Hands-On Genetic Algorithms with Python

Improving the speed with an increased mutation rate

If we zoom in at the lower part of the fitness axis, we will notice a relatively large improvement of the best result found (red line) around generation 180, accompanied by a large swing of the average results (green line):

Enlarged section of the first program's stats graph

One way to interpret this observation is that perhaps introducing more noise can lead to better results faster. This could be another manifestation of the familiar principle of exploration versus exploitation we've discussed several times before—increasing the exploration (which manifests itself as noise in the diagram) ...