Chapter 12

1. DQN computes the Q value directly whereas Dueling DQN breaks down the Q value computation into value function and advantage function.
2. Refer section Replay memory.
3. When we use the same network for predicting target value and predicted value there will lot of divergence so we use separate target network.
4. Refer section Replay memory.
5. Refer section Dueling network.
6. Dueling DQN breaks down the Q value computation into value function and advantage function whereas double DQN uses two Q function to avoid overestimation.
7. Refer section Dueling network.