|
Activity Number:
|
124
- Algorithms for Threat Detection
|
|
Type:
|
Topic-Contributed
|
|
Date/Time:
|
Monday, August 9, 2021 : 1:30 PM to 3:20 PM
|
|
Sponsor:
|
Section on Statistics in Defense and National Security
|
|
Abstract #317134
|
|
|
Title:
|
Matérn Gaussian Fields on Graphs: Theory and Applications
|
|
Author(s):
|
Daniel Sanz-Alonso and Ruiyi Yang*
|
|
Companies:
|
University of Chicago and University of Chicago
|
|
Keywords:
|
Matérn Gaussian fields;
Latent Gaussian models;
graph-Laplacians
|
|
Abstract:
|
In this talk we will consider Matérn Gaussian fields on graphs, derived from a graphical approximation of the stochastic partial differential equation representation of the usual Matérn Gaussian fields on Euclidean domains. We formalize a convergence analysis under a manifold assumption. The graph Matérn Gaussian fields have sparse precision matrices that allow fast sampling and inference. We demonstrate through examples their applications in spatial statistics, Bayesian inverse problems and graph-based machine learning. We will see that the graph representations facilitates exchange of ideas across these disciplines.
|
Authors who are presenting talks have a * after their name.
