Modeling Interacting Galaxies Using a Parallel Genetic Algorithm

Modeling of interacting galaxies suffers from an extended parameter space prohibiting traditional grid based search strategies. As an alternative approach a combination of a Genetic Algorithm (GA) with fast restricted N-body simulations can be applied. A typical fit takes about 3-6 CPU-hours on a PentiumII processor. Here we present a parallel implementation of our GA which reduces the CPU-requirement of a parameter determination to a few minutes on 100 nodes of a CRAY T3E.