def test_make_parameterized_classification():
# Simple case
mu0 = theano.shared(0.)
mu1 = theano.shared(1.)
p0 = Normal(mu=mu0)
p1 = Normal(mu=mu1)
X, y = make_parameterized_classification(p0, p1, 100, [mu0, mu1])
assert X.shape == (100, 1+2)
assert_array_almost_equal(X[:, 1], np.zeros(100))
assert_array_almost_equal(X[:, 2], np.ones(100))
# Grid of parameter values
X, y = make_parameterized_classification(p0, p1, 100,
[(mu0, [0, 0.5]),
(mu1, [0.5, 1.5])])
assert X.shape == (100, 1+2)
assert_array_equal(np.unique(X[:, 1]), [0, 0.5])
assert_array_equal(np.unique(X[:, 2]), [0.5, 1.5])
d = set()
for row in X[:, 1:]:
d.add(tuple(row))
assert_array_equal(np.array(sorted(d)), [[0., 0.5], [0., 1.5],
[0.5, 0.5], [0.5, 1.5]])
def test_parameterized_classifier():
mu0 = theano.shared(0)
mu1 = theano.shared(1)
p0 = Normal(mu=mu0)
p1 = Normal(mu=mu1)
X, y = make_parameterized_classification(p0, p1, 100, [mu0, mu1])
clf = ParameterizedClassifier(DecisionTreeClassifier(), params=[mu0, mu1])
clf.fit(X, y)
assert clf.n_features_ == 1
assert_array_almost_equal(y, clf.predict(X))
