Title: Approximation to stochastic variance reduced gradient Langevin dynamics by stochastic delay differential equations

Abstract: We study in this paper a weak approximation to stochastic variance reduced gradient Langevin dynamics by stochastic delay differential equations in Wasserstein-1 distance, and obtain a uniform error bound. Our approach is via a refined Lindeberg principle and Malliavin calculus.