- NUMERISCHE MATHEMATIK Journal
- October 2009
Digital Object Identifier (DOI)
International Standard Serial Number (ISSN)
Electronic International Standard Serial Number (EISSN)
- We prove that the Jacobi algorithm applied implicitly on a decompositionA = XDXT of the symmetric matrix A, where D is diagonal, and X is well conditioned, computes all eigenvalues of A to high relative accuracy. The relative error in every eigenvalue is bounded by O(Ekappa(X)), where E is the machine precision and kappa(X) ≡ ||X||2 · ||X−1||2 is the spectral condition number of X. The eigenvectors are also computed accurately in the appropriate sense.We believe that this is the first algorithm to compute accurate eigenvalues of symmetric (indefinite)matrices that respects and preserves the symmetry of the problem and uses only orthogonal transformations.