Algorithmic models a la Darwin can be dated back to the beginning of the digital computer era. They mimic evolutionary principles like mutation, recombination, natural selection, and sometimes more, as operators working on a data structure, called individual, representing a possible solution to a given problem. According to the population principle, evolutionary algorithms are inherently parallel by handling several individuals at the same time. Besides of basic similarities, there are some strong differences between algorithms of this class, e.g., between genetic algorithms and evolution strategies. Essential for the latter is the collective auto- adaptation of internal strategy parameters.
Though the underlying theories are still quite weak, the past decade has witnessed an exponential increase in diverse applications, from design synthesis, planning and control processes, to various other adaptation and optimization tasks. They cannot be reported adequately in an overview like this. But a summary of existing theoretical results will be given. A look back from existing practice of evolutionary computation to the natural prototype reveals a still substantial gap between current models of organic evolution and reality. This opens prospects for further advancements in the applicability and efficiency of the algorithms as well as a better understanding of some important real world phenomena, e.g. in ecosystems.
Página principal del curso