In the previous chapter, we implemented the Deep Q-Network (DQN) model published by DeepMind in 2015. This paper had a significant effect on the Reinforcement Learning (RL) field by demonstrating that, despite common belief, it's possible to use nonlinear approximators in RL. This proof of concept stimulated large interest in the deep Q-learning field in particular and in deep RL in general.

Since then, many improvements have been proposed, along with tweaks to the basic architecture, which significantly improve convergence, stability and sample efficiency of the basic DQN invented by DeepMind. In this chapter, we'll take a deeper look at some of those ideas. Very conveniently, in October 2017, DeepMind published a paper called Rainbow: Combining Improvements in Deep Reinforcement Learning ([1] Hessel and others, 2017), which presented the seven most important improvements to DQN, some of which were invented in 2015, but some of which are very...