Asymptotic behavior of evolutionary algorithms

This dissertation proposes a new approach to model and study the asymptotic behavior of evolutionary algorithms. Evolutionary algorithms are modeled here as a particular class of random dynamical systems. Two models are proposed: one using random iterated maps, and a second one using diffusion processes, which allow a complete characterization of their long-term behavior. This characterization explains to some extent the success of this type of algorithms in practical applications. The work in this dissertation is a contribution towards establishing a sound basis for the practical use of evolutionary algorithms.