Python lightfm.LightFM类代码示例

def test_user_supplied_features_accuracy():

    model = LightFM(random_state=SEED)
    model.fit_partial(
        train,
        user_features=train_user_features,
        item_features=train_item_features,
        epochs=10,
    )

    train_predictions = model.predict(
        train.row,
        train.col,
        user_features=train_user_features,
        item_features=train_item_features,
    )
    test_predictions = model.predict(
        test.row,
        test.col,
        user_features=test_user_features,
        item_features=test_item_features,
    )

    assert roc_auc_score(train.data, train_predictions) > 0.84
    assert roc_auc_score(test.data, test_predictions) > 0.76

def test_recall_at_k():

    no_users, no_items = (10, 100)

    train, test = _generate_data(no_users, no_items)

    model = LightFM(loss="bpr")
    model.fit_partial(test)

    for k in (10, 5, 1):

        # Without omitting train interactions
        recall = evaluation.recall_at_k(model, test, k=k)
        expected_mean_recall = _recall_at_k(model, test, k)

        assert np.allclose(recall.mean(), expected_mean_recall)
        assert len(recall) == (test.getnnz(axis=1) > 0).sum()
        assert (
            len(evaluation.recall_at_k(model, train, preserve_rows=True))
            == test.shape[0]
        )

        # With omitting train interactions
        recall = evaluation.recall_at_k(model, test, k=k, train_interactions=train)
        expected_mean_recall = _recall_at_k(model, test, k, train=train)

        assert np.allclose(recall.mean(), expected_mean_recall)

def test_precision_at_k():

    no_users, no_items = (10, 100)

    train, test = _generate_data(no_users, no_items)

    model = LightFM(loss="bpr")

    # We want a high precision to catch the k=1 case
    model.fit_partial(test)

    for k in (10, 5, 1):

        # Without omitting train interactions
        precision = evaluation.precision_at_k(model, test, k=k)
        expected_mean_precision = _precision_at_k(model, test, k)

        assert np.allclose(precision.mean(), expected_mean_precision)
        assert len(precision) == (test.getnnz(axis=1) > 0).sum()
        assert (
            len(evaluation.precision_at_k(model, train, preserve_rows=True))
            == test.shape[0]
        )

        # With omitting train interactions
        precision = evaluation.precision_at_k(
            model, test, k=k, train_interactions=train
        )
        expected_mean_precision = _precision_at_k(model, test, k, train=train)

        assert np.allclose(precision.mean(), expected_mean_precision)

def test_empty_matrix():

    no_users, no_items = (10, 100)

    train = sp.coo_matrix((no_users, no_items), dtype=np.int32)

    model = LightFM()
    model.fit_partial(train)

def test_intersections_check():

    no_users, no_items = (10, 100)

    train, test = _generate_data(no_users, no_items)

    model = LightFM(loss="bpr")
    model.fit_partial(train)

    # check error is raised when train and test have interactions in common
    with pytest.raises(ValueError):
        evaluation.auc_score(
            model, train, train_interactions=train, check_intersections=True
        )

    with pytest.raises(ValueError):
        evaluation.recall_at_k(
            model, train, train_interactions=train, check_intersections=True
        )

    with pytest.raises(ValueError):
        evaluation.precision_at_k(
            model, train, train_interactions=train, check_intersections=True
        )

    with pytest.raises(ValueError):
        evaluation.reciprocal_rank(
            model, train, train_interactions=train, check_intersections=True
        )

    # check no errors raised when train and test have no interactions in common
    evaluation.auc_score(
        model, test, train_interactions=train, check_intersections=True
    )
    evaluation.recall_at_k(
        model, test, train_interactions=train, check_intersections=True
    )
    evaluation.precision_at_k(
        model, test, train_interactions=train, check_intersections=True
    )
    evaluation.reciprocal_rank(
        model, test, train_interactions=train, check_intersections=True
    )

    # check no error is raised when there are intersections but flag is False
    evaluation.auc_score(
        model, train, train_interactions=train, check_intersections=False
    )
    evaluation.recall_at_k(
        model, train, train_interactions=train, check_intersections=False
    )
    evaluation.precision_at_k(
        model, train, train_interactions=train, check_intersections=False
    )
    evaluation.reciprocal_rank(
        model, train, train_interactions=train, check_intersections=False
    )

def test_warp_few_items():

    no_users, no_items = (1000, 2)

    train = sp.rand(no_users, no_items, format="csr", random_state=42)

    model = LightFM(loss="warp", max_sampled=10)

    model.fit(train)

def test_return_self():

    no_users, no_items = (10, 100)

    train = sp.coo_matrix((no_users, no_items), dtype=np.int32)

    model = LightFM()
    assert model.fit_partial(train) is model
    assert model.fit(train) is model

def test_movielens_accuracy():

    model = LightFM(random_state=SEED)
    model.fit_partial(train, epochs=10)

    train_predictions = model.predict(train.row, train.col)
    test_predictions = model.predict(test.row, test.col)

    assert roc_auc_score(train.data, train_predictions) > 0.84
    assert roc_auc_score(test.data, test_predictions) > 0.76

def test_exception_on_divergence():

    no_users, no_items = (1000, 1000)

    train = sp.rand(no_users, no_items, format="csr", random_state=42)

    model = LightFM(learning_rate=10000000.0, loss="warp")

    with pytest.raises(ValueError):
        model.fit(train, epochs=10)

def test_state_reset():

    model = LightFM(random_state=SEED)

    model.fit(train, epochs=1)

    assert np.mean(model.user_embedding_gradients) > 1.0

    model.fit(train, epochs=0)
    assert np.all(model.user_embedding_gradients == 1.0)

def test_hogwild_accuracy():

    # Should get comparable accuracy with 2 threads
    model = LightFM(random_state=SEED)
    model.fit_partial(train, epochs=10, num_threads=2)

    train_predictions = model.predict(train.row, train.col, num_threads=2)
    test_predictions = model.predict(test.row, test.col, num_threads=2)

    assert roc_auc_score(train.data, train_predictions) > 0.84
    assert roc_auc_score(test.data, test_predictions) > 0.76

def test_nan_interactions():

    no_users, no_items = (1000, 1000)

    train = sp.rand(no_users, no_items, format="csr", random_state=42)
    train.data *= np.nan

    model = LightFM(loss="warp")

    with pytest.raises(ValueError):
        model.fit(train)

def test_warp_stability():

    learning_rates = (0.05, 0.1, 0.5)

    for lrate in learning_rates:

        model = LightFM(learning_rate=lrate, loss="warp", random_state=SEED)
        model.fit_partial(train, epochs=10)

        assert not np.isnan(model.user_embeddings).any()
        assert not np.isnan(model.item_embeddings).any()

def test_movielens_accuracy_pickle():

    model = LightFM(random_state=SEED)
    model.fit(train, epochs=10)

    model = pickle.loads(pickle.dumps(model))

    train_predictions = model.predict(train.row, train.col)
    test_predictions = model.predict(test.row, test.col)

    assert roc_auc_score(train.data, train_predictions) > 0.84
    assert roc_auc_score(test.data, test_predictions) > 0.76

def test_regularization():

    # Let's regularize
    model = LightFM(
        no_components=50, item_alpha=0.0001, user_alpha=0.0001, random_state=SEED
    )
    model.fit_partial(train, epochs=30)

    train_predictions = model.predict(train.row, train.col)
    test_predictions = model.predict(test.row, test.col)

    assert roc_auc_score(train.data, train_predictions) > 0.80
    assert roc_auc_score(test.data, test_predictions) > 0.75