def test_parallel_train(self):
if not hasattr(SGDClassifier, 'partial_fit'):
raise SkipTest('sklearn >= 0.13 is required to run this test')
model = SGDClassifier(loss='log', alpha=1e-5, random_state=2)
model = parallel_train(model, self.blocked_data, self.classes)
assert_greater(model.score(self.X, self.y), 0.90)
def test_parallel_train(self):
if not hasattr(SGDClassifier, 'partial_fit'):
raise SkipTest('sklearn >= 0.13 is required to run this test')
model = SGDClassifier(loss='log', alpha=1e-5, random_state=2)
model = parallel_train(model, self.blocked_data, self.classes)
assert_greater(model.score(self.X, self.y), 0.90)
def test_parallel_train_sum_model_non_blocked(self):
n_iter = 2
model = parallel_train(SumModel(), self.data, self.classes, n_iter)
expected_coef = self.X.sum(axis=0) * n_iter / self.n_partitions
assert_array_almost_equal(model.coef_, expected_coef , 5)
def test_parallel_train_sum_model_non_blocked(self):
n_iter = 2
model = parallel_train(SumModel(), self.data, self.classes, n_iter)
expected_coef = self.X.sum(axis=0) * n_iter / self.n_partitions
assert_array_almost_equal(model.coef_, expected_coef, 5)
