A simplified algorithm for inference by lower conditional probabilities (original) (raw)

Thanks to the notion of locally strong coherence, the satisfiability of proper logical conditions on subfamilies of the initial domain helps to simplify inferential processes based on lower conditional assessments. Actually, these conditions avoid also round errors that, on the other hand, appear solving numerical systems. In this paper we introduce new conditions to be applied to sets of particular pairs of events. With respect to more general conditions already proposed, they avoid an exhaustive search, so that a sensible time-complexity reduction is possible. The usefulness of these rules in inferential processes is shown by a diagnostic medical problem with thyroid pathology.