| Popis: |
This study updates a novel technique for evolving parameters that specify fractal images. Example parameter sets are provided as an information resource to evolution, following an earlier study. Instead of choosing parameters with high average compatibility with all other parameters, this study clusters the parameters using a graph clustering algorithm within a network where the adjacency relation of the network is derived from co-fertility, i.e. genetic compatibility values. The result of using the new types of sets of parameters as information resources is studied and compared to evolution that uses the previous type of information resource. The new technique of selecting information resources presented here yields higher fitness values. The new results are on the high end of the fitness distribution, and so the new information resources tested give similar improvements in fitness. However, their variability vary substantially and the resulting fractals have different appearances. |
