Hands-On Neuroevolution with Python

Experiment runner implementation

The experiment runner implementation creates the concurrent evaluation environment and runs the evolution loop over the population of individuals. Let's discuss the essential implementation details:

1. We start by setting up the evaluation environment by loading the controller ANN model and creating the concurrent workers to execute the evaluation:

    Model = neuroevolution.models.__dict__[config['model']]    all_tstart = time.time()    def make_env(b):        return gym_tensorflow.make(game=config["game"],                                    batch_size=b)    worker = ConcurrentWorkers(make_env, Model, batch_size=64)

2. Next, we create a table with random noise values, which will be used as random seeds, and define the function to create offspring for the next ...