Details

Autor(en) / Beteiligte

• Müller, Martin Eric

Titel

Towards Finding Maximal Subrelations with Desired Properties

Ist Teil von

• Relational and Algebraic Methods in Computer Science, p.344-361

Ort / Verlag

Cham: Springer International Publishing

Link zum Volltext

Quelle

Alma/SFX Local Collection

Beschreibungen/Notizen

• As soon as data is noisy, knowledge as it is represented in an information system becomes unreliable. Features in databases induce equivalence relations—but knowledge discovery takes the other way round: given a relation, what could be a suitable functional description? But the relations we work on are noisy again. If we expect to record data for learning a classification of objects then it can well be the real data does not create a reflexive, symmetric and transitive relation although we know it should be. The usual approach taken here is to build the closure in order to ensure desired properties. This, however, leads to overgeneralisation rather quickly. In this paper we present our first steps towards finding maximal subrelations that satisfy the desired properties. This includes a discussion of different properties and their representations, several simple measures on relations that help us comparing them and a few distance measures that we expect to be useful when searching for maximal subrelations.