seniorSupervised Learning
What is residual learning in supervised learning?
Updated May 17, 2026
Short answer
Residual learning focuses on learning the difference between predicted and actual values.
Deep explanation
Instead of learning the full mapping, residual learning models the error (residual) between prediction and true value. This idea is used in boosting and ResNet architectures. Learning residuals simplifies optimization because correcting mistakes is easier than learning full transformations.
Real-world example
XGBoost improving predictions iteratively by correcting previous errors.
Common mistakes
- Thinking residual learning is only used in neural networks.
Follow-up questions
- Why do residuals help optimization?
- Where else is residual learning used?