def train_unsupervised_hierarchy(X1, X2, n_epochs=30):
g1 = gwr_h_unimodal(n_layers=2, window_size=[3, 3], gwr_pars=pars)
g2 = gwr_h_unimodal(n_layers=2, window_size=[3, 3], gwr_pars=pars)
g3 = gwr(**pars)
X1_ = g1.train(X1, n_epochs=n_epochs)
X2_ = g2.train(X2, n_epochs=n_epochs)
X = np.hstack((X1_, X2_))
g3.train(X, n_epochs=n_epochs)
return g1, g2, g3
def test_activation_trajectories(self):
X = np.array([[1],
[2],
[3],
[4],
[5],
[6]])
ossh = gwr_h_unimodal(window_size=[1, 2, 3], lab_ratio=2 / 3)
ossh.H[0].G.add_node(0, attr_dict={'pos': np.array([1]),
'lab': 1})
ossh.H[0].G.add_node(1, attr_dict={'pos': np.array([4]),
'lab': 2})
ossh.H[0].G.add_node(2, attr_dict={'pos': np.array([7]),
'lab': 3})
ossh.H[1].G.add_node(0, attr_dict={'pos': np.array([1, 1]),
'lab': 1})
ossh.H[1].G.add_node(1, attr_dict={'pos': np.array([5, 7]),
'lab': 2})
tr0_fp = ossh._get_trajectories_from_previous(X,0)
tr0 = ossh._get_activation_trajectories_old(X, 0)
tr0_test = X
np.testing.assert_array_equal(tr0_fp, tr0)
np.testing.assert_array_equal(tr0, tr0_test)
tr1_fp = ossh._get_trajectories_from_previous(tr0,1)
tr1 = ossh._get_activation_trajectories_old(X, 1)
tr1_test = np.array([[1, 1],
[1, 4],
[4, 4],
[4, 4],
[4, 7]])
np.testing.assert_array_equal(tr1_fp, tr1)
np.testing.assert_array_equal(tr1, tr1_test)
tr2_fp = ossh._get_trajectories_from_previous(tr1, 2)
tr2 = ossh._get_activation_trajectories_old(X, 2)
tr2_test = np.array([[1, 1, 1, 1, 5, 7],
[1, 1, 5, 7, 5, 7],
[5, 7, 5, 7, 5, 7]])
np.testing.assert_array_equal(tr2_fp,tr2)
np.testing.assert_array_equal(tr2,tr2_test)
a1_raw = pd.read_csv(dir + 'a1_raw.csv')
a1_va3 = np.genfromtxt(dir + 'a1_va3.csv', delimiter=',')
y = np.array(a1_raw['phase'].astype('category').cat.codes)
pos = np.array(a1_raw.iloc[:, :18])
dif = a1_raw.shape[0] - a1_va3.shape[0]
pos = pos[dif // 2:-dif // 2, :18]
y = y[dif // 2:-dif // 2]
vel = a1_va3[:, :32]
pars = {'act_thr': 0.75, 'max_size': 200}
g1 = gwr_h_unimodal(n_layers=2, window_size=[3, 3], gwr_pars=pars)
g2 = gwr_h_unimodal(n_layers=2, window_size=[3, 3], gwr_pars=pars)
pos_ = g1.train(pos, n_epochs=40)
vel_ = g2.train(vel, n_epochs=40)
#posc = g1._get_activation_trajectories(pos)
X_ = np.hstack((pos_, vel_))
# TODO add an extra unsupervised layer here
y_ = -np.ones(X_.shape[0])
for i in range(5, y.shape[0]):
c = collections.Counter(y[i - 5:i])
lab = c.most_common()[0][0]
import matplotlib.pyplot as plt
from neuralgas.oss_gwr import gwr
from neuralgas.oss_gwr_h import gwr_h_unimodal
from neuralgas.oss_gwr_h import _propagate_trajectories
iris = sklearn.datasets.load_iris()
X = iris.data
X1 = X[:, :2]
X2 = X[:, 2:]
y = iris.target
g1 = gwr_h_unimodal(n_layers=2,
window_size=[1, 2],
gwr_pars=[{
'act_thr': 0.7
}, {
'act_thr': 0.7
}])
g2 = gwr_h_unimodal(n_layers=2,
window_size=[1, 2],
gwr_pars=[{
'act_thr': 0.7
}, {
'act_thr': 0.7
}])
g3 = gwr()
XX1 = g1.train(X1)