[pymvpa] Is it possible to predict a linear "categories"?

[Michael Hanke]

Hi,

On Tue, Mar 25, 2014 at 12:58:27PM -0400, Jason Ozubko wrote:
> The classifiers in PyMVPA all seem to be targeted at classifying patterns
> into nominal categories.  Is there anything that can be done when you have
> a linear "category"?

Two ideas:

1. Use a regression. PyMVPA handles regressions and classifiers in very
   similar ways, hence in most cases you can just replace a classifier
   instance with a regression. Here is an example on how to use any
   regression algorithm implemented in scikit-learn within PyMVPA:

   http://www.pymvpa.org/examples/skl_regression_demo.html

2. Keep doing classification, but use a custom error function. So if the
   "distance" from the target value within you linear "category" is
   meaningful, you can assign an error function like this one:

   def eucd(targets, predictions):
       return targets - predictions

Does that make sense in your context?

Michael

-- 
J.-Prof. Dr. Michael Hanke
Psychoinformatik Labor,    Institut für  Psychologie II
Otto-von-Guericke-Universität Magdeburg,  Universitätsplatz 2, Geb.24
Tel.: +49(0)391-67-18481 Fax: +49(0)391-67-11947  GPG: 4096R/7FFB9E9B