Perte binaire Entropie Perte (BCE Loss) is a widely used fonction de perte in apprentissage automatique, particularly for classification binaire problems. It measures the performance of a model whose output is a probability value between 0 and 1. The goal of using BCE Loss is to minimize the difference between the predicted probabilities and the actual binary class labels (0 or 1).
Mathématiquement, la perte d'entropie croisée binaire est définie comme :
BCE Loss = – (1/N) * Σ [y * log(p) + (1 – y) * log(1 – p)]
Où :
- N = nombre d'échantillons
- y = étiquette binaire réelle (0 ou 1)
- p = probabilité prédite de la classe positive
Cette formule calcule la perte moyenne sur tous les échantillons dans le dataset. The log function penalizes incorrect predictions more heavily, which helps the model to learn effectively. If the predicted probability is close to the actual class, the loss is small, and if it is far off, the loss is significant.
La perte d'entropie croisée binaire est particulièrement utile dans des scénarios tels que régression logistique, neural networks for binary classification, and other machine learning models where outputs can be interpreted as probabilities. By optimizing this loss function during training, models can improve their accuracy in classifying binary outcomes.