Memetic Algorithm with GPU Optimization

Abstract

Implementation of memetic algorithms on GPU was proven as an efficient way to speed up the time required to find a solution for diverse problems. In recent years, thanks to the evolution of GPUs, heavily parallelization can be used and the most time-consuming parts of the algorithm were redefined. This paper aims to prove that. It implements memetic algorithms on GPU using CUDA and identifies the computationally demanding parts. It shows that generally known drawbacks are slowly shifted towards new parts of the algorithm. Obtained results are a step forward in the applicability of memetic algorithms on various and high dimensional problems, where they could not be used due to time and computational resource restrictions.