I am not sure whether it can be randomly initialized many times and pick
the best just like in k-means? As an approximation to an integer
programming problem, I think it may subject to poor local minimal,
especially when the problem is quite complex.
Best,
Wei
On Thu, Jan 31, 2013 at 2:29 PM, Gael Varoquaux <
gael.varoqu...@normalesup.org> wrote:
> On Wed, Jan 30, 2013 at 08:37:14PM -0500, Satrajit Ghosh wrote:
> > i have been playing with the discretize option in spectral clustering
> and it
> > seems to be quite sensitive to the random state.
> > [...]
> > i haven't read the literature behind it, but i thought i would ask if
> this is
> > the expected behavior? or if there were any thoughts on the list.
>
> I would expect this behavior. In terms of clustering, these solution are
> probably mostly degenerate. On real data this is often the case, unlike
> on the simple simulation that people show in article for which there are
> easily separable clusters.
>
> G
>
>
> ------------------------------------------------------------------------------
> Everyone hates slow websites. So do we.
> Make your web apps faster with AppDynamics
> Download AppDynamics Lite for free today:
> http://p.sf.net/sfu/appdyn_d2d_jan
> _______________________________________________
> Scikit-learn-general mailing list
> Scikit-learn-general@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>
--
LI, Wei
Tsinghua/CUHK
http://kuantkid.github.com/
------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_d2d_jan
_______________________________________________
Scikit-learn-general mailing list
Scikit-learn-general@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
