Algorithms and Applications of Credal Networks

GLOBAL RESEARCH INSTITUTES

  • Algorithms and Applications of Credal Networks

Algorithms and Applications of Credal Networks

Principal Supervisor: Dr. Cassio P. de Campos

 

 

+ Project Description

Standard approaches to uncertainty modelling assume that the lack of knowledge about the actual state of a quantity is described by probabilities over its possible states. Sharp values are typically used to quantify these probabilities. In many cases there are no compelling reasons for that and a set-valued specification of probabilities might offer a better, or at least more cautious, description. Such approach is based on the so-called imprecise-probabilistic framework. An important model within such framework is the robust Bayesian network, or simply credal network, a sound robust model that has been used successfully to represent and to reason with uncertainty. It can be roughly defined as a compact representation of a set of Bayesian networks sharing the same graph. While powerful in terms of representation, algorithms for credal networks suffer from high computational complexity. This project aims at improving existing methods towards particular applications, in order to bring the existing theoretical developments to real-world problems. Interesting real-world applications arise in a number of fields. We have already worked with disease classification and medical prognosis, missing data treatment in clinical and genomics cohorts, image recognition and segmentation, feature extraction, etc. The goal is to enhance and extend the work on some of these applications. For further details on some applications we refer to our article that appeared in The IEEE Intelligent Informatics Bulletin:

http://www.comp.hkbu.edu.hk/~iib/2015/Dec/article5/iib_vol16no1_article5.pdf

+ How to Apply

Applicants should apply electronically through the Queen’s online application portal at: https://dap.qub.ac.uk/portal/

+ Contact Details

Supervisor Name: Dr. Cassio P de Campos
Address:

Queens University of Belfast
School of EEECS
18 Malone Road, CSB 03.026

Email: c.decampos@qub.ac.uk

 

Tel:

Tel: +44 (0)28 9097 6795