Bayesian networks

A preliminary model for barley developed by Kristian Kristensen. Exploitation of the domain knowledge inherent in this model is permitted only by written accept from Kristian. 17,140,796*
A preliminary model for insulin dose adjustment. The model consists of 24 structurally identical subnetworks interconnected via temporal links. The model has been developed by Steen Andreassen, Roman Hovorka, Jonathan Benn, Kristian G. Olesen, and Ewart R. Carson. See Andreassen et al. (1991), A Model-based Approach to Insulin Adjustment, In Stefanelliet al. (eds.), Proceedings of the Third Conference on Artificial Intelligence in Medicine, 239-248, Springer-Verlag. 9,825,960
Pedigree for linkage analysis created by Claus Skaanning Jensen and Augustine Kong. See Jensen and Kong (1996), Blocking Gibbs Sampling for Linkage Analysis in Large Pedigrees with Many Loops, Research Report R-96-2048, Department of Computer Science, Aalborg University, Denmark 23,983,962
A preliminary model for deciding on the amount of fungicides to be used against attack of mildew in wheat. The model has been developed by Finn V. Jensen, Jørgen Olesen, and Uffe Kjærulff. 3,400,464*
Subset of the Munin network (189 nodes). For a description of the Munin network see Andreasen et al. (1989), MUNIN - An Expert EMG Assistant, In John E. Desmedt (ed.), Computer-Aided Electromyography and Expert Systems, ch. 21, Elsevier Science Publishers, Amsterdam. 83,735,758
Subset of the Munin network (1003 nodes). 2,049,942
Subset of the Munin network (1045 nodes). 3,077,688
Subset of the Munin network (1041 nodes). 8,860,074
Power plant
A model of the correlations among the sensors in a coal-driven powerplant. There are 46 sparsely connected nodes, all of which are ternary. See also Nielsen and Jensen (2006), On-line alert systems for production plants: A conflict based approach. International Journal of Approximate Reasoning, In press. 502
Preliminary model of the biological processes of a water purification plant. The network consists of four structurally identical subnetworks, each representing a time slice of 15 minutes. The network was developed by Finn V. Jensen, Uffe Kjærulff, Kristian G. Olesen, and Jan Pedersen. 3,028,305*

NB: All networks are specified in the Hugin language. The size of a network is the sum of the clique table sizes. The elimination orders associated with the indicated sizes can be downloaded by clicking the numbers. The numbers marked with an asterisk indicate elimination orders that are known to be optimal (obtained by Hugin's Total Weight algorithm with exhaustive search for minimal separators). The remaining elimination ordes are believed to be optimal or nearly so.

There are also several Bayesian network repositories available on the net. Some examples include:

If you are aware of other sites that should be referenced, please send a mail.
Last modified: Tue Jun 19 11:01:50 2007