One of the challenges for modern search methods is resolving multidimensional tasks where optimization parameters are hundreds, thousands and more. Many evolutionary, swarm and adaptive methods, which perform well on numerical test with up to 10 dimensions are suffering insuperable stagnation when are applied to the same tests extended to 50, 100 and more dimensions. This article presents an original investigation on Free Search, Differential Evolution and Particle Swarm Optimization applied to multidimensional versions of several heterogeneous real-value numerical tests. The aim is to identify how dimensionality reflects on the search space complexity, in particular to evaluate relation between tasks’ dimensions’ number and corresponding iterations’’ number required by used methods for reaching acceptable solution with nonzero probability. Experimental results are presented and analyzed.
|Publication status||Published - 7 Jul 2013|
|Event||International Conference on Large-Scale Scientific Computing - |
Duration: 3 Jul 2013 → 7 Jul 2013
Conference number: 9th
|Conference||International Conference on Large-Scale Scientific Computing|
|Period||3/07/13 → 7/07/13|