"""
Module description:
"""
__version__ = '0.3.1'
__author__ = 'Vito Walter Anelli, Claudio Pomo'
__email__ = 'vitowalter.anelli@poliba.it, claudio.pomo@poliba.it'
import numpy as np
from ast import literal_eval as make_tuple
from tqdm import tqdm
from elliot.dataset.samplers import pointwise_pos_neg_ratio_ratings_sampler as pws
from elliot.recommender.neural.DMF.deep_matrix_factorization_model import DeepMatrixFactorizationModel
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 DMF(RecMixin, BaseRecommenderModel):
r"""
Deep Matrix Factorization Models for Recommender Systems.
For further details, please refer to the `paper <https://www.ijcai.org/Proceedings/2017/0447.pdf>`_
Args:
lr: Learning rate
reg: Regularization coefficient
user_mlp: List of units for each layer
item_mlp: List of activation functions
similarity: Number of factors dimension
To include the recommendation model, add it to the config file adopting the following pattern:
.. code:: yaml
models:
DMF:
meta:
save_recs: True
epochs: 10
batch_size: 512
lr: 0.0001
reg: 0.001
user_mlp: (64,32)
item_mlp: (64,32)
similarity: cosine
"""
@init_charger
def __init__(self, data, config, params, *args, **kwargs):
self._params_list = [
("_learning_rate", "lr", "lr", 0.0001, None, None),
("_user_mlp", "user_mlp", "umlp", "(64,32)", lambda x: list(make_tuple(str(x))), lambda x: self._batch_remove(str(x), " []").replace(",", "-")),
("_item_mlp", "item_mlp", "imlp", "(64,32)", lambda x: list(make_tuple(str(x))), lambda x: self._batch_remove(str(x), " []").replace(",", "-")),
("_neg_ratio", "neg_ratio", "negratio", 5, None, None),
("_reg", "reg", "reg", 0.001, None, None),
("_similarity", "similarity", "sim", "cosine", None, None)
]
self.autoset_params()
self._max_ratings = np.max(self._data.sp_i_train_ratings)
self._transactions_per_epoch = self._data.transactions + self._neg_ratio * self._data.transactions
if self._batch_size < 1:
self._batch_size = self._data.transactions + self._neg_ratio * self._data.transactions
self._sampler = pws.Sampler(self._data.i_train_dict, self._data.sp_i_train_ratings, self._neg_ratio)
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._model = DeepMatrixFactorizationModel(self._num_users, self._num_items, self._user_mlp,
self._item_mlp, self._reg,
self._similarity, self._max_ratings,
self._data.sp_i_train_ratings, self._learning_rate,
self._seed)
@property
def name(self):
return "DMF"\
+ f"_{self.get_base_params_shortcut()}" \
+ f"_{self.get_params_shortcut()}"
[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._transactions_per_epoch // self._batch_size), disable=not self._verbose) as t:
for batch in self._sampler.step(self._transactions_per_epoch, 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