def test_sparse():
X = [
"One and two", "One only", "Three and four, nothing else",
"Two nothing else", "Two and three"
]
Y = csr_matrix(
np.array([[1, 1, 0, 0], [1, 0, 0, 0], [0, 0, 1, 1], [0, 1, 0, 0],
[0, 1, 1, 0]]))
model = Pipeline([('vec', KerasVectorizer()),
('clf',
CNNClassifier(multilabel=True,
batch_size=2,
sparse_y=True))])
model.fit(X, Y)
assert model.score(X, Y) > 0.4
assert model.predict(X).shape == (5, 4)
def test_threshold():
X = [
"One",
"One only",
"Two nothing else",
"Two and three"
]
Y = np.array([0, 0, 1, 1])
model = Pipeline([
('vec', KerasVectorizer()),
('clf', CNNClassifier(threshold=0.1))
])
model.fit(X, Y)
Y_pred_expected = model.predict_proba(X) > 0.1
Y_pred = model.predict(X)
assert np.array_equal(Y_pred_expected, Y_pred)
def test_predict_proba():
X = [
"One",
"One only",
"Two nothing else",
"Two and three"
]
Y = np.array([0, 0, 1, 1])
model = Pipeline([
('vec', KerasVectorizer()),
('clf', CNNClassifier())
])
model.fit(X, Y)
Y_pred_prob = model.predict_proba(X)
assert sum(Y_pred_prob >= 0) == Y.shape[0]
assert sum(Y_pred_prob <= 1) == Y.shape[0]
def test_attention():
X = [
"One",
"One only",
"Two nothing else",
"Two and three"
]
Y = np.array([0, 0, 1, 1])
model = Pipeline([
('vec', KerasVectorizer()),
('clf', CNNClassifier(
attention=True,
attention_heads=10))
])
model.fit(X, Y)
assert model.score(X, Y) > 0.6
def test_early_stopping():
X = [
"One",
"One only",
"Two nothing else",
"Two and three"
]
Y = np.array([0, 0, 1, 1])
model = Pipeline([
('vec', KerasVectorizer()),
('clf', CNNClassifier(
early_stopping=True,
nb_epochs=10000
))
])
# if early_stopping is not working it will take
# a lot of time to finish running this test
model.fit(X, Y)
assert model.score(X, Y) > 0.6
from wellcomeml.ml import CNNClassifier, KerasVectorizer
from sklearn.pipeline import Pipeline
import numpy as np
X = ["One", "three", "one", "two", "four"]
Y = np.array([1, 0, 1, 0, 0])
cnn_pipeline = Pipeline([("vec", KerasVectorizer()), ("clf", CNNClassifier())])
cnn_pipeline.fit(X, Y)
print(cnn_pipeline.score(X, Y))
X = ["One, three", "one", "two, three"]
Y = np.array([[1, 0, 1], [1, 0, 0], [0, 1, 1]])
cnn_pipeline = Pipeline([("vec", KerasVectorizer()),
("clf", CNNClassifier(multilabel=True))])
cnn_pipeline.fit(X, Y)
print(cnn_pipeline.score(X, Y))
from wellcomeml.ml import BiLSTMClassifier, KerasVectorizer
from sklearn.pipeline import Pipeline
import numpy as np
X = ["One", "three", "one", "two", "four"]
Y = np.array([1, 0, 1, 0, 0])
bilstm_pipeline = Pipeline([("vec", KerasVectorizer()),
("clf", BiLSTMClassifier())])
bilstm_pipeline.fit(X, Y)
print(bilstm_pipeline.score(X, Y))
X = ["One, three", "one", "two, three"]
Y = np.array([[1, 0, 1], [1, 0, 0], [0, 1, 1]])
bilstm_pipeline = Pipeline([("vec", KerasVectorizer()),
("clf", BiLSTMClassifier(multilabel=True))])
bilstm_pipeline.fit(X, Y)
print(bilstm_pipeline.score(X, Y))