Laplace approximation for rough differential equation driven by fractional Brownian motion (Q1942114): Difference between revisions

The author introduces rough differential equations (RDEs) with small parameter, namely, for \(\varepsilon>0\), \[ dY^\varepsilon_t=\sigma(Y^\varepsilon_t)\varepsilon dW_t^H+\beta(\varepsilon,Y^\varepsilon_t)dt, \] and solution starts from 0. Here, \(W^H\) is the fractional Brownian rough path, \(H\in (1/4,1/2]\), and the coefficients are sufficiently smooth. The main purpose of the paper is to prove the Laplace approximation for the first level path of \(Y^\varepsilon\) as \(\varepsilon\downarrow 0\). As preliminaries, rough path theory and fractional Brownian rough paths are reviewed. Then, the Hilbert-Schmidt property of the Hessian of the Itō map restricted on the Cameron-Martin space \(\mathcal{H}^H\) of fractional Brownian motions is proved. It is mentioned that, thanks to Friz and Victoir's result, such Cameron-Martin paths are Young integrable, and, therefore, their Hessian is computable. A probabilistic representation of the stochastic extension of the Hessian is given. The Laplace approximation for an RDE, which involves a fractional-order term of \(\varepsilon>0\), is considered. This has an application to the short-time asymptotics of integral quantities of the solution of a fixed RDE driven by a fractional Brownian motion.