[pymvpa] LinearSVM Classification Warning
Geethmala
geethmala at gmail.com
Tue Feb 2 19:40:11 UTC 2010
code snippet:
########
dataset = dataset.selectSamples(
N.array([l in [1, 2, 3, 4, 5] for l in dataset.labels],
dtype='bool'))
cv = CrossValidatedTransferError(
TransferError(LinearCSVMC()),
CustomSplitter([([0,1],[2,3,4,5,6,7]),
([2,3],[0,1,4,5,6,7]), ([4,5],[0,1,2,3,6,7]), ([6,7],[0,1,2,3,4,5])],
attr='chunks')
)
sl = Searchlight(cv, radius=11.0)
# run searchlight on example dataset and retrieve error map
sl_map = sl(dataset)
print 'Best performing sphere error:', max(sl_map)
############
Thanks,
Geethmala
On Tue, Feb 2, 2010 at 2:33 PM, Matthias Ekman <Matthias.Ekman at nf.mpg.de>wrote:
> wild guess :):
> because you did some selection like:
>
> ds = ds.selectSamples(N.array([l in [3,4] for l in
> ds.labels],dtype='bool'))
>
> before training/testing?
> May be you could post all relevant code parts?
>
> btw, is there any reason why your data is not counterbalanced?
>
> cheers,
> Matthias
>
> Geethmala wrote:
> > Yes, but what is interesting is why is it not showing the warning for
> other
> > labels? Why only 3?
> >
> > Thanks,
> > Geethmala
> >
> > On Tue, Feb 2, 2010 at 2:19 PM, Matthias Ekman <Matthias.Ekman at nf.mpg.de
> >wrote:
> >
> >> Hi,
> >>
> >> i assume you already tracked the problem, right? ;-) ... since there a
> >> no samples (of class 3) in chunk, 0, 4, 5... exactly what
> >>
> >>>>>> Classifier LinearCSVMC(kernel_type='linear', svm_impl='C_SVC')
> >>>>>> wasn't trained to classify labels Set([3.0]) present in testing
> >>>>>> dataset. Make sure that you have not mixed order/names of the
> >>>>>> arguments anywhere
> >> says.
> >>
> >> cheers,
> >> Matthias
> >>
> >> Geethmala wrote:
> >>> Here you go,
> >>>
> >>> Dataset / float32 153 x 40656
> >>> uniq: 8 chunks 5 labels
> >>> stats: mean=-0.0255406 std=0.994264 var=0.98856 min=-6.94855
> max=6.71638
> >>>
> >>> Counts of labels in each chunk:
> >>> chunks\labels 1.0 2.0 3.0 4.0 5.0
> >>> --- --- --- --- ---
> >>> 0.0 3 6 0 9 6
> >>> 1.0 3 0 6 3 6
> >>> 2.0 3 6 3 3 3
> >>> 3.0 0 6 9 0 6
> >>> 4.0 6 0 0 3 3
> >>> 5.0 6 3 0 9 0
> >>> 6.0 6 6 3 6 0
> >>> 7.0 3 3 9 0 6
> >>>
> >>> Summary per label across chunks
> >>> label mean std min max #chunks
> >>> 1 3.75 1.98 0 6 7
> >>> 2 3.75 2.49 0 6 6
> >>> 3 3.75 3.6 0 9 5
> >>> 4 4.12 3.33 0 9 6
> >>> 5 3.75 2.49 0 6 6
> >>>
> >>> Summary per chunk across labels
> >>> chunk mean std min max #labels
> >>> 0 4.8 3.06 0 9 4
> >>> 1 3.6 2.24 0 6 4
> >>> 2 3.6 1.2 3 6 5
> >>> 3 4.2 3.6 0 9 3
> >>> 4 2.4 2.24 0 6 3
> >>> 5 3.6 3.5 0 9 3
> >>> 6 4.2 2.4 0 6 4
> >>> 7 4.2 3.06 0 9 4
> >>>
> >>>
> >>> Thanks,
> >>> Geethmala
> >>>
> >>> On Tue, Feb 2, 2010 at 2:11 PM, Matthias Ekman <
> Matthias.Ekman at nf.mpg.de
> >>> wrote:
> >>>
> >>>> Hi,
> >>>>
> >>>> could you please post:
> >>>> print ds.summary()
> >>>>
> >>>> .. just to make sure, that there are samples belonging to class 3 :)
> >>>>
> >>>>
> >>>> Matthias
> >>>>
> >>>> Geethmala wrote:
> >>>>> No, I don't have a mix of them. They are all integer values.
> >>>>>
> >>>>> Thanks,
> >>>>> Geethmala
> >>>>>
> >>>>> On Tue, Feb 2, 2010 at 1:59 PM, Yaroslav Halchenko <
> >>>> debian at onerussian.com>wrote:
> >>>>>> it means that
> >>>>>>
> >>>>>> Classifier LinearCSVMC(kernel_type='linear', svm_impl='C_SVC')
> >>>>>> wasn't trained to classify labels Set([3.0]) present in testing
> >>>>>> dataset. Make sure that you have not mixed order/names of the
> >>>>>> arguments anywhere
> >>>>>>
> >>>>>> Also test if you don't have a mix of float and int labels in your
> >>>>>> datasets (as I remember you are manually composing those).
> >>>>>>
> >>>>>>
> >>>>>> On Tue, 02 Feb 2010, Geethmala wrote:
> >>>>>>
> >>>>>>> Hi,
> >>>>>>> I get the following warning when I run LinearCSVMC.
> >>>>>>> WARNING: Classifier LinearCSVMC(kernel_type='linear',
> >>>>>> svm_impl='C_SVC')
> >>>>>>> wasn't trained to classify labels Set([3.0]) present in testing
> >>>>>>> dataset. Make sure that you have not mixed order/names of the
> >>>>>> arguments
> >>>>>>> anywhere
> >>>>>>> What does this warning mean?
> >>>>>>> Thanks,
> >>>>>>> Geethmala
> >>>>>>> _______________________________________________
> >>>>>>> Pkg-ExpPsy-PyMVPA mailing list
> >>>>>>> Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org
> >>>>>>> http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa
> >>>>>> --
> >>>>>> .-.
> >>>>>> =------------------------------ /v\ ----------------------------=
> >>>>>> Keep in touch // \\ (yoh@|www.)
> onerussian.com
> >>>>>> Yaroslav Halchenko /( )\ ICQ#: 60653192
> >>>>>> Linux User ^^-^^ [175555]
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>> _______________________________________________
> >>>>>> Pkg-ExpPsy-PyMVPA mailing list
> >>>>>> Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org
> >>>>>> http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa
> >>>>>>
> >>>>>
> >> ------------------------------------------------------------------------
> >>>>> _______________________________________________
> >>>>> Pkg-ExpPsy-PyMVPA mailing list
> >>>>> Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org
> >>>>> http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa
> >>>
> >>>
> ------------------------------------------------------------------------
> >>>
> >>> _______________________________________________
> >>> Pkg-ExpPsy-PyMVPA mailing list
> >>> Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org
> >>> http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa
> >
> >
> > ------------------------------------------------------------------------
> >
> > _______________________________________________
> > Pkg-ExpPsy-PyMVPA mailing list
> > Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org
> > http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-exppsy-pymvpa/attachments/20100202/08f2f422/attachment.htm>
More information about the Pkg-ExpPsy-PyMVPA
mailing list