NIENHUYS-CHENG

Logic and Computation
FOUNDATIONS OF INDUCTIVE LOGIC PROGRAMMING

Introductory course

SHAN-HWEI NIENHUYS-CHENG

Department of Computer Science, Erasmus University of Rotterdam

Second week
Cheng@cs.few.eur.nl

Course description

Inductive Logic Programming (ILP) is the intersection of Logic Programming and Machine Learning. Most ILP researchers are rather oriented towards implementation and application. However, together with Ronald de Wolf Irecently published a book which gives a unified and rigorous treatment of the theoretical foundations of ILP (S-H. Nienhuys-Cheng & R. de Wolf, 'Foundations of Inductive Logic Programming' Springer-Verlag, LNAI-tutorial 1228, May 1997).

This course will contain the essentials of this book, focussing, however, on examples rather than complete proofs of results. Our book has two parts, one on logic and logic programming, the second on ILP. I intend to treat the first part fairly quickly in the first two sessions (in particular if someone else already gives a course on logic or logic programming), devoting the last three sessions to various ILP techniques and their properties.

Session 1: Syntax and semantics of first order logic; interpretations and models; normal forms; Herbrand models.

Session 2: Substitution and unification; resolution-based theorem proving; subsumption theorem and refutation completeness; SLD-resolution.

Session 3: What is Inductive Logic Programming; Shapiro's framework for model inference; how to find a false clause by backtracing; how to find a correct program by unfolding.

Session 4: The lattice structure of the sets of atoms and clauses under the generality order of subsumption (particularly least generalizations); existence and nonexistence of covers in various search spaces and generality orders.

Session 5: Completeness and properness of refinement operators; how to search for a correct theory by means of refinement operators.

Prerequisites
None

Literature
No specific recommendation

HOME PROGRAMME CONTACT REGISTRATION

Logic and Computation	FOUNDATIONS OF INDUCTIVE LOGIC PROGRAMMING
Introductory course	SHAN-HWEI NIENHUYS-CHENG Department of Computer Science, Erasmus University of Rotterdam
Second week	Cheng@cs.few.eur.nl
Course description	Inductive Logic Programming (ILP) is the intersection of Logic Programming and Machine Learning. Most ILP researchers are rather oriented towards implementation and application. However, together with Ronald de Wolf Irecently published a book which gives a unified and rigorous treatment of the theoretical foundations of ILP (S-H. Nienhuys-Cheng & R. de Wolf, 'Foundations of Inductive Logic Programming' Springer-Verlag, LNAI-tutorial 1228, May 1997). This course will contain the essentials of this book, focussing, however, on examples rather than complete proofs of results. Our book has two parts, one on logic and logic programming, the second on ILP. I intend to treat the first part fairly quickly in the first two sessions (in particular if someone else already gives a course on logic or logic programming), devoting the last three sessions to various ILP techniques and their properties. Session 1: Syntax and semantics of first order logic; interpretations and models; normal forms; Herbrand models. Session 2: Substitution and unification; resolution-based theorem proving; subsumption theorem and refutation completeness; SLD-resolution. Session 3: What is Inductive Logic Programming; Shapiro's framework for model inference; how to find a false clause by backtracing; how to find a correct program by unfolding. Session 4: The lattice structure of the sets of atoms and clauses under the generality order of subsumption (particularly least generalizations); existence and nonexistence of covers in various search spaces and generality orders. Session 5: Completeness and properness of refinement operators; how to search for a correct theory by means of refinement operators.
Prerequisites	None
Literature	No specific recommendation