Supervised Learning: Definition and Examples

Supervised learning is one of the fundamental approaches of machine learning. Its principle is simple: we provide the model with a training set consisting of input-output pairs, where each input is associated with the expected answer (called a label). The model progressively learns to identify patterns that link inputs to outputs.

Concretely, imagine a teacher grading a student's tests. At each exercise, the student proposes an answer and the teacher indicates whether it is correct or not. Over time, the student learns the underlying rules and becomes able to solve new similar exercises. That is exactly how supervised learning works: the model adjusts its internal parameters to minimize the gap between its predictions and the correct answers.

There are two main categories of tasks in supervised learning. Classification consists of assigning a category to a data point (e.g., determining whether an email is spam or not). Regression aims to predict a continuous numerical value (e.g., estimating the price of a property based on its features).

Supervised learning is ubiquitous today: image recognition, machine translation, medical diagnosis, bank fraud detection, content recommendation. Its main constraint is that it requires a sufficient amount of labeled data, which can represent a significant cost in time and human resources for annotation.