def test_train_different_inputs(self):
self.assertInvalidVectorTrain(
algorithms.Kohonen(
n_inputs=1,
n_outputs=2,
step=0.5,
verbose=False
),
np.array([1, 2, 3])
)
def test_kohonen_success(self):
kh = algorithms.Kohonen(self.conn, step=0.5, verbose=False)
# test one iteration update
data = np.reshape(input_data[0, :], (1, input_data.shape[1]))
kh.train(data, epochs=1)
self.assertTrue(
np.all(kh.input_layer.weight == np.array([
[0.7071, 0.4516, -1.0000],
[-0.7071, 0.84385, 0.0000],
])))
def test_predict_different_inputs(self):
knet = algorithms.Kohonen(
LinearLayer(1) > CompetitiveOutputLayer(2),
step=0.5,
verbose=False,
)
data = np.array([[1, 1, 1]]).T
target = np.array([
[1, 0],
[1, 0],
[1, 0],
])
knet.train(data, epochs=100)
self.assertInvalidVectorPred(knet, data.ravel(), target, decimal=2)
def test_kohonen_success(self):
kh = algorithms.Kohonen(
n_inputs=2,
n_outputs=3,
weight=np.array([
[0.7071, 0.7071, -1.0000],
[-0.7071, 0.7071, 0.0000],
]),
step=0.5,
verbose=False,
)
# test one iteration update
data = np.reshape(input_data[0, :], (1, input_data.shape[1]))
kh.train(data, epochs=1)
np.testing.assert_array_almost_equal(kh.weight,
np.array([
[0.7071, 0.4516, -1.0000],
[-0.7071, 0.84385, 0.0000],
]),
decimal=4)
def test_train_different_inputs(self):
self.assertInvalidVectorTrain(
algorithms.Kohonen(LinearLayer(1) > CompetitiveOutputLayer(2),
step=0.5,
verbose=False), np.array([1, 2, 3]))