Supervised learning

noun

Supervised learning is a machine learning technique where an algorithm is trained on a labeled dataset, meaning it learns from input-output pairs. For example, if you teach an AI to recognize cats, you’d show it thousands of labeled images of cats and non-cats.

It works by feeding the model data with correct answers, allowing it to adjust its internal rules to minimize errors. This contrasts with unsupervised learning, where the AI finds patterns without predefined labels.

For developers, supervised learning is powerful but requires high-quality, labeled data, which can be expensive and time-consuming to gather. For instance, training a language model like a chatbot typically requires millions of annotated sentences, costing thousands of dollars in data preparation.

A spam filter uses supervised learning to classify emails as 'spam' or 'not spam' based on labeled examples.

See also: Unsupervised learning

Découvrez nos autres services

ElpisIA, c’est tout un univers — explorez nos autres services.