深度推荐模型——FNN[ECIR16]

视频讲解&＃xff1a;https://www.yuque.com/chudi/tzqav9/ny150b#aalY8

import tensorflow as tf
from tensorflow import keras
from utils import *EPOCH &＃61; 10
BATCH_SIZE &＃61; 32
VEC_DIM &＃61; 10
DNN_LAYERS &＃61; [64, 128, 64]
DROPOUT_RATE &＃61; 0.5base, test &＃61; loadData()
# 所有的特征各个类别值个数之和
FEAT_CATE_NUM &＃61; base.shape[1] - 1
K &＃61; tf.keras.backenddef run():# 将所有的特征的各个类别值统一id化。x中每行为各特征的类别值的idval_x, val_y &＃61; getAllData(test)train_x, train_y &＃61; getAllData(base)model &＃61; keras.models.Sequential()model.add(keras.layers.Embedding(FEAT_CATE_NUM, VEC_DIM, input_length&＃61;val_x[0].shape[0]))model.add(keras.layers.Flatten())for units in DNN_LAYERS:model.add(keras.layers.Dense(units, activation&＃61;&＃39;relu&＃39;))model.add(keras.layers.Dropout(DROPOUT_RATE))model.add(keras.layers.Dense(1, activation&＃61;&＃39;sigmoid&＃39;))model.compile(loss&＃61;&＃39;binary_crossentropy&＃39;, optimizer&＃61;tf.train.AdamOptimizer(0.001), metrics&＃61;[keras.metrics.AUC()])tbCallBack &＃61; keras.callbacks.TensorBoard(log_dir&＃61;&＃39;./logs&＃39;,histogram_freq&＃61;0,write_graph&＃61;True,write_grads&＃61;True,write_images&＃61;True,embeddings_freq&＃61;0,embeddings_layer_names&＃61;None,embeddings_metadata&＃61;None)model.fit(train_x, train_y, batch_size&＃61;BATCH_SIZE, epochs&＃61;EPOCH, verbose&＃61;2, validation_data&＃61;(val_x, val_y),callbacks&＃61;[tbCallBack])run()