Abstract
The utility and capability of simulated annealing algorithm for generalpurpose engineering optimization is well established since introduced by Kirkpatrick et. al1. Numerous augmentations are proposed to make the algorithm effective in solving specific problems or classes of problems. Some proposed modifications were intended to enhance the performance of the algorithm in certain situations. Some specific research has been devoted to augment the convergence and related behavior of annealing algorithms by modifying its parameters, otherwise known as cooling schedule. Here we introduce an approach to tune the simulated annealing algorithm by combining algorithmic and parametric augmentations. Such tuned algorithm harnesses the benefits inherent in both types of augmentations resulting in a robust optimizer. The concept of ‘reheat’ in SA, is also used as another tune up strategy for the annealing algorithm. The beneficial effects of ‘reheat’ for escaping local optima are demonstrated by the solution of a multimodal optimization problem. Specific augmentations include handling of constraints, fast recovery from infeasible design space, immunization against premature convergence, and a simple but effective cooling schedule. Several representative optimization problems are solved to demonstrate effectiveness of tuning annealing algorithms.
| Original language | English (US) |
|---|---|
| Title of host publication | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
| Publisher | Springer Verlag |
| Pages | 669-679 |
| Number of pages | 11 |
| Volume | 2074 |
| ISBN (Print) | 3540422331, 9783540422334 |
| DOIs | |
| State | Published - 2001 |
| Event | International Conference on Computational Science, ICCS 2001 - San Francisco, United States Duration: May 28 2001 → May 30 2001 |
Publication series
| Name | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
|---|---|
| Volume | 2074 |
| ISSN (Print) | 03029743 |
| ISSN (Electronic) | 16113349 |
Other
| Other | International Conference on Computational Science, ICCS 2001 |
|---|---|
| Country | United States |
| City | San Francisco |
| Period | 5/28/01 → 5/30/01 |
Fingerprint
Keywords
- Constrained optimization
- Cooling schedule
- Design optimization
- Simulated annealing
- Tuned annealing
ASJC Scopus subject areas
- Computer Science(all)
- Theoretical Computer Science
Cite this
Tuned annealing for optimization. / Atiqullah, Mir M.; Rao, Singiresu S.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 2074 Springer Verlag, 2001. p. 669-679 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 2074).Research output: Chapter in Book/Report/Conference proceeding › Conference contribution
}
TY - GEN
T1 - Tuned annealing for optimization
AU - Atiqullah, Mir M.
AU - Rao, Singiresu S
PY - 2001
Y1 - 2001
N2 - The utility and capability of simulated annealing algorithm for generalpurpose engineering optimization is well established since introduced by Kirkpatrick et. al1. Numerous augmentations are proposed to make the algorithm effective in solving specific problems or classes of problems. Some proposed modifications were intended to enhance the performance of the algorithm in certain situations. Some specific research has been devoted to augment the convergence and related behavior of annealing algorithms by modifying its parameters, otherwise known as cooling schedule. Here we introduce an approach to tune the simulated annealing algorithm by combining algorithmic and parametric augmentations. Such tuned algorithm harnesses the benefits inherent in both types of augmentations resulting in a robust optimizer. The concept of ‘reheat’ in SA, is also used as another tune up strategy for the annealing algorithm. The beneficial effects of ‘reheat’ for escaping local optima are demonstrated by the solution of a multimodal optimization problem. Specific augmentations include handling of constraints, fast recovery from infeasible design space, immunization against premature convergence, and a simple but effective cooling schedule. Several representative optimization problems are solved to demonstrate effectiveness of tuning annealing algorithms.
AB - The utility and capability of simulated annealing algorithm for generalpurpose engineering optimization is well established since introduced by Kirkpatrick et. al1. Numerous augmentations are proposed to make the algorithm effective in solving specific problems or classes of problems. Some proposed modifications were intended to enhance the performance of the algorithm in certain situations. Some specific research has been devoted to augment the convergence and related behavior of annealing algorithms by modifying its parameters, otherwise known as cooling schedule. Here we introduce an approach to tune the simulated annealing algorithm by combining algorithmic and parametric augmentations. Such tuned algorithm harnesses the benefits inherent in both types of augmentations resulting in a robust optimizer. The concept of ‘reheat’ in SA, is also used as another tune up strategy for the annealing algorithm. The beneficial effects of ‘reheat’ for escaping local optima are demonstrated by the solution of a multimodal optimization problem. Specific augmentations include handling of constraints, fast recovery from infeasible design space, immunization against premature convergence, and a simple but effective cooling schedule. Several representative optimization problems are solved to demonstrate effectiveness of tuning annealing algorithms.
KW - Constrained optimization
KW - Cooling schedule
KW - Design optimization
KW - Simulated annealing
KW - Tuned annealing
UR - http://www.scopus.com/inward/record.url?scp=48849095463&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=48849095463&partnerID=8YFLogxK
U2 - 10.1007/3-540-45718-6_72
DO - 10.1007/3-540-45718-6_72
M3 - Conference contribution
SN - 3540422331
SN - 9783540422334
VL - 2074
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 669
EP - 679
BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PB - Springer Verlag
ER -