Hands-On Neuroevolution with Python

Creating the population of maze solvers

The maze-solver agent needs to get inputs from 11 sensors and generate two control signals, which affect the angular and linear velocity of the robot. Thus, the genome encoding the phenotype of the maze solver must yield phenotype configurations that include 11 input nodes and two output nodes. You can see how the creation of the initial population of genomes for the maze-solver agent is implemented by taking a look at the create_robot function:

    params = create_robot_params()    # Genome has 11 inputs and two outputs    genome = NEAT.Genome(0, 11, 0, 2, False,                         NEAT.ActivationFunction.UNSIGNED_SIGMOID,                         NEAT.ActivationFunction.UNSIGNED_SIGMOID,                         0, params, 0) pop = NEAT.Population(genome, params, True, 1.0, ...