On compatibility and support functions in probabilistic relaxation

In this paper a close relationship between probabilistic relaxation and a compound statistical decision rule is established. This relationship exposes a number of problems associated with the heuristic compatibility and support functions. It is discussed how the choice of these functions should depend on the type of contextual information to be exploited (spatial relationship between objects, correlation of observations). A number of examples of compatibility and support functions corresponding to specific contextual decision rules are given.