O que é ResNet?
ResNet, ou Rede Residual, é um tipo de artificial arquitetura de redes neurais that was introduced by researchers at Microsoft in 2015. It is designed to tackle the problem of training very deep redes neurais utilizando uma estrutura única conhecida como conexões residuais.
Tradicional aprendizado profundo models often struggle with the vanishing gradient problem, where gradients (used to update the model’s weights) become too small for effective learning as they propagate back through many layers. ResNet addresses this issue by allowing gradients to flow through the network without being diminished, thanks to its skip connections.
A residual connection skips one or more layers and feeds the output of a previous layer directly into a later one. This means that the network can learn an ‘identity function’ if it needs to, which helps preserve information and makes it easier to train deeper networks. ResNet architectures can have hundreds or even thousands of layers, significantly improving their performance on tasks such as classificação de imagens e detecção de objetos.
ResNet achieved remarkable results in various competitions, including the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) in 2015, where it won first place. The architecture is now widely used in many applications beyond image processing, including processamento de linguagem natural e aprendizado por reforço.
No geral, ResNet teve um impacto significativo no campo do aprendizado profundo, demonstrando que redes mais profundas podem ser treinadas com sucesso e usadas para alcançar desempenho de ponta.