def test_vectorizer6():
text = ['foo bar', 'baz']
v = PooledVectorizer(2, sent_start='<s>')
v.fit(text)
bounds1, X1 = v.transform(text)
text2 = ['foo alpha', 'foo bravo', 'foo charlie', 'foo delta']
bounds2, X2 = v.transform(text2)
assert X1.max() >= X2.max()
def test_pooled_net():
cats = ['alt.atheism', 'sci.space']
newsgroups_train = fetch_20newsgroups(subset='train', categories=cats)
newsgroups_test = fetch_20newsgroups(subset='test', categories=cats)
v = PooledVectorizer(3, 1)
bounds, X = v.fit_transform(newsgroups_train.data)
y = newsgroups_train.target
test_bounds, test_X = v.transform(newsgroups_test.data)
test_y = newsgroups_test.target
clsf = MyPooledNetwork2(n_epochs=1, learning_rate=0.1)
clsf.fit((bounds, X), y)
pred_y = clsf.predict((test_bounds, test_X))
print accuracy_score(test_y, pred_y)
def test_vectorizer4():
text = ['foo bar baz foo', 'foo baz']
v = PooledVectorizer(2)
bounds, X = v.transform(text)
