Stability analysis of the method of seminormal equations for linear least squares problems

摘要: Abstract Consider the linear least squares problem minx‖Ax−b‖2. When A is large and sparse, then often only R-factor in QR factorization of known.The solution x can be computed from seminormal equations RTRx = ATb. For this method error shown to same order as for normal equations. We show that by adding a correction step using single precision we get which under mild conditions accurate method. The application updating sparse when appending column discussed.