Reinforcement Learning Algorithms with Python

The main function

This is defined in a function called ES that has the following arguments: the name of the Gym environment, the size of the neural network's hidden layers, the total number of generations, the number of workers, the Adam learning rate, the batch size, and the standard deviation noise:

def ES(env_name, hidden_sizes=[8,8], number_iter=1000, num_workers=4, lr=0.01, batch_size=50, std_noise=0.01):

Then, we set an initial seed that is shared among the workers to initialize the parameters with the same weights. Moreover, we calculate the number of individuals that a worker has to generate and evaluate on each iteration and create two multiprocessing.Queue queues. These queues are the entry and exit points for the variables that ...