Many thanks to everyone who decided to participate in the Breeding Objects Experiment 01. My initial hypothesis was that configurators with artificial selection in a GA implemented may prove to be a better tool to navigate through solution spaces with larger amount of dimensions. In other words, genetic algorithms are better tools to look for values of parameters in case of more complicated parametric definitions, that is definitions driven by large set of parameters.

This was confirmed by more than 300 sessions with the configurators, which took place during last three weeks. Parametric configurators with 4 and 18 customizable parameters (4P and 18P) where more popular than their genetic algorithms counterparts (4GA and 18GA), but genetic algorithm was a much more popular mean of driving the 37th parameters definition (37GA turned out to be more popular than 37P). I’m currently writing these findings up into a paper and some further, more in-depth analysis of the results will follow.

For the next iteration of the experiment, Federico Weber has generously offered to share his customizable table design with me. The work called Xylem is based on Voronoi diagram and will become a phenotype in Breeding Objects Experiment 2. Federico has already developed a parametric configurator for Xylem. It’s downloadable from here.