While the full DRTLS method converges within 5 and the GKS-DRTLS method with preconditioner M = [[?
A few words concerning the zero-finders for the full DRTLS method and the GKSDRTLS Algorithm 4.
2 show the number of outer and inner iterations as well as the iterations required for the zero-finder within one inner iteration for the full DRTLS method and the generalized Krylov subspace DRTLS method with and without preconditioner.
The first outer iteration of the GKS-DRTLS method is treated separately since it corresponds to solving the projected DRTLS with the starting basis [V.
2 the number of iterations required for the full DRTLS algorithm is compared to the GKS-DRTLS method when the preconditioner M = [[?
The constraint condition within the DRTLS methods is fulfilled with almost machine precision while for the used implementations of the RTLS methods this quantity varies with the underlying problem.
The DRTLS method outperforms the other three algorithms, i.
true][parallel], whereas the DRTLS solution has a norm of [?
55%, while the DRTLS restoration used in the second image has required a search space of dim(V) = 42 with a relative error of [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].