Data science & AI · Reference
Supervised vs unsupervised learning
Supervised and unsupervised learning are the two foundational paradigms of machine learning, distinguished by whether the training data is labelled with the correct answers or not.
Supervised learning
In supervised learning, each training example is paired with a known correct output, or label. The model learns a mapping from inputs to labels so it can predict the label of new, unseen inputs. The two main supervised tasks are classification, where the output is a category (such as spam or not-spam), and regression, where the output is a continuous value (such as a temperature). Supervised learning needs labelled data, which can be costly to obtain, but it allows clear measurement of accuracy against the known answers.
Unsupervised learning
In unsupervised learning, the training data has no labels. The aim is to discover structure within the data itself. Clustering groups similar examples together; dimensionality reduction, such as principal component analysis, finds a compact representation that retains the data's main variation.
Because there are no correct answers to compare against, evaluating unsupervised results is harder and often relies on indirect measures or domain judgement. Unsupervised methods are valuable precisely when labels are unavailable or too expensive to create.
The key differences
The defining contrast is the presence of labels: supervised learning has them, unsupervised learning does not. This drives further differences. Supervised learning answers a specific predictive question and can be evaluated against ground truth; unsupervised learning explores data without predefined targets and is harder to evaluate. Supervised tasks split into classification and regression; unsupervised tasks include clustering and dimensionality reduction. Between them sit semi-supervised learning (a mix of labelled and unlabelled data) and self-supervised learning (labels derived automatically from the data), which power many modern models.
Choosing in research
Which paradigm to use follows from the data and the question. If labelled examples exist and the goal is prediction, supervised learning is the natural choice; if the goal is to explore or summarise unlabelled data, unsupervised methods fit. Researchers must guard against common pitfalls: data leakage and inappropriate metrics in supervised work; over-reading clusters that may be artefacts in unsupervised work. In both cases, validating findings and reporting methods clearly are essential for reproducible results.
Key facts
At a glance
- Key difference: labelled data (supervised) vs unlabelled (unsupervised)
- Supervised tasks: classification and regression
- Unsupervised tasks: clustering and dimensionality reduction
- Supervised: evaluated against known correct answers
- Unsupervised: harder to evaluate, no ground truth
- In between: semi-supervised and self-supervised learning
Common questions
FAQ
What is the main difference between supervised and unsupervised learning?+
The main difference is whether the training data is labelled. Supervised learning uses labelled examples to learn to predict outputs, while unsupervised learning uses unlabelled data to discover structure such as clusters or compact representations.
When should I use unsupervised learning?+
Unsupervised learning suits situations where labels are unavailable or too costly to obtain, and where the goal is to explore or summarise data — for example, grouping similar items or reducing dimensionality — rather than to predict a specific known output.
What is semi-supervised learning?+
Semi-supervised learning uses a mix of a small amount of labelled data and a larger amount of unlabelled data. It sits between the two main paradigms and is useful when labelling everything would be too expensive.
Going deeper
Related on CASRAI
- What is machine learning? →
- What is PCA? →
- What is deep learning? →
- What is data science? →
- Computer science, data science & AI →
Sources
The step most authors miss
Doing CRediT right? Don’t stop at the statement.
A CRediT statement credits you inside one paper. The recognition CRediT was built for happens when those roles are tied to you, persistently. Sign in with your ORCID — free — and claim your CRediT contributions on casrai.org, the home of the standard. They become a verified, portable part of your identity, not a line that disappears into one PDF.
Free: claim your contributions, then export a journal-ready CRediT statement, schema.org structured data, JATS XML, CSV or BibTeX — and preview your public profile. A membership publishes that profile publicly and verifies the journals you serve.
