Details
| Original language | English |
|---|---|
| Title of host publication | Architecture of Computing Systems - ARCS 2019 |
| Subtitle of host publication | 32nd International Conference, Copenhagen, Denmark, May 20–23, 2019, Proceedings |
| Editors | Martin Schoeberl, Jürgen Brehm, Christian Hochberger, Sascha Uhrig, Thilo Pionteck |
| Place of Publication | Cham |
| Publisher | Springer Verlag |
| Pages | 100-111 |
| Number of pages | 12 |
| Edition | 1. |
| ISBN (electronic) | 978-3-030-18656-2 |
| ISBN (print) | 978-3-030-18655-5 |
| Publication status | Published - 25 Apr 2019 |
| Event | 32nd International Conference on Architecture of Computing Systems, ARCS 2019 - Copenhagen, Denmark Duration: 20 May 2019 → 23 May 2019 |
Publication series
| Name | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
|---|---|
| Volume | 11479 |
| ISSN (Print) | 0302-9743 |
| ISSN (electronic) | 1611-3349 |
Abstract
This paper presents a framework to support parallel swarm search algorithms for solving black-box optimization problems. Looking at swarm based optimization, it is important to find a well fitted set of parameters to increase the convergence rate for finding the optimum. This fitting is problem dependent and time-consuming. The presented framework automates this fitting. After finding parameters for the best algorithm, a good mapping of algorithmic properties onto a parallel hardware is crucial for the overall efficiency of a parallel implementation. Swarm based algorithms are population based, the best number of individuals per swarm and, in the parallel case, the best number of swarms in terms of efficiency and/or performance has to be found. Data dependencies result in communication patterns that have to be cheaper in terms of execution times than the computing in between communications. Taking all this into account, the presented framework enables the programmer to implement efficient and adaptive parallel swarm search algorithms. The approach is evaluated through benchmarks and real world problems.
Keywords
- Adaptive algorithm, Interplanetary space trajectory, Optimization problems, Parallelization, Particle swarm optimization
ASJC Scopus subject areas
- Mathematics(all)
- Theoretical Computer Science
- Computer Science(all)
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
Architecture of Computing Systems - ARCS 2019: 32nd International Conference, Copenhagen, Denmark, May 20–23, 2019, Proceedings. ed. / Martin Schoeberl; Jürgen Brehm; Christian Hochberger; Sascha Uhrig; Thilo Pionteck. 1. ed. Cham: Springer Verlag, 2019. p. 100-111 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11479).
Research output: Chapter in book/report/conference proceeding › Conference contribution › Research › peer review
}
TY - GEN
T1 - A Parallel Adaptive Swarm Search Framework for Solving Black-Box Optimization Problems
AU - Shuka, Romeo
AU - Brehm, Jürgen
PY - 2019/4/25
Y1 - 2019/4/25
N2 - This paper presents a framework to support parallel swarm search algorithms for solving black-box optimization problems. Looking at swarm based optimization, it is important to find a well fitted set of parameters to increase the convergence rate for finding the optimum. This fitting is problem dependent and time-consuming. The presented framework automates this fitting. After finding parameters for the best algorithm, a good mapping of algorithmic properties onto a parallel hardware is crucial for the overall efficiency of a parallel implementation. Swarm based algorithms are population based, the best number of individuals per swarm and, in the parallel case, the best number of swarms in terms of efficiency and/or performance has to be found. Data dependencies result in communication patterns that have to be cheaper in terms of execution times than the computing in between communications. Taking all this into account, the presented framework enables the programmer to implement efficient and adaptive parallel swarm search algorithms. The approach is evaluated through benchmarks and real world problems.
AB - This paper presents a framework to support parallel swarm search algorithms for solving black-box optimization problems. Looking at swarm based optimization, it is important to find a well fitted set of parameters to increase the convergence rate for finding the optimum. This fitting is problem dependent and time-consuming. The presented framework automates this fitting. After finding parameters for the best algorithm, a good mapping of algorithmic properties onto a parallel hardware is crucial for the overall efficiency of a parallel implementation. Swarm based algorithms are population based, the best number of individuals per swarm and, in the parallel case, the best number of swarms in terms of efficiency and/or performance has to be found. Data dependencies result in communication patterns that have to be cheaper in terms of execution times than the computing in between communications. Taking all this into account, the presented framework enables the programmer to implement efficient and adaptive parallel swarm search algorithms. The approach is evaluated through benchmarks and real world problems.
KW - Adaptive algorithm
KW - Interplanetary space trajectory
KW - Optimization problems
KW - Parallelization
KW - Particle swarm optimization
UR - http://www.scopus.com/inward/record.url?scp=85065895057&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-18656-2_8
DO - 10.1007/978-3-030-18656-2_8
M3 - Conference contribution
AN - SCOPUS:85065895057
SN - 978-3-030-18655-5
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 100
EP - 111
BT - Architecture of Computing Systems - ARCS 2019
A2 - Schoeberl, Martin
A2 - Brehm, Jürgen
A2 - Hochberger, Christian
A2 - Uhrig, Sascha
A2 - Pionteck, Thilo
PB - Springer Verlag
CY - Cham
T2 - 32nd International Conference on Architecture of Computing Systems, ARCS 2019
Y2 - 20 May 2019 through 23 May 2019
ER -