Non-asymptotic analysis of Langevin-type Monte Carlo algorithms

We study the Langevin-type algorithms for Gibbs distributions such that the potentials are dissipative and their weak gradients have the finite moduli of continuity. Our main result is a non-asymptotic upper bound of the 2-Wasserstein distance between the Gibbs distribution and the law of general Langevin-type algorithms based on the Liptser--Shiryaev theory and functional inequalities. We apply this bound to show that the dissipativity of the potential and the $\alpha$-H\"{o}lder continuity of the gradient with $\alpha>1/3$ are sufficient for the convergence of the Langevin Monte Carlo algorithm with appropriate control of the parameters. We also propose Langevin-type algorithms with spherical smoothing for potentials without convexity or continuous differentiability.