Computational & Technology Resources
an online resource for computational,
engineering & technology publications |
|
Civil-Comp Conferences
ISSN 2753-3239 CCC: 2
PROCEEDINGS OF THE ELEVENTH INTERNATIONAL CONFERENCE ON ENGINEERING COMPUTATIONAL TECHNOLOGY Edited by: B.H.V. Topping and P. Iványi
Paper 8.1
Solving Optimization Problems with MPI-ACO J. Banda-Almeida1,2 and I. Pineda1,2
1School of Mathematical and Computational Sciences, Yachay
Tech University, UrcuquÃ, Ecuador J. Banda-Almeida, I. Pineda, "Solving Optimization Problems with MPI-ACO", in B.H.V. Topping, P. Iványi, (Editors), "Proceedings of the Eleventh International Conference on Engineering Computational Technology", Civil-Comp Press, Edinburgh, UK,
Online volume: CCC 2, Paper 8.1, 2022, doi:10.4203/ccc.2.8.1
Keywords: metaheuristics, parallel computing, high-performance computing, ant
colony optimization, message passing interface, combinatorial optimization
problems.
Abstract
Parallelization of metaheuristics using High-Performance Computing (HPC) provides
a suitable environment to approximate large NP-hard combinatorial optimization
problems (COPs). The Ant Colony Optimization (ACO) is a population metaheuristic
with outstanding time and performance results. This algorithm mimics the indirect
communication and self-organization capabilities of ants. In ACO, each ant is an
autonomous construction procedure, and builds partial solutions to share with the rest
of the colony. The combination of ants results in inherent parallel behaviour that
enables the evaluation of complex problems. This behaviour has motivated the
creation of algorithms that exploit parallel architectures. This paper accelerates ACO
using Message Passing Interface (MPI) and HPC infrastructure. The MPI-ACO
parallelization follows the master-slave model with coarse granularity. The algorithm
divides the number of ants into different processors that simultaneously create local
solutions and iteratively approximate the optimal solution. We evaluate the algorithm
using three COPs, the travelling salesman problem (TSP), the job shop scheduling
problem (JSP), and image segmentation. Each problem is encoded in MPI-ACO using
the appropriate heuristic information and selection policies. The speedup, efficiency,
and Karp-Flatt metric are used to evaluate the acceleration of the MPI-ACO using up
to 32 cores, demonstrating the scalability of the MPI-ACO algorithm.
download the full-text of this paper (PDF, 7 pages, 232 Kb)
go to the previous paper |
|