Principal Components Analysis: Transformation to new space where features are decorrelated
Linear Discriminant Analysis: Projection of d-dimensional data onto a line; Dimensionality reduction by mapping L distributions to (L-1)-dimensional subspace; Maximize class separability