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?

More PCA interview questions

View all →