A modified Gram-Schmidt algorithm with iterative orthogonalization and column pivoting (Q1978115)

A row oriented version of the modified Gram-Schmidt orthogonalization process for computing the QR-factorization of a matrix is implemented that incorporates column pivoting and is capable of iterative orthogonalization. In addition to introducing the algorithm, the author also discusses related issues such as the need for reorthogonalization, connections with other orthogonalization techniques, and an error analysis. In addition, a modification of the standard least-squares algorithm is proposed that leads to certain desirable scaling properties of the computed residual.