ResNet: Residual Networks
ResNet, introduced by Microsoft Research in 2015, is one of the most influential deep learning architectures ever published — with over 100,000 citations. The core idea is surprisingly simple: instead of learning a mapping from input to output, learn only the difference between them. The Super-Resolution Intuition The easiest way to understand ResNet is through image super-resolution. Say you have a low-resolution image $X$ and you want to produce a high-resolution version $Y$. ...