Dear pyMVPA experts,
<br />
<br /> I just ran the RFE by copying and pasting the code in the manual, but nothing pops up after it was run.
<br /> Is there any way to see the transfer error rate?
<br /> Below is the full code that I modified the one in the example.
<br />
<br /> Thanks,
<br />
<br />
<br />--------------------------------------------------------
<br /> from mvpa.suite import *
<br />
<br />#load nifti dataset
<br />attr = SampleAttributes('label_25jun08MG.txt')
<br />dataset = NiftiDataset(samples='wrbold_25jun08MG.nii.gz', labels=attr.labels,
<br /> chunks=attr.chunks, mask='mask.nii.gz')
<br />
<br />#linearly detrend the data chunk-wise
<br />detrend(dataset, perchunk=True, model="linear", polyord = 1)
<br />
<br /># z-score the data relative to the baseline trials in each chunk:
<br />zscore(dataset, perchunk=True, baselinelabels = [0], pervoxel = True, targetdtype = 'float64')
<br />
<br />
<br /># define and run classifier
<br />
<br />rfesvm_split = SplitClassifier(LinearCSVMC())
<br />clf = \
<br />FeatureSelectionClassifier(
<br />clf = LinearCSVMC(),
<br /># on features selected via RFE
<br />feature_selection = RFE(
<br /># based on sensitivity of a clf which does splitting internally
<br />sensitivity_analyzer=rfesvm_split.getSensitivityAnalyzer(),
<br />transfer_error=ConfusionBasedError(
<br />rfesvm_split,
<br />confusion_state="confusion"),
<br /># and whose internal error we use
<br />feature_selector=FractionTailSelector(
<br />0.2, mode= 'discard', tail='lower'),
<br /># remove 20% of features at each step
<br />update_sensitivity=True),
<br /># update sensitivity at each step
<br />descr= 'LinSVM+RFE(splits_avg)')
<br />-----------------------------------------------------------