Reference · Glossary
Unsupervised learning
Last updated
Finding **patterns in data without labels** — grouping similar items, detecting anomalies, or reducing dimensions.
When to use
Customer segmentation, log anomaly detection, exploratory data analysis, or pre-training representations before fine-tuning.
When not to
When you need a specific yes/no decision with clear ground truth — supervised learning is usually clearer and easier to evaluate.
Example
Cluster support tickets by embedding text — discover recurring themes nobody manually tagged yet.