A posteriori error estimates of Krylov subspace approximations to matrix functions

a posteriori error estimates error bounds error expansion Hermitian matrices Krylov subspace method Krylov-like approximation matrix cosine matrix exponential matrix functions matrix sine numerical examples

Mathematics Subject Classification ID

Matrix exponential and similar functions of matrices (15A16) Iterative numerical methods for linear systems (65F10) Numerical computation of matrix exponential and similar matrix functions (65F60)

Abstract: Krylov subspace methods for approximating a matrix function

f (A)

times a vector

v

are analyzed in this paper. For the Arnoldi approximation to

e^{- a u A} v

, two reliable a posteriori error estimates are derived from the new bounds and generalized error expansion we establish. One of them is similar to the residual norm of an approximate solution of the linear system, and the other one is determined critically by the first term of the error expansion of the Arnoldi approximation to

e^{- a u A} v

due to Saad. We prove that each of the two estimates is reliable to measure the true error norm, and the second one theoretically justifies an empirical claim by Saad. In the paper, by introducing certain functions

p h i_{k} (z)

defined recursively by the given function

f (z)

for certain nodes, we obtain the error expansion of the Krylov-like approximation for

f (z)

sufficiently smooth, which generalizes Saad's result on the Arnoldi approximation to

e^{- a u A} v

. Similarly, it is shown that the first term of the generalized error expansion can be used as a reliable a posteriori estimate for the Krylov-like approximation to some other matrix functions times

v

. Numerical examples are reported to demonstrate the effectiveness of the a posteriori error estimates for the Krylov-like approximations to

e^{- a u A} v

,

c o s (A) v

and

s i n (A) v

.

Review text

The paper deals with the computation of a matrix function times a vector, \(f(A)v\), by Krylov subspace methods, where \(A\) is a square matrix, possibly large and sparse or structured, with \(f\) satisfying certain smoothness conditions. Let \(\hat{f}_m\) denote the approximation to \(f(A)v\) given the Krylov-like subspace method and let \(E_m(f)=f(A)v-\hat{f}_m\) be the corresponding error. By introducing certain functions \(\phi_k(z)\) defined recursively from \(f(z)\), a series expansion to the error \(E_m(f)\) is derived and in this way an old result by Saad on the Arnoldi approximation to the matrix exponential times a vector is generalized. An important contribution of the paper is to show that the first term of such an expansion provides a reliable estimate to the error. Particular emphasis is devoted to the matrix exponential and to Hermitian matrices. Some numerical examples are presented to confirm that the proposed a posteriori error estimates are effective at least for the matrix exponential, matrix cosine and matrix sine functions. It should be noticed that theorems are not numbered, which makes some parts of the paper difficult to understand.

Recommendations

Cites work

Cited in

(15)

Describes a project that uses

Uses Software

This page was built for publication: A posteriori error estimates of Krylov subspace approximations to matrix functions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2346271)