[pymvpa] Help needed untangling searchlight accuracy bias

[Mike E. Klein]

Hi again,

Yes, it meaned 3 across one dimension with the second dimension
collapsed together.

I've pasted below the Errors from PyMVPA if I try to run the script
without run averaging. Strangely, this error disappears if I use
LinearNuSVMC() in place of LinearCSVMC(). I'm very much a newbie in
terms of programming and advanced stats, so I have no idea how these
two differ.

Best and thanks again,
Mike


[SLC] DBG:     Starting computing block for 46174 elements
Traceback (most recent call last):
  File "searchlight.py", line 48, in <module>
    s1_map = sl(dataset)
  File "/Library/Frameworks/EPD64.framework/Versions/7.1/lib/python2.7/site-packages/mvpa/base/learner.py",
line 228, in __call__
    return super(Learner, self).__call__(ds)
  File "/Library/Frameworks/EPD64.framework/Versions/7.1/lib/python2.7/site-packages/mvpa/base/node.py",
line 76, in __call__
    result = self._call(ds)
  File "/Library/Frameworks/EPD64.framework/Versions/7.1/lib/python2.7/site-packages/mvpa/measures/searchlight.py",
line 125, in _call
    results, roi_sizes = self._sl_call(dataset, roi_ids, nproc)
  File "/Library/Frameworks/EPD64.framework/Versions/7.1/lib/python2.7/site-packages/mvpa/measures/searchlight.py",
line 241, in _sl_call
    self._proc_block(roi_ids, dataset, self.__datameasure)
  File "/Library/Frameworks/EPD64.framework/Versions/7.1/lib/python2.7/site-packages/mvpa/measures/searchlight.py",
line 293, in _proc_block
    results.append(measure(roi))
  File "/Library/Frameworks/EPD64.framework/Versions/7.1/lib/python2.7/site-packages/mvpa/base/learner.py",
line 228, in __call__
    return super(Learner, self).__call__(ds)
  File "/Library/Frameworks/EPD64.framework/Versions/7.1/lib/python2.7/site-packages/mvpa/base/node.py",
line 76, in __call__
    result = self._call(ds)
  File "/Library/Frameworks/EPD64.framework/Versions/7.1/lib/python2.7/site-packages/mvpa/measures/base.py",
line 469, in _call
    return super(CrossValidation, self)._call(ds)
  File "/Library/Frameworks/EPD64.framework/Versions/7.1/lib/python2.7/site-packages/mvpa/measures/base.py",
line 302, in _call
    result = node(sds)
  File "/Library/Frameworks/EPD64.framework/Versions/7.1/lib/python2.7/site-packages/mvpa/base/learner.py",
line 228, in __call__
    return super(Learner, self).__call__(ds)
  File "/Library/Frameworks/EPD64.framework/Versions/7.1/lib/python2.7/site-packages/mvpa/base/node.py",
line 76, in __call__
    result = self._call(ds)
  File "/Library/Frameworks/EPD64.framework/Versions/7.1/lib/python2.7/site-packages/mvpa/measures/base.py",
line 557, in _call
    measure.train(dstrain)
  File "/Library/Frameworks/EPD64.framework/Versions/7.1/lib/python2.7/site-packages/mvpa/base/learner.py",
line 121, in train
    result = self._train(ds)
  File "/Library/Frameworks/EPD64.framework/Versions/7.1/lib/python2.7/site-packages/mvpa/clfs/libsvmc/svm.py",
line 186, in _train
    raise FailedToTrainError(str(e))
mvpa.base.learner.FailedToTrainError: in method 'svm_parameter_C_set',
argument 2 of type 'double'



On Thu, Oct 20, 2011 at 3:11 PM, Yaroslav Halchenko
<debian at onerussian.com> wrote:
>
> On Thu, 20 Oct 2011, Mike E. Klein wrote:
>> "little" runs or 9 "big" runs. I've initially gone with the 27-run
>> route (i.e. created an attribute file with 27 chunks), mainly for the
>> somewhat strange reason that (a) I couldn't get the LinearCSVMC
>> classifier to work without using pymvpa's "mean_group_sample" first
>> and
>
> what was meaned actually here since in each of those 27 chunks
> you had only 1 sample per each of the 9 sounds, right?  or it meaned 3
> across one dimension? thus collapsing 3 samples together?
>
> and how SVM complained otherwise? ;)
>
> --
> =------------------------------------------------------------------=
> Keep in touch                                     www.onerussian.com
> Yaroslav Halchenko                 www.ohloh.net/accounts/yarikoptic
>
> _______________________________________________
> Pkg-ExpPsy-PyMVPA mailing list
> Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org
> http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-exppsy-pymvpa
>