Supracompact inference operations
When a proposition α is cumulatively entailed by a finite set A of premisses, there exists, trivially, a finite subset B of A such that B ∪ B′ entails α for all finite subsets B′ that are entailed by A. This property is no longer valid when A is taken to be an arbitrary infinite set, even when the considered inference operation is supposed to be compact. This leads to a refinement of the classical definition of compactness. We call supracompact the inference operations that satisfy the non-finitary analogue of the above property. We show that for any arbitrary cumulative operation C, there exists a supracompact cumulative operation K(C) which is smaller than C and agrees with C on finite sets. Moreover, K(C) inherits most of the properties that C may enjoy, like monotonicity or distributivity. The main part of the paper concerns distributive supracompact operations. We prove that such operations satisfy a simple functional equation, and that there exists a representation theorem which provides a semantic characterization of this family of operations.
Unable to display preview. Download preview PDF.
- M.Freund and D.Lehmann, Deductive inference operations, Lecture Notes in Artificial Intelligence, European Workshop JELIA'90, 227–233, Logics in A.I., Springer-Verlag.Google Scholar
- M.Freund and D.Lehmann, Non-monotonic inference operations, Manuscript.Google Scholar
- M.Freund,D.Lehmann and D.Makinson, Canonical extensions to the infinite case of finitary non-monotonic inference relations, Arbeitespapiere der GMD n∘443: Proceedings of the Workshop on Non-monotonic Reasoning,133–138,1990.Google Scholar
- S.Kraus,D.Lehmann and M.Magidor, Non-monotonic reasoning, preferential models and cumulative logics, Artificial Intelligence 44,167–207.Google Scholar
- D.MakinsonGeneral patterns in non-monotonic reasoning, chapter 2 of Handbook of Logic in Artificial Intelligence and Logic Programming,Volume II, Oxford University Press. (To appear)Google Scholar