Martingale representation theorem for the \(G\)-expectation (Q550131): Difference between revisions

The authors prove a martingale representation theorem for the \(G\)-expectation as introduced by \textit{S. Peng} [in: Stochastic analysis and applications. The Abel symposium 2005. Proceedings of the second Abel symposium, Oslo, Norway, 2005, held in honor of Kiyosi Itô. Berlin: Springer. Abel Symposia 2, 541--567 (2007; Zbl 1131.60057)]. More precisely, they prove that for every square-integrable \(\xi\) (in an appropriate sense), there exist a unique stochastic integrand \(H\) and a unique increasing \(K\) for which \(-K\) is a \(G\)-martingale, such that for the conditional \(G\)-expectation \(E^G_t[\xi]\), the following holds: \[ E^G_t[\xi]= \xi- \int_t^1 H_s \,dB_s+ K_1-K_t= E^G[\xi]+ \int_0^t H_s \,dB_s-K_t; \] here, \(B\) is the canonical process on Wiener space. They prove that, in general, the increasing process \(K\) will always be present, except in the case of symmetric \(G\)-martingales \(M\), i.e., \(G\)-martingales for which also \(-M\) is a \(G\)-martingale. By the nonlinearity of \(G\)-expectation, this class does not include all \(G\)-martingales.