A Kernel Approach to the Stinespring and Kraus Representations
James Tian
TL;DR
The paper develops a self-contained kernel-based derivation of the Stinespring and Kraus representations for completely positive maps. It introduces two scalar positive-definite kernels: the first generates the dilation space and a unital *-representation, and the second exposes a canonical tensor-product structure from which Kraus operators arise. This approach avoids heavy C*-algebra machinery, clarifies the geometric meaning of dilation, and shows how the Kraus decomposition emerges from orthogonal components of the ancillary space. The resulting framework provides a transparent link between positive definite kernels, reproducing kernel Hilbert spaces, and the canonical forms of CP maps, with convergence statements in the strong operator topology.
Abstract
We give a self-contained derivation of the Stinespring and Kraus structure theorems for completely positive maps using only scalar positive-definite kernels.
