Next: Error Bounds for the Up: Error Bounds for the Previous: Error Bounds for the   Contents   Index

## Further Details: Error Bounds for the Generalized Symmetric Definite Eigenproblem

The error analysis of the driver routines for the generalized symmetric definite eigenproblem goes as follows. In all cases is the absolute gap between and the nearest other eigenvalue.

1.
. The computed eigenvalues can differ from true eigenvalues by at most about

The angular difference between the computed eigenvector and a true eigenvector zi is

2.
or . The computed eigenvalues can differ from true eigenvalues by at most about

The angular difference between the computed eigenvector and a true eigenvector zi is

The code fragments above replace p(n) by 1, and makes sure neither RCONDB nor RCONDZ is so small as to cause overflow when used as divisors in the expressions for error bounds.

These error bounds are large when B is ill-conditioned with respect to inversion ( is large). It is often the case that the eigenvalues and eigenvectors are much better conditioned than indicated here. We mention three ways to get tighter bounds. The first way is effective when the diagonal entries of B differ widely in magnitude4.1:

1.
. Let be a diagonal matrix. Then replace B by DBD and A by DAD in the above bounds.
2.
or . Let be a diagonal matrix. Then replace B by DBD and A by D-1AD-1 in the above bounds.

The second way to get tighter bounds does not actually supply guaranteed bounds, but its estimates are often better in practice. It is not guaranteed because it assumes the algorithm is backward stable, which is not necessarily true when B is ill-conditioned. It estimates the chordal distance between a true eigenvalue and a computed eigenvalue :

To interpret this measure we write and . Then . In other words, if represents the one-dimensional subspace consisting of the line through the origin with slope , and represents the analogous subspace , then is the sine of the acute angle between these subspaces. Thus is bounded by one, and is small when both arguments are large4.2. It applies only to the first problem, :

Suppose a computed eigenvalue of is the exact eigenvalue of a perturbed problem . Let xi be the unit eigenvector (|xi|2=1) for the exact eigenvalue . Then if |E| is small compared to |A|, and if |F| is small compared to |B|, we have

Thus is a condition number for eigenvalue .

The third way applies only to the first problem , and only when A is positive definite. We use a different algorithm:

1.
Compute the Cholesky factorization of A = UTA UA, using xPOTRF.
2.
Compute the Cholesky factorization of B = UTB UB, using xPOTRF.
3.
Compute the generalized singular value decomposition of the pair UA, UB using xTGSJA. The squares of the generalized singular values are the desired eigenvalues.
See sections 2.3.5.3 and  2.4.9 for a discussion of the generalized singular value decomposition, and section 4.12 for a discussion of the relevant error bound. This approach can give a tighter error bound than the above bounds when B is ill conditioned but A+B is well-conditioned.

Other yet more refined algorithms and error bounds are discussed in [14,95,103].

Next: Error Bounds for the Up: Error Bounds for the Previous: Error Bounds for the   Contents   Index
Susan Blackford
1999-10-01