AlgoPlus//machine learning / q-learning
Read the theory

Q-Learning

Learn action-values by trial and error — no model of the world required.

Episode
Last reward
Legend
Agent
Learned value
AI Tutor Workspace
In a nutshell
Q-learning learns good behaviour without ever being told how the world works. It keeps a Q-value for each state-and-action pair — an estimate of the total future reward that move is worth — and refines it from experience using the Bellman update: nudge the old estimate toward the reward just received plus the discounted best value of the next state. To keep discovering, the agent sometimes acts at random (exploration) and otherwise takes its best-known move (exploitation). Given enough episodes the Q-values settle and the best move everywhere becomes the optimal policy.
Ready
Press play to begin the cinematic walkthrough.
Learn by bruising. Try things, remember which moves near the goal paid off, and slowly the remembered scores point the way — no map of the world needed.
Key terms
Go deeper in the lesson
Read the full theory, intuition & complexity for Q-Learning.