def test_enet_float_precision():
    # Generate dataset
    X, y, X_test, y_test = build_dataset(n_samples=20, n_features=10)
    # Here we have a small number of iterations, and thus the
    # ElasticNet might not converge. This is to speed up tests

    for normalize in [True, False]:
        for fit_intercept in [True, False]:
            coef = {}
            intercept = {}
            clf = ElasticNet(alpha=0.5, max_iter=100, precompute=False,
                            fit_intercept=fit_intercept, normalize=normalize)
            for dtype in [np.float64, np.float32]:
                X = dtype(X)
                y = dtype(y)
                ignore_warnings(clf.fit)(X, y)

                coef[dtype] = clf.coef_
                intercept[dtype] = clf.intercept_

                assert_equal(clf.coef_.dtype, dtype)

            assert_array_almost_equal(coef[np.float32], coef[np.float64],
                                    decimal=4)
            assert_array_almost_equal(intercept[np.float32],
                                    intercept[np.float64],
                                    decimal=4)

def test_enet_toy_list_input():
    # Test ElasticNet for various values of alpha and l1_ratio with list X

    X = np.array([[-1], [0], [1]])
    X = sp.csc_matrix(X)
    Y = [-1, 0, 1]       # just a straight line
    T = np.array([[2], [3], [4]])  # test sample

    # this should be the same as unregularized least squares
    clf = ElasticNet(alpha=0, l1_ratio=1.0)
    # catch warning about alpha=0.
    # this is discouraged but should work.
    ignore_warnings(clf.fit)(X, Y)
    pred = clf.predict(T)
    assert_array_almost_equal(clf.coef_, [1])
    assert_array_almost_equal(pred, [2, 3, 4])
    assert_almost_equal(clf.dual_gap_, 0)

    clf = ElasticNet(alpha=0.5, l1_ratio=0.3, max_iter=1000)
    clf.fit(X, Y)
    pred = clf.predict(T)
    assert_array_almost_equal(clf.coef_, [0.50819], decimal=3)
    assert_array_almost_equal(pred, [1.0163,  1.5245,  2.0327], decimal=3)
    assert_almost_equal(clf.dual_gap_, 0)

    clf = ElasticNet(alpha=0.5, l1_ratio=0.5)
    clf.fit(X, Y)
    pred = clf.predict(T)
    assert_array_almost_equal(clf.coef_, [0.45454], 3)
    assert_array_almost_equal(pred, [0.9090,  1.3636,  1.8181], 3)
    assert_almost_equal(clf.dual_gap_, 0)

def test_enet_path():
    # We use a large number of samples and of informative features so that
    # the l1_ratio selected is more toward ridge than lasso
    X, y, X_test, y_test = build_dataset(n_samples=200, n_features=100,
                                         n_informative_features=100)
    max_iter = 150

    # Here we have a small number of iterations, and thus the
    # ElasticNet might not converge. This is to speed up tests
    clf = ElasticNetCV(n_alphas=5, eps=2e-3, l1_ratio=[0.5, 0.7], cv=3,
                       max_iter=max_iter)
    ignore_warnings(clf.fit)(X, y)
    # Well-conditioned settings, we should have selected our
    # smallest penalty
    assert_almost_equal(clf.alpha_, min(clf.alphas_))
    # Non-sparse ground truth: we should have seleted an elastic-net
    # that is closer to ridge than to lasso
    assert_equal(clf.l1_ratio_, min(clf.l1_ratio))

    clf = ElasticNetCV(n_alphas=5, eps=2e-3, l1_ratio=[0.5, 0.7], cv=3,
                       max_iter=max_iter, precompute=True)
    ignore_warnings(clf.fit)(X, y)


    # Well-conditioned settings, we should have selected our
    # smallest penalty
    assert_almost_equal(clf.alpha_, min(clf.alphas_))
    # Non-sparse ground truth: we should have seleted an elastic-net
    # that is closer to ridge than to lasso
    assert_equal(clf.l1_ratio_, min(clf.l1_ratio))

    # We are in well-conditioned settings with low noise: we should
    # have a good test-set performance
    assert_greater(clf.score(X_test, y_test), 0.99)

def test_hash_functions():
    # Checks randomness of hash functions.
    # Variance and mean of each hash function (projection vector)
    # should be different from flattened array of hash functions.
    # If hash functions are not randomly built (seeded with
    # same value), variances and means of all functions are equal.
    n_samples = 12
    n_features = 2
    n_estimators = 5
    rng = np.random.RandomState(42)
    X = rng.rand(n_samples, n_features)

    lshf = ignore_warnings(LSHForest, category=DeprecationWarning)(
        n_estimators=n_estimators,
        random_state=rng.randint(0, np.iinfo(np.int32).max))
    ignore_warnings(lshf.fit)(X)

    hash_functions = []
    for i in range(n_estimators):
        hash_functions.append(lshf.hash_functions_[i].components_)

    for i in range(n_estimators):
        assert_not_equal(np.var(hash_functions),
                         np.var(lshf.hash_functions_[i].components_))

    for i in range(n_estimators):
        assert_not_equal(np.mean(hash_functions),
                         np.mean(lshf.hash_functions_[i].components_))

def test_enet_l1_ratio():
    # Test that an error message is raised if an estimator that
    # uses _alpha_grid is called with l1_ratio=0
    msg = ("Automatic alpha grid generation is not supported for l1_ratio=0. "
           "Please supply a grid by providing your estimator with the "
           "appropriate `alphas=` argument.")
    X = np.array([[1, 2, 4, 5, 8], [3, 5, 7, 7, 8]]).T
    y = np.array([12, 10, 11, 21, 5])

    assert_raise_message(ValueError, msg, ElasticNetCV(
        l1_ratio=0, random_state=42).fit, X, y)
    assert_raise_message(ValueError, msg, MultiTaskElasticNetCV(
        l1_ratio=0, random_state=42).fit, X, y[:, None])

    # Test that l1_ratio=0 is allowed if we supply a grid manually
    alphas = [0.1, 10]
    estkwds = {'alphas': alphas, 'random_state': 42}
    est_desired = ElasticNetCV(l1_ratio=0.00001, **estkwds)
    est = ElasticNetCV(l1_ratio=0, **estkwds)
    with ignore_warnings():
        est_desired.fit(X, y)
        est.fit(X, y)
    assert_array_almost_equal(est.coef_, est_desired.coef_, decimal=5)

    est_desired = MultiTaskElasticNetCV(l1_ratio=0.00001, **estkwds)
    est = MultiTaskElasticNetCV(l1_ratio=0, **estkwds)
    with ignore_warnings():
        est.fit(X, y[:, None])
        est_desired.fit(X, y[:, None])
    assert_array_almost_equal(est.coef_, est_desired.coef_, decimal=5)

def test_fit():
    # Checks whether `fit` method sets all attribute values correctly.
    n_samples = 12
    n_features = 2
    n_estimators = 5
    rng = np.random.RandomState(42)
    X = rng.rand(n_samples, n_features)

    lshf = ignore_warnings(LSHForest, category=DeprecationWarning)(
        n_estimators=n_estimators)
    ignore_warnings(lshf.fit)(X)

    # _input_array = X
    assert_array_equal(X, lshf._fit_X)
    # A hash function g(p) for each tree
    assert_equal(n_estimators, len(lshf.hash_functions_))
    # Hash length = 32
    assert_equal(32, lshf.hash_functions_[0].components_.shape[0])
    # Number of trees_ in the forest
    assert_equal(n_estimators, len(lshf.trees_))
    # Each tree has entries for every data point
    assert_equal(n_samples, len(lshf.trees_[0]))
    # Original indices after sorting the hashes
    assert_equal(n_estimators, len(lshf.original_indices_))
    # Each set of original indices in a tree has entries for every data point
    assert_equal(n_samples, len(lshf.original_indices_[0]))

def test_check_estimator_clones():
    # check that check_estimator doesn't modify the estimator it receives
    from sklearn.datasets import load_iris
    iris = load_iris()

    for Estimator in [GaussianMixture, LinearRegression,
                      RandomForestClassifier, NMF, SGDClassifier,
                      MiniBatchKMeans]:
        with ignore_warnings(category=FutureWarning):
            # when 'est = SGDClassifier()'
            est = Estimator()
        set_checking_parameters(est)
        set_random_state(est)
        # without fitting
        old_hash = joblib.hash(est)
        check_estimator(est)
        assert_equal(old_hash, joblib.hash(est))

        with ignore_warnings(category=FutureWarning):
            # when 'est = SGDClassifier()'
            est = Estimator()
        set_checking_parameters(est)
        set_random_state(est)
        # with fitting
        est.fit(iris.data + 10, iris.target)
        old_hash = joblib.hash(est)
        check_estimator(est)
        assert_equal(old_hash, joblib.hash(est))

def test_neighbors_accuracy_with_n_estimators():
    # Checks whether accuracy increases as `n_estimators` increases.
    n_estimators = np.array([1, 10, 100])
    n_samples = 100
    n_features = 10
    n_iter = 10
    n_points = 5
    rng = np.random.RandomState(42)
    accuracies = np.zeros(n_estimators.shape[0], dtype=float)
    X = rng.rand(n_samples, n_features)

    for i, t in enumerate(n_estimators):
        lshf = ignore_warnings(LSHForest, category=DeprecationWarning)(
            n_candidates=500, n_estimators=t)
        ignore_warnings(lshf.fit)(X)
        for j in range(n_iter):
            query = X[rng.randint(0, n_samples)].reshape(1, -1)
            neighbors = lshf.kneighbors(query, n_neighbors=n_points,
                                        return_distance=False)
            distances = pairwise_distances(query, X, metric='cosine')
            ranks = np.argsort(distances)[0, :n_points]

            intersection = np.intersect1d(ranks, neighbors).shape[0]
            ratio = intersection / float(n_points)
            accuracies[i] = accuracies[i] + ratio

        accuracies[i] = accuracies[i] / float(n_iter)
    # Sorted accuracies should be equal to original accuracies
    assert_true(np.all(np.diff(accuracies) >= 0),
                msg="Accuracies are not non-decreasing.")
    # Highest accuracy should be strictly greater than the lowest
    assert_true(np.ptp(accuracies) > 0,
                msg="Highest accuracy is not strictly greater than lowest.")

def test_no_atoms():
    y_empty = np.zeros_like(y)
    Xy_empty = np.dot(X.T, y_empty)
    gamma_empty = ignore_warnings(orthogonal_mp)(X, y_empty, 1)
    gamma_empty_gram = ignore_warnings(orthogonal_mp)(G, Xy_empty, 1)
    assert_equal(np.all(gamma_empty == 0), True)
    assert_equal(np.all(gamma_empty_gram == 0), True)

def test_sparse_input():
    # note: Fixed random state in sp.rand is not supported in older scipy.
    #       The test should succeed regardless.
    X1 = sp.rand(50, 100)
    X2 = sp.rand(10, 100)
    forest_sparse = ignore_warnings(LSHForest, category=DeprecationWarning)(
        radius=1, random_state=0).fit(X1)
    forest_dense = ignore_warnings(LSHForest, category=DeprecationWarning)(
        radius=1, random_state=0).fit(X1.A)

    d_sparse, i_sparse = forest_sparse.kneighbors(X2, return_distance=True)
    d_dense, i_dense = forest_dense.kneighbors(X2.A, return_distance=True)

    assert_almost_equal(d_sparse, d_dense)
    assert_almost_equal(i_sparse, i_dense)

    d_sparse, i_sparse = forest_sparse.radius_neighbors(X2,
                                                        return_distance=True)
    d_dense, i_dense = forest_dense.radius_neighbors(X2.A,
                                                     return_distance=True)
    assert_equal(d_sparse.shape, d_dense.shape)
    for a, b in zip(d_sparse, d_dense):
        assert_almost_equal(a, b)
    for a, b in zip(i_sparse, i_dense):
        assert_almost_equal(a, b)

def check_warm_start_oob(name):
    # Test that the warm start computes oob score when asked.
    X, y = hastie_X, hastie_y
    ForestEstimator = FOREST_ESTIMATORS[name]
    # Use 15 estimators to avoid 'some inputs do not have OOB scores' warning.
    clf = ForestEstimator(n_estimators=15, max_depth=3, warm_start=False,
                          random_state=1, bootstrap=True, oob_score=True)
    clf.fit(X, y)

    clf_2 = ForestEstimator(n_estimators=5, max_depth=3, warm_start=False,
                            random_state=1, bootstrap=True, oob_score=False)
    clf_2.fit(X, y)

    clf_2.set_params(warm_start=True, oob_score=True, n_estimators=15)
    clf_2.fit(X, y)

    assert_true(hasattr(clf_2, 'oob_score_'))
    assert_equal(clf.oob_score_, clf_2.oob_score_)

    # Test that oob_score is computed even if we don't need to train
    # additional trees.
    clf_3 = ForestEstimator(n_estimators=15, max_depth=3, warm_start=True,
                            random_state=1, bootstrap=True, oob_score=False)
    clf_3.fit(X, y)
    assert_true(not(hasattr(clf_3, 'oob_score_')))

    clf_3.set_params(oob_score=True)
    ignore_warnings(clf_3.fit)(X, y)

    assert_equal(clf.oob_score_, clf_3.oob_score_)

def test_cross_val_predict_input_types():
    clf = Ridge()
    # Smoke test
    predictions = cval.cross_val_predict(clf, X, y)
    assert_equal(predictions.shape, (10,))

    # test with multioutput y
    with ignore_warnings(category=ConvergenceWarning):
        predictions = cval.cross_val_predict(clf, X_sparse, X)
    assert_equal(predictions.shape, (10, 2))

    predictions = cval.cross_val_predict(clf, X_sparse, y)
    assert_array_equal(predictions.shape, (10,))

    # test with multioutput y
    with ignore_warnings(category=ConvergenceWarning):
        predictions = cval.cross_val_predict(clf, X_sparse, X)
    assert_array_equal(predictions.shape, (10, 2))

    # test with X and y as list
    list_check = lambda x: isinstance(x, list)
    clf = CheckingClassifier(check_X=list_check)
    predictions = cval.cross_val_predict(clf, X.tolist(), y.tolist())

    clf = CheckingClassifier(check_y=list_check)
    predictions = cval.cross_val_predict(clf, X, y.tolist())

    # test with 3d X and
    X_3d = X[:, :, np.newaxis]
    check_3d = lambda x: x.ndim == 3
    clf = CheckingClassifier(check_X=check_3d)
    predictions = cval.cross_val_predict(clf, X_3d, y)
    assert_array_equal(predictions.shape, (10,))

def test_rfecv():
    generator = check_random_state(0)
    iris = load_iris()
    X = np.c_[iris.data, generator.normal(size=(len(iris.data), 6))]
    y = list(iris.target)  # regression test: list should be supported

    # Test using the score function
    rfecv = RFECV(estimator=SVC(kernel="linear"), step=1, cv=5)
    rfecv.fit(X, y)
    # non-regression test for missing worst feature:
    assert_equal(len(rfecv.grid_scores_), X.shape[1])
    assert_equal(len(rfecv.ranking_), X.shape[1])
    X_r = rfecv.transform(X)

    # All the noisy variable were filtered out
    assert_array_equal(X_r, iris.data)

    # same in sparse
    rfecv_sparse = RFECV(estimator=SVC(kernel="linear"), step=1, cv=5)
    X_sparse = sparse.csr_matrix(X)
    rfecv_sparse.fit(X_sparse, y)
    X_r_sparse = rfecv_sparse.transform(X_sparse)
    assert_array_equal(X_r_sparse.toarray(), iris.data)

    # Test using a customized loss function
    scoring = make_scorer(zero_one_loss, greater_is_better=False)
    rfecv = RFECV(estimator=SVC(kernel="linear"), step=1, cv=5, scoring=scoring)
    ignore_warnings(rfecv.fit)(X, y)
    X_r = rfecv.transform(X)
    assert_array_equal(X_r, iris.data)

    # Test using a scorer
    scorer = get_scorer("accuracy")
    rfecv = RFECV(estimator=SVC(kernel="linear"), step=1, cv=5, scoring=scorer)
    rfecv.fit(X, y)
    X_r = rfecv.transform(X)
    assert_array_equal(X_r, iris.data)

    # Test fix on grid_scores
    def test_scorer(estimator, X, y):
        return 1.0

    rfecv = RFECV(estimator=SVC(kernel="linear"), step=1, cv=5, scoring=test_scorer)
    rfecv.fit(X, y)
    assert_array_equal(rfecv.grid_scores_, np.ones(len(rfecv.grid_scores_)))

    # Same as the first two tests, but with step=2
    rfecv = RFECV(estimator=SVC(kernel="linear"), step=2, cv=5)
    rfecv.fit(X, y)
    assert_equal(len(rfecv.grid_scores_), 6)
    assert_equal(len(rfecv.ranking_), X.shape[1])
    X_r = rfecv.transform(X)
    assert_array_equal(X_r, iris.data)

    rfecv_sparse = RFECV(estimator=SVC(kernel="linear"), step=2, cv=5)
    X_sparse = sparse.csr_matrix(X)
    rfecv_sparse.fit(X_sparse, y)
    X_r_sparse = rfecv_sparse.transform(X_sparse)
    assert_array_equal(X_r_sparse.toarray(), iris.data)

def test_radius_neighbors():
    # Checks whether Returned distances are less than `radius`
    # At least one point should be returned when the `radius` is set
    # to mean distance from the considering point to other points in
    # the database.
    # Moreover, this test compares the radius neighbors of LSHForest
    # with the `sklearn.neighbors.NearestNeighbors`.
    n_samples = 12
    n_features = 2
    n_iter = 10
    rng = np.random.RandomState(42)
    X = rng.rand(n_samples, n_features)

    lshf = ignore_warnings(LSHForest, category=DeprecationWarning)()
    # Test unfitted estimator
    assert_raises(ValueError, lshf.radius_neighbors, X[0])

    ignore_warnings(lshf.fit)(X)

    for i in range(n_iter):
        # Select a random point in the dataset as the query
        query = X[rng.randint(0, n_samples)].reshape(1, -1)

        # At least one neighbor should be returned when the radius is the
        # mean distance from the query to the points of the dataset.
        mean_dist = np.mean(pairwise_distances(query, X, metric='cosine'))
        neighbors = lshf.radius_neighbors(query, radius=mean_dist,
                                          return_distance=False)

        assert_equal(neighbors.shape, (1,))
        assert_equal(neighbors.dtype, object)
        assert_greater(neighbors[0].shape[0], 0)
        # All distances to points in the results of the radius query should
        # be less than mean_dist
        distances, neighbors = lshf.radius_neighbors(query,
                                                     radius=mean_dist,
                                                     return_distance=True)
        assert_array_less(distances[0], mean_dist)

    # Multiple points
    n_queries = 5
    queries = X[rng.randint(0, n_samples, n_queries)]
    distances, neighbors = lshf.radius_neighbors(queries,
                                                 return_distance=True)

    # dists and inds should not be 1D arrays or arrays of variable lengths
    # hence the use of the object dtype.
    assert_equal(distances.shape, (n_queries,))
    assert_equal(distances.dtype, object)
    assert_equal(neighbors.shape, (n_queries,))
    assert_equal(neighbors.dtype, object)

    # Compare with exact neighbor search
    query = X[rng.randint(0, n_samples)].reshape(1, -1)
    mean_dist = np.mean(pairwise_distances(query, X, metric='cosine'))
    nbrs = NearestNeighbors(algorithm='brute', metric='cosine').fit(X)

    distances_exact, _ = nbrs.radius_neighbors(query, radius=mean_dist)
    distances_approx, _ = lshf.radius_neighbors(query, radius=mean_dist)

def test_consistent_proba():
    a = svm.SVC(probability=True, max_iter=1, random_state=0)
    with ignore_warnings(category=ConvergenceWarning):
        proba_1 = a.fit(X, Y).predict_proba(X)
    a = svm.SVC(probability=True, max_iter=1, random_state=0)
    with ignore_warnings(category=ConvergenceWarning):
        proba_2 = a.fit(X, Y).predict_proba(X)
    assert_array_almost_equal(proba_1, proba_2)

def test_kernel_ridge_singular_kernel():
    # alpha=0 causes a LinAlgError in computing the dual coefficients,
    # which causes a fallback to a lstsq solver. This is tested here.
    pred = Ridge(alpha=0, fit_intercept=False).fit(X, y).predict(X)
    kr = KernelRidge(kernel="linear", alpha=0)
    ignore_warnings(kr.fit)(X, y)
    pred2 = kr.predict(X)
    assert_array_almost_equal(pred, pred2)

def test_nans():
    # Assert that SelectKBest and SelectPercentile can handle NaNs.
    # First feature has zero variance to confuse f_classif (ANOVA) and
    # make it return a NaN.
    X = [[0, 1, 0], [0, -1, -1], [0, 0.5, 0.5]]
    y = [1, 0, 1]

    for select in (SelectKBest(f_classif, 2), SelectPercentile(f_classif, percentile=67)):
        ignore_warnings(select.fit)(X, y)
        assert_array_equal(select.get_support(indices=True), np.array([1, 2]))

def _run_one_hot(X, X2, cat):
    # enc = assert_warns(
    #     DeprecationWarning,
    #     OneHotEncoder, categorical_features=cat)
    enc = OneHotEncoder(categorical_features=cat)
    with ignore_warnings(category=(DeprecationWarning, FutureWarning)):
        Xtr = enc.fit_transform(X)
    with ignore_warnings(category=(DeprecationWarning, FutureWarning)):
        X2tr = enc.fit(X).transform(X2)
    return Xtr, X2tr

def test_path_parameters():
    X, y = make_sparse_data()
    max_iter = 50
    n_alphas = 10
    clf = ElasticNetCV(n_alphas=n_alphas, eps=1e-3, max_iter=max_iter, l1_ratio=0.5, fit_intercept=False)
    ignore_warnings(clf.fit)(X, y)  # new params
    assert_almost_equal(0.5, clf.l1_ratio)
    assert_equal(n_alphas, clf.n_alphas)
    assert_equal(n_alphas, len(clf.alphas_))
    sparse_mse_path = clf.mse_path_
    ignore_warnings(clf.fit)(X.toarray(), y)  # compare with dense data
    assert_almost_equal(clf.mse_path_, sparse_mse_path)

def test_enet_float_precision():
    # Generate dataset
    X, y, X_test, y_test = build_dataset(n_samples=20, n_features=10)
    # Here we have a small number of iterations, and thus the
    # ElasticNet might not converge. This is to speed up tests

    for normalize in [True, False]:
        for fit_intercept in [True, False]:
            coef = {}
            intercept = {}
            for dtype in [np.float64, np.float32]:
                clf = ElasticNet(alpha=0.5, max_iter=100, precompute=False,
                                 fit_intercept=fit_intercept,
                                 normalize=normalize)

                X = dtype(X)
                y = dtype(y)
                ignore_warnings(clf.fit)(X, y)

                coef[('simple', dtype)] = clf.coef_
                intercept[('simple', dtype)] = clf.intercept_

                assert_equal(clf.coef_.dtype, dtype)

                # test precompute Gram array
                Gram = X.T.dot(X)
                clf_precompute = ElasticNet(alpha=0.5, max_iter=100,
                                            precompute=Gram,
                                            fit_intercept=fit_intercept,
                                            normalize=normalize)
                ignore_warnings(clf_precompute.fit)(X, y)
                assert_array_almost_equal(clf.coef_, clf_precompute.coef_)
                assert_array_almost_equal(clf.intercept_,
                                          clf_precompute.intercept_)

                # test multi task enet
                multi_y = np.hstack((y[:, np.newaxis], y[:, np.newaxis]))
                clf_multioutput = MultiTaskElasticNet(
                    alpha=0.5, max_iter=100, fit_intercept=fit_intercept,
                    normalize=normalize)
                clf_multioutput.fit(X, multi_y)
                coef[('multi', dtype)] = clf_multioutput.coef_
                intercept[('multi', dtype)] = clf_multioutput.intercept_
                assert_equal(clf.coef_.dtype, dtype)

            for v in ['simple', 'multi']:
                assert_array_almost_equal(coef[(v, np.float32)],
                                          coef[(v, np.float64)],
                                          decimal=4)
                assert_array_almost_equal(intercept[(v, np.float32)],
                                          intercept[(v, np.float64)],
                                          decimal=4)