Bayesian Methods for Prediction of Atomic Migration Barriers for Quantum-Mechanics-Based Material Design

Show full item record



Permalink

http://urn.fi/URN:NBN:fi-fe201804208659
Title: Bayesian Methods for Prediction of Atomic Migration Barriers for Quantum-Mechanics-Based Material Design
Author: Santana Vega, Carlos
Contributor: University of Helsinki, Faculty of Science, Department of Computer Science
Publisher: Helsingin yliopisto
Date: 2018
Language: eng
URI: http://urn.fi/URN:NBN:fi-fe201804208659
http://hdl.handle.net/10138/273610
Thesis level: master's thesis
Abstract: The scope of this project is to provide a set of Bayesian methods to be applied to the task of potential energy barriers prediction. Energy barriers define a physical property of atoms that can be used to characterise their molecular dynamics, with applications in quantum-mechanics simulations for the design of new materials. The goal is to replace the currently used artificial neural network (ANN) with a method that apart of providing accurate predictions, can also assess the predictive certainty of the model. We propose several Bayesian methods and evaluate them on this task, demonstrating that sparse Gaussian process (SGP) are capable of providing predictions, and their confidence intervals, with a level of accuracy equivalent to the current ANN, in a bounded computational complexity time.


Files in this item

Total number of downloads: Loading...

Files Size Format View
Thesis - Final document.pdf 3.380Mb PDF View/Open

This item appears in the following Collection(s)

Show full item record