Graduate Seminars

back Back to all Graduate Seminars

AMCS/STAT Graduate Seminars| Gaussians on graphs: The coolest kid on the block!

Start Date: February 15, 2018
End Date: February 15, 2018

By Prof. Haavard Rue and Dr. Haakon Bakka (KAUST)

The Gaussian distribution is central in statistics, and so is the extension to the multivariate Gaussian distribution. This multivariate distribution is most often, and traditionally, described using the covariance matrix. This matrix describes the pairwise covariance/correlation between each pair of variables.
This representation is not very useful when doing computations, nor is it able to represent important conditional independence properties. The way out is to build  the multivariate Gaussian distribution using an undirected graph, and to use the inverse covariance matrix; hence the phrase 'Gaussian on graphs'.
In this talk we will explore the huge benefits of working with Gaussian on graphs, and demonstrate one application where this is used: How to create spatial dependency models which obeys complicated geometry like coastlines and islands.
 
Biography: Professor Rue's research interests lie in computational Bayesian statistics (see http://bayescomp.kaust.edu.sa) and Bayesian methodology such as priors, sensitivity and robustness.
His main body of research is built around the R-INLA project (www.r-inl a.org), which aims to provide a practical tool for approximate Bayesian analysis of latent Gaussian models, often at extreme data scales. This project also includes efforts to use stochastic partial differential equations to represent Gaussian fields, for the use in spatial statistics. 
 
Dr. Haakon Bakka is a Postdoc in Rue's research group with a PhD from the Norwegian University of Science and Technology (2017).
 

More Information:

For more info contact: Prof. Haavard Rue: email: Haavard.Rue@kaust.edu.sa

Date: Thursday 15th Feb 2018
Time:12:00 PM - 01:00 PM
Location: Building 9, Lecture Hall 1 Room 2322
Refreshments: Brown bag Lunch will be provided at 11:45 AM