A classifieur linéaire is a type of apprentissage automatique model used for classification tasks. It works by finding a combinaison linéaire of features that can separate different classes in a dataset. The core idea is to create a decision boundary—a straight line in two-dimensional space (or a hyperplane in higher dimensions)—that best separates the data points of different classes.
Linear classifiers operate under the assumption that the classes can be separated by a linear function. For example, in a classification binaire problem, if we have two classes, the goal is to determine the weights that define a line (in 2D) or a hyperplane (in higher dimensions) such that data points from one class are on one side, and data points from the other class are on the opposite side.
Exemples courants de classifieurs linéaires incluent :
- Régression logistique: Although it is used for binary classification, it predicts probabilities using a logistic function.
- Machines à vecteurs de support (SVM) : A linear SVM finds the hyperplane that maximizes the margin between the closest data points of each class.
- Perceptron : An early type of réseau neuronal qui peut être utilisé pour la classification binaire.
Les classifieurs linéaires sont particulièrement efficaces lorsque les données sont linéairement séparables, meaning that it can be perfectly divided by a straight line or a hyperplane. However, they may struggle with complex datasets where the relationship between features and classes is non-linear. In such cases, techniques like kernel methods or using more complex models (like neural networks) may be preferred.