Generalized Version Space Trees

We introduce generalized version space trees, a novel data structure that serves as a condensed representation in inductive databases for graph mining. Generalized version space trees allow for a comfortable representation of version spaces and a natural way to efficiently process inductive queries and operations on version spaces. In particular, we focus on using generalized version space trees in the field of mining free (i.e. unrooted) trees in graph databases. Experiments with two data sets from the National Cancer Institute’s Developmental Therapeutics Program (DTP) indicate that generalized version space trees can boost inductive queries considerably while featuring moderately increased space requirements in comparison to a version space representation based on boundary sets.

• Type: Conference
• Year: 2003
• Where: KDID
• Authors: Ulrich Rückert, Stefan Kramer
• Keywords: Database, Inductive Queries, KDID 2003, Version Space, Version Space Trees