ResNet

ResNet, short for Residual Network, is a type of convolutional neural network (CNN) architecture designed to handle very deep networks. Introduced by Microsoft in 2015, it won the ImageNet competition by a significant margin.

The Innovation of ResNet

ResNet's key innovation is the introduction of "skip connections" or "shortcut connections" that allow the gradient to bypass one or more layers. This design combats the vanishing gradient problem and enables the training of networks with hundreds or even thousands of layers, significantly improving performance on tasks like image recognition.