Explicit contraction rates for a class of degenerate and infinite-dimensional diffusions

Abstract

Given a separable and real Hilbert space $\mathbb{H}$ and a trace-class, symmetric and non-negative operator $\mathcal{G}:\mathbb{H}\to \mathbb{H}$, we examine the equation \begin{align*} dX_t = -X_t\, dt + b(X_t) \, dt + \sqrt{2} \, dW_t, \qquad X_0=x\in\mathbb{H}, \end{align*} where $(W_t)$ is a $\mathcal{G}$-Wiener process on $\mathbb{H}$ and $b:\mathbb{H}\to \mathbb{H}$ is Lipschitz. We assume there is a splitting of $\mathbb{H}$ into a finite-dimensional space ${\mathbb{H}}^l$ and its orthogonal complement ${\mathbb{H}}^h$ such that $\mathcal{G}$ is strictly positive definite on ${\mathbb{H}}^l$ and the non-linearity $b$ admits a contraction property on ${\mathbb{H}}^h$. Assuming a geometric drift condition, we derive a Kantorovich ($L^1$ Wasserstein) contraction with an explicit rate for the corresponding Markov kernels. The estimates for the rate are based on the eigenvalues of $\mathcal{G}$ on the space ${\mathbb{H}}^l$, a Lipschitz bound on $b$ and a geometric drift condition. The results are derived using coupling methods.

To be published in 'Stoch PDE: Anal Comp'. The final publication is available at http://link.springer.com

References in corpus (3)

• Spectral gaps in Wasserstein distances and the 2D stochastic Navier--Stokes equations
• Ergodicity of Approximate MCMC Chains with Applications to Large Data Sets
• Theoretical guarantees for approximate sampling from smooth and log-concave densities

Cited by in corpus (3)

• Coupling and Convergence for Hamiltonian Monte Carlo
• Transportation inequalities for non-globally dissipative SDEs with jumps via Malliavin calculus and coupling
• Continuum limit and preconditioned Langevin sampling of the path integral molecular dynamics