truncated singular value decomposition (svd)

A type of matrix factorization that is faster and more memory-efficient than traditional SVD when the input is sparse and only a few singular values are needed.