Sensitivity Analysis for Mean-Field SDEs with Jumps by Malliavin Calculus: Chaos Expansion Approach

Samaneh Sojudi and Mahdieh Tahmasebi · 2025

Read Sensitivity Analysis for Mean-Field SDEs with Jumps by Malliavin Calculus: Chaos Expansion Approach by Samaneh Sojudi and Mahdieh Tahmasebi (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-00793.txt.

Introduction

Samaneh Sojudi and Mahdieh Tahmasebi write, in twenty twenty-five, a math paper on sensitivity for a class of random equations. The equations describe systems where each part feels the whole group's average state, and where sudden jumps can occur. Such models arise in finance. The authors ask how a result shifts when a starting value shifts. That shift is a key risk measure, known as a Greek in finance.

They build an explicit weight formula using Malliavin calculus, a tool for random systems. They show the equations have unique solutions that can be varied in that sense. They also prove that a stepwise simulation method nears the true shift. Tests show their Monte Carlo method beats a plain finite-difference method.

Catalog record

CREATORSamaneh Sojudi and Mahdieh Tahmasebi
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE99 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-00793

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Sensitivity Analysis for Mean-Field SDEs with Jumps by Malliavin Calculus: Chaos Expansion Approach Samaneh Sojudi and Mahdieh Tahmasebi Abstract In this paper, we describe an explicit extension formula in sensitivity analysis regarding the Malliavin weight for jump-diffusion mean-field stochastic differential equations whose local Lipschitz drift coefficients are influenced by the product of the solution and its law. We state that these extended equations have unique Malliavin differentiable solutions in the Wiener-Poisson space and establish the sensitivity analysis of path-dependent discontinuous payoff functions. It will be realized after finding a relation between the stochastic flow of the solutions and their derivatives. The Malliavin derivatives are defined in a chaos approach that the chain rule is not held. The convergence of the Euler method to approximate Delta Greek is proved. The simulation experiment illustrates our results to compute the Delta, in the context of financial mathematics, and demonstrates that the Malliavin Monte-Carlo computations applied in our formula are more efficient than using the finite difference method directly. Journal: Mathematics and Computers in Simulation 1 Introduction In 1999, Fournié et al. [ 1 ] introduced the Greeks criteria using Malliavin calculus on the Wiener space. They demonstrated that the convergence rate of their method is more efficient than the finite difference method, particularly for discontinuous payoff functions. This innovative technique has since been applied by various researchers across different models. Here are some examples of what has been done. In 2004, Youssef Al-Khatib and Nickolas Perfila in [ 2 ] extended this approach to calculate the Greeks in financial markets modeled by discontinuous processes with Poisson jump times and random jump sizes. Similarly, Davis and Johanson [ 3 ] computed these quantities for jump-diffusion processes satisfying separability conditions. Bavouzet-Morel and Messaoud [ 4 ] analyzed the sensitivity of European and Asian options in jump-diffusion models using Malliavin calculus, showing that the variance of the Brownian-jump Malliavin Monte Carlo method is significantly smaller than the standard Brownian Monte Carlo method. Subsequent works have further advanced the use of Malliavin calculus in jump-diffusion models. For example, Forster et al. [ 5 ] investigated Greeks on spaces generated by independent Wiener and Poisson processes under specific conditions on the jump part in finite and infinite dimensions. Each of these approaches has been only investigated and covered by a subgroup of Lévy processes until Petrou [ 6 ] considered a general class of Lévy processes. Petrou’s approach is based on the fact that every Lévy process can be decomposed into an integral of a Wiener process and an integral with respect to a Poisson random measure. With the chaos expansion in hand, he introduced derivatives and their dual operators for processes in the Wiener-Poisson space. By his generalization, he can bypass the separability condition and the chain rule needed in the sensitivity analysis and apply integration by parts formula. In [ 7 ] , Benth et, al. introduced an innovative approach. Benth states that the Malliavin calculus can be used in cases without continuous martingale components in jump-diffusion dynamics. He explained his idea in such a way that small jumps can be approximated by a continuous martingale with appropriate variance. He also states that the Delta obtained based on this approximation is close to the real Delta. This method was later extended by Khedher [ 8 ] to multidimensional diffusion Lévy processes and applied it to Delta calculation for stochastic models with volatility. In 2023, Hude and Rosendorf [ 9 ] calculated Greeks for exponential models with vanishing Brownian components, demonstrating that the Malliavin Monte Carlo method yields a significantly lower variance than traditional methods. Given the efficiency and effectiveness of Malliavin calculus, it has gained significant attention in financial mathematics (see for instance, [ 10 ] , [ 11 ] , [ 12 ] , [ 13 ] ). Recently, this subject has been explored specifically for mean-field stochastic differential equations (SDEs). The mean-field SDEs, or McKean-Vlasov SDEs or distribution-dependent SDEs, were first studied by Kac [ 14 ] in 1956 and it was followed by McKean [ 15 ] in 1966 on stochastic systems with a large number of interacting particles and then by Graham [ 16 ] in 1992 on McKean-Vlasov Itô-Skorohod equations and nonlinear diffusions with discrete jump sets. Mckean-Vlasov SDEs allow us to describe the collective behavior of a large number of particles in statistical physics and analysis the large-scale social interactions, especially within the framework of mean-field games and financial derivatives pricing, see e.g. ( [ 17 ] , [ 18 ] , [ 19 ] , [ 20 ] ) and references therein. In this manuscript, we undertake a rigorous sensitivity analysis of mean-field models through the sophisticated methodology of Malliavin calculus. Over recent years, notable advancements have been made in this area. For instance, in 2018, David Baños [ 21 ] investigated the Bismut formula for multi-dimensional mean-field SDEs with multiplicative Gaussian noise, emphasizing smooth drift and volatility coefficients. Similarly, the implementation of the Bismut formula for mean-field SDEs under additive noise was detailed in [ 22 ] . Song and Wang [ 23 ] , examined the regularity properties of multidimensional distribution-dependent SDEs driven by a pure jump process, deriving a Bismut-type derivative formula based on stochastic gradient and Malliavin derivative operators (as outlined in Chapter 11 of [ 24 ] ), which adheres to the chain rule. Moreover, an extension of the Bismut-Elworthy-Li formula for semi-linear mean-field SDEs driven by fractional Brownian motion was presented in [ 25 ] , with applications demonstrated in the pricing of variance swaps and other derivatives. To illustrate the application of mean-field SDEs with jumps, especially, we can reference [ 26 ] and [ 27 ] . In [ 26 ] , the authors explore the optimal stopping problem for conditional McKean-Vlasov jump diffusions and illustrate their results to find the optimal time to sell a property like a house modeled in this type of SDEs. In [ 27 ] , the authors concentrate on optimizing the control of mean-field jump diffusions by implementing of deep learning methodologies. This study examines multi-particle jump-diffusion dynamics in the linear quadratic conditional mean-field (LQCMF) framework. They introduce a deep learning algorithm that integrates neural networks with path signatures to estimate the conditional expectations. As a practical scenario, including LQCMF, interbank systemic risk (borrowing/lending) were discussed in the numerical outcomes. The dependence of the coefficients on both the solution itself and its probabilistic law inherently introduces significant complexities into the analysis of these equations. These challenges are further compounded by the assumptions and conditions imposed on the coefficients, such as smoothness, boundedness, or Lipschitz continuity, which are often required to ensure the well-posedness and regularity of the solutions. Such constraints not only complicate the mathematical treatment of the equations but also limit the generality of the models, making the analysis highly intricate and technically demanding. Li and Hui in [ 28 ] explore stochastic differential equations defined by a drift coefficient b(s,X_{t\wedge s},Q_{X_{s}}) with globally Lipschitz condition in 1-Wasserstein distance. For locally Lipschitz conditions, the existing literature remains relatively limited. To the best of our knowledge, the first local condition was introduced in the pioneering and elegant work of Carmona and Delarue (2015). In [

Read the complete text