[pymvpa] Can't access an enabled classifier attribute.

[Yaroslav Halchenko]

On Tue, 27 Nov 2012, Roberto Guidotti wrote:
>    Well the snippet probabilities = clf.ca.probabilities is only an example
>    that I've build to reveal the exception (I didn't think is a version
>    problem, I had 2.0.1 mvpa)!!�

2.0.1 ... hm... I do not remember how life was back then ;)

>    BTW I was quite sure that the problem relied on the cross validation :)!
>    �I would like to gather all the predictions/probabilities for each cv
>    fold.

then provide callback to the CrossValidation and there you can collect
anything you like from your classifier upon completion of every split.
e.g. look at
https://github.com/PyMVPA/PyMVPA/blob/HEAD/mvpa2/tests/test_rfe.py#L403


>    This why using the ca 'training_stats' during the definition of the
>    classifier, I had the probabilities of the samples during training?

I guess ;)

>    So I can't have access to every cv fold probabilities/predictions using
>    CrossValidation object? 

you can with above mentioned callback

> Or I need a manual CrossValidation (and sometimes
>    a Feature Selection)?

if you decide to gather probabilities on both training and testing
samples, then you might need indeed manual loop with explicit .train
.predict so you get needed probabilities at the right moment.  Otherwise
-- callback should be sufficient
-- 
Yaroslav O. Halchenko
Postdoctoral Fellow,   Department of Psychological and Brain Sciences
Dartmouth College, 419 Moore Hall, Hinman Box 6207, Hanover, NH 03755
Phone: +1 (603) 646-9834                       Fax: +1 (603) 646-1419
WWW:   http://www.linkedin.com/in/yarik