What is whitening in dimensionality reduction?
Updated May 16, 2026
Short answer
Whitening transforms data to have zero mean and unit variance with decorrelated features.
Deep explanation
It scales PCA components so that each has unit variance, removing correlations and making features independent in transformed space.
Real-world example
Used in signal processing for independent feature representation.
Common mistakes
- Assuming whitening preserves original feature meaning.
Follow-up questions
- Does whitening affect interpretability?
- Is whitening always required after PCA?