Hi Guillermo.
What do you mean with "get the hyperplane" for other kernels?
The reason that "coef_" only works for the linear kernel is that it
has no correspondence for non-linear kernels.

And yes, you need to multiply dual_coef_ with X to get coef_ for the 
linear kernel.

Hth,
Andy

On 02/20/2013 07:12 PM, [email protected] wrote:
> Hi,
>
> I'm using sklearn for training a linear Support Vector Machine, but I cant
> match dual_coef_ with coef_. Here goes the code:
>
> X = np.array([[0, 0], [1, 1]])
> y = np.array([0, 1])
> clf = svm.SVC(kernel = "linear")
> clf.fit(X, y)
>
> this gives me a value for clf.dual_coef_ = [[ 1., -1.]] and clf.coef_ = [[
> 1.,  1.]]. How do I get from the dual_coef_ to the normal vector coef_
> (w)? I understand from
> http://scikit-learn.org/stable/modules/svm.html#svc
> that the dual_coef_ are y_i * alpha_i. In that sense I would only need to
> multiply by X and sum in each coordinate in order to get w, but this gives
> me [1, -1].
>
> This is just a test. I'm planning to use different kernels, and I
> understand coef_ only works for linear kernels.
>
> Thanks in advance,
> Guillermo.
>
>
>
>
> ------------------------------------------------------------------------------
> 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_feb
> _______________________________________________
> Scikit-learn-general mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general


------------------------------------------------------------------------------
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_feb
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

Reply via email to