Realizations and Factorizations of Positive Definite Kernels

TL;DR

This paper investigates the structure and factorizations of positive definite kernels on sigma-finite measure spaces, with applications to stochastic processes like Gaussian fields and Markov processes, using spectral and probability theory.

Contribution

It provides necessary and sufficient conditions for kernels to have realizations and factorizations in L^2 spaces, extending understanding of their role in stochastic process modeling.

Findings

Characterization of kernels with realizations in L^2(ν)

Conditions for kernel factorizations in Hilbert spaces

Applications to Gaussian fields and Markov processes

Abstract

Given a fixed sigma-finite measure space $\left(X,\mathscr{B},\nu\right)$, we shall study an associated family of positive definite kernels $K$. Their factorizations will be studied with view to their role as covariance kernels of a variety of stochastic processes. In the interesting cases, the given measure $\nu$ is infinite, but sigma-finite. We introduce such positive definite kernels $K\left(\cdot,\cdot\right)$ with the two variables from the subset of the sigma-algebra $\mathscr{B}$, sets having finite $\nu$ measure. Our setting and results are motivated by applications. The latter are covered in the second half of the paper. We first make precise the notions of realizations and factorizations for $K$; and we give necessary and sufficient conditions for $K$ to have realizations and factorizations in $L^{2}\left(\nu\right)$. Tools in the proofs rely on probability theory and on spectral theory for unbounded operators in Hilbert space. Applications discussed here include the study of reversible Markov processes, and realizations of Gaussian fields, and their Ito-integrals.