Parallelism: an effective genetic programming implementation on low-powered mathematica workstations

Abstract

Mathematica has proven itself to be a suitable platform on which to develop prototype Genetic Programming applications. However, due to the sheer complexity of genetic programming calculations, non-trivial problems cannot be solved on a single Mathematica workstation. A distributed algorithm is suggested to eliminate such restrictions on the problem domain. A client-server network is utilised to model a system of multiple populations under simultaneous evolution. Regular migration of members, through the medium of the server, results in suitable genetic material being filtered through to other populations. This multi-population model is contrasted with the single population standard approach in terms of its performance and utility value.