Pith. sign in

REVIEW 4 major objections 5 minor 2 cited by

Optimal response for stochastic differential equations by local kernel perturbations

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read For a dissipative SDE on R^d, the invariant density responds to local kernel perturbations with an explicit resolvent formula, and the optimal-perturbation problem has a unique solution.

desk verdict Useful extension of optimal linear response to non-compact R^d with a concrete numerical scheme, but the key spectral gap proof has a hole and one uniqueness proof contains a false inference; likely repairable, but needs major revision. read the letter →

arxiv 2502.09300 v2 pith:I6ZKIEGW submitted 2025-02-13 math.DS math.OC

Gianmarco del Sarto , Stefano Galatolo , Sakshi Jain This is my paper · ORCID
classification math.DSmath.OC MSC 37H1037C3037M0537N3549N4560H10
keywords stochasticdifferentialequationstransferoperatorslinearresponseoptimalkernelperturbationspectralgapinvariantdensityFokker-Planckequation
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 studies an inverse control problem for stochastic differential equations: given a dissipative SDE on $\mathbb{R}^d$ and an observable $\varphi$, which small local change to the transition kernel produces the largest change in the expectation of $\varphi$? The authors prove that the invariant density responds differentiably to such kernel perturbations, with the response given explicitly by $(\mathrm{Id}-L_0)^{-1}\dot{L}f_0$ in $L^1$, and that over a closed, bounded, strictly convex set of allowed perturbations the optimal perturbation exists and is unique. This extends optimal-response theory from compact phase spaces to the noncompact space $\mathbb{R}^d$, the natural setting for many SDE models. The paper also gives a numerical recipe and demonstrates it on a double-well gradient SDE.

What carries the argument

The load-bearing object is the annealed transfer operator $L_0 f(y)=\int \kappa(x,y)f(x)\,dx$, whose kernel $\kappa$ is the time-1 transition density of the SDE; two-sided Gaussian bounds and gradient estimates give it enough regularity to act on a strong space $B=\{f\in L^1_2(\mathbb{R}^d): \|f\|_{L^1_2}+\|\mathbf{1}_D f\|_2<\infty\}$. On this space a Lasota-Yorke inequality together with a compact embedding into $L^1$ yields a spectral gap, with the resolvent $(\mathrm{Id}-L_0)^{-1}$ bounded on zero-average densities. The response operator $R(\dot{\kappa})=(\mathrm{Id}-L_0)^{-1}\dot{L}f_0$ is then a bounded linear map $L^2(D\times D)\to L^1$, so maximising $\int \varphi\,R(\dot{\kappa})$ over a closed bounded strictly convex set is a strictly convex optimisation problem solved by the normalised Riesz representer $g/\|g\|_2$, whose Fourier coefficients are computable as $G_r = J(h_r)$ for an orthonormal basis $(h_r)$ of $L^2(D\times D)$.

What would settle it

For the one-dimensional Ornstein-Uhlenbeck SDE $dX=-X\,dt+dW$, compute or bound $\|L|_{V_B}\|_s$, the strong-norm operator norm of the transfer operator restricted to zero-average densities. If this norm is not strictly less than 1, the contraction premise behind Lemma 14 is false and the spectral gap, hence linear response and uniqueness of the optimal perturbation, must be established by another mechanism; the paper's own Remark 18 shows the analogous $L^2$ norm does not contract.

Watch

Extended reading notes

Core claim

The central claim is Theorem 25: for the time-1 annealed transfer operator $L_0$ of a dissipative SDE on $\mathbb{R}^d$, a local kernel perturbation $\kappa_\delta = \kappa_0 + \delta\dot{\kappa} + r_\delta$ with $r_\delta = o(\delta)$ supported on a compact set $D\times D$ produces an invariant-density response satisfying $\|(f_\delta - f_0)/\delta - (\mathrm{Id}-L_0)^{-1}\dot{L}f_0\|_1 \to 0$, where $\dot{L}f(y)=\int \dot{\kappa}(x,y)f(x)\,dx$ and $f_0$ is the unperturbed invariant density. The second main claim, Proposition 28, states that for any $\varphi\in L^\infty$ and any closed, bounded, strictly convex set $P\subset L^2(D\times D)$ containing zero in its relative interior, the problem of maximising the observable's response over $P$ has a unique solution, given by the normalised Riesz representer of the linear functional $J(\dot{\kappa})=\int \varphi\,(\mathrm{Id}-L_0)^{-1}\dot{L}f_0$. These results rest on constructing a strong space $B$ of densities with weighted decay at infinity and local $L^2$ control on $D$, on which $L_0$ has a spectral gap. The paper derives a Fourier-approximation scheme for the optimal perturbation and illustrates it on symmetric and asymmetric double-well SDE examples with Gaussian observables.

Load-bearing premise

The spectral-gap mechanism rests on the assertion, stated without proof in Lemma 14, that the unperturbed transfer operator strictly contracts the strong norm on zero-average densities; the standard positivity argument gives strict contraction only in $L^1$, so if the strong-norm contraction fails, the linear-response and optimisation conclusions would not follow.

Editorial extensions

If this is right

  • The invariant-density derivative has an explicit, numerically evaluable form $(\mathrm{Id}-L_0)^{-1}\dot{L}f_0$, so linear response for dissipative SDEs on $\mathbb{R}^d$ becomes a computational procedure rather than an existence statement.
  • The optimal kernel-perturbation problem is well-posed: for any closed, bounded, strictly convex set of admissible perturbations containing zero in its relative interior, a unique maximiser exists.
  • A truncated Fourier basis gives an implementable approximation of the optimal perturbation, with convergence as the basis grows; the experiments show the optimal kernel adjustment is localised where the observable is significant.
  • The results apply to any kernel transfer operator on $\mathbb{R}^d$ with the same Gaussian bounds and strong-space properties, so they transfer to other dissipative SDEs beyond the double-well examples.
  • The perturbed kernel need not correspond to an SDE with a modified drift; the authors frame this local, model-independent perturbation as a first step toward treating drift perturbations in the same framework.

Reading between the lines

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

  • The authors note that existing drift-perturbation estimates converge in $L^2$, which is insufficient for their spectral-gap strategy; if Frechet differentiability of the kernel in the drift were proved in the strong norm, the same optimisation framework would likely cover perturbations that genuinely correspond to SDEs.
  • Uniqueness requires strict convexity of $P$; for a feasible set like a box the maximum may be attained on a face, although the Riesz-representer formula still gives the steepest direction at the origin.
  • Computing each Fourier coefficient $G_r$ by solving the resolvent system separately is costly; a fast adjoint method that solves one dual equation per observable would produce all coefficients at once.
  • A numerical check of the contraction constant on the Ornstein-Uhlenbeck example, where the paper already shows the $L^2$ analogue fails, would directly test whether the spectral-gap mechanism is the right one.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper studies a dissipative stochastic differential equation on R^d and its annealed transfer operator L at time one, viewed as a kernel operator. The authors construct a 'strong' Banach space B of densities with a weighted L1 norm and an L2 norm on a compact set D, and cite results from prior work for Lasota-Yorke inequalities and BV-type compactness. They then consider infinitesimal perturbations of the kernel supported in D, prove a linear-response formula of the form (Id - L0)^{-1} \dot L f0 in L1 (Theorem 25), and show that the problem of maximizing the response of an observable over a closed bounded strictly convex set P of perturbations has a unique solution (Proposition 28). A Fourier-based numerical scheme for approximating the optimal perturbation is presented and illustrated on a double-well SDE.

Significance. If fully justified, the paper extends the optimal-response framework from compact phase spaces to noncompact R^d for SDEs, which is a genuinely useful step for climate and control applications. The explicit linear-response formula and the constructive numerical algorithm, with code available on Zenodo, are notable strengths. The mathematical architecture is appropriate: Lasota-Yorke inequalities, resolvent bounds, and convex optimization are the right tools, and most of Sections 4 and 5 are standard once the spectral gap is granted. The main caveat is that two load-bearing proofs, the spectral-gap proof in Lemma 14 and the uniqueness proof in Proposition 23, are not rigorous as written. These should be fixed before the central claims can be considered established.

major comments (4)
  1. [Lemma 14] The simplicity argument for the eigenvalue 1 asserts the inequality ||L(u-v)||_s < ||u-v||_s from positivity of the kernel. This does not follow. Positivity gives pointwise |Lg| < L|g| for nonzero zero-mean g and hence strict contraction in L1, but it gives no such strict contraction in the strong norm ||g||_s = ||g||_{L1_2} + ||1_D g||_2; the L2(D)-component is unrelated to the positivity argument and can increase. In fact, if u-v is a fixed point, then ||L(u-v)||_s = ||u-v||_s, contradicting the asserted strict inequality. The simplicity of the eigenvalue 1 can likely be proved using the strict L1 contraction instead, but as written the proof is invalid.
  2. [Lemma 14 / Proposition 12] The proof that L has a spectral gap on B is incomplete. The cited Hennion theorem, in its standard form, requires a compact embedding of the strong space into the weak space; that embedding is false here. For x0 outside D, the sequence f_n = n 1_{[x0, x0+1/n]} satisfies ||f_n||_s = O(rho_2(x0)) < infinity yet has no subsequence convergent in L1, since it converges weakly to a Dirac mass. Proposition 12 proves only compactness of L(B) in L1, not compactness of the embedding B into L1, and its proof itself contains an invalid step: a Cauchy sequence in L1 is asserted to converge in BV2, and the closed unit ball of B is asserted to be sequentially compact in ||.||_s. If the authors intend a variant of Hennion's theorem that requires only relative compactness of L(B) in L1, that variant and its hypotheses should be stated explicitly and verified. As written, the spectral gap and hence Proposition 17, Theorem 25, and Proposition 28 rest on an unproved assumption.
  3. [Proposition 23] The uniqueness paragraph contains a false implication: from int(f_delta - g_delta) = 0 the text concludes ||f_delta - g_delta||_1 = 0. This is not valid; a nonzero function with zero integral exists in any nontrivial L1 space. Moreover, the perturbed kernel kappa_delta is not shown to be positive, so the positivity-based contraction argument for L0 cannot be applied to L_delta. Uniqueness of the invariant density f_delta may be recoverable from the simplicity of the eigenvalue 1 obtained from the Keller-Liverani stability statement immediately preceding, but that connection is not made, and the proof as written is unsupported.
  4. [Proposition 23 / Lemma 24] The text states that 'B is compactly immersed inside L1 (Lemma 12)', but Lemma 12 is the compactness of L(B) in L1, not the compactness of the embedding B into L1, which is false. This matters because the application of [25, Theorem 1] in Proposition 23 and Lemma 24 requires a correct compactness hypothesis for the perturbed operators. The authors should spell out the uniform conditions, for example a uniform Lasota-Yorke inequality together with relative compactness of L_delta(B) in L1, that justify the eigenvalue stability and the resolvent convergence. Without this, the perturbed spectral theory used in Theorem 25 is not rigorously grounded.
minor comments (5)
  1. [Lemma 24] The statement contains a typo: the limit expression at the end repeats (Id - L_delta)^{-1} twice; the second occurrence should be (Id - L0)^{-1}.
  2. [Lemma 22] In the proof, 'By Theorem 13' should read 'By Lemma 13'.
  3. [Lemma 20] The last display writes a bound of the form C/delta ||1_D f0||_2 -> 0, which is inconsistent with the hypothesis ||r_delta||_2 = o(delta). The correct factor is ||r_delta/delta||_2 -> 0; please correct this line.
  4. [Section 3.2] The phrase 'compact inclusion of the strong space B in L1' in the introduction to Section 3.2 is misleading, since B is not compactly embedded in L1; what is needed is the compactness of L(B) in L1.
  5. [Section 6 / Code availability] The Zenodo code is reported to run with Delta x = 8e-3 while the figures use Delta x = 2e-3; the footnote explains how to modify the code, but for reproducibility it would be preferable to either provide the exact code used or explicitly state that the published figures require a parameter change.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the linear-response formula and optimization uniqueness are obtained from cited external theorems and standard convex analysis, not from the paper's own conclusions.

full rationale

The paper's derivation chain is not circular. The central objects — the transfer operator L0, the strong space B, the response R(kappa_dot) = lim (f_delta - f0)/delta, and the objective J(kappa_dot) = -∫ phi R(kappa_dot) — are defined independently of the results they are used to prove. Theorem 25 derives the linear-response formula (Id - L0)^{-1} Ldot f0 from the spectral setup, relying on cited external results ([11] for weighted-BV compactness and Lasota-Yorke estimates, [25] for spectral stability, [10] for Hennion's theorem) rather than on the formula itself. Proposition 28 obtains uniqueness of the optimal perturbation from general convex-analysis propositions in Appendix A, and no fitted constants are renamed as predictions. The self-citations to [2], [11], and [12] are prior theorems with proofs or stated assumptions that do not include the present conclusion; they are load-bearing but not circular. The numerical experiments independently solve the Fokker-Planck equation and verify the predicted shift of the invariant density, providing an external consistency check. The main vulnerability is a correctness gap, not a circular one: Lemma 14 asserts strict contraction ||Lw||_s < ||w||_s for zero-mean w from positivity of the kernel, but positivity yields strict L1 contraction and not strict contraction in the weighted strong norm; Proposition 12 also does not establish the compact embedding B -> L1. These are proof gaps affecting the spectral-gap hypothesis, not reductions of the conclusions to their inputs.

Assumptions & free parameters 1 free parameters · 6 assumptions · 0 invented entities

The central claim rests on standard stochastic analysis and operator theory: two-sided Gaussian density bounds for the SDE (Theorem 1 from [30]), weighted BV/L1 spectral gap results from [11], and Keller-Liverani spectral stability [25]. No parameters are fitted to data; alpha=2 in the weight and the compact set D are a priori choices. No invented entities.

free parameters (1)
  • alpha (weight exponent)
    Exponent in the weight rho_alpha(|x|) = (1+|x|^2)^(alpha/2) used to define the strong space B; chosen a priori as alpha=2, not fitted to data.
assumptions (6)
  • domain assumption Drift b is locally Lipschitz and dissipative (Assumptions A and B)
    Used throughout Section 3 to guarantee existence and uniqueness of the SDE solution and its stationary measure.
  • domain assumption Two-sided Gaussian bounds and gradient estimates for the transition density (Theorem 1, from Menozzi-Pesce-Zhang [30])
    Provides strict positivity and regularity of kappa, used for L1 strict contraction and boundedness of the transfer operator.
  • domain assumption Weighted BV compact embedding and Lasota-Yorke estimates for L from Flandoli-Galatolo-Giulietti-Vaienti [11]
    Basis for the spectral gap of L on the strong space B, quoted as Proposition 9 and Lemmas 10-11 of this paper.
  • standard math Keller-Liverani spectral stability theorem ([25, Theorem 1])
    Used to pass spectral gap and simplicity of the leading eigenvalue from L0 to Ldelta.
  • standard math Hennion theorem on quasi-compactness of operators
    Used in Lemma 14 to infer spectral gap from Lasota-Yorke and compact embedding.
  • standard math Riesz representation theorem and strict convexity of the unit ball in Hilbert space
    Yields the representer g and uniqueness of the optimum in Section 5.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Optimal response for stochastic differential equations by local kernel perturbations." pith.science (2026). https://pith.science/paper/I6ZKIEGW

@misc{pith2026250209300,
  author       = {Pith},
  title        = {Pith review of: Optimal response for stochastic differential equations by local kernel perturbations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/I6ZKIEGW}},
  note         = {Machine review of arXiv:2502.09300}
}
abstract

We consider a random dynamical system on $\mathbb{R}^d$, whose dynamics is defined by a stochastic differential equation. The annealed transfer operator associated with such systems is a kernel operator. Given a set of feasible infinitesimal perturbations $P$ to this kernel, with support in a certain compact set, and a specified observable function $\phi: \mathbb{R}^d \to \mathbb{R}$, we study which infinitesimal perturbation in $P$ produces the greatest change in expectation of $\phi$. We establish conditions under which the optimal perturbation uniquely exists and present a numerical method to approximate the optimal infinitesimal kernel perturbation. Finally, we numerically illustrate our findings with concrete examples.

Figures

Figures reproduced from arXiv: 2502.09300 by the authors.

Figure 1
Figure 1. Numerical approximation for (a) kernel κ(x, y) = p x (y, T ), where p is a solution of the FPE (36), and (b) invari￾ant density for the transfer operator L0. The final time is T = 1, the noise intensity is ε = 0.25, the domain is Ω = (−2, 2), while the mesh sizes are ∆x = ∆x = 2 · 10−3 . probability mass can escape the domain Ω = (−a, a) ([17]). This leads to (36)    ∂tp x − ε 2 2 ∆yp x + ∂y (pb) … view at source ↗
Figure 2
Figure 2. Symmetric experiment, ∆x = ∆t = 2 · 10−3 , T = 1, D = [−1.2, 1.2], I = 35, J = 35. (a) Symmetric observable y 7→ φ(y). (b) Optimal perturbation gI,J = P r :hr∈BI,J Grhr P r :hr∈BI,J G2 r 1/2 . (c) Perturbed kernel k1/2 = κ + 1 2 gI,J . (d) Invariant densities f0 and f1/2, for L0 and L1/2 respectively. of this work consists of the implementation of the Simpson’s rule for numerical integration by Damien Garcia (link… view at source ↗
Figure 3
Figure 3. Asymmetric experiment, ∆x = ∆t = 2 · 10−3 , T = 1, D = [−1.2, 1.2], I = 35, J = 35. (a) Asymmetric observable y 7→ φ(y). (b) Optimal perturbation gI,J = P r :hr∈BI,J Grhr P r :hr∈BI,J G2 r 1/2 . (c) Perturbed kernel k1/2 = κ + 1 2 gI,J . (d) Invariant densities f0 and f1/2, for L0 and L1/2 respectively [PITH_FULL_IMAGE:figures/full_fig_p027_3.png] view at source ↗

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. Markov matrix perturbations to optimize dynamical and entropy functionals

    math.DS 2025-07 conditional novelty 6.0 of 10

    Linear-response optimization algorithms are derived for entropy, KL divergence, and entropy production on Markov chains, with a drift-reconstruction protocol linking matrix perturbations to vector field forcing.

  2. Divergence-Kernel method for scores of random systems

    math.PR 2025-07 conditional novelty 6.0 of 10

    New divergence-kernel formulas compute scores of SDEs with multiplicative noise via pathwise Monte Carlo, without hyperbolicity assumptions.

Reference graph

Works this paper leans on

37 extracted references · 36 canonical work pages · cited by 2 Pith papers

  1. [25]

    Stability of t he spectrum for transfer operators

    Gerhard Keller and Carlangelo Liverani. Stability of t he spectrum for transfer operators. Ann. Scuola Norm. Sup. Pisa Cl. Sci. (4) , 28(1):141–152, 1999. 30 OPTIMAL RESPONSE FOR SDE BY LOCAL KERNEL PERTURBATIONS

    work page 1999

  2. [1]

    Optim al linear responses for markov chains and stochastically perturbed dynamical systems

    Fadi Antown, Davor Dragičević, and Gary Froyland. Optim al linear responses for markov chains and stochastically perturbed dynamical systems. Journal of Statistical Physics , 170:1051–1087, 3 2018

    work page 2018

  3. [2]

    Optim al linear response for markov hilbert–schmidt integral operators and stochastic dynami cal systems

    Fadi Antown, Gary Froyland, and Stefano Galatolo. Optim al linear response for markov hilbert–schmidt integral operators and stochastic dynami cal systems. Journal of Nonlinear Science, 32(79):60, 2022

    work page 2022

  4. [3]

    Linea r response for random dynamical systems

    W ael Bahsoun, Marks Ruziboev, and Benoît Saussol. Linea r response for random dynamical systems. Advances in Mathematics , 364:107011, 2020

    work page 2020

  5. [4]

    Linear response, or else

    Viviane Baladi. Linear response, or else. ICM Seoul 2014 talk , 2014

    work page 2014

  6. [5]

    Stochastic calculus

    Paolo Baldi. Stochastic calculus. Universitext. Springer, Cham, 2017. An introduction thro ugh theory and exercises

    work page 2017

  7. [6]

    D. A. Bini, G. Latouche, and B. Meini. Numerical methods for structured Markov chains . Numerical Mathematics and Scientific Computation. Oxford U niversity Press, New York,

    work page

  8. [7]

    Can we use linear response theory to assess geoengineering strategies? Chaos: An Interdisciplinary Journal of Nonlinear Science , 30(2), February 2020

    Tamás Bódai, Valerio Lucarini, and Frank Lunkeit. Can we use linear response theory to assess geoengineering strategies? Chaos: An Interdisciplinary Journal of Nonlinear Science , 30(2), February 2020

    work page 2020

Show all 37 references
  1. [8]

    Linear a nd fractional response for nonlinear dissipative SPDEs

    Giulia Carigi, Tobias Kuna, and Jochen Bröcker. Linear a nd fractional response for nonlinear dissipative SPDEs. Nonlinearity, 37(10):Paper No. 105002, 41, 2024

    2024

  2. [9]

    John B. Conway. A course in functional analysis , volume 96 of Graduate Texts in Mathe- matics. Springer-Verlag, New York, second edition, 1990

    1990

  3. [10]

    Demers, Niloofar Kiamari, and Carlangelo Liver ani

    Mark F. Demers, Niloofar Kiamari, and Carlangelo Liver ani. Transfer operators in hyperbolic dynamics—an introduction . 33 o Colóquio Brasileiro de Matemática. Instituto Nacional de Matemática Pura e Aplicada (IMPA), Rio de Janeiro, 2021

    2021

  4. [11]

    Flandoli, S

    F. Flandoli, S. Galatolo, P. Giulietti, and S. Vaienti. Extreme value theory and poisson statistics for discrete time samplings of stochastic differ ential equations. arXiv:2310.13972, 2024

    2024 arXiv

  5. [12]

    Optimal linear res ponse for expanding circle maps

    Gary Froyland and Stefano Galatolo. Optimal linear res ponse for expanding circle maps. Nonlinearity, 38, 2025

    2025

  6. [13]

    A linear response for dynami cal systems with additive noise

    S Galatolo and P Giulietti. A linear response for dynami cal systems with additive noise. Nonlinearity, 32:2269–2301, 6 2019

    2019

  7. [14]

    Self-consistent transfer operator s: Invariant measures, convergence to equi- librium, linear response and control of the statistical pro perties

    Stefano Galatolo. Self-consistent transfer operator s: Invariant measures, convergence to equi- librium, linear response and control of the statistical pro perties. Communications in Mathe- matical Physics , 395:715–772, 10 2022

    2022

  8. [15]

    Optimal response for hy perbolic systems by the fast adjoint response method

    Stefano Galatolo and Angxiu Ni. Optimal response for hy perbolic systems by the fast adjoint response method. arXiv:2501.02395, 2025

    2025 arXiv

  9. [16]

    Controlling the s tatistical properties of expanding maps

    Stefano Galatolo and Mark Pollicott. Controlling the s tatistical properties of expanding maps. Nonlinearity, 30:2737–2751, 7 2017

    2017

  10. [17]

    Stochastic methods

    Crispin Gardiner. Stochastic methods. Springer Series in Synergetics. Springer-Verlag, Berlin , fourth edition, 2009. A handbook for the natural and social s ciences

    2009

  11. [18]

    Gardiner

    Crispin W. Gardiner. Handbook of Stochastic Methods for Physics, Chemistry and th e Natural Sciences. Springer Series in Synergetics. Springer, Berlin, Heidel berg, 3rd edition, 2004

    2004

  12. [19]

    Ghil and V

    M. Ghil and V. Lucarini. The physics of climate variabil ity and climate change. Rev. Mod. Phys, 92:035002, 2020

    2020

  13. [20]

    Convergence of markov processes

    Martin Hairer. Convergence of markov processes. Lecture notes, 18(26):13, 2010

    2010

  14. [21]

    A simple framework to j ustify linear response theory

    Martin Hairer and Andrew J Majda. A simple framework to j ustify linear response theory. Nonlinearity, 23:909–922, 4 2010

    2010

  15. [22]

    Hypoelliptic second order differentia l equations

    Lars Hörmander. Hypoelliptic second order differentia l equations. Acta Mathematica , 119(0):147–171, 1967

    1967

  16. [23]

    Th e variational formulation of the Fokker-Planck equation

    Richard Jordan, David Kinderlehrer, and Felix Otto. Th e variational formulation of the Fokker-Planck equation. SIAM J. Math. Anal. , 29(1):1–17, 1998

    1998

  17. [24]

    Generalized bounded variation and app lications to piecewise monotonic transformations

    Gerhard Keller. Generalized bounded variation and app lications to piecewise monotonic transformations. Zeitschrift für Wahrscheinlichkeitstheorie und Verwandt e Gebiete , 69:461– 478, 9 1985

    1985

  18. [26]

    The linear request problem

    Benoît Kloeckner. The linear request problem. Proceedings of the American Mathematical Society, 146:2953–2962, 3 2018

    2018

  19. [27]

    Fréchet differentiable drift dependence of perron–frobenius and koopman operators for non-determi nistic dynamics

    Péter Koltai, Han Cheng Lie, and Martin Plonka. Fréchet differentiable drift dependence of perron–frobenius and koopman operators for non-determi nistic dynamics. Nonlinearity, 32:4232, 2019

    2019

  20. [28]

    Management of complex dynamical systems

    R S MacKay. Management of complex dynamical systems. Nonlinearity, 31:R52–R65, 2 2018

    2018

  21. [29]

    Stochastic differential equations and their applications

    Xuerong Mao. Stochastic differential equations and their applications . Horwood Publishing Series in Mathematics & Applications. Horwood Publishing L imited, Chichester, 1997

    1997

  22. [30]

    Menozzi, A

    S. Menozzi, A. Pesce, and X. Zhang. Density and gradient estimates for non degenerate Brownian SDEs with unbounded measurable drift. J. Differential Equations , 272:330–369, 2021

    2021

  23. [31]

    Numerical Mathematics

    Alfio Quarteroni, Riccardo Sacco, and Fausto Saleri. Numerical Mathematics . Springer New York, 2007

    2007

  24. [32]

    Numerical approximation of partial differential equations , volume 23 of Springer Series in Computational Mathematics

    Alfio Quarteroni and Alberto Valli. Numerical approximation of partial differential equations , volume 23 of Springer Series in Computational Mathematics . Springer-Verlag, Berlin, 1994

    1994

  25. [33]

    H. Risken. The Fokker-Planck equation , volume 18 of Springer Series in Synergetics . Springer-Verlag, Berlin, 1984. Methods of solution and app lications

    1984

  26. [34]

    Differentiation of srb states

    David Ruelle. Differentiation of srb states. Communications in Mathematical Physics , 187:227–241, 7 1997

    1997

  27. [35]

    Absolutely continuous invariant meas ures for multidimensional expanding maps

    Benoît Saussol. Absolutely continuous invariant meas ures for multidimensional expanding maps. Israel Journal of Mathematics , 116:223–248, 2000

    2000

  28. [36]

    J. W. Thomas. Numerical partial differential equations: finite difference methods, volume 22 of Texts in Applied Mathematics . Springer-Verlag, New York, 1995

    1995

  29. [2005]

    Oxford Science Publications

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.