Summary

In this chapter, we learned how to implement reinforcement learning algorithms in Keras. For the sake of keeping the examples simple, we used Keras; you can implement the same networks and models with TensorFlow as well. We only used a one-layer MLP, as our example game was very simple, but for complex examples, you may end up using complex CNN, RNN, or Sequence to Sequence models.

We also learned about OpenAI Gym, a framework that provides an environment to simulate many popular games in order to implement and practice the reinforcement learning algorithms. We touched on deep reinforcement learning concepts, and we encourage you to explore books specifically written about reinforcement learning to learn deeply about the theories and concepts.

Reinforcement Learning is an advanced technique that you will find is often used for solving complex problems. In the next chapter, we shall learn another family of advanced deep learning techniques: Generative Adversarial Networks.