def test_objective(self):
from orangecontrib.recommendation.ranking.climf import compute_loss
# Load data
data = Orange.data.Table(__dataset__)
steps = [1, 10, 30]
objectives = []
learner = CLiMFLearner(num_factors=10, random_state=42, verbose=0)
for step in steps:
learner.num_iter = step
recommender = learner(data)
# Set parameters
low_rank_matrices = (recommender.U, recommender.V)
params = learner.lmbda
objective = compute_loss(data, low_rank_matrices, params)
objectives.append(objective)
# Assert objective values decrease
test = list(
map(lambda t: t[0] <= t[1], zip(objectives, objectives[1:])))
self.assertTrue(all(test))
def test_objective(self):
from orangecontrib.recommendation.ranking.climf import compute_loss
# Load data
data = Orange.data.Table(__dataset__)
steps = [1, 10, 30]
objectives = []
learner = CLiMFLearner(num_factors=10, random_state=42, verbose=0)
for step in steps:
learner.num_iter = step
recommender = learner(data)
# Set parameters
low_rank_matrices = (recommender.U, recommender.V)
params = learner.lmbda
objective = compute_loss(data, low_rank_matrices, params)
objectives.append(objective)
# Assert objective values decrease
test = list(
map(lambda t: t[0] <= t[1], zip(objectives, objectives[1:])))
self.assertTrue(all(test))
def test_input_data_continuous(self, *args):
learner = CLiMFLearner(num_factors=2, num_iter=1, verbose=3,
callback=lambda x: None)
# Test SGD optimizers too
for opt in __optimizers__:
learner.optimizer = opt
print(learner.optimizer)
super().test_input_data_continuous(learner, filename=__dataset__)
def test_input_data_continuous(self, *args):
learner = CLiMFLearner(num_factors=2,
num_iter=1,
verbose=3,
callback=lambda x: None)
# Test SGD optimizers too
for opt in __optimizers__:
learner.optimizer = opt
print(learner.optimizer)
super().test_input_data_continuous(learner, filename=__dataset__)
def test_outputs(self):
# Load data
data = Orange.data.Table(__dataset__)
# Train recommender
learner = CLiMFLearner(num_factors=2, num_iter=1)
# Train recommender
recommender = learner(data)
# Check tables P, Q
U = recommender.getUTable()
V = recommender.getVTable()
diff = len({U.X.shape[1], V.X.shape[1]})
self.assertEqual(diff, 1)
def test_mrr2(self):
learner = CLiMFLearner(num_factors=10, num_iter=10, verbose=3)
super().test_mrr(learner,
filename='epinions_train.tab',
testdata='epinions_test.tab')
def test_mrr(self, *args):
learner = CLiMFLearner(num_factors=2, num_iter=1, verbose=0)
super().test_mrr(learner, filename=__dataset__)
def test_divergence(self, *args):
learner = CLiMFLearner(num_factors=2, num_iter=1, learning_rate=1e20)
super().test_divergence(learner, filename=__dataset__)
def test_warnings(self, *args):
learner = CLiMFLearner(num_factors=2, num_iter=1, learning_rate=0.0)
super().test_warnings(learner, filename=__dataset__)
def test_CV(self, *args):
learner = CLiMFLearner(num_factors=2, num_iter=1)
super().test_CV(learner, filename=__dataset__)
def test_input_data_discrete(self, *args):
learner = CLiMFLearner(num_factors=2, num_iter=1)
super().test_input_data_discrete(learner, filename=__dataset2__)