Explain principal component analysis.
Whether you can give the geometric picture and say what PCA assumes.
Machine learningStandard
What a strong answer does
Geometrically, PCA rotates the coordinate system so that its new axes line up with the directions in which the data varies the most. Instead of describing each point using the original features, you describe it by its coordinates along these principal directions. The first principal component is the direction of maximum variance, the second is the best remaining orthogonal direction, and so on. Dimensionality reduction comes from keeping only the first few components and dropping the rest, preserving as much variance as possible in fewer coordinates.