def testBinds(self):
ds = normalFeatureDataset()
ds_data = ds.samples.copy()
ds_chunks = ds.chunks.copy()
self.failUnless(N.all(ds.samples == ds_data)) # sanity check
funcs = ['zscore', 'coarsenChunks']
if externals.exists('scipy'):
funcs.append('detrend')
for f in funcs:
eval('ds.%s()' % f)
self.failUnless(N.any(ds.samples != ds_data) or
N.any(ds.chunks != ds_chunks),
msg="We should have modified original dataset with %s" % f)
ds.samples = ds_data.copy()
ds.chunks = ds_chunks.copy()
# and some which should just return results
for f in ['aggregateFeatures', 'removeInvariantFeatures',
'getSamplesPerChunkLabel']:
res = eval('ds.%s()' % f)
self.failUnless(res is not None,
msg='We should have got result from function %s' % f)
self.failUnless(N.all(ds.samples == ds_data),
msg="Function %s should have not modified original dataset" % f)
def testTScoredAnalyzer(self, svm):
self.dataset = normalFeatureDataset(perlabel=100,
nlabels=2,
nchunks=20,
nonbogus_features=[0,1],
nfeatures=4,
snr=10)
svm_weigths = svm.getSensitivityAnalyzer()
sana = TScoredFeaturewiseMeasure(
svm_weigths,
NFoldSplitter(cvtype=1),
enable_states=['maps'])
t = sana(self.dataset)
# correct size?
self.failUnlessEqual(t.shape, (4,))
# check reasonable sensitivities
t = Absolute(t)
self.failUnless(N.mean(t[:2]) > N.mean(t[2:]))
# check whether SplitSensitivityAnalyzer 'maps' state is accessible
self.failUnless(sana.states.isKnown('maps'))
self.failUnless(N.array(sana.maps).shape == (20,4))
def testLARSSensitivities(self):
data = normalFeatureDataset(perlabel=10, nlabels=2, nfeatures=4)
# use LARS on binary problem
clf = LARS()
clf.train(data)
# now ask for the sensitivities WITHOUT having to pass the dataset
# again
sens = clf.getSensitivityAnalyzer(force_training=False)()
self.failUnless(sens.shape == (data.nfeatures,))
def testSMLRSensitivities(self):
data = normalFeatureDataset(perlabel=10, nlabels=2, nfeatures=4)
# use SMLR on binary problem, but not fitting all weights
clf = SMLR(fit_all_weights=False)
clf.train(data)
# now ask for the sensitivities WITHOUT having to pass the dataset
# again
sens = clf.getSensitivityAnalyzer(force_training=False)()
self.failUnless(sens.shape == (data.nfeatures,))
