Parallel evolutionary algorithms

Evolutionary Strategies, a member of the family of Evolutionary Algorithms, provide a generic means of finding optimal solutions to large-scale numerical problems. Their inherent parallelism makes them an outstanding candidate for parallel execution, on devices ranging from simple SMP machines to the GRID. A new implementation of Evolutionary Strategies presented in this paper offers transparent parallelization and, by exploiting common features, offers the prospect of including functionality of Genetic Algorithms, thereby reducing the overall code complexity and size. While the implementation tried to take into account the needs of modern particle physics experiments, the code has been designed to be useful beyond their realm.