Supervised, unsupervised and reinforcement learning

Supervised learning uses labelled examples $(x,y)$ and learns to predict $y$ from $x$. Regression and classification are supervised. Examples: spam detection, price prediction, diagnosing disease from an X-ray.

Unsupervised learning uses only inputs $x$, with no labels, and looks for structure: groups (clustering), lower-dimensional summaries (PCA), unusual points (anomaly detection) or a model of the data distribution itself (generative models).

Reinforcement learning (RL) has an agent that acts in an environment and receives rewards. There are no correct answers to copy — only a reward signal, often delayed. The agent learns a policy that maximises total reward. Examples: game-playing agents, robot control, recommendation that adapts to user feedback.