Gated Linear Unit (GLU)
Eine Gated Linear Unit (GLU) ist eine Aktivierungsfunktion für neuronale Netzwerke designed to enhance the model’s ability to capture complex relationships in data by incorporating gating mechanisms. Introduced in a paper by Yann N. Dauphin et al. in 2017, GLUs help in improving the performance of deep learning architectures, particularly in tasks related to der Verarbeitung natürlicher Sprache (NLP) und andere sequenzielle Daten.
GLUs arbeiten, indem sie lineare Transformationen mit Toren verbinden, die den Informationsfluss steuern. Die Grundformel für eine GLU kann wie folgt ausgedrückt werden:
GLU(x) = (W_1 * x) ⊗ σ(W_2 * x)In this formula, W_1 and W_2 are weight matrices, x is the input data, and σ represents the sigmoid activation function. The output of the GLU is the element-wise product of a linearen Transformation and a gating mechanism, which allows the model to learn which features to focus on while ignoring others.
The gating mechanism provides a way to control the information from the input that gets passed forward through the network, allowing for more effective learning and improved gradient flow. This is particularly useful in deep networks where verschwindende Gradienten kann ein erhebliches Problem sein.
GLUs are often used in combination with other neural network layers, such as convolutional or recurrent layers, to enhance their performance. They can be seen as an evolution of traditional Aktivierungsfunktionen like ReLU, as they add a layer of complexity and adaptability that helps in various applications, including language modeling, machine translation, and more.