[pymvpa] Maximal Vote Issue

[Yaroslav Halchenko]

Sorry about the delay:  indeed was a bug :-/  fixed in GIT:

$> git show
commit 1857ad78d0d1fe16237fc4003339dea8d0c54d86
Author: Yaroslav Halchenko <debian at onerussian.com>
Date:   Fri Jul 8 15:36:56 2011 -0400

    BF: MaximalVote should pass **kwargs along (thanks Andrei Rusu for the report)


would you be comfortable trying it from GIT?

On Thu, 30 Jun 2011, Andrei Rusu wrote:

>    Hello Everyone,
>    I have a question about MaximalVote. I'm getting an error on:
>    voting=MaximalVote(enable_ca=['estimates'])
>    TypeError: __init__() got an unexpected keyword argument 'enable_ca'
>    The documentation page
>    : http://www.pymvpa.org/generated/mvpa.clfs.meta.MaximalVote.html#mvpa.
>    clfs.meta.MaximalVote
>    shows "enable_ca" as a valid argument.
>    I am using python-pymvpa-snapshot from neuro-debian in ubuntu 10.10,
>    64bit.
>    I am using this object in the following context:
>    # inner_part is a custom partitioner which provides random subsets of
>    the training set
>    voting = MaximalVote()

>    # voting=MaximalVote(enable_ca=['estimates'])    # desired but produces
>    error
>         comb_clf = SplitClassifier(LinearCSVMC(C=1),
>    partitioner=inner_part,
>    splitter=splitter,
>    combiner=voting)
>      spartitions = partitions[:, indices]
>       tm = TransferMeasure(comb_clf, splitter,
>    postproc=BinaryFxNode(mean_mismatch_error, space='targets'),
>    enable_ca=['stats','raw_results'])
>      tm(spartitions)
>    I would like to see the "estimates: Estimates keep counts across
>    classifiers for each label/sample"
>    As you might have guessed, this intends to be example-wise bagging.
>    Any suggestions on how to circumvent this issue or make sure the voting
>    procedure operates as expected?
-- 
=------------------------------------------------------------------=
Keep in touch                                     www.onerussian.com
Yaroslav Halchenko                 www.ohloh.net/accounts/yarikoptic