Sharp large deviation estimates for a certain class of sets on the Wiener space (Q1591578)

Starting with the work of Schilder on Wiener integrals, asymptotics of the form \[ P(X_s \in A) = \exp \left(-{m \over s}\right) \left(\sum_{l=0}^L s^{l/2} q_l + o (s^{M/2})\right), \] where \(A\) is a Borel subset of a Banach space, have been obtained by several authors for various types of random processes \((X_s)_{s\geq 0}\), usually under a strong smoothness assumption on \(\partial A\). This boundary condition is weakened by considering sets of the form \(A_s = \{ \theta \in \Theta: G(s,\theta) \geq 0\}\), where \(G(s,\cdot)\) is a completely \(P\)-regular functional on an abstract Wiener space \((\Theta, H, \mu)\). The Azencott method of Girsanov transform and stochastic Taylor expansions, as well as some results of Kusuoka and Stroock on large deviations for abstract Wiener spaces, are used to prove an analog of the above estimate for \(P(\omega_s \in A_s) = E[1_{\{G(s,\omega_s) \geq 0 \}}]\), where \((\omega_s)_{s\geq 0}\) is the \(\Theta\)-valued Wiener process. Examples of completely \(P\)-regular (path-valued) random functionals, such as diffusion processes and multiple stochastic integral processes, are provided. This paper is closely related to another work by the same authors [``Sharp large deviation estimates for the stochastic heat equation'' (to appear in Potential Anal.)], to which the reader is referred for the computation of the coefficients \(q_1,\ldots ,q_l\).