An efficient total least squares algorithm based on a rank-revealing two- sided orthogonal decomposition (Q1208058)

A ``rank-revealing two-sided orthogonal decomposition'' decomposes a matrix into a product of a unitary matrix, a triangular matrix, and another unitary matrix in such a way that one can see the numerical rank of the matrix. The authors study such decompositions and base a new algorithm for total least-squares problems (also multidimensional and non-generic ones) on it. They carefully consider a parallel implementation of their method and compare its complexity to other total least squares algorithms.