A Milstein-based free knot spline approximation for stochastic differential equations (Q657649)

This paper presents a pathwise approximation of scalar stochastic differential equations by polynomial splines with free knots. The pathwise distance between the solution and its approximation is measured globally on the unit interval in the \(L_{\infty}\)-norm, and the \(L_q\)-expectation of this distance is of concern here. An asymptotically optimal spline approximation of the given scalar Brownian motion is first constructed and then used together with a Milstein approximation scheme to obtain the pathwise approximation \(\hat{X}_k\) of the equation. For general stochastic differential equations an upper bound of order \(1/\sqrt{k}\) is obtained with an explicit constant for the approximation error of \(\hat{X}_k\). In the particular case of equations with additive noise this asymptotic upper bound is sharp.