As always, try at least one or two of the following exercises on your own for your own enjoyment and learning:

1. Open the BananaCollectors example Banana scene and run it in training mode.
2. Modify the BananaCollectors | Banana scene so that it uses five separate learning brains and then run it in training mode.
3. Modify the reward functions in the last SoccerTwos exercise to use exponential or logarithmic functions.
4. Modify the reward function in the last SoccerTwos exercise to use non-inverse related and non-linear functions. This way, the mean modifying the positive and negative rewards is different for each personality.
5. Modify the SoccerTwos scene with different characters and personalities. Model new rewards functions as well, and then train the agents.
6. Modify the BananaCollectors example Banana scene to use the same personalities and custom reward functions as we did with the SoccerTwos example.
7. Do exercise 3 with the BananaCollectors example.
8. Do exercise 4 with the BananaCollectors example.
9. Do exercise 5 with the BananaCollectors example.
10. Build a new multi-agent environment using one of the current samples as a template or create your own.  This last exercise could very likely turn into your very own game.

You may have noticed by now that as we progress through the book, the exercises become more time-consuming and difficult. Please try for your own personal benefit to complete at least a couple of the exercises.