An algorithm to reconstruct probabilistic distributions of multivariate random functions from the distributions of their projections (Q1850768)

This paper investigates the reconstruction problem of probabilistic distributions of random functions from distributions of their projections. Without loss of generality (concerning the number of variables), there is considered the class \(T\) made up of (only) bivariate random functions whose realizations are integrable with probability one, and whose compact support is the unit circle. For the class \(T\) of random functions, the distribution of any random function is uniquely determined by the distributions of each infinite set of its projections. An algorithm for the projection classification is presented. When a random function from \(T\) has an infinite number of realizations, truncating the distributions of projections and using the algorithm of reconstruction for functions with a finite number of states, one can reconstruct the distributions which are very close to the original function distribution. A stochastic generalization of the class \(T\) is also proposed and examined, together with the corresponding algorithm of projection classification and distribution reconstruction.