We finally kicked off our MLT Reinforcement Learning sessions with an “Intro to RL” study group at the Tokyo Metropolitan Library. Anugraha Sinha walked us through the following:
- Introduction to RL
- Important elements of an RL problem
- Description of Markov Decision Process (MDP) and and Markov Assumption.
- Importance of parametrization of State, Action, Reward and Environment.
- Model Based and Model Free Methods
- Meaning of Control Problem and Evaluation Problem.
- Algorithm of Policy Evaluation and Value iteration methods
- Finding the best route through a maze/obstruction avoidance using policy iteration algorithm.
- Above problem statement with value iterations algorithm.
- Code exercise