Let A be an m x n matrix with m greater than or equal to n. Then one form of the singular-value decomposition of A is A = U-T SigmaV, where U and V are orthogonal and Sigma is square diagonal. That is, UUT = I-rank(A), VVT ...