Source code of Reinforcement Leanring, 2023, Spring

Lecture-01: Monte Carlo Simulation Method

• The function of Monte Carlo Methods;
• Use MC to caculuate complicated probabilities;
• Use MC to solve BlackJack
• Important terminologies in Reinforcement Learning:
  • Environment, Obvervation
  • State, Action, Reward, Goal
  • epsilo, gamma
  • Value-Function, Q-Value-Function
• off-line learning
• first-visit learning
• on-line learning

Lecture-02: Markov Decision Process, Dynamic Programming, RL Evaluation, Self-Playing Training

• What is Markov Decision Process.
  • Memoryless, $Pr(s_t | s_{t-1} s_{t - 2}) = Pr(s_t | s_{t-1})$
• The dynamic programming character in Reinforcement Learning
• How to Evaluation
  • Mean-Reward
  • Win-rate
• Self-Playing
  • BlackJack-PK
  • Five in Row