Metric duality between positive definite kernels and boundary processes

Abstract

We study representations of positive definite kernels $K$ in a general setting, but with view to applications to harmonic analysis, to metric geometry, and to realizations of certain stochastic processes. Our initial results are stated for the most general given positive definite kernel, but are then subsequently specialized to the above mentioned applications. Given a positive definite kernel $K$ on $S\times S$ where $S$ is a fixed set, we first study families of factorizations of $K$. By a factorization (or representation) we mean a probability space $\left(B,\mu\right)$ and an associated stochastic process indexed by $S$ which has $K$ as its covariance kernel. For each realization we identify a co-isometric transform from $L^{2}\left(\mu\right)$ onto $\mathscr{H}\left(K\right)$, where $\mathscr{H}\left(K\right)$ denotes the reproducing kernel Hilbert space of $K$. In some cases, this entails a certain renormalization of $K$. Our emphasis is on such realizations which are minimal in a sense we make precise. By minimal we mean roughly that $B$ may be realized as a certain $K$-boundary of the given set $S$. We prove existence of minimal realizations in a general setting.