Search

Particle swarm optimization (PSO) is an evolutionary heuristics-based method used for continuous function optimization. PSO is stochastic yet very robust. Nevertheless, real-world optimizations require a high computational effort to converge to a good solution for the problem. In general, parallel PSO implementations provide good performance. However, this depends heavily on the parallelization strategy used as well as the number and characteristics of the exploited processors. In this paper, we propose a cooperative strategy, which consists of subdividing an optimization problem into many simpler sub-problems. Each of these focuses on a distinct subset of the problem dimensions. The optimization work for all the selected sub-problems is done in parallel. We map the work onto four different parallel high-performance multiprocessors, which are based on multi- and many-core architectures. The performance of the strategy thus implemented is evaluated for four well known benchmark functions with high-dimension and different complexity. The obtained speedups are compared to that yielded by a serial PSO implementation. [ABSTRACT FROM AUTHOR]

Particle swarm optimization (PSO) is an evolutionary heuristics-based method used for continuous function optimization. Compared with existing stochastic methods, PSO is very robust. Nevertheless, for real-world optimizations, it requires a high computational effort. In general, parallel implementations of PSO provide better performance. However, this depends heavily on the parallelization strategy engineered as well as the number and characteristics of the exploited processors. In this paper, we analyze three different parallelization strategies: a Particle-oriented Strategy (PoS); a Dimension-oriented Strategy (DoS) and a Cooperation-oriented Strategy (CoS). PoS parallelizes the particle's work. DoS focuses on the work done with respect to each of the problem dimensions and does it in parallel. CoS subdivides the optimization problem into many simpler subproblems, each of which focuses on a distinct subset of the original problem dimensions. The optimization work for all the yielded subproblems is done in parallel. Note that in the second and third strategies, all particles act in parallel too. We map the three strategies onto a Graphics Processing Units (GPU)-based architecture. The performance of the implementations is evaluated using four benchmark functions, considering high-dimension instances. We compare the speedups achieved by the GPU-based implementations of the considered parallelization strategies to the reference sequential PSO implementation as well as to existing PSO implementation on Graphics Processing Units. [ABSTRACT FROM AUTHOR]