Skip to content

Latest commit

 

History

History
 
 

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 
 
 
 
 

Deep & Cross Network (DCN)

1. 论文

Deep & Cross Network for Ad Click Predictions

创新:Cross Network部分,特征交叉相乘

原文笔记:https://mp.weixin.qq.com/s/DkoaMaXhlgQv1NhZHF-7og

2. 模型结构

3. 实验数据集

采用Criteo数据集进行测试。数据集的处理见../data_process文件,主要分为:

  1. 考虑到Criteo文件过大,因此可以通过read_partsample_sum读取部分数据进行测试;
  2. 对缺失数据进行填充;
  3. 对密集数据I1-I13进行离散化分桶(bins=100),对稀疏数据C1-C26进行重新编码LabelEncoder
  4. 整理得到feature_columns
  5. 切分数据集,最后返回feature_columns, (train_X, train_y), (test_X, test_y)

4. 模型API

class DCN(Model):
    def __init__(self, feature_columns, hidden_units, activation='relu',
                 dnn_dropout=0., embed_reg=1e-6, cross_w_reg=1e-6, cross_b_reg=1e-6):
        """
        Deep&Cross Network
        :param feature_columns: A list. sparse column feature information.
        :param hidden_units: A list. Neural network hidden units.
        :param activation: A string. Activation function of dnn.
        :param dnn_dropout: A scalar. Dropout of dnn.
        :param embed_reg: A scalar. The regularizer of embedding.
        :param cross_w_reg: A scalar. The regularizer of cross network.
        :param cross_b_reg: A scalar. The regularizer of cross network.
        """

5. 实验超参数

  • file:Criteo文件;
  • read_part:是否读取部分数据,True
  • sample_num:读取部分时,样本数量,5000000
  • test_size:测试集比例,0.2
  • embed_dim:Embedding维度,8
  • dnn_dropout:Dropout, 0.5
  • hidden_unit:DNN的隐藏单元,[256, 128, 64]
  • learning_rate:学习率,0.001
  • batch_size:4096
  • epoch:10

6. 实验结果

  1. 采用Criteo数据集中前500w条数据,最终测试集的结果为:AUC: 0.782310
  2. 采用Criteo数据集全部内容:
    • 学习参数:235,236,225;
    • 单个Epoch运行时间【GPU:Tesla V100S-PCI】:344s;
    • 测试集结果:AUC: 0.792959, loss: 0.4691