In Fong, Saunders: LSMR: An iterative algorithm for sparse least-squares problems, section 3.3 one reads

Since only the last diagonal of $\tilde{R}_k$ and the bottom 2x2 part of $\hat{R}_k$ change each iteration, this estimate of $x_k$ can again be updated cheaply. The pseudocode, omitted here, can be derived as in section 3.1.1.

As a consequence of the omission, all LSMR codes I've found compute $\|x\|$ explicitly (and not cheaply) via a dot product.

Can anyone fill in the details here?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.