[pymvpa] Announcement: 0.4.4 release (git/sources/Debian/Ubuntu/Windows/Fedora/SuSe/RedHat/CentOS)

[Matthias Ekman]

Hi,

Yaroslav Halchenko wrote:
> On Tue, 02 Feb 2010, Matthias Ekman wrote:
> 
>> Is this only relevant for automatic scaling of value according to the
>> norm of the data (C=-1)?
> yes -- for any C < 0, which triggers automagic scaling
> 
>> is this problem 'always' leading to never finished SVM training, or do
>> you suggest to re-run previous (finished) analysis?
> 
> I guess if they finished -- you are fine ;)

Good :-)

> it was only relevant to cases when SVM was asked to train on data where
> all values are 0's... and even it managed to train/predict (not sure if
> that could happen -- didn't wait long enough) -- it would provided as
> random result as you could from data where it is all 0s ;)

in one case I waited 3 weeks (during my holiday) so I guess it could not
happen - but who knows. To figure this out I would need to have much
longer holidays ;-)