[pymvpa] request for review OUPS WRONG PATCH

[Yaroslav Halchenko]

also why not just prep a new metric, which would take care about taking
a single time-point... ie originally you have 5D dataset:
samples x FIR-timepoints x 3D volume
(instead of current default samples x 3D volume), which is similar in
nature to what we get now from BoxcarMapper when we take multiple
volumes after some onsets.

Later on those datasets get mapped into 2D but if original dataset's
mapper has a proper metric to select neighbors it can spit out only
neighbors within the same time-offset.

This way there is no need in modification of searchlight at all.

or am I missing the point?


On Fri, 24 Apr 2009, Michael Hanke wrote:

> Hi,

> On Fri, Apr 24, 2009 at 11:14:41AM +0200, Valentin Haenel wrote:
> > er, i think i attached the wrong patch. fracking git commit --amend :-)

> > here, use this one instead.

> It looks sane to me. I see the advantages of running multiple datasets
> for the same sphere setup at once. On the other hand this adds some
> magic that just applies to the searchlight measure -- not sure whether
> we want that kind of variability among the different measures. I'm not
> necessarily opposed to the change...

> Did you do some comparison what you actually gain from moving the
> dataset loop inside of searchlight in comparison to a simple outer loop
> (in terms of speed)? I guess it could be quite a bit...

> Oppinions?


> Michael
-- 
Yaroslav Halchenko
Research Assistant, Psychology Department, Rutgers-Newark
Student  Ph.D. @ CS Dept. NJIT
Office: (973) 353-1412 | FWD: 82823 | Fax: (973) 353-1171
        101 Warren Str, Smith Hall, Rm 4-105, Newark NJ 07102
WWW:     http://www.linkedin.com/in/yarik