Fig 3. Description of the genetic algorithm.

The figure summarizes all the steps of the genetic algorithm during a single generation: 1) an initial random population of 100 individuals composed of 8-bit genotypes is generated; 2) the genotypes of the initial population are expressed into phenotypes for each individual, i.e. the actual networks’ weights and biases; 3) the individuals’ phenotypes are evaluated during the training phase by measuring their behavior with the fitness function; 4) the fitness values obtained by each population’s individual are ranked; 5) the 20 best individuals are allowed to reproduce and to generate an offspring of 100 individuals by flipping the original genotypes bits with a 0,01% mutation rate; 6) the offspring constitutes the new generation which can go through the selection cycle again.