXiv preprint arXiv:2603.02429, 2026

  66. [66]

    Im- proved discretization analysis for underdamped Langevin Monte Carlo

    Shunshi Zhang, Sinho Chewi, Mufan Li, Krishna Balasubramanian, and Murat A Erdogdu. Im- proved discretization analysis for underdamped Langevin Monte Carlo. InThe Thirty Sixth Annual Conference on Learning Theory, pages 36–71. PMLR, 2023

  67. [67]

    Nonasymptotic esti- mates for stochastic gradient Langevin dynamics under local conditions in nonconvex optimization

    Ying Zhang, Ömer Deniz Akyildiz, Theodoros Damoulas, and Sotirios Sabanis. Nonasymptotic esti- mates for stochastic gradient Langevin dynamics under local conditions in nonconvex optimization. Applied Mathematics & Optimization, 87(2):25, 2023. FINANCIALTECHNOLOGYTHRUST, SOCIETYHUB, THEHONGKONGUNIVERSITY OFSCIENCE ANDTECHNOLOGY (GUANGZHOU), GUANGZHOU, CHI...

This paper was first reviewed by deepseek-v4-flash on August 2, 2026.