Stochastic mappings and random distribution fields: a correlation approach (Q1757400)

This paper is a nice result of the common effort of the authors to extend the prediction problems in a general abstract framework. Generally speaking, a stochastic process is a function \(\Phi:\Lambda\rightarrow\mathcal H\) which satisfies some appropriate conditions, \(\Lambda\) taking various forms from \(\mathbb Z,\mathbb R,\mathbb C\) till an abstract set, and \(\mathcal H\) being a scalar or operator set. Various situations were studied in the past, and in this paper a general form is found which includes the classical ones as particular cases. Here, generalizing the multivariate second order random fields, which are processes defined on a locally compact abelian group \(\Lambda\), the generalized multivariate second order stochastic mappings are introduced in the second section of the paper, where \(\Lambda\) is an arbitrary set. Also other preliminaries and basic notations are given, as the notion of gramian and the generalized multivariate second order random distribution fields, when \(\Lambda\) is the space of test functions in theory of distributions. In the third section, the set of continuous multivariate second order stochastic mappings indexed over an abstract topological space \(\Lambda\) is considered. To such stochastic mappings the vector domain, the modular domain, the measurement space, the operator covariance and scalar covariance are associated using a general reproducing kernel technique. Also a Kolmogorov type factorization is given. In the fourth section, the previous obtained results are applied to the case of multivariate second order random distribution fields, all the associated elements being obtained and the properties being proved using a distributional framework. In the fifth section, it is explained how from the correlation theory of multivariate second order random distribution fields the classical theory for multivariate second order random fields can be obtained. Also, the results of the previous sections are transposed to the subclass of multivariate second order stochastic regular measures as stochastic mappings indexed by some \(\delta\)-rings of Borel sets in \({\mathbb R}^d\). In the considered general frame, the concepts of determinism and nondeterminism are introduced and a general Wold type decomposition for multivariate second order random distribution fields into deterministic and nondeterministic parts is given in the sixth section. The paper is finished with some concluding remarks about the research in such a general framework.