Implementation of a fixing strategy and parallelization in a recent global optimization method

Öztoprak, Figen and Birbil, Ş. İlker (2008) Implementation of a fixing strategy and parallelization in a recent global optimization method. In: Euro Mini Conference on Continuous Optimization and Knowledge-Based Technologies (EurOPT-2008), Neringa, Lithuania

[thumbnail of paper_europt.pdf] PDF
paper_europt.pdf

Download (1MB)

Abstract

Electromagnetism-like Mechanism (EM) heuristic is a population-based stochastic global optimization method inspired by the attraction-repulsion mechanism of the electromagnetism theory. EM was originally proposed for solving continuous global optimization problems with bound constraints and it has been shown that the algorithm performs quite well compared to some other global optimization methods. In this work, we propose two extensions to improve the performance of the original algorithm: First, we introduce a fixing strategy that provides a mechanism for not being trapped in local minima, and thus, improves the effectiveness of the search. Second, we use the proposed fixing strategy to parallelize the algorithm and utilize a cooperative parallel search on the solution space. We then evaluate the performance of our study under three criteria: the quality of the solutions, the number of function evaluations and the number of local minima obtained. Test problems are generated by an algorithm suggested in the literature that builds test problems with varying degrees of difficulty. Finally, we benchmark our results with that of the Knitro solver with the multistart option set.
Item Type: Papers in Conference Proceedings
Subjects: T Technology > T Technology (General) > T055.4-60.8 Industrial engineering. Management engineering > T57.6-57.97 Operations research. Systems analysis
Divisions: Faculty of Engineering and Natural Sciences > Academic programs > Manufacturing Systems Eng.
Depositing User: Ş. İlker Birbil
Date Deposited: 08 Nov 2008 14:54
Last Modified: 26 Apr 2022 08:47
URI: https://research.sabanciuniv.edu/id/eprint/10186

Actions (login required)

View Item
View Item