Learning Relational Concepts with Decision Trees

In this paper, we describe two di erent learning tasks for relational structures. When learning a classi er for structures, the relational structures in the training sets are classi ed as a whole. Contrarily, when learning a context dependent classi er for elementary objects, the elementary objects of the relational structures in the training set are classi ed. In general, the class of an elementary object will not only depend on its elementary properties, but also on its context, which has to be learned, too. We investigate the question how such classi cations can be induced automatically from a given training set containing classi ed structures or classi ed elementary objects respectively. We present a graph theoretic algorithm that allows the description of the objects in the training set by automatically constructed attributes. This allows us to employ well-known methods of decision tree inductiontoconstruct a hypothesis. We present the system INDIGO and compare it with the LINUS-...

• Type: Conference
• Year: 1996
• Where: ICML
• Authors: Peter Geibel, Fritz Wysotzki
• Keywords: Classi Cations, Classi Ed Structures, Dependent Classi Er, ICML 1996, Machine Learning