Metric duality between positive definite kernels and boundary processes (Q1700429): Difference between revisions

Given a positive definite kernel \(K(s,t)\) on a set \(S\), the authors study the factorization of the kernel \(K(s,t)=\int_{B} k_s(x) \overline{k_t(x)}\,d\mu(x)\), where \(k:S\to L^2(\mu)\) and \((B,\mu)\) is a measure space equipped with a probability measure \(\mu\). Motivated by questions in harmonic analysis and metric geometry, they seek to see \(B\) as a certain ``boundary space'' to the set \(S\). The authors prove that the boundary space exists in their setting and that it is possible to have Gaussian solutions to this factorization problem (see the paper under review for the definition at hand). Applications and examples in the theory of analytic functions are given to demonstrate their theorems.