So-called Evolutionary Algorithms are not truly evolutionary.
# Thanks to Victor Venema of the University of Bonn for the image depicting an evolutionary search algorithm
Current-day so-called Evolutionary Algorithms (EA) are more
about Mendelian selective breeding than about Darwinian evolution.
Nature has no preconceived objectives. Evolution is
open-ended, and is more about adaptation to the ever-evolving environment than
about survival of the fittest. Although the “fittest” can be defined as those
that have adapted in the best to the changing environment.
Using Evolutionary Algorithms such as Genetic Algorithms to
find optimal solutions to decision-making is just directing the algorithm to a static goal. We cannot model the richness, complexity and
dynamism of Nature’s meandering adaptive walks.
The road of evolution knows not its final destination. It twists
and winds its way; the direction of the next step is the net sum of the interaction
(actions and reactions) between all participants at that point in time plus the
lingering effects of points further backwards in time.
Thus Evolution is a Markov chain with coevolution and feedback
loops at the core of the evolutionary process.
Comments
Post a Comment