Attributed grammatical evolution using shared memory spaces and dynamically typed semantic function specification

In this paper we introduce a new Grammatical Evolution (GE) system designed to support the speci cation of problem semantics in the form of attribute grammars (AG). We discuss the motivations behind our system design, from its use of shared memory spaces for attribute storage to the use of a dynamically type programming language, Python, to specify grammar semantics. After a brief analysis of some of the existing GE AG system we outline two sets of experiments carried out on four symbolic regression type (SR) problems. The rst set using a context free grammar (CFG) and second using an AG. After presenting the results of our experiments we highlight some of the potential areas for future performance improvements, using the new functionality that access to Python interpreter and storage of attributes in shared memory space provides.