Source Code for Module mvpa.tests.test_datasetfx

  1  # emacs: -*- mode: python; py-indent-offset: 4; indent-tabs-mode: nil -*- 
  2  # vi: set ft=python sts=4 ts=4 sw=4 et: 
  3  ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ## 
  4  # 
  5  #   See COPYING file distributed along with the PyMVPA package for the 
  6  #   copyright and license terms. 
  7  # 
  8  ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ### ## 
  9  """Unit tests for PyMVPA miscelaneouse functions operating on datasets""" 
 10   
 11  import unittest 
 12   
 13  import numpy as N 
 14   
 15  from mvpa.base import externals 
 16  from mvpa.datasets import Dataset 
 17  from mvpa.datasets.miscfx import removeInvariantFeatures, coarsenChunks, \ 
 18       SequenceStats 
 19   
 20  from mvpa.misc.data_generators import normalFeatureDataset 
 21   
 22 -class MiscDatasetFxTests(unittest.TestCase): 
 23   
 24 -    def testInvarFeaturesRemoval(self): 
 25          r = N.random.normal(size=(3,1)) 
 26          ds = Dataset(samples=N.hstack((N.zeros((3,2)), r)), 
 27                       labels=1) 
 28   
 29          self.failUnless(ds.nfeatures == 3) 
 30   
 31          dsc = removeInvariantFeatures(ds) 
 32   
 33          self.failUnless(dsc.nfeatures == 1) 
 34          self.failUnless((dsc.samples == r).all()) 
 35   
 36   
 37 -    def testCoarsenChunks(self): 
 38          """Just basic testing for now""" 
 39          chunks = [1,1,2,2,3,3,4,4] 
 40          ds = Dataset(samples=N.arange(len(chunks)).reshape( 
 41              (len(chunks),1)), labels=[1]*8, chunks=chunks) 
 42          coarsenChunks(ds, nchunks=2) 
 43          chunks1 = coarsenChunks(chunks, nchunks=2) 
 44          self.failUnless((chunks1 == ds.chunks).all()) 
 45          self.failUnless((chunks1 == N.asarray([0,0,0,0,1,1,1,1])).all()) 
 46   
 47          ds2 = Dataset(samples=N.arange(len(chunks)).reshape( 
 48              (len(chunks),1)), labels=[1]*8) 
 49          coarsenChunks(ds2, nchunks=2) 
 50          self.failUnless((chunks1 == ds.chunks).all()) 
 51   
 52 -    def testBinds(self): 
 53          ds = normalFeatureDataset() 
 54          ds_data = ds.samples.copy() 
 55          ds_chunks = ds.chunks.copy() 
 56          self.failUnless(N.all(ds.samples == ds_data)) # sanity check 
 57   
 58          funcs = ['zscore', 'coarsenChunks'] 
 59          if externals.exists('scipy'): 
 60              funcs.append('detrend') 
 61   
 62          for f in funcs: 
 63              eval('ds.%s()' % f) 
 64              self.failUnless(N.any(ds.samples != ds_data) or 
 65                              N.any(ds.chunks != ds_chunks), 
 66                  msg="We should have modified original dataset with %s" % f) 
 67              ds.samples = ds_data.copy() 
 68              ds.chunks = ds_chunks.copy() 
 69   
 70          # and some which should just return results 
 71          for f in ['aggregateFeatures', 'removeInvariantFeatures', 
 72                    'getSamplesPerChunkLabel']: 
 73              res = eval('ds.%s()' % f) 
 74              self.failUnless(res is not None, 
 75                  msg='We should have got result from function %s' % f) 
 76              self.failUnless(N.all(ds.samples == ds_data), 
 77                  msg="Function %s should have not modified original dataset" % f) 
 78   
 79 -    def testSequenceStat(self): 
 80          """Test sequence statistics 
 81          """ 
 82          order = 3 
 83          # Close to perfectly balanced one 
 84          sp = N.array([-1,  1,  1, -1,  1, -1, -1,  1, -1, -1, -1, 
 85                        -1,  1, -1,  1, -1,  1,  1,  1, -1,  1,  1, 
 86                        -1, -1, -1,  1,  1,  1,  1,  1, -1], dtype=int) 
 87          rp = SequenceStats(sp, order=order) 
 88          self.failUnlessAlmostEqual(rp['sumabscorr'], 1.0) 
 89          self.failUnlessAlmostEqual(N.max(rp['corrcoef'] * (len(sp)-1) + 1.0), 0.0) 
 90   
 91          # Now some random but loong one still binary (boolean) 
 92          sb = (N.random.random_sample((1000,)) >= 0.5) 
 93          rb = SequenceStats(sb, order=order) 
 94   
 95          # Now lets do multiclass with literal labels 
 96          s5 = N.random.permutation(['f', 'bu', 'd', 0, 'zz']*200) 
 97          r5 = SequenceStats(s5, order=order) 
 98   
 99          # Degenerate one but still should be valid 
100          s1 = ['aaa']*100 
101          r1 = SequenceStats(s1, order=order) 
102   
103          # Generic conformance tests 
104          for r in (rp, rb, r5, r1): 
105              ulabels = r['ulabels'] 
106              nlabels = len(r['ulabels']) 
107              cbcounts = r['cbcounts'] 
108              self.failUnlessEqual(len(cbcounts), order) 
109              for cb in cbcounts: 
110                  self.failUnlessEqual(N.asarray(cb).shape, (nlabels, nlabels)) 
111              # Check if str works fine 
112              sr = str(r) 
113              # TODO: check the content 
114   
115 -def suite(): 
116      return unittest.makeSuite(MiscDatasetFxTests) 
117   
118   
119  if __name__ == '__main__': 
120      import runner 
121