La perte L2, communément appelée Erreur quadratique moyenne (MSE), is a popular fonction de perte utilisé en apprentissage automatique and modélisation statistique to measure the accuracy of a model’s predictions. It quantifies the difference between the actual values and the values predicted by the model.
Mathématiquement, la perte L2 est calculée en prenant la moyenne des différences au carré entre chaque valeur prédite et la valeur réelle correspondante. La formule est donnée par :
L2 Loss = (1/n) * Σ(actual – predicted)²
where n is the number of observations, actual is the actual value, and predicted est la valeur prédite.
The key characteristic of L2 Loss is that it heavily penalizes larger errors due to the squaring of the differences. This property makes it sensitive to outliers, which can significantly affect the perte globale value. As a result, L2 Loss is often used in regression tasks where the goal is to minimize the error between predicted and actual values, leading to more accurate models.
While L2 Loss is widely used, it may not always be the best choice, especially in situations where outliers are present. In such cases, alternative loss functions like L1 Loss (Erreur Absolue Moyenne) ou la perte Huber peuvent être plus appropriées.