HUGIN: a shell for building Bayesian belief universes for expert systems

摘要: Causal probabilistic networks have proved to be a useful knowledge representation tool for modelling domains where causal relations in broad sense are natural way of relating domain objects and uncertainty is inherited these relations. This paper outlines an implementation the HUGIN shell - handling model expressed by network. The only topological restriction imposed on network that, it must not contain any directed loops. approach illustrated step solving genetic breeding problem. A graph interactively created using instances basic components-- nodes arcs--as building blocks. structure, together with quantitative between their immediate causes as conditional probabilities, automatically transformed into tree junction tree. Here computationally efficient conceptually simple algebra Bayesian belief universes supports incorporation new evidence, propagation information, calculation revised beliefs states Finally, exam ple real world application, MUNIN expert system electromyography discussed.