Hyperrelations in version space

A version space is a set of all hypotheses consistent with a given set of training examples, delimited by the specific boundary and the general boundary. In existing studies [5, 6, 8] a hypothesis is a conjunction of attribute-value pairs, which is shown to have limited expressive power [9]. In a more expressive hypothesis space, e.g., disjunction of conjunction of attribute-value pairs, a general version space becomes uninteresting unless some restriction (inductive bias) is imposed [9]. In this paper we investigate version space in a hypothesis space where a hypothesis is a hyperrelation, which is in effect a disjunction of conjunctions of disjunctions of attribute-value pairs. Such a hypothesis space is more expressive than the conjunction of attribute-value pairs and the disjunction of conjunction of attribute-value pairs. However, given a dataset, we focus our attention only on those hypotheses which are consistent with given data and are maximal in the sense that the elements in...

• Type: Journal
• Year: 2002
• Where: SAC
• Authors: Hui Wang, Ivo Düntsch, Günther Gediga, Andrzej Skowron
• Keywords: Applied Computing, Attribute-value Pairs, Hypothesis Space, SAC 2002, Version Space