1. Reinforcement Learning

2. Overview Tabular Methods Approximate Methods
Deep Reinforcement Learning

3. Tabular Methods

4. Model: Mathematical models of dynamics and reward
Policy: function mapping agent’s states to action
Value function: future rewards from being in a state and/or action when following a particular policy

5. MDP

6. Markov Reward Process

7. Markov Reward Process

10. MDP = MRP + Action

11. MDP + Policy

12. Compare

13. How to Control?

14. Policy Search

16. State-Action Value Q

17. Policy Iteration

21. Worst Case
Policy Iteration Can Take At Most |A|^|S| Iterations* (Size of # Policies)

22. Value Iteration