Next: Residual Vector.
Up: Stability and Accuracy Assessments
Previous: Stability and Accuracy Assessments
  Contents
  Index
We reduce it to an equivalent standard HEP.
It is done as follows. Choose a decomposition for :
|
(90) |
Then the generalized eigenvalue problem for
is equivalent to the standard HEP
for . Both share the same eigenvalues since
which also says that if is an eigenvector for the pair,
is an eigenvector for the matrix , and on the
other hand if is an eigenvector for ,
is an eigenvector for the pair.
Common choices for are:
- , the unique positive definite square root of .
In this case, . This choice is good enough for
theoretical investigations.
- is the Cholesky factor; optionally with pivoting,
i.e., is lower triangular with positive diagonal entries. This choice is preferred for
numerical computations.
- Analogously is upper triangular with positive diagonal
entries. It shares the same advantage of the second choice.
In what follows, sometimes it is more convenient to use the inner
product
induced by a positive definite matrix ,
the corresponding vector norm , and the two-vector angle function
(more precisely, angle between the subspaces spanned by two vectors)
.
In our case, or . They are defined as follows.
When , all three reduce to the usual
definitions. It is rather easy to see that
|
(91) |
With some extra work,
we can relate to the usual angle function, e.g., for ,
as follows.
|
(92) |
Subsections
Next: Residual Vector.
Up: Stability and Accuracy Assessments
Previous: Stability and Accuracy Assessments
  Contents
  Index
Susan Blackford
2000-11-20