"""
Module description:
"""
__version__ = '0.3.1'
__author__ = 'Vito Walter Anelli, Claudio Pomo, Daniele Malitesta, Felice Antonio Merra'
__email__ = 'vitowalter.anelli@poliba.it, claudio.pomo@poliba.it, daniele.malitesta@poliba.it, felice.merra@poliba.it'
import os
import numpy as np
import tensorflow as tf
from tensorflow import keras
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
[docs]class NPRModel(keras.Model):
def __init__(self,
num_users,
num_items,
embed_mf_size, l_w, mlp_hidden_size, dropout, learning_rate=0.01,
random_seed=42,
name="NPR",
**kwargs):
super().__init__(name=name, **kwargs)
tf.random.set_seed(random_seed)
self.num_users = num_users
self.num_items = num_items
self.embed_mf_size = embed_mf_size
self.l_w = l_w
self.mlp_hidden_size = mlp_hidden_size
self.dropout = dropout
self.initializer = tf.initializers.GlorotUniform()
self.user_mf_embedding = keras.layers.Embedding(input_dim=self.num_users, output_dim=self.embed_mf_size,
embeddings_initializer=self.initializer, name='U_MF',
dtype=tf.float32)
self.item_mf_embedding_1 = keras.layers.Embedding(input_dim=self.num_items, output_dim=self.embed_mf_size,
embeddings_initializer=self.initializer, name='I_MF_1',
dtype=tf.float32)
self.item_mf_embedding_2 = keras.layers.Embedding(input_dim=self.num_items, output_dim=self.embed_mf_size,
embeddings_initializer=self.initializer, name='I_MF_2',
dtype=tf.float32)
self.mlp_layers_1 = keras.Sequential()
for units in mlp_hidden_size:
# We can have a deeper MLP. In the paper is directly to 1
self.mlp_layers_1.add(keras.layers.Dropout(dropout))
self.mlp_layers_1.add(keras.layers.Dense(units, activation='relu'))
self.mlp_layers_2 = keras.Sequential()
for units in mlp_hidden_size:
# We can have a deeper MLP. In the paper is directly to 1
self.mlp_layers_2.add(keras.layers.Dropout(dropout))
self.mlp_layers_2.add(keras.layers.Dense(units, activation='relu'))
self.optimizer = tf.optimizers.Adam(learning_rate)
[docs] @tf.function
def call(self, inputs, training=None, mask=None):
user, item1, item2 = inputs
user_mf_e = self.user_mf_embedding(user)
item_mf_e_1 = self.item_mf_embedding_1(item1)
item_mf_e_2 = self.item_mf_embedding_2(item2)
embedding_input_1 = user_mf_e * item_mf_e_1 # [batch_size, embedding_size]
mlp_output_1 = self.mlp_layers_1(embedding_input_1) # [batch_size, 1]
embedding_input_2 = user_mf_e * item_mf_e_2
mlp_output_2 = self.mlp_layers_2(embedding_input_2) # [batch_size, 1]
return tf.squeeze(mlp_output_1), tf.squeeze(mlp_output_2), user_mf_e, item_mf_e_1, item_mf_e_2
#@tf.function
[docs] def train_step(self, batch):
with tf.GradientTape() as tape:
user, pos, neg = batch
# Clean Inference
mlp_output_1, mlp_output_2, user_mf_e, item_mf_e_1, item_mf_e_2 = self.call(inputs=(user, pos, neg),
training=True)
difference = tf.clip_by_value(mlp_output_1 - mlp_output_2, -80.0, 1e8)
loss = tf.reduce_sum(tf.nn.softplus(-difference))
# Regularization Component
reg_loss = self.l_w * tf.reduce_sum([tf.nn.l2_loss(user_mf_e),
tf.nn.l2_loss(item_mf_e_1),
tf.nn.l2_loss(item_mf_e_2)])
# Loss to be optimized
loss += reg_loss
grads = tape.gradient(loss, self.trainable_variables)
self.optimizer.apply_gradients(zip(grads, self.trainable_variables))
return loss
[docs] @tf.function
def predict(self, inputs, training=False, **kwargs):
"""
Get full predictions on the whole users/items matrix.
Returns:
The matrix of predicted values.
"""
u, i = inputs
output_1, output_2, _, _, _ = self.call(inputs=(u, i, i), training=training)
return (output_1 + output_2) * 0.5
[docs] @tf.function
def get_recs(self, inputs, training=False, **kwargs):
"""
Get full predictions on the whole users/items matrix.
Returns:
The matrix of predicted values.
"""
user, item = inputs
user_mf_e = self.user_mf_embedding(user)
item_mf_e_1 = self.item_mf_embedding_1(item)
item_mf_e_2 = self.item_mf_embedding_2(item)
mf_output_1 = user_mf_e * item_mf_e_1 # [batch_size, embedding_size]
mf_output_2 = user_mf_e * item_mf_e_2 # [batch_size, embedding_size]
mlp_output_1 = self.mlp_layers_1(mf_output_1) # [batch_size, 1]
mlp_output_2 = self.mlp_layers_2(mf_output_2) # [batch_size, 1]
return tf.squeeze((mlp_output_1+mlp_output_2)/2)
[docs] @tf.function
def get_top_k(self, preds, train_mask, k=100):
return tf.nn.top_k(tf.where(train_mask, preds, -np.inf), k=k, sorted=True)