Talk:Minimal residual method

The printable version is no longer supported and may have rendering errors. Please update your browser bookmarks and please use the default browser print function instead.

Feedback from New Page Review process

I left the following feedback for the creator/future reviewers while reviewing this article: Great work! Ideally per MOS:PSEUDOCODE, algorithms should be presented in a language-agnostic format..

Ovinus (talk) 19:24, 28 July 2022 (UTC)[reply]

This article is written confusingly and is unintentionally misleading

I'm wondering why this article provides the Conjugate Residual algorithm instead of the MINRES algorithm. CR is equivalent in terms of the residuals and search direction to MINRES for SPD systems, but why are we providing this algorithm?

The statement in the beginning claims that GMRES tends to suffer less from loss of orthogonality. This is only true because you can literally orthogonalize at each iteration. If one wanted to store the entire subspace in MINRES, one could do so and orthogonalize at each iteration the same as GMRES. I fail to see why this is a comment added to the article.

While we're at it, why don't we state that restarted GMRES converges at a much slower rate as a function of iterations compared to MINRES. Or that full unrestarted GMRES takes orders of magnitude more work and time to converge?

It seems to me that this article is a collection of comments written by people that do not have a background in iterative methods. Poopbutt68 (talk) 17:58, 2 June 2025 (UTC)[reply]