Source Code for Module mvpa.tests.test_lars

 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 least angle regression (LARS) classifier""" 
10   
11  from mvpa import cfg 
12  from mvpa.clfs.lars import LARS 
13  from scipy.stats import pearsonr 
14  from tests_warehouse import * 
15  from mvpa.misc.data_generators import normalFeatureDataset 
16   
17 -class LARSTests(unittest.TestCase): 
18   
19 -    def testLARS(self): 
20          # not the perfect dataset with which to test, but 
21          # it will do for now. 
22          #data = datasets['dumb2'] 
23          # for some reason the R code fails with the dumb data 
24          data = datasets['chirp_linear'] 
25   
26   
27          clf = LARS(regression=True) 
28   
29          clf.train(data) 
30   
31          # prediction has to be almost perfect 
32          # test with a correlation 
33          pre = clf.predict(data.samples) 
34          cor = pearsonr(pre, data.labels) 
35          if cfg.getboolean('tests', 'labile', default='yes'): 
36              self.failUnless(cor[0] > .8) 
37   
38 -    def testLARSState(self): 
39          #data = datasets['dumb2'] 
40          # for some reason the R code fails with the dumb data 
41          data = datasets['chirp_linear'] 
42   
43   
44          clf = LARS() 
45   
46          clf.train(data) 
47   
48          clf.states.enable('predictions') 
49   
50          p = clf.predict(data.samples) 
51   
52          self.failUnless((p == clf.predictions).all()) 
53   
54   
55 -    def testLARSSensitivities(self): 
56          data = datasets['chirp_linear'] 
57   
58          # use LARS on binary problem 
59          clf = LARS() 
60          clf.train(data) 
61   
62          # now ask for the sensitivities WITHOUT having to pass the dataset 
63          # again 
64          sens = clf.getSensitivityAnalyzer(force_training=False)() 
65   
66          self.failUnless(sens.shape == (data.nfeatures,)) 
67   
68   
69 -def suite(): 
70      return unittest.makeSuite(LARSTests) 
71   
72   
73  if __name__ == '__main__': 
74      import runner 
75