• 19: low rank approximations
    • motivation: main use of SVD in practice is to approx data by a low rank matrix