Pith. sign in

REVIEW 3 major objections 4 minor 41 references

Euler-type methods for Levy-driven McKean-Vlasov SDEs with super-linear coefficients: mean-square error analysis

T0 review · 3 major / 4 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read This paper proves that a unified family of explicit Euler-type schemes—tanh, tamed, sine, and mixed—approximate Lévy-driven McKean–Vlasov particle systems with super-linear coefficients at mean-square order arbitrarily close to 1/2, without

desk verdict Solid unified tamed/tanh/sine Euler framework for Lévy-driven McKean–Vlasov, but the advertised 'arbitrarily close to 1/2' rate is contingent on a high-moment condition that Theorem 3.1 omits and the paper's own Example 4.1 fails for small ε. read the letter →

arxiv 2509.09302 v1 pith:BZ4UMMGL submitted 2025-09-11 math.NA cs.NA

Jingtao Zhu , Yuying Zhao , Siqing Gan This is my paper · ORCID
classification math.NAcs.NA MSC 65C3060H35
keywords McKean-VlasovSDELévynoisesuper-linearcoefficientsEuler-typeschemestamedEulermethodmean-squareconvergenceinteractingparticlesystemspropagationofchaos
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

The paper sets out to show that a single class of explicit Euler-type schemes can reliably simulate the interacting particle system associated with a Lévy-driven McKean–Vlasov SDE even when drift, diffusion, and jump coefficients grow faster than linearly. The schemes are built by passing each coefficient through a bounded nonlinear transformation—tanh, taming, sine, or a mix—before the Euler update. The main theorem states that the mean-square error against the true particle system is at most $C \Delta t^{2/(2+\varepsilon)}$ for any $\varepsilon>0$, so the convergence order is arbitrarily close to $\frac{1}{2}$. This matters because prior explicit methods needed a numerical coercivity condition to control moments; the new analysis removes that restriction and covers all four schemes in one framework. If correct, it gives practitioners a choice of explicit, stable, easy-to-implement integrators for superlinear jump models.

What carries the argument

The machinery is a family of transformation operators $\Gamma_l(F,\Delta t)$ applied to drift, diffusion, and jump coefficients before each Euler update. Assumption 3.1 requires $|\Gamma_l(F,\Delta t)| \leq C \Delta t^{-\alpha_l} \wedge |F|$, which caps the magnitude of large coefficients and prevents moment explosion; Assumption 3.3 quantifies the consistency error as $|\Gamma_l(F,\Delta t)-F| \leq C \Delta t^{\delta_l}|F|^{\gamma_l}$. The moment-bound proof chooses a truncation radius $R(\Delta t)=\Delta t^{-1/G}$ so that interactions between superlinear growth and the bounded transform contribute at most $O(\Delta t)$, and the final error estimate uses an enhanced one-sided Lipschitz condition (Assumption 3.4, $\eta>1$) to control the Itô cross terms. For the tanh, tamed, sine, and mixed ex

What would settle it

Take a one-dimensional Lévy-driven McKean–Vlasov SDE satisfying all stated assumptions but with growth parameters chosen so that the Lemma B.3 lower bound on $\bar p$ exceeds the value for which Assumption 2.3 can be verified; simulate the tanh-Euler scheme and measure the empirical $L^2$ error at $T=1$ over many paths. If the observed rate drops below $1/2 - \varepsilon$ for a fixed small $\varepsilon$ while the theorem's hypotheses are claimed to hold, the central claim is refuted; conversely, checking that the moment threshold is actually met would isolate the gap between Theorem 3.1 and its proof.

Watch

Extended reading notes

Core claim

For the interacting particle system (2.6) approximating the Lévy-driven McKean–Vlasov SDE (1.1), the continuous-time Euler-type scheme (3.4) satisfies $\sup_i \sup_t E|X^{i,N}(t)-Y^{i,N}(t)|^2 \leq C \Delta t^{2/(2+\varepsilon)} (1+(E|Y_0|^{2\bar p})^\beta)$. The proof combines a local-event truncation that yields uniform moment bounds with an Itô-formula error decomposition; the key estimate (Lemma B.3) controls the difference between each coefficient and its transformed value at the previous grid point. Since $\delta_l=1$ and $\gamma_l=2$ for the tanh, tamed, sine, and mixed operators, all four schemes inherit the same rate, and the propagation-of-chaos bound extends the result to the McKean–Vlasov limit with an additional N^{-1/

Load-bearing premise

The whole argument rests on the initial condition having sufficiently high finite moments—exactly how high is spelled out only in an appendix lemma, not in the main theorem; if that moment threshold is not met, the stated error bound does not follow.

Editorial extensions

If this is right

  • The tanh-Euler, tamed-Euler, sine-Euler, and mixed-Euler schemes all converge in mean square with order 1/2−ε for Lévy-driven McKean–Vlasov interacting particle systems with superlinear coefficients, so practitioners can choose based on implementation convenience.
  • The full discretization error for the McKean–Vlasov SDE itself is the sum of the particle error (order N^{-1/2} in low dimension) and the time-step error (order Δt^{1/2−ε}), so the method is consistent in both N and Δt.
  • Explicit simulation of superlinear jump models no longer requires the numerical coercivity condition (3.2), removing a restrictive assumption used in earlier tamed schemes.
  • Because Poisson increments contribute O(Δt) rather than O(Δt^{1/2}) once accumulated in these estimates, the near-1/2 rate is the natural ceiling for the general case; improved regularity can raise it to exactly 1/2.
  • The numerical experiments on a jump-extended 3/2-volatility model and a double-well model show empirical rates close to 1/2, consistent with the theorem.

Reading between the lines

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

  • Inference: the same Γ-transformation framework may extend to Milstein-type or higher-order schemes for Lévy-driven McKean–Vlasov SDEs, since the consistency parameters are modular; however, higher-order rates would require refined jump estimates not given here.
  • Inference: the hidden moment-order lower bound in Lemma B.3 means practical reliability depends on coefficients whose growth parameters are modest relative to available moments; for stiff superlinear models users may need to verify high moments of the initial law.
  • Inference: because the bounded transformations are state-independent and explicit, the approach could be adapted to adaptive time stepping or random-batch approximations without changing the core moment-bound argument, though the paper does not examine these extensions.
  • Inference: the rate 1/2−ε holds for any ε>0, but the constant implicitly grows with ε^{-1}; on finite meshes with very small ε the practical rate may be hard to observe.
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

3 major / 4 minor

Summary. The paper develops a unified family of Euler-type schemes for Lévy-driven McKean–Vlasov SDEs with superlinearly growing drift, diffusion, and jump coefficients. The schemes are based on bounded nonlinear transformations (tanh, tamed, sine, and mixed variants) and are analyzed for the associated interacting particle system. The main theoretical result, Theorem 3.1, claims a mean-square error of order Δt^{2/(2+ε)} for arbitrary ε>0, i.e. a rate arbitrarily close to 1/2, under a set of monotonicity, polynomial-growth, and moment assumptions. A propagation-of-chaos corollary extends the error bound to the McKean–Vlasov equation. The proofs are detailed and follow a standard localization/truncation strategy, with lemmas in the appendix providing moment bounds and stability estimates. The paper also contains two numerical examples intended to confirm the predicted rates.

Significance. If the stated result were fully correct as written, it would be a useful contribution: it provides a general framework covering several known explicit schemes, removes a coercivity assumption that is common in earlier tamed-Euler analyses, and gives explicit mean-square rates for superlinear Lévy-driven McKean–Vlasov systems. The proof structure is self-contained and largely conventional, and the special cases in Section 3.3 are carefully verified against the abstract assumptions. However, the central convergence claim is currently over-stated: Theorem 3.1 omits a lower bound on the moment parameter \bar p that is explicitly required by the proof in Lemma B.3. Since the paper's own examples satisfy Assumption 2.3 only for finite \bar p, the advertised 'arbitrarily close to 1/2' rate is not actually delivered for those examples. The result is likely repairable by adding the missing hypothesis and reformulating the rate as a trade-off between ε and \bar p, but this is a substantive correction rather than a cosmetic one.

major comments (3)
  1. [Section 3.2, Theorem 3.1 and Lemma B.3] Theorem 3.1 states that for any ε>0, under Assumptions 2.1, 2.3, 2.5–2.7, 3.1, 3.3–3.4, the mean-square error is O(Δt^{2/(2+ε)}), with '\bar p from Assumption 2.3'. However, the proof relies on Lemma B.3, whose statement imposes the additional lower bound \bar p ≥ [γ(2+ε)/ε ∨ γ1(γ+1) ∨ γ2(γ/2+1) ∨ (γ3+γ/2)](1+\bar αG)(γ+1)+G/2. This condition is not listed in Theorem 3.1. It is load-bearing: for a fixed finite \bar p satisfying Assumption 2.3, the bound forces ε to be bounded below by a quantity of order 1/\bar p. Consequently the claimed 'arbitrarily close to 1/2' rate cannot be achieved as ε→0 unless Assumption 2.3 holds for arbitrarily large \bar p. The problem is not merely a loose constant: in Example 4.1, Appendix D verifies Assumption 2.3 only up to \bar p<300.5 (the leading Lyapunov coefficient −a1+b^2(2\bar p−1) changes sign at \bar p≈300.5), whereas Lemma B.3 requires \bar p=O(
  2. [Lemma 3.1 and Lemmas B.1–B.2 vs. Theorem 3.1] Theorem 3.1 also omits the additional moment lower bounds that the supporting lemmas require. Lemma 3.1 (Eq. (3.3)) assumes \bar p ≥ 1+(\bar α+1/2)G, and Lemmas B.1–B.2 assume the stronger condition \bar p ≥ γ+1+G(\bar αγ+\bar α+1/2). These conditions are not part of Assumption 2.3 as stated, and Theorem 3.1 does not include them. Without these bounds, the moment estimates used in the proof are not available. This is another omitted hypothesis in the main theorem, and it should be incorporated explicitly into Theorem 3.1 and Corollary 3.1.
  3. [Theorem 3.1 / Lemma B.3 proof] Even setting aside the omitted hypothesis, the proof of Lemma B.3 says the estimates hold for 'sufficient small ε>0', while Theorem 3.1 quantifies 'for any ε>0'. These quantifiers are in tension. The lower bound on \bar p in Lemma B.3 depends on ε in a way that makes 'sufficiently small ε' impossible for a fixed \bar p. The paper should either make the dependence of the error constant on ε and \bar p explicit, or rephrase the main result as a rate valid for ε in a range determined by the available moment \bar p. As written, the statement 'for any ε>0' is not justified by the proof.
minor comments (4)
  1. [Section 4, MSE definition] The reported mean-square error is computed as E|Y^{i,N,nδt}_T − Y^{i,N,nΔt}_T|, i.e. the difference between two numerical solutions with different step sizes. This measures self-convergence of the scheme, not the error E|X^{i,N}(T)−Y^{i,N}(T)| appearing in Theorem 3.1. The authors should clarify that this is a reference-solution error and, if possible, add a test with an explicitly known solution or a genuinely finer reference solution to strengthen the validation.
  2. [Section 4, numerical experiments] All simulations use a single seed (rng(100), Mersenne Twister) and no confidence intervals or repeated-seed variability are reported. Since the error curves fluctuate, the visual parallelism with the 1/2-slope line is not by itself strong evidence of the predicted rate. Reporting multiple seeds and error bars would make the numerical evidence more convincing.
  3. [Abstract and Section 3.3] The phrase 'convergence rates arbitrarily close to 1/2' is stronger than what the theorem actually establishes once the \bar p-dependence is taken into account. The rate is Δt^{1/2−ε/4} up to constants, and the admissible ε is tied to the available moment \bar p. The abstract and introduction should reflect this trade-off.
  4. [Throughout] There are minor typographical issues, including 'Itˆof' in Section 2.1 and inconsistent accent usage for 'Lévy'. The generic constant C is used with many different meanings; this is standard but the paper could benefit from a short note that constants may change from line to line.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Theorem 3.1 is proved from stated assumptions and in-paper lemmas; self-citations are motivational only.

full rationale

The central convergence claim (Theorem 3.1) is obtained in Appendix B by a chain of in-paper estimates: Lemma 3.1 supplies moment bounds for the Euler-type scheme from Assumptions 2.1, 2.3, 2.5–2.7, 3.1 and 3.2; Lemmas B.1–B.2 bound time-increment and continuous-time moments; Lemma B.3 controls the differences f−Γ1(f), g−Γ2(g), h−Γ3(h) using Assumptions 2.5, 2.7, 3.3 and the moment bounds. The final Gronwall argument uses only these estimates plus Assumption 3.4. No parameter is fitted to the target error and no 'prediction' is a renamed fitted quantity; the rate Δt^{2/(2+ε)} is an explicit consequence of Hölder interpolation, not an input. The operators in Examples 3.1–3.4 are verified in Appendix C to satisfy Assumptions 3.1 and 3.3 with α=1, δ=1, γ=2, so the theorem applies to them by checking hypotheses. The only external load-bearing result, Proposition 2.1 (propagation of chaos), is cited from [4] by Biswas et al. and is independent of the present authors; Corollary 3.1 combines it with Theorem 3.1 rather than importing the theorem's conclusion. The self-citations [19] and [40] appear only as 'Motivated by' remarks for the moment-bound technique and are not used to justify any step; the proofs are self-contained. The numerical examples compare step sizes and do not feed empirical rates back into the proof. A separate correctness concern—Theorem 3.1 states any ε>0 while Lemma B.3 imposes a ar p lower bound growing like 1/ε—is an omitted-hypothesis/quantification gap, not a circular reduction: the claimed rate is not defined in terms of the assumptions and the proof does not assume its own conclusion.

Assumptions & free parameters 0 free parameters · 12 assumptions · 0 invented entities

The central theorem is conditional on a long list of coefficient and transformation assumptions that are imposed rather than derived: monotonicity, coercivity with high \bar p, polynomial growth, time regularity, boundedness of Γ_l, and the propagation-of-chaos estimate. No free parameters are fitted to data, and no new physical or mathematical entities are introduced.

assumptions (12)
  • domain assumption Assumption 2.1: E|X_0|^{2\bar p} < ∞ for the fixed \bar p of Assumption 2.3.
    Finite high-order initial moments are required for all moment-bound arguments in Lemma 3.1 and the appendices.
  • domain assumption Assumption 2.2: coupled monotonicity for f, g, h with respect to state and measure.
    This enters existence, uniqueness, and the error analysis of the interacting particle system.
  • domain assumption Assumption 2.3: coercivity for f, g, h with moment exponent \bar p.
    Provides bounded moments of the exact solution and is used throughout the moment-bound proof.
  • domain assumption Assumption 2.4: f is uniformly continuous in y.
    Used for well-posedness of the McKean-Vlasov equation.
  • domain assumption Assumption 2.5: polynomial growth of f in y with exponent γ.
    Inherited by g and h through Assumption 2.2; sets the algebraic growth scale used in Lemma 3.1.
  • domain assumption Assumption 2.6: f, g, h are bounded when evaluated at (t,0,δ0).
    Used to convert polynomial-difference bounds into polynomial growth bounds for f, g, h.
  • domain assumption Assumption 2.7: 1/2-Hölder continuity in time for f, g, and the L1 jump integral of h.
    Controls the time-discretization error in Lemma B.3.
  • domain assumption Assumption 3.1: each Γ_l is bounded by |F_l| and by C Δt^{-α_l}.
    The boundedness of the transformed coefficients is what replaces the coercivity condition on the numerical scheme.
  • domain assumption Assumption 3.2: Γ1 approximates f with error C Δt^{\hatδ}|f|^{\hatγ}.
    Needed for the main moment-bound lemma to control the drift transformation error.
  • domain assumption Assumption 3.3: Γ_l approximates F_l with rates δ_l ≥ 1/2 and polynomial factors γ_l ≥ 1.
    This is the quantitative approximation condition that produces the near-1/2 convergence rate.
  • domain assumption Assumption 3.4: enhanced coupled monotonicity with η > 1.
    Stronger than Assumption 2.2 and is used in the final Gronwall step of the proof of Theorem 3.1.
  • standard math Well-posedness of (1.1) and the propagation-of-chaos estimate from [4, Theorem 2.1 and Proposition 3.1].
    Invoked for Corollary 3.1; the paper does not re-prove these background results.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Euler-type methods for Levy-driven McKean-Vlasov SDEs with super-linear coefficients: mean-square error analysis." pith.science (2026). https://pith.science/paper/BZ4UMMGL

@misc{pith2026250909302,
  author       = {Pith},
  title        = {Pith review of: Euler-type methods for Levy-driven McKean-Vlasov SDEs with super-linear coefficients: mean-square error analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BZ4UMMGL}},
  note         = {Machine review of arXiv:2509.09302}
}
read the original abstract

We develop and analyze a general class of Euler-type numerical schemes for Levy-driven McKean-Vlasov stochastic differential equations (SDEs), where the drift, diffusion and jump coefficients grow super-linearly in the state variable. These numerical schemes are derived by incorporating projections or nonlinear transformations into the classical Euler method, with the primary objective of establishing moment bounds for the numerical solutions. This class of schemes includes the tanh-Euler, tamed-Euler and sine-Euler schemes as special cases. In contrast to existing approaches that rely on a coercivity condition (e.g., Assumption B-1 in Kumar et al., arXiv:2010.08585), the proposed schemes remove such a restrictive assumption. We provide a rigorous mean-square convergence analysis and establish that the proposed schemes achieve convergence rates arbitrarily close to 1/2 for the interacting particle systems associated with Levy-driven McKean-Vlasov SDEs. Several numerical examples are presented to illustrate the convergence behavior and validate the theoretical results.

Figures

Figures reproduced from arXiv: 2509.09302 by the authors.

Figure 1
Figure 1. Convergence rates of Euler-type schemes for Example 4.1 [PITH_FULL_IMAGE:figures/full_fig_p012_1.png] view at source ↗
Figure 2
Figure 2. Paths of TameEM (left) and TanhEM methods (right) for Example 4.1 [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗
Figure 3
Figure 3. Paths of SineEM (left) and MixEM methods (right) for Example 4.1 [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Convergence rates of Euler-type schemes for Example 4.2 [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 5 linked inside Pith

  1. [1]

    Cambridge University Press, Cambridge, second edition, 2009

    David Applebaum.Lévy Processes and Stochastic Calculus, volume 116 ofCambridge Studies in Advanced Mathematics. Cambridge University Press, Cambridge, second edition, 2009

    2009

  2. [2]

    Mean-field description and prop- agation of chaos in networks of Hodgkin-Huxley and Fitzhugh-Nagumo neurons.The Journal of Mathe- matical Neuroscience, 2:1–50, 2012

    Javier Baladron, Diego Fasoli, Olivier Faugeras, and Jonathan Touboul. Mean-field description and prop- agation of chaos in networks of Hodgkin-Huxley and Fitzhugh-Nagumo neurons.The Journal of Mathe- matical Neuroscience, 2:1–50, 2012

    2012

  3. [3]

    Approximations of McKean-Vlasov stochastic differential equations with irregular coefficients.Journal of Theoretical Probability, 35:1187–1215, 2022

    Jianhai Bao and Xing Huang. Approximations of McKean-Vlasov stochastic differential equations with irregular coefficients.Journal of Theoretical Probability, 35:1187–1215, 2022

    2022

  4. [4]

    Well-posedness and tamed Euler schemes for McKean-Vlasov equations driven by Lévy noise.arXiv:2010.08585, 2020

    Sani Biswas, Chaman Kumar, Gonçalo dos Reis, and Christoph Reisinger. Well-posedness and tamed Euler schemes for McKean-Vlasov equations driven by Lévy noise.arXiv:2010.08585, 2020

    arXiv 2010

  5. [5]

    Milstein-type schemes for McKean-Vlasov SDEs driven by Brownian motion and Poisson random measure (with super-linear co- efficients).arXiv:2411.11759, 2025

    Sani Biswas, Chaman Kumar, Christoph Reisinger, and Verena Schwarz. Milstein-type schemes for McKean-Vlasov SDEs driven by Brownian motion and Poisson random measure (with super-linear co- efficients).arXiv:2411.11759, 2025

    arXiv 2025

  6. [6]

    A stochastic particle method for the McKean-Vlasov and the Burgers equation.Mathematics of Computation, 66(217):157–192, 1997

    Mireille Bossy and Denis Talay. A stochastic particle method for the McKean-Vlasov and the Burgers equation.Mathematics of Computation, 66(217):157–192, 1997

    1997

  7. [7]

    A flexible split-step scheme for solving McKean-Vlasov stochastic differential equations.Applied Mathematics and Computation, 427:127180, 2022

    Xingyuan Chen and Gonçalo dos Reis. A flexible split-step scheme for solving McKean-Vlasov stochastic differential equations.Applied Mathematics and Computation, 427:127180, 2022

    2022

  8. [8]

    Xingyuan Chen and Gonçalo Dos Reis. Euler simulation of interacting particle systems and McKean-Vlasov SDEs with fully super-linear growth drifts in space and interaction.IMA Journal of Numerical Analysis, 44(2):751–796, 2024

    2024

Show all 41 references
  1. [9]

    Xingyuan Chen, Gonçalo Dos Reis, and Wolfgang Stockinger. Wellposedness, exponential ergodicity and numerical approximation of fully super-linear McKean–Vlasov SDEs and associated particle systems.Elec- tronic Journal of Probability, 30:1–50, 2025

    2025

  2. [10]

    Ziheng Chen, Siqing Gan, and Xiaojie Wang. Mean-square approximations of Lévy noise driven SDEs with super-linearly growing diffusion and jump coefficients.Discrete and Continuous Dynamical Systems-Series B, 24(8):4513–4545, 2019. 14

    2019

  3. [11]

    Explicit numerical approximations for McKean-Vlasov stochastic differential equations in finite and infinite time.arXiv:2401.02878, 2024

    Yuanping Cui, Xiaoyue Li, and Liu Li. Explicit numerical approximations for McKean-Vlasov stochastic differential equations in finite and infinite time.arXiv:2401.02878, 2024

    2024 arXiv

  4. [12]

    On tamed Euler approximations of SDEs driven by Lévy noise with applications to delay equations.SIAM Journal on Numerical Analy- sis, 54(3):1840–1872, 2016

    Konstantinos Dareiotis, Chaman Kumar, and Sotirios Sabanis. On tamed Euler approximations of SDEs driven by Lévy noise with applications to delay equations.SIAM Journal on Numerical Analy- sis, 54(3):1840–1872, 2016

    2016

  5. [13]

    Shounian Deng, Chen Fei, Weiyin Fei, and Xuerong Mao. Generalized Ait-Sahalia-type interest rate model with Poisson jumps and convergence of the numerical approximation.Physica A: Statistical Mechanics and its Applications, 533:122057, 2019

    2019

  6. [14]

    Euler-Maruyama approximations for stochastic McKean-Vlasov equations with non-Lipschitz coefficients.Journal of Theoretical Probability, 34(3):1408–1425, 2021

    Xiaojie Ding and Huijie Qiao. Euler-Maruyama approximations for stochastic McKean-Vlasov equations with non-Lipschitz coefficients.Journal of Theoretical Probability, 34(3):1408–1425, 2021

    2021

  7. [15]

    Simulation of McKean-Vlasov SDEs with super- linear growth.IMA Journal of Numerical Analysis, 42(1):874–922, 2022

    Gonçalo dos Reis, Stefan Engelhardt, and Greig Smith. Simulation of McKean-Vlasov SDEs with super- linear growth.IMA Journal of Numerical Analysis, 42(1):874–922, 2022

    2022

  8. [16]

    Cambridge University Press, 2015

    Jinqiao Duan.An Introduction to Stochastic Dynamics, volume 51. Cambridge University Press, 2015

    2015

  9. [17]

    Qian Guo, Jie He, and Lei Li. Convergence analysis of an explicit method and its random batch approx- imation for the McKean-Vlasov equations with non-globally Lipschitz conditions.ESAIM: Mathematical Modelling and Numerical Analysis, 58(2):639–671, 2024

    2024

  10. [18]

    Strong and weak divergence in finite time of Euler’s method for stochastic differential equations with non-globally Lipschitz continuous coefficients

    Martin Hutzenthaler, Arnulf Jentzen, and Peter E Kloeden. Strong and weak divergence in finite time of Euler’s method for stochastic differential equations with non-globally Lipschitz continuous coefficients. Proceedings of the Royal Society A: Mathematical, Physical and Engin...

    2011

  11. [19]

    On modified Euler methods for McKean-Vlasov stochastic differential equations with super-linear coefficients.Automatica, 177:112284, 2025

    Jiamin Jian, Qingshuo Song, Xiaojie Wang, Zhongqiang Zhang, and Yuying Zhao. On modified Euler methods for McKean-Vlasov stochastic differential equations with super-linear coefficients.Automatica, 177:112284, 2025

    2025

  12. [20]

    Initiation of slime mold aggregation viewed as an instability.Journal of theoretical biology, 26(3):399–415, 1970

    Evelyn F Keller and Lee A Segel. Initiation of slime mold aggregation viewed as an instability.Journal of theoretical biology, 26(3):399–415, 1970

    1970

  13. [21]

    Jump-adapted discretization schemes for Lévy-driven SDEs

    Arturo Kohatsu-Higa and Peter Tankov. Jump-adapted discretization schemes for Lévy-driven SDEs. Stochastic Processes and their Applications, 120(11):2258–2285, 2010

    2010

  14. [22]

    On Milstein-type scheme for SDE driven by Lévy noise with super-linear coefficients

    Chaman Kumar. On Milstein-type scheme for SDE driven by Lévy noise with super-linear coefficients. Discrete and Continuous Dynamical Systems-Series B, 26(3), 2021

    2021

  15. [23]

    On explicit Milstein-type scheme for McKean-Vlasov stochastic differential equations with super-linear drift coefficient.Electronic Journal of Probability, 26:1–32, 2021

    Chaman Kumar and Neelima. On explicit Milstein-type scheme for McKean-Vlasov stochastic differential equations with super-linear drift coefficient.Electronic Journal of Probability, 26:1–32, 2021. 15

    2021

  16. [24]

    Well-posedness and tamed schemes for McKean-Vlasov equations with common noise.The Annals of Applied Probability, 32(5):3283– 3330, 2022

    Chaman Kumar, Neelima, Christoph Reisinger, and Wolfgang Stockinger. Well-posedness and tamed schemes for McKean-Vlasov equations with common noise.The Annals of Applied Probability, 32(5):3283– 3330, 2022

    2022

  17. [25]

    On explicit approximations for Lévy driven SDEs with super-linear diffusion coefficients.Electronic Journal of Probability, 22:1–19, 2017

    Chaman Kumar and Sotirios Sabanis. On explicit approximations for Lévy driven SDEs with super-linear diffusion coefficients.Electronic Journal of Probability, 22:1–19, 2017

    2017

  18. [26]

    First order strong approximation of Ait–Sahalia-type interest rate model with Poisson jumps.Numerical Algorithms, 94(1):93–130, 2023

    Ziyi Lei, Siqing Gan, and Jing Liu. First order strong approximation of Ait–Sahalia-type interest rate model with Poisson jumps.Numerical Algorithms, 94(1):93–130, 2023

    2023

  19. [27]

    Yun Li, Xuerong Mao, Qingshuo Song, Fuke Wu, and George Yin. Strong convergence of Euler-Maruyama schemes for McKean-Vlasov stochastic differential equations under local Lipschitz conditions of state vari- ables.IMA Journal of Numerical Analysis, 43(2):1001–1035, 2023

    2023

  20. [28]

    Huagui Liu, Banban Shi, and Fuke Wu. Tamed Euler-Maruyama approximation of McKean-Vlasov stochas- tic differential equations with super-linear drift and Hölder diffusion coefficients.Applied Numerical Math- ematics, 183:56–85, 2023

    2023

  21. [29]

    Huagui Liu, Fuke Wu, and Minyu Wu. The tamed Euler-Maruyama approximation of McKean-Vlasov stochastic differential equations and asymptotic error analysis.Discrete and Continuous Dynamical Systems-Series S, 16(5), 2023

    2023

  22. [30]

    Large and moderate deviation principles for McKean-Vlasov SDEs with jumps.Potential Analysis, 59(3):1141–1190, 2023

    Wei Liu, Yulin Song, Jianliang Zhai, and Tusheng Zhang. Large and moderate deviation principles for McKean-Vlasov SDEs with jumps.Potential Analysis, 59(3):1141–1190, 2023

    2023

  23. [31]

    Ergodicity for SDEs and approxima- tions: locally Lipschitz vector fields and degenerate noise.Stochastic Processes and Their Applications, 101(2):185–232, 2002

    Jonathan C Mattingly, Andrew M Stuart, and Desmond J Higham. Ergodicity for SDEs and approxima- tions: locally Lipschitz vector fields and degenerate noise.Stochastic Processes and Their Applications, 101(2):185–232, 2002

    2002

  24. [32]

    Springer-Verlag, Berlin, 2010

    Eckhard Platen and Nicola Bruti-Liberati.Numerical Solution of Stochastic Differential Equations With Jumps in Finance, volume 64 ofStochastic Modelling and Applied Probability. Springer-Verlag, Berlin, 2010

    2010

  25. [33]

    Christoph Reisinger and Wolfgang Stockinger. An adaptive Euler-Maruyama scheme for McKean-Vlasov SDEs with super-linear growth and application to the mean-field FitzHugh-Nagumo model.Journal of Computational and Applied Mathematics, 400:113725, 2022

    2022

  26. [34]

    Chapman and Hall/CRC, 2003

    Peter Tankov.Financial Modelling with Jump Processes. Chapman and Hall/CRC, 2003

    2003

  27. [35]

    Ngoc Khue Tran, Trung-Thuy Kieu, Duc-Trong Luong, and Hoang-Long Ngo. On the infinite time horizon approximation for Lévy-driven McKean-Vlasov SDEs with non-globally Lipschitz continuous and super- linearly growth drift and diffusion coefficients.arXiv:2401.03977, 2024. 16

    2024 arXiv

  28. [36]

    Compensated stochastic theta methods for stochastic differential equations with jumps.Applied Numerical Mathematics, 60(9):877–887, 2010

    Xiaojie Wang and Siqing Gan. Compensated stochastic theta methods for stochastic differential equations with jumps.Applied Numerical Mathematics, 60(9):877–887, 2010

    2010

  29. [37]

    Mean-square convergence and stability of compensated stochastic theta methods for jump-diffusion SDEs with super-linearly growing coefficients

    Yiling Wang, Ziheng Chen, Mengyao Niu, and Yuanling Niu. Mean-square convergence and stability of compensated stochastic theta methods for jump-diffusion SDEs with super-linearly growing coefficients. Calcolo, 60(2):30, 2023

    2023

  30. [38]

    Strong convergence of jump-adapted implicit Milstein method for a class of nonlinear jump-diffusion problems.Journal of Computational Mathematics, 42(1):248–270, 2024

    Xu Yang and Weidong Zhao. Strong convergence of jump-adapted implicit Milstein method for a class of nonlinear jump-diffusion problems.Journal of Computational Mathematics, 42(1):248–270, 2024

    2024

  31. [39]

    Order-preserving strong schemes for SDEs with locally Lipschitz coefficients.Applied Numerical Mathematics, 112:1–16, 2017

    Zhongqiang Zhang and Heping Ma. Order-preserving strong schemes for SDEs with locally Lipschitz coefficients.Applied Numerical Mathematics, 112:1–16, 2017

    2017

  32. [40]

    1Ωi R,k |Y i,N tk |2 ¯p−1 1 + Y i,N tk γ+1 +W 2 ρY,N tk , δ0 ˆγ# ∆t1+ˆδ | {z } J1,1 +CE

    Yuying Zhao, Xiaojie Wang, and Zhongqiang Zhang. Weak error analysis for strong approximation schemes of SDEs with super-linear coefficients II: finite moments and higher-order schemes.arXiv:2406.14065, 2024. Appendix A. Proof of Lemma 3.1 Proof.LetR>0be a sufficiently large c...

    2024 arXiv

  33. [41]

    With the preceding lemmas in hand, we now prove Theorem 3.1

    The proof is concluded by performing a similar calculation forhwithδ3 ≥ 1 2. With the preceding lemmas in hand, we now prove Theorem 3.1. Proof of Theorem 3.1.According to (2.6) and (3.4), one can use theItˆ oformula to acquire X i,N (t)−Y i,N (t) 2 = Z t 0 2 X i,N (s)−Y i,N (...

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.