How does PCA mathematically reduce dimensionality?
Updated May 15, 2026
Short answer
It projects data onto eigenvectors of covariance matrix.
Deep explanation
PCA finds orthogonal axes maximizing variance using eigen decomposition or SVD.
Real-world example
Compression of image datasets.
Common mistakes
- Ignoring scaling before PCA.
Follow-up questions
- What are eigenvalues?
- Why orthogonality matters?