sounds good to me. You could even use a Pipeline. A more philosophical is
why would you want to use Isomap as it's blind to the labels (it's
unsupervised). I would bench with a svm with gaussian kernel and a label
propagation.
in theory you should indeed not use the full data in isomap.
Alex
On Thu, May 24, 2012 at 1:00 PM, Buddy Rich <[email protected]> wrote:
> Hi, I am using scikit 0.11 for classification task.
>
> I would like to test svm after Isomap reduction on my dataset, previously
> splitted into a training/test set with StratifiedKFold.
>
> Is it correct this procedure?
>
> isomap = manifold.Isomap(n_neighbors, n_components).fit(x_train)
> iso_train = isomap.transform(x_train)
> iso_test = isomap.transform(x_test)
>
> # SVM code
> .
> .
> .
>
> In other words, is the test set correctly mapped on the reduced space
> induced with respect to the training set?
>
> Should I use isomap on the whole dataset and then split it into training
> and test set?
> Thank you in advance!
>
>
> ------------------------------------------------------------------------------
> Live Security Virtual Conference
> Exclusive live event will cover all the ways today's security and
> threat landscape has changed and how IT managers can respond. Discussions
> will include endpoint security, mobile security and the latest in malware
> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
> _______________________________________________
> Scikit-learn-general mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>
>
--
Alexandre Gramfort, PhD
[email protected]
INRIA Parietal Project Team @ NeuroSpin CEA Saclay
Bat. 145, PC 156
91191 Gif-sur-Yvette, France
http://www-sop.inria.fr/members/Alexandre.Gramfort/
------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
