INFORMATION FUSION, CAUSAL PROBABILISTIC NETWORK AND PROBANET I: Information Fusion Infrastructure and Probabilistic Knowledge Representation

Information fusion is the engine that drives or pulls all related signal/image/data processing in all the information engineering applications. Causal probabilistic networks (CPN) or say Bayesian networks provide a coherent framework for fusing information coming from diverse sources and for maintaining a dynamic equilibrium of a probabilistic world model. Information fusion is realized through updating joint probabilities of the variables upon arrival of new evidence or new hypotheses. This paper is the rst in our review series, which contains three main parts: (1) an overview on the infrastructure of information fusion, (2) probabilistic knowledge representation through constructing CPN's, (3) a comprehensive bibliography derived from the vast but discontiguous literature on CPN's dispersed in arti cial intelligence, probability and statistics and decision analysis. It is worth noting that constructing a CPN for a problem domain is the most individualistic part of the job for each real application of CPNs while probabilistic inferences can be done using proved general inference algorithms developed for special or general network structures.