Peer-to-Peer Based Parallel Genetic Algorithm

Abstract

In an attempt to reduce communication overhead while maintaining high quality Genetic Algorithm results, this paper presents a Peer-to-Peer based Genetic Algorithm that suites both Parallel and Distributed environments. In order to improve our approach's applicability on both distributed and parallel environments, we experimented a set of different individual exchange intervals when running some well-known hard optimization problems. The proposed approach has been applied in different exchange rates and benchmarked with a regular Master-Slave based PGA on result quality and executions time. Experimental results show that our approach managed to find high quality results in shorter execution time compared to Master-Slave based PGA.