Chapter 8, Further Q-Learning Research and Future Projects

1. The RL equivalent of a labeled training dataset is a standardized set of environments used to train models built by different researchers using different algorithms so that the models can be meaningfully compared to each other.
2. As one example, enumerating a different set of states or actions for the same environment can greatly increase the difficulty of solving that environment. Gym simplifies this process by standardizing state and action spaces for all environments and giving researchers a level playing field on which to compare results.
3. The movement of an actuated joint can be controlled, such as by a motor, while an unactuated joint moves freely and is not controlled by an outside source.
4. Being able to find a general solution to a state space that applies to more than one space gives us the experience and potential...