Fish School Search with Algorithmic Skeletons

Wrede Fabian, Menezes Breno, Kuchen Herbert


Abstract
Low-level parallel programming is a tedious and error-prone task, especially when combining several programming models such as OpenMP, MPI, and CUDA. Algorithmic skeletons are a well-known high-level solution to these issues. They provide recurring building blocks such as map, fold, and zip, which are used by the application programmer and executed in parallel. In the present paper, we use the skeleton library Muesli in order to solve hard optimization problems by applying swarm intelligence (SI)-based metaheuristics. We investigate, how much hardware can reasonably be employed in order to find quickly a good solution using Fish School Search (FSS), which is a rather new and innovative SI-based metaheuristic. Moreover, we compare the implementation effort and performance of low-level and high-level implementations of FSS.

Keywords
algorithmic skeletons; metaheuristics; swarm intelligence; fish shool search



Publication type
Article in Journal

Peer reviewed
Yes

Publication status
Published

Year
2018

Journal
International Journal of Parallel Programming

Volume
-

Issue
-

Start page
1

End page
19

DOI

Full text