A functional extension of the Ito formula (Q847101): Difference between revisions

The authors introduce a non-anticipative calculus for certain optional càdlàg processes adapted to the filtration of an Itô process \(X = \int \mu du + \int \sigma dW\). They interpret such a process as a functional of the paths of \(X\) and its quadratic variation \(A\): \[ Y(t) = F_t\left(\left\{X(u): u \leq t\right\}, \left\{A(u): u \leq t\right\}\right) \] where \(F_t\) is for every \(t\) a functional from the respective Skorokhod spaces of paths to \(\mathbb{R}\). If those functionals \((F_t)_{t\geq0}\) are Lipschitz continuous and twice continuously differentiable (in an appropriate sense), then the authors are able to derive a functional Itô formula for \(Y\). In particular in the case where \(X = \int \sigma dW\) is a Brownian martingale, they are able to derive an integral representation of certain martingales \(Y\): \[ Y_T = E(Y_T) + \int_0^T \nabla_X Y(t) dX(t) \] This is a Clark-Haussmann-Ocone formula, where the martingale representation of \(Y\) is non-anticipative. This note only summarizes the results, the proofs will be given in an upcoming article which at the moment can be found under \url{arXiv:1002.2446}.