Benchmarking Parallel Evolutionary Algorithms

This chapter presents a simulation framework and it is used to examine the effectiveness of various asynchronous optimization methods on simulated distributed computing environments. Four benchmark functions were used to evaluate asynchronous versions of differential evolution, genetic algorithms, and particle swarm optimization. Given large-scale homogeneous and heterogeneous computing environments, asynchronous optimization is shown to have superior scalability and performance compared to synchronous implementations, even scaling to potentially millions of processors.