[SciPy-dev] Trouble with optimize.fmin_ncg

Nathan Bell wnbell@gmail....
Wed Jul 25 16:15:50 CDT 2007


On 7/25/07, Alan G Isaac <aisaac@american.edu> wrote:
> There is another reason not to do this.
> This was not really a scipy.optimize bug.
> I suppose the need is rather for a test in scipy.sparse,
> so Nathan can decide what if anything to do about this.

The specific error in sparse can (and should) be tested by something
simpler than the script Nil's used to demonstrate the problem.  I'll
add the appropriate unittests for this bug to sparse later today.

I'll defer to the optimization people regarding whether to include it
in their battery of tests.  Personally, I find that testing against a
few small, but non-trivial real-world data sets is a good thing.  The
matrix in question is rather small (9KB) so it's not unreasonable to
include it in the source tree.

-- 
Nathan Bell wnbell@gmail.com


More information about the Scipy-dev mailing list