Feasibility analysis of the use of GPU to improve the efficiency of metaheuristics optimization algorithms

Abstract

Currently, several real-world optimization problems have been mathematically modeled. The modeling process considers as much information as possible to provide valid results, and the obtained model is commonly computationally solved. However, as information increases, complexity also increases. Consequently, a larger computational capacity is needed to solve complex and scalable problems. As a result, meta-heuristic algorithms have been developed to solve complex optimization problems. These algorithms are commonly used for two or more dimensions in which vector and matrix operations are involved. Therefore, it is helpful to carry out parallel processes that reduce the runtime to solve this problem. Currently, multi-core central processing units (CPUs) manage to solve small problems with parallel calculations easily. However, the Graphics Processing Unit (GPU) improves performance because it integrates a more significant number of cores than the CPU. It is very useful for solving problems using several processes in parallel. The matrix operations, the Travelling Salesman Problem (TSP), and the electric transmission expansion planning (TEP) problem have been implemented using the GPU to verify the processor's contribution to the performance of scientific calculations. In the results, the GPU helped solve the TSP. Because more solutions or candidate particles were analyzed in less time. Because of these results, it was assumed that there would be a better performance in solving the TEP problem by using the GPU and analyzing a more significant number of candidate topologies in less time. However, this was not the case; according to the results, the use of the GPU takes longer when analyzing more particles.