深度推荐模型——DCN [KDD 17][Google]

import tensorflow as tf
from tensorflow import keras
from utils import *

EPOCH = 10
BATCH_SIZE = 32
VEC_DIM = 10
DNN_LAYERS = [64, 128, 64]
CROSS_LAYER_NUM = 4
DROPOUT_RATE = 0.5

base, test = loadData()
# 所有的特征各个类别值个数之和
FEAT_CATE_NUM = base.shape[1] - 1
K = tf.keras.backend


class CrossLayer(keras.layers.Layer):
    def __init__(self, x0, dim, **kwargs):
        self.x0 = x0
        self.dim = dim
        super(CrossLayer, self).__init__(**kwargs)

    def build(self, input_shape):
        self.w = self.add_weight(name='w', shape=(self.dim, 1), initializer='uniform', trainable=True)
        self.b = self.add_weight(name='b', shape=(1, self.dim), initializer='uniform', trainable=True)
        super(CrossLayer, self).build(input_shape)

    def call(self, xl, **kwargs):
        xl_w = K.dot(xl, self.w)
        x0_xl_w = tf.multiply(self.x0, xl_w)
        return x0_xl_w + self.b + xl


def 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id
    val_x, val_y = getAllData(test)
    train_x, train_y = getAllData(base)
    cate_num = val_x[0].shape[0]

    inputs = keras.Input((cate_num,))
    emb = keras.layers.Embedding(FEAT_CATE_NUM, VEC_DIM, input_length=cate_num)(inputs)
    x0 = keras.layers.Flatten()(emb)
    cross = x0
    # Deep 部分
    deep = keras.layers.Dropout(DROPOUT_RATE)(x0)
    for units in DNN_LAYERS:
        deep = keras.layers.Dense(units, activation='relu')(deep)
        deep = keras.layers.Dropout(DROPOUT_RATE)(deep)

    # Cross 部分
    for layer_num in range(CROSS_LAYER_NUM):
        cross = CrossLayer(x0=x0, dim=VEC_DIM * cate_num)(cross)
    cross = keras.layers.Dropout(DROPOUT_RATE)(cross)

    dcn = keras.layers.concatenate([deep] + [cross])

    dcn = keras.layers.Dropout(DROPOUT_RATE)(dcn)
    outputs = keras.layers.Dense(1, activation='sigmoid', kernel_regularizer=keras.regularizers.l2(0.001))(dcn)
    model = keras.Model(inputs=inputs, outputs=outputs)
    model.compile(loss='binary_crossentropy', optimizer=tf.train.AdamOptimizer(0.001), metrics=[keras.metrics.AUC()])
    tbCallBack = keras.callbacks.TensorBoard(log_dir='./logs',
                                             histogram_freq=0,
                                             write_graph=True,
                                             write_grads=True,
                                             write_images=True,
                                             embeddings_freq=0,
                                             embeddings_layer_names=None,
                                             embeddings_metadata=None)

    model.fit(train_x, train_y, batch_size=BATCH_SIZE, epochs=EPOCH, verbose=2, validation_data=(val_x, val_y),
              callbacks=[tbCallBack])


run()