Logic and Learning

Nada Lavrac and Luc De Raedt

Course Description

Since the very start of machine learning, logic has been very popular as a representation language for inductive concept- learning and the possibilities for learning in a first order representation have been investigated. This is due to the fact that first order logic extends propositional representations and therefore also the scope of machine learning. In database terminology, propositional techniques learn from a single relation in a relational database, whereas first order approaches cope with multiple relations. Learning in first-order logic is therefore of special interest to researchers and practitioners in machine learning, data mining, knowledge discovery in databases, knowledge representation and logic programming.

The tutorial will provide a complete survey of "Logic and Learning" and will concentrate on learning in first order logic. Early research involved structural matching, least general generalization, model inference, and theory restructuring. Recently, the area of logic and learning has concentrated on Inductive Logic Programming, which studies inductive machine learning within the representation of logic programming.

The course gives an overview of the history, techniques and applications of logic and learning. Logical and algorithmic foundations of this field are emphasized, and illustrated by means of well-known systems and algorithms. This includes: learning as search, logical representations, operators and settings for induction, state-of-the-art methods and systems, and applications in knowledge discovery and logic programming. For use in data mining and knowledge discovery, pointers to the available public domain systems are provided and discussed.

Prerequisite Knowledge

The course assumes basic knowledge of artificial intelligence and logic (e.g. some notions about Prolog and/or the predicate calculus).

About the Lecturers

Nada Lavrac is a research associate at the J. Stefan Institute, Ljubljana (Slovenia). Her main research interest is in inductive logic programming and medical applications of machine learning. She is co-author and editor of several books published by Sigma Press, MIT Press, Kluwer and Springer, including "Inductive Logic Programming: Techniques and Applications" (Ellis Horwood 1994). She was a coordinator of ILPNET, the European Scientific Network in Inductive Logic Programming (1993-96).

Luc De Raedt is a post-doctoral researcher and a (part-time) assistant professor at the Katholieke Universiteit Leuven (Belgium). His main interest is in inductive logic programming. He is a coordinator of the European ESPRIT III and IV projects on Inductive Logic Programming, was chair of the ILP-95 workshop, and he has published two books on inductive logic programming. He has given tutorials and invited talks on ILP at ISMIS-93, ILPS-93, MSL-96, LOPSTR 96, etc.

Last modified: Thu Feb 20 13:50:38 JST 1997