Hands-On Neuroevolution with Python

Hyperparameters for the objective function candidates population

We create the hyperparameters for the evolution of the population of objective function candidates in the create_objective_fun_params function. Here, we discuss the most critical hyperparameters:

1. We decided to start with a small population to reduce computational costs. Also, the genotypes of the objective function candidate are not expected to be very complicated. Thus, a small population should be sufficient:

    params.PopulationSize = 100

2. As with the maze solvers, we are interested in producing compact genomes. Thus, the probabilities of adding new nodes and connections are kept very small:

    params.MutateAddNeuronProb = 0.03    params.MutateAddLinkProb = 0.05

We are not expecting ...