Hands-On Neuroevolution with Python

Agent record visualization

In this experiment, we presented a new visualization method that allows us to visually discern the performance of various species in the evolutionary process. The visualization can be performed using the following command, executed from the working directory of the experiment:

$ python visualize.py -m medium -r out/maze_objective/medium/data.pickle --width 300 --height 150

The command loads records about the fitness evaluation of each maze-solving agent during the evolution, which is stored in the data.pickle file. After that, it draws the final positions of the agents on the maze map at the end of the maze-solving simulation. Each agent's final position is presented as a color-coded circle. The color of the circle ...