[pymvpa] Individual measures for subjects
Yaroslav Halchenko
debian at onerussian.com
Fri Jan 31 19:47:52 UTC 2014
On Fri, 31 Jan 2014, Arman Eshaghi wrote:
> and would it be possible to get predicted probabilities instead of just
> labeled predictions? Something similar to 'predict_prob' �from SVS in
> scikit-learn?
well -- for SVM those are just 'estimates' of probabilities since SVM do
not deal with probabilities per se. You could enable/collect them from
a SVMs .ca.probabilities via enable_ca=['probabilities']. For other
classifiers, e.g. SMLR, they would really be probabilities it relies
upon to make a decision, thus accessible from 'estimates' ca.
you could pass those .ca into resultant datasets collected by
CrossValidation by specifying to your classifier
pass_attr=['ca.probabilities']
but this is not yet present in released version :-/ "we are working on
it" (i.e. on next releaase)
--
Yaroslav O. Halchenko, Ph.D.
http://neuro.debian.net http://www.pymvpa.org http://www.fail2ban.org
Senior Research Associate, Psychological and Brain Sciences Dept.
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
More information about the Pkg-ExpPsy-PyMVPA
mailing list