“Ce n’est qu’en essayant continuellement que l’on finit par réussir. Autrement dit: plus ça rate, plus on a de chances que ça marche.”
“Only by continually trying we finally succeed. In other words: the more it fails, the more likely it will work.”
The above motto is a good summary for evolutionary algorithms (EAs) and for stochastic searches in general: a repetition of random trials till the optimum is reached. Of course, there is a bit more to it than that, a hint of “intelligent” control in these algorithms.
More seriously, EAs, also known as genetic algorithms (GAs), evolution strategies (ESs), evolutionary programming (EP) or artificial evolution, are stochastic optimization methods based on a simplified model of natural evolution, according to Darwin’s theory. This chapter proposes an overview of these methods, with a focus on their extreme versatility, which is one of the reasons for their success in a large variety of application domains. It has to be noted that artificial evolution is not limited to pure optimization applications, as there are other uses of these techniques, in particular when they are embedded in an interactive framework. Implementations of EAs are, however, computationally expensive, and a fine apperception of artificial evolution mechanisms helps to efficiently tune their various components. The most efficient applications of EAs are often based on hybridizations with ...