midPCA
What is whitening in PCA?
Updated May 17, 2026
Short answer
Whitening scales PCA components to have unit variance and removes correlations.
Deep explanation
Whitening transforms PCA outputs so each component has zero mean, unit variance, and is uncorrelated. It is useful in preprocessing for neural networks.
Real-world example
Preprocessing input for deep learning models.
Common mistakes
- Using whitening without understanding noise amplification.
Follow-up questions
- Does whitening improve all models?
- What is drawback?