The original question about eigenvectors was prompted by the
following.  Eigenvectors over RDF and CDF are computed by SciPy and
returned with their eigenvalues as if each eigenvalue has algebraic
and geometric multiplicity 1.  Decisions about "equal" eigenvalues are
left to the caller.  SciPy normalizes to length 1.

So far there are two possible output eigenvectors in the RDF case,
negatives of each other.  I have not yet determined if SciPy intends
to always return one of the two.  But it would appear this *could* be
the case, since new doctests for 4.4.alpha0 were only failing on one
machine (sextus on Skynet).

I have fixed the "test" part of the doctests, at the expense of the
"doc" part.  Information or advice on the following would be helpful
for a decision about actually fiddling with Sage code:

(a) Does anybody know if SciPy intends to guarantee a canonical choice
between the two length 1 multiples?

(b) If not, is there a good approach to deciding ourselves?

(c) Is the situation worse over CDF given the multitude of norm 1
scalars to act as multiples?

See 
http://groups.google.com/group/sage-release/browse_thread/thread/9136569bd1c67f6
for background and good arguments you don't want to repeat.

IANANAE, but did take a course once,
Rob

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

Reply via email to