def test_simple_som(self):
colors = np.array([[0., 0., 0.], [0., 0., 1.], [0., 1., 0.],
[1., 0., 0.], [0., 1., 1.], [1., 0., 1.],
[1., 1., 0.], [1., 1., 1.]])
distance_measures = (None, lambda x, y:(x ** 3 + y ** 3) ** (1. / 3))
for distance_measure in distance_measures:
# only small SOM for speed reasons
som = SimpleSOMMapper((10, 5), 200, learning_rate=0.05)
# no acces when nothing is there
self.assertRaises(RuntimeError, som._access_kohonen)
som.train(colors)
fmapped = som.forward(colors)
self.assertTrue(fmapped.shape == (8, 2))
# reverse mapping
rmapped = som.reverse(fmapped)
if cfg.getboolean('tests', 'labile', default='yes'):
# should approximately restore the input, but could fail
# with bad initialisation
self.assertTrue((np.round(rmapped) == colors).all())
def test_kohonen_update(self):
# before update error occured when learning_rate*number of samples > 1
# here use extreme learning_rate to force bad behaviour
som = SimpleSOMMapper((10, 5), 200, learning_rate=1.0)
trainer = np.ones([8, 3])
som.train(trainer)
# use 10 instead of 4 to allow for some randomness in the training
# fail values tend to be closer to 10^30
self.assertTrue((np.abs(som.K) <= 10).all())
def test_kohonen_update(self):
# before update error occurred when learning_rate*number of samples > 1
# here use extreme learning_rate to force bad behavior
som = SimpleSOMMapper((10, 5), 200, learning_rate=1.0)
trainer = np.ones([8,3])
som.train(trainer)
# use 10 instead of 4 to allow for some randomness in the training
# fail values tend to be closer to 10^30
self.assertTrue((np.abs(som.K) <= 10).all())
def test_periodic_boundaries(self):
som = SimpleSOMMapper((10, 5), 200, learning_rate=0.05)
test_dqdshape = np.array([5, 2, 5, 3])
# som._dqdshape only defined in newer version
# this is not explicitly linked to the periodic boundary conditions,
# but had trouble coming up with a simple test for them
self.assertTrue((som._dqdshape == test_dqdshape).all())
def test_simple_som(self):
colors = np.array([[0., 0., 0.], [0., 0., 1.], [0., 1., 0.],
[1., 0., 0.], [0., 1., 1.], [1., 0., 1.],
[1., 1., 0.], [1., 1., 1.]])
distance_measures = (None, lambda x, y: (x**3 + y**3)**(1. / 3))
for distance_measure in distance_measures:
# only small SOM for speed reasons
som = SimpleSOMMapper((10, 5), 200, learning_rate=0.05)
# no acces when nothing is there
self.assertRaises(RuntimeError, som._access_kohonen)
som.train(colors)
fmapped = som.forward(colors)
self.assertTrue(fmapped.shape == (8, 2))
# reverse mapping
rmapped = som.reverse(fmapped)
if cfg.getboolean('tests', 'labile', default='yes'):
# should approximately restore the input, but could fail
# with bad initialisation
self.assertTrue((np.round(rmapped) == colors).all())
import numpy as np
colors = np.array([[0., 0., 0.], [0., 0., 1.], [0., 0., 0.5],
[0.125, 0.529, 1.0], [0.33, 0.4, 0.67], [0.6, 0.5, 1.0],
[0., 1., 0.], [1., 0., 0.], [0., 1., 1.], [1., 0., 1.],
[1., 1., 0.], [1., 1., 1.], [.33, .33, .33], [.5, .5, .5],
[.66, .66, .66]])
# store the names of the colors for visualization later on
color_names = \
['black', 'blue', 'darkblue', 'skyblue',
'greyblue', 'lilac', 'green', 'red',
'cyan', 'violet', 'yellow', 'white',
'darkgrey', 'mediumgrey', 'lightgrey']
som = SimpleSOMMapper((20, 30), 400, learning_rate=0.05)
som.train(colors)
plt.imshow(som.K, origin='lower')
mapped = som(colors)
plt.title('Color SOM')
# SOM's kshape is (rows x columns), while matplotlib wants (X x Y)
for i, m in enumerate(mapped):
plt.text(m[1],
m[0],
color_names[i],
ha='center',
va='center',
bbox=dict(facecolor='white', alpha=0.5, lw=0))
