A shortcut in neural network architectures that adds the input of a layer to its output, facilitating the training of deeper networks by mitigating vanishing gradients.
Last updated 10 months ago