1Complex Aggregates in Relational Learning

In relational learning, one learns patterns from rela-tional databases, which usually contain multiple tables that are interconnected via relations. Thus, an exam-ple for which a prediction is to be given may be related to a set of objects that are possibly relevant for that prediction. Relational classi¯ers di®er with respect to how they handle these sets: some use properties of the set as a whole (using aggregation), some refer to prop-erties of speci¯c individuals, however, most classi¯ers do not combine both. This imposes an undesirable bias on these learners. This dissertation describes a learn-ing approach that avoids this bias, using complex ag-gregates, i.e., aggregates that impose selection condi-tions on the set to aggregate on