Guest editors’ introduction: Special issue on Inductive Logic Programming (ILP-2008)

Abstract

The 18th International Conference on Inductive Logic Programming was held in Prague, September 10–12, 2008. Apart from four invited talks and a tutorial, the presented peerreviewed papers consisted of 20 submissions accepted as long papers for the main technical track, and 21 papers accepted for the work-in-progress track. The long papers can be found in Volume 5194 of Springer’s Lecture Notes in Artificial Intelligence series. Based on the reviews of the papers as well as their presentations at the conference, the ILP-2008 PC chairs invited the authors of 7 selected papers to submit a revised and significantly extended version for this special issue. As a result of an additional peer-review adopting the journal criteria, 5 of these papers were finally accepted. While the last MLJ special issue on ILP emphasized approaches combining relational representations with probabilistic modeling and inference, the current issue reflects recent advances in the logical and graphical foundations of relational learning as well as novel applications of ILP in theory revision and natural language processing. A number of first-order logic frameworks have been proposed by the ILP community and used for concept learning. The most popular ones are known as learning from interpretations, learning from entailment, and learning from satisfiability. In the article “Brave Induction: a Logical Framework for Learning from Incomplete Information”, Chiaki Sakama and Katsumi Inoue show that none of these frameworks is directly suited to elegantly deal with a certain type of incompleteness in training data. Inspired by brave inference known in nonmonotonic logics, the authors propose a novel logical setting termed brave induction for learning first-order clauses. They rely on minimal-model semantics instead of using the classical entailment relation to define when an example is covered by a hypothesized clause and background knowledge. The proposed framework is also applied to induction from nonmonotonic logic programs.