# 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
from ...exception import ScoreException
from .recom_ncf_base import NCFBase
[docs]
class NeuMF(NCFBase):
"""Neural Matrix Factorization.
Parameters
----------
num_factors: int, optional, default: 8
Embedding size of MF model.
layers: list, optional, default: [64, 32, 16, 8]
MLP layers. Note that the first layer is the concatenation of
user and item embeddings. So layers[0]/2 is the embedding size.
act_fn: str, default: 'relu'
Name of the activation function used for the MLP layers.
Supported functions: ['sigmoid', 'tanh', 'elu', 'relu', 'selu, 'relu6', 'leaky_relu']
reg: float, optional, default: 0.
Regularization (weight_decay).
reg_layers: list, optional, default: [0., 0., 0., 0.]
Regularization for each MLP layer,
reg_layers[0] is the regularization for embeddings.
num_epochs: int, optional, default: 20
Number of epochs.
batch_size: int, optional, default: 256
Batch size.
num_neg: int, optional, default: 4
Number of negative instances to pair with a positive instance.
lr: float, optional, default: 0.001
Learning rate.
learner: str, optional, default: 'adam'
Specify an optimizer: adagrad, adam, rmsprop, sgd
backend: str, optional, default: 'tensorflow'
Backend used for model training: tensorflow, pytorch
early_stopping: {min_delta: float, patience: int}, optional, default: None
If `None`, no early stopping. Meaning of the arguments:
- `min_delta`: the minimum increase in monitored value on validation set to be considered as improvement, \
i.e. an increment of less than min_delta will count as no improvement.
- `patience`: number of epochs with no improvement after which training should be stopped.
name: string, optional, default: 'NeuMF'
Name of the recommender model.
trainable: boolean, optional, default: True
When False, the model is not trained and Cornac assumes that the model is already \
pre-trained.
verbose: boolean, optional, default: False
When True, some running logs are displayed.
seed: int, optional, default: None
Random seed for parameters initialization.
References
----------
* He, X., Liao, L., Zhang, H., Nie, L., Hu, X., & Chua, T. S. (2017, April). Neural collaborative filtering. \
In Proceedings of the 26th international conference on world wide web (pp. 173-182).
"""
def __init__(
self,
name="NeuMF",
num_factors=8,
layers=(64, 32, 16, 8),
act_fn="relu",
reg=0.0,
num_epochs=20,
batch_size=256,
num_neg=4,
lr=0.001,
learner="adam",
backend="tensorflow",
early_stopping=None,
trainable=True,
verbose=True,
seed=None,
):
super().__init__(
name=name,
trainable=trainable,
verbose=verbose,
num_epochs=num_epochs,
batch_size=batch_size,
num_neg=num_neg,
lr=lr,
learner=learner,
backend=backend,
early_stopping=early_stopping,
seed=seed,
)
self.num_factors = num_factors
self.layers = layers
self.act_fn = act_fn
self.reg = reg
self.pretrained = False
self.ignored_attrs.extend(
[
"gmf_user_id",
"mlp_user_id",
"pretrained_gmf",
"pretrained_mlp",
"alpha",
]
)
[docs]
def from_pretrained(self, pretrained_gmf, pretrained_mlp, alpha=0.5):
"""Provide pre-trained GMF and MLP models. Section 3.4.1 of the paper.
Parameters
----------
pretrained_gmf: object of type GMF, required
Reference to trained/fitted GMF model.
pretrained_mlp: object of type MLP, required
Reference to trained/fitted MLP model.
alpha: float, optional, default: 0.5
Hyper-parameter determining the trade-off between the two pre-trained models.
Details are described in the section 3.4.1 of the paper.
"""
self.pretrained = True
self.pretrained_gmf = pretrained_gmf
self.pretrained_mlp = pretrained_mlp
self.alpha = alpha
return self
########################
## TensorFlow backend ##
########################
def _build_model_tf(self):
import tensorflow as tf
from .backend_tf import GMFLayer, MLPLayer
# Define inputs
user_input = tf.keras.layers.Input(shape=(1,), dtype=tf.int32, name="user_input")
item_input = tf.keras.layers.Input(shape=(1,), dtype=tf.int32, name="item_input")
# GMF layer
gmf_layer = GMFLayer(
num_users=self.num_users,
num_items=self.num_items,
emb_size=self.num_factors,
reg_user=self.reg,
reg_item=self.reg,
seed=self.seed,
name="gmf_layer"
)
# MLP layer
mlp_layer = MLPLayer(
num_users=self.num_users,
num_items=self.num_items,
layers=self.layers,
reg_layers=[self.reg] * len(self.layers),
act_fn=self.act_fn,
seed=self.seed,
name="mlp_layer"
)
# Get embeddings and element-wise product
gmf_vector = gmf_layer([user_input, item_input])
mlp_vector = mlp_layer([user_input, item_input])
# Concatenate GMF and MLP vectors
concat_vector = tf.keras.layers.Concatenate(axis=-1)([gmf_vector, mlp_vector])
# Output layer
logits = tf.keras.layers.Dense(
1,
kernel_initializer=tf.keras.initializers.LecunUniform(seed=self.seed),
name="logits"
)(concat_vector)
prediction = tf.keras.layers.Activation('sigmoid', name="prediction")(logits)
# Create model
model = tf.keras.Model(
inputs=[user_input, item_input],
outputs=prediction,
name="NeuMF"
)
# Handle pretrained models
if self.pretrained:
# Get GMF and MLP models
gmf_model = self.pretrained_gmf.model
mlp_model = self.pretrained_mlp.model
# Copy GMF embeddings
model.get_layer('gmf_layer').user_embedding.set_weights(
gmf_model.get_layer('gmf_layer').user_embedding.get_weights()
)
model.get_layer('gmf_layer').item_embedding.set_weights(
gmf_model.get_layer('gmf_layer').item_embedding.get_weights()
)
# Copy MLP embeddings and layers
model.get_layer('mlp_layer').user_embedding.set_weights(
mlp_model.get_layer('mlp_layer').user_embedding.get_weights()
)
model.get_layer('mlp_layer').item_embedding.set_weights(
mlp_model.get_layer('mlp_layer').item_embedding.get_weights()
)
# Copy dense layers in MLP
for i, layer in enumerate(model.get_layer('mlp_layer').dense_layers):
layer.set_weights(mlp_model.get_layer('mlp_layer').dense_layers[i].get_weights())
# Combine weights for output layer
gmf_logits_weights = gmf_model.get_layer('logits').get_weights()
mlp_logits_weights = mlp_model.get_layer('logits').get_weights()
# Combine kernel weights
combined_kernel = np.concatenate([
self.alpha * gmf_logits_weights[0],
(1.0 - self.alpha) * mlp_logits_weights[0]
], axis=0)
# Combine bias weights
combined_bias = self.alpha * gmf_logits_weights[1] + (1.0 - self.alpha) * mlp_logits_weights[1]
# Set combined weights to output layer
model.get_layer('logits').set_weights([combined_kernel, combined_bias])
return model
#####################
## PyTorch backend ##
#####################
def _build_model_pt(self):
from .backend_pt import NeuMF
model = NeuMF(
num_users=self.num_users,
num_items=self.num_items,
layers=self.layers,
num_factors=self.num_factors,
act_fn=self.act_fn,
)
if self.pretrained:
model.from_pretrained(
self.pretrained_gmf.model, self.pretrained_mlp.model, self.alpha
)
return model
def _score_pt(self, user_idx, item_idx):
import torch
with torch.no_grad():
if item_idx is None:
users = torch.from_numpy(np.ones(self.num_items, dtype=int) * user_idx)
items = (torch.from_numpy(np.arange(self.num_items))).to(self.device)
else:
users = torch.tensor(user_idx).unsqueeze(0)
items = torch.tensor(item_idx).unsqueeze(0)
gmf_users = torch.tensor(user_idx).unsqueeze(0).to(self.device)
output = self.model(
users.to(self.device), items.to(self.device), gmf_users.to(self.device)
)
return output.squeeze().cpu().numpy()
