Learning in a three-valued setting

We discuss the adoption of a three-valued setting for inductive concept learning. Distinguishing between what is true, what is false and what is unknown is necessary in situations where decisions have to be taken on the basis of scarce information. We propose a learning algorithm that adopts extended logic programs under a well-founded semantics as the representation formalism and learns a definition for both the target concept and its opposite, considering positive and negative examples as instances of two disjoint classes. In the target program, default negation is used to ensure consistency and to handle exceptions to general rules. Exceptions to a positive concept are identified from negative examples, whereas exceptions to a negative concept are identified from positive examples. Exceptions can be generalized, in their turn, resulting in a hierarchy of defaults