def test_predict(self):
"""
Test ordering and log score using hand-validated values.
"""
bs = BayesianSets()
bs.fit(np.array(self.X))
computed_ix = bs.predict([0, 1])
expected_ix = [0, 1, 2]
self.assertTrue(np.allclose(computed_ix[:3], expected_ix))
self.assertTrue(np.allclose(bs.log_scores_,
[-4.06, -4.41, -5.44, -6.59, -6.72, -6.72],
rtol=1E-2, atol=1E-2))
def test_fit(self):
"""
Test computation of mean, alpha and beta.
"""
c = 2
mean_X = np.array([2, 2, 3, 1, 2]) / 6.
bs = BayesianSets()
bs.fit(np.array(self.X))
self.assertTrue(np.allclose(bs.mean_X, mean_X))
alpha = mean_X * c
self.assertTrue(np.allclose(bs.alpha, alpha))
beta = c * (1 - mean_X)
self.assertTrue(np.allclose(bs.beta, beta))
def test_fit(self):
"""
Test computation of mean, alpha and beta.
"""
c = 2
mean_X = np.array([2, 2, 3, 1, 2]) / 6.
bs = BayesianSets()
bs.fit(np.array(self.X))
self.assertTrue(np.allclose(bs.mean_X, mean_X))
alpha = mean_X * c
self.assertTrue(np.allclose(bs.alpha, alpha))
beta = c * (1 - mean_X)
self.assertTrue(np.allclose(bs.beta, beta))
def test_check_params(self):
"""
Test checking of X.
"""
# Invalid value.
X = np.array([[1, 0, 2], [1, 1, 1]])
bs = BayesianSets()
with self.assertRaises(ValueError):
bs.fit(X)
# Columns containing all zeros.
X = np.array([[1, 0, 0], [0, 0, 1]])
bs = BayesianSets()
with warnings.catch_warnings(True) as w:
bs.fit(X)
self.assertTrue(len(w) >= 1)
self.assertTrue(np.allclose(bs.X.todense(), [[1, 0], [0, 1]]))
# Columns containing all ones.
X = np.array([[1, 1, 0], [1, 0, 1]])
bs = BayesianSets()
with warnings.catch_warnings(True) as w:
bs.fit(X)
self.assertTrue(len(w) >= 1)
self.assertTrue(np.allclose(bs.X.todense(), [[1, 0], [0, 1]]))
def test_check_params(self):
"""
Test checking of X.
"""
# Invalid value.
X = np.array([[1, 0, 2], [1, 1, 1]])
bs = BayesianSets()
with self.assertRaises(ValueError):
bs.fit(X)
# Columns containing all zeros.
X = np.array([[1, 0, 0], [0, 0, 1]])
bs = BayesianSets()
with warnings.catch_warnings(True) as w:
bs.fit(X)
self.assertTrue(len(w) >= 1)
self.assertTrue(np.allclose(bs.X.todense(), [[1, 0], [0, 1]]))
# Columns containing all ones.
X = np.array([[1, 1, 0], [1, 0, 1]])
bs = BayesianSets()
with warnings.catch_warnings(True) as w:
bs.fit(X)
self.assertTrue(len(w) >= 1)
self.assertTrue(np.allclose(bs.X.todense(), [[1, 0], [0, 1]]))