On Fri, Mar 30, 2012 at 6:20 PM, Gael Varoquaux <
[email protected]> wrote:
> One remark on the algorithmic side, rather than on the project management
> side: it seemed to me when I read the Yuan paper than the strategy to
> choose the order in which the variables where updated could make a fair
> difference on convergence speed. Maybe I read the paper too fast (I
> really scanned through), but I am not sure that I see this appearing in
> your list.
>
>
In the Yuan paper, they just permute the coordinates before each epoch and
they say that it's better than cycling through the coordinates with
permuting. They don't greedily select the coordinates.
Mathieu
------------------------------------------------------------------------------
This SF email is sponsosred by:
Try Windows Azure free for 90 days Click Here
http://p.sf.net/sfu/sfd2d-msazure
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
