Many-Threaded Differential Evolution on the GPU

Abstract

Differential evolution (DE) is an efficient populational meta-heuristic optimization algorithm that has been applied to many difficult real world problems. Due to the relative simplicity of its operations and real encoded data structures, it is very suitable for a parallel implementation on multicore systems and on the GPUs that nowadays reach peak performance of hundreds and thousands of giga FLOPS (floating-point operations per second). In this chapter, we present a simple yet highly parallel implementation of the differential evolution on the GPU using the CUDA architecture and demonstrate its performance on selected test problems.

A Parallel Version of the JADE Algorithm using GPUS

“…A limitation that is immersed when using sequential versions of DE algorithms when it is necessary to solve real optimization problems is their execution time. In works such as [4], [7], [12], [3] parallel versions of the algorithms have been implemented to reduce this inconvenience. Therefore, the use of Graphical Processing Units (GPUs) used as a means of parallelization are gaining more and more importance due to their great computing potential, scalability, and low cost.…”

A Micro-Differential Evolution Algorithm for Continuous Complex Functions

“…DE was initially developed to solve global optimization problems in continuous spaces [10] and it quickly became a widely used optimization algorithm due to its characteristics of fast convergence and high capacity for exploration of feasible solutions to hard optimization problems [11]- [14].…”