A conexão residual is a technique used in aprendizado profundo, particularly in redes neurais, to help improve their training and performance. The concept was popularized by the ResNet (Residual Network) architecture, which won the ImageNet competição em 2015.
In a typical neural network, data flows sequentially through layers, where each layer applies certain transformations. However, as networks become deeper (with more layers), they can experience issues such as gradientes que desaparecem, where the gradients used to update weights during training become very small, hindering learning.
As conexões residuais resolvem esse problema permitindo que a entrada de uma camada bypassie uma ou mais camadas e seja adicionada diretamente à saída dessas camadas. Isso é representado matematicamente como:
Saída = F(Entrada) + Entrada
Here, F(Input) represents the transformation applied by the layers being bypassed. By including the original input in the output, residual connections help maintain the flow of information and gradients, making it easier for the network to learn complex padrões.
These connections also allow for the training of much deeper networks, leading to better performance on various tasks like image recognition, processamento de linguagem natural, and more. Overall, residual connections are a crucial innovation in modern deep learning, facilitating the development of more sophisticated AI models.