def test_fit_index():
model = FE(Ridge, fit_intercept=False)
model.fit(X, y, range(75, 100), range(75))
assert isinstance(model.estimator_, Ridge)
assert len(model.estimator_.coef_) == 3 * X.shape[1]
assert np.abs(np.sum(model.estimator_.coef_[:X.shape[1]]) +
np.sum(model.estimator_.coef_[-X.shape[1]:])) < 0.01
assert np.abs(np.sum(model.estimator_.coef_[X.shape[1]:]) - 1) < 0.01
def test_fit_only_on_given_index():
y_corrupted = np.copy(y)
y_corrupted[25: 75] = np.nan
X_corrupted = np.copy(X)
X_corrupted[25: 75] = np.nan
model = FE(Ridge, fit_intercept=False)
model.fit(X_corrupted, y_corrupted,
range(25), range(75, 100))
assert isinstance(model.estimator_, Ridge)
def test_predict():
model = FE(Ridge)
model.fit(X, y, range(75), range(75, 100))
y_pred = model.predict(X)
assert np.all(y_pred < 0.01)
y_pred = model.predict(X, domain="target")
assert np.all(y_pred < 0.01)
y_pred = model.predict(X, domain="source")
assert np.all(y_pred - 1 < 0.01)
with pytest.raises(ValueError):
model.predict(X, domain="tirelipimpon")
def test_fit_params():
model = FE(Ridge, alpha=123)
model.fit(X, y, range(75), range(75, 100))
assert model.estimator_.alpha == 123
def test_fit_sample_weight():
model = FE(Ridge)
model.fit(X, y, range(75), range(75, 100),
sample_weight=np.array([0] * 75 + [1e4] * 25))
assert np.all(model.estimator_.coef_ == 0)
def test_fit_default():
model = FE()
model.fit(X, y, range(75), range(75, 100))
assert isinstance(model.estimator_, LinearRegression)