Pith. sign in

REVIEW 1 major objections 5 minor 2 cited by

Hypocoercivity properties of adaptive Langevin dynamics

T0 review · 1 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read Adaptive Langevin sampling converges exponentially fast to its target measure.

desk verdict Solid hypocoercivity proof for Adaptive Langevin with an honest parameter-dependent rate; the main gap is the idealized state-independent noise model that limits how far the theorem reaches into stochastic-gradient MCMC. read the letter →

arxiv 1908.09363 v3 pith:OEW3W7KB submitted 2019-08-25 math.PR cs.NAmath.FAmath.NAstat.CO

Benedict Leimkuhler , Matthias Sachs , Gabriel Stoltz This is my paper · ORCID
classification math.PRcs.NAmath.FAmath.NAstat.CO MSC 60J7035B4046N3035Q8465C30
keywords AdaptiveLangevindynamicshypocoercivityexponentialconvergencecentrallimittheoremstochasticgradientnoiseBayesianposteriorsamplingspectralgapthermostat
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 Langevin dynamics (ALD) is a sampling scheme that promotes the friction coefficient to a dynamical variable updated by a feedback control law, so that unknown noise corrupting the gradient of a target potential does not destroy the invariant measure. This paper establishes that, for smooth target measures with a spectral-gap inequality on the position marginal and gradient noise of additive white-noise type, the law of ALD converges exponentially fast in $L^2(\pi)$ to the target measure. The exponential rate is explicitly bounded below by a constant times $\min(\gamma, \gamma^{-1}, \gamma\varepsilon^2, (\gamma\varepsilon^2)^{-1})$, where $\gamma$ is the total noise intensity and $\varepsilon=\sqrt{\nu}$ is the square root of the thermostat mass. A central limit theorem for single-trajectory time averages follows, with asymptotic variance bounded by the reciprocal of that rate. The result matters because ALD is used in Bayesian inference with stochastic minibatch gradients, where previously no quantitative $L^2$ convergence rate was available.

What carries the argument

The argument is carried by a modified norm $\mathcal{H}(\phi)=\tfrac12\|\phi\|^2_{L^2(\pi)}+a_{\varepsilon,\gamma}\langle A_\varepsilon\phi,\phi\rangle_{L^2(\pi)}$, where $A_\varepsilon$ is a regularizing operator constructed from the antisymmetric part of the generator, that is, the Hamiltonian part plus the scaled feedback-control part. The definition of $A_\varepsilon$ switches at $\varepsilon=1$: for $\varepsilon\le 1$ it regularizes the Hamiltonian transport, while for $\varepsilon>1$ the generator is first multiplied by $\varepsilon$ and the feedback part is regularized instead. This switch is what makes the partial coercivity estimate hold with the correct $\varepsilon$-dependent rate. The resulting coercivity of $-L_{\mathrm{AdL}}$ in the modified scalar product gives the exponential decay via a differential-inequality argument, and the same coercivity bounds the inverse generator, which is the input to the central limit theorem.

What would settle it

On the quadratic potential $U(q)=q^2/2$, compute the spectral gap of a finite spectral discretization of $-L_{\mathrm{AdL}}$ along a curve with $\gamma\varepsilon^2$ fixed and $\varepsilon\to0$; Theorem 2.1 requires the gap to stay at least a fixed fraction of $\gamma\varepsilon^2$, so a gap decaying faster than that bound would falsify the rate. Separately, replacing the additive gradient noise in a simulation by state-dependent noise of the same mean and comparing the empirical stationary distribution with $e^{-\beta U}$ would test whether the invariant-measure assumption carries over.

Watch

Extended reading notes

Core claim

The central claim is Theorem 2.1: for any $\varepsilon,\gamma>0$, the semigroup $e^{tL_{\mathrm{AdL}}}$ contracts in $L^2(\pi)$ toward $\int\phi\,d\pi$ with rate at least $\lambda\min(\gamma,\gamma^{-1},\gamma\varepsilon^2,(\gamma\varepsilon^2)^{-1})$. The proof exhibits a modified norm in which the generator is coercive, and then applies a differential-inequality estimate. The invertibility of the generator on mean-zero functions follows, and with it a functional central limit theorem: $\sqrt{t}(\hat\phi_t-\mathbb{E}_\pi\phi)$ converges to a centered normal distribution with variance $2\langle -L_{\mathrm{AdL}}^{-1}\Pi_0\phi,\Pi_0\phi\rangle_{L^2(\pi)}$, bounded above by $O(\max(\gamma,\gamma^{-1},\gamma\varepsilon^2,(\gamma\varepsilon^2)^{-1}))$. In the large-thermostat-mass limit $\varepsilon\to\infty$, the asymptotic variance approaches the underdamped Langevin variance at order $1/\varepsilon$, so the added thermostat does not degrade efficiency in that regime. Numerical experiments on a double-well model and on Bayesian logistic regression for handwritten-digit classification confirm the predicted scalings and the normal limit law.

Load-bearing premise

The load-bearing premise is that gradient corruption is an additive white-noise term with constant covariance matrix $\sigma_G^2 I$, independent of position; the invariant measure and the whole generator decomposition are built on that form, and real minibatch gradient noise only approximates it.

Editorial extensions

If this is right

  • Users of ALD can estimate posterior expectations from one long trajectory and attach asymptotically valid normal error bars, provided the gradient noise is approximately additive and white.
  • The rate formula identifies the dangerous regimes: mixing degrades as $\gamma\to0$ or $\gamma\to\infty$, and as $\nu=\varepsilon^2\to\infty$ or $\nu\to0$, with the precise power law $\min(\gamma,\gamma^{-1},\gamma\varepsilon^2,(\gamma\varepsilon^2)^{-1})$.
  • The inverse-generator bound of order $\max(\gamma,\gamma^{-1},\gamma\varepsilon^2,(\gamma\varepsilon^2)^{-1})$ tells practitioners how long to integrate: of order that maximum will bring the variance of a time average to order one.
  • As $\varepsilon\to\infty$, ALD inherits the sampling efficiency of underdamped Langevin dynamics, since the asymptotic variance converges to the Langevin value at rate $1/\varepsilon$.

Reading between the lines

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

  • In the actual minibatch setting the gradient residual is state-dependent and only approximately normal, so the theorem does not by itself certify the Bayesian posterior; a natural test would be to measure the bias introduced by a state-dependent $\sigma_G^2(q)$ and compare with the white-noise idealization.
  • The $\varepsilon\to0$ singular limit, where the feedback dynamics dominates on level sets of $\Phi=\xi^2+|p|^2-(2n/\beta)\ln|p|^2$, suggests a homogenized slow description that could yield sharp asymptotic variance predictions; the paper does not carry this out.
  • The same modified-norm machinery should transfer to other thermostatted sampling schemes whose generator splits into symmetric and antisymmetric parts, giving explicit rates in terms of their parameters.
  • A quantitative check of the theorem's boundary: increasing minibatch size lowers gradient-noise variance; the theory predicts the rate improves as $\gamma$ decreases toward order one, which is testable on a fixed data set.
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

1 major / 5 minor

Summary. The paper analyzes the Adaptive Langevin dynamics SDE, a sous-thermostatted Langevin system with an additional gradient-noise term of unknown, constant intensity. After rescaling the thermostat variable, the authors decompose the generator and prove (Theorem 2.1) exponential convergence in L^2(pi) for the semigroup, with a rate lower bound proportional to min(gamma, 1/gamma, gamma epsilon^2, 1/(gamma epsilon^2)). They then infer invertibility of the generator on the mean-zero subspace and derive a central limit theorem for ergodic averages with bounds on the asymptotic variance. A companion result (Proposition 3.1) shows that, for observables depending only on (q,p), the asymptotic variance converges to that of standard underdamped Langevin dynamics as epsilon -> infinity. The theoretical scaling is tested numerically by Galerkin computation of the spectral gap, by simulation of asymptotic variances, and by a Bayesian logistic-regression example on MNIST using a minibatch gradient estimator.

Significance. If the results are correct, this is a substantial contribution: it gives the first quantitative hypocoercivity estimate for Adaptive Langevin dynamics with explicit dependence on the friction and thermostat-mass parameters, and it upgrades earlier qualitative results (e.g., Herzog's Lyapunov-function analysis) to an L^2 rate and a CLT. The proof is self-contained and analytic, with no fitted parameters; the derived scaling is concrete and falsifiable, and the Galerkin numerics in Section 4.1 support the predicted exponents. The Bayesian-sampling application illustrates the intended use, although the bridge from the idealized model to the minibatch setting requires a modeling assumption that is not exactly satisfied.

major comments (1)
  1. [Section 2.4, Proposition 2.2 and equation (2.13)] The statement of Proposition 2.2 fixes the regularization parameter as a_{\varepsilon,\gamma} = a \min(\gamma, \gamma^{-1}, \gamma\varepsilon^2, (\gamma\varepsilon^2)^{-1}), but the proof in the paragraph "For \varepsilon > 1" instead chooses a_{\varepsilon,\gamma} = a \min(\gamma\varepsilon, (\gamma\varepsilon)^{-1}). These two choices differ by a factor \varepsilon in the regime \varepsilon > 1: for \gamma\varepsilon < 1 the stated formula gives a\gamma while the proof uses a\gamma\varepsilon, and for \gamma\varepsilon > 1 it gives a/(\gamma\varepsilon^2) while the proof uses a/(\gamma\varepsilon). With the stated formula, the determinant condition (2.24) yields a lower bound of order \gamma/\varepsilon (respectively 1/(\gamma\varepsilon^3)) instead of \gamma (respectively 1/(\gamma\varepsilon^2)) in these regimes, so the uniform constant \tilde{\lambda} in Proposition 2.2 cannot exist. The proposition statement must be corrected, for example by giving the piecewise definition of a_{\varepsilon,\gamma} that is actually used in the proof, so that the statement and proof are consistent.
minor comments (5)
  1. [Appendix C.2, ODABADO algorithm] In the last momentum update of the ODABADO algorithm, the text reads p_{k+1} = \alpha(\zeta_{k+1}, \Delta t/2) p_k + G(\sigma_A, \zeta_{k+1}, \Delta t/2) R_{k+1/2}; for consistency with the symmetric splitting (C.12), the argument should be \hat{p}_{k+1/2} (the value after the gradient-noise substep), not p_k.
  2. [Section 4.3 and Appendix C.2] The numerical CLT demonstration for Bayesian logistic regression uses a minibatch gradient estimator, whose residual noise is state-dependent and only approximately Gaussian, whereas the ODABADO integrator is derived under the assumption in Appendix C.2 that the residual is Gaussian and independent of q_k. The text should state explicitly in Section 4.3 that the experiment is an empirical check under this modeling assumption and is not a direct consequence of Corollary 2 for the actual subsampled process.
  3. [Section 3, Corollaries 1 and 2] The word "Collorary" appears in place of "Corollary" in the headings/callouts for Corollary 1 and Corollary 2; also "trajecotry" appears in Section 4.2 and "different" appears in Section 4.2.
  4. [Section 3 title] The section title "Pathwise ergodicity and functional central limit theorem" promises a functional CLT, but only the ordinary CLT is stated in Corollary 2; either add the functional CLT or change the title.
  5. [Equation (3.5)] In the solvability condition (3.5), the notation \Pi_0\phi is ambiguous: it should be made clear that the integral is a partial average over (q,p), which vanishes because \phi is restricted to functions of (q,p) only in Proposition 3.1; otherwise it can be mistaken for the global projection defined in (2.8).

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the exponential convergence theorem is derived analytically from the generator decomposition and the external hypocoercivity framework, with numerical experiments testing the derived scalings.

full rationale

The paper's central claim (Theorem 2.1) is an analytically derived L2 hypocoercivity estimate for the adaptive Langevin SDE (2.4). The proof chain is self-contained: the generator is decomposed as L_AdL = L_H + gamma L_O + epsilon^{-1} L_NH; the invariant measure (1.3) is verified by direct adjoint computations; the modified norm H and regularization operator A_epsilon are constructed from that decomposition; Proposition 2.2 establishes coercivity, and Theorem 2.1 follows by a Grönwall argument. No parameter is fitted to data, and the rate lower bound (2.12) follows from the proof rather than being imposed. The numerical sections test the predicted scalings (e.g., the spectral gap in the Hermite Galerkin subspace and the CLT residuals) rather than tune parameters to force agreement. Citations to the authors' earlier work [16, 33, 22] introduce the dynamics and numerical integrators, but they do not carry the convergence proof; the key external inputs are the general hypocoercivity framework of Dolbeault–Mouhot–Schmeiser and the CLT criterion of Bhattacharya. The acknowledged assumption in Appendix C.2 that minibatch gradient noise is Gaussian and state-independent is a limitation on the application to subsampled Bayesian inference, not a circular step in the mathematical derivation.

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

No free parameters are fitted; gamma and epsilon are model parameters. The central claim rests on the white-noise model of gradient error and on Assumption 1. The paper introduces no new physical entities; the auxiliary friction variable zeta is part of the Nose-Hoover-based method from prior work.

assumptions (4)
  • domain assumption Assumption 1: U is smooth, pi_q satisfies a Poincare inequality, and Delta U <= c1 + (c2/2)|grad U|^2 with c2 in [0,1), |grad^2 U| <= c3(1+|grad U|).
    Used throughout the hypocoercivity proof, e.g., for boundedness of partial^2_{q_i,q_j}(1-P_q)G_epsilon in Lemma 2.9, relying on [9, Section 3].
  • domain assumption Gradient perturbation is an additive Ito white-noise process sigma_G dW_G with constant sigma_G.
    Defines the SDE (1.2) and the invariant measure (1.3)-(1.4); see Section 1 and Appendix C.2. Without this, the adaptive method does not sample the exact target for general state-dependent gradient noise.
  • standard math Technical boundedness estimates from [9, Section 3] for operators such as partial^2_{q_i,q_j}(1-P_q)(1+Pi grad*_q grad_q Pi)^-1.
    Invoked in the proof of Lemma 2.9. These are prior theorems, not proved in this paper.
  • standard math Bhattacharya's functional central limit theorem [2] for ergodic Markov processes.
    Used in Corollary 2 to convert the bounded inverse of the generator into a CLT for additive functionals.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hypocoercivity properties of adaptive Langevin dynamics." pith.science (2026). https://pith.science/paper/OEW3W7KB

@misc{pith2026190809363,
  author       = {Pith},
  title        = {Pith review of: Hypocoercivity properties of adaptive Langevin dynamics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OEW3W7KB}},
  note         = {Machine review of arXiv:1908.09363}
}
read the original abstract

Adaptive Langevin dynamics is a method for sampling the Boltzmann-Gibbs distribution at prescribed temperature in cases where the potential gradient is subject to stochastic perturbation of unknown magnitude. The method replaces the friction in underdamped Langevin dynamics with a dynamical variable, updated according to a negative feedback loop control law as in the Nos\'e-Hoover thermostat. Using a hypocoercivity analysis we show that the law of Adaptive Langevin dynamics converges exponentially rapidly to the stationary distribution, with a rate that can be quantified in terms of the key parameters of the dynamics. This allows us in particular to obtain a central limit theorem with respect to the time averages computed along a stochastic path. Our theoretical findings are illustrated by numerical simulations involving classification of the MNIST data set of handwritten digits using Bayesian logistic regression.

Figures

Figures reproduced from arXiv: 1908.09363 by the authors.

Figure 1
Figure 1. Spectral gap, λbε,γ, of −LAdL when considered as an operator on G L 0 , with L = 10. Panel A shows λbε,γ as a function of ε for fixed γ. Panel B shows λbε,γ as a function of γ for fixed ε. Panel C shows λbα,α as a function of the scalar α. 4.2. Scaling of asymptotic variance and demonstration of CLT. We next consider a simple skewed double-well potential U : R → R, of the form (4.4) U(q) = b a [PITH_FULL_IMAGE:figu… view at source ↗
Figure 2
Figure 2. Estimated asymptotic variance σb 2 ε,γ(K) for various observables with fixed γ = 1 as a function of ε (Panel A), and with fixed ε = 1 as a function of γ (Panel B), respectively. The dashed line in Panel A corresponds to the slope of a quadratic function in ε. The dashed line in Panel B indicates the slope of a linear function in γ. centering the mean of each pixel, we retain the first 100 principal components and wh… view at source ↗
Figure 3
Figure 3. Empirical probability distribution (EPDF) of the rescaled residual error at different times [PITH_FULL_IMAGE:figures/full_fig_p023_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Examples of images from the MNIST data set. The upper row shows the original images as [PITH_FULL_IMAGE:figures/full_fig_p024_4.png]
Figure 5
Figure 5. Figure 5: Empirical probability distribution (EPDF) of the rescaled residual error at different times [PITH_FULL_IMAGE:figures/full_fig_p024_5.png]
Figure 6
Figure 6. Figure 6: shows the time evolution of the corresponding Monte Carlo estimates of the mean of the 65-th component (Panel A), and the average likelihood over the test set (Panel B). As one may have anticipated based on the asymptotic scaling of the spectral gap as O(ν −1 ) as ν → …
Figure 7
Figure 7. Figure 7: Estimated asymptotic variance σb 2 ε,γ(K) for various observables with fixed ε = 1 as a function of γ. The dashed line corresponds to the slope of a linear function in γ −1 . Appendix C. Numerical integrators. In this section we briefly describe the construction of the…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Hypocoercivity meets lifts

    math.PR 2024-12 conditional novelty 6.0 of 10

    A unified hypocoercivity framework shows adaptive Langevin dynamics is a near-optimal lift, and proves the generalized Langevin equation cannot beat square-root speedup.

  2. Partitioned integrators for thermodynamic parameterization of neural networks

    cs.LG 2019-08 conditional novelty 6.0 of 10

    Layer-partitioned Langevin integrators (LOL and AdLaLa) train single hidden layer perceptrons faster, more accurately, and more robustly than SGD/Adam on hard spiral and trigonometric classification problems.

Reference graph

Works this paper leans on

37 extracted references · 37 canonical work pages · cited by 2 Pith papers

  1. [1]

    Bakry, F

    D. Bakry, F. Barthe, P. Cattiaux, and A. Guillin , A simple proof of the Poincar´ e inequality for a large class of probability measures, Electronic Communications in Probability, 13 (2008), pp. 60–66

    work page 2008

  2. [2]

    R. N. Bhattacharya, On the functional central limit theorem and the law of the iterated logarithm for Markov processes, Zeitschrift f¨ ur Wahrscheinlichkeitstheorie und verwandte Gebiete, 60 (1982), pp. 185–201

    work page 1982

  3. [3]

    Bou-Rabee and H

    N. Bou-Rabee and H. Owhadi , Long-run accuracy of variational integrators in the stochastic context , SIAM Journal on Numerical Analysis, 48 (2010), pp. 278–297

    work page 2010

  4. [4]

    Cances, F

    E. Cances, F. Legoll, and G. Stoltz , Theoretical and numerical comparison of some sampling methods for molecular dynamics, ESAIM Math. Model. Numer. Anal., 41 (2007), pp. 351–389

    work page 2007

  5. [5]

    T. Chen, E. Fox, and C. Guestrin, Stochastic gradient Hamiltonian Monte Carlo, in International Conference on Machine Learning, 2014, pp. 1683–1691

    work page 2014

  6. [6]

    Cheng, N

    X. Cheng, N. S. Chatterji, P. L. Bartlett, and M. I. Jordan , Underdamped Langevin MCMC: A non- asymptotic analysis, in Proceedings of the 31st Conference On Learning Theory, vol. 75 of Proceedings of Machine Learning Research, PMLR, 06–09 Jul 2018, pp. 300–323

    work page 2018

  7. [7]

    N. Ding, Y. Fang, R. Babbush, C. Chen, R. D. Skeel, and H. Neven , Bayesian sampling using stochastic gradient thermostats, in Advances in neural information processing systems, 2014, pp. 3203–3211

    work page 2014

  8. [8]

    Dolbeault, C

    J. Dolbeault, C. Mouhot, and C. Schmeiser , Hypocoercivity for kinetic equations with linear relaxation terms, C. R. Math. Acad. Sci. Paris, 347 (2009), pp. 511–516

    work page 2009

Show all 37 references
  1. [9]

    Dolbeault, C

    J. Dolbeault, C. Mouhot, and C. Schmeiser , Hypocoercivity for linear kinetic equations conserving mass , Transactions of the American Mathematical Society, 367 (2015), pp. 3807–3828

    2015

  2. [10]

    Eberle, Reflection coupling and Wasserstein contractivity without convexity , Comptes Rendus Mathema- tique, 349 (2011), pp

    A. Eberle, Reflection coupling and Wasserstein contractivity without convexity , Comptes Rendus Mathema- tique, 349 (2011), pp. 1101–1104

    2011

  3. [11]

    Eberle, A

    A. Eberle, A. Guillin, and R. Zimmer, Couplings and quantitative contraction rates for Langevin dynamics , Ann. Probab., 47 (2019), pp. 1982–2010

    2019

  4. [12]

    Hairer and J

    M. Hairer and J. C. Mattingly , Yet another look at Harris ergodic theorem for Markov chains , in Seminar on Stochastic Analysis, Random Fields and Applications VI, vol. 63, Springer, 2011, pp. 109–117

    2011

  5. [13]

    D. P. Herzog, Exponential relaxation of the Nos´ e-Hoover equation under Brownian heating, Communications in Mathematical Sciences, 16 (2018), pp. 2231–2260

    2018

  6. [14]

    H¨ormander, Hypoelliptic second order differential equations, Acta Math., 119 (1967), pp

    L. H¨ormander, Hypoelliptic second order differential equations, Acta Math., 119 (1967), pp. 147–171

    1967

  7. [15]

    Iacobucci, S

    A. Iacobucci, S. Olla, and G. Stoltz , Convergence rates for nonequilibrium Langevin dynamics , Annales math´ ematiques du Qu´ ebec, 43 (2019), pp. 73–98

    2019

  8. [16]

    Jones and B

    A. Jones and B. Leimkuhler, Adaptive stochastic methods for sampling driven molecular systems, The Journal of Chemical Physics, 135 (2011), p. 084125

    2011

  9. [17]

    Kliemann, Recurrence and invariant measures for degenerate diffusions, Ann

    W. Kliemann, Recurrence and invariant measures for degenerate diffusions, Ann. Probab., 15 (1987), pp. 690– 707

    1987

  10. [18]

    Kopec , Weak backward error analysis for Langevin process , BIT Numerical Mathematics, 55 (2015), pp

    M. Kopec , Weak backward error analysis for Langevin process , BIT Numerical Mathematics, 55 (2015), pp. 1057–1103

    2015

  11. [19]

    LeCun and C

    Y. LeCun and C. Cortes , MNIST handwritten digit database , (2010), http://yann.lecun.com/exdb/mnist/

    2010

  12. [20]

    Leimkuhler and C

    B. Leimkuhler and C. Matthews, Molecular Dynamics: With Deterministic and Stochastic Numerical Meth- ods, Interdisciplinary Applied Mathematics, Springer, 2015

    2015

  13. [21]

    Leimkuhler, C

    B. Leimkuhler, C. Matthews, and G. Stoltz , The computation of averages from equilibrium and nonequi- librium Langevin molecular dynamics , IMA Journal of Numerical Analysis, 36 (2015), pp. 13–79

    2015

  14. [22]

    Leimkuhler and X

    B. Leimkuhler and X. Shang , Adaptive thermostats for noisy gradient systems , SIAM Journal on Scientific Computing, 38 (2016), pp. A712–A736

    2016

  15. [23]

    Leli `evre and G

    T. Leli `evre and G. Stoltz , Partial differential equations and stochastic methods in molecular dynamics , Acta Numerica, 25 (2016), pp. 681–880

    2016

  16. [24]

    J. C. Mattingly, A. M. Stuart, and D. J. Higham, Ergodicity for SDEs and approximations: locally Lipschitz vector fields and degenerate noise , Stochastic Processes and their Applications, 101 (2002), pp. 185–232

    2002

  17. [25]

    S. P. Meyn and R. L. Tweedie , Stability of Markovian processes II: Continuous-time processes and sampled chains, Advances in Applied Probability, 25 (1993), pp. 487–517

    1993

  18. [26]

    S. P. Meyn and R. L. Tweedie , Markov Chains and Stochastic Stability , Springer Science & Business Media, HYPOCOERCIVITY PROPERTIES OF ADAPTIVE LANGEVIN DYNAMICS 27 2012

    2012

  19. [27]

    Nos ´e, A unified formulation of the constant temperature molecular dynamics methods , The Journal of Chemical Physics, 81 (1984), pp

    S. Nos ´e, A unified formulation of the constant temperature molecular dynamics methods , The Journal of Chemical Physics, 81 (1984), pp. 511–519

    1984

  20. [28]

    G. A. Pavliotis, Stochastic Processes and Applications, Springer, 2016

    2016

  21. [29]

    G. A. Pavliotis and A. Vogiannou, Diffusive transport in periodic potentials: underdamped dynamics , Fluc- tuation and Noise Letters, 08 (2008), pp. L155–L173

    2008

  22. [30]

    Rey-Bellet, Ergodic properties of Markov processes, in Open Quantum Systems II, S

    L. Rey-Bellet, Ergodic properties of Markov processes, in Open Quantum Systems II, S. Attal, A. Joye, and C.-A. Pillet, eds., vol. 1881 of Lecture Notes in Mathematics, Springer, 2006, pp. 1–39

    2006

  23. [31]

    Roussel and G

    J. Roussel and G. Stoltz , Spectral methods for Langevin dynamics and associated error estimates , ESAIM Math. Model. Numer. Anal., 52 (2018), pp. 1051–1083

    2018

  24. [32]

    Sachs, B

    M. Sachs, B. Leimkuhler, and V. Danos , Langevin dynamics with variable coefficients and nonconservative forces: From stationary states to numerical methods , Entropy, 19 (2017), p. 647

    2017

  25. [33]

    Shang, Z

    X. Shang, Z. Zhu, B. Leimkuhler, and A. J. Storkey , Covariance-controlled adaptive Langevin thermostat for large-scale Bayesian sampling, in Advances in Neural Information Processing Systems, 2015, pp. 37–45

    2015

  26. [34]

    D. Talay, Stochastic Hamiltonian systems: exponential convergence to the invariant measure, and discretiza- tion by the implicit Euler scheme , Markov Processes and Related Fields, 8 (2002), pp. 163–198

    2002

  27. [35]

    V´arnai, N

    C. V´arnai, N. Bernstein, L. Mones, and G. Cs´anyi, Tests of an adaptive QM/MM calculation on free energy profiles of chemical reactions in solution , The Journal of Physical Chemistry B, 117 (2013), pp. 12202– 12211

    2013

  28. [36]

    Villani, Hypocoercivity, Mem

    C. Villani, Hypocoercivity, Mem. Amer. Math. Soc., 202 (2009)

    2009

  29. [37]

    Welling and Y

    M. Welling and Y. W. Teh, Bayesian learning via stochastic gradient Langevin dynamics , in Proceedings of the 28th International Conference on Machine Learning (ICML-11), 2011, pp. 681–688. Appendix A. Derivation of the stiffness matrix in the Hermite Galerkin projection. In th...

    2011

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.