def test_gaussian_mixture_fit_predict_n_init():
# Check that fit_predict is equivalent to fit.predict, when n_init > 1
X = np.random.RandomState(0).randn(1000, 5)
gm = GaussianMixture(n_components=5, n_init=5, random_state=0)
y_pred1 = gm.fit_predict(X)
y_pred2 = gm.predict(X)
assert_array_equal(y_pred1, y_pred2)
def test_gaussian_mixture_fit_predict(seed, max_iter, tol):
rng = np.random.RandomState(seed)
rand_data = RandomData(rng)
for covar_type in COVARIANCE_TYPE:
X = rand_data.X[covar_type]
Y = rand_data.Y
g = GaussianMixture(n_components=rand_data.n_components,
random_state=rng,
weights_init=rand_data.weights,
means_init=rand_data.means,
precisions_init=rand_data.precisions[covar_type],
covariance_type=covar_type,
max_iter=max_iter,
tol=tol)
# check if fit_predict(X) is equivalent to fit(X).predict(X)
f = copy.deepcopy(g)
Y_pred1 = f.fit(X).predict(X)
Y_pred2 = g.fit_predict(X)
assert_array_equal(Y_pred1, Y_pred2)
assert adjusted_rand_score(Y, Y_pred2) > .95