Back

G-2013-41

Comparative Analysis of Continuous Global Optimization Methods

, , , , and

BibTeX reference

In this paper we evaluate the performance and compare 19 different heuristics for solving continuous global optimization. They are all based on the following metaheuristics: Simulated annealing, Variable neighborhood search, Particle swarm optimization, and Differential evolution. Codes of methods are taken from their authors. The comparison on usual test instances (convex and non-convex) is performed on the same computer. Dimensions of test functions are changed from 10 to 100, thus effectively covering small and large scale problems. The results measured by computational efforts and ranked statistics show that the recent DE-VNS heuristic outperforms the other 18 algorithms on selected problems. Its better performances are noted in solving non-convex problems.

, 28 pages

Document

G-2013-41.pdf (2 MB)