Questions

1. What is the RL equivalent of a labeled training dataset in supervised learning?
2. What is one of the difficulties of not having a standardized set of environments for developing RL algorithms? How does Gym attempt to solve this problem?
3. What is the difference between an actuated joint and an unactuated joint?
4. What is the benefit of being able to use a single algorithm to solve more than one environment? Explain in two to three sentences.
5. What is the importance of being able to solve generalized control problems in robotics motion?
6. Briefly describe the relationship between a probability distribution and our current estimation of the likelihood of an event.

7. Explain what difference it makes to have a state space available in a contextual bandit problem.
8. Describe the differences in the results of A/B testing versus multi-armed bandit testing in two to three sentences.
...