In this section, we discuss the tools to assess the accuracy of computed eigenvalues and corresponding eigenvectors of the GNHEP of a regular matrix pair . We only assume the availability of residual vectors which are usually available upon the exit of a successful computation or cost marginal to compute afterwards. For the treatment of error estimation for the computed eigenvalues, eigenvectors, and deflating subspaces of dense GNHEPs, see Chapter 4 of the LAPACK Users' Guide [12].
The situation for general regular pairs is more
complicated than the standard NHEP
discussed in §7.13 (p. ),
especially when is singular, in which case the
characteristic polynomial
no longer
has degree , the dimension of the matrices and .
Even when is mathematically
nonsingular but nearly singular, problems arise when one tries to
convert it to a standard eigenvalue problem for , which then
could have huge eigenvalues and consequently cause numerical instability.
To account for all possibilities, a homogeneous representation
of an eigenvalue by a nonzero pair of numbers
has been proposed:
With this new representation of an eigenvalue, the characteristic polynomial takes the form , which does have total degree of in and . (In fact the th term in its expansion is a multiple of .)
But how do we measure the difference of two eigenvalues, given the fact of
non-uniqueness in their representations? We resort to
the chordal metric for
and
; their
distance in chordal metric is defined as