Parallel Multi-Population Technique for Meta-Heuristic Algorithms on Multi Core Processor

Abstract

This paper investigates parallelization method for meta-heuristic algorithms such as Particle Swarm optimization (PSO) and Differential Evolution (DE) on multicore processor to reach eventually fast execution and stable result. In PSO or DE algorithm, all of member in initial population are created to search the best place in which the value of member in that place is satisfied the output criteria. As the parallelization method, the searching region is separated into many sub-regions which are executed with optimized algorithm on multi-core processor. The structure of meta-heuristic algorithms is rebuilt to execute in parallel multi population mode. The benchmark functions such as Rosenbrock, Griewank, Ackley and Michalewicz are used to test those proposed algorithms. The results show that the proposed parallel multi-population technique applied on PSO and DE algorithm has a competitive performance compared to the standard ones. The parallel multi-population technique shows better result which proves more precise and stable. Especially, meta-heuristic algorithms running in parallel multipopulation mode execute quite convincingly faster than standard ones.