A Framework for Learning Rules from Multiple Instance Data

Abstract. This paper proposes a generic extension to propositional rule learners to handle multiple-instance data. In a multiple-instance representation, each learning example is represented by a \bag" of fixed-length \feature vectors". Such a representation, lying somewhere between propositional and first-order representation, offers a tradeoff between the two. Naive-RipperMi is one implementation of this extension on the rule learning algorithm Ripper. Several pitfalls encountered by this naive extension during induction are explained. A new multiple-instance search bias based on decision tree techniques is then used to avoid these pitfalls. Experimental results show the benefits of this approach for solving propositionalized relational problems in terms of speed and accuracy.

• Type: Conference
• Year: 2001
• Where: ECML
• Authors: Yann Chevaleyre, Jean-Daniel Zucker
• Keywords: ECML 2001, Machine Learning, Propositional Rule Learners, Rule Learning Algorithm, fixed-length \feature Vectors