Skip to content

Bootstrapping Algorithm Settings

Kilian Brachtendorf edited this page Jan 28, 2019 · 2 revisions

Finding the correct settings for a genetic algorithm is a trial and error process. While domain knowledge and experience decrease the time needed to tweak the settings to get reasonable results never the less it is still a search problem. Isn't our genetic algorithm just doing this? Search for an optimal solution in a search place? How about we let Darwin bootstrap it's own parameters. Let a genetic algorithm instance search for the optimal parameter settings for another genetic algorithm.

A word of caution: This example works for problem which can quickly be computed. Algorithms that run for minutes or even seconds at a time are not practical to evolve.

Factorial design.

TODO