A Kriging procedure for processes indexed by graphs

TL;DR

This paper introduces a novel kriging method for Gaussian processes indexed by graphs, extending traditional spatial prediction techniques to graph-structured data with explicit estimators and error control.

Contribution

It develops a new kriging framework for processes on graphs, including explicit estimators and error bounds, expanding spatial prediction methods to graph-based data.

Findings

Derived explicit kriging estimators for graph-indexed processes

Provided prediction error bounds for the proposed method

Extended spatial prediction techniques to graph-structured data

Abstract

We provide a new kriging procedure of processes on graphs. Based on the construction of Gaussian random processes indexed by graphs, we extend to this framework the usual linear prediction method for spatial random fields, known as kriging. We provide the expression of the estimator of such a random field at unobserved locations as well as a control for the prediction error.