The module includes configurations used by the Controller and the child networks. Here, you can adjust several training parameters, such as the number of episodes, the learning rate, and the number of child networks generated by the Controller per epoch. You can also experiment with child network sizes, but do note that the larger the child network, the longer training takes for both the Controller and the child network:

child_network_params = {    "learning_rate": 3e-5,    "max_epochs": 100,    "beta": 1e-3,    "batch_size": 20}controller_params = {    "max_layers": 3,    "components_per_layer": 4,    'beta': 1e-4,    'max_episodes': 2000,    "num_children_per_episode": 10}

Some of these numbers (such as max_episodes) are arbitrarily chosen. We ...

Get Python Reinforcement Learning Projects now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.