def test_pmf_convergence(self):
n_user = 100
n_item = 200
n_feature = self.n_feature
ratings = make_ratings(n_user,
n_item,
20,
30,
self.rating_choices,
seed=self.seed)
pmf1 = PMF(n_user,
n_item,
n_feature,
seed=0,
max_rating=self.max_rat,
min_rating=self.min_rat,
converge=1e-2)
pmf1.fit(ratings, n_iters=5)
rmse_1 = RMSE(pmf1.predict(ratings[:, :2]), ratings[:, 2])
pmf2 = PMF(n_user,
n_item,
n_feature,
seed=0,
max_rating=self.max_rat,
min_rating=self.min_rat,
converge=1e-1)
pmf2.fit(ratings, n_iters=5)
rmse_2 = RMSE(pmf2.predict(ratings[:, :2]), ratings[:, 2])
self.assertTrue(rmse_1 < rmse_2)
def test_pmf_with_random_data(self):
n_user = 1000
n_item = 2000
n_feature = self.n_feature
ratings = make_ratings(n_user,
n_item,
20,
30,
self.rating_choices,
seed=self.seed)
pmf1 = PMF(n_user,
n_item,
n_feature,
batch_size=1000.,
epsilon=10.,
seed=0,
max_rating=self.max_rat,
min_rating=self.min_rat)
pmf1.fit(ratings, n_iters=1)
rmse_1 = RMSE(pmf1.predict(ratings[:, :2]), ratings[:, 2])
pmf2 = PMF(n_user,
n_item,
n_feature,
batch_size=1000.,
epsilon=10.,
seed=0,
max_rating=self.max_rat,
min_rating=self.min_rat)
pmf2.fit(ratings, n_iters=3)
rmse_2 = RMSE(pmf2.predict(ratings[:, :2]), ratings[:, 2])
self.assertTrue(rmse_1 > rmse_2)
def test_pmf_not_fitted_err(self):
with self.assertRaises(NotFittedError):
ratings = make_ratings(10,
10,
1,
5,
self.rating_choices,
seed=self.seed)
bpmf = PMF(10, 10, self.n_feature)
bpmf.predict(ratings[:, :2])
def test_pmf_with_ml_100k_rating(self):
n_user = 943
n_item = 1682
n_feature = 10
ratings = self.ratings
pmf = PMF(n_user,
n_item,
n_feature,
batch_size=1e4,
epsilon=20.,
reg=1e-4,
max_rating=5.,
min_rating=1.,
seed=self.seed)
pmf.fit(ratings, n_iters=15)
rmse = RMSE(pmf.predict(ratings[:, :2]), ratings[:, 2])
self.assertTrue(rmse < 0.85)
ratings[:, (0, 1)] -= 1
# split data to training & testing
train_pct = 0.9
np.random.shuffle(ratings)
train_size = int(train_pct * ratings.shape[0])
train = ratings[:train_size]
validation = ratings[train_size:]
# models settings
n_feature = 10
eval_iters = 20
print(
"n_user: %d, n_item: %d, n_feature: %d, training size: %d, validation size: %d"
% (n_user, n_item, n_feature, train.shape[0], validation.shape[0]))
pmf = PMF(n_user=n_user,
n_item=n_item,
n_feature=n_feature,
epsilon=25.,
max_rating=5.,
min_rating=1.,
seed=0)
pmf.fit(train, n_iters=eval_iters)
train_preds = pmf.predict(train[:, :2])
train_rmse = RMSE(train_preds, train[:, 2])
val_preds = pmf.predict(validation[:, :2])
val_rmse = RMSE(val_preds, validation[:, 2])
print("after %d iterations, train RMSE: %.6f, validation RMSE: %.6f" %
(eval_iters, train_rmse, val_rmse))
print("Iteration: %d/%d" % (i + 1, Iteration_time))
print("Time cost: %f" % (time.time() - c))
"""
Probabilistic Matrix Factorization
"""
logging.basicConfig(format='%(levelname)s: %(message)s', level=logging.INFO)
rand_state = RandomState(0)
n_feature = 10
eval_iters = 20
print("training PMF model ...")
pmf = PMF(n_user=user_count,
n_item=song_origin_count + 1,
n_feature=n_feature,
epsilon=15.,
converge=1e-8,
momentum=0.4,
max_rating=5.0,
min_rating=0.,
seed=100,
reg=0.01)
pmf.fit(train, n_iters=eval_iters)
"""
Mix 2 model and write result to file
"""
print("Begin Writing result to file ...")
f = open('submission.txt', 'w')
userGroup = 0
GroupSize = 1000
GroupNumber = 200
print("Finished: %d / %d" % (0, GroupNumber * GroupSize))
while (userGroup < GroupNumber):