A Entièrement connecté Couche (often abbreviated as FC layer) is a fundamental component of artificial réseaux neuronaux, particularly in apprentissage profond architectures. In this layer, each neuron is connected to every neuron in the subsequent layer, creating a dense network of connections. This structure allows the network to learn complex representations of the input data by combining features extracted from previous layers.
Le fonctionnement d'une couche entièrement connectée implique deux étapes principales : transformation linéaire and activation. First, the inputs are multiplied by a weight matrix, followed by the addition of a bias vector. The resulting values are then passed through an fonction d'activation, which introduces non-linearity into the model. Common activation functions used include ReLU (Rectified Linear Unit), sigmoid, and tanh, each providing different benefits depending on the application.
Les couches entièrement connectées se trouvent généralement vers la fin de réseaux de neurones convolutifs (CNNs), where they serve to integrate the high-level features extracted by preceding layers into final predictions or classifications. While they are powerful in capturing relationships in the data, fully connected layers come with a high computational cost and can lead to overfitting, especially when the number of neurons is large relative to the amount of training data.
Pour atténuer le surapprentissage, des techniques telles que dropout can be employed, randomly deactivating a subset of neurons during training to encourage the network to learn more robust features. Overall, fully connected layers play a crucial role in the success of many AI applications, including image recognition, traitement du langage naturel, and more.