Unadjusted Langevin Algorithms for SDEs with Hoelder Drift

TL;DR

This paper analyzes the convergence of unadjusted Langevin algorithms for SDEs with Hoelder continuous drifts, providing explicit rates under ergodicity conditions, extending previous results that required smoother drifts.

Contribution

It extends convergence rate results of Langevin algorithms to cases with Hoelder continuous drifts, relaxing the smoothness assumptions of prior work.

Findings

Explicit convergence rates in Wasserstein distance are established.

Results apply to Hoelder continuous drifts, not just smooth ones.

The estimates can be sharp in specific scenarios.

Abstract

Consider the following stochastic differential equation for $(X_t)_{t\ge 0}$ on $\mathbb R^d$ and its Euler-Maruyama (EM) approximation $(Y_{t_n})_{n\in \mathbb Z^+}$: \begin{align*} &d X_t=b( X_t) d t+\sigma(X_t) d B_t, \\ & Y_{t_{n+1}}=Y_{t_{n}}+\eta_{n+1} b(Y_{t_{n}})+\sigma(Y_{t_{n}})\left(B_{t_{n+1}}-B_{t_{n}}\right), \end{align*} where $b:\mathbb{R}^d \rightarrow \mathbb{R}^d,\ \ \sigma: \mathbb R^d \rightarrow \mathbb{R}^{d \times d}$ are measurable, $B_t$ is the $d$-dimensional Brownian motion, $t_0:=0,t_{n}:=\sum_{k=1}^{n} \eta_{k}$ for constants $\eta_k>0$ satisfying $\lim_{k \rightarrow \infty} \eta_k=0$ and $\sum_{k=1}^\infty\eta_k =\infty$. Under (partial) dissipation conditions ensuring the ergodicity, we obtain explicit convergence rates of $\mathbb W_p(\mathscr{L}(Y_{t_n}), \mathscr{L}(X_{t_n}))+\mathbb W_p(\mathscr{L}(Y_{t_n}), \mu)\rightarrow 0$ as $n\rightarrow \infty$, where $\mathbb W_p$ is the $L^p$-Wasserstein distance for certain $p\in [0,\infty)$, $\mathscr{L}(\xi)$ is the distribution of random variable $\xi$, and $\mu$ is the unique invariant probability measure of $(X_t)_{t \ge 0}$. Comparing with the existing results where $b$ is at least $C^2$-smooth, our estimates apply to Hoelder continuous drift and can be sharp in several specific situations.