Free knot linear interpolation and the Milstein scheme for stochastic differential equations

TL;DR

This paper introduces a practical nonlinear approximation method for scalar stochastic differential equations that achieves the optimal convergence order of 1/√k, improving upon standard methods and providing sharp error bounds.

Contribution

It presents an implementable nonlinear scheme combining Milstein and free-knot linear interpolation, achieving optimal approximation order for SDEs, unlike previous purely theoretical results.

Findings

Achieves the optimal order 1/√k for SDE approximation.

Provides sharp lower and upper error bounds with explicit constants.

Demonstrates the method's practical implementability.

Abstract

The main purpose of this paper is to give a solution to a long-standing unsolved problem concerning the pathwise strong approximation of stochastic differential equations with respect to the global error in the $L_{\infty}$-norm. Typically, one has average supnorm error of order $\left(\ln k/k\right)^{1/2}$ for standard approximations of SDEs with $k$ discretization points, like piecewise interpolated Ito-Taylor schemes. On the other hand there is a lower bound, which indicates that the order $1/\sqrt{k}$ is best possible for spline approximation of SDEs with $k$ free knots. The present paper deals with the question of how to get an implementable method, which achieves the order $1/\sqrt{k}$. Up to now, papers with regard to this issue give only pure existence results and so are inappropriate for practical use. In this paper we introduce a nonlinear method for approximating a scalar SDE. The method combines a Milstein scheme with a pieceweise linear interpolation of the Brownian motion with free knots and is easy to implement. Moreover, we establish sharp lower and upper error bounds with specified constants which exhibit the influence of the coefficients of the equation.