What is supervised learning (vs unsupervised)?

Supervised learning is when a computer studies examples that already have correct answers attached. It practices until it can give the right answers to new examples on its own.

7 min read min de lecture

~$ man apprentissage-supervise

What is supervised learning (vs unsupervised)?

Machine & Deep Learning gneurone encyclopedia
Supervised learning is when a computer studies examples that already have correct answers attached. It practices until it can give the right answers to new examples on its own.

definition

Supervised learning is a machine learning approach where models are trained on labeled datasets. Each training example includes both input features and the correct output label.

The algorithm learns a mapping from inputs to outputs by minimizing prediction errors during training. Common tasks include classification for discrete labels and regression for continuous values.

Unlike unsupervised learning, which finds hidden patterns in unlabeled data, supervised learning requires human-provided labels and is used when the goal is prediction or classification.

Think of teaching a child to identify fruits. You show many pictures of apples, bananas and oranges, each labeled with the correct name. After enough examples the child can name a new fruit picture correctly without help.

key takeaways

  • Requires labeled training data created or verified by humans.
  • Main goal is accurate prediction on unseen data.
  • Includes two main task types: classification and regression.
  • Common algorithms are decision trees, support vector machines and neural networks.
  • Performance is measured with metrics such as accuracy, precision and mean squared error.

the 2026 job market

In 2026 supervised learning remains a core skill for data scientists and machine learning engineers building prediction systems in healthcare, finance and recommendation engines. Job postings increasingly list experience with labeled datasets and evaluation of supervised models as required qualifications.

Data Scientist · $115000-$165000 / CAD 95000-CAD 145000 / £62000-£92000Machine Learning Engineer · $130000-$185000 / CAD 105000-CAD 160000 / £70000-£105000

frequently asked questions

What are common algorithms used in supervised learning?

Popular algorithms include linear regression, logistic regression, decision trees, random forests and gradient boosting. Neural networks are also widely used for complex supervised tasks. Choice depends on data size, feature types and whether the task is classification or regression.

How is training data prepared for supervised learning?

Data preparation involves collecting raw examples, cleaning errors, handling missing values and assigning accurate labels. Features are often scaled or encoded before training. The dataset is usually split into training, validation and test sets.

What is the difference between classification and regression?

Classification predicts discrete categories such as spam or not spam. Regression predicts continuous numeric values such as house prices. Both are supervised learning tasks but use different loss functions and evaluation metrics.

When should supervised learning not be used?

Avoid supervised learning when labeled data is unavailable or too expensive to create. It is also unsuitable for discovering unknown patterns without a clear prediction goal. In those cases unsupervised or semi-supervised methods may be better.

courses to go further

$ cat ./full-guide.mdLance-toi en Machine Learning Débutants : ton premier pas concret aujourd'huiread the guide →

related terms

< back to the encyclopedia

Auteur(s)

R

REHOUMA Haythem

Haythem Rehouma est un ingénieur et architecte IA et cloud, formateur et enseignant technique, avec un profil orienté IA médicale, AWS, MLOps, LLM/RAG et vision par ordinateur.