Developing the next generation of robust Bayesian networks: theory and efficient inference algorithms for mixed credal networks

01 October 2015 → 30 September 2017
Research Foundation - Flanders (FWO)
Research disciplines
  • Natural sciences
    • Applied mathematics in specific fields
    • Statistics and numerical methods
    • Artificial intelligence
    • Computer architecture and networks
    • Distributed computing
    • Information sciences
    • Information systems
    • Programming languages
    • Scientific computing
    • Theoretical computer science
    • Visual computing
    • Other information and computing sciences
  • Social sciences
    • Cognitive science and intelligent systems
algorithms credal networks imprecise probability
Project description

Credal networks are Bayesian networks with imprecise (interval-valued) local probabilities, thereby allowing for robust inferences. This project develops a new type of credal networks, called mixed credal networks. The advantage of this new type is that they do not suffer from the typical computational problems that occur for other types, which allows us to develop efficient inference algorithms.