I wish to solve for $x$ in

$$ (A+B)x=y $$

given square symmetric matrices $A$ and $B$. For certain reasons I have already computed the Cholesky decompositions for A and B:

$$ A = L^T L $$ $$ B = M^T M $$

Can I use these solve for $x$ more efficiently than by naively computing the Cholesky decomposition of $A+B$?

  • $\begingroup$ For the Cholesky decomposition we should also assume that $A$ and $B$ are positive definite, I thought. $\endgroup$ May 22, 2013 at 19:13
  • $\begingroup$ In general, no. One can make low-rank updates of factorization easily, but for anything different the usual answer is "you have to compute everything from scratch". Maybe you can use CG with $A$ (or $B$) as a preconditioner, but that will be numerically convenient only in some special cases. $\endgroup$ May 22, 2013 at 19:50


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.