The investigation of lamarckian inheritance with classifier systems in a massively parallel simulation environment

Abstract

In contrast to simulators for ecological processes as they are designed and implemented today, the ParalLife system which is introduced in this article can make use of problem inherent parallehsm to speed up the simulation process. We describe how the simulated environment can be distributed over massively parallel computer architectures and what can be gained by doing so. We then describe Classifier Systems (CS) as one of the decision models available for Animats in ParalLiye. A second topic of this paper is an experiment where Animats with CS improve their performance with Lamarckian Inheritance by recombining parental brain substructures.