While your motivation may vary as to why you are reading this book, hopefully by now you can appreciate the value of just doing things on your own. As always, we present these exercises for your enjoyment and learning, and hope you have fun completing them:

1. Select another sample scene that uses discrete actions and write the reward functions that go with it. Yes, that means you will need to open up and look at the code.
2. Select a continuous action scene and try writing the reward functions for it. While this one may be difficult, it is essential if you want to build your own control training agent.
3. Add Curriculum Learning to one of the other discrete action samples we have explored. Decide on how you can break the training into levels of difficulty and create parameters for controlling the evolution of the training.
4. Add Curriculum Learning to a continuous action sample. This is more difficult, and you likely want to perform exercise number two first.
5. Implement actual Backplay on the Hallway environment by placing the agent starting at the goal and then, as the agent trains, move it back to the desired start with Curriculum Learning. 
6. Implement Backplay on another discrete action example you have run and see the effect it has on training.
7. Implement Curiosity Learning on the VisualPyramids example and notice the difference in training.
8. Implement Curiosity Learning on a continuous action example and notice the effect it has on training. Is it what you expected?
9.  Disable Curiosity Learning on the Pyramids example and see what effect this has on agent training.
10. Think of a way in which you could add Backplay to the VisualPyramids example. You'll get bonus points if you actually build it.

As you can see, the exercises are getting more demanding as we progress through the book. Remember, even completing one or two of these exercises will make a difference in your take-away knowledge.