# Copyright 2018 The Cornac Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ============================================================================
import numpy as np
import scipy
from ..recommender import Recommender
from ..recommender import ANNMixin, MEASURE_DOT
from ...utils.common import sigmoid
from ...utils.common import scale
from ...exception import ScoreException
[docs]
class SoRec(Recommender, ANNMixin):
"""Social recommendation using Probabilistic Matrix Factorization.
Parameters
----------
k: int, optional, default: 5
The dimension of the latent factors.
max_iter: int, optional, default: 100
Maximum number of iterations or the number of epochs for SGD.
learning_rate: float, optional, default: 0.001
The learning rate for SGD_RMSProp.
gamma: float, optional, default: 0.9
The weight for previous/current gradient in RMSProp.
lambda_c: float, optional, default: 10
The parameter balancing the information from the user-item rating matrix and the user social network.
lambda_reg: float, optional, default: 0.001
The regularization parameter.
weight_link: boolean, optional, default: True
When true the social network links are weighted according to eq. (4) in the original paper.
name: string, optional, default: 'SoRec'
The name of the recommender model.
trainable: boolean, optional, default: True
When False, the model is not trained and Cornac assumes that the model already
pre-trained (U, V and Z are not None).
verbose: boolean, optional, default: False
When True, some running logs are displayed.
init_params: dictionary, optional, default: None
List of initial parameters, e.g., init_params = {'U':U, 'V':V, 'Z':Z}.
U: a ndarray of shape (n_users, k)
Containing the user latent factors.
V: a ndarray of shape (n_items, k)
Containing the item latent factors.
Z: a ndarray of shape (n_users, k)
Containing the social network latent factors.
seed: int, optional, default: None
Random seed for parameters initialization.
References
----------
* H. Ma, H. Yang, M. R. Lyu, and I. King. SoRec:Social recommendation using probabilistic matrix factorization. \
CIKM ’08, pages 931–940, Napa Valley, USA, 2008.
"""
def __init__(
self,
name="SoRec",
k=5,
max_iter=100,
learning_rate=0.001,
lambda_c=10,
lambda_reg=0.001,
gamma=0.9,
weight_link=True,
trainable=True,
verbose=False,
init_params=None,
seed=None,
):
Recommender.__init__(self, name=name, trainable=trainable, verbose=verbose)
self.k = k
self.max_iter = max_iter
self.learning_rate = learning_rate
self.lambda_c = lambda_c
self.lambda_reg = lambda_reg
self.gamma = gamma
self.weight_link = weight_link
self.ll = np.full(max_iter, 0)
self.eps = 0.000000001
self.seed = seed
# Init params if provided
self.init_params = {} if init_params is None else init_params
self.U = self.init_params.get("U", None) # matrix of user factors
self.V = self.init_params.get("V", None) # matrix of item factors
self.Z = self.init_params.get("Z", None) # matrix of social network factors
if self.U is not None and self.U.shape[1] != self.k:
raise ValueError("initial parameters U dimension error")
if self.V is not None and self.V.shape[1] != self.k:
raise ValueError("initial parameters V dimension error")
if self.Z is not None and self.Z.shape[1] != self.k:
raise ValueError("initial parameters Z dimension error")
[docs]
def fit(self, train_set, val_set=None):
"""Fit the model to observations.
Parameters
----------
train_set: :obj:`cornac.data.Dataset`, required
User-Item preference data as well as additional modalities.
val_set: :obj:`cornac.data.Dataset`, optional, default: None
User-Item preference data for model selection purposes (e.g., early stopping).
Returns
-------
self : object
"""
Recommender.fit(self, train_set, val_set)
import math
from cornac.models.sorec import sorec
if self.trainable:
# user-item interactions
(rat_uid, rat_iid, rat_val) = train_set.uir_tuple
# user social network
train_user_indices = set(train_set.uir_tuple[0])
(net_uid, net_jid, net_val) = train_set.user_graph.get_train_triplet(
train_user_indices, train_user_indices
)
if self.weight_link:
degree = train_set.user_graph.get_node_degree(
train_user_indices, train_user_indices
)
weighted_net_val = []
for u, j, val in zip(net_uid, net_jid, net_val):
u_out = degree[int(u)][1]
j_in = degree[int(j)][0]
val_weighted = math.sqrt(j_in / (j_in + u_out)) * val
weighted_net_val.append(val_weighted)
net_val = weighted_net_val
if [self.min_rating, self.max_rating] != [0, 1]:
if self.min_rating == self.max_rating:
rat_val = scale(rat_val, 0.0, 1.0, 0.0, self.max_rating)
else:
rat_val = scale(rat_val, 0.0, 1.0, self.min_rating, self.max_rating)
rat_val = np.array(rat_val, dtype="float32")
rat_uid = np.array(rat_uid, dtype="int32")
rat_iid = np.array(rat_iid, dtype="int32")
net_val = np.array(net_val, dtype="float32")
net_uid = np.array(net_uid, dtype="int32")
net_jid = np.array(net_jid, dtype="int32")
if self.verbose:
print("Learning...")
res = sorec.sorec(
rat_uid,
rat_iid,
rat_val,
net_uid,
net_jid,
net_val,
k=self.k,
n_users=train_set.num_users,
n_items=train_set.num_items,
n_ratings=len(rat_val),
n_edges=len(net_val),
n_epochs=self.max_iter,
lambda_c=self.lambda_c,
lambda_reg=self.lambda_reg,
learning_rate=self.learning_rate,
gamma=self.gamma,
init_params={"U": self.U, "V": self.V, "Z": self.Z},
verbose=self.verbose,
seed=self.seed,
)
self.U = np.asarray(res["U"])
self.V = np.asarray(res["V"])
self.Z = np.asarray(res["Z"])
if self.verbose:
print("Learning completed")
elif self.verbose:
print("%s is trained already (trainable = False)" % self.name)
return self
[docs]
def score(self, user_idx, item_idx=None):
"""Predict the scores/ratings of a user for an item.
Parameters
----------
user_idx: int, required
The index of the user for whom to perform score prediction.
item_idx: int, optional, default: None
The index of the item for which to perform score prediction.
If None, scores for all known items will be returned.
Returns
-------
res : A scalar or a Numpy array
Relative scores that the user gives to the item or to all known items
"""
if self.is_unknown_user(user_idx):
raise ScoreException("Can't make score prediction for user %d" % user_idx)
if item_idx is not None and self.is_unknown_item(item_idx):
raise ScoreException("Can't make score prediction for item %d" % item_idx)
if item_idx is None:
return self.V.dot(self.U[user_idx, :])
user_pred = self.V[item_idx, :].dot(self.U[user_idx, :])
user_pred = sigmoid(user_pred)
if self.min_rating == self.max_rating:
user_pred = scale(user_pred, 0.0, self.max_rating, 0.0, 1.0)
else:
user_pred = scale(user_pred, self.min_rating, self.max_rating, 0.0, 1.0)
return user_pred
[docs]
def get_vector_measure(self):
"""Getting a valid choice of vector measurement in ANNMixin._measures.
Returns
-------
measure: MEASURE_DOT
Dot product aka. inner product
"""
return MEASURE_DOT
[docs]
def get_user_vectors(self):
"""Getting a matrix of user vectors serving as query for ANN search.
Returns
-------
out: numpy.array
Matrix of user vectors for all users available in the model.
"""
return self.U
[docs]
def get_item_vectors(self):
"""Getting a matrix of item vectors used for building the index for ANN search.
Returns
-------
out: numpy.array
Matrix of item vectors for all items available in the model.
"""
return self.V