Performance Analysis of a Parallel Genetic Algorithm: A Case Study of the Traveling Salesman Problem

Abstract

Genetic Algorithm (GA) is one of the most popular optimization techniques. Inspired by the theory of evolution and natural selection, it is also famous for its simplicity and versatility. Hence, it has been applied in diverse fields and domains. However, since it involves iterative and evolutionary processes, it takes a long time to obtain optimal solutions. To improve its performance, in this research work, we had parallelized GA processes to enable searching through the solution space with concurrent efforts. We had experimented with both CPU and GPU architectures. Speedups of GA solutions on CPU architecture range from 7.2 to 22.2, depending on the number of processing cores in the CPU. By contrast, speed-ups of GA solutions on GPU architecture can reach up to 172.4.