Checking for zeros on diagonal when doing cholesky factorization

[Boyce Byerly <BByerly@newcourse.com>]

There are some clear conditions one needs to verify before doing a cholesky
factorization
on a matrix.  For example, the matrix tends to be symmetric and positive
definite.

I think it should also be the case that there are no zeros on the diagonal
of the input
matrix, but there's not a check for this in the code.  If you input a matrix
with a zero
on the diagonal, you get a solution back out with "?" in the matrix (I
assume this is
what the .print method throws out when a Java NaN (Not-a-number) appears
under
div-by-zero and such circumstances.)

Should there be such a check?  I'm not enough of a mathematician to think
through
the theoretical basis for this, and would be interested to hear back from
those who
are.  In the meantime, I've put an exception throwing check in the wrapper I
have
around my matrix classes and am making do with that.



_______________________________________

W. Boyce Byerly, Ph.D.

Vice President of Research and Development

New Course, Inc., (919) 967-0322

New Course Analytics, (919) 967-0322 * 

Makers of ProCourse www.procourse.com

ProCourse is the industry's first true ROI tool. ProCourse can isolate the
effects of business interventions such as training from the total business
impact. ProCourse not only identifies which interventions are working, but
also optimizes the impact of those interventions, adding significant revenue
to your bottom line. You can view a Flash demo of ProCourse here:
http://www.procourse.com/demo/

This message (including any attachments) contains confidential information
intended for a specific individual and purpose, and is protected by law. If
you are not the intended recipient, you should delete this message and are
hereby notified that any disclosure, copying, or distribution of this
message, or the taking of any action based on it, is strictly prohibited.

..