Representative subset selection using genetic algorithms

The selection of representative subsets was performed using genetic algorithms (GAs). The dissimilarity (Dis) of selected samples within a subset and the mean of the product–moment correlation coefficients (MP) were used as fitness function of the GAs. We applied the GAs to select representative subsets from the dataset containing 5000 samples which were randomly selected from Maybridge catalog. Two fitness functions didnʹt always select corresponding subsets; higher Dis didnʹt always lead higher MP and vice versa. The data structures of selected subsets were compared by using box counting analysis. The comparison suggested that the data structures of the selected subsets with higher MP reflect original data structure more correctly.