Hands-On Reinforcement Learning for Games

Exercises

The exercises in this section are a bit wider in scope in this chapter in hopes you look through several frameworks on your own:

1. Take some time and look at one of the frameworks listed earlier but not reviewed in this chapter.
2. Use SimpleRL to solve a grid-world MDP that is different than the one in the example. Be sure to take the time to tune hyperparameters.
3. Use Google Dopamine to train an agent to play the LunarLander environment. The best choice is likely RainbowDQN or a variation of that.
4. Use Keras-RL to train an agent to play the Lunar Lander environment; make sure to spend time tuning hyperparameters.
5. Use RL Lib to train an agent to play the Lunar Lander environment; make sure to spend time tuning hyperparameters.
6. Modify ...