Authors:
Vitaliy Feoktistov
and
Stefan Janaqi
Affiliation:
Centre de Recherche LGI2P, Site EERIE, EMA, France
Keyword(s):
differential evolution, evolutionary algorithms, heuristics, optimization, selection.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence and Decision Support Systems
;
Enterprise Information Systems
;
Evolutionary Programming
Abstract:
The Differential Evolution algorithm goes back to the class of Evolutionary Algorithms and inherits its philosophy and concept. Possessing only three control parameters (size of population, differentiation and recombination constants) Differential Evolution has promising characteristics of robustness and convergence. In this paper we introduce a new principle of Energetic Selection. It consists in both decreasing the population size and the computation efforts according to an energetic barrier function which depends on the number of generation. The value of this function acts as an energetic filter, through which can pass only individuals with lower fitness. Furthermore, this approach allows us to initialize the population of a sufficient (large) size. This method leads us to an improvement of algorithm convergence.