Source Code for Module dimer.nnet.nnet_tests

 1   
 2  import unittest 
 3   
 4  import numpy as np 
 5  rng = np.random.RandomState() 
 6  import pandas as pd 
 7   
 8  from . import adjust_lr, shift_data, scale_data, fit_data 
 9   
10 -class TestLr( unittest.TestCase ): 
11 -    def setUp(self): 
12          rng = np.random.RandomState(10) 
13          self.rx = np.array( rng.rand( 15, 2, 10) ) 
14          self.xpn = pd.Panel(self.rx, items=map(lambda i: "anchor_%d" % i, range(self.rx.shape[0])), 
15                  major_axis=map(lambda i: "track_%d" % i, range(self.rx.shape[1])), 
16                  minor_axis=map(lambda i: "position_%d" % i, range(self.rx.shape[2])) ) 
17   
18          self.trials = range( rng.randint(0, self.rx.shape[1]*self.rx.shape[2] ) ) 
19   
20 -    def test_seqmonotonicity(self): 
21          self.assertEqual( adjust_lr([4], 0.1), 0.1 ) 
22   
23          self.assertLess( adjust_lr([4, 5], 0.1), 0.1 ) 
24          self.assertEqual( adjust_lr([5, 3], 0.1) , 0.1 ) 
25          self.assertEqual( adjust_lr([5, 5], 0.1) , 0.1 ) 
26   
27   
28 -    def test_shift(self): 
29          oxpn = self.xpn.copy() 
30          xpn, meandf = shift_data(self.xpn) 
31   
32          for trial in self.trials: 
33              track = rng.choice( xrange(self.rx.shape[1]) ) 
34              position = rng.choice( xrange(self.rx.shape[2]) ) 
35   
36              self.assertAlmostEqual(xpn.values[:,track, position].mean(), 0) 
37              self.assertAlmostEqual(oxpn.values[:,track, position].mean(), meandf[track, position]) 
38   
39 -    def test_scale(self): 
40          oxpn = self.xpn.copy() 
41          xpn, sddf = scale_data(self.xpn) 
42   
43          for trial in self.trials: 
44              track = rng.choice( xrange(self.rx.shape[1]) ) 
45              position = rng.choice( xrange(self.rx.shape[2]) ) 
46   
47              self.assertAlmostEqual(xpn.values[:,track, position].std(), 1) 
48              self.assertAlmostEqual(oxpn.values[:,track, position].std(), sddf[track, position]) 
49   
50 -    def test_fit(self): 
51          xpn = fit_data(self.xpn) 
52   
53          for trial in self.trials: 
54              track = rng.choice( xrange(self.rx.shape[1]) ) 
55              position = rng.choice( xrange(self.rx.shape[2]) ) 
56   
57              self.assertLessEqual(np.max( np.abs(xpn.values[:,track, position]) ), 1) 
58