Applying logic grammars to induce sub-functions in genetic programming

Abstract

Genetic Programming (GP) is a method of automatically inducing S-expression in LISP to perform specified tasks. The problem of inducing programs can be reformulated as a search for a highly fit program in the space of all possible programs. This paper presents a framework in which the search space can be specified declaratively by a user. Its application in inducing sub-functions is detailed. The framework is based on a formalism of logic grammars and it is implemented as a system called LOGENPRO (the LOgic grammar based GENetic PROgramming system). The formalism is powerful enough to represent context-sensitive information and domain-dependent knowledge. This knowledge can be used to accelerate the learning speed and/or improve the quality of the programs induced. The system is also very flexible and programs in various programming languages can be acquired. Automatic discovery of sub-functions is one of the most important research areas in Genetic Programming. An experiment is used to demonstrate that LOGENPRO can emulate Koza's Automatically Defined Functions (ADF). Moreover, LOGENPRO can employ knowledge such as argument types in a unified framework. The experiment shows that LOGENPRO has superior performance to that of Koza's ADF when more domain-dependent knowledge is available.