[pymvpa] regression vs. classification

[Michael Hanke]

On Mon, May 26, 2008 at 03:55:31PM +0200, Emanuele Olivetti wrote:
> Dear all,
> 
> I've a simple question in mind since a while. Why the base
> class for classification and regression is "Classifier"?
> Why not a simply "LearningModel" which could then be
> subclassed as Classifier and RegressionModel (or similar
> names). Even more: why not adopting the full taxonomy,
> taking into account supervised and unsupervised methods?
> 
> Something like (in my humble my ascii art):
> 
> LearningModel
> |            \
> |             \
> Supervised     Unsupervised
> |         \
> |          RegressionModel
> |
> Classifier
> 
> Unsupervised could be the place for clustering and PCA/ICA
> that could be included in future (is it?).
Yeah, we also had something similar in mind (our intermediate baseclass
name was ML for machine learning). The reason why it is not
there is that nobody ever really touched it. Anyway, as it extends the
hierarchy on the top, the user-visible interface should be unaffected,
thus we can add this at anytime.

Feel free to pave this road ;-)

(but maybe let's get this release done without it)

Cheers,

Michael

-- 
GPG key:  1024D/3144BE0F Michael Hanke
http://apsy.gse.uni-magdeburg.de/hanke
ICQ: 48230050