Algorithmes évolutionnaires pour la programmation non-linéaire globale

Abstract

Evolutionary algorithms are stochastic methods for global optimization based on Darwin's theory of natural biological evolution. There are some variants of these methods but the main idea is similar : the principle is to mimic a population's evolution in a search space with the help of three operators : selection, recombination, mutation. The principal rule is the next one : the fittest individuals survive and produce children that are better suited to their environment. To this idea can be added the notion of mass extinction that considerably contributed to the biological evolution phenomenon. Therefore, this could also improve the ecolution proposed by the evolutionary algorithms, setting a mutual agreement between exploration and exploitation.

Date of Award	2004
Original language	French
Supervisor	Philippe TOINT (Supervisor), Annick Sartenaer (Jury) & Fabian Bastin (Jury)