Master-Slave Parallel Genetic Algorithms

Abstract

Master-slave parallel GAs are easy to implement, often yield considerable improvements in performance, and all the theory available for simple GAs can be used to choose adequate values for the search parameters. The analysis of this chapter showed that, for many applications, the reduction in computation time is sufficient to overcome the cost of communications. The calculations presented here should be useful to design fast master-slave GAs that utilize the computing resources in the best possible way.

The chapter discussed the similarities between asynchronous master-slave GAs and GAs with a generation gap. The chapter also presented a G A with a single distributed population. This algorithm is less efficient than the simple master-slave, but it is important because it resembles a GA with multiple communicating populations. The analysis of this algorithm reveals that the optimal number of processors is of the same order as the master-slave.

The main contribution of this chapter was to present a lower bound on the performance gains that are acceptable in any parallel GAs. More sophisticated single-population GAs or multi-population parallel GAs should do better than the simple case examined here or they should be abandoned. The calculations are simple and they are easy to calibrate to consider the hardware and the particular problem.