On splitting strategies for the numerical solution of stochastic delay differential equations with correlated noises
Cónall Kelly, Wenshi Tang · 2026
Read On splitting strategies for the numerical solution of stochastic delay differential equations with correlated noises by Cónall Kelly, Wenshi Tang (2026) 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-2603-21858.txt.
Introduction
Kelly and Tang study splitting methods for stochastic delay differential equations with two correlated noise sources. Such equations mix a fast linear part with a delayed nonlinear part, and splitting solves each in turn. They bound the mean-square error of Lie-Trotter splitting. With uncorrelated noises, the method converges at order 0.5. With correlation, a stubborn error term appears that does not shrink with step size. Tests confirm the clean case and the rapid loss of accuracy as correlation grows. Reordering the split recovers some range, they find.
Catalog record
| CREATOR | Cónall Kelly, Wenshi Tang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 50713 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.21858
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
On splitting strategies for the numerical solution of stochastic delay differential equations with correlated noises Cónall Kelly, Wenshi Tang Abstract . In this article we investigate the numerical solution of a scalar semilinear stochastic delay differential equation (SDDE) where the linear instantaneous feedback and nonlinear delayed feedback terms are perturbed by a pair of standard Brownian motions with correlation ρ \rho . Such SDDEs may be naturally decomposed into two subsystems: a linear stochastic differential equation (SDE) without delay, and a nonlinear SDDE. Splitting methods work by solving each subsystem separately and composing the results over a single step. Our main theoretical result provides a bound on the mean-square error of a particular strategy for doing this, known as Lie-Trotter splitting. This bound implies that the method is mean-square strongly convergent with order 1 / 2 1/2 when ρ = 0 \rho=0 , so that the noises are uncorrelated, but assurances of convergence are lost when ρ ≠ 0 \rho\neq 0 . Indeed we develop an upper bound on the global mean-square error with a term depends linearly on the magnitude of the correlation, and is independent of the stepsize. While our theoretical error bound is an estimate from above, we conduct numerical experiments that confirm the order of mean-square strong convergence of Lie-Trotter splitting in the ρ = 0 \rho=0 case, and demonstrate a rapid fall-off to effectively zero as | ρ | |\rho| increases. Similar numerical results are observed for an alternative commonly used strategy known as Strang splitting. 2020 Mathematics Subject Classification 34K50, 60H10, 60H35, 65C30 1. Introduction 1.1. General background: stochastic differential equations (SDEs) Differential equations are used to model systems that evolve dynamically according to one or more feedback mechanisms. Real-world systems such as financial markets are subject to uncertainty and it is necessary to incorporate the possibility of random effects disturbing the system. In this work, we model the random perturbations by Brownian motion processes, leading to an SDE of Itô-type. A one-dimensional linear SDE perturbed by a single scalar standard Brownian motion B may be written in the first instance as an integral equation (1) X(t)=X_{0}+\int_{0}^{t}\mu X(s)ds+\int_{0}^{t}\sigma X(s)dB(s),\quad t\in[0,T], for some constants \mu,\sigma\in\mathbb{R} and initial value X(0)=X_{0}\in\mathbb{R} . The second integral on the RHS of ( 1 ) is taken with respect to the Brownian motion, and is known as an Itô integral. Equation ( 1 ) is said to have stochastic differential form when it is written (2) dX(t)=\mu X(t)dt+\sigma X(t)dB(t),\quad t\in[0,T]. The SDE ( 2 ) is widely used to model asset prices in financial markets, and is a rare example of an SDE that admits an explicit solution (see Section 2 ). In the real-world form of this model, \mu represents the constant underlying growth rate of an asset and \sigma its constant volatility. Nonetheless, most SDEs cannot be solved explicitly, and so it is necessary to develop effective and efficient numerical solution techniques. An excellent technical introduction to Itô integrals and SDEs may be found in Mao [ 14 ] , with a more applied treatment available in [ 13 , Chapter 10] . The SDE ( 2 ) has solutions that are Markovian. In many applications it is desirable to model non-Markovian dynamics. In finance, we may be motivated to model time-varying asset-price volatility with historical dependencies as a stochastic process. Alternatively we may be motivated by the fact that, despite classical assumptions about market efficiency, there is evidence that past prices of an asset may sometimes influence its future evolution: see the discussion in [ 1 , 2 ] , and [ 15 ] . Applications of SDEs to the modelling of population dynamics naturally prompt us to incorporate the effects of gestation delay (see [ 3 ] ). An SDE which, like ( 2 ), incorporates only instantaneous feedback will not capture these properties. 1.2. Numerical solution of stochastic delay differential equations (SDDEs) Motivated by this discussion, let us extend ( 2 ) to include additional nonlinear feedback terms with delay in both the drift and diffusion coefficients. Consider the scalar semilinear SDDE of Itô-type \displaystyle dX(t)= \displaystyle\left[\mu X(t)+f(X(t-\tau))\right]dt (3) \displaystyle\qquad\qquad\qquad\qquad+\sigma X(t)dW_{1}(t)+g(X(t-\tau))dW_{2}(t),\quad t\in(0,T], \displaystyle X(t)= \displaystyle\psi(t),\quad t\in[-\tau,0], with delay length \tau>0 , and linear coefficients \mu,\sigma\in\mathbb{R} . An SDDE of the form ( 3 ) describes a system with two feedback channels, one of which is linear and instantaneous, the other nonlinear and subject to feedback delay of length \tau . Both channels are noisy, and the noise perturbations, though not identical on each channel, may be interdependent. Therefore we suppose that W=[W_{1},W_{2}]^{T} is a correlated pair of standard Brownian motions such that W(t)\sim\mathcal{N}\left(0,\Sigma\right),\quad\Sigma=t\cdot\begin{pmatrix}1&\rho\\ \rho&1\end{pmatrix},\quad\rho\in(-1,1). Due to the presence of both noise and feedback delays, SDDE solutions can oscillate about an equilibrium point. We say that a nontrivial continuous function y:[0,\infty)\to\mathbb{R} is oscillatory (about zero) if the set Z_{y}=\{t\geq 0\,:\,y(t)=0\} satisfies \sup Z_{y}=\infty . This definition may be applied pathwise to solutions of ( 3 ), which are said to be almost surely (a.s.) oscillatory if and only if there exists an a.s. event on which all individual trajectories are oscillatory functions. The joint role of noise and feedback delays in oscillation of SDDE solutions is explored in [ 5 , 6 , 7 , 8 ] , building on the analysis of delay-induced oscillation in deterministic differential equations in Gopalsamy [ 11 ] . In this article, we are interested in the efficient numerical solutions of ( 3 ) via splitting. Splitting methods work by decomposing an equation into two or more subequations, each of which may be solved separately over a single discretisation step in a manner most suited to its structure, before being recombined. Readers interested in a general introduction to numerical solution of SDEs may refer to Higham & Kloeden [ 12 ] . Those seeking an introductory reference focused only on numerical solution of SDDEs may consult Buckwar [ 9 ] . By Cholesky factorisation of the covariance matrix \Sigma and by the linear transformation property of the Gaussian distribution, it is equivalent to sample from the SDDE (4) dX(t)=\left[\mu X(t)+f(X(t-\tau))\right]dt\\ +\sigma X(t)\left[\sqrt{1-\rho^{2}}\,dB_{1}(t)+\rho\,dB_{2}(t)\right]+g(X(t-\tau))dB_{2}(t),\quad t\in(0,T], where B=[B_{1},B_{2}]^{T} is a pair of independent standard Brownian motions, and therefore B(t)\sim\mathcal{N}(0,\mathbb{I}_{2}t) . We are motivated by the observation that terms on the RHS of ( 3 ) (and ( 4 )) may be naturally divided into linear terms without a feedback delay, and nonlinear terms with a feedback delay. This induces a natural decomposition of the SDDE into two subsystems: a linear SDE without delay that can be solved exactly, and a nonlinear SDDE, which must be solved by some suitably tailored SDDE approximation method. Our results indicate that care must be taken when following such a strategy, since the correlation of W_{1} and W_{2} can have significant and potentially detrimental effects on the rate at which a splitting method converges in the strong stochastic sense to the true solution of the SDDE. In this article we will examine two specific discretisation strategies of splitting type that are distinguished by the manner in which they recombine these subsystems. They are referred to as Lie-Trotter and Strang splitting respectively. In the case of Lie-Trotter, we find that when W_{1} and W_{2} are independent sources of noise,