[pymvpa] MulticlassClassifier Queries

[Geethmala]

Hi,
I have a couple of queries about multiclass classification.
I have 3 classes and I want to classify them using SVM with Searchlight.

Code snippet:
############################
dataset = dataset.selectSamples(
                N.array([l in [1, 2, 3] for l in dataset.labels],
                        dtype='bool'))

mlclf = MulticlassClassifier(clf=LinearCSVMC(),bclf_type="1-vs-1")

cv = CrossValidatedTransferError(
                        TransferError(mlclf),
                        CustomSplitter([([0,1],[2,3,4,5,6,7,8,9,10,11])],
                        attr='chunks')
                        )

sl = Searchlight(cv, radius=3)

# run searchlight on example dataset and retrieve error map
sl_map1 = sl(dataset)

################################
Queries:

1) I am getting the following warning. What does it mean?

WARNING: Boosted classifier
MulticlassClassifier(clfs=[BinaryClassifier(clf=LinearCSVMC(kernel_type='linear',
svm_impl='C_SVC'), poslabels=[1.0], neglabels=[2.0]),...],
combiner=MaximalVote(), bclf_type='1-vs-1',
clf=LinearCSVMC(kernel_type='linear', svm_impl='C_SVC')) has 'values' state
enabled, but combiner has it active, thus no values could be provided
directly, access .clfs


2) If I don't use Multiclass, there is no problem. Should I use Multiclass
at all?

Thanks,
Geethmala
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-exppsy-pymvpa/attachments/20100126/ecd1c7f2/attachment.htm>