Hands-On Neuroevolution with Python

Hyperparameter selection

According to the objective function definition, the maximum value of the navigator agent fitness score that can be obtained by reaching the maze exit area is 1.0. We are also expecting that the initial configuration of the controller ANN is more complicated than in the previous experiments described earlier in the book, and this will impact the speed of the algorithm's execution. Due to this, it would take too long on a moderate PC to complete the neuroevolution algorithm with a significantly large genome population. But at the same time, the task at hand is much more complicated than in previous experiments and requires using a wider search area for successful solution exploration. Thus, through trial and error, ...