Pith. sign in

REVIEW 2 major objections 5 minor 80 references

Model Reduction of Multivariate Geometric Brownian Motions and Localization in a Two-State Quantum System

T0 review · 2 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper shows that a multivariate geometric Brownian motion with multiplicative noise can be reduced to a one-dimensional mixed Ornstein–Uhlenbeck and geometric process that preserves the localization properties of the original…

desk verdict The reduction scheme holds up and the derivations are consistent, but the Wasserstein error bounds in Proposition 2 are stated for the wrong covariance and need to be recomputed for the reduced SDE (61) the paper actually defines. read the letter →

arxiv 2507.09413 v3 pith:I4CFUEDJ submitted 2025-07-12 math-ph cond-mat.stat-mechmath.MP

C. Chen , M. Colangeli , M. H. Duong , M. Serva This is my paper · ORCID
classification math-phcond-mat.stat-mechmath.MP MSC 34C4560H1060J60
keywords modelreductiongeometricBrownianmotioninvariantmanifoldsadiabaticeliminationfluctuation-dissipationrelationtwo-statequantumsystemlocalizationWassersteindistance
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 develops a systematic way to shrink a multivariate geometric Brownian motion (a stochastic process whose noise multiplies the state) to a lower-dimensional surrogate without giving up the dynamics that matter. The scheme first reduces the deterministic drift using invariant-manifold or adiabatic-elimination closures, then fixes the reduced noise through a generalized fluctuation-dissipation relation combined with a closest-second-moment optimization. Applied to the localization variable $z$ of a two-state quantum system driven by white noise, the scheme yields a single driven SDE, $dz_{\mathrm{red}} = \bar A_z z_{\mathrm{red}}\, dt + \bar D_z\, du$ with $\bar D_z = \sqrt{-(2/3)\bar A_z}$, and the paper argues that this one-variable process preserves the localization behavior of the full three-variable model. This matters because multiplicative-noise models appear throughout finance, population biology, and statistical physics, and the reduction gives an explicit, closed-form route from a coupled system to a solvable scalar process.

What carries the argument

The load-bearing identity is the generalized fluctuation-dissipation, or Lyapunov, relation $(2\bar A + \bar B^2)\bar p^{\infty} + \bar D^2 = 0$ for a one-dimensional mixed Ornstein–Uhlenbeck and geometric process, which determines the reduced noise once the reduced drift and the exact stationary second moment are known. Around it sit the invariance equations (43) and (47), which fix the reduced drift through the closure $m_y = a m_z$ and the affine second-moment closure $p_{xx} = a_1(p_{zz}-1/3)+1/3$, $p_{yy} = a_2(p_{zz}-1/3)+1/3$, $p_{yz} = a_3(p_{zz}-1/3)$; and Lemmas 1–2, which show that the closest-exponential minimization in (60) is achieved at the boundary $b = b_0$, forcing $\bar B_z = 0$. Together these convert a coupled three-variable SDE into a single autonomous SDE while matching the stationary second moment and the slow drift, with the error controlled by the covariance estimates of Proposition 2.

What would settle it

Solve the exact four-variable second-moment system (41) for parameter values such as $\alpha=0.5$, $\beta=1$, and $\epsilon=0.5$, and compare the resulting $p_{xx}(t)$, $p_{yy}(t)$, $p_{yz}(t)$ with the affine closure (46) evaluated at the computed $a^*_i(\epsilon)$; if the mismatch exceeds the error allowed by the Wasserstein bounds of Proposition 2, the reduction is not sustaining the localization claim at that noise strength.

Watch

Extended reading notes

Core claim

The central claim is that for the two-state quantum GBM (12), the resolved localization variable $z$ can be represented by the one-dimensional mixed Ornstein–Uhlenbeck and geometric process $dz_{\mathrm{red}} = \bar A_z z_{\mathrm{red}}\, dt + \bar D_z\, du$, where $\bar A_z$ is either $-2\alpha^2/\beta^2$ (adiabatic elimination) or $-\bigl(\beta^2 - \sqrt{\beta^4 - 4\alpha^2}\bigr)$ (invariant manifold reduction), and $\bar D_z = \sqrt{-(2/3)\bar A_z}$. The noise coefficient is fixed so that the stationary second moment of the reduced process equals the exact value $p^{\infty}_{zz}=1/3$, and the minimization of the $L^\infty$ (or $L^2$) distance between the reduced and the already-reduced second-moment dynamics selects $\bar B_z=0$, so the reduced process is driven only by additive noise. The paper states that this reduced dynamics preserves the localization property of the original system characterized by Eq. (19), and supplies explicit Wasserstein-distance bounds (Proposition 2) between the full distribution of $z$ and the distribution of $z_{\mathrm{red}}$.

Load-bearing premise

The reduction assumes that the three unresolved second moments $p_{xx}$, $p_{yy}$, and $p_{yz}$ stay locked to the resolved second moment $p_{zz}$ through a fixed affine relation, and that the numerically selected branch of the invariance equations is the one the true dynamics follows; if this manifold fails to attract the exact moment flow at finite noise strength, the reduced process and its error bounds inherit the wrong covariance.

Editorial extensions

If this is right

  • The one-dimensional process (61) reproduces the localization relaxation rate of the full model in the large-noise regime, and in the invariant-manifold version it remains valid across the whole interval $0<\epsilon\le \epsilon''_c$, beyond what adiabatic elimination offers.
  • Because the reduced process is scalar and driven by additive noise, closed-form expressions for the mean, second moment, and localization probability follow directly, avoiding simulation of the full three-variable system.
  • Both the $L^\infty$ and $L^2$ refinements select $\bar B_z=0$, so the optimal scalar surrogate carries no multiplicative noise of its own; the multiplicative structure of the original GBM is fully absorbed by the drift and the additive noise scale.
  • Proposition 2 gives an a priori, computable two-sided bound on the Wasserstein distance between the original and reduced distributions in terms of their means and variances, so the reduction comes with a quantitative error certificate.
  • For $k$-dimensional reduced descriptions the same construction requires solving a matrix Lyapunov equation and the corresponding invariance equations, which is feasible numerically but no longer closed-form.

Reading between the lines

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

  • An extension the paper leaves implicit is that the same recipe applies to any scalar observable of a GBM whose slow drift and stationary second moment can be computed: minimal additive driving is the universal way to match a given equilibrium variance and a prescribed relaxation rate, without carrying over multiplicative noise.
  • The appearance of the critical values $\epsilon'_c$ and $\epsilon''_c$ suggests a testable prediction the paper does not pursue: near those noise strengths the reduced model should lose quantitative accuracy precisely because the invariant-manifold branch ceases to be real, and measured localization relaxation should show a qualitative change there.
  • In the editor's reading, the affine second-moment closure is the part most likely to limit the scheme in practice, and the paper's own Fig. 1 shows that the analogous adiabatic closure fails for moderate noise, so a finite-$\epsilon$ check of the ansatz against the exact moment equations would be the natural stress test before porting the method to higher-dimensional observables.
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

2 major / 5 minor

Summary. The paper develops a model-reduction framework for multivariate geometric Brownian motions with multiplicative noise, combining invariant-manifold (IM) closures and adiabatic elimination for the deterministic drift with a fluctuation-dissipation condition for the noise. The framework is applied to a two-state quantum model whose localization variable z satisfies the three-dimensional GBM system (12). The authors derive reduced ODEs for the mean and second moment of z (Eqs. (31c), (37), (45), (49)), assemble a one-dimensional surrogate SDE dz_red = A_bar_z z_red dt + D_bar_z du (Eq. (61)), and claim that this reduced process preserves the localization properties of the original dynamics. Section 5.3 then states Wasserstein error bounds between the laws of z(t) and z_red(t).

Significance. If the error estimates were correct, the paper would provide a useful, non-fitted reduction scheme for multiplicative-noise GBMs: the reduced drift and diffusion coefficients are derived from invariance and Lyapunov-type conditions rather than calibrated to the target solution, and the Chapman-Enskog expansion (51) correctly reproduces the adiabatic limit. The quantum localization application is also physically meaningful. The deterministic part of the reduction is internally consistent: the invariance equations (47) follow from the closure ansatz, and the large-beta asymptotics of the exact spectrum match the adiabatic chain. However, as detailed below, the main quantitative error statement (Proposition 2) is tied to the wrong covariance for the actual reduced process, so the current manuscript overstates the rigor of its central claim.

major comments (2)
  1. [Section 4.5, Eqs. (46)-(49)] The proof of Proposition 2 identifies the covariance of z_red(t) with p_bar_zz(t)-z_bar(t)^2, where p_bar_zz is the IM/adiabatic second moment. But the reduced SDE actually derived in Section 5.2 is (61), with B_z=0, and its second moment solves d p_tilde/dt = 2 A_bar_z (p_tilde - 1/3), so p_tilde(t)=1/3+(p_tilde_0-1/3)e^{2 A_bar_z t}. Since 2 A_bar_z is generally different from 4 alpha a_3^*(epsilon) (for instance, in the large-beta limit 2A_bar_z ≈ -4 alpha^2/beta^2 while 4 alpha a_3^* ≈ -6 alpha^2/beta^2), p_bar_zz is not the second moment of z_red. Therefore the stated Wasserstein bounds apply to an auxiliary law with covariance p_bar_zz, not to Law(z_red(t)). The bounds need to be recomputed with Sigma' = p_tilde - z_bar^2; the same Lemma 4 argument would yield corrected bounds, but Proposition 2 as written is incorrect.
  2. [Section 4.5, Eqs. (46)-(49)] The IM reduction of the variance rests on the affine closure ansatz (46) and on selecting the branch a^* by continuity and the condition (48). The paper only states that 'numerically, we identify two such solutions' and gives no existence or uniqueness argument for the physically relevant branch over (0, epsilon''_c). Since Eq. (49) and the reduced second moment used in Section 5.2 feed on this branch, a proof or a precise continuation argument for a^* is needed to justify the IM part of the reduction, not merely the numerical evidence in Fig. 2.
minor comments (5)
  1. [Section 5.2, Eq. (57)] The IM drift is written as -(beta^2 - sqrt(beta^4 - 4 alpha^2)), whereas Eq. (45) gives 2 epsilon alpha a_+(epsilon) = -(beta^2 - sqrt(beta^4 - 4 epsilon^2 alpha^2)). The epsilon dependence (or the precise rescaling convention) should be restated here to make the two equations match.
  2. [Section 5.2, before Eq. (59)] The phrase 'The second-moment equation for \bar z' should read 'for \tilde p_{zz}' or 'for the second moment', since the quantity being evolved is not the mean \bar z.
  3. [Section 5.3, Lemma 4] Lemma 4 is quoted from [23], which concerns Gaussian measures, while Proposition 2 applies it to arbitrary probability measures. The two-sided estimate (64) should either be proved for general laws or attributed to a source that states it in that generality.
  4. [Abstract and Conclusion] The text sometimes says 'multivariate Brownian motions' where 'geometric Brownian motions' is meant; please align the terminology throughout, including the abstract and the opening of the Conclusion.
  5. [Appendix E] There is a typo, 'straightofrwardly' for 'straightforwardly', which should be corrected.

Circularity Check

1 steps flagged · score 4.0 of 10

Core reduction is derived from the original dynamics rather than fitted, but Proposition 2's error bounds substitute the input closure moment p_zz-bar for the covariance of the constructed SDE (61), making the main quantitative claim circular in part.

  1. fitted input called prediction [Section 5.3, Proposition 2 (proof); Section 5.2 Eqs. (57)-(61); Eq. (49); minimization (60)]
    "We apply Lemmas 1 and 2 for a= Ã, b= 2 Āz + B̄z^2 ≥ 2 Āz =: b0 and deduce that the optimal solution for the minimization problem (60) is obtained at b=b0, that is B̄z = 0. ... Σ′ = Σ′(t) = E(z red(t)^2) − (⟨z red(t)⟩)^2 = p̄zz (t) − z̄(t)^2."

    With B̄z = 0, the second moment of the reduced SDE (61) solves dp̃/dt = 2Āz(p̃ − 1/3), so E(z_red^2) = 1/3 + (p̃0 − 1/3) e^{2Āz t}. The closure moment p̄zz used in Proposition 2 solves (49) with rate 4αa3*, and these rates are not equal (adiabatic case: 2Āz ≈ −4α²/β² vs 4αa3* ≈ −6α²/β² by (51); IM case likewise). Hence p̄zz ≠ p̃ for finite t. The proof nevertheless identifies the covariance of the predicted output process z_red with p̄zz — the same closure input that enters the minimization (60) and is plotted in Fig. 3 — instead of with p̃. Consequently the stated W2 bounds are, by construction, bounds between Law(z) and an auxiliary law with covariance p̄zz, not between Law(z) and Law(z_red) of Eq. (61). The paper's quantitative evidence for the claim that Eq.

full rationale

Walk-through of the derivation chain: (1) Deterministic reduction (Sections 3-4) is derived, not fitted: the reduced drift Āz is obtained by solving the invariance equation (43) (or via adiabatic elimination of the exact ODEs (16)/(40)), and the affine closure (46) coefficients solve the algebraic system (47) with branch rule (48), giving p̄zz in (49). The small-ε expansion (51) independently recovers the adiabatic limit. (2) Stochastic reduction (Sections 5.1-5.2): D̄z = sqrt(−(2/3)Āz) is fixed by the generalized fluctuation-dissipation/Lyapunov relation (59) — enforcing the stationary marginal second moment 1/3 computed from the original system in Section 2.5 — together with the explicit L∞-minimization (60) solved in Lemmas 1-2. No parameter is calibrated to the target localization behavior; the mean-decay rate is reproduced because Āz is the dominant eigenvalue of the original deterministic dynamics (45), and the stationary variance is matched by (59). These are genuine derived properties, not tautologies. (3) Self-citations [15,16,17,20] (Colangeli/Duong/Muntean, overlapping with the present authors) frame the approach and rescaling convention, but the entire scheme is re-derived in Sections 4-5 and Appendix B, so they are not load-bearing; no uniqueness theorem is imported from prior work. (4) The flagged defect: Proposition 2 substitutes the closure moment p̄zz for the covariance of z_red, whereas by the paper's own Section 5.2 the covariance of the SDE (61) is p̃(t) = 1/3 + (p̃0 − 1/3)e^{2Āz t}, which differs from p̄zz at finite times. Thus the stated W2 bounds and the Fig. 3 validation (which plots p̄zz, not p̃) apply to the input closure, not to the actual reduced process. This is a quotable, by-construction identification of the claimed output property with an input quantity, giving partial circularity of the central error claim — but the reduced SDE itself is independently derived, so the paper is not fundamentally circular. (5) Correctness caveats, not circularity: the affine closure (46)/branch (48) is selected numerically without a proof of existence or uniqueness over (0, ε''c), and Fig. 1 shows the adiabatic closure deviating significantly at moderate β; these limit the validity of the ansatz but do not make the derivation circular. Overall score 4: the central construction is self-contained, while the main quantitative error estimate is stated for the construction input rather than the output process.

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

No fitted constants appear anywhere in the scheme: alpha and beta are physical parameters inherited from the model of Ref. [7], epsilon is a rescaling device, and the IM closure coefficients (a1, a2, a3) are determined by solving the algebraic invariance equations (47) plus a continuity and branch rule (48), not by data fitting. The load-bearing ingredients are therefore assumptions rather than fitted numbers: the affine closure ansatz, the spectral-gap condition, and the assumed reduced-SDE class.

assumptions (6)
  • standard math Ito calculus applies and the SDE system has unique strong solutions.
    Used throughout Section 2 (Eqs. (8)-(16)) and Appendix B.1 to close the mean and second-moment equations; standard for linear SDEs with additive coupling.
  • domain assumption Unit trace pxx + pyy + pzz = 1 is preserved (unitary normalization).
    Inherited from the Schrodinger model (Section 2.1); used in Section 2.5 to fix p_inf = 1/3 and in Appendix D to reduce the moment system to (83).
  • ad hoc to paper Affine closure ansatz for the IM second-moment reduction (Eq. (46)).
    Section 4.5 assumes pxx, pyy, pyz are affine functions of (pzz - 1/3); this is a modeling choice, not implied by the dynamics, and it defines the reduced second moment p_bar_zz.
  • ad hoc to paper Branch selection rule: continuity in epsilon and lim epsilon->0 of epsilon a3(epsilon) = 0 (Eq. (48)).
    Among solutions of the polynomial system (47), the physically relevant branch is chosen by this rule and by numerical continuation; no existence or uniqueness proof over (0, epsilon''_c) is given.
  • domain assumption Spectral separation epsilon < min(epsilon'_c, epsilon''_c); real and distinct eigenvalues of Q_epsilon and M_epsilon.
    Sections 4.4-4.5 and Appendices C and E: the invariant manifold and the reduced description break down when eigenvalues merge or become complex.
  • ad hoc to paper The reduced process class (58) with independent Wiener noises is sufficient.
    Section 5.1-5.2 assume a mixed OU plus GBM with independent w and u; the original system has one shared noise, so the surrogate's noise structure is matched only through the covariance or Lyapunov equation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Model Reduction of Multivariate Geometric Brownian Motions and Localization in a Two-State Quantum System." pith.science (2026). https://pith.science/paper/I4CFUEDJ

@misc{pith2026250709413,
  author       = {Pith},
  title        = {Pith review of: Model Reduction of Multivariate Geometric Brownian Motions and Localization in a Two-State Quantum System},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/I4CFUEDJ}},
  note         = {Machine review of arXiv:2507.09413}
}
read the original abstract

We develop a systematic framework for the model reduction of multivariate geometric Brownian motions (GBMs), a fundamental class of stochastic processes with broad applications in mathematical finance, population biology, and statistical physics. Our approach leverages the interplay between the method of invariant manifolds and adiabatic elimination to derive closed-form reduced equations for the deterministic drift. An extended formulation of the fluctuation-dissipation theorem is subsequently employed to characterize the stochastic component of the reduced description. As a concrete application, we apply our reduction scheme to a GBM arising from a two-state quantum system, showing that the reduced dynamics accurately capture the localization properties of the original model while significantly simplifying the analysis.

Figures

Figures reproduced from arXiv: 2507.09413 by the authors.

Figure 1
Figure 1. Time evolution of pzz(t), computed from the the third-order ODE (34) (black solid line) and from the adiabatic elimination (37) (black dashed line), with initial conditions pzz(0) = 0, p′ zz(0) = 0, p′′ zz(0) = 0. The dotted horizontal line marks the asymptotic value 1/3. The parameters are set to the values α = 0.5 and β 2 = 2 (left panel) and β 2 = 100 (right panel). In Section 4, we contrast this strategy with an… view at source ↗
Figure 2
Figure 2. Behavior of the eigenvalues of the matrix [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗
Figure 3
Figure 3. Time evolution of pzz(t) computed from the original system (41) (black solid line), and time evolution of the reduced dynamics ˜pzz(t) computed from the invariant manifold reduction (49) (gray short-dashed line) and from the adiabatic elimination (50) (black dashed line), with initial condition pzz(0) = 0. Results are shown for ϵ = 0.5 < ϵ′′ c (left panel) and ϵ = 0.01 (right panel). The inset in the right panel off… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

80 extracted references · 78 canonical work pages

  1. [1]

    A. C. Antoulas.Approximation of Large-Scale Dynamical Systems. SIAM, 2005

    work page 2005

  2. [2]

    V. I. Arnold. Proof of a theorem of A.N. Kolmogorov on the invariance of quasi-periodic motions under small perturbations of the Hamiltonian.Russian Math Surveys, 18:9–36, 1963

    work page 1963

  3. [3]

    Benner and T

    P. Benner and T. Damm. Lyapunov equations, energy functionals, and model order reduction of bilinear and stochastic systems.SIAM journal on control and optimization, 49(2):686–711, 2011

    work page 2011

  4. [4]

    Benner, M

    P. Benner, M. Ohlberger, A. Cohen, and K. Willcox.Model reduction and approximation: theory and algorithms. SIAM, 2017

    work page 2017

  5. [5]

    Benner and M

    P. Benner and M. Redmann. Model reduction for stochastic systems.Stochastic Partial Differential Equations: Analysis and Computations, 3:291–338, 2015

    work page 2015

  6. [6]

    Bianconi, A

    G. Bianconi, A. Arenas, J. Biamonte, L. D. Carr, B. Kahng, J. Kertesz, J. Kurths, L. L¨ u, C. Masoller, A. M. E, et al. Complex systems in the spotlight: next steps after the 2021 nobel prize in physics.Journal of Physics: Complexity, 4(1):010201, 2023

    work page 2021

  7. [7]

    Blanchard, G

    P. Blanchard, G. Bolz, M. Cini, G. De Angelis, and M. Serva. Localization stabilized by noise.Journal of statistical physics, 75:749–755, 1994

    work page 1994

  8. [8]

    Blanchard, M

    P. Blanchard, M. Pasquini, and M. Serva. Classical limit: Localization induced by noise. In The Foundations Of Quantum Mechanics, pages 63–72. World Scientific, 2000

    work page 2000

Show all 80 references
  1. [9]

    Blanchard, M

    P. Blanchard, M. Pasquini, and M. Serva. Effective localization induced by noise and nonlin- earity.Physica D: Nonlinear Phenomena, 141(3-4):214–220, 2000

    2000

  2. [10]

    A. V. Bobylev. Instabilities in the Chapman-Enskog Expansion and Hyperbolic Burnett Equations.Journal of Statistical Physics, 124(2):371–399, 2006

    2006

  3. [11]

    N. N. Bogoliubov.Problems of Dynamical Theory in Statistical Physics. Studies in Statistical Mechanics, 1960

    1960

  4. [12]

    Chapman and T

    S. Chapman and T. G. Cowling.The Mathematical Theory of Non-Uniform Gases. Cambridge University Press, 1990

    1990

  5. [13]

    M. D. Checkroun, A. Tantet, H. A. Dijkstra, and J. D. Neelin. Ruelle-Pollicott Resonances of Stochastic Systems in Reduced State Space. Part I: Theory.Journal of Statistical Physics, 179:1366–1402, 2020

    2020

  6. [14]

    A. J. Chorin, O. Hald, and R. Kupferman. Optimal prediction with memory.Physica D: Nonlinear Phenomena, 166:239–257, 2000

    2000

  7. [15]

    Colangeli, M

    M. Colangeli, M. H. Duong, and A. Muntean. A reduction scheme for coupled Brownian harmonic oscillators.Journal of Physics A: Mathematical and Theoretical, 55:505002, 2022. 28

    2022

  8. [16]

    Colangeli, M

    M. Colangeli, M. H. Duong, and A. Muntean. Model reduction of brownian oscillators: quantification of errors and long-time behavior.Journal of Physics A: Mathematical and Theoretical, 56(34):345003, 2023

    2023

  9. [17]

    Colangeli, M

    M. Colangeli, M. H. Duong, and A. Muntean. A hybrid approach to model reduction of Generalized Langevin Dynamics.Journal of Statistical Physics, 192(2):1–20, 2025

    2025

  10. [18]

    Colangeli, I

    M. Colangeli, I. V. Karlin, and M. Kr¨ oger. From hyperbolic regularization to exact hydrody- namics for linearized Grad’s equations.Phys. Rev. E, 75:051204, 2007

    2007

  11. [19]

    Colangeli, M

    M. Colangeli, M. Kr¨ oger, and H. C. ¨Ottinger. Boltzmann equation and hydrodynamic fluc- tuations.Phys. Rev. E, 80:051202, 2009

    2009

  12. [20]

    Colangeli and A

    M. Colangeli and A. Muntean. Reduced markovian descriptions of brownian dynamics: To- ward an exact theory.Frontiers in Physics, 10, 2022

    2022

  13. [21]

    Colangeli, L

    M. Colangeli, L. Rondoni, and A. Vulpiani. Fluctuation-dissipation relation for chaotic non-hamiltonian systems.Journal of Statistical Mechanics: Theory and Experiment, 2012(04):L04002, apr 2012

    2012

  14. [22]

    Damm and P

    T. Damm and P. Benner. Balanced truncation for stochastic linear systems with guaranteed error bound. In21st International Symposium on Mathematical Theory of Networks and Systems, pages 1492–1497, 2014

    2014

  15. [23]

    Dowson and B

    D. Dowson and B. Landau. The Fr´ echet distance between multivariate normal distributions. Journal of multivariate analysis, 12(3):450–455, 1982

    1982

  16. [24]

    Duong, A

    M. Duong, A. Lamacz, M. A. Peletier, and U. Sharma. Variational approach to coarse-graining of generalized gradient flows.Calc. Var. Partial Differ. Equ., 56:1–65, 2017

    2017

  17. [25]

    M. H. Duong, C. Hartmann, and M. Ottobre. Coarse graining of stochastic differential equations: averaging and projection method.arXiv preprint arXiv:2506.14939, 2025

    2025 arXiv

  18. [26]

    M. H. Duong, A. Lamacz, M. A. Peletier, A. Schlichting, and U. Sharma. Quantification of coarse-graining error in Langevin and overdamped Langevin dynamics.Nonlinearity, 31(10):4517–4566, aug 2018

    2018

  19. [27]

    Fenichel

    N. Fenichel. Geometric singular perturbation theory for ordinary differential equations.Jour- nal of differential equations, 31(1):53–98, 1979

    1979

  20. [28]

    J.-M. Ginoux. The slow invariant manifold of the lorenz–krishnamurthy model.Qualitative Theory of Dynamical Systems, 13(1):19–37, 2014

    2014

  21. [29]

    Giordano, F

    S. Giordano, F. Cleri, and R. Blossey. Infinite ergodicity in generalized geometric brownian motions with nonlinear drift.Physical Review E, 107(4):044111, 2023

    2023

  22. [30]

    Givon, R

    D. Givon, R. Kupferman, and A. Stuart. Extracting macroscopic dynamics: model problems and algorithms.Nonlinearity, 17(6):R55, 2004

    2004

  23. [31]

    A. Gorban. Model reduction in chemical dynamics: slow invariant manifolds, singular per- turbations, thermodynamic estimates, and analysis of reaction graph.Current Opinion in Chemical Engineering, 21:48–59, 2018

    2018

  24. [32]

    A. N. Gorban and I. V. Karlin.Invariant Manifolds for Physical and Chemical Kinetics, volume 660 ofLect. Notes Phys.Springer-Verlag, Berlin, 2005

    2005

  25. [33]

    A. N. Gorban and I. V. Karlin. Hilbert’s 6th problem: Exact and approximate manifolds for kinetic equations.Bulletin of the American Mathematical Society, 51:187–246, 2013. 29

    2013

  26. [34]

    A. N. Gorban, I. V. Karlin, and A. Y. Zinovyev. Constructive methods of invariant manifolds for kinetic problems.Physics Reports, 396(4):197–403, 2004

    2004

  27. [35]

    A. N. Gorban, N. K. Kazantzis, I. G. Kevrekidis, H. C. ¨Ottinger, and C. Theodoropoulos. Model reduction and coarse-graining approaches for multiscale phenomena. Springer, 2006

    2006

  28. [36]

    M. S. Guti´ errez, V. Lucarini, M. D. Chekroun, and M. Ghil. Reduced-order models for coupled dynamical systems: Data-driven methods and the Koopman operator.Chaos, 31:053116, 2021

    2021

  29. [37]

    Haken.Synergetics: An Introduction

    H. Haken.Synergetics: An Introduction. Springer, 1983

    1983

  30. [38]

    Hamm and A

    K. Hamm and A. Korzeniowski. On Wasserstein distances for affine transformations of random vectors.arXiv preprint arXiv:2310.03945, 2023

    2023 arXiv

  31. [39]

    H¨ anggi and H

    P. H¨ anggi and H. Thomas. Stochastic processes: Time evolution, symmetries and linear response.Physics Reports, 88(4):207–319, 1982

    1982

  32. [40]

    Hartmann, L

    C. Hartmann, L. Neureither, and U. Sharma. Coarse graining of nonreversible stochastic differential equations: Quantitative results and connections to averaging.SIAM Journal on Mathematical Analysis, 52(3):2689–2733, 2020

    2020

  33. [41]

    Hartmann, C

    C. Hartmann, C. Sch¨ utte, and W. Zhang. Model reduction algorithms for optimal control and importance sampling of diffusions.Nonlinearity, 29(8):2298–2326, 2016

    2016

  34. [42]

    Hasselmann

    K. Hasselmann. Stochastic climate models Part I. Theory.Tellus, 28(6):473–485, 1976

    1976

  35. [43]

    I. V. Karlin, M. Colangeli, and M. Kr¨ oger. Exact linear hydrodynamics from the Boltzmann equation.Phys. Rev. Lett., 100:214503, 2008

    2008

  36. [44]

    I. V. Karlin and A. N. Gorban. Hydrodynamics from Grad’s equations: What can we learn from exact solutions?Annalen der Physik, 11:783–833, 2002

    2002

  37. [45]

    I. G. Kevrekidis, C. W. Gear, J. M. Hyman, et al. Equation-free, coarse-grained multiscale computation: enabling microscopic simulators to perform system-level analysis.Communi- cations in Mathematical Sciences, 1(4):715–762, 2003

    2003

  38. [46]

    Kirsten and V

    G. Kirsten and V. Simoncini. Order reduction methods for solving large-scale differential matrix riccati equations.SIAM Journal on Scientific Computing, 42(4):A2182–A2205, 2020

    2020

  39. [47]

    Kloeden and E

    P. Kloeden and E. Platen.Numerical Solution of Stochastic Differential Equations. Stochastic Modelling and Applied Probability. Springer Berlin Heidelberg, 2013

    2013

  40. [48]

    A. N. Kolmogorov. On conservation of conditionally periodic motions under small perturba- tions of the Hamiltonian.Dokl. Akad. Nauk SSSR, 98:527–530, 1954

    1954

  41. [49]

    R. Kubo. The fluctuation-dissipation theorem.Rep. Prog. Phys., 29:255, 1966

    1966

  42. [50]

    R. Kubo, M. Toda, and N. Hashitsume.Statistical Physics II. Nonequilibrium Statistical Mechanics. Springer-Verlag, Berlin, 1985

    1985

  43. [51]

    Kuˇ cera

    V. Kuˇ cera. A review of the matrix riccati equation.Kybernetika, 9(1):42–61, 1973

    1973

  44. [52]

    S. H. Lam and D. A. Goussis. Dynamic simplification of reaction mechanisms.Combustion Science and Technology, 89(5-6):375–404, 1994

    1994

  45. [53]

    Legoll and T

    F. Legoll and T. Leli` evre. Effective dynamics using conditional expectations.Nonlinearity, 23(9):2131, 2010

    2010

  46. [54]

    Legoll, T

    F. Legoll, T. Leli` evre, and U. Sharma. Effective dynamics for non-reversible stochastic dif- ferential equations: a quantitative study.Nonlinearity, 32(12):4779–4816, oct 2019. 30

    2019

  47. [55]

    Leli` evre and W

    T. Leli` evre and W. Zhang. Pathwise estimates for effective dynamics: The case of nonlinear vectorial reaction coordinates.Multiscale Modeling & Simulation, 17(3):1019–1051, 2019

    2019

  48. [56]

    Lu and E

    J. Lu and E. Vanden-Eijnden. Exact dynamical coarse-graining without time-scale separation. The Journal of Chemical Physics, 141(4):044109, 2014

    2014

  49. [57]

    Lucarini and M

    V. Lucarini and M. D. Chekroun. Theoretical tools for understanding the climate crisis from hasselmann’s programme and beyond.Nature Reviews Physics, 5(12):744–765, 2023

    2023

  50. [58]

    C. Maes. On the Second Fluctuation–Dissipation Theorem for Nonequilibrium Baths.Journal of Statistical Physics, 154(3):705–722, 2014

    2014

  51. [59]

    A. J. Majda, I. Timofeyev, and E. Vanden-Eijnden. Mathematical strategies for filtering turbulent dynamical systems.Journal of the Atmospheric Sciences, 58(6):913–930, 2001

    2001

  52. [60]

    Mao.Stochastic differential equations and applications

    X. Mao.Stochastic differential equations and applications. Elsevier, 2007

    2007

  53. [61]

    U. B. M. Marconi, A. Puglisi, L. Rondoni, and A. Vulpiani. Fluctuation-Dissipation: Response Theory in Statistical Physics.Phys. Rep., 461:111–195, 2008

    2008

  54. [62]

    H. Mori. Transport, collective motion, and brownian motion.Progress of Theoretical Physics, 33(3):423–455, 1965

    1965

  55. [63]

    J. Moser. Convergent series expansions for quasi-periodic motions.Math. Ann., 169:136–176, 1967

    1967

  56. [64]

    Ott and T

    E. Ott and T. M. Antonsen. Low dimensional behavior of large systems of globally coupled oscillators.Chaos: An Interdisciplinary Journal of Nonlinear Science, 18(3):037113, 2008

    2008

  57. [65]

    Pavliotis and A

    G. Pavliotis and A. Stuart.Multiscale methods: averaging and homogenization. Springer Science & Business Media, 2008

    2008

  58. [66]

    G. A. Pavliotis.Stochastic Processes and Applications. Diffusion Processes, the Fokker-Planck and Langevin Equations. Springer-Verlag, 2014

    2014

  59. [67]

    M. Redmann. Bilinear systems—a new link to\mathcalh 2-norms, relations to stochastic systems, and further properties.SIAM Journal on Control and Optimization, 59(4):2477– 2497, 2021

    2021

  60. [68]

    Risken.The Fokker-Planck Equation

    H. Risken.The Fokker-Planck Equation. Springer Verlag, Berlin, 1996

    1996

  61. [69]

    A. J. Roberts. Model reduction by invariant manifold theory.Lecture Notes, 2008. Available online

    2008

  62. [70]

    A. J. Roberts.Model Emergent Dynamics in Complex Systems. SIAM, 2015

    2015

  63. [71]

    C. W. Rowley. Model reduction for fluids using balanced proper orthogonal decomposition. International Journal of Bifurcation and Chaos, 15(3):997–1013, 2005

    2005

  64. [72]

    T. J. Snowden, P. H. van der Graaf, and M. J. Tindall. Methods of model reduction for large- scale biological systems: a survey of current methods and trends.Bulletin of Mathematical Biology, 79(7):1449–1486, 2017

    2017

  65. [73]

    S. V. Stasenko and A. N. Kirdin. Preservation of dissipativity in dimensionality reduction. Communications in Nonlinear Science and Numerical Simulation, 142:108553, 2025

    2025

  66. [74]

    Tantet, M

    A. Tantet, M. D. Checkroun, H. A. Dijkstra, and J. D. Neelin. Ruelle-Pollicott Resonances of Stochastic Systems in Reduced State Space. Part II: Stochastic Hopf Bifurcation.Journal of Statistical Physics, 179:1403–1448, 2020. 31

    2020

  67. [75]

    Villamaina, A

    D. Villamaina, A. Baldassarri, A. Puglisi, and A. Vulpiani. The fluctuation-dissipation re- lation: how does one compare correlation functions and responses?Journal of Statistical Mechanics: Theory and Experiment, 2009(07):P07024, jul 2009

    2009

  68. [76]

    Wouters and G

    J. Wouters and G. A. Gottwald. Edgeworth expansions for slow-fast systems with finite time- scale separation.Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 475(2223):20180358, 2019

    2019

  69. [77]

    Wouters and G

    J. Wouters and G. A. Gottwald. Stochastic model reduction for slow-fast systems with moderate time scale separation.Multiscale Modeling & Simulation, 17(4):1172–1188, 2019

    2019

  70. [78]

    Zhang, C

    W. Zhang, C. Hartmann, and C. Sch¨ utte. Effective dynamics along given reaction coordinates, and reaction rate theory.Faraday discussions, 195:365–394, 2016

    2016

  71. [79]

    R. Zwanzig. Ensemble method in the theory of irreversibility.The Journal of Chemical Physics, 33(5):1338–1341, 1960

    1960

  72. [80]

    Zwanzig.Nonequilibrium Statistical Mechanics

    R. Zwanzig.Nonequilibrium Statistical Mechanics. Oxford University Press, 2001. 32

    2001

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.