Reference · Glossary

Supervised learning

Last updated

Machine learning from **labeled examples** — each input comes with the correct answer the model should learn to predict.

When to use

Spam vs not spam, cat vs dog photos, sentiment labels, fraud yes/no — whenever you have trustworthy labels at scale.

When not to

When labels are missing, expensive, or wrong — consider unsupervised clustering or semi-supervised approaches instead.

Example

10,000 emails marked "spam" or "not spam" → train a classifier → flag new inbox messages automatically.