A genetic and evolutionary programming environment with spatially structured populations and built-in parallelism

Abstract

The recent development of the Genetic and Evolutionary Computation field lead to a kaleidoscope of approaches to problem solving, which are based on a common background. These shared principles are used in order to develop a programming environment that enhances modularity, in terms of software design and implementation. The system's core encapsulates the main features of the Genetic and Evolutionary Algorithms, by identifying the entities at stake and implementing them as hierarchies of software modules. This architecture is enriched with the parallelization of the algorithms, based on spatially structured populations, following coarse-grained (Island Model) and fine-grained (Neighborhood Model) strategies. A distributed physical implementation, under the PVM environment, running in a local network, is described.