Reinforcement Learning

October 4, 2026 machine-learning

Handwritten notes on reinforcement learning: Markov decision processes, Q-learning, policy gradients, reward shaping, and exploration vs exploitation.

These notes cover the fundamentals of reinforcement learning — agents learning through interaction:

Markov Decision Processes: States, actions, rewards, transitions, discount factors, and the Bellman equation.
Value-Based Methods: Q-learning, SARSA, value iteration, and temporal difference learning.
Policy-Based Methods: Policy gradients, REINFORCE, actor-critic architectures, and advantage functions.
Exploration vs Exploitation: Epsilon-greedy, UCB, Thompson sampling, and the exploration-exploitation tradeoff.