Partial total least squares algorithm (Q751180)

The paper contains a detailed description of a subroutine, PTLS (partial total least squares), for the total least squares (TLS) solution of the overdetermined system of linear equations \(AX=B\) by using a partial singular value decomposition. It is claimed that the computational efficiency is improved by a factor of about two over the classical TLS algorithm. The computational improvement of PTLS is based on the observation that the TLS solution can be computed from any orthogonal basis of the right singular subspace corresponding to the smallest singular values of the augmented matrix [A;B]. The routine PTLS can accommodate more than one vector in B. It can also be used to solve determined and underdetermined sets of equations by computing the minimum norm solution.