Chapter 10

1. A3C is the Asynchronous Advantage Actor Critic network which uses several agents to learn parallel.
2. Three A's are Asynchronous, Advantage, Actor Critic.
3. A3C requires less computation power and training time than DQN.
4. All agents (workers) works in copies of the environment and then global network aggregate their experience.
5. Entropy is used to ensure enough exploration.
6. Refer section How A3C works.