Classical Reinforcement Learning

Classical reinforcement learning studies how agents learn optimal policies through interaction with the environment, forming the theoretical foundation of deep reinforcement learning.

Contents:

• Introduction to Classical RL — MDP framework, value functions, policies
• Multi-armed Bandits — Exploration vs. exploitation, UCB, Thompson sampling
• Finite MDP — Bellman equations, optimal policies
• Dynamic Programming — Policy iteration, value iteration
• Monte Carlo Methods — MC prediction, MC control, importance sampling
• TD(0) — Temporal difference learning, SARSA, Q-Learning
• N-step TD — Multi-step bootstrapping, bias-variance tradeoff
• Learning & Planning — Dyna architecture, model learning
• Approximation Methods — Function approximation, linear methods
• TD(lambda) — Eligibility traces, forward view & backward view
• Policy Gradient — REINFORCE, baseline functions, Actor-Critic

---

评论