Uncertainty Quantification for Numerical Models with two Regions of Solution

Lead Research Organisation: UNIVERSITY OF EXETER
Department Name: Engineering Computer Science and Maths

Abstract

Abstracts are not currently available in GtR for all funded research. This is normally because the abstract was not required at the time of proposal submission, but may be because it included sensitive information such as personal details.

Publications

10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/N509656/1 30/09/2016 29/09/2021
1783352 Studentship EP/N509656/1 30/09/2016 30/03/2020 Louise Kimpton
 
Description My main area of research has been in uncertainty quantification. Complex numerical models are used in science to represent real life physical systems, and I am particularly interested in models with two distinct regions in output space where classification is required. For example, a computer model may fail to complete for specific input regions, and we'd like to predict where to avoid running the model, or incorrectly running an emulator. My first result was to produce a latent Gaussian process model for correlated classification.

A common classification method is logistic regression, which produces a distribution for the predictive class membership of being in either region. When sampling from this to make predictions, current practice is to draw from an independent Bernoulli distribution; drawing marginally loses any correlation between data and can result in large numbers of misclassifications. If simulating chains or fields of 0's and 1's, it is hard to control the 'stickiness' of like symbols. My current research is aimed at generating a correlated Bernoulli process using de Bruijn graphs to create chains of 0's and 1's, for which like symbols cluster together. De Bruijn Graphs are a generalisation of Markov chains, where the 'word' length controls the number of states that each individual state is dependent on, hence increasing correlation over a wider area.
Exploitation Route I am keen to publish my current work and hopefully one day to apply for funding for a postdoc or fellowship to carry on the work.
Sectors Digital/Communication/Information Technologies (including Software)

Environment

Other