A parallel genetic algorithm for optimizing morphological filters oninhomogenous workstation clusters
Peter Kraft, Michael Nölle, Gerald Schreiber, Stephen Marshall, Hans Burkhardt · 1995
In this paper a modification of a standard parallel genetic algorithm (SPGA) is introduced which can be run efficiently on different types of parallel computers. The purpose of the algorithm is to find optimal morphological filters for grey scale image processing tasks. The structure of the developed General Parallel Genetic Algorithm (GPGA) is based on a new subpopulation model which uses an integral load balancing and soft synchronization mechanism. It is designed to lead to good parallelization efficiencies even on distributed workstation clusters with multi-user operating systems. This is especially important for user groups which have no access to massively parallel computers to speed up their algorithms. Run time results from tests on a massively parallel computer and a workstation cluster are shown.