def test_set_resps(self):
data = np.array([[1, 3], [1, 6]])
mean1 = np.array([1, 5])
C_det1 = 0.75
C_inv1 = np.array([[4, -2], [-2, 4]]) / 3
mean2 = np.array([2, 2])
C_det2 = 1
C_inv2 = np.array([[1, 0], [0, 1]])
mixture = PCAMixture(data, 2, 2)
model1 = PCAModel(mixture, 2)
model2 = PCAModel(mixture, 2)
mixture.models = [model1, model2]
model1.mean = mean1
model1.C_det = C_det1
model1.C_inv = C_inv1
model2.mean = mean2
model2.C_det = C_det2
model2.C_inv = C_inv2
mixture.set_resps()
resps1 = mixture.resps[0]
resps2 = mixture.resps[1]
self.assertAlmostEqual(resps1[0], 0.1790458078)
self.assertAlmostEqual(resps1[1], 0.99965691)
self.assertAlmostEqual(resps2[0], 0.8209541918)
self.assertAlmostEqual(resps2[1], 3.43090081e-04)
def test_set_S(self):
data = np.array([[1, 3], [1, 6]])
mixture = PCAMixture(data, 1, 2)
model = PCAModel(mixture, 2)
model_resps = np.array([1, 2])
mixture.resps = [model_resps]
model.set_mean(model_resps)
mixture.mixing_coeffs = [3 / 2]
mixing_coeff = mixture.mixing_coeffs[0]
model_resps = mixture.resps[0]
model.set_S(mixing_coeff, model_resps)
self.assertTrue(np.array_equal(model.S, np.array([[0, 0], [0, 2]])))
def test_calc_mixing_coeff(self):
data = np.random.multivariate_normal([0, 0], [[2, 1], [1, 1]], 3)
mixture = PCAMixture(data, 1, 2)
model = PCAModel(mixture, 2)
resps = np.array([1, 2, 3])
self.assertEqual(model.calc_mixing_coeff(resps), 2)
def test_set_mean(self):
data = np.array([[1, 3], [1, 6]])
mixture = PCAMixture(data, 1, 2)
model = PCAModel(mixture, 2)
model_resps = np.array([1, 2])
model.set_mean(model_resps)
self.assertTrue(np.array_equal(model.mean, np.array([1, 5])))
def test_set_mixing_coefficients(self):
data = np.array([[1, 3], [1, 6]])
mean1 = np.array([1, 5])
C_det1 = 0.75
C_inv1 = np.array([[4, -2], [-2, 4]]) / 3
mean2 = np.array([2, 2])
C_det2 = 1
C_inv2 = np.array([[1, 0], [0, 1]])
mixture = PCAMixture(data, 2, 2)
model1 = PCAModel(mixture, 2)
model2 = PCAModel(mixture, 2)
mixture.models = [model1, model2]
model1.mean = mean1
model1.C_det = C_det1
model1.C_inv = C_inv1
model2.mean = mean2
model2.C_det = C_det2
model2.C_inv = C_inv2
mixture.set_resps()
mixture.set_mixing_coefficients()
mixing_coeffs = mixture.mixing_coeffs
self.assertAlmostEqual(mixing_coeffs[0], 0.5893513589)
self.assertAlmostEqual(mixing_coeffs[1], 0.41064864094)
def test_update_M_inv(self):
data = np.array([[1, 3], [1, 6]])
mixture = PCAMixture(data, 1, 2)
model = PCAModel(mixture, 2)
model.var = 3
model.W = np.array([[1, 2], [2, 1], [1, 3]])
model.update_M_inv()
M_inv = np.array([[17, -7], [-7, 9]]) / 104
res = np.isclose(model.M_inv, M_inv)
self.assertTrue(res.all())
def test_set_C_det(self):
data = np.array([[1, 3], [1, 6]])
var = 2
W = np.array([[2, 3], [1, 0], [2, 1]])
mixture = PCAMixture(data, 1, 2)
model = PCAModel(mixture, 2)
model.var = var
model.W = W
model.set_C_det()
C_det = 136
self.assertAlmostEqual(model.C_det, C_det)
def test_calc_prob_data(self):
data = np.array([[1, 3], [1, 6]])
mean = np.array([1, 5])
C_det = 0.75
C_inv = np.array([[4, -2], [-2, 4]]) / 3
mixture = PCAMixture(data, 1, 2)
model = PCAModel(mixture, 2)
model.mean = mean
model.C_det = C_det
model.C_inv = C_inv
prob = model.calc_prob_data(data[0])
correct_prob = 0.0127694115
self.assertAlmostEqual(prob, correct_prob)
def test_set_var(self):
S = np.array([[0, 0], [0, 2]])
data = np.array([[1, 3], [1, 6]])
mixture = PCAMixture(data, 1, 2)
model = PCAModel(mixture, 2)
model.S = S
W_old = np.array([[1, 1], [3, 2]])
M_inv_old = np.array([[17, -7], [-7, 9]]) / 104
model.W = np.array([[0, 0], [104 / 87, 208 / 261]])
model.set_var(W_old, M_inv_old)
self.assertAlmostEqual(model.var, 52 / 87)
def test_set_C_inv(self):
data = np.array([[1, 3], [1, 6]])
M_inv = np.array([[17, -7], [-7, 9]]) / 104
W = np.array([[0, 0], [104 / 87, 208 / 261]])
var = 1 + 2 / 87
mixture = PCAMixture(data, 1, 2)
model = PCAModel(mixture, 2)
model.M_inv = M_inv
model.W = W
model.var = var
model.set_C_inv()
C_inv = np.array([[1, 0], [0, 19067 / 22707]]) / var
res = np.isclose(model.C_inv, C_inv)
self.assertTrue(res.all())
def test_set_W(self):
S = np.array([[0, 0], [0, 2]])
M_inv = np.array([[17, -7], [-7, 9]]) / 104
W = np.array([[1, 1], [3, 2]])
data = np.array([[1, 3], [1, 6]])
mixture = PCAMixture(data, 1, 2)
model = PCAModel(mixture, 2)
model.var = 3
model.S = S
model.M_inv = M_inv
model.W = W
model.set_W()
new_W = model.W
#print(new_W)
res = np.isclose(new_W, np.array([[0, 0], [104 / 87, 208 / 261]]))
self.assertTrue(res.all())
def test_calc_prob_data(self):
data = np.array([[1, 3], [1, 6]])
mean = np.array([1, 5])
C_det = 0.75
C_inv = np.array([[4, -2], [-2, 4]]) / 3
mixture = PCAMixture(data, 1, 2)
model = PCAModel(mixture, 2)
mixture.models = [model]
model.mean = mean
model.C_det = C_det
model.C_inv = C_inv
prob = mixture.calc_prob_data(data[0])
real_prob = 0.0127694115
self.assertAlmostEqual(prob, real_prob)
data = np.array([[1, 3], [1, 6]])
mean1 = np.array([1, 5])
C_det1 = 0.75
C_inv1 = np.array([[4, -2], [-2, 4]]) / 3
mean2 = np.array([2, 2])
C_det2 = 1
C_inv2 = np.array([[1, 0], [0, 1]])
mixture = PCAMixture(data, 2, 2)
model1 = PCAModel(mixture, 2)
model2 = PCAModel(mixture, 2)
mixture.models = [model1, model2]
model1.mean = mean1
model1.C_det = C_det1
model1.C_inv = C_inv1
model2.mean = mean2
model2.C_det = C_det2
model2.C_inv = C_inv2
prob = mixture.calc_prob_data(data[0])
real_prob = 0.0356596215
self.assertAlmostEqual(real_prob, prob)
import numpy as np
from pca_mixtures.funcs import PCAMixture
import matplotlib.pyplot as plt
data = np.loadtxt("dataset/tobomovirus.txt")
#print(data)
mixture = PCAMixture(data, 1, 2)
mixture.fit()
class Plotter(object):
def __init__(self, mixture):
self.mixture = mixture
def plot_all_data(self, skip_improbable):
for model in self.mixture.models:
self._plot_data(data, model, skip_improbable)
def _plot_data(self, data, model, skip_improbable):
latent_positions = model.posterior(data)
plt.figure()
X = latent_positions[:, 0]
Y = latent_positions[:, 1]
plt.scatter(X, Y)
plt.figure()
nums = [str(i) for i in range(len(data))]
for x, y, txt in zip(X, Y, nums):
plt.text(x, y, txt)
plotter = Plotter(mixture)
def test_calc_resp(self):
data = np.random.multivariate_normal([0, 0], [[2, 1], [1, 1]], 2)
mixture = PCAMixture(data, 1, 2)
model = PCAModel(mixture, 2)
self.assertEqual(model.calc_resp(0.5, 2, 0.5), 0.125)
def test_get_model_resps(self):
# Test 1
# ----------------------------------
data = np.array([[1, 3], [1, 6]])
mixture = PCAMixture(data, 1, 2)
model = PCAModel(mixture, 2)
mean = np.array([1, 5])
C_det = 0.75
C_inv = np.array([[4, -2], [-2, 4]]) / 3
mixture = PCAMixture(data, 1, 2)
model = PCAModel(mixture, 2)
mixture.models = [model]
model.mean = mean
model.C_det = C_det
model.C_inv = C_inv
resps = mixture.get_model_resps(model)
res = np.array_equal(resps, np.array([1, 1]))
self.assertTrue(res)
# Test 2
# ---------------------------------
data = np.array([[1, 3], [1, 6]])
mean1 = np.array([1, 5])
C_det1 = 0.75
C_inv1 = np.array([[4, -2], [-2, 4]]) / 3
mean2 = np.array([2, 2])
C_det2 = 1
C_inv2 = np.array([[1, 0], [0, 1]])
mixture = PCAMixture(data, 2, 2)
model1 = PCAModel(mixture, 2)
model2 = PCAModel(mixture, 2)
mixture.models = [model1, model2]
model1.mean = mean1
model1.C_det = C_det1
model1.C_inv = C_inv1
model2.mean = mean2
model2.C_det = C_det2
model2.C_inv = C_inv2
resps1 = mixture.get_model_resps(model1)
self.assertAlmostEqual(resps1[0], 0.1790458078)
self.assertAlmostEqual(resps1[1], 0.99965691)
resps2 = mixture.get_model_resps(model2)
self.assertAlmostEqual(resps2[0], 0.8209541918)
self.assertAlmostEqual(resps2[1], 3.43090081e-04)