[Pkg-exppsy-pymvpa-dev] pymvpa and spatially related features

[Satrajit Ghosh]

hi,

i have a set of features that share geometrical/spatial properties. is there
a way i can tell pymvpa that these features are tied together?

for example, i might get curvature and cortical thickness from a single
vertex as features, but currently i don't encode the fact that these two
features are related. the ds is just a nsubj x nfeat matrix.

and related to this question, are there specific classifiers that can take
advantage of this relation?

and to extend it beyond simple grouping based on location, can i include
neighborhood relations on surfaces  or to generalize (connectivity on
graphs).

cheers,

satra
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-exppsy-pymvpa-dev/attachments/20100506/5ae26367/attachment.htm>