[pymvpa] how to serialize/pickle/save a trained SVM?

[Emanuele Olivetti]

Dear All,

I'm playing with SVMs (libSVM backend) and after a very long training
I'd like to exploit the outcome for a while. Is there a way to save to
file (i.e., pickle/serialization) the trained SVM?

Currently pickle does not work since it relies on a PySwigObject
which is... "unpicklable" :-)

As far as I remember libsvm alone can dump the training results
(support vectors etc.) to file for later use. Is this wrapped in some way?

Any idea?

Thanks a lot,

Emanuele