$L^2$-Wasserstein contraction of modified Euler schemes for SDEs with high diffusivity and applications

TL;DR

This paper establishes $L^2$-Wasserstein contraction properties for a modified Euler scheme applied to SDEs with high diffusivity, leading to improved convergence rates and new applications in inequalities and invariant measure approximation.

Contribution

It introduces a novel approach to prove $L^2$-Wasserstein contraction for modified Euler schemes under high diffusivity, with explicit conditions and enhanced convergence results.

Findings

Proves $L^2$-Wasserstein contraction for modified Euler schemes with large diffusivity.

Provides improved non-asymptotic $L^2$-Wasserstein error bounds for invariant measures.

Enhances convergence rates in the strong law of large numbers for additive functionals.

Abstract

In this paper, we are concerned with a modified Euler scheme for the SDE under consideration, where the drift is of super-linear growth and dissipative merely outside a closed ball. By adopting the synchronous coupling, along with the construction of an equivalent quasi-metric, the $L^2$-Wasserstein contraction of the modified Euler scheme is addressed provided that the diffusivity is large enough. In particular, as a by-product, the $L^2$ Wasserstein contraction of the projected (truncated) Euler scheme and the tamed Euler algorithm is treated under much more explicit conditions imposed on drifts. The theory derived on the $L^2$-Wasserstein contraction has numerous applications on various aspects. In addition to applications on Poincar\'{e} inequalities (with respect to the numerical transition kernel and the numerical invariant probability measure), concentration inequalities for empirical averages, and bounds concerning the KL-divergence, in this paper we present another two potential applications. One concerns the non-asymptotic $L^2$-Wasserstein bound corresponding to the projected Euler scheme and the tamed Euler recursion, respectively, which further implies the $L^2$-Wasserstein error bound between the exact invariant probability measure and the numerical counterpart. It is worthy to emphasize that the associated convergence rate is improved greatly in contrast to the existing literature. Another application is devoted to the strong law of large numbers of additive functionals related to the modified Euler algorithm, where the observable functions involved are allowed to be of polynomial growth, and the associated convergence rate is also enhanced remarkably.