Let's evaluate our understanding of DQN and its variants by answering the following questions:
- Why do we need a DQN?
- What is the replay buffer?
- Why do we need the target network?
- How does a double DQN differ from a DQN?
- Why do we have to prioritize the transitions?
- What is the advantage function?
- Why do we need LSTM layers in a DRQN?