What is ResNet and why are residual connections important?
Updated May 15, 2026
Short answer
ResNet introduces residual connections to solve vanishing gradient problems in deep networks.
Deep explanation
ResNet (Residual Networks) uses skip connections that allow gradients to flow directly through layers by learning residual mappings instead of full transformations. Instead of learning H(x), the network learns F(x) = H(x) - x, so output becomes F(x) + x. This helps train very deep networks (50, 101, 152+ layers) without degradation.
Real-world example
Used in medical imaging models to classify tumors with very deep architectures.
Common mistakes
- Assuming deeper networks always improve accuracy without residual connections.
Follow-up questions
- Why do deep networks suffer from vanishing gradients?
- What is identity mapping in ResNet?