A characterization of reciprocal processes via an integration by parts formula on the path space (Q1849494): Difference between revisions

Reciprocal processes are time random fields defined on a compact time interval. They are Markovian fields with respect to the time parameter and therefore a generalization of Markov processes. The interest of these processes was motivated by \textit{E. Schrödinger} [Ann. Inst. Henri Poincaré 2, 269-310 (1932; Zbl 0004.42505)] about the most probable dynamics for a Brownian particle whose laws at two different times are given. The present paper deals with reciprocal processes which are characterized by a simple functional equation, an integration by parts formula, on the space of continuous paths. The main result of the paper is that, for real-valued processes, each class of reciprocal processes with reciprocal characteristics \((1,F)\) coincides with the set of solutions of a functional equation in which the function plays a similar role as the Hamilton function associated to a set of Gibbs measures. This functional equation is an integration by parts formula on the path space \(C([0,1]; R).\) This equation can be interpreted as a perturbed duality equation between Malliavin derivative operator and stochastic integration. To illustrate the approach of reciprocal processes, the authors consider some stochastic differential equations with time boundary conditions (initial and final times). Solutions of such stochastic equations form a wide class of non adapted (then anticipative) non Markovian processes. An application to periodic Ornstein-Uhlenbeck processes is presented.