Artificial Intelligence Laboratory Working Papers. . .

• Recovery of Full Belief from a Single Factoring of a Bayesian Network by Mark Bloemeke: Artificial Intelligence Laboratory Working Paper 1997-1, August 8, 1997.

  ABSTRACT: This memo describes a propagation algorithm for all beliefs obtained from a single factoring of a Bayesian network.

• Optimal Factoring Problem is NP_Hard by Mark Bloemeke and Marco Valtorta: Artificial Intelligence Laboratory Working Paper 1997-2, August 13, 1997.

  ABSTRACT: This working paper shows the NP-Hardness of the Optimal Factoring Problem.

Recent papers from the University of South Carolina Bayesian Network Group. . .

• Bayesian Model Building from a large Medical Dataset: A Practical Study By Subramani Mani and Marco Valtorta: Working Notes of the Fifth International Workshop on Artificial Intelligence and Statistics (AIS-95), 366-271, Ft. Lauderdale, FL, January 1995.

  ABSTRACT: An experiment in Bayesian model building from a large medical dataset for Mental Retardation is discussed in this paper. We give a step by step approach to the practical aspects of building a Bayesian Network from a dataset. We enumerate the tools required, address the problem of missing values in big datasets resulting from surveys, and elaborate on our solution to the problem. We advance some reasons why imputation is a more desirable approach for model building than some other ad hoc methods suggested in literature. We use the CHDS dataset, Hugin software, CB algorithm, IMP algorithm and CondProb program for generating our model. The network structure and the conditional probabilities are modified under the guidance of the domain expert. We also present validation results and give some suggestions for improvement of the model.

• Construction of Bayesian Network Structures from Data: a Brief Survey and an Efficient Algorithm By Moninder Singh and Marco Valtorta: International Journal of Approximate Reasoning, 12:111-131, New York, NY 1995.

  ABSTRACT: Previous algorithms for the recovery of Bayesian belief network structures from data have been either highly dependent on conditional independence (CI) tests, or have required an ordering on the nodes to be supplied by the user. We present an alogrithm that integrates these two approaches - CI tests are used to generate an ordering on the nodes from the database which is then used to recover the underlying Bayesian network structure using non CI based methods. Results of the evaluation of the algortihm on a number of networks (ex. ALARM, LED and SOYBEAN) are presented. We also discuss some algorithm performance issues and open problems.

• On the Detection of Conflicts in Diagnostic Bayesian Networks Using Abstraction By Young-Gyun Kim and Marco Valtorta: Proceedings of the Eleventh International Conference on Uncertainty in Artificial Intelligence. Montreal, August 1995 (To Appear).

  ABSTRACT: An important issue in the use of expert systems is the so-called brittleness problem. Expert systems model only a limited part of the world. While the explicit management of uncertainty in expert systems mitigates the brittleness problem, it is still possible for a system to be used, unwittingly, in ways that the system is not prepared to address. Such situations may be detected by the method of straw models,first presented by Jensen et al. [1990] and later generalized and justified by Laskey [1991]. We describe an algorithm, which we have implemented, that takes as input an annotated diagnostic Bayesian network (the base model) and constructs, without assistance, a bipartide network to be used as a straw model. We show that in some cases this straw model is better than the independent straw model of Jensen et al., the only other straw model for which a construction algorithm has been designed and implemented.