Anticipative calculus with respect to filtered Poisson processes. (Q2490802)

This article develops a stochastic (Malliavin) calculus for so-called filtered Poisson processes, that is, processes having the form \(N_t^k: =\int^t_0 \int_{\mathbb{R}^d}zK(t,s)w(ds,dz)\), where \(K\) is a deterministic kernel, and \(w\) is a marked Poisson measure. Thus there are constructed, beginning by the more or less classical case of Poisson space (when \(K \equiv 1)\), successively: a closed gradient, a divergence (or Skorokhod anticipative integral, for which the sample-paths regularity is studied), and a covariant derivative. A Weitzenböck formula and an Itô formula are established.