A direct method for sparse least squares problems with lower and upper bounds (Q1112557): Difference between revisions

The least squares problem \(\| Ax-b\|_ 2\to Min\). (the solution of which being subject to the additional restriction \(1\leq x\leq u)\) is solved by QR-factorization (using SPARSEPAK), followed by a stable updating procedure for R. The main point is that the updating avoids fill-in and uses the fixed data structure of the factor R. Comparing numerical tests between the new method and the algorithm NNLS of \textit{C. L. Lawson} and \textit{R. J. Hanson} (Solving least squares problems. (1974; M.R. 51.2270)] show remarkable savings both in CPU-time and in storage requirements.