Week 2 - Reinforcement learning basics

When evaluating different learners we normally evaluate them by the policy they produce. However, different methods of learning can create policies in different ways - therefore we may need to also consider:

• Computation complexity: The time it takes for that learner to come up with that policy.
• Sample complexity: The amount of interactions with its environment it needs to come up with that policy.

We don’t normally think about space complexity as with other subjects - as that is not normally a limiting factor.