[pymvpa] HMM

[Michael Hanke]

Hey,

On Mon, Jan 17, 2011 at 05:09:15PM +0800, brice rebsamen wrote:
> scikit seems very useful and well written and well documented.
> Actually it seems to be covering quite a lot of what MVPA is
> providing. So I was wondering: how do you position yourself? More on
> the data handling part (i.e. with the Dataset class and related
> tools)?

Well, when we started there was no scikit.learn so we had to do
everything ourselves. Now there is scikit.learn and we'll use it. pymvpa
0.6 is probably going to be the last release where we ship this number
of built-in implementations. Subsequent releases will wrap scikit.learn.

PyMVPA focuses on a different audience than the scikit. Moreover the
scikit presumably wants to be much more general purpose than pymvpa,
while we are focusing on the neuroscience/imaging field.

All the time that is freed up by wrapping the scikit will be spend
drinking cocktails in the Caribbean ... ;-)

> Btw, on my dataset, the SMLR classifier is performing really well
> compared to the other classifiers provided by MVPA. I see that scikit
> provides some more, I'll compare them and eventually will post the
> results.

Cool -- looking forward to it.

Michael

-- 
Michael Hanke
http://mih.voxindeserto.de