See Slides and recorded Video for the lecture on youtube.

• MDP framework, Terminology, Bellman equation
  • Markov Decision Process
  • Bellman equation
  • Bellman optimal equation
  • Infinite horizon discounted problem
  • Fix point iteration (Contraction mapping)
• Dynamic Programming
  • Value iteration (VI)
  • Policy iteration (PI)
    • Policy Evaluation
    • Policy Improvement
• Approximate PI
  • Bellman Error
  • Tabular TD(0)-learning
• Q-factor
  • Q-learning as stochastic VI (off policy)
  • optimistic PI for Q-factors: SARSA (on policy)

An overview of modern (Deep) Reinforcement Learning Algorithms:

• Things not yet covered
  • On-policy/Off-policy
  • Safe exploration

Discuss in the google groups!