Principal Component Analysis

Data matrix X centered; SVD → principal axes = right singular vectors. Projections maximize variance. Dimensionality reduction.