Supervised Learning
Learning a mapping from inputs to known labels using labeled training examples.
- Needs (input, label) pairs as the training signal.
- Classification = discrete labels; regression = continuous labels.
- Dominates production ML because the signal is clear.
When to use: When you have labels (or can get them) and the goal is to predict that same label on new data.
Example: Predict whether a credit card transaction is fraudulent given amount, merchant, and hour, trained on 100k historically labeled transactions.