Deep Residual Learning for Image Recognition

ResNet is the skip-connection paper that makes depth trainable in practice. It is beginner-friendly because the core idea is simple: learn a residual change instead of relearning the whole representation.

Reading focus: Why very deep networks can degrade even before overfitting is the issue. How identity shortcuts improve gradient flow and optimization. Why residual thinking reappears in modern Transformer blocks.

CVPR 2016. He et al.. 35 min read, very easy difficulty.