This chapter looks at various enhancements and variations to DQN. Specifically, it looks at Prioritized Replay, DDQN (Double Q-Learning), Dueling DQN, NoisyNets DQN, C-51 (Categorical 51-Atom DQN), Quantile Regression DQN, and Hindsight Experience Replay. All the examples in this chapter are coded using PyTorch. This is an optional chapter with each variant of DQN as a standalone topic. You can skip this chapter in the first pass and come back to it when you want to explore specific variants of DQN.
The first ...
