Combining genetic programming and inductive logic programming using logic grammars

Abstract

Genetic programming (GP) and inductive logic programming (ILP) have received increasing interest recently. Since their formalisms are so different, these two approaches cannot be integrated easily, though they share many common goals and functionalities. A unification would greatly enhance their problem-solving power. In this paper, a framework to combine GP and ILP is presented. The framework is based on a formalism of logic grammars, and a system called LOGENPRO (LOgic grammar-based GENetic PROgramming system) is developed. It is so flexible that programs in different programming languages such as LISP, Prolog and C can be induced. The performance of LOGENPRO in inducing logic programs from noisy examples is also evaluated. A detailed comparison with FOIL and mFOIL has been conducted. The experiment demonstrates that LOGENPRO is a promising alternative to other inductive logic programming systems and is sometimes superior for handling noisy data.