Change of variable formulas for non-anticipative functionals on path space (Q984411): Difference between revisions

The purpose of the authors is to provide a general version of functional Itô formula. In the spirit of \textit{H. Föllmer} [in: Séminaire de probabilités XV. Lect. Notes Math. 850, 143--150 (1981; Zbl 0461.60074) and in: Stochastic integrals. Lect. Notes Math. 851, 476--478 (1981; Zbl 0462.60046)], càdlàg trajectories \(x\) having finite quadratic variation are considered, without use of probability. The functionals that are considered are families \((F_t)_{0\leq t\leq T}\) which: (i) are adapted, in the sense that \(F_t(x)\) depends only on the restriction \(x_t\) of the trajectory \(x\) to the interval \([0,t]\); (ii) are continuous (roughly, with respect to the uniform norm); (iii) possess so-called ``horizontal'' derivatives \(D_t F_t\) with respect to time \(t\); (iv) possess so-called ``vertical'' derivatives \(\nabla_x F_t\) with respect to the increase of the path \(x_t\) by \(E1_{\{t\}}\). Then, under some more regularity assumption on \((F_t)\), a pathwise functional Itô formula is established, which yields an integral expansion for \(F_t(x_t)- F_0(x_0)\), having nearly the normal form. This formula applies in particular to càdlàg semimartingales and to adapted continuous processes of zero quadratic variation.