The condition number of equivalence transformations that block diagonalize matrix pencils
How ill-conditioned must a matrix S be if its columns are constrained to span certain subspaces? We answer this question in order to find nearly best conditioned matrices SR and SL that block diagonalize a given matrix pencil T=A+λB, i.e. SL−1TSR=Θ is bloc diagonal. We show that the best conditioned SR has a condition number approximately equal to the cosecant of the smallest angle between right subspaces belonging to different diagonal blocks of Θ. Thus, the more nearly the right subspaces overlap the more ill-conditioned SR must be. The same is true of SL and the left subspaces. For the standard eigenproblem (T=A−λI), SL = SR and the cosecant of the angle between subspaces turns out equal to an earlier estimate of the smallest condition number, namely the norm of the projection matrix associated with one of the subspaces. We apply this result to bound the error in an algorithm to compute analytic functions of matrices, for instance exp(T).
Unable to display preview. Download preview PDF.
- J. DEMMEL, The Condition Number of Similarities that Diagonalize Matrices, Electronics Research Laboratory Memorandum, University of California, Berkeley, 1982Google Scholar
- F. R. GANTMACHER, The Theory of Matrices, trans. K. A. Hirsch, Chelsea, 1959, vol. 2Google Scholar
- E. ISAACSON, H. B. KELLER, Analysis of Numerical Methods, Wiley, 1966Google Scholar
- W. KAHAN, Conserving Confluence Curbs Ill-Condition, Technical Report 6, Computer Science Dept., University of California, Berkeley, August 4, 1972Google Scholar
- T. KATO, Perturbation Theory for Linear Operators, Springer-Verlag, 1966Google Scholar
- J. G. SUN, The Perturbation Bounds for Eigenspaces of Definite Matrix Pairs, to appearGoogle Scholar
- J. H. WILKINSON, Rounding Errors in Algebraic Processes, Prentice Hall, 1963Google Scholar
- J. H. WILKINSON, The Algebraic Eigenvalue Problem, Oxford University Press, 1965Google Scholar