"""
Module description:
Mnih, Andriy, and Russ R. Salakhutdinov. "Probabilistic matrix factorization." Advances in neural information processing systems 20 (2007)
"""
__version__ = '0.3.1'
__author__ = 'Vito Walter Anelli, Claudio Pomo'
__email__ = 'vitowalter.anelli@poliba.it, claudio.pomo@poliba.it'
import numpy as np
import pickle
from tqdm import tqdm
from elliot.dataset.samplers import pointwise_pos_neg_sampler as pws
from elliot.recommender.latent_factor_models.PMF.probabilistic_matrix_factorization_model import ProbabilisticMatrixFactorizationModel
from elliot.recommender.recommender_utils_mixin import RecMixin
from elliot.utils.write import store_recommendation
from elliot.recommender.base_recommender_model import BaseRecommenderModel
from elliot.recommender.base_recommender_model import init_charger
[docs]class PMF(RecMixin, BaseRecommenderModel):
r"""
Probabilistic Matrix Factorization
For further details, please refer to the `paper <https://papers.nips.cc/paper/2007/file/d7322ed717dedf1eb4e6e52a37ea7bcd-Paper.pdf>`_
Args:
factors: Number of latent factors
lr: Learning rate
reg: Regularization coefficient
gaussian_variance: Variance of the Gaussian distribution
To include the recommendation model, add it to the config file adopting the following pattern:
.. code:: yaml
models:
PMF:
meta:
save_recs: True
epochs: 10
batch_size: 512
factors: 50
lr: 0.001
reg: 0.0025
gaussian_variance: 0.1
"""
@init_charger
def __init__(self, data, config, params, *args, **kwargs):
self._params_list = [
("_learning_rate", "lr", "lr", 0.001, None, None),
("_factors", "factors", "factors", 50, None, None),
("_l_w", "reg", "reg", 0.0025, None, None),
("_gvar", "gaussian_variance", "gvar", 0.1, None, None),
]
self.autoset_params()
if self._batch_size < 1:
self._batch_size = self._data.transactions
self._ratings = self._data.train_dict
self._sp_i_train = self._data.sp_i_train
self._i_items_set = list(range(self._num_items))
self._sampler = pws.Sampler(self._data.i_train_dict)
self._model = ProbabilisticMatrixFactorizationModel(self._num_users,
self._num_items,
self._factors,
self._l_w,
self._gvar,
self._learning_rate,
self._seed)
@property
def name(self):
return "PMF"\
+ f"_{self.get_base_params_shortcut()}" \
+ f"_{self.get_params_shortcut()}"
[docs] def predict(self, u: int, i: int):
pass
[docs] def train(self):
if self._restore:
return self.restore_weights()
for it in self.iterate(self._epochs):
loss = 0
steps = 0
with tqdm(total=int(self._data.transactions // self._batch_size), disable=not self._verbose) as t:
for batch in self._sampler.step(self._data.transactions, self._batch_size):
steps += 1
loss += self._model.train_step(batch)
t.set_postfix({'loss': f'{loss.numpy() / steps:.5f}'})
t.update()
self.evaluate(it, loss.numpy()/(it + 1))
[docs] def get_recommendations(self, k: int = 100):
predictions_top_k_test = {}
predictions_top_k_val = {}
for index, offset in enumerate(range(0, self._num_users, self._batch_size)):
offset_stop = min(offset + self._batch_size, self._num_users)
predictions = self._model.get_recs(
(
np.repeat(np.array(list(range(offset, offset_stop)))[:, None], repeats=self._num_items, axis=1),
np.array([self._i_items_set for _ in range(offset, offset_stop)])
)
)
recs_val, recs_test = self.process_protocol(k, predictions, offset, offset_stop)
predictions_top_k_val.update(recs_val)
predictions_top_k_test.update(recs_test)
return predictions_top_k_val, predictions_top_k_test