Abstract:In the field of computational finance, one is commonly interested in the expected value of a financial derivative whose payoff depends on the solution of stochastic differential equations (SDEs). For multi-dimensional SDEs with non-commutative diffusion coefficients in the globally Lipschitz setting, a kind of one-half order truncated Milstein-type scheme without Lévy areas was recently introduced by Giles and Szpruch (2014), which combined with the antithetic multilevel Monte Carlo (MLMC) gives the optimal overall computational cost $\mathcal{O}(\epsilon^{-2})$ for the required target accuracy $\epsilon$. Nevertheless, many nonlinear SDEs in applications have non-globally Lipschitz continuous coefficients and the corresponding theoretical guarantees for antithetic MLMC are absent in the literature. In the present work, we aim to fill the gap and analyze antithetic MLMC in a non-globally Lipschitz setting. First, we propose a family of modified Milstein-type schemes without Lévy areas to approximate SDEs with non-globally Lipschitz continuous coefficients. The expected one-half order of strong convergence is recovered in a non-globally Lipschitz setting, where even the diffusion coefficients are allowed to grow superlinearly. This then helps us to analyze the relevant variance of the multilevel estimator and the optimal computational cost is finally achieved for the antithetic MLMC. Since getting rid of the Lévy areas destroys the martingale properties of the scheme, the analysis of both the convergence rate and the desired variance becomes highly non-trivial in the non-globally Lipschitz setting. By introducing an auxiliary approximation process, we develop non-standard arguments to overcome the essential difficulties. Numerical experiments are provided to confirm the theoretical findings.

What problem does this paper attempt to address?

This paper attempts to solve the problem of approximating the expected value of solutions to stochastic differential equations (SDEs) using the antithetic multilevel Monte Carlo (MLMC) method under the condition of non - globally Lipschitz continuous coefficients. Specifically, the goal of the paper is to analyze the convergence and variance of the antithetic MLMC method by proposing a modified Milstein - type scheme (excluding Lévy area) under non - globally Lipschitz conditions, in order to achieve the optimal computational complexity $O(\epsilon^{-2})$. ### Background and Problem In fields such as computational finance, it is usually necessary to calculate the expected values of certain financial derivatives that depend on the solutions of SDEs. For multi - dimensional SDEs with non - commutative diffusion coefficients, under the globally Lipschitz condition, Giles and Szpruch (2014) proposed a truncated Milstein - type scheme. Combined with the antithetic MLMC method, it can achieve the optimal overall computational cost $O(\epsilon^{-2})$. However, many nonlinear SDEs in practical applications have non - globally Lipschitz continuous coefficients, and there is a lack of theoretical guarantees for the antithetic MLMC method in these cases in the literature. ### Main Contributions of the Paper 1. **Proposing a Modified Milstein - type Scheme**: - The authors propose a class of modified Milstein - type schemes that do not include Lévy area for approximating SDEs with non - globally Lipschitz continuous coefficients. - This scheme recovers half of the order of strong convergence under non - globally Lipschitz conditions, even when the diffusion coefficient allows super - linear growth. 2. **Analyzing the Antithetic MLMC Method**: - By introducing an auxiliary approximation process, the authors develop non - standard argumentation methods to overcome the essential difficulties under non - globally Lipschitz conditions. - The relevant variance of the multilevel estimator is analyzed, and finally the optimal computational complexity $O(\epsilon^{-2})$ of the antithetic MLMC method is achieved. ### Specific Technical Details - **Modified Milstein - type Scheme**: \[ Y_{n + 1}=P(Y_n)+\mu_h(P(Y_n))h+\sigma_h(P(Y_n))\Delta W_n+\sum_{j_1, j_2 = 1}^m(L_{j_1}\sigma_{j_2})_h(P(Y_n))\Pi_{t_n, t_{n + 1}}^{j_1, j_2}, \] where $h=\frac{T}{N}$, $P:\mathbb{R}^d\rightarrow\mathbb{R}^d$ is a certain projection operator, and $\mu_h(\cdot)$, $\sigma_h(\cdot)$, $(L_{j_1}\sigma_{j_2})_h(\cdot)$ are certain modifications of the coefficients $\mu$, $\sigma$ and $L_{j_1}\sigma_{j_2}$. - **Strong Convergence Analysis**: - By introducing an auxiliary process $\tilde{Y}_n$, the strong error is decomposed into two parts and estimated separately. - High - order moment estimates are derived using discrete strategies and Taylor expansions. - **Variance Analysis**: - By non - standard argumentation methods, the variance of the multilevel estimator is derived as $O(h^2)$, thereby achieving the optimal computational complexity $O(\epsilon^{-2})$. ### Conclusion The paper successfully solves the problem of how to use the antithetic MLMC method to approximate the expected value of SDEs solutions under non - globally Lipschitz conditions, and provides theoretical guarantees and numerical experimental verification. This result expands the application range of the MLMC method and provides new tools and methods for dealing with a wider range of SDEs.

Antithetic multilevel Monte Carlo method for approximations of SDEs with non-globally Lipschitz continuous coefficients

Antithetic multilevel Monte Carlo method for approximations of SDEs with non-globally Lipschitz continuous coefficients

An Antithetic Multilevel Monte Carlo-Milstein Scheme for Stochastic Partial Differential Equations

Multilevel Monte Carlo methods for positivity-preserving approximations of the Heston 3/2-model

Three Improvements to Multi-Level Monte Carlo Simulation of SDE Systems

Weak antithetic MLMC estimation of SDEs with the Milstein scheme for low-dimensional Wiener processes

Multilevel Monte Carlo with Numerical Smoothing for Robust and Efficient Computation of Probabilities and Densities

An Explicit Positivity-Preserving Scheme for the Heston 3/2-Model with Order-One Strong Convergence

Mean-square convergence rates of implicit Milstein type methods for SDEs with non-Lipschitz coefficients

Antithetic Multilevel Methods for Elliptic and Hypo-Elliptic Diffusions with Applications

Goal-oriented adaptive finite element multilevel Monte Carlo with convergence rates

Unconditionally positivity-preserving approximations of the Ait-Sahalia type model: Explicit Milstein-type schemes

A Multilevel Monte Carlo Ensemble and Hybridizable Discontinuous Galerkin Method for a Stochastic Parabolic Problem

Multilevel Sequential Monte Carlo Samplers

Multilevel Monte Carlo methods for stochastic convection-diffusion eigenvalue problems

Convergence and stability of the semi-tamed Milstein method for commutative stochastic differential equations with non-globally Lipschitz continuous coefficients

Higher order numerical methods for SDEs without globally monotone coefficients

Milstein schemes and antithetic multilevel Monte Carlo sampling for delay McKean-Vlasov equations and interacting particle systems

A multilevel Monte Carlo algorithm for SDEs driven by countably dimensional Wiener process and Poisson random measure

Multi-level Monte Carlo methods with the Truncated Euler-Maruyama Scheme for Stochastic Differential Equations

Higher order time discretization method for a class of semilinear stochastic partial differential equations with multiplicative noise