In this chapter, we introduced some advanced RL techniques, starting with deep Q-learning. Then, we used DQN to teach an agent to play the Atari Breakout game with moderate success. Next, we introduced policy-based RL methods, which approximate the optimal policy instead of the true value functions. Then, we used A2C to teach an agent how to play the cart pole game. Finally, we introduced model-based RL methods and MCTS in particular.
In the next chapter, we'll explore how to apply deep learning in the challenging and at the same time exciting area of autonomous vehicles.