Approximation Technique for Belief Revision in Timed Influence Nets

摘要

The paper presents an approach for belief updating in Timed Influence Nets. Influence Nets provide graphical representation of causal or influencing relationships in complex situations. They are used to model and evaluate courses of actions in certain domains and to compare the performance of actions based on the desired outcome. In Timed Influence Nets, the impact or effect of these actions on target variables is not instantaneous. This is modeled by adding communication and processing delays in the model. The paper provides a technique for updating the beliefs of variables in the model over time once new evidence is received about some of the variables in the model. The objective is to assess the behavior of the variables of interest as a function of both the timing of actions and the receipt of evidence on indicators, thus providing aid to decision makers in the revision of the planned courses of actions.