Institute of Electrical & Electronic Engineers
Solving probabilistic problems on graphs via belief propagation has become a very promising paradigm in many fields such as communications, artificial intelligence and digital signal processing. The bayesian graphical approach shows great potential when the authors need to integrate smoothly observations and previous knowledge. The idea of "Injecting" in a graph their current observations, and "Collecting" the response of the system after belief propagation, can be very useful in providing dynamic inference and support to human decision making. The bayesian paradigm presented in the literature under many names, such as Bayesian Networks, Generative Models, Factor Graphs, Markov Random Fields, Associative Memories, etc., reflects a very common intuition about how a "Natural" information system should work.