A small-time approximation of Girsanov's exponential

TL;DR

This paper develops a small-time approximation for Girsanov's exponential, describing it via a deterministic PDE with a convergence rate of order one, and applies it to estimate short-time transition densities of Langevin dynamics.

Contribution

It introduces a novel PDE-based approximation for Girsanov's exponential over short intervals, linking stochastic and deterministic dynamics.

Findings

Approximation has order one convergence rate.

Can estimate short-time transition densities of Langevin equations.

Aligns with discretization methods like Euler-Maruyama.

Abstract

The main result of this article regards a small time approximation for the Girsanov's exponential. We prove that the latter is well described over short time intervals by the solution of a deterministic partial differential equation.The rate of convergence of the approximation is of order one in the length of the interval. As a possible application we show that our approximation can be used to obtain an estimation of the short-time transition density of a Langevin equation representing the dynamics of a Brownian particle moving under the influence of an external, non-linear force. Using this approach is equivalent to consider a random ordinary differential equation, where the dynamics of the particle is deterministic and given by the aforementioned force and the stochasticity enters through the initial condition. The result is in accordance with those obtained by discretization methods such as Euler-Maruyama, implicit Euler and by approximating the associated Fokker-Planck-Smoluchowski equation.