def test_grid_search_sparse():
# Test that grid search works with both dense and sparse matrices
X, y = make_classification(n_samples=200, n_features=100, random_state=0)
cv = GridSearchCV(LinearSVC(), {'C': [0.1, 1.0]})
cv.fit(X[:180], y[:180])
y_pred = cv.predict(X[180:])
C = cv.best_estimator_.C
X = sparse.csr_matrix(X)
cv.fit(X[:180], y[:180])
y_pred2 = cv.predict(X[180:])
C2 = cv.best_estimator_.C
assert np.mean(y_pred == y_pred2) >= .9
assert C == C2
def test_grid_search_dask_inputs():
# Test that the best estimator contains the right value for foo_param
dX = da.from_array(X, chunks=2)
dy = da.from_array(y, chunks=2)
clf = MockClassifier()
grid_search = GridSearchCV(clf, {'foo_param': [1, 2, 3]})
# make sure it selects the smallest parameter in case of ties
grid_search.fit(dX, dy)
assert grid_search.best_estimator_.foo_param == 2
for i, foo_i in enumerate([1, 2, 3]):
assert grid_search.grid_scores_[i][0] == {'foo_param': foo_i}
y_pred = grid_search.predict(dX)
assert isinstance(y_pred, da.Array)
tm.assert_array_equal(y_pred, X.sum(axis=1))
y_pred = grid_search.predict(X)
assert isinstance(y_pred, np.ndarray)
tm.assert_array_equal(y_pred, X.sum(axis=1))