def setUp(self):
if skip_all:
raise nose.SkipTest()
self.xv = ContestCrossValid(
n_train=ContestCrossValid.max_n_train - 7500,
n_valid=7500,
ds=ds)
def test_protocol_svm(self):
if 'sklearn' not in globals():
raise nose.SkipTest()
self.xv = ContestCrossValid(
n_train=200,
n_valid=100,
ds=ds)
algo = SklearnClassifier(
partial(sklearn.svm.SVC, kernel='linear'))
self.xv.protocol(algo)
print algo.results
class TestContestXV(TestCase):
def setUp(self):
if skip_all:
raise nose.SkipTest()
self.xv = ContestCrossValid(
n_train=ContestCrossValid.max_n_train - 7500,
n_valid=7500,
ds=ds)
def test_protocol_smoke(self):
# -- smoke test that it just runs
class Algo(object):
def best_model(algo_self, train, valid=None):
# -- all training labels should be legit
assert np.all(train.all_labels[train.idxs] < 7)
assert np.all(train.all_labels[train.idxs] >= 0)
# -- N.B. test labels might be unknown, and
# replaced with dummy ones which may or may
# not be in range(7)
return None
def loss(algo_self, model, task):
return 1.0
algo = Algo()
self.xv.protocol(algo)
def test_protocol_svm(self):
if 'sklearn' not in globals():
raise nose.SkipTest()
self.xv = ContestCrossValid(
n_train=200,
n_valid=100,
ds=ds)
algo = SklearnClassifier(
partial(sklearn.svm.SVC, kernel='linear'))
self.xv.protocol(algo)
print algo.results
